]>
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 | \ | |
418f9b36 N |
84 | MPB_ATTRIB_EXP_STRIPE_SIZE) |
85 | ||
86 | /* Define attributes that are unused but not harmful */ | |
87 | #define MPB_ATTRIB_IGNORED (MPB_ATTRIB_NEVER_USE) | |
fe7ed8cb | 88 | |
8e59f3d8 | 89 | #define MPB_SECTOR_CNT 2210 |
c2c087e6 | 90 | #define IMSM_RESERVED_SECTORS 4096 |
b81221b7 | 91 | #define NUM_BLOCKS_DIRTY_STRIPE_REGION 2056 |
979d38be | 92 | #define SECT_PER_MB_SHIFT 11 |
cdddbdbc DW |
93 | |
94 | /* Disk configuration info. */ | |
95 | #define IMSM_MAX_DEVICES 255 | |
96 | struct imsm_disk { | |
97 | __u8 serial[MAX_RAID_SERIAL_LEN];/* 0xD8 - 0xE7 ascii serial number */ | |
98 | __u32 total_blocks; /* 0xE8 - 0xEB total blocks */ | |
99 | __u32 scsi_id; /* 0xEC - 0xEF scsi ID */ | |
f2f27e63 DW |
100 | #define SPARE_DISK __cpu_to_le32(0x01) /* Spare */ |
101 | #define CONFIGURED_DISK __cpu_to_le32(0x02) /* Member of some RaidDev */ | |
102 | #define FAILED_DISK __cpu_to_le32(0x04) /* Permanent failure */ | |
cdddbdbc | 103 | __u32 status; /* 0xF0 - 0xF3 */ |
fe7ed8cb DW |
104 | __u32 owner_cfg_num; /* which config 0,1,2... owns this disk */ |
105 | #define IMSM_DISK_FILLERS 4 | |
cdddbdbc DW |
106 | __u32 filler[IMSM_DISK_FILLERS]; /* 0xF4 - 0x107 MPB_DISK_FILLERS for future expansion */ |
107 | }; | |
108 | ||
3b451610 AK |
109 | /* map selector for map managment |
110 | */ | |
238c0a71 AK |
111 | #define MAP_0 0 |
112 | #define MAP_1 1 | |
113 | #define MAP_X -1 | |
3b451610 | 114 | |
cdddbdbc DW |
115 | /* RAID map configuration infos. */ |
116 | struct imsm_map { | |
117 | __u32 pba_of_lba0; /* start address of partition */ | |
118 | __u32 blocks_per_member;/* blocks per member */ | |
119 | __u32 num_data_stripes; /* number of data stripes */ | |
120 | __u16 blocks_per_strip; | |
121 | __u8 map_state; /* Normal, Uninitialized, Degraded, Failed */ | |
122 | #define IMSM_T_STATE_NORMAL 0 | |
123 | #define IMSM_T_STATE_UNINITIALIZED 1 | |
e3bba0e0 DW |
124 | #define IMSM_T_STATE_DEGRADED 2 |
125 | #define IMSM_T_STATE_FAILED 3 | |
cdddbdbc DW |
126 | __u8 raid_level; |
127 | #define IMSM_T_RAID0 0 | |
128 | #define IMSM_T_RAID1 1 | |
129 | #define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */ | |
130 | __u8 num_members; /* number of member disks */ | |
fe7ed8cb DW |
131 | __u8 num_domains; /* number of parity domains */ |
132 | __u8 failed_disk_num; /* valid only when state is degraded */ | |
252d23c0 | 133 | __u8 ddf; |
cdddbdbc | 134 | __u32 filler[7]; /* expansion area */ |
7eef0453 | 135 | #define IMSM_ORD_REBUILD (1 << 24) |
cdddbdbc | 136 | __u32 disk_ord_tbl[1]; /* disk_ord_tbl[num_members], |
7eef0453 DW |
137 | * top byte contains some flags |
138 | */ | |
cdddbdbc DW |
139 | } __attribute__ ((packed)); |
140 | ||
141 | struct imsm_vol { | |
f8f603f1 | 142 | __u32 curr_migr_unit; |
fe7ed8cb | 143 | __u32 checkpoint_id; /* id to access curr_migr_unit */ |
cdddbdbc | 144 | __u8 migr_state; /* Normal or Migrating */ |
e3bba0e0 DW |
145 | #define MIGR_INIT 0 |
146 | #define MIGR_REBUILD 1 | |
147 | #define MIGR_VERIFY 2 /* analagous to echo check > sync_action */ | |
148 | #define MIGR_GEN_MIGR 3 | |
149 | #define MIGR_STATE_CHANGE 4 | |
1484e727 | 150 | #define MIGR_REPAIR 5 |
cdddbdbc DW |
151 | __u8 migr_type; /* Initializing, Rebuilding, ... */ |
152 | __u8 dirty; | |
fe7ed8cb DW |
153 | __u8 fs_state; /* fast-sync state for CnG (0xff == disabled) */ |
154 | __u16 verify_errors; /* number of mismatches */ | |
155 | __u16 bad_blocks; /* number of bad blocks during verify */ | |
156 | __u32 filler[4]; | |
cdddbdbc DW |
157 | struct imsm_map map[1]; |
158 | /* here comes another one if migr_state */ | |
159 | } __attribute__ ((packed)); | |
160 | ||
161 | struct imsm_dev { | |
fe7ed8cb | 162 | __u8 volume[MAX_RAID_SERIAL_LEN]; |
cdddbdbc DW |
163 | __u32 size_low; |
164 | __u32 size_high; | |
fe7ed8cb DW |
165 | #define DEV_BOOTABLE __cpu_to_le32(0x01) |
166 | #define DEV_BOOT_DEVICE __cpu_to_le32(0x02) | |
167 | #define DEV_READ_COALESCING __cpu_to_le32(0x04) | |
168 | #define DEV_WRITE_COALESCING __cpu_to_le32(0x08) | |
169 | #define DEV_LAST_SHUTDOWN_DIRTY __cpu_to_le32(0x10) | |
170 | #define DEV_HIDDEN_AT_BOOT __cpu_to_le32(0x20) | |
171 | #define DEV_CURRENTLY_HIDDEN __cpu_to_le32(0x40) | |
172 | #define DEV_VERIFY_AND_FIX __cpu_to_le32(0x80) | |
173 | #define DEV_MAP_STATE_UNINIT __cpu_to_le32(0x100) | |
174 | #define DEV_NO_AUTO_RECOVERY __cpu_to_le32(0x200) | |
175 | #define DEV_CLONE_N_GO __cpu_to_le32(0x400) | |
176 | #define DEV_CLONE_MAN_SYNC __cpu_to_le32(0x800) | |
177 | #define DEV_CNG_MASTER_DISK_NUM __cpu_to_le32(0x1000) | |
cdddbdbc DW |
178 | __u32 status; /* Persistent RaidDev status */ |
179 | __u32 reserved_blocks; /* Reserved blocks at beginning of volume */ | |
fe7ed8cb DW |
180 | __u8 migr_priority; |
181 | __u8 num_sub_vols; | |
182 | __u8 tid; | |
183 | __u8 cng_master_disk; | |
184 | __u16 cache_policy; | |
185 | __u8 cng_state; | |
186 | __u8 cng_sub_state; | |
187 | #define IMSM_DEV_FILLERS 10 | |
cdddbdbc DW |
188 | __u32 filler[IMSM_DEV_FILLERS]; |
189 | struct imsm_vol vol; | |
190 | } __attribute__ ((packed)); | |
191 | ||
192 | struct imsm_super { | |
193 | __u8 sig[MAX_SIGNATURE_LENGTH]; /* 0x00 - 0x1F */ | |
194 | __u32 check_sum; /* 0x20 - 0x23 MPB Checksum */ | |
195 | __u32 mpb_size; /* 0x24 - 0x27 Size of MPB */ | |
196 | __u32 family_num; /* 0x28 - 0x2B Checksum from first time this config was written */ | |
197 | __u32 generation_num; /* 0x2C - 0x2F Incremented each time this array's MPB is written */ | |
604b746f JD |
198 | __u32 error_log_size; /* 0x30 - 0x33 in bytes */ |
199 | __u32 attributes; /* 0x34 - 0x37 */ | |
cdddbdbc DW |
200 | __u8 num_disks; /* 0x38 Number of configured disks */ |
201 | __u8 num_raid_devs; /* 0x39 Number of configured volumes */ | |
604b746f JD |
202 | __u8 error_log_pos; /* 0x3A */ |
203 | __u8 fill[1]; /* 0x3B */ | |
204 | __u32 cache_size; /* 0x3c - 0x40 in mb */ | |
205 | __u32 orig_family_num; /* 0x40 - 0x43 original family num */ | |
206 | __u32 pwr_cycle_count; /* 0x44 - 0x47 simulated power cycle count for array */ | |
207 | __u32 bbm_log_size; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */ | |
208 | #define IMSM_FILLERS 35 | |
209 | __u32 filler[IMSM_FILLERS]; /* 0x4C - 0xD7 RAID_MPB_FILLERS */ | |
cdddbdbc DW |
210 | struct imsm_disk disk[1]; /* 0xD8 diskTbl[numDisks] */ |
211 | /* here comes imsm_dev[num_raid_devs] */ | |
604b746f | 212 | /* here comes BBM logs */ |
cdddbdbc DW |
213 | } __attribute__ ((packed)); |
214 | ||
604b746f JD |
215 | #define BBM_LOG_MAX_ENTRIES 254 |
216 | ||
217 | struct bbm_log_entry { | |
218 | __u64 defective_block_start; | |
219 | #define UNREADABLE 0xFFFFFFFF | |
220 | __u32 spare_block_offset; | |
221 | __u16 remapped_marked_count; | |
222 | __u16 disk_ordinal; | |
223 | } __attribute__ ((__packed__)); | |
224 | ||
225 | struct bbm_log { | |
226 | __u32 signature; /* 0xABADB10C */ | |
227 | __u32 entry_count; | |
228 | __u32 reserved_spare_block_count; /* 0 */ | |
229 | __u32 reserved; /* 0xFFFF */ | |
230 | __u64 first_spare_lba; | |
231 | struct bbm_log_entry mapped_block_entries[BBM_LOG_MAX_ENTRIES]; | |
232 | } __attribute__ ((__packed__)); | |
233 | ||
234 | ||
cdddbdbc DW |
235 | #ifndef MDASSEMBLE |
236 | static char *map_state_str[] = { "normal", "uninitialized", "degraded", "failed" }; | |
237 | #endif | |
238 | ||
8e59f3d8 AK |
239 | #define RAID_DISK_RESERVED_BLOCKS_IMSM_HI 2209 |
240 | ||
241 | #define GEN_MIGR_AREA_SIZE 2048 /* General Migration Copy Area size in blocks */ | |
242 | ||
17a4eaf9 AK |
243 | #define MIGR_REC_BUF_SIZE 512 /* size of migr_record i/o buffer */ |
244 | #define MIGR_REC_POSITION 512 /* migr_record position offset on disk, | |
245 | * MIGR_REC_BUF_SIZE <= MIGR_REC_POSITION | |
246 | */ | |
247 | ||
248 | ||
8e59f3d8 AK |
249 | #define UNIT_SRC_NORMAL 0 /* Source data for curr_migr_unit must |
250 | * be recovered using srcMap */ | |
251 | #define UNIT_SRC_IN_CP_AREA 1 /* Source data for curr_migr_unit has | |
252 | * already been migrated and must | |
253 | * be recovered from checkpoint area */ | |
254 | struct migr_record { | |
255 | __u32 rec_status; /* Status used to determine how to restart | |
256 | * migration in case it aborts | |
257 | * in some fashion */ | |
258 | __u32 curr_migr_unit; /* 0..numMigrUnits-1 */ | |
259 | __u32 family_num; /* Family number of MPB | |
260 | * containing the RaidDev | |
261 | * that is migrating */ | |
262 | __u32 ascending_migr; /* True if migrating in increasing | |
263 | * order of lbas */ | |
264 | __u32 blocks_per_unit; /* Num disk blocks per unit of operation */ | |
265 | __u32 dest_depth_per_unit; /* Num member blocks each destMap | |
266 | * member disk | |
267 | * advances per unit-of-operation */ | |
268 | __u32 ckpt_area_pba; /* Pba of first block of ckpt copy area */ | |
269 | __u32 dest_1st_member_lba; /* First member lba on first | |
270 | * stripe of destination */ | |
271 | __u32 num_migr_units; /* Total num migration units-of-op */ | |
272 | __u32 post_migr_vol_cap; /* Size of volume after | |
273 | * migration completes */ | |
274 | __u32 post_migr_vol_cap_hi; /* Expansion space for LBA64 */ | |
275 | __u32 ckpt_read_disk_num; /* Which member disk in destSubMap[0] the | |
276 | * migration ckpt record was read from | |
277 | * (for recovered migrations) */ | |
278 | } __attribute__ ((__packed__)); | |
279 | ||
ec50f7b6 LM |
280 | struct md_list { |
281 | /* usage marker: | |
282 | * 1: load metadata | |
283 | * 2: metadata does not match | |
284 | * 4: already checked | |
285 | */ | |
286 | int used; | |
287 | char *devname; | |
288 | int found; | |
289 | int container; | |
290 | dev_t st_rdev; | |
291 | struct md_list *next; | |
292 | }; | |
293 | ||
294 | #define pr_vrb(fmt, arg...) (void) (verbose && fprintf(stderr, Name fmt, ##arg)) | |
295 | ||
1484e727 DW |
296 | static __u8 migr_type(struct imsm_dev *dev) |
297 | { | |
298 | if (dev->vol.migr_type == MIGR_VERIFY && | |
299 | dev->status & DEV_VERIFY_AND_FIX) | |
300 | return MIGR_REPAIR; | |
301 | else | |
302 | return dev->vol.migr_type; | |
303 | } | |
304 | ||
305 | static void set_migr_type(struct imsm_dev *dev, __u8 migr_type) | |
306 | { | |
307 | /* for compatibility with older oroms convert MIGR_REPAIR, into | |
308 | * MIGR_VERIFY w/ DEV_VERIFY_AND_FIX status | |
309 | */ | |
310 | if (migr_type == MIGR_REPAIR) { | |
311 | dev->vol.migr_type = MIGR_VERIFY; | |
312 | dev->status |= DEV_VERIFY_AND_FIX; | |
313 | } else { | |
314 | dev->vol.migr_type = migr_type; | |
315 | dev->status &= ~DEV_VERIFY_AND_FIX; | |
316 | } | |
317 | } | |
318 | ||
87eb16df | 319 | static unsigned int sector_count(__u32 bytes) |
cdddbdbc | 320 | { |
87eb16df DW |
321 | return ((bytes + (512-1)) & (~(512-1))) / 512; |
322 | } | |
cdddbdbc | 323 | |
87eb16df DW |
324 | static unsigned int mpb_sectors(struct imsm_super *mpb) |
325 | { | |
326 | return sector_count(__le32_to_cpu(mpb->mpb_size)); | |
cdddbdbc DW |
327 | } |
328 | ||
ba2de7ba DW |
329 | struct intel_dev { |
330 | struct imsm_dev *dev; | |
331 | struct intel_dev *next; | |
f21e18ca | 332 | unsigned index; |
ba2de7ba DW |
333 | }; |
334 | ||
88654014 LM |
335 | struct intel_hba { |
336 | enum sys_dev_type type; | |
337 | char *path; | |
338 | char *pci_id; | |
339 | struct intel_hba *next; | |
340 | }; | |
341 | ||
1a64be56 LM |
342 | enum action { |
343 | DISK_REMOVE = 1, | |
344 | DISK_ADD | |
345 | }; | |
cdddbdbc DW |
346 | /* internal representation of IMSM metadata */ |
347 | struct intel_super { | |
348 | union { | |
949c47a0 DW |
349 | void *buf; /* O_DIRECT buffer for reading/writing metadata */ |
350 | struct imsm_super *anchor; /* immovable parameters */ | |
cdddbdbc | 351 | }; |
8e59f3d8 AK |
352 | union { |
353 | void *migr_rec_buf; /* buffer for I/O operations */ | |
354 | struct migr_record *migr_rec; /* migration record */ | |
355 | }; | |
949c47a0 | 356 | size_t len; /* size of the 'buf' allocation */ |
4d7b1503 DW |
357 | void *next_buf; /* for realloc'ing buf from the manager */ |
358 | size_t next_len; | |
c2c087e6 | 359 | int updates_pending; /* count of pending updates for mdmon */ |
bf5a934a | 360 | int current_vol; /* index of raid device undergoing creation */ |
0dcecb2e | 361 | __u32 create_offset; /* common start for 'current_vol' */ |
148acb7b | 362 | __u32 random; /* random data for seeding new family numbers */ |
ba2de7ba | 363 | struct intel_dev *devlist; |
cdddbdbc DW |
364 | struct dl { |
365 | struct dl *next; | |
366 | int index; | |
367 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
368 | int major, minor; | |
369 | char *devname; | |
b9f594fe | 370 | struct imsm_disk disk; |
cdddbdbc | 371 | int fd; |
0dcecb2e DW |
372 | int extent_cnt; |
373 | struct extent *e; /* for determining freespace @ create */ | |
efb30e7f | 374 | int raiddisk; /* slot to fill in autolayout */ |
1a64be56 | 375 | enum action action; |
ca0748fa | 376 | } *disks, *current_disk; |
1a64be56 LM |
377 | struct dl *disk_mgmt_list; /* list of disks to add/remove while mdmon |
378 | active */ | |
47ee5a45 | 379 | struct dl *missing; /* disks removed while we weren't looking */ |
43dad3d6 | 380 | struct bbm_log *bbm_log; |
88654014 | 381 | struct intel_hba *hba; /* device path of the raid controller for this metadata */ |
88c32bb1 | 382 | const struct imsm_orom *orom; /* platform firmware support */ |
a2b97981 DW |
383 | struct intel_super *next; /* (temp) list for disambiguating family_num */ |
384 | }; | |
385 | ||
386 | struct intel_disk { | |
387 | struct imsm_disk disk; | |
388 | #define IMSM_UNKNOWN_OWNER (-1) | |
389 | int owner; | |
390 | struct intel_disk *next; | |
cdddbdbc DW |
391 | }; |
392 | ||
c2c087e6 DW |
393 | struct extent { |
394 | unsigned long long start, size; | |
395 | }; | |
396 | ||
694575e7 KW |
397 | /* definitions of reshape process types */ |
398 | enum imsm_reshape_type { | |
399 | CH_TAKEOVER, | |
b5347799 | 400 | CH_MIGRATION, |
694575e7 KW |
401 | }; |
402 | ||
88758e9d DW |
403 | /* definition of messages passed to imsm_process_update */ |
404 | enum imsm_update_type { | |
405 | update_activate_spare, | |
8273f55e | 406 | update_create_array, |
33414a01 | 407 | update_kill_array, |
aa534678 | 408 | update_rename_array, |
1a64be56 | 409 | update_add_remove_disk, |
78b10e66 | 410 | update_reshape_container_disks, |
48c5303a | 411 | update_reshape_migration, |
2d40f3a1 AK |
412 | update_takeover, |
413 | update_general_migration_checkpoint, | |
88758e9d DW |
414 | }; |
415 | ||
416 | struct imsm_update_activate_spare { | |
417 | enum imsm_update_type type; | |
d23fe947 | 418 | struct dl *dl; |
88758e9d DW |
419 | int slot; |
420 | int array; | |
421 | struct imsm_update_activate_spare *next; | |
422 | }; | |
423 | ||
78b10e66 N |
424 | struct geo_params { |
425 | int dev_id; | |
426 | char *dev_name; | |
427 | long long size; | |
428 | int level; | |
429 | int layout; | |
430 | int chunksize; | |
431 | int raid_disks; | |
432 | }; | |
433 | ||
bb025c2f KW |
434 | enum takeover_direction { |
435 | R10_TO_R0, | |
436 | R0_TO_R10 | |
437 | }; | |
438 | struct imsm_update_takeover { | |
439 | enum imsm_update_type type; | |
440 | int subarray; | |
441 | enum takeover_direction direction; | |
442 | }; | |
78b10e66 N |
443 | |
444 | struct imsm_update_reshape { | |
445 | enum imsm_update_type type; | |
446 | int old_raid_disks; | |
447 | int new_raid_disks; | |
48c5303a PC |
448 | |
449 | int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */ | |
450 | }; | |
451 | ||
452 | struct imsm_update_reshape_migration { | |
453 | enum imsm_update_type type; | |
454 | int old_raid_disks; | |
455 | int new_raid_disks; | |
456 | /* fields for array migration changes | |
457 | */ | |
458 | int subdev; | |
459 | int new_level; | |
460 | int new_layout; | |
4bba0439 | 461 | int new_chunksize; |
48c5303a | 462 | |
d195167d | 463 | int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */ |
78b10e66 N |
464 | }; |
465 | ||
2d40f3a1 AK |
466 | struct imsm_update_general_migration_checkpoint { |
467 | enum imsm_update_type type; | |
468 | __u32 curr_migr_unit; | |
469 | }; | |
470 | ||
54c2c1ea DW |
471 | struct disk_info { |
472 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
473 | }; | |
474 | ||
8273f55e DW |
475 | struct imsm_update_create_array { |
476 | enum imsm_update_type type; | |
8273f55e | 477 | int dev_idx; |
6a3e913e | 478 | struct imsm_dev dev; |
8273f55e DW |
479 | }; |
480 | ||
33414a01 DW |
481 | struct imsm_update_kill_array { |
482 | enum imsm_update_type type; | |
483 | int dev_idx; | |
484 | }; | |
485 | ||
aa534678 DW |
486 | struct imsm_update_rename_array { |
487 | enum imsm_update_type type; | |
488 | __u8 name[MAX_RAID_SERIAL_LEN]; | |
489 | int dev_idx; | |
490 | }; | |
491 | ||
1a64be56 | 492 | struct imsm_update_add_remove_disk { |
43dad3d6 DW |
493 | enum imsm_update_type type; |
494 | }; | |
495 | ||
88654014 LM |
496 | |
497 | static const char *_sys_dev_type[] = { | |
498 | [SYS_DEV_UNKNOWN] = "Unknown", | |
499 | [SYS_DEV_SAS] = "SAS", | |
500 | [SYS_DEV_SATA] = "SATA" | |
501 | }; | |
502 | ||
503 | const char *get_sys_dev_type(enum sys_dev_type type) | |
504 | { | |
505 | if (type >= SYS_DEV_MAX) | |
506 | type = SYS_DEV_UNKNOWN; | |
507 | ||
508 | return _sys_dev_type[type]; | |
509 | } | |
510 | ||
511 | static struct intel_hba * alloc_intel_hba(struct sys_dev *device) | |
512 | { | |
513 | struct intel_hba *result = malloc(sizeof(*result)); | |
514 | if (result) { | |
515 | result->type = device->type; | |
516 | result->path = strdup(device->path); | |
517 | result->next = NULL; | |
518 | if (result->path && (result->pci_id = strrchr(result->path, '/')) != NULL) | |
519 | result->pci_id++; | |
520 | } | |
521 | return result; | |
522 | } | |
523 | ||
524 | static struct intel_hba * find_intel_hba(struct intel_hba *hba, struct sys_dev *device) | |
525 | { | |
526 | struct intel_hba *result=NULL; | |
527 | for (result = hba; result; result = result->next) { | |
528 | if (result->type == device->type && strcmp(result->path, device->path) == 0) | |
529 | break; | |
530 | } | |
531 | return result; | |
532 | } | |
533 | ||
b4cf4cba | 534 | static int attach_hba_to_super(struct intel_super *super, struct sys_dev *device) |
88654014 LM |
535 | { |
536 | struct intel_hba *hba; | |
537 | ||
538 | /* check if disk attached to Intel HBA */ | |
539 | hba = find_intel_hba(super->hba, device); | |
540 | if (hba != NULL) | |
541 | return 1; | |
542 | /* Check if HBA is already attached to super */ | |
543 | if (super->hba == NULL) { | |
544 | super->hba = alloc_intel_hba(device); | |
545 | return 1; | |
546 | } | |
547 | ||
548 | hba = super->hba; | |
549 | /* Intel metadata allows for all disks attached to the same type HBA. | |
550 | * Do not sypport odf HBA types mixing | |
551 | */ | |
552 | if (device->type != hba->type) | |
553 | return 2; | |
554 | ||
555 | while (hba->next) | |
556 | hba = hba->next; | |
557 | ||
558 | hba->next = alloc_intel_hba(device); | |
559 | return 1; | |
560 | } | |
561 | ||
562 | static struct sys_dev* find_disk_attached_hba(int fd, const char *devname) | |
563 | { | |
564 | struct sys_dev *list, *elem, *prev; | |
565 | char *disk_path; | |
566 | ||
567 | if ((list = find_intel_devices()) == NULL) | |
568 | return 0; | |
569 | ||
570 | if (fd < 0) | |
571 | disk_path = (char *) devname; | |
572 | else | |
573 | disk_path = diskfd_to_devpath(fd); | |
574 | ||
575 | if (!disk_path) { | |
576 | free_sys_dev(&list); | |
577 | return 0; | |
578 | } | |
579 | ||
580 | for (prev = NULL, elem = list; elem; prev = elem, elem = elem->next) { | |
581 | if (path_attached_to_hba(disk_path, elem->path)) { | |
582 | if (prev == NULL) | |
583 | list = list->next; | |
584 | else | |
585 | prev->next = elem->next; | |
586 | elem->next = NULL; | |
587 | if (disk_path != devname) | |
588 | free(disk_path); | |
589 | free_sys_dev(&list); | |
590 | return elem; | |
591 | } | |
592 | } | |
593 | if (disk_path != devname) | |
594 | free(disk_path); | |
595 | free_sys_dev(&list); | |
596 | ||
597 | return NULL; | |
598 | } | |
599 | ||
600 | ||
d424212e N |
601 | static int find_intel_hba_capability(int fd, struct intel_super *super, |
602 | char *devname); | |
f2f5c343 | 603 | |
cdddbdbc DW |
604 | static struct supertype *match_metadata_desc_imsm(char *arg) |
605 | { | |
606 | struct supertype *st; | |
607 | ||
608 | if (strcmp(arg, "imsm") != 0 && | |
609 | strcmp(arg, "default") != 0 | |
610 | ) | |
611 | return NULL; | |
612 | ||
613 | st = malloc(sizeof(*st)); | |
4e9d2186 AW |
614 | if (!st) |
615 | return NULL; | |
ef609477 | 616 | memset(st, 0, sizeof(*st)); |
d1d599ea | 617 | st->container_dev = NoMdDev; |
cdddbdbc DW |
618 | st->ss = &super_imsm; |
619 | st->max_devs = IMSM_MAX_DEVICES; | |
620 | st->minor_version = 0; | |
621 | st->sb = NULL; | |
622 | return st; | |
623 | } | |
624 | ||
0e600426 | 625 | #ifndef MDASSEMBLE |
cdddbdbc DW |
626 | static __u8 *get_imsm_version(struct imsm_super *mpb) |
627 | { | |
628 | return &mpb->sig[MPB_SIG_LEN]; | |
629 | } | |
9e2d750d | 630 | #endif |
cdddbdbc | 631 | |
949c47a0 DW |
632 | /* retrieve a disk directly from the anchor when the anchor is known to be |
633 | * up-to-date, currently only at load time | |
634 | */ | |
635 | static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index) | |
cdddbdbc | 636 | { |
949c47a0 | 637 | if (index >= mpb->num_disks) |
cdddbdbc DW |
638 | return NULL; |
639 | return &mpb->disk[index]; | |
640 | } | |
641 | ||
95d07a2c LM |
642 | /* retrieve the disk description based on a index of the disk |
643 | * in the sub-array | |
644 | */ | |
645 | static struct dl *get_imsm_dl_disk(struct intel_super *super, __u8 index) | |
949c47a0 | 646 | { |
b9f594fe DW |
647 | struct dl *d; |
648 | ||
649 | for (d = super->disks; d; d = d->next) | |
650 | if (d->index == index) | |
95d07a2c LM |
651 | return d; |
652 | ||
653 | return NULL; | |
654 | } | |
655 | /* retrieve a disk from the parsed metadata */ | |
656 | static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index) | |
657 | { | |
658 | struct dl *dl; | |
659 | ||
660 | dl = get_imsm_dl_disk(super, index); | |
661 | if (dl) | |
662 | return &dl->disk; | |
663 | ||
b9f594fe | 664 | return NULL; |
949c47a0 DW |
665 | } |
666 | ||
667 | /* generate a checksum directly from the anchor when the anchor is known to be | |
668 | * up-to-date, currently only at load or write_super after coalescing | |
669 | */ | |
670 | static __u32 __gen_imsm_checksum(struct imsm_super *mpb) | |
cdddbdbc DW |
671 | { |
672 | __u32 end = mpb->mpb_size / sizeof(end); | |
673 | __u32 *p = (__u32 *) mpb; | |
674 | __u32 sum = 0; | |
675 | ||
97f734fd N |
676 | while (end--) { |
677 | sum += __le32_to_cpu(*p); | |
678 | p++; | |
679 | } | |
cdddbdbc DW |
680 | |
681 | return sum - __le32_to_cpu(mpb->check_sum); | |
682 | } | |
683 | ||
a965f303 DW |
684 | static size_t sizeof_imsm_map(struct imsm_map *map) |
685 | { | |
686 | return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1); | |
687 | } | |
688 | ||
689 | struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map) | |
cdddbdbc | 690 | { |
5e7b0330 AK |
691 | /* A device can have 2 maps if it is in the middle of a migration. |
692 | * If second_map is: | |
238c0a71 AK |
693 | * MAP_0 - we return the first map |
694 | * MAP_1 - we return the second map if it exists, else NULL | |
695 | * MAP_X - we return the second map if it exists, else the first | |
5e7b0330 | 696 | */ |
a965f303 | 697 | struct imsm_map *map = &dev->vol.map[0]; |
9535fc47 | 698 | struct imsm_map *map2 = NULL; |
a965f303 | 699 | |
9535fc47 AK |
700 | if (dev->vol.migr_state) |
701 | map2 = (void *)map + sizeof_imsm_map(map); | |
a965f303 | 702 | |
9535fc47 | 703 | switch (second_map) { |
3b451610 | 704 | case MAP_0: |
9535fc47 | 705 | break; |
3b451610 | 706 | case MAP_1: |
9535fc47 AK |
707 | map = map2; |
708 | break; | |
238c0a71 | 709 | case MAP_X: |
9535fc47 AK |
710 | if (map2) |
711 | map = map2; | |
712 | break; | |
9535fc47 AK |
713 | default: |
714 | map = NULL; | |
715 | } | |
716 | return map; | |
5e7b0330 | 717 | |
a965f303 | 718 | } |
cdddbdbc | 719 | |
3393c6af DW |
720 | /* return the size of the device. |
721 | * migr_state increases the returned size if map[0] were to be duplicated | |
722 | */ | |
723 | static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state) | |
a965f303 DW |
724 | { |
725 | size_t size = sizeof(*dev) - sizeof(struct imsm_map) + | |
238c0a71 | 726 | sizeof_imsm_map(get_imsm_map(dev, MAP_0)); |
cdddbdbc DW |
727 | |
728 | /* migrating means an additional map */ | |
a965f303 | 729 | if (dev->vol.migr_state) |
238c0a71 | 730 | size += sizeof_imsm_map(get_imsm_map(dev, MAP_1)); |
3393c6af | 731 | else if (migr_state) |
238c0a71 | 732 | size += sizeof_imsm_map(get_imsm_map(dev, MAP_0)); |
cdddbdbc DW |
733 | |
734 | return size; | |
735 | } | |
736 | ||
54c2c1ea DW |
737 | #ifndef MDASSEMBLE |
738 | /* retrieve disk serial number list from a metadata update */ | |
739 | static struct disk_info *get_disk_info(struct imsm_update_create_array *update) | |
740 | { | |
741 | void *u = update; | |
742 | struct disk_info *inf; | |
743 | ||
744 | inf = u + sizeof(*update) - sizeof(struct imsm_dev) + | |
745 | sizeof_imsm_dev(&update->dev, 0); | |
746 | ||
747 | return inf; | |
748 | } | |
749 | #endif | |
750 | ||
949c47a0 | 751 | static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index) |
cdddbdbc DW |
752 | { |
753 | int offset; | |
754 | int i; | |
755 | void *_mpb = mpb; | |
756 | ||
949c47a0 | 757 | if (index >= mpb->num_raid_devs) |
cdddbdbc DW |
758 | return NULL; |
759 | ||
760 | /* devices start after all disks */ | |
761 | offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb; | |
762 | ||
763 | for (i = 0; i <= index; i++) | |
764 | if (i == index) | |
765 | return _mpb + offset; | |
766 | else | |
3393c6af | 767 | offset += sizeof_imsm_dev(_mpb + offset, 0); |
cdddbdbc DW |
768 | |
769 | return NULL; | |
770 | } | |
771 | ||
949c47a0 DW |
772 | static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index) |
773 | { | |
ba2de7ba DW |
774 | struct intel_dev *dv; |
775 | ||
949c47a0 DW |
776 | if (index >= super->anchor->num_raid_devs) |
777 | return NULL; | |
ba2de7ba DW |
778 | for (dv = super->devlist; dv; dv = dv->next) |
779 | if (dv->index == index) | |
780 | return dv->dev; | |
781 | return NULL; | |
949c47a0 DW |
782 | } |
783 | ||
98130f40 AK |
784 | /* |
785 | * for second_map: | |
238c0a71 AK |
786 | * == MAP_0 get first map |
787 | * == MAP_1 get second map | |
788 | * == MAP_X than get map according to the current migr_state | |
98130f40 AK |
789 | */ |
790 | static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev, | |
791 | int slot, | |
792 | int second_map) | |
7eef0453 DW |
793 | { |
794 | struct imsm_map *map; | |
795 | ||
5e7b0330 | 796 | map = get_imsm_map(dev, second_map); |
7eef0453 | 797 | |
ff077194 DW |
798 | /* top byte identifies disk under rebuild */ |
799 | return __le32_to_cpu(map->disk_ord_tbl[slot]); | |
800 | } | |
801 | ||
802 | #define ord_to_idx(ord) (((ord) << 8) >> 8) | |
98130f40 | 803 | static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot, int second_map) |
ff077194 | 804 | { |
98130f40 | 805 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, second_map); |
ff077194 DW |
806 | |
807 | return ord_to_idx(ord); | |
7eef0453 DW |
808 | } |
809 | ||
be73972f DW |
810 | static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord) |
811 | { | |
812 | map->disk_ord_tbl[slot] = __cpu_to_le32(ord); | |
813 | } | |
814 | ||
f21e18ca | 815 | static int get_imsm_disk_slot(struct imsm_map *map, unsigned idx) |
620b1713 DW |
816 | { |
817 | int slot; | |
818 | __u32 ord; | |
819 | ||
820 | for (slot = 0; slot < map->num_members; slot++) { | |
821 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
822 | if (ord_to_idx(ord) == idx) | |
823 | return slot; | |
824 | } | |
825 | ||
826 | return -1; | |
827 | } | |
828 | ||
cdddbdbc DW |
829 | static int get_imsm_raid_level(struct imsm_map *map) |
830 | { | |
831 | if (map->raid_level == 1) { | |
832 | if (map->num_members == 2) | |
833 | return 1; | |
834 | else | |
835 | return 10; | |
836 | } | |
837 | ||
838 | return map->raid_level; | |
839 | } | |
840 | ||
c2c087e6 DW |
841 | static int cmp_extent(const void *av, const void *bv) |
842 | { | |
843 | const struct extent *a = av; | |
844 | const struct extent *b = bv; | |
845 | if (a->start < b->start) | |
846 | return -1; | |
847 | if (a->start > b->start) | |
848 | return 1; | |
849 | return 0; | |
850 | } | |
851 | ||
0dcecb2e | 852 | static int count_memberships(struct dl *dl, struct intel_super *super) |
c2c087e6 | 853 | { |
c2c087e6 | 854 | int memberships = 0; |
620b1713 | 855 | int i; |
c2c087e6 | 856 | |
949c47a0 DW |
857 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
858 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 859 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
c2c087e6 | 860 | |
620b1713 DW |
861 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
862 | memberships++; | |
c2c087e6 | 863 | } |
0dcecb2e DW |
864 | |
865 | return memberships; | |
866 | } | |
867 | ||
b81221b7 CA |
868 | static __u32 imsm_min_reserved_sectors(struct intel_super *super); |
869 | ||
0dcecb2e DW |
870 | static struct extent *get_extents(struct intel_super *super, struct dl *dl) |
871 | { | |
872 | /* find a list of used extents on the given physical device */ | |
873 | struct extent *rv, *e; | |
620b1713 | 874 | int i; |
0dcecb2e | 875 | int memberships = count_memberships(dl, super); |
b276dd33 DW |
876 | __u32 reservation; |
877 | ||
878 | /* trim the reserved area for spares, so they can join any array | |
879 | * regardless of whether the OROM has assigned sectors from the | |
880 | * IMSM_RESERVED_SECTORS region | |
881 | */ | |
882 | if (dl->index == -1) | |
b81221b7 | 883 | reservation = imsm_min_reserved_sectors(super); |
b276dd33 DW |
884 | else |
885 | reservation = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
0dcecb2e | 886 | |
c2c087e6 DW |
887 | rv = malloc(sizeof(struct extent) * (memberships + 1)); |
888 | if (!rv) | |
889 | return NULL; | |
890 | e = rv; | |
891 | ||
949c47a0 DW |
892 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
893 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 894 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
c2c087e6 | 895 | |
620b1713 DW |
896 | if (get_imsm_disk_slot(map, dl->index) >= 0) { |
897 | e->start = __le32_to_cpu(map->pba_of_lba0); | |
898 | e->size = __le32_to_cpu(map->blocks_per_member); | |
899 | e++; | |
c2c087e6 DW |
900 | } |
901 | } | |
902 | qsort(rv, memberships, sizeof(*rv), cmp_extent); | |
903 | ||
14e8215b DW |
904 | /* determine the start of the metadata |
905 | * when no raid devices are defined use the default | |
906 | * ...otherwise allow the metadata to truncate the value | |
907 | * as is the case with older versions of imsm | |
908 | */ | |
909 | if (memberships) { | |
910 | struct extent *last = &rv[memberships - 1]; | |
911 | __u32 remainder; | |
912 | ||
913 | remainder = __le32_to_cpu(dl->disk.total_blocks) - | |
914 | (last->start + last->size); | |
dda5855f DW |
915 | /* round down to 1k block to satisfy precision of the kernel |
916 | * 'size' interface | |
917 | */ | |
918 | remainder &= ~1UL; | |
919 | /* make sure remainder is still sane */ | |
f21e18ca | 920 | if (remainder < (unsigned)ROUND_UP(super->len, 512) >> 9) |
dda5855f | 921 | remainder = ROUND_UP(super->len, 512) >> 9; |
14e8215b DW |
922 | if (reservation > remainder) |
923 | reservation = remainder; | |
924 | } | |
925 | e->start = __le32_to_cpu(dl->disk.total_blocks) - reservation; | |
c2c087e6 DW |
926 | e->size = 0; |
927 | return rv; | |
928 | } | |
929 | ||
14e8215b DW |
930 | /* try to determine how much space is reserved for metadata from |
931 | * the last get_extents() entry, otherwise fallback to the | |
932 | * default | |
933 | */ | |
934 | static __u32 imsm_reserved_sectors(struct intel_super *super, struct dl *dl) | |
935 | { | |
936 | struct extent *e; | |
937 | int i; | |
938 | __u32 rv; | |
939 | ||
940 | /* for spares just return a minimal reservation which will grow | |
941 | * once the spare is picked up by an array | |
942 | */ | |
943 | if (dl->index == -1) | |
944 | return MPB_SECTOR_CNT; | |
945 | ||
946 | e = get_extents(super, dl); | |
947 | if (!e) | |
948 | return MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
949 | ||
950 | /* scroll to last entry */ | |
951 | for (i = 0; e[i].size; i++) | |
952 | continue; | |
953 | ||
954 | rv = __le32_to_cpu(dl->disk.total_blocks) - e[i].start; | |
955 | ||
956 | free(e); | |
957 | ||
958 | return rv; | |
959 | } | |
960 | ||
25ed7e59 DW |
961 | static int is_spare(struct imsm_disk *disk) |
962 | { | |
963 | return (disk->status & SPARE_DISK) == SPARE_DISK; | |
964 | } | |
965 | ||
966 | static int is_configured(struct imsm_disk *disk) | |
967 | { | |
968 | return (disk->status & CONFIGURED_DISK) == CONFIGURED_DISK; | |
969 | } | |
970 | ||
971 | static int is_failed(struct imsm_disk *disk) | |
972 | { | |
973 | return (disk->status & FAILED_DISK) == FAILED_DISK; | |
974 | } | |
975 | ||
b81221b7 CA |
976 | /* try to determine how much space is reserved for metadata from |
977 | * the last get_extents() entry on the smallest active disk, | |
978 | * otherwise fallback to the default | |
979 | */ | |
980 | static __u32 imsm_min_reserved_sectors(struct intel_super *super) | |
981 | { | |
982 | struct extent *e; | |
983 | int i; | |
984 | __u32 min_active, remainder; | |
985 | __u32 rv = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
986 | struct dl *dl, *dl_min = NULL; | |
987 | ||
988 | if (!super) | |
989 | return rv; | |
990 | ||
991 | min_active = 0; | |
992 | for (dl = super->disks; dl; dl = dl->next) { | |
993 | if (dl->index < 0) | |
994 | continue; | |
995 | if (dl->disk.total_blocks < min_active || min_active == 0) { | |
996 | dl_min = dl; | |
997 | min_active = dl->disk.total_blocks; | |
998 | } | |
999 | } | |
1000 | if (!dl_min) | |
1001 | return rv; | |
1002 | ||
1003 | /* find last lba used by subarrays on the smallest active disk */ | |
1004 | e = get_extents(super, dl_min); | |
1005 | if (!e) | |
1006 | return rv; | |
1007 | for (i = 0; e[i].size; i++) | |
1008 | continue; | |
1009 | ||
1010 | remainder = min_active - e[i].start; | |
1011 | free(e); | |
1012 | ||
1013 | /* to give priority to recovery we should not require full | |
1014 | IMSM_RESERVED_SECTORS from the spare */ | |
1015 | rv = MPB_SECTOR_CNT + NUM_BLOCKS_DIRTY_STRIPE_REGION; | |
1016 | ||
1017 | /* if real reservation is smaller use that value */ | |
1018 | return (remainder < rv) ? remainder : rv; | |
1019 | } | |
1020 | ||
80e7f8c3 AC |
1021 | /* Return minimum size of a spare that can be used in this array*/ |
1022 | static unsigned long long min_acceptable_spare_size_imsm(struct supertype *st) | |
1023 | { | |
1024 | struct intel_super *super = st->sb; | |
1025 | struct dl *dl; | |
1026 | struct extent *e; | |
1027 | int i; | |
1028 | unsigned long long rv = 0; | |
1029 | ||
1030 | if (!super) | |
1031 | return rv; | |
1032 | /* find first active disk in array */ | |
1033 | dl = super->disks; | |
1034 | while (dl && (is_failed(&dl->disk) || dl->index == -1)) | |
1035 | dl = dl->next; | |
1036 | if (!dl) | |
1037 | return rv; | |
1038 | /* find last lba used by subarrays */ | |
1039 | e = get_extents(super, dl); | |
1040 | if (!e) | |
1041 | return rv; | |
1042 | for (i = 0; e[i].size; i++) | |
1043 | continue; | |
1044 | if (i > 0) | |
1045 | rv = e[i-1].start + e[i-1].size; | |
1046 | free(e); | |
b81221b7 | 1047 | |
80e7f8c3 | 1048 | /* add the amount of space needed for metadata */ |
b81221b7 CA |
1049 | rv = rv + imsm_min_reserved_sectors(super); |
1050 | ||
80e7f8c3 AC |
1051 | return rv * 512; |
1052 | } | |
1053 | ||
d1e02575 AK |
1054 | static int is_gen_migration(struct imsm_dev *dev); |
1055 | ||
1799c9e8 | 1056 | #ifndef MDASSEMBLE |
c47b0ff6 AK |
1057 | static __u64 blocks_per_migr_unit(struct intel_super *super, |
1058 | struct imsm_dev *dev); | |
1e5c6983 | 1059 | |
c47b0ff6 AK |
1060 | static void print_imsm_dev(struct intel_super *super, |
1061 | struct imsm_dev *dev, | |
1062 | char *uuid, | |
1063 | int disk_idx) | |
cdddbdbc DW |
1064 | { |
1065 | __u64 sz; | |
0d80bb2f | 1066 | int slot, i; |
238c0a71 AK |
1067 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1068 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); | |
b10b37b8 | 1069 | __u32 ord; |
cdddbdbc DW |
1070 | |
1071 | printf("\n"); | |
1e7bc0ed | 1072 | printf("[%.16s]:\n", dev->volume); |
44470971 | 1073 | printf(" UUID : %s\n", uuid); |
dd8bcb3b AK |
1074 | printf(" RAID Level : %d", get_imsm_raid_level(map)); |
1075 | if (map2) | |
1076 | printf(" <-- %d", get_imsm_raid_level(map2)); | |
1077 | printf("\n"); | |
1078 | printf(" Members : %d", map->num_members); | |
1079 | if (map2) | |
1080 | printf(" <-- %d", map2->num_members); | |
1081 | printf("\n"); | |
0d80bb2f DW |
1082 | printf(" Slots : ["); |
1083 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 1084 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_0); |
0d80bb2f DW |
1085 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
1086 | } | |
dd8bcb3b AK |
1087 | printf("]"); |
1088 | if (map2) { | |
1089 | printf(" <-- ["); | |
1090 | for (i = 0; i < map2->num_members; i++) { | |
238c0a71 | 1091 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_1); |
dd8bcb3b AK |
1092 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
1093 | } | |
1094 | printf("]"); | |
1095 | } | |
1096 | printf("\n"); | |
7095bccb AK |
1097 | printf(" Failed disk : "); |
1098 | if (map->failed_disk_num == 0xff) | |
1099 | printf("none"); | |
1100 | else | |
1101 | printf("%i", map->failed_disk_num); | |
1102 | printf("\n"); | |
620b1713 DW |
1103 | slot = get_imsm_disk_slot(map, disk_idx); |
1104 | if (slot >= 0) { | |
238c0a71 | 1105 | ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); |
b10b37b8 DW |
1106 | printf(" This Slot : %d%s\n", slot, |
1107 | ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : ""); | |
1108 | } else | |
cdddbdbc DW |
1109 | printf(" This Slot : ?\n"); |
1110 | sz = __le32_to_cpu(dev->size_high); | |
1111 | sz <<= 32; | |
1112 | sz += __le32_to_cpu(dev->size_low); | |
1113 | printf(" Array Size : %llu%s\n", (unsigned long long)sz, | |
1114 | human_size(sz * 512)); | |
1115 | sz = __le32_to_cpu(map->blocks_per_member); | |
1116 | printf(" Per Dev Size : %llu%s\n", (unsigned long long)sz, | |
1117 | human_size(sz * 512)); | |
1118 | printf(" Sector Offset : %u\n", | |
1119 | __le32_to_cpu(map->pba_of_lba0)); | |
1120 | printf(" Num Stripes : %u\n", | |
1121 | __le32_to_cpu(map->num_data_stripes)); | |
dd8bcb3b | 1122 | printf(" Chunk Size : %u KiB", |
cdddbdbc | 1123 | __le16_to_cpu(map->blocks_per_strip) / 2); |
dd8bcb3b AK |
1124 | if (map2) |
1125 | printf(" <-- %u KiB", | |
1126 | __le16_to_cpu(map2->blocks_per_strip) / 2); | |
1127 | printf("\n"); | |
cdddbdbc | 1128 | printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks)); |
8655a7b1 | 1129 | printf(" Migrate State : "); |
1484e727 DW |
1130 | if (dev->vol.migr_state) { |
1131 | if (migr_type(dev) == MIGR_INIT) | |
8655a7b1 | 1132 | printf("initialize\n"); |
1484e727 | 1133 | else if (migr_type(dev) == MIGR_REBUILD) |
8655a7b1 | 1134 | printf("rebuild\n"); |
1484e727 | 1135 | else if (migr_type(dev) == MIGR_VERIFY) |
8655a7b1 | 1136 | printf("check\n"); |
1484e727 | 1137 | else if (migr_type(dev) == MIGR_GEN_MIGR) |
8655a7b1 | 1138 | printf("general migration\n"); |
1484e727 | 1139 | else if (migr_type(dev) == MIGR_STATE_CHANGE) |
8655a7b1 | 1140 | printf("state change\n"); |
1484e727 | 1141 | else if (migr_type(dev) == MIGR_REPAIR) |
8655a7b1 | 1142 | printf("repair\n"); |
1484e727 | 1143 | else |
8655a7b1 DW |
1144 | printf("<unknown:%d>\n", migr_type(dev)); |
1145 | } else | |
1146 | printf("idle\n"); | |
3393c6af DW |
1147 | printf(" Map State : %s", map_state_str[map->map_state]); |
1148 | if (dev->vol.migr_state) { | |
238c0a71 | 1149 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 | 1150 | |
b10b37b8 | 1151 | printf(" <-- %s", map_state_str[map->map_state]); |
464d40e8 LD |
1152 | printf("\n Checkpoint : %u ", |
1153 | __le32_to_cpu(dev->vol.curr_migr_unit)); | |
3136abe5 | 1154 | if ((is_gen_migration(dev)) && ((slot > 1) || (slot < 0))) |
464d40e8 LD |
1155 | printf("(N/A)"); |
1156 | else | |
1157 | printf("(%llu)", (unsigned long long) | |
1158 | blocks_per_migr_unit(super, dev)); | |
3393c6af DW |
1159 | } |
1160 | printf("\n"); | |
cdddbdbc | 1161 | printf(" Dirty State : %s\n", dev->vol.dirty ? "dirty" : "clean"); |
cdddbdbc DW |
1162 | } |
1163 | ||
0ec1f4e8 | 1164 | static void print_imsm_disk(struct imsm_disk *disk, int index, __u32 reserved) |
cdddbdbc | 1165 | { |
1f24f035 | 1166 | char str[MAX_RAID_SERIAL_LEN + 1]; |
cdddbdbc DW |
1167 | __u64 sz; |
1168 | ||
0ec1f4e8 | 1169 | if (index < -1 || !disk) |
e9d82038 DW |
1170 | return; |
1171 | ||
cdddbdbc | 1172 | printf("\n"); |
1f24f035 | 1173 | snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial); |
0ec1f4e8 DW |
1174 | if (index >= 0) |
1175 | printf(" Disk%02d Serial : %s\n", index, str); | |
1176 | else | |
1177 | printf(" Disk Serial : %s\n", str); | |
25ed7e59 DW |
1178 | printf(" State :%s%s%s\n", is_spare(disk) ? " spare" : "", |
1179 | is_configured(disk) ? " active" : "", | |
1180 | is_failed(disk) ? " failed" : ""); | |
cdddbdbc | 1181 | printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id)); |
14e8215b | 1182 | sz = __le32_to_cpu(disk->total_blocks) - reserved; |
cdddbdbc DW |
1183 | printf(" Usable Size : %llu%s\n", (unsigned long long)sz, |
1184 | human_size(sz * 512)); | |
1185 | } | |
1186 | ||
520e69e2 AK |
1187 | void examine_migr_rec_imsm(struct intel_super *super) |
1188 | { | |
1189 | struct migr_record *migr_rec = super->migr_rec; | |
1190 | struct imsm_super *mpb = super->anchor; | |
1191 | int i; | |
1192 | ||
1193 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
1194 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
3136abe5 | 1195 | struct imsm_map *map; |
b4ab44d8 | 1196 | int slot = -1; |
3136abe5 | 1197 | |
520e69e2 AK |
1198 | if (is_gen_migration(dev) == 0) |
1199 | continue; | |
1200 | ||
1201 | printf("\nMigration Record Information:"); | |
3136abe5 | 1202 | |
44bfe6df AK |
1203 | /* first map under migration */ |
1204 | map = get_imsm_map(dev, MAP_0); | |
3136abe5 AK |
1205 | if (map) |
1206 | slot = get_imsm_disk_slot(map, super->disks->index); | |
1207 | if ((map == NULL) || (slot > 1) || (slot < 0)) { | |
520e69e2 AK |
1208 | printf(" Empty\n "); |
1209 | printf("Examine one of first two disks in array\n"); | |
1210 | break; | |
1211 | } | |
1212 | printf("\n Status : "); | |
1213 | if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL) | |
1214 | printf("Normal\n"); | |
1215 | else | |
1216 | printf("Contains Data\n"); | |
1217 | printf(" Current Unit : %u\n", | |
1218 | __le32_to_cpu(migr_rec->curr_migr_unit)); | |
1219 | printf(" Family : %u\n", | |
1220 | __le32_to_cpu(migr_rec->family_num)); | |
1221 | printf(" Ascending : %u\n", | |
1222 | __le32_to_cpu(migr_rec->ascending_migr)); | |
1223 | printf(" Blocks Per Unit : %u\n", | |
1224 | __le32_to_cpu(migr_rec->blocks_per_unit)); | |
1225 | printf(" Dest. Depth Per Unit : %u\n", | |
1226 | __le32_to_cpu(migr_rec->dest_depth_per_unit)); | |
1227 | printf(" Checkpoint Area pba : %u\n", | |
1228 | __le32_to_cpu(migr_rec->ckpt_area_pba)); | |
1229 | printf(" First member lba : %u\n", | |
1230 | __le32_to_cpu(migr_rec->dest_1st_member_lba)); | |
1231 | printf(" Total Number of Units : %u\n", | |
1232 | __le32_to_cpu(migr_rec->num_migr_units)); | |
1233 | printf(" Size of volume : %u\n", | |
1234 | __le32_to_cpu(migr_rec->post_migr_vol_cap)); | |
1235 | printf(" Expansion space for LBA64 : %u\n", | |
1236 | __le32_to_cpu(migr_rec->post_migr_vol_cap_hi)); | |
1237 | printf(" Record was read from : %u\n", | |
1238 | __le32_to_cpu(migr_rec->ckpt_read_disk_num)); | |
1239 | ||
1240 | break; | |
1241 | } | |
1242 | } | |
9e2d750d | 1243 | #endif /* MDASSEMBLE */ |
19482bcc AK |
1244 | /******************************************************************************* |
1245 | * function: imsm_check_attributes | |
1246 | * Description: Function checks if features represented by attributes flags | |
1247 | * are supported by mdadm. | |
1248 | * Parameters: | |
1249 | * attributes - Attributes read from metadata | |
1250 | * Returns: | |
1251 | * 0 - passed attributes contains unsupported features flags | |
1252 | * 1 - all features are supported | |
1253 | ******************************************************************************/ | |
1254 | static int imsm_check_attributes(__u32 attributes) | |
1255 | { | |
1256 | int ret_val = 1; | |
418f9b36 N |
1257 | __u32 not_supported = MPB_ATTRIB_SUPPORTED^0xffffffff; |
1258 | ||
1259 | not_supported &= ~MPB_ATTRIB_IGNORED; | |
19482bcc AK |
1260 | |
1261 | not_supported &= attributes; | |
1262 | if (not_supported) { | |
418f9b36 N |
1263 | fprintf(stderr, Name "(IMSM): Unsupported attributes : %x\n", |
1264 | (unsigned)__le32_to_cpu(not_supported)); | |
19482bcc AK |
1265 | if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) { |
1266 | dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY \n"); | |
1267 | not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY; | |
1268 | } | |
1269 | if (not_supported & MPB_ATTRIB_2TB) { | |
1270 | dprintf("\t\tMPB_ATTRIB_2TB\n"); | |
1271 | not_supported ^= MPB_ATTRIB_2TB; | |
1272 | } | |
1273 | if (not_supported & MPB_ATTRIB_RAID0) { | |
1274 | dprintf("\t\tMPB_ATTRIB_RAID0\n"); | |
1275 | not_supported ^= MPB_ATTRIB_RAID0; | |
1276 | } | |
1277 | if (not_supported & MPB_ATTRIB_RAID1) { | |
1278 | dprintf("\t\tMPB_ATTRIB_RAID1\n"); | |
1279 | not_supported ^= MPB_ATTRIB_RAID1; | |
1280 | } | |
1281 | if (not_supported & MPB_ATTRIB_RAID10) { | |
1282 | dprintf("\t\tMPB_ATTRIB_RAID10\n"); | |
1283 | not_supported ^= MPB_ATTRIB_RAID10; | |
1284 | } | |
1285 | if (not_supported & MPB_ATTRIB_RAID1E) { | |
1286 | dprintf("\t\tMPB_ATTRIB_RAID1E\n"); | |
1287 | not_supported ^= MPB_ATTRIB_RAID1E; | |
1288 | } | |
1289 | if (not_supported & MPB_ATTRIB_RAID5) { | |
1290 | dprintf("\t\tMPB_ATTRIB_RAID5\n"); | |
1291 | not_supported ^= MPB_ATTRIB_RAID5; | |
1292 | } | |
1293 | if (not_supported & MPB_ATTRIB_RAIDCNG) { | |
1294 | dprintf("\t\tMPB_ATTRIB_RAIDCNG\n"); | |
1295 | not_supported ^= MPB_ATTRIB_RAIDCNG; | |
1296 | } | |
1297 | if (not_supported & MPB_ATTRIB_BBM) { | |
1298 | dprintf("\t\tMPB_ATTRIB_BBM\n"); | |
1299 | not_supported ^= MPB_ATTRIB_BBM; | |
1300 | } | |
1301 | if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) { | |
1302 | dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY (== MPB_ATTRIB_LEGACY)\n"); | |
1303 | not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY; | |
1304 | } | |
1305 | if (not_supported & MPB_ATTRIB_EXP_STRIPE_SIZE) { | |
1306 | dprintf("\t\tMPB_ATTRIB_EXP_STRIP_SIZE\n"); | |
1307 | not_supported ^= MPB_ATTRIB_EXP_STRIPE_SIZE; | |
1308 | } | |
1309 | if (not_supported & MPB_ATTRIB_2TB_DISK) { | |
1310 | dprintf("\t\tMPB_ATTRIB_2TB_DISK\n"); | |
1311 | not_supported ^= MPB_ATTRIB_2TB_DISK; | |
1312 | } | |
1313 | if (not_supported & MPB_ATTRIB_NEVER_USE2) { | |
1314 | dprintf("\t\tMPB_ATTRIB_NEVER_USE2\n"); | |
1315 | not_supported ^= MPB_ATTRIB_NEVER_USE2; | |
1316 | } | |
1317 | if (not_supported & MPB_ATTRIB_NEVER_USE) { | |
1318 | dprintf("\t\tMPB_ATTRIB_NEVER_USE\n"); | |
1319 | not_supported ^= MPB_ATTRIB_NEVER_USE; | |
1320 | } | |
1321 | ||
1322 | if (not_supported) | |
1323 | dprintf(Name "(IMSM): Unknown attributes : %x\n", not_supported); | |
1324 | ||
1325 | ret_val = 0; | |
1326 | } | |
1327 | ||
1328 | return ret_val; | |
1329 | } | |
1330 | ||
9e2d750d | 1331 | #ifndef MDASSEMBLE |
a5d85af7 | 1332 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map); |
44470971 | 1333 | |
cdddbdbc DW |
1334 | static void examine_super_imsm(struct supertype *st, char *homehost) |
1335 | { | |
1336 | struct intel_super *super = st->sb; | |
949c47a0 | 1337 | struct imsm_super *mpb = super->anchor; |
cdddbdbc DW |
1338 | char str[MAX_SIGNATURE_LENGTH]; |
1339 | int i; | |
27fd6274 DW |
1340 | struct mdinfo info; |
1341 | char nbuf[64]; | |
cdddbdbc | 1342 | __u32 sum; |
14e8215b | 1343 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
94827db3 | 1344 | struct dl *dl; |
27fd6274 | 1345 | |
cdddbdbc DW |
1346 | snprintf(str, MPB_SIG_LEN, "%s", mpb->sig); |
1347 | printf(" Magic : %s\n", str); | |
1348 | snprintf(str, strlen(MPB_VERSION_RAID0), "%s", get_imsm_version(mpb)); | |
1349 | printf(" Version : %s\n", get_imsm_version(mpb)); | |
148acb7b | 1350 | printf(" Orig Family : %08x\n", __le32_to_cpu(mpb->orig_family_num)); |
cdddbdbc DW |
1351 | printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num)); |
1352 | printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num)); | |
19482bcc AK |
1353 | printf(" Attributes : "); |
1354 | if (imsm_check_attributes(mpb->attributes)) | |
1355 | printf("All supported\n"); | |
1356 | else | |
1357 | printf("not supported\n"); | |
a5d85af7 | 1358 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1359 | fname_from_uuid(st, &info, nbuf, ':'); |
27fd6274 | 1360 | printf(" UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
1361 | sum = __le32_to_cpu(mpb->check_sum); |
1362 | printf(" Checksum : %08x %s\n", sum, | |
949c47a0 | 1363 | __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect"); |
87eb16df | 1364 | printf(" MPB Sectors : %d\n", mpb_sectors(mpb)); |
cdddbdbc DW |
1365 | printf(" Disks : %d\n", mpb->num_disks); |
1366 | printf(" RAID Devices : %d\n", mpb->num_raid_devs); | |
0ec1f4e8 | 1367 | print_imsm_disk(__get_imsm_disk(mpb, super->disks->index), super->disks->index, reserved); |
604b746f JD |
1368 | if (super->bbm_log) { |
1369 | struct bbm_log *log = super->bbm_log; | |
1370 | ||
1371 | printf("\n"); | |
1372 | printf("Bad Block Management Log:\n"); | |
1373 | printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size)); | |
1374 | printf(" Signature : %x\n", __le32_to_cpu(log->signature)); | |
1375 | printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count)); | |
1376 | printf(" Spare Blocks : %d\n", __le32_to_cpu(log->reserved_spare_block_count)); | |
13a3b65d N |
1377 | printf(" First Spare : %llx\n", |
1378 | (unsigned long long) __le64_to_cpu(log->first_spare_lba)); | |
604b746f | 1379 | } |
44470971 DW |
1380 | for (i = 0; i < mpb->num_raid_devs; i++) { |
1381 | struct mdinfo info; | |
1382 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
1383 | ||
1384 | super->current_vol = i; | |
a5d85af7 | 1385 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1386 | fname_from_uuid(st, &info, nbuf, ':'); |
c47b0ff6 | 1387 | print_imsm_dev(super, dev, nbuf + 5, super->disks->index); |
44470971 | 1388 | } |
cdddbdbc DW |
1389 | for (i = 0; i < mpb->num_disks; i++) { |
1390 | if (i == super->disks->index) | |
1391 | continue; | |
0ec1f4e8 | 1392 | print_imsm_disk(__get_imsm_disk(mpb, i), i, reserved); |
cdddbdbc | 1393 | } |
94827db3 | 1394 | |
0ec1f4e8 DW |
1395 | for (dl = super->disks; dl; dl = dl->next) |
1396 | if (dl->index == -1) | |
1397 | print_imsm_disk(&dl->disk, -1, reserved); | |
520e69e2 AK |
1398 | |
1399 | examine_migr_rec_imsm(super); | |
cdddbdbc DW |
1400 | } |
1401 | ||
061f2c6a | 1402 | static void brief_examine_super_imsm(struct supertype *st, int verbose) |
cdddbdbc | 1403 | { |
27fd6274 | 1404 | /* We just write a generic IMSM ARRAY entry */ |
ff54de6e N |
1405 | struct mdinfo info; |
1406 | char nbuf[64]; | |
1e7bc0ed | 1407 | struct intel_super *super = st->sb; |
1e7bc0ed | 1408 | |
0d5a423f DW |
1409 | if (!super->anchor->num_raid_devs) { |
1410 | printf("ARRAY metadata=imsm\n"); | |
1e7bc0ed | 1411 | return; |
0d5a423f | 1412 | } |
ff54de6e | 1413 | |
a5d85af7 | 1414 | getinfo_super_imsm(st, &info, NULL); |
4737ae25 N |
1415 | fname_from_uuid(st, &info, nbuf, ':'); |
1416 | printf("ARRAY metadata=imsm UUID=%s\n", nbuf + 5); | |
1417 | } | |
1418 | ||
1419 | static void brief_examine_subarrays_imsm(struct supertype *st, int verbose) | |
1420 | { | |
1421 | /* We just write a generic IMSM ARRAY entry */ | |
1422 | struct mdinfo info; | |
1423 | char nbuf[64]; | |
1424 | char nbuf1[64]; | |
1425 | struct intel_super *super = st->sb; | |
1426 | int i; | |
1427 | ||
1428 | if (!super->anchor->num_raid_devs) | |
1429 | return; | |
1430 | ||
a5d85af7 | 1431 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1432 | fname_from_uuid(st, &info, nbuf, ':'); |
1e7bc0ed DW |
1433 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
1434 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
1435 | ||
1436 | super->current_vol = i; | |
a5d85af7 | 1437 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1438 | fname_from_uuid(st, &info, nbuf1, ':'); |
1124b3cf | 1439 | printf("ARRAY /dev/md/%.16s container=%s member=%d UUID=%s\n", |
cf8de691 | 1440 | dev->volume, nbuf + 5, i, nbuf1 + 5); |
1e7bc0ed | 1441 | } |
cdddbdbc DW |
1442 | } |
1443 | ||
9d84c8ea DW |
1444 | static void export_examine_super_imsm(struct supertype *st) |
1445 | { | |
1446 | struct intel_super *super = st->sb; | |
1447 | struct imsm_super *mpb = super->anchor; | |
1448 | struct mdinfo info; | |
1449 | char nbuf[64]; | |
1450 | ||
a5d85af7 | 1451 | getinfo_super_imsm(st, &info, NULL); |
9d84c8ea DW |
1452 | fname_from_uuid(st, &info, nbuf, ':'); |
1453 | printf("MD_METADATA=imsm\n"); | |
1454 | printf("MD_LEVEL=container\n"); | |
1455 | printf("MD_UUID=%s\n", nbuf+5); | |
1456 | printf("MD_DEVICES=%u\n", mpb->num_disks); | |
1457 | } | |
1458 | ||
cdddbdbc DW |
1459 | static void detail_super_imsm(struct supertype *st, char *homehost) |
1460 | { | |
3ebe00a1 DW |
1461 | struct mdinfo info; |
1462 | char nbuf[64]; | |
1463 | ||
a5d85af7 | 1464 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1465 | fname_from_uuid(st, &info, nbuf, ':'); |
3ebe00a1 | 1466 | printf("\n UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
1467 | } |
1468 | ||
1469 | static void brief_detail_super_imsm(struct supertype *st) | |
1470 | { | |
ff54de6e N |
1471 | struct mdinfo info; |
1472 | char nbuf[64]; | |
a5d85af7 | 1473 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1474 | fname_from_uuid(st, &info, nbuf, ':'); |
ff54de6e | 1475 | printf(" UUID=%s", nbuf + 5); |
cdddbdbc | 1476 | } |
d665cc31 DW |
1477 | |
1478 | static int imsm_read_serial(int fd, char *devname, __u8 *serial); | |
1479 | static void fd2devname(int fd, char *name); | |
1480 | ||
120dc887 | 1481 | static int ahci_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose) |
d665cc31 | 1482 | { |
120dc887 LM |
1483 | /* dump an unsorted list of devices attached to AHCI Intel storage |
1484 | * controller, as well as non-connected ports | |
d665cc31 DW |
1485 | */ |
1486 | int hba_len = strlen(hba_path) + 1; | |
1487 | struct dirent *ent; | |
1488 | DIR *dir; | |
1489 | char *path = NULL; | |
1490 | int err = 0; | |
1491 | unsigned long port_mask = (1 << port_count) - 1; | |
1492 | ||
f21e18ca | 1493 | if (port_count > (int)sizeof(port_mask) * 8) { |
d665cc31 DW |
1494 | if (verbose) |
1495 | fprintf(stderr, Name ": port_count %d out of range\n", port_count); | |
1496 | return 2; | |
1497 | } | |
1498 | ||
1499 | /* scroll through /sys/dev/block looking for devices attached to | |
1500 | * this hba | |
1501 | */ | |
1502 | dir = opendir("/sys/dev/block"); | |
1503 | for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) { | |
1504 | int fd; | |
1505 | char model[64]; | |
1506 | char vendor[64]; | |
1507 | char buf[1024]; | |
1508 | int major, minor; | |
1509 | char *device; | |
1510 | char *c; | |
1511 | int port; | |
1512 | int type; | |
1513 | ||
1514 | if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2) | |
1515 | continue; | |
1516 | path = devt_to_devpath(makedev(major, minor)); | |
1517 | if (!path) | |
1518 | continue; | |
1519 | if (!path_attached_to_hba(path, hba_path)) { | |
1520 | free(path); | |
1521 | path = NULL; | |
1522 | continue; | |
1523 | } | |
1524 | ||
1525 | /* retrieve the scsi device type */ | |
1526 | if (asprintf(&device, "/sys/dev/block/%d:%d/device/xxxxxxx", major, minor) < 0) { | |
1527 | if (verbose) | |
1528 | fprintf(stderr, Name ": failed to allocate 'device'\n"); | |
1529 | err = 2; | |
1530 | break; | |
1531 | } | |
1532 | sprintf(device, "/sys/dev/block/%d:%d/device/type", major, minor); | |
1533 | if (load_sys(device, buf) != 0) { | |
1534 | if (verbose) | |
1535 | fprintf(stderr, Name ": failed to read device type for %s\n", | |
1536 | path); | |
1537 | err = 2; | |
1538 | free(device); | |
1539 | break; | |
1540 | } | |
1541 | type = strtoul(buf, NULL, 10); | |
1542 | ||
1543 | /* if it's not a disk print the vendor and model */ | |
1544 | if (!(type == 0 || type == 7 || type == 14)) { | |
1545 | vendor[0] = '\0'; | |
1546 | model[0] = '\0'; | |
1547 | sprintf(device, "/sys/dev/block/%d:%d/device/vendor", major, minor); | |
1548 | if (load_sys(device, buf) == 0) { | |
1549 | strncpy(vendor, buf, sizeof(vendor)); | |
1550 | vendor[sizeof(vendor) - 1] = '\0'; | |
1551 | c = (char *) &vendor[sizeof(vendor) - 1]; | |
1552 | while (isspace(*c) || *c == '\0') | |
1553 | *c-- = '\0'; | |
1554 | ||
1555 | } | |
1556 | sprintf(device, "/sys/dev/block/%d:%d/device/model", major, minor); | |
1557 | if (load_sys(device, buf) == 0) { | |
1558 | strncpy(model, buf, sizeof(model)); | |
1559 | model[sizeof(model) - 1] = '\0'; | |
1560 | c = (char *) &model[sizeof(model) - 1]; | |
1561 | while (isspace(*c) || *c == '\0') | |
1562 | *c-- = '\0'; | |
1563 | } | |
1564 | ||
1565 | if (vendor[0] && model[0]) | |
1566 | sprintf(buf, "%.64s %.64s", vendor, model); | |
1567 | else | |
1568 | switch (type) { /* numbers from hald/linux/device.c */ | |
1569 | case 1: sprintf(buf, "tape"); break; | |
1570 | case 2: sprintf(buf, "printer"); break; | |
1571 | case 3: sprintf(buf, "processor"); break; | |
1572 | case 4: | |
1573 | case 5: sprintf(buf, "cdrom"); break; | |
1574 | case 6: sprintf(buf, "scanner"); break; | |
1575 | case 8: sprintf(buf, "media_changer"); break; | |
1576 | case 9: sprintf(buf, "comm"); break; | |
1577 | case 12: sprintf(buf, "raid"); break; | |
1578 | default: sprintf(buf, "unknown"); | |
1579 | } | |
1580 | } else | |
1581 | buf[0] = '\0'; | |
1582 | free(device); | |
1583 | ||
1584 | /* chop device path to 'host%d' and calculate the port number */ | |
1585 | c = strchr(&path[hba_len], '/'); | |
4e5e717d AW |
1586 | if (!c) { |
1587 | if (verbose) | |
1588 | fprintf(stderr, Name ": %s - invalid path name\n", path + hba_len); | |
1589 | err = 2; | |
1590 | break; | |
1591 | } | |
d665cc31 DW |
1592 | *c = '\0'; |
1593 | if (sscanf(&path[hba_len], "host%d", &port) == 1) | |
1594 | port -= host_base; | |
1595 | else { | |
1596 | if (verbose) { | |
1597 | *c = '/'; /* repair the full string */ | |
1598 | fprintf(stderr, Name ": failed to determine port number for %s\n", | |
1599 | path); | |
1600 | } | |
1601 | err = 2; | |
1602 | break; | |
1603 | } | |
1604 | ||
1605 | /* mark this port as used */ | |
1606 | port_mask &= ~(1 << port); | |
1607 | ||
1608 | /* print out the device information */ | |
1609 | if (buf[0]) { | |
1610 | printf(" Port%d : - non-disk device (%s) -\n", port, buf); | |
1611 | continue; | |
1612 | } | |
1613 | ||
1614 | fd = dev_open(ent->d_name, O_RDONLY); | |
1615 | if (fd < 0) | |
1616 | printf(" Port%d : - disk info unavailable -\n", port); | |
1617 | else { | |
1618 | fd2devname(fd, buf); | |
1619 | printf(" Port%d : %s", port, buf); | |
1620 | if (imsm_read_serial(fd, NULL, (__u8 *) buf) == 0) | |
664d5325 | 1621 | printf(" (%.*s)\n", MAX_RAID_SERIAL_LEN, buf); |
d665cc31 | 1622 | else |
664d5325 | 1623 | printf(" ()\n"); |
4dab422a | 1624 | close(fd); |
d665cc31 | 1625 | } |
d665cc31 DW |
1626 | free(path); |
1627 | path = NULL; | |
1628 | } | |
1629 | if (path) | |
1630 | free(path); | |
1631 | if (dir) | |
1632 | closedir(dir); | |
1633 | if (err == 0) { | |
1634 | int i; | |
1635 | ||
1636 | for (i = 0; i < port_count; i++) | |
1637 | if (port_mask & (1 << i)) | |
1638 | printf(" Port%d : - no device attached -\n", i); | |
1639 | } | |
1640 | ||
1641 | return err; | |
1642 | } | |
1643 | ||
120dc887 LM |
1644 | static void print_found_intel_controllers(struct sys_dev *elem) |
1645 | { | |
1646 | for (; elem; elem = elem->next) { | |
1647 | fprintf(stderr, Name ": found Intel(R) "); | |
1648 | if (elem->type == SYS_DEV_SATA) | |
1649 | fprintf(stderr, "SATA "); | |
155cbb4c LM |
1650 | else if (elem->type == SYS_DEV_SAS) |
1651 | fprintf(stderr, "SAS "); | |
120dc887 LM |
1652 | fprintf(stderr, "RAID controller"); |
1653 | if (elem->pci_id) | |
1654 | fprintf(stderr, " at %s", elem->pci_id); | |
1655 | fprintf(stderr, ".\n"); | |
1656 | } | |
1657 | fflush(stderr); | |
1658 | } | |
1659 | ||
120dc887 LM |
1660 | static int ahci_get_port_count(const char *hba_path, int *port_count) |
1661 | { | |
1662 | struct dirent *ent; | |
1663 | DIR *dir; | |
1664 | int host_base = -1; | |
1665 | ||
1666 | *port_count = 0; | |
1667 | if ((dir = opendir(hba_path)) == NULL) | |
1668 | return -1; | |
1669 | ||
1670 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
1671 | int host; | |
1672 | ||
1673 | if (sscanf(ent->d_name, "host%d", &host) != 1) | |
1674 | continue; | |
1675 | if (*port_count == 0) | |
1676 | host_base = host; | |
1677 | else if (host < host_base) | |
1678 | host_base = host; | |
1679 | ||
1680 | if (host + 1 > *port_count + host_base) | |
1681 | *port_count = host + 1 - host_base; | |
1682 | } | |
1683 | closedir(dir); | |
1684 | return host_base; | |
1685 | } | |
1686 | ||
a891a3c2 LM |
1687 | static void print_imsm_capability(const struct imsm_orom *orom) |
1688 | { | |
1689 | printf(" Platform : Intel(R) Matrix Storage Manager\n"); | |
1690 | printf(" Version : %d.%d.%d.%d\n", orom->major_ver, orom->minor_ver, | |
1691 | orom->hotfix_ver, orom->build); | |
1692 | printf(" RAID Levels :%s%s%s%s%s\n", | |
1693 | imsm_orom_has_raid0(orom) ? " raid0" : "", | |
1694 | imsm_orom_has_raid1(orom) ? " raid1" : "", | |
1695 | imsm_orom_has_raid1e(orom) ? " raid1e" : "", | |
1696 | imsm_orom_has_raid10(orom) ? " raid10" : "", | |
1697 | imsm_orom_has_raid5(orom) ? " raid5" : ""); | |
1698 | printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n", | |
1699 | imsm_orom_has_chunk(orom, 2) ? " 2k" : "", | |
1700 | imsm_orom_has_chunk(orom, 4) ? " 4k" : "", | |
1701 | imsm_orom_has_chunk(orom, 8) ? " 8k" : "", | |
1702 | imsm_orom_has_chunk(orom, 16) ? " 16k" : "", | |
1703 | imsm_orom_has_chunk(orom, 32) ? " 32k" : "", | |
1704 | imsm_orom_has_chunk(orom, 64) ? " 64k" : "", | |
1705 | imsm_orom_has_chunk(orom, 128) ? " 128k" : "", | |
1706 | imsm_orom_has_chunk(orom, 256) ? " 256k" : "", | |
1707 | imsm_orom_has_chunk(orom, 512) ? " 512k" : "", | |
1708 | imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "", | |
1709 | imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "", | |
1710 | imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "", | |
1711 | imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "", | |
1712 | imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "", | |
1713 | imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "", | |
1714 | imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : ""); | |
1715 | printf(" Max Disks : %d\n", orom->tds); | |
1716 | printf(" Max Volumes : %d\n", orom->vpa); | |
1717 | return; | |
1718 | } | |
1719 | ||
5615172f | 1720 | static int detail_platform_imsm(int verbose, int enumerate_only) |
d665cc31 DW |
1721 | { |
1722 | /* There are two components to imsm platform support, the ahci SATA | |
1723 | * controller and the option-rom. To find the SATA controller we | |
1724 | * simply look in /sys/bus/pci/drivers/ahci to see if an ahci | |
1725 | * controller with the Intel vendor id is present. This approach | |
1726 | * allows mdadm to leverage the kernel's ahci detection logic, with the | |
1727 | * caveat that if ahci.ko is not loaded mdadm will not be able to | |
1728 | * detect platform raid capabilities. The option-rom resides in a | |
1729 | * platform "Adapter ROM". We scan for its signature to retrieve the | |
1730 | * platform capabilities. If raid support is disabled in the BIOS the | |
1731 | * option-rom capability structure will not be available. | |
1732 | */ | |
1733 | const struct imsm_orom *orom; | |
1734 | struct sys_dev *list, *hba; | |
d665cc31 DW |
1735 | int host_base = 0; |
1736 | int port_count = 0; | |
120dc887 | 1737 | int result=0; |
d665cc31 | 1738 | |
5615172f | 1739 | if (enumerate_only) { |
a891a3c2 | 1740 | if (check_env("IMSM_NO_PLATFORM")) |
5615172f | 1741 | return 0; |
a891a3c2 LM |
1742 | list = find_intel_devices(); |
1743 | if (!list) | |
1744 | return 2; | |
1745 | for (hba = list; hba; hba = hba->next) { | |
1746 | orom = find_imsm_capability(hba->type); | |
1747 | if (!orom) { | |
1748 | result = 2; | |
1749 | break; | |
1750 | } | |
1751 | } | |
1752 | free_sys_dev(&list); | |
1753 | return result; | |
5615172f DW |
1754 | } |
1755 | ||
155cbb4c LM |
1756 | list = find_intel_devices(); |
1757 | if (!list) { | |
d665cc31 | 1758 | if (verbose) |
155cbb4c LM |
1759 | fprintf(stderr, Name ": no active Intel(R) RAID " |
1760 | "controller found.\n"); | |
d665cc31 DW |
1761 | free_sys_dev(&list); |
1762 | return 2; | |
1763 | } else if (verbose) | |
155cbb4c | 1764 | print_found_intel_controllers(list); |
d665cc31 | 1765 | |
a891a3c2 LM |
1766 | for (hba = list; hba; hba = hba->next) { |
1767 | orom = find_imsm_capability(hba->type); | |
1768 | if (!orom) | |
1769 | fprintf(stderr, Name ": imsm capabilities not found for controller: %s (type %s)\n", | |
1770 | hba->path, get_sys_dev_type(hba->type)); | |
1771 | else | |
1772 | print_imsm_capability(orom); | |
d665cc31 DW |
1773 | } |
1774 | ||
120dc887 LM |
1775 | for (hba = list; hba; hba = hba->next) { |
1776 | printf(" I/O Controller : %s (%s)\n", | |
1777 | hba->path, get_sys_dev_type(hba->type)); | |
d665cc31 | 1778 | |
120dc887 LM |
1779 | if (hba->type == SYS_DEV_SATA) { |
1780 | host_base = ahci_get_port_count(hba->path, &port_count); | |
1781 | if (ahci_enumerate_ports(hba->path, port_count, host_base, verbose)) { | |
1782 | if (verbose) | |
1783 | fprintf(stderr, Name ": failed to enumerate " | |
1784 | "ports on SATA controller at %s.", hba->pci_id); | |
1785 | result |= 2; | |
1786 | } | |
1787 | } | |
d665cc31 | 1788 | } |
155cbb4c | 1789 | |
120dc887 LM |
1790 | free_sys_dev(&list); |
1791 | return result; | |
d665cc31 | 1792 | } |
cdddbdbc DW |
1793 | #endif |
1794 | ||
1795 | static int match_home_imsm(struct supertype *st, char *homehost) | |
1796 | { | |
5115ca67 DW |
1797 | /* the imsm metadata format does not specify any host |
1798 | * identification information. We return -1 since we can never | |
1799 | * confirm nor deny whether a given array is "meant" for this | |
148acb7b | 1800 | * host. We rely on compare_super and the 'family_num' fields to |
5115ca67 DW |
1801 | * exclude member disks that do not belong, and we rely on |
1802 | * mdadm.conf to specify the arrays that should be assembled. | |
1803 | * Auto-assembly may still pick up "foreign" arrays. | |
1804 | */ | |
cdddbdbc | 1805 | |
9362c1c8 | 1806 | return -1; |
cdddbdbc DW |
1807 | } |
1808 | ||
1809 | static void uuid_from_super_imsm(struct supertype *st, int uuid[4]) | |
1810 | { | |
51006d85 N |
1811 | /* The uuid returned here is used for: |
1812 | * uuid to put into bitmap file (Create, Grow) | |
1813 | * uuid for backup header when saving critical section (Grow) | |
1814 | * comparing uuids when re-adding a device into an array | |
1815 | * In these cases the uuid required is that of the data-array, | |
1816 | * not the device-set. | |
1817 | * uuid to recognise same set when adding a missing device back | |
1818 | * to an array. This is a uuid for the device-set. | |
1819 | * | |
1820 | * For each of these we can make do with a truncated | |
1821 | * or hashed uuid rather than the original, as long as | |
1822 | * everyone agrees. | |
1823 | * In each case the uuid required is that of the data-array, | |
1824 | * not the device-set. | |
43dad3d6 | 1825 | */ |
51006d85 N |
1826 | /* imsm does not track uuid's so we synthesis one using sha1 on |
1827 | * - The signature (Which is constant for all imsm array, but no matter) | |
148acb7b | 1828 | * - the orig_family_num of the container |
51006d85 N |
1829 | * - the index number of the volume |
1830 | * - the 'serial' number of the volume. | |
1831 | * Hopefully these are all constant. | |
1832 | */ | |
1833 | struct intel_super *super = st->sb; | |
43dad3d6 | 1834 | |
51006d85 N |
1835 | char buf[20]; |
1836 | struct sha1_ctx ctx; | |
1837 | struct imsm_dev *dev = NULL; | |
148acb7b | 1838 | __u32 family_num; |
51006d85 | 1839 | |
148acb7b DW |
1840 | /* some mdadm versions failed to set ->orig_family_num, in which |
1841 | * case fall back to ->family_num. orig_family_num will be | |
1842 | * fixed up with the first metadata update. | |
1843 | */ | |
1844 | family_num = super->anchor->orig_family_num; | |
1845 | if (family_num == 0) | |
1846 | family_num = super->anchor->family_num; | |
51006d85 | 1847 | sha1_init_ctx(&ctx); |
92bd8f8d | 1848 | sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx); |
148acb7b | 1849 | sha1_process_bytes(&family_num, sizeof(__u32), &ctx); |
51006d85 N |
1850 | if (super->current_vol >= 0) |
1851 | dev = get_imsm_dev(super, super->current_vol); | |
1852 | if (dev) { | |
1853 | __u32 vol = super->current_vol; | |
1854 | sha1_process_bytes(&vol, sizeof(vol), &ctx); | |
1855 | sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx); | |
1856 | } | |
1857 | sha1_finish_ctx(&ctx, buf); | |
1858 | memcpy(uuid, buf, 4*4); | |
cdddbdbc DW |
1859 | } |
1860 | ||
0d481d37 | 1861 | #if 0 |
4f5bc454 DW |
1862 | static void |
1863 | get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p) | |
cdddbdbc | 1864 | { |
cdddbdbc DW |
1865 | __u8 *v = get_imsm_version(mpb); |
1866 | __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH; | |
1867 | char major[] = { 0, 0, 0 }; | |
1868 | char minor[] = { 0 ,0, 0 }; | |
1869 | char patch[] = { 0, 0, 0 }; | |
1870 | char *ver_parse[] = { major, minor, patch }; | |
1871 | int i, j; | |
1872 | ||
1873 | i = j = 0; | |
1874 | while (*v != '\0' && v < end) { | |
1875 | if (*v != '.' && j < 2) | |
1876 | ver_parse[i][j++] = *v; | |
1877 | else { | |
1878 | i++; | |
1879 | j = 0; | |
1880 | } | |
1881 | v++; | |
1882 | } | |
1883 | ||
4f5bc454 DW |
1884 | *m = strtol(minor, NULL, 0); |
1885 | *p = strtol(patch, NULL, 0); | |
1886 | } | |
0d481d37 | 1887 | #endif |
4f5bc454 | 1888 | |
1e5c6983 DW |
1889 | static __u32 migr_strip_blocks_resync(struct imsm_dev *dev) |
1890 | { | |
1891 | /* migr_strip_size when repairing or initializing parity */ | |
238c0a71 | 1892 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
1893 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
1894 | ||
1895 | switch (get_imsm_raid_level(map)) { | |
1896 | case 5: | |
1897 | case 10: | |
1898 | return chunk; | |
1899 | default: | |
1900 | return 128*1024 >> 9; | |
1901 | } | |
1902 | } | |
1903 | ||
1904 | static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev) | |
1905 | { | |
1906 | /* migr_strip_size when rebuilding a degraded disk, no idea why | |
1907 | * this is different than migr_strip_size_resync(), but it's good | |
1908 | * to be compatible | |
1909 | */ | |
238c0a71 | 1910 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
1911 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
1912 | ||
1913 | switch (get_imsm_raid_level(map)) { | |
1914 | case 1: | |
1915 | case 10: | |
1916 | if (map->num_members % map->num_domains == 0) | |
1917 | return 128*1024 >> 9; | |
1918 | else | |
1919 | return chunk; | |
1920 | case 5: | |
1921 | return max((__u32) 64*1024 >> 9, chunk); | |
1922 | default: | |
1923 | return 128*1024 >> 9; | |
1924 | } | |
1925 | } | |
1926 | ||
1927 | static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev) | |
1928 | { | |
238c0a71 AK |
1929 | struct imsm_map *lo = get_imsm_map(dev, MAP_0); |
1930 | struct imsm_map *hi = get_imsm_map(dev, MAP_1); | |
1e5c6983 DW |
1931 | __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip); |
1932 | __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip); | |
1933 | ||
1934 | return max((__u32) 1, hi_chunk / lo_chunk); | |
1935 | } | |
1936 | ||
1937 | static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev) | |
1938 | { | |
238c0a71 | 1939 | struct imsm_map *lo = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
1940 | int level = get_imsm_raid_level(lo); |
1941 | ||
1942 | if (level == 1 || level == 10) { | |
238c0a71 | 1943 | struct imsm_map *hi = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
1944 | |
1945 | return hi->num_domains; | |
1946 | } else | |
1947 | return num_stripes_per_unit_resync(dev); | |
1948 | } | |
1949 | ||
98130f40 | 1950 | static __u8 imsm_num_data_members(struct imsm_dev *dev, int second_map) |
1e5c6983 DW |
1951 | { |
1952 | /* named 'imsm_' because raid0, raid1 and raid10 | |
1953 | * counter-intuitively have the same number of data disks | |
1954 | */ | |
98130f40 | 1955 | struct imsm_map *map = get_imsm_map(dev, second_map); |
1e5c6983 DW |
1956 | |
1957 | switch (get_imsm_raid_level(map)) { | |
1958 | case 0: | |
1959 | case 1: | |
1960 | case 10: | |
1961 | return map->num_members; | |
1962 | case 5: | |
1963 | return map->num_members - 1; | |
1964 | default: | |
1965 | dprintf("%s: unsupported raid level\n", __func__); | |
1966 | return 0; | |
1967 | } | |
1968 | } | |
1969 | ||
1970 | static __u32 parity_segment_depth(struct imsm_dev *dev) | |
1971 | { | |
238c0a71 | 1972 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
1973 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
1974 | ||
1975 | switch(get_imsm_raid_level(map)) { | |
1976 | case 1: | |
1977 | case 10: | |
1978 | return chunk * map->num_domains; | |
1979 | case 5: | |
1980 | return chunk * map->num_members; | |
1981 | default: | |
1982 | return chunk; | |
1983 | } | |
1984 | } | |
1985 | ||
1986 | static __u32 map_migr_block(struct imsm_dev *dev, __u32 block) | |
1987 | { | |
238c0a71 | 1988 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
1989 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
1990 | __u32 strip = block / chunk; | |
1991 | ||
1992 | switch (get_imsm_raid_level(map)) { | |
1993 | case 1: | |
1994 | case 10: { | |
1995 | __u32 vol_strip = (strip * map->num_domains) + 1; | |
1996 | __u32 vol_stripe = vol_strip / map->num_members; | |
1997 | ||
1998 | return vol_stripe * chunk + block % chunk; | |
1999 | } case 5: { | |
2000 | __u32 stripe = strip / (map->num_members - 1); | |
2001 | ||
2002 | return stripe * chunk + block % chunk; | |
2003 | } | |
2004 | default: | |
2005 | return 0; | |
2006 | } | |
2007 | } | |
2008 | ||
c47b0ff6 AK |
2009 | static __u64 blocks_per_migr_unit(struct intel_super *super, |
2010 | struct imsm_dev *dev) | |
1e5c6983 DW |
2011 | { |
2012 | /* calculate the conversion factor between per member 'blocks' | |
2013 | * (md/{resync,rebuild}_start) and imsm migration units, return | |
2014 | * 0 for the 'not migrating' and 'unsupported migration' cases | |
2015 | */ | |
2016 | if (!dev->vol.migr_state) | |
2017 | return 0; | |
2018 | ||
2019 | switch (migr_type(dev)) { | |
c47b0ff6 AK |
2020 | case MIGR_GEN_MIGR: { |
2021 | struct migr_record *migr_rec = super->migr_rec; | |
2022 | return __le32_to_cpu(migr_rec->blocks_per_unit); | |
2023 | } | |
1e5c6983 DW |
2024 | case MIGR_VERIFY: |
2025 | case MIGR_REPAIR: | |
2026 | case MIGR_INIT: { | |
238c0a71 | 2027 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2028 | __u32 stripes_per_unit; |
2029 | __u32 blocks_per_unit; | |
2030 | __u32 parity_depth; | |
2031 | __u32 migr_chunk; | |
2032 | __u32 block_map; | |
2033 | __u32 block_rel; | |
2034 | __u32 segment; | |
2035 | __u32 stripe; | |
2036 | __u8 disks; | |
2037 | ||
2038 | /* yes, this is really the translation of migr_units to | |
2039 | * per-member blocks in the 'resync' case | |
2040 | */ | |
2041 | stripes_per_unit = num_stripes_per_unit_resync(dev); | |
2042 | migr_chunk = migr_strip_blocks_resync(dev); | |
238c0a71 | 2043 | disks = imsm_num_data_members(dev, MAP_0); |
1e5c6983 | 2044 | blocks_per_unit = stripes_per_unit * migr_chunk * disks; |
7b1ab482 | 2045 | stripe = __le16_to_cpu(map->blocks_per_strip) * disks; |
1e5c6983 DW |
2046 | segment = blocks_per_unit / stripe; |
2047 | block_rel = blocks_per_unit - segment * stripe; | |
2048 | parity_depth = parity_segment_depth(dev); | |
2049 | block_map = map_migr_block(dev, block_rel); | |
2050 | return block_map + parity_depth * segment; | |
2051 | } | |
2052 | case MIGR_REBUILD: { | |
2053 | __u32 stripes_per_unit; | |
2054 | __u32 migr_chunk; | |
2055 | ||
2056 | stripes_per_unit = num_stripes_per_unit_rebuild(dev); | |
2057 | migr_chunk = migr_strip_blocks_rebuild(dev); | |
2058 | return migr_chunk * stripes_per_unit; | |
2059 | } | |
1e5c6983 DW |
2060 | case MIGR_STATE_CHANGE: |
2061 | default: | |
2062 | return 0; | |
2063 | } | |
2064 | } | |
2065 | ||
c2c087e6 DW |
2066 | static int imsm_level_to_layout(int level) |
2067 | { | |
2068 | switch (level) { | |
2069 | case 0: | |
2070 | case 1: | |
2071 | return 0; | |
2072 | case 5: | |
2073 | case 6: | |
a380c027 | 2074 | return ALGORITHM_LEFT_ASYMMETRIC; |
c2c087e6 | 2075 | case 10: |
c92a2527 | 2076 | return 0x102; |
c2c087e6 | 2077 | } |
a18a888e | 2078 | return UnSet; |
c2c087e6 DW |
2079 | } |
2080 | ||
8e59f3d8 AK |
2081 | /******************************************************************************* |
2082 | * Function: read_imsm_migr_rec | |
2083 | * Description: Function reads imsm migration record from last sector of disk | |
2084 | * Parameters: | |
2085 | * fd : disk descriptor | |
2086 | * super : metadata info | |
2087 | * Returns: | |
2088 | * 0 : success, | |
2089 | * -1 : fail | |
2090 | ******************************************************************************/ | |
2091 | static int read_imsm_migr_rec(int fd, struct intel_super *super) | |
2092 | { | |
2093 | int ret_val = -1; | |
2094 | unsigned long long dsize; | |
2095 | ||
2096 | get_dev_size(fd, NULL, &dsize); | |
17a4eaf9 | 2097 | if (lseek64(fd, dsize - MIGR_REC_POSITION, SEEK_SET) < 0) { |
8e59f3d8 AK |
2098 | fprintf(stderr, |
2099 | Name ": Cannot seek to anchor block: %s\n", | |
2100 | strerror(errno)); | |
2101 | goto out; | |
2102 | } | |
17a4eaf9 AK |
2103 | if (read(fd, super->migr_rec_buf, MIGR_REC_BUF_SIZE) != |
2104 | MIGR_REC_BUF_SIZE) { | |
8e59f3d8 AK |
2105 | fprintf(stderr, |
2106 | Name ": Cannot read migr record block: %s\n", | |
2107 | strerror(errno)); | |
2108 | goto out; | |
2109 | } | |
2110 | ret_val = 0; | |
2111 | ||
2112 | out: | |
2113 | return ret_val; | |
2114 | } | |
2115 | ||
3136abe5 AK |
2116 | static struct imsm_dev *imsm_get_device_during_migration( |
2117 | struct intel_super *super) | |
2118 | { | |
2119 | ||
2120 | struct intel_dev *dv; | |
2121 | ||
2122 | for (dv = super->devlist; dv; dv = dv->next) { | |
2123 | if (is_gen_migration(dv->dev)) | |
2124 | return dv->dev; | |
2125 | } | |
2126 | return NULL; | |
2127 | } | |
2128 | ||
8e59f3d8 AK |
2129 | /******************************************************************************* |
2130 | * Function: load_imsm_migr_rec | |
2131 | * Description: Function reads imsm migration record (it is stored at the last | |
2132 | * sector of disk) | |
2133 | * Parameters: | |
2134 | * super : imsm internal array info | |
2135 | * info : general array info | |
2136 | * Returns: | |
2137 | * 0 : success | |
2138 | * -1 : fail | |
4c965cc9 | 2139 | * -2 : no migration in progress |
8e59f3d8 AK |
2140 | ******************************************************************************/ |
2141 | static int load_imsm_migr_rec(struct intel_super *super, struct mdinfo *info) | |
2142 | { | |
2143 | struct mdinfo *sd; | |
2144 | struct dl *dl = NULL; | |
2145 | char nm[30]; | |
2146 | int retval = -1; | |
2147 | int fd = -1; | |
3136abe5 AK |
2148 | struct imsm_dev *dev; |
2149 | struct imsm_map *map = NULL; | |
b4ab44d8 | 2150 | int slot = -1; |
3136abe5 AK |
2151 | |
2152 | /* find map under migration */ | |
2153 | dev = imsm_get_device_during_migration(super); | |
2154 | /* nothing to load,no migration in progress? | |
2155 | */ | |
2156 | if (dev == NULL) | |
4c965cc9 | 2157 | return -2; |
44bfe6df | 2158 | map = get_imsm_map(dev, MAP_0); |
8e59f3d8 AK |
2159 | |
2160 | if (info) { | |
2161 | for (sd = info->devs ; sd ; sd = sd->next) { | |
3136abe5 AK |
2162 | /* skip spare and failed disks |
2163 | */ | |
2164 | if (sd->disk.raid_disk < 0) | |
2165 | continue; | |
8e59f3d8 | 2166 | /* read only from one of the first two slots */ |
3136abe5 AK |
2167 | if (map) |
2168 | slot = get_imsm_disk_slot(map, | |
2169 | sd->disk.raid_disk); | |
2170 | if ((map == NULL) || (slot > 1) || (slot < 0)) | |
8e59f3d8 | 2171 | continue; |
3136abe5 | 2172 | |
8e59f3d8 AK |
2173 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); |
2174 | fd = dev_open(nm, O_RDONLY); | |
2175 | if (fd >= 0) | |
2176 | break; | |
2177 | } | |
2178 | } | |
2179 | if (fd < 0) { | |
2180 | for (dl = super->disks; dl; dl = dl->next) { | |
3136abe5 AK |
2181 | /* skip spare and failed disks |
2182 | */ | |
2183 | if (dl->index < 0) | |
2184 | continue; | |
8e59f3d8 | 2185 | /* read only from one of the first two slots */ |
3136abe5 AK |
2186 | if (map) |
2187 | slot = get_imsm_disk_slot(map, dl->index); | |
2188 | if ((map == NULL) || (slot > 1) || (slot < 0)) | |
8e59f3d8 AK |
2189 | continue; |
2190 | sprintf(nm, "%d:%d", dl->major, dl->minor); | |
2191 | fd = dev_open(nm, O_RDONLY); | |
2192 | if (fd >= 0) | |
2193 | break; | |
2194 | } | |
2195 | } | |
2196 | if (fd < 0) | |
2197 | goto out; | |
2198 | retval = read_imsm_migr_rec(fd, super); | |
2199 | ||
2200 | out: | |
2201 | if (fd >= 0) | |
2202 | close(fd); | |
2203 | return retval; | |
2204 | } | |
2205 | ||
9e2d750d | 2206 | #ifndef MDASSEMBLE |
c17608ea AK |
2207 | /******************************************************************************* |
2208 | * function: imsm_create_metadata_checkpoint_update | |
2209 | * Description: It creates update for checkpoint change. | |
2210 | * Parameters: | |
2211 | * super : imsm internal array info | |
2212 | * u : pointer to prepared update | |
2213 | * Returns: | |
2214 | * Uptate length. | |
2215 | * If length is equal to 0, input pointer u contains no update | |
2216 | ******************************************************************************/ | |
2217 | static int imsm_create_metadata_checkpoint_update( | |
2218 | struct intel_super *super, | |
2219 | struct imsm_update_general_migration_checkpoint **u) | |
2220 | { | |
2221 | ||
2222 | int update_memory_size = 0; | |
2223 | ||
2224 | dprintf("imsm_create_metadata_checkpoint_update(enter)\n"); | |
2225 | ||
2226 | if (u == NULL) | |
2227 | return 0; | |
2228 | *u = NULL; | |
2229 | ||
2230 | /* size of all update data without anchor */ | |
2231 | update_memory_size = | |
2232 | sizeof(struct imsm_update_general_migration_checkpoint); | |
2233 | ||
2234 | *u = calloc(1, update_memory_size); | |
2235 | if (*u == NULL) { | |
2236 | dprintf("error: cannot get memory for " | |
2237 | "imsm_create_metadata_checkpoint_update update\n"); | |
2238 | return 0; | |
2239 | } | |
2240 | (*u)->type = update_general_migration_checkpoint; | |
2241 | (*u)->curr_migr_unit = __le32_to_cpu(super->migr_rec->curr_migr_unit); | |
2242 | dprintf("imsm_create_metadata_checkpoint_update: prepared for %u\n", | |
2243 | (*u)->curr_migr_unit); | |
2244 | ||
2245 | return update_memory_size; | |
2246 | } | |
2247 | ||
2248 | ||
2249 | static void imsm_update_metadata_locally(struct supertype *st, | |
2250 | void *buf, int len); | |
2251 | ||
687629c2 AK |
2252 | /******************************************************************************* |
2253 | * Function: write_imsm_migr_rec | |
2254 | * Description: Function writes imsm migration record | |
2255 | * (at the last sector of disk) | |
2256 | * Parameters: | |
2257 | * super : imsm internal array info | |
2258 | * Returns: | |
2259 | * 0 : success | |
2260 | * -1 : if fail | |
2261 | ******************************************************************************/ | |
2262 | static int write_imsm_migr_rec(struct supertype *st) | |
2263 | { | |
2264 | struct intel_super *super = st->sb; | |
2265 | unsigned long long dsize; | |
2266 | char nm[30]; | |
2267 | int fd = -1; | |
2268 | int retval = -1; | |
2269 | struct dl *sd; | |
c17608ea AK |
2270 | int len; |
2271 | struct imsm_update_general_migration_checkpoint *u; | |
3136abe5 AK |
2272 | struct imsm_dev *dev; |
2273 | struct imsm_map *map = NULL; | |
2274 | ||
2275 | /* find map under migration */ | |
2276 | dev = imsm_get_device_during_migration(super); | |
2277 | /* if no migration, write buffer anyway to clear migr_record | |
2278 | * on disk based on first available device | |
2279 | */ | |
2280 | if (dev == NULL) | |
2281 | dev = get_imsm_dev(super, super->current_vol < 0 ? 0 : | |
2282 | super->current_vol); | |
2283 | ||
44bfe6df | 2284 | map = get_imsm_map(dev, MAP_0); |
687629c2 AK |
2285 | |
2286 | for (sd = super->disks ; sd ; sd = sd->next) { | |
b4ab44d8 | 2287 | int slot = -1; |
3136abe5 AK |
2288 | |
2289 | /* skip failed and spare devices */ | |
2290 | if (sd->index < 0) | |
2291 | continue; | |
687629c2 | 2292 | /* write to 2 first slots only */ |
3136abe5 AK |
2293 | if (map) |
2294 | slot = get_imsm_disk_slot(map, sd->index); | |
2295 | if ((map == NULL) || (slot > 1) || (slot < 0)) | |
687629c2 | 2296 | continue; |
3136abe5 | 2297 | |
687629c2 AK |
2298 | sprintf(nm, "%d:%d", sd->major, sd->minor); |
2299 | fd = dev_open(nm, O_RDWR); | |
2300 | if (fd < 0) | |
2301 | continue; | |
2302 | get_dev_size(fd, NULL, &dsize); | |
17a4eaf9 | 2303 | if (lseek64(fd, dsize - MIGR_REC_POSITION, SEEK_SET) < 0) { |
687629c2 AK |
2304 | fprintf(stderr, |
2305 | Name ": Cannot seek to anchor block: %s\n", | |
2306 | strerror(errno)); | |
2307 | goto out; | |
2308 | } | |
17a4eaf9 AK |
2309 | if (write(fd, super->migr_rec_buf, MIGR_REC_BUF_SIZE) != |
2310 | MIGR_REC_BUF_SIZE) { | |
687629c2 AK |
2311 | fprintf(stderr, |
2312 | Name ": Cannot write migr record block: %s\n", | |
2313 | strerror(errno)); | |
2314 | goto out; | |
2315 | } | |
2316 | close(fd); | |
2317 | fd = -1; | |
2318 | } | |
c17608ea AK |
2319 | /* update checkpoint information in metadata */ |
2320 | len = imsm_create_metadata_checkpoint_update(super, &u); | |
2321 | ||
2322 | if (len <= 0) { | |
2323 | dprintf("imsm: Cannot prepare update\n"); | |
2324 | goto out; | |
2325 | } | |
2326 | /* update metadata locally */ | |
2327 | imsm_update_metadata_locally(st, u, len); | |
2328 | /* and possibly remotely */ | |
2329 | if (st->update_tail) { | |
2330 | append_metadata_update(st, u, len); | |
2331 | /* during reshape we do all work inside metadata handler | |
2332 | * manage_reshape(), so metadata update has to be triggered | |
2333 | * insida it | |
2334 | */ | |
2335 | flush_metadata_updates(st); | |
2336 | st->update_tail = &st->updates; | |
2337 | } else | |
2338 | free(u); | |
687629c2 AK |
2339 | |
2340 | retval = 0; | |
2341 | out: | |
2342 | if (fd >= 0) | |
2343 | close(fd); | |
2344 | return retval; | |
2345 | } | |
9e2d750d | 2346 | #endif /* MDASSEMBLE */ |
687629c2 | 2347 | |
e2962bfc AK |
2348 | /* spare/missing disks activations are not allowe when |
2349 | * array/container performs reshape operation, because | |
2350 | * all arrays in container works on the same disks set | |
2351 | */ | |
2352 | int imsm_reshape_blocks_arrays_changes(struct intel_super *super) | |
2353 | { | |
2354 | int rv = 0; | |
2355 | struct intel_dev *i_dev; | |
2356 | struct imsm_dev *dev; | |
2357 | ||
2358 | /* check whole container | |
2359 | */ | |
2360 | for (i_dev = super->devlist; i_dev; i_dev = i_dev->next) { | |
2361 | dev = i_dev->dev; | |
3ad25638 | 2362 | if (is_gen_migration(dev)) { |
e2962bfc AK |
2363 | /* No repair during any migration in container |
2364 | */ | |
2365 | rv = 1; | |
2366 | break; | |
2367 | } | |
2368 | } | |
2369 | return rv; | |
2370 | } | |
2371 | ||
a5d85af7 | 2372 | static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap) |
bf5a934a DW |
2373 | { |
2374 | struct intel_super *super = st->sb; | |
c47b0ff6 | 2375 | struct migr_record *migr_rec = super->migr_rec; |
949c47a0 | 2376 | struct imsm_dev *dev = get_imsm_dev(super, super->current_vol); |
238c0a71 AK |
2377 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
2378 | struct imsm_map *prev_map = get_imsm_map(dev, MAP_1); | |
b335e593 | 2379 | struct imsm_map *map_to_analyse = map; |
efb30e7f | 2380 | struct dl *dl; |
e207da2f | 2381 | char *devname; |
139dae11 | 2382 | unsigned int component_size_alligment; |
a5d85af7 | 2383 | int map_disks = info->array.raid_disks; |
bf5a934a | 2384 | |
95eeceeb | 2385 | memset(info, 0, sizeof(*info)); |
b335e593 AK |
2386 | if (prev_map) |
2387 | map_to_analyse = prev_map; | |
2388 | ||
ca0748fa | 2389 | dl = super->current_disk; |
9894ec0d | 2390 | |
bf5a934a | 2391 | info->container_member = super->current_vol; |
cd0430a1 | 2392 | info->array.raid_disks = map->num_members; |
b335e593 | 2393 | info->array.level = get_imsm_raid_level(map_to_analyse); |
bf5a934a DW |
2394 | info->array.layout = imsm_level_to_layout(info->array.level); |
2395 | info->array.md_minor = -1; | |
2396 | info->array.ctime = 0; | |
2397 | info->array.utime = 0; | |
b335e593 AK |
2398 | info->array.chunk_size = |
2399 | __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9; | |
301406c9 | 2400 | info->array.state = !dev->vol.dirty; |
da9b4a62 DW |
2401 | info->custom_array_size = __le32_to_cpu(dev->size_high); |
2402 | info->custom_array_size <<= 32; | |
2403 | info->custom_array_size |= __le32_to_cpu(dev->size_low); | |
3ad25638 AK |
2404 | info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb); |
2405 | ||
3f510843 | 2406 | if (is_gen_migration(dev)) { |
3f83228a | 2407 | info->reshape_active = 1; |
b335e593 AK |
2408 | info->new_level = get_imsm_raid_level(map); |
2409 | info->new_layout = imsm_level_to_layout(info->new_level); | |
2410 | info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9; | |
3f83228a | 2411 | info->delta_disks = map->num_members - prev_map->num_members; |
493f5dd6 N |
2412 | if (info->delta_disks) { |
2413 | /* this needs to be applied to every array | |
2414 | * in the container. | |
2415 | */ | |
81219e70 | 2416 | info->reshape_active = CONTAINER_RESHAPE; |
493f5dd6 | 2417 | } |
3f83228a N |
2418 | /* We shape information that we give to md might have to be |
2419 | * modify to cope with md's requirement for reshaping arrays. | |
2420 | * For example, when reshaping a RAID0, md requires it to be | |
2421 | * presented as a degraded RAID4. | |
2422 | * Also if a RAID0 is migrating to a RAID5 we need to specify | |
2423 | * the array as already being RAID5, but the 'before' layout | |
2424 | * is a RAID4-like layout. | |
2425 | */ | |
2426 | switch (info->array.level) { | |
2427 | case 0: | |
2428 | switch(info->new_level) { | |
2429 | case 0: | |
2430 | /* conversion is happening as RAID4 */ | |
2431 | info->array.level = 4; | |
2432 | info->array.raid_disks += 1; | |
2433 | break; | |
2434 | case 5: | |
2435 | /* conversion is happening as RAID5 */ | |
2436 | info->array.level = 5; | |
2437 | info->array.layout = ALGORITHM_PARITY_N; | |
3f83228a N |
2438 | info->delta_disks -= 1; |
2439 | break; | |
2440 | default: | |
2441 | /* FIXME error message */ | |
2442 | info->array.level = UnSet; | |
2443 | break; | |
2444 | } | |
2445 | break; | |
2446 | } | |
b335e593 AK |
2447 | } else { |
2448 | info->new_level = UnSet; | |
2449 | info->new_layout = UnSet; | |
2450 | info->new_chunk = info->array.chunk_size; | |
3f83228a | 2451 | info->delta_disks = 0; |
b335e593 | 2452 | } |
ca0748fa | 2453 | |
efb30e7f DW |
2454 | if (dl) { |
2455 | info->disk.major = dl->major; | |
2456 | info->disk.minor = dl->minor; | |
ca0748fa | 2457 | info->disk.number = dl->index; |
656b6b5a N |
2458 | info->disk.raid_disk = get_imsm_disk_slot(map_to_analyse, |
2459 | dl->index); | |
efb30e7f | 2460 | } |
bf5a934a | 2461 | |
b335e593 AK |
2462 | info->data_offset = __le32_to_cpu(map_to_analyse->pba_of_lba0); |
2463 | info->component_size = | |
2464 | __le32_to_cpu(map_to_analyse->blocks_per_member); | |
139dae11 AK |
2465 | |
2466 | /* check component size aligment | |
2467 | */ | |
2468 | component_size_alligment = | |
2469 | info->component_size % (info->array.chunk_size/512); | |
2470 | ||
2471 | if (component_size_alligment && | |
2472 | (info->array.level != 1) && (info->array.level != UnSet)) { | |
2473 | dprintf("imsm: reported component size alligned from %llu ", | |
2474 | info->component_size); | |
2475 | info->component_size -= component_size_alligment; | |
2476 | dprintf("to %llu (%i).\n", | |
2477 | info->component_size, component_size_alligment); | |
2478 | } | |
2479 | ||
301406c9 | 2480 | memset(info->uuid, 0, sizeof(info->uuid)); |
921d9e16 | 2481 | info->recovery_start = MaxSector; |
bf5a934a | 2482 | |
d2e6d5d6 | 2483 | info->reshape_progress = 0; |
b6796ce1 | 2484 | info->resync_start = MaxSector; |
b9172665 AK |
2485 | if ((map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED || |
2486 | dev->vol.dirty) && | |
2487 | imsm_reshape_blocks_arrays_changes(super) == 0) { | |
301406c9 | 2488 | info->resync_start = 0; |
b6796ce1 AK |
2489 | } |
2490 | if (dev->vol.migr_state) { | |
1e5c6983 DW |
2491 | switch (migr_type(dev)) { |
2492 | case MIGR_REPAIR: | |
2493 | case MIGR_INIT: { | |
c47b0ff6 AK |
2494 | __u64 blocks_per_unit = blocks_per_migr_unit(super, |
2495 | dev); | |
1e5c6983 DW |
2496 | __u64 units = __le32_to_cpu(dev->vol.curr_migr_unit); |
2497 | ||
2498 | info->resync_start = blocks_per_unit * units; | |
2499 | break; | |
2500 | } | |
d2e6d5d6 | 2501 | case MIGR_GEN_MIGR: { |
c47b0ff6 AK |
2502 | __u64 blocks_per_unit = blocks_per_migr_unit(super, |
2503 | dev); | |
2504 | __u64 units = __le32_to_cpu(migr_rec->curr_migr_unit); | |
04fa9523 AK |
2505 | unsigned long long array_blocks; |
2506 | int used_disks; | |
d2e6d5d6 | 2507 | |
befb629b AK |
2508 | if (__le32_to_cpu(migr_rec->ascending_migr) && |
2509 | (units < | |
2510 | (__le32_to_cpu(migr_rec->num_migr_units)-1)) && | |
2511 | (super->migr_rec->rec_status == | |
2512 | __cpu_to_le32(UNIT_SRC_IN_CP_AREA))) | |
2513 | units++; | |
2514 | ||
d2e6d5d6 | 2515 | info->reshape_progress = blocks_per_unit * units; |
6289d1e0 | 2516 | |
d2e6d5d6 AK |
2517 | dprintf("IMSM: General Migration checkpoint : %llu " |
2518 | "(%llu) -> read reshape progress : %llu\n", | |
19986c72 MB |
2519 | (unsigned long long)units, |
2520 | (unsigned long long)blocks_per_unit, | |
2521 | info->reshape_progress); | |
75156c46 | 2522 | |
238c0a71 | 2523 | used_disks = imsm_num_data_members(dev, MAP_1); |
75156c46 AK |
2524 | if (used_disks > 0) { |
2525 | array_blocks = map->blocks_per_member * | |
2526 | used_disks; | |
2527 | /* round array size down to closest MB | |
2528 | */ | |
2529 | info->custom_array_size = (array_blocks | |
2530 | >> SECT_PER_MB_SHIFT) | |
2531 | << SECT_PER_MB_SHIFT; | |
2532 | } | |
d2e6d5d6 | 2533 | } |
1e5c6983 DW |
2534 | case MIGR_VERIFY: |
2535 | /* we could emulate the checkpointing of | |
2536 | * 'sync_action=check' migrations, but for now | |
2537 | * we just immediately complete them | |
2538 | */ | |
2539 | case MIGR_REBUILD: | |
2540 | /* this is handled by container_content_imsm() */ | |
1e5c6983 DW |
2541 | case MIGR_STATE_CHANGE: |
2542 | /* FIXME handle other migrations */ | |
2543 | default: | |
2544 | /* we are not dirty, so... */ | |
2545 | info->resync_start = MaxSector; | |
2546 | } | |
b6796ce1 | 2547 | } |
301406c9 DW |
2548 | |
2549 | strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN); | |
2550 | info->name[MAX_RAID_SERIAL_LEN] = 0; | |
bf5a934a | 2551 | |
f35f2525 N |
2552 | info->array.major_version = -1; |
2553 | info->array.minor_version = -2; | |
e207da2f AW |
2554 | devname = devnum2devname(st->container_dev); |
2555 | *info->text_version = '\0'; | |
2556 | if (devname) | |
2557 | sprintf(info->text_version, "/%s/%d", devname, info->container_member); | |
2558 | free(devname); | |
a67dd8cc | 2559 | info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */ |
51006d85 | 2560 | uuid_from_super_imsm(st, info->uuid); |
a5d85af7 N |
2561 | |
2562 | if (dmap) { | |
2563 | int i, j; | |
2564 | for (i=0; i<map_disks; i++) { | |
2565 | dmap[i] = 0; | |
2566 | if (i < info->array.raid_disks) { | |
2567 | struct imsm_disk *dsk; | |
238c0a71 | 2568 | j = get_imsm_disk_idx(dev, i, MAP_X); |
a5d85af7 N |
2569 | dsk = get_imsm_disk(super, j); |
2570 | if (dsk && (dsk->status & CONFIGURED_DISK)) | |
2571 | dmap[i] = 1; | |
2572 | } | |
2573 | } | |
2574 | } | |
81ac8b4d | 2575 | } |
bf5a934a | 2576 | |
3b451610 AK |
2577 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, |
2578 | int failed, int look_in_map); | |
2579 | ||
2580 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev, | |
2581 | int look_in_map); | |
2582 | ||
b4ab44d8 AK |
2583 | |
2584 | #ifndef MDASSEMBLE | |
3b451610 AK |
2585 | static void manage_second_map(struct intel_super *super, struct imsm_dev *dev) |
2586 | { | |
2587 | if (is_gen_migration(dev)) { | |
2588 | int failed; | |
2589 | __u8 map_state; | |
2590 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); | |
2591 | ||
2592 | failed = imsm_count_failed(super, dev, MAP_1); | |
238c0a71 | 2593 | map_state = imsm_check_degraded(super, dev, failed, MAP_1); |
3b451610 AK |
2594 | if (map2->map_state != map_state) { |
2595 | map2->map_state = map_state; | |
2596 | super->updates_pending++; | |
2597 | } | |
2598 | } | |
2599 | } | |
b4ab44d8 | 2600 | #endif |
97b4d0e9 DW |
2601 | |
2602 | static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index) | |
2603 | { | |
2604 | struct dl *d; | |
2605 | ||
2606 | for (d = super->missing; d; d = d->next) | |
2607 | if (d->index == index) | |
2608 | return &d->disk; | |
2609 | return NULL; | |
2610 | } | |
2611 | ||
a5d85af7 | 2612 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map) |
4f5bc454 DW |
2613 | { |
2614 | struct intel_super *super = st->sb; | |
4f5bc454 | 2615 | struct imsm_disk *disk; |
a5d85af7 | 2616 | int map_disks = info->array.raid_disks; |
ab3cb6b3 N |
2617 | int max_enough = -1; |
2618 | int i; | |
2619 | struct imsm_super *mpb; | |
4f5bc454 | 2620 | |
bf5a934a | 2621 | if (super->current_vol >= 0) { |
a5d85af7 | 2622 | getinfo_super_imsm_volume(st, info, map); |
bf5a934a DW |
2623 | return; |
2624 | } | |
95eeceeb | 2625 | memset(info, 0, sizeof(*info)); |
d23fe947 DW |
2626 | |
2627 | /* Set raid_disks to zero so that Assemble will always pull in valid | |
2628 | * spares | |
2629 | */ | |
2630 | info->array.raid_disks = 0; | |
cdddbdbc DW |
2631 | info->array.level = LEVEL_CONTAINER; |
2632 | info->array.layout = 0; | |
2633 | info->array.md_minor = -1; | |
c2c087e6 | 2634 | info->array.ctime = 0; /* N/A for imsm */ |
cdddbdbc DW |
2635 | info->array.utime = 0; |
2636 | info->array.chunk_size = 0; | |
2637 | ||
2638 | info->disk.major = 0; | |
2639 | info->disk.minor = 0; | |
cdddbdbc | 2640 | info->disk.raid_disk = -1; |
c2c087e6 | 2641 | info->reshape_active = 0; |
f35f2525 N |
2642 | info->array.major_version = -1; |
2643 | info->array.minor_version = -2; | |
c2c087e6 | 2644 | strcpy(info->text_version, "imsm"); |
a67dd8cc | 2645 | info->safe_mode_delay = 0; |
c2c087e6 DW |
2646 | info->disk.number = -1; |
2647 | info->disk.state = 0; | |
c5afc314 | 2648 | info->name[0] = 0; |
921d9e16 | 2649 | info->recovery_start = MaxSector; |
3ad25638 | 2650 | info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb); |
c2c087e6 | 2651 | |
97b4d0e9 | 2652 | /* do we have the all the insync disks that we expect? */ |
ab3cb6b3 | 2653 | mpb = super->anchor; |
97b4d0e9 | 2654 | |
ab3cb6b3 N |
2655 | for (i = 0; i < mpb->num_raid_devs; i++) { |
2656 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
2657 | int failed, enough, j, missing = 0; | |
2658 | struct imsm_map *map; | |
2659 | __u8 state; | |
97b4d0e9 | 2660 | |
3b451610 AK |
2661 | failed = imsm_count_failed(super, dev, MAP_0); |
2662 | state = imsm_check_degraded(super, dev, failed, MAP_0); | |
238c0a71 | 2663 | map = get_imsm_map(dev, MAP_0); |
ab3cb6b3 N |
2664 | |
2665 | /* any newly missing disks? | |
2666 | * (catches single-degraded vs double-degraded) | |
2667 | */ | |
2668 | for (j = 0; j < map->num_members; j++) { | |
238c0a71 | 2669 | __u32 ord = get_imsm_ord_tbl_ent(dev, j, MAP_0); |
ab3cb6b3 N |
2670 | __u32 idx = ord_to_idx(ord); |
2671 | ||
2672 | if (!(ord & IMSM_ORD_REBUILD) && | |
2673 | get_imsm_missing(super, idx)) { | |
2674 | missing = 1; | |
2675 | break; | |
2676 | } | |
97b4d0e9 | 2677 | } |
ab3cb6b3 N |
2678 | |
2679 | if (state == IMSM_T_STATE_FAILED) | |
2680 | enough = -1; | |
2681 | else if (state == IMSM_T_STATE_DEGRADED && | |
2682 | (state != map->map_state || missing)) | |
2683 | enough = 0; | |
2684 | else /* we're normal, or already degraded */ | |
2685 | enough = 1; | |
2686 | ||
2687 | /* in the missing/failed disk case check to see | |
2688 | * if at least one array is runnable | |
2689 | */ | |
2690 | max_enough = max(max_enough, enough); | |
2691 | } | |
2692 | dprintf("%s: enough: %d\n", __func__, max_enough); | |
2693 | info->container_enough = max_enough; | |
97b4d0e9 | 2694 | |
4a04ec6c | 2695 | if (super->disks) { |
14e8215b DW |
2696 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
2697 | ||
b9f594fe | 2698 | disk = &super->disks->disk; |
14e8215b DW |
2699 | info->data_offset = __le32_to_cpu(disk->total_blocks) - reserved; |
2700 | info->component_size = reserved; | |
25ed7e59 | 2701 | info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0; |
df474657 DW |
2702 | /* we don't change info->disk.raid_disk here because |
2703 | * this state will be finalized in mdmon after we have | |
2704 | * found the 'most fresh' version of the metadata | |
2705 | */ | |
25ed7e59 DW |
2706 | info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; |
2707 | info->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC); | |
cdddbdbc | 2708 | } |
a575e2a7 DW |
2709 | |
2710 | /* only call uuid_from_super_imsm when this disk is part of a populated container, | |
2711 | * ->compare_super may have updated the 'num_raid_devs' field for spares | |
2712 | */ | |
2713 | if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs) | |
36ba7d48 | 2714 | uuid_from_super_imsm(st, info->uuid); |
22e263f6 AC |
2715 | else |
2716 | memcpy(info->uuid, uuid_zero, sizeof(uuid_zero)); | |
a5d85af7 N |
2717 | |
2718 | /* I don't know how to compute 'map' on imsm, so use safe default */ | |
2719 | if (map) { | |
2720 | int i; | |
2721 | for (i = 0; i < map_disks; i++) | |
2722 | map[i] = 1; | |
2723 | } | |
2724 | ||
cdddbdbc DW |
2725 | } |
2726 | ||
5c4cd5da AC |
2727 | /* allocates memory and fills disk in mdinfo structure |
2728 | * for each disk in array */ | |
2729 | struct mdinfo *getinfo_super_disks_imsm(struct supertype *st) | |
2730 | { | |
2731 | struct mdinfo *mddev = NULL; | |
2732 | struct intel_super *super = st->sb; | |
2733 | struct imsm_disk *disk; | |
2734 | int count = 0; | |
2735 | struct dl *dl; | |
2736 | if (!super || !super->disks) | |
2737 | return NULL; | |
2738 | dl = super->disks; | |
2739 | mddev = malloc(sizeof(*mddev)); | |
2740 | if (!mddev) { | |
2741 | fprintf(stderr, Name ": Failed to allocate memory.\n"); | |
2742 | return NULL; | |
2743 | } | |
2744 | memset(mddev, 0, sizeof(*mddev)); | |
2745 | while (dl) { | |
2746 | struct mdinfo *tmp; | |
2747 | disk = &dl->disk; | |
2748 | tmp = malloc(sizeof(*tmp)); | |
2749 | if (!tmp) { | |
2750 | fprintf(stderr, Name ": Failed to allocate memory.\n"); | |
2751 | if (mddev) | |
2752 | sysfs_free(mddev); | |
2753 | return NULL; | |
2754 | } | |
2755 | memset(tmp, 0, sizeof(*tmp)); | |
2756 | if (mddev->devs) | |
2757 | tmp->next = mddev->devs; | |
2758 | mddev->devs = tmp; | |
2759 | tmp->disk.number = count++; | |
2760 | tmp->disk.major = dl->major; | |
2761 | tmp->disk.minor = dl->minor; | |
2762 | tmp->disk.state = is_configured(disk) ? | |
2763 | (1 << MD_DISK_ACTIVE) : 0; | |
2764 | tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; | |
2765 | tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC); | |
2766 | tmp->disk.raid_disk = -1; | |
2767 | dl = dl->next; | |
2768 | } | |
2769 | return mddev; | |
2770 | } | |
2771 | ||
cdddbdbc DW |
2772 | static int update_super_imsm(struct supertype *st, struct mdinfo *info, |
2773 | char *update, char *devname, int verbose, | |
2774 | int uuid_set, char *homehost) | |
2775 | { | |
f352c545 DW |
2776 | /* For 'assemble' and 'force' we need to return non-zero if any |
2777 | * change was made. For others, the return value is ignored. | |
2778 | * Update options are: | |
2779 | * force-one : This device looks a bit old but needs to be included, | |
2780 | * update age info appropriately. | |
2781 | * assemble: clear any 'faulty' flag to allow this device to | |
2782 | * be assembled. | |
2783 | * force-array: Array is degraded but being forced, mark it clean | |
2784 | * if that will be needed to assemble it. | |
2785 | * | |
2786 | * newdev: not used ???? | |
2787 | * grow: Array has gained a new device - this is currently for | |
2788 | * linear only | |
2789 | * resync: mark as dirty so a resync will happen. | |
2790 | * name: update the name - preserving the homehost | |
6e46bf34 | 2791 | * uuid: Change the uuid of the array to match watch is given |
f352c545 DW |
2792 | * |
2793 | * Following are not relevant for this imsm: | |
2794 | * sparc2.2 : update from old dodgey metadata | |
2795 | * super-minor: change the preferred_minor number | |
2796 | * summaries: update redundant counters. | |
f352c545 DW |
2797 | * homehost: update the recorded homehost |
2798 | * _reshape_progress: record new reshape_progress position. | |
2799 | */ | |
6e46bf34 DW |
2800 | int rv = 1; |
2801 | struct intel_super *super = st->sb; | |
2802 | struct imsm_super *mpb; | |
f352c545 | 2803 | |
6e46bf34 DW |
2804 | /* we can only update container info */ |
2805 | if (!super || super->current_vol >= 0 || !super->anchor) | |
2806 | return 1; | |
2807 | ||
2808 | mpb = super->anchor; | |
2809 | ||
81a5b4f5 N |
2810 | if (strcmp(update, "uuid") == 0) { |
2811 | /* We take this to mean that the family_num should be updated. | |
2812 | * However that is much smaller than the uuid so we cannot really | |
2813 | * allow an explicit uuid to be given. And it is hard to reliably | |
2814 | * know if one was. | |
2815 | * So if !uuid_set we know the current uuid is random and just used | |
2816 | * the first 'int' and copy it to the other 3 positions. | |
2817 | * Otherwise we require the 4 'int's to be the same as would be the | |
2818 | * case if we are using a random uuid. So an explicit uuid will be | |
2819 | * accepted as long as all for ints are the same... which shouldn't hurt | |
6e46bf34 | 2820 | */ |
81a5b4f5 N |
2821 | if (!uuid_set) { |
2822 | info->uuid[1] = info->uuid[2] = info->uuid[3] = info->uuid[0]; | |
6e46bf34 | 2823 | rv = 0; |
81a5b4f5 N |
2824 | } else { |
2825 | if (info->uuid[0] != info->uuid[1] || | |
2826 | info->uuid[1] != info->uuid[2] || | |
2827 | info->uuid[2] != info->uuid[3]) | |
2828 | rv = -1; | |
2829 | else | |
2830 | rv = 0; | |
6e46bf34 | 2831 | } |
81a5b4f5 N |
2832 | if (rv == 0) |
2833 | mpb->orig_family_num = info->uuid[0]; | |
6e46bf34 DW |
2834 | } else if (strcmp(update, "assemble") == 0) |
2835 | rv = 0; | |
2836 | else | |
1e2b2765 | 2837 | rv = -1; |
f352c545 | 2838 | |
6e46bf34 DW |
2839 | /* successful update? recompute checksum */ |
2840 | if (rv == 0) | |
2841 | mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb)); | |
f352c545 DW |
2842 | |
2843 | return rv; | |
cdddbdbc DW |
2844 | } |
2845 | ||
c2c087e6 | 2846 | static size_t disks_to_mpb_size(int disks) |
cdddbdbc | 2847 | { |
c2c087e6 | 2848 | size_t size; |
cdddbdbc | 2849 | |
c2c087e6 DW |
2850 | size = sizeof(struct imsm_super); |
2851 | size += (disks - 1) * sizeof(struct imsm_disk); | |
2852 | size += 2 * sizeof(struct imsm_dev); | |
2853 | /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */ | |
2854 | size += (4 - 2) * sizeof(struct imsm_map); | |
2855 | /* 4 possible disk_ord_tbl's */ | |
2856 | size += 4 * (disks - 1) * sizeof(__u32); | |
2857 | ||
2858 | return size; | |
2859 | } | |
2860 | ||
2861 | static __u64 avail_size_imsm(struct supertype *st, __u64 devsize) | |
2862 | { | |
2863 | if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS)) | |
2864 | return 0; | |
2865 | ||
2866 | return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS); | |
cdddbdbc DW |
2867 | } |
2868 | ||
ba2de7ba DW |
2869 | static void free_devlist(struct intel_super *super) |
2870 | { | |
2871 | struct intel_dev *dv; | |
2872 | ||
2873 | while (super->devlist) { | |
2874 | dv = super->devlist->next; | |
2875 | free(super->devlist->dev); | |
2876 | free(super->devlist); | |
2877 | super->devlist = dv; | |
2878 | } | |
2879 | } | |
2880 | ||
2881 | static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src) | |
2882 | { | |
2883 | memcpy(dest, src, sizeof_imsm_dev(src, 0)); | |
2884 | } | |
2885 | ||
cdddbdbc DW |
2886 | static int compare_super_imsm(struct supertype *st, struct supertype *tst) |
2887 | { | |
2888 | /* | |
2889 | * return: | |
2890 | * 0 same, or first was empty, and second was copied | |
2891 | * 1 second had wrong number | |
2892 | * 2 wrong uuid | |
2893 | * 3 wrong other info | |
2894 | */ | |
2895 | struct intel_super *first = st->sb; | |
2896 | struct intel_super *sec = tst->sb; | |
2897 | ||
2898 | if (!first) { | |
2899 | st->sb = tst->sb; | |
2900 | tst->sb = NULL; | |
2901 | return 0; | |
2902 | } | |
8603ea6f LM |
2903 | /* in platform dependent environment test if the disks |
2904 | * use the same Intel hba | |
2905 | */ | |
2906 | if (!check_env("IMSM_NO_PLATFORM")) { | |
ea2bc72b LM |
2907 | if (!first->hba || !sec->hba || |
2908 | (first->hba->type != sec->hba->type)) { | |
8603ea6f LM |
2909 | fprintf(stderr, |
2910 | "HBAs of devices does not match %s != %s\n", | |
ea2bc72b LM |
2911 | first->hba ? get_sys_dev_type(first->hba->type) : NULL, |
2912 | sec->hba ? get_sys_dev_type(sec->hba->type) : NULL); | |
8603ea6f LM |
2913 | return 3; |
2914 | } | |
2915 | } | |
cdddbdbc | 2916 | |
d23fe947 DW |
2917 | /* if an anchor does not have num_raid_devs set then it is a free |
2918 | * floating spare | |
2919 | */ | |
2920 | if (first->anchor->num_raid_devs > 0 && | |
2921 | sec->anchor->num_raid_devs > 0) { | |
a2b97981 DW |
2922 | /* Determine if these disks might ever have been |
2923 | * related. Further disambiguation can only take place | |
2924 | * in load_super_imsm_all | |
2925 | */ | |
2926 | __u32 first_family = first->anchor->orig_family_num; | |
2927 | __u32 sec_family = sec->anchor->orig_family_num; | |
2928 | ||
f796af5d DW |
2929 | if (memcmp(first->anchor->sig, sec->anchor->sig, |
2930 | MAX_SIGNATURE_LENGTH) != 0) | |
2931 | return 3; | |
2932 | ||
a2b97981 DW |
2933 | if (first_family == 0) |
2934 | first_family = first->anchor->family_num; | |
2935 | if (sec_family == 0) | |
2936 | sec_family = sec->anchor->family_num; | |
2937 | ||
2938 | if (first_family != sec_family) | |
d23fe947 | 2939 | return 3; |
f796af5d | 2940 | |
d23fe947 | 2941 | } |
cdddbdbc | 2942 | |
f796af5d | 2943 | |
3e372e5a DW |
2944 | /* if 'first' is a spare promote it to a populated mpb with sec's |
2945 | * family number | |
2946 | */ | |
2947 | if (first->anchor->num_raid_devs == 0 && | |
2948 | sec->anchor->num_raid_devs > 0) { | |
78d30f94 | 2949 | int i; |
ba2de7ba DW |
2950 | struct intel_dev *dv; |
2951 | struct imsm_dev *dev; | |
78d30f94 DW |
2952 | |
2953 | /* we need to copy raid device info from sec if an allocation | |
2954 | * fails here we don't associate the spare | |
2955 | */ | |
2956 | for (i = 0; i < sec->anchor->num_raid_devs; i++) { | |
ba2de7ba DW |
2957 | dv = malloc(sizeof(*dv)); |
2958 | if (!dv) | |
2959 | break; | |
2960 | dev = malloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1)); | |
2961 | if (!dev) { | |
2962 | free(dv); | |
2963 | break; | |
78d30f94 | 2964 | } |
ba2de7ba DW |
2965 | dv->dev = dev; |
2966 | dv->index = i; | |
2967 | dv->next = first->devlist; | |
2968 | first->devlist = dv; | |
78d30f94 | 2969 | } |
709743c5 | 2970 | if (i < sec->anchor->num_raid_devs) { |
ba2de7ba DW |
2971 | /* allocation failure */ |
2972 | free_devlist(first); | |
2973 | fprintf(stderr, "imsm: failed to associate spare\n"); | |
2974 | return 3; | |
78d30f94 | 2975 | } |
3e372e5a | 2976 | first->anchor->num_raid_devs = sec->anchor->num_raid_devs; |
148acb7b | 2977 | first->anchor->orig_family_num = sec->anchor->orig_family_num; |
3e372e5a | 2978 | first->anchor->family_num = sec->anchor->family_num; |
ac6449be | 2979 | memcpy(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH); |
709743c5 DW |
2980 | for (i = 0; i < sec->anchor->num_raid_devs; i++) |
2981 | imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i)); | |
3e372e5a DW |
2982 | } |
2983 | ||
cdddbdbc DW |
2984 | return 0; |
2985 | } | |
2986 | ||
0030e8d6 DW |
2987 | static void fd2devname(int fd, char *name) |
2988 | { | |
2989 | struct stat st; | |
2990 | char path[256]; | |
33a6535d | 2991 | char dname[PATH_MAX]; |
0030e8d6 DW |
2992 | char *nm; |
2993 | int rv; | |
2994 | ||
2995 | name[0] = '\0'; | |
2996 | if (fstat(fd, &st) != 0) | |
2997 | return; | |
2998 | sprintf(path, "/sys/dev/block/%d:%d", | |
2999 | major(st.st_rdev), minor(st.st_rdev)); | |
3000 | ||
9cf014ec | 3001 | rv = readlink(path, dname, sizeof(dname)-1); |
0030e8d6 DW |
3002 | if (rv <= 0) |
3003 | return; | |
9587c373 | 3004 | |
0030e8d6 DW |
3005 | dname[rv] = '\0'; |
3006 | nm = strrchr(dname, '/'); | |
7897de29 JS |
3007 | if (nm) { |
3008 | nm++; | |
3009 | snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm); | |
3010 | } | |
0030e8d6 DW |
3011 | } |
3012 | ||
cdddbdbc DW |
3013 | extern int scsi_get_serial(int fd, void *buf, size_t buf_len); |
3014 | ||
3015 | static int imsm_read_serial(int fd, char *devname, | |
3016 | __u8 serial[MAX_RAID_SERIAL_LEN]) | |
3017 | { | |
3018 | unsigned char scsi_serial[255]; | |
cdddbdbc DW |
3019 | int rv; |
3020 | int rsp_len; | |
1f24f035 | 3021 | int len; |
316e2bf4 DW |
3022 | char *dest; |
3023 | char *src; | |
3024 | char *rsp_buf; | |
3025 | int i; | |
cdddbdbc DW |
3026 | |
3027 | memset(scsi_serial, 0, sizeof(scsi_serial)); | |
cdddbdbc | 3028 | |
f9ba0ff1 DW |
3029 | rv = scsi_get_serial(fd, scsi_serial, sizeof(scsi_serial)); |
3030 | ||
40ebbb9c | 3031 | if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) { |
f9ba0ff1 DW |
3032 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
3033 | fd2devname(fd, (char *) serial); | |
0030e8d6 DW |
3034 | return 0; |
3035 | } | |
3036 | ||
cdddbdbc DW |
3037 | if (rv != 0) { |
3038 | if (devname) | |
3039 | fprintf(stderr, | |
3040 | Name ": Failed to retrieve serial for %s\n", | |
3041 | devname); | |
3042 | return rv; | |
3043 | } | |
3044 | ||
3045 | rsp_len = scsi_serial[3]; | |
03cd4cc8 DW |
3046 | if (!rsp_len) { |
3047 | if (devname) | |
3048 | fprintf(stderr, | |
3049 | Name ": Failed to retrieve serial for %s\n", | |
3050 | devname); | |
3051 | return 2; | |
3052 | } | |
1f24f035 | 3053 | rsp_buf = (char *) &scsi_serial[4]; |
5c3db629 | 3054 | |
316e2bf4 DW |
3055 | /* trim all whitespace and non-printable characters and convert |
3056 | * ':' to ';' | |
3057 | */ | |
3058 | for (i = 0, dest = rsp_buf; i < rsp_len; i++) { | |
3059 | src = &rsp_buf[i]; | |
3060 | if (*src > 0x20) { | |
3061 | /* ':' is reserved for use in placeholder serial | |
3062 | * numbers for missing disks | |
3063 | */ | |
3064 | if (*src == ':') | |
3065 | *dest++ = ';'; | |
3066 | else | |
3067 | *dest++ = *src; | |
3068 | } | |
3069 | } | |
3070 | len = dest - rsp_buf; | |
3071 | dest = rsp_buf; | |
3072 | ||
3073 | /* truncate leading characters */ | |
3074 | if (len > MAX_RAID_SERIAL_LEN) { | |
3075 | dest += len - MAX_RAID_SERIAL_LEN; | |
1f24f035 | 3076 | len = MAX_RAID_SERIAL_LEN; |
316e2bf4 | 3077 | } |
5c3db629 | 3078 | |
5c3db629 | 3079 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
316e2bf4 | 3080 | memcpy(serial, dest, len); |
cdddbdbc DW |
3081 | |
3082 | return 0; | |
3083 | } | |
3084 | ||
1f24f035 DW |
3085 | static int serialcmp(__u8 *s1, __u8 *s2) |
3086 | { | |
3087 | return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN); | |
3088 | } | |
3089 | ||
3090 | static void serialcpy(__u8 *dest, __u8 *src) | |
3091 | { | |
3092 | strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN); | |
3093 | } | |
3094 | ||
54c2c1ea DW |
3095 | static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super) |
3096 | { | |
3097 | struct dl *dl; | |
3098 | ||
3099 | for (dl = super->disks; dl; dl = dl->next) | |
3100 | if (serialcmp(dl->serial, serial) == 0) | |
3101 | break; | |
3102 | ||
3103 | return dl; | |
3104 | } | |
3105 | ||
a2b97981 DW |
3106 | static struct imsm_disk * |
3107 | __serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx) | |
3108 | { | |
3109 | int i; | |
3110 | ||
3111 | for (i = 0; i < mpb->num_disks; i++) { | |
3112 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
3113 | ||
3114 | if (serialcmp(disk->serial, serial) == 0) { | |
3115 | if (idx) | |
3116 | *idx = i; | |
3117 | return disk; | |
3118 | } | |
3119 | } | |
3120 | ||
3121 | return NULL; | |
3122 | } | |
3123 | ||
cdddbdbc DW |
3124 | static int |
3125 | load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd) | |
3126 | { | |
a2b97981 | 3127 | struct imsm_disk *disk; |
cdddbdbc DW |
3128 | struct dl *dl; |
3129 | struct stat stb; | |
cdddbdbc | 3130 | int rv; |
a2b97981 | 3131 | char name[40]; |
d23fe947 DW |
3132 | __u8 serial[MAX_RAID_SERIAL_LEN]; |
3133 | ||
3134 | rv = imsm_read_serial(fd, devname, serial); | |
3135 | ||
3136 | if (rv != 0) | |
3137 | return 2; | |
3138 | ||
a2b97981 | 3139 | dl = calloc(1, sizeof(*dl)); |
b9f594fe | 3140 | if (!dl) { |
cdddbdbc DW |
3141 | if (devname) |
3142 | fprintf(stderr, | |
3143 | Name ": failed to allocate disk buffer for %s\n", | |
3144 | devname); | |
3145 | return 2; | |
3146 | } | |
cdddbdbc | 3147 | |
a2b97981 DW |
3148 | fstat(fd, &stb); |
3149 | dl->major = major(stb.st_rdev); | |
3150 | dl->minor = minor(stb.st_rdev); | |
3151 | dl->next = super->disks; | |
3152 | dl->fd = keep_fd ? fd : -1; | |
3153 | assert(super->disks == NULL); | |
3154 | super->disks = dl; | |
3155 | serialcpy(dl->serial, serial); | |
3156 | dl->index = -2; | |
3157 | dl->e = NULL; | |
3158 | fd2devname(fd, name); | |
3159 | if (devname) | |
3160 | dl->devname = strdup(devname); | |
3161 | else | |
3162 | dl->devname = strdup(name); | |
cdddbdbc | 3163 | |
d23fe947 | 3164 | /* look up this disk's index in the current anchor */ |
a2b97981 DW |
3165 | disk = __serial_to_disk(dl->serial, super->anchor, &dl->index); |
3166 | if (disk) { | |
3167 | dl->disk = *disk; | |
3168 | /* only set index on disks that are a member of a | |
3169 | * populated contianer, i.e. one with raid_devs | |
3170 | */ | |
3171 | if (is_failed(&dl->disk)) | |
3f6efecc | 3172 | dl->index = -2; |
a2b97981 DW |
3173 | else if (is_spare(&dl->disk)) |
3174 | dl->index = -1; | |
3f6efecc DW |
3175 | } |
3176 | ||
949c47a0 DW |
3177 | return 0; |
3178 | } | |
3179 | ||
0e600426 | 3180 | #ifndef MDASSEMBLE |
0c046afd DW |
3181 | /* When migrating map0 contains the 'destination' state while map1 |
3182 | * contains the current state. When not migrating map0 contains the | |
3183 | * current state. This routine assumes that map[0].map_state is set to | |
3184 | * the current array state before being called. | |
3185 | * | |
3186 | * Migration is indicated by one of the following states | |
3187 | * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed) | |
e3bba0e0 | 3188 | * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal |
0c046afd | 3189 | * map1state=unitialized) |
1484e727 | 3190 | * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal |
0c046afd | 3191 | * map1state=normal) |
e3bba0e0 | 3192 | * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal |
0c046afd | 3193 | * map1state=degraded) |
8e59f3d8 AK |
3194 | * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal |
3195 | * map1state=normal) | |
0c046afd | 3196 | */ |
8e59f3d8 AK |
3197 | static void migrate(struct imsm_dev *dev, struct intel_super *super, |
3198 | __u8 to_state, int migr_type) | |
3393c6af | 3199 | { |
0c046afd | 3200 | struct imsm_map *dest; |
238c0a71 | 3201 | struct imsm_map *src = get_imsm_map(dev, MAP_0); |
3393c6af | 3202 | |
0c046afd | 3203 | dev->vol.migr_state = 1; |
1484e727 | 3204 | set_migr_type(dev, migr_type); |
f8f603f1 | 3205 | dev->vol.curr_migr_unit = 0; |
238c0a71 | 3206 | dest = get_imsm_map(dev, MAP_1); |
0c046afd | 3207 | |
0556e1a2 | 3208 | /* duplicate and then set the target end state in map[0] */ |
3393c6af | 3209 | memcpy(dest, src, sizeof_imsm_map(src)); |
28bce06f AK |
3210 | if ((migr_type == MIGR_REBUILD) || |
3211 | (migr_type == MIGR_GEN_MIGR)) { | |
0556e1a2 DW |
3212 | __u32 ord; |
3213 | int i; | |
3214 | ||
3215 | for (i = 0; i < src->num_members; i++) { | |
3216 | ord = __le32_to_cpu(src->disk_ord_tbl[i]); | |
3217 | set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord)); | |
3218 | } | |
3219 | } | |
3220 | ||
8e59f3d8 AK |
3221 | if (migr_type == MIGR_GEN_MIGR) |
3222 | /* Clear migration record */ | |
3223 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
3224 | ||
0c046afd | 3225 | src->map_state = to_state; |
949c47a0 | 3226 | } |
f8f603f1 | 3227 | |
809da78e AK |
3228 | static void end_migration(struct imsm_dev *dev, struct intel_super *super, |
3229 | __u8 map_state) | |
f8f603f1 | 3230 | { |
238c0a71 AK |
3231 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
3232 | struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state == 0 ? | |
3233 | MAP_0 : MAP_1); | |
28bce06f | 3234 | int i, j; |
0556e1a2 DW |
3235 | |
3236 | /* merge any IMSM_ORD_REBUILD bits that were not successfully | |
3237 | * completed in the last migration. | |
3238 | * | |
28bce06f | 3239 | * FIXME add support for raid-level-migration |
0556e1a2 | 3240 | */ |
809da78e AK |
3241 | if ((map_state != map->map_state) && (is_gen_migration(dev) == 0) && |
3242 | (prev->map_state != IMSM_T_STATE_UNINITIALIZED)) { | |
3243 | /* when final map state is other than expected | |
3244 | * merge maps (not for migration) | |
3245 | */ | |
3246 | int failed; | |
3247 | ||
3248 | for (i = 0; i < prev->num_members; i++) | |
3249 | for (j = 0; j < map->num_members; j++) | |
3250 | /* during online capacity expansion | |
3251 | * disks position can be changed | |
3252 | * if takeover is used | |
3253 | */ | |
3254 | if (ord_to_idx(map->disk_ord_tbl[j]) == | |
3255 | ord_to_idx(prev->disk_ord_tbl[i])) { | |
3256 | map->disk_ord_tbl[j] |= | |
3257 | prev->disk_ord_tbl[i]; | |
3258 | break; | |
3259 | } | |
3260 | failed = imsm_count_failed(super, dev, MAP_0); | |
3261 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
3262 | } | |
f8f603f1 DW |
3263 | |
3264 | dev->vol.migr_state = 0; | |
ea672ee1 | 3265 | set_migr_type(dev, 0); |
f8f603f1 DW |
3266 | dev->vol.curr_migr_unit = 0; |
3267 | map->map_state = map_state; | |
3268 | } | |
0e600426 | 3269 | #endif |
949c47a0 DW |
3270 | |
3271 | static int parse_raid_devices(struct intel_super *super) | |
3272 | { | |
3273 | int i; | |
3274 | struct imsm_dev *dev_new; | |
4d7b1503 | 3275 | size_t len, len_migr; |
401d313b | 3276 | size_t max_len = 0; |
4d7b1503 DW |
3277 | size_t space_needed = 0; |
3278 | struct imsm_super *mpb = super->anchor; | |
949c47a0 DW |
3279 | |
3280 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
3281 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
ba2de7ba | 3282 | struct intel_dev *dv; |
949c47a0 | 3283 | |
4d7b1503 DW |
3284 | len = sizeof_imsm_dev(dev_iter, 0); |
3285 | len_migr = sizeof_imsm_dev(dev_iter, 1); | |
3286 | if (len_migr > len) | |
3287 | space_needed += len_migr - len; | |
ca9de185 | 3288 | |
ba2de7ba DW |
3289 | dv = malloc(sizeof(*dv)); |
3290 | if (!dv) | |
3291 | return 1; | |
401d313b AK |
3292 | if (max_len < len_migr) |
3293 | max_len = len_migr; | |
3294 | if (max_len > len_migr) | |
3295 | space_needed += max_len - len_migr; | |
3296 | dev_new = malloc(max_len); | |
ba2de7ba DW |
3297 | if (!dev_new) { |
3298 | free(dv); | |
949c47a0 | 3299 | return 1; |
ba2de7ba | 3300 | } |
949c47a0 | 3301 | imsm_copy_dev(dev_new, dev_iter); |
ba2de7ba DW |
3302 | dv->dev = dev_new; |
3303 | dv->index = i; | |
3304 | dv->next = super->devlist; | |
3305 | super->devlist = dv; | |
949c47a0 | 3306 | } |
cdddbdbc | 3307 | |
4d7b1503 DW |
3308 | /* ensure that super->buf is large enough when all raid devices |
3309 | * are migrating | |
3310 | */ | |
3311 | if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) { | |
3312 | void *buf; | |
3313 | ||
3314 | len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed, 512); | |
3315 | if (posix_memalign(&buf, 512, len) != 0) | |
3316 | return 1; | |
3317 | ||
1f45a8ad DW |
3318 | memcpy(buf, super->buf, super->len); |
3319 | memset(buf + super->len, 0, len - super->len); | |
4d7b1503 DW |
3320 | free(super->buf); |
3321 | super->buf = buf; | |
3322 | super->len = len; | |
3323 | } | |
ca9de185 | 3324 | |
cdddbdbc DW |
3325 | return 0; |
3326 | } | |
3327 | ||
604b746f JD |
3328 | /* retrieve a pointer to the bbm log which starts after all raid devices */ |
3329 | struct bbm_log *__get_imsm_bbm_log(struct imsm_super *mpb) | |
3330 | { | |
3331 | void *ptr = NULL; | |
3332 | ||
3333 | if (__le32_to_cpu(mpb->bbm_log_size)) { | |
3334 | ptr = mpb; | |
3335 | ptr += mpb->mpb_size - __le32_to_cpu(mpb->bbm_log_size); | |
3336 | } | |
3337 | ||
3338 | return ptr; | |
3339 | } | |
3340 | ||
e2f41b2c AK |
3341 | /******************************************************************************* |
3342 | * Function: check_mpb_migr_compatibility | |
3343 | * Description: Function checks for unsupported migration features: | |
3344 | * - migration optimization area (pba_of_lba0) | |
3345 | * - descending reshape (ascending_migr) | |
3346 | * Parameters: | |
3347 | * super : imsm metadata information | |
3348 | * Returns: | |
3349 | * 0 : migration is compatible | |
3350 | * -1 : migration is not compatible | |
3351 | ******************************************************************************/ | |
3352 | int check_mpb_migr_compatibility(struct intel_super *super) | |
3353 | { | |
3354 | struct imsm_map *map0, *map1; | |
3355 | struct migr_record *migr_rec = super->migr_rec; | |
3356 | int i; | |
3357 | ||
3358 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
3359 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
3360 | ||
3361 | if (dev_iter && | |
3362 | dev_iter->vol.migr_state == 1 && | |
3363 | dev_iter->vol.migr_type == MIGR_GEN_MIGR) { | |
3364 | /* This device is migrating */ | |
238c0a71 AK |
3365 | map0 = get_imsm_map(dev_iter, MAP_0); |
3366 | map1 = get_imsm_map(dev_iter, MAP_1); | |
e2f41b2c AK |
3367 | if (map0->pba_of_lba0 != map1->pba_of_lba0) |
3368 | /* migration optimization area was used */ | |
3369 | return -1; | |
3370 | if (migr_rec->ascending_migr == 0 | |
3371 | && migr_rec->dest_depth_per_unit > 0) | |
3372 | /* descending reshape not supported yet */ | |
3373 | return -1; | |
3374 | } | |
3375 | } | |
3376 | return 0; | |
3377 | } | |
3378 | ||
d23fe947 | 3379 | static void __free_imsm(struct intel_super *super, int free_disks); |
9ca2c81c | 3380 | |
cdddbdbc | 3381 | /* load_imsm_mpb - read matrix metadata |
f2f5c343 | 3382 | * allocates super->mpb to be freed by free_imsm |
cdddbdbc DW |
3383 | */ |
3384 | static int load_imsm_mpb(int fd, struct intel_super *super, char *devname) | |
3385 | { | |
3386 | unsigned long long dsize; | |
cdddbdbc DW |
3387 | unsigned long long sectors; |
3388 | struct stat; | |
6416d527 | 3389 | struct imsm_super *anchor; |
cdddbdbc DW |
3390 | __u32 check_sum; |
3391 | ||
cdddbdbc | 3392 | get_dev_size(fd, NULL, &dsize); |
64436f06 N |
3393 | if (dsize < 1024) { |
3394 | if (devname) | |
3395 | fprintf(stderr, | |
3396 | Name ": %s: device to small for imsm\n", | |
3397 | devname); | |
3398 | return 1; | |
3399 | } | |
cdddbdbc DW |
3400 | |
3401 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) { | |
3402 | if (devname) | |
2e062e82 AK |
3403 | fprintf(stderr, Name |
3404 | ": Cannot seek to anchor block on %s: %s\n", | |
cdddbdbc DW |
3405 | devname, strerror(errno)); |
3406 | return 1; | |
3407 | } | |
3408 | ||
949c47a0 | 3409 | if (posix_memalign((void**)&anchor, 512, 512) != 0) { |
ad97895e DW |
3410 | if (devname) |
3411 | fprintf(stderr, | |
3412 | Name ": Failed to allocate imsm anchor buffer" | |
3413 | " on %s\n", devname); | |
3414 | return 1; | |
3415 | } | |
949c47a0 | 3416 | if (read(fd, anchor, 512) != 512) { |
cdddbdbc DW |
3417 | if (devname) |
3418 | fprintf(stderr, | |
3419 | Name ": Cannot read anchor block on %s: %s\n", | |
3420 | devname, strerror(errno)); | |
6416d527 | 3421 | free(anchor); |
cdddbdbc DW |
3422 | return 1; |
3423 | } | |
3424 | ||
6416d527 | 3425 | if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) { |
cdddbdbc DW |
3426 | if (devname) |
3427 | fprintf(stderr, | |
3428 | Name ": no IMSM anchor on %s\n", devname); | |
6416d527 | 3429 | free(anchor); |
cdddbdbc DW |
3430 | return 2; |
3431 | } | |
3432 | ||
d23fe947 | 3433 | __free_imsm(super, 0); |
f2f5c343 LM |
3434 | /* reload capability and hba */ |
3435 | ||
3436 | /* capability and hba must be updated with new super allocation */ | |
d424212e | 3437 | find_intel_hba_capability(fd, super, devname); |
949c47a0 DW |
3438 | super->len = ROUND_UP(anchor->mpb_size, 512); |
3439 | if (posix_memalign(&super->buf, 512, super->len) != 0) { | |
cdddbdbc DW |
3440 | if (devname) |
3441 | fprintf(stderr, | |
3442 | Name ": unable to allocate %zu byte mpb buffer\n", | |
949c47a0 | 3443 | super->len); |
6416d527 | 3444 | free(anchor); |
cdddbdbc DW |
3445 | return 2; |
3446 | } | |
949c47a0 | 3447 | memcpy(super->buf, anchor, 512); |
cdddbdbc | 3448 | |
6416d527 NB |
3449 | sectors = mpb_sectors(anchor) - 1; |
3450 | free(anchor); | |
8e59f3d8 | 3451 | |
17a4eaf9 | 3452 | if (posix_memalign(&super->migr_rec_buf, 512, MIGR_REC_BUF_SIZE) != 0) { |
8e59f3d8 AK |
3453 | fprintf(stderr, Name |
3454 | ": %s could not allocate migr_rec buffer\n", __func__); | |
3455 | free(super->buf); | |
3456 | return 2; | |
3457 | } | |
3458 | ||
949c47a0 | 3459 | if (!sectors) { |
ecf45690 DW |
3460 | check_sum = __gen_imsm_checksum(super->anchor); |
3461 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
3462 | if (devname) | |
3463 | fprintf(stderr, | |
3464 | Name ": IMSM checksum %x != %x on %s\n", | |
3465 | check_sum, | |
3466 | __le32_to_cpu(super->anchor->check_sum), | |
3467 | devname); | |
3468 | return 2; | |
3469 | } | |
3470 | ||
a2b97981 | 3471 | return 0; |
949c47a0 | 3472 | } |
cdddbdbc DW |
3473 | |
3474 | /* read the extended mpb */ | |
3475 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) { | |
3476 | if (devname) | |
3477 | fprintf(stderr, | |
3478 | Name ": Cannot seek to extended mpb on %s: %s\n", | |
3479 | devname, strerror(errno)); | |
3480 | return 1; | |
3481 | } | |
3482 | ||
f21e18ca | 3483 | if ((unsigned)read(fd, super->buf + 512, super->len - 512) != super->len - 512) { |
cdddbdbc DW |
3484 | if (devname) |
3485 | fprintf(stderr, | |
3486 | Name ": Cannot read extended mpb on %s: %s\n", | |
3487 | devname, strerror(errno)); | |
3488 | return 2; | |
3489 | } | |
3490 | ||
949c47a0 DW |
3491 | check_sum = __gen_imsm_checksum(super->anchor); |
3492 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
cdddbdbc DW |
3493 | if (devname) |
3494 | fprintf(stderr, | |
3495 | Name ": IMSM checksum %x != %x on %s\n", | |
949c47a0 | 3496 | check_sum, __le32_to_cpu(super->anchor->check_sum), |
cdddbdbc | 3497 | devname); |
db575f3b | 3498 | return 3; |
cdddbdbc DW |
3499 | } |
3500 | ||
604b746f JD |
3501 | /* FIXME the BBM log is disk specific so we cannot use this global |
3502 | * buffer for all disks. Ok for now since we only look at the global | |
3503 | * bbm_log_size parameter to gate assembly | |
3504 | */ | |
3505 | super->bbm_log = __get_imsm_bbm_log(super->anchor); | |
3506 | ||
a2b97981 DW |
3507 | return 0; |
3508 | } | |
3509 | ||
8e59f3d8 AK |
3510 | static int read_imsm_migr_rec(int fd, struct intel_super *super); |
3511 | ||
a2b97981 DW |
3512 | static int |
3513 | load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd) | |
3514 | { | |
3515 | int err; | |
3516 | ||
3517 | err = load_imsm_mpb(fd, super, devname); | |
3518 | if (err) | |
3519 | return err; | |
3520 | err = load_imsm_disk(fd, super, devname, keep_fd); | |
3521 | if (err) | |
3522 | return err; | |
3523 | err = parse_raid_devices(super); | |
4d7b1503 | 3524 | |
a2b97981 | 3525 | return err; |
cdddbdbc DW |
3526 | } |
3527 | ||
ae6aad82 DW |
3528 | static void __free_imsm_disk(struct dl *d) |
3529 | { | |
3530 | if (d->fd >= 0) | |
3531 | close(d->fd); | |
3532 | if (d->devname) | |
3533 | free(d->devname); | |
0dcecb2e DW |
3534 | if (d->e) |
3535 | free(d->e); | |
ae6aad82 DW |
3536 | free(d); |
3537 | ||
3538 | } | |
1a64be56 | 3539 | |
cdddbdbc DW |
3540 | static void free_imsm_disks(struct intel_super *super) |
3541 | { | |
47ee5a45 | 3542 | struct dl *d; |
cdddbdbc | 3543 | |
47ee5a45 DW |
3544 | while (super->disks) { |
3545 | d = super->disks; | |
cdddbdbc | 3546 | super->disks = d->next; |
ae6aad82 | 3547 | __free_imsm_disk(d); |
cdddbdbc | 3548 | } |
cb82edca AK |
3549 | while (super->disk_mgmt_list) { |
3550 | d = super->disk_mgmt_list; | |
3551 | super->disk_mgmt_list = d->next; | |
3552 | __free_imsm_disk(d); | |
3553 | } | |
47ee5a45 DW |
3554 | while (super->missing) { |
3555 | d = super->missing; | |
3556 | super->missing = d->next; | |
3557 | __free_imsm_disk(d); | |
3558 | } | |
3559 | ||
cdddbdbc DW |
3560 | } |
3561 | ||
9ca2c81c | 3562 | /* free all the pieces hanging off of a super pointer */ |
d23fe947 | 3563 | static void __free_imsm(struct intel_super *super, int free_disks) |
cdddbdbc | 3564 | { |
88654014 LM |
3565 | struct intel_hba *elem, *next; |
3566 | ||
9ca2c81c | 3567 | if (super->buf) { |
949c47a0 | 3568 | free(super->buf); |
9ca2c81c DW |
3569 | super->buf = NULL; |
3570 | } | |
f2f5c343 LM |
3571 | /* unlink capability description */ |
3572 | super->orom = NULL; | |
8e59f3d8 AK |
3573 | if (super->migr_rec_buf) { |
3574 | free(super->migr_rec_buf); | |
3575 | super->migr_rec_buf = NULL; | |
3576 | } | |
d23fe947 DW |
3577 | if (free_disks) |
3578 | free_imsm_disks(super); | |
ba2de7ba | 3579 | free_devlist(super); |
88654014 LM |
3580 | elem = super->hba; |
3581 | while (elem) { | |
3582 | if (elem->path) | |
3583 | free((void *)elem->path); | |
3584 | next = elem->next; | |
3585 | free(elem); | |
3586 | elem = next; | |
88c32bb1 | 3587 | } |
88654014 | 3588 | super->hba = NULL; |
cdddbdbc DW |
3589 | } |
3590 | ||
9ca2c81c DW |
3591 | static void free_imsm(struct intel_super *super) |
3592 | { | |
d23fe947 | 3593 | __free_imsm(super, 1); |
9ca2c81c DW |
3594 | free(super); |
3595 | } | |
cdddbdbc DW |
3596 | |
3597 | static void free_super_imsm(struct supertype *st) | |
3598 | { | |
3599 | struct intel_super *super = st->sb; | |
3600 | ||
3601 | if (!super) | |
3602 | return; | |
3603 | ||
3604 | free_imsm(super); | |
3605 | st->sb = NULL; | |
3606 | } | |
3607 | ||
49133e57 | 3608 | static struct intel_super *alloc_super(void) |
c2c087e6 DW |
3609 | { |
3610 | struct intel_super *super = malloc(sizeof(*super)); | |
3611 | ||
3612 | if (super) { | |
3613 | memset(super, 0, sizeof(*super)); | |
bf5a934a | 3614 | super->current_vol = -1; |
0dcecb2e | 3615 | super->create_offset = ~((__u32 ) 0); |
c2c087e6 | 3616 | } |
c2c087e6 DW |
3617 | return super; |
3618 | } | |
3619 | ||
f0f5a016 LM |
3620 | /* |
3621 | * find and allocate hba and OROM/EFI based on valid fd of RAID component device | |
3622 | */ | |
d424212e | 3623 | static int find_intel_hba_capability(int fd, struct intel_super *super, char *devname) |
f0f5a016 LM |
3624 | { |
3625 | struct sys_dev *hba_name; | |
3626 | int rv = 0; | |
3627 | ||
3628 | if ((fd < 0) || check_env("IMSM_NO_PLATFORM")) { | |
f2f5c343 | 3629 | super->orom = NULL; |
f0f5a016 LM |
3630 | super->hba = NULL; |
3631 | return 0; | |
3632 | } | |
3633 | hba_name = find_disk_attached_hba(fd, NULL); | |
3634 | if (!hba_name) { | |
d424212e | 3635 | if (devname) |
f0f5a016 LM |
3636 | fprintf(stderr, |
3637 | Name ": %s is not attached to Intel(R) RAID controller.\n", | |
d424212e | 3638 | devname); |
f0f5a016 LM |
3639 | return 1; |
3640 | } | |
3641 | rv = attach_hba_to_super(super, hba_name); | |
3642 | if (rv == 2) { | |
d424212e N |
3643 | if (devname) { |
3644 | struct intel_hba *hba = super->hba; | |
f0f5a016 | 3645 | |
f0f5a016 LM |
3646 | fprintf(stderr, Name ": %s is attached to Intel(R) %s RAID " |
3647 | "controller (%s),\n" | |
3648 | " but the container is assigned to Intel(R) " | |
3649 | "%s RAID controller (", | |
d424212e | 3650 | devname, |
f0f5a016 LM |
3651 | hba_name->path, |
3652 | hba_name->pci_id ? : "Err!", | |
3653 | get_sys_dev_type(hba_name->type)); | |
3654 | ||
f0f5a016 LM |
3655 | while (hba) { |
3656 | fprintf(stderr, "%s", hba->pci_id ? : "Err!"); | |
3657 | if (hba->next) | |
3658 | fprintf(stderr, ", "); | |
3659 | hba = hba->next; | |
3660 | } | |
3661 | ||
3662 | fprintf(stderr, ").\n" | |
3663 | " Mixing devices attached to different controllers " | |
3664 | "is not allowed.\n"); | |
3665 | } | |
3666 | free_sys_dev(&hba_name); | |
3667 | return 2; | |
3668 | } | |
f2f5c343 | 3669 | super->orom = find_imsm_capability(hba_name->type); |
f0f5a016 | 3670 | free_sys_dev(&hba_name); |
f2f5c343 LM |
3671 | if (!super->orom) |
3672 | return 3; | |
f0f5a016 LM |
3673 | return 0; |
3674 | } | |
3675 | ||
47ee5a45 DW |
3676 | /* find_missing - helper routine for load_super_imsm_all that identifies |
3677 | * disks that have disappeared from the system. This routine relies on | |
3678 | * the mpb being uptodate, which it is at load time. | |
3679 | */ | |
3680 | static int find_missing(struct intel_super *super) | |
3681 | { | |
3682 | int i; | |
3683 | struct imsm_super *mpb = super->anchor; | |
3684 | struct dl *dl; | |
3685 | struct imsm_disk *disk; | |
47ee5a45 DW |
3686 | |
3687 | for (i = 0; i < mpb->num_disks; i++) { | |
3688 | disk = __get_imsm_disk(mpb, i); | |
54c2c1ea | 3689 | dl = serial_to_dl(disk->serial, super); |
47ee5a45 DW |
3690 | if (dl) |
3691 | continue; | |
47ee5a45 DW |
3692 | |
3693 | dl = malloc(sizeof(*dl)); | |
3694 | if (!dl) | |
3695 | return 1; | |
3696 | dl->major = 0; | |
3697 | dl->minor = 0; | |
3698 | dl->fd = -1; | |
3699 | dl->devname = strdup("missing"); | |
3700 | dl->index = i; | |
3701 | serialcpy(dl->serial, disk->serial); | |
3702 | dl->disk = *disk; | |
689c9bf3 | 3703 | dl->e = NULL; |
47ee5a45 DW |
3704 | dl->next = super->missing; |
3705 | super->missing = dl; | |
3706 | } | |
3707 | ||
3708 | return 0; | |
3709 | } | |
3710 | ||
3960e579 | 3711 | #ifndef MDASSEMBLE |
a2b97981 DW |
3712 | static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list) |
3713 | { | |
3714 | struct intel_disk *idisk = disk_list; | |
3715 | ||
3716 | while (idisk) { | |
3717 | if (serialcmp(idisk->disk.serial, serial) == 0) | |
3718 | break; | |
3719 | idisk = idisk->next; | |
3720 | } | |
3721 | ||
3722 | return idisk; | |
3723 | } | |
3724 | ||
3725 | static int __prep_thunderdome(struct intel_super **table, int tbl_size, | |
3726 | struct intel_super *super, | |
3727 | struct intel_disk **disk_list) | |
3728 | { | |
3729 | struct imsm_disk *d = &super->disks->disk; | |
3730 | struct imsm_super *mpb = super->anchor; | |
3731 | int i, j; | |
3732 | ||
3733 | for (i = 0; i < tbl_size; i++) { | |
3734 | struct imsm_super *tbl_mpb = table[i]->anchor; | |
3735 | struct imsm_disk *tbl_d = &table[i]->disks->disk; | |
3736 | ||
3737 | if (tbl_mpb->family_num == mpb->family_num) { | |
3738 | if (tbl_mpb->check_sum == mpb->check_sum) { | |
3739 | dprintf("%s: mpb from %d:%d matches %d:%d\n", | |
3740 | __func__, super->disks->major, | |
3741 | super->disks->minor, | |
3742 | table[i]->disks->major, | |
3743 | table[i]->disks->minor); | |
3744 | break; | |
3745 | } | |
3746 | ||
3747 | if (((is_configured(d) && !is_configured(tbl_d)) || | |
3748 | is_configured(d) == is_configured(tbl_d)) && | |
3749 | tbl_mpb->generation_num < mpb->generation_num) { | |
3750 | /* current version of the mpb is a | |
3751 | * better candidate than the one in | |
3752 | * super_table, but copy over "cross | |
3753 | * generational" status | |
3754 | */ | |
3755 | struct intel_disk *idisk; | |
3756 | ||
3757 | dprintf("%s: mpb from %d:%d replaces %d:%d\n", | |
3758 | __func__, super->disks->major, | |
3759 | super->disks->minor, | |
3760 | table[i]->disks->major, | |
3761 | table[i]->disks->minor); | |
3762 | ||
3763 | idisk = disk_list_get(tbl_d->serial, *disk_list); | |
3764 | if (idisk && is_failed(&idisk->disk)) | |
3765 | tbl_d->status |= FAILED_DISK; | |
3766 | break; | |
3767 | } else { | |
3768 | struct intel_disk *idisk; | |
3769 | struct imsm_disk *disk; | |
3770 | ||
3771 | /* tbl_mpb is more up to date, but copy | |
3772 | * over cross generational status before | |
3773 | * returning | |
3774 | */ | |
3775 | disk = __serial_to_disk(d->serial, mpb, NULL); | |
3776 | if (disk && is_failed(disk)) | |
3777 | d->status |= FAILED_DISK; | |
3778 | ||
3779 | idisk = disk_list_get(d->serial, *disk_list); | |
3780 | if (idisk) { | |
3781 | idisk->owner = i; | |
3782 | if (disk && is_configured(disk)) | |
3783 | idisk->disk.status |= CONFIGURED_DISK; | |
3784 | } | |
3785 | ||
3786 | dprintf("%s: mpb from %d:%d prefer %d:%d\n", | |
3787 | __func__, super->disks->major, | |
3788 | super->disks->minor, | |
3789 | table[i]->disks->major, | |
3790 | table[i]->disks->minor); | |
3791 | ||
3792 | return tbl_size; | |
3793 | } | |
3794 | } | |
3795 | } | |
3796 | ||
3797 | if (i >= tbl_size) | |
3798 | table[tbl_size++] = super; | |
3799 | else | |
3800 | table[i] = super; | |
3801 | ||
3802 | /* update/extend the merged list of imsm_disk records */ | |
3803 | for (j = 0; j < mpb->num_disks; j++) { | |
3804 | struct imsm_disk *disk = __get_imsm_disk(mpb, j); | |
3805 | struct intel_disk *idisk; | |
3806 | ||
3807 | idisk = disk_list_get(disk->serial, *disk_list); | |
3808 | if (idisk) { | |
3809 | idisk->disk.status |= disk->status; | |
3810 | if (is_configured(&idisk->disk) || | |
3811 | is_failed(&idisk->disk)) | |
3812 | idisk->disk.status &= ~(SPARE_DISK); | |
3813 | } else { | |
3814 | idisk = calloc(1, sizeof(*idisk)); | |
3815 | if (!idisk) | |
3816 | return -1; | |
3817 | idisk->owner = IMSM_UNKNOWN_OWNER; | |
3818 | idisk->disk = *disk; | |
3819 | idisk->next = *disk_list; | |
3820 | *disk_list = idisk; | |
3821 | } | |
3822 | ||
3823 | if (serialcmp(idisk->disk.serial, d->serial) == 0) | |
3824 | idisk->owner = i; | |
3825 | } | |
3826 | ||
3827 | return tbl_size; | |
3828 | } | |
3829 | ||
3830 | static struct intel_super * | |
3831 | validate_members(struct intel_super *super, struct intel_disk *disk_list, | |
3832 | const int owner) | |
3833 | { | |
3834 | struct imsm_super *mpb = super->anchor; | |
3835 | int ok_count = 0; | |
3836 | int i; | |
3837 | ||
3838 | for (i = 0; i < mpb->num_disks; i++) { | |
3839 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
3840 | struct intel_disk *idisk; | |
3841 | ||
3842 | idisk = disk_list_get(disk->serial, disk_list); | |
3843 | if (idisk) { | |
3844 | if (idisk->owner == owner || | |
3845 | idisk->owner == IMSM_UNKNOWN_OWNER) | |
3846 | ok_count++; | |
3847 | else | |
3848 | dprintf("%s: '%.16s' owner %d != %d\n", | |
3849 | __func__, disk->serial, idisk->owner, | |
3850 | owner); | |
3851 | } else { | |
3852 | dprintf("%s: unknown disk %x [%d]: %.16s\n", | |
3853 | __func__, __le32_to_cpu(mpb->family_num), i, | |
3854 | disk->serial); | |
3855 | break; | |
3856 | } | |
3857 | } | |
3858 | ||
3859 | if (ok_count == mpb->num_disks) | |
3860 | return super; | |
3861 | return NULL; | |
3862 | } | |
3863 | ||
3864 | static void show_conflicts(__u32 family_num, struct intel_super *super_list) | |
3865 | { | |
3866 | struct intel_super *s; | |
3867 | ||
3868 | for (s = super_list; s; s = s->next) { | |
3869 | if (family_num != s->anchor->family_num) | |
3870 | continue; | |
3871 | fprintf(stderr, "Conflict, offlining family %#x on '%s'\n", | |
3872 | __le32_to_cpu(family_num), s->disks->devname); | |
3873 | } | |
3874 | } | |
3875 | ||
3876 | static struct intel_super * | |
3877 | imsm_thunderdome(struct intel_super **super_list, int len) | |
3878 | { | |
3879 | struct intel_super *super_table[len]; | |
3880 | struct intel_disk *disk_list = NULL; | |
3881 | struct intel_super *champion, *spare; | |
3882 | struct intel_super *s, **del; | |
3883 | int tbl_size = 0; | |
3884 | int conflict; | |
3885 | int i; | |
3886 | ||
3887 | memset(super_table, 0, sizeof(super_table)); | |
3888 | for (s = *super_list; s; s = s->next) | |
3889 | tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list); | |
3890 | ||
3891 | for (i = 0; i < tbl_size; i++) { | |
3892 | struct imsm_disk *d; | |
3893 | struct intel_disk *idisk; | |
3894 | struct imsm_super *mpb = super_table[i]->anchor; | |
3895 | ||
3896 | s = super_table[i]; | |
3897 | d = &s->disks->disk; | |
3898 | ||
3899 | /* 'd' must appear in merged disk list for its | |
3900 | * configuration to be valid | |
3901 | */ | |
3902 | idisk = disk_list_get(d->serial, disk_list); | |
3903 | if (idisk && idisk->owner == i) | |
3904 | s = validate_members(s, disk_list, i); | |
3905 | else | |
3906 | s = NULL; | |
3907 | ||
3908 | if (!s) | |
3909 | dprintf("%s: marking family: %#x from %d:%d offline\n", | |
3910 | __func__, mpb->family_num, | |
3911 | super_table[i]->disks->major, | |
3912 | super_table[i]->disks->minor); | |
3913 | super_table[i] = s; | |
3914 | } | |
3915 | ||
3916 | /* This is where the mdadm implementation differs from the Windows | |
3917 | * driver which has no strict concept of a container. We can only | |
3918 | * assemble one family from a container, so when returning a prodigal | |
3919 | * array member to this system the code will not be able to disambiguate | |
3920 | * the container contents that should be assembled ("foreign" versus | |
3921 | * "local"). It requires user intervention to set the orig_family_num | |
3922 | * to a new value to establish a new container. The Windows driver in | |
3923 | * this situation fixes up the volume name in place and manages the | |
3924 | * foreign array as an independent entity. | |
3925 | */ | |
3926 | s = NULL; | |
3927 | spare = NULL; | |
3928 | conflict = 0; | |
3929 | for (i = 0; i < tbl_size; i++) { | |
3930 | struct intel_super *tbl_ent = super_table[i]; | |
3931 | int is_spare = 0; | |
3932 | ||
3933 | if (!tbl_ent) | |
3934 | continue; | |
3935 | ||
3936 | if (tbl_ent->anchor->num_raid_devs == 0) { | |
3937 | spare = tbl_ent; | |
3938 | is_spare = 1; | |
3939 | } | |
3940 | ||
3941 | if (s && !is_spare) { | |
3942 | show_conflicts(tbl_ent->anchor->family_num, *super_list); | |
3943 | conflict++; | |
3944 | } else if (!s && !is_spare) | |
3945 | s = tbl_ent; | |
3946 | } | |
3947 | ||
3948 | if (!s) | |
3949 | s = spare; | |
3950 | if (!s) { | |
3951 | champion = NULL; | |
3952 | goto out; | |
3953 | } | |
3954 | champion = s; | |
3955 | ||
3956 | if (conflict) | |
3957 | fprintf(stderr, "Chose family %#x on '%s', " | |
3958 | "assemble conflicts to new container with '--update=uuid'\n", | |
3959 | __le32_to_cpu(s->anchor->family_num), s->disks->devname); | |
3960 | ||
3961 | /* collect all dl's onto 'champion', and update them to | |
3962 | * champion's version of the status | |
3963 | */ | |
3964 | for (s = *super_list; s; s = s->next) { | |
3965 | struct imsm_super *mpb = champion->anchor; | |
3966 | struct dl *dl = s->disks; | |
3967 | ||
3968 | if (s == champion) | |
3969 | continue; | |
3970 | ||
3971 | for (i = 0; i < mpb->num_disks; i++) { | |
3972 | struct imsm_disk *disk; | |
3973 | ||
3974 | disk = __serial_to_disk(dl->serial, mpb, &dl->index); | |
3975 | if (disk) { | |
3976 | dl->disk = *disk; | |
3977 | /* only set index on disks that are a member of | |
3978 | * a populated contianer, i.e. one with | |
3979 | * raid_devs | |
3980 | */ | |
3981 | if (is_failed(&dl->disk)) | |
3982 | dl->index = -2; | |
3983 | else if (is_spare(&dl->disk)) | |
3984 | dl->index = -1; | |
3985 | break; | |
3986 | } | |
3987 | } | |
3988 | ||
3989 | if (i >= mpb->num_disks) { | |
3990 | struct intel_disk *idisk; | |
3991 | ||
3992 | idisk = disk_list_get(dl->serial, disk_list); | |
ecf408e9 | 3993 | if (idisk && is_spare(&idisk->disk) && |
a2b97981 DW |
3994 | !is_failed(&idisk->disk) && !is_configured(&idisk->disk)) |
3995 | dl->index = -1; | |
3996 | else { | |
3997 | dl->index = -2; | |
3998 | continue; | |
3999 | } | |
4000 | } | |
4001 | ||
4002 | dl->next = champion->disks; | |
4003 | champion->disks = dl; | |
4004 | s->disks = NULL; | |
4005 | } | |
4006 | ||
4007 | /* delete 'champion' from super_list */ | |
4008 | for (del = super_list; *del; ) { | |
4009 | if (*del == champion) { | |
4010 | *del = (*del)->next; | |
4011 | break; | |
4012 | } else | |
4013 | del = &(*del)->next; | |
4014 | } | |
4015 | champion->next = NULL; | |
4016 | ||
4017 | out: | |
4018 | while (disk_list) { | |
4019 | struct intel_disk *idisk = disk_list; | |
4020 | ||
4021 | disk_list = disk_list->next; | |
4022 | free(idisk); | |
4023 | } | |
4024 | ||
4025 | return champion; | |
4026 | } | |
4027 | ||
9587c373 LM |
4028 | |
4029 | static int | |
4030 | get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd); | |
9587c373 LM |
4031 | static int get_super_block(struct intel_super **super_list, int devnum, char *devname, |
4032 | int major, int minor, int keep_fd); | |
ec50f7b6 LM |
4033 | static int |
4034 | get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list, | |
4035 | int *max, int keep_fd); | |
4036 | ||
9587c373 | 4037 | |
cdddbdbc | 4038 | static int load_super_imsm_all(struct supertype *st, int fd, void **sbp, |
ec50f7b6 LM |
4039 | char *devname, struct md_list *devlist, |
4040 | int keep_fd) | |
cdddbdbc | 4041 | { |
a2b97981 DW |
4042 | struct intel_super *super_list = NULL; |
4043 | struct intel_super *super = NULL; | |
a2b97981 | 4044 | int err = 0; |
9587c373 | 4045 | int i = 0; |
dab4a513 | 4046 | |
9587c373 LM |
4047 | if (fd >= 0) |
4048 | /* 'fd' is an opened container */ | |
4049 | err = get_sra_super_block(fd, &super_list, devname, &i, keep_fd); | |
4050 | else | |
ec50f7b6 LM |
4051 | /* get super block from devlist devices */ |
4052 | err = get_devlist_super_block(devlist, &super_list, &i, keep_fd); | |
9587c373 | 4053 | if (err) |
1602d52c | 4054 | goto error; |
a2b97981 DW |
4055 | /* all mpbs enter, maybe one leaves */ |
4056 | super = imsm_thunderdome(&super_list, i); | |
4057 | if (!super) { | |
4058 | err = 1; | |
4059 | goto error; | |
cdddbdbc DW |
4060 | } |
4061 | ||
47ee5a45 DW |
4062 | if (find_missing(super) != 0) { |
4063 | free_imsm(super); | |
a2b97981 DW |
4064 | err = 2; |
4065 | goto error; | |
47ee5a45 | 4066 | } |
8e59f3d8 AK |
4067 | |
4068 | /* load migration record */ | |
4069 | err = load_imsm_migr_rec(super, NULL); | |
4c965cc9 AK |
4070 | if (err == -1) { |
4071 | /* migration is in progress, | |
4072 | * but migr_rec cannot be loaded, | |
4073 | */ | |
8e59f3d8 AK |
4074 | err = 4; |
4075 | goto error; | |
4076 | } | |
e2f41b2c AK |
4077 | |
4078 | /* Check migration compatibility */ | |
4c965cc9 | 4079 | if ((err == 0) && (check_mpb_migr_compatibility(super) != 0)) { |
e2f41b2c AK |
4080 | fprintf(stderr, Name ": Unsupported migration detected"); |
4081 | if (devname) | |
4082 | fprintf(stderr, " on %s\n", devname); | |
4083 | else | |
4084 | fprintf(stderr, " (IMSM).\n"); | |
4085 | ||
4086 | err = 5; | |
4087 | goto error; | |
4088 | } | |
4089 | ||
a2b97981 DW |
4090 | err = 0; |
4091 | ||
4092 | error: | |
4093 | while (super_list) { | |
4094 | struct intel_super *s = super_list; | |
4095 | ||
4096 | super_list = super_list->next; | |
4097 | free_imsm(s); | |
4098 | } | |
9587c373 | 4099 | |
a2b97981 DW |
4100 | |
4101 | if (err) | |
4102 | return err; | |
f7e7067b | 4103 | |
cdddbdbc | 4104 | *sbp = super; |
9587c373 LM |
4105 | if (fd >= 0) |
4106 | st->container_dev = fd2devnum(fd); | |
4107 | else | |
4108 | st->container_dev = NoMdDev; | |
a2b97981 | 4109 | if (err == 0 && st->ss == NULL) { |
bf5a934a | 4110 | st->ss = &super_imsm; |
cdddbdbc DW |
4111 | st->minor_version = 0; |
4112 | st->max_devs = IMSM_MAX_DEVICES; | |
4113 | } | |
cdddbdbc DW |
4114 | return 0; |
4115 | } | |
2b959fbf | 4116 | |
9587c373 | 4117 | |
ec50f7b6 LM |
4118 | static int |
4119 | get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list, | |
4120 | int *max, int keep_fd) | |
4121 | { | |
4122 | struct md_list *tmpdev; | |
4123 | int err = 0; | |
4124 | int i = 0; | |
9587c373 | 4125 | |
ec50f7b6 LM |
4126 | for (i = 0, tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { |
4127 | if (tmpdev->used != 1) | |
4128 | continue; | |
4129 | if (tmpdev->container == 1) { | |
ca9de185 | 4130 | int lmax = 0; |
ec50f7b6 LM |
4131 | int fd = dev_open(tmpdev->devname, O_RDONLY|O_EXCL); |
4132 | if (fd < 0) { | |
4133 | fprintf(stderr, Name ": cannot open device %s: %s\n", | |
4134 | tmpdev->devname, strerror(errno)); | |
4135 | err = 8; | |
4136 | goto error; | |
4137 | } | |
4138 | err = get_sra_super_block(fd, super_list, | |
4139 | tmpdev->devname, &lmax, | |
4140 | keep_fd); | |
4141 | i += lmax; | |
4142 | close(fd); | |
4143 | if (err) { | |
4144 | err = 7; | |
4145 | goto error; | |
4146 | } | |
4147 | } else { | |
4148 | int major = major(tmpdev->st_rdev); | |
4149 | int minor = minor(tmpdev->st_rdev); | |
4150 | err = get_super_block(super_list, | |
4151 | -1, | |
4152 | tmpdev->devname, | |
4153 | major, minor, | |
4154 | keep_fd); | |
4155 | i++; | |
4156 | if (err) { | |
4157 | err = 6; | |
4158 | goto error; | |
4159 | } | |
4160 | } | |
4161 | } | |
4162 | error: | |
4163 | *max = i; | |
4164 | return err; | |
4165 | } | |
9587c373 LM |
4166 | |
4167 | static int get_super_block(struct intel_super **super_list, int devnum, char *devname, | |
4168 | int major, int minor, int keep_fd) | |
4169 | { | |
4170 | struct intel_super*s = NULL; | |
4171 | char nm[32]; | |
4172 | int dfd = -1; | |
4173 | int rv; | |
4174 | int err = 0; | |
4175 | int retry; | |
4176 | ||
4177 | s = alloc_super(); | |
4178 | if (!s) { | |
4179 | err = 1; | |
4180 | goto error; | |
4181 | } | |
4182 | ||
4183 | sprintf(nm, "%d:%d", major, minor); | |
4184 | dfd = dev_open(nm, O_RDWR); | |
4185 | if (dfd < 0) { | |
4186 | err = 2; | |
4187 | goto error; | |
4188 | } | |
4189 | ||
4190 | rv = find_intel_hba_capability(dfd, s, devname); | |
4191 | /* no orom/efi or non-intel hba of the disk */ | |
4192 | if (rv != 0) { | |
4193 | err = 4; | |
4194 | goto error; | |
4195 | } | |
4196 | ||
4197 | err = load_and_parse_mpb(dfd, s, NULL, keep_fd); | |
4198 | ||
4199 | /* retry the load if we might have raced against mdmon */ | |
4200 | if (err == 3 && (devnum != -1) && mdmon_running(devnum)) | |
4201 | for (retry = 0; retry < 3; retry++) { | |
4202 | usleep(3000); | |
4203 | err = load_and_parse_mpb(dfd, s, NULL, keep_fd); | |
4204 | if (err != 3) | |
4205 | break; | |
4206 | } | |
4207 | error: | |
4208 | if (!err) { | |
4209 | s->next = *super_list; | |
4210 | *super_list = s; | |
4211 | } else { | |
4212 | if (s) | |
4213 | free(s); | |
4214 | if (dfd) | |
4215 | close(dfd); | |
4216 | } | |
4217 | if ((dfd >= 0) && (!keep_fd)) | |
4218 | close(dfd); | |
4219 | return err; | |
4220 | ||
4221 | } | |
4222 | ||
4223 | static int | |
4224 | get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd) | |
4225 | { | |
4226 | struct mdinfo *sra; | |
4227 | int devnum; | |
4228 | struct mdinfo *sd; | |
4229 | int err = 0; | |
4230 | int i = 0; | |
4231 | sra = sysfs_read(fd, 0, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE); | |
4232 | if (!sra) | |
4233 | return 1; | |
4234 | ||
4235 | if (sra->array.major_version != -1 || | |
4236 | sra->array.minor_version != -2 || | |
4237 | strcmp(sra->text_version, "imsm") != 0) { | |
4238 | err = 1; | |
4239 | goto error; | |
4240 | } | |
4241 | /* load all mpbs */ | |
4242 | devnum = fd2devnum(fd); | |
4243 | for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) { | |
4244 | if (get_super_block(super_list, devnum, devname, | |
4245 | sd->disk.major, sd->disk.minor, keep_fd) != 0) { | |
4246 | err = 7; | |
4247 | goto error; | |
4248 | } | |
4249 | } | |
4250 | error: | |
4251 | sysfs_free(sra); | |
4252 | *max = i; | |
4253 | return err; | |
4254 | } | |
4255 | ||
2b959fbf N |
4256 | static int load_container_imsm(struct supertype *st, int fd, char *devname) |
4257 | { | |
ec50f7b6 | 4258 | return load_super_imsm_all(st, fd, &st->sb, devname, NULL, 1); |
2b959fbf | 4259 | } |
cdddbdbc DW |
4260 | #endif |
4261 | ||
4262 | static int load_super_imsm(struct supertype *st, int fd, char *devname) | |
4263 | { | |
4264 | struct intel_super *super; | |
4265 | int rv; | |
4266 | ||
691c6ee1 N |
4267 | if (test_partition(fd)) |
4268 | /* IMSM not allowed on partitions */ | |
4269 | return 1; | |
4270 | ||
37424f13 DW |
4271 | free_super_imsm(st); |
4272 | ||
49133e57 | 4273 | super = alloc_super(); |
cdddbdbc DW |
4274 | if (!super) { |
4275 | fprintf(stderr, | |
4276 | Name ": malloc of %zu failed.\n", | |
4277 | sizeof(*super)); | |
4278 | return 1; | |
4279 | } | |
ea2bc72b LM |
4280 | /* Load hba and capabilities if they exist. |
4281 | * But do not preclude loading metadata in case capabilities or hba are | |
4282 | * non-compliant and ignore_hw_compat is set. | |
4283 | */ | |
d424212e | 4284 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 4285 | /* no orom/efi or non-intel hba of the disk */ |
ea2bc72b | 4286 | if ((rv != 0) && (st->ignore_hw_compat == 0)) { |
f2f5c343 LM |
4287 | if (devname) |
4288 | fprintf(stderr, | |
4289 | Name ": No OROM/EFI properties for %s\n", devname); | |
4290 | free_imsm(super); | |
4291 | return 2; | |
4292 | } | |
a2b97981 | 4293 | rv = load_and_parse_mpb(fd, super, devname, 0); |
cdddbdbc DW |
4294 | |
4295 | if (rv) { | |
4296 | if (devname) | |
4297 | fprintf(stderr, | |
4298 | Name ": Failed to load all information " | |
4299 | "sections on %s\n", devname); | |
4300 | free_imsm(super); | |
4301 | return rv; | |
4302 | } | |
4303 | ||
4304 | st->sb = super; | |
4305 | if (st->ss == NULL) { | |
4306 | st->ss = &super_imsm; | |
4307 | st->minor_version = 0; | |
4308 | st->max_devs = IMSM_MAX_DEVICES; | |
4309 | } | |
8e59f3d8 AK |
4310 | |
4311 | /* load migration record */ | |
2e062e82 AK |
4312 | if (load_imsm_migr_rec(super, NULL) == 0) { |
4313 | /* Check for unsupported migration features */ | |
4314 | if (check_mpb_migr_compatibility(super) != 0) { | |
4315 | fprintf(stderr, | |
4316 | Name ": Unsupported migration detected"); | |
4317 | if (devname) | |
4318 | fprintf(stderr, " on %s\n", devname); | |
4319 | else | |
4320 | fprintf(stderr, " (IMSM).\n"); | |
4321 | return 3; | |
4322 | } | |
e2f41b2c AK |
4323 | } |
4324 | ||
cdddbdbc DW |
4325 | return 0; |
4326 | } | |
4327 | ||
ef6ffade DW |
4328 | static __u16 info_to_blocks_per_strip(mdu_array_info_t *info) |
4329 | { | |
4330 | if (info->level == 1) | |
4331 | return 128; | |
4332 | return info->chunk_size >> 9; | |
4333 | } | |
4334 | ||
ff596308 | 4335 | static __u32 info_to_num_data_stripes(mdu_array_info_t *info, int num_domains) |
ef6ffade DW |
4336 | { |
4337 | __u32 num_stripes; | |
4338 | ||
4339 | num_stripes = (info->size * 2) / info_to_blocks_per_strip(info); | |
ff596308 | 4340 | num_stripes /= num_domains; |
ef6ffade DW |
4341 | |
4342 | return num_stripes; | |
4343 | } | |
4344 | ||
fcfd9599 DW |
4345 | static __u32 info_to_blocks_per_member(mdu_array_info_t *info) |
4346 | { | |
4025c288 DW |
4347 | if (info->level == 1) |
4348 | return info->size * 2; | |
4349 | else | |
4350 | return (info->size * 2) & ~(info_to_blocks_per_strip(info) - 1); | |
fcfd9599 DW |
4351 | } |
4352 | ||
4d1313e9 DW |
4353 | static void imsm_update_version_info(struct intel_super *super) |
4354 | { | |
4355 | /* update the version and attributes */ | |
4356 | struct imsm_super *mpb = super->anchor; | |
4357 | char *version; | |
4358 | struct imsm_dev *dev; | |
4359 | struct imsm_map *map; | |
4360 | int i; | |
4361 | ||
4362 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
4363 | dev = get_imsm_dev(super, i); | |
238c0a71 | 4364 | map = get_imsm_map(dev, MAP_0); |
4d1313e9 DW |
4365 | if (__le32_to_cpu(dev->size_high) > 0) |
4366 | mpb->attributes |= MPB_ATTRIB_2TB; | |
4367 | ||
4368 | /* FIXME detect when an array spans a port multiplier */ | |
4369 | #if 0 | |
4370 | mpb->attributes |= MPB_ATTRIB_PM; | |
4371 | #endif | |
4372 | ||
4373 | if (mpb->num_raid_devs > 1 || | |
4374 | mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) { | |
4375 | version = MPB_VERSION_ATTRIBS; | |
4376 | switch (get_imsm_raid_level(map)) { | |
4377 | case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break; | |
4378 | case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break; | |
4379 | case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break; | |
4380 | case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break; | |
4381 | } | |
4382 | } else { | |
4383 | if (map->num_members >= 5) | |
4384 | version = MPB_VERSION_5OR6_DISK_ARRAY; | |
4385 | else if (dev->status == DEV_CLONE_N_GO) | |
4386 | version = MPB_VERSION_CNG; | |
4387 | else if (get_imsm_raid_level(map) == 5) | |
4388 | version = MPB_VERSION_RAID5; | |
4389 | else if (map->num_members >= 3) | |
4390 | version = MPB_VERSION_3OR4_DISK_ARRAY; | |
4391 | else if (get_imsm_raid_level(map) == 1) | |
4392 | version = MPB_VERSION_RAID1; | |
4393 | else | |
4394 | version = MPB_VERSION_RAID0; | |
4395 | } | |
4396 | strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version); | |
4397 | } | |
4398 | } | |
4399 | ||
aa534678 DW |
4400 | static int check_name(struct intel_super *super, char *name, int quiet) |
4401 | { | |
4402 | struct imsm_super *mpb = super->anchor; | |
4403 | char *reason = NULL; | |
4404 | int i; | |
4405 | ||
4406 | if (strlen(name) > MAX_RAID_SERIAL_LEN) | |
4407 | reason = "must be 16 characters or less"; | |
4408 | ||
4409 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
4410 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
4411 | ||
4412 | if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) { | |
4413 | reason = "already exists"; | |
4414 | break; | |
4415 | } | |
4416 | } | |
4417 | ||
4418 | if (reason && !quiet) | |
4419 | fprintf(stderr, Name ": imsm volume name %s\n", reason); | |
4420 | ||
4421 | return !reason; | |
4422 | } | |
4423 | ||
8b353278 DW |
4424 | static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info, |
4425 | unsigned long long size, char *name, | |
4426 | char *homehost, int *uuid) | |
cdddbdbc | 4427 | { |
c2c087e6 DW |
4428 | /* We are creating a volume inside a pre-existing container. |
4429 | * so st->sb is already set. | |
4430 | */ | |
4431 | struct intel_super *super = st->sb; | |
949c47a0 | 4432 | struct imsm_super *mpb = super->anchor; |
ba2de7ba | 4433 | struct intel_dev *dv; |
c2c087e6 DW |
4434 | struct imsm_dev *dev; |
4435 | struct imsm_vol *vol; | |
4436 | struct imsm_map *map; | |
4437 | int idx = mpb->num_raid_devs; | |
4438 | int i; | |
4439 | unsigned long long array_blocks; | |
2c092cad | 4440 | size_t size_old, size_new; |
ff596308 | 4441 | __u32 num_data_stripes; |
cdddbdbc | 4442 | |
88c32bb1 | 4443 | if (super->orom && mpb->num_raid_devs >= super->orom->vpa) { |
c2c087e6 | 4444 | fprintf(stderr, Name": This imsm-container already has the " |
88c32bb1 | 4445 | "maximum of %d volumes\n", super->orom->vpa); |
c2c087e6 DW |
4446 | return 0; |
4447 | } | |
4448 | ||
2c092cad DW |
4449 | /* ensure the mpb is large enough for the new data */ |
4450 | size_old = __le32_to_cpu(mpb->mpb_size); | |
4451 | size_new = disks_to_mpb_size(info->nr_disks); | |
4452 | if (size_new > size_old) { | |
4453 | void *mpb_new; | |
4454 | size_t size_round = ROUND_UP(size_new, 512); | |
4455 | ||
4456 | if (posix_memalign(&mpb_new, 512, size_round) != 0) { | |
4457 | fprintf(stderr, Name": could not allocate new mpb\n"); | |
4458 | return 0; | |
4459 | } | |
17a4eaf9 AK |
4460 | if (posix_memalign(&super->migr_rec_buf, 512, |
4461 | MIGR_REC_BUF_SIZE) != 0) { | |
8e59f3d8 AK |
4462 | fprintf(stderr, Name |
4463 | ": %s could not allocate migr_rec buffer\n", | |
4464 | __func__); | |
4465 | free(super->buf); | |
4466 | free(super); | |
ea944c8f | 4467 | free(mpb_new); |
8e59f3d8 AK |
4468 | return 0; |
4469 | } | |
2c092cad DW |
4470 | memcpy(mpb_new, mpb, size_old); |
4471 | free(mpb); | |
4472 | mpb = mpb_new; | |
949c47a0 | 4473 | super->anchor = mpb_new; |
2c092cad DW |
4474 | mpb->mpb_size = __cpu_to_le32(size_new); |
4475 | memset(mpb_new + size_old, 0, size_round - size_old); | |
4476 | } | |
bf5a934a | 4477 | super->current_vol = idx; |
3960e579 DW |
4478 | |
4479 | /* handle 'failed_disks' by either: | |
4480 | * a) create dummy disk entries in the table if this the first | |
4481 | * volume in the array. We add them here as this is the only | |
4482 | * opportunity to add them. add_to_super_imsm_volume() | |
4483 | * handles the non-failed disks and continues incrementing | |
4484 | * mpb->num_disks. | |
4485 | * b) validate that 'failed_disks' matches the current number | |
4486 | * of missing disks if the container is populated | |
d23fe947 | 4487 | */ |
3960e579 | 4488 | if (super->current_vol == 0) { |
d23fe947 | 4489 | mpb->num_disks = 0; |
3960e579 DW |
4490 | for (i = 0; i < info->failed_disks; i++) { |
4491 | struct imsm_disk *disk; | |
4492 | ||
4493 | mpb->num_disks++; | |
4494 | disk = __get_imsm_disk(mpb, i); | |
4495 | disk->status = CONFIGURED_DISK | FAILED_DISK; | |
4496 | disk->scsi_id = __cpu_to_le32(~(__u32)0); | |
4497 | snprintf((char *) disk->serial, MAX_RAID_SERIAL_LEN, | |
4498 | "missing:%d", i); | |
4499 | } | |
4500 | find_missing(super); | |
4501 | } else { | |
4502 | int missing = 0; | |
4503 | struct dl *d; | |
4504 | ||
4505 | for (d = super->missing; d; d = d->next) | |
4506 | missing++; | |
4507 | if (info->failed_disks > missing) { | |
4508 | fprintf(stderr, Name": unable to add 'missing' disk to container\n"); | |
4509 | return 0; | |
4510 | } | |
4511 | } | |
5a038140 | 4512 | |
aa534678 DW |
4513 | if (!check_name(super, name, 0)) |
4514 | return 0; | |
ba2de7ba DW |
4515 | dv = malloc(sizeof(*dv)); |
4516 | if (!dv) { | |
4517 | fprintf(stderr, Name ": failed to allocate device list entry\n"); | |
4518 | return 0; | |
4519 | } | |
1a2487c2 | 4520 | dev = calloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1)); |
949c47a0 | 4521 | if (!dev) { |
ba2de7ba | 4522 | free(dv); |
949c47a0 DW |
4523 | fprintf(stderr, Name": could not allocate raid device\n"); |
4524 | return 0; | |
4525 | } | |
1a2487c2 | 4526 | |
c2c087e6 | 4527 | strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN); |
03bcbc65 DW |
4528 | if (info->level == 1) |
4529 | array_blocks = info_to_blocks_per_member(info); | |
4530 | else | |
4531 | array_blocks = calc_array_size(info->level, info->raid_disks, | |
4532 | info->layout, info->chunk_size, | |
4533 | info->size*2); | |
979d38be DW |
4534 | /* round array size down to closest MB */ |
4535 | array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT; | |
4536 | ||
c2c087e6 DW |
4537 | dev->size_low = __cpu_to_le32((__u32) array_blocks); |
4538 | dev->size_high = __cpu_to_le32((__u32) (array_blocks >> 32)); | |
1a2487c2 | 4539 | dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING); |
c2c087e6 DW |
4540 | vol = &dev->vol; |
4541 | vol->migr_state = 0; | |
1484e727 | 4542 | set_migr_type(dev, MIGR_INIT); |
3960e579 | 4543 | vol->dirty = !info->state; |
f8f603f1 | 4544 | vol->curr_migr_unit = 0; |
238c0a71 | 4545 | map = get_imsm_map(dev, MAP_0); |
0dcecb2e | 4546 | map->pba_of_lba0 = __cpu_to_le32(super->create_offset); |
fcfd9599 | 4547 | map->blocks_per_member = __cpu_to_le32(info_to_blocks_per_member(info)); |
ef6ffade | 4548 | map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info)); |
0556e1a2 | 4549 | map->failed_disk_num = ~0; |
bf4442ab AK |
4550 | if (info->level > 0) |
4551 | map->map_state = IMSM_T_STATE_UNINITIALIZED; | |
4552 | else | |
4553 | map->map_state = info->failed_disks ? IMSM_T_STATE_FAILED : | |
4554 | IMSM_T_STATE_NORMAL; | |
252d23c0 | 4555 | map->ddf = 1; |
ef6ffade DW |
4556 | |
4557 | if (info->level == 1 && info->raid_disks > 2) { | |
38950822 AW |
4558 | free(dev); |
4559 | free(dv); | |
ef6ffade DW |
4560 | fprintf(stderr, Name": imsm does not support more than 2 disks" |
4561 | "in a raid1 volume\n"); | |
4562 | return 0; | |
4563 | } | |
81062a36 DW |
4564 | |
4565 | map->raid_level = info->level; | |
4d1313e9 | 4566 | if (info->level == 10) { |
c2c087e6 | 4567 | map->raid_level = 1; |
4d1313e9 | 4568 | map->num_domains = info->raid_disks / 2; |
81062a36 DW |
4569 | } else if (info->level == 1) |
4570 | map->num_domains = info->raid_disks; | |
4571 | else | |
ff596308 | 4572 | map->num_domains = 1; |
81062a36 | 4573 | |
ff596308 DW |
4574 | num_data_stripes = info_to_num_data_stripes(info, map->num_domains); |
4575 | map->num_data_stripes = __cpu_to_le32(num_data_stripes); | |
ef6ffade | 4576 | |
c2c087e6 DW |
4577 | map->num_members = info->raid_disks; |
4578 | for (i = 0; i < map->num_members; i++) { | |
4579 | /* initialized in add_to_super */ | |
4eb26970 | 4580 | set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD); |
c2c087e6 | 4581 | } |
949c47a0 | 4582 | mpb->num_raid_devs++; |
ba2de7ba DW |
4583 | |
4584 | dv->dev = dev; | |
4585 | dv->index = super->current_vol; | |
4586 | dv->next = super->devlist; | |
4587 | super->devlist = dv; | |
c2c087e6 | 4588 | |
4d1313e9 DW |
4589 | imsm_update_version_info(super); |
4590 | ||
c2c087e6 | 4591 | return 1; |
cdddbdbc DW |
4592 | } |
4593 | ||
bf5a934a DW |
4594 | static int init_super_imsm(struct supertype *st, mdu_array_info_t *info, |
4595 | unsigned long long size, char *name, | |
4596 | char *homehost, int *uuid) | |
4597 | { | |
4598 | /* This is primarily called by Create when creating a new array. | |
4599 | * We will then get add_to_super called for each component, and then | |
4600 | * write_init_super called to write it out to each device. | |
4601 | * For IMSM, Create can create on fresh devices or on a pre-existing | |
4602 | * array. | |
4603 | * To create on a pre-existing array a different method will be called. | |
4604 | * This one is just for fresh drives. | |
4605 | */ | |
4606 | struct intel_super *super; | |
4607 | struct imsm_super *mpb; | |
4608 | size_t mpb_size; | |
4d1313e9 | 4609 | char *version; |
bf5a934a | 4610 | |
bf5a934a | 4611 | if (st->sb) |
e683ca88 DW |
4612 | return init_super_imsm_volume(st, info, size, name, homehost, uuid); |
4613 | ||
4614 | if (info) | |
4615 | mpb_size = disks_to_mpb_size(info->nr_disks); | |
4616 | else | |
4617 | mpb_size = 512; | |
bf5a934a | 4618 | |
49133e57 | 4619 | super = alloc_super(); |
e683ca88 | 4620 | if (super && posix_memalign(&super->buf, 512, mpb_size) != 0) { |
bf5a934a | 4621 | free(super); |
e683ca88 DW |
4622 | super = NULL; |
4623 | } | |
4624 | if (!super) { | |
4625 | fprintf(stderr, Name | |
4626 | ": %s could not allocate superblock\n", __func__); | |
bf5a934a DW |
4627 | return 0; |
4628 | } | |
17a4eaf9 | 4629 | if (posix_memalign(&super->migr_rec_buf, 512, MIGR_REC_BUF_SIZE) != 0) { |
8e59f3d8 AK |
4630 | fprintf(stderr, Name |
4631 | ": %s could not allocate migr_rec buffer\n", __func__); | |
4632 | free(super->buf); | |
4633 | free(super); | |
4634 | return 0; | |
4635 | } | |
e683ca88 | 4636 | memset(super->buf, 0, mpb_size); |
ef649044 | 4637 | mpb = super->buf; |
e683ca88 DW |
4638 | mpb->mpb_size = __cpu_to_le32(mpb_size); |
4639 | st->sb = super; | |
4640 | ||
4641 | if (info == NULL) { | |
4642 | /* zeroing superblock */ | |
4643 | return 0; | |
4644 | } | |
bf5a934a | 4645 | |
4d1313e9 DW |
4646 | mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; |
4647 | ||
4648 | version = (char *) mpb->sig; | |
4649 | strcpy(version, MPB_SIGNATURE); | |
4650 | version += strlen(MPB_SIGNATURE); | |
4651 | strcpy(version, MPB_VERSION_RAID0); | |
bf5a934a | 4652 | |
bf5a934a DW |
4653 | return 1; |
4654 | } | |
4655 | ||
0e600426 | 4656 | #ifndef MDASSEMBLE |
f20c3968 | 4657 | static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk, |
bf5a934a DW |
4658 | int fd, char *devname) |
4659 | { | |
4660 | struct intel_super *super = st->sb; | |
d23fe947 | 4661 | struct imsm_super *mpb = super->anchor; |
3960e579 | 4662 | struct imsm_disk *_disk; |
bf5a934a DW |
4663 | struct imsm_dev *dev; |
4664 | struct imsm_map *map; | |
3960e579 | 4665 | struct dl *dl, *df; |
4eb26970 | 4666 | int slot; |
bf5a934a | 4667 | |
949c47a0 | 4668 | dev = get_imsm_dev(super, super->current_vol); |
238c0a71 | 4669 | map = get_imsm_map(dev, MAP_0); |
bf5a934a | 4670 | |
208933a7 N |
4671 | if (! (dk->state & (1<<MD_DISK_SYNC))) { |
4672 | fprintf(stderr, Name ": %s: Cannot add spare devices to IMSM volume\n", | |
4673 | devname); | |
4674 | return 1; | |
4675 | } | |
4676 | ||
efb30e7f DW |
4677 | if (fd == -1) { |
4678 | /* we're doing autolayout so grab the pre-marked (in | |
4679 | * validate_geometry) raid_disk | |
4680 | */ | |
4681 | for (dl = super->disks; dl; dl = dl->next) | |
4682 | if (dl->raiddisk == dk->raid_disk) | |
4683 | break; | |
4684 | } else { | |
4685 | for (dl = super->disks; dl ; dl = dl->next) | |
4686 | if (dl->major == dk->major && | |
4687 | dl->minor == dk->minor) | |
4688 | break; | |
4689 | } | |
d23fe947 | 4690 | |
208933a7 N |
4691 | if (!dl) { |
4692 | fprintf(stderr, Name ": %s is not a member of the same container\n", devname); | |
f20c3968 | 4693 | return 1; |
208933a7 | 4694 | } |
bf5a934a | 4695 | |
d23fe947 DW |
4696 | /* add a pristine spare to the metadata */ |
4697 | if (dl->index < 0) { | |
4698 | dl->index = super->anchor->num_disks; | |
4699 | super->anchor->num_disks++; | |
4700 | } | |
4eb26970 DW |
4701 | /* Check the device has not already been added */ |
4702 | slot = get_imsm_disk_slot(map, dl->index); | |
4703 | if (slot >= 0 && | |
238c0a71 | 4704 | (get_imsm_ord_tbl_ent(dev, slot, MAP_X) & IMSM_ORD_REBUILD) == 0) { |
4eb26970 DW |
4705 | fprintf(stderr, Name ": %s has been included in this array twice\n", |
4706 | devname); | |
4707 | return 1; | |
4708 | } | |
656b6b5a | 4709 | set_imsm_ord_tbl_ent(map, dk->raid_disk, dl->index); |
ee5aad5a | 4710 | dl->disk.status = CONFIGURED_DISK; |
d23fe947 | 4711 | |
3960e579 DW |
4712 | /* update size of 'missing' disks to be at least as large as the |
4713 | * largest acitve member (we only have dummy missing disks when | |
4714 | * creating the first volume) | |
4715 | */ | |
4716 | if (super->current_vol == 0) { | |
4717 | for (df = super->missing; df; df = df->next) { | |
4718 | if (dl->disk.total_blocks > df->disk.total_blocks) | |
4719 | df->disk.total_blocks = dl->disk.total_blocks; | |
4720 | _disk = __get_imsm_disk(mpb, df->index); | |
4721 | *_disk = df->disk; | |
4722 | } | |
4723 | } | |
4724 | ||
4725 | /* refresh unset/failed slots to point to valid 'missing' entries */ | |
4726 | for (df = super->missing; df; df = df->next) | |
4727 | for (slot = 0; slot < mpb->num_disks; slot++) { | |
238c0a71 | 4728 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); |
3960e579 DW |
4729 | |
4730 | if ((ord & IMSM_ORD_REBUILD) == 0) | |
4731 | continue; | |
4732 | set_imsm_ord_tbl_ent(map, slot, df->index | IMSM_ORD_REBUILD); | |
1ace8403 | 4733 | if (is_gen_migration(dev)) { |
238c0a71 AK |
4734 | struct imsm_map *map2 = get_imsm_map(dev, |
4735 | MAP_1); | |
0a108d63 AK |
4736 | int slot2 = get_imsm_disk_slot(map2, df->index); |
4737 | if ((slot2 < map2->num_members) && | |
4738 | (slot2 >= 0)) { | |
1ace8403 | 4739 | __u32 ord2 = get_imsm_ord_tbl_ent(dev, |
238c0a71 AK |
4740 | slot2, |
4741 | MAP_1); | |
1ace8403 AK |
4742 | if ((unsigned)df->index == |
4743 | ord_to_idx(ord2)) | |
4744 | set_imsm_ord_tbl_ent(map2, | |
0a108d63 | 4745 | slot2, |
1ace8403 AK |
4746 | df->index | |
4747 | IMSM_ORD_REBUILD); | |
4748 | } | |
4749 | } | |
3960e579 DW |
4750 | dprintf("set slot:%d to missing disk:%d\n", slot, df->index); |
4751 | break; | |
4752 | } | |
4753 | ||
d23fe947 DW |
4754 | /* if we are creating the first raid device update the family number */ |
4755 | if (super->current_vol == 0) { | |
4756 | __u32 sum; | |
4757 | struct imsm_dev *_dev = __get_imsm_dev(mpb, 0); | |
d23fe947 | 4758 | |
3960e579 | 4759 | _disk = __get_imsm_disk(mpb, dl->index); |
791b666a AW |
4760 | if (!_dev || !_disk) { |
4761 | fprintf(stderr, Name ": BUG mpb setup error\n"); | |
4762 | return 1; | |
4763 | } | |
d23fe947 DW |
4764 | *_dev = *dev; |
4765 | *_disk = dl->disk; | |
148acb7b DW |
4766 | sum = random32(); |
4767 | sum += __gen_imsm_checksum(mpb); | |
d23fe947 | 4768 | mpb->family_num = __cpu_to_le32(sum); |
148acb7b | 4769 | mpb->orig_family_num = mpb->family_num; |
d23fe947 | 4770 | } |
ca0748fa | 4771 | super->current_disk = dl; |
f20c3968 | 4772 | return 0; |
bf5a934a DW |
4773 | } |
4774 | ||
a8619d23 AK |
4775 | /* mark_spare() |
4776 | * Function marks disk as spare and restores disk serial | |
4777 | * in case it was previously marked as failed by takeover operation | |
4778 | * reruns: | |
4779 | * -1 : critical error | |
4780 | * 0 : disk is marked as spare but serial is not set | |
4781 | * 1 : success | |
4782 | */ | |
4783 | int mark_spare(struct dl *disk) | |
4784 | { | |
4785 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
4786 | int ret_val = -1; | |
4787 | ||
4788 | if (!disk) | |
4789 | return ret_val; | |
4790 | ||
4791 | ret_val = 0; | |
4792 | if (!imsm_read_serial(disk->fd, NULL, serial)) { | |
4793 | /* Restore disk serial number, because takeover marks disk | |
4794 | * as failed and adds to serial ':0' before it becomes | |
4795 | * a spare disk. | |
4796 | */ | |
4797 | serialcpy(disk->serial, serial); | |
4798 | serialcpy(disk->disk.serial, serial); | |
4799 | ret_val = 1; | |
4800 | } | |
4801 | disk->disk.status = SPARE_DISK; | |
4802 | disk->index = -1; | |
4803 | ||
4804 | return ret_val; | |
4805 | } | |
88654014 | 4806 | |
f20c3968 | 4807 | static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk, |
88654014 | 4808 | int fd, char *devname) |
cdddbdbc | 4809 | { |
c2c087e6 | 4810 | struct intel_super *super = st->sb; |
c2c087e6 DW |
4811 | struct dl *dd; |
4812 | unsigned long long size; | |
f2f27e63 | 4813 | __u32 id; |
c2c087e6 DW |
4814 | int rv; |
4815 | struct stat stb; | |
4816 | ||
88654014 LM |
4817 | /* If we are on an RAID enabled platform check that the disk is |
4818 | * attached to the raid controller. | |
4819 | * We do not need to test disks attachment for container based additions, | |
4820 | * they shall be already tested when container was created/assembled. | |
88c32bb1 | 4821 | */ |
d424212e | 4822 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 4823 | /* no orom/efi or non-intel hba of the disk */ |
f0f5a016 LM |
4824 | if (rv != 0) { |
4825 | dprintf("capability: %p fd: %d ret: %d\n", | |
4826 | super->orom, fd, rv); | |
4827 | return 1; | |
88c32bb1 DW |
4828 | } |
4829 | ||
f20c3968 DW |
4830 | if (super->current_vol >= 0) |
4831 | return add_to_super_imsm_volume(st, dk, fd, devname); | |
bf5a934a | 4832 | |
c2c087e6 DW |
4833 | fstat(fd, &stb); |
4834 | dd = malloc(sizeof(*dd)); | |
b9f594fe | 4835 | if (!dd) { |
c2c087e6 DW |
4836 | fprintf(stderr, |
4837 | Name ": malloc failed %s:%d.\n", __func__, __LINE__); | |
f20c3968 | 4838 | return 1; |
c2c087e6 DW |
4839 | } |
4840 | memset(dd, 0, sizeof(*dd)); | |
4841 | dd->major = major(stb.st_rdev); | |
4842 | dd->minor = minor(stb.st_rdev); | |
c2c087e6 | 4843 | dd->devname = devname ? strdup(devname) : NULL; |
c2c087e6 | 4844 | dd->fd = fd; |
689c9bf3 | 4845 | dd->e = NULL; |
1a64be56 | 4846 | dd->action = DISK_ADD; |
c2c087e6 | 4847 | rv = imsm_read_serial(fd, devname, dd->serial); |
32ba9157 | 4848 | if (rv) { |
c2c087e6 | 4849 | fprintf(stderr, |
0030e8d6 | 4850 | Name ": failed to retrieve scsi serial, aborting\n"); |
949c47a0 | 4851 | free(dd); |
0030e8d6 | 4852 | abort(); |
c2c087e6 DW |
4853 | } |
4854 | ||
c2c087e6 DW |
4855 | get_dev_size(fd, NULL, &size); |
4856 | size /= 512; | |
1f24f035 | 4857 | serialcpy(dd->disk.serial, dd->serial); |
b9f594fe | 4858 | dd->disk.total_blocks = __cpu_to_le32(size); |
a8619d23 | 4859 | mark_spare(dd); |
c2c087e6 | 4860 | if (sysfs_disk_to_scsi_id(fd, &id) == 0) |
b9f594fe | 4861 | dd->disk.scsi_id = __cpu_to_le32(id); |
c2c087e6 | 4862 | else |
b9f594fe | 4863 | dd->disk.scsi_id = __cpu_to_le32(0); |
43dad3d6 DW |
4864 | |
4865 | if (st->update_tail) { | |
1a64be56 LM |
4866 | dd->next = super->disk_mgmt_list; |
4867 | super->disk_mgmt_list = dd; | |
43dad3d6 DW |
4868 | } else { |
4869 | dd->next = super->disks; | |
4870 | super->disks = dd; | |
ceaf0ee1 | 4871 | super->updates_pending++; |
43dad3d6 | 4872 | } |
f20c3968 DW |
4873 | |
4874 | return 0; | |
cdddbdbc DW |
4875 | } |
4876 | ||
1a64be56 LM |
4877 | |
4878 | static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk) | |
4879 | { | |
4880 | struct intel_super *super = st->sb; | |
4881 | struct dl *dd; | |
4882 | ||
4883 | /* remove from super works only in mdmon - for communication | |
4884 | * manager - monitor. Check if communication memory buffer | |
4885 | * is prepared. | |
4886 | */ | |
4887 | if (!st->update_tail) { | |
4888 | fprintf(stderr, | |
4889 | Name ": %s shall be used in mdmon context only" | |
4890 | "(line %d).\n", __func__, __LINE__); | |
4891 | return 1; | |
4892 | } | |
4893 | dd = malloc(sizeof(*dd)); | |
4894 | if (!dd) { | |
4895 | fprintf(stderr, | |
4896 | Name ": malloc failed %s:%d.\n", __func__, __LINE__); | |
4897 | return 1; | |
4898 | } | |
4899 | memset(dd, 0, sizeof(*dd)); | |
4900 | dd->major = dk->major; | |
4901 | dd->minor = dk->minor; | |
1a64be56 | 4902 | dd->fd = -1; |
a8619d23 | 4903 | mark_spare(dd); |
1a64be56 LM |
4904 | dd->action = DISK_REMOVE; |
4905 | ||
4906 | dd->next = super->disk_mgmt_list; | |
4907 | super->disk_mgmt_list = dd; | |
4908 | ||
4909 | ||
4910 | return 0; | |
4911 | } | |
4912 | ||
f796af5d DW |
4913 | static int store_imsm_mpb(int fd, struct imsm_super *mpb); |
4914 | ||
4915 | static union { | |
4916 | char buf[512]; | |
4917 | struct imsm_super anchor; | |
4918 | } spare_record __attribute__ ((aligned(512))); | |
c2c087e6 | 4919 | |
d23fe947 DW |
4920 | /* spare records have their own family number and do not have any defined raid |
4921 | * devices | |
4922 | */ | |
4923 | static int write_super_imsm_spares(struct intel_super *super, int doclose) | |
4924 | { | |
d23fe947 | 4925 | struct imsm_super *mpb = super->anchor; |
f796af5d | 4926 | struct imsm_super *spare = &spare_record.anchor; |
d23fe947 DW |
4927 | __u32 sum; |
4928 | struct dl *d; | |
4929 | ||
f796af5d DW |
4930 | spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super)), |
4931 | spare->generation_num = __cpu_to_le32(1UL), | |
4932 | spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; | |
4933 | spare->num_disks = 1, | |
4934 | spare->num_raid_devs = 0, | |
4935 | spare->cache_size = mpb->cache_size, | |
4936 | spare->pwr_cycle_count = __cpu_to_le32(1), | |
4937 | ||
4938 | snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH, | |
4939 | MPB_SIGNATURE MPB_VERSION_RAID0); | |
d23fe947 DW |
4940 | |
4941 | for (d = super->disks; d; d = d->next) { | |
8796fdc4 | 4942 | if (d->index != -1) |
d23fe947 DW |
4943 | continue; |
4944 | ||
f796af5d DW |
4945 | spare->disk[0] = d->disk; |
4946 | sum = __gen_imsm_checksum(spare); | |
4947 | spare->family_num = __cpu_to_le32(sum); | |
4948 | spare->orig_family_num = 0; | |
4949 | sum = __gen_imsm_checksum(spare); | |
4950 | spare->check_sum = __cpu_to_le32(sum); | |
d23fe947 | 4951 | |
f796af5d | 4952 | if (store_imsm_mpb(d->fd, spare)) { |
d23fe947 DW |
4953 | fprintf(stderr, "%s: failed for device %d:%d %s\n", |
4954 | __func__, d->major, d->minor, strerror(errno)); | |
e74255d9 | 4955 | return 1; |
d23fe947 DW |
4956 | } |
4957 | if (doclose) { | |
4958 | close(d->fd); | |
4959 | d->fd = -1; | |
4960 | } | |
4961 | } | |
4962 | ||
e74255d9 | 4963 | return 0; |
d23fe947 DW |
4964 | } |
4965 | ||
36988a3d | 4966 | static int write_super_imsm(struct supertype *st, int doclose) |
cdddbdbc | 4967 | { |
36988a3d | 4968 | struct intel_super *super = st->sb; |
949c47a0 | 4969 | struct imsm_super *mpb = super->anchor; |
c2c087e6 DW |
4970 | struct dl *d; |
4971 | __u32 generation; | |
4972 | __u32 sum; | |
d23fe947 | 4973 | int spares = 0; |
949c47a0 | 4974 | int i; |
a48ac0a8 | 4975 | __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk); |
36988a3d | 4976 | int num_disks = 0; |
146c6260 | 4977 | int clear_migration_record = 1; |
cdddbdbc | 4978 | |
c2c087e6 DW |
4979 | /* 'generation' is incremented everytime the metadata is written */ |
4980 | generation = __le32_to_cpu(mpb->generation_num); | |
4981 | generation++; | |
4982 | mpb->generation_num = __cpu_to_le32(generation); | |
4983 | ||
148acb7b DW |
4984 | /* fix up cases where previous mdadm releases failed to set |
4985 | * orig_family_num | |
4986 | */ | |
4987 | if (mpb->orig_family_num == 0) | |
4988 | mpb->orig_family_num = mpb->family_num; | |
4989 | ||
d23fe947 | 4990 | for (d = super->disks; d; d = d->next) { |
8796fdc4 | 4991 | if (d->index == -1) |
d23fe947 | 4992 | spares++; |
36988a3d | 4993 | else { |
d23fe947 | 4994 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
4995 | num_disks++; |
4996 | } | |
d23fe947 | 4997 | } |
36988a3d | 4998 | for (d = super->missing; d; d = d->next) { |
47ee5a45 | 4999 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
5000 | num_disks++; |
5001 | } | |
5002 | mpb->num_disks = num_disks; | |
5003 | mpb_size += sizeof(struct imsm_disk) * mpb->num_disks; | |
b9f594fe | 5004 | |
949c47a0 DW |
5005 | for (i = 0; i < mpb->num_raid_devs; i++) { |
5006 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
36988a3d AK |
5007 | struct imsm_dev *dev2 = get_imsm_dev(super, i); |
5008 | if (dev && dev2) { | |
5009 | imsm_copy_dev(dev, dev2); | |
5010 | mpb_size += sizeof_imsm_dev(dev, 0); | |
5011 | } | |
146c6260 AK |
5012 | if (is_gen_migration(dev2)) |
5013 | clear_migration_record = 0; | |
949c47a0 | 5014 | } |
a48ac0a8 DW |
5015 | mpb_size += __le32_to_cpu(mpb->bbm_log_size); |
5016 | mpb->mpb_size = __cpu_to_le32(mpb_size); | |
949c47a0 | 5017 | |
c2c087e6 | 5018 | /* recalculate checksum */ |
949c47a0 | 5019 | sum = __gen_imsm_checksum(mpb); |
c2c087e6 DW |
5020 | mpb->check_sum = __cpu_to_le32(sum); |
5021 | ||
146c6260 | 5022 | if (clear_migration_record) |
17a4eaf9 | 5023 | memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SIZE); |
146c6260 | 5024 | |
d23fe947 | 5025 | /* write the mpb for disks that compose raid devices */ |
c2c087e6 | 5026 | for (d = super->disks; d ; d = d->next) { |
86c54047 | 5027 | if (d->index < 0 || is_failed(&d->disk)) |
d23fe947 | 5028 | continue; |
f796af5d | 5029 | if (store_imsm_mpb(d->fd, mpb)) |
c2c087e6 DW |
5030 | fprintf(stderr, "%s: failed for device %d:%d %s\n", |
5031 | __func__, d->major, d->minor, strerror(errno)); | |
146c6260 AK |
5032 | if (clear_migration_record) { |
5033 | unsigned long long dsize; | |
5034 | ||
5035 | get_dev_size(d->fd, NULL, &dsize); | |
5036 | if (lseek64(d->fd, dsize - 512, SEEK_SET) >= 0) { | |
17a4eaf9 AK |
5037 | if (write(d->fd, super->migr_rec_buf, |
5038 | MIGR_REC_BUF_SIZE) != MIGR_REC_BUF_SIZE) | |
9e2d750d | 5039 | perror("Write migr_rec failed"); |
146c6260 AK |
5040 | } |
5041 | } | |
c2c087e6 DW |
5042 | if (doclose) { |
5043 | close(d->fd); | |
5044 | d->fd = -1; | |
5045 | } | |
5046 | } | |
5047 | ||
d23fe947 DW |
5048 | if (spares) |
5049 | return write_super_imsm_spares(super, doclose); | |
5050 | ||
e74255d9 | 5051 | return 0; |
c2c087e6 DW |
5052 | } |
5053 | ||
0e600426 | 5054 | |
9b1fb677 | 5055 | static int create_array(struct supertype *st, int dev_idx) |
43dad3d6 DW |
5056 | { |
5057 | size_t len; | |
5058 | struct imsm_update_create_array *u; | |
5059 | struct intel_super *super = st->sb; | |
9b1fb677 | 5060 | struct imsm_dev *dev = get_imsm_dev(super, dev_idx); |
238c0a71 | 5061 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
54c2c1ea DW |
5062 | struct disk_info *inf; |
5063 | struct imsm_disk *disk; | |
5064 | int i; | |
43dad3d6 | 5065 | |
54c2c1ea DW |
5066 | len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) + |
5067 | sizeof(*inf) * map->num_members; | |
43dad3d6 DW |
5068 | u = malloc(len); |
5069 | if (!u) { | |
5070 | fprintf(stderr, "%s: failed to allocate update buffer\n", | |
5071 | __func__); | |
5072 | return 1; | |
5073 | } | |
5074 | ||
5075 | u->type = update_create_array; | |
9b1fb677 | 5076 | u->dev_idx = dev_idx; |
43dad3d6 | 5077 | imsm_copy_dev(&u->dev, dev); |
54c2c1ea DW |
5078 | inf = get_disk_info(u); |
5079 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 5080 | int idx = get_imsm_disk_idx(dev, i, MAP_X); |
9b1fb677 | 5081 | |
54c2c1ea DW |
5082 | disk = get_imsm_disk(super, idx); |
5083 | serialcpy(inf[i].serial, disk->serial); | |
5084 | } | |
43dad3d6 DW |
5085 | append_metadata_update(st, u, len); |
5086 | ||
5087 | return 0; | |
5088 | } | |
5089 | ||
1a64be56 | 5090 | static int mgmt_disk(struct supertype *st) |
43dad3d6 DW |
5091 | { |
5092 | struct intel_super *super = st->sb; | |
5093 | size_t len; | |
1a64be56 | 5094 | struct imsm_update_add_remove_disk *u; |
43dad3d6 | 5095 | |
1a64be56 | 5096 | if (!super->disk_mgmt_list) |
43dad3d6 DW |
5097 | return 0; |
5098 | ||
5099 | len = sizeof(*u); | |
5100 | u = malloc(len); | |
5101 | if (!u) { | |
5102 | fprintf(stderr, "%s: failed to allocate update buffer\n", | |
5103 | __func__); | |
5104 | return 1; | |
5105 | } | |
5106 | ||
1a64be56 | 5107 | u->type = update_add_remove_disk; |
43dad3d6 DW |
5108 | append_metadata_update(st, u, len); |
5109 | ||
5110 | return 0; | |
5111 | } | |
5112 | ||
c2c087e6 DW |
5113 | static int write_init_super_imsm(struct supertype *st) |
5114 | { | |
9b1fb677 DW |
5115 | struct intel_super *super = st->sb; |
5116 | int current_vol = super->current_vol; | |
5117 | ||
5118 | /* we are done with current_vol reset it to point st at the container */ | |
5119 | super->current_vol = -1; | |
5120 | ||
8273f55e | 5121 | if (st->update_tail) { |
43dad3d6 DW |
5122 | /* queue the recently created array / added disk |
5123 | * as a metadata update */ | |
43dad3d6 | 5124 | int rv; |
8273f55e | 5125 | |
43dad3d6 | 5126 | /* determine if we are creating a volume or adding a disk */ |
9b1fb677 | 5127 | if (current_vol < 0) { |
1a64be56 LM |
5128 | /* in the mgmt (add/remove) disk case we are running |
5129 | * in mdmon context, so don't close fd's | |
43dad3d6 | 5130 | */ |
1a64be56 | 5131 | return mgmt_disk(st); |
43dad3d6 | 5132 | } else |
9b1fb677 | 5133 | rv = create_array(st, current_vol); |
8273f55e | 5134 | |
43dad3d6 | 5135 | return rv; |
d682f344 N |
5136 | } else { |
5137 | struct dl *d; | |
5138 | for (d = super->disks; d; d = d->next) | |
5139 | Kill(d->devname, NULL, 0, 1, 1); | |
36988a3d | 5140 | return write_super_imsm(st, 1); |
d682f344 | 5141 | } |
cdddbdbc | 5142 | } |
0e600426 | 5143 | #endif |
cdddbdbc | 5144 | |
e683ca88 | 5145 | static int store_super_imsm(struct supertype *st, int fd) |
cdddbdbc | 5146 | { |
e683ca88 DW |
5147 | struct intel_super *super = st->sb; |
5148 | struct imsm_super *mpb = super ? super->anchor : NULL; | |
551c80c1 | 5149 | |
e683ca88 | 5150 | if (!mpb) |
ad97895e DW |
5151 | return 1; |
5152 | ||
1799c9e8 | 5153 | #ifndef MDASSEMBLE |
e683ca88 | 5154 | return store_imsm_mpb(fd, mpb); |
1799c9e8 N |
5155 | #else |
5156 | return 1; | |
5157 | #endif | |
cdddbdbc DW |
5158 | } |
5159 | ||
0e600426 N |
5160 | static int imsm_bbm_log_size(struct imsm_super *mpb) |
5161 | { | |
5162 | return __le32_to_cpu(mpb->bbm_log_size); | |
5163 | } | |
5164 | ||
5165 | #ifndef MDASSEMBLE | |
cdddbdbc DW |
5166 | static int validate_geometry_imsm_container(struct supertype *st, int level, |
5167 | int layout, int raiddisks, int chunk, | |
c2c087e6 | 5168 | unsigned long long size, char *dev, |
2c514b71 NB |
5169 | unsigned long long *freesize, |
5170 | int verbose) | |
cdddbdbc | 5171 | { |
c2c087e6 DW |
5172 | int fd; |
5173 | unsigned long long ldsize; | |
f2f5c343 LM |
5174 | struct intel_super *super=NULL; |
5175 | int rv = 0; | |
cdddbdbc | 5176 | |
c2c087e6 DW |
5177 | if (level != LEVEL_CONTAINER) |
5178 | return 0; | |
5179 | if (!dev) | |
5180 | return 1; | |
5181 | ||
5182 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
5183 | if (fd < 0) { | |
2c514b71 NB |
5184 | if (verbose) |
5185 | fprintf(stderr, Name ": imsm: Cannot open %s: %s\n", | |
5186 | dev, strerror(errno)); | |
c2c087e6 DW |
5187 | return 0; |
5188 | } | |
5189 | if (!get_dev_size(fd, dev, &ldsize)) { | |
5190 | close(fd); | |
5191 | return 0; | |
5192 | } | |
f2f5c343 LM |
5193 | |
5194 | /* capabilities retrieve could be possible | |
5195 | * note that there is no fd for the disks in array. | |
5196 | */ | |
5197 | super = alloc_super(); | |
5198 | if (!super) { | |
5199 | fprintf(stderr, | |
5200 | Name ": malloc of %zu failed.\n", | |
5201 | sizeof(*super)); | |
5202 | close(fd); | |
5203 | return 0; | |
5204 | } | |
5205 | ||
d424212e | 5206 | rv = find_intel_hba_capability(fd, super, verbose ? dev : NULL); |
f2f5c343 LM |
5207 | if (rv != 0) { |
5208 | #if DEBUG | |
5209 | char str[256]; | |
5210 | fd2devname(fd, str); | |
5211 | dprintf("validate_geometry_imsm_container: fd: %d %s orom: %p rv: %d raiddisk: %d\n", | |
5212 | fd, str, super->orom, rv, raiddisks); | |
5213 | #endif | |
5214 | /* no orom/efi or non-intel hba of the disk */ | |
5215 | close(fd); | |
5216 | free_imsm(super); | |
5217 | return 0; | |
5218 | } | |
c2c087e6 | 5219 | close(fd); |
f2f5c343 LM |
5220 | if (super->orom && raiddisks > super->orom->tds) { |
5221 | if (verbose) | |
5222 | fprintf(stderr, Name ": %d exceeds maximum number of" | |
5223 | " platform supported disks: %d\n", | |
5224 | raiddisks, super->orom->tds); | |
5225 | ||
5226 | free_imsm(super); | |
5227 | return 0; | |
5228 | } | |
c2c087e6 DW |
5229 | |
5230 | *freesize = avail_size_imsm(st, ldsize >> 9); | |
f2f5c343 | 5231 | free_imsm(super); |
c2c087e6 DW |
5232 | |
5233 | return 1; | |
cdddbdbc DW |
5234 | } |
5235 | ||
0dcecb2e DW |
5236 | static unsigned long long find_size(struct extent *e, int *idx, int num_extents) |
5237 | { | |
5238 | const unsigned long long base_start = e[*idx].start; | |
5239 | unsigned long long end = base_start + e[*idx].size; | |
5240 | int i; | |
5241 | ||
5242 | if (base_start == end) | |
5243 | return 0; | |
5244 | ||
5245 | *idx = *idx + 1; | |
5246 | for (i = *idx; i < num_extents; i++) { | |
5247 | /* extend overlapping extents */ | |
5248 | if (e[i].start >= base_start && | |
5249 | e[i].start <= end) { | |
5250 | if (e[i].size == 0) | |
5251 | return 0; | |
5252 | if (e[i].start + e[i].size > end) | |
5253 | end = e[i].start + e[i].size; | |
5254 | } else if (e[i].start > end) { | |
5255 | *idx = i; | |
5256 | break; | |
5257 | } | |
5258 | } | |
5259 | ||
5260 | return end - base_start; | |
5261 | } | |
5262 | ||
5263 | static unsigned long long merge_extents(struct intel_super *super, int sum_extents) | |
5264 | { | |
5265 | /* build a composite disk with all known extents and generate a new | |
5266 | * 'maxsize' given the "all disks in an array must share a common start | |
5267 | * offset" constraint | |
5268 | */ | |
5269 | struct extent *e = calloc(sum_extents, sizeof(*e)); | |
5270 | struct dl *dl; | |
5271 | int i, j; | |
5272 | int start_extent; | |
5273 | unsigned long long pos; | |
b9d77223 | 5274 | unsigned long long start = 0; |
0dcecb2e DW |
5275 | unsigned long long maxsize; |
5276 | unsigned long reserve; | |
5277 | ||
5278 | if (!e) | |
a7dd165b | 5279 | return 0; |
0dcecb2e DW |
5280 | |
5281 | /* coalesce and sort all extents. also, check to see if we need to | |
5282 | * reserve space between member arrays | |
5283 | */ | |
5284 | j = 0; | |
5285 | for (dl = super->disks; dl; dl = dl->next) { | |
5286 | if (!dl->e) | |
5287 | continue; | |
5288 | for (i = 0; i < dl->extent_cnt; i++) | |
5289 | e[j++] = dl->e[i]; | |
5290 | } | |
5291 | qsort(e, sum_extents, sizeof(*e), cmp_extent); | |
5292 | ||
5293 | /* merge extents */ | |
5294 | i = 0; | |
5295 | j = 0; | |
5296 | while (i < sum_extents) { | |
5297 | e[j].start = e[i].start; | |
5298 | e[j].size = find_size(e, &i, sum_extents); | |
5299 | j++; | |
5300 | if (e[j-1].size == 0) | |
5301 | break; | |
5302 | } | |
5303 | ||
5304 | pos = 0; | |
5305 | maxsize = 0; | |
5306 | start_extent = 0; | |
5307 | i = 0; | |
5308 | do { | |
5309 | unsigned long long esize; | |
5310 | ||
5311 | esize = e[i].start - pos; | |
5312 | if (esize >= maxsize) { | |
5313 | maxsize = esize; | |
5314 | start = pos; | |
5315 | start_extent = i; | |
5316 | } | |
5317 | pos = e[i].start + e[i].size; | |
5318 | i++; | |
5319 | } while (e[i-1].size); | |
5320 | free(e); | |
5321 | ||
a7dd165b DW |
5322 | if (maxsize == 0) |
5323 | return 0; | |
5324 | ||
5325 | /* FIXME assumes volume at offset 0 is the first volume in a | |
5326 | * container | |
5327 | */ | |
0dcecb2e DW |
5328 | if (start_extent > 0) |
5329 | reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */ | |
5330 | else | |
5331 | reserve = 0; | |
5332 | ||
5333 | if (maxsize < reserve) | |
a7dd165b | 5334 | return 0; |
0dcecb2e DW |
5335 | |
5336 | super->create_offset = ~((__u32) 0); | |
5337 | if (start + reserve > super->create_offset) | |
a7dd165b | 5338 | return 0; /* start overflows create_offset */ |
0dcecb2e DW |
5339 | super->create_offset = start + reserve; |
5340 | ||
5341 | return maxsize - reserve; | |
5342 | } | |
5343 | ||
88c32bb1 DW |
5344 | static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks) |
5345 | { | |
5346 | if (level < 0 || level == 6 || level == 4) | |
5347 | return 0; | |
5348 | ||
5349 | /* if we have an orom prevent invalid raid levels */ | |
5350 | if (orom) | |
5351 | switch (level) { | |
5352 | case 0: return imsm_orom_has_raid0(orom); | |
5353 | case 1: | |
5354 | if (raiddisks > 2) | |
5355 | return imsm_orom_has_raid1e(orom); | |
1c556e92 DW |
5356 | return imsm_orom_has_raid1(orom) && raiddisks == 2; |
5357 | case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4; | |
5358 | case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2; | |
88c32bb1 DW |
5359 | } |
5360 | else | |
5361 | return 1; /* not on an Intel RAID platform so anything goes */ | |
5362 | ||
5363 | return 0; | |
5364 | } | |
5365 | ||
ca9de185 LM |
5366 | |
5367 | static int | |
5368 | active_arrays_by_format(char *name, char* hba, struct md_list **devlist, | |
5369 | int dpa, int verbose) | |
5370 | { | |
5371 | struct mdstat_ent *mdstat = mdstat_read(0, 0); | |
5372 | struct mdstat_ent *memb = NULL; | |
5373 | int count = 0; | |
5374 | int num = 0; | |
5375 | struct md_list *dv = NULL; | |
5376 | int found; | |
5377 | ||
5378 | for (memb = mdstat ; memb ; memb = memb->next) { | |
5379 | if (memb->metadata_version && | |
5380 | (strncmp(memb->metadata_version, "external:", 9) == 0) && | |
5381 | (strcmp(&memb->metadata_version[9], name) == 0) && | |
5382 | !is_subarray(memb->metadata_version+9) && | |
5383 | memb->members) { | |
5384 | struct dev_member *dev = memb->members; | |
5385 | int fd = -1; | |
5386 | while(dev && (fd < 0)) { | |
5387 | char *path = malloc(strlen(dev->name) + strlen("/dev/") + 1); | |
5388 | if (path) { | |
5389 | num = sprintf(path, "%s%s", "/dev/", dev->name); | |
5390 | if (num > 0) | |
5391 | fd = open(path, O_RDONLY, 0); | |
5392 | if ((num <= 0) || (fd < 0)) { | |
5393 | pr_vrb(": Cannot open %s: %s\n", | |
5394 | dev->name, strerror(errno)); | |
5395 | } | |
5396 | free(path); | |
5397 | } | |
5398 | dev = dev->next; | |
5399 | } | |
5400 | found = 0; | |
5401 | if ((fd >= 0) && disk_attached_to_hba(fd, hba)) { | |
5402 | struct mdstat_ent *vol; | |
5403 | for (vol = mdstat ; vol ; vol = vol->next) { | |
5404 | if ((vol->active > 0) && | |
5405 | vol->metadata_version && | |
5406 | is_container_member(vol, memb->dev)) { | |
5407 | found++; | |
5408 | count++; | |
5409 | } | |
5410 | } | |
5411 | if (*devlist && (found < dpa)) { | |
5412 | dv = calloc(1, sizeof(*dv)); | |
5413 | if (dv == NULL) | |
5414 | fprintf(stderr, Name ": calloc failed\n"); | |
5415 | else { | |
5416 | dv->devname = malloc(strlen(memb->dev) + strlen("/dev/") + 1); | |
5417 | if (dv->devname != NULL) { | |
5418 | sprintf(dv->devname, "%s%s", "/dev/", memb->dev); | |
5419 | dv->found = found; | |
5420 | dv->used = 0; | |
5421 | dv->next = *devlist; | |
5422 | *devlist = dv; | |
5423 | } else | |
5424 | free(dv); | |
5425 | } | |
5426 | } | |
5427 | } | |
5428 | if (fd >= 0) | |
5429 | close(fd); | |
5430 | } | |
5431 | } | |
5432 | free_mdstat(mdstat); | |
5433 | return count; | |
5434 | } | |
5435 | ||
5436 | #ifdef DEBUG_LOOP | |
5437 | static struct md_list* | |
5438 | get_loop_devices(void) | |
5439 | { | |
5440 | int i; | |
5441 | struct md_list *devlist = NULL; | |
5442 | struct md_list *dv = NULL; | |
5443 | ||
5444 | for(i = 0; i < 12; i++) { | |
5445 | dv = calloc(1, sizeof(*dv)); | |
5446 | if (dv == NULL) { | |
5447 | fprintf(stderr, Name ": calloc failed\n"); | |
5448 | break; | |
5449 | } | |
5450 | dv->devname = malloc(40); | |
5451 | if (dv->devname == NULL) { | |
5452 | fprintf(stderr, Name ": malloc failed\n"); | |
5453 | free(dv); | |
5454 | break; | |
5455 | } | |
5456 | sprintf(dv->devname, "/dev/loop%d", i); | |
5457 | dv->next = devlist; | |
5458 | devlist = dv; | |
5459 | } | |
5460 | return devlist; | |
5461 | } | |
5462 | #endif | |
5463 | ||
5464 | static struct md_list* | |
5465 | get_devices(const char *hba_path) | |
5466 | { | |
5467 | struct md_list *devlist = NULL; | |
5468 | struct md_list *dv = NULL; | |
5469 | struct dirent *ent; | |
5470 | DIR *dir; | |
5471 | int err = 0; | |
5472 | ||
5473 | #if DEBUG_LOOP | |
5474 | devlist = get_loop_devices(); | |
5475 | return devlist; | |
5476 | #endif | |
5477 | /* scroll through /sys/dev/block looking for devices attached to | |
5478 | * this hba | |
5479 | */ | |
5480 | dir = opendir("/sys/dev/block"); | |
5481 | for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) { | |
5482 | int fd; | |
5483 | char buf[1024]; | |
5484 | int major, minor; | |
5485 | char *path = NULL; | |
5486 | if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2) | |
5487 | continue; | |
5488 | path = devt_to_devpath(makedev(major, minor)); | |
5489 | if (!path) | |
5490 | continue; | |
5491 | if (!path_attached_to_hba(path, hba_path)) { | |
5492 | free(path); | |
5493 | path = NULL; | |
5494 | continue; | |
5495 | } | |
5496 | free(path); | |
5497 | path = NULL; | |
5498 | fd = dev_open(ent->d_name, O_RDONLY); | |
5499 | if (fd >= 0) { | |
5500 | fd2devname(fd, buf); | |
5501 | close(fd); | |
5502 | } else { | |
5503 | fprintf(stderr, Name ": cannot open device: %s\n", | |
5504 | ent->d_name); | |
5505 | continue; | |
5506 | } | |
5507 | ||
5508 | ||
5509 | dv = calloc(1, sizeof(*dv)); | |
5510 | if (dv == NULL) { | |
5511 | fprintf(stderr, Name ": malloc failed\n"); | |
5512 | err = 1; | |
5513 | break; | |
5514 | } | |
5515 | dv->devname = strdup(buf); | |
5516 | if (dv->devname == NULL) { | |
5517 | fprintf(stderr, Name ": malloc failed\n"); | |
5518 | err = 1; | |
5519 | free(dv); | |
5520 | break; | |
5521 | } | |
5522 | dv->next = devlist; | |
5523 | devlist = dv; | |
5524 | } | |
5525 | if (err) { | |
5526 | while(devlist) { | |
5527 | dv = devlist; | |
5528 | devlist = devlist->next; | |
5529 | free(dv->devname); | |
5530 | free(dv); | |
5531 | } | |
5532 | } | |
5533 | return devlist; | |
5534 | } | |
5535 | ||
5536 | static int | |
5537 | count_volumes_list(struct md_list *devlist, char *homehost, | |
5538 | int verbose, int *found) | |
5539 | { | |
5540 | struct md_list *tmpdev; | |
5541 | int count = 0; | |
5542 | struct supertype *st = NULL; | |
5543 | ||
5544 | /* first walk the list of devices to find a consistent set | |
5545 | * that match the criterea, if that is possible. | |
5546 | * We flag the ones we like with 'used'. | |
5547 | */ | |
5548 | *found = 0; | |
5549 | st = match_metadata_desc_imsm("imsm"); | |
5550 | if (st == NULL) { | |
5551 | pr_vrb(": cannot allocate memory for imsm supertype\n"); | |
5552 | return 0; | |
5553 | } | |
5554 | ||
5555 | for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { | |
5556 | char *devname = tmpdev->devname; | |
5557 | struct stat stb; | |
5558 | struct supertype *tst; | |
5559 | int dfd; | |
5560 | if (tmpdev->used > 1) | |
5561 | continue; | |
5562 | tst = dup_super(st); | |
5563 | if (tst == NULL) { | |
5564 | pr_vrb(": cannot allocate memory for imsm supertype\n"); | |
5565 | goto err_1; | |
5566 | } | |
5567 | tmpdev->container = 0; | |
5568 | dfd = dev_open(devname, O_RDONLY|O_EXCL); | |
5569 | if (dfd < 0) { | |
5570 | dprintf(": cannot open device %s: %s\n", | |
5571 | devname, strerror(errno)); | |
5572 | tmpdev->used = 2; | |
5573 | } else if (fstat(dfd, &stb)< 0) { | |
5574 | /* Impossible! */ | |
5575 | dprintf(": fstat failed for %s: %s\n", | |
5576 | devname, strerror(errno)); | |
5577 | tmpdev->used = 2; | |
5578 | } else if ((stb.st_mode & S_IFMT) != S_IFBLK) { | |
5579 | dprintf(": %s is not a block device.\n", | |
5580 | devname); | |
5581 | tmpdev->used = 2; | |
5582 | } else if (must_be_container(dfd)) { | |
5583 | struct supertype *cst; | |
5584 | cst = super_by_fd(dfd, NULL); | |
5585 | if (cst == NULL) { | |
5586 | dprintf(": cannot recognize container type %s\n", | |
5587 | devname); | |
5588 | tmpdev->used = 2; | |
5589 | } else if (tst->ss != st->ss) { | |
5590 | dprintf(": non-imsm container - ignore it: %s\n", | |
5591 | devname); | |
5592 | tmpdev->used = 2; | |
5593 | } else if (!tst->ss->load_container || | |
5594 | tst->ss->load_container(tst, dfd, NULL)) | |
5595 | tmpdev->used = 2; | |
5596 | else { | |
5597 | tmpdev->container = 1; | |
5598 | } | |
5599 | if (cst) | |
5600 | cst->ss->free_super(cst); | |
5601 | } else { | |
5602 | tmpdev->st_rdev = stb.st_rdev; | |
5603 | if (tst->ss->load_super(tst,dfd, NULL)) { | |
5604 | dprintf(": no RAID superblock on %s\n", | |
5605 | devname); | |
5606 | tmpdev->used = 2; | |
5607 | } else if (tst->ss->compare_super == NULL) { | |
5608 | dprintf(": Cannot assemble %s metadata on %s\n", | |
5609 | tst->ss->name, devname); | |
5610 | tmpdev->used = 2; | |
5611 | } | |
5612 | } | |
5613 | if (dfd >= 0) | |
5614 | close(dfd); | |
5615 | if (tmpdev->used == 2 || tmpdev->used == 4) { | |
5616 | /* Ignore unrecognised devices during auto-assembly */ | |
5617 | goto loop; | |
5618 | } | |
5619 | else { | |
5620 | struct mdinfo info; | |
5621 | tst->ss->getinfo_super(tst, &info, NULL); | |
5622 | ||
5623 | if (st->minor_version == -1) | |
5624 | st->minor_version = tst->minor_version; | |
5625 | ||
5626 | if (memcmp(info.uuid, uuid_zero, | |
5627 | sizeof(int[4])) == 0) { | |
5628 | /* this is a floating spare. It cannot define | |
5629 | * an array unless there are no more arrays of | |
5630 | * this type to be found. It can be included | |
5631 | * in an array of this type though. | |
5632 | */ | |
5633 | tmpdev->used = 3; | |
5634 | goto loop; | |
5635 | } | |
5636 | ||
5637 | if (st->ss != tst->ss || | |
5638 | st->minor_version != tst->minor_version || | |
5639 | st->ss->compare_super(st, tst) != 0) { | |
5640 | /* Some mismatch. If exactly one array matches this host, | |
5641 | * we can resolve on that one. | |
5642 | * Or, if we are auto assembling, we just ignore the second | |
5643 | * for now. | |
5644 | */ | |
5645 | dprintf(": superblock on %s doesn't match others - assembly aborted\n", | |
5646 | devname); | |
5647 | goto loop; | |
5648 | } | |
5649 | tmpdev->used = 1; | |
5650 | *found = 1; | |
5651 | dprintf("found: devname: %s\n", devname); | |
5652 | } | |
5653 | loop: | |
5654 | if (tst) | |
5655 | tst->ss->free_super(tst); | |
5656 | } | |
5657 | if (*found != 0) { | |
5658 | int err; | |
5659 | if ((err = load_super_imsm_all(st, -1, &st->sb, NULL, devlist, 0)) == 0) { | |
5660 | struct mdinfo *iter, *head = st->ss->container_content(st, NULL); | |
5661 | for (iter = head; iter; iter = iter->next) { | |
5662 | dprintf("content->text_version: %s vol\n", | |
5663 | iter->text_version); | |
5664 | if (iter->array.state & (1<<MD_SB_BLOCK_VOLUME)) { | |
5665 | /* do not assemble arrays with unsupported | |
5666 | configurations */ | |
5667 | dprintf(": Cannot activate member %s.\n", | |
5668 | iter->text_version); | |
5669 | } else | |
5670 | count++; | |
5671 | } | |
5672 | sysfs_free(head); | |
5673 | ||
5674 | } else { | |
5675 | dprintf(" no valid super block on device list: err: %d %p\n", | |
5676 | err, st->sb); | |
5677 | } | |
5678 | } else { | |
5679 | dprintf(" no more devices to examin\n"); | |
5680 | } | |
5681 | ||
5682 | for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { | |
5683 | if ((tmpdev->used == 1) && (tmpdev->found)) { | |
5684 | if (count) { | |
5685 | if (count < tmpdev->found) | |
5686 | count = 0; | |
5687 | else | |
5688 | count -= tmpdev->found; | |
5689 | } | |
5690 | } | |
5691 | if (tmpdev->used == 1) | |
5692 | tmpdev->used = 4; | |
5693 | } | |
5694 | err_1: | |
5695 | if (st) | |
5696 | st->ss->free_super(st); | |
5697 | return count; | |
5698 | } | |
5699 | ||
5700 | ||
5701 | static int | |
5702 | count_volumes(char *hba, int dpa, int verbose) | |
5703 | { | |
5704 | struct md_list *devlist = NULL; | |
5705 | int count = 0; | |
5706 | int found = 0;; | |
5707 | ||
5708 | devlist = get_devices(hba); | |
5709 | /* if no intel devices return zero volumes */ | |
5710 | if (devlist == NULL) | |
5711 | return 0; | |
5712 | ||
5713 | count = active_arrays_by_format("imsm", hba, &devlist, dpa, verbose); | |
5714 | dprintf(" path: %s active arrays: %d\n", hba, count); | |
5715 | if (devlist == NULL) | |
5716 | return 0; | |
5717 | do { | |
5718 | found = 0; | |
5719 | count += count_volumes_list(devlist, | |
5720 | NULL, | |
5721 | verbose, | |
5722 | &found); | |
5723 | dprintf("found %d count: %d\n", found, count); | |
5724 | } while (found); | |
5725 | ||
5726 | dprintf("path: %s total number of volumes: %d\n", hba, count); | |
5727 | ||
5728 | while(devlist) { | |
5729 | struct md_list *dv = devlist; | |
5730 | devlist = devlist->next; | |
5731 | free(dv->devname); | |
5732 | free(dv); | |
5733 | } | |
5734 | return count; | |
5735 | } | |
5736 | ||
cd9d1ac7 DW |
5737 | static int imsm_default_chunk(const struct imsm_orom *orom) |
5738 | { | |
5739 | /* up to 512 if the plaform supports it, otherwise the platform max. | |
5740 | * 128 if no platform detected | |
5741 | */ | |
5742 | int fs = max(7, orom ? fls(orom->sss) : 0); | |
5743 | ||
5744 | return min(512, (1 << fs)); | |
5745 | } | |
73408129 | 5746 | |
6592ce37 DW |
5747 | static int |
5748 | validate_geometry_imsm_orom(struct intel_super *super, int level, int layout, | |
c21e737b | 5749 | int raiddisks, int *chunk, int verbose) |
6592ce37 | 5750 | { |
660260d0 DW |
5751 | /* check/set platform and metadata limits/defaults */ |
5752 | if (super->orom && raiddisks > super->orom->dpa) { | |
5753 | pr_vrb(": platform supports a maximum of %d disks per array\n", | |
5754 | super->orom->dpa); | |
73408129 LM |
5755 | return 0; |
5756 | } | |
5757 | ||
5758 | /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */ | |
660260d0 | 5759 | if (!is_raid_level_supported(super->orom, level, raiddisks)) { |
6592ce37 DW |
5760 | pr_vrb(": platform does not support raid%d with %d disk%s\n", |
5761 | level, raiddisks, raiddisks > 1 ? "s" : ""); | |
5762 | return 0; | |
5763 | } | |
cd9d1ac7 DW |
5764 | |
5765 | if (chunk && (*chunk == 0 || *chunk == UnSet)) | |
5766 | *chunk = imsm_default_chunk(super->orom); | |
5767 | ||
5768 | if (super->orom && chunk && !imsm_orom_has_chunk(super->orom, *chunk)) { | |
5769 | pr_vrb(": platform does not support a chunk size of: " | |
5770 | "%d\n", *chunk); | |
5771 | return 0; | |
6592ce37 | 5772 | } |
cd9d1ac7 | 5773 | |
6592ce37 DW |
5774 | if (layout != imsm_level_to_layout(level)) { |
5775 | if (level == 5) | |
5776 | pr_vrb(": imsm raid 5 only supports the left-asymmetric layout\n"); | |
5777 | else if (level == 10) | |
5778 | pr_vrb(": imsm raid 10 only supports the n2 layout\n"); | |
5779 | else | |
5780 | pr_vrb(": imsm unknown layout %#x for this raid level %d\n", | |
5781 | layout, level); | |
5782 | return 0; | |
5783 | } | |
6592ce37 DW |
5784 | return 1; |
5785 | } | |
5786 | ||
c2c087e6 DW |
5787 | /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd |
5788 | * FIX ME add ahci details | |
5789 | */ | |
8b353278 | 5790 | static int validate_geometry_imsm_volume(struct supertype *st, int level, |
c21e737b | 5791 | int layout, int raiddisks, int *chunk, |
c2c087e6 | 5792 | unsigned long long size, char *dev, |
2c514b71 NB |
5793 | unsigned long long *freesize, |
5794 | int verbose) | |
cdddbdbc | 5795 | { |
c2c087e6 DW |
5796 | struct stat stb; |
5797 | struct intel_super *super = st->sb; | |
b2916f25 | 5798 | struct imsm_super *mpb; |
c2c087e6 DW |
5799 | struct dl *dl; |
5800 | unsigned long long pos = 0; | |
5801 | unsigned long long maxsize; | |
5802 | struct extent *e; | |
5803 | int i; | |
cdddbdbc | 5804 | |
88c32bb1 DW |
5805 | /* We must have the container info already read in. */ |
5806 | if (!super) | |
c2c087e6 DW |
5807 | return 0; |
5808 | ||
b2916f25 JS |
5809 | mpb = super->anchor; |
5810 | ||
d54559f0 LM |
5811 | if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, verbose)) { |
5812 | fprintf(stderr, Name ": RAID gemetry validation failed. " | |
5813 | "Cannot proceed with the action(s).\n"); | |
c2c087e6 | 5814 | return 0; |
d54559f0 | 5815 | } |
c2c087e6 DW |
5816 | if (!dev) { |
5817 | /* General test: make sure there is space for | |
2da8544a DW |
5818 | * 'raiddisks' device extents of size 'size' at a given |
5819 | * offset | |
c2c087e6 | 5820 | */ |
e46273eb | 5821 | unsigned long long minsize = size; |
b7528a20 | 5822 | unsigned long long start_offset = MaxSector; |
c2c087e6 DW |
5823 | int dcnt = 0; |
5824 | if (minsize == 0) | |
5825 | minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
5826 | for (dl = super->disks; dl ; dl = dl->next) { | |
5827 | int found = 0; | |
5828 | ||
bf5a934a | 5829 | pos = 0; |
c2c087e6 DW |
5830 | i = 0; |
5831 | e = get_extents(super, dl); | |
5832 | if (!e) continue; | |
5833 | do { | |
5834 | unsigned long long esize; | |
5835 | esize = e[i].start - pos; | |
5836 | if (esize >= minsize) | |
5837 | found = 1; | |
b7528a20 | 5838 | if (found && start_offset == MaxSector) { |
2da8544a DW |
5839 | start_offset = pos; |
5840 | break; | |
5841 | } else if (found && pos != start_offset) { | |
5842 | found = 0; | |
5843 | break; | |
5844 | } | |
c2c087e6 DW |
5845 | pos = e[i].start + e[i].size; |
5846 | i++; | |
5847 | } while (e[i-1].size); | |
5848 | if (found) | |
5849 | dcnt++; | |
5850 | free(e); | |
5851 | } | |
5852 | if (dcnt < raiddisks) { | |
2c514b71 NB |
5853 | if (verbose) |
5854 | fprintf(stderr, Name ": imsm: Not enough " | |
5855 | "devices with space for this array " | |
5856 | "(%d < %d)\n", | |
5857 | dcnt, raiddisks); | |
c2c087e6 DW |
5858 | return 0; |
5859 | } | |
5860 | return 1; | |
5861 | } | |
0dcecb2e | 5862 | |
c2c087e6 DW |
5863 | /* This device must be a member of the set */ |
5864 | if (stat(dev, &stb) < 0) | |
5865 | return 0; | |
5866 | if ((S_IFMT & stb.st_mode) != S_IFBLK) | |
5867 | return 0; | |
5868 | for (dl = super->disks ; dl ; dl = dl->next) { | |
f21e18ca N |
5869 | if (dl->major == (int)major(stb.st_rdev) && |
5870 | dl->minor == (int)minor(stb.st_rdev)) | |
c2c087e6 DW |
5871 | break; |
5872 | } | |
5873 | if (!dl) { | |
2c514b71 NB |
5874 | if (verbose) |
5875 | fprintf(stderr, Name ": %s is not in the " | |
5876 | "same imsm set\n", dev); | |
c2c087e6 | 5877 | return 0; |
a20d2ba5 DW |
5878 | } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) { |
5879 | /* If a volume is present then the current creation attempt | |
5880 | * cannot incorporate new spares because the orom may not | |
5881 | * understand this configuration (all member disks must be | |
5882 | * members of each array in the container). | |
5883 | */ | |
5884 | fprintf(stderr, Name ": %s is a spare and a volume" | |
5885 | " is already defined for this container\n", dev); | |
5886 | fprintf(stderr, Name ": The option-rom requires all member" | |
5887 | " disks to be a member of all volumes\n"); | |
5888 | return 0; | |
5fe62b94 WD |
5889 | } else if (super->orom && mpb->num_raid_devs > 0 && |
5890 | mpb->num_disks != raiddisks) { | |
5891 | fprintf(stderr, Name ": The option-rom requires all member" | |
5892 | " disks to be a member of all volumes\n"); | |
5893 | return 0; | |
c2c087e6 | 5894 | } |
0dcecb2e DW |
5895 | |
5896 | /* retrieve the largest free space block */ | |
c2c087e6 DW |
5897 | e = get_extents(super, dl); |
5898 | maxsize = 0; | |
5899 | i = 0; | |
0dcecb2e DW |
5900 | if (e) { |
5901 | do { | |
5902 | unsigned long long esize; | |
5903 | ||
5904 | esize = e[i].start - pos; | |
5905 | if (esize >= maxsize) | |
5906 | maxsize = esize; | |
5907 | pos = e[i].start + e[i].size; | |
5908 | i++; | |
5909 | } while (e[i-1].size); | |
5910 | dl->e = e; | |
5911 | dl->extent_cnt = i; | |
5912 | } else { | |
5913 | if (verbose) | |
5914 | fprintf(stderr, Name ": unable to determine free space for: %s\n", | |
5915 | dev); | |
5916 | return 0; | |
5917 | } | |
5918 | if (maxsize < size) { | |
5919 | if (verbose) | |
5920 | fprintf(stderr, Name ": %s not enough space (%llu < %llu)\n", | |
5921 | dev, maxsize, size); | |
5922 | return 0; | |
5923 | } | |
5924 | ||
5925 | /* count total number of extents for merge */ | |
5926 | i = 0; | |
5927 | for (dl = super->disks; dl; dl = dl->next) | |
5928 | if (dl->e) | |
5929 | i += dl->extent_cnt; | |
5930 | ||
5931 | maxsize = merge_extents(super, i); | |
3baa56ab LO |
5932 | |
5933 | if (!check_env("IMSM_NO_PLATFORM") && | |
5934 | mpb->num_raid_devs > 0 && size && size != maxsize) { | |
5935 | fprintf(stderr, Name ": attempting to create a second " | |
5936 | "volume with size less then remaining space. " | |
5937 | "Aborting...\n"); | |
5938 | return 0; | |
5939 | } | |
5940 | ||
a7dd165b | 5941 | if (maxsize < size || maxsize == 0) { |
b3071342 LD |
5942 | if (verbose) { |
5943 | if (maxsize == 0) | |
5944 | fprintf(stderr, Name ": no free space" | |
5945 | " left on device. Aborting...\n"); | |
5946 | else | |
5947 | fprintf(stderr, Name ": not enough space" | |
5948 | " to create volume of given size" | |
5949 | " (%llu < %llu). Aborting...\n", | |
5950 | maxsize, size); | |
5951 | } | |
0dcecb2e | 5952 | return 0; |
0dcecb2e DW |
5953 | } |
5954 | ||
c2c087e6 DW |
5955 | *freesize = maxsize; |
5956 | ||
ca9de185 LM |
5957 | if (super->orom) { |
5958 | int count = count_volumes(super->hba->path, | |
5959 | super->orom->dpa, verbose); | |
5960 | if (super->orom->vphba <= count) { | |
5961 | pr_vrb(": platform does not support more then %d raid volumes.\n", | |
5962 | super->orom->vphba); | |
5963 | return 0; | |
5964 | } | |
5965 | } | |
c2c087e6 | 5966 | return 1; |
cdddbdbc DW |
5967 | } |
5968 | ||
efb30e7f DW |
5969 | static int reserve_space(struct supertype *st, int raiddisks, |
5970 | unsigned long long size, int chunk, | |
5971 | unsigned long long *freesize) | |
5972 | { | |
5973 | struct intel_super *super = st->sb; | |
5974 | struct imsm_super *mpb = super->anchor; | |
5975 | struct dl *dl; | |
5976 | int i; | |
5977 | int extent_cnt; | |
5978 | struct extent *e; | |
5979 | unsigned long long maxsize; | |
5980 | unsigned long long minsize; | |
5981 | int cnt; | |
5982 | int used; | |
5983 | ||
5984 | /* find the largest common start free region of the possible disks */ | |
5985 | used = 0; | |
5986 | extent_cnt = 0; | |
5987 | cnt = 0; | |
5988 | for (dl = super->disks; dl; dl = dl->next) { | |
5989 | dl->raiddisk = -1; | |
5990 | ||
5991 | if (dl->index >= 0) | |
5992 | used++; | |
5993 | ||
5994 | /* don't activate new spares if we are orom constrained | |
5995 | * and there is already a volume active in the container | |
5996 | */ | |
5997 | if (super->orom && dl->index < 0 && mpb->num_raid_devs) | |
5998 | continue; | |
5999 | ||
6000 | e = get_extents(super, dl); | |
6001 | if (!e) | |
6002 | continue; | |
6003 | for (i = 1; e[i-1].size; i++) | |
6004 | ; | |
6005 | dl->e = e; | |
6006 | dl->extent_cnt = i; | |
6007 | extent_cnt += i; | |
6008 | cnt++; | |
6009 | } | |
6010 | ||
6011 | maxsize = merge_extents(super, extent_cnt); | |
6012 | minsize = size; | |
6013 | if (size == 0) | |
612e59d8 CA |
6014 | /* chunk is in K */ |
6015 | minsize = chunk * 2; | |
efb30e7f DW |
6016 | |
6017 | if (cnt < raiddisks || | |
6018 | (super->orom && used && used != raiddisks) || | |
a7dd165b DW |
6019 | maxsize < minsize || |
6020 | maxsize == 0) { | |
efb30e7f DW |
6021 | fprintf(stderr, Name ": not enough devices with space to create array.\n"); |
6022 | return 0; /* No enough free spaces large enough */ | |
6023 | } | |
6024 | ||
6025 | if (size == 0) { | |
6026 | size = maxsize; | |
6027 | if (chunk) { | |
612e59d8 CA |
6028 | size /= 2 * chunk; |
6029 | size *= 2 * chunk; | |
efb30e7f DW |
6030 | } |
6031 | } | |
6032 | ||
6033 | cnt = 0; | |
6034 | for (dl = super->disks; dl; dl = dl->next) | |
6035 | if (dl->e) | |
6036 | dl->raiddisk = cnt++; | |
6037 | ||
6038 | *freesize = size; | |
6039 | ||
6040 | return 1; | |
6041 | } | |
6042 | ||
bf5a934a | 6043 | static int validate_geometry_imsm(struct supertype *st, int level, int layout, |
c21e737b | 6044 | int raiddisks, int *chunk, unsigned long long size, |
bf5a934a DW |
6045 | char *dev, unsigned long long *freesize, |
6046 | int verbose) | |
6047 | { | |
6048 | int fd, cfd; | |
6049 | struct mdinfo *sra; | |
20cbe8d2 | 6050 | int is_member = 0; |
bf5a934a | 6051 | |
d54559f0 LM |
6052 | /* load capability |
6053 | * if given unused devices create a container | |
bf5a934a DW |
6054 | * if given given devices in a container create a member volume |
6055 | */ | |
6056 | if (level == LEVEL_CONTAINER) { | |
6057 | /* Must be a fresh device to add to a container */ | |
6058 | return validate_geometry_imsm_container(st, level, layout, | |
c21e737b CA |
6059 | raiddisks, |
6060 | chunk?*chunk:0, size, | |
bf5a934a DW |
6061 | dev, freesize, |
6062 | verbose); | |
6063 | } | |
9587c373 | 6064 | |
8592f29d | 6065 | if (!dev) { |
e91a3bad | 6066 | if (st->sb) { |
ca9de185 | 6067 | struct intel_super *super = st->sb; |
e91a3bad LM |
6068 | if (!validate_geometry_imsm_orom(st->sb, level, layout, |
6069 | raiddisks, chunk, | |
6070 | verbose)) | |
6071 | return 0; | |
efb30e7f DW |
6072 | /* we are being asked to automatically layout a |
6073 | * new volume based on the current contents of | |
6074 | * the container. If the the parameters can be | |
6075 | * satisfied reserve_space will record the disks, | |
6076 | * start offset, and size of the volume to be | |
6077 | * created. add_to_super and getinfo_super | |
6078 | * detect when autolayout is in progress. | |
6079 | */ | |
ca9de185 LM |
6080 | /* assuming that freesize is always given when array is |
6081 | created */ | |
6082 | if (super->orom && freesize) { | |
6083 | int count; | |
6084 | count = count_volumes(super->hba->path, | |
6085 | super->orom->dpa, verbose); | |
6086 | if (super->orom->vphba <= count) { | |
6087 | pr_vrb(": platform does not support more" | |
6088 | "then %d raid volumes.\n", | |
6089 | super->orom->vphba); | |
6090 | return 0; | |
6091 | } | |
6092 | } | |
e91a3bad LM |
6093 | if (freesize) |
6094 | return reserve_space(st, raiddisks, size, | |
6095 | chunk?*chunk:0, freesize); | |
8592f29d N |
6096 | } |
6097 | return 1; | |
6098 | } | |
bf5a934a DW |
6099 | if (st->sb) { |
6100 | /* creating in a given container */ | |
6101 | return validate_geometry_imsm_volume(st, level, layout, | |
6102 | raiddisks, chunk, size, | |
6103 | dev, freesize, verbose); | |
6104 | } | |
6105 | ||
bf5a934a DW |
6106 | /* This device needs to be a device in an 'imsm' container */ |
6107 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
6108 | if (fd >= 0) { | |
6109 | if (verbose) | |
6110 | fprintf(stderr, | |
6111 | Name ": Cannot create this array on device %s\n", | |
6112 | dev); | |
6113 | close(fd); | |
6114 | return 0; | |
6115 | } | |
6116 | if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) { | |
6117 | if (verbose) | |
6118 | fprintf(stderr, Name ": Cannot open %s: %s\n", | |
6119 | dev, strerror(errno)); | |
6120 | return 0; | |
6121 | } | |
6122 | /* Well, it is in use by someone, maybe an 'imsm' container. */ | |
6123 | cfd = open_container(fd); | |
20cbe8d2 | 6124 | close(fd); |
bf5a934a | 6125 | if (cfd < 0) { |
bf5a934a DW |
6126 | if (verbose) |
6127 | fprintf(stderr, Name ": Cannot use %s: It is busy\n", | |
6128 | dev); | |
6129 | return 0; | |
6130 | } | |
6131 | sra = sysfs_read(cfd, 0, GET_VERSION); | |
bf5a934a | 6132 | if (sra && sra->array.major_version == -1 && |
20cbe8d2 AW |
6133 | strcmp(sra->text_version, "imsm") == 0) |
6134 | is_member = 1; | |
6135 | sysfs_free(sra); | |
6136 | if (is_member) { | |
bf5a934a DW |
6137 | /* This is a member of a imsm container. Load the container |
6138 | * and try to create a volume | |
6139 | */ | |
6140 | struct intel_super *super; | |
6141 | ||
ec50f7b6 | 6142 | if (load_super_imsm_all(st, cfd, (void **) &super, NULL, NULL, 1) == 0) { |
bf5a934a DW |
6143 | st->sb = super; |
6144 | st->container_dev = fd2devnum(cfd); | |
6145 | close(cfd); | |
6146 | return validate_geometry_imsm_volume(st, level, layout, | |
6147 | raiddisks, chunk, | |
6148 | size, dev, | |
ecbd9e81 N |
6149 | freesize, 1) |
6150 | ? 1 : -1; | |
bf5a934a | 6151 | } |
20cbe8d2 | 6152 | } |
bf5a934a | 6153 | |
20cbe8d2 AW |
6154 | if (verbose) |
6155 | fprintf(stderr, Name ": failed container membership check\n"); | |
6156 | ||
6157 | close(cfd); | |
6158 | return 0; | |
bf5a934a | 6159 | } |
0bd16cf2 | 6160 | |
30f58b22 | 6161 | static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk) |
0bd16cf2 DJ |
6162 | { |
6163 | struct intel_super *super = st->sb; | |
6164 | ||
30f58b22 DW |
6165 | if (level && *level == UnSet) |
6166 | *level = LEVEL_CONTAINER; | |
6167 | ||
6168 | if (level && layout && *layout == UnSet) | |
6169 | *layout = imsm_level_to_layout(*level); | |
0bd16cf2 | 6170 | |
cd9d1ac7 DW |
6171 | if (chunk && (*chunk == UnSet || *chunk == 0)) |
6172 | *chunk = imsm_default_chunk(super->orom); | |
0bd16cf2 DJ |
6173 | } |
6174 | ||
33414a01 DW |
6175 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev); |
6176 | ||
6177 | static int kill_subarray_imsm(struct supertype *st) | |
6178 | { | |
6179 | /* remove the subarray currently referenced by ->current_vol */ | |
6180 | __u8 i; | |
6181 | struct intel_dev **dp; | |
6182 | struct intel_super *super = st->sb; | |
6183 | __u8 current_vol = super->current_vol; | |
6184 | struct imsm_super *mpb = super->anchor; | |
6185 | ||
6186 | if (super->current_vol < 0) | |
6187 | return 2; | |
6188 | super->current_vol = -1; /* invalidate subarray cursor */ | |
6189 | ||
6190 | /* block deletions that would change the uuid of active subarrays | |
6191 | * | |
6192 | * FIXME when immutable ids are available, but note that we'll | |
6193 | * also need to fixup the invalidated/active subarray indexes in | |
6194 | * mdstat | |
6195 | */ | |
6196 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
6197 | char subarray[4]; | |
6198 | ||
6199 | if (i < current_vol) | |
6200 | continue; | |
6201 | sprintf(subarray, "%u", i); | |
6202 | if (is_subarray_active(subarray, st->devname)) { | |
6203 | fprintf(stderr, | |
6204 | Name ": deleting subarray-%d would change the UUID of active subarray-%d, aborting\n", | |
6205 | current_vol, i); | |
6206 | ||
6207 | return 2; | |
6208 | } | |
6209 | } | |
6210 | ||
6211 | if (st->update_tail) { | |
6212 | struct imsm_update_kill_array *u = malloc(sizeof(*u)); | |
6213 | ||
6214 | if (!u) | |
6215 | return 2; | |
6216 | u->type = update_kill_array; | |
6217 | u->dev_idx = current_vol; | |
6218 | append_metadata_update(st, u, sizeof(*u)); | |
6219 | ||
6220 | return 0; | |
6221 | } | |
6222 | ||
6223 | for (dp = &super->devlist; *dp;) | |
6224 | if ((*dp)->index == current_vol) { | |
6225 | *dp = (*dp)->next; | |
6226 | } else { | |
6227 | handle_missing(super, (*dp)->dev); | |
6228 | if ((*dp)->index > current_vol) | |
6229 | (*dp)->index--; | |
6230 | dp = &(*dp)->next; | |
6231 | } | |
6232 | ||
6233 | /* no more raid devices, all active components are now spares, | |
6234 | * but of course failed are still failed | |
6235 | */ | |
6236 | if (--mpb->num_raid_devs == 0) { | |
6237 | struct dl *d; | |
6238 | ||
6239 | for (d = super->disks; d; d = d->next) | |
a8619d23 AK |
6240 | if (d->index > -2) |
6241 | mark_spare(d); | |
33414a01 DW |
6242 | } |
6243 | ||
6244 | super->updates_pending++; | |
6245 | ||
6246 | return 0; | |
6247 | } | |
aa534678 | 6248 | |
a951a4f7 | 6249 | static int update_subarray_imsm(struct supertype *st, char *subarray, |
fa56eddb | 6250 | char *update, struct mddev_ident *ident) |
aa534678 DW |
6251 | { |
6252 | /* update the subarray currently referenced by ->current_vol */ | |
6253 | struct intel_super *super = st->sb; | |
6254 | struct imsm_super *mpb = super->anchor; | |
6255 | ||
aa534678 DW |
6256 | if (strcmp(update, "name") == 0) { |
6257 | char *name = ident->name; | |
a951a4f7 N |
6258 | char *ep; |
6259 | int vol; | |
aa534678 | 6260 | |
a951a4f7 | 6261 | if (is_subarray_active(subarray, st->devname)) { |
aa534678 DW |
6262 | fprintf(stderr, |
6263 | Name ": Unable to update name of active subarray\n"); | |
6264 | return 2; | |
6265 | } | |
6266 | ||
6267 | if (!check_name(super, name, 0)) | |
6268 | return 2; | |
6269 | ||
a951a4f7 N |
6270 | vol = strtoul(subarray, &ep, 10); |
6271 | if (*ep != '\0' || vol >= super->anchor->num_raid_devs) | |
6272 | return 2; | |
6273 | ||
aa534678 DW |
6274 | if (st->update_tail) { |
6275 | struct imsm_update_rename_array *u = malloc(sizeof(*u)); | |
6276 | ||
6277 | if (!u) | |
6278 | return 2; | |
6279 | u->type = update_rename_array; | |
a951a4f7 | 6280 | u->dev_idx = vol; |
aa534678 DW |
6281 | snprintf((char *) u->name, MAX_RAID_SERIAL_LEN, "%s", name); |
6282 | append_metadata_update(st, u, sizeof(*u)); | |
6283 | } else { | |
6284 | struct imsm_dev *dev; | |
6285 | int i; | |
6286 | ||
a951a4f7 | 6287 | dev = get_imsm_dev(super, vol); |
aa534678 DW |
6288 | snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name); |
6289 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
6290 | dev = get_imsm_dev(super, i); | |
6291 | handle_missing(super, dev); | |
6292 | } | |
6293 | super->updates_pending++; | |
6294 | } | |
6295 | } else | |
6296 | return 2; | |
6297 | ||
6298 | return 0; | |
6299 | } | |
d1e02575 | 6300 | #endif /* MDASSEMBLE */ |
bf5a934a | 6301 | |
28bce06f AK |
6302 | static int is_gen_migration(struct imsm_dev *dev) |
6303 | { | |
7534230b AK |
6304 | if (dev == NULL) |
6305 | return 0; | |
6306 | ||
28bce06f AK |
6307 | if (!dev->vol.migr_state) |
6308 | return 0; | |
6309 | ||
6310 | if (migr_type(dev) == MIGR_GEN_MIGR) | |
6311 | return 1; | |
6312 | ||
6313 | return 0; | |
6314 | } | |
6315 | ||
1e5c6983 DW |
6316 | static int is_rebuilding(struct imsm_dev *dev) |
6317 | { | |
6318 | struct imsm_map *migr_map; | |
6319 | ||
6320 | if (!dev->vol.migr_state) | |
6321 | return 0; | |
6322 | ||
6323 | if (migr_type(dev) != MIGR_REBUILD) | |
6324 | return 0; | |
6325 | ||
238c0a71 | 6326 | migr_map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
6327 | |
6328 | if (migr_map->map_state == IMSM_T_STATE_DEGRADED) | |
6329 | return 1; | |
6330 | else | |
6331 | return 0; | |
6332 | } | |
6333 | ||
b4ab44d8 | 6334 | #ifndef MDASSEMBLE |
6ce1fbf1 AK |
6335 | static int is_initializing(struct imsm_dev *dev) |
6336 | { | |
6337 | struct imsm_map *migr_map; | |
6338 | ||
6339 | if (!dev->vol.migr_state) | |
6340 | return 0; | |
6341 | ||
6342 | if (migr_type(dev) != MIGR_INIT) | |
6343 | return 0; | |
6344 | ||
238c0a71 | 6345 | migr_map = get_imsm_map(dev, MAP_1); |
6ce1fbf1 AK |
6346 | |
6347 | if (migr_map->map_state == IMSM_T_STATE_UNINITIALIZED) | |
6348 | return 1; | |
6349 | ||
6350 | return 0; | |
6ce1fbf1 | 6351 | } |
b4ab44d8 | 6352 | #endif |
6ce1fbf1 | 6353 | |
c47b0ff6 AK |
6354 | static void update_recovery_start(struct intel_super *super, |
6355 | struct imsm_dev *dev, | |
6356 | struct mdinfo *array) | |
1e5c6983 DW |
6357 | { |
6358 | struct mdinfo *rebuild = NULL; | |
6359 | struct mdinfo *d; | |
6360 | __u32 units; | |
6361 | ||
6362 | if (!is_rebuilding(dev)) | |
6363 | return; | |
6364 | ||
6365 | /* Find the rebuild target, but punt on the dual rebuild case */ | |
6366 | for (d = array->devs; d; d = d->next) | |
6367 | if (d->recovery_start == 0) { | |
6368 | if (rebuild) | |
6369 | return; | |
6370 | rebuild = d; | |
6371 | } | |
6372 | ||
4363fd80 DW |
6373 | if (!rebuild) { |
6374 | /* (?) none of the disks are marked with | |
6375 | * IMSM_ORD_REBUILD, so assume they are missing and the | |
6376 | * disk_ord_tbl was not correctly updated | |
6377 | */ | |
6378 | dprintf("%s: failed to locate out-of-sync disk\n", __func__); | |
6379 | return; | |
6380 | } | |
6381 | ||
1e5c6983 | 6382 | units = __le32_to_cpu(dev->vol.curr_migr_unit); |
c47b0ff6 | 6383 | rebuild->recovery_start = units * blocks_per_migr_unit(super, dev); |
1e5c6983 DW |
6384 | } |
6385 | ||
9e2d750d | 6386 | #ifndef MDASSEMBLE |
276d77db | 6387 | static int recover_backup_imsm(struct supertype *st, struct mdinfo *info); |
9e2d750d | 6388 | #endif |
1e5c6983 | 6389 | |
00bbdbda | 6390 | static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray) |
cdddbdbc | 6391 | { |
4f5bc454 DW |
6392 | /* Given a container loaded by load_super_imsm_all, |
6393 | * extract information about all the arrays into | |
6394 | * an mdinfo tree. | |
00bbdbda | 6395 | * If 'subarray' is given, just extract info about that array. |
4f5bc454 DW |
6396 | * |
6397 | * For each imsm_dev create an mdinfo, fill it in, | |
6398 | * then look for matching devices in super->disks | |
6399 | * and create appropriate device mdinfo. | |
6400 | */ | |
6401 | struct intel_super *super = st->sb; | |
949c47a0 | 6402 | struct imsm_super *mpb = super->anchor; |
4f5bc454 | 6403 | struct mdinfo *rest = NULL; |
00bbdbda | 6404 | unsigned int i; |
81219e70 | 6405 | int sb_errors = 0; |
abef11a3 AK |
6406 | struct dl *d; |
6407 | int spare_disks = 0; | |
cdddbdbc | 6408 | |
19482bcc AK |
6409 | /* do not assemble arrays when not all attributes are supported */ |
6410 | if (imsm_check_attributes(mpb->attributes) == 0) { | |
81219e70 LM |
6411 | sb_errors = 1; |
6412 | fprintf(stderr, Name ": Unsupported attributes in IMSM metadata." | |
6413 | "Arrays activation is blocked.\n"); | |
19482bcc AK |
6414 | } |
6415 | ||
a06d022d | 6416 | /* check for bad blocks */ |
81219e70 LM |
6417 | if (imsm_bbm_log_size(super->anchor)) { |
6418 | fprintf(stderr, Name ": BBM log found in IMSM metadata." | |
6419 | "Arrays activation is blocked.\n"); | |
6420 | sb_errors = 1; | |
6421 | } | |
6422 | ||
604b746f | 6423 | |
abef11a3 AK |
6424 | /* count spare devices, not used in maps |
6425 | */ | |
6426 | for (d = super->disks; d; d = d->next) | |
6427 | if (d->index == -1) | |
6428 | spare_disks++; | |
6429 | ||
4f5bc454 | 6430 | for (i = 0; i < mpb->num_raid_devs; i++) { |
00bbdbda N |
6431 | struct imsm_dev *dev; |
6432 | struct imsm_map *map; | |
86e3692b | 6433 | struct imsm_map *map2; |
4f5bc454 | 6434 | struct mdinfo *this; |
a6482415 N |
6435 | int slot; |
6436 | #ifndef MDASSEMBLE | |
6437 | int chunk; | |
6438 | #endif | |
00bbdbda N |
6439 | char *ep; |
6440 | ||
6441 | if (subarray && | |
6442 | (i != strtoul(subarray, &ep, 10) || *ep != '\0')) | |
6443 | continue; | |
6444 | ||
6445 | dev = get_imsm_dev(super, i); | |
238c0a71 AK |
6446 | map = get_imsm_map(dev, MAP_0); |
6447 | map2 = get_imsm_map(dev, MAP_1); | |
4f5bc454 | 6448 | |
1ce0101c DW |
6449 | /* do not publish arrays that are in the middle of an |
6450 | * unsupported migration | |
6451 | */ | |
6452 | if (dev->vol.migr_state && | |
28bce06f | 6453 | (migr_type(dev) == MIGR_STATE_CHANGE)) { |
1ce0101c DW |
6454 | fprintf(stderr, Name ": cannot assemble volume '%.16s':" |
6455 | " unsupported migration in progress\n", | |
6456 | dev->volume); | |
6457 | continue; | |
6458 | } | |
2db86302 LM |
6459 | /* do not publish arrays that are not support by controller's |
6460 | * OROM/EFI | |
6461 | */ | |
1ce0101c | 6462 | |
4f5bc454 | 6463 | this = malloc(sizeof(*this)); |
0fbd635c | 6464 | if (!this) { |
cf1be220 | 6465 | fprintf(stderr, Name ": failed to allocate %zu bytes\n", |
0fbd635c AW |
6466 | sizeof(*this)); |
6467 | break; | |
6468 | } | |
4f5bc454 | 6469 | |
301406c9 | 6470 | super->current_vol = i; |
a5d85af7 | 6471 | getinfo_super_imsm_volume(st, this, NULL); |
9894ec0d | 6472 | this->next = rest; |
81219e70 | 6473 | #ifndef MDASSEMBLE |
a6482415 | 6474 | chunk = __le16_to_cpu(map->blocks_per_strip) >> 1; |
81219e70 LM |
6475 | /* mdadm does not support all metadata features- set the bit in all arrays state */ |
6476 | if (!validate_geometry_imsm_orom(super, | |
6477 | get_imsm_raid_level(map), /* RAID level */ | |
6478 | imsm_level_to_layout(get_imsm_raid_level(map)), | |
6479 | map->num_members, /* raid disks */ | |
6480 | &chunk, | |
6481 | 1 /* verbose */)) { | |
446894ea N |
6482 | fprintf(stderr, Name ": IMSM RAID geometry validation" |
6483 | " failed. Array %s activation is blocked.\n", | |
81219e70 LM |
6484 | dev->volume); |
6485 | this->array.state |= | |
6486 | (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) | | |
6487 | (1<<MD_SB_BLOCK_VOLUME); | |
6488 | } | |
6489 | #endif | |
6490 | ||
6491 | /* if array has bad blocks, set suitable bit in all arrays state */ | |
6492 | if (sb_errors) | |
6493 | this->array.state |= | |
6494 | (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) | | |
6495 | (1<<MD_SB_BLOCK_VOLUME); | |
6496 | ||
4f5bc454 | 6497 | for (slot = 0 ; slot < map->num_members; slot++) { |
1e5c6983 | 6498 | unsigned long long recovery_start; |
4f5bc454 DW |
6499 | struct mdinfo *info_d; |
6500 | struct dl *d; | |
6501 | int idx; | |
9a1608e5 | 6502 | int skip; |
7eef0453 | 6503 | __u32 ord; |
4f5bc454 | 6504 | |
9a1608e5 | 6505 | skip = 0; |
238c0a71 AK |
6506 | idx = get_imsm_disk_idx(dev, slot, MAP_0); |
6507 | ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); | |
4f5bc454 DW |
6508 | for (d = super->disks; d ; d = d->next) |
6509 | if (d->index == idx) | |
0fbd635c | 6510 | break; |
4f5bc454 | 6511 | |
1e5c6983 | 6512 | recovery_start = MaxSector; |
4f5bc454 | 6513 | if (d == NULL) |
9a1608e5 | 6514 | skip = 1; |
25ed7e59 | 6515 | if (d && is_failed(&d->disk)) |
9a1608e5 | 6516 | skip = 1; |
7eef0453 | 6517 | if (ord & IMSM_ORD_REBUILD) |
1e5c6983 | 6518 | recovery_start = 0; |
9a1608e5 DW |
6519 | |
6520 | /* | |
6521 | * if we skip some disks the array will be assmebled degraded; | |
1e5c6983 DW |
6522 | * reset resync start to avoid a dirty-degraded |
6523 | * situation when performing the intial sync | |
9a1608e5 DW |
6524 | * |
6525 | * FIXME handle dirty degraded | |
6526 | */ | |
1e5c6983 | 6527 | if ((skip || recovery_start == 0) && !dev->vol.dirty) |
b7528a20 | 6528 | this->resync_start = MaxSector; |
9a1608e5 DW |
6529 | if (skip) |
6530 | continue; | |
4f5bc454 | 6531 | |
1e5c6983 | 6532 | info_d = calloc(1, sizeof(*info_d)); |
9a1608e5 DW |
6533 | if (!info_d) { |
6534 | fprintf(stderr, Name ": failed to allocate disk" | |
1ce0101c | 6535 | " for volume %.16s\n", dev->volume); |
1e5c6983 DW |
6536 | info_d = this->devs; |
6537 | while (info_d) { | |
6538 | struct mdinfo *d = info_d->next; | |
6539 | ||
6540 | free(info_d); | |
6541 | info_d = d; | |
6542 | } | |
9a1608e5 DW |
6543 | free(this); |
6544 | this = rest; | |
6545 | break; | |
6546 | } | |
4f5bc454 DW |
6547 | info_d->next = this->devs; |
6548 | this->devs = info_d; | |
6549 | ||
4f5bc454 DW |
6550 | info_d->disk.number = d->index; |
6551 | info_d->disk.major = d->major; | |
6552 | info_d->disk.minor = d->minor; | |
6553 | info_d->disk.raid_disk = slot; | |
1e5c6983 | 6554 | info_d->recovery_start = recovery_start; |
86e3692b AK |
6555 | if (map2) { |
6556 | if (slot < map2->num_members) | |
6557 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
6558 | else |
6559 | this->array.spare_disks++; | |
86e3692b AK |
6560 | } else { |
6561 | if (slot < map->num_members) | |
6562 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
6563 | else |
6564 | this->array.spare_disks++; | |
86e3692b | 6565 | } |
1e5c6983 DW |
6566 | if (info_d->recovery_start == MaxSector) |
6567 | this->array.working_disks++; | |
4f5bc454 DW |
6568 | |
6569 | info_d->events = __le32_to_cpu(mpb->generation_num); | |
6570 | info_d->data_offset = __le32_to_cpu(map->pba_of_lba0); | |
6571 | info_d->component_size = __le32_to_cpu(map->blocks_per_member); | |
4f5bc454 | 6572 | } |
1e5c6983 | 6573 | /* now that the disk list is up-to-date fixup recovery_start */ |
c47b0ff6 | 6574 | update_recovery_start(super, dev, this); |
abef11a3 | 6575 | this->array.spare_disks += spare_disks; |
276d77db | 6576 | |
9e2d750d | 6577 | #ifndef MDASSEMBLE |
276d77db AK |
6578 | /* check for reshape */ |
6579 | if (this->reshape_active == 1) | |
6580 | recover_backup_imsm(st, this); | |
9e2d750d | 6581 | #endif |
9a1608e5 | 6582 | rest = this; |
4f5bc454 DW |
6583 | } |
6584 | ||
6585 | return rest; | |
cdddbdbc DW |
6586 | } |
6587 | ||
845dea95 | 6588 | |
3b451610 AK |
6589 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, |
6590 | int failed, int look_in_map) | |
c2a1e7da | 6591 | { |
3b451610 AK |
6592 | struct imsm_map *map; |
6593 | ||
6594 | map = get_imsm_map(dev, look_in_map); | |
c2a1e7da DW |
6595 | |
6596 | if (!failed) | |
3393c6af DW |
6597 | return map->map_state == IMSM_T_STATE_UNINITIALIZED ? |
6598 | IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL; | |
c2a1e7da DW |
6599 | |
6600 | switch (get_imsm_raid_level(map)) { | |
6601 | case 0: | |
6602 | return IMSM_T_STATE_FAILED; | |
6603 | break; | |
6604 | case 1: | |
6605 | if (failed < map->num_members) | |
6606 | return IMSM_T_STATE_DEGRADED; | |
6607 | else | |
6608 | return IMSM_T_STATE_FAILED; | |
6609 | break; | |
6610 | case 10: | |
6611 | { | |
6612 | /** | |
c92a2527 DW |
6613 | * check to see if any mirrors have failed, otherwise we |
6614 | * are degraded. Even numbered slots are mirrored on | |
6615 | * slot+1 | |
c2a1e7da | 6616 | */ |
c2a1e7da | 6617 | int i; |
d9b420a5 N |
6618 | /* gcc -Os complains that this is unused */ |
6619 | int insync = insync; | |
c2a1e7da DW |
6620 | |
6621 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 6622 | __u32 ord = get_imsm_ord_tbl_ent(dev, i, MAP_X); |
c92a2527 DW |
6623 | int idx = ord_to_idx(ord); |
6624 | struct imsm_disk *disk; | |
c2a1e7da | 6625 | |
c92a2527 DW |
6626 | /* reset the potential in-sync count on even-numbered |
6627 | * slots. num_copies is always 2 for imsm raid10 | |
6628 | */ | |
6629 | if ((i & 1) == 0) | |
6630 | insync = 2; | |
c2a1e7da | 6631 | |
c92a2527 | 6632 | disk = get_imsm_disk(super, idx); |
25ed7e59 | 6633 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) |
c92a2527 | 6634 | insync--; |
c2a1e7da | 6635 | |
c92a2527 DW |
6636 | /* no in-sync disks left in this mirror the |
6637 | * array has failed | |
6638 | */ | |
6639 | if (insync == 0) | |
6640 | return IMSM_T_STATE_FAILED; | |
c2a1e7da DW |
6641 | } |
6642 | ||
6643 | return IMSM_T_STATE_DEGRADED; | |
6644 | } | |
6645 | case 5: | |
6646 | if (failed < 2) | |
6647 | return IMSM_T_STATE_DEGRADED; | |
6648 | else | |
6649 | return IMSM_T_STATE_FAILED; | |
6650 | break; | |
6651 | default: | |
6652 | break; | |
6653 | } | |
6654 | ||
6655 | return map->map_state; | |
6656 | } | |
6657 | ||
3b451610 AK |
6658 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev, |
6659 | int look_in_map) | |
c2a1e7da DW |
6660 | { |
6661 | int i; | |
6662 | int failed = 0; | |
6663 | struct imsm_disk *disk; | |
d5985138 AK |
6664 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
6665 | struct imsm_map *prev = get_imsm_map(dev, MAP_1); | |
68fe4598 | 6666 | struct imsm_map *map_for_loop; |
0556e1a2 DW |
6667 | __u32 ord; |
6668 | int idx; | |
d5985138 | 6669 | int idx_1; |
c2a1e7da | 6670 | |
0556e1a2 DW |
6671 | /* at the beginning of migration we set IMSM_ORD_REBUILD on |
6672 | * disks that are being rebuilt. New failures are recorded to | |
6673 | * map[0]. So we look through all the disks we started with and | |
6674 | * see if any failures are still present, or if any new ones | |
6675 | * have arrived | |
0556e1a2 | 6676 | */ |
d5985138 AK |
6677 | map_for_loop = map; |
6678 | if (prev && (map->num_members < prev->num_members)) | |
6679 | map_for_loop = prev; | |
68fe4598 LD |
6680 | |
6681 | for (i = 0; i < map_for_loop->num_members; i++) { | |
d5985138 | 6682 | idx_1 = -255; |
238c0a71 AK |
6683 | /* when MAP_X is passed both maps failures are counted |
6684 | */ | |
d5985138 | 6685 | if (prev && |
238c0a71 AK |
6686 | ((look_in_map == MAP_1) || (look_in_map == MAP_X)) && |
6687 | (i < prev->num_members)) { | |
d5985138 AK |
6688 | ord = __le32_to_cpu(prev->disk_ord_tbl[i]); |
6689 | idx_1 = ord_to_idx(ord); | |
c2a1e7da | 6690 | |
d5985138 AK |
6691 | disk = get_imsm_disk(super, idx_1); |
6692 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) | |
6693 | failed++; | |
6694 | } | |
238c0a71 AK |
6695 | if (((look_in_map == MAP_0) || (look_in_map == MAP_X)) && |
6696 | (i < map->num_members)) { | |
d5985138 AK |
6697 | ord = __le32_to_cpu(map->disk_ord_tbl[i]); |
6698 | idx = ord_to_idx(ord); | |
6699 | ||
6700 | if (idx != idx_1) { | |
6701 | disk = get_imsm_disk(super, idx); | |
6702 | if (!disk || is_failed(disk) || | |
6703 | ord & IMSM_ORD_REBUILD) | |
6704 | failed++; | |
6705 | } | |
6706 | } | |
c2a1e7da DW |
6707 | } |
6708 | ||
6709 | return failed; | |
845dea95 NB |
6710 | } |
6711 | ||
97b4d0e9 DW |
6712 | #ifndef MDASSEMBLE |
6713 | static int imsm_open_new(struct supertype *c, struct active_array *a, | |
6714 | char *inst) | |
6715 | { | |
6716 | struct intel_super *super = c->sb; | |
6717 | struct imsm_super *mpb = super->anchor; | |
9587c373 | 6718 | |
97b4d0e9 DW |
6719 | if (atoi(inst) >= mpb->num_raid_devs) { |
6720 | fprintf(stderr, "%s: subarry index %d, out of range\n", | |
6721 | __func__, atoi(inst)); | |
6722 | return -ENODEV; | |
6723 | } | |
6724 | ||
6725 | dprintf("imsm: open_new %s\n", inst); | |
6726 | a->info.container_member = atoi(inst); | |
6727 | return 0; | |
6728 | } | |
6729 | ||
0c046afd DW |
6730 | static int is_resyncing(struct imsm_dev *dev) |
6731 | { | |
6732 | struct imsm_map *migr_map; | |
6733 | ||
6734 | if (!dev->vol.migr_state) | |
6735 | return 0; | |
6736 | ||
1484e727 DW |
6737 | if (migr_type(dev) == MIGR_INIT || |
6738 | migr_type(dev) == MIGR_REPAIR) | |
0c046afd DW |
6739 | return 1; |
6740 | ||
4c9bc37b AK |
6741 | if (migr_type(dev) == MIGR_GEN_MIGR) |
6742 | return 0; | |
6743 | ||
238c0a71 | 6744 | migr_map = get_imsm_map(dev, MAP_1); |
0c046afd | 6745 | |
4c9bc37b AK |
6746 | if ((migr_map->map_state == IMSM_T_STATE_NORMAL) && |
6747 | (dev->vol.migr_type != MIGR_GEN_MIGR)) | |
0c046afd DW |
6748 | return 1; |
6749 | else | |
6750 | return 0; | |
6751 | } | |
6752 | ||
0556e1a2 DW |
6753 | /* return true if we recorded new information */ |
6754 | static int mark_failure(struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
47ee5a45 | 6755 | { |
0556e1a2 DW |
6756 | __u32 ord; |
6757 | int slot; | |
6758 | struct imsm_map *map; | |
86c54047 DW |
6759 | char buf[MAX_RAID_SERIAL_LEN+3]; |
6760 | unsigned int len, shift = 0; | |
0556e1a2 DW |
6761 | |
6762 | /* new failures are always set in map[0] */ | |
238c0a71 | 6763 | map = get_imsm_map(dev, MAP_0); |
0556e1a2 DW |
6764 | |
6765 | slot = get_imsm_disk_slot(map, idx); | |
6766 | if (slot < 0) | |
6767 | return 0; | |
6768 | ||
6769 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
25ed7e59 | 6770 | if (is_failed(disk) && (ord & IMSM_ORD_REBUILD)) |
0556e1a2 DW |
6771 | return 0; |
6772 | ||
7d0c5e24 LD |
6773 | memcpy(buf, disk->serial, MAX_RAID_SERIAL_LEN); |
6774 | buf[MAX_RAID_SERIAL_LEN] = '\000'; | |
6775 | strcat(buf, ":0"); | |
86c54047 DW |
6776 | if ((len = strlen(buf)) >= MAX_RAID_SERIAL_LEN) |
6777 | shift = len - MAX_RAID_SERIAL_LEN + 1; | |
6778 | strncpy((char *)disk->serial, &buf[shift], MAX_RAID_SERIAL_LEN); | |
6779 | ||
f2f27e63 | 6780 | disk->status |= FAILED_DISK; |
0556e1a2 | 6781 | set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD); |
17788994 AK |
6782 | /* mark failures in second map if second map exists and this disk |
6783 | * in this slot. | |
6784 | * This is valid for migration, initialization and rebuild | |
6785 | */ | |
6786 | if (dev->vol.migr_state) { | |
238c0a71 | 6787 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); |
0a108d63 AK |
6788 | int slot2 = get_imsm_disk_slot(map2, idx); |
6789 | ||
6790 | if ((slot2 < map2->num_members) && | |
6791 | (slot2 >= 0)) | |
6792 | set_imsm_ord_tbl_ent(map2, slot2, | |
1ace8403 AK |
6793 | idx | IMSM_ORD_REBUILD); |
6794 | } | |
f21e18ca | 6795 | if (map->failed_disk_num == 0xff) |
0556e1a2 DW |
6796 | map->failed_disk_num = slot; |
6797 | return 1; | |
6798 | } | |
6799 | ||
6800 | static void mark_missing(struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
6801 | { | |
6802 | mark_failure(dev, disk, idx); | |
6803 | ||
6804 | if (disk->scsi_id == __cpu_to_le32(~(__u32)0)) | |
6805 | return; | |
6806 | ||
47ee5a45 DW |
6807 | disk->scsi_id = __cpu_to_le32(~(__u32)0); |
6808 | memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1); | |
6809 | } | |
6810 | ||
33414a01 DW |
6811 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev) |
6812 | { | |
33414a01 | 6813 | struct dl *dl; |
33414a01 DW |
6814 | |
6815 | if (!super->missing) | |
6816 | return; | |
33414a01 DW |
6817 | |
6818 | dprintf("imsm: mark missing\n"); | |
3d59f0c0 AK |
6819 | /* end process for initialization and rebuild only |
6820 | */ | |
6821 | if (is_gen_migration(dev) == 0) { | |
6822 | __u8 map_state; | |
6823 | int failed; | |
6824 | ||
6825 | failed = imsm_count_failed(super, dev, MAP_0); | |
6826 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
6827 | ||
6828 | end_migration(dev, super, map_state); | |
6829 | } | |
33414a01 DW |
6830 | for (dl = super->missing; dl; dl = dl->next) |
6831 | mark_missing(dev, &dl->disk, dl->index); | |
6832 | super->updates_pending++; | |
6833 | } | |
6834 | ||
70bdf0dc AK |
6835 | static unsigned long long imsm_set_array_size(struct imsm_dev *dev) |
6836 | { | |
238c0a71 | 6837 | int used_disks = imsm_num_data_members(dev, MAP_0); |
70bdf0dc AK |
6838 | unsigned long long array_blocks; |
6839 | struct imsm_map *map; | |
6840 | ||
6841 | if (used_disks == 0) { | |
6842 | /* when problems occures | |
6843 | * return current array_blocks value | |
6844 | */ | |
6845 | array_blocks = __le32_to_cpu(dev->size_high); | |
6846 | array_blocks = array_blocks << 32; | |
6847 | array_blocks += __le32_to_cpu(dev->size_low); | |
6848 | ||
6849 | return array_blocks; | |
6850 | } | |
6851 | ||
6852 | /* set array size in metadata | |
6853 | */ | |
238c0a71 | 6854 | map = get_imsm_map(dev, MAP_0); |
70bdf0dc AK |
6855 | array_blocks = map->blocks_per_member * used_disks; |
6856 | ||
6857 | /* round array size down to closest MB | |
6858 | */ | |
6859 | array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT; | |
6860 | dev->size_low = __cpu_to_le32((__u32)array_blocks); | |
6861 | dev->size_high = __cpu_to_le32((__u32)(array_blocks >> 32)); | |
6862 | ||
6863 | return array_blocks; | |
6864 | } | |
6865 | ||
28bce06f AK |
6866 | static void imsm_set_disk(struct active_array *a, int n, int state); |
6867 | ||
0e2d1a4e AK |
6868 | static void imsm_progress_container_reshape(struct intel_super *super) |
6869 | { | |
6870 | /* if no device has a migr_state, but some device has a | |
6871 | * different number of members than the previous device, start | |
6872 | * changing the number of devices in this device to match | |
6873 | * previous. | |
6874 | */ | |
6875 | struct imsm_super *mpb = super->anchor; | |
6876 | int prev_disks = -1; | |
6877 | int i; | |
1dfaa380 | 6878 | int copy_map_size; |
0e2d1a4e AK |
6879 | |
6880 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
6881 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 6882 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
0e2d1a4e AK |
6883 | struct imsm_map *map2; |
6884 | int prev_num_members; | |
0e2d1a4e AK |
6885 | |
6886 | if (dev->vol.migr_state) | |
6887 | return; | |
6888 | ||
6889 | if (prev_disks == -1) | |
6890 | prev_disks = map->num_members; | |
6891 | if (prev_disks == map->num_members) | |
6892 | continue; | |
6893 | ||
6894 | /* OK, this array needs to enter reshape mode. | |
6895 | * i.e it needs a migr_state | |
6896 | */ | |
6897 | ||
1dfaa380 | 6898 | copy_map_size = sizeof_imsm_map(map); |
0e2d1a4e AK |
6899 | prev_num_members = map->num_members; |
6900 | map->num_members = prev_disks; | |
6901 | dev->vol.migr_state = 1; | |
6902 | dev->vol.curr_migr_unit = 0; | |
ea672ee1 | 6903 | set_migr_type(dev, MIGR_GEN_MIGR); |
0e2d1a4e AK |
6904 | for (i = prev_num_members; |
6905 | i < map->num_members; i++) | |
6906 | set_imsm_ord_tbl_ent(map, i, i); | |
238c0a71 | 6907 | map2 = get_imsm_map(dev, MAP_1); |
0e2d1a4e | 6908 | /* Copy the current map */ |
1dfaa380 | 6909 | memcpy(map2, map, copy_map_size); |
0e2d1a4e AK |
6910 | map2->num_members = prev_num_members; |
6911 | ||
70bdf0dc | 6912 | imsm_set_array_size(dev); |
0e2d1a4e AK |
6913 | super->updates_pending++; |
6914 | } | |
6915 | } | |
6916 | ||
aad6f216 | 6917 | /* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild |
0c046afd DW |
6918 | * states are handled in imsm_set_disk() with one exception, when a |
6919 | * resync is stopped due to a new failure this routine will set the | |
6920 | * 'degraded' state for the array. | |
6921 | */ | |
01f157d7 | 6922 | static int imsm_set_array_state(struct active_array *a, int consistent) |
a862209d DW |
6923 | { |
6924 | int inst = a->info.container_member; | |
6925 | struct intel_super *super = a->container->sb; | |
949c47a0 | 6926 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 6927 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
3b451610 AK |
6928 | int failed = imsm_count_failed(super, dev, MAP_0); |
6929 | __u8 map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
1e5c6983 | 6930 | __u32 blocks_per_unit; |
a862209d | 6931 | |
1af97990 AK |
6932 | if (dev->vol.migr_state && |
6933 | dev->vol.migr_type == MIGR_GEN_MIGR) { | |
6934 | /* array state change is blocked due to reshape action | |
aad6f216 N |
6935 | * We might need to |
6936 | * - abort the reshape (if last_checkpoint is 0 and action!= reshape) | |
6937 | * - finish the reshape (if last_checkpoint is big and action != reshape) | |
6938 | * - update curr_migr_unit | |
1af97990 | 6939 | */ |
aad6f216 N |
6940 | if (a->curr_action == reshape) { |
6941 | /* still reshaping, maybe update curr_migr_unit */ | |
633b5610 | 6942 | goto mark_checkpoint; |
aad6f216 N |
6943 | } else { |
6944 | if (a->last_checkpoint == 0 && a->prev_action == reshape) { | |
6945 | /* for some reason we aborted the reshape. | |
b66e591b AK |
6946 | * |
6947 | * disable automatic metadata rollback | |
6948 | * user action is required to recover process | |
aad6f216 | 6949 | */ |
b66e591b | 6950 | if (0) { |
238c0a71 AK |
6951 | struct imsm_map *map2 = |
6952 | get_imsm_map(dev, MAP_1); | |
6953 | dev->vol.migr_state = 0; | |
6954 | set_migr_type(dev, 0); | |
6955 | dev->vol.curr_migr_unit = 0; | |
6956 | memcpy(map, map2, | |
6957 | sizeof_imsm_map(map2)); | |
6958 | super->updates_pending++; | |
b66e591b | 6959 | } |
aad6f216 N |
6960 | } |
6961 | if (a->last_checkpoint >= a->info.component_size) { | |
6962 | unsigned long long array_blocks; | |
6963 | int used_disks; | |
e154ced3 | 6964 | struct mdinfo *mdi; |
aad6f216 | 6965 | |
238c0a71 | 6966 | used_disks = imsm_num_data_members(dev, MAP_0); |
d55adef9 AK |
6967 | if (used_disks > 0) { |
6968 | array_blocks = | |
6969 | map->blocks_per_member * | |
6970 | used_disks; | |
6971 | /* round array size down to closest MB | |
6972 | */ | |
6973 | array_blocks = (array_blocks | |
6974 | >> SECT_PER_MB_SHIFT) | |
6975 | << SECT_PER_MB_SHIFT; | |
d55adef9 AK |
6976 | a->info.custom_array_size = array_blocks; |
6977 | /* encourage manager to update array | |
6978 | * size | |
6979 | */ | |
e154ced3 | 6980 | |
d55adef9 | 6981 | a->check_reshape = 1; |
633b5610 | 6982 | } |
e154ced3 AK |
6983 | /* finalize online capacity expansion/reshape */ |
6984 | for (mdi = a->info.devs; mdi; mdi = mdi->next) | |
6985 | imsm_set_disk(a, | |
6986 | mdi->disk.raid_disk, | |
6987 | mdi->curr_state); | |
6988 | ||
0e2d1a4e | 6989 | imsm_progress_container_reshape(super); |
e154ced3 | 6990 | } |
aad6f216 | 6991 | } |
1af97990 AK |
6992 | } |
6993 | ||
47ee5a45 | 6994 | /* before we activate this array handle any missing disks */ |
33414a01 DW |
6995 | if (consistent == 2) |
6996 | handle_missing(super, dev); | |
1e5c6983 | 6997 | |
0c046afd | 6998 | if (consistent == 2 && |
b7941fd6 | 6999 | (!is_resync_complete(&a->info) || |
0c046afd DW |
7000 | map_state != IMSM_T_STATE_NORMAL || |
7001 | dev->vol.migr_state)) | |
01f157d7 | 7002 | consistent = 0; |
272906ef | 7003 | |
b7941fd6 | 7004 | if (is_resync_complete(&a->info)) { |
0c046afd | 7005 | /* complete intialization / resync, |
0556e1a2 DW |
7006 | * recovery and interrupted recovery is completed in |
7007 | * ->set_disk | |
0c046afd DW |
7008 | */ |
7009 | if (is_resyncing(dev)) { | |
7010 | dprintf("imsm: mark resync done\n"); | |
809da78e | 7011 | end_migration(dev, super, map_state); |
115c3803 | 7012 | super->updates_pending++; |
484240d8 | 7013 | a->last_checkpoint = 0; |
115c3803 | 7014 | } |
b9172665 AK |
7015 | } else if ((!is_resyncing(dev) && !failed) && |
7016 | (imsm_reshape_blocks_arrays_changes(super) == 0)) { | |
0c046afd | 7017 | /* mark the start of the init process if nothing is failed */ |
b7941fd6 | 7018 | dprintf("imsm: mark resync start\n"); |
1484e727 | 7019 | if (map->map_state == IMSM_T_STATE_UNINITIALIZED) |
8e59f3d8 | 7020 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_INIT); |
1484e727 | 7021 | else |
8e59f3d8 | 7022 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_REPAIR); |
3393c6af | 7023 | super->updates_pending++; |
115c3803 | 7024 | } |
a862209d | 7025 | |
633b5610 | 7026 | mark_checkpoint: |
5b83bacf AK |
7027 | /* skip checkpointing for general migration, |
7028 | * it is controlled in mdadm | |
7029 | */ | |
7030 | if (is_gen_migration(dev)) | |
7031 | goto skip_mark_checkpoint; | |
7032 | ||
1e5c6983 | 7033 | /* check if we can update curr_migr_unit from resync_start, recovery_start */ |
c47b0ff6 | 7034 | blocks_per_unit = blocks_per_migr_unit(super, dev); |
4f0a7acc | 7035 | if (blocks_per_unit) { |
1e5c6983 DW |
7036 | __u32 units32; |
7037 | __u64 units; | |
7038 | ||
4f0a7acc | 7039 | units = a->last_checkpoint / blocks_per_unit; |
1e5c6983 DW |
7040 | units32 = units; |
7041 | ||
7042 | /* check that we did not overflow 32-bits, and that | |
7043 | * curr_migr_unit needs updating | |
7044 | */ | |
7045 | if (units32 == units && | |
bfd80a56 | 7046 | units32 != 0 && |
1e5c6983 DW |
7047 | __le32_to_cpu(dev->vol.curr_migr_unit) != units32) { |
7048 | dprintf("imsm: mark checkpoint (%u)\n", units32); | |
7049 | dev->vol.curr_migr_unit = __cpu_to_le32(units32); | |
7050 | super->updates_pending++; | |
7051 | } | |
7052 | } | |
f8f603f1 | 7053 | |
5b83bacf | 7054 | skip_mark_checkpoint: |
3393c6af | 7055 | /* mark dirty / clean */ |
0c046afd | 7056 | if (dev->vol.dirty != !consistent) { |
b7941fd6 | 7057 | dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty"); |
0c046afd DW |
7058 | if (consistent) |
7059 | dev->vol.dirty = 0; | |
7060 | else | |
7061 | dev->vol.dirty = 1; | |
a862209d DW |
7062 | super->updates_pending++; |
7063 | } | |
28bce06f | 7064 | |
01f157d7 | 7065 | return consistent; |
a862209d DW |
7066 | } |
7067 | ||
8d45d196 | 7068 | static void imsm_set_disk(struct active_array *a, int n, int state) |
845dea95 | 7069 | { |
8d45d196 DW |
7070 | int inst = a->info.container_member; |
7071 | struct intel_super *super = a->container->sb; | |
949c47a0 | 7072 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 7073 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
8d45d196 | 7074 | struct imsm_disk *disk; |
0c046afd | 7075 | int failed; |
b10b37b8 | 7076 | __u32 ord; |
0c046afd | 7077 | __u8 map_state; |
8d45d196 DW |
7078 | |
7079 | if (n > map->num_members) | |
7080 | fprintf(stderr, "imsm: set_disk %d out of range 0..%d\n", | |
7081 | n, map->num_members - 1); | |
7082 | ||
7083 | if (n < 0) | |
7084 | return; | |
7085 | ||
4e6e574a | 7086 | dprintf("imsm: set_disk %d:%x\n", n, state); |
8d45d196 | 7087 | |
238c0a71 | 7088 | ord = get_imsm_ord_tbl_ent(dev, n, MAP_0); |
b10b37b8 | 7089 | disk = get_imsm_disk(super, ord_to_idx(ord)); |
8d45d196 | 7090 | |
5802a811 | 7091 | /* check for new failures */ |
0556e1a2 DW |
7092 | if (state & DS_FAULTY) { |
7093 | if (mark_failure(dev, disk, ord_to_idx(ord))) | |
7094 | super->updates_pending++; | |
8d45d196 | 7095 | } |
47ee5a45 | 7096 | |
19859edc | 7097 | /* check if in_sync */ |
0556e1a2 | 7098 | if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) { |
238c0a71 | 7099 | struct imsm_map *migr_map = get_imsm_map(dev, MAP_1); |
b10b37b8 DW |
7100 | |
7101 | set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord)); | |
19859edc DW |
7102 | super->updates_pending++; |
7103 | } | |
8d45d196 | 7104 | |
3b451610 AK |
7105 | failed = imsm_count_failed(super, dev, MAP_0); |
7106 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
5802a811 | 7107 | |
0c046afd | 7108 | /* check if recovery complete, newly degraded, or failed */ |
94002678 AK |
7109 | dprintf("imsm: Detected transition to state "); |
7110 | switch (map_state) { | |
7111 | case IMSM_T_STATE_NORMAL: /* transition to normal state */ | |
7112 | dprintf("normal: "); | |
7113 | if (is_rebuilding(dev)) { | |
7114 | dprintf("while rebuilding"); | |
7115 | end_migration(dev, super, map_state); | |
238c0a71 | 7116 | map = get_imsm_map(dev, MAP_0); |
94002678 AK |
7117 | map->failed_disk_num = ~0; |
7118 | super->updates_pending++; | |
7119 | a->last_checkpoint = 0; | |
7120 | break; | |
7121 | } | |
7122 | if (is_gen_migration(dev)) { | |
7123 | dprintf("while general migration"); | |
bf2f0071 | 7124 | if (a->last_checkpoint >= a->info.component_size) |
809da78e | 7125 | end_migration(dev, super, map_state); |
94002678 AK |
7126 | else |
7127 | map->map_state = map_state; | |
238c0a71 | 7128 | map = get_imsm_map(dev, MAP_0); |
28bce06f | 7129 | map->failed_disk_num = ~0; |
94002678 | 7130 | super->updates_pending++; |
bf2f0071 | 7131 | break; |
94002678 AK |
7132 | } |
7133 | break; | |
7134 | case IMSM_T_STATE_DEGRADED: /* transition to degraded state */ | |
7135 | dprintf("degraded: "); | |
7136 | if ((map->map_state != map_state) && | |
7137 | !dev->vol.migr_state) { | |
7138 | dprintf("mark degraded"); | |
7139 | map->map_state = map_state; | |
7140 | super->updates_pending++; | |
7141 | a->last_checkpoint = 0; | |
7142 | break; | |
7143 | } | |
7144 | if (is_rebuilding(dev)) { | |
7145 | dprintf("while rebuilding."); | |
7146 | if (map->map_state != map_state) { | |
7147 | dprintf(" Map state change"); | |
7148 | end_migration(dev, super, map_state); | |
7149 | super->updates_pending++; | |
7150 | } | |
7151 | break; | |
7152 | } | |
7153 | if (is_gen_migration(dev)) { | |
7154 | dprintf("while general migration"); | |
bf2f0071 | 7155 | if (a->last_checkpoint >= a->info.component_size) |
809da78e | 7156 | end_migration(dev, super, map_state); |
94002678 AK |
7157 | else { |
7158 | map->map_state = map_state; | |
3b451610 | 7159 | manage_second_map(super, dev); |
94002678 AK |
7160 | } |
7161 | super->updates_pending++; | |
bf2f0071 | 7162 | break; |
28bce06f | 7163 | } |
6ce1fbf1 AK |
7164 | if (is_initializing(dev)) { |
7165 | dprintf("while initialization."); | |
7166 | map->map_state = map_state; | |
7167 | super->updates_pending++; | |
7168 | break; | |
7169 | } | |
94002678 AK |
7170 | break; |
7171 | case IMSM_T_STATE_FAILED: /* transition to failed state */ | |
7172 | dprintf("failed: "); | |
7173 | if (is_gen_migration(dev)) { | |
7174 | dprintf("while general migration"); | |
7175 | map->map_state = map_state; | |
7176 | super->updates_pending++; | |
7177 | break; | |
7178 | } | |
7179 | if (map->map_state != map_state) { | |
7180 | dprintf("mark failed"); | |
7181 | end_migration(dev, super, map_state); | |
7182 | super->updates_pending++; | |
7183 | a->last_checkpoint = 0; | |
7184 | break; | |
7185 | } | |
7186 | break; | |
7187 | default: | |
7188 | dprintf("state %i\n", map_state); | |
5802a811 | 7189 | } |
94002678 AK |
7190 | dprintf("\n"); |
7191 | ||
845dea95 NB |
7192 | } |
7193 | ||
f796af5d | 7194 | static int store_imsm_mpb(int fd, struct imsm_super *mpb) |
c2a1e7da | 7195 | { |
f796af5d | 7196 | void *buf = mpb; |
c2a1e7da DW |
7197 | __u32 mpb_size = __le32_to_cpu(mpb->mpb_size); |
7198 | unsigned long long dsize; | |
7199 | unsigned long long sectors; | |
7200 | ||
7201 | get_dev_size(fd, NULL, &dsize); | |
7202 | ||
272f648f DW |
7203 | if (mpb_size > 512) { |
7204 | /* -1 to account for anchor */ | |
7205 | sectors = mpb_sectors(mpb) - 1; | |
c2a1e7da | 7206 | |
272f648f DW |
7207 | /* write the extended mpb to the sectors preceeding the anchor */ |
7208 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) | |
7209 | return 1; | |
c2a1e7da | 7210 | |
f21e18ca N |
7211 | if ((unsigned long long)write(fd, buf + 512, 512 * sectors) |
7212 | != 512 * sectors) | |
272f648f DW |
7213 | return 1; |
7214 | } | |
c2a1e7da | 7215 | |
272f648f DW |
7216 | /* first block is stored on second to last sector of the disk */ |
7217 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) | |
c2a1e7da DW |
7218 | return 1; |
7219 | ||
f796af5d | 7220 | if (write(fd, buf, 512) != 512) |
c2a1e7da DW |
7221 | return 1; |
7222 | ||
c2a1e7da DW |
7223 | return 0; |
7224 | } | |
7225 | ||
2e735d19 | 7226 | static void imsm_sync_metadata(struct supertype *container) |
845dea95 | 7227 | { |
2e735d19 | 7228 | struct intel_super *super = container->sb; |
c2a1e7da | 7229 | |
1a64be56 | 7230 | dprintf("sync metadata: %d\n", super->updates_pending); |
c2a1e7da DW |
7231 | if (!super->updates_pending) |
7232 | return; | |
7233 | ||
36988a3d | 7234 | write_super_imsm(container, 0); |
c2a1e7da DW |
7235 | |
7236 | super->updates_pending = 0; | |
845dea95 NB |
7237 | } |
7238 | ||
272906ef DW |
7239 | static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a) |
7240 | { | |
7241 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 7242 | int i = get_imsm_disk_idx(dev, idx, MAP_X); |
272906ef DW |
7243 | struct dl *dl; |
7244 | ||
7245 | for (dl = super->disks; dl; dl = dl->next) | |
7246 | if (dl->index == i) | |
7247 | break; | |
7248 | ||
25ed7e59 | 7249 | if (dl && is_failed(&dl->disk)) |
272906ef DW |
7250 | dl = NULL; |
7251 | ||
7252 | if (dl) | |
7253 | dprintf("%s: found %x:%x\n", __func__, dl->major, dl->minor); | |
7254 | ||
7255 | return dl; | |
7256 | } | |
7257 | ||
a20d2ba5 | 7258 | static struct dl *imsm_add_spare(struct intel_super *super, int slot, |
8ba77d32 AK |
7259 | struct active_array *a, int activate_new, |
7260 | struct mdinfo *additional_test_list) | |
272906ef DW |
7261 | { |
7262 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 7263 | int idx = get_imsm_disk_idx(dev, slot, MAP_X); |
a20d2ba5 DW |
7264 | struct imsm_super *mpb = super->anchor; |
7265 | struct imsm_map *map; | |
272906ef DW |
7266 | unsigned long long pos; |
7267 | struct mdinfo *d; | |
7268 | struct extent *ex; | |
a20d2ba5 | 7269 | int i, j; |
272906ef | 7270 | int found; |
569cc43f DW |
7271 | __u32 array_start = 0; |
7272 | __u32 array_end = 0; | |
272906ef | 7273 | struct dl *dl; |
6c932028 | 7274 | struct mdinfo *test_list; |
272906ef DW |
7275 | |
7276 | for (dl = super->disks; dl; dl = dl->next) { | |
7277 | /* If in this array, skip */ | |
7278 | for (d = a->info.devs ; d ; d = d->next) | |
e553d2a4 DW |
7279 | if (d->state_fd >= 0 && |
7280 | d->disk.major == dl->major && | |
272906ef | 7281 | d->disk.minor == dl->minor) { |
8ba77d32 AK |
7282 | dprintf("%x:%x already in array\n", |
7283 | dl->major, dl->minor); | |
272906ef DW |
7284 | break; |
7285 | } | |
7286 | if (d) | |
7287 | continue; | |
6c932028 AK |
7288 | test_list = additional_test_list; |
7289 | while (test_list) { | |
7290 | if (test_list->disk.major == dl->major && | |
7291 | test_list->disk.minor == dl->minor) { | |
8ba77d32 AK |
7292 | dprintf("%x:%x already in additional test list\n", |
7293 | dl->major, dl->minor); | |
7294 | break; | |
7295 | } | |
6c932028 | 7296 | test_list = test_list->next; |
8ba77d32 | 7297 | } |
6c932028 | 7298 | if (test_list) |
8ba77d32 | 7299 | continue; |
272906ef | 7300 | |
e553d2a4 | 7301 | /* skip in use or failed drives */ |
25ed7e59 | 7302 | if (is_failed(&dl->disk) || idx == dl->index || |
df474657 DW |
7303 | dl->index == -2) { |
7304 | dprintf("%x:%x status (failed: %d index: %d)\n", | |
25ed7e59 | 7305 | dl->major, dl->minor, is_failed(&dl->disk), idx); |
9a1608e5 DW |
7306 | continue; |
7307 | } | |
7308 | ||
a20d2ba5 DW |
7309 | /* skip pure spares when we are looking for partially |
7310 | * assimilated drives | |
7311 | */ | |
7312 | if (dl->index == -1 && !activate_new) | |
7313 | continue; | |
7314 | ||
272906ef | 7315 | /* Does this unused device have the requisite free space? |
a20d2ba5 | 7316 | * It needs to be able to cover all member volumes |
272906ef DW |
7317 | */ |
7318 | ex = get_extents(super, dl); | |
7319 | if (!ex) { | |
7320 | dprintf("cannot get extents\n"); | |
7321 | continue; | |
7322 | } | |
a20d2ba5 DW |
7323 | for (i = 0; i < mpb->num_raid_devs; i++) { |
7324 | dev = get_imsm_dev(super, i); | |
238c0a71 | 7325 | map = get_imsm_map(dev, MAP_0); |
272906ef | 7326 | |
a20d2ba5 DW |
7327 | /* check if this disk is already a member of |
7328 | * this array | |
272906ef | 7329 | */ |
620b1713 | 7330 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
a20d2ba5 DW |
7331 | continue; |
7332 | ||
7333 | found = 0; | |
7334 | j = 0; | |
7335 | pos = 0; | |
7336 | array_start = __le32_to_cpu(map->pba_of_lba0); | |
329c8278 DW |
7337 | array_end = array_start + |
7338 | __le32_to_cpu(map->blocks_per_member) - 1; | |
a20d2ba5 DW |
7339 | |
7340 | do { | |
7341 | /* check that we can start at pba_of_lba0 with | |
7342 | * blocks_per_member of space | |
7343 | */ | |
329c8278 | 7344 | if (array_start >= pos && array_end < ex[j].start) { |
a20d2ba5 DW |
7345 | found = 1; |
7346 | break; | |
7347 | } | |
7348 | pos = ex[j].start + ex[j].size; | |
7349 | j++; | |
7350 | } while (ex[j-1].size); | |
7351 | ||
7352 | if (!found) | |
272906ef | 7353 | break; |
a20d2ba5 | 7354 | } |
272906ef DW |
7355 | |
7356 | free(ex); | |
a20d2ba5 | 7357 | if (i < mpb->num_raid_devs) { |
329c8278 DW |
7358 | dprintf("%x:%x does not have %u to %u available\n", |
7359 | dl->major, dl->minor, array_start, array_end); | |
272906ef DW |
7360 | /* No room */ |
7361 | continue; | |
a20d2ba5 DW |
7362 | } |
7363 | return dl; | |
272906ef DW |
7364 | } |
7365 | ||
7366 | return dl; | |
7367 | } | |
7368 | ||
95d07a2c LM |
7369 | |
7370 | static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed) | |
7371 | { | |
7372 | struct imsm_dev *dev2; | |
7373 | struct imsm_map *map; | |
7374 | struct dl *idisk; | |
7375 | int slot; | |
7376 | int idx; | |
7377 | __u8 state; | |
7378 | ||
7379 | dev2 = get_imsm_dev(cont->sb, dev_idx); | |
7380 | if (dev2) { | |
238c0a71 | 7381 | state = imsm_check_degraded(cont->sb, dev2, failed, MAP_0); |
95d07a2c | 7382 | if (state == IMSM_T_STATE_FAILED) { |
238c0a71 | 7383 | map = get_imsm_map(dev2, MAP_0); |
95d07a2c LM |
7384 | if (!map) |
7385 | return 1; | |
7386 | for (slot = 0; slot < map->num_members; slot++) { | |
7387 | /* | |
7388 | * Check if failed disks are deleted from intel | |
7389 | * disk list or are marked to be deleted | |
7390 | */ | |
238c0a71 | 7391 | idx = get_imsm_disk_idx(dev2, slot, MAP_X); |
95d07a2c LM |
7392 | idisk = get_imsm_dl_disk(cont->sb, idx); |
7393 | /* | |
7394 | * Do not rebuild the array if failed disks | |
7395 | * from failed sub-array are not removed from | |
7396 | * container. | |
7397 | */ | |
7398 | if (idisk && | |
7399 | is_failed(&idisk->disk) && | |
7400 | (idisk->action != DISK_REMOVE)) | |
7401 | return 0; | |
7402 | } | |
7403 | } | |
7404 | } | |
7405 | return 1; | |
7406 | } | |
7407 | ||
88758e9d DW |
7408 | static struct mdinfo *imsm_activate_spare(struct active_array *a, |
7409 | struct metadata_update **updates) | |
7410 | { | |
7411 | /** | |
d23fe947 DW |
7412 | * Find a device with unused free space and use it to replace a |
7413 | * failed/vacant region in an array. We replace failed regions one a | |
7414 | * array at a time. The result is that a new spare disk will be added | |
7415 | * to the first failed array and after the monitor has finished | |
7416 | * propagating failures the remainder will be consumed. | |
88758e9d | 7417 | * |
d23fe947 DW |
7418 | * FIXME add a capability for mdmon to request spares from another |
7419 | * container. | |
88758e9d DW |
7420 | */ |
7421 | ||
7422 | struct intel_super *super = a->container->sb; | |
88758e9d | 7423 | int inst = a->info.container_member; |
949c47a0 | 7424 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 7425 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
88758e9d DW |
7426 | int failed = a->info.array.raid_disks; |
7427 | struct mdinfo *rv = NULL; | |
7428 | struct mdinfo *d; | |
7429 | struct mdinfo *di; | |
7430 | struct metadata_update *mu; | |
7431 | struct dl *dl; | |
7432 | struct imsm_update_activate_spare *u; | |
7433 | int num_spares = 0; | |
7434 | int i; | |
95d07a2c | 7435 | int allowed; |
88758e9d DW |
7436 | |
7437 | for (d = a->info.devs ; d ; d = d->next) { | |
7438 | if ((d->curr_state & DS_FAULTY) && | |
7439 | d->state_fd >= 0) | |
7440 | /* wait for Removal to happen */ | |
7441 | return NULL; | |
7442 | if (d->state_fd >= 0) | |
7443 | failed--; | |
7444 | } | |
7445 | ||
7446 | dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n", | |
7447 | inst, failed, a->info.array.raid_disks, a->info.array.level); | |
1af97990 | 7448 | |
e2962bfc AK |
7449 | if (imsm_reshape_blocks_arrays_changes(super)) |
7450 | return NULL; | |
1af97990 | 7451 | |
fc8ca064 AK |
7452 | /* Cannot activate another spare if rebuild is in progress already |
7453 | */ | |
7454 | if (is_rebuilding(dev)) { | |
7455 | dprintf("imsm: No spare activation allowed. " | |
7456 | "Rebuild in progress already.\n"); | |
7457 | return NULL; | |
7458 | } | |
7459 | ||
89c67882 AK |
7460 | if (a->info.array.level == 4) |
7461 | /* No repair for takeovered array | |
7462 | * imsm doesn't support raid4 | |
7463 | */ | |
7464 | return NULL; | |
7465 | ||
3b451610 AK |
7466 | if (imsm_check_degraded(super, dev, failed, MAP_0) != |
7467 | IMSM_T_STATE_DEGRADED) | |
88758e9d DW |
7468 | return NULL; |
7469 | ||
95d07a2c LM |
7470 | /* |
7471 | * If there are any failed disks check state of the other volume. | |
7472 | * Block rebuild if the another one is failed until failed disks | |
7473 | * are removed from container. | |
7474 | */ | |
7475 | if (failed) { | |
c4acd1e5 | 7476 | dprintf("found failed disks in %.*s, check if there another" |
95d07a2c | 7477 | "failed sub-array.\n", |
c4acd1e5 | 7478 | MAX_RAID_SERIAL_LEN, dev->volume); |
95d07a2c LM |
7479 | /* check if states of the other volumes allow for rebuild */ |
7480 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
7481 | if (i != inst) { | |
7482 | allowed = imsm_rebuild_allowed(a->container, | |
7483 | i, failed); | |
7484 | if (!allowed) | |
7485 | return NULL; | |
7486 | } | |
7487 | } | |
7488 | } | |
7489 | ||
88758e9d | 7490 | /* For each slot, if it is not working, find a spare */ |
88758e9d DW |
7491 | for (i = 0; i < a->info.array.raid_disks; i++) { |
7492 | for (d = a->info.devs ; d ; d = d->next) | |
7493 | if (d->disk.raid_disk == i) | |
7494 | break; | |
7495 | dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0); | |
7496 | if (d && (d->state_fd >= 0)) | |
7497 | continue; | |
7498 | ||
272906ef | 7499 | /* |
a20d2ba5 DW |
7500 | * OK, this device needs recovery. Try to re-add the |
7501 | * previous occupant of this slot, if this fails see if | |
7502 | * we can continue the assimilation of a spare that was | |
7503 | * partially assimilated, finally try to activate a new | |
7504 | * spare. | |
272906ef DW |
7505 | */ |
7506 | dl = imsm_readd(super, i, a); | |
7507 | if (!dl) | |
b303fe21 | 7508 | dl = imsm_add_spare(super, i, a, 0, rv); |
a20d2ba5 | 7509 | if (!dl) |
b303fe21 | 7510 | dl = imsm_add_spare(super, i, a, 1, rv); |
272906ef DW |
7511 | if (!dl) |
7512 | continue; | |
7513 | ||
7514 | /* found a usable disk with enough space */ | |
7515 | di = malloc(sizeof(*di)); | |
79244939 DW |
7516 | if (!di) |
7517 | continue; | |
272906ef DW |
7518 | memset(di, 0, sizeof(*di)); |
7519 | ||
7520 | /* dl->index will be -1 in the case we are activating a | |
7521 | * pristine spare. imsm_process_update() will create a | |
7522 | * new index in this case. Once a disk is found to be | |
7523 | * failed in all member arrays it is kicked from the | |
7524 | * metadata | |
7525 | */ | |
7526 | di->disk.number = dl->index; | |
d23fe947 | 7527 | |
272906ef DW |
7528 | /* (ab)use di->devs to store a pointer to the device |
7529 | * we chose | |
7530 | */ | |
7531 | di->devs = (struct mdinfo *) dl; | |
7532 | ||
7533 | di->disk.raid_disk = i; | |
7534 | di->disk.major = dl->major; | |
7535 | di->disk.minor = dl->minor; | |
7536 | di->disk.state = 0; | |
d23534e4 | 7537 | di->recovery_start = 0; |
272906ef DW |
7538 | di->data_offset = __le32_to_cpu(map->pba_of_lba0); |
7539 | di->component_size = a->info.component_size; | |
7540 | di->container_member = inst; | |
148acb7b | 7541 | super->random = random32(); |
272906ef DW |
7542 | di->next = rv; |
7543 | rv = di; | |
7544 | num_spares++; | |
7545 | dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor, | |
7546 | i, di->data_offset); | |
88758e9d DW |
7547 | } |
7548 | ||
7549 | if (!rv) | |
7550 | /* No spares found */ | |
7551 | return rv; | |
7552 | /* Now 'rv' has a list of devices to return. | |
7553 | * Create a metadata_update record to update the | |
7554 | * disk_ord_tbl for the array | |
7555 | */ | |
7556 | mu = malloc(sizeof(*mu)); | |
79244939 DW |
7557 | if (mu) { |
7558 | mu->buf = malloc(sizeof(struct imsm_update_activate_spare) * num_spares); | |
7559 | if (mu->buf == NULL) { | |
7560 | free(mu); | |
7561 | mu = NULL; | |
7562 | } | |
7563 | } | |
7564 | if (!mu) { | |
7565 | while (rv) { | |
7566 | struct mdinfo *n = rv->next; | |
7567 | ||
7568 | free(rv); | |
7569 | rv = n; | |
7570 | } | |
7571 | return NULL; | |
7572 | } | |
ca9de185 | 7573 | |
88758e9d | 7574 | mu->space = NULL; |
cb23f1f4 | 7575 | mu->space_list = NULL; |
88758e9d DW |
7576 | mu->len = sizeof(struct imsm_update_activate_spare) * num_spares; |
7577 | mu->next = *updates; | |
7578 | u = (struct imsm_update_activate_spare *) mu->buf; | |
7579 | ||
7580 | for (di = rv ; di ; di = di->next) { | |
7581 | u->type = update_activate_spare; | |
d23fe947 DW |
7582 | u->dl = (struct dl *) di->devs; |
7583 | di->devs = NULL; | |
88758e9d DW |
7584 | u->slot = di->disk.raid_disk; |
7585 | u->array = inst; | |
7586 | u->next = u + 1; | |
7587 | u++; | |
7588 | } | |
7589 | (u-1)->next = NULL; | |
7590 | *updates = mu; | |
7591 | ||
7592 | return rv; | |
7593 | } | |
7594 | ||
54c2c1ea | 7595 | static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u) |
8273f55e | 7596 | { |
54c2c1ea | 7597 | struct imsm_dev *dev = get_imsm_dev(super, idx); |
238c0a71 AK |
7598 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
7599 | struct imsm_map *new_map = get_imsm_map(&u->dev, MAP_0); | |
54c2c1ea DW |
7600 | struct disk_info *inf = get_disk_info(u); |
7601 | struct imsm_disk *disk; | |
8273f55e DW |
7602 | int i; |
7603 | int j; | |
8273f55e | 7604 | |
54c2c1ea | 7605 | for (i = 0; i < map->num_members; i++) { |
238c0a71 | 7606 | disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, MAP_X)); |
54c2c1ea DW |
7607 | for (j = 0; j < new_map->num_members; j++) |
7608 | if (serialcmp(disk->serial, inf[j].serial) == 0) | |
8273f55e DW |
7609 | return 1; |
7610 | } | |
7611 | ||
7612 | return 0; | |
7613 | } | |
7614 | ||
1a64be56 LM |
7615 | |
7616 | static struct dl *get_disk_super(struct intel_super *super, int major, int minor) | |
7617 | { | |
7618 | struct dl *dl = NULL; | |
7619 | for (dl = super->disks; dl; dl = dl->next) | |
7620 | if ((dl->major == major) && (dl->minor == minor)) | |
7621 | return dl; | |
7622 | return NULL; | |
7623 | } | |
7624 | ||
7625 | static int remove_disk_super(struct intel_super *super, int major, int minor) | |
7626 | { | |
7627 | struct dl *prev = NULL; | |
7628 | struct dl *dl; | |
7629 | ||
7630 | prev = NULL; | |
7631 | for (dl = super->disks; dl; dl = dl->next) { | |
7632 | if ((dl->major == major) && (dl->minor == minor)) { | |
7633 | /* remove */ | |
7634 | if (prev) | |
7635 | prev->next = dl->next; | |
7636 | else | |
7637 | super->disks = dl->next; | |
7638 | dl->next = NULL; | |
7639 | __free_imsm_disk(dl); | |
7640 | dprintf("%s: removed %x:%x\n", | |
7641 | __func__, major, minor); | |
7642 | break; | |
7643 | } | |
7644 | prev = dl; | |
7645 | } | |
7646 | return 0; | |
7647 | } | |
7648 | ||
f21e18ca | 7649 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index); |
ae6aad82 | 7650 | |
1a64be56 LM |
7651 | static int add_remove_disk_update(struct intel_super *super) |
7652 | { | |
7653 | int check_degraded = 0; | |
7654 | struct dl *disk = NULL; | |
7655 | /* add/remove some spares to/from the metadata/contrainer */ | |
7656 | while (super->disk_mgmt_list) { | |
7657 | struct dl *disk_cfg; | |
7658 | ||
7659 | disk_cfg = super->disk_mgmt_list; | |
7660 | super->disk_mgmt_list = disk_cfg->next; | |
7661 | disk_cfg->next = NULL; | |
7662 | ||
7663 | if (disk_cfg->action == DISK_ADD) { | |
7664 | disk_cfg->next = super->disks; | |
7665 | super->disks = disk_cfg; | |
7666 | check_degraded = 1; | |
7667 | dprintf("%s: added %x:%x\n", | |
7668 | __func__, disk_cfg->major, | |
7669 | disk_cfg->minor); | |
7670 | } else if (disk_cfg->action == DISK_REMOVE) { | |
7671 | dprintf("Disk remove action processed: %x.%x\n", | |
7672 | disk_cfg->major, disk_cfg->minor); | |
7673 | disk = get_disk_super(super, | |
7674 | disk_cfg->major, | |
7675 | disk_cfg->minor); | |
7676 | if (disk) { | |
7677 | /* store action status */ | |
7678 | disk->action = DISK_REMOVE; | |
7679 | /* remove spare disks only */ | |
7680 | if (disk->index == -1) { | |
7681 | remove_disk_super(super, | |
7682 | disk_cfg->major, | |
7683 | disk_cfg->minor); | |
7684 | } | |
7685 | } | |
7686 | /* release allocate disk structure */ | |
7687 | __free_imsm_disk(disk_cfg); | |
7688 | } | |
7689 | } | |
7690 | return check_degraded; | |
7691 | } | |
7692 | ||
a29911da PC |
7693 | |
7694 | static int apply_reshape_migration_update(struct imsm_update_reshape_migration *u, | |
7695 | struct intel_super *super, | |
7696 | void ***space_list) | |
7697 | { | |
7698 | struct intel_dev *id; | |
7699 | void **tofree = NULL; | |
7700 | int ret_val = 0; | |
7701 | ||
7702 | dprintf("apply_reshape_migration_update()\n"); | |
7703 | if ((u->subdev < 0) || | |
7704 | (u->subdev > 1)) { | |
7705 | dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev); | |
7706 | return ret_val; | |
7707 | } | |
7708 | if ((space_list == NULL) || (*space_list == NULL)) { | |
7709 | dprintf("imsm: Error: Memory is not allocated\n"); | |
7710 | return ret_val; | |
7711 | } | |
7712 | ||
7713 | for (id = super->devlist ; id; id = id->next) { | |
7714 | if (id->index == (unsigned)u->subdev) { | |
7715 | struct imsm_dev *dev = get_imsm_dev(super, u->subdev); | |
7716 | struct imsm_map *map; | |
7717 | struct imsm_dev *new_dev = | |
7718 | (struct imsm_dev *)*space_list; | |
238c0a71 | 7719 | struct imsm_map *migr_map = get_imsm_map(dev, MAP_1); |
a29911da PC |
7720 | int to_state; |
7721 | struct dl *new_disk; | |
7722 | ||
7723 | if (new_dev == NULL) | |
7724 | return ret_val; | |
7725 | *space_list = **space_list; | |
7726 | memcpy(new_dev, dev, sizeof_imsm_dev(dev, 0)); | |
238c0a71 | 7727 | map = get_imsm_map(new_dev, MAP_0); |
a29911da PC |
7728 | if (migr_map) { |
7729 | dprintf("imsm: Error: migration in progress"); | |
7730 | return ret_val; | |
7731 | } | |
7732 | ||
7733 | to_state = map->map_state; | |
7734 | if ((u->new_level == 5) && (map->raid_level == 0)) { | |
7735 | map->num_members++; | |
7736 | /* this should not happen */ | |
7737 | if (u->new_disks[0] < 0) { | |
7738 | map->failed_disk_num = | |
7739 | map->num_members - 1; | |
7740 | to_state = IMSM_T_STATE_DEGRADED; | |
7741 | } else | |
7742 | to_state = IMSM_T_STATE_NORMAL; | |
7743 | } | |
8e59f3d8 | 7744 | migrate(new_dev, super, to_state, MIGR_GEN_MIGR); |
a29911da PC |
7745 | if (u->new_level > -1) |
7746 | map->raid_level = u->new_level; | |
238c0a71 | 7747 | migr_map = get_imsm_map(new_dev, MAP_1); |
a29911da PC |
7748 | if ((u->new_level == 5) && |
7749 | (migr_map->raid_level == 0)) { | |
7750 | int ord = map->num_members - 1; | |
7751 | migr_map->num_members--; | |
7752 | if (u->new_disks[0] < 0) | |
7753 | ord |= IMSM_ORD_REBUILD; | |
7754 | set_imsm_ord_tbl_ent(map, | |
7755 | map->num_members - 1, | |
7756 | ord); | |
7757 | } | |
7758 | id->dev = new_dev; | |
7759 | tofree = (void **)dev; | |
7760 | ||
4bba0439 PC |
7761 | /* update chunk size |
7762 | */ | |
7763 | if (u->new_chunksize > 0) | |
7764 | map->blocks_per_strip = | |
7765 | __cpu_to_le16(u->new_chunksize * 2); | |
7766 | ||
a29911da PC |
7767 | /* add disk |
7768 | */ | |
7769 | if ((u->new_level != 5) || | |
7770 | (migr_map->raid_level != 0) || | |
7771 | (migr_map->raid_level == map->raid_level)) | |
7772 | goto skip_disk_add; | |
7773 | ||
7774 | if (u->new_disks[0] >= 0) { | |
7775 | /* use passes spare | |
7776 | */ | |
7777 | new_disk = get_disk_super(super, | |
7778 | major(u->new_disks[0]), | |
7779 | minor(u->new_disks[0])); | |
7780 | dprintf("imsm: new disk for reshape is: %i:%i " | |
7781 | "(%p, index = %i)\n", | |
7782 | major(u->new_disks[0]), | |
7783 | minor(u->new_disks[0]), | |
7784 | new_disk, new_disk->index); | |
7785 | if (new_disk == NULL) | |
7786 | goto error_disk_add; | |
7787 | ||
7788 | new_disk->index = map->num_members - 1; | |
7789 | /* slot to fill in autolayout | |
7790 | */ | |
7791 | new_disk->raiddisk = new_disk->index; | |
7792 | new_disk->disk.status |= CONFIGURED_DISK; | |
7793 | new_disk->disk.status &= ~SPARE_DISK; | |
7794 | } else | |
7795 | goto error_disk_add; | |
7796 | ||
7797 | skip_disk_add: | |
7798 | *tofree = *space_list; | |
7799 | /* calculate new size | |
7800 | */ | |
7801 | imsm_set_array_size(new_dev); | |
7802 | ||
7803 | ret_val = 1; | |
7804 | } | |
7805 | } | |
7806 | ||
7807 | if (tofree) | |
7808 | *space_list = tofree; | |
7809 | return ret_val; | |
7810 | ||
7811 | error_disk_add: | |
7812 | dprintf("Error: imsm: Cannot find disk.\n"); | |
7813 | return ret_val; | |
7814 | } | |
7815 | ||
061d7da3 | 7816 | static int apply_update_activate_spare(struct imsm_update_activate_spare *u, |
ca9de185 | 7817 | struct intel_super *super, |
061d7da3 LO |
7818 | struct active_array *active_array) |
7819 | { | |
7820 | struct imsm_super *mpb = super->anchor; | |
7821 | struct imsm_dev *dev = get_imsm_dev(super, u->array); | |
238c0a71 | 7822 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
061d7da3 LO |
7823 | struct imsm_map *migr_map; |
7824 | struct active_array *a; | |
7825 | struct imsm_disk *disk; | |
7826 | __u8 to_state; | |
7827 | struct dl *dl; | |
7828 | unsigned int found; | |
7829 | int failed; | |
5961eeec | 7830 | int victim; |
061d7da3 | 7831 | int i; |
5961eeec | 7832 | int second_map_created = 0; |
061d7da3 | 7833 | |
5961eeec | 7834 | for (; u; u = u->next) { |
238c0a71 | 7835 | victim = get_imsm_disk_idx(dev, u->slot, MAP_X); |
061d7da3 | 7836 | |
5961eeec | 7837 | if (victim < 0) |
7838 | return 0; | |
061d7da3 | 7839 | |
5961eeec | 7840 | for (dl = super->disks; dl; dl = dl->next) |
7841 | if (dl == u->dl) | |
7842 | break; | |
061d7da3 | 7843 | |
5961eeec | 7844 | if (!dl) { |
7845 | fprintf(stderr, "error: imsm_activate_spare passed " | |
7846 | "an unknown disk (index: %d)\n", | |
7847 | u->dl->index); | |
7848 | return 0; | |
7849 | } | |
061d7da3 | 7850 | |
5961eeec | 7851 | /* count failures (excluding rebuilds and the victim) |
7852 | * to determine map[0] state | |
7853 | */ | |
7854 | failed = 0; | |
7855 | for (i = 0; i < map->num_members; i++) { | |
7856 | if (i == u->slot) | |
7857 | continue; | |
7858 | disk = get_imsm_disk(super, | |
238c0a71 | 7859 | get_imsm_disk_idx(dev, i, MAP_X)); |
5961eeec | 7860 | if (!disk || is_failed(disk)) |
7861 | failed++; | |
7862 | } | |
061d7da3 | 7863 | |
5961eeec | 7864 | /* adding a pristine spare, assign a new index */ |
7865 | if (dl->index < 0) { | |
7866 | dl->index = super->anchor->num_disks; | |
7867 | super->anchor->num_disks++; | |
7868 | } | |
7869 | disk = &dl->disk; | |
7870 | disk->status |= CONFIGURED_DISK; | |
7871 | disk->status &= ~SPARE_DISK; | |
7872 | ||
7873 | /* mark rebuild */ | |
238c0a71 | 7874 | to_state = imsm_check_degraded(super, dev, failed, MAP_0); |
5961eeec | 7875 | if (!second_map_created) { |
7876 | second_map_created = 1; | |
7877 | map->map_state = IMSM_T_STATE_DEGRADED; | |
7878 | migrate(dev, super, to_state, MIGR_REBUILD); | |
7879 | } else | |
7880 | map->map_state = to_state; | |
238c0a71 | 7881 | migr_map = get_imsm_map(dev, MAP_1); |
5961eeec | 7882 | set_imsm_ord_tbl_ent(map, u->slot, dl->index); |
7883 | set_imsm_ord_tbl_ent(migr_map, u->slot, | |
7884 | dl->index | IMSM_ORD_REBUILD); | |
7885 | ||
7886 | /* update the family_num to mark a new container | |
7887 | * generation, being careful to record the existing | |
7888 | * family_num in orig_family_num to clean up after | |
7889 | * earlier mdadm versions that neglected to set it. | |
7890 | */ | |
7891 | if (mpb->orig_family_num == 0) | |
7892 | mpb->orig_family_num = mpb->family_num; | |
7893 | mpb->family_num += super->random; | |
7894 | ||
7895 | /* count arrays using the victim in the metadata */ | |
7896 | found = 0; | |
7897 | for (a = active_array; a ; a = a->next) { | |
7898 | dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 7899 | map = get_imsm_map(dev, MAP_0); |
061d7da3 | 7900 | |
5961eeec | 7901 | if (get_imsm_disk_slot(map, victim) >= 0) |
7902 | found++; | |
7903 | } | |
061d7da3 | 7904 | |
5961eeec | 7905 | /* delete the victim if it is no longer being |
7906 | * utilized anywhere | |
061d7da3 | 7907 | */ |
5961eeec | 7908 | if (!found) { |
7909 | struct dl **dlp; | |
061d7da3 | 7910 | |
5961eeec | 7911 | /* We know that 'manager' isn't touching anything, |
7912 | * so it is safe to delete | |
7913 | */ | |
7914 | for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next) | |
061d7da3 LO |
7915 | if ((*dlp)->index == victim) |
7916 | break; | |
5961eeec | 7917 | |
7918 | /* victim may be on the missing list */ | |
7919 | if (!*dlp) | |
7920 | for (dlp = &super->missing; *dlp; | |
7921 | dlp = &(*dlp)->next) | |
7922 | if ((*dlp)->index == victim) | |
7923 | break; | |
7924 | imsm_delete(super, dlp, victim); | |
7925 | } | |
061d7da3 LO |
7926 | } |
7927 | ||
7928 | return 1; | |
7929 | } | |
a29911da | 7930 | |
2e5dc010 N |
7931 | static int apply_reshape_container_disks_update(struct imsm_update_reshape *u, |
7932 | struct intel_super *super, | |
7933 | void ***space_list) | |
7934 | { | |
7935 | struct dl *new_disk; | |
7936 | struct intel_dev *id; | |
7937 | int i; | |
7938 | int delta_disks = u->new_raid_disks - u->old_raid_disks; | |
ee4beede | 7939 | int disk_count = u->old_raid_disks; |
2e5dc010 N |
7940 | void **tofree = NULL; |
7941 | int devices_to_reshape = 1; | |
7942 | struct imsm_super *mpb = super->anchor; | |
7943 | int ret_val = 0; | |
d098291a | 7944 | unsigned int dev_id; |
2e5dc010 | 7945 | |
ed7333bd | 7946 | dprintf("imsm: apply_reshape_container_disks_update()\n"); |
2e5dc010 N |
7947 | |
7948 | /* enable spares to use in array */ | |
7949 | for (i = 0; i < delta_disks; i++) { | |
7950 | new_disk = get_disk_super(super, | |
7951 | major(u->new_disks[i]), | |
7952 | minor(u->new_disks[i])); | |
ed7333bd AK |
7953 | dprintf("imsm: new disk for reshape is: %i:%i " |
7954 | "(%p, index = %i)\n", | |
2e5dc010 N |
7955 | major(u->new_disks[i]), minor(u->new_disks[i]), |
7956 | new_disk, new_disk->index); | |
7957 | if ((new_disk == NULL) || | |
7958 | ((new_disk->index >= 0) && | |
7959 | (new_disk->index < u->old_raid_disks))) | |
7960 | goto update_reshape_exit; | |
ee4beede | 7961 | new_disk->index = disk_count++; |
2e5dc010 N |
7962 | /* slot to fill in autolayout |
7963 | */ | |
7964 | new_disk->raiddisk = new_disk->index; | |
7965 | new_disk->disk.status |= | |
7966 | CONFIGURED_DISK; | |
7967 | new_disk->disk.status &= ~SPARE_DISK; | |
7968 | } | |
7969 | ||
ed7333bd AK |
7970 | dprintf("imsm: volume set mpb->num_raid_devs = %i\n", |
7971 | mpb->num_raid_devs); | |
2e5dc010 N |
7972 | /* manage changes in volume |
7973 | */ | |
d098291a | 7974 | for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) { |
2e5dc010 N |
7975 | void **sp = *space_list; |
7976 | struct imsm_dev *newdev; | |
7977 | struct imsm_map *newmap, *oldmap; | |
7978 | ||
d098291a AK |
7979 | for (id = super->devlist ; id; id = id->next) { |
7980 | if (id->index == dev_id) | |
7981 | break; | |
7982 | } | |
7983 | if (id == NULL) | |
7984 | break; | |
2e5dc010 N |
7985 | if (!sp) |
7986 | continue; | |
7987 | *space_list = *sp; | |
7988 | newdev = (void*)sp; | |
7989 | /* Copy the dev, but not (all of) the map */ | |
7990 | memcpy(newdev, id->dev, sizeof(*newdev)); | |
238c0a71 AK |
7991 | oldmap = get_imsm_map(id->dev, MAP_0); |
7992 | newmap = get_imsm_map(newdev, MAP_0); | |
2e5dc010 N |
7993 | /* Copy the current map */ |
7994 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); | |
7995 | /* update one device only | |
7996 | */ | |
7997 | if (devices_to_reshape) { | |
ed7333bd AK |
7998 | dprintf("imsm: modifying subdev: %i\n", |
7999 | id->index); | |
2e5dc010 N |
8000 | devices_to_reshape--; |
8001 | newdev->vol.migr_state = 1; | |
8002 | newdev->vol.curr_migr_unit = 0; | |
ea672ee1 | 8003 | set_migr_type(newdev, MIGR_GEN_MIGR); |
2e5dc010 N |
8004 | newmap->num_members = u->new_raid_disks; |
8005 | for (i = 0; i < delta_disks; i++) { | |
8006 | set_imsm_ord_tbl_ent(newmap, | |
8007 | u->old_raid_disks + i, | |
8008 | u->old_raid_disks + i); | |
8009 | } | |
8010 | /* New map is correct, now need to save old map | |
8011 | */ | |
238c0a71 | 8012 | newmap = get_imsm_map(newdev, MAP_1); |
2e5dc010 N |
8013 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); |
8014 | ||
70bdf0dc | 8015 | imsm_set_array_size(newdev); |
2e5dc010 N |
8016 | } |
8017 | ||
8018 | sp = (void **)id->dev; | |
8019 | id->dev = newdev; | |
8020 | *sp = tofree; | |
8021 | tofree = sp; | |
8e59f3d8 AK |
8022 | |
8023 | /* Clear migration record */ | |
8024 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
2e5dc010 | 8025 | } |
819bc634 AK |
8026 | if (tofree) |
8027 | *space_list = tofree; | |
2e5dc010 N |
8028 | ret_val = 1; |
8029 | ||
8030 | update_reshape_exit: | |
8031 | ||
8032 | return ret_val; | |
8033 | } | |
8034 | ||
bb025c2f | 8035 | static int apply_takeover_update(struct imsm_update_takeover *u, |
8ca6df95 KW |
8036 | struct intel_super *super, |
8037 | void ***space_list) | |
bb025c2f KW |
8038 | { |
8039 | struct imsm_dev *dev = NULL; | |
8ca6df95 KW |
8040 | struct intel_dev *dv; |
8041 | struct imsm_dev *dev_new; | |
bb025c2f KW |
8042 | struct imsm_map *map; |
8043 | struct dl *dm, *du; | |
8ca6df95 | 8044 | int i; |
bb025c2f KW |
8045 | |
8046 | for (dv = super->devlist; dv; dv = dv->next) | |
8047 | if (dv->index == (unsigned int)u->subarray) { | |
8048 | dev = dv->dev; | |
8049 | break; | |
8050 | } | |
8051 | ||
8052 | if (dev == NULL) | |
8053 | return 0; | |
8054 | ||
238c0a71 | 8055 | map = get_imsm_map(dev, MAP_0); |
bb025c2f KW |
8056 | |
8057 | if (u->direction == R10_TO_R0) { | |
43d5ec18 | 8058 | /* Number of failed disks must be half of initial disk number */ |
3b451610 AK |
8059 | if (imsm_count_failed(super, dev, MAP_0) != |
8060 | (map->num_members / 2)) | |
43d5ec18 KW |
8061 | return 0; |
8062 | ||
bb025c2f KW |
8063 | /* iterate through devices to mark removed disks as spare */ |
8064 | for (dm = super->disks; dm; dm = dm->next) { | |
8065 | if (dm->disk.status & FAILED_DISK) { | |
8066 | int idx = dm->index; | |
8067 | /* update indexes on the disk list */ | |
8068 | /* FIXME this loop-with-the-loop looks wrong, I'm not convinced | |
8069 | the index values will end up being correct.... NB */ | |
8070 | for (du = super->disks; du; du = du->next) | |
8071 | if (du->index > idx) | |
8072 | du->index--; | |
8073 | /* mark as spare disk */ | |
a8619d23 | 8074 | mark_spare(dm); |
bb025c2f KW |
8075 | } |
8076 | } | |
bb025c2f KW |
8077 | /* update map */ |
8078 | map->num_members = map->num_members / 2; | |
8079 | map->map_state = IMSM_T_STATE_NORMAL; | |
8080 | map->num_domains = 1; | |
8081 | map->raid_level = 0; | |
8082 | map->failed_disk_num = -1; | |
8083 | } | |
8084 | ||
8ca6df95 KW |
8085 | if (u->direction == R0_TO_R10) { |
8086 | void **space; | |
8087 | /* update slots in current disk list */ | |
8088 | for (dm = super->disks; dm; dm = dm->next) { | |
8089 | if (dm->index >= 0) | |
8090 | dm->index *= 2; | |
8091 | } | |
8092 | /* create new *missing* disks */ | |
8093 | for (i = 0; i < map->num_members; i++) { | |
8094 | space = *space_list; | |
8095 | if (!space) | |
8096 | continue; | |
8097 | *space_list = *space; | |
8098 | du = (void *)space; | |
8099 | memcpy(du, super->disks, sizeof(*du)); | |
8ca6df95 KW |
8100 | du->fd = -1; |
8101 | du->minor = 0; | |
8102 | du->major = 0; | |
8103 | du->index = (i * 2) + 1; | |
8104 | sprintf((char *)du->disk.serial, | |
8105 | " MISSING_%d", du->index); | |
8106 | sprintf((char *)du->serial, | |
8107 | "MISSING_%d", du->index); | |
8108 | du->next = super->missing; | |
8109 | super->missing = du; | |
8110 | } | |
8111 | /* create new dev and map */ | |
8112 | space = *space_list; | |
8113 | if (!space) | |
8114 | return 0; | |
8115 | *space_list = *space; | |
8116 | dev_new = (void *)space; | |
8117 | memcpy(dev_new, dev, sizeof(*dev)); | |
8118 | /* update new map */ | |
238c0a71 | 8119 | map = get_imsm_map(dev_new, MAP_0); |
8ca6df95 | 8120 | map->num_members = map->num_members * 2; |
1a2487c2 | 8121 | map->map_state = IMSM_T_STATE_DEGRADED; |
8ca6df95 KW |
8122 | map->num_domains = 2; |
8123 | map->raid_level = 1; | |
8124 | /* replace dev<->dev_new */ | |
8125 | dv->dev = dev_new; | |
8126 | } | |
bb025c2f KW |
8127 | /* update disk order table */ |
8128 | for (du = super->disks; du; du = du->next) | |
8129 | if (du->index >= 0) | |
8130 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
8ca6df95 | 8131 | for (du = super->missing; du; du = du->next) |
1a2487c2 KW |
8132 | if (du->index >= 0) { |
8133 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
e4c72d1d | 8134 | mark_missing(dv->dev, &du->disk, du->index); |
1a2487c2 | 8135 | } |
bb025c2f KW |
8136 | |
8137 | return 1; | |
8138 | } | |
8139 | ||
e8319a19 DW |
8140 | static void imsm_process_update(struct supertype *st, |
8141 | struct metadata_update *update) | |
8142 | { | |
8143 | /** | |
8144 | * crack open the metadata_update envelope to find the update record | |
8145 | * update can be one of: | |
d195167d AK |
8146 | * update_reshape_container_disks - all the arrays in the container |
8147 | * are being reshaped to have more devices. We need to mark | |
8148 | * the arrays for general migration and convert selected spares | |
8149 | * into active devices. | |
8150 | * update_activate_spare - a spare device has replaced a failed | |
e8319a19 DW |
8151 | * device in an array, update the disk_ord_tbl. If this disk is |
8152 | * present in all member arrays then also clear the SPARE_DISK | |
8153 | * flag | |
d195167d AK |
8154 | * update_create_array |
8155 | * update_kill_array | |
8156 | * update_rename_array | |
8157 | * update_add_remove_disk | |
e8319a19 DW |
8158 | */ |
8159 | struct intel_super *super = st->sb; | |
4d7b1503 | 8160 | struct imsm_super *mpb; |
e8319a19 DW |
8161 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
8162 | ||
4d7b1503 DW |
8163 | /* update requires a larger buf but the allocation failed */ |
8164 | if (super->next_len && !super->next_buf) { | |
8165 | super->next_len = 0; | |
8166 | return; | |
8167 | } | |
8168 | ||
8169 | if (super->next_buf) { | |
8170 | memcpy(super->next_buf, super->buf, super->len); | |
8171 | free(super->buf); | |
8172 | super->len = super->next_len; | |
8173 | super->buf = super->next_buf; | |
8174 | ||
8175 | super->next_len = 0; | |
8176 | super->next_buf = NULL; | |
8177 | } | |
8178 | ||
8179 | mpb = super->anchor; | |
8180 | ||
e8319a19 | 8181 | switch (type) { |
0ec5d470 AK |
8182 | case update_general_migration_checkpoint: { |
8183 | struct intel_dev *id; | |
8184 | struct imsm_update_general_migration_checkpoint *u = | |
8185 | (void *)update->buf; | |
8186 | ||
8187 | dprintf("imsm: process_update() " | |
8188 | "for update_general_migration_checkpoint called\n"); | |
8189 | ||
8190 | /* find device under general migration */ | |
8191 | for (id = super->devlist ; id; id = id->next) { | |
8192 | if (is_gen_migration(id->dev)) { | |
8193 | id->dev->vol.curr_migr_unit = | |
8194 | __cpu_to_le32(u->curr_migr_unit); | |
8195 | super->updates_pending++; | |
8196 | } | |
8197 | } | |
8198 | break; | |
8199 | } | |
bb025c2f KW |
8200 | case update_takeover: { |
8201 | struct imsm_update_takeover *u = (void *)update->buf; | |
1a2487c2 KW |
8202 | if (apply_takeover_update(u, super, &update->space_list)) { |
8203 | imsm_update_version_info(super); | |
bb025c2f | 8204 | super->updates_pending++; |
1a2487c2 | 8205 | } |
bb025c2f KW |
8206 | break; |
8207 | } | |
8208 | ||
78b10e66 | 8209 | case update_reshape_container_disks: { |
d195167d | 8210 | struct imsm_update_reshape *u = (void *)update->buf; |
2e5dc010 N |
8211 | if (apply_reshape_container_disks_update( |
8212 | u, super, &update->space_list)) | |
8213 | super->updates_pending++; | |
78b10e66 N |
8214 | break; |
8215 | } | |
48c5303a | 8216 | case update_reshape_migration: { |
a29911da PC |
8217 | struct imsm_update_reshape_migration *u = (void *)update->buf; |
8218 | if (apply_reshape_migration_update( | |
8219 | u, super, &update->space_list)) | |
8220 | super->updates_pending++; | |
48c5303a PC |
8221 | break; |
8222 | } | |
e8319a19 DW |
8223 | case update_activate_spare: { |
8224 | struct imsm_update_activate_spare *u = (void *) update->buf; | |
061d7da3 LO |
8225 | if (apply_update_activate_spare(u, super, st->arrays)) |
8226 | super->updates_pending++; | |
8273f55e DW |
8227 | break; |
8228 | } | |
8229 | case update_create_array: { | |
8230 | /* someone wants to create a new array, we need to be aware of | |
8231 | * a few races/collisions: | |
8232 | * 1/ 'Create' called by two separate instances of mdadm | |
8233 | * 2/ 'Create' versus 'activate_spare': mdadm has chosen | |
8234 | * devices that have since been assimilated via | |
8235 | * activate_spare. | |
8236 | * In the event this update can not be carried out mdadm will | |
8237 | * (FIX ME) notice that its update did not take hold. | |
8238 | */ | |
8239 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 8240 | struct intel_dev *dv; |
8273f55e DW |
8241 | struct imsm_dev *dev; |
8242 | struct imsm_map *map, *new_map; | |
8243 | unsigned long long start, end; | |
8244 | unsigned long long new_start, new_end; | |
8245 | int i; | |
54c2c1ea DW |
8246 | struct disk_info *inf; |
8247 | struct dl *dl; | |
8273f55e DW |
8248 | |
8249 | /* handle racing creates: first come first serve */ | |
8250 | if (u->dev_idx < mpb->num_raid_devs) { | |
8251 | dprintf("%s: subarray %d already defined\n", | |
8252 | __func__, u->dev_idx); | |
ba2de7ba | 8253 | goto create_error; |
8273f55e DW |
8254 | } |
8255 | ||
8256 | /* check update is next in sequence */ | |
8257 | if (u->dev_idx != mpb->num_raid_devs) { | |
6a3e913e DW |
8258 | dprintf("%s: can not create array %d expected index %d\n", |
8259 | __func__, u->dev_idx, mpb->num_raid_devs); | |
ba2de7ba | 8260 | goto create_error; |
8273f55e DW |
8261 | } |
8262 | ||
238c0a71 | 8263 | new_map = get_imsm_map(&u->dev, MAP_0); |
8273f55e DW |
8264 | new_start = __le32_to_cpu(new_map->pba_of_lba0); |
8265 | new_end = new_start + __le32_to_cpu(new_map->blocks_per_member); | |
54c2c1ea | 8266 | inf = get_disk_info(u); |
8273f55e DW |
8267 | |
8268 | /* handle activate_spare versus create race: | |
8269 | * check to make sure that overlapping arrays do not include | |
8270 | * overalpping disks | |
8271 | */ | |
8272 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
949c47a0 | 8273 | dev = get_imsm_dev(super, i); |
238c0a71 | 8274 | map = get_imsm_map(dev, MAP_0); |
8273f55e DW |
8275 | start = __le32_to_cpu(map->pba_of_lba0); |
8276 | end = start + __le32_to_cpu(map->blocks_per_member); | |
8277 | if ((new_start >= start && new_start <= end) || | |
8278 | (start >= new_start && start <= new_end)) | |
54c2c1ea DW |
8279 | /* overlap */; |
8280 | else | |
8281 | continue; | |
8282 | ||
8283 | if (disks_overlap(super, i, u)) { | |
8273f55e | 8284 | dprintf("%s: arrays overlap\n", __func__); |
ba2de7ba | 8285 | goto create_error; |
8273f55e DW |
8286 | } |
8287 | } | |
8273f55e | 8288 | |
949c47a0 DW |
8289 | /* check that prepare update was successful */ |
8290 | if (!update->space) { | |
8291 | dprintf("%s: prepare update failed\n", __func__); | |
ba2de7ba | 8292 | goto create_error; |
949c47a0 DW |
8293 | } |
8294 | ||
54c2c1ea DW |
8295 | /* check that all disks are still active before committing |
8296 | * changes. FIXME: could we instead handle this by creating a | |
8297 | * degraded array? That's probably not what the user expects, | |
8298 | * so better to drop this update on the floor. | |
8299 | */ | |
8300 | for (i = 0; i < new_map->num_members; i++) { | |
8301 | dl = serial_to_dl(inf[i].serial, super); | |
8302 | if (!dl) { | |
8303 | dprintf("%s: disk disappeared\n", __func__); | |
ba2de7ba | 8304 | goto create_error; |
54c2c1ea | 8305 | } |
949c47a0 DW |
8306 | } |
8307 | ||
8273f55e | 8308 | super->updates_pending++; |
54c2c1ea DW |
8309 | |
8310 | /* convert spares to members and fixup ord_tbl */ | |
8311 | for (i = 0; i < new_map->num_members; i++) { | |
8312 | dl = serial_to_dl(inf[i].serial, super); | |
8313 | if (dl->index == -1) { | |
8314 | dl->index = mpb->num_disks; | |
8315 | mpb->num_disks++; | |
8316 | dl->disk.status |= CONFIGURED_DISK; | |
8317 | dl->disk.status &= ~SPARE_DISK; | |
8318 | } | |
8319 | set_imsm_ord_tbl_ent(new_map, i, dl->index); | |
8320 | } | |
8321 | ||
ba2de7ba DW |
8322 | dv = update->space; |
8323 | dev = dv->dev; | |
949c47a0 DW |
8324 | update->space = NULL; |
8325 | imsm_copy_dev(dev, &u->dev); | |
ba2de7ba DW |
8326 | dv->index = u->dev_idx; |
8327 | dv->next = super->devlist; | |
8328 | super->devlist = dv; | |
8273f55e | 8329 | mpb->num_raid_devs++; |
8273f55e | 8330 | |
4d1313e9 | 8331 | imsm_update_version_info(super); |
8273f55e | 8332 | break; |
ba2de7ba DW |
8333 | create_error: |
8334 | /* mdmon knows how to release update->space, but not | |
8335 | * ((struct intel_dev *) update->space)->dev | |
8336 | */ | |
8337 | if (update->space) { | |
8338 | dv = update->space; | |
8339 | free(dv->dev); | |
8340 | } | |
8273f55e | 8341 | break; |
e8319a19 | 8342 | } |
33414a01 DW |
8343 | case update_kill_array: { |
8344 | struct imsm_update_kill_array *u = (void *) update->buf; | |
8345 | int victim = u->dev_idx; | |
8346 | struct active_array *a; | |
8347 | struct intel_dev **dp; | |
8348 | struct imsm_dev *dev; | |
8349 | ||
8350 | /* sanity check that we are not affecting the uuid of | |
8351 | * active arrays, or deleting an active array | |
8352 | * | |
8353 | * FIXME when immutable ids are available, but note that | |
8354 | * we'll also need to fixup the invalidated/active | |
8355 | * subarray indexes in mdstat | |
8356 | */ | |
8357 | for (a = st->arrays; a; a = a->next) | |
8358 | if (a->info.container_member >= victim) | |
8359 | break; | |
8360 | /* by definition if mdmon is running at least one array | |
8361 | * is active in the container, so checking | |
8362 | * mpb->num_raid_devs is just extra paranoia | |
8363 | */ | |
8364 | dev = get_imsm_dev(super, victim); | |
8365 | if (a || !dev || mpb->num_raid_devs == 1) { | |
8366 | dprintf("failed to delete subarray-%d\n", victim); | |
8367 | break; | |
8368 | } | |
8369 | ||
8370 | for (dp = &super->devlist; *dp;) | |
f21e18ca | 8371 | if ((*dp)->index == (unsigned)super->current_vol) { |
33414a01 DW |
8372 | *dp = (*dp)->next; |
8373 | } else { | |
f21e18ca | 8374 | if ((*dp)->index > (unsigned)victim) |
33414a01 DW |
8375 | (*dp)->index--; |
8376 | dp = &(*dp)->next; | |
8377 | } | |
8378 | mpb->num_raid_devs--; | |
8379 | super->updates_pending++; | |
8380 | break; | |
8381 | } | |
aa534678 DW |
8382 | case update_rename_array: { |
8383 | struct imsm_update_rename_array *u = (void *) update->buf; | |
8384 | char name[MAX_RAID_SERIAL_LEN+1]; | |
8385 | int target = u->dev_idx; | |
8386 | struct active_array *a; | |
8387 | struct imsm_dev *dev; | |
8388 | ||
8389 | /* sanity check that we are not affecting the uuid of | |
8390 | * an active array | |
8391 | */ | |
8392 | snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name); | |
8393 | name[MAX_RAID_SERIAL_LEN] = '\0'; | |
8394 | for (a = st->arrays; a; a = a->next) | |
8395 | if (a->info.container_member == target) | |
8396 | break; | |
8397 | dev = get_imsm_dev(super, u->dev_idx); | |
8398 | if (a || !dev || !check_name(super, name, 1)) { | |
8399 | dprintf("failed to rename subarray-%d\n", target); | |
8400 | break; | |
8401 | } | |
8402 | ||
cdbe98cd | 8403 | snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name); |
aa534678 DW |
8404 | super->updates_pending++; |
8405 | break; | |
8406 | } | |
1a64be56 | 8407 | case update_add_remove_disk: { |
43dad3d6 | 8408 | /* we may be able to repair some arrays if disks are |
1a64be56 LM |
8409 | * being added, check teh status of add_remove_disk |
8410 | * if discs has been added. | |
8411 | */ | |
8412 | if (add_remove_disk_update(super)) { | |
43dad3d6 | 8413 | struct active_array *a; |
072b727f DW |
8414 | |
8415 | super->updates_pending++; | |
1a64be56 | 8416 | for (a = st->arrays; a; a = a->next) |
43dad3d6 DW |
8417 | a->check_degraded = 1; |
8418 | } | |
43dad3d6 | 8419 | break; |
e8319a19 | 8420 | } |
1a64be56 LM |
8421 | default: |
8422 | fprintf(stderr, "error: unsuported process update type:" | |
8423 | "(type: %d)\n", type); | |
8424 | } | |
e8319a19 | 8425 | } |
88758e9d | 8426 | |
bc0b9d34 PC |
8427 | static struct mdinfo *get_spares_for_grow(struct supertype *st); |
8428 | ||
8273f55e DW |
8429 | static void imsm_prepare_update(struct supertype *st, |
8430 | struct metadata_update *update) | |
8431 | { | |
949c47a0 | 8432 | /** |
4d7b1503 DW |
8433 | * Allocate space to hold new disk entries, raid-device entries or a new |
8434 | * mpb if necessary. The manager synchronously waits for updates to | |
8435 | * complete in the monitor, so new mpb buffers allocated here can be | |
8436 | * integrated by the monitor thread without worrying about live pointers | |
8437 | * in the manager thread. | |
8273f55e | 8438 | */ |
949c47a0 | 8439 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
4d7b1503 DW |
8440 | struct intel_super *super = st->sb; |
8441 | struct imsm_super *mpb = super->anchor; | |
8442 | size_t buf_len; | |
8443 | size_t len = 0; | |
949c47a0 DW |
8444 | |
8445 | switch (type) { | |
0ec5d470 AK |
8446 | case update_general_migration_checkpoint: |
8447 | dprintf("imsm: prepare_update() " | |
8448 | "for update_general_migration_checkpoint called\n"); | |
8449 | break; | |
abedf5fc KW |
8450 | case update_takeover: { |
8451 | struct imsm_update_takeover *u = (void *)update->buf; | |
8452 | if (u->direction == R0_TO_R10) { | |
8453 | void **tail = (void **)&update->space_list; | |
8454 | struct imsm_dev *dev = get_imsm_dev(super, u->subarray); | |
238c0a71 | 8455 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
abedf5fc KW |
8456 | int num_members = map->num_members; |
8457 | void *space; | |
8458 | int size, i; | |
8459 | int err = 0; | |
8460 | /* allocate memory for added disks */ | |
8461 | for (i = 0; i < num_members; i++) { | |
8462 | size = sizeof(struct dl); | |
8463 | space = malloc(size); | |
8464 | if (!space) { | |
8465 | err++; | |
8466 | break; | |
8467 | } | |
8468 | *tail = space; | |
8469 | tail = space; | |
8470 | *tail = NULL; | |
8471 | } | |
8472 | /* allocate memory for new device */ | |
8473 | size = sizeof_imsm_dev(super->devlist->dev, 0) + | |
8474 | (num_members * sizeof(__u32)); | |
8475 | space = malloc(size); | |
8476 | if (!space) | |
8477 | err++; | |
8478 | else { | |
8479 | *tail = space; | |
8480 | tail = space; | |
8481 | *tail = NULL; | |
8482 | } | |
8483 | if (!err) { | |
8484 | len = disks_to_mpb_size(num_members * 2); | |
8485 | } else { | |
8486 | /* if allocation didn't success, free buffer */ | |
8487 | while (update->space_list) { | |
8488 | void **sp = update->space_list; | |
8489 | update->space_list = *sp; | |
8490 | free(sp); | |
8491 | } | |
8492 | } | |
8493 | } | |
8494 | ||
8495 | break; | |
8496 | } | |
78b10e66 | 8497 | case update_reshape_container_disks: { |
d195167d AK |
8498 | /* Every raid device in the container is about to |
8499 | * gain some more devices, and we will enter a | |
8500 | * reconfiguration. | |
8501 | * So each 'imsm_map' will be bigger, and the imsm_vol | |
8502 | * will now hold 2 of them. | |
8503 | * Thus we need new 'struct imsm_dev' allocations sized | |
8504 | * as sizeof_imsm_dev but with more devices in both maps. | |
8505 | */ | |
8506 | struct imsm_update_reshape *u = (void *)update->buf; | |
8507 | struct intel_dev *dl; | |
8508 | void **space_tail = (void**)&update->space_list; | |
8509 | ||
8510 | dprintf("imsm: imsm_prepare_update() for update_reshape\n"); | |
8511 | ||
8512 | for (dl = super->devlist; dl; dl = dl->next) { | |
8513 | int size = sizeof_imsm_dev(dl->dev, 1); | |
8514 | void *s; | |
d677e0b8 AK |
8515 | if (u->new_raid_disks > u->old_raid_disks) |
8516 | size += sizeof(__u32)*2* | |
8517 | (u->new_raid_disks - u->old_raid_disks); | |
d195167d AK |
8518 | s = malloc(size); |
8519 | if (!s) | |
8520 | break; | |
8521 | *space_tail = s; | |
8522 | space_tail = s; | |
8523 | *space_tail = NULL; | |
8524 | } | |
8525 | ||
8526 | len = disks_to_mpb_size(u->new_raid_disks); | |
8527 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
78b10e66 N |
8528 | break; |
8529 | } | |
48c5303a | 8530 | case update_reshape_migration: { |
bc0b9d34 PC |
8531 | /* for migration level 0->5 we need to add disks |
8532 | * so the same as for container operation we will copy | |
8533 | * device to the bigger location. | |
8534 | * in memory prepared device and new disk area are prepared | |
8535 | * for usage in process update | |
8536 | */ | |
8537 | struct imsm_update_reshape_migration *u = (void *)update->buf; | |
8538 | struct intel_dev *id; | |
8539 | void **space_tail = (void **)&update->space_list; | |
8540 | int size; | |
8541 | void *s; | |
8542 | int current_level = -1; | |
8543 | ||
8544 | dprintf("imsm: imsm_prepare_update() for update_reshape\n"); | |
8545 | ||
8546 | /* add space for bigger array in update | |
8547 | */ | |
8548 | for (id = super->devlist; id; id = id->next) { | |
8549 | if (id->index == (unsigned)u->subdev) { | |
8550 | size = sizeof_imsm_dev(id->dev, 1); | |
8551 | if (u->new_raid_disks > u->old_raid_disks) | |
8552 | size += sizeof(__u32)*2* | |
8553 | (u->new_raid_disks - u->old_raid_disks); | |
8554 | s = malloc(size); | |
8555 | if (!s) | |
8556 | break; | |
8557 | *space_tail = s; | |
8558 | space_tail = s; | |
8559 | *space_tail = NULL; | |
8560 | break; | |
8561 | } | |
8562 | } | |
8563 | if (update->space_list == NULL) | |
8564 | break; | |
8565 | ||
8566 | /* add space for disk in update | |
8567 | */ | |
8568 | size = sizeof(struct dl); | |
8569 | s = malloc(size); | |
8570 | if (!s) { | |
8571 | free(update->space_list); | |
8572 | update->space_list = NULL; | |
8573 | break; | |
8574 | } | |
8575 | *space_tail = s; | |
8576 | space_tail = s; | |
8577 | *space_tail = NULL; | |
8578 | ||
8579 | /* add spare device to update | |
8580 | */ | |
8581 | for (id = super->devlist ; id; id = id->next) | |
8582 | if (id->index == (unsigned)u->subdev) { | |
8583 | struct imsm_dev *dev; | |
8584 | struct imsm_map *map; | |
8585 | ||
8586 | dev = get_imsm_dev(super, u->subdev); | |
238c0a71 | 8587 | map = get_imsm_map(dev, MAP_0); |
bc0b9d34 PC |
8588 | current_level = map->raid_level; |
8589 | break; | |
8590 | } | |
8591 | if ((u->new_level == 5) && (u->new_level != current_level)) { | |
8592 | struct mdinfo *spares; | |
8593 | ||
8594 | spares = get_spares_for_grow(st); | |
8595 | if (spares) { | |
8596 | struct dl *dl; | |
8597 | struct mdinfo *dev; | |
8598 | ||
8599 | dev = spares->devs; | |
8600 | if (dev) { | |
8601 | u->new_disks[0] = | |
8602 | makedev(dev->disk.major, | |
8603 | dev->disk.minor); | |
8604 | dl = get_disk_super(super, | |
8605 | dev->disk.major, | |
8606 | dev->disk.minor); | |
8607 | dl->index = u->old_raid_disks; | |
8608 | dev = dev->next; | |
8609 | } | |
8610 | sysfs_free(spares); | |
8611 | } | |
8612 | } | |
8613 | len = disks_to_mpb_size(u->new_raid_disks); | |
8614 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
48c5303a PC |
8615 | break; |
8616 | } | |
949c47a0 DW |
8617 | case update_create_array: { |
8618 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 8619 | struct intel_dev *dv; |
54c2c1ea | 8620 | struct imsm_dev *dev = &u->dev; |
238c0a71 | 8621 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
54c2c1ea DW |
8622 | struct dl *dl; |
8623 | struct disk_info *inf; | |
8624 | int i; | |
8625 | int activate = 0; | |
949c47a0 | 8626 | |
54c2c1ea DW |
8627 | inf = get_disk_info(u); |
8628 | len = sizeof_imsm_dev(dev, 1); | |
ba2de7ba DW |
8629 | /* allocate a new super->devlist entry */ |
8630 | dv = malloc(sizeof(*dv)); | |
8631 | if (dv) { | |
8632 | dv->dev = malloc(len); | |
8633 | if (dv->dev) | |
8634 | update->space = dv; | |
8635 | else { | |
8636 | free(dv); | |
8637 | update->space = NULL; | |
8638 | } | |
8639 | } | |
949c47a0 | 8640 | |
54c2c1ea DW |
8641 | /* count how many spares will be converted to members */ |
8642 | for (i = 0; i < map->num_members; i++) { | |
8643 | dl = serial_to_dl(inf[i].serial, super); | |
8644 | if (!dl) { | |
8645 | /* hmm maybe it failed?, nothing we can do about | |
8646 | * it here | |
8647 | */ | |
8648 | continue; | |
8649 | } | |
8650 | if (count_memberships(dl, super) == 0) | |
8651 | activate++; | |
8652 | } | |
8653 | len += activate * sizeof(struct imsm_disk); | |
949c47a0 DW |
8654 | break; |
8655 | default: | |
8656 | break; | |
8657 | } | |
8658 | } | |
8273f55e | 8659 | |
4d7b1503 DW |
8660 | /* check if we need a larger metadata buffer */ |
8661 | if (super->next_buf) | |
8662 | buf_len = super->next_len; | |
8663 | else | |
8664 | buf_len = super->len; | |
8665 | ||
8666 | if (__le32_to_cpu(mpb->mpb_size) + len > buf_len) { | |
8667 | /* ok we need a larger buf than what is currently allocated | |
8668 | * if this allocation fails process_update will notice that | |
8669 | * ->next_len is set and ->next_buf is NULL | |
8670 | */ | |
8671 | buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + len, 512); | |
8672 | if (super->next_buf) | |
8673 | free(super->next_buf); | |
8674 | ||
8675 | super->next_len = buf_len; | |
1f45a8ad DW |
8676 | if (posix_memalign(&super->next_buf, 512, buf_len) == 0) |
8677 | memset(super->next_buf, 0, buf_len); | |
8678 | else | |
4d7b1503 DW |
8679 | super->next_buf = NULL; |
8680 | } | |
8273f55e DW |
8681 | } |
8682 | ||
ae6aad82 | 8683 | /* must be called while manager is quiesced */ |
f21e18ca | 8684 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index) |
ae6aad82 DW |
8685 | { |
8686 | struct imsm_super *mpb = super->anchor; | |
ae6aad82 DW |
8687 | struct dl *iter; |
8688 | struct imsm_dev *dev; | |
8689 | struct imsm_map *map; | |
24565c9a DW |
8690 | int i, j, num_members; |
8691 | __u32 ord; | |
ae6aad82 | 8692 | |
24565c9a DW |
8693 | dprintf("%s: deleting device[%d] from imsm_super\n", |
8694 | __func__, index); | |
ae6aad82 DW |
8695 | |
8696 | /* shift all indexes down one */ | |
8697 | for (iter = super->disks; iter; iter = iter->next) | |
f21e18ca | 8698 | if (iter->index > (int)index) |
ae6aad82 | 8699 | iter->index--; |
47ee5a45 | 8700 | for (iter = super->missing; iter; iter = iter->next) |
f21e18ca | 8701 | if (iter->index > (int)index) |
47ee5a45 | 8702 | iter->index--; |
ae6aad82 DW |
8703 | |
8704 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
8705 | dev = get_imsm_dev(super, i); | |
238c0a71 | 8706 | map = get_imsm_map(dev, MAP_0); |
24565c9a DW |
8707 | num_members = map->num_members; |
8708 | for (j = 0; j < num_members; j++) { | |
8709 | /* update ord entries being careful not to propagate | |
8710 | * ord-flags to the first map | |
8711 | */ | |
238c0a71 | 8712 | ord = get_imsm_ord_tbl_ent(dev, j, MAP_X); |
ae6aad82 | 8713 | |
24565c9a DW |
8714 | if (ord_to_idx(ord) <= index) |
8715 | continue; | |
ae6aad82 | 8716 | |
238c0a71 | 8717 | map = get_imsm_map(dev, MAP_0); |
24565c9a | 8718 | set_imsm_ord_tbl_ent(map, j, ord_to_idx(ord - 1)); |
238c0a71 | 8719 | map = get_imsm_map(dev, MAP_1); |
24565c9a DW |
8720 | if (map) |
8721 | set_imsm_ord_tbl_ent(map, j, ord - 1); | |
ae6aad82 DW |
8722 | } |
8723 | } | |
8724 | ||
8725 | mpb->num_disks--; | |
8726 | super->updates_pending++; | |
24565c9a DW |
8727 | if (*dlp) { |
8728 | struct dl *dl = *dlp; | |
8729 | ||
8730 | *dlp = (*dlp)->next; | |
8731 | __free_imsm_disk(dl); | |
8732 | } | |
ae6aad82 | 8733 | } |
9e2d750d | 8734 | #endif /* MDASSEMBLE */ |
9a717282 AK |
8735 | |
8736 | static void close_targets(int *targets, int new_disks) | |
8737 | { | |
8738 | int i; | |
8739 | ||
8740 | if (!targets) | |
8741 | return; | |
8742 | ||
8743 | for (i = 0; i < new_disks; i++) { | |
8744 | if (targets[i] >= 0) { | |
8745 | close(targets[i]); | |
8746 | targets[i] = -1; | |
8747 | } | |
8748 | } | |
8749 | } | |
8750 | ||
8751 | static int imsm_get_allowed_degradation(int level, int raid_disks, | |
8752 | struct intel_super *super, | |
8753 | struct imsm_dev *dev) | |
8754 | { | |
8755 | switch (level) { | |
8756 | case 10:{ | |
8757 | int ret_val = 0; | |
8758 | struct imsm_map *map; | |
8759 | int i; | |
8760 | ||
8761 | ret_val = raid_disks/2; | |
8762 | /* check map if all disks pairs not failed | |
8763 | * in both maps | |
8764 | */ | |
238c0a71 | 8765 | map = get_imsm_map(dev, MAP_0); |
9a717282 AK |
8766 | for (i = 0; i < ret_val; i++) { |
8767 | int degradation = 0; | |
8768 | if (get_imsm_disk(super, i) == NULL) | |
8769 | degradation++; | |
8770 | if (get_imsm_disk(super, i + 1) == NULL) | |
8771 | degradation++; | |
8772 | if (degradation == 2) | |
8773 | return 0; | |
8774 | } | |
238c0a71 | 8775 | map = get_imsm_map(dev, MAP_1); |
9a717282 AK |
8776 | /* if there is no second map |
8777 | * result can be returned | |
8778 | */ | |
8779 | if (map == NULL) | |
8780 | return ret_val; | |
8781 | /* check degradation in second map | |
8782 | */ | |
8783 | for (i = 0; i < ret_val; i++) { | |
8784 | int degradation = 0; | |
8785 | if (get_imsm_disk(super, i) == NULL) | |
8786 | degradation++; | |
8787 | if (get_imsm_disk(super, i + 1) == NULL) | |
8788 | degradation++; | |
8789 | if (degradation == 2) | |
8790 | return 0; | |
8791 | } | |
8792 | return ret_val; | |
8793 | } | |
8794 | case 5: | |
8795 | return 1; | |
8796 | case 6: | |
8797 | return 2; | |
8798 | default: | |
8799 | return 0; | |
8800 | } | |
8801 | } | |
8802 | ||
8803 | ||
687629c2 AK |
8804 | /******************************************************************************* |
8805 | * Function: open_backup_targets | |
8806 | * Description: Function opens file descriptors for all devices given in | |
8807 | * info->devs | |
8808 | * Parameters: | |
8809 | * info : general array info | |
8810 | * raid_disks : number of disks | |
8811 | * raid_fds : table of device's file descriptors | |
9a717282 AK |
8812 | * super : intel super for raid10 degradation check |
8813 | * dev : intel device for raid10 degradation check | |
687629c2 AK |
8814 | * Returns: |
8815 | * 0 : success | |
8816 | * -1 : fail | |
8817 | ******************************************************************************/ | |
9a717282 AK |
8818 | int open_backup_targets(struct mdinfo *info, int raid_disks, int *raid_fds, |
8819 | struct intel_super *super, struct imsm_dev *dev) | |
687629c2 AK |
8820 | { |
8821 | struct mdinfo *sd; | |
f627f5ad | 8822 | int i; |
9a717282 | 8823 | int opened = 0; |
f627f5ad AK |
8824 | |
8825 | for (i = 0; i < raid_disks; i++) | |
8826 | raid_fds[i] = -1; | |
687629c2 AK |
8827 | |
8828 | for (sd = info->devs ; sd ; sd = sd->next) { | |
8829 | char *dn; | |
8830 | ||
8831 | if (sd->disk.state & (1<<MD_DISK_FAULTY)) { | |
8832 | dprintf("disk is faulty!!\n"); | |
8833 | continue; | |
8834 | } | |
8835 | ||
8836 | if ((sd->disk.raid_disk >= raid_disks) || | |
8837 | (sd->disk.raid_disk < 0)) | |
8838 | continue; | |
8839 | ||
8840 | dn = map_dev(sd->disk.major, | |
8841 | sd->disk.minor, 1); | |
8842 | raid_fds[sd->disk.raid_disk] = dev_open(dn, O_RDWR); | |
8843 | if (raid_fds[sd->disk.raid_disk] < 0) { | |
8844 | fprintf(stderr, "cannot open component\n"); | |
9a717282 | 8845 | continue; |
687629c2 | 8846 | } |
9a717282 AK |
8847 | opened++; |
8848 | } | |
8849 | /* check if maximum array degradation level is not exceeded | |
8850 | */ | |
8851 | if ((raid_disks - opened) > | |
8852 | imsm_get_allowed_degradation(info->new_level, | |
8853 | raid_disks, | |
8854 | super, dev)) { | |
8855 | fprintf(stderr, "Not enough disks can be opened.\n"); | |
8856 | close_targets(raid_fds, raid_disks); | |
8857 | return -2; | |
687629c2 AK |
8858 | } |
8859 | return 0; | |
8860 | } | |
8861 | ||
9e2d750d | 8862 | #ifndef MDASSEMBLE |
687629c2 AK |
8863 | /******************************************************************************* |
8864 | * Function: init_migr_record_imsm | |
8865 | * Description: Function inits imsm migration record | |
8866 | * Parameters: | |
8867 | * super : imsm internal array info | |
8868 | * dev : device under migration | |
8869 | * info : general array info to find the smallest device | |
8870 | * Returns: | |
8871 | * none | |
8872 | ******************************************************************************/ | |
8873 | void init_migr_record_imsm(struct supertype *st, struct imsm_dev *dev, | |
8874 | struct mdinfo *info) | |
8875 | { | |
8876 | struct intel_super *super = st->sb; | |
8877 | struct migr_record *migr_rec = super->migr_rec; | |
8878 | int new_data_disks; | |
8879 | unsigned long long dsize, dev_sectors; | |
8880 | long long unsigned min_dev_sectors = -1LLU; | |
8881 | struct mdinfo *sd; | |
8882 | char nm[30]; | |
8883 | int fd; | |
238c0a71 AK |
8884 | struct imsm_map *map_dest = get_imsm_map(dev, MAP_0); |
8885 | struct imsm_map *map_src = get_imsm_map(dev, MAP_1); | |
687629c2 | 8886 | unsigned long long num_migr_units; |
3ef4403c | 8887 | unsigned long long array_blocks; |
687629c2 AK |
8888 | |
8889 | memset(migr_rec, 0, sizeof(struct migr_record)); | |
8890 | migr_rec->family_num = __cpu_to_le32(super->anchor->family_num); | |
8891 | ||
8892 | /* only ascending reshape supported now */ | |
8893 | migr_rec->ascending_migr = __cpu_to_le32(1); | |
8894 | ||
8895 | migr_rec->dest_depth_per_unit = GEN_MIGR_AREA_SIZE / | |
8896 | max(map_dest->blocks_per_strip, map_src->blocks_per_strip); | |
8897 | migr_rec->dest_depth_per_unit *= map_dest->blocks_per_strip; | |
238c0a71 | 8898 | new_data_disks = imsm_num_data_members(dev, MAP_0); |
687629c2 AK |
8899 | migr_rec->blocks_per_unit = |
8900 | __cpu_to_le32(migr_rec->dest_depth_per_unit * new_data_disks); | |
8901 | migr_rec->dest_depth_per_unit = | |
8902 | __cpu_to_le32(migr_rec->dest_depth_per_unit); | |
3ef4403c | 8903 | array_blocks = info->component_size * new_data_disks; |
687629c2 AK |
8904 | num_migr_units = |
8905 | array_blocks / __le32_to_cpu(migr_rec->blocks_per_unit); | |
8906 | ||
8907 | if (array_blocks % __le32_to_cpu(migr_rec->blocks_per_unit)) | |
8908 | num_migr_units++; | |
8909 | migr_rec->num_migr_units = __cpu_to_le32(num_migr_units); | |
8910 | ||
8911 | migr_rec->post_migr_vol_cap = dev->size_low; | |
8912 | migr_rec->post_migr_vol_cap_hi = dev->size_high; | |
8913 | ||
8914 | ||
8915 | /* Find the smallest dev */ | |
8916 | for (sd = info->devs ; sd ; sd = sd->next) { | |
8917 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); | |
8918 | fd = dev_open(nm, O_RDONLY); | |
8919 | if (fd < 0) | |
8920 | continue; | |
8921 | get_dev_size(fd, NULL, &dsize); | |
8922 | dev_sectors = dsize / 512; | |
8923 | if (dev_sectors < min_dev_sectors) | |
8924 | min_dev_sectors = dev_sectors; | |
8925 | close(fd); | |
8926 | } | |
8927 | migr_rec->ckpt_area_pba = __cpu_to_le32(min_dev_sectors - | |
8928 | RAID_DISK_RESERVED_BLOCKS_IMSM_HI); | |
8929 | ||
8930 | write_imsm_migr_rec(st); | |
8931 | ||
8932 | return; | |
8933 | } | |
8934 | ||
8935 | /******************************************************************************* | |
8936 | * Function: save_backup_imsm | |
8937 | * Description: Function saves critical data stripes to Migration Copy Area | |
8938 | * and updates the current migration unit status. | |
8939 | * Use restore_stripes() to form a destination stripe, | |
8940 | * and to write it to the Copy Area. | |
8941 | * Parameters: | |
8942 | * st : supertype information | |
aea93171 | 8943 | * dev : imsm device that backup is saved for |
687629c2 AK |
8944 | * info : general array info |
8945 | * buf : input buffer | |
687629c2 AK |
8946 | * length : length of data to backup (blocks_per_unit) |
8947 | * Returns: | |
8948 | * 0 : success | |
8949 | *, -1 : fail | |
8950 | ******************************************************************************/ | |
8951 | int save_backup_imsm(struct supertype *st, | |
8952 | struct imsm_dev *dev, | |
8953 | struct mdinfo *info, | |
8954 | void *buf, | |
687629c2 AK |
8955 | int length) |
8956 | { | |
8957 | int rv = -1; | |
8958 | struct intel_super *super = st->sb; | |
8959 | unsigned long long *target_offsets = NULL; | |
8960 | int *targets = NULL; | |
8961 | int i; | |
238c0a71 | 8962 | struct imsm_map *map_dest = get_imsm_map(dev, MAP_0); |
687629c2 | 8963 | int new_disks = map_dest->num_members; |
ab724b98 AK |
8964 | int dest_layout = 0; |
8965 | int dest_chunk; | |
d1877f69 | 8966 | unsigned long long start; |
238c0a71 | 8967 | int data_disks = imsm_num_data_members(dev, MAP_0); |
687629c2 AK |
8968 | |
8969 | targets = malloc(new_disks * sizeof(int)); | |
8970 | if (!targets) | |
8971 | goto abort; | |
8972 | ||
7e45b550 AK |
8973 | for (i = 0; i < new_disks; i++) |
8974 | targets[i] = -1; | |
8975 | ||
687629c2 AK |
8976 | target_offsets = malloc(new_disks * sizeof(unsigned long long)); |
8977 | if (!target_offsets) | |
8978 | goto abort; | |
8979 | ||
d1877f69 | 8980 | start = info->reshape_progress * 512; |
687629c2 | 8981 | for (i = 0; i < new_disks; i++) { |
687629c2 AK |
8982 | target_offsets[i] = (unsigned long long) |
8983 | __le32_to_cpu(super->migr_rec->ckpt_area_pba) * 512; | |
d1877f69 AK |
8984 | /* move back copy area adderss, it will be moved forward |
8985 | * in restore_stripes() using start input variable | |
8986 | */ | |
8987 | target_offsets[i] -= start/data_disks; | |
687629c2 AK |
8988 | } |
8989 | ||
9a717282 AK |
8990 | if (open_backup_targets(info, new_disks, targets, |
8991 | super, dev)) | |
687629c2 AK |
8992 | goto abort; |
8993 | ||
68eb8bc6 | 8994 | dest_layout = imsm_level_to_layout(map_dest->raid_level); |
ab724b98 AK |
8995 | dest_chunk = __le16_to_cpu(map_dest->blocks_per_strip) * 512; |
8996 | ||
687629c2 AK |
8997 | if (restore_stripes(targets, /* list of dest devices */ |
8998 | target_offsets, /* migration record offsets */ | |
8999 | new_disks, | |
ab724b98 AK |
9000 | dest_chunk, |
9001 | map_dest->raid_level, | |
9002 | dest_layout, | |
9003 | -1, /* source backup file descriptor */ | |
9004 | 0, /* input buf offset | |
9005 | * always 0 buf is already offseted */ | |
d1877f69 | 9006 | start, |
687629c2 AK |
9007 | length, |
9008 | buf) != 0) { | |
9009 | fprintf(stderr, Name ": Error restoring stripes\n"); | |
9010 | goto abort; | |
9011 | } | |
9012 | ||
9013 | rv = 0; | |
9014 | ||
9015 | abort: | |
9016 | if (targets) { | |
9a717282 | 9017 | close_targets(targets, new_disks); |
687629c2 AK |
9018 | free(targets); |
9019 | } | |
9020 | free(target_offsets); | |
9021 | ||
9022 | return rv; | |
9023 | } | |
9024 | ||
9025 | /******************************************************************************* | |
9026 | * Function: save_checkpoint_imsm | |
9027 | * Description: Function called for current unit status update | |
9028 | * in the migration record. It writes it to disk. | |
9029 | * Parameters: | |
9030 | * super : imsm internal array info | |
9031 | * info : general array info | |
9032 | * Returns: | |
9033 | * 0: success | |
9034 | * 1: failure | |
0228d92c AK |
9035 | * 2: failure, means no valid migration record |
9036 | * / no general migration in progress / | |
687629c2 AK |
9037 | ******************************************************************************/ |
9038 | int save_checkpoint_imsm(struct supertype *st, struct mdinfo *info, int state) | |
9039 | { | |
9040 | struct intel_super *super = st->sb; | |
f8b72ef5 AK |
9041 | unsigned long long blocks_per_unit; |
9042 | unsigned long long curr_migr_unit; | |
9043 | ||
2e062e82 AK |
9044 | if (load_imsm_migr_rec(super, info) != 0) { |
9045 | dprintf("imsm: ERROR: Cannot read migration record " | |
9046 | "for checkpoint save.\n"); | |
9047 | return 1; | |
9048 | } | |
9049 | ||
f8b72ef5 AK |
9050 | blocks_per_unit = __le32_to_cpu(super->migr_rec->blocks_per_unit); |
9051 | if (blocks_per_unit == 0) { | |
0228d92c AK |
9052 | dprintf("imsm: no migration in progress.\n"); |
9053 | return 2; | |
687629c2 | 9054 | } |
f8b72ef5 AK |
9055 | curr_migr_unit = info->reshape_progress / blocks_per_unit; |
9056 | /* check if array is alligned to copy area | |
9057 | * if it is not alligned, add one to current migration unit value | |
9058 | * this can happend on array reshape finish only | |
9059 | */ | |
9060 | if (info->reshape_progress % blocks_per_unit) | |
9061 | curr_migr_unit++; | |
687629c2 AK |
9062 | |
9063 | super->migr_rec->curr_migr_unit = | |
f8b72ef5 | 9064 | __cpu_to_le32(curr_migr_unit); |
687629c2 AK |
9065 | super->migr_rec->rec_status = __cpu_to_le32(state); |
9066 | super->migr_rec->dest_1st_member_lba = | |
f8b72ef5 AK |
9067 | __cpu_to_le32(curr_migr_unit * |
9068 | __le32_to_cpu(super->migr_rec->dest_depth_per_unit)); | |
687629c2 AK |
9069 | if (write_imsm_migr_rec(st) < 0) { |
9070 | dprintf("imsm: Cannot write migration record " | |
9071 | "outside backup area\n"); | |
9072 | return 1; | |
9073 | } | |
9074 | ||
9075 | return 0; | |
9076 | } | |
9077 | ||
276d77db AK |
9078 | /******************************************************************************* |
9079 | * Function: recover_backup_imsm | |
9080 | * Description: Function recovers critical data from the Migration Copy Area | |
9081 | * while assembling an array. | |
9082 | * Parameters: | |
9083 | * super : imsm internal array info | |
9084 | * info : general array info | |
9085 | * Returns: | |
9086 | * 0 : success (or there is no data to recover) | |
9087 | * 1 : fail | |
9088 | ******************************************************************************/ | |
9089 | int recover_backup_imsm(struct supertype *st, struct mdinfo *info) | |
9090 | { | |
9091 | struct intel_super *super = st->sb; | |
9092 | struct migr_record *migr_rec = super->migr_rec; | |
9093 | struct imsm_map *map_dest = NULL; | |
9094 | struct intel_dev *id = NULL; | |
9095 | unsigned long long read_offset; | |
9096 | unsigned long long write_offset; | |
9097 | unsigned unit_len; | |
9098 | int *targets = NULL; | |
9099 | int new_disks, i, err; | |
9100 | char *buf = NULL; | |
9101 | int retval = 1; | |
9102 | unsigned long curr_migr_unit = __le32_to_cpu(migr_rec->curr_migr_unit); | |
9103 | unsigned long num_migr_units = __le32_to_cpu(migr_rec->num_migr_units); | |
276d77db | 9104 | char buffer[20]; |
6c3560c0 | 9105 | int skipped_disks = 0; |
276d77db AK |
9106 | |
9107 | err = sysfs_get_str(info, NULL, "array_state", (char *)buffer, 20); | |
9108 | if (err < 1) | |
9109 | return 1; | |
9110 | ||
9111 | /* recover data only during assemblation */ | |
9112 | if (strncmp(buffer, "inactive", 8) != 0) | |
9113 | return 0; | |
9114 | /* no data to recover */ | |
9115 | if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL) | |
9116 | return 0; | |
9117 | if (curr_migr_unit >= num_migr_units) | |
9118 | return 1; | |
9119 | ||
9120 | /* find device during reshape */ | |
9121 | for (id = super->devlist; id; id = id->next) | |
9122 | if (is_gen_migration(id->dev)) | |
9123 | break; | |
9124 | if (id == NULL) | |
9125 | return 1; | |
9126 | ||
238c0a71 | 9127 | map_dest = get_imsm_map(id->dev, MAP_0); |
276d77db AK |
9128 | new_disks = map_dest->num_members; |
9129 | ||
9130 | read_offset = (unsigned long long) | |
9131 | __le32_to_cpu(migr_rec->ckpt_area_pba) * 512; | |
9132 | ||
9133 | write_offset = ((unsigned long long) | |
9134 | __le32_to_cpu(migr_rec->dest_1st_member_lba) + | |
75b69ea4 | 9135 | __le32_to_cpu(map_dest->pba_of_lba0)) * 512; |
276d77db AK |
9136 | |
9137 | unit_len = __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
9138 | if (posix_memalign((void **)&buf, 512, unit_len) != 0) | |
9139 | goto abort; | |
9140 | targets = malloc(new_disks * sizeof(int)); | |
9141 | if (!targets) | |
9142 | goto abort; | |
9143 | ||
9a717282 | 9144 | if (open_backup_targets(info, new_disks, targets, super, id->dev)) { |
f627f5ad AK |
9145 | fprintf(stderr, |
9146 | Name ": Cannot open some devices belonging to array.\n"); | |
9147 | goto abort; | |
9148 | } | |
276d77db AK |
9149 | |
9150 | for (i = 0; i < new_disks; i++) { | |
6c3560c0 AK |
9151 | if (targets[i] < 0) { |
9152 | skipped_disks++; | |
9153 | continue; | |
9154 | } | |
276d77db AK |
9155 | if (lseek64(targets[i], read_offset, SEEK_SET) < 0) { |
9156 | fprintf(stderr, | |
9157 | Name ": Cannot seek to block: %s\n", | |
9158 | strerror(errno)); | |
137debce AK |
9159 | skipped_disks++; |
9160 | continue; | |
276d77db | 9161 | } |
9ec11d1a | 9162 | if ((unsigned)read(targets[i], buf, unit_len) != unit_len) { |
276d77db AK |
9163 | fprintf(stderr, |
9164 | Name ": Cannot read copy area block: %s\n", | |
9165 | strerror(errno)); | |
137debce AK |
9166 | skipped_disks++; |
9167 | continue; | |
276d77db AK |
9168 | } |
9169 | if (lseek64(targets[i], write_offset, SEEK_SET) < 0) { | |
9170 | fprintf(stderr, | |
9171 | Name ": Cannot seek to block: %s\n", | |
9172 | strerror(errno)); | |
137debce AK |
9173 | skipped_disks++; |
9174 | continue; | |
276d77db | 9175 | } |
9ec11d1a | 9176 | if ((unsigned)write(targets[i], buf, unit_len) != unit_len) { |
276d77db AK |
9177 | fprintf(stderr, |
9178 | Name ": Cannot restore block: %s\n", | |
9179 | strerror(errno)); | |
137debce AK |
9180 | skipped_disks++; |
9181 | continue; | |
276d77db AK |
9182 | } |
9183 | } | |
9184 | ||
137debce AK |
9185 | if (skipped_disks > imsm_get_allowed_degradation(info->new_level, |
9186 | new_disks, | |
9187 | super, | |
9188 | id->dev)) { | |
6c3560c0 AK |
9189 | fprintf(stderr, |
9190 | Name ": Cannot restore data from backup." | |
9191 | " Too many failed disks\n"); | |
9192 | goto abort; | |
9193 | } | |
9194 | ||
befb629b AK |
9195 | if (save_checkpoint_imsm(st, info, UNIT_SRC_NORMAL)) { |
9196 | /* ignore error == 2, this can mean end of reshape here | |
9197 | */ | |
9198 | dprintf("imsm: Cannot write checkpoint to " | |
9199 | "migration record (UNIT_SRC_NORMAL) during restart\n"); | |
9200 | } else | |
276d77db | 9201 | retval = 0; |
276d77db AK |
9202 | |
9203 | abort: | |
9204 | if (targets) { | |
9205 | for (i = 0; i < new_disks; i++) | |
9206 | if (targets[i]) | |
9207 | close(targets[i]); | |
9208 | free(targets); | |
9209 | } | |
9210 | free(buf); | |
9211 | return retval; | |
9212 | } | |
9213 | ||
2cda7640 ML |
9214 | static char disk_by_path[] = "/dev/disk/by-path/"; |
9215 | ||
9216 | static const char *imsm_get_disk_controller_domain(const char *path) | |
9217 | { | |
2cda7640 | 9218 | char disk_path[PATH_MAX]; |
96234762 LM |
9219 | char *drv=NULL; |
9220 | struct stat st; | |
2cda7640 | 9221 | |
96234762 LM |
9222 | strncpy(disk_path, disk_by_path, PATH_MAX - 1); |
9223 | strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1); | |
9224 | if (stat(disk_path, &st) == 0) { | |
9225 | struct sys_dev* hba; | |
9226 | char *path=NULL; | |
9227 | ||
9228 | path = devt_to_devpath(st.st_rdev); | |
9229 | if (path == NULL) | |
9230 | return "unknown"; | |
9231 | hba = find_disk_attached_hba(-1, path); | |
9232 | if (hba && hba->type == SYS_DEV_SAS) | |
9233 | drv = "isci"; | |
9234 | else if (hba && hba->type == SYS_DEV_SATA) | |
9235 | drv = "ahci"; | |
9236 | else | |
9237 | drv = "unknown"; | |
9238 | dprintf("path: %s hba: %s attached: %s\n", | |
9239 | path, (hba) ? hba->path : "NULL", drv); | |
9240 | free(path); | |
9241 | if (hba) | |
9242 | free_sys_dev(&hba); | |
2cda7640 | 9243 | } |
96234762 | 9244 | return drv; |
2cda7640 ML |
9245 | } |
9246 | ||
78b10e66 N |
9247 | static int imsm_find_array_minor_by_subdev(int subdev, int container, int *minor) |
9248 | { | |
9249 | char subdev_name[20]; | |
9250 | struct mdstat_ent *mdstat; | |
9251 | ||
9252 | sprintf(subdev_name, "%d", subdev); | |
9253 | mdstat = mdstat_by_subdev(subdev_name, container); | |
9254 | if (!mdstat) | |
9255 | return -1; | |
9256 | ||
9257 | *minor = mdstat->devnum; | |
9258 | free_mdstat(mdstat); | |
9259 | return 0; | |
9260 | } | |
9261 | ||
9262 | static int imsm_reshape_is_allowed_on_container(struct supertype *st, | |
9263 | struct geo_params *geo, | |
9264 | int *old_raid_disks) | |
9265 | { | |
694575e7 KW |
9266 | /* currently we only support increasing the number of devices |
9267 | * for a container. This increases the number of device for each | |
9268 | * member array. They must all be RAID0 or RAID5. | |
9269 | */ | |
78b10e66 N |
9270 | int ret_val = 0; |
9271 | struct mdinfo *info, *member; | |
9272 | int devices_that_can_grow = 0; | |
9273 | ||
9274 | dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): " | |
9275 | "st->devnum = (%i)\n", | |
9276 | st->devnum); | |
9277 | ||
9278 | if (geo->size != -1 || | |
9279 | geo->level != UnSet || | |
9280 | geo->layout != UnSet || | |
9281 | geo->chunksize != 0 || | |
9282 | geo->raid_disks == UnSet) { | |
9283 | dprintf("imsm: Container operation is allowed for " | |
9284 | "raid disks number change only.\n"); | |
9285 | return ret_val; | |
9286 | } | |
9287 | ||
9288 | info = container_content_imsm(st, NULL); | |
9289 | for (member = info; member; member = member->next) { | |
9290 | int result; | |
9291 | int minor; | |
9292 | ||
9293 | dprintf("imsm: checking device_num: %i\n", | |
9294 | member->container_member); | |
9295 | ||
d7d205bd | 9296 | if (geo->raid_disks <= member->array.raid_disks) { |
78b10e66 N |
9297 | /* we work on container for Online Capacity Expansion |
9298 | * only so raid_disks has to grow | |
9299 | */ | |
9300 | dprintf("imsm: for container operation raid disks " | |
9301 | "increase is required\n"); | |
9302 | break; | |
9303 | } | |
9304 | ||
9305 | if ((info->array.level != 0) && | |
9306 | (info->array.level != 5)) { | |
9307 | /* we cannot use this container with other raid level | |
9308 | */ | |
690aae1a | 9309 | dprintf("imsm: for container operation wrong" |
78b10e66 N |
9310 | " raid level (%i) detected\n", |
9311 | info->array.level); | |
9312 | break; | |
9313 | } else { | |
9314 | /* check for platform support | |
9315 | * for this raid level configuration | |
9316 | */ | |
9317 | struct intel_super *super = st->sb; | |
9318 | if (!is_raid_level_supported(super->orom, | |
9319 | member->array.level, | |
9320 | geo->raid_disks)) { | |
690aae1a | 9321 | dprintf("platform does not support raid%d with" |
78b10e66 N |
9322 | " %d disk%s\n", |
9323 | info->array.level, | |
9324 | geo->raid_disks, | |
9325 | geo->raid_disks > 1 ? "s" : ""); | |
9326 | break; | |
9327 | } | |
2a4a08e7 AK |
9328 | /* check if component size is aligned to chunk size |
9329 | */ | |
9330 | if (info->component_size % | |
9331 | (info->array.chunk_size/512)) { | |
9332 | dprintf("Component size is not aligned to " | |
9333 | "chunk size\n"); | |
9334 | break; | |
9335 | } | |
78b10e66 N |
9336 | } |
9337 | ||
9338 | if (*old_raid_disks && | |
9339 | info->array.raid_disks != *old_raid_disks) | |
9340 | break; | |
9341 | *old_raid_disks = info->array.raid_disks; | |
9342 | ||
9343 | /* All raid5 and raid0 volumes in container | |
9344 | * have to be ready for Online Capacity Expansion | |
9345 | * so they need to be assembled. We have already | |
9346 | * checked that no recovery etc is happening. | |
9347 | */ | |
9348 | result = imsm_find_array_minor_by_subdev(member->container_member, | |
9349 | st->container_dev, | |
9350 | &minor); | |
9351 | if (result < 0) { | |
9352 | dprintf("imsm: cannot find array\n"); | |
9353 | break; | |
9354 | } | |
9355 | devices_that_can_grow++; | |
9356 | } | |
9357 | sysfs_free(info); | |
9358 | if (!member && devices_that_can_grow) | |
9359 | ret_val = 1; | |
9360 | ||
9361 | if (ret_val) | |
9362 | dprintf("\tContainer operation allowed\n"); | |
9363 | else | |
9364 | dprintf("\tError: %i\n", ret_val); | |
9365 | ||
9366 | return ret_val; | |
9367 | } | |
9368 | ||
9369 | /* Function: get_spares_for_grow | |
9370 | * Description: Allocates memory and creates list of spare devices | |
9371 | * avaliable in container. Checks if spare drive size is acceptable. | |
9372 | * Parameters: Pointer to the supertype structure | |
9373 | * Returns: Pointer to the list of spare devices (mdinfo structure) on success, | |
9374 | * NULL if fail | |
9375 | */ | |
9376 | static struct mdinfo *get_spares_for_grow(struct supertype *st) | |
9377 | { | |
78b10e66 | 9378 | unsigned long long min_size = min_acceptable_spare_size_imsm(st); |
326727d9 | 9379 | return container_choose_spares(st, min_size, NULL, NULL, NULL, 0); |
78b10e66 N |
9380 | } |
9381 | ||
9382 | /****************************************************************************** | |
9383 | * function: imsm_create_metadata_update_for_reshape | |
9384 | * Function creates update for whole IMSM container. | |
9385 | * | |
9386 | ******************************************************************************/ | |
9387 | static int imsm_create_metadata_update_for_reshape( | |
9388 | struct supertype *st, | |
9389 | struct geo_params *geo, | |
9390 | int old_raid_disks, | |
9391 | struct imsm_update_reshape **updatep) | |
9392 | { | |
9393 | struct intel_super *super = st->sb; | |
9394 | struct imsm_super *mpb = super->anchor; | |
9395 | int update_memory_size = 0; | |
9396 | struct imsm_update_reshape *u = NULL; | |
9397 | struct mdinfo *spares = NULL; | |
9398 | int i; | |
9399 | int delta_disks = 0; | |
bbd24d86 | 9400 | struct mdinfo *dev; |
78b10e66 N |
9401 | |
9402 | dprintf("imsm_update_metadata_for_reshape(enter) raid_disks = %i\n", | |
9403 | geo->raid_disks); | |
9404 | ||
9405 | delta_disks = geo->raid_disks - old_raid_disks; | |
9406 | ||
9407 | /* size of all update data without anchor */ | |
9408 | update_memory_size = sizeof(struct imsm_update_reshape); | |
9409 | ||
9410 | /* now add space for spare disks that we need to add. */ | |
9411 | update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1); | |
9412 | ||
9413 | u = calloc(1, update_memory_size); | |
9414 | if (u == NULL) { | |
9415 | dprintf("error: " | |
9416 | "cannot get memory for imsm_update_reshape update\n"); | |
9417 | return 0; | |
9418 | } | |
9419 | u->type = update_reshape_container_disks; | |
9420 | u->old_raid_disks = old_raid_disks; | |
9421 | u->new_raid_disks = geo->raid_disks; | |
9422 | ||
9423 | /* now get spare disks list | |
9424 | */ | |
9425 | spares = get_spares_for_grow(st); | |
9426 | ||
9427 | if (spares == NULL | |
9428 | || delta_disks > spares->array.spare_disks) { | |
e14e5960 KW |
9429 | fprintf(stderr, Name ": imsm: ERROR: Cannot get spare devices " |
9430 | "for %s.\n", geo->dev_name); | |
e4c72d1d | 9431 | i = -1; |
78b10e66 N |
9432 | goto abort; |
9433 | } | |
9434 | ||
9435 | /* we have got spares | |
9436 | * update disk list in imsm_disk list table in anchor | |
9437 | */ | |
9438 | dprintf("imsm: %i spares are available.\n\n", | |
9439 | spares->array.spare_disks); | |
9440 | ||
bbd24d86 | 9441 | dev = spares->devs; |
78b10e66 | 9442 | for (i = 0; i < delta_disks; i++) { |
78b10e66 N |
9443 | struct dl *dl; |
9444 | ||
bbd24d86 AK |
9445 | if (dev == NULL) |
9446 | break; | |
78b10e66 N |
9447 | u->new_disks[i] = makedev(dev->disk.major, |
9448 | dev->disk.minor); | |
9449 | dl = get_disk_super(super, dev->disk.major, dev->disk.minor); | |
ee4beede AK |
9450 | dl->index = mpb->num_disks; |
9451 | mpb->num_disks++; | |
bbd24d86 | 9452 | dev = dev->next; |
78b10e66 | 9453 | } |
78b10e66 N |
9454 | |
9455 | abort: | |
9456 | /* free spares | |
9457 | */ | |
9458 | sysfs_free(spares); | |
9459 | ||
d677e0b8 | 9460 | dprintf("imsm: reshape update preparation :"); |
78b10e66 | 9461 | if (i == delta_disks) { |
d677e0b8 | 9462 | dprintf(" OK\n"); |
78b10e66 N |
9463 | *updatep = u; |
9464 | return update_memory_size; | |
9465 | } | |
9466 | free(u); | |
d677e0b8 | 9467 | dprintf(" Error\n"); |
78b10e66 N |
9468 | |
9469 | return 0; | |
9470 | } | |
9471 | ||
48c5303a PC |
9472 | /****************************************************************************** |
9473 | * function: imsm_create_metadata_update_for_migration() | |
9474 | * Creates update for IMSM array. | |
9475 | * | |
9476 | ******************************************************************************/ | |
9477 | static int imsm_create_metadata_update_for_migration( | |
9478 | struct supertype *st, | |
9479 | struct geo_params *geo, | |
9480 | struct imsm_update_reshape_migration **updatep) | |
9481 | { | |
9482 | struct intel_super *super = st->sb; | |
9483 | int update_memory_size = 0; | |
9484 | struct imsm_update_reshape_migration *u = NULL; | |
9485 | struct imsm_dev *dev; | |
9486 | int previous_level = -1; | |
9487 | ||
9488 | dprintf("imsm_create_metadata_update_for_migration(enter)" | |
9489 | " New Level = %i\n", geo->level); | |
9490 | ||
9491 | /* size of all update data without anchor */ | |
9492 | update_memory_size = sizeof(struct imsm_update_reshape_migration); | |
9493 | ||
9494 | u = calloc(1, update_memory_size); | |
9495 | if (u == NULL) { | |
9496 | dprintf("error: cannot get memory for " | |
9497 | "imsm_create_metadata_update_for_migration\n"); | |
9498 | return 0; | |
9499 | } | |
9500 | u->type = update_reshape_migration; | |
9501 | u->subdev = super->current_vol; | |
9502 | u->new_level = geo->level; | |
9503 | u->new_layout = geo->layout; | |
9504 | u->new_raid_disks = u->old_raid_disks = geo->raid_disks; | |
9505 | u->new_disks[0] = -1; | |
4bba0439 | 9506 | u->new_chunksize = -1; |
48c5303a PC |
9507 | |
9508 | dev = get_imsm_dev(super, u->subdev); | |
9509 | if (dev) { | |
9510 | struct imsm_map *map; | |
9511 | ||
238c0a71 | 9512 | map = get_imsm_map(dev, MAP_0); |
4bba0439 PC |
9513 | if (map) { |
9514 | int current_chunk_size = | |
9515 | __le16_to_cpu(map->blocks_per_strip) / 2; | |
9516 | ||
9517 | if (geo->chunksize != current_chunk_size) { | |
9518 | u->new_chunksize = geo->chunksize / 1024; | |
9519 | dprintf("imsm: " | |
9520 | "chunk size change from %i to %i\n", | |
9521 | current_chunk_size, u->new_chunksize); | |
9522 | } | |
48c5303a | 9523 | previous_level = map->raid_level; |
4bba0439 | 9524 | } |
48c5303a PC |
9525 | } |
9526 | if ((geo->level == 5) && (previous_level == 0)) { | |
9527 | struct mdinfo *spares = NULL; | |
9528 | ||
9529 | u->new_raid_disks++; | |
9530 | spares = get_spares_for_grow(st); | |
9531 | if ((spares == NULL) || (spares->array.spare_disks < 1)) { | |
9532 | free(u); | |
9533 | sysfs_free(spares); | |
9534 | update_memory_size = 0; | |
9535 | dprintf("error: cannot get spare device " | |
9536 | "for requested migration"); | |
9537 | return 0; | |
9538 | } | |
9539 | sysfs_free(spares); | |
9540 | } | |
9541 | dprintf("imsm: reshape update preparation : OK\n"); | |
9542 | *updatep = u; | |
9543 | ||
9544 | return update_memory_size; | |
9545 | } | |
9546 | ||
8dd70bce AK |
9547 | static void imsm_update_metadata_locally(struct supertype *st, |
9548 | void *buf, int len) | |
9549 | { | |
9550 | struct metadata_update mu; | |
9551 | ||
9552 | mu.buf = buf; | |
9553 | mu.len = len; | |
9554 | mu.space = NULL; | |
9555 | mu.space_list = NULL; | |
9556 | mu.next = NULL; | |
9557 | imsm_prepare_update(st, &mu); | |
9558 | imsm_process_update(st, &mu); | |
9559 | ||
9560 | while (mu.space_list) { | |
9561 | void **space = mu.space_list; | |
9562 | mu.space_list = *space; | |
9563 | free(space); | |
9564 | } | |
9565 | } | |
78b10e66 | 9566 | |
471bceb6 | 9567 | /*************************************************************************** |
694575e7 | 9568 | * Function: imsm_analyze_change |
471bceb6 KW |
9569 | * Description: Function analyze change for single volume |
9570 | * and validate if transition is supported | |
694575e7 KW |
9571 | * Parameters: Geometry parameters, supertype structure |
9572 | * Returns: Operation type code on success, -1 if fail | |
471bceb6 KW |
9573 | ****************************************************************************/ |
9574 | enum imsm_reshape_type imsm_analyze_change(struct supertype *st, | |
9575 | struct geo_params *geo) | |
694575e7 | 9576 | { |
471bceb6 KW |
9577 | struct mdinfo info; |
9578 | int change = -1; | |
9579 | int check_devs = 0; | |
c21e737b | 9580 | int chunk; |
67a2db32 AK |
9581 | /* number of added/removed disks in operation result */ |
9582 | int devNumChange = 0; | |
9583 | /* imsm compatible layout value for array geometry verification */ | |
9584 | int imsm_layout = -1; | |
471bceb6 KW |
9585 | |
9586 | getinfo_super_imsm_volume(st, &info, NULL); | |
471bceb6 KW |
9587 | if ((geo->level != info.array.level) && |
9588 | (geo->level >= 0) && | |
9589 | (geo->level != UnSet)) { | |
9590 | switch (info.array.level) { | |
9591 | case 0: | |
9592 | if (geo->level == 5) { | |
b5347799 | 9593 | change = CH_MIGRATION; |
e13ce846 AK |
9594 | if (geo->layout != ALGORITHM_LEFT_ASYMMETRIC) { |
9595 | fprintf(stderr, | |
9596 | Name " Error. Requested Layout " | |
9597 | "not supported (left-asymmetric layout " | |
9598 | "is supported only)!\n"); | |
9599 | change = -1; | |
9600 | goto analyse_change_exit; | |
9601 | } | |
67a2db32 | 9602 | imsm_layout = geo->layout; |
471bceb6 | 9603 | check_devs = 1; |
e91a3bad LM |
9604 | devNumChange = 1; /* parity disk added */ |
9605 | } else if (geo->level == 10) { | |
471bceb6 KW |
9606 | change = CH_TAKEOVER; |
9607 | check_devs = 1; | |
e91a3bad | 9608 | devNumChange = 2; /* two mirrors added */ |
67a2db32 | 9609 | imsm_layout = 0x102; /* imsm supported layout */ |
471bceb6 | 9610 | } |
dfe77a9e KW |
9611 | break; |
9612 | case 1: | |
471bceb6 KW |
9613 | case 10: |
9614 | if (geo->level == 0) { | |
9615 | change = CH_TAKEOVER; | |
9616 | check_devs = 1; | |
e91a3bad | 9617 | devNumChange = -(geo->raid_disks/2); |
67a2db32 | 9618 | imsm_layout = 0; /* imsm raid0 layout */ |
471bceb6 KW |
9619 | } |
9620 | break; | |
9621 | } | |
9622 | if (change == -1) { | |
9623 | fprintf(stderr, | |
9624 | Name " Error. Level Migration from %d to %d " | |
9625 | "not supported!\n", | |
9626 | info.array.level, geo->level); | |
9627 | goto analyse_change_exit; | |
9628 | } | |
9629 | } else | |
9630 | geo->level = info.array.level; | |
9631 | ||
9632 | if ((geo->layout != info.array.layout) | |
9633 | && ((geo->layout != UnSet) && (geo->layout != -1))) { | |
b5347799 | 9634 | change = CH_MIGRATION; |
471bceb6 KW |
9635 | if ((info.array.layout == 0) |
9636 | && (info.array.level == 5) | |
9637 | && (geo->layout == 5)) { | |
9638 | /* reshape 5 -> 4 */ | |
9639 | } else if ((info.array.layout == 5) | |
9640 | && (info.array.level == 5) | |
9641 | && (geo->layout == 0)) { | |
9642 | /* reshape 4 -> 5 */ | |
9643 | geo->layout = 0; | |
9644 | geo->level = 5; | |
9645 | } else { | |
9646 | fprintf(stderr, | |
9647 | Name " Error. Layout Migration from %d to %d " | |
9648 | "not supported!\n", | |
9649 | info.array.layout, geo->layout); | |
9650 | change = -1; | |
9651 | goto analyse_change_exit; | |
9652 | } | |
67a2db32 | 9653 | } else { |
471bceb6 | 9654 | geo->layout = info.array.layout; |
67a2db32 AK |
9655 | if (imsm_layout == -1) |
9656 | imsm_layout = info.array.layout; | |
9657 | } | |
471bceb6 KW |
9658 | |
9659 | if ((geo->chunksize > 0) && (geo->chunksize != UnSet) | |
9660 | && (geo->chunksize != info.array.chunk_size)) | |
b5347799 | 9661 | change = CH_MIGRATION; |
471bceb6 KW |
9662 | else |
9663 | geo->chunksize = info.array.chunk_size; | |
9664 | ||
c21e737b | 9665 | chunk = geo->chunksize / 1024; |
471bceb6 KW |
9666 | if (!validate_geometry_imsm(st, |
9667 | geo->level, | |
67a2db32 | 9668 | imsm_layout, |
e91a3bad | 9669 | geo->raid_disks + devNumChange, |
c21e737b | 9670 | &chunk, |
471bceb6 KW |
9671 | geo->size, |
9672 | 0, 0, 1)) | |
9673 | change = -1; | |
9674 | ||
9675 | if (check_devs) { | |
9676 | struct intel_super *super = st->sb; | |
9677 | struct imsm_super *mpb = super->anchor; | |
9678 | ||
9679 | if (mpb->num_raid_devs > 1) { | |
9680 | fprintf(stderr, | |
9681 | Name " Error. Cannot perform operation on %s" | |
9682 | "- for this operation it MUST be single " | |
9683 | "array in container\n", | |
9684 | geo->dev_name); | |
9685 | change = -1; | |
9686 | } | |
9687 | } | |
9688 | ||
9689 | analyse_change_exit: | |
9690 | ||
9691 | return change; | |
694575e7 KW |
9692 | } |
9693 | ||
bb025c2f KW |
9694 | int imsm_takeover(struct supertype *st, struct geo_params *geo) |
9695 | { | |
9696 | struct intel_super *super = st->sb; | |
9697 | struct imsm_update_takeover *u; | |
9698 | ||
9699 | u = malloc(sizeof(struct imsm_update_takeover)); | |
9700 | if (u == NULL) | |
9701 | return 1; | |
9702 | ||
9703 | u->type = update_takeover; | |
9704 | u->subarray = super->current_vol; | |
9705 | ||
9706 | /* 10->0 transition */ | |
9707 | if (geo->level == 0) | |
9708 | u->direction = R10_TO_R0; | |
9709 | ||
0529c688 KW |
9710 | /* 0->10 transition */ |
9711 | if (geo->level == 10) | |
9712 | u->direction = R0_TO_R10; | |
9713 | ||
bb025c2f KW |
9714 | /* update metadata locally */ |
9715 | imsm_update_metadata_locally(st, u, | |
9716 | sizeof(struct imsm_update_takeover)); | |
9717 | /* and possibly remotely */ | |
9718 | if (st->update_tail) | |
9719 | append_metadata_update(st, u, | |
9720 | sizeof(struct imsm_update_takeover)); | |
9721 | else | |
9722 | free(u); | |
9723 | ||
9724 | return 0; | |
9725 | } | |
9726 | ||
78b10e66 N |
9727 | static int imsm_reshape_super(struct supertype *st, long long size, int level, |
9728 | int layout, int chunksize, int raid_disks, | |
41784c88 AK |
9729 | int delta_disks, char *backup, char *dev, |
9730 | int verbose) | |
78b10e66 | 9731 | { |
78b10e66 N |
9732 | int ret_val = 1; |
9733 | struct geo_params geo; | |
9734 | ||
9735 | dprintf("imsm: reshape_super called.\n"); | |
9736 | ||
71204a50 | 9737 | memset(&geo, 0, sizeof(struct geo_params)); |
78b10e66 N |
9738 | |
9739 | geo.dev_name = dev; | |
694575e7 | 9740 | geo.dev_id = st->devnum; |
78b10e66 N |
9741 | geo.size = size; |
9742 | geo.level = level; | |
9743 | geo.layout = layout; | |
9744 | geo.chunksize = chunksize; | |
9745 | geo.raid_disks = raid_disks; | |
41784c88 AK |
9746 | if (delta_disks != UnSet) |
9747 | geo.raid_disks += delta_disks; | |
78b10e66 N |
9748 | |
9749 | dprintf("\tfor level : %i\n", geo.level); | |
9750 | dprintf("\tfor raid_disks : %i\n", geo.raid_disks); | |
9751 | ||
9752 | if (experimental() == 0) | |
9753 | return ret_val; | |
9754 | ||
78b10e66 | 9755 | if (st->container_dev == st->devnum) { |
694575e7 KW |
9756 | /* On container level we can only increase number of devices. */ |
9757 | dprintf("imsm: info: Container operation\n"); | |
78b10e66 | 9758 | int old_raid_disks = 0; |
6dc0be30 | 9759 | |
78b10e66 N |
9760 | if (imsm_reshape_is_allowed_on_container( |
9761 | st, &geo, &old_raid_disks)) { | |
9762 | struct imsm_update_reshape *u = NULL; | |
9763 | int len; | |
9764 | ||
9765 | len = imsm_create_metadata_update_for_reshape( | |
9766 | st, &geo, old_raid_disks, &u); | |
9767 | ||
ed08d51c AK |
9768 | if (len <= 0) { |
9769 | dprintf("imsm: Cannot prepare update\n"); | |
9770 | goto exit_imsm_reshape_super; | |
9771 | } | |
9772 | ||
8dd70bce AK |
9773 | ret_val = 0; |
9774 | /* update metadata locally */ | |
9775 | imsm_update_metadata_locally(st, u, len); | |
9776 | /* and possibly remotely */ | |
9777 | if (st->update_tail) | |
9778 | append_metadata_update(st, u, len); | |
9779 | else | |
ed08d51c | 9780 | free(u); |
8dd70bce | 9781 | |
694575e7 | 9782 | } else { |
e7ff7e40 AK |
9783 | fprintf(stderr, Name ": (imsm) Operation " |
9784 | "is not allowed on this container\n"); | |
694575e7 KW |
9785 | } |
9786 | } else { | |
9787 | /* On volume level we support following operations | |
471bceb6 KW |
9788 | * - takeover: raid10 -> raid0; raid0 -> raid10 |
9789 | * - chunk size migration | |
9790 | * - migration: raid5 -> raid0; raid0 -> raid5 | |
9791 | */ | |
9792 | struct intel_super *super = st->sb; | |
9793 | struct intel_dev *dev = super->devlist; | |
9794 | int change, devnum; | |
694575e7 | 9795 | dprintf("imsm: info: Volume operation\n"); |
471bceb6 KW |
9796 | /* find requested device */ |
9797 | while (dev) { | |
19986c72 MB |
9798 | if (imsm_find_array_minor_by_subdev( |
9799 | dev->index, st->container_dev, &devnum) == 0 | |
9800 | && devnum == geo.dev_id) | |
471bceb6 KW |
9801 | break; |
9802 | dev = dev->next; | |
9803 | } | |
9804 | if (dev == NULL) { | |
9805 | fprintf(stderr, Name " Cannot find %s (%i) subarray\n", | |
9806 | geo.dev_name, geo.dev_id); | |
9807 | goto exit_imsm_reshape_super; | |
9808 | } | |
9809 | super->current_vol = dev->index; | |
694575e7 KW |
9810 | change = imsm_analyze_change(st, &geo); |
9811 | switch (change) { | |
471bceb6 | 9812 | case CH_TAKEOVER: |
bb025c2f | 9813 | ret_val = imsm_takeover(st, &geo); |
694575e7 | 9814 | break; |
48c5303a PC |
9815 | case CH_MIGRATION: { |
9816 | struct imsm_update_reshape_migration *u = NULL; | |
9817 | int len = | |
9818 | imsm_create_metadata_update_for_migration( | |
9819 | st, &geo, &u); | |
9820 | if (len < 1) { | |
9821 | dprintf("imsm: " | |
9822 | "Cannot prepare update\n"); | |
9823 | break; | |
9824 | } | |
471bceb6 | 9825 | ret_val = 0; |
48c5303a PC |
9826 | /* update metadata locally */ |
9827 | imsm_update_metadata_locally(st, u, len); | |
9828 | /* and possibly remotely */ | |
9829 | if (st->update_tail) | |
9830 | append_metadata_update(st, u, len); | |
9831 | else | |
9832 | free(u); | |
9833 | } | |
9834 | break; | |
471bceb6 KW |
9835 | default: |
9836 | ret_val = 1; | |
694575e7 | 9837 | } |
694575e7 | 9838 | } |
78b10e66 | 9839 | |
ed08d51c | 9840 | exit_imsm_reshape_super: |
78b10e66 N |
9841 | dprintf("imsm: reshape_super Exit code = %i\n", ret_val); |
9842 | return ret_val; | |
9843 | } | |
2cda7640 | 9844 | |
eee67a47 AK |
9845 | /******************************************************************************* |
9846 | * Function: wait_for_reshape_imsm | |
9847 | * Description: Function writes new sync_max value and waits until | |
9848 | * reshape process reach new position | |
9849 | * Parameters: | |
9850 | * sra : general array info | |
eee67a47 AK |
9851 | * ndata : number of disks in new array's layout |
9852 | * Returns: | |
9853 | * 0 : success, | |
9854 | * 1 : there is no reshape in progress, | |
9855 | * -1 : fail | |
9856 | ******************************************************************************/ | |
ae9f01f8 | 9857 | int wait_for_reshape_imsm(struct mdinfo *sra, int ndata) |
eee67a47 AK |
9858 | { |
9859 | int fd = sysfs_get_fd(sra, NULL, "reshape_position"); | |
9860 | unsigned long long completed; | |
ae9f01f8 AK |
9861 | /* to_complete : new sync_max position */ |
9862 | unsigned long long to_complete = sra->reshape_progress; | |
9863 | unsigned long long position_to_set = to_complete / ndata; | |
eee67a47 | 9864 | |
ae9f01f8 AK |
9865 | if (fd < 0) { |
9866 | dprintf("imsm: wait_for_reshape_imsm() " | |
9867 | "cannot open reshape_position\n"); | |
eee67a47 | 9868 | return 1; |
ae9f01f8 | 9869 | } |
eee67a47 | 9870 | |
ae9f01f8 AK |
9871 | if (sysfs_fd_get_ll(fd, &completed) < 0) { |
9872 | dprintf("imsm: wait_for_reshape_imsm() " | |
9873 | "cannot read reshape_position (no reshape in progres)\n"); | |
9874 | close(fd); | |
9875 | return 0; | |
9876 | } | |
eee67a47 | 9877 | |
ae9f01f8 AK |
9878 | if (completed > to_complete) { |
9879 | dprintf("imsm: wait_for_reshape_imsm() " | |
9880 | "wrong next position to set %llu (%llu)\n", | |
9881 | to_complete, completed); | |
9882 | close(fd); | |
9883 | return -1; | |
9884 | } | |
9885 | dprintf("Position set: %llu\n", position_to_set); | |
9886 | if (sysfs_set_num(sra, NULL, "sync_max", | |
9887 | position_to_set) != 0) { | |
9888 | dprintf("imsm: wait_for_reshape_imsm() " | |
9889 | "cannot set reshape position to %llu\n", | |
9890 | position_to_set); | |
9891 | close(fd); | |
9892 | return -1; | |
eee67a47 AK |
9893 | } |
9894 | ||
eee67a47 AK |
9895 | do { |
9896 | char action[20]; | |
9897 | fd_set rfds; | |
9898 | FD_ZERO(&rfds); | |
9899 | FD_SET(fd, &rfds); | |
a47e44fb AK |
9900 | select(fd+1, &rfds, NULL, NULL, NULL); |
9901 | if (sysfs_get_str(sra, NULL, "sync_action", | |
9902 | action, 20) > 0 && | |
9903 | strncmp(action, "reshape", 7) != 0) | |
9904 | break; | |
eee67a47 | 9905 | if (sysfs_fd_get_ll(fd, &completed) < 0) { |
ae9f01f8 AK |
9906 | dprintf("imsm: wait_for_reshape_imsm() " |
9907 | "cannot read reshape_position (in loop)\n"); | |
eee67a47 AK |
9908 | close(fd); |
9909 | return 1; | |
9910 | } | |
eee67a47 AK |
9911 | } while (completed < to_complete); |
9912 | close(fd); | |
9913 | return 0; | |
9914 | ||
9915 | } | |
9916 | ||
b915c95f AK |
9917 | /******************************************************************************* |
9918 | * Function: check_degradation_change | |
9919 | * Description: Check that array hasn't become failed. | |
9920 | * Parameters: | |
9921 | * info : for sysfs access | |
9922 | * sources : source disks descriptors | |
9923 | * degraded: previous degradation level | |
9924 | * Returns: | |
9925 | * degradation level | |
9926 | ******************************************************************************/ | |
9927 | int check_degradation_change(struct mdinfo *info, | |
9928 | int *sources, | |
9929 | int degraded) | |
9930 | { | |
9931 | unsigned long long new_degraded; | |
9932 | sysfs_get_ll(info, NULL, "degraded", &new_degraded); | |
9933 | if (new_degraded != (unsigned long long)degraded) { | |
9934 | /* check each device to ensure it is still working */ | |
9935 | struct mdinfo *sd; | |
9936 | new_degraded = 0; | |
9937 | for (sd = info->devs ; sd ; sd = sd->next) { | |
9938 | if (sd->disk.state & (1<<MD_DISK_FAULTY)) | |
9939 | continue; | |
9940 | if (sd->disk.state & (1<<MD_DISK_SYNC)) { | |
9941 | char sbuf[20]; | |
9942 | if (sysfs_get_str(info, | |
9943 | sd, "state", sbuf, 20) < 0 || | |
9944 | strstr(sbuf, "faulty") || | |
9945 | strstr(sbuf, "in_sync") == NULL) { | |
9946 | /* this device is dead */ | |
9947 | sd->disk.state = (1<<MD_DISK_FAULTY); | |
9948 | if (sd->disk.raid_disk >= 0 && | |
9949 | sources[sd->disk.raid_disk] >= 0) { | |
9950 | close(sources[ | |
9951 | sd->disk.raid_disk]); | |
9952 | sources[sd->disk.raid_disk] = | |
9953 | -1; | |
9954 | } | |
9955 | new_degraded++; | |
9956 | } | |
9957 | } | |
9958 | } | |
9959 | } | |
9960 | ||
9961 | return new_degraded; | |
9962 | } | |
9963 | ||
10f22854 AK |
9964 | /******************************************************************************* |
9965 | * Function: imsm_manage_reshape | |
9966 | * Description: Function finds array under reshape and it manages reshape | |
9967 | * process. It creates stripes backups (if required) and sets | |
9968 | * checheckpoits. | |
9969 | * Parameters: | |
9970 | * afd : Backup handle (nattive) - not used | |
9971 | * sra : general array info | |
9972 | * reshape : reshape parameters - not used | |
9973 | * st : supertype structure | |
9974 | * blocks : size of critical section [blocks] | |
9975 | * fds : table of source device descriptor | |
9976 | * offsets : start of array (offest per devices) | |
9977 | * dests : not used | |
9978 | * destfd : table of destination device descriptor | |
9979 | * destoffsets : table of destination offsets (per device) | |
9980 | * Returns: | |
9981 | * 1 : success, reshape is done | |
9982 | * 0 : fail | |
9983 | ******************************************************************************/ | |
999b4972 N |
9984 | static int imsm_manage_reshape( |
9985 | int afd, struct mdinfo *sra, struct reshape *reshape, | |
10f22854 | 9986 | struct supertype *st, unsigned long backup_blocks, |
999b4972 N |
9987 | int *fds, unsigned long long *offsets, |
9988 | int dests, int *destfd, unsigned long long *destoffsets) | |
9989 | { | |
10f22854 AK |
9990 | int ret_val = 0; |
9991 | struct intel_super *super = st->sb; | |
9992 | struct intel_dev *dv = NULL; | |
9993 | struct imsm_dev *dev = NULL; | |
a6b6d984 | 9994 | struct imsm_map *map_src; |
10f22854 AK |
9995 | int migr_vol_qan = 0; |
9996 | int ndata, odata; /* [bytes] */ | |
9997 | int chunk; /* [bytes] */ | |
9998 | struct migr_record *migr_rec; | |
9999 | char *buf = NULL; | |
10000 | unsigned int buf_size; /* [bytes] */ | |
10001 | unsigned long long max_position; /* array size [bytes] */ | |
10002 | unsigned long long next_step; /* [blocks]/[bytes] */ | |
10003 | unsigned long long old_data_stripe_length; | |
10f22854 AK |
10004 | unsigned long long start_src; /* [bytes] */ |
10005 | unsigned long long start; /* [bytes] */ | |
10006 | unsigned long long start_buf_shift; /* [bytes] */ | |
b915c95f | 10007 | int degraded = 0; |
ab724b98 | 10008 | int source_layout = 0; |
10f22854 | 10009 | |
1ab242d8 | 10010 | if (!fds || !offsets || !sra) |
10f22854 AK |
10011 | goto abort; |
10012 | ||
10013 | /* Find volume during the reshape */ | |
10014 | for (dv = super->devlist; dv; dv = dv->next) { | |
10015 | if (dv->dev->vol.migr_type == MIGR_GEN_MIGR | |
10016 | && dv->dev->vol.migr_state == 1) { | |
10017 | dev = dv->dev; | |
10018 | migr_vol_qan++; | |
10019 | } | |
10020 | } | |
10021 | /* Only one volume can migrate at the same time */ | |
10022 | if (migr_vol_qan != 1) { | |
10023 | fprintf(stderr, Name " : %s", migr_vol_qan ? | |
10024 | "Number of migrating volumes greater than 1\n" : | |
10025 | "There is no volume during migrationg\n"); | |
10026 | goto abort; | |
10027 | } | |
10028 | ||
238c0a71 | 10029 | map_src = get_imsm_map(dev, MAP_1); |
10f22854 AK |
10030 | if (map_src == NULL) |
10031 | goto abort; | |
10f22854 | 10032 | |
238c0a71 AK |
10033 | ndata = imsm_num_data_members(dev, MAP_0); |
10034 | odata = imsm_num_data_members(dev, MAP_1); | |
10f22854 | 10035 | |
7b1ab482 | 10036 | chunk = __le16_to_cpu(map_src->blocks_per_strip) * 512; |
10f22854 AK |
10037 | old_data_stripe_length = odata * chunk; |
10038 | ||
10039 | migr_rec = super->migr_rec; | |
10040 | ||
10f22854 AK |
10041 | /* initialize migration record for start condition */ |
10042 | if (sra->reshape_progress == 0) | |
10043 | init_migr_record_imsm(st, dev, sra); | |
b2c59438 AK |
10044 | else { |
10045 | if (__le32_to_cpu(migr_rec->rec_status) != UNIT_SRC_NORMAL) { | |
10046 | dprintf("imsm: cannot restart migration when data " | |
10047 | "are present in copy area.\n"); | |
10048 | goto abort; | |
10049 | } | |
10050 | } | |
10f22854 AK |
10051 | |
10052 | /* size for data */ | |
10053 | buf_size = __le32_to_cpu(migr_rec->blocks_per_unit) * 512; | |
10054 | /* extend buffer size for parity disk */ | |
10055 | buf_size += __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
10056 | /* add space for stripe aligment */ | |
10057 | buf_size += old_data_stripe_length; | |
10058 | if (posix_memalign((void **)&buf, 4096, buf_size)) { | |
10059 | dprintf("imsm: Cannot allocate checpoint buffer\n"); | |
10060 | goto abort; | |
10061 | } | |
10062 | ||
3ef4403c | 10063 | max_position = sra->component_size * ndata; |
68eb8bc6 | 10064 | source_layout = imsm_level_to_layout(map_src->raid_level); |
10f22854 AK |
10065 | |
10066 | while (__le32_to_cpu(migr_rec->curr_migr_unit) < | |
10067 | __le32_to_cpu(migr_rec->num_migr_units)) { | |
10068 | /* current reshape position [blocks] */ | |
10069 | unsigned long long current_position = | |
10070 | __le32_to_cpu(migr_rec->blocks_per_unit) | |
10071 | * __le32_to_cpu(migr_rec->curr_migr_unit); | |
10072 | unsigned long long border; | |
10073 | ||
b915c95f AK |
10074 | /* Check that array hasn't become failed. |
10075 | */ | |
10076 | degraded = check_degradation_change(sra, fds, degraded); | |
10077 | if (degraded > 1) { | |
10078 | dprintf("imsm: Abort reshape due to degradation" | |
10079 | " level (%i)\n", degraded); | |
10080 | goto abort; | |
10081 | } | |
10082 | ||
10f22854 AK |
10083 | next_step = __le32_to_cpu(migr_rec->blocks_per_unit); |
10084 | ||
10085 | if ((current_position + next_step) > max_position) | |
10086 | next_step = max_position - current_position; | |
10087 | ||
92144abf | 10088 | start = current_position * 512; |
10f22854 AK |
10089 | |
10090 | /* allign reading start to old geometry */ | |
10091 | start_buf_shift = start % old_data_stripe_length; | |
10092 | start_src = start - start_buf_shift; | |
10093 | ||
10094 | border = (start_src / odata) - (start / ndata); | |
10095 | border /= 512; | |
10096 | if (border <= __le32_to_cpu(migr_rec->dest_depth_per_unit)) { | |
10097 | /* save critical stripes to buf | |
10098 | * start - start address of current unit | |
10099 | * to backup [bytes] | |
10100 | * start_src - start address of current unit | |
10101 | * to backup alligned to source array | |
10102 | * [bytes] | |
10103 | */ | |
10104 | unsigned long long next_step_filler = 0; | |
10105 | unsigned long long copy_length = next_step * 512; | |
10106 | ||
10107 | /* allign copy area length to stripe in old geometry */ | |
10108 | next_step_filler = ((copy_length + start_buf_shift) | |
10109 | % old_data_stripe_length); | |
10110 | if (next_step_filler) | |
10111 | next_step_filler = (old_data_stripe_length | |
10112 | - next_step_filler); | |
10113 | dprintf("save_stripes() parameters: start = %llu," | |
10114 | "\tstart_src = %llu,\tnext_step*512 = %llu," | |
10115 | "\tstart_in_buf_shift = %llu," | |
10116 | "\tnext_step_filler = %llu\n", | |
10117 | start, start_src, copy_length, | |
10118 | start_buf_shift, next_step_filler); | |
10119 | ||
10120 | if (save_stripes(fds, offsets, map_src->num_members, | |
ab724b98 AK |
10121 | chunk, map_src->raid_level, |
10122 | source_layout, 0, NULL, start_src, | |
10f22854 AK |
10123 | copy_length + |
10124 | next_step_filler + start_buf_shift, | |
10125 | buf)) { | |
10126 | dprintf("imsm: Cannot save stripes" | |
10127 | " to buffer\n"); | |
10128 | goto abort; | |
10129 | } | |
10130 | /* Convert data to destination format and store it | |
10131 | * in backup general migration area | |
10132 | */ | |
10133 | if (save_backup_imsm(st, dev, sra, | |
aea93171 | 10134 | buf + start_buf_shift, copy_length)) { |
10f22854 AK |
10135 | dprintf("imsm: Cannot save stripes to " |
10136 | "target devices\n"); | |
10137 | goto abort; | |
10138 | } | |
10139 | if (save_checkpoint_imsm(st, sra, | |
10140 | UNIT_SRC_IN_CP_AREA)) { | |
10141 | dprintf("imsm: Cannot write checkpoint to " | |
10142 | "migration record (UNIT_SRC_IN_CP_AREA)\n"); | |
10143 | goto abort; | |
10144 | } | |
8016a6d4 AK |
10145 | } else { |
10146 | /* set next step to use whole border area */ | |
10147 | border /= next_step; | |
10148 | if (border > 1) | |
10149 | next_step *= border; | |
10f22854 AK |
10150 | } |
10151 | /* When data backed up, checkpoint stored, | |
10152 | * kick the kernel to reshape unit of data | |
10153 | */ | |
10154 | next_step = next_step + sra->reshape_progress; | |
8016a6d4 AK |
10155 | /* limit next step to array max position */ |
10156 | if (next_step > max_position) | |
10157 | next_step = max_position; | |
10f22854 AK |
10158 | sysfs_set_num(sra, NULL, "suspend_lo", sra->reshape_progress); |
10159 | sysfs_set_num(sra, NULL, "suspend_hi", next_step); | |
ae9f01f8 | 10160 | sra->reshape_progress = next_step; |
10f22854 AK |
10161 | |
10162 | /* wait until reshape finish */ | |
ae9f01f8 | 10163 | if (wait_for_reshape_imsm(sra, ndata) < 0) { |
c47b0ff6 AK |
10164 | dprintf("wait_for_reshape_imsm returned error!\n"); |
10165 | goto abort; | |
10166 | } | |
10f22854 | 10167 | |
0228d92c AK |
10168 | if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) { |
10169 | /* ignore error == 2, this can mean end of reshape here | |
10170 | */ | |
10f22854 AK |
10171 | dprintf("imsm: Cannot write checkpoint to " |
10172 | "migration record (UNIT_SRC_NORMAL)\n"); | |
10173 | goto abort; | |
10174 | } | |
10175 | ||
10176 | } | |
10177 | ||
10178 | /* return '1' if done */ | |
10179 | ret_val = 1; | |
10180 | abort: | |
10181 | free(buf); | |
10182 | abort_reshape(sra); | |
10183 | ||
10184 | return ret_val; | |
999b4972 | 10185 | } |
71204a50 | 10186 | #endif /* MDASSEMBLE */ |
999b4972 | 10187 | |
cdddbdbc DW |
10188 | struct superswitch super_imsm = { |
10189 | #ifndef MDASSEMBLE | |
10190 | .examine_super = examine_super_imsm, | |
10191 | .brief_examine_super = brief_examine_super_imsm, | |
4737ae25 | 10192 | .brief_examine_subarrays = brief_examine_subarrays_imsm, |
9d84c8ea | 10193 | .export_examine_super = export_examine_super_imsm, |
cdddbdbc DW |
10194 | .detail_super = detail_super_imsm, |
10195 | .brief_detail_super = brief_detail_super_imsm, | |
bf5a934a | 10196 | .write_init_super = write_init_super_imsm, |
0e600426 N |
10197 | .validate_geometry = validate_geometry_imsm, |
10198 | .add_to_super = add_to_super_imsm, | |
1a64be56 | 10199 | .remove_from_super = remove_from_super_imsm, |
d665cc31 | 10200 | .detail_platform = detail_platform_imsm, |
33414a01 | 10201 | .kill_subarray = kill_subarray_imsm, |
aa534678 | 10202 | .update_subarray = update_subarray_imsm, |
2b959fbf | 10203 | .load_container = load_container_imsm, |
71204a50 N |
10204 | .default_geometry = default_geometry_imsm, |
10205 | .get_disk_controller_domain = imsm_get_disk_controller_domain, | |
10206 | .reshape_super = imsm_reshape_super, | |
10207 | .manage_reshape = imsm_manage_reshape, | |
9e2d750d | 10208 | .recover_backup = recover_backup_imsm, |
cdddbdbc DW |
10209 | #endif |
10210 | .match_home = match_home_imsm, | |
10211 | .uuid_from_super= uuid_from_super_imsm, | |
10212 | .getinfo_super = getinfo_super_imsm, | |
5c4cd5da | 10213 | .getinfo_super_disks = getinfo_super_disks_imsm, |
cdddbdbc DW |
10214 | .update_super = update_super_imsm, |
10215 | ||
10216 | .avail_size = avail_size_imsm, | |
80e7f8c3 | 10217 | .min_acceptable_spare_size = min_acceptable_spare_size_imsm, |
cdddbdbc DW |
10218 | |
10219 | .compare_super = compare_super_imsm, | |
10220 | ||
10221 | .load_super = load_super_imsm, | |
bf5a934a | 10222 | .init_super = init_super_imsm, |
e683ca88 | 10223 | .store_super = store_super_imsm, |
cdddbdbc DW |
10224 | .free_super = free_super_imsm, |
10225 | .match_metadata_desc = match_metadata_desc_imsm, | |
bf5a934a | 10226 | .container_content = container_content_imsm, |
cdddbdbc | 10227 | |
276d77db | 10228 | |
cdddbdbc | 10229 | .external = 1, |
4cce4069 | 10230 | .name = "imsm", |
845dea95 | 10231 | |
0e600426 | 10232 | #ifndef MDASSEMBLE |
845dea95 NB |
10233 | /* for mdmon */ |
10234 | .open_new = imsm_open_new, | |
ed9d66aa | 10235 | .set_array_state= imsm_set_array_state, |
845dea95 NB |
10236 | .set_disk = imsm_set_disk, |
10237 | .sync_metadata = imsm_sync_metadata, | |
88758e9d | 10238 | .activate_spare = imsm_activate_spare, |
e8319a19 | 10239 | .process_update = imsm_process_update, |
8273f55e | 10240 | .prepare_update = imsm_prepare_update, |
0e600426 | 10241 | #endif /* MDASSEMBLE */ |
cdddbdbc | 10242 | }; |