]>
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 |
c2c087e6 | 91 | #define IMSM_RESERVED_SECTORS 4096 |
b81221b7 | 92 | #define NUM_BLOCKS_DIRTY_STRIPE_REGION 2056 |
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 */ | |
299 | __u32 curr_migr_unit; /* 0..numMigrUnits-1 */ | |
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 */ | |
309 | __u32 ckpt_area_pba; /* Pba of first block of ckpt copy area */ | |
310 | __u32 dest_1st_member_lba; /* First member lba on first | |
311 | * stripe of destination */ | |
312 | __u32 num_migr_units; /* Total num migration units-of-op */ | |
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) */ | |
319 | } __attribute__ ((__packed__)); | |
320 | ||
ec50f7b6 LM |
321 | struct md_list { |
322 | /* usage marker: | |
323 | * 1: load metadata | |
324 | * 2: metadata does not match | |
325 | * 4: already checked | |
326 | */ | |
327 | int used; | |
328 | char *devname; | |
329 | int found; | |
330 | int container; | |
331 | dev_t st_rdev; | |
332 | struct md_list *next; | |
333 | }; | |
334 | ||
e7b84f9d | 335 | #define pr_vrb(fmt, arg...) (void) (verbose && pr_err(fmt, ##arg)) |
ec50f7b6 | 336 | |
1484e727 DW |
337 | static __u8 migr_type(struct imsm_dev *dev) |
338 | { | |
339 | if (dev->vol.migr_type == MIGR_VERIFY && | |
340 | dev->status & DEV_VERIFY_AND_FIX) | |
341 | return MIGR_REPAIR; | |
342 | else | |
343 | return dev->vol.migr_type; | |
344 | } | |
345 | ||
346 | static void set_migr_type(struct imsm_dev *dev, __u8 migr_type) | |
347 | { | |
348 | /* for compatibility with older oroms convert MIGR_REPAIR, into | |
349 | * MIGR_VERIFY w/ DEV_VERIFY_AND_FIX status | |
350 | */ | |
351 | if (migr_type == MIGR_REPAIR) { | |
352 | dev->vol.migr_type = MIGR_VERIFY; | |
353 | dev->status |= DEV_VERIFY_AND_FIX; | |
354 | } else { | |
355 | dev->vol.migr_type = migr_type; | |
356 | dev->status &= ~DEV_VERIFY_AND_FIX; | |
357 | } | |
358 | } | |
359 | ||
f36a9ecd | 360 | static unsigned int sector_count(__u32 bytes, unsigned int sector_size) |
cdddbdbc | 361 | { |
f36a9ecd | 362 | return ROUND_UP(bytes, sector_size) / sector_size; |
87eb16df | 363 | } |
cdddbdbc | 364 | |
f36a9ecd PB |
365 | static unsigned int mpb_sectors(struct imsm_super *mpb, |
366 | unsigned int sector_size) | |
87eb16df | 367 | { |
f36a9ecd | 368 | return sector_count(__le32_to_cpu(mpb->mpb_size), sector_size); |
cdddbdbc DW |
369 | } |
370 | ||
ba2de7ba DW |
371 | struct intel_dev { |
372 | struct imsm_dev *dev; | |
373 | struct intel_dev *next; | |
f21e18ca | 374 | unsigned index; |
ba2de7ba DW |
375 | }; |
376 | ||
88654014 LM |
377 | struct intel_hba { |
378 | enum sys_dev_type type; | |
379 | char *path; | |
380 | char *pci_id; | |
381 | struct intel_hba *next; | |
382 | }; | |
383 | ||
1a64be56 LM |
384 | enum action { |
385 | DISK_REMOVE = 1, | |
386 | DISK_ADD | |
387 | }; | |
cdddbdbc DW |
388 | /* internal representation of IMSM metadata */ |
389 | struct intel_super { | |
390 | union { | |
949c47a0 DW |
391 | void *buf; /* O_DIRECT buffer for reading/writing metadata */ |
392 | struct imsm_super *anchor; /* immovable parameters */ | |
cdddbdbc | 393 | }; |
8e59f3d8 AK |
394 | union { |
395 | void *migr_rec_buf; /* buffer for I/O operations */ | |
396 | struct migr_record *migr_rec; /* migration record */ | |
397 | }; | |
51d83f5d AK |
398 | int clean_migration_record_by_mdmon; /* when reshape is switched to next |
399 | array, it indicates that mdmon is allowed to clean migration | |
400 | record */ | |
949c47a0 | 401 | size_t len; /* size of the 'buf' allocation */ |
bbab0940 | 402 | size_t extra_space; /* extra space in 'buf' that is not used yet */ |
4d7b1503 DW |
403 | void *next_buf; /* for realloc'ing buf from the manager */ |
404 | size_t next_len; | |
c2c087e6 | 405 | int updates_pending; /* count of pending updates for mdmon */ |
bf5a934a | 406 | int current_vol; /* index of raid device undergoing creation */ |
5551b113 | 407 | unsigned long long create_offset; /* common start for 'current_vol' */ |
148acb7b | 408 | __u32 random; /* random data for seeding new family numbers */ |
ba2de7ba | 409 | struct intel_dev *devlist; |
fa7bb6f8 | 410 | unsigned int sector_size; /* sector size of used member drives */ |
cdddbdbc DW |
411 | struct dl { |
412 | struct dl *next; | |
413 | int index; | |
414 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
415 | int major, minor; | |
416 | char *devname; | |
b9f594fe | 417 | struct imsm_disk disk; |
cdddbdbc | 418 | int fd; |
0dcecb2e DW |
419 | int extent_cnt; |
420 | struct extent *e; /* for determining freespace @ create */ | |
efb30e7f | 421 | int raiddisk; /* slot to fill in autolayout */ |
1a64be56 | 422 | enum action action; |
ca0748fa | 423 | } *disks, *current_disk; |
1a64be56 LM |
424 | struct dl *disk_mgmt_list; /* list of disks to add/remove while mdmon |
425 | active */ | |
47ee5a45 | 426 | struct dl *missing; /* disks removed while we weren't looking */ |
43dad3d6 | 427 | struct bbm_log *bbm_log; |
88654014 | 428 | struct intel_hba *hba; /* device path of the raid controller for this metadata */ |
88c32bb1 | 429 | const struct imsm_orom *orom; /* platform firmware support */ |
a2b97981 | 430 | struct intel_super *next; /* (temp) list for disambiguating family_num */ |
928f1424 | 431 | struct md_bb bb; /* memory for get_bad_blocks call */ |
a2b97981 DW |
432 | }; |
433 | ||
434 | struct intel_disk { | |
435 | struct imsm_disk disk; | |
436 | #define IMSM_UNKNOWN_OWNER (-1) | |
437 | int owner; | |
438 | struct intel_disk *next; | |
cdddbdbc DW |
439 | }; |
440 | ||
c2c087e6 DW |
441 | struct extent { |
442 | unsigned long long start, size; | |
443 | }; | |
444 | ||
694575e7 KW |
445 | /* definitions of reshape process types */ |
446 | enum imsm_reshape_type { | |
447 | CH_TAKEOVER, | |
b5347799 | 448 | CH_MIGRATION, |
7abc9871 | 449 | CH_ARRAY_SIZE, |
694575e7 KW |
450 | }; |
451 | ||
88758e9d DW |
452 | /* definition of messages passed to imsm_process_update */ |
453 | enum imsm_update_type { | |
454 | update_activate_spare, | |
8273f55e | 455 | update_create_array, |
33414a01 | 456 | update_kill_array, |
aa534678 | 457 | update_rename_array, |
1a64be56 | 458 | update_add_remove_disk, |
78b10e66 | 459 | update_reshape_container_disks, |
48c5303a | 460 | update_reshape_migration, |
2d40f3a1 AK |
461 | update_takeover, |
462 | update_general_migration_checkpoint, | |
f3871fdc | 463 | update_size_change, |
bbab0940 | 464 | update_prealloc_badblocks_mem, |
e6e9dd3f | 465 | update_rwh_policy, |
88758e9d DW |
466 | }; |
467 | ||
468 | struct imsm_update_activate_spare { | |
469 | enum imsm_update_type type; | |
d23fe947 | 470 | struct dl *dl; |
88758e9d DW |
471 | int slot; |
472 | int array; | |
473 | struct imsm_update_activate_spare *next; | |
474 | }; | |
475 | ||
78b10e66 | 476 | struct geo_params { |
4dd2df09 | 477 | char devnm[32]; |
78b10e66 | 478 | char *dev_name; |
d04f65f4 | 479 | unsigned long long size; |
78b10e66 N |
480 | int level; |
481 | int layout; | |
482 | int chunksize; | |
483 | int raid_disks; | |
484 | }; | |
485 | ||
bb025c2f KW |
486 | enum takeover_direction { |
487 | R10_TO_R0, | |
488 | R0_TO_R10 | |
489 | }; | |
490 | struct imsm_update_takeover { | |
491 | enum imsm_update_type type; | |
492 | int subarray; | |
493 | enum takeover_direction direction; | |
494 | }; | |
78b10e66 N |
495 | |
496 | struct imsm_update_reshape { | |
497 | enum imsm_update_type type; | |
498 | int old_raid_disks; | |
499 | int new_raid_disks; | |
48c5303a PC |
500 | |
501 | int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */ | |
502 | }; | |
503 | ||
504 | struct imsm_update_reshape_migration { | |
505 | enum imsm_update_type type; | |
506 | int old_raid_disks; | |
507 | int new_raid_disks; | |
508 | /* fields for array migration changes | |
509 | */ | |
510 | int subdev; | |
511 | int new_level; | |
512 | int new_layout; | |
4bba0439 | 513 | int new_chunksize; |
48c5303a | 514 | |
d195167d | 515 | int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */ |
78b10e66 N |
516 | }; |
517 | ||
f3871fdc AK |
518 | struct imsm_update_size_change { |
519 | enum imsm_update_type type; | |
520 | int subdev; | |
521 | long long new_size; | |
522 | }; | |
523 | ||
2d40f3a1 AK |
524 | struct imsm_update_general_migration_checkpoint { |
525 | enum imsm_update_type type; | |
526 | __u32 curr_migr_unit; | |
527 | }; | |
528 | ||
54c2c1ea DW |
529 | struct disk_info { |
530 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
531 | }; | |
532 | ||
8273f55e DW |
533 | struct imsm_update_create_array { |
534 | enum imsm_update_type type; | |
8273f55e | 535 | int dev_idx; |
6a3e913e | 536 | struct imsm_dev dev; |
8273f55e DW |
537 | }; |
538 | ||
33414a01 DW |
539 | struct imsm_update_kill_array { |
540 | enum imsm_update_type type; | |
541 | int dev_idx; | |
542 | }; | |
543 | ||
aa534678 DW |
544 | struct imsm_update_rename_array { |
545 | enum imsm_update_type type; | |
546 | __u8 name[MAX_RAID_SERIAL_LEN]; | |
547 | int dev_idx; | |
548 | }; | |
549 | ||
1a64be56 | 550 | struct imsm_update_add_remove_disk { |
43dad3d6 DW |
551 | enum imsm_update_type type; |
552 | }; | |
553 | ||
bbab0940 TM |
554 | struct imsm_update_prealloc_bb_mem { |
555 | enum imsm_update_type type; | |
556 | }; | |
557 | ||
e6e9dd3f AP |
558 | struct imsm_update_rwh_policy { |
559 | enum imsm_update_type type; | |
560 | int new_policy; | |
561 | int dev_idx; | |
562 | }; | |
563 | ||
88654014 LM |
564 | static const char *_sys_dev_type[] = { |
565 | [SYS_DEV_UNKNOWN] = "Unknown", | |
566 | [SYS_DEV_SAS] = "SAS", | |
614902f6 | 567 | [SYS_DEV_SATA] = "SATA", |
60f0f54d PB |
568 | [SYS_DEV_NVME] = "NVMe", |
569 | [SYS_DEV_VMD] = "VMD" | |
88654014 LM |
570 | }; |
571 | ||
572 | const char *get_sys_dev_type(enum sys_dev_type type) | |
573 | { | |
574 | if (type >= SYS_DEV_MAX) | |
575 | type = SYS_DEV_UNKNOWN; | |
576 | ||
577 | return _sys_dev_type[type]; | |
578 | } | |
579 | ||
580 | static struct intel_hba * alloc_intel_hba(struct sys_dev *device) | |
581 | { | |
503975b9 N |
582 | struct intel_hba *result = xmalloc(sizeof(*result)); |
583 | ||
584 | result->type = device->type; | |
585 | result->path = xstrdup(device->path); | |
586 | result->next = NULL; | |
587 | if (result->path && (result->pci_id = strrchr(result->path, '/')) != NULL) | |
588 | result->pci_id++; | |
589 | ||
88654014 LM |
590 | return result; |
591 | } | |
592 | ||
593 | static struct intel_hba * find_intel_hba(struct intel_hba *hba, struct sys_dev *device) | |
594 | { | |
594dc1b8 JS |
595 | struct intel_hba *result; |
596 | ||
88654014 LM |
597 | for (result = hba; result; result = result->next) { |
598 | if (result->type == device->type && strcmp(result->path, device->path) == 0) | |
599 | break; | |
600 | } | |
601 | return result; | |
602 | } | |
603 | ||
b4cf4cba | 604 | static int attach_hba_to_super(struct intel_super *super, struct sys_dev *device) |
88654014 LM |
605 | { |
606 | struct intel_hba *hba; | |
607 | ||
608 | /* check if disk attached to Intel HBA */ | |
609 | hba = find_intel_hba(super->hba, device); | |
610 | if (hba != NULL) | |
611 | return 1; | |
612 | /* Check if HBA is already attached to super */ | |
613 | if (super->hba == NULL) { | |
614 | super->hba = alloc_intel_hba(device); | |
615 | return 1; | |
6b781d33 AP |
616 | } |
617 | ||
618 | hba = super->hba; | |
619 | /* Intel metadata allows for all disks attached to the same type HBA. | |
614902f6 | 620 | * Do not support HBA types mixing |
6b781d33 AP |
621 | */ |
622 | if (device->type != hba->type) | |
88654014 | 623 | return 2; |
6b781d33 AP |
624 | |
625 | /* Multiple same type HBAs can be used if they share the same OROM */ | |
626 | const struct imsm_orom *device_orom = get_orom_by_device_id(device->dev_id); | |
627 | ||
628 | if (device_orom != super->orom) | |
629 | return 2; | |
630 | ||
631 | while (hba->next) | |
632 | hba = hba->next; | |
633 | ||
634 | hba->next = alloc_intel_hba(device); | |
635 | return 1; | |
88654014 LM |
636 | } |
637 | ||
638 | static struct sys_dev* find_disk_attached_hba(int fd, const char *devname) | |
639 | { | |
9bc4ae77 | 640 | struct sys_dev *list, *elem; |
88654014 LM |
641 | char *disk_path; |
642 | ||
643 | if ((list = find_intel_devices()) == NULL) | |
644 | return 0; | |
645 | ||
646 | if (fd < 0) | |
647 | disk_path = (char *) devname; | |
648 | else | |
649 | disk_path = diskfd_to_devpath(fd); | |
650 | ||
9bc4ae77 | 651 | if (!disk_path) |
88654014 | 652 | return 0; |
88654014 | 653 | |
9bc4ae77 N |
654 | for (elem = list; elem; elem = elem->next) |
655 | if (path_attached_to_hba(disk_path, elem->path)) | |
88654014 | 656 | return elem; |
9bc4ae77 | 657 | |
88654014 LM |
658 | if (disk_path != devname) |
659 | free(disk_path); | |
88654014 LM |
660 | |
661 | return NULL; | |
662 | } | |
663 | ||
d424212e N |
664 | static int find_intel_hba_capability(int fd, struct intel_super *super, |
665 | char *devname); | |
f2f5c343 | 666 | |
cdddbdbc DW |
667 | static struct supertype *match_metadata_desc_imsm(char *arg) |
668 | { | |
669 | struct supertype *st; | |
670 | ||
671 | if (strcmp(arg, "imsm") != 0 && | |
672 | strcmp(arg, "default") != 0 | |
673 | ) | |
674 | return NULL; | |
675 | ||
503975b9 | 676 | st = xcalloc(1, sizeof(*st)); |
cdddbdbc DW |
677 | st->ss = &super_imsm; |
678 | st->max_devs = IMSM_MAX_DEVICES; | |
679 | st->minor_version = 0; | |
680 | st->sb = NULL; | |
681 | return st; | |
682 | } | |
683 | ||
cdddbdbc DW |
684 | static __u8 *get_imsm_version(struct imsm_super *mpb) |
685 | { | |
686 | return &mpb->sig[MPB_SIG_LEN]; | |
687 | } | |
688 | ||
949c47a0 DW |
689 | /* retrieve a disk directly from the anchor when the anchor is known to be |
690 | * up-to-date, currently only at load time | |
691 | */ | |
692 | static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index) | |
cdddbdbc | 693 | { |
949c47a0 | 694 | if (index >= mpb->num_disks) |
cdddbdbc DW |
695 | return NULL; |
696 | return &mpb->disk[index]; | |
697 | } | |
698 | ||
95d07a2c LM |
699 | /* retrieve the disk description based on a index of the disk |
700 | * in the sub-array | |
701 | */ | |
702 | static struct dl *get_imsm_dl_disk(struct intel_super *super, __u8 index) | |
949c47a0 | 703 | { |
b9f594fe DW |
704 | struct dl *d; |
705 | ||
706 | for (d = super->disks; d; d = d->next) | |
707 | if (d->index == index) | |
95d07a2c LM |
708 | return d; |
709 | ||
710 | return NULL; | |
711 | } | |
712 | /* retrieve a disk from the parsed metadata */ | |
713 | static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index) | |
714 | { | |
715 | struct dl *dl; | |
716 | ||
717 | dl = get_imsm_dl_disk(super, index); | |
718 | if (dl) | |
719 | return &dl->disk; | |
720 | ||
b9f594fe | 721 | return NULL; |
949c47a0 DW |
722 | } |
723 | ||
724 | /* generate a checksum directly from the anchor when the anchor is known to be | |
725 | * up-to-date, currently only at load or write_super after coalescing | |
726 | */ | |
727 | static __u32 __gen_imsm_checksum(struct imsm_super *mpb) | |
cdddbdbc DW |
728 | { |
729 | __u32 end = mpb->mpb_size / sizeof(end); | |
730 | __u32 *p = (__u32 *) mpb; | |
731 | __u32 sum = 0; | |
732 | ||
5d500228 N |
733 | while (end--) { |
734 | sum += __le32_to_cpu(*p); | |
97f734fd N |
735 | p++; |
736 | } | |
cdddbdbc | 737 | |
5d500228 | 738 | return sum - __le32_to_cpu(mpb->check_sum); |
cdddbdbc DW |
739 | } |
740 | ||
a965f303 DW |
741 | static size_t sizeof_imsm_map(struct imsm_map *map) |
742 | { | |
743 | return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1); | |
744 | } | |
745 | ||
746 | struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map) | |
cdddbdbc | 747 | { |
5e7b0330 AK |
748 | /* A device can have 2 maps if it is in the middle of a migration. |
749 | * If second_map is: | |
238c0a71 AK |
750 | * MAP_0 - we return the first map |
751 | * MAP_1 - we return the second map if it exists, else NULL | |
752 | * MAP_X - we return the second map if it exists, else the first | |
5e7b0330 | 753 | */ |
a965f303 | 754 | struct imsm_map *map = &dev->vol.map[0]; |
9535fc47 | 755 | struct imsm_map *map2 = NULL; |
a965f303 | 756 | |
9535fc47 AK |
757 | if (dev->vol.migr_state) |
758 | map2 = (void *)map + sizeof_imsm_map(map); | |
a965f303 | 759 | |
9535fc47 | 760 | switch (second_map) { |
3b451610 | 761 | case MAP_0: |
9535fc47 | 762 | break; |
3b451610 | 763 | case MAP_1: |
9535fc47 AK |
764 | map = map2; |
765 | break; | |
238c0a71 | 766 | case MAP_X: |
9535fc47 AK |
767 | if (map2) |
768 | map = map2; | |
769 | break; | |
9535fc47 AK |
770 | default: |
771 | map = NULL; | |
772 | } | |
773 | return map; | |
5e7b0330 | 774 | |
a965f303 | 775 | } |
cdddbdbc | 776 | |
3393c6af DW |
777 | /* return the size of the device. |
778 | * migr_state increases the returned size if map[0] were to be duplicated | |
779 | */ | |
780 | static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state) | |
a965f303 DW |
781 | { |
782 | size_t size = sizeof(*dev) - sizeof(struct imsm_map) + | |
238c0a71 | 783 | sizeof_imsm_map(get_imsm_map(dev, MAP_0)); |
cdddbdbc DW |
784 | |
785 | /* migrating means an additional map */ | |
a965f303 | 786 | if (dev->vol.migr_state) |
238c0a71 | 787 | size += sizeof_imsm_map(get_imsm_map(dev, MAP_1)); |
3393c6af | 788 | else if (migr_state) |
238c0a71 | 789 | size += sizeof_imsm_map(get_imsm_map(dev, MAP_0)); |
cdddbdbc DW |
790 | |
791 | return size; | |
792 | } | |
793 | ||
54c2c1ea DW |
794 | /* retrieve disk serial number list from a metadata update */ |
795 | static struct disk_info *get_disk_info(struct imsm_update_create_array *update) | |
796 | { | |
797 | void *u = update; | |
798 | struct disk_info *inf; | |
799 | ||
800 | inf = u + sizeof(*update) - sizeof(struct imsm_dev) + | |
801 | sizeof_imsm_dev(&update->dev, 0); | |
802 | ||
803 | return inf; | |
804 | } | |
54c2c1ea | 805 | |
949c47a0 | 806 | static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index) |
cdddbdbc DW |
807 | { |
808 | int offset; | |
809 | int i; | |
810 | void *_mpb = mpb; | |
811 | ||
949c47a0 | 812 | if (index >= mpb->num_raid_devs) |
cdddbdbc DW |
813 | return NULL; |
814 | ||
815 | /* devices start after all disks */ | |
816 | offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb; | |
817 | ||
818 | for (i = 0; i <= index; i++) | |
819 | if (i == index) | |
820 | return _mpb + offset; | |
821 | else | |
3393c6af | 822 | offset += sizeof_imsm_dev(_mpb + offset, 0); |
cdddbdbc DW |
823 | |
824 | return NULL; | |
825 | } | |
826 | ||
949c47a0 DW |
827 | static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index) |
828 | { | |
ba2de7ba DW |
829 | struct intel_dev *dv; |
830 | ||
949c47a0 DW |
831 | if (index >= super->anchor->num_raid_devs) |
832 | return NULL; | |
ba2de7ba DW |
833 | for (dv = super->devlist; dv; dv = dv->next) |
834 | if (dv->index == index) | |
835 | return dv->dev; | |
836 | return NULL; | |
949c47a0 DW |
837 | } |
838 | ||
8d67477f TM |
839 | static inline unsigned long long __le48_to_cpu(const struct bbm_log_block_addr |
840 | *addr) | |
841 | { | |
842 | return ((((__u64)__le32_to_cpu(addr->dw1)) << 16) | | |
843 | __le16_to_cpu(addr->w1)); | |
844 | } | |
845 | ||
846 | static inline struct bbm_log_block_addr __cpu_to_le48(unsigned long long sec) | |
847 | { | |
848 | struct bbm_log_block_addr addr; | |
849 | ||
850 | addr.w1 = __cpu_to_le16((__u16)(sec & 0xffff)); | |
851 | addr.dw1 = __cpu_to_le32((__u32)(sec >> 16) & 0xffffffff); | |
852 | return addr; | |
853 | } | |
854 | ||
8d67477f TM |
855 | /* get size of the bbm log */ |
856 | static __u32 get_imsm_bbm_log_size(struct bbm_log *log) | |
857 | { | |
858 | if (!log || log->entry_count == 0) | |
859 | return 0; | |
860 | ||
861 | return sizeof(log->signature) + | |
862 | sizeof(log->entry_count) + | |
863 | log->entry_count * sizeof(struct bbm_log_entry); | |
864 | } | |
6f50473f TM |
865 | |
866 | /* check if bad block is not partially stored in bbm log */ | |
867 | static int is_stored_in_bbm(struct bbm_log *log, const __u8 idx, const unsigned | |
868 | long long sector, const int length, __u32 *pos) | |
869 | { | |
870 | __u32 i; | |
871 | ||
872 | for (i = *pos; i < log->entry_count; i++) { | |
873 | struct bbm_log_entry *entry = &log->marked_block_entries[i]; | |
874 | unsigned long long bb_start; | |
875 | unsigned long long bb_end; | |
876 | ||
877 | bb_start = __le48_to_cpu(&entry->defective_block_start); | |
878 | bb_end = bb_start + (entry->marked_count + 1); | |
879 | ||
880 | if ((entry->disk_ordinal == idx) && (bb_start >= sector) && | |
881 | (bb_end <= sector + length)) { | |
882 | *pos = i; | |
883 | return 1; | |
884 | } | |
885 | } | |
886 | return 0; | |
887 | } | |
888 | ||
889 | /* record new bad block in bbm log */ | |
890 | static int record_new_badblock(struct bbm_log *log, const __u8 idx, unsigned | |
891 | long long sector, int length) | |
892 | { | |
893 | int new_bb = 0; | |
894 | __u32 pos = 0; | |
895 | struct bbm_log_entry *entry = NULL; | |
896 | ||
897 | while (is_stored_in_bbm(log, idx, sector, length, &pos)) { | |
898 | struct bbm_log_entry *e = &log->marked_block_entries[pos]; | |
899 | ||
900 | if ((e->marked_count + 1 == BBM_LOG_MAX_LBA_ENTRY_VAL) && | |
901 | (__le48_to_cpu(&e->defective_block_start) == sector)) { | |
902 | sector += BBM_LOG_MAX_LBA_ENTRY_VAL; | |
903 | length -= BBM_LOG_MAX_LBA_ENTRY_VAL; | |
904 | pos = pos + 1; | |
905 | continue; | |
906 | } | |
907 | entry = e; | |
908 | break; | |
909 | } | |
910 | ||
911 | if (entry) { | |
912 | int cnt = (length <= BBM_LOG_MAX_LBA_ENTRY_VAL) ? length : | |
913 | BBM_LOG_MAX_LBA_ENTRY_VAL; | |
914 | entry->defective_block_start = __cpu_to_le48(sector); | |
915 | entry->marked_count = cnt - 1; | |
916 | if (cnt == length) | |
917 | return 1; | |
918 | sector += cnt; | |
919 | length -= cnt; | |
920 | } | |
921 | ||
922 | new_bb = ROUND_UP(length, BBM_LOG_MAX_LBA_ENTRY_VAL) / | |
923 | BBM_LOG_MAX_LBA_ENTRY_VAL; | |
924 | if (log->entry_count + new_bb > BBM_LOG_MAX_ENTRIES) | |
925 | return 0; | |
926 | ||
927 | while (length > 0) { | |
928 | int cnt = (length <= BBM_LOG_MAX_LBA_ENTRY_VAL) ? length : | |
929 | BBM_LOG_MAX_LBA_ENTRY_VAL; | |
930 | struct bbm_log_entry *entry = | |
931 | &log->marked_block_entries[log->entry_count]; | |
932 | ||
933 | entry->defective_block_start = __cpu_to_le48(sector); | |
934 | entry->marked_count = cnt - 1; | |
935 | entry->disk_ordinal = idx; | |
936 | ||
937 | sector += cnt; | |
938 | length -= cnt; | |
939 | ||
940 | log->entry_count++; | |
941 | } | |
942 | ||
943 | return new_bb; | |
944 | } | |
c07a5a4f | 945 | |
4c9e8c1e TM |
946 | /* clear all bad blocks for given disk */ |
947 | static void clear_disk_badblocks(struct bbm_log *log, const __u8 idx) | |
948 | { | |
949 | __u32 i = 0; | |
950 | ||
951 | while (i < log->entry_count) { | |
952 | struct bbm_log_entry *entries = log->marked_block_entries; | |
953 | ||
954 | if (entries[i].disk_ordinal == idx) { | |
955 | if (i < log->entry_count - 1) | |
956 | entries[i] = entries[log->entry_count - 1]; | |
957 | log->entry_count--; | |
958 | } else { | |
959 | i++; | |
960 | } | |
961 | } | |
962 | } | |
963 | ||
c07a5a4f TM |
964 | /* clear given bad block */ |
965 | static int clear_badblock(struct bbm_log *log, const __u8 idx, const unsigned | |
966 | long long sector, const int length) { | |
967 | __u32 i = 0; | |
968 | ||
969 | while (i < log->entry_count) { | |
970 | struct bbm_log_entry *entries = log->marked_block_entries; | |
971 | ||
972 | if ((entries[i].disk_ordinal == idx) && | |
973 | (__le48_to_cpu(&entries[i].defective_block_start) == | |
974 | sector) && (entries[i].marked_count + 1 == length)) { | |
975 | if (i < log->entry_count - 1) | |
976 | entries[i] = entries[log->entry_count - 1]; | |
977 | log->entry_count--; | |
978 | break; | |
979 | } | |
980 | i++; | |
981 | } | |
982 | ||
983 | return 1; | |
984 | } | |
8d67477f TM |
985 | |
986 | /* allocate and load BBM log from metadata */ | |
987 | static int load_bbm_log(struct intel_super *super) | |
988 | { | |
989 | struct imsm_super *mpb = super->anchor; | |
990 | __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size); | |
991 | ||
992 | super->bbm_log = xcalloc(1, sizeof(struct bbm_log)); | |
993 | if (!super->bbm_log) | |
994 | return 1; | |
995 | ||
996 | if (bbm_log_size) { | |
997 | struct bbm_log *log = (void *)mpb + | |
998 | __le32_to_cpu(mpb->mpb_size) - bbm_log_size; | |
999 | ||
1000 | __u32 entry_count; | |
1001 | ||
1002 | if (bbm_log_size < sizeof(log->signature) + | |
1003 | sizeof(log->entry_count)) | |
1004 | return 2; | |
1005 | ||
1006 | entry_count = __le32_to_cpu(log->entry_count); | |
1007 | if ((__le32_to_cpu(log->signature) != BBM_LOG_SIGNATURE) || | |
1008 | (entry_count > BBM_LOG_MAX_ENTRIES)) | |
1009 | return 3; | |
1010 | ||
1011 | if (bbm_log_size != | |
1012 | sizeof(log->signature) + sizeof(log->entry_count) + | |
1013 | entry_count * sizeof(struct bbm_log_entry)) | |
1014 | return 4; | |
1015 | ||
1016 | memcpy(super->bbm_log, log, bbm_log_size); | |
1017 | } else { | |
1018 | super->bbm_log->signature = __cpu_to_le32(BBM_LOG_SIGNATURE); | |
1019 | super->bbm_log->entry_count = 0; | |
1020 | } | |
1021 | ||
1022 | return 0; | |
1023 | } | |
1024 | ||
b12796be TM |
1025 | /* checks if bad block is within volume boundaries */ |
1026 | static int is_bad_block_in_volume(const struct bbm_log_entry *entry, | |
1027 | const unsigned long long start_sector, | |
1028 | const unsigned long long size) | |
1029 | { | |
1030 | unsigned long long bb_start; | |
1031 | unsigned long long bb_end; | |
1032 | ||
1033 | bb_start = __le48_to_cpu(&entry->defective_block_start); | |
1034 | bb_end = bb_start + (entry->marked_count + 1); | |
1035 | ||
1036 | if (((bb_start >= start_sector) && (bb_start < start_sector + size)) || | |
1037 | ((bb_end >= start_sector) && (bb_end <= start_sector + size))) | |
1038 | return 1; | |
1039 | ||
1040 | return 0; | |
1041 | } | |
1042 | ||
1043 | /* get list of bad blocks on a drive for a volume */ | |
1044 | static void get_volume_badblocks(const struct bbm_log *log, const __u8 idx, | |
1045 | const unsigned long long start_sector, | |
1046 | const unsigned long long size, | |
1047 | struct md_bb *bbs) | |
1048 | { | |
1049 | __u32 count = 0; | |
1050 | __u32 i; | |
1051 | ||
1052 | for (i = 0; i < log->entry_count; i++) { | |
1053 | const struct bbm_log_entry *ent = | |
1054 | &log->marked_block_entries[i]; | |
1055 | struct md_bb_entry *bb; | |
1056 | ||
1057 | if ((ent->disk_ordinal == idx) && | |
1058 | is_bad_block_in_volume(ent, start_sector, size)) { | |
1059 | ||
1060 | if (!bbs->entries) { | |
1061 | bbs->entries = xmalloc(BBM_LOG_MAX_ENTRIES * | |
1062 | sizeof(*bb)); | |
1063 | if (!bbs->entries) | |
1064 | break; | |
1065 | } | |
1066 | ||
1067 | bb = &bbs->entries[count++]; | |
1068 | bb->sector = __le48_to_cpu(&ent->defective_block_start); | |
1069 | bb->length = ent->marked_count + 1; | |
1070 | } | |
1071 | } | |
1072 | bbs->count = count; | |
1073 | } | |
1074 | ||
98130f40 AK |
1075 | /* |
1076 | * for second_map: | |
238c0a71 AK |
1077 | * == MAP_0 get first map |
1078 | * == MAP_1 get second map | |
1079 | * == MAP_X than get map according to the current migr_state | |
98130f40 AK |
1080 | */ |
1081 | static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev, | |
1082 | int slot, | |
1083 | int second_map) | |
7eef0453 DW |
1084 | { |
1085 | struct imsm_map *map; | |
1086 | ||
5e7b0330 | 1087 | map = get_imsm_map(dev, second_map); |
7eef0453 | 1088 | |
ff077194 DW |
1089 | /* top byte identifies disk under rebuild */ |
1090 | return __le32_to_cpu(map->disk_ord_tbl[slot]); | |
1091 | } | |
1092 | ||
1093 | #define ord_to_idx(ord) (((ord) << 8) >> 8) | |
98130f40 | 1094 | static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot, int second_map) |
ff077194 | 1095 | { |
98130f40 | 1096 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, second_map); |
ff077194 DW |
1097 | |
1098 | return ord_to_idx(ord); | |
7eef0453 DW |
1099 | } |
1100 | ||
be73972f DW |
1101 | static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord) |
1102 | { | |
1103 | map->disk_ord_tbl[slot] = __cpu_to_le32(ord); | |
1104 | } | |
1105 | ||
f21e18ca | 1106 | static int get_imsm_disk_slot(struct imsm_map *map, unsigned idx) |
620b1713 DW |
1107 | { |
1108 | int slot; | |
1109 | __u32 ord; | |
1110 | ||
1111 | for (slot = 0; slot < map->num_members; slot++) { | |
1112 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
1113 | if (ord_to_idx(ord) == idx) | |
1114 | return slot; | |
1115 | } | |
1116 | ||
1117 | return -1; | |
1118 | } | |
1119 | ||
cdddbdbc DW |
1120 | static int get_imsm_raid_level(struct imsm_map *map) |
1121 | { | |
1122 | if (map->raid_level == 1) { | |
1123 | if (map->num_members == 2) | |
1124 | return 1; | |
1125 | else | |
1126 | return 10; | |
1127 | } | |
1128 | ||
1129 | return map->raid_level; | |
1130 | } | |
1131 | ||
c2c087e6 DW |
1132 | static int cmp_extent(const void *av, const void *bv) |
1133 | { | |
1134 | const struct extent *a = av; | |
1135 | const struct extent *b = bv; | |
1136 | if (a->start < b->start) | |
1137 | return -1; | |
1138 | if (a->start > b->start) | |
1139 | return 1; | |
1140 | return 0; | |
1141 | } | |
1142 | ||
0dcecb2e | 1143 | static int count_memberships(struct dl *dl, struct intel_super *super) |
c2c087e6 | 1144 | { |
c2c087e6 | 1145 | int memberships = 0; |
620b1713 | 1146 | int i; |
c2c087e6 | 1147 | |
949c47a0 DW |
1148 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
1149 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 1150 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
c2c087e6 | 1151 | |
620b1713 DW |
1152 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
1153 | memberships++; | |
c2c087e6 | 1154 | } |
0dcecb2e DW |
1155 | |
1156 | return memberships; | |
1157 | } | |
1158 | ||
b81221b7 CA |
1159 | static __u32 imsm_min_reserved_sectors(struct intel_super *super); |
1160 | ||
5551b113 CA |
1161 | static int split_ull(unsigned long long n, __u32 *lo, __u32 *hi) |
1162 | { | |
1163 | if (lo == 0 || hi == 0) | |
1164 | return 1; | |
1165 | *lo = __le32_to_cpu((unsigned)n); | |
1166 | *hi = __le32_to_cpu((unsigned)(n >> 32)); | |
1167 | return 0; | |
1168 | } | |
1169 | ||
1170 | static unsigned long long join_u32(__u32 lo, __u32 hi) | |
1171 | { | |
1172 | return (unsigned long long)__le32_to_cpu(lo) | | |
1173 | (((unsigned long long)__le32_to_cpu(hi)) << 32); | |
1174 | } | |
1175 | ||
1176 | static unsigned long long total_blocks(struct imsm_disk *disk) | |
1177 | { | |
1178 | if (disk == NULL) | |
1179 | return 0; | |
1180 | return join_u32(disk->total_blocks_lo, disk->total_blocks_hi); | |
1181 | } | |
1182 | ||
1183 | static unsigned long long pba_of_lba0(struct imsm_map *map) | |
1184 | { | |
1185 | if (map == NULL) | |
1186 | return 0; | |
1187 | return join_u32(map->pba_of_lba0_lo, map->pba_of_lba0_hi); | |
1188 | } | |
1189 | ||
1190 | static unsigned long long blocks_per_member(struct imsm_map *map) | |
1191 | { | |
1192 | if (map == NULL) | |
1193 | return 0; | |
1194 | return join_u32(map->blocks_per_member_lo, map->blocks_per_member_hi); | |
1195 | } | |
1196 | ||
1197 | static unsigned long long num_data_stripes(struct imsm_map *map) | |
1198 | { | |
1199 | if (map == NULL) | |
1200 | return 0; | |
1201 | return join_u32(map->num_data_stripes_lo, map->num_data_stripes_hi); | |
1202 | } | |
1203 | ||
1204 | static void set_total_blocks(struct imsm_disk *disk, unsigned long long n) | |
1205 | { | |
1206 | split_ull(n, &disk->total_blocks_lo, &disk->total_blocks_hi); | |
1207 | } | |
1208 | ||
1209 | static void set_pba_of_lba0(struct imsm_map *map, unsigned long long n) | |
1210 | { | |
1211 | split_ull(n, &map->pba_of_lba0_lo, &map->pba_of_lba0_hi); | |
1212 | } | |
1213 | ||
1214 | static void set_blocks_per_member(struct imsm_map *map, unsigned long long n) | |
1215 | { | |
1216 | split_ull(n, &map->blocks_per_member_lo, &map->blocks_per_member_hi); | |
1217 | } | |
1218 | ||
1219 | static void set_num_data_stripes(struct imsm_map *map, unsigned long long n) | |
1220 | { | |
1221 | split_ull(n, &map->num_data_stripes_lo, &map->num_data_stripes_hi); | |
1222 | } | |
1223 | ||
0dcecb2e DW |
1224 | static struct extent *get_extents(struct intel_super *super, struct dl *dl) |
1225 | { | |
1226 | /* find a list of used extents on the given physical device */ | |
1227 | struct extent *rv, *e; | |
620b1713 | 1228 | int i; |
0dcecb2e | 1229 | int memberships = count_memberships(dl, super); |
b276dd33 DW |
1230 | __u32 reservation; |
1231 | ||
1232 | /* trim the reserved area for spares, so they can join any array | |
1233 | * regardless of whether the OROM has assigned sectors from the | |
1234 | * IMSM_RESERVED_SECTORS region | |
1235 | */ | |
1236 | if (dl->index == -1) | |
b81221b7 | 1237 | reservation = imsm_min_reserved_sectors(super); |
b276dd33 DW |
1238 | else |
1239 | reservation = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
0dcecb2e | 1240 | |
503975b9 | 1241 | rv = xcalloc(sizeof(struct extent), (memberships + 1)); |
c2c087e6 DW |
1242 | e = rv; |
1243 | ||
949c47a0 DW |
1244 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
1245 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 1246 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
c2c087e6 | 1247 | |
620b1713 | 1248 | if (get_imsm_disk_slot(map, dl->index) >= 0) { |
5551b113 CA |
1249 | e->start = pba_of_lba0(map); |
1250 | e->size = blocks_per_member(map); | |
620b1713 | 1251 | e++; |
c2c087e6 DW |
1252 | } |
1253 | } | |
1254 | qsort(rv, memberships, sizeof(*rv), cmp_extent); | |
1255 | ||
1011e834 | 1256 | /* determine the start of the metadata |
14e8215b DW |
1257 | * when no raid devices are defined use the default |
1258 | * ...otherwise allow the metadata to truncate the value | |
1259 | * as is the case with older versions of imsm | |
1260 | */ | |
1261 | if (memberships) { | |
1262 | struct extent *last = &rv[memberships - 1]; | |
5551b113 | 1263 | unsigned long long remainder; |
14e8215b | 1264 | |
5551b113 | 1265 | remainder = total_blocks(&dl->disk) - (last->start + last->size); |
dda5855f DW |
1266 | /* round down to 1k block to satisfy precision of the kernel |
1267 | * 'size' interface | |
1268 | */ | |
1269 | remainder &= ~1UL; | |
1270 | /* make sure remainder is still sane */ | |
f21e18ca | 1271 | if (remainder < (unsigned)ROUND_UP(super->len, 512) >> 9) |
dda5855f | 1272 | remainder = ROUND_UP(super->len, 512) >> 9; |
14e8215b DW |
1273 | if (reservation > remainder) |
1274 | reservation = remainder; | |
1275 | } | |
5551b113 | 1276 | e->start = total_blocks(&dl->disk) - reservation; |
c2c087e6 DW |
1277 | e->size = 0; |
1278 | return rv; | |
1279 | } | |
1280 | ||
14e8215b DW |
1281 | /* try to determine how much space is reserved for metadata from |
1282 | * the last get_extents() entry, otherwise fallback to the | |
1283 | * default | |
1284 | */ | |
1285 | static __u32 imsm_reserved_sectors(struct intel_super *super, struct dl *dl) | |
1286 | { | |
1287 | struct extent *e; | |
1288 | int i; | |
1289 | __u32 rv; | |
1290 | ||
1291 | /* for spares just return a minimal reservation which will grow | |
1292 | * once the spare is picked up by an array | |
1293 | */ | |
1294 | if (dl->index == -1) | |
1295 | return MPB_SECTOR_CNT; | |
1296 | ||
1297 | e = get_extents(super, dl); | |
1298 | if (!e) | |
1299 | return MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
1300 | ||
1301 | /* scroll to last entry */ | |
1302 | for (i = 0; e[i].size; i++) | |
1303 | continue; | |
1304 | ||
5551b113 | 1305 | rv = total_blocks(&dl->disk) - e[i].start; |
14e8215b DW |
1306 | |
1307 | free(e); | |
1308 | ||
1309 | return rv; | |
1310 | } | |
1311 | ||
25ed7e59 DW |
1312 | static int is_spare(struct imsm_disk *disk) |
1313 | { | |
1314 | return (disk->status & SPARE_DISK) == SPARE_DISK; | |
1315 | } | |
1316 | ||
1317 | static int is_configured(struct imsm_disk *disk) | |
1318 | { | |
1319 | return (disk->status & CONFIGURED_DISK) == CONFIGURED_DISK; | |
1320 | } | |
1321 | ||
1322 | static int is_failed(struct imsm_disk *disk) | |
1323 | { | |
1324 | return (disk->status & FAILED_DISK) == FAILED_DISK; | |
1325 | } | |
1326 | ||
2432ce9b AP |
1327 | static int is_journal(struct imsm_disk *disk) |
1328 | { | |
1329 | return (disk->status & JOURNAL_DISK) == JOURNAL_DISK; | |
1330 | } | |
1331 | ||
b53bfba6 TM |
1332 | /* round array size down to closest MB and ensure it splits evenly |
1333 | * between members | |
1334 | */ | |
1335 | static unsigned long long round_size_to_mb(unsigned long long size, unsigned int | |
1336 | disk_count) | |
1337 | { | |
1338 | size /= disk_count; | |
1339 | size = (size >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT; | |
1340 | size *= disk_count; | |
1341 | ||
1342 | return size; | |
1343 | } | |
1344 | ||
b81221b7 CA |
1345 | /* try to determine how much space is reserved for metadata from |
1346 | * the last get_extents() entry on the smallest active disk, | |
1347 | * otherwise fallback to the default | |
1348 | */ | |
1349 | static __u32 imsm_min_reserved_sectors(struct intel_super *super) | |
1350 | { | |
1351 | struct extent *e; | |
1352 | int i; | |
5551b113 CA |
1353 | unsigned long long min_active; |
1354 | __u32 remainder; | |
b81221b7 CA |
1355 | __u32 rv = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; |
1356 | struct dl *dl, *dl_min = NULL; | |
1357 | ||
1358 | if (!super) | |
1359 | return rv; | |
1360 | ||
1361 | min_active = 0; | |
1362 | for (dl = super->disks; dl; dl = dl->next) { | |
1363 | if (dl->index < 0) | |
1364 | continue; | |
5551b113 CA |
1365 | unsigned long long blocks = total_blocks(&dl->disk); |
1366 | if (blocks < min_active || min_active == 0) { | |
b81221b7 | 1367 | dl_min = dl; |
5551b113 | 1368 | min_active = blocks; |
b81221b7 CA |
1369 | } |
1370 | } | |
1371 | if (!dl_min) | |
1372 | return rv; | |
1373 | ||
1374 | /* find last lba used by subarrays on the smallest active disk */ | |
1375 | e = get_extents(super, dl_min); | |
1376 | if (!e) | |
1377 | return rv; | |
1378 | for (i = 0; e[i].size; i++) | |
1379 | continue; | |
1380 | ||
1381 | remainder = min_active - e[i].start; | |
1382 | free(e); | |
1383 | ||
1384 | /* to give priority to recovery we should not require full | |
1385 | IMSM_RESERVED_SECTORS from the spare */ | |
1386 | rv = MPB_SECTOR_CNT + NUM_BLOCKS_DIRTY_STRIPE_REGION; | |
1387 | ||
1388 | /* if real reservation is smaller use that value */ | |
1389 | return (remainder < rv) ? remainder : rv; | |
1390 | } | |
1391 | ||
fbfdcb06 AO |
1392 | /* |
1393 | * Return minimum size of a spare and sector size | |
1394 | * that can be used in this array | |
1395 | */ | |
1396 | int get_spare_criteria_imsm(struct supertype *st, struct spare_criteria *c) | |
80e7f8c3 AC |
1397 | { |
1398 | struct intel_super *super = st->sb; | |
1399 | struct dl *dl; | |
1400 | struct extent *e; | |
1401 | int i; | |
fbfdcb06 AO |
1402 | unsigned long long size = 0; |
1403 | ||
1404 | c->min_size = 0; | |
4b57ecf6 | 1405 | c->sector_size = 0; |
80e7f8c3 AC |
1406 | |
1407 | if (!super) | |
fbfdcb06 | 1408 | return -EINVAL; |
80e7f8c3 AC |
1409 | /* find first active disk in array */ |
1410 | dl = super->disks; | |
1411 | while (dl && (is_failed(&dl->disk) || dl->index == -1)) | |
1412 | dl = dl->next; | |
1413 | if (!dl) | |
fbfdcb06 | 1414 | return -EINVAL; |
80e7f8c3 AC |
1415 | /* find last lba used by subarrays */ |
1416 | e = get_extents(super, dl); | |
1417 | if (!e) | |
fbfdcb06 | 1418 | return -EINVAL; |
80e7f8c3 AC |
1419 | for (i = 0; e[i].size; i++) |
1420 | continue; | |
1421 | if (i > 0) | |
fbfdcb06 | 1422 | size = e[i-1].start + e[i-1].size; |
80e7f8c3 | 1423 | free(e); |
b81221b7 | 1424 | |
80e7f8c3 | 1425 | /* add the amount of space needed for metadata */ |
fbfdcb06 AO |
1426 | size += imsm_min_reserved_sectors(super); |
1427 | ||
1428 | c->min_size = size * 512; | |
4b57ecf6 | 1429 | c->sector_size = super->sector_size; |
b81221b7 | 1430 | |
fbfdcb06 | 1431 | return 0; |
80e7f8c3 AC |
1432 | } |
1433 | ||
d1e02575 AK |
1434 | static int is_gen_migration(struct imsm_dev *dev); |
1435 | ||
f36a9ecd PB |
1436 | #define IMSM_4K_DIV 8 |
1437 | ||
c47b0ff6 AK |
1438 | static __u64 blocks_per_migr_unit(struct intel_super *super, |
1439 | struct imsm_dev *dev); | |
1e5c6983 | 1440 | |
c47b0ff6 AK |
1441 | static void print_imsm_dev(struct intel_super *super, |
1442 | struct imsm_dev *dev, | |
1443 | char *uuid, | |
1444 | int disk_idx) | |
cdddbdbc DW |
1445 | { |
1446 | __u64 sz; | |
0d80bb2f | 1447 | int slot, i; |
238c0a71 AK |
1448 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1449 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); | |
b10b37b8 | 1450 | __u32 ord; |
cdddbdbc DW |
1451 | |
1452 | printf("\n"); | |
1e7bc0ed | 1453 | printf("[%.16s]:\n", dev->volume); |
44470971 | 1454 | printf(" UUID : %s\n", uuid); |
dd8bcb3b AK |
1455 | printf(" RAID Level : %d", get_imsm_raid_level(map)); |
1456 | if (map2) | |
1457 | printf(" <-- %d", get_imsm_raid_level(map2)); | |
1458 | printf("\n"); | |
1459 | printf(" Members : %d", map->num_members); | |
1460 | if (map2) | |
1461 | printf(" <-- %d", map2->num_members); | |
1462 | printf("\n"); | |
0d80bb2f DW |
1463 | printf(" Slots : ["); |
1464 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 1465 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_0); |
0d80bb2f DW |
1466 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
1467 | } | |
dd8bcb3b AK |
1468 | printf("]"); |
1469 | if (map2) { | |
1470 | printf(" <-- ["); | |
1471 | for (i = 0; i < map2->num_members; i++) { | |
238c0a71 | 1472 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_1); |
dd8bcb3b AK |
1473 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
1474 | } | |
1475 | printf("]"); | |
1476 | } | |
1477 | printf("\n"); | |
7095bccb AK |
1478 | printf(" Failed disk : "); |
1479 | if (map->failed_disk_num == 0xff) | |
1480 | printf("none"); | |
1481 | else | |
1482 | printf("%i", map->failed_disk_num); | |
1483 | printf("\n"); | |
620b1713 DW |
1484 | slot = get_imsm_disk_slot(map, disk_idx); |
1485 | if (slot >= 0) { | |
238c0a71 | 1486 | ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); |
b10b37b8 DW |
1487 | printf(" This Slot : %d%s\n", slot, |
1488 | ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : ""); | |
1489 | } else | |
cdddbdbc | 1490 | printf(" This Slot : ?\n"); |
84918897 | 1491 | printf(" Sector Size : %u\n", super->sector_size); |
cdddbdbc DW |
1492 | sz = __le32_to_cpu(dev->size_high); |
1493 | sz <<= 32; | |
1494 | sz += __le32_to_cpu(dev->size_low); | |
84918897 MK |
1495 | printf(" Array Size : %llu%s\n", |
1496 | (unsigned long long)sz * 512 / super->sector_size, | |
cdddbdbc | 1497 | human_size(sz * 512)); |
5551b113 | 1498 | sz = blocks_per_member(map); |
84918897 MK |
1499 | printf(" Per Dev Size : %llu%s\n", |
1500 | (unsigned long long)sz * 512 / super->sector_size, | |
cdddbdbc | 1501 | human_size(sz * 512)); |
5551b113 CA |
1502 | printf(" Sector Offset : %llu\n", |
1503 | pba_of_lba0(map)); | |
1504 | printf(" Num Stripes : %llu\n", | |
1505 | num_data_stripes(map)); | |
dd8bcb3b | 1506 | printf(" Chunk Size : %u KiB", |
cdddbdbc | 1507 | __le16_to_cpu(map->blocks_per_strip) / 2); |
dd8bcb3b AK |
1508 | if (map2) |
1509 | printf(" <-- %u KiB", | |
1510 | __le16_to_cpu(map2->blocks_per_strip) / 2); | |
1511 | printf("\n"); | |
cdddbdbc | 1512 | printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks)); |
8655a7b1 | 1513 | printf(" Migrate State : "); |
1484e727 DW |
1514 | if (dev->vol.migr_state) { |
1515 | if (migr_type(dev) == MIGR_INIT) | |
8655a7b1 | 1516 | printf("initialize\n"); |
1484e727 | 1517 | else if (migr_type(dev) == MIGR_REBUILD) |
8655a7b1 | 1518 | printf("rebuild\n"); |
1484e727 | 1519 | else if (migr_type(dev) == MIGR_VERIFY) |
8655a7b1 | 1520 | printf("check\n"); |
1484e727 | 1521 | else if (migr_type(dev) == MIGR_GEN_MIGR) |
8655a7b1 | 1522 | printf("general migration\n"); |
1484e727 | 1523 | else if (migr_type(dev) == MIGR_STATE_CHANGE) |
8655a7b1 | 1524 | printf("state change\n"); |
1484e727 | 1525 | else if (migr_type(dev) == MIGR_REPAIR) |
8655a7b1 | 1526 | printf("repair\n"); |
1484e727 | 1527 | else |
8655a7b1 DW |
1528 | printf("<unknown:%d>\n", migr_type(dev)); |
1529 | } else | |
1530 | printf("idle\n"); | |
3393c6af DW |
1531 | printf(" Map State : %s", map_state_str[map->map_state]); |
1532 | if (dev->vol.migr_state) { | |
238c0a71 | 1533 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 | 1534 | |
b10b37b8 | 1535 | printf(" <-- %s", map_state_str[map->map_state]); |
464d40e8 LD |
1536 | printf("\n Checkpoint : %u ", |
1537 | __le32_to_cpu(dev->vol.curr_migr_unit)); | |
089f9d79 | 1538 | if (is_gen_migration(dev) && (slot > 1 || slot < 0)) |
464d40e8 LD |
1539 | printf("(N/A)"); |
1540 | else | |
1541 | printf("(%llu)", (unsigned long long) | |
1542 | blocks_per_migr_unit(super, dev)); | |
3393c6af DW |
1543 | } |
1544 | printf("\n"); | |
2432ce9b AP |
1545 | printf(" Dirty State : %s\n", (dev->vol.dirty & RAIDVOL_DIRTY) ? |
1546 | "dirty" : "clean"); | |
1547 | printf(" RWH Policy : "); | |
c2462068 | 1548 | if (dev->rwh_policy == RWH_OFF || dev->rwh_policy == RWH_MULTIPLE_OFF) |
2432ce9b AP |
1549 | printf("off\n"); |
1550 | else if (dev->rwh_policy == RWH_DISTRIBUTED) | |
1551 | printf("PPL distributed\n"); | |
1552 | else if (dev->rwh_policy == RWH_JOURNALING_DRIVE) | |
1553 | printf("PPL journaling drive\n"); | |
c2462068 PB |
1554 | else if (dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED) |
1555 | printf("Multiple distributed PPLs\n"); | |
1556 | else if (dev->rwh_policy == RWH_MULTIPLE_PPLS_JOURNALING_DRIVE) | |
1557 | printf("Multiple PPLs on journaling drive\n"); | |
2432ce9b AP |
1558 | else |
1559 | printf("<unknown:%d>\n", dev->rwh_policy); | |
cdddbdbc DW |
1560 | } |
1561 | ||
ef5c214e MK |
1562 | static void print_imsm_disk(struct imsm_disk *disk, |
1563 | int index, | |
1564 | __u32 reserved, | |
1565 | unsigned int sector_size) { | |
1f24f035 | 1566 | char str[MAX_RAID_SERIAL_LEN + 1]; |
cdddbdbc DW |
1567 | __u64 sz; |
1568 | ||
0ec1f4e8 | 1569 | if (index < -1 || !disk) |
e9d82038 DW |
1570 | return; |
1571 | ||
cdddbdbc | 1572 | printf("\n"); |
1f24f035 | 1573 | snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial); |
0ec1f4e8 DW |
1574 | if (index >= 0) |
1575 | printf(" Disk%02d Serial : %s\n", index, str); | |
1576 | else | |
1577 | printf(" Disk Serial : %s\n", str); | |
2432ce9b AP |
1578 | printf(" State :%s%s%s%s\n", is_spare(disk) ? " spare" : "", |
1579 | is_configured(disk) ? " active" : "", | |
1580 | is_failed(disk) ? " failed" : "", | |
1581 | is_journal(disk) ? " journal" : ""); | |
cdddbdbc | 1582 | printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id)); |
5551b113 | 1583 | sz = total_blocks(disk) - reserved; |
ef5c214e MK |
1584 | printf(" Usable Size : %llu%s\n", |
1585 | (unsigned long long)sz * 512 / sector_size, | |
cdddbdbc DW |
1586 | human_size(sz * 512)); |
1587 | } | |
1588 | ||
de44e46f PB |
1589 | void convert_to_4k_imsm_migr_rec(struct intel_super *super) |
1590 | { | |
1591 | struct migr_record *migr_rec = super->migr_rec; | |
1592 | ||
1593 | migr_rec->blocks_per_unit /= IMSM_4K_DIV; | |
1594 | migr_rec->ckpt_area_pba /= IMSM_4K_DIV; | |
1595 | migr_rec->dest_1st_member_lba /= IMSM_4K_DIV; | |
1596 | migr_rec->dest_depth_per_unit /= IMSM_4K_DIV; | |
1597 | split_ull((join_u32(migr_rec->post_migr_vol_cap, | |
1598 | migr_rec->post_migr_vol_cap_hi) / IMSM_4K_DIV), | |
1599 | &migr_rec->post_migr_vol_cap, &migr_rec->post_migr_vol_cap_hi); | |
1600 | } | |
1601 | ||
f36a9ecd PB |
1602 | void convert_to_4k_imsm_disk(struct imsm_disk *disk) |
1603 | { | |
1604 | set_total_blocks(disk, (total_blocks(disk)/IMSM_4K_DIV)); | |
1605 | } | |
1606 | ||
1607 | void convert_to_4k(struct intel_super *super) | |
1608 | { | |
1609 | struct imsm_super *mpb = super->anchor; | |
1610 | struct imsm_disk *disk; | |
1611 | int i; | |
e4467bc7 | 1612 | __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size); |
f36a9ecd PB |
1613 | |
1614 | for (i = 0; i < mpb->num_disks ; i++) { | |
1615 | disk = __get_imsm_disk(mpb, i); | |
1616 | /* disk */ | |
1617 | convert_to_4k_imsm_disk(disk); | |
1618 | } | |
1619 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
1620 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
1621 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
1622 | /* dev */ | |
1623 | split_ull((join_u32(dev->size_low, dev->size_high)/IMSM_4K_DIV), | |
1624 | &dev->size_low, &dev->size_high); | |
1625 | dev->vol.curr_migr_unit /= IMSM_4K_DIV; | |
1626 | ||
1627 | /* map0 */ | |
1628 | set_blocks_per_member(map, blocks_per_member(map)/IMSM_4K_DIV); | |
1629 | map->blocks_per_strip /= IMSM_4K_DIV; | |
1630 | set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV); | |
1631 | ||
1632 | if (dev->vol.migr_state) { | |
1633 | /* map1 */ | |
1634 | map = get_imsm_map(dev, MAP_1); | |
1635 | set_blocks_per_member(map, | |
1636 | blocks_per_member(map)/IMSM_4K_DIV); | |
1637 | map->blocks_per_strip /= IMSM_4K_DIV; | |
1638 | set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV); | |
1639 | } | |
1640 | } | |
e4467bc7 TM |
1641 | if (bbm_log_size) { |
1642 | struct bbm_log *log = (void *)mpb + | |
1643 | __le32_to_cpu(mpb->mpb_size) - bbm_log_size; | |
1644 | __u32 i; | |
1645 | ||
1646 | for (i = 0; i < log->entry_count; i++) { | |
1647 | struct bbm_log_entry *entry = | |
1648 | &log->marked_block_entries[i]; | |
1649 | ||
1650 | __u8 count = entry->marked_count + 1; | |
1651 | unsigned long long sector = | |
1652 | __le48_to_cpu(&entry->defective_block_start); | |
1653 | ||
1654 | entry->defective_block_start = | |
1655 | __cpu_to_le48(sector/IMSM_4K_DIV); | |
1656 | entry->marked_count = max(count/IMSM_4K_DIV, 1) - 1; | |
1657 | } | |
1658 | } | |
f36a9ecd PB |
1659 | |
1660 | mpb->check_sum = __gen_imsm_checksum(mpb); | |
1661 | } | |
1662 | ||
520e69e2 AK |
1663 | void examine_migr_rec_imsm(struct intel_super *super) |
1664 | { | |
1665 | struct migr_record *migr_rec = super->migr_rec; | |
1666 | struct imsm_super *mpb = super->anchor; | |
1667 | int i; | |
1668 | ||
1669 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
1670 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
3136abe5 | 1671 | struct imsm_map *map; |
b4ab44d8 | 1672 | int slot = -1; |
3136abe5 | 1673 | |
520e69e2 AK |
1674 | if (is_gen_migration(dev) == 0) |
1675 | continue; | |
1676 | ||
1677 | printf("\nMigration Record Information:"); | |
3136abe5 | 1678 | |
44bfe6df AK |
1679 | /* first map under migration */ |
1680 | map = get_imsm_map(dev, MAP_0); | |
3136abe5 AK |
1681 | if (map) |
1682 | slot = get_imsm_disk_slot(map, super->disks->index); | |
089f9d79 | 1683 | if (map == NULL || slot > 1 || slot < 0) { |
520e69e2 AK |
1684 | printf(" Empty\n "); |
1685 | printf("Examine one of first two disks in array\n"); | |
1686 | break; | |
1687 | } | |
1688 | printf("\n Status : "); | |
1689 | if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL) | |
1690 | printf("Normal\n"); | |
1691 | else | |
1692 | printf("Contains Data\n"); | |
1693 | printf(" Current Unit : %u\n", | |
1694 | __le32_to_cpu(migr_rec->curr_migr_unit)); | |
1695 | printf(" Family : %u\n", | |
1696 | __le32_to_cpu(migr_rec->family_num)); | |
1697 | printf(" Ascending : %u\n", | |
1698 | __le32_to_cpu(migr_rec->ascending_migr)); | |
1699 | printf(" Blocks Per Unit : %u\n", | |
1700 | __le32_to_cpu(migr_rec->blocks_per_unit)); | |
1701 | printf(" Dest. Depth Per Unit : %u\n", | |
1702 | __le32_to_cpu(migr_rec->dest_depth_per_unit)); | |
1703 | printf(" Checkpoint Area pba : %u\n", | |
1704 | __le32_to_cpu(migr_rec->ckpt_area_pba)); | |
1705 | printf(" First member lba : %u\n", | |
1706 | __le32_to_cpu(migr_rec->dest_1st_member_lba)); | |
1707 | printf(" Total Number of Units : %u\n", | |
1708 | __le32_to_cpu(migr_rec->num_migr_units)); | |
1709 | printf(" Size of volume : %u\n", | |
1710 | __le32_to_cpu(migr_rec->post_migr_vol_cap)); | |
1711 | printf(" Expansion space for LBA64 : %u\n", | |
1712 | __le32_to_cpu(migr_rec->post_migr_vol_cap_hi)); | |
1713 | printf(" Record was read from : %u\n", | |
1714 | __le32_to_cpu(migr_rec->ckpt_read_disk_num)); | |
1715 | ||
1716 | break; | |
1717 | } | |
1718 | } | |
f36a9ecd | 1719 | |
de44e46f PB |
1720 | void convert_from_4k_imsm_migr_rec(struct intel_super *super) |
1721 | { | |
1722 | struct migr_record *migr_rec = super->migr_rec; | |
1723 | ||
1724 | migr_rec->blocks_per_unit *= IMSM_4K_DIV; | |
1725 | migr_rec->ckpt_area_pba *= IMSM_4K_DIV; | |
1726 | migr_rec->dest_1st_member_lba *= IMSM_4K_DIV; | |
1727 | migr_rec->dest_depth_per_unit *= IMSM_4K_DIV; | |
1728 | split_ull((join_u32(migr_rec->post_migr_vol_cap, | |
1729 | migr_rec->post_migr_vol_cap_hi) * IMSM_4K_DIV), | |
1730 | &migr_rec->post_migr_vol_cap, | |
1731 | &migr_rec->post_migr_vol_cap_hi); | |
1732 | } | |
1733 | ||
f36a9ecd PB |
1734 | void convert_from_4k(struct intel_super *super) |
1735 | { | |
1736 | struct imsm_super *mpb = super->anchor; | |
1737 | struct imsm_disk *disk; | |
1738 | int i; | |
e4467bc7 | 1739 | __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size); |
f36a9ecd PB |
1740 | |
1741 | for (i = 0; i < mpb->num_disks ; i++) { | |
1742 | disk = __get_imsm_disk(mpb, i); | |
1743 | /* disk */ | |
1744 | set_total_blocks(disk, (total_blocks(disk)*IMSM_4K_DIV)); | |
1745 | } | |
1746 | ||
1747 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
1748 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
1749 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
1750 | /* dev */ | |
1751 | split_ull((join_u32(dev->size_low, dev->size_high)*IMSM_4K_DIV), | |
1752 | &dev->size_low, &dev->size_high); | |
1753 | dev->vol.curr_migr_unit *= IMSM_4K_DIV; | |
1754 | ||
1755 | /* map0 */ | |
1756 | set_blocks_per_member(map, blocks_per_member(map)*IMSM_4K_DIV); | |
1757 | map->blocks_per_strip *= IMSM_4K_DIV; | |
1758 | set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV); | |
1759 | ||
1760 | if (dev->vol.migr_state) { | |
1761 | /* map1 */ | |
1762 | map = get_imsm_map(dev, MAP_1); | |
1763 | set_blocks_per_member(map, | |
1764 | blocks_per_member(map)*IMSM_4K_DIV); | |
1765 | map->blocks_per_strip *= IMSM_4K_DIV; | |
1766 | set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV); | |
1767 | } | |
1768 | } | |
e4467bc7 TM |
1769 | if (bbm_log_size) { |
1770 | struct bbm_log *log = (void *)mpb + | |
1771 | __le32_to_cpu(mpb->mpb_size) - bbm_log_size; | |
1772 | __u32 i; | |
1773 | ||
1774 | for (i = 0; i < log->entry_count; i++) { | |
1775 | struct bbm_log_entry *entry = | |
1776 | &log->marked_block_entries[i]; | |
1777 | ||
1778 | __u8 count = entry->marked_count + 1; | |
1779 | unsigned long long sector = | |
1780 | __le48_to_cpu(&entry->defective_block_start); | |
1781 | ||
1782 | entry->defective_block_start = | |
1783 | __cpu_to_le48(sector*IMSM_4K_DIV); | |
1784 | entry->marked_count = count*IMSM_4K_DIV - 1; | |
1785 | } | |
1786 | } | |
f36a9ecd PB |
1787 | |
1788 | mpb->check_sum = __gen_imsm_checksum(mpb); | |
1789 | } | |
1790 | ||
19482bcc AK |
1791 | /******************************************************************************* |
1792 | * function: imsm_check_attributes | |
1793 | * Description: Function checks if features represented by attributes flags | |
1011e834 | 1794 | * are supported by mdadm. |
19482bcc AK |
1795 | * Parameters: |
1796 | * attributes - Attributes read from metadata | |
1797 | * Returns: | |
1011e834 N |
1798 | * 0 - passed attributes contains unsupported features flags |
1799 | * 1 - all features are supported | |
19482bcc AK |
1800 | ******************************************************************************/ |
1801 | static int imsm_check_attributes(__u32 attributes) | |
1802 | { | |
1803 | int ret_val = 1; | |
418f9b36 N |
1804 | __u32 not_supported = MPB_ATTRIB_SUPPORTED^0xffffffff; |
1805 | ||
1806 | not_supported &= ~MPB_ATTRIB_IGNORED; | |
19482bcc AK |
1807 | |
1808 | not_supported &= attributes; | |
1809 | if (not_supported) { | |
e7b84f9d | 1810 | pr_err("(IMSM): Unsupported attributes : %x\n", |
418f9b36 | 1811 | (unsigned)__le32_to_cpu(not_supported)); |
19482bcc AK |
1812 | if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) { |
1813 | dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY \n"); | |
1814 | not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY; | |
1815 | } | |
1816 | if (not_supported & MPB_ATTRIB_2TB) { | |
1817 | dprintf("\t\tMPB_ATTRIB_2TB\n"); | |
1818 | not_supported ^= MPB_ATTRIB_2TB; | |
1819 | } | |
1820 | if (not_supported & MPB_ATTRIB_RAID0) { | |
1821 | dprintf("\t\tMPB_ATTRIB_RAID0\n"); | |
1822 | not_supported ^= MPB_ATTRIB_RAID0; | |
1823 | } | |
1824 | if (not_supported & MPB_ATTRIB_RAID1) { | |
1825 | dprintf("\t\tMPB_ATTRIB_RAID1\n"); | |
1826 | not_supported ^= MPB_ATTRIB_RAID1; | |
1827 | } | |
1828 | if (not_supported & MPB_ATTRIB_RAID10) { | |
1829 | dprintf("\t\tMPB_ATTRIB_RAID10\n"); | |
1830 | not_supported ^= MPB_ATTRIB_RAID10; | |
1831 | } | |
1832 | if (not_supported & MPB_ATTRIB_RAID1E) { | |
1833 | dprintf("\t\tMPB_ATTRIB_RAID1E\n"); | |
1834 | not_supported ^= MPB_ATTRIB_RAID1E; | |
1835 | } | |
1836 | if (not_supported & MPB_ATTRIB_RAID5) { | |
1837 | dprintf("\t\tMPB_ATTRIB_RAID5\n"); | |
1838 | not_supported ^= MPB_ATTRIB_RAID5; | |
1839 | } | |
1840 | if (not_supported & MPB_ATTRIB_RAIDCNG) { | |
1841 | dprintf("\t\tMPB_ATTRIB_RAIDCNG\n"); | |
1842 | not_supported ^= MPB_ATTRIB_RAIDCNG; | |
1843 | } | |
1844 | if (not_supported & MPB_ATTRIB_BBM) { | |
1845 | dprintf("\t\tMPB_ATTRIB_BBM\n"); | |
1846 | not_supported ^= MPB_ATTRIB_BBM; | |
1847 | } | |
1848 | if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) { | |
1849 | dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY (== MPB_ATTRIB_LEGACY)\n"); | |
1850 | not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY; | |
1851 | } | |
1852 | if (not_supported & MPB_ATTRIB_EXP_STRIPE_SIZE) { | |
1853 | dprintf("\t\tMPB_ATTRIB_EXP_STRIP_SIZE\n"); | |
1854 | not_supported ^= MPB_ATTRIB_EXP_STRIPE_SIZE; | |
1855 | } | |
1856 | if (not_supported & MPB_ATTRIB_2TB_DISK) { | |
1857 | dprintf("\t\tMPB_ATTRIB_2TB_DISK\n"); | |
1858 | not_supported ^= MPB_ATTRIB_2TB_DISK; | |
1859 | } | |
1860 | if (not_supported & MPB_ATTRIB_NEVER_USE2) { | |
1861 | dprintf("\t\tMPB_ATTRIB_NEVER_USE2\n"); | |
1862 | not_supported ^= MPB_ATTRIB_NEVER_USE2; | |
1863 | } | |
1864 | if (not_supported & MPB_ATTRIB_NEVER_USE) { | |
1865 | dprintf("\t\tMPB_ATTRIB_NEVER_USE\n"); | |
1866 | not_supported ^= MPB_ATTRIB_NEVER_USE; | |
1867 | } | |
1868 | ||
1869 | if (not_supported) | |
1ade5cc1 | 1870 | dprintf("(IMSM): Unknown attributes : %x\n", not_supported); |
19482bcc AK |
1871 | |
1872 | ret_val = 0; | |
1873 | } | |
1874 | ||
1875 | return ret_val; | |
1876 | } | |
1877 | ||
a5d85af7 | 1878 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map); |
44470971 | 1879 | |
cdddbdbc DW |
1880 | static void examine_super_imsm(struct supertype *st, char *homehost) |
1881 | { | |
1882 | struct intel_super *super = st->sb; | |
949c47a0 | 1883 | struct imsm_super *mpb = super->anchor; |
cdddbdbc DW |
1884 | char str[MAX_SIGNATURE_LENGTH]; |
1885 | int i; | |
27fd6274 DW |
1886 | struct mdinfo info; |
1887 | char nbuf[64]; | |
cdddbdbc | 1888 | __u32 sum; |
14e8215b | 1889 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
94827db3 | 1890 | struct dl *dl; |
27fd6274 | 1891 | |
618f4e6d XN |
1892 | strncpy(str, (char *)mpb->sig, MPB_SIG_LEN); |
1893 | str[MPB_SIG_LEN-1] = '\0'; | |
cdddbdbc DW |
1894 | printf(" Magic : %s\n", str); |
1895 | snprintf(str, strlen(MPB_VERSION_RAID0), "%s", get_imsm_version(mpb)); | |
1896 | printf(" Version : %s\n", get_imsm_version(mpb)); | |
148acb7b | 1897 | printf(" Orig Family : %08x\n", __le32_to_cpu(mpb->orig_family_num)); |
cdddbdbc DW |
1898 | printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num)); |
1899 | printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num)); | |
19482bcc AK |
1900 | printf(" Attributes : "); |
1901 | if (imsm_check_attributes(mpb->attributes)) | |
1902 | printf("All supported\n"); | |
1903 | else | |
1904 | printf("not supported\n"); | |
a5d85af7 | 1905 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1906 | fname_from_uuid(st, &info, nbuf, ':'); |
27fd6274 | 1907 | printf(" UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
1908 | sum = __le32_to_cpu(mpb->check_sum); |
1909 | printf(" Checksum : %08x %s\n", sum, | |
949c47a0 | 1910 | __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect"); |
f36a9ecd | 1911 | printf(" MPB Sectors : %d\n", mpb_sectors(mpb, super->sector_size)); |
cdddbdbc DW |
1912 | printf(" Disks : %d\n", mpb->num_disks); |
1913 | printf(" RAID Devices : %d\n", mpb->num_raid_devs); | |
ef5c214e MK |
1914 | print_imsm_disk(__get_imsm_disk(mpb, super->disks->index), |
1915 | super->disks->index, reserved, super->sector_size); | |
8d67477f | 1916 | if (get_imsm_bbm_log_size(super->bbm_log)) { |
604b746f JD |
1917 | struct bbm_log *log = super->bbm_log; |
1918 | ||
1919 | printf("\n"); | |
1920 | printf("Bad Block Management Log:\n"); | |
1921 | printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size)); | |
1922 | printf(" Signature : %x\n", __le32_to_cpu(log->signature)); | |
1923 | printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count)); | |
604b746f | 1924 | } |
44470971 DW |
1925 | for (i = 0; i < mpb->num_raid_devs; i++) { |
1926 | struct mdinfo info; | |
1927 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
1928 | ||
1929 | super->current_vol = i; | |
a5d85af7 | 1930 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1931 | fname_from_uuid(st, &info, nbuf, ':'); |
c47b0ff6 | 1932 | print_imsm_dev(super, dev, nbuf + 5, super->disks->index); |
44470971 | 1933 | } |
cdddbdbc DW |
1934 | for (i = 0; i < mpb->num_disks; i++) { |
1935 | if (i == super->disks->index) | |
1936 | continue; | |
ef5c214e MK |
1937 | print_imsm_disk(__get_imsm_disk(mpb, i), i, reserved, |
1938 | super->sector_size); | |
cdddbdbc | 1939 | } |
94827db3 | 1940 | |
0ec1f4e8 DW |
1941 | for (dl = super->disks; dl; dl = dl->next) |
1942 | if (dl->index == -1) | |
ef5c214e MK |
1943 | print_imsm_disk(&dl->disk, -1, reserved, |
1944 | super->sector_size); | |
520e69e2 AK |
1945 | |
1946 | examine_migr_rec_imsm(super); | |
cdddbdbc DW |
1947 | } |
1948 | ||
061f2c6a | 1949 | static void brief_examine_super_imsm(struct supertype *st, int verbose) |
cdddbdbc | 1950 | { |
27fd6274 | 1951 | /* We just write a generic IMSM ARRAY entry */ |
ff54de6e N |
1952 | struct mdinfo info; |
1953 | char nbuf[64]; | |
1e7bc0ed | 1954 | struct intel_super *super = st->sb; |
1e7bc0ed | 1955 | |
0d5a423f DW |
1956 | if (!super->anchor->num_raid_devs) { |
1957 | printf("ARRAY metadata=imsm\n"); | |
1e7bc0ed | 1958 | return; |
0d5a423f | 1959 | } |
ff54de6e | 1960 | |
a5d85af7 | 1961 | getinfo_super_imsm(st, &info, NULL); |
4737ae25 N |
1962 | fname_from_uuid(st, &info, nbuf, ':'); |
1963 | printf("ARRAY metadata=imsm UUID=%s\n", nbuf + 5); | |
1964 | } | |
1965 | ||
1966 | static void brief_examine_subarrays_imsm(struct supertype *st, int verbose) | |
1967 | { | |
1968 | /* We just write a generic IMSM ARRAY entry */ | |
1969 | struct mdinfo info; | |
1970 | char nbuf[64]; | |
1971 | char nbuf1[64]; | |
1972 | struct intel_super *super = st->sb; | |
1973 | int i; | |
1974 | ||
1975 | if (!super->anchor->num_raid_devs) | |
1976 | return; | |
1977 | ||
a5d85af7 | 1978 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1979 | fname_from_uuid(st, &info, nbuf, ':'); |
1e7bc0ed DW |
1980 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
1981 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
1982 | ||
1983 | super->current_vol = i; | |
a5d85af7 | 1984 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1985 | fname_from_uuid(st, &info, nbuf1, ':'); |
1124b3cf | 1986 | printf("ARRAY /dev/md/%.16s container=%s member=%d UUID=%s\n", |
cf8de691 | 1987 | dev->volume, nbuf + 5, i, nbuf1 + 5); |
1e7bc0ed | 1988 | } |
cdddbdbc DW |
1989 | } |
1990 | ||
9d84c8ea DW |
1991 | static void export_examine_super_imsm(struct supertype *st) |
1992 | { | |
1993 | struct intel_super *super = st->sb; | |
1994 | struct imsm_super *mpb = super->anchor; | |
1995 | struct mdinfo info; | |
1996 | char nbuf[64]; | |
1997 | ||
a5d85af7 | 1998 | getinfo_super_imsm(st, &info, NULL); |
9d84c8ea DW |
1999 | fname_from_uuid(st, &info, nbuf, ':'); |
2000 | printf("MD_METADATA=imsm\n"); | |
2001 | printf("MD_LEVEL=container\n"); | |
2002 | printf("MD_UUID=%s\n", nbuf+5); | |
2003 | printf("MD_DEVICES=%u\n", mpb->num_disks); | |
2004 | } | |
2005 | ||
74db60b0 N |
2006 | static int copy_metadata_imsm(struct supertype *st, int from, int to) |
2007 | { | |
f36a9ecd | 2008 | /* The second last sector of the device contains |
74db60b0 N |
2009 | * the "struct imsm_super" metadata. |
2010 | * This contains mpb_size which is the size in bytes of the | |
2011 | * extended metadata. This is located immediately before | |
2012 | * the imsm_super. | |
2013 | * We want to read all that, plus the last sector which | |
2014 | * may contain a migration record, and write it all | |
2015 | * to the target. | |
2016 | */ | |
2017 | void *buf; | |
2018 | unsigned long long dsize, offset; | |
2019 | int sectors; | |
2020 | struct imsm_super *sb; | |
f36a9ecd PB |
2021 | struct intel_super *super = st->sb; |
2022 | unsigned int sector_size = super->sector_size; | |
2023 | unsigned int written = 0; | |
74db60b0 | 2024 | |
de44e46f | 2025 | if (posix_memalign(&buf, MAX_SECTOR_SIZE, MAX_SECTOR_SIZE) != 0) |
74db60b0 N |
2026 | return 1; |
2027 | ||
2028 | if (!get_dev_size(from, NULL, &dsize)) | |
2029 | goto err; | |
2030 | ||
f36a9ecd | 2031 | if (lseek64(from, dsize-(2*sector_size), 0) < 0) |
74db60b0 | 2032 | goto err; |
466070ad | 2033 | if ((unsigned int)read(from, buf, sector_size) != sector_size) |
74db60b0 N |
2034 | goto err; |
2035 | sb = buf; | |
2036 | if (strncmp((char*)sb->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) | |
2037 | goto err; | |
2038 | ||
f36a9ecd PB |
2039 | sectors = mpb_sectors(sb, sector_size) + 2; |
2040 | offset = dsize - sectors * sector_size; | |
74db60b0 N |
2041 | if (lseek64(from, offset, 0) < 0 || |
2042 | lseek64(to, offset, 0) < 0) | |
2043 | goto err; | |
f36a9ecd PB |
2044 | while (written < sectors * sector_size) { |
2045 | int n = sectors*sector_size - written; | |
74db60b0 N |
2046 | if (n > 4096) |
2047 | n = 4096; | |
2048 | if (read(from, buf, n) != n) | |
2049 | goto err; | |
2050 | if (write(to, buf, n) != n) | |
2051 | goto err; | |
2052 | written += n; | |
2053 | } | |
2054 | free(buf); | |
2055 | return 0; | |
2056 | err: | |
2057 | free(buf); | |
2058 | return 1; | |
2059 | } | |
2060 | ||
cdddbdbc DW |
2061 | static void detail_super_imsm(struct supertype *st, char *homehost) |
2062 | { | |
3ebe00a1 DW |
2063 | struct mdinfo info; |
2064 | char nbuf[64]; | |
2065 | ||
a5d85af7 | 2066 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 2067 | fname_from_uuid(st, &info, nbuf, ':'); |
65884368 | 2068 | printf("\n UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
2069 | } |
2070 | ||
2071 | static void brief_detail_super_imsm(struct supertype *st) | |
2072 | { | |
ff54de6e N |
2073 | struct mdinfo info; |
2074 | char nbuf[64]; | |
a5d85af7 | 2075 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 2076 | fname_from_uuid(st, &info, nbuf, ':'); |
ff54de6e | 2077 | printf(" UUID=%s", nbuf + 5); |
cdddbdbc | 2078 | } |
d665cc31 DW |
2079 | |
2080 | static int imsm_read_serial(int fd, char *devname, __u8 *serial); | |
2081 | static void fd2devname(int fd, char *name); | |
2082 | ||
120dc887 | 2083 | static int ahci_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose) |
d665cc31 | 2084 | { |
120dc887 LM |
2085 | /* dump an unsorted list of devices attached to AHCI Intel storage |
2086 | * controller, as well as non-connected ports | |
d665cc31 DW |
2087 | */ |
2088 | int hba_len = strlen(hba_path) + 1; | |
2089 | struct dirent *ent; | |
2090 | DIR *dir; | |
2091 | char *path = NULL; | |
2092 | int err = 0; | |
2093 | unsigned long port_mask = (1 << port_count) - 1; | |
2094 | ||
f21e18ca | 2095 | if (port_count > (int)sizeof(port_mask) * 8) { |
ba728be7 | 2096 | if (verbose > 0) |
e7b84f9d | 2097 | pr_err("port_count %d out of range\n", port_count); |
d665cc31 DW |
2098 | return 2; |
2099 | } | |
2100 | ||
2101 | /* scroll through /sys/dev/block looking for devices attached to | |
2102 | * this hba | |
2103 | */ | |
2104 | dir = opendir("/sys/dev/block"); | |
1a6dd6b9 PB |
2105 | if (!dir) |
2106 | return 1; | |
2107 | ||
2108 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
d665cc31 DW |
2109 | int fd; |
2110 | char model[64]; | |
2111 | char vendor[64]; | |
2112 | char buf[1024]; | |
2113 | int major, minor; | |
2114 | char *device; | |
2115 | char *c; | |
2116 | int port; | |
2117 | int type; | |
2118 | ||
2119 | if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2) | |
2120 | continue; | |
2121 | path = devt_to_devpath(makedev(major, minor)); | |
2122 | if (!path) | |
2123 | continue; | |
2124 | if (!path_attached_to_hba(path, hba_path)) { | |
2125 | free(path); | |
2126 | path = NULL; | |
2127 | continue; | |
2128 | } | |
2129 | ||
2130 | /* retrieve the scsi device type */ | |
2131 | if (asprintf(&device, "/sys/dev/block/%d:%d/device/xxxxxxx", major, minor) < 0) { | |
ba728be7 | 2132 | if (verbose > 0) |
e7b84f9d | 2133 | pr_err("failed to allocate 'device'\n"); |
d665cc31 DW |
2134 | err = 2; |
2135 | break; | |
2136 | } | |
2137 | sprintf(device, "/sys/dev/block/%d:%d/device/type", major, minor); | |
193b6c0b | 2138 | if (load_sys(device, buf, sizeof(buf)) != 0) { |
ba728be7 | 2139 | if (verbose > 0) |
e7b84f9d | 2140 | pr_err("failed to read device type for %s\n", |
d665cc31 DW |
2141 | path); |
2142 | err = 2; | |
2143 | free(device); | |
2144 | break; | |
2145 | } | |
2146 | type = strtoul(buf, NULL, 10); | |
2147 | ||
2148 | /* if it's not a disk print the vendor and model */ | |
2149 | if (!(type == 0 || type == 7 || type == 14)) { | |
2150 | vendor[0] = '\0'; | |
2151 | model[0] = '\0'; | |
2152 | sprintf(device, "/sys/dev/block/%d:%d/device/vendor", major, minor); | |
193b6c0b | 2153 | if (load_sys(device, buf, sizeof(buf)) == 0) { |
d665cc31 DW |
2154 | strncpy(vendor, buf, sizeof(vendor)); |
2155 | vendor[sizeof(vendor) - 1] = '\0'; | |
2156 | c = (char *) &vendor[sizeof(vendor) - 1]; | |
2157 | while (isspace(*c) || *c == '\0') | |
2158 | *c-- = '\0'; | |
2159 | ||
2160 | } | |
2161 | sprintf(device, "/sys/dev/block/%d:%d/device/model", major, minor); | |
193b6c0b | 2162 | if (load_sys(device, buf, sizeof(buf)) == 0) { |
d665cc31 DW |
2163 | strncpy(model, buf, sizeof(model)); |
2164 | model[sizeof(model) - 1] = '\0'; | |
2165 | c = (char *) &model[sizeof(model) - 1]; | |
2166 | while (isspace(*c) || *c == '\0') | |
2167 | *c-- = '\0'; | |
2168 | } | |
2169 | ||
2170 | if (vendor[0] && model[0]) | |
2171 | sprintf(buf, "%.64s %.64s", vendor, model); | |
2172 | else | |
2173 | switch (type) { /* numbers from hald/linux/device.c */ | |
2174 | case 1: sprintf(buf, "tape"); break; | |
2175 | case 2: sprintf(buf, "printer"); break; | |
2176 | case 3: sprintf(buf, "processor"); break; | |
2177 | case 4: | |
2178 | case 5: sprintf(buf, "cdrom"); break; | |
2179 | case 6: sprintf(buf, "scanner"); break; | |
2180 | case 8: sprintf(buf, "media_changer"); break; | |
2181 | case 9: sprintf(buf, "comm"); break; | |
2182 | case 12: sprintf(buf, "raid"); break; | |
2183 | default: sprintf(buf, "unknown"); | |
2184 | } | |
2185 | } else | |
2186 | buf[0] = '\0'; | |
2187 | free(device); | |
2188 | ||
2189 | /* chop device path to 'host%d' and calculate the port number */ | |
2190 | c = strchr(&path[hba_len], '/'); | |
4e5e717d | 2191 | if (!c) { |
ba728be7 | 2192 | if (verbose > 0) |
e7b84f9d | 2193 | pr_err("%s - invalid path name\n", path + hba_len); |
4e5e717d AW |
2194 | err = 2; |
2195 | break; | |
2196 | } | |
d665cc31 | 2197 | *c = '\0'; |
0858eccf AP |
2198 | if ((sscanf(&path[hba_len], "ata%d", &port) == 1) || |
2199 | ((sscanf(&path[hba_len], "host%d", &port) == 1))) | |
d665cc31 DW |
2200 | port -= host_base; |
2201 | else { | |
ba728be7 | 2202 | if (verbose > 0) { |
d665cc31 | 2203 | *c = '/'; /* repair the full string */ |
e7b84f9d | 2204 | pr_err("failed to determine port number for %s\n", |
d665cc31 DW |
2205 | path); |
2206 | } | |
2207 | err = 2; | |
2208 | break; | |
2209 | } | |
2210 | ||
2211 | /* mark this port as used */ | |
2212 | port_mask &= ~(1 << port); | |
2213 | ||
2214 | /* print out the device information */ | |
2215 | if (buf[0]) { | |
2216 | printf(" Port%d : - non-disk device (%s) -\n", port, buf); | |
2217 | continue; | |
2218 | } | |
2219 | ||
2220 | fd = dev_open(ent->d_name, O_RDONLY); | |
2221 | if (fd < 0) | |
2222 | printf(" Port%d : - disk info unavailable -\n", port); | |
2223 | else { | |
2224 | fd2devname(fd, buf); | |
2225 | printf(" Port%d : %s", port, buf); | |
2226 | if (imsm_read_serial(fd, NULL, (__u8 *) buf) == 0) | |
664d5325 | 2227 | printf(" (%.*s)\n", MAX_RAID_SERIAL_LEN, buf); |
d665cc31 | 2228 | else |
664d5325 | 2229 | printf(" ()\n"); |
4dab422a | 2230 | close(fd); |
d665cc31 | 2231 | } |
d665cc31 DW |
2232 | free(path); |
2233 | path = NULL; | |
2234 | } | |
2235 | if (path) | |
2236 | free(path); | |
2237 | if (dir) | |
2238 | closedir(dir); | |
2239 | if (err == 0) { | |
2240 | int i; | |
2241 | ||
2242 | for (i = 0; i < port_count; i++) | |
2243 | if (port_mask & (1 << i)) | |
2244 | printf(" Port%d : - no device attached -\n", i); | |
2245 | } | |
2246 | ||
2247 | return err; | |
2248 | } | |
2249 | ||
b5eece69 | 2250 | static int print_vmd_attached_devs(struct sys_dev *hba) |
60f0f54d PB |
2251 | { |
2252 | struct dirent *ent; | |
2253 | DIR *dir; | |
2254 | char path[292]; | |
2255 | char link[256]; | |
2256 | char *c, *rp; | |
2257 | ||
2258 | if (hba->type != SYS_DEV_VMD) | |
b5eece69 | 2259 | return 1; |
60f0f54d PB |
2260 | |
2261 | /* scroll through /sys/dev/block looking for devices attached to | |
2262 | * this hba | |
2263 | */ | |
2264 | dir = opendir("/sys/bus/pci/drivers/nvme"); | |
b9135011 | 2265 | if (!dir) |
b5eece69 | 2266 | return 1; |
b9135011 JS |
2267 | |
2268 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
60f0f54d PB |
2269 | int n; |
2270 | ||
2271 | /* is 'ent' a device? check that the 'subsystem' link exists and | |
2272 | * that its target matches 'bus' | |
2273 | */ | |
2274 | sprintf(path, "/sys/bus/pci/drivers/nvme/%s/subsystem", | |
2275 | ent->d_name); | |
2276 | n = readlink(path, link, sizeof(link)); | |
2277 | if (n < 0 || n >= (int)sizeof(link)) | |
2278 | continue; | |
2279 | link[n] = '\0'; | |
2280 | c = strrchr(link, '/'); | |
2281 | if (!c) | |
2282 | continue; | |
2283 | if (strncmp("pci", c+1, strlen("pci")) != 0) | |
2284 | continue; | |
2285 | ||
2286 | sprintf(path, "/sys/bus/pci/drivers/nvme/%s", ent->d_name); | |
60f0f54d PB |
2287 | |
2288 | rp = realpath(path, NULL); | |
2289 | if (!rp) | |
2290 | continue; | |
2291 | ||
2292 | if (path_attached_to_hba(rp, hba->path)) { | |
2293 | printf(" NVMe under VMD : %s\n", rp); | |
2294 | } | |
2295 | free(rp); | |
2296 | } | |
2297 | ||
b9135011 | 2298 | closedir(dir); |
b5eece69 | 2299 | return 0; |
60f0f54d PB |
2300 | } |
2301 | ||
120dc887 LM |
2302 | static void print_found_intel_controllers(struct sys_dev *elem) |
2303 | { | |
2304 | for (; elem; elem = elem->next) { | |
e7b84f9d | 2305 | pr_err("found Intel(R) "); |
120dc887 LM |
2306 | if (elem->type == SYS_DEV_SATA) |
2307 | fprintf(stderr, "SATA "); | |
155cbb4c LM |
2308 | else if (elem->type == SYS_DEV_SAS) |
2309 | fprintf(stderr, "SAS "); | |
0858eccf AP |
2310 | else if (elem->type == SYS_DEV_NVME) |
2311 | fprintf(stderr, "NVMe "); | |
60f0f54d PB |
2312 | |
2313 | if (elem->type == SYS_DEV_VMD) | |
2314 | fprintf(stderr, "VMD domain"); | |
2315 | else | |
2316 | fprintf(stderr, "RAID controller"); | |
2317 | ||
120dc887 LM |
2318 | if (elem->pci_id) |
2319 | fprintf(stderr, " at %s", elem->pci_id); | |
2320 | fprintf(stderr, ".\n"); | |
2321 | } | |
2322 | fflush(stderr); | |
2323 | } | |
2324 | ||
120dc887 LM |
2325 | static int ahci_get_port_count(const char *hba_path, int *port_count) |
2326 | { | |
2327 | struct dirent *ent; | |
2328 | DIR *dir; | |
2329 | int host_base = -1; | |
2330 | ||
2331 | *port_count = 0; | |
2332 | if ((dir = opendir(hba_path)) == NULL) | |
2333 | return -1; | |
2334 | ||
2335 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
2336 | int host; | |
2337 | ||
0858eccf AP |
2338 | if ((sscanf(ent->d_name, "ata%d", &host) != 1) && |
2339 | ((sscanf(ent->d_name, "host%d", &host) != 1))) | |
120dc887 LM |
2340 | continue; |
2341 | if (*port_count == 0) | |
2342 | host_base = host; | |
2343 | else if (host < host_base) | |
2344 | host_base = host; | |
2345 | ||
2346 | if (host + 1 > *port_count + host_base) | |
2347 | *port_count = host + 1 - host_base; | |
2348 | } | |
2349 | closedir(dir); | |
2350 | return host_base; | |
2351 | } | |
2352 | ||
a891a3c2 LM |
2353 | static void print_imsm_capability(const struct imsm_orom *orom) |
2354 | { | |
0858eccf AP |
2355 | printf(" Platform : Intel(R) "); |
2356 | if (orom->capabilities == 0 && orom->driver_features == 0) | |
2357 | printf("Matrix Storage Manager\n"); | |
2358 | else | |
2359 | printf("Rapid Storage Technology%s\n", | |
2360 | imsm_orom_is_enterprise(orom) ? " enterprise" : ""); | |
2361 | if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build) | |
2362 | printf(" Version : %d.%d.%d.%d\n", orom->major_ver, | |
2363 | orom->minor_ver, orom->hotfix_ver, orom->build); | |
a891a3c2 LM |
2364 | printf(" RAID Levels :%s%s%s%s%s\n", |
2365 | imsm_orom_has_raid0(orom) ? " raid0" : "", | |
2366 | imsm_orom_has_raid1(orom) ? " raid1" : "", | |
2367 | imsm_orom_has_raid1e(orom) ? " raid1e" : "", | |
2368 | imsm_orom_has_raid10(orom) ? " raid10" : "", | |
2369 | imsm_orom_has_raid5(orom) ? " raid5" : ""); | |
2370 | printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n", | |
2371 | imsm_orom_has_chunk(orom, 2) ? " 2k" : "", | |
2372 | imsm_orom_has_chunk(orom, 4) ? " 4k" : "", | |
2373 | imsm_orom_has_chunk(orom, 8) ? " 8k" : "", | |
2374 | imsm_orom_has_chunk(orom, 16) ? " 16k" : "", | |
2375 | imsm_orom_has_chunk(orom, 32) ? " 32k" : "", | |
2376 | imsm_orom_has_chunk(orom, 64) ? " 64k" : "", | |
2377 | imsm_orom_has_chunk(orom, 128) ? " 128k" : "", | |
2378 | imsm_orom_has_chunk(orom, 256) ? " 256k" : "", | |
2379 | imsm_orom_has_chunk(orom, 512) ? " 512k" : "", | |
2380 | imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "", | |
2381 | imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "", | |
2382 | imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "", | |
2383 | imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "", | |
2384 | imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "", | |
2385 | imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "", | |
2386 | imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : ""); | |
29cd0821 CA |
2387 | printf(" 2TB volumes :%s supported\n", |
2388 | (orom->attr & IMSM_OROM_ATTR_2TB)?"":" not"); | |
2389 | printf(" 2TB disks :%s supported\n", | |
2390 | (orom->attr & IMSM_OROM_ATTR_2TB_DISK)?"":" not"); | |
0e7f69a8 | 2391 | printf(" Max Disks : %d\n", orom->tds); |
0858eccf AP |
2392 | printf(" Max Volumes : %d per array, %d per %s\n", |
2393 | orom->vpa, orom->vphba, | |
2394 | imsm_orom_is_nvme(orom) ? "platform" : "controller"); | |
a891a3c2 LM |
2395 | return; |
2396 | } | |
2397 | ||
e50cf220 MN |
2398 | static void print_imsm_capability_export(const struct imsm_orom *orom) |
2399 | { | |
2400 | printf("MD_FIRMWARE_TYPE=imsm\n"); | |
0858eccf AP |
2401 | if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build) |
2402 | printf("IMSM_VERSION=%d.%d.%d.%d\n", orom->major_ver, orom->minor_ver, | |
2403 | orom->hotfix_ver, orom->build); | |
e50cf220 MN |
2404 | printf("IMSM_SUPPORTED_RAID_LEVELS=%s%s%s%s%s\n", |
2405 | imsm_orom_has_raid0(orom) ? "raid0 " : "", | |
2406 | imsm_orom_has_raid1(orom) ? "raid1 " : "", | |
2407 | imsm_orom_has_raid1e(orom) ? "raid1e " : "", | |
2408 | imsm_orom_has_raid5(orom) ? "raid10 " : "", | |
2409 | imsm_orom_has_raid10(orom) ? "raid5 " : ""); | |
2410 | printf("IMSM_SUPPORTED_CHUNK_SIZES=%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n", | |
2411 | imsm_orom_has_chunk(orom, 2) ? "2k " : "", | |
2412 | imsm_orom_has_chunk(orom, 4) ? "4k " : "", | |
2413 | imsm_orom_has_chunk(orom, 8) ? "8k " : "", | |
2414 | imsm_orom_has_chunk(orom, 16) ? "16k " : "", | |
2415 | imsm_orom_has_chunk(orom, 32) ? "32k " : "", | |
2416 | imsm_orom_has_chunk(orom, 64) ? "64k " : "", | |
2417 | imsm_orom_has_chunk(orom, 128) ? "128k " : "", | |
2418 | imsm_orom_has_chunk(orom, 256) ? "256k " : "", | |
2419 | imsm_orom_has_chunk(orom, 512) ? "512k " : "", | |
2420 | imsm_orom_has_chunk(orom, 1024*1) ? "1M " : "", | |
2421 | imsm_orom_has_chunk(orom, 1024*2) ? "2M " : "", | |
2422 | imsm_orom_has_chunk(orom, 1024*4) ? "4M " : "", | |
2423 | imsm_orom_has_chunk(orom, 1024*8) ? "8M " : "", | |
2424 | imsm_orom_has_chunk(orom, 1024*16) ? "16M " : "", | |
2425 | imsm_orom_has_chunk(orom, 1024*32) ? "32M " : "", | |
2426 | imsm_orom_has_chunk(orom, 1024*64) ? "64M " : ""); | |
2427 | printf("IMSM_2TB_VOLUMES=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB) ? "yes" : "no"); | |
2428 | printf("IMSM_2TB_DISKS=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB_DISK) ? "yes" : "no"); | |
2429 | printf("IMSM_MAX_DISKS=%d\n",orom->tds); | |
2430 | printf("IMSM_MAX_VOLUMES_PER_ARRAY=%d\n",orom->vpa); | |
2431 | printf("IMSM_MAX_VOLUMES_PER_CONTROLLER=%d\n",orom->vphba); | |
2432 | } | |
2433 | ||
9eafa1de | 2434 | static int detail_platform_imsm(int verbose, int enumerate_only, char *controller_path) |
d665cc31 DW |
2435 | { |
2436 | /* There are two components to imsm platform support, the ahci SATA | |
2437 | * controller and the option-rom. To find the SATA controller we | |
2438 | * simply look in /sys/bus/pci/drivers/ahci to see if an ahci | |
2439 | * controller with the Intel vendor id is present. This approach | |
2440 | * allows mdadm to leverage the kernel's ahci detection logic, with the | |
2441 | * caveat that if ahci.ko is not loaded mdadm will not be able to | |
2442 | * detect platform raid capabilities. The option-rom resides in a | |
2443 | * platform "Adapter ROM". We scan for its signature to retrieve the | |
2444 | * platform capabilities. If raid support is disabled in the BIOS the | |
2445 | * option-rom capability structure will not be available. | |
2446 | */ | |
d665cc31 | 2447 | struct sys_dev *list, *hba; |
d665cc31 DW |
2448 | int host_base = 0; |
2449 | int port_count = 0; | |
9eafa1de | 2450 | int result=1; |
d665cc31 | 2451 | |
5615172f | 2452 | if (enumerate_only) { |
a891a3c2 | 2453 | if (check_env("IMSM_NO_PLATFORM")) |
5615172f | 2454 | return 0; |
a891a3c2 LM |
2455 | list = find_intel_devices(); |
2456 | if (!list) | |
2457 | return 2; | |
2458 | for (hba = list; hba; hba = hba->next) { | |
6b781d33 AP |
2459 | if (find_imsm_capability(hba)) { |
2460 | result = 0; | |
a891a3c2 LM |
2461 | break; |
2462 | } | |
9eafa1de | 2463 | else |
6b781d33 | 2464 | result = 2; |
a891a3c2 | 2465 | } |
a891a3c2 | 2466 | return result; |
5615172f DW |
2467 | } |
2468 | ||
155cbb4c LM |
2469 | list = find_intel_devices(); |
2470 | if (!list) { | |
ba728be7 | 2471 | if (verbose > 0) |
7a862a02 | 2472 | pr_err("no active Intel(R) RAID controller found.\n"); |
d665cc31 | 2473 | return 2; |
ba728be7 | 2474 | } else if (verbose > 0) |
155cbb4c | 2475 | print_found_intel_controllers(list); |
d665cc31 | 2476 | |
a891a3c2 | 2477 | for (hba = list; hba; hba = hba->next) { |
0858eccf | 2478 | if (controller_path && (compare_paths(hba->path, controller_path) != 0)) |
9eafa1de | 2479 | continue; |
0858eccf | 2480 | if (!find_imsm_capability(hba)) { |
60f0f54d | 2481 | char buf[PATH_MAX]; |
e7b84f9d | 2482 | pr_err("imsm capabilities not found for controller: %s (type %s)\n", |
60f0f54d PB |
2483 | hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path, |
2484 | get_sys_dev_type(hba->type)); | |
0858eccf AP |
2485 | continue; |
2486 | } | |
2487 | result = 0; | |
2488 | } | |
2489 | ||
2490 | if (controller_path && result == 1) { | |
2491 | pr_err("no active Intel(R) RAID controller found under %s\n", | |
2492 | controller_path); | |
2493 | return result; | |
2494 | } | |
2495 | ||
5e1d6128 | 2496 | const struct orom_entry *entry; |
0858eccf | 2497 | |
5e1d6128 | 2498 | for (entry = orom_entries; entry; entry = entry->next) { |
60f0f54d | 2499 | if (entry->type == SYS_DEV_VMD) { |
07cb1e57 | 2500 | print_imsm_capability(&entry->orom); |
32716c51 PB |
2501 | printf(" 3rd party NVMe :%s supported\n", |
2502 | imsm_orom_has_tpv_support(&entry->orom)?"":" not"); | |
60f0f54d PB |
2503 | for (hba = list; hba; hba = hba->next) { |
2504 | if (hba->type == SYS_DEV_VMD) { | |
2505 | char buf[PATH_MAX]; | |
60f0f54d PB |
2506 | printf(" I/O Controller : %s (%s)\n", |
2507 | vmd_domain_to_controller(hba, buf), get_sys_dev_type(hba->type)); | |
b5eece69 PB |
2508 | if (print_vmd_attached_devs(hba)) { |
2509 | if (verbose > 0) | |
2510 | pr_err("failed to get devices attached to VMD domain.\n"); | |
2511 | result |= 2; | |
2512 | } | |
60f0f54d PB |
2513 | } |
2514 | } | |
07cb1e57 | 2515 | printf("\n"); |
60f0f54d PB |
2516 | continue; |
2517 | } | |
0858eccf | 2518 | |
60f0f54d PB |
2519 | print_imsm_capability(&entry->orom); |
2520 | if (entry->type == SYS_DEV_NVME) { | |
0858eccf AP |
2521 | for (hba = list; hba; hba = hba->next) { |
2522 | if (hba->type == SYS_DEV_NVME) | |
2523 | printf(" NVMe Device : %s\n", hba->path); | |
2524 | } | |
60f0f54d | 2525 | printf("\n"); |
0858eccf AP |
2526 | continue; |
2527 | } | |
2528 | ||
2529 | struct devid_list *devid; | |
5e1d6128 | 2530 | for (devid = entry->devid_list; devid; devid = devid->next) { |
0858eccf AP |
2531 | hba = device_by_id(devid->devid); |
2532 | if (!hba) | |
2533 | continue; | |
2534 | ||
9eafa1de MN |
2535 | printf(" I/O Controller : %s (%s)\n", |
2536 | hba->path, get_sys_dev_type(hba->type)); | |
2537 | if (hba->type == SYS_DEV_SATA) { | |
2538 | host_base = ahci_get_port_count(hba->path, &port_count); | |
2539 | if (ahci_enumerate_ports(hba->path, port_count, host_base, verbose)) { | |
2540 | if (verbose > 0) | |
7a862a02 | 2541 | pr_err("failed to enumerate ports on SATA controller at %s.\n", hba->pci_id); |
9eafa1de MN |
2542 | result |= 2; |
2543 | } | |
120dc887 LM |
2544 | } |
2545 | } | |
0858eccf | 2546 | printf("\n"); |
d665cc31 | 2547 | } |
155cbb4c | 2548 | |
120dc887 | 2549 | return result; |
d665cc31 | 2550 | } |
e50cf220 | 2551 | |
9eafa1de | 2552 | static int export_detail_platform_imsm(int verbose, char *controller_path) |
e50cf220 | 2553 | { |
e50cf220 MN |
2554 | struct sys_dev *list, *hba; |
2555 | int result=1; | |
2556 | ||
2557 | list = find_intel_devices(); | |
2558 | if (!list) { | |
2559 | if (verbose > 0) | |
2560 | pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_INTEL_DEVICES\n"); | |
2561 | result = 2; | |
e50cf220 MN |
2562 | return result; |
2563 | } | |
2564 | ||
2565 | for (hba = list; hba; hba = hba->next) { | |
9eafa1de MN |
2566 | if (controller_path && (compare_paths(hba->path,controller_path) != 0)) |
2567 | continue; | |
60f0f54d PB |
2568 | if (!find_imsm_capability(hba) && verbose > 0) { |
2569 | char buf[PATH_MAX]; | |
2570 | pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_IMSM_CAPABLE_DEVICE_UNDER_%s\n", | |
2571 | hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path); | |
2572 | } | |
0858eccf | 2573 | else |
e50cf220 | 2574 | result = 0; |
e50cf220 MN |
2575 | } |
2576 | ||
5e1d6128 | 2577 | const struct orom_entry *entry; |
0858eccf | 2578 | |
60f0f54d PB |
2579 | for (entry = orom_entries; entry; entry = entry->next) { |
2580 | if (entry->type == SYS_DEV_VMD) { | |
2581 | for (hba = list; hba; hba = hba->next) | |
2582 | print_imsm_capability_export(&entry->orom); | |
2583 | continue; | |
2584 | } | |
5e1d6128 | 2585 | print_imsm_capability_export(&entry->orom); |
60f0f54d | 2586 | } |
0858eccf | 2587 | |
e50cf220 MN |
2588 | return result; |
2589 | } | |
2590 | ||
cdddbdbc DW |
2591 | static int match_home_imsm(struct supertype *st, char *homehost) |
2592 | { | |
5115ca67 DW |
2593 | /* the imsm metadata format does not specify any host |
2594 | * identification information. We return -1 since we can never | |
2595 | * confirm nor deny whether a given array is "meant" for this | |
148acb7b | 2596 | * host. We rely on compare_super and the 'family_num' fields to |
5115ca67 DW |
2597 | * exclude member disks that do not belong, and we rely on |
2598 | * mdadm.conf to specify the arrays that should be assembled. | |
2599 | * Auto-assembly may still pick up "foreign" arrays. | |
2600 | */ | |
cdddbdbc | 2601 | |
9362c1c8 | 2602 | return -1; |
cdddbdbc DW |
2603 | } |
2604 | ||
2605 | static void uuid_from_super_imsm(struct supertype *st, int uuid[4]) | |
2606 | { | |
51006d85 N |
2607 | /* The uuid returned here is used for: |
2608 | * uuid to put into bitmap file (Create, Grow) | |
2609 | * uuid for backup header when saving critical section (Grow) | |
2610 | * comparing uuids when re-adding a device into an array | |
2611 | * In these cases the uuid required is that of the data-array, | |
2612 | * not the device-set. | |
2613 | * uuid to recognise same set when adding a missing device back | |
2614 | * to an array. This is a uuid for the device-set. | |
1011e834 | 2615 | * |
51006d85 N |
2616 | * For each of these we can make do with a truncated |
2617 | * or hashed uuid rather than the original, as long as | |
2618 | * everyone agrees. | |
2619 | * In each case the uuid required is that of the data-array, | |
2620 | * not the device-set. | |
43dad3d6 | 2621 | */ |
51006d85 N |
2622 | /* imsm does not track uuid's so we synthesis one using sha1 on |
2623 | * - The signature (Which is constant for all imsm array, but no matter) | |
148acb7b | 2624 | * - the orig_family_num of the container |
51006d85 N |
2625 | * - the index number of the volume |
2626 | * - the 'serial' number of the volume. | |
2627 | * Hopefully these are all constant. | |
2628 | */ | |
2629 | struct intel_super *super = st->sb; | |
43dad3d6 | 2630 | |
51006d85 N |
2631 | char buf[20]; |
2632 | struct sha1_ctx ctx; | |
2633 | struct imsm_dev *dev = NULL; | |
148acb7b | 2634 | __u32 family_num; |
51006d85 | 2635 | |
148acb7b DW |
2636 | /* some mdadm versions failed to set ->orig_family_num, in which |
2637 | * case fall back to ->family_num. orig_family_num will be | |
2638 | * fixed up with the first metadata update. | |
2639 | */ | |
2640 | family_num = super->anchor->orig_family_num; | |
2641 | if (family_num == 0) | |
2642 | family_num = super->anchor->family_num; | |
51006d85 | 2643 | sha1_init_ctx(&ctx); |
92bd8f8d | 2644 | sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx); |
148acb7b | 2645 | sha1_process_bytes(&family_num, sizeof(__u32), &ctx); |
51006d85 N |
2646 | if (super->current_vol >= 0) |
2647 | dev = get_imsm_dev(super, super->current_vol); | |
2648 | if (dev) { | |
2649 | __u32 vol = super->current_vol; | |
2650 | sha1_process_bytes(&vol, sizeof(vol), &ctx); | |
2651 | sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx); | |
2652 | } | |
2653 | sha1_finish_ctx(&ctx, buf); | |
2654 | memcpy(uuid, buf, 4*4); | |
cdddbdbc DW |
2655 | } |
2656 | ||
0d481d37 | 2657 | #if 0 |
4f5bc454 DW |
2658 | static void |
2659 | get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p) | |
cdddbdbc | 2660 | { |
cdddbdbc DW |
2661 | __u8 *v = get_imsm_version(mpb); |
2662 | __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH; | |
2663 | char major[] = { 0, 0, 0 }; | |
2664 | char minor[] = { 0 ,0, 0 }; | |
2665 | char patch[] = { 0, 0, 0 }; | |
2666 | char *ver_parse[] = { major, minor, patch }; | |
2667 | int i, j; | |
2668 | ||
2669 | i = j = 0; | |
2670 | while (*v != '\0' && v < end) { | |
2671 | if (*v != '.' && j < 2) | |
2672 | ver_parse[i][j++] = *v; | |
2673 | else { | |
2674 | i++; | |
2675 | j = 0; | |
2676 | } | |
2677 | v++; | |
2678 | } | |
2679 | ||
4f5bc454 DW |
2680 | *m = strtol(minor, NULL, 0); |
2681 | *p = strtol(patch, NULL, 0); | |
2682 | } | |
0d481d37 | 2683 | #endif |
4f5bc454 | 2684 | |
1e5c6983 DW |
2685 | static __u32 migr_strip_blocks_resync(struct imsm_dev *dev) |
2686 | { | |
2687 | /* migr_strip_size when repairing or initializing parity */ | |
238c0a71 | 2688 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2689 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2690 | ||
2691 | switch (get_imsm_raid_level(map)) { | |
2692 | case 5: | |
2693 | case 10: | |
2694 | return chunk; | |
2695 | default: | |
2696 | return 128*1024 >> 9; | |
2697 | } | |
2698 | } | |
2699 | ||
2700 | static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev) | |
2701 | { | |
2702 | /* migr_strip_size when rebuilding a degraded disk, no idea why | |
2703 | * this is different than migr_strip_size_resync(), but it's good | |
2704 | * to be compatible | |
2705 | */ | |
238c0a71 | 2706 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
2707 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2708 | ||
2709 | switch (get_imsm_raid_level(map)) { | |
2710 | case 1: | |
2711 | case 10: | |
2712 | if (map->num_members % map->num_domains == 0) | |
2713 | return 128*1024 >> 9; | |
2714 | else | |
2715 | return chunk; | |
2716 | case 5: | |
2717 | return max((__u32) 64*1024 >> 9, chunk); | |
2718 | default: | |
2719 | return 128*1024 >> 9; | |
2720 | } | |
2721 | } | |
2722 | ||
2723 | static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev) | |
2724 | { | |
238c0a71 AK |
2725 | struct imsm_map *lo = get_imsm_map(dev, MAP_0); |
2726 | struct imsm_map *hi = get_imsm_map(dev, MAP_1); | |
1e5c6983 DW |
2727 | __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip); |
2728 | __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip); | |
2729 | ||
2730 | return max((__u32) 1, hi_chunk / lo_chunk); | |
2731 | } | |
2732 | ||
2733 | static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev) | |
2734 | { | |
238c0a71 | 2735 | struct imsm_map *lo = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2736 | int level = get_imsm_raid_level(lo); |
2737 | ||
2738 | if (level == 1 || level == 10) { | |
238c0a71 | 2739 | struct imsm_map *hi = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
2740 | |
2741 | return hi->num_domains; | |
2742 | } else | |
2743 | return num_stripes_per_unit_resync(dev); | |
2744 | } | |
2745 | ||
98130f40 | 2746 | static __u8 imsm_num_data_members(struct imsm_dev *dev, int second_map) |
1e5c6983 DW |
2747 | { |
2748 | /* named 'imsm_' because raid0, raid1 and raid10 | |
2749 | * counter-intuitively have the same number of data disks | |
2750 | */ | |
98130f40 | 2751 | struct imsm_map *map = get_imsm_map(dev, second_map); |
1e5c6983 DW |
2752 | |
2753 | switch (get_imsm_raid_level(map)) { | |
2754 | case 0: | |
36fd8ccc AK |
2755 | return map->num_members; |
2756 | break; | |
1e5c6983 DW |
2757 | case 1: |
2758 | case 10: | |
36fd8ccc | 2759 | return map->num_members/2; |
1e5c6983 DW |
2760 | case 5: |
2761 | return map->num_members - 1; | |
2762 | default: | |
1ade5cc1 | 2763 | dprintf("unsupported raid level\n"); |
1e5c6983 DW |
2764 | return 0; |
2765 | } | |
2766 | } | |
2767 | ||
2768 | static __u32 parity_segment_depth(struct imsm_dev *dev) | |
2769 | { | |
238c0a71 | 2770 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2771 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2772 | ||
2773 | switch(get_imsm_raid_level(map)) { | |
2774 | case 1: | |
2775 | case 10: | |
2776 | return chunk * map->num_domains; | |
2777 | case 5: | |
2778 | return chunk * map->num_members; | |
2779 | default: | |
2780 | return chunk; | |
2781 | } | |
2782 | } | |
2783 | ||
2784 | static __u32 map_migr_block(struct imsm_dev *dev, __u32 block) | |
2785 | { | |
238c0a71 | 2786 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
2787 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2788 | __u32 strip = block / chunk; | |
2789 | ||
2790 | switch (get_imsm_raid_level(map)) { | |
2791 | case 1: | |
2792 | case 10: { | |
2793 | __u32 vol_strip = (strip * map->num_domains) + 1; | |
2794 | __u32 vol_stripe = vol_strip / map->num_members; | |
2795 | ||
2796 | return vol_stripe * chunk + block % chunk; | |
2797 | } case 5: { | |
2798 | __u32 stripe = strip / (map->num_members - 1); | |
2799 | ||
2800 | return stripe * chunk + block % chunk; | |
2801 | } | |
2802 | default: | |
2803 | return 0; | |
2804 | } | |
2805 | } | |
2806 | ||
c47b0ff6 AK |
2807 | static __u64 blocks_per_migr_unit(struct intel_super *super, |
2808 | struct imsm_dev *dev) | |
1e5c6983 DW |
2809 | { |
2810 | /* calculate the conversion factor between per member 'blocks' | |
2811 | * (md/{resync,rebuild}_start) and imsm migration units, return | |
2812 | * 0 for the 'not migrating' and 'unsupported migration' cases | |
2813 | */ | |
2814 | if (!dev->vol.migr_state) | |
2815 | return 0; | |
2816 | ||
2817 | switch (migr_type(dev)) { | |
c47b0ff6 AK |
2818 | case MIGR_GEN_MIGR: { |
2819 | struct migr_record *migr_rec = super->migr_rec; | |
2820 | return __le32_to_cpu(migr_rec->blocks_per_unit); | |
2821 | } | |
1e5c6983 DW |
2822 | case MIGR_VERIFY: |
2823 | case MIGR_REPAIR: | |
2824 | case MIGR_INIT: { | |
238c0a71 | 2825 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2826 | __u32 stripes_per_unit; |
2827 | __u32 blocks_per_unit; | |
2828 | __u32 parity_depth; | |
2829 | __u32 migr_chunk; | |
2830 | __u32 block_map; | |
2831 | __u32 block_rel; | |
2832 | __u32 segment; | |
2833 | __u32 stripe; | |
2834 | __u8 disks; | |
2835 | ||
2836 | /* yes, this is really the translation of migr_units to | |
2837 | * per-member blocks in the 'resync' case | |
2838 | */ | |
2839 | stripes_per_unit = num_stripes_per_unit_resync(dev); | |
2840 | migr_chunk = migr_strip_blocks_resync(dev); | |
238c0a71 | 2841 | disks = imsm_num_data_members(dev, MAP_0); |
1e5c6983 | 2842 | blocks_per_unit = stripes_per_unit * migr_chunk * disks; |
7b1ab482 | 2843 | stripe = __le16_to_cpu(map->blocks_per_strip) * disks; |
1e5c6983 DW |
2844 | segment = blocks_per_unit / stripe; |
2845 | block_rel = blocks_per_unit - segment * stripe; | |
2846 | parity_depth = parity_segment_depth(dev); | |
2847 | block_map = map_migr_block(dev, block_rel); | |
2848 | return block_map + parity_depth * segment; | |
2849 | } | |
2850 | case MIGR_REBUILD: { | |
2851 | __u32 stripes_per_unit; | |
2852 | __u32 migr_chunk; | |
2853 | ||
2854 | stripes_per_unit = num_stripes_per_unit_rebuild(dev); | |
2855 | migr_chunk = migr_strip_blocks_rebuild(dev); | |
2856 | return migr_chunk * stripes_per_unit; | |
2857 | } | |
1e5c6983 DW |
2858 | case MIGR_STATE_CHANGE: |
2859 | default: | |
2860 | return 0; | |
2861 | } | |
2862 | } | |
2863 | ||
c2c087e6 DW |
2864 | static int imsm_level_to_layout(int level) |
2865 | { | |
2866 | switch (level) { | |
2867 | case 0: | |
2868 | case 1: | |
2869 | return 0; | |
2870 | case 5: | |
2871 | case 6: | |
a380c027 | 2872 | return ALGORITHM_LEFT_ASYMMETRIC; |
c2c087e6 | 2873 | case 10: |
c92a2527 | 2874 | return 0x102; |
c2c087e6 | 2875 | } |
a18a888e | 2876 | return UnSet; |
c2c087e6 DW |
2877 | } |
2878 | ||
8e59f3d8 AK |
2879 | /******************************************************************************* |
2880 | * Function: read_imsm_migr_rec | |
2881 | * Description: Function reads imsm migration record from last sector of disk | |
2882 | * Parameters: | |
2883 | * fd : disk descriptor | |
2884 | * super : metadata info | |
2885 | * Returns: | |
2886 | * 0 : success, | |
2887 | * -1 : fail | |
2888 | ******************************************************************************/ | |
2889 | static int read_imsm_migr_rec(int fd, struct intel_super *super) | |
2890 | { | |
2891 | int ret_val = -1; | |
de44e46f | 2892 | unsigned int sector_size = super->sector_size; |
8e59f3d8 AK |
2893 | unsigned long long dsize; |
2894 | ||
2895 | get_dev_size(fd, NULL, &dsize); | |
de44e46f PB |
2896 | if (lseek64(fd, dsize - (sector_size*MIGR_REC_SECTOR_POSITION), |
2897 | SEEK_SET) < 0) { | |
e7b84f9d N |
2898 | pr_err("Cannot seek to anchor block: %s\n", |
2899 | strerror(errno)); | |
8e59f3d8 AK |
2900 | goto out; |
2901 | } | |
466070ad | 2902 | if ((unsigned int)read(fd, super->migr_rec_buf, |
de44e46f PB |
2903 | MIGR_REC_BUF_SECTORS*sector_size) != |
2904 | MIGR_REC_BUF_SECTORS*sector_size) { | |
e7b84f9d N |
2905 | pr_err("Cannot read migr record block: %s\n", |
2906 | strerror(errno)); | |
8e59f3d8 AK |
2907 | goto out; |
2908 | } | |
2909 | ret_val = 0; | |
de44e46f PB |
2910 | if (sector_size == 4096) |
2911 | convert_from_4k_imsm_migr_rec(super); | |
8e59f3d8 AK |
2912 | |
2913 | out: | |
2914 | return ret_val; | |
2915 | } | |
2916 | ||
3136abe5 AK |
2917 | static struct imsm_dev *imsm_get_device_during_migration( |
2918 | struct intel_super *super) | |
2919 | { | |
2920 | ||
2921 | struct intel_dev *dv; | |
2922 | ||
2923 | for (dv = super->devlist; dv; dv = dv->next) { | |
2924 | if (is_gen_migration(dv->dev)) | |
2925 | return dv->dev; | |
2926 | } | |
2927 | return NULL; | |
2928 | } | |
2929 | ||
8e59f3d8 AK |
2930 | /******************************************************************************* |
2931 | * Function: load_imsm_migr_rec | |
2932 | * Description: Function reads imsm migration record (it is stored at the last | |
2933 | * sector of disk) | |
2934 | * Parameters: | |
2935 | * super : imsm internal array info | |
2936 | * info : general array info | |
2937 | * Returns: | |
2938 | * 0 : success | |
2939 | * -1 : fail | |
4c965cc9 | 2940 | * -2 : no migration in progress |
8e59f3d8 AK |
2941 | ******************************************************************************/ |
2942 | static int load_imsm_migr_rec(struct intel_super *super, struct mdinfo *info) | |
2943 | { | |
2944 | struct mdinfo *sd; | |
594dc1b8 | 2945 | struct dl *dl; |
8e59f3d8 AK |
2946 | char nm[30]; |
2947 | int retval = -1; | |
2948 | int fd = -1; | |
3136abe5 | 2949 | struct imsm_dev *dev; |
594dc1b8 | 2950 | struct imsm_map *map; |
b4ab44d8 | 2951 | int slot = -1; |
3136abe5 AK |
2952 | |
2953 | /* find map under migration */ | |
2954 | dev = imsm_get_device_during_migration(super); | |
2955 | /* nothing to load,no migration in progress? | |
2956 | */ | |
2957 | if (dev == NULL) | |
4c965cc9 | 2958 | return -2; |
8e59f3d8 AK |
2959 | |
2960 | if (info) { | |
2961 | for (sd = info->devs ; sd ; sd = sd->next) { | |
2962 | /* read only from one of the first two slots */ | |
12fe93e9 TM |
2963 | if ((sd->disk.raid_disk < 0) || |
2964 | (sd->disk.raid_disk > 1)) | |
8e59f3d8 | 2965 | continue; |
3136abe5 | 2966 | |
8e59f3d8 AK |
2967 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); |
2968 | fd = dev_open(nm, O_RDONLY); | |
2969 | if (fd >= 0) | |
2970 | break; | |
2971 | } | |
2972 | } | |
2973 | if (fd < 0) { | |
12fe93e9 | 2974 | map = get_imsm_map(dev, MAP_0); |
8e59f3d8 | 2975 | for (dl = super->disks; dl; dl = dl->next) { |
3136abe5 AK |
2976 | /* skip spare and failed disks |
2977 | */ | |
2978 | if (dl->index < 0) | |
2979 | continue; | |
8e59f3d8 | 2980 | /* read only from one of the first two slots */ |
3136abe5 AK |
2981 | if (map) |
2982 | slot = get_imsm_disk_slot(map, dl->index); | |
089f9d79 | 2983 | if (map == NULL || slot > 1 || slot < 0) |
8e59f3d8 AK |
2984 | continue; |
2985 | sprintf(nm, "%d:%d", dl->major, dl->minor); | |
2986 | fd = dev_open(nm, O_RDONLY); | |
2987 | if (fd >= 0) | |
2988 | break; | |
2989 | } | |
2990 | } | |
2991 | if (fd < 0) | |
2992 | goto out; | |
2993 | retval = read_imsm_migr_rec(fd, super); | |
2994 | ||
2995 | out: | |
2996 | if (fd >= 0) | |
2997 | close(fd); | |
2998 | return retval; | |
2999 | } | |
3000 | ||
c17608ea AK |
3001 | /******************************************************************************* |
3002 | * function: imsm_create_metadata_checkpoint_update | |
3003 | * Description: It creates update for checkpoint change. | |
3004 | * Parameters: | |
3005 | * super : imsm internal array info | |
3006 | * u : pointer to prepared update | |
3007 | * Returns: | |
3008 | * Uptate length. | |
3009 | * If length is equal to 0, input pointer u contains no update | |
3010 | ******************************************************************************/ | |
3011 | static int imsm_create_metadata_checkpoint_update( | |
3012 | struct intel_super *super, | |
3013 | struct imsm_update_general_migration_checkpoint **u) | |
3014 | { | |
3015 | ||
3016 | int update_memory_size = 0; | |
3017 | ||
1ade5cc1 | 3018 | dprintf("(enter)\n"); |
c17608ea AK |
3019 | |
3020 | if (u == NULL) | |
3021 | return 0; | |
3022 | *u = NULL; | |
3023 | ||
3024 | /* size of all update data without anchor */ | |
3025 | update_memory_size = | |
3026 | sizeof(struct imsm_update_general_migration_checkpoint); | |
3027 | ||
503975b9 | 3028 | *u = xcalloc(1, update_memory_size); |
c17608ea | 3029 | if (*u == NULL) { |
1ade5cc1 | 3030 | dprintf("error: cannot get memory\n"); |
c17608ea AK |
3031 | return 0; |
3032 | } | |
3033 | (*u)->type = update_general_migration_checkpoint; | |
3034 | (*u)->curr_migr_unit = __le32_to_cpu(super->migr_rec->curr_migr_unit); | |
1ade5cc1 | 3035 | dprintf("prepared for %u\n", (*u)->curr_migr_unit); |
c17608ea AK |
3036 | |
3037 | return update_memory_size; | |
3038 | } | |
3039 | ||
c17608ea AK |
3040 | static void imsm_update_metadata_locally(struct supertype *st, |
3041 | void *buf, int len); | |
3042 | ||
687629c2 AK |
3043 | /******************************************************************************* |
3044 | * Function: write_imsm_migr_rec | |
3045 | * Description: Function writes imsm migration record | |
3046 | * (at the last sector of disk) | |
3047 | * Parameters: | |
3048 | * super : imsm internal array info | |
3049 | * Returns: | |
3050 | * 0 : success | |
3051 | * -1 : if fail | |
3052 | ******************************************************************************/ | |
3053 | static int write_imsm_migr_rec(struct supertype *st) | |
3054 | { | |
3055 | struct intel_super *super = st->sb; | |
de44e46f | 3056 | unsigned int sector_size = super->sector_size; |
687629c2 AK |
3057 | unsigned long long dsize; |
3058 | char nm[30]; | |
3059 | int fd = -1; | |
3060 | int retval = -1; | |
3061 | struct dl *sd; | |
c17608ea AK |
3062 | int len; |
3063 | struct imsm_update_general_migration_checkpoint *u; | |
3136abe5 | 3064 | struct imsm_dev *dev; |
594dc1b8 | 3065 | struct imsm_map *map; |
3136abe5 AK |
3066 | |
3067 | /* find map under migration */ | |
3068 | dev = imsm_get_device_during_migration(super); | |
3069 | /* if no migration, write buffer anyway to clear migr_record | |
3070 | * on disk based on first available device | |
3071 | */ | |
3072 | if (dev == NULL) | |
3073 | dev = get_imsm_dev(super, super->current_vol < 0 ? 0 : | |
3074 | super->current_vol); | |
3075 | ||
44bfe6df | 3076 | map = get_imsm_map(dev, MAP_0); |
687629c2 | 3077 | |
de44e46f PB |
3078 | if (sector_size == 4096) |
3079 | convert_to_4k_imsm_migr_rec(super); | |
687629c2 | 3080 | for (sd = super->disks ; sd ; sd = sd->next) { |
b4ab44d8 | 3081 | int slot = -1; |
3136abe5 AK |
3082 | |
3083 | /* skip failed and spare devices */ | |
3084 | if (sd->index < 0) | |
3085 | continue; | |
687629c2 | 3086 | /* write to 2 first slots only */ |
3136abe5 AK |
3087 | if (map) |
3088 | slot = get_imsm_disk_slot(map, sd->index); | |
089f9d79 | 3089 | if (map == NULL || slot > 1 || slot < 0) |
687629c2 | 3090 | continue; |
3136abe5 | 3091 | |
687629c2 AK |
3092 | sprintf(nm, "%d:%d", sd->major, sd->minor); |
3093 | fd = dev_open(nm, O_RDWR); | |
3094 | if (fd < 0) | |
3095 | continue; | |
3096 | get_dev_size(fd, NULL, &dsize); | |
de44e46f PB |
3097 | if (lseek64(fd, dsize - (MIGR_REC_SECTOR_POSITION*sector_size), |
3098 | SEEK_SET) < 0) { | |
e7b84f9d N |
3099 | pr_err("Cannot seek to anchor block: %s\n", |
3100 | strerror(errno)); | |
687629c2 AK |
3101 | goto out; |
3102 | } | |
466070ad | 3103 | if ((unsigned int)write(fd, super->migr_rec_buf, |
de44e46f PB |
3104 | MIGR_REC_BUF_SECTORS*sector_size) != |
3105 | MIGR_REC_BUF_SECTORS*sector_size) { | |
e7b84f9d N |
3106 | pr_err("Cannot write migr record block: %s\n", |
3107 | strerror(errno)); | |
687629c2 AK |
3108 | goto out; |
3109 | } | |
3110 | close(fd); | |
3111 | fd = -1; | |
3112 | } | |
de44e46f PB |
3113 | if (sector_size == 4096) |
3114 | convert_from_4k_imsm_migr_rec(super); | |
c17608ea AK |
3115 | /* update checkpoint information in metadata */ |
3116 | len = imsm_create_metadata_checkpoint_update(super, &u); | |
c17608ea AK |
3117 | if (len <= 0) { |
3118 | dprintf("imsm: Cannot prepare update\n"); | |
3119 | goto out; | |
3120 | } | |
3121 | /* update metadata locally */ | |
3122 | imsm_update_metadata_locally(st, u, len); | |
3123 | /* and possibly remotely */ | |
3124 | if (st->update_tail) { | |
3125 | append_metadata_update(st, u, len); | |
3126 | /* during reshape we do all work inside metadata handler | |
3127 | * manage_reshape(), so metadata update has to be triggered | |
3128 | * insida it | |
3129 | */ | |
3130 | flush_metadata_updates(st); | |
3131 | st->update_tail = &st->updates; | |
3132 | } else | |
3133 | free(u); | |
687629c2 AK |
3134 | |
3135 | retval = 0; | |
3136 | out: | |
3137 | if (fd >= 0) | |
3138 | close(fd); | |
3139 | return retval; | |
3140 | } | |
3141 | ||
e2962bfc AK |
3142 | /* spare/missing disks activations are not allowe when |
3143 | * array/container performs reshape operation, because | |
3144 | * all arrays in container works on the same disks set | |
3145 | */ | |
3146 | int imsm_reshape_blocks_arrays_changes(struct intel_super *super) | |
3147 | { | |
3148 | int rv = 0; | |
3149 | struct intel_dev *i_dev; | |
3150 | struct imsm_dev *dev; | |
3151 | ||
3152 | /* check whole container | |
3153 | */ | |
3154 | for (i_dev = super->devlist; i_dev; i_dev = i_dev->next) { | |
3155 | dev = i_dev->dev; | |
3ad25638 | 3156 | if (is_gen_migration(dev)) { |
e2962bfc AK |
3157 | /* No repair during any migration in container |
3158 | */ | |
3159 | rv = 1; | |
3160 | break; | |
3161 | } | |
3162 | } | |
3163 | return rv; | |
3164 | } | |
c41e00b2 AK |
3165 | static unsigned long long imsm_component_size_aligment_check(int level, |
3166 | int chunk_size, | |
f36a9ecd | 3167 | unsigned int sector_size, |
c41e00b2 AK |
3168 | unsigned long long component_size) |
3169 | { | |
3170 | unsigned int component_size_alligment; | |
3171 | ||
3172 | /* check component size aligment | |
3173 | */ | |
f36a9ecd | 3174 | component_size_alligment = component_size % (chunk_size/sector_size); |
c41e00b2 | 3175 | |
1ade5cc1 | 3176 | dprintf("(Level: %i, chunk_size = %i, component_size = %llu), component_size_alligment = %u\n", |
c41e00b2 AK |
3177 | level, chunk_size, component_size, |
3178 | component_size_alligment); | |
3179 | ||
3180 | if (component_size_alligment && (level != 1) && (level != UnSet)) { | |
3181 | dprintf("imsm: reported component size alligned from %llu ", | |
3182 | component_size); | |
3183 | component_size -= component_size_alligment; | |
1ade5cc1 | 3184 | dprintf_cont("to %llu (%i).\n", |
c41e00b2 AK |
3185 | component_size, component_size_alligment); |
3186 | } | |
3187 | ||
3188 | return component_size; | |
3189 | } | |
e2962bfc | 3190 | |
2432ce9b AP |
3191 | static unsigned long long get_ppl_sector(struct intel_super *super, int dev_idx) |
3192 | { | |
3193 | struct imsm_dev *dev = get_imsm_dev(super, dev_idx); | |
3194 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
3195 | ||
3196 | return pba_of_lba0(map) + | |
3197 | (num_data_stripes(map) * map->blocks_per_strip); | |
3198 | } | |
3199 | ||
a5d85af7 | 3200 | static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap) |
bf5a934a DW |
3201 | { |
3202 | struct intel_super *super = st->sb; | |
c47b0ff6 | 3203 | struct migr_record *migr_rec = super->migr_rec; |
949c47a0 | 3204 | struct imsm_dev *dev = get_imsm_dev(super, super->current_vol); |
238c0a71 AK |
3205 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
3206 | struct imsm_map *prev_map = get_imsm_map(dev, MAP_1); | |
b335e593 | 3207 | struct imsm_map *map_to_analyse = map; |
efb30e7f | 3208 | struct dl *dl; |
a5d85af7 | 3209 | int map_disks = info->array.raid_disks; |
bf5a934a | 3210 | |
95eeceeb | 3211 | memset(info, 0, sizeof(*info)); |
b335e593 AK |
3212 | if (prev_map) |
3213 | map_to_analyse = prev_map; | |
3214 | ||
ca0748fa | 3215 | dl = super->current_disk; |
9894ec0d | 3216 | |
bf5a934a | 3217 | info->container_member = super->current_vol; |
cd0430a1 | 3218 | info->array.raid_disks = map->num_members; |
b335e593 | 3219 | info->array.level = get_imsm_raid_level(map_to_analyse); |
bf5a934a DW |
3220 | info->array.layout = imsm_level_to_layout(info->array.level); |
3221 | info->array.md_minor = -1; | |
3222 | info->array.ctime = 0; | |
3223 | info->array.utime = 0; | |
b335e593 AK |
3224 | info->array.chunk_size = |
3225 | __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9; | |
2432ce9b | 3226 | info->array.state = !(dev->vol.dirty & RAIDVOL_DIRTY); |
da9b4a62 DW |
3227 | info->custom_array_size = __le32_to_cpu(dev->size_high); |
3228 | info->custom_array_size <<= 32; | |
3229 | info->custom_array_size |= __le32_to_cpu(dev->size_low); | |
3ad25638 AK |
3230 | info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb); |
3231 | ||
3f510843 | 3232 | if (is_gen_migration(dev)) { |
3f83228a | 3233 | info->reshape_active = 1; |
b335e593 AK |
3234 | info->new_level = get_imsm_raid_level(map); |
3235 | info->new_layout = imsm_level_to_layout(info->new_level); | |
3236 | info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9; | |
3f83228a | 3237 | info->delta_disks = map->num_members - prev_map->num_members; |
493f5dd6 N |
3238 | if (info->delta_disks) { |
3239 | /* this needs to be applied to every array | |
3240 | * in the container. | |
3241 | */ | |
81219e70 | 3242 | info->reshape_active = CONTAINER_RESHAPE; |
493f5dd6 | 3243 | } |
3f83228a N |
3244 | /* We shape information that we give to md might have to be |
3245 | * modify to cope with md's requirement for reshaping arrays. | |
3246 | * For example, when reshaping a RAID0, md requires it to be | |
3247 | * presented as a degraded RAID4. | |
3248 | * Also if a RAID0 is migrating to a RAID5 we need to specify | |
3249 | * the array as already being RAID5, but the 'before' layout | |
3250 | * is a RAID4-like layout. | |
3251 | */ | |
3252 | switch (info->array.level) { | |
3253 | case 0: | |
3254 | switch(info->new_level) { | |
3255 | case 0: | |
3256 | /* conversion is happening as RAID4 */ | |
3257 | info->array.level = 4; | |
3258 | info->array.raid_disks += 1; | |
3259 | break; | |
3260 | case 5: | |
3261 | /* conversion is happening as RAID5 */ | |
3262 | info->array.level = 5; | |
3263 | info->array.layout = ALGORITHM_PARITY_N; | |
3f83228a N |
3264 | info->delta_disks -= 1; |
3265 | break; | |
3266 | default: | |
3267 | /* FIXME error message */ | |
3268 | info->array.level = UnSet; | |
3269 | break; | |
3270 | } | |
3271 | break; | |
3272 | } | |
b335e593 AK |
3273 | } else { |
3274 | info->new_level = UnSet; | |
3275 | info->new_layout = UnSet; | |
3276 | info->new_chunk = info->array.chunk_size; | |
3f83228a | 3277 | info->delta_disks = 0; |
b335e593 | 3278 | } |
ca0748fa | 3279 | |
efb30e7f DW |
3280 | if (dl) { |
3281 | info->disk.major = dl->major; | |
3282 | info->disk.minor = dl->minor; | |
ca0748fa | 3283 | info->disk.number = dl->index; |
656b6b5a N |
3284 | info->disk.raid_disk = get_imsm_disk_slot(map_to_analyse, |
3285 | dl->index); | |
efb30e7f | 3286 | } |
bf5a934a | 3287 | |
5551b113 | 3288 | info->data_offset = pba_of_lba0(map_to_analyse); |
06fb291a PB |
3289 | |
3290 | if (info->array.level == 5) { | |
3291 | info->component_size = num_data_stripes(map_to_analyse) * | |
3292 | map_to_analyse->blocks_per_strip; | |
3293 | } else { | |
3294 | info->component_size = blocks_per_member(map_to_analyse); | |
3295 | } | |
139dae11 | 3296 | |
c41e00b2 AK |
3297 | info->component_size = imsm_component_size_aligment_check( |
3298 | info->array.level, | |
3299 | info->array.chunk_size, | |
f36a9ecd | 3300 | super->sector_size, |
c41e00b2 | 3301 | info->component_size); |
5e46202e | 3302 | info->bb.supported = 1; |
139dae11 | 3303 | |
301406c9 | 3304 | memset(info->uuid, 0, sizeof(info->uuid)); |
921d9e16 | 3305 | info->recovery_start = MaxSector; |
bf5a934a | 3306 | |
c2462068 PB |
3307 | if (info->array.level == 5 && |
3308 | (dev->rwh_policy == RWH_DISTRIBUTED || | |
3309 | dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)) { | |
2432ce9b AP |
3310 | info->consistency_policy = CONSISTENCY_POLICY_PPL; |
3311 | info->ppl_sector = get_ppl_sector(super, super->current_vol); | |
c2462068 PB |
3312 | if (dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED) |
3313 | info->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9; | |
3314 | else | |
3315 | info->ppl_size = (PPL_HEADER_SIZE + PPL_ENTRY_SPACE) | |
3316 | >> 9; | |
2432ce9b AP |
3317 | } else if (info->array.level <= 0) { |
3318 | info->consistency_policy = CONSISTENCY_POLICY_NONE; | |
3319 | } else { | |
3320 | info->consistency_policy = CONSISTENCY_POLICY_RESYNC; | |
3321 | } | |
3322 | ||
d2e6d5d6 | 3323 | info->reshape_progress = 0; |
b6796ce1 | 3324 | info->resync_start = MaxSector; |
b9172665 | 3325 | if ((map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED || |
2432ce9b | 3326 | !(info->array.state & 1)) && |
b9172665 | 3327 | imsm_reshape_blocks_arrays_changes(super) == 0) { |
301406c9 | 3328 | info->resync_start = 0; |
b6796ce1 AK |
3329 | } |
3330 | if (dev->vol.migr_state) { | |
1e5c6983 DW |
3331 | switch (migr_type(dev)) { |
3332 | case MIGR_REPAIR: | |
3333 | case MIGR_INIT: { | |
c47b0ff6 AK |
3334 | __u64 blocks_per_unit = blocks_per_migr_unit(super, |
3335 | dev); | |
1e5c6983 DW |
3336 | __u64 units = __le32_to_cpu(dev->vol.curr_migr_unit); |
3337 | ||
3338 | info->resync_start = blocks_per_unit * units; | |
3339 | break; | |
3340 | } | |
d2e6d5d6 | 3341 | case MIGR_GEN_MIGR: { |
c47b0ff6 AK |
3342 | __u64 blocks_per_unit = blocks_per_migr_unit(super, |
3343 | dev); | |
3344 | __u64 units = __le32_to_cpu(migr_rec->curr_migr_unit); | |
04fa9523 AK |
3345 | unsigned long long array_blocks; |
3346 | int used_disks; | |
d2e6d5d6 | 3347 | |
befb629b AK |
3348 | if (__le32_to_cpu(migr_rec->ascending_migr) && |
3349 | (units < | |
3350 | (__le32_to_cpu(migr_rec->num_migr_units)-1)) && | |
3351 | (super->migr_rec->rec_status == | |
3352 | __cpu_to_le32(UNIT_SRC_IN_CP_AREA))) | |
3353 | units++; | |
3354 | ||
d2e6d5d6 | 3355 | info->reshape_progress = blocks_per_unit * units; |
6289d1e0 | 3356 | |
7a862a02 | 3357 | dprintf("IMSM: General Migration checkpoint : %llu (%llu) -> read reshape progress : %llu\n", |
19986c72 MB |
3358 | (unsigned long long)units, |
3359 | (unsigned long long)blocks_per_unit, | |
3360 | info->reshape_progress); | |
75156c46 | 3361 | |
238c0a71 | 3362 | used_disks = imsm_num_data_members(dev, MAP_1); |
75156c46 | 3363 | if (used_disks > 0) { |
5551b113 | 3364 | array_blocks = blocks_per_member(map) * |
75156c46 | 3365 | used_disks; |
b53bfba6 TM |
3366 | info->custom_array_size = |
3367 | round_size_to_mb(array_blocks, | |
3368 | used_disks); | |
3369 | ||
75156c46 | 3370 | } |
d2e6d5d6 | 3371 | } |
1e5c6983 DW |
3372 | case MIGR_VERIFY: |
3373 | /* we could emulate the checkpointing of | |
3374 | * 'sync_action=check' migrations, but for now | |
3375 | * we just immediately complete them | |
3376 | */ | |
3377 | case MIGR_REBUILD: | |
3378 | /* this is handled by container_content_imsm() */ | |
1e5c6983 DW |
3379 | case MIGR_STATE_CHANGE: |
3380 | /* FIXME handle other migrations */ | |
3381 | default: | |
3382 | /* we are not dirty, so... */ | |
3383 | info->resync_start = MaxSector; | |
3384 | } | |
b6796ce1 | 3385 | } |
301406c9 DW |
3386 | |
3387 | strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN); | |
3388 | info->name[MAX_RAID_SERIAL_LEN] = 0; | |
bf5a934a | 3389 | |
f35f2525 N |
3390 | info->array.major_version = -1; |
3391 | info->array.minor_version = -2; | |
4dd2df09 | 3392 | sprintf(info->text_version, "/%s/%d", st->container_devnm, info->container_member); |
a67dd8cc | 3393 | info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */ |
51006d85 | 3394 | uuid_from_super_imsm(st, info->uuid); |
a5d85af7 N |
3395 | |
3396 | if (dmap) { | |
3397 | int i, j; | |
3398 | for (i=0; i<map_disks; i++) { | |
3399 | dmap[i] = 0; | |
3400 | if (i < info->array.raid_disks) { | |
3401 | struct imsm_disk *dsk; | |
238c0a71 | 3402 | j = get_imsm_disk_idx(dev, i, MAP_X); |
a5d85af7 N |
3403 | dsk = get_imsm_disk(super, j); |
3404 | if (dsk && (dsk->status & CONFIGURED_DISK)) | |
3405 | dmap[i] = 1; | |
3406 | } | |
3407 | } | |
3408 | } | |
81ac8b4d | 3409 | } |
bf5a934a | 3410 | |
3b451610 AK |
3411 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, |
3412 | int failed, int look_in_map); | |
3413 | ||
3414 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev, | |
3415 | int look_in_map); | |
3416 | ||
3417 | static void manage_second_map(struct intel_super *super, struct imsm_dev *dev) | |
3418 | { | |
3419 | if (is_gen_migration(dev)) { | |
3420 | int failed; | |
3421 | __u8 map_state; | |
3422 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); | |
3423 | ||
3424 | failed = imsm_count_failed(super, dev, MAP_1); | |
238c0a71 | 3425 | map_state = imsm_check_degraded(super, dev, failed, MAP_1); |
3b451610 AK |
3426 | if (map2->map_state != map_state) { |
3427 | map2->map_state = map_state; | |
3428 | super->updates_pending++; | |
3429 | } | |
3430 | } | |
3431 | } | |
97b4d0e9 DW |
3432 | |
3433 | static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index) | |
3434 | { | |
3435 | struct dl *d; | |
3436 | ||
3437 | for (d = super->missing; d; d = d->next) | |
3438 | if (d->index == index) | |
3439 | return &d->disk; | |
3440 | return NULL; | |
3441 | } | |
3442 | ||
a5d85af7 | 3443 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map) |
4f5bc454 DW |
3444 | { |
3445 | struct intel_super *super = st->sb; | |
4f5bc454 | 3446 | struct imsm_disk *disk; |
a5d85af7 | 3447 | int map_disks = info->array.raid_disks; |
ab3cb6b3 N |
3448 | int max_enough = -1; |
3449 | int i; | |
3450 | struct imsm_super *mpb; | |
4f5bc454 | 3451 | |
bf5a934a | 3452 | if (super->current_vol >= 0) { |
a5d85af7 | 3453 | getinfo_super_imsm_volume(st, info, map); |
bf5a934a DW |
3454 | return; |
3455 | } | |
95eeceeb | 3456 | memset(info, 0, sizeof(*info)); |
d23fe947 DW |
3457 | |
3458 | /* Set raid_disks to zero so that Assemble will always pull in valid | |
3459 | * spares | |
3460 | */ | |
3461 | info->array.raid_disks = 0; | |
cdddbdbc DW |
3462 | info->array.level = LEVEL_CONTAINER; |
3463 | info->array.layout = 0; | |
3464 | info->array.md_minor = -1; | |
1011e834 | 3465 | info->array.ctime = 0; /* N/A for imsm */ |
cdddbdbc DW |
3466 | info->array.utime = 0; |
3467 | info->array.chunk_size = 0; | |
3468 | ||
3469 | info->disk.major = 0; | |
3470 | info->disk.minor = 0; | |
cdddbdbc | 3471 | info->disk.raid_disk = -1; |
c2c087e6 | 3472 | info->reshape_active = 0; |
f35f2525 N |
3473 | info->array.major_version = -1; |
3474 | info->array.minor_version = -2; | |
c2c087e6 | 3475 | strcpy(info->text_version, "imsm"); |
a67dd8cc | 3476 | info->safe_mode_delay = 0; |
c2c087e6 DW |
3477 | info->disk.number = -1; |
3478 | info->disk.state = 0; | |
c5afc314 | 3479 | info->name[0] = 0; |
921d9e16 | 3480 | info->recovery_start = MaxSector; |
3ad25638 | 3481 | info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb); |
5e46202e | 3482 | info->bb.supported = 1; |
c2c087e6 | 3483 | |
97b4d0e9 | 3484 | /* do we have the all the insync disks that we expect? */ |
ab3cb6b3 | 3485 | mpb = super->anchor; |
b7d81a38 | 3486 | info->events = __le32_to_cpu(mpb->generation_num); |
97b4d0e9 | 3487 | |
ab3cb6b3 N |
3488 | for (i = 0; i < mpb->num_raid_devs; i++) { |
3489 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
3490 | int failed, enough, j, missing = 0; | |
3491 | struct imsm_map *map; | |
3492 | __u8 state; | |
97b4d0e9 | 3493 | |
3b451610 AK |
3494 | failed = imsm_count_failed(super, dev, MAP_0); |
3495 | state = imsm_check_degraded(super, dev, failed, MAP_0); | |
238c0a71 | 3496 | map = get_imsm_map(dev, MAP_0); |
ab3cb6b3 N |
3497 | |
3498 | /* any newly missing disks? | |
3499 | * (catches single-degraded vs double-degraded) | |
3500 | */ | |
3501 | for (j = 0; j < map->num_members; j++) { | |
238c0a71 | 3502 | __u32 ord = get_imsm_ord_tbl_ent(dev, j, MAP_0); |
ab3cb6b3 N |
3503 | __u32 idx = ord_to_idx(ord); |
3504 | ||
20dc76d1 MT |
3505 | if (super->disks && super->disks->index == (int)idx) |
3506 | info->disk.raid_disk = j; | |
3507 | ||
ab3cb6b3 N |
3508 | if (!(ord & IMSM_ORD_REBUILD) && |
3509 | get_imsm_missing(super, idx)) { | |
3510 | missing = 1; | |
3511 | break; | |
3512 | } | |
97b4d0e9 | 3513 | } |
ab3cb6b3 N |
3514 | |
3515 | if (state == IMSM_T_STATE_FAILED) | |
3516 | enough = -1; | |
3517 | else if (state == IMSM_T_STATE_DEGRADED && | |
3518 | (state != map->map_state || missing)) | |
3519 | enough = 0; | |
3520 | else /* we're normal, or already degraded */ | |
3521 | enough = 1; | |
d2bde6d3 AK |
3522 | if (is_gen_migration(dev) && missing) { |
3523 | /* during general migration we need all disks | |
3524 | * that process is running on. | |
3525 | * No new missing disk is allowed. | |
3526 | */ | |
3527 | max_enough = -1; | |
3528 | enough = -1; | |
3529 | /* no more checks necessary | |
3530 | */ | |
3531 | break; | |
3532 | } | |
ab3cb6b3 N |
3533 | /* in the missing/failed disk case check to see |
3534 | * if at least one array is runnable | |
3535 | */ | |
3536 | max_enough = max(max_enough, enough); | |
3537 | } | |
1ade5cc1 | 3538 | dprintf("enough: %d\n", max_enough); |
ab3cb6b3 | 3539 | info->container_enough = max_enough; |
97b4d0e9 | 3540 | |
4a04ec6c | 3541 | if (super->disks) { |
14e8215b DW |
3542 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
3543 | ||
b9f594fe | 3544 | disk = &super->disks->disk; |
5551b113 | 3545 | info->data_offset = total_blocks(&super->disks->disk) - reserved; |
14e8215b | 3546 | info->component_size = reserved; |
25ed7e59 | 3547 | info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0; |
df474657 DW |
3548 | /* we don't change info->disk.raid_disk here because |
3549 | * this state will be finalized in mdmon after we have | |
3550 | * found the 'most fresh' version of the metadata | |
3551 | */ | |
25ed7e59 | 3552 | info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; |
2432ce9b AP |
3553 | info->disk.state |= (is_spare(disk) || is_journal(disk)) ? |
3554 | 0 : (1 << MD_DISK_SYNC); | |
cdddbdbc | 3555 | } |
a575e2a7 DW |
3556 | |
3557 | /* only call uuid_from_super_imsm when this disk is part of a populated container, | |
3558 | * ->compare_super may have updated the 'num_raid_devs' field for spares | |
3559 | */ | |
3560 | if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs) | |
36ba7d48 | 3561 | uuid_from_super_imsm(st, info->uuid); |
22e263f6 AC |
3562 | else |
3563 | memcpy(info->uuid, uuid_zero, sizeof(uuid_zero)); | |
a5d85af7 N |
3564 | |
3565 | /* I don't know how to compute 'map' on imsm, so use safe default */ | |
3566 | if (map) { | |
3567 | int i; | |
3568 | for (i = 0; i < map_disks; i++) | |
3569 | map[i] = 1; | |
3570 | } | |
3571 | ||
cdddbdbc DW |
3572 | } |
3573 | ||
5c4cd5da AC |
3574 | /* allocates memory and fills disk in mdinfo structure |
3575 | * for each disk in array */ | |
3576 | struct mdinfo *getinfo_super_disks_imsm(struct supertype *st) | |
3577 | { | |
594dc1b8 | 3578 | struct mdinfo *mddev; |
5c4cd5da AC |
3579 | struct intel_super *super = st->sb; |
3580 | struct imsm_disk *disk; | |
3581 | int count = 0; | |
3582 | struct dl *dl; | |
3583 | if (!super || !super->disks) | |
3584 | return NULL; | |
3585 | dl = super->disks; | |
503975b9 | 3586 | mddev = xcalloc(1, sizeof(*mddev)); |
5c4cd5da AC |
3587 | while (dl) { |
3588 | struct mdinfo *tmp; | |
3589 | disk = &dl->disk; | |
503975b9 | 3590 | tmp = xcalloc(1, sizeof(*tmp)); |
5c4cd5da AC |
3591 | if (mddev->devs) |
3592 | tmp->next = mddev->devs; | |
3593 | mddev->devs = tmp; | |
3594 | tmp->disk.number = count++; | |
3595 | tmp->disk.major = dl->major; | |
3596 | tmp->disk.minor = dl->minor; | |
3597 | tmp->disk.state = is_configured(disk) ? | |
3598 | (1 << MD_DISK_ACTIVE) : 0; | |
3599 | tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; | |
3600 | tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC); | |
3601 | tmp->disk.raid_disk = -1; | |
3602 | dl = dl->next; | |
3603 | } | |
3604 | return mddev; | |
3605 | } | |
3606 | ||
cdddbdbc DW |
3607 | static int update_super_imsm(struct supertype *st, struct mdinfo *info, |
3608 | char *update, char *devname, int verbose, | |
3609 | int uuid_set, char *homehost) | |
3610 | { | |
f352c545 DW |
3611 | /* For 'assemble' and 'force' we need to return non-zero if any |
3612 | * change was made. For others, the return value is ignored. | |
3613 | * Update options are: | |
3614 | * force-one : This device looks a bit old but needs to be included, | |
3615 | * update age info appropriately. | |
3616 | * assemble: clear any 'faulty' flag to allow this device to | |
3617 | * be assembled. | |
3618 | * force-array: Array is degraded but being forced, mark it clean | |
3619 | * if that will be needed to assemble it. | |
3620 | * | |
3621 | * newdev: not used ???? | |
3622 | * grow: Array has gained a new device - this is currently for | |
3623 | * linear only | |
3624 | * resync: mark as dirty so a resync will happen. | |
3625 | * name: update the name - preserving the homehost | |
6e46bf34 | 3626 | * uuid: Change the uuid of the array to match watch is given |
f352c545 DW |
3627 | * |
3628 | * Following are not relevant for this imsm: | |
3629 | * sparc2.2 : update from old dodgey metadata | |
3630 | * super-minor: change the preferred_minor number | |
3631 | * summaries: update redundant counters. | |
f352c545 DW |
3632 | * homehost: update the recorded homehost |
3633 | * _reshape_progress: record new reshape_progress position. | |
3634 | */ | |
6e46bf34 DW |
3635 | int rv = 1; |
3636 | struct intel_super *super = st->sb; | |
3637 | struct imsm_super *mpb; | |
f352c545 | 3638 | |
6e46bf34 DW |
3639 | /* we can only update container info */ |
3640 | if (!super || super->current_vol >= 0 || !super->anchor) | |
3641 | return 1; | |
3642 | ||
3643 | mpb = super->anchor; | |
3644 | ||
81a5b4f5 N |
3645 | if (strcmp(update, "uuid") == 0) { |
3646 | /* We take this to mean that the family_num should be updated. | |
3647 | * However that is much smaller than the uuid so we cannot really | |
3648 | * allow an explicit uuid to be given. And it is hard to reliably | |
3649 | * know if one was. | |
3650 | * So if !uuid_set we know the current uuid is random and just used | |
3651 | * the first 'int' and copy it to the other 3 positions. | |
3652 | * Otherwise we require the 4 'int's to be the same as would be the | |
3653 | * case if we are using a random uuid. So an explicit uuid will be | |
3654 | * accepted as long as all for ints are the same... which shouldn't hurt | |
6e46bf34 | 3655 | */ |
81a5b4f5 N |
3656 | if (!uuid_set) { |
3657 | info->uuid[1] = info->uuid[2] = info->uuid[3] = info->uuid[0]; | |
6e46bf34 | 3658 | rv = 0; |
81a5b4f5 N |
3659 | } else { |
3660 | if (info->uuid[0] != info->uuid[1] || | |
3661 | info->uuid[1] != info->uuid[2] || | |
3662 | info->uuid[2] != info->uuid[3]) | |
3663 | rv = -1; | |
3664 | else | |
3665 | rv = 0; | |
6e46bf34 | 3666 | } |
81a5b4f5 N |
3667 | if (rv == 0) |
3668 | mpb->orig_family_num = info->uuid[0]; | |
6e46bf34 DW |
3669 | } else if (strcmp(update, "assemble") == 0) |
3670 | rv = 0; | |
3671 | else | |
1e2b2765 | 3672 | rv = -1; |
f352c545 | 3673 | |
6e46bf34 DW |
3674 | /* successful update? recompute checksum */ |
3675 | if (rv == 0) | |
3676 | mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb)); | |
f352c545 DW |
3677 | |
3678 | return rv; | |
cdddbdbc DW |
3679 | } |
3680 | ||
c2c087e6 | 3681 | static size_t disks_to_mpb_size(int disks) |
cdddbdbc | 3682 | { |
c2c087e6 | 3683 | size_t size; |
cdddbdbc | 3684 | |
c2c087e6 DW |
3685 | size = sizeof(struct imsm_super); |
3686 | size += (disks - 1) * sizeof(struct imsm_disk); | |
3687 | size += 2 * sizeof(struct imsm_dev); | |
3688 | /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */ | |
3689 | size += (4 - 2) * sizeof(struct imsm_map); | |
3690 | /* 4 possible disk_ord_tbl's */ | |
3691 | size += 4 * (disks - 1) * sizeof(__u32); | |
bbab0940 TM |
3692 | /* maximum bbm log */ |
3693 | size += sizeof(struct bbm_log); | |
c2c087e6 DW |
3694 | |
3695 | return size; | |
3696 | } | |
3697 | ||
387fcd59 N |
3698 | static __u64 avail_size_imsm(struct supertype *st, __u64 devsize, |
3699 | unsigned long long data_offset) | |
c2c087e6 DW |
3700 | { |
3701 | if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS)) | |
3702 | return 0; | |
3703 | ||
3704 | return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS); | |
cdddbdbc DW |
3705 | } |
3706 | ||
ba2de7ba DW |
3707 | static void free_devlist(struct intel_super *super) |
3708 | { | |
3709 | struct intel_dev *dv; | |
3710 | ||
3711 | while (super->devlist) { | |
3712 | dv = super->devlist->next; | |
3713 | free(super->devlist->dev); | |
3714 | free(super->devlist); | |
3715 | super->devlist = dv; | |
3716 | } | |
3717 | } | |
3718 | ||
3719 | static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src) | |
3720 | { | |
3721 | memcpy(dest, src, sizeof_imsm_dev(src, 0)); | |
3722 | } | |
3723 | ||
cdddbdbc DW |
3724 | static int compare_super_imsm(struct supertype *st, struct supertype *tst) |
3725 | { | |
3726 | /* | |
3727 | * return: | |
3728 | * 0 same, or first was empty, and second was copied | |
3729 | * 1 second had wrong number | |
3730 | * 2 wrong uuid | |
3731 | * 3 wrong other info | |
3732 | */ | |
3733 | struct intel_super *first = st->sb; | |
3734 | struct intel_super *sec = tst->sb; | |
3735 | ||
5d500228 N |
3736 | if (!first) { |
3737 | st->sb = tst->sb; | |
3738 | tst->sb = NULL; | |
3739 | return 0; | |
3740 | } | |
8603ea6f LM |
3741 | /* in platform dependent environment test if the disks |
3742 | * use the same Intel hba | |
cb8f6859 | 3743 | * If not on Intel hba at all, allow anything. |
8603ea6f | 3744 | */ |
6b781d33 AP |
3745 | if (!check_env("IMSM_NO_PLATFORM") && first->hba && sec->hba) { |
3746 | if (first->hba->type != sec->hba->type) { | |
8603ea6f | 3747 | fprintf(stderr, |
6b781d33 AP |
3748 | "HBAs of devices do not match %s != %s\n", |
3749 | get_sys_dev_type(first->hba->type), | |
3750 | get_sys_dev_type(sec->hba->type)); | |
3751 | return 3; | |
3752 | } | |
3753 | if (first->orom != sec->orom) { | |
3754 | fprintf(stderr, | |
3755 | "HBAs of devices do not match %s != %s\n", | |
3756 | first->hba->pci_id, sec->hba->pci_id); | |
8603ea6f LM |
3757 | return 3; |
3758 | } | |
3759 | } | |
cdddbdbc | 3760 | |
d23fe947 DW |
3761 | /* if an anchor does not have num_raid_devs set then it is a free |
3762 | * floating spare | |
3763 | */ | |
3764 | if (first->anchor->num_raid_devs > 0 && | |
3765 | sec->anchor->num_raid_devs > 0) { | |
a2b97981 DW |
3766 | /* Determine if these disks might ever have been |
3767 | * related. Further disambiguation can only take place | |
3768 | * in load_super_imsm_all | |
3769 | */ | |
3770 | __u32 first_family = first->anchor->orig_family_num; | |
3771 | __u32 sec_family = sec->anchor->orig_family_num; | |
3772 | ||
f796af5d DW |
3773 | if (memcmp(first->anchor->sig, sec->anchor->sig, |
3774 | MAX_SIGNATURE_LENGTH) != 0) | |
3775 | return 3; | |
3776 | ||
a2b97981 DW |
3777 | if (first_family == 0) |
3778 | first_family = first->anchor->family_num; | |
3779 | if (sec_family == 0) | |
3780 | sec_family = sec->anchor->family_num; | |
3781 | ||
3782 | if (first_family != sec_family) | |
d23fe947 | 3783 | return 3; |
f796af5d | 3784 | |
d23fe947 | 3785 | } |
cdddbdbc | 3786 | |
3e372e5a DW |
3787 | /* if 'first' is a spare promote it to a populated mpb with sec's |
3788 | * family number | |
3789 | */ | |
3790 | if (first->anchor->num_raid_devs == 0 && | |
3791 | sec->anchor->num_raid_devs > 0) { | |
78d30f94 | 3792 | int i; |
ba2de7ba DW |
3793 | struct intel_dev *dv; |
3794 | struct imsm_dev *dev; | |
78d30f94 DW |
3795 | |
3796 | /* we need to copy raid device info from sec if an allocation | |
3797 | * fails here we don't associate the spare | |
3798 | */ | |
3799 | for (i = 0; i < sec->anchor->num_raid_devs; i++) { | |
503975b9 N |
3800 | dv = xmalloc(sizeof(*dv)); |
3801 | dev = xmalloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1)); | |
ba2de7ba DW |
3802 | dv->dev = dev; |
3803 | dv->index = i; | |
3804 | dv->next = first->devlist; | |
3805 | first->devlist = dv; | |
78d30f94 | 3806 | } |
709743c5 | 3807 | if (i < sec->anchor->num_raid_devs) { |
ba2de7ba DW |
3808 | /* allocation failure */ |
3809 | free_devlist(first); | |
e12b3daa | 3810 | pr_err("imsm: failed to associate spare\n"); |
ba2de7ba | 3811 | return 3; |
78d30f94 | 3812 | } |
3e372e5a | 3813 | first->anchor->num_raid_devs = sec->anchor->num_raid_devs; |
148acb7b | 3814 | first->anchor->orig_family_num = sec->anchor->orig_family_num; |
3e372e5a | 3815 | first->anchor->family_num = sec->anchor->family_num; |
ac6449be | 3816 | memcpy(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH); |
709743c5 DW |
3817 | for (i = 0; i < sec->anchor->num_raid_devs; i++) |
3818 | imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i)); | |
3e372e5a DW |
3819 | } |
3820 | ||
cdddbdbc DW |
3821 | return 0; |
3822 | } | |
3823 | ||
0030e8d6 DW |
3824 | static void fd2devname(int fd, char *name) |
3825 | { | |
3826 | struct stat st; | |
3827 | char path[256]; | |
33a6535d | 3828 | char dname[PATH_MAX]; |
0030e8d6 DW |
3829 | char *nm; |
3830 | int rv; | |
3831 | ||
3832 | name[0] = '\0'; | |
3833 | if (fstat(fd, &st) != 0) | |
3834 | return; | |
3835 | sprintf(path, "/sys/dev/block/%d:%d", | |
3836 | major(st.st_rdev), minor(st.st_rdev)); | |
3837 | ||
9cf014ec | 3838 | rv = readlink(path, dname, sizeof(dname)-1); |
0030e8d6 DW |
3839 | if (rv <= 0) |
3840 | return; | |
9587c373 | 3841 | |
0030e8d6 DW |
3842 | dname[rv] = '\0'; |
3843 | nm = strrchr(dname, '/'); | |
7897de29 JS |
3844 | if (nm) { |
3845 | nm++; | |
3846 | snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm); | |
3847 | } | |
0030e8d6 DW |
3848 | } |
3849 | ||
21e9380b AP |
3850 | static int nvme_get_serial(int fd, void *buf, size_t buf_len) |
3851 | { | |
3852 | char path[60]; | |
3853 | char *name = fd2kname(fd); | |
3854 | ||
3855 | if (!name) | |
3856 | return 1; | |
3857 | ||
3858 | if (strncmp(name, "nvme", 4) != 0) | |
3859 | return 1; | |
3860 | ||
3861 | snprintf(path, sizeof(path) - 1, "/sys/block/%s/device/serial", name); | |
3862 | ||
3863 | return load_sys(path, buf, buf_len); | |
3864 | } | |
3865 | ||
cdddbdbc DW |
3866 | extern int scsi_get_serial(int fd, void *buf, size_t buf_len); |
3867 | ||
3868 | static int imsm_read_serial(int fd, char *devname, | |
3869 | __u8 serial[MAX_RAID_SERIAL_LEN]) | |
3870 | { | |
21e9380b | 3871 | char buf[50]; |
cdddbdbc | 3872 | int rv; |
1f24f035 | 3873 | int len; |
316e2bf4 DW |
3874 | char *dest; |
3875 | char *src; | |
21e9380b AP |
3876 | unsigned int i; |
3877 | ||
3878 | memset(buf, 0, sizeof(buf)); | |
cdddbdbc | 3879 | |
21e9380b | 3880 | rv = nvme_get_serial(fd, buf, sizeof(buf)); |
cdddbdbc | 3881 | |
21e9380b AP |
3882 | if (rv) |
3883 | rv = scsi_get_serial(fd, buf, sizeof(buf)); | |
f9ba0ff1 | 3884 | |
40ebbb9c | 3885 | if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) { |
f9ba0ff1 DW |
3886 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
3887 | fd2devname(fd, (char *) serial); | |
0030e8d6 DW |
3888 | return 0; |
3889 | } | |
3890 | ||
cdddbdbc DW |
3891 | if (rv != 0) { |
3892 | if (devname) | |
e7b84f9d N |
3893 | pr_err("Failed to retrieve serial for %s\n", |
3894 | devname); | |
cdddbdbc DW |
3895 | return rv; |
3896 | } | |
3897 | ||
316e2bf4 DW |
3898 | /* trim all whitespace and non-printable characters and convert |
3899 | * ':' to ';' | |
3900 | */ | |
21e9380b AP |
3901 | for (i = 0, dest = buf; i < sizeof(buf) && buf[i]; i++) { |
3902 | src = &buf[i]; | |
316e2bf4 DW |
3903 | if (*src > 0x20) { |
3904 | /* ':' is reserved for use in placeholder serial | |
3905 | * numbers for missing disks | |
3906 | */ | |
3907 | if (*src == ':') | |
3908 | *dest++ = ';'; | |
3909 | else | |
3910 | *dest++ = *src; | |
3911 | } | |
3912 | } | |
21e9380b AP |
3913 | len = dest - buf; |
3914 | dest = buf; | |
316e2bf4 DW |
3915 | |
3916 | /* truncate leading characters */ | |
3917 | if (len > MAX_RAID_SERIAL_LEN) { | |
3918 | dest += len - MAX_RAID_SERIAL_LEN; | |
1f24f035 | 3919 | len = MAX_RAID_SERIAL_LEN; |
316e2bf4 | 3920 | } |
5c3db629 | 3921 | |
5c3db629 | 3922 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
316e2bf4 | 3923 | memcpy(serial, dest, len); |
cdddbdbc DW |
3924 | |
3925 | return 0; | |
3926 | } | |
3927 | ||
1f24f035 DW |
3928 | static int serialcmp(__u8 *s1, __u8 *s2) |
3929 | { | |
3930 | return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN); | |
3931 | } | |
3932 | ||
3933 | static void serialcpy(__u8 *dest, __u8 *src) | |
3934 | { | |
3935 | strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN); | |
3936 | } | |
3937 | ||
54c2c1ea DW |
3938 | static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super) |
3939 | { | |
3940 | struct dl *dl; | |
3941 | ||
3942 | for (dl = super->disks; dl; dl = dl->next) | |
3943 | if (serialcmp(dl->serial, serial) == 0) | |
3944 | break; | |
3945 | ||
3946 | return dl; | |
3947 | } | |
3948 | ||
a2b97981 DW |
3949 | static struct imsm_disk * |
3950 | __serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx) | |
3951 | { | |
3952 | int i; | |
3953 | ||
3954 | for (i = 0; i < mpb->num_disks; i++) { | |
3955 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
3956 | ||
3957 | if (serialcmp(disk->serial, serial) == 0) { | |
3958 | if (idx) | |
3959 | *idx = i; | |
3960 | return disk; | |
3961 | } | |
3962 | } | |
3963 | ||
3964 | return NULL; | |
3965 | } | |
3966 | ||
cdddbdbc DW |
3967 | static int |
3968 | load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd) | |
3969 | { | |
a2b97981 | 3970 | struct imsm_disk *disk; |
cdddbdbc DW |
3971 | struct dl *dl; |
3972 | struct stat stb; | |
cdddbdbc | 3973 | int rv; |
a2b97981 | 3974 | char name[40]; |
d23fe947 DW |
3975 | __u8 serial[MAX_RAID_SERIAL_LEN]; |
3976 | ||
3977 | rv = imsm_read_serial(fd, devname, serial); | |
3978 | ||
3979 | if (rv != 0) | |
3980 | return 2; | |
3981 | ||
503975b9 | 3982 | dl = xcalloc(1, sizeof(*dl)); |
cdddbdbc | 3983 | |
a2b97981 DW |
3984 | fstat(fd, &stb); |
3985 | dl->major = major(stb.st_rdev); | |
3986 | dl->minor = minor(stb.st_rdev); | |
3987 | dl->next = super->disks; | |
3988 | dl->fd = keep_fd ? fd : -1; | |
3989 | assert(super->disks == NULL); | |
3990 | super->disks = dl; | |
3991 | serialcpy(dl->serial, serial); | |
3992 | dl->index = -2; | |
3993 | dl->e = NULL; | |
3994 | fd2devname(fd, name); | |
3995 | if (devname) | |
503975b9 | 3996 | dl->devname = xstrdup(devname); |
a2b97981 | 3997 | else |
503975b9 | 3998 | dl->devname = xstrdup(name); |
cdddbdbc | 3999 | |
d23fe947 | 4000 | /* look up this disk's index in the current anchor */ |
a2b97981 DW |
4001 | disk = __serial_to_disk(dl->serial, super->anchor, &dl->index); |
4002 | if (disk) { | |
4003 | dl->disk = *disk; | |
4004 | /* only set index on disks that are a member of a | |
4005 | * populated contianer, i.e. one with raid_devs | |
4006 | */ | |
4007 | if (is_failed(&dl->disk)) | |
3f6efecc | 4008 | dl->index = -2; |
2432ce9b | 4009 | else if (is_spare(&dl->disk) || is_journal(&dl->disk)) |
a2b97981 | 4010 | dl->index = -1; |
3f6efecc DW |
4011 | } |
4012 | ||
949c47a0 DW |
4013 | return 0; |
4014 | } | |
4015 | ||
0c046afd DW |
4016 | /* When migrating map0 contains the 'destination' state while map1 |
4017 | * contains the current state. When not migrating map0 contains the | |
4018 | * current state. This routine assumes that map[0].map_state is set to | |
4019 | * the current array state before being called. | |
4020 | * | |
4021 | * Migration is indicated by one of the following states | |
4022 | * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed) | |
e3bba0e0 | 4023 | * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal |
0c046afd | 4024 | * map1state=unitialized) |
1484e727 | 4025 | * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal |
0c046afd | 4026 | * map1state=normal) |
e3bba0e0 | 4027 | * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal |
0c046afd | 4028 | * map1state=degraded) |
8e59f3d8 AK |
4029 | * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal |
4030 | * map1state=normal) | |
0c046afd | 4031 | */ |
8e59f3d8 AK |
4032 | static void migrate(struct imsm_dev *dev, struct intel_super *super, |
4033 | __u8 to_state, int migr_type) | |
3393c6af | 4034 | { |
0c046afd | 4035 | struct imsm_map *dest; |
238c0a71 | 4036 | struct imsm_map *src = get_imsm_map(dev, MAP_0); |
3393c6af | 4037 | |
0c046afd | 4038 | dev->vol.migr_state = 1; |
1484e727 | 4039 | set_migr_type(dev, migr_type); |
f8f603f1 | 4040 | dev->vol.curr_migr_unit = 0; |
238c0a71 | 4041 | dest = get_imsm_map(dev, MAP_1); |
0c046afd | 4042 | |
0556e1a2 | 4043 | /* duplicate and then set the target end state in map[0] */ |
3393c6af | 4044 | memcpy(dest, src, sizeof_imsm_map(src)); |
fb12a745 | 4045 | if (migr_type == MIGR_GEN_MIGR) { |
0556e1a2 DW |
4046 | __u32 ord; |
4047 | int i; | |
4048 | ||
4049 | for (i = 0; i < src->num_members; i++) { | |
4050 | ord = __le32_to_cpu(src->disk_ord_tbl[i]); | |
4051 | set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord)); | |
4052 | } | |
4053 | } | |
4054 | ||
8e59f3d8 AK |
4055 | if (migr_type == MIGR_GEN_MIGR) |
4056 | /* Clear migration record */ | |
4057 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
4058 | ||
0c046afd | 4059 | src->map_state = to_state; |
949c47a0 | 4060 | } |
f8f603f1 | 4061 | |
809da78e AK |
4062 | static void end_migration(struct imsm_dev *dev, struct intel_super *super, |
4063 | __u8 map_state) | |
f8f603f1 | 4064 | { |
238c0a71 AK |
4065 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
4066 | struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state == 0 ? | |
4067 | MAP_0 : MAP_1); | |
28bce06f | 4068 | int i, j; |
0556e1a2 DW |
4069 | |
4070 | /* merge any IMSM_ORD_REBUILD bits that were not successfully | |
4071 | * completed in the last migration. | |
4072 | * | |
28bce06f | 4073 | * FIXME add support for raid-level-migration |
0556e1a2 | 4074 | */ |
089f9d79 JS |
4075 | if (map_state != map->map_state && (is_gen_migration(dev) == 0) && |
4076 | prev->map_state != IMSM_T_STATE_UNINITIALIZED) { | |
809da78e AK |
4077 | /* when final map state is other than expected |
4078 | * merge maps (not for migration) | |
4079 | */ | |
4080 | int failed; | |
4081 | ||
4082 | for (i = 0; i < prev->num_members; i++) | |
4083 | for (j = 0; j < map->num_members; j++) | |
4084 | /* during online capacity expansion | |
4085 | * disks position can be changed | |
4086 | * if takeover is used | |
4087 | */ | |
4088 | if (ord_to_idx(map->disk_ord_tbl[j]) == | |
4089 | ord_to_idx(prev->disk_ord_tbl[i])) { | |
4090 | map->disk_ord_tbl[j] |= | |
4091 | prev->disk_ord_tbl[i]; | |
4092 | break; | |
4093 | } | |
4094 | failed = imsm_count_failed(super, dev, MAP_0); | |
4095 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
4096 | } | |
f8f603f1 DW |
4097 | |
4098 | dev->vol.migr_state = 0; | |
ea672ee1 | 4099 | set_migr_type(dev, 0); |
f8f603f1 DW |
4100 | dev->vol.curr_migr_unit = 0; |
4101 | map->map_state = map_state; | |
4102 | } | |
949c47a0 DW |
4103 | |
4104 | static int parse_raid_devices(struct intel_super *super) | |
4105 | { | |
4106 | int i; | |
4107 | struct imsm_dev *dev_new; | |
4d7b1503 | 4108 | size_t len, len_migr; |
401d313b | 4109 | size_t max_len = 0; |
4d7b1503 DW |
4110 | size_t space_needed = 0; |
4111 | struct imsm_super *mpb = super->anchor; | |
949c47a0 DW |
4112 | |
4113 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
4114 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
ba2de7ba | 4115 | struct intel_dev *dv; |
949c47a0 | 4116 | |
4d7b1503 DW |
4117 | len = sizeof_imsm_dev(dev_iter, 0); |
4118 | len_migr = sizeof_imsm_dev(dev_iter, 1); | |
4119 | if (len_migr > len) | |
4120 | space_needed += len_migr - len; | |
ca9de185 | 4121 | |
503975b9 | 4122 | dv = xmalloc(sizeof(*dv)); |
401d313b AK |
4123 | if (max_len < len_migr) |
4124 | max_len = len_migr; | |
4125 | if (max_len > len_migr) | |
4126 | space_needed += max_len - len_migr; | |
503975b9 | 4127 | dev_new = xmalloc(max_len); |
949c47a0 | 4128 | imsm_copy_dev(dev_new, dev_iter); |
ba2de7ba DW |
4129 | dv->dev = dev_new; |
4130 | dv->index = i; | |
4131 | dv->next = super->devlist; | |
4132 | super->devlist = dv; | |
949c47a0 | 4133 | } |
cdddbdbc | 4134 | |
4d7b1503 DW |
4135 | /* ensure that super->buf is large enough when all raid devices |
4136 | * are migrating | |
4137 | */ | |
4138 | if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) { | |
4139 | void *buf; | |
4140 | ||
f36a9ecd PB |
4141 | len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed, |
4142 | super->sector_size); | |
4143 | if (posix_memalign(&buf, MAX_SECTOR_SIZE, len) != 0) | |
4d7b1503 DW |
4144 | return 1; |
4145 | ||
1f45a8ad DW |
4146 | memcpy(buf, super->buf, super->len); |
4147 | memset(buf + super->len, 0, len - super->len); | |
4d7b1503 DW |
4148 | free(super->buf); |
4149 | super->buf = buf; | |
4150 | super->len = len; | |
4151 | } | |
ca9de185 | 4152 | |
bbab0940 TM |
4153 | super->extra_space += space_needed; |
4154 | ||
cdddbdbc DW |
4155 | return 0; |
4156 | } | |
4157 | ||
e2f41b2c AK |
4158 | /******************************************************************************* |
4159 | * Function: check_mpb_migr_compatibility | |
4160 | * Description: Function checks for unsupported migration features: | |
4161 | * - migration optimization area (pba_of_lba0) | |
4162 | * - descending reshape (ascending_migr) | |
4163 | * Parameters: | |
4164 | * super : imsm metadata information | |
4165 | * Returns: | |
4166 | * 0 : migration is compatible | |
4167 | * -1 : migration is not compatible | |
4168 | ******************************************************************************/ | |
4169 | int check_mpb_migr_compatibility(struct intel_super *super) | |
4170 | { | |
4171 | struct imsm_map *map0, *map1; | |
4172 | struct migr_record *migr_rec = super->migr_rec; | |
4173 | int i; | |
4174 | ||
4175 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
4176 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
4177 | ||
4178 | if (dev_iter && | |
4179 | dev_iter->vol.migr_state == 1 && | |
4180 | dev_iter->vol.migr_type == MIGR_GEN_MIGR) { | |
4181 | /* This device is migrating */ | |
238c0a71 AK |
4182 | map0 = get_imsm_map(dev_iter, MAP_0); |
4183 | map1 = get_imsm_map(dev_iter, MAP_1); | |
5551b113 | 4184 | if (pba_of_lba0(map0) != pba_of_lba0(map1)) |
e2f41b2c AK |
4185 | /* migration optimization area was used */ |
4186 | return -1; | |
fc54fe7a JS |
4187 | if (migr_rec->ascending_migr == 0 && |
4188 | migr_rec->dest_depth_per_unit > 0) | |
e2f41b2c AK |
4189 | /* descending reshape not supported yet */ |
4190 | return -1; | |
4191 | } | |
4192 | } | |
4193 | return 0; | |
4194 | } | |
4195 | ||
d23fe947 | 4196 | static void __free_imsm(struct intel_super *super, int free_disks); |
9ca2c81c | 4197 | |
cdddbdbc | 4198 | /* load_imsm_mpb - read matrix metadata |
f2f5c343 | 4199 | * allocates super->mpb to be freed by free_imsm |
cdddbdbc DW |
4200 | */ |
4201 | static int load_imsm_mpb(int fd, struct intel_super *super, char *devname) | |
4202 | { | |
4203 | unsigned long long dsize; | |
cdddbdbc | 4204 | unsigned long long sectors; |
f36a9ecd | 4205 | unsigned int sector_size = super->sector_size; |
cdddbdbc | 4206 | struct stat; |
6416d527 | 4207 | struct imsm_super *anchor; |
cdddbdbc DW |
4208 | __u32 check_sum; |
4209 | ||
cdddbdbc | 4210 | get_dev_size(fd, NULL, &dsize); |
f36a9ecd | 4211 | if (dsize < 2*sector_size) { |
64436f06 | 4212 | if (devname) |
e7b84f9d N |
4213 | pr_err("%s: device to small for imsm\n", |
4214 | devname); | |
64436f06 N |
4215 | return 1; |
4216 | } | |
cdddbdbc | 4217 | |
f36a9ecd | 4218 | if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0) { |
cdddbdbc | 4219 | if (devname) |
e7b84f9d N |
4220 | pr_err("Cannot seek to anchor block on %s: %s\n", |
4221 | devname, strerror(errno)); | |
cdddbdbc DW |
4222 | return 1; |
4223 | } | |
4224 | ||
f36a9ecd | 4225 | if (posix_memalign((void **)&anchor, sector_size, sector_size) != 0) { |
ad97895e | 4226 | if (devname) |
7a862a02 | 4227 | pr_err("Failed to allocate imsm anchor buffer on %s\n", devname); |
ad97895e DW |
4228 | return 1; |
4229 | } | |
466070ad | 4230 | if ((unsigned int)read(fd, anchor, sector_size) != sector_size) { |
cdddbdbc | 4231 | if (devname) |
e7b84f9d N |
4232 | pr_err("Cannot read anchor block on %s: %s\n", |
4233 | devname, strerror(errno)); | |
6416d527 | 4234 | free(anchor); |
cdddbdbc DW |
4235 | return 1; |
4236 | } | |
4237 | ||
6416d527 | 4238 | if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) { |
cdddbdbc | 4239 | if (devname) |
e7b84f9d | 4240 | pr_err("no IMSM anchor on %s\n", devname); |
6416d527 | 4241 | free(anchor); |
cdddbdbc DW |
4242 | return 2; |
4243 | } | |
4244 | ||
d23fe947 | 4245 | __free_imsm(super, 0); |
f2f5c343 LM |
4246 | /* reload capability and hba */ |
4247 | ||
4248 | /* capability and hba must be updated with new super allocation */ | |
d424212e | 4249 | find_intel_hba_capability(fd, super, devname); |
f36a9ecd PB |
4250 | super->len = ROUND_UP(anchor->mpb_size, sector_size); |
4251 | if (posix_memalign(&super->buf, MAX_SECTOR_SIZE, super->len) != 0) { | |
cdddbdbc | 4252 | if (devname) |
e7b84f9d N |
4253 | pr_err("unable to allocate %zu byte mpb buffer\n", |
4254 | super->len); | |
6416d527 | 4255 | free(anchor); |
cdddbdbc DW |
4256 | return 2; |
4257 | } | |
f36a9ecd | 4258 | memcpy(super->buf, anchor, sector_size); |
cdddbdbc | 4259 | |
f36a9ecd | 4260 | sectors = mpb_sectors(anchor, sector_size) - 1; |
6416d527 | 4261 | free(anchor); |
8e59f3d8 | 4262 | |
85337573 AO |
4263 | if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE, |
4264 | MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) { | |
1ade5cc1 | 4265 | pr_err("could not allocate migr_rec buffer\n"); |
8e59f3d8 AK |
4266 | free(super->buf); |
4267 | return 2; | |
4268 | } | |
51d83f5d | 4269 | super->clean_migration_record_by_mdmon = 0; |
8e59f3d8 | 4270 | |
949c47a0 | 4271 | if (!sectors) { |
ecf45690 DW |
4272 | check_sum = __gen_imsm_checksum(super->anchor); |
4273 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
4274 | if (devname) | |
e7b84f9d N |
4275 | pr_err("IMSM checksum %x != %x on %s\n", |
4276 | check_sum, | |
4277 | __le32_to_cpu(super->anchor->check_sum), | |
4278 | devname); | |
ecf45690 DW |
4279 | return 2; |
4280 | } | |
4281 | ||
a2b97981 | 4282 | return 0; |
949c47a0 | 4283 | } |
cdddbdbc DW |
4284 | |
4285 | /* read the extended mpb */ | |
f36a9ecd | 4286 | if (lseek64(fd, dsize - (sector_size * (2 + sectors)), SEEK_SET) < 0) { |
cdddbdbc | 4287 | if (devname) |
e7b84f9d N |
4288 | pr_err("Cannot seek to extended mpb on %s: %s\n", |
4289 | devname, strerror(errno)); | |
cdddbdbc DW |
4290 | return 1; |
4291 | } | |
4292 | ||
f36a9ecd PB |
4293 | if ((unsigned int)read(fd, super->buf + sector_size, |
4294 | super->len - sector_size) != super->len - sector_size) { | |
cdddbdbc | 4295 | if (devname) |
e7b84f9d N |
4296 | pr_err("Cannot read extended mpb on %s: %s\n", |
4297 | devname, strerror(errno)); | |
cdddbdbc DW |
4298 | return 2; |
4299 | } | |
4300 | ||
949c47a0 DW |
4301 | check_sum = __gen_imsm_checksum(super->anchor); |
4302 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
cdddbdbc | 4303 | if (devname) |
e7b84f9d N |
4304 | pr_err("IMSM checksum %x != %x on %s\n", |
4305 | check_sum, __le32_to_cpu(super->anchor->check_sum), | |
4306 | devname); | |
db575f3b | 4307 | return 3; |
cdddbdbc DW |
4308 | } |
4309 | ||
a2b97981 DW |
4310 | return 0; |
4311 | } | |
4312 | ||
8e59f3d8 AK |
4313 | static int read_imsm_migr_rec(int fd, struct intel_super *super); |
4314 | ||
97f81ee2 CA |
4315 | /* clears hi bits in metadata if MPB_ATTRIB_2TB_DISK not set */ |
4316 | static void clear_hi(struct intel_super *super) | |
4317 | { | |
4318 | struct imsm_super *mpb = super->anchor; | |
4319 | int i, n; | |
4320 | if (mpb->attributes & MPB_ATTRIB_2TB_DISK) | |
4321 | return; | |
4322 | for (i = 0; i < mpb->num_disks; ++i) { | |
4323 | struct imsm_disk *disk = &mpb->disk[i]; | |
4324 | disk->total_blocks_hi = 0; | |
4325 | } | |
4326 | for (i = 0; i < mpb->num_raid_devs; ++i) { | |
4327 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
4328 | if (!dev) | |
4329 | return; | |
4330 | for (n = 0; n < 2; ++n) { | |
4331 | struct imsm_map *map = get_imsm_map(dev, n); | |
4332 | if (!map) | |
4333 | continue; | |
4334 | map->pba_of_lba0_hi = 0; | |
4335 | map->blocks_per_member_hi = 0; | |
4336 | map->num_data_stripes_hi = 0; | |
4337 | } | |
4338 | } | |
4339 | } | |
4340 | ||
a2b97981 DW |
4341 | static int |
4342 | load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd) | |
4343 | { | |
4344 | int err; | |
4345 | ||
4346 | err = load_imsm_mpb(fd, super, devname); | |
4347 | if (err) | |
4348 | return err; | |
f36a9ecd PB |
4349 | if (super->sector_size == 4096) |
4350 | convert_from_4k(super); | |
a2b97981 DW |
4351 | err = load_imsm_disk(fd, super, devname, keep_fd); |
4352 | if (err) | |
4353 | return err; | |
4354 | err = parse_raid_devices(super); | |
8d67477f TM |
4355 | if (err) |
4356 | return err; | |
4357 | err = load_bbm_log(super); | |
97f81ee2 | 4358 | clear_hi(super); |
a2b97981 | 4359 | return err; |
cdddbdbc DW |
4360 | } |
4361 | ||
ae6aad82 DW |
4362 | static void __free_imsm_disk(struct dl *d) |
4363 | { | |
4364 | if (d->fd >= 0) | |
4365 | close(d->fd); | |
4366 | if (d->devname) | |
4367 | free(d->devname); | |
0dcecb2e DW |
4368 | if (d->e) |
4369 | free(d->e); | |
ae6aad82 DW |
4370 | free(d); |
4371 | ||
4372 | } | |
1a64be56 | 4373 | |
cdddbdbc DW |
4374 | static void free_imsm_disks(struct intel_super *super) |
4375 | { | |
47ee5a45 | 4376 | struct dl *d; |
cdddbdbc | 4377 | |
47ee5a45 DW |
4378 | while (super->disks) { |
4379 | d = super->disks; | |
cdddbdbc | 4380 | super->disks = d->next; |
ae6aad82 | 4381 | __free_imsm_disk(d); |
cdddbdbc | 4382 | } |
cb82edca AK |
4383 | while (super->disk_mgmt_list) { |
4384 | d = super->disk_mgmt_list; | |
4385 | super->disk_mgmt_list = d->next; | |
4386 | __free_imsm_disk(d); | |
4387 | } | |
47ee5a45 DW |
4388 | while (super->missing) { |
4389 | d = super->missing; | |
4390 | super->missing = d->next; | |
4391 | __free_imsm_disk(d); | |
4392 | } | |
4393 | ||
cdddbdbc DW |
4394 | } |
4395 | ||
9ca2c81c | 4396 | /* free all the pieces hanging off of a super pointer */ |
d23fe947 | 4397 | static void __free_imsm(struct intel_super *super, int free_disks) |
cdddbdbc | 4398 | { |
88654014 LM |
4399 | struct intel_hba *elem, *next; |
4400 | ||
9ca2c81c | 4401 | if (super->buf) { |
949c47a0 | 4402 | free(super->buf); |
9ca2c81c DW |
4403 | super->buf = NULL; |
4404 | } | |
f2f5c343 LM |
4405 | /* unlink capability description */ |
4406 | super->orom = NULL; | |
8e59f3d8 AK |
4407 | if (super->migr_rec_buf) { |
4408 | free(super->migr_rec_buf); | |
4409 | super->migr_rec_buf = NULL; | |
4410 | } | |
d23fe947 DW |
4411 | if (free_disks) |
4412 | free_imsm_disks(super); | |
ba2de7ba | 4413 | free_devlist(super); |
88654014 LM |
4414 | elem = super->hba; |
4415 | while (elem) { | |
4416 | if (elem->path) | |
4417 | free((void *)elem->path); | |
4418 | next = elem->next; | |
4419 | free(elem); | |
4420 | elem = next; | |
88c32bb1 | 4421 | } |
8d67477f TM |
4422 | if (super->bbm_log) |
4423 | free(super->bbm_log); | |
88654014 | 4424 | super->hba = NULL; |
cdddbdbc DW |
4425 | } |
4426 | ||
9ca2c81c DW |
4427 | static void free_imsm(struct intel_super *super) |
4428 | { | |
d23fe947 | 4429 | __free_imsm(super, 1); |
928f1424 | 4430 | free(super->bb.entries); |
9ca2c81c DW |
4431 | free(super); |
4432 | } | |
cdddbdbc DW |
4433 | |
4434 | static void free_super_imsm(struct supertype *st) | |
4435 | { | |
4436 | struct intel_super *super = st->sb; | |
4437 | ||
4438 | if (!super) | |
4439 | return; | |
4440 | ||
4441 | free_imsm(super); | |
4442 | st->sb = NULL; | |
4443 | } | |
4444 | ||
49133e57 | 4445 | static struct intel_super *alloc_super(void) |
c2c087e6 | 4446 | { |
503975b9 | 4447 | struct intel_super *super = xcalloc(1, sizeof(*super)); |
c2c087e6 | 4448 | |
503975b9 N |
4449 | super->current_vol = -1; |
4450 | super->create_offset = ~((unsigned long long) 0); | |
928f1424 TM |
4451 | |
4452 | super->bb.entries = xmalloc(BBM_LOG_MAX_ENTRIES * | |
4453 | sizeof(struct md_bb_entry)); | |
4454 | if (!super->bb.entries) { | |
4455 | free(super); | |
4456 | return NULL; | |
4457 | } | |
4458 | ||
c2c087e6 DW |
4459 | return super; |
4460 | } | |
4461 | ||
f0f5a016 LM |
4462 | /* |
4463 | * find and allocate hba and OROM/EFI based on valid fd of RAID component device | |
4464 | */ | |
d424212e | 4465 | static int find_intel_hba_capability(int fd, struct intel_super *super, char *devname) |
f0f5a016 LM |
4466 | { |
4467 | struct sys_dev *hba_name; | |
4468 | int rv = 0; | |
4469 | ||
089f9d79 | 4470 | if (fd < 0 || check_env("IMSM_NO_PLATFORM")) { |
f2f5c343 | 4471 | super->orom = NULL; |
f0f5a016 LM |
4472 | super->hba = NULL; |
4473 | return 0; | |
4474 | } | |
4475 | hba_name = find_disk_attached_hba(fd, NULL); | |
4476 | if (!hba_name) { | |
d424212e | 4477 | if (devname) |
e7b84f9d N |
4478 | pr_err("%s is not attached to Intel(R) RAID controller.\n", |
4479 | devname); | |
f0f5a016 LM |
4480 | return 1; |
4481 | } | |
4482 | rv = attach_hba_to_super(super, hba_name); | |
4483 | if (rv == 2) { | |
d424212e N |
4484 | if (devname) { |
4485 | struct intel_hba *hba = super->hba; | |
f0f5a016 | 4486 | |
60f0f54d PB |
4487 | pr_err("%s is attached to Intel(R) %s %s (%s),\n" |
4488 | " but the container is assigned to Intel(R) %s %s (", | |
d424212e | 4489 | devname, |
614902f6 | 4490 | get_sys_dev_type(hba_name->type), |
60f0f54d | 4491 | hba_name->type == SYS_DEV_VMD ? "domain" : "RAID controller", |
f0f5a016 | 4492 | hba_name->pci_id ? : "Err!", |
60f0f54d PB |
4493 | get_sys_dev_type(super->hba->type), |
4494 | hba->type == SYS_DEV_VMD ? "domain" : "RAID controller"); | |
f0f5a016 | 4495 | |
f0f5a016 LM |
4496 | while (hba) { |
4497 | fprintf(stderr, "%s", hba->pci_id ? : "Err!"); | |
4498 | if (hba->next) | |
4499 | fprintf(stderr, ", "); | |
4500 | hba = hba->next; | |
4501 | } | |
6b781d33 | 4502 | fprintf(stderr, ").\n" |
60f0f54d PB |
4503 | " Mixing devices attached to different %s is not allowed.\n", |
4504 | hba_name->type == SYS_DEV_VMD ? "VMD domains" : "controllers"); | |
f0f5a016 | 4505 | } |
f0f5a016 LM |
4506 | return 2; |
4507 | } | |
6b781d33 | 4508 | super->orom = find_imsm_capability(hba_name); |
f2f5c343 LM |
4509 | if (!super->orom) |
4510 | return 3; | |
614902f6 | 4511 | |
f0f5a016 LM |
4512 | return 0; |
4513 | } | |
4514 | ||
47ee5a45 DW |
4515 | /* find_missing - helper routine for load_super_imsm_all that identifies |
4516 | * disks that have disappeared from the system. This routine relies on | |
4517 | * the mpb being uptodate, which it is at load time. | |
4518 | */ | |
4519 | static int find_missing(struct intel_super *super) | |
4520 | { | |
4521 | int i; | |
4522 | struct imsm_super *mpb = super->anchor; | |
4523 | struct dl *dl; | |
4524 | struct imsm_disk *disk; | |
47ee5a45 DW |
4525 | |
4526 | for (i = 0; i < mpb->num_disks; i++) { | |
4527 | disk = __get_imsm_disk(mpb, i); | |
54c2c1ea | 4528 | dl = serial_to_dl(disk->serial, super); |
47ee5a45 DW |
4529 | if (dl) |
4530 | continue; | |
47ee5a45 | 4531 | |
503975b9 | 4532 | dl = xmalloc(sizeof(*dl)); |
47ee5a45 DW |
4533 | dl->major = 0; |
4534 | dl->minor = 0; | |
4535 | dl->fd = -1; | |
503975b9 | 4536 | dl->devname = xstrdup("missing"); |
47ee5a45 DW |
4537 | dl->index = i; |
4538 | serialcpy(dl->serial, disk->serial); | |
4539 | dl->disk = *disk; | |
689c9bf3 | 4540 | dl->e = NULL; |
47ee5a45 DW |
4541 | dl->next = super->missing; |
4542 | super->missing = dl; | |
4543 | } | |
4544 | ||
4545 | return 0; | |
4546 | } | |
4547 | ||
a2b97981 DW |
4548 | static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list) |
4549 | { | |
4550 | struct intel_disk *idisk = disk_list; | |
4551 | ||
4552 | while (idisk) { | |
4553 | if (serialcmp(idisk->disk.serial, serial) == 0) | |
4554 | break; | |
4555 | idisk = idisk->next; | |
4556 | } | |
4557 | ||
4558 | return idisk; | |
4559 | } | |
4560 | ||
4561 | static int __prep_thunderdome(struct intel_super **table, int tbl_size, | |
4562 | struct intel_super *super, | |
4563 | struct intel_disk **disk_list) | |
4564 | { | |
4565 | struct imsm_disk *d = &super->disks->disk; | |
4566 | struct imsm_super *mpb = super->anchor; | |
4567 | int i, j; | |
4568 | ||
4569 | for (i = 0; i < tbl_size; i++) { | |
4570 | struct imsm_super *tbl_mpb = table[i]->anchor; | |
4571 | struct imsm_disk *tbl_d = &table[i]->disks->disk; | |
4572 | ||
4573 | if (tbl_mpb->family_num == mpb->family_num) { | |
4574 | if (tbl_mpb->check_sum == mpb->check_sum) { | |
1ade5cc1 N |
4575 | dprintf("mpb from %d:%d matches %d:%d\n", |
4576 | super->disks->major, | |
a2b97981 DW |
4577 | super->disks->minor, |
4578 | table[i]->disks->major, | |
4579 | table[i]->disks->minor); | |
4580 | break; | |
4581 | } | |
4582 | ||
4583 | if (((is_configured(d) && !is_configured(tbl_d)) || | |
4584 | is_configured(d) == is_configured(tbl_d)) && | |
4585 | tbl_mpb->generation_num < mpb->generation_num) { | |
4586 | /* current version of the mpb is a | |
4587 | * better candidate than the one in | |
4588 | * super_table, but copy over "cross | |
4589 | * generational" status | |
4590 | */ | |
4591 | struct intel_disk *idisk; | |
4592 | ||
1ade5cc1 N |
4593 | dprintf("mpb from %d:%d replaces %d:%d\n", |
4594 | super->disks->major, | |
a2b97981 DW |
4595 | super->disks->minor, |
4596 | table[i]->disks->major, | |
4597 | table[i]->disks->minor); | |
4598 | ||
4599 | idisk = disk_list_get(tbl_d->serial, *disk_list); | |
4600 | if (idisk && is_failed(&idisk->disk)) | |
4601 | tbl_d->status |= FAILED_DISK; | |
4602 | break; | |
4603 | } else { | |
4604 | struct intel_disk *idisk; | |
4605 | struct imsm_disk *disk; | |
4606 | ||
4607 | /* tbl_mpb is more up to date, but copy | |
4608 | * over cross generational status before | |
4609 | * returning | |
4610 | */ | |
4611 | disk = __serial_to_disk(d->serial, mpb, NULL); | |
4612 | if (disk && is_failed(disk)) | |
4613 | d->status |= FAILED_DISK; | |
4614 | ||
4615 | idisk = disk_list_get(d->serial, *disk_list); | |
4616 | if (idisk) { | |
4617 | idisk->owner = i; | |
4618 | if (disk && is_configured(disk)) | |
4619 | idisk->disk.status |= CONFIGURED_DISK; | |
4620 | } | |
4621 | ||
1ade5cc1 N |
4622 | dprintf("mpb from %d:%d prefer %d:%d\n", |
4623 | super->disks->major, | |
a2b97981 DW |
4624 | super->disks->minor, |
4625 | table[i]->disks->major, | |
4626 | table[i]->disks->minor); | |
4627 | ||
4628 | return tbl_size; | |
4629 | } | |
4630 | } | |
4631 | } | |
4632 | ||
4633 | if (i >= tbl_size) | |
4634 | table[tbl_size++] = super; | |
4635 | else | |
4636 | table[i] = super; | |
4637 | ||
4638 | /* update/extend the merged list of imsm_disk records */ | |
4639 | for (j = 0; j < mpb->num_disks; j++) { | |
4640 | struct imsm_disk *disk = __get_imsm_disk(mpb, j); | |
4641 | struct intel_disk *idisk; | |
4642 | ||
4643 | idisk = disk_list_get(disk->serial, *disk_list); | |
4644 | if (idisk) { | |
4645 | idisk->disk.status |= disk->status; | |
4646 | if (is_configured(&idisk->disk) || | |
4647 | is_failed(&idisk->disk)) | |
4648 | idisk->disk.status &= ~(SPARE_DISK); | |
4649 | } else { | |
503975b9 | 4650 | idisk = xcalloc(1, sizeof(*idisk)); |
a2b97981 DW |
4651 | idisk->owner = IMSM_UNKNOWN_OWNER; |
4652 | idisk->disk = *disk; | |
4653 | idisk->next = *disk_list; | |
4654 | *disk_list = idisk; | |
4655 | } | |
4656 | ||
4657 | if (serialcmp(idisk->disk.serial, d->serial) == 0) | |
4658 | idisk->owner = i; | |
4659 | } | |
4660 | ||
4661 | return tbl_size; | |
4662 | } | |
4663 | ||
4664 | static struct intel_super * | |
4665 | validate_members(struct intel_super *super, struct intel_disk *disk_list, | |
4666 | const int owner) | |
4667 | { | |
4668 | struct imsm_super *mpb = super->anchor; | |
4669 | int ok_count = 0; | |
4670 | int i; | |
4671 | ||
4672 | for (i = 0; i < mpb->num_disks; i++) { | |
4673 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
4674 | struct intel_disk *idisk; | |
4675 | ||
4676 | idisk = disk_list_get(disk->serial, disk_list); | |
4677 | if (idisk) { | |
4678 | if (idisk->owner == owner || | |
4679 | idisk->owner == IMSM_UNKNOWN_OWNER) | |
4680 | ok_count++; | |
4681 | else | |
1ade5cc1 N |
4682 | dprintf("'%.16s' owner %d != %d\n", |
4683 | disk->serial, idisk->owner, | |
a2b97981 DW |
4684 | owner); |
4685 | } else { | |
1ade5cc1 N |
4686 | dprintf("unknown disk %x [%d]: %.16s\n", |
4687 | __le32_to_cpu(mpb->family_num), i, | |
a2b97981 DW |
4688 | disk->serial); |
4689 | break; | |
4690 | } | |
4691 | } | |
4692 | ||
4693 | if (ok_count == mpb->num_disks) | |
4694 | return super; | |
4695 | return NULL; | |
4696 | } | |
4697 | ||
4698 | static void show_conflicts(__u32 family_num, struct intel_super *super_list) | |
4699 | { | |
4700 | struct intel_super *s; | |
4701 | ||
4702 | for (s = super_list; s; s = s->next) { | |
4703 | if (family_num != s->anchor->family_num) | |
4704 | continue; | |
e12b3daa | 4705 | pr_err("Conflict, offlining family %#x on '%s'\n", |
a2b97981 DW |
4706 | __le32_to_cpu(family_num), s->disks->devname); |
4707 | } | |
4708 | } | |
4709 | ||
4710 | static struct intel_super * | |
4711 | imsm_thunderdome(struct intel_super **super_list, int len) | |
4712 | { | |
4713 | struct intel_super *super_table[len]; | |
4714 | struct intel_disk *disk_list = NULL; | |
4715 | struct intel_super *champion, *spare; | |
4716 | struct intel_super *s, **del; | |
4717 | int tbl_size = 0; | |
4718 | int conflict; | |
4719 | int i; | |
4720 | ||
4721 | memset(super_table, 0, sizeof(super_table)); | |
4722 | for (s = *super_list; s; s = s->next) | |
4723 | tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list); | |
4724 | ||
4725 | for (i = 0; i < tbl_size; i++) { | |
4726 | struct imsm_disk *d; | |
4727 | struct intel_disk *idisk; | |
4728 | struct imsm_super *mpb = super_table[i]->anchor; | |
4729 | ||
4730 | s = super_table[i]; | |
4731 | d = &s->disks->disk; | |
4732 | ||
4733 | /* 'd' must appear in merged disk list for its | |
4734 | * configuration to be valid | |
4735 | */ | |
4736 | idisk = disk_list_get(d->serial, disk_list); | |
4737 | if (idisk && idisk->owner == i) | |
4738 | s = validate_members(s, disk_list, i); | |
4739 | else | |
4740 | s = NULL; | |
4741 | ||
4742 | if (!s) | |
1ade5cc1 N |
4743 | dprintf("marking family: %#x from %d:%d offline\n", |
4744 | mpb->family_num, | |
a2b97981 DW |
4745 | super_table[i]->disks->major, |
4746 | super_table[i]->disks->minor); | |
4747 | super_table[i] = s; | |
4748 | } | |
4749 | ||
4750 | /* This is where the mdadm implementation differs from the Windows | |
4751 | * driver which has no strict concept of a container. We can only | |
4752 | * assemble one family from a container, so when returning a prodigal | |
4753 | * array member to this system the code will not be able to disambiguate | |
4754 | * the container contents that should be assembled ("foreign" versus | |
4755 | * "local"). It requires user intervention to set the orig_family_num | |
4756 | * to a new value to establish a new container. The Windows driver in | |
4757 | * this situation fixes up the volume name in place and manages the | |
4758 | * foreign array as an independent entity. | |
4759 | */ | |
4760 | s = NULL; | |
4761 | spare = NULL; | |
4762 | conflict = 0; | |
4763 | for (i = 0; i < tbl_size; i++) { | |
4764 | struct intel_super *tbl_ent = super_table[i]; | |
4765 | int is_spare = 0; | |
4766 | ||
4767 | if (!tbl_ent) | |
4768 | continue; | |
4769 | ||
4770 | if (tbl_ent->anchor->num_raid_devs == 0) { | |
4771 | spare = tbl_ent; | |
4772 | is_spare = 1; | |
4773 | } | |
4774 | ||
4775 | if (s && !is_spare) { | |
4776 | show_conflicts(tbl_ent->anchor->family_num, *super_list); | |
4777 | conflict++; | |
4778 | } else if (!s && !is_spare) | |
4779 | s = tbl_ent; | |
4780 | } | |
4781 | ||
4782 | if (!s) | |
4783 | s = spare; | |
4784 | if (!s) { | |
4785 | champion = NULL; | |
4786 | goto out; | |
4787 | } | |
4788 | champion = s; | |
4789 | ||
4790 | if (conflict) | |
7a862a02 | 4791 | pr_err("Chose family %#x on '%s', assemble conflicts to new container with '--update=uuid'\n", |
a2b97981 DW |
4792 | __le32_to_cpu(s->anchor->family_num), s->disks->devname); |
4793 | ||
4794 | /* collect all dl's onto 'champion', and update them to | |
4795 | * champion's version of the status | |
4796 | */ | |
4797 | for (s = *super_list; s; s = s->next) { | |
4798 | struct imsm_super *mpb = champion->anchor; | |
4799 | struct dl *dl = s->disks; | |
4800 | ||
4801 | if (s == champion) | |
4802 | continue; | |
4803 | ||
5d7b407a CA |
4804 | mpb->attributes |= s->anchor->attributes & MPB_ATTRIB_2TB_DISK; |
4805 | ||
a2b97981 DW |
4806 | for (i = 0; i < mpb->num_disks; i++) { |
4807 | struct imsm_disk *disk; | |
4808 | ||
4809 | disk = __serial_to_disk(dl->serial, mpb, &dl->index); | |
4810 | if (disk) { | |
4811 | dl->disk = *disk; | |
4812 | /* only set index on disks that are a member of | |
4813 | * a populated contianer, i.e. one with | |
4814 | * raid_devs | |
4815 | */ | |
4816 | if (is_failed(&dl->disk)) | |
4817 | dl->index = -2; | |
4818 | else if (is_spare(&dl->disk)) | |
4819 | dl->index = -1; | |
4820 | break; | |
4821 | } | |
4822 | } | |
4823 | ||
4824 | if (i >= mpb->num_disks) { | |
4825 | struct intel_disk *idisk; | |
4826 | ||
4827 | idisk = disk_list_get(dl->serial, disk_list); | |
ecf408e9 | 4828 | if (idisk && is_spare(&idisk->disk) && |
a2b97981 DW |
4829 | !is_failed(&idisk->disk) && !is_configured(&idisk->disk)) |
4830 | dl->index = -1; | |
4831 | else { | |
4832 | dl->index = -2; | |
4833 | continue; | |
4834 | } | |
4835 | } | |
4836 | ||
4837 | dl->next = champion->disks; | |
4838 | champion->disks = dl; | |
4839 | s->disks = NULL; | |
4840 | } | |
4841 | ||
4842 | /* delete 'champion' from super_list */ | |
4843 | for (del = super_list; *del; ) { | |
4844 | if (*del == champion) { | |
4845 | *del = (*del)->next; | |
4846 | break; | |
4847 | } else | |
4848 | del = &(*del)->next; | |
4849 | } | |
4850 | champion->next = NULL; | |
4851 | ||
4852 | out: | |
4853 | while (disk_list) { | |
4854 | struct intel_disk *idisk = disk_list; | |
4855 | ||
4856 | disk_list = disk_list->next; | |
4857 | free(idisk); | |
4858 | } | |
4859 | ||
4860 | return champion; | |
4861 | } | |
4862 | ||
9587c373 LM |
4863 | static int |
4864 | get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd); | |
4dd2df09 | 4865 | static int get_super_block(struct intel_super **super_list, char *devnm, char *devname, |
9587c373 | 4866 | int major, int minor, int keep_fd); |
ec50f7b6 LM |
4867 | static int |
4868 | get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list, | |
4869 | int *max, int keep_fd); | |
4870 | ||
cdddbdbc | 4871 | static int load_super_imsm_all(struct supertype *st, int fd, void **sbp, |
ec50f7b6 LM |
4872 | char *devname, struct md_list *devlist, |
4873 | int keep_fd) | |
cdddbdbc | 4874 | { |
a2b97981 DW |
4875 | struct intel_super *super_list = NULL; |
4876 | struct intel_super *super = NULL; | |
a2b97981 | 4877 | int err = 0; |
9587c373 | 4878 | int i = 0; |
dab4a513 | 4879 | |
9587c373 LM |
4880 | if (fd >= 0) |
4881 | /* 'fd' is an opened container */ | |
4882 | err = get_sra_super_block(fd, &super_list, devname, &i, keep_fd); | |
4883 | else | |
ec50f7b6 LM |
4884 | /* get super block from devlist devices */ |
4885 | err = get_devlist_super_block(devlist, &super_list, &i, keep_fd); | |
9587c373 | 4886 | if (err) |
1602d52c | 4887 | goto error; |
a2b97981 DW |
4888 | /* all mpbs enter, maybe one leaves */ |
4889 | super = imsm_thunderdome(&super_list, i); | |
4890 | if (!super) { | |
4891 | err = 1; | |
4892 | goto error; | |
cdddbdbc DW |
4893 | } |
4894 | ||
47ee5a45 DW |
4895 | if (find_missing(super) != 0) { |
4896 | free_imsm(super); | |
a2b97981 DW |
4897 | err = 2; |
4898 | goto error; | |
47ee5a45 | 4899 | } |
8e59f3d8 AK |
4900 | |
4901 | /* load migration record */ | |
4902 | err = load_imsm_migr_rec(super, NULL); | |
4c965cc9 AK |
4903 | if (err == -1) { |
4904 | /* migration is in progress, | |
4905 | * but migr_rec cannot be loaded, | |
4906 | */ | |
8e59f3d8 AK |
4907 | err = 4; |
4908 | goto error; | |
4909 | } | |
e2f41b2c AK |
4910 | |
4911 | /* Check migration compatibility */ | |
089f9d79 | 4912 | if (err == 0 && check_mpb_migr_compatibility(super) != 0) { |
e7b84f9d | 4913 | pr_err("Unsupported migration detected"); |
e2f41b2c AK |
4914 | if (devname) |
4915 | fprintf(stderr, " on %s\n", devname); | |
4916 | else | |
4917 | fprintf(stderr, " (IMSM).\n"); | |
4918 | ||
4919 | err = 5; | |
4920 | goto error; | |
4921 | } | |
4922 | ||
a2b97981 DW |
4923 | err = 0; |
4924 | ||
4925 | error: | |
4926 | while (super_list) { | |
4927 | struct intel_super *s = super_list; | |
4928 | ||
4929 | super_list = super_list->next; | |
4930 | free_imsm(s); | |
4931 | } | |
9587c373 | 4932 | |
a2b97981 DW |
4933 | if (err) |
4934 | return err; | |
f7e7067b | 4935 | |
cdddbdbc | 4936 | *sbp = super; |
9587c373 | 4937 | if (fd >= 0) |
4dd2df09 | 4938 | strcpy(st->container_devnm, fd2devnm(fd)); |
9587c373 | 4939 | else |
4dd2df09 | 4940 | st->container_devnm[0] = 0; |
a2b97981 | 4941 | if (err == 0 && st->ss == NULL) { |
bf5a934a | 4942 | st->ss = &super_imsm; |
cdddbdbc DW |
4943 | st->minor_version = 0; |
4944 | st->max_devs = IMSM_MAX_DEVICES; | |
4945 | } | |
cdddbdbc DW |
4946 | return 0; |
4947 | } | |
2b959fbf | 4948 | |
ec50f7b6 LM |
4949 | static int |
4950 | get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list, | |
4951 | int *max, int keep_fd) | |
4952 | { | |
4953 | struct md_list *tmpdev; | |
4954 | int err = 0; | |
4955 | int i = 0; | |
9587c373 | 4956 | |
ec50f7b6 LM |
4957 | for (i = 0, tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { |
4958 | if (tmpdev->used != 1) | |
4959 | continue; | |
4960 | if (tmpdev->container == 1) { | |
ca9de185 | 4961 | int lmax = 0; |
ec50f7b6 LM |
4962 | int fd = dev_open(tmpdev->devname, O_RDONLY|O_EXCL); |
4963 | if (fd < 0) { | |
e7b84f9d | 4964 | pr_err("cannot open device %s: %s\n", |
ec50f7b6 LM |
4965 | tmpdev->devname, strerror(errno)); |
4966 | err = 8; | |
4967 | goto error; | |
4968 | } | |
4969 | err = get_sra_super_block(fd, super_list, | |
4970 | tmpdev->devname, &lmax, | |
4971 | keep_fd); | |
4972 | i += lmax; | |
4973 | close(fd); | |
4974 | if (err) { | |
4975 | err = 7; | |
4976 | goto error; | |
4977 | } | |
4978 | } else { | |
4979 | int major = major(tmpdev->st_rdev); | |
4980 | int minor = minor(tmpdev->st_rdev); | |
4981 | err = get_super_block(super_list, | |
4dd2df09 | 4982 | NULL, |
ec50f7b6 LM |
4983 | tmpdev->devname, |
4984 | major, minor, | |
4985 | keep_fd); | |
4986 | i++; | |
4987 | if (err) { | |
4988 | err = 6; | |
4989 | goto error; | |
4990 | } | |
4991 | } | |
4992 | } | |
4993 | error: | |
4994 | *max = i; | |
4995 | return err; | |
4996 | } | |
9587c373 | 4997 | |
4dd2df09 | 4998 | static int get_super_block(struct intel_super **super_list, char *devnm, char *devname, |
9587c373 LM |
4999 | int major, int minor, int keep_fd) |
5000 | { | |
594dc1b8 | 5001 | struct intel_super *s; |
9587c373 LM |
5002 | char nm[32]; |
5003 | int dfd = -1; | |
9587c373 LM |
5004 | int err = 0; |
5005 | int retry; | |
5006 | ||
5007 | s = alloc_super(); | |
5008 | if (!s) { | |
5009 | err = 1; | |
5010 | goto error; | |
5011 | } | |
5012 | ||
5013 | sprintf(nm, "%d:%d", major, minor); | |
5014 | dfd = dev_open(nm, O_RDWR); | |
5015 | if (dfd < 0) { | |
5016 | err = 2; | |
5017 | goto error; | |
5018 | } | |
5019 | ||
fa7bb6f8 | 5020 | get_dev_sector_size(dfd, NULL, &s->sector_size); |
cb8f6859 | 5021 | find_intel_hba_capability(dfd, s, devname); |
9587c373 LM |
5022 | err = load_and_parse_mpb(dfd, s, NULL, keep_fd); |
5023 | ||
5024 | /* retry the load if we might have raced against mdmon */ | |
4dd2df09 | 5025 | if (err == 3 && devnm && mdmon_running(devnm)) |
9587c373 LM |
5026 | for (retry = 0; retry < 3; retry++) { |
5027 | usleep(3000); | |
5028 | err = load_and_parse_mpb(dfd, s, NULL, keep_fd); | |
5029 | if (err != 3) | |
5030 | break; | |
5031 | } | |
5032 | error: | |
5033 | if (!err) { | |
5034 | s->next = *super_list; | |
5035 | *super_list = s; | |
5036 | } else { | |
5037 | if (s) | |
8d67477f | 5038 | free_imsm(s); |
36614e95 | 5039 | if (dfd >= 0) |
9587c373 LM |
5040 | close(dfd); |
5041 | } | |
089f9d79 | 5042 | if (dfd >= 0 && !keep_fd) |
9587c373 LM |
5043 | close(dfd); |
5044 | return err; | |
5045 | ||
5046 | } | |
5047 | ||
5048 | static int | |
5049 | get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd) | |
5050 | { | |
5051 | struct mdinfo *sra; | |
4dd2df09 | 5052 | char *devnm; |
9587c373 LM |
5053 | struct mdinfo *sd; |
5054 | int err = 0; | |
5055 | int i = 0; | |
4dd2df09 | 5056 | sra = sysfs_read(fd, NULL, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE); |
9587c373 LM |
5057 | if (!sra) |
5058 | return 1; | |
5059 | ||
5060 | if (sra->array.major_version != -1 || | |
5061 | sra->array.minor_version != -2 || | |
5062 | strcmp(sra->text_version, "imsm") != 0) { | |
5063 | err = 1; | |
5064 | goto error; | |
5065 | } | |
5066 | /* load all mpbs */ | |
4dd2df09 | 5067 | devnm = fd2devnm(fd); |
9587c373 | 5068 | for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) { |
4dd2df09 | 5069 | if (get_super_block(super_list, devnm, devname, |
9587c373 LM |
5070 | sd->disk.major, sd->disk.minor, keep_fd) != 0) { |
5071 | err = 7; | |
5072 | goto error; | |
5073 | } | |
5074 | } | |
5075 | error: | |
5076 | sysfs_free(sra); | |
5077 | *max = i; | |
5078 | return err; | |
5079 | } | |
5080 | ||
2b959fbf N |
5081 | static int load_container_imsm(struct supertype *st, int fd, char *devname) |
5082 | { | |
ec50f7b6 | 5083 | return load_super_imsm_all(st, fd, &st->sb, devname, NULL, 1); |
2b959fbf | 5084 | } |
cdddbdbc DW |
5085 | |
5086 | static int load_super_imsm(struct supertype *st, int fd, char *devname) | |
5087 | { | |
5088 | struct intel_super *super; | |
5089 | int rv; | |
8a3544f8 | 5090 | int retry; |
cdddbdbc | 5091 | |
357ac106 | 5092 | if (test_partition(fd)) |
691c6ee1 N |
5093 | /* IMSM not allowed on partitions */ |
5094 | return 1; | |
5095 | ||
37424f13 DW |
5096 | free_super_imsm(st); |
5097 | ||
49133e57 | 5098 | super = alloc_super(); |
fa7bb6f8 | 5099 | get_dev_sector_size(fd, NULL, &super->sector_size); |
8d67477f TM |
5100 | if (!super) |
5101 | return 1; | |
ea2bc72b LM |
5102 | /* Load hba and capabilities if they exist. |
5103 | * But do not preclude loading metadata in case capabilities or hba are | |
5104 | * non-compliant and ignore_hw_compat is set. | |
5105 | */ | |
d424212e | 5106 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 5107 | /* no orom/efi or non-intel hba of the disk */ |
089f9d79 | 5108 | if (rv != 0 && st->ignore_hw_compat == 0) { |
f2f5c343 | 5109 | if (devname) |
e7b84f9d | 5110 | pr_err("No OROM/EFI properties for %s\n", devname); |
f2f5c343 LM |
5111 | free_imsm(super); |
5112 | return 2; | |
5113 | } | |
a2b97981 | 5114 | rv = load_and_parse_mpb(fd, super, devname, 0); |
cdddbdbc | 5115 | |
8a3544f8 AP |
5116 | /* retry the load if we might have raced against mdmon */ |
5117 | if (rv == 3) { | |
f96b1302 AP |
5118 | struct mdstat_ent *mdstat = NULL; |
5119 | char *name = fd2kname(fd); | |
5120 | ||
5121 | if (name) | |
5122 | mdstat = mdstat_by_component(name); | |
8a3544f8 AP |
5123 | |
5124 | if (mdstat && mdmon_running(mdstat->devnm) && getpid() != mdmon_pid(mdstat->devnm)) { | |
5125 | for (retry = 0; retry < 3; retry++) { | |
5126 | usleep(3000); | |
5127 | rv = load_and_parse_mpb(fd, super, devname, 0); | |
5128 | if (rv != 3) | |
5129 | break; | |
5130 | } | |
5131 | } | |
5132 | ||
5133 | free_mdstat(mdstat); | |
5134 | } | |
5135 | ||
cdddbdbc DW |
5136 | if (rv) { |
5137 | if (devname) | |
7a862a02 | 5138 | pr_err("Failed to load all information sections on %s\n", devname); |
cdddbdbc DW |
5139 | free_imsm(super); |
5140 | return rv; | |
5141 | } | |
5142 | ||
5143 | st->sb = super; | |
5144 | if (st->ss == NULL) { | |
5145 | st->ss = &super_imsm; | |
5146 | st->minor_version = 0; | |
5147 | st->max_devs = IMSM_MAX_DEVICES; | |
5148 | } | |
8e59f3d8 AK |
5149 | |
5150 | /* load migration record */ | |
2e062e82 AK |
5151 | if (load_imsm_migr_rec(super, NULL) == 0) { |
5152 | /* Check for unsupported migration features */ | |
5153 | if (check_mpb_migr_compatibility(super) != 0) { | |
e7b84f9d | 5154 | pr_err("Unsupported migration detected"); |
2e062e82 AK |
5155 | if (devname) |
5156 | fprintf(stderr, " on %s\n", devname); | |
5157 | else | |
5158 | fprintf(stderr, " (IMSM).\n"); | |
5159 | return 3; | |
5160 | } | |
e2f41b2c AK |
5161 | } |
5162 | ||
cdddbdbc DW |
5163 | return 0; |
5164 | } | |
5165 | ||
ef6ffade DW |
5166 | static __u16 info_to_blocks_per_strip(mdu_array_info_t *info) |
5167 | { | |
5168 | if (info->level == 1) | |
5169 | return 128; | |
5170 | return info->chunk_size >> 9; | |
5171 | } | |
5172 | ||
5551b113 CA |
5173 | static unsigned long long info_to_blocks_per_member(mdu_array_info_t *info, |
5174 | unsigned long long size) | |
fcfd9599 | 5175 | { |
4025c288 | 5176 | if (info->level == 1) |
5551b113 | 5177 | return size * 2; |
4025c288 | 5178 | else |
5551b113 | 5179 | return (size * 2) & ~(info_to_blocks_per_strip(info) - 1); |
fcfd9599 DW |
5180 | } |
5181 | ||
4d1313e9 DW |
5182 | static void imsm_update_version_info(struct intel_super *super) |
5183 | { | |
5184 | /* update the version and attributes */ | |
5185 | struct imsm_super *mpb = super->anchor; | |
5186 | char *version; | |
5187 | struct imsm_dev *dev; | |
5188 | struct imsm_map *map; | |
5189 | int i; | |
5190 | ||
5191 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
5192 | dev = get_imsm_dev(super, i); | |
238c0a71 | 5193 | map = get_imsm_map(dev, MAP_0); |
4d1313e9 DW |
5194 | if (__le32_to_cpu(dev->size_high) > 0) |
5195 | mpb->attributes |= MPB_ATTRIB_2TB; | |
5196 | ||
5197 | /* FIXME detect when an array spans a port multiplier */ | |
5198 | #if 0 | |
5199 | mpb->attributes |= MPB_ATTRIB_PM; | |
5200 | #endif | |
5201 | ||
5202 | if (mpb->num_raid_devs > 1 || | |
5203 | mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) { | |
5204 | version = MPB_VERSION_ATTRIBS; | |
5205 | switch (get_imsm_raid_level(map)) { | |
5206 | case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break; | |
5207 | case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break; | |
5208 | case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break; | |
5209 | case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break; | |
5210 | } | |
5211 | } else { | |
5212 | if (map->num_members >= 5) | |
5213 | version = MPB_VERSION_5OR6_DISK_ARRAY; | |
5214 | else if (dev->status == DEV_CLONE_N_GO) | |
5215 | version = MPB_VERSION_CNG; | |
5216 | else if (get_imsm_raid_level(map) == 5) | |
5217 | version = MPB_VERSION_RAID5; | |
5218 | else if (map->num_members >= 3) | |
5219 | version = MPB_VERSION_3OR4_DISK_ARRAY; | |
5220 | else if (get_imsm_raid_level(map) == 1) | |
5221 | version = MPB_VERSION_RAID1; | |
5222 | else | |
5223 | version = MPB_VERSION_RAID0; | |
5224 | } | |
5225 | strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version); | |
5226 | } | |
5227 | } | |
5228 | ||
aa534678 DW |
5229 | static int check_name(struct intel_super *super, char *name, int quiet) |
5230 | { | |
5231 | struct imsm_super *mpb = super->anchor; | |
5232 | char *reason = NULL; | |
5233 | int i; | |
5234 | ||
5235 | if (strlen(name) > MAX_RAID_SERIAL_LEN) | |
5236 | reason = "must be 16 characters or less"; | |
5237 | ||
5238 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
5239 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
5240 | ||
5241 | if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) { | |
5242 | reason = "already exists"; | |
5243 | break; | |
5244 | } | |
5245 | } | |
5246 | ||
5247 | if (reason && !quiet) | |
e7b84f9d | 5248 | pr_err("imsm volume name %s\n", reason); |
aa534678 DW |
5249 | |
5250 | return !reason; | |
5251 | } | |
5252 | ||
8b353278 | 5253 | static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info, |
5308f117 | 5254 | struct shape *s, char *name, |
83cd1e97 N |
5255 | char *homehost, int *uuid, |
5256 | long long data_offset) | |
cdddbdbc | 5257 | { |
c2c087e6 DW |
5258 | /* We are creating a volume inside a pre-existing container. |
5259 | * so st->sb is already set. | |
5260 | */ | |
5261 | struct intel_super *super = st->sb; | |
f36a9ecd | 5262 | unsigned int sector_size = super->sector_size; |
949c47a0 | 5263 | struct imsm_super *mpb = super->anchor; |
ba2de7ba | 5264 | struct intel_dev *dv; |
c2c087e6 DW |
5265 | struct imsm_dev *dev; |
5266 | struct imsm_vol *vol; | |
5267 | struct imsm_map *map; | |
5268 | int idx = mpb->num_raid_devs; | |
5269 | int i; | |
5270 | unsigned long long array_blocks; | |
2c092cad | 5271 | size_t size_old, size_new; |
5551b113 | 5272 | unsigned long long num_data_stripes; |
b53bfba6 TM |
5273 | unsigned int data_disks; |
5274 | unsigned long long size_per_member; | |
cdddbdbc | 5275 | |
88c32bb1 | 5276 | if (super->orom && mpb->num_raid_devs >= super->orom->vpa) { |
7a862a02 | 5277 | pr_err("This imsm-container already has the maximum of %d volumes\n", super->orom->vpa); |
c2c087e6 DW |
5278 | return 0; |
5279 | } | |
5280 | ||
2c092cad DW |
5281 | /* ensure the mpb is large enough for the new data */ |
5282 | size_old = __le32_to_cpu(mpb->mpb_size); | |
5283 | size_new = disks_to_mpb_size(info->nr_disks); | |
5284 | if (size_new > size_old) { | |
5285 | void *mpb_new; | |
f36a9ecd | 5286 | size_t size_round = ROUND_UP(size_new, sector_size); |
2c092cad | 5287 | |
f36a9ecd | 5288 | if (posix_memalign(&mpb_new, sector_size, size_round) != 0) { |
e7b84f9d | 5289 | pr_err("could not allocate new mpb\n"); |
2c092cad DW |
5290 | return 0; |
5291 | } | |
85337573 AO |
5292 | if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE, |
5293 | MIGR_REC_BUF_SECTORS* | |
5294 | MAX_SECTOR_SIZE) != 0) { | |
1ade5cc1 | 5295 | pr_err("could not allocate migr_rec buffer\n"); |
8e59f3d8 AK |
5296 | free(super->buf); |
5297 | free(super); | |
ea944c8f | 5298 | free(mpb_new); |
8e59f3d8 AK |
5299 | return 0; |
5300 | } | |
2c092cad DW |
5301 | memcpy(mpb_new, mpb, size_old); |
5302 | free(mpb); | |
5303 | mpb = mpb_new; | |
949c47a0 | 5304 | super->anchor = mpb_new; |
2c092cad DW |
5305 | mpb->mpb_size = __cpu_to_le32(size_new); |
5306 | memset(mpb_new + size_old, 0, size_round - size_old); | |
bbab0940 | 5307 | super->len = size_round; |
2c092cad | 5308 | } |
bf5a934a | 5309 | super->current_vol = idx; |
3960e579 DW |
5310 | |
5311 | /* handle 'failed_disks' by either: | |
5312 | * a) create dummy disk entries in the table if this the first | |
5313 | * volume in the array. We add them here as this is the only | |
5314 | * opportunity to add them. add_to_super_imsm_volume() | |
5315 | * handles the non-failed disks and continues incrementing | |
5316 | * mpb->num_disks. | |
5317 | * b) validate that 'failed_disks' matches the current number | |
5318 | * of missing disks if the container is populated | |
d23fe947 | 5319 | */ |
3960e579 | 5320 | if (super->current_vol == 0) { |
d23fe947 | 5321 | mpb->num_disks = 0; |
3960e579 DW |
5322 | for (i = 0; i < info->failed_disks; i++) { |
5323 | struct imsm_disk *disk; | |
5324 | ||
5325 | mpb->num_disks++; | |
5326 | disk = __get_imsm_disk(mpb, i); | |
5327 | disk->status = CONFIGURED_DISK | FAILED_DISK; | |
5328 | disk->scsi_id = __cpu_to_le32(~(__u32)0); | |
5329 | snprintf((char *) disk->serial, MAX_RAID_SERIAL_LEN, | |
5b975129 | 5330 | "missing:%d", (__u8)i); |
3960e579 DW |
5331 | } |
5332 | find_missing(super); | |
5333 | } else { | |
5334 | int missing = 0; | |
5335 | struct dl *d; | |
5336 | ||
5337 | for (d = super->missing; d; d = d->next) | |
5338 | missing++; | |
5339 | if (info->failed_disks > missing) { | |
e7b84f9d | 5340 | pr_err("unable to add 'missing' disk to container\n"); |
3960e579 DW |
5341 | return 0; |
5342 | } | |
5343 | } | |
5a038140 | 5344 | |
aa534678 DW |
5345 | if (!check_name(super, name, 0)) |
5346 | return 0; | |
503975b9 N |
5347 | dv = xmalloc(sizeof(*dv)); |
5348 | dev = xcalloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1)); | |
c2c087e6 | 5349 | strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN); |
e03640bd | 5350 | array_blocks = calc_array_size(info->level, info->raid_disks, |
03bcbc65 | 5351 | info->layout, info->chunk_size, |
b53bfba6 TM |
5352 | s->size * BLOCKS_PER_KB); |
5353 | data_disks = get_data_disks(info->level, info->layout, | |
5354 | info->raid_disks); | |
5355 | array_blocks = round_size_to_mb(array_blocks, data_disks); | |
5356 | size_per_member = array_blocks / data_disks; | |
979d38be | 5357 | |
c2c087e6 DW |
5358 | dev->size_low = __cpu_to_le32((__u32) array_blocks); |
5359 | dev->size_high = __cpu_to_le32((__u32) (array_blocks >> 32)); | |
1a2487c2 | 5360 | dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING); |
c2c087e6 DW |
5361 | vol = &dev->vol; |
5362 | vol->migr_state = 0; | |
1484e727 | 5363 | set_migr_type(dev, MIGR_INIT); |
3960e579 | 5364 | vol->dirty = !info->state; |
f8f603f1 | 5365 | vol->curr_migr_unit = 0; |
238c0a71 | 5366 | map = get_imsm_map(dev, MAP_0); |
5551b113 | 5367 | set_pba_of_lba0(map, super->create_offset); |
b53bfba6 TM |
5368 | set_blocks_per_member(map, info_to_blocks_per_member(info, |
5369 | size_per_member / | |
5370 | BLOCKS_PER_KB)); | |
ef6ffade | 5371 | map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info)); |
0556e1a2 | 5372 | map->failed_disk_num = ~0; |
bf4442ab | 5373 | if (info->level > 0) |
fffaf1ff N |
5374 | map->map_state = (info->state ? IMSM_T_STATE_NORMAL |
5375 | : IMSM_T_STATE_UNINITIALIZED); | |
bf4442ab AK |
5376 | else |
5377 | map->map_state = info->failed_disks ? IMSM_T_STATE_FAILED : | |
5378 | IMSM_T_STATE_NORMAL; | |
252d23c0 | 5379 | map->ddf = 1; |
ef6ffade DW |
5380 | |
5381 | if (info->level == 1 && info->raid_disks > 2) { | |
38950822 AW |
5382 | free(dev); |
5383 | free(dv); | |
7a862a02 | 5384 | pr_err("imsm does not support more than 2 disksin a raid1 volume\n"); |
ef6ffade DW |
5385 | return 0; |
5386 | } | |
81062a36 DW |
5387 | |
5388 | map->raid_level = info->level; | |
4d1313e9 | 5389 | if (info->level == 10) { |
c2c087e6 | 5390 | map->raid_level = 1; |
4d1313e9 | 5391 | map->num_domains = info->raid_disks / 2; |
81062a36 DW |
5392 | } else if (info->level == 1) |
5393 | map->num_domains = info->raid_disks; | |
5394 | else | |
ff596308 | 5395 | map->num_domains = 1; |
81062a36 | 5396 | |
5551b113 | 5397 | /* info->size is only int so use the 'size' parameter instead */ |
b53bfba6 | 5398 | num_data_stripes = size_per_member / info_to_blocks_per_strip(info); |
5551b113 CA |
5399 | num_data_stripes /= map->num_domains; |
5400 | set_num_data_stripes(map, num_data_stripes); | |
ef6ffade | 5401 | |
c2c087e6 DW |
5402 | map->num_members = info->raid_disks; |
5403 | for (i = 0; i < map->num_members; i++) { | |
5404 | /* initialized in add_to_super */ | |
4eb26970 | 5405 | set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD); |
c2c087e6 | 5406 | } |
949c47a0 | 5407 | mpb->num_raid_devs++; |
2a24dc1b PB |
5408 | mpb->num_raid_devs_created++; |
5409 | dev->my_vol_raid_dev_num = mpb->num_raid_devs_created; | |
ba2de7ba | 5410 | |
b7580566 | 5411 | if (s->consistency_policy <= CONSISTENCY_POLICY_RESYNC) { |
c2462068 | 5412 | dev->rwh_policy = RWH_MULTIPLE_OFF; |
2432ce9b | 5413 | } else if (s->consistency_policy == CONSISTENCY_POLICY_PPL) { |
c2462068 | 5414 | dev->rwh_policy = RWH_MULTIPLE_DISTRIBUTED; |
2432ce9b AP |
5415 | } else { |
5416 | free(dev); | |
5417 | free(dv); | |
5418 | pr_err("imsm does not support consistency policy %s\n", | |
5419 | map_num(consistency_policies, s->consistency_policy)); | |
5420 | return 0; | |
5421 | } | |
5422 | ||
ba2de7ba DW |
5423 | dv->dev = dev; |
5424 | dv->index = super->current_vol; | |
5425 | dv->next = super->devlist; | |
5426 | super->devlist = dv; | |
c2c087e6 | 5427 | |
4d1313e9 DW |
5428 | imsm_update_version_info(super); |
5429 | ||
c2c087e6 | 5430 | return 1; |
cdddbdbc DW |
5431 | } |
5432 | ||
bf5a934a | 5433 | static int init_super_imsm(struct supertype *st, mdu_array_info_t *info, |
5308f117 | 5434 | struct shape *s, char *name, |
83cd1e97 N |
5435 | char *homehost, int *uuid, |
5436 | unsigned long long data_offset) | |
bf5a934a DW |
5437 | { |
5438 | /* This is primarily called by Create when creating a new array. | |
5439 | * We will then get add_to_super called for each component, and then | |
5440 | * write_init_super called to write it out to each device. | |
5441 | * For IMSM, Create can create on fresh devices or on a pre-existing | |
5442 | * array. | |
5443 | * To create on a pre-existing array a different method will be called. | |
5444 | * This one is just for fresh drives. | |
5445 | */ | |
5446 | struct intel_super *super; | |
5447 | struct imsm_super *mpb; | |
5448 | size_t mpb_size; | |
4d1313e9 | 5449 | char *version; |
bf5a934a | 5450 | |
83cd1e97 | 5451 | if (data_offset != INVALID_SECTORS) { |
ed503f89 | 5452 | pr_err("data-offset not supported by imsm\n"); |
83cd1e97 N |
5453 | return 0; |
5454 | } | |
5455 | ||
bf5a934a | 5456 | if (st->sb) |
5308f117 | 5457 | return init_super_imsm_volume(st, info, s, name, homehost, uuid, |
83cd1e97 | 5458 | data_offset); |
e683ca88 DW |
5459 | |
5460 | if (info) | |
5461 | mpb_size = disks_to_mpb_size(info->nr_disks); | |
5462 | else | |
f36a9ecd | 5463 | mpb_size = MAX_SECTOR_SIZE; |
bf5a934a | 5464 | |
49133e57 | 5465 | super = alloc_super(); |
f36a9ecd PB |
5466 | if (super && |
5467 | posix_memalign(&super->buf, MAX_SECTOR_SIZE, mpb_size) != 0) { | |
8d67477f | 5468 | free_imsm(super); |
e683ca88 DW |
5469 | super = NULL; |
5470 | } | |
5471 | if (!super) { | |
1ade5cc1 | 5472 | pr_err("could not allocate superblock\n"); |
bf5a934a DW |
5473 | return 0; |
5474 | } | |
de44e46f PB |
5475 | if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE, |
5476 | MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) { | |
1ade5cc1 | 5477 | pr_err("could not allocate migr_rec buffer\n"); |
8e59f3d8 | 5478 | free(super->buf); |
8d67477f | 5479 | free_imsm(super); |
8e59f3d8 AK |
5480 | return 0; |
5481 | } | |
e683ca88 | 5482 | memset(super->buf, 0, mpb_size); |
ef649044 | 5483 | mpb = super->buf; |
e683ca88 DW |
5484 | mpb->mpb_size = __cpu_to_le32(mpb_size); |
5485 | st->sb = super; | |
5486 | ||
5487 | if (info == NULL) { | |
5488 | /* zeroing superblock */ | |
5489 | return 0; | |
5490 | } | |
bf5a934a | 5491 | |
4d1313e9 DW |
5492 | mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; |
5493 | ||
5494 | version = (char *) mpb->sig; | |
5495 | strcpy(version, MPB_SIGNATURE); | |
5496 | version += strlen(MPB_SIGNATURE); | |
5497 | strcpy(version, MPB_VERSION_RAID0); | |
bf5a934a | 5498 | |
bf5a934a DW |
5499 | return 1; |
5500 | } | |
5501 | ||
f2cc4f7d AO |
5502 | static int drive_validate_sector_size(struct intel_super *super, struct dl *dl) |
5503 | { | |
5504 | unsigned int member_sector_size; | |
5505 | ||
5506 | if (dl->fd < 0) { | |
5507 | pr_err("Invalid file descriptor for %s\n", dl->devname); | |
5508 | return 0; | |
5509 | } | |
5510 | ||
5511 | if (!get_dev_sector_size(dl->fd, dl->devname, &member_sector_size)) | |
5512 | return 0; | |
5513 | if (member_sector_size != super->sector_size) | |
5514 | return 0; | |
5515 | return 1; | |
5516 | } | |
5517 | ||
f20c3968 | 5518 | static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk, |
bf5a934a DW |
5519 | int fd, char *devname) |
5520 | { | |
5521 | struct intel_super *super = st->sb; | |
d23fe947 | 5522 | struct imsm_super *mpb = super->anchor; |
3960e579 | 5523 | struct imsm_disk *_disk; |
bf5a934a DW |
5524 | struct imsm_dev *dev; |
5525 | struct imsm_map *map; | |
3960e579 | 5526 | struct dl *dl, *df; |
4eb26970 | 5527 | int slot; |
bf5a934a | 5528 | |
949c47a0 | 5529 | dev = get_imsm_dev(super, super->current_vol); |
238c0a71 | 5530 | map = get_imsm_map(dev, MAP_0); |
bf5a934a | 5531 | |
208933a7 | 5532 | if (! (dk->state & (1<<MD_DISK_SYNC))) { |
e7b84f9d | 5533 | pr_err("%s: Cannot add spare devices to IMSM volume\n", |
208933a7 N |
5534 | devname); |
5535 | return 1; | |
5536 | } | |
5537 | ||
efb30e7f DW |
5538 | if (fd == -1) { |
5539 | /* we're doing autolayout so grab the pre-marked (in | |
5540 | * validate_geometry) raid_disk | |
5541 | */ | |
5542 | for (dl = super->disks; dl; dl = dl->next) | |
5543 | if (dl->raiddisk == dk->raid_disk) | |
5544 | break; | |
5545 | } else { | |
5546 | for (dl = super->disks; dl ; dl = dl->next) | |
5547 | if (dl->major == dk->major && | |
5548 | dl->minor == dk->minor) | |
5549 | break; | |
5550 | } | |
d23fe947 | 5551 | |
208933a7 | 5552 | if (!dl) { |
e7b84f9d | 5553 | pr_err("%s is not a member of the same container\n", devname); |
f20c3968 | 5554 | return 1; |
208933a7 | 5555 | } |
bf5a934a | 5556 | |
f2cc4f7d AO |
5557 | if (!drive_validate_sector_size(super, dl)) { |
5558 | pr_err("Combining drives of different sector size in one volume is not allowed\n"); | |
5559 | return 1; | |
5560 | } | |
5561 | ||
d23fe947 DW |
5562 | /* add a pristine spare to the metadata */ |
5563 | if (dl->index < 0) { | |
5564 | dl->index = super->anchor->num_disks; | |
5565 | super->anchor->num_disks++; | |
5566 | } | |
4eb26970 DW |
5567 | /* Check the device has not already been added */ |
5568 | slot = get_imsm_disk_slot(map, dl->index); | |
5569 | if (slot >= 0 && | |
238c0a71 | 5570 | (get_imsm_ord_tbl_ent(dev, slot, MAP_X) & IMSM_ORD_REBUILD) == 0) { |
e7b84f9d | 5571 | pr_err("%s has been included in this array twice\n", |
4eb26970 DW |
5572 | devname); |
5573 | return 1; | |
5574 | } | |
656b6b5a | 5575 | set_imsm_ord_tbl_ent(map, dk->raid_disk, dl->index); |
ee5aad5a | 5576 | dl->disk.status = CONFIGURED_DISK; |
d23fe947 | 5577 | |
3960e579 DW |
5578 | /* update size of 'missing' disks to be at least as large as the |
5579 | * largest acitve member (we only have dummy missing disks when | |
5580 | * creating the first volume) | |
5581 | */ | |
5582 | if (super->current_vol == 0) { | |
5583 | for (df = super->missing; df; df = df->next) { | |
5551b113 CA |
5584 | if (total_blocks(&dl->disk) > total_blocks(&df->disk)) |
5585 | set_total_blocks(&df->disk, total_blocks(&dl->disk)); | |
3960e579 DW |
5586 | _disk = __get_imsm_disk(mpb, df->index); |
5587 | *_disk = df->disk; | |
5588 | } | |
5589 | } | |
5590 | ||
5591 | /* refresh unset/failed slots to point to valid 'missing' entries */ | |
5592 | for (df = super->missing; df; df = df->next) | |
5593 | for (slot = 0; slot < mpb->num_disks; slot++) { | |
238c0a71 | 5594 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); |
3960e579 DW |
5595 | |
5596 | if ((ord & IMSM_ORD_REBUILD) == 0) | |
5597 | continue; | |
5598 | set_imsm_ord_tbl_ent(map, slot, df->index | IMSM_ORD_REBUILD); | |
1ace8403 | 5599 | if (is_gen_migration(dev)) { |
238c0a71 AK |
5600 | struct imsm_map *map2 = get_imsm_map(dev, |
5601 | MAP_1); | |
0a108d63 | 5602 | int slot2 = get_imsm_disk_slot(map2, df->index); |
089f9d79 | 5603 | if (slot2 < map2->num_members && slot2 >= 0) { |
1ace8403 | 5604 | __u32 ord2 = get_imsm_ord_tbl_ent(dev, |
238c0a71 AK |
5605 | slot2, |
5606 | MAP_1); | |
1ace8403 AK |
5607 | if ((unsigned)df->index == |
5608 | ord_to_idx(ord2)) | |
5609 | set_imsm_ord_tbl_ent(map2, | |
0a108d63 | 5610 | slot2, |
1ace8403 AK |
5611 | df->index | |
5612 | IMSM_ORD_REBUILD); | |
5613 | } | |
5614 | } | |
3960e579 DW |
5615 | dprintf("set slot:%d to missing disk:%d\n", slot, df->index); |
5616 | break; | |
5617 | } | |
5618 | ||
d23fe947 DW |
5619 | /* if we are creating the first raid device update the family number */ |
5620 | if (super->current_vol == 0) { | |
5621 | __u32 sum; | |
5622 | struct imsm_dev *_dev = __get_imsm_dev(mpb, 0); | |
d23fe947 | 5623 | |
3960e579 | 5624 | _disk = __get_imsm_disk(mpb, dl->index); |
791b666a | 5625 | if (!_dev || !_disk) { |
e7b84f9d | 5626 | pr_err("BUG mpb setup error\n"); |
791b666a AW |
5627 | return 1; |
5628 | } | |
d23fe947 DW |
5629 | *_dev = *dev; |
5630 | *_disk = dl->disk; | |
148acb7b DW |
5631 | sum = random32(); |
5632 | sum += __gen_imsm_checksum(mpb); | |
d23fe947 | 5633 | mpb->family_num = __cpu_to_le32(sum); |
148acb7b | 5634 | mpb->orig_family_num = mpb->family_num; |
d23fe947 | 5635 | } |
ca0748fa | 5636 | super->current_disk = dl; |
f20c3968 | 5637 | return 0; |
bf5a934a DW |
5638 | } |
5639 | ||
a8619d23 AK |
5640 | /* mark_spare() |
5641 | * Function marks disk as spare and restores disk serial | |
5642 | * in case it was previously marked as failed by takeover operation | |
5643 | * reruns: | |
5644 | * -1 : critical error | |
5645 | * 0 : disk is marked as spare but serial is not set | |
5646 | * 1 : success | |
5647 | */ | |
5648 | int mark_spare(struct dl *disk) | |
5649 | { | |
5650 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
5651 | int ret_val = -1; | |
5652 | ||
5653 | if (!disk) | |
5654 | return ret_val; | |
5655 | ||
5656 | ret_val = 0; | |
5657 | if (!imsm_read_serial(disk->fd, NULL, serial)) { | |
5658 | /* Restore disk serial number, because takeover marks disk | |
5659 | * as failed and adds to serial ':0' before it becomes | |
5660 | * a spare disk. | |
5661 | */ | |
5662 | serialcpy(disk->serial, serial); | |
5663 | serialcpy(disk->disk.serial, serial); | |
5664 | ret_val = 1; | |
5665 | } | |
5666 | disk->disk.status = SPARE_DISK; | |
5667 | disk->index = -1; | |
5668 | ||
5669 | return ret_val; | |
5670 | } | |
88654014 | 5671 | |
f20c3968 | 5672 | static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk, |
72ca9bcf N |
5673 | int fd, char *devname, |
5674 | unsigned long long data_offset) | |
cdddbdbc | 5675 | { |
c2c087e6 | 5676 | struct intel_super *super = st->sb; |
c2c087e6 DW |
5677 | struct dl *dd; |
5678 | unsigned long long size; | |
fa7bb6f8 | 5679 | unsigned int member_sector_size; |
f2f27e63 | 5680 | __u32 id; |
c2c087e6 DW |
5681 | int rv; |
5682 | struct stat stb; | |
5683 | ||
88654014 LM |
5684 | /* If we are on an RAID enabled platform check that the disk is |
5685 | * attached to the raid controller. | |
5686 | * We do not need to test disks attachment for container based additions, | |
5687 | * they shall be already tested when container was created/assembled. | |
88c32bb1 | 5688 | */ |
d424212e | 5689 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 5690 | /* no orom/efi or non-intel hba of the disk */ |
f0f5a016 LM |
5691 | if (rv != 0) { |
5692 | dprintf("capability: %p fd: %d ret: %d\n", | |
5693 | super->orom, fd, rv); | |
5694 | return 1; | |
88c32bb1 DW |
5695 | } |
5696 | ||
f20c3968 DW |
5697 | if (super->current_vol >= 0) |
5698 | return add_to_super_imsm_volume(st, dk, fd, devname); | |
bf5a934a | 5699 | |
c2c087e6 | 5700 | fstat(fd, &stb); |
503975b9 | 5701 | dd = xcalloc(sizeof(*dd), 1); |
c2c087e6 DW |
5702 | dd->major = major(stb.st_rdev); |
5703 | dd->minor = minor(stb.st_rdev); | |
503975b9 | 5704 | dd->devname = devname ? xstrdup(devname) : NULL; |
c2c087e6 | 5705 | dd->fd = fd; |
689c9bf3 | 5706 | dd->e = NULL; |
1a64be56 | 5707 | dd->action = DISK_ADD; |
c2c087e6 | 5708 | rv = imsm_read_serial(fd, devname, dd->serial); |
32ba9157 | 5709 | if (rv) { |
e7b84f9d | 5710 | pr_err("failed to retrieve scsi serial, aborting\n"); |
20bee0f8 PB |
5711 | if (dd->devname) |
5712 | free(dd->devname); | |
949c47a0 | 5713 | free(dd); |
0030e8d6 | 5714 | abort(); |
c2c087e6 | 5715 | } |
20bee0f8 PB |
5716 | if (super->hba && ((super->hba->type == SYS_DEV_NVME) || |
5717 | (super->hba->type == SYS_DEV_VMD))) { | |
5718 | int i; | |
5719 | char *devpath = diskfd_to_devpath(fd); | |
5720 | char controller_path[PATH_MAX]; | |
5721 | ||
5722 | if (!devpath) { | |
5723 | pr_err("failed to get devpath, aborting\n"); | |
5724 | if (dd->devname) | |
5725 | free(dd->devname); | |
5726 | free(dd); | |
5727 | return 1; | |
5728 | } | |
5729 | ||
5730 | snprintf(controller_path, PATH_MAX-1, "%s/device", devpath); | |
5731 | free(devpath); | |
5732 | ||
5733 | if (devpath_to_vendor(controller_path) == 0x8086) { | |
5734 | /* | |
5735 | * If Intel's NVMe drive has serial ended with | |
5736 | * "-A","-B","-1" or "-2" it means that this is "x8" | |
5737 | * device (double drive on single PCIe card). | |
5738 | * User should be warned about potential data loss. | |
5739 | */ | |
5740 | for (i = MAX_RAID_SERIAL_LEN-1; i > 0; i--) { | |
5741 | /* Skip empty character at the end */ | |
5742 | if (dd->serial[i] == 0) | |
5743 | continue; | |
5744 | ||
5745 | if (((dd->serial[i] == 'A') || | |
5746 | (dd->serial[i] == 'B') || | |
5747 | (dd->serial[i] == '1') || | |
5748 | (dd->serial[i] == '2')) && | |
5749 | (dd->serial[i-1] == '-')) | |
5750 | pr_err("\tThe action you are about to take may put your data at risk.\n" | |
5751 | "\tPlease note that x8 devices may consist of two separate x4 devices " | |
5752 | "located on a single PCIe port.\n" | |
5753 | "\tRAID 0 is the only supported configuration for this type of x8 device.\n"); | |
5754 | break; | |
5755 | } | |
32716c51 PB |
5756 | } else if (super->hba->type == SYS_DEV_VMD && super->orom && |
5757 | !imsm_orom_has_tpv_support(super->orom)) { | |
5758 | pr_err("\tPlatform configuration does not support non-Intel NVMe drives.\n" | |
5759 | "\tPlease refer to Intel(R) RSTe user guide.\n"); | |
5760 | free(dd->devname); | |
5761 | free(dd); | |
5762 | return 1; | |
20bee0f8 PB |
5763 | } |
5764 | } | |
c2c087e6 | 5765 | |
c2c087e6 | 5766 | get_dev_size(fd, NULL, &size); |
fa7bb6f8 PB |
5767 | get_dev_sector_size(fd, NULL, &member_sector_size); |
5768 | ||
5769 | if (super->sector_size == 0) { | |
5770 | /* this a first device, so sector_size is not set yet */ | |
5771 | super->sector_size = member_sector_size; | |
fa7bb6f8 PB |
5772 | } |
5773 | ||
71e5411e | 5774 | /* clear migr_rec when adding disk to container */ |
85337573 AO |
5775 | memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE); |
5776 | if (lseek64(fd, size - MIGR_REC_SECTOR_POSITION*member_sector_size, | |
de44e46f | 5777 | SEEK_SET) >= 0) { |
466070ad | 5778 | if ((unsigned int)write(fd, super->migr_rec_buf, |
85337573 AO |
5779 | MIGR_REC_BUF_SECTORS*member_sector_size) != |
5780 | MIGR_REC_BUF_SECTORS*member_sector_size) | |
71e5411e PB |
5781 | perror("Write migr_rec failed"); |
5782 | } | |
5783 | ||
c2c087e6 | 5784 | size /= 512; |
1f24f035 | 5785 | serialcpy(dd->disk.serial, dd->serial); |
5551b113 CA |
5786 | set_total_blocks(&dd->disk, size); |
5787 | if (__le32_to_cpu(dd->disk.total_blocks_hi) > 0) { | |
5788 | struct imsm_super *mpb = super->anchor; | |
5789 | mpb->attributes |= MPB_ATTRIB_2TB_DISK; | |
5790 | } | |
a8619d23 | 5791 | mark_spare(dd); |
c2c087e6 | 5792 | if (sysfs_disk_to_scsi_id(fd, &id) == 0) |
b9f594fe | 5793 | dd->disk.scsi_id = __cpu_to_le32(id); |
c2c087e6 | 5794 | else |
b9f594fe | 5795 | dd->disk.scsi_id = __cpu_to_le32(0); |
43dad3d6 DW |
5796 | |
5797 | if (st->update_tail) { | |
1a64be56 LM |
5798 | dd->next = super->disk_mgmt_list; |
5799 | super->disk_mgmt_list = dd; | |
43dad3d6 DW |
5800 | } else { |
5801 | dd->next = super->disks; | |
5802 | super->disks = dd; | |
ceaf0ee1 | 5803 | super->updates_pending++; |
43dad3d6 | 5804 | } |
f20c3968 DW |
5805 | |
5806 | return 0; | |
cdddbdbc DW |
5807 | } |
5808 | ||
1a64be56 LM |
5809 | static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk) |
5810 | { | |
5811 | struct intel_super *super = st->sb; | |
5812 | struct dl *dd; | |
5813 | ||
5814 | /* remove from super works only in mdmon - for communication | |
5815 | * manager - monitor. Check if communication memory buffer | |
5816 | * is prepared. | |
5817 | */ | |
5818 | if (!st->update_tail) { | |
1ade5cc1 | 5819 | pr_err("shall be used in mdmon context only\n"); |
1a64be56 LM |
5820 | return 1; |
5821 | } | |
503975b9 | 5822 | dd = xcalloc(1, sizeof(*dd)); |
1a64be56 LM |
5823 | dd->major = dk->major; |
5824 | dd->minor = dk->minor; | |
1a64be56 | 5825 | dd->fd = -1; |
a8619d23 | 5826 | mark_spare(dd); |
1a64be56 LM |
5827 | dd->action = DISK_REMOVE; |
5828 | ||
5829 | dd->next = super->disk_mgmt_list; | |
5830 | super->disk_mgmt_list = dd; | |
5831 | ||
1a64be56 LM |
5832 | return 0; |
5833 | } | |
5834 | ||
f796af5d DW |
5835 | static int store_imsm_mpb(int fd, struct imsm_super *mpb); |
5836 | ||
5837 | static union { | |
f36a9ecd | 5838 | char buf[MAX_SECTOR_SIZE]; |
f796af5d | 5839 | struct imsm_super anchor; |
f36a9ecd | 5840 | } spare_record __attribute__ ((aligned(MAX_SECTOR_SIZE))); |
c2c087e6 | 5841 | |
d23fe947 DW |
5842 | /* spare records have their own family number and do not have any defined raid |
5843 | * devices | |
5844 | */ | |
5845 | static int write_super_imsm_spares(struct intel_super *super, int doclose) | |
5846 | { | |
d23fe947 | 5847 | struct imsm_super *mpb = super->anchor; |
f796af5d | 5848 | struct imsm_super *spare = &spare_record.anchor; |
d23fe947 DW |
5849 | __u32 sum; |
5850 | struct dl *d; | |
5851 | ||
68641cdb JS |
5852 | spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super)); |
5853 | spare->generation_num = __cpu_to_le32(1UL); | |
f796af5d | 5854 | spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; |
68641cdb JS |
5855 | spare->num_disks = 1; |
5856 | spare->num_raid_devs = 0; | |
5857 | spare->cache_size = mpb->cache_size; | |
5858 | spare->pwr_cycle_count = __cpu_to_le32(1); | |
f796af5d DW |
5859 | |
5860 | snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH, | |
5861 | MPB_SIGNATURE MPB_VERSION_RAID0); | |
d23fe947 DW |
5862 | |
5863 | for (d = super->disks; d; d = d->next) { | |
8796fdc4 | 5864 | if (d->index != -1) |
d23fe947 DW |
5865 | continue; |
5866 | ||
f796af5d | 5867 | spare->disk[0] = d->disk; |
027c374f CA |
5868 | if (__le32_to_cpu(d->disk.total_blocks_hi) > 0) |
5869 | spare->attributes |= MPB_ATTRIB_2TB_DISK; | |
5870 | ||
f36a9ecd PB |
5871 | if (super->sector_size == 4096) |
5872 | convert_to_4k_imsm_disk(&spare->disk[0]); | |
5873 | ||
f796af5d DW |
5874 | sum = __gen_imsm_checksum(spare); |
5875 | spare->family_num = __cpu_to_le32(sum); | |
5876 | spare->orig_family_num = 0; | |
5877 | sum = __gen_imsm_checksum(spare); | |
5878 | spare->check_sum = __cpu_to_le32(sum); | |
d23fe947 | 5879 | |
f796af5d | 5880 | if (store_imsm_mpb(d->fd, spare)) { |
1ade5cc1 N |
5881 | pr_err("failed for device %d:%d %s\n", |
5882 | d->major, d->minor, strerror(errno)); | |
e74255d9 | 5883 | return 1; |
d23fe947 DW |
5884 | } |
5885 | if (doclose) { | |
5886 | close(d->fd); | |
5887 | d->fd = -1; | |
5888 | } | |
5889 | } | |
5890 | ||
e74255d9 | 5891 | return 0; |
d23fe947 DW |
5892 | } |
5893 | ||
36988a3d | 5894 | static int write_super_imsm(struct supertype *st, int doclose) |
cdddbdbc | 5895 | { |
36988a3d | 5896 | struct intel_super *super = st->sb; |
f36a9ecd | 5897 | unsigned int sector_size = super->sector_size; |
949c47a0 | 5898 | struct imsm_super *mpb = super->anchor; |
c2c087e6 DW |
5899 | struct dl *d; |
5900 | __u32 generation; | |
5901 | __u32 sum; | |
d23fe947 | 5902 | int spares = 0; |
949c47a0 | 5903 | int i; |
a48ac0a8 | 5904 | __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk); |
36988a3d | 5905 | int num_disks = 0; |
146c6260 | 5906 | int clear_migration_record = 1; |
bbab0940 | 5907 | __u32 bbm_log_size; |
cdddbdbc | 5908 | |
c2c087e6 DW |
5909 | /* 'generation' is incremented everytime the metadata is written */ |
5910 | generation = __le32_to_cpu(mpb->generation_num); | |
5911 | generation++; | |
5912 | mpb->generation_num = __cpu_to_le32(generation); | |
5913 | ||
148acb7b DW |
5914 | /* fix up cases where previous mdadm releases failed to set |
5915 | * orig_family_num | |
5916 | */ | |
5917 | if (mpb->orig_family_num == 0) | |
5918 | mpb->orig_family_num = mpb->family_num; | |
5919 | ||
d23fe947 | 5920 | for (d = super->disks; d; d = d->next) { |
8796fdc4 | 5921 | if (d->index == -1) |
d23fe947 | 5922 | spares++; |
36988a3d | 5923 | else { |
d23fe947 | 5924 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
5925 | num_disks++; |
5926 | } | |
d23fe947 | 5927 | } |
36988a3d | 5928 | for (d = super->missing; d; d = d->next) { |
47ee5a45 | 5929 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
5930 | num_disks++; |
5931 | } | |
5932 | mpb->num_disks = num_disks; | |
5933 | mpb_size += sizeof(struct imsm_disk) * mpb->num_disks; | |
b9f594fe | 5934 | |
949c47a0 DW |
5935 | for (i = 0; i < mpb->num_raid_devs; i++) { |
5936 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
36988a3d AK |
5937 | struct imsm_dev *dev2 = get_imsm_dev(super, i); |
5938 | if (dev && dev2) { | |
5939 | imsm_copy_dev(dev, dev2); | |
5940 | mpb_size += sizeof_imsm_dev(dev, 0); | |
5941 | } | |
146c6260 AK |
5942 | if (is_gen_migration(dev2)) |
5943 | clear_migration_record = 0; | |
949c47a0 | 5944 | } |
bbab0940 TM |
5945 | |
5946 | bbm_log_size = get_imsm_bbm_log_size(super->bbm_log); | |
5947 | ||
5948 | if (bbm_log_size) { | |
5949 | memcpy((void *)mpb + mpb_size, super->bbm_log, bbm_log_size); | |
5950 | mpb->attributes |= MPB_ATTRIB_BBM; | |
5951 | } else | |
5952 | mpb->attributes &= ~MPB_ATTRIB_BBM; | |
5953 | ||
5954 | super->anchor->bbm_log_size = __cpu_to_le32(bbm_log_size); | |
5955 | mpb_size += bbm_log_size; | |
a48ac0a8 | 5956 | mpb->mpb_size = __cpu_to_le32(mpb_size); |
949c47a0 | 5957 | |
bbab0940 TM |
5958 | #ifdef DEBUG |
5959 | assert(super->len == 0 || mpb_size <= super->len); | |
5960 | #endif | |
5961 | ||
c2c087e6 | 5962 | /* recalculate checksum */ |
949c47a0 | 5963 | sum = __gen_imsm_checksum(mpb); |
c2c087e6 DW |
5964 | mpb->check_sum = __cpu_to_le32(sum); |
5965 | ||
51d83f5d AK |
5966 | if (super->clean_migration_record_by_mdmon) { |
5967 | clear_migration_record = 1; | |
5968 | super->clean_migration_record_by_mdmon = 0; | |
5969 | } | |
146c6260 | 5970 | if (clear_migration_record) |
de44e46f | 5971 | memset(super->migr_rec_buf, 0, |
85337573 | 5972 | MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE); |
146c6260 | 5973 | |
f36a9ecd PB |
5974 | if (sector_size == 4096) |
5975 | convert_to_4k(super); | |
5976 | ||
d23fe947 | 5977 | /* write the mpb for disks that compose raid devices */ |
c2c087e6 | 5978 | for (d = super->disks; d ; d = d->next) { |
86c54047 | 5979 | if (d->index < 0 || is_failed(&d->disk)) |
d23fe947 | 5980 | continue; |
30602f53 | 5981 | |
146c6260 AK |
5982 | if (clear_migration_record) { |
5983 | unsigned long long dsize; | |
5984 | ||
5985 | get_dev_size(d->fd, NULL, &dsize); | |
de44e46f PB |
5986 | if (lseek64(d->fd, dsize - sector_size, |
5987 | SEEK_SET) >= 0) { | |
466070ad PB |
5988 | if ((unsigned int)write(d->fd, |
5989 | super->migr_rec_buf, | |
de44e46f PB |
5990 | MIGR_REC_BUF_SECTORS*sector_size) != |
5991 | MIGR_REC_BUF_SECTORS*sector_size) | |
9e2d750d | 5992 | perror("Write migr_rec failed"); |
146c6260 AK |
5993 | } |
5994 | } | |
51d83f5d AK |
5995 | |
5996 | if (store_imsm_mpb(d->fd, mpb)) | |
5997 | fprintf(stderr, | |
1ade5cc1 N |
5998 | "failed for device %d:%d (fd: %d)%s\n", |
5999 | d->major, d->minor, | |
51d83f5d AK |
6000 | d->fd, strerror(errno)); |
6001 | ||
c2c087e6 DW |
6002 | if (doclose) { |
6003 | close(d->fd); | |
6004 | d->fd = -1; | |
6005 | } | |
6006 | } | |
6007 | ||
d23fe947 DW |
6008 | if (spares) |
6009 | return write_super_imsm_spares(super, doclose); | |
6010 | ||
e74255d9 | 6011 | return 0; |
c2c087e6 DW |
6012 | } |
6013 | ||
9b1fb677 | 6014 | static int create_array(struct supertype *st, int dev_idx) |
43dad3d6 DW |
6015 | { |
6016 | size_t len; | |
6017 | struct imsm_update_create_array *u; | |
6018 | struct intel_super *super = st->sb; | |
9b1fb677 | 6019 | struct imsm_dev *dev = get_imsm_dev(super, dev_idx); |
238c0a71 | 6020 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
54c2c1ea DW |
6021 | struct disk_info *inf; |
6022 | struct imsm_disk *disk; | |
6023 | int i; | |
43dad3d6 | 6024 | |
54c2c1ea DW |
6025 | len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) + |
6026 | sizeof(*inf) * map->num_members; | |
503975b9 | 6027 | u = xmalloc(len); |
43dad3d6 | 6028 | u->type = update_create_array; |
9b1fb677 | 6029 | u->dev_idx = dev_idx; |
43dad3d6 | 6030 | imsm_copy_dev(&u->dev, dev); |
54c2c1ea DW |
6031 | inf = get_disk_info(u); |
6032 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 6033 | int idx = get_imsm_disk_idx(dev, i, MAP_X); |
9b1fb677 | 6034 | |
54c2c1ea | 6035 | disk = get_imsm_disk(super, idx); |
1ca5c8e0 N |
6036 | if (!disk) |
6037 | disk = get_imsm_missing(super, idx); | |
54c2c1ea DW |
6038 | serialcpy(inf[i].serial, disk->serial); |
6039 | } | |
43dad3d6 DW |
6040 | append_metadata_update(st, u, len); |
6041 | ||
6042 | return 0; | |
6043 | } | |
6044 | ||
1a64be56 | 6045 | static int mgmt_disk(struct supertype *st) |
43dad3d6 DW |
6046 | { |
6047 | struct intel_super *super = st->sb; | |
6048 | size_t len; | |
1a64be56 | 6049 | struct imsm_update_add_remove_disk *u; |
43dad3d6 | 6050 | |
1a64be56 | 6051 | if (!super->disk_mgmt_list) |
43dad3d6 DW |
6052 | return 0; |
6053 | ||
6054 | len = sizeof(*u); | |
503975b9 | 6055 | u = xmalloc(len); |
1a64be56 | 6056 | u->type = update_add_remove_disk; |
43dad3d6 DW |
6057 | append_metadata_update(st, u, len); |
6058 | ||
6059 | return 0; | |
6060 | } | |
2432ce9b AP |
6061 | |
6062 | __u32 crc32c_le(__u32 crc, unsigned char const *p, size_t len); | |
6063 | ||
6064 | static int write_init_ppl_imsm(struct supertype *st, struct mdinfo *info, int fd) | |
6065 | { | |
6066 | struct intel_super *super = st->sb; | |
6067 | void *buf; | |
6068 | struct ppl_header *ppl_hdr; | |
6069 | int ret; | |
6070 | ||
b2514242 PB |
6071 | /* first clear entire ppl space */ |
6072 | ret = zero_disk_range(fd, info->ppl_sector, info->ppl_size); | |
6073 | if (ret) | |
6074 | return ret; | |
6075 | ||
6076 | ret = posix_memalign(&buf, MAX_SECTOR_SIZE, PPL_HEADER_SIZE); | |
2432ce9b AP |
6077 | if (ret) { |
6078 | pr_err("Failed to allocate PPL header buffer\n"); | |
6079 | return ret; | |
6080 | } | |
6081 | ||
6082 | memset(buf, 0, PPL_HEADER_SIZE); | |
6083 | ppl_hdr = buf; | |
6084 | memset(ppl_hdr->reserved, 0xff, PPL_HDR_RESERVED); | |
6085 | ppl_hdr->signature = __cpu_to_le32(super->anchor->orig_family_num); | |
b23d0750 AP |
6086 | |
6087 | if (info->mismatch_cnt) { | |
6088 | /* | |
6089 | * We are overwriting an invalid ppl. Make one entry with wrong | |
6090 | * checksum to prevent the kernel from skipping resync. | |
6091 | */ | |
6092 | ppl_hdr->entries_count = __cpu_to_le32(1); | |
6093 | ppl_hdr->entries[0].checksum = ~0; | |
6094 | } | |
6095 | ||
2432ce9b AP |
6096 | ppl_hdr->checksum = __cpu_to_le32(~crc32c_le(~0, buf, PPL_HEADER_SIZE)); |
6097 | ||
6098 | if (lseek64(fd, info->ppl_sector * 512, SEEK_SET) < 0) { | |
6099 | ret = errno; | |
6100 | perror("Failed to seek to PPL header location"); | |
6101 | } | |
6102 | ||
6103 | if (!ret && write(fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) { | |
6104 | ret = errno; | |
6105 | perror("Write PPL header failed"); | |
6106 | } | |
6107 | ||
6108 | if (!ret) | |
6109 | fsync(fd); | |
6110 | ||
6111 | free(buf); | |
6112 | return ret; | |
6113 | } | |
6114 | ||
6115 | static int validate_ppl_imsm(struct supertype *st, struct mdinfo *info, | |
6116 | struct mdinfo *disk) | |
6117 | { | |
6118 | struct intel_super *super = st->sb; | |
6119 | struct dl *d; | |
6120 | void *buf; | |
6121 | int ret = 0; | |
6122 | struct ppl_header *ppl_hdr; | |
6123 | __u32 crc; | |
6124 | struct imsm_dev *dev; | |
2432ce9b | 6125 | __u32 idx; |
44b6b876 PB |
6126 | unsigned int i; |
6127 | unsigned long long ppl_offset = 0; | |
6128 | unsigned long long prev_gen_num = 0; | |
2432ce9b AP |
6129 | |
6130 | if (disk->disk.raid_disk < 0) | |
6131 | return 0; | |
6132 | ||
44b6b876 | 6133 | if (posix_memalign(&buf, MAX_SECTOR_SIZE, PPL_HEADER_SIZE)) { |
2432ce9b AP |
6134 | pr_err("Failed to allocate PPL header buffer\n"); |
6135 | return -1; | |
6136 | } | |
6137 | ||
6138 | dev = get_imsm_dev(super, info->container_member); | |
2fc0fc63 | 6139 | idx = get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_0); |
2432ce9b AP |
6140 | d = get_imsm_dl_disk(super, idx); |
6141 | ||
6142 | if (!d || d->index < 0 || is_failed(&d->disk)) | |
6143 | goto out; | |
6144 | ||
44b6b876 PB |
6145 | ret = 1; |
6146 | while (ppl_offset < MULTIPLE_PPL_AREA_SIZE_IMSM) { | |
6147 | dprintf("Checking potential PPL at offset: %llu\n", ppl_offset); | |
2432ce9b | 6148 | |
44b6b876 PB |
6149 | if (lseek64(d->fd, info->ppl_sector * 512 + ppl_offset, |
6150 | SEEK_SET) < 0) { | |
6151 | perror("Failed to seek to PPL header location"); | |
6152 | ret = -1; | |
6153 | goto out; | |
6154 | } | |
2432ce9b | 6155 | |
44b6b876 PB |
6156 | if (read(d->fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) { |
6157 | perror("Read PPL header failed"); | |
6158 | ret = -1; | |
6159 | goto out; | |
6160 | } | |
2432ce9b | 6161 | |
44b6b876 | 6162 | ppl_hdr = buf; |
2432ce9b | 6163 | |
44b6b876 PB |
6164 | crc = __le32_to_cpu(ppl_hdr->checksum); |
6165 | ppl_hdr->checksum = 0; | |
6166 | ||
6167 | if (crc != ~crc32c_le(~0, buf, PPL_HEADER_SIZE)) { | |
6168 | dprintf("Wrong PPL header checksum on %s\n", | |
6169 | d->devname); | |
6170 | goto out; | |
6171 | } | |
6172 | ||
6173 | if (prev_gen_num > __le64_to_cpu(ppl_hdr->generation)) { | |
6174 | /* previous was newest, it was already checked */ | |
6175 | goto out; | |
6176 | } | |
6177 | ||
6178 | if ((__le32_to_cpu(ppl_hdr->signature) != | |
6179 | super->anchor->orig_family_num)) { | |
6180 | dprintf("Wrong PPL header signature on %s\n", | |
6181 | d->devname); | |
6182 | ret = 1; | |
6183 | goto out; | |
6184 | } | |
6185 | ||
6186 | ret = 0; | |
6187 | prev_gen_num = __le64_to_cpu(ppl_hdr->generation); | |
2432ce9b | 6188 | |
44b6b876 PB |
6189 | ppl_offset += PPL_HEADER_SIZE; |
6190 | for (i = 0; i < __le32_to_cpu(ppl_hdr->entries_count); i++) | |
6191 | ppl_offset += | |
6192 | __le32_to_cpu(ppl_hdr->entries[i].pp_size); | |
2432ce9b AP |
6193 | } |
6194 | ||
6195 | out: | |
6196 | free(buf); | |
6197 | ||
54148aba PB |
6198 | /* |
6199 | * Update metadata to use mutliple PPLs area (1MB). | |
6200 | * This is done once for all RAID members | |
6201 | */ | |
6202 | if (info->consistency_policy == CONSISTENCY_POLICY_PPL && | |
6203 | info->ppl_size != (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9)) { | |
6204 | char subarray[20]; | |
6205 | struct mdinfo *member_dev; | |
6206 | ||
6207 | sprintf(subarray, "%d", info->container_member); | |
6208 | ||
6209 | if (mdmon_running(st->container_devnm)) | |
6210 | st->update_tail = &st->updates; | |
6211 | ||
6212 | if (st->ss->update_subarray(st, subarray, "ppl", NULL)) { | |
6213 | pr_err("Failed to update subarray %s\n", | |
6214 | subarray); | |
6215 | } else { | |
6216 | if (st->update_tail) | |
6217 | flush_metadata_updates(st); | |
6218 | else | |
6219 | st->ss->sync_metadata(st); | |
6220 | info->ppl_size = (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9); | |
6221 | for (member_dev = info->devs; member_dev; | |
6222 | member_dev = member_dev->next) | |
6223 | member_dev->ppl_size = | |
6224 | (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9); | |
6225 | } | |
6226 | } | |
6227 | ||
b23d0750 | 6228 | if (ret == 1) { |
2fc0fc63 AP |
6229 | struct imsm_map *map = get_imsm_map(dev, MAP_X); |
6230 | ||
50b9c10d PB |
6231 | if (map->map_state == IMSM_T_STATE_UNINITIALIZED || |
6232 | (map->map_state == IMSM_T_STATE_NORMAL && | |
2ec9d182 AP |
6233 | !(dev->vol.dirty & RAIDVOL_DIRTY)) || |
6234 | (dev->vol.migr_state == MIGR_REBUILD && | |
6235 | dev->vol.curr_migr_unit == 0 && | |
6236 | get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_1) != idx)) | |
b23d0750 AP |
6237 | ret = st->ss->write_init_ppl(st, info, d->fd); |
6238 | else | |
6239 | info->mismatch_cnt++; | |
6240 | } | |
2432ce9b AP |
6241 | |
6242 | return ret; | |
6243 | } | |
6244 | ||
2432ce9b AP |
6245 | static int write_init_ppl_imsm_all(struct supertype *st, struct mdinfo *info) |
6246 | { | |
6247 | struct intel_super *super = st->sb; | |
6248 | struct dl *d; | |
6249 | int ret = 0; | |
6250 | ||
6251 | if (info->consistency_policy != CONSISTENCY_POLICY_PPL || | |
6252 | info->array.level != 5) | |
6253 | return 0; | |
6254 | ||
6255 | for (d = super->disks; d ; d = d->next) { | |
6256 | if (d->index < 0 || is_failed(&d->disk)) | |
6257 | continue; | |
6258 | ||
6259 | ret = st->ss->write_init_ppl(st, info, d->fd); | |
6260 | if (ret) | |
6261 | break; | |
6262 | } | |
6263 | ||
6264 | return ret; | |
6265 | } | |
43dad3d6 | 6266 | |
c2c087e6 DW |
6267 | static int write_init_super_imsm(struct supertype *st) |
6268 | { | |
9b1fb677 DW |
6269 | struct intel_super *super = st->sb; |
6270 | int current_vol = super->current_vol; | |
2432ce9b AP |
6271 | int rv = 0; |
6272 | struct mdinfo info; | |
6273 | ||
6274 | getinfo_super_imsm(st, &info, NULL); | |
9b1fb677 DW |
6275 | |
6276 | /* we are done with current_vol reset it to point st at the container */ | |
6277 | super->current_vol = -1; | |
6278 | ||
8273f55e | 6279 | if (st->update_tail) { |
43dad3d6 DW |
6280 | /* queue the recently created array / added disk |
6281 | * as a metadata update */ | |
8273f55e | 6282 | |
43dad3d6 | 6283 | /* determine if we are creating a volume or adding a disk */ |
9b1fb677 | 6284 | if (current_vol < 0) { |
1a64be56 LM |
6285 | /* in the mgmt (add/remove) disk case we are running |
6286 | * in mdmon context, so don't close fd's | |
43dad3d6 | 6287 | */ |
2432ce9b AP |
6288 | rv = mgmt_disk(st); |
6289 | } else { | |
6290 | rv = write_init_ppl_imsm_all(st, &info); | |
6291 | if (!rv) | |
6292 | rv = create_array(st, current_vol); | |
6293 | } | |
d682f344 N |
6294 | } else { |
6295 | struct dl *d; | |
6296 | for (d = super->disks; d; d = d->next) | |
ba728be7 | 6297 | Kill(d->devname, NULL, 0, -1, 1); |
2432ce9b AP |
6298 | if (current_vol >= 0) |
6299 | rv = write_init_ppl_imsm_all(st, &info); | |
6300 | if (!rv) | |
6301 | rv = write_super_imsm(st, 1); | |
d682f344 | 6302 | } |
2432ce9b AP |
6303 | |
6304 | return rv; | |
cdddbdbc DW |
6305 | } |
6306 | ||
e683ca88 | 6307 | static int store_super_imsm(struct supertype *st, int fd) |
cdddbdbc | 6308 | { |
e683ca88 DW |
6309 | struct intel_super *super = st->sb; |
6310 | struct imsm_super *mpb = super ? super->anchor : NULL; | |
551c80c1 | 6311 | |
e683ca88 | 6312 | if (!mpb) |
ad97895e DW |
6313 | return 1; |
6314 | ||
f36a9ecd PB |
6315 | if (super->sector_size == 4096) |
6316 | convert_to_4k(super); | |
e683ca88 | 6317 | return store_imsm_mpb(fd, mpb); |
cdddbdbc DW |
6318 | } |
6319 | ||
cdddbdbc DW |
6320 | static int validate_geometry_imsm_container(struct supertype *st, int level, |
6321 | int layout, int raiddisks, int chunk, | |
af4348dd N |
6322 | unsigned long long size, |
6323 | unsigned long long data_offset, | |
6324 | char *dev, | |
2c514b71 NB |
6325 | unsigned long long *freesize, |
6326 | int verbose) | |
cdddbdbc | 6327 | { |
c2c087e6 DW |
6328 | int fd; |
6329 | unsigned long long ldsize; | |
594dc1b8 | 6330 | struct intel_super *super; |
f2f5c343 | 6331 | int rv = 0; |
cdddbdbc | 6332 | |
c2c087e6 DW |
6333 | if (level != LEVEL_CONTAINER) |
6334 | return 0; | |
6335 | if (!dev) | |
6336 | return 1; | |
6337 | ||
6338 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
6339 | if (fd < 0) { | |
ba728be7 | 6340 | if (verbose > 0) |
e7b84f9d | 6341 | pr_err("imsm: Cannot open %s: %s\n", |
2c514b71 | 6342 | dev, strerror(errno)); |
c2c087e6 DW |
6343 | return 0; |
6344 | } | |
6345 | if (!get_dev_size(fd, dev, &ldsize)) { | |
6346 | close(fd); | |
6347 | return 0; | |
6348 | } | |
f2f5c343 LM |
6349 | |
6350 | /* capabilities retrieve could be possible | |
6351 | * note that there is no fd for the disks in array. | |
6352 | */ | |
6353 | super = alloc_super(); | |
8d67477f TM |
6354 | if (!super) { |
6355 | close(fd); | |
6356 | return 0; | |
6357 | } | |
fa7bb6f8 PB |
6358 | if (!get_dev_sector_size(fd, NULL, &super->sector_size)) { |
6359 | close(fd); | |
6360 | free_imsm(super); | |
6361 | return 0; | |
6362 | } | |
6363 | ||
ba728be7 | 6364 | rv = find_intel_hba_capability(fd, super, verbose > 0 ? dev : NULL); |
f2f5c343 LM |
6365 | if (rv != 0) { |
6366 | #if DEBUG | |
6367 | char str[256]; | |
6368 | fd2devname(fd, str); | |
1ade5cc1 | 6369 | dprintf("fd: %d %s orom: %p rv: %d raiddisk: %d\n", |
f2f5c343 LM |
6370 | fd, str, super->orom, rv, raiddisks); |
6371 | #endif | |
6372 | /* no orom/efi or non-intel hba of the disk */ | |
6373 | close(fd); | |
6374 | free_imsm(super); | |
6375 | return 0; | |
6376 | } | |
c2c087e6 | 6377 | close(fd); |
9126b9a8 CA |
6378 | if (super->orom) { |
6379 | if (raiddisks > super->orom->tds) { | |
6380 | if (verbose) | |
7a862a02 | 6381 | pr_err("%d exceeds maximum number of platform supported disks: %d\n", |
9126b9a8 CA |
6382 | raiddisks, super->orom->tds); |
6383 | free_imsm(super); | |
6384 | return 0; | |
6385 | } | |
6386 | if ((super->orom->attr & IMSM_OROM_ATTR_2TB_DISK) == 0 && | |
6387 | (ldsize >> 9) >> 32 > 0) { | |
6388 | if (verbose) | |
e7b84f9d | 6389 | pr_err("%s exceeds maximum platform supported size\n", dev); |
9126b9a8 CA |
6390 | free_imsm(super); |
6391 | return 0; | |
6392 | } | |
f2f5c343 | 6393 | } |
c2c087e6 | 6394 | |
af4348dd | 6395 | *freesize = avail_size_imsm(st, ldsize >> 9, data_offset); |
f2f5c343 | 6396 | free_imsm(super); |
c2c087e6 DW |
6397 | |
6398 | return 1; | |
cdddbdbc DW |
6399 | } |
6400 | ||
0dcecb2e DW |
6401 | static unsigned long long find_size(struct extent *e, int *idx, int num_extents) |
6402 | { | |
6403 | const unsigned long long base_start = e[*idx].start; | |
6404 | unsigned long long end = base_start + e[*idx].size; | |
6405 | int i; | |
6406 | ||
6407 | if (base_start == end) | |
6408 | return 0; | |
6409 | ||
6410 | *idx = *idx + 1; | |
6411 | for (i = *idx; i < num_extents; i++) { | |
6412 | /* extend overlapping extents */ | |
6413 | if (e[i].start >= base_start && | |
6414 | e[i].start <= end) { | |
6415 | if (e[i].size == 0) | |
6416 | return 0; | |
6417 | if (e[i].start + e[i].size > end) | |
6418 | end = e[i].start + e[i].size; | |
6419 | } else if (e[i].start > end) { | |
6420 | *idx = i; | |
6421 | break; | |
6422 | } | |
6423 | } | |
6424 | ||
6425 | return end - base_start; | |
6426 | } | |
6427 | ||
6428 | static unsigned long long merge_extents(struct intel_super *super, int sum_extents) | |
6429 | { | |
6430 | /* build a composite disk with all known extents and generate a new | |
6431 | * 'maxsize' given the "all disks in an array must share a common start | |
6432 | * offset" constraint | |
6433 | */ | |
503975b9 | 6434 | struct extent *e = xcalloc(sum_extents, sizeof(*e)); |
0dcecb2e DW |
6435 | struct dl *dl; |
6436 | int i, j; | |
6437 | int start_extent; | |
6438 | unsigned long long pos; | |
b9d77223 | 6439 | unsigned long long start = 0; |
0dcecb2e DW |
6440 | unsigned long long maxsize; |
6441 | unsigned long reserve; | |
6442 | ||
0dcecb2e DW |
6443 | /* coalesce and sort all extents. also, check to see if we need to |
6444 | * reserve space between member arrays | |
6445 | */ | |
6446 | j = 0; | |
6447 | for (dl = super->disks; dl; dl = dl->next) { | |
6448 | if (!dl->e) | |
6449 | continue; | |
6450 | for (i = 0; i < dl->extent_cnt; i++) | |
6451 | e[j++] = dl->e[i]; | |
6452 | } | |
6453 | qsort(e, sum_extents, sizeof(*e), cmp_extent); | |
6454 | ||
6455 | /* merge extents */ | |
6456 | i = 0; | |
6457 | j = 0; | |
6458 | while (i < sum_extents) { | |
6459 | e[j].start = e[i].start; | |
6460 | e[j].size = find_size(e, &i, sum_extents); | |
6461 | j++; | |
6462 | if (e[j-1].size == 0) | |
6463 | break; | |
6464 | } | |
6465 | ||
6466 | pos = 0; | |
6467 | maxsize = 0; | |
6468 | start_extent = 0; | |
6469 | i = 0; | |
6470 | do { | |
6471 | unsigned long long esize; | |
6472 | ||
6473 | esize = e[i].start - pos; | |
6474 | if (esize >= maxsize) { | |
6475 | maxsize = esize; | |
6476 | start = pos; | |
6477 | start_extent = i; | |
6478 | } | |
6479 | pos = e[i].start + e[i].size; | |
6480 | i++; | |
6481 | } while (e[i-1].size); | |
6482 | free(e); | |
6483 | ||
a7dd165b DW |
6484 | if (maxsize == 0) |
6485 | return 0; | |
6486 | ||
6487 | /* FIXME assumes volume at offset 0 is the first volume in a | |
6488 | * container | |
6489 | */ | |
0dcecb2e DW |
6490 | if (start_extent > 0) |
6491 | reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */ | |
6492 | else | |
6493 | reserve = 0; | |
6494 | ||
6495 | if (maxsize < reserve) | |
a7dd165b | 6496 | return 0; |
0dcecb2e | 6497 | |
5551b113 | 6498 | super->create_offset = ~((unsigned long long) 0); |
0dcecb2e | 6499 | if (start + reserve > super->create_offset) |
a7dd165b | 6500 | return 0; /* start overflows create_offset */ |
0dcecb2e DW |
6501 | super->create_offset = start + reserve; |
6502 | ||
6503 | return maxsize - reserve; | |
6504 | } | |
6505 | ||
88c32bb1 DW |
6506 | static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks) |
6507 | { | |
6508 | if (level < 0 || level == 6 || level == 4) | |
6509 | return 0; | |
6510 | ||
6511 | /* if we have an orom prevent invalid raid levels */ | |
6512 | if (orom) | |
6513 | switch (level) { | |
6514 | case 0: return imsm_orom_has_raid0(orom); | |
6515 | case 1: | |
6516 | if (raiddisks > 2) | |
6517 | return imsm_orom_has_raid1e(orom); | |
1c556e92 DW |
6518 | return imsm_orom_has_raid1(orom) && raiddisks == 2; |
6519 | case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4; | |
6520 | case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2; | |
88c32bb1 DW |
6521 | } |
6522 | else | |
6523 | return 1; /* not on an Intel RAID platform so anything goes */ | |
6524 | ||
6525 | return 0; | |
6526 | } | |
6527 | ||
ca9de185 LM |
6528 | static int |
6529 | active_arrays_by_format(char *name, char* hba, struct md_list **devlist, | |
6530 | int dpa, int verbose) | |
6531 | { | |
6532 | struct mdstat_ent *mdstat = mdstat_read(0, 0); | |
594dc1b8 | 6533 | struct mdstat_ent *memb; |
ca9de185 LM |
6534 | int count = 0; |
6535 | int num = 0; | |
594dc1b8 | 6536 | struct md_list *dv; |
ca9de185 LM |
6537 | int found; |
6538 | ||
6539 | for (memb = mdstat ; memb ; memb = memb->next) { | |
6540 | if (memb->metadata_version && | |
fc54fe7a | 6541 | (strncmp(memb->metadata_version, "external:", 9) == 0) && |
ca9de185 LM |
6542 | (strcmp(&memb->metadata_version[9], name) == 0) && |
6543 | !is_subarray(memb->metadata_version+9) && | |
6544 | memb->members) { | |
6545 | struct dev_member *dev = memb->members; | |
6546 | int fd = -1; | |
6547 | while(dev && (fd < 0)) { | |
503975b9 N |
6548 | char *path = xmalloc(strlen(dev->name) + strlen("/dev/") + 1); |
6549 | num = sprintf(path, "%s%s", "/dev/", dev->name); | |
6550 | if (num > 0) | |
6551 | fd = open(path, O_RDONLY, 0); | |
089f9d79 | 6552 | if (num <= 0 || fd < 0) { |
676e87a8 | 6553 | pr_vrb("Cannot open %s: %s\n", |
503975b9 | 6554 | dev->name, strerror(errno)); |
ca9de185 | 6555 | } |
503975b9 | 6556 | free(path); |
ca9de185 LM |
6557 | dev = dev->next; |
6558 | } | |
6559 | found = 0; | |
089f9d79 | 6560 | if (fd >= 0 && disk_attached_to_hba(fd, hba)) { |
ca9de185 LM |
6561 | struct mdstat_ent *vol; |
6562 | for (vol = mdstat ; vol ; vol = vol->next) { | |
089f9d79 | 6563 | if (vol->active > 0 && |
ca9de185 | 6564 | vol->metadata_version && |
9581efb1 | 6565 | is_container_member(vol, memb->devnm)) { |
ca9de185 LM |
6566 | found++; |
6567 | count++; | |
6568 | } | |
6569 | } | |
6570 | if (*devlist && (found < dpa)) { | |
503975b9 | 6571 | dv = xcalloc(1, sizeof(*dv)); |
9581efb1 N |
6572 | dv->devname = xmalloc(strlen(memb->devnm) + strlen("/dev/") + 1); |
6573 | sprintf(dv->devname, "%s%s", "/dev/", memb->devnm); | |
503975b9 N |
6574 | dv->found = found; |
6575 | dv->used = 0; | |
6576 | dv->next = *devlist; | |
6577 | *devlist = dv; | |
ca9de185 LM |
6578 | } |
6579 | } | |
6580 | if (fd >= 0) | |
6581 | close(fd); | |
6582 | } | |
6583 | } | |
6584 | free_mdstat(mdstat); | |
6585 | return count; | |
6586 | } | |
6587 | ||
6588 | #ifdef DEBUG_LOOP | |
6589 | static struct md_list* | |
6590 | get_loop_devices(void) | |
6591 | { | |
6592 | int i; | |
6593 | struct md_list *devlist = NULL; | |
594dc1b8 | 6594 | struct md_list *dv; |
ca9de185 LM |
6595 | |
6596 | for(i = 0; i < 12; i++) { | |
503975b9 N |
6597 | dv = xcalloc(1, sizeof(*dv)); |
6598 | dv->devname = xmalloc(40); | |
ca9de185 LM |
6599 | sprintf(dv->devname, "/dev/loop%d", i); |
6600 | dv->next = devlist; | |
6601 | devlist = dv; | |
6602 | } | |
6603 | return devlist; | |
6604 | } | |
6605 | #endif | |
6606 | ||
6607 | static struct md_list* | |
6608 | get_devices(const char *hba_path) | |
6609 | { | |
6610 | struct md_list *devlist = NULL; | |
594dc1b8 | 6611 | struct md_list *dv; |
ca9de185 LM |
6612 | struct dirent *ent; |
6613 | DIR *dir; | |
6614 | int err = 0; | |
6615 | ||
6616 | #if DEBUG_LOOP | |
6617 | devlist = get_loop_devices(); | |
6618 | return devlist; | |
6619 | #endif | |
6620 | /* scroll through /sys/dev/block looking for devices attached to | |
6621 | * this hba | |
6622 | */ | |
6623 | dir = opendir("/sys/dev/block"); | |
6624 | for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) { | |
6625 | int fd; | |
6626 | char buf[1024]; | |
6627 | int major, minor; | |
6628 | char *path = NULL; | |
6629 | if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2) | |
6630 | continue; | |
6631 | path = devt_to_devpath(makedev(major, minor)); | |
6632 | if (!path) | |
6633 | continue; | |
6634 | if (!path_attached_to_hba(path, hba_path)) { | |
6635 | free(path); | |
6636 | path = NULL; | |
6637 | continue; | |
6638 | } | |
6639 | free(path); | |
6640 | path = NULL; | |
6641 | fd = dev_open(ent->d_name, O_RDONLY); | |
6642 | if (fd >= 0) { | |
6643 | fd2devname(fd, buf); | |
6644 | close(fd); | |
6645 | } else { | |
e7b84f9d | 6646 | pr_err("cannot open device: %s\n", |
ca9de185 LM |
6647 | ent->d_name); |
6648 | continue; | |
6649 | } | |
6650 | ||
503975b9 N |
6651 | dv = xcalloc(1, sizeof(*dv)); |
6652 | dv->devname = xstrdup(buf); | |
ca9de185 LM |
6653 | dv->next = devlist; |
6654 | devlist = dv; | |
6655 | } | |
6656 | if (err) { | |
6657 | while(devlist) { | |
6658 | dv = devlist; | |
6659 | devlist = devlist->next; | |
6660 | free(dv->devname); | |
6661 | free(dv); | |
6662 | } | |
6663 | } | |
562aa102 | 6664 | closedir(dir); |
ca9de185 LM |
6665 | return devlist; |
6666 | } | |
6667 | ||
6668 | static int | |
6669 | count_volumes_list(struct md_list *devlist, char *homehost, | |
6670 | int verbose, int *found) | |
6671 | { | |
6672 | struct md_list *tmpdev; | |
6673 | int count = 0; | |
594dc1b8 | 6674 | struct supertype *st; |
ca9de185 LM |
6675 | |
6676 | /* first walk the list of devices to find a consistent set | |
6677 | * that match the criterea, if that is possible. | |
6678 | * We flag the ones we like with 'used'. | |
6679 | */ | |
6680 | *found = 0; | |
6681 | st = match_metadata_desc_imsm("imsm"); | |
6682 | if (st == NULL) { | |
676e87a8 | 6683 | pr_vrb("cannot allocate memory for imsm supertype\n"); |
ca9de185 LM |
6684 | return 0; |
6685 | } | |
6686 | ||
6687 | for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { | |
6688 | char *devname = tmpdev->devname; | |
0a6bff09 | 6689 | dev_t rdev; |
ca9de185 LM |
6690 | struct supertype *tst; |
6691 | int dfd; | |
6692 | if (tmpdev->used > 1) | |
6693 | continue; | |
6694 | tst = dup_super(st); | |
6695 | if (tst == NULL) { | |
676e87a8 | 6696 | pr_vrb("cannot allocate memory for imsm supertype\n"); |
ca9de185 LM |
6697 | goto err_1; |
6698 | } | |
6699 | tmpdev->container = 0; | |
6700 | dfd = dev_open(devname, O_RDONLY|O_EXCL); | |
6701 | if (dfd < 0) { | |
1ade5cc1 | 6702 | dprintf("cannot open device %s: %s\n", |
ca9de185 LM |
6703 | devname, strerror(errno)); |
6704 | tmpdev->used = 2; | |
0a6bff09 | 6705 | } else if (!fstat_is_blkdev(dfd, devname, &rdev)) { |
ca9de185 LM |
6706 | tmpdev->used = 2; |
6707 | } else if (must_be_container(dfd)) { | |
6708 | struct supertype *cst; | |
6709 | cst = super_by_fd(dfd, NULL); | |
6710 | if (cst == NULL) { | |
1ade5cc1 | 6711 | dprintf("cannot recognize container type %s\n", |
ca9de185 LM |
6712 | devname); |
6713 | tmpdev->used = 2; | |
6714 | } else if (tst->ss != st->ss) { | |
1ade5cc1 | 6715 | dprintf("non-imsm container - ignore it: %s\n", |
ca9de185 LM |
6716 | devname); |
6717 | tmpdev->used = 2; | |
6718 | } else if (!tst->ss->load_container || | |
6719 | tst->ss->load_container(tst, dfd, NULL)) | |
6720 | tmpdev->used = 2; | |
6721 | else { | |
6722 | tmpdev->container = 1; | |
6723 | } | |
6724 | if (cst) | |
6725 | cst->ss->free_super(cst); | |
6726 | } else { | |
0a6bff09 | 6727 | tmpdev->st_rdev = rdev; |
ca9de185 | 6728 | if (tst->ss->load_super(tst,dfd, NULL)) { |
1ade5cc1 | 6729 | dprintf("no RAID superblock on %s\n", |
ca9de185 LM |
6730 | devname); |
6731 | tmpdev->used = 2; | |
6732 | } else if (tst->ss->compare_super == NULL) { | |
1ade5cc1 | 6733 | dprintf("Cannot assemble %s metadata on %s\n", |
ca9de185 LM |
6734 | tst->ss->name, devname); |
6735 | tmpdev->used = 2; | |
6736 | } | |
6737 | } | |
6738 | if (dfd >= 0) | |
6739 | close(dfd); | |
6740 | if (tmpdev->used == 2 || tmpdev->used == 4) { | |
6741 | /* Ignore unrecognised devices during auto-assembly */ | |
6742 | goto loop; | |
6743 | } | |
6744 | else { | |
6745 | struct mdinfo info; | |
6746 | tst->ss->getinfo_super(tst, &info, NULL); | |
6747 | ||
6748 | if (st->minor_version == -1) | |
6749 | st->minor_version = tst->minor_version; | |
6750 | ||
6751 | if (memcmp(info.uuid, uuid_zero, | |
6752 | sizeof(int[4])) == 0) { | |
6753 | /* this is a floating spare. It cannot define | |
6754 | * an array unless there are no more arrays of | |
6755 | * this type to be found. It can be included | |
6756 | * in an array of this type though. | |
6757 | */ | |
6758 | tmpdev->used = 3; | |
6759 | goto loop; | |
6760 | } | |
6761 | ||
6762 | if (st->ss != tst->ss || | |
6763 | st->minor_version != tst->minor_version || | |
6764 | st->ss->compare_super(st, tst) != 0) { | |
6765 | /* Some mismatch. If exactly one array matches this host, | |
6766 | * we can resolve on that one. | |
6767 | * Or, if we are auto assembling, we just ignore the second | |
6768 | * for now. | |
6769 | */ | |
1ade5cc1 | 6770 | dprintf("superblock on %s doesn't match others - assembly aborted\n", |
ca9de185 LM |
6771 | devname); |
6772 | goto loop; | |
6773 | } | |
6774 | tmpdev->used = 1; | |
6775 | *found = 1; | |
6776 | dprintf("found: devname: %s\n", devname); | |
6777 | } | |
6778 | loop: | |
6779 | if (tst) | |
6780 | tst->ss->free_super(tst); | |
6781 | } | |
6782 | if (*found != 0) { | |
6783 | int err; | |
6784 | if ((err = load_super_imsm_all(st, -1, &st->sb, NULL, devlist, 0)) == 0) { | |
6785 | struct mdinfo *iter, *head = st->ss->container_content(st, NULL); | |
6786 | for (iter = head; iter; iter = iter->next) { | |
6787 | dprintf("content->text_version: %s vol\n", | |
6788 | iter->text_version); | |
6789 | if (iter->array.state & (1<<MD_SB_BLOCK_VOLUME)) { | |
6790 | /* do not assemble arrays with unsupported | |
6791 | configurations */ | |
1ade5cc1 | 6792 | dprintf("Cannot activate member %s.\n", |
ca9de185 LM |
6793 | iter->text_version); |
6794 | } else | |
6795 | count++; | |
6796 | } | |
6797 | sysfs_free(head); | |
6798 | ||
6799 | } else { | |
1ade5cc1 | 6800 | dprintf("No valid super block on device list: err: %d %p\n", |
ca9de185 LM |
6801 | err, st->sb); |
6802 | } | |
6803 | } else { | |
1ade5cc1 | 6804 | dprintf("no more devices to examine\n"); |
ca9de185 LM |
6805 | } |
6806 | ||
6807 | for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { | |
089f9d79 | 6808 | if (tmpdev->used == 1 && tmpdev->found) { |
ca9de185 LM |
6809 | if (count) { |
6810 | if (count < tmpdev->found) | |
6811 | count = 0; | |
6812 | else | |
6813 | count -= tmpdev->found; | |
6814 | } | |
6815 | } | |
6816 | if (tmpdev->used == 1) | |
6817 | tmpdev->used = 4; | |
6818 | } | |
6819 | err_1: | |
6820 | if (st) | |
6821 | st->ss->free_super(st); | |
6822 | return count; | |
6823 | } | |
6824 | ||
d3c11416 AO |
6825 | static int __count_volumes(char *hba_path, int dpa, int verbose, |
6826 | int cmp_hba_path) | |
ca9de185 | 6827 | { |
72a45777 | 6828 | struct sys_dev *idev, *intel_devices = find_intel_devices(); |
ca9de185 | 6829 | int count = 0; |
72a45777 PB |
6830 | const struct orom_entry *entry; |
6831 | struct devid_list *dv, *devid_list; | |
ca9de185 | 6832 | |
d3c11416 | 6833 | if (!hba_path) |
ca9de185 LM |
6834 | return 0; |
6835 | ||
72a45777 | 6836 | for (idev = intel_devices; idev; idev = idev->next) { |
d3c11416 AO |
6837 | if (strstr(idev->path, hba_path)) |
6838 | break; | |
72a45777 PB |
6839 | } |
6840 | ||
6841 | if (!idev || !idev->dev_id) | |
ca9de185 | 6842 | return 0; |
72a45777 PB |
6843 | |
6844 | entry = get_orom_entry_by_device_id(idev->dev_id); | |
6845 | ||
6846 | if (!entry || !entry->devid_list) | |
6847 | return 0; | |
6848 | ||
6849 | devid_list = entry->devid_list; | |
6850 | for (dv = devid_list; dv; dv = dv->next) { | |
594dc1b8 | 6851 | struct md_list *devlist; |
d3c11416 AO |
6852 | struct sys_dev *device = NULL; |
6853 | char *hpath; | |
72a45777 PB |
6854 | int found = 0; |
6855 | ||
d3c11416 AO |
6856 | if (cmp_hba_path) |
6857 | device = device_by_id_and_path(dv->devid, hba_path); | |
6858 | else | |
6859 | device = device_by_id(dv->devid); | |
6860 | ||
72a45777 | 6861 | if (device) |
d3c11416 | 6862 | hpath = device->path; |
72a45777 PB |
6863 | else |
6864 | return 0; | |
6865 | ||
d3c11416 | 6866 | devlist = get_devices(hpath); |
72a45777 PB |
6867 | /* if no intel devices return zero volumes */ |
6868 | if (devlist == NULL) | |
6869 | return 0; | |
6870 | ||
d3c11416 AO |
6871 | count += active_arrays_by_format("imsm", hpath, &devlist, dpa, |
6872 | verbose); | |
6873 | dprintf("path: %s active arrays: %d\n", hpath, count); | |
72a45777 PB |
6874 | if (devlist == NULL) |
6875 | return 0; | |
6876 | do { | |
6877 | found = 0; | |
6878 | count += count_volumes_list(devlist, | |
6879 | NULL, | |
6880 | verbose, | |
6881 | &found); | |
6882 | dprintf("found %d count: %d\n", found, count); | |
6883 | } while (found); | |
6884 | ||
d3c11416 | 6885 | dprintf("path: %s total number of volumes: %d\n", hpath, count); |
72a45777 PB |
6886 | |
6887 | while (devlist) { | |
6888 | struct md_list *dv = devlist; | |
6889 | devlist = devlist->next; | |
6890 | free(dv->devname); | |
6891 | free(dv); | |
6892 | } | |
ca9de185 LM |
6893 | } |
6894 | return count; | |
6895 | } | |
6896 | ||
d3c11416 AO |
6897 | static int count_volumes(struct intel_hba *hba, int dpa, int verbose) |
6898 | { | |
6899 | if (!hba) | |
6900 | return 0; | |
6901 | if (hba->type == SYS_DEV_VMD) { | |
6902 | struct sys_dev *dev; | |
6903 | int count = 0; | |
6904 | ||
6905 | for (dev = find_intel_devices(); dev; dev = dev->next) { | |
6906 | if (dev->type == SYS_DEV_VMD) | |
6907 | count += __count_volumes(dev->path, dpa, | |
6908 | verbose, 1); | |
6909 | } | |
6910 | return count; | |
6911 | } | |
6912 | return __count_volumes(hba->path, dpa, verbose, 0); | |
6913 | } | |
6914 | ||
cd9d1ac7 DW |
6915 | static int imsm_default_chunk(const struct imsm_orom *orom) |
6916 | { | |
6917 | /* up to 512 if the plaform supports it, otherwise the platform max. | |
6918 | * 128 if no platform detected | |
6919 | */ | |
6920 | int fs = max(7, orom ? fls(orom->sss) : 0); | |
6921 | ||
6922 | return min(512, (1 << fs)); | |
6923 | } | |
73408129 | 6924 | |
6592ce37 DW |
6925 | static int |
6926 | validate_geometry_imsm_orom(struct intel_super *super, int level, int layout, | |
2cc699af | 6927 | int raiddisks, int *chunk, unsigned long long size, int verbose) |
6592ce37 | 6928 | { |
660260d0 DW |
6929 | /* check/set platform and metadata limits/defaults */ |
6930 | if (super->orom && raiddisks > super->orom->dpa) { | |
676e87a8 | 6931 | pr_vrb("platform supports a maximum of %d disks per array\n", |
660260d0 | 6932 | super->orom->dpa); |
73408129 LM |
6933 | return 0; |
6934 | } | |
6935 | ||
5d500228 | 6936 | /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */ |
660260d0 | 6937 | if (!is_raid_level_supported(super->orom, level, raiddisks)) { |
676e87a8 | 6938 | pr_vrb("platform does not support raid%d with %d disk%s\n", |
6592ce37 DW |
6939 | level, raiddisks, raiddisks > 1 ? "s" : ""); |
6940 | return 0; | |
6941 | } | |
cd9d1ac7 | 6942 | |
7ccc4cc4 | 6943 | if (*chunk == 0 || *chunk == UnSet) |
cd9d1ac7 DW |
6944 | *chunk = imsm_default_chunk(super->orom); |
6945 | ||
7ccc4cc4 | 6946 | if (super->orom && !imsm_orom_has_chunk(super->orom, *chunk)) { |
676e87a8 | 6947 | pr_vrb("platform does not support a chunk size of: %d\n", *chunk); |
cd9d1ac7 | 6948 | return 0; |
6592ce37 | 6949 | } |
cd9d1ac7 | 6950 | |
6592ce37 DW |
6951 | if (layout != imsm_level_to_layout(level)) { |
6952 | if (level == 5) | |
676e87a8 | 6953 | pr_vrb("imsm raid 5 only supports the left-asymmetric layout\n"); |
6592ce37 | 6954 | else if (level == 10) |
676e87a8 | 6955 | pr_vrb("imsm raid 10 only supports the n2 layout\n"); |
6592ce37 | 6956 | else |
676e87a8 | 6957 | pr_vrb("imsm unknown layout %#x for this raid level %d\n", |
6592ce37 DW |
6958 | layout, level); |
6959 | return 0; | |
6960 | } | |
2cc699af | 6961 | |
7ccc4cc4 | 6962 | if (super->orom && (super->orom->attr & IMSM_OROM_ATTR_2TB) == 0 && |
2cc699af | 6963 | (calc_array_size(level, raiddisks, layout, *chunk, size) >> 32) > 0) { |
676e87a8 | 6964 | pr_vrb("platform does not support a volume size over 2TB\n"); |
2cc699af CA |
6965 | return 0; |
6966 | } | |
614902f6 | 6967 | |
6592ce37 DW |
6968 | return 1; |
6969 | } | |
6970 | ||
1011e834 | 6971 | /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd |
c2c087e6 DW |
6972 | * FIX ME add ahci details |
6973 | */ | |
8b353278 | 6974 | static int validate_geometry_imsm_volume(struct supertype *st, int level, |
c21e737b | 6975 | int layout, int raiddisks, int *chunk, |
af4348dd N |
6976 | unsigned long long size, |
6977 | unsigned long long data_offset, | |
6978 | char *dev, | |
2c514b71 NB |
6979 | unsigned long long *freesize, |
6980 | int verbose) | |
cdddbdbc | 6981 | { |
9e04ac1c | 6982 | dev_t rdev; |
c2c087e6 | 6983 | struct intel_super *super = st->sb; |
b2916f25 | 6984 | struct imsm_super *mpb; |
c2c087e6 DW |
6985 | struct dl *dl; |
6986 | unsigned long long pos = 0; | |
6987 | unsigned long long maxsize; | |
6988 | struct extent *e; | |
6989 | int i; | |
cdddbdbc | 6990 | |
88c32bb1 DW |
6991 | /* We must have the container info already read in. */ |
6992 | if (!super) | |
c2c087e6 DW |
6993 | return 0; |
6994 | ||
b2916f25 JS |
6995 | mpb = super->anchor; |
6996 | ||
2cc699af | 6997 | if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, size, verbose)) { |
7a862a02 | 6998 | pr_err("RAID gemetry validation failed. Cannot proceed with the action(s).\n"); |
c2c087e6 | 6999 | return 0; |
d54559f0 | 7000 | } |
c2c087e6 DW |
7001 | if (!dev) { |
7002 | /* General test: make sure there is space for | |
2da8544a DW |
7003 | * 'raiddisks' device extents of size 'size' at a given |
7004 | * offset | |
c2c087e6 | 7005 | */ |
e46273eb | 7006 | unsigned long long minsize = size; |
b7528a20 | 7007 | unsigned long long start_offset = MaxSector; |
c2c087e6 DW |
7008 | int dcnt = 0; |
7009 | if (minsize == 0) | |
7010 | minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
7011 | for (dl = super->disks; dl ; dl = dl->next) { | |
7012 | int found = 0; | |
7013 | ||
bf5a934a | 7014 | pos = 0; |
c2c087e6 DW |
7015 | i = 0; |
7016 | e = get_extents(super, dl); | |
7017 | if (!e) continue; | |
7018 | do { | |
7019 | unsigned long long esize; | |
7020 | esize = e[i].start - pos; | |
7021 | if (esize >= minsize) | |
7022 | found = 1; | |
b7528a20 | 7023 | if (found && start_offset == MaxSector) { |
2da8544a DW |
7024 | start_offset = pos; |
7025 | break; | |
7026 | } else if (found && pos != start_offset) { | |
7027 | found = 0; | |
7028 | break; | |
7029 | } | |
c2c087e6 DW |
7030 | pos = e[i].start + e[i].size; |
7031 | i++; | |
7032 | } while (e[i-1].size); | |
7033 | if (found) | |
7034 | dcnt++; | |
7035 | free(e); | |
7036 | } | |
7037 | if (dcnt < raiddisks) { | |
2c514b71 | 7038 | if (verbose) |
7a862a02 | 7039 | pr_err("imsm: Not enough devices with space for this array (%d < %d)\n", |
2c514b71 | 7040 | dcnt, raiddisks); |
c2c087e6 DW |
7041 | return 0; |
7042 | } | |
7043 | return 1; | |
7044 | } | |
0dcecb2e | 7045 | |
c2c087e6 | 7046 | /* This device must be a member of the set */ |
9e04ac1c | 7047 | if (!stat_is_blkdev(dev, &rdev)) |
c2c087e6 DW |
7048 | return 0; |
7049 | for (dl = super->disks ; dl ; dl = dl->next) { | |
9e04ac1c ZL |
7050 | if (dl->major == (int)major(rdev) && |
7051 | dl->minor == (int)minor(rdev)) | |
c2c087e6 DW |
7052 | break; |
7053 | } | |
7054 | if (!dl) { | |
2c514b71 | 7055 | if (verbose) |
7a862a02 | 7056 | pr_err("%s is not in the same imsm set\n", dev); |
c2c087e6 | 7057 | return 0; |
a20d2ba5 DW |
7058 | } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) { |
7059 | /* If a volume is present then the current creation attempt | |
7060 | * cannot incorporate new spares because the orom may not | |
7061 | * understand this configuration (all member disks must be | |
7062 | * members of each array in the container). | |
7063 | */ | |
7a862a02 N |
7064 | pr_err("%s is a spare and a volume is already defined for this container\n", dev); |
7065 | pr_err("The option-rom requires all member disks to be a member of all volumes\n"); | |
a20d2ba5 | 7066 | return 0; |
5fe62b94 WD |
7067 | } else if (super->orom && mpb->num_raid_devs > 0 && |
7068 | mpb->num_disks != raiddisks) { | |
7a862a02 | 7069 | pr_err("The option-rom requires all member disks to be a member of all volumes\n"); |
5fe62b94 | 7070 | return 0; |
c2c087e6 | 7071 | } |
0dcecb2e DW |
7072 | |
7073 | /* retrieve the largest free space block */ | |
c2c087e6 DW |
7074 | e = get_extents(super, dl); |
7075 | maxsize = 0; | |
7076 | i = 0; | |
0dcecb2e DW |
7077 | if (e) { |
7078 | do { | |
7079 | unsigned long long esize; | |
7080 | ||
7081 | esize = e[i].start - pos; | |
7082 | if (esize >= maxsize) | |
7083 | maxsize = esize; | |
7084 | pos = e[i].start + e[i].size; | |
7085 | i++; | |
7086 | } while (e[i-1].size); | |
7087 | dl->e = e; | |
7088 | dl->extent_cnt = i; | |
7089 | } else { | |
7090 | if (verbose) | |
e7b84f9d | 7091 | pr_err("unable to determine free space for: %s\n", |
0dcecb2e DW |
7092 | dev); |
7093 | return 0; | |
7094 | } | |
7095 | if (maxsize < size) { | |
7096 | if (verbose) | |
e7b84f9d | 7097 | pr_err("%s not enough space (%llu < %llu)\n", |
0dcecb2e DW |
7098 | dev, maxsize, size); |
7099 | return 0; | |
7100 | } | |
7101 | ||
7102 | /* count total number of extents for merge */ | |
7103 | i = 0; | |
7104 | for (dl = super->disks; dl; dl = dl->next) | |
7105 | if (dl->e) | |
7106 | i += dl->extent_cnt; | |
7107 | ||
7108 | maxsize = merge_extents(super, i); | |
3baa56ab LO |
7109 | |
7110 | if (!check_env("IMSM_NO_PLATFORM") && | |
7111 | mpb->num_raid_devs > 0 && size && size != maxsize) { | |
7a862a02 | 7112 | pr_err("attempting to create a second volume with size less then remaining space. Aborting...\n"); |
3baa56ab LO |
7113 | return 0; |
7114 | } | |
7115 | ||
a7dd165b | 7116 | if (maxsize < size || maxsize == 0) { |
b3071342 LD |
7117 | if (verbose) { |
7118 | if (maxsize == 0) | |
7a862a02 | 7119 | pr_err("no free space left on device. Aborting...\n"); |
b3071342 | 7120 | else |
7a862a02 | 7121 | pr_err("not enough space to create volume of given size (%llu < %llu). Aborting...\n", |
b3071342 LD |
7122 | maxsize, size); |
7123 | } | |
0dcecb2e | 7124 | return 0; |
0dcecb2e DW |
7125 | } |
7126 | ||
c2c087e6 DW |
7127 | *freesize = maxsize; |
7128 | ||
ca9de185 | 7129 | if (super->orom) { |
72a45777 | 7130 | int count = count_volumes(super->hba, |
ca9de185 LM |
7131 | super->orom->dpa, verbose); |
7132 | if (super->orom->vphba <= count) { | |
676e87a8 | 7133 | pr_vrb("platform does not support more than %d raid volumes.\n", |
ca9de185 LM |
7134 | super->orom->vphba); |
7135 | return 0; | |
7136 | } | |
7137 | } | |
c2c087e6 | 7138 | return 1; |
cdddbdbc DW |
7139 | } |
7140 | ||
13bcac90 | 7141 | static int imsm_get_free_size(struct supertype *st, int raiddisks, |
efb30e7f DW |
7142 | unsigned long long size, int chunk, |
7143 | unsigned long long *freesize) | |
7144 | { | |
7145 | struct intel_super *super = st->sb; | |
7146 | struct imsm_super *mpb = super->anchor; | |
7147 | struct dl *dl; | |
7148 | int i; | |
7149 | int extent_cnt; | |
7150 | struct extent *e; | |
7151 | unsigned long long maxsize; | |
7152 | unsigned long long minsize; | |
7153 | int cnt; | |
7154 | int used; | |
7155 | ||
7156 | /* find the largest common start free region of the possible disks */ | |
7157 | used = 0; | |
7158 | extent_cnt = 0; | |
7159 | cnt = 0; | |
7160 | for (dl = super->disks; dl; dl = dl->next) { | |
7161 | dl->raiddisk = -1; | |
7162 | ||
7163 | if (dl->index >= 0) | |
7164 | used++; | |
7165 | ||
7166 | /* don't activate new spares if we are orom constrained | |
7167 | * and there is already a volume active in the container | |
7168 | */ | |
7169 | if (super->orom && dl->index < 0 && mpb->num_raid_devs) | |
7170 | continue; | |
7171 | ||
7172 | e = get_extents(super, dl); | |
7173 | if (!e) | |
7174 | continue; | |
7175 | for (i = 1; e[i-1].size; i++) | |
7176 | ; | |
7177 | dl->e = e; | |
7178 | dl->extent_cnt = i; | |
7179 | extent_cnt += i; | |
7180 | cnt++; | |
7181 | } | |
7182 | ||
7183 | maxsize = merge_extents(super, extent_cnt); | |
7184 | minsize = size; | |
7185 | if (size == 0) | |
612e59d8 CA |
7186 | /* chunk is in K */ |
7187 | minsize = chunk * 2; | |
efb30e7f DW |
7188 | |
7189 | if (cnt < raiddisks || | |
7190 | (super->orom && used && used != raiddisks) || | |
a7dd165b DW |
7191 | maxsize < minsize || |
7192 | maxsize == 0) { | |
e7b84f9d | 7193 | pr_err("not enough devices with space to create array.\n"); |
efb30e7f DW |
7194 | return 0; /* No enough free spaces large enough */ |
7195 | } | |
7196 | ||
7197 | if (size == 0) { | |
7198 | size = maxsize; | |
7199 | if (chunk) { | |
612e59d8 CA |
7200 | size /= 2 * chunk; |
7201 | size *= 2 * chunk; | |
efb30e7f | 7202 | } |
f878b242 LM |
7203 | maxsize = size; |
7204 | } | |
7205 | if (!check_env("IMSM_NO_PLATFORM") && | |
7206 | mpb->num_raid_devs > 0 && size && size != maxsize) { | |
7a862a02 | 7207 | pr_err("attempting to create a second volume with size less then remaining space. Aborting...\n"); |
f878b242 | 7208 | return 0; |
efb30e7f | 7209 | } |
efb30e7f DW |
7210 | cnt = 0; |
7211 | for (dl = super->disks; dl; dl = dl->next) | |
7212 | if (dl->e) | |
7213 | dl->raiddisk = cnt++; | |
7214 | ||
7215 | *freesize = size; | |
7216 | ||
13bcac90 AK |
7217 | dprintf("imsm: imsm_get_free_size() returns : %llu\n", size); |
7218 | ||
efb30e7f DW |
7219 | return 1; |
7220 | } | |
7221 | ||
13bcac90 AK |
7222 | static int reserve_space(struct supertype *st, int raiddisks, |
7223 | unsigned long long size, int chunk, | |
7224 | unsigned long long *freesize) | |
7225 | { | |
7226 | struct intel_super *super = st->sb; | |
7227 | struct dl *dl; | |
7228 | int cnt; | |
7229 | int rv = 0; | |
7230 | ||
7231 | rv = imsm_get_free_size(st, raiddisks, size, chunk, freesize); | |
7232 | if (rv) { | |
7233 | cnt = 0; | |
7234 | for (dl = super->disks; dl; dl = dl->next) | |
7235 | if (dl->e) | |
7236 | dl->raiddisk = cnt++; | |
7237 | rv = 1; | |
7238 | } | |
7239 | ||
7240 | return rv; | |
7241 | } | |
7242 | ||
bf5a934a | 7243 | static int validate_geometry_imsm(struct supertype *st, int level, int layout, |
c21e737b | 7244 | int raiddisks, int *chunk, unsigned long long size, |
af4348dd | 7245 | unsigned long long data_offset, |
bf5a934a | 7246 | char *dev, unsigned long long *freesize, |
5308f117 | 7247 | int consistency_policy, int verbose) |
bf5a934a DW |
7248 | { |
7249 | int fd, cfd; | |
7250 | struct mdinfo *sra; | |
20cbe8d2 | 7251 | int is_member = 0; |
bf5a934a | 7252 | |
d54559f0 LM |
7253 | /* load capability |
7254 | * if given unused devices create a container | |
bf5a934a DW |
7255 | * if given given devices in a container create a member volume |
7256 | */ | |
7257 | if (level == LEVEL_CONTAINER) { | |
7258 | /* Must be a fresh device to add to a container */ | |
7259 | return validate_geometry_imsm_container(st, level, layout, | |
c21e737b | 7260 | raiddisks, |
7ccc4cc4 | 7261 | *chunk, |
af4348dd | 7262 | size, data_offset, |
bf5a934a DW |
7263 | dev, freesize, |
7264 | verbose); | |
7265 | } | |
9587c373 | 7266 | |
8592f29d | 7267 | if (!dev) { |
e91a3bad | 7268 | if (st->sb) { |
ca9de185 | 7269 | struct intel_super *super = st->sb; |
e91a3bad | 7270 | if (!validate_geometry_imsm_orom(st->sb, level, layout, |
2cc699af | 7271 | raiddisks, chunk, size, |
e91a3bad LM |
7272 | verbose)) |
7273 | return 0; | |
efb30e7f DW |
7274 | /* we are being asked to automatically layout a |
7275 | * new volume based on the current contents of | |
7276 | * the container. If the the parameters can be | |
7277 | * satisfied reserve_space will record the disks, | |
7278 | * start offset, and size of the volume to be | |
7279 | * created. add_to_super and getinfo_super | |
7280 | * detect when autolayout is in progress. | |
7281 | */ | |
ca9de185 LM |
7282 | /* assuming that freesize is always given when array is |
7283 | created */ | |
7284 | if (super->orom && freesize) { | |
7285 | int count; | |
72a45777 | 7286 | count = count_volumes(super->hba, |
ca9de185 LM |
7287 | super->orom->dpa, verbose); |
7288 | if (super->orom->vphba <= count) { | |
676e87a8 | 7289 | pr_vrb("platform does not support more than %d raid volumes.\n", |
ca9de185 LM |
7290 | super->orom->vphba); |
7291 | return 0; | |
7292 | } | |
7293 | } | |
e91a3bad LM |
7294 | if (freesize) |
7295 | return reserve_space(st, raiddisks, size, | |
7ccc4cc4 | 7296 | *chunk, freesize); |
8592f29d N |
7297 | } |
7298 | return 1; | |
7299 | } | |
bf5a934a DW |
7300 | if (st->sb) { |
7301 | /* creating in a given container */ | |
7302 | return validate_geometry_imsm_volume(st, level, layout, | |
7303 | raiddisks, chunk, size, | |
af4348dd | 7304 | data_offset, |
bf5a934a DW |
7305 | dev, freesize, verbose); |
7306 | } | |
7307 | ||
bf5a934a DW |
7308 | /* This device needs to be a device in an 'imsm' container */ |
7309 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
7310 | if (fd >= 0) { | |
7311 | if (verbose) | |
e7b84f9d N |
7312 | pr_err("Cannot create this array on device %s\n", |
7313 | dev); | |
bf5a934a DW |
7314 | close(fd); |
7315 | return 0; | |
7316 | } | |
7317 | if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) { | |
7318 | if (verbose) | |
e7b84f9d | 7319 | pr_err("Cannot open %s: %s\n", |
bf5a934a DW |
7320 | dev, strerror(errno)); |
7321 | return 0; | |
7322 | } | |
7323 | /* Well, it is in use by someone, maybe an 'imsm' container. */ | |
7324 | cfd = open_container(fd); | |
20cbe8d2 | 7325 | close(fd); |
bf5a934a | 7326 | if (cfd < 0) { |
bf5a934a | 7327 | if (verbose) |
e7b84f9d | 7328 | pr_err("Cannot use %s: It is busy\n", |
bf5a934a DW |
7329 | dev); |
7330 | return 0; | |
7331 | } | |
4dd2df09 | 7332 | sra = sysfs_read(cfd, NULL, GET_VERSION); |
bf5a934a | 7333 | if (sra && sra->array.major_version == -1 && |
20cbe8d2 AW |
7334 | strcmp(sra->text_version, "imsm") == 0) |
7335 | is_member = 1; | |
7336 | sysfs_free(sra); | |
7337 | if (is_member) { | |
bf5a934a DW |
7338 | /* This is a member of a imsm container. Load the container |
7339 | * and try to create a volume | |
7340 | */ | |
7341 | struct intel_super *super; | |
7342 | ||
ec50f7b6 | 7343 | if (load_super_imsm_all(st, cfd, (void **) &super, NULL, NULL, 1) == 0) { |
bf5a934a | 7344 | st->sb = super; |
4dd2df09 | 7345 | strcpy(st->container_devnm, fd2devnm(cfd)); |
bf5a934a DW |
7346 | close(cfd); |
7347 | return validate_geometry_imsm_volume(st, level, layout, | |
7348 | raiddisks, chunk, | |
af4348dd | 7349 | size, data_offset, dev, |
ecbd9e81 N |
7350 | freesize, 1) |
7351 | ? 1 : -1; | |
bf5a934a | 7352 | } |
20cbe8d2 | 7353 | } |
bf5a934a | 7354 | |
20cbe8d2 | 7355 | if (verbose) |
e7b84f9d | 7356 | pr_err("failed container membership check\n"); |
20cbe8d2 AW |
7357 | |
7358 | close(cfd); | |
7359 | return 0; | |
bf5a934a | 7360 | } |
0bd16cf2 | 7361 | |
30f58b22 | 7362 | static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk) |
0bd16cf2 DJ |
7363 | { |
7364 | struct intel_super *super = st->sb; | |
7365 | ||
30f58b22 DW |
7366 | if (level && *level == UnSet) |
7367 | *level = LEVEL_CONTAINER; | |
7368 | ||
7369 | if (level && layout && *layout == UnSet) | |
7370 | *layout = imsm_level_to_layout(*level); | |
0bd16cf2 | 7371 | |
cd9d1ac7 DW |
7372 | if (chunk && (*chunk == UnSet || *chunk == 0)) |
7373 | *chunk = imsm_default_chunk(super->orom); | |
0bd16cf2 DJ |
7374 | } |
7375 | ||
33414a01 DW |
7376 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev); |
7377 | ||
7378 | static int kill_subarray_imsm(struct supertype *st) | |
7379 | { | |
7380 | /* remove the subarray currently referenced by ->current_vol */ | |
7381 | __u8 i; | |
7382 | struct intel_dev **dp; | |
7383 | struct intel_super *super = st->sb; | |
7384 | __u8 current_vol = super->current_vol; | |
7385 | struct imsm_super *mpb = super->anchor; | |
7386 | ||
7387 | if (super->current_vol < 0) | |
7388 | return 2; | |
7389 | super->current_vol = -1; /* invalidate subarray cursor */ | |
7390 | ||
7391 | /* block deletions that would change the uuid of active subarrays | |
7392 | * | |
7393 | * FIXME when immutable ids are available, but note that we'll | |
7394 | * also need to fixup the invalidated/active subarray indexes in | |
7395 | * mdstat | |
7396 | */ | |
7397 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
7398 | char subarray[4]; | |
7399 | ||
7400 | if (i < current_vol) | |
7401 | continue; | |
7402 | sprintf(subarray, "%u", i); | |
4dd2df09 | 7403 | if (is_subarray_active(subarray, st->devnm)) { |
e7b84f9d N |
7404 | pr_err("deleting subarray-%d would change the UUID of active subarray-%d, aborting\n", |
7405 | current_vol, i); | |
33414a01 DW |
7406 | |
7407 | return 2; | |
7408 | } | |
7409 | } | |
7410 | ||
7411 | if (st->update_tail) { | |
503975b9 | 7412 | struct imsm_update_kill_array *u = xmalloc(sizeof(*u)); |
33414a01 | 7413 | |
33414a01 DW |
7414 | u->type = update_kill_array; |
7415 | u->dev_idx = current_vol; | |
7416 | append_metadata_update(st, u, sizeof(*u)); | |
7417 | ||
7418 | return 0; | |
7419 | } | |
7420 | ||
7421 | for (dp = &super->devlist; *dp;) | |
7422 | if ((*dp)->index == current_vol) { | |
7423 | *dp = (*dp)->next; | |
7424 | } else { | |
7425 | handle_missing(super, (*dp)->dev); | |
7426 | if ((*dp)->index > current_vol) | |
7427 | (*dp)->index--; | |
7428 | dp = &(*dp)->next; | |
7429 | } | |
7430 | ||
7431 | /* no more raid devices, all active components are now spares, | |
7432 | * but of course failed are still failed | |
7433 | */ | |
7434 | if (--mpb->num_raid_devs == 0) { | |
7435 | struct dl *d; | |
7436 | ||
7437 | for (d = super->disks; d; d = d->next) | |
a8619d23 AK |
7438 | if (d->index > -2) |
7439 | mark_spare(d); | |
33414a01 DW |
7440 | } |
7441 | ||
7442 | super->updates_pending++; | |
7443 | ||
7444 | return 0; | |
7445 | } | |
aa534678 | 7446 | |
a951a4f7 | 7447 | static int update_subarray_imsm(struct supertype *st, char *subarray, |
fa56eddb | 7448 | char *update, struct mddev_ident *ident) |
aa534678 DW |
7449 | { |
7450 | /* update the subarray currently referenced by ->current_vol */ | |
7451 | struct intel_super *super = st->sb; | |
7452 | struct imsm_super *mpb = super->anchor; | |
7453 | ||
aa534678 DW |
7454 | if (strcmp(update, "name") == 0) { |
7455 | char *name = ident->name; | |
a951a4f7 N |
7456 | char *ep; |
7457 | int vol; | |
aa534678 | 7458 | |
4dd2df09 | 7459 | if (is_subarray_active(subarray, st->devnm)) { |
e7b84f9d | 7460 | pr_err("Unable to update name of active subarray\n"); |
aa534678 DW |
7461 | return 2; |
7462 | } | |
7463 | ||
7464 | if (!check_name(super, name, 0)) | |
7465 | return 2; | |
7466 | ||
a951a4f7 N |
7467 | vol = strtoul(subarray, &ep, 10); |
7468 | if (*ep != '\0' || vol >= super->anchor->num_raid_devs) | |
7469 | return 2; | |
7470 | ||
aa534678 | 7471 | if (st->update_tail) { |
503975b9 | 7472 | struct imsm_update_rename_array *u = xmalloc(sizeof(*u)); |
aa534678 | 7473 | |
aa534678 | 7474 | u->type = update_rename_array; |
a951a4f7 | 7475 | u->dev_idx = vol; |
618f4e6d XN |
7476 | strncpy((char *) u->name, name, MAX_RAID_SERIAL_LEN); |
7477 | u->name[MAX_RAID_SERIAL_LEN-1] = '\0'; | |
aa534678 DW |
7478 | append_metadata_update(st, u, sizeof(*u)); |
7479 | } else { | |
7480 | struct imsm_dev *dev; | |
7481 | int i; | |
7482 | ||
a951a4f7 | 7483 | dev = get_imsm_dev(super, vol); |
618f4e6d XN |
7484 | strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN); |
7485 | dev->volume[MAX_RAID_SERIAL_LEN-1] = '\0'; | |
aa534678 DW |
7486 | for (i = 0; i < mpb->num_raid_devs; i++) { |
7487 | dev = get_imsm_dev(super, i); | |
7488 | handle_missing(super, dev); | |
7489 | } | |
7490 | super->updates_pending++; | |
7491 | } | |
e6e9dd3f AP |
7492 | } else if (strcmp(update, "ppl") == 0 || |
7493 | strcmp(update, "no-ppl") == 0) { | |
7494 | int new_policy; | |
7495 | char *ep; | |
7496 | int vol = strtoul(subarray, &ep, 10); | |
7497 | ||
7498 | if (*ep != '\0' || vol >= super->anchor->num_raid_devs) | |
7499 | return 2; | |
7500 | ||
7501 | if (strcmp(update, "ppl") == 0) | |
c2462068 | 7502 | new_policy = RWH_MULTIPLE_DISTRIBUTED; |
e6e9dd3f | 7503 | else |
c2462068 | 7504 | new_policy = RWH_MULTIPLE_OFF; |
e6e9dd3f AP |
7505 | |
7506 | if (st->update_tail) { | |
7507 | struct imsm_update_rwh_policy *u = xmalloc(sizeof(*u)); | |
7508 | ||
7509 | u->type = update_rwh_policy; | |
7510 | u->dev_idx = vol; | |
7511 | u->new_policy = new_policy; | |
7512 | append_metadata_update(st, u, sizeof(*u)); | |
7513 | } else { | |
7514 | struct imsm_dev *dev; | |
7515 | ||
7516 | dev = get_imsm_dev(super, vol); | |
7517 | dev->rwh_policy = new_policy; | |
7518 | super->updates_pending++; | |
7519 | } | |
aa534678 DW |
7520 | } else |
7521 | return 2; | |
7522 | ||
7523 | return 0; | |
7524 | } | |
bf5a934a | 7525 | |
28bce06f AK |
7526 | static int is_gen_migration(struct imsm_dev *dev) |
7527 | { | |
7534230b AK |
7528 | if (dev == NULL) |
7529 | return 0; | |
7530 | ||
28bce06f AK |
7531 | if (!dev->vol.migr_state) |
7532 | return 0; | |
7533 | ||
7534 | if (migr_type(dev) == MIGR_GEN_MIGR) | |
7535 | return 1; | |
7536 | ||
7537 | return 0; | |
7538 | } | |
7539 | ||
1e5c6983 DW |
7540 | static int is_rebuilding(struct imsm_dev *dev) |
7541 | { | |
7542 | struct imsm_map *migr_map; | |
7543 | ||
7544 | if (!dev->vol.migr_state) | |
7545 | return 0; | |
7546 | ||
7547 | if (migr_type(dev) != MIGR_REBUILD) | |
7548 | return 0; | |
7549 | ||
238c0a71 | 7550 | migr_map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
7551 | |
7552 | if (migr_map->map_state == IMSM_T_STATE_DEGRADED) | |
7553 | return 1; | |
7554 | else | |
7555 | return 0; | |
7556 | } | |
7557 | ||
6ce1fbf1 AK |
7558 | static int is_initializing(struct imsm_dev *dev) |
7559 | { | |
7560 | struct imsm_map *migr_map; | |
7561 | ||
7562 | if (!dev->vol.migr_state) | |
7563 | return 0; | |
7564 | ||
7565 | if (migr_type(dev) != MIGR_INIT) | |
7566 | return 0; | |
7567 | ||
238c0a71 | 7568 | migr_map = get_imsm_map(dev, MAP_1); |
6ce1fbf1 AK |
7569 | |
7570 | if (migr_map->map_state == IMSM_T_STATE_UNINITIALIZED) | |
7571 | return 1; | |
7572 | ||
7573 | return 0; | |
6ce1fbf1 AK |
7574 | } |
7575 | ||
c47b0ff6 AK |
7576 | static void update_recovery_start(struct intel_super *super, |
7577 | struct imsm_dev *dev, | |
7578 | struct mdinfo *array) | |
1e5c6983 DW |
7579 | { |
7580 | struct mdinfo *rebuild = NULL; | |
7581 | struct mdinfo *d; | |
7582 | __u32 units; | |
7583 | ||
7584 | if (!is_rebuilding(dev)) | |
7585 | return; | |
7586 | ||
7587 | /* Find the rebuild target, but punt on the dual rebuild case */ | |
7588 | for (d = array->devs; d; d = d->next) | |
7589 | if (d->recovery_start == 0) { | |
7590 | if (rebuild) | |
7591 | return; | |
7592 | rebuild = d; | |
7593 | } | |
7594 | ||
4363fd80 DW |
7595 | if (!rebuild) { |
7596 | /* (?) none of the disks are marked with | |
7597 | * IMSM_ORD_REBUILD, so assume they are missing and the | |
7598 | * disk_ord_tbl was not correctly updated | |
7599 | */ | |
1ade5cc1 | 7600 | dprintf("failed to locate out-of-sync disk\n"); |
4363fd80 DW |
7601 | return; |
7602 | } | |
7603 | ||
1e5c6983 | 7604 | units = __le32_to_cpu(dev->vol.curr_migr_unit); |
c47b0ff6 | 7605 | rebuild->recovery_start = units * blocks_per_migr_unit(super, dev); |
1e5c6983 DW |
7606 | } |
7607 | ||
276d77db | 7608 | static int recover_backup_imsm(struct supertype *st, struct mdinfo *info); |
1e5c6983 | 7609 | |
00bbdbda | 7610 | static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray) |
cdddbdbc | 7611 | { |
4f5bc454 DW |
7612 | /* Given a container loaded by load_super_imsm_all, |
7613 | * extract information about all the arrays into | |
7614 | * an mdinfo tree. | |
00bbdbda | 7615 | * If 'subarray' is given, just extract info about that array. |
4f5bc454 DW |
7616 | * |
7617 | * For each imsm_dev create an mdinfo, fill it in, | |
7618 | * then look for matching devices in super->disks | |
7619 | * and create appropriate device mdinfo. | |
7620 | */ | |
7621 | struct intel_super *super = st->sb; | |
949c47a0 | 7622 | struct imsm_super *mpb = super->anchor; |
4f5bc454 | 7623 | struct mdinfo *rest = NULL; |
00bbdbda | 7624 | unsigned int i; |
81219e70 | 7625 | int sb_errors = 0; |
abef11a3 AK |
7626 | struct dl *d; |
7627 | int spare_disks = 0; | |
cdddbdbc | 7628 | |
19482bcc AK |
7629 | /* do not assemble arrays when not all attributes are supported */ |
7630 | if (imsm_check_attributes(mpb->attributes) == 0) { | |
81219e70 | 7631 | sb_errors = 1; |
7a862a02 | 7632 | pr_err("Unsupported attributes in IMSM metadata.Arrays activation is blocked.\n"); |
19482bcc AK |
7633 | } |
7634 | ||
abef11a3 AK |
7635 | /* count spare devices, not used in maps |
7636 | */ | |
7637 | for (d = super->disks; d; d = d->next) | |
7638 | if (d->index == -1) | |
7639 | spare_disks++; | |
7640 | ||
4f5bc454 | 7641 | for (i = 0; i < mpb->num_raid_devs; i++) { |
00bbdbda N |
7642 | struct imsm_dev *dev; |
7643 | struct imsm_map *map; | |
86e3692b | 7644 | struct imsm_map *map2; |
4f5bc454 | 7645 | struct mdinfo *this; |
a6482415 | 7646 | int slot; |
a6482415 | 7647 | int chunk; |
00bbdbda N |
7648 | char *ep; |
7649 | ||
7650 | if (subarray && | |
7651 | (i != strtoul(subarray, &ep, 10) || *ep != '\0')) | |
7652 | continue; | |
7653 | ||
7654 | dev = get_imsm_dev(super, i); | |
238c0a71 AK |
7655 | map = get_imsm_map(dev, MAP_0); |
7656 | map2 = get_imsm_map(dev, MAP_1); | |
4f5bc454 | 7657 | |
1ce0101c DW |
7658 | /* do not publish arrays that are in the middle of an |
7659 | * unsupported migration | |
7660 | */ | |
7661 | if (dev->vol.migr_state && | |
28bce06f | 7662 | (migr_type(dev) == MIGR_STATE_CHANGE)) { |
7a862a02 | 7663 | pr_err("cannot assemble volume '%.16s': unsupported migration in progress\n", |
1ce0101c DW |
7664 | dev->volume); |
7665 | continue; | |
7666 | } | |
2db86302 LM |
7667 | /* do not publish arrays that are not support by controller's |
7668 | * OROM/EFI | |
7669 | */ | |
1ce0101c | 7670 | |
503975b9 | 7671 | this = xmalloc(sizeof(*this)); |
4f5bc454 | 7672 | |
301406c9 | 7673 | super->current_vol = i; |
a5d85af7 | 7674 | getinfo_super_imsm_volume(st, this, NULL); |
9894ec0d | 7675 | this->next = rest; |
a6482415 | 7676 | chunk = __le16_to_cpu(map->blocks_per_strip) >> 1; |
81219e70 LM |
7677 | /* mdadm does not support all metadata features- set the bit in all arrays state */ |
7678 | if (!validate_geometry_imsm_orom(super, | |
7679 | get_imsm_raid_level(map), /* RAID level */ | |
7680 | imsm_level_to_layout(get_imsm_raid_level(map)), | |
7681 | map->num_members, /* raid disks */ | |
2cc699af | 7682 | &chunk, join_u32(dev->size_low, dev->size_high), |
81219e70 | 7683 | 1 /* verbose */)) { |
7a862a02 | 7684 | pr_err("IMSM RAID geometry validation failed. Array %s activation is blocked.\n", |
81219e70 LM |
7685 | dev->volume); |
7686 | this->array.state |= | |
7687 | (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) | | |
7688 | (1<<MD_SB_BLOCK_VOLUME); | |
7689 | } | |
81219e70 LM |
7690 | |
7691 | /* if array has bad blocks, set suitable bit in all arrays state */ | |
7692 | if (sb_errors) | |
7693 | this->array.state |= | |
7694 | (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) | | |
7695 | (1<<MD_SB_BLOCK_VOLUME); | |
7696 | ||
4f5bc454 | 7697 | for (slot = 0 ; slot < map->num_members; slot++) { |
1e5c6983 | 7698 | unsigned long long recovery_start; |
4f5bc454 DW |
7699 | struct mdinfo *info_d; |
7700 | struct dl *d; | |
7701 | int idx; | |
9a1608e5 | 7702 | int skip; |
7eef0453 | 7703 | __u32 ord; |
4f5bc454 | 7704 | |
9a1608e5 | 7705 | skip = 0; |
238c0a71 AK |
7706 | idx = get_imsm_disk_idx(dev, slot, MAP_0); |
7707 | ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); | |
4f5bc454 DW |
7708 | for (d = super->disks; d ; d = d->next) |
7709 | if (d->index == idx) | |
0fbd635c | 7710 | break; |
4f5bc454 | 7711 | |
1e5c6983 | 7712 | recovery_start = MaxSector; |
4f5bc454 | 7713 | if (d == NULL) |
9a1608e5 | 7714 | skip = 1; |
25ed7e59 | 7715 | if (d && is_failed(&d->disk)) |
9a1608e5 | 7716 | skip = 1; |
7eef0453 | 7717 | if (ord & IMSM_ORD_REBUILD) |
1e5c6983 | 7718 | recovery_start = 0; |
9a1608e5 | 7719 | |
1011e834 | 7720 | /* |
9a1608e5 | 7721 | * if we skip some disks the array will be assmebled degraded; |
1e5c6983 DW |
7722 | * reset resync start to avoid a dirty-degraded |
7723 | * situation when performing the intial sync | |
9a1608e5 DW |
7724 | * |
7725 | * FIXME handle dirty degraded | |
7726 | */ | |
2432ce9b AP |
7727 | if ((skip || recovery_start == 0) && |
7728 | !(dev->vol.dirty & RAIDVOL_DIRTY)) | |
b7528a20 | 7729 | this->resync_start = MaxSector; |
9a1608e5 DW |
7730 | if (skip) |
7731 | continue; | |
4f5bc454 | 7732 | |
503975b9 | 7733 | info_d = xcalloc(1, sizeof(*info_d)); |
4f5bc454 DW |
7734 | info_d->next = this->devs; |
7735 | this->devs = info_d; | |
7736 | ||
4f5bc454 DW |
7737 | info_d->disk.number = d->index; |
7738 | info_d->disk.major = d->major; | |
7739 | info_d->disk.minor = d->minor; | |
7740 | info_d->disk.raid_disk = slot; | |
1e5c6983 | 7741 | info_d->recovery_start = recovery_start; |
86e3692b AK |
7742 | if (map2) { |
7743 | if (slot < map2->num_members) | |
7744 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
7745 | else |
7746 | this->array.spare_disks++; | |
86e3692b AK |
7747 | } else { |
7748 | if (slot < map->num_members) | |
7749 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
7750 | else |
7751 | this->array.spare_disks++; | |
86e3692b | 7752 | } |
1e5c6983 DW |
7753 | if (info_d->recovery_start == MaxSector) |
7754 | this->array.working_disks++; | |
4f5bc454 DW |
7755 | |
7756 | info_d->events = __le32_to_cpu(mpb->generation_num); | |
5551b113 | 7757 | info_d->data_offset = pba_of_lba0(map); |
06fb291a PB |
7758 | |
7759 | if (map->raid_level == 5) { | |
7760 | info_d->component_size = | |
7761 | num_data_stripes(map) * | |
7762 | map->blocks_per_strip; | |
2432ce9b AP |
7763 | info_d->ppl_sector = this->ppl_sector; |
7764 | info_d->ppl_size = this->ppl_size; | |
98e96bdb AP |
7765 | if (this->consistency_policy == CONSISTENCY_POLICY_PPL && |
7766 | recovery_start == 0) | |
7767 | this->resync_start = 0; | |
06fb291a PB |
7768 | } else { |
7769 | info_d->component_size = blocks_per_member(map); | |
7770 | } | |
b12796be | 7771 | |
5e46202e | 7772 | info_d->bb.supported = 1; |
b12796be TM |
7773 | get_volume_badblocks(super->bbm_log, ord_to_idx(ord), |
7774 | info_d->data_offset, | |
7775 | info_d->component_size, | |
7776 | &info_d->bb); | |
4f5bc454 | 7777 | } |
1e5c6983 | 7778 | /* now that the disk list is up-to-date fixup recovery_start */ |
c47b0ff6 | 7779 | update_recovery_start(super, dev, this); |
abef11a3 | 7780 | this->array.spare_disks += spare_disks; |
276d77db AK |
7781 | |
7782 | /* check for reshape */ | |
7783 | if (this->reshape_active == 1) | |
7784 | recover_backup_imsm(st, this); | |
9a1608e5 | 7785 | rest = this; |
4f5bc454 DW |
7786 | } |
7787 | ||
7788 | return rest; | |
cdddbdbc DW |
7789 | } |
7790 | ||
3b451610 AK |
7791 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, |
7792 | int failed, int look_in_map) | |
c2a1e7da | 7793 | { |
3b451610 AK |
7794 | struct imsm_map *map; |
7795 | ||
7796 | map = get_imsm_map(dev, look_in_map); | |
c2a1e7da DW |
7797 | |
7798 | if (!failed) | |
1011e834 | 7799 | return map->map_state == IMSM_T_STATE_UNINITIALIZED ? |
3393c6af | 7800 | IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL; |
c2a1e7da DW |
7801 | |
7802 | switch (get_imsm_raid_level(map)) { | |
7803 | case 0: | |
7804 | return IMSM_T_STATE_FAILED; | |
7805 | break; | |
7806 | case 1: | |
7807 | if (failed < map->num_members) | |
7808 | return IMSM_T_STATE_DEGRADED; | |
7809 | else | |
7810 | return IMSM_T_STATE_FAILED; | |
7811 | break; | |
7812 | case 10: | |
7813 | { | |
7814 | /** | |
c92a2527 DW |
7815 | * check to see if any mirrors have failed, otherwise we |
7816 | * are degraded. Even numbered slots are mirrored on | |
7817 | * slot+1 | |
c2a1e7da | 7818 | */ |
c2a1e7da | 7819 | int i; |
d9b420a5 N |
7820 | /* gcc -Os complains that this is unused */ |
7821 | int insync = insync; | |
c2a1e7da DW |
7822 | |
7823 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 7824 | __u32 ord = get_imsm_ord_tbl_ent(dev, i, MAP_X); |
c92a2527 DW |
7825 | int idx = ord_to_idx(ord); |
7826 | struct imsm_disk *disk; | |
c2a1e7da | 7827 | |
c92a2527 | 7828 | /* reset the potential in-sync count on even-numbered |
1011e834 | 7829 | * slots. num_copies is always 2 for imsm raid10 |
c92a2527 DW |
7830 | */ |
7831 | if ((i & 1) == 0) | |
7832 | insync = 2; | |
c2a1e7da | 7833 | |
c92a2527 | 7834 | disk = get_imsm_disk(super, idx); |
25ed7e59 | 7835 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) |
c92a2527 | 7836 | insync--; |
c2a1e7da | 7837 | |
c92a2527 DW |
7838 | /* no in-sync disks left in this mirror the |
7839 | * array has failed | |
7840 | */ | |
7841 | if (insync == 0) | |
7842 | return IMSM_T_STATE_FAILED; | |
c2a1e7da DW |
7843 | } |
7844 | ||
7845 | return IMSM_T_STATE_DEGRADED; | |
7846 | } | |
7847 | case 5: | |
7848 | if (failed < 2) | |
7849 | return IMSM_T_STATE_DEGRADED; | |
7850 | else | |
7851 | return IMSM_T_STATE_FAILED; | |
7852 | break; | |
7853 | default: | |
7854 | break; | |
7855 | } | |
7856 | ||
7857 | return map->map_state; | |
7858 | } | |
7859 | ||
3b451610 AK |
7860 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev, |
7861 | int look_in_map) | |
c2a1e7da DW |
7862 | { |
7863 | int i; | |
7864 | int failed = 0; | |
7865 | struct imsm_disk *disk; | |
d5985138 AK |
7866 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
7867 | struct imsm_map *prev = get_imsm_map(dev, MAP_1); | |
68fe4598 | 7868 | struct imsm_map *map_for_loop; |
0556e1a2 DW |
7869 | __u32 ord; |
7870 | int idx; | |
d5985138 | 7871 | int idx_1; |
c2a1e7da | 7872 | |
0556e1a2 DW |
7873 | /* at the beginning of migration we set IMSM_ORD_REBUILD on |
7874 | * disks that are being rebuilt. New failures are recorded to | |
7875 | * map[0]. So we look through all the disks we started with and | |
7876 | * see if any failures are still present, or if any new ones | |
7877 | * have arrived | |
0556e1a2 | 7878 | */ |
d5985138 AK |
7879 | map_for_loop = map; |
7880 | if (prev && (map->num_members < prev->num_members)) | |
7881 | map_for_loop = prev; | |
68fe4598 LD |
7882 | |
7883 | for (i = 0; i < map_for_loop->num_members; i++) { | |
d5985138 | 7884 | idx_1 = -255; |
238c0a71 AK |
7885 | /* when MAP_X is passed both maps failures are counted |
7886 | */ | |
d5985138 | 7887 | if (prev && |
089f9d79 JS |
7888 | (look_in_map == MAP_1 || look_in_map == MAP_X) && |
7889 | i < prev->num_members) { | |
d5985138 AK |
7890 | ord = __le32_to_cpu(prev->disk_ord_tbl[i]); |
7891 | idx_1 = ord_to_idx(ord); | |
c2a1e7da | 7892 | |
d5985138 AK |
7893 | disk = get_imsm_disk(super, idx_1); |
7894 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) | |
7895 | failed++; | |
7896 | } | |
089f9d79 JS |
7897 | if ((look_in_map == MAP_0 || look_in_map == MAP_X) && |
7898 | i < map->num_members) { | |
d5985138 AK |
7899 | ord = __le32_to_cpu(map->disk_ord_tbl[i]); |
7900 | idx = ord_to_idx(ord); | |
7901 | ||
7902 | if (idx != idx_1) { | |
7903 | disk = get_imsm_disk(super, idx); | |
7904 | if (!disk || is_failed(disk) || | |
7905 | ord & IMSM_ORD_REBUILD) | |
7906 | failed++; | |
7907 | } | |
7908 | } | |
c2a1e7da DW |
7909 | } |
7910 | ||
7911 | return failed; | |
845dea95 NB |
7912 | } |
7913 | ||
97b4d0e9 DW |
7914 | static int imsm_open_new(struct supertype *c, struct active_array *a, |
7915 | char *inst) | |
7916 | { | |
7917 | struct intel_super *super = c->sb; | |
7918 | struct imsm_super *mpb = super->anchor; | |
bbab0940 | 7919 | struct imsm_update_prealloc_bb_mem u; |
9587c373 | 7920 | |
97b4d0e9 | 7921 | if (atoi(inst) >= mpb->num_raid_devs) { |
1ade5cc1 | 7922 | pr_err("subarry index %d, out of range\n", atoi(inst)); |
97b4d0e9 DW |
7923 | return -ENODEV; |
7924 | } | |
7925 | ||
7926 | dprintf("imsm: open_new %s\n", inst); | |
7927 | a->info.container_member = atoi(inst); | |
bbab0940 TM |
7928 | |
7929 | u.type = update_prealloc_badblocks_mem; | |
7930 | imsm_update_metadata_locally(c, &u, sizeof(u)); | |
7931 | ||
97b4d0e9 DW |
7932 | return 0; |
7933 | } | |
7934 | ||
0c046afd DW |
7935 | static int is_resyncing(struct imsm_dev *dev) |
7936 | { | |
7937 | struct imsm_map *migr_map; | |
7938 | ||
7939 | if (!dev->vol.migr_state) | |
7940 | return 0; | |
7941 | ||
1484e727 DW |
7942 | if (migr_type(dev) == MIGR_INIT || |
7943 | migr_type(dev) == MIGR_REPAIR) | |
0c046afd DW |
7944 | return 1; |
7945 | ||
4c9bc37b AK |
7946 | if (migr_type(dev) == MIGR_GEN_MIGR) |
7947 | return 0; | |
7948 | ||
238c0a71 | 7949 | migr_map = get_imsm_map(dev, MAP_1); |
0c046afd | 7950 | |
089f9d79 JS |
7951 | if (migr_map->map_state == IMSM_T_STATE_NORMAL && |
7952 | dev->vol.migr_type != MIGR_GEN_MIGR) | |
0c046afd DW |
7953 | return 1; |
7954 | else | |
7955 | return 0; | |
7956 | } | |
7957 | ||
0556e1a2 | 7958 | /* return true if we recorded new information */ |
4c9e8c1e TM |
7959 | static int mark_failure(struct intel_super *super, |
7960 | struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
47ee5a45 | 7961 | { |
0556e1a2 DW |
7962 | __u32 ord; |
7963 | int slot; | |
7964 | struct imsm_map *map; | |
86c54047 DW |
7965 | char buf[MAX_RAID_SERIAL_LEN+3]; |
7966 | unsigned int len, shift = 0; | |
0556e1a2 DW |
7967 | |
7968 | /* new failures are always set in map[0] */ | |
238c0a71 | 7969 | map = get_imsm_map(dev, MAP_0); |
0556e1a2 DW |
7970 | |
7971 | slot = get_imsm_disk_slot(map, idx); | |
7972 | if (slot < 0) | |
7973 | return 0; | |
7974 | ||
7975 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
25ed7e59 | 7976 | if (is_failed(disk) && (ord & IMSM_ORD_REBUILD)) |
0556e1a2 DW |
7977 | return 0; |
7978 | ||
7d0c5e24 LD |
7979 | memcpy(buf, disk->serial, MAX_RAID_SERIAL_LEN); |
7980 | buf[MAX_RAID_SERIAL_LEN] = '\000'; | |
7981 | strcat(buf, ":0"); | |
86c54047 DW |
7982 | if ((len = strlen(buf)) >= MAX_RAID_SERIAL_LEN) |
7983 | shift = len - MAX_RAID_SERIAL_LEN + 1; | |
7984 | strncpy((char *)disk->serial, &buf[shift], MAX_RAID_SERIAL_LEN); | |
7985 | ||
f2f27e63 | 7986 | disk->status |= FAILED_DISK; |
0556e1a2 | 7987 | set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD); |
17788994 AK |
7988 | /* mark failures in second map if second map exists and this disk |
7989 | * in this slot. | |
7990 | * This is valid for migration, initialization and rebuild | |
7991 | */ | |
7992 | if (dev->vol.migr_state) { | |
238c0a71 | 7993 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); |
0a108d63 AK |
7994 | int slot2 = get_imsm_disk_slot(map2, idx); |
7995 | ||
089f9d79 | 7996 | if (slot2 < map2->num_members && slot2 >= 0) |
0a108d63 | 7997 | set_imsm_ord_tbl_ent(map2, slot2, |
1ace8403 AK |
7998 | idx | IMSM_ORD_REBUILD); |
7999 | } | |
f21e18ca | 8000 | if (map->failed_disk_num == 0xff) |
0556e1a2 | 8001 | map->failed_disk_num = slot; |
4c9e8c1e TM |
8002 | |
8003 | clear_disk_badblocks(super->bbm_log, ord_to_idx(ord)); | |
8004 | ||
0556e1a2 DW |
8005 | return 1; |
8006 | } | |
8007 | ||
4c9e8c1e TM |
8008 | static void mark_missing(struct intel_super *super, |
8009 | struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
0556e1a2 | 8010 | { |
4c9e8c1e | 8011 | mark_failure(super, dev, disk, idx); |
0556e1a2 DW |
8012 | |
8013 | if (disk->scsi_id == __cpu_to_le32(~(__u32)0)) | |
8014 | return; | |
8015 | ||
47ee5a45 DW |
8016 | disk->scsi_id = __cpu_to_le32(~(__u32)0); |
8017 | memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1); | |
8018 | } | |
8019 | ||
33414a01 DW |
8020 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev) |
8021 | { | |
33414a01 | 8022 | struct dl *dl; |
33414a01 DW |
8023 | |
8024 | if (!super->missing) | |
8025 | return; | |
33414a01 | 8026 | |
79b68f1b PC |
8027 | /* When orom adds replacement for missing disk it does |
8028 | * not remove entry of missing disk, but just updates map with | |
8029 | * new added disk. So it is not enough just to test if there is | |
8030 | * any missing disk, we have to look if there are any failed disks | |
8031 | * in map to stop migration */ | |
8032 | ||
33414a01 | 8033 | dprintf("imsm: mark missing\n"); |
3d59f0c0 AK |
8034 | /* end process for initialization and rebuild only |
8035 | */ | |
8036 | if (is_gen_migration(dev) == 0) { | |
fb12a745 | 8037 | int failed = imsm_count_failed(super, dev, MAP_0); |
3d59f0c0 | 8038 | |
fb12a745 TM |
8039 | if (failed) { |
8040 | __u8 map_state; | |
8041 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
8042 | struct imsm_map *map1; | |
8043 | int i, ord, ord_map1; | |
8044 | int rebuilt = 1; | |
3d59f0c0 | 8045 | |
fb12a745 TM |
8046 | for (i = 0; i < map->num_members; i++) { |
8047 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_0); | |
8048 | if (!(ord & IMSM_ORD_REBUILD)) | |
8049 | continue; | |
8050 | ||
8051 | map1 = get_imsm_map(dev, MAP_1); | |
8052 | if (!map1) | |
8053 | continue; | |
8054 | ||
8055 | ord_map1 = __le32_to_cpu(map1->disk_ord_tbl[i]); | |
8056 | if (ord_map1 & IMSM_ORD_REBUILD) | |
8057 | rebuilt = 0; | |
8058 | } | |
8059 | ||
8060 | if (rebuilt) { | |
8061 | map_state = imsm_check_degraded(super, dev, | |
8062 | failed, MAP_0); | |
8063 | end_migration(dev, super, map_state); | |
8064 | } | |
8065 | } | |
3d59f0c0 | 8066 | } |
33414a01 | 8067 | for (dl = super->missing; dl; dl = dl->next) |
4c9e8c1e | 8068 | mark_missing(super, dev, &dl->disk, dl->index); |
33414a01 DW |
8069 | super->updates_pending++; |
8070 | } | |
8071 | ||
f3871fdc AK |
8072 | static unsigned long long imsm_set_array_size(struct imsm_dev *dev, |
8073 | long long new_size) | |
70bdf0dc | 8074 | { |
238c0a71 | 8075 | int used_disks = imsm_num_data_members(dev, MAP_0); |
70bdf0dc AK |
8076 | unsigned long long array_blocks; |
8077 | struct imsm_map *map; | |
8078 | ||
8079 | if (used_disks == 0) { | |
8080 | /* when problems occures | |
8081 | * return current array_blocks value | |
8082 | */ | |
8083 | array_blocks = __le32_to_cpu(dev->size_high); | |
8084 | array_blocks = array_blocks << 32; | |
8085 | array_blocks += __le32_to_cpu(dev->size_low); | |
8086 | ||
8087 | return array_blocks; | |
8088 | } | |
8089 | ||
8090 | /* set array size in metadata | |
8091 | */ | |
f3871fdc AK |
8092 | if (new_size <= 0) { |
8093 | /* OLCE size change is caused by added disks | |
8094 | */ | |
8095 | map = get_imsm_map(dev, MAP_0); | |
8096 | array_blocks = blocks_per_member(map) * used_disks; | |
8097 | } else { | |
8098 | /* Online Volume Size Change | |
8099 | * Using available free space | |
8100 | */ | |
8101 | array_blocks = new_size; | |
8102 | } | |
70bdf0dc | 8103 | |
b53bfba6 | 8104 | array_blocks = round_size_to_mb(array_blocks, used_disks); |
70bdf0dc AK |
8105 | dev->size_low = __cpu_to_le32((__u32)array_blocks); |
8106 | dev->size_high = __cpu_to_le32((__u32)(array_blocks >> 32)); | |
8107 | ||
8108 | return array_blocks; | |
8109 | } | |
8110 | ||
28bce06f AK |
8111 | static void imsm_set_disk(struct active_array *a, int n, int state); |
8112 | ||
0e2d1a4e AK |
8113 | static void imsm_progress_container_reshape(struct intel_super *super) |
8114 | { | |
8115 | /* if no device has a migr_state, but some device has a | |
8116 | * different number of members than the previous device, start | |
8117 | * changing the number of devices in this device to match | |
8118 | * previous. | |
8119 | */ | |
8120 | struct imsm_super *mpb = super->anchor; | |
8121 | int prev_disks = -1; | |
8122 | int i; | |
1dfaa380 | 8123 | int copy_map_size; |
0e2d1a4e AK |
8124 | |
8125 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
8126 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 8127 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
0e2d1a4e AK |
8128 | struct imsm_map *map2; |
8129 | int prev_num_members; | |
0e2d1a4e AK |
8130 | |
8131 | if (dev->vol.migr_state) | |
8132 | return; | |
8133 | ||
8134 | if (prev_disks == -1) | |
8135 | prev_disks = map->num_members; | |
8136 | if (prev_disks == map->num_members) | |
8137 | continue; | |
8138 | ||
8139 | /* OK, this array needs to enter reshape mode. | |
8140 | * i.e it needs a migr_state | |
8141 | */ | |
8142 | ||
1dfaa380 | 8143 | copy_map_size = sizeof_imsm_map(map); |
0e2d1a4e AK |
8144 | prev_num_members = map->num_members; |
8145 | map->num_members = prev_disks; | |
8146 | dev->vol.migr_state = 1; | |
8147 | dev->vol.curr_migr_unit = 0; | |
ea672ee1 | 8148 | set_migr_type(dev, MIGR_GEN_MIGR); |
0e2d1a4e AK |
8149 | for (i = prev_num_members; |
8150 | i < map->num_members; i++) | |
8151 | set_imsm_ord_tbl_ent(map, i, i); | |
238c0a71 | 8152 | map2 = get_imsm_map(dev, MAP_1); |
0e2d1a4e | 8153 | /* Copy the current map */ |
1dfaa380 | 8154 | memcpy(map2, map, copy_map_size); |
0e2d1a4e AK |
8155 | map2->num_members = prev_num_members; |
8156 | ||
f3871fdc | 8157 | imsm_set_array_size(dev, -1); |
51d83f5d | 8158 | super->clean_migration_record_by_mdmon = 1; |
0e2d1a4e AK |
8159 | super->updates_pending++; |
8160 | } | |
8161 | } | |
8162 | ||
aad6f216 | 8163 | /* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild |
0c046afd DW |
8164 | * states are handled in imsm_set_disk() with one exception, when a |
8165 | * resync is stopped due to a new failure this routine will set the | |
8166 | * 'degraded' state for the array. | |
8167 | */ | |
01f157d7 | 8168 | static int imsm_set_array_state(struct active_array *a, int consistent) |
a862209d DW |
8169 | { |
8170 | int inst = a->info.container_member; | |
8171 | struct intel_super *super = a->container->sb; | |
949c47a0 | 8172 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 8173 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
3b451610 AK |
8174 | int failed = imsm_count_failed(super, dev, MAP_0); |
8175 | __u8 map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
1e5c6983 | 8176 | __u32 blocks_per_unit; |
a862209d | 8177 | |
1af97990 AK |
8178 | if (dev->vol.migr_state && |
8179 | dev->vol.migr_type == MIGR_GEN_MIGR) { | |
8180 | /* array state change is blocked due to reshape action | |
aad6f216 N |
8181 | * We might need to |
8182 | * - abort the reshape (if last_checkpoint is 0 and action!= reshape) | |
8183 | * - finish the reshape (if last_checkpoint is big and action != reshape) | |
8184 | * - update curr_migr_unit | |
1af97990 | 8185 | */ |
aad6f216 N |
8186 | if (a->curr_action == reshape) { |
8187 | /* still reshaping, maybe update curr_migr_unit */ | |
633b5610 | 8188 | goto mark_checkpoint; |
aad6f216 N |
8189 | } else { |
8190 | if (a->last_checkpoint == 0 && a->prev_action == reshape) { | |
8191 | /* for some reason we aborted the reshape. | |
b66e591b AK |
8192 | * |
8193 | * disable automatic metadata rollback | |
8194 | * user action is required to recover process | |
aad6f216 | 8195 | */ |
b66e591b | 8196 | if (0) { |
238c0a71 AK |
8197 | struct imsm_map *map2 = |
8198 | get_imsm_map(dev, MAP_1); | |
8199 | dev->vol.migr_state = 0; | |
8200 | set_migr_type(dev, 0); | |
8201 | dev->vol.curr_migr_unit = 0; | |
8202 | memcpy(map, map2, | |
8203 | sizeof_imsm_map(map2)); | |
8204 | super->updates_pending++; | |
b66e591b | 8205 | } |
aad6f216 N |
8206 | } |
8207 | if (a->last_checkpoint >= a->info.component_size) { | |
8208 | unsigned long long array_blocks; | |
8209 | int used_disks; | |
e154ced3 | 8210 | struct mdinfo *mdi; |
aad6f216 | 8211 | |
238c0a71 | 8212 | used_disks = imsm_num_data_members(dev, MAP_0); |
d55adef9 AK |
8213 | if (used_disks > 0) { |
8214 | array_blocks = | |
5551b113 | 8215 | blocks_per_member(map) * |
d55adef9 | 8216 | used_disks; |
b53bfba6 TM |
8217 | array_blocks = |
8218 | round_size_to_mb(array_blocks, | |
8219 | used_disks); | |
d55adef9 AK |
8220 | a->info.custom_array_size = array_blocks; |
8221 | /* encourage manager to update array | |
8222 | * size | |
8223 | */ | |
e154ced3 | 8224 | |
d55adef9 | 8225 | a->check_reshape = 1; |
633b5610 | 8226 | } |
e154ced3 AK |
8227 | /* finalize online capacity expansion/reshape */ |
8228 | for (mdi = a->info.devs; mdi; mdi = mdi->next) | |
8229 | imsm_set_disk(a, | |
8230 | mdi->disk.raid_disk, | |
8231 | mdi->curr_state); | |
8232 | ||
0e2d1a4e | 8233 | imsm_progress_container_reshape(super); |
e154ced3 | 8234 | } |
aad6f216 | 8235 | } |
1af97990 AK |
8236 | } |
8237 | ||
47ee5a45 | 8238 | /* before we activate this array handle any missing disks */ |
33414a01 DW |
8239 | if (consistent == 2) |
8240 | handle_missing(super, dev); | |
1e5c6983 | 8241 | |
0c046afd | 8242 | if (consistent == 2 && |
b7941fd6 | 8243 | (!is_resync_complete(&a->info) || |
0c046afd DW |
8244 | map_state != IMSM_T_STATE_NORMAL || |
8245 | dev->vol.migr_state)) | |
01f157d7 | 8246 | consistent = 0; |
272906ef | 8247 | |
b7941fd6 | 8248 | if (is_resync_complete(&a->info)) { |
0c046afd | 8249 | /* complete intialization / resync, |
0556e1a2 DW |
8250 | * recovery and interrupted recovery is completed in |
8251 | * ->set_disk | |
0c046afd DW |
8252 | */ |
8253 | if (is_resyncing(dev)) { | |
8254 | dprintf("imsm: mark resync done\n"); | |
809da78e | 8255 | end_migration(dev, super, map_state); |
115c3803 | 8256 | super->updates_pending++; |
484240d8 | 8257 | a->last_checkpoint = 0; |
115c3803 | 8258 | } |
b9172665 AK |
8259 | } else if ((!is_resyncing(dev) && !failed) && |
8260 | (imsm_reshape_blocks_arrays_changes(super) == 0)) { | |
0c046afd | 8261 | /* mark the start of the init process if nothing is failed */ |
b7941fd6 | 8262 | dprintf("imsm: mark resync start\n"); |
1484e727 | 8263 | if (map->map_state == IMSM_T_STATE_UNINITIALIZED) |
8e59f3d8 | 8264 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_INIT); |
1484e727 | 8265 | else |
8e59f3d8 | 8266 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_REPAIR); |
3393c6af | 8267 | super->updates_pending++; |
115c3803 | 8268 | } |
a862209d | 8269 | |
633b5610 | 8270 | mark_checkpoint: |
5b83bacf AK |
8271 | /* skip checkpointing for general migration, |
8272 | * it is controlled in mdadm | |
8273 | */ | |
8274 | if (is_gen_migration(dev)) | |
8275 | goto skip_mark_checkpoint; | |
8276 | ||
1e5c6983 | 8277 | /* check if we can update curr_migr_unit from resync_start, recovery_start */ |
c47b0ff6 | 8278 | blocks_per_unit = blocks_per_migr_unit(super, dev); |
4f0a7acc | 8279 | if (blocks_per_unit) { |
1e5c6983 DW |
8280 | __u32 units32; |
8281 | __u64 units; | |
8282 | ||
4f0a7acc | 8283 | units = a->last_checkpoint / blocks_per_unit; |
1e5c6983 DW |
8284 | units32 = units; |
8285 | ||
8286 | /* check that we did not overflow 32-bits, and that | |
8287 | * curr_migr_unit needs updating | |
8288 | */ | |
8289 | if (units32 == units && | |
bfd80a56 | 8290 | units32 != 0 && |
1e5c6983 DW |
8291 | __le32_to_cpu(dev->vol.curr_migr_unit) != units32) { |
8292 | dprintf("imsm: mark checkpoint (%u)\n", units32); | |
8293 | dev->vol.curr_migr_unit = __cpu_to_le32(units32); | |
8294 | super->updates_pending++; | |
8295 | } | |
8296 | } | |
f8f603f1 | 8297 | |
5b83bacf | 8298 | skip_mark_checkpoint: |
3393c6af | 8299 | /* mark dirty / clean */ |
2432ce9b AP |
8300 | if (((dev->vol.dirty & RAIDVOL_DIRTY) && consistent) || |
8301 | (!(dev->vol.dirty & RAIDVOL_DIRTY) && !consistent)) { | |
b7941fd6 | 8302 | dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty"); |
2432ce9b AP |
8303 | if (consistent) { |
8304 | dev->vol.dirty = RAIDVOL_CLEAN; | |
8305 | } else { | |
8306 | dev->vol.dirty = RAIDVOL_DIRTY; | |
c2462068 PB |
8307 | if (dev->rwh_policy == RWH_DISTRIBUTED || |
8308 | dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED) | |
2432ce9b AP |
8309 | dev->vol.dirty |= RAIDVOL_DSRECORD_VALID; |
8310 | } | |
a862209d DW |
8311 | super->updates_pending++; |
8312 | } | |
28bce06f | 8313 | |
01f157d7 | 8314 | return consistent; |
a862209d DW |
8315 | } |
8316 | ||
6f50473f TM |
8317 | static int imsm_disk_slot_to_ord(struct active_array *a, int slot) |
8318 | { | |
8319 | int inst = a->info.container_member; | |
8320 | struct intel_super *super = a->container->sb; | |
8321 | struct imsm_dev *dev = get_imsm_dev(super, inst); | |
8322 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
8323 | ||
8324 | if (slot > map->num_members) { | |
8325 | pr_err("imsm: imsm_disk_slot_to_ord %d out of range 0..%d\n", | |
8326 | slot, map->num_members - 1); | |
8327 | return -1; | |
8328 | } | |
8329 | ||
8330 | if (slot < 0) | |
8331 | return -1; | |
8332 | ||
8333 | return get_imsm_ord_tbl_ent(dev, slot, MAP_0); | |
8334 | } | |
8335 | ||
8d45d196 | 8336 | static void imsm_set_disk(struct active_array *a, int n, int state) |
845dea95 | 8337 | { |
8d45d196 DW |
8338 | int inst = a->info.container_member; |
8339 | struct intel_super *super = a->container->sb; | |
949c47a0 | 8340 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 8341 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
8d45d196 | 8342 | struct imsm_disk *disk; |
7ce05701 LD |
8343 | struct mdinfo *mdi; |
8344 | int recovery_not_finished = 0; | |
0c046afd | 8345 | int failed; |
6f50473f | 8346 | int ord; |
0c046afd | 8347 | __u8 map_state; |
fb12a745 TM |
8348 | int rebuild_done = 0; |
8349 | int i; | |
8d45d196 | 8350 | |
fb12a745 | 8351 | ord = get_imsm_ord_tbl_ent(dev, n, MAP_X); |
6f50473f | 8352 | if (ord < 0) |
8d45d196 DW |
8353 | return; |
8354 | ||
4e6e574a | 8355 | dprintf("imsm: set_disk %d:%x\n", n, state); |
b10b37b8 | 8356 | disk = get_imsm_disk(super, ord_to_idx(ord)); |
8d45d196 | 8357 | |
5802a811 | 8358 | /* check for new failures */ |
0556e1a2 | 8359 | if (state & DS_FAULTY) { |
4c9e8c1e | 8360 | if (mark_failure(super, dev, disk, ord_to_idx(ord))) |
0556e1a2 | 8361 | super->updates_pending++; |
8d45d196 | 8362 | } |
47ee5a45 | 8363 | |
19859edc | 8364 | /* check if in_sync */ |
0556e1a2 | 8365 | if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) { |
238c0a71 | 8366 | struct imsm_map *migr_map = get_imsm_map(dev, MAP_1); |
b10b37b8 DW |
8367 | |
8368 | set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord)); | |
fb12a745 | 8369 | rebuild_done = 1; |
19859edc DW |
8370 | super->updates_pending++; |
8371 | } | |
8d45d196 | 8372 | |
3b451610 AK |
8373 | failed = imsm_count_failed(super, dev, MAP_0); |
8374 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
5802a811 | 8375 | |
0c046afd | 8376 | /* check if recovery complete, newly degraded, or failed */ |
94002678 AK |
8377 | dprintf("imsm: Detected transition to state "); |
8378 | switch (map_state) { | |
8379 | case IMSM_T_STATE_NORMAL: /* transition to normal state */ | |
8380 | dprintf("normal: "); | |
8381 | if (is_rebuilding(dev)) { | |
1ade5cc1 | 8382 | dprintf_cont("while rebuilding"); |
7ce05701 LD |
8383 | /* check if recovery is really finished */ |
8384 | for (mdi = a->info.devs; mdi ; mdi = mdi->next) | |
8385 | if (mdi->recovery_start != MaxSector) { | |
8386 | recovery_not_finished = 1; | |
8387 | break; | |
8388 | } | |
8389 | if (recovery_not_finished) { | |
1ade5cc1 N |
8390 | dprintf_cont("\n"); |
8391 | dprintf("Rebuild has not finished yet, state not changed"); | |
7ce05701 LD |
8392 | if (a->last_checkpoint < mdi->recovery_start) { |
8393 | a->last_checkpoint = mdi->recovery_start; | |
8394 | super->updates_pending++; | |
8395 | } | |
8396 | break; | |
8397 | } | |
94002678 | 8398 | end_migration(dev, super, map_state); |
238c0a71 | 8399 | map = get_imsm_map(dev, MAP_0); |
94002678 AK |
8400 | map->failed_disk_num = ~0; |
8401 | super->updates_pending++; | |
8402 | a->last_checkpoint = 0; | |
8403 | break; | |
8404 | } | |
8405 | if (is_gen_migration(dev)) { | |
1ade5cc1 | 8406 | dprintf_cont("while general migration"); |
bf2f0071 | 8407 | if (a->last_checkpoint >= a->info.component_size) |
809da78e | 8408 | end_migration(dev, super, map_state); |
94002678 AK |
8409 | else |
8410 | map->map_state = map_state; | |
238c0a71 | 8411 | map = get_imsm_map(dev, MAP_0); |
28bce06f | 8412 | map->failed_disk_num = ~0; |
94002678 | 8413 | super->updates_pending++; |
bf2f0071 | 8414 | break; |
94002678 AK |
8415 | } |
8416 | break; | |
8417 | case IMSM_T_STATE_DEGRADED: /* transition to degraded state */ | |
1ade5cc1 | 8418 | dprintf_cont("degraded: "); |
089f9d79 | 8419 | if (map->map_state != map_state && !dev->vol.migr_state) { |
1ade5cc1 | 8420 | dprintf_cont("mark degraded"); |
94002678 AK |
8421 | map->map_state = map_state; |
8422 | super->updates_pending++; | |
8423 | a->last_checkpoint = 0; | |
8424 | break; | |
8425 | } | |
8426 | if (is_rebuilding(dev)) { | |
1ade5cc1 | 8427 | dprintf_cont("while rebuilding."); |
94002678 | 8428 | if (map->map_state != map_state) { |
1ade5cc1 | 8429 | dprintf_cont(" Map state change"); |
94002678 AK |
8430 | end_migration(dev, super, map_state); |
8431 | super->updates_pending++; | |
fb12a745 TM |
8432 | } else if (!rebuild_done) { |
8433 | break; | |
8434 | } | |
8435 | ||
8436 | /* check if recovery is really finished */ | |
8437 | for (mdi = a->info.devs; mdi ; mdi = mdi->next) | |
8438 | if (mdi->recovery_start != MaxSector) { | |
8439 | recovery_not_finished = 1; | |
8440 | break; | |
8441 | } | |
8442 | if (recovery_not_finished) { | |
8443 | dprintf_cont("\n"); | |
8444 | dprintf("Rebuild has not finished yet, state not changed"); | |
8445 | if (a->last_checkpoint < mdi->recovery_start) { | |
8446 | a->last_checkpoint = | |
8447 | mdi->recovery_start; | |
8448 | super->updates_pending++; | |
8449 | } | |
8450 | break; | |
94002678 | 8451 | } |
fb12a745 TM |
8452 | |
8453 | dprintf_cont(" Rebuild done, still degraded"); | |
8454 | dev->vol.migr_state = 0; | |
8455 | set_migr_type(dev, 0); | |
8456 | dev->vol.curr_migr_unit = 0; | |
8457 | ||
8458 | for (i = 0; i < map->num_members; i++) { | |
8459 | int idx = get_imsm_ord_tbl_ent(dev, i, MAP_0); | |
8460 | ||
8461 | if (idx & IMSM_ORD_REBUILD) | |
8462 | map->failed_disk_num = i; | |
8463 | } | |
8464 | super->updates_pending++; | |
94002678 AK |
8465 | break; |
8466 | } | |
8467 | if (is_gen_migration(dev)) { | |
1ade5cc1 | 8468 | dprintf_cont("while general migration"); |
bf2f0071 | 8469 | if (a->last_checkpoint >= a->info.component_size) |
809da78e | 8470 | end_migration(dev, super, map_state); |
94002678 AK |
8471 | else { |
8472 | map->map_state = map_state; | |
3b451610 | 8473 | manage_second_map(super, dev); |
94002678 AK |
8474 | } |
8475 | super->updates_pending++; | |
bf2f0071 | 8476 | break; |
28bce06f | 8477 | } |
6ce1fbf1 | 8478 | if (is_initializing(dev)) { |
1ade5cc1 | 8479 | dprintf_cont("while initialization."); |
6ce1fbf1 AK |
8480 | map->map_state = map_state; |
8481 | super->updates_pending++; | |
8482 | break; | |
8483 | } | |
94002678 AK |
8484 | break; |
8485 | case IMSM_T_STATE_FAILED: /* transition to failed state */ | |
1ade5cc1 | 8486 | dprintf_cont("failed: "); |
94002678 | 8487 | if (is_gen_migration(dev)) { |
1ade5cc1 | 8488 | dprintf_cont("while general migration"); |
94002678 AK |
8489 | map->map_state = map_state; |
8490 | super->updates_pending++; | |
8491 | break; | |
8492 | } | |
8493 | if (map->map_state != map_state) { | |
1ade5cc1 | 8494 | dprintf_cont("mark failed"); |
94002678 AK |
8495 | end_migration(dev, super, map_state); |
8496 | super->updates_pending++; | |
8497 | a->last_checkpoint = 0; | |
8498 | break; | |
8499 | } | |
8500 | break; | |
8501 | default: | |
1ade5cc1 | 8502 | dprintf_cont("state %i\n", map_state); |
5802a811 | 8503 | } |
1ade5cc1 | 8504 | dprintf_cont("\n"); |
845dea95 NB |
8505 | } |
8506 | ||
f796af5d | 8507 | static int store_imsm_mpb(int fd, struct imsm_super *mpb) |
c2a1e7da | 8508 | { |
f796af5d | 8509 | void *buf = mpb; |
c2a1e7da DW |
8510 | __u32 mpb_size = __le32_to_cpu(mpb->mpb_size); |
8511 | unsigned long long dsize; | |
8512 | unsigned long long sectors; | |
f36a9ecd | 8513 | unsigned int sector_size; |
c2a1e7da | 8514 | |
f36a9ecd | 8515 | get_dev_sector_size(fd, NULL, §or_size); |
c2a1e7da DW |
8516 | get_dev_size(fd, NULL, &dsize); |
8517 | ||
f36a9ecd | 8518 | if (mpb_size > sector_size) { |
272f648f | 8519 | /* -1 to account for anchor */ |
f36a9ecd | 8520 | sectors = mpb_sectors(mpb, sector_size) - 1; |
c2a1e7da | 8521 | |
272f648f | 8522 | /* write the extended mpb to the sectors preceeding the anchor */ |
f36a9ecd PB |
8523 | if (lseek64(fd, dsize - (sector_size * (2 + sectors)), |
8524 | SEEK_SET) < 0) | |
272f648f | 8525 | return 1; |
c2a1e7da | 8526 | |
f36a9ecd PB |
8527 | if ((unsigned long long)write(fd, buf + sector_size, |
8528 | sector_size * sectors) != sector_size * sectors) | |
272f648f DW |
8529 | return 1; |
8530 | } | |
c2a1e7da | 8531 | |
272f648f | 8532 | /* first block is stored on second to last sector of the disk */ |
f36a9ecd | 8533 | if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0) |
c2a1e7da DW |
8534 | return 1; |
8535 | ||
466070ad | 8536 | if ((unsigned int)write(fd, buf, sector_size) != sector_size) |
c2a1e7da DW |
8537 | return 1; |
8538 | ||
c2a1e7da DW |
8539 | return 0; |
8540 | } | |
8541 | ||
2e735d19 | 8542 | static void imsm_sync_metadata(struct supertype *container) |
845dea95 | 8543 | { |
2e735d19 | 8544 | struct intel_super *super = container->sb; |
c2a1e7da | 8545 | |
1a64be56 | 8546 | dprintf("sync metadata: %d\n", super->updates_pending); |
c2a1e7da DW |
8547 | if (!super->updates_pending) |
8548 | return; | |
8549 | ||
36988a3d | 8550 | write_super_imsm(container, 0); |
c2a1e7da DW |
8551 | |
8552 | super->updates_pending = 0; | |
845dea95 NB |
8553 | } |
8554 | ||
272906ef DW |
8555 | static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a) |
8556 | { | |
8557 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 8558 | int i = get_imsm_disk_idx(dev, idx, MAP_X); |
272906ef DW |
8559 | struct dl *dl; |
8560 | ||
8561 | for (dl = super->disks; dl; dl = dl->next) | |
8562 | if (dl->index == i) | |
8563 | break; | |
8564 | ||
25ed7e59 | 8565 | if (dl && is_failed(&dl->disk)) |
272906ef DW |
8566 | dl = NULL; |
8567 | ||
8568 | if (dl) | |
1ade5cc1 | 8569 | dprintf("found %x:%x\n", dl->major, dl->minor); |
272906ef DW |
8570 | |
8571 | return dl; | |
8572 | } | |
8573 | ||
a20d2ba5 | 8574 | static struct dl *imsm_add_spare(struct intel_super *super, int slot, |
8ba77d32 AK |
8575 | struct active_array *a, int activate_new, |
8576 | struct mdinfo *additional_test_list) | |
272906ef DW |
8577 | { |
8578 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 8579 | int idx = get_imsm_disk_idx(dev, slot, MAP_X); |
a20d2ba5 DW |
8580 | struct imsm_super *mpb = super->anchor; |
8581 | struct imsm_map *map; | |
272906ef DW |
8582 | unsigned long long pos; |
8583 | struct mdinfo *d; | |
8584 | struct extent *ex; | |
a20d2ba5 | 8585 | int i, j; |
272906ef | 8586 | int found; |
569cc43f DW |
8587 | __u32 array_start = 0; |
8588 | __u32 array_end = 0; | |
272906ef | 8589 | struct dl *dl; |
6c932028 | 8590 | struct mdinfo *test_list; |
272906ef DW |
8591 | |
8592 | for (dl = super->disks; dl; dl = dl->next) { | |
8593 | /* If in this array, skip */ | |
8594 | for (d = a->info.devs ; d ; d = d->next) | |
e553d2a4 DW |
8595 | if (d->state_fd >= 0 && |
8596 | d->disk.major == dl->major && | |
272906ef | 8597 | d->disk.minor == dl->minor) { |
8ba77d32 AK |
8598 | dprintf("%x:%x already in array\n", |
8599 | dl->major, dl->minor); | |
272906ef DW |
8600 | break; |
8601 | } | |
8602 | if (d) | |
8603 | continue; | |
6c932028 AK |
8604 | test_list = additional_test_list; |
8605 | while (test_list) { | |
8606 | if (test_list->disk.major == dl->major && | |
8607 | test_list->disk.minor == dl->minor) { | |
8ba77d32 AK |
8608 | dprintf("%x:%x already in additional test list\n", |
8609 | dl->major, dl->minor); | |
8610 | break; | |
8611 | } | |
6c932028 | 8612 | test_list = test_list->next; |
8ba77d32 | 8613 | } |
6c932028 | 8614 | if (test_list) |
8ba77d32 | 8615 | continue; |
272906ef | 8616 | |
e553d2a4 | 8617 | /* skip in use or failed drives */ |
25ed7e59 | 8618 | if (is_failed(&dl->disk) || idx == dl->index || |
df474657 DW |
8619 | dl->index == -2) { |
8620 | dprintf("%x:%x status (failed: %d index: %d)\n", | |
25ed7e59 | 8621 | dl->major, dl->minor, is_failed(&dl->disk), idx); |
9a1608e5 DW |
8622 | continue; |
8623 | } | |
8624 | ||
a20d2ba5 DW |
8625 | /* skip pure spares when we are looking for partially |
8626 | * assimilated drives | |
8627 | */ | |
8628 | if (dl->index == -1 && !activate_new) | |
8629 | continue; | |
8630 | ||
f2cc4f7d AO |
8631 | if (!drive_validate_sector_size(super, dl)) |
8632 | continue; | |
8633 | ||
272906ef | 8634 | /* Does this unused device have the requisite free space? |
a20d2ba5 | 8635 | * It needs to be able to cover all member volumes |
272906ef DW |
8636 | */ |
8637 | ex = get_extents(super, dl); | |
8638 | if (!ex) { | |
8639 | dprintf("cannot get extents\n"); | |
8640 | continue; | |
8641 | } | |
a20d2ba5 DW |
8642 | for (i = 0; i < mpb->num_raid_devs; i++) { |
8643 | dev = get_imsm_dev(super, i); | |
238c0a71 | 8644 | map = get_imsm_map(dev, MAP_0); |
272906ef | 8645 | |
a20d2ba5 DW |
8646 | /* check if this disk is already a member of |
8647 | * this array | |
272906ef | 8648 | */ |
620b1713 | 8649 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
a20d2ba5 DW |
8650 | continue; |
8651 | ||
8652 | found = 0; | |
8653 | j = 0; | |
8654 | pos = 0; | |
5551b113 | 8655 | array_start = pba_of_lba0(map); |
329c8278 | 8656 | array_end = array_start + |
5551b113 | 8657 | blocks_per_member(map) - 1; |
a20d2ba5 DW |
8658 | |
8659 | do { | |
8660 | /* check that we can start at pba_of_lba0 with | |
8661 | * blocks_per_member of space | |
8662 | */ | |
329c8278 | 8663 | if (array_start >= pos && array_end < ex[j].start) { |
a20d2ba5 DW |
8664 | found = 1; |
8665 | break; | |
8666 | } | |
8667 | pos = ex[j].start + ex[j].size; | |
8668 | j++; | |
8669 | } while (ex[j-1].size); | |
8670 | ||
8671 | if (!found) | |
272906ef | 8672 | break; |
a20d2ba5 | 8673 | } |
272906ef DW |
8674 | |
8675 | free(ex); | |
a20d2ba5 | 8676 | if (i < mpb->num_raid_devs) { |
329c8278 DW |
8677 | dprintf("%x:%x does not have %u to %u available\n", |
8678 | dl->major, dl->minor, array_start, array_end); | |
272906ef DW |
8679 | /* No room */ |
8680 | continue; | |
a20d2ba5 DW |
8681 | } |
8682 | return dl; | |
272906ef DW |
8683 | } |
8684 | ||
8685 | return dl; | |
8686 | } | |
8687 | ||
95d07a2c LM |
8688 | static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed) |
8689 | { | |
8690 | struct imsm_dev *dev2; | |
8691 | struct imsm_map *map; | |
8692 | struct dl *idisk; | |
8693 | int slot; | |
8694 | int idx; | |
8695 | __u8 state; | |
8696 | ||
8697 | dev2 = get_imsm_dev(cont->sb, dev_idx); | |
8698 | if (dev2) { | |
238c0a71 | 8699 | state = imsm_check_degraded(cont->sb, dev2, failed, MAP_0); |
95d07a2c | 8700 | if (state == IMSM_T_STATE_FAILED) { |
238c0a71 | 8701 | map = get_imsm_map(dev2, MAP_0); |
95d07a2c LM |
8702 | if (!map) |
8703 | return 1; | |
8704 | for (slot = 0; slot < map->num_members; slot++) { | |
8705 | /* | |
8706 | * Check if failed disks are deleted from intel | |
8707 | * disk list or are marked to be deleted | |
8708 | */ | |
238c0a71 | 8709 | idx = get_imsm_disk_idx(dev2, slot, MAP_X); |
95d07a2c LM |
8710 | idisk = get_imsm_dl_disk(cont->sb, idx); |
8711 | /* | |
8712 | * Do not rebuild the array if failed disks | |
8713 | * from failed sub-array are not removed from | |
8714 | * container. | |
8715 | */ | |
8716 | if (idisk && | |
8717 | is_failed(&idisk->disk) && | |
8718 | (idisk->action != DISK_REMOVE)) | |
8719 | return 0; | |
8720 | } | |
8721 | } | |
8722 | } | |
8723 | return 1; | |
8724 | } | |
8725 | ||
88758e9d DW |
8726 | static struct mdinfo *imsm_activate_spare(struct active_array *a, |
8727 | struct metadata_update **updates) | |
8728 | { | |
8729 | /** | |
d23fe947 DW |
8730 | * Find a device with unused free space and use it to replace a |
8731 | * failed/vacant region in an array. We replace failed regions one a | |
8732 | * array at a time. The result is that a new spare disk will be added | |
8733 | * to the first failed array and after the monitor has finished | |
8734 | * propagating failures the remainder will be consumed. | |
88758e9d | 8735 | * |
d23fe947 DW |
8736 | * FIXME add a capability for mdmon to request spares from another |
8737 | * container. | |
88758e9d DW |
8738 | */ |
8739 | ||
8740 | struct intel_super *super = a->container->sb; | |
88758e9d | 8741 | int inst = a->info.container_member; |
949c47a0 | 8742 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 8743 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
88758e9d DW |
8744 | int failed = a->info.array.raid_disks; |
8745 | struct mdinfo *rv = NULL; | |
8746 | struct mdinfo *d; | |
8747 | struct mdinfo *di; | |
8748 | struct metadata_update *mu; | |
8749 | struct dl *dl; | |
8750 | struct imsm_update_activate_spare *u; | |
8751 | int num_spares = 0; | |
8752 | int i; | |
95d07a2c | 8753 | int allowed; |
88758e9d DW |
8754 | |
8755 | for (d = a->info.devs ; d ; d = d->next) { | |
8756 | if ((d->curr_state & DS_FAULTY) && | |
8757 | d->state_fd >= 0) | |
8758 | /* wait for Removal to happen */ | |
8759 | return NULL; | |
8760 | if (d->state_fd >= 0) | |
8761 | failed--; | |
8762 | } | |
8763 | ||
8764 | dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n", | |
8765 | inst, failed, a->info.array.raid_disks, a->info.array.level); | |
1af97990 | 8766 | |
e2962bfc AK |
8767 | if (imsm_reshape_blocks_arrays_changes(super)) |
8768 | return NULL; | |
1af97990 | 8769 | |
fc8ca064 AK |
8770 | /* Cannot activate another spare if rebuild is in progress already |
8771 | */ | |
8772 | if (is_rebuilding(dev)) { | |
7a862a02 | 8773 | dprintf("imsm: No spare activation allowed. Rebuild in progress already.\n"); |
fc8ca064 AK |
8774 | return NULL; |
8775 | } | |
8776 | ||
89c67882 AK |
8777 | if (a->info.array.level == 4) |
8778 | /* No repair for takeovered array | |
8779 | * imsm doesn't support raid4 | |
8780 | */ | |
8781 | return NULL; | |
8782 | ||
3b451610 AK |
8783 | if (imsm_check_degraded(super, dev, failed, MAP_0) != |
8784 | IMSM_T_STATE_DEGRADED) | |
88758e9d DW |
8785 | return NULL; |
8786 | ||
83ca7d45 AP |
8787 | if (get_imsm_map(dev, MAP_0)->map_state == IMSM_T_STATE_UNINITIALIZED) { |
8788 | dprintf("imsm: No spare activation allowed. Volume is not initialized.\n"); | |
8789 | return NULL; | |
8790 | } | |
8791 | ||
95d07a2c LM |
8792 | /* |
8793 | * If there are any failed disks check state of the other volume. | |
8794 | * Block rebuild if the another one is failed until failed disks | |
8795 | * are removed from container. | |
8796 | */ | |
8797 | if (failed) { | |
7a862a02 | 8798 | dprintf("found failed disks in %.*s, check if there anotherfailed sub-array.\n", |
c4acd1e5 | 8799 | MAX_RAID_SERIAL_LEN, dev->volume); |
95d07a2c LM |
8800 | /* check if states of the other volumes allow for rebuild */ |
8801 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
8802 | if (i != inst) { | |
8803 | allowed = imsm_rebuild_allowed(a->container, | |
8804 | i, failed); | |
8805 | if (!allowed) | |
8806 | return NULL; | |
8807 | } | |
8808 | } | |
8809 | } | |
8810 | ||
88758e9d | 8811 | /* For each slot, if it is not working, find a spare */ |
88758e9d DW |
8812 | for (i = 0; i < a->info.array.raid_disks; i++) { |
8813 | for (d = a->info.devs ; d ; d = d->next) | |
8814 | if (d->disk.raid_disk == i) | |
8815 | break; | |
8816 | dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0); | |
8817 | if (d && (d->state_fd >= 0)) | |
8818 | continue; | |
8819 | ||
272906ef | 8820 | /* |
a20d2ba5 DW |
8821 | * OK, this device needs recovery. Try to re-add the |
8822 | * previous occupant of this slot, if this fails see if | |
8823 | * we can continue the assimilation of a spare that was | |
8824 | * partially assimilated, finally try to activate a new | |
8825 | * spare. | |
272906ef DW |
8826 | */ |
8827 | dl = imsm_readd(super, i, a); | |
8828 | if (!dl) | |
b303fe21 | 8829 | dl = imsm_add_spare(super, i, a, 0, rv); |
a20d2ba5 | 8830 | if (!dl) |
b303fe21 | 8831 | dl = imsm_add_spare(super, i, a, 1, rv); |
272906ef DW |
8832 | if (!dl) |
8833 | continue; | |
1011e834 | 8834 | |
272906ef | 8835 | /* found a usable disk with enough space */ |
503975b9 | 8836 | di = xcalloc(1, sizeof(*di)); |
272906ef DW |
8837 | |
8838 | /* dl->index will be -1 in the case we are activating a | |
8839 | * pristine spare. imsm_process_update() will create a | |
8840 | * new index in this case. Once a disk is found to be | |
8841 | * failed in all member arrays it is kicked from the | |
8842 | * metadata | |
8843 | */ | |
8844 | di->disk.number = dl->index; | |
d23fe947 | 8845 | |
272906ef DW |
8846 | /* (ab)use di->devs to store a pointer to the device |
8847 | * we chose | |
8848 | */ | |
8849 | di->devs = (struct mdinfo *) dl; | |
8850 | ||
8851 | di->disk.raid_disk = i; | |
8852 | di->disk.major = dl->major; | |
8853 | di->disk.minor = dl->minor; | |
8854 | di->disk.state = 0; | |
d23534e4 | 8855 | di->recovery_start = 0; |
5551b113 | 8856 | di->data_offset = pba_of_lba0(map); |
272906ef DW |
8857 | di->component_size = a->info.component_size; |
8858 | di->container_member = inst; | |
5e46202e | 8859 | di->bb.supported = 1; |
2c8890e9 | 8860 | if (a->info.consistency_policy == CONSISTENCY_POLICY_PPL) { |
2432ce9b | 8861 | di->ppl_sector = get_ppl_sector(super, inst); |
c2462068 | 8862 | di->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9; |
2432ce9b | 8863 | } |
148acb7b | 8864 | super->random = random32(); |
272906ef DW |
8865 | di->next = rv; |
8866 | rv = di; | |
8867 | num_spares++; | |
8868 | dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor, | |
8869 | i, di->data_offset); | |
88758e9d DW |
8870 | } |
8871 | ||
8872 | if (!rv) | |
8873 | /* No spares found */ | |
8874 | return rv; | |
8875 | /* Now 'rv' has a list of devices to return. | |
8876 | * Create a metadata_update record to update the | |
8877 | * disk_ord_tbl for the array | |
8878 | */ | |
503975b9 | 8879 | mu = xmalloc(sizeof(*mu)); |
1011e834 | 8880 | mu->buf = xcalloc(num_spares, |
503975b9 | 8881 | sizeof(struct imsm_update_activate_spare)); |
88758e9d | 8882 | mu->space = NULL; |
cb23f1f4 | 8883 | mu->space_list = NULL; |
88758e9d DW |
8884 | mu->len = sizeof(struct imsm_update_activate_spare) * num_spares; |
8885 | mu->next = *updates; | |
8886 | u = (struct imsm_update_activate_spare *) mu->buf; | |
8887 | ||
8888 | for (di = rv ; di ; di = di->next) { | |
8889 | u->type = update_activate_spare; | |
d23fe947 DW |
8890 | u->dl = (struct dl *) di->devs; |
8891 | di->devs = NULL; | |
88758e9d DW |
8892 | u->slot = di->disk.raid_disk; |
8893 | u->array = inst; | |
8894 | u->next = u + 1; | |
8895 | u++; | |
8896 | } | |
8897 | (u-1)->next = NULL; | |
8898 | *updates = mu; | |
8899 | ||
8900 | return rv; | |
8901 | } | |
8902 | ||
54c2c1ea | 8903 | static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u) |
8273f55e | 8904 | { |
54c2c1ea | 8905 | struct imsm_dev *dev = get_imsm_dev(super, idx); |
238c0a71 AK |
8906 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
8907 | struct imsm_map *new_map = get_imsm_map(&u->dev, MAP_0); | |
54c2c1ea DW |
8908 | struct disk_info *inf = get_disk_info(u); |
8909 | struct imsm_disk *disk; | |
8273f55e DW |
8910 | int i; |
8911 | int j; | |
8273f55e | 8912 | |
54c2c1ea | 8913 | for (i = 0; i < map->num_members; i++) { |
238c0a71 | 8914 | disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, MAP_X)); |
54c2c1ea DW |
8915 | for (j = 0; j < new_map->num_members; j++) |
8916 | if (serialcmp(disk->serial, inf[j].serial) == 0) | |
8273f55e DW |
8917 | return 1; |
8918 | } | |
8919 | ||
8920 | return 0; | |
8921 | } | |
8922 | ||
1a64be56 LM |
8923 | static struct dl *get_disk_super(struct intel_super *super, int major, int minor) |
8924 | { | |
594dc1b8 JS |
8925 | struct dl *dl; |
8926 | ||
1a64be56 | 8927 | for (dl = super->disks; dl; dl = dl->next) |
089f9d79 | 8928 | if (dl->major == major && dl->minor == minor) |
1a64be56 LM |
8929 | return dl; |
8930 | return NULL; | |
8931 | } | |
8932 | ||
8933 | static int remove_disk_super(struct intel_super *super, int major, int minor) | |
8934 | { | |
594dc1b8 | 8935 | struct dl *prev; |
1a64be56 LM |
8936 | struct dl *dl; |
8937 | ||
8938 | prev = NULL; | |
8939 | for (dl = super->disks; dl; dl = dl->next) { | |
089f9d79 | 8940 | if (dl->major == major && dl->minor == minor) { |
1a64be56 LM |
8941 | /* remove */ |
8942 | if (prev) | |
8943 | prev->next = dl->next; | |
8944 | else | |
8945 | super->disks = dl->next; | |
8946 | dl->next = NULL; | |
8947 | __free_imsm_disk(dl); | |
1ade5cc1 | 8948 | dprintf("removed %x:%x\n", major, minor); |
1a64be56 LM |
8949 | break; |
8950 | } | |
8951 | prev = dl; | |
8952 | } | |
8953 | return 0; | |
8954 | } | |
8955 | ||
f21e18ca | 8956 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index); |
ae6aad82 | 8957 | |
1a64be56 LM |
8958 | static int add_remove_disk_update(struct intel_super *super) |
8959 | { | |
8960 | int check_degraded = 0; | |
594dc1b8 JS |
8961 | struct dl *disk; |
8962 | ||
1a64be56 LM |
8963 | /* add/remove some spares to/from the metadata/contrainer */ |
8964 | while (super->disk_mgmt_list) { | |
8965 | struct dl *disk_cfg; | |
8966 | ||
8967 | disk_cfg = super->disk_mgmt_list; | |
8968 | super->disk_mgmt_list = disk_cfg->next; | |
8969 | disk_cfg->next = NULL; | |
8970 | ||
8971 | if (disk_cfg->action == DISK_ADD) { | |
8972 | disk_cfg->next = super->disks; | |
8973 | super->disks = disk_cfg; | |
8974 | check_degraded = 1; | |
1ade5cc1 N |
8975 | dprintf("added %x:%x\n", |
8976 | disk_cfg->major, disk_cfg->minor); | |
1a64be56 LM |
8977 | } else if (disk_cfg->action == DISK_REMOVE) { |
8978 | dprintf("Disk remove action processed: %x.%x\n", | |
8979 | disk_cfg->major, disk_cfg->minor); | |
8980 | disk = get_disk_super(super, | |
8981 | disk_cfg->major, | |
8982 | disk_cfg->minor); | |
8983 | if (disk) { | |
8984 | /* store action status */ | |
8985 | disk->action = DISK_REMOVE; | |
8986 | /* remove spare disks only */ | |
8987 | if (disk->index == -1) { | |
8988 | remove_disk_super(super, | |
8989 | disk_cfg->major, | |
8990 | disk_cfg->minor); | |
8991 | } | |
8992 | } | |
8993 | /* release allocate disk structure */ | |
8994 | __free_imsm_disk(disk_cfg); | |
8995 | } | |
8996 | } | |
8997 | return check_degraded; | |
8998 | } | |
8999 | ||
a29911da PC |
9000 | static int apply_reshape_migration_update(struct imsm_update_reshape_migration *u, |
9001 | struct intel_super *super, | |
9002 | void ***space_list) | |
9003 | { | |
9004 | struct intel_dev *id; | |
9005 | void **tofree = NULL; | |
9006 | int ret_val = 0; | |
9007 | ||
1ade5cc1 | 9008 | dprintf("(enter)\n"); |
089f9d79 | 9009 | if (u->subdev < 0 || u->subdev > 1) { |
a29911da PC |
9010 | dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev); |
9011 | return ret_val; | |
9012 | } | |
089f9d79 | 9013 | if (space_list == NULL || *space_list == NULL) { |
a29911da PC |
9014 | dprintf("imsm: Error: Memory is not allocated\n"); |
9015 | return ret_val; | |
9016 | } | |
9017 | ||
9018 | for (id = super->devlist ; id; id = id->next) { | |
9019 | if (id->index == (unsigned)u->subdev) { | |
9020 | struct imsm_dev *dev = get_imsm_dev(super, u->subdev); | |
9021 | struct imsm_map *map; | |
9022 | struct imsm_dev *new_dev = | |
9023 | (struct imsm_dev *)*space_list; | |
238c0a71 | 9024 | struct imsm_map *migr_map = get_imsm_map(dev, MAP_1); |
a29911da PC |
9025 | int to_state; |
9026 | struct dl *new_disk; | |
9027 | ||
9028 | if (new_dev == NULL) | |
9029 | return ret_val; | |
9030 | *space_list = **space_list; | |
9031 | memcpy(new_dev, dev, sizeof_imsm_dev(dev, 0)); | |
238c0a71 | 9032 | map = get_imsm_map(new_dev, MAP_0); |
a29911da PC |
9033 | if (migr_map) { |
9034 | dprintf("imsm: Error: migration in progress"); | |
9035 | return ret_val; | |
9036 | } | |
9037 | ||
9038 | to_state = map->map_state; | |
9039 | if ((u->new_level == 5) && (map->raid_level == 0)) { | |
9040 | map->num_members++; | |
9041 | /* this should not happen */ | |
9042 | if (u->new_disks[0] < 0) { | |
9043 | map->failed_disk_num = | |
9044 | map->num_members - 1; | |
9045 | to_state = IMSM_T_STATE_DEGRADED; | |
9046 | } else | |
9047 | to_state = IMSM_T_STATE_NORMAL; | |
9048 | } | |
8e59f3d8 | 9049 | migrate(new_dev, super, to_state, MIGR_GEN_MIGR); |
a29911da PC |
9050 | if (u->new_level > -1) |
9051 | map->raid_level = u->new_level; | |
238c0a71 | 9052 | migr_map = get_imsm_map(new_dev, MAP_1); |
a29911da PC |
9053 | if ((u->new_level == 5) && |
9054 | (migr_map->raid_level == 0)) { | |
9055 | int ord = map->num_members - 1; | |
9056 | migr_map->num_members--; | |
9057 | if (u->new_disks[0] < 0) | |
9058 | ord |= IMSM_ORD_REBUILD; | |
9059 | set_imsm_ord_tbl_ent(map, | |
9060 | map->num_members - 1, | |
9061 | ord); | |
9062 | } | |
9063 | id->dev = new_dev; | |
9064 | tofree = (void **)dev; | |
9065 | ||
4bba0439 PC |
9066 | /* update chunk size |
9067 | */ | |
06fb291a PB |
9068 | if (u->new_chunksize > 0) { |
9069 | unsigned long long num_data_stripes; | |
9070 | int used_disks = | |
9071 | imsm_num_data_members(dev, MAP_0); | |
9072 | ||
9073 | if (used_disks == 0) | |
9074 | return ret_val; | |
9075 | ||
4bba0439 PC |
9076 | map->blocks_per_strip = |
9077 | __cpu_to_le16(u->new_chunksize * 2); | |
06fb291a PB |
9078 | num_data_stripes = |
9079 | (join_u32(dev->size_low, dev->size_high) | |
9080 | / used_disks); | |
9081 | num_data_stripes /= map->blocks_per_strip; | |
9082 | num_data_stripes /= map->num_domains; | |
9083 | set_num_data_stripes(map, num_data_stripes); | |
9084 | } | |
4bba0439 | 9085 | |
a29911da PC |
9086 | /* add disk |
9087 | */ | |
089f9d79 JS |
9088 | if (u->new_level != 5 || migr_map->raid_level != 0 || |
9089 | migr_map->raid_level == map->raid_level) | |
a29911da PC |
9090 | goto skip_disk_add; |
9091 | ||
9092 | if (u->new_disks[0] >= 0) { | |
9093 | /* use passes spare | |
9094 | */ | |
9095 | new_disk = get_disk_super(super, | |
9096 | major(u->new_disks[0]), | |
9097 | minor(u->new_disks[0])); | |
7a862a02 | 9098 | dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n", |
a29911da PC |
9099 | major(u->new_disks[0]), |
9100 | minor(u->new_disks[0]), | |
9101 | new_disk, new_disk->index); | |
9102 | if (new_disk == NULL) | |
9103 | goto error_disk_add; | |
9104 | ||
9105 | new_disk->index = map->num_members - 1; | |
9106 | /* slot to fill in autolayout | |
9107 | */ | |
9108 | new_disk->raiddisk = new_disk->index; | |
9109 | new_disk->disk.status |= CONFIGURED_DISK; | |
9110 | new_disk->disk.status &= ~SPARE_DISK; | |
9111 | } else | |
9112 | goto error_disk_add; | |
9113 | ||
9114 | skip_disk_add: | |
9115 | *tofree = *space_list; | |
9116 | /* calculate new size | |
9117 | */ | |
f3871fdc | 9118 | imsm_set_array_size(new_dev, -1); |
a29911da PC |
9119 | |
9120 | ret_val = 1; | |
9121 | } | |
9122 | } | |
9123 | ||
9124 | if (tofree) | |
9125 | *space_list = tofree; | |
9126 | return ret_val; | |
9127 | ||
9128 | error_disk_add: | |
9129 | dprintf("Error: imsm: Cannot find disk.\n"); | |
9130 | return ret_val; | |
9131 | } | |
9132 | ||
f3871fdc AK |
9133 | static int apply_size_change_update(struct imsm_update_size_change *u, |
9134 | struct intel_super *super) | |
9135 | { | |
9136 | struct intel_dev *id; | |
9137 | int ret_val = 0; | |
9138 | ||
1ade5cc1 | 9139 | dprintf("(enter)\n"); |
089f9d79 | 9140 | if (u->subdev < 0 || u->subdev > 1) { |
f3871fdc AK |
9141 | dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev); |
9142 | return ret_val; | |
9143 | } | |
9144 | ||
9145 | for (id = super->devlist ; id; id = id->next) { | |
9146 | if (id->index == (unsigned)u->subdev) { | |
9147 | struct imsm_dev *dev = get_imsm_dev(super, u->subdev); | |
9148 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
9149 | int used_disks = imsm_num_data_members(dev, MAP_0); | |
9150 | unsigned long long blocks_per_member; | |
06fb291a | 9151 | unsigned long long num_data_stripes; |
f3871fdc AK |
9152 | |
9153 | /* calculate new size | |
9154 | */ | |
9155 | blocks_per_member = u->new_size / used_disks; | |
06fb291a PB |
9156 | num_data_stripes = blocks_per_member / |
9157 | map->blocks_per_strip; | |
9158 | num_data_stripes /= map->num_domains; | |
9159 | dprintf("(size: %llu, blocks per member: %llu, num_data_stipes: %llu)\n", | |
9160 | u->new_size, blocks_per_member, | |
9161 | num_data_stripes); | |
f3871fdc | 9162 | set_blocks_per_member(map, blocks_per_member); |
06fb291a | 9163 | set_num_data_stripes(map, num_data_stripes); |
f3871fdc AK |
9164 | imsm_set_array_size(dev, u->new_size); |
9165 | ||
9166 | ret_val = 1; | |
9167 | break; | |
9168 | } | |
9169 | } | |
9170 | ||
9171 | return ret_val; | |
9172 | } | |
9173 | ||
061d7da3 | 9174 | static int apply_update_activate_spare(struct imsm_update_activate_spare *u, |
ca9de185 | 9175 | struct intel_super *super, |
061d7da3 LO |
9176 | struct active_array *active_array) |
9177 | { | |
9178 | struct imsm_super *mpb = super->anchor; | |
9179 | struct imsm_dev *dev = get_imsm_dev(super, u->array); | |
238c0a71 | 9180 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
061d7da3 LO |
9181 | struct imsm_map *migr_map; |
9182 | struct active_array *a; | |
9183 | struct imsm_disk *disk; | |
9184 | __u8 to_state; | |
9185 | struct dl *dl; | |
9186 | unsigned int found; | |
9187 | int failed; | |
5961eeec | 9188 | int victim; |
061d7da3 | 9189 | int i; |
5961eeec | 9190 | int second_map_created = 0; |
061d7da3 | 9191 | |
5961eeec | 9192 | for (; u; u = u->next) { |
238c0a71 | 9193 | victim = get_imsm_disk_idx(dev, u->slot, MAP_X); |
061d7da3 | 9194 | |
5961eeec | 9195 | if (victim < 0) |
9196 | return 0; | |
061d7da3 | 9197 | |
5961eeec | 9198 | for (dl = super->disks; dl; dl = dl->next) |
9199 | if (dl == u->dl) | |
9200 | break; | |
061d7da3 | 9201 | |
5961eeec | 9202 | if (!dl) { |
7a862a02 | 9203 | pr_err("error: imsm_activate_spare passed an unknown disk (index: %d)\n", |
5961eeec | 9204 | u->dl->index); |
9205 | return 0; | |
9206 | } | |
061d7da3 | 9207 | |
5961eeec | 9208 | /* count failures (excluding rebuilds and the victim) |
9209 | * to determine map[0] state | |
9210 | */ | |
9211 | failed = 0; | |
9212 | for (i = 0; i < map->num_members; i++) { | |
9213 | if (i == u->slot) | |
9214 | continue; | |
9215 | disk = get_imsm_disk(super, | |
238c0a71 | 9216 | get_imsm_disk_idx(dev, i, MAP_X)); |
5961eeec | 9217 | if (!disk || is_failed(disk)) |
9218 | failed++; | |
9219 | } | |
061d7da3 | 9220 | |
5961eeec | 9221 | /* adding a pristine spare, assign a new index */ |
9222 | if (dl->index < 0) { | |
9223 | dl->index = super->anchor->num_disks; | |
9224 | super->anchor->num_disks++; | |
9225 | } | |
9226 | disk = &dl->disk; | |
9227 | disk->status |= CONFIGURED_DISK; | |
9228 | disk->status &= ~SPARE_DISK; | |
9229 | ||
9230 | /* mark rebuild */ | |
238c0a71 | 9231 | to_state = imsm_check_degraded(super, dev, failed, MAP_0); |
5961eeec | 9232 | if (!second_map_created) { |
9233 | second_map_created = 1; | |
9234 | map->map_state = IMSM_T_STATE_DEGRADED; | |
9235 | migrate(dev, super, to_state, MIGR_REBUILD); | |
9236 | } else | |
9237 | map->map_state = to_state; | |
238c0a71 | 9238 | migr_map = get_imsm_map(dev, MAP_1); |
5961eeec | 9239 | set_imsm_ord_tbl_ent(map, u->slot, dl->index); |
9240 | set_imsm_ord_tbl_ent(migr_map, u->slot, | |
9241 | dl->index | IMSM_ORD_REBUILD); | |
9242 | ||
9243 | /* update the family_num to mark a new container | |
9244 | * generation, being careful to record the existing | |
9245 | * family_num in orig_family_num to clean up after | |
9246 | * earlier mdadm versions that neglected to set it. | |
9247 | */ | |
9248 | if (mpb->orig_family_num == 0) | |
9249 | mpb->orig_family_num = mpb->family_num; | |
9250 | mpb->family_num += super->random; | |
9251 | ||
9252 | /* count arrays using the victim in the metadata */ | |
9253 | found = 0; | |
9254 | for (a = active_array; a ; a = a->next) { | |
9255 | dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 9256 | map = get_imsm_map(dev, MAP_0); |
061d7da3 | 9257 | |
5961eeec | 9258 | if (get_imsm_disk_slot(map, victim) >= 0) |
9259 | found++; | |
9260 | } | |
061d7da3 | 9261 | |
5961eeec | 9262 | /* delete the victim if it is no longer being |
9263 | * utilized anywhere | |
061d7da3 | 9264 | */ |
5961eeec | 9265 | if (!found) { |
9266 | struct dl **dlp; | |
061d7da3 | 9267 | |
5961eeec | 9268 | /* We know that 'manager' isn't touching anything, |
9269 | * so it is safe to delete | |
9270 | */ | |
9271 | for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next) | |
061d7da3 LO |
9272 | if ((*dlp)->index == victim) |
9273 | break; | |
5961eeec | 9274 | |
9275 | /* victim may be on the missing list */ | |
9276 | if (!*dlp) | |
9277 | for (dlp = &super->missing; *dlp; | |
9278 | dlp = &(*dlp)->next) | |
9279 | if ((*dlp)->index == victim) | |
9280 | break; | |
9281 | imsm_delete(super, dlp, victim); | |
9282 | } | |
061d7da3 LO |
9283 | } |
9284 | ||
9285 | return 1; | |
9286 | } | |
a29911da | 9287 | |
2e5dc010 N |
9288 | static int apply_reshape_container_disks_update(struct imsm_update_reshape *u, |
9289 | struct intel_super *super, | |
9290 | void ***space_list) | |
9291 | { | |
9292 | struct dl *new_disk; | |
9293 | struct intel_dev *id; | |
9294 | int i; | |
9295 | int delta_disks = u->new_raid_disks - u->old_raid_disks; | |
ee4beede | 9296 | int disk_count = u->old_raid_disks; |
2e5dc010 N |
9297 | void **tofree = NULL; |
9298 | int devices_to_reshape = 1; | |
9299 | struct imsm_super *mpb = super->anchor; | |
9300 | int ret_val = 0; | |
d098291a | 9301 | unsigned int dev_id; |
2e5dc010 | 9302 | |
1ade5cc1 | 9303 | dprintf("(enter)\n"); |
2e5dc010 N |
9304 | |
9305 | /* enable spares to use in array */ | |
9306 | for (i = 0; i < delta_disks; i++) { | |
9307 | new_disk = get_disk_super(super, | |
9308 | major(u->new_disks[i]), | |
9309 | minor(u->new_disks[i])); | |
7a862a02 | 9310 | dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n", |
2e5dc010 N |
9311 | major(u->new_disks[i]), minor(u->new_disks[i]), |
9312 | new_disk, new_disk->index); | |
089f9d79 JS |
9313 | if (new_disk == NULL || |
9314 | (new_disk->index >= 0 && | |
9315 | new_disk->index < u->old_raid_disks)) | |
2e5dc010 | 9316 | goto update_reshape_exit; |
ee4beede | 9317 | new_disk->index = disk_count++; |
2e5dc010 N |
9318 | /* slot to fill in autolayout |
9319 | */ | |
9320 | new_disk->raiddisk = new_disk->index; | |
9321 | new_disk->disk.status |= | |
9322 | CONFIGURED_DISK; | |
9323 | new_disk->disk.status &= ~SPARE_DISK; | |
9324 | } | |
9325 | ||
ed7333bd AK |
9326 | dprintf("imsm: volume set mpb->num_raid_devs = %i\n", |
9327 | mpb->num_raid_devs); | |
2e5dc010 N |
9328 | /* manage changes in volume |
9329 | */ | |
d098291a | 9330 | for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) { |
2e5dc010 N |
9331 | void **sp = *space_list; |
9332 | struct imsm_dev *newdev; | |
9333 | struct imsm_map *newmap, *oldmap; | |
9334 | ||
d098291a AK |
9335 | for (id = super->devlist ; id; id = id->next) { |
9336 | if (id->index == dev_id) | |
9337 | break; | |
9338 | } | |
9339 | if (id == NULL) | |
9340 | break; | |
2e5dc010 N |
9341 | if (!sp) |
9342 | continue; | |
9343 | *space_list = *sp; | |
9344 | newdev = (void*)sp; | |
9345 | /* Copy the dev, but not (all of) the map */ | |
9346 | memcpy(newdev, id->dev, sizeof(*newdev)); | |
238c0a71 AK |
9347 | oldmap = get_imsm_map(id->dev, MAP_0); |
9348 | newmap = get_imsm_map(newdev, MAP_0); | |
2e5dc010 N |
9349 | /* Copy the current map */ |
9350 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); | |
9351 | /* update one device only | |
9352 | */ | |
9353 | if (devices_to_reshape) { | |
ed7333bd AK |
9354 | dprintf("imsm: modifying subdev: %i\n", |
9355 | id->index); | |
2e5dc010 N |
9356 | devices_to_reshape--; |
9357 | newdev->vol.migr_state = 1; | |
9358 | newdev->vol.curr_migr_unit = 0; | |
ea672ee1 | 9359 | set_migr_type(newdev, MIGR_GEN_MIGR); |
2e5dc010 N |
9360 | newmap->num_members = u->new_raid_disks; |
9361 | for (i = 0; i < delta_disks; i++) { | |
9362 | set_imsm_ord_tbl_ent(newmap, | |
9363 | u->old_raid_disks + i, | |
9364 | u->old_raid_disks + i); | |
9365 | } | |
9366 | /* New map is correct, now need to save old map | |
9367 | */ | |
238c0a71 | 9368 | newmap = get_imsm_map(newdev, MAP_1); |
2e5dc010 N |
9369 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); |
9370 | ||
f3871fdc | 9371 | imsm_set_array_size(newdev, -1); |
2e5dc010 N |
9372 | } |
9373 | ||
9374 | sp = (void **)id->dev; | |
9375 | id->dev = newdev; | |
9376 | *sp = tofree; | |
9377 | tofree = sp; | |
8e59f3d8 AK |
9378 | |
9379 | /* Clear migration record */ | |
9380 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
2e5dc010 | 9381 | } |
819bc634 AK |
9382 | if (tofree) |
9383 | *space_list = tofree; | |
2e5dc010 N |
9384 | ret_val = 1; |
9385 | ||
9386 | update_reshape_exit: | |
9387 | ||
9388 | return ret_val; | |
9389 | } | |
9390 | ||
bb025c2f | 9391 | static int apply_takeover_update(struct imsm_update_takeover *u, |
8ca6df95 KW |
9392 | struct intel_super *super, |
9393 | void ***space_list) | |
bb025c2f KW |
9394 | { |
9395 | struct imsm_dev *dev = NULL; | |
8ca6df95 KW |
9396 | struct intel_dev *dv; |
9397 | struct imsm_dev *dev_new; | |
bb025c2f KW |
9398 | struct imsm_map *map; |
9399 | struct dl *dm, *du; | |
8ca6df95 | 9400 | int i; |
bb025c2f KW |
9401 | |
9402 | for (dv = super->devlist; dv; dv = dv->next) | |
9403 | if (dv->index == (unsigned int)u->subarray) { | |
9404 | dev = dv->dev; | |
9405 | break; | |
9406 | } | |
9407 | ||
9408 | if (dev == NULL) | |
9409 | return 0; | |
9410 | ||
238c0a71 | 9411 | map = get_imsm_map(dev, MAP_0); |
bb025c2f KW |
9412 | |
9413 | if (u->direction == R10_TO_R0) { | |
06fb291a PB |
9414 | unsigned long long num_data_stripes; |
9415 | ||
9416 | map->num_domains = 1; | |
9417 | num_data_stripes = blocks_per_member(map); | |
9418 | num_data_stripes /= map->blocks_per_strip; | |
9419 | num_data_stripes /= map->num_domains; | |
9420 | set_num_data_stripes(map, num_data_stripes); | |
9421 | ||
43d5ec18 | 9422 | /* Number of failed disks must be half of initial disk number */ |
3b451610 AK |
9423 | if (imsm_count_failed(super, dev, MAP_0) != |
9424 | (map->num_members / 2)) | |
43d5ec18 KW |
9425 | return 0; |
9426 | ||
bb025c2f KW |
9427 | /* iterate through devices to mark removed disks as spare */ |
9428 | for (dm = super->disks; dm; dm = dm->next) { | |
9429 | if (dm->disk.status & FAILED_DISK) { | |
9430 | int idx = dm->index; | |
9431 | /* update indexes on the disk list */ | |
9432 | /* FIXME this loop-with-the-loop looks wrong, I'm not convinced | |
9433 | the index values will end up being correct.... NB */ | |
9434 | for (du = super->disks; du; du = du->next) | |
9435 | if (du->index > idx) | |
9436 | du->index--; | |
9437 | /* mark as spare disk */ | |
a8619d23 | 9438 | mark_spare(dm); |
bb025c2f KW |
9439 | } |
9440 | } | |
bb025c2f KW |
9441 | /* update map */ |
9442 | map->num_members = map->num_members / 2; | |
9443 | map->map_state = IMSM_T_STATE_NORMAL; | |
9444 | map->num_domains = 1; | |
9445 | map->raid_level = 0; | |
9446 | map->failed_disk_num = -1; | |
9447 | } | |
9448 | ||
8ca6df95 KW |
9449 | if (u->direction == R0_TO_R10) { |
9450 | void **space; | |
9451 | /* update slots in current disk list */ | |
9452 | for (dm = super->disks; dm; dm = dm->next) { | |
9453 | if (dm->index >= 0) | |
9454 | dm->index *= 2; | |
9455 | } | |
9456 | /* create new *missing* disks */ | |
9457 | for (i = 0; i < map->num_members; i++) { | |
9458 | space = *space_list; | |
9459 | if (!space) | |
9460 | continue; | |
9461 | *space_list = *space; | |
9462 | du = (void *)space; | |
9463 | memcpy(du, super->disks, sizeof(*du)); | |
8ca6df95 KW |
9464 | du->fd = -1; |
9465 | du->minor = 0; | |
9466 | du->major = 0; | |
9467 | du->index = (i * 2) + 1; | |
9468 | sprintf((char *)du->disk.serial, | |
9469 | " MISSING_%d", du->index); | |
9470 | sprintf((char *)du->serial, | |
9471 | "MISSING_%d", du->index); | |
9472 | du->next = super->missing; | |
9473 | super->missing = du; | |
9474 | } | |
9475 | /* create new dev and map */ | |
9476 | space = *space_list; | |
9477 | if (!space) | |
9478 | return 0; | |
9479 | *space_list = *space; | |
9480 | dev_new = (void *)space; | |
9481 | memcpy(dev_new, dev, sizeof(*dev)); | |
9482 | /* update new map */ | |
238c0a71 | 9483 | map = get_imsm_map(dev_new, MAP_0); |
8ca6df95 | 9484 | map->num_members = map->num_members * 2; |
1a2487c2 | 9485 | map->map_state = IMSM_T_STATE_DEGRADED; |
8ca6df95 KW |
9486 | map->num_domains = 2; |
9487 | map->raid_level = 1; | |
9488 | /* replace dev<->dev_new */ | |
9489 | dv->dev = dev_new; | |
9490 | } | |
bb025c2f KW |
9491 | /* update disk order table */ |
9492 | for (du = super->disks; du; du = du->next) | |
9493 | if (du->index >= 0) | |
9494 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
8ca6df95 | 9495 | for (du = super->missing; du; du = du->next) |
1a2487c2 KW |
9496 | if (du->index >= 0) { |
9497 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
4c9e8c1e | 9498 | mark_missing(super, dv->dev, &du->disk, du->index); |
1a2487c2 | 9499 | } |
bb025c2f KW |
9500 | |
9501 | return 1; | |
9502 | } | |
9503 | ||
e8319a19 DW |
9504 | static void imsm_process_update(struct supertype *st, |
9505 | struct metadata_update *update) | |
9506 | { | |
9507 | /** | |
9508 | * crack open the metadata_update envelope to find the update record | |
9509 | * update can be one of: | |
d195167d AK |
9510 | * update_reshape_container_disks - all the arrays in the container |
9511 | * are being reshaped to have more devices. We need to mark | |
9512 | * the arrays for general migration and convert selected spares | |
9513 | * into active devices. | |
9514 | * update_activate_spare - a spare device has replaced a failed | |
1011e834 N |
9515 | * device in an array, update the disk_ord_tbl. If this disk is |
9516 | * present in all member arrays then also clear the SPARE_DISK | |
9517 | * flag | |
d195167d AK |
9518 | * update_create_array |
9519 | * update_kill_array | |
9520 | * update_rename_array | |
9521 | * update_add_remove_disk | |
e8319a19 DW |
9522 | */ |
9523 | struct intel_super *super = st->sb; | |
4d7b1503 | 9524 | struct imsm_super *mpb; |
e8319a19 DW |
9525 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
9526 | ||
4d7b1503 DW |
9527 | /* update requires a larger buf but the allocation failed */ |
9528 | if (super->next_len && !super->next_buf) { | |
9529 | super->next_len = 0; | |
9530 | return; | |
9531 | } | |
9532 | ||
9533 | if (super->next_buf) { | |
9534 | memcpy(super->next_buf, super->buf, super->len); | |
9535 | free(super->buf); | |
9536 | super->len = super->next_len; | |
9537 | super->buf = super->next_buf; | |
9538 | ||
9539 | super->next_len = 0; | |
9540 | super->next_buf = NULL; | |
9541 | } | |
9542 | ||
9543 | mpb = super->anchor; | |
9544 | ||
e8319a19 | 9545 | switch (type) { |
0ec5d470 AK |
9546 | case update_general_migration_checkpoint: { |
9547 | struct intel_dev *id; | |
9548 | struct imsm_update_general_migration_checkpoint *u = | |
9549 | (void *)update->buf; | |
9550 | ||
1ade5cc1 | 9551 | dprintf("called for update_general_migration_checkpoint\n"); |
0ec5d470 AK |
9552 | |
9553 | /* find device under general migration */ | |
9554 | for (id = super->devlist ; id; id = id->next) { | |
9555 | if (is_gen_migration(id->dev)) { | |
9556 | id->dev->vol.curr_migr_unit = | |
9557 | __cpu_to_le32(u->curr_migr_unit); | |
9558 | super->updates_pending++; | |
9559 | } | |
9560 | } | |
9561 | break; | |
9562 | } | |
bb025c2f KW |
9563 | case update_takeover: { |
9564 | struct imsm_update_takeover *u = (void *)update->buf; | |
1a2487c2 KW |
9565 | if (apply_takeover_update(u, super, &update->space_list)) { |
9566 | imsm_update_version_info(super); | |
bb025c2f | 9567 | super->updates_pending++; |
1a2487c2 | 9568 | } |
bb025c2f KW |
9569 | break; |
9570 | } | |
9571 | ||
78b10e66 | 9572 | case update_reshape_container_disks: { |
d195167d | 9573 | struct imsm_update_reshape *u = (void *)update->buf; |
2e5dc010 N |
9574 | if (apply_reshape_container_disks_update( |
9575 | u, super, &update->space_list)) | |
9576 | super->updates_pending++; | |
78b10e66 N |
9577 | break; |
9578 | } | |
48c5303a | 9579 | case update_reshape_migration: { |
a29911da PC |
9580 | struct imsm_update_reshape_migration *u = (void *)update->buf; |
9581 | if (apply_reshape_migration_update( | |
9582 | u, super, &update->space_list)) | |
9583 | super->updates_pending++; | |
48c5303a PC |
9584 | break; |
9585 | } | |
f3871fdc AK |
9586 | case update_size_change: { |
9587 | struct imsm_update_size_change *u = (void *)update->buf; | |
9588 | if (apply_size_change_update(u, super)) | |
9589 | super->updates_pending++; | |
9590 | break; | |
9591 | } | |
e8319a19 | 9592 | case update_activate_spare: { |
1011e834 | 9593 | struct imsm_update_activate_spare *u = (void *) update->buf; |
061d7da3 LO |
9594 | if (apply_update_activate_spare(u, super, st->arrays)) |
9595 | super->updates_pending++; | |
8273f55e DW |
9596 | break; |
9597 | } | |
9598 | case update_create_array: { | |
9599 | /* someone wants to create a new array, we need to be aware of | |
9600 | * a few races/collisions: | |
9601 | * 1/ 'Create' called by two separate instances of mdadm | |
9602 | * 2/ 'Create' versus 'activate_spare': mdadm has chosen | |
9603 | * devices that have since been assimilated via | |
9604 | * activate_spare. | |
9605 | * In the event this update can not be carried out mdadm will | |
9606 | * (FIX ME) notice that its update did not take hold. | |
9607 | */ | |
9608 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 9609 | struct intel_dev *dv; |
8273f55e DW |
9610 | struct imsm_dev *dev; |
9611 | struct imsm_map *map, *new_map; | |
9612 | unsigned long long start, end; | |
9613 | unsigned long long new_start, new_end; | |
9614 | int i; | |
54c2c1ea DW |
9615 | struct disk_info *inf; |
9616 | struct dl *dl; | |
8273f55e DW |
9617 | |
9618 | /* handle racing creates: first come first serve */ | |
9619 | if (u->dev_idx < mpb->num_raid_devs) { | |
1ade5cc1 | 9620 | dprintf("subarray %d already defined\n", u->dev_idx); |
ba2de7ba | 9621 | goto create_error; |
8273f55e DW |
9622 | } |
9623 | ||
9624 | /* check update is next in sequence */ | |
9625 | if (u->dev_idx != mpb->num_raid_devs) { | |
1ade5cc1 N |
9626 | dprintf("can not create array %d expected index %d\n", |
9627 | u->dev_idx, mpb->num_raid_devs); | |
ba2de7ba | 9628 | goto create_error; |
8273f55e DW |
9629 | } |
9630 | ||
238c0a71 | 9631 | new_map = get_imsm_map(&u->dev, MAP_0); |
5551b113 CA |
9632 | new_start = pba_of_lba0(new_map); |
9633 | new_end = new_start + blocks_per_member(new_map); | |
54c2c1ea | 9634 | inf = get_disk_info(u); |
8273f55e DW |
9635 | |
9636 | /* handle activate_spare versus create race: | |
9637 | * check to make sure that overlapping arrays do not include | |
9638 | * overalpping disks | |
9639 | */ | |
9640 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
949c47a0 | 9641 | dev = get_imsm_dev(super, i); |
238c0a71 | 9642 | map = get_imsm_map(dev, MAP_0); |
5551b113 CA |
9643 | start = pba_of_lba0(map); |
9644 | end = start + blocks_per_member(map); | |
8273f55e DW |
9645 | if ((new_start >= start && new_start <= end) || |
9646 | (start >= new_start && start <= new_end)) | |
54c2c1ea DW |
9647 | /* overlap */; |
9648 | else | |
9649 | continue; | |
9650 | ||
9651 | if (disks_overlap(super, i, u)) { | |
1ade5cc1 | 9652 | dprintf("arrays overlap\n"); |
ba2de7ba | 9653 | goto create_error; |
8273f55e DW |
9654 | } |
9655 | } | |
8273f55e | 9656 | |
949c47a0 DW |
9657 | /* check that prepare update was successful */ |
9658 | if (!update->space) { | |
1ade5cc1 | 9659 | dprintf("prepare update failed\n"); |
ba2de7ba | 9660 | goto create_error; |
949c47a0 DW |
9661 | } |
9662 | ||
54c2c1ea DW |
9663 | /* check that all disks are still active before committing |
9664 | * changes. FIXME: could we instead handle this by creating a | |
9665 | * degraded array? That's probably not what the user expects, | |
9666 | * so better to drop this update on the floor. | |
9667 | */ | |
9668 | for (i = 0; i < new_map->num_members; i++) { | |
9669 | dl = serial_to_dl(inf[i].serial, super); | |
9670 | if (!dl) { | |
1ade5cc1 | 9671 | dprintf("disk disappeared\n"); |
ba2de7ba | 9672 | goto create_error; |
54c2c1ea | 9673 | } |
949c47a0 DW |
9674 | } |
9675 | ||
8273f55e | 9676 | super->updates_pending++; |
54c2c1ea DW |
9677 | |
9678 | /* convert spares to members and fixup ord_tbl */ | |
9679 | for (i = 0; i < new_map->num_members; i++) { | |
9680 | dl = serial_to_dl(inf[i].serial, super); | |
9681 | if (dl->index == -1) { | |
9682 | dl->index = mpb->num_disks; | |
9683 | mpb->num_disks++; | |
9684 | dl->disk.status |= CONFIGURED_DISK; | |
9685 | dl->disk.status &= ~SPARE_DISK; | |
9686 | } | |
9687 | set_imsm_ord_tbl_ent(new_map, i, dl->index); | |
9688 | } | |
9689 | ||
ba2de7ba DW |
9690 | dv = update->space; |
9691 | dev = dv->dev; | |
949c47a0 DW |
9692 | update->space = NULL; |
9693 | imsm_copy_dev(dev, &u->dev); | |
ba2de7ba DW |
9694 | dv->index = u->dev_idx; |
9695 | dv->next = super->devlist; | |
9696 | super->devlist = dv; | |
8273f55e | 9697 | mpb->num_raid_devs++; |
8273f55e | 9698 | |
4d1313e9 | 9699 | imsm_update_version_info(super); |
8273f55e | 9700 | break; |
ba2de7ba DW |
9701 | create_error: |
9702 | /* mdmon knows how to release update->space, but not | |
9703 | * ((struct intel_dev *) update->space)->dev | |
9704 | */ | |
9705 | if (update->space) { | |
9706 | dv = update->space; | |
9707 | free(dv->dev); | |
9708 | } | |
8273f55e | 9709 | break; |
e8319a19 | 9710 | } |
33414a01 DW |
9711 | case update_kill_array: { |
9712 | struct imsm_update_kill_array *u = (void *) update->buf; | |
9713 | int victim = u->dev_idx; | |
9714 | struct active_array *a; | |
9715 | struct intel_dev **dp; | |
9716 | struct imsm_dev *dev; | |
9717 | ||
9718 | /* sanity check that we are not affecting the uuid of | |
9719 | * active arrays, or deleting an active array | |
9720 | * | |
9721 | * FIXME when immutable ids are available, but note that | |
9722 | * we'll also need to fixup the invalidated/active | |
9723 | * subarray indexes in mdstat | |
9724 | */ | |
9725 | for (a = st->arrays; a; a = a->next) | |
9726 | if (a->info.container_member >= victim) | |
9727 | break; | |
9728 | /* by definition if mdmon is running at least one array | |
9729 | * is active in the container, so checking | |
9730 | * mpb->num_raid_devs is just extra paranoia | |
9731 | */ | |
9732 | dev = get_imsm_dev(super, victim); | |
9733 | if (a || !dev || mpb->num_raid_devs == 1) { | |
9734 | dprintf("failed to delete subarray-%d\n", victim); | |
9735 | break; | |
9736 | } | |
9737 | ||
9738 | for (dp = &super->devlist; *dp;) | |
f21e18ca | 9739 | if ((*dp)->index == (unsigned)super->current_vol) { |
33414a01 DW |
9740 | *dp = (*dp)->next; |
9741 | } else { | |
f21e18ca | 9742 | if ((*dp)->index > (unsigned)victim) |
33414a01 DW |
9743 | (*dp)->index--; |
9744 | dp = &(*dp)->next; | |
9745 | } | |
9746 | mpb->num_raid_devs--; | |
9747 | super->updates_pending++; | |
9748 | break; | |
9749 | } | |
aa534678 DW |
9750 | case update_rename_array: { |
9751 | struct imsm_update_rename_array *u = (void *) update->buf; | |
9752 | char name[MAX_RAID_SERIAL_LEN+1]; | |
9753 | int target = u->dev_idx; | |
9754 | struct active_array *a; | |
9755 | struct imsm_dev *dev; | |
9756 | ||
9757 | /* sanity check that we are not affecting the uuid of | |
9758 | * an active array | |
9759 | */ | |
9760 | snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name); | |
9761 | name[MAX_RAID_SERIAL_LEN] = '\0'; | |
9762 | for (a = st->arrays; a; a = a->next) | |
9763 | if (a->info.container_member == target) | |
9764 | break; | |
9765 | dev = get_imsm_dev(super, u->dev_idx); | |
9766 | if (a || !dev || !check_name(super, name, 1)) { | |
9767 | dprintf("failed to rename subarray-%d\n", target); | |
9768 | break; | |
9769 | } | |
9770 | ||
cdbe98cd | 9771 | snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name); |
aa534678 DW |
9772 | super->updates_pending++; |
9773 | break; | |
9774 | } | |
1a64be56 | 9775 | case update_add_remove_disk: { |
43dad3d6 | 9776 | /* we may be able to repair some arrays if disks are |
095b8088 | 9777 | * being added, check the status of add_remove_disk |
1a64be56 LM |
9778 | * if discs has been added. |
9779 | */ | |
9780 | if (add_remove_disk_update(super)) { | |
43dad3d6 | 9781 | struct active_array *a; |
072b727f DW |
9782 | |
9783 | super->updates_pending++; | |
1a64be56 | 9784 | for (a = st->arrays; a; a = a->next) |
43dad3d6 DW |
9785 | a->check_degraded = 1; |
9786 | } | |
43dad3d6 | 9787 | break; |
e8319a19 | 9788 | } |
bbab0940 TM |
9789 | case update_prealloc_badblocks_mem: |
9790 | break; | |
e6e9dd3f AP |
9791 | case update_rwh_policy: { |
9792 | struct imsm_update_rwh_policy *u = (void *)update->buf; | |
9793 | int target = u->dev_idx; | |
9794 | struct imsm_dev *dev = get_imsm_dev(super, target); | |
9795 | if (!dev) { | |
9796 | dprintf("could not find subarray-%d\n", target); | |
9797 | break; | |
9798 | } | |
9799 | ||
9800 | if (dev->rwh_policy != u->new_policy) { | |
9801 | dev->rwh_policy = u->new_policy; | |
9802 | super->updates_pending++; | |
9803 | } | |
9804 | break; | |
9805 | } | |
1a64be56 | 9806 | default: |
7a862a02 | 9807 | pr_err("error: unsuported process update type:(type: %d)\n", type); |
1a64be56 | 9808 | } |
e8319a19 | 9809 | } |
88758e9d | 9810 | |
bc0b9d34 PC |
9811 | static struct mdinfo *get_spares_for_grow(struct supertype *st); |
9812 | ||
5fe6f031 N |
9813 | static int imsm_prepare_update(struct supertype *st, |
9814 | struct metadata_update *update) | |
8273f55e | 9815 | { |
949c47a0 | 9816 | /** |
4d7b1503 DW |
9817 | * Allocate space to hold new disk entries, raid-device entries or a new |
9818 | * mpb if necessary. The manager synchronously waits for updates to | |
9819 | * complete in the monitor, so new mpb buffers allocated here can be | |
9820 | * integrated by the monitor thread without worrying about live pointers | |
9821 | * in the manager thread. | |
8273f55e | 9822 | */ |
095b8088 | 9823 | enum imsm_update_type type; |
4d7b1503 | 9824 | struct intel_super *super = st->sb; |
f36a9ecd | 9825 | unsigned int sector_size = super->sector_size; |
4d7b1503 DW |
9826 | struct imsm_super *mpb = super->anchor; |
9827 | size_t buf_len; | |
9828 | size_t len = 0; | |
949c47a0 | 9829 | |
095b8088 N |
9830 | if (update->len < (int)sizeof(type)) |
9831 | return 0; | |
9832 | ||
9833 | type = *(enum imsm_update_type *) update->buf; | |
9834 | ||
949c47a0 | 9835 | switch (type) { |
0ec5d470 | 9836 | case update_general_migration_checkpoint: |
095b8088 N |
9837 | if (update->len < (int)sizeof(struct imsm_update_general_migration_checkpoint)) |
9838 | return 0; | |
1ade5cc1 | 9839 | dprintf("called for update_general_migration_checkpoint\n"); |
0ec5d470 | 9840 | break; |
abedf5fc KW |
9841 | case update_takeover: { |
9842 | struct imsm_update_takeover *u = (void *)update->buf; | |
095b8088 N |
9843 | if (update->len < (int)sizeof(*u)) |
9844 | return 0; | |
abedf5fc KW |
9845 | if (u->direction == R0_TO_R10) { |
9846 | void **tail = (void **)&update->space_list; | |
9847 | struct imsm_dev *dev = get_imsm_dev(super, u->subarray); | |
238c0a71 | 9848 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
abedf5fc KW |
9849 | int num_members = map->num_members; |
9850 | void *space; | |
9851 | int size, i; | |
abedf5fc KW |
9852 | /* allocate memory for added disks */ |
9853 | for (i = 0; i < num_members; i++) { | |
9854 | size = sizeof(struct dl); | |
503975b9 | 9855 | space = xmalloc(size); |
abedf5fc KW |
9856 | *tail = space; |
9857 | tail = space; | |
9858 | *tail = NULL; | |
9859 | } | |
9860 | /* allocate memory for new device */ | |
9861 | size = sizeof_imsm_dev(super->devlist->dev, 0) + | |
9862 | (num_members * sizeof(__u32)); | |
503975b9 N |
9863 | space = xmalloc(size); |
9864 | *tail = space; | |
9865 | tail = space; | |
9866 | *tail = NULL; | |
9867 | len = disks_to_mpb_size(num_members * 2); | |
abedf5fc KW |
9868 | } |
9869 | ||
9870 | break; | |
9871 | } | |
78b10e66 | 9872 | case update_reshape_container_disks: { |
d195167d AK |
9873 | /* Every raid device in the container is about to |
9874 | * gain some more devices, and we will enter a | |
9875 | * reconfiguration. | |
9876 | * So each 'imsm_map' will be bigger, and the imsm_vol | |
9877 | * will now hold 2 of them. | |
9878 | * Thus we need new 'struct imsm_dev' allocations sized | |
9879 | * as sizeof_imsm_dev but with more devices in both maps. | |
9880 | */ | |
9881 | struct imsm_update_reshape *u = (void *)update->buf; | |
9882 | struct intel_dev *dl; | |
9883 | void **space_tail = (void**)&update->space_list; | |
9884 | ||
095b8088 N |
9885 | if (update->len < (int)sizeof(*u)) |
9886 | return 0; | |
9887 | ||
1ade5cc1 | 9888 | dprintf("for update_reshape\n"); |
d195167d AK |
9889 | |
9890 | for (dl = super->devlist; dl; dl = dl->next) { | |
9891 | int size = sizeof_imsm_dev(dl->dev, 1); | |
9892 | void *s; | |
d677e0b8 AK |
9893 | if (u->new_raid_disks > u->old_raid_disks) |
9894 | size += sizeof(__u32)*2* | |
9895 | (u->new_raid_disks - u->old_raid_disks); | |
503975b9 | 9896 | s = xmalloc(size); |
d195167d AK |
9897 | *space_tail = s; |
9898 | space_tail = s; | |
9899 | *space_tail = NULL; | |
9900 | } | |
9901 | ||
9902 | len = disks_to_mpb_size(u->new_raid_disks); | |
9903 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
78b10e66 N |
9904 | break; |
9905 | } | |
48c5303a | 9906 | case update_reshape_migration: { |
bc0b9d34 PC |
9907 | /* for migration level 0->5 we need to add disks |
9908 | * so the same as for container operation we will copy | |
9909 | * device to the bigger location. | |
9910 | * in memory prepared device and new disk area are prepared | |
9911 | * for usage in process update | |
9912 | */ | |
9913 | struct imsm_update_reshape_migration *u = (void *)update->buf; | |
9914 | struct intel_dev *id; | |
9915 | void **space_tail = (void **)&update->space_list; | |
9916 | int size; | |
9917 | void *s; | |
9918 | int current_level = -1; | |
9919 | ||
095b8088 N |
9920 | if (update->len < (int)sizeof(*u)) |
9921 | return 0; | |
9922 | ||
1ade5cc1 | 9923 | dprintf("for update_reshape\n"); |
bc0b9d34 PC |
9924 | |
9925 | /* add space for bigger array in update | |
9926 | */ | |
9927 | for (id = super->devlist; id; id = id->next) { | |
9928 | if (id->index == (unsigned)u->subdev) { | |
9929 | size = sizeof_imsm_dev(id->dev, 1); | |
9930 | if (u->new_raid_disks > u->old_raid_disks) | |
9931 | size += sizeof(__u32)*2* | |
9932 | (u->new_raid_disks - u->old_raid_disks); | |
503975b9 | 9933 | s = xmalloc(size); |
bc0b9d34 PC |
9934 | *space_tail = s; |
9935 | space_tail = s; | |
9936 | *space_tail = NULL; | |
9937 | break; | |
9938 | } | |
9939 | } | |
9940 | if (update->space_list == NULL) | |
9941 | break; | |
9942 | ||
9943 | /* add space for disk in update | |
9944 | */ | |
9945 | size = sizeof(struct dl); | |
503975b9 | 9946 | s = xmalloc(size); |
bc0b9d34 PC |
9947 | *space_tail = s; |
9948 | space_tail = s; | |
9949 | *space_tail = NULL; | |
9950 | ||
9951 | /* add spare device to update | |
9952 | */ | |
9953 | for (id = super->devlist ; id; id = id->next) | |
9954 | if (id->index == (unsigned)u->subdev) { | |
9955 | struct imsm_dev *dev; | |
9956 | struct imsm_map *map; | |
9957 | ||
9958 | dev = get_imsm_dev(super, u->subdev); | |
238c0a71 | 9959 | map = get_imsm_map(dev, MAP_0); |
bc0b9d34 PC |
9960 | current_level = map->raid_level; |
9961 | break; | |
9962 | } | |
089f9d79 | 9963 | if (u->new_level == 5 && u->new_level != current_level) { |
bc0b9d34 PC |
9964 | struct mdinfo *spares; |
9965 | ||
9966 | spares = get_spares_for_grow(st); | |
9967 | if (spares) { | |
9968 | struct dl *dl; | |
9969 | struct mdinfo *dev; | |
9970 | ||
9971 | dev = spares->devs; | |
9972 | if (dev) { | |
9973 | u->new_disks[0] = | |
9974 | makedev(dev->disk.major, | |
9975 | dev->disk.minor); | |
9976 | dl = get_disk_super(super, | |
9977 | dev->disk.major, | |
9978 | dev->disk.minor); | |
9979 | dl->index = u->old_raid_disks; | |
9980 | dev = dev->next; | |
9981 | } | |
9982 | sysfs_free(spares); | |
9983 | } | |
9984 | } | |
9985 | len = disks_to_mpb_size(u->new_raid_disks); | |
9986 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
48c5303a PC |
9987 | break; |
9988 | } | |
f3871fdc | 9989 | case update_size_change: { |
095b8088 N |
9990 | if (update->len < (int)sizeof(struct imsm_update_size_change)) |
9991 | return 0; | |
9992 | break; | |
9993 | } | |
9994 | case update_activate_spare: { | |
9995 | if (update->len < (int)sizeof(struct imsm_update_activate_spare)) | |
9996 | return 0; | |
f3871fdc AK |
9997 | break; |
9998 | } | |
949c47a0 DW |
9999 | case update_create_array: { |
10000 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 10001 | struct intel_dev *dv; |
54c2c1ea | 10002 | struct imsm_dev *dev = &u->dev; |
238c0a71 | 10003 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
54c2c1ea DW |
10004 | struct dl *dl; |
10005 | struct disk_info *inf; | |
10006 | int i; | |
10007 | int activate = 0; | |
949c47a0 | 10008 | |
095b8088 N |
10009 | if (update->len < (int)sizeof(*u)) |
10010 | return 0; | |
10011 | ||
54c2c1ea DW |
10012 | inf = get_disk_info(u); |
10013 | len = sizeof_imsm_dev(dev, 1); | |
ba2de7ba | 10014 | /* allocate a new super->devlist entry */ |
503975b9 N |
10015 | dv = xmalloc(sizeof(*dv)); |
10016 | dv->dev = xmalloc(len); | |
10017 | update->space = dv; | |
949c47a0 | 10018 | |
54c2c1ea DW |
10019 | /* count how many spares will be converted to members */ |
10020 | for (i = 0; i < map->num_members; i++) { | |
10021 | dl = serial_to_dl(inf[i].serial, super); | |
10022 | if (!dl) { | |
10023 | /* hmm maybe it failed?, nothing we can do about | |
10024 | * it here | |
10025 | */ | |
10026 | continue; | |
10027 | } | |
10028 | if (count_memberships(dl, super) == 0) | |
10029 | activate++; | |
10030 | } | |
10031 | len += activate * sizeof(struct imsm_disk); | |
949c47a0 | 10032 | break; |
095b8088 N |
10033 | } |
10034 | case update_kill_array: { | |
10035 | if (update->len < (int)sizeof(struct imsm_update_kill_array)) | |
10036 | return 0; | |
949c47a0 DW |
10037 | break; |
10038 | } | |
095b8088 N |
10039 | case update_rename_array: { |
10040 | if (update->len < (int)sizeof(struct imsm_update_rename_array)) | |
10041 | return 0; | |
10042 | break; | |
10043 | } | |
10044 | case update_add_remove_disk: | |
10045 | /* no update->len needed */ | |
10046 | break; | |
bbab0940 TM |
10047 | case update_prealloc_badblocks_mem: |
10048 | super->extra_space += sizeof(struct bbm_log) - | |
10049 | get_imsm_bbm_log_size(super->bbm_log); | |
10050 | break; | |
e6e9dd3f AP |
10051 | case update_rwh_policy: { |
10052 | if (update->len < (int)sizeof(struct imsm_update_rwh_policy)) | |
10053 | return 0; | |
10054 | break; | |
10055 | } | |
095b8088 N |
10056 | default: |
10057 | return 0; | |
949c47a0 | 10058 | } |
8273f55e | 10059 | |
4d7b1503 DW |
10060 | /* check if we need a larger metadata buffer */ |
10061 | if (super->next_buf) | |
10062 | buf_len = super->next_len; | |
10063 | else | |
10064 | buf_len = super->len; | |
10065 | ||
bbab0940 | 10066 | if (__le32_to_cpu(mpb->mpb_size) + super->extra_space + len > buf_len) { |
4d7b1503 DW |
10067 | /* ok we need a larger buf than what is currently allocated |
10068 | * if this allocation fails process_update will notice that | |
10069 | * ->next_len is set and ->next_buf is NULL | |
10070 | */ | |
bbab0940 TM |
10071 | buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + |
10072 | super->extra_space + len, sector_size); | |
4d7b1503 DW |
10073 | if (super->next_buf) |
10074 | free(super->next_buf); | |
10075 | ||
10076 | super->next_len = buf_len; | |
f36a9ecd | 10077 | if (posix_memalign(&super->next_buf, sector_size, buf_len) == 0) |
1f45a8ad DW |
10078 | memset(super->next_buf, 0, buf_len); |
10079 | else | |
4d7b1503 DW |
10080 | super->next_buf = NULL; |
10081 | } | |
5fe6f031 | 10082 | return 1; |
8273f55e DW |
10083 | } |
10084 | ||
ae6aad82 | 10085 | /* must be called while manager is quiesced */ |
f21e18ca | 10086 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index) |
ae6aad82 DW |
10087 | { |
10088 | struct imsm_super *mpb = super->anchor; | |
ae6aad82 DW |
10089 | struct dl *iter; |
10090 | struct imsm_dev *dev; | |
10091 | struct imsm_map *map; | |
4c9e8c1e | 10092 | unsigned int i, j, num_members; |
fb12a745 | 10093 | __u32 ord, ord_map0; |
4c9e8c1e | 10094 | struct bbm_log *log = super->bbm_log; |
ae6aad82 | 10095 | |
1ade5cc1 | 10096 | dprintf("deleting device[%d] from imsm_super\n", index); |
ae6aad82 DW |
10097 | |
10098 | /* shift all indexes down one */ | |
10099 | for (iter = super->disks; iter; iter = iter->next) | |
f21e18ca | 10100 | if (iter->index > (int)index) |
ae6aad82 | 10101 | iter->index--; |
47ee5a45 | 10102 | for (iter = super->missing; iter; iter = iter->next) |
f21e18ca | 10103 | if (iter->index > (int)index) |
47ee5a45 | 10104 | iter->index--; |
ae6aad82 DW |
10105 | |
10106 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
10107 | dev = get_imsm_dev(super, i); | |
238c0a71 | 10108 | map = get_imsm_map(dev, MAP_0); |
24565c9a DW |
10109 | num_members = map->num_members; |
10110 | for (j = 0; j < num_members; j++) { | |
10111 | /* update ord entries being careful not to propagate | |
10112 | * ord-flags to the first map | |
10113 | */ | |
238c0a71 | 10114 | ord = get_imsm_ord_tbl_ent(dev, j, MAP_X); |
fb12a745 | 10115 | ord_map0 = get_imsm_ord_tbl_ent(dev, j, MAP_0); |
ae6aad82 | 10116 | |
24565c9a DW |
10117 | if (ord_to_idx(ord) <= index) |
10118 | continue; | |
ae6aad82 | 10119 | |
238c0a71 | 10120 | map = get_imsm_map(dev, MAP_0); |
fb12a745 | 10121 | set_imsm_ord_tbl_ent(map, j, ord_map0 - 1); |
238c0a71 | 10122 | map = get_imsm_map(dev, MAP_1); |
24565c9a DW |
10123 | if (map) |
10124 | set_imsm_ord_tbl_ent(map, j, ord - 1); | |
ae6aad82 DW |
10125 | } |
10126 | } | |
10127 | ||
4c9e8c1e TM |
10128 | for (i = 0; i < log->entry_count; i++) { |
10129 | struct bbm_log_entry *entry = &log->marked_block_entries[i]; | |
10130 | ||
10131 | if (entry->disk_ordinal <= index) | |
10132 | continue; | |
10133 | entry->disk_ordinal--; | |
10134 | } | |
10135 | ||
ae6aad82 DW |
10136 | mpb->num_disks--; |
10137 | super->updates_pending++; | |
24565c9a DW |
10138 | if (*dlp) { |
10139 | struct dl *dl = *dlp; | |
10140 | ||
10141 | *dlp = (*dlp)->next; | |
10142 | __free_imsm_disk(dl); | |
10143 | } | |
ae6aad82 | 10144 | } |
9a717282 AK |
10145 | |
10146 | static void close_targets(int *targets, int new_disks) | |
10147 | { | |
10148 | int i; | |
10149 | ||
10150 | if (!targets) | |
10151 | return; | |
10152 | ||
10153 | for (i = 0; i < new_disks; i++) { | |
10154 | if (targets[i] >= 0) { | |
10155 | close(targets[i]); | |
10156 | targets[i] = -1; | |
10157 | } | |
10158 | } | |
10159 | } | |
10160 | ||
10161 | static int imsm_get_allowed_degradation(int level, int raid_disks, | |
10162 | struct intel_super *super, | |
10163 | struct imsm_dev *dev) | |
10164 | { | |
10165 | switch (level) { | |
bf5cf7c7 | 10166 | case 1: |
9a717282 AK |
10167 | case 10:{ |
10168 | int ret_val = 0; | |
10169 | struct imsm_map *map; | |
10170 | int i; | |
10171 | ||
10172 | ret_val = raid_disks/2; | |
10173 | /* check map if all disks pairs not failed | |
10174 | * in both maps | |
10175 | */ | |
238c0a71 | 10176 | map = get_imsm_map(dev, MAP_0); |
9a717282 AK |
10177 | for (i = 0; i < ret_val; i++) { |
10178 | int degradation = 0; | |
10179 | if (get_imsm_disk(super, i) == NULL) | |
10180 | degradation++; | |
10181 | if (get_imsm_disk(super, i + 1) == NULL) | |
10182 | degradation++; | |
10183 | if (degradation == 2) | |
10184 | return 0; | |
10185 | } | |
238c0a71 | 10186 | map = get_imsm_map(dev, MAP_1); |
9a717282 AK |
10187 | /* if there is no second map |
10188 | * result can be returned | |
10189 | */ | |
10190 | if (map == NULL) | |
10191 | return ret_val; | |
10192 | /* check degradation in second map | |
10193 | */ | |
10194 | for (i = 0; i < ret_val; i++) { | |
10195 | int degradation = 0; | |
10196 | if (get_imsm_disk(super, i) == NULL) | |
10197 | degradation++; | |
10198 | if (get_imsm_disk(super, i + 1) == NULL) | |
10199 | degradation++; | |
10200 | if (degradation == 2) | |
10201 | return 0; | |
10202 | } | |
10203 | return ret_val; | |
10204 | } | |
10205 | case 5: | |
10206 | return 1; | |
10207 | case 6: | |
10208 | return 2; | |
10209 | default: | |
10210 | return 0; | |
10211 | } | |
10212 | } | |
10213 | ||
687629c2 AK |
10214 | /******************************************************************************* |
10215 | * Function: open_backup_targets | |
10216 | * Description: Function opens file descriptors for all devices given in | |
10217 | * info->devs | |
10218 | * Parameters: | |
10219 | * info : general array info | |
10220 | * raid_disks : number of disks | |
10221 | * raid_fds : table of device's file descriptors | |
9a717282 AK |
10222 | * super : intel super for raid10 degradation check |
10223 | * dev : intel device for raid10 degradation check | |
687629c2 AK |
10224 | * Returns: |
10225 | * 0 : success | |
10226 | * -1 : fail | |
10227 | ******************************************************************************/ | |
9a717282 AK |
10228 | int open_backup_targets(struct mdinfo *info, int raid_disks, int *raid_fds, |
10229 | struct intel_super *super, struct imsm_dev *dev) | |
687629c2 AK |
10230 | { |
10231 | struct mdinfo *sd; | |
f627f5ad | 10232 | int i; |
9a717282 | 10233 | int opened = 0; |
f627f5ad AK |
10234 | |
10235 | for (i = 0; i < raid_disks; i++) | |
10236 | raid_fds[i] = -1; | |
687629c2 AK |
10237 | |
10238 | for (sd = info->devs ; sd ; sd = sd->next) { | |
10239 | char *dn; | |
10240 | ||
10241 | if (sd->disk.state & (1<<MD_DISK_FAULTY)) { | |
10242 | dprintf("disk is faulty!!\n"); | |
10243 | continue; | |
10244 | } | |
10245 | ||
089f9d79 | 10246 | if (sd->disk.raid_disk >= raid_disks || sd->disk.raid_disk < 0) |
687629c2 AK |
10247 | continue; |
10248 | ||
10249 | dn = map_dev(sd->disk.major, | |
10250 | sd->disk.minor, 1); | |
10251 | raid_fds[sd->disk.raid_disk] = dev_open(dn, O_RDWR); | |
10252 | if (raid_fds[sd->disk.raid_disk] < 0) { | |
e12b3daa | 10253 | pr_err("cannot open component\n"); |
9a717282 | 10254 | continue; |
687629c2 | 10255 | } |
9a717282 AK |
10256 | opened++; |
10257 | } | |
10258 | /* check if maximum array degradation level is not exceeded | |
10259 | */ | |
10260 | if ((raid_disks - opened) > | |
089f9d79 JS |
10261 | imsm_get_allowed_degradation(info->new_level, raid_disks, |
10262 | super, dev)) { | |
e12b3daa | 10263 | pr_err("Not enough disks can be opened.\n"); |
9a717282 AK |
10264 | close_targets(raid_fds, raid_disks); |
10265 | return -2; | |
687629c2 AK |
10266 | } |
10267 | return 0; | |
10268 | } | |
10269 | ||
d31ad643 PB |
10270 | /******************************************************************************* |
10271 | * Function: validate_container_imsm | |
10272 | * Description: This routine validates container after assemble, | |
10273 | * eg. if devices in container are under the same controller. | |
10274 | * | |
10275 | * Parameters: | |
10276 | * info : linked list with info about devices used in array | |
10277 | * Returns: | |
10278 | * 1 : HBA mismatch | |
10279 | * 0 : Success | |
10280 | ******************************************************************************/ | |
10281 | int validate_container_imsm(struct mdinfo *info) | |
10282 | { | |
6b781d33 AP |
10283 | if (check_env("IMSM_NO_PLATFORM")) |
10284 | return 0; | |
d31ad643 | 10285 | |
6b781d33 AP |
10286 | struct sys_dev *idev; |
10287 | struct sys_dev *hba = NULL; | |
10288 | struct sys_dev *intel_devices = find_intel_devices(); | |
10289 | char *dev_path = devt_to_devpath(makedev(info->disk.major, | |
10290 | info->disk.minor)); | |
10291 | ||
10292 | for (idev = intel_devices; idev; idev = idev->next) { | |
10293 | if (dev_path && strstr(dev_path, idev->path)) { | |
10294 | hba = idev; | |
10295 | break; | |
d31ad643 | 10296 | } |
6b781d33 AP |
10297 | } |
10298 | if (dev_path) | |
d31ad643 PB |
10299 | free(dev_path); |
10300 | ||
6b781d33 AP |
10301 | if (!hba) { |
10302 | pr_err("WARNING - Cannot detect HBA for device %s!\n", | |
10303 | devid2kname(makedev(info->disk.major, info->disk.minor))); | |
10304 | return 1; | |
10305 | } | |
10306 | ||
10307 | const struct imsm_orom *orom = get_orom_by_device_id(hba->dev_id); | |
10308 | struct mdinfo *dev; | |
10309 | ||
10310 | for (dev = info->next; dev; dev = dev->next) { | |
10311 | dev_path = devt_to_devpath(makedev(dev->disk.major, dev->disk.minor)); | |
10312 | ||
10313 | struct sys_dev *hba2 = NULL; | |
10314 | for (idev = intel_devices; idev; idev = idev->next) { | |
10315 | if (dev_path && strstr(dev_path, idev->path)) { | |
10316 | hba2 = idev; | |
10317 | break; | |
d31ad643 PB |
10318 | } |
10319 | } | |
6b781d33 AP |
10320 | if (dev_path) |
10321 | free(dev_path); | |
10322 | ||
10323 | const struct imsm_orom *orom2 = hba2 == NULL ? NULL : | |
10324 | get_orom_by_device_id(hba2->dev_id); | |
10325 | ||
10326 | if (hba2 && hba->type != hba2->type) { | |
10327 | pr_err("WARNING - HBAs of devices do not match %s != %s\n", | |
10328 | get_sys_dev_type(hba->type), get_sys_dev_type(hba2->type)); | |
10329 | return 1; | |
10330 | } | |
10331 | ||
07cb1e57 | 10332 | if (orom != orom2) { |
6b781d33 AP |
10333 | pr_err("WARNING - IMSM container assembled with disks under different HBAs!\n" |
10334 | " This operation is not supported and can lead to data loss.\n"); | |
10335 | return 1; | |
10336 | } | |
10337 | ||
10338 | if (!orom) { | |
10339 | pr_err("WARNING - IMSM container assembled with disks under HBAs without IMSM platform support!\n" | |
10340 | " This operation is not supported and can lead to data loss.\n"); | |
10341 | return 1; | |
10342 | } | |
d31ad643 | 10343 | } |
6b781d33 | 10344 | |
d31ad643 PB |
10345 | return 0; |
10346 | } | |
32141c17 | 10347 | |
6f50473f TM |
10348 | /******************************************************************************* |
10349 | * Function: imsm_record_badblock | |
10350 | * Description: This routine stores new bad block record in BBM log | |
10351 | * | |
10352 | * Parameters: | |
10353 | * a : array containing a bad block | |
10354 | * slot : disk number containing a bad block | |
10355 | * sector : bad block sector | |
10356 | * length : bad block sectors range | |
10357 | * Returns: | |
10358 | * 1 : Success | |
10359 | * 0 : Error | |
10360 | ******************************************************************************/ | |
10361 | static int imsm_record_badblock(struct active_array *a, int slot, | |
10362 | unsigned long long sector, int length) | |
10363 | { | |
10364 | struct intel_super *super = a->container->sb; | |
10365 | int ord; | |
10366 | int ret; | |
10367 | ||
10368 | ord = imsm_disk_slot_to_ord(a, slot); | |
10369 | if (ord < 0) | |
10370 | return 0; | |
10371 | ||
10372 | ret = record_new_badblock(super->bbm_log, ord_to_idx(ord), sector, | |
10373 | length); | |
10374 | if (ret) | |
10375 | super->updates_pending++; | |
10376 | ||
10377 | return ret; | |
10378 | } | |
c07a5a4f TM |
10379 | /******************************************************************************* |
10380 | * Function: imsm_clear_badblock | |
10381 | * Description: This routine clears bad block record from BBM log | |
10382 | * | |
10383 | * Parameters: | |
10384 | * a : array containing a bad block | |
10385 | * slot : disk number containing a bad block | |
10386 | * sector : bad block sector | |
10387 | * length : bad block sectors range | |
10388 | * Returns: | |
10389 | * 1 : Success | |
10390 | * 0 : Error | |
10391 | ******************************************************************************/ | |
10392 | static int imsm_clear_badblock(struct active_array *a, int slot, | |
10393 | unsigned long long sector, int length) | |
10394 | { | |
10395 | struct intel_super *super = a->container->sb; | |
10396 | int ord; | |
10397 | int ret; | |
10398 | ||
10399 | ord = imsm_disk_slot_to_ord(a, slot); | |
10400 | if (ord < 0) | |
10401 | return 0; | |
10402 | ||
10403 | ret = clear_badblock(super->bbm_log, ord_to_idx(ord), sector, length); | |
10404 | if (ret) | |
10405 | super->updates_pending++; | |
10406 | ||
10407 | return ret; | |
10408 | } | |
928f1424 TM |
10409 | /******************************************************************************* |
10410 | * Function: imsm_get_badblocks | |
10411 | * Description: This routine get list of bad blocks for an array | |
10412 | * | |
10413 | * Parameters: | |
10414 | * a : array | |
10415 | * slot : disk number | |
10416 | * Returns: | |
10417 | * bb : structure containing bad blocks | |
10418 | * NULL : error | |
10419 | ******************************************************************************/ | |
10420 | static struct md_bb *imsm_get_badblocks(struct active_array *a, int slot) | |
10421 | { | |
10422 | int inst = a->info.container_member; | |
10423 | struct intel_super *super = a->container->sb; | |
10424 | struct imsm_dev *dev = get_imsm_dev(super, inst); | |
10425 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
10426 | int ord; | |
10427 | ||
10428 | ord = imsm_disk_slot_to_ord(a, slot); | |
10429 | if (ord < 0) | |
10430 | return NULL; | |
10431 | ||
10432 | get_volume_badblocks(super->bbm_log, ord_to_idx(ord), pba_of_lba0(map), | |
10433 | blocks_per_member(map), &super->bb); | |
10434 | ||
10435 | return &super->bb; | |
10436 | } | |
27156a57 TM |
10437 | /******************************************************************************* |
10438 | * Function: examine_badblocks_imsm | |
10439 | * Description: Prints list of bad blocks on a disk to the standard output | |
10440 | * | |
10441 | * Parameters: | |
10442 | * st : metadata handler | |
10443 | * fd : open file descriptor for device | |
10444 | * devname : device name | |
10445 | * Returns: | |
10446 | * 0 : Success | |
10447 | * 1 : Error | |
10448 | ******************************************************************************/ | |
10449 | static int examine_badblocks_imsm(struct supertype *st, int fd, char *devname) | |
10450 | { | |
10451 | struct intel_super *super = st->sb; | |
10452 | struct bbm_log *log = super->bbm_log; | |
10453 | struct dl *d = NULL; | |
10454 | int any = 0; | |
10455 | ||
10456 | for (d = super->disks; d ; d = d->next) { | |
10457 | if (strcmp(d->devname, devname) == 0) | |
10458 | break; | |
10459 | } | |
10460 | ||
10461 | if ((d == NULL) || (d->index < 0)) { /* serial mismatch probably */ | |
10462 | pr_err("%s doesn't appear to be part of a raid array\n", | |
10463 | devname); | |
10464 | return 1; | |
10465 | } | |
10466 | ||
10467 | if (log != NULL) { | |
10468 | unsigned int i; | |
10469 | struct bbm_log_entry *entry = &log->marked_block_entries[0]; | |
10470 | ||
10471 | for (i = 0; i < log->entry_count; i++) { | |
10472 | if (entry[i].disk_ordinal == d->index) { | |
10473 | unsigned long long sector = __le48_to_cpu( | |
10474 | &entry[i].defective_block_start); | |
10475 | int cnt = entry[i].marked_count + 1; | |
10476 | ||
10477 | if (!any) { | |
10478 | printf("Bad-blocks on %s:\n", devname); | |
10479 | any = 1; | |
10480 | } | |
10481 | ||
10482 | printf("%20llu for %d sectors\n", sector, cnt); | |
10483 | } | |
10484 | } | |
10485 | } | |
10486 | ||
10487 | if (!any) | |
10488 | printf("No bad-blocks list configured on %s\n", devname); | |
10489 | ||
10490 | return 0; | |
10491 | } | |
687629c2 AK |
10492 | /******************************************************************************* |
10493 | * Function: init_migr_record_imsm | |
10494 | * Description: Function inits imsm migration record | |
10495 | * Parameters: | |
10496 | * super : imsm internal array info | |
10497 | * dev : device under migration | |
10498 | * info : general array info to find the smallest device | |
10499 | * Returns: | |
10500 | * none | |
10501 | ******************************************************************************/ | |
10502 | void init_migr_record_imsm(struct supertype *st, struct imsm_dev *dev, | |
10503 | struct mdinfo *info) | |
10504 | { | |
10505 | struct intel_super *super = st->sb; | |
10506 | struct migr_record *migr_rec = super->migr_rec; | |
10507 | int new_data_disks; | |
10508 | unsigned long long dsize, dev_sectors; | |
10509 | long long unsigned min_dev_sectors = -1LLU; | |
10510 | struct mdinfo *sd; | |
10511 | char nm[30]; | |
10512 | int fd; | |
238c0a71 AK |
10513 | struct imsm_map *map_dest = get_imsm_map(dev, MAP_0); |
10514 | struct imsm_map *map_src = get_imsm_map(dev, MAP_1); | |
687629c2 | 10515 | unsigned long long num_migr_units; |
3ef4403c | 10516 | unsigned long long array_blocks; |
687629c2 AK |
10517 | |
10518 | memset(migr_rec, 0, sizeof(struct migr_record)); | |
10519 | migr_rec->family_num = __cpu_to_le32(super->anchor->family_num); | |
10520 | ||
10521 | /* only ascending reshape supported now */ | |
10522 | migr_rec->ascending_migr = __cpu_to_le32(1); | |
10523 | ||
10524 | migr_rec->dest_depth_per_unit = GEN_MIGR_AREA_SIZE / | |
10525 | max(map_dest->blocks_per_strip, map_src->blocks_per_strip); | |
e1742195 AK |
10526 | migr_rec->dest_depth_per_unit *= |
10527 | max(map_dest->blocks_per_strip, map_src->blocks_per_strip); | |
238c0a71 | 10528 | new_data_disks = imsm_num_data_members(dev, MAP_0); |
687629c2 AK |
10529 | migr_rec->blocks_per_unit = |
10530 | __cpu_to_le32(migr_rec->dest_depth_per_unit * new_data_disks); | |
10531 | migr_rec->dest_depth_per_unit = | |
10532 | __cpu_to_le32(migr_rec->dest_depth_per_unit); | |
3ef4403c | 10533 | array_blocks = info->component_size * new_data_disks; |
687629c2 AK |
10534 | num_migr_units = |
10535 | array_blocks / __le32_to_cpu(migr_rec->blocks_per_unit); | |
10536 | ||
10537 | if (array_blocks % __le32_to_cpu(migr_rec->blocks_per_unit)) | |
10538 | num_migr_units++; | |
10539 | migr_rec->num_migr_units = __cpu_to_le32(num_migr_units); | |
10540 | ||
10541 | migr_rec->post_migr_vol_cap = dev->size_low; | |
10542 | migr_rec->post_migr_vol_cap_hi = dev->size_high; | |
10543 | ||
687629c2 AK |
10544 | /* Find the smallest dev */ |
10545 | for (sd = info->devs ; sd ; sd = sd->next) { | |
10546 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); | |
10547 | fd = dev_open(nm, O_RDONLY); | |
10548 | if (fd < 0) | |
10549 | continue; | |
10550 | get_dev_size(fd, NULL, &dsize); | |
10551 | dev_sectors = dsize / 512; | |
10552 | if (dev_sectors < min_dev_sectors) | |
10553 | min_dev_sectors = dev_sectors; | |
10554 | close(fd); | |
10555 | } | |
10556 | migr_rec->ckpt_area_pba = __cpu_to_le32(min_dev_sectors - | |
10557 | RAID_DISK_RESERVED_BLOCKS_IMSM_HI); | |
10558 | ||
10559 | write_imsm_migr_rec(st); | |
10560 | ||
10561 | return; | |
10562 | } | |
10563 | ||
10564 | /******************************************************************************* | |
10565 | * Function: save_backup_imsm | |
10566 | * Description: Function saves critical data stripes to Migration Copy Area | |
10567 | * and updates the current migration unit status. | |
10568 | * Use restore_stripes() to form a destination stripe, | |
10569 | * and to write it to the Copy Area. | |
10570 | * Parameters: | |
10571 | * st : supertype information | |
aea93171 | 10572 | * dev : imsm device that backup is saved for |
687629c2 AK |
10573 | * info : general array info |
10574 | * buf : input buffer | |
687629c2 AK |
10575 | * length : length of data to backup (blocks_per_unit) |
10576 | * Returns: | |
10577 | * 0 : success | |
10578 | *, -1 : fail | |
10579 | ******************************************************************************/ | |
10580 | int save_backup_imsm(struct supertype *st, | |
10581 | struct imsm_dev *dev, | |
10582 | struct mdinfo *info, | |
10583 | void *buf, | |
687629c2 AK |
10584 | int length) |
10585 | { | |
10586 | int rv = -1; | |
10587 | struct intel_super *super = st->sb; | |
594dc1b8 JS |
10588 | unsigned long long *target_offsets; |
10589 | int *targets; | |
687629c2 | 10590 | int i; |
238c0a71 | 10591 | struct imsm_map *map_dest = get_imsm_map(dev, MAP_0); |
687629c2 | 10592 | int new_disks = map_dest->num_members; |
ab724b98 AK |
10593 | int dest_layout = 0; |
10594 | int dest_chunk; | |
d1877f69 | 10595 | unsigned long long start; |
238c0a71 | 10596 | int data_disks = imsm_num_data_members(dev, MAP_0); |
687629c2 | 10597 | |
503975b9 | 10598 | targets = xmalloc(new_disks * sizeof(int)); |
687629c2 | 10599 | |
7e45b550 AK |
10600 | for (i = 0; i < new_disks; i++) |
10601 | targets[i] = -1; | |
10602 | ||
503975b9 | 10603 | target_offsets = xcalloc(new_disks, sizeof(unsigned long long)); |
687629c2 | 10604 | |
d1877f69 | 10605 | start = info->reshape_progress * 512; |
687629c2 | 10606 | for (i = 0; i < new_disks; i++) { |
687629c2 AK |
10607 | target_offsets[i] = (unsigned long long) |
10608 | __le32_to_cpu(super->migr_rec->ckpt_area_pba) * 512; | |
d1877f69 AK |
10609 | /* move back copy area adderss, it will be moved forward |
10610 | * in restore_stripes() using start input variable | |
10611 | */ | |
10612 | target_offsets[i] -= start/data_disks; | |
687629c2 AK |
10613 | } |
10614 | ||
9a717282 AK |
10615 | if (open_backup_targets(info, new_disks, targets, |
10616 | super, dev)) | |
687629c2 AK |
10617 | goto abort; |
10618 | ||
68eb8bc6 | 10619 | dest_layout = imsm_level_to_layout(map_dest->raid_level); |
ab724b98 AK |
10620 | dest_chunk = __le16_to_cpu(map_dest->blocks_per_strip) * 512; |
10621 | ||
687629c2 AK |
10622 | if (restore_stripes(targets, /* list of dest devices */ |
10623 | target_offsets, /* migration record offsets */ | |
10624 | new_disks, | |
ab724b98 AK |
10625 | dest_chunk, |
10626 | map_dest->raid_level, | |
10627 | dest_layout, | |
10628 | -1, /* source backup file descriptor */ | |
10629 | 0, /* input buf offset | |
10630 | * always 0 buf is already offseted */ | |
d1877f69 | 10631 | start, |
687629c2 AK |
10632 | length, |
10633 | buf) != 0) { | |
e7b84f9d | 10634 | pr_err("Error restoring stripes\n"); |
687629c2 AK |
10635 | goto abort; |
10636 | } | |
10637 | ||
10638 | rv = 0; | |
10639 | ||
10640 | abort: | |
10641 | if (targets) { | |
9a717282 | 10642 | close_targets(targets, new_disks); |
687629c2 AK |
10643 | free(targets); |
10644 | } | |
10645 | free(target_offsets); | |
10646 | ||
10647 | return rv; | |
10648 | } | |
10649 | ||
10650 | /******************************************************************************* | |
10651 | * Function: save_checkpoint_imsm | |
10652 | * Description: Function called for current unit status update | |
10653 | * in the migration record. It writes it to disk. | |
10654 | * Parameters: | |
10655 | * super : imsm internal array info | |
10656 | * info : general array info | |
10657 | * Returns: | |
10658 | * 0: success | |
10659 | * 1: failure | |
0228d92c AK |
10660 | * 2: failure, means no valid migration record |
10661 | * / no general migration in progress / | |
687629c2 AK |
10662 | ******************************************************************************/ |
10663 | int save_checkpoint_imsm(struct supertype *st, struct mdinfo *info, int state) | |
10664 | { | |
10665 | struct intel_super *super = st->sb; | |
f8b72ef5 AK |
10666 | unsigned long long blocks_per_unit; |
10667 | unsigned long long curr_migr_unit; | |
10668 | ||
2e062e82 | 10669 | if (load_imsm_migr_rec(super, info) != 0) { |
7a862a02 | 10670 | dprintf("imsm: ERROR: Cannot read migration record for checkpoint save.\n"); |
2e062e82 AK |
10671 | return 1; |
10672 | } | |
10673 | ||
f8b72ef5 AK |
10674 | blocks_per_unit = __le32_to_cpu(super->migr_rec->blocks_per_unit); |
10675 | if (blocks_per_unit == 0) { | |
0228d92c AK |
10676 | dprintf("imsm: no migration in progress.\n"); |
10677 | return 2; | |
687629c2 | 10678 | } |
f8b72ef5 AK |
10679 | curr_migr_unit = info->reshape_progress / blocks_per_unit; |
10680 | /* check if array is alligned to copy area | |
10681 | * if it is not alligned, add one to current migration unit value | |
10682 | * this can happend on array reshape finish only | |
10683 | */ | |
10684 | if (info->reshape_progress % blocks_per_unit) | |
10685 | curr_migr_unit++; | |
687629c2 AK |
10686 | |
10687 | super->migr_rec->curr_migr_unit = | |
f8b72ef5 | 10688 | __cpu_to_le32(curr_migr_unit); |
687629c2 AK |
10689 | super->migr_rec->rec_status = __cpu_to_le32(state); |
10690 | super->migr_rec->dest_1st_member_lba = | |
f8b72ef5 AK |
10691 | __cpu_to_le32(curr_migr_unit * |
10692 | __le32_to_cpu(super->migr_rec->dest_depth_per_unit)); | |
687629c2 | 10693 | if (write_imsm_migr_rec(st) < 0) { |
7a862a02 | 10694 | dprintf("imsm: Cannot write migration record outside backup area\n"); |
687629c2 AK |
10695 | return 1; |
10696 | } | |
10697 | ||
10698 | return 0; | |
10699 | } | |
10700 | ||
276d77db AK |
10701 | /******************************************************************************* |
10702 | * Function: recover_backup_imsm | |
10703 | * Description: Function recovers critical data from the Migration Copy Area | |
10704 | * while assembling an array. | |
10705 | * Parameters: | |
10706 | * super : imsm internal array info | |
10707 | * info : general array info | |
10708 | * Returns: | |
10709 | * 0 : success (or there is no data to recover) | |
10710 | * 1 : fail | |
10711 | ******************************************************************************/ | |
10712 | int recover_backup_imsm(struct supertype *st, struct mdinfo *info) | |
10713 | { | |
10714 | struct intel_super *super = st->sb; | |
10715 | struct migr_record *migr_rec = super->migr_rec; | |
594dc1b8 | 10716 | struct imsm_map *map_dest; |
276d77db AK |
10717 | struct intel_dev *id = NULL; |
10718 | unsigned long long read_offset; | |
10719 | unsigned long long write_offset; | |
10720 | unsigned unit_len; | |
10721 | int *targets = NULL; | |
10722 | int new_disks, i, err; | |
10723 | char *buf = NULL; | |
10724 | int retval = 1; | |
f36a9ecd | 10725 | unsigned int sector_size = super->sector_size; |
276d77db AK |
10726 | unsigned long curr_migr_unit = __le32_to_cpu(migr_rec->curr_migr_unit); |
10727 | unsigned long num_migr_units = __le32_to_cpu(migr_rec->num_migr_units); | |
276d77db | 10728 | char buffer[20]; |
6c3560c0 | 10729 | int skipped_disks = 0; |
276d77db AK |
10730 | |
10731 | err = sysfs_get_str(info, NULL, "array_state", (char *)buffer, 20); | |
10732 | if (err < 1) | |
10733 | return 1; | |
10734 | ||
10735 | /* recover data only during assemblation */ | |
10736 | if (strncmp(buffer, "inactive", 8) != 0) | |
10737 | return 0; | |
10738 | /* no data to recover */ | |
10739 | if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL) | |
10740 | return 0; | |
10741 | if (curr_migr_unit >= num_migr_units) | |
10742 | return 1; | |
10743 | ||
10744 | /* find device during reshape */ | |
10745 | for (id = super->devlist; id; id = id->next) | |
10746 | if (is_gen_migration(id->dev)) | |
10747 | break; | |
10748 | if (id == NULL) | |
10749 | return 1; | |
10750 | ||
238c0a71 | 10751 | map_dest = get_imsm_map(id->dev, MAP_0); |
276d77db AK |
10752 | new_disks = map_dest->num_members; |
10753 | ||
10754 | read_offset = (unsigned long long) | |
10755 | __le32_to_cpu(migr_rec->ckpt_area_pba) * 512; | |
10756 | ||
10757 | write_offset = ((unsigned long long) | |
10758 | __le32_to_cpu(migr_rec->dest_1st_member_lba) + | |
5551b113 | 10759 | pba_of_lba0(map_dest)) * 512; |
276d77db AK |
10760 | |
10761 | unit_len = __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
f36a9ecd | 10762 | if (posix_memalign((void **)&buf, sector_size, unit_len) != 0) |
276d77db | 10763 | goto abort; |
503975b9 | 10764 | targets = xcalloc(new_disks, sizeof(int)); |
276d77db | 10765 | |
9a717282 | 10766 | if (open_backup_targets(info, new_disks, targets, super, id->dev)) { |
e7b84f9d | 10767 | pr_err("Cannot open some devices belonging to array.\n"); |
f627f5ad AK |
10768 | goto abort; |
10769 | } | |
276d77db AK |
10770 | |
10771 | for (i = 0; i < new_disks; i++) { | |
6c3560c0 AK |
10772 | if (targets[i] < 0) { |
10773 | skipped_disks++; | |
10774 | continue; | |
10775 | } | |
276d77db | 10776 | if (lseek64(targets[i], read_offset, SEEK_SET) < 0) { |
e7b84f9d N |
10777 | pr_err("Cannot seek to block: %s\n", |
10778 | strerror(errno)); | |
137debce AK |
10779 | skipped_disks++; |
10780 | continue; | |
276d77db | 10781 | } |
9ec11d1a | 10782 | if ((unsigned)read(targets[i], buf, unit_len) != unit_len) { |
e7b84f9d N |
10783 | pr_err("Cannot read copy area block: %s\n", |
10784 | strerror(errno)); | |
137debce AK |
10785 | skipped_disks++; |
10786 | continue; | |
276d77db AK |
10787 | } |
10788 | if (lseek64(targets[i], write_offset, SEEK_SET) < 0) { | |
e7b84f9d N |
10789 | pr_err("Cannot seek to block: %s\n", |
10790 | strerror(errno)); | |
137debce AK |
10791 | skipped_disks++; |
10792 | continue; | |
276d77db | 10793 | } |
9ec11d1a | 10794 | if ((unsigned)write(targets[i], buf, unit_len) != unit_len) { |
e7b84f9d N |
10795 | pr_err("Cannot restore block: %s\n", |
10796 | strerror(errno)); | |
137debce AK |
10797 | skipped_disks++; |
10798 | continue; | |
276d77db AK |
10799 | } |
10800 | } | |
10801 | ||
137debce AK |
10802 | if (skipped_disks > imsm_get_allowed_degradation(info->new_level, |
10803 | new_disks, | |
10804 | super, | |
10805 | id->dev)) { | |
7a862a02 | 10806 | pr_err("Cannot restore data from backup. Too many failed disks\n"); |
6c3560c0 AK |
10807 | goto abort; |
10808 | } | |
10809 | ||
befb629b AK |
10810 | if (save_checkpoint_imsm(st, info, UNIT_SRC_NORMAL)) { |
10811 | /* ignore error == 2, this can mean end of reshape here | |
10812 | */ | |
7a862a02 | 10813 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL) during restart\n"); |
befb629b | 10814 | } else |
276d77db | 10815 | retval = 0; |
276d77db AK |
10816 | |
10817 | abort: | |
10818 | if (targets) { | |
10819 | for (i = 0; i < new_disks; i++) | |
10820 | if (targets[i]) | |
10821 | close(targets[i]); | |
10822 | free(targets); | |
10823 | } | |
10824 | free(buf); | |
10825 | return retval; | |
10826 | } | |
10827 | ||
2cda7640 ML |
10828 | static char disk_by_path[] = "/dev/disk/by-path/"; |
10829 | ||
10830 | static const char *imsm_get_disk_controller_domain(const char *path) | |
10831 | { | |
2cda7640 | 10832 | char disk_path[PATH_MAX]; |
96234762 LM |
10833 | char *drv=NULL; |
10834 | struct stat st; | |
2cda7640 | 10835 | |
6d8d290a | 10836 | strcpy(disk_path, disk_by_path); |
96234762 LM |
10837 | strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1); |
10838 | if (stat(disk_path, &st) == 0) { | |
10839 | struct sys_dev* hba; | |
594dc1b8 | 10840 | char *path; |
96234762 LM |
10841 | |
10842 | path = devt_to_devpath(st.st_rdev); | |
10843 | if (path == NULL) | |
10844 | return "unknown"; | |
10845 | hba = find_disk_attached_hba(-1, path); | |
10846 | if (hba && hba->type == SYS_DEV_SAS) | |
10847 | drv = "isci"; | |
10848 | else if (hba && hba->type == SYS_DEV_SATA) | |
10849 | drv = "ahci"; | |
c6839718 MT |
10850 | else if (hba && hba->type == SYS_DEV_VMD) |
10851 | drv = "vmd"; | |
10852 | else if (hba && hba->type == SYS_DEV_NVME) | |
10853 | drv = "nvme"; | |
1011e834 | 10854 | else |
96234762 LM |
10855 | drv = "unknown"; |
10856 | dprintf("path: %s hba: %s attached: %s\n", | |
10857 | path, (hba) ? hba->path : "NULL", drv); | |
10858 | free(path); | |
2cda7640 | 10859 | } |
96234762 | 10860 | return drv; |
2cda7640 ML |
10861 | } |
10862 | ||
4dd2df09 | 10863 | static char *imsm_find_array_devnm_by_subdev(int subdev, char *container) |
78b10e66 | 10864 | { |
4dd2df09 | 10865 | static char devnm[32]; |
78b10e66 N |
10866 | char subdev_name[20]; |
10867 | struct mdstat_ent *mdstat; | |
10868 | ||
10869 | sprintf(subdev_name, "%d", subdev); | |
10870 | mdstat = mdstat_by_subdev(subdev_name, container); | |
10871 | if (!mdstat) | |
4dd2df09 | 10872 | return NULL; |
78b10e66 | 10873 | |
4dd2df09 | 10874 | strcpy(devnm, mdstat->devnm); |
78b10e66 | 10875 | free_mdstat(mdstat); |
4dd2df09 | 10876 | return devnm; |
78b10e66 N |
10877 | } |
10878 | ||
10879 | static int imsm_reshape_is_allowed_on_container(struct supertype *st, | |
10880 | struct geo_params *geo, | |
fbf3d202 AK |
10881 | int *old_raid_disks, |
10882 | int direction) | |
78b10e66 | 10883 | { |
694575e7 KW |
10884 | /* currently we only support increasing the number of devices |
10885 | * for a container. This increases the number of device for each | |
10886 | * member array. They must all be RAID0 or RAID5. | |
10887 | */ | |
78b10e66 N |
10888 | int ret_val = 0; |
10889 | struct mdinfo *info, *member; | |
10890 | int devices_that_can_grow = 0; | |
10891 | ||
7a862a02 | 10892 | dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): st->devnm = (%s)\n", st->devnm); |
78b10e66 | 10893 | |
d04f65f4 | 10894 | if (geo->size > 0 || |
78b10e66 N |
10895 | geo->level != UnSet || |
10896 | geo->layout != UnSet || | |
10897 | geo->chunksize != 0 || | |
10898 | geo->raid_disks == UnSet) { | |
7a862a02 | 10899 | dprintf("imsm: Container operation is allowed for raid disks number change only.\n"); |
78b10e66 N |
10900 | return ret_val; |
10901 | } | |
10902 | ||
fbf3d202 | 10903 | if (direction == ROLLBACK_METADATA_CHANGES) { |
7a862a02 | 10904 | dprintf("imsm: Metadata changes rollback is not supported for container operation.\n"); |
fbf3d202 AK |
10905 | return ret_val; |
10906 | } | |
10907 | ||
78b10e66 N |
10908 | info = container_content_imsm(st, NULL); |
10909 | for (member = info; member; member = member->next) { | |
4dd2df09 | 10910 | char *result; |
78b10e66 N |
10911 | |
10912 | dprintf("imsm: checking device_num: %i\n", | |
10913 | member->container_member); | |
10914 | ||
d7d205bd | 10915 | if (geo->raid_disks <= member->array.raid_disks) { |
78b10e66 N |
10916 | /* we work on container for Online Capacity Expansion |
10917 | * only so raid_disks has to grow | |
10918 | */ | |
7a862a02 | 10919 | dprintf("imsm: for container operation raid disks increase is required\n"); |
78b10e66 N |
10920 | break; |
10921 | } | |
10922 | ||
089f9d79 | 10923 | if (info->array.level != 0 && info->array.level != 5) { |
78b10e66 N |
10924 | /* we cannot use this container with other raid level |
10925 | */ | |
7a862a02 | 10926 | dprintf("imsm: for container operation wrong raid level (%i) detected\n", |
78b10e66 N |
10927 | info->array.level); |
10928 | break; | |
10929 | } else { | |
10930 | /* check for platform support | |
10931 | * for this raid level configuration | |
10932 | */ | |
10933 | struct intel_super *super = st->sb; | |
10934 | if (!is_raid_level_supported(super->orom, | |
10935 | member->array.level, | |
10936 | geo->raid_disks)) { | |
7a862a02 | 10937 | dprintf("platform does not support raid%d with %d disk%s\n", |
78b10e66 N |
10938 | info->array.level, |
10939 | geo->raid_disks, | |
10940 | geo->raid_disks > 1 ? "s" : ""); | |
10941 | break; | |
10942 | } | |
2a4a08e7 AK |
10943 | /* check if component size is aligned to chunk size |
10944 | */ | |
10945 | if (info->component_size % | |
10946 | (info->array.chunk_size/512)) { | |
7a862a02 | 10947 | dprintf("Component size is not aligned to chunk size\n"); |
2a4a08e7 AK |
10948 | break; |
10949 | } | |
78b10e66 N |
10950 | } |
10951 | ||
10952 | if (*old_raid_disks && | |
10953 | info->array.raid_disks != *old_raid_disks) | |
10954 | break; | |
10955 | *old_raid_disks = info->array.raid_disks; | |
10956 | ||
10957 | /* All raid5 and raid0 volumes in container | |
10958 | * have to be ready for Online Capacity Expansion | |
10959 | * so they need to be assembled. We have already | |
10960 | * checked that no recovery etc is happening. | |
10961 | */ | |
4dd2df09 N |
10962 | result = imsm_find_array_devnm_by_subdev(member->container_member, |
10963 | st->container_devnm); | |
10964 | if (result == NULL) { | |
78b10e66 N |
10965 | dprintf("imsm: cannot find array\n"); |
10966 | break; | |
10967 | } | |
10968 | devices_that_can_grow++; | |
10969 | } | |
10970 | sysfs_free(info); | |
10971 | if (!member && devices_that_can_grow) | |
10972 | ret_val = 1; | |
10973 | ||
10974 | if (ret_val) | |
1ade5cc1 | 10975 | dprintf("Container operation allowed\n"); |
78b10e66 | 10976 | else |
1ade5cc1 | 10977 | dprintf("Error: %i\n", ret_val); |
78b10e66 N |
10978 | |
10979 | return ret_val; | |
10980 | } | |
10981 | ||
10982 | /* Function: get_spares_for_grow | |
10983 | * Description: Allocates memory and creates list of spare devices | |
1011e834 | 10984 | * avaliable in container. Checks if spare drive size is acceptable. |
78b10e66 N |
10985 | * Parameters: Pointer to the supertype structure |
10986 | * Returns: Pointer to the list of spare devices (mdinfo structure) on success, | |
1011e834 | 10987 | * NULL if fail |
78b10e66 N |
10988 | */ |
10989 | static struct mdinfo *get_spares_for_grow(struct supertype *st) | |
10990 | { | |
fbfdcb06 AO |
10991 | struct spare_criteria sc; |
10992 | ||
10993 | get_spare_criteria_imsm(st, &sc); | |
10994 | return container_choose_spares(st, &sc, NULL, NULL, NULL, 0); | |
78b10e66 N |
10995 | } |
10996 | ||
10997 | /****************************************************************************** | |
10998 | * function: imsm_create_metadata_update_for_reshape | |
10999 | * Function creates update for whole IMSM container. | |
11000 | * | |
11001 | ******************************************************************************/ | |
11002 | static int imsm_create_metadata_update_for_reshape( | |
11003 | struct supertype *st, | |
11004 | struct geo_params *geo, | |
11005 | int old_raid_disks, | |
11006 | struct imsm_update_reshape **updatep) | |
11007 | { | |
11008 | struct intel_super *super = st->sb; | |
11009 | struct imsm_super *mpb = super->anchor; | |
594dc1b8 JS |
11010 | int update_memory_size; |
11011 | struct imsm_update_reshape *u; | |
11012 | struct mdinfo *spares; | |
78b10e66 | 11013 | int i; |
594dc1b8 | 11014 | int delta_disks; |
bbd24d86 | 11015 | struct mdinfo *dev; |
78b10e66 | 11016 | |
1ade5cc1 | 11017 | dprintf("(enter) raid_disks = %i\n", geo->raid_disks); |
78b10e66 N |
11018 | |
11019 | delta_disks = geo->raid_disks - old_raid_disks; | |
11020 | ||
11021 | /* size of all update data without anchor */ | |
11022 | update_memory_size = sizeof(struct imsm_update_reshape); | |
11023 | ||
11024 | /* now add space for spare disks that we need to add. */ | |
11025 | update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1); | |
11026 | ||
503975b9 | 11027 | u = xcalloc(1, update_memory_size); |
78b10e66 N |
11028 | u->type = update_reshape_container_disks; |
11029 | u->old_raid_disks = old_raid_disks; | |
11030 | u->new_raid_disks = geo->raid_disks; | |
11031 | ||
11032 | /* now get spare disks list | |
11033 | */ | |
11034 | spares = get_spares_for_grow(st); | |
11035 | ||
d7be7d87 | 11036 | if (spares == NULL || delta_disks > spares->array.spare_disks) { |
7a862a02 | 11037 | pr_err("imsm: ERROR: Cannot get spare devices for %s.\n", geo->dev_name); |
e4c72d1d | 11038 | i = -1; |
78b10e66 N |
11039 | goto abort; |
11040 | } | |
11041 | ||
11042 | /* we have got spares | |
11043 | * update disk list in imsm_disk list table in anchor | |
11044 | */ | |
11045 | dprintf("imsm: %i spares are available.\n\n", | |
11046 | spares->array.spare_disks); | |
11047 | ||
bbd24d86 | 11048 | dev = spares->devs; |
78b10e66 | 11049 | for (i = 0; i < delta_disks; i++) { |
78b10e66 N |
11050 | struct dl *dl; |
11051 | ||
bbd24d86 AK |
11052 | if (dev == NULL) |
11053 | break; | |
78b10e66 N |
11054 | u->new_disks[i] = makedev(dev->disk.major, |
11055 | dev->disk.minor); | |
11056 | dl = get_disk_super(super, dev->disk.major, dev->disk.minor); | |
ee4beede AK |
11057 | dl->index = mpb->num_disks; |
11058 | mpb->num_disks++; | |
bbd24d86 | 11059 | dev = dev->next; |
78b10e66 | 11060 | } |
78b10e66 N |
11061 | |
11062 | abort: | |
11063 | /* free spares | |
11064 | */ | |
11065 | sysfs_free(spares); | |
11066 | ||
d677e0b8 | 11067 | dprintf("imsm: reshape update preparation :"); |
78b10e66 | 11068 | if (i == delta_disks) { |
1ade5cc1 | 11069 | dprintf_cont(" OK\n"); |
78b10e66 N |
11070 | *updatep = u; |
11071 | return update_memory_size; | |
11072 | } | |
11073 | free(u); | |
1ade5cc1 | 11074 | dprintf_cont(" Error\n"); |
78b10e66 N |
11075 | |
11076 | return 0; | |
11077 | } | |
11078 | ||
f3871fdc AK |
11079 | /****************************************************************************** |
11080 | * function: imsm_create_metadata_update_for_size_change() | |
11081 | * Creates update for IMSM array for array size change. | |
11082 | * | |
11083 | ******************************************************************************/ | |
11084 | static int imsm_create_metadata_update_for_size_change( | |
11085 | struct supertype *st, | |
11086 | struct geo_params *geo, | |
11087 | struct imsm_update_size_change **updatep) | |
11088 | { | |
11089 | struct intel_super *super = st->sb; | |
594dc1b8 JS |
11090 | int update_memory_size; |
11091 | struct imsm_update_size_change *u; | |
f3871fdc | 11092 | |
1ade5cc1 | 11093 | dprintf("(enter) New size = %llu\n", geo->size); |
f3871fdc AK |
11094 | |
11095 | /* size of all update data without anchor */ | |
11096 | update_memory_size = sizeof(struct imsm_update_size_change); | |
11097 | ||
503975b9 | 11098 | u = xcalloc(1, update_memory_size); |
f3871fdc AK |
11099 | u->type = update_size_change; |
11100 | u->subdev = super->current_vol; | |
11101 | u->new_size = geo->size; | |
11102 | ||
11103 | dprintf("imsm: reshape update preparation : OK\n"); | |
11104 | *updatep = u; | |
11105 | ||
11106 | return update_memory_size; | |
11107 | } | |
11108 | ||
48c5303a PC |
11109 | /****************************************************************************** |
11110 | * function: imsm_create_metadata_update_for_migration() | |
11111 | * Creates update for IMSM array. | |
11112 | * | |
11113 | ******************************************************************************/ | |
11114 | static int imsm_create_metadata_update_for_migration( | |
11115 | struct supertype *st, | |
11116 | struct geo_params *geo, | |
11117 | struct imsm_update_reshape_migration **updatep) | |
11118 | { | |
11119 | struct intel_super *super = st->sb; | |
594dc1b8 JS |
11120 | int update_memory_size; |
11121 | struct imsm_update_reshape_migration *u; | |
48c5303a PC |
11122 | struct imsm_dev *dev; |
11123 | int previous_level = -1; | |
11124 | ||
1ade5cc1 | 11125 | dprintf("(enter) New Level = %i\n", geo->level); |
48c5303a PC |
11126 | |
11127 | /* size of all update data without anchor */ | |
11128 | update_memory_size = sizeof(struct imsm_update_reshape_migration); | |
11129 | ||
503975b9 | 11130 | u = xcalloc(1, update_memory_size); |
48c5303a PC |
11131 | u->type = update_reshape_migration; |
11132 | u->subdev = super->current_vol; | |
11133 | u->new_level = geo->level; | |
11134 | u->new_layout = geo->layout; | |
11135 | u->new_raid_disks = u->old_raid_disks = geo->raid_disks; | |
11136 | u->new_disks[0] = -1; | |
4bba0439 | 11137 | u->new_chunksize = -1; |
48c5303a PC |
11138 | |
11139 | dev = get_imsm_dev(super, u->subdev); | |
11140 | if (dev) { | |
11141 | struct imsm_map *map; | |
11142 | ||
238c0a71 | 11143 | map = get_imsm_map(dev, MAP_0); |
4bba0439 PC |
11144 | if (map) { |
11145 | int current_chunk_size = | |
11146 | __le16_to_cpu(map->blocks_per_strip) / 2; | |
11147 | ||
11148 | if (geo->chunksize != current_chunk_size) { | |
11149 | u->new_chunksize = geo->chunksize / 1024; | |
7a862a02 | 11150 | dprintf("imsm: chunk size change from %i to %i\n", |
4bba0439 PC |
11151 | current_chunk_size, u->new_chunksize); |
11152 | } | |
48c5303a | 11153 | previous_level = map->raid_level; |
4bba0439 | 11154 | } |
48c5303a | 11155 | } |
089f9d79 | 11156 | if (geo->level == 5 && previous_level == 0) { |
48c5303a PC |
11157 | struct mdinfo *spares = NULL; |
11158 | ||
11159 | u->new_raid_disks++; | |
11160 | spares = get_spares_for_grow(st); | |
089f9d79 | 11161 | if (spares == NULL || spares->array.spare_disks < 1) { |
48c5303a PC |
11162 | free(u); |
11163 | sysfs_free(spares); | |
11164 | update_memory_size = 0; | |
565cc99e | 11165 | pr_err("cannot get spare device for requested migration\n"); |
48c5303a PC |
11166 | return 0; |
11167 | } | |
11168 | sysfs_free(spares); | |
11169 | } | |
11170 | dprintf("imsm: reshape update preparation : OK\n"); | |
11171 | *updatep = u; | |
11172 | ||
11173 | return update_memory_size; | |
11174 | } | |
11175 | ||
8dd70bce AK |
11176 | static void imsm_update_metadata_locally(struct supertype *st, |
11177 | void *buf, int len) | |
11178 | { | |
11179 | struct metadata_update mu; | |
11180 | ||
11181 | mu.buf = buf; | |
11182 | mu.len = len; | |
11183 | mu.space = NULL; | |
11184 | mu.space_list = NULL; | |
11185 | mu.next = NULL; | |
5fe6f031 N |
11186 | if (imsm_prepare_update(st, &mu)) |
11187 | imsm_process_update(st, &mu); | |
8dd70bce AK |
11188 | |
11189 | while (mu.space_list) { | |
11190 | void **space = mu.space_list; | |
11191 | mu.space_list = *space; | |
11192 | free(space); | |
11193 | } | |
11194 | } | |
78b10e66 | 11195 | |
471bceb6 | 11196 | /*************************************************************************** |
694575e7 | 11197 | * Function: imsm_analyze_change |
471bceb6 | 11198 | * Description: Function analyze change for single volume |
1011e834 | 11199 | * and validate if transition is supported |
fbf3d202 AK |
11200 | * Parameters: Geometry parameters, supertype structure, |
11201 | * metadata change direction (apply/rollback) | |
694575e7 | 11202 | * Returns: Operation type code on success, -1 if fail |
471bceb6 KW |
11203 | ****************************************************************************/ |
11204 | enum imsm_reshape_type imsm_analyze_change(struct supertype *st, | |
fbf3d202 AK |
11205 | struct geo_params *geo, |
11206 | int direction) | |
694575e7 | 11207 | { |
471bceb6 KW |
11208 | struct mdinfo info; |
11209 | int change = -1; | |
11210 | int check_devs = 0; | |
c21e737b | 11211 | int chunk; |
67a2db32 AK |
11212 | /* number of added/removed disks in operation result */ |
11213 | int devNumChange = 0; | |
11214 | /* imsm compatible layout value for array geometry verification */ | |
11215 | int imsm_layout = -1; | |
7abc9871 AK |
11216 | int data_disks; |
11217 | struct imsm_dev *dev; | |
11218 | struct intel_super *super; | |
d04f65f4 | 11219 | unsigned long long current_size; |
65d38cca | 11220 | unsigned long long free_size; |
d04f65f4 | 11221 | unsigned long long max_size; |
65d38cca | 11222 | int rv; |
471bceb6 KW |
11223 | |
11224 | getinfo_super_imsm_volume(st, &info, NULL); | |
089f9d79 JS |
11225 | if (geo->level != info.array.level && geo->level >= 0 && |
11226 | geo->level != UnSet) { | |
471bceb6 KW |
11227 | switch (info.array.level) { |
11228 | case 0: | |
11229 | if (geo->level == 5) { | |
b5347799 | 11230 | change = CH_MIGRATION; |
e13ce846 | 11231 | if (geo->layout != ALGORITHM_LEFT_ASYMMETRIC) { |
7a862a02 | 11232 | pr_err("Error. Requested Layout not supported (left-asymmetric layout is supported only)!\n"); |
e13ce846 AK |
11233 | change = -1; |
11234 | goto analyse_change_exit; | |
11235 | } | |
67a2db32 | 11236 | imsm_layout = geo->layout; |
471bceb6 | 11237 | check_devs = 1; |
e91a3bad LM |
11238 | devNumChange = 1; /* parity disk added */ |
11239 | } else if (geo->level == 10) { | |
471bceb6 KW |
11240 | change = CH_TAKEOVER; |
11241 | check_devs = 1; | |
e91a3bad | 11242 | devNumChange = 2; /* two mirrors added */ |
67a2db32 | 11243 | imsm_layout = 0x102; /* imsm supported layout */ |
471bceb6 | 11244 | } |
dfe77a9e KW |
11245 | break; |
11246 | case 1: | |
471bceb6 KW |
11247 | case 10: |
11248 | if (geo->level == 0) { | |
11249 | change = CH_TAKEOVER; | |
11250 | check_devs = 1; | |
e91a3bad | 11251 | devNumChange = -(geo->raid_disks/2); |
67a2db32 | 11252 | imsm_layout = 0; /* imsm raid0 layout */ |
471bceb6 KW |
11253 | } |
11254 | break; | |
11255 | } | |
11256 | if (change == -1) { | |
7a862a02 | 11257 | pr_err("Error. Level Migration from %d to %d not supported!\n", |
e7b84f9d | 11258 | info.array.level, geo->level); |
471bceb6 KW |
11259 | goto analyse_change_exit; |
11260 | } | |
11261 | } else | |
11262 | geo->level = info.array.level; | |
11263 | ||
089f9d79 JS |
11264 | if (geo->layout != info.array.layout && |
11265 | (geo->layout != UnSet && geo->layout != -1)) { | |
b5347799 | 11266 | change = CH_MIGRATION; |
089f9d79 JS |
11267 | if (info.array.layout == 0 && info.array.level == 5 && |
11268 | geo->layout == 5) { | |
471bceb6 | 11269 | /* reshape 5 -> 4 */ |
089f9d79 JS |
11270 | } else if (info.array.layout == 5 && info.array.level == 5 && |
11271 | geo->layout == 0) { | |
471bceb6 KW |
11272 | /* reshape 4 -> 5 */ |
11273 | geo->layout = 0; | |
11274 | geo->level = 5; | |
11275 | } else { | |
7a862a02 | 11276 | pr_err("Error. Layout Migration from %d to %d not supported!\n", |
e7b84f9d | 11277 | info.array.layout, geo->layout); |
471bceb6 KW |
11278 | change = -1; |
11279 | goto analyse_change_exit; | |
11280 | } | |
67a2db32 | 11281 | } else { |
471bceb6 | 11282 | geo->layout = info.array.layout; |
67a2db32 AK |
11283 | if (imsm_layout == -1) |
11284 | imsm_layout = info.array.layout; | |
11285 | } | |
471bceb6 | 11286 | |
089f9d79 JS |
11287 | if (geo->chunksize > 0 && geo->chunksize != UnSet && |
11288 | geo->chunksize != info.array.chunk_size) { | |
2d2b0eb7 MD |
11289 | if (info.array.level == 10) { |
11290 | pr_err("Error. Chunk size change for RAID 10 is not supported.\n"); | |
11291 | change = -1; | |
11292 | goto analyse_change_exit; | |
1e9b2c3f PB |
11293 | } else if (info.component_size % (geo->chunksize/512)) { |
11294 | pr_err("New chunk size (%dK) does not evenly divide device size (%lluk). Aborting...\n", | |
11295 | geo->chunksize/1024, info.component_size/2); | |
11296 | change = -1; | |
11297 | goto analyse_change_exit; | |
2d2b0eb7 | 11298 | } |
b5347799 | 11299 | change = CH_MIGRATION; |
2d2b0eb7 | 11300 | } else { |
471bceb6 | 11301 | geo->chunksize = info.array.chunk_size; |
2d2b0eb7 | 11302 | } |
471bceb6 | 11303 | |
c21e737b | 11304 | chunk = geo->chunksize / 1024; |
7abc9871 AK |
11305 | |
11306 | super = st->sb; | |
11307 | dev = get_imsm_dev(super, super->current_vol); | |
11308 | data_disks = imsm_num_data_members(dev , MAP_0); | |
c41e00b2 | 11309 | /* compute current size per disk member |
7abc9871 | 11310 | */ |
c41e00b2 AK |
11311 | current_size = info.custom_array_size / data_disks; |
11312 | ||
089f9d79 | 11313 | if (geo->size > 0 && geo->size != MAX_SIZE) { |
c41e00b2 AK |
11314 | /* align component size |
11315 | */ | |
11316 | geo->size = imsm_component_size_aligment_check( | |
11317 | get_imsm_raid_level(dev->vol.map), | |
f36a9ecd | 11318 | chunk * 1024, super->sector_size, |
c41e00b2 | 11319 | geo->size * 2); |
65d0b4ce | 11320 | if (geo->size == 0) { |
7a862a02 | 11321 | pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is 0).\n", |
65d0b4ce LD |
11322 | current_size); |
11323 | goto analyse_change_exit; | |
11324 | } | |
c41e00b2 | 11325 | } |
7abc9871 | 11326 | |
089f9d79 | 11327 | if (current_size != geo->size && geo->size > 0) { |
7abc9871 | 11328 | if (change != -1) { |
7a862a02 | 11329 | pr_err("Error. Size change should be the only one at a time.\n"); |
7abc9871 AK |
11330 | change = -1; |
11331 | goto analyse_change_exit; | |
11332 | } | |
11333 | if ((super->current_vol + 1) != super->anchor->num_raid_devs) { | |
7a862a02 | 11334 | pr_err("Error. The last volume in container can be expanded only (%i/%s).\n", |
4dd2df09 | 11335 | super->current_vol, st->devnm); |
7abc9871 AK |
11336 | goto analyse_change_exit; |
11337 | } | |
65d38cca LD |
11338 | /* check the maximum available size |
11339 | */ | |
11340 | rv = imsm_get_free_size(st, dev->vol.map->num_members, | |
11341 | 0, chunk, &free_size); | |
11342 | if (rv == 0) | |
11343 | /* Cannot find maximum available space | |
11344 | */ | |
11345 | max_size = 0; | |
11346 | else { | |
11347 | max_size = free_size + current_size; | |
11348 | /* align component size | |
11349 | */ | |
11350 | max_size = imsm_component_size_aligment_check( | |
11351 | get_imsm_raid_level(dev->vol.map), | |
f36a9ecd | 11352 | chunk * 1024, super->sector_size, |
65d38cca LD |
11353 | max_size); |
11354 | } | |
d04f65f4 | 11355 | if (geo->size == MAX_SIZE) { |
b130333f AK |
11356 | /* requested size change to the maximum available size |
11357 | */ | |
65d38cca | 11358 | if (max_size == 0) { |
7a862a02 | 11359 | pr_err("Error. Cannot find maximum available space.\n"); |
b130333f AK |
11360 | change = -1; |
11361 | goto analyse_change_exit; | |
65d38cca LD |
11362 | } else |
11363 | geo->size = max_size; | |
c41e00b2 | 11364 | } |
b130333f | 11365 | |
681b7ae2 | 11366 | if (direction == ROLLBACK_METADATA_CHANGES) { |
fbf3d202 AK |
11367 | /* accept size for rollback only |
11368 | */ | |
11369 | } else { | |
11370 | /* round size due to metadata compatibility | |
11371 | */ | |
11372 | geo->size = (geo->size >> SECT_PER_MB_SHIFT) | |
11373 | << SECT_PER_MB_SHIFT; | |
11374 | dprintf("Prepare update for size change to %llu\n", | |
11375 | geo->size ); | |
11376 | if (current_size >= geo->size) { | |
7a862a02 | 11377 | pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is %llu).\n", |
e7b84f9d | 11378 | current_size, geo->size); |
fbf3d202 AK |
11379 | goto analyse_change_exit; |
11380 | } | |
65d38cca | 11381 | if (max_size && geo->size > max_size) { |
7a862a02 | 11382 | pr_err("Error. Requested size is larger than maximum available size (maximum available size is %llu, requested size /rounded/ is %llu).\n", |
e7b84f9d | 11383 | max_size, geo->size); |
65d38cca LD |
11384 | goto analyse_change_exit; |
11385 | } | |
7abc9871 AK |
11386 | } |
11387 | geo->size *= data_disks; | |
11388 | geo->raid_disks = dev->vol.map->num_members; | |
11389 | change = CH_ARRAY_SIZE; | |
11390 | } | |
471bceb6 KW |
11391 | if (!validate_geometry_imsm(st, |
11392 | geo->level, | |
67a2db32 | 11393 | imsm_layout, |
e91a3bad | 11394 | geo->raid_disks + devNumChange, |
c21e737b | 11395 | &chunk, |
af4348dd | 11396 | geo->size, INVALID_SECTORS, |
5308f117 | 11397 | 0, 0, info.consistency_policy, 1)) |
471bceb6 KW |
11398 | change = -1; |
11399 | ||
11400 | if (check_devs) { | |
11401 | struct intel_super *super = st->sb; | |
11402 | struct imsm_super *mpb = super->anchor; | |
11403 | ||
11404 | if (mpb->num_raid_devs > 1) { | |
7a862a02 | 11405 | pr_err("Error. Cannot perform operation on %s- for this operation it MUST be single array in container\n", |
e7b84f9d | 11406 | geo->dev_name); |
471bceb6 KW |
11407 | change = -1; |
11408 | } | |
11409 | } | |
11410 | ||
11411 | analyse_change_exit: | |
089f9d79 JS |
11412 | if (direction == ROLLBACK_METADATA_CHANGES && |
11413 | (change == CH_MIGRATION || change == CH_TAKEOVER)) { | |
7a862a02 | 11414 | dprintf("imsm: Metadata changes rollback is not supported for migration and takeover operations.\n"); |
fbf3d202 AK |
11415 | change = -1; |
11416 | } | |
471bceb6 | 11417 | return change; |
694575e7 KW |
11418 | } |
11419 | ||
bb025c2f KW |
11420 | int imsm_takeover(struct supertype *st, struct geo_params *geo) |
11421 | { | |
11422 | struct intel_super *super = st->sb; | |
11423 | struct imsm_update_takeover *u; | |
11424 | ||
503975b9 | 11425 | u = xmalloc(sizeof(struct imsm_update_takeover)); |
bb025c2f KW |
11426 | |
11427 | u->type = update_takeover; | |
11428 | u->subarray = super->current_vol; | |
11429 | ||
11430 | /* 10->0 transition */ | |
11431 | if (geo->level == 0) | |
11432 | u->direction = R10_TO_R0; | |
11433 | ||
0529c688 KW |
11434 | /* 0->10 transition */ |
11435 | if (geo->level == 10) | |
11436 | u->direction = R0_TO_R10; | |
11437 | ||
bb025c2f KW |
11438 | /* update metadata locally */ |
11439 | imsm_update_metadata_locally(st, u, | |
11440 | sizeof(struct imsm_update_takeover)); | |
11441 | /* and possibly remotely */ | |
11442 | if (st->update_tail) | |
11443 | append_metadata_update(st, u, | |
11444 | sizeof(struct imsm_update_takeover)); | |
11445 | else | |
11446 | free(u); | |
11447 | ||
11448 | return 0; | |
11449 | } | |
11450 | ||
d04f65f4 N |
11451 | static int imsm_reshape_super(struct supertype *st, unsigned long long size, |
11452 | int level, | |
78b10e66 | 11453 | int layout, int chunksize, int raid_disks, |
41784c88 | 11454 | int delta_disks, char *backup, char *dev, |
016e00f5 | 11455 | int direction, int verbose) |
78b10e66 | 11456 | { |
78b10e66 N |
11457 | int ret_val = 1; |
11458 | struct geo_params geo; | |
11459 | ||
1ade5cc1 | 11460 | dprintf("(enter)\n"); |
78b10e66 | 11461 | |
71204a50 | 11462 | memset(&geo, 0, sizeof(struct geo_params)); |
78b10e66 N |
11463 | |
11464 | geo.dev_name = dev; | |
4dd2df09 | 11465 | strcpy(geo.devnm, st->devnm); |
78b10e66 N |
11466 | geo.size = size; |
11467 | geo.level = level; | |
11468 | geo.layout = layout; | |
11469 | geo.chunksize = chunksize; | |
11470 | geo.raid_disks = raid_disks; | |
41784c88 AK |
11471 | if (delta_disks != UnSet) |
11472 | geo.raid_disks += delta_disks; | |
78b10e66 | 11473 | |
1ade5cc1 N |
11474 | dprintf("for level : %i\n", geo.level); |
11475 | dprintf("for raid_disks : %i\n", geo.raid_disks); | |
78b10e66 N |
11476 | |
11477 | if (experimental() == 0) | |
11478 | return ret_val; | |
11479 | ||
4dd2df09 | 11480 | if (strcmp(st->container_devnm, st->devnm) == 0) { |
694575e7 KW |
11481 | /* On container level we can only increase number of devices. */ |
11482 | dprintf("imsm: info: Container operation\n"); | |
78b10e66 | 11483 | int old_raid_disks = 0; |
6dc0be30 | 11484 | |
78b10e66 | 11485 | if (imsm_reshape_is_allowed_on_container( |
fbf3d202 | 11486 | st, &geo, &old_raid_disks, direction)) { |
78b10e66 N |
11487 | struct imsm_update_reshape *u = NULL; |
11488 | int len; | |
11489 | ||
11490 | len = imsm_create_metadata_update_for_reshape( | |
11491 | st, &geo, old_raid_disks, &u); | |
11492 | ||
ed08d51c AK |
11493 | if (len <= 0) { |
11494 | dprintf("imsm: Cannot prepare update\n"); | |
11495 | goto exit_imsm_reshape_super; | |
11496 | } | |
11497 | ||
8dd70bce AK |
11498 | ret_val = 0; |
11499 | /* update metadata locally */ | |
11500 | imsm_update_metadata_locally(st, u, len); | |
11501 | /* and possibly remotely */ | |
11502 | if (st->update_tail) | |
11503 | append_metadata_update(st, u, len); | |
11504 | else | |
ed08d51c | 11505 | free(u); |
8dd70bce | 11506 | |
694575e7 | 11507 | } else { |
7a862a02 | 11508 | pr_err("(imsm) Operation is not allowed on this container\n"); |
694575e7 KW |
11509 | } |
11510 | } else { | |
11511 | /* On volume level we support following operations | |
471bceb6 KW |
11512 | * - takeover: raid10 -> raid0; raid0 -> raid10 |
11513 | * - chunk size migration | |
11514 | * - migration: raid5 -> raid0; raid0 -> raid5 | |
11515 | */ | |
11516 | struct intel_super *super = st->sb; | |
11517 | struct intel_dev *dev = super->devlist; | |
4dd2df09 | 11518 | int change; |
694575e7 | 11519 | dprintf("imsm: info: Volume operation\n"); |
471bceb6 KW |
11520 | /* find requested device */ |
11521 | while (dev) { | |
1011e834 | 11522 | char *devnm = |
4dd2df09 N |
11523 | imsm_find_array_devnm_by_subdev( |
11524 | dev->index, st->container_devnm); | |
11525 | if (devnm && strcmp(devnm, geo.devnm) == 0) | |
471bceb6 KW |
11526 | break; |
11527 | dev = dev->next; | |
11528 | } | |
11529 | if (dev == NULL) { | |
4dd2df09 N |
11530 | pr_err("Cannot find %s (%s) subarray\n", |
11531 | geo.dev_name, geo.devnm); | |
471bceb6 KW |
11532 | goto exit_imsm_reshape_super; |
11533 | } | |
11534 | super->current_vol = dev->index; | |
fbf3d202 | 11535 | change = imsm_analyze_change(st, &geo, direction); |
694575e7 | 11536 | switch (change) { |
471bceb6 | 11537 | case CH_TAKEOVER: |
bb025c2f | 11538 | ret_val = imsm_takeover(st, &geo); |
694575e7 | 11539 | break; |
48c5303a PC |
11540 | case CH_MIGRATION: { |
11541 | struct imsm_update_reshape_migration *u = NULL; | |
11542 | int len = | |
11543 | imsm_create_metadata_update_for_migration( | |
11544 | st, &geo, &u); | |
11545 | if (len < 1) { | |
7a862a02 | 11546 | dprintf("imsm: Cannot prepare update\n"); |
48c5303a PC |
11547 | break; |
11548 | } | |
471bceb6 | 11549 | ret_val = 0; |
48c5303a PC |
11550 | /* update metadata locally */ |
11551 | imsm_update_metadata_locally(st, u, len); | |
11552 | /* and possibly remotely */ | |
11553 | if (st->update_tail) | |
11554 | append_metadata_update(st, u, len); | |
11555 | else | |
11556 | free(u); | |
11557 | } | |
11558 | break; | |
7abc9871 | 11559 | case CH_ARRAY_SIZE: { |
f3871fdc AK |
11560 | struct imsm_update_size_change *u = NULL; |
11561 | int len = | |
11562 | imsm_create_metadata_update_for_size_change( | |
11563 | st, &geo, &u); | |
11564 | if (len < 1) { | |
7a862a02 | 11565 | dprintf("imsm: Cannot prepare update\n"); |
f3871fdc AK |
11566 | break; |
11567 | } | |
11568 | ret_val = 0; | |
11569 | /* update metadata locally */ | |
11570 | imsm_update_metadata_locally(st, u, len); | |
11571 | /* and possibly remotely */ | |
11572 | if (st->update_tail) | |
11573 | append_metadata_update(st, u, len); | |
11574 | else | |
11575 | free(u); | |
7abc9871 AK |
11576 | } |
11577 | break; | |
471bceb6 KW |
11578 | default: |
11579 | ret_val = 1; | |
694575e7 | 11580 | } |
694575e7 | 11581 | } |
78b10e66 | 11582 | |
ed08d51c | 11583 | exit_imsm_reshape_super: |
78b10e66 N |
11584 | dprintf("imsm: reshape_super Exit code = %i\n", ret_val); |
11585 | return ret_val; | |
11586 | } | |
2cda7640 | 11587 | |
0febb20c AO |
11588 | #define COMPLETED_OK 0 |
11589 | #define COMPLETED_NONE 1 | |
11590 | #define COMPLETED_DELAYED 2 | |
11591 | ||
11592 | static int read_completed(int fd, unsigned long long *val) | |
11593 | { | |
11594 | int ret; | |
11595 | char buf[50]; | |
11596 | ||
11597 | ret = sysfs_fd_get_str(fd, buf, 50); | |
11598 | if (ret < 0) | |
11599 | return ret; | |
11600 | ||
11601 | ret = COMPLETED_OK; | |
11602 | if (strncmp(buf, "none", 4) == 0) { | |
11603 | ret = COMPLETED_NONE; | |
11604 | } else if (strncmp(buf, "delayed", 7) == 0) { | |
11605 | ret = COMPLETED_DELAYED; | |
11606 | } else { | |
11607 | char *ep; | |
11608 | *val = strtoull(buf, &ep, 0); | |
11609 | if (ep == buf || (*ep != 0 && *ep != '\n' && *ep != ' ')) | |
11610 | ret = -1; | |
11611 | } | |
11612 | return ret; | |
11613 | } | |
11614 | ||
eee67a47 AK |
11615 | /******************************************************************************* |
11616 | * Function: wait_for_reshape_imsm | |
11617 | * Description: Function writes new sync_max value and waits until | |
11618 | * reshape process reach new position | |
11619 | * Parameters: | |
11620 | * sra : general array info | |
eee67a47 AK |
11621 | * ndata : number of disks in new array's layout |
11622 | * Returns: | |
11623 | * 0 : success, | |
11624 | * 1 : there is no reshape in progress, | |
11625 | * -1 : fail | |
11626 | ******************************************************************************/ | |
ae9f01f8 | 11627 | int wait_for_reshape_imsm(struct mdinfo *sra, int ndata) |
eee67a47 | 11628 | { |
85ca499c | 11629 | int fd = sysfs_get_fd(sra, NULL, "sync_completed"); |
df2647fa | 11630 | int retry = 3; |
eee67a47 | 11631 | unsigned long long completed; |
ae9f01f8 AK |
11632 | /* to_complete : new sync_max position */ |
11633 | unsigned long long to_complete = sra->reshape_progress; | |
11634 | unsigned long long position_to_set = to_complete / ndata; | |
eee67a47 | 11635 | |
ae9f01f8 | 11636 | if (fd < 0) { |
1ade5cc1 | 11637 | dprintf("cannot open reshape_position\n"); |
eee67a47 | 11638 | return 1; |
ae9f01f8 | 11639 | } |
eee67a47 | 11640 | |
df2647fa PB |
11641 | do { |
11642 | if (sysfs_fd_get_ll(fd, &completed) < 0) { | |
11643 | if (!retry) { | |
11644 | dprintf("cannot read reshape_position (no reshape in progres)\n"); | |
11645 | close(fd); | |
11646 | return 1; | |
11647 | } | |
11648 | usleep(30000); | |
11649 | } else | |
11650 | break; | |
11651 | } while (retry--); | |
eee67a47 | 11652 | |
85ca499c | 11653 | if (completed > position_to_set) { |
1ade5cc1 | 11654 | dprintf("wrong next position to set %llu (%llu)\n", |
85ca499c | 11655 | to_complete, position_to_set); |
ae9f01f8 AK |
11656 | close(fd); |
11657 | return -1; | |
11658 | } | |
11659 | dprintf("Position set: %llu\n", position_to_set); | |
11660 | if (sysfs_set_num(sra, NULL, "sync_max", | |
11661 | position_to_set) != 0) { | |
1ade5cc1 | 11662 | dprintf("cannot set reshape position to %llu\n", |
ae9f01f8 AK |
11663 | position_to_set); |
11664 | close(fd); | |
11665 | return -1; | |
eee67a47 AK |
11666 | } |
11667 | ||
eee67a47 | 11668 | do { |
0febb20c | 11669 | int rc; |
eee67a47 | 11670 | char action[20]; |
5ff3a780 | 11671 | int timeout = 3000; |
0febb20c | 11672 | |
5ff3a780 | 11673 | sysfs_wait(fd, &timeout); |
a47e44fb AK |
11674 | if (sysfs_get_str(sra, NULL, "sync_action", |
11675 | action, 20) > 0 && | |
d7d3809a | 11676 | strncmp(action, "reshape", 7) != 0) { |
b2be2b62 AO |
11677 | if (strncmp(action, "idle", 4) == 0) |
11678 | break; | |
d7d3809a AP |
11679 | close(fd); |
11680 | return -1; | |
11681 | } | |
0febb20c AO |
11682 | |
11683 | rc = read_completed(fd, &completed); | |
11684 | if (rc < 0) { | |
1ade5cc1 | 11685 | dprintf("cannot read reshape_position (in loop)\n"); |
eee67a47 AK |
11686 | close(fd); |
11687 | return 1; | |
0febb20c AO |
11688 | } else if (rc == COMPLETED_NONE) |
11689 | break; | |
85ca499c | 11690 | } while (completed < position_to_set); |
b2be2b62 | 11691 | |
eee67a47 AK |
11692 | close(fd); |
11693 | return 0; | |
eee67a47 AK |
11694 | } |
11695 | ||
b915c95f AK |
11696 | /******************************************************************************* |
11697 | * Function: check_degradation_change | |
11698 | * Description: Check that array hasn't become failed. | |
11699 | * Parameters: | |
11700 | * info : for sysfs access | |
11701 | * sources : source disks descriptors | |
11702 | * degraded: previous degradation level | |
11703 | * Returns: | |
11704 | * degradation level | |
11705 | ******************************************************************************/ | |
11706 | int check_degradation_change(struct mdinfo *info, | |
11707 | int *sources, | |
11708 | int degraded) | |
11709 | { | |
11710 | unsigned long long new_degraded; | |
e1993023 LD |
11711 | int rv; |
11712 | ||
11713 | rv = sysfs_get_ll(info, NULL, "degraded", &new_degraded); | |
089f9d79 | 11714 | if (rv == -1 || (new_degraded != (unsigned long long)degraded)) { |
b915c95f AK |
11715 | /* check each device to ensure it is still working */ |
11716 | struct mdinfo *sd; | |
11717 | new_degraded = 0; | |
11718 | for (sd = info->devs ; sd ; sd = sd->next) { | |
11719 | if (sd->disk.state & (1<<MD_DISK_FAULTY)) | |
11720 | continue; | |
11721 | if (sd->disk.state & (1<<MD_DISK_SYNC)) { | |
cf52eff5 TM |
11722 | char sbuf[100]; |
11723 | ||
b915c95f | 11724 | if (sysfs_get_str(info, |
cf52eff5 | 11725 | sd, "state", sbuf, sizeof(sbuf)) < 0 || |
b915c95f AK |
11726 | strstr(sbuf, "faulty") || |
11727 | strstr(sbuf, "in_sync") == NULL) { | |
11728 | /* this device is dead */ | |
11729 | sd->disk.state = (1<<MD_DISK_FAULTY); | |
11730 | if (sd->disk.raid_disk >= 0 && | |
11731 | sources[sd->disk.raid_disk] >= 0) { | |
11732 | close(sources[ | |
11733 | sd->disk.raid_disk]); | |
11734 | sources[sd->disk.raid_disk] = | |
11735 | -1; | |
11736 | } | |
11737 | new_degraded++; | |
11738 | } | |
11739 | } | |
11740 | } | |
11741 | } | |
11742 | ||
11743 | return new_degraded; | |
11744 | } | |
11745 | ||
10f22854 AK |
11746 | /******************************************************************************* |
11747 | * Function: imsm_manage_reshape | |
11748 | * Description: Function finds array under reshape and it manages reshape | |
11749 | * process. It creates stripes backups (if required) and sets | |
942e1cdb | 11750 | * checkpoints. |
10f22854 AK |
11751 | * Parameters: |
11752 | * afd : Backup handle (nattive) - not used | |
11753 | * sra : general array info | |
11754 | * reshape : reshape parameters - not used | |
11755 | * st : supertype structure | |
11756 | * blocks : size of critical section [blocks] | |
11757 | * fds : table of source device descriptor | |
11758 | * offsets : start of array (offest per devices) | |
11759 | * dests : not used | |
11760 | * destfd : table of destination device descriptor | |
11761 | * destoffsets : table of destination offsets (per device) | |
11762 | * Returns: | |
11763 | * 1 : success, reshape is done | |
11764 | * 0 : fail | |
11765 | ******************************************************************************/ | |
999b4972 N |
11766 | static int imsm_manage_reshape( |
11767 | int afd, struct mdinfo *sra, struct reshape *reshape, | |
10f22854 | 11768 | struct supertype *st, unsigned long backup_blocks, |
999b4972 N |
11769 | int *fds, unsigned long long *offsets, |
11770 | int dests, int *destfd, unsigned long long *destoffsets) | |
11771 | { | |
10f22854 AK |
11772 | int ret_val = 0; |
11773 | struct intel_super *super = st->sb; | |
594dc1b8 | 11774 | struct intel_dev *dv; |
de44e46f | 11775 | unsigned int sector_size = super->sector_size; |
10f22854 | 11776 | struct imsm_dev *dev = NULL; |
a6b6d984 | 11777 | struct imsm_map *map_src; |
10f22854 AK |
11778 | int migr_vol_qan = 0; |
11779 | int ndata, odata; /* [bytes] */ | |
11780 | int chunk; /* [bytes] */ | |
11781 | struct migr_record *migr_rec; | |
11782 | char *buf = NULL; | |
11783 | unsigned int buf_size; /* [bytes] */ | |
11784 | unsigned long long max_position; /* array size [bytes] */ | |
11785 | unsigned long long next_step; /* [blocks]/[bytes] */ | |
11786 | unsigned long long old_data_stripe_length; | |
10f22854 AK |
11787 | unsigned long long start_src; /* [bytes] */ |
11788 | unsigned long long start; /* [bytes] */ | |
11789 | unsigned long long start_buf_shift; /* [bytes] */ | |
b915c95f | 11790 | int degraded = 0; |
ab724b98 | 11791 | int source_layout = 0; |
10f22854 | 11792 | |
79a16a9b JS |
11793 | if (!sra) |
11794 | return ret_val; | |
11795 | ||
11796 | if (!fds || !offsets) | |
10f22854 AK |
11797 | goto abort; |
11798 | ||
11799 | /* Find volume during the reshape */ | |
11800 | for (dv = super->devlist; dv; dv = dv->next) { | |
fc54fe7a JS |
11801 | if (dv->dev->vol.migr_type == MIGR_GEN_MIGR && |
11802 | dv->dev->vol.migr_state == 1) { | |
10f22854 AK |
11803 | dev = dv->dev; |
11804 | migr_vol_qan++; | |
11805 | } | |
11806 | } | |
11807 | /* Only one volume can migrate at the same time */ | |
11808 | if (migr_vol_qan != 1) { | |
676e87a8 | 11809 | pr_err("%s", migr_vol_qan ? |
10f22854 AK |
11810 | "Number of migrating volumes greater than 1\n" : |
11811 | "There is no volume during migrationg\n"); | |
11812 | goto abort; | |
11813 | } | |
11814 | ||
238c0a71 | 11815 | map_src = get_imsm_map(dev, MAP_1); |
10f22854 AK |
11816 | if (map_src == NULL) |
11817 | goto abort; | |
10f22854 | 11818 | |
238c0a71 AK |
11819 | ndata = imsm_num_data_members(dev, MAP_0); |
11820 | odata = imsm_num_data_members(dev, MAP_1); | |
10f22854 | 11821 | |
7b1ab482 | 11822 | chunk = __le16_to_cpu(map_src->blocks_per_strip) * 512; |
10f22854 AK |
11823 | old_data_stripe_length = odata * chunk; |
11824 | ||
11825 | migr_rec = super->migr_rec; | |
11826 | ||
10f22854 AK |
11827 | /* initialize migration record for start condition */ |
11828 | if (sra->reshape_progress == 0) | |
11829 | init_migr_record_imsm(st, dev, sra); | |
b2c59438 AK |
11830 | else { |
11831 | if (__le32_to_cpu(migr_rec->rec_status) != UNIT_SRC_NORMAL) { | |
7a862a02 | 11832 | dprintf("imsm: cannot restart migration when data are present in copy area.\n"); |
b2c59438 AK |
11833 | goto abort; |
11834 | } | |
6a75c8ca AK |
11835 | /* Save checkpoint to update migration record for current |
11836 | * reshape position (in md). It can be farther than current | |
11837 | * reshape position in metadata. | |
11838 | */ | |
11839 | if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) { | |
11840 | /* ignore error == 2, this can mean end of reshape here | |
11841 | */ | |
7a862a02 | 11842 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL, initial save)\n"); |
6a75c8ca AK |
11843 | goto abort; |
11844 | } | |
b2c59438 | 11845 | } |
10f22854 AK |
11846 | |
11847 | /* size for data */ | |
11848 | buf_size = __le32_to_cpu(migr_rec->blocks_per_unit) * 512; | |
11849 | /* extend buffer size for parity disk */ | |
11850 | buf_size += __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
11851 | /* add space for stripe aligment */ | |
11852 | buf_size += old_data_stripe_length; | |
de44e46f PB |
11853 | if (posix_memalign((void **)&buf, MAX_SECTOR_SIZE, buf_size)) { |
11854 | dprintf("imsm: Cannot allocate checkpoint buffer\n"); | |
10f22854 AK |
11855 | goto abort; |
11856 | } | |
11857 | ||
3ef4403c | 11858 | max_position = sra->component_size * ndata; |
68eb8bc6 | 11859 | source_layout = imsm_level_to_layout(map_src->raid_level); |
10f22854 AK |
11860 | |
11861 | while (__le32_to_cpu(migr_rec->curr_migr_unit) < | |
11862 | __le32_to_cpu(migr_rec->num_migr_units)) { | |
11863 | /* current reshape position [blocks] */ | |
11864 | unsigned long long current_position = | |
11865 | __le32_to_cpu(migr_rec->blocks_per_unit) | |
11866 | * __le32_to_cpu(migr_rec->curr_migr_unit); | |
11867 | unsigned long long border; | |
11868 | ||
b915c95f AK |
11869 | /* Check that array hasn't become failed. |
11870 | */ | |
11871 | degraded = check_degradation_change(sra, fds, degraded); | |
11872 | if (degraded > 1) { | |
7a862a02 | 11873 | dprintf("imsm: Abort reshape due to degradation level (%i)\n", degraded); |
b915c95f AK |
11874 | goto abort; |
11875 | } | |
11876 | ||
10f22854 AK |
11877 | next_step = __le32_to_cpu(migr_rec->blocks_per_unit); |
11878 | ||
11879 | if ((current_position + next_step) > max_position) | |
11880 | next_step = max_position - current_position; | |
11881 | ||
92144abf | 11882 | start = current_position * 512; |
10f22854 | 11883 | |
942e1cdb | 11884 | /* align reading start to old geometry */ |
10f22854 AK |
11885 | start_buf_shift = start % old_data_stripe_length; |
11886 | start_src = start - start_buf_shift; | |
11887 | ||
11888 | border = (start_src / odata) - (start / ndata); | |
11889 | border /= 512; | |
11890 | if (border <= __le32_to_cpu(migr_rec->dest_depth_per_unit)) { | |
11891 | /* save critical stripes to buf | |
11892 | * start - start address of current unit | |
11893 | * to backup [bytes] | |
11894 | * start_src - start address of current unit | |
11895 | * to backup alligned to source array | |
11896 | * [bytes] | |
11897 | */ | |
594dc1b8 | 11898 | unsigned long long next_step_filler; |
10f22854 AK |
11899 | unsigned long long copy_length = next_step * 512; |
11900 | ||
11901 | /* allign copy area length to stripe in old geometry */ | |
11902 | next_step_filler = ((copy_length + start_buf_shift) | |
11903 | % old_data_stripe_length); | |
11904 | if (next_step_filler) | |
11905 | next_step_filler = (old_data_stripe_length | |
11906 | - next_step_filler); | |
7a862a02 | 11907 | 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 |
11908 | start, start_src, copy_length, |
11909 | start_buf_shift, next_step_filler); | |
11910 | ||
11911 | if (save_stripes(fds, offsets, map_src->num_members, | |
ab724b98 AK |
11912 | chunk, map_src->raid_level, |
11913 | source_layout, 0, NULL, start_src, | |
10f22854 AK |
11914 | copy_length + |
11915 | next_step_filler + start_buf_shift, | |
11916 | buf)) { | |
7a862a02 | 11917 | dprintf("imsm: Cannot save stripes to buffer\n"); |
10f22854 AK |
11918 | goto abort; |
11919 | } | |
11920 | /* Convert data to destination format and store it | |
11921 | * in backup general migration area | |
11922 | */ | |
11923 | if (save_backup_imsm(st, dev, sra, | |
aea93171 | 11924 | buf + start_buf_shift, copy_length)) { |
7a862a02 | 11925 | dprintf("imsm: Cannot save stripes to target devices\n"); |
10f22854 AK |
11926 | goto abort; |
11927 | } | |
11928 | if (save_checkpoint_imsm(st, sra, | |
11929 | UNIT_SRC_IN_CP_AREA)) { | |
7a862a02 | 11930 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_IN_CP_AREA)\n"); |
10f22854 AK |
11931 | goto abort; |
11932 | } | |
8016a6d4 AK |
11933 | } else { |
11934 | /* set next step to use whole border area */ | |
11935 | border /= next_step; | |
11936 | if (border > 1) | |
11937 | next_step *= border; | |
10f22854 AK |
11938 | } |
11939 | /* When data backed up, checkpoint stored, | |
11940 | * kick the kernel to reshape unit of data | |
11941 | */ | |
11942 | next_step = next_step + sra->reshape_progress; | |
8016a6d4 AK |
11943 | /* limit next step to array max position */ |
11944 | if (next_step > max_position) | |
11945 | next_step = max_position; | |
10f22854 AK |
11946 | sysfs_set_num(sra, NULL, "suspend_lo", sra->reshape_progress); |
11947 | sysfs_set_num(sra, NULL, "suspend_hi", next_step); | |
ae9f01f8 | 11948 | sra->reshape_progress = next_step; |
10f22854 AK |
11949 | |
11950 | /* wait until reshape finish */ | |
c85338c6 | 11951 | if (wait_for_reshape_imsm(sra, ndata)) { |
c47b0ff6 AK |
11952 | dprintf("wait_for_reshape_imsm returned error!\n"); |
11953 | goto abort; | |
11954 | } | |
84d11e6c N |
11955 | if (sigterm) |
11956 | goto abort; | |
10f22854 | 11957 | |
0228d92c AK |
11958 | if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) { |
11959 | /* ignore error == 2, this can mean end of reshape here | |
11960 | */ | |
7a862a02 | 11961 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL)\n"); |
10f22854 AK |
11962 | goto abort; |
11963 | } | |
11964 | ||
11965 | } | |
11966 | ||
71e5411e PB |
11967 | /* clear migr_rec on disks after successful migration */ |
11968 | struct dl *d; | |
11969 | ||
85337573 | 11970 | memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE); |
71e5411e PB |
11971 | for (d = super->disks; d; d = d->next) { |
11972 | if (d->index < 0 || is_failed(&d->disk)) | |
11973 | continue; | |
11974 | unsigned long long dsize; | |
11975 | ||
11976 | get_dev_size(d->fd, NULL, &dsize); | |
de44e46f | 11977 | if (lseek64(d->fd, dsize - MIGR_REC_SECTOR_POSITION*sector_size, |
71e5411e | 11978 | SEEK_SET) >= 0) { |
466070ad | 11979 | if ((unsigned int)write(d->fd, super->migr_rec_buf, |
de44e46f PB |
11980 | MIGR_REC_BUF_SECTORS*sector_size) != |
11981 | MIGR_REC_BUF_SECTORS*sector_size) | |
71e5411e PB |
11982 | perror("Write migr_rec failed"); |
11983 | } | |
11984 | } | |
11985 | ||
10f22854 AK |
11986 | /* return '1' if done */ |
11987 | ret_val = 1; | |
11988 | abort: | |
11989 | free(buf); | |
942e1cdb N |
11990 | /* See Grow.c: abort_reshape() for further explanation */ |
11991 | sysfs_set_num(sra, NULL, "suspend_lo", 0x7FFFFFFFFFFFFFFFULL); | |
11992 | sysfs_set_num(sra, NULL, "suspend_hi", 0); | |
11993 | sysfs_set_num(sra, NULL, "suspend_lo", 0); | |
10f22854 AK |
11994 | |
11995 | return ret_val; | |
999b4972 | 11996 | } |
0c21b485 | 11997 | |
cdddbdbc | 11998 | struct superswitch super_imsm = { |
cdddbdbc DW |
11999 | .examine_super = examine_super_imsm, |
12000 | .brief_examine_super = brief_examine_super_imsm, | |
4737ae25 | 12001 | .brief_examine_subarrays = brief_examine_subarrays_imsm, |
9d84c8ea | 12002 | .export_examine_super = export_examine_super_imsm, |
cdddbdbc DW |
12003 | .detail_super = detail_super_imsm, |
12004 | .brief_detail_super = brief_detail_super_imsm, | |
bf5a934a | 12005 | .write_init_super = write_init_super_imsm, |
0e600426 N |
12006 | .validate_geometry = validate_geometry_imsm, |
12007 | .add_to_super = add_to_super_imsm, | |
1a64be56 | 12008 | .remove_from_super = remove_from_super_imsm, |
d665cc31 | 12009 | .detail_platform = detail_platform_imsm, |
e50cf220 | 12010 | .export_detail_platform = export_detail_platform_imsm, |
33414a01 | 12011 | .kill_subarray = kill_subarray_imsm, |
aa534678 | 12012 | .update_subarray = update_subarray_imsm, |
2b959fbf | 12013 | .load_container = load_container_imsm, |
71204a50 N |
12014 | .default_geometry = default_geometry_imsm, |
12015 | .get_disk_controller_domain = imsm_get_disk_controller_domain, | |
12016 | .reshape_super = imsm_reshape_super, | |
12017 | .manage_reshape = imsm_manage_reshape, | |
9e2d750d | 12018 | .recover_backup = recover_backup_imsm, |
74db60b0 | 12019 | .copy_metadata = copy_metadata_imsm, |
27156a57 | 12020 | .examine_badblocks = examine_badblocks_imsm, |
cdddbdbc DW |
12021 | .match_home = match_home_imsm, |
12022 | .uuid_from_super= uuid_from_super_imsm, | |
12023 | .getinfo_super = getinfo_super_imsm, | |
5c4cd5da | 12024 | .getinfo_super_disks = getinfo_super_disks_imsm, |
cdddbdbc DW |
12025 | .update_super = update_super_imsm, |
12026 | ||
12027 | .avail_size = avail_size_imsm, | |
fbfdcb06 | 12028 | .get_spare_criteria = get_spare_criteria_imsm, |
cdddbdbc DW |
12029 | |
12030 | .compare_super = compare_super_imsm, | |
12031 | ||
12032 | .load_super = load_super_imsm, | |
bf5a934a | 12033 | .init_super = init_super_imsm, |
e683ca88 | 12034 | .store_super = store_super_imsm, |
cdddbdbc DW |
12035 | .free_super = free_super_imsm, |
12036 | .match_metadata_desc = match_metadata_desc_imsm, | |
bf5a934a | 12037 | .container_content = container_content_imsm, |
0c21b485 | 12038 | .validate_container = validate_container_imsm, |
cdddbdbc | 12039 | |
2432ce9b AP |
12040 | .write_init_ppl = write_init_ppl_imsm, |
12041 | .validate_ppl = validate_ppl_imsm, | |
12042 | ||
cdddbdbc | 12043 | .external = 1, |
4cce4069 | 12044 | .name = "imsm", |
845dea95 NB |
12045 | |
12046 | /* for mdmon */ | |
12047 | .open_new = imsm_open_new, | |
ed9d66aa | 12048 | .set_array_state= imsm_set_array_state, |
845dea95 NB |
12049 | .set_disk = imsm_set_disk, |
12050 | .sync_metadata = imsm_sync_metadata, | |
88758e9d | 12051 | .activate_spare = imsm_activate_spare, |
e8319a19 | 12052 | .process_update = imsm_process_update, |
8273f55e | 12053 | .prepare_update = imsm_prepare_update, |
6f50473f | 12054 | .record_bad_block = imsm_record_badblock, |
c07a5a4f | 12055 | .clear_bad_block = imsm_clear_badblock, |
928f1424 | 12056 | .get_bad_blocks = imsm_get_badblocks, |
cdddbdbc | 12057 | }; |