]>
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 | ||
fbfdcb06 AO |
1739 | /* |
1740 | * Return minimum size of a spare and sector size | |
1741 | * that can be used in this array | |
1742 | */ | |
1743 | int get_spare_criteria_imsm(struct supertype *st, struct spare_criteria *c) | |
80e7f8c3 AC |
1744 | { |
1745 | struct intel_super *super = st->sb; | |
1746 | struct dl *dl; | |
1747 | struct extent *e; | |
1748 | int i; | |
fbfdcb06 AO |
1749 | unsigned long long size = 0; |
1750 | ||
1751 | c->min_size = 0; | |
4b57ecf6 | 1752 | c->sector_size = 0; |
80e7f8c3 AC |
1753 | |
1754 | if (!super) | |
fbfdcb06 | 1755 | return -EINVAL; |
80e7f8c3 AC |
1756 | /* find first active disk in array */ |
1757 | dl = super->disks; | |
1758 | while (dl && (is_failed(&dl->disk) || dl->index == -1)) | |
1759 | dl = dl->next; | |
1760 | if (!dl) | |
fbfdcb06 | 1761 | return -EINVAL; |
80e7f8c3 | 1762 | /* find last lba used by subarrays */ |
05501181 | 1763 | e = get_extents(super, dl, 0); |
80e7f8c3 | 1764 | if (!e) |
fbfdcb06 | 1765 | return -EINVAL; |
80e7f8c3 AC |
1766 | for (i = 0; e[i].size; i++) |
1767 | continue; | |
1768 | if (i > 0) | |
fbfdcb06 | 1769 | size = e[i-1].start + e[i-1].size; |
80e7f8c3 | 1770 | free(e); |
b81221b7 | 1771 | |
80e7f8c3 | 1772 | /* add the amount of space needed for metadata */ |
fbfdcb06 AO |
1773 | size += imsm_min_reserved_sectors(super); |
1774 | ||
1775 | c->min_size = size * 512; | |
4b57ecf6 | 1776 | c->sector_size = super->sector_size; |
b81221b7 | 1777 | |
fbfdcb06 | 1778 | return 0; |
80e7f8c3 AC |
1779 | } |
1780 | ||
195d1d76 | 1781 | static bool is_gen_migration(struct imsm_dev *dev); |
d1e02575 | 1782 | |
f36a9ecd PB |
1783 | #define IMSM_4K_DIV 8 |
1784 | ||
c47b0ff6 AK |
1785 | static __u64 blocks_per_migr_unit(struct intel_super *super, |
1786 | struct imsm_dev *dev); | |
1e5c6983 | 1787 | |
c47b0ff6 AK |
1788 | static void print_imsm_dev(struct intel_super *super, |
1789 | struct imsm_dev *dev, | |
1790 | char *uuid, | |
1791 | int disk_idx) | |
cdddbdbc DW |
1792 | { |
1793 | __u64 sz; | |
0d80bb2f | 1794 | int slot, i; |
238c0a71 AK |
1795 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1796 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); | |
b10b37b8 | 1797 | __u32 ord; |
cdddbdbc DW |
1798 | |
1799 | printf("\n"); | |
1e7bc0ed | 1800 | printf("[%.16s]:\n", dev->volume); |
ba1b3bc8 | 1801 | printf(" Subarray : %d\n", super->current_vol); |
44470971 | 1802 | printf(" UUID : %s\n", uuid); |
dd8bcb3b AK |
1803 | printf(" RAID Level : %d", get_imsm_raid_level(map)); |
1804 | if (map2) | |
1805 | printf(" <-- %d", get_imsm_raid_level(map2)); | |
1806 | printf("\n"); | |
1807 | printf(" Members : %d", map->num_members); | |
1808 | if (map2) | |
1809 | printf(" <-- %d", map2->num_members); | |
1810 | printf("\n"); | |
0d80bb2f DW |
1811 | printf(" Slots : ["); |
1812 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 1813 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_0); |
0d80bb2f DW |
1814 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
1815 | } | |
dd8bcb3b AK |
1816 | printf("]"); |
1817 | if (map2) { | |
1818 | printf(" <-- ["); | |
1819 | for (i = 0; i < map2->num_members; i++) { | |
238c0a71 | 1820 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_1); |
dd8bcb3b AK |
1821 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
1822 | } | |
1823 | printf("]"); | |
1824 | } | |
1825 | printf("\n"); | |
7095bccb AK |
1826 | printf(" Failed disk : "); |
1827 | if (map->failed_disk_num == 0xff) | |
1828 | printf("none"); | |
1829 | else | |
1830 | printf("%i", map->failed_disk_num); | |
1831 | printf("\n"); | |
620b1713 DW |
1832 | slot = get_imsm_disk_slot(map, disk_idx); |
1833 | if (slot >= 0) { | |
238c0a71 | 1834 | ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); |
b10b37b8 DW |
1835 | printf(" This Slot : %d%s\n", slot, |
1836 | ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : ""); | |
1837 | } else | |
cdddbdbc | 1838 | printf(" This Slot : ?\n"); |
84918897 | 1839 | printf(" Sector Size : %u\n", super->sector_size); |
fcc2c9da | 1840 | sz = imsm_dev_size(dev); |
84918897 MK |
1841 | printf(" Array Size : %llu%s\n", |
1842 | (unsigned long long)sz * 512 / super->sector_size, | |
cdddbdbc | 1843 | human_size(sz * 512)); |
5551b113 | 1844 | sz = blocks_per_member(map); |
84918897 MK |
1845 | printf(" Per Dev Size : %llu%s\n", |
1846 | (unsigned long long)sz * 512 / super->sector_size, | |
cdddbdbc | 1847 | human_size(sz * 512)); |
5551b113 | 1848 | printf(" Sector Offset : %llu\n", |
7d8935cb | 1849 | pba_of_lba0(map) * 512 / super->sector_size); |
5551b113 CA |
1850 | printf(" Num Stripes : %llu\n", |
1851 | num_data_stripes(map)); | |
dd8bcb3b | 1852 | printf(" Chunk Size : %u KiB", |
cdddbdbc | 1853 | __le16_to_cpu(map->blocks_per_strip) / 2); |
dd8bcb3b AK |
1854 | if (map2) |
1855 | printf(" <-- %u KiB", | |
1856 | __le16_to_cpu(map2->blocks_per_strip) / 2); | |
1857 | printf("\n"); | |
cdddbdbc | 1858 | printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks)); |
8655a7b1 | 1859 | printf(" Migrate State : "); |
1484e727 DW |
1860 | if (dev->vol.migr_state) { |
1861 | if (migr_type(dev) == MIGR_INIT) | |
8655a7b1 | 1862 | printf("initialize\n"); |
1484e727 | 1863 | else if (migr_type(dev) == MIGR_REBUILD) |
8655a7b1 | 1864 | printf("rebuild\n"); |
1484e727 | 1865 | else if (migr_type(dev) == MIGR_VERIFY) |
8655a7b1 | 1866 | printf("check\n"); |
1484e727 | 1867 | else if (migr_type(dev) == MIGR_GEN_MIGR) |
8655a7b1 | 1868 | printf("general migration\n"); |
1484e727 | 1869 | else if (migr_type(dev) == MIGR_STATE_CHANGE) |
8655a7b1 | 1870 | printf("state change\n"); |
1484e727 | 1871 | else if (migr_type(dev) == MIGR_REPAIR) |
8655a7b1 | 1872 | printf("repair\n"); |
1484e727 | 1873 | else |
8655a7b1 DW |
1874 | printf("<unknown:%d>\n", migr_type(dev)); |
1875 | } else | |
1876 | printf("idle\n"); | |
3393c6af DW |
1877 | printf(" Map State : %s", map_state_str[map->map_state]); |
1878 | if (dev->vol.migr_state) { | |
238c0a71 | 1879 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 | 1880 | |
b10b37b8 | 1881 | printf(" <-- %s", map_state_str[map->map_state]); |
4036e7ee | 1882 | printf("\n Checkpoint : %llu ", vol_curr_migr_unit(dev)); |
089f9d79 | 1883 | if (is_gen_migration(dev) && (slot > 1 || slot < 0)) |
464d40e8 LD |
1884 | printf("(N/A)"); |
1885 | else | |
1886 | printf("(%llu)", (unsigned long long) | |
1887 | blocks_per_migr_unit(super, dev)); | |
3393c6af DW |
1888 | } |
1889 | printf("\n"); | |
2432ce9b AP |
1890 | printf(" Dirty State : %s\n", (dev->vol.dirty & RAIDVOL_DIRTY) ? |
1891 | "dirty" : "clean"); | |
1892 | printf(" RWH Policy : "); | |
c2462068 | 1893 | if (dev->rwh_policy == RWH_OFF || dev->rwh_policy == RWH_MULTIPLE_OFF) |
2432ce9b AP |
1894 | printf("off\n"); |
1895 | else if (dev->rwh_policy == RWH_DISTRIBUTED) | |
1896 | printf("PPL distributed\n"); | |
1897 | else if (dev->rwh_policy == RWH_JOURNALING_DRIVE) | |
1898 | printf("PPL journaling drive\n"); | |
c2462068 PB |
1899 | else if (dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED) |
1900 | printf("Multiple distributed PPLs\n"); | |
1901 | else if (dev->rwh_policy == RWH_MULTIPLE_PPLS_JOURNALING_DRIVE) | |
1902 | printf("Multiple PPLs on journaling drive\n"); | |
fbc42556 JR |
1903 | else if (dev->rwh_policy == RWH_BITMAP) |
1904 | printf("Write-intent bitmap\n"); | |
2432ce9b AP |
1905 | else |
1906 | printf("<unknown:%d>\n", dev->rwh_policy); | |
ba1b3bc8 AP |
1907 | |
1908 | printf(" Volume ID : %u\n", dev->my_vol_raid_dev_num); | |
cdddbdbc DW |
1909 | } |
1910 | ||
ef5c214e MK |
1911 | static void print_imsm_disk(struct imsm_disk *disk, |
1912 | int index, | |
1913 | __u32 reserved, | |
1914 | unsigned int sector_size) { | |
1f24f035 | 1915 | char str[MAX_RAID_SERIAL_LEN + 1]; |
cdddbdbc DW |
1916 | __u64 sz; |
1917 | ||
0ec1f4e8 | 1918 | if (index < -1 || !disk) |
e9d82038 DW |
1919 | return; |
1920 | ||
cdddbdbc | 1921 | printf("\n"); |
1f24f035 | 1922 | snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial); |
0ec1f4e8 DW |
1923 | if (index >= 0) |
1924 | printf(" Disk%02d Serial : %s\n", index, str); | |
1925 | else | |
1926 | printf(" Disk Serial : %s\n", str); | |
2432ce9b AP |
1927 | printf(" State :%s%s%s%s\n", is_spare(disk) ? " spare" : "", |
1928 | is_configured(disk) ? " active" : "", | |
1929 | is_failed(disk) ? " failed" : "", | |
1930 | is_journal(disk) ? " journal" : ""); | |
cdddbdbc | 1931 | printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id)); |
5551b113 | 1932 | sz = total_blocks(disk) - reserved; |
ef5c214e MK |
1933 | printf(" Usable Size : %llu%s\n", |
1934 | (unsigned long long)sz * 512 / sector_size, | |
cdddbdbc DW |
1935 | human_size(sz * 512)); |
1936 | } | |
1937 | ||
de44e46f PB |
1938 | void convert_to_4k_imsm_migr_rec(struct intel_super *super) |
1939 | { | |
1940 | struct migr_record *migr_rec = super->migr_rec; | |
1941 | ||
1942 | migr_rec->blocks_per_unit /= IMSM_4K_DIV; | |
de44e46f PB |
1943 | migr_rec->dest_depth_per_unit /= IMSM_4K_DIV; |
1944 | split_ull((join_u32(migr_rec->post_migr_vol_cap, | |
1945 | migr_rec->post_migr_vol_cap_hi) / IMSM_4K_DIV), | |
1946 | &migr_rec->post_migr_vol_cap, &migr_rec->post_migr_vol_cap_hi); | |
9f421827 PB |
1947 | set_migr_chkp_area_pba(migr_rec, |
1948 | migr_chkp_area_pba(migr_rec) / IMSM_4K_DIV); | |
1949 | set_migr_dest_1st_member_lba(migr_rec, | |
1950 | migr_dest_1st_member_lba(migr_rec) / IMSM_4K_DIV); | |
de44e46f PB |
1951 | } |
1952 | ||
f36a9ecd PB |
1953 | void convert_to_4k_imsm_disk(struct imsm_disk *disk) |
1954 | { | |
1955 | set_total_blocks(disk, (total_blocks(disk)/IMSM_4K_DIV)); | |
1956 | } | |
1957 | ||
1958 | void convert_to_4k(struct intel_super *super) | |
1959 | { | |
1960 | struct imsm_super *mpb = super->anchor; | |
1961 | struct imsm_disk *disk; | |
1962 | int i; | |
e4467bc7 | 1963 | __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size); |
f36a9ecd PB |
1964 | |
1965 | for (i = 0; i < mpb->num_disks ; i++) { | |
1966 | disk = __get_imsm_disk(mpb, i); | |
1967 | /* disk */ | |
1968 | convert_to_4k_imsm_disk(disk); | |
1969 | } | |
1970 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
1971 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
1972 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
1973 | /* dev */ | |
fcc2c9da | 1974 | set_imsm_dev_size(dev, imsm_dev_size(dev)/IMSM_4K_DIV); |
4036e7ee MT |
1975 | set_vol_curr_migr_unit(dev, |
1976 | vol_curr_migr_unit(dev) / IMSM_4K_DIV); | |
f36a9ecd PB |
1977 | |
1978 | /* map0 */ | |
1979 | set_blocks_per_member(map, blocks_per_member(map)/IMSM_4K_DIV); | |
1980 | map->blocks_per_strip /= IMSM_4K_DIV; | |
1981 | set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV); | |
1982 | ||
1983 | if (dev->vol.migr_state) { | |
1984 | /* map1 */ | |
1985 | map = get_imsm_map(dev, MAP_1); | |
1986 | set_blocks_per_member(map, | |
1987 | blocks_per_member(map)/IMSM_4K_DIV); | |
1988 | map->blocks_per_strip /= IMSM_4K_DIV; | |
1989 | set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV); | |
1990 | } | |
1991 | } | |
e4467bc7 TM |
1992 | if (bbm_log_size) { |
1993 | struct bbm_log *log = (void *)mpb + | |
1994 | __le32_to_cpu(mpb->mpb_size) - bbm_log_size; | |
1995 | __u32 i; | |
1996 | ||
1997 | for (i = 0; i < log->entry_count; i++) { | |
1998 | struct bbm_log_entry *entry = | |
1999 | &log->marked_block_entries[i]; | |
2000 | ||
2001 | __u8 count = entry->marked_count + 1; | |
2002 | unsigned long long sector = | |
2003 | __le48_to_cpu(&entry->defective_block_start); | |
2004 | ||
2005 | entry->defective_block_start = | |
2006 | __cpu_to_le48(sector/IMSM_4K_DIV); | |
2007 | entry->marked_count = max(count/IMSM_4K_DIV, 1) - 1; | |
2008 | } | |
2009 | } | |
f36a9ecd PB |
2010 | |
2011 | mpb->check_sum = __gen_imsm_checksum(mpb); | |
2012 | } | |
2013 | ||
520e69e2 AK |
2014 | void examine_migr_rec_imsm(struct intel_super *super) |
2015 | { | |
2016 | struct migr_record *migr_rec = super->migr_rec; | |
2017 | struct imsm_super *mpb = super->anchor; | |
2018 | int i; | |
2019 | ||
2020 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
2021 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
3136abe5 | 2022 | struct imsm_map *map; |
b4ab44d8 | 2023 | int slot = -1; |
3136abe5 | 2024 | |
195d1d76 | 2025 | if (is_gen_migration(dev) == false) |
520e69e2 AK |
2026 | continue; |
2027 | ||
2028 | printf("\nMigration Record Information:"); | |
3136abe5 | 2029 | |
44bfe6df AK |
2030 | /* first map under migration */ |
2031 | map = get_imsm_map(dev, MAP_0); | |
76c152ca | 2032 | |
3136abe5 AK |
2033 | if (map) |
2034 | slot = get_imsm_disk_slot(map, super->disks->index); | |
089f9d79 | 2035 | if (map == NULL || slot > 1 || slot < 0) { |
520e69e2 AK |
2036 | printf(" Empty\n "); |
2037 | printf("Examine one of first two disks in array\n"); | |
2038 | break; | |
2039 | } | |
2040 | printf("\n Status : "); | |
2041 | if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL) | |
2042 | printf("Normal\n"); | |
2043 | else | |
2044 | printf("Contains Data\n"); | |
9f421827 PB |
2045 | printf(" Current Unit : %llu\n", |
2046 | current_migr_unit(migr_rec)); | |
520e69e2 AK |
2047 | printf(" Family : %u\n", |
2048 | __le32_to_cpu(migr_rec->family_num)); | |
2049 | printf(" Ascending : %u\n", | |
2050 | __le32_to_cpu(migr_rec->ascending_migr)); | |
2051 | printf(" Blocks Per Unit : %u\n", | |
2052 | __le32_to_cpu(migr_rec->blocks_per_unit)); | |
2053 | printf(" Dest. Depth Per Unit : %u\n", | |
2054 | __le32_to_cpu(migr_rec->dest_depth_per_unit)); | |
9f421827 PB |
2055 | printf(" Checkpoint Area pba : %llu\n", |
2056 | migr_chkp_area_pba(migr_rec)); | |
2057 | printf(" First member lba : %llu\n", | |
2058 | migr_dest_1st_member_lba(migr_rec)); | |
2059 | printf(" Total Number of Units : %llu\n", | |
2060 | get_num_migr_units(migr_rec)); | |
2061 | printf(" Size of volume : %llu\n", | |
2062 | join_u32(migr_rec->post_migr_vol_cap, | |
2063 | migr_rec->post_migr_vol_cap_hi)); | |
520e69e2 AK |
2064 | printf(" Record was read from : %u\n", |
2065 | __le32_to_cpu(migr_rec->ckpt_read_disk_num)); | |
2066 | ||
2067 | break; | |
2068 | } | |
2069 | } | |
f36a9ecd | 2070 | |
de44e46f PB |
2071 | void convert_from_4k_imsm_migr_rec(struct intel_super *super) |
2072 | { | |
2073 | struct migr_record *migr_rec = super->migr_rec; | |
2074 | ||
2075 | migr_rec->blocks_per_unit *= IMSM_4K_DIV; | |
de44e46f PB |
2076 | migr_rec->dest_depth_per_unit *= IMSM_4K_DIV; |
2077 | split_ull((join_u32(migr_rec->post_migr_vol_cap, | |
2078 | migr_rec->post_migr_vol_cap_hi) * IMSM_4K_DIV), | |
2079 | &migr_rec->post_migr_vol_cap, | |
2080 | &migr_rec->post_migr_vol_cap_hi); | |
9f421827 PB |
2081 | set_migr_chkp_area_pba(migr_rec, |
2082 | migr_chkp_area_pba(migr_rec) * IMSM_4K_DIV); | |
2083 | set_migr_dest_1st_member_lba(migr_rec, | |
2084 | migr_dest_1st_member_lba(migr_rec) * IMSM_4K_DIV); | |
de44e46f PB |
2085 | } |
2086 | ||
f36a9ecd PB |
2087 | void convert_from_4k(struct intel_super *super) |
2088 | { | |
2089 | struct imsm_super *mpb = super->anchor; | |
2090 | struct imsm_disk *disk; | |
2091 | int i; | |
e4467bc7 | 2092 | __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size); |
f36a9ecd PB |
2093 | |
2094 | for (i = 0; i < mpb->num_disks ; i++) { | |
2095 | disk = __get_imsm_disk(mpb, i); | |
2096 | /* disk */ | |
2097 | set_total_blocks(disk, (total_blocks(disk)*IMSM_4K_DIV)); | |
2098 | } | |
2099 | ||
2100 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
2101 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
2102 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
2103 | /* dev */ | |
fcc2c9da | 2104 | set_imsm_dev_size(dev, imsm_dev_size(dev)*IMSM_4K_DIV); |
4036e7ee MT |
2105 | set_vol_curr_migr_unit(dev, |
2106 | vol_curr_migr_unit(dev) * IMSM_4K_DIV); | |
f36a9ecd PB |
2107 | |
2108 | /* map0 */ | |
2109 | set_blocks_per_member(map, blocks_per_member(map)*IMSM_4K_DIV); | |
2110 | map->blocks_per_strip *= IMSM_4K_DIV; | |
2111 | set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV); | |
2112 | ||
2113 | if (dev->vol.migr_state) { | |
2114 | /* map1 */ | |
2115 | map = get_imsm_map(dev, MAP_1); | |
2116 | set_blocks_per_member(map, | |
2117 | blocks_per_member(map)*IMSM_4K_DIV); | |
2118 | map->blocks_per_strip *= IMSM_4K_DIV; | |
2119 | set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV); | |
2120 | } | |
2121 | } | |
e4467bc7 TM |
2122 | if (bbm_log_size) { |
2123 | struct bbm_log *log = (void *)mpb + | |
2124 | __le32_to_cpu(mpb->mpb_size) - bbm_log_size; | |
2125 | __u32 i; | |
2126 | ||
2127 | for (i = 0; i < log->entry_count; i++) { | |
2128 | struct bbm_log_entry *entry = | |
2129 | &log->marked_block_entries[i]; | |
2130 | ||
2131 | __u8 count = entry->marked_count + 1; | |
2132 | unsigned long long sector = | |
2133 | __le48_to_cpu(&entry->defective_block_start); | |
2134 | ||
2135 | entry->defective_block_start = | |
2136 | __cpu_to_le48(sector*IMSM_4K_DIV); | |
2137 | entry->marked_count = count*IMSM_4K_DIV - 1; | |
2138 | } | |
2139 | } | |
f36a9ecd PB |
2140 | |
2141 | mpb->check_sum = __gen_imsm_checksum(mpb); | |
2142 | } | |
2143 | ||
19482bcc AK |
2144 | /******************************************************************************* |
2145 | * function: imsm_check_attributes | |
2146 | * Description: Function checks if features represented by attributes flags | |
1011e834 | 2147 | * are supported by mdadm. |
19482bcc AK |
2148 | * Parameters: |
2149 | * attributes - Attributes read from metadata | |
2150 | * Returns: | |
1011e834 N |
2151 | * 0 - passed attributes contains unsupported features flags |
2152 | * 1 - all features are supported | |
19482bcc AK |
2153 | ******************************************************************************/ |
2154 | static int imsm_check_attributes(__u32 attributes) | |
2155 | { | |
2156 | int ret_val = 1; | |
418f9b36 N |
2157 | __u32 not_supported = MPB_ATTRIB_SUPPORTED^0xffffffff; |
2158 | ||
2159 | not_supported &= ~MPB_ATTRIB_IGNORED; | |
19482bcc AK |
2160 | |
2161 | not_supported &= attributes; | |
2162 | if (not_supported) { | |
e7b84f9d | 2163 | pr_err("(IMSM): Unsupported attributes : %x\n", |
418f9b36 | 2164 | (unsigned)__le32_to_cpu(not_supported)); |
19482bcc AK |
2165 | if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) { |
2166 | dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY \n"); | |
2167 | not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY; | |
2168 | } | |
2169 | if (not_supported & MPB_ATTRIB_2TB) { | |
2170 | dprintf("\t\tMPB_ATTRIB_2TB\n"); | |
2171 | not_supported ^= MPB_ATTRIB_2TB; | |
2172 | } | |
2173 | if (not_supported & MPB_ATTRIB_RAID0) { | |
2174 | dprintf("\t\tMPB_ATTRIB_RAID0\n"); | |
2175 | not_supported ^= MPB_ATTRIB_RAID0; | |
2176 | } | |
2177 | if (not_supported & MPB_ATTRIB_RAID1) { | |
2178 | dprintf("\t\tMPB_ATTRIB_RAID1\n"); | |
2179 | not_supported ^= MPB_ATTRIB_RAID1; | |
2180 | } | |
2181 | if (not_supported & MPB_ATTRIB_RAID10) { | |
2182 | dprintf("\t\tMPB_ATTRIB_RAID10\n"); | |
2183 | not_supported ^= MPB_ATTRIB_RAID10; | |
2184 | } | |
2185 | if (not_supported & MPB_ATTRIB_RAID1E) { | |
2186 | dprintf("\t\tMPB_ATTRIB_RAID1E\n"); | |
2187 | not_supported ^= MPB_ATTRIB_RAID1E; | |
2188 | } | |
2189 | if (not_supported & MPB_ATTRIB_RAID5) { | |
2190 | dprintf("\t\tMPB_ATTRIB_RAID5\n"); | |
2191 | not_supported ^= MPB_ATTRIB_RAID5; | |
2192 | } | |
2193 | if (not_supported & MPB_ATTRIB_RAIDCNG) { | |
2194 | dprintf("\t\tMPB_ATTRIB_RAIDCNG\n"); | |
2195 | not_supported ^= MPB_ATTRIB_RAIDCNG; | |
2196 | } | |
2197 | if (not_supported & MPB_ATTRIB_BBM) { | |
2198 | dprintf("\t\tMPB_ATTRIB_BBM\n"); | |
2199 | not_supported ^= MPB_ATTRIB_BBM; | |
2200 | } | |
2201 | if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) { | |
2202 | dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY (== MPB_ATTRIB_LEGACY)\n"); | |
2203 | not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY; | |
2204 | } | |
2205 | if (not_supported & MPB_ATTRIB_EXP_STRIPE_SIZE) { | |
2206 | dprintf("\t\tMPB_ATTRIB_EXP_STRIP_SIZE\n"); | |
2207 | not_supported ^= MPB_ATTRIB_EXP_STRIPE_SIZE; | |
2208 | } | |
2209 | if (not_supported & MPB_ATTRIB_2TB_DISK) { | |
2210 | dprintf("\t\tMPB_ATTRIB_2TB_DISK\n"); | |
2211 | not_supported ^= MPB_ATTRIB_2TB_DISK; | |
2212 | } | |
2213 | if (not_supported & MPB_ATTRIB_NEVER_USE2) { | |
2214 | dprintf("\t\tMPB_ATTRIB_NEVER_USE2\n"); | |
2215 | not_supported ^= MPB_ATTRIB_NEVER_USE2; | |
2216 | } | |
2217 | if (not_supported & MPB_ATTRIB_NEVER_USE) { | |
2218 | dprintf("\t\tMPB_ATTRIB_NEVER_USE\n"); | |
2219 | not_supported ^= MPB_ATTRIB_NEVER_USE; | |
2220 | } | |
2221 | ||
2222 | if (not_supported) | |
1ade5cc1 | 2223 | dprintf("(IMSM): Unknown attributes : %x\n", not_supported); |
19482bcc AK |
2224 | |
2225 | ret_val = 0; | |
2226 | } | |
2227 | ||
2228 | return ret_val; | |
2229 | } | |
2230 | ||
a5d85af7 | 2231 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map); |
44470971 | 2232 | |
cdddbdbc DW |
2233 | static void examine_super_imsm(struct supertype *st, char *homehost) |
2234 | { | |
2235 | struct intel_super *super = st->sb; | |
949c47a0 | 2236 | struct imsm_super *mpb = super->anchor; |
cdddbdbc DW |
2237 | char str[MAX_SIGNATURE_LENGTH]; |
2238 | int i; | |
27fd6274 DW |
2239 | struct mdinfo info; |
2240 | char nbuf[64]; | |
cdddbdbc | 2241 | __u32 sum; |
14e8215b | 2242 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
94827db3 | 2243 | struct dl *dl; |
e48aed3c | 2244 | time_t creation_time; |
27fd6274 | 2245 | |
618f4e6d XN |
2246 | strncpy(str, (char *)mpb->sig, MPB_SIG_LEN); |
2247 | str[MPB_SIG_LEN-1] = '\0'; | |
cdddbdbc | 2248 | printf(" Magic : %s\n", str); |
cdddbdbc | 2249 | printf(" Version : %s\n", get_imsm_version(mpb)); |
148acb7b | 2250 | printf(" Orig Family : %08x\n", __le32_to_cpu(mpb->orig_family_num)); |
cdddbdbc DW |
2251 | printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num)); |
2252 | printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num)); | |
e48aed3c AP |
2253 | creation_time = __le64_to_cpu(mpb->creation_time); |
2254 | printf(" Creation Time : %.24s\n", | |
2255 | creation_time ? ctime(&creation_time) : "Unknown"); | |
19482bcc AK |
2256 | printf(" Attributes : "); |
2257 | if (imsm_check_attributes(mpb->attributes)) | |
2258 | printf("All supported\n"); | |
2259 | else | |
2260 | printf("not supported\n"); | |
a5d85af7 | 2261 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 2262 | fname_from_uuid(st, &info, nbuf, ':'); |
27fd6274 | 2263 | printf(" UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
2264 | sum = __le32_to_cpu(mpb->check_sum); |
2265 | printf(" Checksum : %08x %s\n", sum, | |
949c47a0 | 2266 | __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect"); |
f36a9ecd | 2267 | printf(" MPB Sectors : %d\n", mpb_sectors(mpb, super->sector_size)); |
cdddbdbc DW |
2268 | printf(" Disks : %d\n", mpb->num_disks); |
2269 | printf(" RAID Devices : %d\n", mpb->num_raid_devs); | |
ef5c214e MK |
2270 | print_imsm_disk(__get_imsm_disk(mpb, super->disks->index), |
2271 | super->disks->index, reserved, super->sector_size); | |
8d67477f | 2272 | if (get_imsm_bbm_log_size(super->bbm_log)) { |
604b746f JD |
2273 | struct bbm_log *log = super->bbm_log; |
2274 | ||
2275 | printf("\n"); | |
2276 | printf("Bad Block Management Log:\n"); | |
2277 | printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size)); | |
2278 | printf(" Signature : %x\n", __le32_to_cpu(log->signature)); | |
2279 | printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count)); | |
604b746f | 2280 | } |
44470971 DW |
2281 | for (i = 0; i < mpb->num_raid_devs; i++) { |
2282 | struct mdinfo info; | |
2283 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
2284 | ||
2285 | super->current_vol = i; | |
a5d85af7 | 2286 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 2287 | fname_from_uuid(st, &info, nbuf, ':'); |
c47b0ff6 | 2288 | print_imsm_dev(super, dev, nbuf + 5, super->disks->index); |
44470971 | 2289 | } |
cdddbdbc DW |
2290 | for (i = 0; i < mpb->num_disks; i++) { |
2291 | if (i == super->disks->index) | |
2292 | continue; | |
ef5c214e MK |
2293 | print_imsm_disk(__get_imsm_disk(mpb, i), i, reserved, |
2294 | super->sector_size); | |
cdddbdbc | 2295 | } |
94827db3 | 2296 | |
0ec1f4e8 DW |
2297 | for (dl = super->disks; dl; dl = dl->next) |
2298 | if (dl->index == -1) | |
ef5c214e MK |
2299 | print_imsm_disk(&dl->disk, -1, reserved, |
2300 | super->sector_size); | |
520e69e2 AK |
2301 | |
2302 | examine_migr_rec_imsm(super); | |
cdddbdbc DW |
2303 | } |
2304 | ||
061f2c6a | 2305 | static void brief_examine_super_imsm(struct supertype *st, int verbose) |
cdddbdbc | 2306 | { |
27fd6274 | 2307 | /* We just write a generic IMSM ARRAY entry */ |
ff54de6e N |
2308 | struct mdinfo info; |
2309 | char nbuf[64]; | |
2310 | ||
a5d85af7 | 2311 | getinfo_super_imsm(st, &info, NULL); |
4737ae25 N |
2312 | fname_from_uuid(st, &info, nbuf, ':'); |
2313 | printf("ARRAY metadata=imsm UUID=%s\n", nbuf + 5); | |
2314 | } | |
2315 | ||
2316 | static void brief_examine_subarrays_imsm(struct supertype *st, int verbose) | |
2317 | { | |
2318 | /* We just write a generic IMSM ARRAY entry */ | |
2319 | struct mdinfo info; | |
2320 | char nbuf[64]; | |
2321 | char nbuf1[64]; | |
2322 | struct intel_super *super = st->sb; | |
2323 | int i; | |
2324 | ||
2325 | if (!super->anchor->num_raid_devs) | |
2326 | return; | |
2327 | ||
a5d85af7 | 2328 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 2329 | fname_from_uuid(st, &info, nbuf, ':'); |
1e7bc0ed DW |
2330 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
2331 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
2332 | ||
2333 | super->current_vol = i; | |
a5d85af7 | 2334 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 2335 | fname_from_uuid(st, &info, nbuf1, ':'); |
b9ce7ab0 | 2336 | printf("ARRAY " DEV_MD_DIR "%.16s container=%s member=%d UUID=%s\n", |
cf8de691 | 2337 | dev->volume, nbuf + 5, i, nbuf1 + 5); |
1e7bc0ed | 2338 | } |
cdddbdbc DW |
2339 | } |
2340 | ||
9d84c8ea DW |
2341 | static void export_examine_super_imsm(struct supertype *st) |
2342 | { | |
2343 | struct intel_super *super = st->sb; | |
2344 | struct imsm_super *mpb = super->anchor; | |
2345 | struct mdinfo info; | |
2346 | char nbuf[64]; | |
2347 | ||
a5d85af7 | 2348 | getinfo_super_imsm(st, &info, NULL); |
9d84c8ea DW |
2349 | fname_from_uuid(st, &info, nbuf, ':'); |
2350 | printf("MD_METADATA=imsm\n"); | |
2351 | printf("MD_LEVEL=container\n"); | |
2352 | printf("MD_UUID=%s\n", nbuf+5); | |
2353 | printf("MD_DEVICES=%u\n", mpb->num_disks); | |
e48aed3c | 2354 | printf("MD_CREATION_TIME=%llu\n", __le64_to_cpu(mpb->creation_time)); |
9d84c8ea DW |
2355 | } |
2356 | ||
b771faef BK |
2357 | static void detail_super_imsm(struct supertype *st, char *homehost, |
2358 | char *subarray) | |
cdddbdbc | 2359 | { |
3ebe00a1 DW |
2360 | struct mdinfo info; |
2361 | char nbuf[64]; | |
b771faef BK |
2362 | struct intel_super *super = st->sb; |
2363 | int temp_vol = super->current_vol; | |
2364 | ||
2365 | if (subarray) | |
2366 | super->current_vol = strtoul(subarray, NULL, 10); | |
3ebe00a1 | 2367 | |
a5d85af7 | 2368 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 2369 | fname_from_uuid(st, &info, nbuf, ':'); |
65884368 | 2370 | printf("\n UUID : %s\n", nbuf + 5); |
b771faef BK |
2371 | |
2372 | super->current_vol = temp_vol; | |
cdddbdbc DW |
2373 | } |
2374 | ||
b771faef | 2375 | static void brief_detail_super_imsm(struct supertype *st, char *subarray) |
cdddbdbc | 2376 | { |
ff54de6e N |
2377 | struct mdinfo info; |
2378 | char nbuf[64]; | |
b771faef BK |
2379 | struct intel_super *super = st->sb; |
2380 | int temp_vol = super->current_vol; | |
2381 | ||
2382 | if (subarray) | |
2383 | super->current_vol = strtoul(subarray, NULL, 10); | |
2384 | ||
a5d85af7 | 2385 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 2386 | fname_from_uuid(st, &info, nbuf, ':'); |
ff54de6e | 2387 | printf(" UUID=%s", nbuf + 5); |
b771faef BK |
2388 | |
2389 | super->current_vol = temp_vol; | |
cdddbdbc | 2390 | } |
d665cc31 | 2391 | |
6da53c0e BK |
2392 | static int imsm_read_serial(int fd, char *devname, __u8 *serial, |
2393 | size_t serial_buf_len); | |
d665cc31 DW |
2394 | static void fd2devname(int fd, char *name); |
2395 | ||
120dc887 | 2396 | static int ahci_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose) |
d665cc31 | 2397 | { |
120dc887 LM |
2398 | /* dump an unsorted list of devices attached to AHCI Intel storage |
2399 | * controller, as well as non-connected ports | |
d665cc31 DW |
2400 | */ |
2401 | int hba_len = strlen(hba_path) + 1; | |
2402 | struct dirent *ent; | |
2403 | DIR *dir; | |
2404 | char *path = NULL; | |
2405 | int err = 0; | |
2406 | unsigned long port_mask = (1 << port_count) - 1; | |
2407 | ||
f21e18ca | 2408 | if (port_count > (int)sizeof(port_mask) * 8) { |
ba728be7 | 2409 | if (verbose > 0) |
e7b84f9d | 2410 | pr_err("port_count %d out of range\n", port_count); |
d665cc31 DW |
2411 | return 2; |
2412 | } | |
2413 | ||
2414 | /* scroll through /sys/dev/block looking for devices attached to | |
2415 | * this hba | |
2416 | */ | |
2417 | dir = opendir("/sys/dev/block"); | |
1a6dd6b9 PB |
2418 | if (!dir) |
2419 | return 1; | |
2420 | ||
2421 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
d665cc31 DW |
2422 | int fd; |
2423 | char model[64]; | |
2424 | char vendor[64]; | |
2425 | char buf[1024]; | |
2426 | int major, minor; | |
fcebeb77 | 2427 | char device[PATH_MAX]; |
d665cc31 DW |
2428 | char *c; |
2429 | int port; | |
2430 | int type; | |
2431 | ||
2432 | if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2) | |
2433 | continue; | |
7c798f87 | 2434 | path = devt_to_devpath(makedev(major, minor), 1, NULL); |
d665cc31 DW |
2435 | if (!path) |
2436 | continue; | |
2437 | if (!path_attached_to_hba(path, hba_path)) { | |
2438 | free(path); | |
2439 | path = NULL; | |
2440 | continue; | |
2441 | } | |
2442 | ||
fcebeb77 MT |
2443 | /* retrieve the scsi device */ |
2444 | if (!devt_to_devpath(makedev(major, minor), 1, device)) { | |
ba728be7 | 2445 | if (verbose > 0) |
fcebeb77 | 2446 | pr_err("failed to get device\n"); |
d665cc31 DW |
2447 | err = 2; |
2448 | break; | |
2449 | } | |
fcebeb77 | 2450 | if (devpath_to_char(device, "type", buf, sizeof(buf), 0)) { |
d665cc31 | 2451 | err = 2; |
d665cc31 DW |
2452 | break; |
2453 | } | |
2454 | type = strtoul(buf, NULL, 10); | |
2455 | ||
2456 | /* if it's not a disk print the vendor and model */ | |
2457 | if (!(type == 0 || type == 7 || type == 14)) { | |
2458 | vendor[0] = '\0'; | |
2459 | model[0] = '\0'; | |
fcebeb77 MT |
2460 | |
2461 | if (devpath_to_char(device, "vendor", buf, | |
2462 | sizeof(buf), 0) == 0) { | |
d665cc31 DW |
2463 | strncpy(vendor, buf, sizeof(vendor)); |
2464 | vendor[sizeof(vendor) - 1] = '\0'; | |
2465 | c = (char *) &vendor[sizeof(vendor) - 1]; | |
2466 | while (isspace(*c) || *c == '\0') | |
2467 | *c-- = '\0'; | |
2468 | ||
2469 | } | |
fcebeb77 MT |
2470 | |
2471 | if (devpath_to_char(device, "model", buf, | |
2472 | sizeof(buf), 0) == 0) { | |
d665cc31 DW |
2473 | strncpy(model, buf, sizeof(model)); |
2474 | model[sizeof(model) - 1] = '\0'; | |
2475 | c = (char *) &model[sizeof(model) - 1]; | |
2476 | while (isspace(*c) || *c == '\0') | |
2477 | *c-- = '\0'; | |
2478 | } | |
2479 | ||
2480 | if (vendor[0] && model[0]) | |
2481 | sprintf(buf, "%.64s %.64s", vendor, model); | |
2482 | else | |
2483 | switch (type) { /* numbers from hald/linux/device.c */ | |
2484 | case 1: sprintf(buf, "tape"); break; | |
2485 | case 2: sprintf(buf, "printer"); break; | |
2486 | case 3: sprintf(buf, "processor"); break; | |
2487 | case 4: | |
2488 | case 5: sprintf(buf, "cdrom"); break; | |
2489 | case 6: sprintf(buf, "scanner"); break; | |
2490 | case 8: sprintf(buf, "media_changer"); break; | |
2491 | case 9: sprintf(buf, "comm"); break; | |
2492 | case 12: sprintf(buf, "raid"); break; | |
2493 | default: sprintf(buf, "unknown"); | |
2494 | } | |
2495 | } else | |
2496 | buf[0] = '\0'; | |
d665cc31 DW |
2497 | |
2498 | /* chop device path to 'host%d' and calculate the port number */ | |
2499 | c = strchr(&path[hba_len], '/'); | |
4e5e717d | 2500 | if (!c) { |
ba728be7 | 2501 | if (verbose > 0) |
e7b84f9d | 2502 | pr_err("%s - invalid path name\n", path + hba_len); |
4e5e717d AW |
2503 | err = 2; |
2504 | break; | |
2505 | } | |
d665cc31 | 2506 | *c = '\0'; |
0858eccf AP |
2507 | if ((sscanf(&path[hba_len], "ata%d", &port) == 1) || |
2508 | ((sscanf(&path[hba_len], "host%d", &port) == 1))) | |
d665cc31 DW |
2509 | port -= host_base; |
2510 | else { | |
ba728be7 | 2511 | if (verbose > 0) { |
d665cc31 | 2512 | *c = '/'; /* repair the full string */ |
e7b84f9d | 2513 | pr_err("failed to determine port number for %s\n", |
d665cc31 DW |
2514 | path); |
2515 | } | |
2516 | err = 2; | |
2517 | break; | |
2518 | } | |
2519 | ||
2520 | /* mark this port as used */ | |
2521 | port_mask &= ~(1 << port); | |
2522 | ||
2523 | /* print out the device information */ | |
2524 | if (buf[0]) { | |
2525 | printf(" Port%d : - non-disk device (%s) -\n", port, buf); | |
2526 | continue; | |
2527 | } | |
2528 | ||
2529 | fd = dev_open(ent->d_name, O_RDONLY); | |
4389ce73 | 2530 | if (!is_fd_valid(fd)) |
d665cc31 DW |
2531 | printf(" Port%d : - disk info unavailable -\n", port); |
2532 | else { | |
2533 | fd2devname(fd, buf); | |
2534 | printf(" Port%d : %s", port, buf); | |
6da53c0e BK |
2535 | if (imsm_read_serial(fd, NULL, (__u8 *)buf, |
2536 | sizeof(buf)) == 0) | |
2537 | printf(" (%s)\n", buf); | |
d665cc31 | 2538 | else |
664d5325 | 2539 | printf(" ()\n"); |
4dab422a | 2540 | close(fd); |
d665cc31 | 2541 | } |
d665cc31 DW |
2542 | free(path); |
2543 | path = NULL; | |
2544 | } | |
2545 | if (path) | |
2546 | free(path); | |
2547 | if (dir) | |
2548 | closedir(dir); | |
2549 | if (err == 0) { | |
2550 | int i; | |
2551 | ||
2552 | for (i = 0; i < port_count; i++) | |
2553 | if (port_mask & (1 << i)) | |
2554 | printf(" Port%d : - no device attached -\n", i); | |
2555 | } | |
2556 | ||
2557 | return err; | |
2558 | } | |
2559 | ||
6da53c0e | 2560 | static int print_nvme_info(struct sys_dev *hba) |
60f0f54d PB |
2561 | { |
2562 | struct dirent *ent; | |
2563 | DIR *dir; | |
60f0f54d | 2564 | |
6da53c0e | 2565 | dir = opendir("/sys/block/"); |
b9135011 | 2566 | if (!dir) |
b5eece69 | 2567 | return 1; |
b9135011 JS |
2568 | |
2569 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
8662f92d MT |
2570 | char ns_path[PATH_MAX]; |
2571 | char cntrl_path[PATH_MAX]; | |
2572 | char buf[PATH_MAX]; | |
2573 | int fd = -1; | |
60f0f54d | 2574 | |
8662f92d MT |
2575 | if (!strstr(ent->d_name, "nvme")) |
2576 | goto skip; | |
d835518b | 2577 | |
8662f92d | 2578 | fd = open_dev(ent->d_name); |
4389ce73 | 2579 | if (!is_fd_valid(fd)) |
8662f92d | 2580 | goto skip; |
d835518b | 2581 | |
8662f92d MT |
2582 | if (!diskfd_to_devpath(fd, 0, ns_path) || |
2583 | !diskfd_to_devpath(fd, 1, cntrl_path)) | |
2584 | goto skip; | |
2585 | ||
2586 | if (!path_attached_to_hba(cntrl_path, hba->path)) | |
2587 | goto skip; | |
2588 | ||
2589 | if (!imsm_is_nvme_namespace_supported(fd, 0)) | |
2590 | goto skip; | |
2591 | ||
2592 | fd2devname(fd, buf); | |
2593 | if (hba->type == SYS_DEV_VMD) | |
2594 | printf(" NVMe under VMD : %s", buf); | |
2595 | else if (hba->type == SYS_DEV_NVME) | |
2596 | printf(" NVMe Device : %s", buf); | |
2597 | ||
2598 | if (!imsm_read_serial(fd, NULL, (__u8 *)buf, | |
2599 | sizeof(buf))) | |
2600 | printf(" (%s)\n", buf); | |
2601 | else | |
2602 | printf("()\n"); | |
2603 | ||
2604 | skip: | |
4389ce73 | 2605 | close_fd(&fd); |
60f0f54d PB |
2606 | } |
2607 | ||
b9135011 | 2608 | closedir(dir); |
b5eece69 | 2609 | return 0; |
60f0f54d PB |
2610 | } |
2611 | ||
120dc887 LM |
2612 | static void print_found_intel_controllers(struct sys_dev *elem) |
2613 | { | |
2614 | for (; elem; elem = elem->next) { | |
e7b84f9d | 2615 | pr_err("found Intel(R) "); |
120dc887 LM |
2616 | if (elem->type == SYS_DEV_SATA) |
2617 | fprintf(stderr, "SATA "); | |
155cbb4c LM |
2618 | else if (elem->type == SYS_DEV_SAS) |
2619 | fprintf(stderr, "SAS "); | |
0858eccf AP |
2620 | else if (elem->type == SYS_DEV_NVME) |
2621 | fprintf(stderr, "NVMe "); | |
60f0f54d PB |
2622 | |
2623 | if (elem->type == SYS_DEV_VMD) | |
2624 | fprintf(stderr, "VMD domain"); | |
75350d87 KF |
2625 | else if (elem->type == SYS_DEV_SATA_VMD) |
2626 | fprintf(stderr, "SATA VMD domain"); | |
60f0f54d PB |
2627 | else |
2628 | fprintf(stderr, "RAID controller"); | |
2629 | ||
120dc887 LM |
2630 | if (elem->pci_id) |
2631 | fprintf(stderr, " at %s", elem->pci_id); | |
2632 | fprintf(stderr, ".\n"); | |
2633 | } | |
2634 | fflush(stderr); | |
2635 | } | |
2636 | ||
120dc887 LM |
2637 | static int ahci_get_port_count(const char *hba_path, int *port_count) |
2638 | { | |
2639 | struct dirent *ent; | |
2640 | DIR *dir; | |
2641 | int host_base = -1; | |
2642 | ||
2643 | *port_count = 0; | |
2644 | if ((dir = opendir(hba_path)) == NULL) | |
2645 | return -1; | |
2646 | ||
2647 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
2648 | int host; | |
2649 | ||
0858eccf AP |
2650 | if ((sscanf(ent->d_name, "ata%d", &host) != 1) && |
2651 | ((sscanf(ent->d_name, "host%d", &host) != 1))) | |
120dc887 LM |
2652 | continue; |
2653 | if (*port_count == 0) | |
2654 | host_base = host; | |
2655 | else if (host < host_base) | |
2656 | host_base = host; | |
2657 | ||
2658 | if (host + 1 > *port_count + host_base) | |
2659 | *port_count = host + 1 - host_base; | |
2660 | } | |
2661 | closedir(dir); | |
2662 | return host_base; | |
2663 | } | |
2664 | ||
a891a3c2 LM |
2665 | static void print_imsm_capability(const struct imsm_orom *orom) |
2666 | { | |
0858eccf AP |
2667 | printf(" Platform : Intel(R) "); |
2668 | if (orom->capabilities == 0 && orom->driver_features == 0) | |
2669 | printf("Matrix Storage Manager\n"); | |
ab0c6bb9 AP |
2670 | else if (imsm_orom_is_enterprise(orom) && orom->major_ver >= 6) |
2671 | printf("Virtual RAID on CPU\n"); | |
0858eccf AP |
2672 | else |
2673 | printf("Rapid Storage Technology%s\n", | |
2674 | imsm_orom_is_enterprise(orom) ? " enterprise" : ""); | |
8d1114be MG |
2675 | if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build) { |
2676 | if (imsm_orom_is_vmd_without_efi(orom)) | |
2677 | printf(" Version : %d.%d\n", orom->major_ver, | |
2678 | orom->minor_ver); | |
2679 | else | |
2680 | printf(" Version : %d.%d.%d.%d\n", orom->major_ver, | |
2681 | orom->minor_ver, orom->hotfix_ver, orom->build); | |
2682 | } | |
a891a3c2 LM |
2683 | printf(" RAID Levels :%s%s%s%s%s\n", |
2684 | imsm_orom_has_raid0(orom) ? " raid0" : "", | |
2685 | imsm_orom_has_raid1(orom) ? " raid1" : "", | |
2686 | imsm_orom_has_raid1e(orom) ? " raid1e" : "", | |
2687 | imsm_orom_has_raid10(orom) ? " raid10" : "", | |
2688 | imsm_orom_has_raid5(orom) ? " raid5" : ""); | |
2689 | printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n", | |
2690 | imsm_orom_has_chunk(orom, 2) ? " 2k" : "", | |
2691 | imsm_orom_has_chunk(orom, 4) ? " 4k" : "", | |
2692 | imsm_orom_has_chunk(orom, 8) ? " 8k" : "", | |
2693 | imsm_orom_has_chunk(orom, 16) ? " 16k" : "", | |
2694 | imsm_orom_has_chunk(orom, 32) ? " 32k" : "", | |
2695 | imsm_orom_has_chunk(orom, 64) ? " 64k" : "", | |
2696 | imsm_orom_has_chunk(orom, 128) ? " 128k" : "", | |
2697 | imsm_orom_has_chunk(orom, 256) ? " 256k" : "", | |
2698 | imsm_orom_has_chunk(orom, 512) ? " 512k" : "", | |
2699 | imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "", | |
2700 | imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "", | |
2701 | imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "", | |
2702 | imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "", | |
2703 | imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "", | |
2704 | imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "", | |
2705 | imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : ""); | |
29cd0821 CA |
2706 | printf(" 2TB volumes :%s supported\n", |
2707 | (orom->attr & IMSM_OROM_ATTR_2TB)?"":" not"); | |
2708 | printf(" 2TB disks :%s supported\n", | |
2709 | (orom->attr & IMSM_OROM_ATTR_2TB_DISK)?"":" not"); | |
0e7f69a8 | 2710 | printf(" Max Disks : %d\n", orom->tds); |
0858eccf AP |
2711 | printf(" Max Volumes : %d per array, %d per %s\n", |
2712 | orom->vpa, orom->vphba, | |
2713 | imsm_orom_is_nvme(orom) ? "platform" : "controller"); | |
a891a3c2 LM |
2714 | return; |
2715 | } | |
2716 | ||
e50cf220 MN |
2717 | static void print_imsm_capability_export(const struct imsm_orom *orom) |
2718 | { | |
2719 | printf("MD_FIRMWARE_TYPE=imsm\n"); | |
0858eccf AP |
2720 | if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build) |
2721 | printf("IMSM_VERSION=%d.%d.%d.%d\n", orom->major_ver, orom->minor_ver, | |
2722 | orom->hotfix_ver, orom->build); | |
e50cf220 MN |
2723 | printf("IMSM_SUPPORTED_RAID_LEVELS=%s%s%s%s%s\n", |
2724 | imsm_orom_has_raid0(orom) ? "raid0 " : "", | |
2725 | imsm_orom_has_raid1(orom) ? "raid1 " : "", | |
2726 | imsm_orom_has_raid1e(orom) ? "raid1e " : "", | |
2727 | imsm_orom_has_raid5(orom) ? "raid10 " : "", | |
2728 | imsm_orom_has_raid10(orom) ? "raid5 " : ""); | |
2729 | printf("IMSM_SUPPORTED_CHUNK_SIZES=%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n", | |
2730 | imsm_orom_has_chunk(orom, 2) ? "2k " : "", | |
2731 | imsm_orom_has_chunk(orom, 4) ? "4k " : "", | |
2732 | imsm_orom_has_chunk(orom, 8) ? "8k " : "", | |
2733 | imsm_orom_has_chunk(orom, 16) ? "16k " : "", | |
2734 | imsm_orom_has_chunk(orom, 32) ? "32k " : "", | |
2735 | imsm_orom_has_chunk(orom, 64) ? "64k " : "", | |
2736 | imsm_orom_has_chunk(orom, 128) ? "128k " : "", | |
2737 | imsm_orom_has_chunk(orom, 256) ? "256k " : "", | |
2738 | imsm_orom_has_chunk(orom, 512) ? "512k " : "", | |
2739 | imsm_orom_has_chunk(orom, 1024*1) ? "1M " : "", | |
2740 | imsm_orom_has_chunk(orom, 1024*2) ? "2M " : "", | |
2741 | imsm_orom_has_chunk(orom, 1024*4) ? "4M " : "", | |
2742 | imsm_orom_has_chunk(orom, 1024*8) ? "8M " : "", | |
2743 | imsm_orom_has_chunk(orom, 1024*16) ? "16M " : "", | |
2744 | imsm_orom_has_chunk(orom, 1024*32) ? "32M " : "", | |
2745 | imsm_orom_has_chunk(orom, 1024*64) ? "64M " : ""); | |
2746 | printf("IMSM_2TB_VOLUMES=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB) ? "yes" : "no"); | |
2747 | printf("IMSM_2TB_DISKS=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB_DISK) ? "yes" : "no"); | |
2748 | printf("IMSM_MAX_DISKS=%d\n",orom->tds); | |
2749 | printf("IMSM_MAX_VOLUMES_PER_ARRAY=%d\n",orom->vpa); | |
2750 | printf("IMSM_MAX_VOLUMES_PER_CONTROLLER=%d\n",orom->vphba); | |
2751 | } | |
2752 | ||
9eafa1de | 2753 | static int detail_platform_imsm(int verbose, int enumerate_only, char *controller_path) |
d665cc31 DW |
2754 | { |
2755 | /* There are two components to imsm platform support, the ahci SATA | |
2756 | * controller and the option-rom. To find the SATA controller we | |
2757 | * simply look in /sys/bus/pci/drivers/ahci to see if an ahci | |
2758 | * controller with the Intel vendor id is present. This approach | |
2759 | * allows mdadm to leverage the kernel's ahci detection logic, with the | |
2760 | * caveat that if ahci.ko is not loaded mdadm will not be able to | |
2761 | * detect platform raid capabilities. The option-rom resides in a | |
2762 | * platform "Adapter ROM". We scan for its signature to retrieve the | |
2763 | * platform capabilities. If raid support is disabled in the BIOS the | |
2764 | * option-rom capability structure will not be available. | |
2765 | */ | |
d665cc31 | 2766 | struct sys_dev *list, *hba; |
d665cc31 DW |
2767 | int host_base = 0; |
2768 | int port_count = 0; | |
9eafa1de | 2769 | int result=1; |
d665cc31 | 2770 | |
5615172f | 2771 | if (enumerate_only) { |
420dafcd | 2772 | if (check_no_platform()) |
5615172f | 2773 | return 0; |
a891a3c2 LM |
2774 | list = find_intel_devices(); |
2775 | if (!list) | |
2776 | return 2; | |
2777 | for (hba = list; hba; hba = hba->next) { | |
6b781d33 AP |
2778 | if (find_imsm_capability(hba)) { |
2779 | result = 0; | |
a891a3c2 LM |
2780 | break; |
2781 | } | |
9eafa1de | 2782 | else |
6b781d33 | 2783 | result = 2; |
a891a3c2 | 2784 | } |
a891a3c2 | 2785 | return result; |
5615172f DW |
2786 | } |
2787 | ||
155cbb4c LM |
2788 | list = find_intel_devices(); |
2789 | if (!list) { | |
ba728be7 | 2790 | if (verbose > 0) |
7a862a02 | 2791 | pr_err("no active Intel(R) RAID controller found.\n"); |
d665cc31 | 2792 | return 2; |
ba728be7 | 2793 | } else if (verbose > 0) |
155cbb4c | 2794 | print_found_intel_controllers(list); |
d665cc31 | 2795 | |
a891a3c2 | 2796 | for (hba = list; hba; hba = hba->next) { |
0858eccf | 2797 | if (controller_path && (compare_paths(hba->path, controller_path) != 0)) |
9eafa1de | 2798 | continue; |
0858eccf | 2799 | if (!find_imsm_capability(hba)) { |
60f0f54d | 2800 | char buf[PATH_MAX]; |
e7b84f9d | 2801 | pr_err("imsm capabilities not found for controller: %s (type %s)\n", |
75350d87 KF |
2802 | hba->type == SYS_DEV_VMD || hba->type == SYS_DEV_SATA_VMD ? |
2803 | vmd_domain_to_controller(hba, buf) : | |
2804 | hba->path, get_sys_dev_type(hba->type)); | |
0858eccf AP |
2805 | continue; |
2806 | } | |
2807 | result = 0; | |
2808 | } | |
2809 | ||
2810 | if (controller_path && result == 1) { | |
2811 | pr_err("no active Intel(R) RAID controller found under %s\n", | |
2812 | controller_path); | |
2813 | return result; | |
2814 | } | |
2815 | ||
5e1d6128 | 2816 | const struct orom_entry *entry; |
0858eccf | 2817 | |
5e1d6128 | 2818 | for (entry = orom_entries; entry; entry = entry->next) { |
60f0f54d | 2819 | if (entry->type == SYS_DEV_VMD) { |
07cb1e57 | 2820 | print_imsm_capability(&entry->orom); |
32716c51 PB |
2821 | printf(" 3rd party NVMe :%s supported\n", |
2822 | imsm_orom_has_tpv_support(&entry->orom)?"":" not"); | |
60f0f54d PB |
2823 | for (hba = list; hba; hba = hba->next) { |
2824 | if (hba->type == SYS_DEV_VMD) { | |
2825 | char buf[PATH_MAX]; | |
60f0f54d PB |
2826 | printf(" I/O Controller : %s (%s)\n", |
2827 | vmd_domain_to_controller(hba, buf), get_sys_dev_type(hba->type)); | |
6da53c0e | 2828 | if (print_nvme_info(hba)) { |
b5eece69 PB |
2829 | if (verbose > 0) |
2830 | pr_err("failed to get devices attached to VMD domain.\n"); | |
2831 | result |= 2; | |
2832 | } | |
60f0f54d PB |
2833 | } |
2834 | } | |
07cb1e57 | 2835 | printf("\n"); |
60f0f54d PB |
2836 | continue; |
2837 | } | |
0858eccf | 2838 | |
60f0f54d PB |
2839 | print_imsm_capability(&entry->orom); |
2840 | if (entry->type == SYS_DEV_NVME) { | |
0858eccf AP |
2841 | for (hba = list; hba; hba = hba->next) { |
2842 | if (hba->type == SYS_DEV_NVME) | |
6da53c0e | 2843 | print_nvme_info(hba); |
0858eccf | 2844 | } |
60f0f54d | 2845 | printf("\n"); |
0858eccf AP |
2846 | continue; |
2847 | } | |
2848 | ||
2849 | struct devid_list *devid; | |
5e1d6128 | 2850 | for (devid = entry->devid_list; devid; devid = devid->next) { |
0858eccf AP |
2851 | hba = device_by_id(devid->devid); |
2852 | if (!hba) | |
2853 | continue; | |
2854 | ||
9eafa1de MN |
2855 | printf(" I/O Controller : %s (%s)\n", |
2856 | hba->path, get_sys_dev_type(hba->type)); | |
75350d87 | 2857 | if (hba->type == SYS_DEV_SATA || hba->type == SYS_DEV_SATA_VMD) { |
9eafa1de MN |
2858 | host_base = ahci_get_port_count(hba->path, &port_count); |
2859 | if (ahci_enumerate_ports(hba->path, port_count, host_base, verbose)) { | |
2860 | if (verbose > 0) | |
75350d87 KF |
2861 | pr_err("failed to enumerate ports on %s controller at %s.\n", |
2862 | get_sys_dev_type(hba->type), hba->pci_id); | |
9eafa1de MN |
2863 | result |= 2; |
2864 | } | |
120dc887 LM |
2865 | } |
2866 | } | |
0858eccf | 2867 | printf("\n"); |
d665cc31 | 2868 | } |
155cbb4c | 2869 | |
120dc887 | 2870 | return result; |
d665cc31 | 2871 | } |
e50cf220 | 2872 | |
9eafa1de | 2873 | static int export_detail_platform_imsm(int verbose, char *controller_path) |
e50cf220 | 2874 | { |
e50cf220 MN |
2875 | struct sys_dev *list, *hba; |
2876 | int result=1; | |
2877 | ||
2878 | list = find_intel_devices(); | |
2879 | if (!list) { | |
2880 | if (verbose > 0) | |
2881 | pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_INTEL_DEVICES\n"); | |
2882 | result = 2; | |
e50cf220 MN |
2883 | return result; |
2884 | } | |
2885 | ||
2886 | for (hba = list; hba; hba = hba->next) { | |
9eafa1de MN |
2887 | if (controller_path && (compare_paths(hba->path,controller_path) != 0)) |
2888 | continue; | |
60f0f54d PB |
2889 | if (!find_imsm_capability(hba) && verbose > 0) { |
2890 | char buf[PATH_MAX]; | |
2891 | pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_IMSM_CAPABLE_DEVICE_UNDER_%s\n", | |
75350d87 KF |
2892 | hba->type == SYS_DEV_VMD || hba->type == SYS_DEV_SATA_VMD ? |
2893 | vmd_domain_to_controller(hba, buf) : hba->path); | |
60f0f54d | 2894 | } |
0858eccf | 2895 | else |
e50cf220 | 2896 | result = 0; |
e50cf220 MN |
2897 | } |
2898 | ||
5e1d6128 | 2899 | const struct orom_entry *entry; |
0858eccf | 2900 | |
60f0f54d | 2901 | for (entry = orom_entries; entry; entry = entry->next) { |
75350d87 | 2902 | if (entry->type == SYS_DEV_VMD || entry->type == SYS_DEV_SATA_VMD) { |
60f0f54d PB |
2903 | for (hba = list; hba; hba = hba->next) |
2904 | print_imsm_capability_export(&entry->orom); | |
2905 | continue; | |
2906 | } | |
5e1d6128 | 2907 | print_imsm_capability_export(&entry->orom); |
60f0f54d | 2908 | } |
0858eccf | 2909 | |
e50cf220 MN |
2910 | return result; |
2911 | } | |
2912 | ||
cdddbdbc DW |
2913 | static int match_home_imsm(struct supertype *st, char *homehost) |
2914 | { | |
5115ca67 DW |
2915 | /* the imsm metadata format does not specify any host |
2916 | * identification information. We return -1 since we can never | |
2917 | * confirm nor deny whether a given array is "meant" for this | |
148acb7b | 2918 | * host. We rely on compare_super and the 'family_num' fields to |
5115ca67 DW |
2919 | * exclude member disks that do not belong, and we rely on |
2920 | * mdadm.conf to specify the arrays that should be assembled. | |
2921 | * Auto-assembly may still pick up "foreign" arrays. | |
2922 | */ | |
cdddbdbc | 2923 | |
9362c1c8 | 2924 | return -1; |
cdddbdbc DW |
2925 | } |
2926 | ||
2927 | static void uuid_from_super_imsm(struct supertype *st, int uuid[4]) | |
2928 | { | |
51006d85 N |
2929 | /* The uuid returned here is used for: |
2930 | * uuid to put into bitmap file (Create, Grow) | |
2931 | * uuid for backup header when saving critical section (Grow) | |
2932 | * comparing uuids when re-adding a device into an array | |
2933 | * In these cases the uuid required is that of the data-array, | |
2934 | * not the device-set. | |
2935 | * uuid to recognise same set when adding a missing device back | |
2936 | * to an array. This is a uuid for the device-set. | |
1011e834 | 2937 | * |
51006d85 N |
2938 | * For each of these we can make do with a truncated |
2939 | * or hashed uuid rather than the original, as long as | |
2940 | * everyone agrees. | |
2941 | * In each case the uuid required is that of the data-array, | |
2942 | * not the device-set. | |
43dad3d6 | 2943 | */ |
51006d85 N |
2944 | /* imsm does not track uuid's so we synthesis one using sha1 on |
2945 | * - The signature (Which is constant for all imsm array, but no matter) | |
148acb7b | 2946 | * - the orig_family_num of the container |
51006d85 N |
2947 | * - the index number of the volume |
2948 | * - the 'serial' number of the volume. | |
2949 | * Hopefully these are all constant. | |
2950 | */ | |
2951 | struct intel_super *super = st->sb; | |
43dad3d6 | 2952 | |
51006d85 N |
2953 | char buf[20]; |
2954 | struct sha1_ctx ctx; | |
2955 | struct imsm_dev *dev = NULL; | |
148acb7b | 2956 | __u32 family_num; |
51006d85 | 2957 | |
148acb7b DW |
2958 | /* some mdadm versions failed to set ->orig_family_num, in which |
2959 | * case fall back to ->family_num. orig_family_num will be | |
2960 | * fixed up with the first metadata update. | |
2961 | */ | |
2962 | family_num = super->anchor->orig_family_num; | |
2963 | if (family_num == 0) | |
2964 | family_num = super->anchor->family_num; | |
51006d85 | 2965 | sha1_init_ctx(&ctx); |
92bd8f8d | 2966 | sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx); |
148acb7b | 2967 | sha1_process_bytes(&family_num, sizeof(__u32), &ctx); |
51006d85 N |
2968 | if (super->current_vol >= 0) |
2969 | dev = get_imsm_dev(super, super->current_vol); | |
2970 | if (dev) { | |
2971 | __u32 vol = super->current_vol; | |
2972 | sha1_process_bytes(&vol, sizeof(vol), &ctx); | |
2973 | sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx); | |
2974 | } | |
2975 | sha1_finish_ctx(&ctx, buf); | |
2976 | memcpy(uuid, buf, 4*4); | |
cdddbdbc DW |
2977 | } |
2978 | ||
0d481d37 | 2979 | #if 0 |
4f5bc454 DW |
2980 | static void |
2981 | get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p) | |
cdddbdbc | 2982 | { |
cdddbdbc DW |
2983 | __u8 *v = get_imsm_version(mpb); |
2984 | __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH; | |
2985 | char major[] = { 0, 0, 0 }; | |
2986 | char minor[] = { 0 ,0, 0 }; | |
2987 | char patch[] = { 0, 0, 0 }; | |
2988 | char *ver_parse[] = { major, minor, patch }; | |
2989 | int i, j; | |
2990 | ||
2991 | i = j = 0; | |
2992 | while (*v != '\0' && v < end) { | |
2993 | if (*v != '.' && j < 2) | |
2994 | ver_parse[i][j++] = *v; | |
2995 | else { | |
2996 | i++; | |
2997 | j = 0; | |
2998 | } | |
2999 | v++; | |
3000 | } | |
3001 | ||
4f5bc454 DW |
3002 | *m = strtol(minor, NULL, 0); |
3003 | *p = strtol(patch, NULL, 0); | |
3004 | } | |
0d481d37 | 3005 | #endif |
4f5bc454 | 3006 | |
1e5c6983 DW |
3007 | static __u32 migr_strip_blocks_resync(struct imsm_dev *dev) |
3008 | { | |
3009 | /* migr_strip_size when repairing or initializing parity */ | |
238c0a71 | 3010 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
3011 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
3012 | ||
3013 | switch (get_imsm_raid_level(map)) { | |
3014 | case 5: | |
3015 | case 10: | |
3016 | return chunk; | |
3017 | default: | |
3018 | return 128*1024 >> 9; | |
3019 | } | |
3020 | } | |
3021 | ||
3022 | static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev) | |
3023 | { | |
3024 | /* migr_strip_size when rebuilding a degraded disk, no idea why | |
3025 | * this is different than migr_strip_size_resync(), but it's good | |
3026 | * to be compatible | |
3027 | */ | |
238c0a71 | 3028 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
3029 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
3030 | ||
3031 | switch (get_imsm_raid_level(map)) { | |
3032 | case 1: | |
3033 | case 10: | |
3034 | if (map->num_members % map->num_domains == 0) | |
3035 | return 128*1024 >> 9; | |
3036 | else | |
3037 | return chunk; | |
3038 | case 5: | |
3039 | return max((__u32) 64*1024 >> 9, chunk); | |
3040 | default: | |
3041 | return 128*1024 >> 9; | |
3042 | } | |
3043 | } | |
3044 | ||
3045 | static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev) | |
3046 | { | |
238c0a71 AK |
3047 | struct imsm_map *lo = get_imsm_map(dev, MAP_0); |
3048 | struct imsm_map *hi = get_imsm_map(dev, MAP_1); | |
1e5c6983 DW |
3049 | __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip); |
3050 | __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip); | |
3051 | ||
3052 | return max((__u32) 1, hi_chunk / lo_chunk); | |
3053 | } | |
3054 | ||
3055 | static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev) | |
3056 | { | |
238c0a71 | 3057 | struct imsm_map *lo = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
3058 | int level = get_imsm_raid_level(lo); |
3059 | ||
3060 | if (level == 1 || level == 10) { | |
238c0a71 | 3061 | struct imsm_map *hi = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
3062 | |
3063 | return hi->num_domains; | |
3064 | } else | |
3065 | return num_stripes_per_unit_resync(dev); | |
3066 | } | |
3067 | ||
44490938 MD |
3068 | static unsigned long long calc_component_size(struct imsm_map *map, |
3069 | struct imsm_dev *dev) | |
3070 | { | |
3071 | unsigned long long component_size; | |
3072 | unsigned long long dev_size = imsm_dev_size(dev); | |
a4f7290c | 3073 | long long calc_dev_size = 0; |
44490938 MD |
3074 | unsigned int member_disks = imsm_num_data_members(map); |
3075 | ||
3076 | if (member_disks == 0) | |
3077 | return 0; | |
3078 | ||
3079 | component_size = per_dev_array_size(map); | |
3080 | calc_dev_size = component_size * member_disks; | |
3081 | ||
3082 | /* Component size is rounded to 1MB so difference between size from | |
3083 | * metadata and size calculated from num_data_stripes equals up to | |
3084 | * 2048 blocks per each device. If the difference is higher it means | |
3085 | * that array size was expanded and num_data_stripes was not updated. | |
3086 | */ | |
a4f7290c | 3087 | if (llabs(calc_dev_size - (long long)dev_size) > |
44490938 MD |
3088 | (1 << SECT_PER_MB_SHIFT) * member_disks) { |
3089 | component_size = dev_size / member_disks; | |
3090 | dprintf("Invalid num_data_stripes in metadata; expected=%llu, found=%llu\n", | |
3091 | component_size / map->blocks_per_strip, | |
3092 | num_data_stripes(map)); | |
3093 | } | |
3094 | ||
3095 | return component_size; | |
3096 | } | |
3097 | ||
1e5c6983 DW |
3098 | static __u32 parity_segment_depth(struct imsm_dev *dev) |
3099 | { | |
238c0a71 | 3100 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
3101 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
3102 | ||
3103 | switch(get_imsm_raid_level(map)) { | |
3104 | case 1: | |
3105 | case 10: | |
3106 | return chunk * map->num_domains; | |
3107 | case 5: | |
3108 | return chunk * map->num_members; | |
3109 | default: | |
3110 | return chunk; | |
3111 | } | |
3112 | } | |
3113 | ||
3114 | static __u32 map_migr_block(struct imsm_dev *dev, __u32 block) | |
3115 | { | |
238c0a71 | 3116 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
3117 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
3118 | __u32 strip = block / chunk; | |
3119 | ||
3120 | switch (get_imsm_raid_level(map)) { | |
3121 | case 1: | |
3122 | case 10: { | |
3123 | __u32 vol_strip = (strip * map->num_domains) + 1; | |
3124 | __u32 vol_stripe = vol_strip / map->num_members; | |
3125 | ||
3126 | return vol_stripe * chunk + block % chunk; | |
3127 | } case 5: { | |
3128 | __u32 stripe = strip / (map->num_members - 1); | |
3129 | ||
3130 | return stripe * chunk + block % chunk; | |
3131 | } | |
3132 | default: | |
3133 | return 0; | |
3134 | } | |
3135 | } | |
3136 | ||
c47b0ff6 AK |
3137 | static __u64 blocks_per_migr_unit(struct intel_super *super, |
3138 | struct imsm_dev *dev) | |
1e5c6983 DW |
3139 | { |
3140 | /* calculate the conversion factor between per member 'blocks' | |
3141 | * (md/{resync,rebuild}_start) and imsm migration units, return | |
3142 | * 0 for the 'not migrating' and 'unsupported migration' cases | |
3143 | */ | |
3144 | if (!dev->vol.migr_state) | |
3145 | return 0; | |
3146 | ||
3147 | switch (migr_type(dev)) { | |
c47b0ff6 AK |
3148 | case MIGR_GEN_MIGR: { |
3149 | struct migr_record *migr_rec = super->migr_rec; | |
3150 | return __le32_to_cpu(migr_rec->blocks_per_unit); | |
3151 | } | |
1e5c6983 DW |
3152 | case MIGR_VERIFY: |
3153 | case MIGR_REPAIR: | |
3154 | case MIGR_INIT: { | |
238c0a71 | 3155 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
3156 | __u32 stripes_per_unit; |
3157 | __u32 blocks_per_unit; | |
3158 | __u32 parity_depth; | |
3159 | __u32 migr_chunk; | |
3160 | __u32 block_map; | |
3161 | __u32 block_rel; | |
3162 | __u32 segment; | |
3163 | __u32 stripe; | |
3164 | __u8 disks; | |
3165 | ||
3166 | /* yes, this is really the translation of migr_units to | |
3167 | * per-member blocks in the 'resync' case | |
3168 | */ | |
3169 | stripes_per_unit = num_stripes_per_unit_resync(dev); | |
3170 | migr_chunk = migr_strip_blocks_resync(dev); | |
9529d343 | 3171 | disks = imsm_num_data_members(map); |
1e5c6983 | 3172 | blocks_per_unit = stripes_per_unit * migr_chunk * disks; |
7b1ab482 | 3173 | stripe = __le16_to_cpu(map->blocks_per_strip) * disks; |
1e5c6983 DW |
3174 | segment = blocks_per_unit / stripe; |
3175 | block_rel = blocks_per_unit - segment * stripe; | |
3176 | parity_depth = parity_segment_depth(dev); | |
3177 | block_map = map_migr_block(dev, block_rel); | |
3178 | return block_map + parity_depth * segment; | |
3179 | } | |
3180 | case MIGR_REBUILD: { | |
3181 | __u32 stripes_per_unit; | |
3182 | __u32 migr_chunk; | |
3183 | ||
3184 | stripes_per_unit = num_stripes_per_unit_rebuild(dev); | |
3185 | migr_chunk = migr_strip_blocks_rebuild(dev); | |
3186 | return migr_chunk * stripes_per_unit; | |
3187 | } | |
1e5c6983 DW |
3188 | case MIGR_STATE_CHANGE: |
3189 | default: | |
3190 | return 0; | |
3191 | } | |
3192 | } | |
3193 | ||
c2c087e6 DW |
3194 | static int imsm_level_to_layout(int level) |
3195 | { | |
3196 | switch (level) { | |
3197 | case 0: | |
3198 | case 1: | |
3199 | return 0; | |
3200 | case 5: | |
3201 | case 6: | |
a380c027 | 3202 | return ALGORITHM_LEFT_ASYMMETRIC; |
c2c087e6 | 3203 | case 10: |
c92a2527 | 3204 | return 0x102; |
c2c087e6 | 3205 | } |
a18a888e | 3206 | return UnSet; |
c2c087e6 DW |
3207 | } |
3208 | ||
8e59f3d8 AK |
3209 | /******************************************************************************* |
3210 | * Function: read_imsm_migr_rec | |
3211 | * Description: Function reads imsm migration record from last sector of disk | |
3212 | * Parameters: | |
3213 | * fd : disk descriptor | |
3214 | * super : metadata info | |
3215 | * Returns: | |
3216 | * 0 : success, | |
3217 | * -1 : fail | |
3218 | ******************************************************************************/ | |
3219 | static int read_imsm_migr_rec(int fd, struct intel_super *super) | |
3220 | { | |
3221 | int ret_val = -1; | |
de44e46f | 3222 | unsigned int sector_size = super->sector_size; |
8e59f3d8 AK |
3223 | unsigned long long dsize; |
3224 | ||
3225 | get_dev_size(fd, NULL, &dsize); | |
de44e46f PB |
3226 | if (lseek64(fd, dsize - (sector_size*MIGR_REC_SECTOR_POSITION), |
3227 | SEEK_SET) < 0) { | |
e7b84f9d N |
3228 | pr_err("Cannot seek to anchor block: %s\n", |
3229 | strerror(errno)); | |
8e59f3d8 AK |
3230 | goto out; |
3231 | } | |
466070ad | 3232 | if ((unsigned int)read(fd, super->migr_rec_buf, |
de44e46f PB |
3233 | MIGR_REC_BUF_SECTORS*sector_size) != |
3234 | MIGR_REC_BUF_SECTORS*sector_size) { | |
e7b84f9d N |
3235 | pr_err("Cannot read migr record block: %s\n", |
3236 | strerror(errno)); | |
8e59f3d8 AK |
3237 | goto out; |
3238 | } | |
3239 | ret_val = 0; | |
de44e46f PB |
3240 | if (sector_size == 4096) |
3241 | convert_from_4k_imsm_migr_rec(super); | |
8e59f3d8 AK |
3242 | |
3243 | out: | |
3244 | return ret_val; | |
3245 | } | |
3246 | ||
3136abe5 AK |
3247 | static struct imsm_dev *imsm_get_device_during_migration( |
3248 | struct intel_super *super) | |
3249 | { | |
3250 | ||
3251 | struct intel_dev *dv; | |
3252 | ||
3253 | for (dv = super->devlist; dv; dv = dv->next) { | |
3254 | if (is_gen_migration(dv->dev)) | |
3255 | return dv->dev; | |
3256 | } | |
3257 | return NULL; | |
3258 | } | |
3259 | ||
8e59f3d8 AK |
3260 | /******************************************************************************* |
3261 | * Function: load_imsm_migr_rec | |
3262 | * Description: Function reads imsm migration record (it is stored at the last | |
3263 | * sector of disk) | |
3264 | * Parameters: | |
3265 | * super : imsm internal array info | |
8e59f3d8 AK |
3266 | * Returns: |
3267 | * 0 : success | |
3268 | * -1 : fail | |
4c965cc9 | 3269 | * -2 : no migration in progress |
8e59f3d8 | 3270 | ******************************************************************************/ |
2f86fda3 | 3271 | static int load_imsm_migr_rec(struct intel_super *super) |
8e59f3d8 | 3272 | { |
594dc1b8 | 3273 | struct dl *dl; |
8e59f3d8 AK |
3274 | char nm[30]; |
3275 | int retval = -1; | |
3276 | int fd = -1; | |
3136abe5 | 3277 | struct imsm_dev *dev; |
594dc1b8 | 3278 | struct imsm_map *map; |
b4ab44d8 | 3279 | int slot = -1; |
2f86fda3 | 3280 | int keep_fd = 1; |
3136abe5 AK |
3281 | |
3282 | /* find map under migration */ | |
3283 | dev = imsm_get_device_during_migration(super); | |
3284 | /* nothing to load,no migration in progress? | |
3285 | */ | |
3286 | if (dev == NULL) | |
4c965cc9 | 3287 | return -2; |
8e59f3d8 | 3288 | |
2f86fda3 MT |
3289 | map = get_imsm_map(dev, MAP_0); |
3290 | if (!map) | |
3291 | return -1; | |
3136abe5 | 3292 | |
2f86fda3 MT |
3293 | for (dl = super->disks; dl; dl = dl->next) { |
3294 | /* skip spare and failed disks | |
3295 | */ | |
3296 | if (dl->index < 0) | |
3297 | continue; | |
3298 | /* read only from one of the first two slots | |
3299 | */ | |
3300 | slot = get_imsm_disk_slot(map, dl->index); | |
3301 | if (slot > 1 || slot < 0) | |
3302 | continue; | |
3303 | ||
4389ce73 | 3304 | if (!is_fd_valid(dl->fd)) { |
8e59f3d8 AK |
3305 | sprintf(nm, "%d:%d", dl->major, dl->minor); |
3306 | fd = dev_open(nm, O_RDONLY); | |
4389ce73 MT |
3307 | |
3308 | if (is_fd_valid(fd)) { | |
2f86fda3 | 3309 | keep_fd = 0; |
8e59f3d8 | 3310 | break; |
2f86fda3 MT |
3311 | } |
3312 | } else { | |
3313 | fd = dl->fd; | |
3314 | break; | |
8e59f3d8 AK |
3315 | } |
3316 | } | |
2f86fda3 | 3317 | |
4389ce73 | 3318 | if (!is_fd_valid(fd)) |
2f86fda3 | 3319 | return retval; |
8e59f3d8 | 3320 | retval = read_imsm_migr_rec(fd, super); |
2f86fda3 | 3321 | if (!keep_fd) |
8e59f3d8 | 3322 | close(fd); |
2f86fda3 | 3323 | |
8e59f3d8 AK |
3324 | return retval; |
3325 | } | |
3326 | ||
c17608ea AK |
3327 | /******************************************************************************* |
3328 | * function: imsm_create_metadata_checkpoint_update | |
3329 | * Description: It creates update for checkpoint change. | |
3330 | * Parameters: | |
3331 | * super : imsm internal array info | |
3332 | * u : pointer to prepared update | |
3333 | * Returns: | |
3334 | * Uptate length. | |
3335 | * If length is equal to 0, input pointer u contains no update | |
3336 | ******************************************************************************/ | |
3337 | static int imsm_create_metadata_checkpoint_update( | |
3338 | struct intel_super *super, | |
3339 | struct imsm_update_general_migration_checkpoint **u) | |
3340 | { | |
3341 | ||
3342 | int update_memory_size = 0; | |
3343 | ||
1ade5cc1 | 3344 | dprintf("(enter)\n"); |
c17608ea AK |
3345 | |
3346 | if (u == NULL) | |
3347 | return 0; | |
3348 | *u = NULL; | |
3349 | ||
3350 | /* size of all update data without anchor */ | |
3351 | update_memory_size = | |
3352 | sizeof(struct imsm_update_general_migration_checkpoint); | |
3353 | ||
503975b9 | 3354 | *u = xcalloc(1, update_memory_size); |
c17608ea | 3355 | if (*u == NULL) { |
1ade5cc1 | 3356 | dprintf("error: cannot get memory\n"); |
c17608ea AK |
3357 | return 0; |
3358 | } | |
3359 | (*u)->type = update_general_migration_checkpoint; | |
9f421827 | 3360 | (*u)->curr_migr_unit = current_migr_unit(super->migr_rec); |
83b3de77 | 3361 | dprintf("prepared for %llu\n", (unsigned long long)(*u)->curr_migr_unit); |
c17608ea AK |
3362 | |
3363 | return update_memory_size; | |
3364 | } | |
3365 | ||
c17608ea AK |
3366 | static void imsm_update_metadata_locally(struct supertype *st, |
3367 | void *buf, int len); | |
3368 | ||
687629c2 AK |
3369 | /******************************************************************************* |
3370 | * Function: write_imsm_migr_rec | |
3371 | * Description: Function writes imsm migration record | |
3372 | * (at the last sector of disk) | |
3373 | * Parameters: | |
3374 | * super : imsm internal array info | |
3375 | * Returns: | |
3376 | * 0 : success | |
3377 | * -1 : if fail | |
3378 | ******************************************************************************/ | |
3379 | static int write_imsm_migr_rec(struct supertype *st) | |
3380 | { | |
3381 | struct intel_super *super = st->sb; | |
de44e46f | 3382 | unsigned int sector_size = super->sector_size; |
687629c2 | 3383 | unsigned long long dsize; |
687629c2 AK |
3384 | int retval = -1; |
3385 | struct dl *sd; | |
c17608ea AK |
3386 | int len; |
3387 | struct imsm_update_general_migration_checkpoint *u; | |
3136abe5 | 3388 | struct imsm_dev *dev; |
594dc1b8 | 3389 | struct imsm_map *map; |
3136abe5 AK |
3390 | |
3391 | /* find map under migration */ | |
3392 | dev = imsm_get_device_during_migration(super); | |
3393 | /* if no migration, write buffer anyway to clear migr_record | |
3394 | * on disk based on first available device | |
3395 | */ | |
3396 | if (dev == NULL) | |
3397 | dev = get_imsm_dev(super, super->current_vol < 0 ? 0 : | |
3398 | super->current_vol); | |
3399 | ||
44bfe6df | 3400 | map = get_imsm_map(dev, MAP_0); |
687629c2 | 3401 | |
de44e46f PB |
3402 | if (sector_size == 4096) |
3403 | convert_to_4k_imsm_migr_rec(super); | |
687629c2 | 3404 | for (sd = super->disks ; sd ; sd = sd->next) { |
b4ab44d8 | 3405 | int slot = -1; |
3136abe5 AK |
3406 | |
3407 | /* skip failed and spare devices */ | |
3408 | if (sd->index < 0) | |
3409 | continue; | |
687629c2 | 3410 | /* write to 2 first slots only */ |
3136abe5 AK |
3411 | if (map) |
3412 | slot = get_imsm_disk_slot(map, sd->index); | |
089f9d79 | 3413 | if (map == NULL || slot > 1 || slot < 0) |
687629c2 | 3414 | continue; |
3136abe5 | 3415 | |
2f86fda3 MT |
3416 | get_dev_size(sd->fd, NULL, &dsize); |
3417 | if (lseek64(sd->fd, dsize - (MIGR_REC_SECTOR_POSITION * | |
3418 | sector_size), | |
de44e46f | 3419 | SEEK_SET) < 0) { |
e7b84f9d N |
3420 | pr_err("Cannot seek to anchor block: %s\n", |
3421 | strerror(errno)); | |
687629c2 AK |
3422 | goto out; |
3423 | } | |
2f86fda3 | 3424 | if ((unsigned int)write(sd->fd, super->migr_rec_buf, |
de44e46f PB |
3425 | MIGR_REC_BUF_SECTORS*sector_size) != |
3426 | MIGR_REC_BUF_SECTORS*sector_size) { | |
e7b84f9d N |
3427 | pr_err("Cannot write migr record block: %s\n", |
3428 | strerror(errno)); | |
687629c2 AK |
3429 | goto out; |
3430 | } | |
687629c2 | 3431 | } |
de44e46f PB |
3432 | if (sector_size == 4096) |
3433 | convert_from_4k_imsm_migr_rec(super); | |
c17608ea AK |
3434 | /* update checkpoint information in metadata */ |
3435 | len = imsm_create_metadata_checkpoint_update(super, &u); | |
c17608ea AK |
3436 | if (len <= 0) { |
3437 | dprintf("imsm: Cannot prepare update\n"); | |
3438 | goto out; | |
3439 | } | |
3440 | /* update metadata locally */ | |
3441 | imsm_update_metadata_locally(st, u, len); | |
3442 | /* and possibly remotely */ | |
3443 | if (st->update_tail) { | |
3444 | append_metadata_update(st, u, len); | |
3445 | /* during reshape we do all work inside metadata handler | |
3446 | * manage_reshape(), so metadata update has to be triggered | |
3447 | * insida it | |
3448 | */ | |
3449 | flush_metadata_updates(st); | |
3450 | st->update_tail = &st->updates; | |
3451 | } else | |
3452 | free(u); | |
687629c2 AK |
3453 | |
3454 | retval = 0; | |
3455 | out: | |
687629c2 AK |
3456 | return retval; |
3457 | } | |
3458 | ||
e2962bfc AK |
3459 | /* spare/missing disks activations are not allowe when |
3460 | * array/container performs reshape operation, because | |
3461 | * all arrays in container works on the same disks set | |
3462 | */ | |
3463 | int imsm_reshape_blocks_arrays_changes(struct intel_super *super) | |
3464 | { | |
3465 | int rv = 0; | |
3466 | struct intel_dev *i_dev; | |
3467 | struct imsm_dev *dev; | |
3468 | ||
3469 | /* check whole container | |
3470 | */ | |
3471 | for (i_dev = super->devlist; i_dev; i_dev = i_dev->next) { | |
3472 | dev = i_dev->dev; | |
3ad25638 | 3473 | if (is_gen_migration(dev)) { |
e2962bfc AK |
3474 | /* No repair during any migration in container |
3475 | */ | |
3476 | rv = 1; | |
3477 | break; | |
3478 | } | |
3479 | } | |
3480 | return rv; | |
3481 | } | |
3e684231 | 3482 | static unsigned long long imsm_component_size_alignment_check(int level, |
c41e00b2 | 3483 | int chunk_size, |
f36a9ecd | 3484 | unsigned int sector_size, |
c41e00b2 AK |
3485 | unsigned long long component_size) |
3486 | { | |
3e684231 | 3487 | unsigned int component_size_alignment; |
c41e00b2 | 3488 | |
3e684231 | 3489 | /* check component size alignment |
c41e00b2 | 3490 | */ |
3e684231 | 3491 | component_size_alignment = component_size % (chunk_size/sector_size); |
c41e00b2 | 3492 | |
3e684231 | 3493 | dprintf("(Level: %i, chunk_size = %i, component_size = %llu), component_size_alignment = %u\n", |
c41e00b2 | 3494 | level, chunk_size, component_size, |
3e684231 | 3495 | component_size_alignment); |
c41e00b2 | 3496 | |
3e684231 MZ |
3497 | if (component_size_alignment && (level != 1) && (level != UnSet)) { |
3498 | dprintf("imsm: reported component size aligned from %llu ", | |
c41e00b2 | 3499 | component_size); |
3e684231 | 3500 | component_size -= component_size_alignment; |
1ade5cc1 | 3501 | dprintf_cont("to %llu (%i).\n", |
3e684231 | 3502 | component_size, component_size_alignment); |
c41e00b2 AK |
3503 | } |
3504 | ||
3505 | return component_size; | |
3506 | } | |
e2962bfc | 3507 | |
fbc42556 JR |
3508 | /******************************************************************************* |
3509 | * Function: get_bitmap_header_sector | |
3510 | * Description: Returns the sector where the bitmap header is placed. | |
3511 | * Parameters: | |
3512 | * st : supertype information | |
3513 | * dev_idx : index of the device with bitmap | |
3514 | * | |
3515 | * Returns: | |
3516 | * The sector where the bitmap header is placed | |
3517 | ******************************************************************************/ | |
3518 | static unsigned long long get_bitmap_header_sector(struct intel_super *super, | |
3519 | int dev_idx) | |
3520 | { | |
3521 | struct imsm_dev *dev = get_imsm_dev(super, dev_idx); | |
3522 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
3523 | ||
3524 | if (!super->sector_size) { | |
3525 | dprintf("sector size is not set\n"); | |
3526 | return 0; | |
3527 | } | |
3528 | ||
3529 | return pba_of_lba0(map) + calc_component_size(map, dev) + | |
3530 | (IMSM_BITMAP_HEADER_OFFSET / super->sector_size); | |
3531 | } | |
3532 | ||
3533 | /******************************************************************************* | |
3534 | * Function: get_bitmap_sector | |
3535 | * Description: Returns the sector where the bitmap is placed. | |
3536 | * Parameters: | |
3537 | * st : supertype information | |
3538 | * dev_idx : index of the device with bitmap | |
3539 | * | |
3540 | * Returns: | |
3541 | * The sector where the bitmap is placed | |
3542 | ******************************************************************************/ | |
3543 | static unsigned long long get_bitmap_sector(struct intel_super *super, | |
3544 | int dev_idx) | |
3545 | { | |
3546 | if (!super->sector_size) { | |
3547 | dprintf("sector size is not set\n"); | |
3548 | return 0; | |
3549 | } | |
3550 | ||
3551 | return get_bitmap_header_sector(super, dev_idx) + | |
3552 | (IMSM_BITMAP_HEADER_SIZE / super->sector_size); | |
3553 | } | |
3554 | ||
2432ce9b AP |
3555 | static unsigned long long get_ppl_sector(struct intel_super *super, int dev_idx) |
3556 | { | |
3557 | struct imsm_dev *dev = get_imsm_dev(super, dev_idx); | |
3558 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
3559 | ||
3560 | return pba_of_lba0(map) + | |
3561 | (num_data_stripes(map) * map->blocks_per_strip); | |
3562 | } | |
3563 | ||
a5d85af7 | 3564 | static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap) |
bf5a934a DW |
3565 | { |
3566 | struct intel_super *super = st->sb; | |
c47b0ff6 | 3567 | struct migr_record *migr_rec = super->migr_rec; |
949c47a0 | 3568 | struct imsm_dev *dev = get_imsm_dev(super, super->current_vol); |
238c0a71 AK |
3569 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
3570 | struct imsm_map *prev_map = get_imsm_map(dev, MAP_1); | |
b335e593 | 3571 | struct imsm_map *map_to_analyse = map; |
efb30e7f | 3572 | struct dl *dl; |
a5d85af7 | 3573 | int map_disks = info->array.raid_disks; |
bf5a934a | 3574 | |
95eeceeb | 3575 | memset(info, 0, sizeof(*info)); |
b335e593 AK |
3576 | if (prev_map) |
3577 | map_to_analyse = prev_map; | |
3578 | ||
ca0748fa | 3579 | dl = super->current_disk; |
9894ec0d | 3580 | |
bf5a934a | 3581 | info->container_member = super->current_vol; |
cd0430a1 | 3582 | info->array.raid_disks = map->num_members; |
b335e593 | 3583 | info->array.level = get_imsm_raid_level(map_to_analyse); |
bf5a934a DW |
3584 | info->array.layout = imsm_level_to_layout(info->array.level); |
3585 | info->array.md_minor = -1; | |
3586 | info->array.ctime = 0; | |
3587 | info->array.utime = 0; | |
b335e593 AK |
3588 | info->array.chunk_size = |
3589 | __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9; | |
2432ce9b | 3590 | info->array.state = !(dev->vol.dirty & RAIDVOL_DIRTY); |
fcc2c9da | 3591 | info->custom_array_size = imsm_dev_size(dev); |
3ad25638 AK |
3592 | info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb); |
3593 | ||
3f510843 | 3594 | if (is_gen_migration(dev)) { |
195d1d76 PP |
3595 | /* |
3596 | * device prev_map should be added if it is in the middle | |
3597 | * of migration | |
3598 | */ | |
3599 | assert(prev_map); | |
3600 | ||
3f83228a | 3601 | info->reshape_active = 1; |
b335e593 AK |
3602 | info->new_level = get_imsm_raid_level(map); |
3603 | info->new_layout = imsm_level_to_layout(info->new_level); | |
3604 | info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9; | |
3f83228a | 3605 | info->delta_disks = map->num_members - prev_map->num_members; |
493f5dd6 N |
3606 | if (info->delta_disks) { |
3607 | /* this needs to be applied to every array | |
3608 | * in the container. | |
3609 | */ | |
81219e70 | 3610 | info->reshape_active = CONTAINER_RESHAPE; |
493f5dd6 | 3611 | } |
3f83228a N |
3612 | /* We shape information that we give to md might have to be |
3613 | * modify to cope with md's requirement for reshaping arrays. | |
3614 | * For example, when reshaping a RAID0, md requires it to be | |
3615 | * presented as a degraded RAID4. | |
3616 | * Also if a RAID0 is migrating to a RAID5 we need to specify | |
3617 | * the array as already being RAID5, but the 'before' layout | |
3618 | * is a RAID4-like layout. | |
3619 | */ | |
3620 | switch (info->array.level) { | |
3621 | case 0: | |
3622 | switch(info->new_level) { | |
3623 | case 0: | |
3624 | /* conversion is happening as RAID4 */ | |
3625 | info->array.level = 4; | |
3626 | info->array.raid_disks += 1; | |
3627 | break; | |
3628 | case 5: | |
3629 | /* conversion is happening as RAID5 */ | |
3630 | info->array.level = 5; | |
3631 | info->array.layout = ALGORITHM_PARITY_N; | |
3f83228a N |
3632 | info->delta_disks -= 1; |
3633 | break; | |
3634 | default: | |
3635 | /* FIXME error message */ | |
3636 | info->array.level = UnSet; | |
3637 | break; | |
3638 | } | |
3639 | break; | |
3640 | } | |
b335e593 AK |
3641 | } else { |
3642 | info->new_level = UnSet; | |
3643 | info->new_layout = UnSet; | |
3644 | info->new_chunk = info->array.chunk_size; | |
3f83228a | 3645 | info->delta_disks = 0; |
b335e593 | 3646 | } |
ca0748fa | 3647 | |
efb30e7f DW |
3648 | if (dl) { |
3649 | info->disk.major = dl->major; | |
3650 | info->disk.minor = dl->minor; | |
ca0748fa | 3651 | info->disk.number = dl->index; |
656b6b5a N |
3652 | info->disk.raid_disk = get_imsm_disk_slot(map_to_analyse, |
3653 | dl->index); | |
efb30e7f | 3654 | } |
bf5a934a | 3655 | |
5551b113 | 3656 | info->data_offset = pba_of_lba0(map_to_analyse); |
44490938 | 3657 | info->component_size = calc_component_size(map, dev); |
3e684231 | 3658 | info->component_size = imsm_component_size_alignment_check( |
c41e00b2 AK |
3659 | info->array.level, |
3660 | info->array.chunk_size, | |
f36a9ecd | 3661 | super->sector_size, |
c41e00b2 | 3662 | info->component_size); |
5e46202e | 3663 | info->bb.supported = 1; |
139dae11 | 3664 | |
301406c9 | 3665 | memset(info->uuid, 0, sizeof(info->uuid)); |
921d9e16 | 3666 | info->recovery_start = MaxSector; |
bf5a934a | 3667 | |
c2462068 PB |
3668 | if (info->array.level == 5 && |
3669 | (dev->rwh_policy == RWH_DISTRIBUTED || | |
3670 | dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)) { | |
2432ce9b AP |
3671 | info->consistency_policy = CONSISTENCY_POLICY_PPL; |
3672 | info->ppl_sector = get_ppl_sector(super, super->current_vol); | |
c2462068 PB |
3673 | if (dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED) |
3674 | info->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9; | |
3675 | else | |
3676 | info->ppl_size = (PPL_HEADER_SIZE + PPL_ENTRY_SPACE) | |
3677 | >> 9; | |
2432ce9b AP |
3678 | } else if (info->array.level <= 0) { |
3679 | info->consistency_policy = CONSISTENCY_POLICY_NONE; | |
3680 | } else { | |
fbc42556 JR |
3681 | if (dev->rwh_policy == RWH_BITMAP) { |
3682 | info->bitmap_offset = get_bitmap_sector(super, super->current_vol); | |
3683 | info->consistency_policy = CONSISTENCY_POLICY_BITMAP; | |
3684 | } else { | |
3685 | info->consistency_policy = CONSISTENCY_POLICY_RESYNC; | |
3686 | } | |
2432ce9b AP |
3687 | } |
3688 | ||
d2e6d5d6 | 3689 | info->reshape_progress = 0; |
b6796ce1 | 3690 | info->resync_start = MaxSector; |
b9172665 | 3691 | if ((map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED || |
2432ce9b | 3692 | !(info->array.state & 1)) && |
b9172665 | 3693 | imsm_reshape_blocks_arrays_changes(super) == 0) { |
301406c9 | 3694 | info->resync_start = 0; |
b6796ce1 AK |
3695 | } |
3696 | if (dev->vol.migr_state) { | |
1e5c6983 DW |
3697 | switch (migr_type(dev)) { |
3698 | case MIGR_REPAIR: | |
3699 | case MIGR_INIT: { | |
c47b0ff6 AK |
3700 | __u64 blocks_per_unit = blocks_per_migr_unit(super, |
3701 | dev); | |
4036e7ee | 3702 | __u64 units = vol_curr_migr_unit(dev); |
1e5c6983 DW |
3703 | |
3704 | info->resync_start = blocks_per_unit * units; | |
3705 | break; | |
3706 | } | |
d2e6d5d6 | 3707 | case MIGR_GEN_MIGR: { |
c47b0ff6 AK |
3708 | __u64 blocks_per_unit = blocks_per_migr_unit(super, |
3709 | dev); | |
9f421827 | 3710 | __u64 units = current_migr_unit(migr_rec); |
04fa9523 | 3711 | int used_disks; |
d2e6d5d6 | 3712 | |
befb629b AK |
3713 | if (__le32_to_cpu(migr_rec->ascending_migr) && |
3714 | (units < | |
9f421827 | 3715 | (get_num_migr_units(migr_rec)-1)) && |
befb629b AK |
3716 | (super->migr_rec->rec_status == |
3717 | __cpu_to_le32(UNIT_SRC_IN_CP_AREA))) | |
3718 | units++; | |
3719 | ||
d2e6d5d6 | 3720 | info->reshape_progress = blocks_per_unit * units; |
6289d1e0 | 3721 | |
7a862a02 | 3722 | dprintf("IMSM: General Migration checkpoint : %llu (%llu) -> read reshape progress : %llu\n", |
19986c72 MB |
3723 | (unsigned long long)units, |
3724 | (unsigned long long)blocks_per_unit, | |
3725 | info->reshape_progress); | |
75156c46 | 3726 | |
9529d343 | 3727 | used_disks = imsm_num_data_members(prev_map); |
75156c46 | 3728 | if (used_disks > 0) { |
895ffd99 | 3729 | info->custom_array_size = per_dev_array_size(map) * |
75156c46 | 3730 | used_disks; |
75156c46 | 3731 | } |
d2e6d5d6 | 3732 | } |
1e5c6983 DW |
3733 | case MIGR_VERIFY: |
3734 | /* we could emulate the checkpointing of | |
3735 | * 'sync_action=check' migrations, but for now | |
3736 | * we just immediately complete them | |
3737 | */ | |
3738 | case MIGR_REBUILD: | |
3739 | /* this is handled by container_content_imsm() */ | |
1e5c6983 DW |
3740 | case MIGR_STATE_CHANGE: |
3741 | /* FIXME handle other migrations */ | |
3742 | default: | |
3743 | /* we are not dirty, so... */ | |
3744 | info->resync_start = MaxSector; | |
3745 | } | |
b6796ce1 | 3746 | } |
301406c9 DW |
3747 | |
3748 | strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN); | |
3749 | info->name[MAX_RAID_SERIAL_LEN] = 0; | |
bf5a934a | 3750 | |
f35f2525 N |
3751 | info->array.major_version = -1; |
3752 | info->array.minor_version = -2; | |
4dd2df09 | 3753 | sprintf(info->text_version, "/%s/%d", st->container_devnm, info->container_member); |
a67dd8cc | 3754 | info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */ |
51006d85 | 3755 | uuid_from_super_imsm(st, info->uuid); |
a5d85af7 N |
3756 | |
3757 | if (dmap) { | |
3758 | int i, j; | |
3759 | for (i=0; i<map_disks; i++) { | |
3760 | dmap[i] = 0; | |
3761 | if (i < info->array.raid_disks) { | |
3762 | struct imsm_disk *dsk; | |
238c0a71 | 3763 | j = get_imsm_disk_idx(dev, i, MAP_X); |
a5d85af7 N |
3764 | dsk = get_imsm_disk(super, j); |
3765 | if (dsk && (dsk->status & CONFIGURED_DISK)) | |
3766 | dmap[i] = 1; | |
3767 | } | |
3768 | } | |
3769 | } | |
81ac8b4d | 3770 | } |
bf5a934a | 3771 | |
3b451610 AK |
3772 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, |
3773 | int failed, int look_in_map); | |
3774 | ||
3775 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev, | |
3776 | int look_in_map); | |
3777 | ||
3778 | static void manage_second_map(struct intel_super *super, struct imsm_dev *dev) | |
3779 | { | |
3780 | if (is_gen_migration(dev)) { | |
3781 | int failed; | |
3782 | __u8 map_state; | |
3783 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); | |
3784 | ||
3785 | failed = imsm_count_failed(super, dev, MAP_1); | |
238c0a71 | 3786 | map_state = imsm_check_degraded(super, dev, failed, MAP_1); |
3b451610 AK |
3787 | if (map2->map_state != map_state) { |
3788 | map2->map_state = map_state; | |
3789 | super->updates_pending++; | |
3790 | } | |
3791 | } | |
3792 | } | |
97b4d0e9 DW |
3793 | |
3794 | static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index) | |
3795 | { | |
3796 | struct dl *d; | |
3797 | ||
3798 | for (d = super->missing; d; d = d->next) | |
3799 | if (d->index == index) | |
3800 | return &d->disk; | |
3801 | return NULL; | |
3802 | } | |
3803 | ||
a5d85af7 | 3804 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map) |
4f5bc454 DW |
3805 | { |
3806 | struct intel_super *super = st->sb; | |
4f5bc454 | 3807 | struct imsm_disk *disk; |
a5d85af7 | 3808 | int map_disks = info->array.raid_disks; |
ab3cb6b3 N |
3809 | int max_enough = -1; |
3810 | int i; | |
3811 | struct imsm_super *mpb; | |
4f5bc454 | 3812 | |
bf5a934a | 3813 | if (super->current_vol >= 0) { |
a5d85af7 | 3814 | getinfo_super_imsm_volume(st, info, map); |
bf5a934a DW |
3815 | return; |
3816 | } | |
95eeceeb | 3817 | memset(info, 0, sizeof(*info)); |
d23fe947 DW |
3818 | |
3819 | /* Set raid_disks to zero so that Assemble will always pull in valid | |
3820 | * spares | |
3821 | */ | |
3822 | info->array.raid_disks = 0; | |
cdddbdbc DW |
3823 | info->array.level = LEVEL_CONTAINER; |
3824 | info->array.layout = 0; | |
3825 | info->array.md_minor = -1; | |
1011e834 | 3826 | info->array.ctime = 0; /* N/A for imsm */ |
cdddbdbc DW |
3827 | info->array.utime = 0; |
3828 | info->array.chunk_size = 0; | |
3829 | ||
3830 | info->disk.major = 0; | |
3831 | info->disk.minor = 0; | |
cdddbdbc | 3832 | info->disk.raid_disk = -1; |
c2c087e6 | 3833 | info->reshape_active = 0; |
f35f2525 N |
3834 | info->array.major_version = -1; |
3835 | info->array.minor_version = -2; | |
c2c087e6 | 3836 | strcpy(info->text_version, "imsm"); |
a67dd8cc | 3837 | info->safe_mode_delay = 0; |
c2c087e6 DW |
3838 | info->disk.number = -1; |
3839 | info->disk.state = 0; | |
c5afc314 | 3840 | info->name[0] = 0; |
921d9e16 | 3841 | info->recovery_start = MaxSector; |
3ad25638 | 3842 | info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb); |
5e46202e | 3843 | info->bb.supported = 1; |
c2c087e6 | 3844 | |
97b4d0e9 | 3845 | /* do we have the all the insync disks that we expect? */ |
ab3cb6b3 | 3846 | mpb = super->anchor; |
b7d81a38 | 3847 | info->events = __le32_to_cpu(mpb->generation_num); |
97b4d0e9 | 3848 | |
ab3cb6b3 N |
3849 | for (i = 0; i < mpb->num_raid_devs; i++) { |
3850 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
3851 | int failed, enough, j, missing = 0; | |
3852 | struct imsm_map *map; | |
3853 | __u8 state; | |
97b4d0e9 | 3854 | |
3b451610 AK |
3855 | failed = imsm_count_failed(super, dev, MAP_0); |
3856 | state = imsm_check_degraded(super, dev, failed, MAP_0); | |
238c0a71 | 3857 | map = get_imsm_map(dev, MAP_0); |
ab3cb6b3 N |
3858 | |
3859 | /* any newly missing disks? | |
3860 | * (catches single-degraded vs double-degraded) | |
3861 | */ | |
3862 | for (j = 0; j < map->num_members; j++) { | |
238c0a71 | 3863 | __u32 ord = get_imsm_ord_tbl_ent(dev, j, MAP_0); |
ab3cb6b3 N |
3864 | __u32 idx = ord_to_idx(ord); |
3865 | ||
20dc76d1 MT |
3866 | if (super->disks && super->disks->index == (int)idx) |
3867 | info->disk.raid_disk = j; | |
3868 | ||
ab3cb6b3 N |
3869 | if (!(ord & IMSM_ORD_REBUILD) && |
3870 | get_imsm_missing(super, idx)) { | |
3871 | missing = 1; | |
3872 | break; | |
3873 | } | |
97b4d0e9 | 3874 | } |
ab3cb6b3 N |
3875 | |
3876 | if (state == IMSM_T_STATE_FAILED) | |
3877 | enough = -1; | |
3878 | else if (state == IMSM_T_STATE_DEGRADED && | |
3879 | (state != map->map_state || missing)) | |
3880 | enough = 0; | |
3881 | else /* we're normal, or already degraded */ | |
3882 | enough = 1; | |
d2bde6d3 AK |
3883 | if (is_gen_migration(dev) && missing) { |
3884 | /* during general migration we need all disks | |
3885 | * that process is running on. | |
3886 | * No new missing disk is allowed. | |
3887 | */ | |
3888 | max_enough = -1; | |
3889 | enough = -1; | |
3890 | /* no more checks necessary | |
3891 | */ | |
3892 | break; | |
3893 | } | |
ab3cb6b3 N |
3894 | /* in the missing/failed disk case check to see |
3895 | * if at least one array is runnable | |
3896 | */ | |
3897 | max_enough = max(max_enough, enough); | |
3898 | } | |
1ade5cc1 | 3899 | dprintf("enough: %d\n", max_enough); |
ab3cb6b3 | 3900 | info->container_enough = max_enough; |
97b4d0e9 | 3901 | |
4a04ec6c | 3902 | if (super->disks) { |
14e8215b DW |
3903 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
3904 | ||
b9f594fe | 3905 | disk = &super->disks->disk; |
5551b113 | 3906 | info->data_offset = total_blocks(&super->disks->disk) - reserved; |
14e8215b | 3907 | info->component_size = reserved; |
25ed7e59 | 3908 | info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0; |
df474657 DW |
3909 | /* we don't change info->disk.raid_disk here because |
3910 | * this state will be finalized in mdmon after we have | |
3911 | * found the 'most fresh' version of the metadata | |
3912 | */ | |
25ed7e59 | 3913 | info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; |
2432ce9b AP |
3914 | info->disk.state |= (is_spare(disk) || is_journal(disk)) ? |
3915 | 0 : (1 << MD_DISK_SYNC); | |
cdddbdbc | 3916 | } |
a575e2a7 DW |
3917 | |
3918 | /* only call uuid_from_super_imsm when this disk is part of a populated container, | |
3919 | * ->compare_super may have updated the 'num_raid_devs' field for spares | |
3920 | */ | |
3921 | if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs) | |
36ba7d48 | 3922 | uuid_from_super_imsm(st, info->uuid); |
22e263f6 AC |
3923 | else |
3924 | memcpy(info->uuid, uuid_zero, sizeof(uuid_zero)); | |
a5d85af7 N |
3925 | |
3926 | /* I don't know how to compute 'map' on imsm, so use safe default */ | |
3927 | if (map) { | |
3928 | int i; | |
3929 | for (i = 0; i < map_disks; i++) | |
3930 | map[i] = 1; | |
3931 | } | |
3932 | ||
cdddbdbc DW |
3933 | } |
3934 | ||
5c4cd5da AC |
3935 | /* allocates memory and fills disk in mdinfo structure |
3936 | * for each disk in array */ | |
3937 | struct mdinfo *getinfo_super_disks_imsm(struct supertype *st) | |
3938 | { | |
594dc1b8 | 3939 | struct mdinfo *mddev; |
5c4cd5da AC |
3940 | struct intel_super *super = st->sb; |
3941 | struct imsm_disk *disk; | |
3942 | int count = 0; | |
3943 | struct dl *dl; | |
3944 | if (!super || !super->disks) | |
3945 | return NULL; | |
3946 | dl = super->disks; | |
503975b9 | 3947 | mddev = xcalloc(1, sizeof(*mddev)); |
5c4cd5da AC |
3948 | while (dl) { |
3949 | struct mdinfo *tmp; | |
3950 | disk = &dl->disk; | |
503975b9 | 3951 | tmp = xcalloc(1, sizeof(*tmp)); |
5c4cd5da AC |
3952 | if (mddev->devs) |
3953 | tmp->next = mddev->devs; | |
3954 | mddev->devs = tmp; | |
3955 | tmp->disk.number = count++; | |
3956 | tmp->disk.major = dl->major; | |
3957 | tmp->disk.minor = dl->minor; | |
3958 | tmp->disk.state = is_configured(disk) ? | |
3959 | (1 << MD_DISK_ACTIVE) : 0; | |
3960 | tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; | |
3961 | tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC); | |
3962 | tmp->disk.raid_disk = -1; | |
3963 | dl = dl->next; | |
3964 | } | |
3965 | return mddev; | |
3966 | } | |
3967 | ||
cdddbdbc | 3968 | static int update_super_imsm(struct supertype *st, struct mdinfo *info, |
03312b52 MK |
3969 | enum update_opt update, char *devname, |
3970 | int verbose, int uuid_set, char *homehost) | |
cdddbdbc | 3971 | { |
f352c545 DW |
3972 | /* For 'assemble' and 'force' we need to return non-zero if any |
3973 | * change was made. For others, the return value is ignored. | |
3974 | * Update options are: | |
3975 | * force-one : This device looks a bit old but needs to be included, | |
3976 | * update age info appropriately. | |
3977 | * assemble: clear any 'faulty' flag to allow this device to | |
3978 | * be assembled. | |
3979 | * force-array: Array is degraded but being forced, mark it clean | |
3980 | * if that will be needed to assemble it. | |
3981 | * | |
3982 | * newdev: not used ???? | |
3983 | * grow: Array has gained a new device - this is currently for | |
3984 | * linear only | |
3985 | * resync: mark as dirty so a resync will happen. | |
3986 | * name: update the name - preserving the homehost | |
6e46bf34 | 3987 | * uuid: Change the uuid of the array to match watch is given |
f352c545 DW |
3988 | * |
3989 | * Following are not relevant for this imsm: | |
3990 | * sparc2.2 : update from old dodgey metadata | |
3991 | * super-minor: change the preferred_minor number | |
3992 | * summaries: update redundant counters. | |
f352c545 DW |
3993 | * homehost: update the recorded homehost |
3994 | * _reshape_progress: record new reshape_progress position. | |
3995 | */ | |
6e46bf34 DW |
3996 | int rv = 1; |
3997 | struct intel_super *super = st->sb; | |
3998 | struct imsm_super *mpb; | |
f352c545 | 3999 | |
6e46bf34 DW |
4000 | /* we can only update container info */ |
4001 | if (!super || super->current_vol >= 0 || !super->anchor) | |
4002 | return 1; | |
4003 | ||
4004 | mpb = super->anchor; | |
4005 | ||
03312b52 | 4006 | switch (update) { |
4345e135 | 4007 | case UOPT_UUID: |
81a5b4f5 N |
4008 | /* We take this to mean that the family_num should be updated. |
4009 | * However that is much smaller than the uuid so we cannot really | |
4010 | * allow an explicit uuid to be given. And it is hard to reliably | |
4011 | * know if one was. | |
4012 | * So if !uuid_set we know the current uuid is random and just used | |
4013 | * the first 'int' and copy it to the other 3 positions. | |
4014 | * Otherwise we require the 4 'int's to be the same as would be the | |
4015 | * case if we are using a random uuid. So an explicit uuid will be | |
4016 | * accepted as long as all for ints are the same... which shouldn't hurt | |
6e46bf34 | 4017 | */ |
81a5b4f5 N |
4018 | if (!uuid_set) { |
4019 | info->uuid[1] = info->uuid[2] = info->uuid[3] = info->uuid[0]; | |
6e46bf34 | 4020 | rv = 0; |
81a5b4f5 N |
4021 | } else { |
4022 | if (info->uuid[0] != info->uuid[1] || | |
4023 | info->uuid[1] != info->uuid[2] || | |
4024 | info->uuid[2] != info->uuid[3]) | |
4025 | rv = -1; | |
4026 | else | |
4027 | rv = 0; | |
6e46bf34 | 4028 | } |
81a5b4f5 N |
4029 | if (rv == 0) |
4030 | mpb->orig_family_num = info->uuid[0]; | |
4345e135 MK |
4031 | break; |
4032 | case UOPT_SPEC_ASSEMBLE: | |
6e46bf34 | 4033 | rv = 0; |
4345e135 MK |
4034 | break; |
4035 | default: | |
1e2b2765 | 4036 | rv = -1; |
4345e135 MK |
4037 | break; |
4038 | } | |
f352c545 | 4039 | |
6e46bf34 DW |
4040 | /* successful update? recompute checksum */ |
4041 | if (rv == 0) | |
4042 | mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb)); | |
f352c545 DW |
4043 | |
4044 | return rv; | |
cdddbdbc DW |
4045 | } |
4046 | ||
c2c087e6 | 4047 | static size_t disks_to_mpb_size(int disks) |
cdddbdbc | 4048 | { |
c2c087e6 | 4049 | size_t size; |
cdddbdbc | 4050 | |
c2c087e6 DW |
4051 | size = sizeof(struct imsm_super); |
4052 | size += (disks - 1) * sizeof(struct imsm_disk); | |
4053 | size += 2 * sizeof(struct imsm_dev); | |
4054 | /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */ | |
4055 | size += (4 - 2) * sizeof(struct imsm_map); | |
4056 | /* 4 possible disk_ord_tbl's */ | |
4057 | size += 4 * (disks - 1) * sizeof(__u32); | |
bbab0940 TM |
4058 | /* maximum bbm log */ |
4059 | size += sizeof(struct bbm_log); | |
c2c087e6 DW |
4060 | |
4061 | return size; | |
4062 | } | |
4063 | ||
387fcd59 N |
4064 | static __u64 avail_size_imsm(struct supertype *st, __u64 devsize, |
4065 | unsigned long long data_offset) | |
c2c087e6 DW |
4066 | { |
4067 | if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS)) | |
4068 | return 0; | |
4069 | ||
4070 | return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS); | |
cdddbdbc DW |
4071 | } |
4072 | ||
ba2de7ba DW |
4073 | static void free_devlist(struct intel_super *super) |
4074 | { | |
4075 | struct intel_dev *dv; | |
4076 | ||
4077 | while (super->devlist) { | |
4078 | dv = super->devlist->next; | |
4079 | free(super->devlist->dev); | |
4080 | free(super->devlist); | |
4081 | super->devlist = dv; | |
4082 | } | |
4083 | } | |
4084 | ||
4085 | static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src) | |
4086 | { | |
4087 | memcpy(dest, src, sizeof_imsm_dev(src, 0)); | |
4088 | } | |
4089 | ||
c7b8547c MT |
4090 | static int compare_super_imsm(struct supertype *st, struct supertype *tst, |
4091 | int verbose) | |
cdddbdbc | 4092 | { |
601ffa78 | 4093 | /* return: |
cdddbdbc | 4094 | * 0 same, or first was empty, and second was copied |
601ffa78 | 4095 | * 1 sb are different |
cdddbdbc DW |
4096 | */ |
4097 | struct intel_super *first = st->sb; | |
4098 | struct intel_super *sec = tst->sb; | |
4099 | ||
5d500228 N |
4100 | if (!first) { |
4101 | st->sb = tst->sb; | |
4102 | tst->sb = NULL; | |
4103 | return 0; | |
4104 | } | |
601ffa78 | 4105 | |
8603ea6f LM |
4106 | /* in platform dependent environment test if the disks |
4107 | * use the same Intel hba | |
601ffa78 OS |
4108 | * if not on Intel hba at all, allow anything. |
4109 | * doesn't check HBAs if num_raid_devs is not set, as it means | |
4110 | * it is a free floating spare, and all spares regardless of HBA type | |
4111 | * will fall into separate container during the assembly | |
8603ea6f | 4112 | */ |
601ffa78 | 4113 | if (first->hba && sec->hba && first->anchor->num_raid_devs != 0) { |
6b781d33 | 4114 | if (first->hba->type != sec->hba->type) { |
c7b8547c MT |
4115 | if (verbose) |
4116 | pr_err("HBAs of devices do not match %s != %s\n", | |
4117 | get_sys_dev_type(first->hba->type), | |
4118 | get_sys_dev_type(sec->hba->type)); | |
601ffa78 | 4119 | return 1; |
6b781d33 AP |
4120 | } |
4121 | if (first->orom != sec->orom) { | |
c7b8547c MT |
4122 | if (verbose) |
4123 | pr_err("HBAs of devices do not match %s != %s\n", | |
4124 | first->hba->pci_id, sec->hba->pci_id); | |
601ffa78 | 4125 | return 1; |
8603ea6f LM |
4126 | } |
4127 | } | |
cdddbdbc | 4128 | |
d23fe947 DW |
4129 | if (first->anchor->num_raid_devs > 0 && |
4130 | sec->anchor->num_raid_devs > 0) { | |
a2b97981 DW |
4131 | /* Determine if these disks might ever have been |
4132 | * related. Further disambiguation can only take place | |
4133 | * in load_super_imsm_all | |
4134 | */ | |
4135 | __u32 first_family = first->anchor->orig_family_num; | |
4136 | __u32 sec_family = sec->anchor->orig_family_num; | |
4137 | ||
f796af5d DW |
4138 | if (memcmp(first->anchor->sig, sec->anchor->sig, |
4139 | MAX_SIGNATURE_LENGTH) != 0) | |
601ffa78 | 4140 | return 1; |
f796af5d | 4141 | |
a2b97981 DW |
4142 | if (first_family == 0) |
4143 | first_family = first->anchor->family_num; | |
4144 | if (sec_family == 0) | |
4145 | sec_family = sec->anchor->family_num; | |
4146 | ||
4147 | if (first_family != sec_family) | |
601ffa78 | 4148 | return 1; |
f796af5d | 4149 | |
d23fe947 | 4150 | } |
cdddbdbc | 4151 | |
601ffa78 OS |
4152 | /* if an anchor does not have num_raid_devs set then it is a free |
4153 | * floating spare. don't assosiate spare with any array, as during assembly | |
4154 | * spares shall fall into separate container, from which they can be moved | |
4155 | * when necessary | |
4156 | */ | |
4157 | if (first->anchor->num_raid_devs ^ sec->anchor->num_raid_devs) | |
4158 | return 1; | |
3e372e5a | 4159 | |
cdddbdbc DW |
4160 | return 0; |
4161 | } | |
4162 | ||
0030e8d6 DW |
4163 | static void fd2devname(int fd, char *name) |
4164 | { | |
0030e8d6 | 4165 | char *nm; |
0030e8d6 | 4166 | |
7c798f87 MT |
4167 | nm = fd2kname(fd); |
4168 | if (!nm) | |
0030e8d6 | 4169 | return; |
9587c373 | 4170 | |
7c798f87 | 4171 | snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm); |
0030e8d6 DW |
4172 | } |
4173 | ||
21e9380b AP |
4174 | static int nvme_get_serial(int fd, void *buf, size_t buf_len) |
4175 | { | |
fcebeb77 | 4176 | char path[PATH_MAX]; |
21e9380b AP |
4177 | char *name = fd2kname(fd); |
4178 | ||
4179 | if (!name) | |
4180 | return 1; | |
4181 | ||
4182 | if (strncmp(name, "nvme", 4) != 0) | |
4183 | return 1; | |
4184 | ||
fcebeb77 MT |
4185 | if (!diskfd_to_devpath(fd, 1, path)) |
4186 | return 1; | |
21e9380b | 4187 | |
fcebeb77 | 4188 | return devpath_to_char(path, "serial", buf, buf_len, 0); |
21e9380b AP |
4189 | } |
4190 | ||
cdddbdbc DW |
4191 | extern int scsi_get_serial(int fd, void *buf, size_t buf_len); |
4192 | ||
4193 | static int imsm_read_serial(int fd, char *devname, | |
6da53c0e | 4194 | __u8 *serial, size_t serial_buf_len) |
cdddbdbc | 4195 | { |
21e9380b | 4196 | char buf[50]; |
cdddbdbc | 4197 | int rv; |
6da53c0e | 4198 | size_t len; |
316e2bf4 DW |
4199 | char *dest; |
4200 | char *src; | |
21e9380b AP |
4201 | unsigned int i; |
4202 | ||
4203 | memset(buf, 0, sizeof(buf)); | |
cdddbdbc | 4204 | |
21e9380b | 4205 | rv = nvme_get_serial(fd, buf, sizeof(buf)); |
cdddbdbc | 4206 | |
21e9380b AP |
4207 | if (rv) |
4208 | rv = scsi_get_serial(fd, buf, sizeof(buf)); | |
f9ba0ff1 | 4209 | |
40ebbb9c | 4210 | if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) { |
f9ba0ff1 DW |
4211 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
4212 | fd2devname(fd, (char *) serial); | |
0030e8d6 DW |
4213 | return 0; |
4214 | } | |
4215 | ||
cdddbdbc DW |
4216 | if (rv != 0) { |
4217 | if (devname) | |
e7b84f9d N |
4218 | pr_err("Failed to retrieve serial for %s\n", |
4219 | devname); | |
cdddbdbc DW |
4220 | return rv; |
4221 | } | |
4222 | ||
316e2bf4 DW |
4223 | /* trim all whitespace and non-printable characters and convert |
4224 | * ':' to ';' | |
4225 | */ | |
21e9380b AP |
4226 | for (i = 0, dest = buf; i < sizeof(buf) && buf[i]; i++) { |
4227 | src = &buf[i]; | |
316e2bf4 DW |
4228 | if (*src > 0x20) { |
4229 | /* ':' is reserved for use in placeholder serial | |
4230 | * numbers for missing disks | |
4231 | */ | |
4232 | if (*src == ':') | |
4233 | *dest++ = ';'; | |
4234 | else | |
4235 | *dest++ = *src; | |
4236 | } | |
4237 | } | |
21e9380b AP |
4238 | len = dest - buf; |
4239 | dest = buf; | |
316e2bf4 | 4240 | |
6da53c0e BK |
4241 | if (len > serial_buf_len) { |
4242 | /* truncate leading characters */ | |
4243 | dest += len - serial_buf_len; | |
4244 | len = serial_buf_len; | |
316e2bf4 | 4245 | } |
5c3db629 | 4246 | |
6da53c0e | 4247 | memset(serial, 0, serial_buf_len); |
316e2bf4 | 4248 | memcpy(serial, dest, len); |
cdddbdbc DW |
4249 | |
4250 | return 0; | |
4251 | } | |
4252 | ||
1f24f035 DW |
4253 | static int serialcmp(__u8 *s1, __u8 *s2) |
4254 | { | |
4255 | return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN); | |
4256 | } | |
4257 | ||
4258 | static void serialcpy(__u8 *dest, __u8 *src) | |
4259 | { | |
4260 | strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN); | |
4261 | } | |
4262 | ||
54c2c1ea DW |
4263 | static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super) |
4264 | { | |
4265 | struct dl *dl; | |
4266 | ||
4267 | for (dl = super->disks; dl; dl = dl->next) | |
4268 | if (serialcmp(dl->serial, serial) == 0) | |
4269 | break; | |
4270 | ||
4271 | return dl; | |
4272 | } | |
4273 | ||
a2b97981 DW |
4274 | static struct imsm_disk * |
4275 | __serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx) | |
4276 | { | |
4277 | int i; | |
4278 | ||
4279 | for (i = 0; i < mpb->num_disks; i++) { | |
4280 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
4281 | ||
4282 | if (serialcmp(disk->serial, serial) == 0) { | |
4283 | if (idx) | |
4284 | *idx = i; | |
4285 | return disk; | |
4286 | } | |
4287 | } | |
4288 | ||
4289 | return NULL; | |
4290 | } | |
4291 | ||
cdddbdbc DW |
4292 | static int |
4293 | load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd) | |
4294 | { | |
a2b97981 | 4295 | struct imsm_disk *disk; |
cdddbdbc DW |
4296 | struct dl *dl; |
4297 | struct stat stb; | |
cdddbdbc | 4298 | int rv; |
a2b97981 | 4299 | char name[40]; |
d23fe947 DW |
4300 | __u8 serial[MAX_RAID_SERIAL_LEN]; |
4301 | ||
6da53c0e | 4302 | rv = imsm_read_serial(fd, devname, serial, MAX_RAID_SERIAL_LEN); |
d23fe947 DW |
4303 | |
4304 | if (rv != 0) | |
4305 | return 2; | |
4306 | ||
503975b9 | 4307 | dl = xcalloc(1, sizeof(*dl)); |
cdddbdbc | 4308 | |
a2b97981 DW |
4309 | fstat(fd, &stb); |
4310 | dl->major = major(stb.st_rdev); | |
4311 | dl->minor = minor(stb.st_rdev); | |
4312 | dl->next = super->disks; | |
4313 | dl->fd = keep_fd ? fd : -1; | |
4314 | assert(super->disks == NULL); | |
4315 | super->disks = dl; | |
4316 | serialcpy(dl->serial, serial); | |
4317 | dl->index = -2; | |
4318 | dl->e = NULL; | |
4319 | fd2devname(fd, name); | |
4320 | if (devname) | |
503975b9 | 4321 | dl->devname = xstrdup(devname); |
a2b97981 | 4322 | else |
503975b9 | 4323 | dl->devname = xstrdup(name); |
cdddbdbc | 4324 | |
d23fe947 | 4325 | /* look up this disk's index in the current anchor */ |
a2b97981 DW |
4326 | disk = __serial_to_disk(dl->serial, super->anchor, &dl->index); |
4327 | if (disk) { | |
4328 | dl->disk = *disk; | |
4329 | /* only set index on disks that are a member of a | |
4330 | * populated contianer, i.e. one with raid_devs | |
4331 | */ | |
4332 | if (is_failed(&dl->disk)) | |
3f6efecc | 4333 | dl->index = -2; |
2432ce9b | 4334 | else if (is_spare(&dl->disk) || is_journal(&dl->disk)) |
a2b97981 | 4335 | dl->index = -1; |
3f6efecc DW |
4336 | } |
4337 | ||
949c47a0 DW |
4338 | return 0; |
4339 | } | |
4340 | ||
0c046afd DW |
4341 | /* When migrating map0 contains the 'destination' state while map1 |
4342 | * contains the current state. When not migrating map0 contains the | |
4343 | * current state. This routine assumes that map[0].map_state is set to | |
4344 | * the current array state before being called. | |
4345 | * | |
4346 | * Migration is indicated by one of the following states | |
4347 | * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed) | |
e3bba0e0 | 4348 | * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal |
0c046afd | 4349 | * map1state=unitialized) |
1484e727 | 4350 | * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal |
0c046afd | 4351 | * map1state=normal) |
e3bba0e0 | 4352 | * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal |
0c046afd | 4353 | * map1state=degraded) |
8e59f3d8 AK |
4354 | * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal |
4355 | * map1state=normal) | |
0c046afd | 4356 | */ |
8e59f3d8 AK |
4357 | static void migrate(struct imsm_dev *dev, struct intel_super *super, |
4358 | __u8 to_state, int migr_type) | |
3393c6af | 4359 | { |
0c046afd | 4360 | struct imsm_map *dest; |
238c0a71 | 4361 | struct imsm_map *src = get_imsm_map(dev, MAP_0); |
3393c6af | 4362 | |
0c046afd | 4363 | dev->vol.migr_state = 1; |
1484e727 | 4364 | set_migr_type(dev, migr_type); |
4036e7ee | 4365 | set_vol_curr_migr_unit(dev, 0); |
238c0a71 | 4366 | dest = get_imsm_map(dev, MAP_1); |
0c046afd | 4367 | |
0556e1a2 | 4368 | /* duplicate and then set the target end state in map[0] */ |
3393c6af | 4369 | memcpy(dest, src, sizeof_imsm_map(src)); |
fb12a745 | 4370 | if (migr_type == MIGR_GEN_MIGR) { |
0556e1a2 DW |
4371 | __u32 ord; |
4372 | int i; | |
4373 | ||
4374 | for (i = 0; i < src->num_members; i++) { | |
4375 | ord = __le32_to_cpu(src->disk_ord_tbl[i]); | |
4376 | set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord)); | |
4377 | } | |
4378 | } | |
4379 | ||
8e59f3d8 AK |
4380 | if (migr_type == MIGR_GEN_MIGR) |
4381 | /* Clear migration record */ | |
4382 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
4383 | ||
0c046afd | 4384 | src->map_state = to_state; |
949c47a0 | 4385 | } |
f8f603f1 | 4386 | |
809da78e AK |
4387 | static void end_migration(struct imsm_dev *dev, struct intel_super *super, |
4388 | __u8 map_state) | |
f8f603f1 | 4389 | { |
238c0a71 AK |
4390 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
4391 | struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state == 0 ? | |
4392 | MAP_0 : MAP_1); | |
28bce06f | 4393 | int i, j; |
0556e1a2 DW |
4394 | |
4395 | /* merge any IMSM_ORD_REBUILD bits that were not successfully | |
4396 | * completed in the last migration. | |
4397 | * | |
28bce06f | 4398 | * FIXME add support for raid-level-migration |
0556e1a2 | 4399 | */ |
195d1d76 | 4400 | if (map_state != map->map_state && (is_gen_migration(dev) == false) && |
089f9d79 | 4401 | prev->map_state != IMSM_T_STATE_UNINITIALIZED) { |
809da78e AK |
4402 | /* when final map state is other than expected |
4403 | * merge maps (not for migration) | |
4404 | */ | |
4405 | int failed; | |
4406 | ||
4407 | for (i = 0; i < prev->num_members; i++) | |
4408 | for (j = 0; j < map->num_members; j++) | |
4409 | /* during online capacity expansion | |
4410 | * disks position can be changed | |
4411 | * if takeover is used | |
4412 | */ | |
4413 | if (ord_to_idx(map->disk_ord_tbl[j]) == | |
4414 | ord_to_idx(prev->disk_ord_tbl[i])) { | |
4415 | map->disk_ord_tbl[j] |= | |
4416 | prev->disk_ord_tbl[i]; | |
4417 | break; | |
4418 | } | |
4419 | failed = imsm_count_failed(super, dev, MAP_0); | |
4420 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
4421 | } | |
f8f603f1 DW |
4422 | |
4423 | dev->vol.migr_state = 0; | |
ea672ee1 | 4424 | set_migr_type(dev, 0); |
4036e7ee | 4425 | set_vol_curr_migr_unit(dev, 0); |
f8f603f1 DW |
4426 | map->map_state = map_state; |
4427 | } | |
949c47a0 DW |
4428 | |
4429 | static int parse_raid_devices(struct intel_super *super) | |
4430 | { | |
4431 | int i; | |
4432 | struct imsm_dev *dev_new; | |
4d7b1503 | 4433 | size_t len, len_migr; |
401d313b | 4434 | size_t max_len = 0; |
4d7b1503 DW |
4435 | size_t space_needed = 0; |
4436 | struct imsm_super *mpb = super->anchor; | |
949c47a0 DW |
4437 | |
4438 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
4439 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
ba2de7ba | 4440 | struct intel_dev *dv; |
949c47a0 | 4441 | |
4d7b1503 DW |
4442 | len = sizeof_imsm_dev(dev_iter, 0); |
4443 | len_migr = sizeof_imsm_dev(dev_iter, 1); | |
4444 | if (len_migr > len) | |
4445 | space_needed += len_migr - len; | |
ca9de185 | 4446 | |
503975b9 | 4447 | dv = xmalloc(sizeof(*dv)); |
401d313b AK |
4448 | if (max_len < len_migr) |
4449 | max_len = len_migr; | |
4450 | if (max_len > len_migr) | |
4451 | space_needed += max_len - len_migr; | |
503975b9 | 4452 | dev_new = xmalloc(max_len); |
949c47a0 | 4453 | imsm_copy_dev(dev_new, dev_iter); |
ba2de7ba DW |
4454 | dv->dev = dev_new; |
4455 | dv->index = i; | |
4456 | dv->next = super->devlist; | |
4457 | super->devlist = dv; | |
949c47a0 | 4458 | } |
cdddbdbc | 4459 | |
4d7b1503 DW |
4460 | /* ensure that super->buf is large enough when all raid devices |
4461 | * are migrating | |
4462 | */ | |
4463 | if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) { | |
4464 | void *buf; | |
4465 | ||
f36a9ecd PB |
4466 | len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed, |
4467 | super->sector_size); | |
4468 | if (posix_memalign(&buf, MAX_SECTOR_SIZE, len) != 0) | |
4d7b1503 DW |
4469 | return 1; |
4470 | ||
1f45a8ad DW |
4471 | memcpy(buf, super->buf, super->len); |
4472 | memset(buf + super->len, 0, len - super->len); | |
4d7b1503 DW |
4473 | free(super->buf); |
4474 | super->buf = buf; | |
4475 | super->len = len; | |
4476 | } | |
ca9de185 | 4477 | |
bbab0940 TM |
4478 | super->extra_space += space_needed; |
4479 | ||
cdddbdbc DW |
4480 | return 0; |
4481 | } | |
4482 | ||
e2f41b2c AK |
4483 | /******************************************************************************* |
4484 | * Function: check_mpb_migr_compatibility | |
4485 | * Description: Function checks for unsupported migration features: | |
4486 | * - migration optimization area (pba_of_lba0) | |
4487 | * - descending reshape (ascending_migr) | |
4488 | * Parameters: | |
4489 | * super : imsm metadata information | |
4490 | * Returns: | |
4491 | * 0 : migration is compatible | |
4492 | * -1 : migration is not compatible | |
4493 | ******************************************************************************/ | |
4494 | int check_mpb_migr_compatibility(struct intel_super *super) | |
4495 | { | |
4496 | struct imsm_map *map0, *map1; | |
4497 | struct migr_record *migr_rec = super->migr_rec; | |
4498 | int i; | |
4499 | ||
4500 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
4501 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
4502 | ||
756a15f3 | 4503 | if (dev_iter->vol.migr_state == 1 && |
e2f41b2c AK |
4504 | dev_iter->vol.migr_type == MIGR_GEN_MIGR) { |
4505 | /* This device is migrating */ | |
238c0a71 AK |
4506 | map0 = get_imsm_map(dev_iter, MAP_0); |
4507 | map1 = get_imsm_map(dev_iter, MAP_1); | |
5551b113 | 4508 | if (pba_of_lba0(map0) != pba_of_lba0(map1)) |
e2f41b2c AK |
4509 | /* migration optimization area was used */ |
4510 | return -1; | |
fc54fe7a JS |
4511 | if (migr_rec->ascending_migr == 0 && |
4512 | migr_rec->dest_depth_per_unit > 0) | |
e2f41b2c AK |
4513 | /* descending reshape not supported yet */ |
4514 | return -1; | |
4515 | } | |
4516 | } | |
4517 | return 0; | |
4518 | } | |
4519 | ||
d23fe947 | 4520 | static void __free_imsm(struct intel_super *super, int free_disks); |
9ca2c81c | 4521 | |
cdddbdbc | 4522 | /* load_imsm_mpb - read matrix metadata |
f2f5c343 | 4523 | * allocates super->mpb to be freed by free_imsm |
cdddbdbc DW |
4524 | */ |
4525 | static int load_imsm_mpb(int fd, struct intel_super *super, char *devname) | |
4526 | { | |
4527 | unsigned long long dsize; | |
cdddbdbc | 4528 | unsigned long long sectors; |
f36a9ecd | 4529 | unsigned int sector_size = super->sector_size; |
cdddbdbc | 4530 | struct stat; |
6416d527 | 4531 | struct imsm_super *anchor; |
cdddbdbc DW |
4532 | __u32 check_sum; |
4533 | ||
cdddbdbc | 4534 | get_dev_size(fd, NULL, &dsize); |
f36a9ecd | 4535 | if (dsize < 2*sector_size) { |
64436f06 | 4536 | if (devname) |
e7b84f9d N |
4537 | pr_err("%s: device to small for imsm\n", |
4538 | devname); | |
64436f06 N |
4539 | return 1; |
4540 | } | |
cdddbdbc | 4541 | |
f36a9ecd | 4542 | if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0) { |
cdddbdbc | 4543 | if (devname) |
e7b84f9d N |
4544 | pr_err("Cannot seek to anchor block on %s: %s\n", |
4545 | devname, strerror(errno)); | |
cdddbdbc DW |
4546 | return 1; |
4547 | } | |
4548 | ||
f36a9ecd | 4549 | if (posix_memalign((void **)&anchor, sector_size, sector_size) != 0) { |
ad97895e | 4550 | if (devname) |
7a862a02 | 4551 | pr_err("Failed to allocate imsm anchor buffer on %s\n", devname); |
ad97895e DW |
4552 | return 1; |
4553 | } | |
466070ad | 4554 | if ((unsigned int)read(fd, anchor, sector_size) != sector_size) { |
cdddbdbc | 4555 | if (devname) |
e7b84f9d N |
4556 | pr_err("Cannot read anchor block on %s: %s\n", |
4557 | devname, strerror(errno)); | |
6416d527 | 4558 | free(anchor); |
cdddbdbc DW |
4559 | return 1; |
4560 | } | |
4561 | ||
6416d527 | 4562 | if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) { |
cdddbdbc | 4563 | if (devname) |
e7b84f9d | 4564 | pr_err("no IMSM anchor on %s\n", devname); |
6416d527 | 4565 | free(anchor); |
cdddbdbc DW |
4566 | return 2; |
4567 | } | |
4568 | ||
d23fe947 | 4569 | __free_imsm(super, 0); |
f2f5c343 LM |
4570 | /* reload capability and hba */ |
4571 | ||
4572 | /* capability and hba must be updated with new super allocation */ | |
d424212e | 4573 | find_intel_hba_capability(fd, super, devname); |
f36a9ecd PB |
4574 | super->len = ROUND_UP(anchor->mpb_size, sector_size); |
4575 | if (posix_memalign(&super->buf, MAX_SECTOR_SIZE, super->len) != 0) { | |
cdddbdbc | 4576 | if (devname) |
e7b84f9d N |
4577 | pr_err("unable to allocate %zu byte mpb buffer\n", |
4578 | super->len); | |
6416d527 | 4579 | free(anchor); |
cdddbdbc DW |
4580 | return 2; |
4581 | } | |
f36a9ecd | 4582 | memcpy(super->buf, anchor, sector_size); |
cdddbdbc | 4583 | |
f36a9ecd | 4584 | sectors = mpb_sectors(anchor, sector_size) - 1; |
6416d527 | 4585 | free(anchor); |
8e59f3d8 | 4586 | |
85337573 AO |
4587 | if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE, |
4588 | MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) { | |
1ade5cc1 | 4589 | pr_err("could not allocate migr_rec buffer\n"); |
8e59f3d8 | 4590 | free(super->buf); |
50cd06b4 | 4591 | super->buf = NULL; |
8e59f3d8 AK |
4592 | return 2; |
4593 | } | |
51d83f5d | 4594 | super->clean_migration_record_by_mdmon = 0; |
8e59f3d8 | 4595 | |
949c47a0 | 4596 | if (!sectors) { |
ecf45690 DW |
4597 | check_sum = __gen_imsm_checksum(super->anchor); |
4598 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
4599 | if (devname) | |
e7b84f9d N |
4600 | pr_err("IMSM checksum %x != %x on %s\n", |
4601 | check_sum, | |
4602 | __le32_to_cpu(super->anchor->check_sum), | |
4603 | devname); | |
ecf45690 DW |
4604 | return 2; |
4605 | } | |
4606 | ||
a2b97981 | 4607 | return 0; |
949c47a0 | 4608 | } |
cdddbdbc DW |
4609 | |
4610 | /* read the extended mpb */ | |
f36a9ecd | 4611 | if (lseek64(fd, dsize - (sector_size * (2 + sectors)), SEEK_SET) < 0) { |
cdddbdbc | 4612 | if (devname) |
e7b84f9d N |
4613 | pr_err("Cannot seek to extended mpb on %s: %s\n", |
4614 | devname, strerror(errno)); | |
cdddbdbc DW |
4615 | return 1; |
4616 | } | |
4617 | ||
f36a9ecd PB |
4618 | if ((unsigned int)read(fd, super->buf + sector_size, |
4619 | super->len - sector_size) != super->len - sector_size) { | |
cdddbdbc | 4620 | if (devname) |
e7b84f9d N |
4621 | pr_err("Cannot read extended mpb on %s: %s\n", |
4622 | devname, strerror(errno)); | |
cdddbdbc DW |
4623 | return 2; |
4624 | } | |
4625 | ||
949c47a0 DW |
4626 | check_sum = __gen_imsm_checksum(super->anchor); |
4627 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
cdddbdbc | 4628 | if (devname) |
e7b84f9d N |
4629 | pr_err("IMSM checksum %x != %x on %s\n", |
4630 | check_sum, __le32_to_cpu(super->anchor->check_sum), | |
4631 | devname); | |
db575f3b | 4632 | return 3; |
cdddbdbc DW |
4633 | } |
4634 | ||
a2b97981 DW |
4635 | return 0; |
4636 | } | |
4637 | ||
8e59f3d8 AK |
4638 | static int read_imsm_migr_rec(int fd, struct intel_super *super); |
4639 | ||
97f81ee2 CA |
4640 | /* clears hi bits in metadata if MPB_ATTRIB_2TB_DISK not set */ |
4641 | static void clear_hi(struct intel_super *super) | |
4642 | { | |
4643 | struct imsm_super *mpb = super->anchor; | |
4644 | int i, n; | |
4645 | if (mpb->attributes & MPB_ATTRIB_2TB_DISK) | |
4646 | return; | |
4647 | for (i = 0; i < mpb->num_disks; ++i) { | |
4648 | struct imsm_disk *disk = &mpb->disk[i]; | |
4649 | disk->total_blocks_hi = 0; | |
4650 | } | |
4651 | for (i = 0; i < mpb->num_raid_devs; ++i) { | |
4652 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
97f81ee2 CA |
4653 | for (n = 0; n < 2; ++n) { |
4654 | struct imsm_map *map = get_imsm_map(dev, n); | |
4655 | if (!map) | |
4656 | continue; | |
4657 | map->pba_of_lba0_hi = 0; | |
4658 | map->blocks_per_member_hi = 0; | |
4659 | map->num_data_stripes_hi = 0; | |
4660 | } | |
4661 | } | |
4662 | } | |
4663 | ||
a2b97981 DW |
4664 | static int |
4665 | load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd) | |
4666 | { | |
4667 | int err; | |
4668 | ||
4669 | err = load_imsm_mpb(fd, super, devname); | |
4670 | if (err) | |
4671 | return err; | |
f36a9ecd PB |
4672 | if (super->sector_size == 4096) |
4673 | convert_from_4k(super); | |
a2b97981 DW |
4674 | err = load_imsm_disk(fd, super, devname, keep_fd); |
4675 | if (err) | |
4676 | return err; | |
4677 | err = parse_raid_devices(super); | |
8d67477f TM |
4678 | if (err) |
4679 | return err; | |
4680 | err = load_bbm_log(super); | |
97f81ee2 | 4681 | clear_hi(super); |
a2b97981 | 4682 | return err; |
cdddbdbc DW |
4683 | } |
4684 | ||
4389ce73 | 4685 | static void __free_imsm_disk(struct dl *d, int do_close) |
ae6aad82 | 4686 | { |
4389ce73 MT |
4687 | if (do_close) |
4688 | close_fd(&d->fd); | |
ae6aad82 DW |
4689 | if (d->devname) |
4690 | free(d->devname); | |
0dcecb2e DW |
4691 | if (d->e) |
4692 | free(d->e); | |
ae6aad82 DW |
4693 | free(d); |
4694 | ||
4695 | } | |
1a64be56 | 4696 | |
cdddbdbc DW |
4697 | static void free_imsm_disks(struct intel_super *super) |
4698 | { | |
47ee5a45 | 4699 | struct dl *d; |
cdddbdbc | 4700 | |
47ee5a45 DW |
4701 | while (super->disks) { |
4702 | d = super->disks; | |
cdddbdbc | 4703 | super->disks = d->next; |
3a85bf0e | 4704 | __free_imsm_disk(d, 1); |
cdddbdbc | 4705 | } |
cb82edca AK |
4706 | while (super->disk_mgmt_list) { |
4707 | d = super->disk_mgmt_list; | |
4708 | super->disk_mgmt_list = d->next; | |
3a85bf0e | 4709 | __free_imsm_disk(d, 1); |
cb82edca | 4710 | } |
47ee5a45 DW |
4711 | while (super->missing) { |
4712 | d = super->missing; | |
4713 | super->missing = d->next; | |
3a85bf0e | 4714 | __free_imsm_disk(d, 1); |
47ee5a45 DW |
4715 | } |
4716 | ||
cdddbdbc DW |
4717 | } |
4718 | ||
9ca2c81c | 4719 | /* free all the pieces hanging off of a super pointer */ |
d23fe947 | 4720 | static void __free_imsm(struct intel_super *super, int free_disks) |
cdddbdbc | 4721 | { |
88654014 LM |
4722 | struct intel_hba *elem, *next; |
4723 | ||
9ca2c81c | 4724 | if (super->buf) { |
949c47a0 | 4725 | free(super->buf); |
9ca2c81c DW |
4726 | super->buf = NULL; |
4727 | } | |
f2f5c343 LM |
4728 | /* unlink capability description */ |
4729 | super->orom = NULL; | |
8e59f3d8 AK |
4730 | if (super->migr_rec_buf) { |
4731 | free(super->migr_rec_buf); | |
4732 | super->migr_rec_buf = NULL; | |
4733 | } | |
d23fe947 DW |
4734 | if (free_disks) |
4735 | free_imsm_disks(super); | |
ba2de7ba | 4736 | free_devlist(super); |
88654014 LM |
4737 | elem = super->hba; |
4738 | while (elem) { | |
4739 | if (elem->path) | |
4740 | free((void *)elem->path); | |
4741 | next = elem->next; | |
4742 | free(elem); | |
4743 | elem = next; | |
88c32bb1 | 4744 | } |
8d67477f TM |
4745 | if (super->bbm_log) |
4746 | free(super->bbm_log); | |
88654014 | 4747 | super->hba = NULL; |
cdddbdbc DW |
4748 | } |
4749 | ||
9ca2c81c DW |
4750 | static void free_imsm(struct intel_super *super) |
4751 | { | |
d23fe947 | 4752 | __free_imsm(super, 1); |
928f1424 | 4753 | free(super->bb.entries); |
9ca2c81c DW |
4754 | free(super); |
4755 | } | |
cdddbdbc DW |
4756 | |
4757 | static void free_super_imsm(struct supertype *st) | |
4758 | { | |
4759 | struct intel_super *super = st->sb; | |
4760 | ||
4761 | if (!super) | |
4762 | return; | |
4763 | ||
4764 | free_imsm(super); | |
4765 | st->sb = NULL; | |
4766 | } | |
4767 | ||
49133e57 | 4768 | static struct intel_super *alloc_super(void) |
c2c087e6 | 4769 | { |
503975b9 | 4770 | struct intel_super *super = xcalloc(1, sizeof(*super)); |
c2c087e6 | 4771 | |
503975b9 N |
4772 | super->current_vol = -1; |
4773 | super->create_offset = ~((unsigned long long) 0); | |
928f1424 TM |
4774 | |
4775 | super->bb.entries = xmalloc(BBM_LOG_MAX_ENTRIES * | |
4776 | sizeof(struct md_bb_entry)); | |
4777 | if (!super->bb.entries) { | |
4778 | free(super); | |
4779 | return NULL; | |
4780 | } | |
4781 | ||
c2c087e6 DW |
4782 | return super; |
4783 | } | |
4784 | ||
f0f5a016 LM |
4785 | /* |
4786 | * find and allocate hba and OROM/EFI based on valid fd of RAID component device | |
4787 | */ | |
d424212e | 4788 | static int find_intel_hba_capability(int fd, struct intel_super *super, char *devname) |
f0f5a016 LM |
4789 | { |
4790 | struct sys_dev *hba_name; | |
4791 | int rv = 0; | |
4792 | ||
4389ce73 | 4793 | if (is_fd_valid(fd) && test_partition(fd)) { |
3a30e28e MT |
4794 | pr_err("imsm: %s is a partition, cannot be used in IMSM\n", |
4795 | devname); | |
4796 | return 1; | |
4797 | } | |
420dafcd | 4798 | if (!is_fd_valid(fd) || check_no_platform()) { |
f2f5c343 | 4799 | super->orom = NULL; |
f0f5a016 LM |
4800 | super->hba = NULL; |
4801 | return 0; | |
4802 | } | |
4803 | hba_name = find_disk_attached_hba(fd, NULL); | |
4804 | if (!hba_name) { | |
d424212e | 4805 | if (devname) |
e7b84f9d N |
4806 | pr_err("%s is not attached to Intel(R) RAID controller.\n", |
4807 | devname); | |
f0f5a016 LM |
4808 | return 1; |
4809 | } | |
4810 | rv = attach_hba_to_super(super, hba_name); | |
4811 | if (rv == 2) { | |
d424212e N |
4812 | if (devname) { |
4813 | struct intel_hba *hba = super->hba; | |
f0f5a016 | 4814 | |
60f0f54d PB |
4815 | pr_err("%s is attached to Intel(R) %s %s (%s),\n" |
4816 | " but the container is assigned to Intel(R) %s %s (", | |
d424212e | 4817 | devname, |
614902f6 | 4818 | get_sys_dev_type(hba_name->type), |
75350d87 KF |
4819 | hba_name->type == SYS_DEV_VMD || hba_name->type == SYS_DEV_SATA_VMD ? |
4820 | "domain" : "RAID controller", | |
f0f5a016 | 4821 | hba_name->pci_id ? : "Err!", |
60f0f54d | 4822 | get_sys_dev_type(super->hba->type), |
75350d87 KF |
4823 | hba->type == SYS_DEV_VMD || hba_name->type == SYS_DEV_SATA_VMD ? |
4824 | "domain" : "RAID controller"); | |
f0f5a016 | 4825 | |
f0f5a016 LM |
4826 | while (hba) { |
4827 | fprintf(stderr, "%s", hba->pci_id ? : "Err!"); | |
4828 | if (hba->next) | |
4829 | fprintf(stderr, ", "); | |
4830 | hba = hba->next; | |
4831 | } | |
6b781d33 | 4832 | fprintf(stderr, ").\n" |
cca67208 | 4833 | " Mixing devices attached to different controllers is not allowed.\n"); |
f0f5a016 | 4834 | } |
f0f5a016 LM |
4835 | return 2; |
4836 | } | |
6b781d33 | 4837 | super->orom = find_imsm_capability(hba_name); |
f2f5c343 LM |
4838 | if (!super->orom) |
4839 | return 3; | |
614902f6 | 4840 | |
f0f5a016 LM |
4841 | return 0; |
4842 | } | |
4843 | ||
47ee5a45 DW |
4844 | /* find_missing - helper routine for load_super_imsm_all that identifies |
4845 | * disks that have disappeared from the system. This routine relies on | |
4846 | * the mpb being uptodate, which it is at load time. | |
4847 | */ | |
4848 | static int find_missing(struct intel_super *super) | |
4849 | { | |
4850 | int i; | |
4851 | struct imsm_super *mpb = super->anchor; | |
4852 | struct dl *dl; | |
4853 | struct imsm_disk *disk; | |
47ee5a45 DW |
4854 | |
4855 | for (i = 0; i < mpb->num_disks; i++) { | |
4856 | disk = __get_imsm_disk(mpb, i); | |
54c2c1ea | 4857 | dl = serial_to_dl(disk->serial, super); |
47ee5a45 DW |
4858 | if (dl) |
4859 | continue; | |
47ee5a45 | 4860 | |
503975b9 | 4861 | dl = xmalloc(sizeof(*dl)); |
47ee5a45 DW |
4862 | dl->major = 0; |
4863 | dl->minor = 0; | |
4864 | dl->fd = -1; | |
503975b9 | 4865 | dl->devname = xstrdup("missing"); |
47ee5a45 DW |
4866 | dl->index = i; |
4867 | serialcpy(dl->serial, disk->serial); | |
4868 | dl->disk = *disk; | |
689c9bf3 | 4869 | dl->e = NULL; |
47ee5a45 DW |
4870 | dl->next = super->missing; |
4871 | super->missing = dl; | |
4872 | } | |
4873 | ||
4874 | return 0; | |
4875 | } | |
4876 | ||
a2b97981 DW |
4877 | static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list) |
4878 | { | |
4879 | struct intel_disk *idisk = disk_list; | |
4880 | ||
4881 | while (idisk) { | |
4882 | if (serialcmp(idisk->disk.serial, serial) == 0) | |
4883 | break; | |
4884 | idisk = idisk->next; | |
4885 | } | |
4886 | ||
4887 | return idisk; | |
4888 | } | |
4889 | ||
4890 | static int __prep_thunderdome(struct intel_super **table, int tbl_size, | |
4891 | struct intel_super *super, | |
4892 | struct intel_disk **disk_list) | |
4893 | { | |
4894 | struct imsm_disk *d = &super->disks->disk; | |
4895 | struct imsm_super *mpb = super->anchor; | |
4896 | int i, j; | |
4897 | ||
4898 | for (i = 0; i < tbl_size; i++) { | |
4899 | struct imsm_super *tbl_mpb = table[i]->anchor; | |
4900 | struct imsm_disk *tbl_d = &table[i]->disks->disk; | |
4901 | ||
4902 | if (tbl_mpb->family_num == mpb->family_num) { | |
4903 | if (tbl_mpb->check_sum == mpb->check_sum) { | |
1ade5cc1 N |
4904 | dprintf("mpb from %d:%d matches %d:%d\n", |
4905 | super->disks->major, | |
a2b97981 DW |
4906 | super->disks->minor, |
4907 | table[i]->disks->major, | |
4908 | table[i]->disks->minor); | |
4909 | break; | |
4910 | } | |
4911 | ||
4912 | if (((is_configured(d) && !is_configured(tbl_d)) || | |
4913 | is_configured(d) == is_configured(tbl_d)) && | |
4914 | tbl_mpb->generation_num < mpb->generation_num) { | |
4915 | /* current version of the mpb is a | |
4916 | * better candidate than the one in | |
4917 | * super_table, but copy over "cross | |
4918 | * generational" status | |
4919 | */ | |
4920 | struct intel_disk *idisk; | |
4921 | ||
1ade5cc1 N |
4922 | dprintf("mpb from %d:%d replaces %d:%d\n", |
4923 | super->disks->major, | |
a2b97981 DW |
4924 | super->disks->minor, |
4925 | table[i]->disks->major, | |
4926 | table[i]->disks->minor); | |
4927 | ||
4928 | idisk = disk_list_get(tbl_d->serial, *disk_list); | |
4929 | if (idisk && is_failed(&idisk->disk)) | |
4930 | tbl_d->status |= FAILED_DISK; | |
4931 | break; | |
4932 | } else { | |
4933 | struct intel_disk *idisk; | |
4934 | struct imsm_disk *disk; | |
4935 | ||
4936 | /* tbl_mpb is more up to date, but copy | |
4937 | * over cross generational status before | |
4938 | * returning | |
4939 | */ | |
4940 | disk = __serial_to_disk(d->serial, mpb, NULL); | |
4941 | if (disk && is_failed(disk)) | |
4942 | d->status |= FAILED_DISK; | |
4943 | ||
4944 | idisk = disk_list_get(d->serial, *disk_list); | |
4945 | if (idisk) { | |
4946 | idisk->owner = i; | |
4947 | if (disk && is_configured(disk)) | |
4948 | idisk->disk.status |= CONFIGURED_DISK; | |
4949 | } | |
4950 | ||
1ade5cc1 N |
4951 | dprintf("mpb from %d:%d prefer %d:%d\n", |
4952 | super->disks->major, | |
a2b97981 DW |
4953 | super->disks->minor, |
4954 | table[i]->disks->major, | |
4955 | table[i]->disks->minor); | |
4956 | ||
4957 | return tbl_size; | |
4958 | } | |
4959 | } | |
4960 | } | |
4961 | ||
4962 | if (i >= tbl_size) | |
4963 | table[tbl_size++] = super; | |
4964 | else | |
4965 | table[i] = super; | |
4966 | ||
4967 | /* update/extend the merged list of imsm_disk records */ | |
4968 | for (j = 0; j < mpb->num_disks; j++) { | |
4969 | struct imsm_disk *disk = __get_imsm_disk(mpb, j); | |
4970 | struct intel_disk *idisk; | |
4971 | ||
4972 | idisk = disk_list_get(disk->serial, *disk_list); | |
4973 | if (idisk) { | |
4974 | idisk->disk.status |= disk->status; | |
4975 | if (is_configured(&idisk->disk) || | |
4976 | is_failed(&idisk->disk)) | |
4977 | idisk->disk.status &= ~(SPARE_DISK); | |
4978 | } else { | |
503975b9 | 4979 | idisk = xcalloc(1, sizeof(*idisk)); |
a2b97981 DW |
4980 | idisk->owner = IMSM_UNKNOWN_OWNER; |
4981 | idisk->disk = *disk; | |
4982 | idisk->next = *disk_list; | |
4983 | *disk_list = idisk; | |
4984 | } | |
4985 | ||
4986 | if (serialcmp(idisk->disk.serial, d->serial) == 0) | |
4987 | idisk->owner = i; | |
4988 | } | |
4989 | ||
4990 | return tbl_size; | |
4991 | } | |
4992 | ||
4993 | static struct intel_super * | |
4994 | validate_members(struct intel_super *super, struct intel_disk *disk_list, | |
4995 | const int owner) | |
4996 | { | |
4997 | struct imsm_super *mpb = super->anchor; | |
4998 | int ok_count = 0; | |
4999 | int i; | |
5000 | ||
5001 | for (i = 0; i < mpb->num_disks; i++) { | |
5002 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
5003 | struct intel_disk *idisk; | |
5004 | ||
5005 | idisk = disk_list_get(disk->serial, disk_list); | |
5006 | if (idisk) { | |
5007 | if (idisk->owner == owner || | |
5008 | idisk->owner == IMSM_UNKNOWN_OWNER) | |
5009 | ok_count++; | |
5010 | else | |
1ade5cc1 N |
5011 | dprintf("'%.16s' owner %d != %d\n", |
5012 | disk->serial, idisk->owner, | |
a2b97981 DW |
5013 | owner); |
5014 | } else { | |
1ade5cc1 N |
5015 | dprintf("unknown disk %x [%d]: %.16s\n", |
5016 | __le32_to_cpu(mpb->family_num), i, | |
a2b97981 DW |
5017 | disk->serial); |
5018 | break; | |
5019 | } | |
5020 | } | |
5021 | ||
5022 | if (ok_count == mpb->num_disks) | |
5023 | return super; | |
5024 | return NULL; | |
5025 | } | |
5026 | ||
5027 | static void show_conflicts(__u32 family_num, struct intel_super *super_list) | |
5028 | { | |
5029 | struct intel_super *s; | |
5030 | ||
5031 | for (s = super_list; s; s = s->next) { | |
5032 | if (family_num != s->anchor->family_num) | |
5033 | continue; | |
e12b3daa | 5034 | pr_err("Conflict, offlining family %#x on '%s'\n", |
a2b97981 DW |
5035 | __le32_to_cpu(family_num), s->disks->devname); |
5036 | } | |
5037 | } | |
5038 | ||
5039 | static struct intel_super * | |
5040 | imsm_thunderdome(struct intel_super **super_list, int len) | |
5041 | { | |
5042 | struct intel_super *super_table[len]; | |
5043 | struct intel_disk *disk_list = NULL; | |
5044 | struct intel_super *champion, *spare; | |
5045 | struct intel_super *s, **del; | |
5046 | int tbl_size = 0; | |
5047 | int conflict; | |
5048 | int i; | |
5049 | ||
5050 | memset(super_table, 0, sizeof(super_table)); | |
5051 | for (s = *super_list; s; s = s->next) | |
5052 | tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list); | |
5053 | ||
5054 | for (i = 0; i < tbl_size; i++) { | |
5055 | struct imsm_disk *d; | |
5056 | struct intel_disk *idisk; | |
5057 | struct imsm_super *mpb = super_table[i]->anchor; | |
5058 | ||
5059 | s = super_table[i]; | |
5060 | d = &s->disks->disk; | |
5061 | ||
5062 | /* 'd' must appear in merged disk list for its | |
5063 | * configuration to be valid | |
5064 | */ | |
5065 | idisk = disk_list_get(d->serial, disk_list); | |
5066 | if (idisk && idisk->owner == i) | |
5067 | s = validate_members(s, disk_list, i); | |
5068 | else | |
5069 | s = NULL; | |
5070 | ||
5071 | if (!s) | |
1ade5cc1 N |
5072 | dprintf("marking family: %#x from %d:%d offline\n", |
5073 | mpb->family_num, | |
a2b97981 DW |
5074 | super_table[i]->disks->major, |
5075 | super_table[i]->disks->minor); | |
5076 | super_table[i] = s; | |
5077 | } | |
5078 | ||
5079 | /* This is where the mdadm implementation differs from the Windows | |
5080 | * driver which has no strict concept of a container. We can only | |
5081 | * assemble one family from a container, so when returning a prodigal | |
5082 | * array member to this system the code will not be able to disambiguate | |
5083 | * the container contents that should be assembled ("foreign" versus | |
5084 | * "local"). It requires user intervention to set the orig_family_num | |
5085 | * to a new value to establish a new container. The Windows driver in | |
5086 | * this situation fixes up the volume name in place and manages the | |
5087 | * foreign array as an independent entity. | |
5088 | */ | |
5089 | s = NULL; | |
5090 | spare = NULL; | |
5091 | conflict = 0; | |
5092 | for (i = 0; i < tbl_size; i++) { | |
5093 | struct intel_super *tbl_ent = super_table[i]; | |
5094 | int is_spare = 0; | |
5095 | ||
5096 | if (!tbl_ent) | |
5097 | continue; | |
5098 | ||
5099 | if (tbl_ent->anchor->num_raid_devs == 0) { | |
5100 | spare = tbl_ent; | |
5101 | is_spare = 1; | |
5102 | } | |
5103 | ||
5104 | if (s && !is_spare) { | |
5105 | show_conflicts(tbl_ent->anchor->family_num, *super_list); | |
5106 | conflict++; | |
5107 | } else if (!s && !is_spare) | |
5108 | s = tbl_ent; | |
5109 | } | |
5110 | ||
5111 | if (!s) | |
5112 | s = spare; | |
5113 | if (!s) { | |
5114 | champion = NULL; | |
5115 | goto out; | |
5116 | } | |
5117 | champion = s; | |
5118 | ||
5119 | if (conflict) | |
7a862a02 | 5120 | pr_err("Chose family %#x on '%s', assemble conflicts to new container with '--update=uuid'\n", |
a2b97981 DW |
5121 | __le32_to_cpu(s->anchor->family_num), s->disks->devname); |
5122 | ||
5123 | /* collect all dl's onto 'champion', and update them to | |
5124 | * champion's version of the status | |
5125 | */ | |
5126 | for (s = *super_list; s; s = s->next) { | |
5127 | struct imsm_super *mpb = champion->anchor; | |
5128 | struct dl *dl = s->disks; | |
5129 | ||
5130 | if (s == champion) | |
5131 | continue; | |
5132 | ||
5d7b407a CA |
5133 | mpb->attributes |= s->anchor->attributes & MPB_ATTRIB_2TB_DISK; |
5134 | ||
a2b97981 DW |
5135 | for (i = 0; i < mpb->num_disks; i++) { |
5136 | struct imsm_disk *disk; | |
5137 | ||
5138 | disk = __serial_to_disk(dl->serial, mpb, &dl->index); | |
5139 | if (disk) { | |
5140 | dl->disk = *disk; | |
5141 | /* only set index on disks that are a member of | |
5142 | * a populated contianer, i.e. one with | |
5143 | * raid_devs | |
5144 | */ | |
5145 | if (is_failed(&dl->disk)) | |
5146 | dl->index = -2; | |
5147 | else if (is_spare(&dl->disk)) | |
5148 | dl->index = -1; | |
5149 | break; | |
5150 | } | |
5151 | } | |
5152 | ||
5153 | if (i >= mpb->num_disks) { | |
5154 | struct intel_disk *idisk; | |
5155 | ||
5156 | idisk = disk_list_get(dl->serial, disk_list); | |
ecf408e9 | 5157 | if (idisk && is_spare(&idisk->disk) && |
a2b97981 DW |
5158 | !is_failed(&idisk->disk) && !is_configured(&idisk->disk)) |
5159 | dl->index = -1; | |
5160 | else { | |
5161 | dl->index = -2; | |
5162 | continue; | |
5163 | } | |
5164 | } | |
5165 | ||
5166 | dl->next = champion->disks; | |
5167 | champion->disks = dl; | |
5168 | s->disks = NULL; | |
5169 | } | |
5170 | ||
5171 | /* delete 'champion' from super_list */ | |
5172 | for (del = super_list; *del; ) { | |
5173 | if (*del == champion) { | |
5174 | *del = (*del)->next; | |
5175 | break; | |
5176 | } else | |
5177 | del = &(*del)->next; | |
5178 | } | |
5179 | champion->next = NULL; | |
5180 | ||
5181 | out: | |
5182 | while (disk_list) { | |
5183 | struct intel_disk *idisk = disk_list; | |
5184 | ||
5185 | disk_list = disk_list->next; | |
5186 | free(idisk); | |
5187 | } | |
5188 | ||
5189 | return champion; | |
5190 | } | |
5191 | ||
9587c373 LM |
5192 | static int |
5193 | get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd); | |
4dd2df09 | 5194 | static int get_super_block(struct intel_super **super_list, char *devnm, char *devname, |
9587c373 | 5195 | int major, int minor, int keep_fd); |
ec50f7b6 LM |
5196 | static int |
5197 | get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list, | |
5198 | int *max, int keep_fd); | |
5199 | ||
cdddbdbc | 5200 | static int load_super_imsm_all(struct supertype *st, int fd, void **sbp, |
ec50f7b6 LM |
5201 | char *devname, struct md_list *devlist, |
5202 | int keep_fd) | |
cdddbdbc | 5203 | { |
a2b97981 DW |
5204 | struct intel_super *super_list = NULL; |
5205 | struct intel_super *super = NULL; | |
a2b97981 | 5206 | int err = 0; |
9587c373 | 5207 | int i = 0; |
dab4a513 | 5208 | |
4389ce73 | 5209 | if (is_fd_valid(fd)) |
9587c373 LM |
5210 | /* 'fd' is an opened container */ |
5211 | err = get_sra_super_block(fd, &super_list, devname, &i, keep_fd); | |
5212 | else | |
ec50f7b6 LM |
5213 | /* get super block from devlist devices */ |
5214 | err = get_devlist_super_block(devlist, &super_list, &i, keep_fd); | |
9587c373 | 5215 | if (err) |
1602d52c | 5216 | goto error; |
a2b97981 DW |
5217 | /* all mpbs enter, maybe one leaves */ |
5218 | super = imsm_thunderdome(&super_list, i); | |
5219 | if (!super) { | |
5220 | err = 1; | |
5221 | goto error; | |
cdddbdbc DW |
5222 | } |
5223 | ||
47ee5a45 DW |
5224 | if (find_missing(super) != 0) { |
5225 | free_imsm(super); | |
a2b97981 DW |
5226 | err = 2; |
5227 | goto error; | |
47ee5a45 | 5228 | } |
8e59f3d8 AK |
5229 | |
5230 | /* load migration record */ | |
2f86fda3 | 5231 | err = load_imsm_migr_rec(super); |
4c965cc9 AK |
5232 | if (err == -1) { |
5233 | /* migration is in progress, | |
5234 | * but migr_rec cannot be loaded, | |
5235 | */ | |
8e59f3d8 AK |
5236 | err = 4; |
5237 | goto error; | |
5238 | } | |
e2f41b2c AK |
5239 | |
5240 | /* Check migration compatibility */ | |
089f9d79 | 5241 | if (err == 0 && check_mpb_migr_compatibility(super) != 0) { |
e7b84f9d | 5242 | pr_err("Unsupported migration detected"); |
e2f41b2c AK |
5243 | if (devname) |
5244 | fprintf(stderr, " on %s\n", devname); | |
5245 | else | |
5246 | fprintf(stderr, " (IMSM).\n"); | |
5247 | ||
5248 | err = 5; | |
5249 | goto error; | |
5250 | } | |
5251 | ||
a2b97981 DW |
5252 | err = 0; |
5253 | ||
5254 | error: | |
5255 | while (super_list) { | |
5256 | struct intel_super *s = super_list; | |
5257 | ||
5258 | super_list = super_list->next; | |
5259 | free_imsm(s); | |
5260 | } | |
9587c373 | 5261 | |
a2b97981 DW |
5262 | if (err) |
5263 | return err; | |
f7e7067b | 5264 | |
cdddbdbc | 5265 | *sbp = super; |
4389ce73 | 5266 | if (is_fd_valid(fd)) |
4dd2df09 | 5267 | strcpy(st->container_devnm, fd2devnm(fd)); |
9587c373 | 5268 | else |
4dd2df09 | 5269 | st->container_devnm[0] = 0; |
a2b97981 | 5270 | if (err == 0 && st->ss == NULL) { |
bf5a934a | 5271 | st->ss = &super_imsm; |
cdddbdbc DW |
5272 | st->minor_version = 0; |
5273 | st->max_devs = IMSM_MAX_DEVICES; | |
5274 | } | |
cdddbdbc DW |
5275 | return 0; |
5276 | } | |
2b959fbf | 5277 | |
ec50f7b6 LM |
5278 | static int |
5279 | get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list, | |
5280 | int *max, int keep_fd) | |
5281 | { | |
5282 | struct md_list *tmpdev; | |
5283 | int err = 0; | |
5284 | int i = 0; | |
9587c373 | 5285 | |
ec50f7b6 LM |
5286 | for (i = 0, tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { |
5287 | if (tmpdev->used != 1) | |
5288 | continue; | |
5289 | if (tmpdev->container == 1) { | |
ca9de185 | 5290 | int lmax = 0; |
ec50f7b6 | 5291 | int fd = dev_open(tmpdev->devname, O_RDONLY|O_EXCL); |
4389ce73 | 5292 | if (!is_fd_valid(fd)) { |
e7b84f9d | 5293 | pr_err("cannot open device %s: %s\n", |
ec50f7b6 LM |
5294 | tmpdev->devname, strerror(errno)); |
5295 | err = 8; | |
5296 | goto error; | |
5297 | } | |
5298 | err = get_sra_super_block(fd, super_list, | |
5299 | tmpdev->devname, &lmax, | |
5300 | keep_fd); | |
5301 | i += lmax; | |
5302 | close(fd); | |
5303 | if (err) { | |
5304 | err = 7; | |
5305 | goto error; | |
5306 | } | |
5307 | } else { | |
5308 | int major = major(tmpdev->st_rdev); | |
5309 | int minor = minor(tmpdev->st_rdev); | |
5310 | err = get_super_block(super_list, | |
4dd2df09 | 5311 | NULL, |
ec50f7b6 LM |
5312 | tmpdev->devname, |
5313 | major, minor, | |
5314 | keep_fd); | |
5315 | i++; | |
5316 | if (err) { | |
5317 | err = 6; | |
5318 | goto error; | |
5319 | } | |
5320 | } | |
5321 | } | |
5322 | error: | |
5323 | *max = i; | |
5324 | return err; | |
5325 | } | |
9587c373 | 5326 | |
4dd2df09 | 5327 | static int get_super_block(struct intel_super **super_list, char *devnm, char *devname, |
9587c373 LM |
5328 | int major, int minor, int keep_fd) |
5329 | { | |
594dc1b8 | 5330 | struct intel_super *s; |
9587c373 LM |
5331 | char nm[32]; |
5332 | int dfd = -1; | |
9587c373 LM |
5333 | int err = 0; |
5334 | int retry; | |
5335 | ||
5336 | s = alloc_super(); | |
5337 | if (!s) { | |
5338 | err = 1; | |
5339 | goto error; | |
5340 | } | |
5341 | ||
5342 | sprintf(nm, "%d:%d", major, minor); | |
5343 | dfd = dev_open(nm, O_RDWR); | |
4389ce73 | 5344 | if (!is_fd_valid(dfd)) { |
9587c373 LM |
5345 | err = 2; |
5346 | goto error; | |
5347 | } | |
5348 | ||
aec01630 JS |
5349 | if (!get_dev_sector_size(dfd, NULL, &s->sector_size)) { |
5350 | err = 2; | |
5351 | goto error; | |
5352 | } | |
cb8f6859 | 5353 | find_intel_hba_capability(dfd, s, devname); |
9587c373 LM |
5354 | err = load_and_parse_mpb(dfd, s, NULL, keep_fd); |
5355 | ||
5356 | /* retry the load if we might have raced against mdmon */ | |
4dd2df09 | 5357 | if (err == 3 && devnm && mdmon_running(devnm)) |
9587c373 | 5358 | for (retry = 0; retry < 3; retry++) { |
239b3cc0 | 5359 | sleep_for(0, MSEC_TO_NSEC(3), true); |
9587c373 LM |
5360 | err = load_and_parse_mpb(dfd, s, NULL, keep_fd); |
5361 | if (err != 3) | |
5362 | break; | |
5363 | } | |
5364 | error: | |
5365 | if (!err) { | |
5366 | s->next = *super_list; | |
5367 | *super_list = s; | |
5368 | } else { | |
5369 | if (s) | |
8d67477f | 5370 | free_imsm(s); |
4389ce73 | 5371 | close_fd(&dfd); |
9587c373 | 5372 | } |
4389ce73 MT |
5373 | if (!keep_fd) |
5374 | close_fd(&dfd); | |
9587c373 LM |
5375 | return err; |
5376 | ||
5377 | } | |
5378 | ||
5379 | static int | |
5380 | get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd) | |
5381 | { | |
5382 | struct mdinfo *sra; | |
4dd2df09 | 5383 | char *devnm; |
9587c373 LM |
5384 | struct mdinfo *sd; |
5385 | int err = 0; | |
5386 | int i = 0; | |
4dd2df09 | 5387 | sra = sysfs_read(fd, NULL, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE); |
9587c373 LM |
5388 | if (!sra) |
5389 | return 1; | |
5390 | ||
5391 | if (sra->array.major_version != -1 || | |
5392 | sra->array.minor_version != -2 || | |
5393 | strcmp(sra->text_version, "imsm") != 0) { | |
5394 | err = 1; | |
5395 | goto error; | |
5396 | } | |
5397 | /* load all mpbs */ | |
4dd2df09 | 5398 | devnm = fd2devnm(fd); |
9587c373 | 5399 | for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) { |
4dd2df09 | 5400 | if (get_super_block(super_list, devnm, devname, |
9587c373 LM |
5401 | sd->disk.major, sd->disk.minor, keep_fd) != 0) { |
5402 | err = 7; | |
5403 | goto error; | |
5404 | } | |
5405 | } | |
5406 | error: | |
5407 | sysfs_free(sra); | |
5408 | *max = i; | |
5409 | return err; | |
5410 | } | |
5411 | ||
2b959fbf N |
5412 | static int load_container_imsm(struct supertype *st, int fd, char *devname) |
5413 | { | |
ec50f7b6 | 5414 | return load_super_imsm_all(st, fd, &st->sb, devname, NULL, 1); |
2b959fbf | 5415 | } |
cdddbdbc DW |
5416 | |
5417 | static int load_super_imsm(struct supertype *st, int fd, char *devname) | |
5418 | { | |
5419 | struct intel_super *super; | |
5420 | int rv; | |
8a3544f8 | 5421 | int retry; |
cdddbdbc | 5422 | |
357ac106 | 5423 | if (test_partition(fd)) |
691c6ee1 N |
5424 | /* IMSM not allowed on partitions */ |
5425 | return 1; | |
5426 | ||
37424f13 DW |
5427 | free_super_imsm(st); |
5428 | ||
49133e57 | 5429 | super = alloc_super(); |
8d67477f TM |
5430 | if (!super) |
5431 | return 1; | |
3a85bf0e MG |
5432 | |
5433 | if (!get_dev_sector_size(fd, NULL, &super->sector_size)) { | |
5434 | free_imsm(super); | |
5435 | return 1; | |
5436 | } | |
ea2bc72b LM |
5437 | /* Load hba and capabilities if they exist. |
5438 | * But do not preclude loading metadata in case capabilities or hba are | |
5439 | * non-compliant and ignore_hw_compat is set. | |
5440 | */ | |
d424212e | 5441 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 5442 | /* no orom/efi or non-intel hba of the disk */ |
089f9d79 | 5443 | if (rv != 0 && st->ignore_hw_compat == 0) { |
f2f5c343 | 5444 | if (devname) |
e7b84f9d | 5445 | pr_err("No OROM/EFI properties for %s\n", devname); |
f2f5c343 LM |
5446 | free_imsm(super); |
5447 | return 2; | |
5448 | } | |
a2b97981 | 5449 | rv = load_and_parse_mpb(fd, super, devname, 0); |
cdddbdbc | 5450 | |
8a3544f8 AP |
5451 | /* retry the load if we might have raced against mdmon */ |
5452 | if (rv == 3) { | |
f96b1302 AP |
5453 | struct mdstat_ent *mdstat = NULL; |
5454 | char *name = fd2kname(fd); | |
5455 | ||
5456 | if (name) | |
5457 | mdstat = mdstat_by_component(name); | |
8a3544f8 AP |
5458 | |
5459 | if (mdstat && mdmon_running(mdstat->devnm) && getpid() != mdmon_pid(mdstat->devnm)) { | |
5460 | for (retry = 0; retry < 3; retry++) { | |
239b3cc0 | 5461 | sleep_for(0, MSEC_TO_NSEC(3), true); |
8a3544f8 AP |
5462 | rv = load_and_parse_mpb(fd, super, devname, 0); |
5463 | if (rv != 3) | |
5464 | break; | |
5465 | } | |
5466 | } | |
5467 | ||
5468 | free_mdstat(mdstat); | |
5469 | } | |
5470 | ||
cdddbdbc DW |
5471 | if (rv) { |
5472 | if (devname) | |
7a862a02 | 5473 | pr_err("Failed to load all information sections on %s\n", devname); |
cdddbdbc DW |
5474 | free_imsm(super); |
5475 | return rv; | |
5476 | } | |
5477 | ||
5478 | st->sb = super; | |
5479 | if (st->ss == NULL) { | |
5480 | st->ss = &super_imsm; | |
5481 | st->minor_version = 0; | |
5482 | st->max_devs = IMSM_MAX_DEVICES; | |
5483 | } | |
8e59f3d8 AK |
5484 | |
5485 | /* load migration record */ | |
2f86fda3 | 5486 | if (load_imsm_migr_rec(super) == 0) { |
2e062e82 AK |
5487 | /* Check for unsupported migration features */ |
5488 | if (check_mpb_migr_compatibility(super) != 0) { | |
e7b84f9d | 5489 | pr_err("Unsupported migration detected"); |
2e062e82 AK |
5490 | if (devname) |
5491 | fprintf(stderr, " on %s\n", devname); | |
5492 | else | |
5493 | fprintf(stderr, " (IMSM).\n"); | |
5494 | return 3; | |
5495 | } | |
e2f41b2c AK |
5496 | } |
5497 | ||
cdddbdbc DW |
5498 | return 0; |
5499 | } | |
5500 | ||
ef6ffade DW |
5501 | static __u16 info_to_blocks_per_strip(mdu_array_info_t *info) |
5502 | { | |
5503 | if (info->level == 1) | |
5504 | return 128; | |
5505 | return info->chunk_size >> 9; | |
5506 | } | |
5507 | ||
5551b113 CA |
5508 | static unsigned long long info_to_blocks_per_member(mdu_array_info_t *info, |
5509 | unsigned long long size) | |
fcfd9599 | 5510 | { |
4025c288 | 5511 | if (info->level == 1) |
5551b113 | 5512 | return size * 2; |
4025c288 | 5513 | else |
5551b113 | 5514 | return (size * 2) & ~(info_to_blocks_per_strip(info) - 1); |
fcfd9599 DW |
5515 | } |
5516 | ||
4d1313e9 DW |
5517 | static void imsm_update_version_info(struct intel_super *super) |
5518 | { | |
5519 | /* update the version and attributes */ | |
5520 | struct imsm_super *mpb = super->anchor; | |
5521 | char *version; | |
5522 | struct imsm_dev *dev; | |
5523 | struct imsm_map *map; | |
5524 | int i; | |
5525 | ||
5526 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
5527 | dev = get_imsm_dev(super, i); | |
238c0a71 | 5528 | map = get_imsm_map(dev, MAP_0); |
4d1313e9 DW |
5529 | if (__le32_to_cpu(dev->size_high) > 0) |
5530 | mpb->attributes |= MPB_ATTRIB_2TB; | |
5531 | ||
5532 | /* FIXME detect when an array spans a port multiplier */ | |
5533 | #if 0 | |
5534 | mpb->attributes |= MPB_ATTRIB_PM; | |
5535 | #endif | |
5536 | ||
5537 | if (mpb->num_raid_devs > 1 || | |
5538 | mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) { | |
5539 | version = MPB_VERSION_ATTRIBS; | |
5540 | switch (get_imsm_raid_level(map)) { | |
5541 | case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break; | |
5542 | case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break; | |
5543 | case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break; | |
5544 | case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break; | |
5545 | } | |
5546 | } else { | |
5547 | if (map->num_members >= 5) | |
5548 | version = MPB_VERSION_5OR6_DISK_ARRAY; | |
5549 | else if (dev->status == DEV_CLONE_N_GO) | |
5550 | version = MPB_VERSION_CNG; | |
5551 | else if (get_imsm_raid_level(map) == 5) | |
5552 | version = MPB_VERSION_RAID5; | |
5553 | else if (map->num_members >= 3) | |
5554 | version = MPB_VERSION_3OR4_DISK_ARRAY; | |
5555 | else if (get_imsm_raid_level(map) == 1) | |
5556 | version = MPB_VERSION_RAID1; | |
5557 | else | |
5558 | version = MPB_VERSION_RAID0; | |
5559 | } | |
5560 | strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version); | |
5561 | } | |
5562 | } | |
5563 | ||
aa534678 DW |
5564 | static int check_name(struct intel_super *super, char *name, int quiet) |
5565 | { | |
5566 | struct imsm_super *mpb = super->anchor; | |
5567 | char *reason = NULL; | |
9bd99a90 RS |
5568 | char *start = name; |
5569 | size_t len = strlen(name); | |
aa534678 DW |
5570 | int i; |
5571 | ||
9bd99a90 RS |
5572 | if (len > 0) { |
5573 | while (isspace(start[len - 1])) | |
5574 | start[--len] = 0; | |
5575 | while (*start && isspace(*start)) | |
5576 | ++start, --len; | |
5577 | memmove(name, start, len + 1); | |
5578 | } | |
5579 | ||
5580 | if (len > MAX_RAID_SERIAL_LEN) | |
aa534678 | 5581 | reason = "must be 16 characters or less"; |
9bd99a90 RS |
5582 | else if (len == 0) |
5583 | reason = "must be a non-empty string"; | |
aa534678 DW |
5584 | |
5585 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
5586 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
5587 | ||
5588 | if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) { | |
5589 | reason = "already exists"; | |
5590 | break; | |
5591 | } | |
5592 | } | |
5593 | ||
5594 | if (reason && !quiet) | |
e7b84f9d | 5595 | pr_err("imsm volume name %s\n", reason); |
aa534678 DW |
5596 | |
5597 | return !reason; | |
5598 | } | |
5599 | ||
8b353278 | 5600 | static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info, |
5308f117 | 5601 | struct shape *s, char *name, |
83cd1e97 N |
5602 | char *homehost, int *uuid, |
5603 | long long data_offset) | |
cdddbdbc | 5604 | { |
c2c087e6 DW |
5605 | /* We are creating a volume inside a pre-existing container. |
5606 | * so st->sb is already set. | |
5607 | */ | |
5608 | struct intel_super *super = st->sb; | |
f36a9ecd | 5609 | unsigned int sector_size = super->sector_size; |
949c47a0 | 5610 | struct imsm_super *mpb = super->anchor; |
ba2de7ba | 5611 | struct intel_dev *dv; |
c2c087e6 DW |
5612 | struct imsm_dev *dev; |
5613 | struct imsm_vol *vol; | |
5614 | struct imsm_map *map; | |
5615 | int idx = mpb->num_raid_devs; | |
5616 | int i; | |
760365f9 | 5617 | int namelen; |
c2c087e6 | 5618 | unsigned long long array_blocks; |
2c092cad | 5619 | size_t size_old, size_new; |
b53bfba6 TM |
5620 | unsigned int data_disks; |
5621 | unsigned long long size_per_member; | |
cdddbdbc | 5622 | |
88c32bb1 | 5623 | if (super->orom && mpb->num_raid_devs >= super->orom->vpa) { |
7a862a02 | 5624 | pr_err("This imsm-container already has the maximum of %d volumes\n", super->orom->vpa); |
c2c087e6 DW |
5625 | return 0; |
5626 | } | |
5627 | ||
2c092cad DW |
5628 | /* ensure the mpb is large enough for the new data */ |
5629 | size_old = __le32_to_cpu(mpb->mpb_size); | |
5630 | size_new = disks_to_mpb_size(info->nr_disks); | |
5631 | if (size_new > size_old) { | |
5632 | void *mpb_new; | |
f36a9ecd | 5633 | size_t size_round = ROUND_UP(size_new, sector_size); |
2c092cad | 5634 | |
f36a9ecd | 5635 | if (posix_memalign(&mpb_new, sector_size, size_round) != 0) { |
e7b84f9d | 5636 | pr_err("could not allocate new mpb\n"); |
2c092cad DW |
5637 | return 0; |
5638 | } | |
85337573 AO |
5639 | if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE, |
5640 | MIGR_REC_BUF_SECTORS* | |
5641 | MAX_SECTOR_SIZE) != 0) { | |
1ade5cc1 | 5642 | pr_err("could not allocate migr_rec buffer\n"); |
8e59f3d8 AK |
5643 | free(super->buf); |
5644 | free(super); | |
ea944c8f | 5645 | free(mpb_new); |
8e59f3d8 AK |
5646 | return 0; |
5647 | } | |
2c092cad DW |
5648 | memcpy(mpb_new, mpb, size_old); |
5649 | free(mpb); | |
5650 | mpb = mpb_new; | |
949c47a0 | 5651 | super->anchor = mpb_new; |
2c092cad DW |
5652 | mpb->mpb_size = __cpu_to_le32(size_new); |
5653 | memset(mpb_new + size_old, 0, size_round - size_old); | |
bbab0940 | 5654 | super->len = size_round; |
2c092cad | 5655 | } |
bf5a934a | 5656 | super->current_vol = idx; |
3960e579 DW |
5657 | |
5658 | /* handle 'failed_disks' by either: | |
5659 | * a) create dummy disk entries in the table if this the first | |
5660 | * volume in the array. We add them here as this is the only | |
5661 | * opportunity to add them. add_to_super_imsm_volume() | |
5662 | * handles the non-failed disks and continues incrementing | |
5663 | * mpb->num_disks. | |
5664 | * b) validate that 'failed_disks' matches the current number | |
5665 | * of missing disks if the container is populated | |
d23fe947 | 5666 | */ |
3960e579 | 5667 | if (super->current_vol == 0) { |
d23fe947 | 5668 | mpb->num_disks = 0; |
3960e579 DW |
5669 | for (i = 0; i < info->failed_disks; i++) { |
5670 | struct imsm_disk *disk; | |
5671 | ||
5672 | mpb->num_disks++; | |
5673 | disk = __get_imsm_disk(mpb, i); | |
5674 | disk->status = CONFIGURED_DISK | FAILED_DISK; | |
5675 | disk->scsi_id = __cpu_to_le32(~(__u32)0); | |
5676 | snprintf((char *) disk->serial, MAX_RAID_SERIAL_LEN, | |
5b975129 | 5677 | "missing:%d", (__u8)i); |
3960e579 DW |
5678 | } |
5679 | find_missing(super); | |
5680 | } else { | |
5681 | int missing = 0; | |
5682 | struct dl *d; | |
5683 | ||
5684 | for (d = super->missing; d; d = d->next) | |
5685 | missing++; | |
5686 | if (info->failed_disks > missing) { | |
e7b84f9d | 5687 | pr_err("unable to add 'missing' disk to container\n"); |
3960e579 DW |
5688 | return 0; |
5689 | } | |
5690 | } | |
5a038140 | 5691 | |
aa534678 DW |
5692 | if (!check_name(super, name, 0)) |
5693 | return 0; | |
503975b9 N |
5694 | dv = xmalloc(sizeof(*dv)); |
5695 | dev = xcalloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1)); | |
760365f9 JS |
5696 | /* |
5697 | * Explicitly allow truncating to not confuse gcc's | |
5698 | * -Werror=stringop-truncation | |
5699 | */ | |
5700 | namelen = min((int) strlen(name), MAX_RAID_SERIAL_LEN); | |
5701 | memcpy(dev->volume, name, namelen); | |
e03640bd | 5702 | array_blocks = calc_array_size(info->level, info->raid_disks, |
03bcbc65 | 5703 | info->layout, info->chunk_size, |
b53bfba6 TM |
5704 | s->size * BLOCKS_PER_KB); |
5705 | data_disks = get_data_disks(info->level, info->layout, | |
5706 | info->raid_disks); | |
5707 | array_blocks = round_size_to_mb(array_blocks, data_disks); | |
5708 | size_per_member = array_blocks / data_disks; | |
979d38be | 5709 | |
fcc2c9da | 5710 | set_imsm_dev_size(dev, array_blocks); |
1a2487c2 | 5711 | dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING); |
c2c087e6 DW |
5712 | vol = &dev->vol; |
5713 | vol->migr_state = 0; | |
1484e727 | 5714 | set_migr_type(dev, MIGR_INIT); |
3960e579 | 5715 | vol->dirty = !info->state; |
4036e7ee | 5716 | set_vol_curr_migr_unit(dev, 0); |
238c0a71 | 5717 | map = get_imsm_map(dev, MAP_0); |
5551b113 | 5718 | set_pba_of_lba0(map, super->create_offset); |
ef6ffade | 5719 | map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info)); |
0556e1a2 | 5720 | map->failed_disk_num = ~0; |
bf4442ab | 5721 | if (info->level > 0) |
fffaf1ff N |
5722 | map->map_state = (info->state ? IMSM_T_STATE_NORMAL |
5723 | : IMSM_T_STATE_UNINITIALIZED); | |
bf4442ab AK |
5724 | else |
5725 | map->map_state = info->failed_disks ? IMSM_T_STATE_FAILED : | |
5726 | IMSM_T_STATE_NORMAL; | |
252d23c0 | 5727 | map->ddf = 1; |
ef6ffade DW |
5728 | |
5729 | if (info->level == 1 && info->raid_disks > 2) { | |
38950822 AW |
5730 | free(dev); |
5731 | free(dv); | |
7a862a02 | 5732 | pr_err("imsm does not support more than 2 disksin a raid1 volume\n"); |
ef6ffade DW |
5733 | return 0; |
5734 | } | |
81062a36 DW |
5735 | |
5736 | map->raid_level = info->level; | |
1c275381 | 5737 | if (info->level == 10) |
c2c087e6 | 5738 | map->raid_level = 1; |
1c275381 | 5739 | set_num_domains(map); |
ef6ffade | 5740 | |
44490938 MD |
5741 | size_per_member += NUM_BLOCKS_DIRTY_STRIPE_REGION; |
5742 | set_blocks_per_member(map, info_to_blocks_per_member(info, | |
5743 | size_per_member / | |
5744 | BLOCKS_PER_KB)); | |
5745 | ||
c2c087e6 | 5746 | map->num_members = info->raid_disks; |
1c275381 | 5747 | update_num_data_stripes(map, array_blocks); |
c2c087e6 DW |
5748 | for (i = 0; i < map->num_members; i++) { |
5749 | /* initialized in add_to_super */ | |
4eb26970 | 5750 | set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD); |
c2c087e6 | 5751 | } |
949c47a0 | 5752 | mpb->num_raid_devs++; |
2a24dc1b PB |
5753 | mpb->num_raid_devs_created++; |
5754 | dev->my_vol_raid_dev_num = mpb->num_raid_devs_created; | |
ba2de7ba | 5755 | |
b7580566 | 5756 | if (s->consistency_policy <= CONSISTENCY_POLICY_RESYNC) { |
c2462068 | 5757 | dev->rwh_policy = RWH_MULTIPLE_OFF; |
2432ce9b | 5758 | } else if (s->consistency_policy == CONSISTENCY_POLICY_PPL) { |
c2462068 | 5759 | dev->rwh_policy = RWH_MULTIPLE_DISTRIBUTED; |
2432ce9b AP |
5760 | } else { |
5761 | free(dev); | |
5762 | free(dv); | |
5763 | pr_err("imsm does not support consistency policy %s\n", | |
5f21d674 | 5764 | map_num_s(consistency_policies, s->consistency_policy)); |
2432ce9b AP |
5765 | return 0; |
5766 | } | |
5767 | ||
ba2de7ba DW |
5768 | dv->dev = dev; |
5769 | dv->index = super->current_vol; | |
5770 | dv->next = super->devlist; | |
5771 | super->devlist = dv; | |
c2c087e6 | 5772 | |
4d1313e9 DW |
5773 | imsm_update_version_info(super); |
5774 | ||
c2c087e6 | 5775 | return 1; |
cdddbdbc DW |
5776 | } |
5777 | ||
bf5a934a | 5778 | static int init_super_imsm(struct supertype *st, mdu_array_info_t *info, |
5308f117 | 5779 | struct shape *s, char *name, |
83cd1e97 N |
5780 | char *homehost, int *uuid, |
5781 | unsigned long long data_offset) | |
bf5a934a DW |
5782 | { |
5783 | /* This is primarily called by Create when creating a new array. | |
5784 | * We will then get add_to_super called for each component, and then | |
5785 | * write_init_super called to write it out to each device. | |
5786 | * For IMSM, Create can create on fresh devices or on a pre-existing | |
5787 | * array. | |
5788 | * To create on a pre-existing array a different method will be called. | |
5789 | * This one is just for fresh drives. | |
5790 | */ | |
5791 | struct intel_super *super; | |
5792 | struct imsm_super *mpb; | |
5793 | size_t mpb_size; | |
4d1313e9 | 5794 | char *version; |
bf5a934a | 5795 | |
83cd1e97 | 5796 | if (data_offset != INVALID_SECTORS) { |
ed503f89 | 5797 | pr_err("data-offset not supported by imsm\n"); |
83cd1e97 N |
5798 | return 0; |
5799 | } | |
5800 | ||
bf5a934a | 5801 | if (st->sb) |
5308f117 | 5802 | return init_super_imsm_volume(st, info, s, name, homehost, uuid, |
83cd1e97 | 5803 | data_offset); |
e683ca88 DW |
5804 | |
5805 | if (info) | |
5806 | mpb_size = disks_to_mpb_size(info->nr_disks); | |
5807 | else | |
f36a9ecd | 5808 | mpb_size = MAX_SECTOR_SIZE; |
bf5a934a | 5809 | |
49133e57 | 5810 | super = alloc_super(); |
f36a9ecd PB |
5811 | if (super && |
5812 | posix_memalign(&super->buf, MAX_SECTOR_SIZE, mpb_size) != 0) { | |
8d67477f | 5813 | free_imsm(super); |
e683ca88 DW |
5814 | super = NULL; |
5815 | } | |
5816 | if (!super) { | |
1ade5cc1 | 5817 | pr_err("could not allocate superblock\n"); |
bf5a934a DW |
5818 | return 0; |
5819 | } | |
de44e46f PB |
5820 | if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE, |
5821 | MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) { | |
1ade5cc1 | 5822 | pr_err("could not allocate migr_rec buffer\n"); |
8e59f3d8 | 5823 | free(super->buf); |
8d67477f | 5824 | free_imsm(super); |
8e59f3d8 AK |
5825 | return 0; |
5826 | } | |
e683ca88 | 5827 | memset(super->buf, 0, mpb_size); |
ef649044 | 5828 | mpb = super->buf; |
e683ca88 DW |
5829 | mpb->mpb_size = __cpu_to_le32(mpb_size); |
5830 | st->sb = super; | |
5831 | ||
5832 | if (info == NULL) { | |
5833 | /* zeroing superblock */ | |
5834 | return 0; | |
5835 | } | |
bf5a934a | 5836 | |
4d1313e9 DW |
5837 | mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; |
5838 | ||
5839 | version = (char *) mpb->sig; | |
5840 | strcpy(version, MPB_SIGNATURE); | |
5841 | version += strlen(MPB_SIGNATURE); | |
5842 | strcpy(version, MPB_VERSION_RAID0); | |
bf5a934a | 5843 | |
bf5a934a DW |
5844 | return 1; |
5845 | } | |
5846 | ||
f2cc4f7d AO |
5847 | static int drive_validate_sector_size(struct intel_super *super, struct dl *dl) |
5848 | { | |
5849 | unsigned int member_sector_size; | |
5850 | ||
4389ce73 | 5851 | if (!is_fd_valid(dl->fd)) { |
f2cc4f7d AO |
5852 | pr_err("Invalid file descriptor for %s\n", dl->devname); |
5853 | return 0; | |
5854 | } | |
5855 | ||
5856 | if (!get_dev_sector_size(dl->fd, dl->devname, &member_sector_size)) | |
5857 | return 0; | |
5858 | if (member_sector_size != super->sector_size) | |
5859 | return 0; | |
5860 | return 1; | |
5861 | } | |
5862 | ||
f20c3968 | 5863 | static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk, |
bf5a934a DW |
5864 | int fd, char *devname) |
5865 | { | |
5866 | struct intel_super *super = st->sb; | |
d23fe947 | 5867 | struct imsm_super *mpb = super->anchor; |
3960e579 | 5868 | struct imsm_disk *_disk; |
bf5a934a DW |
5869 | struct imsm_dev *dev; |
5870 | struct imsm_map *map; | |
3960e579 | 5871 | struct dl *dl, *df; |
4eb26970 | 5872 | int slot; |
9a7df595 MT |
5873 | int autolayout = 0; |
5874 | ||
5875 | if (!is_fd_valid(fd)) | |
5876 | autolayout = 1; | |
bf5a934a | 5877 | |
949c47a0 | 5878 | dev = get_imsm_dev(super, super->current_vol); |
238c0a71 | 5879 | map = get_imsm_map(dev, MAP_0); |
bf5a934a | 5880 | |
208933a7 | 5881 | if (! (dk->state & (1<<MD_DISK_SYNC))) { |
e7b84f9d | 5882 | pr_err("%s: Cannot add spare devices to IMSM volume\n", |
208933a7 N |
5883 | devname); |
5884 | return 1; | |
5885 | } | |
5886 | ||
9a7df595 MT |
5887 | for (dl = super->disks; dl ; dl = dl->next) { |
5888 | if (autolayout) { | |
efb30e7f DW |
5889 | if (dl->raiddisk == dk->raid_disk) |
5890 | break; | |
9a7df595 MT |
5891 | } else if (dl->major == dk->major && dl->minor == dk->minor) |
5892 | break; | |
efb30e7f | 5893 | } |
d23fe947 | 5894 | |
208933a7 | 5895 | if (!dl) { |
9a7df595 MT |
5896 | if (!autolayout) |
5897 | pr_err("%s is not a member of the same container.\n", | |
5898 | devname); | |
f20c3968 | 5899 | return 1; |
208933a7 | 5900 | } |
bf5a934a | 5901 | |
9a7df595 MT |
5902 | if (!autolayout && super->current_vol > 0) { |
5903 | int _slot = get_disk_slot_in_dev(super, 0, dl->index); | |
5904 | ||
5905 | if (_slot != dk->raid_disk) { | |
5906 | pr_err("Member %s is in %d slot for the first volume, but is in %d slot for a new volume.\n", | |
5907 | dl->devname, _slot, dk->raid_disk); | |
5908 | pr_err("Raid members are in different order than for the first volume, aborting.\n"); | |
5909 | return 1; | |
5910 | } | |
5911 | } | |
5912 | ||
59632db9 MZ |
5913 | if (mpb->num_disks == 0) |
5914 | if (!get_dev_sector_size(dl->fd, dl->devname, | |
5915 | &super->sector_size)) | |
5916 | return 1; | |
5917 | ||
f2cc4f7d AO |
5918 | if (!drive_validate_sector_size(super, dl)) { |
5919 | pr_err("Combining drives of different sector size in one volume is not allowed\n"); | |
5920 | return 1; | |
5921 | } | |
5922 | ||
d23fe947 DW |
5923 | /* add a pristine spare to the metadata */ |
5924 | if (dl->index < 0) { | |
5925 | dl->index = super->anchor->num_disks; | |
5926 | super->anchor->num_disks++; | |
5927 | } | |
4eb26970 DW |
5928 | /* Check the device has not already been added */ |
5929 | slot = get_imsm_disk_slot(map, dl->index); | |
5930 | if (slot >= 0 && | |
238c0a71 | 5931 | (get_imsm_ord_tbl_ent(dev, slot, MAP_X) & IMSM_ORD_REBUILD) == 0) { |
e7b84f9d | 5932 | pr_err("%s has been included in this array twice\n", |
4eb26970 DW |
5933 | devname); |
5934 | return 1; | |
5935 | } | |
656b6b5a | 5936 | set_imsm_ord_tbl_ent(map, dk->raid_disk, dl->index); |
ee5aad5a | 5937 | dl->disk.status = CONFIGURED_DISK; |
d23fe947 | 5938 | |
3960e579 DW |
5939 | /* update size of 'missing' disks to be at least as large as the |
5940 | * largest acitve member (we only have dummy missing disks when | |
5941 | * creating the first volume) | |
5942 | */ | |
5943 | if (super->current_vol == 0) { | |
5944 | for (df = super->missing; df; df = df->next) { | |
5551b113 CA |
5945 | if (total_blocks(&dl->disk) > total_blocks(&df->disk)) |
5946 | set_total_blocks(&df->disk, total_blocks(&dl->disk)); | |
3960e579 DW |
5947 | _disk = __get_imsm_disk(mpb, df->index); |
5948 | *_disk = df->disk; | |
5949 | } | |
5950 | } | |
5951 | ||
5952 | /* refresh unset/failed slots to point to valid 'missing' entries */ | |
5953 | for (df = super->missing; df; df = df->next) | |
5954 | for (slot = 0; slot < mpb->num_disks; slot++) { | |
238c0a71 | 5955 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); |
3960e579 DW |
5956 | |
5957 | if ((ord & IMSM_ORD_REBUILD) == 0) | |
5958 | continue; | |
5959 | set_imsm_ord_tbl_ent(map, slot, df->index | IMSM_ORD_REBUILD); | |
1ace8403 | 5960 | if (is_gen_migration(dev)) { |
238c0a71 AK |
5961 | struct imsm_map *map2 = get_imsm_map(dev, |
5962 | MAP_1); | |
0a108d63 | 5963 | int slot2 = get_imsm_disk_slot(map2, df->index); |
089f9d79 | 5964 | if (slot2 < map2->num_members && slot2 >= 0) { |
1ace8403 | 5965 | __u32 ord2 = get_imsm_ord_tbl_ent(dev, |
238c0a71 AK |
5966 | slot2, |
5967 | MAP_1); | |
1ace8403 AK |
5968 | if ((unsigned)df->index == |
5969 | ord_to_idx(ord2)) | |
5970 | set_imsm_ord_tbl_ent(map2, | |
0a108d63 | 5971 | slot2, |
1ace8403 AK |
5972 | df->index | |
5973 | IMSM_ORD_REBUILD); | |
5974 | } | |
5975 | } | |
3960e579 DW |
5976 | dprintf("set slot:%d to missing disk:%d\n", slot, df->index); |
5977 | break; | |
5978 | } | |
5979 | ||
d23fe947 DW |
5980 | /* if we are creating the first raid device update the family number */ |
5981 | if (super->current_vol == 0) { | |
5982 | __u32 sum; | |
5983 | struct imsm_dev *_dev = __get_imsm_dev(mpb, 0); | |
d23fe947 | 5984 | |
3960e579 | 5985 | _disk = __get_imsm_disk(mpb, dl->index); |
756a15f3 | 5986 | if (!_disk) { |
e7b84f9d | 5987 | pr_err("BUG mpb setup error\n"); |
791b666a AW |
5988 | return 1; |
5989 | } | |
d23fe947 DW |
5990 | *_dev = *dev; |
5991 | *_disk = dl->disk; | |
148acb7b DW |
5992 | sum = random32(); |
5993 | sum += __gen_imsm_checksum(mpb); | |
d23fe947 | 5994 | mpb->family_num = __cpu_to_le32(sum); |
148acb7b | 5995 | mpb->orig_family_num = mpb->family_num; |
e48aed3c | 5996 | mpb->creation_time = __cpu_to_le64((__u64)time(NULL)); |
d23fe947 | 5997 | } |
ca0748fa | 5998 | super->current_disk = dl; |
f20c3968 | 5999 | return 0; |
bf5a934a DW |
6000 | } |
6001 | ||
a8619d23 AK |
6002 | /* mark_spare() |
6003 | * Function marks disk as spare and restores disk serial | |
6004 | * in case it was previously marked as failed by takeover operation | |
6005 | * reruns: | |
6006 | * -1 : critical error | |
6007 | * 0 : disk is marked as spare but serial is not set | |
6008 | * 1 : success | |
6009 | */ | |
6010 | int mark_spare(struct dl *disk) | |
6011 | { | |
6012 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
6013 | int ret_val = -1; | |
6014 | ||
6015 | if (!disk) | |
6016 | return ret_val; | |
6017 | ||
6018 | ret_val = 0; | |
6da53c0e | 6019 | if (!imsm_read_serial(disk->fd, NULL, serial, MAX_RAID_SERIAL_LEN)) { |
a8619d23 AK |
6020 | /* Restore disk serial number, because takeover marks disk |
6021 | * as failed and adds to serial ':0' before it becomes | |
6022 | * a spare disk. | |
6023 | */ | |
6024 | serialcpy(disk->serial, serial); | |
6025 | serialcpy(disk->disk.serial, serial); | |
6026 | ret_val = 1; | |
6027 | } | |
6028 | disk->disk.status = SPARE_DISK; | |
6029 | disk->index = -1; | |
6030 | ||
6031 | return ret_val; | |
6032 | } | |
88654014 | 6033 | |
12724c01 TM |
6034 | |
6035 | static int write_super_imsm_spare(struct intel_super *super, struct dl *d); | |
6036 | ||
f20c3968 | 6037 | static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk, |
72ca9bcf N |
6038 | int fd, char *devname, |
6039 | unsigned long long data_offset) | |
cdddbdbc | 6040 | { |
c2c087e6 | 6041 | struct intel_super *super = st->sb; |
c2c087e6 DW |
6042 | struct dl *dd; |
6043 | unsigned long long size; | |
fa7bb6f8 | 6044 | unsigned int member_sector_size; |
f2f27e63 | 6045 | __u32 id; |
c2c087e6 DW |
6046 | int rv; |
6047 | struct stat stb; | |
6048 | ||
88654014 LM |
6049 | /* If we are on an RAID enabled platform check that the disk is |
6050 | * attached to the raid controller. | |
6051 | * We do not need to test disks attachment for container based additions, | |
6052 | * they shall be already tested when container was created/assembled. | |
88c32bb1 | 6053 | */ |
d424212e | 6054 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 6055 | /* no orom/efi or non-intel hba of the disk */ |
f0f5a016 LM |
6056 | if (rv != 0) { |
6057 | dprintf("capability: %p fd: %d ret: %d\n", | |
6058 | super->orom, fd, rv); | |
6059 | return 1; | |
88c32bb1 DW |
6060 | } |
6061 | ||
f20c3968 DW |
6062 | if (super->current_vol >= 0) |
6063 | return add_to_super_imsm_volume(st, dk, fd, devname); | |
bf5a934a | 6064 | |
c2c087e6 | 6065 | fstat(fd, &stb); |
503975b9 | 6066 | dd = xcalloc(sizeof(*dd), 1); |
c2c087e6 DW |
6067 | dd->major = major(stb.st_rdev); |
6068 | dd->minor = minor(stb.st_rdev); | |
503975b9 | 6069 | dd->devname = devname ? xstrdup(devname) : NULL; |
c2c087e6 | 6070 | dd->fd = fd; |
689c9bf3 | 6071 | dd->e = NULL; |
1a64be56 | 6072 | dd->action = DISK_ADD; |
6da53c0e | 6073 | rv = imsm_read_serial(fd, devname, dd->serial, MAX_RAID_SERIAL_LEN); |
32ba9157 | 6074 | if (rv) { |
e7b84f9d | 6075 | pr_err("failed to retrieve scsi serial, aborting\n"); |
3a85bf0e | 6076 | __free_imsm_disk(dd, 0); |
0030e8d6 | 6077 | abort(); |
c2c087e6 | 6078 | } |
7c798f87 | 6079 | |
20bee0f8 PB |
6080 | if (super->hba && ((super->hba->type == SYS_DEV_NVME) || |
6081 | (super->hba->type == SYS_DEV_VMD))) { | |
6082 | int i; | |
7c798f87 MT |
6083 | char cntrl_path[PATH_MAX]; |
6084 | char *cntrl_name; | |
6085 | char pci_dev_path[PATH_MAX]; | |
20bee0f8 | 6086 | |
7c798f87 MT |
6087 | if (!diskfd_to_devpath(fd, 2, pci_dev_path) || |
6088 | !diskfd_to_devpath(fd, 1, cntrl_path)) { | |
8662f92d | 6089 | pr_err("failed to get dev paths, aborting\n"); |
3a85bf0e | 6090 | __free_imsm_disk(dd, 0); |
a8f3cfd5 MT |
6091 | return 1; |
6092 | } | |
6093 | ||
7c798f87 MT |
6094 | cntrl_name = basename(cntrl_path); |
6095 | if (is_multipath_nvme(fd)) | |
6096 | pr_err("%s controller supports Multi-Path I/O, Intel (R) VROC does not support multipathing\n", | |
6097 | cntrl_name); | |
6098 | ||
6099 | if (devpath_to_vendor(pci_dev_path) == 0x8086) { | |
20bee0f8 PB |
6100 | /* |
6101 | * If Intel's NVMe drive has serial ended with | |
6102 | * "-A","-B","-1" or "-2" it means that this is "x8" | |
6103 | * device (double drive on single PCIe card). | |
6104 | * User should be warned about potential data loss. | |
6105 | */ | |
6106 | for (i = MAX_RAID_SERIAL_LEN-1; i > 0; i--) { | |
6107 | /* Skip empty character at the end */ | |
6108 | if (dd->serial[i] == 0) | |
6109 | continue; | |
6110 | ||
6111 | if (((dd->serial[i] == 'A') || | |
6112 | (dd->serial[i] == 'B') || | |
6113 | (dd->serial[i] == '1') || | |
6114 | (dd->serial[i] == '2')) && | |
6115 | (dd->serial[i-1] == '-')) | |
6116 | pr_err("\tThe action you are about to take may put your data at risk.\n" | |
6117 | "\tPlease note that x8 devices may consist of two separate x4 devices " | |
6118 | "located on a single PCIe port.\n" | |
6119 | "\tRAID 0 is the only supported configuration for this type of x8 device.\n"); | |
6120 | break; | |
6121 | } | |
32716c51 PB |
6122 | } else if (super->hba->type == SYS_DEV_VMD && super->orom && |
6123 | !imsm_orom_has_tpv_support(super->orom)) { | |
6124 | pr_err("\tPlatform configuration does not support non-Intel NVMe drives.\n" | |
8b751247 | 6125 | "\tPlease refer to Intel(R) RSTe/VROC user guide.\n"); |
3a85bf0e | 6126 | __free_imsm_disk(dd, 0); |
32716c51 | 6127 | return 1; |
20bee0f8 PB |
6128 | } |
6129 | } | |
c2c087e6 | 6130 | |
c2c087e6 | 6131 | get_dev_size(fd, NULL, &size); |
3a85bf0e MG |
6132 | if (!get_dev_sector_size(fd, NULL, &member_sector_size)) { |
6133 | __free_imsm_disk(dd, 0); | |
aec01630 | 6134 | return 1; |
3a85bf0e | 6135 | } |
fa7bb6f8 PB |
6136 | |
6137 | if (super->sector_size == 0) { | |
6138 | /* this a first device, so sector_size is not set yet */ | |
6139 | super->sector_size = member_sector_size; | |
fa7bb6f8 PB |
6140 | } |
6141 | ||
71e5411e | 6142 | /* clear migr_rec when adding disk to container */ |
85337573 AO |
6143 | memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE); |
6144 | if (lseek64(fd, size - MIGR_REC_SECTOR_POSITION*member_sector_size, | |
de44e46f | 6145 | SEEK_SET) >= 0) { |
466070ad | 6146 | if ((unsigned int)write(fd, super->migr_rec_buf, |
85337573 AO |
6147 | MIGR_REC_BUF_SECTORS*member_sector_size) != |
6148 | MIGR_REC_BUF_SECTORS*member_sector_size) | |
71e5411e PB |
6149 | perror("Write migr_rec failed"); |
6150 | } | |
6151 | ||
c2c087e6 | 6152 | size /= 512; |
1f24f035 | 6153 | serialcpy(dd->disk.serial, dd->serial); |
5551b113 CA |
6154 | set_total_blocks(&dd->disk, size); |
6155 | if (__le32_to_cpu(dd->disk.total_blocks_hi) > 0) { | |
6156 | struct imsm_super *mpb = super->anchor; | |
6157 | mpb->attributes |= MPB_ATTRIB_2TB_DISK; | |
6158 | } | |
a8619d23 | 6159 | mark_spare(dd); |
c2c087e6 | 6160 | if (sysfs_disk_to_scsi_id(fd, &id) == 0) |
b9f594fe | 6161 | dd->disk.scsi_id = __cpu_to_le32(id); |
c2c087e6 | 6162 | else |
b9f594fe | 6163 | dd->disk.scsi_id = __cpu_to_le32(0); |
43dad3d6 DW |
6164 | |
6165 | if (st->update_tail) { | |
1a64be56 LM |
6166 | dd->next = super->disk_mgmt_list; |
6167 | super->disk_mgmt_list = dd; | |
43dad3d6 | 6168 | } else { |
12724c01 TM |
6169 | /* this is called outside of mdmon |
6170 | * write initial spare metadata | |
6171 | * mdmon will overwrite it. | |
6172 | */ | |
43dad3d6 DW |
6173 | dd->next = super->disks; |
6174 | super->disks = dd; | |
12724c01 | 6175 | write_super_imsm_spare(super, dd); |
43dad3d6 | 6176 | } |
f20c3968 DW |
6177 | |
6178 | return 0; | |
cdddbdbc DW |
6179 | } |
6180 | ||
1a64be56 LM |
6181 | static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk) |
6182 | { | |
6183 | struct intel_super *super = st->sb; | |
6184 | struct dl *dd; | |
6185 | ||
6186 | /* remove from super works only in mdmon - for communication | |
6187 | * manager - monitor. Check if communication memory buffer | |
6188 | * is prepared. | |
6189 | */ | |
6190 | if (!st->update_tail) { | |
1ade5cc1 | 6191 | pr_err("shall be used in mdmon context only\n"); |
1a64be56 LM |
6192 | return 1; |
6193 | } | |
503975b9 | 6194 | dd = xcalloc(1, sizeof(*dd)); |
1a64be56 LM |
6195 | dd->major = dk->major; |
6196 | dd->minor = dk->minor; | |
1a64be56 | 6197 | dd->fd = -1; |
a8619d23 | 6198 | mark_spare(dd); |
1a64be56 LM |
6199 | dd->action = DISK_REMOVE; |
6200 | ||
6201 | dd->next = super->disk_mgmt_list; | |
6202 | super->disk_mgmt_list = dd; | |
6203 | ||
1a64be56 LM |
6204 | return 0; |
6205 | } | |
6206 | ||
f796af5d DW |
6207 | static int store_imsm_mpb(int fd, struct imsm_super *mpb); |
6208 | ||
6209 | static union { | |
f36a9ecd | 6210 | char buf[MAX_SECTOR_SIZE]; |
f796af5d | 6211 | struct imsm_super anchor; |
f36a9ecd | 6212 | } spare_record __attribute__ ((aligned(MAX_SECTOR_SIZE))); |
c2c087e6 | 6213 | |
12724c01 TM |
6214 | |
6215 | static int write_super_imsm_spare(struct intel_super *super, struct dl *d) | |
d23fe947 | 6216 | { |
d23fe947 | 6217 | struct imsm_super *mpb = super->anchor; |
f796af5d | 6218 | struct imsm_super *spare = &spare_record.anchor; |
d23fe947 | 6219 | __u32 sum; |
12724c01 TM |
6220 | |
6221 | if (d->index != -1) | |
6222 | return 1; | |
d23fe947 | 6223 | |
68641cdb JS |
6224 | spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super)); |
6225 | spare->generation_num = __cpu_to_le32(1UL); | |
f796af5d | 6226 | spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; |
68641cdb JS |
6227 | spare->num_disks = 1; |
6228 | spare->num_raid_devs = 0; | |
6229 | spare->cache_size = mpb->cache_size; | |
6230 | spare->pwr_cycle_count = __cpu_to_le32(1); | |
f796af5d DW |
6231 | |
6232 | snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH, | |
6233 | MPB_SIGNATURE MPB_VERSION_RAID0); | |
d23fe947 | 6234 | |
12724c01 TM |
6235 | spare->disk[0] = d->disk; |
6236 | if (__le32_to_cpu(d->disk.total_blocks_hi) > 0) | |
6237 | spare->attributes |= MPB_ATTRIB_2TB_DISK; | |
6238 | ||
6239 | if (super->sector_size == 4096) | |
6240 | convert_to_4k_imsm_disk(&spare->disk[0]); | |
d23fe947 | 6241 | |
12724c01 TM |
6242 | sum = __gen_imsm_checksum(spare); |
6243 | spare->family_num = __cpu_to_le32(sum); | |
6244 | spare->orig_family_num = 0; | |
6245 | sum = __gen_imsm_checksum(spare); | |
6246 | spare->check_sum = __cpu_to_le32(sum); | |
027c374f | 6247 | |
12724c01 TM |
6248 | if (store_imsm_mpb(d->fd, spare)) { |
6249 | pr_err("failed for device %d:%d %s\n", | |
6250 | d->major, d->minor, strerror(errno)); | |
6251 | return 1; | |
6252 | } | |
6253 | ||
6254 | return 0; | |
6255 | } | |
6256 | /* spare records have their own family number and do not have any defined raid | |
6257 | * devices | |
6258 | */ | |
6259 | static int write_super_imsm_spares(struct intel_super *super, int doclose) | |
6260 | { | |
6261 | struct dl *d; | |
f36a9ecd | 6262 | |
12724c01 TM |
6263 | for (d = super->disks; d; d = d->next) { |
6264 | if (d->index != -1) | |
6265 | continue; | |
d23fe947 | 6266 | |
12724c01 | 6267 | if (write_super_imsm_spare(super, d)) |
e74255d9 | 6268 | return 1; |
12724c01 | 6269 | |
4389ce73 MT |
6270 | if (doclose) |
6271 | close_fd(&d->fd); | |
d23fe947 DW |
6272 | } |
6273 | ||
e74255d9 | 6274 | return 0; |
d23fe947 DW |
6275 | } |
6276 | ||
36988a3d | 6277 | static int write_super_imsm(struct supertype *st, int doclose) |
cdddbdbc | 6278 | { |
36988a3d | 6279 | struct intel_super *super = st->sb; |
f36a9ecd | 6280 | unsigned int sector_size = super->sector_size; |
949c47a0 | 6281 | struct imsm_super *mpb = super->anchor; |
c2c087e6 DW |
6282 | struct dl *d; |
6283 | __u32 generation; | |
6284 | __u32 sum; | |
d23fe947 | 6285 | int spares = 0; |
949c47a0 | 6286 | int i; |
a48ac0a8 | 6287 | __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk); |
36988a3d | 6288 | int num_disks = 0; |
146c6260 | 6289 | int clear_migration_record = 1; |
bbab0940 | 6290 | __u32 bbm_log_size; |
cdddbdbc | 6291 | |
c2c087e6 DW |
6292 | /* 'generation' is incremented everytime the metadata is written */ |
6293 | generation = __le32_to_cpu(mpb->generation_num); | |
6294 | generation++; | |
6295 | mpb->generation_num = __cpu_to_le32(generation); | |
6296 | ||
148acb7b DW |
6297 | /* fix up cases where previous mdadm releases failed to set |
6298 | * orig_family_num | |
6299 | */ | |
6300 | if (mpb->orig_family_num == 0) | |
6301 | mpb->orig_family_num = mpb->family_num; | |
6302 | ||
d23fe947 | 6303 | for (d = super->disks; d; d = d->next) { |
8796fdc4 | 6304 | if (d->index == -1) |
d23fe947 | 6305 | spares++; |
36988a3d | 6306 | else { |
d23fe947 | 6307 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
6308 | num_disks++; |
6309 | } | |
d23fe947 | 6310 | } |
36988a3d | 6311 | for (d = super->missing; d; d = d->next) { |
47ee5a45 | 6312 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
6313 | num_disks++; |
6314 | } | |
6315 | mpb->num_disks = num_disks; | |
6316 | mpb_size += sizeof(struct imsm_disk) * mpb->num_disks; | |
b9f594fe | 6317 | |
949c47a0 DW |
6318 | for (i = 0; i < mpb->num_raid_devs; i++) { |
6319 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
36988a3d | 6320 | struct imsm_dev *dev2 = get_imsm_dev(super, i); |
756a15f3 MG |
6321 | |
6322 | imsm_copy_dev(dev, dev2); | |
6323 | mpb_size += sizeof_imsm_dev(dev, 0); | |
6324 | ||
146c6260 AK |
6325 | if (is_gen_migration(dev2)) |
6326 | clear_migration_record = 0; | |
949c47a0 | 6327 | } |
bbab0940 TM |
6328 | |
6329 | bbm_log_size = get_imsm_bbm_log_size(super->bbm_log); | |
6330 | ||
6331 | if (bbm_log_size) { | |
6332 | memcpy((void *)mpb + mpb_size, super->bbm_log, bbm_log_size); | |
6333 | mpb->attributes |= MPB_ATTRIB_BBM; | |
6334 | } else | |
6335 | mpb->attributes &= ~MPB_ATTRIB_BBM; | |
6336 | ||
6337 | super->anchor->bbm_log_size = __cpu_to_le32(bbm_log_size); | |
6338 | mpb_size += bbm_log_size; | |
a48ac0a8 | 6339 | mpb->mpb_size = __cpu_to_le32(mpb_size); |
949c47a0 | 6340 | |
bbab0940 TM |
6341 | #ifdef DEBUG |
6342 | assert(super->len == 0 || mpb_size <= super->len); | |
6343 | #endif | |
6344 | ||
c2c087e6 | 6345 | /* recalculate checksum */ |
949c47a0 | 6346 | sum = __gen_imsm_checksum(mpb); |
c2c087e6 DW |
6347 | mpb->check_sum = __cpu_to_le32(sum); |
6348 | ||
51d83f5d AK |
6349 | if (super->clean_migration_record_by_mdmon) { |
6350 | clear_migration_record = 1; | |
6351 | super->clean_migration_record_by_mdmon = 0; | |
6352 | } | |
146c6260 | 6353 | if (clear_migration_record) |
de44e46f | 6354 | memset(super->migr_rec_buf, 0, |
85337573 | 6355 | MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE); |
146c6260 | 6356 | |
f36a9ecd PB |
6357 | if (sector_size == 4096) |
6358 | convert_to_4k(super); | |
6359 | ||
d23fe947 | 6360 | /* write the mpb for disks that compose raid devices */ |
c2c087e6 | 6361 | for (d = super->disks; d ; d = d->next) { |
86c54047 | 6362 | if (d->index < 0 || is_failed(&d->disk)) |
d23fe947 | 6363 | continue; |
30602f53 | 6364 | |
146c6260 AK |
6365 | if (clear_migration_record) { |
6366 | unsigned long long dsize; | |
6367 | ||
6368 | get_dev_size(d->fd, NULL, &dsize); | |
de44e46f PB |
6369 | if (lseek64(d->fd, dsize - sector_size, |
6370 | SEEK_SET) >= 0) { | |
466070ad PB |
6371 | if ((unsigned int)write(d->fd, |
6372 | super->migr_rec_buf, | |
de44e46f PB |
6373 | MIGR_REC_BUF_SECTORS*sector_size) != |
6374 | MIGR_REC_BUF_SECTORS*sector_size) | |
9e2d750d | 6375 | perror("Write migr_rec failed"); |
146c6260 AK |
6376 | } |
6377 | } | |
51d83f5d AK |
6378 | |
6379 | if (store_imsm_mpb(d->fd, mpb)) | |
6380 | fprintf(stderr, | |
1ade5cc1 N |
6381 | "failed for device %d:%d (fd: %d)%s\n", |
6382 | d->major, d->minor, | |
51d83f5d AK |
6383 | d->fd, strerror(errno)); |
6384 | ||
4389ce73 MT |
6385 | if (doclose) |
6386 | close_fd(&d->fd); | |
c2c087e6 DW |
6387 | } |
6388 | ||
d23fe947 DW |
6389 | if (spares) |
6390 | return write_super_imsm_spares(super, doclose); | |
6391 | ||
e74255d9 | 6392 | return 0; |
c2c087e6 DW |
6393 | } |
6394 | ||
9b1fb677 | 6395 | static int create_array(struct supertype *st, int dev_idx) |
43dad3d6 DW |
6396 | { |
6397 | size_t len; | |
6398 | struct imsm_update_create_array *u; | |
6399 | struct intel_super *super = st->sb; | |
9b1fb677 | 6400 | struct imsm_dev *dev = get_imsm_dev(super, dev_idx); |
238c0a71 | 6401 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
54c2c1ea DW |
6402 | struct disk_info *inf; |
6403 | struct imsm_disk *disk; | |
6404 | int i; | |
43dad3d6 | 6405 | |
54c2c1ea DW |
6406 | len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) + |
6407 | sizeof(*inf) * map->num_members; | |
503975b9 | 6408 | u = xmalloc(len); |
43dad3d6 | 6409 | u->type = update_create_array; |
9b1fb677 | 6410 | u->dev_idx = dev_idx; |
43dad3d6 | 6411 | imsm_copy_dev(&u->dev, dev); |
54c2c1ea DW |
6412 | inf = get_disk_info(u); |
6413 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 6414 | int idx = get_imsm_disk_idx(dev, i, MAP_X); |
9b1fb677 | 6415 | |
54c2c1ea | 6416 | disk = get_imsm_disk(super, idx); |
1ca5c8e0 N |
6417 | if (!disk) |
6418 | disk = get_imsm_missing(super, idx); | |
54c2c1ea DW |
6419 | serialcpy(inf[i].serial, disk->serial); |
6420 | } | |
43dad3d6 DW |
6421 | append_metadata_update(st, u, len); |
6422 | ||
6423 | return 0; | |
6424 | } | |
6425 | ||
1a64be56 | 6426 | static int mgmt_disk(struct supertype *st) |
43dad3d6 DW |
6427 | { |
6428 | struct intel_super *super = st->sb; | |
6429 | size_t len; | |
1a64be56 | 6430 | struct imsm_update_add_remove_disk *u; |
43dad3d6 | 6431 | |
1a64be56 | 6432 | if (!super->disk_mgmt_list) |
43dad3d6 DW |
6433 | return 0; |
6434 | ||
6435 | len = sizeof(*u); | |
503975b9 | 6436 | u = xmalloc(len); |
1a64be56 | 6437 | u->type = update_add_remove_disk; |
43dad3d6 DW |
6438 | append_metadata_update(st, u, len); |
6439 | ||
6440 | return 0; | |
6441 | } | |
2432ce9b AP |
6442 | |
6443 | __u32 crc32c_le(__u32 crc, unsigned char const *p, size_t len); | |
6444 | ||
e397cefe AP |
6445 | static int write_ppl_header(unsigned long long ppl_sector, int fd, void *buf) |
6446 | { | |
6447 | struct ppl_header *ppl_hdr = buf; | |
6448 | int ret; | |
6449 | ||
6450 | ppl_hdr->checksum = __cpu_to_le32(~crc32c_le(~0, buf, PPL_HEADER_SIZE)); | |
6451 | ||
6452 | if (lseek64(fd, ppl_sector * 512, SEEK_SET) < 0) { | |
6453 | ret = -errno; | |
6454 | perror("Failed to seek to PPL header location"); | |
6455 | return ret; | |
6456 | } | |
6457 | ||
6458 | if (write(fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) { | |
6459 | ret = -errno; | |
6460 | perror("Write PPL header failed"); | |
6461 | return ret; | |
6462 | } | |
6463 | ||
6464 | fsync(fd); | |
6465 | ||
6466 | return 0; | |
6467 | } | |
6468 | ||
2432ce9b AP |
6469 | static int write_init_ppl_imsm(struct supertype *st, struct mdinfo *info, int fd) |
6470 | { | |
6471 | struct intel_super *super = st->sb; | |
6472 | void *buf; | |
6473 | struct ppl_header *ppl_hdr; | |
6474 | int ret; | |
6475 | ||
b2514242 PB |
6476 | /* first clear entire ppl space */ |
6477 | ret = zero_disk_range(fd, info->ppl_sector, info->ppl_size); | |
6478 | if (ret) | |
6479 | return ret; | |
6480 | ||
6481 | ret = posix_memalign(&buf, MAX_SECTOR_SIZE, PPL_HEADER_SIZE); | |
2432ce9b AP |
6482 | if (ret) { |
6483 | pr_err("Failed to allocate PPL header buffer\n"); | |
e397cefe | 6484 | return -ret; |
2432ce9b AP |
6485 | } |
6486 | ||
6487 | memset(buf, 0, PPL_HEADER_SIZE); | |
6488 | ppl_hdr = buf; | |
6489 | memset(ppl_hdr->reserved, 0xff, PPL_HDR_RESERVED); | |
6490 | ppl_hdr->signature = __cpu_to_le32(super->anchor->orig_family_num); | |
b23d0750 AP |
6491 | |
6492 | if (info->mismatch_cnt) { | |
6493 | /* | |
6494 | * We are overwriting an invalid ppl. Make one entry with wrong | |
6495 | * checksum to prevent the kernel from skipping resync. | |
6496 | */ | |
6497 | ppl_hdr->entries_count = __cpu_to_le32(1); | |
6498 | ppl_hdr->entries[0].checksum = ~0; | |
6499 | } | |
6500 | ||
e397cefe | 6501 | ret = write_ppl_header(info->ppl_sector, fd, buf); |
2432ce9b AP |
6502 | |
6503 | free(buf); | |
6504 | return ret; | |
6505 | } | |
6506 | ||
e397cefe AP |
6507 | static int is_rebuilding(struct imsm_dev *dev); |
6508 | ||
2432ce9b AP |
6509 | static int validate_ppl_imsm(struct supertype *st, struct mdinfo *info, |
6510 | struct mdinfo *disk) | |
6511 | { | |
6512 | struct intel_super *super = st->sb; | |
6513 | struct dl *d; | |
e397cefe | 6514 | void *buf_orig, *buf, *buf_prev = NULL; |
2432ce9b | 6515 | int ret = 0; |
e397cefe | 6516 | struct ppl_header *ppl_hdr = NULL; |
2432ce9b AP |
6517 | __u32 crc; |
6518 | struct imsm_dev *dev; | |
2432ce9b | 6519 | __u32 idx; |
44b6b876 PB |
6520 | unsigned int i; |
6521 | unsigned long long ppl_offset = 0; | |
6522 | unsigned long long prev_gen_num = 0; | |
2432ce9b AP |
6523 | |
6524 | if (disk->disk.raid_disk < 0) | |
6525 | return 0; | |
6526 | ||
2432ce9b | 6527 | dev = get_imsm_dev(super, info->container_member); |
2fc0fc63 | 6528 | idx = get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_0); |
2432ce9b AP |
6529 | d = get_imsm_dl_disk(super, idx); |
6530 | ||
6531 | if (!d || d->index < 0 || is_failed(&d->disk)) | |
e397cefe AP |
6532 | return 0; |
6533 | ||
6534 | if (posix_memalign(&buf_orig, MAX_SECTOR_SIZE, PPL_HEADER_SIZE * 2)) { | |
6535 | pr_err("Failed to allocate PPL header buffer\n"); | |
6536 | return -1; | |
6537 | } | |
6538 | buf = buf_orig; | |
2432ce9b | 6539 | |
44b6b876 PB |
6540 | ret = 1; |
6541 | while (ppl_offset < MULTIPLE_PPL_AREA_SIZE_IMSM) { | |
e397cefe AP |
6542 | void *tmp; |
6543 | ||
44b6b876 | 6544 | dprintf("Checking potential PPL at offset: %llu\n", ppl_offset); |
2432ce9b | 6545 | |
44b6b876 PB |
6546 | if (lseek64(d->fd, info->ppl_sector * 512 + ppl_offset, |
6547 | SEEK_SET) < 0) { | |
6548 | perror("Failed to seek to PPL header location"); | |
6549 | ret = -1; | |
e397cefe | 6550 | break; |
44b6b876 | 6551 | } |
2432ce9b | 6552 | |
44b6b876 PB |
6553 | if (read(d->fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) { |
6554 | perror("Read PPL header failed"); | |
6555 | ret = -1; | |
e397cefe | 6556 | break; |
44b6b876 | 6557 | } |
2432ce9b | 6558 | |
44b6b876 | 6559 | ppl_hdr = buf; |
2432ce9b | 6560 | |
44b6b876 PB |
6561 | crc = __le32_to_cpu(ppl_hdr->checksum); |
6562 | ppl_hdr->checksum = 0; | |
6563 | ||
6564 | if (crc != ~crc32c_le(~0, buf, PPL_HEADER_SIZE)) { | |
6565 | dprintf("Wrong PPL header checksum on %s\n", | |
6566 | d->devname); | |
e397cefe | 6567 | break; |
44b6b876 PB |
6568 | } |
6569 | ||
6570 | if (prev_gen_num > __le64_to_cpu(ppl_hdr->generation)) { | |
6571 | /* previous was newest, it was already checked */ | |
e397cefe | 6572 | break; |
44b6b876 PB |
6573 | } |
6574 | ||
6575 | if ((__le32_to_cpu(ppl_hdr->signature) != | |
6576 | super->anchor->orig_family_num)) { | |
6577 | dprintf("Wrong PPL header signature on %s\n", | |
6578 | d->devname); | |
6579 | ret = 1; | |
e397cefe | 6580 | break; |
44b6b876 PB |
6581 | } |
6582 | ||
6583 | ret = 0; | |
6584 | prev_gen_num = __le64_to_cpu(ppl_hdr->generation); | |
2432ce9b | 6585 | |
44b6b876 PB |
6586 | ppl_offset += PPL_HEADER_SIZE; |
6587 | for (i = 0; i < __le32_to_cpu(ppl_hdr->entries_count); i++) | |
6588 | ppl_offset += | |
6589 | __le32_to_cpu(ppl_hdr->entries[i].pp_size); | |
e397cefe AP |
6590 | |
6591 | if (!buf_prev) | |
6592 | buf_prev = buf + PPL_HEADER_SIZE; | |
6593 | tmp = buf_prev; | |
6594 | buf_prev = buf; | |
6595 | buf = tmp; | |
2432ce9b AP |
6596 | } |
6597 | ||
e397cefe AP |
6598 | if (buf_prev) { |
6599 | buf = buf_prev; | |
6600 | ppl_hdr = buf_prev; | |
6601 | } | |
2432ce9b | 6602 | |
54148aba PB |
6603 | /* |
6604 | * Update metadata to use mutliple PPLs area (1MB). | |
6605 | * This is done once for all RAID members | |
6606 | */ | |
6607 | if (info->consistency_policy == CONSISTENCY_POLICY_PPL && | |
6608 | info->ppl_size != (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9)) { | |
6609 | char subarray[20]; | |
6610 | struct mdinfo *member_dev; | |
6611 | ||
6612 | sprintf(subarray, "%d", info->container_member); | |
6613 | ||
6614 | if (mdmon_running(st->container_devnm)) | |
6615 | st->update_tail = &st->updates; | |
6616 | ||
03312b52 | 6617 | if (st->ss->update_subarray(st, subarray, UOPT_PPL, NULL)) { |
54148aba PB |
6618 | pr_err("Failed to update subarray %s\n", |
6619 | subarray); | |
6620 | } else { | |
6621 | if (st->update_tail) | |
6622 | flush_metadata_updates(st); | |
6623 | else | |
6624 | st->ss->sync_metadata(st); | |
6625 | info->ppl_size = (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9); | |
6626 | for (member_dev = info->devs; member_dev; | |
6627 | member_dev = member_dev->next) | |
6628 | member_dev->ppl_size = | |
6629 | (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9); | |
6630 | } | |
6631 | } | |
6632 | ||
b23d0750 | 6633 | if (ret == 1) { |
2fc0fc63 AP |
6634 | struct imsm_map *map = get_imsm_map(dev, MAP_X); |
6635 | ||
50b9c10d PB |
6636 | if (map->map_state == IMSM_T_STATE_UNINITIALIZED || |
6637 | (map->map_state == IMSM_T_STATE_NORMAL && | |
2ec9d182 | 6638 | !(dev->vol.dirty & RAIDVOL_DIRTY)) || |
e397cefe | 6639 | (is_rebuilding(dev) && |
4036e7ee | 6640 | vol_curr_migr_unit(dev) == 0 && |
2ec9d182 | 6641 | get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_1) != idx)) |
b23d0750 AP |
6642 | ret = st->ss->write_init_ppl(st, info, d->fd); |
6643 | else | |
6644 | info->mismatch_cnt++; | |
e397cefe AP |
6645 | } else if (ret == 0 && |
6646 | ppl_hdr->entries_count == 0 && | |
6647 | is_rebuilding(dev) && | |
6648 | info->resync_start == 0) { | |
6649 | /* | |
6650 | * The header has no entries - add a single empty entry and | |
6651 | * rewrite the header to prevent the kernel from going into | |
6652 | * resync after an interrupted rebuild. | |
6653 | */ | |
6654 | ppl_hdr->entries_count = __cpu_to_le32(1); | |
6655 | ret = write_ppl_header(info->ppl_sector, d->fd, buf); | |
b23d0750 | 6656 | } |
2432ce9b | 6657 | |
e397cefe AP |
6658 | free(buf_orig); |
6659 | ||
2432ce9b AP |
6660 | return ret; |
6661 | } | |
6662 | ||
2432ce9b AP |
6663 | static int write_init_ppl_imsm_all(struct supertype *st, struct mdinfo *info) |
6664 | { | |
6665 | struct intel_super *super = st->sb; | |
6666 | struct dl *d; | |
6667 | int ret = 0; | |
6668 | ||
6669 | if (info->consistency_policy != CONSISTENCY_POLICY_PPL || | |
6670 | info->array.level != 5) | |
6671 | return 0; | |
6672 | ||
6673 | for (d = super->disks; d ; d = d->next) { | |
6674 | if (d->index < 0 || is_failed(&d->disk)) | |
6675 | continue; | |
6676 | ||
6677 | ret = st->ss->write_init_ppl(st, info, d->fd); | |
6678 | if (ret) | |
6679 | break; | |
6680 | } | |
6681 | ||
6682 | return ret; | |
6683 | } | |
43dad3d6 | 6684 | |
fbc42556 JR |
6685 | /******************************************************************************* |
6686 | * Function: write_init_bitmap_imsm_vol | |
6687 | * Description: Write a bitmap header and prepares the area for the bitmap. | |
6688 | * Parameters: | |
6689 | * st : supertype information | |
6690 | * vol_idx : the volume index to use | |
6691 | * | |
6692 | * Returns: | |
6693 | * 0 : success | |
6694 | * -1 : fail | |
6695 | ******************************************************************************/ | |
6696 | static int write_init_bitmap_imsm_vol(struct supertype *st, int vol_idx) | |
6697 | { | |
6698 | struct intel_super *super = st->sb; | |
6699 | int prev_current_vol = super->current_vol; | |
6700 | struct dl *d; | |
6701 | int ret = 0; | |
6702 | ||
6703 | super->current_vol = vol_idx; | |
6704 | for (d = super->disks; d; d = d->next) { | |
6705 | if (d->index < 0 || is_failed(&d->disk)) | |
6706 | continue; | |
6707 | ret = st->ss->write_bitmap(st, d->fd, NoUpdate); | |
6708 | if (ret) | |
6709 | break; | |
6710 | } | |
6711 | super->current_vol = prev_current_vol; | |
6712 | return ret; | |
6713 | } | |
6714 | ||
6715 | /******************************************************************************* | |
6716 | * Function: write_init_bitmap_imsm_all | |
6717 | * Description: Write a bitmap header and prepares the area for the bitmap. | |
6718 | * Operation is executed for volumes with CONSISTENCY_POLICY_BITMAP. | |
6719 | * Parameters: | |
6720 | * st : supertype information | |
6721 | * info : info about the volume where the bitmap should be written | |
6722 | * vol_idx : the volume index to use | |
6723 | * | |
6724 | * Returns: | |
6725 | * 0 : success | |
6726 | * -1 : fail | |
6727 | ******************************************************************************/ | |
6728 | static int write_init_bitmap_imsm_all(struct supertype *st, struct mdinfo *info, | |
6729 | int vol_idx) | |
6730 | { | |
6731 | int ret = 0; | |
6732 | ||
6733 | if (info && (info->consistency_policy == CONSISTENCY_POLICY_BITMAP)) | |
6734 | ret = write_init_bitmap_imsm_vol(st, vol_idx); | |
6735 | ||
6736 | return ret; | |
6737 | } | |
6738 | ||
c2c087e6 DW |
6739 | static int write_init_super_imsm(struct supertype *st) |
6740 | { | |
9b1fb677 DW |
6741 | struct intel_super *super = st->sb; |
6742 | int current_vol = super->current_vol; | |
2432ce9b AP |
6743 | int rv = 0; |
6744 | struct mdinfo info; | |
6745 | ||
6746 | getinfo_super_imsm(st, &info, NULL); | |
9b1fb677 DW |
6747 | |
6748 | /* we are done with current_vol reset it to point st at the container */ | |
6749 | super->current_vol = -1; | |
6750 | ||
8273f55e | 6751 | if (st->update_tail) { |
43dad3d6 DW |
6752 | /* queue the recently created array / added disk |
6753 | * as a metadata update */ | |
8273f55e | 6754 | |
43dad3d6 | 6755 | /* determine if we are creating a volume or adding a disk */ |
9b1fb677 | 6756 | if (current_vol < 0) { |
1a64be56 LM |
6757 | /* in the mgmt (add/remove) disk case we are running |
6758 | * in mdmon context, so don't close fd's | |
43dad3d6 | 6759 | */ |
2432ce9b AP |
6760 | rv = mgmt_disk(st); |
6761 | } else { | |
fbc42556 | 6762 | /* adding the second volume to the array */ |
2432ce9b | 6763 | rv = write_init_ppl_imsm_all(st, &info); |
fbc42556 JR |
6764 | if (!rv) |
6765 | rv = write_init_bitmap_imsm_all(st, &info, current_vol); | |
2432ce9b AP |
6766 | if (!rv) |
6767 | rv = create_array(st, current_vol); | |
6768 | } | |
d682f344 N |
6769 | } else { |
6770 | struct dl *d; | |
6771 | for (d = super->disks; d; d = d->next) | |
ba728be7 | 6772 | Kill(d->devname, NULL, 0, -1, 1); |
fbc42556 | 6773 | if (current_vol >= 0) { |
2432ce9b | 6774 | rv = write_init_ppl_imsm_all(st, &info); |
fbc42556 JR |
6775 | if (!rv) |
6776 | rv = write_init_bitmap_imsm_all(st, &info, current_vol); | |
6777 | } | |
6778 | ||
2432ce9b AP |
6779 | if (!rv) |
6780 | rv = write_super_imsm(st, 1); | |
d682f344 | 6781 | } |
2432ce9b AP |
6782 | |
6783 | return rv; | |
cdddbdbc DW |
6784 | } |
6785 | ||
e683ca88 | 6786 | static int store_super_imsm(struct supertype *st, int fd) |
cdddbdbc | 6787 | { |
e683ca88 DW |
6788 | struct intel_super *super = st->sb; |
6789 | struct imsm_super *mpb = super ? super->anchor : NULL; | |
551c80c1 | 6790 | |
e683ca88 | 6791 | if (!mpb) |
ad97895e DW |
6792 | return 1; |
6793 | ||
f36a9ecd PB |
6794 | if (super->sector_size == 4096) |
6795 | convert_to_4k(super); | |
e683ca88 | 6796 | return store_imsm_mpb(fd, mpb); |
cdddbdbc DW |
6797 | } |
6798 | ||
cdddbdbc | 6799 | static int validate_geometry_imsm_container(struct supertype *st, int level, |
1f5d54a0 | 6800 | int raiddisks, |
af4348dd N |
6801 | unsigned long long data_offset, |
6802 | char *dev, | |
2c514b71 NB |
6803 | unsigned long long *freesize, |
6804 | int verbose) | |
cdddbdbc | 6805 | { |
c2c087e6 DW |
6806 | int fd; |
6807 | unsigned long long ldsize; | |
8662f92d | 6808 | struct intel_super *super = NULL; |
f2f5c343 | 6809 | int rv = 0; |
cdddbdbc | 6810 | |
6f2af6a4 | 6811 | if (!is_container(level)) |
c2c087e6 DW |
6812 | return 0; |
6813 | if (!dev) | |
6814 | return 1; | |
6815 | ||
dca80fcd | 6816 | fd = dev_open(dev, O_RDONLY|O_EXCL); |
4389ce73 MT |
6817 | if (!is_fd_valid(fd)) { |
6818 | pr_vrb("imsm: Cannot open %s: %s\n", dev, strerror(errno)); | |
c2c087e6 DW |
6819 | return 0; |
6820 | } | |
8662f92d MT |
6821 | if (!get_dev_size(fd, dev, &ldsize)) |
6822 | goto exit; | |
f2f5c343 LM |
6823 | |
6824 | /* capabilities retrieve could be possible | |
6825 | * note that there is no fd for the disks in array. | |
6826 | */ | |
6827 | super = alloc_super(); | |
8662f92d MT |
6828 | if (!super) |
6829 | goto exit; | |
6830 | ||
6831 | if (!get_dev_sector_size(fd, NULL, &super->sector_size)) | |
6832 | goto exit; | |
fa7bb6f8 | 6833 | |
ba728be7 | 6834 | rv = find_intel_hba_capability(fd, super, verbose > 0 ? dev : NULL); |
f2f5c343 LM |
6835 | if (rv != 0) { |
6836 | #if DEBUG | |
6837 | char str[256]; | |
6838 | fd2devname(fd, str); | |
1ade5cc1 | 6839 | dprintf("fd: %d %s orom: %p rv: %d raiddisk: %d\n", |
f2f5c343 LM |
6840 | fd, str, super->orom, rv, raiddisks); |
6841 | #endif | |
6842 | /* no orom/efi or non-intel hba of the disk */ | |
8662f92d MT |
6843 | rv = 0; |
6844 | goto exit; | |
f2f5c343 | 6845 | } |
9126b9a8 CA |
6846 | if (super->orom) { |
6847 | if (raiddisks > super->orom->tds) { | |
6848 | if (verbose) | |
7a862a02 | 6849 | pr_err("%d exceeds maximum number of platform supported disks: %d\n", |
9126b9a8 | 6850 | raiddisks, super->orom->tds); |
8662f92d | 6851 | goto exit; |
9126b9a8 CA |
6852 | } |
6853 | if ((super->orom->attr & IMSM_OROM_ATTR_2TB_DISK) == 0 && | |
6854 | (ldsize >> 9) >> 32 > 0) { | |
6855 | if (verbose) | |
e7b84f9d | 6856 | pr_err("%s exceeds maximum platform supported size\n", dev); |
8662f92d MT |
6857 | goto exit; |
6858 | } | |
6859 | ||
6860 | if (super->hba->type == SYS_DEV_VMD || | |
6861 | super->hba->type == SYS_DEV_NVME) { | |
6862 | if (!imsm_is_nvme_namespace_supported(fd, 1)) { | |
6863 | if (verbose) | |
6864 | pr_err("NVMe namespace %s is not supported by IMSM\n", | |
6865 | basename(dev)); | |
6866 | goto exit; | |
6867 | } | |
9126b9a8 | 6868 | } |
f2f5c343 | 6869 | } |
1f5d54a0 MT |
6870 | if (freesize) |
6871 | *freesize = avail_size_imsm(st, ldsize >> 9, data_offset); | |
8662f92d MT |
6872 | rv = 1; |
6873 | exit: | |
6874 | if (super) | |
6875 | free_imsm(super); | |
6876 | close(fd); | |
c2c087e6 | 6877 | |
8662f92d | 6878 | return rv; |
cdddbdbc DW |
6879 | } |
6880 | ||
0dcecb2e DW |
6881 | static unsigned long long find_size(struct extent *e, int *idx, int num_extents) |
6882 | { | |
6883 | const unsigned long long base_start = e[*idx].start; | |
6884 | unsigned long long end = base_start + e[*idx].size; | |
6885 | int i; | |
6886 | ||
6887 | if (base_start == end) | |
6888 | return 0; | |
6889 | ||
6890 | *idx = *idx + 1; | |
6891 | for (i = *idx; i < num_extents; i++) { | |
6892 | /* extend overlapping extents */ | |
6893 | if (e[i].start >= base_start && | |
6894 | e[i].start <= end) { | |
6895 | if (e[i].size == 0) | |
6896 | return 0; | |
6897 | if (e[i].start + e[i].size > end) | |
6898 | end = e[i].start + e[i].size; | |
6899 | } else if (e[i].start > end) { | |
6900 | *idx = i; | |
6901 | break; | |
6902 | } | |
6903 | } | |
6904 | ||
6905 | return end - base_start; | |
6906 | } | |
6907 | ||
aa19fdd4 | 6908 | /** merge_extents() - analyze extents and get free size. |
9bc426fa | 6909 | * @super: Intel metadata, not NULL. |
aa19fdd4 | 6910 | * @expanding: if set, we are expanding &super->current_vol. |
9bc426fa | 6911 | * |
aa19fdd4 MT |
6912 | * Build a composite disk with all known extents and generate a size given the |
6913 | * "all disks in an array must share a common start offset" constraint. | |
6914 | * If a volume is expanded, then return free space after the volume. | |
9bc426fa | 6915 | * |
aa19fdd4 | 6916 | * Return: Free space or 0 on failure. |
9bc426fa | 6917 | */ |
aa19fdd4 | 6918 | static unsigned long long merge_extents(struct intel_super *super, const bool expanding) |
0dcecb2e | 6919 | { |
9bc426fa | 6920 | struct extent *e; |
0dcecb2e | 6921 | struct dl *dl; |
aa19fdd4 MT |
6922 | int i, j, pos_vol_idx = -1; |
6923 | int extent_idx = 0; | |
6924 | int sum_extents = 0; | |
6925 | unsigned long long pos = 0; | |
b9d77223 | 6926 | unsigned long long start = 0; |
1dea84ae MT |
6927 | unsigned long long free_size = 0; |
6928 | ||
6929 | unsigned long pre_reservation = 0; | |
6930 | unsigned long post_reservation = IMSM_RESERVED_SECTORS; | |
6931 | unsigned long reservation_size; | |
0dcecb2e | 6932 | |
9bc426fa MT |
6933 | for (dl = super->disks; dl; dl = dl->next) |
6934 | if (dl->e) | |
6935 | sum_extents += dl->extent_cnt; | |
6936 | e = xcalloc(sum_extents, sizeof(struct extent)); | |
6937 | ||
0dcecb2e DW |
6938 | /* coalesce and sort all extents. also, check to see if we need to |
6939 | * reserve space between member arrays | |
6940 | */ | |
6941 | j = 0; | |
6942 | for (dl = super->disks; dl; dl = dl->next) { | |
6943 | if (!dl->e) | |
6944 | continue; | |
6945 | for (i = 0; i < dl->extent_cnt; i++) | |
6946 | e[j++] = dl->e[i]; | |
6947 | } | |
6948 | qsort(e, sum_extents, sizeof(*e), cmp_extent); | |
6949 | ||
6950 | /* merge extents */ | |
6951 | i = 0; | |
6952 | j = 0; | |
6953 | while (i < sum_extents) { | |
6954 | e[j].start = e[i].start; | |
aa19fdd4 | 6955 | e[j].vol = e[i].vol; |
0dcecb2e DW |
6956 | e[j].size = find_size(e, &i, sum_extents); |
6957 | j++; | |
6958 | if (e[j-1].size == 0) | |
6959 | break; | |
6960 | } | |
6961 | ||
0dcecb2e DW |
6962 | i = 0; |
6963 | do { | |
aa19fdd4 | 6964 | unsigned long long esize = e[i].start - pos; |
0dcecb2e | 6965 | |
1dea84ae MT |
6966 | if (expanding ? pos_vol_idx == super->current_vol : esize >= free_size) { |
6967 | free_size = esize; | |
0dcecb2e | 6968 | start = pos; |
aa19fdd4 | 6969 | extent_idx = i; |
0dcecb2e | 6970 | } |
aa19fdd4 | 6971 | |
0dcecb2e | 6972 | pos = e[i].start + e[i].size; |
aa19fdd4 MT |
6973 | pos_vol_idx = e[i].vol; |
6974 | ||
0dcecb2e DW |
6975 | i++; |
6976 | } while (e[i-1].size); | |
0dcecb2e | 6977 | |
1dea84ae MT |
6978 | if (free_size == 0) { |
6979 | dprintf("imsm: Cannot find free size.\n"); | |
6980 | free(e); | |
a7dd165b | 6981 | return 0; |
1dea84ae | 6982 | } |
a7dd165b | 6983 | |
1dea84ae MT |
6984 | if (!expanding && extent_idx != 0) |
6985 | /* | |
6986 | * Not a real first volume in a container is created, pre_reservation is needed. | |
6987 | */ | |
6988 | pre_reservation = IMSM_RESERVED_SECTORS; | |
0dcecb2e | 6989 | |
1dea84ae MT |
6990 | if (e[extent_idx].size == 0) |
6991 | /* | |
6992 | * extent_idx points to the metadata, post_reservation is allready done. | |
6993 | */ | |
6994 | post_reservation = 0; | |
6995 | free(e); | |
0dcecb2e | 6996 | |
1dea84ae MT |
6997 | reservation_size = pre_reservation + post_reservation; |
6998 | ||
6999 | if (free_size < reservation_size) { | |
7000 | dprintf("imsm: Reservation size is greater than free space.\n"); | |
7001 | return 0; | |
7002 | } | |
0dcecb2e | 7003 | |
1dea84ae MT |
7004 | super->create_offset = start + pre_reservation; |
7005 | return free_size - reservation_size; | |
0dcecb2e DW |
7006 | } |
7007 | ||
88c32bb1 DW |
7008 | static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks) |
7009 | { | |
7010 | if (level < 0 || level == 6 || level == 4) | |
7011 | return 0; | |
7012 | ||
7013 | /* if we have an orom prevent invalid raid levels */ | |
7014 | if (orom) | |
7015 | switch (level) { | |
7016 | case 0: return imsm_orom_has_raid0(orom); | |
7017 | case 1: | |
7018 | if (raiddisks > 2) | |
7019 | return imsm_orom_has_raid1e(orom); | |
1c556e92 DW |
7020 | return imsm_orom_has_raid1(orom) && raiddisks == 2; |
7021 | case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4; | |
7022 | case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2; | |
88c32bb1 DW |
7023 | } |
7024 | else | |
7025 | return 1; /* not on an Intel RAID platform so anything goes */ | |
7026 | ||
7027 | return 0; | |
7028 | } | |
7029 | ||
ca9de185 LM |
7030 | static int |
7031 | active_arrays_by_format(char *name, char* hba, struct md_list **devlist, | |
7032 | int dpa, int verbose) | |
7033 | { | |
7034 | struct mdstat_ent *mdstat = mdstat_read(0, 0); | |
594dc1b8 | 7035 | struct mdstat_ent *memb; |
ca9de185 LM |
7036 | int count = 0; |
7037 | int num = 0; | |
594dc1b8 | 7038 | struct md_list *dv; |
ca9de185 LM |
7039 | int found; |
7040 | ||
7041 | for (memb = mdstat ; memb ; memb = memb->next) { | |
7042 | if (memb->metadata_version && | |
fc54fe7a | 7043 | (strncmp(memb->metadata_version, "external:", 9) == 0) && |
ca9de185 LM |
7044 | (strcmp(&memb->metadata_version[9], name) == 0) && |
7045 | !is_subarray(memb->metadata_version+9) && | |
7046 | memb->members) { | |
7047 | struct dev_member *dev = memb->members; | |
7048 | int fd = -1; | |
4389ce73 | 7049 | while (dev && !is_fd_valid(fd)) { |
503975b9 | 7050 | char *path = xmalloc(strlen(dev->name) + strlen("/dev/") + 1); |
dd5ab402 | 7051 | num = snprintf(path, PATH_MAX, "%s%s", "/dev/", dev->name); |
503975b9 N |
7052 | if (num > 0) |
7053 | fd = open(path, O_RDONLY, 0); | |
4389ce73 | 7054 | if (num <= 0 || !is_fd_valid(fd)) { |
676e87a8 | 7055 | pr_vrb("Cannot open %s: %s\n", |
503975b9 | 7056 | dev->name, strerror(errno)); |
ca9de185 | 7057 | } |
503975b9 | 7058 | free(path); |
ca9de185 LM |
7059 | dev = dev->next; |
7060 | } | |
7061 | found = 0; | |
4389ce73 | 7062 | if (is_fd_valid(fd) && disk_attached_to_hba(fd, hba)) { |
ca9de185 LM |
7063 | struct mdstat_ent *vol; |
7064 | for (vol = mdstat ; vol ; vol = vol->next) { | |
089f9d79 | 7065 | if (vol->active > 0 && |
ca9de185 | 7066 | vol->metadata_version && |
9581efb1 | 7067 | is_container_member(vol, memb->devnm)) { |
ca9de185 LM |
7068 | found++; |
7069 | count++; | |
7070 | } | |
7071 | } | |
7072 | if (*devlist && (found < dpa)) { | |
503975b9 | 7073 | dv = xcalloc(1, sizeof(*dv)); |
9581efb1 N |
7074 | dv->devname = xmalloc(strlen(memb->devnm) + strlen("/dev/") + 1); |
7075 | sprintf(dv->devname, "%s%s", "/dev/", memb->devnm); | |
503975b9 N |
7076 | dv->found = found; |
7077 | dv->used = 0; | |
7078 | dv->next = *devlist; | |
7079 | *devlist = dv; | |
ca9de185 LM |
7080 | } |
7081 | } | |
4389ce73 | 7082 | close_fd(&fd); |
ca9de185 LM |
7083 | } |
7084 | } | |
7085 | free_mdstat(mdstat); | |
7086 | return count; | |
7087 | } | |
7088 | ||
7089 | #ifdef DEBUG_LOOP | |
7090 | static struct md_list* | |
7091 | get_loop_devices(void) | |
7092 | { | |
7093 | int i; | |
7094 | struct md_list *devlist = NULL; | |
594dc1b8 | 7095 | struct md_list *dv; |
ca9de185 LM |
7096 | |
7097 | for(i = 0; i < 12; i++) { | |
503975b9 N |
7098 | dv = xcalloc(1, sizeof(*dv)); |
7099 | dv->devname = xmalloc(40); | |
ca9de185 LM |
7100 | sprintf(dv->devname, "/dev/loop%d", i); |
7101 | dv->next = devlist; | |
7102 | devlist = dv; | |
7103 | } | |
7104 | return devlist; | |
7105 | } | |
7106 | #endif | |
7107 | ||
7108 | static struct md_list* | |
7109 | get_devices(const char *hba_path) | |
7110 | { | |
7111 | struct md_list *devlist = NULL; | |
594dc1b8 | 7112 | struct md_list *dv; |
ca9de185 LM |
7113 | struct dirent *ent; |
7114 | DIR *dir; | |
7115 | int err = 0; | |
7116 | ||
7117 | #if DEBUG_LOOP | |
7118 | devlist = get_loop_devices(); | |
7119 | return devlist; | |
7120 | #endif | |
7121 | /* scroll through /sys/dev/block looking for devices attached to | |
7122 | * this hba | |
7123 | */ | |
7124 | dir = opendir("/sys/dev/block"); | |
7125 | for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) { | |
7126 | int fd; | |
7127 | char buf[1024]; | |
7128 | int major, minor; | |
7129 | char *path = NULL; | |
7130 | if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2) | |
7131 | continue; | |
7c798f87 | 7132 | path = devt_to_devpath(makedev(major, minor), 1, NULL); |
ca9de185 LM |
7133 | if (!path) |
7134 | continue; | |
7135 | if (!path_attached_to_hba(path, hba_path)) { | |
7136 | free(path); | |
7137 | path = NULL; | |
7138 | continue; | |
7139 | } | |
7140 | free(path); | |
7141 | path = NULL; | |
7142 | fd = dev_open(ent->d_name, O_RDONLY); | |
4389ce73 | 7143 | if (is_fd_valid(fd)) { |
ca9de185 LM |
7144 | fd2devname(fd, buf); |
7145 | close(fd); | |
7146 | } else { | |
e7b84f9d | 7147 | pr_err("cannot open device: %s\n", |
ca9de185 LM |
7148 | ent->d_name); |
7149 | continue; | |
7150 | } | |
7151 | ||
503975b9 N |
7152 | dv = xcalloc(1, sizeof(*dv)); |
7153 | dv->devname = xstrdup(buf); | |
ca9de185 LM |
7154 | dv->next = devlist; |
7155 | devlist = dv; | |
7156 | } | |
7157 | if (err) { | |
7158 | while(devlist) { | |
7159 | dv = devlist; | |
7160 | devlist = devlist->next; | |
7161 | free(dv->devname); | |
7162 | free(dv); | |
7163 | } | |
7164 | } | |
562aa102 | 7165 | closedir(dir); |
ca9de185 LM |
7166 | return devlist; |
7167 | } | |
7168 | ||
7169 | static int | |
7170 | count_volumes_list(struct md_list *devlist, char *homehost, | |
7171 | int verbose, int *found) | |
7172 | { | |
7173 | struct md_list *tmpdev; | |
7174 | int count = 0; | |
594dc1b8 | 7175 | struct supertype *st; |
ca9de185 LM |
7176 | |
7177 | /* first walk the list of devices to find a consistent set | |
7178 | * that match the criterea, if that is possible. | |
7179 | * We flag the ones we like with 'used'. | |
7180 | */ | |
7181 | *found = 0; | |
7182 | st = match_metadata_desc_imsm("imsm"); | |
7183 | if (st == NULL) { | |
676e87a8 | 7184 | pr_vrb("cannot allocate memory for imsm supertype\n"); |
ca9de185 LM |
7185 | return 0; |
7186 | } | |
7187 | ||
7188 | for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { | |
7189 | char *devname = tmpdev->devname; | |
0a6bff09 | 7190 | dev_t rdev; |
ca9de185 LM |
7191 | struct supertype *tst; |
7192 | int dfd; | |
7193 | if (tmpdev->used > 1) | |
7194 | continue; | |
7195 | tst = dup_super(st); | |
7196 | if (tst == NULL) { | |
676e87a8 | 7197 | pr_vrb("cannot allocate memory for imsm supertype\n"); |
ca9de185 LM |
7198 | goto err_1; |
7199 | } | |
7200 | tmpdev->container = 0; | |
7201 | dfd = dev_open(devname, O_RDONLY|O_EXCL); | |
4389ce73 | 7202 | if (!is_fd_valid(dfd)) { |
1ade5cc1 | 7203 | dprintf("cannot open device %s: %s\n", |
ca9de185 LM |
7204 | devname, strerror(errno)); |
7205 | tmpdev->used = 2; | |
0a6bff09 | 7206 | } else if (!fstat_is_blkdev(dfd, devname, &rdev)) { |
ca9de185 LM |
7207 | tmpdev->used = 2; |
7208 | } else if (must_be_container(dfd)) { | |
7209 | struct supertype *cst; | |
7210 | cst = super_by_fd(dfd, NULL); | |
7211 | if (cst == NULL) { | |
1ade5cc1 | 7212 | dprintf("cannot recognize container type %s\n", |
ca9de185 LM |
7213 | devname); |
7214 | tmpdev->used = 2; | |
7215 | } else if (tst->ss != st->ss) { | |
1ade5cc1 | 7216 | dprintf("non-imsm container - ignore it: %s\n", |
ca9de185 LM |
7217 | devname); |
7218 | tmpdev->used = 2; | |
7219 | } else if (!tst->ss->load_container || | |
7220 | tst->ss->load_container(tst, dfd, NULL)) | |
7221 | tmpdev->used = 2; | |
7222 | else { | |
7223 | tmpdev->container = 1; | |
7224 | } | |
7225 | if (cst) | |
7226 | cst->ss->free_super(cst); | |
7227 | } else { | |
0a6bff09 | 7228 | tmpdev->st_rdev = rdev; |
ca9de185 | 7229 | if (tst->ss->load_super(tst,dfd, NULL)) { |
1ade5cc1 | 7230 | dprintf("no RAID superblock on %s\n", |
ca9de185 LM |
7231 | devname); |
7232 | tmpdev->used = 2; | |
7233 | } else if (tst->ss->compare_super == NULL) { | |
1ade5cc1 | 7234 | dprintf("Cannot assemble %s metadata on %s\n", |
ca9de185 LM |
7235 | tst->ss->name, devname); |
7236 | tmpdev->used = 2; | |
7237 | } | |
7238 | } | |
4389ce73 MT |
7239 | close_fd(&dfd); |
7240 | ||
ca9de185 LM |
7241 | if (tmpdev->used == 2 || tmpdev->used == 4) { |
7242 | /* Ignore unrecognised devices during auto-assembly */ | |
7243 | goto loop; | |
7244 | } | |
7245 | else { | |
7246 | struct mdinfo info; | |
7247 | tst->ss->getinfo_super(tst, &info, NULL); | |
7248 | ||
7249 | if (st->minor_version == -1) | |
7250 | st->minor_version = tst->minor_version; | |
7251 | ||
7252 | if (memcmp(info.uuid, uuid_zero, | |
7253 | sizeof(int[4])) == 0) { | |
7254 | /* this is a floating spare. It cannot define | |
7255 | * an array unless there are no more arrays of | |
7256 | * this type to be found. It can be included | |
7257 | * in an array of this type though. | |
7258 | */ | |
7259 | tmpdev->used = 3; | |
7260 | goto loop; | |
7261 | } | |
7262 | ||
7263 | if (st->ss != tst->ss || | |
7264 | st->minor_version != tst->minor_version || | |
c7b8547c | 7265 | st->ss->compare_super(st, tst, 1) != 0) { |
ca9de185 LM |
7266 | /* Some mismatch. If exactly one array matches this host, |
7267 | * we can resolve on that one. | |
7268 | * Or, if we are auto assembling, we just ignore the second | |
7269 | * for now. | |
7270 | */ | |
1ade5cc1 | 7271 | dprintf("superblock on %s doesn't match others - assembly aborted\n", |
ca9de185 LM |
7272 | devname); |
7273 | goto loop; | |
7274 | } | |
7275 | tmpdev->used = 1; | |
7276 | *found = 1; | |
7277 | dprintf("found: devname: %s\n", devname); | |
7278 | } | |
7279 | loop: | |
7280 | if (tst) | |
7281 | tst->ss->free_super(tst); | |
7282 | } | |
7283 | if (*found != 0) { | |
7284 | int err; | |
7285 | if ((err = load_super_imsm_all(st, -1, &st->sb, NULL, devlist, 0)) == 0) { | |
7286 | struct mdinfo *iter, *head = st->ss->container_content(st, NULL); | |
7287 | for (iter = head; iter; iter = iter->next) { | |
7288 | dprintf("content->text_version: %s vol\n", | |
7289 | iter->text_version); | |
7290 | if (iter->array.state & (1<<MD_SB_BLOCK_VOLUME)) { | |
7291 | /* do not assemble arrays with unsupported | |
7292 | configurations */ | |
1ade5cc1 | 7293 | dprintf("Cannot activate member %s.\n", |
ca9de185 LM |
7294 | iter->text_version); |
7295 | } else | |
7296 | count++; | |
7297 | } | |
7298 | sysfs_free(head); | |
7299 | ||
7300 | } else { | |
1ade5cc1 | 7301 | dprintf("No valid super block on device list: err: %d %p\n", |
ca9de185 LM |
7302 | err, st->sb); |
7303 | } | |
7304 | } else { | |
1ade5cc1 | 7305 | dprintf("no more devices to examine\n"); |
ca9de185 LM |
7306 | } |
7307 | ||
7308 | for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { | |
089f9d79 | 7309 | if (tmpdev->used == 1 && tmpdev->found) { |
ca9de185 LM |
7310 | if (count) { |
7311 | if (count < tmpdev->found) | |
7312 | count = 0; | |
7313 | else | |
7314 | count -= tmpdev->found; | |
7315 | } | |
7316 | } | |
7317 | if (tmpdev->used == 1) | |
7318 | tmpdev->used = 4; | |
7319 | } | |
7320 | err_1: | |
7321 | if (st) | |
7322 | st->ss->free_super(st); | |
7323 | return count; | |
7324 | } | |
7325 | ||
d3c11416 AO |
7326 | static int __count_volumes(char *hba_path, int dpa, int verbose, |
7327 | int cmp_hba_path) | |
ca9de185 | 7328 | { |
72a45777 | 7329 | struct sys_dev *idev, *intel_devices = find_intel_devices(); |
ca9de185 | 7330 | int count = 0; |
72a45777 PB |
7331 | const struct orom_entry *entry; |
7332 | struct devid_list *dv, *devid_list; | |
ca9de185 | 7333 | |
d3c11416 | 7334 | if (!hba_path) |
ca9de185 LM |
7335 | return 0; |
7336 | ||
72a45777 | 7337 | for (idev = intel_devices; idev; idev = idev->next) { |
d3c11416 AO |
7338 | if (strstr(idev->path, hba_path)) |
7339 | break; | |
72a45777 PB |
7340 | } |
7341 | ||
7342 | if (!idev || !idev->dev_id) | |
ca9de185 | 7343 | return 0; |
72a45777 PB |
7344 | |
7345 | entry = get_orom_entry_by_device_id(idev->dev_id); | |
7346 | ||
7347 | if (!entry || !entry->devid_list) | |
7348 | return 0; | |
7349 | ||
7350 | devid_list = entry->devid_list; | |
7351 | for (dv = devid_list; dv; dv = dv->next) { | |
594dc1b8 | 7352 | struct md_list *devlist; |
d3c11416 AO |
7353 | struct sys_dev *device = NULL; |
7354 | char *hpath; | |
72a45777 PB |
7355 | int found = 0; |
7356 | ||
d3c11416 AO |
7357 | if (cmp_hba_path) |
7358 | device = device_by_id_and_path(dv->devid, hba_path); | |
7359 | else | |
7360 | device = device_by_id(dv->devid); | |
7361 | ||
72a45777 | 7362 | if (device) |
d3c11416 | 7363 | hpath = device->path; |
72a45777 PB |
7364 | else |
7365 | return 0; | |
7366 | ||
d3c11416 | 7367 | devlist = get_devices(hpath); |
72a45777 PB |
7368 | /* if no intel devices return zero volumes */ |
7369 | if (devlist == NULL) | |
7370 | return 0; | |
7371 | ||
d3c11416 AO |
7372 | count += active_arrays_by_format("imsm", hpath, &devlist, dpa, |
7373 | verbose); | |
7374 | dprintf("path: %s active arrays: %d\n", hpath, count); | |
72a45777 PB |
7375 | if (devlist == NULL) |
7376 | return 0; | |
7377 | do { | |
7378 | found = 0; | |
7379 | count += count_volumes_list(devlist, | |
7380 | NULL, | |
7381 | verbose, | |
7382 | &found); | |
7383 | dprintf("found %d count: %d\n", found, count); | |
7384 | } while (found); | |
7385 | ||
d3c11416 | 7386 | dprintf("path: %s total number of volumes: %d\n", hpath, count); |
72a45777 PB |
7387 | |
7388 | while (devlist) { | |
7389 | struct md_list *dv = devlist; | |
7390 | devlist = devlist->next; | |
7391 | free(dv->devname); | |
7392 | free(dv); | |
7393 | } | |
ca9de185 LM |
7394 | } |
7395 | return count; | |
7396 | } | |
7397 | ||
d3c11416 AO |
7398 | static int count_volumes(struct intel_hba *hba, int dpa, int verbose) |
7399 | { | |
7400 | if (!hba) | |
7401 | return 0; | |
7402 | if (hba->type == SYS_DEV_VMD) { | |
7403 | struct sys_dev *dev; | |
7404 | int count = 0; | |
7405 | ||
7406 | for (dev = find_intel_devices(); dev; dev = dev->next) { | |
7407 | if (dev->type == SYS_DEV_VMD) | |
7408 | count += __count_volumes(dev->path, dpa, | |
7409 | verbose, 1); | |
7410 | } | |
7411 | return count; | |
7412 | } | |
7413 | return __count_volumes(hba->path, dpa, verbose, 0); | |
7414 | } | |
7415 | ||
cd9d1ac7 DW |
7416 | static int imsm_default_chunk(const struct imsm_orom *orom) |
7417 | { | |
7418 | /* up to 512 if the plaform supports it, otherwise the platform max. | |
7419 | * 128 if no platform detected | |
7420 | */ | |
7421 | int fs = max(7, orom ? fls(orom->sss) : 0); | |
7422 | ||
7423 | return min(512, (1 << fs)); | |
7424 | } | |
73408129 | 7425 | |
6592ce37 DW |
7426 | static int |
7427 | validate_geometry_imsm_orom(struct intel_super *super, int level, int layout, | |
2cc699af | 7428 | int raiddisks, int *chunk, unsigned long long size, int verbose) |
6592ce37 | 7429 | { |
660260d0 DW |
7430 | /* check/set platform and metadata limits/defaults */ |
7431 | if (super->orom && raiddisks > super->orom->dpa) { | |
676e87a8 | 7432 | pr_vrb("platform supports a maximum of %d disks per array\n", |
660260d0 | 7433 | super->orom->dpa); |
73408129 LM |
7434 | return 0; |
7435 | } | |
7436 | ||
5d500228 | 7437 | /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */ |
660260d0 | 7438 | if (!is_raid_level_supported(super->orom, level, raiddisks)) { |
676e87a8 | 7439 | pr_vrb("platform does not support raid%d with %d disk%s\n", |
6592ce37 DW |
7440 | level, raiddisks, raiddisks > 1 ? "s" : ""); |
7441 | return 0; | |
7442 | } | |
cd9d1ac7 | 7443 | |
7ccc4cc4 | 7444 | if (*chunk == 0 || *chunk == UnSet) |
cd9d1ac7 DW |
7445 | *chunk = imsm_default_chunk(super->orom); |
7446 | ||
7ccc4cc4 | 7447 | if (super->orom && !imsm_orom_has_chunk(super->orom, *chunk)) { |
676e87a8 | 7448 | pr_vrb("platform does not support a chunk size of: %d\n", *chunk); |
cd9d1ac7 | 7449 | return 0; |
6592ce37 | 7450 | } |
cd9d1ac7 | 7451 | |
6592ce37 DW |
7452 | if (layout != imsm_level_to_layout(level)) { |
7453 | if (level == 5) | |
676e87a8 | 7454 | pr_vrb("imsm raid 5 only supports the left-asymmetric layout\n"); |
6592ce37 | 7455 | else if (level == 10) |
676e87a8 | 7456 | pr_vrb("imsm raid 10 only supports the n2 layout\n"); |
6592ce37 | 7457 | else |
676e87a8 | 7458 | pr_vrb("imsm unknown layout %#x for this raid level %d\n", |
6592ce37 DW |
7459 | layout, level); |
7460 | return 0; | |
7461 | } | |
2cc699af | 7462 | |
7ccc4cc4 | 7463 | if (super->orom && (super->orom->attr & IMSM_OROM_ATTR_2TB) == 0 && |
2cc699af | 7464 | (calc_array_size(level, raiddisks, layout, *chunk, size) >> 32) > 0) { |
676e87a8 | 7465 | pr_vrb("platform does not support a volume size over 2TB\n"); |
2cc699af CA |
7466 | return 0; |
7467 | } | |
614902f6 | 7468 | |
6592ce37 DW |
7469 | return 1; |
7470 | } | |
7471 | ||
1011e834 | 7472 | /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd |
c2c087e6 DW |
7473 | * FIX ME add ahci details |
7474 | */ | |
8b353278 | 7475 | static int validate_geometry_imsm_volume(struct supertype *st, int level, |
c21e737b | 7476 | int layout, int raiddisks, int *chunk, |
af4348dd N |
7477 | unsigned long long size, |
7478 | unsigned long long data_offset, | |
7479 | char *dev, | |
2c514b71 NB |
7480 | unsigned long long *freesize, |
7481 | int verbose) | |
cdddbdbc | 7482 | { |
9e04ac1c | 7483 | dev_t rdev; |
c2c087e6 | 7484 | struct intel_super *super = st->sb; |
b2916f25 | 7485 | struct imsm_super *mpb; |
c2c087e6 DW |
7486 | struct dl *dl; |
7487 | unsigned long long pos = 0; | |
7488 | unsigned long long maxsize; | |
7489 | struct extent *e; | |
7490 | int i; | |
cdddbdbc | 7491 | |
88c32bb1 DW |
7492 | /* We must have the container info already read in. */ |
7493 | if (!super) | |
c2c087e6 DW |
7494 | return 0; |
7495 | ||
b2916f25 JS |
7496 | mpb = super->anchor; |
7497 | ||
2cc699af | 7498 | if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, size, verbose)) { |
3e684231 | 7499 | pr_err("RAID geometry validation failed. Cannot proceed with the action(s).\n"); |
c2c087e6 | 7500 | return 0; |
d54559f0 | 7501 | } |
c2c087e6 DW |
7502 | if (!dev) { |
7503 | /* General test: make sure there is space for | |
2da8544a DW |
7504 | * 'raiddisks' device extents of size 'size' at a given |
7505 | * offset | |
c2c087e6 | 7506 | */ |
e46273eb | 7507 | unsigned long long minsize = size; |
b7528a20 | 7508 | unsigned long long start_offset = MaxSector; |
c2c087e6 DW |
7509 | int dcnt = 0; |
7510 | if (minsize == 0) | |
7511 | minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
7512 | for (dl = super->disks; dl ; dl = dl->next) { | |
7513 | int found = 0; | |
7514 | ||
bf5a934a | 7515 | pos = 0; |
c2c087e6 | 7516 | i = 0; |
05501181 | 7517 | e = get_extents(super, dl, 0); |
c2c087e6 DW |
7518 | if (!e) continue; |
7519 | do { | |
7520 | unsigned long long esize; | |
7521 | esize = e[i].start - pos; | |
7522 | if (esize >= minsize) | |
7523 | found = 1; | |
b7528a20 | 7524 | if (found && start_offset == MaxSector) { |
2da8544a DW |
7525 | start_offset = pos; |
7526 | break; | |
7527 | } else if (found && pos != start_offset) { | |
7528 | found = 0; | |
7529 | break; | |
7530 | } | |
c2c087e6 DW |
7531 | pos = e[i].start + e[i].size; |
7532 | i++; | |
7533 | } while (e[i-1].size); | |
7534 | if (found) | |
7535 | dcnt++; | |
7536 | free(e); | |
7537 | } | |
7538 | if (dcnt < raiddisks) { | |
2c514b71 | 7539 | if (verbose) |
7a862a02 | 7540 | pr_err("imsm: Not enough devices with space for this array (%d < %d)\n", |
2c514b71 | 7541 | dcnt, raiddisks); |
c2c087e6 DW |
7542 | return 0; |
7543 | } | |
7544 | return 1; | |
7545 | } | |
0dcecb2e | 7546 | |
c2c087e6 | 7547 | /* This device must be a member of the set */ |
9e04ac1c | 7548 | if (!stat_is_blkdev(dev, &rdev)) |
c2c087e6 DW |
7549 | return 0; |
7550 | for (dl = super->disks ; dl ; dl = dl->next) { | |
9e04ac1c ZL |
7551 | if (dl->major == (int)major(rdev) && |
7552 | dl->minor == (int)minor(rdev)) | |
c2c087e6 DW |
7553 | break; |
7554 | } | |
7555 | if (!dl) { | |
2c514b71 | 7556 | if (verbose) |
7a862a02 | 7557 | pr_err("%s is not in the same imsm set\n", dev); |
c2c087e6 | 7558 | return 0; |
a20d2ba5 DW |
7559 | } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) { |
7560 | /* If a volume is present then the current creation attempt | |
7561 | * cannot incorporate new spares because the orom may not | |
7562 | * understand this configuration (all member disks must be | |
7563 | * members of each array in the container). | |
7564 | */ | |
7a862a02 N |
7565 | pr_err("%s is a spare and a volume is already defined for this container\n", dev); |
7566 | pr_err("The option-rom requires all member disks to be a member of all volumes\n"); | |
a20d2ba5 | 7567 | return 0; |
5fe62b94 WD |
7568 | } else if (super->orom && mpb->num_raid_devs > 0 && |
7569 | mpb->num_disks != raiddisks) { | |
7a862a02 | 7570 | pr_err("The option-rom requires all member disks to be a member of all volumes\n"); |
5fe62b94 | 7571 | return 0; |
c2c087e6 | 7572 | } |
0dcecb2e DW |
7573 | |
7574 | /* retrieve the largest free space block */ | |
05501181 | 7575 | e = get_extents(super, dl, 0); |
c2c087e6 DW |
7576 | maxsize = 0; |
7577 | i = 0; | |
0dcecb2e DW |
7578 | if (e) { |
7579 | do { | |
7580 | unsigned long long esize; | |
7581 | ||
7582 | esize = e[i].start - pos; | |
7583 | if (esize >= maxsize) | |
7584 | maxsize = esize; | |
7585 | pos = e[i].start + e[i].size; | |
7586 | i++; | |
7587 | } while (e[i-1].size); | |
7588 | dl->e = e; | |
7589 | dl->extent_cnt = i; | |
7590 | } else { | |
7591 | if (verbose) | |
e7b84f9d | 7592 | pr_err("unable to determine free space for: %s\n", |
0dcecb2e DW |
7593 | dev); |
7594 | return 0; | |
7595 | } | |
7596 | if (maxsize < size) { | |
7597 | if (verbose) | |
e7b84f9d | 7598 | pr_err("%s not enough space (%llu < %llu)\n", |
0dcecb2e DW |
7599 | dev, maxsize, size); |
7600 | return 0; | |
7601 | } | |
7602 | ||
aa19fdd4 | 7603 | maxsize = merge_extents(super, false); |
3baa56ab | 7604 | |
1a1ced1e KS |
7605 | if (mpb->num_raid_devs > 0 && size && size != maxsize) |
7606 | pr_err("attempting to create a second volume with size less then remaining space.\n"); | |
3baa56ab | 7607 | |
a7dd165b | 7608 | if (maxsize < size || maxsize == 0) { |
b3071342 LD |
7609 | if (verbose) { |
7610 | if (maxsize == 0) | |
7a862a02 | 7611 | pr_err("no free space left on device. Aborting...\n"); |
b3071342 | 7612 | else |
7a862a02 | 7613 | pr_err("not enough space to create volume of given size (%llu < %llu). Aborting...\n", |
b3071342 LD |
7614 | maxsize, size); |
7615 | } | |
0dcecb2e | 7616 | return 0; |
0dcecb2e DW |
7617 | } |
7618 | ||
c2c087e6 DW |
7619 | *freesize = maxsize; |
7620 | ||
ca9de185 | 7621 | if (super->orom) { |
72a45777 | 7622 | int count = count_volumes(super->hba, |
ca9de185 LM |
7623 | super->orom->dpa, verbose); |
7624 | if (super->orom->vphba <= count) { | |
676e87a8 | 7625 | pr_vrb("platform does not support more than %d raid volumes.\n", |
ca9de185 LM |
7626 | super->orom->vphba); |
7627 | return 0; | |
7628 | } | |
7629 | } | |
c2c087e6 | 7630 | return 1; |
cdddbdbc DW |
7631 | } |
7632 | ||
6d4d9ab2 MT |
7633 | /** |
7634 | * imsm_get_free_size() - get the biggest, common free space from members. | |
7635 | * @super: &intel_super pointer, not NULL. | |
7636 | * @raiddisks: number of raid disks. | |
7637 | * @size: requested size, could be 0 (means max size). | |
5f027b93 | 7638 | * @chunk: requested chunk size in KiB. |
6d4d9ab2 MT |
7639 | * @freesize: pointer for returned size value. |
7640 | * | |
7641 | * Return: &IMSM_STATUS_OK or &IMSM_STATUS_ERROR. | |
7642 | * | |
7643 | * @freesize is set to meaningful value, this can be @size, or calculated | |
7644 | * max free size. | |
7645 | * super->create_offset value is modified and set appropriately in | |
7646 | * merge_extends() for further creation. | |
7647 | */ | |
7648 | static imsm_status_t imsm_get_free_size(struct intel_super *super, | |
7649 | const int raiddisks, | |
7650 | unsigned long long size, | |
7651 | const int chunk, | |
aa19fdd4 MT |
7652 | unsigned long long *freesize, |
7653 | bool expanding) | |
efb30e7f | 7654 | { |
efb30e7f DW |
7655 | struct imsm_super *mpb = super->anchor; |
7656 | struct dl *dl; | |
7657 | int i; | |
efb30e7f | 7658 | struct extent *e; |
5f027b93 MT |
7659 | int cnt = 0; |
7660 | int used = 0; | |
efb30e7f | 7661 | unsigned long long maxsize; |
5f027b93 MT |
7662 | unsigned long long minsize = size; |
7663 | ||
7664 | if (minsize == 0) | |
7665 | minsize = chunk * 2; | |
efb30e7f DW |
7666 | |
7667 | /* find the largest common start free region of the possible disks */ | |
efb30e7f DW |
7668 | for (dl = super->disks; dl; dl = dl->next) { |
7669 | dl->raiddisk = -1; | |
7670 | ||
7671 | if (dl->index >= 0) | |
7672 | used++; | |
7673 | ||
7674 | /* don't activate new spares if we are orom constrained | |
7675 | * and there is already a volume active in the container | |
7676 | */ | |
7677 | if (super->orom && dl->index < 0 && mpb->num_raid_devs) | |
7678 | continue; | |
7679 | ||
05501181 | 7680 | e = get_extents(super, dl, 0); |
efb30e7f DW |
7681 | if (!e) |
7682 | continue; | |
7683 | for (i = 1; e[i-1].size; i++) | |
7684 | ; | |
7685 | dl->e = e; | |
7686 | dl->extent_cnt = i; | |
efb30e7f DW |
7687 | cnt++; |
7688 | } | |
7689 | ||
aa19fdd4 | 7690 | maxsize = merge_extents(super, expanding); |
5f027b93 MT |
7691 | if (maxsize < minsize) { |
7692 | pr_err("imsm: Free space is %llu but must be equal or larger than %llu.\n", | |
7693 | maxsize, minsize); | |
7694 | return IMSM_STATUS_ERROR; | |
7695 | } | |
efb30e7f | 7696 | |
5f027b93 MT |
7697 | if (cnt < raiddisks || (super->orom && used && used != raiddisks)) { |
7698 | pr_err("imsm: Not enough devices with space to create array.\n"); | |
6d4d9ab2 | 7699 | return IMSM_STATUS_ERROR; |
efb30e7f DW |
7700 | } |
7701 | ||
7702 | if (size == 0) { | |
7703 | size = maxsize; | |
7704 | if (chunk) { | |
612e59d8 CA |
7705 | size /= 2 * chunk; |
7706 | size *= 2 * chunk; | |
efb30e7f | 7707 | } |
f878b242 LM |
7708 | maxsize = size; |
7709 | } | |
1a1ced1e KS |
7710 | if (mpb->num_raid_devs > 0 && size && size != maxsize) |
7711 | pr_err("attempting to create a second volume with size less then remaining space.\n"); | |
efb30e7f DW |
7712 | *freesize = size; |
7713 | ||
13bcac90 AK |
7714 | dprintf("imsm: imsm_get_free_size() returns : %llu\n", size); |
7715 | ||
6d4d9ab2 | 7716 | return IMSM_STATUS_OK; |
efb30e7f DW |
7717 | } |
7718 | ||
6d4d9ab2 MT |
7719 | /** |
7720 | * autolayout_imsm() - automatically layout a new volume. | |
7721 | * @super: &intel_super pointer, not NULL. | |
7722 | * @raiddisks: number of raid disks. | |
7723 | * @size: requested size, could be 0 (means max size). | |
7724 | * @chunk: requested chunk. | |
7725 | * @freesize: pointer for returned size value. | |
7726 | * | |
7727 | * We are being asked to automatically layout a new volume based on the current | |
7728 | * contents of the container. If the parameters can be satisfied autolayout_imsm | |
7729 | * will record the disks, start offset, and will return size of the volume to | |
7730 | * be created. See imsm_get_free_size() for details. | |
7731 | * add_to_super() and getinfo_super() detect when autolayout is in progress. | |
7732 | * If first volume exists, slots are set consistently to it. | |
7733 | * | |
7734 | * Return: &IMSM_STATUS_OK on success, &IMSM_STATUS_ERROR otherwise. | |
7735 | * | |
7736 | * Disks are marked for creation via dl->raiddisk. | |
7737 | */ | |
7738 | static imsm_status_t autolayout_imsm(struct intel_super *super, | |
7739 | const int raiddisks, | |
7740 | unsigned long long size, const int chunk, | |
7741 | unsigned long long *freesize) | |
13bcac90 | 7742 | { |
6d4d9ab2 MT |
7743 | int curr_slot = 0; |
7744 | struct dl *disk; | |
7745 | int vol_cnt = super->anchor->num_raid_devs; | |
7746 | imsm_status_t rv; | |
13bcac90 | 7747 | |
aa19fdd4 | 7748 | rv = imsm_get_free_size(super, raiddisks, size, chunk, freesize, false); |
6d4d9ab2 MT |
7749 | if (rv != IMSM_STATUS_OK) |
7750 | return IMSM_STATUS_ERROR; | |
7751 | ||
7752 | for (disk = super->disks; disk; disk = disk->next) { | |
7753 | if (!disk->e) | |
7754 | continue; | |
7755 | ||
7756 | if (curr_slot == raiddisks) | |
7757 | break; | |
7758 | ||
7759 | if (vol_cnt == 0) { | |
7760 | disk->raiddisk = curr_slot; | |
7761 | } else { | |
7762 | int _slot = get_disk_slot_in_dev(super, 0, disk->index); | |
7763 | ||
7764 | if (_slot == -1) { | |
7765 | pr_err("Disk %s is not used in first volume, aborting\n", | |
7766 | disk->devname); | |
7767 | return IMSM_STATUS_ERROR; | |
7768 | } | |
7769 | disk->raiddisk = _slot; | |
7770 | } | |
7771 | curr_slot++; | |
13bcac90 AK |
7772 | } |
7773 | ||
6d4d9ab2 | 7774 | return IMSM_STATUS_OK; |
13bcac90 AK |
7775 | } |
7776 | ||
bf5a934a | 7777 | static int validate_geometry_imsm(struct supertype *st, int level, int layout, |
c21e737b | 7778 | int raiddisks, int *chunk, unsigned long long size, |
af4348dd | 7779 | unsigned long long data_offset, |
bf5a934a | 7780 | char *dev, unsigned long long *freesize, |
5308f117 | 7781 | int consistency_policy, int verbose) |
bf5a934a DW |
7782 | { |
7783 | int fd, cfd; | |
7784 | struct mdinfo *sra; | |
20cbe8d2 | 7785 | int is_member = 0; |
bf5a934a | 7786 | |
d54559f0 LM |
7787 | /* load capability |
7788 | * if given unused devices create a container | |
bf5a934a DW |
7789 | * if given given devices in a container create a member volume |
7790 | */ | |
6f2af6a4 | 7791 | if (is_container(level)) |
bf5a934a | 7792 | /* Must be a fresh device to add to a container */ |
1f5d54a0 MT |
7793 | return validate_geometry_imsm_container(st, level, raiddisks, |
7794 | data_offset, dev, | |
7795 | freesize, verbose); | |
9587c373 | 7796 | |
06a6101c BK |
7797 | /* |
7798 | * Size is given in sectors. | |
7799 | */ | |
7800 | if (size && (size < 2048)) { | |
22dc741f | 7801 | pr_err("Given size must be greater than 1M.\n"); |
54865c30 RS |
7802 | /* Depends on algorithm in Create.c : |
7803 | * if container was given (dev == NULL) return -1, | |
7804 | * if block device was given ( dev != NULL) return 0. | |
7805 | */ | |
7806 | return dev ? -1 : 0; | |
7807 | } | |
7808 | ||
8592f29d | 7809 | if (!dev) { |
6d4d9ab2 MT |
7810 | struct intel_super *super = st->sb; |
7811 | ||
7812 | /* | |
071f839e | 7813 | * Autolayout mode, st->sb must be set. |
6d4d9ab2 | 7814 | */ |
071f839e KT |
7815 | if (!super) { |
7816 | pr_vrb("superblock must be set for autolayout, aborting\n"); | |
7817 | return 0; | |
6d4d9ab2 MT |
7818 | } |
7819 | ||
7820 | if (!validate_geometry_imsm_orom(st->sb, level, layout, | |
7821 | raiddisks, chunk, size, | |
7822 | verbose)) | |
7823 | return 0; | |
7824 | ||
071f839e | 7825 | if (super->orom && freesize) { |
6d4d9ab2 MT |
7826 | imsm_status_t rv; |
7827 | int count = count_volumes(super->hba, super->orom->dpa, | |
7828 | verbose); | |
7829 | if (super->orom->vphba <= count) { | |
7830 | pr_vrb("platform does not support more than %d raid volumes.\n", | |
7831 | super->orom->vphba); | |
e91a3bad | 7832 | return 0; |
ca9de185 | 7833 | } |
6d4d9ab2 MT |
7834 | |
7835 | rv = autolayout_imsm(super, raiddisks, size, *chunk, | |
7836 | freesize); | |
7837 | if (rv != IMSM_STATUS_OK) | |
7838 | return 0; | |
8592f29d N |
7839 | } |
7840 | return 1; | |
7841 | } | |
bf5a934a DW |
7842 | if (st->sb) { |
7843 | /* creating in a given container */ | |
7844 | return validate_geometry_imsm_volume(st, level, layout, | |
7845 | raiddisks, chunk, size, | |
af4348dd | 7846 | data_offset, |
bf5a934a DW |
7847 | dev, freesize, verbose); |
7848 | } | |
7849 | ||
bf5a934a DW |
7850 | /* This device needs to be a device in an 'imsm' container */ |
7851 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
4389ce73 MT |
7852 | |
7853 | if (is_fd_valid(fd)) { | |
7854 | pr_vrb("Cannot create this array on device %s\n", dev); | |
bf5a934a DW |
7855 | close(fd); |
7856 | return 0; | |
7857 | } | |
4389ce73 MT |
7858 | if (errno == EBUSY) |
7859 | fd = open(dev, O_RDONLY, 0); | |
7860 | ||
7861 | if (!is_fd_valid(fd)) { | |
7862 | pr_vrb("Cannot open %s: %s\n", dev, strerror(errno)); | |
bf5a934a DW |
7863 | return 0; |
7864 | } | |
4389ce73 | 7865 | |
bf5a934a DW |
7866 | /* Well, it is in use by someone, maybe an 'imsm' container. */ |
7867 | cfd = open_container(fd); | |
4389ce73 MT |
7868 | close_fd(&fd); |
7869 | ||
7870 | if (!is_fd_valid(cfd)) { | |
7871 | pr_vrb("Cannot use %s: It is busy\n", dev); | |
bf5a934a DW |
7872 | return 0; |
7873 | } | |
4dd2df09 | 7874 | sra = sysfs_read(cfd, NULL, GET_VERSION); |
bf5a934a | 7875 | if (sra && sra->array.major_version == -1 && |
20cbe8d2 AW |
7876 | strcmp(sra->text_version, "imsm") == 0) |
7877 | is_member = 1; | |
7878 | sysfs_free(sra); | |
7879 | if (is_member) { | |
bf5a934a DW |
7880 | /* This is a member of a imsm container. Load the container |
7881 | * and try to create a volume | |
7882 | */ | |
7883 | struct intel_super *super; | |
7884 | ||
ec50f7b6 | 7885 | if (load_super_imsm_all(st, cfd, (void **) &super, NULL, NULL, 1) == 0) { |
bf5a934a | 7886 | st->sb = super; |
4dd2df09 | 7887 | strcpy(st->container_devnm, fd2devnm(cfd)); |
bf5a934a DW |
7888 | close(cfd); |
7889 | return validate_geometry_imsm_volume(st, level, layout, | |
7890 | raiddisks, chunk, | |
af4348dd | 7891 | size, data_offset, dev, |
ecbd9e81 N |
7892 | freesize, 1) |
7893 | ? 1 : -1; | |
bf5a934a | 7894 | } |
20cbe8d2 | 7895 | } |
bf5a934a | 7896 | |
20cbe8d2 | 7897 | if (verbose) |
e7b84f9d | 7898 | pr_err("failed container membership check\n"); |
20cbe8d2 AW |
7899 | |
7900 | close(cfd); | |
7901 | return 0; | |
bf5a934a | 7902 | } |
0bd16cf2 | 7903 | |
30f58b22 | 7904 | static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk) |
0bd16cf2 DJ |
7905 | { |
7906 | struct intel_super *super = st->sb; | |
7907 | ||
30f58b22 DW |
7908 | if (level && *level == UnSet) |
7909 | *level = LEVEL_CONTAINER; | |
7910 | ||
7911 | if (level && layout && *layout == UnSet) | |
7912 | *layout = imsm_level_to_layout(*level); | |
0bd16cf2 | 7913 | |
cd9d1ac7 DW |
7914 | if (chunk && (*chunk == UnSet || *chunk == 0)) |
7915 | *chunk = imsm_default_chunk(super->orom); | |
0bd16cf2 DJ |
7916 | } |
7917 | ||
33414a01 DW |
7918 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev); |
7919 | ||
3364781b | 7920 | static int kill_subarray_imsm(struct supertype *st, char *subarray_id) |
33414a01 | 7921 | { |
3364781b | 7922 | /* remove the subarray currently referenced by subarray_id */ |
33414a01 DW |
7923 | __u8 i; |
7924 | struct intel_dev **dp; | |
7925 | struct intel_super *super = st->sb; | |
3364781b | 7926 | __u8 current_vol = strtoul(subarray_id, NULL, 10); |
33414a01 DW |
7927 | struct imsm_super *mpb = super->anchor; |
7928 | ||
3364781b | 7929 | if (mpb->num_raid_devs == 0) |
33414a01 | 7930 | return 2; |
33414a01 DW |
7931 | |
7932 | /* block deletions that would change the uuid of active subarrays | |
7933 | * | |
7934 | * FIXME when immutable ids are available, but note that we'll | |
7935 | * also need to fixup the invalidated/active subarray indexes in | |
7936 | * mdstat | |
7937 | */ | |
7938 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
7939 | char subarray[4]; | |
7940 | ||
7941 | if (i < current_vol) | |
7942 | continue; | |
dd5ab402 | 7943 | snprintf(subarray, sizeof(subarray), "%u", i); |
4dd2df09 | 7944 | if (is_subarray_active(subarray, st->devnm)) { |
e7b84f9d N |
7945 | pr_err("deleting subarray-%d would change the UUID of active subarray-%d, aborting\n", |
7946 | current_vol, i); | |
33414a01 DW |
7947 | |
7948 | return 2; | |
7949 | } | |
7950 | } | |
7951 | ||
7952 | if (st->update_tail) { | |
503975b9 | 7953 | struct imsm_update_kill_array *u = xmalloc(sizeof(*u)); |
33414a01 | 7954 | |
33414a01 DW |
7955 | u->type = update_kill_array; |
7956 | u->dev_idx = current_vol; | |
7957 | append_metadata_update(st, u, sizeof(*u)); | |
7958 | ||
7959 | return 0; | |
7960 | } | |
7961 | ||
7962 | for (dp = &super->devlist; *dp;) | |
7963 | if ((*dp)->index == current_vol) { | |
7964 | *dp = (*dp)->next; | |
7965 | } else { | |
7966 | handle_missing(super, (*dp)->dev); | |
7967 | if ((*dp)->index > current_vol) | |
7968 | (*dp)->index--; | |
7969 | dp = &(*dp)->next; | |
7970 | } | |
7971 | ||
7972 | /* no more raid devices, all active components are now spares, | |
7973 | * but of course failed are still failed | |
7974 | */ | |
7975 | if (--mpb->num_raid_devs == 0) { | |
7976 | struct dl *d; | |
7977 | ||
7978 | for (d = super->disks; d; d = d->next) | |
a8619d23 AK |
7979 | if (d->index > -2) |
7980 | mark_spare(d); | |
33414a01 DW |
7981 | } |
7982 | ||
7983 | super->updates_pending++; | |
7984 | ||
7985 | return 0; | |
7986 | } | |
aa534678 | 7987 | |
4345e135 MK |
7988 | /** |
7989 | * get_rwh_policy_from_update() - Get the rwh policy for update option. | |
7990 | * @update: Update option. | |
7991 | */ | |
7992 | static int get_rwh_policy_from_update(enum update_opt update) | |
19ad203e | 7993 | { |
4345e135 MK |
7994 | switch (update) { |
7995 | case UOPT_PPL: | |
19ad203e | 7996 | return RWH_MULTIPLE_DISTRIBUTED; |
4345e135 | 7997 | case UOPT_NO_PPL: |
19ad203e | 7998 | return RWH_MULTIPLE_OFF; |
4345e135 | 7999 | case UOPT_BITMAP: |
19ad203e | 8000 | return RWH_BITMAP; |
4345e135 | 8001 | case UOPT_NO_BITMAP: |
19ad203e | 8002 | return RWH_OFF; |
4345e135 MK |
8003 | default: |
8004 | break; | |
8005 | } | |
8006 | return UOPT_UNDEFINED; | |
19ad203e JR |
8007 | } |
8008 | ||
a951a4f7 | 8009 | static int update_subarray_imsm(struct supertype *st, char *subarray, |
03312b52 | 8010 | enum update_opt update, struct mddev_ident *ident) |
aa534678 DW |
8011 | { |
8012 | /* update the subarray currently referenced by ->current_vol */ | |
8013 | struct intel_super *super = st->sb; | |
8014 | struct imsm_super *mpb = super->anchor; | |
8015 | ||
03312b52 | 8016 | if (update == UOPT_NAME) { |
aa534678 | 8017 | char *name = ident->name; |
a951a4f7 N |
8018 | char *ep; |
8019 | int vol; | |
aa534678 | 8020 | |
aa534678 DW |
8021 | if (!check_name(super, name, 0)) |
8022 | return 2; | |
8023 | ||
a951a4f7 N |
8024 | vol = strtoul(subarray, &ep, 10); |
8025 | if (*ep != '\0' || vol >= super->anchor->num_raid_devs) | |
8026 | return 2; | |
8027 | ||
aa534678 | 8028 | if (st->update_tail) { |
503975b9 | 8029 | struct imsm_update_rename_array *u = xmalloc(sizeof(*u)); |
aa534678 | 8030 | |
aa534678 | 8031 | u->type = update_rename_array; |
a951a4f7 | 8032 | u->dev_idx = vol; |
618f4e6d XN |
8033 | strncpy((char *) u->name, name, MAX_RAID_SERIAL_LEN); |
8034 | u->name[MAX_RAID_SERIAL_LEN-1] = '\0'; | |
aa534678 DW |
8035 | append_metadata_update(st, u, sizeof(*u)); |
8036 | } else { | |
8037 | struct imsm_dev *dev; | |
ebad3af2 | 8038 | int i, namelen; |
aa534678 | 8039 | |
a951a4f7 | 8040 | dev = get_imsm_dev(super, vol); |
ebad3af2 JS |
8041 | memset(dev->volume, '\0', MAX_RAID_SERIAL_LEN); |
8042 | namelen = min((int)strlen(name), MAX_RAID_SERIAL_LEN); | |
8043 | memcpy(dev->volume, name, namelen); | |
aa534678 DW |
8044 | for (i = 0; i < mpb->num_raid_devs; i++) { |
8045 | dev = get_imsm_dev(super, i); | |
8046 | handle_missing(super, dev); | |
8047 | } | |
8048 | super->updates_pending++; | |
8049 | } | |
03312b52 | 8050 | } else if (get_rwh_policy_from_update(update) != UOPT_UNDEFINED) { |
e6e9dd3f AP |
8051 | int new_policy; |
8052 | char *ep; | |
8053 | int vol = strtoul(subarray, &ep, 10); | |
8054 | ||
8055 | if (*ep != '\0' || vol >= super->anchor->num_raid_devs) | |
8056 | return 2; | |
8057 | ||
03312b52 | 8058 | new_policy = get_rwh_policy_from_update(update); |
e6e9dd3f AP |
8059 | |
8060 | if (st->update_tail) { | |
8061 | struct imsm_update_rwh_policy *u = xmalloc(sizeof(*u)); | |
8062 | ||
8063 | u->type = update_rwh_policy; | |
8064 | u->dev_idx = vol; | |
8065 | u->new_policy = new_policy; | |
8066 | append_metadata_update(st, u, sizeof(*u)); | |
8067 | } else { | |
8068 | struct imsm_dev *dev; | |
8069 | ||
8070 | dev = get_imsm_dev(super, vol); | |
8071 | dev->rwh_policy = new_policy; | |
8072 | super->updates_pending++; | |
8073 | } | |
19ad203e JR |
8074 | if (new_policy == RWH_BITMAP) |
8075 | return write_init_bitmap_imsm_vol(st, vol); | |
aa534678 DW |
8076 | } else |
8077 | return 2; | |
8078 | ||
8079 | return 0; | |
8080 | } | |
bf5a934a | 8081 | |
195d1d76 | 8082 | static bool is_gen_migration(struct imsm_dev *dev) |
28bce06f | 8083 | { |
195d1d76 PP |
8084 | if (dev && dev->vol.migr_state && |
8085 | migr_type(dev) == MIGR_GEN_MIGR) | |
8086 | return true; | |
28bce06f | 8087 | |
195d1d76 | 8088 | return false; |
28bce06f AK |
8089 | } |
8090 | ||
1e5c6983 DW |
8091 | static int is_rebuilding(struct imsm_dev *dev) |
8092 | { | |
8093 | struct imsm_map *migr_map; | |
8094 | ||
8095 | if (!dev->vol.migr_state) | |
8096 | return 0; | |
8097 | ||
8098 | if (migr_type(dev) != MIGR_REBUILD) | |
8099 | return 0; | |
8100 | ||
238c0a71 | 8101 | migr_map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
8102 | |
8103 | if (migr_map->map_state == IMSM_T_STATE_DEGRADED) | |
8104 | return 1; | |
8105 | else | |
8106 | return 0; | |
8107 | } | |
8108 | ||
6ce1fbf1 AK |
8109 | static int is_initializing(struct imsm_dev *dev) |
8110 | { | |
8111 | struct imsm_map *migr_map; | |
8112 | ||
8113 | if (!dev->vol.migr_state) | |
8114 | return 0; | |
8115 | ||
8116 | if (migr_type(dev) != MIGR_INIT) | |
8117 | return 0; | |
8118 | ||
238c0a71 | 8119 | migr_map = get_imsm_map(dev, MAP_1); |
6ce1fbf1 AK |
8120 | |
8121 | if (migr_map->map_state == IMSM_T_STATE_UNINITIALIZED) | |
8122 | return 1; | |
8123 | ||
8124 | return 0; | |
6ce1fbf1 AK |
8125 | } |
8126 | ||
c47b0ff6 AK |
8127 | static void update_recovery_start(struct intel_super *super, |
8128 | struct imsm_dev *dev, | |
8129 | struct mdinfo *array) | |
1e5c6983 DW |
8130 | { |
8131 | struct mdinfo *rebuild = NULL; | |
8132 | struct mdinfo *d; | |
8133 | __u32 units; | |
8134 | ||
8135 | if (!is_rebuilding(dev)) | |
8136 | return; | |
8137 | ||
8138 | /* Find the rebuild target, but punt on the dual rebuild case */ | |
8139 | for (d = array->devs; d; d = d->next) | |
8140 | if (d->recovery_start == 0) { | |
8141 | if (rebuild) | |
8142 | return; | |
8143 | rebuild = d; | |
8144 | } | |
8145 | ||
4363fd80 DW |
8146 | if (!rebuild) { |
8147 | /* (?) none of the disks are marked with | |
8148 | * IMSM_ORD_REBUILD, so assume they are missing and the | |
8149 | * disk_ord_tbl was not correctly updated | |
8150 | */ | |
1ade5cc1 | 8151 | dprintf("failed to locate out-of-sync disk\n"); |
4363fd80 DW |
8152 | return; |
8153 | } | |
8154 | ||
4036e7ee | 8155 | units = vol_curr_migr_unit(dev); |
c47b0ff6 | 8156 | rebuild->recovery_start = units * blocks_per_migr_unit(super, dev); |
1e5c6983 DW |
8157 | } |
8158 | ||
276d77db | 8159 | static int recover_backup_imsm(struct supertype *st, struct mdinfo *info); |
1e5c6983 | 8160 | |
00bbdbda | 8161 | static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray) |
cdddbdbc | 8162 | { |
4f5bc454 DW |
8163 | /* Given a container loaded by load_super_imsm_all, |
8164 | * extract information about all the arrays into | |
8165 | * an mdinfo tree. | |
00bbdbda | 8166 | * If 'subarray' is given, just extract info about that array. |
4f5bc454 DW |
8167 | * |
8168 | * For each imsm_dev create an mdinfo, fill it in, | |
8169 | * then look for matching devices in super->disks | |
8170 | * and create appropriate device mdinfo. | |
8171 | */ | |
8172 | struct intel_super *super = st->sb; | |
949c47a0 | 8173 | struct imsm_super *mpb = super->anchor; |
4f5bc454 | 8174 | struct mdinfo *rest = NULL; |
00bbdbda | 8175 | unsigned int i; |
81219e70 | 8176 | int sb_errors = 0; |
abef11a3 AK |
8177 | struct dl *d; |
8178 | int spare_disks = 0; | |
b6180160 | 8179 | int current_vol = super->current_vol; |
cdddbdbc | 8180 | |
19482bcc AK |
8181 | /* do not assemble arrays when not all attributes are supported */ |
8182 | if (imsm_check_attributes(mpb->attributes) == 0) { | |
81219e70 | 8183 | sb_errors = 1; |
7a862a02 | 8184 | pr_err("Unsupported attributes in IMSM metadata.Arrays activation is blocked.\n"); |
19482bcc AK |
8185 | } |
8186 | ||
abef11a3 AK |
8187 | /* count spare devices, not used in maps |
8188 | */ | |
8189 | for (d = super->disks; d; d = d->next) | |
8190 | if (d->index == -1) | |
8191 | spare_disks++; | |
8192 | ||
4f5bc454 | 8193 | for (i = 0; i < mpb->num_raid_devs; i++) { |
00bbdbda N |
8194 | struct imsm_dev *dev; |
8195 | struct imsm_map *map; | |
86e3692b | 8196 | struct imsm_map *map2; |
4f5bc454 | 8197 | struct mdinfo *this; |
a6482415 | 8198 | int slot; |
a6482415 | 8199 | int chunk; |
00bbdbda | 8200 | char *ep; |
8b9cd157 | 8201 | int level; |
00bbdbda N |
8202 | |
8203 | if (subarray && | |
8204 | (i != strtoul(subarray, &ep, 10) || *ep != '\0')) | |
8205 | continue; | |
8206 | ||
8207 | dev = get_imsm_dev(super, i); | |
238c0a71 AK |
8208 | map = get_imsm_map(dev, MAP_0); |
8209 | map2 = get_imsm_map(dev, MAP_1); | |
8b9cd157 | 8210 | level = get_imsm_raid_level(map); |
4f5bc454 | 8211 | |
1ce0101c DW |
8212 | /* do not publish arrays that are in the middle of an |
8213 | * unsupported migration | |
8214 | */ | |
8215 | if (dev->vol.migr_state && | |
28bce06f | 8216 | (migr_type(dev) == MIGR_STATE_CHANGE)) { |
7a862a02 | 8217 | pr_err("cannot assemble volume '%.16s': unsupported migration in progress\n", |
1ce0101c DW |
8218 | dev->volume); |
8219 | continue; | |
8220 | } | |
2db86302 LM |
8221 | /* do not publish arrays that are not support by controller's |
8222 | * OROM/EFI | |
8223 | */ | |
1ce0101c | 8224 | |
503975b9 | 8225 | this = xmalloc(sizeof(*this)); |
4f5bc454 | 8226 | |
301406c9 | 8227 | super->current_vol = i; |
a5d85af7 | 8228 | getinfo_super_imsm_volume(st, this, NULL); |
9894ec0d | 8229 | this->next = rest; |
a6482415 | 8230 | chunk = __le16_to_cpu(map->blocks_per_strip) >> 1; |
81219e70 LM |
8231 | /* mdadm does not support all metadata features- set the bit in all arrays state */ |
8232 | if (!validate_geometry_imsm_orom(super, | |
8b9cd157 MK |
8233 | level, /* RAID level */ |
8234 | imsm_level_to_layout(level), | |
81219e70 | 8235 | map->num_members, /* raid disks */ |
fcc2c9da | 8236 | &chunk, imsm_dev_size(dev), |
81219e70 | 8237 | 1 /* verbose */)) { |
7a862a02 | 8238 | pr_err("IMSM RAID geometry validation failed. Array %s activation is blocked.\n", |
81219e70 LM |
8239 | dev->volume); |
8240 | this->array.state |= | |
8241 | (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) | | |
8242 | (1<<MD_SB_BLOCK_VOLUME); | |
8243 | } | |
81219e70 LM |
8244 | |
8245 | /* if array has bad blocks, set suitable bit in all arrays state */ | |
8246 | if (sb_errors) | |
8247 | this->array.state |= | |
8248 | (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) | | |
8249 | (1<<MD_SB_BLOCK_VOLUME); | |
8250 | ||
4f5bc454 | 8251 | for (slot = 0 ; slot < map->num_members; slot++) { |
1e5c6983 | 8252 | unsigned long long recovery_start; |
4f5bc454 DW |
8253 | struct mdinfo *info_d; |
8254 | struct dl *d; | |
8255 | int idx; | |
9a1608e5 | 8256 | int skip; |
7eef0453 | 8257 | __u32 ord; |
8b9cd157 | 8258 | int missing = 0; |
4f5bc454 | 8259 | |
9a1608e5 | 8260 | skip = 0; |
238c0a71 AK |
8261 | idx = get_imsm_disk_idx(dev, slot, MAP_0); |
8262 | ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); | |
4f5bc454 DW |
8263 | for (d = super->disks; d ; d = d->next) |
8264 | if (d->index == idx) | |
0fbd635c | 8265 | break; |
4f5bc454 | 8266 | |
1e5c6983 | 8267 | recovery_start = MaxSector; |
4f5bc454 | 8268 | if (d == NULL) |
9a1608e5 | 8269 | skip = 1; |
25ed7e59 | 8270 | if (d && is_failed(&d->disk)) |
9a1608e5 | 8271 | skip = 1; |
8b9cd157 | 8272 | if (!skip && (ord & IMSM_ORD_REBUILD)) |
1e5c6983 | 8273 | recovery_start = 0; |
1e93d0d1 BK |
8274 | if (!(ord & IMSM_ORD_REBUILD)) |
8275 | this->array.working_disks++; | |
1011e834 | 8276 | /* |
9a1608e5 | 8277 | * if we skip some disks the array will be assmebled degraded; |
1e5c6983 DW |
8278 | * reset resync start to avoid a dirty-degraded |
8279 | * situation when performing the intial sync | |
9a1608e5 | 8280 | */ |
8b9cd157 MK |
8281 | if (skip) |
8282 | missing++; | |
8283 | ||
8284 | if (!(dev->vol.dirty & RAIDVOL_DIRTY)) { | |
8285 | if ((!able_to_resync(level, missing) || | |
8286 | recovery_start == 0)) | |
8287 | this->resync_start = MaxSector; | |
8b9cd157 MK |
8288 | } |
8289 | ||
9a1608e5 DW |
8290 | if (skip) |
8291 | continue; | |
4f5bc454 | 8292 | |
503975b9 | 8293 | info_d = xcalloc(1, sizeof(*info_d)); |
4f5bc454 DW |
8294 | info_d->next = this->devs; |
8295 | this->devs = info_d; | |
8296 | ||
4f5bc454 DW |
8297 | info_d->disk.number = d->index; |
8298 | info_d->disk.major = d->major; | |
8299 | info_d->disk.minor = d->minor; | |
8300 | info_d->disk.raid_disk = slot; | |
1e5c6983 | 8301 | info_d->recovery_start = recovery_start; |
86e3692b AK |
8302 | if (map2) { |
8303 | if (slot < map2->num_members) | |
8304 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
8305 | else |
8306 | this->array.spare_disks++; | |
86e3692b AK |
8307 | } else { |
8308 | if (slot < map->num_members) | |
8309 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
8310 | else |
8311 | this->array.spare_disks++; | |
86e3692b | 8312 | } |
4f5bc454 DW |
8313 | |
8314 | info_d->events = __le32_to_cpu(mpb->generation_num); | |
5551b113 | 8315 | info_d->data_offset = pba_of_lba0(map); |
44490938 | 8316 | info_d->component_size = calc_component_size(map, dev); |
06fb291a PB |
8317 | |
8318 | if (map->raid_level == 5) { | |
2432ce9b AP |
8319 | info_d->ppl_sector = this->ppl_sector; |
8320 | info_d->ppl_size = this->ppl_size; | |
98e96bdb AP |
8321 | if (this->consistency_policy == CONSISTENCY_POLICY_PPL && |
8322 | recovery_start == 0) | |
8323 | this->resync_start = 0; | |
06fb291a | 8324 | } |
b12796be | 8325 | |
5e46202e | 8326 | info_d->bb.supported = 1; |
b12796be TM |
8327 | get_volume_badblocks(super->bbm_log, ord_to_idx(ord), |
8328 | info_d->data_offset, | |
8329 | info_d->component_size, | |
8330 | &info_d->bb); | |
4f5bc454 | 8331 | } |
1e5c6983 | 8332 | /* now that the disk list is up-to-date fixup recovery_start */ |
c47b0ff6 | 8333 | update_recovery_start(super, dev, this); |
abef11a3 | 8334 | this->array.spare_disks += spare_disks; |
276d77db AK |
8335 | |
8336 | /* check for reshape */ | |
8337 | if (this->reshape_active == 1) | |
8338 | recover_backup_imsm(st, this); | |
9a1608e5 | 8339 | rest = this; |
4f5bc454 DW |
8340 | } |
8341 | ||
b6180160 | 8342 | super->current_vol = current_vol; |
4f5bc454 | 8343 | return rest; |
cdddbdbc DW |
8344 | } |
8345 | ||
3b451610 AK |
8346 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, |
8347 | int failed, int look_in_map) | |
c2a1e7da | 8348 | { |
3b451610 AK |
8349 | struct imsm_map *map; |
8350 | ||
8351 | map = get_imsm_map(dev, look_in_map); | |
c2a1e7da DW |
8352 | |
8353 | if (!failed) | |
1011e834 | 8354 | return map->map_state == IMSM_T_STATE_UNINITIALIZED ? |
3393c6af | 8355 | IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL; |
c2a1e7da DW |
8356 | |
8357 | switch (get_imsm_raid_level(map)) { | |
8358 | case 0: | |
8359 | return IMSM_T_STATE_FAILED; | |
8360 | break; | |
8361 | case 1: | |
8362 | if (failed < map->num_members) | |
8363 | return IMSM_T_STATE_DEGRADED; | |
8364 | else | |
8365 | return IMSM_T_STATE_FAILED; | |
8366 | break; | |
8367 | case 10: | |
8368 | { | |
8369 | /** | |
c92a2527 DW |
8370 | * check to see if any mirrors have failed, otherwise we |
8371 | * are degraded. Even numbered slots are mirrored on | |
8372 | * slot+1 | |
c2a1e7da | 8373 | */ |
c2a1e7da | 8374 | int i; |
d9b420a5 N |
8375 | /* gcc -Os complains that this is unused */ |
8376 | int insync = insync; | |
c2a1e7da DW |
8377 | |
8378 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 8379 | __u32 ord = get_imsm_ord_tbl_ent(dev, i, MAP_X); |
c92a2527 DW |
8380 | int idx = ord_to_idx(ord); |
8381 | struct imsm_disk *disk; | |
c2a1e7da | 8382 | |
c92a2527 | 8383 | /* reset the potential in-sync count on even-numbered |
1011e834 | 8384 | * slots. num_copies is always 2 for imsm raid10 |
c92a2527 DW |
8385 | */ |
8386 | if ((i & 1) == 0) | |
8387 | insync = 2; | |
c2a1e7da | 8388 | |
c92a2527 | 8389 | disk = get_imsm_disk(super, idx); |
25ed7e59 | 8390 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) |
c92a2527 | 8391 | insync--; |
c2a1e7da | 8392 | |
c92a2527 DW |
8393 | /* no in-sync disks left in this mirror the |
8394 | * array has failed | |
8395 | */ | |
8396 | if (insync == 0) | |
8397 | return IMSM_T_STATE_FAILED; | |
c2a1e7da DW |
8398 | } |
8399 | ||
8400 | return IMSM_T_STATE_DEGRADED; | |
8401 | } | |
8402 | case 5: | |
8403 | if (failed < 2) | |
8404 | return IMSM_T_STATE_DEGRADED; | |
8405 | else | |
8406 | return IMSM_T_STATE_FAILED; | |
8407 | break; | |
8408 | default: | |
8409 | break; | |
8410 | } | |
8411 | ||
8412 | return map->map_state; | |
8413 | } | |
8414 | ||
3b451610 AK |
8415 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev, |
8416 | int look_in_map) | |
c2a1e7da DW |
8417 | { |
8418 | int i; | |
8419 | int failed = 0; | |
8420 | struct imsm_disk *disk; | |
d5985138 AK |
8421 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
8422 | struct imsm_map *prev = get_imsm_map(dev, MAP_1); | |
68fe4598 | 8423 | struct imsm_map *map_for_loop; |
0556e1a2 DW |
8424 | __u32 ord; |
8425 | int idx; | |
d5985138 | 8426 | int idx_1; |
c2a1e7da | 8427 | |
0556e1a2 DW |
8428 | /* at the beginning of migration we set IMSM_ORD_REBUILD on |
8429 | * disks that are being rebuilt. New failures are recorded to | |
8430 | * map[0]. So we look through all the disks we started with and | |
8431 | * see if any failures are still present, or if any new ones | |
8432 | * have arrived | |
0556e1a2 | 8433 | */ |
d5985138 AK |
8434 | map_for_loop = map; |
8435 | if (prev && (map->num_members < prev->num_members)) | |
8436 | map_for_loop = prev; | |
68fe4598 LD |
8437 | |
8438 | for (i = 0; i < map_for_loop->num_members; i++) { | |
d5985138 | 8439 | idx_1 = -255; |
238c0a71 AK |
8440 | /* when MAP_X is passed both maps failures are counted |
8441 | */ | |
d5985138 | 8442 | if (prev && |
089f9d79 JS |
8443 | (look_in_map == MAP_1 || look_in_map == MAP_X) && |
8444 | i < prev->num_members) { | |
d5985138 AK |
8445 | ord = __le32_to_cpu(prev->disk_ord_tbl[i]); |
8446 | idx_1 = ord_to_idx(ord); | |
c2a1e7da | 8447 | |
d5985138 AK |
8448 | disk = get_imsm_disk(super, idx_1); |
8449 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) | |
8450 | failed++; | |
8451 | } | |
089f9d79 JS |
8452 | if ((look_in_map == MAP_0 || look_in_map == MAP_X) && |
8453 | i < map->num_members) { | |
d5985138 AK |
8454 | ord = __le32_to_cpu(map->disk_ord_tbl[i]); |
8455 | idx = ord_to_idx(ord); | |
8456 | ||
8457 | if (idx != idx_1) { | |
8458 | disk = get_imsm_disk(super, idx); | |
8459 | if (!disk || is_failed(disk) || | |
8460 | ord & IMSM_ORD_REBUILD) | |
8461 | failed++; | |
8462 | } | |
8463 | } | |
c2a1e7da DW |
8464 | } |
8465 | ||
8466 | return failed; | |
845dea95 NB |
8467 | } |
8468 | ||
97b4d0e9 | 8469 | static int imsm_open_new(struct supertype *c, struct active_array *a, |
60815698 | 8470 | int inst) |
97b4d0e9 DW |
8471 | { |
8472 | struct intel_super *super = c->sb; | |
8473 | struct imsm_super *mpb = super->anchor; | |
bbab0940 | 8474 | struct imsm_update_prealloc_bb_mem u; |
9587c373 | 8475 | |
60815698 MG |
8476 | if (inst >= mpb->num_raid_devs) { |
8477 | pr_err("subarry index %d, out of range\n", inst); | |
97b4d0e9 DW |
8478 | return -ENODEV; |
8479 | } | |
8480 | ||
60815698 MG |
8481 | dprintf("imsm: open_new %d\n", inst); |
8482 | a->info.container_member = inst; | |
bbab0940 TM |
8483 | |
8484 | u.type = update_prealloc_badblocks_mem; | |
8485 | imsm_update_metadata_locally(c, &u, sizeof(u)); | |
8486 | ||
97b4d0e9 DW |
8487 | return 0; |
8488 | } | |
8489 | ||
0c046afd DW |
8490 | static int is_resyncing(struct imsm_dev *dev) |
8491 | { | |
8492 | struct imsm_map *migr_map; | |
8493 | ||
8494 | if (!dev->vol.migr_state) | |
8495 | return 0; | |
8496 | ||
1484e727 DW |
8497 | if (migr_type(dev) == MIGR_INIT || |
8498 | migr_type(dev) == MIGR_REPAIR) | |
0c046afd DW |
8499 | return 1; |
8500 | ||
4c9bc37b AK |
8501 | if (migr_type(dev) == MIGR_GEN_MIGR) |
8502 | return 0; | |
8503 | ||
238c0a71 | 8504 | migr_map = get_imsm_map(dev, MAP_1); |
0c046afd | 8505 | |
089f9d79 JS |
8506 | if (migr_map->map_state == IMSM_T_STATE_NORMAL && |
8507 | dev->vol.migr_type != MIGR_GEN_MIGR) | |
0c046afd DW |
8508 | return 1; |
8509 | else | |
8510 | return 0; | |
8511 | } | |
8512 | ||
0556e1a2 | 8513 | /* return true if we recorded new information */ |
4c9e8c1e TM |
8514 | static int mark_failure(struct intel_super *super, |
8515 | struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
47ee5a45 | 8516 | { |
0556e1a2 DW |
8517 | __u32 ord; |
8518 | int slot; | |
8519 | struct imsm_map *map; | |
86c54047 DW |
8520 | char buf[MAX_RAID_SERIAL_LEN+3]; |
8521 | unsigned int len, shift = 0; | |
0556e1a2 DW |
8522 | |
8523 | /* new failures are always set in map[0] */ | |
238c0a71 | 8524 | map = get_imsm_map(dev, MAP_0); |
0556e1a2 DW |
8525 | |
8526 | slot = get_imsm_disk_slot(map, idx); | |
8527 | if (slot < 0) | |
8528 | return 0; | |
8529 | ||
8530 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
25ed7e59 | 8531 | if (is_failed(disk) && (ord & IMSM_ORD_REBUILD)) |
0556e1a2 DW |
8532 | return 0; |
8533 | ||
7d0c5e24 LD |
8534 | memcpy(buf, disk->serial, MAX_RAID_SERIAL_LEN); |
8535 | buf[MAX_RAID_SERIAL_LEN] = '\000'; | |
8536 | strcat(buf, ":0"); | |
86c54047 DW |
8537 | if ((len = strlen(buf)) >= MAX_RAID_SERIAL_LEN) |
8538 | shift = len - MAX_RAID_SERIAL_LEN + 1; | |
167d8bb8 | 8539 | memcpy(disk->serial, &buf[shift], len + 1 - shift); |
86c54047 | 8540 | |
f2f27e63 | 8541 | disk->status |= FAILED_DISK; |
0556e1a2 | 8542 | set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD); |
17788994 AK |
8543 | /* mark failures in second map if second map exists and this disk |
8544 | * in this slot. | |
8545 | * This is valid for migration, initialization and rebuild | |
8546 | */ | |
8547 | if (dev->vol.migr_state) { | |
238c0a71 | 8548 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); |
0a108d63 AK |
8549 | int slot2 = get_imsm_disk_slot(map2, idx); |
8550 | ||
089f9d79 | 8551 | if (slot2 < map2->num_members && slot2 >= 0) |
0a108d63 | 8552 | set_imsm_ord_tbl_ent(map2, slot2, |
1ace8403 AK |
8553 | idx | IMSM_ORD_REBUILD); |
8554 | } | |
d7a1fda2 MT |
8555 | if (map->failed_disk_num == 0xff || |
8556 | (!is_rebuilding(dev) && map->failed_disk_num > slot)) | |
0556e1a2 | 8557 | map->failed_disk_num = slot; |
4c9e8c1e TM |
8558 | |
8559 | clear_disk_badblocks(super->bbm_log, ord_to_idx(ord)); | |
8560 | ||
0556e1a2 DW |
8561 | return 1; |
8562 | } | |
8563 | ||
4c9e8c1e TM |
8564 | static void mark_missing(struct intel_super *super, |
8565 | struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
0556e1a2 | 8566 | { |
4c9e8c1e | 8567 | mark_failure(super, dev, disk, idx); |
0556e1a2 DW |
8568 | |
8569 | if (disk->scsi_id == __cpu_to_le32(~(__u32)0)) | |
8570 | return; | |
8571 | ||
47ee5a45 DW |
8572 | disk->scsi_id = __cpu_to_le32(~(__u32)0); |
8573 | memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1); | |
8574 | } | |
8575 | ||
33414a01 DW |
8576 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev) |
8577 | { | |
33414a01 | 8578 | struct dl *dl; |
33414a01 DW |
8579 | |
8580 | if (!super->missing) | |
8581 | return; | |
33414a01 | 8582 | |
79b68f1b PC |
8583 | /* When orom adds replacement for missing disk it does |
8584 | * not remove entry of missing disk, but just updates map with | |
8585 | * new added disk. So it is not enough just to test if there is | |
8586 | * any missing disk, we have to look if there are any failed disks | |
8587 | * in map to stop migration */ | |
8588 | ||
33414a01 | 8589 | dprintf("imsm: mark missing\n"); |
3d59f0c0 AK |
8590 | /* end process for initialization and rebuild only |
8591 | */ | |
195d1d76 | 8592 | if (is_gen_migration(dev) == false) { |
fb12a745 | 8593 | int failed = imsm_count_failed(super, dev, MAP_0); |
3d59f0c0 | 8594 | |
fb12a745 TM |
8595 | if (failed) { |
8596 | __u8 map_state; | |
8597 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
8598 | struct imsm_map *map1; | |
8599 | int i, ord, ord_map1; | |
8600 | int rebuilt = 1; | |
3d59f0c0 | 8601 | |
fb12a745 TM |
8602 | for (i = 0; i < map->num_members; i++) { |
8603 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_0); | |
8604 | if (!(ord & IMSM_ORD_REBUILD)) | |
8605 | continue; | |
8606 | ||
8607 | map1 = get_imsm_map(dev, MAP_1); | |
8608 | if (!map1) | |
8609 | continue; | |
8610 | ||
8611 | ord_map1 = __le32_to_cpu(map1->disk_ord_tbl[i]); | |
8612 | if (ord_map1 & IMSM_ORD_REBUILD) | |
8613 | rebuilt = 0; | |
8614 | } | |
8615 | ||
8616 | if (rebuilt) { | |
8617 | map_state = imsm_check_degraded(super, dev, | |
8618 | failed, MAP_0); | |
8619 | end_migration(dev, super, map_state); | |
8620 | } | |
8621 | } | |
3d59f0c0 | 8622 | } |
33414a01 | 8623 | for (dl = super->missing; dl; dl = dl->next) |
4c9e8c1e | 8624 | mark_missing(super, dev, &dl->disk, dl->index); |
33414a01 DW |
8625 | super->updates_pending++; |
8626 | } | |
8627 | ||
f3871fdc AK |
8628 | static unsigned long long imsm_set_array_size(struct imsm_dev *dev, |
8629 | long long new_size) | |
70bdf0dc | 8630 | { |
70bdf0dc | 8631 | unsigned long long array_blocks; |
9529d343 MD |
8632 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
8633 | int used_disks = imsm_num_data_members(map); | |
70bdf0dc AK |
8634 | |
8635 | if (used_disks == 0) { | |
8636 | /* when problems occures | |
8637 | * return current array_blocks value | |
8638 | */ | |
fcc2c9da | 8639 | array_blocks = imsm_dev_size(dev); |
70bdf0dc AK |
8640 | |
8641 | return array_blocks; | |
8642 | } | |
8643 | ||
8644 | /* set array size in metadata | |
8645 | */ | |
9529d343 | 8646 | if (new_size <= 0) |
f3871fdc AK |
8647 | /* OLCE size change is caused by added disks |
8648 | */ | |
44490938 | 8649 | array_blocks = per_dev_array_size(map) * used_disks; |
9529d343 | 8650 | else |
f3871fdc AK |
8651 | /* Online Volume Size Change |
8652 | * Using available free space | |
8653 | */ | |
8654 | array_blocks = new_size; | |
70bdf0dc | 8655 | |
b53bfba6 | 8656 | array_blocks = round_size_to_mb(array_blocks, used_disks); |
fcc2c9da | 8657 | set_imsm_dev_size(dev, array_blocks); |
70bdf0dc AK |
8658 | |
8659 | return array_blocks; | |
8660 | } | |
8661 | ||
28bce06f AK |
8662 | static void imsm_set_disk(struct active_array *a, int n, int state); |
8663 | ||
0e2d1a4e AK |
8664 | static void imsm_progress_container_reshape(struct intel_super *super) |
8665 | { | |
8666 | /* if no device has a migr_state, but some device has a | |
8667 | * different number of members than the previous device, start | |
8668 | * changing the number of devices in this device to match | |
8669 | * previous. | |
8670 | */ | |
8671 | struct imsm_super *mpb = super->anchor; | |
8672 | int prev_disks = -1; | |
8673 | int i; | |
1dfaa380 | 8674 | int copy_map_size; |
0e2d1a4e AK |
8675 | |
8676 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
8677 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 8678 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
0e2d1a4e AK |
8679 | struct imsm_map *map2; |
8680 | int prev_num_members; | |
0e2d1a4e AK |
8681 | |
8682 | if (dev->vol.migr_state) | |
8683 | return; | |
8684 | ||
8685 | if (prev_disks == -1) | |
8686 | prev_disks = map->num_members; | |
8687 | if (prev_disks == map->num_members) | |
8688 | continue; | |
8689 | ||
8690 | /* OK, this array needs to enter reshape mode. | |
8691 | * i.e it needs a migr_state | |
8692 | */ | |
8693 | ||
1dfaa380 | 8694 | copy_map_size = sizeof_imsm_map(map); |
0e2d1a4e AK |
8695 | prev_num_members = map->num_members; |
8696 | map->num_members = prev_disks; | |
8697 | dev->vol.migr_state = 1; | |
4036e7ee | 8698 | set_vol_curr_migr_unit(dev, 0); |
ea672ee1 | 8699 | set_migr_type(dev, MIGR_GEN_MIGR); |
0e2d1a4e AK |
8700 | for (i = prev_num_members; |
8701 | i < map->num_members; i++) | |
8702 | set_imsm_ord_tbl_ent(map, i, i); | |
238c0a71 | 8703 | map2 = get_imsm_map(dev, MAP_1); |
0e2d1a4e | 8704 | /* Copy the current map */ |
1dfaa380 | 8705 | memcpy(map2, map, copy_map_size); |
0e2d1a4e AK |
8706 | map2->num_members = prev_num_members; |
8707 | ||
f3871fdc | 8708 | imsm_set_array_size(dev, -1); |
51d83f5d | 8709 | super->clean_migration_record_by_mdmon = 1; |
0e2d1a4e AK |
8710 | super->updates_pending++; |
8711 | } | |
8712 | } | |
8713 | ||
aad6f216 | 8714 | /* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild |
0c046afd DW |
8715 | * states are handled in imsm_set_disk() with one exception, when a |
8716 | * resync is stopped due to a new failure this routine will set the | |
8717 | * 'degraded' state for the array. | |
8718 | */ | |
01f157d7 | 8719 | static int imsm_set_array_state(struct active_array *a, int consistent) |
a862209d DW |
8720 | { |
8721 | int inst = a->info.container_member; | |
8722 | struct intel_super *super = a->container->sb; | |
949c47a0 | 8723 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 8724 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
3b451610 AK |
8725 | int failed = imsm_count_failed(super, dev, MAP_0); |
8726 | __u8 map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
1e5c6983 | 8727 | __u32 blocks_per_unit; |
a862209d | 8728 | |
1af97990 AK |
8729 | if (dev->vol.migr_state && |
8730 | dev->vol.migr_type == MIGR_GEN_MIGR) { | |
8731 | /* array state change is blocked due to reshape action | |
aad6f216 N |
8732 | * We might need to |
8733 | * - abort the reshape (if last_checkpoint is 0 and action!= reshape) | |
8734 | * - finish the reshape (if last_checkpoint is big and action != reshape) | |
4036e7ee | 8735 | * - update vol_curr_migr_unit |
1af97990 | 8736 | */ |
aad6f216 | 8737 | if (a->curr_action == reshape) { |
4036e7ee | 8738 | /* still reshaping, maybe update vol_curr_migr_unit */ |
633b5610 | 8739 | goto mark_checkpoint; |
aad6f216 N |
8740 | } else { |
8741 | if (a->last_checkpoint == 0 && a->prev_action == reshape) { | |
8742 | /* for some reason we aborted the reshape. | |
b66e591b AK |
8743 | * |
8744 | * disable automatic metadata rollback | |
8745 | * user action is required to recover process | |
aad6f216 | 8746 | */ |
b66e591b | 8747 | if (0) { |
238c0a71 AK |
8748 | struct imsm_map *map2 = |
8749 | get_imsm_map(dev, MAP_1); | |
8750 | dev->vol.migr_state = 0; | |
8751 | set_migr_type(dev, 0); | |
4036e7ee | 8752 | set_vol_curr_migr_unit(dev, 0); |
238c0a71 AK |
8753 | memcpy(map, map2, |
8754 | sizeof_imsm_map(map2)); | |
8755 | super->updates_pending++; | |
b66e591b | 8756 | } |
aad6f216 N |
8757 | } |
8758 | if (a->last_checkpoint >= a->info.component_size) { | |
8759 | unsigned long long array_blocks; | |
8760 | int used_disks; | |
e154ced3 | 8761 | struct mdinfo *mdi; |
aad6f216 | 8762 | |
9529d343 | 8763 | used_disks = imsm_num_data_members(map); |
d55adef9 AK |
8764 | if (used_disks > 0) { |
8765 | array_blocks = | |
44490938 | 8766 | per_dev_array_size(map) * |
d55adef9 | 8767 | used_disks; |
b53bfba6 TM |
8768 | array_blocks = |
8769 | round_size_to_mb(array_blocks, | |
8770 | used_disks); | |
d55adef9 AK |
8771 | a->info.custom_array_size = array_blocks; |
8772 | /* encourage manager to update array | |
8773 | * size | |
8774 | */ | |
e154ced3 | 8775 | |
d55adef9 | 8776 | a->check_reshape = 1; |
633b5610 | 8777 | } |
e154ced3 AK |
8778 | /* finalize online capacity expansion/reshape */ |
8779 | for (mdi = a->info.devs; mdi; mdi = mdi->next) | |
8780 | imsm_set_disk(a, | |
8781 | mdi->disk.raid_disk, | |
8782 | mdi->curr_state); | |
8783 | ||
0e2d1a4e | 8784 | imsm_progress_container_reshape(super); |
e154ced3 | 8785 | } |
aad6f216 | 8786 | } |
1af97990 AK |
8787 | } |
8788 | ||
47ee5a45 | 8789 | /* before we activate this array handle any missing disks */ |
33414a01 DW |
8790 | if (consistent == 2) |
8791 | handle_missing(super, dev); | |
1e5c6983 | 8792 | |
0c046afd | 8793 | if (consistent == 2 && |
b7941fd6 | 8794 | (!is_resync_complete(&a->info) || |
0c046afd DW |
8795 | map_state != IMSM_T_STATE_NORMAL || |
8796 | dev->vol.migr_state)) | |
01f157d7 | 8797 | consistent = 0; |
272906ef | 8798 | |
b7941fd6 | 8799 | if (is_resync_complete(&a->info)) { |
0c046afd | 8800 | /* complete intialization / resync, |
0556e1a2 DW |
8801 | * recovery and interrupted recovery is completed in |
8802 | * ->set_disk | |
0c046afd DW |
8803 | */ |
8804 | if (is_resyncing(dev)) { | |
8805 | dprintf("imsm: mark resync done\n"); | |
809da78e | 8806 | end_migration(dev, super, map_state); |
115c3803 | 8807 | super->updates_pending++; |
484240d8 | 8808 | a->last_checkpoint = 0; |
115c3803 | 8809 | } |
b9172665 AK |
8810 | } else if ((!is_resyncing(dev) && !failed) && |
8811 | (imsm_reshape_blocks_arrays_changes(super) == 0)) { | |
0c046afd | 8812 | /* mark the start of the init process if nothing is failed */ |
b7941fd6 | 8813 | dprintf("imsm: mark resync start\n"); |
1484e727 | 8814 | if (map->map_state == IMSM_T_STATE_UNINITIALIZED) |
8e59f3d8 | 8815 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_INIT); |
1484e727 | 8816 | else |
8e59f3d8 | 8817 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_REPAIR); |
3393c6af | 8818 | super->updates_pending++; |
115c3803 | 8819 | } |
a862209d | 8820 | |
633b5610 | 8821 | mark_checkpoint: |
5b83bacf AK |
8822 | /* skip checkpointing for general migration, |
8823 | * it is controlled in mdadm | |
8824 | */ | |
8825 | if (is_gen_migration(dev)) | |
8826 | goto skip_mark_checkpoint; | |
8827 | ||
4036e7ee MT |
8828 | /* check if we can update vol_curr_migr_unit from resync_start, |
8829 | * recovery_start | |
8830 | */ | |
c47b0ff6 | 8831 | blocks_per_unit = blocks_per_migr_unit(super, dev); |
4f0a7acc | 8832 | if (blocks_per_unit) { |
4036e7ee MT |
8833 | set_vol_curr_migr_unit(dev, |
8834 | a->last_checkpoint / blocks_per_unit); | |
8835 | dprintf("imsm: mark checkpoint (%llu)\n", | |
8836 | vol_curr_migr_unit(dev)); | |
8837 | super->updates_pending++; | |
1e5c6983 | 8838 | } |
f8f603f1 | 8839 | |
5b83bacf | 8840 | skip_mark_checkpoint: |
3393c6af | 8841 | /* mark dirty / clean */ |
2432ce9b AP |
8842 | if (((dev->vol.dirty & RAIDVOL_DIRTY) && consistent) || |
8843 | (!(dev->vol.dirty & RAIDVOL_DIRTY) && !consistent)) { | |
b7941fd6 | 8844 | dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty"); |
2432ce9b AP |
8845 | if (consistent) { |
8846 | dev->vol.dirty = RAIDVOL_CLEAN; | |
8847 | } else { | |
8848 | dev->vol.dirty = RAIDVOL_DIRTY; | |
c2462068 PB |
8849 | if (dev->rwh_policy == RWH_DISTRIBUTED || |
8850 | dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED) | |
2432ce9b AP |
8851 | dev->vol.dirty |= RAIDVOL_DSRECORD_VALID; |
8852 | } | |
a862209d DW |
8853 | super->updates_pending++; |
8854 | } | |
28bce06f | 8855 | |
01f157d7 | 8856 | return consistent; |
a862209d DW |
8857 | } |
8858 | ||
6f50473f TM |
8859 | static int imsm_disk_slot_to_ord(struct active_array *a, int slot) |
8860 | { | |
8861 | int inst = a->info.container_member; | |
8862 | struct intel_super *super = a->container->sb; | |
8863 | struct imsm_dev *dev = get_imsm_dev(super, inst); | |
8864 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
8865 | ||
8866 | if (slot > map->num_members) { | |
8867 | pr_err("imsm: imsm_disk_slot_to_ord %d out of range 0..%d\n", | |
8868 | slot, map->num_members - 1); | |
8869 | return -1; | |
8870 | } | |
8871 | ||
8872 | if (slot < 0) | |
8873 | return -1; | |
8874 | ||
8875 | return get_imsm_ord_tbl_ent(dev, slot, MAP_0); | |
8876 | } | |
8877 | ||
8d45d196 | 8878 | static void imsm_set_disk(struct active_array *a, int n, int state) |
845dea95 | 8879 | { |
8d45d196 DW |
8880 | int inst = a->info.container_member; |
8881 | struct intel_super *super = a->container->sb; | |
949c47a0 | 8882 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 8883 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
8d45d196 | 8884 | struct imsm_disk *disk; |
7ce05701 LD |
8885 | struct mdinfo *mdi; |
8886 | int recovery_not_finished = 0; | |
0c046afd | 8887 | int failed; |
6f50473f | 8888 | int ord; |
0c046afd | 8889 | __u8 map_state; |
fb12a745 TM |
8890 | int rebuild_done = 0; |
8891 | int i; | |
8d45d196 | 8892 | |
fb12a745 | 8893 | ord = get_imsm_ord_tbl_ent(dev, n, MAP_X); |
6f50473f | 8894 | if (ord < 0) |
8d45d196 DW |
8895 | return; |
8896 | ||
4e6e574a | 8897 | dprintf("imsm: set_disk %d:%x\n", n, state); |
b10b37b8 | 8898 | disk = get_imsm_disk(super, ord_to_idx(ord)); |
8d45d196 | 8899 | |
5802a811 | 8900 | /* check for new failures */ |
ae7d61e3 | 8901 | if (disk && (state & DS_FAULTY)) { |
4c9e8c1e | 8902 | if (mark_failure(super, dev, disk, ord_to_idx(ord))) |
0556e1a2 | 8903 | super->updates_pending++; |
8d45d196 | 8904 | } |
47ee5a45 | 8905 | |
19859edc | 8906 | /* check if in_sync */ |
0556e1a2 | 8907 | if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) { |
238c0a71 | 8908 | struct imsm_map *migr_map = get_imsm_map(dev, MAP_1); |
b10b37b8 DW |
8909 | |
8910 | set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord)); | |
fb12a745 | 8911 | rebuild_done = 1; |
19859edc DW |
8912 | super->updates_pending++; |
8913 | } | |
8d45d196 | 8914 | |
3b451610 AK |
8915 | failed = imsm_count_failed(super, dev, MAP_0); |
8916 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
5802a811 | 8917 | |
0c046afd | 8918 | /* check if recovery complete, newly degraded, or failed */ |
94002678 AK |
8919 | dprintf("imsm: Detected transition to state "); |
8920 | switch (map_state) { | |
8921 | case IMSM_T_STATE_NORMAL: /* transition to normal state */ | |
8922 | dprintf("normal: "); | |
8923 | if (is_rebuilding(dev)) { | |
1ade5cc1 | 8924 | dprintf_cont("while rebuilding"); |
7ce05701 LD |
8925 | /* check if recovery is really finished */ |
8926 | for (mdi = a->info.devs; mdi ; mdi = mdi->next) | |
8927 | if (mdi->recovery_start != MaxSector) { | |
8928 | recovery_not_finished = 1; | |
8929 | break; | |
8930 | } | |
8931 | if (recovery_not_finished) { | |
1ade5cc1 N |
8932 | dprintf_cont("\n"); |
8933 | dprintf("Rebuild has not finished yet, state not changed"); | |
7ce05701 LD |
8934 | if (a->last_checkpoint < mdi->recovery_start) { |
8935 | a->last_checkpoint = mdi->recovery_start; | |
8936 | super->updates_pending++; | |
8937 | } | |
8938 | break; | |
8939 | } | |
94002678 | 8940 | end_migration(dev, super, map_state); |
94002678 AK |
8941 | map->failed_disk_num = ~0; |
8942 | super->updates_pending++; | |
8943 | a->last_checkpoint = 0; | |
8944 | break; | |
8945 | } | |
8946 | if (is_gen_migration(dev)) { | |
1ade5cc1 | 8947 | dprintf_cont("while general migration"); |
bf2f0071 | 8948 | if (a->last_checkpoint >= a->info.component_size) |
809da78e | 8949 | end_migration(dev, super, map_state); |
94002678 AK |
8950 | else |
8951 | map->map_state = map_state; | |
28bce06f | 8952 | map->failed_disk_num = ~0; |
94002678 | 8953 | super->updates_pending++; |
bf2f0071 | 8954 | break; |
94002678 AK |
8955 | } |
8956 | break; | |
8957 | case IMSM_T_STATE_DEGRADED: /* transition to degraded state */ | |
1ade5cc1 | 8958 | dprintf_cont("degraded: "); |
089f9d79 | 8959 | if (map->map_state != map_state && !dev->vol.migr_state) { |
1ade5cc1 | 8960 | dprintf_cont("mark degraded"); |
94002678 AK |
8961 | map->map_state = map_state; |
8962 | super->updates_pending++; | |
8963 | a->last_checkpoint = 0; | |
8964 | break; | |
8965 | } | |
8966 | if (is_rebuilding(dev)) { | |
d7a1fda2 | 8967 | dprintf_cont("while rebuilding "); |
a4e96fd8 MT |
8968 | if (state & DS_FAULTY) { |
8969 | dprintf_cont("removing failed drive "); | |
d7a1fda2 MT |
8970 | if (n == map->failed_disk_num) { |
8971 | dprintf_cont("end migration"); | |
8972 | end_migration(dev, super, map_state); | |
a4e96fd8 | 8973 | a->last_checkpoint = 0; |
d7a1fda2 | 8974 | } else { |
a4e96fd8 | 8975 | dprintf_cont("fail detected during rebuild, changing map state"); |
d7a1fda2 MT |
8976 | map->map_state = map_state; |
8977 | } | |
94002678 | 8978 | super->updates_pending++; |
fb12a745 TM |
8979 | } |
8980 | ||
a4e96fd8 MT |
8981 | if (!rebuild_done) |
8982 | break; | |
8983 | ||
fb12a745 TM |
8984 | /* check if recovery is really finished */ |
8985 | for (mdi = a->info.devs; mdi ; mdi = mdi->next) | |
8986 | if (mdi->recovery_start != MaxSector) { | |
8987 | recovery_not_finished = 1; | |
8988 | break; | |
8989 | } | |
8990 | if (recovery_not_finished) { | |
8991 | dprintf_cont("\n"); | |
a4e96fd8 | 8992 | dprintf_cont("Rebuild has not finished yet"); |
fb12a745 TM |
8993 | if (a->last_checkpoint < mdi->recovery_start) { |
8994 | a->last_checkpoint = | |
8995 | mdi->recovery_start; | |
8996 | super->updates_pending++; | |
8997 | } | |
8998 | break; | |
94002678 | 8999 | } |
fb12a745 TM |
9000 | |
9001 | dprintf_cont(" Rebuild done, still degraded"); | |
a4e96fd8 MT |
9002 | end_migration(dev, super, map_state); |
9003 | a->last_checkpoint = 0; | |
9004 | super->updates_pending++; | |
fb12a745 TM |
9005 | |
9006 | for (i = 0; i < map->num_members; i++) { | |
9007 | int idx = get_imsm_ord_tbl_ent(dev, i, MAP_0); | |
9008 | ||
9009 | if (idx & IMSM_ORD_REBUILD) | |
9010 | map->failed_disk_num = i; | |
9011 | } | |
9012 | super->updates_pending++; | |
94002678 AK |
9013 | break; |
9014 | } | |
9015 | if (is_gen_migration(dev)) { | |
1ade5cc1 | 9016 | dprintf_cont("while general migration"); |
bf2f0071 | 9017 | if (a->last_checkpoint >= a->info.component_size) |
809da78e | 9018 | end_migration(dev, super, map_state); |
94002678 AK |
9019 | else { |
9020 | map->map_state = map_state; | |
3b451610 | 9021 | manage_second_map(super, dev); |
94002678 AK |
9022 | } |
9023 | super->updates_pending++; | |
bf2f0071 | 9024 | break; |
28bce06f | 9025 | } |
6ce1fbf1 | 9026 | if (is_initializing(dev)) { |
1ade5cc1 | 9027 | dprintf_cont("while initialization."); |
6ce1fbf1 AK |
9028 | map->map_state = map_state; |
9029 | super->updates_pending++; | |
9030 | break; | |
9031 | } | |
94002678 AK |
9032 | break; |
9033 | case IMSM_T_STATE_FAILED: /* transition to failed state */ | |
1ade5cc1 | 9034 | dprintf_cont("failed: "); |
94002678 | 9035 | if (is_gen_migration(dev)) { |
1ade5cc1 | 9036 | dprintf_cont("while general migration"); |
94002678 AK |
9037 | map->map_state = map_state; |
9038 | super->updates_pending++; | |
9039 | break; | |
9040 | } | |
9041 | if (map->map_state != map_state) { | |
1ade5cc1 | 9042 | dprintf_cont("mark failed"); |
94002678 AK |
9043 | end_migration(dev, super, map_state); |
9044 | super->updates_pending++; | |
9045 | a->last_checkpoint = 0; | |
9046 | break; | |
9047 | } | |
9048 | break; | |
9049 | default: | |
1ade5cc1 | 9050 | dprintf_cont("state %i\n", map_state); |
5802a811 | 9051 | } |
1ade5cc1 | 9052 | dprintf_cont("\n"); |
845dea95 NB |
9053 | } |
9054 | ||
f796af5d | 9055 | static int store_imsm_mpb(int fd, struct imsm_super *mpb) |
c2a1e7da | 9056 | { |
f796af5d | 9057 | void *buf = mpb; |
c2a1e7da DW |
9058 | __u32 mpb_size = __le32_to_cpu(mpb->mpb_size); |
9059 | unsigned long long dsize; | |
9060 | unsigned long long sectors; | |
f36a9ecd | 9061 | unsigned int sector_size; |
c2a1e7da | 9062 | |
aec01630 JS |
9063 | if (!get_dev_sector_size(fd, NULL, §or_size)) |
9064 | return 1; | |
c2a1e7da DW |
9065 | get_dev_size(fd, NULL, &dsize); |
9066 | ||
f36a9ecd | 9067 | if (mpb_size > sector_size) { |
272f648f | 9068 | /* -1 to account for anchor */ |
f36a9ecd | 9069 | sectors = mpb_sectors(mpb, sector_size) - 1; |
c2a1e7da | 9070 | |
272f648f | 9071 | /* write the extended mpb to the sectors preceeding the anchor */ |
f36a9ecd PB |
9072 | if (lseek64(fd, dsize - (sector_size * (2 + sectors)), |
9073 | SEEK_SET) < 0) | |
272f648f | 9074 | return 1; |
c2a1e7da | 9075 | |
f36a9ecd PB |
9076 | if ((unsigned long long)write(fd, buf + sector_size, |
9077 | sector_size * sectors) != sector_size * sectors) | |
272f648f DW |
9078 | return 1; |
9079 | } | |
c2a1e7da | 9080 | |
272f648f | 9081 | /* first block is stored on second to last sector of the disk */ |
f36a9ecd | 9082 | if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0) |
c2a1e7da DW |
9083 | return 1; |
9084 | ||
466070ad | 9085 | if ((unsigned int)write(fd, buf, sector_size) != sector_size) |
c2a1e7da DW |
9086 | return 1; |
9087 | ||
c2a1e7da DW |
9088 | return 0; |
9089 | } | |
9090 | ||
2e735d19 | 9091 | static void imsm_sync_metadata(struct supertype *container) |
845dea95 | 9092 | { |
2e735d19 | 9093 | struct intel_super *super = container->sb; |
c2a1e7da | 9094 | |
1a64be56 | 9095 | dprintf("sync metadata: %d\n", super->updates_pending); |
c2a1e7da DW |
9096 | if (!super->updates_pending) |
9097 | return; | |
9098 | ||
36988a3d | 9099 | write_super_imsm(container, 0); |
c2a1e7da DW |
9100 | |
9101 | super->updates_pending = 0; | |
845dea95 NB |
9102 | } |
9103 | ||
272906ef DW |
9104 | static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a) |
9105 | { | |
9106 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 9107 | int i = get_imsm_disk_idx(dev, idx, MAP_X); |
272906ef DW |
9108 | struct dl *dl; |
9109 | ||
9110 | for (dl = super->disks; dl; dl = dl->next) | |
9111 | if (dl->index == i) | |
9112 | break; | |
9113 | ||
25ed7e59 | 9114 | if (dl && is_failed(&dl->disk)) |
272906ef DW |
9115 | dl = NULL; |
9116 | ||
9117 | if (dl) | |
1ade5cc1 | 9118 | dprintf("found %x:%x\n", dl->major, dl->minor); |
272906ef DW |
9119 | |
9120 | return dl; | |
9121 | } | |
9122 | ||
a20d2ba5 | 9123 | static struct dl *imsm_add_spare(struct intel_super *super, int slot, |
8ba77d32 AK |
9124 | struct active_array *a, int activate_new, |
9125 | struct mdinfo *additional_test_list) | |
272906ef DW |
9126 | { |
9127 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 9128 | int idx = get_imsm_disk_idx(dev, slot, MAP_X); |
a20d2ba5 DW |
9129 | struct imsm_super *mpb = super->anchor; |
9130 | struct imsm_map *map; | |
272906ef DW |
9131 | unsigned long long pos; |
9132 | struct mdinfo *d; | |
9133 | struct extent *ex; | |
a20d2ba5 | 9134 | int i, j; |
272906ef | 9135 | int found; |
569cc43f DW |
9136 | __u32 array_start = 0; |
9137 | __u32 array_end = 0; | |
272906ef | 9138 | struct dl *dl; |
6c932028 | 9139 | struct mdinfo *test_list; |
272906ef DW |
9140 | |
9141 | for (dl = super->disks; dl; dl = dl->next) { | |
9142 | /* If in this array, skip */ | |
9143 | for (d = a->info.devs ; d ; d = d->next) | |
4389ce73 | 9144 | if (is_fd_valid(d->state_fd) && |
e553d2a4 | 9145 | d->disk.major == dl->major && |
272906ef | 9146 | d->disk.minor == dl->minor) { |
8ba77d32 AK |
9147 | dprintf("%x:%x already in array\n", |
9148 | dl->major, dl->minor); | |
272906ef DW |
9149 | break; |
9150 | } | |
9151 | if (d) | |
9152 | continue; | |
6c932028 AK |
9153 | test_list = additional_test_list; |
9154 | while (test_list) { | |
9155 | if (test_list->disk.major == dl->major && | |
9156 | test_list->disk.minor == dl->minor) { | |
8ba77d32 AK |
9157 | dprintf("%x:%x already in additional test list\n", |
9158 | dl->major, dl->minor); | |
9159 | break; | |
9160 | } | |
6c932028 | 9161 | test_list = test_list->next; |
8ba77d32 | 9162 | } |
6c932028 | 9163 | if (test_list) |
8ba77d32 | 9164 | continue; |
272906ef | 9165 | |
e553d2a4 | 9166 | /* skip in use or failed drives */ |
25ed7e59 | 9167 | if (is_failed(&dl->disk) || idx == dl->index || |
df474657 DW |
9168 | dl->index == -2) { |
9169 | dprintf("%x:%x status (failed: %d index: %d)\n", | |
25ed7e59 | 9170 | dl->major, dl->minor, is_failed(&dl->disk), idx); |
9a1608e5 DW |
9171 | continue; |
9172 | } | |
9173 | ||
a20d2ba5 DW |
9174 | /* skip pure spares when we are looking for partially |
9175 | * assimilated drives | |
9176 | */ | |
9177 | if (dl->index == -1 && !activate_new) | |
9178 | continue; | |
9179 | ||
f2cc4f7d AO |
9180 | if (!drive_validate_sector_size(super, dl)) |
9181 | continue; | |
9182 | ||
272906ef | 9183 | /* Does this unused device have the requisite free space? |
a20d2ba5 | 9184 | * It needs to be able to cover all member volumes |
272906ef | 9185 | */ |
05501181 | 9186 | ex = get_extents(super, dl, 1); |
272906ef DW |
9187 | if (!ex) { |
9188 | dprintf("cannot get extents\n"); | |
9189 | continue; | |
9190 | } | |
a20d2ba5 DW |
9191 | for (i = 0; i < mpb->num_raid_devs; i++) { |
9192 | dev = get_imsm_dev(super, i); | |
238c0a71 | 9193 | map = get_imsm_map(dev, MAP_0); |
272906ef | 9194 | |
a20d2ba5 DW |
9195 | /* check if this disk is already a member of |
9196 | * this array | |
272906ef | 9197 | */ |
620b1713 | 9198 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
a20d2ba5 DW |
9199 | continue; |
9200 | ||
9201 | found = 0; | |
9202 | j = 0; | |
9203 | pos = 0; | |
5551b113 | 9204 | array_start = pba_of_lba0(map); |
329c8278 | 9205 | array_end = array_start + |
44490938 | 9206 | per_dev_array_size(map) - 1; |
a20d2ba5 DW |
9207 | |
9208 | do { | |
9209 | /* check that we can start at pba_of_lba0 with | |
44490938 | 9210 | * num_data_stripes*blocks_per_stripe of space |
a20d2ba5 | 9211 | */ |
329c8278 | 9212 | if (array_start >= pos && array_end < ex[j].start) { |
a20d2ba5 DW |
9213 | found = 1; |
9214 | break; | |
9215 | } | |
9216 | pos = ex[j].start + ex[j].size; | |
9217 | j++; | |
9218 | } while (ex[j-1].size); | |
9219 | ||
9220 | if (!found) | |
272906ef | 9221 | break; |
a20d2ba5 | 9222 | } |
272906ef DW |
9223 | |
9224 | free(ex); | |
a20d2ba5 | 9225 | if (i < mpb->num_raid_devs) { |
329c8278 DW |
9226 | dprintf("%x:%x does not have %u to %u available\n", |
9227 | dl->major, dl->minor, array_start, array_end); | |
272906ef DW |
9228 | /* No room */ |
9229 | continue; | |
a20d2ba5 DW |
9230 | } |
9231 | return dl; | |
272906ef DW |
9232 | } |
9233 | ||
9234 | return dl; | |
9235 | } | |
9236 | ||
95d07a2c LM |
9237 | static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed) |
9238 | { | |
9239 | struct imsm_dev *dev2; | |
9240 | struct imsm_map *map; | |
9241 | struct dl *idisk; | |
9242 | int slot; | |
9243 | int idx; | |
9244 | __u8 state; | |
9245 | ||
9246 | dev2 = get_imsm_dev(cont->sb, dev_idx); | |
756a15f3 MG |
9247 | |
9248 | state = imsm_check_degraded(cont->sb, dev2, failed, MAP_0); | |
9249 | if (state == IMSM_T_STATE_FAILED) { | |
9250 | map = get_imsm_map(dev2, MAP_0); | |
9251 | for (slot = 0; slot < map->num_members; slot++) { | |
9252 | /* | |
9253 | * Check if failed disks are deleted from intel | |
9254 | * disk list or are marked to be deleted | |
9255 | */ | |
9256 | idx = get_imsm_disk_idx(dev2, slot, MAP_X); | |
9257 | idisk = get_imsm_dl_disk(cont->sb, idx); | |
9258 | /* | |
9259 | * Do not rebuild the array if failed disks | |
9260 | * from failed sub-array are not removed from | |
9261 | * container. | |
9262 | */ | |
9263 | if (idisk && | |
9264 | is_failed(&idisk->disk) && | |
9265 | (idisk->action != DISK_REMOVE)) | |
9266 | return 0; | |
95d07a2c LM |
9267 | } |
9268 | } | |
9269 | return 1; | |
9270 | } | |
9271 | ||
88758e9d DW |
9272 | static struct mdinfo *imsm_activate_spare(struct active_array *a, |
9273 | struct metadata_update **updates) | |
9274 | { | |
9275 | /** | |
d23fe947 DW |
9276 | * Find a device with unused free space and use it to replace a |
9277 | * failed/vacant region in an array. We replace failed regions one a | |
9278 | * array at a time. The result is that a new spare disk will be added | |
9279 | * to the first failed array and after the monitor has finished | |
9280 | * propagating failures the remainder will be consumed. | |
88758e9d | 9281 | * |
d23fe947 DW |
9282 | * FIXME add a capability for mdmon to request spares from another |
9283 | * container. | |
88758e9d DW |
9284 | */ |
9285 | ||
9286 | struct intel_super *super = a->container->sb; | |
88758e9d | 9287 | int inst = a->info.container_member; |
949c47a0 | 9288 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 9289 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
88758e9d DW |
9290 | int failed = a->info.array.raid_disks; |
9291 | struct mdinfo *rv = NULL; | |
9292 | struct mdinfo *d; | |
9293 | struct mdinfo *di; | |
9294 | struct metadata_update *mu; | |
9295 | struct dl *dl; | |
9296 | struct imsm_update_activate_spare *u; | |
9297 | int num_spares = 0; | |
9298 | int i; | |
95d07a2c | 9299 | int allowed; |
88758e9d | 9300 | |
4389ce73 MT |
9301 | for (d = a->info.devs ; d; d = d->next) { |
9302 | if (!is_fd_valid(d->state_fd)) | |
9303 | continue; | |
9304 | ||
9305 | if (d->curr_state & DS_FAULTY) | |
88758e9d DW |
9306 | /* wait for Removal to happen */ |
9307 | return NULL; | |
4389ce73 MT |
9308 | |
9309 | failed--; | |
88758e9d DW |
9310 | } |
9311 | ||
9312 | dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n", | |
9313 | inst, failed, a->info.array.raid_disks, a->info.array.level); | |
1af97990 | 9314 | |
e2962bfc AK |
9315 | if (imsm_reshape_blocks_arrays_changes(super)) |
9316 | return NULL; | |
1af97990 | 9317 | |
fc8ca064 AK |
9318 | /* Cannot activate another spare if rebuild is in progress already |
9319 | */ | |
9320 | if (is_rebuilding(dev)) { | |
7a862a02 | 9321 | dprintf("imsm: No spare activation allowed. Rebuild in progress already.\n"); |
fc8ca064 AK |
9322 | return NULL; |
9323 | } | |
9324 | ||
89c67882 AK |
9325 | if (a->info.array.level == 4) |
9326 | /* No repair for takeovered array | |
9327 | * imsm doesn't support raid4 | |
9328 | */ | |
9329 | return NULL; | |
9330 | ||
3b451610 AK |
9331 | if (imsm_check_degraded(super, dev, failed, MAP_0) != |
9332 | IMSM_T_STATE_DEGRADED) | |
88758e9d DW |
9333 | return NULL; |
9334 | ||
83ca7d45 AP |
9335 | if (get_imsm_map(dev, MAP_0)->map_state == IMSM_T_STATE_UNINITIALIZED) { |
9336 | dprintf("imsm: No spare activation allowed. Volume is not initialized.\n"); | |
9337 | return NULL; | |
9338 | } | |
9339 | ||
95d07a2c LM |
9340 | /* |
9341 | * If there are any failed disks check state of the other volume. | |
9342 | * Block rebuild if the another one is failed until failed disks | |
9343 | * are removed from container. | |
9344 | */ | |
9345 | if (failed) { | |
7a862a02 | 9346 | dprintf("found failed disks in %.*s, check if there anotherfailed sub-array.\n", |
c4acd1e5 | 9347 | MAX_RAID_SERIAL_LEN, dev->volume); |
95d07a2c LM |
9348 | /* check if states of the other volumes allow for rebuild */ |
9349 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
9350 | if (i != inst) { | |
9351 | allowed = imsm_rebuild_allowed(a->container, | |
9352 | i, failed); | |
9353 | if (!allowed) | |
9354 | return NULL; | |
9355 | } | |
9356 | } | |
9357 | } | |
9358 | ||
88758e9d | 9359 | /* For each slot, if it is not working, find a spare */ |
88758e9d DW |
9360 | for (i = 0; i < a->info.array.raid_disks; i++) { |
9361 | for (d = a->info.devs ; d ; d = d->next) | |
9362 | if (d->disk.raid_disk == i) | |
9363 | break; | |
9364 | dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0); | |
4389ce73 | 9365 | if (d && is_fd_valid(d->state_fd)) |
88758e9d DW |
9366 | continue; |
9367 | ||
272906ef | 9368 | /* |
a20d2ba5 DW |
9369 | * OK, this device needs recovery. Try to re-add the |
9370 | * previous occupant of this slot, if this fails see if | |
9371 | * we can continue the assimilation of a spare that was | |
9372 | * partially assimilated, finally try to activate a new | |
9373 | * spare. | |
272906ef DW |
9374 | */ |
9375 | dl = imsm_readd(super, i, a); | |
9376 | if (!dl) | |
b303fe21 | 9377 | dl = imsm_add_spare(super, i, a, 0, rv); |
a20d2ba5 | 9378 | if (!dl) |
b303fe21 | 9379 | dl = imsm_add_spare(super, i, a, 1, rv); |
272906ef DW |
9380 | if (!dl) |
9381 | continue; | |
1011e834 | 9382 | |
272906ef | 9383 | /* found a usable disk with enough space */ |
503975b9 | 9384 | di = xcalloc(1, sizeof(*di)); |
272906ef DW |
9385 | |
9386 | /* dl->index will be -1 in the case we are activating a | |
9387 | * pristine spare. imsm_process_update() will create a | |
9388 | * new index in this case. Once a disk is found to be | |
9389 | * failed in all member arrays it is kicked from the | |
9390 | * metadata | |
9391 | */ | |
9392 | di->disk.number = dl->index; | |
d23fe947 | 9393 | |
272906ef DW |
9394 | /* (ab)use di->devs to store a pointer to the device |
9395 | * we chose | |
9396 | */ | |
9397 | di->devs = (struct mdinfo *) dl; | |
9398 | ||
9399 | di->disk.raid_disk = i; | |
9400 | di->disk.major = dl->major; | |
9401 | di->disk.minor = dl->minor; | |
9402 | di->disk.state = 0; | |
d23534e4 | 9403 | di->recovery_start = 0; |
5551b113 | 9404 | di->data_offset = pba_of_lba0(map); |
272906ef DW |
9405 | di->component_size = a->info.component_size; |
9406 | di->container_member = inst; | |
5e46202e | 9407 | di->bb.supported = 1; |
2c8890e9 | 9408 | if (a->info.consistency_policy == CONSISTENCY_POLICY_PPL) { |
2432ce9b | 9409 | di->ppl_sector = get_ppl_sector(super, inst); |
c2462068 | 9410 | di->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9; |
2432ce9b | 9411 | } |
148acb7b | 9412 | super->random = random32(); |
272906ef DW |
9413 | di->next = rv; |
9414 | rv = di; | |
9415 | num_spares++; | |
9416 | dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor, | |
9417 | i, di->data_offset); | |
88758e9d DW |
9418 | } |
9419 | ||
9420 | if (!rv) | |
9421 | /* No spares found */ | |
9422 | return rv; | |
9423 | /* Now 'rv' has a list of devices to return. | |
9424 | * Create a metadata_update record to update the | |
9425 | * disk_ord_tbl for the array | |
9426 | */ | |
503975b9 | 9427 | mu = xmalloc(sizeof(*mu)); |
1011e834 | 9428 | mu->buf = xcalloc(num_spares, |
503975b9 | 9429 | sizeof(struct imsm_update_activate_spare)); |
88758e9d | 9430 | mu->space = NULL; |
cb23f1f4 | 9431 | mu->space_list = NULL; |
88758e9d DW |
9432 | mu->len = sizeof(struct imsm_update_activate_spare) * num_spares; |
9433 | mu->next = *updates; | |
9434 | u = (struct imsm_update_activate_spare *) mu->buf; | |
9435 | ||
9436 | for (di = rv ; di ; di = di->next) { | |
9437 | u->type = update_activate_spare; | |
d23fe947 DW |
9438 | u->dl = (struct dl *) di->devs; |
9439 | di->devs = NULL; | |
88758e9d DW |
9440 | u->slot = di->disk.raid_disk; |
9441 | u->array = inst; | |
9442 | u->next = u + 1; | |
9443 | u++; | |
9444 | } | |
9445 | (u-1)->next = NULL; | |
9446 | *updates = mu; | |
9447 | ||
9448 | return rv; | |
9449 | } | |
9450 | ||
54c2c1ea | 9451 | static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u) |
8273f55e | 9452 | { |
54c2c1ea | 9453 | struct imsm_dev *dev = get_imsm_dev(super, idx); |
238c0a71 AK |
9454 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
9455 | struct imsm_map *new_map = get_imsm_map(&u->dev, MAP_0); | |
54c2c1ea DW |
9456 | struct disk_info *inf = get_disk_info(u); |
9457 | struct imsm_disk *disk; | |
8273f55e DW |
9458 | int i; |
9459 | int j; | |
8273f55e | 9460 | |
54c2c1ea | 9461 | for (i = 0; i < map->num_members; i++) { |
238c0a71 | 9462 | disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, MAP_X)); |
54c2c1ea DW |
9463 | for (j = 0; j < new_map->num_members; j++) |
9464 | if (serialcmp(disk->serial, inf[j].serial) == 0) | |
8273f55e DW |
9465 | return 1; |
9466 | } | |
9467 | ||
9468 | return 0; | |
9469 | } | |
9470 | ||
1a64be56 LM |
9471 | static struct dl *get_disk_super(struct intel_super *super, int major, int minor) |
9472 | { | |
594dc1b8 JS |
9473 | struct dl *dl; |
9474 | ||
1a64be56 | 9475 | for (dl = super->disks; dl; dl = dl->next) |
089f9d79 | 9476 | if (dl->major == major && dl->minor == minor) |
1a64be56 LM |
9477 | return dl; |
9478 | return NULL; | |
9479 | } | |
9480 | ||
9481 | static int remove_disk_super(struct intel_super *super, int major, int minor) | |
9482 | { | |
594dc1b8 | 9483 | struct dl *prev; |
1a64be56 LM |
9484 | struct dl *dl; |
9485 | ||
9486 | prev = NULL; | |
9487 | for (dl = super->disks; dl; dl = dl->next) { | |
089f9d79 | 9488 | if (dl->major == major && dl->minor == minor) { |
1a64be56 LM |
9489 | /* remove */ |
9490 | if (prev) | |
9491 | prev->next = dl->next; | |
9492 | else | |
9493 | super->disks = dl->next; | |
9494 | dl->next = NULL; | |
3a85bf0e | 9495 | __free_imsm_disk(dl, 1); |
1ade5cc1 | 9496 | dprintf("removed %x:%x\n", major, minor); |
1a64be56 LM |
9497 | break; |
9498 | } | |
9499 | prev = dl; | |
9500 | } | |
9501 | return 0; | |
9502 | } | |
9503 | ||
f21e18ca | 9504 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index); |
ae6aad82 | 9505 | |
1a64be56 LM |
9506 | static int add_remove_disk_update(struct intel_super *super) |
9507 | { | |
9508 | int check_degraded = 0; | |
594dc1b8 JS |
9509 | struct dl *disk; |
9510 | ||
1a64be56 LM |
9511 | /* add/remove some spares to/from the metadata/contrainer */ |
9512 | while (super->disk_mgmt_list) { | |
9513 | struct dl *disk_cfg; | |
9514 | ||
9515 | disk_cfg = super->disk_mgmt_list; | |
9516 | super->disk_mgmt_list = disk_cfg->next; | |
9517 | disk_cfg->next = NULL; | |
9518 | ||
9519 | if (disk_cfg->action == DISK_ADD) { | |
9520 | disk_cfg->next = super->disks; | |
9521 | super->disks = disk_cfg; | |
9522 | check_degraded = 1; | |
1ade5cc1 N |
9523 | dprintf("added %x:%x\n", |
9524 | disk_cfg->major, disk_cfg->minor); | |
1a64be56 LM |
9525 | } else if (disk_cfg->action == DISK_REMOVE) { |
9526 | dprintf("Disk remove action processed: %x.%x\n", | |
9527 | disk_cfg->major, disk_cfg->minor); | |
9528 | disk = get_disk_super(super, | |
9529 | disk_cfg->major, | |
9530 | disk_cfg->minor); | |
9531 | if (disk) { | |
9532 | /* store action status */ | |
9533 | disk->action = DISK_REMOVE; | |
9534 | /* remove spare disks only */ | |
9535 | if (disk->index == -1) { | |
9536 | remove_disk_super(super, | |
9537 | disk_cfg->major, | |
9538 | disk_cfg->minor); | |
91c97c54 MT |
9539 | } else { |
9540 | disk_cfg->fd = disk->fd; | |
9541 | disk->fd = -1; | |
1a64be56 LM |
9542 | } |
9543 | } | |
9544 | /* release allocate disk structure */ | |
3a85bf0e | 9545 | __free_imsm_disk(disk_cfg, 1); |
1a64be56 LM |
9546 | } |
9547 | } | |
9548 | return check_degraded; | |
9549 | } | |
9550 | ||
a29911da PC |
9551 | static int apply_reshape_migration_update(struct imsm_update_reshape_migration *u, |
9552 | struct intel_super *super, | |
9553 | void ***space_list) | |
9554 | { | |
9555 | struct intel_dev *id; | |
9556 | void **tofree = NULL; | |
9557 | int ret_val = 0; | |
9558 | ||
1ade5cc1 | 9559 | dprintf("(enter)\n"); |
089f9d79 | 9560 | if (u->subdev < 0 || u->subdev > 1) { |
a29911da PC |
9561 | dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev); |
9562 | return ret_val; | |
9563 | } | |
089f9d79 | 9564 | if (space_list == NULL || *space_list == NULL) { |
a29911da PC |
9565 | dprintf("imsm: Error: Memory is not allocated\n"); |
9566 | return ret_val; | |
9567 | } | |
9568 | ||
9569 | for (id = super->devlist ; id; id = id->next) { | |
9570 | if (id->index == (unsigned)u->subdev) { | |
9571 | struct imsm_dev *dev = get_imsm_dev(super, u->subdev); | |
9572 | struct imsm_map *map; | |
9573 | struct imsm_dev *new_dev = | |
9574 | (struct imsm_dev *)*space_list; | |
238c0a71 | 9575 | struct imsm_map *migr_map = get_imsm_map(dev, MAP_1); |
a29911da PC |
9576 | int to_state; |
9577 | struct dl *new_disk; | |
9578 | ||
9579 | if (new_dev == NULL) | |
9580 | return ret_val; | |
9581 | *space_list = **space_list; | |
9582 | memcpy(new_dev, dev, sizeof_imsm_dev(dev, 0)); | |
238c0a71 | 9583 | map = get_imsm_map(new_dev, MAP_0); |
a29911da PC |
9584 | if (migr_map) { |
9585 | dprintf("imsm: Error: migration in progress"); | |
9586 | return ret_val; | |
9587 | } | |
9588 | ||
9589 | to_state = map->map_state; | |
9590 | if ((u->new_level == 5) && (map->raid_level == 0)) { | |
9591 | map->num_members++; | |
9592 | /* this should not happen */ | |
9593 | if (u->new_disks[0] < 0) { | |
9594 | map->failed_disk_num = | |
9595 | map->num_members - 1; | |
9596 | to_state = IMSM_T_STATE_DEGRADED; | |
9597 | } else | |
9598 | to_state = IMSM_T_STATE_NORMAL; | |
9599 | } | |
8e59f3d8 | 9600 | migrate(new_dev, super, to_state, MIGR_GEN_MIGR); |
a29911da PC |
9601 | if (u->new_level > -1) |
9602 | map->raid_level = u->new_level; | |
238c0a71 | 9603 | migr_map = get_imsm_map(new_dev, MAP_1); |
a29911da PC |
9604 | if ((u->new_level == 5) && |
9605 | (migr_map->raid_level == 0)) { | |
9606 | int ord = map->num_members - 1; | |
9607 | migr_map->num_members--; | |
9608 | if (u->new_disks[0] < 0) | |
9609 | ord |= IMSM_ORD_REBUILD; | |
9610 | set_imsm_ord_tbl_ent(map, | |
9611 | map->num_members - 1, | |
9612 | ord); | |
9613 | } | |
9614 | id->dev = new_dev; | |
9615 | tofree = (void **)dev; | |
9616 | ||
4bba0439 PC |
9617 | /* update chunk size |
9618 | */ | |
06fb291a | 9619 | if (u->new_chunksize > 0) { |
9529d343 MD |
9620 | struct imsm_map *dest_map = |
9621 | get_imsm_map(dev, MAP_0); | |
06fb291a | 9622 | int used_disks = |
9529d343 | 9623 | imsm_num_data_members(dest_map); |
06fb291a PB |
9624 | |
9625 | if (used_disks == 0) | |
9626 | return ret_val; | |
9627 | ||
4bba0439 PC |
9628 | map->blocks_per_strip = |
9629 | __cpu_to_le16(u->new_chunksize * 2); | |
1c275381 | 9630 | update_num_data_stripes(map, imsm_dev_size(dev)); |
06fb291a | 9631 | } |
4bba0439 | 9632 | |
44490938 MD |
9633 | /* ensure blocks_per_member has valid value |
9634 | */ | |
9635 | set_blocks_per_member(map, | |
9636 | per_dev_array_size(map) + | |
9637 | NUM_BLOCKS_DIRTY_STRIPE_REGION); | |
9638 | ||
a29911da PC |
9639 | /* add disk |
9640 | */ | |
089f9d79 JS |
9641 | if (u->new_level != 5 || migr_map->raid_level != 0 || |
9642 | migr_map->raid_level == map->raid_level) | |
a29911da PC |
9643 | goto skip_disk_add; |
9644 | ||
9645 | if (u->new_disks[0] >= 0) { | |
9646 | /* use passes spare | |
9647 | */ | |
9648 | new_disk = get_disk_super(super, | |
9649 | major(u->new_disks[0]), | |
9650 | minor(u->new_disks[0])); | |
7a862a02 | 9651 | dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n", |
a29911da PC |
9652 | major(u->new_disks[0]), |
9653 | minor(u->new_disks[0]), | |
9654 | new_disk, new_disk->index); | |
9655 | if (new_disk == NULL) | |
9656 | goto error_disk_add; | |
9657 | ||
9658 | new_disk->index = map->num_members - 1; | |
9659 | /* slot to fill in autolayout | |
9660 | */ | |
9661 | new_disk->raiddisk = new_disk->index; | |
9662 | new_disk->disk.status |= CONFIGURED_DISK; | |
9663 | new_disk->disk.status &= ~SPARE_DISK; | |
9664 | } else | |
9665 | goto error_disk_add; | |
9666 | ||
9667 | skip_disk_add: | |
9668 | *tofree = *space_list; | |
9669 | /* calculate new size | |
9670 | */ | |
f3871fdc | 9671 | imsm_set_array_size(new_dev, -1); |
a29911da PC |
9672 | |
9673 | ret_val = 1; | |
9674 | } | |
9675 | } | |
9676 | ||
9677 | if (tofree) | |
9678 | *space_list = tofree; | |
9679 | return ret_val; | |
9680 | ||
9681 | error_disk_add: | |
9682 | dprintf("Error: imsm: Cannot find disk.\n"); | |
9683 | return ret_val; | |
9684 | } | |
9685 | ||
f3871fdc AK |
9686 | static int apply_size_change_update(struct imsm_update_size_change *u, |
9687 | struct intel_super *super) | |
9688 | { | |
9689 | struct intel_dev *id; | |
9690 | int ret_val = 0; | |
9691 | ||
1ade5cc1 | 9692 | dprintf("(enter)\n"); |
089f9d79 | 9693 | if (u->subdev < 0 || u->subdev > 1) { |
f3871fdc AK |
9694 | dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev); |
9695 | return ret_val; | |
9696 | } | |
9697 | ||
9698 | for (id = super->devlist ; id; id = id->next) { | |
9699 | if (id->index == (unsigned)u->subdev) { | |
9700 | struct imsm_dev *dev = get_imsm_dev(super, u->subdev); | |
9701 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
9529d343 | 9702 | int used_disks = imsm_num_data_members(map); |
f3871fdc | 9703 | unsigned long long blocks_per_member; |
44490938 MD |
9704 | unsigned long long new_size_per_disk; |
9705 | ||
9706 | if (used_disks == 0) | |
9707 | return 0; | |
f3871fdc AK |
9708 | |
9709 | /* calculate new size | |
9710 | */ | |
44490938 MD |
9711 | new_size_per_disk = u->new_size / used_disks; |
9712 | blocks_per_member = new_size_per_disk + | |
9713 | NUM_BLOCKS_DIRTY_STRIPE_REGION; | |
f3871fdc | 9714 | |
1c275381 MT |
9715 | imsm_set_array_size(dev, u->new_size); |
9716 | set_blocks_per_member(map, blocks_per_member); | |
9717 | update_num_data_stripes(map, u->new_size); | |
f3871fdc AK |
9718 | ret_val = 1; |
9719 | break; | |
9720 | } | |
9721 | } | |
9722 | ||
9723 | return ret_val; | |
9724 | } | |
9725 | ||
69d40de4 JR |
9726 | static int prepare_spare_to_activate(struct supertype *st, |
9727 | struct imsm_update_activate_spare *u) | |
9728 | { | |
9729 | struct intel_super *super = st->sb; | |
9730 | int prev_current_vol = super->current_vol; | |
9731 | struct active_array *a; | |
9732 | int ret = 1; | |
9733 | ||
9734 | for (a = st->arrays; a; a = a->next) | |
9735 | /* | |
9736 | * Additional initialization (adding bitmap header, filling | |
9737 | * the bitmap area with '1's to force initial rebuild for a whole | |
9738 | * data-area) is required when adding the spare to the volume | |
9739 | * with write-intent bitmap. | |
9740 | */ | |
9741 | if (a->info.container_member == u->array && | |
9742 | a->info.consistency_policy == CONSISTENCY_POLICY_BITMAP) { | |
9743 | struct dl *dl; | |
9744 | ||
9745 | for (dl = super->disks; dl; dl = dl->next) | |
9746 | if (dl == u->dl) | |
9747 | break; | |
9748 | if (!dl) | |
9749 | break; | |
9750 | ||
9751 | super->current_vol = u->array; | |
9752 | if (st->ss->write_bitmap(st, dl->fd, NoUpdate)) | |
9753 | ret = 0; | |
9754 | super->current_vol = prev_current_vol; | |
9755 | } | |
9756 | return ret; | |
9757 | } | |
9758 | ||
061d7da3 | 9759 | static int apply_update_activate_spare(struct imsm_update_activate_spare *u, |
ca9de185 | 9760 | struct intel_super *super, |
061d7da3 LO |
9761 | struct active_array *active_array) |
9762 | { | |
9763 | struct imsm_super *mpb = super->anchor; | |
9764 | struct imsm_dev *dev = get_imsm_dev(super, u->array); | |
238c0a71 | 9765 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
061d7da3 LO |
9766 | struct imsm_map *migr_map; |
9767 | struct active_array *a; | |
9768 | struct imsm_disk *disk; | |
9769 | __u8 to_state; | |
9770 | struct dl *dl; | |
9771 | unsigned int found; | |
9772 | int failed; | |
5961eeec | 9773 | int victim; |
061d7da3 | 9774 | int i; |
5961eeec | 9775 | int second_map_created = 0; |
061d7da3 | 9776 | |
5961eeec | 9777 | for (; u; u = u->next) { |
238c0a71 | 9778 | victim = get_imsm_disk_idx(dev, u->slot, MAP_X); |
061d7da3 | 9779 | |
5961eeec | 9780 | if (victim < 0) |
9781 | return 0; | |
061d7da3 | 9782 | |
5961eeec | 9783 | for (dl = super->disks; dl; dl = dl->next) |
9784 | if (dl == u->dl) | |
9785 | break; | |
061d7da3 | 9786 | |
5961eeec | 9787 | if (!dl) { |
7a862a02 | 9788 | pr_err("error: imsm_activate_spare passed an unknown disk (index: %d)\n", |
5961eeec | 9789 | u->dl->index); |
9790 | return 0; | |
9791 | } | |
061d7da3 | 9792 | |
5961eeec | 9793 | /* count failures (excluding rebuilds and the victim) |
9794 | * to determine map[0] state | |
9795 | */ | |
9796 | failed = 0; | |
9797 | for (i = 0; i < map->num_members; i++) { | |
9798 | if (i == u->slot) | |
9799 | continue; | |
9800 | disk = get_imsm_disk(super, | |
238c0a71 | 9801 | get_imsm_disk_idx(dev, i, MAP_X)); |
5961eeec | 9802 | if (!disk || is_failed(disk)) |
9803 | failed++; | |
9804 | } | |
061d7da3 | 9805 | |
5961eeec | 9806 | /* adding a pristine spare, assign a new index */ |
9807 | if (dl->index < 0) { | |
9808 | dl->index = super->anchor->num_disks; | |
9809 | super->anchor->num_disks++; | |
9810 | } | |
9811 | disk = &dl->disk; | |
9812 | disk->status |= CONFIGURED_DISK; | |
9813 | disk->status &= ~SPARE_DISK; | |
9814 | ||
9815 | /* mark rebuild */ | |
238c0a71 | 9816 | to_state = imsm_check_degraded(super, dev, failed, MAP_0); |
5961eeec | 9817 | if (!second_map_created) { |
9818 | second_map_created = 1; | |
9819 | map->map_state = IMSM_T_STATE_DEGRADED; | |
9820 | migrate(dev, super, to_state, MIGR_REBUILD); | |
9821 | } else | |
9822 | map->map_state = to_state; | |
238c0a71 | 9823 | migr_map = get_imsm_map(dev, MAP_1); |
5961eeec | 9824 | set_imsm_ord_tbl_ent(map, u->slot, dl->index); |
9825 | set_imsm_ord_tbl_ent(migr_map, u->slot, | |
9826 | dl->index | IMSM_ORD_REBUILD); | |
9827 | ||
9828 | /* update the family_num to mark a new container | |
9829 | * generation, being careful to record the existing | |
9830 | * family_num in orig_family_num to clean up after | |
9831 | * earlier mdadm versions that neglected to set it. | |
9832 | */ | |
9833 | if (mpb->orig_family_num == 0) | |
9834 | mpb->orig_family_num = mpb->family_num; | |
9835 | mpb->family_num += super->random; | |
9836 | ||
9837 | /* count arrays using the victim in the metadata */ | |
9838 | found = 0; | |
9839 | for (a = active_array; a ; a = a->next) { | |
76c152ca | 9840 | int dev_idx = a->info.container_member; |
061d7da3 | 9841 | |
76c152ca | 9842 | if (get_disk_slot_in_dev(super, dev_idx, victim) >= 0) |
5961eeec | 9843 | found++; |
9844 | } | |
061d7da3 | 9845 | |
5961eeec | 9846 | /* delete the victim if it is no longer being |
9847 | * utilized anywhere | |
061d7da3 | 9848 | */ |
5961eeec | 9849 | if (!found) { |
9850 | struct dl **dlp; | |
061d7da3 | 9851 | |
5961eeec | 9852 | /* We know that 'manager' isn't touching anything, |
9853 | * so it is safe to delete | |
9854 | */ | |
9855 | for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next) | |
061d7da3 LO |
9856 | if ((*dlp)->index == victim) |
9857 | break; | |
5961eeec | 9858 | |
9859 | /* victim may be on the missing list */ | |
9860 | if (!*dlp) | |
9861 | for (dlp = &super->missing; *dlp; | |
9862 | dlp = &(*dlp)->next) | |
9863 | if ((*dlp)->index == victim) | |
9864 | break; | |
9865 | imsm_delete(super, dlp, victim); | |
9866 | } | |
061d7da3 LO |
9867 | } |
9868 | ||
9869 | return 1; | |
9870 | } | |
a29911da | 9871 | |
2e5dc010 N |
9872 | static int apply_reshape_container_disks_update(struct imsm_update_reshape *u, |
9873 | struct intel_super *super, | |
9874 | void ***space_list) | |
9875 | { | |
9876 | struct dl *new_disk; | |
9877 | struct intel_dev *id; | |
9878 | int i; | |
9879 | int delta_disks = u->new_raid_disks - u->old_raid_disks; | |
ee4beede | 9880 | int disk_count = u->old_raid_disks; |
2e5dc010 N |
9881 | void **tofree = NULL; |
9882 | int devices_to_reshape = 1; | |
9883 | struct imsm_super *mpb = super->anchor; | |
9884 | int ret_val = 0; | |
d098291a | 9885 | unsigned int dev_id; |
2e5dc010 | 9886 | |
1ade5cc1 | 9887 | dprintf("(enter)\n"); |
2e5dc010 N |
9888 | |
9889 | /* enable spares to use in array */ | |
9890 | for (i = 0; i < delta_disks; i++) { | |
9891 | new_disk = get_disk_super(super, | |
9892 | major(u->new_disks[i]), | |
9893 | minor(u->new_disks[i])); | |
7a862a02 | 9894 | dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n", |
2e5dc010 N |
9895 | major(u->new_disks[i]), minor(u->new_disks[i]), |
9896 | new_disk, new_disk->index); | |
089f9d79 JS |
9897 | if (new_disk == NULL || |
9898 | (new_disk->index >= 0 && | |
9899 | new_disk->index < u->old_raid_disks)) | |
2e5dc010 | 9900 | goto update_reshape_exit; |
ee4beede | 9901 | new_disk->index = disk_count++; |
2e5dc010 N |
9902 | /* slot to fill in autolayout |
9903 | */ | |
9904 | new_disk->raiddisk = new_disk->index; | |
9905 | new_disk->disk.status |= | |
9906 | CONFIGURED_DISK; | |
9907 | new_disk->disk.status &= ~SPARE_DISK; | |
9908 | } | |
9909 | ||
ed7333bd AK |
9910 | dprintf("imsm: volume set mpb->num_raid_devs = %i\n", |
9911 | mpb->num_raid_devs); | |
2e5dc010 N |
9912 | /* manage changes in volume |
9913 | */ | |
d098291a | 9914 | for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) { |
2e5dc010 N |
9915 | void **sp = *space_list; |
9916 | struct imsm_dev *newdev; | |
9917 | struct imsm_map *newmap, *oldmap; | |
9918 | ||
d098291a AK |
9919 | for (id = super->devlist ; id; id = id->next) { |
9920 | if (id->index == dev_id) | |
9921 | break; | |
9922 | } | |
9923 | if (id == NULL) | |
9924 | break; | |
2e5dc010 N |
9925 | if (!sp) |
9926 | continue; | |
9927 | *space_list = *sp; | |
9928 | newdev = (void*)sp; | |
9929 | /* Copy the dev, but not (all of) the map */ | |
9930 | memcpy(newdev, id->dev, sizeof(*newdev)); | |
238c0a71 AK |
9931 | oldmap = get_imsm_map(id->dev, MAP_0); |
9932 | newmap = get_imsm_map(newdev, MAP_0); | |
2e5dc010 N |
9933 | /* Copy the current map */ |
9934 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); | |
9935 | /* update one device only | |
9936 | */ | |
9937 | if (devices_to_reshape) { | |
ed7333bd AK |
9938 | dprintf("imsm: modifying subdev: %i\n", |
9939 | id->index); | |
2e5dc010 N |
9940 | devices_to_reshape--; |
9941 | newdev->vol.migr_state = 1; | |
4036e7ee | 9942 | set_vol_curr_migr_unit(newdev, 0); |
ea672ee1 | 9943 | set_migr_type(newdev, MIGR_GEN_MIGR); |
2e5dc010 N |
9944 | newmap->num_members = u->new_raid_disks; |
9945 | for (i = 0; i < delta_disks; i++) { | |
9946 | set_imsm_ord_tbl_ent(newmap, | |
9947 | u->old_raid_disks + i, | |
9948 | u->old_raid_disks + i); | |
9949 | } | |
9950 | /* New map is correct, now need to save old map | |
9951 | */ | |
238c0a71 | 9952 | newmap = get_imsm_map(newdev, MAP_1); |
2e5dc010 N |
9953 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); |
9954 | ||
f3871fdc | 9955 | imsm_set_array_size(newdev, -1); |
2e5dc010 N |
9956 | } |
9957 | ||
9958 | sp = (void **)id->dev; | |
9959 | id->dev = newdev; | |
9960 | *sp = tofree; | |
9961 | tofree = sp; | |
8e59f3d8 AK |
9962 | |
9963 | /* Clear migration record */ | |
9964 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
2e5dc010 | 9965 | } |
819bc634 AK |
9966 | if (tofree) |
9967 | *space_list = tofree; | |
2e5dc010 N |
9968 | ret_val = 1; |
9969 | ||
9970 | update_reshape_exit: | |
9971 | ||
9972 | return ret_val; | |
9973 | } | |
9974 | ||
bb025c2f | 9975 | static int apply_takeover_update(struct imsm_update_takeover *u, |
8ca6df95 KW |
9976 | struct intel_super *super, |
9977 | void ***space_list) | |
bb025c2f KW |
9978 | { |
9979 | struct imsm_dev *dev = NULL; | |
8ca6df95 KW |
9980 | struct intel_dev *dv; |
9981 | struct imsm_dev *dev_new; | |
bb025c2f KW |
9982 | struct imsm_map *map; |
9983 | struct dl *dm, *du; | |
8ca6df95 | 9984 | int i; |
bb025c2f KW |
9985 | |
9986 | for (dv = super->devlist; dv; dv = dv->next) | |
9987 | if (dv->index == (unsigned int)u->subarray) { | |
9988 | dev = dv->dev; | |
9989 | break; | |
9990 | } | |
9991 | ||
9992 | if (dev == NULL) | |
9993 | return 0; | |
9994 | ||
238c0a71 | 9995 | map = get_imsm_map(dev, MAP_0); |
bb025c2f KW |
9996 | |
9997 | if (u->direction == R10_TO_R0) { | |
43d5ec18 | 9998 | /* Number of failed disks must be half of initial disk number */ |
3b451610 AK |
9999 | if (imsm_count_failed(super, dev, MAP_0) != |
10000 | (map->num_members / 2)) | |
43d5ec18 KW |
10001 | return 0; |
10002 | ||
bb025c2f KW |
10003 | /* iterate through devices to mark removed disks as spare */ |
10004 | for (dm = super->disks; dm; dm = dm->next) { | |
10005 | if (dm->disk.status & FAILED_DISK) { | |
10006 | int idx = dm->index; | |
10007 | /* update indexes on the disk list */ | |
10008 | /* FIXME this loop-with-the-loop looks wrong, I'm not convinced | |
10009 | the index values will end up being correct.... NB */ | |
10010 | for (du = super->disks; du; du = du->next) | |
10011 | if (du->index > idx) | |
10012 | du->index--; | |
10013 | /* mark as spare disk */ | |
a8619d23 | 10014 | mark_spare(dm); |
bb025c2f KW |
10015 | } |
10016 | } | |
bb025c2f | 10017 | /* update map */ |
1c275381 | 10018 | map->num_members /= map->num_domains; |
bb025c2f | 10019 | map->map_state = IMSM_T_STATE_NORMAL; |
bb025c2f | 10020 | map->raid_level = 0; |
1c275381 MT |
10021 | set_num_domains(map); |
10022 | update_num_data_stripes(map, imsm_dev_size(dev)); | |
bb025c2f KW |
10023 | map->failed_disk_num = -1; |
10024 | } | |
10025 | ||
8ca6df95 KW |
10026 | if (u->direction == R0_TO_R10) { |
10027 | void **space; | |
4a353e6e | 10028 | |
8ca6df95 KW |
10029 | /* update slots in current disk list */ |
10030 | for (dm = super->disks; dm; dm = dm->next) { | |
10031 | if (dm->index >= 0) | |
10032 | dm->index *= 2; | |
10033 | } | |
10034 | /* create new *missing* disks */ | |
10035 | for (i = 0; i < map->num_members; i++) { | |
10036 | space = *space_list; | |
10037 | if (!space) | |
10038 | continue; | |
10039 | *space_list = *space; | |
10040 | du = (void *)space; | |
10041 | memcpy(du, super->disks, sizeof(*du)); | |
8ca6df95 KW |
10042 | du->fd = -1; |
10043 | du->minor = 0; | |
10044 | du->major = 0; | |
10045 | du->index = (i * 2) + 1; | |
10046 | sprintf((char *)du->disk.serial, | |
10047 | " MISSING_%d", du->index); | |
10048 | sprintf((char *)du->serial, | |
10049 | "MISSING_%d", du->index); | |
10050 | du->next = super->missing; | |
10051 | super->missing = du; | |
10052 | } | |
10053 | /* create new dev and map */ | |
10054 | space = *space_list; | |
10055 | if (!space) | |
10056 | return 0; | |
10057 | *space_list = *space; | |
10058 | dev_new = (void *)space; | |
10059 | memcpy(dev_new, dev, sizeof(*dev)); | |
10060 | /* update new map */ | |
238c0a71 | 10061 | map = get_imsm_map(dev_new, MAP_0); |
1c275381 | 10062 | |
1a2487c2 | 10063 | map->map_state = IMSM_T_STATE_DEGRADED; |
8ca6df95 | 10064 | map->raid_level = 1; |
1c275381 MT |
10065 | set_num_domains(map); |
10066 | map->num_members = map->num_members * map->num_domains; | |
10067 | update_num_data_stripes(map, imsm_dev_size(dev)); | |
4a353e6e | 10068 | |
8ca6df95 KW |
10069 | /* replace dev<->dev_new */ |
10070 | dv->dev = dev_new; | |
10071 | } | |
bb025c2f KW |
10072 | /* update disk order table */ |
10073 | for (du = super->disks; du; du = du->next) | |
10074 | if (du->index >= 0) | |
10075 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
8ca6df95 | 10076 | for (du = super->missing; du; du = du->next) |
1a2487c2 KW |
10077 | if (du->index >= 0) { |
10078 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
4c9e8c1e | 10079 | mark_missing(super, dv->dev, &du->disk, du->index); |
1a2487c2 | 10080 | } |
bb025c2f KW |
10081 | |
10082 | return 1; | |
10083 | } | |
10084 | ||
e8319a19 DW |
10085 | static void imsm_process_update(struct supertype *st, |
10086 | struct metadata_update *update) | |
10087 | { | |
10088 | /** | |
10089 | * crack open the metadata_update envelope to find the update record | |
10090 | * update can be one of: | |
d195167d AK |
10091 | * update_reshape_container_disks - all the arrays in the container |
10092 | * are being reshaped to have more devices. We need to mark | |
10093 | * the arrays for general migration and convert selected spares | |
10094 | * into active devices. | |
10095 | * update_activate_spare - a spare device has replaced a failed | |
1011e834 N |
10096 | * device in an array, update the disk_ord_tbl. If this disk is |
10097 | * present in all member arrays then also clear the SPARE_DISK | |
10098 | * flag | |
d195167d AK |
10099 | * update_create_array |
10100 | * update_kill_array | |
10101 | * update_rename_array | |
10102 | * update_add_remove_disk | |
e8319a19 DW |
10103 | */ |
10104 | struct intel_super *super = st->sb; | |
4d7b1503 | 10105 | struct imsm_super *mpb; |
e8319a19 DW |
10106 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
10107 | ||
4d7b1503 DW |
10108 | /* update requires a larger buf but the allocation failed */ |
10109 | if (super->next_len && !super->next_buf) { | |
10110 | super->next_len = 0; | |
10111 | return; | |
10112 | } | |
10113 | ||
10114 | if (super->next_buf) { | |
10115 | memcpy(super->next_buf, super->buf, super->len); | |
10116 | free(super->buf); | |
10117 | super->len = super->next_len; | |
10118 | super->buf = super->next_buf; | |
10119 | ||
10120 | super->next_len = 0; | |
10121 | super->next_buf = NULL; | |
10122 | } | |
10123 | ||
10124 | mpb = super->anchor; | |
10125 | ||
e8319a19 | 10126 | switch (type) { |
0ec5d470 AK |
10127 | case update_general_migration_checkpoint: { |
10128 | struct intel_dev *id; | |
10129 | struct imsm_update_general_migration_checkpoint *u = | |
10130 | (void *)update->buf; | |
10131 | ||
1ade5cc1 | 10132 | dprintf("called for update_general_migration_checkpoint\n"); |
0ec5d470 AK |
10133 | |
10134 | /* find device under general migration */ | |
10135 | for (id = super->devlist ; id; id = id->next) { | |
10136 | if (is_gen_migration(id->dev)) { | |
4036e7ee MT |
10137 | set_vol_curr_migr_unit(id->dev, |
10138 | u->curr_migr_unit); | |
0ec5d470 AK |
10139 | super->updates_pending++; |
10140 | } | |
10141 | } | |
10142 | break; | |
10143 | } | |
bb025c2f KW |
10144 | case update_takeover: { |
10145 | struct imsm_update_takeover *u = (void *)update->buf; | |
1a2487c2 KW |
10146 | if (apply_takeover_update(u, super, &update->space_list)) { |
10147 | imsm_update_version_info(super); | |
bb025c2f | 10148 | super->updates_pending++; |
1a2487c2 | 10149 | } |
bb025c2f KW |
10150 | break; |
10151 | } | |
10152 | ||
78b10e66 | 10153 | case update_reshape_container_disks: { |
d195167d | 10154 | struct imsm_update_reshape *u = (void *)update->buf; |
2e5dc010 N |
10155 | if (apply_reshape_container_disks_update( |
10156 | u, super, &update->space_list)) | |
10157 | super->updates_pending++; | |
78b10e66 N |
10158 | break; |
10159 | } | |
48c5303a | 10160 | case update_reshape_migration: { |
a29911da PC |
10161 | struct imsm_update_reshape_migration *u = (void *)update->buf; |
10162 | if (apply_reshape_migration_update( | |
10163 | u, super, &update->space_list)) | |
10164 | super->updates_pending++; | |
48c5303a PC |
10165 | break; |
10166 | } | |
f3871fdc AK |
10167 | case update_size_change: { |
10168 | struct imsm_update_size_change *u = (void *)update->buf; | |
10169 | if (apply_size_change_update(u, super)) | |
10170 | super->updates_pending++; | |
10171 | break; | |
10172 | } | |
e8319a19 | 10173 | case update_activate_spare: { |
1011e834 | 10174 | struct imsm_update_activate_spare *u = (void *) update->buf; |
69d40de4 JR |
10175 | |
10176 | if (prepare_spare_to_activate(st, u) && | |
10177 | apply_update_activate_spare(u, super, st->arrays)) | |
061d7da3 | 10178 | super->updates_pending++; |
8273f55e DW |
10179 | break; |
10180 | } | |
10181 | case update_create_array: { | |
10182 | /* someone wants to create a new array, we need to be aware of | |
10183 | * a few races/collisions: | |
10184 | * 1/ 'Create' called by two separate instances of mdadm | |
10185 | * 2/ 'Create' versus 'activate_spare': mdadm has chosen | |
10186 | * devices that have since been assimilated via | |
10187 | * activate_spare. | |
10188 | * In the event this update can not be carried out mdadm will | |
10189 | * (FIX ME) notice that its update did not take hold. | |
10190 | */ | |
10191 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 10192 | struct intel_dev *dv; |
8273f55e DW |
10193 | struct imsm_dev *dev; |
10194 | struct imsm_map *map, *new_map; | |
10195 | unsigned long long start, end; | |
10196 | unsigned long long new_start, new_end; | |
10197 | int i; | |
54c2c1ea DW |
10198 | struct disk_info *inf; |
10199 | struct dl *dl; | |
8273f55e DW |
10200 | |
10201 | /* handle racing creates: first come first serve */ | |
10202 | if (u->dev_idx < mpb->num_raid_devs) { | |
1ade5cc1 | 10203 | dprintf("subarray %d already defined\n", u->dev_idx); |
ba2de7ba | 10204 | goto create_error; |
8273f55e DW |
10205 | } |
10206 | ||
10207 | /* check update is next in sequence */ | |
10208 | if (u->dev_idx != mpb->num_raid_devs) { | |
1ade5cc1 N |
10209 | dprintf("can not create array %d expected index %d\n", |
10210 | u->dev_idx, mpb->num_raid_devs); | |
ba2de7ba | 10211 | goto create_error; |
8273f55e DW |
10212 | } |
10213 | ||
238c0a71 | 10214 | new_map = get_imsm_map(&u->dev, MAP_0); |
5551b113 | 10215 | new_start = pba_of_lba0(new_map); |
44490938 | 10216 | new_end = new_start + per_dev_array_size(new_map); |
54c2c1ea | 10217 | inf = get_disk_info(u); |
8273f55e DW |
10218 | |
10219 | /* handle activate_spare versus create race: | |
10220 | * check to make sure that overlapping arrays do not include | |
10221 | * overalpping disks | |
10222 | */ | |
10223 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
949c47a0 | 10224 | dev = get_imsm_dev(super, i); |
238c0a71 | 10225 | map = get_imsm_map(dev, MAP_0); |
5551b113 | 10226 | start = pba_of_lba0(map); |
44490938 | 10227 | end = start + per_dev_array_size(map); |
8273f55e DW |
10228 | if ((new_start >= start && new_start <= end) || |
10229 | (start >= new_start && start <= new_end)) | |
54c2c1ea DW |
10230 | /* overlap */; |
10231 | else | |
10232 | continue; | |
10233 | ||
10234 | if (disks_overlap(super, i, u)) { | |
1ade5cc1 | 10235 | dprintf("arrays overlap\n"); |
ba2de7ba | 10236 | goto create_error; |
8273f55e DW |
10237 | } |
10238 | } | |
8273f55e | 10239 | |
949c47a0 DW |
10240 | /* check that prepare update was successful */ |
10241 | if (!update->space) { | |
1ade5cc1 | 10242 | dprintf("prepare update failed\n"); |
ba2de7ba | 10243 | goto create_error; |
949c47a0 DW |
10244 | } |
10245 | ||
54c2c1ea DW |
10246 | /* check that all disks are still active before committing |
10247 | * changes. FIXME: could we instead handle this by creating a | |
10248 | * degraded array? That's probably not what the user expects, | |
10249 | * so better to drop this update on the floor. | |
10250 | */ | |
10251 | for (i = 0; i < new_map->num_members; i++) { | |
10252 | dl = serial_to_dl(inf[i].serial, super); | |
10253 | if (!dl) { | |
1ade5cc1 | 10254 | dprintf("disk disappeared\n"); |
ba2de7ba | 10255 | goto create_error; |
54c2c1ea | 10256 | } |
949c47a0 DW |
10257 | } |
10258 | ||
8273f55e | 10259 | super->updates_pending++; |
54c2c1ea DW |
10260 | |
10261 | /* convert spares to members and fixup ord_tbl */ | |
10262 | for (i = 0; i < new_map->num_members; i++) { | |
10263 | dl = serial_to_dl(inf[i].serial, super); | |
10264 | if (dl->index == -1) { | |
10265 | dl->index = mpb->num_disks; | |
10266 | mpb->num_disks++; | |
10267 | dl->disk.status |= CONFIGURED_DISK; | |
10268 | dl->disk.status &= ~SPARE_DISK; | |
10269 | } | |
10270 | set_imsm_ord_tbl_ent(new_map, i, dl->index); | |
10271 | } | |
10272 | ||
ba2de7ba DW |
10273 | dv = update->space; |
10274 | dev = dv->dev; | |
949c47a0 DW |
10275 | update->space = NULL; |
10276 | imsm_copy_dev(dev, &u->dev); | |
ba2de7ba DW |
10277 | dv->index = u->dev_idx; |
10278 | dv->next = super->devlist; | |
10279 | super->devlist = dv; | |
8273f55e | 10280 | mpb->num_raid_devs++; |
8273f55e | 10281 | |
4d1313e9 | 10282 | imsm_update_version_info(super); |
8273f55e | 10283 | break; |
ba2de7ba DW |
10284 | create_error: |
10285 | /* mdmon knows how to release update->space, but not | |
10286 | * ((struct intel_dev *) update->space)->dev | |
10287 | */ | |
10288 | if (update->space) { | |
10289 | dv = update->space; | |
10290 | free(dv->dev); | |
10291 | } | |
8273f55e | 10292 | break; |
e8319a19 | 10293 | } |
33414a01 DW |
10294 | case update_kill_array: { |
10295 | struct imsm_update_kill_array *u = (void *) update->buf; | |
10296 | int victim = u->dev_idx; | |
10297 | struct active_array *a; | |
10298 | struct intel_dev **dp; | |
33414a01 DW |
10299 | |
10300 | /* sanity check that we are not affecting the uuid of | |
10301 | * active arrays, or deleting an active array | |
10302 | * | |
10303 | * FIXME when immutable ids are available, but note that | |
10304 | * we'll also need to fixup the invalidated/active | |
10305 | * subarray indexes in mdstat | |
10306 | */ | |
10307 | for (a = st->arrays; a; a = a->next) | |
10308 | if (a->info.container_member >= victim) | |
10309 | break; | |
10310 | /* by definition if mdmon is running at least one array | |
10311 | * is active in the container, so checking | |
10312 | * mpb->num_raid_devs is just extra paranoia | |
10313 | */ | |
756a15f3 | 10314 | if (a || mpb->num_raid_devs == 1 || victim >= super->anchor->num_raid_devs) { |
33414a01 DW |
10315 | dprintf("failed to delete subarray-%d\n", victim); |
10316 | break; | |
10317 | } | |
10318 | ||
10319 | for (dp = &super->devlist; *dp;) | |
f21e18ca | 10320 | if ((*dp)->index == (unsigned)super->current_vol) { |
33414a01 DW |
10321 | *dp = (*dp)->next; |
10322 | } else { | |
f21e18ca | 10323 | if ((*dp)->index > (unsigned)victim) |
33414a01 DW |
10324 | (*dp)->index--; |
10325 | dp = &(*dp)->next; | |
10326 | } | |
10327 | mpb->num_raid_devs--; | |
10328 | super->updates_pending++; | |
10329 | break; | |
10330 | } | |
aa534678 DW |
10331 | case update_rename_array: { |
10332 | struct imsm_update_rename_array *u = (void *) update->buf; | |
10333 | char name[MAX_RAID_SERIAL_LEN+1]; | |
10334 | int target = u->dev_idx; | |
10335 | struct active_array *a; | |
10336 | struct imsm_dev *dev; | |
10337 | ||
10338 | /* sanity check that we are not affecting the uuid of | |
10339 | * an active array | |
10340 | */ | |
40659392 | 10341 | memset(name, 0, sizeof(name)); |
aa534678 DW |
10342 | snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name); |
10343 | name[MAX_RAID_SERIAL_LEN] = '\0'; | |
10344 | for (a = st->arrays; a; a = a->next) | |
10345 | if (a->info.container_member == target) | |
10346 | break; | |
10347 | dev = get_imsm_dev(super, u->dev_idx); | |
756a15f3 | 10348 | if (a || !check_name(super, name, 1)) { |
aa534678 DW |
10349 | dprintf("failed to rename subarray-%d\n", target); |
10350 | break; | |
10351 | } | |
10352 | ||
40659392 | 10353 | memcpy(dev->volume, name, MAX_RAID_SERIAL_LEN); |
aa534678 DW |
10354 | super->updates_pending++; |
10355 | break; | |
10356 | } | |
1a64be56 | 10357 | case update_add_remove_disk: { |
43dad3d6 | 10358 | /* we may be able to repair some arrays if disks are |
095b8088 | 10359 | * being added, check the status of add_remove_disk |
1a64be56 LM |
10360 | * if discs has been added. |
10361 | */ | |
10362 | if (add_remove_disk_update(super)) { | |
43dad3d6 | 10363 | struct active_array *a; |
072b727f DW |
10364 | |
10365 | super->updates_pending++; | |
1a64be56 | 10366 | for (a = st->arrays; a; a = a->next) |
43dad3d6 DW |
10367 | a->check_degraded = 1; |
10368 | } | |
43dad3d6 | 10369 | break; |
e8319a19 | 10370 | } |
bbab0940 TM |
10371 | case update_prealloc_badblocks_mem: |
10372 | break; | |
e6e9dd3f AP |
10373 | case update_rwh_policy: { |
10374 | struct imsm_update_rwh_policy *u = (void *)update->buf; | |
10375 | int target = u->dev_idx; | |
10376 | struct imsm_dev *dev = get_imsm_dev(super, target); | |
e6e9dd3f AP |
10377 | |
10378 | if (dev->rwh_policy != u->new_policy) { | |
10379 | dev->rwh_policy = u->new_policy; | |
10380 | super->updates_pending++; | |
10381 | } | |
10382 | break; | |
10383 | } | |
1a64be56 | 10384 | default: |
ebf3be99 | 10385 | pr_err("error: unsupported process update type:(type: %d)\n", type); |
1a64be56 | 10386 | } |
e8319a19 | 10387 | } |
88758e9d | 10388 | |
bc0b9d34 PC |
10389 | static struct mdinfo *get_spares_for_grow(struct supertype *st); |
10390 | ||
5fe6f031 N |
10391 | static int imsm_prepare_update(struct supertype *st, |
10392 | struct metadata_update *update) | |
8273f55e | 10393 | { |
949c47a0 | 10394 | /** |
4d7b1503 DW |
10395 | * Allocate space to hold new disk entries, raid-device entries or a new |
10396 | * mpb if necessary. The manager synchronously waits for updates to | |
10397 | * complete in the monitor, so new mpb buffers allocated here can be | |
10398 | * integrated by the monitor thread without worrying about live pointers | |
10399 | * in the manager thread. | |
8273f55e | 10400 | */ |
095b8088 | 10401 | enum imsm_update_type type; |
4d7b1503 | 10402 | struct intel_super *super = st->sb; |
f36a9ecd | 10403 | unsigned int sector_size = super->sector_size; |
4d7b1503 DW |
10404 | struct imsm_super *mpb = super->anchor; |
10405 | size_t buf_len; | |
10406 | size_t len = 0; | |
949c47a0 | 10407 | |
095b8088 N |
10408 | if (update->len < (int)sizeof(type)) |
10409 | return 0; | |
10410 | ||
10411 | type = *(enum imsm_update_type *) update->buf; | |
10412 | ||
949c47a0 | 10413 | switch (type) { |
0ec5d470 | 10414 | case update_general_migration_checkpoint: |
095b8088 N |
10415 | if (update->len < (int)sizeof(struct imsm_update_general_migration_checkpoint)) |
10416 | return 0; | |
1ade5cc1 | 10417 | dprintf("called for update_general_migration_checkpoint\n"); |
0ec5d470 | 10418 | break; |
abedf5fc KW |
10419 | case update_takeover: { |
10420 | struct imsm_update_takeover *u = (void *)update->buf; | |
095b8088 N |
10421 | if (update->len < (int)sizeof(*u)) |
10422 | return 0; | |
abedf5fc KW |
10423 | if (u->direction == R0_TO_R10) { |
10424 | void **tail = (void **)&update->space_list; | |
10425 | struct imsm_dev *dev = get_imsm_dev(super, u->subarray); | |
238c0a71 | 10426 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
abedf5fc KW |
10427 | int num_members = map->num_members; |
10428 | void *space; | |
10429 | int size, i; | |
abedf5fc KW |
10430 | /* allocate memory for added disks */ |
10431 | for (i = 0; i < num_members; i++) { | |
10432 | size = sizeof(struct dl); | |
503975b9 | 10433 | space = xmalloc(size); |
abedf5fc KW |
10434 | *tail = space; |
10435 | tail = space; | |
10436 | *tail = NULL; | |
10437 | } | |
10438 | /* allocate memory for new device */ | |
10439 | size = sizeof_imsm_dev(super->devlist->dev, 0) + | |
10440 | (num_members * sizeof(__u32)); | |
503975b9 N |
10441 | space = xmalloc(size); |
10442 | *tail = space; | |
10443 | tail = space; | |
10444 | *tail = NULL; | |
10445 | len = disks_to_mpb_size(num_members * 2); | |
abedf5fc KW |
10446 | } |
10447 | ||
10448 | break; | |
10449 | } | |
78b10e66 | 10450 | case update_reshape_container_disks: { |
d195167d AK |
10451 | /* Every raid device in the container is about to |
10452 | * gain some more devices, and we will enter a | |
10453 | * reconfiguration. | |
10454 | * So each 'imsm_map' will be bigger, and the imsm_vol | |
10455 | * will now hold 2 of them. | |
10456 | * Thus we need new 'struct imsm_dev' allocations sized | |
10457 | * as sizeof_imsm_dev but with more devices in both maps. | |
10458 | */ | |
10459 | struct imsm_update_reshape *u = (void *)update->buf; | |
10460 | struct intel_dev *dl; | |
10461 | void **space_tail = (void**)&update->space_list; | |
10462 | ||
095b8088 N |
10463 | if (update->len < (int)sizeof(*u)) |
10464 | return 0; | |
10465 | ||
1ade5cc1 | 10466 | dprintf("for update_reshape\n"); |
d195167d AK |
10467 | |
10468 | for (dl = super->devlist; dl; dl = dl->next) { | |
10469 | int size = sizeof_imsm_dev(dl->dev, 1); | |
10470 | void *s; | |
d677e0b8 AK |
10471 | if (u->new_raid_disks > u->old_raid_disks) |
10472 | size += sizeof(__u32)*2* | |
10473 | (u->new_raid_disks - u->old_raid_disks); | |
503975b9 | 10474 | s = xmalloc(size); |
d195167d AK |
10475 | *space_tail = s; |
10476 | space_tail = s; | |
10477 | *space_tail = NULL; | |
10478 | } | |
10479 | ||
10480 | len = disks_to_mpb_size(u->new_raid_disks); | |
10481 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
78b10e66 N |
10482 | break; |
10483 | } | |
48c5303a | 10484 | case update_reshape_migration: { |
bc0b9d34 PC |
10485 | /* for migration level 0->5 we need to add disks |
10486 | * so the same as for container operation we will copy | |
10487 | * device to the bigger location. | |
10488 | * in memory prepared device and new disk area are prepared | |
10489 | * for usage in process update | |
10490 | */ | |
10491 | struct imsm_update_reshape_migration *u = (void *)update->buf; | |
10492 | struct intel_dev *id; | |
10493 | void **space_tail = (void **)&update->space_list; | |
10494 | int size; | |
10495 | void *s; | |
10496 | int current_level = -1; | |
10497 | ||
095b8088 N |
10498 | if (update->len < (int)sizeof(*u)) |
10499 | return 0; | |
10500 | ||
1ade5cc1 | 10501 | dprintf("for update_reshape\n"); |
bc0b9d34 PC |
10502 | |
10503 | /* add space for bigger array in update | |
10504 | */ | |
10505 | for (id = super->devlist; id; id = id->next) { | |
10506 | if (id->index == (unsigned)u->subdev) { | |
10507 | size = sizeof_imsm_dev(id->dev, 1); | |
10508 | if (u->new_raid_disks > u->old_raid_disks) | |
10509 | size += sizeof(__u32)*2* | |
10510 | (u->new_raid_disks - u->old_raid_disks); | |
503975b9 | 10511 | s = xmalloc(size); |
bc0b9d34 PC |
10512 | *space_tail = s; |
10513 | space_tail = s; | |
10514 | *space_tail = NULL; | |
10515 | break; | |
10516 | } | |
10517 | } | |
10518 | if (update->space_list == NULL) | |
10519 | break; | |
10520 | ||
10521 | /* add space for disk in update | |
10522 | */ | |
10523 | size = sizeof(struct dl); | |
503975b9 | 10524 | s = xmalloc(size); |
bc0b9d34 PC |
10525 | *space_tail = s; |
10526 | space_tail = s; | |
10527 | *space_tail = NULL; | |
10528 | ||
10529 | /* add spare device to update | |
10530 | */ | |
10531 | for (id = super->devlist ; id; id = id->next) | |
10532 | if (id->index == (unsigned)u->subdev) { | |
10533 | struct imsm_dev *dev; | |
10534 | struct imsm_map *map; | |
10535 | ||
10536 | dev = get_imsm_dev(super, u->subdev); | |
238c0a71 | 10537 | map = get_imsm_map(dev, MAP_0); |
bc0b9d34 PC |
10538 | current_level = map->raid_level; |
10539 | break; | |
10540 | } | |
089f9d79 | 10541 | if (u->new_level == 5 && u->new_level != current_level) { |
bc0b9d34 PC |
10542 | struct mdinfo *spares; |
10543 | ||
10544 | spares = get_spares_for_grow(st); | |
10545 | if (spares) { | |
10546 | struct dl *dl; | |
10547 | struct mdinfo *dev; | |
10548 | ||
10549 | dev = spares->devs; | |
10550 | if (dev) { | |
10551 | u->new_disks[0] = | |
10552 | makedev(dev->disk.major, | |
10553 | dev->disk.minor); | |
10554 | dl = get_disk_super(super, | |
10555 | dev->disk.major, | |
10556 | dev->disk.minor); | |
10557 | dl->index = u->old_raid_disks; | |
10558 | dev = dev->next; | |
10559 | } | |
10560 | sysfs_free(spares); | |
10561 | } | |
10562 | } | |
10563 | len = disks_to_mpb_size(u->new_raid_disks); | |
10564 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
48c5303a PC |
10565 | break; |
10566 | } | |
f3871fdc | 10567 | case update_size_change: { |
095b8088 N |
10568 | if (update->len < (int)sizeof(struct imsm_update_size_change)) |
10569 | return 0; | |
10570 | break; | |
10571 | } | |
10572 | case update_activate_spare: { | |
10573 | if (update->len < (int)sizeof(struct imsm_update_activate_spare)) | |
10574 | return 0; | |
f3871fdc AK |
10575 | break; |
10576 | } | |
949c47a0 DW |
10577 | case update_create_array: { |
10578 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 10579 | struct intel_dev *dv; |
54c2c1ea | 10580 | struct imsm_dev *dev = &u->dev; |
238c0a71 | 10581 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
54c2c1ea DW |
10582 | struct dl *dl; |
10583 | struct disk_info *inf; | |
10584 | int i; | |
10585 | int activate = 0; | |
949c47a0 | 10586 | |
095b8088 N |
10587 | if (update->len < (int)sizeof(*u)) |
10588 | return 0; | |
10589 | ||
54c2c1ea DW |
10590 | inf = get_disk_info(u); |
10591 | len = sizeof_imsm_dev(dev, 1); | |
ba2de7ba | 10592 | /* allocate a new super->devlist entry */ |
503975b9 N |
10593 | dv = xmalloc(sizeof(*dv)); |
10594 | dv->dev = xmalloc(len); | |
10595 | update->space = dv; | |
949c47a0 | 10596 | |
54c2c1ea DW |
10597 | /* count how many spares will be converted to members */ |
10598 | for (i = 0; i < map->num_members; i++) { | |
10599 | dl = serial_to_dl(inf[i].serial, super); | |
10600 | if (!dl) { | |
10601 | /* hmm maybe it failed?, nothing we can do about | |
10602 | * it here | |
10603 | */ | |
10604 | continue; | |
10605 | } | |
10606 | if (count_memberships(dl, super) == 0) | |
10607 | activate++; | |
10608 | } | |
10609 | len += activate * sizeof(struct imsm_disk); | |
949c47a0 | 10610 | break; |
095b8088 N |
10611 | } |
10612 | case update_kill_array: { | |
10613 | if (update->len < (int)sizeof(struct imsm_update_kill_array)) | |
10614 | return 0; | |
949c47a0 DW |
10615 | break; |
10616 | } | |
095b8088 N |
10617 | case update_rename_array: { |
10618 | if (update->len < (int)sizeof(struct imsm_update_rename_array)) | |
10619 | return 0; | |
10620 | break; | |
10621 | } | |
10622 | case update_add_remove_disk: | |
10623 | /* no update->len needed */ | |
10624 | break; | |
bbab0940 TM |
10625 | case update_prealloc_badblocks_mem: |
10626 | super->extra_space += sizeof(struct bbm_log) - | |
10627 | get_imsm_bbm_log_size(super->bbm_log); | |
10628 | break; | |
e6e9dd3f AP |
10629 | case update_rwh_policy: { |
10630 | if (update->len < (int)sizeof(struct imsm_update_rwh_policy)) | |
10631 | return 0; | |
10632 | break; | |
10633 | } | |
095b8088 N |
10634 | default: |
10635 | return 0; | |
949c47a0 | 10636 | } |
8273f55e | 10637 | |
4d7b1503 DW |
10638 | /* check if we need a larger metadata buffer */ |
10639 | if (super->next_buf) | |
10640 | buf_len = super->next_len; | |
10641 | else | |
10642 | buf_len = super->len; | |
10643 | ||
bbab0940 | 10644 | if (__le32_to_cpu(mpb->mpb_size) + super->extra_space + len > buf_len) { |
4d7b1503 DW |
10645 | /* ok we need a larger buf than what is currently allocated |
10646 | * if this allocation fails process_update will notice that | |
10647 | * ->next_len is set and ->next_buf is NULL | |
10648 | */ | |
bbab0940 TM |
10649 | buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + |
10650 | super->extra_space + len, sector_size); | |
4d7b1503 DW |
10651 | if (super->next_buf) |
10652 | free(super->next_buf); | |
10653 | ||
10654 | super->next_len = buf_len; | |
f36a9ecd | 10655 | if (posix_memalign(&super->next_buf, sector_size, buf_len) == 0) |
1f45a8ad DW |
10656 | memset(super->next_buf, 0, buf_len); |
10657 | else | |
4d7b1503 DW |
10658 | super->next_buf = NULL; |
10659 | } | |
5fe6f031 | 10660 | return 1; |
8273f55e DW |
10661 | } |
10662 | ||
ae6aad82 | 10663 | /* must be called while manager is quiesced */ |
f21e18ca | 10664 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index) |
ae6aad82 DW |
10665 | { |
10666 | struct imsm_super *mpb = super->anchor; | |
ae6aad82 DW |
10667 | struct dl *iter; |
10668 | struct imsm_dev *dev; | |
10669 | struct imsm_map *map; | |
4c9e8c1e | 10670 | unsigned int i, j, num_members; |
fb12a745 | 10671 | __u32 ord, ord_map0; |
4c9e8c1e | 10672 | struct bbm_log *log = super->bbm_log; |
ae6aad82 | 10673 | |
1ade5cc1 | 10674 | dprintf("deleting device[%d] from imsm_super\n", index); |
ae6aad82 DW |
10675 | |
10676 | /* shift all indexes down one */ | |
10677 | for (iter = super->disks; iter; iter = iter->next) | |
f21e18ca | 10678 | if (iter->index > (int)index) |
ae6aad82 | 10679 | iter->index--; |
47ee5a45 | 10680 | for (iter = super->missing; iter; iter = iter->next) |
f21e18ca | 10681 | if (iter->index > (int)index) |
47ee5a45 | 10682 | iter->index--; |
ae6aad82 DW |
10683 | |
10684 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
10685 | dev = get_imsm_dev(super, i); | |
238c0a71 | 10686 | map = get_imsm_map(dev, MAP_0); |
24565c9a DW |
10687 | num_members = map->num_members; |
10688 | for (j = 0; j < num_members; j++) { | |
10689 | /* update ord entries being careful not to propagate | |
10690 | * ord-flags to the first map | |
10691 | */ | |
238c0a71 | 10692 | ord = get_imsm_ord_tbl_ent(dev, j, MAP_X); |
fb12a745 | 10693 | ord_map0 = get_imsm_ord_tbl_ent(dev, j, MAP_0); |
ae6aad82 | 10694 | |
24565c9a DW |
10695 | if (ord_to_idx(ord) <= index) |
10696 | continue; | |
ae6aad82 | 10697 | |
238c0a71 | 10698 | map = get_imsm_map(dev, MAP_0); |
fb12a745 | 10699 | set_imsm_ord_tbl_ent(map, j, ord_map0 - 1); |
238c0a71 | 10700 | map = get_imsm_map(dev, MAP_1); |
24565c9a DW |
10701 | if (map) |
10702 | set_imsm_ord_tbl_ent(map, j, ord - 1); | |
ae6aad82 DW |
10703 | } |
10704 | } | |
10705 | ||
4c9e8c1e TM |
10706 | for (i = 0; i < log->entry_count; i++) { |
10707 | struct bbm_log_entry *entry = &log->marked_block_entries[i]; | |
10708 | ||
10709 | if (entry->disk_ordinal <= index) | |
10710 | continue; | |
10711 | entry->disk_ordinal--; | |
10712 | } | |
10713 | ||
ae6aad82 DW |
10714 | mpb->num_disks--; |
10715 | super->updates_pending++; | |
24565c9a DW |
10716 | if (*dlp) { |
10717 | struct dl *dl = *dlp; | |
10718 | ||
10719 | *dlp = (*dlp)->next; | |
3a85bf0e | 10720 | __free_imsm_disk(dl, 1); |
24565c9a | 10721 | } |
ae6aad82 | 10722 | } |
9a717282 | 10723 | |
9a717282 AK |
10724 | static int imsm_get_allowed_degradation(int level, int raid_disks, |
10725 | struct intel_super *super, | |
10726 | struct imsm_dev *dev) | |
10727 | { | |
10728 | switch (level) { | |
bf5cf7c7 | 10729 | case 1: |
9a717282 AK |
10730 | case 10:{ |
10731 | int ret_val = 0; | |
10732 | struct imsm_map *map; | |
10733 | int i; | |
10734 | ||
10735 | ret_val = raid_disks/2; | |
10736 | /* check map if all disks pairs not failed | |
10737 | * in both maps | |
10738 | */ | |
238c0a71 | 10739 | map = get_imsm_map(dev, MAP_0); |
9a717282 AK |
10740 | for (i = 0; i < ret_val; i++) { |
10741 | int degradation = 0; | |
10742 | if (get_imsm_disk(super, i) == NULL) | |
10743 | degradation++; | |
10744 | if (get_imsm_disk(super, i + 1) == NULL) | |
10745 | degradation++; | |
10746 | if (degradation == 2) | |
10747 | return 0; | |
10748 | } | |
238c0a71 | 10749 | map = get_imsm_map(dev, MAP_1); |
9a717282 AK |
10750 | /* if there is no second map |
10751 | * result can be returned | |
10752 | */ | |
10753 | if (map == NULL) | |
10754 | return ret_val; | |
10755 | /* check degradation in second map | |
10756 | */ | |
10757 | for (i = 0; i < ret_val; i++) { | |
10758 | int degradation = 0; | |
10759 | if (get_imsm_disk(super, i) == NULL) | |
10760 | degradation++; | |
10761 | if (get_imsm_disk(super, i + 1) == NULL) | |
10762 | degradation++; | |
10763 | if (degradation == 2) | |
10764 | return 0; | |
10765 | } | |
10766 | return ret_val; | |
10767 | } | |
10768 | case 5: | |
10769 | return 1; | |
10770 | case 6: | |
10771 | return 2; | |
10772 | default: | |
10773 | return 0; | |
10774 | } | |
10775 | } | |
10776 | ||
d31ad643 PB |
10777 | /******************************************************************************* |
10778 | * Function: validate_container_imsm | |
10779 | * Description: This routine validates container after assemble, | |
10780 | * eg. if devices in container are under the same controller. | |
10781 | * | |
10782 | * Parameters: | |
10783 | * info : linked list with info about devices used in array | |
10784 | * Returns: | |
10785 | * 1 : HBA mismatch | |
10786 | * 0 : Success | |
10787 | ******************************************************************************/ | |
10788 | int validate_container_imsm(struct mdinfo *info) | |
10789 | { | |
420dafcd | 10790 | if (check_no_platform()) |
6b781d33 | 10791 | return 0; |
d31ad643 | 10792 | |
6b781d33 AP |
10793 | struct sys_dev *idev; |
10794 | struct sys_dev *hba = NULL; | |
10795 | struct sys_dev *intel_devices = find_intel_devices(); | |
10796 | char *dev_path = devt_to_devpath(makedev(info->disk.major, | |
7c798f87 | 10797 | info->disk.minor), 1, NULL); |
6b781d33 AP |
10798 | |
10799 | for (idev = intel_devices; idev; idev = idev->next) { | |
10800 | if (dev_path && strstr(dev_path, idev->path)) { | |
10801 | hba = idev; | |
10802 | break; | |
d31ad643 | 10803 | } |
6b781d33 AP |
10804 | } |
10805 | if (dev_path) | |
d31ad643 PB |
10806 | free(dev_path); |
10807 | ||
6b781d33 AP |
10808 | if (!hba) { |
10809 | pr_err("WARNING - Cannot detect HBA for device %s!\n", | |
10810 | devid2kname(makedev(info->disk.major, info->disk.minor))); | |
10811 | return 1; | |
10812 | } | |
10813 | ||
10814 | const struct imsm_orom *orom = get_orom_by_device_id(hba->dev_id); | |
10815 | struct mdinfo *dev; | |
10816 | ||
10817 | for (dev = info->next; dev; dev = dev->next) { | |
7c798f87 MT |
10818 | dev_path = devt_to_devpath(makedev(dev->disk.major, |
10819 | dev->disk.minor), 1, NULL); | |
6b781d33 AP |
10820 | |
10821 | struct sys_dev *hba2 = NULL; | |
10822 | for (idev = intel_devices; idev; idev = idev->next) { | |
10823 | if (dev_path && strstr(dev_path, idev->path)) { | |
10824 | hba2 = idev; | |
10825 | break; | |
d31ad643 PB |
10826 | } |
10827 | } | |
6b781d33 AP |
10828 | if (dev_path) |
10829 | free(dev_path); | |
10830 | ||
10831 | const struct imsm_orom *orom2 = hba2 == NULL ? NULL : | |
10832 | get_orom_by_device_id(hba2->dev_id); | |
10833 | ||
10834 | if (hba2 && hba->type != hba2->type) { | |
10835 | pr_err("WARNING - HBAs of devices do not match %s != %s\n", | |
10836 | get_sys_dev_type(hba->type), get_sys_dev_type(hba2->type)); | |
10837 | return 1; | |
10838 | } | |
10839 | ||
07cb1e57 | 10840 | if (orom != orom2) { |
6b781d33 AP |
10841 | pr_err("WARNING - IMSM container assembled with disks under different HBAs!\n" |
10842 | " This operation is not supported and can lead to data loss.\n"); | |
10843 | return 1; | |
10844 | } | |
10845 | ||
10846 | if (!orom) { | |
10847 | pr_err("WARNING - IMSM container assembled with disks under HBAs without IMSM platform support!\n" | |
10848 | " This operation is not supported and can lead to data loss.\n"); | |
10849 | return 1; | |
10850 | } | |
d31ad643 | 10851 | } |
6b781d33 | 10852 | |
d31ad643 PB |
10853 | return 0; |
10854 | } | |
32141c17 | 10855 | |
6f50473f TM |
10856 | /******************************************************************************* |
10857 | * Function: imsm_record_badblock | |
10858 | * Description: This routine stores new bad block record in BBM log | |
10859 | * | |
10860 | * Parameters: | |
10861 | * a : array containing a bad block | |
10862 | * slot : disk number containing a bad block | |
10863 | * sector : bad block sector | |
10864 | * length : bad block sectors range | |
10865 | * Returns: | |
10866 | * 1 : Success | |
10867 | * 0 : Error | |
10868 | ******************************************************************************/ | |
10869 | static int imsm_record_badblock(struct active_array *a, int slot, | |
10870 | unsigned long long sector, int length) | |
10871 | { | |
10872 | struct intel_super *super = a->container->sb; | |
10873 | int ord; | |
10874 | int ret; | |
10875 | ||
10876 | ord = imsm_disk_slot_to_ord(a, slot); | |
10877 | if (ord < 0) | |
10878 | return 0; | |
10879 | ||
10880 | ret = record_new_badblock(super->bbm_log, ord_to_idx(ord), sector, | |
10881 | length); | |
10882 | if (ret) | |
10883 | super->updates_pending++; | |
10884 | ||
10885 | return ret; | |
10886 | } | |
c07a5a4f TM |
10887 | /******************************************************************************* |
10888 | * Function: imsm_clear_badblock | |
10889 | * Description: This routine clears bad block record from BBM log | |
10890 | * | |
10891 | * Parameters: | |
10892 | * a : array containing a bad block | |
10893 | * slot : disk number containing a bad block | |
10894 | * sector : bad block sector | |
10895 | * length : bad block sectors range | |
10896 | * Returns: | |
10897 | * 1 : Success | |
10898 | * 0 : Error | |
10899 | ******************************************************************************/ | |
10900 | static int imsm_clear_badblock(struct active_array *a, int slot, | |
10901 | unsigned long long sector, int length) | |
10902 | { | |
10903 | struct intel_super *super = a->container->sb; | |
10904 | int ord; | |
10905 | int ret; | |
10906 | ||
10907 | ord = imsm_disk_slot_to_ord(a, slot); | |
10908 | if (ord < 0) | |
10909 | return 0; | |
10910 | ||
10911 | ret = clear_badblock(super->bbm_log, ord_to_idx(ord), sector, length); | |
10912 | if (ret) | |
10913 | super->updates_pending++; | |
10914 | ||
10915 | return ret; | |
10916 | } | |
928f1424 TM |
10917 | /******************************************************************************* |
10918 | * Function: imsm_get_badblocks | |
10919 | * Description: This routine get list of bad blocks for an array | |
10920 | * | |
10921 | * Parameters: | |
10922 | * a : array | |
10923 | * slot : disk number | |
10924 | * Returns: | |
10925 | * bb : structure containing bad blocks | |
10926 | * NULL : error | |
10927 | ******************************************************************************/ | |
10928 | static struct md_bb *imsm_get_badblocks(struct active_array *a, int slot) | |
10929 | { | |
10930 | int inst = a->info.container_member; | |
10931 | struct intel_super *super = a->container->sb; | |
10932 | struct imsm_dev *dev = get_imsm_dev(super, inst); | |
10933 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
10934 | int ord; | |
10935 | ||
10936 | ord = imsm_disk_slot_to_ord(a, slot); | |
10937 | if (ord < 0) | |
10938 | return NULL; | |
10939 | ||
10940 | get_volume_badblocks(super->bbm_log, ord_to_idx(ord), pba_of_lba0(map), | |
44490938 | 10941 | per_dev_array_size(map), &super->bb); |
928f1424 TM |
10942 | |
10943 | return &super->bb; | |
10944 | } | |
27156a57 TM |
10945 | /******************************************************************************* |
10946 | * Function: examine_badblocks_imsm | |
10947 | * Description: Prints list of bad blocks on a disk to the standard output | |
10948 | * | |
10949 | * Parameters: | |
10950 | * st : metadata handler | |
10951 | * fd : open file descriptor for device | |
10952 | * devname : device name | |
10953 | * Returns: | |
10954 | * 0 : Success | |
10955 | * 1 : Error | |
10956 | ******************************************************************************/ | |
10957 | static int examine_badblocks_imsm(struct supertype *st, int fd, char *devname) | |
10958 | { | |
10959 | struct intel_super *super = st->sb; | |
10960 | struct bbm_log *log = super->bbm_log; | |
10961 | struct dl *d = NULL; | |
10962 | int any = 0; | |
10963 | ||
10964 | for (d = super->disks; d ; d = d->next) { | |
10965 | if (strcmp(d->devname, devname) == 0) | |
10966 | break; | |
10967 | } | |
10968 | ||
10969 | if ((d == NULL) || (d->index < 0)) { /* serial mismatch probably */ | |
10970 | pr_err("%s doesn't appear to be part of a raid array\n", | |
10971 | devname); | |
10972 | return 1; | |
10973 | } | |
10974 | ||
10975 | if (log != NULL) { | |
10976 | unsigned int i; | |
10977 | struct bbm_log_entry *entry = &log->marked_block_entries[0]; | |
10978 | ||
10979 | for (i = 0; i < log->entry_count; i++) { | |
10980 | if (entry[i].disk_ordinal == d->index) { | |
10981 | unsigned long long sector = __le48_to_cpu( | |
10982 | &entry[i].defective_block_start); | |
10983 | int cnt = entry[i].marked_count + 1; | |
10984 | ||
10985 | if (!any) { | |
10986 | printf("Bad-blocks on %s:\n", devname); | |
10987 | any = 1; | |
10988 | } | |
10989 | ||
10990 | printf("%20llu for %d sectors\n", sector, cnt); | |
10991 | } | |
10992 | } | |
10993 | } | |
10994 | ||
10995 | if (!any) | |
10996 | printf("No bad-blocks list configured on %s\n", devname); | |
10997 | ||
10998 | return 0; | |
10999 | } | |
687629c2 AK |
11000 | /******************************************************************************* |
11001 | * Function: init_migr_record_imsm | |
11002 | * Description: Function inits imsm migration record | |
11003 | * Parameters: | |
11004 | * super : imsm internal array info | |
11005 | * dev : device under migration | |
11006 | * info : general array info to find the smallest device | |
11007 | * Returns: | |
11008 | * none | |
11009 | ******************************************************************************/ | |
11010 | void init_migr_record_imsm(struct supertype *st, struct imsm_dev *dev, | |
11011 | struct mdinfo *info) | |
11012 | { | |
11013 | struct intel_super *super = st->sb; | |
11014 | struct migr_record *migr_rec = super->migr_rec; | |
11015 | int new_data_disks; | |
11016 | unsigned long long dsize, dev_sectors; | |
11017 | long long unsigned min_dev_sectors = -1LLU; | |
238c0a71 AK |
11018 | struct imsm_map *map_dest = get_imsm_map(dev, MAP_0); |
11019 | struct imsm_map *map_src = get_imsm_map(dev, MAP_1); | |
687629c2 | 11020 | unsigned long long num_migr_units; |
3ef4403c | 11021 | unsigned long long array_blocks; |
2f86fda3 | 11022 | struct dl *dl_disk = NULL; |
687629c2 AK |
11023 | |
11024 | memset(migr_rec, 0, sizeof(struct migr_record)); | |
11025 | migr_rec->family_num = __cpu_to_le32(super->anchor->family_num); | |
11026 | ||
11027 | /* only ascending reshape supported now */ | |
11028 | migr_rec->ascending_migr = __cpu_to_le32(1); | |
11029 | ||
11030 | migr_rec->dest_depth_per_unit = GEN_MIGR_AREA_SIZE / | |
11031 | max(map_dest->blocks_per_strip, map_src->blocks_per_strip); | |
e1742195 AK |
11032 | migr_rec->dest_depth_per_unit *= |
11033 | max(map_dest->blocks_per_strip, map_src->blocks_per_strip); | |
9529d343 | 11034 | new_data_disks = imsm_num_data_members(map_dest); |
687629c2 AK |
11035 | migr_rec->blocks_per_unit = |
11036 | __cpu_to_le32(migr_rec->dest_depth_per_unit * new_data_disks); | |
11037 | migr_rec->dest_depth_per_unit = | |
11038 | __cpu_to_le32(migr_rec->dest_depth_per_unit); | |
3ef4403c | 11039 | array_blocks = info->component_size * new_data_disks; |
687629c2 AK |
11040 | num_migr_units = |
11041 | array_blocks / __le32_to_cpu(migr_rec->blocks_per_unit); | |
11042 | ||
11043 | if (array_blocks % __le32_to_cpu(migr_rec->blocks_per_unit)) | |
11044 | num_migr_units++; | |
9f421827 | 11045 | set_num_migr_units(migr_rec, num_migr_units); |
687629c2 AK |
11046 | |
11047 | migr_rec->post_migr_vol_cap = dev->size_low; | |
11048 | migr_rec->post_migr_vol_cap_hi = dev->size_high; | |
11049 | ||
687629c2 | 11050 | /* Find the smallest dev */ |
2f86fda3 MT |
11051 | for (dl_disk = super->disks; dl_disk ; dl_disk = dl_disk->next) { |
11052 | /* ignore spares in container */ | |
11053 | if (dl_disk->index < 0) | |
687629c2 | 11054 | continue; |
2f86fda3 | 11055 | get_dev_size(dl_disk->fd, NULL, &dsize); |
687629c2 AK |
11056 | dev_sectors = dsize / 512; |
11057 | if (dev_sectors < min_dev_sectors) | |
11058 | min_dev_sectors = dev_sectors; | |
687629c2 | 11059 | } |
9f421827 | 11060 | set_migr_chkp_area_pba(migr_rec, min_dev_sectors - |
687629c2 AK |
11061 | RAID_DISK_RESERVED_BLOCKS_IMSM_HI); |
11062 | ||
11063 | write_imsm_migr_rec(st); | |
11064 | ||
11065 | return; | |
11066 | } | |
11067 | ||
11068 | /******************************************************************************* | |
11069 | * Function: save_backup_imsm | |
11070 | * Description: Function saves critical data stripes to Migration Copy Area | |
11071 | * and updates the current migration unit status. | |
11072 | * Use restore_stripes() to form a destination stripe, | |
11073 | * and to write it to the Copy Area. | |
11074 | * Parameters: | |
11075 | * st : supertype information | |
aea93171 | 11076 | * dev : imsm device that backup is saved for |
687629c2 AK |
11077 | * info : general array info |
11078 | * buf : input buffer | |
687629c2 AK |
11079 | * length : length of data to backup (blocks_per_unit) |
11080 | * Returns: | |
11081 | * 0 : success | |
11082 | *, -1 : fail | |
11083 | ******************************************************************************/ | |
11084 | int save_backup_imsm(struct supertype *st, | |
11085 | struct imsm_dev *dev, | |
11086 | struct mdinfo *info, | |
11087 | void *buf, | |
687629c2 AK |
11088 | int length) |
11089 | { | |
11090 | int rv = -1; | |
11091 | struct intel_super *super = st->sb; | |
687629c2 | 11092 | int i; |
238c0a71 | 11093 | struct imsm_map *map_dest = get_imsm_map(dev, MAP_0); |
687629c2 | 11094 | int new_disks = map_dest->num_members; |
ab724b98 | 11095 | int dest_layout = 0; |
4389ce73 MT |
11096 | int dest_chunk, targets[new_disks]; |
11097 | unsigned long long start, target_offsets[new_disks]; | |
9529d343 | 11098 | int data_disks = imsm_num_data_members(map_dest); |
687629c2 | 11099 | |
2f86fda3 MT |
11100 | for (i = 0; i < new_disks; i++) { |
11101 | struct dl *dl_disk = get_imsm_dl_disk(super, i); | |
4389ce73 MT |
11102 | if (dl_disk && is_fd_valid(dl_disk->fd)) |
11103 | targets[i] = dl_disk->fd; | |
11104 | else | |
11105 | goto abort; | |
2f86fda3 | 11106 | } |
7e45b550 | 11107 | |
d1877f69 | 11108 | start = info->reshape_progress * 512; |
687629c2 | 11109 | for (i = 0; i < new_disks; i++) { |
9f421827 | 11110 | target_offsets[i] = migr_chkp_area_pba(super->migr_rec) * 512; |
d1877f69 AK |
11111 | /* move back copy area adderss, it will be moved forward |
11112 | * in restore_stripes() using start input variable | |
11113 | */ | |
11114 | target_offsets[i] -= start/data_disks; | |
687629c2 AK |
11115 | } |
11116 | ||
68eb8bc6 | 11117 | dest_layout = imsm_level_to_layout(map_dest->raid_level); |
ab724b98 AK |
11118 | dest_chunk = __le16_to_cpu(map_dest->blocks_per_strip) * 512; |
11119 | ||
687629c2 AK |
11120 | if (restore_stripes(targets, /* list of dest devices */ |
11121 | target_offsets, /* migration record offsets */ | |
11122 | new_disks, | |
ab724b98 AK |
11123 | dest_chunk, |
11124 | map_dest->raid_level, | |
11125 | dest_layout, | |
11126 | -1, /* source backup file descriptor */ | |
11127 | 0, /* input buf offset | |
11128 | * always 0 buf is already offseted */ | |
d1877f69 | 11129 | start, |
687629c2 AK |
11130 | length, |
11131 | buf) != 0) { | |
e7b84f9d | 11132 | pr_err("Error restoring stripes\n"); |
687629c2 AK |
11133 | goto abort; |
11134 | } | |
11135 | ||
11136 | rv = 0; | |
11137 | ||
11138 | abort: | |
687629c2 AK |
11139 | return rv; |
11140 | } | |
11141 | ||
11142 | /******************************************************************************* | |
11143 | * Function: save_checkpoint_imsm | |
11144 | * Description: Function called for current unit status update | |
11145 | * in the migration record. It writes it to disk. | |
11146 | * Parameters: | |
11147 | * super : imsm internal array info | |
11148 | * info : general array info | |
11149 | * Returns: | |
11150 | * 0: success | |
11151 | * 1: failure | |
0228d92c AK |
11152 | * 2: failure, means no valid migration record |
11153 | * / no general migration in progress / | |
687629c2 AK |
11154 | ******************************************************************************/ |
11155 | int save_checkpoint_imsm(struct supertype *st, struct mdinfo *info, int state) | |
11156 | { | |
11157 | struct intel_super *super = st->sb; | |
f8b72ef5 AK |
11158 | unsigned long long blocks_per_unit; |
11159 | unsigned long long curr_migr_unit; | |
11160 | ||
2f86fda3 | 11161 | if (load_imsm_migr_rec(super) != 0) { |
7a862a02 | 11162 | dprintf("imsm: ERROR: Cannot read migration record for checkpoint save.\n"); |
2e062e82 AK |
11163 | return 1; |
11164 | } | |
11165 | ||
f8b72ef5 AK |
11166 | blocks_per_unit = __le32_to_cpu(super->migr_rec->blocks_per_unit); |
11167 | if (blocks_per_unit == 0) { | |
0228d92c AK |
11168 | dprintf("imsm: no migration in progress.\n"); |
11169 | return 2; | |
687629c2 | 11170 | } |
f8b72ef5 AK |
11171 | curr_migr_unit = info->reshape_progress / blocks_per_unit; |
11172 | /* check if array is alligned to copy area | |
11173 | * if it is not alligned, add one to current migration unit value | |
11174 | * this can happend on array reshape finish only | |
11175 | */ | |
11176 | if (info->reshape_progress % blocks_per_unit) | |
11177 | curr_migr_unit++; | |
687629c2 | 11178 | |
9f421827 | 11179 | set_current_migr_unit(super->migr_rec, curr_migr_unit); |
687629c2 | 11180 | super->migr_rec->rec_status = __cpu_to_le32(state); |
9f421827 PB |
11181 | set_migr_dest_1st_member_lba(super->migr_rec, |
11182 | super->migr_rec->dest_depth_per_unit * curr_migr_unit); | |
11183 | ||
687629c2 | 11184 | if (write_imsm_migr_rec(st) < 0) { |
7a862a02 | 11185 | dprintf("imsm: Cannot write migration record outside backup area\n"); |
687629c2 AK |
11186 | return 1; |
11187 | } | |
11188 | ||
11189 | return 0; | |
11190 | } | |
11191 | ||
276d77db AK |
11192 | /******************************************************************************* |
11193 | * Function: recover_backup_imsm | |
11194 | * Description: Function recovers critical data from the Migration Copy Area | |
11195 | * while assembling an array. | |
11196 | * Parameters: | |
11197 | * super : imsm internal array info | |
11198 | * info : general array info | |
11199 | * Returns: | |
11200 | * 0 : success (or there is no data to recover) | |
11201 | * 1 : fail | |
11202 | ******************************************************************************/ | |
11203 | int recover_backup_imsm(struct supertype *st, struct mdinfo *info) | |
11204 | { | |
11205 | struct intel_super *super = st->sb; | |
11206 | struct migr_record *migr_rec = super->migr_rec; | |
594dc1b8 | 11207 | struct imsm_map *map_dest; |
276d77db AK |
11208 | struct intel_dev *id = NULL; |
11209 | unsigned long long read_offset; | |
11210 | unsigned long long write_offset; | |
11211 | unsigned unit_len; | |
2f86fda3 | 11212 | int new_disks, err; |
276d77db AK |
11213 | char *buf = NULL; |
11214 | int retval = 1; | |
f36a9ecd | 11215 | unsigned int sector_size = super->sector_size; |
4036e7ee MT |
11216 | unsigned long long curr_migr_unit = current_migr_unit(migr_rec); |
11217 | unsigned long long num_migr_units = get_num_migr_units(migr_rec); | |
276d77db | 11218 | char buffer[20]; |
6c3560c0 | 11219 | int skipped_disks = 0; |
2f86fda3 | 11220 | struct dl *dl_disk; |
276d77db AK |
11221 | |
11222 | err = sysfs_get_str(info, NULL, "array_state", (char *)buffer, 20); | |
11223 | if (err < 1) | |
11224 | return 1; | |
11225 | ||
11226 | /* recover data only during assemblation */ | |
11227 | if (strncmp(buffer, "inactive", 8) != 0) | |
11228 | return 0; | |
11229 | /* no data to recover */ | |
11230 | if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL) | |
11231 | return 0; | |
11232 | if (curr_migr_unit >= num_migr_units) | |
11233 | return 1; | |
11234 | ||
11235 | /* find device during reshape */ | |
11236 | for (id = super->devlist; id; id = id->next) | |
11237 | if (is_gen_migration(id->dev)) | |
11238 | break; | |
11239 | if (id == NULL) | |
11240 | return 1; | |
11241 | ||
238c0a71 | 11242 | map_dest = get_imsm_map(id->dev, MAP_0); |
276d77db AK |
11243 | new_disks = map_dest->num_members; |
11244 | ||
9f421827 | 11245 | read_offset = migr_chkp_area_pba(migr_rec) * 512; |
276d77db | 11246 | |
9f421827 | 11247 | write_offset = (migr_dest_1st_member_lba(migr_rec) + |
5551b113 | 11248 | pba_of_lba0(map_dest)) * 512; |
276d77db AK |
11249 | |
11250 | unit_len = __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
f36a9ecd | 11251 | if (posix_memalign((void **)&buf, sector_size, unit_len) != 0) |
276d77db | 11252 | goto abort; |
276d77db | 11253 | |
2f86fda3 MT |
11254 | for (dl_disk = super->disks; dl_disk; dl_disk = dl_disk->next) { |
11255 | if (dl_disk->index < 0) | |
11256 | continue; | |
276d77db | 11257 | |
4389ce73 | 11258 | if (!is_fd_valid(dl_disk->fd)) { |
6c3560c0 AK |
11259 | skipped_disks++; |
11260 | continue; | |
11261 | } | |
2f86fda3 | 11262 | if (lseek64(dl_disk->fd, read_offset, SEEK_SET) < 0) { |
e7b84f9d N |
11263 | pr_err("Cannot seek to block: %s\n", |
11264 | strerror(errno)); | |
137debce AK |
11265 | skipped_disks++; |
11266 | continue; | |
276d77db | 11267 | } |
83b3de77 | 11268 | if (read(dl_disk->fd, buf, unit_len) != (ssize_t)unit_len) { |
e7b84f9d N |
11269 | pr_err("Cannot read copy area block: %s\n", |
11270 | strerror(errno)); | |
137debce AK |
11271 | skipped_disks++; |
11272 | continue; | |
276d77db | 11273 | } |
2f86fda3 | 11274 | if (lseek64(dl_disk->fd, write_offset, SEEK_SET) < 0) { |
e7b84f9d N |
11275 | pr_err("Cannot seek to block: %s\n", |
11276 | strerror(errno)); | |
137debce AK |
11277 | skipped_disks++; |
11278 | continue; | |
276d77db | 11279 | } |
83b3de77 | 11280 | if (write(dl_disk->fd, buf, unit_len) != (ssize_t)unit_len) { |
e7b84f9d N |
11281 | pr_err("Cannot restore block: %s\n", |
11282 | strerror(errno)); | |
137debce AK |
11283 | skipped_disks++; |
11284 | continue; | |
276d77db AK |
11285 | } |
11286 | } | |
11287 | ||
137debce AK |
11288 | if (skipped_disks > imsm_get_allowed_degradation(info->new_level, |
11289 | new_disks, | |
11290 | super, | |
11291 | id->dev)) { | |
7a862a02 | 11292 | pr_err("Cannot restore data from backup. Too many failed disks\n"); |
6c3560c0 AK |
11293 | goto abort; |
11294 | } | |
11295 | ||
befb629b AK |
11296 | if (save_checkpoint_imsm(st, info, UNIT_SRC_NORMAL)) { |
11297 | /* ignore error == 2, this can mean end of reshape here | |
11298 | */ | |
7a862a02 | 11299 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL) during restart\n"); |
befb629b | 11300 | } else |
276d77db | 11301 | retval = 0; |
276d77db AK |
11302 | |
11303 | abort: | |
276d77db AK |
11304 | free(buf); |
11305 | return retval; | |
11306 | } | |
11307 | ||
2cda7640 ML |
11308 | static char disk_by_path[] = "/dev/disk/by-path/"; |
11309 | ||
11310 | static const char *imsm_get_disk_controller_domain(const char *path) | |
11311 | { | |
2cda7640 | 11312 | char disk_path[PATH_MAX]; |
96234762 LM |
11313 | char *drv=NULL; |
11314 | struct stat st; | |
2cda7640 | 11315 | |
dd5ab402 | 11316 | strncpy(disk_path, disk_by_path, PATH_MAX); |
96234762 LM |
11317 | strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1); |
11318 | if (stat(disk_path, &st) == 0) { | |
11319 | struct sys_dev* hba; | |
594dc1b8 | 11320 | char *path; |
96234762 | 11321 | |
7c798f87 | 11322 | path = devt_to_devpath(st.st_rdev, 1, NULL); |
96234762 LM |
11323 | if (path == NULL) |
11324 | return "unknown"; | |
11325 | hba = find_disk_attached_hba(-1, path); | |
11326 | if (hba && hba->type == SYS_DEV_SAS) | |
11327 | drv = "isci"; | |
75350d87 | 11328 | else if (hba && (hba->type == SYS_DEV_SATA || hba->type == SYS_DEV_SATA_VMD)) |
96234762 | 11329 | drv = "ahci"; |
c6839718 MT |
11330 | else if (hba && hba->type == SYS_DEV_VMD) |
11331 | drv = "vmd"; | |
11332 | else if (hba && hba->type == SYS_DEV_NVME) | |
11333 | drv = "nvme"; | |
1011e834 | 11334 | else |
96234762 LM |
11335 | drv = "unknown"; |
11336 | dprintf("path: %s hba: %s attached: %s\n", | |
11337 | path, (hba) ? hba->path : "NULL", drv); | |
11338 | free(path); | |
2cda7640 | 11339 | } |
96234762 | 11340 | return drv; |
2cda7640 ML |
11341 | } |
11342 | ||
4dd2df09 | 11343 | static char *imsm_find_array_devnm_by_subdev(int subdev, char *container) |
78b10e66 | 11344 | { |
4dd2df09 | 11345 | static char devnm[32]; |
78b10e66 N |
11346 | char subdev_name[20]; |
11347 | struct mdstat_ent *mdstat; | |
11348 | ||
11349 | sprintf(subdev_name, "%d", subdev); | |
11350 | mdstat = mdstat_by_subdev(subdev_name, container); | |
11351 | if (!mdstat) | |
4dd2df09 | 11352 | return NULL; |
78b10e66 | 11353 | |
4dd2df09 | 11354 | strcpy(devnm, mdstat->devnm); |
78b10e66 | 11355 | free_mdstat(mdstat); |
4dd2df09 | 11356 | return devnm; |
78b10e66 N |
11357 | } |
11358 | ||
11359 | static int imsm_reshape_is_allowed_on_container(struct supertype *st, | |
11360 | struct geo_params *geo, | |
fbf3d202 AK |
11361 | int *old_raid_disks, |
11362 | int direction) | |
78b10e66 | 11363 | { |
694575e7 KW |
11364 | /* currently we only support increasing the number of devices |
11365 | * for a container. This increases the number of device for each | |
11366 | * member array. They must all be RAID0 or RAID5. | |
11367 | */ | |
78b10e66 N |
11368 | int ret_val = 0; |
11369 | struct mdinfo *info, *member; | |
11370 | int devices_that_can_grow = 0; | |
11371 | ||
7a862a02 | 11372 | dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): st->devnm = (%s)\n", st->devnm); |
78b10e66 | 11373 | |
d04f65f4 | 11374 | if (geo->size > 0 || |
78b10e66 N |
11375 | geo->level != UnSet || |
11376 | geo->layout != UnSet || | |
11377 | geo->chunksize != 0 || | |
11378 | geo->raid_disks == UnSet) { | |
7a862a02 | 11379 | dprintf("imsm: Container operation is allowed for raid disks number change only.\n"); |
78b10e66 N |
11380 | return ret_val; |
11381 | } | |
11382 | ||
fbf3d202 | 11383 | if (direction == ROLLBACK_METADATA_CHANGES) { |
7a862a02 | 11384 | dprintf("imsm: Metadata changes rollback is not supported for container operation.\n"); |
fbf3d202 AK |
11385 | return ret_val; |
11386 | } | |
11387 | ||
78b10e66 N |
11388 | info = container_content_imsm(st, NULL); |
11389 | for (member = info; member; member = member->next) { | |
4dd2df09 | 11390 | char *result; |
78b10e66 N |
11391 | |
11392 | dprintf("imsm: checking device_num: %i\n", | |
11393 | member->container_member); | |
11394 | ||
d7d205bd | 11395 | if (geo->raid_disks <= member->array.raid_disks) { |
78b10e66 N |
11396 | /* we work on container for Online Capacity Expansion |
11397 | * only so raid_disks has to grow | |
11398 | */ | |
7a862a02 | 11399 | dprintf("imsm: for container operation raid disks increase is required\n"); |
78b10e66 N |
11400 | break; |
11401 | } | |
11402 | ||
089f9d79 | 11403 | if (info->array.level != 0 && info->array.level != 5) { |
78b10e66 N |
11404 | /* we cannot use this container with other raid level |
11405 | */ | |
7a862a02 | 11406 | dprintf("imsm: for container operation wrong raid level (%i) detected\n", |
78b10e66 N |
11407 | info->array.level); |
11408 | break; | |
11409 | } else { | |
11410 | /* check for platform support | |
11411 | * for this raid level configuration | |
11412 | */ | |
11413 | struct intel_super *super = st->sb; | |
11414 | if (!is_raid_level_supported(super->orom, | |
11415 | member->array.level, | |
11416 | geo->raid_disks)) { | |
7a862a02 | 11417 | dprintf("platform does not support raid%d with %d disk%s\n", |
78b10e66 N |
11418 | info->array.level, |
11419 | geo->raid_disks, | |
11420 | geo->raid_disks > 1 ? "s" : ""); | |
11421 | break; | |
11422 | } | |
2a4a08e7 AK |
11423 | /* check if component size is aligned to chunk size |
11424 | */ | |
11425 | if (info->component_size % | |
11426 | (info->array.chunk_size/512)) { | |
7a862a02 | 11427 | dprintf("Component size is not aligned to chunk size\n"); |
2a4a08e7 AK |
11428 | break; |
11429 | } | |
78b10e66 N |
11430 | } |
11431 | ||
11432 | if (*old_raid_disks && | |
11433 | info->array.raid_disks != *old_raid_disks) | |
11434 | break; | |
11435 | *old_raid_disks = info->array.raid_disks; | |
11436 | ||
11437 | /* All raid5 and raid0 volumes in container | |
11438 | * have to be ready for Online Capacity Expansion | |
11439 | * so they need to be assembled. We have already | |
11440 | * checked that no recovery etc is happening. | |
11441 | */ | |
4dd2df09 N |
11442 | result = imsm_find_array_devnm_by_subdev(member->container_member, |
11443 | st->container_devnm); | |
11444 | if (result == NULL) { | |
78b10e66 N |
11445 | dprintf("imsm: cannot find array\n"); |
11446 | break; | |
11447 | } | |
11448 | devices_that_can_grow++; | |
11449 | } | |
11450 | sysfs_free(info); | |
11451 | if (!member && devices_that_can_grow) | |
11452 | ret_val = 1; | |
11453 | ||
11454 | if (ret_val) | |
1ade5cc1 | 11455 | dprintf("Container operation allowed\n"); |
78b10e66 | 11456 | else |
1ade5cc1 | 11457 | dprintf("Error: %i\n", ret_val); |
78b10e66 N |
11458 | |
11459 | return ret_val; | |
11460 | } | |
11461 | ||
11462 | /* Function: get_spares_for_grow | |
11463 | * Description: Allocates memory and creates list of spare devices | |
1011e834 | 11464 | * avaliable in container. Checks if spare drive size is acceptable. |
78b10e66 N |
11465 | * Parameters: Pointer to the supertype structure |
11466 | * Returns: Pointer to the list of spare devices (mdinfo structure) on success, | |
1011e834 | 11467 | * NULL if fail |
78b10e66 N |
11468 | */ |
11469 | static struct mdinfo *get_spares_for_grow(struct supertype *st) | |
11470 | { | |
fbfdcb06 AO |
11471 | struct spare_criteria sc; |
11472 | ||
11473 | get_spare_criteria_imsm(st, &sc); | |
11474 | return container_choose_spares(st, &sc, NULL, NULL, NULL, 0); | |
78b10e66 N |
11475 | } |
11476 | ||
11477 | /****************************************************************************** | |
11478 | * function: imsm_create_metadata_update_for_reshape | |
11479 | * Function creates update for whole IMSM container. | |
11480 | * | |
11481 | ******************************************************************************/ | |
11482 | static int imsm_create_metadata_update_for_reshape( | |
11483 | struct supertype *st, | |
11484 | struct geo_params *geo, | |
11485 | int old_raid_disks, | |
11486 | struct imsm_update_reshape **updatep) | |
11487 | { | |
11488 | struct intel_super *super = st->sb; | |
11489 | struct imsm_super *mpb = super->anchor; | |
594dc1b8 JS |
11490 | int update_memory_size; |
11491 | struct imsm_update_reshape *u; | |
11492 | struct mdinfo *spares; | |
78b10e66 | 11493 | int i; |
594dc1b8 | 11494 | int delta_disks; |
bbd24d86 | 11495 | struct mdinfo *dev; |
78b10e66 | 11496 | |
1ade5cc1 | 11497 | dprintf("(enter) raid_disks = %i\n", geo->raid_disks); |
78b10e66 N |
11498 | |
11499 | delta_disks = geo->raid_disks - old_raid_disks; | |
11500 | ||
11501 | /* size of all update data without anchor */ | |
11502 | update_memory_size = sizeof(struct imsm_update_reshape); | |
11503 | ||
11504 | /* now add space for spare disks that we need to add. */ | |
11505 | update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1); | |
11506 | ||
503975b9 | 11507 | u = xcalloc(1, update_memory_size); |
78b10e66 N |
11508 | u->type = update_reshape_container_disks; |
11509 | u->old_raid_disks = old_raid_disks; | |
11510 | u->new_raid_disks = geo->raid_disks; | |
11511 | ||
11512 | /* now get spare disks list | |
11513 | */ | |
11514 | spares = get_spares_for_grow(st); | |
11515 | ||
d7be7d87 | 11516 | if (spares == NULL || delta_disks > spares->array.spare_disks) { |
7a862a02 | 11517 | pr_err("imsm: ERROR: Cannot get spare devices for %s.\n", geo->dev_name); |
e4c72d1d | 11518 | i = -1; |
78b10e66 N |
11519 | goto abort; |
11520 | } | |
11521 | ||
11522 | /* we have got spares | |
11523 | * update disk list in imsm_disk list table in anchor | |
11524 | */ | |
11525 | dprintf("imsm: %i spares are available.\n\n", | |
11526 | spares->array.spare_disks); | |
11527 | ||
bbd24d86 | 11528 | dev = spares->devs; |
78b10e66 | 11529 | for (i = 0; i < delta_disks; i++) { |
78b10e66 N |
11530 | struct dl *dl; |
11531 | ||
bbd24d86 AK |
11532 | if (dev == NULL) |
11533 | break; | |
78b10e66 N |
11534 | u->new_disks[i] = makedev(dev->disk.major, |
11535 | dev->disk.minor); | |
11536 | dl = get_disk_super(super, dev->disk.major, dev->disk.minor); | |
ee4beede AK |
11537 | dl->index = mpb->num_disks; |
11538 | mpb->num_disks++; | |
bbd24d86 | 11539 | dev = dev->next; |
78b10e66 | 11540 | } |
78b10e66 N |
11541 | |
11542 | abort: | |
11543 | /* free spares | |
11544 | */ | |
11545 | sysfs_free(spares); | |
11546 | ||
d677e0b8 | 11547 | dprintf("imsm: reshape update preparation :"); |
78b10e66 | 11548 | if (i == delta_disks) { |
1ade5cc1 | 11549 | dprintf_cont(" OK\n"); |
78b10e66 N |
11550 | *updatep = u; |
11551 | return update_memory_size; | |
11552 | } | |
11553 | free(u); | |
1ade5cc1 | 11554 | dprintf_cont(" Error\n"); |
78b10e66 N |
11555 | |
11556 | return 0; | |
11557 | } | |
11558 | ||
f3871fdc AK |
11559 | /****************************************************************************** |
11560 | * function: imsm_create_metadata_update_for_size_change() | |
11561 | * Creates update for IMSM array for array size change. | |
11562 | * | |
11563 | ******************************************************************************/ | |
11564 | static int imsm_create_metadata_update_for_size_change( | |
11565 | struct supertype *st, | |
11566 | struct geo_params *geo, | |
11567 | struct imsm_update_size_change **updatep) | |
11568 | { | |
11569 | struct intel_super *super = st->sb; | |
594dc1b8 JS |
11570 | int update_memory_size; |
11571 | struct imsm_update_size_change *u; | |
f3871fdc | 11572 | |
1ade5cc1 | 11573 | dprintf("(enter) New size = %llu\n", geo->size); |
f3871fdc AK |
11574 | |
11575 | /* size of all update data without anchor */ | |
11576 | update_memory_size = sizeof(struct imsm_update_size_change); | |
11577 | ||
503975b9 | 11578 | u = xcalloc(1, update_memory_size); |
f3871fdc AK |
11579 | u->type = update_size_change; |
11580 | u->subdev = super->current_vol; | |
11581 | u->new_size = geo->size; | |
11582 | ||
11583 | dprintf("imsm: reshape update preparation : OK\n"); | |
11584 | *updatep = u; | |
11585 | ||
11586 | return update_memory_size; | |
11587 | } | |
11588 | ||
48c5303a PC |
11589 | /****************************************************************************** |
11590 | * function: imsm_create_metadata_update_for_migration() | |
11591 | * Creates update for IMSM array. | |
11592 | * | |
11593 | ******************************************************************************/ | |
11594 | static int imsm_create_metadata_update_for_migration( | |
11595 | struct supertype *st, | |
11596 | struct geo_params *geo, | |
11597 | struct imsm_update_reshape_migration **updatep) | |
11598 | { | |
11599 | struct intel_super *super = st->sb; | |
594dc1b8 | 11600 | int update_memory_size; |
756a15f3 | 11601 | int current_chunk_size; |
594dc1b8 | 11602 | struct imsm_update_reshape_migration *u; |
756a15f3 MG |
11603 | struct imsm_dev *dev = get_imsm_dev(super, super->current_vol); |
11604 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
48c5303a PC |
11605 | int previous_level = -1; |
11606 | ||
1ade5cc1 | 11607 | dprintf("(enter) New Level = %i\n", geo->level); |
48c5303a PC |
11608 | |
11609 | /* size of all update data without anchor */ | |
11610 | update_memory_size = sizeof(struct imsm_update_reshape_migration); | |
11611 | ||
503975b9 | 11612 | u = xcalloc(1, update_memory_size); |
48c5303a PC |
11613 | u->type = update_reshape_migration; |
11614 | u->subdev = super->current_vol; | |
11615 | u->new_level = geo->level; | |
11616 | u->new_layout = geo->layout; | |
11617 | u->new_raid_disks = u->old_raid_disks = geo->raid_disks; | |
11618 | u->new_disks[0] = -1; | |
4bba0439 | 11619 | u->new_chunksize = -1; |
48c5303a | 11620 | |
756a15f3 | 11621 | current_chunk_size = __le16_to_cpu(map->blocks_per_strip) / 2; |
48c5303a | 11622 | |
756a15f3 MG |
11623 | if (geo->chunksize != current_chunk_size) { |
11624 | u->new_chunksize = geo->chunksize / 1024; | |
11625 | dprintf("imsm: chunk size change from %i to %i\n", | |
11626 | current_chunk_size, u->new_chunksize); | |
48c5303a | 11627 | } |
756a15f3 MG |
11628 | previous_level = map->raid_level; |
11629 | ||
089f9d79 | 11630 | if (geo->level == 5 && previous_level == 0) { |
48c5303a PC |
11631 | struct mdinfo *spares = NULL; |
11632 | ||
11633 | u->new_raid_disks++; | |
11634 | spares = get_spares_for_grow(st); | |
089f9d79 | 11635 | if (spares == NULL || spares->array.spare_disks < 1) { |
48c5303a PC |
11636 | free(u); |
11637 | sysfs_free(spares); | |
11638 | update_memory_size = 0; | |
565cc99e | 11639 | pr_err("cannot get spare device for requested migration\n"); |
48c5303a PC |
11640 | return 0; |
11641 | } | |
11642 | sysfs_free(spares); | |
11643 | } | |
11644 | dprintf("imsm: reshape update preparation : OK\n"); | |
11645 | *updatep = u; | |
11646 | ||
11647 | return update_memory_size; | |
11648 | } | |
11649 | ||
8dd70bce AK |
11650 | static void imsm_update_metadata_locally(struct supertype *st, |
11651 | void *buf, int len) | |
11652 | { | |
11653 | struct metadata_update mu; | |
11654 | ||
11655 | mu.buf = buf; | |
11656 | mu.len = len; | |
11657 | mu.space = NULL; | |
11658 | mu.space_list = NULL; | |
11659 | mu.next = NULL; | |
5fe6f031 N |
11660 | if (imsm_prepare_update(st, &mu)) |
11661 | imsm_process_update(st, &mu); | |
8dd70bce AK |
11662 | |
11663 | while (mu.space_list) { | |
11664 | void **space = mu.space_list; | |
11665 | mu.space_list = *space; | |
11666 | free(space); | |
11667 | } | |
11668 | } | |
78b10e66 | 11669 | |
cbaa7904 MT |
11670 | /** |
11671 | * imsm_analyze_expand() - check expand properties and calculate new size. | |
11672 | * @st: imsm supertype. | |
11673 | * @geo: new geometry params. | |
11674 | * @array: array info. | |
11675 | * @direction: reshape direction. | |
11676 | * | |
11677 | * Obtain free space after the &array and verify if expand to requested size is | |
11678 | * possible. If geo->size is set to %MAX_SIZE, assume that max free size is | |
11679 | * requested. | |
11680 | * | |
11681 | * Return: | |
11682 | * On success %IMSM_STATUS_OK is returned, geo->size and geo->raid_disks are | |
11683 | * updated. | |
11684 | * On error, %IMSM_STATUS_ERROR is returned. | |
11685 | */ | |
11686 | static imsm_status_t imsm_analyze_expand(struct supertype *st, | |
11687 | struct geo_params *geo, | |
11688 | struct mdinfo *array, | |
11689 | int direction) | |
11690 | { | |
11691 | struct intel_super *super = st->sb; | |
11692 | struct imsm_dev *dev = get_imsm_dev(super, super->current_vol); | |
11693 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
11694 | int data_disks = imsm_num_data_members(map); | |
11695 | ||
11696 | unsigned long long current_size; | |
11697 | unsigned long long free_size; | |
11698 | unsigned long long new_size; | |
11699 | unsigned long long max_size; | |
11700 | ||
11701 | const int chunk_kib = geo->chunksize / 1024; | |
11702 | imsm_status_t rv; | |
11703 | ||
11704 | if (direction == ROLLBACK_METADATA_CHANGES) { | |
11705 | /** | |
11706 | * Accept size for rollback only. | |
11707 | */ | |
11708 | new_size = geo->size * 2; | |
11709 | goto success; | |
11710 | } | |
11711 | ||
cbaa7904 MT |
11712 | if (data_disks == 0) { |
11713 | pr_err("imsm: Cannot retrieve data disks.\n"); | |
11714 | return IMSM_STATUS_ERROR; | |
11715 | } | |
11716 | current_size = array->custom_array_size / data_disks; | |
11717 | ||
aa19fdd4 | 11718 | rv = imsm_get_free_size(super, dev->vol.map->num_members, 0, chunk_kib, &free_size, true); |
cbaa7904 MT |
11719 | if (rv != IMSM_STATUS_OK) { |
11720 | pr_err("imsm: Cannot find free space for expand.\n"); | |
11721 | return IMSM_STATUS_ERROR; | |
11722 | } | |
11723 | max_size = round_member_size_to_mb(free_size + current_size); | |
11724 | ||
11725 | if (geo->size == MAX_SIZE) | |
11726 | new_size = max_size; | |
11727 | else | |
11728 | new_size = round_member_size_to_mb(geo->size * 2); | |
11729 | ||
11730 | if (new_size == 0) { | |
11731 | pr_err("imsm: Rounded requested size is 0.\n"); | |
11732 | return IMSM_STATUS_ERROR; | |
11733 | } | |
11734 | ||
11735 | if (new_size > max_size) { | |
11736 | pr_err("imsm: Rounded requested size (%llu) is larger than free space available (%llu).\n", | |
11737 | new_size, max_size); | |
11738 | return IMSM_STATUS_ERROR; | |
11739 | } | |
11740 | ||
11741 | if (new_size == current_size) { | |
11742 | pr_err("imsm: Rounded requested size (%llu) is same as current size (%llu).\n", | |
11743 | new_size, current_size); | |
11744 | return IMSM_STATUS_ERROR; | |
11745 | } | |
11746 | ||
11747 | if (new_size < current_size) { | |
11748 | pr_err("imsm: Size reduction is not supported, rounded requested size (%llu) is smaller than current (%llu).\n", | |
11749 | new_size, current_size); | |
11750 | return IMSM_STATUS_ERROR; | |
11751 | } | |
11752 | ||
11753 | success: | |
11754 | dprintf("imsm: New size per member is %llu.\n", new_size); | |
11755 | geo->size = data_disks * new_size; | |
11756 | geo->raid_disks = dev->vol.map->num_members; | |
11757 | return IMSM_STATUS_OK; | |
11758 | } | |
11759 | ||
471bceb6 | 11760 | /*************************************************************************** |
694575e7 | 11761 | * Function: imsm_analyze_change |
471bceb6 | 11762 | * Description: Function analyze change for single volume |
1011e834 | 11763 | * and validate if transition is supported |
fbf3d202 AK |
11764 | * Parameters: Geometry parameters, supertype structure, |
11765 | * metadata change direction (apply/rollback) | |
694575e7 | 11766 | * Returns: Operation type code on success, -1 if fail |
471bceb6 KW |
11767 | ****************************************************************************/ |
11768 | enum imsm_reshape_type imsm_analyze_change(struct supertype *st, | |
fbf3d202 AK |
11769 | struct geo_params *geo, |
11770 | int direction) | |
694575e7 | 11771 | { |
471bceb6 KW |
11772 | struct mdinfo info; |
11773 | int change = -1; | |
11774 | int check_devs = 0; | |
c21e737b | 11775 | int chunk; |
67a2db32 AK |
11776 | /* number of added/removed disks in operation result */ |
11777 | int devNumChange = 0; | |
11778 | /* imsm compatible layout value for array geometry verification */ | |
11779 | int imsm_layout = -1; | |
6d4d9ab2 | 11780 | imsm_status_t rv; |
471bceb6 KW |
11781 | |
11782 | getinfo_super_imsm_volume(st, &info, NULL); | |
089f9d79 JS |
11783 | if (geo->level != info.array.level && geo->level >= 0 && |
11784 | geo->level != UnSet) { | |
471bceb6 KW |
11785 | switch (info.array.level) { |
11786 | case 0: | |
11787 | if (geo->level == 5) { | |
b5347799 | 11788 | change = CH_MIGRATION; |
e13ce846 | 11789 | if (geo->layout != ALGORITHM_LEFT_ASYMMETRIC) { |
7a862a02 | 11790 | pr_err("Error. Requested Layout not supported (left-asymmetric layout is supported only)!\n"); |
e13ce846 AK |
11791 | change = -1; |
11792 | goto analyse_change_exit; | |
11793 | } | |
67a2db32 | 11794 | imsm_layout = geo->layout; |
471bceb6 | 11795 | check_devs = 1; |
e91a3bad LM |
11796 | devNumChange = 1; /* parity disk added */ |
11797 | } else if (geo->level == 10) { | |
471bceb6 KW |
11798 | change = CH_TAKEOVER; |
11799 | check_devs = 1; | |
e91a3bad | 11800 | devNumChange = 2; /* two mirrors added */ |
67a2db32 | 11801 | imsm_layout = 0x102; /* imsm supported layout */ |
471bceb6 | 11802 | } |
dfe77a9e KW |
11803 | break; |
11804 | case 1: | |
471bceb6 KW |
11805 | case 10: |
11806 | if (geo->level == 0) { | |
11807 | change = CH_TAKEOVER; | |
11808 | check_devs = 1; | |
e91a3bad | 11809 | devNumChange = -(geo->raid_disks/2); |
67a2db32 | 11810 | imsm_layout = 0; /* imsm raid0 layout */ |
471bceb6 KW |
11811 | } |
11812 | break; | |
11813 | } | |
11814 | if (change == -1) { | |
7a862a02 | 11815 | pr_err("Error. Level Migration from %d to %d not supported!\n", |
e7b84f9d | 11816 | info.array.level, geo->level); |
471bceb6 KW |
11817 | goto analyse_change_exit; |
11818 | } | |
11819 | } else | |
11820 | geo->level = info.array.level; | |
11821 | ||
089f9d79 JS |
11822 | if (geo->layout != info.array.layout && |
11823 | (geo->layout != UnSet && geo->layout != -1)) { | |
b5347799 | 11824 | change = CH_MIGRATION; |
089f9d79 JS |
11825 | if (info.array.layout == 0 && info.array.level == 5 && |
11826 | geo->layout == 5) { | |
471bceb6 | 11827 | /* reshape 5 -> 4 */ |
089f9d79 JS |
11828 | } else if (info.array.layout == 5 && info.array.level == 5 && |
11829 | geo->layout == 0) { | |
471bceb6 KW |
11830 | /* reshape 4 -> 5 */ |
11831 | geo->layout = 0; | |
11832 | geo->level = 5; | |
11833 | } else { | |
7a862a02 | 11834 | pr_err("Error. Layout Migration from %d to %d not supported!\n", |
e7b84f9d | 11835 | info.array.layout, geo->layout); |
471bceb6 KW |
11836 | change = -1; |
11837 | goto analyse_change_exit; | |
11838 | } | |
67a2db32 | 11839 | } else { |
471bceb6 | 11840 | geo->layout = info.array.layout; |
67a2db32 AK |
11841 | if (imsm_layout == -1) |
11842 | imsm_layout = info.array.layout; | |
11843 | } | |
471bceb6 | 11844 | |
089f9d79 JS |
11845 | if (geo->chunksize > 0 && geo->chunksize != UnSet && |
11846 | geo->chunksize != info.array.chunk_size) { | |
2d2b0eb7 MD |
11847 | if (info.array.level == 10) { |
11848 | pr_err("Error. Chunk size change for RAID 10 is not supported.\n"); | |
11849 | change = -1; | |
11850 | goto analyse_change_exit; | |
1e9b2c3f PB |
11851 | } else if (info.component_size % (geo->chunksize/512)) { |
11852 | pr_err("New chunk size (%dK) does not evenly divide device size (%lluk). Aborting...\n", | |
11853 | geo->chunksize/1024, info.component_size/2); | |
11854 | change = -1; | |
11855 | goto analyse_change_exit; | |
2d2b0eb7 | 11856 | } |
b5347799 | 11857 | change = CH_MIGRATION; |
2d2b0eb7 | 11858 | } else { |
471bceb6 | 11859 | geo->chunksize = info.array.chunk_size; |
2d2b0eb7 | 11860 | } |
471bceb6 | 11861 | |
cbaa7904 | 11862 | if (geo->size > 0) { |
7abc9871 | 11863 | if (change != -1) { |
7a862a02 | 11864 | pr_err("Error. Size change should be the only one at a time.\n"); |
7abc9871 AK |
11865 | change = -1; |
11866 | goto analyse_change_exit; | |
11867 | } | |
6d4d9ab2 | 11868 | |
cbaa7904 | 11869 | rv = imsm_analyze_expand(st, geo, &info, direction); |
6d4d9ab2 | 11870 | if (rv != IMSM_STATUS_OK) |
cbaa7904 | 11871 | goto analyse_change_exit; |
7abc9871 AK |
11872 | change = CH_ARRAY_SIZE; |
11873 | } | |
cbaa7904 MT |
11874 | |
11875 | chunk = geo->chunksize / 1024; | |
471bceb6 KW |
11876 | if (!validate_geometry_imsm(st, |
11877 | geo->level, | |
67a2db32 | 11878 | imsm_layout, |
e91a3bad | 11879 | geo->raid_disks + devNumChange, |
c21e737b | 11880 | &chunk, |
af4348dd | 11881 | geo->size, INVALID_SECTORS, |
5308f117 | 11882 | 0, 0, info.consistency_policy, 1)) |
471bceb6 KW |
11883 | change = -1; |
11884 | ||
11885 | if (check_devs) { | |
11886 | struct intel_super *super = st->sb; | |
11887 | struct imsm_super *mpb = super->anchor; | |
11888 | ||
11889 | if (mpb->num_raid_devs > 1) { | |
f1cc8ab9 LF |
11890 | pr_err("Error. Cannot perform operation on %s- for this operation " |
11891 | "it MUST be single array in container\n", geo->dev_name); | |
471bceb6 KW |
11892 | change = -1; |
11893 | } | |
11894 | } | |
11895 | ||
11896 | analyse_change_exit: | |
089f9d79 JS |
11897 | if (direction == ROLLBACK_METADATA_CHANGES && |
11898 | (change == CH_MIGRATION || change == CH_TAKEOVER)) { | |
7a862a02 | 11899 | dprintf("imsm: Metadata changes rollback is not supported for migration and takeover operations.\n"); |
fbf3d202 AK |
11900 | change = -1; |
11901 | } | |
471bceb6 | 11902 | return change; |
694575e7 KW |
11903 | } |
11904 | ||
bb025c2f KW |
11905 | int imsm_takeover(struct supertype *st, struct geo_params *geo) |
11906 | { | |
11907 | struct intel_super *super = st->sb; | |
11908 | struct imsm_update_takeover *u; | |
11909 | ||
503975b9 | 11910 | u = xmalloc(sizeof(struct imsm_update_takeover)); |
bb025c2f KW |
11911 | |
11912 | u->type = update_takeover; | |
11913 | u->subarray = super->current_vol; | |
11914 | ||
11915 | /* 10->0 transition */ | |
11916 | if (geo->level == 0) | |
11917 | u->direction = R10_TO_R0; | |
11918 | ||
0529c688 KW |
11919 | /* 0->10 transition */ |
11920 | if (geo->level == 10) | |
11921 | u->direction = R0_TO_R10; | |
11922 | ||
bb025c2f KW |
11923 | /* update metadata locally */ |
11924 | imsm_update_metadata_locally(st, u, | |
11925 | sizeof(struct imsm_update_takeover)); | |
11926 | /* and possibly remotely */ | |
11927 | if (st->update_tail) | |
11928 | append_metadata_update(st, u, | |
11929 | sizeof(struct imsm_update_takeover)); | |
11930 | else | |
11931 | free(u); | |
11932 | ||
11933 | return 0; | |
11934 | } | |
11935 | ||
895ffd99 MT |
11936 | /* Flush size update if size calculated by num_data_stripes is higher than |
11937 | * imsm_dev_size to eliminate differences during reshape. | |
11938 | * Mdmon will recalculate them correctly. | |
11939 | * If subarray index is not set then check whole container. | |
11940 | * Returns: | |
11941 | * 0 - no error occurred | |
11942 | * 1 - error detected | |
11943 | */ | |
11944 | static int imsm_fix_size_mismatch(struct supertype *st, int subarray_index) | |
11945 | { | |
11946 | struct intel_super *super = st->sb; | |
11947 | int tmp = super->current_vol; | |
11948 | int ret_val = 1; | |
11949 | int i; | |
11950 | ||
11951 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
11952 | if (subarray_index >= 0 && i != subarray_index) | |
11953 | continue; | |
11954 | super->current_vol = i; | |
11955 | struct imsm_dev *dev = get_imsm_dev(super, super->current_vol); | |
11956 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
11957 | unsigned int disc_count = imsm_num_data_members(map); | |
11958 | struct geo_params geo; | |
11959 | struct imsm_update_size_change *update; | |
11960 | unsigned long long calc_size = per_dev_array_size(map) * disc_count; | |
11961 | unsigned long long d_size = imsm_dev_size(dev); | |
11962 | int u_size; | |
11963 | ||
42e02e61 | 11964 | if (calc_size == d_size) |
895ffd99 MT |
11965 | continue; |
11966 | ||
ff904202 MT |
11967 | /* There is a difference, confirm that imsm_dev_size is |
11968 | * smaller and push update. | |
895ffd99 | 11969 | */ |
ff904202 MT |
11970 | if (d_size > calc_size) { |
11971 | pr_err("imsm: dev size of subarray %d is incorrect\n", | |
11972 | i); | |
895ffd99 MT |
11973 | goto exit; |
11974 | } | |
11975 | memset(&geo, 0, sizeof(struct geo_params)); | |
11976 | geo.size = d_size; | |
11977 | u_size = imsm_create_metadata_update_for_size_change(st, &geo, | |
11978 | &update); | |
895ffd99 MT |
11979 | imsm_update_metadata_locally(st, update, u_size); |
11980 | if (st->update_tail) { | |
11981 | append_metadata_update(st, update, u_size); | |
11982 | flush_metadata_updates(st); | |
11983 | st->update_tail = &st->updates; | |
11984 | } else { | |
11985 | imsm_sync_metadata(st); | |
5ce5a15f | 11986 | free(update); |
895ffd99 MT |
11987 | } |
11988 | } | |
11989 | ret_val = 0; | |
11990 | exit: | |
11991 | super->current_vol = tmp; | |
11992 | return ret_val; | |
11993 | } | |
11994 | ||
d04f65f4 N |
11995 | static int imsm_reshape_super(struct supertype *st, unsigned long long size, |
11996 | int level, | |
78b10e66 | 11997 | int layout, int chunksize, int raid_disks, |
41784c88 | 11998 | int delta_disks, char *backup, char *dev, |
016e00f5 | 11999 | int direction, int verbose) |
78b10e66 | 12000 | { |
78b10e66 N |
12001 | int ret_val = 1; |
12002 | struct geo_params geo; | |
12003 | ||
1ade5cc1 | 12004 | dprintf("(enter)\n"); |
78b10e66 | 12005 | |
71204a50 | 12006 | memset(&geo, 0, sizeof(struct geo_params)); |
78b10e66 N |
12007 | |
12008 | geo.dev_name = dev; | |
4dd2df09 | 12009 | strcpy(geo.devnm, st->devnm); |
78b10e66 N |
12010 | geo.size = size; |
12011 | geo.level = level; | |
12012 | geo.layout = layout; | |
12013 | geo.chunksize = chunksize; | |
12014 | geo.raid_disks = raid_disks; | |
41784c88 AK |
12015 | if (delta_disks != UnSet) |
12016 | geo.raid_disks += delta_disks; | |
78b10e66 | 12017 | |
1ade5cc1 N |
12018 | dprintf("for level : %i\n", geo.level); |
12019 | dprintf("for raid_disks : %i\n", geo.raid_disks); | |
78b10e66 | 12020 | |
4dd2df09 | 12021 | if (strcmp(st->container_devnm, st->devnm) == 0) { |
694575e7 KW |
12022 | /* On container level we can only increase number of devices. */ |
12023 | dprintf("imsm: info: Container operation\n"); | |
78b10e66 | 12024 | int old_raid_disks = 0; |
6dc0be30 | 12025 | |
78b10e66 | 12026 | if (imsm_reshape_is_allowed_on_container( |
fbf3d202 | 12027 | st, &geo, &old_raid_disks, direction)) { |
78b10e66 N |
12028 | struct imsm_update_reshape *u = NULL; |
12029 | int len; | |
12030 | ||
895ffd99 MT |
12031 | if (imsm_fix_size_mismatch(st, -1)) { |
12032 | dprintf("imsm: Cannot fix size mismatch\n"); | |
12033 | goto exit_imsm_reshape_super; | |
12034 | } | |
12035 | ||
78b10e66 N |
12036 | len = imsm_create_metadata_update_for_reshape( |
12037 | st, &geo, old_raid_disks, &u); | |
12038 | ||
ed08d51c AK |
12039 | if (len <= 0) { |
12040 | dprintf("imsm: Cannot prepare update\n"); | |
12041 | goto exit_imsm_reshape_super; | |
12042 | } | |
12043 | ||
8dd70bce AK |
12044 | ret_val = 0; |
12045 | /* update metadata locally */ | |
12046 | imsm_update_metadata_locally(st, u, len); | |
12047 | /* and possibly remotely */ | |
12048 | if (st->update_tail) | |
12049 | append_metadata_update(st, u, len); | |
12050 | else | |
ed08d51c | 12051 | free(u); |
8dd70bce | 12052 | |
694575e7 | 12053 | } else { |
7a862a02 | 12054 | pr_err("(imsm) Operation is not allowed on this container\n"); |
694575e7 KW |
12055 | } |
12056 | } else { | |
12057 | /* On volume level we support following operations | |
471bceb6 KW |
12058 | * - takeover: raid10 -> raid0; raid0 -> raid10 |
12059 | * - chunk size migration | |
12060 | * - migration: raid5 -> raid0; raid0 -> raid5 | |
12061 | */ | |
12062 | struct intel_super *super = st->sb; | |
12063 | struct intel_dev *dev = super->devlist; | |
4dd2df09 | 12064 | int change; |
694575e7 | 12065 | dprintf("imsm: info: Volume operation\n"); |
471bceb6 KW |
12066 | /* find requested device */ |
12067 | while (dev) { | |
1011e834 | 12068 | char *devnm = |
4dd2df09 N |
12069 | imsm_find_array_devnm_by_subdev( |
12070 | dev->index, st->container_devnm); | |
12071 | if (devnm && strcmp(devnm, geo.devnm) == 0) | |
471bceb6 KW |
12072 | break; |
12073 | dev = dev->next; | |
12074 | } | |
12075 | if (dev == NULL) { | |
4dd2df09 N |
12076 | pr_err("Cannot find %s (%s) subarray\n", |
12077 | geo.dev_name, geo.devnm); | |
471bceb6 KW |
12078 | goto exit_imsm_reshape_super; |
12079 | } | |
12080 | super->current_vol = dev->index; | |
fbf3d202 | 12081 | change = imsm_analyze_change(st, &geo, direction); |
694575e7 | 12082 | switch (change) { |
471bceb6 | 12083 | case CH_TAKEOVER: |
bb025c2f | 12084 | ret_val = imsm_takeover(st, &geo); |
694575e7 | 12085 | break; |
48c5303a PC |
12086 | case CH_MIGRATION: { |
12087 | struct imsm_update_reshape_migration *u = NULL; | |
12088 | int len = | |
12089 | imsm_create_metadata_update_for_migration( | |
12090 | st, &geo, &u); | |
12091 | if (len < 1) { | |
7a862a02 | 12092 | dprintf("imsm: Cannot prepare update\n"); |
48c5303a PC |
12093 | break; |
12094 | } | |
471bceb6 | 12095 | ret_val = 0; |
48c5303a PC |
12096 | /* update metadata locally */ |
12097 | imsm_update_metadata_locally(st, u, len); | |
12098 | /* and possibly remotely */ | |
12099 | if (st->update_tail) | |
12100 | append_metadata_update(st, u, len); | |
12101 | else | |
12102 | free(u); | |
12103 | } | |
12104 | break; | |
7abc9871 | 12105 | case CH_ARRAY_SIZE: { |
f3871fdc AK |
12106 | struct imsm_update_size_change *u = NULL; |
12107 | int len = | |
12108 | imsm_create_metadata_update_for_size_change( | |
12109 | st, &geo, &u); | |
12110 | if (len < 1) { | |
7a862a02 | 12111 | dprintf("imsm: Cannot prepare update\n"); |
f3871fdc AK |
12112 | break; |
12113 | } | |
12114 | ret_val = 0; | |
12115 | /* update metadata locally */ | |
12116 | imsm_update_metadata_locally(st, u, len); | |
12117 | /* and possibly remotely */ | |
12118 | if (st->update_tail) | |
12119 | append_metadata_update(st, u, len); | |
12120 | else | |
12121 | free(u); | |
7abc9871 AK |
12122 | } |
12123 | break; | |
471bceb6 KW |
12124 | default: |
12125 | ret_val = 1; | |
694575e7 | 12126 | } |
694575e7 | 12127 | } |
78b10e66 | 12128 | |
ed08d51c | 12129 | exit_imsm_reshape_super: |
78b10e66 N |
12130 | dprintf("imsm: reshape_super Exit code = %i\n", ret_val); |
12131 | return ret_val; | |
12132 | } | |
2cda7640 | 12133 | |
0febb20c AO |
12134 | #define COMPLETED_OK 0 |
12135 | #define COMPLETED_NONE 1 | |
12136 | #define COMPLETED_DELAYED 2 | |
12137 | ||
12138 | static int read_completed(int fd, unsigned long long *val) | |
12139 | { | |
12140 | int ret; | |
12141 | char buf[50]; | |
12142 | ||
12143 | ret = sysfs_fd_get_str(fd, buf, 50); | |
12144 | if (ret < 0) | |
12145 | return ret; | |
12146 | ||
12147 | ret = COMPLETED_OK; | |
12148 | if (strncmp(buf, "none", 4) == 0) { | |
12149 | ret = COMPLETED_NONE; | |
12150 | } else if (strncmp(buf, "delayed", 7) == 0) { | |
12151 | ret = COMPLETED_DELAYED; | |
12152 | } else { | |
12153 | char *ep; | |
12154 | *val = strtoull(buf, &ep, 0); | |
12155 | if (ep == buf || (*ep != 0 && *ep != '\n' && *ep != ' ')) | |
12156 | ret = -1; | |
12157 | } | |
12158 | return ret; | |
12159 | } | |
12160 | ||
eee67a47 AK |
12161 | /******************************************************************************* |
12162 | * Function: wait_for_reshape_imsm | |
12163 | * Description: Function writes new sync_max value and waits until | |
12164 | * reshape process reach new position | |
12165 | * Parameters: | |
12166 | * sra : general array info | |
eee67a47 AK |
12167 | * ndata : number of disks in new array's layout |
12168 | * Returns: | |
12169 | * 0 : success, | |
12170 | * 1 : there is no reshape in progress, | |
12171 | * -1 : fail | |
12172 | ******************************************************************************/ | |
ae9f01f8 | 12173 | int wait_for_reshape_imsm(struct mdinfo *sra, int ndata) |
eee67a47 | 12174 | { |
85ca499c | 12175 | int fd = sysfs_get_fd(sra, NULL, "sync_completed"); |
df2647fa | 12176 | int retry = 3; |
eee67a47 | 12177 | unsigned long long completed; |
ae9f01f8 AK |
12178 | /* to_complete : new sync_max position */ |
12179 | unsigned long long to_complete = sra->reshape_progress; | |
12180 | unsigned long long position_to_set = to_complete / ndata; | |
eee67a47 | 12181 | |
4389ce73 | 12182 | if (!is_fd_valid(fd)) { |
1ade5cc1 | 12183 | dprintf("cannot open reshape_position\n"); |
eee67a47 | 12184 | return 1; |
ae9f01f8 | 12185 | } |
eee67a47 | 12186 | |
df2647fa PB |
12187 | do { |
12188 | if (sysfs_fd_get_ll(fd, &completed) < 0) { | |
12189 | if (!retry) { | |
12190 | dprintf("cannot read reshape_position (no reshape in progres)\n"); | |
12191 | close(fd); | |
12192 | return 1; | |
12193 | } | |
239b3cc0 | 12194 | sleep_for(0, MSEC_TO_NSEC(30), true); |
df2647fa PB |
12195 | } else |
12196 | break; | |
12197 | } while (retry--); | |
eee67a47 | 12198 | |
85ca499c | 12199 | if (completed > position_to_set) { |
1ade5cc1 | 12200 | dprintf("wrong next position to set %llu (%llu)\n", |
85ca499c | 12201 | to_complete, position_to_set); |
ae9f01f8 AK |
12202 | close(fd); |
12203 | return -1; | |
12204 | } | |
12205 | dprintf("Position set: %llu\n", position_to_set); | |
12206 | if (sysfs_set_num(sra, NULL, "sync_max", | |
12207 | position_to_set) != 0) { | |
1ade5cc1 | 12208 | dprintf("cannot set reshape position to %llu\n", |
ae9f01f8 AK |
12209 | position_to_set); |
12210 | close(fd); | |
12211 | return -1; | |
eee67a47 AK |
12212 | } |
12213 | ||
eee67a47 | 12214 | do { |
0febb20c | 12215 | int rc; |
eee67a47 | 12216 | char action[20]; |
5ff3a780 | 12217 | int timeout = 3000; |
0febb20c | 12218 | |
5ff3a780 | 12219 | sysfs_wait(fd, &timeout); |
a47e44fb AK |
12220 | if (sysfs_get_str(sra, NULL, "sync_action", |
12221 | action, 20) > 0 && | |
d7d3809a | 12222 | strncmp(action, "reshape", 7) != 0) { |
b2be2b62 AO |
12223 | if (strncmp(action, "idle", 4) == 0) |
12224 | break; | |
d7d3809a AP |
12225 | close(fd); |
12226 | return -1; | |
12227 | } | |
0febb20c AO |
12228 | |
12229 | rc = read_completed(fd, &completed); | |
12230 | if (rc < 0) { | |
1ade5cc1 | 12231 | dprintf("cannot read reshape_position (in loop)\n"); |
eee67a47 AK |
12232 | close(fd); |
12233 | return 1; | |
0febb20c AO |
12234 | } else if (rc == COMPLETED_NONE) |
12235 | break; | |
85ca499c | 12236 | } while (completed < position_to_set); |
b2be2b62 | 12237 | |
eee67a47 AK |
12238 | close(fd); |
12239 | return 0; | |
eee67a47 AK |
12240 | } |
12241 | ||
b915c95f AK |
12242 | /******************************************************************************* |
12243 | * Function: check_degradation_change | |
12244 | * Description: Check that array hasn't become failed. | |
12245 | * Parameters: | |
12246 | * info : for sysfs access | |
12247 | * sources : source disks descriptors | |
12248 | * degraded: previous degradation level | |
12249 | * Returns: | |
12250 | * degradation level | |
12251 | ******************************************************************************/ | |
12252 | int check_degradation_change(struct mdinfo *info, | |
12253 | int *sources, | |
12254 | int degraded) | |
12255 | { | |
12256 | unsigned long long new_degraded; | |
e1993023 LD |
12257 | int rv; |
12258 | ||
12259 | rv = sysfs_get_ll(info, NULL, "degraded", &new_degraded); | |
089f9d79 | 12260 | if (rv == -1 || (new_degraded != (unsigned long long)degraded)) { |
b915c95f AK |
12261 | /* check each device to ensure it is still working */ |
12262 | struct mdinfo *sd; | |
12263 | new_degraded = 0; | |
12264 | for (sd = info->devs ; sd ; sd = sd->next) { | |
12265 | if (sd->disk.state & (1<<MD_DISK_FAULTY)) | |
12266 | continue; | |
12267 | if (sd->disk.state & (1<<MD_DISK_SYNC)) { | |
cf52eff5 | 12268 | char sbuf[100]; |
4389ce73 | 12269 | int raid_disk = sd->disk.raid_disk; |
cf52eff5 | 12270 | |
b915c95f | 12271 | if (sysfs_get_str(info, |
cf52eff5 | 12272 | sd, "state", sbuf, sizeof(sbuf)) < 0 || |
b915c95f AK |
12273 | strstr(sbuf, "faulty") || |
12274 | strstr(sbuf, "in_sync") == NULL) { | |
12275 | /* this device is dead */ | |
12276 | sd->disk.state = (1<<MD_DISK_FAULTY); | |
4389ce73 MT |
12277 | if (raid_disk >= 0) |
12278 | close_fd(&sources[raid_disk]); | |
b915c95f AK |
12279 | new_degraded++; |
12280 | } | |
12281 | } | |
12282 | } | |
12283 | } | |
12284 | ||
12285 | return new_degraded; | |
12286 | } | |
12287 | ||
10f22854 AK |
12288 | /******************************************************************************* |
12289 | * Function: imsm_manage_reshape | |
12290 | * Description: Function finds array under reshape and it manages reshape | |
12291 | * process. It creates stripes backups (if required) and sets | |
942e1cdb | 12292 | * checkpoints. |
10f22854 AK |
12293 | * Parameters: |
12294 | * afd : Backup handle (nattive) - not used | |
12295 | * sra : general array info | |
12296 | * reshape : reshape parameters - not used | |
12297 | * st : supertype structure | |
12298 | * blocks : size of critical section [blocks] | |
12299 | * fds : table of source device descriptor | |
12300 | * offsets : start of array (offest per devices) | |
12301 | * dests : not used | |
12302 | * destfd : table of destination device descriptor | |
12303 | * destoffsets : table of destination offsets (per device) | |
12304 | * Returns: | |
12305 | * 1 : success, reshape is done | |
12306 | * 0 : fail | |
12307 | ******************************************************************************/ | |
999b4972 N |
12308 | static int imsm_manage_reshape( |
12309 | int afd, struct mdinfo *sra, struct reshape *reshape, | |
10f22854 | 12310 | struct supertype *st, unsigned long backup_blocks, |
999b4972 N |
12311 | int *fds, unsigned long long *offsets, |
12312 | int dests, int *destfd, unsigned long long *destoffsets) | |
12313 | { | |
10f22854 AK |
12314 | int ret_val = 0; |
12315 | struct intel_super *super = st->sb; | |
594dc1b8 | 12316 | struct intel_dev *dv; |
de44e46f | 12317 | unsigned int sector_size = super->sector_size; |
10f22854 | 12318 | struct imsm_dev *dev = NULL; |
9529d343 | 12319 | struct imsm_map *map_src, *map_dest; |
10f22854 AK |
12320 | int migr_vol_qan = 0; |
12321 | int ndata, odata; /* [bytes] */ | |
12322 | int chunk; /* [bytes] */ | |
12323 | struct migr_record *migr_rec; | |
12324 | char *buf = NULL; | |
12325 | unsigned int buf_size; /* [bytes] */ | |
12326 | unsigned long long max_position; /* array size [bytes] */ | |
12327 | unsigned long long next_step; /* [blocks]/[bytes] */ | |
12328 | unsigned long long old_data_stripe_length; | |
10f22854 AK |
12329 | unsigned long long start_src; /* [bytes] */ |
12330 | unsigned long long start; /* [bytes] */ | |
12331 | unsigned long long start_buf_shift; /* [bytes] */ | |
b915c95f | 12332 | int degraded = 0; |
ab724b98 | 12333 | int source_layout = 0; |
895ffd99 | 12334 | int subarray_index = -1; |
10f22854 | 12335 | |
79a16a9b JS |
12336 | if (!sra) |
12337 | return ret_val; | |
12338 | ||
12339 | if (!fds || !offsets) | |
10f22854 AK |
12340 | goto abort; |
12341 | ||
12342 | /* Find volume during the reshape */ | |
12343 | for (dv = super->devlist; dv; dv = dv->next) { | |
fc54fe7a JS |
12344 | if (dv->dev->vol.migr_type == MIGR_GEN_MIGR && |
12345 | dv->dev->vol.migr_state == 1) { | |
10f22854 AK |
12346 | dev = dv->dev; |
12347 | migr_vol_qan++; | |
895ffd99 | 12348 | subarray_index = dv->index; |
10f22854 AK |
12349 | } |
12350 | } | |
12351 | /* Only one volume can migrate at the same time */ | |
12352 | if (migr_vol_qan != 1) { | |
676e87a8 | 12353 | pr_err("%s", migr_vol_qan ? |
10f22854 AK |
12354 | "Number of migrating volumes greater than 1\n" : |
12355 | "There is no volume during migrationg\n"); | |
12356 | goto abort; | |
12357 | } | |
12358 | ||
9529d343 | 12359 | map_dest = get_imsm_map(dev, MAP_0); |
238c0a71 | 12360 | map_src = get_imsm_map(dev, MAP_1); |
10f22854 AK |
12361 | if (map_src == NULL) |
12362 | goto abort; | |
10f22854 | 12363 | |
9529d343 MD |
12364 | ndata = imsm_num_data_members(map_dest); |
12365 | odata = imsm_num_data_members(map_src); | |
10f22854 | 12366 | |
7b1ab482 | 12367 | chunk = __le16_to_cpu(map_src->blocks_per_strip) * 512; |
10f22854 AK |
12368 | old_data_stripe_length = odata * chunk; |
12369 | ||
12370 | migr_rec = super->migr_rec; | |
12371 | ||
10f22854 AK |
12372 | /* initialize migration record for start condition */ |
12373 | if (sra->reshape_progress == 0) | |
12374 | init_migr_record_imsm(st, dev, sra); | |
b2c59438 AK |
12375 | else { |
12376 | if (__le32_to_cpu(migr_rec->rec_status) != UNIT_SRC_NORMAL) { | |
7a862a02 | 12377 | dprintf("imsm: cannot restart migration when data are present in copy area.\n"); |
b2c59438 AK |
12378 | goto abort; |
12379 | } | |
6a75c8ca AK |
12380 | /* Save checkpoint to update migration record for current |
12381 | * reshape position (in md). It can be farther than current | |
12382 | * reshape position in metadata. | |
12383 | */ | |
12384 | if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) { | |
12385 | /* ignore error == 2, this can mean end of reshape here | |
12386 | */ | |
7a862a02 | 12387 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL, initial save)\n"); |
6a75c8ca AK |
12388 | goto abort; |
12389 | } | |
b2c59438 | 12390 | } |
10f22854 AK |
12391 | |
12392 | /* size for data */ | |
12393 | buf_size = __le32_to_cpu(migr_rec->blocks_per_unit) * 512; | |
12394 | /* extend buffer size for parity disk */ | |
12395 | buf_size += __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
3e684231 | 12396 | /* add space for stripe alignment */ |
10f22854 | 12397 | buf_size += old_data_stripe_length; |
de44e46f PB |
12398 | if (posix_memalign((void **)&buf, MAX_SECTOR_SIZE, buf_size)) { |
12399 | dprintf("imsm: Cannot allocate checkpoint buffer\n"); | |
10f22854 AK |
12400 | goto abort; |
12401 | } | |
12402 | ||
3ef4403c | 12403 | max_position = sra->component_size * ndata; |
68eb8bc6 | 12404 | source_layout = imsm_level_to_layout(map_src->raid_level); |
10f22854 | 12405 | |
9f421827 PB |
12406 | while (current_migr_unit(migr_rec) < |
12407 | get_num_migr_units(migr_rec)) { | |
10f22854 AK |
12408 | /* current reshape position [blocks] */ |
12409 | unsigned long long current_position = | |
12410 | __le32_to_cpu(migr_rec->blocks_per_unit) | |
9f421827 | 12411 | * current_migr_unit(migr_rec); |
10f22854 AK |
12412 | unsigned long long border; |
12413 | ||
b915c95f AK |
12414 | /* Check that array hasn't become failed. |
12415 | */ | |
12416 | degraded = check_degradation_change(sra, fds, degraded); | |
12417 | if (degraded > 1) { | |
7a862a02 | 12418 | dprintf("imsm: Abort reshape due to degradation level (%i)\n", degraded); |
b915c95f AK |
12419 | goto abort; |
12420 | } | |
12421 | ||
10f22854 AK |
12422 | next_step = __le32_to_cpu(migr_rec->blocks_per_unit); |
12423 | ||
12424 | if ((current_position + next_step) > max_position) | |
12425 | next_step = max_position - current_position; | |
12426 | ||
92144abf | 12427 | start = current_position * 512; |
10f22854 | 12428 | |
942e1cdb | 12429 | /* align reading start to old geometry */ |
10f22854 AK |
12430 | start_buf_shift = start % old_data_stripe_length; |
12431 | start_src = start - start_buf_shift; | |
12432 | ||
12433 | border = (start_src / odata) - (start / ndata); | |
12434 | border /= 512; | |
12435 | if (border <= __le32_to_cpu(migr_rec->dest_depth_per_unit)) { | |
12436 | /* save critical stripes to buf | |
12437 | * start - start address of current unit | |
12438 | * to backup [bytes] | |
12439 | * start_src - start address of current unit | |
12440 | * to backup alligned to source array | |
12441 | * [bytes] | |
12442 | */ | |
594dc1b8 | 12443 | unsigned long long next_step_filler; |
10f22854 AK |
12444 | unsigned long long copy_length = next_step * 512; |
12445 | ||
12446 | /* allign copy area length to stripe in old geometry */ | |
12447 | next_step_filler = ((copy_length + start_buf_shift) | |
12448 | % old_data_stripe_length); | |
12449 | if (next_step_filler) | |
12450 | next_step_filler = (old_data_stripe_length | |
12451 | - next_step_filler); | |
7a862a02 | 12452 | 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 |
12453 | start, start_src, copy_length, |
12454 | start_buf_shift, next_step_filler); | |
12455 | ||
12456 | if (save_stripes(fds, offsets, map_src->num_members, | |
ab724b98 AK |
12457 | chunk, map_src->raid_level, |
12458 | source_layout, 0, NULL, start_src, | |
10f22854 AK |
12459 | copy_length + |
12460 | next_step_filler + start_buf_shift, | |
12461 | buf)) { | |
7a862a02 | 12462 | dprintf("imsm: Cannot save stripes to buffer\n"); |
10f22854 AK |
12463 | goto abort; |
12464 | } | |
12465 | /* Convert data to destination format and store it | |
12466 | * in backup general migration area | |
12467 | */ | |
12468 | if (save_backup_imsm(st, dev, sra, | |
aea93171 | 12469 | buf + start_buf_shift, copy_length)) { |
7a862a02 | 12470 | dprintf("imsm: Cannot save stripes to target devices\n"); |
10f22854 AK |
12471 | goto abort; |
12472 | } | |
12473 | if (save_checkpoint_imsm(st, sra, | |
12474 | UNIT_SRC_IN_CP_AREA)) { | |
7a862a02 | 12475 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_IN_CP_AREA)\n"); |
10f22854 AK |
12476 | goto abort; |
12477 | } | |
8016a6d4 AK |
12478 | } else { |
12479 | /* set next step to use whole border area */ | |
12480 | border /= next_step; | |
12481 | if (border > 1) | |
12482 | next_step *= border; | |
10f22854 AK |
12483 | } |
12484 | /* When data backed up, checkpoint stored, | |
12485 | * kick the kernel to reshape unit of data | |
12486 | */ | |
12487 | next_step = next_step + sra->reshape_progress; | |
8016a6d4 AK |
12488 | /* limit next step to array max position */ |
12489 | if (next_step > max_position) | |
12490 | next_step = max_position; | |
10f22854 AK |
12491 | sysfs_set_num(sra, NULL, "suspend_lo", sra->reshape_progress); |
12492 | sysfs_set_num(sra, NULL, "suspend_hi", next_step); | |
ae9f01f8 | 12493 | sra->reshape_progress = next_step; |
10f22854 AK |
12494 | |
12495 | /* wait until reshape finish */ | |
c85338c6 | 12496 | if (wait_for_reshape_imsm(sra, ndata)) { |
c47b0ff6 AK |
12497 | dprintf("wait_for_reshape_imsm returned error!\n"); |
12498 | goto abort; | |
12499 | } | |
84d11e6c N |
12500 | if (sigterm) |
12501 | goto abort; | |
10f22854 | 12502 | |
0228d92c AK |
12503 | if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) { |
12504 | /* ignore error == 2, this can mean end of reshape here | |
12505 | */ | |
7a862a02 | 12506 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL)\n"); |
10f22854 AK |
12507 | goto abort; |
12508 | } | |
12509 | ||
12510 | } | |
12511 | ||
71e5411e PB |
12512 | /* clear migr_rec on disks after successful migration */ |
12513 | struct dl *d; | |
12514 | ||
85337573 | 12515 | memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE); |
71e5411e PB |
12516 | for (d = super->disks; d; d = d->next) { |
12517 | if (d->index < 0 || is_failed(&d->disk)) | |
12518 | continue; | |
12519 | unsigned long long dsize; | |
12520 | ||
12521 | get_dev_size(d->fd, NULL, &dsize); | |
de44e46f | 12522 | if (lseek64(d->fd, dsize - MIGR_REC_SECTOR_POSITION*sector_size, |
71e5411e | 12523 | SEEK_SET) >= 0) { |
466070ad | 12524 | if ((unsigned int)write(d->fd, super->migr_rec_buf, |
de44e46f PB |
12525 | MIGR_REC_BUF_SECTORS*sector_size) != |
12526 | MIGR_REC_BUF_SECTORS*sector_size) | |
71e5411e PB |
12527 | perror("Write migr_rec failed"); |
12528 | } | |
12529 | } | |
12530 | ||
10f22854 AK |
12531 | /* return '1' if done */ |
12532 | ret_val = 1; | |
895ffd99 MT |
12533 | |
12534 | /* After the reshape eliminate size mismatch in metadata. | |
12535 | * Don't update md/component_size here, volume hasn't | |
12536 | * to take whole space. It is allowed by kernel. | |
12537 | * md/component_size will be set propoperly after next assembly. | |
12538 | */ | |
12539 | imsm_fix_size_mismatch(st, subarray_index); | |
12540 | ||
10f22854 AK |
12541 | abort: |
12542 | free(buf); | |
942e1cdb N |
12543 | /* See Grow.c: abort_reshape() for further explanation */ |
12544 | sysfs_set_num(sra, NULL, "suspend_lo", 0x7FFFFFFFFFFFFFFFULL); | |
12545 | sysfs_set_num(sra, NULL, "suspend_hi", 0); | |
12546 | sysfs_set_num(sra, NULL, "suspend_lo", 0); | |
10f22854 AK |
12547 | |
12548 | return ret_val; | |
999b4972 | 12549 | } |
0c21b485 | 12550 | |
fbc42556 JR |
12551 | /******************************************************************************* |
12552 | * Function: calculate_bitmap_min_chunksize | |
12553 | * Description: Calculates the minimal valid bitmap chunk size | |
12554 | * Parameters: | |
12555 | * max_bits : indicate how many bits can be used for the bitmap | |
12556 | * data_area_size : the size of the data area covered by the bitmap | |
12557 | * | |
12558 | * Returns: | |
12559 | * The bitmap chunk size | |
12560 | ******************************************************************************/ | |
12561 | static unsigned long long | |
12562 | calculate_bitmap_min_chunksize(unsigned long long max_bits, | |
12563 | unsigned long long data_area_size) | |
12564 | { | |
12565 | unsigned long long min_chunk = | |
12566 | 4096; /* sub-page chunks don't work yet.. */ | |
12567 | unsigned long long bits = data_area_size / min_chunk + 1; | |
12568 | ||
12569 | while (bits > max_bits) { | |
12570 | min_chunk *= 2; | |
12571 | bits = (bits + 1) / 2; | |
12572 | } | |
12573 | return min_chunk; | |
12574 | } | |
12575 | ||
12576 | /******************************************************************************* | |
12577 | * Function: calculate_bitmap_chunksize | |
12578 | * Description: Calculates the bitmap chunk size for the given device | |
12579 | * Parameters: | |
12580 | * st : supertype information | |
12581 | * dev : device for the bitmap | |
12582 | * | |
12583 | * Returns: | |
12584 | * The bitmap chunk size | |
12585 | ******************************************************************************/ | |
12586 | static unsigned long long calculate_bitmap_chunksize(struct supertype *st, | |
12587 | struct imsm_dev *dev) | |
12588 | { | |
12589 | struct intel_super *super = st->sb; | |
12590 | unsigned long long min_chunksize; | |
12591 | unsigned long long result = IMSM_DEFAULT_BITMAP_CHUNKSIZE; | |
12592 | size_t dev_size = imsm_dev_size(dev); | |
12593 | ||
12594 | min_chunksize = calculate_bitmap_min_chunksize( | |
12595 | IMSM_BITMAP_AREA_SIZE * super->sector_size, dev_size); | |
12596 | ||
12597 | if (result < min_chunksize) | |
12598 | result = min_chunksize; | |
12599 | ||
12600 | return result; | |
12601 | } | |
12602 | ||
12603 | /******************************************************************************* | |
12604 | * Function: init_bitmap_header | |
12605 | * Description: Initialize the bitmap header structure | |
12606 | * Parameters: | |
12607 | * st : supertype information | |
12608 | * bms : bitmap header struct to initialize | |
12609 | * dev : device for the bitmap | |
12610 | * | |
12611 | * Returns: | |
12612 | * 0 : success | |
12613 | * -1 : fail | |
12614 | ******************************************************************************/ | |
12615 | static int init_bitmap_header(struct supertype *st, struct bitmap_super_s *bms, | |
12616 | struct imsm_dev *dev) | |
12617 | { | |
12618 | int vol_uuid[4]; | |
12619 | ||
12620 | if (!bms || !dev) | |
12621 | return -1; | |
12622 | ||
12623 | bms->magic = __cpu_to_le32(BITMAP_MAGIC); | |
12624 | bms->version = __cpu_to_le32(BITMAP_MAJOR_HI); | |
12625 | bms->daemon_sleep = __cpu_to_le32(IMSM_DEFAULT_BITMAP_DAEMON_SLEEP); | |
12626 | bms->sync_size = __cpu_to_le64(IMSM_BITMAP_AREA_SIZE); | |
12627 | bms->write_behind = __cpu_to_le32(0); | |
12628 | ||
12629 | uuid_from_super_imsm(st, vol_uuid); | |
12630 | memcpy(bms->uuid, vol_uuid, 16); | |
12631 | ||
12632 | bms->chunksize = calculate_bitmap_chunksize(st, dev); | |
12633 | ||
12634 | return 0; | |
12635 | } | |
12636 | ||
12637 | /******************************************************************************* | |
12638 | * Function: validate_internal_bitmap_for_drive | |
12639 | * Description: Verify if the bitmap header for a given drive. | |
12640 | * Parameters: | |
12641 | * st : supertype information | |
12642 | * offset : The offset from the beginning of the drive where to look for | |
12643 | * the bitmap header. | |
12644 | * d : the drive info | |
12645 | * | |
12646 | * Returns: | |
12647 | * 0 : success | |
12648 | * -1 : fail | |
12649 | ******************************************************************************/ | |
12650 | static int validate_internal_bitmap_for_drive(struct supertype *st, | |
12651 | unsigned long long offset, | |
12652 | struct dl *d) | |
12653 | { | |
12654 | struct intel_super *super = st->sb; | |
12655 | int ret = -1; | |
12656 | int vol_uuid[4]; | |
12657 | bitmap_super_t *bms; | |
12658 | int fd; | |
12659 | ||
12660 | if (!d) | |
12661 | return -1; | |
12662 | ||
12663 | void *read_buf; | |
12664 | ||
12665 | if (posix_memalign(&read_buf, MAX_SECTOR_SIZE, IMSM_BITMAP_HEADER_SIZE)) | |
12666 | return -1; | |
12667 | ||
12668 | fd = d->fd; | |
4389ce73 | 12669 | if (!is_fd_valid(fd)) { |
fbc42556 | 12670 | fd = open(d->devname, O_RDONLY, 0); |
4389ce73 MT |
12671 | |
12672 | if (!is_fd_valid(fd)) { | |
fbc42556 JR |
12673 | dprintf("cannot open the device %s\n", d->devname); |
12674 | goto abort; | |
12675 | } | |
12676 | } | |
12677 | ||
12678 | if (lseek64(fd, offset * super->sector_size, SEEK_SET) < 0) | |
12679 | goto abort; | |
12680 | if (read(fd, read_buf, IMSM_BITMAP_HEADER_SIZE) != | |
12681 | IMSM_BITMAP_HEADER_SIZE) | |
12682 | goto abort; | |
12683 | ||
12684 | uuid_from_super_imsm(st, vol_uuid); | |
12685 | ||
12686 | bms = read_buf; | |
12687 | if ((bms->magic != __cpu_to_le32(BITMAP_MAGIC)) || | |
12688 | (bms->version != __cpu_to_le32(BITMAP_MAJOR_HI)) || | |
12689 | (!same_uuid((int *)bms->uuid, vol_uuid, st->ss->swapuuid))) { | |
12690 | dprintf("wrong bitmap header detected\n"); | |
12691 | goto abort; | |
12692 | } | |
12693 | ||
12694 | ret = 0; | |
12695 | abort: | |
4389ce73 MT |
12696 | if (!is_fd_valid(d->fd)) |
12697 | close_fd(&fd); | |
12698 | ||
fbc42556 JR |
12699 | if (read_buf) |
12700 | free(read_buf); | |
12701 | ||
12702 | return ret; | |
12703 | } | |
12704 | ||
12705 | /******************************************************************************* | |
12706 | * Function: validate_internal_bitmap_imsm | |
12707 | * Description: Verify if the bitmap header is in place and with proper data. | |
12708 | * Parameters: | |
12709 | * st : supertype information | |
12710 | * | |
12711 | * Returns: | |
12712 | * 0 : success or device w/o RWH_BITMAP | |
12713 | * -1 : fail | |
12714 | ******************************************************************************/ | |
12715 | static int validate_internal_bitmap_imsm(struct supertype *st) | |
12716 | { | |
12717 | struct intel_super *super = st->sb; | |
12718 | struct imsm_dev *dev = get_imsm_dev(super, super->current_vol); | |
12719 | unsigned long long offset; | |
12720 | struct dl *d; | |
12721 | ||
fbc42556 JR |
12722 | if (dev->rwh_policy != RWH_BITMAP) |
12723 | return 0; | |
12724 | ||
12725 | offset = get_bitmap_header_sector(super, super->current_vol); | |
12726 | for (d = super->disks; d; d = d->next) { | |
12727 | if (d->index < 0 || is_failed(&d->disk)) | |
12728 | continue; | |
12729 | ||
12730 | if (validate_internal_bitmap_for_drive(st, offset, d)) { | |
12731 | pr_err("imsm: bitmap validation failed\n"); | |
12732 | return -1; | |
12733 | } | |
12734 | } | |
12735 | return 0; | |
12736 | } | |
12737 | ||
12738 | /******************************************************************************* | |
12739 | * Function: add_internal_bitmap_imsm | |
12740 | * Description: Mark the volume to use the bitmap and updates the chunk size value. | |
12741 | * Parameters: | |
12742 | * st : supertype information | |
12743 | * chunkp : bitmap chunk size | |
12744 | * delay : not used for imsm | |
12745 | * write_behind : not used for imsm | |
12746 | * size : not used for imsm | |
12747 | * may_change : not used for imsm | |
12748 | * amajor : not used for imsm | |
12749 | * | |
12750 | * Returns: | |
12751 | * 0 : success | |
12752 | * -1 : fail | |
12753 | ******************************************************************************/ | |
12754 | static int add_internal_bitmap_imsm(struct supertype *st, int *chunkp, | |
12755 | int delay, int write_behind, | |
12756 | unsigned long long size, int may_change, | |
12757 | int amajor) | |
12758 | { | |
12759 | struct intel_super *super = st->sb; | |
12760 | int vol_idx = super->current_vol; | |
12761 | struct imsm_dev *dev; | |
12762 | ||
12763 | if (!super->devlist || vol_idx == -1 || !chunkp) | |
12764 | return -1; | |
12765 | ||
12766 | dev = get_imsm_dev(super, vol_idx); | |
fbc42556 | 12767 | dev->rwh_policy = RWH_BITMAP; |
fbc42556 | 12768 | *chunkp = calculate_bitmap_chunksize(st, dev); |
fbc42556 JR |
12769 | return 0; |
12770 | } | |
12771 | ||
12772 | /******************************************************************************* | |
12773 | * Function: locate_bitmap_imsm | |
12774 | * Description: Seek 'fd' to start of write-intent-bitmap. | |
12775 | * Parameters: | |
12776 | * st : supertype information | |
12777 | * fd : file descriptor for the device | |
12778 | * node_num : not used for imsm | |
12779 | * | |
12780 | * Returns: | |
12781 | * 0 : success | |
12782 | * -1 : fail | |
12783 | ******************************************************************************/ | |
12784 | static int locate_bitmap_imsm(struct supertype *st, int fd, int node_num) | |
12785 | { | |
12786 | struct intel_super *super = st->sb; | |
12787 | unsigned long long offset; | |
12788 | int vol_idx = super->current_vol; | |
12789 | ||
12790 | if (!super->devlist || vol_idx == -1) | |
12791 | return -1; | |
12792 | ||
12793 | offset = get_bitmap_header_sector(super, super->current_vol); | |
12794 | dprintf("bitmap header offset is %llu\n", offset); | |
12795 | ||
12796 | lseek64(fd, offset << 9, 0); | |
12797 | ||
12798 | return 0; | |
12799 | } | |
12800 | ||
12801 | /******************************************************************************* | |
12802 | * Function: write_init_bitmap_imsm | |
12803 | * Description: Write a bitmap header and prepares the area for the bitmap. | |
12804 | * Parameters: | |
12805 | * st : supertype information | |
12806 | * fd : file descriptor for the device | |
12807 | * update : not used for imsm | |
12808 | * | |
12809 | * Returns: | |
12810 | * 0 : success | |
12811 | * -1 : fail | |
12812 | ******************************************************************************/ | |
12813 | static int write_init_bitmap_imsm(struct supertype *st, int fd, | |
12814 | enum bitmap_update update) | |
12815 | { | |
12816 | struct intel_super *super = st->sb; | |
12817 | int vol_idx = super->current_vol; | |
12818 | int ret = 0; | |
12819 | unsigned long long offset; | |
12820 | bitmap_super_t bms = { 0 }; | |
12821 | size_t written = 0; | |
12822 | size_t to_write; | |
12823 | ssize_t rv_num; | |
12824 | void *buf; | |
12825 | ||
12826 | if (!super->devlist || !super->sector_size || vol_idx == -1) | |
12827 | return -1; | |
12828 | ||
12829 | struct imsm_dev *dev = get_imsm_dev(super, vol_idx); | |
12830 | ||
12831 | /* first clear the space for bitmap header */ | |
12832 | unsigned long long bitmap_area_start = | |
12833 | get_bitmap_header_sector(super, vol_idx); | |
12834 | ||
12835 | dprintf("zeroing area start (%llu) and size (%u)\n", bitmap_area_start, | |
12836 | IMSM_BITMAP_AND_HEADER_SIZE / super->sector_size); | |
12837 | if (zero_disk_range(fd, bitmap_area_start, | |
12838 | IMSM_BITMAP_HEADER_SIZE / super->sector_size)) { | |
12839 | pr_err("imsm: cannot zeroing the space for the bitmap\n"); | |
12840 | return -1; | |
12841 | } | |
12842 | ||
12843 | /* The bitmap area should be filled with "1"s to perform initial | |
12844 | * synchronization. | |
12845 | */ | |
12846 | if (posix_memalign(&buf, MAX_SECTOR_SIZE, MAX_SECTOR_SIZE)) | |
12847 | return -1; | |
12848 | memset(buf, 0xFF, MAX_SECTOR_SIZE); | |
12849 | offset = get_bitmap_sector(super, vol_idx); | |
12850 | lseek64(fd, offset << 9, 0); | |
12851 | while (written < IMSM_BITMAP_AREA_SIZE) { | |
12852 | to_write = IMSM_BITMAP_AREA_SIZE - written; | |
12853 | if (to_write > MAX_SECTOR_SIZE) | |
12854 | to_write = MAX_SECTOR_SIZE; | |
12855 | rv_num = write(fd, buf, MAX_SECTOR_SIZE); | |
12856 | if (rv_num != MAX_SECTOR_SIZE) { | |
12857 | ret = -1; | |
12858 | dprintf("cannot initialize bitmap area\n"); | |
12859 | goto abort; | |
12860 | } | |
12861 | written += rv_num; | |
12862 | } | |
12863 | ||
12864 | /* write a bitmap header */ | |
12865 | init_bitmap_header(st, &bms, dev); | |
12866 | memset(buf, 0, MAX_SECTOR_SIZE); | |
12867 | memcpy(buf, &bms, sizeof(bitmap_super_t)); | |
12868 | if (locate_bitmap_imsm(st, fd, 0)) { | |
12869 | ret = -1; | |
12870 | dprintf("cannot locate the bitmap\n"); | |
12871 | goto abort; | |
12872 | } | |
12873 | if (write(fd, buf, MAX_SECTOR_SIZE) != MAX_SECTOR_SIZE) { | |
12874 | ret = -1; | |
12875 | dprintf("cannot write the bitmap header\n"); | |
12876 | goto abort; | |
12877 | } | |
12878 | fsync(fd); | |
12879 | ||
12880 | abort: | |
12881 | free(buf); | |
12882 | ||
12883 | return ret; | |
12884 | } | |
12885 | ||
12886 | /******************************************************************************* | |
12887 | * Function: is_vol_to_setup_bitmap | |
12888 | * Description: Checks if a bitmap should be activated on the dev. | |
12889 | * Parameters: | |
12890 | * info : info about the volume to setup the bitmap | |
12891 | * dev : the device to check against bitmap creation | |
12892 | * | |
12893 | * Returns: | |
12894 | * 0 : bitmap should be set up on the device | |
12895 | * -1 : otherwise | |
12896 | ******************************************************************************/ | |
12897 | static int is_vol_to_setup_bitmap(struct mdinfo *info, struct imsm_dev *dev) | |
12898 | { | |
12899 | if (!dev || !info) | |
12900 | return -1; | |
12901 | ||
12902 | if ((strcmp((char *)dev->volume, info->name) == 0) && | |
12903 | (dev->rwh_policy == RWH_BITMAP)) | |
12904 | return -1; | |
12905 | ||
12906 | return 0; | |
12907 | } | |
12908 | ||
12909 | /******************************************************************************* | |
12910 | * Function: set_bitmap_sysfs | |
12911 | * Description: Set the sysfs atributes of a given volume to activate the bitmap. | |
12912 | * Parameters: | |
12913 | * info : info about the volume where the bitmap should be setup | |
12914 | * chunksize : bitmap chunk size | |
12915 | * location : location of the bitmap | |
12916 | * | |
12917 | * Returns: | |
12918 | * 0 : success | |
12919 | * -1 : fail | |
12920 | ******************************************************************************/ | |
12921 | static int set_bitmap_sysfs(struct mdinfo *info, unsigned long long chunksize, | |
12922 | char *location) | |
12923 | { | |
12924 | /* The bitmap/metadata is set to external to allow changing of value for | |
12925 | * bitmap/location. When external is used, the kernel will treat an offset | |
12926 | * related to the device's first lba (in opposition to the "internal" case | |
12927 | * when this value is related to the beginning of the superblock). | |
12928 | */ | |
12929 | if (sysfs_set_str(info, NULL, "bitmap/metadata", "external")) { | |
12930 | dprintf("failed to set bitmap/metadata\n"); | |
12931 | return -1; | |
12932 | } | |
12933 | ||
12934 | /* It can only be changed when no bitmap is active. | |
12935 | * Should be bigger than 512 and must be power of 2. | |
12936 | * It is expecting the value in bytes. | |
12937 | */ | |
12938 | if (sysfs_set_num(info, NULL, "bitmap/chunksize", | |
12939 | __cpu_to_le32(chunksize))) { | |
12940 | dprintf("failed to set bitmap/chunksize\n"); | |
12941 | return -1; | |
12942 | } | |
12943 | ||
12944 | /* It is expecting the value in sectors. */ | |
12945 | if (sysfs_set_num(info, NULL, "bitmap/space", | |
12946 | __cpu_to_le64(IMSM_BITMAP_AREA_SIZE))) { | |
12947 | dprintf("failed to set bitmap/space\n"); | |
12948 | return -1; | |
12949 | } | |
12950 | ||
12951 | /* Determines the delay between the bitmap updates. | |
12952 | * It is expecting the value in seconds. | |
12953 | */ | |
12954 | if (sysfs_set_num(info, NULL, "bitmap/time_base", | |
12955 | __cpu_to_le64(IMSM_DEFAULT_BITMAP_DAEMON_SLEEP))) { | |
12956 | dprintf("failed to set bitmap/time_base\n"); | |
12957 | return -1; | |
12958 | } | |
12959 | ||
12960 | /* It is expecting the value in sectors with a sign at the beginning. */ | |
12961 | if (sysfs_set_str(info, NULL, "bitmap/location", location)) { | |
12962 | dprintf("failed to set bitmap/location\n"); | |
12963 | return -1; | |
12964 | } | |
12965 | ||
12966 | return 0; | |
12967 | } | |
12968 | ||
12969 | /******************************************************************************* | |
12970 | * Function: set_bitmap_imsm | |
12971 | * Description: Setup the bitmap for the given volume | |
12972 | * Parameters: | |
12973 | * st : supertype information | |
12974 | * info : info about the volume where the bitmap should be setup | |
12975 | * | |
12976 | * Returns: | |
12977 | * 0 : success | |
12978 | * -1 : fail | |
12979 | ******************************************************************************/ | |
12980 | static int set_bitmap_imsm(struct supertype *st, struct mdinfo *info) | |
12981 | { | |
12982 | struct intel_super *super = st->sb; | |
12983 | int prev_current_vol = super->current_vol; | |
12984 | struct imsm_dev *dev; | |
12985 | int ret = -1; | |
12986 | char location[16] = ""; | |
12987 | unsigned long long chunksize; | |
12988 | struct intel_dev *dev_it; | |
12989 | ||
12990 | for (dev_it = super->devlist; dev_it; dev_it = dev_it->next) { | |
12991 | super->current_vol = dev_it->index; | |
12992 | dev = get_imsm_dev(super, super->current_vol); | |
12993 | ||
12994 | if (is_vol_to_setup_bitmap(info, dev)) { | |
12995 | if (validate_internal_bitmap_imsm(st)) { | |
12996 | dprintf("bitmap header validation failed\n"); | |
12997 | goto abort; | |
12998 | } | |
12999 | ||
13000 | chunksize = calculate_bitmap_chunksize(st, dev); | |
13001 | dprintf("chunk size is %llu\n", chunksize); | |
13002 | ||
13003 | snprintf(location, sizeof(location), "+%llu", | |
13004 | get_bitmap_sector(super, super->current_vol)); | |
13005 | dprintf("bitmap offset is %s\n", location); | |
13006 | ||
13007 | if (set_bitmap_sysfs(info, chunksize, location)) { | |
13008 | dprintf("cannot setup the bitmap\n"); | |
13009 | goto abort; | |
13010 | } | |
13011 | } | |
13012 | } | |
13013 | ret = 0; | |
13014 | abort: | |
13015 | super->current_vol = prev_current_vol; | |
13016 | return ret; | |
13017 | } | |
13018 | ||
cdddbdbc | 13019 | struct superswitch super_imsm = { |
cdddbdbc DW |
13020 | .examine_super = examine_super_imsm, |
13021 | .brief_examine_super = brief_examine_super_imsm, | |
4737ae25 | 13022 | .brief_examine_subarrays = brief_examine_subarrays_imsm, |
9d84c8ea | 13023 | .export_examine_super = export_examine_super_imsm, |
cdddbdbc DW |
13024 | .detail_super = detail_super_imsm, |
13025 | .brief_detail_super = brief_detail_super_imsm, | |
bf5a934a | 13026 | .write_init_super = write_init_super_imsm, |
0e600426 N |
13027 | .validate_geometry = validate_geometry_imsm, |
13028 | .add_to_super = add_to_super_imsm, | |
1a64be56 | 13029 | .remove_from_super = remove_from_super_imsm, |
d665cc31 | 13030 | .detail_platform = detail_platform_imsm, |
e50cf220 | 13031 | .export_detail_platform = export_detail_platform_imsm, |
33414a01 | 13032 | .kill_subarray = kill_subarray_imsm, |
aa534678 | 13033 | .update_subarray = update_subarray_imsm, |
2b959fbf | 13034 | .load_container = load_container_imsm, |
71204a50 N |
13035 | .default_geometry = default_geometry_imsm, |
13036 | .get_disk_controller_domain = imsm_get_disk_controller_domain, | |
13037 | .reshape_super = imsm_reshape_super, | |
13038 | .manage_reshape = imsm_manage_reshape, | |
9e2d750d | 13039 | .recover_backup = recover_backup_imsm, |
27156a57 | 13040 | .examine_badblocks = examine_badblocks_imsm, |
cdddbdbc DW |
13041 | .match_home = match_home_imsm, |
13042 | .uuid_from_super= uuid_from_super_imsm, | |
13043 | .getinfo_super = getinfo_super_imsm, | |
5c4cd5da | 13044 | .getinfo_super_disks = getinfo_super_disks_imsm, |
cdddbdbc DW |
13045 | .update_super = update_super_imsm, |
13046 | ||
13047 | .avail_size = avail_size_imsm, | |
fbfdcb06 | 13048 | .get_spare_criteria = get_spare_criteria_imsm, |
cdddbdbc DW |
13049 | |
13050 | .compare_super = compare_super_imsm, | |
13051 | ||
13052 | .load_super = load_super_imsm, | |
bf5a934a | 13053 | .init_super = init_super_imsm, |
e683ca88 | 13054 | .store_super = store_super_imsm, |
cdddbdbc DW |
13055 | .free_super = free_super_imsm, |
13056 | .match_metadata_desc = match_metadata_desc_imsm, | |
bf5a934a | 13057 | .container_content = container_content_imsm, |
0c21b485 | 13058 | .validate_container = validate_container_imsm, |
cdddbdbc | 13059 | |
fbc42556 JR |
13060 | .add_internal_bitmap = add_internal_bitmap_imsm, |
13061 | .locate_bitmap = locate_bitmap_imsm, | |
13062 | .write_bitmap = write_init_bitmap_imsm, | |
13063 | .set_bitmap = set_bitmap_imsm, | |
13064 | ||
2432ce9b AP |
13065 | .write_init_ppl = write_init_ppl_imsm, |
13066 | .validate_ppl = validate_ppl_imsm, | |
13067 | ||
cdddbdbc | 13068 | .external = 1, |
4cce4069 | 13069 | .name = "imsm", |
845dea95 NB |
13070 | |
13071 | /* for mdmon */ | |
13072 | .open_new = imsm_open_new, | |
ed9d66aa | 13073 | .set_array_state= imsm_set_array_state, |
845dea95 NB |
13074 | .set_disk = imsm_set_disk, |
13075 | .sync_metadata = imsm_sync_metadata, | |
88758e9d | 13076 | .activate_spare = imsm_activate_spare, |
e8319a19 | 13077 | .process_update = imsm_process_update, |
8273f55e | 13078 | .prepare_update = imsm_prepare_update, |
6f50473f | 13079 | .record_bad_block = imsm_record_badblock, |
c07a5a4f | 13080 | .clear_bad_block = imsm_clear_badblock, |
928f1424 | 13081 | .get_bad_blocks = imsm_get_badblocks, |
cdddbdbc | 13082 | }; |