]>
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" |
51006d85 | 23 | #include "sha1.h" |
88c32bb1 | 24 | #include "platform-intel.h" |
cdddbdbc DW |
25 | #include <values.h> |
26 | #include <scsi/sg.h> | |
27 | #include <ctype.h> | |
d665cc31 | 28 | #include <dirent.h> |
cdddbdbc DW |
29 | |
30 | /* MPB == Metadata Parameter Block */ | |
31 | #define MPB_SIGNATURE "Intel Raid ISM Cfg Sig. " | |
32 | #define MPB_SIG_LEN (strlen(MPB_SIGNATURE)) | |
33 | #define MPB_VERSION_RAID0 "1.0.00" | |
34 | #define MPB_VERSION_RAID1 "1.1.00" | |
fe7ed8cb DW |
35 | #define MPB_VERSION_MANY_VOLUMES_PER_ARRAY "1.2.00" |
36 | #define MPB_VERSION_3OR4_DISK_ARRAY "1.2.01" | |
cdddbdbc | 37 | #define MPB_VERSION_RAID5 "1.2.02" |
fe7ed8cb DW |
38 | #define MPB_VERSION_5OR6_DISK_ARRAY "1.2.04" |
39 | #define MPB_VERSION_CNG "1.2.06" | |
40 | #define MPB_VERSION_ATTRIBS "1.3.00" | |
cdddbdbc DW |
41 | #define MAX_SIGNATURE_LENGTH 32 |
42 | #define MAX_RAID_SERIAL_LEN 16 | |
fe7ed8cb | 43 | |
19482bcc AK |
44 | /* supports RAID0 */ |
45 | #define MPB_ATTRIB_RAID0 __cpu_to_le32(0x00000001) | |
46 | /* supports RAID1 */ | |
47 | #define MPB_ATTRIB_RAID1 __cpu_to_le32(0x00000002) | |
48 | /* supports RAID10 */ | |
49 | #define MPB_ATTRIB_RAID10 __cpu_to_le32(0x00000004) | |
50 | /* supports RAID1E */ | |
51 | #define MPB_ATTRIB_RAID1E __cpu_to_le32(0x00000008) | |
52 | /* supports RAID5 */ | |
53 | #define MPB_ATTRIB_RAID5 __cpu_to_le32(0x00000010) | |
54 | /* supports RAID CNG */ | |
55 | #define MPB_ATTRIB_RAIDCNG __cpu_to_le32(0x00000020) | |
56 | /* supports expanded stripe sizes of 256K, 512K and 1MB */ | |
57 | #define MPB_ATTRIB_EXP_STRIPE_SIZE __cpu_to_le32(0x00000040) | |
58 | ||
59 | /* The OROM Support RST Caching of Volumes */ | |
60 | #define MPB_ATTRIB_NVM __cpu_to_le32(0x02000000) | |
61 | /* The OROM supports creating disks greater than 2TB */ | |
62 | #define MPB_ATTRIB_2TB_DISK __cpu_to_le32(0x04000000) | |
63 | /* The OROM supports Bad Block Management */ | |
64 | #define MPB_ATTRIB_BBM __cpu_to_le32(0x08000000) | |
65 | ||
66 | /* THe OROM Supports NVM Caching of Volumes */ | |
67 | #define MPB_ATTRIB_NEVER_USE2 __cpu_to_le32(0x10000000) | |
68 | /* The OROM supports creating volumes greater than 2TB */ | |
69 | #define MPB_ATTRIB_2TB __cpu_to_le32(0x20000000) | |
70 | /* originally for PMP, now it's wasted b/c. Never use this bit! */ | |
71 | #define MPB_ATTRIB_NEVER_USE __cpu_to_le32(0x40000000) | |
72 | /* Verify MPB contents against checksum after reading MPB */ | |
73 | #define MPB_ATTRIB_CHECKSUM_VERIFY __cpu_to_le32(0x80000000) | |
74 | ||
75 | /* Define all supported attributes that have to be accepted by mdadm | |
76 | */ | |
418f9b36 | 77 | #define MPB_ATTRIB_SUPPORTED (MPB_ATTRIB_CHECKSUM_VERIFY | \ |
19482bcc AK |
78 | MPB_ATTRIB_2TB | \ |
79 | MPB_ATTRIB_2TB_DISK | \ | |
80 | MPB_ATTRIB_RAID0 | \ | |
81 | MPB_ATTRIB_RAID1 | \ | |
82 | MPB_ATTRIB_RAID10 | \ | |
83 | MPB_ATTRIB_RAID5 | \ | |
bbab0940 TM |
84 | MPB_ATTRIB_EXP_STRIPE_SIZE | \ |
85 | MPB_ATTRIB_BBM) | |
418f9b36 N |
86 | |
87 | /* Define attributes that are unused but not harmful */ | |
88 | #define MPB_ATTRIB_IGNORED (MPB_ATTRIB_NEVER_USE) | |
fe7ed8cb | 89 | |
8e59f3d8 | 90 | #define MPB_SECTOR_CNT 2210 |
611d9529 MD |
91 | #define IMSM_RESERVED_SECTORS 8192 |
92 | #define NUM_BLOCKS_DIRTY_STRIPE_REGION 2048 | |
979d38be | 93 | #define SECT_PER_MB_SHIFT 11 |
f36a9ecd | 94 | #define MAX_SECTOR_SIZE 4096 |
c2462068 PB |
95 | #define MULTIPLE_PPL_AREA_SIZE_IMSM (1024 * 1024) /* Size of the whole |
96 | * mutliple PPL area | |
97 | */ | |
cdddbdbc DW |
98 | |
99 | /* Disk configuration info. */ | |
100 | #define IMSM_MAX_DEVICES 255 | |
101 | struct imsm_disk { | |
102 | __u8 serial[MAX_RAID_SERIAL_LEN];/* 0xD8 - 0xE7 ascii serial number */ | |
5551b113 | 103 | __u32 total_blocks_lo; /* 0xE8 - 0xEB total blocks lo */ |
cdddbdbc | 104 | __u32 scsi_id; /* 0xEC - 0xEF scsi ID */ |
f2f27e63 DW |
105 | #define SPARE_DISK __cpu_to_le32(0x01) /* Spare */ |
106 | #define CONFIGURED_DISK __cpu_to_le32(0x02) /* Member of some RaidDev */ | |
107 | #define FAILED_DISK __cpu_to_le32(0x04) /* Permanent failure */ | |
2432ce9b | 108 | #define JOURNAL_DISK __cpu_to_le32(0x2000000) /* Device marked as Journaling Drive */ |
cdddbdbc | 109 | __u32 status; /* 0xF0 - 0xF3 */ |
1011e834 | 110 | __u32 owner_cfg_num; /* which config 0,1,2... owns this disk */ |
5551b113 CA |
111 | __u32 total_blocks_hi; /* 0xF4 - 0xF5 total blocks hi */ |
112 | #define IMSM_DISK_FILLERS 3 | |
113 | __u32 filler[IMSM_DISK_FILLERS]; /* 0xF5 - 0x107 MPB_DISK_FILLERS for future expansion */ | |
cdddbdbc DW |
114 | }; |
115 | ||
3b451610 AK |
116 | /* map selector for map managment |
117 | */ | |
238c0a71 AK |
118 | #define MAP_0 0 |
119 | #define MAP_1 1 | |
120 | #define MAP_X -1 | |
3b451610 | 121 | |
cdddbdbc DW |
122 | /* RAID map configuration infos. */ |
123 | struct imsm_map { | |
5551b113 CA |
124 | __u32 pba_of_lba0_lo; /* start address of partition */ |
125 | __u32 blocks_per_member_lo;/* blocks per member */ | |
126 | __u32 num_data_stripes_lo; /* number of data stripes */ | |
cdddbdbc DW |
127 | __u16 blocks_per_strip; |
128 | __u8 map_state; /* Normal, Uninitialized, Degraded, Failed */ | |
129 | #define IMSM_T_STATE_NORMAL 0 | |
130 | #define IMSM_T_STATE_UNINITIALIZED 1 | |
e3bba0e0 DW |
131 | #define IMSM_T_STATE_DEGRADED 2 |
132 | #define IMSM_T_STATE_FAILED 3 | |
cdddbdbc DW |
133 | __u8 raid_level; |
134 | #define IMSM_T_RAID0 0 | |
135 | #define IMSM_T_RAID1 1 | |
136 | #define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */ | |
137 | __u8 num_members; /* number of member disks */ | |
fe7ed8cb DW |
138 | __u8 num_domains; /* number of parity domains */ |
139 | __u8 failed_disk_num; /* valid only when state is degraded */ | |
252d23c0 | 140 | __u8 ddf; |
5551b113 CA |
141 | __u32 pba_of_lba0_hi; |
142 | __u32 blocks_per_member_hi; | |
143 | __u32 num_data_stripes_hi; | |
144 | __u32 filler[4]; /* expansion area */ | |
7eef0453 | 145 | #define IMSM_ORD_REBUILD (1 << 24) |
cdddbdbc | 146 | __u32 disk_ord_tbl[1]; /* disk_ord_tbl[num_members], |
7eef0453 DW |
147 | * top byte contains some flags |
148 | */ | |
cdddbdbc DW |
149 | } __attribute__ ((packed)); |
150 | ||
151 | struct imsm_vol { | |
f8f603f1 | 152 | __u32 curr_migr_unit; |
fe7ed8cb | 153 | __u32 checkpoint_id; /* id to access curr_migr_unit */ |
cdddbdbc | 154 | __u8 migr_state; /* Normal or Migrating */ |
e3bba0e0 DW |
155 | #define MIGR_INIT 0 |
156 | #define MIGR_REBUILD 1 | |
157 | #define MIGR_VERIFY 2 /* analagous to echo check > sync_action */ | |
158 | #define MIGR_GEN_MIGR 3 | |
159 | #define MIGR_STATE_CHANGE 4 | |
1484e727 | 160 | #define MIGR_REPAIR 5 |
cdddbdbc | 161 | __u8 migr_type; /* Initializing, Rebuilding, ... */ |
2432ce9b AP |
162 | #define RAIDVOL_CLEAN 0 |
163 | #define RAIDVOL_DIRTY 1 | |
164 | #define RAIDVOL_DSRECORD_VALID 2 | |
cdddbdbc | 165 | __u8 dirty; |
fe7ed8cb DW |
166 | __u8 fs_state; /* fast-sync state for CnG (0xff == disabled) */ |
167 | __u16 verify_errors; /* number of mismatches */ | |
168 | __u16 bad_blocks; /* number of bad blocks during verify */ | |
169 | __u32 filler[4]; | |
cdddbdbc DW |
170 | struct imsm_map map[1]; |
171 | /* here comes another one if migr_state */ | |
172 | } __attribute__ ((packed)); | |
173 | ||
174 | struct imsm_dev { | |
fe7ed8cb | 175 | __u8 volume[MAX_RAID_SERIAL_LEN]; |
cdddbdbc DW |
176 | __u32 size_low; |
177 | __u32 size_high; | |
fe7ed8cb DW |
178 | #define DEV_BOOTABLE __cpu_to_le32(0x01) |
179 | #define DEV_BOOT_DEVICE __cpu_to_le32(0x02) | |
180 | #define DEV_READ_COALESCING __cpu_to_le32(0x04) | |
181 | #define DEV_WRITE_COALESCING __cpu_to_le32(0x08) | |
182 | #define DEV_LAST_SHUTDOWN_DIRTY __cpu_to_le32(0x10) | |
183 | #define DEV_HIDDEN_AT_BOOT __cpu_to_le32(0x20) | |
184 | #define DEV_CURRENTLY_HIDDEN __cpu_to_le32(0x40) | |
185 | #define DEV_VERIFY_AND_FIX __cpu_to_le32(0x80) | |
186 | #define DEV_MAP_STATE_UNINIT __cpu_to_le32(0x100) | |
187 | #define DEV_NO_AUTO_RECOVERY __cpu_to_le32(0x200) | |
188 | #define DEV_CLONE_N_GO __cpu_to_le32(0x400) | |
189 | #define DEV_CLONE_MAN_SYNC __cpu_to_le32(0x800) | |
190 | #define DEV_CNG_MASTER_DISK_NUM __cpu_to_le32(0x1000) | |
cdddbdbc DW |
191 | __u32 status; /* Persistent RaidDev status */ |
192 | __u32 reserved_blocks; /* Reserved blocks at beginning of volume */ | |
fe7ed8cb DW |
193 | __u8 migr_priority; |
194 | __u8 num_sub_vols; | |
195 | __u8 tid; | |
196 | __u8 cng_master_disk; | |
197 | __u16 cache_policy; | |
198 | __u8 cng_state; | |
199 | __u8 cng_sub_state; | |
2432ce9b AP |
200 | __u16 my_vol_raid_dev_num; /* Used in Unique volume Id for this RaidDev */ |
201 | ||
202 | /* NVM_EN */ | |
203 | __u8 nv_cache_mode; | |
204 | __u8 nv_cache_flags; | |
205 | ||
206 | /* Unique Volume Id of the NvCache Volume associated with this volume */ | |
207 | __u32 nvc_vol_orig_family_num; | |
208 | __u16 nvc_vol_raid_dev_num; | |
209 | ||
210 | #define RWH_OFF 0 | |
211 | #define RWH_DISTRIBUTED 1 | |
212 | #define RWH_JOURNALING_DRIVE 2 | |
c2462068 PB |
213 | #define RWH_MULTIPLE_DISTRIBUTED 3 |
214 | #define RWH_MULTIPLE_PPLS_JOURNALING_DRIVE 4 | |
215 | #define RWH_MULTIPLE_OFF 5 | |
2432ce9b AP |
216 | __u8 rwh_policy; /* Raid Write Hole Policy */ |
217 | __u8 jd_serial[MAX_RAID_SERIAL_LEN]; /* Journal Drive serial number */ | |
218 | __u8 filler1; | |
219 | ||
220 | #define IMSM_DEV_FILLERS 3 | |
cdddbdbc DW |
221 | __u32 filler[IMSM_DEV_FILLERS]; |
222 | struct imsm_vol vol; | |
223 | } __attribute__ ((packed)); | |
224 | ||
225 | struct imsm_super { | |
226 | __u8 sig[MAX_SIGNATURE_LENGTH]; /* 0x00 - 0x1F */ | |
227 | __u32 check_sum; /* 0x20 - 0x23 MPB Checksum */ | |
228 | __u32 mpb_size; /* 0x24 - 0x27 Size of MPB */ | |
229 | __u32 family_num; /* 0x28 - 0x2B Checksum from first time this config was written */ | |
230 | __u32 generation_num; /* 0x2C - 0x2F Incremented each time this array's MPB is written */ | |
604b746f JD |
231 | __u32 error_log_size; /* 0x30 - 0x33 in bytes */ |
232 | __u32 attributes; /* 0x34 - 0x37 */ | |
cdddbdbc DW |
233 | __u8 num_disks; /* 0x38 Number of configured disks */ |
234 | __u8 num_raid_devs; /* 0x39 Number of configured volumes */ | |
604b746f JD |
235 | __u8 error_log_pos; /* 0x3A */ |
236 | __u8 fill[1]; /* 0x3B */ | |
237 | __u32 cache_size; /* 0x3c - 0x40 in mb */ | |
238 | __u32 orig_family_num; /* 0x40 - 0x43 original family num */ | |
239 | __u32 pwr_cycle_count; /* 0x44 - 0x47 simulated power cycle count for array */ | |
240 | __u32 bbm_log_size; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */ | |
2a24dc1b PB |
241 | __u16 num_raid_devs_created; /* 0x4C - 0x4D Used for generating unique |
242 | * volume IDs for raid_dev created in this array | |
243 | * (starts at 1) | |
244 | */ | |
245 | __u16 filler1; /* 0x4E - 0x4F */ | |
246 | #define IMSM_FILLERS 34 | |
247 | __u32 filler[IMSM_FILLERS]; /* 0x50 - 0xD7 RAID_MPB_FILLERS */ | |
cdddbdbc DW |
248 | struct imsm_disk disk[1]; /* 0xD8 diskTbl[numDisks] */ |
249 | /* here comes imsm_dev[num_raid_devs] */ | |
604b746f | 250 | /* here comes BBM logs */ |
cdddbdbc DW |
251 | } __attribute__ ((packed)); |
252 | ||
604b746f | 253 | #define BBM_LOG_MAX_ENTRIES 254 |
8d67477f TM |
254 | #define BBM_LOG_MAX_LBA_ENTRY_VAL 256 /* Represents 256 LBAs */ |
255 | #define BBM_LOG_SIGNATURE 0xabadb10c | |
256 | ||
257 | struct bbm_log_block_addr { | |
258 | __u16 w1; | |
259 | __u32 dw1; | |
260 | } __attribute__ ((__packed__)); | |
604b746f JD |
261 | |
262 | struct bbm_log_entry { | |
8d67477f TM |
263 | __u8 marked_count; /* Number of blocks marked - 1 */ |
264 | __u8 disk_ordinal; /* Disk entry within the imsm_super */ | |
265 | struct bbm_log_block_addr defective_block_start; | |
604b746f JD |
266 | } __attribute__ ((__packed__)); |
267 | ||
268 | struct bbm_log { | |
269 | __u32 signature; /* 0xABADB10C */ | |
270 | __u32 entry_count; | |
8d67477f | 271 | struct bbm_log_entry marked_block_entries[BBM_LOG_MAX_ENTRIES]; |
604b746f JD |
272 | } __attribute__ ((__packed__)); |
273 | ||
cdddbdbc | 274 | static char *map_state_str[] = { "normal", "uninitialized", "degraded", "failed" }; |
cdddbdbc | 275 | |
b53bfba6 TM |
276 | #define BLOCKS_PER_KB (1024/512) |
277 | ||
8e59f3d8 AK |
278 | #define RAID_DISK_RESERVED_BLOCKS_IMSM_HI 2209 |
279 | ||
280 | #define GEN_MIGR_AREA_SIZE 2048 /* General Migration Copy Area size in blocks */ | |
281 | ||
de44e46f PB |
282 | #define MIGR_REC_BUF_SECTORS 1 /* size of migr_record i/o buffer in sectors */ |
283 | #define MIGR_REC_SECTOR_POSITION 1 /* migr_record position offset on disk, | |
284 | * MIGR_REC_BUF_SECTORS <= MIGR_REC_SECTOR_POS | |
17a4eaf9 AK |
285 | */ |
286 | ||
8e59f3d8 AK |
287 | #define UNIT_SRC_NORMAL 0 /* Source data for curr_migr_unit must |
288 | * be recovered using srcMap */ | |
289 | #define UNIT_SRC_IN_CP_AREA 1 /* Source data for curr_migr_unit has | |
290 | * already been migrated and must | |
291 | * be recovered from checkpoint area */ | |
2432ce9b | 292 | |
c2462068 | 293 | #define PPL_ENTRY_SPACE (128 * 1024) /* Size of single PPL, without the header */ |
2432ce9b | 294 | |
8e59f3d8 AK |
295 | struct migr_record { |
296 | __u32 rec_status; /* Status used to determine how to restart | |
297 | * migration in case it aborts | |
298 | * in some fashion */ | |
9f421827 | 299 | __u32 curr_migr_unit_lo; /* 0..numMigrUnits-1 */ |
8e59f3d8 AK |
300 | __u32 family_num; /* Family number of MPB |
301 | * containing the RaidDev | |
302 | * that is migrating */ | |
303 | __u32 ascending_migr; /* True if migrating in increasing | |
304 | * order of lbas */ | |
305 | __u32 blocks_per_unit; /* Num disk blocks per unit of operation */ | |
306 | __u32 dest_depth_per_unit; /* Num member blocks each destMap | |
307 | * member disk | |
308 | * advances per unit-of-operation */ | |
9f421827 PB |
309 | __u32 ckpt_area_pba_lo; /* Pba of first block of ckpt copy area */ |
310 | __u32 dest_1st_member_lba_lo; /* First member lba on first | |
311 | * stripe of destination */ | |
312 | __u32 num_migr_units_lo; /* Total num migration units-of-op */ | |
8e59f3d8 AK |
313 | __u32 post_migr_vol_cap; /* Size of volume after |
314 | * migration completes */ | |
315 | __u32 post_migr_vol_cap_hi; /* Expansion space for LBA64 */ | |
316 | __u32 ckpt_read_disk_num; /* Which member disk in destSubMap[0] the | |
317 | * migration ckpt record was read from | |
318 | * (for recovered migrations) */ | |
9f421827 PB |
319 | __u32 curr_migr_unit_hi; /* 0..numMigrUnits-1 high order 32 bits */ |
320 | __u32 ckpt_area_pba_hi; /* Pba of first block of ckpt copy area | |
321 | * high order 32 bits */ | |
322 | __u32 dest_1st_member_lba_hi; /* First member lba on first stripe of | |
323 | * destination - high order 32 bits */ | |
324 | __u32 num_migr_units_hi; /* Total num migration units-of-op | |
325 | * high order 32 bits */ | |
8e59f3d8 AK |
326 | } __attribute__ ((__packed__)); |
327 | ||
ec50f7b6 LM |
328 | struct md_list { |
329 | /* usage marker: | |
330 | * 1: load metadata | |
331 | * 2: metadata does not match | |
332 | * 4: already checked | |
333 | */ | |
334 | int used; | |
335 | char *devname; | |
336 | int found; | |
337 | int container; | |
338 | dev_t st_rdev; | |
339 | struct md_list *next; | |
340 | }; | |
341 | ||
e7b84f9d | 342 | #define pr_vrb(fmt, arg...) (void) (verbose && pr_err(fmt, ##arg)) |
ec50f7b6 | 343 | |
1484e727 DW |
344 | static __u8 migr_type(struct imsm_dev *dev) |
345 | { | |
346 | if (dev->vol.migr_type == MIGR_VERIFY && | |
347 | dev->status & DEV_VERIFY_AND_FIX) | |
348 | return MIGR_REPAIR; | |
349 | else | |
350 | return dev->vol.migr_type; | |
351 | } | |
352 | ||
353 | static void set_migr_type(struct imsm_dev *dev, __u8 migr_type) | |
354 | { | |
355 | /* for compatibility with older oroms convert MIGR_REPAIR, into | |
356 | * MIGR_VERIFY w/ DEV_VERIFY_AND_FIX status | |
357 | */ | |
358 | if (migr_type == MIGR_REPAIR) { | |
359 | dev->vol.migr_type = MIGR_VERIFY; | |
360 | dev->status |= DEV_VERIFY_AND_FIX; | |
361 | } else { | |
362 | dev->vol.migr_type = migr_type; | |
363 | dev->status &= ~DEV_VERIFY_AND_FIX; | |
364 | } | |
365 | } | |
366 | ||
f36a9ecd | 367 | static unsigned int sector_count(__u32 bytes, unsigned int sector_size) |
cdddbdbc | 368 | { |
f36a9ecd | 369 | return ROUND_UP(bytes, sector_size) / sector_size; |
87eb16df | 370 | } |
cdddbdbc | 371 | |
f36a9ecd PB |
372 | static unsigned int mpb_sectors(struct imsm_super *mpb, |
373 | unsigned int sector_size) | |
87eb16df | 374 | { |
f36a9ecd | 375 | return sector_count(__le32_to_cpu(mpb->mpb_size), sector_size); |
cdddbdbc DW |
376 | } |
377 | ||
ba2de7ba DW |
378 | struct intel_dev { |
379 | struct imsm_dev *dev; | |
380 | struct intel_dev *next; | |
f21e18ca | 381 | unsigned index; |
ba2de7ba DW |
382 | }; |
383 | ||
88654014 LM |
384 | struct intel_hba { |
385 | enum sys_dev_type type; | |
386 | char *path; | |
387 | char *pci_id; | |
388 | struct intel_hba *next; | |
389 | }; | |
390 | ||
1a64be56 LM |
391 | enum action { |
392 | DISK_REMOVE = 1, | |
393 | DISK_ADD | |
394 | }; | |
cdddbdbc DW |
395 | /* internal representation of IMSM metadata */ |
396 | struct intel_super { | |
397 | union { | |
949c47a0 DW |
398 | void *buf; /* O_DIRECT buffer for reading/writing metadata */ |
399 | struct imsm_super *anchor; /* immovable parameters */ | |
cdddbdbc | 400 | }; |
8e59f3d8 AK |
401 | union { |
402 | void *migr_rec_buf; /* buffer for I/O operations */ | |
403 | struct migr_record *migr_rec; /* migration record */ | |
404 | }; | |
51d83f5d AK |
405 | int clean_migration_record_by_mdmon; /* when reshape is switched to next |
406 | array, it indicates that mdmon is allowed to clean migration | |
407 | record */ | |
949c47a0 | 408 | size_t len; /* size of the 'buf' allocation */ |
bbab0940 | 409 | size_t extra_space; /* extra space in 'buf' that is not used yet */ |
4d7b1503 DW |
410 | void *next_buf; /* for realloc'ing buf from the manager */ |
411 | size_t next_len; | |
c2c087e6 | 412 | int updates_pending; /* count of pending updates for mdmon */ |
bf5a934a | 413 | int current_vol; /* index of raid device undergoing creation */ |
5551b113 | 414 | unsigned long long create_offset; /* common start for 'current_vol' */ |
148acb7b | 415 | __u32 random; /* random data for seeding new family numbers */ |
ba2de7ba | 416 | struct intel_dev *devlist; |
fa7bb6f8 | 417 | unsigned int sector_size; /* sector size of used member drives */ |
cdddbdbc DW |
418 | struct dl { |
419 | struct dl *next; | |
420 | int index; | |
421 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
422 | int major, minor; | |
423 | char *devname; | |
b9f594fe | 424 | struct imsm_disk disk; |
cdddbdbc | 425 | int fd; |
0dcecb2e DW |
426 | int extent_cnt; |
427 | struct extent *e; /* for determining freespace @ create */ | |
efb30e7f | 428 | int raiddisk; /* slot to fill in autolayout */ |
1a64be56 | 429 | enum action action; |
ca0748fa | 430 | } *disks, *current_disk; |
1a64be56 LM |
431 | struct dl *disk_mgmt_list; /* list of disks to add/remove while mdmon |
432 | active */ | |
47ee5a45 | 433 | struct dl *missing; /* disks removed while we weren't looking */ |
43dad3d6 | 434 | struct bbm_log *bbm_log; |
88654014 | 435 | struct intel_hba *hba; /* device path of the raid controller for this metadata */ |
88c32bb1 | 436 | const struct imsm_orom *orom; /* platform firmware support */ |
a2b97981 | 437 | struct intel_super *next; /* (temp) list for disambiguating family_num */ |
928f1424 | 438 | struct md_bb bb; /* memory for get_bad_blocks call */ |
a2b97981 DW |
439 | }; |
440 | ||
441 | struct intel_disk { | |
442 | struct imsm_disk disk; | |
443 | #define IMSM_UNKNOWN_OWNER (-1) | |
444 | int owner; | |
445 | struct intel_disk *next; | |
cdddbdbc DW |
446 | }; |
447 | ||
c2c087e6 DW |
448 | struct extent { |
449 | unsigned long long start, size; | |
450 | }; | |
451 | ||
694575e7 KW |
452 | /* definitions of reshape process types */ |
453 | enum imsm_reshape_type { | |
454 | CH_TAKEOVER, | |
b5347799 | 455 | CH_MIGRATION, |
7abc9871 | 456 | CH_ARRAY_SIZE, |
694575e7 KW |
457 | }; |
458 | ||
88758e9d DW |
459 | /* definition of messages passed to imsm_process_update */ |
460 | enum imsm_update_type { | |
461 | update_activate_spare, | |
8273f55e | 462 | update_create_array, |
33414a01 | 463 | update_kill_array, |
aa534678 | 464 | update_rename_array, |
1a64be56 | 465 | update_add_remove_disk, |
78b10e66 | 466 | update_reshape_container_disks, |
48c5303a | 467 | update_reshape_migration, |
2d40f3a1 AK |
468 | update_takeover, |
469 | update_general_migration_checkpoint, | |
f3871fdc | 470 | update_size_change, |
bbab0940 | 471 | update_prealloc_badblocks_mem, |
e6e9dd3f | 472 | update_rwh_policy, |
88758e9d DW |
473 | }; |
474 | ||
475 | struct imsm_update_activate_spare { | |
476 | enum imsm_update_type type; | |
d23fe947 | 477 | struct dl *dl; |
88758e9d DW |
478 | int slot; |
479 | int array; | |
480 | struct imsm_update_activate_spare *next; | |
481 | }; | |
482 | ||
78b10e66 | 483 | struct geo_params { |
4dd2df09 | 484 | char devnm[32]; |
78b10e66 | 485 | char *dev_name; |
d04f65f4 | 486 | unsigned long long size; |
78b10e66 N |
487 | int level; |
488 | int layout; | |
489 | int chunksize; | |
490 | int raid_disks; | |
491 | }; | |
492 | ||
bb025c2f KW |
493 | enum takeover_direction { |
494 | R10_TO_R0, | |
495 | R0_TO_R10 | |
496 | }; | |
497 | struct imsm_update_takeover { | |
498 | enum imsm_update_type type; | |
499 | int subarray; | |
500 | enum takeover_direction direction; | |
501 | }; | |
78b10e66 N |
502 | |
503 | struct imsm_update_reshape { | |
504 | enum imsm_update_type type; | |
505 | int old_raid_disks; | |
506 | int new_raid_disks; | |
48c5303a PC |
507 | |
508 | int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */ | |
509 | }; | |
510 | ||
511 | struct imsm_update_reshape_migration { | |
512 | enum imsm_update_type type; | |
513 | int old_raid_disks; | |
514 | int new_raid_disks; | |
515 | /* fields for array migration changes | |
516 | */ | |
517 | int subdev; | |
518 | int new_level; | |
519 | int new_layout; | |
4bba0439 | 520 | int new_chunksize; |
48c5303a | 521 | |
d195167d | 522 | int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */ |
78b10e66 N |
523 | }; |
524 | ||
f3871fdc AK |
525 | struct imsm_update_size_change { |
526 | enum imsm_update_type type; | |
527 | int subdev; | |
528 | long long new_size; | |
529 | }; | |
530 | ||
2d40f3a1 AK |
531 | struct imsm_update_general_migration_checkpoint { |
532 | enum imsm_update_type type; | |
533 | __u32 curr_migr_unit; | |
534 | }; | |
535 | ||
54c2c1ea DW |
536 | struct disk_info { |
537 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
538 | }; | |
539 | ||
8273f55e DW |
540 | struct imsm_update_create_array { |
541 | enum imsm_update_type type; | |
8273f55e | 542 | int dev_idx; |
6a3e913e | 543 | struct imsm_dev dev; |
8273f55e DW |
544 | }; |
545 | ||
33414a01 DW |
546 | struct imsm_update_kill_array { |
547 | enum imsm_update_type type; | |
548 | int dev_idx; | |
549 | }; | |
550 | ||
aa534678 DW |
551 | struct imsm_update_rename_array { |
552 | enum imsm_update_type type; | |
553 | __u8 name[MAX_RAID_SERIAL_LEN]; | |
554 | int dev_idx; | |
555 | }; | |
556 | ||
1a64be56 | 557 | struct imsm_update_add_remove_disk { |
43dad3d6 DW |
558 | enum imsm_update_type type; |
559 | }; | |
560 | ||
bbab0940 TM |
561 | struct imsm_update_prealloc_bb_mem { |
562 | enum imsm_update_type type; | |
563 | }; | |
564 | ||
e6e9dd3f AP |
565 | struct imsm_update_rwh_policy { |
566 | enum imsm_update_type type; | |
567 | int new_policy; | |
568 | int dev_idx; | |
569 | }; | |
570 | ||
88654014 LM |
571 | static const char *_sys_dev_type[] = { |
572 | [SYS_DEV_UNKNOWN] = "Unknown", | |
573 | [SYS_DEV_SAS] = "SAS", | |
614902f6 | 574 | [SYS_DEV_SATA] = "SATA", |
60f0f54d PB |
575 | [SYS_DEV_NVME] = "NVMe", |
576 | [SYS_DEV_VMD] = "VMD" | |
88654014 LM |
577 | }; |
578 | ||
579 | const char *get_sys_dev_type(enum sys_dev_type type) | |
580 | { | |
581 | if (type >= SYS_DEV_MAX) | |
582 | type = SYS_DEV_UNKNOWN; | |
583 | ||
584 | return _sys_dev_type[type]; | |
585 | } | |
586 | ||
587 | static struct intel_hba * alloc_intel_hba(struct sys_dev *device) | |
588 | { | |
503975b9 N |
589 | struct intel_hba *result = xmalloc(sizeof(*result)); |
590 | ||
591 | result->type = device->type; | |
592 | result->path = xstrdup(device->path); | |
593 | result->next = NULL; | |
594 | if (result->path && (result->pci_id = strrchr(result->path, '/')) != NULL) | |
595 | result->pci_id++; | |
596 | ||
88654014 LM |
597 | return result; |
598 | } | |
599 | ||
600 | static struct intel_hba * find_intel_hba(struct intel_hba *hba, struct sys_dev *device) | |
601 | { | |
594dc1b8 JS |
602 | struct intel_hba *result; |
603 | ||
88654014 LM |
604 | for (result = hba; result; result = result->next) { |
605 | if (result->type == device->type && strcmp(result->path, device->path) == 0) | |
606 | break; | |
607 | } | |
608 | return result; | |
609 | } | |
610 | ||
b4cf4cba | 611 | static int attach_hba_to_super(struct intel_super *super, struct sys_dev *device) |
88654014 LM |
612 | { |
613 | struct intel_hba *hba; | |
614 | ||
615 | /* check if disk attached to Intel HBA */ | |
616 | hba = find_intel_hba(super->hba, device); | |
617 | if (hba != NULL) | |
618 | return 1; | |
619 | /* Check if HBA is already attached to super */ | |
620 | if (super->hba == NULL) { | |
621 | super->hba = alloc_intel_hba(device); | |
622 | return 1; | |
6b781d33 AP |
623 | } |
624 | ||
625 | hba = super->hba; | |
626 | /* Intel metadata allows for all disks attached to the same type HBA. | |
614902f6 | 627 | * Do not support HBA types mixing |
6b781d33 AP |
628 | */ |
629 | if (device->type != hba->type) | |
88654014 | 630 | return 2; |
6b781d33 AP |
631 | |
632 | /* Multiple same type HBAs can be used if they share the same OROM */ | |
633 | const struct imsm_orom *device_orom = get_orom_by_device_id(device->dev_id); | |
634 | ||
635 | if (device_orom != super->orom) | |
636 | return 2; | |
637 | ||
638 | while (hba->next) | |
639 | hba = hba->next; | |
640 | ||
641 | hba->next = alloc_intel_hba(device); | |
642 | return 1; | |
88654014 LM |
643 | } |
644 | ||
645 | static struct sys_dev* find_disk_attached_hba(int fd, const char *devname) | |
646 | { | |
9bc4ae77 | 647 | struct sys_dev *list, *elem; |
88654014 LM |
648 | char *disk_path; |
649 | ||
650 | if ((list = find_intel_devices()) == NULL) | |
651 | return 0; | |
652 | ||
653 | if (fd < 0) | |
654 | disk_path = (char *) devname; | |
655 | else | |
656 | disk_path = diskfd_to_devpath(fd); | |
657 | ||
9bc4ae77 | 658 | if (!disk_path) |
88654014 | 659 | return 0; |
88654014 | 660 | |
9bc4ae77 N |
661 | for (elem = list; elem; elem = elem->next) |
662 | if (path_attached_to_hba(disk_path, elem->path)) | |
88654014 | 663 | return elem; |
9bc4ae77 | 664 | |
88654014 LM |
665 | if (disk_path != devname) |
666 | free(disk_path); | |
88654014 LM |
667 | |
668 | return NULL; | |
669 | } | |
670 | ||
d424212e N |
671 | static int find_intel_hba_capability(int fd, struct intel_super *super, |
672 | char *devname); | |
f2f5c343 | 673 | |
cdddbdbc DW |
674 | static struct supertype *match_metadata_desc_imsm(char *arg) |
675 | { | |
676 | struct supertype *st; | |
677 | ||
678 | if (strcmp(arg, "imsm") != 0 && | |
679 | strcmp(arg, "default") != 0 | |
680 | ) | |
681 | return NULL; | |
682 | ||
503975b9 | 683 | st = xcalloc(1, sizeof(*st)); |
cdddbdbc DW |
684 | st->ss = &super_imsm; |
685 | st->max_devs = IMSM_MAX_DEVICES; | |
686 | st->minor_version = 0; | |
687 | st->sb = NULL; | |
688 | return st; | |
689 | } | |
690 | ||
cdddbdbc DW |
691 | static __u8 *get_imsm_version(struct imsm_super *mpb) |
692 | { | |
693 | return &mpb->sig[MPB_SIG_LEN]; | |
694 | } | |
695 | ||
949c47a0 DW |
696 | /* retrieve a disk directly from the anchor when the anchor is known to be |
697 | * up-to-date, currently only at load time | |
698 | */ | |
699 | static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index) | |
cdddbdbc | 700 | { |
949c47a0 | 701 | if (index >= mpb->num_disks) |
cdddbdbc DW |
702 | return NULL; |
703 | return &mpb->disk[index]; | |
704 | } | |
705 | ||
95d07a2c LM |
706 | /* retrieve the disk description based on a index of the disk |
707 | * in the sub-array | |
708 | */ | |
709 | static struct dl *get_imsm_dl_disk(struct intel_super *super, __u8 index) | |
949c47a0 | 710 | { |
b9f594fe DW |
711 | struct dl *d; |
712 | ||
713 | for (d = super->disks; d; d = d->next) | |
714 | if (d->index == index) | |
95d07a2c LM |
715 | return d; |
716 | ||
717 | return NULL; | |
718 | } | |
719 | /* retrieve a disk from the parsed metadata */ | |
720 | static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index) | |
721 | { | |
722 | struct dl *dl; | |
723 | ||
724 | dl = get_imsm_dl_disk(super, index); | |
725 | if (dl) | |
726 | return &dl->disk; | |
727 | ||
b9f594fe | 728 | return NULL; |
949c47a0 DW |
729 | } |
730 | ||
731 | /* generate a checksum directly from the anchor when the anchor is known to be | |
732 | * up-to-date, currently only at load or write_super after coalescing | |
733 | */ | |
734 | static __u32 __gen_imsm_checksum(struct imsm_super *mpb) | |
cdddbdbc DW |
735 | { |
736 | __u32 end = mpb->mpb_size / sizeof(end); | |
737 | __u32 *p = (__u32 *) mpb; | |
738 | __u32 sum = 0; | |
739 | ||
5d500228 N |
740 | while (end--) { |
741 | sum += __le32_to_cpu(*p); | |
97f734fd N |
742 | p++; |
743 | } | |
cdddbdbc | 744 | |
5d500228 | 745 | return sum - __le32_to_cpu(mpb->check_sum); |
cdddbdbc DW |
746 | } |
747 | ||
a965f303 DW |
748 | static size_t sizeof_imsm_map(struct imsm_map *map) |
749 | { | |
750 | return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1); | |
751 | } | |
752 | ||
753 | struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map) | |
cdddbdbc | 754 | { |
5e7b0330 AK |
755 | /* A device can have 2 maps if it is in the middle of a migration. |
756 | * If second_map is: | |
238c0a71 AK |
757 | * MAP_0 - we return the first map |
758 | * MAP_1 - we return the second map if it exists, else NULL | |
759 | * MAP_X - we return the second map if it exists, else the first | |
5e7b0330 | 760 | */ |
a965f303 | 761 | struct imsm_map *map = &dev->vol.map[0]; |
9535fc47 | 762 | struct imsm_map *map2 = NULL; |
a965f303 | 763 | |
9535fc47 AK |
764 | if (dev->vol.migr_state) |
765 | map2 = (void *)map + sizeof_imsm_map(map); | |
a965f303 | 766 | |
9535fc47 | 767 | switch (second_map) { |
3b451610 | 768 | case MAP_0: |
9535fc47 | 769 | break; |
3b451610 | 770 | case MAP_1: |
9535fc47 AK |
771 | map = map2; |
772 | break; | |
238c0a71 | 773 | case MAP_X: |
9535fc47 AK |
774 | if (map2) |
775 | map = map2; | |
776 | break; | |
9535fc47 AK |
777 | default: |
778 | map = NULL; | |
779 | } | |
780 | return map; | |
5e7b0330 | 781 | |
a965f303 | 782 | } |
cdddbdbc | 783 | |
3393c6af DW |
784 | /* return the size of the device. |
785 | * migr_state increases the returned size if map[0] were to be duplicated | |
786 | */ | |
787 | static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state) | |
a965f303 DW |
788 | { |
789 | size_t size = sizeof(*dev) - sizeof(struct imsm_map) + | |
238c0a71 | 790 | sizeof_imsm_map(get_imsm_map(dev, MAP_0)); |
cdddbdbc DW |
791 | |
792 | /* migrating means an additional map */ | |
a965f303 | 793 | if (dev->vol.migr_state) |
238c0a71 | 794 | size += sizeof_imsm_map(get_imsm_map(dev, MAP_1)); |
3393c6af | 795 | else if (migr_state) |
238c0a71 | 796 | size += sizeof_imsm_map(get_imsm_map(dev, MAP_0)); |
cdddbdbc DW |
797 | |
798 | return size; | |
799 | } | |
800 | ||
54c2c1ea DW |
801 | /* retrieve disk serial number list from a metadata update */ |
802 | static struct disk_info *get_disk_info(struct imsm_update_create_array *update) | |
803 | { | |
804 | void *u = update; | |
805 | struct disk_info *inf; | |
806 | ||
807 | inf = u + sizeof(*update) - sizeof(struct imsm_dev) + | |
808 | sizeof_imsm_dev(&update->dev, 0); | |
809 | ||
810 | return inf; | |
811 | } | |
54c2c1ea | 812 | |
949c47a0 | 813 | static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index) |
cdddbdbc DW |
814 | { |
815 | int offset; | |
816 | int i; | |
817 | void *_mpb = mpb; | |
818 | ||
949c47a0 | 819 | if (index >= mpb->num_raid_devs) |
cdddbdbc DW |
820 | return NULL; |
821 | ||
822 | /* devices start after all disks */ | |
823 | offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb; | |
824 | ||
825 | for (i = 0; i <= index; i++) | |
826 | if (i == index) | |
827 | return _mpb + offset; | |
828 | else | |
3393c6af | 829 | offset += sizeof_imsm_dev(_mpb + offset, 0); |
cdddbdbc DW |
830 | |
831 | return NULL; | |
832 | } | |
833 | ||
949c47a0 DW |
834 | static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index) |
835 | { | |
ba2de7ba DW |
836 | struct intel_dev *dv; |
837 | ||
949c47a0 DW |
838 | if (index >= super->anchor->num_raid_devs) |
839 | return NULL; | |
ba2de7ba DW |
840 | for (dv = super->devlist; dv; dv = dv->next) |
841 | if (dv->index == index) | |
842 | return dv->dev; | |
843 | return NULL; | |
949c47a0 DW |
844 | } |
845 | ||
8d67477f TM |
846 | static inline unsigned long long __le48_to_cpu(const struct bbm_log_block_addr |
847 | *addr) | |
848 | { | |
849 | return ((((__u64)__le32_to_cpu(addr->dw1)) << 16) | | |
850 | __le16_to_cpu(addr->w1)); | |
851 | } | |
852 | ||
853 | static inline struct bbm_log_block_addr __cpu_to_le48(unsigned long long sec) | |
854 | { | |
855 | struct bbm_log_block_addr addr; | |
856 | ||
857 | addr.w1 = __cpu_to_le16((__u16)(sec & 0xffff)); | |
858 | addr.dw1 = __cpu_to_le32((__u32)(sec >> 16) & 0xffffffff); | |
859 | return addr; | |
860 | } | |
861 | ||
8d67477f TM |
862 | /* get size of the bbm log */ |
863 | static __u32 get_imsm_bbm_log_size(struct bbm_log *log) | |
864 | { | |
865 | if (!log || log->entry_count == 0) | |
866 | return 0; | |
867 | ||
868 | return sizeof(log->signature) + | |
869 | sizeof(log->entry_count) + | |
870 | log->entry_count * sizeof(struct bbm_log_entry); | |
871 | } | |
6f50473f TM |
872 | |
873 | /* check if bad block is not partially stored in bbm log */ | |
874 | static int is_stored_in_bbm(struct bbm_log *log, const __u8 idx, const unsigned | |
875 | long long sector, const int length, __u32 *pos) | |
876 | { | |
877 | __u32 i; | |
878 | ||
879 | for (i = *pos; i < log->entry_count; i++) { | |
880 | struct bbm_log_entry *entry = &log->marked_block_entries[i]; | |
881 | unsigned long long bb_start; | |
882 | unsigned long long bb_end; | |
883 | ||
884 | bb_start = __le48_to_cpu(&entry->defective_block_start); | |
885 | bb_end = bb_start + (entry->marked_count + 1); | |
886 | ||
887 | if ((entry->disk_ordinal == idx) && (bb_start >= sector) && | |
888 | (bb_end <= sector + length)) { | |
889 | *pos = i; | |
890 | return 1; | |
891 | } | |
892 | } | |
893 | return 0; | |
894 | } | |
895 | ||
896 | /* record new bad block in bbm log */ | |
897 | static int record_new_badblock(struct bbm_log *log, const __u8 idx, unsigned | |
898 | long long sector, int length) | |
899 | { | |
900 | int new_bb = 0; | |
901 | __u32 pos = 0; | |
902 | struct bbm_log_entry *entry = NULL; | |
903 | ||
904 | while (is_stored_in_bbm(log, idx, sector, length, &pos)) { | |
905 | struct bbm_log_entry *e = &log->marked_block_entries[pos]; | |
906 | ||
907 | if ((e->marked_count + 1 == BBM_LOG_MAX_LBA_ENTRY_VAL) && | |
908 | (__le48_to_cpu(&e->defective_block_start) == sector)) { | |
909 | sector += BBM_LOG_MAX_LBA_ENTRY_VAL; | |
910 | length -= BBM_LOG_MAX_LBA_ENTRY_VAL; | |
911 | pos = pos + 1; | |
912 | continue; | |
913 | } | |
914 | entry = e; | |
915 | break; | |
916 | } | |
917 | ||
918 | if (entry) { | |
919 | int cnt = (length <= BBM_LOG_MAX_LBA_ENTRY_VAL) ? length : | |
920 | BBM_LOG_MAX_LBA_ENTRY_VAL; | |
921 | entry->defective_block_start = __cpu_to_le48(sector); | |
922 | entry->marked_count = cnt - 1; | |
923 | if (cnt == length) | |
924 | return 1; | |
925 | sector += cnt; | |
926 | length -= cnt; | |
927 | } | |
928 | ||
929 | new_bb = ROUND_UP(length, BBM_LOG_MAX_LBA_ENTRY_VAL) / | |
930 | BBM_LOG_MAX_LBA_ENTRY_VAL; | |
931 | if (log->entry_count + new_bb > BBM_LOG_MAX_ENTRIES) | |
932 | return 0; | |
933 | ||
934 | while (length > 0) { | |
935 | int cnt = (length <= BBM_LOG_MAX_LBA_ENTRY_VAL) ? length : | |
936 | BBM_LOG_MAX_LBA_ENTRY_VAL; | |
937 | struct bbm_log_entry *entry = | |
938 | &log->marked_block_entries[log->entry_count]; | |
939 | ||
940 | entry->defective_block_start = __cpu_to_le48(sector); | |
941 | entry->marked_count = cnt - 1; | |
942 | entry->disk_ordinal = idx; | |
943 | ||
944 | sector += cnt; | |
945 | length -= cnt; | |
946 | ||
947 | log->entry_count++; | |
948 | } | |
949 | ||
950 | return new_bb; | |
951 | } | |
c07a5a4f | 952 | |
4c9e8c1e TM |
953 | /* clear all bad blocks for given disk */ |
954 | static void clear_disk_badblocks(struct bbm_log *log, const __u8 idx) | |
955 | { | |
956 | __u32 i = 0; | |
957 | ||
958 | while (i < log->entry_count) { | |
959 | struct bbm_log_entry *entries = log->marked_block_entries; | |
960 | ||
961 | if (entries[i].disk_ordinal == idx) { | |
962 | if (i < log->entry_count - 1) | |
963 | entries[i] = entries[log->entry_count - 1]; | |
964 | log->entry_count--; | |
965 | } else { | |
966 | i++; | |
967 | } | |
968 | } | |
969 | } | |
970 | ||
c07a5a4f TM |
971 | /* clear given bad block */ |
972 | static int clear_badblock(struct bbm_log *log, const __u8 idx, const unsigned | |
973 | long long sector, const int length) { | |
974 | __u32 i = 0; | |
975 | ||
976 | while (i < log->entry_count) { | |
977 | struct bbm_log_entry *entries = log->marked_block_entries; | |
978 | ||
979 | if ((entries[i].disk_ordinal == idx) && | |
980 | (__le48_to_cpu(&entries[i].defective_block_start) == | |
981 | sector) && (entries[i].marked_count + 1 == length)) { | |
982 | if (i < log->entry_count - 1) | |
983 | entries[i] = entries[log->entry_count - 1]; | |
984 | log->entry_count--; | |
985 | break; | |
986 | } | |
987 | i++; | |
988 | } | |
989 | ||
990 | return 1; | |
991 | } | |
8d67477f TM |
992 | |
993 | /* allocate and load BBM log from metadata */ | |
994 | static int load_bbm_log(struct intel_super *super) | |
995 | { | |
996 | struct imsm_super *mpb = super->anchor; | |
997 | __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size); | |
998 | ||
999 | super->bbm_log = xcalloc(1, sizeof(struct bbm_log)); | |
1000 | if (!super->bbm_log) | |
1001 | return 1; | |
1002 | ||
1003 | if (bbm_log_size) { | |
1004 | struct bbm_log *log = (void *)mpb + | |
1005 | __le32_to_cpu(mpb->mpb_size) - bbm_log_size; | |
1006 | ||
1007 | __u32 entry_count; | |
1008 | ||
1009 | if (bbm_log_size < sizeof(log->signature) + | |
1010 | sizeof(log->entry_count)) | |
1011 | return 2; | |
1012 | ||
1013 | entry_count = __le32_to_cpu(log->entry_count); | |
1014 | if ((__le32_to_cpu(log->signature) != BBM_LOG_SIGNATURE) || | |
1015 | (entry_count > BBM_LOG_MAX_ENTRIES)) | |
1016 | return 3; | |
1017 | ||
1018 | if (bbm_log_size != | |
1019 | sizeof(log->signature) + sizeof(log->entry_count) + | |
1020 | entry_count * sizeof(struct bbm_log_entry)) | |
1021 | return 4; | |
1022 | ||
1023 | memcpy(super->bbm_log, log, bbm_log_size); | |
1024 | } else { | |
1025 | super->bbm_log->signature = __cpu_to_le32(BBM_LOG_SIGNATURE); | |
1026 | super->bbm_log->entry_count = 0; | |
1027 | } | |
1028 | ||
1029 | return 0; | |
1030 | } | |
1031 | ||
b12796be TM |
1032 | /* checks if bad block is within volume boundaries */ |
1033 | static int is_bad_block_in_volume(const struct bbm_log_entry *entry, | |
1034 | const unsigned long long start_sector, | |
1035 | const unsigned long long size) | |
1036 | { | |
1037 | unsigned long long bb_start; | |
1038 | unsigned long long bb_end; | |
1039 | ||
1040 | bb_start = __le48_to_cpu(&entry->defective_block_start); | |
1041 | bb_end = bb_start + (entry->marked_count + 1); | |
1042 | ||
1043 | if (((bb_start >= start_sector) && (bb_start < start_sector + size)) || | |
1044 | ((bb_end >= start_sector) && (bb_end <= start_sector + size))) | |
1045 | return 1; | |
1046 | ||
1047 | return 0; | |
1048 | } | |
1049 | ||
1050 | /* get list of bad blocks on a drive for a volume */ | |
1051 | static void get_volume_badblocks(const struct bbm_log *log, const __u8 idx, | |
1052 | const unsigned long long start_sector, | |
1053 | const unsigned long long size, | |
1054 | struct md_bb *bbs) | |
1055 | { | |
1056 | __u32 count = 0; | |
1057 | __u32 i; | |
1058 | ||
1059 | for (i = 0; i < log->entry_count; i++) { | |
1060 | const struct bbm_log_entry *ent = | |
1061 | &log->marked_block_entries[i]; | |
1062 | struct md_bb_entry *bb; | |
1063 | ||
1064 | if ((ent->disk_ordinal == idx) && | |
1065 | is_bad_block_in_volume(ent, start_sector, size)) { | |
1066 | ||
1067 | if (!bbs->entries) { | |
1068 | bbs->entries = xmalloc(BBM_LOG_MAX_ENTRIES * | |
1069 | sizeof(*bb)); | |
1070 | if (!bbs->entries) | |
1071 | break; | |
1072 | } | |
1073 | ||
1074 | bb = &bbs->entries[count++]; | |
1075 | bb->sector = __le48_to_cpu(&ent->defective_block_start); | |
1076 | bb->length = ent->marked_count + 1; | |
1077 | } | |
1078 | } | |
1079 | bbs->count = count; | |
1080 | } | |
1081 | ||
98130f40 AK |
1082 | /* |
1083 | * for second_map: | |
238c0a71 AK |
1084 | * == MAP_0 get first map |
1085 | * == MAP_1 get second map | |
1086 | * == MAP_X than get map according to the current migr_state | |
98130f40 AK |
1087 | */ |
1088 | static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev, | |
1089 | int slot, | |
1090 | int second_map) | |
7eef0453 DW |
1091 | { |
1092 | struct imsm_map *map; | |
1093 | ||
5e7b0330 | 1094 | map = get_imsm_map(dev, second_map); |
7eef0453 | 1095 | |
ff077194 DW |
1096 | /* top byte identifies disk under rebuild */ |
1097 | return __le32_to_cpu(map->disk_ord_tbl[slot]); | |
1098 | } | |
1099 | ||
1100 | #define ord_to_idx(ord) (((ord) << 8) >> 8) | |
98130f40 | 1101 | static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot, int second_map) |
ff077194 | 1102 | { |
98130f40 | 1103 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, second_map); |
ff077194 DW |
1104 | |
1105 | return ord_to_idx(ord); | |
7eef0453 DW |
1106 | } |
1107 | ||
be73972f DW |
1108 | static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord) |
1109 | { | |
1110 | map->disk_ord_tbl[slot] = __cpu_to_le32(ord); | |
1111 | } | |
1112 | ||
f21e18ca | 1113 | static int get_imsm_disk_slot(struct imsm_map *map, unsigned idx) |
620b1713 DW |
1114 | { |
1115 | int slot; | |
1116 | __u32 ord; | |
1117 | ||
1118 | for (slot = 0; slot < map->num_members; slot++) { | |
1119 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
1120 | if (ord_to_idx(ord) == idx) | |
1121 | return slot; | |
1122 | } | |
1123 | ||
1124 | return -1; | |
1125 | } | |
1126 | ||
cdddbdbc DW |
1127 | static int get_imsm_raid_level(struct imsm_map *map) |
1128 | { | |
1129 | if (map->raid_level == 1) { | |
1130 | if (map->num_members == 2) | |
1131 | return 1; | |
1132 | else | |
1133 | return 10; | |
1134 | } | |
1135 | ||
1136 | return map->raid_level; | |
1137 | } | |
1138 | ||
c2c087e6 DW |
1139 | static int cmp_extent(const void *av, const void *bv) |
1140 | { | |
1141 | const struct extent *a = av; | |
1142 | const struct extent *b = bv; | |
1143 | if (a->start < b->start) | |
1144 | return -1; | |
1145 | if (a->start > b->start) | |
1146 | return 1; | |
1147 | return 0; | |
1148 | } | |
1149 | ||
0dcecb2e | 1150 | static int count_memberships(struct dl *dl, struct intel_super *super) |
c2c087e6 | 1151 | { |
c2c087e6 | 1152 | int memberships = 0; |
620b1713 | 1153 | int i; |
c2c087e6 | 1154 | |
949c47a0 DW |
1155 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
1156 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 1157 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
c2c087e6 | 1158 | |
620b1713 DW |
1159 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
1160 | memberships++; | |
c2c087e6 | 1161 | } |
0dcecb2e DW |
1162 | |
1163 | return memberships; | |
1164 | } | |
1165 | ||
b81221b7 CA |
1166 | static __u32 imsm_min_reserved_sectors(struct intel_super *super); |
1167 | ||
486720e0 | 1168 | static int split_ull(unsigned long long n, void *lo, void *hi) |
5551b113 CA |
1169 | { |
1170 | if (lo == 0 || hi == 0) | |
1171 | return 1; | |
486720e0 JS |
1172 | __put_unaligned32(__cpu_to_le32((__u32)n), lo); |
1173 | __put_unaligned32(__cpu_to_le32((n >> 32)), hi); | |
5551b113 CA |
1174 | return 0; |
1175 | } | |
1176 | ||
1177 | static unsigned long long join_u32(__u32 lo, __u32 hi) | |
1178 | { | |
1179 | return (unsigned long long)__le32_to_cpu(lo) | | |
1180 | (((unsigned long long)__le32_to_cpu(hi)) << 32); | |
1181 | } | |
1182 | ||
1183 | static unsigned long long total_blocks(struct imsm_disk *disk) | |
1184 | { | |
1185 | if (disk == NULL) | |
1186 | return 0; | |
1187 | return join_u32(disk->total_blocks_lo, disk->total_blocks_hi); | |
1188 | } | |
1189 | ||
1190 | static unsigned long long pba_of_lba0(struct imsm_map *map) | |
1191 | { | |
1192 | if (map == NULL) | |
1193 | return 0; | |
1194 | return join_u32(map->pba_of_lba0_lo, map->pba_of_lba0_hi); | |
1195 | } | |
1196 | ||
1197 | static unsigned long long blocks_per_member(struct imsm_map *map) | |
1198 | { | |
1199 | if (map == NULL) | |
1200 | return 0; | |
1201 | return join_u32(map->blocks_per_member_lo, map->blocks_per_member_hi); | |
1202 | } | |
1203 | ||
1204 | static unsigned long long num_data_stripes(struct imsm_map *map) | |
1205 | { | |
1206 | if (map == NULL) | |
1207 | return 0; | |
1208 | return join_u32(map->num_data_stripes_lo, map->num_data_stripes_hi); | |
1209 | } | |
1210 | ||
fcc2c9da MD |
1211 | static unsigned long long imsm_dev_size(struct imsm_dev *dev) |
1212 | { | |
1213 | if (dev == NULL) | |
1214 | return 0; | |
1215 | return join_u32(dev->size_low, dev->size_high); | |
1216 | } | |
1217 | ||
9f421827 PB |
1218 | static unsigned long long migr_chkp_area_pba(struct migr_record *migr_rec) |
1219 | { | |
1220 | if (migr_rec == NULL) | |
1221 | return 0; | |
1222 | return join_u32(migr_rec->ckpt_area_pba_lo, | |
1223 | migr_rec->ckpt_area_pba_hi); | |
1224 | } | |
1225 | ||
1226 | static unsigned long long current_migr_unit(struct migr_record *migr_rec) | |
1227 | { | |
1228 | if (migr_rec == NULL) | |
1229 | return 0; | |
1230 | return join_u32(migr_rec->curr_migr_unit_lo, | |
1231 | migr_rec->curr_migr_unit_hi); | |
1232 | } | |
1233 | ||
1234 | static unsigned long long migr_dest_1st_member_lba(struct migr_record *migr_rec) | |
1235 | { | |
1236 | if (migr_rec == NULL) | |
1237 | return 0; | |
1238 | return join_u32(migr_rec->dest_1st_member_lba_lo, | |
1239 | migr_rec->dest_1st_member_lba_hi); | |
1240 | } | |
1241 | ||
1242 | static unsigned long long get_num_migr_units(struct migr_record *migr_rec) | |
1243 | { | |
1244 | if (migr_rec == NULL) | |
1245 | return 0; | |
1246 | return join_u32(migr_rec->num_migr_units_lo, | |
1247 | migr_rec->num_migr_units_hi); | |
1248 | } | |
1249 | ||
5551b113 CA |
1250 | static void set_total_blocks(struct imsm_disk *disk, unsigned long long n) |
1251 | { | |
1252 | split_ull(n, &disk->total_blocks_lo, &disk->total_blocks_hi); | |
1253 | } | |
1254 | ||
1255 | static void set_pba_of_lba0(struct imsm_map *map, unsigned long long n) | |
1256 | { | |
1257 | split_ull(n, &map->pba_of_lba0_lo, &map->pba_of_lba0_hi); | |
1258 | } | |
1259 | ||
1260 | static void set_blocks_per_member(struct imsm_map *map, unsigned long long n) | |
1261 | { | |
1262 | split_ull(n, &map->blocks_per_member_lo, &map->blocks_per_member_hi); | |
1263 | } | |
1264 | ||
1265 | static void set_num_data_stripes(struct imsm_map *map, unsigned long long n) | |
1266 | { | |
1267 | split_ull(n, &map->num_data_stripes_lo, &map->num_data_stripes_hi); | |
1268 | } | |
1269 | ||
fcc2c9da MD |
1270 | static void set_imsm_dev_size(struct imsm_dev *dev, unsigned long long n) |
1271 | { | |
1272 | split_ull(n, &dev->size_low, &dev->size_high); | |
1273 | } | |
1274 | ||
9f421827 PB |
1275 | static void set_migr_chkp_area_pba(struct migr_record *migr_rec, |
1276 | unsigned long long n) | |
1277 | { | |
1278 | split_ull(n, &migr_rec->ckpt_area_pba_lo, &migr_rec->ckpt_area_pba_hi); | |
1279 | } | |
1280 | ||
1281 | static void set_current_migr_unit(struct migr_record *migr_rec, | |
1282 | unsigned long long n) | |
1283 | { | |
1284 | split_ull(n, &migr_rec->curr_migr_unit_lo, | |
1285 | &migr_rec->curr_migr_unit_hi); | |
1286 | } | |
1287 | ||
1288 | static void set_migr_dest_1st_member_lba(struct migr_record *migr_rec, | |
1289 | unsigned long long n) | |
1290 | { | |
1291 | split_ull(n, &migr_rec->dest_1st_member_lba_lo, | |
1292 | &migr_rec->dest_1st_member_lba_hi); | |
1293 | } | |
1294 | ||
1295 | static void set_num_migr_units(struct migr_record *migr_rec, | |
1296 | unsigned long long n) | |
1297 | { | |
1298 | split_ull(n, &migr_rec->num_migr_units_lo, | |
1299 | &migr_rec->num_migr_units_hi); | |
1300 | } | |
1301 | ||
44490938 MD |
1302 | static unsigned long long per_dev_array_size(struct imsm_map *map) |
1303 | { | |
1304 | unsigned long long array_size = 0; | |
1305 | ||
1306 | if (map == NULL) | |
1307 | return array_size; | |
1308 | ||
1309 | array_size = num_data_stripes(map) * map->blocks_per_strip; | |
1310 | if (get_imsm_raid_level(map) == 1 || get_imsm_raid_level(map) == 10) | |
1311 | array_size *= 2; | |
1312 | ||
1313 | return array_size; | |
1314 | } | |
1315 | ||
05501181 PB |
1316 | static struct extent *get_extents(struct intel_super *super, struct dl *dl, |
1317 | int get_minimal_reservation) | |
0dcecb2e DW |
1318 | { |
1319 | /* find a list of used extents on the given physical device */ | |
1320 | struct extent *rv, *e; | |
620b1713 | 1321 | int i; |
0dcecb2e | 1322 | int memberships = count_memberships(dl, super); |
b276dd33 DW |
1323 | __u32 reservation; |
1324 | ||
1325 | /* trim the reserved area for spares, so they can join any array | |
1326 | * regardless of whether the OROM has assigned sectors from the | |
1327 | * IMSM_RESERVED_SECTORS region | |
1328 | */ | |
05501181 | 1329 | if (dl->index == -1 || get_minimal_reservation) |
b81221b7 | 1330 | reservation = imsm_min_reserved_sectors(super); |
b276dd33 DW |
1331 | else |
1332 | reservation = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
0dcecb2e | 1333 | |
503975b9 | 1334 | rv = xcalloc(sizeof(struct extent), (memberships + 1)); |
c2c087e6 DW |
1335 | e = rv; |
1336 | ||
949c47a0 DW |
1337 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
1338 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 1339 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
c2c087e6 | 1340 | |
620b1713 | 1341 | if (get_imsm_disk_slot(map, dl->index) >= 0) { |
5551b113 | 1342 | e->start = pba_of_lba0(map); |
44490938 | 1343 | e->size = per_dev_array_size(map); |
620b1713 | 1344 | e++; |
c2c087e6 DW |
1345 | } |
1346 | } | |
1347 | qsort(rv, memberships, sizeof(*rv), cmp_extent); | |
1348 | ||
1011e834 | 1349 | /* determine the start of the metadata |
14e8215b DW |
1350 | * when no raid devices are defined use the default |
1351 | * ...otherwise allow the metadata to truncate the value | |
1352 | * as is the case with older versions of imsm | |
1353 | */ | |
1354 | if (memberships) { | |
1355 | struct extent *last = &rv[memberships - 1]; | |
5551b113 | 1356 | unsigned long long remainder; |
14e8215b | 1357 | |
5551b113 | 1358 | remainder = total_blocks(&dl->disk) - (last->start + last->size); |
dda5855f DW |
1359 | /* round down to 1k block to satisfy precision of the kernel |
1360 | * 'size' interface | |
1361 | */ | |
1362 | remainder &= ~1UL; | |
1363 | /* make sure remainder is still sane */ | |
f21e18ca | 1364 | if (remainder < (unsigned)ROUND_UP(super->len, 512) >> 9) |
dda5855f | 1365 | remainder = ROUND_UP(super->len, 512) >> 9; |
14e8215b DW |
1366 | if (reservation > remainder) |
1367 | reservation = remainder; | |
1368 | } | |
5551b113 | 1369 | e->start = total_blocks(&dl->disk) - reservation; |
c2c087e6 DW |
1370 | e->size = 0; |
1371 | return rv; | |
1372 | } | |
1373 | ||
14e8215b DW |
1374 | /* try to determine how much space is reserved for metadata from |
1375 | * the last get_extents() entry, otherwise fallback to the | |
1376 | * default | |
1377 | */ | |
1378 | static __u32 imsm_reserved_sectors(struct intel_super *super, struct dl *dl) | |
1379 | { | |
1380 | struct extent *e; | |
1381 | int i; | |
1382 | __u32 rv; | |
1383 | ||
1384 | /* for spares just return a minimal reservation which will grow | |
1385 | * once the spare is picked up by an array | |
1386 | */ | |
1387 | if (dl->index == -1) | |
1388 | return MPB_SECTOR_CNT; | |
1389 | ||
05501181 | 1390 | e = get_extents(super, dl, 0); |
14e8215b DW |
1391 | if (!e) |
1392 | return MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
1393 | ||
1394 | /* scroll to last entry */ | |
1395 | for (i = 0; e[i].size; i++) | |
1396 | continue; | |
1397 | ||
5551b113 | 1398 | rv = total_blocks(&dl->disk) - e[i].start; |
14e8215b DW |
1399 | |
1400 | free(e); | |
1401 | ||
1402 | return rv; | |
1403 | } | |
1404 | ||
25ed7e59 DW |
1405 | static int is_spare(struct imsm_disk *disk) |
1406 | { | |
1407 | return (disk->status & SPARE_DISK) == SPARE_DISK; | |
1408 | } | |
1409 | ||
1410 | static int is_configured(struct imsm_disk *disk) | |
1411 | { | |
1412 | return (disk->status & CONFIGURED_DISK) == CONFIGURED_DISK; | |
1413 | } | |
1414 | ||
1415 | static int is_failed(struct imsm_disk *disk) | |
1416 | { | |
1417 | return (disk->status & FAILED_DISK) == FAILED_DISK; | |
1418 | } | |
1419 | ||
2432ce9b AP |
1420 | static int is_journal(struct imsm_disk *disk) |
1421 | { | |
1422 | return (disk->status & JOURNAL_DISK) == JOURNAL_DISK; | |
1423 | } | |
1424 | ||
b53bfba6 TM |
1425 | /* round array size down to closest MB and ensure it splits evenly |
1426 | * between members | |
1427 | */ | |
1428 | static unsigned long long round_size_to_mb(unsigned long long size, unsigned int | |
1429 | disk_count) | |
1430 | { | |
1431 | size /= disk_count; | |
1432 | size = (size >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT; | |
1433 | size *= disk_count; | |
1434 | ||
1435 | return size; | |
1436 | } | |
1437 | ||
8b9cd157 MK |
1438 | static int able_to_resync(int raid_level, int missing_disks) |
1439 | { | |
1440 | int max_missing_disks = 0; | |
1441 | ||
1442 | switch (raid_level) { | |
1443 | case 10: | |
1444 | max_missing_disks = 1; | |
1445 | break; | |
1446 | default: | |
1447 | max_missing_disks = 0; | |
1448 | } | |
1449 | return missing_disks <= max_missing_disks; | |
1450 | } | |
1451 | ||
b81221b7 CA |
1452 | /* try to determine how much space is reserved for metadata from |
1453 | * the last get_extents() entry on the smallest active disk, | |
1454 | * otherwise fallback to the default | |
1455 | */ | |
1456 | static __u32 imsm_min_reserved_sectors(struct intel_super *super) | |
1457 | { | |
1458 | struct extent *e; | |
1459 | int i; | |
5551b113 CA |
1460 | unsigned long long min_active; |
1461 | __u32 remainder; | |
b81221b7 CA |
1462 | __u32 rv = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; |
1463 | struct dl *dl, *dl_min = NULL; | |
1464 | ||
1465 | if (!super) | |
1466 | return rv; | |
1467 | ||
1468 | min_active = 0; | |
1469 | for (dl = super->disks; dl; dl = dl->next) { | |
1470 | if (dl->index < 0) | |
1471 | continue; | |
5551b113 CA |
1472 | unsigned long long blocks = total_blocks(&dl->disk); |
1473 | if (blocks < min_active || min_active == 0) { | |
b81221b7 | 1474 | dl_min = dl; |
5551b113 | 1475 | min_active = blocks; |
b81221b7 CA |
1476 | } |
1477 | } | |
1478 | if (!dl_min) | |
1479 | return rv; | |
1480 | ||
1481 | /* find last lba used by subarrays on the smallest active disk */ | |
05501181 | 1482 | e = get_extents(super, dl_min, 0); |
b81221b7 CA |
1483 | if (!e) |
1484 | return rv; | |
1485 | for (i = 0; e[i].size; i++) | |
1486 | continue; | |
1487 | ||
1488 | remainder = min_active - e[i].start; | |
1489 | free(e); | |
1490 | ||
1491 | /* to give priority to recovery we should not require full | |
1492 | IMSM_RESERVED_SECTORS from the spare */ | |
1493 | rv = MPB_SECTOR_CNT + NUM_BLOCKS_DIRTY_STRIPE_REGION; | |
1494 | ||
1495 | /* if real reservation is smaller use that value */ | |
1496 | return (remainder < rv) ? remainder : rv; | |
1497 | } | |
1498 | ||
fbfdcb06 AO |
1499 | /* |
1500 | * Return minimum size of a spare and sector size | |
1501 | * that can be used in this array | |
1502 | */ | |
1503 | int get_spare_criteria_imsm(struct supertype *st, struct spare_criteria *c) | |
80e7f8c3 AC |
1504 | { |
1505 | struct intel_super *super = st->sb; | |
1506 | struct dl *dl; | |
1507 | struct extent *e; | |
1508 | int i; | |
fbfdcb06 AO |
1509 | unsigned long long size = 0; |
1510 | ||
1511 | c->min_size = 0; | |
4b57ecf6 | 1512 | c->sector_size = 0; |
80e7f8c3 AC |
1513 | |
1514 | if (!super) | |
fbfdcb06 | 1515 | return -EINVAL; |
80e7f8c3 AC |
1516 | /* find first active disk in array */ |
1517 | dl = super->disks; | |
1518 | while (dl && (is_failed(&dl->disk) || dl->index == -1)) | |
1519 | dl = dl->next; | |
1520 | if (!dl) | |
fbfdcb06 | 1521 | return -EINVAL; |
80e7f8c3 | 1522 | /* find last lba used by subarrays */ |
05501181 | 1523 | e = get_extents(super, dl, 0); |
80e7f8c3 | 1524 | if (!e) |
fbfdcb06 | 1525 | return -EINVAL; |
80e7f8c3 AC |
1526 | for (i = 0; e[i].size; i++) |
1527 | continue; | |
1528 | if (i > 0) | |
fbfdcb06 | 1529 | size = e[i-1].start + e[i-1].size; |
80e7f8c3 | 1530 | free(e); |
b81221b7 | 1531 | |
80e7f8c3 | 1532 | /* add the amount of space needed for metadata */ |
fbfdcb06 AO |
1533 | size += imsm_min_reserved_sectors(super); |
1534 | ||
1535 | c->min_size = size * 512; | |
4b57ecf6 | 1536 | c->sector_size = super->sector_size; |
b81221b7 | 1537 | |
fbfdcb06 | 1538 | return 0; |
80e7f8c3 AC |
1539 | } |
1540 | ||
d1e02575 AK |
1541 | static int is_gen_migration(struct imsm_dev *dev); |
1542 | ||
f36a9ecd PB |
1543 | #define IMSM_4K_DIV 8 |
1544 | ||
c47b0ff6 AK |
1545 | static __u64 blocks_per_migr_unit(struct intel_super *super, |
1546 | struct imsm_dev *dev); | |
1e5c6983 | 1547 | |
c47b0ff6 AK |
1548 | static void print_imsm_dev(struct intel_super *super, |
1549 | struct imsm_dev *dev, | |
1550 | char *uuid, | |
1551 | int disk_idx) | |
cdddbdbc DW |
1552 | { |
1553 | __u64 sz; | |
0d80bb2f | 1554 | int slot, i; |
238c0a71 AK |
1555 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1556 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); | |
b10b37b8 | 1557 | __u32 ord; |
cdddbdbc DW |
1558 | |
1559 | printf("\n"); | |
1e7bc0ed | 1560 | printf("[%.16s]:\n", dev->volume); |
44470971 | 1561 | printf(" UUID : %s\n", uuid); |
dd8bcb3b AK |
1562 | printf(" RAID Level : %d", get_imsm_raid_level(map)); |
1563 | if (map2) | |
1564 | printf(" <-- %d", get_imsm_raid_level(map2)); | |
1565 | printf("\n"); | |
1566 | printf(" Members : %d", map->num_members); | |
1567 | if (map2) | |
1568 | printf(" <-- %d", map2->num_members); | |
1569 | printf("\n"); | |
0d80bb2f DW |
1570 | printf(" Slots : ["); |
1571 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 1572 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_0); |
0d80bb2f DW |
1573 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
1574 | } | |
dd8bcb3b AK |
1575 | printf("]"); |
1576 | if (map2) { | |
1577 | printf(" <-- ["); | |
1578 | for (i = 0; i < map2->num_members; i++) { | |
238c0a71 | 1579 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_1); |
dd8bcb3b AK |
1580 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
1581 | } | |
1582 | printf("]"); | |
1583 | } | |
1584 | printf("\n"); | |
7095bccb AK |
1585 | printf(" Failed disk : "); |
1586 | if (map->failed_disk_num == 0xff) | |
1587 | printf("none"); | |
1588 | else | |
1589 | printf("%i", map->failed_disk_num); | |
1590 | printf("\n"); | |
620b1713 DW |
1591 | slot = get_imsm_disk_slot(map, disk_idx); |
1592 | if (slot >= 0) { | |
238c0a71 | 1593 | ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); |
b10b37b8 DW |
1594 | printf(" This Slot : %d%s\n", slot, |
1595 | ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : ""); | |
1596 | } else | |
cdddbdbc | 1597 | printf(" This Slot : ?\n"); |
84918897 | 1598 | printf(" Sector Size : %u\n", super->sector_size); |
fcc2c9da | 1599 | sz = imsm_dev_size(dev); |
84918897 MK |
1600 | printf(" Array Size : %llu%s\n", |
1601 | (unsigned long long)sz * 512 / super->sector_size, | |
cdddbdbc | 1602 | human_size(sz * 512)); |
5551b113 | 1603 | sz = blocks_per_member(map); |
84918897 MK |
1604 | printf(" Per Dev Size : %llu%s\n", |
1605 | (unsigned long long)sz * 512 / super->sector_size, | |
cdddbdbc | 1606 | human_size(sz * 512)); |
5551b113 CA |
1607 | printf(" Sector Offset : %llu\n", |
1608 | pba_of_lba0(map)); | |
1609 | printf(" Num Stripes : %llu\n", | |
1610 | num_data_stripes(map)); | |
dd8bcb3b | 1611 | printf(" Chunk Size : %u KiB", |
cdddbdbc | 1612 | __le16_to_cpu(map->blocks_per_strip) / 2); |
dd8bcb3b AK |
1613 | if (map2) |
1614 | printf(" <-- %u KiB", | |
1615 | __le16_to_cpu(map2->blocks_per_strip) / 2); | |
1616 | printf("\n"); | |
cdddbdbc | 1617 | printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks)); |
8655a7b1 | 1618 | printf(" Migrate State : "); |
1484e727 DW |
1619 | if (dev->vol.migr_state) { |
1620 | if (migr_type(dev) == MIGR_INIT) | |
8655a7b1 | 1621 | printf("initialize\n"); |
1484e727 | 1622 | else if (migr_type(dev) == MIGR_REBUILD) |
8655a7b1 | 1623 | printf("rebuild\n"); |
1484e727 | 1624 | else if (migr_type(dev) == MIGR_VERIFY) |
8655a7b1 | 1625 | printf("check\n"); |
1484e727 | 1626 | else if (migr_type(dev) == MIGR_GEN_MIGR) |
8655a7b1 | 1627 | printf("general migration\n"); |
1484e727 | 1628 | else if (migr_type(dev) == MIGR_STATE_CHANGE) |
8655a7b1 | 1629 | printf("state change\n"); |
1484e727 | 1630 | else if (migr_type(dev) == MIGR_REPAIR) |
8655a7b1 | 1631 | printf("repair\n"); |
1484e727 | 1632 | else |
8655a7b1 DW |
1633 | printf("<unknown:%d>\n", migr_type(dev)); |
1634 | } else | |
1635 | printf("idle\n"); | |
3393c6af DW |
1636 | printf(" Map State : %s", map_state_str[map->map_state]); |
1637 | if (dev->vol.migr_state) { | |
238c0a71 | 1638 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 | 1639 | |
b10b37b8 | 1640 | printf(" <-- %s", map_state_str[map->map_state]); |
464d40e8 LD |
1641 | printf("\n Checkpoint : %u ", |
1642 | __le32_to_cpu(dev->vol.curr_migr_unit)); | |
089f9d79 | 1643 | if (is_gen_migration(dev) && (slot > 1 || slot < 0)) |
464d40e8 LD |
1644 | printf("(N/A)"); |
1645 | else | |
1646 | printf("(%llu)", (unsigned long long) | |
1647 | blocks_per_migr_unit(super, dev)); | |
3393c6af DW |
1648 | } |
1649 | printf("\n"); | |
2432ce9b AP |
1650 | printf(" Dirty State : %s\n", (dev->vol.dirty & RAIDVOL_DIRTY) ? |
1651 | "dirty" : "clean"); | |
1652 | printf(" RWH Policy : "); | |
c2462068 | 1653 | if (dev->rwh_policy == RWH_OFF || dev->rwh_policy == RWH_MULTIPLE_OFF) |
2432ce9b AP |
1654 | printf("off\n"); |
1655 | else if (dev->rwh_policy == RWH_DISTRIBUTED) | |
1656 | printf("PPL distributed\n"); | |
1657 | else if (dev->rwh_policy == RWH_JOURNALING_DRIVE) | |
1658 | printf("PPL journaling drive\n"); | |
c2462068 PB |
1659 | else if (dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED) |
1660 | printf("Multiple distributed PPLs\n"); | |
1661 | else if (dev->rwh_policy == RWH_MULTIPLE_PPLS_JOURNALING_DRIVE) | |
1662 | printf("Multiple PPLs on journaling drive\n"); | |
2432ce9b AP |
1663 | else |
1664 | printf("<unknown:%d>\n", dev->rwh_policy); | |
cdddbdbc DW |
1665 | } |
1666 | ||
ef5c214e MK |
1667 | static void print_imsm_disk(struct imsm_disk *disk, |
1668 | int index, | |
1669 | __u32 reserved, | |
1670 | unsigned int sector_size) { | |
1f24f035 | 1671 | char str[MAX_RAID_SERIAL_LEN + 1]; |
cdddbdbc DW |
1672 | __u64 sz; |
1673 | ||
0ec1f4e8 | 1674 | if (index < -1 || !disk) |
e9d82038 DW |
1675 | return; |
1676 | ||
cdddbdbc | 1677 | printf("\n"); |
1f24f035 | 1678 | snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial); |
0ec1f4e8 DW |
1679 | if (index >= 0) |
1680 | printf(" Disk%02d Serial : %s\n", index, str); | |
1681 | else | |
1682 | printf(" Disk Serial : %s\n", str); | |
2432ce9b AP |
1683 | printf(" State :%s%s%s%s\n", is_spare(disk) ? " spare" : "", |
1684 | is_configured(disk) ? " active" : "", | |
1685 | is_failed(disk) ? " failed" : "", | |
1686 | is_journal(disk) ? " journal" : ""); | |
cdddbdbc | 1687 | printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id)); |
5551b113 | 1688 | sz = total_blocks(disk) - reserved; |
ef5c214e MK |
1689 | printf(" Usable Size : %llu%s\n", |
1690 | (unsigned long long)sz * 512 / sector_size, | |
cdddbdbc DW |
1691 | human_size(sz * 512)); |
1692 | } | |
1693 | ||
de44e46f PB |
1694 | void convert_to_4k_imsm_migr_rec(struct intel_super *super) |
1695 | { | |
1696 | struct migr_record *migr_rec = super->migr_rec; | |
1697 | ||
1698 | migr_rec->blocks_per_unit /= IMSM_4K_DIV; | |
de44e46f PB |
1699 | migr_rec->dest_depth_per_unit /= IMSM_4K_DIV; |
1700 | split_ull((join_u32(migr_rec->post_migr_vol_cap, | |
1701 | migr_rec->post_migr_vol_cap_hi) / IMSM_4K_DIV), | |
1702 | &migr_rec->post_migr_vol_cap, &migr_rec->post_migr_vol_cap_hi); | |
9f421827 PB |
1703 | set_migr_chkp_area_pba(migr_rec, |
1704 | migr_chkp_area_pba(migr_rec) / IMSM_4K_DIV); | |
1705 | set_migr_dest_1st_member_lba(migr_rec, | |
1706 | migr_dest_1st_member_lba(migr_rec) / IMSM_4K_DIV); | |
de44e46f PB |
1707 | } |
1708 | ||
f36a9ecd PB |
1709 | void convert_to_4k_imsm_disk(struct imsm_disk *disk) |
1710 | { | |
1711 | set_total_blocks(disk, (total_blocks(disk)/IMSM_4K_DIV)); | |
1712 | } | |
1713 | ||
1714 | void convert_to_4k(struct intel_super *super) | |
1715 | { | |
1716 | struct imsm_super *mpb = super->anchor; | |
1717 | struct imsm_disk *disk; | |
1718 | int i; | |
e4467bc7 | 1719 | __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size); |
f36a9ecd PB |
1720 | |
1721 | for (i = 0; i < mpb->num_disks ; i++) { | |
1722 | disk = __get_imsm_disk(mpb, i); | |
1723 | /* disk */ | |
1724 | convert_to_4k_imsm_disk(disk); | |
1725 | } | |
1726 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
1727 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
1728 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
1729 | /* dev */ | |
fcc2c9da | 1730 | set_imsm_dev_size(dev, imsm_dev_size(dev)/IMSM_4K_DIV); |
f36a9ecd PB |
1731 | dev->vol.curr_migr_unit /= IMSM_4K_DIV; |
1732 | ||
1733 | /* map0 */ | |
1734 | set_blocks_per_member(map, blocks_per_member(map)/IMSM_4K_DIV); | |
1735 | map->blocks_per_strip /= IMSM_4K_DIV; | |
1736 | set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV); | |
1737 | ||
1738 | if (dev->vol.migr_state) { | |
1739 | /* map1 */ | |
1740 | map = get_imsm_map(dev, MAP_1); | |
1741 | set_blocks_per_member(map, | |
1742 | blocks_per_member(map)/IMSM_4K_DIV); | |
1743 | map->blocks_per_strip /= IMSM_4K_DIV; | |
1744 | set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV); | |
1745 | } | |
1746 | } | |
e4467bc7 TM |
1747 | if (bbm_log_size) { |
1748 | struct bbm_log *log = (void *)mpb + | |
1749 | __le32_to_cpu(mpb->mpb_size) - bbm_log_size; | |
1750 | __u32 i; | |
1751 | ||
1752 | for (i = 0; i < log->entry_count; i++) { | |
1753 | struct bbm_log_entry *entry = | |
1754 | &log->marked_block_entries[i]; | |
1755 | ||
1756 | __u8 count = entry->marked_count + 1; | |
1757 | unsigned long long sector = | |
1758 | __le48_to_cpu(&entry->defective_block_start); | |
1759 | ||
1760 | entry->defective_block_start = | |
1761 | __cpu_to_le48(sector/IMSM_4K_DIV); | |
1762 | entry->marked_count = max(count/IMSM_4K_DIV, 1) - 1; | |
1763 | } | |
1764 | } | |
f36a9ecd PB |
1765 | |
1766 | mpb->check_sum = __gen_imsm_checksum(mpb); | |
1767 | } | |
1768 | ||
520e69e2 AK |
1769 | void examine_migr_rec_imsm(struct intel_super *super) |
1770 | { | |
1771 | struct migr_record *migr_rec = super->migr_rec; | |
1772 | struct imsm_super *mpb = super->anchor; | |
1773 | int i; | |
1774 | ||
1775 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
1776 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
3136abe5 | 1777 | struct imsm_map *map; |
b4ab44d8 | 1778 | int slot = -1; |
3136abe5 | 1779 | |
520e69e2 AK |
1780 | if (is_gen_migration(dev) == 0) |
1781 | continue; | |
1782 | ||
1783 | printf("\nMigration Record Information:"); | |
3136abe5 | 1784 | |
44bfe6df AK |
1785 | /* first map under migration */ |
1786 | map = get_imsm_map(dev, MAP_0); | |
3136abe5 AK |
1787 | if (map) |
1788 | slot = get_imsm_disk_slot(map, super->disks->index); | |
089f9d79 | 1789 | if (map == NULL || slot > 1 || slot < 0) { |
520e69e2 AK |
1790 | printf(" Empty\n "); |
1791 | printf("Examine one of first two disks in array\n"); | |
1792 | break; | |
1793 | } | |
1794 | printf("\n Status : "); | |
1795 | if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL) | |
1796 | printf("Normal\n"); | |
1797 | else | |
1798 | printf("Contains Data\n"); | |
9f421827 PB |
1799 | printf(" Current Unit : %llu\n", |
1800 | current_migr_unit(migr_rec)); | |
520e69e2 AK |
1801 | printf(" Family : %u\n", |
1802 | __le32_to_cpu(migr_rec->family_num)); | |
1803 | printf(" Ascending : %u\n", | |
1804 | __le32_to_cpu(migr_rec->ascending_migr)); | |
1805 | printf(" Blocks Per Unit : %u\n", | |
1806 | __le32_to_cpu(migr_rec->blocks_per_unit)); | |
1807 | printf(" Dest. Depth Per Unit : %u\n", | |
1808 | __le32_to_cpu(migr_rec->dest_depth_per_unit)); | |
9f421827 PB |
1809 | printf(" Checkpoint Area pba : %llu\n", |
1810 | migr_chkp_area_pba(migr_rec)); | |
1811 | printf(" First member lba : %llu\n", | |
1812 | migr_dest_1st_member_lba(migr_rec)); | |
1813 | printf(" Total Number of Units : %llu\n", | |
1814 | get_num_migr_units(migr_rec)); | |
1815 | printf(" Size of volume : %llu\n", | |
1816 | join_u32(migr_rec->post_migr_vol_cap, | |
1817 | migr_rec->post_migr_vol_cap_hi)); | |
520e69e2 AK |
1818 | printf(" Record was read from : %u\n", |
1819 | __le32_to_cpu(migr_rec->ckpt_read_disk_num)); | |
1820 | ||
1821 | break; | |
1822 | } | |
1823 | } | |
f36a9ecd | 1824 | |
de44e46f PB |
1825 | void convert_from_4k_imsm_migr_rec(struct intel_super *super) |
1826 | { | |
1827 | struct migr_record *migr_rec = super->migr_rec; | |
1828 | ||
1829 | migr_rec->blocks_per_unit *= IMSM_4K_DIV; | |
de44e46f PB |
1830 | migr_rec->dest_depth_per_unit *= IMSM_4K_DIV; |
1831 | split_ull((join_u32(migr_rec->post_migr_vol_cap, | |
1832 | migr_rec->post_migr_vol_cap_hi) * IMSM_4K_DIV), | |
1833 | &migr_rec->post_migr_vol_cap, | |
1834 | &migr_rec->post_migr_vol_cap_hi); | |
9f421827 PB |
1835 | set_migr_chkp_area_pba(migr_rec, |
1836 | migr_chkp_area_pba(migr_rec) * IMSM_4K_DIV); | |
1837 | set_migr_dest_1st_member_lba(migr_rec, | |
1838 | migr_dest_1st_member_lba(migr_rec) * IMSM_4K_DIV); | |
de44e46f PB |
1839 | } |
1840 | ||
f36a9ecd PB |
1841 | void convert_from_4k(struct intel_super *super) |
1842 | { | |
1843 | struct imsm_super *mpb = super->anchor; | |
1844 | struct imsm_disk *disk; | |
1845 | int i; | |
e4467bc7 | 1846 | __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size); |
f36a9ecd PB |
1847 | |
1848 | for (i = 0; i < mpb->num_disks ; i++) { | |
1849 | disk = __get_imsm_disk(mpb, i); | |
1850 | /* disk */ | |
1851 | set_total_blocks(disk, (total_blocks(disk)*IMSM_4K_DIV)); | |
1852 | } | |
1853 | ||
1854 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
1855 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
1856 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
1857 | /* dev */ | |
fcc2c9da | 1858 | set_imsm_dev_size(dev, imsm_dev_size(dev)*IMSM_4K_DIV); |
f36a9ecd PB |
1859 | dev->vol.curr_migr_unit *= IMSM_4K_DIV; |
1860 | ||
1861 | /* map0 */ | |
1862 | set_blocks_per_member(map, blocks_per_member(map)*IMSM_4K_DIV); | |
1863 | map->blocks_per_strip *= IMSM_4K_DIV; | |
1864 | set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV); | |
1865 | ||
1866 | if (dev->vol.migr_state) { | |
1867 | /* map1 */ | |
1868 | map = get_imsm_map(dev, MAP_1); | |
1869 | set_blocks_per_member(map, | |
1870 | blocks_per_member(map)*IMSM_4K_DIV); | |
1871 | map->blocks_per_strip *= IMSM_4K_DIV; | |
1872 | set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV); | |
1873 | } | |
1874 | } | |
e4467bc7 TM |
1875 | if (bbm_log_size) { |
1876 | struct bbm_log *log = (void *)mpb + | |
1877 | __le32_to_cpu(mpb->mpb_size) - bbm_log_size; | |
1878 | __u32 i; | |
1879 | ||
1880 | for (i = 0; i < log->entry_count; i++) { | |
1881 | struct bbm_log_entry *entry = | |
1882 | &log->marked_block_entries[i]; | |
1883 | ||
1884 | __u8 count = entry->marked_count + 1; | |
1885 | unsigned long long sector = | |
1886 | __le48_to_cpu(&entry->defective_block_start); | |
1887 | ||
1888 | entry->defective_block_start = | |
1889 | __cpu_to_le48(sector*IMSM_4K_DIV); | |
1890 | entry->marked_count = count*IMSM_4K_DIV - 1; | |
1891 | } | |
1892 | } | |
f36a9ecd PB |
1893 | |
1894 | mpb->check_sum = __gen_imsm_checksum(mpb); | |
1895 | } | |
1896 | ||
19482bcc AK |
1897 | /******************************************************************************* |
1898 | * function: imsm_check_attributes | |
1899 | * Description: Function checks if features represented by attributes flags | |
1011e834 | 1900 | * are supported by mdadm. |
19482bcc AK |
1901 | * Parameters: |
1902 | * attributes - Attributes read from metadata | |
1903 | * Returns: | |
1011e834 N |
1904 | * 0 - passed attributes contains unsupported features flags |
1905 | * 1 - all features are supported | |
19482bcc AK |
1906 | ******************************************************************************/ |
1907 | static int imsm_check_attributes(__u32 attributes) | |
1908 | { | |
1909 | int ret_val = 1; | |
418f9b36 N |
1910 | __u32 not_supported = MPB_ATTRIB_SUPPORTED^0xffffffff; |
1911 | ||
1912 | not_supported &= ~MPB_ATTRIB_IGNORED; | |
19482bcc AK |
1913 | |
1914 | not_supported &= attributes; | |
1915 | if (not_supported) { | |
e7b84f9d | 1916 | pr_err("(IMSM): Unsupported attributes : %x\n", |
418f9b36 | 1917 | (unsigned)__le32_to_cpu(not_supported)); |
19482bcc AK |
1918 | if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) { |
1919 | dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY \n"); | |
1920 | not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY; | |
1921 | } | |
1922 | if (not_supported & MPB_ATTRIB_2TB) { | |
1923 | dprintf("\t\tMPB_ATTRIB_2TB\n"); | |
1924 | not_supported ^= MPB_ATTRIB_2TB; | |
1925 | } | |
1926 | if (not_supported & MPB_ATTRIB_RAID0) { | |
1927 | dprintf("\t\tMPB_ATTRIB_RAID0\n"); | |
1928 | not_supported ^= MPB_ATTRIB_RAID0; | |
1929 | } | |
1930 | if (not_supported & MPB_ATTRIB_RAID1) { | |
1931 | dprintf("\t\tMPB_ATTRIB_RAID1\n"); | |
1932 | not_supported ^= MPB_ATTRIB_RAID1; | |
1933 | } | |
1934 | if (not_supported & MPB_ATTRIB_RAID10) { | |
1935 | dprintf("\t\tMPB_ATTRIB_RAID10\n"); | |
1936 | not_supported ^= MPB_ATTRIB_RAID10; | |
1937 | } | |
1938 | if (not_supported & MPB_ATTRIB_RAID1E) { | |
1939 | dprintf("\t\tMPB_ATTRIB_RAID1E\n"); | |
1940 | not_supported ^= MPB_ATTRIB_RAID1E; | |
1941 | } | |
1942 | if (not_supported & MPB_ATTRIB_RAID5) { | |
1943 | dprintf("\t\tMPB_ATTRIB_RAID5\n"); | |
1944 | not_supported ^= MPB_ATTRIB_RAID5; | |
1945 | } | |
1946 | if (not_supported & MPB_ATTRIB_RAIDCNG) { | |
1947 | dprintf("\t\tMPB_ATTRIB_RAIDCNG\n"); | |
1948 | not_supported ^= MPB_ATTRIB_RAIDCNG; | |
1949 | } | |
1950 | if (not_supported & MPB_ATTRIB_BBM) { | |
1951 | dprintf("\t\tMPB_ATTRIB_BBM\n"); | |
1952 | not_supported ^= MPB_ATTRIB_BBM; | |
1953 | } | |
1954 | if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) { | |
1955 | dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY (== MPB_ATTRIB_LEGACY)\n"); | |
1956 | not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY; | |
1957 | } | |
1958 | if (not_supported & MPB_ATTRIB_EXP_STRIPE_SIZE) { | |
1959 | dprintf("\t\tMPB_ATTRIB_EXP_STRIP_SIZE\n"); | |
1960 | not_supported ^= MPB_ATTRIB_EXP_STRIPE_SIZE; | |
1961 | } | |
1962 | if (not_supported & MPB_ATTRIB_2TB_DISK) { | |
1963 | dprintf("\t\tMPB_ATTRIB_2TB_DISK\n"); | |
1964 | not_supported ^= MPB_ATTRIB_2TB_DISK; | |
1965 | } | |
1966 | if (not_supported & MPB_ATTRIB_NEVER_USE2) { | |
1967 | dprintf("\t\tMPB_ATTRIB_NEVER_USE2\n"); | |
1968 | not_supported ^= MPB_ATTRIB_NEVER_USE2; | |
1969 | } | |
1970 | if (not_supported & MPB_ATTRIB_NEVER_USE) { | |
1971 | dprintf("\t\tMPB_ATTRIB_NEVER_USE\n"); | |
1972 | not_supported ^= MPB_ATTRIB_NEVER_USE; | |
1973 | } | |
1974 | ||
1975 | if (not_supported) | |
1ade5cc1 | 1976 | dprintf("(IMSM): Unknown attributes : %x\n", not_supported); |
19482bcc AK |
1977 | |
1978 | ret_val = 0; | |
1979 | } | |
1980 | ||
1981 | return ret_val; | |
1982 | } | |
1983 | ||
a5d85af7 | 1984 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map); |
44470971 | 1985 | |
cdddbdbc DW |
1986 | static void examine_super_imsm(struct supertype *st, char *homehost) |
1987 | { | |
1988 | struct intel_super *super = st->sb; | |
949c47a0 | 1989 | struct imsm_super *mpb = super->anchor; |
cdddbdbc DW |
1990 | char str[MAX_SIGNATURE_LENGTH]; |
1991 | int i; | |
27fd6274 DW |
1992 | struct mdinfo info; |
1993 | char nbuf[64]; | |
cdddbdbc | 1994 | __u32 sum; |
14e8215b | 1995 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
94827db3 | 1996 | struct dl *dl; |
27fd6274 | 1997 | |
618f4e6d XN |
1998 | strncpy(str, (char *)mpb->sig, MPB_SIG_LEN); |
1999 | str[MPB_SIG_LEN-1] = '\0'; | |
cdddbdbc | 2000 | printf(" Magic : %s\n", str); |
cdddbdbc | 2001 | printf(" Version : %s\n", get_imsm_version(mpb)); |
148acb7b | 2002 | printf(" Orig Family : %08x\n", __le32_to_cpu(mpb->orig_family_num)); |
cdddbdbc DW |
2003 | printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num)); |
2004 | printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num)); | |
19482bcc AK |
2005 | printf(" Attributes : "); |
2006 | if (imsm_check_attributes(mpb->attributes)) | |
2007 | printf("All supported\n"); | |
2008 | else | |
2009 | printf("not supported\n"); | |
a5d85af7 | 2010 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 2011 | fname_from_uuid(st, &info, nbuf, ':'); |
27fd6274 | 2012 | printf(" UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
2013 | sum = __le32_to_cpu(mpb->check_sum); |
2014 | printf(" Checksum : %08x %s\n", sum, | |
949c47a0 | 2015 | __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect"); |
f36a9ecd | 2016 | printf(" MPB Sectors : %d\n", mpb_sectors(mpb, super->sector_size)); |
cdddbdbc DW |
2017 | printf(" Disks : %d\n", mpb->num_disks); |
2018 | printf(" RAID Devices : %d\n", mpb->num_raid_devs); | |
ef5c214e MK |
2019 | print_imsm_disk(__get_imsm_disk(mpb, super->disks->index), |
2020 | super->disks->index, reserved, super->sector_size); | |
8d67477f | 2021 | if (get_imsm_bbm_log_size(super->bbm_log)) { |
604b746f JD |
2022 | struct bbm_log *log = super->bbm_log; |
2023 | ||
2024 | printf("\n"); | |
2025 | printf("Bad Block Management Log:\n"); | |
2026 | printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size)); | |
2027 | printf(" Signature : %x\n", __le32_to_cpu(log->signature)); | |
2028 | printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count)); | |
604b746f | 2029 | } |
44470971 DW |
2030 | for (i = 0; i < mpb->num_raid_devs; i++) { |
2031 | struct mdinfo info; | |
2032 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
2033 | ||
2034 | super->current_vol = i; | |
a5d85af7 | 2035 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 2036 | fname_from_uuid(st, &info, nbuf, ':'); |
c47b0ff6 | 2037 | print_imsm_dev(super, dev, nbuf + 5, super->disks->index); |
44470971 | 2038 | } |
cdddbdbc DW |
2039 | for (i = 0; i < mpb->num_disks; i++) { |
2040 | if (i == super->disks->index) | |
2041 | continue; | |
ef5c214e MK |
2042 | print_imsm_disk(__get_imsm_disk(mpb, i), i, reserved, |
2043 | super->sector_size); | |
cdddbdbc | 2044 | } |
94827db3 | 2045 | |
0ec1f4e8 DW |
2046 | for (dl = super->disks; dl; dl = dl->next) |
2047 | if (dl->index == -1) | |
ef5c214e MK |
2048 | print_imsm_disk(&dl->disk, -1, reserved, |
2049 | super->sector_size); | |
520e69e2 AK |
2050 | |
2051 | examine_migr_rec_imsm(super); | |
cdddbdbc DW |
2052 | } |
2053 | ||
061f2c6a | 2054 | static void brief_examine_super_imsm(struct supertype *st, int verbose) |
cdddbdbc | 2055 | { |
27fd6274 | 2056 | /* We just write a generic IMSM ARRAY entry */ |
ff54de6e N |
2057 | struct mdinfo info; |
2058 | char nbuf[64]; | |
1e7bc0ed | 2059 | struct intel_super *super = st->sb; |
1e7bc0ed | 2060 | |
0d5a423f DW |
2061 | if (!super->anchor->num_raid_devs) { |
2062 | printf("ARRAY metadata=imsm\n"); | |
1e7bc0ed | 2063 | return; |
0d5a423f | 2064 | } |
ff54de6e | 2065 | |
a5d85af7 | 2066 | getinfo_super_imsm(st, &info, NULL); |
4737ae25 N |
2067 | fname_from_uuid(st, &info, nbuf, ':'); |
2068 | printf("ARRAY metadata=imsm UUID=%s\n", nbuf + 5); | |
2069 | } | |
2070 | ||
2071 | static void brief_examine_subarrays_imsm(struct supertype *st, int verbose) | |
2072 | { | |
2073 | /* We just write a generic IMSM ARRAY entry */ | |
2074 | struct mdinfo info; | |
2075 | char nbuf[64]; | |
2076 | char nbuf1[64]; | |
2077 | struct intel_super *super = st->sb; | |
2078 | int i; | |
2079 | ||
2080 | if (!super->anchor->num_raid_devs) | |
2081 | return; | |
2082 | ||
a5d85af7 | 2083 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 2084 | fname_from_uuid(st, &info, nbuf, ':'); |
1e7bc0ed DW |
2085 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
2086 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
2087 | ||
2088 | super->current_vol = i; | |
a5d85af7 | 2089 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 2090 | fname_from_uuid(st, &info, nbuf1, ':'); |
1124b3cf | 2091 | printf("ARRAY /dev/md/%.16s container=%s member=%d UUID=%s\n", |
cf8de691 | 2092 | dev->volume, nbuf + 5, i, nbuf1 + 5); |
1e7bc0ed | 2093 | } |
cdddbdbc DW |
2094 | } |
2095 | ||
9d84c8ea DW |
2096 | static void export_examine_super_imsm(struct supertype *st) |
2097 | { | |
2098 | struct intel_super *super = st->sb; | |
2099 | struct imsm_super *mpb = super->anchor; | |
2100 | struct mdinfo info; | |
2101 | char nbuf[64]; | |
2102 | ||
a5d85af7 | 2103 | getinfo_super_imsm(st, &info, NULL); |
9d84c8ea DW |
2104 | fname_from_uuid(st, &info, nbuf, ':'); |
2105 | printf("MD_METADATA=imsm\n"); | |
2106 | printf("MD_LEVEL=container\n"); | |
2107 | printf("MD_UUID=%s\n", nbuf+5); | |
2108 | printf("MD_DEVICES=%u\n", mpb->num_disks); | |
2109 | } | |
2110 | ||
74db60b0 N |
2111 | static int copy_metadata_imsm(struct supertype *st, int from, int to) |
2112 | { | |
f36a9ecd | 2113 | /* The second last sector of the device contains |
74db60b0 N |
2114 | * the "struct imsm_super" metadata. |
2115 | * This contains mpb_size which is the size in bytes of the | |
2116 | * extended metadata. This is located immediately before | |
2117 | * the imsm_super. | |
2118 | * We want to read all that, plus the last sector which | |
2119 | * may contain a migration record, and write it all | |
2120 | * to the target. | |
2121 | */ | |
2122 | void *buf; | |
2123 | unsigned long long dsize, offset; | |
2124 | int sectors; | |
2125 | struct imsm_super *sb; | |
f36a9ecd PB |
2126 | struct intel_super *super = st->sb; |
2127 | unsigned int sector_size = super->sector_size; | |
2128 | unsigned int written = 0; | |
74db60b0 | 2129 | |
de44e46f | 2130 | if (posix_memalign(&buf, MAX_SECTOR_SIZE, MAX_SECTOR_SIZE) != 0) |
74db60b0 N |
2131 | return 1; |
2132 | ||
2133 | if (!get_dev_size(from, NULL, &dsize)) | |
2134 | goto err; | |
2135 | ||
f36a9ecd | 2136 | if (lseek64(from, dsize-(2*sector_size), 0) < 0) |
74db60b0 | 2137 | goto err; |
466070ad | 2138 | if ((unsigned int)read(from, buf, sector_size) != sector_size) |
74db60b0 N |
2139 | goto err; |
2140 | sb = buf; | |
2141 | if (strncmp((char*)sb->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) | |
2142 | goto err; | |
2143 | ||
f36a9ecd PB |
2144 | sectors = mpb_sectors(sb, sector_size) + 2; |
2145 | offset = dsize - sectors * sector_size; | |
74db60b0 N |
2146 | if (lseek64(from, offset, 0) < 0 || |
2147 | lseek64(to, offset, 0) < 0) | |
2148 | goto err; | |
f36a9ecd PB |
2149 | while (written < sectors * sector_size) { |
2150 | int n = sectors*sector_size - written; | |
74db60b0 N |
2151 | if (n > 4096) |
2152 | n = 4096; | |
2153 | if (read(from, buf, n) != n) | |
2154 | goto err; | |
2155 | if (write(to, buf, n) != n) | |
2156 | goto err; | |
2157 | written += n; | |
2158 | } | |
2159 | free(buf); | |
2160 | return 0; | |
2161 | err: | |
2162 | free(buf); | |
2163 | return 1; | |
2164 | } | |
2165 | ||
cdddbdbc DW |
2166 | static void detail_super_imsm(struct supertype *st, char *homehost) |
2167 | { | |
3ebe00a1 DW |
2168 | struct mdinfo info; |
2169 | char nbuf[64]; | |
2170 | ||
a5d85af7 | 2171 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 2172 | fname_from_uuid(st, &info, nbuf, ':'); |
65884368 | 2173 | printf("\n UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
2174 | } |
2175 | ||
2176 | static void brief_detail_super_imsm(struct supertype *st) | |
2177 | { | |
ff54de6e N |
2178 | struct mdinfo info; |
2179 | char nbuf[64]; | |
a5d85af7 | 2180 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 2181 | fname_from_uuid(st, &info, nbuf, ':'); |
ff54de6e | 2182 | printf(" UUID=%s", nbuf + 5); |
cdddbdbc | 2183 | } |
d665cc31 DW |
2184 | |
2185 | static int imsm_read_serial(int fd, char *devname, __u8 *serial); | |
2186 | static void fd2devname(int fd, char *name); | |
2187 | ||
120dc887 | 2188 | static int ahci_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose) |
d665cc31 | 2189 | { |
120dc887 LM |
2190 | /* dump an unsorted list of devices attached to AHCI Intel storage |
2191 | * controller, as well as non-connected ports | |
d665cc31 DW |
2192 | */ |
2193 | int hba_len = strlen(hba_path) + 1; | |
2194 | struct dirent *ent; | |
2195 | DIR *dir; | |
2196 | char *path = NULL; | |
2197 | int err = 0; | |
2198 | unsigned long port_mask = (1 << port_count) - 1; | |
2199 | ||
f21e18ca | 2200 | if (port_count > (int)sizeof(port_mask) * 8) { |
ba728be7 | 2201 | if (verbose > 0) |
e7b84f9d | 2202 | pr_err("port_count %d out of range\n", port_count); |
d665cc31 DW |
2203 | return 2; |
2204 | } | |
2205 | ||
2206 | /* scroll through /sys/dev/block looking for devices attached to | |
2207 | * this hba | |
2208 | */ | |
2209 | dir = opendir("/sys/dev/block"); | |
1a6dd6b9 PB |
2210 | if (!dir) |
2211 | return 1; | |
2212 | ||
2213 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
d665cc31 DW |
2214 | int fd; |
2215 | char model[64]; | |
2216 | char vendor[64]; | |
2217 | char buf[1024]; | |
2218 | int major, minor; | |
2219 | char *device; | |
2220 | char *c; | |
2221 | int port; | |
2222 | int type; | |
2223 | ||
2224 | if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2) | |
2225 | continue; | |
2226 | path = devt_to_devpath(makedev(major, minor)); | |
2227 | if (!path) | |
2228 | continue; | |
2229 | if (!path_attached_to_hba(path, hba_path)) { | |
2230 | free(path); | |
2231 | path = NULL; | |
2232 | continue; | |
2233 | } | |
2234 | ||
2235 | /* retrieve the scsi device type */ | |
2236 | if (asprintf(&device, "/sys/dev/block/%d:%d/device/xxxxxxx", major, minor) < 0) { | |
ba728be7 | 2237 | if (verbose > 0) |
e7b84f9d | 2238 | pr_err("failed to allocate 'device'\n"); |
d665cc31 DW |
2239 | err = 2; |
2240 | break; | |
2241 | } | |
2242 | sprintf(device, "/sys/dev/block/%d:%d/device/type", major, minor); | |
193b6c0b | 2243 | if (load_sys(device, buf, sizeof(buf)) != 0) { |
ba728be7 | 2244 | if (verbose > 0) |
e7b84f9d | 2245 | pr_err("failed to read device type for %s\n", |
d665cc31 DW |
2246 | path); |
2247 | err = 2; | |
2248 | free(device); | |
2249 | break; | |
2250 | } | |
2251 | type = strtoul(buf, NULL, 10); | |
2252 | ||
2253 | /* if it's not a disk print the vendor and model */ | |
2254 | if (!(type == 0 || type == 7 || type == 14)) { | |
2255 | vendor[0] = '\0'; | |
2256 | model[0] = '\0'; | |
2257 | sprintf(device, "/sys/dev/block/%d:%d/device/vendor", major, minor); | |
193b6c0b | 2258 | if (load_sys(device, buf, sizeof(buf)) == 0) { |
d665cc31 DW |
2259 | strncpy(vendor, buf, sizeof(vendor)); |
2260 | vendor[sizeof(vendor) - 1] = '\0'; | |
2261 | c = (char *) &vendor[sizeof(vendor) - 1]; | |
2262 | while (isspace(*c) || *c == '\0') | |
2263 | *c-- = '\0'; | |
2264 | ||
2265 | } | |
2266 | sprintf(device, "/sys/dev/block/%d:%d/device/model", major, minor); | |
193b6c0b | 2267 | if (load_sys(device, buf, sizeof(buf)) == 0) { |
d665cc31 DW |
2268 | strncpy(model, buf, sizeof(model)); |
2269 | model[sizeof(model) - 1] = '\0'; | |
2270 | c = (char *) &model[sizeof(model) - 1]; | |
2271 | while (isspace(*c) || *c == '\0') | |
2272 | *c-- = '\0'; | |
2273 | } | |
2274 | ||
2275 | if (vendor[0] && model[0]) | |
2276 | sprintf(buf, "%.64s %.64s", vendor, model); | |
2277 | else | |
2278 | switch (type) { /* numbers from hald/linux/device.c */ | |
2279 | case 1: sprintf(buf, "tape"); break; | |
2280 | case 2: sprintf(buf, "printer"); break; | |
2281 | case 3: sprintf(buf, "processor"); break; | |
2282 | case 4: | |
2283 | case 5: sprintf(buf, "cdrom"); break; | |
2284 | case 6: sprintf(buf, "scanner"); break; | |
2285 | case 8: sprintf(buf, "media_changer"); break; | |
2286 | case 9: sprintf(buf, "comm"); break; | |
2287 | case 12: sprintf(buf, "raid"); break; | |
2288 | default: sprintf(buf, "unknown"); | |
2289 | } | |
2290 | } else | |
2291 | buf[0] = '\0'; | |
2292 | free(device); | |
2293 | ||
2294 | /* chop device path to 'host%d' and calculate the port number */ | |
2295 | c = strchr(&path[hba_len], '/'); | |
4e5e717d | 2296 | if (!c) { |
ba728be7 | 2297 | if (verbose > 0) |
e7b84f9d | 2298 | pr_err("%s - invalid path name\n", path + hba_len); |
4e5e717d AW |
2299 | err = 2; |
2300 | break; | |
2301 | } | |
d665cc31 | 2302 | *c = '\0'; |
0858eccf AP |
2303 | if ((sscanf(&path[hba_len], "ata%d", &port) == 1) || |
2304 | ((sscanf(&path[hba_len], "host%d", &port) == 1))) | |
d665cc31 DW |
2305 | port -= host_base; |
2306 | else { | |
ba728be7 | 2307 | if (verbose > 0) { |
d665cc31 | 2308 | *c = '/'; /* repair the full string */ |
e7b84f9d | 2309 | pr_err("failed to determine port number for %s\n", |
d665cc31 DW |
2310 | path); |
2311 | } | |
2312 | err = 2; | |
2313 | break; | |
2314 | } | |
2315 | ||
2316 | /* mark this port as used */ | |
2317 | port_mask &= ~(1 << port); | |
2318 | ||
2319 | /* print out the device information */ | |
2320 | if (buf[0]) { | |
2321 | printf(" Port%d : - non-disk device (%s) -\n", port, buf); | |
2322 | continue; | |
2323 | } | |
2324 | ||
2325 | fd = dev_open(ent->d_name, O_RDONLY); | |
2326 | if (fd < 0) | |
2327 | printf(" Port%d : - disk info unavailable -\n", port); | |
2328 | else { | |
2329 | fd2devname(fd, buf); | |
2330 | printf(" Port%d : %s", port, buf); | |
2331 | if (imsm_read_serial(fd, NULL, (__u8 *) buf) == 0) | |
664d5325 | 2332 | printf(" (%.*s)\n", MAX_RAID_SERIAL_LEN, buf); |
d665cc31 | 2333 | else |
664d5325 | 2334 | printf(" ()\n"); |
4dab422a | 2335 | close(fd); |
d665cc31 | 2336 | } |
d665cc31 DW |
2337 | free(path); |
2338 | path = NULL; | |
2339 | } | |
2340 | if (path) | |
2341 | free(path); | |
2342 | if (dir) | |
2343 | closedir(dir); | |
2344 | if (err == 0) { | |
2345 | int i; | |
2346 | ||
2347 | for (i = 0; i < port_count; i++) | |
2348 | if (port_mask & (1 << i)) | |
2349 | printf(" Port%d : - no device attached -\n", i); | |
2350 | } | |
2351 | ||
2352 | return err; | |
2353 | } | |
2354 | ||
b5eece69 | 2355 | static int print_vmd_attached_devs(struct sys_dev *hba) |
60f0f54d PB |
2356 | { |
2357 | struct dirent *ent; | |
2358 | DIR *dir; | |
2359 | char path[292]; | |
2360 | char link[256]; | |
2361 | char *c, *rp; | |
2362 | ||
2363 | if (hba->type != SYS_DEV_VMD) | |
b5eece69 | 2364 | return 1; |
60f0f54d PB |
2365 | |
2366 | /* scroll through /sys/dev/block looking for devices attached to | |
2367 | * this hba | |
2368 | */ | |
2369 | dir = opendir("/sys/bus/pci/drivers/nvme"); | |
b9135011 | 2370 | if (!dir) |
b5eece69 | 2371 | return 1; |
b9135011 JS |
2372 | |
2373 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
60f0f54d PB |
2374 | int n; |
2375 | ||
2376 | /* is 'ent' a device? check that the 'subsystem' link exists and | |
2377 | * that its target matches 'bus' | |
2378 | */ | |
2379 | sprintf(path, "/sys/bus/pci/drivers/nvme/%s/subsystem", | |
2380 | ent->d_name); | |
2381 | n = readlink(path, link, sizeof(link)); | |
2382 | if (n < 0 || n >= (int)sizeof(link)) | |
2383 | continue; | |
2384 | link[n] = '\0'; | |
2385 | c = strrchr(link, '/'); | |
2386 | if (!c) | |
2387 | continue; | |
2388 | if (strncmp("pci", c+1, strlen("pci")) != 0) | |
2389 | continue; | |
2390 | ||
2391 | sprintf(path, "/sys/bus/pci/drivers/nvme/%s", ent->d_name); | |
60f0f54d PB |
2392 | |
2393 | rp = realpath(path, NULL); | |
2394 | if (!rp) | |
2395 | continue; | |
2396 | ||
2397 | if (path_attached_to_hba(rp, hba->path)) { | |
2398 | printf(" NVMe under VMD : %s\n", rp); | |
2399 | } | |
2400 | free(rp); | |
2401 | } | |
2402 | ||
b9135011 | 2403 | closedir(dir); |
b5eece69 | 2404 | return 0; |
60f0f54d PB |
2405 | } |
2406 | ||
120dc887 LM |
2407 | static void print_found_intel_controllers(struct sys_dev *elem) |
2408 | { | |
2409 | for (; elem; elem = elem->next) { | |
e7b84f9d | 2410 | pr_err("found Intel(R) "); |
120dc887 LM |
2411 | if (elem->type == SYS_DEV_SATA) |
2412 | fprintf(stderr, "SATA "); | |
155cbb4c LM |
2413 | else if (elem->type == SYS_DEV_SAS) |
2414 | fprintf(stderr, "SAS "); | |
0858eccf AP |
2415 | else if (elem->type == SYS_DEV_NVME) |
2416 | fprintf(stderr, "NVMe "); | |
60f0f54d PB |
2417 | |
2418 | if (elem->type == SYS_DEV_VMD) | |
2419 | fprintf(stderr, "VMD domain"); | |
2420 | else | |
2421 | fprintf(stderr, "RAID controller"); | |
2422 | ||
120dc887 LM |
2423 | if (elem->pci_id) |
2424 | fprintf(stderr, " at %s", elem->pci_id); | |
2425 | fprintf(stderr, ".\n"); | |
2426 | } | |
2427 | fflush(stderr); | |
2428 | } | |
2429 | ||
120dc887 LM |
2430 | static int ahci_get_port_count(const char *hba_path, int *port_count) |
2431 | { | |
2432 | struct dirent *ent; | |
2433 | DIR *dir; | |
2434 | int host_base = -1; | |
2435 | ||
2436 | *port_count = 0; | |
2437 | if ((dir = opendir(hba_path)) == NULL) | |
2438 | return -1; | |
2439 | ||
2440 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
2441 | int host; | |
2442 | ||
0858eccf AP |
2443 | if ((sscanf(ent->d_name, "ata%d", &host) != 1) && |
2444 | ((sscanf(ent->d_name, "host%d", &host) != 1))) | |
120dc887 LM |
2445 | continue; |
2446 | if (*port_count == 0) | |
2447 | host_base = host; | |
2448 | else if (host < host_base) | |
2449 | host_base = host; | |
2450 | ||
2451 | if (host + 1 > *port_count + host_base) | |
2452 | *port_count = host + 1 - host_base; | |
2453 | } | |
2454 | closedir(dir); | |
2455 | return host_base; | |
2456 | } | |
2457 | ||
a891a3c2 LM |
2458 | static void print_imsm_capability(const struct imsm_orom *orom) |
2459 | { | |
0858eccf AP |
2460 | printf(" Platform : Intel(R) "); |
2461 | if (orom->capabilities == 0 && orom->driver_features == 0) | |
2462 | printf("Matrix Storage Manager\n"); | |
ab0c6bb9 AP |
2463 | else if (imsm_orom_is_enterprise(orom) && orom->major_ver >= 6) |
2464 | printf("Virtual RAID on CPU\n"); | |
0858eccf AP |
2465 | else |
2466 | printf("Rapid Storage Technology%s\n", | |
2467 | imsm_orom_is_enterprise(orom) ? " enterprise" : ""); | |
2468 | if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build) | |
2469 | printf(" Version : %d.%d.%d.%d\n", orom->major_ver, | |
2470 | orom->minor_ver, orom->hotfix_ver, orom->build); | |
a891a3c2 LM |
2471 | printf(" RAID Levels :%s%s%s%s%s\n", |
2472 | imsm_orom_has_raid0(orom) ? " raid0" : "", | |
2473 | imsm_orom_has_raid1(orom) ? " raid1" : "", | |
2474 | imsm_orom_has_raid1e(orom) ? " raid1e" : "", | |
2475 | imsm_orom_has_raid10(orom) ? " raid10" : "", | |
2476 | imsm_orom_has_raid5(orom) ? " raid5" : ""); | |
2477 | printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n", | |
2478 | imsm_orom_has_chunk(orom, 2) ? " 2k" : "", | |
2479 | imsm_orom_has_chunk(orom, 4) ? " 4k" : "", | |
2480 | imsm_orom_has_chunk(orom, 8) ? " 8k" : "", | |
2481 | imsm_orom_has_chunk(orom, 16) ? " 16k" : "", | |
2482 | imsm_orom_has_chunk(orom, 32) ? " 32k" : "", | |
2483 | imsm_orom_has_chunk(orom, 64) ? " 64k" : "", | |
2484 | imsm_orom_has_chunk(orom, 128) ? " 128k" : "", | |
2485 | imsm_orom_has_chunk(orom, 256) ? " 256k" : "", | |
2486 | imsm_orom_has_chunk(orom, 512) ? " 512k" : "", | |
2487 | imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "", | |
2488 | imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "", | |
2489 | imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "", | |
2490 | imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "", | |
2491 | imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "", | |
2492 | imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "", | |
2493 | imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : ""); | |
29cd0821 CA |
2494 | printf(" 2TB volumes :%s supported\n", |
2495 | (orom->attr & IMSM_OROM_ATTR_2TB)?"":" not"); | |
2496 | printf(" 2TB disks :%s supported\n", | |
2497 | (orom->attr & IMSM_OROM_ATTR_2TB_DISK)?"":" not"); | |
0e7f69a8 | 2498 | printf(" Max Disks : %d\n", orom->tds); |
0858eccf AP |
2499 | printf(" Max Volumes : %d per array, %d per %s\n", |
2500 | orom->vpa, orom->vphba, | |
2501 | imsm_orom_is_nvme(orom) ? "platform" : "controller"); | |
a891a3c2 LM |
2502 | return; |
2503 | } | |
2504 | ||
e50cf220 MN |
2505 | static void print_imsm_capability_export(const struct imsm_orom *orom) |
2506 | { | |
2507 | printf("MD_FIRMWARE_TYPE=imsm\n"); | |
0858eccf AP |
2508 | if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build) |
2509 | printf("IMSM_VERSION=%d.%d.%d.%d\n", orom->major_ver, orom->minor_ver, | |
2510 | orom->hotfix_ver, orom->build); | |
e50cf220 MN |
2511 | printf("IMSM_SUPPORTED_RAID_LEVELS=%s%s%s%s%s\n", |
2512 | imsm_orom_has_raid0(orom) ? "raid0 " : "", | |
2513 | imsm_orom_has_raid1(orom) ? "raid1 " : "", | |
2514 | imsm_orom_has_raid1e(orom) ? "raid1e " : "", | |
2515 | imsm_orom_has_raid5(orom) ? "raid10 " : "", | |
2516 | imsm_orom_has_raid10(orom) ? "raid5 " : ""); | |
2517 | printf("IMSM_SUPPORTED_CHUNK_SIZES=%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n", | |
2518 | imsm_orom_has_chunk(orom, 2) ? "2k " : "", | |
2519 | imsm_orom_has_chunk(orom, 4) ? "4k " : "", | |
2520 | imsm_orom_has_chunk(orom, 8) ? "8k " : "", | |
2521 | imsm_orom_has_chunk(orom, 16) ? "16k " : "", | |
2522 | imsm_orom_has_chunk(orom, 32) ? "32k " : "", | |
2523 | imsm_orom_has_chunk(orom, 64) ? "64k " : "", | |
2524 | imsm_orom_has_chunk(orom, 128) ? "128k " : "", | |
2525 | imsm_orom_has_chunk(orom, 256) ? "256k " : "", | |
2526 | imsm_orom_has_chunk(orom, 512) ? "512k " : "", | |
2527 | imsm_orom_has_chunk(orom, 1024*1) ? "1M " : "", | |
2528 | imsm_orom_has_chunk(orom, 1024*2) ? "2M " : "", | |
2529 | imsm_orom_has_chunk(orom, 1024*4) ? "4M " : "", | |
2530 | imsm_orom_has_chunk(orom, 1024*8) ? "8M " : "", | |
2531 | imsm_orom_has_chunk(orom, 1024*16) ? "16M " : "", | |
2532 | imsm_orom_has_chunk(orom, 1024*32) ? "32M " : "", | |
2533 | imsm_orom_has_chunk(orom, 1024*64) ? "64M " : ""); | |
2534 | printf("IMSM_2TB_VOLUMES=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB) ? "yes" : "no"); | |
2535 | printf("IMSM_2TB_DISKS=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB_DISK) ? "yes" : "no"); | |
2536 | printf("IMSM_MAX_DISKS=%d\n",orom->tds); | |
2537 | printf("IMSM_MAX_VOLUMES_PER_ARRAY=%d\n",orom->vpa); | |
2538 | printf("IMSM_MAX_VOLUMES_PER_CONTROLLER=%d\n",orom->vphba); | |
2539 | } | |
2540 | ||
9eafa1de | 2541 | static int detail_platform_imsm(int verbose, int enumerate_only, char *controller_path) |
d665cc31 DW |
2542 | { |
2543 | /* There are two components to imsm platform support, the ahci SATA | |
2544 | * controller and the option-rom. To find the SATA controller we | |
2545 | * simply look in /sys/bus/pci/drivers/ahci to see if an ahci | |
2546 | * controller with the Intel vendor id is present. This approach | |
2547 | * allows mdadm to leverage the kernel's ahci detection logic, with the | |
2548 | * caveat that if ahci.ko is not loaded mdadm will not be able to | |
2549 | * detect platform raid capabilities. The option-rom resides in a | |
2550 | * platform "Adapter ROM". We scan for its signature to retrieve the | |
2551 | * platform capabilities. If raid support is disabled in the BIOS the | |
2552 | * option-rom capability structure will not be available. | |
2553 | */ | |
d665cc31 | 2554 | struct sys_dev *list, *hba; |
d665cc31 DW |
2555 | int host_base = 0; |
2556 | int port_count = 0; | |
9eafa1de | 2557 | int result=1; |
d665cc31 | 2558 | |
5615172f | 2559 | if (enumerate_only) { |
a891a3c2 | 2560 | if (check_env("IMSM_NO_PLATFORM")) |
5615172f | 2561 | return 0; |
a891a3c2 LM |
2562 | list = find_intel_devices(); |
2563 | if (!list) | |
2564 | return 2; | |
2565 | for (hba = list; hba; hba = hba->next) { | |
6b781d33 AP |
2566 | if (find_imsm_capability(hba)) { |
2567 | result = 0; | |
a891a3c2 LM |
2568 | break; |
2569 | } | |
9eafa1de | 2570 | else |
6b781d33 | 2571 | result = 2; |
a891a3c2 | 2572 | } |
a891a3c2 | 2573 | return result; |
5615172f DW |
2574 | } |
2575 | ||
155cbb4c LM |
2576 | list = find_intel_devices(); |
2577 | if (!list) { | |
ba728be7 | 2578 | if (verbose > 0) |
7a862a02 | 2579 | pr_err("no active Intel(R) RAID controller found.\n"); |
d665cc31 | 2580 | return 2; |
ba728be7 | 2581 | } else if (verbose > 0) |
155cbb4c | 2582 | print_found_intel_controllers(list); |
d665cc31 | 2583 | |
a891a3c2 | 2584 | for (hba = list; hba; hba = hba->next) { |
0858eccf | 2585 | if (controller_path && (compare_paths(hba->path, controller_path) != 0)) |
9eafa1de | 2586 | continue; |
0858eccf | 2587 | if (!find_imsm_capability(hba)) { |
60f0f54d | 2588 | char buf[PATH_MAX]; |
e7b84f9d | 2589 | pr_err("imsm capabilities not found for controller: %s (type %s)\n", |
60f0f54d PB |
2590 | hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path, |
2591 | get_sys_dev_type(hba->type)); | |
0858eccf AP |
2592 | continue; |
2593 | } | |
2594 | result = 0; | |
2595 | } | |
2596 | ||
2597 | if (controller_path && result == 1) { | |
2598 | pr_err("no active Intel(R) RAID controller found under %s\n", | |
2599 | controller_path); | |
2600 | return result; | |
2601 | } | |
2602 | ||
5e1d6128 | 2603 | const struct orom_entry *entry; |
0858eccf | 2604 | |
5e1d6128 | 2605 | for (entry = orom_entries; entry; entry = entry->next) { |
60f0f54d | 2606 | if (entry->type == SYS_DEV_VMD) { |
07cb1e57 | 2607 | print_imsm_capability(&entry->orom); |
32716c51 PB |
2608 | printf(" 3rd party NVMe :%s supported\n", |
2609 | imsm_orom_has_tpv_support(&entry->orom)?"":" not"); | |
60f0f54d PB |
2610 | for (hba = list; hba; hba = hba->next) { |
2611 | if (hba->type == SYS_DEV_VMD) { | |
2612 | char buf[PATH_MAX]; | |
60f0f54d PB |
2613 | printf(" I/O Controller : %s (%s)\n", |
2614 | vmd_domain_to_controller(hba, buf), get_sys_dev_type(hba->type)); | |
b5eece69 PB |
2615 | if (print_vmd_attached_devs(hba)) { |
2616 | if (verbose > 0) | |
2617 | pr_err("failed to get devices attached to VMD domain.\n"); | |
2618 | result |= 2; | |
2619 | } | |
60f0f54d PB |
2620 | } |
2621 | } | |
07cb1e57 | 2622 | printf("\n"); |
60f0f54d PB |
2623 | continue; |
2624 | } | |
0858eccf | 2625 | |
60f0f54d PB |
2626 | print_imsm_capability(&entry->orom); |
2627 | if (entry->type == SYS_DEV_NVME) { | |
0858eccf AP |
2628 | for (hba = list; hba; hba = hba->next) { |
2629 | if (hba->type == SYS_DEV_NVME) | |
2630 | printf(" NVMe Device : %s\n", hba->path); | |
2631 | } | |
60f0f54d | 2632 | printf("\n"); |
0858eccf AP |
2633 | continue; |
2634 | } | |
2635 | ||
2636 | struct devid_list *devid; | |
5e1d6128 | 2637 | for (devid = entry->devid_list; devid; devid = devid->next) { |
0858eccf AP |
2638 | hba = device_by_id(devid->devid); |
2639 | if (!hba) | |
2640 | continue; | |
2641 | ||
9eafa1de MN |
2642 | printf(" I/O Controller : %s (%s)\n", |
2643 | hba->path, get_sys_dev_type(hba->type)); | |
2644 | if (hba->type == SYS_DEV_SATA) { | |
2645 | host_base = ahci_get_port_count(hba->path, &port_count); | |
2646 | if (ahci_enumerate_ports(hba->path, port_count, host_base, verbose)) { | |
2647 | if (verbose > 0) | |
7a862a02 | 2648 | pr_err("failed to enumerate ports on SATA controller at %s.\n", hba->pci_id); |
9eafa1de MN |
2649 | result |= 2; |
2650 | } | |
120dc887 LM |
2651 | } |
2652 | } | |
0858eccf | 2653 | printf("\n"); |
d665cc31 | 2654 | } |
155cbb4c | 2655 | |
120dc887 | 2656 | return result; |
d665cc31 | 2657 | } |
e50cf220 | 2658 | |
9eafa1de | 2659 | static int export_detail_platform_imsm(int verbose, char *controller_path) |
e50cf220 | 2660 | { |
e50cf220 MN |
2661 | struct sys_dev *list, *hba; |
2662 | int result=1; | |
2663 | ||
2664 | list = find_intel_devices(); | |
2665 | if (!list) { | |
2666 | if (verbose > 0) | |
2667 | pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_INTEL_DEVICES\n"); | |
2668 | result = 2; | |
e50cf220 MN |
2669 | return result; |
2670 | } | |
2671 | ||
2672 | for (hba = list; hba; hba = hba->next) { | |
9eafa1de MN |
2673 | if (controller_path && (compare_paths(hba->path,controller_path) != 0)) |
2674 | continue; | |
60f0f54d PB |
2675 | if (!find_imsm_capability(hba) && verbose > 0) { |
2676 | char buf[PATH_MAX]; | |
2677 | pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_IMSM_CAPABLE_DEVICE_UNDER_%s\n", | |
2678 | hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path); | |
2679 | } | |
0858eccf | 2680 | else |
e50cf220 | 2681 | result = 0; |
e50cf220 MN |
2682 | } |
2683 | ||
5e1d6128 | 2684 | const struct orom_entry *entry; |
0858eccf | 2685 | |
60f0f54d PB |
2686 | for (entry = orom_entries; entry; entry = entry->next) { |
2687 | if (entry->type == SYS_DEV_VMD) { | |
2688 | for (hba = list; hba; hba = hba->next) | |
2689 | print_imsm_capability_export(&entry->orom); | |
2690 | continue; | |
2691 | } | |
5e1d6128 | 2692 | print_imsm_capability_export(&entry->orom); |
60f0f54d | 2693 | } |
0858eccf | 2694 | |
e50cf220 MN |
2695 | return result; |
2696 | } | |
2697 | ||
cdddbdbc DW |
2698 | static int match_home_imsm(struct supertype *st, char *homehost) |
2699 | { | |
5115ca67 DW |
2700 | /* the imsm metadata format does not specify any host |
2701 | * identification information. We return -1 since we can never | |
2702 | * confirm nor deny whether a given array is "meant" for this | |
148acb7b | 2703 | * host. We rely on compare_super and the 'family_num' fields to |
5115ca67 DW |
2704 | * exclude member disks that do not belong, and we rely on |
2705 | * mdadm.conf to specify the arrays that should be assembled. | |
2706 | * Auto-assembly may still pick up "foreign" arrays. | |
2707 | */ | |
cdddbdbc | 2708 | |
9362c1c8 | 2709 | return -1; |
cdddbdbc DW |
2710 | } |
2711 | ||
2712 | static void uuid_from_super_imsm(struct supertype *st, int uuid[4]) | |
2713 | { | |
51006d85 N |
2714 | /* The uuid returned here is used for: |
2715 | * uuid to put into bitmap file (Create, Grow) | |
2716 | * uuid for backup header when saving critical section (Grow) | |
2717 | * comparing uuids when re-adding a device into an array | |
2718 | * In these cases the uuid required is that of the data-array, | |
2719 | * not the device-set. | |
2720 | * uuid to recognise same set when adding a missing device back | |
2721 | * to an array. This is a uuid for the device-set. | |
1011e834 | 2722 | * |
51006d85 N |
2723 | * For each of these we can make do with a truncated |
2724 | * or hashed uuid rather than the original, as long as | |
2725 | * everyone agrees. | |
2726 | * In each case the uuid required is that of the data-array, | |
2727 | * not the device-set. | |
43dad3d6 | 2728 | */ |
51006d85 N |
2729 | /* imsm does not track uuid's so we synthesis one using sha1 on |
2730 | * - The signature (Which is constant for all imsm array, but no matter) | |
148acb7b | 2731 | * - the orig_family_num of the container |
51006d85 N |
2732 | * - the index number of the volume |
2733 | * - the 'serial' number of the volume. | |
2734 | * Hopefully these are all constant. | |
2735 | */ | |
2736 | struct intel_super *super = st->sb; | |
43dad3d6 | 2737 | |
51006d85 N |
2738 | char buf[20]; |
2739 | struct sha1_ctx ctx; | |
2740 | struct imsm_dev *dev = NULL; | |
148acb7b | 2741 | __u32 family_num; |
51006d85 | 2742 | |
148acb7b DW |
2743 | /* some mdadm versions failed to set ->orig_family_num, in which |
2744 | * case fall back to ->family_num. orig_family_num will be | |
2745 | * fixed up with the first metadata update. | |
2746 | */ | |
2747 | family_num = super->anchor->orig_family_num; | |
2748 | if (family_num == 0) | |
2749 | family_num = super->anchor->family_num; | |
51006d85 | 2750 | sha1_init_ctx(&ctx); |
92bd8f8d | 2751 | sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx); |
148acb7b | 2752 | sha1_process_bytes(&family_num, sizeof(__u32), &ctx); |
51006d85 N |
2753 | if (super->current_vol >= 0) |
2754 | dev = get_imsm_dev(super, super->current_vol); | |
2755 | if (dev) { | |
2756 | __u32 vol = super->current_vol; | |
2757 | sha1_process_bytes(&vol, sizeof(vol), &ctx); | |
2758 | sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx); | |
2759 | } | |
2760 | sha1_finish_ctx(&ctx, buf); | |
2761 | memcpy(uuid, buf, 4*4); | |
cdddbdbc DW |
2762 | } |
2763 | ||
0d481d37 | 2764 | #if 0 |
4f5bc454 DW |
2765 | static void |
2766 | get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p) | |
cdddbdbc | 2767 | { |
cdddbdbc DW |
2768 | __u8 *v = get_imsm_version(mpb); |
2769 | __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH; | |
2770 | char major[] = { 0, 0, 0 }; | |
2771 | char minor[] = { 0 ,0, 0 }; | |
2772 | char patch[] = { 0, 0, 0 }; | |
2773 | char *ver_parse[] = { major, minor, patch }; | |
2774 | int i, j; | |
2775 | ||
2776 | i = j = 0; | |
2777 | while (*v != '\0' && v < end) { | |
2778 | if (*v != '.' && j < 2) | |
2779 | ver_parse[i][j++] = *v; | |
2780 | else { | |
2781 | i++; | |
2782 | j = 0; | |
2783 | } | |
2784 | v++; | |
2785 | } | |
2786 | ||
4f5bc454 DW |
2787 | *m = strtol(minor, NULL, 0); |
2788 | *p = strtol(patch, NULL, 0); | |
2789 | } | |
0d481d37 | 2790 | #endif |
4f5bc454 | 2791 | |
1e5c6983 DW |
2792 | static __u32 migr_strip_blocks_resync(struct imsm_dev *dev) |
2793 | { | |
2794 | /* migr_strip_size when repairing or initializing parity */ | |
238c0a71 | 2795 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2796 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2797 | ||
2798 | switch (get_imsm_raid_level(map)) { | |
2799 | case 5: | |
2800 | case 10: | |
2801 | return chunk; | |
2802 | default: | |
2803 | return 128*1024 >> 9; | |
2804 | } | |
2805 | } | |
2806 | ||
2807 | static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev) | |
2808 | { | |
2809 | /* migr_strip_size when rebuilding a degraded disk, no idea why | |
2810 | * this is different than migr_strip_size_resync(), but it's good | |
2811 | * to be compatible | |
2812 | */ | |
238c0a71 | 2813 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
2814 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2815 | ||
2816 | switch (get_imsm_raid_level(map)) { | |
2817 | case 1: | |
2818 | case 10: | |
2819 | if (map->num_members % map->num_domains == 0) | |
2820 | return 128*1024 >> 9; | |
2821 | else | |
2822 | return chunk; | |
2823 | case 5: | |
2824 | return max((__u32) 64*1024 >> 9, chunk); | |
2825 | default: | |
2826 | return 128*1024 >> 9; | |
2827 | } | |
2828 | } | |
2829 | ||
2830 | static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev) | |
2831 | { | |
238c0a71 AK |
2832 | struct imsm_map *lo = get_imsm_map(dev, MAP_0); |
2833 | struct imsm_map *hi = get_imsm_map(dev, MAP_1); | |
1e5c6983 DW |
2834 | __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip); |
2835 | __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip); | |
2836 | ||
2837 | return max((__u32) 1, hi_chunk / lo_chunk); | |
2838 | } | |
2839 | ||
2840 | static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev) | |
2841 | { | |
238c0a71 | 2842 | struct imsm_map *lo = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2843 | int level = get_imsm_raid_level(lo); |
2844 | ||
2845 | if (level == 1 || level == 10) { | |
238c0a71 | 2846 | struct imsm_map *hi = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
2847 | |
2848 | return hi->num_domains; | |
2849 | } else | |
2850 | return num_stripes_per_unit_resync(dev); | |
2851 | } | |
2852 | ||
9529d343 | 2853 | static __u8 imsm_num_data_members(struct imsm_map *map) |
1e5c6983 DW |
2854 | { |
2855 | /* named 'imsm_' because raid0, raid1 and raid10 | |
2856 | * counter-intuitively have the same number of data disks | |
2857 | */ | |
1e5c6983 DW |
2858 | switch (get_imsm_raid_level(map)) { |
2859 | case 0: | |
36fd8ccc AK |
2860 | return map->num_members; |
2861 | break; | |
1e5c6983 DW |
2862 | case 1: |
2863 | case 10: | |
36fd8ccc | 2864 | return map->num_members/2; |
1e5c6983 DW |
2865 | case 5: |
2866 | return map->num_members - 1; | |
2867 | default: | |
1ade5cc1 | 2868 | dprintf("unsupported raid level\n"); |
1e5c6983 DW |
2869 | return 0; |
2870 | } | |
2871 | } | |
2872 | ||
44490938 MD |
2873 | static unsigned long long calc_component_size(struct imsm_map *map, |
2874 | struct imsm_dev *dev) | |
2875 | { | |
2876 | unsigned long long component_size; | |
2877 | unsigned long long dev_size = imsm_dev_size(dev); | |
a4f7290c | 2878 | long long calc_dev_size = 0; |
44490938 MD |
2879 | unsigned int member_disks = imsm_num_data_members(map); |
2880 | ||
2881 | if (member_disks == 0) | |
2882 | return 0; | |
2883 | ||
2884 | component_size = per_dev_array_size(map); | |
2885 | calc_dev_size = component_size * member_disks; | |
2886 | ||
2887 | /* Component size is rounded to 1MB so difference between size from | |
2888 | * metadata and size calculated from num_data_stripes equals up to | |
2889 | * 2048 blocks per each device. If the difference is higher it means | |
2890 | * that array size was expanded and num_data_stripes was not updated. | |
2891 | */ | |
a4f7290c | 2892 | if (llabs(calc_dev_size - (long long)dev_size) > |
44490938 MD |
2893 | (1 << SECT_PER_MB_SHIFT) * member_disks) { |
2894 | component_size = dev_size / member_disks; | |
2895 | dprintf("Invalid num_data_stripes in metadata; expected=%llu, found=%llu\n", | |
2896 | component_size / map->blocks_per_strip, | |
2897 | num_data_stripes(map)); | |
2898 | } | |
2899 | ||
2900 | return component_size; | |
2901 | } | |
2902 | ||
1e5c6983 DW |
2903 | static __u32 parity_segment_depth(struct imsm_dev *dev) |
2904 | { | |
238c0a71 | 2905 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2906 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2907 | ||
2908 | switch(get_imsm_raid_level(map)) { | |
2909 | case 1: | |
2910 | case 10: | |
2911 | return chunk * map->num_domains; | |
2912 | case 5: | |
2913 | return chunk * map->num_members; | |
2914 | default: | |
2915 | return chunk; | |
2916 | } | |
2917 | } | |
2918 | ||
2919 | static __u32 map_migr_block(struct imsm_dev *dev, __u32 block) | |
2920 | { | |
238c0a71 | 2921 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
2922 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2923 | __u32 strip = block / chunk; | |
2924 | ||
2925 | switch (get_imsm_raid_level(map)) { | |
2926 | case 1: | |
2927 | case 10: { | |
2928 | __u32 vol_strip = (strip * map->num_domains) + 1; | |
2929 | __u32 vol_stripe = vol_strip / map->num_members; | |
2930 | ||
2931 | return vol_stripe * chunk + block % chunk; | |
2932 | } case 5: { | |
2933 | __u32 stripe = strip / (map->num_members - 1); | |
2934 | ||
2935 | return stripe * chunk + block % chunk; | |
2936 | } | |
2937 | default: | |
2938 | return 0; | |
2939 | } | |
2940 | } | |
2941 | ||
c47b0ff6 AK |
2942 | static __u64 blocks_per_migr_unit(struct intel_super *super, |
2943 | struct imsm_dev *dev) | |
1e5c6983 DW |
2944 | { |
2945 | /* calculate the conversion factor between per member 'blocks' | |
2946 | * (md/{resync,rebuild}_start) and imsm migration units, return | |
2947 | * 0 for the 'not migrating' and 'unsupported migration' cases | |
2948 | */ | |
2949 | if (!dev->vol.migr_state) | |
2950 | return 0; | |
2951 | ||
2952 | switch (migr_type(dev)) { | |
c47b0ff6 AK |
2953 | case MIGR_GEN_MIGR: { |
2954 | struct migr_record *migr_rec = super->migr_rec; | |
2955 | return __le32_to_cpu(migr_rec->blocks_per_unit); | |
2956 | } | |
1e5c6983 DW |
2957 | case MIGR_VERIFY: |
2958 | case MIGR_REPAIR: | |
2959 | case MIGR_INIT: { | |
238c0a71 | 2960 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2961 | __u32 stripes_per_unit; |
2962 | __u32 blocks_per_unit; | |
2963 | __u32 parity_depth; | |
2964 | __u32 migr_chunk; | |
2965 | __u32 block_map; | |
2966 | __u32 block_rel; | |
2967 | __u32 segment; | |
2968 | __u32 stripe; | |
2969 | __u8 disks; | |
2970 | ||
2971 | /* yes, this is really the translation of migr_units to | |
2972 | * per-member blocks in the 'resync' case | |
2973 | */ | |
2974 | stripes_per_unit = num_stripes_per_unit_resync(dev); | |
2975 | migr_chunk = migr_strip_blocks_resync(dev); | |
9529d343 | 2976 | disks = imsm_num_data_members(map); |
1e5c6983 | 2977 | blocks_per_unit = stripes_per_unit * migr_chunk * disks; |
7b1ab482 | 2978 | stripe = __le16_to_cpu(map->blocks_per_strip) * disks; |
1e5c6983 DW |
2979 | segment = blocks_per_unit / stripe; |
2980 | block_rel = blocks_per_unit - segment * stripe; | |
2981 | parity_depth = parity_segment_depth(dev); | |
2982 | block_map = map_migr_block(dev, block_rel); | |
2983 | return block_map + parity_depth * segment; | |
2984 | } | |
2985 | case MIGR_REBUILD: { | |
2986 | __u32 stripes_per_unit; | |
2987 | __u32 migr_chunk; | |
2988 | ||
2989 | stripes_per_unit = num_stripes_per_unit_rebuild(dev); | |
2990 | migr_chunk = migr_strip_blocks_rebuild(dev); | |
2991 | return migr_chunk * stripes_per_unit; | |
2992 | } | |
1e5c6983 DW |
2993 | case MIGR_STATE_CHANGE: |
2994 | default: | |
2995 | return 0; | |
2996 | } | |
2997 | } | |
2998 | ||
c2c087e6 DW |
2999 | static int imsm_level_to_layout(int level) |
3000 | { | |
3001 | switch (level) { | |
3002 | case 0: | |
3003 | case 1: | |
3004 | return 0; | |
3005 | case 5: | |
3006 | case 6: | |
a380c027 | 3007 | return ALGORITHM_LEFT_ASYMMETRIC; |
c2c087e6 | 3008 | case 10: |
c92a2527 | 3009 | return 0x102; |
c2c087e6 | 3010 | } |
a18a888e | 3011 | return UnSet; |
c2c087e6 DW |
3012 | } |
3013 | ||
8e59f3d8 AK |
3014 | /******************************************************************************* |
3015 | * Function: read_imsm_migr_rec | |
3016 | * Description: Function reads imsm migration record from last sector of disk | |
3017 | * Parameters: | |
3018 | * fd : disk descriptor | |
3019 | * super : metadata info | |
3020 | * Returns: | |
3021 | * 0 : success, | |
3022 | * -1 : fail | |
3023 | ******************************************************************************/ | |
3024 | static int read_imsm_migr_rec(int fd, struct intel_super *super) | |
3025 | { | |
3026 | int ret_val = -1; | |
de44e46f | 3027 | unsigned int sector_size = super->sector_size; |
8e59f3d8 AK |
3028 | unsigned long long dsize; |
3029 | ||
3030 | get_dev_size(fd, NULL, &dsize); | |
de44e46f PB |
3031 | if (lseek64(fd, dsize - (sector_size*MIGR_REC_SECTOR_POSITION), |
3032 | SEEK_SET) < 0) { | |
e7b84f9d N |
3033 | pr_err("Cannot seek to anchor block: %s\n", |
3034 | strerror(errno)); | |
8e59f3d8 AK |
3035 | goto out; |
3036 | } | |
466070ad | 3037 | if ((unsigned int)read(fd, super->migr_rec_buf, |
de44e46f PB |
3038 | MIGR_REC_BUF_SECTORS*sector_size) != |
3039 | MIGR_REC_BUF_SECTORS*sector_size) { | |
e7b84f9d N |
3040 | pr_err("Cannot read migr record block: %s\n", |
3041 | strerror(errno)); | |
8e59f3d8 AK |
3042 | goto out; |
3043 | } | |
3044 | ret_val = 0; | |
de44e46f PB |
3045 | if (sector_size == 4096) |
3046 | convert_from_4k_imsm_migr_rec(super); | |
8e59f3d8 AK |
3047 | |
3048 | out: | |
3049 | return ret_val; | |
3050 | } | |
3051 | ||
3136abe5 AK |
3052 | static struct imsm_dev *imsm_get_device_during_migration( |
3053 | struct intel_super *super) | |
3054 | { | |
3055 | ||
3056 | struct intel_dev *dv; | |
3057 | ||
3058 | for (dv = super->devlist; dv; dv = dv->next) { | |
3059 | if (is_gen_migration(dv->dev)) | |
3060 | return dv->dev; | |
3061 | } | |
3062 | return NULL; | |
3063 | } | |
3064 | ||
8e59f3d8 AK |
3065 | /******************************************************************************* |
3066 | * Function: load_imsm_migr_rec | |
3067 | * Description: Function reads imsm migration record (it is stored at the last | |
3068 | * sector of disk) | |
3069 | * Parameters: | |
3070 | * super : imsm internal array info | |
3071 | * info : general array info | |
3072 | * Returns: | |
3073 | * 0 : success | |
3074 | * -1 : fail | |
4c965cc9 | 3075 | * -2 : no migration in progress |
8e59f3d8 AK |
3076 | ******************************************************************************/ |
3077 | static int load_imsm_migr_rec(struct intel_super *super, struct mdinfo *info) | |
3078 | { | |
3079 | struct mdinfo *sd; | |
594dc1b8 | 3080 | struct dl *dl; |
8e59f3d8 AK |
3081 | char nm[30]; |
3082 | int retval = -1; | |
3083 | int fd = -1; | |
3136abe5 | 3084 | struct imsm_dev *dev; |
594dc1b8 | 3085 | struct imsm_map *map; |
b4ab44d8 | 3086 | int slot = -1; |
3136abe5 AK |
3087 | |
3088 | /* find map under migration */ | |
3089 | dev = imsm_get_device_during_migration(super); | |
3090 | /* nothing to load,no migration in progress? | |
3091 | */ | |
3092 | if (dev == NULL) | |
4c965cc9 | 3093 | return -2; |
8e59f3d8 AK |
3094 | |
3095 | if (info) { | |
3096 | for (sd = info->devs ; sd ; sd = sd->next) { | |
3097 | /* read only from one of the first two slots */ | |
12fe93e9 TM |
3098 | if ((sd->disk.raid_disk < 0) || |
3099 | (sd->disk.raid_disk > 1)) | |
8e59f3d8 | 3100 | continue; |
3136abe5 | 3101 | |
8e59f3d8 AK |
3102 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); |
3103 | fd = dev_open(nm, O_RDONLY); | |
3104 | if (fd >= 0) | |
3105 | break; | |
3106 | } | |
3107 | } | |
3108 | if (fd < 0) { | |
12fe93e9 | 3109 | map = get_imsm_map(dev, MAP_0); |
8e59f3d8 | 3110 | for (dl = super->disks; dl; dl = dl->next) { |
3136abe5 AK |
3111 | /* skip spare and failed disks |
3112 | */ | |
3113 | if (dl->index < 0) | |
3114 | continue; | |
8e59f3d8 | 3115 | /* read only from one of the first two slots */ |
3136abe5 AK |
3116 | if (map) |
3117 | slot = get_imsm_disk_slot(map, dl->index); | |
089f9d79 | 3118 | if (map == NULL || slot > 1 || slot < 0) |
8e59f3d8 AK |
3119 | continue; |
3120 | sprintf(nm, "%d:%d", dl->major, dl->minor); | |
3121 | fd = dev_open(nm, O_RDONLY); | |
3122 | if (fd >= 0) | |
3123 | break; | |
3124 | } | |
3125 | } | |
3126 | if (fd < 0) | |
3127 | goto out; | |
3128 | retval = read_imsm_migr_rec(fd, super); | |
3129 | ||
3130 | out: | |
3131 | if (fd >= 0) | |
3132 | close(fd); | |
3133 | return retval; | |
3134 | } | |
3135 | ||
c17608ea AK |
3136 | /******************************************************************************* |
3137 | * function: imsm_create_metadata_checkpoint_update | |
3138 | * Description: It creates update for checkpoint change. | |
3139 | * Parameters: | |
3140 | * super : imsm internal array info | |
3141 | * u : pointer to prepared update | |
3142 | * Returns: | |
3143 | * Uptate length. | |
3144 | * If length is equal to 0, input pointer u contains no update | |
3145 | ******************************************************************************/ | |
3146 | static int imsm_create_metadata_checkpoint_update( | |
3147 | struct intel_super *super, | |
3148 | struct imsm_update_general_migration_checkpoint **u) | |
3149 | { | |
3150 | ||
3151 | int update_memory_size = 0; | |
3152 | ||
1ade5cc1 | 3153 | dprintf("(enter)\n"); |
c17608ea AK |
3154 | |
3155 | if (u == NULL) | |
3156 | return 0; | |
3157 | *u = NULL; | |
3158 | ||
3159 | /* size of all update data without anchor */ | |
3160 | update_memory_size = | |
3161 | sizeof(struct imsm_update_general_migration_checkpoint); | |
3162 | ||
503975b9 | 3163 | *u = xcalloc(1, update_memory_size); |
c17608ea | 3164 | if (*u == NULL) { |
1ade5cc1 | 3165 | dprintf("error: cannot get memory\n"); |
c17608ea AK |
3166 | return 0; |
3167 | } | |
3168 | (*u)->type = update_general_migration_checkpoint; | |
9f421827 | 3169 | (*u)->curr_migr_unit = current_migr_unit(super->migr_rec); |
1ade5cc1 | 3170 | dprintf("prepared for %u\n", (*u)->curr_migr_unit); |
c17608ea AK |
3171 | |
3172 | return update_memory_size; | |
3173 | } | |
3174 | ||
c17608ea AK |
3175 | static void imsm_update_metadata_locally(struct supertype *st, |
3176 | void *buf, int len); | |
3177 | ||
687629c2 AK |
3178 | /******************************************************************************* |
3179 | * Function: write_imsm_migr_rec | |
3180 | * Description: Function writes imsm migration record | |
3181 | * (at the last sector of disk) | |
3182 | * Parameters: | |
3183 | * super : imsm internal array info | |
3184 | * Returns: | |
3185 | * 0 : success | |
3186 | * -1 : if fail | |
3187 | ******************************************************************************/ | |
3188 | static int write_imsm_migr_rec(struct supertype *st) | |
3189 | { | |
3190 | struct intel_super *super = st->sb; | |
de44e46f | 3191 | unsigned int sector_size = super->sector_size; |
687629c2 AK |
3192 | unsigned long long dsize; |
3193 | char nm[30]; | |
3194 | int fd = -1; | |
3195 | int retval = -1; | |
3196 | struct dl *sd; | |
c17608ea AK |
3197 | int len; |
3198 | struct imsm_update_general_migration_checkpoint *u; | |
3136abe5 | 3199 | struct imsm_dev *dev; |
594dc1b8 | 3200 | struct imsm_map *map; |
3136abe5 AK |
3201 | |
3202 | /* find map under migration */ | |
3203 | dev = imsm_get_device_during_migration(super); | |
3204 | /* if no migration, write buffer anyway to clear migr_record | |
3205 | * on disk based on first available device | |
3206 | */ | |
3207 | if (dev == NULL) | |
3208 | dev = get_imsm_dev(super, super->current_vol < 0 ? 0 : | |
3209 | super->current_vol); | |
3210 | ||
44bfe6df | 3211 | map = get_imsm_map(dev, MAP_0); |
687629c2 | 3212 | |
de44e46f PB |
3213 | if (sector_size == 4096) |
3214 | convert_to_4k_imsm_migr_rec(super); | |
687629c2 | 3215 | for (sd = super->disks ; sd ; sd = sd->next) { |
b4ab44d8 | 3216 | int slot = -1; |
3136abe5 AK |
3217 | |
3218 | /* skip failed and spare devices */ | |
3219 | if (sd->index < 0) | |
3220 | continue; | |
687629c2 | 3221 | /* write to 2 first slots only */ |
3136abe5 AK |
3222 | if (map) |
3223 | slot = get_imsm_disk_slot(map, sd->index); | |
089f9d79 | 3224 | if (map == NULL || slot > 1 || slot < 0) |
687629c2 | 3225 | continue; |
3136abe5 | 3226 | |
687629c2 AK |
3227 | sprintf(nm, "%d:%d", sd->major, sd->minor); |
3228 | fd = dev_open(nm, O_RDWR); | |
3229 | if (fd < 0) | |
3230 | continue; | |
3231 | get_dev_size(fd, NULL, &dsize); | |
de44e46f PB |
3232 | if (lseek64(fd, dsize - (MIGR_REC_SECTOR_POSITION*sector_size), |
3233 | SEEK_SET) < 0) { | |
e7b84f9d N |
3234 | pr_err("Cannot seek to anchor block: %s\n", |
3235 | strerror(errno)); | |
687629c2 AK |
3236 | goto out; |
3237 | } | |
466070ad | 3238 | if ((unsigned int)write(fd, super->migr_rec_buf, |
de44e46f PB |
3239 | MIGR_REC_BUF_SECTORS*sector_size) != |
3240 | MIGR_REC_BUF_SECTORS*sector_size) { | |
e7b84f9d N |
3241 | pr_err("Cannot write migr record block: %s\n", |
3242 | strerror(errno)); | |
687629c2 AK |
3243 | goto out; |
3244 | } | |
3245 | close(fd); | |
3246 | fd = -1; | |
3247 | } | |
de44e46f PB |
3248 | if (sector_size == 4096) |
3249 | convert_from_4k_imsm_migr_rec(super); | |
c17608ea AK |
3250 | /* update checkpoint information in metadata */ |
3251 | len = imsm_create_metadata_checkpoint_update(super, &u); | |
c17608ea AK |
3252 | if (len <= 0) { |
3253 | dprintf("imsm: Cannot prepare update\n"); | |
3254 | goto out; | |
3255 | } | |
3256 | /* update metadata locally */ | |
3257 | imsm_update_metadata_locally(st, u, len); | |
3258 | /* and possibly remotely */ | |
3259 | if (st->update_tail) { | |
3260 | append_metadata_update(st, u, len); | |
3261 | /* during reshape we do all work inside metadata handler | |
3262 | * manage_reshape(), so metadata update has to be triggered | |
3263 | * insida it | |
3264 | */ | |
3265 | flush_metadata_updates(st); | |
3266 | st->update_tail = &st->updates; | |
3267 | } else | |
3268 | free(u); | |
687629c2 AK |
3269 | |
3270 | retval = 0; | |
3271 | out: | |
3272 | if (fd >= 0) | |
3273 | close(fd); | |
3274 | return retval; | |
3275 | } | |
3276 | ||
e2962bfc AK |
3277 | /* spare/missing disks activations are not allowe when |
3278 | * array/container performs reshape operation, because | |
3279 | * all arrays in container works on the same disks set | |
3280 | */ | |
3281 | int imsm_reshape_blocks_arrays_changes(struct intel_super *super) | |
3282 | { | |
3283 | int rv = 0; | |
3284 | struct intel_dev *i_dev; | |
3285 | struct imsm_dev *dev; | |
3286 | ||
3287 | /* check whole container | |
3288 | */ | |
3289 | for (i_dev = super->devlist; i_dev; i_dev = i_dev->next) { | |
3290 | dev = i_dev->dev; | |
3ad25638 | 3291 | if (is_gen_migration(dev)) { |
e2962bfc AK |
3292 | /* No repair during any migration in container |
3293 | */ | |
3294 | rv = 1; | |
3295 | break; | |
3296 | } | |
3297 | } | |
3298 | return rv; | |
3299 | } | |
3e684231 | 3300 | static unsigned long long imsm_component_size_alignment_check(int level, |
c41e00b2 | 3301 | int chunk_size, |
f36a9ecd | 3302 | unsigned int sector_size, |
c41e00b2 AK |
3303 | unsigned long long component_size) |
3304 | { | |
3e684231 | 3305 | unsigned int component_size_alignment; |
c41e00b2 | 3306 | |
3e684231 | 3307 | /* check component size alignment |
c41e00b2 | 3308 | */ |
3e684231 | 3309 | component_size_alignment = component_size % (chunk_size/sector_size); |
c41e00b2 | 3310 | |
3e684231 | 3311 | dprintf("(Level: %i, chunk_size = %i, component_size = %llu), component_size_alignment = %u\n", |
c41e00b2 | 3312 | level, chunk_size, component_size, |
3e684231 | 3313 | component_size_alignment); |
c41e00b2 | 3314 | |
3e684231 MZ |
3315 | if (component_size_alignment && (level != 1) && (level != UnSet)) { |
3316 | dprintf("imsm: reported component size aligned from %llu ", | |
c41e00b2 | 3317 | component_size); |
3e684231 | 3318 | component_size -= component_size_alignment; |
1ade5cc1 | 3319 | dprintf_cont("to %llu (%i).\n", |
3e684231 | 3320 | component_size, component_size_alignment); |
c41e00b2 AK |
3321 | } |
3322 | ||
3323 | return component_size; | |
3324 | } | |
e2962bfc | 3325 | |
2432ce9b AP |
3326 | static unsigned long long get_ppl_sector(struct intel_super *super, int dev_idx) |
3327 | { | |
3328 | struct imsm_dev *dev = get_imsm_dev(super, dev_idx); | |
3329 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
3330 | ||
3331 | return pba_of_lba0(map) + | |
3332 | (num_data_stripes(map) * map->blocks_per_strip); | |
3333 | } | |
3334 | ||
a5d85af7 | 3335 | static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap) |
bf5a934a DW |
3336 | { |
3337 | struct intel_super *super = st->sb; | |
c47b0ff6 | 3338 | struct migr_record *migr_rec = super->migr_rec; |
949c47a0 | 3339 | struct imsm_dev *dev = get_imsm_dev(super, super->current_vol); |
238c0a71 AK |
3340 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
3341 | struct imsm_map *prev_map = get_imsm_map(dev, MAP_1); | |
b335e593 | 3342 | struct imsm_map *map_to_analyse = map; |
efb30e7f | 3343 | struct dl *dl; |
a5d85af7 | 3344 | int map_disks = info->array.raid_disks; |
bf5a934a | 3345 | |
95eeceeb | 3346 | memset(info, 0, sizeof(*info)); |
b335e593 AK |
3347 | if (prev_map) |
3348 | map_to_analyse = prev_map; | |
3349 | ||
ca0748fa | 3350 | dl = super->current_disk; |
9894ec0d | 3351 | |
bf5a934a | 3352 | info->container_member = super->current_vol; |
cd0430a1 | 3353 | info->array.raid_disks = map->num_members; |
b335e593 | 3354 | info->array.level = get_imsm_raid_level(map_to_analyse); |
bf5a934a DW |
3355 | info->array.layout = imsm_level_to_layout(info->array.level); |
3356 | info->array.md_minor = -1; | |
3357 | info->array.ctime = 0; | |
3358 | info->array.utime = 0; | |
b335e593 AK |
3359 | info->array.chunk_size = |
3360 | __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9; | |
2432ce9b | 3361 | info->array.state = !(dev->vol.dirty & RAIDVOL_DIRTY); |
fcc2c9da | 3362 | info->custom_array_size = imsm_dev_size(dev); |
3ad25638 AK |
3363 | info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb); |
3364 | ||
3f510843 | 3365 | if (is_gen_migration(dev)) { |
3f83228a | 3366 | info->reshape_active = 1; |
b335e593 AK |
3367 | info->new_level = get_imsm_raid_level(map); |
3368 | info->new_layout = imsm_level_to_layout(info->new_level); | |
3369 | info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9; | |
3f83228a | 3370 | info->delta_disks = map->num_members - prev_map->num_members; |
493f5dd6 N |
3371 | if (info->delta_disks) { |
3372 | /* this needs to be applied to every array | |
3373 | * in the container. | |
3374 | */ | |
81219e70 | 3375 | info->reshape_active = CONTAINER_RESHAPE; |
493f5dd6 | 3376 | } |
3f83228a N |
3377 | /* We shape information that we give to md might have to be |
3378 | * modify to cope with md's requirement for reshaping arrays. | |
3379 | * For example, when reshaping a RAID0, md requires it to be | |
3380 | * presented as a degraded RAID4. | |
3381 | * Also if a RAID0 is migrating to a RAID5 we need to specify | |
3382 | * the array as already being RAID5, but the 'before' layout | |
3383 | * is a RAID4-like layout. | |
3384 | */ | |
3385 | switch (info->array.level) { | |
3386 | case 0: | |
3387 | switch(info->new_level) { | |
3388 | case 0: | |
3389 | /* conversion is happening as RAID4 */ | |
3390 | info->array.level = 4; | |
3391 | info->array.raid_disks += 1; | |
3392 | break; | |
3393 | case 5: | |
3394 | /* conversion is happening as RAID5 */ | |
3395 | info->array.level = 5; | |
3396 | info->array.layout = ALGORITHM_PARITY_N; | |
3f83228a N |
3397 | info->delta_disks -= 1; |
3398 | break; | |
3399 | default: | |
3400 | /* FIXME error message */ | |
3401 | info->array.level = UnSet; | |
3402 | break; | |
3403 | } | |
3404 | break; | |
3405 | } | |
b335e593 AK |
3406 | } else { |
3407 | info->new_level = UnSet; | |
3408 | info->new_layout = UnSet; | |
3409 | info->new_chunk = info->array.chunk_size; | |
3f83228a | 3410 | info->delta_disks = 0; |
b335e593 | 3411 | } |
ca0748fa | 3412 | |
efb30e7f DW |
3413 | if (dl) { |
3414 | info->disk.major = dl->major; | |
3415 | info->disk.minor = dl->minor; | |
ca0748fa | 3416 | info->disk.number = dl->index; |
656b6b5a N |
3417 | info->disk.raid_disk = get_imsm_disk_slot(map_to_analyse, |
3418 | dl->index); | |
efb30e7f | 3419 | } |
bf5a934a | 3420 | |
5551b113 | 3421 | info->data_offset = pba_of_lba0(map_to_analyse); |
44490938 | 3422 | info->component_size = calc_component_size(map, dev); |
3e684231 | 3423 | info->component_size = imsm_component_size_alignment_check( |
c41e00b2 AK |
3424 | info->array.level, |
3425 | info->array.chunk_size, | |
f36a9ecd | 3426 | super->sector_size, |
c41e00b2 | 3427 | info->component_size); |
5e46202e | 3428 | info->bb.supported = 1; |
139dae11 | 3429 | |
301406c9 | 3430 | memset(info->uuid, 0, sizeof(info->uuid)); |
921d9e16 | 3431 | info->recovery_start = MaxSector; |
bf5a934a | 3432 | |
c2462068 PB |
3433 | if (info->array.level == 5 && |
3434 | (dev->rwh_policy == RWH_DISTRIBUTED || | |
3435 | dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)) { | |
2432ce9b AP |
3436 | info->consistency_policy = CONSISTENCY_POLICY_PPL; |
3437 | info->ppl_sector = get_ppl_sector(super, super->current_vol); | |
c2462068 PB |
3438 | if (dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED) |
3439 | info->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9; | |
3440 | else | |
3441 | info->ppl_size = (PPL_HEADER_SIZE + PPL_ENTRY_SPACE) | |
3442 | >> 9; | |
2432ce9b AP |
3443 | } else if (info->array.level <= 0) { |
3444 | info->consistency_policy = CONSISTENCY_POLICY_NONE; | |
3445 | } else { | |
3446 | info->consistency_policy = CONSISTENCY_POLICY_RESYNC; | |
3447 | } | |
3448 | ||
d2e6d5d6 | 3449 | info->reshape_progress = 0; |
b6796ce1 | 3450 | info->resync_start = MaxSector; |
b9172665 | 3451 | if ((map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED || |
2432ce9b | 3452 | !(info->array.state & 1)) && |
b9172665 | 3453 | imsm_reshape_blocks_arrays_changes(super) == 0) { |
301406c9 | 3454 | info->resync_start = 0; |
b6796ce1 AK |
3455 | } |
3456 | if (dev->vol.migr_state) { | |
1e5c6983 DW |
3457 | switch (migr_type(dev)) { |
3458 | case MIGR_REPAIR: | |
3459 | case MIGR_INIT: { | |
c47b0ff6 AK |
3460 | __u64 blocks_per_unit = blocks_per_migr_unit(super, |
3461 | dev); | |
1e5c6983 DW |
3462 | __u64 units = __le32_to_cpu(dev->vol.curr_migr_unit); |
3463 | ||
3464 | info->resync_start = blocks_per_unit * units; | |
3465 | break; | |
3466 | } | |
d2e6d5d6 | 3467 | case MIGR_GEN_MIGR: { |
c47b0ff6 AK |
3468 | __u64 blocks_per_unit = blocks_per_migr_unit(super, |
3469 | dev); | |
9f421827 | 3470 | __u64 units = current_migr_unit(migr_rec); |
04fa9523 AK |
3471 | unsigned long long array_blocks; |
3472 | int used_disks; | |
d2e6d5d6 | 3473 | |
befb629b AK |
3474 | if (__le32_to_cpu(migr_rec->ascending_migr) && |
3475 | (units < | |
9f421827 | 3476 | (get_num_migr_units(migr_rec)-1)) && |
befb629b AK |
3477 | (super->migr_rec->rec_status == |
3478 | __cpu_to_le32(UNIT_SRC_IN_CP_AREA))) | |
3479 | units++; | |
3480 | ||
d2e6d5d6 | 3481 | info->reshape_progress = blocks_per_unit * units; |
6289d1e0 | 3482 | |
7a862a02 | 3483 | dprintf("IMSM: General Migration checkpoint : %llu (%llu) -> read reshape progress : %llu\n", |
19986c72 MB |
3484 | (unsigned long long)units, |
3485 | (unsigned long long)blocks_per_unit, | |
3486 | info->reshape_progress); | |
75156c46 | 3487 | |
9529d343 | 3488 | used_disks = imsm_num_data_members(prev_map); |
75156c46 | 3489 | if (used_disks > 0) { |
44490938 | 3490 | array_blocks = per_dev_array_size(map) * |
75156c46 | 3491 | used_disks; |
b53bfba6 TM |
3492 | info->custom_array_size = |
3493 | round_size_to_mb(array_blocks, | |
3494 | used_disks); | |
3495 | ||
75156c46 | 3496 | } |
d2e6d5d6 | 3497 | } |
1e5c6983 DW |
3498 | case MIGR_VERIFY: |
3499 | /* we could emulate the checkpointing of | |
3500 | * 'sync_action=check' migrations, but for now | |
3501 | * we just immediately complete them | |
3502 | */ | |
3503 | case MIGR_REBUILD: | |
3504 | /* this is handled by container_content_imsm() */ | |
1e5c6983 DW |
3505 | case MIGR_STATE_CHANGE: |
3506 | /* FIXME handle other migrations */ | |
3507 | default: | |
3508 | /* we are not dirty, so... */ | |
3509 | info->resync_start = MaxSector; | |
3510 | } | |
b6796ce1 | 3511 | } |
301406c9 DW |
3512 | |
3513 | strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN); | |
3514 | info->name[MAX_RAID_SERIAL_LEN] = 0; | |
bf5a934a | 3515 | |
f35f2525 N |
3516 | info->array.major_version = -1; |
3517 | info->array.minor_version = -2; | |
4dd2df09 | 3518 | sprintf(info->text_version, "/%s/%d", st->container_devnm, info->container_member); |
a67dd8cc | 3519 | info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */ |
51006d85 | 3520 | uuid_from_super_imsm(st, info->uuid); |
a5d85af7 N |
3521 | |
3522 | if (dmap) { | |
3523 | int i, j; | |
3524 | for (i=0; i<map_disks; i++) { | |
3525 | dmap[i] = 0; | |
3526 | if (i < info->array.raid_disks) { | |
3527 | struct imsm_disk *dsk; | |
238c0a71 | 3528 | j = get_imsm_disk_idx(dev, i, MAP_X); |
a5d85af7 N |
3529 | dsk = get_imsm_disk(super, j); |
3530 | if (dsk && (dsk->status & CONFIGURED_DISK)) | |
3531 | dmap[i] = 1; | |
3532 | } | |
3533 | } | |
3534 | } | |
81ac8b4d | 3535 | } |
bf5a934a | 3536 | |
3b451610 AK |
3537 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, |
3538 | int failed, int look_in_map); | |
3539 | ||
3540 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev, | |
3541 | int look_in_map); | |
3542 | ||
3543 | static void manage_second_map(struct intel_super *super, struct imsm_dev *dev) | |
3544 | { | |
3545 | if (is_gen_migration(dev)) { | |
3546 | int failed; | |
3547 | __u8 map_state; | |
3548 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); | |
3549 | ||
3550 | failed = imsm_count_failed(super, dev, MAP_1); | |
238c0a71 | 3551 | map_state = imsm_check_degraded(super, dev, failed, MAP_1); |
3b451610 AK |
3552 | if (map2->map_state != map_state) { |
3553 | map2->map_state = map_state; | |
3554 | super->updates_pending++; | |
3555 | } | |
3556 | } | |
3557 | } | |
97b4d0e9 DW |
3558 | |
3559 | static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index) | |
3560 | { | |
3561 | struct dl *d; | |
3562 | ||
3563 | for (d = super->missing; d; d = d->next) | |
3564 | if (d->index == index) | |
3565 | return &d->disk; | |
3566 | return NULL; | |
3567 | } | |
3568 | ||
a5d85af7 | 3569 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map) |
4f5bc454 DW |
3570 | { |
3571 | struct intel_super *super = st->sb; | |
4f5bc454 | 3572 | struct imsm_disk *disk; |
a5d85af7 | 3573 | int map_disks = info->array.raid_disks; |
ab3cb6b3 N |
3574 | int max_enough = -1; |
3575 | int i; | |
3576 | struct imsm_super *mpb; | |
4f5bc454 | 3577 | |
bf5a934a | 3578 | if (super->current_vol >= 0) { |
a5d85af7 | 3579 | getinfo_super_imsm_volume(st, info, map); |
bf5a934a DW |
3580 | return; |
3581 | } | |
95eeceeb | 3582 | memset(info, 0, sizeof(*info)); |
d23fe947 DW |
3583 | |
3584 | /* Set raid_disks to zero so that Assemble will always pull in valid | |
3585 | * spares | |
3586 | */ | |
3587 | info->array.raid_disks = 0; | |
cdddbdbc DW |
3588 | info->array.level = LEVEL_CONTAINER; |
3589 | info->array.layout = 0; | |
3590 | info->array.md_minor = -1; | |
1011e834 | 3591 | info->array.ctime = 0; /* N/A for imsm */ |
cdddbdbc DW |
3592 | info->array.utime = 0; |
3593 | info->array.chunk_size = 0; | |
3594 | ||
3595 | info->disk.major = 0; | |
3596 | info->disk.minor = 0; | |
cdddbdbc | 3597 | info->disk.raid_disk = -1; |
c2c087e6 | 3598 | info->reshape_active = 0; |
f35f2525 N |
3599 | info->array.major_version = -1; |
3600 | info->array.minor_version = -2; | |
c2c087e6 | 3601 | strcpy(info->text_version, "imsm"); |
a67dd8cc | 3602 | info->safe_mode_delay = 0; |
c2c087e6 DW |
3603 | info->disk.number = -1; |
3604 | info->disk.state = 0; | |
c5afc314 | 3605 | info->name[0] = 0; |
921d9e16 | 3606 | info->recovery_start = MaxSector; |
3ad25638 | 3607 | info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb); |
5e46202e | 3608 | info->bb.supported = 1; |
c2c087e6 | 3609 | |
97b4d0e9 | 3610 | /* do we have the all the insync disks that we expect? */ |
ab3cb6b3 | 3611 | mpb = super->anchor; |
b7d81a38 | 3612 | info->events = __le32_to_cpu(mpb->generation_num); |
97b4d0e9 | 3613 | |
ab3cb6b3 N |
3614 | for (i = 0; i < mpb->num_raid_devs; i++) { |
3615 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
3616 | int failed, enough, j, missing = 0; | |
3617 | struct imsm_map *map; | |
3618 | __u8 state; | |
97b4d0e9 | 3619 | |
3b451610 AK |
3620 | failed = imsm_count_failed(super, dev, MAP_0); |
3621 | state = imsm_check_degraded(super, dev, failed, MAP_0); | |
238c0a71 | 3622 | map = get_imsm_map(dev, MAP_0); |
ab3cb6b3 N |
3623 | |
3624 | /* any newly missing disks? | |
3625 | * (catches single-degraded vs double-degraded) | |
3626 | */ | |
3627 | for (j = 0; j < map->num_members; j++) { | |
238c0a71 | 3628 | __u32 ord = get_imsm_ord_tbl_ent(dev, j, MAP_0); |
ab3cb6b3 N |
3629 | __u32 idx = ord_to_idx(ord); |
3630 | ||
20dc76d1 MT |
3631 | if (super->disks && super->disks->index == (int)idx) |
3632 | info->disk.raid_disk = j; | |
3633 | ||
ab3cb6b3 N |
3634 | if (!(ord & IMSM_ORD_REBUILD) && |
3635 | get_imsm_missing(super, idx)) { | |
3636 | missing = 1; | |
3637 | break; | |
3638 | } | |
97b4d0e9 | 3639 | } |
ab3cb6b3 N |
3640 | |
3641 | if (state == IMSM_T_STATE_FAILED) | |
3642 | enough = -1; | |
3643 | else if (state == IMSM_T_STATE_DEGRADED && | |
3644 | (state != map->map_state || missing)) | |
3645 | enough = 0; | |
3646 | else /* we're normal, or already degraded */ | |
3647 | enough = 1; | |
d2bde6d3 AK |
3648 | if (is_gen_migration(dev) && missing) { |
3649 | /* during general migration we need all disks | |
3650 | * that process is running on. | |
3651 | * No new missing disk is allowed. | |
3652 | */ | |
3653 | max_enough = -1; | |
3654 | enough = -1; | |
3655 | /* no more checks necessary | |
3656 | */ | |
3657 | break; | |
3658 | } | |
ab3cb6b3 N |
3659 | /* in the missing/failed disk case check to see |
3660 | * if at least one array is runnable | |
3661 | */ | |
3662 | max_enough = max(max_enough, enough); | |
3663 | } | |
1ade5cc1 | 3664 | dprintf("enough: %d\n", max_enough); |
ab3cb6b3 | 3665 | info->container_enough = max_enough; |
97b4d0e9 | 3666 | |
4a04ec6c | 3667 | if (super->disks) { |
14e8215b DW |
3668 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
3669 | ||
b9f594fe | 3670 | disk = &super->disks->disk; |
5551b113 | 3671 | info->data_offset = total_blocks(&super->disks->disk) - reserved; |
14e8215b | 3672 | info->component_size = reserved; |
25ed7e59 | 3673 | info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0; |
df474657 DW |
3674 | /* we don't change info->disk.raid_disk here because |
3675 | * this state will be finalized in mdmon after we have | |
3676 | * found the 'most fresh' version of the metadata | |
3677 | */ | |
25ed7e59 | 3678 | info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; |
2432ce9b AP |
3679 | info->disk.state |= (is_spare(disk) || is_journal(disk)) ? |
3680 | 0 : (1 << MD_DISK_SYNC); | |
cdddbdbc | 3681 | } |
a575e2a7 DW |
3682 | |
3683 | /* only call uuid_from_super_imsm when this disk is part of a populated container, | |
3684 | * ->compare_super may have updated the 'num_raid_devs' field for spares | |
3685 | */ | |
3686 | if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs) | |
36ba7d48 | 3687 | uuid_from_super_imsm(st, info->uuid); |
22e263f6 AC |
3688 | else |
3689 | memcpy(info->uuid, uuid_zero, sizeof(uuid_zero)); | |
a5d85af7 N |
3690 | |
3691 | /* I don't know how to compute 'map' on imsm, so use safe default */ | |
3692 | if (map) { | |
3693 | int i; | |
3694 | for (i = 0; i < map_disks; i++) | |
3695 | map[i] = 1; | |
3696 | } | |
3697 | ||
cdddbdbc DW |
3698 | } |
3699 | ||
5c4cd5da AC |
3700 | /* allocates memory and fills disk in mdinfo structure |
3701 | * for each disk in array */ | |
3702 | struct mdinfo *getinfo_super_disks_imsm(struct supertype *st) | |
3703 | { | |
594dc1b8 | 3704 | struct mdinfo *mddev; |
5c4cd5da AC |
3705 | struct intel_super *super = st->sb; |
3706 | struct imsm_disk *disk; | |
3707 | int count = 0; | |
3708 | struct dl *dl; | |
3709 | if (!super || !super->disks) | |
3710 | return NULL; | |
3711 | dl = super->disks; | |
503975b9 | 3712 | mddev = xcalloc(1, sizeof(*mddev)); |
5c4cd5da AC |
3713 | while (dl) { |
3714 | struct mdinfo *tmp; | |
3715 | disk = &dl->disk; | |
503975b9 | 3716 | tmp = xcalloc(1, sizeof(*tmp)); |
5c4cd5da AC |
3717 | if (mddev->devs) |
3718 | tmp->next = mddev->devs; | |
3719 | mddev->devs = tmp; | |
3720 | tmp->disk.number = count++; | |
3721 | tmp->disk.major = dl->major; | |
3722 | tmp->disk.minor = dl->minor; | |
3723 | tmp->disk.state = is_configured(disk) ? | |
3724 | (1 << MD_DISK_ACTIVE) : 0; | |
3725 | tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; | |
3726 | tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC); | |
3727 | tmp->disk.raid_disk = -1; | |
3728 | dl = dl->next; | |
3729 | } | |
3730 | return mddev; | |
3731 | } | |
3732 | ||
cdddbdbc DW |
3733 | static int update_super_imsm(struct supertype *st, struct mdinfo *info, |
3734 | char *update, char *devname, int verbose, | |
3735 | int uuid_set, char *homehost) | |
3736 | { | |
f352c545 DW |
3737 | /* For 'assemble' and 'force' we need to return non-zero if any |
3738 | * change was made. For others, the return value is ignored. | |
3739 | * Update options are: | |
3740 | * force-one : This device looks a bit old but needs to be included, | |
3741 | * update age info appropriately. | |
3742 | * assemble: clear any 'faulty' flag to allow this device to | |
3743 | * be assembled. | |
3744 | * force-array: Array is degraded but being forced, mark it clean | |
3745 | * if that will be needed to assemble it. | |
3746 | * | |
3747 | * newdev: not used ???? | |
3748 | * grow: Array has gained a new device - this is currently for | |
3749 | * linear only | |
3750 | * resync: mark as dirty so a resync will happen. | |
3751 | * name: update the name - preserving the homehost | |
6e46bf34 | 3752 | * uuid: Change the uuid of the array to match watch is given |
f352c545 DW |
3753 | * |
3754 | * Following are not relevant for this imsm: | |
3755 | * sparc2.2 : update from old dodgey metadata | |
3756 | * super-minor: change the preferred_minor number | |
3757 | * summaries: update redundant counters. | |
f352c545 DW |
3758 | * homehost: update the recorded homehost |
3759 | * _reshape_progress: record new reshape_progress position. | |
3760 | */ | |
6e46bf34 DW |
3761 | int rv = 1; |
3762 | struct intel_super *super = st->sb; | |
3763 | struct imsm_super *mpb; | |
f352c545 | 3764 | |
6e46bf34 DW |
3765 | /* we can only update container info */ |
3766 | if (!super || super->current_vol >= 0 || !super->anchor) | |
3767 | return 1; | |
3768 | ||
3769 | mpb = super->anchor; | |
3770 | ||
81a5b4f5 N |
3771 | if (strcmp(update, "uuid") == 0) { |
3772 | /* We take this to mean that the family_num should be updated. | |
3773 | * However that is much smaller than the uuid so we cannot really | |
3774 | * allow an explicit uuid to be given. And it is hard to reliably | |
3775 | * know if one was. | |
3776 | * So if !uuid_set we know the current uuid is random and just used | |
3777 | * the first 'int' and copy it to the other 3 positions. | |
3778 | * Otherwise we require the 4 'int's to be the same as would be the | |
3779 | * case if we are using a random uuid. So an explicit uuid will be | |
3780 | * accepted as long as all for ints are the same... which shouldn't hurt | |
6e46bf34 | 3781 | */ |
81a5b4f5 N |
3782 | if (!uuid_set) { |
3783 | info->uuid[1] = info->uuid[2] = info->uuid[3] = info->uuid[0]; | |
6e46bf34 | 3784 | rv = 0; |
81a5b4f5 N |
3785 | } else { |
3786 | if (info->uuid[0] != info->uuid[1] || | |
3787 | info->uuid[1] != info->uuid[2] || | |
3788 | info->uuid[2] != info->uuid[3]) | |
3789 | rv = -1; | |
3790 | else | |
3791 | rv = 0; | |
6e46bf34 | 3792 | } |
81a5b4f5 N |
3793 | if (rv == 0) |
3794 | mpb->orig_family_num = info->uuid[0]; | |
6e46bf34 DW |
3795 | } else if (strcmp(update, "assemble") == 0) |
3796 | rv = 0; | |
3797 | else | |
1e2b2765 | 3798 | rv = -1; |
f352c545 | 3799 | |
6e46bf34 DW |
3800 | /* successful update? recompute checksum */ |
3801 | if (rv == 0) | |
3802 | mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb)); | |
f352c545 DW |
3803 | |
3804 | return rv; | |
cdddbdbc DW |
3805 | } |
3806 | ||
c2c087e6 | 3807 | static size_t disks_to_mpb_size(int disks) |
cdddbdbc | 3808 | { |
c2c087e6 | 3809 | size_t size; |
cdddbdbc | 3810 | |
c2c087e6 DW |
3811 | size = sizeof(struct imsm_super); |
3812 | size += (disks - 1) * sizeof(struct imsm_disk); | |
3813 | size += 2 * sizeof(struct imsm_dev); | |
3814 | /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */ | |
3815 | size += (4 - 2) * sizeof(struct imsm_map); | |
3816 | /* 4 possible disk_ord_tbl's */ | |
3817 | size += 4 * (disks - 1) * sizeof(__u32); | |
bbab0940 TM |
3818 | /* maximum bbm log */ |
3819 | size += sizeof(struct bbm_log); | |
c2c087e6 DW |
3820 | |
3821 | return size; | |
3822 | } | |
3823 | ||
387fcd59 N |
3824 | static __u64 avail_size_imsm(struct supertype *st, __u64 devsize, |
3825 | unsigned long long data_offset) | |
c2c087e6 DW |
3826 | { |
3827 | if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS)) | |
3828 | return 0; | |
3829 | ||
3830 | return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS); | |
cdddbdbc DW |
3831 | } |
3832 | ||
ba2de7ba DW |
3833 | static void free_devlist(struct intel_super *super) |
3834 | { | |
3835 | struct intel_dev *dv; | |
3836 | ||
3837 | while (super->devlist) { | |
3838 | dv = super->devlist->next; | |
3839 | free(super->devlist->dev); | |
3840 | free(super->devlist); | |
3841 | super->devlist = dv; | |
3842 | } | |
3843 | } | |
3844 | ||
3845 | static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src) | |
3846 | { | |
3847 | memcpy(dest, src, sizeof_imsm_dev(src, 0)); | |
3848 | } | |
3849 | ||
cdddbdbc DW |
3850 | static int compare_super_imsm(struct supertype *st, struct supertype *tst) |
3851 | { | |
3852 | /* | |
3853 | * return: | |
3854 | * 0 same, or first was empty, and second was copied | |
3855 | * 1 second had wrong number | |
3856 | * 2 wrong uuid | |
3857 | * 3 wrong other info | |
3858 | */ | |
3859 | struct intel_super *first = st->sb; | |
3860 | struct intel_super *sec = tst->sb; | |
3861 | ||
5d500228 N |
3862 | if (!first) { |
3863 | st->sb = tst->sb; | |
3864 | tst->sb = NULL; | |
3865 | return 0; | |
3866 | } | |
8603ea6f LM |
3867 | /* in platform dependent environment test if the disks |
3868 | * use the same Intel hba | |
cb8f6859 | 3869 | * If not on Intel hba at all, allow anything. |
8603ea6f | 3870 | */ |
6b781d33 AP |
3871 | if (!check_env("IMSM_NO_PLATFORM") && first->hba && sec->hba) { |
3872 | if (first->hba->type != sec->hba->type) { | |
8603ea6f | 3873 | fprintf(stderr, |
6b781d33 AP |
3874 | "HBAs of devices do not match %s != %s\n", |
3875 | get_sys_dev_type(first->hba->type), | |
3876 | get_sys_dev_type(sec->hba->type)); | |
3877 | return 3; | |
3878 | } | |
3879 | if (first->orom != sec->orom) { | |
3880 | fprintf(stderr, | |
3881 | "HBAs of devices do not match %s != %s\n", | |
3882 | first->hba->pci_id, sec->hba->pci_id); | |
8603ea6f LM |
3883 | return 3; |
3884 | } | |
3885 | } | |
cdddbdbc | 3886 | |
d23fe947 DW |
3887 | /* if an anchor does not have num_raid_devs set then it is a free |
3888 | * floating spare | |
3889 | */ | |
3890 | if (first->anchor->num_raid_devs > 0 && | |
3891 | sec->anchor->num_raid_devs > 0) { | |
a2b97981 DW |
3892 | /* Determine if these disks might ever have been |
3893 | * related. Further disambiguation can only take place | |
3894 | * in load_super_imsm_all | |
3895 | */ | |
3896 | __u32 first_family = first->anchor->orig_family_num; | |
3897 | __u32 sec_family = sec->anchor->orig_family_num; | |
3898 | ||
f796af5d DW |
3899 | if (memcmp(first->anchor->sig, sec->anchor->sig, |
3900 | MAX_SIGNATURE_LENGTH) != 0) | |
3901 | return 3; | |
3902 | ||
a2b97981 DW |
3903 | if (first_family == 0) |
3904 | first_family = first->anchor->family_num; | |
3905 | if (sec_family == 0) | |
3906 | sec_family = sec->anchor->family_num; | |
3907 | ||
3908 | if (first_family != sec_family) | |
d23fe947 | 3909 | return 3; |
f796af5d | 3910 | |
d23fe947 | 3911 | } |
cdddbdbc | 3912 | |
3e372e5a DW |
3913 | /* if 'first' is a spare promote it to a populated mpb with sec's |
3914 | * family number | |
3915 | */ | |
3916 | if (first->anchor->num_raid_devs == 0 && | |
3917 | sec->anchor->num_raid_devs > 0) { | |
78d30f94 | 3918 | int i; |
ba2de7ba DW |
3919 | struct intel_dev *dv; |
3920 | struct imsm_dev *dev; | |
78d30f94 DW |
3921 | |
3922 | /* we need to copy raid device info from sec if an allocation | |
3923 | * fails here we don't associate the spare | |
3924 | */ | |
3925 | for (i = 0; i < sec->anchor->num_raid_devs; i++) { | |
503975b9 N |
3926 | dv = xmalloc(sizeof(*dv)); |
3927 | dev = xmalloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1)); | |
ba2de7ba DW |
3928 | dv->dev = dev; |
3929 | dv->index = i; | |
3930 | dv->next = first->devlist; | |
3931 | first->devlist = dv; | |
78d30f94 | 3932 | } |
709743c5 | 3933 | if (i < sec->anchor->num_raid_devs) { |
ba2de7ba DW |
3934 | /* allocation failure */ |
3935 | free_devlist(first); | |
e12b3daa | 3936 | pr_err("imsm: failed to associate spare\n"); |
ba2de7ba | 3937 | return 3; |
78d30f94 | 3938 | } |
3e372e5a | 3939 | first->anchor->num_raid_devs = sec->anchor->num_raid_devs; |
148acb7b | 3940 | first->anchor->orig_family_num = sec->anchor->orig_family_num; |
3e372e5a | 3941 | first->anchor->family_num = sec->anchor->family_num; |
ac6449be | 3942 | memcpy(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH); |
709743c5 DW |
3943 | for (i = 0; i < sec->anchor->num_raid_devs; i++) |
3944 | imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i)); | |
3e372e5a DW |
3945 | } |
3946 | ||
cdddbdbc DW |
3947 | return 0; |
3948 | } | |
3949 | ||
0030e8d6 DW |
3950 | static void fd2devname(int fd, char *name) |
3951 | { | |
3952 | struct stat st; | |
3953 | char path[256]; | |
33a6535d | 3954 | char dname[PATH_MAX]; |
0030e8d6 DW |
3955 | char *nm; |
3956 | int rv; | |
3957 | ||
3958 | name[0] = '\0'; | |
3959 | if (fstat(fd, &st) != 0) | |
3960 | return; | |
3961 | sprintf(path, "/sys/dev/block/%d:%d", | |
3962 | major(st.st_rdev), minor(st.st_rdev)); | |
3963 | ||
9cf014ec | 3964 | rv = readlink(path, dname, sizeof(dname)-1); |
0030e8d6 DW |
3965 | if (rv <= 0) |
3966 | return; | |
9587c373 | 3967 | |
0030e8d6 DW |
3968 | dname[rv] = '\0'; |
3969 | nm = strrchr(dname, '/'); | |
7897de29 JS |
3970 | if (nm) { |
3971 | nm++; | |
3972 | snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm); | |
3973 | } | |
0030e8d6 DW |
3974 | } |
3975 | ||
21e9380b AP |
3976 | static int nvme_get_serial(int fd, void *buf, size_t buf_len) |
3977 | { | |
3978 | char path[60]; | |
3979 | char *name = fd2kname(fd); | |
3980 | ||
3981 | if (!name) | |
3982 | return 1; | |
3983 | ||
3984 | if (strncmp(name, "nvme", 4) != 0) | |
3985 | return 1; | |
3986 | ||
3987 | snprintf(path, sizeof(path) - 1, "/sys/block/%s/device/serial", name); | |
3988 | ||
3989 | return load_sys(path, buf, buf_len); | |
3990 | } | |
3991 | ||
cdddbdbc DW |
3992 | extern int scsi_get_serial(int fd, void *buf, size_t buf_len); |
3993 | ||
3994 | static int imsm_read_serial(int fd, char *devname, | |
3995 | __u8 serial[MAX_RAID_SERIAL_LEN]) | |
3996 | { | |
21e9380b | 3997 | char buf[50]; |
cdddbdbc | 3998 | int rv; |
1f24f035 | 3999 | int len; |
316e2bf4 DW |
4000 | char *dest; |
4001 | char *src; | |
21e9380b AP |
4002 | unsigned int i; |
4003 | ||
4004 | memset(buf, 0, sizeof(buf)); | |
cdddbdbc | 4005 | |
21e9380b | 4006 | rv = nvme_get_serial(fd, buf, sizeof(buf)); |
cdddbdbc | 4007 | |
21e9380b AP |
4008 | if (rv) |
4009 | rv = scsi_get_serial(fd, buf, sizeof(buf)); | |
f9ba0ff1 | 4010 | |
40ebbb9c | 4011 | if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) { |
f9ba0ff1 DW |
4012 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
4013 | fd2devname(fd, (char *) serial); | |
0030e8d6 DW |
4014 | return 0; |
4015 | } | |
4016 | ||
cdddbdbc DW |
4017 | if (rv != 0) { |
4018 | if (devname) | |
e7b84f9d N |
4019 | pr_err("Failed to retrieve serial for %s\n", |
4020 | devname); | |
cdddbdbc DW |
4021 | return rv; |
4022 | } | |
4023 | ||
316e2bf4 DW |
4024 | /* trim all whitespace and non-printable characters and convert |
4025 | * ':' to ';' | |
4026 | */ | |
21e9380b AP |
4027 | for (i = 0, dest = buf; i < sizeof(buf) && buf[i]; i++) { |
4028 | src = &buf[i]; | |
316e2bf4 DW |
4029 | if (*src > 0x20) { |
4030 | /* ':' is reserved for use in placeholder serial | |
4031 | * numbers for missing disks | |
4032 | */ | |
4033 | if (*src == ':') | |
4034 | *dest++ = ';'; | |
4035 | else | |
4036 | *dest++ = *src; | |
4037 | } | |
4038 | } | |
21e9380b AP |
4039 | len = dest - buf; |
4040 | dest = buf; | |
316e2bf4 DW |
4041 | |
4042 | /* truncate leading characters */ | |
4043 | if (len > MAX_RAID_SERIAL_LEN) { | |
4044 | dest += len - MAX_RAID_SERIAL_LEN; | |
1f24f035 | 4045 | len = MAX_RAID_SERIAL_LEN; |
316e2bf4 | 4046 | } |
5c3db629 | 4047 | |
5c3db629 | 4048 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
316e2bf4 | 4049 | memcpy(serial, dest, len); |
cdddbdbc DW |
4050 | |
4051 | return 0; | |
4052 | } | |
4053 | ||
1f24f035 DW |
4054 | static int serialcmp(__u8 *s1, __u8 *s2) |
4055 | { | |
4056 | return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN); | |
4057 | } | |
4058 | ||
4059 | static void serialcpy(__u8 *dest, __u8 *src) | |
4060 | { | |
4061 | strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN); | |
4062 | } | |
4063 | ||
54c2c1ea DW |
4064 | static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super) |
4065 | { | |
4066 | struct dl *dl; | |
4067 | ||
4068 | for (dl = super->disks; dl; dl = dl->next) | |
4069 | if (serialcmp(dl->serial, serial) == 0) | |
4070 | break; | |
4071 | ||
4072 | return dl; | |
4073 | } | |
4074 | ||
a2b97981 DW |
4075 | static struct imsm_disk * |
4076 | __serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx) | |
4077 | { | |
4078 | int i; | |
4079 | ||
4080 | for (i = 0; i < mpb->num_disks; i++) { | |
4081 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
4082 | ||
4083 | if (serialcmp(disk->serial, serial) == 0) { | |
4084 | if (idx) | |
4085 | *idx = i; | |
4086 | return disk; | |
4087 | } | |
4088 | } | |
4089 | ||
4090 | return NULL; | |
4091 | } | |
4092 | ||
cdddbdbc DW |
4093 | static int |
4094 | load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd) | |
4095 | { | |
a2b97981 | 4096 | struct imsm_disk *disk; |
cdddbdbc DW |
4097 | struct dl *dl; |
4098 | struct stat stb; | |
cdddbdbc | 4099 | int rv; |
a2b97981 | 4100 | char name[40]; |
d23fe947 DW |
4101 | __u8 serial[MAX_RAID_SERIAL_LEN]; |
4102 | ||
4103 | rv = imsm_read_serial(fd, devname, serial); | |
4104 | ||
4105 | if (rv != 0) | |
4106 | return 2; | |
4107 | ||
503975b9 | 4108 | dl = xcalloc(1, sizeof(*dl)); |
cdddbdbc | 4109 | |
a2b97981 DW |
4110 | fstat(fd, &stb); |
4111 | dl->major = major(stb.st_rdev); | |
4112 | dl->minor = minor(stb.st_rdev); | |
4113 | dl->next = super->disks; | |
4114 | dl->fd = keep_fd ? fd : -1; | |
4115 | assert(super->disks == NULL); | |
4116 | super->disks = dl; | |
4117 | serialcpy(dl->serial, serial); | |
4118 | dl->index = -2; | |
4119 | dl->e = NULL; | |
4120 | fd2devname(fd, name); | |
4121 | if (devname) | |
503975b9 | 4122 | dl->devname = xstrdup(devname); |
a2b97981 | 4123 | else |
503975b9 | 4124 | dl->devname = xstrdup(name); |
cdddbdbc | 4125 | |
d23fe947 | 4126 | /* look up this disk's index in the current anchor */ |
a2b97981 DW |
4127 | disk = __serial_to_disk(dl->serial, super->anchor, &dl->index); |
4128 | if (disk) { | |
4129 | dl->disk = *disk; | |
4130 | /* only set index on disks that are a member of a | |
4131 | * populated contianer, i.e. one with raid_devs | |
4132 | */ | |
4133 | if (is_failed(&dl->disk)) | |
3f6efecc | 4134 | dl->index = -2; |
2432ce9b | 4135 | else if (is_spare(&dl->disk) || is_journal(&dl->disk)) |
a2b97981 | 4136 | dl->index = -1; |
3f6efecc DW |
4137 | } |
4138 | ||
949c47a0 DW |
4139 | return 0; |
4140 | } | |
4141 | ||
0c046afd DW |
4142 | /* When migrating map0 contains the 'destination' state while map1 |
4143 | * contains the current state. When not migrating map0 contains the | |
4144 | * current state. This routine assumes that map[0].map_state is set to | |
4145 | * the current array state before being called. | |
4146 | * | |
4147 | * Migration is indicated by one of the following states | |
4148 | * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed) | |
e3bba0e0 | 4149 | * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal |
0c046afd | 4150 | * map1state=unitialized) |
1484e727 | 4151 | * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal |
0c046afd | 4152 | * map1state=normal) |
e3bba0e0 | 4153 | * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal |
0c046afd | 4154 | * map1state=degraded) |
8e59f3d8 AK |
4155 | * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal |
4156 | * map1state=normal) | |
0c046afd | 4157 | */ |
8e59f3d8 AK |
4158 | static void migrate(struct imsm_dev *dev, struct intel_super *super, |
4159 | __u8 to_state, int migr_type) | |
3393c6af | 4160 | { |
0c046afd | 4161 | struct imsm_map *dest; |
238c0a71 | 4162 | struct imsm_map *src = get_imsm_map(dev, MAP_0); |
3393c6af | 4163 | |
0c046afd | 4164 | dev->vol.migr_state = 1; |
1484e727 | 4165 | set_migr_type(dev, migr_type); |
f8f603f1 | 4166 | dev->vol.curr_migr_unit = 0; |
238c0a71 | 4167 | dest = get_imsm_map(dev, MAP_1); |
0c046afd | 4168 | |
0556e1a2 | 4169 | /* duplicate and then set the target end state in map[0] */ |
3393c6af | 4170 | memcpy(dest, src, sizeof_imsm_map(src)); |
fb12a745 | 4171 | if (migr_type == MIGR_GEN_MIGR) { |
0556e1a2 DW |
4172 | __u32 ord; |
4173 | int i; | |
4174 | ||
4175 | for (i = 0; i < src->num_members; i++) { | |
4176 | ord = __le32_to_cpu(src->disk_ord_tbl[i]); | |
4177 | set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord)); | |
4178 | } | |
4179 | } | |
4180 | ||
8e59f3d8 AK |
4181 | if (migr_type == MIGR_GEN_MIGR) |
4182 | /* Clear migration record */ | |
4183 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
4184 | ||
0c046afd | 4185 | src->map_state = to_state; |
949c47a0 | 4186 | } |
f8f603f1 | 4187 | |
809da78e AK |
4188 | static void end_migration(struct imsm_dev *dev, struct intel_super *super, |
4189 | __u8 map_state) | |
f8f603f1 | 4190 | { |
238c0a71 AK |
4191 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
4192 | struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state == 0 ? | |
4193 | MAP_0 : MAP_1); | |
28bce06f | 4194 | int i, j; |
0556e1a2 DW |
4195 | |
4196 | /* merge any IMSM_ORD_REBUILD bits that were not successfully | |
4197 | * completed in the last migration. | |
4198 | * | |
28bce06f | 4199 | * FIXME add support for raid-level-migration |
0556e1a2 | 4200 | */ |
089f9d79 JS |
4201 | if (map_state != map->map_state && (is_gen_migration(dev) == 0) && |
4202 | prev->map_state != IMSM_T_STATE_UNINITIALIZED) { | |
809da78e AK |
4203 | /* when final map state is other than expected |
4204 | * merge maps (not for migration) | |
4205 | */ | |
4206 | int failed; | |
4207 | ||
4208 | for (i = 0; i < prev->num_members; i++) | |
4209 | for (j = 0; j < map->num_members; j++) | |
4210 | /* during online capacity expansion | |
4211 | * disks position can be changed | |
4212 | * if takeover is used | |
4213 | */ | |
4214 | if (ord_to_idx(map->disk_ord_tbl[j]) == | |
4215 | ord_to_idx(prev->disk_ord_tbl[i])) { | |
4216 | map->disk_ord_tbl[j] |= | |
4217 | prev->disk_ord_tbl[i]; | |
4218 | break; | |
4219 | } | |
4220 | failed = imsm_count_failed(super, dev, MAP_0); | |
4221 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
4222 | } | |
f8f603f1 DW |
4223 | |
4224 | dev->vol.migr_state = 0; | |
ea672ee1 | 4225 | set_migr_type(dev, 0); |
f8f603f1 DW |
4226 | dev->vol.curr_migr_unit = 0; |
4227 | map->map_state = map_state; | |
4228 | } | |
949c47a0 DW |
4229 | |
4230 | static int parse_raid_devices(struct intel_super *super) | |
4231 | { | |
4232 | int i; | |
4233 | struct imsm_dev *dev_new; | |
4d7b1503 | 4234 | size_t len, len_migr; |
401d313b | 4235 | size_t max_len = 0; |
4d7b1503 DW |
4236 | size_t space_needed = 0; |
4237 | struct imsm_super *mpb = super->anchor; | |
949c47a0 DW |
4238 | |
4239 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
4240 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
ba2de7ba | 4241 | struct intel_dev *dv; |
949c47a0 | 4242 | |
4d7b1503 DW |
4243 | len = sizeof_imsm_dev(dev_iter, 0); |
4244 | len_migr = sizeof_imsm_dev(dev_iter, 1); | |
4245 | if (len_migr > len) | |
4246 | space_needed += len_migr - len; | |
ca9de185 | 4247 | |
503975b9 | 4248 | dv = xmalloc(sizeof(*dv)); |
401d313b AK |
4249 | if (max_len < len_migr) |
4250 | max_len = len_migr; | |
4251 | if (max_len > len_migr) | |
4252 | space_needed += max_len - len_migr; | |
503975b9 | 4253 | dev_new = xmalloc(max_len); |
949c47a0 | 4254 | imsm_copy_dev(dev_new, dev_iter); |
ba2de7ba DW |
4255 | dv->dev = dev_new; |
4256 | dv->index = i; | |
4257 | dv->next = super->devlist; | |
4258 | super->devlist = dv; | |
949c47a0 | 4259 | } |
cdddbdbc | 4260 | |
4d7b1503 DW |
4261 | /* ensure that super->buf is large enough when all raid devices |
4262 | * are migrating | |
4263 | */ | |
4264 | if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) { | |
4265 | void *buf; | |
4266 | ||
f36a9ecd PB |
4267 | len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed, |
4268 | super->sector_size); | |
4269 | if (posix_memalign(&buf, MAX_SECTOR_SIZE, len) != 0) | |
4d7b1503 DW |
4270 | return 1; |
4271 | ||
1f45a8ad DW |
4272 | memcpy(buf, super->buf, super->len); |
4273 | memset(buf + super->len, 0, len - super->len); | |
4d7b1503 DW |
4274 | free(super->buf); |
4275 | super->buf = buf; | |
4276 | super->len = len; | |
4277 | } | |
ca9de185 | 4278 | |
bbab0940 TM |
4279 | super->extra_space += space_needed; |
4280 | ||
cdddbdbc DW |
4281 | return 0; |
4282 | } | |
4283 | ||
e2f41b2c AK |
4284 | /******************************************************************************* |
4285 | * Function: check_mpb_migr_compatibility | |
4286 | * Description: Function checks for unsupported migration features: | |
4287 | * - migration optimization area (pba_of_lba0) | |
4288 | * - descending reshape (ascending_migr) | |
4289 | * Parameters: | |
4290 | * super : imsm metadata information | |
4291 | * Returns: | |
4292 | * 0 : migration is compatible | |
4293 | * -1 : migration is not compatible | |
4294 | ******************************************************************************/ | |
4295 | int check_mpb_migr_compatibility(struct intel_super *super) | |
4296 | { | |
4297 | struct imsm_map *map0, *map1; | |
4298 | struct migr_record *migr_rec = super->migr_rec; | |
4299 | int i; | |
4300 | ||
4301 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
4302 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
4303 | ||
4304 | if (dev_iter && | |
4305 | dev_iter->vol.migr_state == 1 && | |
4306 | dev_iter->vol.migr_type == MIGR_GEN_MIGR) { | |
4307 | /* This device is migrating */ | |
238c0a71 AK |
4308 | map0 = get_imsm_map(dev_iter, MAP_0); |
4309 | map1 = get_imsm_map(dev_iter, MAP_1); | |
5551b113 | 4310 | if (pba_of_lba0(map0) != pba_of_lba0(map1)) |
e2f41b2c AK |
4311 | /* migration optimization area was used */ |
4312 | return -1; | |
fc54fe7a JS |
4313 | if (migr_rec->ascending_migr == 0 && |
4314 | migr_rec->dest_depth_per_unit > 0) | |
e2f41b2c AK |
4315 | /* descending reshape not supported yet */ |
4316 | return -1; | |
4317 | } | |
4318 | } | |
4319 | return 0; | |
4320 | } | |
4321 | ||
d23fe947 | 4322 | static void __free_imsm(struct intel_super *super, int free_disks); |
9ca2c81c | 4323 | |
cdddbdbc | 4324 | /* load_imsm_mpb - read matrix metadata |
f2f5c343 | 4325 | * allocates super->mpb to be freed by free_imsm |
cdddbdbc DW |
4326 | */ |
4327 | static int load_imsm_mpb(int fd, struct intel_super *super, char *devname) | |
4328 | { | |
4329 | unsigned long long dsize; | |
cdddbdbc | 4330 | unsigned long long sectors; |
f36a9ecd | 4331 | unsigned int sector_size = super->sector_size; |
cdddbdbc | 4332 | struct stat; |
6416d527 | 4333 | struct imsm_super *anchor; |
cdddbdbc DW |
4334 | __u32 check_sum; |
4335 | ||
cdddbdbc | 4336 | get_dev_size(fd, NULL, &dsize); |
f36a9ecd | 4337 | if (dsize < 2*sector_size) { |
64436f06 | 4338 | if (devname) |
e7b84f9d N |
4339 | pr_err("%s: device to small for imsm\n", |
4340 | devname); | |
64436f06 N |
4341 | return 1; |
4342 | } | |
cdddbdbc | 4343 | |
f36a9ecd | 4344 | if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0) { |
cdddbdbc | 4345 | if (devname) |
e7b84f9d N |
4346 | pr_err("Cannot seek to anchor block on %s: %s\n", |
4347 | devname, strerror(errno)); | |
cdddbdbc DW |
4348 | return 1; |
4349 | } | |
4350 | ||
f36a9ecd | 4351 | if (posix_memalign((void **)&anchor, sector_size, sector_size) != 0) { |
ad97895e | 4352 | if (devname) |
7a862a02 | 4353 | pr_err("Failed to allocate imsm anchor buffer on %s\n", devname); |
ad97895e DW |
4354 | return 1; |
4355 | } | |
466070ad | 4356 | if ((unsigned int)read(fd, anchor, sector_size) != sector_size) { |
cdddbdbc | 4357 | if (devname) |
e7b84f9d N |
4358 | pr_err("Cannot read anchor block on %s: %s\n", |
4359 | devname, strerror(errno)); | |
6416d527 | 4360 | free(anchor); |
cdddbdbc DW |
4361 | return 1; |
4362 | } | |
4363 | ||
6416d527 | 4364 | if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) { |
cdddbdbc | 4365 | if (devname) |
e7b84f9d | 4366 | pr_err("no IMSM anchor on %s\n", devname); |
6416d527 | 4367 | free(anchor); |
cdddbdbc DW |
4368 | return 2; |
4369 | } | |
4370 | ||
d23fe947 | 4371 | __free_imsm(super, 0); |
f2f5c343 LM |
4372 | /* reload capability and hba */ |
4373 | ||
4374 | /* capability and hba must be updated with new super allocation */ | |
d424212e | 4375 | find_intel_hba_capability(fd, super, devname); |
f36a9ecd PB |
4376 | super->len = ROUND_UP(anchor->mpb_size, sector_size); |
4377 | if (posix_memalign(&super->buf, MAX_SECTOR_SIZE, super->len) != 0) { | |
cdddbdbc | 4378 | if (devname) |
e7b84f9d N |
4379 | pr_err("unable to allocate %zu byte mpb buffer\n", |
4380 | super->len); | |
6416d527 | 4381 | free(anchor); |
cdddbdbc DW |
4382 | return 2; |
4383 | } | |
f36a9ecd | 4384 | memcpy(super->buf, anchor, sector_size); |
cdddbdbc | 4385 | |
f36a9ecd | 4386 | sectors = mpb_sectors(anchor, sector_size) - 1; |
6416d527 | 4387 | free(anchor); |
8e59f3d8 | 4388 | |
85337573 AO |
4389 | if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE, |
4390 | MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) { | |
1ade5cc1 | 4391 | pr_err("could not allocate migr_rec buffer\n"); |
8e59f3d8 AK |
4392 | free(super->buf); |
4393 | return 2; | |
4394 | } | |
51d83f5d | 4395 | super->clean_migration_record_by_mdmon = 0; |
8e59f3d8 | 4396 | |
949c47a0 | 4397 | if (!sectors) { |
ecf45690 DW |
4398 | check_sum = __gen_imsm_checksum(super->anchor); |
4399 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
4400 | if (devname) | |
e7b84f9d N |
4401 | pr_err("IMSM checksum %x != %x on %s\n", |
4402 | check_sum, | |
4403 | __le32_to_cpu(super->anchor->check_sum), | |
4404 | devname); | |
ecf45690 DW |
4405 | return 2; |
4406 | } | |
4407 | ||
a2b97981 | 4408 | return 0; |
949c47a0 | 4409 | } |
cdddbdbc DW |
4410 | |
4411 | /* read the extended mpb */ | |
f36a9ecd | 4412 | if (lseek64(fd, dsize - (sector_size * (2 + sectors)), SEEK_SET) < 0) { |
cdddbdbc | 4413 | if (devname) |
e7b84f9d N |
4414 | pr_err("Cannot seek to extended mpb on %s: %s\n", |
4415 | devname, strerror(errno)); | |
cdddbdbc DW |
4416 | return 1; |
4417 | } | |
4418 | ||
f36a9ecd PB |
4419 | if ((unsigned int)read(fd, super->buf + sector_size, |
4420 | super->len - sector_size) != super->len - sector_size) { | |
cdddbdbc | 4421 | if (devname) |
e7b84f9d N |
4422 | pr_err("Cannot read extended mpb on %s: %s\n", |
4423 | devname, strerror(errno)); | |
cdddbdbc DW |
4424 | return 2; |
4425 | } | |
4426 | ||
949c47a0 DW |
4427 | check_sum = __gen_imsm_checksum(super->anchor); |
4428 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
cdddbdbc | 4429 | if (devname) |
e7b84f9d N |
4430 | pr_err("IMSM checksum %x != %x on %s\n", |
4431 | check_sum, __le32_to_cpu(super->anchor->check_sum), | |
4432 | devname); | |
db575f3b | 4433 | return 3; |
cdddbdbc DW |
4434 | } |
4435 | ||
a2b97981 DW |
4436 | return 0; |
4437 | } | |
4438 | ||
8e59f3d8 AK |
4439 | static int read_imsm_migr_rec(int fd, struct intel_super *super); |
4440 | ||
97f81ee2 CA |
4441 | /* clears hi bits in metadata if MPB_ATTRIB_2TB_DISK not set */ |
4442 | static void clear_hi(struct intel_super *super) | |
4443 | { | |
4444 | struct imsm_super *mpb = super->anchor; | |
4445 | int i, n; | |
4446 | if (mpb->attributes & MPB_ATTRIB_2TB_DISK) | |
4447 | return; | |
4448 | for (i = 0; i < mpb->num_disks; ++i) { | |
4449 | struct imsm_disk *disk = &mpb->disk[i]; | |
4450 | disk->total_blocks_hi = 0; | |
4451 | } | |
4452 | for (i = 0; i < mpb->num_raid_devs; ++i) { | |
4453 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
4454 | if (!dev) | |
4455 | return; | |
4456 | for (n = 0; n < 2; ++n) { | |
4457 | struct imsm_map *map = get_imsm_map(dev, n); | |
4458 | if (!map) | |
4459 | continue; | |
4460 | map->pba_of_lba0_hi = 0; | |
4461 | map->blocks_per_member_hi = 0; | |
4462 | map->num_data_stripes_hi = 0; | |
4463 | } | |
4464 | } | |
4465 | } | |
4466 | ||
a2b97981 DW |
4467 | static int |
4468 | load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd) | |
4469 | { | |
4470 | int err; | |
4471 | ||
4472 | err = load_imsm_mpb(fd, super, devname); | |
4473 | if (err) | |
4474 | return err; | |
f36a9ecd PB |
4475 | if (super->sector_size == 4096) |
4476 | convert_from_4k(super); | |
a2b97981 DW |
4477 | err = load_imsm_disk(fd, super, devname, keep_fd); |
4478 | if (err) | |
4479 | return err; | |
4480 | err = parse_raid_devices(super); | |
8d67477f TM |
4481 | if (err) |
4482 | return err; | |
4483 | err = load_bbm_log(super); | |
97f81ee2 | 4484 | clear_hi(super); |
a2b97981 | 4485 | return err; |
cdddbdbc DW |
4486 | } |
4487 | ||
ae6aad82 DW |
4488 | static void __free_imsm_disk(struct dl *d) |
4489 | { | |
4490 | if (d->fd >= 0) | |
4491 | close(d->fd); | |
4492 | if (d->devname) | |
4493 | free(d->devname); | |
0dcecb2e DW |
4494 | if (d->e) |
4495 | free(d->e); | |
ae6aad82 DW |
4496 | free(d); |
4497 | ||
4498 | } | |
1a64be56 | 4499 | |
cdddbdbc DW |
4500 | static void free_imsm_disks(struct intel_super *super) |
4501 | { | |
47ee5a45 | 4502 | struct dl *d; |
cdddbdbc | 4503 | |
47ee5a45 DW |
4504 | while (super->disks) { |
4505 | d = super->disks; | |
cdddbdbc | 4506 | super->disks = d->next; |
ae6aad82 | 4507 | __free_imsm_disk(d); |
cdddbdbc | 4508 | } |
cb82edca AK |
4509 | while (super->disk_mgmt_list) { |
4510 | d = super->disk_mgmt_list; | |
4511 | super->disk_mgmt_list = d->next; | |
4512 | __free_imsm_disk(d); | |
4513 | } | |
47ee5a45 DW |
4514 | while (super->missing) { |
4515 | d = super->missing; | |
4516 | super->missing = d->next; | |
4517 | __free_imsm_disk(d); | |
4518 | } | |
4519 | ||
cdddbdbc DW |
4520 | } |
4521 | ||
9ca2c81c | 4522 | /* free all the pieces hanging off of a super pointer */ |
d23fe947 | 4523 | static void __free_imsm(struct intel_super *super, int free_disks) |
cdddbdbc | 4524 | { |
88654014 LM |
4525 | struct intel_hba *elem, *next; |
4526 | ||
9ca2c81c | 4527 | if (super->buf) { |
949c47a0 | 4528 | free(super->buf); |
9ca2c81c DW |
4529 | super->buf = NULL; |
4530 | } | |
f2f5c343 LM |
4531 | /* unlink capability description */ |
4532 | super->orom = NULL; | |
8e59f3d8 AK |
4533 | if (super->migr_rec_buf) { |
4534 | free(super->migr_rec_buf); | |
4535 | super->migr_rec_buf = NULL; | |
4536 | } | |
d23fe947 DW |
4537 | if (free_disks) |
4538 | free_imsm_disks(super); | |
ba2de7ba | 4539 | free_devlist(super); |
88654014 LM |
4540 | elem = super->hba; |
4541 | while (elem) { | |
4542 | if (elem->path) | |
4543 | free((void *)elem->path); | |
4544 | next = elem->next; | |
4545 | free(elem); | |
4546 | elem = next; | |
88c32bb1 | 4547 | } |
8d67477f TM |
4548 | if (super->bbm_log) |
4549 | free(super->bbm_log); | |
88654014 | 4550 | super->hba = NULL; |
cdddbdbc DW |
4551 | } |
4552 | ||
9ca2c81c DW |
4553 | static void free_imsm(struct intel_super *super) |
4554 | { | |
d23fe947 | 4555 | __free_imsm(super, 1); |
928f1424 | 4556 | free(super->bb.entries); |
9ca2c81c DW |
4557 | free(super); |
4558 | } | |
cdddbdbc DW |
4559 | |
4560 | static void free_super_imsm(struct supertype *st) | |
4561 | { | |
4562 | struct intel_super *super = st->sb; | |
4563 | ||
4564 | if (!super) | |
4565 | return; | |
4566 | ||
4567 | free_imsm(super); | |
4568 | st->sb = NULL; | |
4569 | } | |
4570 | ||
49133e57 | 4571 | static struct intel_super *alloc_super(void) |
c2c087e6 | 4572 | { |
503975b9 | 4573 | struct intel_super *super = xcalloc(1, sizeof(*super)); |
c2c087e6 | 4574 | |
503975b9 N |
4575 | super->current_vol = -1; |
4576 | super->create_offset = ~((unsigned long long) 0); | |
928f1424 TM |
4577 | |
4578 | super->bb.entries = xmalloc(BBM_LOG_MAX_ENTRIES * | |
4579 | sizeof(struct md_bb_entry)); | |
4580 | if (!super->bb.entries) { | |
4581 | free(super); | |
4582 | return NULL; | |
4583 | } | |
4584 | ||
c2c087e6 DW |
4585 | return super; |
4586 | } | |
4587 | ||
f0f5a016 LM |
4588 | /* |
4589 | * find and allocate hba and OROM/EFI based on valid fd of RAID component device | |
4590 | */ | |
d424212e | 4591 | static int find_intel_hba_capability(int fd, struct intel_super *super, char *devname) |
f0f5a016 LM |
4592 | { |
4593 | struct sys_dev *hba_name; | |
4594 | int rv = 0; | |
4595 | ||
3a30e28e MT |
4596 | if (fd >= 0 && test_partition(fd)) { |
4597 | pr_err("imsm: %s is a partition, cannot be used in IMSM\n", | |
4598 | devname); | |
4599 | return 1; | |
4600 | } | |
089f9d79 | 4601 | if (fd < 0 || check_env("IMSM_NO_PLATFORM")) { |
f2f5c343 | 4602 | super->orom = NULL; |
f0f5a016 LM |
4603 | super->hba = NULL; |
4604 | return 0; | |
4605 | } | |
4606 | hba_name = find_disk_attached_hba(fd, NULL); | |
4607 | if (!hba_name) { | |
d424212e | 4608 | if (devname) |
e7b84f9d N |
4609 | pr_err("%s is not attached to Intel(R) RAID controller.\n", |
4610 | devname); | |
f0f5a016 LM |
4611 | return 1; |
4612 | } | |
4613 | rv = attach_hba_to_super(super, hba_name); | |
4614 | if (rv == 2) { | |
d424212e N |
4615 | if (devname) { |
4616 | struct intel_hba *hba = super->hba; | |
f0f5a016 | 4617 | |
60f0f54d PB |
4618 | pr_err("%s is attached to Intel(R) %s %s (%s),\n" |
4619 | " but the container is assigned to Intel(R) %s %s (", | |
d424212e | 4620 | devname, |
614902f6 | 4621 | get_sys_dev_type(hba_name->type), |
60f0f54d | 4622 | hba_name->type == SYS_DEV_VMD ? "domain" : "RAID controller", |
f0f5a016 | 4623 | hba_name->pci_id ? : "Err!", |
60f0f54d PB |
4624 | get_sys_dev_type(super->hba->type), |
4625 | hba->type == SYS_DEV_VMD ? "domain" : "RAID controller"); | |
f0f5a016 | 4626 | |
f0f5a016 LM |
4627 | while (hba) { |
4628 | fprintf(stderr, "%s", hba->pci_id ? : "Err!"); | |
4629 | if (hba->next) | |
4630 | fprintf(stderr, ", "); | |
4631 | hba = hba->next; | |
4632 | } | |
6b781d33 | 4633 | fprintf(stderr, ").\n" |
cca67208 | 4634 | " Mixing devices attached to different controllers is not allowed.\n"); |
f0f5a016 | 4635 | } |
f0f5a016 LM |
4636 | return 2; |
4637 | } | |
6b781d33 | 4638 | super->orom = find_imsm_capability(hba_name); |
f2f5c343 LM |
4639 | if (!super->orom) |
4640 | return 3; | |
614902f6 | 4641 | |
f0f5a016 LM |
4642 | return 0; |
4643 | } | |
4644 | ||
47ee5a45 DW |
4645 | /* find_missing - helper routine for load_super_imsm_all that identifies |
4646 | * disks that have disappeared from the system. This routine relies on | |
4647 | * the mpb being uptodate, which it is at load time. | |
4648 | */ | |
4649 | static int find_missing(struct intel_super *super) | |
4650 | { | |
4651 | int i; | |
4652 | struct imsm_super *mpb = super->anchor; | |
4653 | struct dl *dl; | |
4654 | struct imsm_disk *disk; | |
47ee5a45 DW |
4655 | |
4656 | for (i = 0; i < mpb->num_disks; i++) { | |
4657 | disk = __get_imsm_disk(mpb, i); | |
54c2c1ea | 4658 | dl = serial_to_dl(disk->serial, super); |
47ee5a45 DW |
4659 | if (dl) |
4660 | continue; | |
47ee5a45 | 4661 | |
503975b9 | 4662 | dl = xmalloc(sizeof(*dl)); |
47ee5a45 DW |
4663 | dl->major = 0; |
4664 | dl->minor = 0; | |
4665 | dl->fd = -1; | |
503975b9 | 4666 | dl->devname = xstrdup("missing"); |
47ee5a45 DW |
4667 | dl->index = i; |
4668 | serialcpy(dl->serial, disk->serial); | |
4669 | dl->disk = *disk; | |
689c9bf3 | 4670 | dl->e = NULL; |
47ee5a45 DW |
4671 | dl->next = super->missing; |
4672 | super->missing = dl; | |
4673 | } | |
4674 | ||
4675 | return 0; | |
4676 | } | |
4677 | ||
a2b97981 DW |
4678 | static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list) |
4679 | { | |
4680 | struct intel_disk *idisk = disk_list; | |
4681 | ||
4682 | while (idisk) { | |
4683 | if (serialcmp(idisk->disk.serial, serial) == 0) | |
4684 | break; | |
4685 | idisk = idisk->next; | |
4686 | } | |
4687 | ||
4688 | return idisk; | |
4689 | } | |
4690 | ||
4691 | static int __prep_thunderdome(struct intel_super **table, int tbl_size, | |
4692 | struct intel_super *super, | |
4693 | struct intel_disk **disk_list) | |
4694 | { | |
4695 | struct imsm_disk *d = &super->disks->disk; | |
4696 | struct imsm_super *mpb = super->anchor; | |
4697 | int i, j; | |
4698 | ||
4699 | for (i = 0; i < tbl_size; i++) { | |
4700 | struct imsm_super *tbl_mpb = table[i]->anchor; | |
4701 | struct imsm_disk *tbl_d = &table[i]->disks->disk; | |
4702 | ||
4703 | if (tbl_mpb->family_num == mpb->family_num) { | |
4704 | if (tbl_mpb->check_sum == mpb->check_sum) { | |
1ade5cc1 N |
4705 | dprintf("mpb from %d:%d matches %d:%d\n", |
4706 | super->disks->major, | |
a2b97981 DW |
4707 | super->disks->minor, |
4708 | table[i]->disks->major, | |
4709 | table[i]->disks->minor); | |
4710 | break; | |
4711 | } | |
4712 | ||
4713 | if (((is_configured(d) && !is_configured(tbl_d)) || | |
4714 | is_configured(d) == is_configured(tbl_d)) && | |
4715 | tbl_mpb->generation_num < mpb->generation_num) { | |
4716 | /* current version of the mpb is a | |
4717 | * better candidate than the one in | |
4718 | * super_table, but copy over "cross | |
4719 | * generational" status | |
4720 | */ | |
4721 | struct intel_disk *idisk; | |
4722 | ||
1ade5cc1 N |
4723 | dprintf("mpb from %d:%d replaces %d:%d\n", |
4724 | super->disks->major, | |
a2b97981 DW |
4725 | super->disks->minor, |
4726 | table[i]->disks->major, | |
4727 | table[i]->disks->minor); | |
4728 | ||
4729 | idisk = disk_list_get(tbl_d->serial, *disk_list); | |
4730 | if (idisk && is_failed(&idisk->disk)) | |
4731 | tbl_d->status |= FAILED_DISK; | |
4732 | break; | |
4733 | } else { | |
4734 | struct intel_disk *idisk; | |
4735 | struct imsm_disk *disk; | |
4736 | ||
4737 | /* tbl_mpb is more up to date, but copy | |
4738 | * over cross generational status before | |
4739 | * returning | |
4740 | */ | |
4741 | disk = __serial_to_disk(d->serial, mpb, NULL); | |
4742 | if (disk && is_failed(disk)) | |
4743 | d->status |= FAILED_DISK; | |
4744 | ||
4745 | idisk = disk_list_get(d->serial, *disk_list); | |
4746 | if (idisk) { | |
4747 | idisk->owner = i; | |
4748 | if (disk && is_configured(disk)) | |
4749 | idisk->disk.status |= CONFIGURED_DISK; | |
4750 | } | |
4751 | ||
1ade5cc1 N |
4752 | dprintf("mpb from %d:%d prefer %d:%d\n", |
4753 | super->disks->major, | |
a2b97981 DW |
4754 | super->disks->minor, |
4755 | table[i]->disks->major, | |
4756 | table[i]->disks->minor); | |
4757 | ||
4758 | return tbl_size; | |
4759 | } | |
4760 | } | |
4761 | } | |
4762 | ||
4763 | if (i >= tbl_size) | |
4764 | table[tbl_size++] = super; | |
4765 | else | |
4766 | table[i] = super; | |
4767 | ||
4768 | /* update/extend the merged list of imsm_disk records */ | |
4769 | for (j = 0; j < mpb->num_disks; j++) { | |
4770 | struct imsm_disk *disk = __get_imsm_disk(mpb, j); | |
4771 | struct intel_disk *idisk; | |
4772 | ||
4773 | idisk = disk_list_get(disk->serial, *disk_list); | |
4774 | if (idisk) { | |
4775 | idisk->disk.status |= disk->status; | |
4776 | if (is_configured(&idisk->disk) || | |
4777 | is_failed(&idisk->disk)) | |
4778 | idisk->disk.status &= ~(SPARE_DISK); | |
4779 | } else { | |
503975b9 | 4780 | idisk = xcalloc(1, sizeof(*idisk)); |
a2b97981 DW |
4781 | idisk->owner = IMSM_UNKNOWN_OWNER; |
4782 | idisk->disk = *disk; | |
4783 | idisk->next = *disk_list; | |
4784 | *disk_list = idisk; | |
4785 | } | |
4786 | ||
4787 | if (serialcmp(idisk->disk.serial, d->serial) == 0) | |
4788 | idisk->owner = i; | |
4789 | } | |
4790 | ||
4791 | return tbl_size; | |
4792 | } | |
4793 | ||
4794 | static struct intel_super * | |
4795 | validate_members(struct intel_super *super, struct intel_disk *disk_list, | |
4796 | const int owner) | |
4797 | { | |
4798 | struct imsm_super *mpb = super->anchor; | |
4799 | int ok_count = 0; | |
4800 | int i; | |
4801 | ||
4802 | for (i = 0; i < mpb->num_disks; i++) { | |
4803 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
4804 | struct intel_disk *idisk; | |
4805 | ||
4806 | idisk = disk_list_get(disk->serial, disk_list); | |
4807 | if (idisk) { | |
4808 | if (idisk->owner == owner || | |
4809 | idisk->owner == IMSM_UNKNOWN_OWNER) | |
4810 | ok_count++; | |
4811 | else | |
1ade5cc1 N |
4812 | dprintf("'%.16s' owner %d != %d\n", |
4813 | disk->serial, idisk->owner, | |
a2b97981 DW |
4814 | owner); |
4815 | } else { | |
1ade5cc1 N |
4816 | dprintf("unknown disk %x [%d]: %.16s\n", |
4817 | __le32_to_cpu(mpb->family_num), i, | |
a2b97981 DW |
4818 | disk->serial); |
4819 | break; | |
4820 | } | |
4821 | } | |
4822 | ||
4823 | if (ok_count == mpb->num_disks) | |
4824 | return super; | |
4825 | return NULL; | |
4826 | } | |
4827 | ||
4828 | static void show_conflicts(__u32 family_num, struct intel_super *super_list) | |
4829 | { | |
4830 | struct intel_super *s; | |
4831 | ||
4832 | for (s = super_list; s; s = s->next) { | |
4833 | if (family_num != s->anchor->family_num) | |
4834 | continue; | |
e12b3daa | 4835 | pr_err("Conflict, offlining family %#x on '%s'\n", |
a2b97981 DW |
4836 | __le32_to_cpu(family_num), s->disks->devname); |
4837 | } | |
4838 | } | |
4839 | ||
4840 | static struct intel_super * | |
4841 | imsm_thunderdome(struct intel_super **super_list, int len) | |
4842 | { | |
4843 | struct intel_super *super_table[len]; | |
4844 | struct intel_disk *disk_list = NULL; | |
4845 | struct intel_super *champion, *spare; | |
4846 | struct intel_super *s, **del; | |
4847 | int tbl_size = 0; | |
4848 | int conflict; | |
4849 | int i; | |
4850 | ||
4851 | memset(super_table, 0, sizeof(super_table)); | |
4852 | for (s = *super_list; s; s = s->next) | |
4853 | tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list); | |
4854 | ||
4855 | for (i = 0; i < tbl_size; i++) { | |
4856 | struct imsm_disk *d; | |
4857 | struct intel_disk *idisk; | |
4858 | struct imsm_super *mpb = super_table[i]->anchor; | |
4859 | ||
4860 | s = super_table[i]; | |
4861 | d = &s->disks->disk; | |
4862 | ||
4863 | /* 'd' must appear in merged disk list for its | |
4864 | * configuration to be valid | |
4865 | */ | |
4866 | idisk = disk_list_get(d->serial, disk_list); | |
4867 | if (idisk && idisk->owner == i) | |
4868 | s = validate_members(s, disk_list, i); | |
4869 | else | |
4870 | s = NULL; | |
4871 | ||
4872 | if (!s) | |
1ade5cc1 N |
4873 | dprintf("marking family: %#x from %d:%d offline\n", |
4874 | mpb->family_num, | |
a2b97981 DW |
4875 | super_table[i]->disks->major, |
4876 | super_table[i]->disks->minor); | |
4877 | super_table[i] = s; | |
4878 | } | |
4879 | ||
4880 | /* This is where the mdadm implementation differs from the Windows | |
4881 | * driver which has no strict concept of a container. We can only | |
4882 | * assemble one family from a container, so when returning a prodigal | |
4883 | * array member to this system the code will not be able to disambiguate | |
4884 | * the container contents that should be assembled ("foreign" versus | |
4885 | * "local"). It requires user intervention to set the orig_family_num | |
4886 | * to a new value to establish a new container. The Windows driver in | |
4887 | * this situation fixes up the volume name in place and manages the | |
4888 | * foreign array as an independent entity. | |
4889 | */ | |
4890 | s = NULL; | |
4891 | spare = NULL; | |
4892 | conflict = 0; | |
4893 | for (i = 0; i < tbl_size; i++) { | |
4894 | struct intel_super *tbl_ent = super_table[i]; | |
4895 | int is_spare = 0; | |
4896 | ||
4897 | if (!tbl_ent) | |
4898 | continue; | |
4899 | ||
4900 | if (tbl_ent->anchor->num_raid_devs == 0) { | |
4901 | spare = tbl_ent; | |
4902 | is_spare = 1; | |
4903 | } | |
4904 | ||
4905 | if (s && !is_spare) { | |
4906 | show_conflicts(tbl_ent->anchor->family_num, *super_list); | |
4907 | conflict++; | |
4908 | } else if (!s && !is_spare) | |
4909 | s = tbl_ent; | |
4910 | } | |
4911 | ||
4912 | if (!s) | |
4913 | s = spare; | |
4914 | if (!s) { | |
4915 | champion = NULL; | |
4916 | goto out; | |
4917 | } | |
4918 | champion = s; | |
4919 | ||
4920 | if (conflict) | |
7a862a02 | 4921 | pr_err("Chose family %#x on '%s', assemble conflicts to new container with '--update=uuid'\n", |
a2b97981 DW |
4922 | __le32_to_cpu(s->anchor->family_num), s->disks->devname); |
4923 | ||
4924 | /* collect all dl's onto 'champion', and update them to | |
4925 | * champion's version of the status | |
4926 | */ | |
4927 | for (s = *super_list; s; s = s->next) { | |
4928 | struct imsm_super *mpb = champion->anchor; | |
4929 | struct dl *dl = s->disks; | |
4930 | ||
4931 | if (s == champion) | |
4932 | continue; | |
4933 | ||
5d7b407a CA |
4934 | mpb->attributes |= s->anchor->attributes & MPB_ATTRIB_2TB_DISK; |
4935 | ||
a2b97981 DW |
4936 | for (i = 0; i < mpb->num_disks; i++) { |
4937 | struct imsm_disk *disk; | |
4938 | ||
4939 | disk = __serial_to_disk(dl->serial, mpb, &dl->index); | |
4940 | if (disk) { | |
4941 | dl->disk = *disk; | |
4942 | /* only set index on disks that are a member of | |
4943 | * a populated contianer, i.e. one with | |
4944 | * raid_devs | |
4945 | */ | |
4946 | if (is_failed(&dl->disk)) | |
4947 | dl->index = -2; | |
4948 | else if (is_spare(&dl->disk)) | |
4949 | dl->index = -1; | |
4950 | break; | |
4951 | } | |
4952 | } | |
4953 | ||
4954 | if (i >= mpb->num_disks) { | |
4955 | struct intel_disk *idisk; | |
4956 | ||
4957 | idisk = disk_list_get(dl->serial, disk_list); | |
ecf408e9 | 4958 | if (idisk && is_spare(&idisk->disk) && |
a2b97981 DW |
4959 | !is_failed(&idisk->disk) && !is_configured(&idisk->disk)) |
4960 | dl->index = -1; | |
4961 | else { | |
4962 | dl->index = -2; | |
4963 | continue; | |
4964 | } | |
4965 | } | |
4966 | ||
4967 | dl->next = champion->disks; | |
4968 | champion->disks = dl; | |
4969 | s->disks = NULL; | |
4970 | } | |
4971 | ||
4972 | /* delete 'champion' from super_list */ | |
4973 | for (del = super_list; *del; ) { | |
4974 | if (*del == champion) { | |
4975 | *del = (*del)->next; | |
4976 | break; | |
4977 | } else | |
4978 | del = &(*del)->next; | |
4979 | } | |
4980 | champion->next = NULL; | |
4981 | ||
4982 | out: | |
4983 | while (disk_list) { | |
4984 | struct intel_disk *idisk = disk_list; | |
4985 | ||
4986 | disk_list = disk_list->next; | |
4987 | free(idisk); | |
4988 | } | |
4989 | ||
4990 | return champion; | |
4991 | } | |
4992 | ||
9587c373 LM |
4993 | static int |
4994 | get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd); | |
4dd2df09 | 4995 | static int get_super_block(struct intel_super **super_list, char *devnm, char *devname, |
9587c373 | 4996 | int major, int minor, int keep_fd); |
ec50f7b6 LM |
4997 | static int |
4998 | get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list, | |
4999 | int *max, int keep_fd); | |
5000 | ||
cdddbdbc | 5001 | static int load_super_imsm_all(struct supertype *st, int fd, void **sbp, |
ec50f7b6 LM |
5002 | char *devname, struct md_list *devlist, |
5003 | int keep_fd) | |
cdddbdbc | 5004 | { |
a2b97981 DW |
5005 | struct intel_super *super_list = NULL; |
5006 | struct intel_super *super = NULL; | |
a2b97981 | 5007 | int err = 0; |
9587c373 | 5008 | int i = 0; |
dab4a513 | 5009 | |
9587c373 LM |
5010 | if (fd >= 0) |
5011 | /* 'fd' is an opened container */ | |
5012 | err = get_sra_super_block(fd, &super_list, devname, &i, keep_fd); | |
5013 | else | |
ec50f7b6 LM |
5014 | /* get super block from devlist devices */ |
5015 | err = get_devlist_super_block(devlist, &super_list, &i, keep_fd); | |
9587c373 | 5016 | if (err) |
1602d52c | 5017 | goto error; |
a2b97981 DW |
5018 | /* all mpbs enter, maybe one leaves */ |
5019 | super = imsm_thunderdome(&super_list, i); | |
5020 | if (!super) { | |
5021 | err = 1; | |
5022 | goto error; | |
cdddbdbc DW |
5023 | } |
5024 | ||
47ee5a45 DW |
5025 | if (find_missing(super) != 0) { |
5026 | free_imsm(super); | |
a2b97981 DW |
5027 | err = 2; |
5028 | goto error; | |
47ee5a45 | 5029 | } |
8e59f3d8 AK |
5030 | |
5031 | /* load migration record */ | |
5032 | err = load_imsm_migr_rec(super, NULL); | |
4c965cc9 AK |
5033 | if (err == -1) { |
5034 | /* migration is in progress, | |
5035 | * but migr_rec cannot be loaded, | |
5036 | */ | |
8e59f3d8 AK |
5037 | err = 4; |
5038 | goto error; | |
5039 | } | |
e2f41b2c AK |
5040 | |
5041 | /* Check migration compatibility */ | |
089f9d79 | 5042 | if (err == 0 && check_mpb_migr_compatibility(super) != 0) { |
e7b84f9d | 5043 | pr_err("Unsupported migration detected"); |
e2f41b2c AK |
5044 | if (devname) |
5045 | fprintf(stderr, " on %s\n", devname); | |
5046 | else | |
5047 | fprintf(stderr, " (IMSM).\n"); | |
5048 | ||
5049 | err = 5; | |
5050 | goto error; | |
5051 | } | |
5052 | ||
a2b97981 DW |
5053 | err = 0; |
5054 | ||
5055 | error: | |
5056 | while (super_list) { | |
5057 | struct intel_super *s = super_list; | |
5058 | ||
5059 | super_list = super_list->next; | |
5060 | free_imsm(s); | |
5061 | } | |
9587c373 | 5062 | |
a2b97981 DW |
5063 | if (err) |
5064 | return err; | |
f7e7067b | 5065 | |
cdddbdbc | 5066 | *sbp = super; |
9587c373 | 5067 | if (fd >= 0) |
4dd2df09 | 5068 | strcpy(st->container_devnm, fd2devnm(fd)); |
9587c373 | 5069 | else |
4dd2df09 | 5070 | st->container_devnm[0] = 0; |
a2b97981 | 5071 | if (err == 0 && st->ss == NULL) { |
bf5a934a | 5072 | st->ss = &super_imsm; |
cdddbdbc DW |
5073 | st->minor_version = 0; |
5074 | st->max_devs = IMSM_MAX_DEVICES; | |
5075 | } | |
cdddbdbc DW |
5076 | return 0; |
5077 | } | |
2b959fbf | 5078 | |
ec50f7b6 LM |
5079 | static int |
5080 | get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list, | |
5081 | int *max, int keep_fd) | |
5082 | { | |
5083 | struct md_list *tmpdev; | |
5084 | int err = 0; | |
5085 | int i = 0; | |
9587c373 | 5086 | |
ec50f7b6 LM |
5087 | for (i = 0, tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { |
5088 | if (tmpdev->used != 1) | |
5089 | continue; | |
5090 | if (tmpdev->container == 1) { | |
ca9de185 | 5091 | int lmax = 0; |
ec50f7b6 LM |
5092 | int fd = dev_open(tmpdev->devname, O_RDONLY|O_EXCL); |
5093 | if (fd < 0) { | |
e7b84f9d | 5094 | pr_err("cannot open device %s: %s\n", |
ec50f7b6 LM |
5095 | tmpdev->devname, strerror(errno)); |
5096 | err = 8; | |
5097 | goto error; | |
5098 | } | |
5099 | err = get_sra_super_block(fd, super_list, | |
5100 | tmpdev->devname, &lmax, | |
5101 | keep_fd); | |
5102 | i += lmax; | |
5103 | close(fd); | |
5104 | if (err) { | |
5105 | err = 7; | |
5106 | goto error; | |
5107 | } | |
5108 | } else { | |
5109 | int major = major(tmpdev->st_rdev); | |
5110 | int minor = minor(tmpdev->st_rdev); | |
5111 | err = get_super_block(super_list, | |
4dd2df09 | 5112 | NULL, |
ec50f7b6 LM |
5113 | tmpdev->devname, |
5114 | major, minor, | |
5115 | keep_fd); | |
5116 | i++; | |
5117 | if (err) { | |
5118 | err = 6; | |
5119 | goto error; | |
5120 | } | |
5121 | } | |
5122 | } | |
5123 | error: | |
5124 | *max = i; | |
5125 | return err; | |
5126 | } | |
9587c373 | 5127 | |
4dd2df09 | 5128 | static int get_super_block(struct intel_super **super_list, char *devnm, char *devname, |
9587c373 LM |
5129 | int major, int minor, int keep_fd) |
5130 | { | |
594dc1b8 | 5131 | struct intel_super *s; |
9587c373 LM |
5132 | char nm[32]; |
5133 | int dfd = -1; | |
9587c373 LM |
5134 | int err = 0; |
5135 | int retry; | |
5136 | ||
5137 | s = alloc_super(); | |
5138 | if (!s) { | |
5139 | err = 1; | |
5140 | goto error; | |
5141 | } | |
5142 | ||
5143 | sprintf(nm, "%d:%d", major, minor); | |
5144 | dfd = dev_open(nm, O_RDWR); | |
5145 | if (dfd < 0) { | |
5146 | err = 2; | |
5147 | goto error; | |
5148 | } | |
5149 | ||
fa7bb6f8 | 5150 | get_dev_sector_size(dfd, NULL, &s->sector_size); |
cb8f6859 | 5151 | find_intel_hba_capability(dfd, s, devname); |
9587c373 LM |
5152 | err = load_and_parse_mpb(dfd, s, NULL, keep_fd); |
5153 | ||
5154 | /* retry the load if we might have raced against mdmon */ | |
4dd2df09 | 5155 | if (err == 3 && devnm && mdmon_running(devnm)) |
9587c373 LM |
5156 | for (retry = 0; retry < 3; retry++) { |
5157 | usleep(3000); | |
5158 | err = load_and_parse_mpb(dfd, s, NULL, keep_fd); | |
5159 | if (err != 3) | |
5160 | break; | |
5161 | } | |
5162 | error: | |
5163 | if (!err) { | |
5164 | s->next = *super_list; | |
5165 | *super_list = s; | |
5166 | } else { | |
5167 | if (s) | |
8d67477f | 5168 | free_imsm(s); |
36614e95 | 5169 | if (dfd >= 0) |
9587c373 LM |
5170 | close(dfd); |
5171 | } | |
089f9d79 | 5172 | if (dfd >= 0 && !keep_fd) |
9587c373 LM |
5173 | close(dfd); |
5174 | return err; | |
5175 | ||
5176 | } | |
5177 | ||
5178 | static int | |
5179 | get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd) | |
5180 | { | |
5181 | struct mdinfo *sra; | |
4dd2df09 | 5182 | char *devnm; |
9587c373 LM |
5183 | struct mdinfo *sd; |
5184 | int err = 0; | |
5185 | int i = 0; | |
4dd2df09 | 5186 | sra = sysfs_read(fd, NULL, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE); |
9587c373 LM |
5187 | if (!sra) |
5188 | return 1; | |
5189 | ||
5190 | if (sra->array.major_version != -1 || | |
5191 | sra->array.minor_version != -2 || | |
5192 | strcmp(sra->text_version, "imsm") != 0) { | |
5193 | err = 1; | |
5194 | goto error; | |
5195 | } | |
5196 | /* load all mpbs */ | |
4dd2df09 | 5197 | devnm = fd2devnm(fd); |
9587c373 | 5198 | for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) { |
4dd2df09 | 5199 | if (get_super_block(super_list, devnm, devname, |
9587c373 LM |
5200 | sd->disk.major, sd->disk.minor, keep_fd) != 0) { |
5201 | err = 7; | |
5202 | goto error; | |
5203 | } | |
5204 | } | |
5205 | error: | |
5206 | sysfs_free(sra); | |
5207 | *max = i; | |
5208 | return err; | |
5209 | } | |
5210 | ||
2b959fbf N |
5211 | static int load_container_imsm(struct supertype *st, int fd, char *devname) |
5212 | { | |
ec50f7b6 | 5213 | return load_super_imsm_all(st, fd, &st->sb, devname, NULL, 1); |
2b959fbf | 5214 | } |
cdddbdbc DW |
5215 | |
5216 | static int load_super_imsm(struct supertype *st, int fd, char *devname) | |
5217 | { | |
5218 | struct intel_super *super; | |
5219 | int rv; | |
8a3544f8 | 5220 | int retry; |
cdddbdbc | 5221 | |
357ac106 | 5222 | if (test_partition(fd)) |
691c6ee1 N |
5223 | /* IMSM not allowed on partitions */ |
5224 | return 1; | |
5225 | ||
37424f13 DW |
5226 | free_super_imsm(st); |
5227 | ||
49133e57 | 5228 | super = alloc_super(); |
fa7bb6f8 | 5229 | get_dev_sector_size(fd, NULL, &super->sector_size); |
8d67477f TM |
5230 | if (!super) |
5231 | return 1; | |
ea2bc72b LM |
5232 | /* Load hba and capabilities if they exist. |
5233 | * But do not preclude loading metadata in case capabilities or hba are | |
5234 | * non-compliant and ignore_hw_compat is set. | |
5235 | */ | |
d424212e | 5236 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 5237 | /* no orom/efi or non-intel hba of the disk */ |
089f9d79 | 5238 | if (rv != 0 && st->ignore_hw_compat == 0) { |
f2f5c343 | 5239 | if (devname) |
e7b84f9d | 5240 | pr_err("No OROM/EFI properties for %s\n", devname); |
f2f5c343 LM |
5241 | free_imsm(super); |
5242 | return 2; | |
5243 | } | |
a2b97981 | 5244 | rv = load_and_parse_mpb(fd, super, devname, 0); |
cdddbdbc | 5245 | |
8a3544f8 AP |
5246 | /* retry the load if we might have raced against mdmon */ |
5247 | if (rv == 3) { | |
f96b1302 AP |
5248 | struct mdstat_ent *mdstat = NULL; |
5249 | char *name = fd2kname(fd); | |
5250 | ||
5251 | if (name) | |
5252 | mdstat = mdstat_by_component(name); | |
8a3544f8 AP |
5253 | |
5254 | if (mdstat && mdmon_running(mdstat->devnm) && getpid() != mdmon_pid(mdstat->devnm)) { | |
5255 | for (retry = 0; retry < 3; retry++) { | |
5256 | usleep(3000); | |
5257 | rv = load_and_parse_mpb(fd, super, devname, 0); | |
5258 | if (rv != 3) | |
5259 | break; | |
5260 | } | |
5261 | } | |
5262 | ||
5263 | free_mdstat(mdstat); | |
5264 | } | |
5265 | ||
cdddbdbc DW |
5266 | if (rv) { |
5267 | if (devname) | |
7a862a02 | 5268 | pr_err("Failed to load all information sections on %s\n", devname); |
cdddbdbc DW |
5269 | free_imsm(super); |
5270 | return rv; | |
5271 | } | |
5272 | ||
5273 | st->sb = super; | |
5274 | if (st->ss == NULL) { | |
5275 | st->ss = &super_imsm; | |
5276 | st->minor_version = 0; | |
5277 | st->max_devs = IMSM_MAX_DEVICES; | |
5278 | } | |
8e59f3d8 AK |
5279 | |
5280 | /* load migration record */ | |
2e062e82 AK |
5281 | if (load_imsm_migr_rec(super, NULL) == 0) { |
5282 | /* Check for unsupported migration features */ | |
5283 | if (check_mpb_migr_compatibility(super) != 0) { | |
e7b84f9d | 5284 | pr_err("Unsupported migration detected"); |
2e062e82 AK |
5285 | if (devname) |
5286 | fprintf(stderr, " on %s\n", devname); | |
5287 | else | |
5288 | fprintf(stderr, " (IMSM).\n"); | |
5289 | return 3; | |
5290 | } | |
e2f41b2c AK |
5291 | } |
5292 | ||
cdddbdbc DW |
5293 | return 0; |
5294 | } | |
5295 | ||
ef6ffade DW |
5296 | static __u16 info_to_blocks_per_strip(mdu_array_info_t *info) |
5297 | { | |
5298 | if (info->level == 1) | |
5299 | return 128; | |
5300 | return info->chunk_size >> 9; | |
5301 | } | |
5302 | ||
5551b113 CA |
5303 | static unsigned long long info_to_blocks_per_member(mdu_array_info_t *info, |
5304 | unsigned long long size) | |
fcfd9599 | 5305 | { |
4025c288 | 5306 | if (info->level == 1) |
5551b113 | 5307 | return size * 2; |
4025c288 | 5308 | else |
5551b113 | 5309 | return (size * 2) & ~(info_to_blocks_per_strip(info) - 1); |
fcfd9599 DW |
5310 | } |
5311 | ||
4d1313e9 DW |
5312 | static void imsm_update_version_info(struct intel_super *super) |
5313 | { | |
5314 | /* update the version and attributes */ | |
5315 | struct imsm_super *mpb = super->anchor; | |
5316 | char *version; | |
5317 | struct imsm_dev *dev; | |
5318 | struct imsm_map *map; | |
5319 | int i; | |
5320 | ||
5321 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
5322 | dev = get_imsm_dev(super, i); | |
238c0a71 | 5323 | map = get_imsm_map(dev, MAP_0); |
4d1313e9 DW |
5324 | if (__le32_to_cpu(dev->size_high) > 0) |
5325 | mpb->attributes |= MPB_ATTRIB_2TB; | |
5326 | ||
5327 | /* FIXME detect when an array spans a port multiplier */ | |
5328 | #if 0 | |
5329 | mpb->attributes |= MPB_ATTRIB_PM; | |
5330 | #endif | |
5331 | ||
5332 | if (mpb->num_raid_devs > 1 || | |
5333 | mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) { | |
5334 | version = MPB_VERSION_ATTRIBS; | |
5335 | switch (get_imsm_raid_level(map)) { | |
5336 | case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break; | |
5337 | case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break; | |
5338 | case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break; | |
5339 | case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break; | |
5340 | } | |
5341 | } else { | |
5342 | if (map->num_members >= 5) | |
5343 | version = MPB_VERSION_5OR6_DISK_ARRAY; | |
5344 | else if (dev->status == DEV_CLONE_N_GO) | |
5345 | version = MPB_VERSION_CNG; | |
5346 | else if (get_imsm_raid_level(map) == 5) | |
5347 | version = MPB_VERSION_RAID5; | |
5348 | else if (map->num_members >= 3) | |
5349 | version = MPB_VERSION_3OR4_DISK_ARRAY; | |
5350 | else if (get_imsm_raid_level(map) == 1) | |
5351 | version = MPB_VERSION_RAID1; | |
5352 | else | |
5353 | version = MPB_VERSION_RAID0; | |
5354 | } | |
5355 | strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version); | |
5356 | } | |
5357 | } | |
5358 | ||
aa534678 DW |
5359 | static int check_name(struct intel_super *super, char *name, int quiet) |
5360 | { | |
5361 | struct imsm_super *mpb = super->anchor; | |
5362 | char *reason = NULL; | |
9bd99a90 RS |
5363 | char *start = name; |
5364 | size_t len = strlen(name); | |
aa534678 DW |
5365 | int i; |
5366 | ||
9bd99a90 RS |
5367 | if (len > 0) { |
5368 | while (isspace(start[len - 1])) | |
5369 | start[--len] = 0; | |
5370 | while (*start && isspace(*start)) | |
5371 | ++start, --len; | |
5372 | memmove(name, start, len + 1); | |
5373 | } | |
5374 | ||
5375 | if (len > MAX_RAID_SERIAL_LEN) | |
aa534678 | 5376 | reason = "must be 16 characters or less"; |
9bd99a90 RS |
5377 | else if (len == 0) |
5378 | reason = "must be a non-empty string"; | |
aa534678 DW |
5379 | |
5380 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
5381 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
5382 | ||
5383 | if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) { | |
5384 | reason = "already exists"; | |
5385 | break; | |
5386 | } | |
5387 | } | |
5388 | ||
5389 | if (reason && !quiet) | |
e7b84f9d | 5390 | pr_err("imsm volume name %s\n", reason); |
aa534678 DW |
5391 | |
5392 | return !reason; | |
5393 | } | |
5394 | ||
8b353278 | 5395 | static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info, |
5308f117 | 5396 | struct shape *s, char *name, |
83cd1e97 N |
5397 | char *homehost, int *uuid, |
5398 | long long data_offset) | |
cdddbdbc | 5399 | { |
c2c087e6 DW |
5400 | /* We are creating a volume inside a pre-existing container. |
5401 | * so st->sb is already set. | |
5402 | */ | |
5403 | struct intel_super *super = st->sb; | |
f36a9ecd | 5404 | unsigned int sector_size = super->sector_size; |
949c47a0 | 5405 | struct imsm_super *mpb = super->anchor; |
ba2de7ba | 5406 | struct intel_dev *dv; |
c2c087e6 DW |
5407 | struct imsm_dev *dev; |
5408 | struct imsm_vol *vol; | |
5409 | struct imsm_map *map; | |
5410 | int idx = mpb->num_raid_devs; | |
5411 | int i; | |
760365f9 | 5412 | int namelen; |
c2c087e6 | 5413 | unsigned long long array_blocks; |
2c092cad | 5414 | size_t size_old, size_new; |
5551b113 | 5415 | unsigned long long num_data_stripes; |
b53bfba6 TM |
5416 | unsigned int data_disks; |
5417 | unsigned long long size_per_member; | |
cdddbdbc | 5418 | |
88c32bb1 | 5419 | if (super->orom && mpb->num_raid_devs >= super->orom->vpa) { |
7a862a02 | 5420 | pr_err("This imsm-container already has the maximum of %d volumes\n", super->orom->vpa); |
c2c087e6 DW |
5421 | return 0; |
5422 | } | |
5423 | ||
2c092cad DW |
5424 | /* ensure the mpb is large enough for the new data */ |
5425 | size_old = __le32_to_cpu(mpb->mpb_size); | |
5426 | size_new = disks_to_mpb_size(info->nr_disks); | |
5427 | if (size_new > size_old) { | |
5428 | void *mpb_new; | |
f36a9ecd | 5429 | size_t size_round = ROUND_UP(size_new, sector_size); |
2c092cad | 5430 | |
f36a9ecd | 5431 | if (posix_memalign(&mpb_new, sector_size, size_round) != 0) { |
e7b84f9d | 5432 | pr_err("could not allocate new mpb\n"); |
2c092cad DW |
5433 | return 0; |
5434 | } | |
85337573 AO |
5435 | if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE, |
5436 | MIGR_REC_BUF_SECTORS* | |
5437 | MAX_SECTOR_SIZE) != 0) { | |
1ade5cc1 | 5438 | pr_err("could not allocate migr_rec buffer\n"); |
8e59f3d8 AK |
5439 | free(super->buf); |
5440 | free(super); | |
ea944c8f | 5441 | free(mpb_new); |
8e59f3d8 AK |
5442 | return 0; |
5443 | } | |
2c092cad DW |
5444 | memcpy(mpb_new, mpb, size_old); |
5445 | free(mpb); | |
5446 | mpb = mpb_new; | |
949c47a0 | 5447 | super->anchor = mpb_new; |
2c092cad DW |
5448 | mpb->mpb_size = __cpu_to_le32(size_new); |
5449 | memset(mpb_new + size_old, 0, size_round - size_old); | |
bbab0940 | 5450 | super->len = size_round; |
2c092cad | 5451 | } |
bf5a934a | 5452 | super->current_vol = idx; |
3960e579 DW |
5453 | |
5454 | /* handle 'failed_disks' by either: | |
5455 | * a) create dummy disk entries in the table if this the first | |
5456 | * volume in the array. We add them here as this is the only | |
5457 | * opportunity to add them. add_to_super_imsm_volume() | |
5458 | * handles the non-failed disks and continues incrementing | |
5459 | * mpb->num_disks. | |
5460 | * b) validate that 'failed_disks' matches the current number | |
5461 | * of missing disks if the container is populated | |
d23fe947 | 5462 | */ |
3960e579 | 5463 | if (super->current_vol == 0) { |
d23fe947 | 5464 | mpb->num_disks = 0; |
3960e579 DW |
5465 | for (i = 0; i < info->failed_disks; i++) { |
5466 | struct imsm_disk *disk; | |
5467 | ||
5468 | mpb->num_disks++; | |
5469 | disk = __get_imsm_disk(mpb, i); | |
5470 | disk->status = CONFIGURED_DISK | FAILED_DISK; | |
5471 | disk->scsi_id = __cpu_to_le32(~(__u32)0); | |
5472 | snprintf((char *) disk->serial, MAX_RAID_SERIAL_LEN, | |
5b975129 | 5473 | "missing:%d", (__u8)i); |
3960e579 DW |
5474 | } |
5475 | find_missing(super); | |
5476 | } else { | |
5477 | int missing = 0; | |
5478 | struct dl *d; | |
5479 | ||
5480 | for (d = super->missing; d; d = d->next) | |
5481 | missing++; | |
5482 | if (info->failed_disks > missing) { | |
e7b84f9d | 5483 | pr_err("unable to add 'missing' disk to container\n"); |
3960e579 DW |
5484 | return 0; |
5485 | } | |
5486 | } | |
5a038140 | 5487 | |
aa534678 DW |
5488 | if (!check_name(super, name, 0)) |
5489 | return 0; | |
503975b9 N |
5490 | dv = xmalloc(sizeof(*dv)); |
5491 | dev = xcalloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1)); | |
760365f9 JS |
5492 | /* |
5493 | * Explicitly allow truncating to not confuse gcc's | |
5494 | * -Werror=stringop-truncation | |
5495 | */ | |
5496 | namelen = min((int) strlen(name), MAX_RAID_SERIAL_LEN); | |
5497 | memcpy(dev->volume, name, namelen); | |
e03640bd | 5498 | array_blocks = calc_array_size(info->level, info->raid_disks, |
03bcbc65 | 5499 | info->layout, info->chunk_size, |
b53bfba6 TM |
5500 | s->size * BLOCKS_PER_KB); |
5501 | data_disks = get_data_disks(info->level, info->layout, | |
5502 | info->raid_disks); | |
5503 | array_blocks = round_size_to_mb(array_blocks, data_disks); | |
5504 | size_per_member = array_blocks / data_disks; | |
979d38be | 5505 | |
fcc2c9da | 5506 | set_imsm_dev_size(dev, array_blocks); |
1a2487c2 | 5507 | dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING); |
c2c087e6 DW |
5508 | vol = &dev->vol; |
5509 | vol->migr_state = 0; | |
1484e727 | 5510 | set_migr_type(dev, MIGR_INIT); |
3960e579 | 5511 | vol->dirty = !info->state; |
f8f603f1 | 5512 | vol->curr_migr_unit = 0; |
238c0a71 | 5513 | map = get_imsm_map(dev, MAP_0); |
5551b113 | 5514 | set_pba_of_lba0(map, super->create_offset); |
ef6ffade | 5515 | map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info)); |
0556e1a2 | 5516 | map->failed_disk_num = ~0; |
bf4442ab | 5517 | if (info->level > 0) |
fffaf1ff N |
5518 | map->map_state = (info->state ? IMSM_T_STATE_NORMAL |
5519 | : IMSM_T_STATE_UNINITIALIZED); | |
bf4442ab AK |
5520 | else |
5521 | map->map_state = info->failed_disks ? IMSM_T_STATE_FAILED : | |
5522 | IMSM_T_STATE_NORMAL; | |
252d23c0 | 5523 | map->ddf = 1; |
ef6ffade DW |
5524 | |
5525 | if (info->level == 1 && info->raid_disks > 2) { | |
38950822 AW |
5526 | free(dev); |
5527 | free(dv); | |
7a862a02 | 5528 | pr_err("imsm does not support more than 2 disksin a raid1 volume\n"); |
ef6ffade DW |
5529 | return 0; |
5530 | } | |
81062a36 DW |
5531 | |
5532 | map->raid_level = info->level; | |
4d1313e9 | 5533 | if (info->level == 10) { |
c2c087e6 | 5534 | map->raid_level = 1; |
4d1313e9 | 5535 | map->num_domains = info->raid_disks / 2; |
81062a36 DW |
5536 | } else if (info->level == 1) |
5537 | map->num_domains = info->raid_disks; | |
5538 | else | |
ff596308 | 5539 | map->num_domains = 1; |
81062a36 | 5540 | |
5551b113 | 5541 | /* info->size is only int so use the 'size' parameter instead */ |
b53bfba6 | 5542 | num_data_stripes = size_per_member / info_to_blocks_per_strip(info); |
5551b113 CA |
5543 | num_data_stripes /= map->num_domains; |
5544 | set_num_data_stripes(map, num_data_stripes); | |
ef6ffade | 5545 | |
44490938 MD |
5546 | size_per_member += NUM_BLOCKS_DIRTY_STRIPE_REGION; |
5547 | set_blocks_per_member(map, info_to_blocks_per_member(info, | |
5548 | size_per_member / | |
5549 | BLOCKS_PER_KB)); | |
5550 | ||
c2c087e6 DW |
5551 | map->num_members = info->raid_disks; |
5552 | for (i = 0; i < map->num_members; i++) { | |
5553 | /* initialized in add_to_super */ | |
4eb26970 | 5554 | set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD); |
c2c087e6 | 5555 | } |
949c47a0 | 5556 | mpb->num_raid_devs++; |
2a24dc1b PB |
5557 | mpb->num_raid_devs_created++; |
5558 | dev->my_vol_raid_dev_num = mpb->num_raid_devs_created; | |
ba2de7ba | 5559 | |
b7580566 | 5560 | if (s->consistency_policy <= CONSISTENCY_POLICY_RESYNC) { |
c2462068 | 5561 | dev->rwh_policy = RWH_MULTIPLE_OFF; |
2432ce9b | 5562 | } else if (s->consistency_policy == CONSISTENCY_POLICY_PPL) { |
c2462068 | 5563 | dev->rwh_policy = RWH_MULTIPLE_DISTRIBUTED; |
2432ce9b AP |
5564 | } else { |
5565 | free(dev); | |
5566 | free(dv); | |
5567 | pr_err("imsm does not support consistency policy %s\n", | |
5568 | map_num(consistency_policies, s->consistency_policy)); | |
5569 | return 0; | |
5570 | } | |
5571 | ||
ba2de7ba DW |
5572 | dv->dev = dev; |
5573 | dv->index = super->current_vol; | |
5574 | dv->next = super->devlist; | |
5575 | super->devlist = dv; | |
c2c087e6 | 5576 | |
4d1313e9 DW |
5577 | imsm_update_version_info(super); |
5578 | ||
c2c087e6 | 5579 | return 1; |
cdddbdbc DW |
5580 | } |
5581 | ||
bf5a934a | 5582 | static int init_super_imsm(struct supertype *st, mdu_array_info_t *info, |
5308f117 | 5583 | struct shape *s, char *name, |
83cd1e97 N |
5584 | char *homehost, int *uuid, |
5585 | unsigned long long data_offset) | |
bf5a934a DW |
5586 | { |
5587 | /* This is primarily called by Create when creating a new array. | |
5588 | * We will then get add_to_super called for each component, and then | |
5589 | * write_init_super called to write it out to each device. | |
5590 | * For IMSM, Create can create on fresh devices or on a pre-existing | |
5591 | * array. | |
5592 | * To create on a pre-existing array a different method will be called. | |
5593 | * This one is just for fresh drives. | |
5594 | */ | |
5595 | struct intel_super *super; | |
5596 | struct imsm_super *mpb; | |
5597 | size_t mpb_size; | |
4d1313e9 | 5598 | char *version; |
bf5a934a | 5599 | |
83cd1e97 | 5600 | if (data_offset != INVALID_SECTORS) { |
ed503f89 | 5601 | pr_err("data-offset not supported by imsm\n"); |
83cd1e97 N |
5602 | return 0; |
5603 | } | |
5604 | ||
bf5a934a | 5605 | if (st->sb) |
5308f117 | 5606 | return init_super_imsm_volume(st, info, s, name, homehost, uuid, |
83cd1e97 | 5607 | data_offset); |
e683ca88 DW |
5608 | |
5609 | if (info) | |
5610 | mpb_size = disks_to_mpb_size(info->nr_disks); | |
5611 | else | |
f36a9ecd | 5612 | mpb_size = MAX_SECTOR_SIZE; |
bf5a934a | 5613 | |
49133e57 | 5614 | super = alloc_super(); |
f36a9ecd PB |
5615 | if (super && |
5616 | posix_memalign(&super->buf, MAX_SECTOR_SIZE, mpb_size) != 0) { | |
8d67477f | 5617 | free_imsm(super); |
e683ca88 DW |
5618 | super = NULL; |
5619 | } | |
5620 | if (!super) { | |
1ade5cc1 | 5621 | pr_err("could not allocate superblock\n"); |
bf5a934a DW |
5622 | return 0; |
5623 | } | |
de44e46f PB |
5624 | if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE, |
5625 | MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) { | |
1ade5cc1 | 5626 | pr_err("could not allocate migr_rec buffer\n"); |
8e59f3d8 | 5627 | free(super->buf); |
8d67477f | 5628 | free_imsm(super); |
8e59f3d8 AK |
5629 | return 0; |
5630 | } | |
e683ca88 | 5631 | memset(super->buf, 0, mpb_size); |
ef649044 | 5632 | mpb = super->buf; |
e683ca88 DW |
5633 | mpb->mpb_size = __cpu_to_le32(mpb_size); |
5634 | st->sb = super; | |
5635 | ||
5636 | if (info == NULL) { | |
5637 | /* zeroing superblock */ | |
5638 | return 0; | |
5639 | } | |
bf5a934a | 5640 | |
4d1313e9 DW |
5641 | mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; |
5642 | ||
5643 | version = (char *) mpb->sig; | |
5644 | strcpy(version, MPB_SIGNATURE); | |
5645 | version += strlen(MPB_SIGNATURE); | |
5646 | strcpy(version, MPB_VERSION_RAID0); | |
bf5a934a | 5647 | |
bf5a934a DW |
5648 | return 1; |
5649 | } | |
5650 | ||
f2cc4f7d AO |
5651 | static int drive_validate_sector_size(struct intel_super *super, struct dl *dl) |
5652 | { | |
5653 | unsigned int member_sector_size; | |
5654 | ||
5655 | if (dl->fd < 0) { | |
5656 | pr_err("Invalid file descriptor for %s\n", dl->devname); | |
5657 | return 0; | |
5658 | } | |
5659 | ||
5660 | if (!get_dev_sector_size(dl->fd, dl->devname, &member_sector_size)) | |
5661 | return 0; | |
5662 | if (member_sector_size != super->sector_size) | |
5663 | return 0; | |
5664 | return 1; | |
5665 | } | |
5666 | ||
f20c3968 | 5667 | static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk, |
bf5a934a DW |
5668 | int fd, char *devname) |
5669 | { | |
5670 | struct intel_super *super = st->sb; | |
d23fe947 | 5671 | struct imsm_super *mpb = super->anchor; |
3960e579 | 5672 | struct imsm_disk *_disk; |
bf5a934a DW |
5673 | struct imsm_dev *dev; |
5674 | struct imsm_map *map; | |
3960e579 | 5675 | struct dl *dl, *df; |
4eb26970 | 5676 | int slot; |
bf5a934a | 5677 | |
949c47a0 | 5678 | dev = get_imsm_dev(super, super->current_vol); |
238c0a71 | 5679 | map = get_imsm_map(dev, MAP_0); |
bf5a934a | 5680 | |
208933a7 | 5681 | if (! (dk->state & (1<<MD_DISK_SYNC))) { |
e7b84f9d | 5682 | pr_err("%s: Cannot add spare devices to IMSM volume\n", |
208933a7 N |
5683 | devname); |
5684 | return 1; | |
5685 | } | |
5686 | ||
efb30e7f DW |
5687 | if (fd == -1) { |
5688 | /* we're doing autolayout so grab the pre-marked (in | |
5689 | * validate_geometry) raid_disk | |
5690 | */ | |
5691 | for (dl = super->disks; dl; dl = dl->next) | |
5692 | if (dl->raiddisk == dk->raid_disk) | |
5693 | break; | |
5694 | } else { | |
5695 | for (dl = super->disks; dl ; dl = dl->next) | |
5696 | if (dl->major == dk->major && | |
5697 | dl->minor == dk->minor) | |
5698 | break; | |
5699 | } | |
d23fe947 | 5700 | |
208933a7 | 5701 | if (!dl) { |
e7b84f9d | 5702 | pr_err("%s is not a member of the same container\n", devname); |
f20c3968 | 5703 | return 1; |
208933a7 | 5704 | } |
bf5a934a | 5705 | |
59632db9 MZ |
5706 | if (mpb->num_disks == 0) |
5707 | if (!get_dev_sector_size(dl->fd, dl->devname, | |
5708 | &super->sector_size)) | |
5709 | return 1; | |
5710 | ||
f2cc4f7d AO |
5711 | if (!drive_validate_sector_size(super, dl)) { |
5712 | pr_err("Combining drives of different sector size in one volume is not allowed\n"); | |
5713 | return 1; | |
5714 | } | |
5715 | ||
d23fe947 DW |
5716 | /* add a pristine spare to the metadata */ |
5717 | if (dl->index < 0) { | |
5718 | dl->index = super->anchor->num_disks; | |
5719 | super->anchor->num_disks++; | |
5720 | } | |
4eb26970 DW |
5721 | /* Check the device has not already been added */ |
5722 | slot = get_imsm_disk_slot(map, dl->index); | |
5723 | if (slot >= 0 && | |
238c0a71 | 5724 | (get_imsm_ord_tbl_ent(dev, slot, MAP_X) & IMSM_ORD_REBUILD) == 0) { |
e7b84f9d | 5725 | pr_err("%s has been included in this array twice\n", |
4eb26970 DW |
5726 | devname); |
5727 | return 1; | |
5728 | } | |
656b6b5a | 5729 | set_imsm_ord_tbl_ent(map, dk->raid_disk, dl->index); |
ee5aad5a | 5730 | dl->disk.status = CONFIGURED_DISK; |
d23fe947 | 5731 | |
3960e579 DW |
5732 | /* update size of 'missing' disks to be at least as large as the |
5733 | * largest acitve member (we only have dummy missing disks when | |
5734 | * creating the first volume) | |
5735 | */ | |
5736 | if (super->current_vol == 0) { | |
5737 | for (df = super->missing; df; df = df->next) { | |
5551b113 CA |
5738 | if (total_blocks(&dl->disk) > total_blocks(&df->disk)) |
5739 | set_total_blocks(&df->disk, total_blocks(&dl->disk)); | |
3960e579 DW |
5740 | _disk = __get_imsm_disk(mpb, df->index); |
5741 | *_disk = df->disk; | |
5742 | } | |
5743 | } | |
5744 | ||
5745 | /* refresh unset/failed slots to point to valid 'missing' entries */ | |
5746 | for (df = super->missing; df; df = df->next) | |
5747 | for (slot = 0; slot < mpb->num_disks; slot++) { | |
238c0a71 | 5748 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); |
3960e579 DW |
5749 | |
5750 | if ((ord & IMSM_ORD_REBUILD) == 0) | |
5751 | continue; | |
5752 | set_imsm_ord_tbl_ent(map, slot, df->index | IMSM_ORD_REBUILD); | |
1ace8403 | 5753 | if (is_gen_migration(dev)) { |
238c0a71 AK |
5754 | struct imsm_map *map2 = get_imsm_map(dev, |
5755 | MAP_1); | |
0a108d63 | 5756 | int slot2 = get_imsm_disk_slot(map2, df->index); |
089f9d79 | 5757 | if (slot2 < map2->num_members && slot2 >= 0) { |
1ace8403 | 5758 | __u32 ord2 = get_imsm_ord_tbl_ent(dev, |
238c0a71 AK |
5759 | slot2, |
5760 | MAP_1); | |
1ace8403 AK |
5761 | if ((unsigned)df->index == |
5762 | ord_to_idx(ord2)) | |
5763 | set_imsm_ord_tbl_ent(map2, | |
0a108d63 | 5764 | slot2, |
1ace8403 AK |
5765 | df->index | |
5766 | IMSM_ORD_REBUILD); | |
5767 | } | |
5768 | } | |
3960e579 DW |
5769 | dprintf("set slot:%d to missing disk:%d\n", slot, df->index); |
5770 | break; | |
5771 | } | |
5772 | ||
d23fe947 DW |
5773 | /* if we are creating the first raid device update the family number */ |
5774 | if (super->current_vol == 0) { | |
5775 | __u32 sum; | |
5776 | struct imsm_dev *_dev = __get_imsm_dev(mpb, 0); | |
d23fe947 | 5777 | |
3960e579 | 5778 | _disk = __get_imsm_disk(mpb, dl->index); |
791b666a | 5779 | if (!_dev || !_disk) { |
e7b84f9d | 5780 | pr_err("BUG mpb setup error\n"); |
791b666a AW |
5781 | return 1; |
5782 | } | |
d23fe947 DW |
5783 | *_dev = *dev; |
5784 | *_disk = dl->disk; | |
148acb7b DW |
5785 | sum = random32(); |
5786 | sum += __gen_imsm_checksum(mpb); | |
d23fe947 | 5787 | mpb->family_num = __cpu_to_le32(sum); |
148acb7b | 5788 | mpb->orig_family_num = mpb->family_num; |
d23fe947 | 5789 | } |
ca0748fa | 5790 | super->current_disk = dl; |
f20c3968 | 5791 | return 0; |
bf5a934a DW |
5792 | } |
5793 | ||
a8619d23 AK |
5794 | /* mark_spare() |
5795 | * Function marks disk as spare and restores disk serial | |
5796 | * in case it was previously marked as failed by takeover operation | |
5797 | * reruns: | |
5798 | * -1 : critical error | |
5799 | * 0 : disk is marked as spare but serial is not set | |
5800 | * 1 : success | |
5801 | */ | |
5802 | int mark_spare(struct dl *disk) | |
5803 | { | |
5804 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
5805 | int ret_val = -1; | |
5806 | ||
5807 | if (!disk) | |
5808 | return ret_val; | |
5809 | ||
5810 | ret_val = 0; | |
5811 | if (!imsm_read_serial(disk->fd, NULL, serial)) { | |
5812 | /* Restore disk serial number, because takeover marks disk | |
5813 | * as failed and adds to serial ':0' before it becomes | |
5814 | * a spare disk. | |
5815 | */ | |
5816 | serialcpy(disk->serial, serial); | |
5817 | serialcpy(disk->disk.serial, serial); | |
5818 | ret_val = 1; | |
5819 | } | |
5820 | disk->disk.status = SPARE_DISK; | |
5821 | disk->index = -1; | |
5822 | ||
5823 | return ret_val; | |
5824 | } | |
88654014 | 5825 | |
f20c3968 | 5826 | static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk, |
72ca9bcf N |
5827 | int fd, char *devname, |
5828 | unsigned long long data_offset) | |
cdddbdbc | 5829 | { |
c2c087e6 | 5830 | struct intel_super *super = st->sb; |
c2c087e6 DW |
5831 | struct dl *dd; |
5832 | unsigned long long size; | |
fa7bb6f8 | 5833 | unsigned int member_sector_size; |
f2f27e63 | 5834 | __u32 id; |
c2c087e6 DW |
5835 | int rv; |
5836 | struct stat stb; | |
5837 | ||
88654014 LM |
5838 | /* If we are on an RAID enabled platform check that the disk is |
5839 | * attached to the raid controller. | |
5840 | * We do not need to test disks attachment for container based additions, | |
5841 | * they shall be already tested when container was created/assembled. | |
88c32bb1 | 5842 | */ |
d424212e | 5843 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 5844 | /* no orom/efi or non-intel hba of the disk */ |
f0f5a016 LM |
5845 | if (rv != 0) { |
5846 | dprintf("capability: %p fd: %d ret: %d\n", | |
5847 | super->orom, fd, rv); | |
5848 | return 1; | |
88c32bb1 DW |
5849 | } |
5850 | ||
f20c3968 DW |
5851 | if (super->current_vol >= 0) |
5852 | return add_to_super_imsm_volume(st, dk, fd, devname); | |
bf5a934a | 5853 | |
c2c087e6 | 5854 | fstat(fd, &stb); |
503975b9 | 5855 | dd = xcalloc(sizeof(*dd), 1); |
c2c087e6 DW |
5856 | dd->major = major(stb.st_rdev); |
5857 | dd->minor = minor(stb.st_rdev); | |
503975b9 | 5858 | dd->devname = devname ? xstrdup(devname) : NULL; |
c2c087e6 | 5859 | dd->fd = fd; |
689c9bf3 | 5860 | dd->e = NULL; |
1a64be56 | 5861 | dd->action = DISK_ADD; |
c2c087e6 | 5862 | rv = imsm_read_serial(fd, devname, dd->serial); |
32ba9157 | 5863 | if (rv) { |
e7b84f9d | 5864 | pr_err("failed to retrieve scsi serial, aborting\n"); |
20bee0f8 PB |
5865 | if (dd->devname) |
5866 | free(dd->devname); | |
949c47a0 | 5867 | free(dd); |
0030e8d6 | 5868 | abort(); |
c2c087e6 | 5869 | } |
20bee0f8 PB |
5870 | if (super->hba && ((super->hba->type == SYS_DEV_NVME) || |
5871 | (super->hba->type == SYS_DEV_VMD))) { | |
5872 | int i; | |
5873 | char *devpath = diskfd_to_devpath(fd); | |
5874 | char controller_path[PATH_MAX]; | |
5875 | ||
5876 | if (!devpath) { | |
5877 | pr_err("failed to get devpath, aborting\n"); | |
5878 | if (dd->devname) | |
5879 | free(dd->devname); | |
5880 | free(dd); | |
5881 | return 1; | |
5882 | } | |
5883 | ||
5884 | snprintf(controller_path, PATH_MAX-1, "%s/device", devpath); | |
5885 | free(devpath); | |
5886 | ||
5887 | if (devpath_to_vendor(controller_path) == 0x8086) { | |
5888 | /* | |
5889 | * If Intel's NVMe drive has serial ended with | |
5890 | * "-A","-B","-1" or "-2" it means that this is "x8" | |
5891 | * device (double drive on single PCIe card). | |
5892 | * User should be warned about potential data loss. | |
5893 | */ | |
5894 | for (i = MAX_RAID_SERIAL_LEN-1; i > 0; i--) { | |
5895 | /* Skip empty character at the end */ | |
5896 | if (dd->serial[i] == 0) | |
5897 | continue; | |
5898 | ||
5899 | if (((dd->serial[i] == 'A') || | |
5900 | (dd->serial[i] == 'B') || | |
5901 | (dd->serial[i] == '1') || | |
5902 | (dd->serial[i] == '2')) && | |
5903 | (dd->serial[i-1] == '-')) | |
5904 | pr_err("\tThe action you are about to take may put your data at risk.\n" | |
5905 | "\tPlease note that x8 devices may consist of two separate x4 devices " | |
5906 | "located on a single PCIe port.\n" | |
5907 | "\tRAID 0 is the only supported configuration for this type of x8 device.\n"); | |
5908 | break; | |
5909 | } | |
32716c51 PB |
5910 | } else if (super->hba->type == SYS_DEV_VMD && super->orom && |
5911 | !imsm_orom_has_tpv_support(super->orom)) { | |
5912 | pr_err("\tPlatform configuration does not support non-Intel NVMe drives.\n" | |
8b751247 | 5913 | "\tPlease refer to Intel(R) RSTe/VROC user guide.\n"); |
32716c51 PB |
5914 | free(dd->devname); |
5915 | free(dd); | |
5916 | return 1; | |
20bee0f8 PB |
5917 | } |
5918 | } | |
c2c087e6 | 5919 | |
c2c087e6 | 5920 | get_dev_size(fd, NULL, &size); |
fa7bb6f8 PB |
5921 | get_dev_sector_size(fd, NULL, &member_sector_size); |
5922 | ||
5923 | if (super->sector_size == 0) { | |
5924 | /* this a first device, so sector_size is not set yet */ | |
5925 | super->sector_size = member_sector_size; | |
fa7bb6f8 PB |
5926 | } |
5927 | ||
71e5411e | 5928 | /* clear migr_rec when adding disk to container */ |
85337573 AO |
5929 | memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE); |
5930 | if (lseek64(fd, size - MIGR_REC_SECTOR_POSITION*member_sector_size, | |
de44e46f | 5931 | SEEK_SET) >= 0) { |
466070ad | 5932 | if ((unsigned int)write(fd, super->migr_rec_buf, |
85337573 AO |
5933 | MIGR_REC_BUF_SECTORS*member_sector_size) != |
5934 | MIGR_REC_BUF_SECTORS*member_sector_size) | |
71e5411e PB |
5935 | perror("Write migr_rec failed"); |
5936 | } | |
5937 | ||
c2c087e6 | 5938 | size /= 512; |
1f24f035 | 5939 | serialcpy(dd->disk.serial, dd->serial); |
5551b113 CA |
5940 | set_total_blocks(&dd->disk, size); |
5941 | if (__le32_to_cpu(dd->disk.total_blocks_hi) > 0) { | |
5942 | struct imsm_super *mpb = super->anchor; | |
5943 | mpb->attributes |= MPB_ATTRIB_2TB_DISK; | |
5944 | } | |
a8619d23 | 5945 | mark_spare(dd); |
c2c087e6 | 5946 | if (sysfs_disk_to_scsi_id(fd, &id) == 0) |
b9f594fe | 5947 | dd->disk.scsi_id = __cpu_to_le32(id); |
c2c087e6 | 5948 | else |
b9f594fe | 5949 | dd->disk.scsi_id = __cpu_to_le32(0); |
43dad3d6 DW |
5950 | |
5951 | if (st->update_tail) { | |
1a64be56 LM |
5952 | dd->next = super->disk_mgmt_list; |
5953 | super->disk_mgmt_list = dd; | |
43dad3d6 DW |
5954 | } else { |
5955 | dd->next = super->disks; | |
5956 | super->disks = dd; | |
ceaf0ee1 | 5957 | super->updates_pending++; |
43dad3d6 | 5958 | } |
f20c3968 DW |
5959 | |
5960 | return 0; | |
cdddbdbc DW |
5961 | } |
5962 | ||
1a64be56 LM |
5963 | static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk) |
5964 | { | |
5965 | struct intel_super *super = st->sb; | |
5966 | struct dl *dd; | |
5967 | ||
5968 | /* remove from super works only in mdmon - for communication | |
5969 | * manager - monitor. Check if communication memory buffer | |
5970 | * is prepared. | |
5971 | */ | |
5972 | if (!st->update_tail) { | |
1ade5cc1 | 5973 | pr_err("shall be used in mdmon context only\n"); |
1a64be56 LM |
5974 | return 1; |
5975 | } | |
503975b9 | 5976 | dd = xcalloc(1, sizeof(*dd)); |
1a64be56 LM |
5977 | dd->major = dk->major; |
5978 | dd->minor = dk->minor; | |
1a64be56 | 5979 | dd->fd = -1; |
a8619d23 | 5980 | mark_spare(dd); |
1a64be56 LM |
5981 | dd->action = DISK_REMOVE; |
5982 | ||
5983 | dd->next = super->disk_mgmt_list; | |
5984 | super->disk_mgmt_list = dd; | |
5985 | ||
1a64be56 LM |
5986 | return 0; |
5987 | } | |
5988 | ||
f796af5d DW |
5989 | static int store_imsm_mpb(int fd, struct imsm_super *mpb); |
5990 | ||
5991 | static union { | |
f36a9ecd | 5992 | char buf[MAX_SECTOR_SIZE]; |
f796af5d | 5993 | struct imsm_super anchor; |
f36a9ecd | 5994 | } spare_record __attribute__ ((aligned(MAX_SECTOR_SIZE))); |
c2c087e6 | 5995 | |
d23fe947 DW |
5996 | /* spare records have their own family number and do not have any defined raid |
5997 | * devices | |
5998 | */ | |
5999 | static int write_super_imsm_spares(struct intel_super *super, int doclose) | |
6000 | { | |
d23fe947 | 6001 | struct imsm_super *mpb = super->anchor; |
f796af5d | 6002 | struct imsm_super *spare = &spare_record.anchor; |
d23fe947 DW |
6003 | __u32 sum; |
6004 | struct dl *d; | |
6005 | ||
68641cdb JS |
6006 | spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super)); |
6007 | spare->generation_num = __cpu_to_le32(1UL); | |
f796af5d | 6008 | spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; |
68641cdb JS |
6009 | spare->num_disks = 1; |
6010 | spare->num_raid_devs = 0; | |
6011 | spare->cache_size = mpb->cache_size; | |
6012 | spare->pwr_cycle_count = __cpu_to_le32(1); | |
f796af5d DW |
6013 | |
6014 | snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH, | |
6015 | MPB_SIGNATURE MPB_VERSION_RAID0); | |
d23fe947 DW |
6016 | |
6017 | for (d = super->disks; d; d = d->next) { | |
8796fdc4 | 6018 | if (d->index != -1) |
d23fe947 DW |
6019 | continue; |
6020 | ||
f796af5d | 6021 | spare->disk[0] = d->disk; |
027c374f CA |
6022 | if (__le32_to_cpu(d->disk.total_blocks_hi) > 0) |
6023 | spare->attributes |= MPB_ATTRIB_2TB_DISK; | |
6024 | ||
f36a9ecd PB |
6025 | if (super->sector_size == 4096) |
6026 | convert_to_4k_imsm_disk(&spare->disk[0]); | |
6027 | ||
f796af5d DW |
6028 | sum = __gen_imsm_checksum(spare); |
6029 | spare->family_num = __cpu_to_le32(sum); | |
6030 | spare->orig_family_num = 0; | |
6031 | sum = __gen_imsm_checksum(spare); | |
6032 | spare->check_sum = __cpu_to_le32(sum); | |
d23fe947 | 6033 | |
f796af5d | 6034 | if (store_imsm_mpb(d->fd, spare)) { |
1ade5cc1 N |
6035 | pr_err("failed for device %d:%d %s\n", |
6036 | d->major, d->minor, strerror(errno)); | |
e74255d9 | 6037 | return 1; |
d23fe947 DW |
6038 | } |
6039 | if (doclose) { | |
6040 | close(d->fd); | |
6041 | d->fd = -1; | |
6042 | } | |
6043 | } | |
6044 | ||
e74255d9 | 6045 | return 0; |
d23fe947 DW |
6046 | } |
6047 | ||
36988a3d | 6048 | static int write_super_imsm(struct supertype *st, int doclose) |
cdddbdbc | 6049 | { |
36988a3d | 6050 | struct intel_super *super = st->sb; |
f36a9ecd | 6051 | unsigned int sector_size = super->sector_size; |
949c47a0 | 6052 | struct imsm_super *mpb = super->anchor; |
c2c087e6 DW |
6053 | struct dl *d; |
6054 | __u32 generation; | |
6055 | __u32 sum; | |
d23fe947 | 6056 | int spares = 0; |
949c47a0 | 6057 | int i; |
a48ac0a8 | 6058 | __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk); |
36988a3d | 6059 | int num_disks = 0; |
146c6260 | 6060 | int clear_migration_record = 1; |
bbab0940 | 6061 | __u32 bbm_log_size; |
cdddbdbc | 6062 | |
c2c087e6 DW |
6063 | /* 'generation' is incremented everytime the metadata is written */ |
6064 | generation = __le32_to_cpu(mpb->generation_num); | |
6065 | generation++; | |
6066 | mpb->generation_num = __cpu_to_le32(generation); | |
6067 | ||
148acb7b DW |
6068 | /* fix up cases where previous mdadm releases failed to set |
6069 | * orig_family_num | |
6070 | */ | |
6071 | if (mpb->orig_family_num == 0) | |
6072 | mpb->orig_family_num = mpb->family_num; | |
6073 | ||
d23fe947 | 6074 | for (d = super->disks; d; d = d->next) { |
8796fdc4 | 6075 | if (d->index == -1) |
d23fe947 | 6076 | spares++; |
36988a3d | 6077 | else { |
d23fe947 | 6078 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
6079 | num_disks++; |
6080 | } | |
d23fe947 | 6081 | } |
36988a3d | 6082 | for (d = super->missing; d; d = d->next) { |
47ee5a45 | 6083 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
6084 | num_disks++; |
6085 | } | |
6086 | mpb->num_disks = num_disks; | |
6087 | mpb_size += sizeof(struct imsm_disk) * mpb->num_disks; | |
b9f594fe | 6088 | |
949c47a0 DW |
6089 | for (i = 0; i < mpb->num_raid_devs; i++) { |
6090 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
36988a3d AK |
6091 | struct imsm_dev *dev2 = get_imsm_dev(super, i); |
6092 | if (dev && dev2) { | |
6093 | imsm_copy_dev(dev, dev2); | |
6094 | mpb_size += sizeof_imsm_dev(dev, 0); | |
6095 | } | |
146c6260 AK |
6096 | if (is_gen_migration(dev2)) |
6097 | clear_migration_record = 0; | |
949c47a0 | 6098 | } |
bbab0940 TM |
6099 | |
6100 | bbm_log_size = get_imsm_bbm_log_size(super->bbm_log); | |
6101 | ||
6102 | if (bbm_log_size) { | |
6103 | memcpy((void *)mpb + mpb_size, super->bbm_log, bbm_log_size); | |
6104 | mpb->attributes |= MPB_ATTRIB_BBM; | |
6105 | } else | |
6106 | mpb->attributes &= ~MPB_ATTRIB_BBM; | |
6107 | ||
6108 | super->anchor->bbm_log_size = __cpu_to_le32(bbm_log_size); | |
6109 | mpb_size += bbm_log_size; | |
a48ac0a8 | 6110 | mpb->mpb_size = __cpu_to_le32(mpb_size); |
949c47a0 | 6111 | |
bbab0940 TM |
6112 | #ifdef DEBUG |
6113 | assert(super->len == 0 || mpb_size <= super->len); | |
6114 | #endif | |
6115 | ||
c2c087e6 | 6116 | /* recalculate checksum */ |
949c47a0 | 6117 | sum = __gen_imsm_checksum(mpb); |
c2c087e6 DW |
6118 | mpb->check_sum = __cpu_to_le32(sum); |
6119 | ||
51d83f5d AK |
6120 | if (super->clean_migration_record_by_mdmon) { |
6121 | clear_migration_record = 1; | |
6122 | super->clean_migration_record_by_mdmon = 0; | |
6123 | } | |
146c6260 | 6124 | if (clear_migration_record) |
de44e46f | 6125 | memset(super->migr_rec_buf, 0, |
85337573 | 6126 | MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE); |
146c6260 | 6127 | |
f36a9ecd PB |
6128 | if (sector_size == 4096) |
6129 | convert_to_4k(super); | |
6130 | ||
d23fe947 | 6131 | /* write the mpb for disks that compose raid devices */ |
c2c087e6 | 6132 | for (d = super->disks; d ; d = d->next) { |
86c54047 | 6133 | if (d->index < 0 || is_failed(&d->disk)) |
d23fe947 | 6134 | continue; |
30602f53 | 6135 | |
146c6260 AK |
6136 | if (clear_migration_record) { |
6137 | unsigned long long dsize; | |
6138 | ||
6139 | get_dev_size(d->fd, NULL, &dsize); | |
de44e46f PB |
6140 | if (lseek64(d->fd, dsize - sector_size, |
6141 | SEEK_SET) >= 0) { | |
466070ad PB |
6142 | if ((unsigned int)write(d->fd, |
6143 | super->migr_rec_buf, | |
de44e46f PB |
6144 | MIGR_REC_BUF_SECTORS*sector_size) != |
6145 | MIGR_REC_BUF_SECTORS*sector_size) | |
9e2d750d | 6146 | perror("Write migr_rec failed"); |
146c6260 AK |
6147 | } |
6148 | } | |
51d83f5d AK |
6149 | |
6150 | if (store_imsm_mpb(d->fd, mpb)) | |
6151 | fprintf(stderr, | |
1ade5cc1 N |
6152 | "failed for device %d:%d (fd: %d)%s\n", |
6153 | d->major, d->minor, | |
51d83f5d AK |
6154 | d->fd, strerror(errno)); |
6155 | ||
c2c087e6 DW |
6156 | if (doclose) { |
6157 | close(d->fd); | |
6158 | d->fd = -1; | |
6159 | } | |
6160 | } | |
6161 | ||
d23fe947 DW |
6162 | if (spares) |
6163 | return write_super_imsm_spares(super, doclose); | |
6164 | ||
e74255d9 | 6165 | return 0; |
c2c087e6 DW |
6166 | } |
6167 | ||
9b1fb677 | 6168 | static int create_array(struct supertype *st, int dev_idx) |
43dad3d6 DW |
6169 | { |
6170 | size_t len; | |
6171 | struct imsm_update_create_array *u; | |
6172 | struct intel_super *super = st->sb; | |
9b1fb677 | 6173 | struct imsm_dev *dev = get_imsm_dev(super, dev_idx); |
238c0a71 | 6174 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
54c2c1ea DW |
6175 | struct disk_info *inf; |
6176 | struct imsm_disk *disk; | |
6177 | int i; | |
43dad3d6 | 6178 | |
54c2c1ea DW |
6179 | len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) + |
6180 | sizeof(*inf) * map->num_members; | |
503975b9 | 6181 | u = xmalloc(len); |
43dad3d6 | 6182 | u->type = update_create_array; |
9b1fb677 | 6183 | u->dev_idx = dev_idx; |
43dad3d6 | 6184 | imsm_copy_dev(&u->dev, dev); |
54c2c1ea DW |
6185 | inf = get_disk_info(u); |
6186 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 6187 | int idx = get_imsm_disk_idx(dev, i, MAP_X); |
9b1fb677 | 6188 | |
54c2c1ea | 6189 | disk = get_imsm_disk(super, idx); |
1ca5c8e0 N |
6190 | if (!disk) |
6191 | disk = get_imsm_missing(super, idx); | |
54c2c1ea DW |
6192 | serialcpy(inf[i].serial, disk->serial); |
6193 | } | |
43dad3d6 DW |
6194 | append_metadata_update(st, u, len); |
6195 | ||
6196 | return 0; | |
6197 | } | |
6198 | ||
1a64be56 | 6199 | static int mgmt_disk(struct supertype *st) |
43dad3d6 DW |
6200 | { |
6201 | struct intel_super *super = st->sb; | |
6202 | size_t len; | |
1a64be56 | 6203 | struct imsm_update_add_remove_disk *u; |
43dad3d6 | 6204 | |
1a64be56 | 6205 | if (!super->disk_mgmt_list) |
43dad3d6 DW |
6206 | return 0; |
6207 | ||
6208 | len = sizeof(*u); | |
503975b9 | 6209 | u = xmalloc(len); |
1a64be56 | 6210 | u->type = update_add_remove_disk; |
43dad3d6 DW |
6211 | append_metadata_update(st, u, len); |
6212 | ||
6213 | return 0; | |
6214 | } | |
2432ce9b AP |
6215 | |
6216 | __u32 crc32c_le(__u32 crc, unsigned char const *p, size_t len); | |
6217 | ||
e397cefe AP |
6218 | static int write_ppl_header(unsigned long long ppl_sector, int fd, void *buf) |
6219 | { | |
6220 | struct ppl_header *ppl_hdr = buf; | |
6221 | int ret; | |
6222 | ||
6223 | ppl_hdr->checksum = __cpu_to_le32(~crc32c_le(~0, buf, PPL_HEADER_SIZE)); | |
6224 | ||
6225 | if (lseek64(fd, ppl_sector * 512, SEEK_SET) < 0) { | |
6226 | ret = -errno; | |
6227 | perror("Failed to seek to PPL header location"); | |
6228 | return ret; | |
6229 | } | |
6230 | ||
6231 | if (write(fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) { | |
6232 | ret = -errno; | |
6233 | perror("Write PPL header failed"); | |
6234 | return ret; | |
6235 | } | |
6236 | ||
6237 | fsync(fd); | |
6238 | ||
6239 | return 0; | |
6240 | } | |
6241 | ||
2432ce9b AP |
6242 | static int write_init_ppl_imsm(struct supertype *st, struct mdinfo *info, int fd) |
6243 | { | |
6244 | struct intel_super *super = st->sb; | |
6245 | void *buf; | |
6246 | struct ppl_header *ppl_hdr; | |
6247 | int ret; | |
6248 | ||
b2514242 PB |
6249 | /* first clear entire ppl space */ |
6250 | ret = zero_disk_range(fd, info->ppl_sector, info->ppl_size); | |
6251 | if (ret) | |
6252 | return ret; | |
6253 | ||
6254 | ret = posix_memalign(&buf, MAX_SECTOR_SIZE, PPL_HEADER_SIZE); | |
2432ce9b AP |
6255 | if (ret) { |
6256 | pr_err("Failed to allocate PPL header buffer\n"); | |
e397cefe | 6257 | return -ret; |
2432ce9b AP |
6258 | } |
6259 | ||
6260 | memset(buf, 0, PPL_HEADER_SIZE); | |
6261 | ppl_hdr = buf; | |
6262 | memset(ppl_hdr->reserved, 0xff, PPL_HDR_RESERVED); | |
6263 | ppl_hdr->signature = __cpu_to_le32(super->anchor->orig_family_num); | |
b23d0750 AP |
6264 | |
6265 | if (info->mismatch_cnt) { | |
6266 | /* | |
6267 | * We are overwriting an invalid ppl. Make one entry with wrong | |
6268 | * checksum to prevent the kernel from skipping resync. | |
6269 | */ | |
6270 | ppl_hdr->entries_count = __cpu_to_le32(1); | |
6271 | ppl_hdr->entries[0].checksum = ~0; | |
6272 | } | |
6273 | ||
e397cefe | 6274 | ret = write_ppl_header(info->ppl_sector, fd, buf); |
2432ce9b AP |
6275 | |
6276 | free(buf); | |
6277 | return ret; | |
6278 | } | |
6279 | ||
e397cefe AP |
6280 | static int is_rebuilding(struct imsm_dev *dev); |
6281 | ||
2432ce9b AP |
6282 | static int validate_ppl_imsm(struct supertype *st, struct mdinfo *info, |
6283 | struct mdinfo *disk) | |
6284 | { | |
6285 | struct intel_super *super = st->sb; | |
6286 | struct dl *d; | |
e397cefe | 6287 | void *buf_orig, *buf, *buf_prev = NULL; |
2432ce9b | 6288 | int ret = 0; |
e397cefe | 6289 | struct ppl_header *ppl_hdr = NULL; |
2432ce9b AP |
6290 | __u32 crc; |
6291 | struct imsm_dev *dev; | |
2432ce9b | 6292 | __u32 idx; |
44b6b876 PB |
6293 | unsigned int i; |
6294 | unsigned long long ppl_offset = 0; | |
6295 | unsigned long long prev_gen_num = 0; | |
2432ce9b AP |
6296 | |
6297 | if (disk->disk.raid_disk < 0) | |
6298 | return 0; | |
6299 | ||
2432ce9b | 6300 | dev = get_imsm_dev(super, info->container_member); |
2fc0fc63 | 6301 | idx = get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_0); |
2432ce9b AP |
6302 | d = get_imsm_dl_disk(super, idx); |
6303 | ||
6304 | if (!d || d->index < 0 || is_failed(&d->disk)) | |
e397cefe AP |
6305 | return 0; |
6306 | ||
6307 | if (posix_memalign(&buf_orig, MAX_SECTOR_SIZE, PPL_HEADER_SIZE * 2)) { | |
6308 | pr_err("Failed to allocate PPL header buffer\n"); | |
6309 | return -1; | |
6310 | } | |
6311 | buf = buf_orig; | |
2432ce9b | 6312 | |
44b6b876 PB |
6313 | ret = 1; |
6314 | while (ppl_offset < MULTIPLE_PPL_AREA_SIZE_IMSM) { | |
e397cefe AP |
6315 | void *tmp; |
6316 | ||
44b6b876 | 6317 | dprintf("Checking potential PPL at offset: %llu\n", ppl_offset); |
2432ce9b | 6318 | |
44b6b876 PB |
6319 | if (lseek64(d->fd, info->ppl_sector * 512 + ppl_offset, |
6320 | SEEK_SET) < 0) { | |
6321 | perror("Failed to seek to PPL header location"); | |
6322 | ret = -1; | |
e397cefe | 6323 | break; |
44b6b876 | 6324 | } |
2432ce9b | 6325 | |
44b6b876 PB |
6326 | if (read(d->fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) { |
6327 | perror("Read PPL header failed"); | |
6328 | ret = -1; | |
e397cefe | 6329 | break; |
44b6b876 | 6330 | } |
2432ce9b | 6331 | |
44b6b876 | 6332 | ppl_hdr = buf; |
2432ce9b | 6333 | |
44b6b876 PB |
6334 | crc = __le32_to_cpu(ppl_hdr->checksum); |
6335 | ppl_hdr->checksum = 0; | |
6336 | ||
6337 | if (crc != ~crc32c_le(~0, buf, PPL_HEADER_SIZE)) { | |
6338 | dprintf("Wrong PPL header checksum on %s\n", | |
6339 | d->devname); | |
e397cefe | 6340 | break; |
44b6b876 PB |
6341 | } |
6342 | ||
6343 | if (prev_gen_num > __le64_to_cpu(ppl_hdr->generation)) { | |
6344 | /* previous was newest, it was already checked */ | |
e397cefe | 6345 | break; |
44b6b876 PB |
6346 | } |
6347 | ||
6348 | if ((__le32_to_cpu(ppl_hdr->signature) != | |
6349 | super->anchor->orig_family_num)) { | |
6350 | dprintf("Wrong PPL header signature on %s\n", | |
6351 | d->devname); | |
6352 | ret = 1; | |
e397cefe | 6353 | break; |
44b6b876 PB |
6354 | } |
6355 | ||
6356 | ret = 0; | |
6357 | prev_gen_num = __le64_to_cpu(ppl_hdr->generation); | |
2432ce9b | 6358 | |
44b6b876 PB |
6359 | ppl_offset += PPL_HEADER_SIZE; |
6360 | for (i = 0; i < __le32_to_cpu(ppl_hdr->entries_count); i++) | |
6361 | ppl_offset += | |
6362 | __le32_to_cpu(ppl_hdr->entries[i].pp_size); | |
e397cefe AP |
6363 | |
6364 | if (!buf_prev) | |
6365 | buf_prev = buf + PPL_HEADER_SIZE; | |
6366 | tmp = buf_prev; | |
6367 | buf_prev = buf; | |
6368 | buf = tmp; | |
2432ce9b AP |
6369 | } |
6370 | ||
e397cefe AP |
6371 | if (buf_prev) { |
6372 | buf = buf_prev; | |
6373 | ppl_hdr = buf_prev; | |
6374 | } | |
2432ce9b | 6375 | |
54148aba PB |
6376 | /* |
6377 | * Update metadata to use mutliple PPLs area (1MB). | |
6378 | * This is done once for all RAID members | |
6379 | */ | |
6380 | if (info->consistency_policy == CONSISTENCY_POLICY_PPL && | |
6381 | info->ppl_size != (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9)) { | |
6382 | char subarray[20]; | |
6383 | struct mdinfo *member_dev; | |
6384 | ||
6385 | sprintf(subarray, "%d", info->container_member); | |
6386 | ||
6387 | if (mdmon_running(st->container_devnm)) | |
6388 | st->update_tail = &st->updates; | |
6389 | ||
6390 | if (st->ss->update_subarray(st, subarray, "ppl", NULL)) { | |
6391 | pr_err("Failed to update subarray %s\n", | |
6392 | subarray); | |
6393 | } else { | |
6394 | if (st->update_tail) | |
6395 | flush_metadata_updates(st); | |
6396 | else | |
6397 | st->ss->sync_metadata(st); | |
6398 | info->ppl_size = (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9); | |
6399 | for (member_dev = info->devs; member_dev; | |
6400 | member_dev = member_dev->next) | |
6401 | member_dev->ppl_size = | |
6402 | (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9); | |
6403 | } | |
6404 | } | |
6405 | ||
b23d0750 | 6406 | if (ret == 1) { |
2fc0fc63 AP |
6407 | struct imsm_map *map = get_imsm_map(dev, MAP_X); |
6408 | ||
50b9c10d PB |
6409 | if (map->map_state == IMSM_T_STATE_UNINITIALIZED || |
6410 | (map->map_state == IMSM_T_STATE_NORMAL && | |
2ec9d182 | 6411 | !(dev->vol.dirty & RAIDVOL_DIRTY)) || |
e397cefe | 6412 | (is_rebuilding(dev) && |
2ec9d182 AP |
6413 | dev->vol.curr_migr_unit == 0 && |
6414 | get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_1) != idx)) | |
b23d0750 AP |
6415 | ret = st->ss->write_init_ppl(st, info, d->fd); |
6416 | else | |
6417 | info->mismatch_cnt++; | |
e397cefe AP |
6418 | } else if (ret == 0 && |
6419 | ppl_hdr->entries_count == 0 && | |
6420 | is_rebuilding(dev) && | |
6421 | info->resync_start == 0) { | |
6422 | /* | |
6423 | * The header has no entries - add a single empty entry and | |
6424 | * rewrite the header to prevent the kernel from going into | |
6425 | * resync after an interrupted rebuild. | |
6426 | */ | |
6427 | ppl_hdr->entries_count = __cpu_to_le32(1); | |
6428 | ret = write_ppl_header(info->ppl_sector, d->fd, buf); | |
b23d0750 | 6429 | } |
2432ce9b | 6430 | |
e397cefe AP |
6431 | free(buf_orig); |
6432 | ||
2432ce9b AP |
6433 | return ret; |
6434 | } | |
6435 | ||
2432ce9b AP |
6436 | static int write_init_ppl_imsm_all(struct supertype *st, struct mdinfo *info) |
6437 | { | |
6438 | struct intel_super *super = st->sb; | |
6439 | struct dl *d; | |
6440 | int ret = 0; | |
6441 | ||
6442 | if (info->consistency_policy != CONSISTENCY_POLICY_PPL || | |
6443 | info->array.level != 5) | |
6444 | return 0; | |
6445 | ||
6446 | for (d = super->disks; d ; d = d->next) { | |
6447 | if (d->index < 0 || is_failed(&d->disk)) | |
6448 | continue; | |
6449 | ||
6450 | ret = st->ss->write_init_ppl(st, info, d->fd); | |
6451 | if (ret) | |
6452 | break; | |
6453 | } | |
6454 | ||
6455 | return ret; | |
6456 | } | |
43dad3d6 | 6457 | |
c2c087e6 DW |
6458 | static int write_init_super_imsm(struct supertype *st) |
6459 | { | |
9b1fb677 DW |
6460 | struct intel_super *super = st->sb; |
6461 | int current_vol = super->current_vol; | |
2432ce9b AP |
6462 | int rv = 0; |
6463 | struct mdinfo info; | |
6464 | ||
6465 | getinfo_super_imsm(st, &info, NULL); | |
9b1fb677 DW |
6466 | |
6467 | /* we are done with current_vol reset it to point st at the container */ | |
6468 | super->current_vol = -1; | |
6469 | ||
8273f55e | 6470 | if (st->update_tail) { |
43dad3d6 DW |
6471 | /* queue the recently created array / added disk |
6472 | * as a metadata update */ | |
8273f55e | 6473 | |
43dad3d6 | 6474 | /* determine if we are creating a volume or adding a disk */ |
9b1fb677 | 6475 | if (current_vol < 0) { |
1a64be56 LM |
6476 | /* in the mgmt (add/remove) disk case we are running |
6477 | * in mdmon context, so don't close fd's | |
43dad3d6 | 6478 | */ |
2432ce9b AP |
6479 | rv = mgmt_disk(st); |
6480 | } else { | |
6481 | rv = write_init_ppl_imsm_all(st, &info); | |
6482 | if (!rv) | |
6483 | rv = create_array(st, current_vol); | |
6484 | } | |
d682f344 N |
6485 | } else { |
6486 | struct dl *d; | |
6487 | for (d = super->disks; d; d = d->next) | |
ba728be7 | 6488 | Kill(d->devname, NULL, 0, -1, 1); |
2432ce9b AP |
6489 | if (current_vol >= 0) |
6490 | rv = write_init_ppl_imsm_all(st, &info); | |
6491 | if (!rv) | |
6492 | rv = write_super_imsm(st, 1); | |
d682f344 | 6493 | } |
2432ce9b AP |
6494 | |
6495 | return rv; | |
cdddbdbc DW |
6496 | } |
6497 | ||
e683ca88 | 6498 | static int store_super_imsm(struct supertype *st, int fd) |
cdddbdbc | 6499 | { |
e683ca88 DW |
6500 | struct intel_super *super = st->sb; |
6501 | struct imsm_super *mpb = super ? super->anchor : NULL; | |
551c80c1 | 6502 | |
e683ca88 | 6503 | if (!mpb) |
ad97895e DW |
6504 | return 1; |
6505 | ||
f36a9ecd PB |
6506 | if (super->sector_size == 4096) |
6507 | convert_to_4k(super); | |
e683ca88 | 6508 | return store_imsm_mpb(fd, mpb); |
cdddbdbc DW |
6509 | } |
6510 | ||
cdddbdbc DW |
6511 | static int validate_geometry_imsm_container(struct supertype *st, int level, |
6512 | int layout, int raiddisks, int chunk, | |
af4348dd N |
6513 | unsigned long long size, |
6514 | unsigned long long data_offset, | |
6515 | char *dev, | |
2c514b71 NB |
6516 | unsigned long long *freesize, |
6517 | int verbose) | |
cdddbdbc | 6518 | { |
c2c087e6 DW |
6519 | int fd; |
6520 | unsigned long long ldsize; | |
594dc1b8 | 6521 | struct intel_super *super; |
f2f5c343 | 6522 | int rv = 0; |
cdddbdbc | 6523 | |
c2c087e6 DW |
6524 | if (level != LEVEL_CONTAINER) |
6525 | return 0; | |
6526 | if (!dev) | |
6527 | return 1; | |
6528 | ||
6529 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
6530 | if (fd < 0) { | |
ba728be7 | 6531 | if (verbose > 0) |
e7b84f9d | 6532 | pr_err("imsm: Cannot open %s: %s\n", |
2c514b71 | 6533 | dev, strerror(errno)); |
c2c087e6 DW |
6534 | return 0; |
6535 | } | |
6536 | if (!get_dev_size(fd, dev, &ldsize)) { | |
6537 | close(fd); | |
6538 | return 0; | |
6539 | } | |
f2f5c343 LM |
6540 | |
6541 | /* capabilities retrieve could be possible | |
6542 | * note that there is no fd for the disks in array. | |
6543 | */ | |
6544 | super = alloc_super(); | |
8d67477f TM |
6545 | if (!super) { |
6546 | close(fd); | |
6547 | return 0; | |
6548 | } | |
fa7bb6f8 PB |
6549 | if (!get_dev_sector_size(fd, NULL, &super->sector_size)) { |
6550 | close(fd); | |
6551 | free_imsm(super); | |
6552 | return 0; | |
6553 | } | |
6554 | ||
ba728be7 | 6555 | rv = find_intel_hba_capability(fd, super, verbose > 0 ? dev : NULL); |
f2f5c343 LM |
6556 | if (rv != 0) { |
6557 | #if DEBUG | |
6558 | char str[256]; | |
6559 | fd2devname(fd, str); | |
1ade5cc1 | 6560 | dprintf("fd: %d %s orom: %p rv: %d raiddisk: %d\n", |
f2f5c343 LM |
6561 | fd, str, super->orom, rv, raiddisks); |
6562 | #endif | |
6563 | /* no orom/efi or non-intel hba of the disk */ | |
6564 | close(fd); | |
6565 | free_imsm(super); | |
6566 | return 0; | |
6567 | } | |
c2c087e6 | 6568 | close(fd); |
9126b9a8 CA |
6569 | if (super->orom) { |
6570 | if (raiddisks > super->orom->tds) { | |
6571 | if (verbose) | |
7a862a02 | 6572 | pr_err("%d exceeds maximum number of platform supported disks: %d\n", |
9126b9a8 CA |
6573 | raiddisks, super->orom->tds); |
6574 | free_imsm(super); | |
6575 | return 0; | |
6576 | } | |
6577 | if ((super->orom->attr & IMSM_OROM_ATTR_2TB_DISK) == 0 && | |
6578 | (ldsize >> 9) >> 32 > 0) { | |
6579 | if (verbose) | |
e7b84f9d | 6580 | pr_err("%s exceeds maximum platform supported size\n", dev); |
9126b9a8 CA |
6581 | free_imsm(super); |
6582 | return 0; | |
6583 | } | |
f2f5c343 | 6584 | } |
c2c087e6 | 6585 | |
af4348dd | 6586 | *freesize = avail_size_imsm(st, ldsize >> 9, data_offset); |
f2f5c343 | 6587 | free_imsm(super); |
c2c087e6 DW |
6588 | |
6589 | return 1; | |
cdddbdbc DW |
6590 | } |
6591 | ||
0dcecb2e DW |
6592 | static unsigned long long find_size(struct extent *e, int *idx, int num_extents) |
6593 | { | |
6594 | const unsigned long long base_start = e[*idx].start; | |
6595 | unsigned long long end = base_start + e[*idx].size; | |
6596 | int i; | |
6597 | ||
6598 | if (base_start == end) | |
6599 | return 0; | |
6600 | ||
6601 | *idx = *idx + 1; | |
6602 | for (i = *idx; i < num_extents; i++) { | |
6603 | /* extend overlapping extents */ | |
6604 | if (e[i].start >= base_start && | |
6605 | e[i].start <= end) { | |
6606 | if (e[i].size == 0) | |
6607 | return 0; | |
6608 | if (e[i].start + e[i].size > end) | |
6609 | end = e[i].start + e[i].size; | |
6610 | } else if (e[i].start > end) { | |
6611 | *idx = i; | |
6612 | break; | |
6613 | } | |
6614 | } | |
6615 | ||
6616 | return end - base_start; | |
6617 | } | |
6618 | ||
6619 | static unsigned long long merge_extents(struct intel_super *super, int sum_extents) | |
6620 | { | |
6621 | /* build a composite disk with all known extents and generate a new | |
6622 | * 'maxsize' given the "all disks in an array must share a common start | |
6623 | * offset" constraint | |
6624 | */ | |
503975b9 | 6625 | struct extent *e = xcalloc(sum_extents, sizeof(*e)); |
0dcecb2e DW |
6626 | struct dl *dl; |
6627 | int i, j; | |
6628 | int start_extent; | |
6629 | unsigned long long pos; | |
b9d77223 | 6630 | unsigned long long start = 0; |
0dcecb2e DW |
6631 | unsigned long long maxsize; |
6632 | unsigned long reserve; | |
6633 | ||
0dcecb2e DW |
6634 | /* coalesce and sort all extents. also, check to see if we need to |
6635 | * reserve space between member arrays | |
6636 | */ | |
6637 | j = 0; | |
6638 | for (dl = super->disks; dl; dl = dl->next) { | |
6639 | if (!dl->e) | |
6640 | continue; | |
6641 | for (i = 0; i < dl->extent_cnt; i++) | |
6642 | e[j++] = dl->e[i]; | |
6643 | } | |
6644 | qsort(e, sum_extents, sizeof(*e), cmp_extent); | |
6645 | ||
6646 | /* merge extents */ | |
6647 | i = 0; | |
6648 | j = 0; | |
6649 | while (i < sum_extents) { | |
6650 | e[j].start = e[i].start; | |
6651 | e[j].size = find_size(e, &i, sum_extents); | |
6652 | j++; | |
6653 | if (e[j-1].size == 0) | |
6654 | break; | |
6655 | } | |
6656 | ||
6657 | pos = 0; | |
6658 | maxsize = 0; | |
6659 | start_extent = 0; | |
6660 | i = 0; | |
6661 | do { | |
6662 | unsigned long long esize; | |
6663 | ||
6664 | esize = e[i].start - pos; | |
6665 | if (esize >= maxsize) { | |
6666 | maxsize = esize; | |
6667 | start = pos; | |
6668 | start_extent = i; | |
6669 | } | |
6670 | pos = e[i].start + e[i].size; | |
6671 | i++; | |
6672 | } while (e[i-1].size); | |
6673 | free(e); | |
6674 | ||
a7dd165b DW |
6675 | if (maxsize == 0) |
6676 | return 0; | |
6677 | ||
6678 | /* FIXME assumes volume at offset 0 is the first volume in a | |
6679 | * container | |
6680 | */ | |
0dcecb2e DW |
6681 | if (start_extent > 0) |
6682 | reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */ | |
6683 | else | |
6684 | reserve = 0; | |
6685 | ||
6686 | if (maxsize < reserve) | |
a7dd165b | 6687 | return 0; |
0dcecb2e | 6688 | |
5551b113 | 6689 | super->create_offset = ~((unsigned long long) 0); |
0dcecb2e | 6690 | if (start + reserve > super->create_offset) |
a7dd165b | 6691 | return 0; /* start overflows create_offset */ |
0dcecb2e DW |
6692 | super->create_offset = start + reserve; |
6693 | ||
6694 | return maxsize - reserve; | |
6695 | } | |
6696 | ||
88c32bb1 DW |
6697 | static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks) |
6698 | { | |
6699 | if (level < 0 || level == 6 || level == 4) | |
6700 | return 0; | |
6701 | ||
6702 | /* if we have an orom prevent invalid raid levels */ | |
6703 | if (orom) | |
6704 | switch (level) { | |
6705 | case 0: return imsm_orom_has_raid0(orom); | |
6706 | case 1: | |
6707 | if (raiddisks > 2) | |
6708 | return imsm_orom_has_raid1e(orom); | |
1c556e92 DW |
6709 | return imsm_orom_has_raid1(orom) && raiddisks == 2; |
6710 | case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4; | |
6711 | case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2; | |
88c32bb1 DW |
6712 | } |
6713 | else | |
6714 | return 1; /* not on an Intel RAID platform so anything goes */ | |
6715 | ||
6716 | return 0; | |
6717 | } | |
6718 | ||
ca9de185 LM |
6719 | static int |
6720 | active_arrays_by_format(char *name, char* hba, struct md_list **devlist, | |
6721 | int dpa, int verbose) | |
6722 | { | |
6723 | struct mdstat_ent *mdstat = mdstat_read(0, 0); | |
594dc1b8 | 6724 | struct mdstat_ent *memb; |
ca9de185 LM |
6725 | int count = 0; |
6726 | int num = 0; | |
594dc1b8 | 6727 | struct md_list *dv; |
ca9de185 LM |
6728 | int found; |
6729 | ||
6730 | for (memb = mdstat ; memb ; memb = memb->next) { | |
6731 | if (memb->metadata_version && | |
fc54fe7a | 6732 | (strncmp(memb->metadata_version, "external:", 9) == 0) && |
ca9de185 LM |
6733 | (strcmp(&memb->metadata_version[9], name) == 0) && |
6734 | !is_subarray(memb->metadata_version+9) && | |
6735 | memb->members) { | |
6736 | struct dev_member *dev = memb->members; | |
6737 | int fd = -1; | |
6738 | while(dev && (fd < 0)) { | |
503975b9 N |
6739 | char *path = xmalloc(strlen(dev->name) + strlen("/dev/") + 1); |
6740 | num = sprintf(path, "%s%s", "/dev/", dev->name); | |
6741 | if (num > 0) | |
6742 | fd = open(path, O_RDONLY, 0); | |
089f9d79 | 6743 | if (num <= 0 || fd < 0) { |
676e87a8 | 6744 | pr_vrb("Cannot open %s: %s\n", |
503975b9 | 6745 | dev->name, strerror(errno)); |
ca9de185 | 6746 | } |
503975b9 | 6747 | free(path); |
ca9de185 LM |
6748 | dev = dev->next; |
6749 | } | |
6750 | found = 0; | |
089f9d79 | 6751 | if (fd >= 0 && disk_attached_to_hba(fd, hba)) { |
ca9de185 LM |
6752 | struct mdstat_ent *vol; |
6753 | for (vol = mdstat ; vol ; vol = vol->next) { | |
089f9d79 | 6754 | if (vol->active > 0 && |
ca9de185 | 6755 | vol->metadata_version && |
9581efb1 | 6756 | is_container_member(vol, memb->devnm)) { |
ca9de185 LM |
6757 | found++; |
6758 | count++; | |
6759 | } | |
6760 | } | |
6761 | if (*devlist && (found < dpa)) { | |
503975b9 | 6762 | dv = xcalloc(1, sizeof(*dv)); |
9581efb1 N |
6763 | dv->devname = xmalloc(strlen(memb->devnm) + strlen("/dev/") + 1); |
6764 | sprintf(dv->devname, "%s%s", "/dev/", memb->devnm); | |
503975b9 N |
6765 | dv->found = found; |
6766 | dv->used = 0; | |
6767 | dv->next = *devlist; | |
6768 | *devlist = dv; | |
ca9de185 LM |
6769 | } |
6770 | } | |
6771 | if (fd >= 0) | |
6772 | close(fd); | |
6773 | } | |
6774 | } | |
6775 | free_mdstat(mdstat); | |
6776 | return count; | |
6777 | } | |
6778 | ||
6779 | #ifdef DEBUG_LOOP | |
6780 | static struct md_list* | |
6781 | get_loop_devices(void) | |
6782 | { | |
6783 | int i; | |
6784 | struct md_list *devlist = NULL; | |
594dc1b8 | 6785 | struct md_list *dv; |
ca9de185 LM |
6786 | |
6787 | for(i = 0; i < 12; i++) { | |
503975b9 N |
6788 | dv = xcalloc(1, sizeof(*dv)); |
6789 | dv->devname = xmalloc(40); | |
ca9de185 LM |
6790 | sprintf(dv->devname, "/dev/loop%d", i); |
6791 | dv->next = devlist; | |
6792 | devlist = dv; | |
6793 | } | |
6794 | return devlist; | |
6795 | } | |
6796 | #endif | |
6797 | ||
6798 | static struct md_list* | |
6799 | get_devices(const char *hba_path) | |
6800 | { | |
6801 | struct md_list *devlist = NULL; | |
594dc1b8 | 6802 | struct md_list *dv; |
ca9de185 LM |
6803 | struct dirent *ent; |
6804 | DIR *dir; | |
6805 | int err = 0; | |
6806 | ||
6807 | #if DEBUG_LOOP | |
6808 | devlist = get_loop_devices(); | |
6809 | return devlist; | |
6810 | #endif | |
6811 | /* scroll through /sys/dev/block looking for devices attached to | |
6812 | * this hba | |
6813 | */ | |
6814 | dir = opendir("/sys/dev/block"); | |
6815 | for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) { | |
6816 | int fd; | |
6817 | char buf[1024]; | |
6818 | int major, minor; | |
6819 | char *path = NULL; | |
6820 | if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2) | |
6821 | continue; | |
6822 | path = devt_to_devpath(makedev(major, minor)); | |
6823 | if (!path) | |
6824 | continue; | |
6825 | if (!path_attached_to_hba(path, hba_path)) { | |
6826 | free(path); | |
6827 | path = NULL; | |
6828 | continue; | |
6829 | } | |
6830 | free(path); | |
6831 | path = NULL; | |
6832 | fd = dev_open(ent->d_name, O_RDONLY); | |
6833 | if (fd >= 0) { | |
6834 | fd2devname(fd, buf); | |
6835 | close(fd); | |
6836 | } else { | |
e7b84f9d | 6837 | pr_err("cannot open device: %s\n", |
ca9de185 LM |
6838 | ent->d_name); |
6839 | continue; | |
6840 | } | |
6841 | ||
503975b9 N |
6842 | dv = xcalloc(1, sizeof(*dv)); |
6843 | dv->devname = xstrdup(buf); | |
ca9de185 LM |
6844 | dv->next = devlist; |
6845 | devlist = dv; | |
6846 | } | |
6847 | if (err) { | |
6848 | while(devlist) { | |
6849 | dv = devlist; | |
6850 | devlist = devlist->next; | |
6851 | free(dv->devname); | |
6852 | free(dv); | |
6853 | } | |
6854 | } | |
562aa102 | 6855 | closedir(dir); |
ca9de185 LM |
6856 | return devlist; |
6857 | } | |
6858 | ||
6859 | static int | |
6860 | count_volumes_list(struct md_list *devlist, char *homehost, | |
6861 | int verbose, int *found) | |
6862 | { | |
6863 | struct md_list *tmpdev; | |
6864 | int count = 0; | |
594dc1b8 | 6865 | struct supertype *st; |
ca9de185 LM |
6866 | |
6867 | /* first walk the list of devices to find a consistent set | |
6868 | * that match the criterea, if that is possible. | |
6869 | * We flag the ones we like with 'used'. | |
6870 | */ | |
6871 | *found = 0; | |
6872 | st = match_metadata_desc_imsm("imsm"); | |
6873 | if (st == NULL) { | |
676e87a8 | 6874 | pr_vrb("cannot allocate memory for imsm supertype\n"); |
ca9de185 LM |
6875 | return 0; |
6876 | } | |
6877 | ||
6878 | for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { | |
6879 | char *devname = tmpdev->devname; | |
0a6bff09 | 6880 | dev_t rdev; |
ca9de185 LM |
6881 | struct supertype *tst; |
6882 | int dfd; | |
6883 | if (tmpdev->used > 1) | |
6884 | continue; | |
6885 | tst = dup_super(st); | |
6886 | if (tst == NULL) { | |
676e87a8 | 6887 | pr_vrb("cannot allocate memory for imsm supertype\n"); |
ca9de185 LM |
6888 | goto err_1; |
6889 | } | |
6890 | tmpdev->container = 0; | |
6891 | dfd = dev_open(devname, O_RDONLY|O_EXCL); | |
6892 | if (dfd < 0) { | |
1ade5cc1 | 6893 | dprintf("cannot open device %s: %s\n", |
ca9de185 LM |
6894 | devname, strerror(errno)); |
6895 | tmpdev->used = 2; | |
0a6bff09 | 6896 | } else if (!fstat_is_blkdev(dfd, devname, &rdev)) { |
ca9de185 LM |
6897 | tmpdev->used = 2; |
6898 | } else if (must_be_container(dfd)) { | |
6899 | struct supertype *cst; | |
6900 | cst = super_by_fd(dfd, NULL); | |
6901 | if (cst == NULL) { | |
1ade5cc1 | 6902 | dprintf("cannot recognize container type %s\n", |
ca9de185 LM |
6903 | devname); |
6904 | tmpdev->used = 2; | |
6905 | } else if (tst->ss != st->ss) { | |
1ade5cc1 | 6906 | dprintf("non-imsm container - ignore it: %s\n", |
ca9de185 LM |
6907 | devname); |
6908 | tmpdev->used = 2; | |
6909 | } else if (!tst->ss->load_container || | |
6910 | tst->ss->load_container(tst, dfd, NULL)) | |
6911 | tmpdev->used = 2; | |
6912 | else { | |
6913 | tmpdev->container = 1; | |
6914 | } | |
6915 | if (cst) | |
6916 | cst->ss->free_super(cst); | |
6917 | } else { | |
0a6bff09 | 6918 | tmpdev->st_rdev = rdev; |
ca9de185 | 6919 | if (tst->ss->load_super(tst,dfd, NULL)) { |
1ade5cc1 | 6920 | dprintf("no RAID superblock on %s\n", |
ca9de185 LM |
6921 | devname); |
6922 | tmpdev->used = 2; | |
6923 | } else if (tst->ss->compare_super == NULL) { | |
1ade5cc1 | 6924 | dprintf("Cannot assemble %s metadata on %s\n", |
ca9de185 LM |
6925 | tst->ss->name, devname); |
6926 | tmpdev->used = 2; | |
6927 | } | |
6928 | } | |
6929 | if (dfd >= 0) | |
6930 | close(dfd); | |
6931 | if (tmpdev->used == 2 || tmpdev->used == 4) { | |
6932 | /* Ignore unrecognised devices during auto-assembly */ | |
6933 | goto loop; | |
6934 | } | |
6935 | else { | |
6936 | struct mdinfo info; | |
6937 | tst->ss->getinfo_super(tst, &info, NULL); | |
6938 | ||
6939 | if (st->minor_version == -1) | |
6940 | st->minor_version = tst->minor_version; | |
6941 | ||
6942 | if (memcmp(info.uuid, uuid_zero, | |
6943 | sizeof(int[4])) == 0) { | |
6944 | /* this is a floating spare. It cannot define | |
6945 | * an array unless there are no more arrays of | |
6946 | * this type to be found. It can be included | |
6947 | * in an array of this type though. | |
6948 | */ | |
6949 | tmpdev->used = 3; | |
6950 | goto loop; | |
6951 | } | |
6952 | ||
6953 | if (st->ss != tst->ss || | |
6954 | st->minor_version != tst->minor_version || | |
6955 | st->ss->compare_super(st, tst) != 0) { | |
6956 | /* Some mismatch. If exactly one array matches this host, | |
6957 | * we can resolve on that one. | |
6958 | * Or, if we are auto assembling, we just ignore the second | |
6959 | * for now. | |
6960 | */ | |
1ade5cc1 | 6961 | dprintf("superblock on %s doesn't match others - assembly aborted\n", |
ca9de185 LM |
6962 | devname); |
6963 | goto loop; | |
6964 | } | |
6965 | tmpdev->used = 1; | |
6966 | *found = 1; | |
6967 | dprintf("found: devname: %s\n", devname); | |
6968 | } | |
6969 | loop: | |
6970 | if (tst) | |
6971 | tst->ss->free_super(tst); | |
6972 | } | |
6973 | if (*found != 0) { | |
6974 | int err; | |
6975 | if ((err = load_super_imsm_all(st, -1, &st->sb, NULL, devlist, 0)) == 0) { | |
6976 | struct mdinfo *iter, *head = st->ss->container_content(st, NULL); | |
6977 | for (iter = head; iter; iter = iter->next) { | |
6978 | dprintf("content->text_version: %s vol\n", | |
6979 | iter->text_version); | |
6980 | if (iter->array.state & (1<<MD_SB_BLOCK_VOLUME)) { | |
6981 | /* do not assemble arrays with unsupported | |
6982 | configurations */ | |
1ade5cc1 | 6983 | dprintf("Cannot activate member %s.\n", |
ca9de185 LM |
6984 | iter->text_version); |
6985 | } else | |
6986 | count++; | |
6987 | } | |
6988 | sysfs_free(head); | |
6989 | ||
6990 | } else { | |
1ade5cc1 | 6991 | dprintf("No valid super block on device list: err: %d %p\n", |
ca9de185 LM |
6992 | err, st->sb); |
6993 | } | |
6994 | } else { | |
1ade5cc1 | 6995 | dprintf("no more devices to examine\n"); |
ca9de185 LM |
6996 | } |
6997 | ||
6998 | for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { | |
089f9d79 | 6999 | if (tmpdev->used == 1 && tmpdev->found) { |
ca9de185 LM |
7000 | if (count) { |
7001 | if (count < tmpdev->found) | |
7002 | count = 0; | |
7003 | else | |
7004 | count -= tmpdev->found; | |
7005 | } | |
7006 | } | |
7007 | if (tmpdev->used == 1) | |
7008 | tmpdev->used = 4; | |
7009 | } | |
7010 | err_1: | |
7011 | if (st) | |
7012 | st->ss->free_super(st); | |
7013 | return count; | |
7014 | } | |
7015 | ||
d3c11416 AO |
7016 | static int __count_volumes(char *hba_path, int dpa, int verbose, |
7017 | int cmp_hba_path) | |
ca9de185 | 7018 | { |
72a45777 | 7019 | struct sys_dev *idev, *intel_devices = find_intel_devices(); |
ca9de185 | 7020 | int count = 0; |
72a45777 PB |
7021 | const struct orom_entry *entry; |
7022 | struct devid_list *dv, *devid_list; | |
ca9de185 | 7023 | |
d3c11416 | 7024 | if (!hba_path) |
ca9de185 LM |
7025 | return 0; |
7026 | ||
72a45777 | 7027 | for (idev = intel_devices; idev; idev = idev->next) { |
d3c11416 AO |
7028 | if (strstr(idev->path, hba_path)) |
7029 | break; | |
72a45777 PB |
7030 | } |
7031 | ||
7032 | if (!idev || !idev->dev_id) | |
ca9de185 | 7033 | return 0; |
72a45777 PB |
7034 | |
7035 | entry = get_orom_entry_by_device_id(idev->dev_id); | |
7036 | ||
7037 | if (!entry || !entry->devid_list) | |
7038 | return 0; | |
7039 | ||
7040 | devid_list = entry->devid_list; | |
7041 | for (dv = devid_list; dv; dv = dv->next) { | |
594dc1b8 | 7042 | struct md_list *devlist; |
d3c11416 AO |
7043 | struct sys_dev *device = NULL; |
7044 | char *hpath; | |
72a45777 PB |
7045 | int found = 0; |
7046 | ||
d3c11416 AO |
7047 | if (cmp_hba_path) |
7048 | device = device_by_id_and_path(dv->devid, hba_path); | |
7049 | else | |
7050 | device = device_by_id(dv->devid); | |
7051 | ||
72a45777 | 7052 | if (device) |
d3c11416 | 7053 | hpath = device->path; |
72a45777 PB |
7054 | else |
7055 | return 0; | |
7056 | ||
d3c11416 | 7057 | devlist = get_devices(hpath); |
72a45777 PB |
7058 | /* if no intel devices return zero volumes */ |
7059 | if (devlist == NULL) | |
7060 | return 0; | |
7061 | ||
d3c11416 AO |
7062 | count += active_arrays_by_format("imsm", hpath, &devlist, dpa, |
7063 | verbose); | |
7064 | dprintf("path: %s active arrays: %d\n", hpath, count); | |
72a45777 PB |
7065 | if (devlist == NULL) |
7066 | return 0; | |
7067 | do { | |
7068 | found = 0; | |
7069 | count += count_volumes_list(devlist, | |
7070 | NULL, | |
7071 | verbose, | |
7072 | &found); | |
7073 | dprintf("found %d count: %d\n", found, count); | |
7074 | } while (found); | |
7075 | ||
d3c11416 | 7076 | dprintf("path: %s total number of volumes: %d\n", hpath, count); |
72a45777 PB |
7077 | |
7078 | while (devlist) { | |
7079 | struct md_list *dv = devlist; | |
7080 | devlist = devlist->next; | |
7081 | free(dv->devname); | |
7082 | free(dv); | |
7083 | } | |
ca9de185 LM |
7084 | } |
7085 | return count; | |
7086 | } | |
7087 | ||
d3c11416 AO |
7088 | static int count_volumes(struct intel_hba *hba, int dpa, int verbose) |
7089 | { | |
7090 | if (!hba) | |
7091 | return 0; | |
7092 | if (hba->type == SYS_DEV_VMD) { | |
7093 | struct sys_dev *dev; | |
7094 | int count = 0; | |
7095 | ||
7096 | for (dev = find_intel_devices(); dev; dev = dev->next) { | |
7097 | if (dev->type == SYS_DEV_VMD) | |
7098 | count += __count_volumes(dev->path, dpa, | |
7099 | verbose, 1); | |
7100 | } | |
7101 | return count; | |
7102 | } | |
7103 | return __count_volumes(hba->path, dpa, verbose, 0); | |
7104 | } | |
7105 | ||
cd9d1ac7 DW |
7106 | static int imsm_default_chunk(const struct imsm_orom *orom) |
7107 | { | |
7108 | /* up to 512 if the plaform supports it, otherwise the platform max. | |
7109 | * 128 if no platform detected | |
7110 | */ | |
7111 | int fs = max(7, orom ? fls(orom->sss) : 0); | |
7112 | ||
7113 | return min(512, (1 << fs)); | |
7114 | } | |
73408129 | 7115 | |
6592ce37 DW |
7116 | static int |
7117 | validate_geometry_imsm_orom(struct intel_super *super, int level, int layout, | |
2cc699af | 7118 | int raiddisks, int *chunk, unsigned long long size, int verbose) |
6592ce37 | 7119 | { |
660260d0 DW |
7120 | /* check/set platform and metadata limits/defaults */ |
7121 | if (super->orom && raiddisks > super->orom->dpa) { | |
676e87a8 | 7122 | pr_vrb("platform supports a maximum of %d disks per array\n", |
660260d0 | 7123 | super->orom->dpa); |
73408129 LM |
7124 | return 0; |
7125 | } | |
7126 | ||
5d500228 | 7127 | /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */ |
660260d0 | 7128 | if (!is_raid_level_supported(super->orom, level, raiddisks)) { |
676e87a8 | 7129 | pr_vrb("platform does not support raid%d with %d disk%s\n", |
6592ce37 DW |
7130 | level, raiddisks, raiddisks > 1 ? "s" : ""); |
7131 | return 0; | |
7132 | } | |
cd9d1ac7 | 7133 | |
7ccc4cc4 | 7134 | if (*chunk == 0 || *chunk == UnSet) |
cd9d1ac7 DW |
7135 | *chunk = imsm_default_chunk(super->orom); |
7136 | ||
7ccc4cc4 | 7137 | if (super->orom && !imsm_orom_has_chunk(super->orom, *chunk)) { |
676e87a8 | 7138 | pr_vrb("platform does not support a chunk size of: %d\n", *chunk); |
cd9d1ac7 | 7139 | return 0; |
6592ce37 | 7140 | } |
cd9d1ac7 | 7141 | |
6592ce37 DW |
7142 | if (layout != imsm_level_to_layout(level)) { |
7143 | if (level == 5) | |
676e87a8 | 7144 | pr_vrb("imsm raid 5 only supports the left-asymmetric layout\n"); |
6592ce37 | 7145 | else if (level == 10) |
676e87a8 | 7146 | pr_vrb("imsm raid 10 only supports the n2 layout\n"); |
6592ce37 | 7147 | else |
676e87a8 | 7148 | pr_vrb("imsm unknown layout %#x for this raid level %d\n", |
6592ce37 DW |
7149 | layout, level); |
7150 | return 0; | |
7151 | } | |
2cc699af | 7152 | |
7ccc4cc4 | 7153 | if (super->orom && (super->orom->attr & IMSM_OROM_ATTR_2TB) == 0 && |
2cc699af | 7154 | (calc_array_size(level, raiddisks, layout, *chunk, size) >> 32) > 0) { |
676e87a8 | 7155 | pr_vrb("platform does not support a volume size over 2TB\n"); |
2cc699af CA |
7156 | return 0; |
7157 | } | |
614902f6 | 7158 | |
6592ce37 DW |
7159 | return 1; |
7160 | } | |
7161 | ||
1011e834 | 7162 | /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd |
c2c087e6 DW |
7163 | * FIX ME add ahci details |
7164 | */ | |
8b353278 | 7165 | static int validate_geometry_imsm_volume(struct supertype *st, int level, |
c21e737b | 7166 | int layout, int raiddisks, int *chunk, |
af4348dd N |
7167 | unsigned long long size, |
7168 | unsigned long long data_offset, | |
7169 | char *dev, | |
2c514b71 NB |
7170 | unsigned long long *freesize, |
7171 | int verbose) | |
cdddbdbc | 7172 | { |
9e04ac1c | 7173 | dev_t rdev; |
c2c087e6 | 7174 | struct intel_super *super = st->sb; |
b2916f25 | 7175 | struct imsm_super *mpb; |
c2c087e6 DW |
7176 | struct dl *dl; |
7177 | unsigned long long pos = 0; | |
7178 | unsigned long long maxsize; | |
7179 | struct extent *e; | |
7180 | int i; | |
cdddbdbc | 7181 | |
88c32bb1 DW |
7182 | /* We must have the container info already read in. */ |
7183 | if (!super) | |
c2c087e6 DW |
7184 | return 0; |
7185 | ||
b2916f25 JS |
7186 | mpb = super->anchor; |
7187 | ||
2cc699af | 7188 | if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, size, verbose)) { |
3e684231 | 7189 | pr_err("RAID geometry validation failed. Cannot proceed with the action(s).\n"); |
c2c087e6 | 7190 | return 0; |
d54559f0 | 7191 | } |
c2c087e6 DW |
7192 | if (!dev) { |
7193 | /* General test: make sure there is space for | |
2da8544a DW |
7194 | * 'raiddisks' device extents of size 'size' at a given |
7195 | * offset | |
c2c087e6 | 7196 | */ |
e46273eb | 7197 | unsigned long long minsize = size; |
b7528a20 | 7198 | unsigned long long start_offset = MaxSector; |
c2c087e6 DW |
7199 | int dcnt = 0; |
7200 | if (minsize == 0) | |
7201 | minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
7202 | for (dl = super->disks; dl ; dl = dl->next) { | |
7203 | int found = 0; | |
7204 | ||
bf5a934a | 7205 | pos = 0; |
c2c087e6 | 7206 | i = 0; |
05501181 | 7207 | e = get_extents(super, dl, 0); |
c2c087e6 DW |
7208 | if (!e) continue; |
7209 | do { | |
7210 | unsigned long long esize; | |
7211 | esize = e[i].start - pos; | |
7212 | if (esize >= minsize) | |
7213 | found = 1; | |
b7528a20 | 7214 | if (found && start_offset == MaxSector) { |
2da8544a DW |
7215 | start_offset = pos; |
7216 | break; | |
7217 | } else if (found && pos != start_offset) { | |
7218 | found = 0; | |
7219 | break; | |
7220 | } | |
c2c087e6 DW |
7221 | pos = e[i].start + e[i].size; |
7222 | i++; | |
7223 | } while (e[i-1].size); | |
7224 | if (found) | |
7225 | dcnt++; | |
7226 | free(e); | |
7227 | } | |
7228 | if (dcnt < raiddisks) { | |
2c514b71 | 7229 | if (verbose) |
7a862a02 | 7230 | pr_err("imsm: Not enough devices with space for this array (%d < %d)\n", |
2c514b71 | 7231 | dcnt, raiddisks); |
c2c087e6 DW |
7232 | return 0; |
7233 | } | |
7234 | return 1; | |
7235 | } | |
0dcecb2e | 7236 | |
c2c087e6 | 7237 | /* This device must be a member of the set */ |
9e04ac1c | 7238 | if (!stat_is_blkdev(dev, &rdev)) |
c2c087e6 DW |
7239 | return 0; |
7240 | for (dl = super->disks ; dl ; dl = dl->next) { | |
9e04ac1c ZL |
7241 | if (dl->major == (int)major(rdev) && |
7242 | dl->minor == (int)minor(rdev)) | |
c2c087e6 DW |
7243 | break; |
7244 | } | |
7245 | if (!dl) { | |
2c514b71 | 7246 | if (verbose) |
7a862a02 | 7247 | pr_err("%s is not in the same imsm set\n", dev); |
c2c087e6 | 7248 | return 0; |
a20d2ba5 DW |
7249 | } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) { |
7250 | /* If a volume is present then the current creation attempt | |
7251 | * cannot incorporate new spares because the orom may not | |
7252 | * understand this configuration (all member disks must be | |
7253 | * members of each array in the container). | |
7254 | */ | |
7a862a02 N |
7255 | pr_err("%s is a spare and a volume is already defined for this container\n", dev); |
7256 | pr_err("The option-rom requires all member disks to be a member of all volumes\n"); | |
a20d2ba5 | 7257 | return 0; |
5fe62b94 WD |
7258 | } else if (super->orom && mpb->num_raid_devs > 0 && |
7259 | mpb->num_disks != raiddisks) { | |
7a862a02 | 7260 | pr_err("The option-rom requires all member disks to be a member of all volumes\n"); |
5fe62b94 | 7261 | return 0; |
c2c087e6 | 7262 | } |
0dcecb2e DW |
7263 | |
7264 | /* retrieve the largest free space block */ | |
05501181 | 7265 | e = get_extents(super, dl, 0); |
c2c087e6 DW |
7266 | maxsize = 0; |
7267 | i = 0; | |
0dcecb2e DW |
7268 | if (e) { |
7269 | do { | |
7270 | unsigned long long esize; | |
7271 | ||
7272 | esize = e[i].start - pos; | |
7273 | if (esize >= maxsize) | |
7274 | maxsize = esize; | |
7275 | pos = e[i].start + e[i].size; | |
7276 | i++; | |
7277 | } while (e[i-1].size); | |
7278 | dl->e = e; | |
7279 | dl->extent_cnt = i; | |
7280 | } else { | |
7281 | if (verbose) | |
e7b84f9d | 7282 | pr_err("unable to determine free space for: %s\n", |
0dcecb2e DW |
7283 | dev); |
7284 | return 0; | |
7285 | } | |
7286 | if (maxsize < size) { | |
7287 | if (verbose) | |
e7b84f9d | 7288 | pr_err("%s not enough space (%llu < %llu)\n", |
0dcecb2e DW |
7289 | dev, maxsize, size); |
7290 | return 0; | |
7291 | } | |
7292 | ||
7293 | /* count total number of extents for merge */ | |
7294 | i = 0; | |
7295 | for (dl = super->disks; dl; dl = dl->next) | |
7296 | if (dl->e) | |
7297 | i += dl->extent_cnt; | |
7298 | ||
7299 | maxsize = merge_extents(super, i); | |
3baa56ab LO |
7300 | |
7301 | if (!check_env("IMSM_NO_PLATFORM") && | |
7302 | mpb->num_raid_devs > 0 && size && size != maxsize) { | |
7a862a02 | 7303 | pr_err("attempting to create a second volume with size less then remaining space. Aborting...\n"); |
3baa56ab LO |
7304 | return 0; |
7305 | } | |
7306 | ||
a7dd165b | 7307 | if (maxsize < size || maxsize == 0) { |
b3071342 LD |
7308 | if (verbose) { |
7309 | if (maxsize == 0) | |
7a862a02 | 7310 | pr_err("no free space left on device. Aborting...\n"); |
b3071342 | 7311 | else |
7a862a02 | 7312 | pr_err("not enough space to create volume of given size (%llu < %llu). Aborting...\n", |
b3071342 LD |
7313 | maxsize, size); |
7314 | } | |
0dcecb2e | 7315 | return 0; |
0dcecb2e DW |
7316 | } |
7317 | ||
c2c087e6 DW |
7318 | *freesize = maxsize; |
7319 | ||
ca9de185 | 7320 | if (super->orom) { |
72a45777 | 7321 | int count = count_volumes(super->hba, |
ca9de185 LM |
7322 | super->orom->dpa, verbose); |
7323 | if (super->orom->vphba <= count) { | |
676e87a8 | 7324 | pr_vrb("platform does not support more than %d raid volumes.\n", |
ca9de185 LM |
7325 | super->orom->vphba); |
7326 | return 0; | |
7327 | } | |
7328 | } | |
c2c087e6 | 7329 | return 1; |
cdddbdbc DW |
7330 | } |
7331 | ||
13bcac90 | 7332 | static int imsm_get_free_size(struct supertype *st, int raiddisks, |
efb30e7f DW |
7333 | unsigned long long size, int chunk, |
7334 | unsigned long long *freesize) | |
7335 | { | |
7336 | struct intel_super *super = st->sb; | |
7337 | struct imsm_super *mpb = super->anchor; | |
7338 | struct dl *dl; | |
7339 | int i; | |
7340 | int extent_cnt; | |
7341 | struct extent *e; | |
7342 | unsigned long long maxsize; | |
7343 | unsigned long long minsize; | |
7344 | int cnt; | |
7345 | int used; | |
7346 | ||
7347 | /* find the largest common start free region of the possible disks */ | |
7348 | used = 0; | |
7349 | extent_cnt = 0; | |
7350 | cnt = 0; | |
7351 | for (dl = super->disks; dl; dl = dl->next) { | |
7352 | dl->raiddisk = -1; | |
7353 | ||
7354 | if (dl->index >= 0) | |
7355 | used++; | |
7356 | ||
7357 | /* don't activate new spares if we are orom constrained | |
7358 | * and there is already a volume active in the container | |
7359 | */ | |
7360 | if (super->orom && dl->index < 0 && mpb->num_raid_devs) | |
7361 | continue; | |
7362 | ||
05501181 | 7363 | e = get_extents(super, dl, 0); |
efb30e7f DW |
7364 | if (!e) |
7365 | continue; | |
7366 | for (i = 1; e[i-1].size; i++) | |
7367 | ; | |
7368 | dl->e = e; | |
7369 | dl->extent_cnt = i; | |
7370 | extent_cnt += i; | |
7371 | cnt++; | |
7372 | } | |
7373 | ||
7374 | maxsize = merge_extents(super, extent_cnt); | |
7375 | minsize = size; | |
7376 | if (size == 0) | |
612e59d8 CA |
7377 | /* chunk is in K */ |
7378 | minsize = chunk * 2; | |
efb30e7f DW |
7379 | |
7380 | if (cnt < raiddisks || | |
7381 | (super->orom && used && used != raiddisks) || | |
a7dd165b DW |
7382 | maxsize < minsize || |
7383 | maxsize == 0) { | |
e7b84f9d | 7384 | pr_err("not enough devices with space to create array.\n"); |
efb30e7f DW |
7385 | return 0; /* No enough free spaces large enough */ |
7386 | } | |
7387 | ||
7388 | if (size == 0) { | |
7389 | size = maxsize; | |
7390 | if (chunk) { | |
612e59d8 CA |
7391 | size /= 2 * chunk; |
7392 | size *= 2 * chunk; | |
efb30e7f | 7393 | } |
f878b242 LM |
7394 | maxsize = size; |
7395 | } | |
7396 | if (!check_env("IMSM_NO_PLATFORM") && | |
7397 | mpb->num_raid_devs > 0 && size && size != maxsize) { | |
7a862a02 | 7398 | pr_err("attempting to create a second volume with size less then remaining space. Aborting...\n"); |
f878b242 | 7399 | return 0; |
efb30e7f | 7400 | } |
efb30e7f DW |
7401 | cnt = 0; |
7402 | for (dl = super->disks; dl; dl = dl->next) | |
7403 | if (dl->e) | |
7404 | dl->raiddisk = cnt++; | |
7405 | ||
7406 | *freesize = size; | |
7407 | ||
13bcac90 AK |
7408 | dprintf("imsm: imsm_get_free_size() returns : %llu\n", size); |
7409 | ||
efb30e7f DW |
7410 | return 1; |
7411 | } | |
7412 | ||
13bcac90 AK |
7413 | static int reserve_space(struct supertype *st, int raiddisks, |
7414 | unsigned long long size, int chunk, | |
7415 | unsigned long long *freesize) | |
7416 | { | |
7417 | struct intel_super *super = st->sb; | |
7418 | struct dl *dl; | |
7419 | int cnt; | |
7420 | int rv = 0; | |
7421 | ||
7422 | rv = imsm_get_free_size(st, raiddisks, size, chunk, freesize); | |
7423 | if (rv) { | |
7424 | cnt = 0; | |
7425 | for (dl = super->disks; dl; dl = dl->next) | |
7426 | if (dl->e) | |
7427 | dl->raiddisk = cnt++; | |
7428 | rv = 1; | |
7429 | } | |
7430 | ||
7431 | return rv; | |
7432 | } | |
7433 | ||
bf5a934a | 7434 | static int validate_geometry_imsm(struct supertype *st, int level, int layout, |
c21e737b | 7435 | int raiddisks, int *chunk, unsigned long long size, |
af4348dd | 7436 | unsigned long long data_offset, |
bf5a934a | 7437 | char *dev, unsigned long long *freesize, |
5308f117 | 7438 | int consistency_policy, int verbose) |
bf5a934a DW |
7439 | { |
7440 | int fd, cfd; | |
7441 | struct mdinfo *sra; | |
20cbe8d2 | 7442 | int is_member = 0; |
bf5a934a | 7443 | |
d54559f0 LM |
7444 | /* load capability |
7445 | * if given unused devices create a container | |
bf5a934a DW |
7446 | * if given given devices in a container create a member volume |
7447 | */ | |
7448 | if (level == LEVEL_CONTAINER) { | |
7449 | /* Must be a fresh device to add to a container */ | |
7450 | return validate_geometry_imsm_container(st, level, layout, | |
c21e737b | 7451 | raiddisks, |
7ccc4cc4 | 7452 | *chunk, |
af4348dd | 7453 | size, data_offset, |
bf5a934a DW |
7454 | dev, freesize, |
7455 | verbose); | |
7456 | } | |
9587c373 | 7457 | |
22dc741f MT |
7458 | if (size && (size < 1024)) { |
7459 | pr_err("Given size must be greater than 1M.\n"); | |
54865c30 RS |
7460 | /* Depends on algorithm in Create.c : |
7461 | * if container was given (dev == NULL) return -1, | |
7462 | * if block device was given ( dev != NULL) return 0. | |
7463 | */ | |
7464 | return dev ? -1 : 0; | |
7465 | } | |
7466 | ||
8592f29d | 7467 | if (!dev) { |
e91a3bad | 7468 | if (st->sb) { |
ca9de185 | 7469 | struct intel_super *super = st->sb; |
e91a3bad | 7470 | if (!validate_geometry_imsm_orom(st->sb, level, layout, |
2cc699af | 7471 | raiddisks, chunk, size, |
e91a3bad LM |
7472 | verbose)) |
7473 | return 0; | |
efb30e7f DW |
7474 | /* we are being asked to automatically layout a |
7475 | * new volume based on the current contents of | |
7476 | * the container. If the the parameters can be | |
7477 | * satisfied reserve_space will record the disks, | |
7478 | * start offset, and size of the volume to be | |
7479 | * created. add_to_super and getinfo_super | |
7480 | * detect when autolayout is in progress. | |
7481 | */ | |
ca9de185 LM |
7482 | /* assuming that freesize is always given when array is |
7483 | created */ | |
7484 | if (super->orom && freesize) { | |
7485 | int count; | |
72a45777 | 7486 | count = count_volumes(super->hba, |
ca9de185 LM |
7487 | super->orom->dpa, verbose); |
7488 | if (super->orom->vphba <= count) { | |
676e87a8 | 7489 | pr_vrb("platform does not support more than %d raid volumes.\n", |
ca9de185 LM |
7490 | super->orom->vphba); |
7491 | return 0; | |
7492 | } | |
7493 | } | |
e91a3bad LM |
7494 | if (freesize) |
7495 | return reserve_space(st, raiddisks, size, | |
7ccc4cc4 | 7496 | *chunk, freesize); |
8592f29d N |
7497 | } |
7498 | return 1; | |
7499 | } | |
bf5a934a DW |
7500 | if (st->sb) { |
7501 | /* creating in a given container */ | |
7502 | return validate_geometry_imsm_volume(st, level, layout, | |
7503 | raiddisks, chunk, size, | |
af4348dd | 7504 | data_offset, |
bf5a934a DW |
7505 | dev, freesize, verbose); |
7506 | } | |
7507 | ||
bf5a934a DW |
7508 | /* This device needs to be a device in an 'imsm' container */ |
7509 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
7510 | if (fd >= 0) { | |
7511 | if (verbose) | |
e7b84f9d N |
7512 | pr_err("Cannot create this array on device %s\n", |
7513 | dev); | |
bf5a934a DW |
7514 | close(fd); |
7515 | return 0; | |
7516 | } | |
7517 | if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) { | |
7518 | if (verbose) | |
e7b84f9d | 7519 | pr_err("Cannot open %s: %s\n", |
bf5a934a DW |
7520 | dev, strerror(errno)); |
7521 | return 0; | |
7522 | } | |
7523 | /* Well, it is in use by someone, maybe an 'imsm' container. */ | |
7524 | cfd = open_container(fd); | |
20cbe8d2 | 7525 | close(fd); |
bf5a934a | 7526 | if (cfd < 0) { |
bf5a934a | 7527 | if (verbose) |
e7b84f9d | 7528 | pr_err("Cannot use %s: It is busy\n", |
bf5a934a DW |
7529 | dev); |
7530 | return 0; | |
7531 | } | |
4dd2df09 | 7532 | sra = sysfs_read(cfd, NULL, GET_VERSION); |
bf5a934a | 7533 | if (sra && sra->array.major_version == -1 && |
20cbe8d2 AW |
7534 | strcmp(sra->text_version, "imsm") == 0) |
7535 | is_member = 1; | |
7536 | sysfs_free(sra); | |
7537 | if (is_member) { | |
bf5a934a DW |
7538 | /* This is a member of a imsm container. Load the container |
7539 | * and try to create a volume | |
7540 | */ | |
7541 | struct intel_super *super; | |
7542 | ||
ec50f7b6 | 7543 | if (load_super_imsm_all(st, cfd, (void **) &super, NULL, NULL, 1) == 0) { |
bf5a934a | 7544 | st->sb = super; |
4dd2df09 | 7545 | strcpy(st->container_devnm, fd2devnm(cfd)); |
bf5a934a DW |
7546 | close(cfd); |
7547 | return validate_geometry_imsm_volume(st, level, layout, | |
7548 | raiddisks, chunk, | |
af4348dd | 7549 | size, data_offset, dev, |
ecbd9e81 N |
7550 | freesize, 1) |
7551 | ? 1 : -1; | |
bf5a934a | 7552 | } |
20cbe8d2 | 7553 | } |
bf5a934a | 7554 | |
20cbe8d2 | 7555 | if (verbose) |
e7b84f9d | 7556 | pr_err("failed container membership check\n"); |
20cbe8d2 AW |
7557 | |
7558 | close(cfd); | |
7559 | return 0; | |
bf5a934a | 7560 | } |
0bd16cf2 | 7561 | |
30f58b22 | 7562 | static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk) |
0bd16cf2 DJ |
7563 | { |
7564 | struct intel_super *super = st->sb; | |
7565 | ||
30f58b22 DW |
7566 | if (level && *level == UnSet) |
7567 | *level = LEVEL_CONTAINER; | |
7568 | ||
7569 | if (level && layout && *layout == UnSet) | |
7570 | *layout = imsm_level_to_layout(*level); | |
0bd16cf2 | 7571 | |
cd9d1ac7 DW |
7572 | if (chunk && (*chunk == UnSet || *chunk == 0)) |
7573 | *chunk = imsm_default_chunk(super->orom); | |
0bd16cf2 DJ |
7574 | } |
7575 | ||
33414a01 DW |
7576 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev); |
7577 | ||
7578 | static int kill_subarray_imsm(struct supertype *st) | |
7579 | { | |
7580 | /* remove the subarray currently referenced by ->current_vol */ | |
7581 | __u8 i; | |
7582 | struct intel_dev **dp; | |
7583 | struct intel_super *super = st->sb; | |
7584 | __u8 current_vol = super->current_vol; | |
7585 | struct imsm_super *mpb = super->anchor; | |
7586 | ||
7587 | if (super->current_vol < 0) | |
7588 | return 2; | |
7589 | super->current_vol = -1; /* invalidate subarray cursor */ | |
7590 | ||
7591 | /* block deletions that would change the uuid of active subarrays | |
7592 | * | |
7593 | * FIXME when immutable ids are available, but note that we'll | |
7594 | * also need to fixup the invalidated/active subarray indexes in | |
7595 | * mdstat | |
7596 | */ | |
7597 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
7598 | char subarray[4]; | |
7599 | ||
7600 | if (i < current_vol) | |
7601 | continue; | |
7602 | sprintf(subarray, "%u", i); | |
4dd2df09 | 7603 | if (is_subarray_active(subarray, st->devnm)) { |
e7b84f9d N |
7604 | pr_err("deleting subarray-%d would change the UUID of active subarray-%d, aborting\n", |
7605 | current_vol, i); | |
33414a01 DW |
7606 | |
7607 | return 2; | |
7608 | } | |
7609 | } | |
7610 | ||
7611 | if (st->update_tail) { | |
503975b9 | 7612 | struct imsm_update_kill_array *u = xmalloc(sizeof(*u)); |
33414a01 | 7613 | |
33414a01 DW |
7614 | u->type = update_kill_array; |
7615 | u->dev_idx = current_vol; | |
7616 | append_metadata_update(st, u, sizeof(*u)); | |
7617 | ||
7618 | return 0; | |
7619 | } | |
7620 | ||
7621 | for (dp = &super->devlist; *dp;) | |
7622 | if ((*dp)->index == current_vol) { | |
7623 | *dp = (*dp)->next; | |
7624 | } else { | |
7625 | handle_missing(super, (*dp)->dev); | |
7626 | if ((*dp)->index > current_vol) | |
7627 | (*dp)->index--; | |
7628 | dp = &(*dp)->next; | |
7629 | } | |
7630 | ||
7631 | /* no more raid devices, all active components are now spares, | |
7632 | * but of course failed are still failed | |
7633 | */ | |
7634 | if (--mpb->num_raid_devs == 0) { | |
7635 | struct dl *d; | |
7636 | ||
7637 | for (d = super->disks; d; d = d->next) | |
a8619d23 AK |
7638 | if (d->index > -2) |
7639 | mark_spare(d); | |
33414a01 DW |
7640 | } |
7641 | ||
7642 | super->updates_pending++; | |
7643 | ||
7644 | return 0; | |
7645 | } | |
aa534678 | 7646 | |
a951a4f7 | 7647 | static int update_subarray_imsm(struct supertype *st, char *subarray, |
fa56eddb | 7648 | char *update, struct mddev_ident *ident) |
aa534678 DW |
7649 | { |
7650 | /* update the subarray currently referenced by ->current_vol */ | |
7651 | struct intel_super *super = st->sb; | |
7652 | struct imsm_super *mpb = super->anchor; | |
7653 | ||
aa534678 DW |
7654 | if (strcmp(update, "name") == 0) { |
7655 | char *name = ident->name; | |
a951a4f7 N |
7656 | char *ep; |
7657 | int vol; | |
aa534678 | 7658 | |
4dd2df09 | 7659 | if (is_subarray_active(subarray, st->devnm)) { |
e7b84f9d | 7660 | pr_err("Unable to update name of active subarray\n"); |
aa534678 DW |
7661 | return 2; |
7662 | } | |
7663 | ||
7664 | if (!check_name(super, name, 0)) | |
7665 | return 2; | |
7666 | ||
a951a4f7 N |
7667 | vol = strtoul(subarray, &ep, 10); |
7668 | if (*ep != '\0' || vol >= super->anchor->num_raid_devs) | |
7669 | return 2; | |
7670 | ||
aa534678 | 7671 | if (st->update_tail) { |
503975b9 | 7672 | struct imsm_update_rename_array *u = xmalloc(sizeof(*u)); |
aa534678 | 7673 | |
aa534678 | 7674 | u->type = update_rename_array; |
a951a4f7 | 7675 | u->dev_idx = vol; |
618f4e6d XN |
7676 | strncpy((char *) u->name, name, MAX_RAID_SERIAL_LEN); |
7677 | u->name[MAX_RAID_SERIAL_LEN-1] = '\0'; | |
aa534678 DW |
7678 | append_metadata_update(st, u, sizeof(*u)); |
7679 | } else { | |
7680 | struct imsm_dev *dev; | |
ebad3af2 | 7681 | int i, namelen; |
aa534678 | 7682 | |
a951a4f7 | 7683 | dev = get_imsm_dev(super, vol); |
ebad3af2 JS |
7684 | memset(dev->volume, '\0', MAX_RAID_SERIAL_LEN); |
7685 | namelen = min((int)strlen(name), MAX_RAID_SERIAL_LEN); | |
7686 | memcpy(dev->volume, name, namelen); | |
aa534678 DW |
7687 | for (i = 0; i < mpb->num_raid_devs; i++) { |
7688 | dev = get_imsm_dev(super, i); | |
7689 | handle_missing(super, dev); | |
7690 | } | |
7691 | super->updates_pending++; | |
7692 | } | |
e6e9dd3f AP |
7693 | } else if (strcmp(update, "ppl") == 0 || |
7694 | strcmp(update, "no-ppl") == 0) { | |
7695 | int new_policy; | |
7696 | char *ep; | |
7697 | int vol = strtoul(subarray, &ep, 10); | |
7698 | ||
7699 | if (*ep != '\0' || vol >= super->anchor->num_raid_devs) | |
7700 | return 2; | |
7701 | ||
7702 | if (strcmp(update, "ppl") == 0) | |
c2462068 | 7703 | new_policy = RWH_MULTIPLE_DISTRIBUTED; |
e6e9dd3f | 7704 | else |
c2462068 | 7705 | new_policy = RWH_MULTIPLE_OFF; |
e6e9dd3f AP |
7706 | |
7707 | if (st->update_tail) { | |
7708 | struct imsm_update_rwh_policy *u = xmalloc(sizeof(*u)); | |
7709 | ||
7710 | u->type = update_rwh_policy; | |
7711 | u->dev_idx = vol; | |
7712 | u->new_policy = new_policy; | |
7713 | append_metadata_update(st, u, sizeof(*u)); | |
7714 | } else { | |
7715 | struct imsm_dev *dev; | |
7716 | ||
7717 | dev = get_imsm_dev(super, vol); | |
7718 | dev->rwh_policy = new_policy; | |
7719 | super->updates_pending++; | |
7720 | } | |
aa534678 DW |
7721 | } else |
7722 | return 2; | |
7723 | ||
7724 | return 0; | |
7725 | } | |
bf5a934a | 7726 | |
28bce06f AK |
7727 | static int is_gen_migration(struct imsm_dev *dev) |
7728 | { | |
7534230b AK |
7729 | if (dev == NULL) |
7730 | return 0; | |
7731 | ||
28bce06f AK |
7732 | if (!dev->vol.migr_state) |
7733 | return 0; | |
7734 | ||
7735 | if (migr_type(dev) == MIGR_GEN_MIGR) | |
7736 | return 1; | |
7737 | ||
7738 | return 0; | |
7739 | } | |
7740 | ||
1e5c6983 DW |
7741 | static int is_rebuilding(struct imsm_dev *dev) |
7742 | { | |
7743 | struct imsm_map *migr_map; | |
7744 | ||
7745 | if (!dev->vol.migr_state) | |
7746 | return 0; | |
7747 | ||
7748 | if (migr_type(dev) != MIGR_REBUILD) | |
7749 | return 0; | |
7750 | ||
238c0a71 | 7751 | migr_map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
7752 | |
7753 | if (migr_map->map_state == IMSM_T_STATE_DEGRADED) | |
7754 | return 1; | |
7755 | else | |
7756 | return 0; | |
7757 | } | |
7758 | ||
6ce1fbf1 AK |
7759 | static int is_initializing(struct imsm_dev *dev) |
7760 | { | |
7761 | struct imsm_map *migr_map; | |
7762 | ||
7763 | if (!dev->vol.migr_state) | |
7764 | return 0; | |
7765 | ||
7766 | if (migr_type(dev) != MIGR_INIT) | |
7767 | return 0; | |
7768 | ||
238c0a71 | 7769 | migr_map = get_imsm_map(dev, MAP_1); |
6ce1fbf1 AK |
7770 | |
7771 | if (migr_map->map_state == IMSM_T_STATE_UNINITIALIZED) | |
7772 | return 1; | |
7773 | ||
7774 | return 0; | |
6ce1fbf1 AK |
7775 | } |
7776 | ||
c47b0ff6 AK |
7777 | static void update_recovery_start(struct intel_super *super, |
7778 | struct imsm_dev *dev, | |
7779 | struct mdinfo *array) | |
1e5c6983 DW |
7780 | { |
7781 | struct mdinfo *rebuild = NULL; | |
7782 | struct mdinfo *d; | |
7783 | __u32 units; | |
7784 | ||
7785 | if (!is_rebuilding(dev)) | |
7786 | return; | |
7787 | ||
7788 | /* Find the rebuild target, but punt on the dual rebuild case */ | |
7789 | for (d = array->devs; d; d = d->next) | |
7790 | if (d->recovery_start == 0) { | |
7791 | if (rebuild) | |
7792 | return; | |
7793 | rebuild = d; | |
7794 | } | |
7795 | ||
4363fd80 DW |
7796 | if (!rebuild) { |
7797 | /* (?) none of the disks are marked with | |
7798 | * IMSM_ORD_REBUILD, so assume they are missing and the | |
7799 | * disk_ord_tbl was not correctly updated | |
7800 | */ | |
1ade5cc1 | 7801 | dprintf("failed to locate out-of-sync disk\n"); |
4363fd80 DW |
7802 | return; |
7803 | } | |
7804 | ||
1e5c6983 | 7805 | units = __le32_to_cpu(dev->vol.curr_migr_unit); |
c47b0ff6 | 7806 | rebuild->recovery_start = units * blocks_per_migr_unit(super, dev); |
1e5c6983 DW |
7807 | } |
7808 | ||
276d77db | 7809 | static int recover_backup_imsm(struct supertype *st, struct mdinfo *info); |
1e5c6983 | 7810 | |
00bbdbda | 7811 | static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray) |
cdddbdbc | 7812 | { |
4f5bc454 DW |
7813 | /* Given a container loaded by load_super_imsm_all, |
7814 | * extract information about all the arrays into | |
7815 | * an mdinfo tree. | |
00bbdbda | 7816 | * If 'subarray' is given, just extract info about that array. |
4f5bc454 DW |
7817 | * |
7818 | * For each imsm_dev create an mdinfo, fill it in, | |
7819 | * then look for matching devices in super->disks | |
7820 | * and create appropriate device mdinfo. | |
7821 | */ | |
7822 | struct intel_super *super = st->sb; | |
949c47a0 | 7823 | struct imsm_super *mpb = super->anchor; |
4f5bc454 | 7824 | struct mdinfo *rest = NULL; |
00bbdbda | 7825 | unsigned int i; |
81219e70 | 7826 | int sb_errors = 0; |
abef11a3 AK |
7827 | struct dl *d; |
7828 | int spare_disks = 0; | |
cdddbdbc | 7829 | |
19482bcc AK |
7830 | /* do not assemble arrays when not all attributes are supported */ |
7831 | if (imsm_check_attributes(mpb->attributes) == 0) { | |
81219e70 | 7832 | sb_errors = 1; |
7a862a02 | 7833 | pr_err("Unsupported attributes in IMSM metadata.Arrays activation is blocked.\n"); |
19482bcc AK |
7834 | } |
7835 | ||
abef11a3 AK |
7836 | /* count spare devices, not used in maps |
7837 | */ | |
7838 | for (d = super->disks; d; d = d->next) | |
7839 | if (d->index == -1) | |
7840 | spare_disks++; | |
7841 | ||
4f5bc454 | 7842 | for (i = 0; i < mpb->num_raid_devs; i++) { |
00bbdbda N |
7843 | struct imsm_dev *dev; |
7844 | struct imsm_map *map; | |
86e3692b | 7845 | struct imsm_map *map2; |
4f5bc454 | 7846 | struct mdinfo *this; |
a6482415 | 7847 | int slot; |
a6482415 | 7848 | int chunk; |
00bbdbda | 7849 | char *ep; |
8b9cd157 | 7850 | int level; |
00bbdbda N |
7851 | |
7852 | if (subarray && | |
7853 | (i != strtoul(subarray, &ep, 10) || *ep != '\0')) | |
7854 | continue; | |
7855 | ||
7856 | dev = get_imsm_dev(super, i); | |
238c0a71 AK |
7857 | map = get_imsm_map(dev, MAP_0); |
7858 | map2 = get_imsm_map(dev, MAP_1); | |
8b9cd157 | 7859 | level = get_imsm_raid_level(map); |
4f5bc454 | 7860 | |
1ce0101c DW |
7861 | /* do not publish arrays that are in the middle of an |
7862 | * unsupported migration | |
7863 | */ | |
7864 | if (dev->vol.migr_state && | |
28bce06f | 7865 | (migr_type(dev) == MIGR_STATE_CHANGE)) { |
7a862a02 | 7866 | pr_err("cannot assemble volume '%.16s': unsupported migration in progress\n", |
1ce0101c DW |
7867 | dev->volume); |
7868 | continue; | |
7869 | } | |
2db86302 LM |
7870 | /* do not publish arrays that are not support by controller's |
7871 | * OROM/EFI | |
7872 | */ | |
1ce0101c | 7873 | |
503975b9 | 7874 | this = xmalloc(sizeof(*this)); |
4f5bc454 | 7875 | |
301406c9 | 7876 | super->current_vol = i; |
a5d85af7 | 7877 | getinfo_super_imsm_volume(st, this, NULL); |
9894ec0d | 7878 | this->next = rest; |
a6482415 | 7879 | chunk = __le16_to_cpu(map->blocks_per_strip) >> 1; |
81219e70 LM |
7880 | /* mdadm does not support all metadata features- set the bit in all arrays state */ |
7881 | if (!validate_geometry_imsm_orom(super, | |
8b9cd157 MK |
7882 | level, /* RAID level */ |
7883 | imsm_level_to_layout(level), | |
81219e70 | 7884 | map->num_members, /* raid disks */ |
fcc2c9da | 7885 | &chunk, imsm_dev_size(dev), |
81219e70 | 7886 | 1 /* verbose */)) { |
7a862a02 | 7887 | pr_err("IMSM RAID geometry validation failed. Array %s activation is blocked.\n", |
81219e70 LM |
7888 | dev->volume); |
7889 | this->array.state |= | |
7890 | (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) | | |
7891 | (1<<MD_SB_BLOCK_VOLUME); | |
7892 | } | |
81219e70 LM |
7893 | |
7894 | /* if array has bad blocks, set suitable bit in all arrays state */ | |
7895 | if (sb_errors) | |
7896 | this->array.state |= | |
7897 | (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) | | |
7898 | (1<<MD_SB_BLOCK_VOLUME); | |
7899 | ||
4f5bc454 | 7900 | for (slot = 0 ; slot < map->num_members; slot++) { |
1e5c6983 | 7901 | unsigned long long recovery_start; |
4f5bc454 DW |
7902 | struct mdinfo *info_d; |
7903 | struct dl *d; | |
7904 | int idx; | |
9a1608e5 | 7905 | int skip; |
7eef0453 | 7906 | __u32 ord; |
8b9cd157 | 7907 | int missing = 0; |
4f5bc454 | 7908 | |
9a1608e5 | 7909 | skip = 0; |
238c0a71 AK |
7910 | idx = get_imsm_disk_idx(dev, slot, MAP_0); |
7911 | ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); | |
4f5bc454 DW |
7912 | for (d = super->disks; d ; d = d->next) |
7913 | if (d->index == idx) | |
0fbd635c | 7914 | break; |
4f5bc454 | 7915 | |
1e5c6983 | 7916 | recovery_start = MaxSector; |
4f5bc454 | 7917 | if (d == NULL) |
9a1608e5 | 7918 | skip = 1; |
25ed7e59 | 7919 | if (d && is_failed(&d->disk)) |
9a1608e5 | 7920 | skip = 1; |
8b9cd157 | 7921 | if (!skip && (ord & IMSM_ORD_REBUILD)) |
1e5c6983 | 7922 | recovery_start = 0; |
9a1608e5 | 7923 | |
1011e834 | 7924 | /* |
9a1608e5 | 7925 | * if we skip some disks the array will be assmebled degraded; |
1e5c6983 DW |
7926 | * reset resync start to avoid a dirty-degraded |
7927 | * situation when performing the intial sync | |
9a1608e5 | 7928 | */ |
8b9cd157 MK |
7929 | if (skip) |
7930 | missing++; | |
7931 | ||
7932 | if (!(dev->vol.dirty & RAIDVOL_DIRTY)) { | |
7933 | if ((!able_to_resync(level, missing) || | |
7934 | recovery_start == 0)) | |
7935 | this->resync_start = MaxSector; | |
7936 | } else { | |
7937 | /* | |
7938 | * FIXME handle dirty degraded | |
7939 | */ | |
7940 | } | |
7941 | ||
9a1608e5 DW |
7942 | if (skip) |
7943 | continue; | |
4f5bc454 | 7944 | |
503975b9 | 7945 | info_d = xcalloc(1, sizeof(*info_d)); |
4f5bc454 DW |
7946 | info_d->next = this->devs; |
7947 | this->devs = info_d; | |
7948 | ||
4f5bc454 DW |
7949 | info_d->disk.number = d->index; |
7950 | info_d->disk.major = d->major; | |
7951 | info_d->disk.minor = d->minor; | |
7952 | info_d->disk.raid_disk = slot; | |
1e5c6983 | 7953 | info_d->recovery_start = recovery_start; |
86e3692b AK |
7954 | if (map2) { |
7955 | if (slot < map2->num_members) | |
7956 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
7957 | else |
7958 | this->array.spare_disks++; | |
86e3692b AK |
7959 | } else { |
7960 | if (slot < map->num_members) | |
7961 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
7962 | else |
7963 | this->array.spare_disks++; | |
86e3692b | 7964 | } |
1e5c6983 DW |
7965 | if (info_d->recovery_start == MaxSector) |
7966 | this->array.working_disks++; | |
4f5bc454 DW |
7967 | |
7968 | info_d->events = __le32_to_cpu(mpb->generation_num); | |
5551b113 | 7969 | info_d->data_offset = pba_of_lba0(map); |
44490938 | 7970 | info_d->component_size = calc_component_size(map, dev); |
06fb291a PB |
7971 | |
7972 | if (map->raid_level == 5) { | |
2432ce9b AP |
7973 | info_d->ppl_sector = this->ppl_sector; |
7974 | info_d->ppl_size = this->ppl_size; | |
98e96bdb AP |
7975 | if (this->consistency_policy == CONSISTENCY_POLICY_PPL && |
7976 | recovery_start == 0) | |
7977 | this->resync_start = 0; | |
06fb291a | 7978 | } |
b12796be | 7979 | |
5e46202e | 7980 | info_d->bb.supported = 1; |
b12796be TM |
7981 | get_volume_badblocks(super->bbm_log, ord_to_idx(ord), |
7982 | info_d->data_offset, | |
7983 | info_d->component_size, | |
7984 | &info_d->bb); | |
4f5bc454 | 7985 | } |
1e5c6983 | 7986 | /* now that the disk list is up-to-date fixup recovery_start */ |
c47b0ff6 | 7987 | update_recovery_start(super, dev, this); |
abef11a3 | 7988 | this->array.spare_disks += spare_disks; |
276d77db AK |
7989 | |
7990 | /* check for reshape */ | |
7991 | if (this->reshape_active == 1) | |
7992 | recover_backup_imsm(st, this); | |
9a1608e5 | 7993 | rest = this; |
4f5bc454 DW |
7994 | } |
7995 | ||
7996 | return rest; | |
cdddbdbc DW |
7997 | } |
7998 | ||
3b451610 AK |
7999 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, |
8000 | int failed, int look_in_map) | |
c2a1e7da | 8001 | { |
3b451610 AK |
8002 | struct imsm_map *map; |
8003 | ||
8004 | map = get_imsm_map(dev, look_in_map); | |
c2a1e7da DW |
8005 | |
8006 | if (!failed) | |
1011e834 | 8007 | return map->map_state == IMSM_T_STATE_UNINITIALIZED ? |
3393c6af | 8008 | IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL; |
c2a1e7da DW |
8009 | |
8010 | switch (get_imsm_raid_level(map)) { | |
8011 | case 0: | |
8012 | return IMSM_T_STATE_FAILED; | |
8013 | break; | |
8014 | case 1: | |
8015 | if (failed < map->num_members) | |
8016 | return IMSM_T_STATE_DEGRADED; | |
8017 | else | |
8018 | return IMSM_T_STATE_FAILED; | |
8019 | break; | |
8020 | case 10: | |
8021 | { | |
8022 | /** | |
c92a2527 DW |
8023 | * check to see if any mirrors have failed, otherwise we |
8024 | * are degraded. Even numbered slots are mirrored on | |
8025 | * slot+1 | |
c2a1e7da | 8026 | */ |
c2a1e7da | 8027 | int i; |
d9b420a5 N |
8028 | /* gcc -Os complains that this is unused */ |
8029 | int insync = insync; | |
c2a1e7da DW |
8030 | |
8031 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 8032 | __u32 ord = get_imsm_ord_tbl_ent(dev, i, MAP_X); |
c92a2527 DW |
8033 | int idx = ord_to_idx(ord); |
8034 | struct imsm_disk *disk; | |
c2a1e7da | 8035 | |
c92a2527 | 8036 | /* reset the potential in-sync count on even-numbered |
1011e834 | 8037 | * slots. num_copies is always 2 for imsm raid10 |
c92a2527 DW |
8038 | */ |
8039 | if ((i & 1) == 0) | |
8040 | insync = 2; | |
c2a1e7da | 8041 | |
c92a2527 | 8042 | disk = get_imsm_disk(super, idx); |
25ed7e59 | 8043 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) |
c92a2527 | 8044 | insync--; |
c2a1e7da | 8045 | |
c92a2527 DW |
8046 | /* no in-sync disks left in this mirror the |
8047 | * array has failed | |
8048 | */ | |
8049 | if (insync == 0) | |
8050 | return IMSM_T_STATE_FAILED; | |
c2a1e7da DW |
8051 | } |
8052 | ||
8053 | return IMSM_T_STATE_DEGRADED; | |
8054 | } | |
8055 | case 5: | |
8056 | if (failed < 2) | |
8057 | return IMSM_T_STATE_DEGRADED; | |
8058 | else | |
8059 | return IMSM_T_STATE_FAILED; | |
8060 | break; | |
8061 | default: | |
8062 | break; | |
8063 | } | |
8064 | ||
8065 | return map->map_state; | |
8066 | } | |
8067 | ||
3b451610 AK |
8068 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev, |
8069 | int look_in_map) | |
c2a1e7da DW |
8070 | { |
8071 | int i; | |
8072 | int failed = 0; | |
8073 | struct imsm_disk *disk; | |
d5985138 AK |
8074 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
8075 | struct imsm_map *prev = get_imsm_map(dev, MAP_1); | |
68fe4598 | 8076 | struct imsm_map *map_for_loop; |
0556e1a2 DW |
8077 | __u32 ord; |
8078 | int idx; | |
d5985138 | 8079 | int idx_1; |
c2a1e7da | 8080 | |
0556e1a2 DW |
8081 | /* at the beginning of migration we set IMSM_ORD_REBUILD on |
8082 | * disks that are being rebuilt. New failures are recorded to | |
8083 | * map[0]. So we look through all the disks we started with and | |
8084 | * see if any failures are still present, or if any new ones | |
8085 | * have arrived | |
0556e1a2 | 8086 | */ |
d5985138 AK |
8087 | map_for_loop = map; |
8088 | if (prev && (map->num_members < prev->num_members)) | |
8089 | map_for_loop = prev; | |
68fe4598 LD |
8090 | |
8091 | for (i = 0; i < map_for_loop->num_members; i++) { | |
d5985138 | 8092 | idx_1 = -255; |
238c0a71 AK |
8093 | /* when MAP_X is passed both maps failures are counted |
8094 | */ | |
d5985138 | 8095 | if (prev && |
089f9d79 JS |
8096 | (look_in_map == MAP_1 || look_in_map == MAP_X) && |
8097 | i < prev->num_members) { | |
d5985138 AK |
8098 | ord = __le32_to_cpu(prev->disk_ord_tbl[i]); |
8099 | idx_1 = ord_to_idx(ord); | |
c2a1e7da | 8100 | |
d5985138 AK |
8101 | disk = get_imsm_disk(super, idx_1); |
8102 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) | |
8103 | failed++; | |
8104 | } | |
089f9d79 JS |
8105 | if ((look_in_map == MAP_0 || look_in_map == MAP_X) && |
8106 | i < map->num_members) { | |
d5985138 AK |
8107 | ord = __le32_to_cpu(map->disk_ord_tbl[i]); |
8108 | idx = ord_to_idx(ord); | |
8109 | ||
8110 | if (idx != idx_1) { | |
8111 | disk = get_imsm_disk(super, idx); | |
8112 | if (!disk || is_failed(disk) || | |
8113 | ord & IMSM_ORD_REBUILD) | |
8114 | failed++; | |
8115 | } | |
8116 | } | |
c2a1e7da DW |
8117 | } |
8118 | ||
8119 | return failed; | |
845dea95 NB |
8120 | } |
8121 | ||
97b4d0e9 DW |
8122 | static int imsm_open_new(struct supertype *c, struct active_array *a, |
8123 | char *inst) | |
8124 | { | |
8125 | struct intel_super *super = c->sb; | |
8126 | struct imsm_super *mpb = super->anchor; | |
bbab0940 | 8127 | struct imsm_update_prealloc_bb_mem u; |
9587c373 | 8128 | |
97b4d0e9 | 8129 | if (atoi(inst) >= mpb->num_raid_devs) { |
1ade5cc1 | 8130 | pr_err("subarry index %d, out of range\n", atoi(inst)); |
97b4d0e9 DW |
8131 | return -ENODEV; |
8132 | } | |
8133 | ||
8134 | dprintf("imsm: open_new %s\n", inst); | |
8135 | a->info.container_member = atoi(inst); | |
bbab0940 TM |
8136 | |
8137 | u.type = update_prealloc_badblocks_mem; | |
8138 | imsm_update_metadata_locally(c, &u, sizeof(u)); | |
8139 | ||
97b4d0e9 DW |
8140 | return 0; |
8141 | } | |
8142 | ||
0c046afd DW |
8143 | static int is_resyncing(struct imsm_dev *dev) |
8144 | { | |
8145 | struct imsm_map *migr_map; | |
8146 | ||
8147 | if (!dev->vol.migr_state) | |
8148 | return 0; | |
8149 | ||
1484e727 DW |
8150 | if (migr_type(dev) == MIGR_INIT || |
8151 | migr_type(dev) == MIGR_REPAIR) | |
0c046afd DW |
8152 | return 1; |
8153 | ||
4c9bc37b AK |
8154 | if (migr_type(dev) == MIGR_GEN_MIGR) |
8155 | return 0; | |
8156 | ||
238c0a71 | 8157 | migr_map = get_imsm_map(dev, MAP_1); |
0c046afd | 8158 | |
089f9d79 JS |
8159 | if (migr_map->map_state == IMSM_T_STATE_NORMAL && |
8160 | dev->vol.migr_type != MIGR_GEN_MIGR) | |
0c046afd DW |
8161 | return 1; |
8162 | else | |
8163 | return 0; | |
8164 | } | |
8165 | ||
0556e1a2 | 8166 | /* return true if we recorded new information */ |
4c9e8c1e TM |
8167 | static int mark_failure(struct intel_super *super, |
8168 | struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
47ee5a45 | 8169 | { |
0556e1a2 DW |
8170 | __u32 ord; |
8171 | int slot; | |
8172 | struct imsm_map *map; | |
86c54047 DW |
8173 | char buf[MAX_RAID_SERIAL_LEN+3]; |
8174 | unsigned int len, shift = 0; | |
0556e1a2 DW |
8175 | |
8176 | /* new failures are always set in map[0] */ | |
238c0a71 | 8177 | map = get_imsm_map(dev, MAP_0); |
0556e1a2 DW |
8178 | |
8179 | slot = get_imsm_disk_slot(map, idx); | |
8180 | if (slot < 0) | |
8181 | return 0; | |
8182 | ||
8183 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
25ed7e59 | 8184 | if (is_failed(disk) && (ord & IMSM_ORD_REBUILD)) |
0556e1a2 DW |
8185 | return 0; |
8186 | ||
7d0c5e24 LD |
8187 | memcpy(buf, disk->serial, MAX_RAID_SERIAL_LEN); |
8188 | buf[MAX_RAID_SERIAL_LEN] = '\000'; | |
8189 | strcat(buf, ":0"); | |
86c54047 DW |
8190 | if ((len = strlen(buf)) >= MAX_RAID_SERIAL_LEN) |
8191 | shift = len - MAX_RAID_SERIAL_LEN + 1; | |
167d8bb8 | 8192 | memcpy(disk->serial, &buf[shift], len + 1 - shift); |
86c54047 | 8193 | |
f2f27e63 | 8194 | disk->status |= FAILED_DISK; |
0556e1a2 | 8195 | set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD); |
17788994 AK |
8196 | /* mark failures in second map if second map exists and this disk |
8197 | * in this slot. | |
8198 | * This is valid for migration, initialization and rebuild | |
8199 | */ | |
8200 | if (dev->vol.migr_state) { | |
238c0a71 | 8201 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); |
0a108d63 AK |
8202 | int slot2 = get_imsm_disk_slot(map2, idx); |
8203 | ||
089f9d79 | 8204 | if (slot2 < map2->num_members && slot2 >= 0) |
0a108d63 | 8205 | set_imsm_ord_tbl_ent(map2, slot2, |
1ace8403 AK |
8206 | idx | IMSM_ORD_REBUILD); |
8207 | } | |
d7a1fda2 MT |
8208 | if (map->failed_disk_num == 0xff || |
8209 | (!is_rebuilding(dev) && map->failed_disk_num > slot)) | |
0556e1a2 | 8210 | map->failed_disk_num = slot; |
4c9e8c1e TM |
8211 | |
8212 | clear_disk_badblocks(super->bbm_log, ord_to_idx(ord)); | |
8213 | ||
0556e1a2 DW |
8214 | return 1; |
8215 | } | |
8216 | ||
4c9e8c1e TM |
8217 | static void mark_missing(struct intel_super *super, |
8218 | struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
0556e1a2 | 8219 | { |
4c9e8c1e | 8220 | mark_failure(super, dev, disk, idx); |
0556e1a2 DW |
8221 | |
8222 | if (disk->scsi_id == __cpu_to_le32(~(__u32)0)) | |
8223 | return; | |
8224 | ||
47ee5a45 DW |
8225 | disk->scsi_id = __cpu_to_le32(~(__u32)0); |
8226 | memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1); | |
8227 | } | |
8228 | ||
33414a01 DW |
8229 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev) |
8230 | { | |
33414a01 | 8231 | struct dl *dl; |
33414a01 DW |
8232 | |
8233 | if (!super->missing) | |
8234 | return; | |
33414a01 | 8235 | |
79b68f1b PC |
8236 | /* When orom adds replacement for missing disk it does |
8237 | * not remove entry of missing disk, but just updates map with | |
8238 | * new added disk. So it is not enough just to test if there is | |
8239 | * any missing disk, we have to look if there are any failed disks | |
8240 | * in map to stop migration */ | |
8241 | ||
33414a01 | 8242 | dprintf("imsm: mark missing\n"); |
3d59f0c0 AK |
8243 | /* end process for initialization and rebuild only |
8244 | */ | |
8245 | if (is_gen_migration(dev) == 0) { | |
fb12a745 | 8246 | int failed = imsm_count_failed(super, dev, MAP_0); |
3d59f0c0 | 8247 | |
fb12a745 TM |
8248 | if (failed) { |
8249 | __u8 map_state; | |
8250 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
8251 | struct imsm_map *map1; | |
8252 | int i, ord, ord_map1; | |
8253 | int rebuilt = 1; | |
3d59f0c0 | 8254 | |
fb12a745 TM |
8255 | for (i = 0; i < map->num_members; i++) { |
8256 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_0); | |
8257 | if (!(ord & IMSM_ORD_REBUILD)) | |
8258 | continue; | |
8259 | ||
8260 | map1 = get_imsm_map(dev, MAP_1); | |
8261 | if (!map1) | |
8262 | continue; | |
8263 | ||
8264 | ord_map1 = __le32_to_cpu(map1->disk_ord_tbl[i]); | |
8265 | if (ord_map1 & IMSM_ORD_REBUILD) | |
8266 | rebuilt = 0; | |
8267 | } | |
8268 | ||
8269 | if (rebuilt) { | |
8270 | map_state = imsm_check_degraded(super, dev, | |
8271 | failed, MAP_0); | |
8272 | end_migration(dev, super, map_state); | |
8273 | } | |
8274 | } | |
3d59f0c0 | 8275 | } |
33414a01 | 8276 | for (dl = super->missing; dl; dl = dl->next) |
4c9e8c1e | 8277 | mark_missing(super, dev, &dl->disk, dl->index); |
33414a01 DW |
8278 | super->updates_pending++; |
8279 | } | |
8280 | ||
f3871fdc AK |
8281 | static unsigned long long imsm_set_array_size(struct imsm_dev *dev, |
8282 | long long new_size) | |
70bdf0dc | 8283 | { |
70bdf0dc | 8284 | unsigned long long array_blocks; |
9529d343 MD |
8285 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
8286 | int used_disks = imsm_num_data_members(map); | |
70bdf0dc AK |
8287 | |
8288 | if (used_disks == 0) { | |
8289 | /* when problems occures | |
8290 | * return current array_blocks value | |
8291 | */ | |
fcc2c9da | 8292 | array_blocks = imsm_dev_size(dev); |
70bdf0dc AK |
8293 | |
8294 | return array_blocks; | |
8295 | } | |
8296 | ||
8297 | /* set array size in metadata | |
8298 | */ | |
9529d343 | 8299 | if (new_size <= 0) |
f3871fdc AK |
8300 | /* OLCE size change is caused by added disks |
8301 | */ | |
44490938 | 8302 | array_blocks = per_dev_array_size(map) * used_disks; |
9529d343 | 8303 | else |
f3871fdc AK |
8304 | /* Online Volume Size Change |
8305 | * Using available free space | |
8306 | */ | |
8307 | array_blocks = new_size; | |
70bdf0dc | 8308 | |
b53bfba6 | 8309 | array_blocks = round_size_to_mb(array_blocks, used_disks); |
fcc2c9da | 8310 | set_imsm_dev_size(dev, array_blocks); |
70bdf0dc AK |
8311 | |
8312 | return array_blocks; | |
8313 | } | |
8314 | ||
28bce06f AK |
8315 | static void imsm_set_disk(struct active_array *a, int n, int state); |
8316 | ||
0e2d1a4e AK |
8317 | static void imsm_progress_container_reshape(struct intel_super *super) |
8318 | { | |
8319 | /* if no device has a migr_state, but some device has a | |
8320 | * different number of members than the previous device, start | |
8321 | * changing the number of devices in this device to match | |
8322 | * previous. | |
8323 | */ | |
8324 | struct imsm_super *mpb = super->anchor; | |
8325 | int prev_disks = -1; | |
8326 | int i; | |
1dfaa380 | 8327 | int copy_map_size; |
0e2d1a4e AK |
8328 | |
8329 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
8330 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 8331 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
0e2d1a4e AK |
8332 | struct imsm_map *map2; |
8333 | int prev_num_members; | |
0e2d1a4e AK |
8334 | |
8335 | if (dev->vol.migr_state) | |
8336 | return; | |
8337 | ||
8338 | if (prev_disks == -1) | |
8339 | prev_disks = map->num_members; | |
8340 | if (prev_disks == map->num_members) | |
8341 | continue; | |
8342 | ||
8343 | /* OK, this array needs to enter reshape mode. | |
8344 | * i.e it needs a migr_state | |
8345 | */ | |
8346 | ||
1dfaa380 | 8347 | copy_map_size = sizeof_imsm_map(map); |
0e2d1a4e AK |
8348 | prev_num_members = map->num_members; |
8349 | map->num_members = prev_disks; | |
8350 | dev->vol.migr_state = 1; | |
8351 | dev->vol.curr_migr_unit = 0; | |
ea672ee1 | 8352 | set_migr_type(dev, MIGR_GEN_MIGR); |
0e2d1a4e AK |
8353 | for (i = prev_num_members; |
8354 | i < map->num_members; i++) | |
8355 | set_imsm_ord_tbl_ent(map, i, i); | |
238c0a71 | 8356 | map2 = get_imsm_map(dev, MAP_1); |
0e2d1a4e | 8357 | /* Copy the current map */ |
1dfaa380 | 8358 | memcpy(map2, map, copy_map_size); |
0e2d1a4e AK |
8359 | map2->num_members = prev_num_members; |
8360 | ||
f3871fdc | 8361 | imsm_set_array_size(dev, -1); |
51d83f5d | 8362 | super->clean_migration_record_by_mdmon = 1; |
0e2d1a4e AK |
8363 | super->updates_pending++; |
8364 | } | |
8365 | } | |
8366 | ||
aad6f216 | 8367 | /* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild |
0c046afd DW |
8368 | * states are handled in imsm_set_disk() with one exception, when a |
8369 | * resync is stopped due to a new failure this routine will set the | |
8370 | * 'degraded' state for the array. | |
8371 | */ | |
01f157d7 | 8372 | static int imsm_set_array_state(struct active_array *a, int consistent) |
a862209d DW |
8373 | { |
8374 | int inst = a->info.container_member; | |
8375 | struct intel_super *super = a->container->sb; | |
949c47a0 | 8376 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 8377 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
3b451610 AK |
8378 | int failed = imsm_count_failed(super, dev, MAP_0); |
8379 | __u8 map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
1e5c6983 | 8380 | __u32 blocks_per_unit; |
a862209d | 8381 | |
1af97990 AK |
8382 | if (dev->vol.migr_state && |
8383 | dev->vol.migr_type == MIGR_GEN_MIGR) { | |
8384 | /* array state change is blocked due to reshape action | |
aad6f216 N |
8385 | * We might need to |
8386 | * - abort the reshape (if last_checkpoint is 0 and action!= reshape) | |
8387 | * - finish the reshape (if last_checkpoint is big and action != reshape) | |
8388 | * - update curr_migr_unit | |
1af97990 | 8389 | */ |
aad6f216 N |
8390 | if (a->curr_action == reshape) { |
8391 | /* still reshaping, maybe update curr_migr_unit */ | |
633b5610 | 8392 | goto mark_checkpoint; |
aad6f216 N |
8393 | } else { |
8394 | if (a->last_checkpoint == 0 && a->prev_action == reshape) { | |
8395 | /* for some reason we aborted the reshape. | |
b66e591b AK |
8396 | * |
8397 | * disable automatic metadata rollback | |
8398 | * user action is required to recover process | |
aad6f216 | 8399 | */ |
b66e591b | 8400 | if (0) { |
238c0a71 AK |
8401 | struct imsm_map *map2 = |
8402 | get_imsm_map(dev, MAP_1); | |
8403 | dev->vol.migr_state = 0; | |
8404 | set_migr_type(dev, 0); | |
8405 | dev->vol.curr_migr_unit = 0; | |
8406 | memcpy(map, map2, | |
8407 | sizeof_imsm_map(map2)); | |
8408 | super->updates_pending++; | |
b66e591b | 8409 | } |
aad6f216 N |
8410 | } |
8411 | if (a->last_checkpoint >= a->info.component_size) { | |
8412 | unsigned long long array_blocks; | |
8413 | int used_disks; | |
e154ced3 | 8414 | struct mdinfo *mdi; |
aad6f216 | 8415 | |
9529d343 | 8416 | used_disks = imsm_num_data_members(map); |
d55adef9 AK |
8417 | if (used_disks > 0) { |
8418 | array_blocks = | |
44490938 | 8419 | per_dev_array_size(map) * |
d55adef9 | 8420 | used_disks; |
b53bfba6 TM |
8421 | array_blocks = |
8422 | round_size_to_mb(array_blocks, | |
8423 | used_disks); | |
d55adef9 AK |
8424 | a->info.custom_array_size = array_blocks; |
8425 | /* encourage manager to update array | |
8426 | * size | |
8427 | */ | |
e154ced3 | 8428 | |
d55adef9 | 8429 | a->check_reshape = 1; |
633b5610 | 8430 | } |
e154ced3 AK |
8431 | /* finalize online capacity expansion/reshape */ |
8432 | for (mdi = a->info.devs; mdi; mdi = mdi->next) | |
8433 | imsm_set_disk(a, | |
8434 | mdi->disk.raid_disk, | |
8435 | mdi->curr_state); | |
8436 | ||
0e2d1a4e | 8437 | imsm_progress_container_reshape(super); |
e154ced3 | 8438 | } |
aad6f216 | 8439 | } |
1af97990 AK |
8440 | } |
8441 | ||
47ee5a45 | 8442 | /* before we activate this array handle any missing disks */ |
33414a01 DW |
8443 | if (consistent == 2) |
8444 | handle_missing(super, dev); | |
1e5c6983 | 8445 | |
0c046afd | 8446 | if (consistent == 2 && |
b7941fd6 | 8447 | (!is_resync_complete(&a->info) || |
0c046afd DW |
8448 | map_state != IMSM_T_STATE_NORMAL || |
8449 | dev->vol.migr_state)) | |
01f157d7 | 8450 | consistent = 0; |
272906ef | 8451 | |
b7941fd6 | 8452 | if (is_resync_complete(&a->info)) { |
0c046afd | 8453 | /* complete intialization / resync, |
0556e1a2 DW |
8454 | * recovery and interrupted recovery is completed in |
8455 | * ->set_disk | |
0c046afd DW |
8456 | */ |
8457 | if (is_resyncing(dev)) { | |
8458 | dprintf("imsm: mark resync done\n"); | |
809da78e | 8459 | end_migration(dev, super, map_state); |
115c3803 | 8460 | super->updates_pending++; |
484240d8 | 8461 | a->last_checkpoint = 0; |
115c3803 | 8462 | } |
b9172665 AK |
8463 | } else if ((!is_resyncing(dev) && !failed) && |
8464 | (imsm_reshape_blocks_arrays_changes(super) == 0)) { | |
0c046afd | 8465 | /* mark the start of the init process if nothing is failed */ |
b7941fd6 | 8466 | dprintf("imsm: mark resync start\n"); |
1484e727 | 8467 | if (map->map_state == IMSM_T_STATE_UNINITIALIZED) |
8e59f3d8 | 8468 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_INIT); |
1484e727 | 8469 | else |
8e59f3d8 | 8470 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_REPAIR); |
3393c6af | 8471 | super->updates_pending++; |
115c3803 | 8472 | } |
a862209d | 8473 | |
633b5610 | 8474 | mark_checkpoint: |
5b83bacf AK |
8475 | /* skip checkpointing for general migration, |
8476 | * it is controlled in mdadm | |
8477 | */ | |
8478 | if (is_gen_migration(dev)) | |
8479 | goto skip_mark_checkpoint; | |
8480 | ||
1e5c6983 | 8481 | /* check if we can update curr_migr_unit from resync_start, recovery_start */ |
c47b0ff6 | 8482 | blocks_per_unit = blocks_per_migr_unit(super, dev); |
4f0a7acc | 8483 | if (blocks_per_unit) { |
1e5c6983 DW |
8484 | __u32 units32; |
8485 | __u64 units; | |
8486 | ||
4f0a7acc | 8487 | units = a->last_checkpoint / blocks_per_unit; |
1e5c6983 DW |
8488 | units32 = units; |
8489 | ||
8490 | /* check that we did not overflow 32-bits, and that | |
8491 | * curr_migr_unit needs updating | |
8492 | */ | |
8493 | if (units32 == units && | |
bfd80a56 | 8494 | units32 != 0 && |
1e5c6983 DW |
8495 | __le32_to_cpu(dev->vol.curr_migr_unit) != units32) { |
8496 | dprintf("imsm: mark checkpoint (%u)\n", units32); | |
8497 | dev->vol.curr_migr_unit = __cpu_to_le32(units32); | |
8498 | super->updates_pending++; | |
8499 | } | |
8500 | } | |
f8f603f1 | 8501 | |
5b83bacf | 8502 | skip_mark_checkpoint: |
3393c6af | 8503 | /* mark dirty / clean */ |
2432ce9b AP |
8504 | if (((dev->vol.dirty & RAIDVOL_DIRTY) && consistent) || |
8505 | (!(dev->vol.dirty & RAIDVOL_DIRTY) && !consistent)) { | |
b7941fd6 | 8506 | dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty"); |
2432ce9b AP |
8507 | if (consistent) { |
8508 | dev->vol.dirty = RAIDVOL_CLEAN; | |
8509 | } else { | |
8510 | dev->vol.dirty = RAIDVOL_DIRTY; | |
c2462068 PB |
8511 | if (dev->rwh_policy == RWH_DISTRIBUTED || |
8512 | dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED) | |
2432ce9b AP |
8513 | dev->vol.dirty |= RAIDVOL_DSRECORD_VALID; |
8514 | } | |
a862209d DW |
8515 | super->updates_pending++; |
8516 | } | |
28bce06f | 8517 | |
01f157d7 | 8518 | return consistent; |
a862209d DW |
8519 | } |
8520 | ||
6f50473f TM |
8521 | static int imsm_disk_slot_to_ord(struct active_array *a, int slot) |
8522 | { | |
8523 | int inst = a->info.container_member; | |
8524 | struct intel_super *super = a->container->sb; | |
8525 | struct imsm_dev *dev = get_imsm_dev(super, inst); | |
8526 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
8527 | ||
8528 | if (slot > map->num_members) { | |
8529 | pr_err("imsm: imsm_disk_slot_to_ord %d out of range 0..%d\n", | |
8530 | slot, map->num_members - 1); | |
8531 | return -1; | |
8532 | } | |
8533 | ||
8534 | if (slot < 0) | |
8535 | return -1; | |
8536 | ||
8537 | return get_imsm_ord_tbl_ent(dev, slot, MAP_0); | |
8538 | } | |
8539 | ||
8d45d196 | 8540 | static void imsm_set_disk(struct active_array *a, int n, int state) |
845dea95 | 8541 | { |
8d45d196 DW |
8542 | int inst = a->info.container_member; |
8543 | struct intel_super *super = a->container->sb; | |
949c47a0 | 8544 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 8545 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
8d45d196 | 8546 | struct imsm_disk *disk; |
7ce05701 LD |
8547 | struct mdinfo *mdi; |
8548 | int recovery_not_finished = 0; | |
0c046afd | 8549 | int failed; |
6f50473f | 8550 | int ord; |
0c046afd | 8551 | __u8 map_state; |
fb12a745 TM |
8552 | int rebuild_done = 0; |
8553 | int i; | |
8d45d196 | 8554 | |
fb12a745 | 8555 | ord = get_imsm_ord_tbl_ent(dev, n, MAP_X); |
6f50473f | 8556 | if (ord < 0) |
8d45d196 DW |
8557 | return; |
8558 | ||
4e6e574a | 8559 | dprintf("imsm: set_disk %d:%x\n", n, state); |
b10b37b8 | 8560 | disk = get_imsm_disk(super, ord_to_idx(ord)); |
8d45d196 | 8561 | |
5802a811 | 8562 | /* check for new failures */ |
ae7d61e3 | 8563 | if (disk && (state & DS_FAULTY)) { |
4c9e8c1e | 8564 | if (mark_failure(super, dev, disk, ord_to_idx(ord))) |
0556e1a2 | 8565 | super->updates_pending++; |
8d45d196 | 8566 | } |
47ee5a45 | 8567 | |
19859edc | 8568 | /* check if in_sync */ |
0556e1a2 | 8569 | if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) { |
238c0a71 | 8570 | struct imsm_map *migr_map = get_imsm_map(dev, MAP_1); |
b10b37b8 DW |
8571 | |
8572 | set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord)); | |
fb12a745 | 8573 | rebuild_done = 1; |
19859edc DW |
8574 | super->updates_pending++; |
8575 | } | |
8d45d196 | 8576 | |
3b451610 AK |
8577 | failed = imsm_count_failed(super, dev, MAP_0); |
8578 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
5802a811 | 8579 | |
0c046afd | 8580 | /* check if recovery complete, newly degraded, or failed */ |
94002678 AK |
8581 | dprintf("imsm: Detected transition to state "); |
8582 | switch (map_state) { | |
8583 | case IMSM_T_STATE_NORMAL: /* transition to normal state */ | |
8584 | dprintf("normal: "); | |
8585 | if (is_rebuilding(dev)) { | |
1ade5cc1 | 8586 | dprintf_cont("while rebuilding"); |
7ce05701 LD |
8587 | /* check if recovery is really finished */ |
8588 | for (mdi = a->info.devs; mdi ; mdi = mdi->next) | |
8589 | if (mdi->recovery_start != MaxSector) { | |
8590 | recovery_not_finished = 1; | |
8591 | break; | |
8592 | } | |
8593 | if (recovery_not_finished) { | |
1ade5cc1 N |
8594 | dprintf_cont("\n"); |
8595 | dprintf("Rebuild has not finished yet, state not changed"); | |
7ce05701 LD |
8596 | if (a->last_checkpoint < mdi->recovery_start) { |
8597 | a->last_checkpoint = mdi->recovery_start; | |
8598 | super->updates_pending++; | |
8599 | } | |
8600 | break; | |
8601 | } | |
94002678 | 8602 | end_migration(dev, super, map_state); |
238c0a71 | 8603 | map = get_imsm_map(dev, MAP_0); |
94002678 AK |
8604 | map->failed_disk_num = ~0; |
8605 | super->updates_pending++; | |
8606 | a->last_checkpoint = 0; | |
8607 | break; | |
8608 | } | |
8609 | if (is_gen_migration(dev)) { | |
1ade5cc1 | 8610 | dprintf_cont("while general migration"); |
bf2f0071 | 8611 | if (a->last_checkpoint >= a->info.component_size) |
809da78e | 8612 | end_migration(dev, super, map_state); |
94002678 AK |
8613 | else |
8614 | map->map_state = map_state; | |
238c0a71 | 8615 | map = get_imsm_map(dev, MAP_0); |
28bce06f | 8616 | map->failed_disk_num = ~0; |
94002678 | 8617 | super->updates_pending++; |
bf2f0071 | 8618 | break; |
94002678 AK |
8619 | } |
8620 | break; | |
8621 | case IMSM_T_STATE_DEGRADED: /* transition to degraded state */ | |
1ade5cc1 | 8622 | dprintf_cont("degraded: "); |
089f9d79 | 8623 | if (map->map_state != map_state && !dev->vol.migr_state) { |
1ade5cc1 | 8624 | dprintf_cont("mark degraded"); |
94002678 AK |
8625 | map->map_state = map_state; |
8626 | super->updates_pending++; | |
8627 | a->last_checkpoint = 0; | |
8628 | break; | |
8629 | } | |
8630 | if (is_rebuilding(dev)) { | |
d7a1fda2 | 8631 | dprintf_cont("while rebuilding "); |
a4e96fd8 MT |
8632 | if (state & DS_FAULTY) { |
8633 | dprintf_cont("removing failed drive "); | |
d7a1fda2 MT |
8634 | if (n == map->failed_disk_num) { |
8635 | dprintf_cont("end migration"); | |
8636 | end_migration(dev, super, map_state); | |
a4e96fd8 | 8637 | a->last_checkpoint = 0; |
d7a1fda2 | 8638 | } else { |
a4e96fd8 | 8639 | dprintf_cont("fail detected during rebuild, changing map state"); |
d7a1fda2 MT |
8640 | map->map_state = map_state; |
8641 | } | |
94002678 | 8642 | super->updates_pending++; |
fb12a745 TM |
8643 | } |
8644 | ||
a4e96fd8 MT |
8645 | if (!rebuild_done) |
8646 | break; | |
8647 | ||
fb12a745 TM |
8648 | /* check if recovery is really finished */ |
8649 | for (mdi = a->info.devs; mdi ; mdi = mdi->next) | |
8650 | if (mdi->recovery_start != MaxSector) { | |
8651 | recovery_not_finished = 1; | |
8652 | break; | |
8653 | } | |
8654 | if (recovery_not_finished) { | |
8655 | dprintf_cont("\n"); | |
a4e96fd8 | 8656 | dprintf_cont("Rebuild has not finished yet"); |
fb12a745 TM |
8657 | if (a->last_checkpoint < mdi->recovery_start) { |
8658 | a->last_checkpoint = | |
8659 | mdi->recovery_start; | |
8660 | super->updates_pending++; | |
8661 | } | |
8662 | break; | |
94002678 | 8663 | } |
fb12a745 TM |
8664 | |
8665 | dprintf_cont(" Rebuild done, still degraded"); | |
a4e96fd8 MT |
8666 | end_migration(dev, super, map_state); |
8667 | a->last_checkpoint = 0; | |
8668 | super->updates_pending++; | |
fb12a745 TM |
8669 | |
8670 | for (i = 0; i < map->num_members; i++) { | |
8671 | int idx = get_imsm_ord_tbl_ent(dev, i, MAP_0); | |
8672 | ||
8673 | if (idx & IMSM_ORD_REBUILD) | |
8674 | map->failed_disk_num = i; | |
8675 | } | |
8676 | super->updates_pending++; | |
94002678 AK |
8677 | break; |
8678 | } | |
8679 | if (is_gen_migration(dev)) { | |
1ade5cc1 | 8680 | dprintf_cont("while general migration"); |
bf2f0071 | 8681 | if (a->last_checkpoint >= a->info.component_size) |
809da78e | 8682 | end_migration(dev, super, map_state); |
94002678 AK |
8683 | else { |
8684 | map->map_state = map_state; | |
3b451610 | 8685 | manage_second_map(super, dev); |
94002678 AK |
8686 | } |
8687 | super->updates_pending++; | |
bf2f0071 | 8688 | break; |
28bce06f | 8689 | } |
6ce1fbf1 | 8690 | if (is_initializing(dev)) { |
1ade5cc1 | 8691 | dprintf_cont("while initialization."); |
6ce1fbf1 AK |
8692 | map->map_state = map_state; |
8693 | super->updates_pending++; | |
8694 | break; | |
8695 | } | |
94002678 AK |
8696 | break; |
8697 | case IMSM_T_STATE_FAILED: /* transition to failed state */ | |
1ade5cc1 | 8698 | dprintf_cont("failed: "); |
94002678 | 8699 | if (is_gen_migration(dev)) { |
1ade5cc1 | 8700 | dprintf_cont("while general migration"); |
94002678 AK |
8701 | map->map_state = map_state; |
8702 | super->updates_pending++; | |
8703 | break; | |
8704 | } | |
8705 | if (map->map_state != map_state) { | |
1ade5cc1 | 8706 | dprintf_cont("mark failed"); |
94002678 AK |
8707 | end_migration(dev, super, map_state); |
8708 | super->updates_pending++; | |
8709 | a->last_checkpoint = 0; | |
8710 | break; | |
8711 | } | |
8712 | break; | |
8713 | default: | |
1ade5cc1 | 8714 | dprintf_cont("state %i\n", map_state); |
5802a811 | 8715 | } |
1ade5cc1 | 8716 | dprintf_cont("\n"); |
845dea95 NB |
8717 | } |
8718 | ||
f796af5d | 8719 | static int store_imsm_mpb(int fd, struct imsm_super *mpb) |
c2a1e7da | 8720 | { |
f796af5d | 8721 | void *buf = mpb; |
c2a1e7da DW |
8722 | __u32 mpb_size = __le32_to_cpu(mpb->mpb_size); |
8723 | unsigned long long dsize; | |
8724 | unsigned long long sectors; | |
f36a9ecd | 8725 | unsigned int sector_size; |
c2a1e7da | 8726 | |
f36a9ecd | 8727 | get_dev_sector_size(fd, NULL, §or_size); |
c2a1e7da DW |
8728 | get_dev_size(fd, NULL, &dsize); |
8729 | ||
f36a9ecd | 8730 | if (mpb_size > sector_size) { |
272f648f | 8731 | /* -1 to account for anchor */ |
f36a9ecd | 8732 | sectors = mpb_sectors(mpb, sector_size) - 1; |
c2a1e7da | 8733 | |
272f648f | 8734 | /* write the extended mpb to the sectors preceeding the anchor */ |
f36a9ecd PB |
8735 | if (lseek64(fd, dsize - (sector_size * (2 + sectors)), |
8736 | SEEK_SET) < 0) | |
272f648f | 8737 | return 1; |
c2a1e7da | 8738 | |
f36a9ecd PB |
8739 | if ((unsigned long long)write(fd, buf + sector_size, |
8740 | sector_size * sectors) != sector_size * sectors) | |
272f648f DW |
8741 | return 1; |
8742 | } | |
c2a1e7da | 8743 | |
272f648f | 8744 | /* first block is stored on second to last sector of the disk */ |
f36a9ecd | 8745 | if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0) |
c2a1e7da DW |
8746 | return 1; |
8747 | ||
466070ad | 8748 | if ((unsigned int)write(fd, buf, sector_size) != sector_size) |
c2a1e7da DW |
8749 | return 1; |
8750 | ||
c2a1e7da DW |
8751 | return 0; |
8752 | } | |
8753 | ||
2e735d19 | 8754 | static void imsm_sync_metadata(struct supertype *container) |
845dea95 | 8755 | { |
2e735d19 | 8756 | struct intel_super *super = container->sb; |
c2a1e7da | 8757 | |
1a64be56 | 8758 | dprintf("sync metadata: %d\n", super->updates_pending); |
c2a1e7da DW |
8759 | if (!super->updates_pending) |
8760 | return; | |
8761 | ||
36988a3d | 8762 | write_super_imsm(container, 0); |
c2a1e7da DW |
8763 | |
8764 | super->updates_pending = 0; | |
845dea95 NB |
8765 | } |
8766 | ||
272906ef DW |
8767 | static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a) |
8768 | { | |
8769 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 8770 | int i = get_imsm_disk_idx(dev, idx, MAP_X); |
272906ef DW |
8771 | struct dl *dl; |
8772 | ||
8773 | for (dl = super->disks; dl; dl = dl->next) | |
8774 | if (dl->index == i) | |
8775 | break; | |
8776 | ||
25ed7e59 | 8777 | if (dl && is_failed(&dl->disk)) |
272906ef DW |
8778 | dl = NULL; |
8779 | ||
8780 | if (dl) | |
1ade5cc1 | 8781 | dprintf("found %x:%x\n", dl->major, dl->minor); |
272906ef DW |
8782 | |
8783 | return dl; | |
8784 | } | |
8785 | ||
a20d2ba5 | 8786 | static struct dl *imsm_add_spare(struct intel_super *super, int slot, |
8ba77d32 AK |
8787 | struct active_array *a, int activate_new, |
8788 | struct mdinfo *additional_test_list) | |
272906ef DW |
8789 | { |
8790 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 8791 | int idx = get_imsm_disk_idx(dev, slot, MAP_X); |
a20d2ba5 DW |
8792 | struct imsm_super *mpb = super->anchor; |
8793 | struct imsm_map *map; | |
272906ef DW |
8794 | unsigned long long pos; |
8795 | struct mdinfo *d; | |
8796 | struct extent *ex; | |
a20d2ba5 | 8797 | int i, j; |
272906ef | 8798 | int found; |
569cc43f DW |
8799 | __u32 array_start = 0; |
8800 | __u32 array_end = 0; | |
272906ef | 8801 | struct dl *dl; |
6c932028 | 8802 | struct mdinfo *test_list; |
272906ef DW |
8803 | |
8804 | for (dl = super->disks; dl; dl = dl->next) { | |
8805 | /* If in this array, skip */ | |
8806 | for (d = a->info.devs ; d ; d = d->next) | |
e553d2a4 DW |
8807 | if (d->state_fd >= 0 && |
8808 | d->disk.major == dl->major && | |
272906ef | 8809 | d->disk.minor == dl->minor) { |
8ba77d32 AK |
8810 | dprintf("%x:%x already in array\n", |
8811 | dl->major, dl->minor); | |
272906ef DW |
8812 | break; |
8813 | } | |
8814 | if (d) | |
8815 | continue; | |
6c932028 AK |
8816 | test_list = additional_test_list; |
8817 | while (test_list) { | |
8818 | if (test_list->disk.major == dl->major && | |
8819 | test_list->disk.minor == dl->minor) { | |
8ba77d32 AK |
8820 | dprintf("%x:%x already in additional test list\n", |
8821 | dl->major, dl->minor); | |
8822 | break; | |
8823 | } | |
6c932028 | 8824 | test_list = test_list->next; |
8ba77d32 | 8825 | } |
6c932028 | 8826 | if (test_list) |
8ba77d32 | 8827 | continue; |
272906ef | 8828 | |
e553d2a4 | 8829 | /* skip in use or failed drives */ |
25ed7e59 | 8830 | if (is_failed(&dl->disk) || idx == dl->index || |
df474657 DW |
8831 | dl->index == -2) { |
8832 | dprintf("%x:%x status (failed: %d index: %d)\n", | |
25ed7e59 | 8833 | dl->major, dl->minor, is_failed(&dl->disk), idx); |
9a1608e5 DW |
8834 | continue; |
8835 | } | |
8836 | ||
a20d2ba5 DW |
8837 | /* skip pure spares when we are looking for partially |
8838 | * assimilated drives | |
8839 | */ | |
8840 | if (dl->index == -1 && !activate_new) | |
8841 | continue; | |
8842 | ||
f2cc4f7d AO |
8843 | if (!drive_validate_sector_size(super, dl)) |
8844 | continue; | |
8845 | ||
272906ef | 8846 | /* Does this unused device have the requisite free space? |
a20d2ba5 | 8847 | * It needs to be able to cover all member volumes |
272906ef | 8848 | */ |
05501181 | 8849 | ex = get_extents(super, dl, 1); |
272906ef DW |
8850 | if (!ex) { |
8851 | dprintf("cannot get extents\n"); | |
8852 | continue; | |
8853 | } | |
a20d2ba5 DW |
8854 | for (i = 0; i < mpb->num_raid_devs; i++) { |
8855 | dev = get_imsm_dev(super, i); | |
238c0a71 | 8856 | map = get_imsm_map(dev, MAP_0); |
272906ef | 8857 | |
a20d2ba5 DW |
8858 | /* check if this disk is already a member of |
8859 | * this array | |
272906ef | 8860 | */ |
620b1713 | 8861 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
a20d2ba5 DW |
8862 | continue; |
8863 | ||
8864 | found = 0; | |
8865 | j = 0; | |
8866 | pos = 0; | |
5551b113 | 8867 | array_start = pba_of_lba0(map); |
329c8278 | 8868 | array_end = array_start + |
44490938 | 8869 | per_dev_array_size(map) - 1; |
a20d2ba5 DW |
8870 | |
8871 | do { | |
8872 | /* check that we can start at pba_of_lba0 with | |
44490938 | 8873 | * num_data_stripes*blocks_per_stripe of space |
a20d2ba5 | 8874 | */ |
329c8278 | 8875 | if (array_start >= pos && array_end < ex[j].start) { |
a20d2ba5 DW |
8876 | found = 1; |
8877 | break; | |
8878 | } | |
8879 | pos = ex[j].start + ex[j].size; | |
8880 | j++; | |
8881 | } while (ex[j-1].size); | |
8882 | ||
8883 | if (!found) | |
272906ef | 8884 | break; |
a20d2ba5 | 8885 | } |
272906ef DW |
8886 | |
8887 | free(ex); | |
a20d2ba5 | 8888 | if (i < mpb->num_raid_devs) { |
329c8278 DW |
8889 | dprintf("%x:%x does not have %u to %u available\n", |
8890 | dl->major, dl->minor, array_start, array_end); | |
272906ef DW |
8891 | /* No room */ |
8892 | continue; | |
a20d2ba5 DW |
8893 | } |
8894 | return dl; | |
272906ef DW |
8895 | } |
8896 | ||
8897 | return dl; | |
8898 | } | |
8899 | ||
95d07a2c LM |
8900 | static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed) |
8901 | { | |
8902 | struct imsm_dev *dev2; | |
8903 | struct imsm_map *map; | |
8904 | struct dl *idisk; | |
8905 | int slot; | |
8906 | int idx; | |
8907 | __u8 state; | |
8908 | ||
8909 | dev2 = get_imsm_dev(cont->sb, dev_idx); | |
8910 | if (dev2) { | |
238c0a71 | 8911 | state = imsm_check_degraded(cont->sb, dev2, failed, MAP_0); |
95d07a2c | 8912 | if (state == IMSM_T_STATE_FAILED) { |
238c0a71 | 8913 | map = get_imsm_map(dev2, MAP_0); |
95d07a2c LM |
8914 | if (!map) |
8915 | return 1; | |
8916 | for (slot = 0; slot < map->num_members; slot++) { | |
8917 | /* | |
8918 | * Check if failed disks are deleted from intel | |
8919 | * disk list or are marked to be deleted | |
8920 | */ | |
238c0a71 | 8921 | idx = get_imsm_disk_idx(dev2, slot, MAP_X); |
95d07a2c LM |
8922 | idisk = get_imsm_dl_disk(cont->sb, idx); |
8923 | /* | |
8924 | * Do not rebuild the array if failed disks | |
8925 | * from failed sub-array are not removed from | |
8926 | * container. | |
8927 | */ | |
8928 | if (idisk && | |
8929 | is_failed(&idisk->disk) && | |
8930 | (idisk->action != DISK_REMOVE)) | |
8931 | return 0; | |
8932 | } | |
8933 | } | |
8934 | } | |
8935 | return 1; | |
8936 | } | |
8937 | ||
88758e9d DW |
8938 | static struct mdinfo *imsm_activate_spare(struct active_array *a, |
8939 | struct metadata_update **updates) | |
8940 | { | |
8941 | /** | |
d23fe947 DW |
8942 | * Find a device with unused free space and use it to replace a |
8943 | * failed/vacant region in an array. We replace failed regions one a | |
8944 | * array at a time. The result is that a new spare disk will be added | |
8945 | * to the first failed array and after the monitor has finished | |
8946 | * propagating failures the remainder will be consumed. | |
88758e9d | 8947 | * |
d23fe947 DW |
8948 | * FIXME add a capability for mdmon to request spares from another |
8949 | * container. | |
88758e9d DW |
8950 | */ |
8951 | ||
8952 | struct intel_super *super = a->container->sb; | |
88758e9d | 8953 | int inst = a->info.container_member; |
949c47a0 | 8954 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 8955 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
88758e9d DW |
8956 | int failed = a->info.array.raid_disks; |
8957 | struct mdinfo *rv = NULL; | |
8958 | struct mdinfo *d; | |
8959 | struct mdinfo *di; | |
8960 | struct metadata_update *mu; | |
8961 | struct dl *dl; | |
8962 | struct imsm_update_activate_spare *u; | |
8963 | int num_spares = 0; | |
8964 | int i; | |
95d07a2c | 8965 | int allowed; |
88758e9d DW |
8966 | |
8967 | for (d = a->info.devs ; d ; d = d->next) { | |
8968 | if ((d->curr_state & DS_FAULTY) && | |
8969 | d->state_fd >= 0) | |
8970 | /* wait for Removal to happen */ | |
8971 | return NULL; | |
8972 | if (d->state_fd >= 0) | |
8973 | failed--; | |
8974 | } | |
8975 | ||
8976 | dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n", | |
8977 | inst, failed, a->info.array.raid_disks, a->info.array.level); | |
1af97990 | 8978 | |
e2962bfc AK |
8979 | if (imsm_reshape_blocks_arrays_changes(super)) |
8980 | return NULL; | |
1af97990 | 8981 | |
fc8ca064 AK |
8982 | /* Cannot activate another spare if rebuild is in progress already |
8983 | */ | |
8984 | if (is_rebuilding(dev)) { | |
7a862a02 | 8985 | dprintf("imsm: No spare activation allowed. Rebuild in progress already.\n"); |
fc8ca064 AK |
8986 | return NULL; |
8987 | } | |
8988 | ||
89c67882 AK |
8989 | if (a->info.array.level == 4) |
8990 | /* No repair for takeovered array | |
8991 | * imsm doesn't support raid4 | |
8992 | */ | |
8993 | return NULL; | |
8994 | ||
3b451610 AK |
8995 | if (imsm_check_degraded(super, dev, failed, MAP_0) != |
8996 | IMSM_T_STATE_DEGRADED) | |
88758e9d DW |
8997 | return NULL; |
8998 | ||
83ca7d45 AP |
8999 | if (get_imsm_map(dev, MAP_0)->map_state == IMSM_T_STATE_UNINITIALIZED) { |
9000 | dprintf("imsm: No spare activation allowed. Volume is not initialized.\n"); | |
9001 | return NULL; | |
9002 | } | |
9003 | ||
95d07a2c LM |
9004 | /* |
9005 | * If there are any failed disks check state of the other volume. | |
9006 | * Block rebuild if the another one is failed until failed disks | |
9007 | * are removed from container. | |
9008 | */ | |
9009 | if (failed) { | |
7a862a02 | 9010 | dprintf("found failed disks in %.*s, check if there anotherfailed sub-array.\n", |
c4acd1e5 | 9011 | MAX_RAID_SERIAL_LEN, dev->volume); |
95d07a2c LM |
9012 | /* check if states of the other volumes allow for rebuild */ |
9013 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
9014 | if (i != inst) { | |
9015 | allowed = imsm_rebuild_allowed(a->container, | |
9016 | i, failed); | |
9017 | if (!allowed) | |
9018 | return NULL; | |
9019 | } | |
9020 | } | |
9021 | } | |
9022 | ||
88758e9d | 9023 | /* For each slot, if it is not working, find a spare */ |
88758e9d DW |
9024 | for (i = 0; i < a->info.array.raid_disks; i++) { |
9025 | for (d = a->info.devs ; d ; d = d->next) | |
9026 | if (d->disk.raid_disk == i) | |
9027 | break; | |
9028 | dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0); | |
9029 | if (d && (d->state_fd >= 0)) | |
9030 | continue; | |
9031 | ||
272906ef | 9032 | /* |
a20d2ba5 DW |
9033 | * OK, this device needs recovery. Try to re-add the |
9034 | * previous occupant of this slot, if this fails see if | |
9035 | * we can continue the assimilation of a spare that was | |
9036 | * partially assimilated, finally try to activate a new | |
9037 | * spare. | |
272906ef DW |
9038 | */ |
9039 | dl = imsm_readd(super, i, a); | |
9040 | if (!dl) | |
b303fe21 | 9041 | dl = imsm_add_spare(super, i, a, 0, rv); |
a20d2ba5 | 9042 | if (!dl) |
b303fe21 | 9043 | dl = imsm_add_spare(super, i, a, 1, rv); |
272906ef DW |
9044 | if (!dl) |
9045 | continue; | |
1011e834 | 9046 | |
272906ef | 9047 | /* found a usable disk with enough space */ |
503975b9 | 9048 | di = xcalloc(1, sizeof(*di)); |
272906ef DW |
9049 | |
9050 | /* dl->index will be -1 in the case we are activating a | |
9051 | * pristine spare. imsm_process_update() will create a | |
9052 | * new index in this case. Once a disk is found to be | |
9053 | * failed in all member arrays it is kicked from the | |
9054 | * metadata | |
9055 | */ | |
9056 | di->disk.number = dl->index; | |
d23fe947 | 9057 | |
272906ef DW |
9058 | /* (ab)use di->devs to store a pointer to the device |
9059 | * we chose | |
9060 | */ | |
9061 | di->devs = (struct mdinfo *) dl; | |
9062 | ||
9063 | di->disk.raid_disk = i; | |
9064 | di->disk.major = dl->major; | |
9065 | di->disk.minor = dl->minor; | |
9066 | di->disk.state = 0; | |
d23534e4 | 9067 | di->recovery_start = 0; |
5551b113 | 9068 | di->data_offset = pba_of_lba0(map); |
272906ef DW |
9069 | di->component_size = a->info.component_size; |
9070 | di->container_member = inst; | |
5e46202e | 9071 | di->bb.supported = 1; |
2c8890e9 | 9072 | if (a->info.consistency_policy == CONSISTENCY_POLICY_PPL) { |
2432ce9b | 9073 | di->ppl_sector = get_ppl_sector(super, inst); |
c2462068 | 9074 | di->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9; |
2432ce9b | 9075 | } |
148acb7b | 9076 | super->random = random32(); |
272906ef DW |
9077 | di->next = rv; |
9078 | rv = di; | |
9079 | num_spares++; | |
9080 | dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor, | |
9081 | i, di->data_offset); | |
88758e9d DW |
9082 | } |
9083 | ||
9084 | if (!rv) | |
9085 | /* No spares found */ | |
9086 | return rv; | |
9087 | /* Now 'rv' has a list of devices to return. | |
9088 | * Create a metadata_update record to update the | |
9089 | * disk_ord_tbl for the array | |
9090 | */ | |
503975b9 | 9091 | mu = xmalloc(sizeof(*mu)); |
1011e834 | 9092 | mu->buf = xcalloc(num_spares, |
503975b9 | 9093 | sizeof(struct imsm_update_activate_spare)); |
88758e9d | 9094 | mu->space = NULL; |
cb23f1f4 | 9095 | mu->space_list = NULL; |
88758e9d DW |
9096 | mu->len = sizeof(struct imsm_update_activate_spare) * num_spares; |
9097 | mu->next = *updates; | |
9098 | u = (struct imsm_update_activate_spare *) mu->buf; | |
9099 | ||
9100 | for (di = rv ; di ; di = di->next) { | |
9101 | u->type = update_activate_spare; | |
d23fe947 DW |
9102 | u->dl = (struct dl *) di->devs; |
9103 | di->devs = NULL; | |
88758e9d DW |
9104 | u->slot = di->disk.raid_disk; |
9105 | u->array = inst; | |
9106 | u->next = u + 1; | |
9107 | u++; | |
9108 | } | |
9109 | (u-1)->next = NULL; | |
9110 | *updates = mu; | |
9111 | ||
9112 | return rv; | |
9113 | } | |
9114 | ||
54c2c1ea | 9115 | static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u) |
8273f55e | 9116 | { |
54c2c1ea | 9117 | struct imsm_dev *dev = get_imsm_dev(super, idx); |
238c0a71 AK |
9118 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
9119 | struct imsm_map *new_map = get_imsm_map(&u->dev, MAP_0); | |
54c2c1ea DW |
9120 | struct disk_info *inf = get_disk_info(u); |
9121 | struct imsm_disk *disk; | |
8273f55e DW |
9122 | int i; |
9123 | int j; | |
8273f55e | 9124 | |
54c2c1ea | 9125 | for (i = 0; i < map->num_members; i++) { |
238c0a71 | 9126 | disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, MAP_X)); |
54c2c1ea DW |
9127 | for (j = 0; j < new_map->num_members; j++) |
9128 | if (serialcmp(disk->serial, inf[j].serial) == 0) | |
8273f55e DW |
9129 | return 1; |
9130 | } | |
9131 | ||
9132 | return 0; | |
9133 | } | |
9134 | ||
1a64be56 LM |
9135 | static struct dl *get_disk_super(struct intel_super *super, int major, int minor) |
9136 | { | |
594dc1b8 JS |
9137 | struct dl *dl; |
9138 | ||
1a64be56 | 9139 | for (dl = super->disks; dl; dl = dl->next) |
089f9d79 | 9140 | if (dl->major == major && dl->minor == minor) |
1a64be56 LM |
9141 | return dl; |
9142 | return NULL; | |
9143 | } | |
9144 | ||
9145 | static int remove_disk_super(struct intel_super *super, int major, int minor) | |
9146 | { | |
594dc1b8 | 9147 | struct dl *prev; |
1a64be56 LM |
9148 | struct dl *dl; |
9149 | ||
9150 | prev = NULL; | |
9151 | for (dl = super->disks; dl; dl = dl->next) { | |
089f9d79 | 9152 | if (dl->major == major && dl->minor == minor) { |
1a64be56 LM |
9153 | /* remove */ |
9154 | if (prev) | |
9155 | prev->next = dl->next; | |
9156 | else | |
9157 | super->disks = dl->next; | |
9158 | dl->next = NULL; | |
9159 | __free_imsm_disk(dl); | |
1ade5cc1 | 9160 | dprintf("removed %x:%x\n", major, minor); |
1a64be56 LM |
9161 | break; |
9162 | } | |
9163 | prev = dl; | |
9164 | } | |
9165 | return 0; | |
9166 | } | |
9167 | ||
f21e18ca | 9168 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index); |
ae6aad82 | 9169 | |
1a64be56 LM |
9170 | static int add_remove_disk_update(struct intel_super *super) |
9171 | { | |
9172 | int check_degraded = 0; | |
594dc1b8 JS |
9173 | struct dl *disk; |
9174 | ||
1a64be56 LM |
9175 | /* add/remove some spares to/from the metadata/contrainer */ |
9176 | while (super->disk_mgmt_list) { | |
9177 | struct dl *disk_cfg; | |
9178 | ||
9179 | disk_cfg = super->disk_mgmt_list; | |
9180 | super->disk_mgmt_list = disk_cfg->next; | |
9181 | disk_cfg->next = NULL; | |
9182 | ||
9183 | if (disk_cfg->action == DISK_ADD) { | |
9184 | disk_cfg->next = super->disks; | |
9185 | super->disks = disk_cfg; | |
9186 | check_degraded = 1; | |
1ade5cc1 N |
9187 | dprintf("added %x:%x\n", |
9188 | disk_cfg->major, disk_cfg->minor); | |
1a64be56 LM |
9189 | } else if (disk_cfg->action == DISK_REMOVE) { |
9190 | dprintf("Disk remove action processed: %x.%x\n", | |
9191 | disk_cfg->major, disk_cfg->minor); | |
9192 | disk = get_disk_super(super, | |
9193 | disk_cfg->major, | |
9194 | disk_cfg->minor); | |
9195 | if (disk) { | |
9196 | /* store action status */ | |
9197 | disk->action = DISK_REMOVE; | |
9198 | /* remove spare disks only */ | |
9199 | if (disk->index == -1) { | |
9200 | remove_disk_super(super, | |
9201 | disk_cfg->major, | |
9202 | disk_cfg->minor); | |
9203 | } | |
9204 | } | |
9205 | /* release allocate disk structure */ | |
9206 | __free_imsm_disk(disk_cfg); | |
9207 | } | |
9208 | } | |
9209 | return check_degraded; | |
9210 | } | |
9211 | ||
a29911da PC |
9212 | static int apply_reshape_migration_update(struct imsm_update_reshape_migration *u, |
9213 | struct intel_super *super, | |
9214 | void ***space_list) | |
9215 | { | |
9216 | struct intel_dev *id; | |
9217 | void **tofree = NULL; | |
9218 | int ret_val = 0; | |
9219 | ||
1ade5cc1 | 9220 | dprintf("(enter)\n"); |
089f9d79 | 9221 | if (u->subdev < 0 || u->subdev > 1) { |
a29911da PC |
9222 | dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev); |
9223 | return ret_val; | |
9224 | } | |
089f9d79 | 9225 | if (space_list == NULL || *space_list == NULL) { |
a29911da PC |
9226 | dprintf("imsm: Error: Memory is not allocated\n"); |
9227 | return ret_val; | |
9228 | } | |
9229 | ||
9230 | for (id = super->devlist ; id; id = id->next) { | |
9231 | if (id->index == (unsigned)u->subdev) { | |
9232 | struct imsm_dev *dev = get_imsm_dev(super, u->subdev); | |
9233 | struct imsm_map *map; | |
9234 | struct imsm_dev *new_dev = | |
9235 | (struct imsm_dev *)*space_list; | |
238c0a71 | 9236 | struct imsm_map *migr_map = get_imsm_map(dev, MAP_1); |
a29911da PC |
9237 | int to_state; |
9238 | struct dl *new_disk; | |
9239 | ||
9240 | if (new_dev == NULL) | |
9241 | return ret_val; | |
9242 | *space_list = **space_list; | |
9243 | memcpy(new_dev, dev, sizeof_imsm_dev(dev, 0)); | |
238c0a71 | 9244 | map = get_imsm_map(new_dev, MAP_0); |
a29911da PC |
9245 | if (migr_map) { |
9246 | dprintf("imsm: Error: migration in progress"); | |
9247 | return ret_val; | |
9248 | } | |
9249 | ||
9250 | to_state = map->map_state; | |
9251 | if ((u->new_level == 5) && (map->raid_level == 0)) { | |
9252 | map->num_members++; | |
9253 | /* this should not happen */ | |
9254 | if (u->new_disks[0] < 0) { | |
9255 | map->failed_disk_num = | |
9256 | map->num_members - 1; | |
9257 | to_state = IMSM_T_STATE_DEGRADED; | |
9258 | } else | |
9259 | to_state = IMSM_T_STATE_NORMAL; | |
9260 | } | |
8e59f3d8 | 9261 | migrate(new_dev, super, to_state, MIGR_GEN_MIGR); |
a29911da PC |
9262 | if (u->new_level > -1) |
9263 | map->raid_level = u->new_level; | |
238c0a71 | 9264 | migr_map = get_imsm_map(new_dev, MAP_1); |
a29911da PC |
9265 | if ((u->new_level == 5) && |
9266 | (migr_map->raid_level == 0)) { | |
9267 | int ord = map->num_members - 1; | |
9268 | migr_map->num_members--; | |
9269 | if (u->new_disks[0] < 0) | |
9270 | ord |= IMSM_ORD_REBUILD; | |
9271 | set_imsm_ord_tbl_ent(map, | |
9272 | map->num_members - 1, | |
9273 | ord); | |
9274 | } | |
9275 | id->dev = new_dev; | |
9276 | tofree = (void **)dev; | |
9277 | ||
4bba0439 PC |
9278 | /* update chunk size |
9279 | */ | |
06fb291a PB |
9280 | if (u->new_chunksize > 0) { |
9281 | unsigned long long num_data_stripes; | |
9529d343 MD |
9282 | struct imsm_map *dest_map = |
9283 | get_imsm_map(dev, MAP_0); | |
06fb291a | 9284 | int used_disks = |
9529d343 | 9285 | imsm_num_data_members(dest_map); |
06fb291a PB |
9286 | |
9287 | if (used_disks == 0) | |
9288 | return ret_val; | |
9289 | ||
4bba0439 PC |
9290 | map->blocks_per_strip = |
9291 | __cpu_to_le16(u->new_chunksize * 2); | |
06fb291a | 9292 | num_data_stripes = |
fcc2c9da | 9293 | imsm_dev_size(dev) / used_disks; |
06fb291a PB |
9294 | num_data_stripes /= map->blocks_per_strip; |
9295 | num_data_stripes /= map->num_domains; | |
9296 | set_num_data_stripes(map, num_data_stripes); | |
9297 | } | |
4bba0439 | 9298 | |
44490938 MD |
9299 | /* ensure blocks_per_member has valid value |
9300 | */ | |
9301 | set_blocks_per_member(map, | |
9302 | per_dev_array_size(map) + | |
9303 | NUM_BLOCKS_DIRTY_STRIPE_REGION); | |
9304 | ||
a29911da PC |
9305 | /* add disk |
9306 | */ | |
089f9d79 JS |
9307 | if (u->new_level != 5 || migr_map->raid_level != 0 || |
9308 | migr_map->raid_level == map->raid_level) | |
a29911da PC |
9309 | goto skip_disk_add; |
9310 | ||
9311 | if (u->new_disks[0] >= 0) { | |
9312 | /* use passes spare | |
9313 | */ | |
9314 | new_disk = get_disk_super(super, | |
9315 | major(u->new_disks[0]), | |
9316 | minor(u->new_disks[0])); | |
7a862a02 | 9317 | dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n", |
a29911da PC |
9318 | major(u->new_disks[0]), |
9319 | minor(u->new_disks[0]), | |
9320 | new_disk, new_disk->index); | |
9321 | if (new_disk == NULL) | |
9322 | goto error_disk_add; | |
9323 | ||
9324 | new_disk->index = map->num_members - 1; | |
9325 | /* slot to fill in autolayout | |
9326 | */ | |
9327 | new_disk->raiddisk = new_disk->index; | |
9328 | new_disk->disk.status |= CONFIGURED_DISK; | |
9329 | new_disk->disk.status &= ~SPARE_DISK; | |
9330 | } else | |
9331 | goto error_disk_add; | |
9332 | ||
9333 | skip_disk_add: | |
9334 | *tofree = *space_list; | |
9335 | /* calculate new size | |
9336 | */ | |
f3871fdc | 9337 | imsm_set_array_size(new_dev, -1); |
a29911da PC |
9338 | |
9339 | ret_val = 1; | |
9340 | } | |
9341 | } | |
9342 | ||
9343 | if (tofree) | |
9344 | *space_list = tofree; | |
9345 | return ret_val; | |
9346 | ||
9347 | error_disk_add: | |
9348 | dprintf("Error: imsm: Cannot find disk.\n"); | |
9349 | return ret_val; | |
9350 | } | |
9351 | ||
f3871fdc AK |
9352 | static int apply_size_change_update(struct imsm_update_size_change *u, |
9353 | struct intel_super *super) | |
9354 | { | |
9355 | struct intel_dev *id; | |
9356 | int ret_val = 0; | |
9357 | ||
1ade5cc1 | 9358 | dprintf("(enter)\n"); |
089f9d79 | 9359 | if (u->subdev < 0 || u->subdev > 1) { |
f3871fdc AK |
9360 | dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev); |
9361 | return ret_val; | |
9362 | } | |
9363 | ||
9364 | for (id = super->devlist ; id; id = id->next) { | |
9365 | if (id->index == (unsigned)u->subdev) { | |
9366 | struct imsm_dev *dev = get_imsm_dev(super, u->subdev); | |
9367 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
9529d343 | 9368 | int used_disks = imsm_num_data_members(map); |
f3871fdc | 9369 | unsigned long long blocks_per_member; |
06fb291a | 9370 | unsigned long long num_data_stripes; |
44490938 MD |
9371 | unsigned long long new_size_per_disk; |
9372 | ||
9373 | if (used_disks == 0) | |
9374 | return 0; | |
f3871fdc AK |
9375 | |
9376 | /* calculate new size | |
9377 | */ | |
44490938 MD |
9378 | new_size_per_disk = u->new_size / used_disks; |
9379 | blocks_per_member = new_size_per_disk + | |
9380 | NUM_BLOCKS_DIRTY_STRIPE_REGION; | |
9381 | num_data_stripes = new_size_per_disk / | |
06fb291a PB |
9382 | map->blocks_per_strip; |
9383 | num_data_stripes /= map->num_domains; | |
9384 | dprintf("(size: %llu, blocks per member: %llu, num_data_stipes: %llu)\n", | |
44490938 | 9385 | u->new_size, new_size_per_disk, |
06fb291a | 9386 | num_data_stripes); |
f3871fdc | 9387 | set_blocks_per_member(map, blocks_per_member); |
06fb291a | 9388 | set_num_data_stripes(map, num_data_stripes); |
f3871fdc AK |
9389 | imsm_set_array_size(dev, u->new_size); |
9390 | ||
9391 | ret_val = 1; | |
9392 | break; | |
9393 | } | |
9394 | } | |
9395 | ||
9396 | return ret_val; | |
9397 | } | |
9398 | ||
061d7da3 | 9399 | static int apply_update_activate_spare(struct imsm_update_activate_spare *u, |
ca9de185 | 9400 | struct intel_super *super, |
061d7da3 LO |
9401 | struct active_array *active_array) |
9402 | { | |
9403 | struct imsm_super *mpb = super->anchor; | |
9404 | struct imsm_dev *dev = get_imsm_dev(super, u->array); | |
238c0a71 | 9405 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
061d7da3 LO |
9406 | struct imsm_map *migr_map; |
9407 | struct active_array *a; | |
9408 | struct imsm_disk *disk; | |
9409 | __u8 to_state; | |
9410 | struct dl *dl; | |
9411 | unsigned int found; | |
9412 | int failed; | |
5961eeec | 9413 | int victim; |
061d7da3 | 9414 | int i; |
5961eeec | 9415 | int second_map_created = 0; |
061d7da3 | 9416 | |
5961eeec | 9417 | for (; u; u = u->next) { |
238c0a71 | 9418 | victim = get_imsm_disk_idx(dev, u->slot, MAP_X); |
061d7da3 | 9419 | |
5961eeec | 9420 | if (victim < 0) |
9421 | return 0; | |
061d7da3 | 9422 | |
5961eeec | 9423 | for (dl = super->disks; dl; dl = dl->next) |
9424 | if (dl == u->dl) | |
9425 | break; | |
061d7da3 | 9426 | |
5961eeec | 9427 | if (!dl) { |
7a862a02 | 9428 | pr_err("error: imsm_activate_spare passed an unknown disk (index: %d)\n", |
5961eeec | 9429 | u->dl->index); |
9430 | return 0; | |
9431 | } | |
061d7da3 | 9432 | |
5961eeec | 9433 | /* count failures (excluding rebuilds and the victim) |
9434 | * to determine map[0] state | |
9435 | */ | |
9436 | failed = 0; | |
9437 | for (i = 0; i < map->num_members; i++) { | |
9438 | if (i == u->slot) | |
9439 | continue; | |
9440 | disk = get_imsm_disk(super, | |
238c0a71 | 9441 | get_imsm_disk_idx(dev, i, MAP_X)); |
5961eeec | 9442 | if (!disk || is_failed(disk)) |
9443 | failed++; | |
9444 | } | |
061d7da3 | 9445 | |
5961eeec | 9446 | /* adding a pristine spare, assign a new index */ |
9447 | if (dl->index < 0) { | |
9448 | dl->index = super->anchor->num_disks; | |
9449 | super->anchor->num_disks++; | |
9450 | } | |
9451 | disk = &dl->disk; | |
9452 | disk->status |= CONFIGURED_DISK; | |
9453 | disk->status &= ~SPARE_DISK; | |
9454 | ||
9455 | /* mark rebuild */ | |
238c0a71 | 9456 | to_state = imsm_check_degraded(super, dev, failed, MAP_0); |
5961eeec | 9457 | if (!second_map_created) { |
9458 | second_map_created = 1; | |
9459 | map->map_state = IMSM_T_STATE_DEGRADED; | |
9460 | migrate(dev, super, to_state, MIGR_REBUILD); | |
9461 | } else | |
9462 | map->map_state = to_state; | |
238c0a71 | 9463 | migr_map = get_imsm_map(dev, MAP_1); |
5961eeec | 9464 | set_imsm_ord_tbl_ent(map, u->slot, dl->index); |
9465 | set_imsm_ord_tbl_ent(migr_map, u->slot, | |
9466 | dl->index | IMSM_ORD_REBUILD); | |
9467 | ||
9468 | /* update the family_num to mark a new container | |
9469 | * generation, being careful to record the existing | |
9470 | * family_num in orig_family_num to clean up after | |
9471 | * earlier mdadm versions that neglected to set it. | |
9472 | */ | |
9473 | if (mpb->orig_family_num == 0) | |
9474 | mpb->orig_family_num = mpb->family_num; | |
9475 | mpb->family_num += super->random; | |
9476 | ||
9477 | /* count arrays using the victim in the metadata */ | |
9478 | found = 0; | |
9479 | for (a = active_array; a ; a = a->next) { | |
9480 | dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 9481 | map = get_imsm_map(dev, MAP_0); |
061d7da3 | 9482 | |
5961eeec | 9483 | if (get_imsm_disk_slot(map, victim) >= 0) |
9484 | found++; | |
9485 | } | |
061d7da3 | 9486 | |
5961eeec | 9487 | /* delete the victim if it is no longer being |
9488 | * utilized anywhere | |
061d7da3 | 9489 | */ |
5961eeec | 9490 | if (!found) { |
9491 | struct dl **dlp; | |
061d7da3 | 9492 | |
5961eeec | 9493 | /* We know that 'manager' isn't touching anything, |
9494 | * so it is safe to delete | |
9495 | */ | |
9496 | for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next) | |
061d7da3 LO |
9497 | if ((*dlp)->index == victim) |
9498 | break; | |
5961eeec | 9499 | |
9500 | /* victim may be on the missing list */ | |
9501 | if (!*dlp) | |
9502 | for (dlp = &super->missing; *dlp; | |
9503 | dlp = &(*dlp)->next) | |
9504 | if ((*dlp)->index == victim) | |
9505 | break; | |
9506 | imsm_delete(super, dlp, victim); | |
9507 | } | |
061d7da3 LO |
9508 | } |
9509 | ||
9510 | return 1; | |
9511 | } | |
a29911da | 9512 | |
2e5dc010 N |
9513 | static int apply_reshape_container_disks_update(struct imsm_update_reshape *u, |
9514 | struct intel_super *super, | |
9515 | void ***space_list) | |
9516 | { | |
9517 | struct dl *new_disk; | |
9518 | struct intel_dev *id; | |
9519 | int i; | |
9520 | int delta_disks = u->new_raid_disks - u->old_raid_disks; | |
ee4beede | 9521 | int disk_count = u->old_raid_disks; |
2e5dc010 N |
9522 | void **tofree = NULL; |
9523 | int devices_to_reshape = 1; | |
9524 | struct imsm_super *mpb = super->anchor; | |
9525 | int ret_val = 0; | |
d098291a | 9526 | unsigned int dev_id; |
2e5dc010 | 9527 | |
1ade5cc1 | 9528 | dprintf("(enter)\n"); |
2e5dc010 N |
9529 | |
9530 | /* enable spares to use in array */ | |
9531 | for (i = 0; i < delta_disks; i++) { | |
9532 | new_disk = get_disk_super(super, | |
9533 | major(u->new_disks[i]), | |
9534 | minor(u->new_disks[i])); | |
7a862a02 | 9535 | dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n", |
2e5dc010 N |
9536 | major(u->new_disks[i]), minor(u->new_disks[i]), |
9537 | new_disk, new_disk->index); | |
089f9d79 JS |
9538 | if (new_disk == NULL || |
9539 | (new_disk->index >= 0 && | |
9540 | new_disk->index < u->old_raid_disks)) | |
2e5dc010 | 9541 | goto update_reshape_exit; |
ee4beede | 9542 | new_disk->index = disk_count++; |
2e5dc010 N |
9543 | /* slot to fill in autolayout |
9544 | */ | |
9545 | new_disk->raiddisk = new_disk->index; | |
9546 | new_disk->disk.status |= | |
9547 | CONFIGURED_DISK; | |
9548 | new_disk->disk.status &= ~SPARE_DISK; | |
9549 | } | |
9550 | ||
ed7333bd AK |
9551 | dprintf("imsm: volume set mpb->num_raid_devs = %i\n", |
9552 | mpb->num_raid_devs); | |
2e5dc010 N |
9553 | /* manage changes in volume |
9554 | */ | |
d098291a | 9555 | for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) { |
2e5dc010 N |
9556 | void **sp = *space_list; |
9557 | struct imsm_dev *newdev; | |
9558 | struct imsm_map *newmap, *oldmap; | |
9559 | ||
d098291a AK |
9560 | for (id = super->devlist ; id; id = id->next) { |
9561 | if (id->index == dev_id) | |
9562 | break; | |
9563 | } | |
9564 | if (id == NULL) | |
9565 | break; | |
2e5dc010 N |
9566 | if (!sp) |
9567 | continue; | |
9568 | *space_list = *sp; | |
9569 | newdev = (void*)sp; | |
9570 | /* Copy the dev, but not (all of) the map */ | |
9571 | memcpy(newdev, id->dev, sizeof(*newdev)); | |
238c0a71 AK |
9572 | oldmap = get_imsm_map(id->dev, MAP_0); |
9573 | newmap = get_imsm_map(newdev, MAP_0); | |
2e5dc010 N |
9574 | /* Copy the current map */ |
9575 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); | |
9576 | /* update one device only | |
9577 | */ | |
9578 | if (devices_to_reshape) { | |
ed7333bd AK |
9579 | dprintf("imsm: modifying subdev: %i\n", |
9580 | id->index); | |
2e5dc010 N |
9581 | devices_to_reshape--; |
9582 | newdev->vol.migr_state = 1; | |
9583 | newdev->vol.curr_migr_unit = 0; | |
ea672ee1 | 9584 | set_migr_type(newdev, MIGR_GEN_MIGR); |
2e5dc010 N |
9585 | newmap->num_members = u->new_raid_disks; |
9586 | for (i = 0; i < delta_disks; i++) { | |
9587 | set_imsm_ord_tbl_ent(newmap, | |
9588 | u->old_raid_disks + i, | |
9589 | u->old_raid_disks + i); | |
9590 | } | |
9591 | /* New map is correct, now need to save old map | |
9592 | */ | |
238c0a71 | 9593 | newmap = get_imsm_map(newdev, MAP_1); |
2e5dc010 N |
9594 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); |
9595 | ||
f3871fdc | 9596 | imsm_set_array_size(newdev, -1); |
2e5dc010 N |
9597 | } |
9598 | ||
9599 | sp = (void **)id->dev; | |
9600 | id->dev = newdev; | |
9601 | *sp = tofree; | |
9602 | tofree = sp; | |
8e59f3d8 AK |
9603 | |
9604 | /* Clear migration record */ | |
9605 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
2e5dc010 | 9606 | } |
819bc634 AK |
9607 | if (tofree) |
9608 | *space_list = tofree; | |
2e5dc010 N |
9609 | ret_val = 1; |
9610 | ||
9611 | update_reshape_exit: | |
9612 | ||
9613 | return ret_val; | |
9614 | } | |
9615 | ||
bb025c2f | 9616 | static int apply_takeover_update(struct imsm_update_takeover *u, |
8ca6df95 KW |
9617 | struct intel_super *super, |
9618 | void ***space_list) | |
bb025c2f KW |
9619 | { |
9620 | struct imsm_dev *dev = NULL; | |
8ca6df95 KW |
9621 | struct intel_dev *dv; |
9622 | struct imsm_dev *dev_new; | |
bb025c2f KW |
9623 | struct imsm_map *map; |
9624 | struct dl *dm, *du; | |
8ca6df95 | 9625 | int i; |
bb025c2f KW |
9626 | |
9627 | for (dv = super->devlist; dv; dv = dv->next) | |
9628 | if (dv->index == (unsigned int)u->subarray) { | |
9629 | dev = dv->dev; | |
9630 | break; | |
9631 | } | |
9632 | ||
9633 | if (dev == NULL) | |
9634 | return 0; | |
9635 | ||
238c0a71 | 9636 | map = get_imsm_map(dev, MAP_0); |
bb025c2f KW |
9637 | |
9638 | if (u->direction == R10_TO_R0) { | |
06fb291a PB |
9639 | unsigned long long num_data_stripes; |
9640 | ||
43d5ec18 | 9641 | /* Number of failed disks must be half of initial disk number */ |
3b451610 AK |
9642 | if (imsm_count_failed(super, dev, MAP_0) != |
9643 | (map->num_members / 2)) | |
43d5ec18 KW |
9644 | return 0; |
9645 | ||
bb025c2f KW |
9646 | /* iterate through devices to mark removed disks as spare */ |
9647 | for (dm = super->disks; dm; dm = dm->next) { | |
9648 | if (dm->disk.status & FAILED_DISK) { | |
9649 | int idx = dm->index; | |
9650 | /* update indexes on the disk list */ | |
9651 | /* FIXME this loop-with-the-loop looks wrong, I'm not convinced | |
9652 | the index values will end up being correct.... NB */ | |
9653 | for (du = super->disks; du; du = du->next) | |
9654 | if (du->index > idx) | |
9655 | du->index--; | |
9656 | /* mark as spare disk */ | |
a8619d23 | 9657 | mark_spare(dm); |
bb025c2f KW |
9658 | } |
9659 | } | |
bb025c2f KW |
9660 | /* update map */ |
9661 | map->num_members = map->num_members / 2; | |
9662 | map->map_state = IMSM_T_STATE_NORMAL; | |
9663 | map->num_domains = 1; | |
9664 | map->raid_level = 0; | |
9665 | map->failed_disk_num = -1; | |
4a353e6e RS |
9666 | num_data_stripes = imsm_dev_size(dev) / 2; |
9667 | num_data_stripes /= map->blocks_per_strip; | |
9668 | set_num_data_stripes(map, num_data_stripes); | |
bb025c2f KW |
9669 | } |
9670 | ||
8ca6df95 KW |
9671 | if (u->direction == R0_TO_R10) { |
9672 | void **space; | |
4a353e6e RS |
9673 | unsigned long long num_data_stripes; |
9674 | ||
8ca6df95 KW |
9675 | /* update slots in current disk list */ |
9676 | for (dm = super->disks; dm; dm = dm->next) { | |
9677 | if (dm->index >= 0) | |
9678 | dm->index *= 2; | |
9679 | } | |
9680 | /* create new *missing* disks */ | |
9681 | for (i = 0; i < map->num_members; i++) { | |
9682 | space = *space_list; | |
9683 | if (!space) | |
9684 | continue; | |
9685 | *space_list = *space; | |
9686 | du = (void *)space; | |
9687 | memcpy(du, super->disks, sizeof(*du)); | |
8ca6df95 KW |
9688 | du->fd = -1; |
9689 | du->minor = 0; | |
9690 | du->major = 0; | |
9691 | du->index = (i * 2) + 1; | |
9692 | sprintf((char *)du->disk.serial, | |
9693 | " MISSING_%d", du->index); | |
9694 | sprintf((char *)du->serial, | |
9695 | "MISSING_%d", du->index); | |
9696 | du->next = super->missing; | |
9697 | super->missing = du; | |
9698 | } | |
9699 | /* create new dev and map */ | |
9700 | space = *space_list; | |
9701 | if (!space) | |
9702 | return 0; | |
9703 | *space_list = *space; | |
9704 | dev_new = (void *)space; | |
9705 | memcpy(dev_new, dev, sizeof(*dev)); | |
9706 | /* update new map */ | |
238c0a71 | 9707 | map = get_imsm_map(dev_new, MAP_0); |
8ca6df95 | 9708 | map->num_members = map->num_members * 2; |
1a2487c2 | 9709 | map->map_state = IMSM_T_STATE_DEGRADED; |
8ca6df95 KW |
9710 | map->num_domains = 2; |
9711 | map->raid_level = 1; | |
4a353e6e RS |
9712 | num_data_stripes = imsm_dev_size(dev) / 2; |
9713 | num_data_stripes /= map->blocks_per_strip; | |
9714 | num_data_stripes /= map->num_domains; | |
9715 | set_num_data_stripes(map, num_data_stripes); | |
9716 | ||
8ca6df95 KW |
9717 | /* replace dev<->dev_new */ |
9718 | dv->dev = dev_new; | |
9719 | } | |
bb025c2f KW |
9720 | /* update disk order table */ |
9721 | for (du = super->disks; du; du = du->next) | |
9722 | if (du->index >= 0) | |
9723 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
8ca6df95 | 9724 | for (du = super->missing; du; du = du->next) |
1a2487c2 KW |
9725 | if (du->index >= 0) { |
9726 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
4c9e8c1e | 9727 | mark_missing(super, dv->dev, &du->disk, du->index); |
1a2487c2 | 9728 | } |
bb025c2f KW |
9729 | |
9730 | return 1; | |
9731 | } | |
9732 | ||
e8319a19 DW |
9733 | static void imsm_process_update(struct supertype *st, |
9734 | struct metadata_update *update) | |
9735 | { | |
9736 | /** | |
9737 | * crack open the metadata_update envelope to find the update record | |
9738 | * update can be one of: | |
d195167d AK |
9739 | * update_reshape_container_disks - all the arrays in the container |
9740 | * are being reshaped to have more devices. We need to mark | |
9741 | * the arrays for general migration and convert selected spares | |
9742 | * into active devices. | |
9743 | * update_activate_spare - a spare device has replaced a failed | |
1011e834 N |
9744 | * device in an array, update the disk_ord_tbl. If this disk is |
9745 | * present in all member arrays then also clear the SPARE_DISK | |
9746 | * flag | |
d195167d AK |
9747 | * update_create_array |
9748 | * update_kill_array | |
9749 | * update_rename_array | |
9750 | * update_add_remove_disk | |
e8319a19 DW |
9751 | */ |
9752 | struct intel_super *super = st->sb; | |
4d7b1503 | 9753 | struct imsm_super *mpb; |
e8319a19 DW |
9754 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
9755 | ||
4d7b1503 DW |
9756 | /* update requires a larger buf but the allocation failed */ |
9757 | if (super->next_len && !super->next_buf) { | |
9758 | super->next_len = 0; | |
9759 | return; | |
9760 | } | |
9761 | ||
9762 | if (super->next_buf) { | |
9763 | memcpy(super->next_buf, super->buf, super->len); | |
9764 | free(super->buf); | |
9765 | super->len = super->next_len; | |
9766 | super->buf = super->next_buf; | |
9767 | ||
9768 | super->next_len = 0; | |
9769 | super->next_buf = NULL; | |
9770 | } | |
9771 | ||
9772 | mpb = super->anchor; | |
9773 | ||
e8319a19 | 9774 | switch (type) { |
0ec5d470 AK |
9775 | case update_general_migration_checkpoint: { |
9776 | struct intel_dev *id; | |
9777 | struct imsm_update_general_migration_checkpoint *u = | |
9778 | (void *)update->buf; | |
9779 | ||
1ade5cc1 | 9780 | dprintf("called for update_general_migration_checkpoint\n"); |
0ec5d470 AK |
9781 | |
9782 | /* find device under general migration */ | |
9783 | for (id = super->devlist ; id; id = id->next) { | |
9784 | if (is_gen_migration(id->dev)) { | |
9785 | id->dev->vol.curr_migr_unit = | |
9786 | __cpu_to_le32(u->curr_migr_unit); | |
9787 | super->updates_pending++; | |
9788 | } | |
9789 | } | |
9790 | break; | |
9791 | } | |
bb025c2f KW |
9792 | case update_takeover: { |
9793 | struct imsm_update_takeover *u = (void *)update->buf; | |
1a2487c2 KW |
9794 | if (apply_takeover_update(u, super, &update->space_list)) { |
9795 | imsm_update_version_info(super); | |
bb025c2f | 9796 | super->updates_pending++; |
1a2487c2 | 9797 | } |
bb025c2f KW |
9798 | break; |
9799 | } | |
9800 | ||
78b10e66 | 9801 | case update_reshape_container_disks: { |
d195167d | 9802 | struct imsm_update_reshape *u = (void *)update->buf; |
2e5dc010 N |
9803 | if (apply_reshape_container_disks_update( |
9804 | u, super, &update->space_list)) | |
9805 | super->updates_pending++; | |
78b10e66 N |
9806 | break; |
9807 | } | |
48c5303a | 9808 | case update_reshape_migration: { |
a29911da PC |
9809 | struct imsm_update_reshape_migration *u = (void *)update->buf; |
9810 | if (apply_reshape_migration_update( | |
9811 | u, super, &update->space_list)) | |
9812 | super->updates_pending++; | |
48c5303a PC |
9813 | break; |
9814 | } | |
f3871fdc AK |
9815 | case update_size_change: { |
9816 | struct imsm_update_size_change *u = (void *)update->buf; | |
9817 | if (apply_size_change_update(u, super)) | |
9818 | super->updates_pending++; | |
9819 | break; | |
9820 | } | |
e8319a19 | 9821 | case update_activate_spare: { |
1011e834 | 9822 | struct imsm_update_activate_spare *u = (void *) update->buf; |
061d7da3 LO |
9823 | if (apply_update_activate_spare(u, super, st->arrays)) |
9824 | super->updates_pending++; | |
8273f55e DW |
9825 | break; |
9826 | } | |
9827 | case update_create_array: { | |
9828 | /* someone wants to create a new array, we need to be aware of | |
9829 | * a few races/collisions: | |
9830 | * 1/ 'Create' called by two separate instances of mdadm | |
9831 | * 2/ 'Create' versus 'activate_spare': mdadm has chosen | |
9832 | * devices that have since been assimilated via | |
9833 | * activate_spare. | |
9834 | * In the event this update can not be carried out mdadm will | |
9835 | * (FIX ME) notice that its update did not take hold. | |
9836 | */ | |
9837 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 9838 | struct intel_dev *dv; |
8273f55e DW |
9839 | struct imsm_dev *dev; |
9840 | struct imsm_map *map, *new_map; | |
9841 | unsigned long long start, end; | |
9842 | unsigned long long new_start, new_end; | |
9843 | int i; | |
54c2c1ea DW |
9844 | struct disk_info *inf; |
9845 | struct dl *dl; | |
8273f55e DW |
9846 | |
9847 | /* handle racing creates: first come first serve */ | |
9848 | if (u->dev_idx < mpb->num_raid_devs) { | |
1ade5cc1 | 9849 | dprintf("subarray %d already defined\n", u->dev_idx); |
ba2de7ba | 9850 | goto create_error; |
8273f55e DW |
9851 | } |
9852 | ||
9853 | /* check update is next in sequence */ | |
9854 | if (u->dev_idx != mpb->num_raid_devs) { | |
1ade5cc1 N |
9855 | dprintf("can not create array %d expected index %d\n", |
9856 | u->dev_idx, mpb->num_raid_devs); | |
ba2de7ba | 9857 | goto create_error; |
8273f55e DW |
9858 | } |
9859 | ||
238c0a71 | 9860 | new_map = get_imsm_map(&u->dev, MAP_0); |
5551b113 | 9861 | new_start = pba_of_lba0(new_map); |
44490938 | 9862 | new_end = new_start + per_dev_array_size(new_map); |
54c2c1ea | 9863 | inf = get_disk_info(u); |
8273f55e DW |
9864 | |
9865 | /* handle activate_spare versus create race: | |
9866 | * check to make sure that overlapping arrays do not include | |
9867 | * overalpping disks | |
9868 | */ | |
9869 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
949c47a0 | 9870 | dev = get_imsm_dev(super, i); |
238c0a71 | 9871 | map = get_imsm_map(dev, MAP_0); |
5551b113 | 9872 | start = pba_of_lba0(map); |
44490938 | 9873 | end = start + per_dev_array_size(map); |
8273f55e DW |
9874 | if ((new_start >= start && new_start <= end) || |
9875 | (start >= new_start && start <= new_end)) | |
54c2c1ea DW |
9876 | /* overlap */; |
9877 | else | |
9878 | continue; | |
9879 | ||
9880 | if (disks_overlap(super, i, u)) { | |
1ade5cc1 | 9881 | dprintf("arrays overlap\n"); |
ba2de7ba | 9882 | goto create_error; |
8273f55e DW |
9883 | } |
9884 | } | |
8273f55e | 9885 | |
949c47a0 DW |
9886 | /* check that prepare update was successful */ |
9887 | if (!update->space) { | |
1ade5cc1 | 9888 | dprintf("prepare update failed\n"); |
ba2de7ba | 9889 | goto create_error; |
949c47a0 DW |
9890 | } |
9891 | ||
54c2c1ea DW |
9892 | /* check that all disks are still active before committing |
9893 | * changes. FIXME: could we instead handle this by creating a | |
9894 | * degraded array? That's probably not what the user expects, | |
9895 | * so better to drop this update on the floor. | |
9896 | */ | |
9897 | for (i = 0; i < new_map->num_members; i++) { | |
9898 | dl = serial_to_dl(inf[i].serial, super); | |
9899 | if (!dl) { | |
1ade5cc1 | 9900 | dprintf("disk disappeared\n"); |
ba2de7ba | 9901 | goto create_error; |
54c2c1ea | 9902 | } |
949c47a0 DW |
9903 | } |
9904 | ||
8273f55e | 9905 | super->updates_pending++; |
54c2c1ea DW |
9906 | |
9907 | /* convert spares to members and fixup ord_tbl */ | |
9908 | for (i = 0; i < new_map->num_members; i++) { | |
9909 | dl = serial_to_dl(inf[i].serial, super); | |
9910 | if (dl->index == -1) { | |
9911 | dl->index = mpb->num_disks; | |
9912 | mpb->num_disks++; | |
9913 | dl->disk.status |= CONFIGURED_DISK; | |
9914 | dl->disk.status &= ~SPARE_DISK; | |
9915 | } | |
9916 | set_imsm_ord_tbl_ent(new_map, i, dl->index); | |
9917 | } | |
9918 | ||
ba2de7ba DW |
9919 | dv = update->space; |
9920 | dev = dv->dev; | |
949c47a0 DW |
9921 | update->space = NULL; |
9922 | imsm_copy_dev(dev, &u->dev); | |
ba2de7ba DW |
9923 | dv->index = u->dev_idx; |
9924 | dv->next = super->devlist; | |
9925 | super->devlist = dv; | |
8273f55e | 9926 | mpb->num_raid_devs++; |
8273f55e | 9927 | |
4d1313e9 | 9928 | imsm_update_version_info(super); |
8273f55e | 9929 | break; |
ba2de7ba DW |
9930 | create_error: |
9931 | /* mdmon knows how to release update->space, but not | |
9932 | * ((struct intel_dev *) update->space)->dev | |
9933 | */ | |
9934 | if (update->space) { | |
9935 | dv = update->space; | |
9936 | free(dv->dev); | |
9937 | } | |
8273f55e | 9938 | break; |
e8319a19 | 9939 | } |
33414a01 DW |
9940 | case update_kill_array: { |
9941 | struct imsm_update_kill_array *u = (void *) update->buf; | |
9942 | int victim = u->dev_idx; | |
9943 | struct active_array *a; | |
9944 | struct intel_dev **dp; | |
9945 | struct imsm_dev *dev; | |
9946 | ||
9947 | /* sanity check that we are not affecting the uuid of | |
9948 | * active arrays, or deleting an active array | |
9949 | * | |
9950 | * FIXME when immutable ids are available, but note that | |
9951 | * we'll also need to fixup the invalidated/active | |
9952 | * subarray indexes in mdstat | |
9953 | */ | |
9954 | for (a = st->arrays; a; a = a->next) | |
9955 | if (a->info.container_member >= victim) | |
9956 | break; | |
9957 | /* by definition if mdmon is running at least one array | |
9958 | * is active in the container, so checking | |
9959 | * mpb->num_raid_devs is just extra paranoia | |
9960 | */ | |
9961 | dev = get_imsm_dev(super, victim); | |
9962 | if (a || !dev || mpb->num_raid_devs == 1) { | |
9963 | dprintf("failed to delete subarray-%d\n", victim); | |
9964 | break; | |
9965 | } | |
9966 | ||
9967 | for (dp = &super->devlist; *dp;) | |
f21e18ca | 9968 | if ((*dp)->index == (unsigned)super->current_vol) { |
33414a01 DW |
9969 | *dp = (*dp)->next; |
9970 | } else { | |
f21e18ca | 9971 | if ((*dp)->index > (unsigned)victim) |
33414a01 DW |
9972 | (*dp)->index--; |
9973 | dp = &(*dp)->next; | |
9974 | } | |
9975 | mpb->num_raid_devs--; | |
9976 | super->updates_pending++; | |
9977 | break; | |
9978 | } | |
aa534678 DW |
9979 | case update_rename_array: { |
9980 | struct imsm_update_rename_array *u = (void *) update->buf; | |
9981 | char name[MAX_RAID_SERIAL_LEN+1]; | |
9982 | int target = u->dev_idx; | |
9983 | struct active_array *a; | |
9984 | struct imsm_dev *dev; | |
9985 | ||
9986 | /* sanity check that we are not affecting the uuid of | |
9987 | * an active array | |
9988 | */ | |
40659392 | 9989 | memset(name, 0, sizeof(name)); |
aa534678 DW |
9990 | snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name); |
9991 | name[MAX_RAID_SERIAL_LEN] = '\0'; | |
9992 | for (a = st->arrays; a; a = a->next) | |
9993 | if (a->info.container_member == target) | |
9994 | break; | |
9995 | dev = get_imsm_dev(super, u->dev_idx); | |
9996 | if (a || !dev || !check_name(super, name, 1)) { | |
9997 | dprintf("failed to rename subarray-%d\n", target); | |
9998 | break; | |
9999 | } | |
10000 | ||
40659392 | 10001 | memcpy(dev->volume, name, MAX_RAID_SERIAL_LEN); |
aa534678 DW |
10002 | super->updates_pending++; |
10003 | break; | |
10004 | } | |
1a64be56 | 10005 | case update_add_remove_disk: { |
43dad3d6 | 10006 | /* we may be able to repair some arrays if disks are |
095b8088 | 10007 | * being added, check the status of add_remove_disk |
1a64be56 LM |
10008 | * if discs has been added. |
10009 | */ | |
10010 | if (add_remove_disk_update(super)) { | |
43dad3d6 | 10011 | struct active_array *a; |
072b727f DW |
10012 | |
10013 | super->updates_pending++; | |
1a64be56 | 10014 | for (a = st->arrays; a; a = a->next) |
43dad3d6 DW |
10015 | a->check_degraded = 1; |
10016 | } | |
43dad3d6 | 10017 | break; |
e8319a19 | 10018 | } |
bbab0940 TM |
10019 | case update_prealloc_badblocks_mem: |
10020 | break; | |
e6e9dd3f AP |
10021 | case update_rwh_policy: { |
10022 | struct imsm_update_rwh_policy *u = (void *)update->buf; | |
10023 | int target = u->dev_idx; | |
10024 | struct imsm_dev *dev = get_imsm_dev(super, target); | |
10025 | if (!dev) { | |
10026 | dprintf("could not find subarray-%d\n", target); | |
10027 | break; | |
10028 | } | |
10029 | ||
10030 | if (dev->rwh_policy != u->new_policy) { | |
10031 | dev->rwh_policy = u->new_policy; | |
10032 | super->updates_pending++; | |
10033 | } | |
10034 | break; | |
10035 | } | |
1a64be56 | 10036 | default: |
ebf3be99 | 10037 | pr_err("error: unsupported process update type:(type: %d)\n", type); |
1a64be56 | 10038 | } |
e8319a19 | 10039 | } |
88758e9d | 10040 | |
bc0b9d34 PC |
10041 | static struct mdinfo *get_spares_for_grow(struct supertype *st); |
10042 | ||
5fe6f031 N |
10043 | static int imsm_prepare_update(struct supertype *st, |
10044 | struct metadata_update *update) | |
8273f55e | 10045 | { |
949c47a0 | 10046 | /** |
4d7b1503 DW |
10047 | * Allocate space to hold new disk entries, raid-device entries or a new |
10048 | * mpb if necessary. The manager synchronously waits for updates to | |
10049 | * complete in the monitor, so new mpb buffers allocated here can be | |
10050 | * integrated by the monitor thread without worrying about live pointers | |
10051 | * in the manager thread. | |
8273f55e | 10052 | */ |
095b8088 | 10053 | enum imsm_update_type type; |
4d7b1503 | 10054 | struct intel_super *super = st->sb; |
f36a9ecd | 10055 | unsigned int sector_size = super->sector_size; |
4d7b1503 DW |
10056 | struct imsm_super *mpb = super->anchor; |
10057 | size_t buf_len; | |
10058 | size_t len = 0; | |
949c47a0 | 10059 | |
095b8088 N |
10060 | if (update->len < (int)sizeof(type)) |
10061 | return 0; | |
10062 | ||
10063 | type = *(enum imsm_update_type *) update->buf; | |
10064 | ||
949c47a0 | 10065 | switch (type) { |
0ec5d470 | 10066 | case update_general_migration_checkpoint: |
095b8088 N |
10067 | if (update->len < (int)sizeof(struct imsm_update_general_migration_checkpoint)) |
10068 | return 0; | |
1ade5cc1 | 10069 | dprintf("called for update_general_migration_checkpoint\n"); |
0ec5d470 | 10070 | break; |
abedf5fc KW |
10071 | case update_takeover: { |
10072 | struct imsm_update_takeover *u = (void *)update->buf; | |
095b8088 N |
10073 | if (update->len < (int)sizeof(*u)) |
10074 | return 0; | |
abedf5fc KW |
10075 | if (u->direction == R0_TO_R10) { |
10076 | void **tail = (void **)&update->space_list; | |
10077 | struct imsm_dev *dev = get_imsm_dev(super, u->subarray); | |
238c0a71 | 10078 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
abedf5fc KW |
10079 | int num_members = map->num_members; |
10080 | void *space; | |
10081 | int size, i; | |
abedf5fc KW |
10082 | /* allocate memory for added disks */ |
10083 | for (i = 0; i < num_members; i++) { | |
10084 | size = sizeof(struct dl); | |
503975b9 | 10085 | space = xmalloc(size); |
abedf5fc KW |
10086 | *tail = space; |
10087 | tail = space; | |
10088 | *tail = NULL; | |
10089 | } | |
10090 | /* allocate memory for new device */ | |
10091 | size = sizeof_imsm_dev(super->devlist->dev, 0) + | |
10092 | (num_members * sizeof(__u32)); | |
503975b9 N |
10093 | space = xmalloc(size); |
10094 | *tail = space; | |
10095 | tail = space; | |
10096 | *tail = NULL; | |
10097 | len = disks_to_mpb_size(num_members * 2); | |
abedf5fc KW |
10098 | } |
10099 | ||
10100 | break; | |
10101 | } | |
78b10e66 | 10102 | case update_reshape_container_disks: { |
d195167d AK |
10103 | /* Every raid device in the container is about to |
10104 | * gain some more devices, and we will enter a | |
10105 | * reconfiguration. | |
10106 | * So each 'imsm_map' will be bigger, and the imsm_vol | |
10107 | * will now hold 2 of them. | |
10108 | * Thus we need new 'struct imsm_dev' allocations sized | |
10109 | * as sizeof_imsm_dev but with more devices in both maps. | |
10110 | */ | |
10111 | struct imsm_update_reshape *u = (void *)update->buf; | |
10112 | struct intel_dev *dl; | |
10113 | void **space_tail = (void**)&update->space_list; | |
10114 | ||
095b8088 N |
10115 | if (update->len < (int)sizeof(*u)) |
10116 | return 0; | |
10117 | ||
1ade5cc1 | 10118 | dprintf("for update_reshape\n"); |
d195167d AK |
10119 | |
10120 | for (dl = super->devlist; dl; dl = dl->next) { | |
10121 | int size = sizeof_imsm_dev(dl->dev, 1); | |
10122 | void *s; | |
d677e0b8 AK |
10123 | if (u->new_raid_disks > u->old_raid_disks) |
10124 | size += sizeof(__u32)*2* | |
10125 | (u->new_raid_disks - u->old_raid_disks); | |
503975b9 | 10126 | s = xmalloc(size); |
d195167d AK |
10127 | *space_tail = s; |
10128 | space_tail = s; | |
10129 | *space_tail = NULL; | |
10130 | } | |
10131 | ||
10132 | len = disks_to_mpb_size(u->new_raid_disks); | |
10133 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
78b10e66 N |
10134 | break; |
10135 | } | |
48c5303a | 10136 | case update_reshape_migration: { |
bc0b9d34 PC |
10137 | /* for migration level 0->5 we need to add disks |
10138 | * so the same as for container operation we will copy | |
10139 | * device to the bigger location. | |
10140 | * in memory prepared device and new disk area are prepared | |
10141 | * for usage in process update | |
10142 | */ | |
10143 | struct imsm_update_reshape_migration *u = (void *)update->buf; | |
10144 | struct intel_dev *id; | |
10145 | void **space_tail = (void **)&update->space_list; | |
10146 | int size; | |
10147 | void *s; | |
10148 | int current_level = -1; | |
10149 | ||
095b8088 N |
10150 | if (update->len < (int)sizeof(*u)) |
10151 | return 0; | |
10152 | ||
1ade5cc1 | 10153 | dprintf("for update_reshape\n"); |
bc0b9d34 PC |
10154 | |
10155 | /* add space for bigger array in update | |
10156 | */ | |
10157 | for (id = super->devlist; id; id = id->next) { | |
10158 | if (id->index == (unsigned)u->subdev) { | |
10159 | size = sizeof_imsm_dev(id->dev, 1); | |
10160 | if (u->new_raid_disks > u->old_raid_disks) | |
10161 | size += sizeof(__u32)*2* | |
10162 | (u->new_raid_disks - u->old_raid_disks); | |
503975b9 | 10163 | s = xmalloc(size); |
bc0b9d34 PC |
10164 | *space_tail = s; |
10165 | space_tail = s; | |
10166 | *space_tail = NULL; | |
10167 | break; | |
10168 | } | |
10169 | } | |
10170 | if (update->space_list == NULL) | |
10171 | break; | |
10172 | ||
10173 | /* add space for disk in update | |
10174 | */ | |
10175 | size = sizeof(struct dl); | |
503975b9 | 10176 | s = xmalloc(size); |
bc0b9d34 PC |
10177 | *space_tail = s; |
10178 | space_tail = s; | |
10179 | *space_tail = NULL; | |
10180 | ||
10181 | /* add spare device to update | |
10182 | */ | |
10183 | for (id = super->devlist ; id; id = id->next) | |
10184 | if (id->index == (unsigned)u->subdev) { | |
10185 | struct imsm_dev *dev; | |
10186 | struct imsm_map *map; | |
10187 | ||
10188 | dev = get_imsm_dev(super, u->subdev); | |
238c0a71 | 10189 | map = get_imsm_map(dev, MAP_0); |
bc0b9d34 PC |
10190 | current_level = map->raid_level; |
10191 | break; | |
10192 | } | |
089f9d79 | 10193 | if (u->new_level == 5 && u->new_level != current_level) { |
bc0b9d34 PC |
10194 | struct mdinfo *spares; |
10195 | ||
10196 | spares = get_spares_for_grow(st); | |
10197 | if (spares) { | |
10198 | struct dl *dl; | |
10199 | struct mdinfo *dev; | |
10200 | ||
10201 | dev = spares->devs; | |
10202 | if (dev) { | |
10203 | u->new_disks[0] = | |
10204 | makedev(dev->disk.major, | |
10205 | dev->disk.minor); | |
10206 | dl = get_disk_super(super, | |
10207 | dev->disk.major, | |
10208 | dev->disk.minor); | |
10209 | dl->index = u->old_raid_disks; | |
10210 | dev = dev->next; | |
10211 | } | |
10212 | sysfs_free(spares); | |
10213 | } | |
10214 | } | |
10215 | len = disks_to_mpb_size(u->new_raid_disks); | |
10216 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
48c5303a PC |
10217 | break; |
10218 | } | |
f3871fdc | 10219 | case update_size_change: { |
095b8088 N |
10220 | if (update->len < (int)sizeof(struct imsm_update_size_change)) |
10221 | return 0; | |
10222 | break; | |
10223 | } | |
10224 | case update_activate_spare: { | |
10225 | if (update->len < (int)sizeof(struct imsm_update_activate_spare)) | |
10226 | return 0; | |
f3871fdc AK |
10227 | break; |
10228 | } | |
949c47a0 DW |
10229 | case update_create_array: { |
10230 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 10231 | struct intel_dev *dv; |
54c2c1ea | 10232 | struct imsm_dev *dev = &u->dev; |
238c0a71 | 10233 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
54c2c1ea DW |
10234 | struct dl *dl; |
10235 | struct disk_info *inf; | |
10236 | int i; | |
10237 | int activate = 0; | |
949c47a0 | 10238 | |
095b8088 N |
10239 | if (update->len < (int)sizeof(*u)) |
10240 | return 0; | |
10241 | ||
54c2c1ea DW |
10242 | inf = get_disk_info(u); |
10243 | len = sizeof_imsm_dev(dev, 1); | |
ba2de7ba | 10244 | /* allocate a new super->devlist entry */ |
503975b9 N |
10245 | dv = xmalloc(sizeof(*dv)); |
10246 | dv->dev = xmalloc(len); | |
10247 | update->space = dv; | |
949c47a0 | 10248 | |
54c2c1ea DW |
10249 | /* count how many spares will be converted to members */ |
10250 | for (i = 0; i < map->num_members; i++) { | |
10251 | dl = serial_to_dl(inf[i].serial, super); | |
10252 | if (!dl) { | |
10253 | /* hmm maybe it failed?, nothing we can do about | |
10254 | * it here | |
10255 | */ | |
10256 | continue; | |
10257 | } | |
10258 | if (count_memberships(dl, super) == 0) | |
10259 | activate++; | |
10260 | } | |
10261 | len += activate * sizeof(struct imsm_disk); | |
949c47a0 | 10262 | break; |
095b8088 N |
10263 | } |
10264 | case update_kill_array: { | |
10265 | if (update->len < (int)sizeof(struct imsm_update_kill_array)) | |
10266 | return 0; | |
949c47a0 DW |
10267 | break; |
10268 | } | |
095b8088 N |
10269 | case update_rename_array: { |
10270 | if (update->len < (int)sizeof(struct imsm_update_rename_array)) | |
10271 | return 0; | |
10272 | break; | |
10273 | } | |
10274 | case update_add_remove_disk: | |
10275 | /* no update->len needed */ | |
10276 | break; | |
bbab0940 TM |
10277 | case update_prealloc_badblocks_mem: |
10278 | super->extra_space += sizeof(struct bbm_log) - | |
10279 | get_imsm_bbm_log_size(super->bbm_log); | |
10280 | break; | |
e6e9dd3f AP |
10281 | case update_rwh_policy: { |
10282 | if (update->len < (int)sizeof(struct imsm_update_rwh_policy)) | |
10283 | return 0; | |
10284 | break; | |
10285 | } | |
095b8088 N |
10286 | default: |
10287 | return 0; | |
949c47a0 | 10288 | } |
8273f55e | 10289 | |
4d7b1503 DW |
10290 | /* check if we need a larger metadata buffer */ |
10291 | if (super->next_buf) | |
10292 | buf_len = super->next_len; | |
10293 | else | |
10294 | buf_len = super->len; | |
10295 | ||
bbab0940 | 10296 | if (__le32_to_cpu(mpb->mpb_size) + super->extra_space + len > buf_len) { |
4d7b1503 DW |
10297 | /* ok we need a larger buf than what is currently allocated |
10298 | * if this allocation fails process_update will notice that | |
10299 | * ->next_len is set and ->next_buf is NULL | |
10300 | */ | |
bbab0940 TM |
10301 | buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + |
10302 | super->extra_space + len, sector_size); | |
4d7b1503 DW |
10303 | if (super->next_buf) |
10304 | free(super->next_buf); | |
10305 | ||
10306 | super->next_len = buf_len; | |
f36a9ecd | 10307 | if (posix_memalign(&super->next_buf, sector_size, buf_len) == 0) |
1f45a8ad DW |
10308 | memset(super->next_buf, 0, buf_len); |
10309 | else | |
4d7b1503 DW |
10310 | super->next_buf = NULL; |
10311 | } | |
5fe6f031 | 10312 | return 1; |
8273f55e DW |
10313 | } |
10314 | ||
ae6aad82 | 10315 | /* must be called while manager is quiesced */ |
f21e18ca | 10316 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index) |
ae6aad82 DW |
10317 | { |
10318 | struct imsm_super *mpb = super->anchor; | |
ae6aad82 DW |
10319 | struct dl *iter; |
10320 | struct imsm_dev *dev; | |
10321 | struct imsm_map *map; | |
4c9e8c1e | 10322 | unsigned int i, j, num_members; |
fb12a745 | 10323 | __u32 ord, ord_map0; |
4c9e8c1e | 10324 | struct bbm_log *log = super->bbm_log; |
ae6aad82 | 10325 | |
1ade5cc1 | 10326 | dprintf("deleting device[%d] from imsm_super\n", index); |
ae6aad82 DW |
10327 | |
10328 | /* shift all indexes down one */ | |
10329 | for (iter = super->disks; iter; iter = iter->next) | |
f21e18ca | 10330 | if (iter->index > (int)index) |
ae6aad82 | 10331 | iter->index--; |
47ee5a45 | 10332 | for (iter = super->missing; iter; iter = iter->next) |
f21e18ca | 10333 | if (iter->index > (int)index) |
47ee5a45 | 10334 | iter->index--; |
ae6aad82 DW |
10335 | |
10336 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
10337 | dev = get_imsm_dev(super, i); | |
238c0a71 | 10338 | map = get_imsm_map(dev, MAP_0); |
24565c9a DW |
10339 | num_members = map->num_members; |
10340 | for (j = 0; j < num_members; j++) { | |
10341 | /* update ord entries being careful not to propagate | |
10342 | * ord-flags to the first map | |
10343 | */ | |
238c0a71 | 10344 | ord = get_imsm_ord_tbl_ent(dev, j, MAP_X); |
fb12a745 | 10345 | ord_map0 = get_imsm_ord_tbl_ent(dev, j, MAP_0); |
ae6aad82 | 10346 | |
24565c9a DW |
10347 | if (ord_to_idx(ord) <= index) |
10348 | continue; | |
ae6aad82 | 10349 | |
238c0a71 | 10350 | map = get_imsm_map(dev, MAP_0); |
fb12a745 | 10351 | set_imsm_ord_tbl_ent(map, j, ord_map0 - 1); |
238c0a71 | 10352 | map = get_imsm_map(dev, MAP_1); |
24565c9a DW |
10353 | if (map) |
10354 | set_imsm_ord_tbl_ent(map, j, ord - 1); | |
ae6aad82 DW |
10355 | } |
10356 | } | |
10357 | ||
4c9e8c1e TM |
10358 | for (i = 0; i < log->entry_count; i++) { |
10359 | struct bbm_log_entry *entry = &log->marked_block_entries[i]; | |
10360 | ||
10361 | if (entry->disk_ordinal <= index) | |
10362 | continue; | |
10363 | entry->disk_ordinal--; | |
10364 | } | |
10365 | ||
ae6aad82 DW |
10366 | mpb->num_disks--; |
10367 | super->updates_pending++; | |
24565c9a DW |
10368 | if (*dlp) { |
10369 | struct dl *dl = *dlp; | |
10370 | ||
10371 | *dlp = (*dlp)->next; | |
10372 | __free_imsm_disk(dl); | |
10373 | } | |
ae6aad82 | 10374 | } |
9a717282 AK |
10375 | |
10376 | static void close_targets(int *targets, int new_disks) | |
10377 | { | |
10378 | int i; | |
10379 | ||
10380 | if (!targets) | |
10381 | return; | |
10382 | ||
10383 | for (i = 0; i < new_disks; i++) { | |
10384 | if (targets[i] >= 0) { | |
10385 | close(targets[i]); | |
10386 | targets[i] = -1; | |
10387 | } | |
10388 | } | |
10389 | } | |
10390 | ||
10391 | static int imsm_get_allowed_degradation(int level, int raid_disks, | |
10392 | struct intel_super *super, | |
10393 | struct imsm_dev *dev) | |
10394 | { | |
10395 | switch (level) { | |
bf5cf7c7 | 10396 | case 1: |
9a717282 AK |
10397 | case 10:{ |
10398 | int ret_val = 0; | |
10399 | struct imsm_map *map; | |
10400 | int i; | |
10401 | ||
10402 | ret_val = raid_disks/2; | |
10403 | /* check map if all disks pairs not failed | |
10404 | * in both maps | |
10405 | */ | |
238c0a71 | 10406 | map = get_imsm_map(dev, MAP_0); |
9a717282 AK |
10407 | for (i = 0; i < ret_val; i++) { |
10408 | int degradation = 0; | |
10409 | if (get_imsm_disk(super, i) == NULL) | |
10410 | degradation++; | |
10411 | if (get_imsm_disk(super, i + 1) == NULL) | |
10412 | degradation++; | |
10413 | if (degradation == 2) | |
10414 | return 0; | |
10415 | } | |
238c0a71 | 10416 | map = get_imsm_map(dev, MAP_1); |
9a717282 AK |
10417 | /* if there is no second map |
10418 | * result can be returned | |
10419 | */ | |
10420 | if (map == NULL) | |
10421 | return ret_val; | |
10422 | /* check degradation in second map | |
10423 | */ | |
10424 | for (i = 0; i < ret_val; i++) { | |
10425 | int degradation = 0; | |
10426 | if (get_imsm_disk(super, i) == NULL) | |
10427 | degradation++; | |
10428 | if (get_imsm_disk(super, i + 1) == NULL) | |
10429 | degradation++; | |
10430 | if (degradation == 2) | |
10431 | return 0; | |
10432 | } | |
10433 | return ret_val; | |
10434 | } | |
10435 | case 5: | |
10436 | return 1; | |
10437 | case 6: | |
10438 | return 2; | |
10439 | default: | |
10440 | return 0; | |
10441 | } | |
10442 | } | |
10443 | ||
687629c2 AK |
10444 | /******************************************************************************* |
10445 | * Function: open_backup_targets | |
10446 | * Description: Function opens file descriptors for all devices given in | |
10447 | * info->devs | |
10448 | * Parameters: | |
10449 | * info : general array info | |
10450 | * raid_disks : number of disks | |
10451 | * raid_fds : table of device's file descriptors | |
9a717282 AK |
10452 | * super : intel super for raid10 degradation check |
10453 | * dev : intel device for raid10 degradation check | |
687629c2 AK |
10454 | * Returns: |
10455 | * 0 : success | |
10456 | * -1 : fail | |
10457 | ******************************************************************************/ | |
9a717282 AK |
10458 | int open_backup_targets(struct mdinfo *info, int raid_disks, int *raid_fds, |
10459 | struct intel_super *super, struct imsm_dev *dev) | |
687629c2 AK |
10460 | { |
10461 | struct mdinfo *sd; | |
f627f5ad | 10462 | int i; |
9a717282 | 10463 | int opened = 0; |
f627f5ad AK |
10464 | |
10465 | for (i = 0; i < raid_disks; i++) | |
10466 | raid_fds[i] = -1; | |
687629c2 AK |
10467 | |
10468 | for (sd = info->devs ; sd ; sd = sd->next) { | |
10469 | char *dn; | |
10470 | ||
10471 | if (sd->disk.state & (1<<MD_DISK_FAULTY)) { | |
10472 | dprintf("disk is faulty!!\n"); | |
10473 | continue; | |
10474 | } | |
10475 | ||
089f9d79 | 10476 | if (sd->disk.raid_disk >= raid_disks || sd->disk.raid_disk < 0) |
687629c2 AK |
10477 | continue; |
10478 | ||
10479 | dn = map_dev(sd->disk.major, | |
10480 | sd->disk.minor, 1); | |
10481 | raid_fds[sd->disk.raid_disk] = dev_open(dn, O_RDWR); | |
10482 | if (raid_fds[sd->disk.raid_disk] < 0) { | |
e12b3daa | 10483 | pr_err("cannot open component\n"); |
9a717282 | 10484 | continue; |
687629c2 | 10485 | } |
9a717282 AK |
10486 | opened++; |
10487 | } | |
10488 | /* check if maximum array degradation level is not exceeded | |
10489 | */ | |
10490 | if ((raid_disks - opened) > | |
089f9d79 JS |
10491 | imsm_get_allowed_degradation(info->new_level, raid_disks, |
10492 | super, dev)) { | |
e12b3daa | 10493 | pr_err("Not enough disks can be opened.\n"); |
9a717282 AK |
10494 | close_targets(raid_fds, raid_disks); |
10495 | return -2; | |
687629c2 AK |
10496 | } |
10497 | return 0; | |
10498 | } | |
10499 | ||
d31ad643 PB |
10500 | /******************************************************************************* |
10501 | * Function: validate_container_imsm | |
10502 | * Description: This routine validates container after assemble, | |
10503 | * eg. if devices in container are under the same controller. | |
10504 | * | |
10505 | * Parameters: | |
10506 | * info : linked list with info about devices used in array | |
10507 | * Returns: | |
10508 | * 1 : HBA mismatch | |
10509 | * 0 : Success | |
10510 | ******************************************************************************/ | |
10511 | int validate_container_imsm(struct mdinfo *info) | |
10512 | { | |
6b781d33 AP |
10513 | if (check_env("IMSM_NO_PLATFORM")) |
10514 | return 0; | |
d31ad643 | 10515 | |
6b781d33 AP |
10516 | struct sys_dev *idev; |
10517 | struct sys_dev *hba = NULL; | |
10518 | struct sys_dev *intel_devices = find_intel_devices(); | |
10519 | char *dev_path = devt_to_devpath(makedev(info->disk.major, | |
10520 | info->disk.minor)); | |
10521 | ||
10522 | for (idev = intel_devices; idev; idev = idev->next) { | |
10523 | if (dev_path && strstr(dev_path, idev->path)) { | |
10524 | hba = idev; | |
10525 | break; | |
d31ad643 | 10526 | } |
6b781d33 AP |
10527 | } |
10528 | if (dev_path) | |
d31ad643 PB |
10529 | free(dev_path); |
10530 | ||
6b781d33 AP |
10531 | if (!hba) { |
10532 | pr_err("WARNING - Cannot detect HBA for device %s!\n", | |
10533 | devid2kname(makedev(info->disk.major, info->disk.minor))); | |
10534 | return 1; | |
10535 | } | |
10536 | ||
10537 | const struct imsm_orom *orom = get_orom_by_device_id(hba->dev_id); | |
10538 | struct mdinfo *dev; | |
10539 | ||
10540 | for (dev = info->next; dev; dev = dev->next) { | |
10541 | dev_path = devt_to_devpath(makedev(dev->disk.major, dev->disk.minor)); | |
10542 | ||
10543 | struct sys_dev *hba2 = NULL; | |
10544 | for (idev = intel_devices; idev; idev = idev->next) { | |
10545 | if (dev_path && strstr(dev_path, idev->path)) { | |
10546 | hba2 = idev; | |
10547 | break; | |
d31ad643 PB |
10548 | } |
10549 | } | |
6b781d33 AP |
10550 | if (dev_path) |
10551 | free(dev_path); | |
10552 | ||
10553 | const struct imsm_orom *orom2 = hba2 == NULL ? NULL : | |
10554 | get_orom_by_device_id(hba2->dev_id); | |
10555 | ||
10556 | if (hba2 && hba->type != hba2->type) { | |
10557 | pr_err("WARNING - HBAs of devices do not match %s != %s\n", | |
10558 | get_sys_dev_type(hba->type), get_sys_dev_type(hba2->type)); | |
10559 | return 1; | |
10560 | } | |
10561 | ||
07cb1e57 | 10562 | if (orom != orom2) { |
6b781d33 AP |
10563 | pr_err("WARNING - IMSM container assembled with disks under different HBAs!\n" |
10564 | " This operation is not supported and can lead to data loss.\n"); | |
10565 | return 1; | |
10566 | } | |
10567 | ||
10568 | if (!orom) { | |
10569 | pr_err("WARNING - IMSM container assembled with disks under HBAs without IMSM platform support!\n" | |
10570 | " This operation is not supported and can lead to data loss.\n"); | |
10571 | return 1; | |
10572 | } | |
d31ad643 | 10573 | } |
6b781d33 | 10574 | |
d31ad643 PB |
10575 | return 0; |
10576 | } | |
32141c17 | 10577 | |
6f50473f TM |
10578 | /******************************************************************************* |
10579 | * Function: imsm_record_badblock | |
10580 | * Description: This routine stores new bad block record in BBM log | |
10581 | * | |
10582 | * Parameters: | |
10583 | * a : array containing a bad block | |
10584 | * slot : disk number containing a bad block | |
10585 | * sector : bad block sector | |
10586 | * length : bad block sectors range | |
10587 | * Returns: | |
10588 | * 1 : Success | |
10589 | * 0 : Error | |
10590 | ******************************************************************************/ | |
10591 | static int imsm_record_badblock(struct active_array *a, int slot, | |
10592 | unsigned long long sector, int length) | |
10593 | { | |
10594 | struct intel_super *super = a->container->sb; | |
10595 | int ord; | |
10596 | int ret; | |
10597 | ||
10598 | ord = imsm_disk_slot_to_ord(a, slot); | |
10599 | if (ord < 0) | |
10600 | return 0; | |
10601 | ||
10602 | ret = record_new_badblock(super->bbm_log, ord_to_idx(ord), sector, | |
10603 | length); | |
10604 | if (ret) | |
10605 | super->updates_pending++; | |
10606 | ||
10607 | return ret; | |
10608 | } | |
c07a5a4f TM |
10609 | /******************************************************************************* |
10610 | * Function: imsm_clear_badblock | |
10611 | * Description: This routine clears bad block record from BBM log | |
10612 | * | |
10613 | * Parameters: | |
10614 | * a : array containing a bad block | |
10615 | * slot : disk number containing a bad block | |
10616 | * sector : bad block sector | |
10617 | * length : bad block sectors range | |
10618 | * Returns: | |
10619 | * 1 : Success | |
10620 | * 0 : Error | |
10621 | ******************************************************************************/ | |
10622 | static int imsm_clear_badblock(struct active_array *a, int slot, | |
10623 | unsigned long long sector, int length) | |
10624 | { | |
10625 | struct intel_super *super = a->container->sb; | |
10626 | int ord; | |
10627 | int ret; | |
10628 | ||
10629 | ord = imsm_disk_slot_to_ord(a, slot); | |
10630 | if (ord < 0) | |
10631 | return 0; | |
10632 | ||
10633 | ret = clear_badblock(super->bbm_log, ord_to_idx(ord), sector, length); | |
10634 | if (ret) | |
10635 | super->updates_pending++; | |
10636 | ||
10637 | return ret; | |
10638 | } | |
928f1424 TM |
10639 | /******************************************************************************* |
10640 | * Function: imsm_get_badblocks | |
10641 | * Description: This routine get list of bad blocks for an array | |
10642 | * | |
10643 | * Parameters: | |
10644 | * a : array | |
10645 | * slot : disk number | |
10646 | * Returns: | |
10647 | * bb : structure containing bad blocks | |
10648 | * NULL : error | |
10649 | ******************************************************************************/ | |
10650 | static struct md_bb *imsm_get_badblocks(struct active_array *a, int slot) | |
10651 | { | |
10652 | int inst = a->info.container_member; | |
10653 | struct intel_super *super = a->container->sb; | |
10654 | struct imsm_dev *dev = get_imsm_dev(super, inst); | |
10655 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
10656 | int ord; | |
10657 | ||
10658 | ord = imsm_disk_slot_to_ord(a, slot); | |
10659 | if (ord < 0) | |
10660 | return NULL; | |
10661 | ||
10662 | get_volume_badblocks(super->bbm_log, ord_to_idx(ord), pba_of_lba0(map), | |
44490938 | 10663 | per_dev_array_size(map), &super->bb); |
928f1424 TM |
10664 | |
10665 | return &super->bb; | |
10666 | } | |
27156a57 TM |
10667 | /******************************************************************************* |
10668 | * Function: examine_badblocks_imsm | |
10669 | * Description: Prints list of bad blocks on a disk to the standard output | |
10670 | * | |
10671 | * Parameters: | |
10672 | * st : metadata handler | |
10673 | * fd : open file descriptor for device | |
10674 | * devname : device name | |
10675 | * Returns: | |
10676 | * 0 : Success | |
10677 | * 1 : Error | |
10678 | ******************************************************************************/ | |
10679 | static int examine_badblocks_imsm(struct supertype *st, int fd, char *devname) | |
10680 | { | |
10681 | struct intel_super *super = st->sb; | |
10682 | struct bbm_log *log = super->bbm_log; | |
10683 | struct dl *d = NULL; | |
10684 | int any = 0; | |
10685 | ||
10686 | for (d = super->disks; d ; d = d->next) { | |
10687 | if (strcmp(d->devname, devname) == 0) | |
10688 | break; | |
10689 | } | |
10690 | ||
10691 | if ((d == NULL) || (d->index < 0)) { /* serial mismatch probably */ | |
10692 | pr_err("%s doesn't appear to be part of a raid array\n", | |
10693 | devname); | |
10694 | return 1; | |
10695 | } | |
10696 | ||
10697 | if (log != NULL) { | |
10698 | unsigned int i; | |
10699 | struct bbm_log_entry *entry = &log->marked_block_entries[0]; | |
10700 | ||
10701 | for (i = 0; i < log->entry_count; i++) { | |
10702 | if (entry[i].disk_ordinal == d->index) { | |
10703 | unsigned long long sector = __le48_to_cpu( | |
10704 | &entry[i].defective_block_start); | |
10705 | int cnt = entry[i].marked_count + 1; | |
10706 | ||
10707 | if (!any) { | |
10708 | printf("Bad-blocks on %s:\n", devname); | |
10709 | any = 1; | |
10710 | } | |
10711 | ||
10712 | printf("%20llu for %d sectors\n", sector, cnt); | |
10713 | } | |
10714 | } | |
10715 | } | |
10716 | ||
10717 | if (!any) | |
10718 | printf("No bad-blocks list configured on %s\n", devname); | |
10719 | ||
10720 | return 0; | |
10721 | } | |
687629c2 AK |
10722 | /******************************************************************************* |
10723 | * Function: init_migr_record_imsm | |
10724 | * Description: Function inits imsm migration record | |
10725 | * Parameters: | |
10726 | * super : imsm internal array info | |
10727 | * dev : device under migration | |
10728 | * info : general array info to find the smallest device | |
10729 | * Returns: | |
10730 | * none | |
10731 | ******************************************************************************/ | |
10732 | void init_migr_record_imsm(struct supertype *st, struct imsm_dev *dev, | |
10733 | struct mdinfo *info) | |
10734 | { | |
10735 | struct intel_super *super = st->sb; | |
10736 | struct migr_record *migr_rec = super->migr_rec; | |
10737 | int new_data_disks; | |
10738 | unsigned long long dsize, dev_sectors; | |
10739 | long long unsigned min_dev_sectors = -1LLU; | |
10740 | struct mdinfo *sd; | |
10741 | char nm[30]; | |
10742 | int fd; | |
238c0a71 AK |
10743 | struct imsm_map *map_dest = get_imsm_map(dev, MAP_0); |
10744 | struct imsm_map *map_src = get_imsm_map(dev, MAP_1); | |
687629c2 | 10745 | unsigned long long num_migr_units; |
3ef4403c | 10746 | unsigned long long array_blocks; |
687629c2 AK |
10747 | |
10748 | memset(migr_rec, 0, sizeof(struct migr_record)); | |
10749 | migr_rec->family_num = __cpu_to_le32(super->anchor->family_num); | |
10750 | ||
10751 | /* only ascending reshape supported now */ | |
10752 | migr_rec->ascending_migr = __cpu_to_le32(1); | |
10753 | ||
10754 | migr_rec->dest_depth_per_unit = GEN_MIGR_AREA_SIZE / | |
10755 | max(map_dest->blocks_per_strip, map_src->blocks_per_strip); | |
e1742195 AK |
10756 | migr_rec->dest_depth_per_unit *= |
10757 | max(map_dest->blocks_per_strip, map_src->blocks_per_strip); | |
9529d343 | 10758 | new_data_disks = imsm_num_data_members(map_dest); |
687629c2 AK |
10759 | migr_rec->blocks_per_unit = |
10760 | __cpu_to_le32(migr_rec->dest_depth_per_unit * new_data_disks); | |
10761 | migr_rec->dest_depth_per_unit = | |
10762 | __cpu_to_le32(migr_rec->dest_depth_per_unit); | |
3ef4403c | 10763 | array_blocks = info->component_size * new_data_disks; |
687629c2 AK |
10764 | num_migr_units = |
10765 | array_blocks / __le32_to_cpu(migr_rec->blocks_per_unit); | |
10766 | ||
10767 | if (array_blocks % __le32_to_cpu(migr_rec->blocks_per_unit)) | |
10768 | num_migr_units++; | |
9f421827 | 10769 | set_num_migr_units(migr_rec, num_migr_units); |
687629c2 AK |
10770 | |
10771 | migr_rec->post_migr_vol_cap = dev->size_low; | |
10772 | migr_rec->post_migr_vol_cap_hi = dev->size_high; | |
10773 | ||
687629c2 AK |
10774 | /* Find the smallest dev */ |
10775 | for (sd = info->devs ; sd ; sd = sd->next) { | |
10776 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); | |
10777 | fd = dev_open(nm, O_RDONLY); | |
10778 | if (fd < 0) | |
10779 | continue; | |
10780 | get_dev_size(fd, NULL, &dsize); | |
10781 | dev_sectors = dsize / 512; | |
10782 | if (dev_sectors < min_dev_sectors) | |
10783 | min_dev_sectors = dev_sectors; | |
10784 | close(fd); | |
10785 | } | |
9f421827 | 10786 | set_migr_chkp_area_pba(migr_rec, min_dev_sectors - |
687629c2 AK |
10787 | RAID_DISK_RESERVED_BLOCKS_IMSM_HI); |
10788 | ||
10789 | write_imsm_migr_rec(st); | |
10790 | ||
10791 | return; | |
10792 | } | |
10793 | ||
10794 | /******************************************************************************* | |
10795 | * Function: save_backup_imsm | |
10796 | * Description: Function saves critical data stripes to Migration Copy Area | |
10797 | * and updates the current migration unit status. | |
10798 | * Use restore_stripes() to form a destination stripe, | |
10799 | * and to write it to the Copy Area. | |
10800 | * Parameters: | |
10801 | * st : supertype information | |
aea93171 | 10802 | * dev : imsm device that backup is saved for |
687629c2 AK |
10803 | * info : general array info |
10804 | * buf : input buffer | |
687629c2 AK |
10805 | * length : length of data to backup (blocks_per_unit) |
10806 | * Returns: | |
10807 | * 0 : success | |
10808 | *, -1 : fail | |
10809 | ******************************************************************************/ | |
10810 | int save_backup_imsm(struct supertype *st, | |
10811 | struct imsm_dev *dev, | |
10812 | struct mdinfo *info, | |
10813 | void *buf, | |
687629c2 AK |
10814 | int length) |
10815 | { | |
10816 | int rv = -1; | |
10817 | struct intel_super *super = st->sb; | |
594dc1b8 JS |
10818 | unsigned long long *target_offsets; |
10819 | int *targets; | |
687629c2 | 10820 | int i; |
238c0a71 | 10821 | struct imsm_map *map_dest = get_imsm_map(dev, MAP_0); |
687629c2 | 10822 | int new_disks = map_dest->num_members; |
ab724b98 AK |
10823 | int dest_layout = 0; |
10824 | int dest_chunk; | |
d1877f69 | 10825 | unsigned long long start; |
9529d343 | 10826 | int data_disks = imsm_num_data_members(map_dest); |
687629c2 | 10827 | |
503975b9 | 10828 | targets = xmalloc(new_disks * sizeof(int)); |
687629c2 | 10829 | |
7e45b550 AK |
10830 | for (i = 0; i < new_disks; i++) |
10831 | targets[i] = -1; | |
10832 | ||
503975b9 | 10833 | target_offsets = xcalloc(new_disks, sizeof(unsigned long long)); |
687629c2 | 10834 | |
d1877f69 | 10835 | start = info->reshape_progress * 512; |
687629c2 | 10836 | for (i = 0; i < new_disks; i++) { |
9f421827 | 10837 | target_offsets[i] = migr_chkp_area_pba(super->migr_rec) * 512; |
d1877f69 AK |
10838 | /* move back copy area adderss, it will be moved forward |
10839 | * in restore_stripes() using start input variable | |
10840 | */ | |
10841 | target_offsets[i] -= start/data_disks; | |
687629c2 AK |
10842 | } |
10843 | ||
9a717282 AK |
10844 | if (open_backup_targets(info, new_disks, targets, |
10845 | super, dev)) | |
687629c2 AK |
10846 | goto abort; |
10847 | ||
68eb8bc6 | 10848 | dest_layout = imsm_level_to_layout(map_dest->raid_level); |
ab724b98 AK |
10849 | dest_chunk = __le16_to_cpu(map_dest->blocks_per_strip) * 512; |
10850 | ||
687629c2 AK |
10851 | if (restore_stripes(targets, /* list of dest devices */ |
10852 | target_offsets, /* migration record offsets */ | |
10853 | new_disks, | |
ab724b98 AK |
10854 | dest_chunk, |
10855 | map_dest->raid_level, | |
10856 | dest_layout, | |
10857 | -1, /* source backup file descriptor */ | |
10858 | 0, /* input buf offset | |
10859 | * always 0 buf is already offseted */ | |
d1877f69 | 10860 | start, |
687629c2 AK |
10861 | length, |
10862 | buf) != 0) { | |
e7b84f9d | 10863 | pr_err("Error restoring stripes\n"); |
687629c2 AK |
10864 | goto abort; |
10865 | } | |
10866 | ||
10867 | rv = 0; | |
10868 | ||
10869 | abort: | |
10870 | if (targets) { | |
9a717282 | 10871 | close_targets(targets, new_disks); |
687629c2 AK |
10872 | free(targets); |
10873 | } | |
10874 | free(target_offsets); | |
10875 | ||
10876 | return rv; | |
10877 | } | |
10878 | ||
10879 | /******************************************************************************* | |
10880 | * Function: save_checkpoint_imsm | |
10881 | * Description: Function called for current unit status update | |
10882 | * in the migration record. It writes it to disk. | |
10883 | * Parameters: | |
10884 | * super : imsm internal array info | |
10885 | * info : general array info | |
10886 | * Returns: | |
10887 | * 0: success | |
10888 | * 1: failure | |
0228d92c AK |
10889 | * 2: failure, means no valid migration record |
10890 | * / no general migration in progress / | |
687629c2 AK |
10891 | ******************************************************************************/ |
10892 | int save_checkpoint_imsm(struct supertype *st, struct mdinfo *info, int state) | |
10893 | { | |
10894 | struct intel_super *super = st->sb; | |
f8b72ef5 AK |
10895 | unsigned long long blocks_per_unit; |
10896 | unsigned long long curr_migr_unit; | |
10897 | ||
2e062e82 | 10898 | if (load_imsm_migr_rec(super, info) != 0) { |
7a862a02 | 10899 | dprintf("imsm: ERROR: Cannot read migration record for checkpoint save.\n"); |
2e062e82 AK |
10900 | return 1; |
10901 | } | |
10902 | ||
f8b72ef5 AK |
10903 | blocks_per_unit = __le32_to_cpu(super->migr_rec->blocks_per_unit); |
10904 | if (blocks_per_unit == 0) { | |
0228d92c AK |
10905 | dprintf("imsm: no migration in progress.\n"); |
10906 | return 2; | |
687629c2 | 10907 | } |
f8b72ef5 AK |
10908 | curr_migr_unit = info->reshape_progress / blocks_per_unit; |
10909 | /* check if array is alligned to copy area | |
10910 | * if it is not alligned, add one to current migration unit value | |
10911 | * this can happend on array reshape finish only | |
10912 | */ | |
10913 | if (info->reshape_progress % blocks_per_unit) | |
10914 | curr_migr_unit++; | |
687629c2 | 10915 | |
9f421827 | 10916 | set_current_migr_unit(super->migr_rec, curr_migr_unit); |
687629c2 | 10917 | super->migr_rec->rec_status = __cpu_to_le32(state); |
9f421827 PB |
10918 | set_migr_dest_1st_member_lba(super->migr_rec, |
10919 | super->migr_rec->dest_depth_per_unit * curr_migr_unit); | |
10920 | ||
687629c2 | 10921 | if (write_imsm_migr_rec(st) < 0) { |
7a862a02 | 10922 | dprintf("imsm: Cannot write migration record outside backup area\n"); |
687629c2 AK |
10923 | return 1; |
10924 | } | |
10925 | ||
10926 | return 0; | |
10927 | } | |
10928 | ||
276d77db AK |
10929 | /******************************************************************************* |
10930 | * Function: recover_backup_imsm | |
10931 | * Description: Function recovers critical data from the Migration Copy Area | |
10932 | * while assembling an array. | |
10933 | * Parameters: | |
10934 | * super : imsm internal array info | |
10935 | * info : general array info | |
10936 | * Returns: | |
10937 | * 0 : success (or there is no data to recover) | |
10938 | * 1 : fail | |
10939 | ******************************************************************************/ | |
10940 | int recover_backup_imsm(struct supertype *st, struct mdinfo *info) | |
10941 | { | |
10942 | struct intel_super *super = st->sb; | |
10943 | struct migr_record *migr_rec = super->migr_rec; | |
594dc1b8 | 10944 | struct imsm_map *map_dest; |
276d77db AK |
10945 | struct intel_dev *id = NULL; |
10946 | unsigned long long read_offset; | |
10947 | unsigned long long write_offset; | |
10948 | unsigned unit_len; | |
10949 | int *targets = NULL; | |
10950 | int new_disks, i, err; | |
10951 | char *buf = NULL; | |
10952 | int retval = 1; | |
f36a9ecd | 10953 | unsigned int sector_size = super->sector_size; |
9f421827 PB |
10954 | unsigned long curr_migr_unit = current_migr_unit(migr_rec); |
10955 | unsigned long num_migr_units = get_num_migr_units(migr_rec); | |
276d77db | 10956 | char buffer[20]; |
6c3560c0 | 10957 | int skipped_disks = 0; |
276d77db AK |
10958 | |
10959 | err = sysfs_get_str(info, NULL, "array_state", (char *)buffer, 20); | |
10960 | if (err < 1) | |
10961 | return 1; | |
10962 | ||
10963 | /* recover data only during assemblation */ | |
10964 | if (strncmp(buffer, "inactive", 8) != 0) | |
10965 | return 0; | |
10966 | /* no data to recover */ | |
10967 | if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL) | |
10968 | return 0; | |
10969 | if (curr_migr_unit >= num_migr_units) | |
10970 | return 1; | |
10971 | ||
10972 | /* find device during reshape */ | |
10973 | for (id = super->devlist; id; id = id->next) | |
10974 | if (is_gen_migration(id->dev)) | |
10975 | break; | |
10976 | if (id == NULL) | |
10977 | return 1; | |
10978 | ||
238c0a71 | 10979 | map_dest = get_imsm_map(id->dev, MAP_0); |
276d77db AK |
10980 | new_disks = map_dest->num_members; |
10981 | ||
9f421827 | 10982 | read_offset = migr_chkp_area_pba(migr_rec) * 512; |
276d77db | 10983 | |
9f421827 | 10984 | write_offset = (migr_dest_1st_member_lba(migr_rec) + |
5551b113 | 10985 | pba_of_lba0(map_dest)) * 512; |
276d77db AK |
10986 | |
10987 | unit_len = __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
f36a9ecd | 10988 | if (posix_memalign((void **)&buf, sector_size, unit_len) != 0) |
276d77db | 10989 | goto abort; |
503975b9 | 10990 | targets = xcalloc(new_disks, sizeof(int)); |
276d77db | 10991 | |
9a717282 | 10992 | if (open_backup_targets(info, new_disks, targets, super, id->dev)) { |
e7b84f9d | 10993 | pr_err("Cannot open some devices belonging to array.\n"); |
f627f5ad AK |
10994 | goto abort; |
10995 | } | |
276d77db AK |
10996 | |
10997 | for (i = 0; i < new_disks; i++) { | |
6c3560c0 AK |
10998 | if (targets[i] < 0) { |
10999 | skipped_disks++; | |
11000 | continue; | |
11001 | } | |
276d77db | 11002 | if (lseek64(targets[i], read_offset, SEEK_SET) < 0) { |
e7b84f9d N |
11003 | pr_err("Cannot seek to block: %s\n", |
11004 | strerror(errno)); | |
137debce AK |
11005 | skipped_disks++; |
11006 | continue; | |
276d77db | 11007 | } |
9ec11d1a | 11008 | if ((unsigned)read(targets[i], buf, unit_len) != unit_len) { |
e7b84f9d N |
11009 | pr_err("Cannot read copy area block: %s\n", |
11010 | strerror(errno)); | |
137debce AK |
11011 | skipped_disks++; |
11012 | continue; | |
276d77db AK |
11013 | } |
11014 | if (lseek64(targets[i], write_offset, SEEK_SET) < 0) { | |
e7b84f9d N |
11015 | pr_err("Cannot seek to block: %s\n", |
11016 | strerror(errno)); | |
137debce AK |
11017 | skipped_disks++; |
11018 | continue; | |
276d77db | 11019 | } |
9ec11d1a | 11020 | if ((unsigned)write(targets[i], buf, unit_len) != unit_len) { |
e7b84f9d N |
11021 | pr_err("Cannot restore block: %s\n", |
11022 | strerror(errno)); | |
137debce AK |
11023 | skipped_disks++; |
11024 | continue; | |
276d77db AK |
11025 | } |
11026 | } | |
11027 | ||
137debce AK |
11028 | if (skipped_disks > imsm_get_allowed_degradation(info->new_level, |
11029 | new_disks, | |
11030 | super, | |
11031 | id->dev)) { | |
7a862a02 | 11032 | pr_err("Cannot restore data from backup. Too many failed disks\n"); |
6c3560c0 AK |
11033 | goto abort; |
11034 | } | |
11035 | ||
befb629b AK |
11036 | if (save_checkpoint_imsm(st, info, UNIT_SRC_NORMAL)) { |
11037 | /* ignore error == 2, this can mean end of reshape here | |
11038 | */ | |
7a862a02 | 11039 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL) during restart\n"); |
befb629b | 11040 | } else |
276d77db | 11041 | retval = 0; |
276d77db AK |
11042 | |
11043 | abort: | |
11044 | if (targets) { | |
11045 | for (i = 0; i < new_disks; i++) | |
11046 | if (targets[i]) | |
11047 | close(targets[i]); | |
11048 | free(targets); | |
11049 | } | |
11050 | free(buf); | |
11051 | return retval; | |
11052 | } | |
11053 | ||
2cda7640 ML |
11054 | static char disk_by_path[] = "/dev/disk/by-path/"; |
11055 | ||
11056 | static const char *imsm_get_disk_controller_domain(const char *path) | |
11057 | { | |
2cda7640 | 11058 | char disk_path[PATH_MAX]; |
96234762 LM |
11059 | char *drv=NULL; |
11060 | struct stat st; | |
2cda7640 | 11061 | |
6d8d290a | 11062 | strcpy(disk_path, disk_by_path); |
96234762 LM |
11063 | strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1); |
11064 | if (stat(disk_path, &st) == 0) { | |
11065 | struct sys_dev* hba; | |
594dc1b8 | 11066 | char *path; |
96234762 LM |
11067 | |
11068 | path = devt_to_devpath(st.st_rdev); | |
11069 | if (path == NULL) | |
11070 | return "unknown"; | |
11071 | hba = find_disk_attached_hba(-1, path); | |
11072 | if (hba && hba->type == SYS_DEV_SAS) | |
11073 | drv = "isci"; | |
11074 | else if (hba && hba->type == SYS_DEV_SATA) | |
11075 | drv = "ahci"; | |
c6839718 MT |
11076 | else if (hba && hba->type == SYS_DEV_VMD) |
11077 | drv = "vmd"; | |
11078 | else if (hba && hba->type == SYS_DEV_NVME) | |
11079 | drv = "nvme"; | |
1011e834 | 11080 | else |
96234762 LM |
11081 | drv = "unknown"; |
11082 | dprintf("path: %s hba: %s attached: %s\n", | |
11083 | path, (hba) ? hba->path : "NULL", drv); | |
11084 | free(path); | |
2cda7640 | 11085 | } |
96234762 | 11086 | return drv; |
2cda7640 ML |
11087 | } |
11088 | ||
4dd2df09 | 11089 | static char *imsm_find_array_devnm_by_subdev(int subdev, char *container) |
78b10e66 | 11090 | { |
4dd2df09 | 11091 | static char devnm[32]; |
78b10e66 N |
11092 | char subdev_name[20]; |
11093 | struct mdstat_ent *mdstat; | |
11094 | ||
11095 | sprintf(subdev_name, "%d", subdev); | |
11096 | mdstat = mdstat_by_subdev(subdev_name, container); | |
11097 | if (!mdstat) | |
4dd2df09 | 11098 | return NULL; |
78b10e66 | 11099 | |
4dd2df09 | 11100 | strcpy(devnm, mdstat->devnm); |
78b10e66 | 11101 | free_mdstat(mdstat); |
4dd2df09 | 11102 | return devnm; |
78b10e66 N |
11103 | } |
11104 | ||
11105 | static int imsm_reshape_is_allowed_on_container(struct supertype *st, | |
11106 | struct geo_params *geo, | |
fbf3d202 AK |
11107 | int *old_raid_disks, |
11108 | int direction) | |
78b10e66 | 11109 | { |
694575e7 KW |
11110 | /* currently we only support increasing the number of devices |
11111 | * for a container. This increases the number of device for each | |
11112 | * member array. They must all be RAID0 or RAID5. | |
11113 | */ | |
78b10e66 N |
11114 | int ret_val = 0; |
11115 | struct mdinfo *info, *member; | |
11116 | int devices_that_can_grow = 0; | |
11117 | ||
7a862a02 | 11118 | dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): st->devnm = (%s)\n", st->devnm); |
78b10e66 | 11119 | |
d04f65f4 | 11120 | if (geo->size > 0 || |
78b10e66 N |
11121 | geo->level != UnSet || |
11122 | geo->layout != UnSet || | |
11123 | geo->chunksize != 0 || | |
11124 | geo->raid_disks == UnSet) { | |
7a862a02 | 11125 | dprintf("imsm: Container operation is allowed for raid disks number change only.\n"); |
78b10e66 N |
11126 | return ret_val; |
11127 | } | |
11128 | ||
fbf3d202 | 11129 | if (direction == ROLLBACK_METADATA_CHANGES) { |
7a862a02 | 11130 | dprintf("imsm: Metadata changes rollback is not supported for container operation.\n"); |
fbf3d202 AK |
11131 | return ret_val; |
11132 | } | |
11133 | ||
78b10e66 N |
11134 | info = container_content_imsm(st, NULL); |
11135 | for (member = info; member; member = member->next) { | |
4dd2df09 | 11136 | char *result; |
78b10e66 N |
11137 | |
11138 | dprintf("imsm: checking device_num: %i\n", | |
11139 | member->container_member); | |
11140 | ||
d7d205bd | 11141 | if (geo->raid_disks <= member->array.raid_disks) { |
78b10e66 N |
11142 | /* we work on container for Online Capacity Expansion |
11143 | * only so raid_disks has to grow | |
11144 | */ | |
7a862a02 | 11145 | dprintf("imsm: for container operation raid disks increase is required\n"); |
78b10e66 N |
11146 | break; |
11147 | } | |
11148 | ||
089f9d79 | 11149 | if (info->array.level != 0 && info->array.level != 5) { |
78b10e66 N |
11150 | /* we cannot use this container with other raid level |
11151 | */ | |
7a862a02 | 11152 | dprintf("imsm: for container operation wrong raid level (%i) detected\n", |
78b10e66 N |
11153 | info->array.level); |
11154 | break; | |
11155 | } else { | |
11156 | /* check for platform support | |
11157 | * for this raid level configuration | |
11158 | */ | |
11159 | struct intel_super *super = st->sb; | |
11160 | if (!is_raid_level_supported(super->orom, | |
11161 | member->array.level, | |
11162 | geo->raid_disks)) { | |
7a862a02 | 11163 | dprintf("platform does not support raid%d with %d disk%s\n", |
78b10e66 N |
11164 | info->array.level, |
11165 | geo->raid_disks, | |
11166 | geo->raid_disks > 1 ? "s" : ""); | |
11167 | break; | |
11168 | } | |
2a4a08e7 AK |
11169 | /* check if component size is aligned to chunk size |
11170 | */ | |
11171 | if (info->component_size % | |
11172 | (info->array.chunk_size/512)) { | |
7a862a02 | 11173 | dprintf("Component size is not aligned to chunk size\n"); |
2a4a08e7 AK |
11174 | break; |
11175 | } | |
78b10e66 N |
11176 | } |
11177 | ||
11178 | if (*old_raid_disks && | |
11179 | info->array.raid_disks != *old_raid_disks) | |
11180 | break; | |
11181 | *old_raid_disks = info->array.raid_disks; | |
11182 | ||
11183 | /* All raid5 and raid0 volumes in container | |
11184 | * have to be ready for Online Capacity Expansion | |
11185 | * so they need to be assembled. We have already | |
11186 | * checked that no recovery etc is happening. | |
11187 | */ | |
4dd2df09 N |
11188 | result = imsm_find_array_devnm_by_subdev(member->container_member, |
11189 | st->container_devnm); | |
11190 | if (result == NULL) { | |
78b10e66 N |
11191 | dprintf("imsm: cannot find array\n"); |
11192 | break; | |
11193 | } | |
11194 | devices_that_can_grow++; | |
11195 | } | |
11196 | sysfs_free(info); | |
11197 | if (!member && devices_that_can_grow) | |
11198 | ret_val = 1; | |
11199 | ||
11200 | if (ret_val) | |
1ade5cc1 | 11201 | dprintf("Container operation allowed\n"); |
78b10e66 | 11202 | else |
1ade5cc1 | 11203 | dprintf("Error: %i\n", ret_val); |
78b10e66 N |
11204 | |
11205 | return ret_val; | |
11206 | } | |
11207 | ||
11208 | /* Function: get_spares_for_grow | |
11209 | * Description: Allocates memory and creates list of spare devices | |
1011e834 | 11210 | * avaliable in container. Checks if spare drive size is acceptable. |
78b10e66 N |
11211 | * Parameters: Pointer to the supertype structure |
11212 | * Returns: Pointer to the list of spare devices (mdinfo structure) on success, | |
1011e834 | 11213 | * NULL if fail |
78b10e66 N |
11214 | */ |
11215 | static struct mdinfo *get_spares_for_grow(struct supertype *st) | |
11216 | { | |
fbfdcb06 AO |
11217 | struct spare_criteria sc; |
11218 | ||
11219 | get_spare_criteria_imsm(st, &sc); | |
11220 | return container_choose_spares(st, &sc, NULL, NULL, NULL, 0); | |
78b10e66 N |
11221 | } |
11222 | ||
11223 | /****************************************************************************** | |
11224 | * function: imsm_create_metadata_update_for_reshape | |
11225 | * Function creates update for whole IMSM container. | |
11226 | * | |
11227 | ******************************************************************************/ | |
11228 | static int imsm_create_metadata_update_for_reshape( | |
11229 | struct supertype *st, | |
11230 | struct geo_params *geo, | |
11231 | int old_raid_disks, | |
11232 | struct imsm_update_reshape **updatep) | |
11233 | { | |
11234 | struct intel_super *super = st->sb; | |
11235 | struct imsm_super *mpb = super->anchor; | |
594dc1b8 JS |
11236 | int update_memory_size; |
11237 | struct imsm_update_reshape *u; | |
11238 | struct mdinfo *spares; | |
78b10e66 | 11239 | int i; |
594dc1b8 | 11240 | int delta_disks; |
bbd24d86 | 11241 | struct mdinfo *dev; |
78b10e66 | 11242 | |
1ade5cc1 | 11243 | dprintf("(enter) raid_disks = %i\n", geo->raid_disks); |
78b10e66 N |
11244 | |
11245 | delta_disks = geo->raid_disks - old_raid_disks; | |
11246 | ||
11247 | /* size of all update data without anchor */ | |
11248 | update_memory_size = sizeof(struct imsm_update_reshape); | |
11249 | ||
11250 | /* now add space for spare disks that we need to add. */ | |
11251 | update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1); | |
11252 | ||
503975b9 | 11253 | u = xcalloc(1, update_memory_size); |
78b10e66 N |
11254 | u->type = update_reshape_container_disks; |
11255 | u->old_raid_disks = old_raid_disks; | |
11256 | u->new_raid_disks = geo->raid_disks; | |
11257 | ||
11258 | /* now get spare disks list | |
11259 | */ | |
11260 | spares = get_spares_for_grow(st); | |
11261 | ||
d7be7d87 | 11262 | if (spares == NULL || delta_disks > spares->array.spare_disks) { |
7a862a02 | 11263 | pr_err("imsm: ERROR: Cannot get spare devices for %s.\n", geo->dev_name); |
e4c72d1d | 11264 | i = -1; |
78b10e66 N |
11265 | goto abort; |
11266 | } | |
11267 | ||
11268 | /* we have got spares | |
11269 | * update disk list in imsm_disk list table in anchor | |
11270 | */ | |
11271 | dprintf("imsm: %i spares are available.\n\n", | |
11272 | spares->array.spare_disks); | |
11273 | ||
bbd24d86 | 11274 | dev = spares->devs; |
78b10e66 | 11275 | for (i = 0; i < delta_disks; i++) { |
78b10e66 N |
11276 | struct dl *dl; |
11277 | ||
bbd24d86 AK |
11278 | if (dev == NULL) |
11279 | break; | |
78b10e66 N |
11280 | u->new_disks[i] = makedev(dev->disk.major, |
11281 | dev->disk.minor); | |
11282 | dl = get_disk_super(super, dev->disk.major, dev->disk.minor); | |
ee4beede AK |
11283 | dl->index = mpb->num_disks; |
11284 | mpb->num_disks++; | |
bbd24d86 | 11285 | dev = dev->next; |
78b10e66 | 11286 | } |
78b10e66 N |
11287 | |
11288 | abort: | |
11289 | /* free spares | |
11290 | */ | |
11291 | sysfs_free(spares); | |
11292 | ||
d677e0b8 | 11293 | dprintf("imsm: reshape update preparation :"); |
78b10e66 | 11294 | if (i == delta_disks) { |
1ade5cc1 | 11295 | dprintf_cont(" OK\n"); |
78b10e66 N |
11296 | *updatep = u; |
11297 | return update_memory_size; | |
11298 | } | |
11299 | free(u); | |
1ade5cc1 | 11300 | dprintf_cont(" Error\n"); |
78b10e66 N |
11301 | |
11302 | return 0; | |
11303 | } | |
11304 | ||
f3871fdc AK |
11305 | /****************************************************************************** |
11306 | * function: imsm_create_metadata_update_for_size_change() | |
11307 | * Creates update for IMSM array for array size change. | |
11308 | * | |
11309 | ******************************************************************************/ | |
11310 | static int imsm_create_metadata_update_for_size_change( | |
11311 | struct supertype *st, | |
11312 | struct geo_params *geo, | |
11313 | struct imsm_update_size_change **updatep) | |
11314 | { | |
11315 | struct intel_super *super = st->sb; | |
594dc1b8 JS |
11316 | int update_memory_size; |
11317 | struct imsm_update_size_change *u; | |
f3871fdc | 11318 | |
1ade5cc1 | 11319 | dprintf("(enter) New size = %llu\n", geo->size); |
f3871fdc AK |
11320 | |
11321 | /* size of all update data without anchor */ | |
11322 | update_memory_size = sizeof(struct imsm_update_size_change); | |
11323 | ||
503975b9 | 11324 | u = xcalloc(1, update_memory_size); |
f3871fdc AK |
11325 | u->type = update_size_change; |
11326 | u->subdev = super->current_vol; | |
11327 | u->new_size = geo->size; | |
11328 | ||
11329 | dprintf("imsm: reshape update preparation : OK\n"); | |
11330 | *updatep = u; | |
11331 | ||
11332 | return update_memory_size; | |
11333 | } | |
11334 | ||
48c5303a PC |
11335 | /****************************************************************************** |
11336 | * function: imsm_create_metadata_update_for_migration() | |
11337 | * Creates update for IMSM array. | |
11338 | * | |
11339 | ******************************************************************************/ | |
11340 | static int imsm_create_metadata_update_for_migration( | |
11341 | struct supertype *st, | |
11342 | struct geo_params *geo, | |
11343 | struct imsm_update_reshape_migration **updatep) | |
11344 | { | |
11345 | struct intel_super *super = st->sb; | |
594dc1b8 JS |
11346 | int update_memory_size; |
11347 | struct imsm_update_reshape_migration *u; | |
48c5303a PC |
11348 | struct imsm_dev *dev; |
11349 | int previous_level = -1; | |
11350 | ||
1ade5cc1 | 11351 | dprintf("(enter) New Level = %i\n", geo->level); |
48c5303a PC |
11352 | |
11353 | /* size of all update data without anchor */ | |
11354 | update_memory_size = sizeof(struct imsm_update_reshape_migration); | |
11355 | ||
503975b9 | 11356 | u = xcalloc(1, update_memory_size); |
48c5303a PC |
11357 | u->type = update_reshape_migration; |
11358 | u->subdev = super->current_vol; | |
11359 | u->new_level = geo->level; | |
11360 | u->new_layout = geo->layout; | |
11361 | u->new_raid_disks = u->old_raid_disks = geo->raid_disks; | |
11362 | u->new_disks[0] = -1; | |
4bba0439 | 11363 | u->new_chunksize = -1; |
48c5303a PC |
11364 | |
11365 | dev = get_imsm_dev(super, u->subdev); | |
11366 | if (dev) { | |
11367 | struct imsm_map *map; | |
11368 | ||
238c0a71 | 11369 | map = get_imsm_map(dev, MAP_0); |
4bba0439 PC |
11370 | if (map) { |
11371 | int current_chunk_size = | |
11372 | __le16_to_cpu(map->blocks_per_strip) / 2; | |
11373 | ||
11374 | if (geo->chunksize != current_chunk_size) { | |
11375 | u->new_chunksize = geo->chunksize / 1024; | |
7a862a02 | 11376 | dprintf("imsm: chunk size change from %i to %i\n", |
4bba0439 PC |
11377 | current_chunk_size, u->new_chunksize); |
11378 | } | |
48c5303a | 11379 | previous_level = map->raid_level; |
4bba0439 | 11380 | } |
48c5303a | 11381 | } |
089f9d79 | 11382 | if (geo->level == 5 && previous_level == 0) { |
48c5303a PC |
11383 | struct mdinfo *spares = NULL; |
11384 | ||
11385 | u->new_raid_disks++; | |
11386 | spares = get_spares_for_grow(st); | |
089f9d79 | 11387 | if (spares == NULL || spares->array.spare_disks < 1) { |
48c5303a PC |
11388 | free(u); |
11389 | sysfs_free(spares); | |
11390 | update_memory_size = 0; | |
565cc99e | 11391 | pr_err("cannot get spare device for requested migration\n"); |
48c5303a PC |
11392 | return 0; |
11393 | } | |
11394 | sysfs_free(spares); | |
11395 | } | |
11396 | dprintf("imsm: reshape update preparation : OK\n"); | |
11397 | *updatep = u; | |
11398 | ||
11399 | return update_memory_size; | |
11400 | } | |
11401 | ||
8dd70bce AK |
11402 | static void imsm_update_metadata_locally(struct supertype *st, |
11403 | void *buf, int len) | |
11404 | { | |
11405 | struct metadata_update mu; | |
11406 | ||
11407 | mu.buf = buf; | |
11408 | mu.len = len; | |
11409 | mu.space = NULL; | |
11410 | mu.space_list = NULL; | |
11411 | mu.next = NULL; | |
5fe6f031 N |
11412 | if (imsm_prepare_update(st, &mu)) |
11413 | imsm_process_update(st, &mu); | |
8dd70bce AK |
11414 | |
11415 | while (mu.space_list) { | |
11416 | void **space = mu.space_list; | |
11417 | mu.space_list = *space; | |
11418 | free(space); | |
11419 | } | |
11420 | } | |
78b10e66 | 11421 | |
471bceb6 | 11422 | /*************************************************************************** |
694575e7 | 11423 | * Function: imsm_analyze_change |
471bceb6 | 11424 | * Description: Function analyze change for single volume |
1011e834 | 11425 | * and validate if transition is supported |
fbf3d202 AK |
11426 | * Parameters: Geometry parameters, supertype structure, |
11427 | * metadata change direction (apply/rollback) | |
694575e7 | 11428 | * Returns: Operation type code on success, -1 if fail |
471bceb6 KW |
11429 | ****************************************************************************/ |
11430 | enum imsm_reshape_type imsm_analyze_change(struct supertype *st, | |
fbf3d202 AK |
11431 | struct geo_params *geo, |
11432 | int direction) | |
694575e7 | 11433 | { |
471bceb6 KW |
11434 | struct mdinfo info; |
11435 | int change = -1; | |
11436 | int check_devs = 0; | |
c21e737b | 11437 | int chunk; |
67a2db32 AK |
11438 | /* number of added/removed disks in operation result */ |
11439 | int devNumChange = 0; | |
11440 | /* imsm compatible layout value for array geometry verification */ | |
11441 | int imsm_layout = -1; | |
7abc9871 AK |
11442 | int data_disks; |
11443 | struct imsm_dev *dev; | |
9529d343 | 11444 | struct imsm_map *map; |
7abc9871 | 11445 | struct intel_super *super; |
d04f65f4 | 11446 | unsigned long long current_size; |
65d38cca | 11447 | unsigned long long free_size; |
d04f65f4 | 11448 | unsigned long long max_size; |
65d38cca | 11449 | int rv; |
471bceb6 KW |
11450 | |
11451 | getinfo_super_imsm_volume(st, &info, NULL); | |
089f9d79 JS |
11452 | if (geo->level != info.array.level && geo->level >= 0 && |
11453 | geo->level != UnSet) { | |
471bceb6 KW |
11454 | switch (info.array.level) { |
11455 | case 0: | |
11456 | if (geo->level == 5) { | |
b5347799 | 11457 | change = CH_MIGRATION; |
e13ce846 | 11458 | if (geo->layout != ALGORITHM_LEFT_ASYMMETRIC) { |
7a862a02 | 11459 | pr_err("Error. Requested Layout not supported (left-asymmetric layout is supported only)!\n"); |
e13ce846 AK |
11460 | change = -1; |
11461 | goto analyse_change_exit; | |
11462 | } | |
67a2db32 | 11463 | imsm_layout = geo->layout; |
471bceb6 | 11464 | check_devs = 1; |
e91a3bad LM |
11465 | devNumChange = 1; /* parity disk added */ |
11466 | } else if (geo->level == 10) { | |
471bceb6 KW |
11467 | change = CH_TAKEOVER; |
11468 | check_devs = 1; | |
e91a3bad | 11469 | devNumChange = 2; /* two mirrors added */ |
67a2db32 | 11470 | imsm_layout = 0x102; /* imsm supported layout */ |
471bceb6 | 11471 | } |
dfe77a9e KW |
11472 | break; |
11473 | case 1: | |
471bceb6 KW |
11474 | case 10: |
11475 | if (geo->level == 0) { | |
11476 | change = CH_TAKEOVER; | |
11477 | check_devs = 1; | |
e91a3bad | 11478 | devNumChange = -(geo->raid_disks/2); |
67a2db32 | 11479 | imsm_layout = 0; /* imsm raid0 layout */ |
471bceb6 KW |
11480 | } |
11481 | break; | |
11482 | } | |
11483 | if (change == -1) { | |
7a862a02 | 11484 | pr_err("Error. Level Migration from %d to %d not supported!\n", |
e7b84f9d | 11485 | info.array.level, geo->level); |
471bceb6 KW |
11486 | goto analyse_change_exit; |
11487 | } | |
11488 | } else | |
11489 | geo->level = info.array.level; | |
11490 | ||
089f9d79 JS |
11491 | if (geo->layout != info.array.layout && |
11492 | (geo->layout != UnSet && geo->layout != -1)) { | |
b5347799 | 11493 | change = CH_MIGRATION; |
089f9d79 JS |
11494 | if (info.array.layout == 0 && info.array.level == 5 && |
11495 | geo->layout == 5) { | |
471bceb6 | 11496 | /* reshape 5 -> 4 */ |
089f9d79 JS |
11497 | } else if (info.array.layout == 5 && info.array.level == 5 && |
11498 | geo->layout == 0) { | |
471bceb6 KW |
11499 | /* reshape 4 -> 5 */ |
11500 | geo->layout = 0; | |
11501 | geo->level = 5; | |
11502 | } else { | |
7a862a02 | 11503 | pr_err("Error. Layout Migration from %d to %d not supported!\n", |
e7b84f9d | 11504 | info.array.layout, geo->layout); |
471bceb6 KW |
11505 | change = -1; |
11506 | goto analyse_change_exit; | |
11507 | } | |
67a2db32 | 11508 | } else { |
471bceb6 | 11509 | geo->layout = info.array.layout; |
67a2db32 AK |
11510 | if (imsm_layout == -1) |
11511 | imsm_layout = info.array.layout; | |
11512 | } | |
471bceb6 | 11513 | |
089f9d79 JS |
11514 | if (geo->chunksize > 0 && geo->chunksize != UnSet && |
11515 | geo->chunksize != info.array.chunk_size) { | |
2d2b0eb7 MD |
11516 | if (info.array.level == 10) { |
11517 | pr_err("Error. Chunk size change for RAID 10 is not supported.\n"); | |
11518 | change = -1; | |
11519 | goto analyse_change_exit; | |
1e9b2c3f PB |
11520 | } else if (info.component_size % (geo->chunksize/512)) { |
11521 | pr_err("New chunk size (%dK) does not evenly divide device size (%lluk). Aborting...\n", | |
11522 | geo->chunksize/1024, info.component_size/2); | |
11523 | change = -1; | |
11524 | goto analyse_change_exit; | |
2d2b0eb7 | 11525 | } |
b5347799 | 11526 | change = CH_MIGRATION; |
2d2b0eb7 | 11527 | } else { |
471bceb6 | 11528 | geo->chunksize = info.array.chunk_size; |
2d2b0eb7 | 11529 | } |
471bceb6 | 11530 | |
c21e737b | 11531 | chunk = geo->chunksize / 1024; |
7abc9871 AK |
11532 | |
11533 | super = st->sb; | |
11534 | dev = get_imsm_dev(super, super->current_vol); | |
9529d343 MD |
11535 | map = get_imsm_map(dev, MAP_0); |
11536 | data_disks = imsm_num_data_members(map); | |
c41e00b2 | 11537 | /* compute current size per disk member |
7abc9871 | 11538 | */ |
c41e00b2 AK |
11539 | current_size = info.custom_array_size / data_disks; |
11540 | ||
089f9d79 | 11541 | if (geo->size > 0 && geo->size != MAX_SIZE) { |
c41e00b2 AK |
11542 | /* align component size |
11543 | */ | |
3e684231 | 11544 | geo->size = imsm_component_size_alignment_check( |
c41e00b2 | 11545 | get_imsm_raid_level(dev->vol.map), |
f36a9ecd | 11546 | chunk * 1024, super->sector_size, |
c41e00b2 | 11547 | geo->size * 2); |
65d0b4ce | 11548 | if (geo->size == 0) { |
7a862a02 | 11549 | pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is 0).\n", |
65d0b4ce LD |
11550 | current_size); |
11551 | goto analyse_change_exit; | |
11552 | } | |
c41e00b2 | 11553 | } |
7abc9871 | 11554 | |
089f9d79 | 11555 | if (current_size != geo->size && geo->size > 0) { |
7abc9871 | 11556 | if (change != -1) { |
7a862a02 | 11557 | pr_err("Error. Size change should be the only one at a time.\n"); |
7abc9871 AK |
11558 | change = -1; |
11559 | goto analyse_change_exit; | |
11560 | } | |
11561 | if ((super->current_vol + 1) != super->anchor->num_raid_devs) { | |
7a862a02 | 11562 | pr_err("Error. The last volume in container can be expanded only (%i/%s).\n", |
4dd2df09 | 11563 | super->current_vol, st->devnm); |
7abc9871 AK |
11564 | goto analyse_change_exit; |
11565 | } | |
65d38cca LD |
11566 | /* check the maximum available size |
11567 | */ | |
11568 | rv = imsm_get_free_size(st, dev->vol.map->num_members, | |
11569 | 0, chunk, &free_size); | |
11570 | if (rv == 0) | |
11571 | /* Cannot find maximum available space | |
11572 | */ | |
11573 | max_size = 0; | |
11574 | else { | |
11575 | max_size = free_size + current_size; | |
11576 | /* align component size | |
11577 | */ | |
3e684231 | 11578 | max_size = imsm_component_size_alignment_check( |
65d38cca | 11579 | get_imsm_raid_level(dev->vol.map), |
f36a9ecd | 11580 | chunk * 1024, super->sector_size, |
65d38cca LD |
11581 | max_size); |
11582 | } | |
d04f65f4 | 11583 | if (geo->size == MAX_SIZE) { |
b130333f AK |
11584 | /* requested size change to the maximum available size |
11585 | */ | |
65d38cca | 11586 | if (max_size == 0) { |
7a862a02 | 11587 | pr_err("Error. Cannot find maximum available space.\n"); |
b130333f AK |
11588 | change = -1; |
11589 | goto analyse_change_exit; | |
65d38cca LD |
11590 | } else |
11591 | geo->size = max_size; | |
c41e00b2 | 11592 | } |
b130333f | 11593 | |
681b7ae2 | 11594 | if (direction == ROLLBACK_METADATA_CHANGES) { |
fbf3d202 AK |
11595 | /* accept size for rollback only |
11596 | */ | |
11597 | } else { | |
11598 | /* round size due to metadata compatibility | |
11599 | */ | |
11600 | geo->size = (geo->size >> SECT_PER_MB_SHIFT) | |
11601 | << SECT_PER_MB_SHIFT; | |
11602 | dprintf("Prepare update for size change to %llu\n", | |
11603 | geo->size ); | |
11604 | if (current_size >= geo->size) { | |
7a862a02 | 11605 | pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is %llu).\n", |
e7b84f9d | 11606 | current_size, geo->size); |
fbf3d202 AK |
11607 | goto analyse_change_exit; |
11608 | } | |
65d38cca | 11609 | if (max_size && geo->size > max_size) { |
7a862a02 | 11610 | pr_err("Error. Requested size is larger than maximum available size (maximum available size is %llu, requested size /rounded/ is %llu).\n", |
e7b84f9d | 11611 | max_size, geo->size); |
65d38cca LD |
11612 | goto analyse_change_exit; |
11613 | } | |
7abc9871 AK |
11614 | } |
11615 | geo->size *= data_disks; | |
11616 | geo->raid_disks = dev->vol.map->num_members; | |
11617 | change = CH_ARRAY_SIZE; | |
11618 | } | |
471bceb6 KW |
11619 | if (!validate_geometry_imsm(st, |
11620 | geo->level, | |
67a2db32 | 11621 | imsm_layout, |
e91a3bad | 11622 | geo->raid_disks + devNumChange, |
c21e737b | 11623 | &chunk, |
af4348dd | 11624 | geo->size, INVALID_SECTORS, |
5308f117 | 11625 | 0, 0, info.consistency_policy, 1)) |
471bceb6 KW |
11626 | change = -1; |
11627 | ||
11628 | if (check_devs) { | |
11629 | struct intel_super *super = st->sb; | |
11630 | struct imsm_super *mpb = super->anchor; | |
11631 | ||
11632 | if (mpb->num_raid_devs > 1) { | |
7a862a02 | 11633 | pr_err("Error. Cannot perform operation on %s- for this operation it MUST be single array in container\n", |
e7b84f9d | 11634 | geo->dev_name); |
471bceb6 KW |
11635 | change = -1; |
11636 | } | |
11637 | } | |
11638 | ||
11639 | analyse_change_exit: | |
089f9d79 JS |
11640 | if (direction == ROLLBACK_METADATA_CHANGES && |
11641 | (change == CH_MIGRATION || change == CH_TAKEOVER)) { | |
7a862a02 | 11642 | dprintf("imsm: Metadata changes rollback is not supported for migration and takeover operations.\n"); |
fbf3d202 AK |
11643 | change = -1; |
11644 | } | |
471bceb6 | 11645 | return change; |
694575e7 KW |
11646 | } |
11647 | ||
bb025c2f KW |
11648 | int imsm_takeover(struct supertype *st, struct geo_params *geo) |
11649 | { | |
11650 | struct intel_super *super = st->sb; | |
11651 | struct imsm_update_takeover *u; | |
11652 | ||
503975b9 | 11653 | u = xmalloc(sizeof(struct imsm_update_takeover)); |
bb025c2f KW |
11654 | |
11655 | u->type = update_takeover; | |
11656 | u->subarray = super->current_vol; | |
11657 | ||
11658 | /* 10->0 transition */ | |
11659 | if (geo->level == 0) | |
11660 | u->direction = R10_TO_R0; | |
11661 | ||
0529c688 KW |
11662 | /* 0->10 transition */ |
11663 | if (geo->level == 10) | |
11664 | u->direction = R0_TO_R10; | |
11665 | ||
bb025c2f KW |
11666 | /* update metadata locally */ |
11667 | imsm_update_metadata_locally(st, u, | |
11668 | sizeof(struct imsm_update_takeover)); | |
11669 | /* and possibly remotely */ | |
11670 | if (st->update_tail) | |
11671 | append_metadata_update(st, u, | |
11672 | sizeof(struct imsm_update_takeover)); | |
11673 | else | |
11674 | free(u); | |
11675 | ||
11676 | return 0; | |
11677 | } | |
11678 | ||
d04f65f4 N |
11679 | static int imsm_reshape_super(struct supertype *st, unsigned long long size, |
11680 | int level, | |
78b10e66 | 11681 | int layout, int chunksize, int raid_disks, |
41784c88 | 11682 | int delta_disks, char *backup, char *dev, |
016e00f5 | 11683 | int direction, int verbose) |
78b10e66 | 11684 | { |
78b10e66 N |
11685 | int ret_val = 1; |
11686 | struct geo_params geo; | |
11687 | ||
1ade5cc1 | 11688 | dprintf("(enter)\n"); |
78b10e66 | 11689 | |
71204a50 | 11690 | memset(&geo, 0, sizeof(struct geo_params)); |
78b10e66 N |
11691 | |
11692 | geo.dev_name = dev; | |
4dd2df09 | 11693 | strcpy(geo.devnm, st->devnm); |
78b10e66 N |
11694 | geo.size = size; |
11695 | geo.level = level; | |
11696 | geo.layout = layout; | |
11697 | geo.chunksize = chunksize; | |
11698 | geo.raid_disks = raid_disks; | |
41784c88 AK |
11699 | if (delta_disks != UnSet) |
11700 | geo.raid_disks += delta_disks; | |
78b10e66 | 11701 | |
1ade5cc1 N |
11702 | dprintf("for level : %i\n", geo.level); |
11703 | dprintf("for raid_disks : %i\n", geo.raid_disks); | |
78b10e66 | 11704 | |
4dd2df09 | 11705 | if (strcmp(st->container_devnm, st->devnm) == 0) { |
694575e7 KW |
11706 | /* On container level we can only increase number of devices. */ |
11707 | dprintf("imsm: info: Container operation\n"); | |
78b10e66 | 11708 | int old_raid_disks = 0; |
6dc0be30 | 11709 | |
78b10e66 | 11710 | if (imsm_reshape_is_allowed_on_container( |
fbf3d202 | 11711 | st, &geo, &old_raid_disks, direction)) { |
78b10e66 N |
11712 | struct imsm_update_reshape *u = NULL; |
11713 | int len; | |
11714 | ||
11715 | len = imsm_create_metadata_update_for_reshape( | |
11716 | st, &geo, old_raid_disks, &u); | |
11717 | ||
ed08d51c AK |
11718 | if (len <= 0) { |
11719 | dprintf("imsm: Cannot prepare update\n"); | |
11720 | goto exit_imsm_reshape_super; | |
11721 | } | |
11722 | ||
8dd70bce AK |
11723 | ret_val = 0; |
11724 | /* update metadata locally */ | |
11725 | imsm_update_metadata_locally(st, u, len); | |
11726 | /* and possibly remotely */ | |
11727 | if (st->update_tail) | |
11728 | append_metadata_update(st, u, len); | |
11729 | else | |
ed08d51c | 11730 | free(u); |
8dd70bce | 11731 | |
694575e7 | 11732 | } else { |
7a862a02 | 11733 | pr_err("(imsm) Operation is not allowed on this container\n"); |
694575e7 KW |
11734 | } |
11735 | } else { | |
11736 | /* On volume level we support following operations | |
471bceb6 KW |
11737 | * - takeover: raid10 -> raid0; raid0 -> raid10 |
11738 | * - chunk size migration | |
11739 | * - migration: raid5 -> raid0; raid0 -> raid5 | |
11740 | */ | |
11741 | struct intel_super *super = st->sb; | |
11742 | struct intel_dev *dev = super->devlist; | |
4dd2df09 | 11743 | int change; |
694575e7 | 11744 | dprintf("imsm: info: Volume operation\n"); |
471bceb6 KW |
11745 | /* find requested device */ |
11746 | while (dev) { | |
1011e834 | 11747 | char *devnm = |
4dd2df09 N |
11748 | imsm_find_array_devnm_by_subdev( |
11749 | dev->index, st->container_devnm); | |
11750 | if (devnm && strcmp(devnm, geo.devnm) == 0) | |
471bceb6 KW |
11751 | break; |
11752 | dev = dev->next; | |
11753 | } | |
11754 | if (dev == NULL) { | |
4dd2df09 N |
11755 | pr_err("Cannot find %s (%s) subarray\n", |
11756 | geo.dev_name, geo.devnm); | |
471bceb6 KW |
11757 | goto exit_imsm_reshape_super; |
11758 | } | |
11759 | super->current_vol = dev->index; | |
fbf3d202 | 11760 | change = imsm_analyze_change(st, &geo, direction); |
694575e7 | 11761 | switch (change) { |
471bceb6 | 11762 | case CH_TAKEOVER: |
bb025c2f | 11763 | ret_val = imsm_takeover(st, &geo); |
694575e7 | 11764 | break; |
48c5303a PC |
11765 | case CH_MIGRATION: { |
11766 | struct imsm_update_reshape_migration *u = NULL; | |
11767 | int len = | |
11768 | imsm_create_metadata_update_for_migration( | |
11769 | st, &geo, &u); | |
11770 | if (len < 1) { | |
7a862a02 | 11771 | dprintf("imsm: Cannot prepare update\n"); |
48c5303a PC |
11772 | break; |
11773 | } | |
471bceb6 | 11774 | ret_val = 0; |
48c5303a PC |
11775 | /* update metadata locally */ |
11776 | imsm_update_metadata_locally(st, u, len); | |
11777 | /* and possibly remotely */ | |
11778 | if (st->update_tail) | |
11779 | append_metadata_update(st, u, len); | |
11780 | else | |
11781 | free(u); | |
11782 | } | |
11783 | break; | |
7abc9871 | 11784 | case CH_ARRAY_SIZE: { |
f3871fdc AK |
11785 | struct imsm_update_size_change *u = NULL; |
11786 | int len = | |
11787 | imsm_create_metadata_update_for_size_change( | |
11788 | st, &geo, &u); | |
11789 | if (len < 1) { | |
7a862a02 | 11790 | dprintf("imsm: Cannot prepare update\n"); |
f3871fdc AK |
11791 | break; |
11792 | } | |
11793 | ret_val = 0; | |
11794 | /* update metadata locally */ | |
11795 | imsm_update_metadata_locally(st, u, len); | |
11796 | /* and possibly remotely */ | |
11797 | if (st->update_tail) | |
11798 | append_metadata_update(st, u, len); | |
11799 | else | |
11800 | free(u); | |
7abc9871 AK |
11801 | } |
11802 | break; | |
471bceb6 KW |
11803 | default: |
11804 | ret_val = 1; | |
694575e7 | 11805 | } |
694575e7 | 11806 | } |
78b10e66 | 11807 | |
ed08d51c | 11808 | exit_imsm_reshape_super: |
78b10e66 N |
11809 | dprintf("imsm: reshape_super Exit code = %i\n", ret_val); |
11810 | return ret_val; | |
11811 | } | |
2cda7640 | 11812 | |
0febb20c AO |
11813 | #define COMPLETED_OK 0 |
11814 | #define COMPLETED_NONE 1 | |
11815 | #define COMPLETED_DELAYED 2 | |
11816 | ||
11817 | static int read_completed(int fd, unsigned long long *val) | |
11818 | { | |
11819 | int ret; | |
11820 | char buf[50]; | |
11821 | ||
11822 | ret = sysfs_fd_get_str(fd, buf, 50); | |
11823 | if (ret < 0) | |
11824 | return ret; | |
11825 | ||
11826 | ret = COMPLETED_OK; | |
11827 | if (strncmp(buf, "none", 4) == 0) { | |
11828 | ret = COMPLETED_NONE; | |
11829 | } else if (strncmp(buf, "delayed", 7) == 0) { | |
11830 | ret = COMPLETED_DELAYED; | |
11831 | } else { | |
11832 | char *ep; | |
11833 | *val = strtoull(buf, &ep, 0); | |
11834 | if (ep == buf || (*ep != 0 && *ep != '\n' && *ep != ' ')) | |
11835 | ret = -1; | |
11836 | } | |
11837 | return ret; | |
11838 | } | |
11839 | ||
eee67a47 AK |
11840 | /******************************************************************************* |
11841 | * Function: wait_for_reshape_imsm | |
11842 | * Description: Function writes new sync_max value and waits until | |
11843 | * reshape process reach new position | |
11844 | * Parameters: | |
11845 | * sra : general array info | |
eee67a47 AK |
11846 | * ndata : number of disks in new array's layout |
11847 | * Returns: | |
11848 | * 0 : success, | |
11849 | * 1 : there is no reshape in progress, | |
11850 | * -1 : fail | |
11851 | ******************************************************************************/ | |
ae9f01f8 | 11852 | int wait_for_reshape_imsm(struct mdinfo *sra, int ndata) |
eee67a47 | 11853 | { |
85ca499c | 11854 | int fd = sysfs_get_fd(sra, NULL, "sync_completed"); |
df2647fa | 11855 | int retry = 3; |
eee67a47 | 11856 | unsigned long long completed; |
ae9f01f8 AK |
11857 | /* to_complete : new sync_max position */ |
11858 | unsigned long long to_complete = sra->reshape_progress; | |
11859 | unsigned long long position_to_set = to_complete / ndata; | |
eee67a47 | 11860 | |
ae9f01f8 | 11861 | if (fd < 0) { |
1ade5cc1 | 11862 | dprintf("cannot open reshape_position\n"); |
eee67a47 | 11863 | return 1; |
ae9f01f8 | 11864 | } |
eee67a47 | 11865 | |
df2647fa PB |
11866 | do { |
11867 | if (sysfs_fd_get_ll(fd, &completed) < 0) { | |
11868 | if (!retry) { | |
11869 | dprintf("cannot read reshape_position (no reshape in progres)\n"); | |
11870 | close(fd); | |
11871 | return 1; | |
11872 | } | |
11873 | usleep(30000); | |
11874 | } else | |
11875 | break; | |
11876 | } while (retry--); | |
eee67a47 | 11877 | |
85ca499c | 11878 | if (completed > position_to_set) { |
1ade5cc1 | 11879 | dprintf("wrong next position to set %llu (%llu)\n", |
85ca499c | 11880 | to_complete, position_to_set); |
ae9f01f8 AK |
11881 | close(fd); |
11882 | return -1; | |
11883 | } | |
11884 | dprintf("Position set: %llu\n", position_to_set); | |
11885 | if (sysfs_set_num(sra, NULL, "sync_max", | |
11886 | position_to_set) != 0) { | |
1ade5cc1 | 11887 | dprintf("cannot set reshape position to %llu\n", |
ae9f01f8 AK |
11888 | position_to_set); |
11889 | close(fd); | |
11890 | return -1; | |
eee67a47 AK |
11891 | } |
11892 | ||
eee67a47 | 11893 | do { |
0febb20c | 11894 | int rc; |
eee67a47 | 11895 | char action[20]; |
5ff3a780 | 11896 | int timeout = 3000; |
0febb20c | 11897 | |
5ff3a780 | 11898 | sysfs_wait(fd, &timeout); |
a47e44fb AK |
11899 | if (sysfs_get_str(sra, NULL, "sync_action", |
11900 | action, 20) > 0 && | |
d7d3809a | 11901 | strncmp(action, "reshape", 7) != 0) { |
b2be2b62 AO |
11902 | if (strncmp(action, "idle", 4) == 0) |
11903 | break; | |
d7d3809a AP |
11904 | close(fd); |
11905 | return -1; | |
11906 | } | |
0febb20c AO |
11907 | |
11908 | rc = read_completed(fd, &completed); | |
11909 | if (rc < 0) { | |
1ade5cc1 | 11910 | dprintf("cannot read reshape_position (in loop)\n"); |
eee67a47 AK |
11911 | close(fd); |
11912 | return 1; | |
0febb20c AO |
11913 | } else if (rc == COMPLETED_NONE) |
11914 | break; | |
85ca499c | 11915 | } while (completed < position_to_set); |
b2be2b62 | 11916 | |
eee67a47 AK |
11917 | close(fd); |
11918 | return 0; | |
eee67a47 AK |
11919 | } |
11920 | ||
b915c95f AK |
11921 | /******************************************************************************* |
11922 | * Function: check_degradation_change | |
11923 | * Description: Check that array hasn't become failed. | |
11924 | * Parameters: | |
11925 | * info : for sysfs access | |
11926 | * sources : source disks descriptors | |
11927 | * degraded: previous degradation level | |
11928 | * Returns: | |
11929 | * degradation level | |
11930 | ******************************************************************************/ | |
11931 | int check_degradation_change(struct mdinfo *info, | |
11932 | int *sources, | |
11933 | int degraded) | |
11934 | { | |
11935 | unsigned long long new_degraded; | |
e1993023 LD |
11936 | int rv; |
11937 | ||
11938 | rv = sysfs_get_ll(info, NULL, "degraded", &new_degraded); | |
089f9d79 | 11939 | if (rv == -1 || (new_degraded != (unsigned long long)degraded)) { |
b915c95f AK |
11940 | /* check each device to ensure it is still working */ |
11941 | struct mdinfo *sd; | |
11942 | new_degraded = 0; | |
11943 | for (sd = info->devs ; sd ; sd = sd->next) { | |
11944 | if (sd->disk.state & (1<<MD_DISK_FAULTY)) | |
11945 | continue; | |
11946 | if (sd->disk.state & (1<<MD_DISK_SYNC)) { | |
cf52eff5 TM |
11947 | char sbuf[100]; |
11948 | ||
b915c95f | 11949 | if (sysfs_get_str(info, |
cf52eff5 | 11950 | sd, "state", sbuf, sizeof(sbuf)) < 0 || |
b915c95f AK |
11951 | strstr(sbuf, "faulty") || |
11952 | strstr(sbuf, "in_sync") == NULL) { | |
11953 | /* this device is dead */ | |
11954 | sd->disk.state = (1<<MD_DISK_FAULTY); | |
11955 | if (sd->disk.raid_disk >= 0 && | |
11956 | sources[sd->disk.raid_disk] >= 0) { | |
11957 | close(sources[ | |
11958 | sd->disk.raid_disk]); | |
11959 | sources[sd->disk.raid_disk] = | |
11960 | -1; | |
11961 | } | |
11962 | new_degraded++; | |
11963 | } | |
11964 | } | |
11965 | } | |
11966 | } | |
11967 | ||
11968 | return new_degraded; | |
11969 | } | |
11970 | ||
10f22854 AK |
11971 | /******************************************************************************* |
11972 | * Function: imsm_manage_reshape | |
11973 | * Description: Function finds array under reshape and it manages reshape | |
11974 | * process. It creates stripes backups (if required) and sets | |
942e1cdb | 11975 | * checkpoints. |
10f22854 AK |
11976 | * Parameters: |
11977 | * afd : Backup handle (nattive) - not used | |
11978 | * sra : general array info | |
11979 | * reshape : reshape parameters - not used | |
11980 | * st : supertype structure | |
11981 | * blocks : size of critical section [blocks] | |
11982 | * fds : table of source device descriptor | |
11983 | * offsets : start of array (offest per devices) | |
11984 | * dests : not used | |
11985 | * destfd : table of destination device descriptor | |
11986 | * destoffsets : table of destination offsets (per device) | |
11987 | * Returns: | |
11988 | * 1 : success, reshape is done | |
11989 | * 0 : fail | |
11990 | ******************************************************************************/ | |
999b4972 N |
11991 | static int imsm_manage_reshape( |
11992 | int afd, struct mdinfo *sra, struct reshape *reshape, | |
10f22854 | 11993 | struct supertype *st, unsigned long backup_blocks, |
999b4972 N |
11994 | int *fds, unsigned long long *offsets, |
11995 | int dests, int *destfd, unsigned long long *destoffsets) | |
11996 | { | |
10f22854 AK |
11997 | int ret_val = 0; |
11998 | struct intel_super *super = st->sb; | |
594dc1b8 | 11999 | struct intel_dev *dv; |
de44e46f | 12000 | unsigned int sector_size = super->sector_size; |
10f22854 | 12001 | struct imsm_dev *dev = NULL; |
9529d343 | 12002 | struct imsm_map *map_src, *map_dest; |
10f22854 AK |
12003 | int migr_vol_qan = 0; |
12004 | int ndata, odata; /* [bytes] */ | |
12005 | int chunk; /* [bytes] */ | |
12006 | struct migr_record *migr_rec; | |
12007 | char *buf = NULL; | |
12008 | unsigned int buf_size; /* [bytes] */ | |
12009 | unsigned long long max_position; /* array size [bytes] */ | |
12010 | unsigned long long next_step; /* [blocks]/[bytes] */ | |
12011 | unsigned long long old_data_stripe_length; | |
10f22854 AK |
12012 | unsigned long long start_src; /* [bytes] */ |
12013 | unsigned long long start; /* [bytes] */ | |
12014 | unsigned long long start_buf_shift; /* [bytes] */ | |
b915c95f | 12015 | int degraded = 0; |
ab724b98 | 12016 | int source_layout = 0; |
10f22854 | 12017 | |
79a16a9b JS |
12018 | if (!sra) |
12019 | return ret_val; | |
12020 | ||
12021 | if (!fds || !offsets) | |
10f22854 AK |
12022 | goto abort; |
12023 | ||
12024 | /* Find volume during the reshape */ | |
12025 | for (dv = super->devlist; dv; dv = dv->next) { | |
fc54fe7a JS |
12026 | if (dv->dev->vol.migr_type == MIGR_GEN_MIGR && |
12027 | dv->dev->vol.migr_state == 1) { | |
10f22854 AK |
12028 | dev = dv->dev; |
12029 | migr_vol_qan++; | |
12030 | } | |
12031 | } | |
12032 | /* Only one volume can migrate at the same time */ | |
12033 | if (migr_vol_qan != 1) { | |
676e87a8 | 12034 | pr_err("%s", migr_vol_qan ? |
10f22854 AK |
12035 | "Number of migrating volumes greater than 1\n" : |
12036 | "There is no volume during migrationg\n"); | |
12037 | goto abort; | |
12038 | } | |
12039 | ||
9529d343 | 12040 | map_dest = get_imsm_map(dev, MAP_0); |
238c0a71 | 12041 | map_src = get_imsm_map(dev, MAP_1); |
10f22854 AK |
12042 | if (map_src == NULL) |
12043 | goto abort; | |
10f22854 | 12044 | |
9529d343 MD |
12045 | ndata = imsm_num_data_members(map_dest); |
12046 | odata = imsm_num_data_members(map_src); | |
10f22854 | 12047 | |
7b1ab482 | 12048 | chunk = __le16_to_cpu(map_src->blocks_per_strip) * 512; |
10f22854 AK |
12049 | old_data_stripe_length = odata * chunk; |
12050 | ||
12051 | migr_rec = super->migr_rec; | |
12052 | ||
10f22854 AK |
12053 | /* initialize migration record for start condition */ |
12054 | if (sra->reshape_progress == 0) | |
12055 | init_migr_record_imsm(st, dev, sra); | |
b2c59438 AK |
12056 | else { |
12057 | if (__le32_to_cpu(migr_rec->rec_status) != UNIT_SRC_NORMAL) { | |
7a862a02 | 12058 | dprintf("imsm: cannot restart migration when data are present in copy area.\n"); |
b2c59438 AK |
12059 | goto abort; |
12060 | } | |
6a75c8ca AK |
12061 | /* Save checkpoint to update migration record for current |
12062 | * reshape position (in md). It can be farther than current | |
12063 | * reshape position in metadata. | |
12064 | */ | |
12065 | if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) { | |
12066 | /* ignore error == 2, this can mean end of reshape here | |
12067 | */ | |
7a862a02 | 12068 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL, initial save)\n"); |
6a75c8ca AK |
12069 | goto abort; |
12070 | } | |
b2c59438 | 12071 | } |
10f22854 AK |
12072 | |
12073 | /* size for data */ | |
12074 | buf_size = __le32_to_cpu(migr_rec->blocks_per_unit) * 512; | |
12075 | /* extend buffer size for parity disk */ | |
12076 | buf_size += __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
3e684231 | 12077 | /* add space for stripe alignment */ |
10f22854 | 12078 | buf_size += old_data_stripe_length; |
de44e46f PB |
12079 | if (posix_memalign((void **)&buf, MAX_SECTOR_SIZE, buf_size)) { |
12080 | dprintf("imsm: Cannot allocate checkpoint buffer\n"); | |
10f22854 AK |
12081 | goto abort; |
12082 | } | |
12083 | ||
3ef4403c | 12084 | max_position = sra->component_size * ndata; |
68eb8bc6 | 12085 | source_layout = imsm_level_to_layout(map_src->raid_level); |
10f22854 | 12086 | |
9f421827 PB |
12087 | while (current_migr_unit(migr_rec) < |
12088 | get_num_migr_units(migr_rec)) { | |
10f22854 AK |
12089 | /* current reshape position [blocks] */ |
12090 | unsigned long long current_position = | |
12091 | __le32_to_cpu(migr_rec->blocks_per_unit) | |
9f421827 | 12092 | * current_migr_unit(migr_rec); |
10f22854 AK |
12093 | unsigned long long border; |
12094 | ||
b915c95f AK |
12095 | /* Check that array hasn't become failed. |
12096 | */ | |
12097 | degraded = check_degradation_change(sra, fds, degraded); | |
12098 | if (degraded > 1) { | |
7a862a02 | 12099 | dprintf("imsm: Abort reshape due to degradation level (%i)\n", degraded); |
b915c95f AK |
12100 | goto abort; |
12101 | } | |
12102 | ||
10f22854 AK |
12103 | next_step = __le32_to_cpu(migr_rec->blocks_per_unit); |
12104 | ||
12105 | if ((current_position + next_step) > max_position) | |
12106 | next_step = max_position - current_position; | |
12107 | ||
92144abf | 12108 | start = current_position * 512; |
10f22854 | 12109 | |
942e1cdb | 12110 | /* align reading start to old geometry */ |
10f22854 AK |
12111 | start_buf_shift = start % old_data_stripe_length; |
12112 | start_src = start - start_buf_shift; | |
12113 | ||
12114 | border = (start_src / odata) - (start / ndata); | |
12115 | border /= 512; | |
12116 | if (border <= __le32_to_cpu(migr_rec->dest_depth_per_unit)) { | |
12117 | /* save critical stripes to buf | |
12118 | * start - start address of current unit | |
12119 | * to backup [bytes] | |
12120 | * start_src - start address of current unit | |
12121 | * to backup alligned to source array | |
12122 | * [bytes] | |
12123 | */ | |
594dc1b8 | 12124 | unsigned long long next_step_filler; |
10f22854 AK |
12125 | unsigned long long copy_length = next_step * 512; |
12126 | ||
12127 | /* allign copy area length to stripe in old geometry */ | |
12128 | next_step_filler = ((copy_length + start_buf_shift) | |
12129 | % old_data_stripe_length); | |
12130 | if (next_step_filler) | |
12131 | next_step_filler = (old_data_stripe_length | |
12132 | - next_step_filler); | |
7a862a02 | 12133 | 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 |
12134 | start, start_src, copy_length, |
12135 | start_buf_shift, next_step_filler); | |
12136 | ||
12137 | if (save_stripes(fds, offsets, map_src->num_members, | |
ab724b98 AK |
12138 | chunk, map_src->raid_level, |
12139 | source_layout, 0, NULL, start_src, | |
10f22854 AK |
12140 | copy_length + |
12141 | next_step_filler + start_buf_shift, | |
12142 | buf)) { | |
7a862a02 | 12143 | dprintf("imsm: Cannot save stripes to buffer\n"); |
10f22854 AK |
12144 | goto abort; |
12145 | } | |
12146 | /* Convert data to destination format and store it | |
12147 | * in backup general migration area | |
12148 | */ | |
12149 | if (save_backup_imsm(st, dev, sra, | |
aea93171 | 12150 | buf + start_buf_shift, copy_length)) { |
7a862a02 | 12151 | dprintf("imsm: Cannot save stripes to target devices\n"); |
10f22854 AK |
12152 | goto abort; |
12153 | } | |
12154 | if (save_checkpoint_imsm(st, sra, | |
12155 | UNIT_SRC_IN_CP_AREA)) { | |
7a862a02 | 12156 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_IN_CP_AREA)\n"); |
10f22854 AK |
12157 | goto abort; |
12158 | } | |
8016a6d4 AK |
12159 | } else { |
12160 | /* set next step to use whole border area */ | |
12161 | border /= next_step; | |
12162 | if (border > 1) | |
12163 | next_step *= border; | |
10f22854 AK |
12164 | } |
12165 | /* When data backed up, checkpoint stored, | |
12166 | * kick the kernel to reshape unit of data | |
12167 | */ | |
12168 | next_step = next_step + sra->reshape_progress; | |
8016a6d4 AK |
12169 | /* limit next step to array max position */ |
12170 | if (next_step > max_position) | |
12171 | next_step = max_position; | |
10f22854 AK |
12172 | sysfs_set_num(sra, NULL, "suspend_lo", sra->reshape_progress); |
12173 | sysfs_set_num(sra, NULL, "suspend_hi", next_step); | |
ae9f01f8 | 12174 | sra->reshape_progress = next_step; |
10f22854 AK |
12175 | |
12176 | /* wait until reshape finish */ | |
c85338c6 | 12177 | if (wait_for_reshape_imsm(sra, ndata)) { |
c47b0ff6 AK |
12178 | dprintf("wait_for_reshape_imsm returned error!\n"); |
12179 | goto abort; | |
12180 | } | |
84d11e6c N |
12181 | if (sigterm) |
12182 | goto abort; | |
10f22854 | 12183 | |
0228d92c AK |
12184 | if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) { |
12185 | /* ignore error == 2, this can mean end of reshape here | |
12186 | */ | |
7a862a02 | 12187 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL)\n"); |
10f22854 AK |
12188 | goto abort; |
12189 | } | |
12190 | ||
12191 | } | |
12192 | ||
71e5411e PB |
12193 | /* clear migr_rec on disks after successful migration */ |
12194 | struct dl *d; | |
12195 | ||
85337573 | 12196 | memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE); |
71e5411e PB |
12197 | for (d = super->disks; d; d = d->next) { |
12198 | if (d->index < 0 || is_failed(&d->disk)) | |
12199 | continue; | |
12200 | unsigned long long dsize; | |
12201 | ||
12202 | get_dev_size(d->fd, NULL, &dsize); | |
de44e46f | 12203 | if (lseek64(d->fd, dsize - MIGR_REC_SECTOR_POSITION*sector_size, |
71e5411e | 12204 | SEEK_SET) >= 0) { |
466070ad | 12205 | if ((unsigned int)write(d->fd, super->migr_rec_buf, |
de44e46f PB |
12206 | MIGR_REC_BUF_SECTORS*sector_size) != |
12207 | MIGR_REC_BUF_SECTORS*sector_size) | |
71e5411e PB |
12208 | perror("Write migr_rec failed"); |
12209 | } | |
12210 | } | |
12211 | ||
10f22854 AK |
12212 | /* return '1' if done */ |
12213 | ret_val = 1; | |
12214 | abort: | |
12215 | free(buf); | |
942e1cdb N |
12216 | /* See Grow.c: abort_reshape() for further explanation */ |
12217 | sysfs_set_num(sra, NULL, "suspend_lo", 0x7FFFFFFFFFFFFFFFULL); | |
12218 | sysfs_set_num(sra, NULL, "suspend_hi", 0); | |
12219 | sysfs_set_num(sra, NULL, "suspend_lo", 0); | |
10f22854 AK |
12220 | |
12221 | return ret_val; | |
999b4972 | 12222 | } |
0c21b485 | 12223 | |
cdddbdbc | 12224 | struct superswitch super_imsm = { |
cdddbdbc DW |
12225 | .examine_super = examine_super_imsm, |
12226 | .brief_examine_super = brief_examine_super_imsm, | |
4737ae25 | 12227 | .brief_examine_subarrays = brief_examine_subarrays_imsm, |
9d84c8ea | 12228 | .export_examine_super = export_examine_super_imsm, |
cdddbdbc DW |
12229 | .detail_super = detail_super_imsm, |
12230 | .brief_detail_super = brief_detail_super_imsm, | |
bf5a934a | 12231 | .write_init_super = write_init_super_imsm, |
0e600426 N |
12232 | .validate_geometry = validate_geometry_imsm, |
12233 | .add_to_super = add_to_super_imsm, | |
1a64be56 | 12234 | .remove_from_super = remove_from_super_imsm, |
d665cc31 | 12235 | .detail_platform = detail_platform_imsm, |
e50cf220 | 12236 | .export_detail_platform = export_detail_platform_imsm, |
33414a01 | 12237 | .kill_subarray = kill_subarray_imsm, |
aa534678 | 12238 | .update_subarray = update_subarray_imsm, |
2b959fbf | 12239 | .load_container = load_container_imsm, |
71204a50 N |
12240 | .default_geometry = default_geometry_imsm, |
12241 | .get_disk_controller_domain = imsm_get_disk_controller_domain, | |
12242 | .reshape_super = imsm_reshape_super, | |
12243 | .manage_reshape = imsm_manage_reshape, | |
9e2d750d | 12244 | .recover_backup = recover_backup_imsm, |
74db60b0 | 12245 | .copy_metadata = copy_metadata_imsm, |
27156a57 | 12246 | .examine_badblocks = examine_badblocks_imsm, |
cdddbdbc DW |
12247 | .match_home = match_home_imsm, |
12248 | .uuid_from_super= uuid_from_super_imsm, | |
12249 | .getinfo_super = getinfo_super_imsm, | |
5c4cd5da | 12250 | .getinfo_super_disks = getinfo_super_disks_imsm, |
cdddbdbc DW |
12251 | .update_super = update_super_imsm, |
12252 | ||
12253 | .avail_size = avail_size_imsm, | |
fbfdcb06 | 12254 | .get_spare_criteria = get_spare_criteria_imsm, |
cdddbdbc DW |
12255 | |
12256 | .compare_super = compare_super_imsm, | |
12257 | ||
12258 | .load_super = load_super_imsm, | |
bf5a934a | 12259 | .init_super = init_super_imsm, |
e683ca88 | 12260 | .store_super = store_super_imsm, |
cdddbdbc DW |
12261 | .free_super = free_super_imsm, |
12262 | .match_metadata_desc = match_metadata_desc_imsm, | |
bf5a934a | 12263 | .container_content = container_content_imsm, |
0c21b485 | 12264 | .validate_container = validate_container_imsm, |
cdddbdbc | 12265 | |
2432ce9b AP |
12266 | .write_init_ppl = write_init_ppl_imsm, |
12267 | .validate_ppl = validate_ppl_imsm, | |
12268 | ||
cdddbdbc | 12269 | .external = 1, |
4cce4069 | 12270 | .name = "imsm", |
845dea95 NB |
12271 | |
12272 | /* for mdmon */ | |
12273 | .open_new = imsm_open_new, | |
ed9d66aa | 12274 | .set_array_state= imsm_set_array_state, |
845dea95 NB |
12275 | .set_disk = imsm_set_disk, |
12276 | .sync_metadata = imsm_sync_metadata, | |
88758e9d | 12277 | .activate_spare = imsm_activate_spare, |
e8319a19 | 12278 | .process_update = imsm_process_update, |
8273f55e | 12279 | .prepare_update = imsm_prepare_update, |
6f50473f | 12280 | .record_bad_block = imsm_record_badblock, |
c07a5a4f | 12281 | .clear_bad_block = imsm_clear_badblock, |
928f1424 | 12282 | .get_bad_blocks = imsm_get_badblocks, |
cdddbdbc | 12283 | }; |