]>
Commit | Line | Data |
---|---|---|
cdddbdbc DW |
1 | /* |
2 | * mdadm - Intel(R) Matrix Storage Manager Support | |
3 | * | |
a54d5262 | 4 | * Copyright (C) 2002-2008 Intel Corporation |
cdddbdbc DW |
5 | * |
6 | * This program is free software; you can redistribute it and/or modify it | |
7 | * under the terms and conditions of the GNU General Public License, | |
8 | * version 2, as published by the Free Software Foundation. | |
9 | * | |
10 | * This program is distributed in the hope it will be useful, but WITHOUT | |
11 | * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or | |
12 | * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for | |
13 | * more details. | |
14 | * | |
15 | * You should have received a copy of the GNU General Public License along with | |
16 | * this program; if not, write to the Free Software Foundation, Inc., | |
17 | * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA. | |
18 | */ | |
19 | ||
51006d85 | 20 | #define HAVE_STDINT_H 1 |
cdddbdbc | 21 | #include "mdadm.h" |
c2a1e7da | 22 | #include "mdmon.h" |
420dafcd | 23 | #include "dlink.h" |
51006d85 | 24 | #include "sha1.h" |
88c32bb1 | 25 | #include "platform-intel.h" |
cdddbdbc DW |
26 | #include <values.h> |
27 | #include <scsi/sg.h> | |
28 | #include <ctype.h> | |
d665cc31 | 29 | #include <dirent.h> |
cdddbdbc DW |
30 | |
31 | /* MPB == Metadata Parameter Block */ | |
32 | #define MPB_SIGNATURE "Intel Raid ISM Cfg Sig. " | |
33 | #define MPB_SIG_LEN (strlen(MPB_SIGNATURE)) | |
34 | #define MPB_VERSION_RAID0 "1.0.00" | |
35 | #define MPB_VERSION_RAID1 "1.1.00" | |
fe7ed8cb DW |
36 | #define MPB_VERSION_MANY_VOLUMES_PER_ARRAY "1.2.00" |
37 | #define MPB_VERSION_3OR4_DISK_ARRAY "1.2.01" | |
cdddbdbc | 38 | #define MPB_VERSION_RAID5 "1.2.02" |
fe7ed8cb DW |
39 | #define MPB_VERSION_5OR6_DISK_ARRAY "1.2.04" |
40 | #define MPB_VERSION_CNG "1.2.06" | |
41 | #define MPB_VERSION_ATTRIBS "1.3.00" | |
cdddbdbc DW |
42 | #define MAX_SIGNATURE_LENGTH 32 |
43 | #define MAX_RAID_SERIAL_LEN 16 | |
fe7ed8cb | 44 | |
19482bcc AK |
45 | /* supports RAID0 */ |
46 | #define MPB_ATTRIB_RAID0 __cpu_to_le32(0x00000001) | |
47 | /* supports RAID1 */ | |
48 | #define MPB_ATTRIB_RAID1 __cpu_to_le32(0x00000002) | |
49 | /* supports RAID10 */ | |
50 | #define MPB_ATTRIB_RAID10 __cpu_to_le32(0x00000004) | |
51 | /* supports RAID1E */ | |
52 | #define MPB_ATTRIB_RAID1E __cpu_to_le32(0x00000008) | |
53 | /* supports RAID5 */ | |
54 | #define MPB_ATTRIB_RAID5 __cpu_to_le32(0x00000010) | |
55 | /* supports RAID CNG */ | |
56 | #define MPB_ATTRIB_RAIDCNG __cpu_to_le32(0x00000020) | |
57 | /* supports expanded stripe sizes of 256K, 512K and 1MB */ | |
58 | #define MPB_ATTRIB_EXP_STRIPE_SIZE __cpu_to_le32(0x00000040) | |
59 | ||
60 | /* The OROM Support RST Caching of Volumes */ | |
61 | #define MPB_ATTRIB_NVM __cpu_to_le32(0x02000000) | |
62 | /* The OROM supports creating disks greater than 2TB */ | |
63 | #define MPB_ATTRIB_2TB_DISK __cpu_to_le32(0x04000000) | |
64 | /* The OROM supports Bad Block Management */ | |
65 | #define MPB_ATTRIB_BBM __cpu_to_le32(0x08000000) | |
66 | ||
67 | /* THe OROM Supports NVM Caching of Volumes */ | |
68 | #define MPB_ATTRIB_NEVER_USE2 __cpu_to_le32(0x10000000) | |
69 | /* The OROM supports creating volumes greater than 2TB */ | |
70 | #define MPB_ATTRIB_2TB __cpu_to_le32(0x20000000) | |
71 | /* originally for PMP, now it's wasted b/c. Never use this bit! */ | |
72 | #define MPB_ATTRIB_NEVER_USE __cpu_to_le32(0x40000000) | |
73 | /* Verify MPB contents against checksum after reading MPB */ | |
74 | #define MPB_ATTRIB_CHECKSUM_VERIFY __cpu_to_le32(0x80000000) | |
75 | ||
76 | /* Define all supported attributes that have to be accepted by mdadm | |
77 | */ | |
418f9b36 | 78 | #define MPB_ATTRIB_SUPPORTED (MPB_ATTRIB_CHECKSUM_VERIFY | \ |
19482bcc AK |
79 | MPB_ATTRIB_2TB | \ |
80 | MPB_ATTRIB_2TB_DISK | \ | |
81 | MPB_ATTRIB_RAID0 | \ | |
82 | MPB_ATTRIB_RAID1 | \ | |
83 | MPB_ATTRIB_RAID10 | \ | |
84 | MPB_ATTRIB_RAID5 | \ | |
bbab0940 TM |
85 | MPB_ATTRIB_EXP_STRIPE_SIZE | \ |
86 | MPB_ATTRIB_BBM) | |
418f9b36 N |
87 | |
88 | /* Define attributes that are unused but not harmful */ | |
89 | #define MPB_ATTRIB_IGNORED (MPB_ATTRIB_NEVER_USE) | |
fe7ed8cb | 90 | |
8e59f3d8 | 91 | #define MPB_SECTOR_CNT 2210 |
611d9529 MD |
92 | #define IMSM_RESERVED_SECTORS 8192 |
93 | #define NUM_BLOCKS_DIRTY_STRIPE_REGION 2048 | |
979d38be | 94 | #define SECT_PER_MB_SHIFT 11 |
f36a9ecd | 95 | #define MAX_SECTOR_SIZE 4096 |
c2462068 PB |
96 | #define MULTIPLE_PPL_AREA_SIZE_IMSM (1024 * 1024) /* Size of the whole |
97 | * mutliple PPL area | |
98 | */ | |
cdddbdbc | 99 | |
fbc42556 JR |
100 | /* |
101 | * Internal Write-intent bitmap is stored in the same area where PPL. | |
102 | * Both features are mutually exclusive, so it is not an issue. | |
103 | * The first 8KiB of the area are reserved and shall not be used. | |
104 | */ | |
105 | #define IMSM_BITMAP_AREA_RESERVED_SIZE 8192 | |
106 | ||
107 | #define IMSM_BITMAP_HEADER_OFFSET (IMSM_BITMAP_AREA_RESERVED_SIZE) | |
108 | #define IMSM_BITMAP_HEADER_SIZE MAX_SECTOR_SIZE | |
109 | ||
110 | #define IMSM_BITMAP_START_OFFSET (IMSM_BITMAP_HEADER_OFFSET + IMSM_BITMAP_HEADER_SIZE) | |
111 | #define IMSM_BITMAP_AREA_SIZE (MULTIPLE_PPL_AREA_SIZE_IMSM - IMSM_BITMAP_START_OFFSET) | |
112 | #define IMSM_BITMAP_AND_HEADER_SIZE (IMSM_BITMAP_AREA_SIZE + IMSM_BITMAP_HEADER_SIZE) | |
113 | ||
114 | #define IMSM_DEFAULT_BITMAP_CHUNKSIZE (64 * 1024 * 1024) | |
115 | #define IMSM_DEFAULT_BITMAP_DAEMON_SLEEP 5 | |
116 | ||
761e3bd9 N |
117 | /* |
118 | * This macro let's us ensure that no-one accidentally | |
119 | * changes the size of a struct | |
120 | */ | |
121 | #define ASSERT_SIZE(_struct, size) \ | |
122 | static inline void __assert_size_##_struct(void) \ | |
123 | { \ | |
124 | switch (0) { \ | |
125 | case 0: break; \ | |
126 | case (sizeof(struct _struct) == size): break; \ | |
127 | } \ | |
128 | } | |
129 | ||
cdddbdbc DW |
130 | /* Disk configuration info. */ |
131 | #define IMSM_MAX_DEVICES 255 | |
132 | struct imsm_disk { | |
133 | __u8 serial[MAX_RAID_SERIAL_LEN];/* 0xD8 - 0xE7 ascii serial number */ | |
5551b113 | 134 | __u32 total_blocks_lo; /* 0xE8 - 0xEB total blocks lo */ |
cdddbdbc | 135 | __u32 scsi_id; /* 0xEC - 0xEF scsi ID */ |
f2f27e63 DW |
136 | #define SPARE_DISK __cpu_to_le32(0x01) /* Spare */ |
137 | #define CONFIGURED_DISK __cpu_to_le32(0x02) /* Member of some RaidDev */ | |
138 | #define FAILED_DISK __cpu_to_le32(0x04) /* Permanent failure */ | |
2432ce9b | 139 | #define JOURNAL_DISK __cpu_to_le32(0x2000000) /* Device marked as Journaling Drive */ |
cdddbdbc | 140 | __u32 status; /* 0xF0 - 0xF3 */ |
1011e834 | 141 | __u32 owner_cfg_num; /* which config 0,1,2... owns this disk */ |
5551b113 CA |
142 | __u32 total_blocks_hi; /* 0xF4 - 0xF5 total blocks hi */ |
143 | #define IMSM_DISK_FILLERS 3 | |
144 | __u32 filler[IMSM_DISK_FILLERS]; /* 0xF5 - 0x107 MPB_DISK_FILLERS for future expansion */ | |
cdddbdbc | 145 | }; |
761e3bd9 | 146 | ASSERT_SIZE(imsm_disk, 48) |
cdddbdbc | 147 | |
3b451610 AK |
148 | /* map selector for map managment |
149 | */ | |
238c0a71 AK |
150 | #define MAP_0 0 |
151 | #define MAP_1 1 | |
152 | #define MAP_X -1 | |
3b451610 | 153 | |
cdddbdbc DW |
154 | /* RAID map configuration infos. */ |
155 | struct imsm_map { | |
5551b113 CA |
156 | __u32 pba_of_lba0_lo; /* start address of partition */ |
157 | __u32 blocks_per_member_lo;/* blocks per member */ | |
158 | __u32 num_data_stripes_lo; /* number of data stripes */ | |
cdddbdbc DW |
159 | __u16 blocks_per_strip; |
160 | __u8 map_state; /* Normal, Uninitialized, Degraded, Failed */ | |
161 | #define IMSM_T_STATE_NORMAL 0 | |
162 | #define IMSM_T_STATE_UNINITIALIZED 1 | |
e3bba0e0 DW |
163 | #define IMSM_T_STATE_DEGRADED 2 |
164 | #define IMSM_T_STATE_FAILED 3 | |
cdddbdbc DW |
165 | __u8 raid_level; |
166 | #define IMSM_T_RAID0 0 | |
167 | #define IMSM_T_RAID1 1 | |
168 | #define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */ | |
169 | __u8 num_members; /* number of member disks */ | |
fe7ed8cb DW |
170 | __u8 num_domains; /* number of parity domains */ |
171 | __u8 failed_disk_num; /* valid only when state is degraded */ | |
252d23c0 | 172 | __u8 ddf; |
5551b113 CA |
173 | __u32 pba_of_lba0_hi; |
174 | __u32 blocks_per_member_hi; | |
175 | __u32 num_data_stripes_hi; | |
176 | __u32 filler[4]; /* expansion area */ | |
7eef0453 | 177 | #define IMSM_ORD_REBUILD (1 << 24) |
cdddbdbc | 178 | __u32 disk_ord_tbl[1]; /* disk_ord_tbl[num_members], |
7eef0453 DW |
179 | * top byte contains some flags |
180 | */ | |
761e3bd9 N |
181 | }; |
182 | ASSERT_SIZE(imsm_map, 52) | |
cdddbdbc DW |
183 | |
184 | struct imsm_vol { | |
4036e7ee | 185 | __u32 curr_migr_unit_lo; |
fe7ed8cb | 186 | __u32 checkpoint_id; /* id to access curr_migr_unit */ |
cdddbdbc | 187 | __u8 migr_state; /* Normal or Migrating */ |
e3bba0e0 DW |
188 | #define MIGR_INIT 0 |
189 | #define MIGR_REBUILD 1 | |
190 | #define MIGR_VERIFY 2 /* analagous to echo check > sync_action */ | |
191 | #define MIGR_GEN_MIGR 3 | |
192 | #define MIGR_STATE_CHANGE 4 | |
1484e727 | 193 | #define MIGR_REPAIR 5 |
cdddbdbc | 194 | __u8 migr_type; /* Initializing, Rebuilding, ... */ |
2432ce9b AP |
195 | #define RAIDVOL_CLEAN 0 |
196 | #define RAIDVOL_DIRTY 1 | |
197 | #define RAIDVOL_DSRECORD_VALID 2 | |
cdddbdbc | 198 | __u8 dirty; |
fe7ed8cb DW |
199 | __u8 fs_state; /* fast-sync state for CnG (0xff == disabled) */ |
200 | __u16 verify_errors; /* number of mismatches */ | |
201 | __u16 bad_blocks; /* number of bad blocks during verify */ | |
4036e7ee MT |
202 | __u32 curr_migr_unit_hi; |
203 | __u32 filler[3]; | |
cdddbdbc DW |
204 | struct imsm_map map[1]; |
205 | /* here comes another one if migr_state */ | |
761e3bd9 N |
206 | }; |
207 | ASSERT_SIZE(imsm_vol, 84) | |
cdddbdbc DW |
208 | |
209 | struct imsm_dev { | |
fe7ed8cb | 210 | __u8 volume[MAX_RAID_SERIAL_LEN]; |
cdddbdbc DW |
211 | __u32 size_low; |
212 | __u32 size_high; | |
fe7ed8cb DW |
213 | #define DEV_BOOTABLE __cpu_to_le32(0x01) |
214 | #define DEV_BOOT_DEVICE __cpu_to_le32(0x02) | |
215 | #define DEV_READ_COALESCING __cpu_to_le32(0x04) | |
216 | #define DEV_WRITE_COALESCING __cpu_to_le32(0x08) | |
217 | #define DEV_LAST_SHUTDOWN_DIRTY __cpu_to_le32(0x10) | |
218 | #define DEV_HIDDEN_AT_BOOT __cpu_to_le32(0x20) | |
219 | #define DEV_CURRENTLY_HIDDEN __cpu_to_le32(0x40) | |
220 | #define DEV_VERIFY_AND_FIX __cpu_to_le32(0x80) | |
221 | #define DEV_MAP_STATE_UNINIT __cpu_to_le32(0x100) | |
222 | #define DEV_NO_AUTO_RECOVERY __cpu_to_le32(0x200) | |
223 | #define DEV_CLONE_N_GO __cpu_to_le32(0x400) | |
224 | #define DEV_CLONE_MAN_SYNC __cpu_to_le32(0x800) | |
225 | #define DEV_CNG_MASTER_DISK_NUM __cpu_to_le32(0x1000) | |
cdddbdbc DW |
226 | __u32 status; /* Persistent RaidDev status */ |
227 | __u32 reserved_blocks; /* Reserved blocks at beginning of volume */ | |
fe7ed8cb DW |
228 | __u8 migr_priority; |
229 | __u8 num_sub_vols; | |
230 | __u8 tid; | |
231 | __u8 cng_master_disk; | |
232 | __u16 cache_policy; | |
233 | __u8 cng_state; | |
234 | __u8 cng_sub_state; | |
2432ce9b AP |
235 | __u16 my_vol_raid_dev_num; /* Used in Unique volume Id for this RaidDev */ |
236 | ||
237 | /* NVM_EN */ | |
238 | __u8 nv_cache_mode; | |
239 | __u8 nv_cache_flags; | |
240 | ||
241 | /* Unique Volume Id of the NvCache Volume associated with this volume */ | |
242 | __u32 nvc_vol_orig_family_num; | |
243 | __u16 nvc_vol_raid_dev_num; | |
244 | ||
245 | #define RWH_OFF 0 | |
246 | #define RWH_DISTRIBUTED 1 | |
247 | #define RWH_JOURNALING_DRIVE 2 | |
c2462068 PB |
248 | #define RWH_MULTIPLE_DISTRIBUTED 3 |
249 | #define RWH_MULTIPLE_PPLS_JOURNALING_DRIVE 4 | |
250 | #define RWH_MULTIPLE_OFF 5 | |
fbc42556 | 251 | #define RWH_BITMAP 6 |
2432ce9b AP |
252 | __u8 rwh_policy; /* Raid Write Hole Policy */ |
253 | __u8 jd_serial[MAX_RAID_SERIAL_LEN]; /* Journal Drive serial number */ | |
254 | __u8 filler1; | |
255 | ||
256 | #define IMSM_DEV_FILLERS 3 | |
cdddbdbc DW |
257 | __u32 filler[IMSM_DEV_FILLERS]; |
258 | struct imsm_vol vol; | |
761e3bd9 N |
259 | }; |
260 | ASSERT_SIZE(imsm_dev, 164) | |
cdddbdbc DW |
261 | |
262 | struct imsm_super { | |
263 | __u8 sig[MAX_SIGNATURE_LENGTH]; /* 0x00 - 0x1F */ | |
264 | __u32 check_sum; /* 0x20 - 0x23 MPB Checksum */ | |
265 | __u32 mpb_size; /* 0x24 - 0x27 Size of MPB */ | |
266 | __u32 family_num; /* 0x28 - 0x2B Checksum from first time this config was written */ | |
267 | __u32 generation_num; /* 0x2C - 0x2F Incremented each time this array's MPB is written */ | |
604b746f JD |
268 | __u32 error_log_size; /* 0x30 - 0x33 in bytes */ |
269 | __u32 attributes; /* 0x34 - 0x37 */ | |
cdddbdbc DW |
270 | __u8 num_disks; /* 0x38 Number of configured disks */ |
271 | __u8 num_raid_devs; /* 0x39 Number of configured volumes */ | |
604b746f JD |
272 | __u8 error_log_pos; /* 0x3A */ |
273 | __u8 fill[1]; /* 0x3B */ | |
274 | __u32 cache_size; /* 0x3c - 0x40 in mb */ | |
275 | __u32 orig_family_num; /* 0x40 - 0x43 original family num */ | |
276 | __u32 pwr_cycle_count; /* 0x44 - 0x47 simulated power cycle count for array */ | |
277 | __u32 bbm_log_size; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */ | |
2a24dc1b PB |
278 | __u16 num_raid_devs_created; /* 0x4C - 0x4D Used for generating unique |
279 | * volume IDs for raid_dev created in this array | |
280 | * (starts at 1) | |
281 | */ | |
282 | __u16 filler1; /* 0x4E - 0x4F */ | |
e48aed3c AP |
283 | __u64 creation_time; /* 0x50 - 0x57 Array creation time */ |
284 | #define IMSM_FILLERS 32 | |
285 | __u32 filler[IMSM_FILLERS]; /* 0x58 - 0xD7 RAID_MPB_FILLERS */ | |
cdddbdbc DW |
286 | struct imsm_disk disk[1]; /* 0xD8 diskTbl[numDisks] */ |
287 | /* here comes imsm_dev[num_raid_devs] */ | |
604b746f | 288 | /* here comes BBM logs */ |
761e3bd9 N |
289 | }; |
290 | ASSERT_SIZE(imsm_super, 264) | |
cdddbdbc | 291 | |
604b746f | 292 | #define BBM_LOG_MAX_ENTRIES 254 |
8d67477f TM |
293 | #define BBM_LOG_MAX_LBA_ENTRY_VAL 256 /* Represents 256 LBAs */ |
294 | #define BBM_LOG_SIGNATURE 0xabadb10c | |
295 | ||
296 | struct bbm_log_block_addr { | |
297 | __u16 w1; | |
298 | __u32 dw1; | |
299 | } __attribute__ ((__packed__)); | |
604b746f JD |
300 | |
301 | struct bbm_log_entry { | |
8d67477f TM |
302 | __u8 marked_count; /* Number of blocks marked - 1 */ |
303 | __u8 disk_ordinal; /* Disk entry within the imsm_super */ | |
304 | struct bbm_log_block_addr defective_block_start; | |
604b746f JD |
305 | } __attribute__ ((__packed__)); |
306 | ||
307 | struct bbm_log { | |
308 | __u32 signature; /* 0xABADB10C */ | |
309 | __u32 entry_count; | |
8d67477f | 310 | struct bbm_log_entry marked_block_entries[BBM_LOG_MAX_ENTRIES]; |
761e3bd9 N |
311 | }; |
312 | ASSERT_SIZE(bbm_log, 2040) | |
604b746f | 313 | |
cdddbdbc | 314 | static char *map_state_str[] = { "normal", "uninitialized", "degraded", "failed" }; |
cdddbdbc | 315 | |
b53bfba6 TM |
316 | #define BLOCKS_PER_KB (1024/512) |
317 | ||
8e59f3d8 AK |
318 | #define RAID_DISK_RESERVED_BLOCKS_IMSM_HI 2209 |
319 | ||
320 | #define GEN_MIGR_AREA_SIZE 2048 /* General Migration Copy Area size in blocks */ | |
321 | ||
de44e46f PB |
322 | #define MIGR_REC_BUF_SECTORS 1 /* size of migr_record i/o buffer in sectors */ |
323 | #define MIGR_REC_SECTOR_POSITION 1 /* migr_record position offset on disk, | |
324 | * MIGR_REC_BUF_SECTORS <= MIGR_REC_SECTOR_POS | |
17a4eaf9 AK |
325 | */ |
326 | ||
8e59f3d8 AK |
327 | #define UNIT_SRC_NORMAL 0 /* Source data for curr_migr_unit must |
328 | * be recovered using srcMap */ | |
329 | #define UNIT_SRC_IN_CP_AREA 1 /* Source data for curr_migr_unit has | |
330 | * already been migrated and must | |
331 | * be recovered from checkpoint area */ | |
2432ce9b | 332 | |
c2462068 | 333 | #define PPL_ENTRY_SPACE (128 * 1024) /* Size of single PPL, without the header */ |
2432ce9b | 334 | |
8e59f3d8 AK |
335 | struct migr_record { |
336 | __u32 rec_status; /* Status used to determine how to restart | |
337 | * migration in case it aborts | |
338 | * in some fashion */ | |
9f421827 | 339 | __u32 curr_migr_unit_lo; /* 0..numMigrUnits-1 */ |
8e59f3d8 AK |
340 | __u32 family_num; /* Family number of MPB |
341 | * containing the RaidDev | |
342 | * that is migrating */ | |
343 | __u32 ascending_migr; /* True if migrating in increasing | |
344 | * order of lbas */ | |
345 | __u32 blocks_per_unit; /* Num disk blocks per unit of operation */ | |
346 | __u32 dest_depth_per_unit; /* Num member blocks each destMap | |
347 | * member disk | |
348 | * advances per unit-of-operation */ | |
9f421827 PB |
349 | __u32 ckpt_area_pba_lo; /* Pba of first block of ckpt copy area */ |
350 | __u32 dest_1st_member_lba_lo; /* First member lba on first | |
351 | * stripe of destination */ | |
352 | __u32 num_migr_units_lo; /* Total num migration units-of-op */ | |
8e59f3d8 AK |
353 | __u32 post_migr_vol_cap; /* Size of volume after |
354 | * migration completes */ | |
355 | __u32 post_migr_vol_cap_hi; /* Expansion space for LBA64 */ | |
356 | __u32 ckpt_read_disk_num; /* Which member disk in destSubMap[0] the | |
357 | * migration ckpt record was read from | |
358 | * (for recovered migrations) */ | |
9f421827 PB |
359 | __u32 curr_migr_unit_hi; /* 0..numMigrUnits-1 high order 32 bits */ |
360 | __u32 ckpt_area_pba_hi; /* Pba of first block of ckpt copy area | |
361 | * high order 32 bits */ | |
362 | __u32 dest_1st_member_lba_hi; /* First member lba on first stripe of | |
363 | * destination - high order 32 bits */ | |
364 | __u32 num_migr_units_hi; /* Total num migration units-of-op | |
365 | * high order 32 bits */ | |
4036e7ee | 366 | __u32 filler[16]; |
761e3bd9 | 367 | }; |
4036e7ee | 368 | ASSERT_SIZE(migr_record, 128) |
8e59f3d8 | 369 | |
76c152ca MT |
370 | /** |
371 | * enum imsm_status - internal IMSM return values representation. | |
372 | * @STATUS_OK: function succeeded. | |
373 | * @STATUS_ERROR: General error ocurred (not specified). | |
374 | * | |
375 | * Typedefed to imsm_status_t. | |
376 | */ | |
377 | typedef enum imsm_status { | |
378 | IMSM_STATUS_ERROR = -1, | |
379 | IMSM_STATUS_OK = 0, | |
380 | } imsm_status_t; | |
381 | ||
ec50f7b6 LM |
382 | struct md_list { |
383 | /* usage marker: | |
384 | * 1: load metadata | |
385 | * 2: metadata does not match | |
386 | * 4: already checked | |
387 | */ | |
388 | int used; | |
389 | char *devname; | |
390 | int found; | |
391 | int container; | |
392 | dev_t st_rdev; | |
393 | struct md_list *next; | |
394 | }; | |
395 | ||
e7b84f9d | 396 | #define pr_vrb(fmt, arg...) (void) (verbose && pr_err(fmt, ##arg)) |
ec50f7b6 | 397 | |
1484e727 DW |
398 | static __u8 migr_type(struct imsm_dev *dev) |
399 | { | |
400 | if (dev->vol.migr_type == MIGR_VERIFY && | |
401 | dev->status & DEV_VERIFY_AND_FIX) | |
402 | return MIGR_REPAIR; | |
403 | else | |
404 | return dev->vol.migr_type; | |
405 | } | |
406 | ||
407 | static void set_migr_type(struct imsm_dev *dev, __u8 migr_type) | |
408 | { | |
409 | /* for compatibility with older oroms convert MIGR_REPAIR, into | |
410 | * MIGR_VERIFY w/ DEV_VERIFY_AND_FIX status | |
411 | */ | |
412 | if (migr_type == MIGR_REPAIR) { | |
413 | dev->vol.migr_type = MIGR_VERIFY; | |
414 | dev->status |= DEV_VERIFY_AND_FIX; | |
415 | } else { | |
416 | dev->vol.migr_type = migr_type; | |
417 | dev->status &= ~DEV_VERIFY_AND_FIX; | |
418 | } | |
419 | } | |
420 | ||
f36a9ecd | 421 | static unsigned int sector_count(__u32 bytes, unsigned int sector_size) |
cdddbdbc | 422 | { |
f36a9ecd | 423 | return ROUND_UP(bytes, sector_size) / sector_size; |
87eb16df | 424 | } |
cdddbdbc | 425 | |
f36a9ecd PB |
426 | static unsigned int mpb_sectors(struct imsm_super *mpb, |
427 | unsigned int sector_size) | |
87eb16df | 428 | { |
f36a9ecd | 429 | return sector_count(__le32_to_cpu(mpb->mpb_size), sector_size); |
cdddbdbc DW |
430 | } |
431 | ||
ba2de7ba DW |
432 | struct intel_dev { |
433 | struct imsm_dev *dev; | |
434 | struct intel_dev *next; | |
f21e18ca | 435 | unsigned index; |
ba2de7ba DW |
436 | }; |
437 | ||
88654014 LM |
438 | struct intel_hba { |
439 | enum sys_dev_type type; | |
440 | char *path; | |
441 | char *pci_id; | |
442 | struct intel_hba *next; | |
443 | }; | |
444 | ||
1a64be56 LM |
445 | enum action { |
446 | DISK_REMOVE = 1, | |
447 | DISK_ADD | |
448 | }; | |
cdddbdbc DW |
449 | /* internal representation of IMSM metadata */ |
450 | struct intel_super { | |
451 | union { | |
949c47a0 DW |
452 | void *buf; /* O_DIRECT buffer for reading/writing metadata */ |
453 | struct imsm_super *anchor; /* immovable parameters */ | |
cdddbdbc | 454 | }; |
8e59f3d8 AK |
455 | union { |
456 | void *migr_rec_buf; /* buffer for I/O operations */ | |
457 | struct migr_record *migr_rec; /* migration record */ | |
458 | }; | |
51d83f5d AK |
459 | int clean_migration_record_by_mdmon; /* when reshape is switched to next |
460 | array, it indicates that mdmon is allowed to clean migration | |
461 | record */ | |
949c47a0 | 462 | size_t len; /* size of the 'buf' allocation */ |
bbab0940 | 463 | size_t extra_space; /* extra space in 'buf' that is not used yet */ |
4d7b1503 DW |
464 | void *next_buf; /* for realloc'ing buf from the manager */ |
465 | size_t next_len; | |
c2c087e6 | 466 | int updates_pending; /* count of pending updates for mdmon */ |
bf5a934a | 467 | int current_vol; /* index of raid device undergoing creation */ |
5551b113 | 468 | unsigned long long create_offset; /* common start for 'current_vol' */ |
148acb7b | 469 | __u32 random; /* random data for seeding new family numbers */ |
ba2de7ba | 470 | struct intel_dev *devlist; |
fa7bb6f8 | 471 | unsigned int sector_size; /* sector size of used member drives */ |
cdddbdbc DW |
472 | struct dl { |
473 | struct dl *next; | |
474 | int index; | |
475 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
476 | int major, minor; | |
477 | char *devname; | |
b9f594fe | 478 | struct imsm_disk disk; |
cdddbdbc | 479 | int fd; |
0dcecb2e DW |
480 | int extent_cnt; |
481 | struct extent *e; /* for determining freespace @ create */ | |
efb30e7f | 482 | int raiddisk; /* slot to fill in autolayout */ |
1a64be56 | 483 | enum action action; |
ca0748fa | 484 | } *disks, *current_disk; |
1a64be56 LM |
485 | struct dl *disk_mgmt_list; /* list of disks to add/remove while mdmon |
486 | active */ | |
47ee5a45 | 487 | struct dl *missing; /* disks removed while we weren't looking */ |
43dad3d6 | 488 | struct bbm_log *bbm_log; |
88654014 | 489 | struct intel_hba *hba; /* device path of the raid controller for this metadata */ |
88c32bb1 | 490 | const struct imsm_orom *orom; /* platform firmware support */ |
a2b97981 | 491 | struct intel_super *next; /* (temp) list for disambiguating family_num */ |
928f1424 | 492 | struct md_bb bb; /* memory for get_bad_blocks call */ |
a2b97981 DW |
493 | }; |
494 | ||
495 | struct intel_disk { | |
496 | struct imsm_disk disk; | |
497 | #define IMSM_UNKNOWN_OWNER (-1) | |
498 | int owner; | |
499 | struct intel_disk *next; | |
cdddbdbc DW |
500 | }; |
501 | ||
aa19fdd4 MT |
502 | /** |
503 | * struct extent - reserved space details. | |
504 | * @start: start offset. | |
505 | * @size: size of reservation, set to 0 for metadata reservation. | |
506 | * @vol: index of the volume, meaningful if &size is set. | |
507 | */ | |
c2c087e6 DW |
508 | struct extent { |
509 | unsigned long long start, size; | |
aa19fdd4 | 510 | int vol; |
c2c087e6 DW |
511 | }; |
512 | ||
694575e7 KW |
513 | /* definitions of reshape process types */ |
514 | enum imsm_reshape_type { | |
515 | CH_TAKEOVER, | |
b5347799 | 516 | CH_MIGRATION, |
7abc9871 | 517 | CH_ARRAY_SIZE, |
694575e7 KW |
518 | }; |
519 | ||
88758e9d DW |
520 | /* definition of messages passed to imsm_process_update */ |
521 | enum imsm_update_type { | |
522 | update_activate_spare, | |
8273f55e | 523 | update_create_array, |
33414a01 | 524 | update_kill_array, |
aa534678 | 525 | update_rename_array, |
1a64be56 | 526 | update_add_remove_disk, |
78b10e66 | 527 | update_reshape_container_disks, |
48c5303a | 528 | update_reshape_migration, |
2d40f3a1 AK |
529 | update_takeover, |
530 | update_general_migration_checkpoint, | |
f3871fdc | 531 | update_size_change, |
bbab0940 | 532 | update_prealloc_badblocks_mem, |
e6e9dd3f | 533 | update_rwh_policy, |
88758e9d DW |
534 | }; |
535 | ||
536 | struct imsm_update_activate_spare { | |
537 | enum imsm_update_type type; | |
d23fe947 | 538 | struct dl *dl; |
88758e9d DW |
539 | int slot; |
540 | int array; | |
541 | struct imsm_update_activate_spare *next; | |
542 | }; | |
543 | ||
78b10e66 | 544 | struct geo_params { |
4dd2df09 | 545 | char devnm[32]; |
78b10e66 | 546 | char *dev_name; |
d04f65f4 | 547 | unsigned long long size; |
78b10e66 N |
548 | int level; |
549 | int layout; | |
550 | int chunksize; | |
551 | int raid_disks; | |
552 | }; | |
553 | ||
bb025c2f KW |
554 | enum takeover_direction { |
555 | R10_TO_R0, | |
556 | R0_TO_R10 | |
557 | }; | |
558 | struct imsm_update_takeover { | |
559 | enum imsm_update_type type; | |
560 | int subarray; | |
561 | enum takeover_direction direction; | |
562 | }; | |
78b10e66 N |
563 | |
564 | struct imsm_update_reshape { | |
565 | enum imsm_update_type type; | |
566 | int old_raid_disks; | |
567 | int new_raid_disks; | |
48c5303a PC |
568 | |
569 | int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */ | |
570 | }; | |
571 | ||
572 | struct imsm_update_reshape_migration { | |
573 | enum imsm_update_type type; | |
574 | int old_raid_disks; | |
575 | int new_raid_disks; | |
576 | /* fields for array migration changes | |
577 | */ | |
578 | int subdev; | |
579 | int new_level; | |
580 | int new_layout; | |
4bba0439 | 581 | int new_chunksize; |
48c5303a | 582 | |
d195167d | 583 | int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */ |
78b10e66 N |
584 | }; |
585 | ||
f3871fdc AK |
586 | struct imsm_update_size_change { |
587 | enum imsm_update_type type; | |
588 | int subdev; | |
589 | long long new_size; | |
590 | }; | |
591 | ||
2d40f3a1 AK |
592 | struct imsm_update_general_migration_checkpoint { |
593 | enum imsm_update_type type; | |
4036e7ee | 594 | __u64 curr_migr_unit; |
2d40f3a1 AK |
595 | }; |
596 | ||
54c2c1ea DW |
597 | struct disk_info { |
598 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
599 | }; | |
600 | ||
8273f55e DW |
601 | struct imsm_update_create_array { |
602 | enum imsm_update_type type; | |
8273f55e | 603 | int dev_idx; |
6a3e913e | 604 | struct imsm_dev dev; |
8273f55e DW |
605 | }; |
606 | ||
33414a01 DW |
607 | struct imsm_update_kill_array { |
608 | enum imsm_update_type type; | |
609 | int dev_idx; | |
610 | }; | |
611 | ||
aa534678 DW |
612 | struct imsm_update_rename_array { |
613 | enum imsm_update_type type; | |
614 | __u8 name[MAX_RAID_SERIAL_LEN]; | |
615 | int dev_idx; | |
616 | }; | |
617 | ||
1a64be56 | 618 | struct imsm_update_add_remove_disk { |
43dad3d6 DW |
619 | enum imsm_update_type type; |
620 | }; | |
621 | ||
bbab0940 TM |
622 | struct imsm_update_prealloc_bb_mem { |
623 | enum imsm_update_type type; | |
624 | }; | |
625 | ||
e6e9dd3f AP |
626 | struct imsm_update_rwh_policy { |
627 | enum imsm_update_type type; | |
628 | int new_policy; | |
629 | int dev_idx; | |
630 | }; | |
631 | ||
88654014 LM |
632 | static const char *_sys_dev_type[] = { |
633 | [SYS_DEV_UNKNOWN] = "Unknown", | |
634 | [SYS_DEV_SAS] = "SAS", | |
614902f6 | 635 | [SYS_DEV_SATA] = "SATA", |
60f0f54d | 636 | [SYS_DEV_NVME] = "NVMe", |
75350d87 KF |
637 | [SYS_DEV_VMD] = "VMD", |
638 | [SYS_DEV_SATA_VMD] = "SATA VMD" | |
88654014 LM |
639 | }; |
640 | ||
420dafcd N |
641 | static int no_platform = -1; |
642 | ||
643 | static int check_no_platform(void) | |
644 | { | |
645 | static const char search[] = "mdadm.imsm.test=1"; | |
646 | FILE *fp; | |
647 | ||
648 | if (no_platform >= 0) | |
649 | return no_platform; | |
650 | ||
651 | if (check_env("IMSM_NO_PLATFORM")) { | |
652 | no_platform = 1; | |
653 | return 1; | |
654 | } | |
655 | fp = fopen("/proc/cmdline", "r"); | |
656 | if (fp) { | |
657 | char *l = conf_line(fp); | |
658 | char *w = l; | |
659 | ||
cf1577bf MG |
660 | if (l == NULL) { |
661 | fclose(fp); | |
662 | return 0; | |
663 | } | |
664 | ||
420dafcd N |
665 | do { |
666 | if (strcmp(w, search) == 0) | |
667 | no_platform = 1; | |
668 | w = dl_next(w); | |
669 | } while (w != l); | |
670 | free_line(l); | |
671 | fclose(fp); | |
672 | if (no_platform >= 0) | |
673 | return no_platform; | |
674 | } | |
675 | no_platform = 0; | |
676 | return 0; | |
677 | } | |
678 | ||
679 | void imsm_set_no_platform(int v) | |
680 | { | |
681 | no_platform = v; | |
682 | } | |
683 | ||
88654014 LM |
684 | const char *get_sys_dev_type(enum sys_dev_type type) |
685 | { | |
686 | if (type >= SYS_DEV_MAX) | |
687 | type = SYS_DEV_UNKNOWN; | |
688 | ||
689 | return _sys_dev_type[type]; | |
690 | } | |
691 | ||
692 | static struct intel_hba * alloc_intel_hba(struct sys_dev *device) | |
693 | { | |
503975b9 N |
694 | struct intel_hba *result = xmalloc(sizeof(*result)); |
695 | ||
696 | result->type = device->type; | |
697 | result->path = xstrdup(device->path); | |
698 | result->next = NULL; | |
699 | if (result->path && (result->pci_id = strrchr(result->path, '/')) != NULL) | |
700 | result->pci_id++; | |
701 | ||
88654014 LM |
702 | return result; |
703 | } | |
704 | ||
705 | static struct intel_hba * find_intel_hba(struct intel_hba *hba, struct sys_dev *device) | |
706 | { | |
594dc1b8 JS |
707 | struct intel_hba *result; |
708 | ||
88654014 LM |
709 | for (result = hba; result; result = result->next) { |
710 | if (result->type == device->type && strcmp(result->path, device->path) == 0) | |
711 | break; | |
712 | } | |
713 | return result; | |
714 | } | |
715 | ||
b4cf4cba | 716 | static int attach_hba_to_super(struct intel_super *super, struct sys_dev *device) |
88654014 LM |
717 | { |
718 | struct intel_hba *hba; | |
719 | ||
720 | /* check if disk attached to Intel HBA */ | |
721 | hba = find_intel_hba(super->hba, device); | |
722 | if (hba != NULL) | |
723 | return 1; | |
724 | /* Check if HBA is already attached to super */ | |
725 | if (super->hba == NULL) { | |
726 | super->hba = alloc_intel_hba(device); | |
727 | return 1; | |
6b781d33 AP |
728 | } |
729 | ||
730 | hba = super->hba; | |
731 | /* Intel metadata allows for all disks attached to the same type HBA. | |
614902f6 | 732 | * Do not support HBA types mixing |
6b781d33 AP |
733 | */ |
734 | if (device->type != hba->type) | |
88654014 | 735 | return 2; |
6b781d33 AP |
736 | |
737 | /* Multiple same type HBAs can be used if they share the same OROM */ | |
738 | const struct imsm_orom *device_orom = get_orom_by_device_id(device->dev_id); | |
739 | ||
740 | if (device_orom != super->orom) | |
741 | return 2; | |
742 | ||
743 | while (hba->next) | |
744 | hba = hba->next; | |
745 | ||
746 | hba->next = alloc_intel_hba(device); | |
747 | return 1; | |
88654014 LM |
748 | } |
749 | ||
750 | static struct sys_dev* find_disk_attached_hba(int fd, const char *devname) | |
751 | { | |
9bc4ae77 | 752 | struct sys_dev *list, *elem; |
88654014 LM |
753 | char *disk_path; |
754 | ||
755 | if ((list = find_intel_devices()) == NULL) | |
756 | return 0; | |
757 | ||
4389ce73 | 758 | if (!is_fd_valid(fd)) |
88654014 LM |
759 | disk_path = (char *) devname; |
760 | else | |
7c798f87 | 761 | disk_path = diskfd_to_devpath(fd, 1, NULL); |
88654014 | 762 | |
9bc4ae77 | 763 | if (!disk_path) |
88654014 | 764 | return 0; |
88654014 | 765 | |
9bc4ae77 N |
766 | for (elem = list; elem; elem = elem->next) |
767 | if (path_attached_to_hba(disk_path, elem->path)) | |
5d2434d1 | 768 | break; |
9bc4ae77 | 769 | |
88654014 LM |
770 | if (disk_path != devname) |
771 | free(disk_path); | |
88654014 | 772 | |
5d2434d1 | 773 | return elem; |
88654014 LM |
774 | } |
775 | ||
d424212e N |
776 | static int find_intel_hba_capability(int fd, struct intel_super *super, |
777 | char *devname); | |
f2f5c343 | 778 | |
cdddbdbc DW |
779 | static struct supertype *match_metadata_desc_imsm(char *arg) |
780 | { | |
781 | struct supertype *st; | |
782 | ||
783 | if (strcmp(arg, "imsm") != 0 && | |
784 | strcmp(arg, "default") != 0 | |
785 | ) | |
786 | return NULL; | |
787 | ||
503975b9 | 788 | st = xcalloc(1, sizeof(*st)); |
cdddbdbc DW |
789 | st->ss = &super_imsm; |
790 | st->max_devs = IMSM_MAX_DEVICES; | |
791 | st->minor_version = 0; | |
792 | st->sb = NULL; | |
793 | return st; | |
794 | } | |
795 | ||
cdddbdbc DW |
796 | static __u8 *get_imsm_version(struct imsm_super *mpb) |
797 | { | |
798 | return &mpb->sig[MPB_SIG_LEN]; | |
799 | } | |
800 | ||
949c47a0 DW |
801 | /* retrieve a disk directly from the anchor when the anchor is known to be |
802 | * up-to-date, currently only at load time | |
803 | */ | |
804 | static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index) | |
cdddbdbc | 805 | { |
949c47a0 | 806 | if (index >= mpb->num_disks) |
cdddbdbc DW |
807 | return NULL; |
808 | return &mpb->disk[index]; | |
809 | } | |
810 | ||
95d07a2c LM |
811 | /* retrieve the disk description based on a index of the disk |
812 | * in the sub-array | |
813 | */ | |
814 | static struct dl *get_imsm_dl_disk(struct intel_super *super, __u8 index) | |
949c47a0 | 815 | { |
b9f594fe DW |
816 | struct dl *d; |
817 | ||
818 | for (d = super->disks; d; d = d->next) | |
819 | if (d->index == index) | |
95d07a2c LM |
820 | return d; |
821 | ||
822 | return NULL; | |
823 | } | |
824 | /* retrieve a disk from the parsed metadata */ | |
825 | static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index) | |
826 | { | |
827 | struct dl *dl; | |
828 | ||
829 | dl = get_imsm_dl_disk(super, index); | |
830 | if (dl) | |
831 | return &dl->disk; | |
832 | ||
b9f594fe | 833 | return NULL; |
949c47a0 DW |
834 | } |
835 | ||
836 | /* generate a checksum directly from the anchor when the anchor is known to be | |
837 | * up-to-date, currently only at load or write_super after coalescing | |
838 | */ | |
839 | static __u32 __gen_imsm_checksum(struct imsm_super *mpb) | |
cdddbdbc DW |
840 | { |
841 | __u32 end = mpb->mpb_size / sizeof(end); | |
842 | __u32 *p = (__u32 *) mpb; | |
843 | __u32 sum = 0; | |
844 | ||
5d500228 N |
845 | while (end--) { |
846 | sum += __le32_to_cpu(*p); | |
97f734fd N |
847 | p++; |
848 | } | |
cdddbdbc | 849 | |
5d500228 | 850 | return sum - __le32_to_cpu(mpb->check_sum); |
cdddbdbc DW |
851 | } |
852 | ||
a965f303 DW |
853 | static size_t sizeof_imsm_map(struct imsm_map *map) |
854 | { | |
855 | return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1); | |
856 | } | |
857 | ||
858 | struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map) | |
cdddbdbc | 859 | { |
5e7b0330 AK |
860 | /* A device can have 2 maps if it is in the middle of a migration. |
861 | * If second_map is: | |
238c0a71 AK |
862 | * MAP_0 - we return the first map |
863 | * MAP_1 - we return the second map if it exists, else NULL | |
864 | * MAP_X - we return the second map if it exists, else the first | |
5e7b0330 | 865 | */ |
a965f303 | 866 | struct imsm_map *map = &dev->vol.map[0]; |
9535fc47 | 867 | struct imsm_map *map2 = NULL; |
a965f303 | 868 | |
9535fc47 AK |
869 | if (dev->vol.migr_state) |
870 | map2 = (void *)map + sizeof_imsm_map(map); | |
a965f303 | 871 | |
9535fc47 | 872 | switch (second_map) { |
3b451610 | 873 | case MAP_0: |
9535fc47 | 874 | break; |
3b451610 | 875 | case MAP_1: |
9535fc47 AK |
876 | map = map2; |
877 | break; | |
238c0a71 | 878 | case MAP_X: |
9535fc47 AK |
879 | if (map2) |
880 | map = map2; | |
881 | break; | |
9535fc47 AK |
882 | default: |
883 | map = NULL; | |
884 | } | |
885 | return map; | |
5e7b0330 | 886 | |
a965f303 | 887 | } |
cdddbdbc | 888 | |
3393c6af DW |
889 | /* return the size of the device. |
890 | * migr_state increases the returned size if map[0] were to be duplicated | |
891 | */ | |
892 | static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state) | |
a965f303 DW |
893 | { |
894 | size_t size = sizeof(*dev) - sizeof(struct imsm_map) + | |
238c0a71 | 895 | sizeof_imsm_map(get_imsm_map(dev, MAP_0)); |
cdddbdbc DW |
896 | |
897 | /* migrating means an additional map */ | |
a965f303 | 898 | if (dev->vol.migr_state) |
238c0a71 | 899 | size += sizeof_imsm_map(get_imsm_map(dev, MAP_1)); |
3393c6af | 900 | else if (migr_state) |
238c0a71 | 901 | size += sizeof_imsm_map(get_imsm_map(dev, MAP_0)); |
cdddbdbc DW |
902 | |
903 | return size; | |
904 | } | |
905 | ||
54c2c1ea DW |
906 | /* retrieve disk serial number list from a metadata update */ |
907 | static struct disk_info *get_disk_info(struct imsm_update_create_array *update) | |
908 | { | |
909 | void *u = update; | |
910 | struct disk_info *inf; | |
911 | ||
912 | inf = u + sizeof(*update) - sizeof(struct imsm_dev) + | |
913 | sizeof_imsm_dev(&update->dev, 0); | |
914 | ||
915 | return inf; | |
916 | } | |
54c2c1ea | 917 | |
756a15f3 MG |
918 | /** |
919 | * __get_imsm_dev() - Get device with index from imsm_super. | |
920 | * @mpb: &imsm_super pointer, not NULL. | |
921 | * @index: Device index. | |
922 | * | |
923 | * Function works as non-NULL, aborting in such a case, | |
924 | * when NULL would be returned. | |
925 | * | |
926 | * Device index should be in range 0 up to num_raid_devs. | |
927 | * Function assumes the index was already verified. | |
928 | * Index must be valid, otherwise abort() is called. | |
929 | * | |
930 | * Return: Pointer to corresponding imsm_dev. | |
931 | * | |
932 | */ | |
949c47a0 | 933 | static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index) |
cdddbdbc DW |
934 | { |
935 | int offset; | |
936 | int i; | |
937 | void *_mpb = mpb; | |
938 | ||
949c47a0 | 939 | if (index >= mpb->num_raid_devs) |
756a15f3 | 940 | goto error; |
cdddbdbc DW |
941 | |
942 | /* devices start after all disks */ | |
943 | offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb; | |
944 | ||
756a15f3 | 945 | for (i = 0; i <= index; i++, offset += sizeof_imsm_dev(_mpb + offset, 0)) |
cdddbdbc DW |
946 | if (i == index) |
947 | return _mpb + offset; | |
756a15f3 MG |
948 | error: |
949 | pr_err("cannot find imsm_dev with index %u in imsm_super\n", index); | |
950 | abort(); | |
cdddbdbc DW |
951 | } |
952 | ||
756a15f3 MG |
953 | /** |
954 | * get_imsm_dev() - Get device with index from intel_super. | |
955 | * @super: &intel_super pointer, not NULL. | |
956 | * @index: Device index. | |
957 | * | |
958 | * Function works as non-NULL, aborting in such a case, | |
959 | * when NULL would be returned. | |
960 | * | |
961 | * Device index should be in range 0 up to num_raid_devs. | |
962 | * Function assumes the index was already verified. | |
963 | * Index must be valid, otherwise abort() is called. | |
964 | * | |
965 | * Return: Pointer to corresponding imsm_dev. | |
966 | * | |
967 | */ | |
949c47a0 DW |
968 | static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index) |
969 | { | |
ba2de7ba DW |
970 | struct intel_dev *dv; |
971 | ||
949c47a0 | 972 | if (index >= super->anchor->num_raid_devs) |
756a15f3 MG |
973 | goto error; |
974 | ||
ba2de7ba DW |
975 | for (dv = super->devlist; dv; dv = dv->next) |
976 | if (dv->index == index) | |
977 | return dv->dev; | |
756a15f3 MG |
978 | error: |
979 | pr_err("cannot find imsm_dev with index %u in intel_super\n", index); | |
980 | abort(); | |
949c47a0 DW |
981 | } |
982 | ||
8d67477f TM |
983 | static inline unsigned long long __le48_to_cpu(const struct bbm_log_block_addr |
984 | *addr) | |
985 | { | |
986 | return ((((__u64)__le32_to_cpu(addr->dw1)) << 16) | | |
987 | __le16_to_cpu(addr->w1)); | |
988 | } | |
989 | ||
990 | static inline struct bbm_log_block_addr __cpu_to_le48(unsigned long long sec) | |
991 | { | |
992 | struct bbm_log_block_addr addr; | |
993 | ||
994 | addr.w1 = __cpu_to_le16((__u16)(sec & 0xffff)); | |
995 | addr.dw1 = __cpu_to_le32((__u32)(sec >> 16) & 0xffffffff); | |
996 | return addr; | |
997 | } | |
998 | ||
8d67477f TM |
999 | /* get size of the bbm log */ |
1000 | static __u32 get_imsm_bbm_log_size(struct bbm_log *log) | |
1001 | { | |
1002 | if (!log || log->entry_count == 0) | |
1003 | return 0; | |
1004 | ||
1005 | return sizeof(log->signature) + | |
1006 | sizeof(log->entry_count) + | |
1007 | log->entry_count * sizeof(struct bbm_log_entry); | |
1008 | } | |
6f50473f TM |
1009 | |
1010 | /* check if bad block is not partially stored in bbm log */ | |
1011 | static int is_stored_in_bbm(struct bbm_log *log, const __u8 idx, const unsigned | |
1012 | long long sector, const int length, __u32 *pos) | |
1013 | { | |
1014 | __u32 i; | |
1015 | ||
1016 | for (i = *pos; i < log->entry_count; i++) { | |
1017 | struct bbm_log_entry *entry = &log->marked_block_entries[i]; | |
1018 | unsigned long long bb_start; | |
1019 | unsigned long long bb_end; | |
1020 | ||
1021 | bb_start = __le48_to_cpu(&entry->defective_block_start); | |
1022 | bb_end = bb_start + (entry->marked_count + 1); | |
1023 | ||
1024 | if ((entry->disk_ordinal == idx) && (bb_start >= sector) && | |
1025 | (bb_end <= sector + length)) { | |
1026 | *pos = i; | |
1027 | return 1; | |
1028 | } | |
1029 | } | |
1030 | return 0; | |
1031 | } | |
1032 | ||
1033 | /* record new bad block in bbm log */ | |
1034 | static int record_new_badblock(struct bbm_log *log, const __u8 idx, unsigned | |
1035 | long long sector, int length) | |
1036 | { | |
1037 | int new_bb = 0; | |
1038 | __u32 pos = 0; | |
1039 | struct bbm_log_entry *entry = NULL; | |
1040 | ||
1041 | while (is_stored_in_bbm(log, idx, sector, length, &pos)) { | |
1042 | struct bbm_log_entry *e = &log->marked_block_entries[pos]; | |
1043 | ||
1044 | if ((e->marked_count + 1 == BBM_LOG_MAX_LBA_ENTRY_VAL) && | |
1045 | (__le48_to_cpu(&e->defective_block_start) == sector)) { | |
1046 | sector += BBM_LOG_MAX_LBA_ENTRY_VAL; | |
1047 | length -= BBM_LOG_MAX_LBA_ENTRY_VAL; | |
1048 | pos = pos + 1; | |
1049 | continue; | |
1050 | } | |
1051 | entry = e; | |
1052 | break; | |
1053 | } | |
1054 | ||
1055 | if (entry) { | |
1056 | int cnt = (length <= BBM_LOG_MAX_LBA_ENTRY_VAL) ? length : | |
1057 | BBM_LOG_MAX_LBA_ENTRY_VAL; | |
1058 | entry->defective_block_start = __cpu_to_le48(sector); | |
1059 | entry->marked_count = cnt - 1; | |
1060 | if (cnt == length) | |
1061 | return 1; | |
1062 | sector += cnt; | |
1063 | length -= cnt; | |
1064 | } | |
1065 | ||
1066 | new_bb = ROUND_UP(length, BBM_LOG_MAX_LBA_ENTRY_VAL) / | |
1067 | BBM_LOG_MAX_LBA_ENTRY_VAL; | |
1068 | if (log->entry_count + new_bb > BBM_LOG_MAX_ENTRIES) | |
1069 | return 0; | |
1070 | ||
1071 | while (length > 0) { | |
1072 | int cnt = (length <= BBM_LOG_MAX_LBA_ENTRY_VAL) ? length : | |
1073 | BBM_LOG_MAX_LBA_ENTRY_VAL; | |
1074 | struct bbm_log_entry *entry = | |
1075 | &log->marked_block_entries[log->entry_count]; | |
1076 | ||
1077 | entry->defective_block_start = __cpu_to_le48(sector); | |
1078 | entry->marked_count = cnt - 1; | |
1079 | entry->disk_ordinal = idx; | |
1080 | ||
1081 | sector += cnt; | |
1082 | length -= cnt; | |
1083 | ||
1084 | log->entry_count++; | |
1085 | } | |
1086 | ||
1087 | return new_bb; | |
1088 | } | |
c07a5a4f | 1089 | |
4c9e8c1e TM |
1090 | /* clear all bad blocks for given disk */ |
1091 | static void clear_disk_badblocks(struct bbm_log *log, const __u8 idx) | |
1092 | { | |
1093 | __u32 i = 0; | |
1094 | ||
1095 | while (i < log->entry_count) { | |
1096 | struct bbm_log_entry *entries = log->marked_block_entries; | |
1097 | ||
1098 | if (entries[i].disk_ordinal == idx) { | |
1099 | if (i < log->entry_count - 1) | |
1100 | entries[i] = entries[log->entry_count - 1]; | |
1101 | log->entry_count--; | |
1102 | } else { | |
1103 | i++; | |
1104 | } | |
1105 | } | |
1106 | } | |
1107 | ||
c07a5a4f TM |
1108 | /* clear given bad block */ |
1109 | static int clear_badblock(struct bbm_log *log, const __u8 idx, const unsigned | |
1110 | long long sector, const int length) { | |
1111 | __u32 i = 0; | |
1112 | ||
1113 | while (i < log->entry_count) { | |
1114 | struct bbm_log_entry *entries = log->marked_block_entries; | |
1115 | ||
1116 | if ((entries[i].disk_ordinal == idx) && | |
1117 | (__le48_to_cpu(&entries[i].defective_block_start) == | |
1118 | sector) && (entries[i].marked_count + 1 == length)) { | |
1119 | if (i < log->entry_count - 1) | |
1120 | entries[i] = entries[log->entry_count - 1]; | |
1121 | log->entry_count--; | |
1122 | break; | |
1123 | } | |
1124 | i++; | |
1125 | } | |
1126 | ||
1127 | return 1; | |
1128 | } | |
8d67477f TM |
1129 | |
1130 | /* allocate and load BBM log from metadata */ | |
1131 | static int load_bbm_log(struct intel_super *super) | |
1132 | { | |
1133 | struct imsm_super *mpb = super->anchor; | |
1134 | __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size); | |
1135 | ||
1136 | super->bbm_log = xcalloc(1, sizeof(struct bbm_log)); | |
1137 | if (!super->bbm_log) | |
1138 | return 1; | |
1139 | ||
1140 | if (bbm_log_size) { | |
1141 | struct bbm_log *log = (void *)mpb + | |
1142 | __le32_to_cpu(mpb->mpb_size) - bbm_log_size; | |
1143 | ||
1144 | __u32 entry_count; | |
1145 | ||
1146 | if (bbm_log_size < sizeof(log->signature) + | |
1147 | sizeof(log->entry_count)) | |
1148 | return 2; | |
1149 | ||
1150 | entry_count = __le32_to_cpu(log->entry_count); | |
1151 | if ((__le32_to_cpu(log->signature) != BBM_LOG_SIGNATURE) || | |
1152 | (entry_count > BBM_LOG_MAX_ENTRIES)) | |
1153 | return 3; | |
1154 | ||
1155 | if (bbm_log_size != | |
1156 | sizeof(log->signature) + sizeof(log->entry_count) + | |
1157 | entry_count * sizeof(struct bbm_log_entry)) | |
1158 | return 4; | |
1159 | ||
1160 | memcpy(super->bbm_log, log, bbm_log_size); | |
1161 | } else { | |
1162 | super->bbm_log->signature = __cpu_to_le32(BBM_LOG_SIGNATURE); | |
1163 | super->bbm_log->entry_count = 0; | |
1164 | } | |
1165 | ||
1166 | return 0; | |
1167 | } | |
1168 | ||
b12796be TM |
1169 | /* checks if bad block is within volume boundaries */ |
1170 | static int is_bad_block_in_volume(const struct bbm_log_entry *entry, | |
1171 | const unsigned long long start_sector, | |
1172 | const unsigned long long size) | |
1173 | { | |
1174 | unsigned long long bb_start; | |
1175 | unsigned long long bb_end; | |
1176 | ||
1177 | bb_start = __le48_to_cpu(&entry->defective_block_start); | |
1178 | bb_end = bb_start + (entry->marked_count + 1); | |
1179 | ||
1180 | if (((bb_start >= start_sector) && (bb_start < start_sector + size)) || | |
1181 | ((bb_end >= start_sector) && (bb_end <= start_sector + size))) | |
1182 | return 1; | |
1183 | ||
1184 | return 0; | |
1185 | } | |
1186 | ||
1187 | /* get list of bad blocks on a drive for a volume */ | |
1188 | static void get_volume_badblocks(const struct bbm_log *log, const __u8 idx, | |
1189 | const unsigned long long start_sector, | |
1190 | const unsigned long long size, | |
1191 | struct md_bb *bbs) | |
1192 | { | |
1193 | __u32 count = 0; | |
1194 | __u32 i; | |
1195 | ||
1196 | for (i = 0; i < log->entry_count; i++) { | |
1197 | const struct bbm_log_entry *ent = | |
1198 | &log->marked_block_entries[i]; | |
1199 | struct md_bb_entry *bb; | |
1200 | ||
1201 | if ((ent->disk_ordinal == idx) && | |
1202 | is_bad_block_in_volume(ent, start_sector, size)) { | |
1203 | ||
1204 | if (!bbs->entries) { | |
1205 | bbs->entries = xmalloc(BBM_LOG_MAX_ENTRIES * | |
1206 | sizeof(*bb)); | |
1207 | if (!bbs->entries) | |
1208 | break; | |
1209 | } | |
1210 | ||
1211 | bb = &bbs->entries[count++]; | |
1212 | bb->sector = __le48_to_cpu(&ent->defective_block_start); | |
1213 | bb->length = ent->marked_count + 1; | |
1214 | } | |
1215 | } | |
1216 | bbs->count = count; | |
1217 | } | |
1218 | ||
98130f40 AK |
1219 | /* |
1220 | * for second_map: | |
238c0a71 AK |
1221 | * == MAP_0 get first map |
1222 | * == MAP_1 get second map | |
1223 | * == MAP_X than get map according to the current migr_state | |
98130f40 AK |
1224 | */ |
1225 | static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev, | |
1226 | int slot, | |
1227 | int second_map) | |
7eef0453 DW |
1228 | { |
1229 | struct imsm_map *map; | |
1230 | ||
5e7b0330 | 1231 | map = get_imsm_map(dev, second_map); |
7eef0453 | 1232 | |
ff077194 DW |
1233 | /* top byte identifies disk under rebuild */ |
1234 | return __le32_to_cpu(map->disk_ord_tbl[slot]); | |
1235 | } | |
1236 | ||
1237 | #define ord_to_idx(ord) (((ord) << 8) >> 8) | |
98130f40 | 1238 | static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot, int second_map) |
ff077194 | 1239 | { |
98130f40 | 1240 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, second_map); |
ff077194 DW |
1241 | |
1242 | return ord_to_idx(ord); | |
7eef0453 DW |
1243 | } |
1244 | ||
be73972f DW |
1245 | static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord) |
1246 | { | |
1247 | map->disk_ord_tbl[slot] = __cpu_to_le32(ord); | |
1248 | } | |
1249 | ||
76c152ca | 1250 | static int get_imsm_disk_slot(struct imsm_map *map, const unsigned int idx) |
620b1713 DW |
1251 | { |
1252 | int slot; | |
1253 | __u32 ord; | |
1254 | ||
1255 | for (slot = 0; slot < map->num_members; slot++) { | |
1256 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
1257 | if (ord_to_idx(ord) == idx) | |
1258 | return slot; | |
1259 | } | |
1260 | ||
76c152ca | 1261 | return IMSM_STATUS_ERROR; |
620b1713 DW |
1262 | } |
1263 | ||
cdddbdbc DW |
1264 | static int get_imsm_raid_level(struct imsm_map *map) |
1265 | { | |
1266 | if (map->raid_level == 1) { | |
1267 | if (map->num_members == 2) | |
1268 | return 1; | |
1269 | else | |
1270 | return 10; | |
1271 | } | |
1272 | ||
1273 | return map->raid_level; | |
1274 | } | |
1275 | ||
76c152ca MT |
1276 | /** |
1277 | * get_disk_slot_in_dev() - retrieve disk slot from &imsm_dev. | |
1278 | * @super: &intel_super pointer, not NULL. | |
1279 | * @dev_idx: imsm device index. | |
1280 | * @idx: disk index. | |
1281 | * | |
1282 | * Return: Slot on success, IMSM_STATUS_ERROR otherwise. | |
1283 | */ | |
1284 | static int get_disk_slot_in_dev(struct intel_super *super, const __u8 dev_idx, | |
1285 | const unsigned int idx) | |
1286 | { | |
1287 | struct imsm_dev *dev = get_imsm_dev(super, dev_idx); | |
1288 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
1289 | ||
1290 | return get_imsm_disk_slot(map, idx); | |
1291 | } | |
1292 | ||
c2c087e6 DW |
1293 | static int cmp_extent(const void *av, const void *bv) |
1294 | { | |
1295 | const struct extent *a = av; | |
1296 | const struct extent *b = bv; | |
1297 | if (a->start < b->start) | |
1298 | return -1; | |
1299 | if (a->start > b->start) | |
1300 | return 1; | |
1301 | return 0; | |
1302 | } | |
1303 | ||
0dcecb2e | 1304 | static int count_memberships(struct dl *dl, struct intel_super *super) |
c2c087e6 | 1305 | { |
c2c087e6 | 1306 | int memberships = 0; |
620b1713 | 1307 | int i; |
c2c087e6 | 1308 | |
76c152ca MT |
1309 | for (i = 0; i < super->anchor->num_raid_devs; i++) |
1310 | if (get_disk_slot_in_dev(super, i, dl->index) >= 0) | |
620b1713 | 1311 | memberships++; |
0dcecb2e DW |
1312 | |
1313 | return memberships; | |
1314 | } | |
1315 | ||
b81221b7 CA |
1316 | static __u32 imsm_min_reserved_sectors(struct intel_super *super); |
1317 | ||
486720e0 | 1318 | static int split_ull(unsigned long long n, void *lo, void *hi) |
5551b113 CA |
1319 | { |
1320 | if (lo == 0 || hi == 0) | |
1321 | return 1; | |
486720e0 JS |
1322 | __put_unaligned32(__cpu_to_le32((__u32)n), lo); |
1323 | __put_unaligned32(__cpu_to_le32((n >> 32)), hi); | |
5551b113 CA |
1324 | return 0; |
1325 | } | |
1326 | ||
1327 | static unsigned long long join_u32(__u32 lo, __u32 hi) | |
1328 | { | |
1329 | return (unsigned long long)__le32_to_cpu(lo) | | |
1330 | (((unsigned long long)__le32_to_cpu(hi)) << 32); | |
1331 | } | |
1332 | ||
1333 | static unsigned long long total_blocks(struct imsm_disk *disk) | |
1334 | { | |
1335 | if (disk == NULL) | |
1336 | return 0; | |
1337 | return join_u32(disk->total_blocks_lo, disk->total_blocks_hi); | |
1338 | } | |
1339 | ||
1c275381 MT |
1340 | /** |
1341 | * imsm_num_data_members() - get data drives count for an array. | |
1342 | * @map: Map to analyze. | |
1343 | * | |
1344 | * num_data_members value represents minimal count of drives for level. | |
1345 | * The name of the property could be misleading for RAID5 with asymmetric layout | |
1346 | * because some data required to be calculated from parity. | |
1347 | * The property is extracted from level and num_members value. | |
1348 | * | |
1349 | * Return: num_data_members value on success, zero otherwise. | |
1350 | */ | |
1351 | static __u8 imsm_num_data_members(struct imsm_map *map) | |
1352 | { | |
1353 | switch (get_imsm_raid_level(map)) { | |
1354 | case 0: | |
1355 | return map->num_members; | |
1356 | case 1: | |
1357 | case 10: | |
1358 | return map->num_members / 2; | |
1359 | case 5: | |
1360 | return map->num_members - 1; | |
1361 | default: | |
1362 | dprintf("unsupported raid level\n"); | |
1363 | return 0; | |
1364 | } | |
1365 | } | |
1366 | ||
5551b113 CA |
1367 | static unsigned long long pba_of_lba0(struct imsm_map *map) |
1368 | { | |
1369 | if (map == NULL) | |
1370 | return 0; | |
1371 | return join_u32(map->pba_of_lba0_lo, map->pba_of_lba0_hi); | |
1372 | } | |
1373 | ||
1374 | static unsigned long long blocks_per_member(struct imsm_map *map) | |
1375 | { | |
1376 | if (map == NULL) | |
1377 | return 0; | |
1378 | return join_u32(map->blocks_per_member_lo, map->blocks_per_member_hi); | |
1379 | } | |
1380 | ||
1381 | static unsigned long long num_data_stripes(struct imsm_map *map) | |
1382 | { | |
1383 | if (map == NULL) | |
1384 | return 0; | |
1385 | return join_u32(map->num_data_stripes_lo, map->num_data_stripes_hi); | |
1386 | } | |
1387 | ||
4036e7ee MT |
1388 | static unsigned long long vol_curr_migr_unit(struct imsm_dev *dev) |
1389 | { | |
1390 | if (dev == NULL) | |
1391 | return 0; | |
1392 | ||
1393 | return join_u32(dev->vol.curr_migr_unit_lo, dev->vol.curr_migr_unit_hi); | |
1394 | } | |
1395 | ||
fcc2c9da MD |
1396 | static unsigned long long imsm_dev_size(struct imsm_dev *dev) |
1397 | { | |
1398 | if (dev == NULL) | |
1399 | return 0; | |
1400 | return join_u32(dev->size_low, dev->size_high); | |
1401 | } | |
1402 | ||
9f421827 PB |
1403 | static unsigned long long migr_chkp_area_pba(struct migr_record *migr_rec) |
1404 | { | |
1405 | if (migr_rec == NULL) | |
1406 | return 0; | |
1407 | return join_u32(migr_rec->ckpt_area_pba_lo, | |
1408 | migr_rec->ckpt_area_pba_hi); | |
1409 | } | |
1410 | ||
1411 | static unsigned long long current_migr_unit(struct migr_record *migr_rec) | |
1412 | { | |
1413 | if (migr_rec == NULL) | |
1414 | return 0; | |
1415 | return join_u32(migr_rec->curr_migr_unit_lo, | |
1416 | migr_rec->curr_migr_unit_hi); | |
1417 | } | |
1418 | ||
1419 | static unsigned long long migr_dest_1st_member_lba(struct migr_record *migr_rec) | |
1420 | { | |
1421 | if (migr_rec == NULL) | |
1422 | return 0; | |
1423 | return join_u32(migr_rec->dest_1st_member_lba_lo, | |
1424 | migr_rec->dest_1st_member_lba_hi); | |
1425 | } | |
1426 | ||
1427 | static unsigned long long get_num_migr_units(struct migr_record *migr_rec) | |
1428 | { | |
1429 | if (migr_rec == NULL) | |
1430 | return 0; | |
1431 | return join_u32(migr_rec->num_migr_units_lo, | |
1432 | migr_rec->num_migr_units_hi); | |
1433 | } | |
1434 | ||
5551b113 CA |
1435 | static void set_total_blocks(struct imsm_disk *disk, unsigned long long n) |
1436 | { | |
1437 | split_ull(n, &disk->total_blocks_lo, &disk->total_blocks_hi); | |
1438 | } | |
1439 | ||
1c275381 MT |
1440 | /** |
1441 | * set_num_domains() - Set number of domains for an array. | |
1442 | * @map: Map to be updated. | |
1443 | * | |
1444 | * num_domains property represents copies count of each data drive, thus make | |
1445 | * it meaningful only for RAID1 and RAID10. IMSM supports two domains for | |
1446 | * raid1 and raid10. | |
1447 | */ | |
1448 | static void set_num_domains(struct imsm_map *map) | |
1449 | { | |
1450 | int level = get_imsm_raid_level(map); | |
1451 | ||
1452 | if (level == 1 || level == 10) | |
1453 | map->num_domains = 2; | |
1454 | else | |
1455 | map->num_domains = 1; | |
1456 | } | |
1457 | ||
5551b113 CA |
1458 | static void set_pba_of_lba0(struct imsm_map *map, unsigned long long n) |
1459 | { | |
1460 | split_ull(n, &map->pba_of_lba0_lo, &map->pba_of_lba0_hi); | |
1461 | } | |
1462 | ||
1463 | static void set_blocks_per_member(struct imsm_map *map, unsigned long long n) | |
1464 | { | |
1465 | split_ull(n, &map->blocks_per_member_lo, &map->blocks_per_member_hi); | |
1466 | } | |
1467 | ||
1468 | static void set_num_data_stripes(struct imsm_map *map, unsigned long long n) | |
1469 | { | |
1470 | split_ull(n, &map->num_data_stripes_lo, &map->num_data_stripes_hi); | |
1471 | } | |
1472 | ||
1c275381 MT |
1473 | /** |
1474 | * update_num_data_stripes() - Calculate and update num_data_stripes value. | |
1475 | * @map: map to be updated. | |
1476 | * @dev_size: size of volume. | |
1477 | * | |
1478 | * num_data_stripes value is addictionally divided by num_domains, therefore for | |
1479 | * levels where num_domains is not 1, nds is a part of real value. | |
1480 | */ | |
1481 | static void update_num_data_stripes(struct imsm_map *map, | |
1482 | unsigned long long dev_size) | |
1483 | { | |
1484 | unsigned long long nds = dev_size / imsm_num_data_members(map); | |
1485 | ||
1486 | nds /= map->num_domains; | |
1487 | nds /= map->blocks_per_strip; | |
1488 | set_num_data_stripes(map, nds); | |
1489 | } | |
1490 | ||
4036e7ee MT |
1491 | static void set_vol_curr_migr_unit(struct imsm_dev *dev, unsigned long long n) |
1492 | { | |
1493 | if (dev == NULL) | |
1494 | return; | |
1495 | ||
1496 | split_ull(n, &dev->vol.curr_migr_unit_lo, &dev->vol.curr_migr_unit_hi); | |
1497 | } | |
1498 | ||
fcc2c9da MD |
1499 | static void set_imsm_dev_size(struct imsm_dev *dev, unsigned long long n) |
1500 | { | |
1501 | split_ull(n, &dev->size_low, &dev->size_high); | |
1502 | } | |
1503 | ||
9f421827 PB |
1504 | static void set_migr_chkp_area_pba(struct migr_record *migr_rec, |
1505 | unsigned long long n) | |
1506 | { | |
1507 | split_ull(n, &migr_rec->ckpt_area_pba_lo, &migr_rec->ckpt_area_pba_hi); | |
1508 | } | |
1509 | ||
1510 | static void set_current_migr_unit(struct migr_record *migr_rec, | |
1511 | unsigned long long n) | |
1512 | { | |
1513 | split_ull(n, &migr_rec->curr_migr_unit_lo, | |
1514 | &migr_rec->curr_migr_unit_hi); | |
1515 | } | |
1516 | ||
1517 | static void set_migr_dest_1st_member_lba(struct migr_record *migr_rec, | |
1518 | unsigned long long n) | |
1519 | { | |
1520 | split_ull(n, &migr_rec->dest_1st_member_lba_lo, | |
1521 | &migr_rec->dest_1st_member_lba_hi); | |
1522 | } | |
1523 | ||
1524 | static void set_num_migr_units(struct migr_record *migr_rec, | |
1525 | unsigned long long n) | |
1526 | { | |
1527 | split_ull(n, &migr_rec->num_migr_units_lo, | |
1528 | &migr_rec->num_migr_units_hi); | |
1529 | } | |
1530 | ||
44490938 MD |
1531 | static unsigned long long per_dev_array_size(struct imsm_map *map) |
1532 | { | |
1533 | unsigned long long array_size = 0; | |
1534 | ||
1535 | if (map == NULL) | |
1536 | return array_size; | |
1537 | ||
1538 | array_size = num_data_stripes(map) * map->blocks_per_strip; | |
1539 | if (get_imsm_raid_level(map) == 1 || get_imsm_raid_level(map) == 10) | |
1540 | array_size *= 2; | |
1541 | ||
1542 | return array_size; | |
1543 | } | |
1544 | ||
05501181 PB |
1545 | static struct extent *get_extents(struct intel_super *super, struct dl *dl, |
1546 | int get_minimal_reservation) | |
0dcecb2e DW |
1547 | { |
1548 | /* find a list of used extents on the given physical device */ | |
0dcecb2e | 1549 | int memberships = count_memberships(dl, super); |
aa19fdd4 MT |
1550 | struct extent *rv = xcalloc(memberships + 1, sizeof(struct extent)); |
1551 | struct extent *e = rv; | |
1552 | int i; | |
b276dd33 DW |
1553 | __u32 reservation; |
1554 | ||
1555 | /* trim the reserved area for spares, so they can join any array | |
1556 | * regardless of whether the OROM has assigned sectors from the | |
1557 | * IMSM_RESERVED_SECTORS region | |
1558 | */ | |
05501181 | 1559 | if (dl->index == -1 || get_minimal_reservation) |
b81221b7 | 1560 | reservation = imsm_min_reserved_sectors(super); |
b276dd33 DW |
1561 | else |
1562 | reservation = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
0dcecb2e | 1563 | |
949c47a0 DW |
1564 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
1565 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 1566 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
c2c087e6 | 1567 | |
620b1713 | 1568 | if (get_imsm_disk_slot(map, dl->index) >= 0) { |
5551b113 | 1569 | e->start = pba_of_lba0(map); |
44490938 | 1570 | e->size = per_dev_array_size(map); |
aa19fdd4 | 1571 | e->vol = i; |
620b1713 | 1572 | e++; |
c2c087e6 DW |
1573 | } |
1574 | } | |
1575 | qsort(rv, memberships, sizeof(*rv), cmp_extent); | |
1576 | ||
1011e834 | 1577 | /* determine the start of the metadata |
14e8215b DW |
1578 | * when no raid devices are defined use the default |
1579 | * ...otherwise allow the metadata to truncate the value | |
1580 | * as is the case with older versions of imsm | |
1581 | */ | |
1582 | if (memberships) { | |
1583 | struct extent *last = &rv[memberships - 1]; | |
5551b113 | 1584 | unsigned long long remainder; |
14e8215b | 1585 | |
5551b113 | 1586 | remainder = total_blocks(&dl->disk) - (last->start + last->size); |
dda5855f DW |
1587 | /* round down to 1k block to satisfy precision of the kernel |
1588 | * 'size' interface | |
1589 | */ | |
1590 | remainder &= ~1UL; | |
1591 | /* make sure remainder is still sane */ | |
f21e18ca | 1592 | if (remainder < (unsigned)ROUND_UP(super->len, 512) >> 9) |
dda5855f | 1593 | remainder = ROUND_UP(super->len, 512) >> 9; |
14e8215b DW |
1594 | if (reservation > remainder) |
1595 | reservation = remainder; | |
1596 | } | |
5551b113 | 1597 | e->start = total_blocks(&dl->disk) - reservation; |
c2c087e6 DW |
1598 | e->size = 0; |
1599 | return rv; | |
1600 | } | |
1601 | ||
14e8215b DW |
1602 | /* try to determine how much space is reserved for metadata from |
1603 | * the last get_extents() entry, otherwise fallback to the | |
1604 | * default | |
1605 | */ | |
1606 | static __u32 imsm_reserved_sectors(struct intel_super *super, struct dl *dl) | |
1607 | { | |
1608 | struct extent *e; | |
1609 | int i; | |
1610 | __u32 rv; | |
1611 | ||
1612 | /* for spares just return a minimal reservation which will grow | |
1613 | * once the spare is picked up by an array | |
1614 | */ | |
1615 | if (dl->index == -1) | |
1616 | return MPB_SECTOR_CNT; | |
1617 | ||
05501181 | 1618 | e = get_extents(super, dl, 0); |
14e8215b DW |
1619 | if (!e) |
1620 | return MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
1621 | ||
1622 | /* scroll to last entry */ | |
1623 | for (i = 0; e[i].size; i++) | |
1624 | continue; | |
1625 | ||
5551b113 | 1626 | rv = total_blocks(&dl->disk) - e[i].start; |
14e8215b DW |
1627 | |
1628 | free(e); | |
1629 | ||
1630 | return rv; | |
1631 | } | |
1632 | ||
25ed7e59 DW |
1633 | static int is_spare(struct imsm_disk *disk) |
1634 | { | |
1635 | return (disk->status & SPARE_DISK) == SPARE_DISK; | |
1636 | } | |
1637 | ||
1638 | static int is_configured(struct imsm_disk *disk) | |
1639 | { | |
1640 | return (disk->status & CONFIGURED_DISK) == CONFIGURED_DISK; | |
1641 | } | |
1642 | ||
1643 | static int is_failed(struct imsm_disk *disk) | |
1644 | { | |
1645 | return (disk->status & FAILED_DISK) == FAILED_DISK; | |
1646 | } | |
1647 | ||
2432ce9b AP |
1648 | static int is_journal(struct imsm_disk *disk) |
1649 | { | |
1650 | return (disk->status & JOURNAL_DISK) == JOURNAL_DISK; | |
1651 | } | |
1652 | ||
78c8028b MT |
1653 | /** |
1654 | * round_member_size_to_mb()- Round given size to closest MiB. | |
1655 | * @size: size to round in sectors. | |
b53bfba6 | 1656 | */ |
78c8028b | 1657 | static inline unsigned long long round_member_size_to_mb(unsigned long long size) |
b53bfba6 | 1658 | { |
78c8028b MT |
1659 | return (size >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT; |
1660 | } | |
b53bfba6 | 1661 | |
78c8028b MT |
1662 | /** |
1663 | * round_size_to_mb()- Round given size. | |
1664 | * @array_size: size to round in sectors. | |
1665 | * @disk_count: count of data members. | |
1666 | * | |
1667 | * Get size per each data member and round it to closest MiB to ensure that data | |
1668 | * splits evenly between members. | |
1669 | * | |
1670 | * Return: Array size, rounded down. | |
1671 | */ | |
1672 | static inline unsigned long long round_size_to_mb(unsigned long long array_size, | |
1673 | unsigned int disk_count) | |
1674 | { | |
1675 | return round_member_size_to_mb(array_size / disk_count) * disk_count; | |
b53bfba6 TM |
1676 | } |
1677 | ||
8b9cd157 MK |
1678 | static int able_to_resync(int raid_level, int missing_disks) |
1679 | { | |
1680 | int max_missing_disks = 0; | |
1681 | ||
1682 | switch (raid_level) { | |
1683 | case 10: | |
1684 | max_missing_disks = 1; | |
1685 | break; | |
1686 | default: | |
1687 | max_missing_disks = 0; | |
1688 | } | |
1689 | return missing_disks <= max_missing_disks; | |
1690 | } | |
1691 | ||
b81221b7 CA |
1692 | /* try to determine how much space is reserved for metadata from |
1693 | * the last get_extents() entry on the smallest active disk, | |
1694 | * otherwise fallback to the default | |
1695 | */ | |
1696 | static __u32 imsm_min_reserved_sectors(struct intel_super *super) | |
1697 | { | |
1698 | struct extent *e; | |
1699 | int i; | |
5551b113 CA |
1700 | unsigned long long min_active; |
1701 | __u32 remainder; | |
b81221b7 CA |
1702 | __u32 rv = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; |
1703 | struct dl *dl, *dl_min = NULL; | |
1704 | ||
1705 | if (!super) | |
1706 | return rv; | |
1707 | ||
1708 | min_active = 0; | |
1709 | for (dl = super->disks; dl; dl = dl->next) { | |
1710 | if (dl->index < 0) | |
1711 | continue; | |
5551b113 CA |
1712 | unsigned long long blocks = total_blocks(&dl->disk); |
1713 | if (blocks < min_active || min_active == 0) { | |
b81221b7 | 1714 | dl_min = dl; |
5551b113 | 1715 | min_active = blocks; |
b81221b7 CA |
1716 | } |
1717 | } | |
1718 | if (!dl_min) | |
1719 | return rv; | |
1720 | ||
1721 | /* find last lba used by subarrays on the smallest active disk */ | |
05501181 | 1722 | e = get_extents(super, dl_min, 0); |
b81221b7 CA |
1723 | if (!e) |
1724 | return rv; | |
1725 | for (i = 0; e[i].size; i++) | |
1726 | continue; | |
1727 | ||
1728 | remainder = min_active - e[i].start; | |
1729 | free(e); | |
1730 | ||
1731 | /* to give priority to recovery we should not require full | |
1732 | IMSM_RESERVED_SECTORS from the spare */ | |
1733 | rv = MPB_SECTOR_CNT + NUM_BLOCKS_DIRTY_STRIPE_REGION; | |
1734 | ||
1735 | /* if real reservation is smaller use that value */ | |
1736 | return (remainder < rv) ? remainder : rv; | |
1737 | } | |
1738 | ||
195d1d76 | 1739 | static bool is_gen_migration(struct imsm_dev *dev); |
d1e02575 | 1740 | |
f36a9ecd PB |
1741 | #define IMSM_4K_DIV 8 |
1742 | ||
c47b0ff6 AK |
1743 | static __u64 blocks_per_migr_unit(struct intel_super *super, |
1744 | struct imsm_dev *dev); | |
1e5c6983 | 1745 | |
c47b0ff6 AK |
1746 | static void print_imsm_dev(struct intel_super *super, |
1747 | struct imsm_dev *dev, | |
1748 | char *uuid, | |
1749 | int disk_idx) | |
cdddbdbc DW |
1750 | { |
1751 | __u64 sz; | |
0d80bb2f | 1752 | int slot, i; |
238c0a71 AK |
1753 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1754 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); | |
b10b37b8 | 1755 | __u32 ord; |
cdddbdbc DW |
1756 | |
1757 | printf("\n"); | |
1e7bc0ed | 1758 | printf("[%.16s]:\n", dev->volume); |
ba1b3bc8 | 1759 | printf(" Subarray : %d\n", super->current_vol); |
44470971 | 1760 | printf(" UUID : %s\n", uuid); |
dd8bcb3b AK |
1761 | printf(" RAID Level : %d", get_imsm_raid_level(map)); |
1762 | if (map2) | |
1763 | printf(" <-- %d", get_imsm_raid_level(map2)); | |
1764 | printf("\n"); | |
1765 | printf(" Members : %d", map->num_members); | |
1766 | if (map2) | |
1767 | printf(" <-- %d", map2->num_members); | |
1768 | printf("\n"); | |
0d80bb2f DW |
1769 | printf(" Slots : ["); |
1770 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 1771 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_0); |
0d80bb2f DW |
1772 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
1773 | } | |
dd8bcb3b AK |
1774 | printf("]"); |
1775 | if (map2) { | |
1776 | printf(" <-- ["); | |
1777 | for (i = 0; i < map2->num_members; i++) { | |
238c0a71 | 1778 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_1); |
dd8bcb3b AK |
1779 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
1780 | } | |
1781 | printf("]"); | |
1782 | } | |
1783 | printf("\n"); | |
7095bccb AK |
1784 | printf(" Failed disk : "); |
1785 | if (map->failed_disk_num == 0xff) | |
b823c8f9 | 1786 | printf(STR_COMMON_NONE); |
7095bccb AK |
1787 | else |
1788 | printf("%i", map->failed_disk_num); | |
1789 | printf("\n"); | |
620b1713 DW |
1790 | slot = get_imsm_disk_slot(map, disk_idx); |
1791 | if (slot >= 0) { | |
238c0a71 | 1792 | ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); |
b10b37b8 DW |
1793 | printf(" This Slot : %d%s\n", slot, |
1794 | ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : ""); | |
1795 | } else | |
cdddbdbc | 1796 | printf(" This Slot : ?\n"); |
84918897 | 1797 | printf(" Sector Size : %u\n", super->sector_size); |
fcc2c9da | 1798 | sz = imsm_dev_size(dev); |
84918897 MK |
1799 | printf(" Array Size : %llu%s\n", |
1800 | (unsigned long long)sz * 512 / super->sector_size, | |
cdddbdbc | 1801 | human_size(sz * 512)); |
5551b113 | 1802 | sz = blocks_per_member(map); |
84918897 MK |
1803 | printf(" Per Dev Size : %llu%s\n", |
1804 | (unsigned long long)sz * 512 / super->sector_size, | |
cdddbdbc | 1805 | human_size(sz * 512)); |
5551b113 | 1806 | printf(" Sector Offset : %llu\n", |
7d8935cb | 1807 | pba_of_lba0(map) * 512 / super->sector_size); |
5551b113 CA |
1808 | printf(" Num Stripes : %llu\n", |
1809 | num_data_stripes(map)); | |
dd8bcb3b | 1810 | printf(" Chunk Size : %u KiB", |
cdddbdbc | 1811 | __le16_to_cpu(map->blocks_per_strip) / 2); |
dd8bcb3b AK |
1812 | if (map2) |
1813 | printf(" <-- %u KiB", | |
1814 | __le16_to_cpu(map2->blocks_per_strip) / 2); | |
1815 | printf("\n"); | |
cdddbdbc | 1816 | printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks)); |
8655a7b1 | 1817 | printf(" Migrate State : "); |
1484e727 DW |
1818 | if (dev->vol.migr_state) { |
1819 | if (migr_type(dev) == MIGR_INIT) | |
8655a7b1 | 1820 | printf("initialize\n"); |
1484e727 | 1821 | else if (migr_type(dev) == MIGR_REBUILD) |
8655a7b1 | 1822 | printf("rebuild\n"); |
1484e727 | 1823 | else if (migr_type(dev) == MIGR_VERIFY) |
8655a7b1 | 1824 | printf("check\n"); |
1484e727 | 1825 | else if (migr_type(dev) == MIGR_GEN_MIGR) |
8655a7b1 | 1826 | printf("general migration\n"); |
1484e727 | 1827 | else if (migr_type(dev) == MIGR_STATE_CHANGE) |
8655a7b1 | 1828 | printf("state change\n"); |
1484e727 | 1829 | else if (migr_type(dev) == MIGR_REPAIR) |
8655a7b1 | 1830 | printf("repair\n"); |
1484e727 | 1831 | else |
8655a7b1 DW |
1832 | printf("<unknown:%d>\n", migr_type(dev)); |
1833 | } else | |
1834 | printf("idle\n"); | |
3393c6af DW |
1835 | printf(" Map State : %s", map_state_str[map->map_state]); |
1836 | if (dev->vol.migr_state) { | |
238c0a71 | 1837 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 | 1838 | |
b10b37b8 | 1839 | printf(" <-- %s", map_state_str[map->map_state]); |
4036e7ee | 1840 | printf("\n Checkpoint : %llu ", vol_curr_migr_unit(dev)); |
089f9d79 | 1841 | if (is_gen_migration(dev) && (slot > 1 || slot < 0)) |
464d40e8 LD |
1842 | printf("(N/A)"); |
1843 | else | |
1844 | printf("(%llu)", (unsigned long long) | |
1845 | blocks_per_migr_unit(super, dev)); | |
3393c6af DW |
1846 | } |
1847 | printf("\n"); | |
2432ce9b AP |
1848 | printf(" Dirty State : %s\n", (dev->vol.dirty & RAIDVOL_DIRTY) ? |
1849 | "dirty" : "clean"); | |
1850 | printf(" RWH Policy : "); | |
c2462068 | 1851 | if (dev->rwh_policy == RWH_OFF || dev->rwh_policy == RWH_MULTIPLE_OFF) |
2432ce9b AP |
1852 | printf("off\n"); |
1853 | else if (dev->rwh_policy == RWH_DISTRIBUTED) | |
1854 | printf("PPL distributed\n"); | |
1855 | else if (dev->rwh_policy == RWH_JOURNALING_DRIVE) | |
1856 | printf("PPL journaling drive\n"); | |
c2462068 PB |
1857 | else if (dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED) |
1858 | printf("Multiple distributed PPLs\n"); | |
1859 | else if (dev->rwh_policy == RWH_MULTIPLE_PPLS_JOURNALING_DRIVE) | |
1860 | printf("Multiple PPLs on journaling drive\n"); | |
fbc42556 JR |
1861 | else if (dev->rwh_policy == RWH_BITMAP) |
1862 | printf("Write-intent bitmap\n"); | |
2432ce9b AP |
1863 | else |
1864 | printf("<unknown:%d>\n", dev->rwh_policy); | |
ba1b3bc8 AP |
1865 | |
1866 | printf(" Volume ID : %u\n", dev->my_vol_raid_dev_num); | |
cdddbdbc DW |
1867 | } |
1868 | ||
ef5c214e MK |
1869 | static void print_imsm_disk(struct imsm_disk *disk, |
1870 | int index, | |
1871 | __u32 reserved, | |
1872 | unsigned int sector_size) { | |
1f24f035 | 1873 | char str[MAX_RAID_SERIAL_LEN + 1]; |
cdddbdbc DW |
1874 | __u64 sz; |
1875 | ||
0ec1f4e8 | 1876 | if (index < -1 || !disk) |
e9d82038 DW |
1877 | return; |
1878 | ||
cdddbdbc | 1879 | printf("\n"); |
1f24f035 | 1880 | snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial); |
0ec1f4e8 DW |
1881 | if (index >= 0) |
1882 | printf(" Disk%02d Serial : %s\n", index, str); | |
1883 | else | |
1884 | printf(" Disk Serial : %s\n", str); | |
2432ce9b AP |
1885 | printf(" State :%s%s%s%s\n", is_spare(disk) ? " spare" : "", |
1886 | is_configured(disk) ? " active" : "", | |
1887 | is_failed(disk) ? " failed" : "", | |
1888 | is_journal(disk) ? " journal" : ""); | |
cdddbdbc | 1889 | printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id)); |
5551b113 | 1890 | sz = total_blocks(disk) - reserved; |
ef5c214e MK |
1891 | printf(" Usable Size : %llu%s\n", |
1892 | (unsigned long long)sz * 512 / sector_size, | |
cdddbdbc DW |
1893 | human_size(sz * 512)); |
1894 | } | |
1895 | ||
de44e46f PB |
1896 | void convert_to_4k_imsm_migr_rec(struct intel_super *super) |
1897 | { | |
1898 | struct migr_record *migr_rec = super->migr_rec; | |
1899 | ||
1900 | migr_rec->blocks_per_unit /= IMSM_4K_DIV; | |
de44e46f PB |
1901 | migr_rec->dest_depth_per_unit /= IMSM_4K_DIV; |
1902 | split_ull((join_u32(migr_rec->post_migr_vol_cap, | |
1903 | migr_rec->post_migr_vol_cap_hi) / IMSM_4K_DIV), | |
1904 | &migr_rec->post_migr_vol_cap, &migr_rec->post_migr_vol_cap_hi); | |
9f421827 PB |
1905 | set_migr_chkp_area_pba(migr_rec, |
1906 | migr_chkp_area_pba(migr_rec) / IMSM_4K_DIV); | |
1907 | set_migr_dest_1st_member_lba(migr_rec, | |
1908 | migr_dest_1st_member_lba(migr_rec) / IMSM_4K_DIV); | |
de44e46f PB |
1909 | } |
1910 | ||
f36a9ecd PB |
1911 | void convert_to_4k_imsm_disk(struct imsm_disk *disk) |
1912 | { | |
1913 | set_total_blocks(disk, (total_blocks(disk)/IMSM_4K_DIV)); | |
1914 | } | |
1915 | ||
1916 | void convert_to_4k(struct intel_super *super) | |
1917 | { | |
1918 | struct imsm_super *mpb = super->anchor; | |
1919 | struct imsm_disk *disk; | |
1920 | int i; | |
e4467bc7 | 1921 | __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size); |
f36a9ecd PB |
1922 | |
1923 | for (i = 0; i < mpb->num_disks ; i++) { | |
1924 | disk = __get_imsm_disk(mpb, i); | |
1925 | /* disk */ | |
1926 | convert_to_4k_imsm_disk(disk); | |
1927 | } | |
1928 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
1929 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
1930 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
1931 | /* dev */ | |
fcc2c9da | 1932 | set_imsm_dev_size(dev, imsm_dev_size(dev)/IMSM_4K_DIV); |
4036e7ee MT |
1933 | set_vol_curr_migr_unit(dev, |
1934 | vol_curr_migr_unit(dev) / IMSM_4K_DIV); | |
f36a9ecd PB |
1935 | |
1936 | /* map0 */ | |
1937 | set_blocks_per_member(map, blocks_per_member(map)/IMSM_4K_DIV); | |
1938 | map->blocks_per_strip /= IMSM_4K_DIV; | |
1939 | set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV); | |
1940 | ||
1941 | if (dev->vol.migr_state) { | |
1942 | /* map1 */ | |
1943 | map = get_imsm_map(dev, MAP_1); | |
1944 | set_blocks_per_member(map, | |
1945 | blocks_per_member(map)/IMSM_4K_DIV); | |
1946 | map->blocks_per_strip /= IMSM_4K_DIV; | |
1947 | set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV); | |
1948 | } | |
1949 | } | |
e4467bc7 TM |
1950 | if (bbm_log_size) { |
1951 | struct bbm_log *log = (void *)mpb + | |
1952 | __le32_to_cpu(mpb->mpb_size) - bbm_log_size; | |
1953 | __u32 i; | |
1954 | ||
1955 | for (i = 0; i < log->entry_count; i++) { | |
1956 | struct bbm_log_entry *entry = | |
1957 | &log->marked_block_entries[i]; | |
1958 | ||
1959 | __u8 count = entry->marked_count + 1; | |
1960 | unsigned long long sector = | |
1961 | __le48_to_cpu(&entry->defective_block_start); | |
1962 | ||
1963 | entry->defective_block_start = | |
1964 | __cpu_to_le48(sector/IMSM_4K_DIV); | |
1965 | entry->marked_count = max(count/IMSM_4K_DIV, 1) - 1; | |
1966 | } | |
1967 | } | |
f36a9ecd PB |
1968 | |
1969 | mpb->check_sum = __gen_imsm_checksum(mpb); | |
1970 | } | |
1971 | ||
520e69e2 AK |
1972 | void examine_migr_rec_imsm(struct intel_super *super) |
1973 | { | |
1974 | struct migr_record *migr_rec = super->migr_rec; | |
1975 | struct imsm_super *mpb = super->anchor; | |
1976 | int i; | |
1977 | ||
1978 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
1979 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
3136abe5 | 1980 | struct imsm_map *map; |
b4ab44d8 | 1981 | int slot = -1; |
3136abe5 | 1982 | |
195d1d76 | 1983 | if (is_gen_migration(dev) == false) |
520e69e2 AK |
1984 | continue; |
1985 | ||
1986 | printf("\nMigration Record Information:"); | |
3136abe5 | 1987 | |
44bfe6df AK |
1988 | /* first map under migration */ |
1989 | map = get_imsm_map(dev, MAP_0); | |
76c152ca | 1990 | |
3136abe5 AK |
1991 | if (map) |
1992 | slot = get_imsm_disk_slot(map, super->disks->index); | |
089f9d79 | 1993 | if (map == NULL || slot > 1 || slot < 0) { |
520e69e2 AK |
1994 | printf(" Empty\n "); |
1995 | printf("Examine one of first two disks in array\n"); | |
1996 | break; | |
1997 | } | |
1998 | printf("\n Status : "); | |
1999 | if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL) | |
2000 | printf("Normal\n"); | |
2001 | else | |
2002 | printf("Contains Data\n"); | |
9f421827 PB |
2003 | printf(" Current Unit : %llu\n", |
2004 | current_migr_unit(migr_rec)); | |
520e69e2 AK |
2005 | printf(" Family : %u\n", |
2006 | __le32_to_cpu(migr_rec->family_num)); | |
2007 | printf(" Ascending : %u\n", | |
2008 | __le32_to_cpu(migr_rec->ascending_migr)); | |
2009 | printf(" Blocks Per Unit : %u\n", | |
2010 | __le32_to_cpu(migr_rec->blocks_per_unit)); | |
2011 | printf(" Dest. Depth Per Unit : %u\n", | |
2012 | __le32_to_cpu(migr_rec->dest_depth_per_unit)); | |
9f421827 PB |
2013 | printf(" Checkpoint Area pba : %llu\n", |
2014 | migr_chkp_area_pba(migr_rec)); | |
2015 | printf(" First member lba : %llu\n", | |
2016 | migr_dest_1st_member_lba(migr_rec)); | |
2017 | printf(" Total Number of Units : %llu\n", | |
2018 | get_num_migr_units(migr_rec)); | |
2019 | printf(" Size of volume : %llu\n", | |
2020 | join_u32(migr_rec->post_migr_vol_cap, | |
2021 | migr_rec->post_migr_vol_cap_hi)); | |
520e69e2 AK |
2022 | printf(" Record was read from : %u\n", |
2023 | __le32_to_cpu(migr_rec->ckpt_read_disk_num)); | |
2024 | ||
2025 | break; | |
2026 | } | |
2027 | } | |
f36a9ecd | 2028 | |
de44e46f PB |
2029 | void convert_from_4k_imsm_migr_rec(struct intel_super *super) |
2030 | { | |
2031 | struct migr_record *migr_rec = super->migr_rec; | |
2032 | ||
2033 | migr_rec->blocks_per_unit *= IMSM_4K_DIV; | |
de44e46f PB |
2034 | migr_rec->dest_depth_per_unit *= IMSM_4K_DIV; |
2035 | split_ull((join_u32(migr_rec->post_migr_vol_cap, | |
2036 | migr_rec->post_migr_vol_cap_hi) * IMSM_4K_DIV), | |
2037 | &migr_rec->post_migr_vol_cap, | |
2038 | &migr_rec->post_migr_vol_cap_hi); | |
9f421827 PB |
2039 | set_migr_chkp_area_pba(migr_rec, |
2040 | migr_chkp_area_pba(migr_rec) * IMSM_4K_DIV); | |
2041 | set_migr_dest_1st_member_lba(migr_rec, | |
2042 | migr_dest_1st_member_lba(migr_rec) * IMSM_4K_DIV); | |
de44e46f PB |
2043 | } |
2044 | ||
f36a9ecd PB |
2045 | void convert_from_4k(struct intel_super *super) |
2046 | { | |
2047 | struct imsm_super *mpb = super->anchor; | |
2048 | struct imsm_disk *disk; | |
2049 | int i; | |
e4467bc7 | 2050 | __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size); |
f36a9ecd PB |
2051 | |
2052 | for (i = 0; i < mpb->num_disks ; i++) { | |
2053 | disk = __get_imsm_disk(mpb, i); | |
2054 | /* disk */ | |
2055 | set_total_blocks(disk, (total_blocks(disk)*IMSM_4K_DIV)); | |
2056 | } | |
2057 | ||
2058 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
2059 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
2060 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
2061 | /* dev */ | |
fcc2c9da | 2062 | set_imsm_dev_size(dev, imsm_dev_size(dev)*IMSM_4K_DIV); |
4036e7ee MT |
2063 | set_vol_curr_migr_unit(dev, |
2064 | vol_curr_migr_unit(dev) * IMSM_4K_DIV); | |
f36a9ecd PB |
2065 | |
2066 | /* map0 */ | |
2067 | set_blocks_per_member(map, blocks_per_member(map)*IMSM_4K_DIV); | |
2068 | map->blocks_per_strip *= IMSM_4K_DIV; | |
2069 | set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV); | |
2070 | ||
2071 | if (dev->vol.migr_state) { | |
2072 | /* map1 */ | |
2073 | map = get_imsm_map(dev, MAP_1); | |
2074 | set_blocks_per_member(map, | |
2075 | blocks_per_member(map)*IMSM_4K_DIV); | |
2076 | map->blocks_per_strip *= IMSM_4K_DIV; | |
2077 | set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV); | |
2078 | } | |
2079 | } | |
e4467bc7 TM |
2080 | if (bbm_log_size) { |
2081 | struct bbm_log *log = (void *)mpb + | |
2082 | __le32_to_cpu(mpb->mpb_size) - bbm_log_size; | |
2083 | __u32 i; | |
2084 | ||
2085 | for (i = 0; i < log->entry_count; i++) { | |
2086 | struct bbm_log_entry *entry = | |
2087 | &log->marked_block_entries[i]; | |
2088 | ||
2089 | __u8 count = entry->marked_count + 1; | |
2090 | unsigned long long sector = | |
2091 | __le48_to_cpu(&entry->defective_block_start); | |
2092 | ||
2093 | entry->defective_block_start = | |
2094 | __cpu_to_le48(sector*IMSM_4K_DIV); | |
2095 | entry->marked_count = count*IMSM_4K_DIV - 1; | |
2096 | } | |
2097 | } | |
f36a9ecd PB |
2098 | |
2099 | mpb->check_sum = __gen_imsm_checksum(mpb); | |
2100 | } | |
2101 | ||
19482bcc AK |
2102 | /******************************************************************************* |
2103 | * function: imsm_check_attributes | |
2104 | * Description: Function checks if features represented by attributes flags | |
1011e834 | 2105 | * are supported by mdadm. |
19482bcc AK |
2106 | * Parameters: |
2107 | * attributes - Attributes read from metadata | |
2108 | * Returns: | |
1011e834 N |
2109 | * 0 - passed attributes contains unsupported features flags |
2110 | * 1 - all features are supported | |
19482bcc AK |
2111 | ******************************************************************************/ |
2112 | static int imsm_check_attributes(__u32 attributes) | |
2113 | { | |
2114 | int ret_val = 1; | |
418f9b36 N |
2115 | __u32 not_supported = MPB_ATTRIB_SUPPORTED^0xffffffff; |
2116 | ||
2117 | not_supported &= ~MPB_ATTRIB_IGNORED; | |
19482bcc AK |
2118 | |
2119 | not_supported &= attributes; | |
2120 | if (not_supported) { | |
e7b84f9d | 2121 | pr_err("(IMSM): Unsupported attributes : %x\n", |
418f9b36 | 2122 | (unsigned)__le32_to_cpu(not_supported)); |
19482bcc AK |
2123 | if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) { |
2124 | dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY \n"); | |
2125 | not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY; | |
2126 | } | |
2127 | if (not_supported & MPB_ATTRIB_2TB) { | |
2128 | dprintf("\t\tMPB_ATTRIB_2TB\n"); | |
2129 | not_supported ^= MPB_ATTRIB_2TB; | |
2130 | } | |
2131 | if (not_supported & MPB_ATTRIB_RAID0) { | |
2132 | dprintf("\t\tMPB_ATTRIB_RAID0\n"); | |
2133 | not_supported ^= MPB_ATTRIB_RAID0; | |
2134 | } | |
2135 | if (not_supported & MPB_ATTRIB_RAID1) { | |
2136 | dprintf("\t\tMPB_ATTRIB_RAID1\n"); | |
2137 | not_supported ^= MPB_ATTRIB_RAID1; | |
2138 | } | |
2139 | if (not_supported & MPB_ATTRIB_RAID10) { | |
2140 | dprintf("\t\tMPB_ATTRIB_RAID10\n"); | |
2141 | not_supported ^= MPB_ATTRIB_RAID10; | |
2142 | } | |
2143 | if (not_supported & MPB_ATTRIB_RAID1E) { | |
2144 | dprintf("\t\tMPB_ATTRIB_RAID1E\n"); | |
2145 | not_supported ^= MPB_ATTRIB_RAID1E; | |
2146 | } | |
2147 | if (not_supported & MPB_ATTRIB_RAID5) { | |
2148 | dprintf("\t\tMPB_ATTRIB_RAID5\n"); | |
2149 | not_supported ^= MPB_ATTRIB_RAID5; | |
2150 | } | |
2151 | if (not_supported & MPB_ATTRIB_RAIDCNG) { | |
2152 | dprintf("\t\tMPB_ATTRIB_RAIDCNG\n"); | |
2153 | not_supported ^= MPB_ATTRIB_RAIDCNG; | |
2154 | } | |
2155 | if (not_supported & MPB_ATTRIB_BBM) { | |
2156 | dprintf("\t\tMPB_ATTRIB_BBM\n"); | |
2157 | not_supported ^= MPB_ATTRIB_BBM; | |
2158 | } | |
2159 | if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) { | |
2160 | dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY (== MPB_ATTRIB_LEGACY)\n"); | |
2161 | not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY; | |
2162 | } | |
2163 | if (not_supported & MPB_ATTRIB_EXP_STRIPE_SIZE) { | |
2164 | dprintf("\t\tMPB_ATTRIB_EXP_STRIP_SIZE\n"); | |
2165 | not_supported ^= MPB_ATTRIB_EXP_STRIPE_SIZE; | |
2166 | } | |
2167 | if (not_supported & MPB_ATTRIB_2TB_DISK) { | |
2168 | dprintf("\t\tMPB_ATTRIB_2TB_DISK\n"); | |
2169 | not_supported ^= MPB_ATTRIB_2TB_DISK; | |
2170 | } | |
2171 | if (not_supported & MPB_ATTRIB_NEVER_USE2) { | |
2172 | dprintf("\t\tMPB_ATTRIB_NEVER_USE2\n"); | |
2173 | not_supported ^= MPB_ATTRIB_NEVER_USE2; | |
2174 | } | |
2175 | if (not_supported & MPB_ATTRIB_NEVER_USE) { | |
2176 | dprintf("\t\tMPB_ATTRIB_NEVER_USE\n"); | |
2177 | not_supported ^= MPB_ATTRIB_NEVER_USE; | |
2178 | } | |
2179 | ||
2180 | if (not_supported) | |
1ade5cc1 | 2181 | dprintf("(IMSM): Unknown attributes : %x\n", not_supported); |
19482bcc AK |
2182 | |
2183 | ret_val = 0; | |
2184 | } | |
2185 | ||
2186 | return ret_val; | |
2187 | } | |
2188 | ||
a5d85af7 | 2189 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map); |
44470971 | 2190 | |
cdddbdbc DW |
2191 | static void examine_super_imsm(struct supertype *st, char *homehost) |
2192 | { | |
2193 | struct intel_super *super = st->sb; | |
949c47a0 | 2194 | struct imsm_super *mpb = super->anchor; |
cdddbdbc DW |
2195 | char str[MAX_SIGNATURE_LENGTH]; |
2196 | int i; | |
27fd6274 DW |
2197 | struct mdinfo info; |
2198 | char nbuf[64]; | |
cdddbdbc | 2199 | __u32 sum; |
14e8215b | 2200 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
94827db3 | 2201 | struct dl *dl; |
e48aed3c | 2202 | time_t creation_time; |
27fd6274 | 2203 | |
618f4e6d XN |
2204 | strncpy(str, (char *)mpb->sig, MPB_SIG_LEN); |
2205 | str[MPB_SIG_LEN-1] = '\0'; | |
cdddbdbc | 2206 | printf(" Magic : %s\n", str); |
cdddbdbc | 2207 | printf(" Version : %s\n", get_imsm_version(mpb)); |
148acb7b | 2208 | printf(" Orig Family : %08x\n", __le32_to_cpu(mpb->orig_family_num)); |
cdddbdbc DW |
2209 | printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num)); |
2210 | printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num)); | |
e48aed3c AP |
2211 | creation_time = __le64_to_cpu(mpb->creation_time); |
2212 | printf(" Creation Time : %.24s\n", | |
2213 | creation_time ? ctime(&creation_time) : "Unknown"); | |
19482bcc AK |
2214 | printf(" Attributes : "); |
2215 | if (imsm_check_attributes(mpb->attributes)) | |
2216 | printf("All supported\n"); | |
2217 | else | |
2218 | printf("not supported\n"); | |
a5d85af7 | 2219 | getinfo_super_imsm(st, &info, NULL); |
ba65d917 | 2220 | fname_from_uuid(&info, nbuf); |
27fd6274 | 2221 | printf(" UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
2222 | sum = __le32_to_cpu(mpb->check_sum); |
2223 | printf(" Checksum : %08x %s\n", sum, | |
949c47a0 | 2224 | __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect"); |
f36a9ecd | 2225 | printf(" MPB Sectors : %d\n", mpb_sectors(mpb, super->sector_size)); |
cdddbdbc DW |
2226 | printf(" Disks : %d\n", mpb->num_disks); |
2227 | printf(" RAID Devices : %d\n", mpb->num_raid_devs); | |
ef5c214e MK |
2228 | print_imsm_disk(__get_imsm_disk(mpb, super->disks->index), |
2229 | super->disks->index, reserved, super->sector_size); | |
8d67477f | 2230 | if (get_imsm_bbm_log_size(super->bbm_log)) { |
604b746f JD |
2231 | struct bbm_log *log = super->bbm_log; |
2232 | ||
2233 | printf("\n"); | |
2234 | printf("Bad Block Management Log:\n"); | |
2235 | printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size)); | |
2236 | printf(" Signature : %x\n", __le32_to_cpu(log->signature)); | |
2237 | printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count)); | |
604b746f | 2238 | } |
44470971 DW |
2239 | for (i = 0; i < mpb->num_raid_devs; i++) { |
2240 | struct mdinfo info; | |
2241 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
2242 | ||
2243 | super->current_vol = i; | |
a5d85af7 | 2244 | getinfo_super_imsm(st, &info, NULL); |
ba65d917 | 2245 | fname_from_uuid(&info, nbuf); |
c47b0ff6 | 2246 | print_imsm_dev(super, dev, nbuf + 5, super->disks->index); |
44470971 | 2247 | } |
cdddbdbc DW |
2248 | for (i = 0; i < mpb->num_disks; i++) { |
2249 | if (i == super->disks->index) | |
2250 | continue; | |
ef5c214e MK |
2251 | print_imsm_disk(__get_imsm_disk(mpb, i), i, reserved, |
2252 | super->sector_size); | |
cdddbdbc | 2253 | } |
94827db3 | 2254 | |
0ec1f4e8 DW |
2255 | for (dl = super->disks; dl; dl = dl->next) |
2256 | if (dl->index == -1) | |
ef5c214e MK |
2257 | print_imsm_disk(&dl->disk, -1, reserved, |
2258 | super->sector_size); | |
520e69e2 AK |
2259 | |
2260 | examine_migr_rec_imsm(super); | |
cdddbdbc DW |
2261 | } |
2262 | ||
061f2c6a | 2263 | static void brief_examine_super_imsm(struct supertype *st, int verbose) |
cdddbdbc | 2264 | { |
27fd6274 | 2265 | /* We just write a generic IMSM ARRAY entry */ |
ff54de6e N |
2266 | struct mdinfo info; |
2267 | char nbuf[64]; | |
2268 | ||
a5d85af7 | 2269 | getinfo_super_imsm(st, &info, NULL); |
ba65d917 | 2270 | fname_from_uuid(&info, nbuf); |
4737ae25 N |
2271 | printf("ARRAY metadata=imsm UUID=%s\n", nbuf + 5); |
2272 | } | |
2273 | ||
2274 | static void brief_examine_subarrays_imsm(struct supertype *st, int verbose) | |
2275 | { | |
2276 | /* We just write a generic IMSM ARRAY entry */ | |
2277 | struct mdinfo info; | |
2278 | char nbuf[64]; | |
2279 | char nbuf1[64]; | |
2280 | struct intel_super *super = st->sb; | |
2281 | int i; | |
2282 | ||
2283 | if (!super->anchor->num_raid_devs) | |
2284 | return; | |
2285 | ||
a5d85af7 | 2286 | getinfo_super_imsm(st, &info, NULL); |
ba65d917 | 2287 | fname_from_uuid(&info, nbuf); |
1e7bc0ed DW |
2288 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
2289 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
2290 | ||
2291 | super->current_vol = i; | |
a5d85af7 | 2292 | getinfo_super_imsm(st, &info, NULL); |
ba65d917 | 2293 | fname_from_uuid(&info, nbuf1); |
b9ce7ab0 | 2294 | printf("ARRAY " DEV_MD_DIR "%.16s container=%s member=%d UUID=%s\n", |
cf8de691 | 2295 | dev->volume, nbuf + 5, i, nbuf1 + 5); |
1e7bc0ed | 2296 | } |
cdddbdbc DW |
2297 | } |
2298 | ||
9d84c8ea DW |
2299 | static void export_examine_super_imsm(struct supertype *st) |
2300 | { | |
2301 | struct intel_super *super = st->sb; | |
2302 | struct imsm_super *mpb = super->anchor; | |
2303 | struct mdinfo info; | |
2304 | char nbuf[64]; | |
2305 | ||
a5d85af7 | 2306 | getinfo_super_imsm(st, &info, NULL); |
ba65d917 | 2307 | fname_from_uuid(&info, nbuf); |
9d84c8ea DW |
2308 | printf("MD_METADATA=imsm\n"); |
2309 | printf("MD_LEVEL=container\n"); | |
2310 | printf("MD_UUID=%s\n", nbuf+5); | |
2311 | printf("MD_DEVICES=%u\n", mpb->num_disks); | |
e48aed3c | 2312 | printf("MD_CREATION_TIME=%llu\n", __le64_to_cpu(mpb->creation_time)); |
9d84c8ea DW |
2313 | } |
2314 | ||
b771faef BK |
2315 | static void detail_super_imsm(struct supertype *st, char *homehost, |
2316 | char *subarray) | |
cdddbdbc | 2317 | { |
3ebe00a1 DW |
2318 | struct mdinfo info; |
2319 | char nbuf[64]; | |
b771faef BK |
2320 | struct intel_super *super = st->sb; |
2321 | int temp_vol = super->current_vol; | |
2322 | ||
2323 | if (subarray) | |
2324 | super->current_vol = strtoul(subarray, NULL, 10); | |
3ebe00a1 | 2325 | |
a5d85af7 | 2326 | getinfo_super_imsm(st, &info, NULL); |
ba65d917 | 2327 | fname_from_uuid(&info, nbuf); |
65884368 | 2328 | printf("\n UUID : %s\n", nbuf + 5); |
b771faef BK |
2329 | |
2330 | super->current_vol = temp_vol; | |
cdddbdbc DW |
2331 | } |
2332 | ||
b771faef | 2333 | static void brief_detail_super_imsm(struct supertype *st, char *subarray) |
cdddbdbc | 2334 | { |
ff54de6e N |
2335 | struct mdinfo info; |
2336 | char nbuf[64]; | |
b771faef BK |
2337 | struct intel_super *super = st->sb; |
2338 | int temp_vol = super->current_vol; | |
2339 | ||
2340 | if (subarray) | |
2341 | super->current_vol = strtoul(subarray, NULL, 10); | |
2342 | ||
a5d85af7 | 2343 | getinfo_super_imsm(st, &info, NULL); |
ba65d917 | 2344 | fname_from_uuid(&info, nbuf); |
ff54de6e | 2345 | printf(" UUID=%s", nbuf + 5); |
b771faef BK |
2346 | |
2347 | super->current_vol = temp_vol; | |
cdddbdbc | 2348 | } |
d665cc31 | 2349 | |
6da53c0e BK |
2350 | static int imsm_read_serial(int fd, char *devname, __u8 *serial, |
2351 | size_t serial_buf_len); | |
d665cc31 DW |
2352 | static void fd2devname(int fd, char *name); |
2353 | ||
120dc887 | 2354 | static int ahci_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose) |
d665cc31 | 2355 | { |
120dc887 LM |
2356 | /* dump an unsorted list of devices attached to AHCI Intel storage |
2357 | * controller, as well as non-connected ports | |
d665cc31 DW |
2358 | */ |
2359 | int hba_len = strlen(hba_path) + 1; | |
2360 | struct dirent *ent; | |
2361 | DIR *dir; | |
2362 | char *path = NULL; | |
2363 | int err = 0; | |
2364 | unsigned long port_mask = (1 << port_count) - 1; | |
2365 | ||
f21e18ca | 2366 | if (port_count > (int)sizeof(port_mask) * 8) { |
ba728be7 | 2367 | if (verbose > 0) |
e7b84f9d | 2368 | pr_err("port_count %d out of range\n", port_count); |
d665cc31 DW |
2369 | return 2; |
2370 | } | |
2371 | ||
2372 | /* scroll through /sys/dev/block looking for devices attached to | |
2373 | * this hba | |
2374 | */ | |
2375 | dir = opendir("/sys/dev/block"); | |
1a6dd6b9 PB |
2376 | if (!dir) |
2377 | return 1; | |
2378 | ||
2379 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
d665cc31 DW |
2380 | int fd; |
2381 | char model[64]; | |
2382 | char vendor[64]; | |
2383 | char buf[1024]; | |
2384 | int major, minor; | |
fcebeb77 | 2385 | char device[PATH_MAX]; |
d665cc31 DW |
2386 | char *c; |
2387 | int port; | |
2388 | int type; | |
2389 | ||
2390 | if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2) | |
2391 | continue; | |
7c798f87 | 2392 | path = devt_to_devpath(makedev(major, minor), 1, NULL); |
d665cc31 DW |
2393 | if (!path) |
2394 | continue; | |
2395 | if (!path_attached_to_hba(path, hba_path)) { | |
2396 | free(path); | |
2397 | path = NULL; | |
2398 | continue; | |
2399 | } | |
2400 | ||
fcebeb77 MT |
2401 | /* retrieve the scsi device */ |
2402 | if (!devt_to_devpath(makedev(major, minor), 1, device)) { | |
ba728be7 | 2403 | if (verbose > 0) |
fcebeb77 | 2404 | pr_err("failed to get device\n"); |
d665cc31 DW |
2405 | err = 2; |
2406 | break; | |
2407 | } | |
fcebeb77 | 2408 | if (devpath_to_char(device, "type", buf, sizeof(buf), 0)) { |
d665cc31 | 2409 | err = 2; |
d665cc31 DW |
2410 | break; |
2411 | } | |
2412 | type = strtoul(buf, NULL, 10); | |
2413 | ||
2414 | /* if it's not a disk print the vendor and model */ | |
2415 | if (!(type == 0 || type == 7 || type == 14)) { | |
2416 | vendor[0] = '\0'; | |
2417 | model[0] = '\0'; | |
fcebeb77 MT |
2418 | |
2419 | if (devpath_to_char(device, "vendor", buf, | |
2420 | sizeof(buf), 0) == 0) { | |
d665cc31 DW |
2421 | strncpy(vendor, buf, sizeof(vendor)); |
2422 | vendor[sizeof(vendor) - 1] = '\0'; | |
2423 | c = (char *) &vendor[sizeof(vendor) - 1]; | |
2424 | while (isspace(*c) || *c == '\0') | |
2425 | *c-- = '\0'; | |
2426 | ||
2427 | } | |
fcebeb77 MT |
2428 | |
2429 | if (devpath_to_char(device, "model", buf, | |
2430 | sizeof(buf), 0) == 0) { | |
d665cc31 DW |
2431 | strncpy(model, buf, sizeof(model)); |
2432 | model[sizeof(model) - 1] = '\0'; | |
2433 | c = (char *) &model[sizeof(model) - 1]; | |
2434 | while (isspace(*c) || *c == '\0') | |
2435 | *c-- = '\0'; | |
2436 | } | |
2437 | ||
2438 | if (vendor[0] && model[0]) | |
2439 | sprintf(buf, "%.64s %.64s", vendor, model); | |
2440 | else | |
2441 | switch (type) { /* numbers from hald/linux/device.c */ | |
2442 | case 1: sprintf(buf, "tape"); break; | |
2443 | case 2: sprintf(buf, "printer"); break; | |
2444 | case 3: sprintf(buf, "processor"); break; | |
2445 | case 4: | |
2446 | case 5: sprintf(buf, "cdrom"); break; | |
2447 | case 6: sprintf(buf, "scanner"); break; | |
2448 | case 8: sprintf(buf, "media_changer"); break; | |
2449 | case 9: sprintf(buf, "comm"); break; | |
2450 | case 12: sprintf(buf, "raid"); break; | |
2451 | default: sprintf(buf, "unknown"); | |
2452 | } | |
2453 | } else | |
2454 | buf[0] = '\0'; | |
d665cc31 DW |
2455 | |
2456 | /* chop device path to 'host%d' and calculate the port number */ | |
2457 | c = strchr(&path[hba_len], '/'); | |
4e5e717d | 2458 | if (!c) { |
ba728be7 | 2459 | if (verbose > 0) |
e7b84f9d | 2460 | pr_err("%s - invalid path name\n", path + hba_len); |
4e5e717d AW |
2461 | err = 2; |
2462 | break; | |
2463 | } | |
d665cc31 | 2464 | *c = '\0'; |
0858eccf AP |
2465 | if ((sscanf(&path[hba_len], "ata%d", &port) == 1) || |
2466 | ((sscanf(&path[hba_len], "host%d", &port) == 1))) | |
d665cc31 DW |
2467 | port -= host_base; |
2468 | else { | |
ba728be7 | 2469 | if (verbose > 0) { |
d665cc31 | 2470 | *c = '/'; /* repair the full string */ |
e7b84f9d | 2471 | pr_err("failed to determine port number for %s\n", |
d665cc31 DW |
2472 | path); |
2473 | } | |
2474 | err = 2; | |
2475 | break; | |
2476 | } | |
2477 | ||
2478 | /* mark this port as used */ | |
2479 | port_mask &= ~(1 << port); | |
2480 | ||
2481 | /* print out the device information */ | |
2482 | if (buf[0]) { | |
2483 | printf(" Port%d : - non-disk device (%s) -\n", port, buf); | |
2484 | continue; | |
2485 | } | |
2486 | ||
2487 | fd = dev_open(ent->d_name, O_RDONLY); | |
4389ce73 | 2488 | if (!is_fd_valid(fd)) |
d665cc31 DW |
2489 | printf(" Port%d : - disk info unavailable -\n", port); |
2490 | else { | |
2491 | fd2devname(fd, buf); | |
2492 | printf(" Port%d : %s", port, buf); | |
6da53c0e BK |
2493 | if (imsm_read_serial(fd, NULL, (__u8 *)buf, |
2494 | sizeof(buf)) == 0) | |
2495 | printf(" (%s)\n", buf); | |
d665cc31 | 2496 | else |
664d5325 | 2497 | printf(" ()\n"); |
4dab422a | 2498 | close(fd); |
d665cc31 | 2499 | } |
d665cc31 DW |
2500 | free(path); |
2501 | path = NULL; | |
2502 | } | |
2503 | if (path) | |
2504 | free(path); | |
2505 | if (dir) | |
2506 | closedir(dir); | |
2507 | if (err == 0) { | |
2508 | int i; | |
2509 | ||
2510 | for (i = 0; i < port_count; i++) | |
2511 | if (port_mask & (1 << i)) | |
2512 | printf(" Port%d : - no device attached -\n", i); | |
2513 | } | |
2514 | ||
2515 | return err; | |
2516 | } | |
2517 | ||
6da53c0e | 2518 | static int print_nvme_info(struct sys_dev *hba) |
60f0f54d PB |
2519 | { |
2520 | struct dirent *ent; | |
2521 | DIR *dir; | |
60f0f54d | 2522 | |
6da53c0e | 2523 | dir = opendir("/sys/block/"); |
b9135011 | 2524 | if (!dir) |
b5eece69 | 2525 | return 1; |
b9135011 JS |
2526 | |
2527 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
8662f92d MT |
2528 | char ns_path[PATH_MAX]; |
2529 | char cntrl_path[PATH_MAX]; | |
2530 | char buf[PATH_MAX]; | |
2531 | int fd = -1; | |
60f0f54d | 2532 | |
8662f92d MT |
2533 | if (!strstr(ent->d_name, "nvme")) |
2534 | goto skip; | |
d835518b | 2535 | |
8662f92d | 2536 | fd = open_dev(ent->d_name); |
4389ce73 | 2537 | if (!is_fd_valid(fd)) |
8662f92d | 2538 | goto skip; |
d835518b | 2539 | |
8662f92d MT |
2540 | if (!diskfd_to_devpath(fd, 0, ns_path) || |
2541 | !diskfd_to_devpath(fd, 1, cntrl_path)) | |
2542 | goto skip; | |
2543 | ||
2544 | if (!path_attached_to_hba(cntrl_path, hba->path)) | |
2545 | goto skip; | |
2546 | ||
2547 | if (!imsm_is_nvme_namespace_supported(fd, 0)) | |
2548 | goto skip; | |
2549 | ||
2550 | fd2devname(fd, buf); | |
2551 | if (hba->type == SYS_DEV_VMD) | |
2552 | printf(" NVMe under VMD : %s", buf); | |
2553 | else if (hba->type == SYS_DEV_NVME) | |
2554 | printf(" NVMe Device : %s", buf); | |
2555 | ||
2556 | if (!imsm_read_serial(fd, NULL, (__u8 *)buf, | |
2557 | sizeof(buf))) | |
2558 | printf(" (%s)\n", buf); | |
2559 | else | |
2560 | printf("()\n"); | |
2561 | ||
2562 | skip: | |
4389ce73 | 2563 | close_fd(&fd); |
60f0f54d PB |
2564 | } |
2565 | ||
b9135011 | 2566 | closedir(dir); |
b5eece69 | 2567 | return 0; |
60f0f54d PB |
2568 | } |
2569 | ||
120dc887 LM |
2570 | static void print_found_intel_controllers(struct sys_dev *elem) |
2571 | { | |
2572 | for (; elem; elem = elem->next) { | |
e7b84f9d | 2573 | pr_err("found Intel(R) "); |
120dc887 LM |
2574 | if (elem->type == SYS_DEV_SATA) |
2575 | fprintf(stderr, "SATA "); | |
155cbb4c LM |
2576 | else if (elem->type == SYS_DEV_SAS) |
2577 | fprintf(stderr, "SAS "); | |
0858eccf AP |
2578 | else if (elem->type == SYS_DEV_NVME) |
2579 | fprintf(stderr, "NVMe "); | |
60f0f54d PB |
2580 | |
2581 | if (elem->type == SYS_DEV_VMD) | |
2582 | fprintf(stderr, "VMD domain"); | |
75350d87 KF |
2583 | else if (elem->type == SYS_DEV_SATA_VMD) |
2584 | fprintf(stderr, "SATA VMD domain"); | |
60f0f54d PB |
2585 | else |
2586 | fprintf(stderr, "RAID controller"); | |
2587 | ||
120dc887 LM |
2588 | if (elem->pci_id) |
2589 | fprintf(stderr, " at %s", elem->pci_id); | |
2590 | fprintf(stderr, ".\n"); | |
2591 | } | |
2592 | fflush(stderr); | |
2593 | } | |
2594 | ||
120dc887 LM |
2595 | static int ahci_get_port_count(const char *hba_path, int *port_count) |
2596 | { | |
2597 | struct dirent *ent; | |
2598 | DIR *dir; | |
2599 | int host_base = -1; | |
2600 | ||
2601 | *port_count = 0; | |
2602 | if ((dir = opendir(hba_path)) == NULL) | |
2603 | return -1; | |
2604 | ||
2605 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
2606 | int host; | |
2607 | ||
0858eccf AP |
2608 | if ((sscanf(ent->d_name, "ata%d", &host) != 1) && |
2609 | ((sscanf(ent->d_name, "host%d", &host) != 1))) | |
120dc887 LM |
2610 | continue; |
2611 | if (*port_count == 0) | |
2612 | host_base = host; | |
2613 | else if (host < host_base) | |
2614 | host_base = host; | |
2615 | ||
2616 | if (host + 1 > *port_count + host_base) | |
2617 | *port_count = host + 1 - host_base; | |
2618 | } | |
2619 | closedir(dir); | |
2620 | return host_base; | |
2621 | } | |
2622 | ||
a891a3c2 LM |
2623 | static void print_imsm_capability(const struct imsm_orom *orom) |
2624 | { | |
0858eccf AP |
2625 | printf(" Platform : Intel(R) "); |
2626 | if (orom->capabilities == 0 && orom->driver_features == 0) | |
2627 | printf("Matrix Storage Manager\n"); | |
ab0c6bb9 AP |
2628 | else if (imsm_orom_is_enterprise(orom) && orom->major_ver >= 6) |
2629 | printf("Virtual RAID on CPU\n"); | |
0858eccf AP |
2630 | else |
2631 | printf("Rapid Storage Technology%s\n", | |
2632 | imsm_orom_is_enterprise(orom) ? " enterprise" : ""); | |
8d1114be MG |
2633 | if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build) { |
2634 | if (imsm_orom_is_vmd_without_efi(orom)) | |
2635 | printf(" Version : %d.%d\n", orom->major_ver, | |
2636 | orom->minor_ver); | |
2637 | else | |
2638 | printf(" Version : %d.%d.%d.%d\n", orom->major_ver, | |
2639 | orom->minor_ver, orom->hotfix_ver, orom->build); | |
2640 | } | |
a891a3c2 LM |
2641 | printf(" RAID Levels :%s%s%s%s%s\n", |
2642 | imsm_orom_has_raid0(orom) ? " raid0" : "", | |
2643 | imsm_orom_has_raid1(orom) ? " raid1" : "", | |
2644 | imsm_orom_has_raid1e(orom) ? " raid1e" : "", | |
2645 | imsm_orom_has_raid10(orom) ? " raid10" : "", | |
2646 | imsm_orom_has_raid5(orom) ? " raid5" : ""); | |
2647 | printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n", | |
2648 | imsm_orom_has_chunk(orom, 2) ? " 2k" : "", | |
2649 | imsm_orom_has_chunk(orom, 4) ? " 4k" : "", | |
2650 | imsm_orom_has_chunk(orom, 8) ? " 8k" : "", | |
2651 | imsm_orom_has_chunk(orom, 16) ? " 16k" : "", | |
2652 | imsm_orom_has_chunk(orom, 32) ? " 32k" : "", | |
2653 | imsm_orom_has_chunk(orom, 64) ? " 64k" : "", | |
2654 | imsm_orom_has_chunk(orom, 128) ? " 128k" : "", | |
2655 | imsm_orom_has_chunk(orom, 256) ? " 256k" : "", | |
2656 | imsm_orom_has_chunk(orom, 512) ? " 512k" : "", | |
2657 | imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "", | |
2658 | imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "", | |
2659 | imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "", | |
2660 | imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "", | |
2661 | imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "", | |
2662 | imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "", | |
2663 | imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : ""); | |
29cd0821 CA |
2664 | printf(" 2TB volumes :%s supported\n", |
2665 | (orom->attr & IMSM_OROM_ATTR_2TB)?"":" not"); | |
2666 | printf(" 2TB disks :%s supported\n", | |
2667 | (orom->attr & IMSM_OROM_ATTR_2TB_DISK)?"":" not"); | |
0e7f69a8 | 2668 | printf(" Max Disks : %d\n", orom->tds); |
0858eccf AP |
2669 | printf(" Max Volumes : %d per array, %d per %s\n", |
2670 | orom->vpa, orom->vphba, | |
2671 | imsm_orom_is_nvme(orom) ? "platform" : "controller"); | |
a891a3c2 LM |
2672 | return; |
2673 | } | |
2674 | ||
e50cf220 MN |
2675 | static void print_imsm_capability_export(const struct imsm_orom *orom) |
2676 | { | |
2677 | printf("MD_FIRMWARE_TYPE=imsm\n"); | |
0858eccf AP |
2678 | if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build) |
2679 | printf("IMSM_VERSION=%d.%d.%d.%d\n", orom->major_ver, orom->minor_ver, | |
2680 | orom->hotfix_ver, orom->build); | |
e50cf220 MN |
2681 | printf("IMSM_SUPPORTED_RAID_LEVELS=%s%s%s%s%s\n", |
2682 | imsm_orom_has_raid0(orom) ? "raid0 " : "", | |
2683 | imsm_orom_has_raid1(orom) ? "raid1 " : "", | |
2684 | imsm_orom_has_raid1e(orom) ? "raid1e " : "", | |
2685 | imsm_orom_has_raid5(orom) ? "raid10 " : "", | |
2686 | imsm_orom_has_raid10(orom) ? "raid5 " : ""); | |
2687 | printf("IMSM_SUPPORTED_CHUNK_SIZES=%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n", | |
2688 | imsm_orom_has_chunk(orom, 2) ? "2k " : "", | |
2689 | imsm_orom_has_chunk(orom, 4) ? "4k " : "", | |
2690 | imsm_orom_has_chunk(orom, 8) ? "8k " : "", | |
2691 | imsm_orom_has_chunk(orom, 16) ? "16k " : "", | |
2692 | imsm_orom_has_chunk(orom, 32) ? "32k " : "", | |
2693 | imsm_orom_has_chunk(orom, 64) ? "64k " : "", | |
2694 | imsm_orom_has_chunk(orom, 128) ? "128k " : "", | |
2695 | imsm_orom_has_chunk(orom, 256) ? "256k " : "", | |
2696 | imsm_orom_has_chunk(orom, 512) ? "512k " : "", | |
2697 | imsm_orom_has_chunk(orom, 1024*1) ? "1M " : "", | |
2698 | imsm_orom_has_chunk(orom, 1024*2) ? "2M " : "", | |
2699 | imsm_orom_has_chunk(orom, 1024*4) ? "4M " : "", | |
2700 | imsm_orom_has_chunk(orom, 1024*8) ? "8M " : "", | |
2701 | imsm_orom_has_chunk(orom, 1024*16) ? "16M " : "", | |
2702 | imsm_orom_has_chunk(orom, 1024*32) ? "32M " : "", | |
2703 | imsm_orom_has_chunk(orom, 1024*64) ? "64M " : ""); | |
2704 | printf("IMSM_2TB_VOLUMES=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB) ? "yes" : "no"); | |
2705 | printf("IMSM_2TB_DISKS=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB_DISK) ? "yes" : "no"); | |
2706 | printf("IMSM_MAX_DISKS=%d\n",orom->tds); | |
2707 | printf("IMSM_MAX_VOLUMES_PER_ARRAY=%d\n",orom->vpa); | |
2708 | printf("IMSM_MAX_VOLUMES_PER_CONTROLLER=%d\n",orom->vphba); | |
2709 | } | |
2710 | ||
9eafa1de | 2711 | static int detail_platform_imsm(int verbose, int enumerate_only, char *controller_path) |
d665cc31 DW |
2712 | { |
2713 | /* There are two components to imsm platform support, the ahci SATA | |
2714 | * controller and the option-rom. To find the SATA controller we | |
2715 | * simply look in /sys/bus/pci/drivers/ahci to see if an ahci | |
2716 | * controller with the Intel vendor id is present. This approach | |
2717 | * allows mdadm to leverage the kernel's ahci detection logic, with the | |
2718 | * caveat that if ahci.ko is not loaded mdadm will not be able to | |
2719 | * detect platform raid capabilities. The option-rom resides in a | |
2720 | * platform "Adapter ROM". We scan for its signature to retrieve the | |
2721 | * platform capabilities. If raid support is disabled in the BIOS the | |
2722 | * option-rom capability structure will not be available. | |
2723 | */ | |
d665cc31 | 2724 | struct sys_dev *list, *hba; |
d665cc31 DW |
2725 | int host_base = 0; |
2726 | int port_count = 0; | |
9eafa1de | 2727 | int result=1; |
d665cc31 | 2728 | |
5615172f | 2729 | if (enumerate_only) { |
420dafcd | 2730 | if (check_no_platform()) |
5615172f | 2731 | return 0; |
a891a3c2 LM |
2732 | list = find_intel_devices(); |
2733 | if (!list) | |
2734 | return 2; | |
2735 | for (hba = list; hba; hba = hba->next) { | |
6b781d33 AP |
2736 | if (find_imsm_capability(hba)) { |
2737 | result = 0; | |
a891a3c2 LM |
2738 | break; |
2739 | } | |
9eafa1de | 2740 | else |
6b781d33 | 2741 | result = 2; |
a891a3c2 | 2742 | } |
a891a3c2 | 2743 | return result; |
5615172f DW |
2744 | } |
2745 | ||
155cbb4c LM |
2746 | list = find_intel_devices(); |
2747 | if (!list) { | |
ba728be7 | 2748 | if (verbose > 0) |
7a862a02 | 2749 | pr_err("no active Intel(R) RAID controller found.\n"); |
d665cc31 | 2750 | return 2; |
ba728be7 | 2751 | } else if (verbose > 0) |
155cbb4c | 2752 | print_found_intel_controllers(list); |
d665cc31 | 2753 | |
a891a3c2 | 2754 | for (hba = list; hba; hba = hba->next) { |
0858eccf | 2755 | if (controller_path && (compare_paths(hba->path, controller_path) != 0)) |
9eafa1de | 2756 | continue; |
0858eccf | 2757 | if (!find_imsm_capability(hba)) { |
60f0f54d | 2758 | char buf[PATH_MAX]; |
e7b84f9d | 2759 | pr_err("imsm capabilities not found for controller: %s (type %s)\n", |
75350d87 KF |
2760 | hba->type == SYS_DEV_VMD || hba->type == SYS_DEV_SATA_VMD ? |
2761 | vmd_domain_to_controller(hba, buf) : | |
2762 | hba->path, get_sys_dev_type(hba->type)); | |
0858eccf AP |
2763 | continue; |
2764 | } | |
2765 | result = 0; | |
2766 | } | |
2767 | ||
2768 | if (controller_path && result == 1) { | |
2769 | pr_err("no active Intel(R) RAID controller found under %s\n", | |
2770 | controller_path); | |
2771 | return result; | |
2772 | } | |
2773 | ||
5e1d6128 | 2774 | const struct orom_entry *entry; |
0858eccf | 2775 | |
5e1d6128 | 2776 | for (entry = orom_entries; entry; entry = entry->next) { |
60f0f54d | 2777 | if (entry->type == SYS_DEV_VMD) { |
07cb1e57 | 2778 | print_imsm_capability(&entry->orom); |
32716c51 PB |
2779 | printf(" 3rd party NVMe :%s supported\n", |
2780 | imsm_orom_has_tpv_support(&entry->orom)?"":" not"); | |
60f0f54d PB |
2781 | for (hba = list; hba; hba = hba->next) { |
2782 | if (hba->type == SYS_DEV_VMD) { | |
2783 | char buf[PATH_MAX]; | |
60f0f54d PB |
2784 | printf(" I/O Controller : %s (%s)\n", |
2785 | vmd_domain_to_controller(hba, buf), get_sys_dev_type(hba->type)); | |
6da53c0e | 2786 | if (print_nvme_info(hba)) { |
b5eece69 PB |
2787 | if (verbose > 0) |
2788 | pr_err("failed to get devices attached to VMD domain.\n"); | |
2789 | result |= 2; | |
2790 | } | |
60f0f54d PB |
2791 | } |
2792 | } | |
07cb1e57 | 2793 | printf("\n"); |
60f0f54d PB |
2794 | continue; |
2795 | } | |
0858eccf | 2796 | |
60f0f54d PB |
2797 | print_imsm_capability(&entry->orom); |
2798 | if (entry->type == SYS_DEV_NVME) { | |
0858eccf AP |
2799 | for (hba = list; hba; hba = hba->next) { |
2800 | if (hba->type == SYS_DEV_NVME) | |
6da53c0e | 2801 | print_nvme_info(hba); |
0858eccf | 2802 | } |
60f0f54d | 2803 | printf("\n"); |
0858eccf AP |
2804 | continue; |
2805 | } | |
2806 | ||
2807 | struct devid_list *devid; | |
5e1d6128 | 2808 | for (devid = entry->devid_list; devid; devid = devid->next) { |
0858eccf AP |
2809 | hba = device_by_id(devid->devid); |
2810 | if (!hba) | |
2811 | continue; | |
2812 | ||
9eafa1de MN |
2813 | printf(" I/O Controller : %s (%s)\n", |
2814 | hba->path, get_sys_dev_type(hba->type)); | |
75350d87 | 2815 | if (hba->type == SYS_DEV_SATA || hba->type == SYS_DEV_SATA_VMD) { |
9eafa1de MN |
2816 | host_base = ahci_get_port_count(hba->path, &port_count); |
2817 | if (ahci_enumerate_ports(hba->path, port_count, host_base, verbose)) { | |
2818 | if (verbose > 0) | |
75350d87 KF |
2819 | pr_err("failed to enumerate ports on %s controller at %s.\n", |
2820 | get_sys_dev_type(hba->type), hba->pci_id); | |
9eafa1de MN |
2821 | result |= 2; |
2822 | } | |
120dc887 LM |
2823 | } |
2824 | } | |
0858eccf | 2825 | printf("\n"); |
d665cc31 | 2826 | } |
155cbb4c | 2827 | |
120dc887 | 2828 | return result; |
d665cc31 | 2829 | } |
e50cf220 | 2830 | |
9eafa1de | 2831 | static int export_detail_platform_imsm(int verbose, char *controller_path) |
e50cf220 | 2832 | { |
e50cf220 MN |
2833 | struct sys_dev *list, *hba; |
2834 | int result=1; | |
2835 | ||
2836 | list = find_intel_devices(); | |
2837 | if (!list) { | |
2838 | if (verbose > 0) | |
2839 | pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_INTEL_DEVICES\n"); | |
2840 | result = 2; | |
e50cf220 MN |
2841 | return result; |
2842 | } | |
2843 | ||
2844 | for (hba = list; hba; hba = hba->next) { | |
9eafa1de MN |
2845 | if (controller_path && (compare_paths(hba->path,controller_path) != 0)) |
2846 | continue; | |
60f0f54d PB |
2847 | if (!find_imsm_capability(hba) && verbose > 0) { |
2848 | char buf[PATH_MAX]; | |
2849 | pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_IMSM_CAPABLE_DEVICE_UNDER_%s\n", | |
75350d87 KF |
2850 | hba->type == SYS_DEV_VMD || hba->type == SYS_DEV_SATA_VMD ? |
2851 | vmd_domain_to_controller(hba, buf) : hba->path); | |
60f0f54d | 2852 | } |
0858eccf | 2853 | else |
e50cf220 | 2854 | result = 0; |
e50cf220 MN |
2855 | } |
2856 | ||
5e1d6128 | 2857 | const struct orom_entry *entry; |
0858eccf | 2858 | |
60f0f54d | 2859 | for (entry = orom_entries; entry; entry = entry->next) { |
75350d87 | 2860 | if (entry->type == SYS_DEV_VMD || entry->type == SYS_DEV_SATA_VMD) { |
60f0f54d PB |
2861 | for (hba = list; hba; hba = hba->next) |
2862 | print_imsm_capability_export(&entry->orom); | |
2863 | continue; | |
2864 | } | |
5e1d6128 | 2865 | print_imsm_capability_export(&entry->orom); |
60f0f54d | 2866 | } |
0858eccf | 2867 | |
e50cf220 MN |
2868 | return result; |
2869 | } | |
2870 | ||
cdddbdbc DW |
2871 | static int match_home_imsm(struct supertype *st, char *homehost) |
2872 | { | |
5115ca67 DW |
2873 | /* the imsm metadata format does not specify any host |
2874 | * identification information. We return -1 since we can never | |
2875 | * confirm nor deny whether a given array is "meant" for this | |
148acb7b | 2876 | * host. We rely on compare_super and the 'family_num' fields to |
5115ca67 DW |
2877 | * exclude member disks that do not belong, and we rely on |
2878 | * mdadm.conf to specify the arrays that should be assembled. | |
2879 | * Auto-assembly may still pick up "foreign" arrays. | |
2880 | */ | |
cdddbdbc | 2881 | |
9362c1c8 | 2882 | return -1; |
cdddbdbc DW |
2883 | } |
2884 | ||
2885 | static void uuid_from_super_imsm(struct supertype *st, int uuid[4]) | |
2886 | { | |
51006d85 N |
2887 | /* The uuid returned here is used for: |
2888 | * uuid to put into bitmap file (Create, Grow) | |
2889 | * uuid for backup header when saving critical section (Grow) | |
2890 | * comparing uuids when re-adding a device into an array | |
2891 | * In these cases the uuid required is that of the data-array, | |
2892 | * not the device-set. | |
2893 | * uuid to recognise same set when adding a missing device back | |
2894 | * to an array. This is a uuid for the device-set. | |
1011e834 | 2895 | * |
51006d85 N |
2896 | * For each of these we can make do with a truncated |
2897 | * or hashed uuid rather than the original, as long as | |
2898 | * everyone agrees. | |
2899 | * In each case the uuid required is that of the data-array, | |
2900 | * not the device-set. | |
43dad3d6 | 2901 | */ |
51006d85 N |
2902 | /* imsm does not track uuid's so we synthesis one using sha1 on |
2903 | * - The signature (Which is constant for all imsm array, but no matter) | |
148acb7b | 2904 | * - the orig_family_num of the container |
51006d85 N |
2905 | * - the index number of the volume |
2906 | * - the 'serial' number of the volume. | |
2907 | * Hopefully these are all constant. | |
2908 | */ | |
2909 | struct intel_super *super = st->sb; | |
43dad3d6 | 2910 | |
51006d85 N |
2911 | char buf[20]; |
2912 | struct sha1_ctx ctx; | |
2913 | struct imsm_dev *dev = NULL; | |
148acb7b | 2914 | __u32 family_num; |
51006d85 | 2915 | |
148acb7b DW |
2916 | /* some mdadm versions failed to set ->orig_family_num, in which |
2917 | * case fall back to ->family_num. orig_family_num will be | |
2918 | * fixed up with the first metadata update. | |
2919 | */ | |
2920 | family_num = super->anchor->orig_family_num; | |
2921 | if (family_num == 0) | |
2922 | family_num = super->anchor->family_num; | |
51006d85 | 2923 | sha1_init_ctx(&ctx); |
92bd8f8d | 2924 | sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx); |
148acb7b | 2925 | sha1_process_bytes(&family_num, sizeof(__u32), &ctx); |
51006d85 N |
2926 | if (super->current_vol >= 0) |
2927 | dev = get_imsm_dev(super, super->current_vol); | |
2928 | if (dev) { | |
2929 | __u32 vol = super->current_vol; | |
2930 | sha1_process_bytes(&vol, sizeof(vol), &ctx); | |
2931 | sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx); | |
2932 | } | |
2933 | sha1_finish_ctx(&ctx, buf); | |
2934 | memcpy(uuid, buf, 4*4); | |
cdddbdbc DW |
2935 | } |
2936 | ||
1e5c6983 DW |
2937 | static __u32 migr_strip_blocks_resync(struct imsm_dev *dev) |
2938 | { | |
2939 | /* migr_strip_size when repairing or initializing parity */ | |
238c0a71 | 2940 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2941 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2942 | ||
2943 | switch (get_imsm_raid_level(map)) { | |
2944 | case 5: | |
2945 | case 10: | |
2946 | return chunk; | |
2947 | default: | |
2948 | return 128*1024 >> 9; | |
2949 | } | |
2950 | } | |
2951 | ||
2952 | static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev) | |
2953 | { | |
2954 | /* migr_strip_size when rebuilding a degraded disk, no idea why | |
2955 | * this is different than migr_strip_size_resync(), but it's good | |
2956 | * to be compatible | |
2957 | */ | |
238c0a71 | 2958 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
2959 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2960 | ||
2961 | switch (get_imsm_raid_level(map)) { | |
2962 | case 1: | |
2963 | case 10: | |
2964 | if (map->num_members % map->num_domains == 0) | |
2965 | return 128*1024 >> 9; | |
2966 | else | |
2967 | return chunk; | |
2968 | case 5: | |
2969 | return max((__u32) 64*1024 >> 9, chunk); | |
2970 | default: | |
2971 | return 128*1024 >> 9; | |
2972 | } | |
2973 | } | |
2974 | ||
2975 | static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev) | |
2976 | { | |
238c0a71 AK |
2977 | struct imsm_map *lo = get_imsm_map(dev, MAP_0); |
2978 | struct imsm_map *hi = get_imsm_map(dev, MAP_1); | |
1e5c6983 DW |
2979 | __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip); |
2980 | __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip); | |
2981 | ||
2982 | return max((__u32) 1, hi_chunk / lo_chunk); | |
2983 | } | |
2984 | ||
2985 | static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev) | |
2986 | { | |
238c0a71 | 2987 | struct imsm_map *lo = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2988 | int level = get_imsm_raid_level(lo); |
2989 | ||
2990 | if (level == 1 || level == 10) { | |
238c0a71 | 2991 | struct imsm_map *hi = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
2992 | |
2993 | return hi->num_domains; | |
2994 | } else | |
2995 | return num_stripes_per_unit_resync(dev); | |
2996 | } | |
2997 | ||
44490938 MD |
2998 | static unsigned long long calc_component_size(struct imsm_map *map, |
2999 | struct imsm_dev *dev) | |
3000 | { | |
3001 | unsigned long long component_size; | |
3002 | unsigned long long dev_size = imsm_dev_size(dev); | |
a4f7290c | 3003 | long long calc_dev_size = 0; |
44490938 MD |
3004 | unsigned int member_disks = imsm_num_data_members(map); |
3005 | ||
3006 | if (member_disks == 0) | |
3007 | return 0; | |
3008 | ||
3009 | component_size = per_dev_array_size(map); | |
3010 | calc_dev_size = component_size * member_disks; | |
3011 | ||
3012 | /* Component size is rounded to 1MB so difference between size from | |
3013 | * metadata and size calculated from num_data_stripes equals up to | |
3014 | * 2048 blocks per each device. If the difference is higher it means | |
3015 | * that array size was expanded and num_data_stripes was not updated. | |
3016 | */ | |
a4f7290c | 3017 | if (llabs(calc_dev_size - (long long)dev_size) > |
44490938 MD |
3018 | (1 << SECT_PER_MB_SHIFT) * member_disks) { |
3019 | component_size = dev_size / member_disks; | |
3020 | dprintf("Invalid num_data_stripes in metadata; expected=%llu, found=%llu\n", | |
3021 | component_size / map->blocks_per_strip, | |
3022 | num_data_stripes(map)); | |
3023 | } | |
3024 | ||
3025 | return component_size; | |
3026 | } | |
3027 | ||
1e5c6983 DW |
3028 | static __u32 parity_segment_depth(struct imsm_dev *dev) |
3029 | { | |
238c0a71 | 3030 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
3031 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
3032 | ||
3033 | switch(get_imsm_raid_level(map)) { | |
3034 | case 1: | |
3035 | case 10: | |
3036 | return chunk * map->num_domains; | |
3037 | case 5: | |
3038 | return chunk * map->num_members; | |
3039 | default: | |
3040 | return chunk; | |
3041 | } | |
3042 | } | |
3043 | ||
3044 | static __u32 map_migr_block(struct imsm_dev *dev, __u32 block) | |
3045 | { | |
238c0a71 | 3046 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
3047 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
3048 | __u32 strip = block / chunk; | |
3049 | ||
3050 | switch (get_imsm_raid_level(map)) { | |
3051 | case 1: | |
3052 | case 10: { | |
3053 | __u32 vol_strip = (strip * map->num_domains) + 1; | |
3054 | __u32 vol_stripe = vol_strip / map->num_members; | |
3055 | ||
3056 | return vol_stripe * chunk + block % chunk; | |
3057 | } case 5: { | |
3058 | __u32 stripe = strip / (map->num_members - 1); | |
3059 | ||
3060 | return stripe * chunk + block % chunk; | |
3061 | } | |
3062 | default: | |
3063 | return 0; | |
3064 | } | |
3065 | } | |
3066 | ||
c47b0ff6 AK |
3067 | static __u64 blocks_per_migr_unit(struct intel_super *super, |
3068 | struct imsm_dev *dev) | |
1e5c6983 DW |
3069 | { |
3070 | /* calculate the conversion factor between per member 'blocks' | |
3071 | * (md/{resync,rebuild}_start) and imsm migration units, return | |
3072 | * 0 for the 'not migrating' and 'unsupported migration' cases | |
3073 | */ | |
3074 | if (!dev->vol.migr_state) | |
3075 | return 0; | |
3076 | ||
3077 | switch (migr_type(dev)) { | |
c47b0ff6 AK |
3078 | case MIGR_GEN_MIGR: { |
3079 | struct migr_record *migr_rec = super->migr_rec; | |
3080 | return __le32_to_cpu(migr_rec->blocks_per_unit); | |
3081 | } | |
1e5c6983 DW |
3082 | case MIGR_VERIFY: |
3083 | case MIGR_REPAIR: | |
3084 | case MIGR_INIT: { | |
238c0a71 | 3085 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
3086 | __u32 stripes_per_unit; |
3087 | __u32 blocks_per_unit; | |
3088 | __u32 parity_depth; | |
3089 | __u32 migr_chunk; | |
3090 | __u32 block_map; | |
3091 | __u32 block_rel; | |
3092 | __u32 segment; | |
3093 | __u32 stripe; | |
3094 | __u8 disks; | |
3095 | ||
3096 | /* yes, this is really the translation of migr_units to | |
3097 | * per-member blocks in the 'resync' case | |
3098 | */ | |
3099 | stripes_per_unit = num_stripes_per_unit_resync(dev); | |
3100 | migr_chunk = migr_strip_blocks_resync(dev); | |
9529d343 | 3101 | disks = imsm_num_data_members(map); |
1e5c6983 | 3102 | blocks_per_unit = stripes_per_unit * migr_chunk * disks; |
7b1ab482 | 3103 | stripe = __le16_to_cpu(map->blocks_per_strip) * disks; |
1e5c6983 DW |
3104 | segment = blocks_per_unit / stripe; |
3105 | block_rel = blocks_per_unit - segment * stripe; | |
3106 | parity_depth = parity_segment_depth(dev); | |
3107 | block_map = map_migr_block(dev, block_rel); | |
3108 | return block_map + parity_depth * segment; | |
3109 | } | |
3110 | case MIGR_REBUILD: { | |
3111 | __u32 stripes_per_unit; | |
3112 | __u32 migr_chunk; | |
3113 | ||
3114 | stripes_per_unit = num_stripes_per_unit_rebuild(dev); | |
3115 | migr_chunk = migr_strip_blocks_rebuild(dev); | |
3116 | return migr_chunk * stripes_per_unit; | |
3117 | } | |
1e5c6983 DW |
3118 | case MIGR_STATE_CHANGE: |
3119 | default: | |
3120 | return 0; | |
3121 | } | |
3122 | } | |
3123 | ||
c2c087e6 DW |
3124 | static int imsm_level_to_layout(int level) |
3125 | { | |
3126 | switch (level) { | |
3127 | case 0: | |
3128 | case 1: | |
3129 | return 0; | |
3130 | case 5: | |
3131 | case 6: | |
a380c027 | 3132 | return ALGORITHM_LEFT_ASYMMETRIC; |
c2c087e6 | 3133 | case 10: |
c92a2527 | 3134 | return 0x102; |
c2c087e6 | 3135 | } |
a18a888e | 3136 | return UnSet; |
c2c087e6 DW |
3137 | } |
3138 | ||
8e59f3d8 AK |
3139 | /******************************************************************************* |
3140 | * Function: read_imsm_migr_rec | |
3141 | * Description: Function reads imsm migration record from last sector of disk | |
3142 | * Parameters: | |
3143 | * fd : disk descriptor | |
3144 | * super : metadata info | |
3145 | * Returns: | |
3146 | * 0 : success, | |
3147 | * -1 : fail | |
3148 | ******************************************************************************/ | |
3149 | static int read_imsm_migr_rec(int fd, struct intel_super *super) | |
3150 | { | |
3151 | int ret_val = -1; | |
de44e46f | 3152 | unsigned int sector_size = super->sector_size; |
8e59f3d8 AK |
3153 | unsigned long long dsize; |
3154 | ||
3155 | get_dev_size(fd, NULL, &dsize); | |
de44e46f PB |
3156 | if (lseek64(fd, dsize - (sector_size*MIGR_REC_SECTOR_POSITION), |
3157 | SEEK_SET) < 0) { | |
e7b84f9d N |
3158 | pr_err("Cannot seek to anchor block: %s\n", |
3159 | strerror(errno)); | |
8e59f3d8 AK |
3160 | goto out; |
3161 | } | |
466070ad | 3162 | if ((unsigned int)read(fd, super->migr_rec_buf, |
de44e46f PB |
3163 | MIGR_REC_BUF_SECTORS*sector_size) != |
3164 | MIGR_REC_BUF_SECTORS*sector_size) { | |
e7b84f9d N |
3165 | pr_err("Cannot read migr record block: %s\n", |
3166 | strerror(errno)); | |
8e59f3d8 AK |
3167 | goto out; |
3168 | } | |
3169 | ret_val = 0; | |
de44e46f PB |
3170 | if (sector_size == 4096) |
3171 | convert_from_4k_imsm_migr_rec(super); | |
8e59f3d8 AK |
3172 | |
3173 | out: | |
3174 | return ret_val; | |
3175 | } | |
3176 | ||
3136abe5 AK |
3177 | static struct imsm_dev *imsm_get_device_during_migration( |
3178 | struct intel_super *super) | |
3179 | { | |
3180 | ||
3181 | struct intel_dev *dv; | |
3182 | ||
3183 | for (dv = super->devlist; dv; dv = dv->next) { | |
3184 | if (is_gen_migration(dv->dev)) | |
3185 | return dv->dev; | |
3186 | } | |
3187 | return NULL; | |
3188 | } | |
3189 | ||
8e59f3d8 AK |
3190 | /******************************************************************************* |
3191 | * Function: load_imsm_migr_rec | |
3192 | * Description: Function reads imsm migration record (it is stored at the last | |
3193 | * sector of disk) | |
3194 | * Parameters: | |
3195 | * super : imsm internal array info | |
8e59f3d8 AK |
3196 | * Returns: |
3197 | * 0 : success | |
3198 | * -1 : fail | |
4c965cc9 | 3199 | * -2 : no migration in progress |
8e59f3d8 | 3200 | ******************************************************************************/ |
2f86fda3 | 3201 | static int load_imsm_migr_rec(struct intel_super *super) |
8e59f3d8 | 3202 | { |
594dc1b8 | 3203 | struct dl *dl; |
8e59f3d8 AK |
3204 | char nm[30]; |
3205 | int retval = -1; | |
3206 | int fd = -1; | |
3136abe5 | 3207 | struct imsm_dev *dev; |
594dc1b8 | 3208 | struct imsm_map *map; |
b4ab44d8 | 3209 | int slot = -1; |
2f86fda3 | 3210 | int keep_fd = 1; |
3136abe5 AK |
3211 | |
3212 | /* find map under migration */ | |
3213 | dev = imsm_get_device_during_migration(super); | |
3214 | /* nothing to load,no migration in progress? | |
3215 | */ | |
3216 | if (dev == NULL) | |
4c965cc9 | 3217 | return -2; |
8e59f3d8 | 3218 | |
2f86fda3 MT |
3219 | map = get_imsm_map(dev, MAP_0); |
3220 | if (!map) | |
3221 | return -1; | |
3136abe5 | 3222 | |
2f86fda3 MT |
3223 | for (dl = super->disks; dl; dl = dl->next) { |
3224 | /* skip spare and failed disks | |
3225 | */ | |
3226 | if (dl->index < 0) | |
3227 | continue; | |
3228 | /* read only from one of the first two slots | |
3229 | */ | |
3230 | slot = get_imsm_disk_slot(map, dl->index); | |
3231 | if (slot > 1 || slot < 0) | |
3232 | continue; | |
3233 | ||
4389ce73 | 3234 | if (!is_fd_valid(dl->fd)) { |
8e59f3d8 AK |
3235 | sprintf(nm, "%d:%d", dl->major, dl->minor); |
3236 | fd = dev_open(nm, O_RDONLY); | |
4389ce73 MT |
3237 | |
3238 | if (is_fd_valid(fd)) { | |
2f86fda3 | 3239 | keep_fd = 0; |
8e59f3d8 | 3240 | break; |
2f86fda3 MT |
3241 | } |
3242 | } else { | |
3243 | fd = dl->fd; | |
3244 | break; | |
8e59f3d8 AK |
3245 | } |
3246 | } | |
2f86fda3 | 3247 | |
4389ce73 | 3248 | if (!is_fd_valid(fd)) |
2f86fda3 | 3249 | return retval; |
8e59f3d8 | 3250 | retval = read_imsm_migr_rec(fd, super); |
2f86fda3 | 3251 | if (!keep_fd) |
8e59f3d8 | 3252 | close(fd); |
2f86fda3 | 3253 | |
8e59f3d8 AK |
3254 | return retval; |
3255 | } | |
3256 | ||
c17608ea AK |
3257 | /******************************************************************************* |
3258 | * function: imsm_create_metadata_checkpoint_update | |
3259 | * Description: It creates update for checkpoint change. | |
3260 | * Parameters: | |
3261 | * super : imsm internal array info | |
3262 | * u : pointer to prepared update | |
3263 | * Returns: | |
3264 | * Uptate length. | |
3265 | * If length is equal to 0, input pointer u contains no update | |
3266 | ******************************************************************************/ | |
3267 | static int imsm_create_metadata_checkpoint_update( | |
3268 | struct intel_super *super, | |
3269 | struct imsm_update_general_migration_checkpoint **u) | |
3270 | { | |
3271 | ||
3272 | int update_memory_size = 0; | |
3273 | ||
1ade5cc1 | 3274 | dprintf("(enter)\n"); |
c17608ea AK |
3275 | |
3276 | if (u == NULL) | |
3277 | return 0; | |
3278 | *u = NULL; | |
3279 | ||
3280 | /* size of all update data without anchor */ | |
3281 | update_memory_size = | |
3282 | sizeof(struct imsm_update_general_migration_checkpoint); | |
3283 | ||
503975b9 | 3284 | *u = xcalloc(1, update_memory_size); |
c17608ea | 3285 | if (*u == NULL) { |
1ade5cc1 | 3286 | dprintf("error: cannot get memory\n"); |
c17608ea AK |
3287 | return 0; |
3288 | } | |
3289 | (*u)->type = update_general_migration_checkpoint; | |
9f421827 | 3290 | (*u)->curr_migr_unit = current_migr_unit(super->migr_rec); |
83b3de77 | 3291 | dprintf("prepared for %llu\n", (unsigned long long)(*u)->curr_migr_unit); |
c17608ea AK |
3292 | |
3293 | return update_memory_size; | |
3294 | } | |
3295 | ||
c17608ea AK |
3296 | static void imsm_update_metadata_locally(struct supertype *st, |
3297 | void *buf, int len); | |
3298 | ||
687629c2 AK |
3299 | /******************************************************************************* |
3300 | * Function: write_imsm_migr_rec | |
3301 | * Description: Function writes imsm migration record | |
3302 | * (at the last sector of disk) | |
3303 | * Parameters: | |
3304 | * super : imsm internal array info | |
3305 | * Returns: | |
3306 | * 0 : success | |
3307 | * -1 : if fail | |
3308 | ******************************************************************************/ | |
3309 | static int write_imsm_migr_rec(struct supertype *st) | |
3310 | { | |
3311 | struct intel_super *super = st->sb; | |
de44e46f | 3312 | unsigned int sector_size = super->sector_size; |
687629c2 | 3313 | unsigned long long dsize; |
687629c2 AK |
3314 | int retval = -1; |
3315 | struct dl *sd; | |
c17608ea AK |
3316 | int len; |
3317 | struct imsm_update_general_migration_checkpoint *u; | |
3136abe5 | 3318 | struct imsm_dev *dev; |
594dc1b8 | 3319 | struct imsm_map *map; |
3136abe5 AK |
3320 | |
3321 | /* find map under migration */ | |
3322 | dev = imsm_get_device_during_migration(super); | |
3323 | /* if no migration, write buffer anyway to clear migr_record | |
3324 | * on disk based on first available device | |
3325 | */ | |
3326 | if (dev == NULL) | |
3327 | dev = get_imsm_dev(super, super->current_vol < 0 ? 0 : | |
3328 | super->current_vol); | |
3329 | ||
44bfe6df | 3330 | map = get_imsm_map(dev, MAP_0); |
687629c2 | 3331 | |
de44e46f PB |
3332 | if (sector_size == 4096) |
3333 | convert_to_4k_imsm_migr_rec(super); | |
687629c2 | 3334 | for (sd = super->disks ; sd ; sd = sd->next) { |
b4ab44d8 | 3335 | int slot = -1; |
3136abe5 AK |
3336 | |
3337 | /* skip failed and spare devices */ | |
3338 | if (sd->index < 0) | |
3339 | continue; | |
687629c2 | 3340 | /* write to 2 first slots only */ |
3136abe5 AK |
3341 | if (map) |
3342 | slot = get_imsm_disk_slot(map, sd->index); | |
089f9d79 | 3343 | if (map == NULL || slot > 1 || slot < 0) |
687629c2 | 3344 | continue; |
3136abe5 | 3345 | |
2f86fda3 MT |
3346 | get_dev_size(sd->fd, NULL, &dsize); |
3347 | if (lseek64(sd->fd, dsize - (MIGR_REC_SECTOR_POSITION * | |
3348 | sector_size), | |
de44e46f | 3349 | SEEK_SET) < 0) { |
e7b84f9d N |
3350 | pr_err("Cannot seek to anchor block: %s\n", |
3351 | strerror(errno)); | |
687629c2 AK |
3352 | goto out; |
3353 | } | |
2f86fda3 | 3354 | if ((unsigned int)write(sd->fd, super->migr_rec_buf, |
de44e46f PB |
3355 | MIGR_REC_BUF_SECTORS*sector_size) != |
3356 | MIGR_REC_BUF_SECTORS*sector_size) { | |
e7b84f9d N |
3357 | pr_err("Cannot write migr record block: %s\n", |
3358 | strerror(errno)); | |
687629c2 AK |
3359 | goto out; |
3360 | } | |
687629c2 | 3361 | } |
de44e46f PB |
3362 | if (sector_size == 4096) |
3363 | convert_from_4k_imsm_migr_rec(super); | |
c17608ea AK |
3364 | /* update checkpoint information in metadata */ |
3365 | len = imsm_create_metadata_checkpoint_update(super, &u); | |
c17608ea AK |
3366 | if (len <= 0) { |
3367 | dprintf("imsm: Cannot prepare update\n"); | |
3368 | goto out; | |
3369 | } | |
3370 | /* update metadata locally */ | |
3371 | imsm_update_metadata_locally(st, u, len); | |
3372 | /* and possibly remotely */ | |
3373 | if (st->update_tail) { | |
3374 | append_metadata_update(st, u, len); | |
3375 | /* during reshape we do all work inside metadata handler | |
3376 | * manage_reshape(), so metadata update has to be triggered | |
3377 | * insida it | |
3378 | */ | |
3379 | flush_metadata_updates(st); | |
3380 | st->update_tail = &st->updates; | |
3381 | } else | |
3382 | free(u); | |
687629c2 AK |
3383 | |
3384 | retval = 0; | |
3385 | out: | |
687629c2 AK |
3386 | return retval; |
3387 | } | |
3388 | ||
e2962bfc AK |
3389 | /* spare/missing disks activations are not allowe when |
3390 | * array/container performs reshape operation, because | |
3391 | * all arrays in container works on the same disks set | |
3392 | */ | |
3393 | int imsm_reshape_blocks_arrays_changes(struct intel_super *super) | |
3394 | { | |
3395 | int rv = 0; | |
3396 | struct intel_dev *i_dev; | |
3397 | struct imsm_dev *dev; | |
3398 | ||
3399 | /* check whole container | |
3400 | */ | |
3401 | for (i_dev = super->devlist; i_dev; i_dev = i_dev->next) { | |
3402 | dev = i_dev->dev; | |
3ad25638 | 3403 | if (is_gen_migration(dev)) { |
e2962bfc AK |
3404 | /* No repair during any migration in container |
3405 | */ | |
3406 | rv = 1; | |
3407 | break; | |
3408 | } | |
3409 | } | |
3410 | return rv; | |
3411 | } | |
3e684231 | 3412 | static unsigned long long imsm_component_size_alignment_check(int level, |
c41e00b2 | 3413 | int chunk_size, |
f36a9ecd | 3414 | unsigned int sector_size, |
c41e00b2 AK |
3415 | unsigned long long component_size) |
3416 | { | |
3e684231 | 3417 | unsigned int component_size_alignment; |
c41e00b2 | 3418 | |
3e684231 | 3419 | /* check component size alignment |
c41e00b2 | 3420 | */ |
3e684231 | 3421 | component_size_alignment = component_size % (chunk_size/sector_size); |
c41e00b2 | 3422 | |
3e684231 | 3423 | dprintf("(Level: %i, chunk_size = %i, component_size = %llu), component_size_alignment = %u\n", |
c41e00b2 | 3424 | level, chunk_size, component_size, |
3e684231 | 3425 | component_size_alignment); |
c41e00b2 | 3426 | |
3e684231 MZ |
3427 | if (component_size_alignment && (level != 1) && (level != UnSet)) { |
3428 | dprintf("imsm: reported component size aligned from %llu ", | |
c41e00b2 | 3429 | component_size); |
3e684231 | 3430 | component_size -= component_size_alignment; |
1ade5cc1 | 3431 | dprintf_cont("to %llu (%i).\n", |
3e684231 | 3432 | component_size, component_size_alignment); |
c41e00b2 AK |
3433 | } |
3434 | ||
3435 | return component_size; | |
3436 | } | |
e2962bfc | 3437 | |
fbc42556 JR |
3438 | /******************************************************************************* |
3439 | * Function: get_bitmap_header_sector | |
3440 | * Description: Returns the sector where the bitmap header is placed. | |
3441 | * Parameters: | |
3442 | * st : supertype information | |
3443 | * dev_idx : index of the device with bitmap | |
3444 | * | |
3445 | * Returns: | |
3446 | * The sector where the bitmap header is placed | |
3447 | ******************************************************************************/ | |
3448 | static unsigned long long get_bitmap_header_sector(struct intel_super *super, | |
3449 | int dev_idx) | |
3450 | { | |
3451 | struct imsm_dev *dev = get_imsm_dev(super, dev_idx); | |
3452 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
3453 | ||
3454 | if (!super->sector_size) { | |
3455 | dprintf("sector size is not set\n"); | |
3456 | return 0; | |
3457 | } | |
3458 | ||
3459 | return pba_of_lba0(map) + calc_component_size(map, dev) + | |
3460 | (IMSM_BITMAP_HEADER_OFFSET / super->sector_size); | |
3461 | } | |
3462 | ||
3463 | /******************************************************************************* | |
3464 | * Function: get_bitmap_sector | |
3465 | * Description: Returns the sector where the bitmap is placed. | |
3466 | * Parameters: | |
3467 | * st : supertype information | |
3468 | * dev_idx : index of the device with bitmap | |
3469 | * | |
3470 | * Returns: | |
3471 | * The sector where the bitmap is placed | |
3472 | ******************************************************************************/ | |
3473 | static unsigned long long get_bitmap_sector(struct intel_super *super, | |
3474 | int dev_idx) | |
3475 | { | |
3476 | if (!super->sector_size) { | |
3477 | dprintf("sector size is not set\n"); | |
3478 | return 0; | |
3479 | } | |
3480 | ||
3481 | return get_bitmap_header_sector(super, dev_idx) + | |
3482 | (IMSM_BITMAP_HEADER_SIZE / super->sector_size); | |
3483 | } | |
3484 | ||
2432ce9b AP |
3485 | static unsigned long long get_ppl_sector(struct intel_super *super, int dev_idx) |
3486 | { | |
3487 | struct imsm_dev *dev = get_imsm_dev(super, dev_idx); | |
3488 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
3489 | ||
3490 | return pba_of_lba0(map) + | |
3491 | (num_data_stripes(map) * map->blocks_per_strip); | |
3492 | } | |
3493 | ||
a5d85af7 | 3494 | static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap) |
bf5a934a DW |
3495 | { |
3496 | struct intel_super *super = st->sb; | |
c47b0ff6 | 3497 | struct migr_record *migr_rec = super->migr_rec; |
949c47a0 | 3498 | struct imsm_dev *dev = get_imsm_dev(super, super->current_vol); |
238c0a71 AK |
3499 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
3500 | struct imsm_map *prev_map = get_imsm_map(dev, MAP_1); | |
b335e593 | 3501 | struct imsm_map *map_to_analyse = map; |
efb30e7f | 3502 | struct dl *dl; |
a5d85af7 | 3503 | int map_disks = info->array.raid_disks; |
bf5a934a | 3504 | |
95eeceeb | 3505 | memset(info, 0, sizeof(*info)); |
b335e593 AK |
3506 | if (prev_map) |
3507 | map_to_analyse = prev_map; | |
3508 | ||
ca0748fa | 3509 | dl = super->current_disk; |
9894ec0d | 3510 | |
bf5a934a | 3511 | info->container_member = super->current_vol; |
cd0430a1 | 3512 | info->array.raid_disks = map->num_members; |
b335e593 | 3513 | info->array.level = get_imsm_raid_level(map_to_analyse); |
bf5a934a DW |
3514 | info->array.layout = imsm_level_to_layout(info->array.level); |
3515 | info->array.md_minor = -1; | |
3516 | info->array.ctime = 0; | |
3517 | info->array.utime = 0; | |
b335e593 AK |
3518 | info->array.chunk_size = |
3519 | __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9; | |
2432ce9b | 3520 | info->array.state = !(dev->vol.dirty & RAIDVOL_DIRTY); |
fcc2c9da | 3521 | info->custom_array_size = imsm_dev_size(dev); |
3ad25638 AK |
3522 | info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb); |
3523 | ||
3f510843 | 3524 | if (is_gen_migration(dev)) { |
195d1d76 PP |
3525 | /* |
3526 | * device prev_map should be added if it is in the middle | |
3527 | * of migration | |
3528 | */ | |
3529 | assert(prev_map); | |
3530 | ||
3f83228a | 3531 | info->reshape_active = 1; |
b335e593 AK |
3532 | info->new_level = get_imsm_raid_level(map); |
3533 | info->new_layout = imsm_level_to_layout(info->new_level); | |
3534 | info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9; | |
3f83228a | 3535 | info->delta_disks = map->num_members - prev_map->num_members; |
493f5dd6 N |
3536 | if (info->delta_disks) { |
3537 | /* this needs to be applied to every array | |
3538 | * in the container. | |
3539 | */ | |
81219e70 | 3540 | info->reshape_active = CONTAINER_RESHAPE; |
493f5dd6 | 3541 | } |
3f83228a N |
3542 | /* We shape information that we give to md might have to be |
3543 | * modify to cope with md's requirement for reshaping arrays. | |
3544 | * For example, when reshaping a RAID0, md requires it to be | |
3545 | * presented as a degraded RAID4. | |
3546 | * Also if a RAID0 is migrating to a RAID5 we need to specify | |
3547 | * the array as already being RAID5, but the 'before' layout | |
3548 | * is a RAID4-like layout. | |
3549 | */ | |
3550 | switch (info->array.level) { | |
3551 | case 0: | |
3552 | switch(info->new_level) { | |
3553 | case 0: | |
3554 | /* conversion is happening as RAID4 */ | |
3555 | info->array.level = 4; | |
3556 | info->array.raid_disks += 1; | |
3557 | break; | |
3558 | case 5: | |
3559 | /* conversion is happening as RAID5 */ | |
3560 | info->array.level = 5; | |
3561 | info->array.layout = ALGORITHM_PARITY_N; | |
3f83228a N |
3562 | info->delta_disks -= 1; |
3563 | break; | |
3564 | default: | |
3565 | /* FIXME error message */ | |
3566 | info->array.level = UnSet; | |
3567 | break; | |
3568 | } | |
3569 | break; | |
3570 | } | |
b335e593 AK |
3571 | } else { |
3572 | info->new_level = UnSet; | |
3573 | info->new_layout = UnSet; | |
3574 | info->new_chunk = info->array.chunk_size; | |
3f83228a | 3575 | info->delta_disks = 0; |
b335e593 | 3576 | } |
ca0748fa | 3577 | |
efb30e7f DW |
3578 | if (dl) { |
3579 | info->disk.major = dl->major; | |
3580 | info->disk.minor = dl->minor; | |
ca0748fa | 3581 | info->disk.number = dl->index; |
656b6b5a N |
3582 | info->disk.raid_disk = get_imsm_disk_slot(map_to_analyse, |
3583 | dl->index); | |
efb30e7f | 3584 | } |
bf5a934a | 3585 | |
5551b113 | 3586 | info->data_offset = pba_of_lba0(map_to_analyse); |
44490938 | 3587 | info->component_size = calc_component_size(map, dev); |
3e684231 | 3588 | info->component_size = imsm_component_size_alignment_check( |
c41e00b2 AK |
3589 | info->array.level, |
3590 | info->array.chunk_size, | |
f36a9ecd | 3591 | super->sector_size, |
c41e00b2 | 3592 | info->component_size); |
5e46202e | 3593 | info->bb.supported = 1; |
139dae11 | 3594 | |
301406c9 | 3595 | memset(info->uuid, 0, sizeof(info->uuid)); |
921d9e16 | 3596 | info->recovery_start = MaxSector; |
bf5a934a | 3597 | |
c2462068 PB |
3598 | if (info->array.level == 5 && |
3599 | (dev->rwh_policy == RWH_DISTRIBUTED || | |
3600 | dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)) { | |
2432ce9b AP |
3601 | info->consistency_policy = CONSISTENCY_POLICY_PPL; |
3602 | info->ppl_sector = get_ppl_sector(super, super->current_vol); | |
c2462068 PB |
3603 | if (dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED) |
3604 | info->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9; | |
3605 | else | |
3606 | info->ppl_size = (PPL_HEADER_SIZE + PPL_ENTRY_SPACE) | |
3607 | >> 9; | |
2432ce9b AP |
3608 | } else if (info->array.level <= 0) { |
3609 | info->consistency_policy = CONSISTENCY_POLICY_NONE; | |
3610 | } else { | |
fbc42556 JR |
3611 | if (dev->rwh_policy == RWH_BITMAP) { |
3612 | info->bitmap_offset = get_bitmap_sector(super, super->current_vol); | |
3613 | info->consistency_policy = CONSISTENCY_POLICY_BITMAP; | |
3614 | } else { | |
3615 | info->consistency_policy = CONSISTENCY_POLICY_RESYNC; | |
3616 | } | |
2432ce9b AP |
3617 | } |
3618 | ||
d2e6d5d6 | 3619 | info->reshape_progress = 0; |
b6796ce1 | 3620 | info->resync_start = MaxSector; |
b9172665 | 3621 | if ((map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED || |
2432ce9b | 3622 | !(info->array.state & 1)) && |
b9172665 | 3623 | imsm_reshape_blocks_arrays_changes(super) == 0) { |
301406c9 | 3624 | info->resync_start = 0; |
b6796ce1 AK |
3625 | } |
3626 | if (dev->vol.migr_state) { | |
1e5c6983 DW |
3627 | switch (migr_type(dev)) { |
3628 | case MIGR_REPAIR: | |
3629 | case MIGR_INIT: { | |
c47b0ff6 AK |
3630 | __u64 blocks_per_unit = blocks_per_migr_unit(super, |
3631 | dev); | |
4036e7ee | 3632 | __u64 units = vol_curr_migr_unit(dev); |
1e5c6983 DW |
3633 | |
3634 | info->resync_start = blocks_per_unit * units; | |
3635 | break; | |
3636 | } | |
d2e6d5d6 | 3637 | case MIGR_GEN_MIGR: { |
c47b0ff6 AK |
3638 | __u64 blocks_per_unit = blocks_per_migr_unit(super, |
3639 | dev); | |
9f421827 | 3640 | __u64 units = current_migr_unit(migr_rec); |
04fa9523 | 3641 | int used_disks; |
d2e6d5d6 | 3642 | |
befb629b AK |
3643 | if (__le32_to_cpu(migr_rec->ascending_migr) && |
3644 | (units < | |
9f421827 | 3645 | (get_num_migr_units(migr_rec)-1)) && |
befb629b AK |
3646 | (super->migr_rec->rec_status == |
3647 | __cpu_to_le32(UNIT_SRC_IN_CP_AREA))) | |
3648 | units++; | |
3649 | ||
d2e6d5d6 | 3650 | info->reshape_progress = blocks_per_unit * units; |
6289d1e0 | 3651 | |
7a862a02 | 3652 | dprintf("IMSM: General Migration checkpoint : %llu (%llu) -> read reshape progress : %llu\n", |
19986c72 MB |
3653 | (unsigned long long)units, |
3654 | (unsigned long long)blocks_per_unit, | |
3655 | info->reshape_progress); | |
75156c46 | 3656 | |
9529d343 | 3657 | used_disks = imsm_num_data_members(prev_map); |
75156c46 | 3658 | if (used_disks > 0) { |
895ffd99 | 3659 | info->custom_array_size = per_dev_array_size(map) * |
75156c46 | 3660 | used_disks; |
75156c46 | 3661 | } |
d2e6d5d6 | 3662 | } |
1e5c6983 DW |
3663 | case MIGR_VERIFY: |
3664 | /* we could emulate the checkpointing of | |
3665 | * 'sync_action=check' migrations, but for now | |
3666 | * we just immediately complete them | |
3667 | */ | |
3668 | case MIGR_REBUILD: | |
3669 | /* this is handled by container_content_imsm() */ | |
1e5c6983 DW |
3670 | case MIGR_STATE_CHANGE: |
3671 | /* FIXME handle other migrations */ | |
3672 | default: | |
3673 | /* we are not dirty, so... */ | |
3674 | info->resync_start = MaxSector; | |
3675 | } | |
b6796ce1 | 3676 | } |
301406c9 DW |
3677 | |
3678 | strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN); | |
3679 | info->name[MAX_RAID_SERIAL_LEN] = 0; | |
bf5a934a | 3680 | |
f35f2525 N |
3681 | info->array.major_version = -1; |
3682 | info->array.minor_version = -2; | |
4dd2df09 | 3683 | sprintf(info->text_version, "/%s/%d", st->container_devnm, info->container_member); |
a67dd8cc | 3684 | info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */ |
51006d85 | 3685 | uuid_from_super_imsm(st, info->uuid); |
a5d85af7 N |
3686 | |
3687 | if (dmap) { | |
3688 | int i, j; | |
3689 | for (i=0; i<map_disks; i++) { | |
3690 | dmap[i] = 0; | |
3691 | if (i < info->array.raid_disks) { | |
3692 | struct imsm_disk *dsk; | |
238c0a71 | 3693 | j = get_imsm_disk_idx(dev, i, MAP_X); |
a5d85af7 N |
3694 | dsk = get_imsm_disk(super, j); |
3695 | if (dsk && (dsk->status & CONFIGURED_DISK)) | |
3696 | dmap[i] = 1; | |
3697 | } | |
3698 | } | |
3699 | } | |
81ac8b4d | 3700 | } |
bf5a934a | 3701 | |
3b451610 AK |
3702 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, |
3703 | int failed, int look_in_map); | |
3704 | ||
3705 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev, | |
3706 | int look_in_map); | |
3707 | ||
3708 | static void manage_second_map(struct intel_super *super, struct imsm_dev *dev) | |
3709 | { | |
3710 | if (is_gen_migration(dev)) { | |
3711 | int failed; | |
3712 | __u8 map_state; | |
3713 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); | |
3714 | ||
3715 | failed = imsm_count_failed(super, dev, MAP_1); | |
238c0a71 | 3716 | map_state = imsm_check_degraded(super, dev, failed, MAP_1); |
3b451610 AK |
3717 | if (map2->map_state != map_state) { |
3718 | map2->map_state = map_state; | |
3719 | super->updates_pending++; | |
3720 | } | |
3721 | } | |
3722 | } | |
97b4d0e9 DW |
3723 | |
3724 | static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index) | |
3725 | { | |
3726 | struct dl *d; | |
3727 | ||
3728 | for (d = super->missing; d; d = d->next) | |
3729 | if (d->index == index) | |
3730 | return &d->disk; | |
3731 | return NULL; | |
3732 | } | |
3733 | ||
a5d85af7 | 3734 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map) |
4f5bc454 DW |
3735 | { |
3736 | struct intel_super *super = st->sb; | |
4f5bc454 | 3737 | struct imsm_disk *disk; |
a5d85af7 | 3738 | int map_disks = info->array.raid_disks; |
476b00bd | 3739 | int max_enough = -1; |
ab3cb6b3 N |
3740 | int i; |
3741 | struct imsm_super *mpb; | |
4f5bc454 | 3742 | |
bf5a934a | 3743 | if (super->current_vol >= 0) { |
a5d85af7 | 3744 | getinfo_super_imsm_volume(st, info, map); |
bf5a934a DW |
3745 | return; |
3746 | } | |
95eeceeb | 3747 | memset(info, 0, sizeof(*info)); |
d23fe947 DW |
3748 | |
3749 | /* Set raid_disks to zero so that Assemble will always pull in valid | |
3750 | * spares | |
3751 | */ | |
3752 | info->array.raid_disks = 0; | |
cdddbdbc DW |
3753 | info->array.level = LEVEL_CONTAINER; |
3754 | info->array.layout = 0; | |
3755 | info->array.md_minor = -1; | |
1011e834 | 3756 | info->array.ctime = 0; /* N/A for imsm */ |
cdddbdbc DW |
3757 | info->array.utime = 0; |
3758 | info->array.chunk_size = 0; | |
3759 | ||
3760 | info->disk.major = 0; | |
3761 | info->disk.minor = 0; | |
cdddbdbc | 3762 | info->disk.raid_disk = -1; |
c2c087e6 | 3763 | info->reshape_active = 0; |
f35f2525 N |
3764 | info->array.major_version = -1; |
3765 | info->array.minor_version = -2; | |
c2c087e6 | 3766 | strcpy(info->text_version, "imsm"); |
a67dd8cc | 3767 | info->safe_mode_delay = 0; |
c2c087e6 DW |
3768 | info->disk.number = -1; |
3769 | info->disk.state = 0; | |
c5afc314 | 3770 | info->name[0] = 0; |
921d9e16 | 3771 | info->recovery_start = MaxSector; |
3ad25638 | 3772 | info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb); |
5e46202e | 3773 | info->bb.supported = 1; |
c2c087e6 | 3774 | |
97b4d0e9 | 3775 | /* do we have the all the insync disks that we expect? */ |
ab3cb6b3 | 3776 | mpb = super->anchor; |
b7d81a38 | 3777 | info->events = __le32_to_cpu(mpb->generation_num); |
97b4d0e9 | 3778 | |
ab3cb6b3 N |
3779 | for (i = 0; i < mpb->num_raid_devs; i++) { |
3780 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
476b00bd | 3781 | int failed, enough, j, missing = 0; |
ab3cb6b3 | 3782 | struct imsm_map *map; |
476b00bd | 3783 | __u8 state; |
97b4d0e9 | 3784 | |
476b00bd MT |
3785 | failed = imsm_count_failed(super, dev, MAP_0); |
3786 | state = imsm_check_degraded(super, dev, failed, MAP_0); | |
238c0a71 | 3787 | map = get_imsm_map(dev, MAP_0); |
ab3cb6b3 N |
3788 | |
3789 | /* any newly missing disks? | |
3790 | * (catches single-degraded vs double-degraded) | |
3791 | */ | |
3792 | for (j = 0; j < map->num_members; j++) { | |
238c0a71 | 3793 | __u32 ord = get_imsm_ord_tbl_ent(dev, j, MAP_0); |
ab3cb6b3 N |
3794 | __u32 idx = ord_to_idx(ord); |
3795 | ||
20dc76d1 MT |
3796 | if (super->disks && super->disks->index == (int)idx) |
3797 | info->disk.raid_disk = j; | |
3798 | ||
ab3cb6b3 N |
3799 | if (!(ord & IMSM_ORD_REBUILD) && |
3800 | get_imsm_missing(super, idx)) { | |
476b00bd | 3801 | missing = 1; |
ab3cb6b3 N |
3802 | break; |
3803 | } | |
97b4d0e9 | 3804 | } |
476b00bd MT |
3805 | |
3806 | if (state == IMSM_T_STATE_FAILED) | |
3807 | enough = -1; | |
3808 | else if (state == IMSM_T_STATE_DEGRADED && | |
3809 | (state != map->map_state || missing)) | |
3810 | enough = 0; | |
3811 | else /* we're normal, or already degraded */ | |
3812 | enough = 1; | |
3813 | if (is_gen_migration(dev) && missing) { | |
3814 | /* during general migration we need all disks | |
3815 | * that process is running on. | |
3816 | * No new missing disk is allowed. | |
3817 | */ | |
3818 | max_enough = -1; | |
3819 | enough = -1; | |
3820 | /* no more checks necessary | |
3821 | */ | |
3822 | break; | |
3823 | } | |
3824 | /* in the missing/failed disk case check to see | |
3825 | * if at least one array is runnable | |
3826 | */ | |
3827 | max_enough = max(max_enough, enough); | |
ab3cb6b3 | 3828 | } |
97b4d0e9 | 3829 | |
476b00bd MT |
3830 | info->container_enough = max_enough; |
3831 | ||
4a04ec6c | 3832 | if (super->disks) { |
14e8215b DW |
3833 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
3834 | ||
b9f594fe | 3835 | disk = &super->disks->disk; |
5551b113 | 3836 | info->data_offset = total_blocks(&super->disks->disk) - reserved; |
14e8215b | 3837 | info->component_size = reserved; |
25ed7e59 | 3838 | info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0; |
df474657 DW |
3839 | /* we don't change info->disk.raid_disk here because |
3840 | * this state will be finalized in mdmon after we have | |
3841 | * found the 'most fresh' version of the metadata | |
3842 | */ | |
25ed7e59 | 3843 | info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; |
2432ce9b AP |
3844 | info->disk.state |= (is_spare(disk) || is_journal(disk)) ? |
3845 | 0 : (1 << MD_DISK_SYNC); | |
cdddbdbc | 3846 | } |
a575e2a7 DW |
3847 | |
3848 | /* only call uuid_from_super_imsm when this disk is part of a populated container, | |
3849 | * ->compare_super may have updated the 'num_raid_devs' field for spares | |
3850 | */ | |
3851 | if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs) | |
36ba7d48 | 3852 | uuid_from_super_imsm(st, info->uuid); |
22e263f6 AC |
3853 | else |
3854 | memcpy(info->uuid, uuid_zero, sizeof(uuid_zero)); | |
a5d85af7 N |
3855 | |
3856 | /* I don't know how to compute 'map' on imsm, so use safe default */ | |
3857 | if (map) { | |
3858 | int i; | |
3859 | for (i = 0; i < map_disks; i++) | |
3860 | map[i] = 1; | |
3861 | } | |
3862 | ||
cdddbdbc DW |
3863 | } |
3864 | ||
5c4cd5da AC |
3865 | /* allocates memory and fills disk in mdinfo structure |
3866 | * for each disk in array */ | |
3867 | struct mdinfo *getinfo_super_disks_imsm(struct supertype *st) | |
3868 | { | |
594dc1b8 | 3869 | struct mdinfo *mddev; |
5c4cd5da AC |
3870 | struct intel_super *super = st->sb; |
3871 | struct imsm_disk *disk; | |
3872 | int count = 0; | |
3873 | struct dl *dl; | |
3874 | if (!super || !super->disks) | |
3875 | return NULL; | |
3876 | dl = super->disks; | |
503975b9 | 3877 | mddev = xcalloc(1, sizeof(*mddev)); |
5c4cd5da AC |
3878 | while (dl) { |
3879 | struct mdinfo *tmp; | |
3880 | disk = &dl->disk; | |
503975b9 | 3881 | tmp = xcalloc(1, sizeof(*tmp)); |
5c4cd5da AC |
3882 | if (mddev->devs) |
3883 | tmp->next = mddev->devs; | |
3884 | mddev->devs = tmp; | |
3885 | tmp->disk.number = count++; | |
3886 | tmp->disk.major = dl->major; | |
3887 | tmp->disk.minor = dl->minor; | |
3888 | tmp->disk.state = is_configured(disk) ? | |
3889 | (1 << MD_DISK_ACTIVE) : 0; | |
3890 | tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; | |
3891 | tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC); | |
3892 | tmp->disk.raid_disk = -1; | |
3893 | dl = dl->next; | |
3894 | } | |
3895 | return mddev; | |
3896 | } | |
3897 | ||
cdddbdbc | 3898 | static int update_super_imsm(struct supertype *st, struct mdinfo *info, |
03312b52 MK |
3899 | enum update_opt update, char *devname, |
3900 | int verbose, int uuid_set, char *homehost) | |
cdddbdbc | 3901 | { |
f352c545 DW |
3902 | /* For 'assemble' and 'force' we need to return non-zero if any |
3903 | * change was made. For others, the return value is ignored. | |
3904 | * Update options are: | |
3905 | * force-one : This device looks a bit old but needs to be included, | |
3906 | * update age info appropriately. | |
3907 | * assemble: clear any 'faulty' flag to allow this device to | |
3908 | * be assembled. | |
3909 | * force-array: Array is degraded but being forced, mark it clean | |
3910 | * if that will be needed to assemble it. | |
3911 | * | |
3912 | * newdev: not used ???? | |
3913 | * grow: Array has gained a new device - this is currently for | |
3914 | * linear only | |
3915 | * resync: mark as dirty so a resync will happen. | |
3916 | * name: update the name - preserving the homehost | |
6e46bf34 | 3917 | * uuid: Change the uuid of the array to match watch is given |
f352c545 DW |
3918 | * |
3919 | * Following are not relevant for this imsm: | |
3920 | * sparc2.2 : update from old dodgey metadata | |
3921 | * super-minor: change the preferred_minor number | |
3922 | * summaries: update redundant counters. | |
f352c545 DW |
3923 | * homehost: update the recorded homehost |
3924 | * _reshape_progress: record new reshape_progress position. | |
3925 | */ | |
6e46bf34 DW |
3926 | int rv = 1; |
3927 | struct intel_super *super = st->sb; | |
3928 | struct imsm_super *mpb; | |
f352c545 | 3929 | |
6e46bf34 DW |
3930 | /* we can only update container info */ |
3931 | if (!super || super->current_vol >= 0 || !super->anchor) | |
3932 | return 1; | |
3933 | ||
3934 | mpb = super->anchor; | |
3935 | ||
03312b52 | 3936 | switch (update) { |
4345e135 | 3937 | case UOPT_UUID: |
81a5b4f5 N |
3938 | /* We take this to mean that the family_num should be updated. |
3939 | * However that is much smaller than the uuid so we cannot really | |
3940 | * allow an explicit uuid to be given. And it is hard to reliably | |
3941 | * know if one was. | |
3942 | * So if !uuid_set we know the current uuid is random and just used | |
3943 | * the first 'int' and copy it to the other 3 positions. | |
3944 | * Otherwise we require the 4 'int's to be the same as would be the | |
3945 | * case if we are using a random uuid. So an explicit uuid will be | |
3946 | * accepted as long as all for ints are the same... which shouldn't hurt | |
6e46bf34 | 3947 | */ |
81a5b4f5 N |
3948 | if (!uuid_set) { |
3949 | info->uuid[1] = info->uuid[2] = info->uuid[3] = info->uuid[0]; | |
6e46bf34 | 3950 | rv = 0; |
81a5b4f5 N |
3951 | } else { |
3952 | if (info->uuid[0] != info->uuid[1] || | |
3953 | info->uuid[1] != info->uuid[2] || | |
3954 | info->uuid[2] != info->uuid[3]) | |
3955 | rv = -1; | |
3956 | else | |
3957 | rv = 0; | |
6e46bf34 | 3958 | } |
81a5b4f5 N |
3959 | if (rv == 0) |
3960 | mpb->orig_family_num = info->uuid[0]; | |
4345e135 MK |
3961 | break; |
3962 | case UOPT_SPEC_ASSEMBLE: | |
6e46bf34 | 3963 | rv = 0; |
4345e135 MK |
3964 | break; |
3965 | default: | |
1e2b2765 | 3966 | rv = -1; |
4345e135 MK |
3967 | break; |
3968 | } | |
f352c545 | 3969 | |
6e46bf34 DW |
3970 | /* successful update? recompute checksum */ |
3971 | if (rv == 0) | |
3972 | mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb)); | |
f352c545 DW |
3973 | |
3974 | return rv; | |
cdddbdbc DW |
3975 | } |
3976 | ||
c2c087e6 | 3977 | static size_t disks_to_mpb_size(int disks) |
cdddbdbc | 3978 | { |
c2c087e6 | 3979 | size_t size; |
cdddbdbc | 3980 | |
c2c087e6 DW |
3981 | size = sizeof(struct imsm_super); |
3982 | size += (disks - 1) * sizeof(struct imsm_disk); | |
3983 | size += 2 * sizeof(struct imsm_dev); | |
3984 | /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */ | |
3985 | size += (4 - 2) * sizeof(struct imsm_map); | |
3986 | /* 4 possible disk_ord_tbl's */ | |
3987 | size += 4 * (disks - 1) * sizeof(__u32); | |
bbab0940 TM |
3988 | /* maximum bbm log */ |
3989 | size += sizeof(struct bbm_log); | |
c2c087e6 DW |
3990 | |
3991 | return size; | |
3992 | } | |
3993 | ||
387fcd59 N |
3994 | static __u64 avail_size_imsm(struct supertype *st, __u64 devsize, |
3995 | unsigned long long data_offset) | |
c2c087e6 DW |
3996 | { |
3997 | if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS)) | |
3998 | return 0; | |
3999 | ||
4000 | return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS); | |
cdddbdbc DW |
4001 | } |
4002 | ||
ba2de7ba DW |
4003 | static void free_devlist(struct intel_super *super) |
4004 | { | |
4005 | struct intel_dev *dv; | |
4006 | ||
4007 | while (super->devlist) { | |
4008 | dv = super->devlist->next; | |
4009 | free(super->devlist->dev); | |
4010 | free(super->devlist); | |
4011 | super->devlist = dv; | |
4012 | } | |
4013 | } | |
4014 | ||
4015 | static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src) | |
4016 | { | |
4017 | memcpy(dest, src, sizeof_imsm_dev(src, 0)); | |
4018 | } | |
4019 | ||
c7b8547c MT |
4020 | static int compare_super_imsm(struct supertype *st, struct supertype *tst, |
4021 | int verbose) | |
cdddbdbc | 4022 | { |
601ffa78 | 4023 | /* return: |
cdddbdbc | 4024 | * 0 same, or first was empty, and second was copied |
601ffa78 | 4025 | * 1 sb are different |
cdddbdbc DW |
4026 | */ |
4027 | struct intel_super *first = st->sb; | |
4028 | struct intel_super *sec = tst->sb; | |
4029 | ||
5d500228 N |
4030 | if (!first) { |
4031 | st->sb = tst->sb; | |
4032 | tst->sb = NULL; | |
4033 | return 0; | |
4034 | } | |
601ffa78 | 4035 | |
8603ea6f LM |
4036 | /* in platform dependent environment test if the disks |
4037 | * use the same Intel hba | |
601ffa78 OS |
4038 | * if not on Intel hba at all, allow anything. |
4039 | * doesn't check HBAs if num_raid_devs is not set, as it means | |
4040 | * it is a free floating spare, and all spares regardless of HBA type | |
4041 | * will fall into separate container during the assembly | |
8603ea6f | 4042 | */ |
601ffa78 | 4043 | if (first->hba && sec->hba && first->anchor->num_raid_devs != 0) { |
6b781d33 | 4044 | if (first->hba->type != sec->hba->type) { |
c7b8547c MT |
4045 | if (verbose) |
4046 | pr_err("HBAs of devices do not match %s != %s\n", | |
4047 | get_sys_dev_type(first->hba->type), | |
4048 | get_sys_dev_type(sec->hba->type)); | |
601ffa78 | 4049 | return 1; |
6b781d33 AP |
4050 | } |
4051 | if (first->orom != sec->orom) { | |
c7b8547c MT |
4052 | if (verbose) |
4053 | pr_err("HBAs of devices do not match %s != %s\n", | |
4054 | first->hba->pci_id, sec->hba->pci_id); | |
601ffa78 | 4055 | return 1; |
8603ea6f LM |
4056 | } |
4057 | } | |
cdddbdbc | 4058 | |
d23fe947 DW |
4059 | if (first->anchor->num_raid_devs > 0 && |
4060 | sec->anchor->num_raid_devs > 0) { | |
a2b97981 DW |
4061 | /* Determine if these disks might ever have been |
4062 | * related. Further disambiguation can only take place | |
4063 | * in load_super_imsm_all | |
4064 | */ | |
4065 | __u32 first_family = first->anchor->orig_family_num; | |
4066 | __u32 sec_family = sec->anchor->orig_family_num; | |
4067 | ||
f796af5d DW |
4068 | if (memcmp(first->anchor->sig, sec->anchor->sig, |
4069 | MAX_SIGNATURE_LENGTH) != 0) | |
601ffa78 | 4070 | return 1; |
f796af5d | 4071 | |
a2b97981 DW |
4072 | if (first_family == 0) |
4073 | first_family = first->anchor->family_num; | |
4074 | if (sec_family == 0) | |
4075 | sec_family = sec->anchor->family_num; | |
4076 | ||
4077 | if (first_family != sec_family) | |
601ffa78 | 4078 | return 1; |
f796af5d | 4079 | |
d23fe947 | 4080 | } |
cdddbdbc | 4081 | |
601ffa78 OS |
4082 | /* if an anchor does not have num_raid_devs set then it is a free |
4083 | * floating spare. don't assosiate spare with any array, as during assembly | |
4084 | * spares shall fall into separate container, from which they can be moved | |
4085 | * when necessary | |
4086 | */ | |
4087 | if (first->anchor->num_raid_devs ^ sec->anchor->num_raid_devs) | |
4088 | return 1; | |
3e372e5a | 4089 | |
cdddbdbc DW |
4090 | return 0; |
4091 | } | |
4092 | ||
0030e8d6 DW |
4093 | static void fd2devname(int fd, char *name) |
4094 | { | |
0030e8d6 | 4095 | char *nm; |
0030e8d6 | 4096 | |
7c798f87 MT |
4097 | nm = fd2kname(fd); |
4098 | if (!nm) | |
0030e8d6 | 4099 | return; |
9587c373 | 4100 | |
7c798f87 | 4101 | snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm); |
0030e8d6 DW |
4102 | } |
4103 | ||
21e9380b AP |
4104 | static int nvme_get_serial(int fd, void *buf, size_t buf_len) |
4105 | { | |
fcebeb77 | 4106 | char path[PATH_MAX]; |
21e9380b AP |
4107 | char *name = fd2kname(fd); |
4108 | ||
4109 | if (!name) | |
4110 | return 1; | |
4111 | ||
4112 | if (strncmp(name, "nvme", 4) != 0) | |
4113 | return 1; | |
4114 | ||
fcebeb77 MT |
4115 | if (!diskfd_to_devpath(fd, 1, path)) |
4116 | return 1; | |
21e9380b | 4117 | |
fcebeb77 | 4118 | return devpath_to_char(path, "serial", buf, buf_len, 0); |
21e9380b AP |
4119 | } |
4120 | ||
cdddbdbc DW |
4121 | extern int scsi_get_serial(int fd, void *buf, size_t buf_len); |
4122 | ||
4123 | static int imsm_read_serial(int fd, char *devname, | |
6da53c0e | 4124 | __u8 *serial, size_t serial_buf_len) |
cdddbdbc | 4125 | { |
21e9380b | 4126 | char buf[50]; |
cdddbdbc | 4127 | int rv; |
6da53c0e | 4128 | size_t len; |
316e2bf4 DW |
4129 | char *dest; |
4130 | char *src; | |
21e9380b AP |
4131 | unsigned int i; |
4132 | ||
4133 | memset(buf, 0, sizeof(buf)); | |
cdddbdbc | 4134 | |
d1cd231a KT |
4135 | if (check_env("IMSM_DEVNAME_AS_SERIAL")) { |
4136 | memset(serial, 0, serial_buf_len); | |
4137 | fd2devname(fd, (char *) serial); | |
4138 | return 0; | |
4139 | } | |
4140 | ||
21e9380b | 4141 | rv = nvme_get_serial(fd, buf, sizeof(buf)); |
cdddbdbc | 4142 | |
21e9380b AP |
4143 | if (rv) |
4144 | rv = scsi_get_serial(fd, buf, sizeof(buf)); | |
f9ba0ff1 | 4145 | |
cdddbdbc DW |
4146 | if (rv != 0) { |
4147 | if (devname) | |
e7b84f9d N |
4148 | pr_err("Failed to retrieve serial for %s\n", |
4149 | devname); | |
cdddbdbc DW |
4150 | return rv; |
4151 | } | |
4152 | ||
316e2bf4 DW |
4153 | /* trim all whitespace and non-printable characters and convert |
4154 | * ':' to ';' | |
4155 | */ | |
21e9380b AP |
4156 | for (i = 0, dest = buf; i < sizeof(buf) && buf[i]; i++) { |
4157 | src = &buf[i]; | |
316e2bf4 DW |
4158 | if (*src > 0x20) { |
4159 | /* ':' is reserved for use in placeholder serial | |
4160 | * numbers for missing disks | |
4161 | */ | |
4162 | if (*src == ':') | |
4163 | *dest++ = ';'; | |
4164 | else | |
4165 | *dest++ = *src; | |
4166 | } | |
4167 | } | |
21e9380b AP |
4168 | len = dest - buf; |
4169 | dest = buf; | |
316e2bf4 | 4170 | |
6da53c0e BK |
4171 | if (len > serial_buf_len) { |
4172 | /* truncate leading characters */ | |
4173 | dest += len - serial_buf_len; | |
4174 | len = serial_buf_len; | |
316e2bf4 | 4175 | } |
5c3db629 | 4176 | |
6da53c0e | 4177 | memset(serial, 0, serial_buf_len); |
316e2bf4 | 4178 | memcpy(serial, dest, len); |
cdddbdbc DW |
4179 | |
4180 | return 0; | |
4181 | } | |
4182 | ||
1f24f035 DW |
4183 | static int serialcmp(__u8 *s1, __u8 *s2) |
4184 | { | |
4185 | return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN); | |
4186 | } | |
4187 | ||
4188 | static void serialcpy(__u8 *dest, __u8 *src) | |
4189 | { | |
4190 | strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN); | |
4191 | } | |
4192 | ||
54c2c1ea DW |
4193 | static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super) |
4194 | { | |
4195 | struct dl *dl; | |
4196 | ||
4197 | for (dl = super->disks; dl; dl = dl->next) | |
4198 | if (serialcmp(dl->serial, serial) == 0) | |
4199 | break; | |
4200 | ||
4201 | return dl; | |
4202 | } | |
4203 | ||
a2b97981 DW |
4204 | static struct imsm_disk * |
4205 | __serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx) | |
4206 | { | |
4207 | int i; | |
4208 | ||
4209 | for (i = 0; i < mpb->num_disks; i++) { | |
4210 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
4211 | ||
4212 | if (serialcmp(disk->serial, serial) == 0) { | |
4213 | if (idx) | |
4214 | *idx = i; | |
4215 | return disk; | |
4216 | } | |
4217 | } | |
4218 | ||
4219 | return NULL; | |
4220 | } | |
4221 | ||
cdddbdbc DW |
4222 | static int |
4223 | load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd) | |
4224 | { | |
a2b97981 | 4225 | struct imsm_disk *disk; |
cdddbdbc DW |
4226 | struct dl *dl; |
4227 | struct stat stb; | |
cdddbdbc | 4228 | int rv; |
a2b97981 | 4229 | char name[40]; |
d23fe947 DW |
4230 | __u8 serial[MAX_RAID_SERIAL_LEN]; |
4231 | ||
6da53c0e | 4232 | rv = imsm_read_serial(fd, devname, serial, MAX_RAID_SERIAL_LEN); |
d23fe947 DW |
4233 | |
4234 | if (rv != 0) | |
4235 | return 2; | |
4236 | ||
503975b9 | 4237 | dl = xcalloc(1, sizeof(*dl)); |
cdddbdbc | 4238 | |
a2b97981 DW |
4239 | fstat(fd, &stb); |
4240 | dl->major = major(stb.st_rdev); | |
4241 | dl->minor = minor(stb.st_rdev); | |
4242 | dl->next = super->disks; | |
4243 | dl->fd = keep_fd ? fd : -1; | |
4244 | assert(super->disks == NULL); | |
4245 | super->disks = dl; | |
4246 | serialcpy(dl->serial, serial); | |
4247 | dl->index = -2; | |
4248 | dl->e = NULL; | |
4249 | fd2devname(fd, name); | |
4250 | if (devname) | |
503975b9 | 4251 | dl->devname = xstrdup(devname); |
a2b97981 | 4252 | else |
503975b9 | 4253 | dl->devname = xstrdup(name); |
cdddbdbc | 4254 | |
d23fe947 | 4255 | /* look up this disk's index in the current anchor */ |
a2b97981 DW |
4256 | disk = __serial_to_disk(dl->serial, super->anchor, &dl->index); |
4257 | if (disk) { | |
4258 | dl->disk = *disk; | |
4259 | /* only set index on disks that are a member of a | |
4260 | * populated contianer, i.e. one with raid_devs | |
4261 | */ | |
4262 | if (is_failed(&dl->disk)) | |
3f6efecc | 4263 | dl->index = -2; |
2432ce9b | 4264 | else if (is_spare(&dl->disk) || is_journal(&dl->disk)) |
a2b97981 | 4265 | dl->index = -1; |
3f6efecc DW |
4266 | } |
4267 | ||
949c47a0 DW |
4268 | return 0; |
4269 | } | |
4270 | ||
0c046afd DW |
4271 | /* When migrating map0 contains the 'destination' state while map1 |
4272 | * contains the current state. When not migrating map0 contains the | |
4273 | * current state. This routine assumes that map[0].map_state is set to | |
4274 | * the current array state before being called. | |
4275 | * | |
4276 | * Migration is indicated by one of the following states | |
4277 | * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed) | |
e3bba0e0 | 4278 | * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal |
0c046afd | 4279 | * map1state=unitialized) |
1484e727 | 4280 | * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal |
0c046afd | 4281 | * map1state=normal) |
e3bba0e0 | 4282 | * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal |
0c046afd | 4283 | * map1state=degraded) |
8e59f3d8 AK |
4284 | * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal |
4285 | * map1state=normal) | |
0c046afd | 4286 | */ |
8e59f3d8 AK |
4287 | static void migrate(struct imsm_dev *dev, struct intel_super *super, |
4288 | __u8 to_state, int migr_type) | |
3393c6af | 4289 | { |
0c046afd | 4290 | struct imsm_map *dest; |
238c0a71 | 4291 | struct imsm_map *src = get_imsm_map(dev, MAP_0); |
3393c6af | 4292 | |
0c046afd | 4293 | dev->vol.migr_state = 1; |
1484e727 | 4294 | set_migr_type(dev, migr_type); |
4036e7ee | 4295 | set_vol_curr_migr_unit(dev, 0); |
238c0a71 | 4296 | dest = get_imsm_map(dev, MAP_1); |
0c046afd | 4297 | |
0556e1a2 | 4298 | /* duplicate and then set the target end state in map[0] */ |
3393c6af | 4299 | memcpy(dest, src, sizeof_imsm_map(src)); |
fb12a745 | 4300 | if (migr_type == MIGR_GEN_MIGR) { |
0556e1a2 DW |
4301 | __u32 ord; |
4302 | int i; | |
4303 | ||
4304 | for (i = 0; i < src->num_members; i++) { | |
4305 | ord = __le32_to_cpu(src->disk_ord_tbl[i]); | |
4306 | set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord)); | |
4307 | } | |
4308 | } | |
4309 | ||
8e59f3d8 AK |
4310 | if (migr_type == MIGR_GEN_MIGR) |
4311 | /* Clear migration record */ | |
4312 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
4313 | ||
0c046afd | 4314 | src->map_state = to_state; |
949c47a0 | 4315 | } |
f8f603f1 | 4316 | |
809da78e AK |
4317 | static void end_migration(struct imsm_dev *dev, struct intel_super *super, |
4318 | __u8 map_state) | |
f8f603f1 | 4319 | { |
238c0a71 AK |
4320 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
4321 | struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state == 0 ? | |
4322 | MAP_0 : MAP_1); | |
28bce06f | 4323 | int i, j; |
0556e1a2 DW |
4324 | |
4325 | /* merge any IMSM_ORD_REBUILD bits that were not successfully | |
4326 | * completed in the last migration. | |
4327 | * | |
28bce06f | 4328 | * FIXME add support for raid-level-migration |
0556e1a2 | 4329 | */ |
195d1d76 | 4330 | if (map_state != map->map_state && (is_gen_migration(dev) == false) && |
089f9d79 | 4331 | prev->map_state != IMSM_T_STATE_UNINITIALIZED) { |
809da78e AK |
4332 | /* when final map state is other than expected |
4333 | * merge maps (not for migration) | |
4334 | */ | |
4335 | int failed; | |
4336 | ||
4337 | for (i = 0; i < prev->num_members; i++) | |
4338 | for (j = 0; j < map->num_members; j++) | |
4339 | /* during online capacity expansion | |
4340 | * disks position can be changed | |
4341 | * if takeover is used | |
4342 | */ | |
4343 | if (ord_to_idx(map->disk_ord_tbl[j]) == | |
4344 | ord_to_idx(prev->disk_ord_tbl[i])) { | |
4345 | map->disk_ord_tbl[j] |= | |
4346 | prev->disk_ord_tbl[i]; | |
4347 | break; | |
4348 | } | |
4349 | failed = imsm_count_failed(super, dev, MAP_0); | |
4350 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
4351 | } | |
f8f603f1 DW |
4352 | |
4353 | dev->vol.migr_state = 0; | |
ea672ee1 | 4354 | set_migr_type(dev, 0); |
4036e7ee | 4355 | set_vol_curr_migr_unit(dev, 0); |
f8f603f1 DW |
4356 | map->map_state = map_state; |
4357 | } | |
949c47a0 DW |
4358 | |
4359 | static int parse_raid_devices(struct intel_super *super) | |
4360 | { | |
4361 | int i; | |
4362 | struct imsm_dev *dev_new; | |
4d7b1503 | 4363 | size_t len, len_migr; |
401d313b | 4364 | size_t max_len = 0; |
4d7b1503 DW |
4365 | size_t space_needed = 0; |
4366 | struct imsm_super *mpb = super->anchor; | |
949c47a0 DW |
4367 | |
4368 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
4369 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
ba2de7ba | 4370 | struct intel_dev *dv; |
949c47a0 | 4371 | |
4d7b1503 DW |
4372 | len = sizeof_imsm_dev(dev_iter, 0); |
4373 | len_migr = sizeof_imsm_dev(dev_iter, 1); | |
4374 | if (len_migr > len) | |
4375 | space_needed += len_migr - len; | |
ca9de185 | 4376 | |
503975b9 | 4377 | dv = xmalloc(sizeof(*dv)); |
401d313b AK |
4378 | if (max_len < len_migr) |
4379 | max_len = len_migr; | |
4380 | if (max_len > len_migr) | |
4381 | space_needed += max_len - len_migr; | |
503975b9 | 4382 | dev_new = xmalloc(max_len); |
949c47a0 | 4383 | imsm_copy_dev(dev_new, dev_iter); |
ba2de7ba DW |
4384 | dv->dev = dev_new; |
4385 | dv->index = i; | |
4386 | dv->next = super->devlist; | |
4387 | super->devlist = dv; | |
949c47a0 | 4388 | } |
cdddbdbc | 4389 | |
4d7b1503 DW |
4390 | /* ensure that super->buf is large enough when all raid devices |
4391 | * are migrating | |
4392 | */ | |
4393 | if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) { | |
4394 | void *buf; | |
4395 | ||
f36a9ecd PB |
4396 | len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed, |
4397 | super->sector_size); | |
4398 | if (posix_memalign(&buf, MAX_SECTOR_SIZE, len) != 0) | |
4d7b1503 DW |
4399 | return 1; |
4400 | ||
1f45a8ad DW |
4401 | memcpy(buf, super->buf, super->len); |
4402 | memset(buf + super->len, 0, len - super->len); | |
4d7b1503 DW |
4403 | free(super->buf); |
4404 | super->buf = buf; | |
4405 | super->len = len; | |
4406 | } | |
ca9de185 | 4407 | |
bbab0940 TM |
4408 | super->extra_space += space_needed; |
4409 | ||
cdddbdbc DW |
4410 | return 0; |
4411 | } | |
4412 | ||
e2f41b2c AK |
4413 | /******************************************************************************* |
4414 | * Function: check_mpb_migr_compatibility | |
4415 | * Description: Function checks for unsupported migration features: | |
4416 | * - migration optimization area (pba_of_lba0) | |
4417 | * - descending reshape (ascending_migr) | |
4418 | * Parameters: | |
4419 | * super : imsm metadata information | |
4420 | * Returns: | |
4421 | * 0 : migration is compatible | |
4422 | * -1 : migration is not compatible | |
4423 | ******************************************************************************/ | |
4424 | int check_mpb_migr_compatibility(struct intel_super *super) | |
4425 | { | |
4426 | struct imsm_map *map0, *map1; | |
4427 | struct migr_record *migr_rec = super->migr_rec; | |
4428 | int i; | |
4429 | ||
4430 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
4431 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
4432 | ||
756a15f3 | 4433 | if (dev_iter->vol.migr_state == 1 && |
e2f41b2c AK |
4434 | dev_iter->vol.migr_type == MIGR_GEN_MIGR) { |
4435 | /* This device is migrating */ | |
238c0a71 AK |
4436 | map0 = get_imsm_map(dev_iter, MAP_0); |
4437 | map1 = get_imsm_map(dev_iter, MAP_1); | |
5551b113 | 4438 | if (pba_of_lba0(map0) != pba_of_lba0(map1)) |
e2f41b2c AK |
4439 | /* migration optimization area was used */ |
4440 | return -1; | |
fc54fe7a JS |
4441 | if (migr_rec->ascending_migr == 0 && |
4442 | migr_rec->dest_depth_per_unit > 0) | |
e2f41b2c AK |
4443 | /* descending reshape not supported yet */ |
4444 | return -1; | |
4445 | } | |
4446 | } | |
4447 | return 0; | |
4448 | } | |
4449 | ||
d23fe947 | 4450 | static void __free_imsm(struct intel_super *super, int free_disks); |
9ca2c81c | 4451 | |
cdddbdbc | 4452 | /* load_imsm_mpb - read matrix metadata |
f2f5c343 | 4453 | * allocates super->mpb to be freed by free_imsm |
cdddbdbc DW |
4454 | */ |
4455 | static int load_imsm_mpb(int fd, struct intel_super *super, char *devname) | |
4456 | { | |
4457 | unsigned long long dsize; | |
cdddbdbc | 4458 | unsigned long long sectors; |
f36a9ecd | 4459 | unsigned int sector_size = super->sector_size; |
cdddbdbc | 4460 | struct stat; |
6416d527 | 4461 | struct imsm_super *anchor; |
cdddbdbc DW |
4462 | __u32 check_sum; |
4463 | ||
cdddbdbc | 4464 | get_dev_size(fd, NULL, &dsize); |
f36a9ecd | 4465 | if (dsize < 2*sector_size) { |
64436f06 | 4466 | if (devname) |
e7b84f9d N |
4467 | pr_err("%s: device to small for imsm\n", |
4468 | devname); | |
64436f06 N |
4469 | return 1; |
4470 | } | |
cdddbdbc | 4471 | |
f36a9ecd | 4472 | if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0) { |
cdddbdbc | 4473 | if (devname) |
e7b84f9d N |
4474 | pr_err("Cannot seek to anchor block on %s: %s\n", |
4475 | devname, strerror(errno)); | |
cdddbdbc DW |
4476 | return 1; |
4477 | } | |
4478 | ||
f36a9ecd | 4479 | if (posix_memalign((void **)&anchor, sector_size, sector_size) != 0) { |
ad97895e | 4480 | if (devname) |
7a862a02 | 4481 | pr_err("Failed to allocate imsm anchor buffer on %s\n", devname); |
ad97895e DW |
4482 | return 1; |
4483 | } | |
466070ad | 4484 | if ((unsigned int)read(fd, anchor, sector_size) != sector_size) { |
cdddbdbc | 4485 | if (devname) |
e7b84f9d N |
4486 | pr_err("Cannot read anchor block on %s: %s\n", |
4487 | devname, strerror(errno)); | |
6416d527 | 4488 | free(anchor); |
cdddbdbc DW |
4489 | return 1; |
4490 | } | |
4491 | ||
6416d527 | 4492 | if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) { |
cdddbdbc | 4493 | if (devname) |
e7b84f9d | 4494 | pr_err("no IMSM anchor on %s\n", devname); |
6416d527 | 4495 | free(anchor); |
cdddbdbc DW |
4496 | return 2; |
4497 | } | |
4498 | ||
d23fe947 | 4499 | __free_imsm(super, 0); |
f2f5c343 LM |
4500 | /* reload capability and hba */ |
4501 | ||
4502 | /* capability and hba must be updated with new super allocation */ | |
d424212e | 4503 | find_intel_hba_capability(fd, super, devname); |
f36a9ecd PB |
4504 | super->len = ROUND_UP(anchor->mpb_size, sector_size); |
4505 | if (posix_memalign(&super->buf, MAX_SECTOR_SIZE, super->len) != 0) { | |
cdddbdbc | 4506 | if (devname) |
e7b84f9d N |
4507 | pr_err("unable to allocate %zu byte mpb buffer\n", |
4508 | super->len); | |
6416d527 | 4509 | free(anchor); |
cdddbdbc DW |
4510 | return 2; |
4511 | } | |
f36a9ecd | 4512 | memcpy(super->buf, anchor, sector_size); |
cdddbdbc | 4513 | |
f36a9ecd | 4514 | sectors = mpb_sectors(anchor, sector_size) - 1; |
6416d527 | 4515 | free(anchor); |
8e59f3d8 | 4516 | |
85337573 AO |
4517 | if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE, |
4518 | MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) { | |
1ade5cc1 | 4519 | pr_err("could not allocate migr_rec buffer\n"); |
8e59f3d8 | 4520 | free(super->buf); |
50cd06b4 | 4521 | super->buf = NULL; |
8e59f3d8 AK |
4522 | return 2; |
4523 | } | |
51d83f5d | 4524 | super->clean_migration_record_by_mdmon = 0; |
8e59f3d8 | 4525 | |
949c47a0 | 4526 | if (!sectors) { |
ecf45690 DW |
4527 | check_sum = __gen_imsm_checksum(super->anchor); |
4528 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
4529 | if (devname) | |
e7b84f9d N |
4530 | pr_err("IMSM checksum %x != %x on %s\n", |
4531 | check_sum, | |
4532 | __le32_to_cpu(super->anchor->check_sum), | |
4533 | devname); | |
ecf45690 DW |
4534 | return 2; |
4535 | } | |
4536 | ||
a2b97981 | 4537 | return 0; |
949c47a0 | 4538 | } |
cdddbdbc DW |
4539 | |
4540 | /* read the extended mpb */ | |
f36a9ecd | 4541 | if (lseek64(fd, dsize - (sector_size * (2 + sectors)), SEEK_SET) < 0) { |
cdddbdbc | 4542 | if (devname) |
e7b84f9d N |
4543 | pr_err("Cannot seek to extended mpb on %s: %s\n", |
4544 | devname, strerror(errno)); | |
cdddbdbc DW |
4545 | return 1; |
4546 | } | |
4547 | ||
f36a9ecd PB |
4548 | if ((unsigned int)read(fd, super->buf + sector_size, |
4549 | super->len - sector_size) != super->len - sector_size) { | |
cdddbdbc | 4550 | if (devname) |
e7b84f9d N |
4551 | pr_err("Cannot read extended mpb on %s: %s\n", |
4552 | devname, strerror(errno)); | |
cdddbdbc DW |
4553 | return 2; |
4554 | } | |
4555 | ||
949c47a0 DW |
4556 | check_sum = __gen_imsm_checksum(super->anchor); |
4557 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
cdddbdbc | 4558 | if (devname) |
e7b84f9d N |
4559 | pr_err("IMSM checksum %x != %x on %s\n", |
4560 | check_sum, __le32_to_cpu(super->anchor->check_sum), | |
4561 | devname); | |
db575f3b | 4562 | return 3; |
cdddbdbc DW |
4563 | } |
4564 | ||
a2b97981 DW |
4565 | return 0; |
4566 | } | |
4567 | ||
8e59f3d8 AK |
4568 | static int read_imsm_migr_rec(int fd, struct intel_super *super); |
4569 | ||
97f81ee2 CA |
4570 | /* clears hi bits in metadata if MPB_ATTRIB_2TB_DISK not set */ |
4571 | static void clear_hi(struct intel_super *super) | |
4572 | { | |
4573 | struct imsm_super *mpb = super->anchor; | |
4574 | int i, n; | |
4575 | if (mpb->attributes & MPB_ATTRIB_2TB_DISK) | |
4576 | return; | |
4577 | for (i = 0; i < mpb->num_disks; ++i) { | |
4578 | struct imsm_disk *disk = &mpb->disk[i]; | |
4579 | disk->total_blocks_hi = 0; | |
4580 | } | |
4581 | for (i = 0; i < mpb->num_raid_devs; ++i) { | |
4582 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
97f81ee2 CA |
4583 | for (n = 0; n < 2; ++n) { |
4584 | struct imsm_map *map = get_imsm_map(dev, n); | |
4585 | if (!map) | |
4586 | continue; | |
4587 | map->pba_of_lba0_hi = 0; | |
4588 | map->blocks_per_member_hi = 0; | |
4589 | map->num_data_stripes_hi = 0; | |
4590 | } | |
4591 | } | |
4592 | } | |
4593 | ||
a2b97981 DW |
4594 | static int |
4595 | load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd) | |
4596 | { | |
4597 | int err; | |
4598 | ||
4599 | err = load_imsm_mpb(fd, super, devname); | |
4600 | if (err) | |
4601 | return err; | |
f36a9ecd PB |
4602 | if (super->sector_size == 4096) |
4603 | convert_from_4k(super); | |
a2b97981 DW |
4604 | err = load_imsm_disk(fd, super, devname, keep_fd); |
4605 | if (err) | |
4606 | return err; | |
4607 | err = parse_raid_devices(super); | |
8d67477f TM |
4608 | if (err) |
4609 | return err; | |
4610 | err = load_bbm_log(super); | |
97f81ee2 | 4611 | clear_hi(super); |
a2b97981 | 4612 | return err; |
cdddbdbc DW |
4613 | } |
4614 | ||
4389ce73 | 4615 | static void __free_imsm_disk(struct dl *d, int do_close) |
ae6aad82 | 4616 | { |
4389ce73 MT |
4617 | if (do_close) |
4618 | close_fd(&d->fd); | |
ae6aad82 DW |
4619 | if (d->devname) |
4620 | free(d->devname); | |
0dcecb2e DW |
4621 | if (d->e) |
4622 | free(d->e); | |
ae6aad82 DW |
4623 | free(d); |
4624 | ||
4625 | } | |
1a64be56 | 4626 | |
cdddbdbc DW |
4627 | static void free_imsm_disks(struct intel_super *super) |
4628 | { | |
47ee5a45 | 4629 | struct dl *d; |
cdddbdbc | 4630 | |
47ee5a45 DW |
4631 | while (super->disks) { |
4632 | d = super->disks; | |
cdddbdbc | 4633 | super->disks = d->next; |
3a85bf0e | 4634 | __free_imsm_disk(d, 1); |
cdddbdbc | 4635 | } |
cb82edca AK |
4636 | while (super->disk_mgmt_list) { |
4637 | d = super->disk_mgmt_list; | |
4638 | super->disk_mgmt_list = d->next; | |
3a85bf0e | 4639 | __free_imsm_disk(d, 1); |
cb82edca | 4640 | } |
47ee5a45 DW |
4641 | while (super->missing) { |
4642 | d = super->missing; | |
4643 | super->missing = d->next; | |
3a85bf0e | 4644 | __free_imsm_disk(d, 1); |
47ee5a45 DW |
4645 | } |
4646 | ||
cdddbdbc DW |
4647 | } |
4648 | ||
9ca2c81c | 4649 | /* free all the pieces hanging off of a super pointer */ |
d23fe947 | 4650 | static void __free_imsm(struct intel_super *super, int free_disks) |
cdddbdbc | 4651 | { |
88654014 LM |
4652 | struct intel_hba *elem, *next; |
4653 | ||
9ca2c81c | 4654 | if (super->buf) { |
949c47a0 | 4655 | free(super->buf); |
9ca2c81c DW |
4656 | super->buf = NULL; |
4657 | } | |
f2f5c343 LM |
4658 | /* unlink capability description */ |
4659 | super->orom = NULL; | |
8e59f3d8 AK |
4660 | if (super->migr_rec_buf) { |
4661 | free(super->migr_rec_buf); | |
4662 | super->migr_rec_buf = NULL; | |
4663 | } | |
d23fe947 DW |
4664 | if (free_disks) |
4665 | free_imsm_disks(super); | |
ba2de7ba | 4666 | free_devlist(super); |
88654014 LM |
4667 | elem = super->hba; |
4668 | while (elem) { | |
4669 | if (elem->path) | |
4670 | free((void *)elem->path); | |
4671 | next = elem->next; | |
4672 | free(elem); | |
4673 | elem = next; | |
88c32bb1 | 4674 | } |
8d67477f TM |
4675 | if (super->bbm_log) |
4676 | free(super->bbm_log); | |
88654014 | 4677 | super->hba = NULL; |
cdddbdbc DW |
4678 | } |
4679 | ||
9ca2c81c DW |
4680 | static void free_imsm(struct intel_super *super) |
4681 | { | |
d23fe947 | 4682 | __free_imsm(super, 1); |
928f1424 | 4683 | free(super->bb.entries); |
9ca2c81c DW |
4684 | free(super); |
4685 | } | |
cdddbdbc DW |
4686 | |
4687 | static void free_super_imsm(struct supertype *st) | |
4688 | { | |
4689 | struct intel_super *super = st->sb; | |
4690 | ||
4691 | if (!super) | |
4692 | return; | |
4693 | ||
4694 | free_imsm(super); | |
4695 | st->sb = NULL; | |
4696 | } | |
4697 | ||
49133e57 | 4698 | static struct intel_super *alloc_super(void) |
c2c087e6 | 4699 | { |
503975b9 | 4700 | struct intel_super *super = xcalloc(1, sizeof(*super)); |
c2c087e6 | 4701 | |
503975b9 N |
4702 | super->current_vol = -1; |
4703 | super->create_offset = ~((unsigned long long) 0); | |
928f1424 TM |
4704 | |
4705 | super->bb.entries = xmalloc(BBM_LOG_MAX_ENTRIES * | |
4706 | sizeof(struct md_bb_entry)); | |
4707 | if (!super->bb.entries) { | |
4708 | free(super); | |
4709 | return NULL; | |
4710 | } | |
4711 | ||
c2c087e6 DW |
4712 | return super; |
4713 | } | |
4714 | ||
f0f5a016 LM |
4715 | /* |
4716 | * find and allocate hba and OROM/EFI based on valid fd of RAID component device | |
4717 | */ | |
d424212e | 4718 | static int find_intel_hba_capability(int fd, struct intel_super *super, char *devname) |
f0f5a016 LM |
4719 | { |
4720 | struct sys_dev *hba_name; | |
4721 | int rv = 0; | |
4722 | ||
4389ce73 | 4723 | if (is_fd_valid(fd) && test_partition(fd)) { |
3a30e28e MT |
4724 | pr_err("imsm: %s is a partition, cannot be used in IMSM\n", |
4725 | devname); | |
4726 | return 1; | |
4727 | } | |
420dafcd | 4728 | if (!is_fd_valid(fd) || check_no_platform()) { |
f2f5c343 | 4729 | super->orom = NULL; |
f0f5a016 LM |
4730 | super->hba = NULL; |
4731 | return 0; | |
4732 | } | |
4733 | hba_name = find_disk_attached_hba(fd, NULL); | |
4734 | if (!hba_name) { | |
d424212e | 4735 | if (devname) |
e7b84f9d N |
4736 | pr_err("%s is not attached to Intel(R) RAID controller.\n", |
4737 | devname); | |
f0f5a016 LM |
4738 | return 1; |
4739 | } | |
4740 | rv = attach_hba_to_super(super, hba_name); | |
4741 | if (rv == 2) { | |
d424212e N |
4742 | if (devname) { |
4743 | struct intel_hba *hba = super->hba; | |
f0f5a016 | 4744 | |
60f0f54d PB |
4745 | pr_err("%s is attached to Intel(R) %s %s (%s),\n" |
4746 | " but the container is assigned to Intel(R) %s %s (", | |
d424212e | 4747 | devname, |
614902f6 | 4748 | get_sys_dev_type(hba_name->type), |
75350d87 KF |
4749 | hba_name->type == SYS_DEV_VMD || hba_name->type == SYS_DEV_SATA_VMD ? |
4750 | "domain" : "RAID controller", | |
f0f5a016 | 4751 | hba_name->pci_id ? : "Err!", |
60f0f54d | 4752 | get_sys_dev_type(super->hba->type), |
75350d87 KF |
4753 | hba->type == SYS_DEV_VMD || hba_name->type == SYS_DEV_SATA_VMD ? |
4754 | "domain" : "RAID controller"); | |
f0f5a016 | 4755 | |
f0f5a016 LM |
4756 | while (hba) { |
4757 | fprintf(stderr, "%s", hba->pci_id ? : "Err!"); | |
4758 | if (hba->next) | |
4759 | fprintf(stderr, ", "); | |
4760 | hba = hba->next; | |
4761 | } | |
6b781d33 | 4762 | fprintf(stderr, ").\n" |
cca67208 | 4763 | " Mixing devices attached to different controllers is not allowed.\n"); |
f0f5a016 | 4764 | } |
f0f5a016 LM |
4765 | return 2; |
4766 | } | |
6b781d33 | 4767 | super->orom = find_imsm_capability(hba_name); |
f2f5c343 LM |
4768 | if (!super->orom) |
4769 | return 3; | |
614902f6 | 4770 | |
f0f5a016 LM |
4771 | return 0; |
4772 | } | |
4773 | ||
47ee5a45 DW |
4774 | /* find_missing - helper routine for load_super_imsm_all that identifies |
4775 | * disks that have disappeared from the system. This routine relies on | |
4776 | * the mpb being uptodate, which it is at load time. | |
4777 | */ | |
4778 | static int find_missing(struct intel_super *super) | |
4779 | { | |
4780 | int i; | |
4781 | struct imsm_super *mpb = super->anchor; | |
4782 | struct dl *dl; | |
4783 | struct imsm_disk *disk; | |
47ee5a45 DW |
4784 | |
4785 | for (i = 0; i < mpb->num_disks; i++) { | |
4786 | disk = __get_imsm_disk(mpb, i); | |
54c2c1ea | 4787 | dl = serial_to_dl(disk->serial, super); |
47ee5a45 DW |
4788 | if (dl) |
4789 | continue; | |
47ee5a45 | 4790 | |
503975b9 | 4791 | dl = xmalloc(sizeof(*dl)); |
47ee5a45 DW |
4792 | dl->major = 0; |
4793 | dl->minor = 0; | |
4794 | dl->fd = -1; | |
503975b9 | 4795 | dl->devname = xstrdup("missing"); |
47ee5a45 DW |
4796 | dl->index = i; |
4797 | serialcpy(dl->serial, disk->serial); | |
4798 | dl->disk = *disk; | |
689c9bf3 | 4799 | dl->e = NULL; |
47ee5a45 DW |
4800 | dl->next = super->missing; |
4801 | super->missing = dl; | |
4802 | } | |
4803 | ||
4804 | return 0; | |
4805 | } | |
4806 | ||
a2b97981 DW |
4807 | static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list) |
4808 | { | |
4809 | struct intel_disk *idisk = disk_list; | |
4810 | ||
4811 | while (idisk) { | |
4812 | if (serialcmp(idisk->disk.serial, serial) == 0) | |
4813 | break; | |
4814 | idisk = idisk->next; | |
4815 | } | |
4816 | ||
4817 | return idisk; | |
4818 | } | |
4819 | ||
4820 | static int __prep_thunderdome(struct intel_super **table, int tbl_size, | |
4821 | struct intel_super *super, | |
4822 | struct intel_disk **disk_list) | |
4823 | { | |
4824 | struct imsm_disk *d = &super->disks->disk; | |
4825 | struct imsm_super *mpb = super->anchor; | |
4826 | int i, j; | |
4827 | ||
4828 | for (i = 0; i < tbl_size; i++) { | |
4829 | struct imsm_super *tbl_mpb = table[i]->anchor; | |
4830 | struct imsm_disk *tbl_d = &table[i]->disks->disk; | |
4831 | ||
4832 | if (tbl_mpb->family_num == mpb->family_num) { | |
4833 | if (tbl_mpb->check_sum == mpb->check_sum) { | |
1ade5cc1 N |
4834 | dprintf("mpb from %d:%d matches %d:%d\n", |
4835 | super->disks->major, | |
a2b97981 DW |
4836 | super->disks->minor, |
4837 | table[i]->disks->major, | |
4838 | table[i]->disks->minor); | |
4839 | break; | |
4840 | } | |
4841 | ||
4842 | if (((is_configured(d) && !is_configured(tbl_d)) || | |
4843 | is_configured(d) == is_configured(tbl_d)) && | |
4844 | tbl_mpb->generation_num < mpb->generation_num) { | |
4845 | /* current version of the mpb is a | |
4846 | * better candidate than the one in | |
4847 | * super_table, but copy over "cross | |
4848 | * generational" status | |
4849 | */ | |
4850 | struct intel_disk *idisk; | |
4851 | ||
1ade5cc1 N |
4852 | dprintf("mpb from %d:%d replaces %d:%d\n", |
4853 | super->disks->major, | |
a2b97981 DW |
4854 | super->disks->minor, |
4855 | table[i]->disks->major, | |
4856 | table[i]->disks->minor); | |
4857 | ||
4858 | idisk = disk_list_get(tbl_d->serial, *disk_list); | |
4859 | if (idisk && is_failed(&idisk->disk)) | |
4860 | tbl_d->status |= FAILED_DISK; | |
4861 | break; | |
4862 | } else { | |
4863 | struct intel_disk *idisk; | |
4864 | struct imsm_disk *disk; | |
4865 | ||
4866 | /* tbl_mpb is more up to date, but copy | |
4867 | * over cross generational status before | |
4868 | * returning | |
4869 | */ | |
4870 | disk = __serial_to_disk(d->serial, mpb, NULL); | |
4871 | if (disk && is_failed(disk)) | |
4872 | d->status |= FAILED_DISK; | |
4873 | ||
4874 | idisk = disk_list_get(d->serial, *disk_list); | |
4875 | if (idisk) { | |
4876 | idisk->owner = i; | |
4877 | if (disk && is_configured(disk)) | |
4878 | idisk->disk.status |= CONFIGURED_DISK; | |
4879 | } | |
4880 | ||
1ade5cc1 N |
4881 | dprintf("mpb from %d:%d prefer %d:%d\n", |
4882 | super->disks->major, | |
a2b97981 DW |
4883 | super->disks->minor, |
4884 | table[i]->disks->major, | |
4885 | table[i]->disks->minor); | |
4886 | ||
4887 | return tbl_size; | |
4888 | } | |
4889 | } | |
4890 | } | |
4891 | ||
4892 | if (i >= tbl_size) | |
4893 | table[tbl_size++] = super; | |
4894 | else | |
4895 | table[i] = super; | |
4896 | ||
4897 | /* update/extend the merged list of imsm_disk records */ | |
4898 | for (j = 0; j < mpb->num_disks; j++) { | |
4899 | struct imsm_disk *disk = __get_imsm_disk(mpb, j); | |
4900 | struct intel_disk *idisk; | |
4901 | ||
4902 | idisk = disk_list_get(disk->serial, *disk_list); | |
4903 | if (idisk) { | |
4904 | idisk->disk.status |= disk->status; | |
4905 | if (is_configured(&idisk->disk) || | |
4906 | is_failed(&idisk->disk)) | |
4907 | idisk->disk.status &= ~(SPARE_DISK); | |
4908 | } else { | |
503975b9 | 4909 | idisk = xcalloc(1, sizeof(*idisk)); |
a2b97981 DW |
4910 | idisk->owner = IMSM_UNKNOWN_OWNER; |
4911 | idisk->disk = *disk; | |
4912 | idisk->next = *disk_list; | |
4913 | *disk_list = idisk; | |
4914 | } | |
4915 | ||
4916 | if (serialcmp(idisk->disk.serial, d->serial) == 0) | |
4917 | idisk->owner = i; | |
4918 | } | |
4919 | ||
4920 | return tbl_size; | |
4921 | } | |
4922 | ||
4923 | static struct intel_super * | |
4924 | validate_members(struct intel_super *super, struct intel_disk *disk_list, | |
4925 | const int owner) | |
4926 | { | |
4927 | struct imsm_super *mpb = super->anchor; | |
4928 | int ok_count = 0; | |
4929 | int i; | |
4930 | ||
4931 | for (i = 0; i < mpb->num_disks; i++) { | |
4932 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
4933 | struct intel_disk *idisk; | |
4934 | ||
4935 | idisk = disk_list_get(disk->serial, disk_list); | |
4936 | if (idisk) { | |
4937 | if (idisk->owner == owner || | |
4938 | idisk->owner == IMSM_UNKNOWN_OWNER) | |
4939 | ok_count++; | |
4940 | else | |
1ade5cc1 N |
4941 | dprintf("'%.16s' owner %d != %d\n", |
4942 | disk->serial, idisk->owner, | |
a2b97981 DW |
4943 | owner); |
4944 | } else { | |
1ade5cc1 N |
4945 | dprintf("unknown disk %x [%d]: %.16s\n", |
4946 | __le32_to_cpu(mpb->family_num), i, | |
a2b97981 DW |
4947 | disk->serial); |
4948 | break; | |
4949 | } | |
4950 | } | |
4951 | ||
4952 | if (ok_count == mpb->num_disks) | |
4953 | return super; | |
4954 | return NULL; | |
4955 | } | |
4956 | ||
4957 | static void show_conflicts(__u32 family_num, struct intel_super *super_list) | |
4958 | { | |
4959 | struct intel_super *s; | |
4960 | ||
4961 | for (s = super_list; s; s = s->next) { | |
4962 | if (family_num != s->anchor->family_num) | |
4963 | continue; | |
e12b3daa | 4964 | pr_err("Conflict, offlining family %#x on '%s'\n", |
a2b97981 DW |
4965 | __le32_to_cpu(family_num), s->disks->devname); |
4966 | } | |
4967 | } | |
4968 | ||
4969 | static struct intel_super * | |
4970 | imsm_thunderdome(struct intel_super **super_list, int len) | |
4971 | { | |
4972 | struct intel_super *super_table[len]; | |
4973 | struct intel_disk *disk_list = NULL; | |
4974 | struct intel_super *champion, *spare; | |
4975 | struct intel_super *s, **del; | |
4976 | int tbl_size = 0; | |
4977 | int conflict; | |
4978 | int i; | |
4979 | ||
4980 | memset(super_table, 0, sizeof(super_table)); | |
4981 | for (s = *super_list; s; s = s->next) | |
4982 | tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list); | |
4983 | ||
4984 | for (i = 0; i < tbl_size; i++) { | |
4985 | struct imsm_disk *d; | |
4986 | struct intel_disk *idisk; | |
4987 | struct imsm_super *mpb = super_table[i]->anchor; | |
4988 | ||
4989 | s = super_table[i]; | |
4990 | d = &s->disks->disk; | |
4991 | ||
4992 | /* 'd' must appear in merged disk list for its | |
4993 | * configuration to be valid | |
4994 | */ | |
4995 | idisk = disk_list_get(d->serial, disk_list); | |
4996 | if (idisk && idisk->owner == i) | |
4997 | s = validate_members(s, disk_list, i); | |
4998 | else | |
4999 | s = NULL; | |
5000 | ||
5001 | if (!s) | |
1ade5cc1 N |
5002 | dprintf("marking family: %#x from %d:%d offline\n", |
5003 | mpb->family_num, | |
a2b97981 DW |
5004 | super_table[i]->disks->major, |
5005 | super_table[i]->disks->minor); | |
5006 | super_table[i] = s; | |
5007 | } | |
5008 | ||
5009 | /* This is where the mdadm implementation differs from the Windows | |
5010 | * driver which has no strict concept of a container. We can only | |
5011 | * assemble one family from a container, so when returning a prodigal | |
5012 | * array member to this system the code will not be able to disambiguate | |
5013 | * the container contents that should be assembled ("foreign" versus | |
5014 | * "local"). It requires user intervention to set the orig_family_num | |
5015 | * to a new value to establish a new container. The Windows driver in | |
5016 | * this situation fixes up the volume name in place and manages the | |
5017 | * foreign array as an independent entity. | |
5018 | */ | |
5019 | s = NULL; | |
5020 | spare = NULL; | |
5021 | conflict = 0; | |
5022 | for (i = 0; i < tbl_size; i++) { | |
5023 | struct intel_super *tbl_ent = super_table[i]; | |
5024 | int is_spare = 0; | |
5025 | ||
5026 | if (!tbl_ent) | |
5027 | continue; | |
5028 | ||
5029 | if (tbl_ent->anchor->num_raid_devs == 0) { | |
5030 | spare = tbl_ent; | |
5031 | is_spare = 1; | |
5032 | } | |
5033 | ||
5034 | if (s && !is_spare) { | |
5035 | show_conflicts(tbl_ent->anchor->family_num, *super_list); | |
5036 | conflict++; | |
5037 | } else if (!s && !is_spare) | |
5038 | s = tbl_ent; | |
5039 | } | |
5040 | ||
5041 | if (!s) | |
5042 | s = spare; | |
5043 | if (!s) { | |
5044 | champion = NULL; | |
5045 | goto out; | |
5046 | } | |
5047 | champion = s; | |
5048 | ||
5049 | if (conflict) | |
7a862a02 | 5050 | pr_err("Chose family %#x on '%s', assemble conflicts to new container with '--update=uuid'\n", |
a2b97981 DW |
5051 | __le32_to_cpu(s->anchor->family_num), s->disks->devname); |
5052 | ||
5053 | /* collect all dl's onto 'champion', and update them to | |
5054 | * champion's version of the status | |
5055 | */ | |
5056 | for (s = *super_list; s; s = s->next) { | |
5057 | struct imsm_super *mpb = champion->anchor; | |
5058 | struct dl *dl = s->disks; | |
5059 | ||
5060 | if (s == champion) | |
5061 | continue; | |
5062 | ||
5d7b407a CA |
5063 | mpb->attributes |= s->anchor->attributes & MPB_ATTRIB_2TB_DISK; |
5064 | ||
a2b97981 DW |
5065 | for (i = 0; i < mpb->num_disks; i++) { |
5066 | struct imsm_disk *disk; | |
5067 | ||
5068 | disk = __serial_to_disk(dl->serial, mpb, &dl->index); | |
5069 | if (disk) { | |
5070 | dl->disk = *disk; | |
5071 | /* only set index on disks that are a member of | |
5072 | * a populated contianer, i.e. one with | |
5073 | * raid_devs | |
5074 | */ | |
5075 | if (is_failed(&dl->disk)) | |
5076 | dl->index = -2; | |
5077 | else if (is_spare(&dl->disk)) | |
5078 | dl->index = -1; | |
5079 | break; | |
5080 | } | |
5081 | } | |
5082 | ||
5083 | if (i >= mpb->num_disks) { | |
5084 | struct intel_disk *idisk; | |
5085 | ||
5086 | idisk = disk_list_get(dl->serial, disk_list); | |
ecf408e9 | 5087 | if (idisk && is_spare(&idisk->disk) && |
a2b97981 DW |
5088 | !is_failed(&idisk->disk) && !is_configured(&idisk->disk)) |
5089 | dl->index = -1; | |
5090 | else { | |
5091 | dl->index = -2; | |
5092 | continue; | |
5093 | } | |
5094 | } | |
5095 | ||
5096 | dl->next = champion->disks; | |
5097 | champion->disks = dl; | |
5098 | s->disks = NULL; | |
5099 | } | |
5100 | ||
5101 | /* delete 'champion' from super_list */ | |
5102 | for (del = super_list; *del; ) { | |
5103 | if (*del == champion) { | |
5104 | *del = (*del)->next; | |
5105 | break; | |
5106 | } else | |
5107 | del = &(*del)->next; | |
5108 | } | |
5109 | champion->next = NULL; | |
5110 | ||
5111 | out: | |
5112 | while (disk_list) { | |
5113 | struct intel_disk *idisk = disk_list; | |
5114 | ||
5115 | disk_list = disk_list->next; | |
5116 | free(idisk); | |
5117 | } | |
5118 | ||
5119 | return champion; | |
5120 | } | |
5121 | ||
9587c373 LM |
5122 | static int |
5123 | get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd); | |
4dd2df09 | 5124 | static int get_super_block(struct intel_super **super_list, char *devnm, char *devname, |
9587c373 | 5125 | int major, int minor, int keep_fd); |
ec50f7b6 LM |
5126 | static int |
5127 | get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list, | |
5128 | int *max, int keep_fd); | |
5129 | ||
cdddbdbc | 5130 | static int load_super_imsm_all(struct supertype *st, int fd, void **sbp, |
ec50f7b6 LM |
5131 | char *devname, struct md_list *devlist, |
5132 | int keep_fd) | |
cdddbdbc | 5133 | { |
a2b97981 DW |
5134 | struct intel_super *super_list = NULL; |
5135 | struct intel_super *super = NULL; | |
a2b97981 | 5136 | int err = 0; |
9587c373 | 5137 | int i = 0; |
dab4a513 | 5138 | |
4389ce73 | 5139 | if (is_fd_valid(fd)) |
9587c373 LM |
5140 | /* 'fd' is an opened container */ |
5141 | err = get_sra_super_block(fd, &super_list, devname, &i, keep_fd); | |
5142 | else | |
ec50f7b6 LM |
5143 | /* get super block from devlist devices */ |
5144 | err = get_devlist_super_block(devlist, &super_list, &i, keep_fd); | |
9587c373 | 5145 | if (err) |
1602d52c | 5146 | goto error; |
a2b97981 DW |
5147 | /* all mpbs enter, maybe one leaves */ |
5148 | super = imsm_thunderdome(&super_list, i); | |
5149 | if (!super) { | |
5150 | err = 1; | |
5151 | goto error; | |
cdddbdbc DW |
5152 | } |
5153 | ||
47ee5a45 DW |
5154 | if (find_missing(super) != 0) { |
5155 | free_imsm(super); | |
a2b97981 DW |
5156 | err = 2; |
5157 | goto error; | |
47ee5a45 | 5158 | } |
8e59f3d8 AK |
5159 | |
5160 | /* load migration record */ | |
2f86fda3 | 5161 | err = load_imsm_migr_rec(super); |
4c965cc9 AK |
5162 | if (err == -1) { |
5163 | /* migration is in progress, | |
5164 | * but migr_rec cannot be loaded, | |
5165 | */ | |
8e59f3d8 AK |
5166 | err = 4; |
5167 | goto error; | |
5168 | } | |
e2f41b2c AK |
5169 | |
5170 | /* Check migration compatibility */ | |
089f9d79 | 5171 | if (err == 0 && check_mpb_migr_compatibility(super) != 0) { |
e7b84f9d | 5172 | pr_err("Unsupported migration detected"); |
e2f41b2c AK |
5173 | if (devname) |
5174 | fprintf(stderr, " on %s\n", devname); | |
5175 | else | |
5176 | fprintf(stderr, " (IMSM).\n"); | |
5177 | ||
5178 | err = 5; | |
5179 | goto error; | |
5180 | } | |
5181 | ||
a2b97981 DW |
5182 | err = 0; |
5183 | ||
5184 | error: | |
5185 | while (super_list) { | |
5186 | struct intel_super *s = super_list; | |
5187 | ||
5188 | super_list = super_list->next; | |
5189 | free_imsm(s); | |
5190 | } | |
9587c373 | 5191 | |
a2b97981 DW |
5192 | if (err) |
5193 | return err; | |
f7e7067b | 5194 | |
cdddbdbc | 5195 | *sbp = super; |
4389ce73 | 5196 | if (is_fd_valid(fd)) |
4dd2df09 | 5197 | strcpy(st->container_devnm, fd2devnm(fd)); |
9587c373 | 5198 | else |
4dd2df09 | 5199 | st->container_devnm[0] = 0; |
a2b97981 | 5200 | if (err == 0 && st->ss == NULL) { |
bf5a934a | 5201 | st->ss = &super_imsm; |
cdddbdbc DW |
5202 | st->minor_version = 0; |
5203 | st->max_devs = IMSM_MAX_DEVICES; | |
5204 | } | |
cdddbdbc DW |
5205 | return 0; |
5206 | } | |
2b959fbf | 5207 | |
ec50f7b6 LM |
5208 | static int |
5209 | get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list, | |
5210 | int *max, int keep_fd) | |
5211 | { | |
5212 | struct md_list *tmpdev; | |
5213 | int err = 0; | |
5214 | int i = 0; | |
9587c373 | 5215 | |
ec50f7b6 LM |
5216 | for (i = 0, tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { |
5217 | if (tmpdev->used != 1) | |
5218 | continue; | |
5219 | if (tmpdev->container == 1) { | |
ca9de185 | 5220 | int lmax = 0; |
ec50f7b6 | 5221 | int fd = dev_open(tmpdev->devname, O_RDONLY|O_EXCL); |
4389ce73 | 5222 | if (!is_fd_valid(fd)) { |
e7b84f9d | 5223 | pr_err("cannot open device %s: %s\n", |
ec50f7b6 LM |
5224 | tmpdev->devname, strerror(errno)); |
5225 | err = 8; | |
5226 | goto error; | |
5227 | } | |
5228 | err = get_sra_super_block(fd, super_list, | |
5229 | tmpdev->devname, &lmax, | |
5230 | keep_fd); | |
5231 | i += lmax; | |
5232 | close(fd); | |
5233 | if (err) { | |
5234 | err = 7; | |
5235 | goto error; | |
5236 | } | |
5237 | } else { | |
5238 | int major = major(tmpdev->st_rdev); | |
5239 | int minor = minor(tmpdev->st_rdev); | |
5240 | err = get_super_block(super_list, | |
4dd2df09 | 5241 | NULL, |
ec50f7b6 LM |
5242 | tmpdev->devname, |
5243 | major, minor, | |
5244 | keep_fd); | |
5245 | i++; | |
5246 | if (err) { | |
5247 | err = 6; | |
5248 | goto error; | |
5249 | } | |
5250 | } | |
5251 | } | |
5252 | error: | |
5253 | *max = i; | |
5254 | return err; | |
5255 | } | |
9587c373 | 5256 | |
4dd2df09 | 5257 | static int get_super_block(struct intel_super **super_list, char *devnm, char *devname, |
9587c373 LM |
5258 | int major, int minor, int keep_fd) |
5259 | { | |
594dc1b8 | 5260 | struct intel_super *s; |
9587c373 LM |
5261 | char nm[32]; |
5262 | int dfd = -1; | |
9587c373 LM |
5263 | int err = 0; |
5264 | int retry; | |
5265 | ||
5266 | s = alloc_super(); | |
5267 | if (!s) { | |
5268 | err = 1; | |
5269 | goto error; | |
5270 | } | |
5271 | ||
5272 | sprintf(nm, "%d:%d", major, minor); | |
5273 | dfd = dev_open(nm, O_RDWR); | |
4389ce73 | 5274 | if (!is_fd_valid(dfd)) { |
9587c373 LM |
5275 | err = 2; |
5276 | goto error; | |
5277 | } | |
5278 | ||
aec01630 JS |
5279 | if (!get_dev_sector_size(dfd, NULL, &s->sector_size)) { |
5280 | err = 2; | |
5281 | goto error; | |
5282 | } | |
cb8f6859 | 5283 | find_intel_hba_capability(dfd, s, devname); |
9587c373 LM |
5284 | err = load_and_parse_mpb(dfd, s, NULL, keep_fd); |
5285 | ||
5286 | /* retry the load if we might have raced against mdmon */ | |
4dd2df09 | 5287 | if (err == 3 && devnm && mdmon_running(devnm)) |
9587c373 | 5288 | for (retry = 0; retry < 3; retry++) { |
239b3cc0 | 5289 | sleep_for(0, MSEC_TO_NSEC(3), true); |
9587c373 LM |
5290 | err = load_and_parse_mpb(dfd, s, NULL, keep_fd); |
5291 | if (err != 3) | |
5292 | break; | |
5293 | } | |
5294 | error: | |
5295 | if (!err) { | |
5296 | s->next = *super_list; | |
5297 | *super_list = s; | |
5298 | } else { | |
5299 | if (s) | |
8d67477f | 5300 | free_imsm(s); |
4389ce73 | 5301 | close_fd(&dfd); |
9587c373 | 5302 | } |
4389ce73 MT |
5303 | if (!keep_fd) |
5304 | close_fd(&dfd); | |
9587c373 LM |
5305 | return err; |
5306 | ||
5307 | } | |
5308 | ||
5309 | static int | |
5310 | get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd) | |
5311 | { | |
5312 | struct mdinfo *sra; | |
4dd2df09 | 5313 | char *devnm; |
9587c373 LM |
5314 | struct mdinfo *sd; |
5315 | int err = 0; | |
5316 | int i = 0; | |
4dd2df09 | 5317 | sra = sysfs_read(fd, NULL, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE); |
9587c373 LM |
5318 | if (!sra) |
5319 | return 1; | |
5320 | ||
5321 | if (sra->array.major_version != -1 || | |
5322 | sra->array.minor_version != -2 || | |
5323 | strcmp(sra->text_version, "imsm") != 0) { | |
5324 | err = 1; | |
5325 | goto error; | |
5326 | } | |
5327 | /* load all mpbs */ | |
4dd2df09 | 5328 | devnm = fd2devnm(fd); |
9587c373 | 5329 | for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) { |
4dd2df09 | 5330 | if (get_super_block(super_list, devnm, devname, |
9587c373 LM |
5331 | sd->disk.major, sd->disk.minor, keep_fd) != 0) { |
5332 | err = 7; | |
5333 | goto error; | |
5334 | } | |
5335 | } | |
5336 | error: | |
5337 | sysfs_free(sra); | |
5338 | *max = i; | |
5339 | return err; | |
5340 | } | |
5341 | ||
2b959fbf N |
5342 | static int load_container_imsm(struct supertype *st, int fd, char *devname) |
5343 | { | |
ec50f7b6 | 5344 | return load_super_imsm_all(st, fd, &st->sb, devname, NULL, 1); |
2b959fbf | 5345 | } |
cdddbdbc DW |
5346 | |
5347 | static int load_super_imsm(struct supertype *st, int fd, char *devname) | |
5348 | { | |
5349 | struct intel_super *super; | |
5350 | int rv; | |
8a3544f8 | 5351 | int retry; |
cdddbdbc | 5352 | |
357ac106 | 5353 | if (test_partition(fd)) |
691c6ee1 N |
5354 | /* IMSM not allowed on partitions */ |
5355 | return 1; | |
5356 | ||
37424f13 DW |
5357 | free_super_imsm(st); |
5358 | ||
49133e57 | 5359 | super = alloc_super(); |
8d67477f TM |
5360 | if (!super) |
5361 | return 1; | |
3a85bf0e MG |
5362 | |
5363 | if (!get_dev_sector_size(fd, NULL, &super->sector_size)) { | |
5364 | free_imsm(super); | |
5365 | return 1; | |
5366 | } | |
ea2bc72b LM |
5367 | /* Load hba and capabilities if they exist. |
5368 | * But do not preclude loading metadata in case capabilities or hba are | |
5369 | * non-compliant and ignore_hw_compat is set. | |
5370 | */ | |
d424212e | 5371 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 5372 | /* no orom/efi or non-intel hba of the disk */ |
089f9d79 | 5373 | if (rv != 0 && st->ignore_hw_compat == 0) { |
f2f5c343 | 5374 | if (devname) |
e7b84f9d | 5375 | pr_err("No OROM/EFI properties for %s\n", devname); |
f2f5c343 LM |
5376 | free_imsm(super); |
5377 | return 2; | |
5378 | } | |
a2b97981 | 5379 | rv = load_and_parse_mpb(fd, super, devname, 0); |
cdddbdbc | 5380 | |
8a3544f8 AP |
5381 | /* retry the load if we might have raced against mdmon */ |
5382 | if (rv == 3) { | |
f96b1302 AP |
5383 | struct mdstat_ent *mdstat = NULL; |
5384 | char *name = fd2kname(fd); | |
5385 | ||
5386 | if (name) | |
5387 | mdstat = mdstat_by_component(name); | |
8a3544f8 AP |
5388 | |
5389 | if (mdstat && mdmon_running(mdstat->devnm) && getpid() != mdmon_pid(mdstat->devnm)) { | |
5390 | for (retry = 0; retry < 3; retry++) { | |
239b3cc0 | 5391 | sleep_for(0, MSEC_TO_NSEC(3), true); |
8a3544f8 AP |
5392 | rv = load_and_parse_mpb(fd, super, devname, 0); |
5393 | if (rv != 3) | |
5394 | break; | |
5395 | } | |
5396 | } | |
5397 | ||
5398 | free_mdstat(mdstat); | |
5399 | } | |
5400 | ||
cdddbdbc DW |
5401 | if (rv) { |
5402 | if (devname) | |
7a862a02 | 5403 | pr_err("Failed to load all information sections on %s\n", devname); |
cdddbdbc DW |
5404 | free_imsm(super); |
5405 | return rv; | |
5406 | } | |
5407 | ||
5408 | st->sb = super; | |
5409 | if (st->ss == NULL) { | |
5410 | st->ss = &super_imsm; | |
5411 | st->minor_version = 0; | |
5412 | st->max_devs = IMSM_MAX_DEVICES; | |
5413 | } | |
8e59f3d8 AK |
5414 | |
5415 | /* load migration record */ | |
2f86fda3 | 5416 | if (load_imsm_migr_rec(super) == 0) { |
2e062e82 AK |
5417 | /* Check for unsupported migration features */ |
5418 | if (check_mpb_migr_compatibility(super) != 0) { | |
e7b84f9d | 5419 | pr_err("Unsupported migration detected"); |
2e062e82 AK |
5420 | if (devname) |
5421 | fprintf(stderr, " on %s\n", devname); | |
5422 | else | |
5423 | fprintf(stderr, " (IMSM).\n"); | |
5424 | return 3; | |
5425 | } | |
e2f41b2c AK |
5426 | } |
5427 | ||
cdddbdbc DW |
5428 | return 0; |
5429 | } | |
5430 | ||
ef6ffade DW |
5431 | static __u16 info_to_blocks_per_strip(mdu_array_info_t *info) |
5432 | { | |
5433 | if (info->level == 1) | |
5434 | return 128; | |
5435 | return info->chunk_size >> 9; | |
5436 | } | |
5437 | ||
5551b113 CA |
5438 | static unsigned long long info_to_blocks_per_member(mdu_array_info_t *info, |
5439 | unsigned long long size) | |
fcfd9599 | 5440 | { |
4025c288 | 5441 | if (info->level == 1) |
5551b113 | 5442 | return size * 2; |
4025c288 | 5443 | else |
5551b113 | 5444 | return (size * 2) & ~(info_to_blocks_per_strip(info) - 1); |
fcfd9599 DW |
5445 | } |
5446 | ||
4d1313e9 DW |
5447 | static void imsm_update_version_info(struct intel_super *super) |
5448 | { | |
5449 | /* update the version and attributes */ | |
5450 | struct imsm_super *mpb = super->anchor; | |
5451 | char *version; | |
5452 | struct imsm_dev *dev; | |
5453 | struct imsm_map *map; | |
5454 | int i; | |
5455 | ||
5456 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
5457 | dev = get_imsm_dev(super, i); | |
238c0a71 | 5458 | map = get_imsm_map(dev, MAP_0); |
4d1313e9 DW |
5459 | if (__le32_to_cpu(dev->size_high) > 0) |
5460 | mpb->attributes |= MPB_ATTRIB_2TB; | |
5461 | ||
5462 | /* FIXME detect when an array spans a port multiplier */ | |
5463 | #if 0 | |
5464 | mpb->attributes |= MPB_ATTRIB_PM; | |
5465 | #endif | |
5466 | ||
5467 | if (mpb->num_raid_devs > 1 || | |
5468 | mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) { | |
5469 | version = MPB_VERSION_ATTRIBS; | |
5470 | switch (get_imsm_raid_level(map)) { | |
5471 | case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break; | |
5472 | case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break; | |
5473 | case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break; | |
5474 | case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break; | |
5475 | } | |
5476 | } else { | |
5477 | if (map->num_members >= 5) | |
5478 | version = MPB_VERSION_5OR6_DISK_ARRAY; | |
5479 | else if (dev->status == DEV_CLONE_N_GO) | |
5480 | version = MPB_VERSION_CNG; | |
5481 | else if (get_imsm_raid_level(map) == 5) | |
5482 | version = MPB_VERSION_RAID5; | |
5483 | else if (map->num_members >= 3) | |
5484 | version = MPB_VERSION_3OR4_DISK_ARRAY; | |
5485 | else if (get_imsm_raid_level(map) == 1) | |
5486 | version = MPB_VERSION_RAID1; | |
5487 | else | |
5488 | version = MPB_VERSION_RAID0; | |
5489 | } | |
5490 | strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version); | |
5491 | } | |
5492 | } | |
5493 | ||
e2eb503b MT |
5494 | /** |
5495 | * imsm_check_name() - check imsm naming criteria. | |
5496 | * @super: &intel_super pointer, not NULL. | |
5497 | * @name: name to check. | |
5498 | * @verbose: verbose level. | |
5499 | * | |
5500 | * Name must be no longer than &MAX_RAID_SERIAL_LEN and must be unique across volumes. | |
5501 | * | |
5502 | * Returns: &true if @name matches, &false otherwise. | |
5503 | */ | |
5504 | static bool imsm_is_name_allowed(struct intel_super *super, const char * const name, | |
5505 | const int verbose) | |
aa534678 DW |
5506 | { |
5507 | struct imsm_super *mpb = super->anchor; | |
aa534678 DW |
5508 | int i; |
5509 | ||
e2eb503b MT |
5510 | if (is_string_lq(name, MAX_RAID_SERIAL_LEN + 1) == false) { |
5511 | pr_vrb("imsm: Name \"%s\" is too long\n", name); | |
5512 | return false; | |
9bd99a90 RS |
5513 | } |
5514 | ||
aa534678 DW |
5515 | for (i = 0; i < mpb->num_raid_devs; i++) { |
5516 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
5517 | ||
5518 | if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) { | |
e2eb503b MT |
5519 | pr_vrb("imsm: Name \"%s\" already exists\n", name); |
5520 | return false; | |
aa534678 DW |
5521 | } |
5522 | } | |
5523 | ||
e2eb503b | 5524 | return true; |
aa534678 DW |
5525 | } |
5526 | ||
8b353278 | 5527 | static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info, |
5308f117 | 5528 | struct shape *s, char *name, |
83cd1e97 N |
5529 | char *homehost, int *uuid, |
5530 | long long data_offset) | |
cdddbdbc | 5531 | { |
c2c087e6 DW |
5532 | /* We are creating a volume inside a pre-existing container. |
5533 | * so st->sb is already set. | |
5534 | */ | |
5535 | struct intel_super *super = st->sb; | |
f36a9ecd | 5536 | unsigned int sector_size = super->sector_size; |
949c47a0 | 5537 | struct imsm_super *mpb = super->anchor; |
ba2de7ba | 5538 | struct intel_dev *dv; |
c2c087e6 DW |
5539 | struct imsm_dev *dev; |
5540 | struct imsm_vol *vol; | |
5541 | struct imsm_map *map; | |
5542 | int idx = mpb->num_raid_devs; | |
5543 | int i; | |
760365f9 | 5544 | int namelen; |
c2c087e6 | 5545 | unsigned long long array_blocks; |
2c092cad | 5546 | size_t size_old, size_new; |
b53bfba6 TM |
5547 | unsigned int data_disks; |
5548 | unsigned long long size_per_member; | |
cdddbdbc | 5549 | |
88c32bb1 | 5550 | if (super->orom && mpb->num_raid_devs >= super->orom->vpa) { |
7a862a02 | 5551 | pr_err("This imsm-container already has the maximum of %d volumes\n", super->orom->vpa); |
c2c087e6 DW |
5552 | return 0; |
5553 | } | |
5554 | ||
2c092cad DW |
5555 | /* ensure the mpb is large enough for the new data */ |
5556 | size_old = __le32_to_cpu(mpb->mpb_size); | |
5557 | size_new = disks_to_mpb_size(info->nr_disks); | |
5558 | if (size_new > size_old) { | |
5559 | void *mpb_new; | |
f36a9ecd | 5560 | size_t size_round = ROUND_UP(size_new, sector_size); |
2c092cad | 5561 | |
f36a9ecd | 5562 | if (posix_memalign(&mpb_new, sector_size, size_round) != 0) { |
e7b84f9d | 5563 | pr_err("could not allocate new mpb\n"); |
2c092cad DW |
5564 | return 0; |
5565 | } | |
85337573 AO |
5566 | if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE, |
5567 | MIGR_REC_BUF_SECTORS* | |
5568 | MAX_SECTOR_SIZE) != 0) { | |
1ade5cc1 | 5569 | pr_err("could not allocate migr_rec buffer\n"); |
8e59f3d8 AK |
5570 | free(super->buf); |
5571 | free(super); | |
ea944c8f | 5572 | free(mpb_new); |
8e59f3d8 AK |
5573 | return 0; |
5574 | } | |
2c092cad DW |
5575 | memcpy(mpb_new, mpb, size_old); |
5576 | free(mpb); | |
5577 | mpb = mpb_new; | |
949c47a0 | 5578 | super->anchor = mpb_new; |
2c092cad DW |
5579 | mpb->mpb_size = __cpu_to_le32(size_new); |
5580 | memset(mpb_new + size_old, 0, size_round - size_old); | |
bbab0940 | 5581 | super->len = size_round; |
2c092cad | 5582 | } |
bf5a934a | 5583 | super->current_vol = idx; |
3960e579 DW |
5584 | |
5585 | /* handle 'failed_disks' by either: | |
5586 | * a) create dummy disk entries in the table if this the first | |
5587 | * volume in the array. We add them here as this is the only | |
5588 | * opportunity to add them. add_to_super_imsm_volume() | |
5589 | * handles the non-failed disks and continues incrementing | |
5590 | * mpb->num_disks. | |
5591 | * b) validate that 'failed_disks' matches the current number | |
5592 | * of missing disks if the container is populated | |
d23fe947 | 5593 | */ |
3960e579 | 5594 | if (super->current_vol == 0) { |
d23fe947 | 5595 | mpb->num_disks = 0; |
3960e579 DW |
5596 | for (i = 0; i < info->failed_disks; i++) { |
5597 | struct imsm_disk *disk; | |
5598 | ||
5599 | mpb->num_disks++; | |
5600 | disk = __get_imsm_disk(mpb, i); | |
5601 | disk->status = CONFIGURED_DISK | FAILED_DISK; | |
5602 | disk->scsi_id = __cpu_to_le32(~(__u32)0); | |
5603 | snprintf((char *) disk->serial, MAX_RAID_SERIAL_LEN, | |
5b975129 | 5604 | "missing:%d", (__u8)i); |
3960e579 DW |
5605 | } |
5606 | find_missing(super); | |
5607 | } else { | |
5608 | int missing = 0; | |
5609 | struct dl *d; | |
5610 | ||
5611 | for (d = super->missing; d; d = d->next) | |
5612 | missing++; | |
5613 | if (info->failed_disks > missing) { | |
e7b84f9d | 5614 | pr_err("unable to add 'missing' disk to container\n"); |
3960e579 DW |
5615 | return 0; |
5616 | } | |
5617 | } | |
5a038140 | 5618 | |
e2eb503b | 5619 | if (imsm_is_name_allowed(super, name, 1) == false) |
aa534678 | 5620 | return 0; |
e2eb503b | 5621 | |
503975b9 N |
5622 | dv = xmalloc(sizeof(*dv)); |
5623 | dev = xcalloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1)); | |
760365f9 JS |
5624 | /* |
5625 | * Explicitly allow truncating to not confuse gcc's | |
5626 | * -Werror=stringop-truncation | |
5627 | */ | |
5628 | namelen = min((int) strlen(name), MAX_RAID_SERIAL_LEN); | |
5629 | memcpy(dev->volume, name, namelen); | |
e03640bd | 5630 | array_blocks = calc_array_size(info->level, info->raid_disks, |
03bcbc65 | 5631 | info->layout, info->chunk_size, |
b53bfba6 TM |
5632 | s->size * BLOCKS_PER_KB); |
5633 | data_disks = get_data_disks(info->level, info->layout, | |
5634 | info->raid_disks); | |
5635 | array_blocks = round_size_to_mb(array_blocks, data_disks); | |
5636 | size_per_member = array_blocks / data_disks; | |
979d38be | 5637 | |
fcc2c9da | 5638 | set_imsm_dev_size(dev, array_blocks); |
1a2487c2 | 5639 | dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING); |
c2c087e6 DW |
5640 | vol = &dev->vol; |
5641 | vol->migr_state = 0; | |
1484e727 | 5642 | set_migr_type(dev, MIGR_INIT); |
3960e579 | 5643 | vol->dirty = !info->state; |
4036e7ee | 5644 | set_vol_curr_migr_unit(dev, 0); |
238c0a71 | 5645 | map = get_imsm_map(dev, MAP_0); |
5551b113 | 5646 | set_pba_of_lba0(map, super->create_offset); |
ef6ffade | 5647 | map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info)); |
0556e1a2 | 5648 | map->failed_disk_num = ~0; |
bf4442ab | 5649 | if (info->level > 0) |
fffaf1ff N |
5650 | map->map_state = (info->state ? IMSM_T_STATE_NORMAL |
5651 | : IMSM_T_STATE_UNINITIALIZED); | |
bf4442ab AK |
5652 | else |
5653 | map->map_state = info->failed_disks ? IMSM_T_STATE_FAILED : | |
5654 | IMSM_T_STATE_NORMAL; | |
252d23c0 | 5655 | map->ddf = 1; |
ef6ffade DW |
5656 | |
5657 | if (info->level == 1 && info->raid_disks > 2) { | |
38950822 AW |
5658 | free(dev); |
5659 | free(dv); | |
7a862a02 | 5660 | pr_err("imsm does not support more than 2 disksin a raid1 volume\n"); |
ef6ffade DW |
5661 | return 0; |
5662 | } | |
81062a36 DW |
5663 | |
5664 | map->raid_level = info->level; | |
1c275381 | 5665 | if (info->level == 10) |
c2c087e6 | 5666 | map->raid_level = 1; |
1c275381 | 5667 | set_num_domains(map); |
ef6ffade | 5668 | |
44490938 MD |
5669 | size_per_member += NUM_BLOCKS_DIRTY_STRIPE_REGION; |
5670 | set_blocks_per_member(map, info_to_blocks_per_member(info, | |
5671 | size_per_member / | |
5672 | BLOCKS_PER_KB)); | |
5673 | ||
c2c087e6 | 5674 | map->num_members = info->raid_disks; |
1c275381 | 5675 | update_num_data_stripes(map, array_blocks); |
c2c087e6 DW |
5676 | for (i = 0; i < map->num_members; i++) { |
5677 | /* initialized in add_to_super */ | |
4eb26970 | 5678 | set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD); |
c2c087e6 | 5679 | } |
949c47a0 | 5680 | mpb->num_raid_devs++; |
2a24dc1b PB |
5681 | mpb->num_raid_devs_created++; |
5682 | dev->my_vol_raid_dev_num = mpb->num_raid_devs_created; | |
ba2de7ba | 5683 | |
b7580566 | 5684 | if (s->consistency_policy <= CONSISTENCY_POLICY_RESYNC) { |
c2462068 | 5685 | dev->rwh_policy = RWH_MULTIPLE_OFF; |
2432ce9b | 5686 | } else if (s->consistency_policy == CONSISTENCY_POLICY_PPL) { |
c2462068 | 5687 | dev->rwh_policy = RWH_MULTIPLE_DISTRIBUTED; |
2432ce9b AP |
5688 | } else { |
5689 | free(dev); | |
5690 | free(dv); | |
5691 | pr_err("imsm does not support consistency policy %s\n", | |
5f21d674 | 5692 | map_num_s(consistency_policies, s->consistency_policy)); |
2432ce9b AP |
5693 | return 0; |
5694 | } | |
5695 | ||
ba2de7ba DW |
5696 | dv->dev = dev; |
5697 | dv->index = super->current_vol; | |
5698 | dv->next = super->devlist; | |
5699 | super->devlist = dv; | |
c2c087e6 | 5700 | |
4d1313e9 DW |
5701 | imsm_update_version_info(super); |
5702 | ||
c2c087e6 | 5703 | return 1; |
cdddbdbc DW |
5704 | } |
5705 | ||
bf5a934a | 5706 | static int init_super_imsm(struct supertype *st, mdu_array_info_t *info, |
5308f117 | 5707 | struct shape *s, char *name, |
83cd1e97 N |
5708 | char *homehost, int *uuid, |
5709 | unsigned long long data_offset) | |
bf5a934a DW |
5710 | { |
5711 | /* This is primarily called by Create when creating a new array. | |
5712 | * We will then get add_to_super called for each component, and then | |
5713 | * write_init_super called to write it out to each device. | |
5714 | * For IMSM, Create can create on fresh devices or on a pre-existing | |
5715 | * array. | |
5716 | * To create on a pre-existing array a different method will be called. | |
5717 | * This one is just for fresh drives. | |
5718 | */ | |
5719 | struct intel_super *super; | |
5720 | struct imsm_super *mpb; | |
5721 | size_t mpb_size; | |
4d1313e9 | 5722 | char *version; |
bf5a934a | 5723 | |
83cd1e97 | 5724 | if (data_offset != INVALID_SECTORS) { |
ed503f89 | 5725 | pr_err("data-offset not supported by imsm\n"); |
83cd1e97 N |
5726 | return 0; |
5727 | } | |
5728 | ||
bf5a934a | 5729 | if (st->sb) |
5308f117 | 5730 | return init_super_imsm_volume(st, info, s, name, homehost, uuid, |
83cd1e97 | 5731 | data_offset); |
e683ca88 DW |
5732 | |
5733 | if (info) | |
5734 | mpb_size = disks_to_mpb_size(info->nr_disks); | |
5735 | else | |
f36a9ecd | 5736 | mpb_size = MAX_SECTOR_SIZE; |
bf5a934a | 5737 | |
49133e57 | 5738 | super = alloc_super(); |
f36a9ecd PB |
5739 | if (super && |
5740 | posix_memalign(&super->buf, MAX_SECTOR_SIZE, mpb_size) != 0) { | |
8d67477f | 5741 | free_imsm(super); |
e683ca88 DW |
5742 | super = NULL; |
5743 | } | |
5744 | if (!super) { | |
1ade5cc1 | 5745 | pr_err("could not allocate superblock\n"); |
bf5a934a DW |
5746 | return 0; |
5747 | } | |
de44e46f PB |
5748 | if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE, |
5749 | MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) { | |
1ade5cc1 | 5750 | pr_err("could not allocate migr_rec buffer\n"); |
8e59f3d8 | 5751 | free(super->buf); |
8d67477f | 5752 | free_imsm(super); |
8e59f3d8 AK |
5753 | return 0; |
5754 | } | |
e683ca88 | 5755 | memset(super->buf, 0, mpb_size); |
ef649044 | 5756 | mpb = super->buf; |
e683ca88 DW |
5757 | mpb->mpb_size = __cpu_to_le32(mpb_size); |
5758 | st->sb = super; | |
5759 | ||
5760 | if (info == NULL) { | |
5761 | /* zeroing superblock */ | |
5762 | return 0; | |
5763 | } | |
bf5a934a | 5764 | |
4d1313e9 DW |
5765 | mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; |
5766 | ||
5767 | version = (char *) mpb->sig; | |
5768 | strcpy(version, MPB_SIGNATURE); | |
5769 | version += strlen(MPB_SIGNATURE); | |
5770 | strcpy(version, MPB_VERSION_RAID0); | |
bf5a934a | 5771 | |
bf5a934a DW |
5772 | return 1; |
5773 | } | |
5774 | ||
f2cc4f7d AO |
5775 | static int drive_validate_sector_size(struct intel_super *super, struct dl *dl) |
5776 | { | |
5777 | unsigned int member_sector_size; | |
5778 | ||
4389ce73 | 5779 | if (!is_fd_valid(dl->fd)) { |
f2cc4f7d AO |
5780 | pr_err("Invalid file descriptor for %s\n", dl->devname); |
5781 | return 0; | |
5782 | } | |
5783 | ||
5784 | if (!get_dev_sector_size(dl->fd, dl->devname, &member_sector_size)) | |
5785 | return 0; | |
5786 | if (member_sector_size != super->sector_size) | |
5787 | return 0; | |
5788 | return 1; | |
5789 | } | |
5790 | ||
f20c3968 | 5791 | static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk, |
bf5a934a DW |
5792 | int fd, char *devname) |
5793 | { | |
5794 | struct intel_super *super = st->sb; | |
d23fe947 | 5795 | struct imsm_super *mpb = super->anchor; |
3960e579 | 5796 | struct imsm_disk *_disk; |
bf5a934a DW |
5797 | struct imsm_dev *dev; |
5798 | struct imsm_map *map; | |
3960e579 | 5799 | struct dl *dl, *df; |
4eb26970 | 5800 | int slot; |
9a7df595 MT |
5801 | int autolayout = 0; |
5802 | ||
5803 | if (!is_fd_valid(fd)) | |
5804 | autolayout = 1; | |
bf5a934a | 5805 | |
949c47a0 | 5806 | dev = get_imsm_dev(super, super->current_vol); |
238c0a71 | 5807 | map = get_imsm_map(dev, MAP_0); |
bf5a934a | 5808 | |
208933a7 | 5809 | if (! (dk->state & (1<<MD_DISK_SYNC))) { |
e7b84f9d | 5810 | pr_err("%s: Cannot add spare devices to IMSM volume\n", |
208933a7 N |
5811 | devname); |
5812 | return 1; | |
5813 | } | |
5814 | ||
9a7df595 MT |
5815 | for (dl = super->disks; dl ; dl = dl->next) { |
5816 | if (autolayout) { | |
efb30e7f DW |
5817 | if (dl->raiddisk == dk->raid_disk) |
5818 | break; | |
9a7df595 MT |
5819 | } else if (dl->major == dk->major && dl->minor == dk->minor) |
5820 | break; | |
efb30e7f | 5821 | } |
d23fe947 | 5822 | |
208933a7 | 5823 | if (!dl) { |
9a7df595 MT |
5824 | if (!autolayout) |
5825 | pr_err("%s is not a member of the same container.\n", | |
5826 | devname); | |
f20c3968 | 5827 | return 1; |
208933a7 | 5828 | } |
bf5a934a | 5829 | |
9a7df595 MT |
5830 | if (!autolayout && super->current_vol > 0) { |
5831 | int _slot = get_disk_slot_in_dev(super, 0, dl->index); | |
5832 | ||
5833 | if (_slot != dk->raid_disk) { | |
5834 | pr_err("Member %s is in %d slot for the first volume, but is in %d slot for a new volume.\n", | |
5835 | dl->devname, _slot, dk->raid_disk); | |
5836 | pr_err("Raid members are in different order than for the first volume, aborting.\n"); | |
5837 | return 1; | |
5838 | } | |
5839 | } | |
5840 | ||
59632db9 MZ |
5841 | if (mpb->num_disks == 0) |
5842 | if (!get_dev_sector_size(dl->fd, dl->devname, | |
5843 | &super->sector_size)) | |
5844 | return 1; | |
5845 | ||
f2cc4f7d AO |
5846 | if (!drive_validate_sector_size(super, dl)) { |
5847 | pr_err("Combining drives of different sector size in one volume is not allowed\n"); | |
5848 | return 1; | |
5849 | } | |
5850 | ||
d23fe947 DW |
5851 | /* add a pristine spare to the metadata */ |
5852 | if (dl->index < 0) { | |
5853 | dl->index = super->anchor->num_disks; | |
5854 | super->anchor->num_disks++; | |
5855 | } | |
4eb26970 DW |
5856 | /* Check the device has not already been added */ |
5857 | slot = get_imsm_disk_slot(map, dl->index); | |
5858 | if (slot >= 0 && | |
238c0a71 | 5859 | (get_imsm_ord_tbl_ent(dev, slot, MAP_X) & IMSM_ORD_REBUILD) == 0) { |
e7b84f9d | 5860 | pr_err("%s has been included in this array twice\n", |
4eb26970 DW |
5861 | devname); |
5862 | return 1; | |
5863 | } | |
656b6b5a | 5864 | set_imsm_ord_tbl_ent(map, dk->raid_disk, dl->index); |
ee5aad5a | 5865 | dl->disk.status = CONFIGURED_DISK; |
d23fe947 | 5866 | |
3960e579 DW |
5867 | /* update size of 'missing' disks to be at least as large as the |
5868 | * largest acitve member (we only have dummy missing disks when | |
5869 | * creating the first volume) | |
5870 | */ | |
5871 | if (super->current_vol == 0) { | |
5872 | for (df = super->missing; df; df = df->next) { | |
5551b113 CA |
5873 | if (total_blocks(&dl->disk) > total_blocks(&df->disk)) |
5874 | set_total_blocks(&df->disk, total_blocks(&dl->disk)); | |
3960e579 DW |
5875 | _disk = __get_imsm_disk(mpb, df->index); |
5876 | *_disk = df->disk; | |
5877 | } | |
5878 | } | |
5879 | ||
5880 | /* refresh unset/failed slots to point to valid 'missing' entries */ | |
5881 | for (df = super->missing; df; df = df->next) | |
5882 | for (slot = 0; slot < mpb->num_disks; slot++) { | |
238c0a71 | 5883 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); |
3960e579 DW |
5884 | |
5885 | if ((ord & IMSM_ORD_REBUILD) == 0) | |
5886 | continue; | |
5887 | set_imsm_ord_tbl_ent(map, slot, df->index | IMSM_ORD_REBUILD); | |
1ace8403 | 5888 | if (is_gen_migration(dev)) { |
238c0a71 AK |
5889 | struct imsm_map *map2 = get_imsm_map(dev, |
5890 | MAP_1); | |
0a108d63 | 5891 | int slot2 = get_imsm_disk_slot(map2, df->index); |
089f9d79 | 5892 | if (slot2 < map2->num_members && slot2 >= 0) { |
1ace8403 | 5893 | __u32 ord2 = get_imsm_ord_tbl_ent(dev, |
238c0a71 AK |
5894 | slot2, |
5895 | MAP_1); | |
1ace8403 AK |
5896 | if ((unsigned)df->index == |
5897 | ord_to_idx(ord2)) | |
5898 | set_imsm_ord_tbl_ent(map2, | |
0a108d63 | 5899 | slot2, |
1ace8403 AK |
5900 | df->index | |
5901 | IMSM_ORD_REBUILD); | |
5902 | } | |
5903 | } | |
3960e579 DW |
5904 | dprintf("set slot:%d to missing disk:%d\n", slot, df->index); |
5905 | break; | |
5906 | } | |
5907 | ||
d23fe947 DW |
5908 | /* if we are creating the first raid device update the family number */ |
5909 | if (super->current_vol == 0) { | |
5910 | __u32 sum; | |
5911 | struct imsm_dev *_dev = __get_imsm_dev(mpb, 0); | |
d23fe947 | 5912 | |
3960e579 | 5913 | _disk = __get_imsm_disk(mpb, dl->index); |
756a15f3 | 5914 | if (!_disk) { |
e7b84f9d | 5915 | pr_err("BUG mpb setup error\n"); |
791b666a AW |
5916 | return 1; |
5917 | } | |
d23fe947 DW |
5918 | *_dev = *dev; |
5919 | *_disk = dl->disk; | |
148acb7b DW |
5920 | sum = random32(); |
5921 | sum += __gen_imsm_checksum(mpb); | |
d23fe947 | 5922 | mpb->family_num = __cpu_to_le32(sum); |
148acb7b | 5923 | mpb->orig_family_num = mpb->family_num; |
e48aed3c | 5924 | mpb->creation_time = __cpu_to_le64((__u64)time(NULL)); |
d23fe947 | 5925 | } |
ca0748fa | 5926 | super->current_disk = dl; |
f20c3968 | 5927 | return 0; |
bf5a934a DW |
5928 | } |
5929 | ||
a8619d23 AK |
5930 | /* mark_spare() |
5931 | * Function marks disk as spare and restores disk serial | |
5932 | * in case it was previously marked as failed by takeover operation | |
5933 | * reruns: | |
5934 | * -1 : critical error | |
5935 | * 0 : disk is marked as spare but serial is not set | |
5936 | * 1 : success | |
5937 | */ | |
5938 | int mark_spare(struct dl *disk) | |
5939 | { | |
5940 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
5941 | int ret_val = -1; | |
5942 | ||
5943 | if (!disk) | |
5944 | return ret_val; | |
5945 | ||
5946 | ret_val = 0; | |
6da53c0e | 5947 | if (!imsm_read_serial(disk->fd, NULL, serial, MAX_RAID_SERIAL_LEN)) { |
a8619d23 AK |
5948 | /* Restore disk serial number, because takeover marks disk |
5949 | * as failed and adds to serial ':0' before it becomes | |
5950 | * a spare disk. | |
5951 | */ | |
5952 | serialcpy(disk->serial, serial); | |
5953 | serialcpy(disk->disk.serial, serial); | |
5954 | ret_val = 1; | |
5955 | } | |
5956 | disk->disk.status = SPARE_DISK; | |
5957 | disk->index = -1; | |
5958 | ||
5959 | return ret_val; | |
5960 | } | |
88654014 | 5961 | |
12724c01 TM |
5962 | |
5963 | static int write_super_imsm_spare(struct intel_super *super, struct dl *d); | |
5964 | ||
f20c3968 | 5965 | static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk, |
72ca9bcf N |
5966 | int fd, char *devname, |
5967 | unsigned long long data_offset) | |
cdddbdbc | 5968 | { |
c2c087e6 | 5969 | struct intel_super *super = st->sb; |
c2c087e6 DW |
5970 | struct dl *dd; |
5971 | unsigned long long size; | |
fa7bb6f8 | 5972 | unsigned int member_sector_size; |
f2f27e63 | 5973 | __u32 id; |
c2c087e6 DW |
5974 | int rv; |
5975 | struct stat stb; | |
5976 | ||
88654014 LM |
5977 | /* If we are on an RAID enabled platform check that the disk is |
5978 | * attached to the raid controller. | |
5979 | * We do not need to test disks attachment for container based additions, | |
5980 | * they shall be already tested when container was created/assembled. | |
88c32bb1 | 5981 | */ |
d424212e | 5982 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 5983 | /* no orom/efi or non-intel hba of the disk */ |
f0f5a016 LM |
5984 | if (rv != 0) { |
5985 | dprintf("capability: %p fd: %d ret: %d\n", | |
5986 | super->orom, fd, rv); | |
5987 | return 1; | |
88c32bb1 DW |
5988 | } |
5989 | ||
f20c3968 DW |
5990 | if (super->current_vol >= 0) |
5991 | return add_to_super_imsm_volume(st, dk, fd, devname); | |
bf5a934a | 5992 | |
c2c087e6 | 5993 | fstat(fd, &stb); |
503975b9 | 5994 | dd = xcalloc(sizeof(*dd), 1); |
c2c087e6 DW |
5995 | dd->major = major(stb.st_rdev); |
5996 | dd->minor = minor(stb.st_rdev); | |
503975b9 | 5997 | dd->devname = devname ? xstrdup(devname) : NULL; |
c2c087e6 | 5998 | dd->fd = fd; |
689c9bf3 | 5999 | dd->e = NULL; |
1a64be56 | 6000 | dd->action = DISK_ADD; |
6da53c0e | 6001 | rv = imsm_read_serial(fd, devname, dd->serial, MAX_RAID_SERIAL_LEN); |
32ba9157 | 6002 | if (rv) { |
e7b84f9d | 6003 | pr_err("failed to retrieve scsi serial, aborting\n"); |
3a85bf0e | 6004 | __free_imsm_disk(dd, 0); |
0030e8d6 | 6005 | abort(); |
c2c087e6 | 6006 | } |
7c798f87 | 6007 | |
20bee0f8 PB |
6008 | if (super->hba && ((super->hba->type == SYS_DEV_NVME) || |
6009 | (super->hba->type == SYS_DEV_VMD))) { | |
6010 | int i; | |
7c798f87 MT |
6011 | char cntrl_path[PATH_MAX]; |
6012 | char *cntrl_name; | |
6013 | char pci_dev_path[PATH_MAX]; | |
20bee0f8 | 6014 | |
7c798f87 MT |
6015 | if (!diskfd_to_devpath(fd, 2, pci_dev_path) || |
6016 | !diskfd_to_devpath(fd, 1, cntrl_path)) { | |
8662f92d | 6017 | pr_err("failed to get dev paths, aborting\n"); |
3a85bf0e | 6018 | __free_imsm_disk(dd, 0); |
a8f3cfd5 MT |
6019 | return 1; |
6020 | } | |
6021 | ||
7c798f87 MT |
6022 | cntrl_name = basename(cntrl_path); |
6023 | if (is_multipath_nvme(fd)) | |
6024 | pr_err("%s controller supports Multi-Path I/O, Intel (R) VROC does not support multipathing\n", | |
6025 | cntrl_name); | |
6026 | ||
6027 | if (devpath_to_vendor(pci_dev_path) == 0x8086) { | |
20bee0f8 PB |
6028 | /* |
6029 | * If Intel's NVMe drive has serial ended with | |
6030 | * "-A","-B","-1" or "-2" it means that this is "x8" | |
6031 | * device (double drive on single PCIe card). | |
6032 | * User should be warned about potential data loss. | |
6033 | */ | |
6034 | for (i = MAX_RAID_SERIAL_LEN-1; i > 0; i--) { | |
6035 | /* Skip empty character at the end */ | |
6036 | if (dd->serial[i] == 0) | |
6037 | continue; | |
6038 | ||
6039 | if (((dd->serial[i] == 'A') || | |
6040 | (dd->serial[i] == 'B') || | |
6041 | (dd->serial[i] == '1') || | |
6042 | (dd->serial[i] == '2')) && | |
6043 | (dd->serial[i-1] == '-')) | |
6044 | pr_err("\tThe action you are about to take may put your data at risk.\n" | |
6045 | "\tPlease note that x8 devices may consist of two separate x4 devices " | |
6046 | "located on a single PCIe port.\n" | |
6047 | "\tRAID 0 is the only supported configuration for this type of x8 device.\n"); | |
6048 | break; | |
6049 | } | |
32716c51 PB |
6050 | } else if (super->hba->type == SYS_DEV_VMD && super->orom && |
6051 | !imsm_orom_has_tpv_support(super->orom)) { | |
6052 | pr_err("\tPlatform configuration does not support non-Intel NVMe drives.\n" | |
8b751247 | 6053 | "\tPlease refer to Intel(R) RSTe/VROC user guide.\n"); |
3a85bf0e | 6054 | __free_imsm_disk(dd, 0); |
32716c51 | 6055 | return 1; |
20bee0f8 PB |
6056 | } |
6057 | } | |
c2c087e6 | 6058 | |
c2c087e6 | 6059 | get_dev_size(fd, NULL, &size); |
3a85bf0e MG |
6060 | if (!get_dev_sector_size(fd, NULL, &member_sector_size)) { |
6061 | __free_imsm_disk(dd, 0); | |
aec01630 | 6062 | return 1; |
3a85bf0e | 6063 | } |
fa7bb6f8 PB |
6064 | |
6065 | if (super->sector_size == 0) { | |
6066 | /* this a first device, so sector_size is not set yet */ | |
6067 | super->sector_size = member_sector_size; | |
fa7bb6f8 PB |
6068 | } |
6069 | ||
71e5411e | 6070 | /* clear migr_rec when adding disk to container */ |
85337573 AO |
6071 | memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE); |
6072 | if (lseek64(fd, size - MIGR_REC_SECTOR_POSITION*member_sector_size, | |
de44e46f | 6073 | SEEK_SET) >= 0) { |
466070ad | 6074 | if ((unsigned int)write(fd, super->migr_rec_buf, |
85337573 AO |
6075 | MIGR_REC_BUF_SECTORS*member_sector_size) != |
6076 | MIGR_REC_BUF_SECTORS*member_sector_size) | |
71e5411e PB |
6077 | perror("Write migr_rec failed"); |
6078 | } | |
6079 | ||
c2c087e6 | 6080 | size /= 512; |
1f24f035 | 6081 | serialcpy(dd->disk.serial, dd->serial); |
5551b113 CA |
6082 | set_total_blocks(&dd->disk, size); |
6083 | if (__le32_to_cpu(dd->disk.total_blocks_hi) > 0) { | |
6084 | struct imsm_super *mpb = super->anchor; | |
6085 | mpb->attributes |= MPB_ATTRIB_2TB_DISK; | |
6086 | } | |
a8619d23 | 6087 | mark_spare(dd); |
c2c087e6 | 6088 | if (sysfs_disk_to_scsi_id(fd, &id) == 0) |
b9f594fe | 6089 | dd->disk.scsi_id = __cpu_to_le32(id); |
c2c087e6 | 6090 | else |
b9f594fe | 6091 | dd->disk.scsi_id = __cpu_to_le32(0); |
43dad3d6 DW |
6092 | |
6093 | if (st->update_tail) { | |
1a64be56 LM |
6094 | dd->next = super->disk_mgmt_list; |
6095 | super->disk_mgmt_list = dd; | |
43dad3d6 | 6096 | } else { |
12724c01 TM |
6097 | /* this is called outside of mdmon |
6098 | * write initial spare metadata | |
6099 | * mdmon will overwrite it. | |
6100 | */ | |
43dad3d6 DW |
6101 | dd->next = super->disks; |
6102 | super->disks = dd; | |
12724c01 | 6103 | write_super_imsm_spare(super, dd); |
43dad3d6 | 6104 | } |
f20c3968 DW |
6105 | |
6106 | return 0; | |
cdddbdbc DW |
6107 | } |
6108 | ||
1a64be56 LM |
6109 | static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk) |
6110 | { | |
6111 | struct intel_super *super = st->sb; | |
6112 | struct dl *dd; | |
6113 | ||
6114 | /* remove from super works only in mdmon - for communication | |
6115 | * manager - monitor. Check if communication memory buffer | |
6116 | * is prepared. | |
6117 | */ | |
6118 | if (!st->update_tail) { | |
1ade5cc1 | 6119 | pr_err("shall be used in mdmon context only\n"); |
1a64be56 LM |
6120 | return 1; |
6121 | } | |
503975b9 | 6122 | dd = xcalloc(1, sizeof(*dd)); |
1a64be56 LM |
6123 | dd->major = dk->major; |
6124 | dd->minor = dk->minor; | |
1a64be56 | 6125 | dd->fd = -1; |
a8619d23 | 6126 | mark_spare(dd); |
1a64be56 LM |
6127 | dd->action = DISK_REMOVE; |
6128 | ||
6129 | dd->next = super->disk_mgmt_list; | |
6130 | super->disk_mgmt_list = dd; | |
6131 | ||
1a64be56 LM |
6132 | return 0; |
6133 | } | |
6134 | ||
f796af5d DW |
6135 | static int store_imsm_mpb(int fd, struct imsm_super *mpb); |
6136 | ||
6137 | static union { | |
f36a9ecd | 6138 | char buf[MAX_SECTOR_SIZE]; |
f796af5d | 6139 | struct imsm_super anchor; |
f36a9ecd | 6140 | } spare_record __attribute__ ((aligned(MAX_SECTOR_SIZE))); |
c2c087e6 | 6141 | |
12724c01 TM |
6142 | |
6143 | static int write_super_imsm_spare(struct intel_super *super, struct dl *d) | |
d23fe947 | 6144 | { |
d23fe947 | 6145 | struct imsm_super *mpb = super->anchor; |
f796af5d | 6146 | struct imsm_super *spare = &spare_record.anchor; |
d23fe947 | 6147 | __u32 sum; |
12724c01 TM |
6148 | |
6149 | if (d->index != -1) | |
6150 | return 1; | |
d23fe947 | 6151 | |
68641cdb JS |
6152 | spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super)); |
6153 | spare->generation_num = __cpu_to_le32(1UL); | |
f796af5d | 6154 | spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; |
68641cdb JS |
6155 | spare->num_disks = 1; |
6156 | spare->num_raid_devs = 0; | |
6157 | spare->cache_size = mpb->cache_size; | |
6158 | spare->pwr_cycle_count = __cpu_to_le32(1); | |
f796af5d DW |
6159 | |
6160 | snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH, | |
6161 | MPB_SIGNATURE MPB_VERSION_RAID0); | |
d23fe947 | 6162 | |
12724c01 TM |
6163 | spare->disk[0] = d->disk; |
6164 | if (__le32_to_cpu(d->disk.total_blocks_hi) > 0) | |
6165 | spare->attributes |= MPB_ATTRIB_2TB_DISK; | |
6166 | ||
6167 | if (super->sector_size == 4096) | |
6168 | convert_to_4k_imsm_disk(&spare->disk[0]); | |
d23fe947 | 6169 | |
12724c01 TM |
6170 | sum = __gen_imsm_checksum(spare); |
6171 | spare->family_num = __cpu_to_le32(sum); | |
6172 | spare->orig_family_num = 0; | |
6173 | sum = __gen_imsm_checksum(spare); | |
6174 | spare->check_sum = __cpu_to_le32(sum); | |
027c374f | 6175 | |
12724c01 TM |
6176 | if (store_imsm_mpb(d->fd, spare)) { |
6177 | pr_err("failed for device %d:%d %s\n", | |
6178 | d->major, d->minor, strerror(errno)); | |
6179 | return 1; | |
6180 | } | |
6181 | ||
6182 | return 0; | |
6183 | } | |
6184 | /* spare records have their own family number and do not have any defined raid | |
6185 | * devices | |
6186 | */ | |
6187 | static int write_super_imsm_spares(struct intel_super *super, int doclose) | |
6188 | { | |
6189 | struct dl *d; | |
f36a9ecd | 6190 | |
12724c01 TM |
6191 | for (d = super->disks; d; d = d->next) { |
6192 | if (d->index != -1) | |
6193 | continue; | |
d23fe947 | 6194 | |
12724c01 | 6195 | if (write_super_imsm_spare(super, d)) |
e74255d9 | 6196 | return 1; |
12724c01 | 6197 | |
4389ce73 MT |
6198 | if (doclose) |
6199 | close_fd(&d->fd); | |
d23fe947 DW |
6200 | } |
6201 | ||
e74255d9 | 6202 | return 0; |
d23fe947 DW |
6203 | } |
6204 | ||
36988a3d | 6205 | static int write_super_imsm(struct supertype *st, int doclose) |
cdddbdbc | 6206 | { |
36988a3d | 6207 | struct intel_super *super = st->sb; |
f36a9ecd | 6208 | unsigned int sector_size = super->sector_size; |
949c47a0 | 6209 | struct imsm_super *mpb = super->anchor; |
c2c087e6 DW |
6210 | struct dl *d; |
6211 | __u32 generation; | |
6212 | __u32 sum; | |
d23fe947 | 6213 | int spares = 0; |
949c47a0 | 6214 | int i; |
a48ac0a8 | 6215 | __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk); |
36988a3d | 6216 | int num_disks = 0; |
146c6260 | 6217 | int clear_migration_record = 1; |
bbab0940 | 6218 | __u32 bbm_log_size; |
cdddbdbc | 6219 | |
c2c087e6 DW |
6220 | /* 'generation' is incremented everytime the metadata is written */ |
6221 | generation = __le32_to_cpu(mpb->generation_num); | |
6222 | generation++; | |
6223 | mpb->generation_num = __cpu_to_le32(generation); | |
6224 | ||
148acb7b DW |
6225 | /* fix up cases where previous mdadm releases failed to set |
6226 | * orig_family_num | |
6227 | */ | |
6228 | if (mpb->orig_family_num == 0) | |
6229 | mpb->orig_family_num = mpb->family_num; | |
6230 | ||
d23fe947 | 6231 | for (d = super->disks; d; d = d->next) { |
8796fdc4 | 6232 | if (d->index == -1) |
d23fe947 | 6233 | spares++; |
36988a3d | 6234 | else { |
d23fe947 | 6235 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
6236 | num_disks++; |
6237 | } | |
d23fe947 | 6238 | } |
36988a3d | 6239 | for (d = super->missing; d; d = d->next) { |
47ee5a45 | 6240 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
6241 | num_disks++; |
6242 | } | |
6243 | mpb->num_disks = num_disks; | |
6244 | mpb_size += sizeof(struct imsm_disk) * mpb->num_disks; | |
b9f594fe | 6245 | |
949c47a0 DW |
6246 | for (i = 0; i < mpb->num_raid_devs; i++) { |
6247 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
36988a3d | 6248 | struct imsm_dev *dev2 = get_imsm_dev(super, i); |
756a15f3 MG |
6249 | |
6250 | imsm_copy_dev(dev, dev2); | |
6251 | mpb_size += sizeof_imsm_dev(dev, 0); | |
6252 | ||
146c6260 AK |
6253 | if (is_gen_migration(dev2)) |
6254 | clear_migration_record = 0; | |
949c47a0 | 6255 | } |
bbab0940 TM |
6256 | |
6257 | bbm_log_size = get_imsm_bbm_log_size(super->bbm_log); | |
6258 | ||
6259 | if (bbm_log_size) { | |
6260 | memcpy((void *)mpb + mpb_size, super->bbm_log, bbm_log_size); | |
6261 | mpb->attributes |= MPB_ATTRIB_BBM; | |
6262 | } else | |
6263 | mpb->attributes &= ~MPB_ATTRIB_BBM; | |
6264 | ||
6265 | super->anchor->bbm_log_size = __cpu_to_le32(bbm_log_size); | |
6266 | mpb_size += bbm_log_size; | |
a48ac0a8 | 6267 | mpb->mpb_size = __cpu_to_le32(mpb_size); |
949c47a0 | 6268 | |
bbab0940 TM |
6269 | #ifdef DEBUG |
6270 | assert(super->len == 0 || mpb_size <= super->len); | |
6271 | #endif | |
6272 | ||
c2c087e6 | 6273 | /* recalculate checksum */ |
949c47a0 | 6274 | sum = __gen_imsm_checksum(mpb); |
c2c087e6 DW |
6275 | mpb->check_sum = __cpu_to_le32(sum); |
6276 | ||
51d83f5d AK |
6277 | if (super->clean_migration_record_by_mdmon) { |
6278 | clear_migration_record = 1; | |
6279 | super->clean_migration_record_by_mdmon = 0; | |
6280 | } | |
146c6260 | 6281 | if (clear_migration_record) |
de44e46f | 6282 | memset(super->migr_rec_buf, 0, |
85337573 | 6283 | MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE); |
146c6260 | 6284 | |
f36a9ecd PB |
6285 | if (sector_size == 4096) |
6286 | convert_to_4k(super); | |
6287 | ||
d23fe947 | 6288 | /* write the mpb for disks that compose raid devices */ |
c2c087e6 | 6289 | for (d = super->disks; d ; d = d->next) { |
86c54047 | 6290 | if (d->index < 0 || is_failed(&d->disk)) |
d23fe947 | 6291 | continue; |
30602f53 | 6292 | |
146c6260 AK |
6293 | if (clear_migration_record) { |
6294 | unsigned long long dsize; | |
6295 | ||
6296 | get_dev_size(d->fd, NULL, &dsize); | |
de44e46f PB |
6297 | if (lseek64(d->fd, dsize - sector_size, |
6298 | SEEK_SET) >= 0) { | |
466070ad PB |
6299 | if ((unsigned int)write(d->fd, |
6300 | super->migr_rec_buf, | |
de44e46f PB |
6301 | MIGR_REC_BUF_SECTORS*sector_size) != |
6302 | MIGR_REC_BUF_SECTORS*sector_size) | |
9e2d750d | 6303 | perror("Write migr_rec failed"); |
146c6260 AK |
6304 | } |
6305 | } | |
51d83f5d AK |
6306 | |
6307 | if (store_imsm_mpb(d->fd, mpb)) | |
6308 | fprintf(stderr, | |
1ade5cc1 N |
6309 | "failed for device %d:%d (fd: %d)%s\n", |
6310 | d->major, d->minor, | |
51d83f5d AK |
6311 | d->fd, strerror(errno)); |
6312 | ||
4389ce73 MT |
6313 | if (doclose) |
6314 | close_fd(&d->fd); | |
c2c087e6 DW |
6315 | } |
6316 | ||
d23fe947 DW |
6317 | if (spares) |
6318 | return write_super_imsm_spares(super, doclose); | |
6319 | ||
e74255d9 | 6320 | return 0; |
c2c087e6 DW |
6321 | } |
6322 | ||
9b1fb677 | 6323 | static int create_array(struct supertype *st, int dev_idx) |
43dad3d6 DW |
6324 | { |
6325 | size_t len; | |
6326 | struct imsm_update_create_array *u; | |
6327 | struct intel_super *super = st->sb; | |
9b1fb677 | 6328 | struct imsm_dev *dev = get_imsm_dev(super, dev_idx); |
238c0a71 | 6329 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
54c2c1ea DW |
6330 | struct disk_info *inf; |
6331 | struct imsm_disk *disk; | |
6332 | int i; | |
43dad3d6 | 6333 | |
54c2c1ea DW |
6334 | len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) + |
6335 | sizeof(*inf) * map->num_members; | |
503975b9 | 6336 | u = xmalloc(len); |
43dad3d6 | 6337 | u->type = update_create_array; |
9b1fb677 | 6338 | u->dev_idx = dev_idx; |
43dad3d6 | 6339 | imsm_copy_dev(&u->dev, dev); |
54c2c1ea DW |
6340 | inf = get_disk_info(u); |
6341 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 6342 | int idx = get_imsm_disk_idx(dev, i, MAP_X); |
9b1fb677 | 6343 | |
54c2c1ea | 6344 | disk = get_imsm_disk(super, idx); |
1ca5c8e0 N |
6345 | if (!disk) |
6346 | disk = get_imsm_missing(super, idx); | |
54c2c1ea DW |
6347 | serialcpy(inf[i].serial, disk->serial); |
6348 | } | |
43dad3d6 DW |
6349 | append_metadata_update(st, u, len); |
6350 | ||
6351 | return 0; | |
6352 | } | |
6353 | ||
1a64be56 | 6354 | static int mgmt_disk(struct supertype *st) |
43dad3d6 DW |
6355 | { |
6356 | struct intel_super *super = st->sb; | |
6357 | size_t len; | |
1a64be56 | 6358 | struct imsm_update_add_remove_disk *u; |
43dad3d6 | 6359 | |
1a64be56 | 6360 | if (!super->disk_mgmt_list) |
43dad3d6 DW |
6361 | return 0; |
6362 | ||
6363 | len = sizeof(*u); | |
503975b9 | 6364 | u = xmalloc(len); |
1a64be56 | 6365 | u->type = update_add_remove_disk; |
43dad3d6 DW |
6366 | append_metadata_update(st, u, len); |
6367 | ||
6368 | return 0; | |
6369 | } | |
2432ce9b AP |
6370 | |
6371 | __u32 crc32c_le(__u32 crc, unsigned char const *p, size_t len); | |
6372 | ||
e397cefe AP |
6373 | static int write_ppl_header(unsigned long long ppl_sector, int fd, void *buf) |
6374 | { | |
6375 | struct ppl_header *ppl_hdr = buf; | |
6376 | int ret; | |
6377 | ||
6378 | ppl_hdr->checksum = __cpu_to_le32(~crc32c_le(~0, buf, PPL_HEADER_SIZE)); | |
6379 | ||
6380 | if (lseek64(fd, ppl_sector * 512, SEEK_SET) < 0) { | |
6381 | ret = -errno; | |
6382 | perror("Failed to seek to PPL header location"); | |
6383 | return ret; | |
6384 | } | |
6385 | ||
6386 | if (write(fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) { | |
6387 | ret = -errno; | |
6388 | perror("Write PPL header failed"); | |
6389 | return ret; | |
6390 | } | |
6391 | ||
6392 | fsync(fd); | |
6393 | ||
6394 | return 0; | |
6395 | } | |
6396 | ||
2432ce9b AP |
6397 | static int write_init_ppl_imsm(struct supertype *st, struct mdinfo *info, int fd) |
6398 | { | |
6399 | struct intel_super *super = st->sb; | |
6400 | void *buf; | |
6401 | struct ppl_header *ppl_hdr; | |
6402 | int ret; | |
6403 | ||
b2514242 PB |
6404 | /* first clear entire ppl space */ |
6405 | ret = zero_disk_range(fd, info->ppl_sector, info->ppl_size); | |
6406 | if (ret) | |
6407 | return ret; | |
6408 | ||
6409 | ret = posix_memalign(&buf, MAX_SECTOR_SIZE, PPL_HEADER_SIZE); | |
2432ce9b AP |
6410 | if (ret) { |
6411 | pr_err("Failed to allocate PPL header buffer\n"); | |
e397cefe | 6412 | return -ret; |
2432ce9b AP |
6413 | } |
6414 | ||
6415 | memset(buf, 0, PPL_HEADER_SIZE); | |
6416 | ppl_hdr = buf; | |
6417 | memset(ppl_hdr->reserved, 0xff, PPL_HDR_RESERVED); | |
6418 | ppl_hdr->signature = __cpu_to_le32(super->anchor->orig_family_num); | |
b23d0750 AP |
6419 | |
6420 | if (info->mismatch_cnt) { | |
6421 | /* | |
6422 | * We are overwriting an invalid ppl. Make one entry with wrong | |
6423 | * checksum to prevent the kernel from skipping resync. | |
6424 | */ | |
6425 | ppl_hdr->entries_count = __cpu_to_le32(1); | |
6426 | ppl_hdr->entries[0].checksum = ~0; | |
6427 | } | |
6428 | ||
e397cefe | 6429 | ret = write_ppl_header(info->ppl_sector, fd, buf); |
2432ce9b AP |
6430 | |
6431 | free(buf); | |
6432 | return ret; | |
6433 | } | |
6434 | ||
e397cefe AP |
6435 | static int is_rebuilding(struct imsm_dev *dev); |
6436 | ||
2432ce9b AP |
6437 | static int validate_ppl_imsm(struct supertype *st, struct mdinfo *info, |
6438 | struct mdinfo *disk) | |
6439 | { | |
6440 | struct intel_super *super = st->sb; | |
6441 | struct dl *d; | |
e397cefe | 6442 | void *buf_orig, *buf, *buf_prev = NULL; |
2432ce9b | 6443 | int ret = 0; |
e397cefe | 6444 | struct ppl_header *ppl_hdr = NULL; |
2432ce9b AP |
6445 | __u32 crc; |
6446 | struct imsm_dev *dev; | |
2432ce9b | 6447 | __u32 idx; |
44b6b876 PB |
6448 | unsigned int i; |
6449 | unsigned long long ppl_offset = 0; | |
6450 | unsigned long long prev_gen_num = 0; | |
2432ce9b AP |
6451 | |
6452 | if (disk->disk.raid_disk < 0) | |
6453 | return 0; | |
6454 | ||
2432ce9b | 6455 | dev = get_imsm_dev(super, info->container_member); |
2fc0fc63 | 6456 | idx = get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_0); |
2432ce9b AP |
6457 | d = get_imsm_dl_disk(super, idx); |
6458 | ||
6459 | if (!d || d->index < 0 || is_failed(&d->disk)) | |
e397cefe AP |
6460 | return 0; |
6461 | ||
6462 | if (posix_memalign(&buf_orig, MAX_SECTOR_SIZE, PPL_HEADER_SIZE * 2)) { | |
6463 | pr_err("Failed to allocate PPL header buffer\n"); | |
6464 | return -1; | |
6465 | } | |
6466 | buf = buf_orig; | |
2432ce9b | 6467 | |
44b6b876 PB |
6468 | ret = 1; |
6469 | while (ppl_offset < MULTIPLE_PPL_AREA_SIZE_IMSM) { | |
e397cefe AP |
6470 | void *tmp; |
6471 | ||
44b6b876 | 6472 | dprintf("Checking potential PPL at offset: %llu\n", ppl_offset); |
2432ce9b | 6473 | |
44b6b876 PB |
6474 | if (lseek64(d->fd, info->ppl_sector * 512 + ppl_offset, |
6475 | SEEK_SET) < 0) { | |
6476 | perror("Failed to seek to PPL header location"); | |
6477 | ret = -1; | |
e397cefe | 6478 | break; |
44b6b876 | 6479 | } |
2432ce9b | 6480 | |
44b6b876 PB |
6481 | if (read(d->fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) { |
6482 | perror("Read PPL header failed"); | |
6483 | ret = -1; | |
e397cefe | 6484 | break; |
44b6b876 | 6485 | } |
2432ce9b | 6486 | |
44b6b876 | 6487 | ppl_hdr = buf; |
2432ce9b | 6488 | |
44b6b876 PB |
6489 | crc = __le32_to_cpu(ppl_hdr->checksum); |
6490 | ppl_hdr->checksum = 0; | |
6491 | ||
6492 | if (crc != ~crc32c_le(~0, buf, PPL_HEADER_SIZE)) { | |
6493 | dprintf("Wrong PPL header checksum on %s\n", | |
6494 | d->devname); | |
e397cefe | 6495 | break; |
44b6b876 PB |
6496 | } |
6497 | ||
6498 | if (prev_gen_num > __le64_to_cpu(ppl_hdr->generation)) { | |
6499 | /* previous was newest, it was already checked */ | |
e397cefe | 6500 | break; |
44b6b876 PB |
6501 | } |
6502 | ||
6503 | if ((__le32_to_cpu(ppl_hdr->signature) != | |
6504 | super->anchor->orig_family_num)) { | |
6505 | dprintf("Wrong PPL header signature on %s\n", | |
6506 | d->devname); | |
6507 | ret = 1; | |
e397cefe | 6508 | break; |
44b6b876 PB |
6509 | } |
6510 | ||
6511 | ret = 0; | |
6512 | prev_gen_num = __le64_to_cpu(ppl_hdr->generation); | |
2432ce9b | 6513 | |
44b6b876 PB |
6514 | ppl_offset += PPL_HEADER_SIZE; |
6515 | for (i = 0; i < __le32_to_cpu(ppl_hdr->entries_count); i++) | |
6516 | ppl_offset += | |
6517 | __le32_to_cpu(ppl_hdr->entries[i].pp_size); | |
e397cefe AP |
6518 | |
6519 | if (!buf_prev) | |
6520 | buf_prev = buf + PPL_HEADER_SIZE; | |
6521 | tmp = buf_prev; | |
6522 | buf_prev = buf; | |
6523 | buf = tmp; | |
2432ce9b AP |
6524 | } |
6525 | ||
e397cefe AP |
6526 | if (buf_prev) { |
6527 | buf = buf_prev; | |
6528 | ppl_hdr = buf_prev; | |
6529 | } | |
2432ce9b | 6530 | |
54148aba PB |
6531 | /* |
6532 | * Update metadata to use mutliple PPLs area (1MB). | |
6533 | * This is done once for all RAID members | |
6534 | */ | |
6535 | if (info->consistency_policy == CONSISTENCY_POLICY_PPL && | |
6536 | info->ppl_size != (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9)) { | |
6537 | char subarray[20]; | |
6538 | struct mdinfo *member_dev; | |
6539 | ||
6540 | sprintf(subarray, "%d", info->container_member); | |
6541 | ||
6542 | if (mdmon_running(st->container_devnm)) | |
6543 | st->update_tail = &st->updates; | |
6544 | ||
03312b52 | 6545 | if (st->ss->update_subarray(st, subarray, UOPT_PPL, NULL)) { |
54148aba PB |
6546 | pr_err("Failed to update subarray %s\n", |
6547 | subarray); | |
6548 | } else { | |
6549 | if (st->update_tail) | |
6550 | flush_metadata_updates(st); | |
6551 | else | |
6552 | st->ss->sync_metadata(st); | |
6553 | info->ppl_size = (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9); | |
6554 | for (member_dev = info->devs; member_dev; | |
6555 | member_dev = member_dev->next) | |
6556 | member_dev->ppl_size = | |
6557 | (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9); | |
6558 | } | |
6559 | } | |
6560 | ||
b23d0750 | 6561 | if (ret == 1) { |
2fc0fc63 AP |
6562 | struct imsm_map *map = get_imsm_map(dev, MAP_X); |
6563 | ||
50b9c10d PB |
6564 | if (map->map_state == IMSM_T_STATE_UNINITIALIZED || |
6565 | (map->map_state == IMSM_T_STATE_NORMAL && | |
2ec9d182 | 6566 | !(dev->vol.dirty & RAIDVOL_DIRTY)) || |
e397cefe | 6567 | (is_rebuilding(dev) && |
4036e7ee | 6568 | vol_curr_migr_unit(dev) == 0 && |
2ec9d182 | 6569 | get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_1) != idx)) |
b23d0750 AP |
6570 | ret = st->ss->write_init_ppl(st, info, d->fd); |
6571 | else | |
6572 | info->mismatch_cnt++; | |
e397cefe AP |
6573 | } else if (ret == 0 && |
6574 | ppl_hdr->entries_count == 0 && | |
6575 | is_rebuilding(dev) && | |
6576 | info->resync_start == 0) { | |
6577 | /* | |
6578 | * The header has no entries - add a single empty entry and | |
6579 | * rewrite the header to prevent the kernel from going into | |
6580 | * resync after an interrupted rebuild. | |
6581 | */ | |
6582 | ppl_hdr->entries_count = __cpu_to_le32(1); | |
6583 | ret = write_ppl_header(info->ppl_sector, d->fd, buf); | |
b23d0750 | 6584 | } |
2432ce9b | 6585 | |
e397cefe AP |
6586 | free(buf_orig); |
6587 | ||
2432ce9b AP |
6588 | return ret; |
6589 | } | |
6590 | ||
2432ce9b AP |
6591 | static int write_init_ppl_imsm_all(struct supertype *st, struct mdinfo *info) |
6592 | { | |
6593 | struct intel_super *super = st->sb; | |
6594 | struct dl *d; | |
6595 | int ret = 0; | |
6596 | ||
6597 | if (info->consistency_policy != CONSISTENCY_POLICY_PPL || | |
6598 | info->array.level != 5) | |
6599 | return 0; | |
6600 | ||
6601 | for (d = super->disks; d ; d = d->next) { | |
6602 | if (d->index < 0 || is_failed(&d->disk)) | |
6603 | continue; | |
6604 | ||
6605 | ret = st->ss->write_init_ppl(st, info, d->fd); | |
6606 | if (ret) | |
6607 | break; | |
6608 | } | |
6609 | ||
6610 | return ret; | |
6611 | } | |
43dad3d6 | 6612 | |
fbc42556 JR |
6613 | /******************************************************************************* |
6614 | * Function: write_init_bitmap_imsm_vol | |
6615 | * Description: Write a bitmap header and prepares the area for the bitmap. | |
6616 | * Parameters: | |
6617 | * st : supertype information | |
6618 | * vol_idx : the volume index to use | |
6619 | * | |
6620 | * Returns: | |
6621 | * 0 : success | |
6622 | * -1 : fail | |
6623 | ******************************************************************************/ | |
6624 | static int write_init_bitmap_imsm_vol(struct supertype *st, int vol_idx) | |
6625 | { | |
6626 | struct intel_super *super = st->sb; | |
6627 | int prev_current_vol = super->current_vol; | |
6628 | struct dl *d; | |
6629 | int ret = 0; | |
6630 | ||
6631 | super->current_vol = vol_idx; | |
6632 | for (d = super->disks; d; d = d->next) { | |
6633 | if (d->index < 0 || is_failed(&d->disk)) | |
6634 | continue; | |
6635 | ret = st->ss->write_bitmap(st, d->fd, NoUpdate); | |
6636 | if (ret) | |
6637 | break; | |
6638 | } | |
6639 | super->current_vol = prev_current_vol; | |
6640 | return ret; | |
6641 | } | |
6642 | ||
6643 | /******************************************************************************* | |
6644 | * Function: write_init_bitmap_imsm_all | |
6645 | * Description: Write a bitmap header and prepares the area for the bitmap. | |
6646 | * Operation is executed for volumes with CONSISTENCY_POLICY_BITMAP. | |
6647 | * Parameters: | |
6648 | * st : supertype information | |
6649 | * info : info about the volume where the bitmap should be written | |
6650 | * vol_idx : the volume index to use | |
6651 | * | |
6652 | * Returns: | |
6653 | * 0 : success | |
6654 | * -1 : fail | |
6655 | ******************************************************************************/ | |
6656 | static int write_init_bitmap_imsm_all(struct supertype *st, struct mdinfo *info, | |
6657 | int vol_idx) | |
6658 | { | |
6659 | int ret = 0; | |
6660 | ||
6661 | if (info && (info->consistency_policy == CONSISTENCY_POLICY_BITMAP)) | |
6662 | ret = write_init_bitmap_imsm_vol(st, vol_idx); | |
6663 | ||
6664 | return ret; | |
6665 | } | |
6666 | ||
c2c087e6 DW |
6667 | static int write_init_super_imsm(struct supertype *st) |
6668 | { | |
9b1fb677 DW |
6669 | struct intel_super *super = st->sb; |
6670 | int current_vol = super->current_vol; | |
2432ce9b AP |
6671 | int rv = 0; |
6672 | struct mdinfo info; | |
6673 | ||
6674 | getinfo_super_imsm(st, &info, NULL); | |
9b1fb677 DW |
6675 | |
6676 | /* we are done with current_vol reset it to point st at the container */ | |
6677 | super->current_vol = -1; | |
6678 | ||
8273f55e | 6679 | if (st->update_tail) { |
43dad3d6 DW |
6680 | /* queue the recently created array / added disk |
6681 | * as a metadata update */ | |
8273f55e | 6682 | |
43dad3d6 | 6683 | /* determine if we are creating a volume or adding a disk */ |
9b1fb677 | 6684 | if (current_vol < 0) { |
1a64be56 LM |
6685 | /* in the mgmt (add/remove) disk case we are running |
6686 | * in mdmon context, so don't close fd's | |
43dad3d6 | 6687 | */ |
2432ce9b AP |
6688 | rv = mgmt_disk(st); |
6689 | } else { | |
fbc42556 | 6690 | /* adding the second volume to the array */ |
2432ce9b | 6691 | rv = write_init_ppl_imsm_all(st, &info); |
fbc42556 JR |
6692 | if (!rv) |
6693 | rv = write_init_bitmap_imsm_all(st, &info, current_vol); | |
2432ce9b AP |
6694 | if (!rv) |
6695 | rv = create_array(st, current_vol); | |
6696 | } | |
d682f344 N |
6697 | } else { |
6698 | struct dl *d; | |
6699 | for (d = super->disks; d; d = d->next) | |
ba728be7 | 6700 | Kill(d->devname, NULL, 0, -1, 1); |
fbc42556 | 6701 | if (current_vol >= 0) { |
2432ce9b | 6702 | rv = write_init_ppl_imsm_all(st, &info); |
fbc42556 JR |
6703 | if (!rv) |
6704 | rv = write_init_bitmap_imsm_all(st, &info, current_vol); | |
6705 | } | |
6706 | ||
2432ce9b AP |
6707 | if (!rv) |
6708 | rv = write_super_imsm(st, 1); | |
d682f344 | 6709 | } |
2432ce9b AP |
6710 | |
6711 | return rv; | |
cdddbdbc DW |
6712 | } |
6713 | ||
e683ca88 | 6714 | static int store_super_imsm(struct supertype *st, int fd) |
cdddbdbc | 6715 | { |
e683ca88 DW |
6716 | struct intel_super *super = st->sb; |
6717 | struct imsm_super *mpb = super ? super->anchor : NULL; | |
551c80c1 | 6718 | |
e683ca88 | 6719 | if (!mpb) |
ad97895e DW |
6720 | return 1; |
6721 | ||
f36a9ecd PB |
6722 | if (super->sector_size == 4096) |
6723 | convert_to_4k(super); | |
e683ca88 | 6724 | return store_imsm_mpb(fd, mpb); |
cdddbdbc DW |
6725 | } |
6726 | ||
cdddbdbc | 6727 | static int validate_geometry_imsm_container(struct supertype *st, int level, |
1f5d54a0 | 6728 | int raiddisks, |
af4348dd N |
6729 | unsigned long long data_offset, |
6730 | char *dev, | |
2c514b71 NB |
6731 | unsigned long long *freesize, |
6732 | int verbose) | |
cdddbdbc | 6733 | { |
c2c087e6 DW |
6734 | int fd; |
6735 | unsigned long long ldsize; | |
8662f92d | 6736 | struct intel_super *super = NULL; |
f2f5c343 | 6737 | int rv = 0; |
cdddbdbc | 6738 | |
6f2af6a4 | 6739 | if (!is_container(level)) |
c2c087e6 DW |
6740 | return 0; |
6741 | if (!dev) | |
6742 | return 1; | |
6743 | ||
dca80fcd | 6744 | fd = dev_open(dev, O_RDONLY|O_EXCL); |
4389ce73 MT |
6745 | if (!is_fd_valid(fd)) { |
6746 | pr_vrb("imsm: Cannot open %s: %s\n", dev, strerror(errno)); | |
c2c087e6 DW |
6747 | return 0; |
6748 | } | |
8662f92d MT |
6749 | if (!get_dev_size(fd, dev, &ldsize)) |
6750 | goto exit; | |
f2f5c343 LM |
6751 | |
6752 | /* capabilities retrieve could be possible | |
6753 | * note that there is no fd for the disks in array. | |
6754 | */ | |
6755 | super = alloc_super(); | |
8662f92d MT |
6756 | if (!super) |
6757 | goto exit; | |
6758 | ||
6759 | if (!get_dev_sector_size(fd, NULL, &super->sector_size)) | |
6760 | goto exit; | |
fa7bb6f8 | 6761 | |
ba728be7 | 6762 | rv = find_intel_hba_capability(fd, super, verbose > 0 ? dev : NULL); |
f2f5c343 LM |
6763 | if (rv != 0) { |
6764 | #if DEBUG | |
6765 | char str[256]; | |
6766 | fd2devname(fd, str); | |
1ade5cc1 | 6767 | dprintf("fd: %d %s orom: %p rv: %d raiddisk: %d\n", |
f2f5c343 LM |
6768 | fd, str, super->orom, rv, raiddisks); |
6769 | #endif | |
6770 | /* no orom/efi or non-intel hba of the disk */ | |
8662f92d MT |
6771 | rv = 0; |
6772 | goto exit; | |
f2f5c343 | 6773 | } |
9126b9a8 CA |
6774 | if (super->orom) { |
6775 | if (raiddisks > super->orom->tds) { | |
6776 | if (verbose) | |
7a862a02 | 6777 | pr_err("%d exceeds maximum number of platform supported disks: %d\n", |
9126b9a8 | 6778 | raiddisks, super->orom->tds); |
8662f92d | 6779 | goto exit; |
9126b9a8 CA |
6780 | } |
6781 | if ((super->orom->attr & IMSM_OROM_ATTR_2TB_DISK) == 0 && | |
6782 | (ldsize >> 9) >> 32 > 0) { | |
6783 | if (verbose) | |
e7b84f9d | 6784 | pr_err("%s exceeds maximum platform supported size\n", dev); |
8662f92d MT |
6785 | goto exit; |
6786 | } | |
6787 | ||
6788 | if (super->hba->type == SYS_DEV_VMD || | |
6789 | super->hba->type == SYS_DEV_NVME) { | |
6790 | if (!imsm_is_nvme_namespace_supported(fd, 1)) { | |
6791 | if (verbose) | |
6792 | pr_err("NVMe namespace %s is not supported by IMSM\n", | |
6793 | basename(dev)); | |
6794 | goto exit; | |
6795 | } | |
9126b9a8 | 6796 | } |
f2f5c343 | 6797 | } |
1f5d54a0 MT |
6798 | if (freesize) |
6799 | *freesize = avail_size_imsm(st, ldsize >> 9, data_offset); | |
8662f92d MT |
6800 | rv = 1; |
6801 | exit: | |
6802 | if (super) | |
6803 | free_imsm(super); | |
6804 | close(fd); | |
c2c087e6 | 6805 | |
8662f92d | 6806 | return rv; |
cdddbdbc DW |
6807 | } |
6808 | ||
0dcecb2e DW |
6809 | static unsigned long long find_size(struct extent *e, int *idx, int num_extents) |
6810 | { | |
6811 | const unsigned long long base_start = e[*idx].start; | |
6812 | unsigned long long end = base_start + e[*idx].size; | |
6813 | int i; | |
6814 | ||
6815 | if (base_start == end) | |
6816 | return 0; | |
6817 | ||
6818 | *idx = *idx + 1; | |
6819 | for (i = *idx; i < num_extents; i++) { | |
6820 | /* extend overlapping extents */ | |
6821 | if (e[i].start >= base_start && | |
6822 | e[i].start <= end) { | |
6823 | if (e[i].size == 0) | |
6824 | return 0; | |
6825 | if (e[i].start + e[i].size > end) | |
6826 | end = e[i].start + e[i].size; | |
6827 | } else if (e[i].start > end) { | |
6828 | *idx = i; | |
6829 | break; | |
6830 | } | |
6831 | } | |
6832 | ||
6833 | return end - base_start; | |
6834 | } | |
6835 | ||
aa19fdd4 | 6836 | /** merge_extents() - analyze extents and get free size. |
9bc426fa | 6837 | * @super: Intel metadata, not NULL. |
aa19fdd4 | 6838 | * @expanding: if set, we are expanding &super->current_vol. |
9bc426fa | 6839 | * |
aa19fdd4 MT |
6840 | * Build a composite disk with all known extents and generate a size given the |
6841 | * "all disks in an array must share a common start offset" constraint. | |
6842 | * If a volume is expanded, then return free space after the volume. | |
9bc426fa | 6843 | * |
aa19fdd4 | 6844 | * Return: Free space or 0 on failure. |
9bc426fa | 6845 | */ |
aa19fdd4 | 6846 | static unsigned long long merge_extents(struct intel_super *super, const bool expanding) |
0dcecb2e | 6847 | { |
9bc426fa | 6848 | struct extent *e; |
0dcecb2e | 6849 | struct dl *dl; |
aa19fdd4 MT |
6850 | int i, j, pos_vol_idx = -1; |
6851 | int extent_idx = 0; | |
6852 | int sum_extents = 0; | |
6853 | unsigned long long pos = 0; | |
b9d77223 | 6854 | unsigned long long start = 0; |
1dea84ae MT |
6855 | unsigned long long free_size = 0; |
6856 | ||
6857 | unsigned long pre_reservation = 0; | |
6858 | unsigned long post_reservation = IMSM_RESERVED_SECTORS; | |
6859 | unsigned long reservation_size; | |
0dcecb2e | 6860 | |
9bc426fa MT |
6861 | for (dl = super->disks; dl; dl = dl->next) |
6862 | if (dl->e) | |
6863 | sum_extents += dl->extent_cnt; | |
6864 | e = xcalloc(sum_extents, sizeof(struct extent)); | |
6865 | ||
0dcecb2e DW |
6866 | /* coalesce and sort all extents. also, check to see if we need to |
6867 | * reserve space between member arrays | |
6868 | */ | |
6869 | j = 0; | |
6870 | for (dl = super->disks; dl; dl = dl->next) { | |
6871 | if (!dl->e) | |
6872 | continue; | |
6873 | for (i = 0; i < dl->extent_cnt; i++) | |
6874 | e[j++] = dl->e[i]; | |
6875 | } | |
6876 | qsort(e, sum_extents, sizeof(*e), cmp_extent); | |
6877 | ||
6878 | /* merge extents */ | |
6879 | i = 0; | |
6880 | j = 0; | |
6881 | while (i < sum_extents) { | |
6882 | e[j].start = e[i].start; | |
aa19fdd4 | 6883 | e[j].vol = e[i].vol; |
0dcecb2e DW |
6884 | e[j].size = find_size(e, &i, sum_extents); |
6885 | j++; | |
6886 | if (e[j-1].size == 0) | |
6887 | break; | |
6888 | } | |
6889 | ||
0dcecb2e DW |
6890 | i = 0; |
6891 | do { | |
aa19fdd4 | 6892 | unsigned long long esize = e[i].start - pos; |
0dcecb2e | 6893 | |
1dea84ae MT |
6894 | if (expanding ? pos_vol_idx == super->current_vol : esize >= free_size) { |
6895 | free_size = esize; | |
0dcecb2e | 6896 | start = pos; |
aa19fdd4 | 6897 | extent_idx = i; |
0dcecb2e | 6898 | } |
aa19fdd4 | 6899 | |
0dcecb2e | 6900 | pos = e[i].start + e[i].size; |
aa19fdd4 MT |
6901 | pos_vol_idx = e[i].vol; |
6902 | ||
0dcecb2e DW |
6903 | i++; |
6904 | } while (e[i-1].size); | |
0dcecb2e | 6905 | |
1dea84ae MT |
6906 | if (free_size == 0) { |
6907 | dprintf("imsm: Cannot find free size.\n"); | |
6908 | free(e); | |
a7dd165b | 6909 | return 0; |
1dea84ae | 6910 | } |
a7dd165b | 6911 | |
1dea84ae MT |
6912 | if (!expanding && extent_idx != 0) |
6913 | /* | |
6914 | * Not a real first volume in a container is created, pre_reservation is needed. | |
6915 | */ | |
6916 | pre_reservation = IMSM_RESERVED_SECTORS; | |
0dcecb2e | 6917 | |
1dea84ae MT |
6918 | if (e[extent_idx].size == 0) |
6919 | /* | |
6920 | * extent_idx points to the metadata, post_reservation is allready done. | |
6921 | */ | |
6922 | post_reservation = 0; | |
6923 | free(e); | |
0dcecb2e | 6924 | |
1dea84ae MT |
6925 | reservation_size = pre_reservation + post_reservation; |
6926 | ||
6927 | if (free_size < reservation_size) { | |
6928 | dprintf("imsm: Reservation size is greater than free space.\n"); | |
6929 | return 0; | |
6930 | } | |
0dcecb2e | 6931 | |
1dea84ae MT |
6932 | super->create_offset = start + pre_reservation; |
6933 | return free_size - reservation_size; | |
0dcecb2e DW |
6934 | } |
6935 | ||
88c32bb1 DW |
6936 | static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks) |
6937 | { | |
6938 | if (level < 0 || level == 6 || level == 4) | |
6939 | return 0; | |
6940 | ||
6941 | /* if we have an orom prevent invalid raid levels */ | |
6942 | if (orom) | |
6943 | switch (level) { | |
6944 | case 0: return imsm_orom_has_raid0(orom); | |
6945 | case 1: | |
6946 | if (raiddisks > 2) | |
6947 | return imsm_orom_has_raid1e(orom); | |
1c556e92 DW |
6948 | return imsm_orom_has_raid1(orom) && raiddisks == 2; |
6949 | case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4; | |
6950 | case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2; | |
88c32bb1 DW |
6951 | } |
6952 | else | |
6953 | return 1; /* not on an Intel RAID platform so anything goes */ | |
6954 | ||
6955 | return 0; | |
6956 | } | |
6957 | ||
ca9de185 LM |
6958 | static int |
6959 | active_arrays_by_format(char *name, char* hba, struct md_list **devlist, | |
6960 | int dpa, int verbose) | |
6961 | { | |
6962 | struct mdstat_ent *mdstat = mdstat_read(0, 0); | |
594dc1b8 | 6963 | struct mdstat_ent *memb; |
ca9de185 LM |
6964 | int count = 0; |
6965 | int num = 0; | |
594dc1b8 | 6966 | struct md_list *dv; |
ca9de185 LM |
6967 | int found; |
6968 | ||
6969 | for (memb = mdstat ; memb ; memb = memb->next) { | |
6970 | if (memb->metadata_version && | |
fc54fe7a | 6971 | (strncmp(memb->metadata_version, "external:", 9) == 0) && |
ca9de185 LM |
6972 | (strcmp(&memb->metadata_version[9], name) == 0) && |
6973 | !is_subarray(memb->metadata_version+9) && | |
6974 | memb->members) { | |
6975 | struct dev_member *dev = memb->members; | |
6976 | int fd = -1; | |
4389ce73 | 6977 | while (dev && !is_fd_valid(fd)) { |
503975b9 | 6978 | char *path = xmalloc(strlen(dev->name) + strlen("/dev/") + 1); |
dd5ab402 | 6979 | num = snprintf(path, PATH_MAX, "%s%s", "/dev/", dev->name); |
503975b9 N |
6980 | if (num > 0) |
6981 | fd = open(path, O_RDONLY, 0); | |
4389ce73 | 6982 | if (num <= 0 || !is_fd_valid(fd)) { |
676e87a8 | 6983 | pr_vrb("Cannot open %s: %s\n", |
503975b9 | 6984 | dev->name, strerror(errno)); |
ca9de185 | 6985 | } |
503975b9 | 6986 | free(path); |
ca9de185 LM |
6987 | dev = dev->next; |
6988 | } | |
6989 | found = 0; | |
4389ce73 | 6990 | if (is_fd_valid(fd) && disk_attached_to_hba(fd, hba)) { |
ca9de185 LM |
6991 | struct mdstat_ent *vol; |
6992 | for (vol = mdstat ; vol ; vol = vol->next) { | |
089f9d79 | 6993 | if (vol->active > 0 && |
ca9de185 | 6994 | vol->metadata_version && |
9581efb1 | 6995 | is_container_member(vol, memb->devnm)) { |
ca9de185 LM |
6996 | found++; |
6997 | count++; | |
6998 | } | |
6999 | } | |
7000 | if (*devlist && (found < dpa)) { | |
503975b9 | 7001 | dv = xcalloc(1, sizeof(*dv)); |
9581efb1 N |
7002 | dv->devname = xmalloc(strlen(memb->devnm) + strlen("/dev/") + 1); |
7003 | sprintf(dv->devname, "%s%s", "/dev/", memb->devnm); | |
503975b9 N |
7004 | dv->found = found; |
7005 | dv->used = 0; | |
7006 | dv->next = *devlist; | |
7007 | *devlist = dv; | |
ca9de185 LM |
7008 | } |
7009 | } | |
4389ce73 | 7010 | close_fd(&fd); |
ca9de185 LM |
7011 | } |
7012 | } | |
7013 | free_mdstat(mdstat); | |
7014 | return count; | |
7015 | } | |
7016 | ||
7017 | #ifdef DEBUG_LOOP | |
7018 | static struct md_list* | |
7019 | get_loop_devices(void) | |
7020 | { | |
7021 | int i; | |
7022 | struct md_list *devlist = NULL; | |
594dc1b8 | 7023 | struct md_list *dv; |
ca9de185 LM |
7024 | |
7025 | for(i = 0; i < 12; i++) { | |
503975b9 N |
7026 | dv = xcalloc(1, sizeof(*dv)); |
7027 | dv->devname = xmalloc(40); | |
ca9de185 LM |
7028 | sprintf(dv->devname, "/dev/loop%d", i); |
7029 | dv->next = devlist; | |
7030 | devlist = dv; | |
7031 | } | |
7032 | return devlist; | |
7033 | } | |
7034 | #endif | |
7035 | ||
7036 | static struct md_list* | |
7037 | get_devices(const char *hba_path) | |
7038 | { | |
7039 | struct md_list *devlist = NULL; | |
594dc1b8 | 7040 | struct md_list *dv; |
ca9de185 LM |
7041 | struct dirent *ent; |
7042 | DIR *dir; | |
7043 | int err = 0; | |
7044 | ||
7045 | #if DEBUG_LOOP | |
7046 | devlist = get_loop_devices(); | |
7047 | return devlist; | |
7048 | #endif | |
7049 | /* scroll through /sys/dev/block looking for devices attached to | |
7050 | * this hba | |
7051 | */ | |
7052 | dir = opendir("/sys/dev/block"); | |
7053 | for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) { | |
7054 | int fd; | |
7055 | char buf[1024]; | |
7056 | int major, minor; | |
7057 | char *path = NULL; | |
7058 | if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2) | |
7059 | continue; | |
7c798f87 | 7060 | path = devt_to_devpath(makedev(major, minor), 1, NULL); |
ca9de185 LM |
7061 | if (!path) |
7062 | continue; | |
7063 | if (!path_attached_to_hba(path, hba_path)) { | |
7064 | free(path); | |
7065 | path = NULL; | |
7066 | continue; | |
7067 | } | |
7068 | free(path); | |
7069 | path = NULL; | |
7070 | fd = dev_open(ent->d_name, O_RDONLY); | |
4389ce73 | 7071 | if (is_fd_valid(fd)) { |
ca9de185 LM |
7072 | fd2devname(fd, buf); |
7073 | close(fd); | |
7074 | } else { | |
e7b84f9d | 7075 | pr_err("cannot open device: %s\n", |
ca9de185 LM |
7076 | ent->d_name); |
7077 | continue; | |
7078 | } | |
7079 | ||
503975b9 N |
7080 | dv = xcalloc(1, sizeof(*dv)); |
7081 | dv->devname = xstrdup(buf); | |
ca9de185 LM |
7082 | dv->next = devlist; |
7083 | devlist = dv; | |
7084 | } | |
7085 | if (err) { | |
7086 | while(devlist) { | |
7087 | dv = devlist; | |
7088 | devlist = devlist->next; | |
7089 | free(dv->devname); | |
7090 | free(dv); | |
7091 | } | |
7092 | } | |
562aa102 | 7093 | closedir(dir); |
ca9de185 LM |
7094 | return devlist; |
7095 | } | |
7096 | ||
7097 | static int | |
7098 | count_volumes_list(struct md_list *devlist, char *homehost, | |
7099 | int verbose, int *found) | |
7100 | { | |
7101 | struct md_list *tmpdev; | |
7102 | int count = 0; | |
594dc1b8 | 7103 | struct supertype *st; |
ca9de185 LM |
7104 | |
7105 | /* first walk the list of devices to find a consistent set | |
7106 | * that match the criterea, if that is possible. | |
7107 | * We flag the ones we like with 'used'. | |
7108 | */ | |
7109 | *found = 0; | |
7110 | st = match_metadata_desc_imsm("imsm"); | |
7111 | if (st == NULL) { | |
676e87a8 | 7112 | pr_vrb("cannot allocate memory for imsm supertype\n"); |
ca9de185 LM |
7113 | return 0; |
7114 | } | |
7115 | ||
7116 | for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { | |
7117 | char *devname = tmpdev->devname; | |
0a6bff09 | 7118 | dev_t rdev; |
ca9de185 LM |
7119 | struct supertype *tst; |
7120 | int dfd; | |
7121 | if (tmpdev->used > 1) | |
7122 | continue; | |
7123 | tst = dup_super(st); | |
7124 | if (tst == NULL) { | |
676e87a8 | 7125 | pr_vrb("cannot allocate memory for imsm supertype\n"); |
ca9de185 LM |
7126 | goto err_1; |
7127 | } | |
7128 | tmpdev->container = 0; | |
7129 | dfd = dev_open(devname, O_RDONLY|O_EXCL); | |
4389ce73 | 7130 | if (!is_fd_valid(dfd)) { |
1ade5cc1 | 7131 | dprintf("cannot open device %s: %s\n", |
ca9de185 LM |
7132 | devname, strerror(errno)); |
7133 | tmpdev->used = 2; | |
0a6bff09 | 7134 | } else if (!fstat_is_blkdev(dfd, devname, &rdev)) { |
ca9de185 LM |
7135 | tmpdev->used = 2; |
7136 | } else if (must_be_container(dfd)) { | |
7137 | struct supertype *cst; | |
7138 | cst = super_by_fd(dfd, NULL); | |
7139 | if (cst == NULL) { | |
1ade5cc1 | 7140 | dprintf("cannot recognize container type %s\n", |
ca9de185 LM |
7141 | devname); |
7142 | tmpdev->used = 2; | |
7143 | } else if (tst->ss != st->ss) { | |
1ade5cc1 | 7144 | dprintf("non-imsm container - ignore it: %s\n", |
ca9de185 LM |
7145 | devname); |
7146 | tmpdev->used = 2; | |
7147 | } else if (!tst->ss->load_container || | |
7148 | tst->ss->load_container(tst, dfd, NULL)) | |
7149 | tmpdev->used = 2; | |
7150 | else { | |
7151 | tmpdev->container = 1; | |
7152 | } | |
7153 | if (cst) | |
7154 | cst->ss->free_super(cst); | |
7155 | } else { | |
0a6bff09 | 7156 | tmpdev->st_rdev = rdev; |
ca9de185 | 7157 | if (tst->ss->load_super(tst,dfd, NULL)) { |
1ade5cc1 | 7158 | dprintf("no RAID superblock on %s\n", |
ca9de185 LM |
7159 | devname); |
7160 | tmpdev->used = 2; | |
7161 | } else if (tst->ss->compare_super == NULL) { | |
1ade5cc1 | 7162 | dprintf("Cannot assemble %s metadata on %s\n", |
ca9de185 LM |
7163 | tst->ss->name, devname); |
7164 | tmpdev->used = 2; | |
7165 | } | |
7166 | } | |
4389ce73 MT |
7167 | close_fd(&dfd); |
7168 | ||
ca9de185 LM |
7169 | if (tmpdev->used == 2 || tmpdev->used == 4) { |
7170 | /* Ignore unrecognised devices during auto-assembly */ | |
7171 | goto loop; | |
7172 | } | |
7173 | else { | |
7174 | struct mdinfo info; | |
7175 | tst->ss->getinfo_super(tst, &info, NULL); | |
7176 | ||
7177 | if (st->minor_version == -1) | |
7178 | st->minor_version = tst->minor_version; | |
7179 | ||
7180 | if (memcmp(info.uuid, uuid_zero, | |
7181 | sizeof(int[4])) == 0) { | |
7182 | /* this is a floating spare. It cannot define | |
7183 | * an array unless there are no more arrays of | |
7184 | * this type to be found. It can be included | |
7185 | * in an array of this type though. | |
7186 | */ | |
7187 | tmpdev->used = 3; | |
7188 | goto loop; | |
7189 | } | |
7190 | ||
7191 | if (st->ss != tst->ss || | |
7192 | st->minor_version != tst->minor_version || | |
c7b8547c | 7193 | st->ss->compare_super(st, tst, 1) != 0) { |
ca9de185 LM |
7194 | /* Some mismatch. If exactly one array matches this host, |
7195 | * we can resolve on that one. | |
7196 | * Or, if we are auto assembling, we just ignore the second | |
7197 | * for now. | |
7198 | */ | |
1ade5cc1 | 7199 | dprintf("superblock on %s doesn't match others - assembly aborted\n", |
ca9de185 LM |
7200 | devname); |
7201 | goto loop; | |
7202 | } | |
7203 | tmpdev->used = 1; | |
7204 | *found = 1; | |
7205 | dprintf("found: devname: %s\n", devname); | |
7206 | } | |
7207 | loop: | |
7208 | if (tst) | |
7209 | tst->ss->free_super(tst); | |
7210 | } | |
7211 | if (*found != 0) { | |
7212 | int err; | |
7213 | if ((err = load_super_imsm_all(st, -1, &st->sb, NULL, devlist, 0)) == 0) { | |
7214 | struct mdinfo *iter, *head = st->ss->container_content(st, NULL); | |
7215 | for (iter = head; iter; iter = iter->next) { | |
7216 | dprintf("content->text_version: %s vol\n", | |
7217 | iter->text_version); | |
7218 | if (iter->array.state & (1<<MD_SB_BLOCK_VOLUME)) { | |
7219 | /* do not assemble arrays with unsupported | |
7220 | configurations */ | |
1ade5cc1 | 7221 | dprintf("Cannot activate member %s.\n", |
ca9de185 LM |
7222 | iter->text_version); |
7223 | } else | |
7224 | count++; | |
7225 | } | |
7226 | sysfs_free(head); | |
7227 | ||
7228 | } else { | |
1ade5cc1 | 7229 | dprintf("No valid super block on device list: err: %d %p\n", |
ca9de185 LM |
7230 | err, st->sb); |
7231 | } | |
7232 | } else { | |
1ade5cc1 | 7233 | dprintf("no more devices to examine\n"); |
ca9de185 LM |
7234 | } |
7235 | ||
7236 | for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { | |
089f9d79 | 7237 | if (tmpdev->used == 1 && tmpdev->found) { |
ca9de185 LM |
7238 | if (count) { |
7239 | if (count < tmpdev->found) | |
7240 | count = 0; | |
7241 | else | |
7242 | count -= tmpdev->found; | |
7243 | } | |
7244 | } | |
7245 | if (tmpdev->used == 1) | |
7246 | tmpdev->used = 4; | |
7247 | } | |
7248 | err_1: | |
7249 | if (st) | |
7250 | st->ss->free_super(st); | |
7251 | return count; | |
7252 | } | |
7253 | ||
d3c11416 AO |
7254 | static int __count_volumes(char *hba_path, int dpa, int verbose, |
7255 | int cmp_hba_path) | |
ca9de185 | 7256 | { |
72a45777 | 7257 | struct sys_dev *idev, *intel_devices = find_intel_devices(); |
ca9de185 | 7258 | int count = 0; |
72a45777 PB |
7259 | const struct orom_entry *entry; |
7260 | struct devid_list *dv, *devid_list; | |
ca9de185 | 7261 | |
d3c11416 | 7262 | if (!hba_path) |
ca9de185 LM |
7263 | return 0; |
7264 | ||
72a45777 | 7265 | for (idev = intel_devices; idev; idev = idev->next) { |
d3c11416 AO |
7266 | if (strstr(idev->path, hba_path)) |
7267 | break; | |
72a45777 PB |
7268 | } |
7269 | ||
7270 | if (!idev || !idev->dev_id) | |
ca9de185 | 7271 | return 0; |
72a45777 PB |
7272 | |
7273 | entry = get_orom_entry_by_device_id(idev->dev_id); | |
7274 | ||
7275 | if (!entry || !entry->devid_list) | |
7276 | return 0; | |
7277 | ||
7278 | devid_list = entry->devid_list; | |
7279 | for (dv = devid_list; dv; dv = dv->next) { | |
594dc1b8 | 7280 | struct md_list *devlist; |
d3c11416 AO |
7281 | struct sys_dev *device = NULL; |
7282 | char *hpath; | |
72a45777 PB |
7283 | int found = 0; |
7284 | ||
d3c11416 AO |
7285 | if (cmp_hba_path) |
7286 | device = device_by_id_and_path(dv->devid, hba_path); | |
7287 | else | |
7288 | device = device_by_id(dv->devid); | |
7289 | ||
72a45777 | 7290 | if (device) |
d3c11416 | 7291 | hpath = device->path; |
72a45777 PB |
7292 | else |
7293 | return 0; | |
7294 | ||
d3c11416 | 7295 | devlist = get_devices(hpath); |
72a45777 PB |
7296 | /* if no intel devices return zero volumes */ |
7297 | if (devlist == NULL) | |
7298 | return 0; | |
7299 | ||
d3c11416 AO |
7300 | count += active_arrays_by_format("imsm", hpath, &devlist, dpa, |
7301 | verbose); | |
7302 | dprintf("path: %s active arrays: %d\n", hpath, count); | |
72a45777 PB |
7303 | if (devlist == NULL) |
7304 | return 0; | |
7305 | do { | |
7306 | found = 0; | |
7307 | count += count_volumes_list(devlist, | |
7308 | NULL, | |
7309 | verbose, | |
7310 | &found); | |
7311 | dprintf("found %d count: %d\n", found, count); | |
7312 | } while (found); | |
7313 | ||
d3c11416 | 7314 | dprintf("path: %s total number of volumes: %d\n", hpath, count); |
72a45777 PB |
7315 | |
7316 | while (devlist) { | |
7317 | struct md_list *dv = devlist; | |
7318 | devlist = devlist->next; | |
7319 | free(dv->devname); | |
7320 | free(dv); | |
7321 | } | |
ca9de185 LM |
7322 | } |
7323 | return count; | |
7324 | } | |
7325 | ||
d3c11416 AO |
7326 | static int count_volumes(struct intel_hba *hba, int dpa, int verbose) |
7327 | { | |
7328 | if (!hba) | |
7329 | return 0; | |
7330 | if (hba->type == SYS_DEV_VMD) { | |
7331 | struct sys_dev *dev; | |
7332 | int count = 0; | |
7333 | ||
7334 | for (dev = find_intel_devices(); dev; dev = dev->next) { | |
7335 | if (dev->type == SYS_DEV_VMD) | |
7336 | count += __count_volumes(dev->path, dpa, | |
7337 | verbose, 1); | |
7338 | } | |
7339 | return count; | |
7340 | } | |
7341 | return __count_volumes(hba->path, dpa, verbose, 0); | |
7342 | } | |
7343 | ||
cd9d1ac7 DW |
7344 | static int imsm_default_chunk(const struct imsm_orom *orom) |
7345 | { | |
7346 | /* up to 512 if the plaform supports it, otherwise the platform max. | |
7347 | * 128 if no platform detected | |
7348 | */ | |
7349 | int fs = max(7, orom ? fls(orom->sss) : 0); | |
7350 | ||
7351 | return min(512, (1 << fs)); | |
7352 | } | |
73408129 | 7353 | |
6592ce37 DW |
7354 | static int |
7355 | validate_geometry_imsm_orom(struct intel_super *super, int level, int layout, | |
2cc699af | 7356 | int raiddisks, int *chunk, unsigned long long size, int verbose) |
6592ce37 | 7357 | { |
660260d0 DW |
7358 | /* check/set platform and metadata limits/defaults */ |
7359 | if (super->orom && raiddisks > super->orom->dpa) { | |
676e87a8 | 7360 | pr_vrb("platform supports a maximum of %d disks per array\n", |
660260d0 | 7361 | super->orom->dpa); |
73408129 LM |
7362 | return 0; |
7363 | } | |
7364 | ||
5d500228 | 7365 | /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */ |
660260d0 | 7366 | if (!is_raid_level_supported(super->orom, level, raiddisks)) { |
676e87a8 | 7367 | pr_vrb("platform does not support raid%d with %d disk%s\n", |
6592ce37 DW |
7368 | level, raiddisks, raiddisks > 1 ? "s" : ""); |
7369 | return 0; | |
7370 | } | |
cd9d1ac7 | 7371 | |
7ccc4cc4 | 7372 | if (*chunk == 0 || *chunk == UnSet) |
cd9d1ac7 DW |
7373 | *chunk = imsm_default_chunk(super->orom); |
7374 | ||
7ccc4cc4 | 7375 | if (super->orom && !imsm_orom_has_chunk(super->orom, *chunk)) { |
676e87a8 | 7376 | pr_vrb("platform does not support a chunk size of: %d\n", *chunk); |
cd9d1ac7 | 7377 | return 0; |
6592ce37 | 7378 | } |
cd9d1ac7 | 7379 | |
6592ce37 DW |
7380 | if (layout != imsm_level_to_layout(level)) { |
7381 | if (level == 5) | |
676e87a8 | 7382 | pr_vrb("imsm raid 5 only supports the left-asymmetric layout\n"); |
6592ce37 | 7383 | else if (level == 10) |
676e87a8 | 7384 | pr_vrb("imsm raid 10 only supports the n2 layout\n"); |
6592ce37 | 7385 | else |
676e87a8 | 7386 | pr_vrb("imsm unknown layout %#x for this raid level %d\n", |
6592ce37 DW |
7387 | layout, level); |
7388 | return 0; | |
7389 | } | |
2cc699af | 7390 | |
7ccc4cc4 | 7391 | if (super->orom && (super->orom->attr & IMSM_OROM_ATTR_2TB) == 0 && |
2cc699af | 7392 | (calc_array_size(level, raiddisks, layout, *chunk, size) >> 32) > 0) { |
676e87a8 | 7393 | pr_vrb("platform does not support a volume size over 2TB\n"); |
2cc699af CA |
7394 | return 0; |
7395 | } | |
614902f6 | 7396 | |
6592ce37 DW |
7397 | return 1; |
7398 | } | |
7399 | ||
1011e834 | 7400 | /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd |
c2c087e6 DW |
7401 | * FIX ME add ahci details |
7402 | */ | |
8b353278 | 7403 | static int validate_geometry_imsm_volume(struct supertype *st, int level, |
c21e737b | 7404 | int layout, int raiddisks, int *chunk, |
af4348dd N |
7405 | unsigned long long size, |
7406 | unsigned long long data_offset, | |
7407 | char *dev, | |
2c514b71 NB |
7408 | unsigned long long *freesize, |
7409 | int verbose) | |
cdddbdbc | 7410 | { |
9e04ac1c | 7411 | dev_t rdev; |
c2c087e6 | 7412 | struct intel_super *super = st->sb; |
b2916f25 | 7413 | struct imsm_super *mpb; |
c2c087e6 DW |
7414 | struct dl *dl; |
7415 | unsigned long long pos = 0; | |
7416 | unsigned long long maxsize; | |
7417 | struct extent *e; | |
7418 | int i; | |
cdddbdbc | 7419 | |
88c32bb1 DW |
7420 | /* We must have the container info already read in. */ |
7421 | if (!super) | |
c2c087e6 DW |
7422 | return 0; |
7423 | ||
b2916f25 JS |
7424 | mpb = super->anchor; |
7425 | ||
2cc699af | 7426 | if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, size, verbose)) { |
3e684231 | 7427 | pr_err("RAID geometry validation failed. Cannot proceed with the action(s).\n"); |
c2c087e6 | 7428 | return 0; |
d54559f0 | 7429 | } |
c2c087e6 DW |
7430 | if (!dev) { |
7431 | /* General test: make sure there is space for | |
2da8544a DW |
7432 | * 'raiddisks' device extents of size 'size' at a given |
7433 | * offset | |
c2c087e6 | 7434 | */ |
e46273eb | 7435 | unsigned long long minsize = size; |
b7528a20 | 7436 | unsigned long long start_offset = MaxSector; |
c2c087e6 DW |
7437 | int dcnt = 0; |
7438 | if (minsize == 0) | |
7439 | minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
7440 | for (dl = super->disks; dl ; dl = dl->next) { | |
7441 | int found = 0; | |
7442 | ||
bf5a934a | 7443 | pos = 0; |
c2c087e6 | 7444 | i = 0; |
05501181 | 7445 | e = get_extents(super, dl, 0); |
c2c087e6 DW |
7446 | if (!e) continue; |
7447 | do { | |
7448 | unsigned long long esize; | |
7449 | esize = e[i].start - pos; | |
7450 | if (esize >= minsize) | |
7451 | found = 1; | |
b7528a20 | 7452 | if (found && start_offset == MaxSector) { |
2da8544a DW |
7453 | start_offset = pos; |
7454 | break; | |
7455 | } else if (found && pos != start_offset) { | |
7456 | found = 0; | |
7457 | break; | |
7458 | } | |
c2c087e6 DW |
7459 | pos = e[i].start + e[i].size; |
7460 | i++; | |
7461 | } while (e[i-1].size); | |
7462 | if (found) | |
7463 | dcnt++; | |
7464 | free(e); | |
7465 | } | |
7466 | if (dcnt < raiddisks) { | |
2c514b71 | 7467 | if (verbose) |
7a862a02 | 7468 | pr_err("imsm: Not enough devices with space for this array (%d < %d)\n", |
2c514b71 | 7469 | dcnt, raiddisks); |
c2c087e6 DW |
7470 | return 0; |
7471 | } | |
7472 | return 1; | |
7473 | } | |
0dcecb2e | 7474 | |
c2c087e6 | 7475 | /* This device must be a member of the set */ |
9e04ac1c | 7476 | if (!stat_is_blkdev(dev, &rdev)) |
c2c087e6 DW |
7477 | return 0; |
7478 | for (dl = super->disks ; dl ; dl = dl->next) { | |
9e04ac1c ZL |
7479 | if (dl->major == (int)major(rdev) && |
7480 | dl->minor == (int)minor(rdev)) | |
c2c087e6 DW |
7481 | break; |
7482 | } | |
7483 | if (!dl) { | |
2c514b71 | 7484 | if (verbose) |
7a862a02 | 7485 | pr_err("%s is not in the same imsm set\n", dev); |
c2c087e6 | 7486 | return 0; |
a20d2ba5 DW |
7487 | } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) { |
7488 | /* If a volume is present then the current creation attempt | |
7489 | * cannot incorporate new spares because the orom may not | |
7490 | * understand this configuration (all member disks must be | |
7491 | * members of each array in the container). | |
7492 | */ | |
7a862a02 N |
7493 | pr_err("%s is a spare and a volume is already defined for this container\n", dev); |
7494 | pr_err("The option-rom requires all member disks to be a member of all volumes\n"); | |
a20d2ba5 | 7495 | return 0; |
5fe62b94 WD |
7496 | } else if (super->orom && mpb->num_raid_devs > 0 && |
7497 | mpb->num_disks != raiddisks) { | |
7a862a02 | 7498 | pr_err("The option-rom requires all member disks to be a member of all volumes\n"); |
5fe62b94 | 7499 | return 0; |
c2c087e6 | 7500 | } |
0dcecb2e DW |
7501 | |
7502 | /* retrieve the largest free space block */ | |
05501181 | 7503 | e = get_extents(super, dl, 0); |
c2c087e6 DW |
7504 | maxsize = 0; |
7505 | i = 0; | |
0dcecb2e DW |
7506 | if (e) { |
7507 | do { | |
7508 | unsigned long long esize; | |
7509 | ||
7510 | esize = e[i].start - pos; | |
7511 | if (esize >= maxsize) | |
7512 | maxsize = esize; | |
7513 | pos = e[i].start + e[i].size; | |
7514 | i++; | |
7515 | } while (e[i-1].size); | |
7516 | dl->e = e; | |
7517 | dl->extent_cnt = i; | |
7518 | } else { | |
7519 | if (verbose) | |
e7b84f9d | 7520 | pr_err("unable to determine free space for: %s\n", |
0dcecb2e DW |
7521 | dev); |
7522 | return 0; | |
7523 | } | |
7524 | if (maxsize < size) { | |
7525 | if (verbose) | |
e7b84f9d | 7526 | pr_err("%s not enough space (%llu < %llu)\n", |
0dcecb2e DW |
7527 | dev, maxsize, size); |
7528 | return 0; | |
7529 | } | |
7530 | ||
aa19fdd4 | 7531 | maxsize = merge_extents(super, false); |
3baa56ab | 7532 | |
1a1ced1e KS |
7533 | if (mpb->num_raid_devs > 0 && size && size != maxsize) |
7534 | pr_err("attempting to create a second volume with size less then remaining space.\n"); | |
3baa56ab | 7535 | |
a7dd165b | 7536 | if (maxsize < size || maxsize == 0) { |
b3071342 LD |
7537 | if (verbose) { |
7538 | if (maxsize == 0) | |
7a862a02 | 7539 | pr_err("no free space left on device. Aborting...\n"); |
b3071342 | 7540 | else |
7a862a02 | 7541 | pr_err("not enough space to create volume of given size (%llu < %llu). Aborting...\n", |
b3071342 LD |
7542 | maxsize, size); |
7543 | } | |
0dcecb2e | 7544 | return 0; |
0dcecb2e DW |
7545 | } |
7546 | ||
c2c087e6 DW |
7547 | *freesize = maxsize; |
7548 | ||
ca9de185 | 7549 | if (super->orom) { |
72a45777 | 7550 | int count = count_volumes(super->hba, |
ca9de185 LM |
7551 | super->orom->dpa, verbose); |
7552 | if (super->orom->vphba <= count) { | |
676e87a8 | 7553 | pr_vrb("platform does not support more than %d raid volumes.\n", |
ca9de185 LM |
7554 | super->orom->vphba); |
7555 | return 0; | |
7556 | } | |
7557 | } | |
c2c087e6 | 7558 | return 1; |
cdddbdbc DW |
7559 | } |
7560 | ||
6d4d9ab2 MT |
7561 | /** |
7562 | * imsm_get_free_size() - get the biggest, common free space from members. | |
7563 | * @super: &intel_super pointer, not NULL. | |
7564 | * @raiddisks: number of raid disks. | |
7565 | * @size: requested size, could be 0 (means max size). | |
5f027b93 | 7566 | * @chunk: requested chunk size in KiB. |
6d4d9ab2 MT |
7567 | * @freesize: pointer for returned size value. |
7568 | * | |
7569 | * Return: &IMSM_STATUS_OK or &IMSM_STATUS_ERROR. | |
7570 | * | |
7571 | * @freesize is set to meaningful value, this can be @size, or calculated | |
7572 | * max free size. | |
7573 | * super->create_offset value is modified and set appropriately in | |
7574 | * merge_extends() for further creation. | |
7575 | */ | |
7576 | static imsm_status_t imsm_get_free_size(struct intel_super *super, | |
7577 | const int raiddisks, | |
7578 | unsigned long long size, | |
7579 | const int chunk, | |
aa19fdd4 MT |
7580 | unsigned long long *freesize, |
7581 | bool expanding) | |
efb30e7f | 7582 | { |
efb30e7f DW |
7583 | struct imsm_super *mpb = super->anchor; |
7584 | struct dl *dl; | |
7585 | int i; | |
efb30e7f | 7586 | struct extent *e; |
5f027b93 MT |
7587 | int cnt = 0; |
7588 | int used = 0; | |
efb30e7f | 7589 | unsigned long long maxsize; |
5f027b93 MT |
7590 | unsigned long long minsize = size; |
7591 | ||
7592 | if (minsize == 0) | |
7593 | minsize = chunk * 2; | |
efb30e7f DW |
7594 | |
7595 | /* find the largest common start free region of the possible disks */ | |
efb30e7f DW |
7596 | for (dl = super->disks; dl; dl = dl->next) { |
7597 | dl->raiddisk = -1; | |
7598 | ||
7599 | if (dl->index >= 0) | |
7600 | used++; | |
7601 | ||
7602 | /* don't activate new spares if we are orom constrained | |
7603 | * and there is already a volume active in the container | |
7604 | */ | |
7605 | if (super->orom && dl->index < 0 && mpb->num_raid_devs) | |
7606 | continue; | |
7607 | ||
05501181 | 7608 | e = get_extents(super, dl, 0); |
efb30e7f DW |
7609 | if (!e) |
7610 | continue; | |
7611 | for (i = 1; e[i-1].size; i++) | |
7612 | ; | |
7613 | dl->e = e; | |
7614 | dl->extent_cnt = i; | |
efb30e7f DW |
7615 | cnt++; |
7616 | } | |
7617 | ||
aa19fdd4 | 7618 | maxsize = merge_extents(super, expanding); |
5f027b93 MT |
7619 | if (maxsize < minsize) { |
7620 | pr_err("imsm: Free space is %llu but must be equal or larger than %llu.\n", | |
7621 | maxsize, minsize); | |
7622 | return IMSM_STATUS_ERROR; | |
7623 | } | |
efb30e7f | 7624 | |
5f027b93 MT |
7625 | if (cnt < raiddisks || (super->orom && used && used != raiddisks)) { |
7626 | pr_err("imsm: Not enough devices with space to create array.\n"); | |
6d4d9ab2 | 7627 | return IMSM_STATUS_ERROR; |
efb30e7f DW |
7628 | } |
7629 | ||
7630 | if (size == 0) { | |
7631 | size = maxsize; | |
7632 | if (chunk) { | |
612e59d8 CA |
7633 | size /= 2 * chunk; |
7634 | size *= 2 * chunk; | |
efb30e7f | 7635 | } |
f878b242 LM |
7636 | maxsize = size; |
7637 | } | |
1a1ced1e KS |
7638 | if (mpb->num_raid_devs > 0 && size && size != maxsize) |
7639 | pr_err("attempting to create a second volume with size less then remaining space.\n"); | |
efb30e7f DW |
7640 | *freesize = size; |
7641 | ||
13bcac90 AK |
7642 | dprintf("imsm: imsm_get_free_size() returns : %llu\n", size); |
7643 | ||
6d4d9ab2 | 7644 | return IMSM_STATUS_OK; |
efb30e7f DW |
7645 | } |
7646 | ||
6d4d9ab2 MT |
7647 | /** |
7648 | * autolayout_imsm() - automatically layout a new volume. | |
7649 | * @super: &intel_super pointer, not NULL. | |
7650 | * @raiddisks: number of raid disks. | |
7651 | * @size: requested size, could be 0 (means max size). | |
7652 | * @chunk: requested chunk. | |
7653 | * @freesize: pointer for returned size value. | |
7654 | * | |
7655 | * We are being asked to automatically layout a new volume based on the current | |
7656 | * contents of the container. If the parameters can be satisfied autolayout_imsm | |
7657 | * will record the disks, start offset, and will return size of the volume to | |
7658 | * be created. See imsm_get_free_size() for details. | |
7659 | * add_to_super() and getinfo_super() detect when autolayout is in progress. | |
7660 | * If first volume exists, slots are set consistently to it. | |
7661 | * | |
7662 | * Return: &IMSM_STATUS_OK on success, &IMSM_STATUS_ERROR otherwise. | |
7663 | * | |
7664 | * Disks are marked for creation via dl->raiddisk. | |
7665 | */ | |
7666 | static imsm_status_t autolayout_imsm(struct intel_super *super, | |
7667 | const int raiddisks, | |
7668 | unsigned long long size, const int chunk, | |
7669 | unsigned long long *freesize) | |
13bcac90 | 7670 | { |
6d4d9ab2 MT |
7671 | int curr_slot = 0; |
7672 | struct dl *disk; | |
7673 | int vol_cnt = super->anchor->num_raid_devs; | |
7674 | imsm_status_t rv; | |
13bcac90 | 7675 | |
aa19fdd4 | 7676 | rv = imsm_get_free_size(super, raiddisks, size, chunk, freesize, false); |
6d4d9ab2 MT |
7677 | if (rv != IMSM_STATUS_OK) |
7678 | return IMSM_STATUS_ERROR; | |
7679 | ||
7680 | for (disk = super->disks; disk; disk = disk->next) { | |
7681 | if (!disk->e) | |
7682 | continue; | |
7683 | ||
7684 | if (curr_slot == raiddisks) | |
7685 | break; | |
7686 | ||
7687 | if (vol_cnt == 0) { | |
7688 | disk->raiddisk = curr_slot; | |
7689 | } else { | |
7690 | int _slot = get_disk_slot_in_dev(super, 0, disk->index); | |
7691 | ||
7692 | if (_slot == -1) { | |
7693 | pr_err("Disk %s is not used in first volume, aborting\n", | |
7694 | disk->devname); | |
7695 | return IMSM_STATUS_ERROR; | |
7696 | } | |
7697 | disk->raiddisk = _slot; | |
7698 | } | |
7699 | curr_slot++; | |
13bcac90 AK |
7700 | } |
7701 | ||
6d4d9ab2 | 7702 | return IMSM_STATUS_OK; |
13bcac90 AK |
7703 | } |
7704 | ||
bf5a934a | 7705 | static int validate_geometry_imsm(struct supertype *st, int level, int layout, |
c21e737b | 7706 | int raiddisks, int *chunk, unsigned long long size, |
af4348dd | 7707 | unsigned long long data_offset, |
bf5a934a | 7708 | char *dev, unsigned long long *freesize, |
5308f117 | 7709 | int consistency_policy, int verbose) |
bf5a934a DW |
7710 | { |
7711 | int fd, cfd; | |
7712 | struct mdinfo *sra; | |
20cbe8d2 | 7713 | int is_member = 0; |
bf5a934a | 7714 | |
d54559f0 LM |
7715 | /* load capability |
7716 | * if given unused devices create a container | |
bf5a934a DW |
7717 | * if given given devices in a container create a member volume |
7718 | */ | |
6f2af6a4 | 7719 | if (is_container(level)) |
bf5a934a | 7720 | /* Must be a fresh device to add to a container */ |
1f5d54a0 MT |
7721 | return validate_geometry_imsm_container(st, level, raiddisks, |
7722 | data_offset, dev, | |
7723 | freesize, verbose); | |
9587c373 | 7724 | |
06a6101c BK |
7725 | /* |
7726 | * Size is given in sectors. | |
7727 | */ | |
7728 | if (size && (size < 2048)) { | |
22dc741f | 7729 | pr_err("Given size must be greater than 1M.\n"); |
54865c30 RS |
7730 | /* Depends on algorithm in Create.c : |
7731 | * if container was given (dev == NULL) return -1, | |
7732 | * if block device was given ( dev != NULL) return 0. | |
7733 | */ | |
7734 | return dev ? -1 : 0; | |
7735 | } | |
7736 | ||
8592f29d | 7737 | if (!dev) { |
6d4d9ab2 MT |
7738 | struct intel_super *super = st->sb; |
7739 | ||
7740 | /* | |
071f839e | 7741 | * Autolayout mode, st->sb must be set. |
6d4d9ab2 | 7742 | */ |
071f839e KT |
7743 | if (!super) { |
7744 | pr_vrb("superblock must be set for autolayout, aborting\n"); | |
7745 | return 0; | |
6d4d9ab2 MT |
7746 | } |
7747 | ||
7748 | if (!validate_geometry_imsm_orom(st->sb, level, layout, | |
7749 | raiddisks, chunk, size, | |
7750 | verbose)) | |
7751 | return 0; | |
7752 | ||
071f839e | 7753 | if (super->orom && freesize) { |
6d4d9ab2 MT |
7754 | imsm_status_t rv; |
7755 | int count = count_volumes(super->hba, super->orom->dpa, | |
7756 | verbose); | |
7757 | if (super->orom->vphba <= count) { | |
7758 | pr_vrb("platform does not support more than %d raid volumes.\n", | |
7759 | super->orom->vphba); | |
e91a3bad | 7760 | return 0; |
ca9de185 | 7761 | } |
6d4d9ab2 MT |
7762 | |
7763 | rv = autolayout_imsm(super, raiddisks, size, *chunk, | |
7764 | freesize); | |
7765 | if (rv != IMSM_STATUS_OK) | |
7766 | return 0; | |
8592f29d N |
7767 | } |
7768 | return 1; | |
7769 | } | |
bf5a934a DW |
7770 | if (st->sb) { |
7771 | /* creating in a given container */ | |
7772 | return validate_geometry_imsm_volume(st, level, layout, | |
7773 | raiddisks, chunk, size, | |
af4348dd | 7774 | data_offset, |
bf5a934a DW |
7775 | dev, freesize, verbose); |
7776 | } | |
7777 | ||
bf5a934a DW |
7778 | /* This device needs to be a device in an 'imsm' container */ |
7779 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
4389ce73 MT |
7780 | |
7781 | if (is_fd_valid(fd)) { | |
7782 | pr_vrb("Cannot create this array on device %s\n", dev); | |
bf5a934a DW |
7783 | close(fd); |
7784 | return 0; | |
7785 | } | |
4389ce73 MT |
7786 | if (errno == EBUSY) |
7787 | fd = open(dev, O_RDONLY, 0); | |
7788 | ||
7789 | if (!is_fd_valid(fd)) { | |
7790 | pr_vrb("Cannot open %s: %s\n", dev, strerror(errno)); | |
bf5a934a DW |
7791 | return 0; |
7792 | } | |
4389ce73 | 7793 | |
bf5a934a DW |
7794 | /* Well, it is in use by someone, maybe an 'imsm' container. */ |
7795 | cfd = open_container(fd); | |
4389ce73 MT |
7796 | close_fd(&fd); |
7797 | ||
7798 | if (!is_fd_valid(cfd)) { | |
7799 | pr_vrb("Cannot use %s: It is busy\n", dev); | |
bf5a934a DW |
7800 | return 0; |
7801 | } | |
4dd2df09 | 7802 | sra = sysfs_read(cfd, NULL, GET_VERSION); |
bf5a934a | 7803 | if (sra && sra->array.major_version == -1 && |
20cbe8d2 AW |
7804 | strcmp(sra->text_version, "imsm") == 0) |
7805 | is_member = 1; | |
7806 | sysfs_free(sra); | |
7807 | if (is_member) { | |
bf5a934a DW |
7808 | /* This is a member of a imsm container. Load the container |
7809 | * and try to create a volume | |
7810 | */ | |
7811 | struct intel_super *super; | |
7812 | ||
ec50f7b6 | 7813 | if (load_super_imsm_all(st, cfd, (void **) &super, NULL, NULL, 1) == 0) { |
bf5a934a | 7814 | st->sb = super; |
4dd2df09 | 7815 | strcpy(st->container_devnm, fd2devnm(cfd)); |
bf5a934a DW |
7816 | close(cfd); |
7817 | return validate_geometry_imsm_volume(st, level, layout, | |
7818 | raiddisks, chunk, | |
af4348dd | 7819 | size, data_offset, dev, |
ecbd9e81 N |
7820 | freesize, 1) |
7821 | ? 1 : -1; | |
bf5a934a | 7822 | } |
20cbe8d2 | 7823 | } |
bf5a934a | 7824 | |
20cbe8d2 | 7825 | if (verbose) |
e7b84f9d | 7826 | pr_err("failed container membership check\n"); |
20cbe8d2 AW |
7827 | |
7828 | close(cfd); | |
7829 | return 0; | |
bf5a934a | 7830 | } |
0bd16cf2 | 7831 | |
30f58b22 | 7832 | static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk) |
0bd16cf2 DJ |
7833 | { |
7834 | struct intel_super *super = st->sb; | |
7835 | ||
30f58b22 DW |
7836 | if (level && *level == UnSet) |
7837 | *level = LEVEL_CONTAINER; | |
7838 | ||
7839 | if (level && layout && *layout == UnSet) | |
7840 | *layout = imsm_level_to_layout(*level); | |
0bd16cf2 | 7841 | |
cd9d1ac7 DW |
7842 | if (chunk && (*chunk == UnSet || *chunk == 0)) |
7843 | *chunk = imsm_default_chunk(super->orom); | |
0bd16cf2 DJ |
7844 | } |
7845 | ||
33414a01 DW |
7846 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev); |
7847 | ||
3364781b | 7848 | static int kill_subarray_imsm(struct supertype *st, char *subarray_id) |
33414a01 | 7849 | { |
3364781b | 7850 | /* remove the subarray currently referenced by subarray_id */ |
33414a01 DW |
7851 | __u8 i; |
7852 | struct intel_dev **dp; | |
7853 | struct intel_super *super = st->sb; | |
3364781b | 7854 | __u8 current_vol = strtoul(subarray_id, NULL, 10); |
33414a01 DW |
7855 | struct imsm_super *mpb = super->anchor; |
7856 | ||
3364781b | 7857 | if (mpb->num_raid_devs == 0) |
33414a01 | 7858 | return 2; |
33414a01 DW |
7859 | |
7860 | /* block deletions that would change the uuid of active subarrays | |
7861 | * | |
7862 | * FIXME when immutable ids are available, but note that we'll | |
7863 | * also need to fixup the invalidated/active subarray indexes in | |
7864 | * mdstat | |
7865 | */ | |
7866 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
7867 | char subarray[4]; | |
7868 | ||
7869 | if (i < current_vol) | |
7870 | continue; | |
dd5ab402 | 7871 | snprintf(subarray, sizeof(subarray), "%u", i); |
4dd2df09 | 7872 | if (is_subarray_active(subarray, st->devnm)) { |
e7b84f9d N |
7873 | pr_err("deleting subarray-%d would change the UUID of active subarray-%d, aborting\n", |
7874 | current_vol, i); | |
33414a01 DW |
7875 | |
7876 | return 2; | |
7877 | } | |
7878 | } | |
7879 | ||
7880 | if (st->update_tail) { | |
503975b9 | 7881 | struct imsm_update_kill_array *u = xmalloc(sizeof(*u)); |
33414a01 | 7882 | |
33414a01 DW |
7883 | u->type = update_kill_array; |
7884 | u->dev_idx = current_vol; | |
7885 | append_metadata_update(st, u, sizeof(*u)); | |
7886 | ||
7887 | return 0; | |
7888 | } | |
7889 | ||
7890 | for (dp = &super->devlist; *dp;) | |
7891 | if ((*dp)->index == current_vol) { | |
7892 | *dp = (*dp)->next; | |
7893 | } else { | |
7894 | handle_missing(super, (*dp)->dev); | |
7895 | if ((*dp)->index > current_vol) | |
7896 | (*dp)->index--; | |
7897 | dp = &(*dp)->next; | |
7898 | } | |
7899 | ||
7900 | /* no more raid devices, all active components are now spares, | |
7901 | * but of course failed are still failed | |
7902 | */ | |
7903 | if (--mpb->num_raid_devs == 0) { | |
7904 | struct dl *d; | |
7905 | ||
7906 | for (d = super->disks; d; d = d->next) | |
a8619d23 AK |
7907 | if (d->index > -2) |
7908 | mark_spare(d); | |
33414a01 DW |
7909 | } |
7910 | ||
7911 | super->updates_pending++; | |
7912 | ||
7913 | return 0; | |
7914 | } | |
aa534678 | 7915 | |
4345e135 MK |
7916 | /** |
7917 | * get_rwh_policy_from_update() - Get the rwh policy for update option. | |
7918 | * @update: Update option. | |
7919 | */ | |
7920 | static int get_rwh_policy_from_update(enum update_opt update) | |
19ad203e | 7921 | { |
4345e135 MK |
7922 | switch (update) { |
7923 | case UOPT_PPL: | |
19ad203e | 7924 | return RWH_MULTIPLE_DISTRIBUTED; |
4345e135 | 7925 | case UOPT_NO_PPL: |
19ad203e | 7926 | return RWH_MULTIPLE_OFF; |
4345e135 | 7927 | case UOPT_BITMAP: |
19ad203e | 7928 | return RWH_BITMAP; |
4345e135 | 7929 | case UOPT_NO_BITMAP: |
19ad203e | 7930 | return RWH_OFF; |
4345e135 MK |
7931 | default: |
7932 | break; | |
7933 | } | |
7934 | return UOPT_UNDEFINED; | |
19ad203e JR |
7935 | } |
7936 | ||
a951a4f7 | 7937 | static int update_subarray_imsm(struct supertype *st, char *subarray, |
03312b52 | 7938 | enum update_opt update, struct mddev_ident *ident) |
aa534678 DW |
7939 | { |
7940 | /* update the subarray currently referenced by ->current_vol */ | |
7941 | struct intel_super *super = st->sb; | |
7942 | struct imsm_super *mpb = super->anchor; | |
7943 | ||
03312b52 | 7944 | if (update == UOPT_NAME) { |
aa534678 | 7945 | char *name = ident->name; |
a951a4f7 N |
7946 | char *ep; |
7947 | int vol; | |
aa534678 | 7948 | |
e2eb503b | 7949 | if (imsm_is_name_allowed(super, name, 1) == false) |
aa534678 DW |
7950 | return 2; |
7951 | ||
a951a4f7 N |
7952 | vol = strtoul(subarray, &ep, 10); |
7953 | if (*ep != '\0' || vol >= super->anchor->num_raid_devs) | |
7954 | return 2; | |
7955 | ||
aa534678 | 7956 | if (st->update_tail) { |
503975b9 | 7957 | struct imsm_update_rename_array *u = xmalloc(sizeof(*u)); |
aa534678 | 7958 | |
aa534678 | 7959 | u->type = update_rename_array; |
a951a4f7 | 7960 | u->dev_idx = vol; |
618f4e6d XN |
7961 | strncpy((char *) u->name, name, MAX_RAID_SERIAL_LEN); |
7962 | u->name[MAX_RAID_SERIAL_LEN-1] = '\0'; | |
aa534678 DW |
7963 | append_metadata_update(st, u, sizeof(*u)); |
7964 | } else { | |
7965 | struct imsm_dev *dev; | |
ebad3af2 | 7966 | int i, namelen; |
aa534678 | 7967 | |
a951a4f7 | 7968 | dev = get_imsm_dev(super, vol); |
ebad3af2 JS |
7969 | memset(dev->volume, '\0', MAX_RAID_SERIAL_LEN); |
7970 | namelen = min((int)strlen(name), MAX_RAID_SERIAL_LEN); | |
7971 | memcpy(dev->volume, name, namelen); | |
aa534678 DW |
7972 | for (i = 0; i < mpb->num_raid_devs; i++) { |
7973 | dev = get_imsm_dev(super, i); | |
7974 | handle_missing(super, dev); | |
7975 | } | |
7976 | super->updates_pending++; | |
7977 | } | |
03312b52 | 7978 | } else if (get_rwh_policy_from_update(update) != UOPT_UNDEFINED) { |
e6e9dd3f AP |
7979 | int new_policy; |
7980 | char *ep; | |
7981 | int vol = strtoul(subarray, &ep, 10); | |
7982 | ||
7983 | if (*ep != '\0' || vol >= super->anchor->num_raid_devs) | |
7984 | return 2; | |
7985 | ||
03312b52 | 7986 | new_policy = get_rwh_policy_from_update(update); |
e6e9dd3f AP |
7987 | |
7988 | if (st->update_tail) { | |
7989 | struct imsm_update_rwh_policy *u = xmalloc(sizeof(*u)); | |
7990 | ||
7991 | u->type = update_rwh_policy; | |
7992 | u->dev_idx = vol; | |
7993 | u->new_policy = new_policy; | |
7994 | append_metadata_update(st, u, sizeof(*u)); | |
7995 | } else { | |
7996 | struct imsm_dev *dev; | |
7997 | ||
7998 | dev = get_imsm_dev(super, vol); | |
7999 | dev->rwh_policy = new_policy; | |
8000 | super->updates_pending++; | |
8001 | } | |
19ad203e JR |
8002 | if (new_policy == RWH_BITMAP) |
8003 | return write_init_bitmap_imsm_vol(st, vol); | |
aa534678 DW |
8004 | } else |
8005 | return 2; | |
8006 | ||
8007 | return 0; | |
8008 | } | |
bf5a934a | 8009 | |
195d1d76 | 8010 | static bool is_gen_migration(struct imsm_dev *dev) |
28bce06f | 8011 | { |
195d1d76 PP |
8012 | if (dev && dev->vol.migr_state && |
8013 | migr_type(dev) == MIGR_GEN_MIGR) | |
8014 | return true; | |
28bce06f | 8015 | |
195d1d76 | 8016 | return false; |
28bce06f AK |
8017 | } |
8018 | ||
1e5c6983 DW |
8019 | static int is_rebuilding(struct imsm_dev *dev) |
8020 | { | |
8021 | struct imsm_map *migr_map; | |
8022 | ||
8023 | if (!dev->vol.migr_state) | |
8024 | return 0; | |
8025 | ||
8026 | if (migr_type(dev) != MIGR_REBUILD) | |
8027 | return 0; | |
8028 | ||
238c0a71 | 8029 | migr_map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
8030 | |
8031 | if (migr_map->map_state == IMSM_T_STATE_DEGRADED) | |
8032 | return 1; | |
8033 | else | |
8034 | return 0; | |
8035 | } | |
8036 | ||
6ce1fbf1 AK |
8037 | static int is_initializing(struct imsm_dev *dev) |
8038 | { | |
8039 | struct imsm_map *migr_map; | |
8040 | ||
8041 | if (!dev->vol.migr_state) | |
8042 | return 0; | |
8043 | ||
8044 | if (migr_type(dev) != MIGR_INIT) | |
8045 | return 0; | |
8046 | ||
238c0a71 | 8047 | migr_map = get_imsm_map(dev, MAP_1); |
6ce1fbf1 AK |
8048 | |
8049 | if (migr_map->map_state == IMSM_T_STATE_UNINITIALIZED) | |
8050 | return 1; | |
8051 | ||
8052 | return 0; | |
6ce1fbf1 AK |
8053 | } |
8054 | ||
c47b0ff6 AK |
8055 | static void update_recovery_start(struct intel_super *super, |
8056 | struct imsm_dev *dev, | |
8057 | struct mdinfo *array) | |
1e5c6983 DW |
8058 | { |
8059 | struct mdinfo *rebuild = NULL; | |
8060 | struct mdinfo *d; | |
8061 | __u32 units; | |
8062 | ||
8063 | if (!is_rebuilding(dev)) | |
8064 | return; | |
8065 | ||
8066 | /* Find the rebuild target, but punt on the dual rebuild case */ | |
8067 | for (d = array->devs; d; d = d->next) | |
8068 | if (d->recovery_start == 0) { | |
8069 | if (rebuild) | |
8070 | return; | |
8071 | rebuild = d; | |
8072 | } | |
8073 | ||
4363fd80 DW |
8074 | if (!rebuild) { |
8075 | /* (?) none of the disks are marked with | |
8076 | * IMSM_ORD_REBUILD, so assume they are missing and the | |
8077 | * disk_ord_tbl was not correctly updated | |
8078 | */ | |
1ade5cc1 | 8079 | dprintf("failed to locate out-of-sync disk\n"); |
4363fd80 DW |
8080 | return; |
8081 | } | |
8082 | ||
4036e7ee | 8083 | units = vol_curr_migr_unit(dev); |
c47b0ff6 | 8084 | rebuild->recovery_start = units * blocks_per_migr_unit(super, dev); |
1e5c6983 DW |
8085 | } |
8086 | ||
276d77db | 8087 | static int recover_backup_imsm(struct supertype *st, struct mdinfo *info); |
1e5c6983 | 8088 | |
00bbdbda | 8089 | static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray) |
cdddbdbc | 8090 | { |
4f5bc454 DW |
8091 | /* Given a container loaded by load_super_imsm_all, |
8092 | * extract information about all the arrays into | |
8093 | * an mdinfo tree. | |
00bbdbda | 8094 | * If 'subarray' is given, just extract info about that array. |
4f5bc454 DW |
8095 | * |
8096 | * For each imsm_dev create an mdinfo, fill it in, | |
8097 | * then look for matching devices in super->disks | |
8098 | * and create appropriate device mdinfo. | |
8099 | */ | |
8100 | struct intel_super *super = st->sb; | |
949c47a0 | 8101 | struct imsm_super *mpb = super->anchor; |
4f5bc454 | 8102 | struct mdinfo *rest = NULL; |
00bbdbda | 8103 | unsigned int i; |
81219e70 | 8104 | int sb_errors = 0; |
abef11a3 AK |
8105 | struct dl *d; |
8106 | int spare_disks = 0; | |
b6180160 | 8107 | int current_vol = super->current_vol; |
cdddbdbc | 8108 | |
19482bcc AK |
8109 | /* do not assemble arrays when not all attributes are supported */ |
8110 | if (imsm_check_attributes(mpb->attributes) == 0) { | |
81219e70 | 8111 | sb_errors = 1; |
7a862a02 | 8112 | pr_err("Unsupported attributes in IMSM metadata.Arrays activation is blocked.\n"); |
19482bcc AK |
8113 | } |
8114 | ||
abef11a3 AK |
8115 | /* count spare devices, not used in maps |
8116 | */ | |
8117 | for (d = super->disks; d; d = d->next) | |
8118 | if (d->index == -1) | |
8119 | spare_disks++; | |
8120 | ||
4f5bc454 | 8121 | for (i = 0; i < mpb->num_raid_devs; i++) { |
00bbdbda N |
8122 | struct imsm_dev *dev; |
8123 | struct imsm_map *map; | |
86e3692b | 8124 | struct imsm_map *map2; |
4f5bc454 | 8125 | struct mdinfo *this; |
a6482415 | 8126 | int slot; |
a6482415 | 8127 | int chunk; |
00bbdbda | 8128 | char *ep; |
8b9cd157 | 8129 | int level; |
00bbdbda N |
8130 | |
8131 | if (subarray && | |
8132 | (i != strtoul(subarray, &ep, 10) || *ep != '\0')) | |
8133 | continue; | |
8134 | ||
8135 | dev = get_imsm_dev(super, i); | |
238c0a71 AK |
8136 | map = get_imsm_map(dev, MAP_0); |
8137 | map2 = get_imsm_map(dev, MAP_1); | |
8b9cd157 | 8138 | level = get_imsm_raid_level(map); |
4f5bc454 | 8139 | |
1ce0101c DW |
8140 | /* do not publish arrays that are in the middle of an |
8141 | * unsupported migration | |
8142 | */ | |
8143 | if (dev->vol.migr_state && | |
28bce06f | 8144 | (migr_type(dev) == MIGR_STATE_CHANGE)) { |
7a862a02 | 8145 | pr_err("cannot assemble volume '%.16s': unsupported migration in progress\n", |
1ce0101c DW |
8146 | dev->volume); |
8147 | continue; | |
8148 | } | |
2db86302 LM |
8149 | /* do not publish arrays that are not support by controller's |
8150 | * OROM/EFI | |
8151 | */ | |
1ce0101c | 8152 | |
503975b9 | 8153 | this = xmalloc(sizeof(*this)); |
4f5bc454 | 8154 | |
301406c9 | 8155 | super->current_vol = i; |
a5d85af7 | 8156 | getinfo_super_imsm_volume(st, this, NULL); |
9894ec0d | 8157 | this->next = rest; |
a6482415 | 8158 | chunk = __le16_to_cpu(map->blocks_per_strip) >> 1; |
81219e70 LM |
8159 | /* mdadm does not support all metadata features- set the bit in all arrays state */ |
8160 | if (!validate_geometry_imsm_orom(super, | |
8b9cd157 MK |
8161 | level, /* RAID level */ |
8162 | imsm_level_to_layout(level), | |
81219e70 | 8163 | map->num_members, /* raid disks */ |
fcc2c9da | 8164 | &chunk, imsm_dev_size(dev), |
81219e70 | 8165 | 1 /* verbose */)) { |
7a862a02 | 8166 | pr_err("IMSM RAID geometry validation failed. Array %s activation is blocked.\n", |
81219e70 LM |
8167 | dev->volume); |
8168 | this->array.state |= | |
8169 | (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) | | |
8170 | (1<<MD_SB_BLOCK_VOLUME); | |
8171 | } | |
81219e70 LM |
8172 | |
8173 | /* if array has bad blocks, set suitable bit in all arrays state */ | |
8174 | if (sb_errors) | |
8175 | this->array.state |= | |
8176 | (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) | | |
8177 | (1<<MD_SB_BLOCK_VOLUME); | |
8178 | ||
4f5bc454 | 8179 | for (slot = 0 ; slot < map->num_members; slot++) { |
1e5c6983 | 8180 | unsigned long long recovery_start; |
4f5bc454 DW |
8181 | struct mdinfo *info_d; |
8182 | struct dl *d; | |
8183 | int idx; | |
9a1608e5 | 8184 | int skip; |
7eef0453 | 8185 | __u32 ord; |
8b9cd157 | 8186 | int missing = 0; |
4f5bc454 | 8187 | |
9a1608e5 | 8188 | skip = 0; |
238c0a71 AK |
8189 | idx = get_imsm_disk_idx(dev, slot, MAP_0); |
8190 | ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); | |
4f5bc454 DW |
8191 | for (d = super->disks; d ; d = d->next) |
8192 | if (d->index == idx) | |
0fbd635c | 8193 | break; |
4f5bc454 | 8194 | |
1e5c6983 | 8195 | recovery_start = MaxSector; |
4f5bc454 | 8196 | if (d == NULL) |
9a1608e5 | 8197 | skip = 1; |
25ed7e59 | 8198 | if (d && is_failed(&d->disk)) |
9a1608e5 | 8199 | skip = 1; |
8b9cd157 | 8200 | if (!skip && (ord & IMSM_ORD_REBUILD)) |
1e5c6983 | 8201 | recovery_start = 0; |
1e93d0d1 BK |
8202 | if (!(ord & IMSM_ORD_REBUILD)) |
8203 | this->array.working_disks++; | |
1011e834 | 8204 | /* |
9a1608e5 | 8205 | * if we skip some disks the array will be assmebled degraded; |
1e5c6983 DW |
8206 | * reset resync start to avoid a dirty-degraded |
8207 | * situation when performing the intial sync | |
9a1608e5 | 8208 | */ |
8b9cd157 MK |
8209 | if (skip) |
8210 | missing++; | |
8211 | ||
8212 | if (!(dev->vol.dirty & RAIDVOL_DIRTY)) { | |
8213 | if ((!able_to_resync(level, missing) || | |
8214 | recovery_start == 0)) | |
8215 | this->resync_start = MaxSector; | |
8b9cd157 MK |
8216 | } |
8217 | ||
9a1608e5 DW |
8218 | if (skip) |
8219 | continue; | |
4f5bc454 | 8220 | |
503975b9 | 8221 | info_d = xcalloc(1, sizeof(*info_d)); |
4f5bc454 DW |
8222 | info_d->next = this->devs; |
8223 | this->devs = info_d; | |
8224 | ||
4f5bc454 DW |
8225 | info_d->disk.number = d->index; |
8226 | info_d->disk.major = d->major; | |
8227 | info_d->disk.minor = d->minor; | |
8228 | info_d->disk.raid_disk = slot; | |
1e5c6983 | 8229 | info_d->recovery_start = recovery_start; |
86e3692b AK |
8230 | if (map2) { |
8231 | if (slot < map2->num_members) | |
8232 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
8233 | else |
8234 | this->array.spare_disks++; | |
86e3692b AK |
8235 | } else { |
8236 | if (slot < map->num_members) | |
8237 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
8238 | else |
8239 | this->array.spare_disks++; | |
86e3692b | 8240 | } |
4f5bc454 DW |
8241 | |
8242 | info_d->events = __le32_to_cpu(mpb->generation_num); | |
5551b113 | 8243 | info_d->data_offset = pba_of_lba0(map); |
44490938 | 8244 | info_d->component_size = calc_component_size(map, dev); |
06fb291a PB |
8245 | |
8246 | if (map->raid_level == 5) { | |
2432ce9b AP |
8247 | info_d->ppl_sector = this->ppl_sector; |
8248 | info_d->ppl_size = this->ppl_size; | |
98e96bdb AP |
8249 | if (this->consistency_policy == CONSISTENCY_POLICY_PPL && |
8250 | recovery_start == 0) | |
8251 | this->resync_start = 0; | |
06fb291a | 8252 | } |
b12796be | 8253 | |
5e46202e | 8254 | info_d->bb.supported = 1; |
b12796be TM |
8255 | get_volume_badblocks(super->bbm_log, ord_to_idx(ord), |
8256 | info_d->data_offset, | |
8257 | info_d->component_size, | |
8258 | &info_d->bb); | |
4f5bc454 | 8259 | } |
1e5c6983 | 8260 | /* now that the disk list is up-to-date fixup recovery_start */ |
c47b0ff6 | 8261 | update_recovery_start(super, dev, this); |
abef11a3 | 8262 | this->array.spare_disks += spare_disks; |
276d77db AK |
8263 | |
8264 | /* check for reshape */ | |
8265 | if (this->reshape_active == 1) | |
8266 | recover_backup_imsm(st, this); | |
9a1608e5 | 8267 | rest = this; |
4f5bc454 DW |
8268 | } |
8269 | ||
b6180160 | 8270 | super->current_vol = current_vol; |
4f5bc454 | 8271 | return rest; |
cdddbdbc DW |
8272 | } |
8273 | ||
3b451610 AK |
8274 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, |
8275 | int failed, int look_in_map) | |
c2a1e7da | 8276 | { |
3b451610 AK |
8277 | struct imsm_map *map; |
8278 | ||
8279 | map = get_imsm_map(dev, look_in_map); | |
c2a1e7da DW |
8280 | |
8281 | if (!failed) | |
1011e834 | 8282 | return map->map_state == IMSM_T_STATE_UNINITIALIZED ? |
3393c6af | 8283 | IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL; |
c2a1e7da DW |
8284 | |
8285 | switch (get_imsm_raid_level(map)) { | |
8286 | case 0: | |
8287 | return IMSM_T_STATE_FAILED; | |
8288 | break; | |
8289 | case 1: | |
8290 | if (failed < map->num_members) | |
8291 | return IMSM_T_STATE_DEGRADED; | |
8292 | else | |
8293 | return IMSM_T_STATE_FAILED; | |
8294 | break; | |
8295 | case 10: | |
8296 | { | |
8297 | /** | |
c92a2527 DW |
8298 | * check to see if any mirrors have failed, otherwise we |
8299 | * are degraded. Even numbered slots are mirrored on | |
8300 | * slot+1 | |
c2a1e7da | 8301 | */ |
c2a1e7da | 8302 | int i; |
d9b420a5 N |
8303 | /* gcc -Os complains that this is unused */ |
8304 | int insync = insync; | |
c2a1e7da DW |
8305 | |
8306 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 8307 | __u32 ord = get_imsm_ord_tbl_ent(dev, i, MAP_X); |
c92a2527 DW |
8308 | int idx = ord_to_idx(ord); |
8309 | struct imsm_disk *disk; | |
c2a1e7da | 8310 | |
c92a2527 | 8311 | /* reset the potential in-sync count on even-numbered |
1011e834 | 8312 | * slots. num_copies is always 2 for imsm raid10 |
c92a2527 DW |
8313 | */ |
8314 | if ((i & 1) == 0) | |
8315 | insync = 2; | |
c2a1e7da | 8316 | |
c92a2527 | 8317 | disk = get_imsm_disk(super, idx); |
25ed7e59 | 8318 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) |
c92a2527 | 8319 | insync--; |
c2a1e7da | 8320 | |
c92a2527 DW |
8321 | /* no in-sync disks left in this mirror the |
8322 | * array has failed | |
8323 | */ | |
8324 | if (insync == 0) | |
8325 | return IMSM_T_STATE_FAILED; | |
c2a1e7da DW |
8326 | } |
8327 | ||
8328 | return IMSM_T_STATE_DEGRADED; | |
8329 | } | |
8330 | case 5: | |
8331 | if (failed < 2) | |
8332 | return IMSM_T_STATE_DEGRADED; | |
8333 | else | |
8334 | return IMSM_T_STATE_FAILED; | |
8335 | break; | |
8336 | default: | |
8337 | break; | |
8338 | } | |
8339 | ||
8340 | return map->map_state; | |
8341 | } | |
8342 | ||
3b451610 AK |
8343 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev, |
8344 | int look_in_map) | |
c2a1e7da DW |
8345 | { |
8346 | int i; | |
8347 | int failed = 0; | |
8348 | struct imsm_disk *disk; | |
d5985138 AK |
8349 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
8350 | struct imsm_map *prev = get_imsm_map(dev, MAP_1); | |
68fe4598 | 8351 | struct imsm_map *map_for_loop; |
0556e1a2 DW |
8352 | __u32 ord; |
8353 | int idx; | |
d5985138 | 8354 | int idx_1; |
c2a1e7da | 8355 | |
0556e1a2 DW |
8356 | /* at the beginning of migration we set IMSM_ORD_REBUILD on |
8357 | * disks that are being rebuilt. New failures are recorded to | |
8358 | * map[0]. So we look through all the disks we started with and | |
8359 | * see if any failures are still present, or if any new ones | |
8360 | * have arrived | |
0556e1a2 | 8361 | */ |
d5985138 AK |
8362 | map_for_loop = map; |
8363 | if (prev && (map->num_members < prev->num_members)) | |
8364 | map_for_loop = prev; | |
68fe4598 LD |
8365 | |
8366 | for (i = 0; i < map_for_loop->num_members; i++) { | |
d5985138 | 8367 | idx_1 = -255; |
238c0a71 AK |
8368 | /* when MAP_X is passed both maps failures are counted |
8369 | */ | |
d5985138 | 8370 | if (prev && |
089f9d79 JS |
8371 | (look_in_map == MAP_1 || look_in_map == MAP_X) && |
8372 | i < prev->num_members) { | |
d5985138 AK |
8373 | ord = __le32_to_cpu(prev->disk_ord_tbl[i]); |
8374 | idx_1 = ord_to_idx(ord); | |
c2a1e7da | 8375 | |
d5985138 AK |
8376 | disk = get_imsm_disk(super, idx_1); |
8377 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) | |
8378 | failed++; | |
8379 | } | |
089f9d79 JS |
8380 | if ((look_in_map == MAP_0 || look_in_map == MAP_X) && |
8381 | i < map->num_members) { | |
d5985138 AK |
8382 | ord = __le32_to_cpu(map->disk_ord_tbl[i]); |
8383 | idx = ord_to_idx(ord); | |
8384 | ||
8385 | if (idx != idx_1) { | |
8386 | disk = get_imsm_disk(super, idx); | |
8387 | if (!disk || is_failed(disk) || | |
8388 | ord & IMSM_ORD_REBUILD) | |
8389 | failed++; | |
8390 | } | |
8391 | } | |
c2a1e7da DW |
8392 | } |
8393 | ||
8394 | return failed; | |
845dea95 NB |
8395 | } |
8396 | ||
97b4d0e9 | 8397 | static int imsm_open_new(struct supertype *c, struct active_array *a, |
60815698 | 8398 | int inst) |
97b4d0e9 DW |
8399 | { |
8400 | struct intel_super *super = c->sb; | |
8401 | struct imsm_super *mpb = super->anchor; | |
bbab0940 | 8402 | struct imsm_update_prealloc_bb_mem u; |
9587c373 | 8403 | |
60815698 MG |
8404 | if (inst >= mpb->num_raid_devs) { |
8405 | pr_err("subarry index %d, out of range\n", inst); | |
97b4d0e9 DW |
8406 | return -ENODEV; |
8407 | } | |
8408 | ||
60815698 MG |
8409 | dprintf("imsm: open_new %d\n", inst); |
8410 | a->info.container_member = inst; | |
bbab0940 TM |
8411 | |
8412 | u.type = update_prealloc_badblocks_mem; | |
8413 | imsm_update_metadata_locally(c, &u, sizeof(u)); | |
8414 | ||
97b4d0e9 DW |
8415 | return 0; |
8416 | } | |
8417 | ||
0c046afd DW |
8418 | static int is_resyncing(struct imsm_dev *dev) |
8419 | { | |
8420 | struct imsm_map *migr_map; | |
8421 | ||
8422 | if (!dev->vol.migr_state) | |
8423 | return 0; | |
8424 | ||
1484e727 DW |
8425 | if (migr_type(dev) == MIGR_INIT || |
8426 | migr_type(dev) == MIGR_REPAIR) | |
0c046afd DW |
8427 | return 1; |
8428 | ||
4c9bc37b AK |
8429 | if (migr_type(dev) == MIGR_GEN_MIGR) |
8430 | return 0; | |
8431 | ||
238c0a71 | 8432 | migr_map = get_imsm_map(dev, MAP_1); |
0c046afd | 8433 | |
089f9d79 JS |
8434 | if (migr_map->map_state == IMSM_T_STATE_NORMAL && |
8435 | dev->vol.migr_type != MIGR_GEN_MIGR) | |
0c046afd DW |
8436 | return 1; |
8437 | else | |
8438 | return 0; | |
8439 | } | |
8440 | ||
0556e1a2 | 8441 | /* return true if we recorded new information */ |
4c9e8c1e TM |
8442 | static int mark_failure(struct intel_super *super, |
8443 | struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
47ee5a45 | 8444 | { |
0556e1a2 DW |
8445 | __u32 ord; |
8446 | int slot; | |
8447 | struct imsm_map *map; | |
86c54047 DW |
8448 | char buf[MAX_RAID_SERIAL_LEN+3]; |
8449 | unsigned int len, shift = 0; | |
0556e1a2 DW |
8450 | |
8451 | /* new failures are always set in map[0] */ | |
238c0a71 | 8452 | map = get_imsm_map(dev, MAP_0); |
0556e1a2 DW |
8453 | |
8454 | slot = get_imsm_disk_slot(map, idx); | |
8455 | if (slot < 0) | |
8456 | return 0; | |
8457 | ||
8458 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
25ed7e59 | 8459 | if (is_failed(disk) && (ord & IMSM_ORD_REBUILD)) |
0556e1a2 DW |
8460 | return 0; |
8461 | ||
7d0c5e24 LD |
8462 | memcpy(buf, disk->serial, MAX_RAID_SERIAL_LEN); |
8463 | buf[MAX_RAID_SERIAL_LEN] = '\000'; | |
8464 | strcat(buf, ":0"); | |
86c54047 DW |
8465 | if ((len = strlen(buf)) >= MAX_RAID_SERIAL_LEN) |
8466 | shift = len - MAX_RAID_SERIAL_LEN + 1; | |
167d8bb8 | 8467 | memcpy(disk->serial, &buf[shift], len + 1 - shift); |
86c54047 | 8468 | |
f2f27e63 | 8469 | disk->status |= FAILED_DISK; |
0556e1a2 | 8470 | set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD); |
17788994 AK |
8471 | /* mark failures in second map if second map exists and this disk |
8472 | * in this slot. | |
8473 | * This is valid for migration, initialization and rebuild | |
8474 | */ | |
8475 | if (dev->vol.migr_state) { | |
238c0a71 | 8476 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); |
0a108d63 AK |
8477 | int slot2 = get_imsm_disk_slot(map2, idx); |
8478 | ||
089f9d79 | 8479 | if (slot2 < map2->num_members && slot2 >= 0) |
0a108d63 | 8480 | set_imsm_ord_tbl_ent(map2, slot2, |
1ace8403 AK |
8481 | idx | IMSM_ORD_REBUILD); |
8482 | } | |
d7a1fda2 MT |
8483 | if (map->failed_disk_num == 0xff || |
8484 | (!is_rebuilding(dev) && map->failed_disk_num > slot)) | |
0556e1a2 | 8485 | map->failed_disk_num = slot; |
4c9e8c1e TM |
8486 | |
8487 | clear_disk_badblocks(super->bbm_log, ord_to_idx(ord)); | |
8488 | ||
0556e1a2 DW |
8489 | return 1; |
8490 | } | |
8491 | ||
4c9e8c1e TM |
8492 | static void mark_missing(struct intel_super *super, |
8493 | struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
0556e1a2 | 8494 | { |
4c9e8c1e | 8495 | mark_failure(super, dev, disk, idx); |
0556e1a2 DW |
8496 | |
8497 | if (disk->scsi_id == __cpu_to_le32(~(__u32)0)) | |
8498 | return; | |
8499 | ||
47ee5a45 DW |
8500 | disk->scsi_id = __cpu_to_le32(~(__u32)0); |
8501 | memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1); | |
8502 | } | |
8503 | ||
33414a01 DW |
8504 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev) |
8505 | { | |
33414a01 | 8506 | struct dl *dl; |
33414a01 DW |
8507 | |
8508 | if (!super->missing) | |
8509 | return; | |
33414a01 | 8510 | |
79b68f1b PC |
8511 | /* When orom adds replacement for missing disk it does |
8512 | * not remove entry of missing disk, but just updates map with | |
8513 | * new added disk. So it is not enough just to test if there is | |
8514 | * any missing disk, we have to look if there are any failed disks | |
8515 | * in map to stop migration */ | |
8516 | ||
33414a01 | 8517 | dprintf("imsm: mark missing\n"); |
3d59f0c0 AK |
8518 | /* end process for initialization and rebuild only |
8519 | */ | |
195d1d76 | 8520 | if (is_gen_migration(dev) == false) { |
fb12a745 | 8521 | int failed = imsm_count_failed(super, dev, MAP_0); |
3d59f0c0 | 8522 | |
fb12a745 TM |
8523 | if (failed) { |
8524 | __u8 map_state; | |
8525 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
8526 | struct imsm_map *map1; | |
8527 | int i, ord, ord_map1; | |
8528 | int rebuilt = 1; | |
3d59f0c0 | 8529 | |
fb12a745 TM |
8530 | for (i = 0; i < map->num_members; i++) { |
8531 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_0); | |
8532 | if (!(ord & IMSM_ORD_REBUILD)) | |
8533 | continue; | |
8534 | ||
8535 | map1 = get_imsm_map(dev, MAP_1); | |
8536 | if (!map1) | |
8537 | continue; | |
8538 | ||
8539 | ord_map1 = __le32_to_cpu(map1->disk_ord_tbl[i]); | |
8540 | if (ord_map1 & IMSM_ORD_REBUILD) | |
8541 | rebuilt = 0; | |
8542 | } | |
8543 | ||
8544 | if (rebuilt) { | |
8545 | map_state = imsm_check_degraded(super, dev, | |
8546 | failed, MAP_0); | |
8547 | end_migration(dev, super, map_state); | |
8548 | } | |
8549 | } | |
3d59f0c0 | 8550 | } |
33414a01 | 8551 | for (dl = super->missing; dl; dl = dl->next) |
4c9e8c1e | 8552 | mark_missing(super, dev, &dl->disk, dl->index); |
33414a01 DW |
8553 | super->updates_pending++; |
8554 | } | |
8555 | ||
f3871fdc AK |
8556 | static unsigned long long imsm_set_array_size(struct imsm_dev *dev, |
8557 | long long new_size) | |
70bdf0dc | 8558 | { |
70bdf0dc | 8559 | unsigned long long array_blocks; |
9529d343 MD |
8560 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
8561 | int used_disks = imsm_num_data_members(map); | |
70bdf0dc AK |
8562 | |
8563 | if (used_disks == 0) { | |
8564 | /* when problems occures | |
8565 | * return current array_blocks value | |
8566 | */ | |
fcc2c9da | 8567 | array_blocks = imsm_dev_size(dev); |
70bdf0dc AK |
8568 | |
8569 | return array_blocks; | |
8570 | } | |
8571 | ||
8572 | /* set array size in metadata | |
8573 | */ | |
9529d343 | 8574 | if (new_size <= 0) |
f3871fdc AK |
8575 | /* OLCE size change is caused by added disks |
8576 | */ | |
44490938 | 8577 | array_blocks = per_dev_array_size(map) * used_disks; |
9529d343 | 8578 | else |
f3871fdc AK |
8579 | /* Online Volume Size Change |
8580 | * Using available free space | |
8581 | */ | |
8582 | array_blocks = new_size; | |
70bdf0dc | 8583 | |
b53bfba6 | 8584 | array_blocks = round_size_to_mb(array_blocks, used_disks); |
fcc2c9da | 8585 | set_imsm_dev_size(dev, array_blocks); |
70bdf0dc AK |
8586 | |
8587 | return array_blocks; | |
8588 | } | |
8589 | ||
28bce06f AK |
8590 | static void imsm_set_disk(struct active_array *a, int n, int state); |
8591 | ||
0e2d1a4e AK |
8592 | static void imsm_progress_container_reshape(struct intel_super *super) |
8593 | { | |
8594 | /* if no device has a migr_state, but some device has a | |
8595 | * different number of members than the previous device, start | |
8596 | * changing the number of devices in this device to match | |
8597 | * previous. | |
8598 | */ | |
8599 | struct imsm_super *mpb = super->anchor; | |
8600 | int prev_disks = -1; | |
8601 | int i; | |
1dfaa380 | 8602 | int copy_map_size; |
0e2d1a4e AK |
8603 | |
8604 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
8605 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 8606 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
0e2d1a4e AK |
8607 | struct imsm_map *map2; |
8608 | int prev_num_members; | |
0e2d1a4e AK |
8609 | |
8610 | if (dev->vol.migr_state) | |
8611 | return; | |
8612 | ||
8613 | if (prev_disks == -1) | |
8614 | prev_disks = map->num_members; | |
8615 | if (prev_disks == map->num_members) | |
8616 | continue; | |
8617 | ||
8618 | /* OK, this array needs to enter reshape mode. | |
8619 | * i.e it needs a migr_state | |
8620 | */ | |
8621 | ||
1dfaa380 | 8622 | copy_map_size = sizeof_imsm_map(map); |
0e2d1a4e AK |
8623 | prev_num_members = map->num_members; |
8624 | map->num_members = prev_disks; | |
8625 | dev->vol.migr_state = 1; | |
4036e7ee | 8626 | set_vol_curr_migr_unit(dev, 0); |
ea672ee1 | 8627 | set_migr_type(dev, MIGR_GEN_MIGR); |
0e2d1a4e AK |
8628 | for (i = prev_num_members; |
8629 | i < map->num_members; i++) | |
8630 | set_imsm_ord_tbl_ent(map, i, i); | |
238c0a71 | 8631 | map2 = get_imsm_map(dev, MAP_1); |
0e2d1a4e | 8632 | /* Copy the current map */ |
1dfaa380 | 8633 | memcpy(map2, map, copy_map_size); |
0e2d1a4e AK |
8634 | map2->num_members = prev_num_members; |
8635 | ||
f3871fdc | 8636 | imsm_set_array_size(dev, -1); |
51d83f5d | 8637 | super->clean_migration_record_by_mdmon = 1; |
0e2d1a4e AK |
8638 | super->updates_pending++; |
8639 | } | |
8640 | } | |
8641 | ||
aad6f216 | 8642 | /* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild |
0c046afd DW |
8643 | * states are handled in imsm_set_disk() with one exception, when a |
8644 | * resync is stopped due to a new failure this routine will set the | |
8645 | * 'degraded' state for the array. | |
8646 | */ | |
01f157d7 | 8647 | static int imsm_set_array_state(struct active_array *a, int consistent) |
a862209d DW |
8648 | { |
8649 | int inst = a->info.container_member; | |
8650 | struct intel_super *super = a->container->sb; | |
949c47a0 | 8651 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 8652 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
3b451610 AK |
8653 | int failed = imsm_count_failed(super, dev, MAP_0); |
8654 | __u8 map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
1e5c6983 | 8655 | __u32 blocks_per_unit; |
a862209d | 8656 | |
1af97990 AK |
8657 | if (dev->vol.migr_state && |
8658 | dev->vol.migr_type == MIGR_GEN_MIGR) { | |
8659 | /* array state change is blocked due to reshape action | |
aad6f216 N |
8660 | * We might need to |
8661 | * - abort the reshape (if last_checkpoint is 0 and action!= reshape) | |
8662 | * - finish the reshape (if last_checkpoint is big and action != reshape) | |
4036e7ee | 8663 | * - update vol_curr_migr_unit |
1af97990 | 8664 | */ |
aad6f216 | 8665 | if (a->curr_action == reshape) { |
4036e7ee | 8666 | /* still reshaping, maybe update vol_curr_migr_unit */ |
633b5610 | 8667 | goto mark_checkpoint; |
aad6f216 | 8668 | } else { |
aad6f216 N |
8669 | if (a->last_checkpoint >= a->info.component_size) { |
8670 | unsigned long long array_blocks; | |
8671 | int used_disks; | |
e154ced3 | 8672 | struct mdinfo *mdi; |
aad6f216 | 8673 | |
9529d343 | 8674 | used_disks = imsm_num_data_members(map); |
d55adef9 AK |
8675 | if (used_disks > 0) { |
8676 | array_blocks = | |
44490938 | 8677 | per_dev_array_size(map) * |
d55adef9 | 8678 | used_disks; |
b53bfba6 TM |
8679 | array_blocks = |
8680 | round_size_to_mb(array_blocks, | |
8681 | used_disks); | |
d55adef9 AK |
8682 | a->info.custom_array_size = array_blocks; |
8683 | /* encourage manager to update array | |
8684 | * size | |
8685 | */ | |
e154ced3 | 8686 | |
d55adef9 | 8687 | a->check_reshape = 1; |
633b5610 | 8688 | } |
e154ced3 AK |
8689 | /* finalize online capacity expansion/reshape */ |
8690 | for (mdi = a->info.devs; mdi; mdi = mdi->next) | |
8691 | imsm_set_disk(a, | |
8692 | mdi->disk.raid_disk, | |
8693 | mdi->curr_state); | |
8694 | ||
0e2d1a4e | 8695 | imsm_progress_container_reshape(super); |
e154ced3 | 8696 | } |
aad6f216 | 8697 | } |
1af97990 AK |
8698 | } |
8699 | ||
47ee5a45 | 8700 | /* before we activate this array handle any missing disks */ |
33414a01 DW |
8701 | if (consistent == 2) |
8702 | handle_missing(super, dev); | |
1e5c6983 | 8703 | |
0c046afd | 8704 | if (consistent == 2 && |
b7941fd6 | 8705 | (!is_resync_complete(&a->info) || |
0c046afd DW |
8706 | map_state != IMSM_T_STATE_NORMAL || |
8707 | dev->vol.migr_state)) | |
01f157d7 | 8708 | consistent = 0; |
272906ef | 8709 | |
b7941fd6 | 8710 | if (is_resync_complete(&a->info)) { |
0c046afd | 8711 | /* complete intialization / resync, |
0556e1a2 DW |
8712 | * recovery and interrupted recovery is completed in |
8713 | * ->set_disk | |
0c046afd DW |
8714 | */ |
8715 | if (is_resyncing(dev)) { | |
8716 | dprintf("imsm: mark resync done\n"); | |
809da78e | 8717 | end_migration(dev, super, map_state); |
115c3803 | 8718 | super->updates_pending++; |
484240d8 | 8719 | a->last_checkpoint = 0; |
115c3803 | 8720 | } |
b9172665 AK |
8721 | } else if ((!is_resyncing(dev) && !failed) && |
8722 | (imsm_reshape_blocks_arrays_changes(super) == 0)) { | |
0c046afd | 8723 | /* mark the start of the init process if nothing is failed */ |
b7941fd6 | 8724 | dprintf("imsm: mark resync start\n"); |
1484e727 | 8725 | if (map->map_state == IMSM_T_STATE_UNINITIALIZED) |
8e59f3d8 | 8726 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_INIT); |
1484e727 | 8727 | else |
8e59f3d8 | 8728 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_REPAIR); |
3393c6af | 8729 | super->updates_pending++; |
115c3803 | 8730 | } |
a862209d | 8731 | |
fdb7e802 MK |
8732 | if (a->prev_action == idle) |
8733 | goto skip_mark_checkpoint; | |
8734 | ||
633b5610 | 8735 | mark_checkpoint: |
5b83bacf AK |
8736 | /* skip checkpointing for general migration, |
8737 | * it is controlled in mdadm | |
8738 | */ | |
8739 | if (is_gen_migration(dev)) | |
8740 | goto skip_mark_checkpoint; | |
8741 | ||
4036e7ee MT |
8742 | /* check if we can update vol_curr_migr_unit from resync_start, |
8743 | * recovery_start | |
8744 | */ | |
c47b0ff6 | 8745 | blocks_per_unit = blocks_per_migr_unit(super, dev); |
4f0a7acc | 8746 | if (blocks_per_unit) { |
4036e7ee MT |
8747 | set_vol_curr_migr_unit(dev, |
8748 | a->last_checkpoint / blocks_per_unit); | |
8749 | dprintf("imsm: mark checkpoint (%llu)\n", | |
8750 | vol_curr_migr_unit(dev)); | |
8751 | super->updates_pending++; | |
1e5c6983 | 8752 | } |
f8f603f1 | 8753 | |
5b83bacf | 8754 | skip_mark_checkpoint: |
3393c6af | 8755 | /* mark dirty / clean */ |
2432ce9b AP |
8756 | if (((dev->vol.dirty & RAIDVOL_DIRTY) && consistent) || |
8757 | (!(dev->vol.dirty & RAIDVOL_DIRTY) && !consistent)) { | |
b7941fd6 | 8758 | dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty"); |
2432ce9b AP |
8759 | if (consistent) { |
8760 | dev->vol.dirty = RAIDVOL_CLEAN; | |
8761 | } else { | |
8762 | dev->vol.dirty = RAIDVOL_DIRTY; | |
c2462068 PB |
8763 | if (dev->rwh_policy == RWH_DISTRIBUTED || |
8764 | dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED) | |
2432ce9b AP |
8765 | dev->vol.dirty |= RAIDVOL_DSRECORD_VALID; |
8766 | } | |
a862209d DW |
8767 | super->updates_pending++; |
8768 | } | |
28bce06f | 8769 | |
01f157d7 | 8770 | return consistent; |
a862209d DW |
8771 | } |
8772 | ||
6f50473f TM |
8773 | static int imsm_disk_slot_to_ord(struct active_array *a, int slot) |
8774 | { | |
8775 | int inst = a->info.container_member; | |
8776 | struct intel_super *super = a->container->sb; | |
8777 | struct imsm_dev *dev = get_imsm_dev(super, inst); | |
8778 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
8779 | ||
8780 | if (slot > map->num_members) { | |
8781 | pr_err("imsm: imsm_disk_slot_to_ord %d out of range 0..%d\n", | |
8782 | slot, map->num_members - 1); | |
8783 | return -1; | |
8784 | } | |
8785 | ||
8786 | if (slot < 0) | |
8787 | return -1; | |
8788 | ||
8789 | return get_imsm_ord_tbl_ent(dev, slot, MAP_0); | |
8790 | } | |
8791 | ||
8d45d196 | 8792 | static void imsm_set_disk(struct active_array *a, int n, int state) |
845dea95 | 8793 | { |
8d45d196 DW |
8794 | int inst = a->info.container_member; |
8795 | struct intel_super *super = a->container->sb; | |
949c47a0 | 8796 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 8797 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
8d45d196 | 8798 | struct imsm_disk *disk; |
7ce05701 LD |
8799 | struct mdinfo *mdi; |
8800 | int recovery_not_finished = 0; | |
0c046afd | 8801 | int failed; |
6f50473f | 8802 | int ord; |
0c046afd | 8803 | __u8 map_state; |
fb12a745 TM |
8804 | int rebuild_done = 0; |
8805 | int i; | |
8d45d196 | 8806 | |
fb12a745 | 8807 | ord = get_imsm_ord_tbl_ent(dev, n, MAP_X); |
6f50473f | 8808 | if (ord < 0) |
8d45d196 DW |
8809 | return; |
8810 | ||
4e6e574a | 8811 | dprintf("imsm: set_disk %d:%x\n", n, state); |
b10b37b8 | 8812 | disk = get_imsm_disk(super, ord_to_idx(ord)); |
8d45d196 | 8813 | |
5802a811 | 8814 | /* check for new failures */ |
ae7d61e3 | 8815 | if (disk && (state & DS_FAULTY)) { |
4c9e8c1e | 8816 | if (mark_failure(super, dev, disk, ord_to_idx(ord))) |
0556e1a2 | 8817 | super->updates_pending++; |
8d45d196 | 8818 | } |
47ee5a45 | 8819 | |
19859edc | 8820 | /* check if in_sync */ |
0556e1a2 | 8821 | if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) { |
238c0a71 | 8822 | struct imsm_map *migr_map = get_imsm_map(dev, MAP_1); |
b10b37b8 DW |
8823 | |
8824 | set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord)); | |
fb12a745 | 8825 | rebuild_done = 1; |
19859edc DW |
8826 | super->updates_pending++; |
8827 | } | |
8d45d196 | 8828 | |
3b451610 AK |
8829 | failed = imsm_count_failed(super, dev, MAP_0); |
8830 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
5802a811 | 8831 | |
0c046afd | 8832 | /* check if recovery complete, newly degraded, or failed */ |
94002678 AK |
8833 | dprintf("imsm: Detected transition to state "); |
8834 | switch (map_state) { | |
8835 | case IMSM_T_STATE_NORMAL: /* transition to normal state */ | |
8836 | dprintf("normal: "); | |
8837 | if (is_rebuilding(dev)) { | |
1ade5cc1 | 8838 | dprintf_cont("while rebuilding"); |
7ce05701 LD |
8839 | /* check if recovery is really finished */ |
8840 | for (mdi = a->info.devs; mdi ; mdi = mdi->next) | |
8841 | if (mdi->recovery_start != MaxSector) { | |
8842 | recovery_not_finished = 1; | |
8843 | break; | |
8844 | } | |
8845 | if (recovery_not_finished) { | |
1ade5cc1 N |
8846 | dprintf_cont("\n"); |
8847 | dprintf("Rebuild has not finished yet, state not changed"); | |
7ce05701 LD |
8848 | if (a->last_checkpoint < mdi->recovery_start) { |
8849 | a->last_checkpoint = mdi->recovery_start; | |
8850 | super->updates_pending++; | |
8851 | } | |
8852 | break; | |
8853 | } | |
94002678 | 8854 | end_migration(dev, super, map_state); |
94002678 AK |
8855 | map->failed_disk_num = ~0; |
8856 | super->updates_pending++; | |
8857 | a->last_checkpoint = 0; | |
8858 | break; | |
8859 | } | |
8860 | if (is_gen_migration(dev)) { | |
1ade5cc1 | 8861 | dprintf_cont("while general migration"); |
bf2f0071 | 8862 | if (a->last_checkpoint >= a->info.component_size) |
809da78e | 8863 | end_migration(dev, super, map_state); |
94002678 AK |
8864 | else |
8865 | map->map_state = map_state; | |
28bce06f | 8866 | map->failed_disk_num = ~0; |
94002678 | 8867 | super->updates_pending++; |
bf2f0071 | 8868 | break; |
94002678 AK |
8869 | } |
8870 | break; | |
8871 | case IMSM_T_STATE_DEGRADED: /* transition to degraded state */ | |
1ade5cc1 | 8872 | dprintf_cont("degraded: "); |
089f9d79 | 8873 | if (map->map_state != map_state && !dev->vol.migr_state) { |
1ade5cc1 | 8874 | dprintf_cont("mark degraded"); |
94002678 AK |
8875 | map->map_state = map_state; |
8876 | super->updates_pending++; | |
8877 | a->last_checkpoint = 0; | |
8878 | break; | |
8879 | } | |
8880 | if (is_rebuilding(dev)) { | |
d7a1fda2 | 8881 | dprintf_cont("while rebuilding "); |
a4e96fd8 MT |
8882 | if (state & DS_FAULTY) { |
8883 | dprintf_cont("removing failed drive "); | |
d7a1fda2 MT |
8884 | if (n == map->failed_disk_num) { |
8885 | dprintf_cont("end migration"); | |
8886 | end_migration(dev, super, map_state); | |
a4e96fd8 | 8887 | a->last_checkpoint = 0; |
d7a1fda2 | 8888 | } else { |
a4e96fd8 | 8889 | dprintf_cont("fail detected during rebuild, changing map state"); |
d7a1fda2 MT |
8890 | map->map_state = map_state; |
8891 | } | |
94002678 | 8892 | super->updates_pending++; |
fb12a745 TM |
8893 | } |
8894 | ||
a4e96fd8 MT |
8895 | if (!rebuild_done) |
8896 | break; | |
8897 | ||
fb12a745 TM |
8898 | /* check if recovery is really finished */ |
8899 | for (mdi = a->info.devs; mdi ; mdi = mdi->next) | |
8900 | if (mdi->recovery_start != MaxSector) { | |
8901 | recovery_not_finished = 1; | |
8902 | break; | |
8903 | } | |
8904 | if (recovery_not_finished) { | |
8905 | dprintf_cont("\n"); | |
a4e96fd8 | 8906 | dprintf_cont("Rebuild has not finished yet"); |
fb12a745 TM |
8907 | if (a->last_checkpoint < mdi->recovery_start) { |
8908 | a->last_checkpoint = | |
8909 | mdi->recovery_start; | |
8910 | super->updates_pending++; | |
8911 | } | |
8912 | break; | |
94002678 | 8913 | } |
fb12a745 TM |
8914 | |
8915 | dprintf_cont(" Rebuild done, still degraded"); | |
a4e96fd8 MT |
8916 | end_migration(dev, super, map_state); |
8917 | a->last_checkpoint = 0; | |
8918 | super->updates_pending++; | |
fb12a745 TM |
8919 | |
8920 | for (i = 0; i < map->num_members; i++) { | |
8921 | int idx = get_imsm_ord_tbl_ent(dev, i, MAP_0); | |
8922 | ||
8923 | if (idx & IMSM_ORD_REBUILD) | |
8924 | map->failed_disk_num = i; | |
8925 | } | |
8926 | super->updates_pending++; | |
94002678 AK |
8927 | break; |
8928 | } | |
8929 | if (is_gen_migration(dev)) { | |
1ade5cc1 | 8930 | dprintf_cont("while general migration"); |
bf2f0071 | 8931 | if (a->last_checkpoint >= a->info.component_size) |
809da78e | 8932 | end_migration(dev, super, map_state); |
94002678 AK |
8933 | else { |
8934 | map->map_state = map_state; | |
3b451610 | 8935 | manage_second_map(super, dev); |
94002678 AK |
8936 | } |
8937 | super->updates_pending++; | |
bf2f0071 | 8938 | break; |
28bce06f | 8939 | } |
6ce1fbf1 | 8940 | if (is_initializing(dev)) { |
1ade5cc1 | 8941 | dprintf_cont("while initialization."); |
6ce1fbf1 AK |
8942 | map->map_state = map_state; |
8943 | super->updates_pending++; | |
8944 | break; | |
8945 | } | |
94002678 AK |
8946 | break; |
8947 | case IMSM_T_STATE_FAILED: /* transition to failed state */ | |
1ade5cc1 | 8948 | dprintf_cont("failed: "); |
94002678 | 8949 | if (is_gen_migration(dev)) { |
1ade5cc1 | 8950 | dprintf_cont("while general migration"); |
94002678 AK |
8951 | map->map_state = map_state; |
8952 | super->updates_pending++; | |
8953 | break; | |
8954 | } | |
8955 | if (map->map_state != map_state) { | |
1ade5cc1 | 8956 | dprintf_cont("mark failed"); |
94002678 AK |
8957 | end_migration(dev, super, map_state); |
8958 | super->updates_pending++; | |
8959 | a->last_checkpoint = 0; | |
8960 | break; | |
8961 | } | |
8962 | break; | |
8963 | default: | |
1ade5cc1 | 8964 | dprintf_cont("state %i\n", map_state); |
5802a811 | 8965 | } |
1ade5cc1 | 8966 | dprintf_cont("\n"); |
845dea95 NB |
8967 | } |
8968 | ||
f796af5d | 8969 | static int store_imsm_mpb(int fd, struct imsm_super *mpb) |
c2a1e7da | 8970 | { |
f796af5d | 8971 | void *buf = mpb; |
c2a1e7da DW |
8972 | __u32 mpb_size = __le32_to_cpu(mpb->mpb_size); |
8973 | unsigned long long dsize; | |
8974 | unsigned long long sectors; | |
f36a9ecd | 8975 | unsigned int sector_size; |
c2a1e7da | 8976 | |
aec01630 JS |
8977 | if (!get_dev_sector_size(fd, NULL, §or_size)) |
8978 | return 1; | |
c2a1e7da DW |
8979 | get_dev_size(fd, NULL, &dsize); |
8980 | ||
f36a9ecd | 8981 | if (mpb_size > sector_size) { |
272f648f | 8982 | /* -1 to account for anchor */ |
f36a9ecd | 8983 | sectors = mpb_sectors(mpb, sector_size) - 1; |
c2a1e7da | 8984 | |
272f648f | 8985 | /* write the extended mpb to the sectors preceeding the anchor */ |
f36a9ecd PB |
8986 | if (lseek64(fd, dsize - (sector_size * (2 + sectors)), |
8987 | SEEK_SET) < 0) | |
272f648f | 8988 | return 1; |
c2a1e7da | 8989 | |
f36a9ecd PB |
8990 | if ((unsigned long long)write(fd, buf + sector_size, |
8991 | sector_size * sectors) != sector_size * sectors) | |
272f648f DW |
8992 | return 1; |
8993 | } | |
c2a1e7da | 8994 | |
272f648f | 8995 | /* first block is stored on second to last sector of the disk */ |
f36a9ecd | 8996 | if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0) |
c2a1e7da DW |
8997 | return 1; |
8998 | ||
466070ad | 8999 | if ((unsigned int)write(fd, buf, sector_size) != sector_size) |
c2a1e7da DW |
9000 | return 1; |
9001 | ||
c2a1e7da DW |
9002 | return 0; |
9003 | } | |
9004 | ||
2e735d19 | 9005 | static void imsm_sync_metadata(struct supertype *container) |
845dea95 | 9006 | { |
2e735d19 | 9007 | struct intel_super *super = container->sb; |
c2a1e7da | 9008 | |
1a64be56 | 9009 | dprintf("sync metadata: %d\n", super->updates_pending); |
c2a1e7da DW |
9010 | if (!super->updates_pending) |
9011 | return; | |
9012 | ||
36988a3d | 9013 | write_super_imsm(container, 0); |
c2a1e7da DW |
9014 | |
9015 | super->updates_pending = 0; | |
845dea95 NB |
9016 | } |
9017 | ||
272906ef DW |
9018 | static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a) |
9019 | { | |
9020 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 9021 | int i = get_imsm_disk_idx(dev, idx, MAP_X); |
272906ef DW |
9022 | struct dl *dl; |
9023 | ||
9024 | for (dl = super->disks; dl; dl = dl->next) | |
9025 | if (dl->index == i) | |
9026 | break; | |
9027 | ||
25ed7e59 | 9028 | if (dl && is_failed(&dl->disk)) |
272906ef DW |
9029 | dl = NULL; |
9030 | ||
9031 | if (dl) | |
1ade5cc1 | 9032 | dprintf("found %x:%x\n", dl->major, dl->minor); |
272906ef DW |
9033 | |
9034 | return dl; | |
9035 | } | |
9036 | ||
a20d2ba5 | 9037 | static struct dl *imsm_add_spare(struct intel_super *super, int slot, |
8ba77d32 AK |
9038 | struct active_array *a, int activate_new, |
9039 | struct mdinfo *additional_test_list) | |
272906ef DW |
9040 | { |
9041 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 9042 | int idx = get_imsm_disk_idx(dev, slot, MAP_X); |
a20d2ba5 DW |
9043 | struct imsm_super *mpb = super->anchor; |
9044 | struct imsm_map *map; | |
272906ef DW |
9045 | unsigned long long pos; |
9046 | struct mdinfo *d; | |
9047 | struct extent *ex; | |
a20d2ba5 | 9048 | int i, j; |
272906ef | 9049 | int found; |
569cc43f DW |
9050 | __u32 array_start = 0; |
9051 | __u32 array_end = 0; | |
272906ef | 9052 | struct dl *dl; |
6c932028 | 9053 | struct mdinfo *test_list; |
272906ef DW |
9054 | |
9055 | for (dl = super->disks; dl; dl = dl->next) { | |
9056 | /* If in this array, skip */ | |
9057 | for (d = a->info.devs ; d ; d = d->next) | |
4389ce73 | 9058 | if (is_fd_valid(d->state_fd) && |
e553d2a4 | 9059 | d->disk.major == dl->major && |
272906ef | 9060 | d->disk.minor == dl->minor) { |
8ba77d32 AK |
9061 | dprintf("%x:%x already in array\n", |
9062 | dl->major, dl->minor); | |
272906ef DW |
9063 | break; |
9064 | } | |
9065 | if (d) | |
9066 | continue; | |
6c932028 AK |
9067 | test_list = additional_test_list; |
9068 | while (test_list) { | |
9069 | if (test_list->disk.major == dl->major && | |
9070 | test_list->disk.minor == dl->minor) { | |
8ba77d32 AK |
9071 | dprintf("%x:%x already in additional test list\n", |
9072 | dl->major, dl->minor); | |
9073 | break; | |
9074 | } | |
6c932028 | 9075 | test_list = test_list->next; |
8ba77d32 | 9076 | } |
6c932028 | 9077 | if (test_list) |
8ba77d32 | 9078 | continue; |
272906ef | 9079 | |
e553d2a4 | 9080 | /* skip in use or failed drives */ |
25ed7e59 | 9081 | if (is_failed(&dl->disk) || idx == dl->index || |
df474657 DW |
9082 | dl->index == -2) { |
9083 | dprintf("%x:%x status (failed: %d index: %d)\n", | |
25ed7e59 | 9084 | dl->major, dl->minor, is_failed(&dl->disk), idx); |
9a1608e5 DW |
9085 | continue; |
9086 | } | |
9087 | ||
a20d2ba5 DW |
9088 | /* skip pure spares when we are looking for partially |
9089 | * assimilated drives | |
9090 | */ | |
9091 | if (dl->index == -1 && !activate_new) | |
9092 | continue; | |
9093 | ||
f2cc4f7d AO |
9094 | if (!drive_validate_sector_size(super, dl)) |
9095 | continue; | |
9096 | ||
272906ef | 9097 | /* Does this unused device have the requisite free space? |
a20d2ba5 | 9098 | * It needs to be able to cover all member volumes |
272906ef | 9099 | */ |
05501181 | 9100 | ex = get_extents(super, dl, 1); |
272906ef DW |
9101 | if (!ex) { |
9102 | dprintf("cannot get extents\n"); | |
9103 | continue; | |
9104 | } | |
a20d2ba5 DW |
9105 | for (i = 0; i < mpb->num_raid_devs; i++) { |
9106 | dev = get_imsm_dev(super, i); | |
238c0a71 | 9107 | map = get_imsm_map(dev, MAP_0); |
272906ef | 9108 | |
a20d2ba5 DW |
9109 | /* check if this disk is already a member of |
9110 | * this array | |
272906ef | 9111 | */ |
620b1713 | 9112 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
a20d2ba5 DW |
9113 | continue; |
9114 | ||
9115 | found = 0; | |
9116 | j = 0; | |
9117 | pos = 0; | |
5551b113 | 9118 | array_start = pba_of_lba0(map); |
329c8278 | 9119 | array_end = array_start + |
44490938 | 9120 | per_dev_array_size(map) - 1; |
a20d2ba5 DW |
9121 | |
9122 | do { | |
9123 | /* check that we can start at pba_of_lba0 with | |
44490938 | 9124 | * num_data_stripes*blocks_per_stripe of space |
a20d2ba5 | 9125 | */ |
329c8278 | 9126 | if (array_start >= pos && array_end < ex[j].start) { |
a20d2ba5 DW |
9127 | found = 1; |
9128 | break; | |
9129 | } | |
9130 | pos = ex[j].start + ex[j].size; | |
9131 | j++; | |
9132 | } while (ex[j-1].size); | |
9133 | ||
9134 | if (!found) | |
272906ef | 9135 | break; |
a20d2ba5 | 9136 | } |
272906ef DW |
9137 | |
9138 | free(ex); | |
a20d2ba5 | 9139 | if (i < mpb->num_raid_devs) { |
329c8278 DW |
9140 | dprintf("%x:%x does not have %u to %u available\n", |
9141 | dl->major, dl->minor, array_start, array_end); | |
272906ef DW |
9142 | /* No room */ |
9143 | continue; | |
a20d2ba5 DW |
9144 | } |
9145 | return dl; | |
272906ef DW |
9146 | } |
9147 | ||
9148 | return dl; | |
9149 | } | |
9150 | ||
95d07a2c LM |
9151 | static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed) |
9152 | { | |
9153 | struct imsm_dev *dev2; | |
9154 | struct imsm_map *map; | |
9155 | struct dl *idisk; | |
9156 | int slot; | |
9157 | int idx; | |
9158 | __u8 state; | |
9159 | ||
9160 | dev2 = get_imsm_dev(cont->sb, dev_idx); | |
756a15f3 MG |
9161 | |
9162 | state = imsm_check_degraded(cont->sb, dev2, failed, MAP_0); | |
9163 | if (state == IMSM_T_STATE_FAILED) { | |
9164 | map = get_imsm_map(dev2, MAP_0); | |
9165 | for (slot = 0; slot < map->num_members; slot++) { | |
9166 | /* | |
9167 | * Check if failed disks are deleted from intel | |
9168 | * disk list or are marked to be deleted | |
9169 | */ | |
9170 | idx = get_imsm_disk_idx(dev2, slot, MAP_X); | |
9171 | idisk = get_imsm_dl_disk(cont->sb, idx); | |
9172 | /* | |
9173 | * Do not rebuild the array if failed disks | |
9174 | * from failed sub-array are not removed from | |
9175 | * container. | |
9176 | */ | |
9177 | if (idisk && | |
9178 | is_failed(&idisk->disk) && | |
9179 | (idisk->action != DISK_REMOVE)) | |
9180 | return 0; | |
95d07a2c LM |
9181 | } |
9182 | } | |
9183 | return 1; | |
9184 | } | |
9185 | ||
88758e9d DW |
9186 | static struct mdinfo *imsm_activate_spare(struct active_array *a, |
9187 | struct metadata_update **updates) | |
9188 | { | |
9189 | /** | |
d23fe947 DW |
9190 | * Find a device with unused free space and use it to replace a |
9191 | * failed/vacant region in an array. We replace failed regions one a | |
9192 | * array at a time. The result is that a new spare disk will be added | |
9193 | * to the first failed array and after the monitor has finished | |
9194 | * propagating failures the remainder will be consumed. | |
88758e9d | 9195 | * |
d23fe947 DW |
9196 | * FIXME add a capability for mdmon to request spares from another |
9197 | * container. | |
88758e9d DW |
9198 | */ |
9199 | ||
9200 | struct intel_super *super = a->container->sb; | |
88758e9d | 9201 | int inst = a->info.container_member; |
949c47a0 | 9202 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 9203 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
88758e9d DW |
9204 | int failed = a->info.array.raid_disks; |
9205 | struct mdinfo *rv = NULL; | |
9206 | struct mdinfo *d; | |
9207 | struct mdinfo *di; | |
9208 | struct metadata_update *mu; | |
9209 | struct dl *dl; | |
9210 | struct imsm_update_activate_spare *u; | |
9211 | int num_spares = 0; | |
9212 | int i; | |
95d07a2c | 9213 | int allowed; |
88758e9d | 9214 | |
4389ce73 MT |
9215 | for (d = a->info.devs ; d; d = d->next) { |
9216 | if (!is_fd_valid(d->state_fd)) | |
9217 | continue; | |
9218 | ||
9219 | if (d->curr_state & DS_FAULTY) | |
88758e9d DW |
9220 | /* wait for Removal to happen */ |
9221 | return NULL; | |
4389ce73 MT |
9222 | |
9223 | failed--; | |
88758e9d DW |
9224 | } |
9225 | ||
9226 | dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n", | |
9227 | inst, failed, a->info.array.raid_disks, a->info.array.level); | |
1af97990 | 9228 | |
e2962bfc AK |
9229 | if (imsm_reshape_blocks_arrays_changes(super)) |
9230 | return NULL; | |
1af97990 | 9231 | |
fc8ca064 AK |
9232 | /* Cannot activate another spare if rebuild is in progress already |
9233 | */ | |
9234 | if (is_rebuilding(dev)) { | |
7a862a02 | 9235 | dprintf("imsm: No spare activation allowed. Rebuild in progress already.\n"); |
fc8ca064 AK |
9236 | return NULL; |
9237 | } | |
9238 | ||
89c67882 AK |
9239 | if (a->info.array.level == 4) |
9240 | /* No repair for takeovered array | |
9241 | * imsm doesn't support raid4 | |
9242 | */ | |
9243 | return NULL; | |
9244 | ||
3b451610 AK |
9245 | if (imsm_check_degraded(super, dev, failed, MAP_0) != |
9246 | IMSM_T_STATE_DEGRADED) | |
88758e9d DW |
9247 | return NULL; |
9248 | ||
83ca7d45 AP |
9249 | if (get_imsm_map(dev, MAP_0)->map_state == IMSM_T_STATE_UNINITIALIZED) { |
9250 | dprintf("imsm: No spare activation allowed. Volume is not initialized.\n"); | |
9251 | return NULL; | |
9252 | } | |
9253 | ||
95d07a2c LM |
9254 | /* |
9255 | * If there are any failed disks check state of the other volume. | |
9256 | * Block rebuild if the another one is failed until failed disks | |
9257 | * are removed from container. | |
9258 | */ | |
9259 | if (failed) { | |
7a862a02 | 9260 | dprintf("found failed disks in %.*s, check if there anotherfailed sub-array.\n", |
c4acd1e5 | 9261 | MAX_RAID_SERIAL_LEN, dev->volume); |
95d07a2c LM |
9262 | /* check if states of the other volumes allow for rebuild */ |
9263 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
9264 | if (i != inst) { | |
9265 | allowed = imsm_rebuild_allowed(a->container, | |
9266 | i, failed); | |
9267 | if (!allowed) | |
9268 | return NULL; | |
9269 | } | |
9270 | } | |
9271 | } | |
9272 | ||
88758e9d | 9273 | /* For each slot, if it is not working, find a spare */ |
88758e9d DW |
9274 | for (i = 0; i < a->info.array.raid_disks; i++) { |
9275 | for (d = a->info.devs ; d ; d = d->next) | |
9276 | if (d->disk.raid_disk == i) | |
9277 | break; | |
9278 | dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0); | |
4389ce73 | 9279 | if (d && is_fd_valid(d->state_fd)) |
88758e9d DW |
9280 | continue; |
9281 | ||
272906ef | 9282 | /* |
a20d2ba5 DW |
9283 | * OK, this device needs recovery. Try to re-add the |
9284 | * previous occupant of this slot, if this fails see if | |
9285 | * we can continue the assimilation of a spare that was | |
9286 | * partially assimilated, finally try to activate a new | |
9287 | * spare. | |
272906ef DW |
9288 | */ |
9289 | dl = imsm_readd(super, i, a); | |
9290 | if (!dl) | |
b303fe21 | 9291 | dl = imsm_add_spare(super, i, a, 0, rv); |
a20d2ba5 | 9292 | if (!dl) |
b303fe21 | 9293 | dl = imsm_add_spare(super, i, a, 1, rv); |
272906ef DW |
9294 | if (!dl) |
9295 | continue; | |
1011e834 | 9296 | |
272906ef | 9297 | /* found a usable disk with enough space */ |
503975b9 | 9298 | di = xcalloc(1, sizeof(*di)); |
272906ef DW |
9299 | |
9300 | /* dl->index will be -1 in the case we are activating a | |
9301 | * pristine spare. imsm_process_update() will create a | |
9302 | * new index in this case. Once a disk is found to be | |
9303 | * failed in all member arrays it is kicked from the | |
9304 | * metadata | |
9305 | */ | |
9306 | di->disk.number = dl->index; | |
d23fe947 | 9307 | |
272906ef DW |
9308 | /* (ab)use di->devs to store a pointer to the device |
9309 | * we chose | |
9310 | */ | |
9311 | di->devs = (struct mdinfo *) dl; | |
9312 | ||
9313 | di->disk.raid_disk = i; | |
9314 | di->disk.major = dl->major; | |
9315 | di->disk.minor = dl->minor; | |
9316 | di->disk.state = 0; | |
d23534e4 | 9317 | di->recovery_start = 0; |
5551b113 | 9318 | di->data_offset = pba_of_lba0(map); |
272906ef DW |
9319 | di->component_size = a->info.component_size; |
9320 | di->container_member = inst; | |
5e46202e | 9321 | di->bb.supported = 1; |
2c8890e9 | 9322 | if (a->info.consistency_policy == CONSISTENCY_POLICY_PPL) { |
2432ce9b | 9323 | di->ppl_sector = get_ppl_sector(super, inst); |
c2462068 | 9324 | di->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9; |
2432ce9b | 9325 | } |
148acb7b | 9326 | super->random = random32(); |
272906ef DW |
9327 | di->next = rv; |
9328 | rv = di; | |
9329 | num_spares++; | |
9330 | dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor, | |
9331 | i, di->data_offset); | |
88758e9d DW |
9332 | } |
9333 | ||
9334 | if (!rv) | |
9335 | /* No spares found */ | |
9336 | return rv; | |
9337 | /* Now 'rv' has a list of devices to return. | |
9338 | * Create a metadata_update record to update the | |
9339 | * disk_ord_tbl for the array | |
9340 | */ | |
503975b9 | 9341 | mu = xmalloc(sizeof(*mu)); |
1011e834 | 9342 | mu->buf = xcalloc(num_spares, |
503975b9 | 9343 | sizeof(struct imsm_update_activate_spare)); |
88758e9d | 9344 | mu->space = NULL; |
cb23f1f4 | 9345 | mu->space_list = NULL; |
88758e9d DW |
9346 | mu->len = sizeof(struct imsm_update_activate_spare) * num_spares; |
9347 | mu->next = *updates; | |
9348 | u = (struct imsm_update_activate_spare *) mu->buf; | |
9349 | ||
9350 | for (di = rv ; di ; di = di->next) { | |
9351 | u->type = update_activate_spare; | |
d23fe947 DW |
9352 | u->dl = (struct dl *) di->devs; |
9353 | di->devs = NULL; | |
88758e9d DW |
9354 | u->slot = di->disk.raid_disk; |
9355 | u->array = inst; | |
9356 | u->next = u + 1; | |
9357 | u++; | |
9358 | } | |
9359 | (u-1)->next = NULL; | |
9360 | *updates = mu; | |
9361 | ||
9362 | return rv; | |
9363 | } | |
9364 | ||
54c2c1ea | 9365 | static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u) |
8273f55e | 9366 | { |
54c2c1ea | 9367 | struct imsm_dev *dev = get_imsm_dev(super, idx); |
238c0a71 AK |
9368 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
9369 | struct imsm_map *new_map = get_imsm_map(&u->dev, MAP_0); | |
54c2c1ea DW |
9370 | struct disk_info *inf = get_disk_info(u); |
9371 | struct imsm_disk *disk; | |
8273f55e DW |
9372 | int i; |
9373 | int j; | |
8273f55e | 9374 | |
54c2c1ea | 9375 | for (i = 0; i < map->num_members; i++) { |
238c0a71 | 9376 | disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, MAP_X)); |
54c2c1ea DW |
9377 | for (j = 0; j < new_map->num_members; j++) |
9378 | if (serialcmp(disk->serial, inf[j].serial) == 0) | |
8273f55e DW |
9379 | return 1; |
9380 | } | |
9381 | ||
9382 | return 0; | |
9383 | } | |
9384 | ||
1a64be56 LM |
9385 | static struct dl *get_disk_super(struct intel_super *super, int major, int minor) |
9386 | { | |
594dc1b8 JS |
9387 | struct dl *dl; |
9388 | ||
1a64be56 | 9389 | for (dl = super->disks; dl; dl = dl->next) |
089f9d79 | 9390 | if (dl->major == major && dl->minor == minor) |
1a64be56 LM |
9391 | return dl; |
9392 | return NULL; | |
9393 | } | |
9394 | ||
9395 | static int remove_disk_super(struct intel_super *super, int major, int minor) | |
9396 | { | |
594dc1b8 | 9397 | struct dl *prev; |
1a64be56 LM |
9398 | struct dl *dl; |
9399 | ||
9400 | prev = NULL; | |
9401 | for (dl = super->disks; dl; dl = dl->next) { | |
089f9d79 | 9402 | if (dl->major == major && dl->minor == minor) { |
1a64be56 LM |
9403 | /* remove */ |
9404 | if (prev) | |
9405 | prev->next = dl->next; | |
9406 | else | |
9407 | super->disks = dl->next; | |
9408 | dl->next = NULL; | |
3a85bf0e | 9409 | __free_imsm_disk(dl, 1); |
1ade5cc1 | 9410 | dprintf("removed %x:%x\n", major, minor); |
1a64be56 LM |
9411 | break; |
9412 | } | |
9413 | prev = dl; | |
9414 | } | |
9415 | return 0; | |
9416 | } | |
9417 | ||
f21e18ca | 9418 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index); |
ae6aad82 | 9419 | |
1a64be56 LM |
9420 | static int add_remove_disk_update(struct intel_super *super) |
9421 | { | |
9422 | int check_degraded = 0; | |
594dc1b8 JS |
9423 | struct dl *disk; |
9424 | ||
1a64be56 LM |
9425 | /* add/remove some spares to/from the metadata/contrainer */ |
9426 | while (super->disk_mgmt_list) { | |
9427 | struct dl *disk_cfg; | |
9428 | ||
9429 | disk_cfg = super->disk_mgmt_list; | |
9430 | super->disk_mgmt_list = disk_cfg->next; | |
9431 | disk_cfg->next = NULL; | |
9432 | ||
9433 | if (disk_cfg->action == DISK_ADD) { | |
9434 | disk_cfg->next = super->disks; | |
9435 | super->disks = disk_cfg; | |
9436 | check_degraded = 1; | |
1ade5cc1 N |
9437 | dprintf("added %x:%x\n", |
9438 | disk_cfg->major, disk_cfg->minor); | |
1a64be56 LM |
9439 | } else if (disk_cfg->action == DISK_REMOVE) { |
9440 | dprintf("Disk remove action processed: %x.%x\n", | |
9441 | disk_cfg->major, disk_cfg->minor); | |
9442 | disk = get_disk_super(super, | |
9443 | disk_cfg->major, | |
9444 | disk_cfg->minor); | |
9445 | if (disk) { | |
9446 | /* store action status */ | |
9447 | disk->action = DISK_REMOVE; | |
9448 | /* remove spare disks only */ | |
9449 | if (disk->index == -1) { | |
9450 | remove_disk_super(super, | |
9451 | disk_cfg->major, | |
9452 | disk_cfg->minor); | |
91c97c54 MT |
9453 | } else { |
9454 | disk_cfg->fd = disk->fd; | |
9455 | disk->fd = -1; | |
1a64be56 LM |
9456 | } |
9457 | } | |
9458 | /* release allocate disk structure */ | |
3a85bf0e | 9459 | __free_imsm_disk(disk_cfg, 1); |
1a64be56 LM |
9460 | } |
9461 | } | |
9462 | return check_degraded; | |
9463 | } | |
9464 | ||
a29911da PC |
9465 | static int apply_reshape_migration_update(struct imsm_update_reshape_migration *u, |
9466 | struct intel_super *super, | |
9467 | void ***space_list) | |
9468 | { | |
9469 | struct intel_dev *id; | |
9470 | void **tofree = NULL; | |
9471 | int ret_val = 0; | |
9472 | ||
1ade5cc1 | 9473 | dprintf("(enter)\n"); |
089f9d79 | 9474 | if (u->subdev < 0 || u->subdev > 1) { |
a29911da PC |
9475 | dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev); |
9476 | return ret_val; | |
9477 | } | |
089f9d79 | 9478 | if (space_list == NULL || *space_list == NULL) { |
a29911da PC |
9479 | dprintf("imsm: Error: Memory is not allocated\n"); |
9480 | return ret_val; | |
9481 | } | |
9482 | ||
9483 | for (id = super->devlist ; id; id = id->next) { | |
9484 | if (id->index == (unsigned)u->subdev) { | |
9485 | struct imsm_dev *dev = get_imsm_dev(super, u->subdev); | |
9486 | struct imsm_map *map; | |
9487 | struct imsm_dev *new_dev = | |
9488 | (struct imsm_dev *)*space_list; | |
238c0a71 | 9489 | struct imsm_map *migr_map = get_imsm_map(dev, MAP_1); |
a29911da PC |
9490 | int to_state; |
9491 | struct dl *new_disk; | |
9492 | ||
9493 | if (new_dev == NULL) | |
9494 | return ret_val; | |
9495 | *space_list = **space_list; | |
9496 | memcpy(new_dev, dev, sizeof_imsm_dev(dev, 0)); | |
238c0a71 | 9497 | map = get_imsm_map(new_dev, MAP_0); |
a29911da PC |
9498 | if (migr_map) { |
9499 | dprintf("imsm: Error: migration in progress"); | |
9500 | return ret_val; | |
9501 | } | |
9502 | ||
9503 | to_state = map->map_state; | |
9504 | if ((u->new_level == 5) && (map->raid_level == 0)) { | |
9505 | map->num_members++; | |
9506 | /* this should not happen */ | |
9507 | if (u->new_disks[0] < 0) { | |
9508 | map->failed_disk_num = | |
9509 | map->num_members - 1; | |
9510 | to_state = IMSM_T_STATE_DEGRADED; | |
9511 | } else | |
9512 | to_state = IMSM_T_STATE_NORMAL; | |
9513 | } | |
8e59f3d8 | 9514 | migrate(new_dev, super, to_state, MIGR_GEN_MIGR); |
a29911da PC |
9515 | if (u->new_level > -1) |
9516 | map->raid_level = u->new_level; | |
238c0a71 | 9517 | migr_map = get_imsm_map(new_dev, MAP_1); |
a29911da PC |
9518 | if ((u->new_level == 5) && |
9519 | (migr_map->raid_level == 0)) { | |
9520 | int ord = map->num_members - 1; | |
9521 | migr_map->num_members--; | |
9522 | if (u->new_disks[0] < 0) | |
9523 | ord |= IMSM_ORD_REBUILD; | |
9524 | set_imsm_ord_tbl_ent(map, | |
9525 | map->num_members - 1, | |
9526 | ord); | |
9527 | } | |
9528 | id->dev = new_dev; | |
9529 | tofree = (void **)dev; | |
9530 | ||
4bba0439 PC |
9531 | /* update chunk size |
9532 | */ | |
06fb291a | 9533 | if (u->new_chunksize > 0) { |
9529d343 MD |
9534 | struct imsm_map *dest_map = |
9535 | get_imsm_map(dev, MAP_0); | |
06fb291a | 9536 | int used_disks = |
9529d343 | 9537 | imsm_num_data_members(dest_map); |
06fb291a PB |
9538 | |
9539 | if (used_disks == 0) | |
9540 | return ret_val; | |
9541 | ||
4bba0439 PC |
9542 | map->blocks_per_strip = |
9543 | __cpu_to_le16(u->new_chunksize * 2); | |
1c275381 | 9544 | update_num_data_stripes(map, imsm_dev_size(dev)); |
06fb291a | 9545 | } |
4bba0439 | 9546 | |
44490938 MD |
9547 | /* ensure blocks_per_member has valid value |
9548 | */ | |
9549 | set_blocks_per_member(map, | |
9550 | per_dev_array_size(map) + | |
9551 | NUM_BLOCKS_DIRTY_STRIPE_REGION); | |
9552 | ||
a29911da PC |
9553 | /* add disk |
9554 | */ | |
089f9d79 JS |
9555 | if (u->new_level != 5 || migr_map->raid_level != 0 || |
9556 | migr_map->raid_level == map->raid_level) | |
a29911da PC |
9557 | goto skip_disk_add; |
9558 | ||
9559 | if (u->new_disks[0] >= 0) { | |
9560 | /* use passes spare | |
9561 | */ | |
9562 | new_disk = get_disk_super(super, | |
9563 | major(u->new_disks[0]), | |
9564 | minor(u->new_disks[0])); | |
7a862a02 | 9565 | dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n", |
a29911da PC |
9566 | major(u->new_disks[0]), |
9567 | minor(u->new_disks[0]), | |
9568 | new_disk, new_disk->index); | |
9569 | if (new_disk == NULL) | |
9570 | goto error_disk_add; | |
9571 | ||
9572 | new_disk->index = map->num_members - 1; | |
9573 | /* slot to fill in autolayout | |
9574 | */ | |
9575 | new_disk->raiddisk = new_disk->index; | |
9576 | new_disk->disk.status |= CONFIGURED_DISK; | |
9577 | new_disk->disk.status &= ~SPARE_DISK; | |
9578 | } else | |
9579 | goto error_disk_add; | |
9580 | ||
9581 | skip_disk_add: | |
9582 | *tofree = *space_list; | |
9583 | /* calculate new size | |
9584 | */ | |
f3871fdc | 9585 | imsm_set_array_size(new_dev, -1); |
a29911da PC |
9586 | |
9587 | ret_val = 1; | |
9588 | } | |
9589 | } | |
9590 | ||
9591 | if (tofree) | |
9592 | *space_list = tofree; | |
9593 | return ret_val; | |
9594 | ||
9595 | error_disk_add: | |
9596 | dprintf("Error: imsm: Cannot find disk.\n"); | |
9597 | return ret_val; | |
9598 | } | |
9599 | ||
f3871fdc AK |
9600 | static int apply_size_change_update(struct imsm_update_size_change *u, |
9601 | struct intel_super *super) | |
9602 | { | |
9603 | struct intel_dev *id; | |
9604 | int ret_val = 0; | |
9605 | ||
1ade5cc1 | 9606 | dprintf("(enter)\n"); |
089f9d79 | 9607 | if (u->subdev < 0 || u->subdev > 1) { |
f3871fdc AK |
9608 | dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev); |
9609 | return ret_val; | |
9610 | } | |
9611 | ||
9612 | for (id = super->devlist ; id; id = id->next) { | |
9613 | if (id->index == (unsigned)u->subdev) { | |
9614 | struct imsm_dev *dev = get_imsm_dev(super, u->subdev); | |
9615 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
9529d343 | 9616 | int used_disks = imsm_num_data_members(map); |
f3871fdc | 9617 | unsigned long long blocks_per_member; |
44490938 MD |
9618 | unsigned long long new_size_per_disk; |
9619 | ||
9620 | if (used_disks == 0) | |
9621 | return 0; | |
f3871fdc AK |
9622 | |
9623 | /* calculate new size | |
9624 | */ | |
44490938 MD |
9625 | new_size_per_disk = u->new_size / used_disks; |
9626 | blocks_per_member = new_size_per_disk + | |
9627 | NUM_BLOCKS_DIRTY_STRIPE_REGION; | |
f3871fdc | 9628 | |
1c275381 MT |
9629 | imsm_set_array_size(dev, u->new_size); |
9630 | set_blocks_per_member(map, blocks_per_member); | |
9631 | update_num_data_stripes(map, u->new_size); | |
f3871fdc AK |
9632 | ret_val = 1; |
9633 | break; | |
9634 | } | |
9635 | } | |
9636 | ||
9637 | return ret_val; | |
9638 | } | |
9639 | ||
69d40de4 JR |
9640 | static int prepare_spare_to_activate(struct supertype *st, |
9641 | struct imsm_update_activate_spare *u) | |
9642 | { | |
9643 | struct intel_super *super = st->sb; | |
9644 | int prev_current_vol = super->current_vol; | |
9645 | struct active_array *a; | |
9646 | int ret = 1; | |
9647 | ||
9648 | for (a = st->arrays; a; a = a->next) | |
9649 | /* | |
9650 | * Additional initialization (adding bitmap header, filling | |
9651 | * the bitmap area with '1's to force initial rebuild for a whole | |
9652 | * data-area) is required when adding the spare to the volume | |
9653 | * with write-intent bitmap. | |
9654 | */ | |
9655 | if (a->info.container_member == u->array && | |
9656 | a->info.consistency_policy == CONSISTENCY_POLICY_BITMAP) { | |
9657 | struct dl *dl; | |
9658 | ||
9659 | for (dl = super->disks; dl; dl = dl->next) | |
9660 | if (dl == u->dl) | |
9661 | break; | |
9662 | if (!dl) | |
9663 | break; | |
9664 | ||
9665 | super->current_vol = u->array; | |
9666 | if (st->ss->write_bitmap(st, dl->fd, NoUpdate)) | |
9667 | ret = 0; | |
9668 | super->current_vol = prev_current_vol; | |
9669 | } | |
9670 | return ret; | |
9671 | } | |
9672 | ||
061d7da3 | 9673 | static int apply_update_activate_spare(struct imsm_update_activate_spare *u, |
ca9de185 | 9674 | struct intel_super *super, |
061d7da3 LO |
9675 | struct active_array *active_array) |
9676 | { | |
9677 | struct imsm_super *mpb = super->anchor; | |
9678 | struct imsm_dev *dev = get_imsm_dev(super, u->array); | |
238c0a71 | 9679 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
061d7da3 LO |
9680 | struct imsm_map *migr_map; |
9681 | struct active_array *a; | |
9682 | struct imsm_disk *disk; | |
9683 | __u8 to_state; | |
9684 | struct dl *dl; | |
9685 | unsigned int found; | |
9686 | int failed; | |
5961eeec | 9687 | int victim; |
061d7da3 | 9688 | int i; |
5961eeec | 9689 | int second_map_created = 0; |
061d7da3 | 9690 | |
5961eeec | 9691 | for (; u; u = u->next) { |
238c0a71 | 9692 | victim = get_imsm_disk_idx(dev, u->slot, MAP_X); |
061d7da3 | 9693 | |
5961eeec | 9694 | if (victim < 0) |
9695 | return 0; | |
061d7da3 | 9696 | |
5961eeec | 9697 | for (dl = super->disks; dl; dl = dl->next) |
9698 | if (dl == u->dl) | |
9699 | break; | |
061d7da3 | 9700 | |
5961eeec | 9701 | if (!dl) { |
7a862a02 | 9702 | pr_err("error: imsm_activate_spare passed an unknown disk (index: %d)\n", |
5961eeec | 9703 | u->dl->index); |
9704 | return 0; | |
9705 | } | |
061d7da3 | 9706 | |
5961eeec | 9707 | /* count failures (excluding rebuilds and the victim) |
9708 | * to determine map[0] state | |
9709 | */ | |
9710 | failed = 0; | |
9711 | for (i = 0; i < map->num_members; i++) { | |
9712 | if (i == u->slot) | |
9713 | continue; | |
9714 | disk = get_imsm_disk(super, | |
238c0a71 | 9715 | get_imsm_disk_idx(dev, i, MAP_X)); |
5961eeec | 9716 | if (!disk || is_failed(disk)) |
9717 | failed++; | |
9718 | } | |
061d7da3 | 9719 | |
5961eeec | 9720 | /* adding a pristine spare, assign a new index */ |
9721 | if (dl->index < 0) { | |
9722 | dl->index = super->anchor->num_disks; | |
9723 | super->anchor->num_disks++; | |
9724 | } | |
9725 | disk = &dl->disk; | |
9726 | disk->status |= CONFIGURED_DISK; | |
9727 | disk->status &= ~SPARE_DISK; | |
9728 | ||
9729 | /* mark rebuild */ | |
238c0a71 | 9730 | to_state = imsm_check_degraded(super, dev, failed, MAP_0); |
5961eeec | 9731 | if (!second_map_created) { |
9732 | second_map_created = 1; | |
9733 | map->map_state = IMSM_T_STATE_DEGRADED; | |
9734 | migrate(dev, super, to_state, MIGR_REBUILD); | |
9735 | } else | |
9736 | map->map_state = to_state; | |
238c0a71 | 9737 | migr_map = get_imsm_map(dev, MAP_1); |
5961eeec | 9738 | set_imsm_ord_tbl_ent(map, u->slot, dl->index); |
9739 | set_imsm_ord_tbl_ent(migr_map, u->slot, | |
9740 | dl->index | IMSM_ORD_REBUILD); | |
9741 | ||
9742 | /* update the family_num to mark a new container | |
9743 | * generation, being careful to record the existing | |
9744 | * family_num in orig_family_num to clean up after | |
9745 | * earlier mdadm versions that neglected to set it. | |
9746 | */ | |
9747 | if (mpb->orig_family_num == 0) | |
9748 | mpb->orig_family_num = mpb->family_num; | |
9749 | mpb->family_num += super->random; | |
9750 | ||
9751 | /* count arrays using the victim in the metadata */ | |
9752 | found = 0; | |
9753 | for (a = active_array; a ; a = a->next) { | |
76c152ca | 9754 | int dev_idx = a->info.container_member; |
061d7da3 | 9755 | |
76c152ca | 9756 | if (get_disk_slot_in_dev(super, dev_idx, victim) >= 0) |
5961eeec | 9757 | found++; |
9758 | } | |
061d7da3 | 9759 | |
5961eeec | 9760 | /* delete the victim if it is no longer being |
9761 | * utilized anywhere | |
061d7da3 | 9762 | */ |
5961eeec | 9763 | if (!found) { |
9764 | struct dl **dlp; | |
061d7da3 | 9765 | |
5961eeec | 9766 | /* We know that 'manager' isn't touching anything, |
9767 | * so it is safe to delete | |
9768 | */ | |
9769 | for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next) | |
061d7da3 LO |
9770 | if ((*dlp)->index == victim) |
9771 | break; | |
5961eeec | 9772 | |
9773 | /* victim may be on the missing list */ | |
9774 | if (!*dlp) | |
9775 | for (dlp = &super->missing; *dlp; | |
9776 | dlp = &(*dlp)->next) | |
9777 | if ((*dlp)->index == victim) | |
9778 | break; | |
9779 | imsm_delete(super, dlp, victim); | |
9780 | } | |
061d7da3 LO |
9781 | } |
9782 | ||
9783 | return 1; | |
9784 | } | |
a29911da | 9785 | |
2e5dc010 N |
9786 | static int apply_reshape_container_disks_update(struct imsm_update_reshape *u, |
9787 | struct intel_super *super, | |
9788 | void ***space_list) | |
9789 | { | |
9790 | struct dl *new_disk; | |
9791 | struct intel_dev *id; | |
9792 | int i; | |
9793 | int delta_disks = u->new_raid_disks - u->old_raid_disks; | |
ee4beede | 9794 | int disk_count = u->old_raid_disks; |
2e5dc010 N |
9795 | void **tofree = NULL; |
9796 | int devices_to_reshape = 1; | |
9797 | struct imsm_super *mpb = super->anchor; | |
9798 | int ret_val = 0; | |
d098291a | 9799 | unsigned int dev_id; |
2e5dc010 | 9800 | |
1ade5cc1 | 9801 | dprintf("(enter)\n"); |
2e5dc010 N |
9802 | |
9803 | /* enable spares to use in array */ | |
9804 | for (i = 0; i < delta_disks; i++) { | |
9805 | new_disk = get_disk_super(super, | |
9806 | major(u->new_disks[i]), | |
9807 | minor(u->new_disks[i])); | |
7a862a02 | 9808 | dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n", |
2e5dc010 N |
9809 | major(u->new_disks[i]), minor(u->new_disks[i]), |
9810 | new_disk, new_disk->index); | |
089f9d79 JS |
9811 | if (new_disk == NULL || |
9812 | (new_disk->index >= 0 && | |
9813 | new_disk->index < u->old_raid_disks)) | |
2e5dc010 | 9814 | goto update_reshape_exit; |
ee4beede | 9815 | new_disk->index = disk_count++; |
2e5dc010 N |
9816 | /* slot to fill in autolayout |
9817 | */ | |
9818 | new_disk->raiddisk = new_disk->index; | |
9819 | new_disk->disk.status |= | |
9820 | CONFIGURED_DISK; | |
9821 | new_disk->disk.status &= ~SPARE_DISK; | |
9822 | } | |
9823 | ||
ed7333bd AK |
9824 | dprintf("imsm: volume set mpb->num_raid_devs = %i\n", |
9825 | mpb->num_raid_devs); | |
2e5dc010 N |
9826 | /* manage changes in volume |
9827 | */ | |
d098291a | 9828 | for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) { |
2e5dc010 N |
9829 | void **sp = *space_list; |
9830 | struct imsm_dev *newdev; | |
9831 | struct imsm_map *newmap, *oldmap; | |
9832 | ||
d098291a AK |
9833 | for (id = super->devlist ; id; id = id->next) { |
9834 | if (id->index == dev_id) | |
9835 | break; | |
9836 | } | |
9837 | if (id == NULL) | |
9838 | break; | |
2e5dc010 N |
9839 | if (!sp) |
9840 | continue; | |
9841 | *space_list = *sp; | |
9842 | newdev = (void*)sp; | |
9843 | /* Copy the dev, but not (all of) the map */ | |
9844 | memcpy(newdev, id->dev, sizeof(*newdev)); | |
238c0a71 AK |
9845 | oldmap = get_imsm_map(id->dev, MAP_0); |
9846 | newmap = get_imsm_map(newdev, MAP_0); | |
2e5dc010 N |
9847 | /* Copy the current map */ |
9848 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); | |
9849 | /* update one device only | |
9850 | */ | |
9851 | if (devices_to_reshape) { | |
ed7333bd AK |
9852 | dprintf("imsm: modifying subdev: %i\n", |
9853 | id->index); | |
2e5dc010 N |
9854 | devices_to_reshape--; |
9855 | newdev->vol.migr_state = 1; | |
4036e7ee | 9856 | set_vol_curr_migr_unit(newdev, 0); |
ea672ee1 | 9857 | set_migr_type(newdev, MIGR_GEN_MIGR); |
2e5dc010 N |
9858 | newmap->num_members = u->new_raid_disks; |
9859 | for (i = 0; i < delta_disks; i++) { | |
9860 | set_imsm_ord_tbl_ent(newmap, | |
9861 | u->old_raid_disks + i, | |
9862 | u->old_raid_disks + i); | |
9863 | } | |
9864 | /* New map is correct, now need to save old map | |
9865 | */ | |
238c0a71 | 9866 | newmap = get_imsm_map(newdev, MAP_1); |
2e5dc010 N |
9867 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); |
9868 | ||
f3871fdc | 9869 | imsm_set_array_size(newdev, -1); |
2e5dc010 N |
9870 | } |
9871 | ||
9872 | sp = (void **)id->dev; | |
9873 | id->dev = newdev; | |
9874 | *sp = tofree; | |
9875 | tofree = sp; | |
8e59f3d8 AK |
9876 | |
9877 | /* Clear migration record */ | |
9878 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
2e5dc010 | 9879 | } |
819bc634 AK |
9880 | if (tofree) |
9881 | *space_list = tofree; | |
2e5dc010 N |
9882 | ret_val = 1; |
9883 | ||
9884 | update_reshape_exit: | |
9885 | ||
9886 | return ret_val; | |
9887 | } | |
9888 | ||
bb025c2f | 9889 | static int apply_takeover_update(struct imsm_update_takeover *u, |
8ca6df95 KW |
9890 | struct intel_super *super, |
9891 | void ***space_list) | |
bb025c2f KW |
9892 | { |
9893 | struct imsm_dev *dev = NULL; | |
8ca6df95 KW |
9894 | struct intel_dev *dv; |
9895 | struct imsm_dev *dev_new; | |
bb025c2f KW |
9896 | struct imsm_map *map; |
9897 | struct dl *dm, *du; | |
8ca6df95 | 9898 | int i; |
bb025c2f KW |
9899 | |
9900 | for (dv = super->devlist; dv; dv = dv->next) | |
9901 | if (dv->index == (unsigned int)u->subarray) { | |
9902 | dev = dv->dev; | |
9903 | break; | |
9904 | } | |
9905 | ||
9906 | if (dev == NULL) | |
9907 | return 0; | |
9908 | ||
238c0a71 | 9909 | map = get_imsm_map(dev, MAP_0); |
bb025c2f KW |
9910 | |
9911 | if (u->direction == R10_TO_R0) { | |
43d5ec18 | 9912 | /* Number of failed disks must be half of initial disk number */ |
3b451610 AK |
9913 | if (imsm_count_failed(super, dev, MAP_0) != |
9914 | (map->num_members / 2)) | |
43d5ec18 KW |
9915 | return 0; |
9916 | ||
bb025c2f KW |
9917 | /* iterate through devices to mark removed disks as spare */ |
9918 | for (dm = super->disks; dm; dm = dm->next) { | |
9919 | if (dm->disk.status & FAILED_DISK) { | |
9920 | int idx = dm->index; | |
9921 | /* update indexes on the disk list */ | |
9922 | /* FIXME this loop-with-the-loop looks wrong, I'm not convinced | |
9923 | the index values will end up being correct.... NB */ | |
9924 | for (du = super->disks; du; du = du->next) | |
9925 | if (du->index > idx) | |
9926 | du->index--; | |
9927 | /* mark as spare disk */ | |
a8619d23 | 9928 | mark_spare(dm); |
bb025c2f KW |
9929 | } |
9930 | } | |
bb025c2f | 9931 | /* update map */ |
1c275381 | 9932 | map->num_members /= map->num_domains; |
bb025c2f | 9933 | map->map_state = IMSM_T_STATE_NORMAL; |
bb025c2f | 9934 | map->raid_level = 0; |
1c275381 MT |
9935 | set_num_domains(map); |
9936 | update_num_data_stripes(map, imsm_dev_size(dev)); | |
bb025c2f KW |
9937 | map->failed_disk_num = -1; |
9938 | } | |
9939 | ||
8ca6df95 KW |
9940 | if (u->direction == R0_TO_R10) { |
9941 | void **space; | |
4a353e6e | 9942 | |
8ca6df95 KW |
9943 | /* update slots in current disk list */ |
9944 | for (dm = super->disks; dm; dm = dm->next) { | |
9945 | if (dm->index >= 0) | |
9946 | dm->index *= 2; | |
9947 | } | |
9948 | /* create new *missing* disks */ | |
9949 | for (i = 0; i < map->num_members; i++) { | |
9950 | space = *space_list; | |
9951 | if (!space) | |
9952 | continue; | |
9953 | *space_list = *space; | |
9954 | du = (void *)space; | |
9955 | memcpy(du, super->disks, sizeof(*du)); | |
8ca6df95 KW |
9956 | du->fd = -1; |
9957 | du->minor = 0; | |
9958 | du->major = 0; | |
9959 | du->index = (i * 2) + 1; | |
9960 | sprintf((char *)du->disk.serial, | |
9961 | " MISSING_%d", du->index); | |
9962 | sprintf((char *)du->serial, | |
9963 | "MISSING_%d", du->index); | |
9964 | du->next = super->missing; | |
9965 | super->missing = du; | |
9966 | } | |
9967 | /* create new dev and map */ | |
9968 | space = *space_list; | |
9969 | if (!space) | |
9970 | return 0; | |
9971 | *space_list = *space; | |
9972 | dev_new = (void *)space; | |
9973 | memcpy(dev_new, dev, sizeof(*dev)); | |
9974 | /* update new map */ | |
238c0a71 | 9975 | map = get_imsm_map(dev_new, MAP_0); |
1c275381 | 9976 | |
1a2487c2 | 9977 | map->map_state = IMSM_T_STATE_DEGRADED; |
8ca6df95 | 9978 | map->raid_level = 1; |
1c275381 MT |
9979 | set_num_domains(map); |
9980 | map->num_members = map->num_members * map->num_domains; | |
9981 | update_num_data_stripes(map, imsm_dev_size(dev)); | |
4a353e6e | 9982 | |
8ca6df95 KW |
9983 | /* replace dev<->dev_new */ |
9984 | dv->dev = dev_new; | |
9985 | } | |
bb025c2f KW |
9986 | /* update disk order table */ |
9987 | for (du = super->disks; du; du = du->next) | |
9988 | if (du->index >= 0) | |
9989 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
8ca6df95 | 9990 | for (du = super->missing; du; du = du->next) |
1a2487c2 KW |
9991 | if (du->index >= 0) { |
9992 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
4c9e8c1e | 9993 | mark_missing(super, dv->dev, &du->disk, du->index); |
1a2487c2 | 9994 | } |
bb025c2f KW |
9995 | |
9996 | return 1; | |
9997 | } | |
9998 | ||
e8319a19 DW |
9999 | static void imsm_process_update(struct supertype *st, |
10000 | struct metadata_update *update) | |
10001 | { | |
10002 | /** | |
10003 | * crack open the metadata_update envelope to find the update record | |
10004 | * update can be one of: | |
d195167d AK |
10005 | * update_reshape_container_disks - all the arrays in the container |
10006 | * are being reshaped to have more devices. We need to mark | |
10007 | * the arrays for general migration and convert selected spares | |
10008 | * into active devices. | |
10009 | * update_activate_spare - a spare device has replaced a failed | |
1011e834 N |
10010 | * device in an array, update the disk_ord_tbl. If this disk is |
10011 | * present in all member arrays then also clear the SPARE_DISK | |
10012 | * flag | |
d195167d AK |
10013 | * update_create_array |
10014 | * update_kill_array | |
10015 | * update_rename_array | |
10016 | * update_add_remove_disk | |
e8319a19 DW |
10017 | */ |
10018 | struct intel_super *super = st->sb; | |
4d7b1503 | 10019 | struct imsm_super *mpb; |
e8319a19 DW |
10020 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
10021 | ||
4d7b1503 DW |
10022 | /* update requires a larger buf but the allocation failed */ |
10023 | if (super->next_len && !super->next_buf) { | |
10024 | super->next_len = 0; | |
10025 | return; | |
10026 | } | |
10027 | ||
10028 | if (super->next_buf) { | |
10029 | memcpy(super->next_buf, super->buf, super->len); | |
10030 | free(super->buf); | |
10031 | super->len = super->next_len; | |
10032 | super->buf = super->next_buf; | |
10033 | ||
10034 | super->next_len = 0; | |
10035 | super->next_buf = NULL; | |
10036 | } | |
10037 | ||
10038 | mpb = super->anchor; | |
10039 | ||
e8319a19 | 10040 | switch (type) { |
0ec5d470 AK |
10041 | case update_general_migration_checkpoint: { |
10042 | struct intel_dev *id; | |
10043 | struct imsm_update_general_migration_checkpoint *u = | |
10044 | (void *)update->buf; | |
10045 | ||
1ade5cc1 | 10046 | dprintf("called for update_general_migration_checkpoint\n"); |
0ec5d470 AK |
10047 | |
10048 | /* find device under general migration */ | |
10049 | for (id = super->devlist ; id; id = id->next) { | |
10050 | if (is_gen_migration(id->dev)) { | |
4036e7ee MT |
10051 | set_vol_curr_migr_unit(id->dev, |
10052 | u->curr_migr_unit); | |
0ec5d470 AK |
10053 | super->updates_pending++; |
10054 | } | |
10055 | } | |
10056 | break; | |
10057 | } | |
bb025c2f KW |
10058 | case update_takeover: { |
10059 | struct imsm_update_takeover *u = (void *)update->buf; | |
1a2487c2 KW |
10060 | if (apply_takeover_update(u, super, &update->space_list)) { |
10061 | imsm_update_version_info(super); | |
bb025c2f | 10062 | super->updates_pending++; |
1a2487c2 | 10063 | } |
bb025c2f KW |
10064 | break; |
10065 | } | |
10066 | ||
78b10e66 | 10067 | case update_reshape_container_disks: { |
d195167d | 10068 | struct imsm_update_reshape *u = (void *)update->buf; |
2e5dc010 N |
10069 | if (apply_reshape_container_disks_update( |
10070 | u, super, &update->space_list)) | |
10071 | super->updates_pending++; | |
78b10e66 N |
10072 | break; |
10073 | } | |
48c5303a | 10074 | case update_reshape_migration: { |
a29911da PC |
10075 | struct imsm_update_reshape_migration *u = (void *)update->buf; |
10076 | if (apply_reshape_migration_update( | |
10077 | u, super, &update->space_list)) | |
10078 | super->updates_pending++; | |
48c5303a PC |
10079 | break; |
10080 | } | |
f3871fdc AK |
10081 | case update_size_change: { |
10082 | struct imsm_update_size_change *u = (void *)update->buf; | |
10083 | if (apply_size_change_update(u, super)) | |
10084 | super->updates_pending++; | |
10085 | break; | |
10086 | } | |
e8319a19 | 10087 | case update_activate_spare: { |
1011e834 | 10088 | struct imsm_update_activate_spare *u = (void *) update->buf; |
69d40de4 JR |
10089 | |
10090 | if (prepare_spare_to_activate(st, u) && | |
10091 | apply_update_activate_spare(u, super, st->arrays)) | |
061d7da3 | 10092 | super->updates_pending++; |
8273f55e DW |
10093 | break; |
10094 | } | |
10095 | case update_create_array: { | |
10096 | /* someone wants to create a new array, we need to be aware of | |
10097 | * a few races/collisions: | |
10098 | * 1/ 'Create' called by two separate instances of mdadm | |
10099 | * 2/ 'Create' versus 'activate_spare': mdadm has chosen | |
10100 | * devices that have since been assimilated via | |
10101 | * activate_spare. | |
10102 | * In the event this update can not be carried out mdadm will | |
10103 | * (FIX ME) notice that its update did not take hold. | |
10104 | */ | |
10105 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 10106 | struct intel_dev *dv; |
8273f55e DW |
10107 | struct imsm_dev *dev; |
10108 | struct imsm_map *map, *new_map; | |
10109 | unsigned long long start, end; | |
10110 | unsigned long long new_start, new_end; | |
10111 | int i; | |
54c2c1ea DW |
10112 | struct disk_info *inf; |
10113 | struct dl *dl; | |
8273f55e DW |
10114 | |
10115 | /* handle racing creates: first come first serve */ | |
10116 | if (u->dev_idx < mpb->num_raid_devs) { | |
1ade5cc1 | 10117 | dprintf("subarray %d already defined\n", u->dev_idx); |
ba2de7ba | 10118 | goto create_error; |
8273f55e DW |
10119 | } |
10120 | ||
10121 | /* check update is next in sequence */ | |
10122 | if (u->dev_idx != mpb->num_raid_devs) { | |
1ade5cc1 N |
10123 | dprintf("can not create array %d expected index %d\n", |
10124 | u->dev_idx, mpb->num_raid_devs); | |
ba2de7ba | 10125 | goto create_error; |
8273f55e DW |
10126 | } |
10127 | ||
238c0a71 | 10128 | new_map = get_imsm_map(&u->dev, MAP_0); |
5551b113 | 10129 | new_start = pba_of_lba0(new_map); |
44490938 | 10130 | new_end = new_start + per_dev_array_size(new_map); |
54c2c1ea | 10131 | inf = get_disk_info(u); |
8273f55e DW |
10132 | |
10133 | /* handle activate_spare versus create race: | |
10134 | * check to make sure that overlapping arrays do not include | |
10135 | * overalpping disks | |
10136 | */ | |
10137 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
949c47a0 | 10138 | dev = get_imsm_dev(super, i); |
238c0a71 | 10139 | map = get_imsm_map(dev, MAP_0); |
5551b113 | 10140 | start = pba_of_lba0(map); |
44490938 | 10141 | end = start + per_dev_array_size(map); |
8273f55e DW |
10142 | if ((new_start >= start && new_start <= end) || |
10143 | (start >= new_start && start <= new_end)) | |
54c2c1ea DW |
10144 | /* overlap */; |
10145 | else | |
10146 | continue; | |
10147 | ||
10148 | if (disks_overlap(super, i, u)) { | |
1ade5cc1 | 10149 | dprintf("arrays overlap\n"); |
ba2de7ba | 10150 | goto create_error; |
8273f55e DW |
10151 | } |
10152 | } | |
8273f55e | 10153 | |
949c47a0 DW |
10154 | /* check that prepare update was successful */ |
10155 | if (!update->space) { | |
1ade5cc1 | 10156 | dprintf("prepare update failed\n"); |
ba2de7ba | 10157 | goto create_error; |
949c47a0 DW |
10158 | } |
10159 | ||
54c2c1ea DW |
10160 | /* check that all disks are still active before committing |
10161 | * changes. FIXME: could we instead handle this by creating a | |
10162 | * degraded array? That's probably not what the user expects, | |
10163 | * so better to drop this update on the floor. | |
10164 | */ | |
10165 | for (i = 0; i < new_map->num_members; i++) { | |
10166 | dl = serial_to_dl(inf[i].serial, super); | |
10167 | if (!dl) { | |
1ade5cc1 | 10168 | dprintf("disk disappeared\n"); |
ba2de7ba | 10169 | goto create_error; |
54c2c1ea | 10170 | } |
949c47a0 DW |
10171 | } |
10172 | ||
8273f55e | 10173 | super->updates_pending++; |
54c2c1ea DW |
10174 | |
10175 | /* convert spares to members and fixup ord_tbl */ | |
10176 | for (i = 0; i < new_map->num_members; i++) { | |
10177 | dl = serial_to_dl(inf[i].serial, super); | |
10178 | if (dl->index == -1) { | |
10179 | dl->index = mpb->num_disks; | |
10180 | mpb->num_disks++; | |
10181 | dl->disk.status |= CONFIGURED_DISK; | |
10182 | dl->disk.status &= ~SPARE_DISK; | |
10183 | } | |
10184 | set_imsm_ord_tbl_ent(new_map, i, dl->index); | |
10185 | } | |
10186 | ||
ba2de7ba DW |
10187 | dv = update->space; |
10188 | dev = dv->dev; | |
949c47a0 DW |
10189 | update->space = NULL; |
10190 | imsm_copy_dev(dev, &u->dev); | |
ba2de7ba DW |
10191 | dv->index = u->dev_idx; |
10192 | dv->next = super->devlist; | |
10193 | super->devlist = dv; | |
8273f55e | 10194 | mpb->num_raid_devs++; |
8273f55e | 10195 | |
4d1313e9 | 10196 | imsm_update_version_info(super); |
8273f55e | 10197 | break; |
ba2de7ba DW |
10198 | create_error: |
10199 | /* mdmon knows how to release update->space, but not | |
10200 | * ((struct intel_dev *) update->space)->dev | |
10201 | */ | |
10202 | if (update->space) { | |
10203 | dv = update->space; | |
10204 | free(dv->dev); | |
10205 | } | |
8273f55e | 10206 | break; |
e8319a19 | 10207 | } |
33414a01 DW |
10208 | case update_kill_array: { |
10209 | struct imsm_update_kill_array *u = (void *) update->buf; | |
10210 | int victim = u->dev_idx; | |
10211 | struct active_array *a; | |
10212 | struct intel_dev **dp; | |
33414a01 DW |
10213 | |
10214 | /* sanity check that we are not affecting the uuid of | |
10215 | * active arrays, or deleting an active array | |
10216 | * | |
10217 | * FIXME when immutable ids are available, but note that | |
10218 | * we'll also need to fixup the invalidated/active | |
10219 | * subarray indexes in mdstat | |
10220 | */ | |
10221 | for (a = st->arrays; a; a = a->next) | |
10222 | if (a->info.container_member >= victim) | |
10223 | break; | |
10224 | /* by definition if mdmon is running at least one array | |
10225 | * is active in the container, so checking | |
10226 | * mpb->num_raid_devs is just extra paranoia | |
10227 | */ | |
756a15f3 | 10228 | if (a || mpb->num_raid_devs == 1 || victim >= super->anchor->num_raid_devs) { |
33414a01 DW |
10229 | dprintf("failed to delete subarray-%d\n", victim); |
10230 | break; | |
10231 | } | |
10232 | ||
10233 | for (dp = &super->devlist; *dp;) | |
f21e18ca | 10234 | if ((*dp)->index == (unsigned)super->current_vol) { |
33414a01 DW |
10235 | *dp = (*dp)->next; |
10236 | } else { | |
f21e18ca | 10237 | if ((*dp)->index > (unsigned)victim) |
33414a01 DW |
10238 | (*dp)->index--; |
10239 | dp = &(*dp)->next; | |
10240 | } | |
10241 | mpb->num_raid_devs--; | |
10242 | super->updates_pending++; | |
10243 | break; | |
10244 | } | |
aa534678 DW |
10245 | case update_rename_array: { |
10246 | struct imsm_update_rename_array *u = (void *) update->buf; | |
10247 | char name[MAX_RAID_SERIAL_LEN+1]; | |
10248 | int target = u->dev_idx; | |
10249 | struct active_array *a; | |
10250 | struct imsm_dev *dev; | |
10251 | ||
10252 | /* sanity check that we are not affecting the uuid of | |
10253 | * an active array | |
10254 | */ | |
40659392 | 10255 | memset(name, 0, sizeof(name)); |
aa534678 DW |
10256 | snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name); |
10257 | name[MAX_RAID_SERIAL_LEN] = '\0'; | |
10258 | for (a = st->arrays; a; a = a->next) | |
10259 | if (a->info.container_member == target) | |
10260 | break; | |
10261 | dev = get_imsm_dev(super, u->dev_idx); | |
e2eb503b MT |
10262 | |
10263 | if (a || !dev || imsm_is_name_allowed(super, name, 0) == false) { | |
aa534678 DW |
10264 | dprintf("failed to rename subarray-%d\n", target); |
10265 | break; | |
10266 | } | |
10267 | ||
40659392 | 10268 | memcpy(dev->volume, name, MAX_RAID_SERIAL_LEN); |
aa534678 DW |
10269 | super->updates_pending++; |
10270 | break; | |
10271 | } | |
1a64be56 | 10272 | case update_add_remove_disk: { |
43dad3d6 | 10273 | /* we may be able to repair some arrays if disks are |
095b8088 | 10274 | * being added, check the status of add_remove_disk |
1a64be56 LM |
10275 | * if discs has been added. |
10276 | */ | |
10277 | if (add_remove_disk_update(super)) { | |
43dad3d6 | 10278 | struct active_array *a; |
072b727f DW |
10279 | |
10280 | super->updates_pending++; | |
1a64be56 | 10281 | for (a = st->arrays; a; a = a->next) |
43dad3d6 DW |
10282 | a->check_degraded = 1; |
10283 | } | |
43dad3d6 | 10284 | break; |
e8319a19 | 10285 | } |
bbab0940 TM |
10286 | case update_prealloc_badblocks_mem: |
10287 | break; | |
e6e9dd3f AP |
10288 | case update_rwh_policy: { |
10289 | struct imsm_update_rwh_policy *u = (void *)update->buf; | |
10290 | int target = u->dev_idx; | |
10291 | struct imsm_dev *dev = get_imsm_dev(super, target); | |
e6e9dd3f AP |
10292 | |
10293 | if (dev->rwh_policy != u->new_policy) { | |
10294 | dev->rwh_policy = u->new_policy; | |
10295 | super->updates_pending++; | |
10296 | } | |
10297 | break; | |
10298 | } | |
1a64be56 | 10299 | default: |
ebf3be99 | 10300 | pr_err("error: unsupported process update type:(type: %d)\n", type); |
1a64be56 | 10301 | } |
e8319a19 | 10302 | } |
88758e9d | 10303 | |
bc0b9d34 PC |
10304 | static struct mdinfo *get_spares_for_grow(struct supertype *st); |
10305 | ||
5fe6f031 N |
10306 | static int imsm_prepare_update(struct supertype *st, |
10307 | struct metadata_update *update) | |
8273f55e | 10308 | { |
949c47a0 | 10309 | /** |
4d7b1503 DW |
10310 | * Allocate space to hold new disk entries, raid-device entries or a new |
10311 | * mpb if necessary. The manager synchronously waits for updates to | |
10312 | * complete in the monitor, so new mpb buffers allocated here can be | |
10313 | * integrated by the monitor thread without worrying about live pointers | |
10314 | * in the manager thread. | |
8273f55e | 10315 | */ |
095b8088 | 10316 | enum imsm_update_type type; |
4d7b1503 | 10317 | struct intel_super *super = st->sb; |
f36a9ecd | 10318 | unsigned int sector_size = super->sector_size; |
4d7b1503 DW |
10319 | struct imsm_super *mpb = super->anchor; |
10320 | size_t buf_len; | |
10321 | size_t len = 0; | |
949c47a0 | 10322 | |
095b8088 N |
10323 | if (update->len < (int)sizeof(type)) |
10324 | return 0; | |
10325 | ||
10326 | type = *(enum imsm_update_type *) update->buf; | |
10327 | ||
949c47a0 | 10328 | switch (type) { |
0ec5d470 | 10329 | case update_general_migration_checkpoint: |
095b8088 N |
10330 | if (update->len < (int)sizeof(struct imsm_update_general_migration_checkpoint)) |
10331 | return 0; | |
1ade5cc1 | 10332 | dprintf("called for update_general_migration_checkpoint\n"); |
0ec5d470 | 10333 | break; |
abedf5fc KW |
10334 | case update_takeover: { |
10335 | struct imsm_update_takeover *u = (void *)update->buf; | |
095b8088 N |
10336 | if (update->len < (int)sizeof(*u)) |
10337 | return 0; | |
abedf5fc KW |
10338 | if (u->direction == R0_TO_R10) { |
10339 | void **tail = (void **)&update->space_list; | |
10340 | struct imsm_dev *dev = get_imsm_dev(super, u->subarray); | |
238c0a71 | 10341 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
abedf5fc KW |
10342 | int num_members = map->num_members; |
10343 | void *space; | |
10344 | int size, i; | |
abedf5fc KW |
10345 | /* allocate memory for added disks */ |
10346 | for (i = 0; i < num_members; i++) { | |
10347 | size = sizeof(struct dl); | |
503975b9 | 10348 | space = xmalloc(size); |
abedf5fc KW |
10349 | *tail = space; |
10350 | tail = space; | |
10351 | *tail = NULL; | |
10352 | } | |
10353 | /* allocate memory for new device */ | |
10354 | size = sizeof_imsm_dev(super->devlist->dev, 0) + | |
10355 | (num_members * sizeof(__u32)); | |
503975b9 N |
10356 | space = xmalloc(size); |
10357 | *tail = space; | |
10358 | tail = space; | |
10359 | *tail = NULL; | |
10360 | len = disks_to_mpb_size(num_members * 2); | |
abedf5fc KW |
10361 | } |
10362 | ||
10363 | break; | |
10364 | } | |
78b10e66 | 10365 | case update_reshape_container_disks: { |
d195167d AK |
10366 | /* Every raid device in the container is about to |
10367 | * gain some more devices, and we will enter a | |
10368 | * reconfiguration. | |
10369 | * So each 'imsm_map' will be bigger, and the imsm_vol | |
10370 | * will now hold 2 of them. | |
10371 | * Thus we need new 'struct imsm_dev' allocations sized | |
10372 | * as sizeof_imsm_dev but with more devices in both maps. | |
10373 | */ | |
10374 | struct imsm_update_reshape *u = (void *)update->buf; | |
10375 | struct intel_dev *dl; | |
10376 | void **space_tail = (void**)&update->space_list; | |
10377 | ||
095b8088 N |
10378 | if (update->len < (int)sizeof(*u)) |
10379 | return 0; | |
10380 | ||
1ade5cc1 | 10381 | dprintf("for update_reshape\n"); |
d195167d AK |
10382 | |
10383 | for (dl = super->devlist; dl; dl = dl->next) { | |
10384 | int size = sizeof_imsm_dev(dl->dev, 1); | |
10385 | void *s; | |
d677e0b8 AK |
10386 | if (u->new_raid_disks > u->old_raid_disks) |
10387 | size += sizeof(__u32)*2* | |
10388 | (u->new_raid_disks - u->old_raid_disks); | |
503975b9 | 10389 | s = xmalloc(size); |
d195167d AK |
10390 | *space_tail = s; |
10391 | space_tail = s; | |
10392 | *space_tail = NULL; | |
10393 | } | |
10394 | ||
10395 | len = disks_to_mpb_size(u->new_raid_disks); | |
10396 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
78b10e66 N |
10397 | break; |
10398 | } | |
48c5303a | 10399 | case update_reshape_migration: { |
bc0b9d34 PC |
10400 | /* for migration level 0->5 we need to add disks |
10401 | * so the same as for container operation we will copy | |
10402 | * device to the bigger location. | |
10403 | * in memory prepared device and new disk area are prepared | |
10404 | * for usage in process update | |
10405 | */ | |
10406 | struct imsm_update_reshape_migration *u = (void *)update->buf; | |
10407 | struct intel_dev *id; | |
10408 | void **space_tail = (void **)&update->space_list; | |
10409 | int size; | |
10410 | void *s; | |
10411 | int current_level = -1; | |
10412 | ||
095b8088 N |
10413 | if (update->len < (int)sizeof(*u)) |
10414 | return 0; | |
10415 | ||
1ade5cc1 | 10416 | dprintf("for update_reshape\n"); |
bc0b9d34 PC |
10417 | |
10418 | /* add space for bigger array in update | |
10419 | */ | |
10420 | for (id = super->devlist; id; id = id->next) { | |
10421 | if (id->index == (unsigned)u->subdev) { | |
10422 | size = sizeof_imsm_dev(id->dev, 1); | |
10423 | if (u->new_raid_disks > u->old_raid_disks) | |
10424 | size += sizeof(__u32)*2* | |
10425 | (u->new_raid_disks - u->old_raid_disks); | |
503975b9 | 10426 | s = xmalloc(size); |
bc0b9d34 PC |
10427 | *space_tail = s; |
10428 | space_tail = s; | |
10429 | *space_tail = NULL; | |
10430 | break; | |
10431 | } | |
10432 | } | |
10433 | if (update->space_list == NULL) | |
10434 | break; | |
10435 | ||
10436 | /* add space for disk in update | |
10437 | */ | |
10438 | size = sizeof(struct dl); | |
503975b9 | 10439 | s = xmalloc(size); |
bc0b9d34 PC |
10440 | *space_tail = s; |
10441 | space_tail = s; | |
10442 | *space_tail = NULL; | |
10443 | ||
10444 | /* add spare device to update | |
10445 | */ | |
10446 | for (id = super->devlist ; id; id = id->next) | |
10447 | if (id->index == (unsigned)u->subdev) { | |
10448 | struct imsm_dev *dev; | |
10449 | struct imsm_map *map; | |
10450 | ||
10451 | dev = get_imsm_dev(super, u->subdev); | |
238c0a71 | 10452 | map = get_imsm_map(dev, MAP_0); |
bc0b9d34 PC |
10453 | current_level = map->raid_level; |
10454 | break; | |
10455 | } | |
089f9d79 | 10456 | if (u->new_level == 5 && u->new_level != current_level) { |
bc0b9d34 PC |
10457 | struct mdinfo *spares; |
10458 | ||
10459 | spares = get_spares_for_grow(st); | |
10460 | if (spares) { | |
10461 | struct dl *dl; | |
10462 | struct mdinfo *dev; | |
10463 | ||
10464 | dev = spares->devs; | |
10465 | if (dev) { | |
10466 | u->new_disks[0] = | |
10467 | makedev(dev->disk.major, | |
10468 | dev->disk.minor); | |
10469 | dl = get_disk_super(super, | |
10470 | dev->disk.major, | |
10471 | dev->disk.minor); | |
10472 | dl->index = u->old_raid_disks; | |
10473 | dev = dev->next; | |
10474 | } | |
10475 | sysfs_free(spares); | |
10476 | } | |
10477 | } | |
10478 | len = disks_to_mpb_size(u->new_raid_disks); | |
10479 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
48c5303a PC |
10480 | break; |
10481 | } | |
f3871fdc | 10482 | case update_size_change: { |
095b8088 N |
10483 | if (update->len < (int)sizeof(struct imsm_update_size_change)) |
10484 | return 0; | |
10485 | break; | |
10486 | } | |
10487 | case update_activate_spare: { | |
10488 | if (update->len < (int)sizeof(struct imsm_update_activate_spare)) | |
10489 | return 0; | |
f3871fdc AK |
10490 | break; |
10491 | } | |
949c47a0 DW |
10492 | case update_create_array: { |
10493 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 10494 | struct intel_dev *dv; |
54c2c1ea | 10495 | struct imsm_dev *dev = &u->dev; |
238c0a71 | 10496 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
54c2c1ea DW |
10497 | struct dl *dl; |
10498 | struct disk_info *inf; | |
10499 | int i; | |
10500 | int activate = 0; | |
949c47a0 | 10501 | |
095b8088 N |
10502 | if (update->len < (int)sizeof(*u)) |
10503 | return 0; | |
10504 | ||
54c2c1ea DW |
10505 | inf = get_disk_info(u); |
10506 | len = sizeof_imsm_dev(dev, 1); | |
ba2de7ba | 10507 | /* allocate a new super->devlist entry */ |
503975b9 N |
10508 | dv = xmalloc(sizeof(*dv)); |
10509 | dv->dev = xmalloc(len); | |
10510 | update->space = dv; | |
949c47a0 | 10511 | |
54c2c1ea DW |
10512 | /* count how many spares will be converted to members */ |
10513 | for (i = 0; i < map->num_members; i++) { | |
10514 | dl = serial_to_dl(inf[i].serial, super); | |
10515 | if (!dl) { | |
10516 | /* hmm maybe it failed?, nothing we can do about | |
10517 | * it here | |
10518 | */ | |
10519 | continue; | |
10520 | } | |
10521 | if (count_memberships(dl, super) == 0) | |
10522 | activate++; | |
10523 | } | |
10524 | len += activate * sizeof(struct imsm_disk); | |
949c47a0 | 10525 | break; |
095b8088 N |
10526 | } |
10527 | case update_kill_array: { | |
10528 | if (update->len < (int)sizeof(struct imsm_update_kill_array)) | |
10529 | return 0; | |
949c47a0 DW |
10530 | break; |
10531 | } | |
095b8088 N |
10532 | case update_rename_array: { |
10533 | if (update->len < (int)sizeof(struct imsm_update_rename_array)) | |
10534 | return 0; | |
10535 | break; | |
10536 | } | |
10537 | case update_add_remove_disk: | |
10538 | /* no update->len needed */ | |
10539 | break; | |
bbab0940 TM |
10540 | case update_prealloc_badblocks_mem: |
10541 | super->extra_space += sizeof(struct bbm_log) - | |
10542 | get_imsm_bbm_log_size(super->bbm_log); | |
10543 | break; | |
e6e9dd3f AP |
10544 | case update_rwh_policy: { |
10545 | if (update->len < (int)sizeof(struct imsm_update_rwh_policy)) | |
10546 | return 0; | |
10547 | break; | |
10548 | } | |
095b8088 N |
10549 | default: |
10550 | return 0; | |
949c47a0 | 10551 | } |
8273f55e | 10552 | |
4d7b1503 DW |
10553 | /* check if we need a larger metadata buffer */ |
10554 | if (super->next_buf) | |
10555 | buf_len = super->next_len; | |
10556 | else | |
10557 | buf_len = super->len; | |
10558 | ||
bbab0940 | 10559 | if (__le32_to_cpu(mpb->mpb_size) + super->extra_space + len > buf_len) { |
4d7b1503 DW |
10560 | /* ok we need a larger buf than what is currently allocated |
10561 | * if this allocation fails process_update will notice that | |
10562 | * ->next_len is set and ->next_buf is NULL | |
10563 | */ | |
bbab0940 TM |
10564 | buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + |
10565 | super->extra_space + len, sector_size); | |
4d7b1503 DW |
10566 | if (super->next_buf) |
10567 | free(super->next_buf); | |
10568 | ||
10569 | super->next_len = buf_len; | |
f36a9ecd | 10570 | if (posix_memalign(&super->next_buf, sector_size, buf_len) == 0) |
1f45a8ad DW |
10571 | memset(super->next_buf, 0, buf_len); |
10572 | else | |
4d7b1503 DW |
10573 | super->next_buf = NULL; |
10574 | } | |
5fe6f031 | 10575 | return 1; |
8273f55e DW |
10576 | } |
10577 | ||
ae6aad82 | 10578 | /* must be called while manager is quiesced */ |
f21e18ca | 10579 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index) |
ae6aad82 DW |
10580 | { |
10581 | struct imsm_super *mpb = super->anchor; | |
ae6aad82 DW |
10582 | struct dl *iter; |
10583 | struct imsm_dev *dev; | |
10584 | struct imsm_map *map; | |
4c9e8c1e | 10585 | unsigned int i, j, num_members; |
fb12a745 | 10586 | __u32 ord, ord_map0; |
4c9e8c1e | 10587 | struct bbm_log *log = super->bbm_log; |
ae6aad82 | 10588 | |
1ade5cc1 | 10589 | dprintf("deleting device[%d] from imsm_super\n", index); |
ae6aad82 DW |
10590 | |
10591 | /* shift all indexes down one */ | |
10592 | for (iter = super->disks; iter; iter = iter->next) | |
f21e18ca | 10593 | if (iter->index > (int)index) |
ae6aad82 | 10594 | iter->index--; |
47ee5a45 | 10595 | for (iter = super->missing; iter; iter = iter->next) |
f21e18ca | 10596 | if (iter->index > (int)index) |
47ee5a45 | 10597 | iter->index--; |
ae6aad82 DW |
10598 | |
10599 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
10600 | dev = get_imsm_dev(super, i); | |
238c0a71 | 10601 | map = get_imsm_map(dev, MAP_0); |
24565c9a DW |
10602 | num_members = map->num_members; |
10603 | for (j = 0; j < num_members; j++) { | |
10604 | /* update ord entries being careful not to propagate | |
10605 | * ord-flags to the first map | |
10606 | */ | |
238c0a71 | 10607 | ord = get_imsm_ord_tbl_ent(dev, j, MAP_X); |
fb12a745 | 10608 | ord_map0 = get_imsm_ord_tbl_ent(dev, j, MAP_0); |
ae6aad82 | 10609 | |
24565c9a DW |
10610 | if (ord_to_idx(ord) <= index) |
10611 | continue; | |
ae6aad82 | 10612 | |
238c0a71 | 10613 | map = get_imsm_map(dev, MAP_0); |
fb12a745 | 10614 | set_imsm_ord_tbl_ent(map, j, ord_map0 - 1); |
238c0a71 | 10615 | map = get_imsm_map(dev, MAP_1); |
24565c9a DW |
10616 | if (map) |
10617 | set_imsm_ord_tbl_ent(map, j, ord - 1); | |
ae6aad82 DW |
10618 | } |
10619 | } | |
10620 | ||
4c9e8c1e TM |
10621 | for (i = 0; i < log->entry_count; i++) { |
10622 | struct bbm_log_entry *entry = &log->marked_block_entries[i]; | |
10623 | ||
10624 | if (entry->disk_ordinal <= index) | |
10625 | continue; | |
10626 | entry->disk_ordinal--; | |
10627 | } | |
10628 | ||
ae6aad82 DW |
10629 | mpb->num_disks--; |
10630 | super->updates_pending++; | |
24565c9a DW |
10631 | if (*dlp) { |
10632 | struct dl *dl = *dlp; | |
10633 | ||
10634 | *dlp = (*dlp)->next; | |
3a85bf0e | 10635 | __free_imsm_disk(dl, 1); |
24565c9a | 10636 | } |
ae6aad82 | 10637 | } |
9a717282 | 10638 | |
9a717282 AK |
10639 | static int imsm_get_allowed_degradation(int level, int raid_disks, |
10640 | struct intel_super *super, | |
10641 | struct imsm_dev *dev) | |
10642 | { | |
10643 | switch (level) { | |
bf5cf7c7 | 10644 | case 1: |
9a717282 AK |
10645 | case 10:{ |
10646 | int ret_val = 0; | |
10647 | struct imsm_map *map; | |
10648 | int i; | |
10649 | ||
10650 | ret_val = raid_disks/2; | |
10651 | /* check map if all disks pairs not failed | |
10652 | * in both maps | |
10653 | */ | |
238c0a71 | 10654 | map = get_imsm_map(dev, MAP_0); |
9a717282 AK |
10655 | for (i = 0; i < ret_val; i++) { |
10656 | int degradation = 0; | |
10657 | if (get_imsm_disk(super, i) == NULL) | |
10658 | degradation++; | |
10659 | if (get_imsm_disk(super, i + 1) == NULL) | |
10660 | degradation++; | |
10661 | if (degradation == 2) | |
10662 | return 0; | |
10663 | } | |
238c0a71 | 10664 | map = get_imsm_map(dev, MAP_1); |
9a717282 AK |
10665 | /* if there is no second map |
10666 | * result can be returned | |
10667 | */ | |
10668 | if (map == NULL) | |
10669 | return ret_val; | |
10670 | /* check degradation in second map | |
10671 | */ | |
10672 | for (i = 0; i < ret_val; i++) { | |
10673 | int degradation = 0; | |
10674 | if (get_imsm_disk(super, i) == NULL) | |
10675 | degradation++; | |
10676 | if (get_imsm_disk(super, i + 1) == NULL) | |
10677 | degradation++; | |
10678 | if (degradation == 2) | |
10679 | return 0; | |
10680 | } | |
10681 | return ret_val; | |
10682 | } | |
10683 | case 5: | |
10684 | return 1; | |
10685 | case 6: | |
10686 | return 2; | |
10687 | default: | |
10688 | return 0; | |
10689 | } | |
10690 | } | |
10691 | ||
d31ad643 PB |
10692 | /******************************************************************************* |
10693 | * Function: validate_container_imsm | |
10694 | * Description: This routine validates container after assemble, | |
10695 | * eg. if devices in container are under the same controller. | |
10696 | * | |
10697 | * Parameters: | |
10698 | * info : linked list with info about devices used in array | |
10699 | * Returns: | |
10700 | * 1 : HBA mismatch | |
10701 | * 0 : Success | |
10702 | ******************************************************************************/ | |
10703 | int validate_container_imsm(struct mdinfo *info) | |
10704 | { | |
420dafcd | 10705 | if (check_no_platform()) |
6b781d33 | 10706 | return 0; |
d31ad643 | 10707 | |
6b781d33 AP |
10708 | struct sys_dev *idev; |
10709 | struct sys_dev *hba = NULL; | |
10710 | struct sys_dev *intel_devices = find_intel_devices(); | |
10711 | char *dev_path = devt_to_devpath(makedev(info->disk.major, | |
7c798f87 | 10712 | info->disk.minor), 1, NULL); |
6b781d33 AP |
10713 | |
10714 | for (idev = intel_devices; idev; idev = idev->next) { | |
10715 | if (dev_path && strstr(dev_path, idev->path)) { | |
10716 | hba = idev; | |
10717 | break; | |
d31ad643 | 10718 | } |
6b781d33 AP |
10719 | } |
10720 | if (dev_path) | |
d31ad643 PB |
10721 | free(dev_path); |
10722 | ||
6b781d33 AP |
10723 | if (!hba) { |
10724 | pr_err("WARNING - Cannot detect HBA for device %s!\n", | |
10725 | devid2kname(makedev(info->disk.major, info->disk.minor))); | |
10726 | return 1; | |
10727 | } | |
10728 | ||
10729 | const struct imsm_orom *orom = get_orom_by_device_id(hba->dev_id); | |
10730 | struct mdinfo *dev; | |
10731 | ||
10732 | for (dev = info->next; dev; dev = dev->next) { | |
7c798f87 MT |
10733 | dev_path = devt_to_devpath(makedev(dev->disk.major, |
10734 | dev->disk.minor), 1, NULL); | |
6b781d33 AP |
10735 | |
10736 | struct sys_dev *hba2 = NULL; | |
10737 | for (idev = intel_devices; idev; idev = idev->next) { | |
10738 | if (dev_path && strstr(dev_path, idev->path)) { | |
10739 | hba2 = idev; | |
10740 | break; | |
d31ad643 PB |
10741 | } |
10742 | } | |
6b781d33 AP |
10743 | if (dev_path) |
10744 | free(dev_path); | |
10745 | ||
10746 | const struct imsm_orom *orom2 = hba2 == NULL ? NULL : | |
10747 | get_orom_by_device_id(hba2->dev_id); | |
10748 | ||
10749 | if (hba2 && hba->type != hba2->type) { | |
10750 | pr_err("WARNING - HBAs of devices do not match %s != %s\n", | |
10751 | get_sys_dev_type(hba->type), get_sys_dev_type(hba2->type)); | |
10752 | return 1; | |
10753 | } | |
10754 | ||
07cb1e57 | 10755 | if (orom != orom2) { |
6b781d33 AP |
10756 | pr_err("WARNING - IMSM container assembled with disks under different HBAs!\n" |
10757 | " This operation is not supported and can lead to data loss.\n"); | |
10758 | return 1; | |
10759 | } | |
10760 | ||
10761 | if (!orom) { | |
10762 | pr_err("WARNING - IMSM container assembled with disks under HBAs without IMSM platform support!\n" | |
10763 | " This operation is not supported and can lead to data loss.\n"); | |
10764 | return 1; | |
10765 | } | |
d31ad643 | 10766 | } |
6b781d33 | 10767 | |
d31ad643 PB |
10768 | return 0; |
10769 | } | |
32141c17 | 10770 | |
6f50473f TM |
10771 | /******************************************************************************* |
10772 | * Function: imsm_record_badblock | |
10773 | * Description: This routine stores new bad block record in BBM log | |
10774 | * | |
10775 | * Parameters: | |
10776 | * a : array containing a bad block | |
10777 | * slot : disk number containing a bad block | |
10778 | * sector : bad block sector | |
10779 | * length : bad block sectors range | |
10780 | * Returns: | |
10781 | * 1 : Success | |
10782 | * 0 : Error | |
10783 | ******************************************************************************/ | |
10784 | static int imsm_record_badblock(struct active_array *a, int slot, | |
10785 | unsigned long long sector, int length) | |
10786 | { | |
10787 | struct intel_super *super = a->container->sb; | |
10788 | int ord; | |
10789 | int ret; | |
10790 | ||
10791 | ord = imsm_disk_slot_to_ord(a, slot); | |
10792 | if (ord < 0) | |
10793 | return 0; | |
10794 | ||
10795 | ret = record_new_badblock(super->bbm_log, ord_to_idx(ord), sector, | |
10796 | length); | |
10797 | if (ret) | |
10798 | super->updates_pending++; | |
10799 | ||
10800 | return ret; | |
10801 | } | |
c07a5a4f TM |
10802 | /******************************************************************************* |
10803 | * Function: imsm_clear_badblock | |
10804 | * Description: This routine clears bad block record from BBM log | |
10805 | * | |
10806 | * Parameters: | |
10807 | * a : array containing a bad block | |
10808 | * slot : disk number containing a bad block | |
10809 | * sector : bad block sector | |
10810 | * length : bad block sectors range | |
10811 | * Returns: | |
10812 | * 1 : Success | |
10813 | * 0 : Error | |
10814 | ******************************************************************************/ | |
10815 | static int imsm_clear_badblock(struct active_array *a, int slot, | |
10816 | unsigned long long sector, int length) | |
10817 | { | |
10818 | struct intel_super *super = a->container->sb; | |
10819 | int ord; | |
10820 | int ret; | |
10821 | ||
10822 | ord = imsm_disk_slot_to_ord(a, slot); | |
10823 | if (ord < 0) | |
10824 | return 0; | |
10825 | ||
10826 | ret = clear_badblock(super->bbm_log, ord_to_idx(ord), sector, length); | |
10827 | if (ret) | |
10828 | super->updates_pending++; | |
10829 | ||
10830 | return ret; | |
10831 | } | |
928f1424 TM |
10832 | /******************************************************************************* |
10833 | * Function: imsm_get_badblocks | |
10834 | * Description: This routine get list of bad blocks for an array | |
10835 | * | |
10836 | * Parameters: | |
10837 | * a : array | |
10838 | * slot : disk number | |
10839 | * Returns: | |
10840 | * bb : structure containing bad blocks | |
10841 | * NULL : error | |
10842 | ******************************************************************************/ | |
10843 | static struct md_bb *imsm_get_badblocks(struct active_array *a, int slot) | |
10844 | { | |
10845 | int inst = a->info.container_member; | |
10846 | struct intel_super *super = a->container->sb; | |
10847 | struct imsm_dev *dev = get_imsm_dev(super, inst); | |
10848 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
10849 | int ord; | |
10850 | ||
10851 | ord = imsm_disk_slot_to_ord(a, slot); | |
10852 | if (ord < 0) | |
10853 | return NULL; | |
10854 | ||
10855 | get_volume_badblocks(super->bbm_log, ord_to_idx(ord), pba_of_lba0(map), | |
44490938 | 10856 | per_dev_array_size(map), &super->bb); |
928f1424 TM |
10857 | |
10858 | return &super->bb; | |
10859 | } | |
27156a57 TM |
10860 | /******************************************************************************* |
10861 | * Function: examine_badblocks_imsm | |
10862 | * Description: Prints list of bad blocks on a disk to the standard output | |
10863 | * | |
10864 | * Parameters: | |
10865 | * st : metadata handler | |
10866 | * fd : open file descriptor for device | |
10867 | * devname : device name | |
10868 | * Returns: | |
10869 | * 0 : Success | |
10870 | * 1 : Error | |
10871 | ******************************************************************************/ | |
10872 | static int examine_badblocks_imsm(struct supertype *st, int fd, char *devname) | |
10873 | { | |
10874 | struct intel_super *super = st->sb; | |
10875 | struct bbm_log *log = super->bbm_log; | |
10876 | struct dl *d = NULL; | |
10877 | int any = 0; | |
10878 | ||
10879 | for (d = super->disks; d ; d = d->next) { | |
10880 | if (strcmp(d->devname, devname) == 0) | |
10881 | break; | |
10882 | } | |
10883 | ||
10884 | if ((d == NULL) || (d->index < 0)) { /* serial mismatch probably */ | |
10885 | pr_err("%s doesn't appear to be part of a raid array\n", | |
10886 | devname); | |
10887 | return 1; | |
10888 | } | |
10889 | ||
10890 | if (log != NULL) { | |
10891 | unsigned int i; | |
10892 | struct bbm_log_entry *entry = &log->marked_block_entries[0]; | |
10893 | ||
10894 | for (i = 0; i < log->entry_count; i++) { | |
10895 | if (entry[i].disk_ordinal == d->index) { | |
10896 | unsigned long long sector = __le48_to_cpu( | |
10897 | &entry[i].defective_block_start); | |
10898 | int cnt = entry[i].marked_count + 1; | |
10899 | ||
10900 | if (!any) { | |
10901 | printf("Bad-blocks on %s:\n", devname); | |
10902 | any = 1; | |
10903 | } | |
10904 | ||
10905 | printf("%20llu for %d sectors\n", sector, cnt); | |
10906 | } | |
10907 | } | |
10908 | } | |
10909 | ||
10910 | if (!any) | |
10911 | printf("No bad-blocks list configured on %s\n", devname); | |
10912 | ||
10913 | return 0; | |
10914 | } | |
687629c2 AK |
10915 | /******************************************************************************* |
10916 | * Function: init_migr_record_imsm | |
10917 | * Description: Function inits imsm migration record | |
10918 | * Parameters: | |
10919 | * super : imsm internal array info | |
10920 | * dev : device under migration | |
10921 | * info : general array info to find the smallest device | |
10922 | * Returns: | |
10923 | * none | |
10924 | ******************************************************************************/ | |
10925 | void init_migr_record_imsm(struct supertype *st, struct imsm_dev *dev, | |
10926 | struct mdinfo *info) | |
10927 | { | |
10928 | struct intel_super *super = st->sb; | |
10929 | struct migr_record *migr_rec = super->migr_rec; | |
10930 | int new_data_disks; | |
10931 | unsigned long long dsize, dev_sectors; | |
10932 | long long unsigned min_dev_sectors = -1LLU; | |
238c0a71 AK |
10933 | struct imsm_map *map_dest = get_imsm_map(dev, MAP_0); |
10934 | struct imsm_map *map_src = get_imsm_map(dev, MAP_1); | |
687629c2 | 10935 | unsigned long long num_migr_units; |
3ef4403c | 10936 | unsigned long long array_blocks; |
2f86fda3 | 10937 | struct dl *dl_disk = NULL; |
687629c2 AK |
10938 | |
10939 | memset(migr_rec, 0, sizeof(struct migr_record)); | |
10940 | migr_rec->family_num = __cpu_to_le32(super->anchor->family_num); | |
10941 | ||
10942 | /* only ascending reshape supported now */ | |
10943 | migr_rec->ascending_migr = __cpu_to_le32(1); | |
10944 | ||
10945 | migr_rec->dest_depth_per_unit = GEN_MIGR_AREA_SIZE / | |
10946 | max(map_dest->blocks_per_strip, map_src->blocks_per_strip); | |
e1742195 AK |
10947 | migr_rec->dest_depth_per_unit *= |
10948 | max(map_dest->blocks_per_strip, map_src->blocks_per_strip); | |
9529d343 | 10949 | new_data_disks = imsm_num_data_members(map_dest); |
687629c2 AK |
10950 | migr_rec->blocks_per_unit = |
10951 | __cpu_to_le32(migr_rec->dest_depth_per_unit * new_data_disks); | |
10952 | migr_rec->dest_depth_per_unit = | |
10953 | __cpu_to_le32(migr_rec->dest_depth_per_unit); | |
3ef4403c | 10954 | array_blocks = info->component_size * new_data_disks; |
687629c2 AK |
10955 | num_migr_units = |
10956 | array_blocks / __le32_to_cpu(migr_rec->blocks_per_unit); | |
10957 | ||
10958 | if (array_blocks % __le32_to_cpu(migr_rec->blocks_per_unit)) | |
10959 | num_migr_units++; | |
9f421827 | 10960 | set_num_migr_units(migr_rec, num_migr_units); |
687629c2 AK |
10961 | |
10962 | migr_rec->post_migr_vol_cap = dev->size_low; | |
10963 | migr_rec->post_migr_vol_cap_hi = dev->size_high; | |
10964 | ||
687629c2 | 10965 | /* Find the smallest dev */ |
2f86fda3 MT |
10966 | for (dl_disk = super->disks; dl_disk ; dl_disk = dl_disk->next) { |
10967 | /* ignore spares in container */ | |
10968 | if (dl_disk->index < 0) | |
687629c2 | 10969 | continue; |
2f86fda3 | 10970 | get_dev_size(dl_disk->fd, NULL, &dsize); |
687629c2 AK |
10971 | dev_sectors = dsize / 512; |
10972 | if (dev_sectors < min_dev_sectors) | |
10973 | min_dev_sectors = dev_sectors; | |
687629c2 | 10974 | } |
9f421827 | 10975 | set_migr_chkp_area_pba(migr_rec, min_dev_sectors - |
687629c2 AK |
10976 | RAID_DISK_RESERVED_BLOCKS_IMSM_HI); |
10977 | ||
10978 | write_imsm_migr_rec(st); | |
10979 | ||
10980 | return; | |
10981 | } | |
10982 | ||
10983 | /******************************************************************************* | |
10984 | * Function: save_backup_imsm | |
10985 | * Description: Function saves critical data stripes to Migration Copy Area | |
10986 | * and updates the current migration unit status. | |
10987 | * Use restore_stripes() to form a destination stripe, | |
10988 | * and to write it to the Copy Area. | |
10989 | * Parameters: | |
10990 | * st : supertype information | |
aea93171 | 10991 | * dev : imsm device that backup is saved for |
687629c2 AK |
10992 | * info : general array info |
10993 | * buf : input buffer | |
687629c2 AK |
10994 | * length : length of data to backup (blocks_per_unit) |
10995 | * Returns: | |
10996 | * 0 : success | |
10997 | *, -1 : fail | |
10998 | ******************************************************************************/ | |
10999 | int save_backup_imsm(struct supertype *st, | |
11000 | struct imsm_dev *dev, | |
11001 | struct mdinfo *info, | |
11002 | void *buf, | |
687629c2 AK |
11003 | int length) |
11004 | { | |
11005 | int rv = -1; | |
11006 | struct intel_super *super = st->sb; | |
687629c2 | 11007 | int i; |
238c0a71 | 11008 | struct imsm_map *map_dest = get_imsm_map(dev, MAP_0); |
687629c2 | 11009 | int new_disks = map_dest->num_members; |
ab724b98 | 11010 | int dest_layout = 0; |
4389ce73 MT |
11011 | int dest_chunk, targets[new_disks]; |
11012 | unsigned long long start, target_offsets[new_disks]; | |
9529d343 | 11013 | int data_disks = imsm_num_data_members(map_dest); |
687629c2 | 11014 | |
2f86fda3 MT |
11015 | for (i = 0; i < new_disks; i++) { |
11016 | struct dl *dl_disk = get_imsm_dl_disk(super, i); | |
4389ce73 MT |
11017 | if (dl_disk && is_fd_valid(dl_disk->fd)) |
11018 | targets[i] = dl_disk->fd; | |
11019 | else | |
11020 | goto abort; | |
2f86fda3 | 11021 | } |
7e45b550 | 11022 | |
d1877f69 | 11023 | start = info->reshape_progress * 512; |
687629c2 | 11024 | for (i = 0; i < new_disks; i++) { |
9f421827 | 11025 | target_offsets[i] = migr_chkp_area_pba(super->migr_rec) * 512; |
d1877f69 AK |
11026 | /* move back copy area adderss, it will be moved forward |
11027 | * in restore_stripes() using start input variable | |
11028 | */ | |
11029 | target_offsets[i] -= start/data_disks; | |
687629c2 AK |
11030 | } |
11031 | ||
68eb8bc6 | 11032 | dest_layout = imsm_level_to_layout(map_dest->raid_level); |
ab724b98 AK |
11033 | dest_chunk = __le16_to_cpu(map_dest->blocks_per_strip) * 512; |
11034 | ||
687629c2 AK |
11035 | if (restore_stripes(targets, /* list of dest devices */ |
11036 | target_offsets, /* migration record offsets */ | |
11037 | new_disks, | |
ab724b98 AK |
11038 | dest_chunk, |
11039 | map_dest->raid_level, | |
11040 | dest_layout, | |
11041 | -1, /* source backup file descriptor */ | |
11042 | 0, /* input buf offset | |
11043 | * always 0 buf is already offseted */ | |
d1877f69 | 11044 | start, |
687629c2 AK |
11045 | length, |
11046 | buf) != 0) { | |
e7b84f9d | 11047 | pr_err("Error restoring stripes\n"); |
687629c2 AK |
11048 | goto abort; |
11049 | } | |
11050 | ||
11051 | rv = 0; | |
11052 | ||
11053 | abort: | |
687629c2 AK |
11054 | return rv; |
11055 | } | |
11056 | ||
11057 | /******************************************************************************* | |
11058 | * Function: save_checkpoint_imsm | |
11059 | * Description: Function called for current unit status update | |
11060 | * in the migration record. It writes it to disk. | |
11061 | * Parameters: | |
11062 | * super : imsm internal array info | |
11063 | * info : general array info | |
11064 | * Returns: | |
11065 | * 0: success | |
11066 | * 1: failure | |
0228d92c AK |
11067 | * 2: failure, means no valid migration record |
11068 | * / no general migration in progress / | |
687629c2 AK |
11069 | ******************************************************************************/ |
11070 | int save_checkpoint_imsm(struct supertype *st, struct mdinfo *info, int state) | |
11071 | { | |
11072 | struct intel_super *super = st->sb; | |
f8b72ef5 AK |
11073 | unsigned long long blocks_per_unit; |
11074 | unsigned long long curr_migr_unit; | |
11075 | ||
2f86fda3 | 11076 | if (load_imsm_migr_rec(super) != 0) { |
7a862a02 | 11077 | dprintf("imsm: ERROR: Cannot read migration record for checkpoint save.\n"); |
2e062e82 AK |
11078 | return 1; |
11079 | } | |
11080 | ||
f8b72ef5 AK |
11081 | blocks_per_unit = __le32_to_cpu(super->migr_rec->blocks_per_unit); |
11082 | if (blocks_per_unit == 0) { | |
0228d92c AK |
11083 | dprintf("imsm: no migration in progress.\n"); |
11084 | return 2; | |
687629c2 | 11085 | } |
f8b72ef5 AK |
11086 | curr_migr_unit = info->reshape_progress / blocks_per_unit; |
11087 | /* check if array is alligned to copy area | |
11088 | * if it is not alligned, add one to current migration unit value | |
11089 | * this can happend on array reshape finish only | |
11090 | */ | |
11091 | if (info->reshape_progress % blocks_per_unit) | |
11092 | curr_migr_unit++; | |
687629c2 | 11093 | |
9f421827 | 11094 | set_current_migr_unit(super->migr_rec, curr_migr_unit); |
687629c2 | 11095 | super->migr_rec->rec_status = __cpu_to_le32(state); |
9f421827 PB |
11096 | set_migr_dest_1st_member_lba(super->migr_rec, |
11097 | super->migr_rec->dest_depth_per_unit * curr_migr_unit); | |
11098 | ||
687629c2 | 11099 | if (write_imsm_migr_rec(st) < 0) { |
7a862a02 | 11100 | dprintf("imsm: Cannot write migration record outside backup area\n"); |
687629c2 AK |
11101 | return 1; |
11102 | } | |
11103 | ||
11104 | return 0; | |
11105 | } | |
11106 | ||
276d77db AK |
11107 | /******************************************************************************* |
11108 | * Function: recover_backup_imsm | |
11109 | * Description: Function recovers critical data from the Migration Copy Area | |
11110 | * while assembling an array. | |
11111 | * Parameters: | |
11112 | * super : imsm internal array info | |
11113 | * info : general array info | |
11114 | * Returns: | |
11115 | * 0 : success (or there is no data to recover) | |
11116 | * 1 : fail | |
11117 | ******************************************************************************/ | |
11118 | int recover_backup_imsm(struct supertype *st, struct mdinfo *info) | |
11119 | { | |
11120 | struct intel_super *super = st->sb; | |
11121 | struct migr_record *migr_rec = super->migr_rec; | |
594dc1b8 | 11122 | struct imsm_map *map_dest; |
276d77db AK |
11123 | struct intel_dev *id = NULL; |
11124 | unsigned long long read_offset; | |
11125 | unsigned long long write_offset; | |
11126 | unsigned unit_len; | |
2f86fda3 | 11127 | int new_disks, err; |
276d77db AK |
11128 | char *buf = NULL; |
11129 | int retval = 1; | |
f36a9ecd | 11130 | unsigned int sector_size = super->sector_size; |
4036e7ee MT |
11131 | unsigned long long curr_migr_unit = current_migr_unit(migr_rec); |
11132 | unsigned long long num_migr_units = get_num_migr_units(migr_rec); | |
90fd7001 | 11133 | char buffer[SYSFS_MAX_BUF_SIZE]; |
6c3560c0 | 11134 | int skipped_disks = 0; |
2f86fda3 | 11135 | struct dl *dl_disk; |
276d77db | 11136 | |
90fd7001 | 11137 | err = sysfs_get_str(info, NULL, "array_state", (char *)buffer, sizeof(buffer)); |
276d77db AK |
11138 | if (err < 1) |
11139 | return 1; | |
11140 | ||
11141 | /* recover data only during assemblation */ | |
11142 | if (strncmp(buffer, "inactive", 8) != 0) | |
11143 | return 0; | |
11144 | /* no data to recover */ | |
11145 | if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL) | |
11146 | return 0; | |
11147 | if (curr_migr_unit >= num_migr_units) | |
11148 | return 1; | |
11149 | ||
11150 | /* find device during reshape */ | |
11151 | for (id = super->devlist; id; id = id->next) | |
11152 | if (is_gen_migration(id->dev)) | |
11153 | break; | |
11154 | if (id == NULL) | |
11155 | return 1; | |
11156 | ||
238c0a71 | 11157 | map_dest = get_imsm_map(id->dev, MAP_0); |
276d77db AK |
11158 | new_disks = map_dest->num_members; |
11159 | ||
9f421827 | 11160 | read_offset = migr_chkp_area_pba(migr_rec) * 512; |
276d77db | 11161 | |
9f421827 | 11162 | write_offset = (migr_dest_1st_member_lba(migr_rec) + |
5551b113 | 11163 | pba_of_lba0(map_dest)) * 512; |
276d77db AK |
11164 | |
11165 | unit_len = __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
f36a9ecd | 11166 | if (posix_memalign((void **)&buf, sector_size, unit_len) != 0) |
276d77db | 11167 | goto abort; |
276d77db | 11168 | |
2f86fda3 MT |
11169 | for (dl_disk = super->disks; dl_disk; dl_disk = dl_disk->next) { |
11170 | if (dl_disk->index < 0) | |
11171 | continue; | |
276d77db | 11172 | |
4389ce73 | 11173 | if (!is_fd_valid(dl_disk->fd)) { |
6c3560c0 AK |
11174 | skipped_disks++; |
11175 | continue; | |
11176 | } | |
2f86fda3 | 11177 | if (lseek64(dl_disk->fd, read_offset, SEEK_SET) < 0) { |
e7b84f9d N |
11178 | pr_err("Cannot seek to block: %s\n", |
11179 | strerror(errno)); | |
137debce AK |
11180 | skipped_disks++; |
11181 | continue; | |
276d77db | 11182 | } |
83b3de77 | 11183 | if (read(dl_disk->fd, buf, unit_len) != (ssize_t)unit_len) { |
e7b84f9d N |
11184 | pr_err("Cannot read copy area block: %s\n", |
11185 | strerror(errno)); | |
137debce AK |
11186 | skipped_disks++; |
11187 | continue; | |
276d77db | 11188 | } |
2f86fda3 | 11189 | if (lseek64(dl_disk->fd, write_offset, SEEK_SET) < 0) { |
e7b84f9d N |
11190 | pr_err("Cannot seek to block: %s\n", |
11191 | strerror(errno)); | |
137debce AK |
11192 | skipped_disks++; |
11193 | continue; | |
276d77db | 11194 | } |
83b3de77 | 11195 | if (write(dl_disk->fd, buf, unit_len) != (ssize_t)unit_len) { |
e7b84f9d N |
11196 | pr_err("Cannot restore block: %s\n", |
11197 | strerror(errno)); | |
137debce AK |
11198 | skipped_disks++; |
11199 | continue; | |
276d77db AK |
11200 | } |
11201 | } | |
11202 | ||
137debce AK |
11203 | if (skipped_disks > imsm_get_allowed_degradation(info->new_level, |
11204 | new_disks, | |
11205 | super, | |
11206 | id->dev)) { | |
7a862a02 | 11207 | pr_err("Cannot restore data from backup. Too many failed disks\n"); |
6c3560c0 AK |
11208 | goto abort; |
11209 | } | |
11210 | ||
befb629b AK |
11211 | if (save_checkpoint_imsm(st, info, UNIT_SRC_NORMAL)) { |
11212 | /* ignore error == 2, this can mean end of reshape here | |
11213 | */ | |
7a862a02 | 11214 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL) during restart\n"); |
befb629b | 11215 | } else |
276d77db | 11216 | retval = 0; |
276d77db AK |
11217 | |
11218 | abort: | |
276d77db AK |
11219 | free(buf); |
11220 | return retval; | |
11221 | } | |
11222 | ||
e21aea08 MT |
11223 | /** |
11224 | * test_and_add_drive_controller_policy_imsm() - add disk controller to policies list. | |
11225 | * @type: Policy type to search on list. | |
11226 | * @pols: List of currently recorded policies. | |
11227 | * @disk_fd: File descriptor of the device to check. | |
11228 | * @hba: The hba disk is attached, could be NULL if verification is disabled. | |
11229 | * @verbose: verbose flag. | |
11230 | * | |
11231 | * IMSM cares about drive physical placement. If @hba is not set, it adds unknown policy. | |
11232 | * If there is no controller policy on pols we are free to add first one. If there is a policy then, | |
11233 | * new must be the same - no controller mixing allowed. | |
11234 | */ | |
11235 | static mdadm_status_t | |
11236 | test_and_add_drive_controller_policy_imsm(const char * const type, dev_policy_t **pols, int disk_fd, | |
11237 | struct sys_dev *hba, const int verbose) | |
11238 | { | |
11239 | const char *controller_policy = get_sys_dev_type(SYS_DEV_UNKNOWN); | |
11240 | struct dev_policy *pol = pol_find(*pols, (char *)type); | |
11241 | char devname[MAX_RAID_SERIAL_LEN]; | |
11242 | ||
11243 | if (hba) | |
11244 | controller_policy = get_sys_dev_type(hba->type); | |
11245 | ||
11246 | if (!pol) { | |
11247 | pol_add(pols, (char *)type, (char *)controller_policy, "imsm"); | |
11248 | return MDADM_STATUS_SUCCESS; | |
2cda7640 | 11249 | } |
e21aea08 MT |
11250 | |
11251 | if (strcmp(pol->value, controller_policy) == 0) | |
11252 | return MDADM_STATUS_SUCCESS; | |
11253 | ||
11254 | fd2devname(disk_fd, devname); | |
11255 | pr_vrb("Intel(R) raid controller \"%s\" found for %s, but \"%s\" was detected earlier\n", | |
11256 | controller_policy, devname, pol->value); | |
11257 | pr_vrb("Disks under different controllers cannot be used, aborting\n"); | |
11258 | ||
11259 | return MDADM_STATUS_ERROR; | |
11260 | } | |
11261 | ||
11262 | struct imsm_drive_policy { | |
11263 | char *type; | |
11264 | mdadm_status_t (*test_and_add_drive_policy)(const char * const type, | |
11265 | struct dev_policy **pols, int disk_fd, | |
11266 | struct sys_dev *hba, const int verbose); | |
11267 | }; | |
11268 | ||
11269 | struct imsm_drive_policy imsm_policies[] = { | |
11270 | {"controller", test_and_add_drive_controller_policy_imsm}, | |
11271 | }; | |
11272 | ||
11273 | mdadm_status_t test_and_add_drive_policies_imsm(struct dev_policy **pols, int disk_fd, | |
11274 | const int verbose) | |
11275 | { | |
11276 | struct imsm_drive_policy *imsm_pol; | |
11277 | struct sys_dev *hba = NULL; | |
11278 | char path[PATH_MAX]; | |
11279 | mdadm_status_t ret; | |
11280 | unsigned int i; | |
11281 | ||
11282 | /* If imsm platform verification is disabled, do not search for hba. */ | |
11283 | if (check_no_platform() != 1) { | |
11284 | if (!diskfd_to_devpath(disk_fd, 1, path)) { | |
11285 | pr_vrb("IMSM: Failed to retrieve device path by file descriptor.\n"); | |
11286 | return MDADM_STATUS_ERROR; | |
11287 | } | |
11288 | ||
11289 | hba = find_disk_attached_hba(disk_fd, path); | |
11290 | if (!hba) { | |
11291 | pr_vrb("IMSM: Failed to find hba for %s\n", path); | |
11292 | return MDADM_STATUS_ERROR; | |
11293 | } | |
11294 | } | |
11295 | ||
11296 | for (i = 0; i < ARRAY_SIZE(imsm_policies); i++) { | |
11297 | imsm_pol = &imsm_policies[i]; | |
11298 | ||
11299 | ret = imsm_pol->test_and_add_drive_policy(imsm_pol->type, pols, disk_fd, hba, | |
11300 | verbose); | |
11301 | if (ret != MDADM_STATUS_SUCCESS) | |
11302 | /* Inherit error code */ | |
11303 | return ret; | |
11304 | } | |
11305 | ||
11306 | return MDADM_STATUS_SUCCESS; | |
2cda7640 ML |
11307 | } |
11308 | ||
f6562011 MT |
11309 | /** |
11310 | * get_spare_criteria_imsm() - set spare criteria. | |
11311 | * @st: supertype. | |
11312 | * @mddev_path: path to md device devnode, it must be container. | |
11313 | * @c: spare_criteria struct to fill, not NULL. | |
11314 | * | |
11315 | * If superblock is not loaded, use mddev_path to load_container. It must be given in this case. | |
11316 | * Filles size and sector size accordingly to superblock. | |
11317 | */ | |
11318 | mdadm_status_t get_spare_criteria_imsm(struct supertype *st, char *mddev_path, | |
11319 | struct spare_criteria *c) | |
11320 | { | |
11321 | mdadm_status_t ret = MDADM_STATUS_ERROR; | |
11322 | bool free_superblock = false; | |
11323 | unsigned long long size = 0; | |
11324 | struct intel_super *super; | |
11325 | struct extent *e; | |
11326 | struct dl *dl; | |
11327 | int i; | |
11328 | ||
11329 | /* If no superblock and no mddev_path, we cannot load superblock. */ | |
11330 | assert(st->sb || mddev_path); | |
11331 | ||
11332 | if (mddev_path) { | |
11333 | int fd = open(mddev_path, O_RDONLY); | |
e21aea08 | 11334 | mdadm_status_t rv; |
f6562011 MT |
11335 | |
11336 | if (!is_fd_valid(fd)) | |
11337 | return MDADM_STATUS_ERROR; | |
11338 | ||
11339 | if (!st->sb) { | |
11340 | if (load_container_imsm(st, fd, st->devnm)) { | |
11341 | close(fd); | |
11342 | return MDADM_STATUS_ERROR; | |
11343 | } | |
11344 | free_superblock = true; | |
11345 | } | |
e21aea08 MT |
11346 | |
11347 | rv = mddev_test_and_add_drive_policies(st, &c->pols, fd, 0); | |
f6562011 | 11348 | close(fd); |
e21aea08 MT |
11349 | |
11350 | if (rv != MDADM_STATUS_SUCCESS) | |
11351 | goto out; | |
f6562011 MT |
11352 | } |
11353 | ||
11354 | super = st->sb; | |
11355 | ||
11356 | /* find first active disk in array */ | |
11357 | dl = super->disks; | |
11358 | while (dl && (is_failed(&dl->disk) || dl->index == -1)) | |
11359 | dl = dl->next; | |
11360 | ||
11361 | if (!dl) | |
11362 | goto out; | |
11363 | ||
11364 | /* find last lba used by subarrays */ | |
11365 | e = get_extents(super, dl, 0); | |
11366 | if (!e) | |
11367 | goto out; | |
11368 | ||
11369 | for (i = 0; e[i].size; i++) | |
11370 | continue; | |
11371 | if (i > 0) | |
11372 | size = e[i - 1].start + e[i - 1].size; | |
11373 | free(e); | |
11374 | ||
11375 | /* add the amount of space needed for metadata */ | |
11376 | size += imsm_min_reserved_sectors(super); | |
11377 | ||
11378 | c->min_size = size * 512; | |
11379 | c->sector_size = super->sector_size; | |
11380 | c->criteria_set = true; | |
11381 | ret = MDADM_STATUS_SUCCESS; | |
11382 | ||
11383 | out: | |
11384 | if (free_superblock) | |
11385 | free_super_imsm(st); | |
11386 | ||
11387 | if (ret != MDADM_STATUS_SUCCESS) | |
11388 | c->criteria_set = false; | |
11389 | ||
11390 | return ret; | |
11391 | } | |
11392 | ||
4dd2df09 | 11393 | static char *imsm_find_array_devnm_by_subdev(int subdev, char *container) |
78b10e66 | 11394 | { |
4dd2df09 | 11395 | static char devnm[32]; |
78b10e66 N |
11396 | char subdev_name[20]; |
11397 | struct mdstat_ent *mdstat; | |
11398 | ||
11399 | sprintf(subdev_name, "%d", subdev); | |
11400 | mdstat = mdstat_by_subdev(subdev_name, container); | |
11401 | if (!mdstat) | |
4dd2df09 | 11402 | return NULL; |
78b10e66 | 11403 | |
4dd2df09 | 11404 | strcpy(devnm, mdstat->devnm); |
78b10e66 | 11405 | free_mdstat(mdstat); |
4dd2df09 | 11406 | return devnm; |
78b10e66 N |
11407 | } |
11408 | ||
11409 | static int imsm_reshape_is_allowed_on_container(struct supertype *st, | |
11410 | struct geo_params *geo, | |
fbf3d202 AK |
11411 | int *old_raid_disks, |
11412 | int direction) | |
78b10e66 | 11413 | { |
694575e7 KW |
11414 | /* currently we only support increasing the number of devices |
11415 | * for a container. This increases the number of device for each | |
11416 | * member array. They must all be RAID0 or RAID5. | |
11417 | */ | |
78b10e66 N |
11418 | int ret_val = 0; |
11419 | struct mdinfo *info, *member; | |
11420 | int devices_that_can_grow = 0; | |
11421 | ||
7a862a02 | 11422 | dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): st->devnm = (%s)\n", st->devnm); |
78b10e66 | 11423 | |
d04f65f4 | 11424 | if (geo->size > 0 || |
78b10e66 N |
11425 | geo->level != UnSet || |
11426 | geo->layout != UnSet || | |
11427 | geo->chunksize != 0 || | |
11428 | geo->raid_disks == UnSet) { | |
7a862a02 | 11429 | dprintf("imsm: Container operation is allowed for raid disks number change only.\n"); |
78b10e66 N |
11430 | return ret_val; |
11431 | } | |
11432 | ||
fbf3d202 | 11433 | if (direction == ROLLBACK_METADATA_CHANGES) { |
7a862a02 | 11434 | dprintf("imsm: Metadata changes rollback is not supported for container operation.\n"); |
fbf3d202 AK |
11435 | return ret_val; |
11436 | } | |
11437 | ||
78b10e66 N |
11438 | info = container_content_imsm(st, NULL); |
11439 | for (member = info; member; member = member->next) { | |
4dd2df09 | 11440 | char *result; |
78b10e66 N |
11441 | |
11442 | dprintf("imsm: checking device_num: %i\n", | |
11443 | member->container_member); | |
11444 | ||
d7d205bd | 11445 | if (geo->raid_disks <= member->array.raid_disks) { |
78b10e66 N |
11446 | /* we work on container for Online Capacity Expansion |
11447 | * only so raid_disks has to grow | |
11448 | */ | |
7a862a02 | 11449 | dprintf("imsm: for container operation raid disks increase is required\n"); |
78b10e66 N |
11450 | break; |
11451 | } | |
11452 | ||
089f9d79 | 11453 | if (info->array.level != 0 && info->array.level != 5) { |
78b10e66 N |
11454 | /* we cannot use this container with other raid level |
11455 | */ | |
7a862a02 | 11456 | dprintf("imsm: for container operation wrong raid level (%i) detected\n", |
78b10e66 N |
11457 | info->array.level); |
11458 | break; | |
11459 | } else { | |
11460 | /* check for platform support | |
11461 | * for this raid level configuration | |
11462 | */ | |
11463 | struct intel_super *super = st->sb; | |
11464 | if (!is_raid_level_supported(super->orom, | |
11465 | member->array.level, | |
11466 | geo->raid_disks)) { | |
7a862a02 | 11467 | dprintf("platform does not support raid%d with %d disk%s\n", |
78b10e66 N |
11468 | info->array.level, |
11469 | geo->raid_disks, | |
11470 | geo->raid_disks > 1 ? "s" : ""); | |
11471 | break; | |
11472 | } | |
2a4a08e7 AK |
11473 | /* check if component size is aligned to chunk size |
11474 | */ | |
11475 | if (info->component_size % | |
11476 | (info->array.chunk_size/512)) { | |
7a862a02 | 11477 | dprintf("Component size is not aligned to chunk size\n"); |
2a4a08e7 AK |
11478 | break; |
11479 | } | |
78b10e66 N |
11480 | } |
11481 | ||
11482 | if (*old_raid_disks && | |
11483 | info->array.raid_disks != *old_raid_disks) | |
11484 | break; | |
11485 | *old_raid_disks = info->array.raid_disks; | |
11486 | ||
11487 | /* All raid5 and raid0 volumes in container | |
11488 | * have to be ready for Online Capacity Expansion | |
11489 | * so they need to be assembled. We have already | |
11490 | * checked that no recovery etc is happening. | |
11491 | */ | |
4dd2df09 N |
11492 | result = imsm_find_array_devnm_by_subdev(member->container_member, |
11493 | st->container_devnm); | |
11494 | if (result == NULL) { | |
78b10e66 N |
11495 | dprintf("imsm: cannot find array\n"); |
11496 | break; | |
11497 | } | |
11498 | devices_that_can_grow++; | |
11499 | } | |
11500 | sysfs_free(info); | |
11501 | if (!member && devices_that_can_grow) | |
11502 | ret_val = 1; | |
11503 | ||
11504 | if (ret_val) | |
1ade5cc1 | 11505 | dprintf("Container operation allowed\n"); |
78b10e66 | 11506 | else |
1ade5cc1 | 11507 | dprintf("Error: %i\n", ret_val); |
78b10e66 N |
11508 | |
11509 | return ret_val; | |
11510 | } | |
11511 | ||
11512 | /* Function: get_spares_for_grow | |
11513 | * Description: Allocates memory and creates list of spare devices | |
1011e834 | 11514 | * avaliable in container. Checks if spare drive size is acceptable. |
78b10e66 N |
11515 | * Parameters: Pointer to the supertype structure |
11516 | * Returns: Pointer to the list of spare devices (mdinfo structure) on success, | |
1011e834 | 11517 | * NULL if fail |
78b10e66 N |
11518 | */ |
11519 | static struct mdinfo *get_spares_for_grow(struct supertype *st) | |
11520 | { | |
ae996e81 MT |
11521 | struct spare_criteria sc = {0}; |
11522 | struct mdinfo *spares; | |
fbfdcb06 | 11523 | |
f6562011 | 11524 | get_spare_criteria_imsm(st, NULL, &sc); |
ae996e81 MT |
11525 | spares = container_choose_spares(st, &sc, NULL, NULL, NULL, 0); |
11526 | ||
11527 | dev_policy_free(sc.pols); | |
11528 | ||
11529 | return spares; | |
78b10e66 N |
11530 | } |
11531 | ||
11532 | /****************************************************************************** | |
11533 | * function: imsm_create_metadata_update_for_reshape | |
11534 | * Function creates update for whole IMSM container. | |
11535 | * | |
11536 | ******************************************************************************/ | |
11537 | static int imsm_create_metadata_update_for_reshape( | |
11538 | struct supertype *st, | |
11539 | struct geo_params *geo, | |
11540 | int old_raid_disks, | |
11541 | struct imsm_update_reshape **updatep) | |
11542 | { | |
11543 | struct intel_super *super = st->sb; | |
11544 | struct imsm_super *mpb = super->anchor; | |
594dc1b8 JS |
11545 | int update_memory_size; |
11546 | struct imsm_update_reshape *u; | |
11547 | struct mdinfo *spares; | |
78b10e66 | 11548 | int i; |
594dc1b8 | 11549 | int delta_disks; |
bbd24d86 | 11550 | struct mdinfo *dev; |
78b10e66 | 11551 | |
1ade5cc1 | 11552 | dprintf("(enter) raid_disks = %i\n", geo->raid_disks); |
78b10e66 N |
11553 | |
11554 | delta_disks = geo->raid_disks - old_raid_disks; | |
11555 | ||
11556 | /* size of all update data without anchor */ | |
11557 | update_memory_size = sizeof(struct imsm_update_reshape); | |
11558 | ||
11559 | /* now add space for spare disks that we need to add. */ | |
11560 | update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1); | |
11561 | ||
503975b9 | 11562 | u = xcalloc(1, update_memory_size); |
78b10e66 N |
11563 | u->type = update_reshape_container_disks; |
11564 | u->old_raid_disks = old_raid_disks; | |
11565 | u->new_raid_disks = geo->raid_disks; | |
11566 | ||
11567 | /* now get spare disks list | |
11568 | */ | |
11569 | spares = get_spares_for_grow(st); | |
11570 | ||
d7be7d87 | 11571 | if (spares == NULL || delta_disks > spares->array.spare_disks) { |
7a862a02 | 11572 | pr_err("imsm: ERROR: Cannot get spare devices for %s.\n", geo->dev_name); |
e4c72d1d | 11573 | i = -1; |
78b10e66 N |
11574 | goto abort; |
11575 | } | |
11576 | ||
11577 | /* we have got spares | |
11578 | * update disk list in imsm_disk list table in anchor | |
11579 | */ | |
11580 | dprintf("imsm: %i spares are available.\n\n", | |
11581 | spares->array.spare_disks); | |
11582 | ||
bbd24d86 | 11583 | dev = spares->devs; |
78b10e66 | 11584 | for (i = 0; i < delta_disks; i++) { |
78b10e66 N |
11585 | struct dl *dl; |
11586 | ||
bbd24d86 AK |
11587 | if (dev == NULL) |
11588 | break; | |
78b10e66 N |
11589 | u->new_disks[i] = makedev(dev->disk.major, |
11590 | dev->disk.minor); | |
11591 | dl = get_disk_super(super, dev->disk.major, dev->disk.minor); | |
ee4beede AK |
11592 | dl->index = mpb->num_disks; |
11593 | mpb->num_disks++; | |
bbd24d86 | 11594 | dev = dev->next; |
78b10e66 | 11595 | } |
78b10e66 N |
11596 | |
11597 | abort: | |
11598 | /* free spares | |
11599 | */ | |
11600 | sysfs_free(spares); | |
11601 | ||
d677e0b8 | 11602 | dprintf("imsm: reshape update preparation :"); |
78b10e66 | 11603 | if (i == delta_disks) { |
1ade5cc1 | 11604 | dprintf_cont(" OK\n"); |
78b10e66 N |
11605 | *updatep = u; |
11606 | return update_memory_size; | |
11607 | } | |
11608 | free(u); | |
1ade5cc1 | 11609 | dprintf_cont(" Error\n"); |
78b10e66 N |
11610 | |
11611 | return 0; | |
11612 | } | |
11613 | ||
f3871fdc AK |
11614 | /****************************************************************************** |
11615 | * function: imsm_create_metadata_update_for_size_change() | |
11616 | * Creates update for IMSM array for array size change. | |
11617 | * | |
11618 | ******************************************************************************/ | |
11619 | static int imsm_create_metadata_update_for_size_change( | |
11620 | struct supertype *st, | |
11621 | struct geo_params *geo, | |
11622 | struct imsm_update_size_change **updatep) | |
11623 | { | |
11624 | struct intel_super *super = st->sb; | |
594dc1b8 JS |
11625 | int update_memory_size; |
11626 | struct imsm_update_size_change *u; | |
f3871fdc | 11627 | |
1ade5cc1 | 11628 | dprintf("(enter) New size = %llu\n", geo->size); |
f3871fdc AK |
11629 | |
11630 | /* size of all update data without anchor */ | |
11631 | update_memory_size = sizeof(struct imsm_update_size_change); | |
11632 | ||
503975b9 | 11633 | u = xcalloc(1, update_memory_size); |
f3871fdc AK |
11634 | u->type = update_size_change; |
11635 | u->subdev = super->current_vol; | |
11636 | u->new_size = geo->size; | |
11637 | ||
11638 | dprintf("imsm: reshape update preparation : OK\n"); | |
11639 | *updatep = u; | |
11640 | ||
11641 | return update_memory_size; | |
11642 | } | |
11643 | ||
48c5303a PC |
11644 | /****************************************************************************** |
11645 | * function: imsm_create_metadata_update_for_migration() | |
11646 | * Creates update for IMSM array. | |
11647 | * | |
11648 | ******************************************************************************/ | |
11649 | static int imsm_create_metadata_update_for_migration( | |
11650 | struct supertype *st, | |
11651 | struct geo_params *geo, | |
11652 | struct imsm_update_reshape_migration **updatep) | |
11653 | { | |
11654 | struct intel_super *super = st->sb; | |
594dc1b8 | 11655 | int update_memory_size; |
756a15f3 | 11656 | int current_chunk_size; |
594dc1b8 | 11657 | struct imsm_update_reshape_migration *u; |
756a15f3 MG |
11658 | struct imsm_dev *dev = get_imsm_dev(super, super->current_vol); |
11659 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
48c5303a PC |
11660 | int previous_level = -1; |
11661 | ||
1ade5cc1 | 11662 | dprintf("(enter) New Level = %i\n", geo->level); |
48c5303a PC |
11663 | |
11664 | /* size of all update data without anchor */ | |
11665 | update_memory_size = sizeof(struct imsm_update_reshape_migration); | |
11666 | ||
503975b9 | 11667 | u = xcalloc(1, update_memory_size); |
48c5303a PC |
11668 | u->type = update_reshape_migration; |
11669 | u->subdev = super->current_vol; | |
11670 | u->new_level = geo->level; | |
11671 | u->new_layout = geo->layout; | |
11672 | u->new_raid_disks = u->old_raid_disks = geo->raid_disks; | |
11673 | u->new_disks[0] = -1; | |
4bba0439 | 11674 | u->new_chunksize = -1; |
48c5303a | 11675 | |
756a15f3 | 11676 | current_chunk_size = __le16_to_cpu(map->blocks_per_strip) / 2; |
48c5303a | 11677 | |
756a15f3 MG |
11678 | if (geo->chunksize != current_chunk_size) { |
11679 | u->new_chunksize = geo->chunksize / 1024; | |
11680 | dprintf("imsm: chunk size change from %i to %i\n", | |
11681 | current_chunk_size, u->new_chunksize); | |
48c5303a | 11682 | } |
756a15f3 MG |
11683 | previous_level = map->raid_level; |
11684 | ||
089f9d79 | 11685 | if (geo->level == 5 && previous_level == 0) { |
48c5303a PC |
11686 | struct mdinfo *spares = NULL; |
11687 | ||
11688 | u->new_raid_disks++; | |
11689 | spares = get_spares_for_grow(st); | |
089f9d79 | 11690 | if (spares == NULL || spares->array.spare_disks < 1) { |
48c5303a PC |
11691 | free(u); |
11692 | sysfs_free(spares); | |
11693 | update_memory_size = 0; | |
565cc99e | 11694 | pr_err("cannot get spare device for requested migration\n"); |
48c5303a PC |
11695 | return 0; |
11696 | } | |
11697 | sysfs_free(spares); | |
11698 | } | |
11699 | dprintf("imsm: reshape update preparation : OK\n"); | |
11700 | *updatep = u; | |
11701 | ||
11702 | return update_memory_size; | |
11703 | } | |
11704 | ||
8dd70bce AK |
11705 | static void imsm_update_metadata_locally(struct supertype *st, |
11706 | void *buf, int len) | |
11707 | { | |
11708 | struct metadata_update mu; | |
11709 | ||
11710 | mu.buf = buf; | |
11711 | mu.len = len; | |
11712 | mu.space = NULL; | |
11713 | mu.space_list = NULL; | |
11714 | mu.next = NULL; | |
5fe6f031 N |
11715 | if (imsm_prepare_update(st, &mu)) |
11716 | imsm_process_update(st, &mu); | |
8dd70bce AK |
11717 | |
11718 | while (mu.space_list) { | |
11719 | void **space = mu.space_list; | |
11720 | mu.space_list = *space; | |
11721 | free(space); | |
11722 | } | |
11723 | } | |
78b10e66 | 11724 | |
cbaa7904 MT |
11725 | /** |
11726 | * imsm_analyze_expand() - check expand properties and calculate new size. | |
11727 | * @st: imsm supertype. | |
11728 | * @geo: new geometry params. | |
11729 | * @array: array info. | |
11730 | * @direction: reshape direction. | |
11731 | * | |
11732 | * Obtain free space after the &array and verify if expand to requested size is | |
11733 | * possible. If geo->size is set to %MAX_SIZE, assume that max free size is | |
11734 | * requested. | |
11735 | * | |
11736 | * Return: | |
11737 | * On success %IMSM_STATUS_OK is returned, geo->size and geo->raid_disks are | |
11738 | * updated. | |
11739 | * On error, %IMSM_STATUS_ERROR is returned. | |
11740 | */ | |
11741 | static imsm_status_t imsm_analyze_expand(struct supertype *st, | |
11742 | struct geo_params *geo, | |
11743 | struct mdinfo *array, | |
11744 | int direction) | |
11745 | { | |
11746 | struct intel_super *super = st->sb; | |
11747 | struct imsm_dev *dev = get_imsm_dev(super, super->current_vol); | |
11748 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
11749 | int data_disks = imsm_num_data_members(map); | |
11750 | ||
11751 | unsigned long long current_size; | |
11752 | unsigned long long free_size; | |
11753 | unsigned long long new_size; | |
11754 | unsigned long long max_size; | |
11755 | ||
11756 | const int chunk_kib = geo->chunksize / 1024; | |
11757 | imsm_status_t rv; | |
11758 | ||
11759 | if (direction == ROLLBACK_METADATA_CHANGES) { | |
11760 | /** | |
11761 | * Accept size for rollback only. | |
11762 | */ | |
11763 | new_size = geo->size * 2; | |
11764 | goto success; | |
11765 | } | |
11766 | ||
cbaa7904 MT |
11767 | if (data_disks == 0) { |
11768 | pr_err("imsm: Cannot retrieve data disks.\n"); | |
11769 | return IMSM_STATUS_ERROR; | |
11770 | } | |
11771 | current_size = array->custom_array_size / data_disks; | |
11772 | ||
aa19fdd4 | 11773 | rv = imsm_get_free_size(super, dev->vol.map->num_members, 0, chunk_kib, &free_size, true); |
cbaa7904 MT |
11774 | if (rv != IMSM_STATUS_OK) { |
11775 | pr_err("imsm: Cannot find free space for expand.\n"); | |
11776 | return IMSM_STATUS_ERROR; | |
11777 | } | |
11778 | max_size = round_member_size_to_mb(free_size + current_size); | |
11779 | ||
11780 | if (geo->size == MAX_SIZE) | |
11781 | new_size = max_size; | |
11782 | else | |
11783 | new_size = round_member_size_to_mb(geo->size * 2); | |
11784 | ||
11785 | if (new_size == 0) { | |
11786 | pr_err("imsm: Rounded requested size is 0.\n"); | |
11787 | return IMSM_STATUS_ERROR; | |
11788 | } | |
11789 | ||
11790 | if (new_size > max_size) { | |
11791 | pr_err("imsm: Rounded requested size (%llu) is larger than free space available (%llu).\n", | |
11792 | new_size, max_size); | |
11793 | return IMSM_STATUS_ERROR; | |
11794 | } | |
11795 | ||
11796 | if (new_size == current_size) { | |
11797 | pr_err("imsm: Rounded requested size (%llu) is same as current size (%llu).\n", | |
11798 | new_size, current_size); | |
11799 | return IMSM_STATUS_ERROR; | |
11800 | } | |
11801 | ||
11802 | if (new_size < current_size) { | |
11803 | pr_err("imsm: Size reduction is not supported, rounded requested size (%llu) is smaller than current (%llu).\n", | |
11804 | new_size, current_size); | |
11805 | return IMSM_STATUS_ERROR; | |
11806 | } | |
11807 | ||
11808 | success: | |
11809 | dprintf("imsm: New size per member is %llu.\n", new_size); | |
11810 | geo->size = data_disks * new_size; | |
11811 | geo->raid_disks = dev->vol.map->num_members; | |
11812 | return IMSM_STATUS_OK; | |
11813 | } | |
11814 | ||
471bceb6 | 11815 | /*************************************************************************** |
694575e7 | 11816 | * Function: imsm_analyze_change |
471bceb6 | 11817 | * Description: Function analyze change for single volume |
1011e834 | 11818 | * and validate if transition is supported |
fbf3d202 AK |
11819 | * Parameters: Geometry parameters, supertype structure, |
11820 | * metadata change direction (apply/rollback) | |
694575e7 | 11821 | * Returns: Operation type code on success, -1 if fail |
471bceb6 KW |
11822 | ****************************************************************************/ |
11823 | enum imsm_reshape_type imsm_analyze_change(struct supertype *st, | |
fbf3d202 AK |
11824 | struct geo_params *geo, |
11825 | int direction) | |
694575e7 | 11826 | { |
471bceb6 KW |
11827 | struct mdinfo info; |
11828 | int change = -1; | |
11829 | int check_devs = 0; | |
c21e737b | 11830 | int chunk; |
67a2db32 AK |
11831 | /* number of added/removed disks in operation result */ |
11832 | int devNumChange = 0; | |
11833 | /* imsm compatible layout value for array geometry verification */ | |
11834 | int imsm_layout = -1; | |
6d4d9ab2 | 11835 | imsm_status_t rv; |
471bceb6 KW |
11836 | |
11837 | getinfo_super_imsm_volume(st, &info, NULL); | |
089f9d79 JS |
11838 | if (geo->level != info.array.level && geo->level >= 0 && |
11839 | geo->level != UnSet) { | |
471bceb6 KW |
11840 | switch (info.array.level) { |
11841 | case 0: | |
11842 | if (geo->level == 5) { | |
b5347799 | 11843 | change = CH_MIGRATION; |
e13ce846 | 11844 | if (geo->layout != ALGORITHM_LEFT_ASYMMETRIC) { |
7a862a02 | 11845 | pr_err("Error. Requested Layout not supported (left-asymmetric layout is supported only)!\n"); |
e13ce846 AK |
11846 | change = -1; |
11847 | goto analyse_change_exit; | |
11848 | } | |
67a2db32 | 11849 | imsm_layout = geo->layout; |
471bceb6 | 11850 | check_devs = 1; |
e91a3bad LM |
11851 | devNumChange = 1; /* parity disk added */ |
11852 | } else if (geo->level == 10) { | |
471bceb6 KW |
11853 | change = CH_TAKEOVER; |
11854 | check_devs = 1; | |
e91a3bad | 11855 | devNumChange = 2; /* two mirrors added */ |
67a2db32 | 11856 | imsm_layout = 0x102; /* imsm supported layout */ |
471bceb6 | 11857 | } |
dfe77a9e KW |
11858 | break; |
11859 | case 1: | |
471bceb6 KW |
11860 | case 10: |
11861 | if (geo->level == 0) { | |
11862 | change = CH_TAKEOVER; | |
11863 | check_devs = 1; | |
e91a3bad | 11864 | devNumChange = -(geo->raid_disks/2); |
67a2db32 | 11865 | imsm_layout = 0; /* imsm raid0 layout */ |
471bceb6 KW |
11866 | } |
11867 | break; | |
11868 | } | |
11869 | if (change == -1) { | |
7a862a02 | 11870 | pr_err("Error. Level Migration from %d to %d not supported!\n", |
e7b84f9d | 11871 | info.array.level, geo->level); |
471bceb6 KW |
11872 | goto analyse_change_exit; |
11873 | } | |
11874 | } else | |
11875 | geo->level = info.array.level; | |
11876 | ||
089f9d79 JS |
11877 | if (geo->layout != info.array.layout && |
11878 | (geo->layout != UnSet && geo->layout != -1)) { | |
b5347799 | 11879 | change = CH_MIGRATION; |
089f9d79 JS |
11880 | if (info.array.layout == 0 && info.array.level == 5 && |
11881 | geo->layout == 5) { | |
471bceb6 | 11882 | /* reshape 5 -> 4 */ |
089f9d79 JS |
11883 | } else if (info.array.layout == 5 && info.array.level == 5 && |
11884 | geo->layout == 0) { | |
471bceb6 KW |
11885 | /* reshape 4 -> 5 */ |
11886 | geo->layout = 0; | |
11887 | geo->level = 5; | |
11888 | } else { | |
7a862a02 | 11889 | pr_err("Error. Layout Migration from %d to %d not supported!\n", |
e7b84f9d | 11890 | info.array.layout, geo->layout); |
471bceb6 KW |
11891 | change = -1; |
11892 | goto analyse_change_exit; | |
11893 | } | |
67a2db32 | 11894 | } else { |
471bceb6 | 11895 | geo->layout = info.array.layout; |
67a2db32 AK |
11896 | if (imsm_layout == -1) |
11897 | imsm_layout = info.array.layout; | |
11898 | } | |
471bceb6 | 11899 | |
089f9d79 JS |
11900 | if (geo->chunksize > 0 && geo->chunksize != UnSet && |
11901 | geo->chunksize != info.array.chunk_size) { | |
2d2b0eb7 MD |
11902 | if (info.array.level == 10) { |
11903 | pr_err("Error. Chunk size change for RAID 10 is not supported.\n"); | |
11904 | change = -1; | |
11905 | goto analyse_change_exit; | |
1e9b2c3f PB |
11906 | } else if (info.component_size % (geo->chunksize/512)) { |
11907 | pr_err("New chunk size (%dK) does not evenly divide device size (%lluk). Aborting...\n", | |
11908 | geo->chunksize/1024, info.component_size/2); | |
11909 | change = -1; | |
11910 | goto analyse_change_exit; | |
2d2b0eb7 | 11911 | } |
b5347799 | 11912 | change = CH_MIGRATION; |
2d2b0eb7 | 11913 | } else { |
471bceb6 | 11914 | geo->chunksize = info.array.chunk_size; |
2d2b0eb7 | 11915 | } |
471bceb6 | 11916 | |
cbaa7904 | 11917 | if (geo->size > 0) { |
7abc9871 | 11918 | if (change != -1) { |
7a862a02 | 11919 | pr_err("Error. Size change should be the only one at a time.\n"); |
7abc9871 AK |
11920 | change = -1; |
11921 | goto analyse_change_exit; | |
11922 | } | |
6d4d9ab2 | 11923 | |
cbaa7904 | 11924 | rv = imsm_analyze_expand(st, geo, &info, direction); |
6d4d9ab2 | 11925 | if (rv != IMSM_STATUS_OK) |
cbaa7904 | 11926 | goto analyse_change_exit; |
7abc9871 AK |
11927 | change = CH_ARRAY_SIZE; |
11928 | } | |
cbaa7904 MT |
11929 | |
11930 | chunk = geo->chunksize / 1024; | |
471bceb6 KW |
11931 | if (!validate_geometry_imsm(st, |
11932 | geo->level, | |
67a2db32 | 11933 | imsm_layout, |
e91a3bad | 11934 | geo->raid_disks + devNumChange, |
c21e737b | 11935 | &chunk, |
af4348dd | 11936 | geo->size, INVALID_SECTORS, |
5308f117 | 11937 | 0, 0, info.consistency_policy, 1)) |
471bceb6 KW |
11938 | change = -1; |
11939 | ||
11940 | if (check_devs) { | |
11941 | struct intel_super *super = st->sb; | |
11942 | struct imsm_super *mpb = super->anchor; | |
11943 | ||
11944 | if (mpb->num_raid_devs > 1) { | |
f1cc8ab9 LF |
11945 | pr_err("Error. Cannot perform operation on %s- for this operation " |
11946 | "it MUST be single array in container\n", geo->dev_name); | |
471bceb6 KW |
11947 | change = -1; |
11948 | } | |
11949 | } | |
11950 | ||
11951 | analyse_change_exit: | |
089f9d79 JS |
11952 | if (direction == ROLLBACK_METADATA_CHANGES && |
11953 | (change == CH_MIGRATION || change == CH_TAKEOVER)) { | |
7a862a02 | 11954 | dprintf("imsm: Metadata changes rollback is not supported for migration and takeover operations.\n"); |
fbf3d202 AK |
11955 | change = -1; |
11956 | } | |
471bceb6 | 11957 | return change; |
694575e7 KW |
11958 | } |
11959 | ||
bb025c2f KW |
11960 | int imsm_takeover(struct supertype *st, struct geo_params *geo) |
11961 | { | |
11962 | struct intel_super *super = st->sb; | |
11963 | struct imsm_update_takeover *u; | |
11964 | ||
503975b9 | 11965 | u = xmalloc(sizeof(struct imsm_update_takeover)); |
bb025c2f KW |
11966 | |
11967 | u->type = update_takeover; | |
11968 | u->subarray = super->current_vol; | |
11969 | ||
11970 | /* 10->0 transition */ | |
11971 | if (geo->level == 0) | |
11972 | u->direction = R10_TO_R0; | |
11973 | ||
0529c688 KW |
11974 | /* 0->10 transition */ |
11975 | if (geo->level == 10) | |
11976 | u->direction = R0_TO_R10; | |
11977 | ||
bb025c2f KW |
11978 | /* update metadata locally */ |
11979 | imsm_update_metadata_locally(st, u, | |
11980 | sizeof(struct imsm_update_takeover)); | |
11981 | /* and possibly remotely */ | |
11982 | if (st->update_tail) | |
11983 | append_metadata_update(st, u, | |
11984 | sizeof(struct imsm_update_takeover)); | |
11985 | else | |
11986 | free(u); | |
11987 | ||
11988 | return 0; | |
11989 | } | |
11990 | ||
895ffd99 MT |
11991 | /* Flush size update if size calculated by num_data_stripes is higher than |
11992 | * imsm_dev_size to eliminate differences during reshape. | |
11993 | * Mdmon will recalculate them correctly. | |
11994 | * If subarray index is not set then check whole container. | |
11995 | * Returns: | |
11996 | * 0 - no error occurred | |
11997 | * 1 - error detected | |
11998 | */ | |
11999 | static int imsm_fix_size_mismatch(struct supertype *st, int subarray_index) | |
12000 | { | |
12001 | struct intel_super *super = st->sb; | |
12002 | int tmp = super->current_vol; | |
12003 | int ret_val = 1; | |
12004 | int i; | |
12005 | ||
12006 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
12007 | if (subarray_index >= 0 && i != subarray_index) | |
12008 | continue; | |
12009 | super->current_vol = i; | |
12010 | struct imsm_dev *dev = get_imsm_dev(super, super->current_vol); | |
12011 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
12012 | unsigned int disc_count = imsm_num_data_members(map); | |
12013 | struct geo_params geo; | |
12014 | struct imsm_update_size_change *update; | |
12015 | unsigned long long calc_size = per_dev_array_size(map) * disc_count; | |
12016 | unsigned long long d_size = imsm_dev_size(dev); | |
12017 | int u_size; | |
12018 | ||
42e02e61 | 12019 | if (calc_size == d_size) |
895ffd99 MT |
12020 | continue; |
12021 | ||
ff904202 MT |
12022 | /* There is a difference, confirm that imsm_dev_size is |
12023 | * smaller and push update. | |
895ffd99 | 12024 | */ |
ff904202 MT |
12025 | if (d_size > calc_size) { |
12026 | pr_err("imsm: dev size of subarray %d is incorrect\n", | |
12027 | i); | |
895ffd99 MT |
12028 | goto exit; |
12029 | } | |
12030 | memset(&geo, 0, sizeof(struct geo_params)); | |
12031 | geo.size = d_size; | |
12032 | u_size = imsm_create_metadata_update_for_size_change(st, &geo, | |
12033 | &update); | |
895ffd99 MT |
12034 | imsm_update_metadata_locally(st, update, u_size); |
12035 | if (st->update_tail) { | |
12036 | append_metadata_update(st, update, u_size); | |
12037 | flush_metadata_updates(st); | |
12038 | st->update_tail = &st->updates; | |
12039 | } else { | |
12040 | imsm_sync_metadata(st); | |
5ce5a15f | 12041 | free(update); |
895ffd99 MT |
12042 | } |
12043 | } | |
12044 | ret_val = 0; | |
12045 | exit: | |
12046 | super->current_vol = tmp; | |
12047 | return ret_val; | |
12048 | } | |
12049 | ||
d04f65f4 N |
12050 | static int imsm_reshape_super(struct supertype *st, unsigned long long size, |
12051 | int level, | |
78b10e66 | 12052 | int layout, int chunksize, int raid_disks, |
41784c88 | 12053 | int delta_disks, char *backup, char *dev, |
016e00f5 | 12054 | int direction, int verbose) |
78b10e66 | 12055 | { |
78b10e66 N |
12056 | int ret_val = 1; |
12057 | struct geo_params geo; | |
12058 | ||
1ade5cc1 | 12059 | dprintf("(enter)\n"); |
78b10e66 | 12060 | |
71204a50 | 12061 | memset(&geo, 0, sizeof(struct geo_params)); |
78b10e66 N |
12062 | |
12063 | geo.dev_name = dev; | |
4dd2df09 | 12064 | strcpy(geo.devnm, st->devnm); |
78b10e66 N |
12065 | geo.size = size; |
12066 | geo.level = level; | |
12067 | geo.layout = layout; | |
12068 | geo.chunksize = chunksize; | |
12069 | geo.raid_disks = raid_disks; | |
41784c88 AK |
12070 | if (delta_disks != UnSet) |
12071 | geo.raid_disks += delta_disks; | |
78b10e66 | 12072 | |
1ade5cc1 N |
12073 | dprintf("for level : %i\n", geo.level); |
12074 | dprintf("for raid_disks : %i\n", geo.raid_disks); | |
78b10e66 | 12075 | |
4dd2df09 | 12076 | if (strcmp(st->container_devnm, st->devnm) == 0) { |
694575e7 KW |
12077 | /* On container level we can only increase number of devices. */ |
12078 | dprintf("imsm: info: Container operation\n"); | |
78b10e66 | 12079 | int old_raid_disks = 0; |
6dc0be30 | 12080 | |
78b10e66 | 12081 | if (imsm_reshape_is_allowed_on_container( |
fbf3d202 | 12082 | st, &geo, &old_raid_disks, direction)) { |
78b10e66 N |
12083 | struct imsm_update_reshape *u = NULL; |
12084 | int len; | |
12085 | ||
895ffd99 MT |
12086 | if (imsm_fix_size_mismatch(st, -1)) { |
12087 | dprintf("imsm: Cannot fix size mismatch\n"); | |
12088 | goto exit_imsm_reshape_super; | |
12089 | } | |
12090 | ||
78b10e66 N |
12091 | len = imsm_create_metadata_update_for_reshape( |
12092 | st, &geo, old_raid_disks, &u); | |
12093 | ||
ed08d51c AK |
12094 | if (len <= 0) { |
12095 | dprintf("imsm: Cannot prepare update\n"); | |
12096 | goto exit_imsm_reshape_super; | |
12097 | } | |
12098 | ||
8dd70bce AK |
12099 | ret_val = 0; |
12100 | /* update metadata locally */ | |
12101 | imsm_update_metadata_locally(st, u, len); | |
12102 | /* and possibly remotely */ | |
12103 | if (st->update_tail) | |
12104 | append_metadata_update(st, u, len); | |
12105 | else | |
ed08d51c | 12106 | free(u); |
8dd70bce | 12107 | |
694575e7 | 12108 | } else { |
7a862a02 | 12109 | pr_err("(imsm) Operation is not allowed on this container\n"); |
694575e7 KW |
12110 | } |
12111 | } else { | |
12112 | /* On volume level we support following operations | |
471bceb6 KW |
12113 | * - takeover: raid10 -> raid0; raid0 -> raid10 |
12114 | * - chunk size migration | |
12115 | * - migration: raid5 -> raid0; raid0 -> raid5 | |
12116 | */ | |
12117 | struct intel_super *super = st->sb; | |
12118 | struct intel_dev *dev = super->devlist; | |
4dd2df09 | 12119 | int change; |
694575e7 | 12120 | dprintf("imsm: info: Volume operation\n"); |
471bceb6 KW |
12121 | /* find requested device */ |
12122 | while (dev) { | |
1011e834 | 12123 | char *devnm = |
4dd2df09 N |
12124 | imsm_find_array_devnm_by_subdev( |
12125 | dev->index, st->container_devnm); | |
12126 | if (devnm && strcmp(devnm, geo.devnm) == 0) | |
471bceb6 KW |
12127 | break; |
12128 | dev = dev->next; | |
12129 | } | |
12130 | if (dev == NULL) { | |
4dd2df09 N |
12131 | pr_err("Cannot find %s (%s) subarray\n", |
12132 | geo.dev_name, geo.devnm); | |
471bceb6 KW |
12133 | goto exit_imsm_reshape_super; |
12134 | } | |
12135 | super->current_vol = dev->index; | |
fbf3d202 | 12136 | change = imsm_analyze_change(st, &geo, direction); |
694575e7 | 12137 | switch (change) { |
471bceb6 | 12138 | case CH_TAKEOVER: |
bb025c2f | 12139 | ret_val = imsm_takeover(st, &geo); |
694575e7 | 12140 | break; |
48c5303a PC |
12141 | case CH_MIGRATION: { |
12142 | struct imsm_update_reshape_migration *u = NULL; | |
12143 | int len = | |
12144 | imsm_create_metadata_update_for_migration( | |
12145 | st, &geo, &u); | |
12146 | if (len < 1) { | |
7a862a02 | 12147 | dprintf("imsm: Cannot prepare update\n"); |
48c5303a PC |
12148 | break; |
12149 | } | |
471bceb6 | 12150 | ret_val = 0; |
48c5303a PC |
12151 | /* update metadata locally */ |
12152 | imsm_update_metadata_locally(st, u, len); | |
12153 | /* and possibly remotely */ | |
12154 | if (st->update_tail) | |
12155 | append_metadata_update(st, u, len); | |
12156 | else | |
12157 | free(u); | |
12158 | } | |
12159 | break; | |
7abc9871 | 12160 | case CH_ARRAY_SIZE: { |
f3871fdc AK |
12161 | struct imsm_update_size_change *u = NULL; |
12162 | int len = | |
12163 | imsm_create_metadata_update_for_size_change( | |
12164 | st, &geo, &u); | |
12165 | if (len < 1) { | |
7a862a02 | 12166 | dprintf("imsm: Cannot prepare update\n"); |
f3871fdc AK |
12167 | break; |
12168 | } | |
12169 | ret_val = 0; | |
12170 | /* update metadata locally */ | |
12171 | imsm_update_metadata_locally(st, u, len); | |
12172 | /* and possibly remotely */ | |
12173 | if (st->update_tail) | |
12174 | append_metadata_update(st, u, len); | |
12175 | else | |
12176 | free(u); | |
7abc9871 AK |
12177 | } |
12178 | break; | |
471bceb6 KW |
12179 | default: |
12180 | ret_val = 1; | |
694575e7 | 12181 | } |
694575e7 | 12182 | } |
78b10e66 | 12183 | |
ed08d51c | 12184 | exit_imsm_reshape_super: |
78b10e66 N |
12185 | dprintf("imsm: reshape_super Exit code = %i\n", ret_val); |
12186 | return ret_val; | |
12187 | } | |
2cda7640 | 12188 | |
0febb20c AO |
12189 | #define COMPLETED_OK 0 |
12190 | #define COMPLETED_NONE 1 | |
12191 | #define COMPLETED_DELAYED 2 | |
12192 | ||
12193 | static int read_completed(int fd, unsigned long long *val) | |
12194 | { | |
12195 | int ret; | |
90fd7001 | 12196 | char buf[SYSFS_MAX_BUF_SIZE]; |
0febb20c | 12197 | |
90fd7001 | 12198 | ret = sysfs_fd_get_str(fd, buf, sizeof(buf)); |
0febb20c AO |
12199 | if (ret < 0) |
12200 | return ret; | |
12201 | ||
12202 | ret = COMPLETED_OK; | |
b823c8f9 | 12203 | if (str_is_none(buf) == true) { |
0febb20c AO |
12204 | ret = COMPLETED_NONE; |
12205 | } else if (strncmp(buf, "delayed", 7) == 0) { | |
12206 | ret = COMPLETED_DELAYED; | |
12207 | } else { | |
12208 | char *ep; | |
12209 | *val = strtoull(buf, &ep, 0); | |
12210 | if (ep == buf || (*ep != 0 && *ep != '\n' && *ep != ' ')) | |
12211 | ret = -1; | |
12212 | } | |
12213 | return ret; | |
12214 | } | |
12215 | ||
eee67a47 AK |
12216 | /******************************************************************************* |
12217 | * Function: wait_for_reshape_imsm | |
12218 | * Description: Function writes new sync_max value and waits until | |
12219 | * reshape process reach new position | |
12220 | * Parameters: | |
12221 | * sra : general array info | |
eee67a47 AK |
12222 | * ndata : number of disks in new array's layout |
12223 | * Returns: | |
12224 | * 0 : success, | |
12225 | * 1 : there is no reshape in progress, | |
12226 | * -1 : fail | |
12227 | ******************************************************************************/ | |
ae9f01f8 | 12228 | int wait_for_reshape_imsm(struct mdinfo *sra, int ndata) |
eee67a47 | 12229 | { |
85ca499c | 12230 | int fd = sysfs_get_fd(sra, NULL, "sync_completed"); |
df2647fa | 12231 | int retry = 3; |
eee67a47 | 12232 | unsigned long long completed; |
ae9f01f8 AK |
12233 | /* to_complete : new sync_max position */ |
12234 | unsigned long long to_complete = sra->reshape_progress; | |
12235 | unsigned long long position_to_set = to_complete / ndata; | |
eee67a47 | 12236 | |
4389ce73 | 12237 | if (!is_fd_valid(fd)) { |
1ade5cc1 | 12238 | dprintf("cannot open reshape_position\n"); |
eee67a47 | 12239 | return 1; |
ae9f01f8 | 12240 | } |
eee67a47 | 12241 | |
df2647fa PB |
12242 | do { |
12243 | if (sysfs_fd_get_ll(fd, &completed) < 0) { | |
12244 | if (!retry) { | |
12245 | dprintf("cannot read reshape_position (no reshape in progres)\n"); | |
12246 | close(fd); | |
12247 | return 1; | |
12248 | } | |
239b3cc0 | 12249 | sleep_for(0, MSEC_TO_NSEC(30), true); |
df2647fa PB |
12250 | } else |
12251 | break; | |
12252 | } while (retry--); | |
eee67a47 | 12253 | |
85ca499c | 12254 | if (completed > position_to_set) { |
1ade5cc1 | 12255 | dprintf("wrong next position to set %llu (%llu)\n", |
85ca499c | 12256 | to_complete, position_to_set); |
ae9f01f8 AK |
12257 | close(fd); |
12258 | return -1; | |
12259 | } | |
12260 | dprintf("Position set: %llu\n", position_to_set); | |
12261 | if (sysfs_set_num(sra, NULL, "sync_max", | |
12262 | position_to_set) != 0) { | |
1ade5cc1 | 12263 | dprintf("cannot set reshape position to %llu\n", |
ae9f01f8 AK |
12264 | position_to_set); |
12265 | close(fd); | |
12266 | return -1; | |
eee67a47 AK |
12267 | } |
12268 | ||
eee67a47 | 12269 | do { |
0febb20c | 12270 | int rc; |
90fd7001 | 12271 | char action[SYSFS_MAX_BUF_SIZE]; |
5ff3a780 | 12272 | int timeout = 3000; |
0febb20c | 12273 | |
5ff3a780 | 12274 | sysfs_wait(fd, &timeout); |
a47e44fb | 12275 | if (sysfs_get_str(sra, NULL, "sync_action", |
90fd7001 | 12276 | action, sizeof(action)) > 0 && |
d7d3809a | 12277 | strncmp(action, "reshape", 7) != 0) { |
b2be2b62 AO |
12278 | if (strncmp(action, "idle", 4) == 0) |
12279 | break; | |
d7d3809a AP |
12280 | close(fd); |
12281 | return -1; | |
12282 | } | |
0febb20c AO |
12283 | |
12284 | rc = read_completed(fd, &completed); | |
12285 | if (rc < 0) { | |
1ade5cc1 | 12286 | dprintf("cannot read reshape_position (in loop)\n"); |
eee67a47 AK |
12287 | close(fd); |
12288 | return 1; | |
0febb20c AO |
12289 | } else if (rc == COMPLETED_NONE) |
12290 | break; | |
85ca499c | 12291 | } while (completed < position_to_set); |
b2be2b62 | 12292 | |
eee67a47 AK |
12293 | close(fd); |
12294 | return 0; | |
eee67a47 AK |
12295 | } |
12296 | ||
b915c95f AK |
12297 | /******************************************************************************* |
12298 | * Function: check_degradation_change | |
12299 | * Description: Check that array hasn't become failed. | |
12300 | * Parameters: | |
12301 | * info : for sysfs access | |
12302 | * sources : source disks descriptors | |
12303 | * degraded: previous degradation level | |
12304 | * Returns: | |
12305 | * degradation level | |
12306 | ******************************************************************************/ | |
12307 | int check_degradation_change(struct mdinfo *info, | |
12308 | int *sources, | |
12309 | int degraded) | |
12310 | { | |
12311 | unsigned long long new_degraded; | |
e1993023 LD |
12312 | int rv; |
12313 | ||
12314 | rv = sysfs_get_ll(info, NULL, "degraded", &new_degraded); | |
089f9d79 | 12315 | if (rv == -1 || (new_degraded != (unsigned long long)degraded)) { |
b915c95f AK |
12316 | /* check each device to ensure it is still working */ |
12317 | struct mdinfo *sd; | |
12318 | new_degraded = 0; | |
12319 | for (sd = info->devs ; sd ; sd = sd->next) { | |
12320 | if (sd->disk.state & (1<<MD_DISK_FAULTY)) | |
12321 | continue; | |
12322 | if (sd->disk.state & (1<<MD_DISK_SYNC)) { | |
90fd7001 | 12323 | char sbuf[SYSFS_MAX_BUF_SIZE]; |
4389ce73 | 12324 | int raid_disk = sd->disk.raid_disk; |
cf52eff5 | 12325 | |
b915c95f | 12326 | if (sysfs_get_str(info, |
cf52eff5 | 12327 | sd, "state", sbuf, sizeof(sbuf)) < 0 || |
b915c95f AK |
12328 | strstr(sbuf, "faulty") || |
12329 | strstr(sbuf, "in_sync") == NULL) { | |
12330 | /* this device is dead */ | |
12331 | sd->disk.state = (1<<MD_DISK_FAULTY); | |
4389ce73 MT |
12332 | if (raid_disk >= 0) |
12333 | close_fd(&sources[raid_disk]); | |
b915c95f AK |
12334 | new_degraded++; |
12335 | } | |
12336 | } | |
12337 | } | |
12338 | } | |
12339 | ||
12340 | return new_degraded; | |
12341 | } | |
12342 | ||
10f22854 AK |
12343 | /******************************************************************************* |
12344 | * Function: imsm_manage_reshape | |
12345 | * Description: Function finds array under reshape and it manages reshape | |
12346 | * process. It creates stripes backups (if required) and sets | |
942e1cdb | 12347 | * checkpoints. |
10f22854 AK |
12348 | * Parameters: |
12349 | * afd : Backup handle (nattive) - not used | |
12350 | * sra : general array info | |
12351 | * reshape : reshape parameters - not used | |
12352 | * st : supertype structure | |
12353 | * blocks : size of critical section [blocks] | |
12354 | * fds : table of source device descriptor | |
12355 | * offsets : start of array (offest per devices) | |
12356 | * dests : not used | |
12357 | * destfd : table of destination device descriptor | |
12358 | * destoffsets : table of destination offsets (per device) | |
12359 | * Returns: | |
12360 | * 1 : success, reshape is done | |
12361 | * 0 : fail | |
12362 | ******************************************************************************/ | |
999b4972 N |
12363 | static int imsm_manage_reshape( |
12364 | int afd, struct mdinfo *sra, struct reshape *reshape, | |
10f22854 | 12365 | struct supertype *st, unsigned long backup_blocks, |
999b4972 N |
12366 | int *fds, unsigned long long *offsets, |
12367 | int dests, int *destfd, unsigned long long *destoffsets) | |
12368 | { | |
10f22854 AK |
12369 | int ret_val = 0; |
12370 | struct intel_super *super = st->sb; | |
594dc1b8 | 12371 | struct intel_dev *dv; |
de44e46f | 12372 | unsigned int sector_size = super->sector_size; |
10f22854 | 12373 | struct imsm_dev *dev = NULL; |
9529d343 | 12374 | struct imsm_map *map_src, *map_dest; |
10f22854 AK |
12375 | int migr_vol_qan = 0; |
12376 | int ndata, odata; /* [bytes] */ | |
12377 | int chunk; /* [bytes] */ | |
12378 | struct migr_record *migr_rec; | |
12379 | char *buf = NULL; | |
12380 | unsigned int buf_size; /* [bytes] */ | |
12381 | unsigned long long max_position; /* array size [bytes] */ | |
12382 | unsigned long long next_step; /* [blocks]/[bytes] */ | |
12383 | unsigned long long old_data_stripe_length; | |
10f22854 AK |
12384 | unsigned long long start_src; /* [bytes] */ |
12385 | unsigned long long start; /* [bytes] */ | |
12386 | unsigned long long start_buf_shift; /* [bytes] */ | |
b915c95f | 12387 | int degraded = 0; |
ab724b98 | 12388 | int source_layout = 0; |
895ffd99 | 12389 | int subarray_index = -1; |
10f22854 | 12390 | |
79a16a9b JS |
12391 | if (!sra) |
12392 | return ret_val; | |
12393 | ||
12394 | if (!fds || !offsets) | |
10f22854 AK |
12395 | goto abort; |
12396 | ||
12397 | /* Find volume during the reshape */ | |
12398 | for (dv = super->devlist; dv; dv = dv->next) { | |
fc54fe7a JS |
12399 | if (dv->dev->vol.migr_type == MIGR_GEN_MIGR && |
12400 | dv->dev->vol.migr_state == 1) { | |
10f22854 AK |
12401 | dev = dv->dev; |
12402 | migr_vol_qan++; | |
895ffd99 | 12403 | subarray_index = dv->index; |
10f22854 AK |
12404 | } |
12405 | } | |
12406 | /* Only one volume can migrate at the same time */ | |
12407 | if (migr_vol_qan != 1) { | |
676e87a8 | 12408 | pr_err("%s", migr_vol_qan ? |
10f22854 AK |
12409 | "Number of migrating volumes greater than 1\n" : |
12410 | "There is no volume during migrationg\n"); | |
12411 | goto abort; | |
12412 | } | |
12413 | ||
9529d343 | 12414 | map_dest = get_imsm_map(dev, MAP_0); |
238c0a71 | 12415 | map_src = get_imsm_map(dev, MAP_1); |
10f22854 AK |
12416 | if (map_src == NULL) |
12417 | goto abort; | |
10f22854 | 12418 | |
9529d343 MD |
12419 | ndata = imsm_num_data_members(map_dest); |
12420 | odata = imsm_num_data_members(map_src); | |
10f22854 | 12421 | |
7b1ab482 | 12422 | chunk = __le16_to_cpu(map_src->blocks_per_strip) * 512; |
10f22854 AK |
12423 | old_data_stripe_length = odata * chunk; |
12424 | ||
12425 | migr_rec = super->migr_rec; | |
12426 | ||
10f22854 AK |
12427 | /* initialize migration record for start condition */ |
12428 | if (sra->reshape_progress == 0) | |
12429 | init_migr_record_imsm(st, dev, sra); | |
b2c59438 AK |
12430 | else { |
12431 | if (__le32_to_cpu(migr_rec->rec_status) != UNIT_SRC_NORMAL) { | |
7a862a02 | 12432 | dprintf("imsm: cannot restart migration when data are present in copy area.\n"); |
b2c59438 AK |
12433 | goto abort; |
12434 | } | |
6a75c8ca AK |
12435 | /* Save checkpoint to update migration record for current |
12436 | * reshape position (in md). It can be farther than current | |
12437 | * reshape position in metadata. | |
12438 | */ | |
12439 | if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) { | |
12440 | /* ignore error == 2, this can mean end of reshape here | |
12441 | */ | |
7a862a02 | 12442 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL, initial save)\n"); |
6a75c8ca AK |
12443 | goto abort; |
12444 | } | |
b2c59438 | 12445 | } |
10f22854 AK |
12446 | |
12447 | /* size for data */ | |
12448 | buf_size = __le32_to_cpu(migr_rec->blocks_per_unit) * 512; | |
12449 | /* extend buffer size for parity disk */ | |
12450 | buf_size += __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
3e684231 | 12451 | /* add space for stripe alignment */ |
10f22854 | 12452 | buf_size += old_data_stripe_length; |
de44e46f PB |
12453 | if (posix_memalign((void **)&buf, MAX_SECTOR_SIZE, buf_size)) { |
12454 | dprintf("imsm: Cannot allocate checkpoint buffer\n"); | |
10f22854 AK |
12455 | goto abort; |
12456 | } | |
12457 | ||
3ef4403c | 12458 | max_position = sra->component_size * ndata; |
68eb8bc6 | 12459 | source_layout = imsm_level_to_layout(map_src->raid_level); |
10f22854 | 12460 | |
9f421827 PB |
12461 | while (current_migr_unit(migr_rec) < |
12462 | get_num_migr_units(migr_rec)) { | |
10f22854 AK |
12463 | /* current reshape position [blocks] */ |
12464 | unsigned long long current_position = | |
12465 | __le32_to_cpu(migr_rec->blocks_per_unit) | |
9f421827 | 12466 | * current_migr_unit(migr_rec); |
10f22854 AK |
12467 | unsigned long long border; |
12468 | ||
b915c95f AK |
12469 | /* Check that array hasn't become failed. |
12470 | */ | |
12471 | degraded = check_degradation_change(sra, fds, degraded); | |
12472 | if (degraded > 1) { | |
7a862a02 | 12473 | dprintf("imsm: Abort reshape due to degradation level (%i)\n", degraded); |
b915c95f AK |
12474 | goto abort; |
12475 | } | |
12476 | ||
10f22854 AK |
12477 | next_step = __le32_to_cpu(migr_rec->blocks_per_unit); |
12478 | ||
12479 | if ((current_position + next_step) > max_position) | |
12480 | next_step = max_position - current_position; | |
12481 | ||
92144abf | 12482 | start = current_position * 512; |
10f22854 | 12483 | |
942e1cdb | 12484 | /* align reading start to old geometry */ |
10f22854 AK |
12485 | start_buf_shift = start % old_data_stripe_length; |
12486 | start_src = start - start_buf_shift; | |
12487 | ||
12488 | border = (start_src / odata) - (start / ndata); | |
12489 | border /= 512; | |
12490 | if (border <= __le32_to_cpu(migr_rec->dest_depth_per_unit)) { | |
12491 | /* save critical stripes to buf | |
12492 | * start - start address of current unit | |
12493 | * to backup [bytes] | |
12494 | * start_src - start address of current unit | |
12495 | * to backup alligned to source array | |
12496 | * [bytes] | |
12497 | */ | |
594dc1b8 | 12498 | unsigned long long next_step_filler; |
10f22854 AK |
12499 | unsigned long long copy_length = next_step * 512; |
12500 | ||
12501 | /* allign copy area length to stripe in old geometry */ | |
12502 | next_step_filler = ((copy_length + start_buf_shift) | |
12503 | % old_data_stripe_length); | |
12504 | if (next_step_filler) | |
12505 | next_step_filler = (old_data_stripe_length | |
12506 | - next_step_filler); | |
7a862a02 | 12507 | dprintf("save_stripes() parameters: start = %llu,\tstart_src = %llu,\tnext_step*512 = %llu,\tstart_in_buf_shift = %llu,\tnext_step_filler = %llu\n", |
10f22854 AK |
12508 | start, start_src, copy_length, |
12509 | start_buf_shift, next_step_filler); | |
12510 | ||
12511 | if (save_stripes(fds, offsets, map_src->num_members, | |
ab724b98 AK |
12512 | chunk, map_src->raid_level, |
12513 | source_layout, 0, NULL, start_src, | |
10f22854 AK |
12514 | copy_length + |
12515 | next_step_filler + start_buf_shift, | |
12516 | buf)) { | |
7a862a02 | 12517 | dprintf("imsm: Cannot save stripes to buffer\n"); |
10f22854 AK |
12518 | goto abort; |
12519 | } | |
12520 | /* Convert data to destination format and store it | |
12521 | * in backup general migration area | |
12522 | */ | |
12523 | if (save_backup_imsm(st, dev, sra, | |
aea93171 | 12524 | buf + start_buf_shift, copy_length)) { |
7a862a02 | 12525 | dprintf("imsm: Cannot save stripes to target devices\n"); |
10f22854 AK |
12526 | goto abort; |
12527 | } | |
12528 | if (save_checkpoint_imsm(st, sra, | |
12529 | UNIT_SRC_IN_CP_AREA)) { | |
7a862a02 | 12530 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_IN_CP_AREA)\n"); |
10f22854 AK |
12531 | goto abort; |
12532 | } | |
8016a6d4 AK |
12533 | } else { |
12534 | /* set next step to use whole border area */ | |
12535 | border /= next_step; | |
12536 | if (border > 1) | |
12537 | next_step *= border; | |
10f22854 AK |
12538 | } |
12539 | /* When data backed up, checkpoint stored, | |
12540 | * kick the kernel to reshape unit of data | |
12541 | */ | |
12542 | next_step = next_step + sra->reshape_progress; | |
8016a6d4 AK |
12543 | /* limit next step to array max position */ |
12544 | if (next_step > max_position) | |
12545 | next_step = max_position; | |
10f22854 AK |
12546 | sysfs_set_num(sra, NULL, "suspend_lo", sra->reshape_progress); |
12547 | sysfs_set_num(sra, NULL, "suspend_hi", next_step); | |
ae9f01f8 | 12548 | sra->reshape_progress = next_step; |
10f22854 AK |
12549 | |
12550 | /* wait until reshape finish */ | |
c85338c6 | 12551 | if (wait_for_reshape_imsm(sra, ndata)) { |
c47b0ff6 AK |
12552 | dprintf("wait_for_reshape_imsm returned error!\n"); |
12553 | goto abort; | |
12554 | } | |
84d11e6c N |
12555 | if (sigterm) |
12556 | goto abort; | |
10f22854 | 12557 | |
0228d92c AK |
12558 | if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) { |
12559 | /* ignore error == 2, this can mean end of reshape here | |
12560 | */ | |
7a862a02 | 12561 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL)\n"); |
10f22854 AK |
12562 | goto abort; |
12563 | } | |
12564 | ||
12565 | } | |
12566 | ||
71e5411e PB |
12567 | /* clear migr_rec on disks after successful migration */ |
12568 | struct dl *d; | |
12569 | ||
85337573 | 12570 | memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE); |
71e5411e PB |
12571 | for (d = super->disks; d; d = d->next) { |
12572 | if (d->index < 0 || is_failed(&d->disk)) | |
12573 | continue; | |
12574 | unsigned long long dsize; | |
12575 | ||
12576 | get_dev_size(d->fd, NULL, &dsize); | |
de44e46f | 12577 | if (lseek64(d->fd, dsize - MIGR_REC_SECTOR_POSITION*sector_size, |
71e5411e | 12578 | SEEK_SET) >= 0) { |
466070ad | 12579 | if ((unsigned int)write(d->fd, super->migr_rec_buf, |
de44e46f PB |
12580 | MIGR_REC_BUF_SECTORS*sector_size) != |
12581 | MIGR_REC_BUF_SECTORS*sector_size) | |
71e5411e PB |
12582 | perror("Write migr_rec failed"); |
12583 | } | |
12584 | } | |
12585 | ||
10f22854 AK |
12586 | /* return '1' if done */ |
12587 | ret_val = 1; | |
895ffd99 MT |
12588 | |
12589 | /* After the reshape eliminate size mismatch in metadata. | |
12590 | * Don't update md/component_size here, volume hasn't | |
12591 | * to take whole space. It is allowed by kernel. | |
12592 | * md/component_size will be set propoperly after next assembly. | |
12593 | */ | |
12594 | imsm_fix_size_mismatch(st, subarray_index); | |
12595 | ||
10f22854 AK |
12596 | abort: |
12597 | free(buf); | |
942e1cdb N |
12598 | /* See Grow.c: abort_reshape() for further explanation */ |
12599 | sysfs_set_num(sra, NULL, "suspend_lo", 0x7FFFFFFFFFFFFFFFULL); | |
12600 | sysfs_set_num(sra, NULL, "suspend_hi", 0); | |
12601 | sysfs_set_num(sra, NULL, "suspend_lo", 0); | |
10f22854 AK |
12602 | |
12603 | return ret_val; | |
999b4972 | 12604 | } |
0c21b485 | 12605 | |
fbc42556 JR |
12606 | /******************************************************************************* |
12607 | * Function: calculate_bitmap_min_chunksize | |
12608 | * Description: Calculates the minimal valid bitmap chunk size | |
12609 | * Parameters: | |
12610 | * max_bits : indicate how many bits can be used for the bitmap | |
12611 | * data_area_size : the size of the data area covered by the bitmap | |
12612 | * | |
12613 | * Returns: | |
12614 | * The bitmap chunk size | |
12615 | ******************************************************************************/ | |
12616 | static unsigned long long | |
12617 | calculate_bitmap_min_chunksize(unsigned long long max_bits, | |
12618 | unsigned long long data_area_size) | |
12619 | { | |
12620 | unsigned long long min_chunk = | |
12621 | 4096; /* sub-page chunks don't work yet.. */ | |
12622 | unsigned long long bits = data_area_size / min_chunk + 1; | |
12623 | ||
12624 | while (bits > max_bits) { | |
12625 | min_chunk *= 2; | |
12626 | bits = (bits + 1) / 2; | |
12627 | } | |
12628 | return min_chunk; | |
12629 | } | |
12630 | ||
12631 | /******************************************************************************* | |
12632 | * Function: calculate_bitmap_chunksize | |
12633 | * Description: Calculates the bitmap chunk size for the given device | |
12634 | * Parameters: | |
12635 | * st : supertype information | |
12636 | * dev : device for the bitmap | |
12637 | * | |
12638 | * Returns: | |
12639 | * The bitmap chunk size | |
12640 | ******************************************************************************/ | |
12641 | static unsigned long long calculate_bitmap_chunksize(struct supertype *st, | |
12642 | struct imsm_dev *dev) | |
12643 | { | |
12644 | struct intel_super *super = st->sb; | |
12645 | unsigned long long min_chunksize; | |
12646 | unsigned long long result = IMSM_DEFAULT_BITMAP_CHUNKSIZE; | |
12647 | size_t dev_size = imsm_dev_size(dev); | |
12648 | ||
12649 | min_chunksize = calculate_bitmap_min_chunksize( | |
12650 | IMSM_BITMAP_AREA_SIZE * super->sector_size, dev_size); | |
12651 | ||
12652 | if (result < min_chunksize) | |
12653 | result = min_chunksize; | |
12654 | ||
12655 | return result; | |
12656 | } | |
12657 | ||
12658 | /******************************************************************************* | |
12659 | * Function: init_bitmap_header | |
12660 | * Description: Initialize the bitmap header structure | |
12661 | * Parameters: | |
12662 | * st : supertype information | |
12663 | * bms : bitmap header struct to initialize | |
12664 | * dev : device for the bitmap | |
12665 | * | |
12666 | * Returns: | |
12667 | * 0 : success | |
12668 | * -1 : fail | |
12669 | ******************************************************************************/ | |
12670 | static int init_bitmap_header(struct supertype *st, struct bitmap_super_s *bms, | |
12671 | struct imsm_dev *dev) | |
12672 | { | |
12673 | int vol_uuid[4]; | |
12674 | ||
12675 | if (!bms || !dev) | |
12676 | return -1; | |
12677 | ||
12678 | bms->magic = __cpu_to_le32(BITMAP_MAGIC); | |
12679 | bms->version = __cpu_to_le32(BITMAP_MAJOR_HI); | |
12680 | bms->daemon_sleep = __cpu_to_le32(IMSM_DEFAULT_BITMAP_DAEMON_SLEEP); | |
12681 | bms->sync_size = __cpu_to_le64(IMSM_BITMAP_AREA_SIZE); | |
12682 | bms->write_behind = __cpu_to_le32(0); | |
12683 | ||
12684 | uuid_from_super_imsm(st, vol_uuid); | |
12685 | memcpy(bms->uuid, vol_uuid, 16); | |
12686 | ||
12687 | bms->chunksize = calculate_bitmap_chunksize(st, dev); | |
12688 | ||
12689 | return 0; | |
12690 | } | |
12691 | ||
12692 | /******************************************************************************* | |
12693 | * Function: validate_internal_bitmap_for_drive | |
12694 | * Description: Verify if the bitmap header for a given drive. | |
12695 | * Parameters: | |
12696 | * st : supertype information | |
12697 | * offset : The offset from the beginning of the drive where to look for | |
12698 | * the bitmap header. | |
12699 | * d : the drive info | |
12700 | * | |
12701 | * Returns: | |
12702 | * 0 : success | |
12703 | * -1 : fail | |
12704 | ******************************************************************************/ | |
12705 | static int validate_internal_bitmap_for_drive(struct supertype *st, | |
12706 | unsigned long long offset, | |
12707 | struct dl *d) | |
12708 | { | |
12709 | struct intel_super *super = st->sb; | |
12710 | int ret = -1; | |
12711 | int vol_uuid[4]; | |
12712 | bitmap_super_t *bms; | |
12713 | int fd; | |
12714 | ||
12715 | if (!d) | |
12716 | return -1; | |
12717 | ||
12718 | void *read_buf; | |
12719 | ||
12720 | if (posix_memalign(&read_buf, MAX_SECTOR_SIZE, IMSM_BITMAP_HEADER_SIZE)) | |
12721 | return -1; | |
12722 | ||
12723 | fd = d->fd; | |
4389ce73 | 12724 | if (!is_fd_valid(fd)) { |
fbc42556 | 12725 | fd = open(d->devname, O_RDONLY, 0); |
4389ce73 MT |
12726 | |
12727 | if (!is_fd_valid(fd)) { | |
fbc42556 JR |
12728 | dprintf("cannot open the device %s\n", d->devname); |
12729 | goto abort; | |
12730 | } | |
12731 | } | |
12732 | ||
12733 | if (lseek64(fd, offset * super->sector_size, SEEK_SET) < 0) | |
12734 | goto abort; | |
12735 | if (read(fd, read_buf, IMSM_BITMAP_HEADER_SIZE) != | |
12736 | IMSM_BITMAP_HEADER_SIZE) | |
12737 | goto abort; | |
12738 | ||
12739 | uuid_from_super_imsm(st, vol_uuid); | |
12740 | ||
12741 | bms = read_buf; | |
12742 | if ((bms->magic != __cpu_to_le32(BITMAP_MAGIC)) || | |
12743 | (bms->version != __cpu_to_le32(BITMAP_MAJOR_HI)) || | |
12744 | (!same_uuid((int *)bms->uuid, vol_uuid, st->ss->swapuuid))) { | |
12745 | dprintf("wrong bitmap header detected\n"); | |
12746 | goto abort; | |
12747 | } | |
12748 | ||
12749 | ret = 0; | |
12750 | abort: | |
4389ce73 MT |
12751 | if (!is_fd_valid(d->fd)) |
12752 | close_fd(&fd); | |
12753 | ||
fbc42556 JR |
12754 | if (read_buf) |
12755 | free(read_buf); | |
12756 | ||
12757 | return ret; | |
12758 | } | |
12759 | ||
12760 | /******************************************************************************* | |
12761 | * Function: validate_internal_bitmap_imsm | |
12762 | * Description: Verify if the bitmap header is in place and with proper data. | |
12763 | * Parameters: | |
12764 | * st : supertype information | |
12765 | * | |
12766 | * Returns: | |
12767 | * 0 : success or device w/o RWH_BITMAP | |
12768 | * -1 : fail | |
12769 | ******************************************************************************/ | |
12770 | static int validate_internal_bitmap_imsm(struct supertype *st) | |
12771 | { | |
12772 | struct intel_super *super = st->sb; | |
12773 | struct imsm_dev *dev = get_imsm_dev(super, super->current_vol); | |
12774 | unsigned long long offset; | |
12775 | struct dl *d; | |
12776 | ||
fbc42556 JR |
12777 | if (dev->rwh_policy != RWH_BITMAP) |
12778 | return 0; | |
12779 | ||
12780 | offset = get_bitmap_header_sector(super, super->current_vol); | |
12781 | for (d = super->disks; d; d = d->next) { | |
12782 | if (d->index < 0 || is_failed(&d->disk)) | |
12783 | continue; | |
12784 | ||
12785 | if (validate_internal_bitmap_for_drive(st, offset, d)) { | |
12786 | pr_err("imsm: bitmap validation failed\n"); | |
12787 | return -1; | |
12788 | } | |
12789 | } | |
12790 | return 0; | |
12791 | } | |
12792 | ||
12793 | /******************************************************************************* | |
12794 | * Function: add_internal_bitmap_imsm | |
12795 | * Description: Mark the volume to use the bitmap and updates the chunk size value. | |
12796 | * Parameters: | |
12797 | * st : supertype information | |
12798 | * chunkp : bitmap chunk size | |
12799 | * delay : not used for imsm | |
12800 | * write_behind : not used for imsm | |
12801 | * size : not used for imsm | |
12802 | * may_change : not used for imsm | |
12803 | * amajor : not used for imsm | |
12804 | * | |
12805 | * Returns: | |
12806 | * 0 : success | |
12807 | * -1 : fail | |
12808 | ******************************************************************************/ | |
12809 | static int add_internal_bitmap_imsm(struct supertype *st, int *chunkp, | |
12810 | int delay, int write_behind, | |
12811 | unsigned long long size, int may_change, | |
12812 | int amajor) | |
12813 | { | |
12814 | struct intel_super *super = st->sb; | |
12815 | int vol_idx = super->current_vol; | |
12816 | struct imsm_dev *dev; | |
12817 | ||
12818 | if (!super->devlist || vol_idx == -1 || !chunkp) | |
12819 | return -1; | |
12820 | ||
12821 | dev = get_imsm_dev(super, vol_idx); | |
fbc42556 | 12822 | dev->rwh_policy = RWH_BITMAP; |
fbc42556 | 12823 | *chunkp = calculate_bitmap_chunksize(st, dev); |
fbc42556 JR |
12824 | return 0; |
12825 | } | |
12826 | ||
12827 | /******************************************************************************* | |
12828 | * Function: locate_bitmap_imsm | |
12829 | * Description: Seek 'fd' to start of write-intent-bitmap. | |
12830 | * Parameters: | |
12831 | * st : supertype information | |
12832 | * fd : file descriptor for the device | |
12833 | * node_num : not used for imsm | |
12834 | * | |
12835 | * Returns: | |
12836 | * 0 : success | |
12837 | * -1 : fail | |
12838 | ******************************************************************************/ | |
12839 | static int locate_bitmap_imsm(struct supertype *st, int fd, int node_num) | |
12840 | { | |
12841 | struct intel_super *super = st->sb; | |
12842 | unsigned long long offset; | |
12843 | int vol_idx = super->current_vol; | |
12844 | ||
12845 | if (!super->devlist || vol_idx == -1) | |
12846 | return -1; | |
12847 | ||
12848 | offset = get_bitmap_header_sector(super, super->current_vol); | |
12849 | dprintf("bitmap header offset is %llu\n", offset); | |
12850 | ||
12851 | lseek64(fd, offset << 9, 0); | |
12852 | ||
12853 | return 0; | |
12854 | } | |
12855 | ||
12856 | /******************************************************************************* | |
12857 | * Function: write_init_bitmap_imsm | |
12858 | * Description: Write a bitmap header and prepares the area for the bitmap. | |
12859 | * Parameters: | |
12860 | * st : supertype information | |
12861 | * fd : file descriptor for the device | |
12862 | * update : not used for imsm | |
12863 | * | |
12864 | * Returns: | |
12865 | * 0 : success | |
12866 | * -1 : fail | |
12867 | ******************************************************************************/ | |
12868 | static int write_init_bitmap_imsm(struct supertype *st, int fd, | |
12869 | enum bitmap_update update) | |
12870 | { | |
12871 | struct intel_super *super = st->sb; | |
12872 | int vol_idx = super->current_vol; | |
12873 | int ret = 0; | |
12874 | unsigned long long offset; | |
12875 | bitmap_super_t bms = { 0 }; | |
12876 | size_t written = 0; | |
12877 | size_t to_write; | |
12878 | ssize_t rv_num; | |
12879 | void *buf; | |
12880 | ||
12881 | if (!super->devlist || !super->sector_size || vol_idx == -1) | |
12882 | return -1; | |
12883 | ||
12884 | struct imsm_dev *dev = get_imsm_dev(super, vol_idx); | |
12885 | ||
12886 | /* first clear the space for bitmap header */ | |
12887 | unsigned long long bitmap_area_start = | |
12888 | get_bitmap_header_sector(super, vol_idx); | |
12889 | ||
12890 | dprintf("zeroing area start (%llu) and size (%u)\n", bitmap_area_start, | |
12891 | IMSM_BITMAP_AND_HEADER_SIZE / super->sector_size); | |
12892 | if (zero_disk_range(fd, bitmap_area_start, | |
12893 | IMSM_BITMAP_HEADER_SIZE / super->sector_size)) { | |
12894 | pr_err("imsm: cannot zeroing the space for the bitmap\n"); | |
12895 | return -1; | |
12896 | } | |
12897 | ||
12898 | /* The bitmap area should be filled with "1"s to perform initial | |
12899 | * synchronization. | |
12900 | */ | |
12901 | if (posix_memalign(&buf, MAX_SECTOR_SIZE, MAX_SECTOR_SIZE)) | |
12902 | return -1; | |
12903 | memset(buf, 0xFF, MAX_SECTOR_SIZE); | |
12904 | offset = get_bitmap_sector(super, vol_idx); | |
12905 | lseek64(fd, offset << 9, 0); | |
12906 | while (written < IMSM_BITMAP_AREA_SIZE) { | |
12907 | to_write = IMSM_BITMAP_AREA_SIZE - written; | |
12908 | if (to_write > MAX_SECTOR_SIZE) | |
12909 | to_write = MAX_SECTOR_SIZE; | |
12910 | rv_num = write(fd, buf, MAX_SECTOR_SIZE); | |
12911 | if (rv_num != MAX_SECTOR_SIZE) { | |
12912 | ret = -1; | |
12913 | dprintf("cannot initialize bitmap area\n"); | |
12914 | goto abort; | |
12915 | } | |
12916 | written += rv_num; | |
12917 | } | |
12918 | ||
12919 | /* write a bitmap header */ | |
12920 | init_bitmap_header(st, &bms, dev); | |
12921 | memset(buf, 0, MAX_SECTOR_SIZE); | |
12922 | memcpy(buf, &bms, sizeof(bitmap_super_t)); | |
12923 | if (locate_bitmap_imsm(st, fd, 0)) { | |
12924 | ret = -1; | |
12925 | dprintf("cannot locate the bitmap\n"); | |
12926 | goto abort; | |
12927 | } | |
12928 | if (write(fd, buf, MAX_SECTOR_SIZE) != MAX_SECTOR_SIZE) { | |
12929 | ret = -1; | |
12930 | dprintf("cannot write the bitmap header\n"); | |
12931 | goto abort; | |
12932 | } | |
12933 | fsync(fd); | |
12934 | ||
12935 | abort: | |
12936 | free(buf); | |
12937 | ||
12938 | return ret; | |
12939 | } | |
12940 | ||
12941 | /******************************************************************************* | |
12942 | * Function: is_vol_to_setup_bitmap | |
12943 | * Description: Checks if a bitmap should be activated on the dev. | |
12944 | * Parameters: | |
12945 | * info : info about the volume to setup the bitmap | |
12946 | * dev : the device to check against bitmap creation | |
12947 | * | |
12948 | * Returns: | |
12949 | * 0 : bitmap should be set up on the device | |
12950 | * -1 : otherwise | |
12951 | ******************************************************************************/ | |
12952 | static int is_vol_to_setup_bitmap(struct mdinfo *info, struct imsm_dev *dev) | |
12953 | { | |
12954 | if (!dev || !info) | |
12955 | return -1; | |
12956 | ||
12957 | if ((strcmp((char *)dev->volume, info->name) == 0) && | |
12958 | (dev->rwh_policy == RWH_BITMAP)) | |
12959 | return -1; | |
12960 | ||
12961 | return 0; | |
12962 | } | |
12963 | ||
12964 | /******************************************************************************* | |
12965 | * Function: set_bitmap_sysfs | |
12966 | * Description: Set the sysfs atributes of a given volume to activate the bitmap. | |
12967 | * Parameters: | |
12968 | * info : info about the volume where the bitmap should be setup | |
12969 | * chunksize : bitmap chunk size | |
12970 | * location : location of the bitmap | |
12971 | * | |
12972 | * Returns: | |
12973 | * 0 : success | |
12974 | * -1 : fail | |
12975 | ******************************************************************************/ | |
12976 | static int set_bitmap_sysfs(struct mdinfo *info, unsigned long long chunksize, | |
12977 | char *location) | |
12978 | { | |
12979 | /* The bitmap/metadata is set to external to allow changing of value for | |
12980 | * bitmap/location. When external is used, the kernel will treat an offset | |
12981 | * related to the device's first lba (in opposition to the "internal" case | |
12982 | * when this value is related to the beginning of the superblock). | |
12983 | */ | |
12984 | if (sysfs_set_str(info, NULL, "bitmap/metadata", "external")) { | |
12985 | dprintf("failed to set bitmap/metadata\n"); | |
12986 | return -1; | |
12987 | } | |
12988 | ||
12989 | /* It can only be changed when no bitmap is active. | |
12990 | * Should be bigger than 512 and must be power of 2. | |
12991 | * It is expecting the value in bytes. | |
12992 | */ | |
12993 | if (sysfs_set_num(info, NULL, "bitmap/chunksize", | |
12994 | __cpu_to_le32(chunksize))) { | |
12995 | dprintf("failed to set bitmap/chunksize\n"); | |
12996 | return -1; | |
12997 | } | |
12998 | ||
12999 | /* It is expecting the value in sectors. */ | |
13000 | if (sysfs_set_num(info, NULL, "bitmap/space", | |
13001 | __cpu_to_le64(IMSM_BITMAP_AREA_SIZE))) { | |
13002 | dprintf("failed to set bitmap/space\n"); | |
13003 | return -1; | |
13004 | } | |
13005 | ||
13006 | /* Determines the delay between the bitmap updates. | |
13007 | * It is expecting the value in seconds. | |
13008 | */ | |
13009 | if (sysfs_set_num(info, NULL, "bitmap/time_base", | |
13010 | __cpu_to_le64(IMSM_DEFAULT_BITMAP_DAEMON_SLEEP))) { | |
13011 | dprintf("failed to set bitmap/time_base\n"); | |
13012 | return -1; | |
13013 | } | |
13014 | ||
13015 | /* It is expecting the value in sectors with a sign at the beginning. */ | |
13016 | if (sysfs_set_str(info, NULL, "bitmap/location", location)) { | |
13017 | dprintf("failed to set bitmap/location\n"); | |
13018 | return -1; | |
13019 | } | |
13020 | ||
13021 | return 0; | |
13022 | } | |
13023 | ||
13024 | /******************************************************************************* | |
13025 | * Function: set_bitmap_imsm | |
13026 | * Description: Setup the bitmap for the given volume | |
13027 | * Parameters: | |
13028 | * st : supertype information | |
13029 | * info : info about the volume where the bitmap should be setup | |
13030 | * | |
13031 | * Returns: | |
13032 | * 0 : success | |
13033 | * -1 : fail | |
13034 | ******************************************************************************/ | |
13035 | static int set_bitmap_imsm(struct supertype *st, struct mdinfo *info) | |
13036 | { | |
13037 | struct intel_super *super = st->sb; | |
13038 | int prev_current_vol = super->current_vol; | |
13039 | struct imsm_dev *dev; | |
13040 | int ret = -1; | |
13041 | char location[16] = ""; | |
13042 | unsigned long long chunksize; | |
13043 | struct intel_dev *dev_it; | |
13044 | ||
13045 | for (dev_it = super->devlist; dev_it; dev_it = dev_it->next) { | |
13046 | super->current_vol = dev_it->index; | |
13047 | dev = get_imsm_dev(super, super->current_vol); | |
13048 | ||
13049 | if (is_vol_to_setup_bitmap(info, dev)) { | |
13050 | if (validate_internal_bitmap_imsm(st)) { | |
13051 | dprintf("bitmap header validation failed\n"); | |
13052 | goto abort; | |
13053 | } | |
13054 | ||
13055 | chunksize = calculate_bitmap_chunksize(st, dev); | |
13056 | dprintf("chunk size is %llu\n", chunksize); | |
13057 | ||
13058 | snprintf(location, sizeof(location), "+%llu", | |
13059 | get_bitmap_sector(super, super->current_vol)); | |
13060 | dprintf("bitmap offset is %s\n", location); | |
13061 | ||
13062 | if (set_bitmap_sysfs(info, chunksize, location)) { | |
13063 | dprintf("cannot setup the bitmap\n"); | |
13064 | goto abort; | |
13065 | } | |
13066 | } | |
13067 | } | |
13068 | ret = 0; | |
13069 | abort: | |
13070 | super->current_vol = prev_current_vol; | |
13071 | return ret; | |
13072 | } | |
13073 | ||
cdddbdbc | 13074 | struct superswitch super_imsm = { |
cdddbdbc DW |
13075 | .examine_super = examine_super_imsm, |
13076 | .brief_examine_super = brief_examine_super_imsm, | |
4737ae25 | 13077 | .brief_examine_subarrays = brief_examine_subarrays_imsm, |
9d84c8ea | 13078 | .export_examine_super = export_examine_super_imsm, |
cdddbdbc DW |
13079 | .detail_super = detail_super_imsm, |
13080 | .brief_detail_super = brief_detail_super_imsm, | |
bf5a934a | 13081 | .write_init_super = write_init_super_imsm, |
0e600426 N |
13082 | .validate_geometry = validate_geometry_imsm, |
13083 | .add_to_super = add_to_super_imsm, | |
1a64be56 | 13084 | .remove_from_super = remove_from_super_imsm, |
d665cc31 | 13085 | .detail_platform = detail_platform_imsm, |
e50cf220 | 13086 | .export_detail_platform = export_detail_platform_imsm, |
33414a01 | 13087 | .kill_subarray = kill_subarray_imsm, |
aa534678 | 13088 | .update_subarray = update_subarray_imsm, |
2b959fbf | 13089 | .load_container = load_container_imsm, |
71204a50 | 13090 | .default_geometry = default_geometry_imsm, |
e21aea08 | 13091 | .test_and_add_drive_policies = test_and_add_drive_policies_imsm, |
71204a50 N |
13092 | .reshape_super = imsm_reshape_super, |
13093 | .manage_reshape = imsm_manage_reshape, | |
9e2d750d | 13094 | .recover_backup = recover_backup_imsm, |
27156a57 | 13095 | .examine_badblocks = examine_badblocks_imsm, |
cdddbdbc DW |
13096 | .match_home = match_home_imsm, |
13097 | .uuid_from_super= uuid_from_super_imsm, | |
13098 | .getinfo_super = getinfo_super_imsm, | |
5c4cd5da | 13099 | .getinfo_super_disks = getinfo_super_disks_imsm, |
cdddbdbc DW |
13100 | .update_super = update_super_imsm, |
13101 | ||
13102 | .avail_size = avail_size_imsm, | |
fbfdcb06 | 13103 | .get_spare_criteria = get_spare_criteria_imsm, |
cdddbdbc DW |
13104 | |
13105 | .compare_super = compare_super_imsm, | |
13106 | ||
13107 | .load_super = load_super_imsm, | |
bf5a934a | 13108 | .init_super = init_super_imsm, |
e683ca88 | 13109 | .store_super = store_super_imsm, |
cdddbdbc DW |
13110 | .free_super = free_super_imsm, |
13111 | .match_metadata_desc = match_metadata_desc_imsm, | |
bf5a934a | 13112 | .container_content = container_content_imsm, |
0c21b485 | 13113 | .validate_container = validate_container_imsm, |
cdddbdbc | 13114 | |
fbc42556 JR |
13115 | .add_internal_bitmap = add_internal_bitmap_imsm, |
13116 | .locate_bitmap = locate_bitmap_imsm, | |
13117 | .write_bitmap = write_init_bitmap_imsm, | |
13118 | .set_bitmap = set_bitmap_imsm, | |
13119 | ||
2432ce9b AP |
13120 | .write_init_ppl = write_init_ppl_imsm, |
13121 | .validate_ppl = validate_ppl_imsm, | |
13122 | ||
cdddbdbc | 13123 | .external = 1, |
1c832795 | 13124 | .swapuuid = 0, |
4cce4069 | 13125 | .name = "imsm", |
845dea95 NB |
13126 | |
13127 | /* for mdmon */ | |
13128 | .open_new = imsm_open_new, | |
ed9d66aa | 13129 | .set_array_state= imsm_set_array_state, |
845dea95 NB |
13130 | .set_disk = imsm_set_disk, |
13131 | .sync_metadata = imsm_sync_metadata, | |
88758e9d | 13132 | .activate_spare = imsm_activate_spare, |
e8319a19 | 13133 | .process_update = imsm_process_update, |
8273f55e | 13134 | .prepare_update = imsm_prepare_update, |
6f50473f | 13135 | .record_bad_block = imsm_record_badblock, |
c07a5a4f | 13136 | .clear_bad_block = imsm_clear_badblock, |
928f1424 | 13137 | .get_bad_blocks = imsm_get_badblocks, |
cdddbdbc | 13138 | }; |