]>
Commit | Line | Data |
---|---|---|
cdddbdbc DW |
1 | /* |
2 | * mdadm - Intel(R) Matrix Storage Manager Support | |
3 | * | |
4 | * Copyright (C) 2002-2007 Intel Corporation | |
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 | ||
20 | #include "mdadm.h" | |
c2a1e7da | 21 | #include "mdmon.h" |
cdddbdbc DW |
22 | #include <values.h> |
23 | #include <scsi/sg.h> | |
24 | #include <ctype.h> | |
25 | ||
26 | /* MPB == Metadata Parameter Block */ | |
27 | #define MPB_SIGNATURE "Intel Raid ISM Cfg Sig. " | |
28 | #define MPB_SIG_LEN (strlen(MPB_SIGNATURE)) | |
29 | #define MPB_VERSION_RAID0 "1.0.00" | |
30 | #define MPB_VERSION_RAID1 "1.1.00" | |
31 | #define MPB_VERSION_RAID5 "1.2.02" | |
32 | #define MAX_SIGNATURE_LENGTH 32 | |
33 | #define MAX_RAID_SERIAL_LEN 16 | |
c2c087e6 DW |
34 | #define MPB_SECTOR_CNT 418 |
35 | #define IMSM_RESERVED_SECTORS 4096 | |
cdddbdbc DW |
36 | |
37 | /* Disk configuration info. */ | |
38 | #define IMSM_MAX_DEVICES 255 | |
39 | struct imsm_disk { | |
40 | __u8 serial[MAX_RAID_SERIAL_LEN];/* 0xD8 - 0xE7 ascii serial number */ | |
41 | __u32 total_blocks; /* 0xE8 - 0xEB total blocks */ | |
42 | __u32 scsi_id; /* 0xEC - 0xEF scsi ID */ | |
43 | __u32 status; /* 0xF0 - 0xF3 */ | |
44 | #define SPARE_DISK 0x01 /* Spare */ | |
45 | #define CONFIGURED_DISK 0x02 /* Member of some RaidDev */ | |
46 | #define FAILED_DISK 0x04 /* Permanent failure */ | |
47 | #define USABLE_DISK 0x08 /* Fully usable unless FAILED_DISK is set */ | |
48 | ||
49 | #define IMSM_DISK_FILLERS 5 | |
50 | __u32 filler[IMSM_DISK_FILLERS]; /* 0xF4 - 0x107 MPB_DISK_FILLERS for future expansion */ | |
51 | }; | |
52 | ||
53 | /* RAID map configuration infos. */ | |
54 | struct imsm_map { | |
55 | __u32 pba_of_lba0; /* start address of partition */ | |
56 | __u32 blocks_per_member;/* blocks per member */ | |
57 | __u32 num_data_stripes; /* number of data stripes */ | |
58 | __u16 blocks_per_strip; | |
59 | __u8 map_state; /* Normal, Uninitialized, Degraded, Failed */ | |
60 | #define IMSM_T_STATE_NORMAL 0 | |
61 | #define IMSM_T_STATE_UNINITIALIZED 1 | |
62 | #define IMSM_T_STATE_DEGRADED 2 /* FIXME: is this correct? */ | |
63 | #define IMSM_T_STATE_FAILED 3 /* FIXME: is this correct? */ | |
64 | __u8 raid_level; | |
65 | #define IMSM_T_RAID0 0 | |
66 | #define IMSM_T_RAID1 1 | |
67 | #define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */ | |
68 | __u8 num_members; /* number of member disks */ | |
69 | __u8 reserved[3]; | |
70 | __u32 filler[7]; /* expansion area */ | |
7eef0453 | 71 | #define IMSM_ORD_REBUILD (1 << 24) |
cdddbdbc | 72 | __u32 disk_ord_tbl[1]; /* disk_ord_tbl[num_members], |
7eef0453 DW |
73 | * top byte contains some flags |
74 | */ | |
cdddbdbc DW |
75 | } __attribute__ ((packed)); |
76 | ||
77 | struct imsm_vol { | |
78 | __u32 reserved[2]; | |
79 | __u8 migr_state; /* Normal or Migrating */ | |
80 | __u8 migr_type; /* Initializing, Rebuilding, ... */ | |
81 | __u8 dirty; | |
82 | __u8 fill[1]; | |
83 | __u32 filler[5]; | |
84 | struct imsm_map map[1]; | |
85 | /* here comes another one if migr_state */ | |
86 | } __attribute__ ((packed)); | |
87 | ||
88 | struct imsm_dev { | |
89 | __u8 volume[MAX_RAID_SERIAL_LEN]; | |
90 | __u32 size_low; | |
91 | __u32 size_high; | |
92 | __u32 status; /* Persistent RaidDev status */ | |
93 | __u32 reserved_blocks; /* Reserved blocks at beginning of volume */ | |
94 | #define IMSM_DEV_FILLERS 12 | |
95 | __u32 filler[IMSM_DEV_FILLERS]; | |
96 | struct imsm_vol vol; | |
97 | } __attribute__ ((packed)); | |
98 | ||
99 | struct imsm_super { | |
100 | __u8 sig[MAX_SIGNATURE_LENGTH]; /* 0x00 - 0x1F */ | |
101 | __u32 check_sum; /* 0x20 - 0x23 MPB Checksum */ | |
102 | __u32 mpb_size; /* 0x24 - 0x27 Size of MPB */ | |
103 | __u32 family_num; /* 0x28 - 0x2B Checksum from first time this config was written */ | |
104 | __u32 generation_num; /* 0x2C - 0x2F Incremented each time this array's MPB is written */ | |
604b746f JD |
105 | __u32 error_log_size; /* 0x30 - 0x33 in bytes */ |
106 | __u32 attributes; /* 0x34 - 0x37 */ | |
cdddbdbc DW |
107 | __u8 num_disks; /* 0x38 Number of configured disks */ |
108 | __u8 num_raid_devs; /* 0x39 Number of configured volumes */ | |
604b746f JD |
109 | __u8 error_log_pos; /* 0x3A */ |
110 | __u8 fill[1]; /* 0x3B */ | |
111 | __u32 cache_size; /* 0x3c - 0x40 in mb */ | |
112 | __u32 orig_family_num; /* 0x40 - 0x43 original family num */ | |
113 | __u32 pwr_cycle_count; /* 0x44 - 0x47 simulated power cycle count for array */ | |
114 | __u32 bbm_log_size; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */ | |
115 | #define IMSM_FILLERS 35 | |
116 | __u32 filler[IMSM_FILLERS]; /* 0x4C - 0xD7 RAID_MPB_FILLERS */ | |
cdddbdbc DW |
117 | struct imsm_disk disk[1]; /* 0xD8 diskTbl[numDisks] */ |
118 | /* here comes imsm_dev[num_raid_devs] */ | |
604b746f | 119 | /* here comes BBM logs */ |
cdddbdbc DW |
120 | } __attribute__ ((packed)); |
121 | ||
604b746f JD |
122 | #define BBM_LOG_MAX_ENTRIES 254 |
123 | ||
124 | struct bbm_log_entry { | |
125 | __u64 defective_block_start; | |
126 | #define UNREADABLE 0xFFFFFFFF | |
127 | __u32 spare_block_offset; | |
128 | __u16 remapped_marked_count; | |
129 | __u16 disk_ordinal; | |
130 | } __attribute__ ((__packed__)); | |
131 | ||
132 | struct bbm_log { | |
133 | __u32 signature; /* 0xABADB10C */ | |
134 | __u32 entry_count; | |
135 | __u32 reserved_spare_block_count; /* 0 */ | |
136 | __u32 reserved; /* 0xFFFF */ | |
137 | __u64 first_spare_lba; | |
138 | struct bbm_log_entry mapped_block_entries[BBM_LOG_MAX_ENTRIES]; | |
139 | } __attribute__ ((__packed__)); | |
140 | ||
141 | ||
cdddbdbc DW |
142 | #ifndef MDASSEMBLE |
143 | static char *map_state_str[] = { "normal", "uninitialized", "degraded", "failed" }; | |
144 | #endif | |
145 | ||
87eb16df | 146 | static unsigned int sector_count(__u32 bytes) |
cdddbdbc | 147 | { |
87eb16df DW |
148 | return ((bytes + (512-1)) & (~(512-1))) / 512; |
149 | } | |
cdddbdbc | 150 | |
87eb16df DW |
151 | static unsigned int mpb_sectors(struct imsm_super *mpb) |
152 | { | |
153 | return sector_count(__le32_to_cpu(mpb->mpb_size)); | |
cdddbdbc DW |
154 | } |
155 | ||
156 | /* internal representation of IMSM metadata */ | |
157 | struct intel_super { | |
158 | union { | |
949c47a0 DW |
159 | void *buf; /* O_DIRECT buffer for reading/writing metadata */ |
160 | struct imsm_super *anchor; /* immovable parameters */ | |
cdddbdbc | 161 | }; |
949c47a0 | 162 | size_t len; /* size of the 'buf' allocation */ |
c2c087e6 DW |
163 | int updates_pending; /* count of pending updates for mdmon */ |
164 | int creating_imsm; /* flag to indicate container creation */ | |
bf5a934a | 165 | int current_vol; /* index of raid device undergoing creation */ |
949c47a0 DW |
166 | #define IMSM_MAX_RAID_DEVS 2 |
167 | struct imsm_dev *dev_tbl[IMSM_MAX_RAID_DEVS]; | |
cdddbdbc DW |
168 | struct dl { |
169 | struct dl *next; | |
170 | int index; | |
171 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
172 | int major, minor; | |
173 | char *devname; | |
b9f594fe | 174 | struct imsm_disk disk; |
cdddbdbc DW |
175 | int fd; |
176 | } *disks; | |
604b746f | 177 | struct bbm_log *bbm_log; |
cdddbdbc DW |
178 | }; |
179 | ||
c2c087e6 DW |
180 | struct extent { |
181 | unsigned long long start, size; | |
182 | }; | |
183 | ||
88758e9d DW |
184 | /* definition of messages passed to imsm_process_update */ |
185 | enum imsm_update_type { | |
186 | update_activate_spare, | |
8273f55e | 187 | update_create_array, |
88758e9d DW |
188 | }; |
189 | ||
190 | struct imsm_update_activate_spare { | |
191 | enum imsm_update_type type; | |
d23fe947 | 192 | struct dl *dl; |
88758e9d DW |
193 | int slot; |
194 | int array; | |
195 | struct imsm_update_activate_spare *next; | |
196 | }; | |
197 | ||
8273f55e DW |
198 | struct imsm_update_create_array { |
199 | enum imsm_update_type type; | |
8273f55e | 200 | int dev_idx; |
6a3e913e | 201 | struct imsm_dev dev; |
8273f55e DW |
202 | }; |
203 | ||
0030e8d6 DW |
204 | static int imsm_env_devname_as_serial(void) |
205 | { | |
206 | char *val = getenv("IMSM_DEVNAME_AS_SERIAL"); | |
207 | ||
208 | if (val && atoi(val) == 1) | |
209 | return 1; | |
210 | ||
211 | return 0; | |
212 | } | |
213 | ||
214 | ||
cdddbdbc DW |
215 | static struct supertype *match_metadata_desc_imsm(char *arg) |
216 | { | |
217 | struct supertype *st; | |
218 | ||
219 | if (strcmp(arg, "imsm") != 0 && | |
220 | strcmp(arg, "default") != 0 | |
221 | ) | |
222 | return NULL; | |
223 | ||
224 | st = malloc(sizeof(*st)); | |
ef609477 | 225 | memset(st, 0, sizeof(*st)); |
cdddbdbc DW |
226 | st->ss = &super_imsm; |
227 | st->max_devs = IMSM_MAX_DEVICES; | |
228 | st->minor_version = 0; | |
229 | st->sb = NULL; | |
230 | return st; | |
231 | } | |
232 | ||
cdddbdbc DW |
233 | static __u8 *get_imsm_version(struct imsm_super *mpb) |
234 | { | |
235 | return &mpb->sig[MPB_SIG_LEN]; | |
236 | } | |
237 | ||
949c47a0 DW |
238 | /* retrieve a disk directly from the anchor when the anchor is known to be |
239 | * up-to-date, currently only at load time | |
240 | */ | |
241 | static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index) | |
cdddbdbc | 242 | { |
949c47a0 | 243 | if (index >= mpb->num_disks) |
cdddbdbc DW |
244 | return NULL; |
245 | return &mpb->disk[index]; | |
246 | } | |
247 | ||
b9f594fe | 248 | /* retrieve a disk from the parsed metadata */ |
949c47a0 DW |
249 | static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index) |
250 | { | |
b9f594fe DW |
251 | struct dl *d; |
252 | ||
253 | for (d = super->disks; d; d = d->next) | |
254 | if (d->index == index) | |
255 | return &d->disk; | |
256 | ||
257 | return NULL; | |
949c47a0 DW |
258 | } |
259 | ||
260 | /* generate a checksum directly from the anchor when the anchor is known to be | |
261 | * up-to-date, currently only at load or write_super after coalescing | |
262 | */ | |
263 | static __u32 __gen_imsm_checksum(struct imsm_super *mpb) | |
cdddbdbc DW |
264 | { |
265 | __u32 end = mpb->mpb_size / sizeof(end); | |
266 | __u32 *p = (__u32 *) mpb; | |
267 | __u32 sum = 0; | |
268 | ||
269 | while (end--) | |
270 | sum += __le32_to_cpu(*p++); | |
271 | ||
272 | return sum - __le32_to_cpu(mpb->check_sum); | |
273 | } | |
274 | ||
a965f303 DW |
275 | static size_t sizeof_imsm_map(struct imsm_map *map) |
276 | { | |
277 | return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1); | |
278 | } | |
279 | ||
280 | struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map) | |
cdddbdbc | 281 | { |
a965f303 DW |
282 | struct imsm_map *map = &dev->vol.map[0]; |
283 | ||
284 | if (second_map && !dev->vol.migr_state) | |
285 | return NULL; | |
286 | else if (second_map) { | |
287 | void *ptr = map; | |
288 | ||
289 | return ptr + sizeof_imsm_map(map); | |
290 | } else | |
291 | return map; | |
292 | ||
293 | } | |
cdddbdbc | 294 | |
3393c6af DW |
295 | /* return the size of the device. |
296 | * migr_state increases the returned size if map[0] were to be duplicated | |
297 | */ | |
298 | static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state) | |
a965f303 DW |
299 | { |
300 | size_t size = sizeof(*dev) - sizeof(struct imsm_map) + | |
301 | sizeof_imsm_map(get_imsm_map(dev, 0)); | |
cdddbdbc DW |
302 | |
303 | /* migrating means an additional map */ | |
a965f303 DW |
304 | if (dev->vol.migr_state) |
305 | size += sizeof_imsm_map(get_imsm_map(dev, 1)); | |
3393c6af DW |
306 | else if (migr_state) |
307 | size += sizeof_imsm_map(get_imsm_map(dev, 0)); | |
cdddbdbc DW |
308 | |
309 | return size; | |
310 | } | |
311 | ||
949c47a0 | 312 | static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index) |
cdddbdbc DW |
313 | { |
314 | int offset; | |
315 | int i; | |
316 | void *_mpb = mpb; | |
317 | ||
949c47a0 | 318 | if (index >= mpb->num_raid_devs) |
cdddbdbc DW |
319 | return NULL; |
320 | ||
321 | /* devices start after all disks */ | |
322 | offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb; | |
323 | ||
324 | for (i = 0; i <= index; i++) | |
325 | if (i == index) | |
326 | return _mpb + offset; | |
327 | else | |
3393c6af | 328 | offset += sizeof_imsm_dev(_mpb + offset, 0); |
cdddbdbc DW |
329 | |
330 | return NULL; | |
331 | } | |
332 | ||
949c47a0 DW |
333 | static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index) |
334 | { | |
335 | if (index >= super->anchor->num_raid_devs) | |
336 | return NULL; | |
337 | return super->dev_tbl[index]; | |
338 | } | |
339 | ||
cdddbdbc DW |
340 | static __u32 get_imsm_disk_idx(struct imsm_map *map, int slot) |
341 | { | |
342 | __u32 *ord_tbl = &map->disk_ord_tbl[slot]; | |
343 | ||
7eef0453 DW |
344 | /* top byte identifies disk under rebuild |
345 | * why not just use the USABLE bit... oh well. | |
346 | */ | |
cdddbdbc DW |
347 | return __le32_to_cpu(*ord_tbl & ~(0xff << 24)); |
348 | } | |
349 | ||
7eef0453 DW |
350 | static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev, int slot) |
351 | { | |
352 | struct imsm_map *map; | |
353 | ||
354 | if (dev->vol.migr_state) | |
355 | map = get_imsm_map(dev, 0); | |
356 | else | |
357 | map = get_imsm_map(dev, 1); | |
358 | ||
359 | return map->disk_ord_tbl[slot]; | |
360 | } | |
361 | ||
cdddbdbc DW |
362 | static int get_imsm_raid_level(struct imsm_map *map) |
363 | { | |
364 | if (map->raid_level == 1) { | |
365 | if (map->num_members == 2) | |
366 | return 1; | |
367 | else | |
368 | return 10; | |
369 | } | |
370 | ||
371 | return map->raid_level; | |
372 | } | |
373 | ||
c2c087e6 DW |
374 | static int cmp_extent(const void *av, const void *bv) |
375 | { | |
376 | const struct extent *a = av; | |
377 | const struct extent *b = bv; | |
378 | if (a->start < b->start) | |
379 | return -1; | |
380 | if (a->start > b->start) | |
381 | return 1; | |
382 | return 0; | |
383 | } | |
384 | ||
385 | static struct extent *get_extents(struct intel_super *super, struct dl *dl) | |
386 | { | |
387 | /* find a list of used extents on the given physical device */ | |
c2c087e6 DW |
388 | struct extent *rv, *e; |
389 | int i, j; | |
390 | int memberships = 0; | |
391 | ||
949c47a0 DW |
392 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
393 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
a965f303 | 394 | struct imsm_map *map = get_imsm_map(dev, 0); |
c2c087e6 DW |
395 | |
396 | for (j = 0; j < map->num_members; j++) { | |
397 | __u32 index = get_imsm_disk_idx(map, j); | |
398 | ||
399 | if (index == dl->index) | |
400 | memberships++; | |
401 | } | |
402 | } | |
403 | rv = malloc(sizeof(struct extent) * (memberships + 1)); | |
404 | if (!rv) | |
405 | return NULL; | |
406 | e = rv; | |
407 | ||
949c47a0 DW |
408 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
409 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
a965f303 | 410 | struct imsm_map *map = get_imsm_map(dev, 0); |
c2c087e6 DW |
411 | |
412 | for (j = 0; j < map->num_members; j++) { | |
413 | __u32 index = get_imsm_disk_idx(map, j); | |
414 | ||
415 | if (index == dl->index) { | |
416 | e->start = __le32_to_cpu(map->pba_of_lba0); | |
417 | e->size = __le32_to_cpu(map->blocks_per_member); | |
418 | e++; | |
419 | } | |
420 | } | |
421 | } | |
422 | qsort(rv, memberships, sizeof(*rv), cmp_extent); | |
423 | ||
b9f594fe | 424 | e->start = __le32_to_cpu(dl->disk.total_blocks) - |
c2c087e6 DW |
425 | (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS); |
426 | e->size = 0; | |
427 | return rv; | |
428 | } | |
429 | ||
4f5bc454 | 430 | #ifndef MDASSEMBLE |
cdddbdbc DW |
431 | static void print_imsm_dev(struct imsm_dev *dev, int index) |
432 | { | |
433 | __u64 sz; | |
434 | int slot; | |
a965f303 | 435 | struct imsm_map *map = get_imsm_map(dev, 0); |
cdddbdbc DW |
436 | |
437 | printf("\n"); | |
438 | printf("[%s]:\n", dev->volume); | |
439 | printf(" RAID Level : %d\n", get_imsm_raid_level(map)); | |
440 | printf(" Members : %d\n", map->num_members); | |
441 | for (slot = 0; slot < map->num_members; slot++) | |
442 | if (index == get_imsm_disk_idx(map, slot)) | |
443 | break; | |
444 | if (slot < map->num_members) | |
445 | printf(" This Slot : %d\n", slot); | |
446 | else | |
447 | printf(" This Slot : ?\n"); | |
448 | sz = __le32_to_cpu(dev->size_high); | |
449 | sz <<= 32; | |
450 | sz += __le32_to_cpu(dev->size_low); | |
451 | printf(" Array Size : %llu%s\n", (unsigned long long)sz, | |
452 | human_size(sz * 512)); | |
453 | sz = __le32_to_cpu(map->blocks_per_member); | |
454 | printf(" Per Dev Size : %llu%s\n", (unsigned long long)sz, | |
455 | human_size(sz * 512)); | |
456 | printf(" Sector Offset : %u\n", | |
457 | __le32_to_cpu(map->pba_of_lba0)); | |
458 | printf(" Num Stripes : %u\n", | |
459 | __le32_to_cpu(map->num_data_stripes)); | |
460 | printf(" Chunk Size : %u KiB\n", | |
461 | __le16_to_cpu(map->blocks_per_strip) / 2); | |
462 | printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks)); | |
3393c6af DW |
463 | printf(" Migrate State : %s", dev->vol.migr_state ? "migrating" : "idle"); |
464 | if (dev->vol.migr_state) | |
465 | printf(": %s", dev->vol.migr_type ? "rebuilding" : "initializing"); | |
466 | printf("\n"); | |
467 | printf(" Map State : %s", map_state_str[map->map_state]); | |
468 | if (dev->vol.migr_state) { | |
469 | struct imsm_map *map = get_imsm_map(dev, 1); | |
470 | printf(", %s", map_state_str[map->map_state]); | |
471 | } | |
472 | printf("\n"); | |
cdddbdbc | 473 | printf(" Dirty State : %s\n", dev->vol.dirty ? "dirty" : "clean"); |
cdddbdbc DW |
474 | } |
475 | ||
476 | static void print_imsm_disk(struct imsm_super *mpb, int index) | |
477 | { | |
949c47a0 | 478 | struct imsm_disk *disk = __get_imsm_disk(mpb, index); |
cdddbdbc DW |
479 | char str[MAX_RAID_SERIAL_LEN]; |
480 | __u32 s; | |
481 | __u64 sz; | |
482 | ||
e9d82038 DW |
483 | if (index < 0) |
484 | return; | |
485 | ||
cdddbdbc DW |
486 | printf("\n"); |
487 | snprintf(str, MAX_RAID_SERIAL_LEN, "%s", disk->serial); | |
488 | printf(" Disk%02d Serial : %s\n", index, str); | |
489 | s = __le32_to_cpu(disk->status); | |
490 | printf(" State :%s%s%s%s\n", s&SPARE_DISK ? " spare" : "", | |
491 | s&CONFIGURED_DISK ? " active" : "", | |
492 | s&FAILED_DISK ? " failed" : "", | |
493 | s&USABLE_DISK ? " usable" : ""); | |
494 | printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id)); | |
c2c087e6 DW |
495 | sz = __le32_to_cpu(disk->total_blocks) - |
496 | (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS * mpb->num_raid_devs); | |
cdddbdbc DW |
497 | printf(" Usable Size : %llu%s\n", (unsigned long long)sz, |
498 | human_size(sz * 512)); | |
499 | } | |
500 | ||
501 | static void examine_super_imsm(struct supertype *st, char *homehost) | |
502 | { | |
503 | struct intel_super *super = st->sb; | |
949c47a0 | 504 | struct imsm_super *mpb = super->anchor; |
cdddbdbc DW |
505 | char str[MAX_SIGNATURE_LENGTH]; |
506 | int i; | |
507 | __u32 sum; | |
508 | ||
509 | snprintf(str, MPB_SIG_LEN, "%s", mpb->sig); | |
510 | printf(" Magic : %s\n", str); | |
511 | snprintf(str, strlen(MPB_VERSION_RAID0), "%s", get_imsm_version(mpb)); | |
512 | printf(" Version : %s\n", get_imsm_version(mpb)); | |
513 | printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num)); | |
514 | printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num)); | |
515 | sum = __le32_to_cpu(mpb->check_sum); | |
516 | printf(" Checksum : %08x %s\n", sum, | |
949c47a0 | 517 | __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect"); |
87eb16df | 518 | printf(" MPB Sectors : %d\n", mpb_sectors(mpb)); |
cdddbdbc DW |
519 | printf(" Disks : %d\n", mpb->num_disks); |
520 | printf(" RAID Devices : %d\n", mpb->num_raid_devs); | |
521 | print_imsm_disk(mpb, super->disks->index); | |
604b746f JD |
522 | if (super->bbm_log) { |
523 | struct bbm_log *log = super->bbm_log; | |
524 | ||
525 | printf("\n"); | |
526 | printf("Bad Block Management Log:\n"); | |
527 | printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size)); | |
528 | printf(" Signature : %x\n", __le32_to_cpu(log->signature)); | |
529 | printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count)); | |
530 | printf(" Spare Blocks : %d\n", __le32_to_cpu(log->reserved_spare_block_count)); | |
531 | printf(" First Spare : %llx\n", __le64_to_cpu(log->first_spare_lba)); | |
532 | } | |
cdddbdbc | 533 | for (i = 0; i < mpb->num_raid_devs; i++) |
949c47a0 | 534 | print_imsm_dev(__get_imsm_dev(mpb, i), super->disks->index); |
cdddbdbc DW |
535 | for (i = 0; i < mpb->num_disks; i++) { |
536 | if (i == super->disks->index) | |
537 | continue; | |
538 | print_imsm_disk(mpb, i); | |
539 | } | |
540 | } | |
541 | ||
542 | static void brief_examine_super_imsm(struct supertype *st) | |
543 | { | |
828408eb | 544 | printf("ARRAY /dev/imsm metadata=imsm\n"); |
cdddbdbc DW |
545 | } |
546 | ||
547 | static void detail_super_imsm(struct supertype *st, char *homehost) | |
548 | { | |
549 | printf("%s\n", __FUNCTION__); | |
550 | } | |
551 | ||
552 | static void brief_detail_super_imsm(struct supertype *st) | |
553 | { | |
554 | printf("%s\n", __FUNCTION__); | |
555 | } | |
556 | #endif | |
557 | ||
558 | static int match_home_imsm(struct supertype *st, char *homehost) | |
559 | { | |
560 | printf("%s\n", __FUNCTION__); | |
561 | ||
562 | return 0; | |
563 | } | |
564 | ||
565 | static void uuid_from_super_imsm(struct supertype *st, int uuid[4]) | |
566 | { | |
567 | printf("%s\n", __FUNCTION__); | |
568 | } | |
569 | ||
0d481d37 | 570 | #if 0 |
4f5bc454 DW |
571 | static void |
572 | get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p) | |
cdddbdbc | 573 | { |
cdddbdbc DW |
574 | __u8 *v = get_imsm_version(mpb); |
575 | __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH; | |
576 | char major[] = { 0, 0, 0 }; | |
577 | char minor[] = { 0 ,0, 0 }; | |
578 | char patch[] = { 0, 0, 0 }; | |
579 | char *ver_parse[] = { major, minor, patch }; | |
580 | int i, j; | |
581 | ||
582 | i = j = 0; | |
583 | while (*v != '\0' && v < end) { | |
584 | if (*v != '.' && j < 2) | |
585 | ver_parse[i][j++] = *v; | |
586 | else { | |
587 | i++; | |
588 | j = 0; | |
589 | } | |
590 | v++; | |
591 | } | |
592 | ||
4f5bc454 DW |
593 | *m = strtol(minor, NULL, 0); |
594 | *p = strtol(patch, NULL, 0); | |
595 | } | |
0d481d37 | 596 | #endif |
4f5bc454 | 597 | |
c2c087e6 DW |
598 | static int imsm_level_to_layout(int level) |
599 | { | |
600 | switch (level) { | |
601 | case 0: | |
602 | case 1: | |
603 | return 0; | |
604 | case 5: | |
605 | case 6: | |
a380c027 | 606 | return ALGORITHM_LEFT_ASYMMETRIC; |
c2c087e6 DW |
607 | case 10: |
608 | return 0x102; //FIXME is this correct? | |
609 | } | |
610 | return -1; | |
611 | } | |
612 | ||
bf5a934a DW |
613 | static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info) |
614 | { | |
615 | struct intel_super *super = st->sb; | |
949c47a0 | 616 | struct imsm_dev *dev = get_imsm_dev(super, super->current_vol); |
a965f303 | 617 | struct imsm_map *map = get_imsm_map(dev, 0); |
bf5a934a DW |
618 | |
619 | info->container_member = super->current_vol; | |
620 | info->array.raid_disks = map->num_members; | |
621 | info->array.level = get_imsm_raid_level(map); | |
622 | info->array.layout = imsm_level_to_layout(info->array.level); | |
623 | info->array.md_minor = -1; | |
624 | info->array.ctime = 0; | |
625 | info->array.utime = 0; | |
626 | info->array.chunk_size = __le16_to_cpu(map->blocks_per_strip * 512); | |
627 | ||
628 | info->data_offset = __le32_to_cpu(map->pba_of_lba0); | |
629 | info->component_size = __le32_to_cpu(map->blocks_per_member); | |
630 | ||
631 | info->disk.major = 0; | |
632 | info->disk.minor = 0; | |
633 | ||
634 | sprintf(info->text_version, "/%s/%d", | |
635 | devnum2devname(st->container_dev), | |
636 | info->container_member); | |
637 | } | |
638 | ||
639 | ||
4f5bc454 DW |
640 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info) |
641 | { | |
642 | struct intel_super *super = st->sb; | |
4f5bc454 DW |
643 | struct imsm_disk *disk; |
644 | __u32 s; | |
4f5bc454 | 645 | |
bf5a934a DW |
646 | if (super->current_vol >= 0) { |
647 | getinfo_super_imsm_volume(st, info); | |
648 | return; | |
649 | } | |
d23fe947 DW |
650 | |
651 | /* Set raid_disks to zero so that Assemble will always pull in valid | |
652 | * spares | |
653 | */ | |
654 | info->array.raid_disks = 0; | |
cdddbdbc DW |
655 | info->array.level = LEVEL_CONTAINER; |
656 | info->array.layout = 0; | |
657 | info->array.md_minor = -1; | |
c2c087e6 | 658 | info->array.ctime = 0; /* N/A for imsm */ |
cdddbdbc DW |
659 | info->array.utime = 0; |
660 | info->array.chunk_size = 0; | |
661 | ||
662 | info->disk.major = 0; | |
663 | info->disk.minor = 0; | |
cdddbdbc | 664 | info->disk.raid_disk = -1; |
c2c087e6 DW |
665 | info->reshape_active = 0; |
666 | strcpy(info->text_version, "imsm"); | |
667 | info->disk.number = -1; | |
668 | info->disk.state = 0; | |
669 | ||
4a04ec6c | 670 | if (super->disks) { |
b9f594fe | 671 | disk = &super->disks->disk; |
4a04ec6c DW |
672 | info->disk.number = super->disks->index; |
673 | info->disk.raid_disk = super->disks->index; | |
bf5a934a DW |
674 | info->data_offset = __le32_to_cpu(disk->total_blocks) - |
675 | (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS); | |
676 | info->component_size = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
4a04ec6c DW |
677 | s = __le32_to_cpu(disk->status); |
678 | info->disk.state = s & CONFIGURED_DISK ? (1 << MD_DISK_ACTIVE) : 0; | |
679 | info->disk.state |= s & FAILED_DISK ? (1 << MD_DISK_FAULTY) : 0; | |
680 | info->disk.state |= s & USABLE_DISK ? (1 << MD_DISK_SYNC) : 0; | |
cdddbdbc | 681 | } |
cdddbdbc DW |
682 | } |
683 | ||
cdddbdbc DW |
684 | static int update_super_imsm(struct supertype *st, struct mdinfo *info, |
685 | char *update, char *devname, int verbose, | |
686 | int uuid_set, char *homehost) | |
687 | { | |
f352c545 DW |
688 | /* FIXME */ |
689 | ||
690 | /* For 'assemble' and 'force' we need to return non-zero if any | |
691 | * change was made. For others, the return value is ignored. | |
692 | * Update options are: | |
693 | * force-one : This device looks a bit old but needs to be included, | |
694 | * update age info appropriately. | |
695 | * assemble: clear any 'faulty' flag to allow this device to | |
696 | * be assembled. | |
697 | * force-array: Array is degraded but being forced, mark it clean | |
698 | * if that will be needed to assemble it. | |
699 | * | |
700 | * newdev: not used ???? | |
701 | * grow: Array has gained a new device - this is currently for | |
702 | * linear only | |
703 | * resync: mark as dirty so a resync will happen. | |
704 | * name: update the name - preserving the homehost | |
705 | * | |
706 | * Following are not relevant for this imsm: | |
707 | * sparc2.2 : update from old dodgey metadata | |
708 | * super-minor: change the preferred_minor number | |
709 | * summaries: update redundant counters. | |
710 | * uuid: Change the uuid of the array to match watch is given | |
711 | * homehost: update the recorded homehost | |
712 | * _reshape_progress: record new reshape_progress position. | |
713 | */ | |
714 | int rv = 0; | |
715 | //struct intel_super *super = st->sb; | |
716 | //struct imsm_super *mpb = super->mpb; | |
717 | ||
718 | if (strcmp(update, "grow") == 0) { | |
719 | } | |
720 | if (strcmp(update, "resync") == 0) { | |
721 | /* dev->vol.dirty = 1; */ | |
722 | } | |
723 | ||
724 | /* IMSM has no concept of UUID or homehost */ | |
725 | ||
726 | return rv; | |
cdddbdbc DW |
727 | } |
728 | ||
c2c087e6 | 729 | static size_t disks_to_mpb_size(int disks) |
cdddbdbc | 730 | { |
c2c087e6 | 731 | size_t size; |
cdddbdbc | 732 | |
c2c087e6 DW |
733 | size = sizeof(struct imsm_super); |
734 | size += (disks - 1) * sizeof(struct imsm_disk); | |
735 | size += 2 * sizeof(struct imsm_dev); | |
736 | /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */ | |
737 | size += (4 - 2) * sizeof(struct imsm_map); | |
738 | /* 4 possible disk_ord_tbl's */ | |
739 | size += 4 * (disks - 1) * sizeof(__u32); | |
740 | ||
741 | return size; | |
742 | } | |
743 | ||
744 | static __u64 avail_size_imsm(struct supertype *st, __u64 devsize) | |
745 | { | |
746 | if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS)) | |
747 | return 0; | |
748 | ||
749 | return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS); | |
cdddbdbc DW |
750 | } |
751 | ||
752 | static int compare_super_imsm(struct supertype *st, struct supertype *tst) | |
753 | { | |
754 | /* | |
755 | * return: | |
756 | * 0 same, or first was empty, and second was copied | |
757 | * 1 second had wrong number | |
758 | * 2 wrong uuid | |
759 | * 3 wrong other info | |
760 | */ | |
761 | struct intel_super *first = st->sb; | |
762 | struct intel_super *sec = tst->sb; | |
763 | ||
764 | if (!first) { | |
765 | st->sb = tst->sb; | |
766 | tst->sb = NULL; | |
767 | return 0; | |
768 | } | |
769 | ||
949c47a0 | 770 | if (memcmp(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH) != 0) |
cdddbdbc | 771 | return 3; |
d23fe947 DW |
772 | |
773 | /* if an anchor does not have num_raid_devs set then it is a free | |
774 | * floating spare | |
775 | */ | |
776 | if (first->anchor->num_raid_devs > 0 && | |
777 | sec->anchor->num_raid_devs > 0) { | |
778 | if (first->anchor->family_num != sec->anchor->family_num) | |
779 | return 3; | |
780 | if (first->anchor->mpb_size != sec->anchor->mpb_size) | |
781 | return 3; | |
782 | if (first->anchor->check_sum != sec->anchor->check_sum) | |
783 | return 3; | |
784 | } | |
cdddbdbc DW |
785 | |
786 | return 0; | |
787 | } | |
788 | ||
0030e8d6 DW |
789 | static void fd2devname(int fd, char *name) |
790 | { | |
791 | struct stat st; | |
792 | char path[256]; | |
793 | char dname[100]; | |
794 | char *nm; | |
795 | int rv; | |
796 | ||
797 | name[0] = '\0'; | |
798 | if (fstat(fd, &st) != 0) | |
799 | return; | |
800 | sprintf(path, "/sys/dev/block/%d:%d", | |
801 | major(st.st_rdev), minor(st.st_rdev)); | |
802 | ||
803 | rv = readlink(path, dname, sizeof(dname)); | |
804 | if (rv <= 0) | |
805 | return; | |
806 | ||
807 | dname[rv] = '\0'; | |
808 | nm = strrchr(dname, '/'); | |
809 | nm++; | |
810 | snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm); | |
811 | } | |
812 | ||
813 | ||
cdddbdbc DW |
814 | extern int scsi_get_serial(int fd, void *buf, size_t buf_len); |
815 | ||
816 | static int imsm_read_serial(int fd, char *devname, | |
817 | __u8 serial[MAX_RAID_SERIAL_LEN]) | |
818 | { | |
819 | unsigned char scsi_serial[255]; | |
cdddbdbc DW |
820 | int rv; |
821 | int rsp_len; | |
822 | int i, cnt; | |
823 | ||
824 | memset(scsi_serial, 0, sizeof(scsi_serial)); | |
cdddbdbc | 825 | |
0030e8d6 DW |
826 | if (imsm_env_devname_as_serial()) { |
827 | char name[MAX_RAID_SERIAL_LEN]; | |
828 | ||
829 | fd2devname(fd, name); | |
830 | strcpy((char *) serial, name); | |
831 | return 0; | |
832 | } | |
833 | ||
755c99fa | 834 | rv = scsi_get_serial(fd, scsi_serial, sizeof(scsi_serial)); |
cdddbdbc DW |
835 | |
836 | if (rv != 0) { | |
837 | if (devname) | |
838 | fprintf(stderr, | |
839 | Name ": Failed to retrieve serial for %s\n", | |
840 | devname); | |
841 | return rv; | |
842 | } | |
843 | ||
844 | rsp_len = scsi_serial[3]; | |
845 | for (i = 0, cnt = 0; i < rsp_len; i++) { | |
846 | if (!isspace(scsi_serial[4 + i])) | |
847 | serial[cnt++] = scsi_serial[4 + i]; | |
848 | if (cnt == MAX_RAID_SERIAL_LEN) | |
849 | break; | |
850 | } | |
851 | ||
852 | serial[MAX_RAID_SERIAL_LEN - 1] = '\0'; | |
853 | ||
854 | return 0; | |
855 | } | |
856 | ||
857 | static int | |
858 | load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd) | |
859 | { | |
cdddbdbc DW |
860 | struct dl *dl; |
861 | struct stat stb; | |
cdddbdbc DW |
862 | int rv; |
863 | int i; | |
d23fe947 DW |
864 | int alloc = 1; |
865 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
866 | ||
867 | rv = imsm_read_serial(fd, devname, serial); | |
868 | ||
869 | if (rv != 0) | |
870 | return 2; | |
871 | ||
872 | /* check if this is a disk we have seen before. it may be a spare in | |
873 | * super->disks while the current anchor believes it is a raid member, | |
874 | * check if we need to update dl->index | |
875 | */ | |
876 | for (dl = super->disks; dl; dl = dl->next) | |
877 | if (memcmp(dl->serial, serial, MAX_RAID_SERIAL_LEN) == 0) | |
878 | break; | |
879 | ||
880 | if (!dl) | |
881 | dl = malloc(sizeof(*dl)); | |
882 | else | |
883 | alloc = 0; | |
cdddbdbc | 884 | |
b9f594fe | 885 | if (!dl) { |
cdddbdbc DW |
886 | if (devname) |
887 | fprintf(stderr, | |
888 | Name ": failed to allocate disk buffer for %s\n", | |
889 | devname); | |
890 | return 2; | |
891 | } | |
cdddbdbc | 892 | |
d23fe947 DW |
893 | if (alloc) { |
894 | fstat(fd, &stb); | |
895 | dl->major = major(stb.st_rdev); | |
896 | dl->minor = minor(stb.st_rdev); | |
897 | dl->next = super->disks; | |
898 | dl->fd = keep_fd ? fd : -1; | |
899 | dl->devname = devname ? strdup(devname) : NULL; | |
900 | strncpy((char *) dl->serial, (char *) serial, MAX_RAID_SERIAL_LEN); | |
901 | } else if (keep_fd) { | |
902 | close(dl->fd); | |
903 | dl->fd = fd; | |
904 | } | |
cdddbdbc | 905 | |
d23fe947 | 906 | /* look up this disk's index in the current anchor */ |
949c47a0 DW |
907 | for (i = 0; i < super->anchor->num_disks; i++) { |
908 | struct imsm_disk *disk_iter; | |
909 | ||
910 | disk_iter = __get_imsm_disk(super->anchor, i); | |
cdddbdbc | 911 | |
949c47a0 DW |
912 | if (memcmp(disk_iter->serial, dl->serial, |
913 | MAX_RAID_SERIAL_LEN) == 0) { | |
d23fe947 DW |
914 | __u32 status; |
915 | ||
b9f594fe | 916 | dl->disk = *disk_iter; |
d23fe947 DW |
917 | status = __le32_to_cpu(dl->disk.status); |
918 | /* only set index on disks that are a member of a | |
919 | * populated contianer, i.e. one with raid_devs | |
920 | */ | |
921 | if (status & SPARE_DISK) | |
922 | dl->index = -1; | |
923 | else | |
924 | dl->index = i; | |
cdddbdbc | 925 | break; |
949c47a0 | 926 | } |
cdddbdbc DW |
927 | } |
928 | ||
d23fe947 | 929 | if (i == super->anchor->num_disks && alloc) { |
e9d82038 DW |
930 | if (devname) |
931 | fprintf(stderr, | |
d23fe947 | 932 | Name ": failed to load disk with serial \'%s\' for %s\n", |
e9d82038 | 933 | dl->serial, devname); |
d23fe947 DW |
934 | free(dl); |
935 | return 1; | |
936 | } | |
937 | if (i == super->anchor->num_disks && dl->index >= 0) { | |
938 | if (devname) | |
939 | fprintf(stderr, | |
940 | Name ": confused... disk %d with serial \'%s\' " | |
941 | "is not listed in the current anchor\n", | |
942 | dl->index, dl->serial); | |
943 | return 1; | |
e9d82038 | 944 | } |
cdddbdbc | 945 | |
d23fe947 DW |
946 | if (alloc) |
947 | super->disks = dl; | |
948 | ||
949c47a0 DW |
949 | return 0; |
950 | } | |
951 | ||
952 | static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src) | |
953 | { | |
3393c6af DW |
954 | memcpy(dest, src, sizeof_imsm_dev(src, 0)); |
955 | } | |
956 | ||
957 | static void dup_map(struct imsm_dev *dev) | |
958 | { | |
959 | struct imsm_map *dest = get_imsm_map(dev, 1); | |
960 | struct imsm_map *src = get_imsm_map(dev, 0); | |
961 | ||
962 | memcpy(dest, src, sizeof_imsm_map(src)); | |
949c47a0 DW |
963 | } |
964 | ||
965 | static int parse_raid_devices(struct intel_super *super) | |
966 | { | |
967 | int i; | |
968 | struct imsm_dev *dev_new; | |
969 | size_t len; | |
970 | ||
971 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
972 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
973 | ||
3393c6af | 974 | len = sizeof_imsm_dev(dev_iter, 1); |
949c47a0 DW |
975 | dev_new = malloc(len); |
976 | if (!dev_new) | |
977 | return 1; | |
978 | imsm_copy_dev(dev_new, dev_iter); | |
979 | super->dev_tbl[i] = dev_new; | |
980 | } | |
cdddbdbc DW |
981 | |
982 | return 0; | |
983 | } | |
984 | ||
604b746f JD |
985 | /* retrieve a pointer to the bbm log which starts after all raid devices */ |
986 | struct bbm_log *__get_imsm_bbm_log(struct imsm_super *mpb) | |
987 | { | |
988 | void *ptr = NULL; | |
989 | ||
990 | if (__le32_to_cpu(mpb->bbm_log_size)) { | |
991 | ptr = mpb; | |
992 | ptr += mpb->mpb_size - __le32_to_cpu(mpb->bbm_log_size); | |
993 | } | |
994 | ||
995 | return ptr; | |
996 | } | |
997 | ||
d23fe947 | 998 | static void __free_imsm(struct intel_super *super, int free_disks); |
9ca2c81c | 999 | |
cdddbdbc DW |
1000 | /* load_imsm_mpb - read matrix metadata |
1001 | * allocates super->mpb to be freed by free_super | |
1002 | */ | |
1003 | static int load_imsm_mpb(int fd, struct intel_super *super, char *devname) | |
1004 | { | |
1005 | unsigned long long dsize; | |
cdddbdbc DW |
1006 | unsigned long long sectors; |
1007 | struct stat; | |
6416d527 | 1008 | struct imsm_super *anchor; |
cdddbdbc | 1009 | __u32 check_sum; |
949c47a0 | 1010 | int rc; |
cdddbdbc | 1011 | |
cdddbdbc DW |
1012 | get_dev_size(fd, NULL, &dsize); |
1013 | ||
1014 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) { | |
1015 | if (devname) | |
1016 | fprintf(stderr, | |
1017 | Name ": Cannot seek to anchor block on %s: %s\n", | |
1018 | devname, strerror(errno)); | |
1019 | return 1; | |
1020 | } | |
1021 | ||
949c47a0 | 1022 | if (posix_memalign((void**)&anchor, 512, 512) != 0) { |
ad97895e DW |
1023 | if (devname) |
1024 | fprintf(stderr, | |
1025 | Name ": Failed to allocate imsm anchor buffer" | |
1026 | " on %s\n", devname); | |
1027 | return 1; | |
1028 | } | |
949c47a0 | 1029 | if (read(fd, anchor, 512) != 512) { |
cdddbdbc DW |
1030 | if (devname) |
1031 | fprintf(stderr, | |
1032 | Name ": Cannot read anchor block on %s: %s\n", | |
1033 | devname, strerror(errno)); | |
6416d527 | 1034 | free(anchor); |
cdddbdbc DW |
1035 | return 1; |
1036 | } | |
1037 | ||
6416d527 | 1038 | if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) { |
cdddbdbc DW |
1039 | if (devname) |
1040 | fprintf(stderr, | |
1041 | Name ": no IMSM anchor on %s\n", devname); | |
6416d527 | 1042 | free(anchor); |
cdddbdbc DW |
1043 | return 2; |
1044 | } | |
1045 | ||
d23fe947 | 1046 | __free_imsm(super, 0); |
949c47a0 DW |
1047 | super->len = __le32_to_cpu(anchor->mpb_size); |
1048 | super->len = ROUND_UP(anchor->mpb_size, 512); | |
1049 | if (posix_memalign(&super->buf, 512, super->len) != 0) { | |
cdddbdbc DW |
1050 | if (devname) |
1051 | fprintf(stderr, | |
1052 | Name ": unable to allocate %zu byte mpb buffer\n", | |
949c47a0 | 1053 | super->len); |
6416d527 | 1054 | free(anchor); |
cdddbdbc DW |
1055 | return 2; |
1056 | } | |
949c47a0 | 1057 | memcpy(super->buf, anchor, 512); |
cdddbdbc | 1058 | |
6416d527 NB |
1059 | sectors = mpb_sectors(anchor) - 1; |
1060 | free(anchor); | |
949c47a0 DW |
1061 | if (!sectors) { |
1062 | rc = load_imsm_disk(fd, super, devname, 0); | |
1063 | if (rc == 0) | |
1064 | rc = parse_raid_devices(super); | |
1065 | return rc; | |
1066 | } | |
cdddbdbc DW |
1067 | |
1068 | /* read the extended mpb */ | |
1069 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) { | |
1070 | if (devname) | |
1071 | fprintf(stderr, | |
1072 | Name ": Cannot seek to extended mpb on %s: %s\n", | |
1073 | devname, strerror(errno)); | |
1074 | return 1; | |
1075 | } | |
1076 | ||
949c47a0 | 1077 | if (read(fd, super->buf + 512, super->len - 512) != super->len - 512) { |
cdddbdbc DW |
1078 | if (devname) |
1079 | fprintf(stderr, | |
1080 | Name ": Cannot read extended mpb on %s: %s\n", | |
1081 | devname, strerror(errno)); | |
1082 | return 2; | |
1083 | } | |
1084 | ||
949c47a0 DW |
1085 | check_sum = __gen_imsm_checksum(super->anchor); |
1086 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
cdddbdbc DW |
1087 | if (devname) |
1088 | fprintf(stderr, | |
1089 | Name ": IMSM checksum %x != %x on %s\n", | |
949c47a0 | 1090 | check_sum, __le32_to_cpu(super->anchor->check_sum), |
cdddbdbc DW |
1091 | devname); |
1092 | return 2; | |
1093 | } | |
1094 | ||
604b746f JD |
1095 | /* FIXME the BBM log is disk specific so we cannot use this global |
1096 | * buffer for all disks. Ok for now since we only look at the global | |
1097 | * bbm_log_size parameter to gate assembly | |
1098 | */ | |
1099 | super->bbm_log = __get_imsm_bbm_log(super->anchor); | |
1100 | ||
949c47a0 DW |
1101 | rc = load_imsm_disk(fd, super, devname, 0); |
1102 | if (rc == 0) | |
1103 | rc = parse_raid_devices(super); | |
1104 | return rc; | |
cdddbdbc DW |
1105 | } |
1106 | ||
cdddbdbc DW |
1107 | static void free_imsm_disks(struct intel_super *super) |
1108 | { | |
1109 | while (super->disks) { | |
1110 | struct dl *d = super->disks; | |
1111 | ||
1112 | super->disks = d->next; | |
1113 | if (d->fd >= 0) | |
1114 | close(d->fd); | |
1115 | if (d->devname) | |
1116 | free(d->devname); | |
1117 | free(d); | |
1118 | } | |
1119 | } | |
1120 | ||
9ca2c81c | 1121 | /* free all the pieces hanging off of a super pointer */ |
d23fe947 | 1122 | static void __free_imsm(struct intel_super *super, int free_disks) |
cdddbdbc | 1123 | { |
949c47a0 DW |
1124 | int i; |
1125 | ||
9ca2c81c | 1126 | if (super->buf) { |
949c47a0 | 1127 | free(super->buf); |
9ca2c81c DW |
1128 | super->buf = NULL; |
1129 | } | |
d23fe947 DW |
1130 | if (free_disks) |
1131 | free_imsm_disks(super); | |
949c47a0 | 1132 | for (i = 0; i < IMSM_MAX_RAID_DEVS; i++) |
9ca2c81c | 1133 | if (super->dev_tbl[i]) { |
949c47a0 | 1134 | free(super->dev_tbl[i]); |
9ca2c81c DW |
1135 | super->dev_tbl[i] = NULL; |
1136 | } | |
cdddbdbc DW |
1137 | } |
1138 | ||
9ca2c81c DW |
1139 | static void free_imsm(struct intel_super *super) |
1140 | { | |
d23fe947 | 1141 | __free_imsm(super, 1); |
9ca2c81c DW |
1142 | free(super); |
1143 | } | |
cdddbdbc DW |
1144 | |
1145 | static void free_super_imsm(struct supertype *st) | |
1146 | { | |
1147 | struct intel_super *super = st->sb; | |
1148 | ||
1149 | if (!super) | |
1150 | return; | |
1151 | ||
1152 | free_imsm(super); | |
1153 | st->sb = NULL; | |
1154 | } | |
1155 | ||
c2c087e6 DW |
1156 | static struct intel_super *alloc_super(int creating_imsm) |
1157 | { | |
1158 | struct intel_super *super = malloc(sizeof(*super)); | |
1159 | ||
1160 | if (super) { | |
1161 | memset(super, 0, sizeof(*super)); | |
1162 | super->creating_imsm = creating_imsm; | |
bf5a934a | 1163 | super->current_vol = -1; |
c2c087e6 DW |
1164 | } |
1165 | ||
1166 | return super; | |
1167 | } | |
1168 | ||
cdddbdbc DW |
1169 | #ifndef MDASSEMBLE |
1170 | static int load_super_imsm_all(struct supertype *st, int fd, void **sbp, | |
1171 | char *devname, int keep_fd) | |
1172 | { | |
1173 | struct mdinfo *sra; | |
1174 | struct intel_super *super; | |
1175 | struct mdinfo *sd, *best = NULL; | |
1176 | __u32 bestgen = 0; | |
1177 | __u32 gen; | |
1178 | char nm[20]; | |
1179 | int dfd; | |
1180 | int rv; | |
1181 | ||
1182 | /* check if this disk is a member of an active array */ | |
1183 | sra = sysfs_read(fd, 0, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE); | |
1184 | if (!sra) | |
1185 | return 1; | |
1186 | ||
1187 | if (sra->array.major_version != -1 || | |
1188 | sra->array.minor_version != -2 || | |
1189 | strcmp(sra->text_version, "imsm") != 0) | |
1190 | return 1; | |
1191 | ||
c2c087e6 | 1192 | super = alloc_super(0); |
cdddbdbc DW |
1193 | if (!super) |
1194 | return 1; | |
1195 | ||
d23fe947 | 1196 | /* find the most up to date disk in this array, skipping spares */ |
cdddbdbc DW |
1197 | for (sd = sra->devs; sd; sd = sd->next) { |
1198 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); | |
1199 | dfd = dev_open(nm, keep_fd ? O_RDWR : O_RDONLY); | |
1200 | if (!dfd) { | |
1201 | free_imsm(super); | |
1202 | return 2; | |
1203 | } | |
1204 | rv = load_imsm_mpb(dfd, super, NULL); | |
1205 | if (!keep_fd) | |
1206 | close(dfd); | |
1207 | if (rv == 0) { | |
d23fe947 DW |
1208 | if (super->anchor->num_raid_devs == 0) |
1209 | gen = 0; | |
1210 | else | |
1211 | gen = __le32_to_cpu(super->anchor->generation_num); | |
cdddbdbc DW |
1212 | if (!best || gen > bestgen) { |
1213 | bestgen = gen; | |
1214 | best = sd; | |
1215 | } | |
1216 | } else { | |
1217 | free_imsm(super); | |
1218 | return 2; | |
1219 | } | |
1220 | } | |
1221 | ||
1222 | if (!best) { | |
1223 | free_imsm(super); | |
1224 | return 1; | |
1225 | } | |
1226 | ||
1227 | /* load the most up to date anchor */ | |
1228 | sprintf(nm, "%d:%d", best->disk.major, best->disk.minor); | |
1229 | dfd = dev_open(nm, O_RDONLY); | |
1230 | if (!dfd) { | |
1231 | free_imsm(super); | |
1232 | return 1; | |
1233 | } | |
1234 | rv = load_imsm_mpb(dfd, super, NULL); | |
1235 | close(dfd); | |
1236 | if (rv != 0) { | |
1237 | free_imsm(super); | |
1238 | return 2; | |
1239 | } | |
1240 | ||
d23fe947 | 1241 | /* re-parse the disk list with the current anchor */ |
cdddbdbc DW |
1242 | for (sd = sra->devs ; sd ; sd = sd->next) { |
1243 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); | |
1244 | dfd = dev_open(nm, keep_fd? O_RDWR : O_RDONLY); | |
1245 | if (!dfd) { | |
1246 | free_imsm(super); | |
1247 | return 2; | |
1248 | } | |
1249 | load_imsm_disk(dfd, super, NULL, keep_fd); | |
1250 | if (!keep_fd) | |
1251 | close(dfd); | |
1252 | } | |
1253 | ||
f7e7067b | 1254 | if (st->subarray[0]) { |
949c47a0 | 1255 | if (atoi(st->subarray) <= super->anchor->num_raid_devs) |
bf5a934a DW |
1256 | super->current_vol = atoi(st->subarray); |
1257 | else | |
1258 | return 1; | |
f7e7067b NB |
1259 | } |
1260 | ||
cdddbdbc DW |
1261 | *sbp = super; |
1262 | if (st->ss == NULL) { | |
bf5a934a | 1263 | st->ss = &super_imsm; |
cdddbdbc DW |
1264 | st->minor_version = 0; |
1265 | st->max_devs = IMSM_MAX_DEVICES; | |
f4d11639 | 1266 | st->container_dev = fd2devnum(fd); |
cdddbdbc DW |
1267 | } |
1268 | ||
1269 | return 0; | |
1270 | } | |
1271 | #endif | |
1272 | ||
1273 | static int load_super_imsm(struct supertype *st, int fd, char *devname) | |
1274 | { | |
1275 | struct intel_super *super; | |
1276 | int rv; | |
1277 | ||
1278 | #ifndef MDASSEMBLE | |
3dbccbcf | 1279 | if (load_super_imsm_all(st, fd, &st->sb, devname, 1) == 0) |
cdddbdbc DW |
1280 | return 0; |
1281 | #endif | |
f7e7067b NB |
1282 | if (st->subarray[0]) |
1283 | return 1; /* FIXME */ | |
cdddbdbc | 1284 | |
c2c087e6 | 1285 | super = alloc_super(0); |
cdddbdbc DW |
1286 | if (!super) { |
1287 | fprintf(stderr, | |
1288 | Name ": malloc of %zu failed.\n", | |
1289 | sizeof(*super)); | |
1290 | return 1; | |
1291 | } | |
1292 | ||
1293 | rv = load_imsm_mpb(fd, super, devname); | |
1294 | ||
1295 | if (rv) { | |
1296 | if (devname) | |
1297 | fprintf(stderr, | |
1298 | Name ": Failed to load all information " | |
1299 | "sections on %s\n", devname); | |
1300 | free_imsm(super); | |
1301 | return rv; | |
1302 | } | |
1303 | ||
1304 | st->sb = super; | |
1305 | if (st->ss == NULL) { | |
1306 | st->ss = &super_imsm; | |
1307 | st->minor_version = 0; | |
1308 | st->max_devs = IMSM_MAX_DEVICES; | |
1309 | } | |
1310 | ||
1311 | return 0; | |
1312 | } | |
1313 | ||
ef6ffade DW |
1314 | static __u16 info_to_blocks_per_strip(mdu_array_info_t *info) |
1315 | { | |
1316 | if (info->level == 1) | |
1317 | return 128; | |
1318 | return info->chunk_size >> 9; | |
1319 | } | |
1320 | ||
1321 | static __u32 info_to_num_data_stripes(mdu_array_info_t *info) | |
1322 | { | |
1323 | __u32 num_stripes; | |
1324 | ||
1325 | num_stripes = (info->size * 2) / info_to_blocks_per_strip(info); | |
1326 | if (info->level == 1) | |
1327 | num_stripes /= 2; | |
1328 | ||
1329 | return num_stripes; | |
1330 | } | |
1331 | ||
fcfd9599 DW |
1332 | static __u32 info_to_blocks_per_member(mdu_array_info_t *info) |
1333 | { | |
1334 | return (info->size * 2) & ~(info_to_blocks_per_strip(info) - 1); | |
1335 | } | |
1336 | ||
8b353278 DW |
1337 | static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info, |
1338 | unsigned long long size, char *name, | |
1339 | char *homehost, int *uuid) | |
cdddbdbc | 1340 | { |
c2c087e6 DW |
1341 | /* We are creating a volume inside a pre-existing container. |
1342 | * so st->sb is already set. | |
1343 | */ | |
1344 | struct intel_super *super = st->sb; | |
949c47a0 | 1345 | struct imsm_super *mpb = super->anchor; |
c2c087e6 DW |
1346 | struct imsm_dev *dev; |
1347 | struct imsm_vol *vol; | |
1348 | struct imsm_map *map; | |
1349 | int idx = mpb->num_raid_devs; | |
1350 | int i; | |
1351 | unsigned long long array_blocks; | |
c2c087e6 | 1352 | __u32 offset = 0; |
2c092cad | 1353 | size_t size_old, size_new; |
cdddbdbc | 1354 | |
c2c087e6 DW |
1355 | if (mpb->num_raid_devs >= 2) { |
1356 | fprintf(stderr, Name": This imsm-container already has the " | |
1357 | "maximum of 2 volumes\n"); | |
1358 | return 0; | |
1359 | } | |
1360 | ||
2c092cad DW |
1361 | /* ensure the mpb is large enough for the new data */ |
1362 | size_old = __le32_to_cpu(mpb->mpb_size); | |
1363 | size_new = disks_to_mpb_size(info->nr_disks); | |
1364 | if (size_new > size_old) { | |
1365 | void *mpb_new; | |
1366 | size_t size_round = ROUND_UP(size_new, 512); | |
1367 | ||
1368 | if (posix_memalign(&mpb_new, 512, size_round) != 0) { | |
1369 | fprintf(stderr, Name": could not allocate new mpb\n"); | |
1370 | return 0; | |
1371 | } | |
1372 | memcpy(mpb_new, mpb, size_old); | |
1373 | free(mpb); | |
1374 | mpb = mpb_new; | |
949c47a0 | 1375 | super->anchor = mpb_new; |
2c092cad DW |
1376 | mpb->mpb_size = __cpu_to_le32(size_new); |
1377 | memset(mpb_new + size_old, 0, size_round - size_old); | |
1378 | } | |
bf5a934a | 1379 | super->current_vol = idx; |
d23fe947 DW |
1380 | /* when creating the first raid device in this container set num_disks |
1381 | * to zero, i.e. delete this spare and add raid member devices in | |
1382 | * add_to_super_imsm_volume() | |
1383 | */ | |
1384 | if (super->current_vol == 0) | |
1385 | mpb->num_disks = 0; | |
bf5a934a | 1386 | sprintf(st->subarray, "%d", idx); |
949c47a0 DW |
1387 | dev = malloc(sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1)); |
1388 | if (!dev) { | |
1389 | fprintf(stderr, Name": could not allocate raid device\n"); | |
1390 | return 0; | |
1391 | } | |
c2c087e6 DW |
1392 | strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN); |
1393 | array_blocks = calc_array_size(info->level, info->raid_disks, | |
1394 | info->layout, info->chunk_size, | |
1395 | info->size*2); | |
1396 | dev->size_low = __cpu_to_le32((__u32) array_blocks); | |
1397 | dev->size_high = __cpu_to_le32((__u32) (array_blocks >> 32)); | |
1398 | dev->status = __cpu_to_le32(0); | |
1399 | dev->reserved_blocks = __cpu_to_le32(0); | |
1400 | vol = &dev->vol; | |
1401 | vol->migr_state = 0; | |
1402 | vol->migr_type = 0; | |
1403 | vol->dirty = 0; | |
1404 | for (i = 0; i < idx; i++) { | |
949c47a0 | 1405 | struct imsm_dev *prev = get_imsm_dev(super, i); |
a965f303 | 1406 | struct imsm_map *pmap = get_imsm_map(prev, 0); |
c2c087e6 DW |
1407 | |
1408 | offset += __le32_to_cpu(pmap->blocks_per_member); | |
1409 | offset += IMSM_RESERVED_SECTORS; | |
1410 | } | |
a965f303 | 1411 | map = get_imsm_map(dev, 0); |
c2c087e6 | 1412 | map->pba_of_lba0 = __cpu_to_le32(offset); |
fcfd9599 | 1413 | map->blocks_per_member = __cpu_to_le32(info_to_blocks_per_member(info)); |
ef6ffade DW |
1414 | map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info)); |
1415 | map->num_data_stripes = __cpu_to_le32(info_to_num_data_stripes(info)); | |
c2c087e6 DW |
1416 | map->map_state = info->level ? IMSM_T_STATE_UNINITIALIZED : |
1417 | IMSM_T_STATE_NORMAL; | |
ef6ffade DW |
1418 | |
1419 | if (info->level == 1 && info->raid_disks > 2) { | |
1420 | fprintf(stderr, Name": imsm does not support more than 2 disks" | |
1421 | "in a raid1 volume\n"); | |
1422 | return 0; | |
1423 | } | |
c2c087e6 DW |
1424 | if (info->level == 10) |
1425 | map->raid_level = 1; | |
1426 | else | |
1427 | map->raid_level = info->level; | |
ef6ffade | 1428 | |
c2c087e6 DW |
1429 | map->num_members = info->raid_disks; |
1430 | for (i = 0; i < map->num_members; i++) { | |
1431 | /* initialized in add_to_super */ | |
1432 | map->disk_ord_tbl[i] = __cpu_to_le32(0); | |
1433 | } | |
949c47a0 DW |
1434 | mpb->num_raid_devs++; |
1435 | super->dev_tbl[super->current_vol] = dev; | |
c2c087e6 DW |
1436 | |
1437 | return 1; | |
cdddbdbc DW |
1438 | } |
1439 | ||
bf5a934a DW |
1440 | static int init_super_imsm(struct supertype *st, mdu_array_info_t *info, |
1441 | unsigned long long size, char *name, | |
1442 | char *homehost, int *uuid) | |
1443 | { | |
1444 | /* This is primarily called by Create when creating a new array. | |
1445 | * We will then get add_to_super called for each component, and then | |
1446 | * write_init_super called to write it out to each device. | |
1447 | * For IMSM, Create can create on fresh devices or on a pre-existing | |
1448 | * array. | |
1449 | * To create on a pre-existing array a different method will be called. | |
1450 | * This one is just for fresh drives. | |
1451 | */ | |
1452 | struct intel_super *super; | |
1453 | struct imsm_super *mpb; | |
1454 | size_t mpb_size; | |
1455 | ||
1456 | if (!info) { | |
1457 | st->sb = NULL; | |
1458 | return 0; | |
1459 | } | |
1460 | if (st->sb) | |
1461 | return init_super_imsm_volume(st, info, size, name, homehost, | |
1462 | uuid); | |
1463 | ||
1464 | super = alloc_super(1); | |
1465 | if (!super) | |
1466 | return 0; | |
1467 | mpb_size = disks_to_mpb_size(info->nr_disks); | |
ef649044 | 1468 | if (posix_memalign(&super->buf, 512, mpb_size) != 0) { |
bf5a934a DW |
1469 | free(super); |
1470 | return 0; | |
1471 | } | |
ef649044 | 1472 | mpb = super->buf; |
bf5a934a DW |
1473 | memset(mpb, 0, mpb_size); |
1474 | ||
1475 | memcpy(mpb->sig, MPB_SIGNATURE, strlen(MPB_SIGNATURE)); | |
1476 | memcpy(mpb->sig + strlen(MPB_SIGNATURE), MPB_VERSION_RAID5, | |
1477 | strlen(MPB_VERSION_RAID5)); | |
1478 | mpb->mpb_size = mpb_size; | |
1479 | ||
bf5a934a DW |
1480 | st->sb = super; |
1481 | return 1; | |
1482 | } | |
1483 | ||
1484 | static void add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk, | |
1485 | int fd, char *devname) | |
1486 | { | |
1487 | struct intel_super *super = st->sb; | |
d23fe947 | 1488 | struct imsm_super *mpb = super->anchor; |
bf5a934a DW |
1489 | struct dl *dl; |
1490 | struct imsm_dev *dev; | |
1491 | struct imsm_map *map; | |
bf5a934a DW |
1492 | __u32 status; |
1493 | ||
949c47a0 | 1494 | dev = get_imsm_dev(super, super->current_vol); |
a965f303 | 1495 | map = get_imsm_map(dev, 0); |
bf5a934a DW |
1496 | |
1497 | for (dl = super->disks; dl ; dl = dl->next) | |
1498 | if (dl->major == dk->major && | |
1499 | dl->minor == dk->minor) | |
1500 | break; | |
d23fe947 | 1501 | |
bf5a934a DW |
1502 | if (!dl || ! (dk->state & (1<<MD_DISK_SYNC))) |
1503 | return; | |
1504 | ||
d23fe947 DW |
1505 | /* add a pristine spare to the metadata */ |
1506 | if (dl->index < 0) { | |
1507 | dl->index = super->anchor->num_disks; | |
1508 | super->anchor->num_disks++; | |
1509 | } | |
bf5a934a | 1510 | map->disk_ord_tbl[dk->number] = __cpu_to_le32(dl->index); |
bf5a934a | 1511 | status = CONFIGURED_DISK | USABLE_DISK; |
d23fe947 DW |
1512 | dl->disk.status = __cpu_to_le32(status); |
1513 | ||
1514 | /* if we are creating the first raid device update the family number */ | |
1515 | if (super->current_vol == 0) { | |
1516 | __u32 sum; | |
1517 | struct imsm_dev *_dev = __get_imsm_dev(mpb, 0); | |
1518 | struct imsm_disk *_disk = __get_imsm_disk(mpb, dl->index); | |
1519 | ||
1520 | *_dev = *dev; | |
1521 | *_disk = dl->disk; | |
1522 | sum = __gen_imsm_checksum(mpb); | |
1523 | mpb->family_num = __cpu_to_le32(sum); | |
1524 | } | |
bf5a934a DW |
1525 | } |
1526 | ||
c2c087e6 | 1527 | static void add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk, |
cdddbdbc DW |
1528 | int fd, char *devname) |
1529 | { | |
c2c087e6 | 1530 | struct intel_super *super = st->sb; |
c2c087e6 DW |
1531 | struct dl *dd; |
1532 | unsigned long long size; | |
1533 | __u32 status, id; | |
1534 | int rv; | |
1535 | struct stat stb; | |
1536 | ||
bf5a934a DW |
1537 | if (super->current_vol >= 0) { |
1538 | add_to_super_imsm_volume(st, dk, fd, devname); | |
1539 | return; | |
1540 | } | |
1541 | ||
c2c087e6 DW |
1542 | fstat(fd, &stb); |
1543 | dd = malloc(sizeof(*dd)); | |
b9f594fe | 1544 | if (!dd) { |
c2c087e6 DW |
1545 | fprintf(stderr, |
1546 | Name ": malloc failed %s:%d.\n", __func__, __LINE__); | |
1547 | abort(); | |
1548 | } | |
1549 | memset(dd, 0, sizeof(*dd)); | |
1550 | dd->major = major(stb.st_rdev); | |
1551 | dd->minor = minor(stb.st_rdev); | |
b9f594fe | 1552 | dd->index = -1; |
c2c087e6 DW |
1553 | dd->devname = devname ? strdup(devname) : NULL; |
1554 | dd->next = super->disks; | |
1555 | dd->fd = fd; | |
1556 | rv = imsm_read_serial(fd, devname, dd->serial); | |
1557 | if (rv) { | |
1558 | fprintf(stderr, | |
0030e8d6 | 1559 | Name ": failed to retrieve scsi serial, aborting\n"); |
949c47a0 | 1560 | free(dd); |
0030e8d6 | 1561 | abort(); |
c2c087e6 DW |
1562 | } |
1563 | ||
c2c087e6 DW |
1564 | get_dev_size(fd, NULL, &size); |
1565 | size /= 512; | |
1566 | status = USABLE_DISK | SPARE_DISK; | |
b9f594fe DW |
1567 | strcpy((char *) dd->disk.serial, (char *) dd->serial); |
1568 | dd->disk.total_blocks = __cpu_to_le32(size); | |
1569 | dd->disk.status = __cpu_to_le32(status); | |
c2c087e6 | 1570 | if (sysfs_disk_to_scsi_id(fd, &id) == 0) |
b9f594fe | 1571 | dd->disk.scsi_id = __cpu_to_le32(id); |
c2c087e6 | 1572 | else |
b9f594fe | 1573 | dd->disk.scsi_id = __cpu_to_le32(0); |
c2c087e6 | 1574 | super->disks = dd; |
cdddbdbc DW |
1575 | } |
1576 | ||
c2c087e6 DW |
1577 | static int store_imsm_mpb(int fd, struct intel_super *super); |
1578 | ||
d23fe947 DW |
1579 | /* spare records have their own family number and do not have any defined raid |
1580 | * devices | |
1581 | */ | |
1582 | static int write_super_imsm_spares(struct intel_super *super, int doclose) | |
1583 | { | |
1584 | struct imsm_super mpb_save; | |
1585 | struct imsm_super *mpb = super->anchor; | |
1586 | __u32 sum; | |
1587 | struct dl *d; | |
1588 | ||
1589 | mpb_save = *mpb; | |
1590 | mpb->num_raid_devs = 0; | |
1591 | mpb->num_disks = 1; | |
1592 | mpb->mpb_size = sizeof(struct imsm_super); | |
1593 | mpb->generation_num = __cpu_to_le32(1UL); | |
1594 | ||
1595 | for (d = super->disks; d; d = d->next) { | |
1596 | if (d->index >= 0) | |
1597 | continue; | |
1598 | ||
1599 | mpb->disk[0] = d->disk; | |
1600 | sum = __gen_imsm_checksum(mpb); | |
1601 | mpb->family_num = __cpu_to_le32(sum); | |
1602 | sum = __gen_imsm_checksum(mpb); | |
1603 | mpb->check_sum = __cpu_to_le32(sum); | |
1604 | ||
1605 | if (store_imsm_mpb(d->fd, super)) { | |
1606 | fprintf(stderr, "%s: failed for device %d:%d %s\n", | |
1607 | __func__, d->major, d->minor, strerror(errno)); | |
1608 | *mpb = mpb_save; | |
e74255d9 | 1609 | return 1; |
d23fe947 DW |
1610 | } |
1611 | if (doclose) { | |
1612 | close(d->fd); | |
1613 | d->fd = -1; | |
1614 | } | |
1615 | } | |
1616 | ||
1617 | *mpb = mpb_save; | |
e74255d9 | 1618 | return 0; |
d23fe947 DW |
1619 | } |
1620 | ||
c2c087e6 | 1621 | static int write_super_imsm(struct intel_super *super, int doclose) |
cdddbdbc | 1622 | { |
949c47a0 | 1623 | struct imsm_super *mpb = super->anchor; |
c2c087e6 DW |
1624 | struct dl *d; |
1625 | __u32 generation; | |
1626 | __u32 sum; | |
d23fe947 DW |
1627 | int spares = 0; |
1628 | int raid_disks = 0; | |
949c47a0 | 1629 | int i; |
a48ac0a8 | 1630 | __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk); |
cdddbdbc | 1631 | |
c2c087e6 DW |
1632 | /* 'generation' is incremented everytime the metadata is written */ |
1633 | generation = __le32_to_cpu(mpb->generation_num); | |
1634 | generation++; | |
1635 | mpb->generation_num = __cpu_to_le32(generation); | |
1636 | ||
d23fe947 DW |
1637 | for (d = super->disks; d; d = d->next) { |
1638 | if (d->index < 0) | |
1639 | spares++; | |
1640 | else { | |
1641 | raid_disks++; | |
1642 | mpb->disk[d->index] = d->disk; | |
a48ac0a8 | 1643 | mpb_size += sizeof(struct imsm_disk); |
d23fe947 DW |
1644 | } |
1645 | } | |
1646 | if (raid_disks != mpb->num_disks) { | |
1647 | fprintf(stderr, "%s: expected %d disks only found %d\n", | |
1648 | __func__, mpb->num_disks, raid_disks); | |
e74255d9 | 1649 | return 1; |
d23fe947 | 1650 | } |
b9f594fe | 1651 | |
949c47a0 DW |
1652 | for (i = 0; i < mpb->num_raid_devs; i++) { |
1653 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
1654 | ||
1655 | imsm_copy_dev(dev, super->dev_tbl[i]); | |
a48ac0a8 | 1656 | mpb_size += sizeof_imsm_dev(dev, 0); |
949c47a0 | 1657 | } |
a48ac0a8 DW |
1658 | mpb_size += __le32_to_cpu(mpb->bbm_log_size); |
1659 | mpb->mpb_size = __cpu_to_le32(mpb_size); | |
949c47a0 | 1660 | |
c2c087e6 | 1661 | /* recalculate checksum */ |
949c47a0 | 1662 | sum = __gen_imsm_checksum(mpb); |
c2c087e6 DW |
1663 | mpb->check_sum = __cpu_to_le32(sum); |
1664 | ||
d23fe947 | 1665 | /* write the mpb for disks that compose raid devices */ |
c2c087e6 | 1666 | for (d = super->disks; d ; d = d->next) { |
d23fe947 DW |
1667 | if (d->index < 0) |
1668 | continue; | |
c2c087e6 DW |
1669 | if (store_imsm_mpb(d->fd, super)) { |
1670 | fprintf(stderr, "%s: failed for device %d:%d %s\n", | |
1671 | __func__, d->major, d->minor, strerror(errno)); | |
e74255d9 | 1672 | return 1; |
c2c087e6 DW |
1673 | } |
1674 | if (doclose) { | |
1675 | close(d->fd); | |
1676 | d->fd = -1; | |
1677 | } | |
1678 | } | |
1679 | ||
d23fe947 DW |
1680 | if (spares) |
1681 | return write_super_imsm_spares(super, doclose); | |
1682 | ||
e74255d9 | 1683 | return 0; |
c2c087e6 DW |
1684 | } |
1685 | ||
1686 | static int write_init_super_imsm(struct supertype *st) | |
1687 | { | |
8273f55e DW |
1688 | if (st->update_tail) { |
1689 | /* queue the recently created array as a metadata update */ | |
1690 | size_t len; | |
1691 | struct imsm_update_create_array *u; | |
1692 | struct intel_super *super = st->sb; | |
8273f55e | 1693 | struct imsm_dev *dev; |
8273f55e DW |
1694 | struct dl *d; |
1695 | ||
1696 | if (super->current_vol < 0 || | |
949c47a0 | 1697 | !(dev = get_imsm_dev(super, super->current_vol))) { |
8273f55e DW |
1698 | fprintf(stderr, "%s: could not determine sub-array\n", |
1699 | __func__); | |
1700 | return 1; | |
1701 | } | |
1702 | ||
1703 | ||
3393c6af | 1704 | len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0); |
8273f55e DW |
1705 | u = malloc(len); |
1706 | if (!u) { | |
1707 | fprintf(stderr, "%s: failed to allocate update buffer\n", | |
1708 | __func__); | |
1709 | return 1; | |
1710 | } | |
1711 | ||
1712 | u->type = update_create_array; | |
1713 | u->dev_idx = super->current_vol; | |
949c47a0 | 1714 | imsm_copy_dev(&u->dev, dev); |
8273f55e DW |
1715 | append_metadata_update(st, u, len); |
1716 | ||
1717 | for (d = super->disks; d ; d = d->next) { | |
1718 | close(d->fd); | |
1719 | d->fd = -1; | |
1720 | } | |
1721 | ||
1722 | return 0; | |
1723 | } else | |
1724 | return write_super_imsm(st->sb, 1); | |
cdddbdbc DW |
1725 | } |
1726 | ||
1727 | static int store_zero_imsm(struct supertype *st, int fd) | |
1728 | { | |
551c80c1 | 1729 | unsigned long long dsize; |
6416d527 | 1730 | void *buf; |
551c80c1 DW |
1731 | |
1732 | get_dev_size(fd, NULL, &dsize); | |
1733 | ||
1734 | /* first block is stored on second to last sector of the disk */ | |
1735 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) | |
1736 | return 1; | |
1737 | ||
ad97895e DW |
1738 | if (posix_memalign(&buf, 512, 512) != 0) |
1739 | return 1; | |
1740 | ||
eb7ea463 DW |
1741 | memset(buf, 0, 512); |
1742 | if (write(fd, buf, 512) != 512) | |
551c80c1 | 1743 | return 1; |
cdddbdbc DW |
1744 | return 0; |
1745 | } | |
1746 | ||
cdddbdbc DW |
1747 | static int validate_geometry_imsm_container(struct supertype *st, int level, |
1748 | int layout, int raiddisks, int chunk, | |
c2c087e6 | 1749 | unsigned long long size, char *dev, |
2c514b71 NB |
1750 | unsigned long long *freesize, |
1751 | int verbose) | |
cdddbdbc | 1752 | { |
c2c087e6 DW |
1753 | int fd; |
1754 | unsigned long long ldsize; | |
cdddbdbc | 1755 | |
c2c087e6 DW |
1756 | if (level != LEVEL_CONTAINER) |
1757 | return 0; | |
1758 | if (!dev) | |
1759 | return 1; | |
1760 | ||
1761 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
1762 | if (fd < 0) { | |
2c514b71 NB |
1763 | if (verbose) |
1764 | fprintf(stderr, Name ": imsm: Cannot open %s: %s\n", | |
1765 | dev, strerror(errno)); | |
c2c087e6 DW |
1766 | return 0; |
1767 | } | |
1768 | if (!get_dev_size(fd, dev, &ldsize)) { | |
1769 | close(fd); | |
1770 | return 0; | |
1771 | } | |
1772 | close(fd); | |
1773 | ||
1774 | *freesize = avail_size_imsm(st, ldsize >> 9); | |
1775 | ||
1776 | return 1; | |
cdddbdbc DW |
1777 | } |
1778 | ||
c2c087e6 DW |
1779 | /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd |
1780 | * FIX ME add ahci details | |
1781 | */ | |
8b353278 DW |
1782 | static int validate_geometry_imsm_volume(struct supertype *st, int level, |
1783 | int layout, int raiddisks, int chunk, | |
c2c087e6 | 1784 | unsigned long long size, char *dev, |
2c514b71 NB |
1785 | unsigned long long *freesize, |
1786 | int verbose) | |
cdddbdbc | 1787 | { |
c2c087e6 DW |
1788 | struct stat stb; |
1789 | struct intel_super *super = st->sb; | |
1790 | struct dl *dl; | |
1791 | unsigned long long pos = 0; | |
1792 | unsigned long long maxsize; | |
1793 | struct extent *e; | |
1794 | int i; | |
cdddbdbc | 1795 | |
c2c087e6 DW |
1796 | if (level == LEVEL_CONTAINER) |
1797 | return 0; | |
1798 | ||
1799 | if (level == 1 && raiddisks > 2) { | |
2c514b71 NB |
1800 | if (verbose) |
1801 | fprintf(stderr, Name ": imsm does not support more " | |
1802 | "than 2 in a raid1 configuration\n"); | |
c2c087e6 DW |
1803 | return 0; |
1804 | } | |
1805 | ||
1806 | /* We must have the container info already read in. */ | |
1807 | if (!super) | |
1808 | return 0; | |
1809 | ||
1810 | if (!dev) { | |
1811 | /* General test: make sure there is space for | |
2da8544a DW |
1812 | * 'raiddisks' device extents of size 'size' at a given |
1813 | * offset | |
c2c087e6 DW |
1814 | */ |
1815 | unsigned long long minsize = size*2 /* convert to blocks */; | |
2da8544a | 1816 | unsigned long long start_offset = ~0ULL; |
c2c087e6 DW |
1817 | int dcnt = 0; |
1818 | if (minsize == 0) | |
1819 | minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
1820 | for (dl = super->disks; dl ; dl = dl->next) { | |
1821 | int found = 0; | |
1822 | ||
bf5a934a | 1823 | pos = 0; |
c2c087e6 DW |
1824 | i = 0; |
1825 | e = get_extents(super, dl); | |
1826 | if (!e) continue; | |
1827 | do { | |
1828 | unsigned long long esize; | |
1829 | esize = e[i].start - pos; | |
1830 | if (esize >= minsize) | |
1831 | found = 1; | |
2da8544a DW |
1832 | if (found && start_offset == ~0ULL) { |
1833 | start_offset = pos; | |
1834 | break; | |
1835 | } else if (found && pos != start_offset) { | |
1836 | found = 0; | |
1837 | break; | |
1838 | } | |
c2c087e6 DW |
1839 | pos = e[i].start + e[i].size; |
1840 | i++; | |
1841 | } while (e[i-1].size); | |
1842 | if (found) | |
1843 | dcnt++; | |
1844 | free(e); | |
1845 | } | |
1846 | if (dcnt < raiddisks) { | |
2c514b71 NB |
1847 | if (verbose) |
1848 | fprintf(stderr, Name ": imsm: Not enough " | |
1849 | "devices with space for this array " | |
1850 | "(%d < %d)\n", | |
1851 | dcnt, raiddisks); | |
c2c087e6 DW |
1852 | return 0; |
1853 | } | |
1854 | return 1; | |
1855 | } | |
1856 | /* This device must be a member of the set */ | |
1857 | if (stat(dev, &stb) < 0) | |
1858 | return 0; | |
1859 | if ((S_IFMT & stb.st_mode) != S_IFBLK) | |
1860 | return 0; | |
1861 | for (dl = super->disks ; dl ; dl = dl->next) { | |
1862 | if (dl->major == major(stb.st_rdev) && | |
1863 | dl->minor == minor(stb.st_rdev)) | |
1864 | break; | |
1865 | } | |
1866 | if (!dl) { | |
2c514b71 NB |
1867 | if (verbose) |
1868 | fprintf(stderr, Name ": %s is not in the " | |
1869 | "same imsm set\n", dev); | |
c2c087e6 DW |
1870 | return 0; |
1871 | } | |
1872 | e = get_extents(super, dl); | |
1873 | maxsize = 0; | |
1874 | i = 0; | |
1875 | if (e) do { | |
1876 | unsigned long long esize; | |
1877 | esize = e[i].start - pos; | |
1878 | if (esize >= maxsize) | |
1879 | maxsize = esize; | |
1880 | pos = e[i].start + e[i].size; | |
1881 | i++; | |
1882 | } while (e[i-1].size); | |
1883 | *freesize = maxsize; | |
1884 | ||
1885 | return 1; | |
cdddbdbc DW |
1886 | } |
1887 | ||
604b746f JD |
1888 | int imsm_bbm_log_size(struct imsm_super *mpb) |
1889 | { | |
1890 | return __le32_to_cpu(mpb->bbm_log_size); | |
1891 | } | |
1892 | ||
bf5a934a DW |
1893 | static int validate_geometry_imsm(struct supertype *st, int level, int layout, |
1894 | int raiddisks, int chunk, unsigned long long size, | |
1895 | char *dev, unsigned long long *freesize, | |
1896 | int verbose) | |
1897 | { | |
1898 | int fd, cfd; | |
1899 | struct mdinfo *sra; | |
1900 | ||
1901 | /* if given unused devices create a container | |
1902 | * if given given devices in a container create a member volume | |
1903 | */ | |
1904 | if (level == LEVEL_CONTAINER) { | |
1905 | /* Must be a fresh device to add to a container */ | |
1906 | return validate_geometry_imsm_container(st, level, layout, | |
1907 | raiddisks, chunk, size, | |
1908 | dev, freesize, | |
1909 | verbose); | |
1910 | } | |
1911 | ||
1912 | if (st->sb) { | |
1913 | /* creating in a given container */ | |
1914 | return validate_geometry_imsm_volume(st, level, layout, | |
1915 | raiddisks, chunk, size, | |
1916 | dev, freesize, verbose); | |
1917 | } | |
1918 | ||
1919 | /* limit creation to the following levels */ | |
1920 | if (!dev) | |
1921 | switch (level) { | |
1922 | case 0: | |
1923 | case 1: | |
1924 | case 10: | |
1925 | case 5: | |
1926 | break; | |
1927 | default: | |
1928 | return 1; | |
1929 | } | |
1930 | ||
1931 | /* This device needs to be a device in an 'imsm' container */ | |
1932 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
1933 | if (fd >= 0) { | |
1934 | if (verbose) | |
1935 | fprintf(stderr, | |
1936 | Name ": Cannot create this array on device %s\n", | |
1937 | dev); | |
1938 | close(fd); | |
1939 | return 0; | |
1940 | } | |
1941 | if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) { | |
1942 | if (verbose) | |
1943 | fprintf(stderr, Name ": Cannot open %s: %s\n", | |
1944 | dev, strerror(errno)); | |
1945 | return 0; | |
1946 | } | |
1947 | /* Well, it is in use by someone, maybe an 'imsm' container. */ | |
1948 | cfd = open_container(fd); | |
1949 | if (cfd < 0) { | |
1950 | close(fd); | |
1951 | if (verbose) | |
1952 | fprintf(stderr, Name ": Cannot use %s: It is busy\n", | |
1953 | dev); | |
1954 | return 0; | |
1955 | } | |
1956 | sra = sysfs_read(cfd, 0, GET_VERSION); | |
1957 | close(fd); | |
1958 | if (sra && sra->array.major_version == -1 && | |
1959 | strcmp(sra->text_version, "imsm") == 0) { | |
1960 | /* This is a member of a imsm container. Load the container | |
1961 | * and try to create a volume | |
1962 | */ | |
1963 | struct intel_super *super; | |
1964 | ||
1965 | if (load_super_imsm_all(st, cfd, (void **) &super, NULL, 1) == 0) { | |
1966 | st->sb = super; | |
1967 | st->container_dev = fd2devnum(cfd); | |
1968 | close(cfd); | |
1969 | return validate_geometry_imsm_volume(st, level, layout, | |
1970 | raiddisks, chunk, | |
1971 | size, dev, | |
1972 | freesize, verbose); | |
1973 | } | |
1974 | close(cfd); | |
1975 | } else /* may belong to another container */ | |
1976 | return 0; | |
1977 | ||
1978 | return 1; | |
1979 | } | |
1980 | ||
cdddbdbc DW |
1981 | static struct mdinfo *container_content_imsm(struct supertype *st) |
1982 | { | |
4f5bc454 DW |
1983 | /* Given a container loaded by load_super_imsm_all, |
1984 | * extract information about all the arrays into | |
1985 | * an mdinfo tree. | |
1986 | * | |
1987 | * For each imsm_dev create an mdinfo, fill it in, | |
1988 | * then look for matching devices in super->disks | |
1989 | * and create appropriate device mdinfo. | |
1990 | */ | |
1991 | struct intel_super *super = st->sb; | |
949c47a0 | 1992 | struct imsm_super *mpb = super->anchor; |
4f5bc454 DW |
1993 | struct mdinfo *rest = NULL; |
1994 | int i; | |
cdddbdbc | 1995 | |
604b746f JD |
1996 | /* do not assemble arrays that might have bad blocks */ |
1997 | if (imsm_bbm_log_size(super->anchor)) { | |
1998 | fprintf(stderr, Name ": BBM log found in metadata. " | |
1999 | "Cannot activate array(s).\n"); | |
2000 | return NULL; | |
2001 | } | |
2002 | ||
4f5bc454 | 2003 | for (i = 0; i < mpb->num_raid_devs; i++) { |
949c47a0 | 2004 | struct imsm_dev *dev = get_imsm_dev(super, i); |
4f5bc454 | 2005 | struct imsm_vol *vol = &dev->vol; |
a965f303 | 2006 | struct imsm_map *map = get_imsm_map(dev, 0); |
4f5bc454 | 2007 | struct mdinfo *this; |
4f5bc454 DW |
2008 | int slot; |
2009 | ||
2010 | this = malloc(sizeof(*this)); | |
2011 | memset(this, 0, sizeof(*this)); | |
2012 | this->next = rest; | |
4f5bc454 | 2013 | |
4f5bc454 DW |
2014 | this->array.level = get_imsm_raid_level(map); |
2015 | this->array.raid_disks = map->num_members; | |
c2c087e6 | 2016 | this->array.layout = imsm_level_to_layout(this->array.level); |
4f5bc454 DW |
2017 | this->array.md_minor = -1; |
2018 | this->array.ctime = 0; | |
2019 | this->array.utime = 0; | |
2020 | this->array.chunk_size = __le16_to_cpu(map->blocks_per_strip) << 9; | |
2021 | this->array.state = !vol->dirty; | |
2022 | this->container_member = i; | |
9a1608e5 DW |
2023 | if (map->map_state == IMSM_T_STATE_UNINITIALIZED || |
2024 | dev->vol.dirty || dev->vol.migr_state) | |
0fd5c350 DW |
2025 | this->resync_start = 0; |
2026 | else | |
2027 | this->resync_start = ~0ULL; | |
2028 | ||
4f5bc454 DW |
2029 | strncpy(this->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN); |
2030 | this->name[MAX_RAID_SERIAL_LEN] = 0; | |
2031 | ||
159c3a1a NB |
2032 | sprintf(this->text_version, "/%s/%d", |
2033 | devnum2devname(st->container_dev), | |
2034 | this->container_member); | |
2035 | ||
4f5bc454 DW |
2036 | memset(this->uuid, 0, sizeof(this->uuid)); |
2037 | ||
f54e6321 | 2038 | this->component_size = __le32_to_cpu(map->blocks_per_member); |
4f5bc454 DW |
2039 | |
2040 | for (slot = 0 ; slot < map->num_members; slot++) { | |
4f5bc454 DW |
2041 | struct mdinfo *info_d; |
2042 | struct dl *d; | |
2043 | int idx; | |
9a1608e5 | 2044 | int skip; |
4f5bc454 | 2045 | __u32 s; |
7eef0453 | 2046 | __u32 ord; |
4f5bc454 | 2047 | |
9a1608e5 | 2048 | skip = 0; |
d23fe947 | 2049 | idx = get_imsm_disk_idx(map, slot); |
7eef0453 | 2050 | ord = get_imsm_ord_tbl_ent(dev, slot); |
4f5bc454 DW |
2051 | for (d = super->disks; d ; d = d->next) |
2052 | if (d->index == idx) | |
2053 | break; | |
2054 | ||
2055 | if (d == NULL) | |
9a1608e5 DW |
2056 | skip = 1; |
2057 | ||
2058 | s = d ? __le32_to_cpu(d->disk.status) : 0; | |
2059 | if (s & FAILED_DISK) | |
2060 | skip = 1; | |
2061 | if (!(s & USABLE_DISK)) | |
2062 | skip = 1; | |
7eef0453 DW |
2063 | if (ord & IMSM_ORD_REBUILD) |
2064 | skip = 1; | |
9a1608e5 DW |
2065 | |
2066 | /* | |
2067 | * if we skip some disks the array will be assmebled degraded; | |
2068 | * reset resync start to avoid a dirty-degraded situation | |
2069 | * | |
2070 | * FIXME handle dirty degraded | |
2071 | */ | |
2072 | if (skip && !dev->vol.dirty) | |
2073 | this->resync_start = ~0ULL; | |
2074 | if (skip) | |
2075 | continue; | |
4f5bc454 DW |
2076 | |
2077 | info_d = malloc(sizeof(*info_d)); | |
9a1608e5 DW |
2078 | if (!info_d) { |
2079 | fprintf(stderr, Name ": failed to allocate disk" | |
2080 | " for volume %s\n", (char *) dev->volume); | |
2081 | free(this); | |
2082 | this = rest; | |
2083 | break; | |
2084 | } | |
4f5bc454 DW |
2085 | memset(info_d, 0, sizeof(*info_d)); |
2086 | info_d->next = this->devs; | |
2087 | this->devs = info_d; | |
2088 | ||
4f5bc454 DW |
2089 | info_d->disk.number = d->index; |
2090 | info_d->disk.major = d->major; | |
2091 | info_d->disk.minor = d->minor; | |
2092 | info_d->disk.raid_disk = slot; | |
4f5bc454 DW |
2093 | |
2094 | this->array.working_disks++; | |
2095 | ||
2096 | info_d->events = __le32_to_cpu(mpb->generation_num); | |
2097 | info_d->data_offset = __le32_to_cpu(map->pba_of_lba0); | |
2098 | info_d->component_size = __le32_to_cpu(map->blocks_per_member); | |
2099 | if (d->devname) | |
2100 | strcpy(info_d->name, d->devname); | |
2101 | } | |
9a1608e5 | 2102 | rest = this; |
4f5bc454 DW |
2103 | } |
2104 | ||
2105 | return rest; | |
cdddbdbc DW |
2106 | } |
2107 | ||
845dea95 | 2108 | |
cba0191b NB |
2109 | static int imsm_open_new(struct supertype *c, struct active_array *a, |
2110 | char *inst) | |
845dea95 | 2111 | { |
0372d5a2 | 2112 | struct intel_super *super = c->sb; |
949c47a0 | 2113 | struct imsm_super *mpb = super->anchor; |
0372d5a2 | 2114 | |
949c47a0 | 2115 | if (atoi(inst) >= mpb->num_raid_devs) { |
0372d5a2 DW |
2116 | fprintf(stderr, "%s: subarry index %d, out of range\n", |
2117 | __func__, atoi(inst)); | |
2118 | return -ENODEV; | |
2119 | } | |
2120 | ||
4e6e574a | 2121 | dprintf("imsm: open_new %s\n", inst); |
cba0191b | 2122 | a->info.container_member = atoi(inst); |
845dea95 NB |
2123 | return 0; |
2124 | } | |
2125 | ||
949c47a0 | 2126 | static __u8 imsm_check_degraded(struct intel_super *super, int n, int failed) |
c2a1e7da | 2127 | { |
949c47a0 | 2128 | struct imsm_dev *dev = get_imsm_dev(super, n); |
a965f303 | 2129 | struct imsm_map *map = get_imsm_map(dev, 0); |
c2a1e7da DW |
2130 | |
2131 | if (!failed) | |
3393c6af DW |
2132 | return map->map_state == IMSM_T_STATE_UNINITIALIZED ? |
2133 | IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL; | |
c2a1e7da DW |
2134 | |
2135 | switch (get_imsm_raid_level(map)) { | |
2136 | case 0: | |
2137 | return IMSM_T_STATE_FAILED; | |
2138 | break; | |
2139 | case 1: | |
2140 | if (failed < map->num_members) | |
2141 | return IMSM_T_STATE_DEGRADED; | |
2142 | else | |
2143 | return IMSM_T_STATE_FAILED; | |
2144 | break; | |
2145 | case 10: | |
2146 | { | |
2147 | /** | |
2148 | * check to see if any mirrors have failed, | |
2149 | * otherwise we are degraded | |
2150 | */ | |
2151 | int device_per_mirror = 2; /* FIXME is this always the case? | |
2152 | * and are they always adjacent? | |
2153 | */ | |
2154 | int failed = 0; | |
2155 | int i; | |
2156 | ||
2157 | for (i = 0; i < map->num_members; i++) { | |
2158 | int idx = get_imsm_disk_idx(map, i); | |
949c47a0 | 2159 | struct imsm_disk *disk = get_imsm_disk(super, idx); |
c2a1e7da DW |
2160 | |
2161 | if (__le32_to_cpu(disk->status) & FAILED_DISK) | |
2162 | failed++; | |
2163 | ||
2164 | if (failed >= device_per_mirror) | |
2165 | return IMSM_T_STATE_FAILED; | |
2166 | ||
2167 | /* reset 'failed' for next mirror set */ | |
2168 | if (!((i + 1) % device_per_mirror)) | |
2169 | failed = 0; | |
2170 | } | |
2171 | ||
2172 | return IMSM_T_STATE_DEGRADED; | |
2173 | } | |
2174 | case 5: | |
2175 | if (failed < 2) | |
2176 | return IMSM_T_STATE_DEGRADED; | |
2177 | else | |
2178 | return IMSM_T_STATE_FAILED; | |
2179 | break; | |
2180 | default: | |
2181 | break; | |
2182 | } | |
2183 | ||
2184 | return map->map_state; | |
2185 | } | |
2186 | ||
949c47a0 | 2187 | static int imsm_count_failed(struct intel_super *super, struct imsm_map *map) |
c2a1e7da DW |
2188 | { |
2189 | int i; | |
2190 | int failed = 0; | |
2191 | struct imsm_disk *disk; | |
2192 | ||
2193 | for (i = 0; i < map->num_members; i++) { | |
2194 | int idx = get_imsm_disk_idx(map, i); | |
2195 | ||
949c47a0 | 2196 | disk = get_imsm_disk(super, idx); |
c2a1e7da DW |
2197 | if (__le32_to_cpu(disk->status) & FAILED_DISK) |
2198 | failed++; | |
fcb84475 DW |
2199 | else if (!(__le32_to_cpu(disk->status) & USABLE_DISK)) |
2200 | failed++; | |
c2a1e7da DW |
2201 | } |
2202 | ||
2203 | return failed; | |
845dea95 NB |
2204 | } |
2205 | ||
a862209d DW |
2206 | static void imsm_set_array_state(struct active_array *a, int consistent) |
2207 | { | |
2208 | int inst = a->info.container_member; | |
2209 | struct intel_super *super = a->container->sb; | |
949c47a0 | 2210 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
a965f303 | 2211 | struct imsm_map *map = get_imsm_map(dev, 0); |
a862209d DW |
2212 | int dirty = !consistent; |
2213 | int failed; | |
2214 | __u8 map_state; | |
2215 | ||
3393c6af DW |
2216 | failed = imsm_count_failed(super, map); |
2217 | map_state = imsm_check_degraded(super, inst, failed); | |
2218 | ||
272906ef DW |
2219 | if (consistent && !dev->vol.dirty && |
2220 | (dev->vol.migr_state || map_state != IMSM_T_STATE_NORMAL)) | |
2221 | a->resync_start = 0ULL; | |
2222 | ||
a862209d | 2223 | if (a->resync_start == ~0ULL) { |
115c3803 | 2224 | /* complete recovery or initial resync */ |
a862209d | 2225 | if (map->map_state != map_state) { |
4e6e574a | 2226 | dprintf("imsm: map_state %d: %d\n", |
a862209d DW |
2227 | inst, map_state); |
2228 | map->map_state = map_state; | |
2229 | super->updates_pending++; | |
2230 | } | |
3393c6af DW |
2231 | if (dev->vol.migr_state) { |
2232 | dprintf("imsm: mark resync complete\n"); | |
2233 | dev->vol.migr_state = 0; | |
2234 | dev->vol.migr_type = 0; | |
115c3803 | 2235 | super->updates_pending++; |
115c3803 | 2236 | } |
3393c6af DW |
2237 | } else if (!dev->vol.migr_state) { |
2238 | dprintf("imsm: mark '%s' (%llu)\n", | |
2239 | failed ? "rebuild" : "initializing", a->resync_start); | |
2240 | /* mark that we are rebuilding */ | |
2241 | map->map_state = failed ? map_state : IMSM_T_STATE_NORMAL; | |
2242 | dev->vol.migr_state = 1; | |
2243 | dev->vol.migr_type = failed ? 1 : 0; | |
2244 | dup_map(dev); | |
272906ef | 2245 | a->check_degraded = 1; |
3393c6af | 2246 | super->updates_pending++; |
115c3803 | 2247 | } |
a862209d | 2248 | |
3393c6af DW |
2249 | /* mark dirty / clean */ |
2250 | if (dirty != dev->vol.dirty) { | |
2251 | dprintf("imsm: mark '%s' (%llu)\n", | |
2252 | dirty ? "dirty" : "clean", a->resync_start); | |
2253 | dev->vol.dirty = dirty; | |
a862209d DW |
2254 | super->updates_pending++; |
2255 | } | |
2256 | } | |
2257 | ||
8d45d196 | 2258 | static void imsm_set_disk(struct active_array *a, int n, int state) |
845dea95 | 2259 | { |
8d45d196 DW |
2260 | int inst = a->info.container_member; |
2261 | struct intel_super *super = a->container->sb; | |
949c47a0 | 2262 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
a965f303 | 2263 | struct imsm_map *map = get_imsm_map(dev, 0); |
8d45d196 DW |
2264 | struct imsm_disk *disk; |
2265 | __u32 status; | |
2266 | int failed = 0; | |
2267 | int new_failure = 0; | |
2268 | ||
2269 | if (n > map->num_members) | |
2270 | fprintf(stderr, "imsm: set_disk %d out of range 0..%d\n", | |
2271 | n, map->num_members - 1); | |
2272 | ||
2273 | if (n < 0) | |
2274 | return; | |
2275 | ||
4e6e574a | 2276 | dprintf("imsm: set_disk %d:%x\n", n, state); |
8d45d196 | 2277 | |
949c47a0 | 2278 | disk = get_imsm_disk(super, get_imsm_disk_idx(map, n)); |
8d45d196 | 2279 | |
5802a811 | 2280 | /* check for new failures */ |
8d45d196 DW |
2281 | status = __le32_to_cpu(disk->status); |
2282 | if ((state & DS_FAULTY) && !(status & FAILED_DISK)) { | |
2283 | status |= FAILED_DISK; | |
2284 | disk->status = __cpu_to_le32(status); | |
2285 | new_failure = 1; | |
5802a811 | 2286 | super->updates_pending++; |
8d45d196 | 2287 | } |
19859edc DW |
2288 | /* check if in_sync */ |
2289 | if ((state & DS_INSYNC) && !(status & USABLE_DISK)) { | |
2290 | status |= USABLE_DISK; | |
2291 | disk->status = __cpu_to_le32(status); | |
2292 | super->updates_pending++; | |
2293 | } | |
8d45d196 | 2294 | |
5802a811 | 2295 | /* the number of failures have changed, count up 'failed' to determine |
8d45d196 DW |
2296 | * degraded / failed status |
2297 | */ | |
2298 | if (new_failure && map->map_state != IMSM_T_STATE_FAILED) | |
949c47a0 | 2299 | failed = imsm_count_failed(super, map); |
8d45d196 | 2300 | |
5802a811 | 2301 | /* determine map_state based on failed or in_sync count */ |
8d45d196 | 2302 | if (failed) |
949c47a0 | 2303 | map->map_state = imsm_check_degraded(super, inst, failed); |
5802a811 DW |
2304 | else if (map->map_state == IMSM_T_STATE_DEGRADED) { |
2305 | struct mdinfo *d; | |
2306 | int working = 0; | |
8d45d196 | 2307 | |
5802a811 DW |
2308 | for (d = a->info.devs ; d ; d = d->next) |
2309 | if (d->curr_state & DS_INSYNC) | |
2310 | working++; | |
2311 | ||
2312 | if (working == a->info.array.raid_disks) { | |
2313 | map->map_state = IMSM_T_STATE_NORMAL; | |
3393c6af DW |
2314 | dev->vol.migr_state = 0; |
2315 | dev->vol.migr_type = 0; | |
5802a811 DW |
2316 | super->updates_pending++; |
2317 | } | |
2318 | } | |
845dea95 NB |
2319 | } |
2320 | ||
c2a1e7da DW |
2321 | static int store_imsm_mpb(int fd, struct intel_super *super) |
2322 | { | |
949c47a0 | 2323 | struct imsm_super *mpb = super->anchor; |
c2a1e7da DW |
2324 | __u32 mpb_size = __le32_to_cpu(mpb->mpb_size); |
2325 | unsigned long long dsize; | |
2326 | unsigned long long sectors; | |
2327 | ||
2328 | get_dev_size(fd, NULL, &dsize); | |
2329 | ||
272f648f DW |
2330 | if (mpb_size > 512) { |
2331 | /* -1 to account for anchor */ | |
2332 | sectors = mpb_sectors(mpb) - 1; | |
c2a1e7da | 2333 | |
272f648f DW |
2334 | /* write the extended mpb to the sectors preceeding the anchor */ |
2335 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) | |
2336 | return 1; | |
c2a1e7da | 2337 | |
99e29264 | 2338 | if (write(fd, super->buf + 512, 512 * sectors) != 512 * sectors) |
272f648f DW |
2339 | return 1; |
2340 | } | |
c2a1e7da | 2341 | |
272f648f DW |
2342 | /* first block is stored on second to last sector of the disk */ |
2343 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) | |
c2a1e7da DW |
2344 | return 1; |
2345 | ||
272f648f | 2346 | if (write(fd, super->buf, 512) != 512) |
c2a1e7da DW |
2347 | return 1; |
2348 | ||
c2a1e7da DW |
2349 | return 0; |
2350 | } | |
2351 | ||
2e735d19 | 2352 | static void imsm_sync_metadata(struct supertype *container) |
845dea95 | 2353 | { |
2e735d19 | 2354 | struct intel_super *super = container->sb; |
c2a1e7da DW |
2355 | |
2356 | if (!super->updates_pending) | |
2357 | return; | |
2358 | ||
c2c087e6 | 2359 | write_super_imsm(super, 0); |
c2a1e7da DW |
2360 | |
2361 | super->updates_pending = 0; | |
845dea95 NB |
2362 | } |
2363 | ||
272906ef DW |
2364 | static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a) |
2365 | { | |
2366 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
2367 | struct imsm_map *map = get_imsm_map(dev, 0); | |
2368 | int i = get_imsm_disk_idx(map, idx); | |
2369 | struct dl *dl; | |
2370 | ||
2371 | for (dl = super->disks; dl; dl = dl->next) | |
2372 | if (dl->index == i) | |
2373 | break; | |
2374 | ||
2375 | if (__le32_to_cpu(dl->disk.status) & FAILED_DISK) | |
2376 | dl = NULL; | |
2377 | ||
2378 | if (dl) | |
2379 | dprintf("%s: found %x:%x\n", __func__, dl->major, dl->minor); | |
2380 | ||
2381 | return dl; | |
2382 | } | |
2383 | ||
2384 | static struct dl *imsm_add_spare(struct intel_super *super, int idx, struct active_array *a) | |
2385 | { | |
2386 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
2387 | struct imsm_map *map = get_imsm_map(dev, 0); | |
2388 | unsigned long long esize; | |
2389 | unsigned long long pos; | |
2390 | struct mdinfo *d; | |
2391 | struct extent *ex; | |
2392 | int j; | |
2393 | int found; | |
2394 | __u32 array_start; | |
9a1608e5 | 2395 | __u32 status; |
272906ef DW |
2396 | struct dl *dl; |
2397 | ||
2398 | for (dl = super->disks; dl; dl = dl->next) { | |
2399 | /* If in this array, skip */ | |
2400 | for (d = a->info.devs ; d ; d = d->next) | |
2401 | if (d->disk.major == dl->major && | |
2402 | d->disk.minor == dl->minor) { | |
2403 | dprintf("%x:%x already in array\n", dl->major, dl->minor); | |
2404 | break; | |
2405 | } | |
2406 | if (d) | |
2407 | continue; | |
2408 | ||
9a1608e5 DW |
2409 | /* skip marked in use or failed drives */ |
2410 | status = __le32_to_cpu(dl->disk.status); | |
2411 | if (status & FAILED_DISK || status & CONFIGURED_DISK) { | |
2412 | dprintf("%x:%x status ( %s%s)\n", | |
2413 | dl->major, dl->minor, | |
2414 | status & FAILED_DISK ? "failed " : "", | |
2415 | status & CONFIGURED_DISK ? "configured " : ""); | |
2416 | continue; | |
2417 | } | |
2418 | ||
272906ef DW |
2419 | /* Does this unused device have the requisite free space? |
2420 | * We need a->info.component_size sectors | |
2421 | */ | |
2422 | ex = get_extents(super, dl); | |
2423 | if (!ex) { | |
2424 | dprintf("cannot get extents\n"); | |
2425 | continue; | |
2426 | } | |
2427 | found = 0; | |
2428 | j = 0; | |
2429 | pos = 0; | |
2430 | array_start = __le32_to_cpu(map->pba_of_lba0); | |
2431 | ||
2432 | do { | |
2433 | /* check that we can start at pba_of_lba0 with | |
2434 | * a->info.component_size of space | |
2435 | */ | |
2436 | esize = ex[j].start - pos; | |
2437 | if (array_start >= pos && | |
2438 | array_start + a->info.component_size < ex[j].start) { | |
2439 | found = 1; | |
2440 | break; | |
2441 | } | |
2442 | pos = ex[j].start + ex[j].size; | |
2443 | j++; | |
2444 | ||
2445 | } while (ex[j-1].size); | |
2446 | ||
2447 | free(ex); | |
2448 | if (!found) { | |
2449 | dprintf("%x:%x does not have %llu at %d\n", | |
2450 | dl->major, dl->minor, | |
2451 | a->info.component_size, | |
2452 | __le32_to_cpu(map->pba_of_lba0)); | |
2453 | /* No room */ | |
2454 | continue; | |
2455 | } else | |
2456 | break; | |
2457 | } | |
2458 | ||
2459 | return dl; | |
2460 | } | |
2461 | ||
88758e9d DW |
2462 | static struct mdinfo *imsm_activate_spare(struct active_array *a, |
2463 | struct metadata_update **updates) | |
2464 | { | |
2465 | /** | |
d23fe947 DW |
2466 | * Find a device with unused free space and use it to replace a |
2467 | * failed/vacant region in an array. We replace failed regions one a | |
2468 | * array at a time. The result is that a new spare disk will be added | |
2469 | * to the first failed array and after the monitor has finished | |
2470 | * propagating failures the remainder will be consumed. | |
88758e9d | 2471 | * |
d23fe947 DW |
2472 | * FIXME add a capability for mdmon to request spares from another |
2473 | * container. | |
88758e9d DW |
2474 | */ |
2475 | ||
2476 | struct intel_super *super = a->container->sb; | |
88758e9d | 2477 | int inst = a->info.container_member; |
949c47a0 | 2478 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
a965f303 | 2479 | struct imsm_map *map = get_imsm_map(dev, 0); |
88758e9d DW |
2480 | int failed = a->info.array.raid_disks; |
2481 | struct mdinfo *rv = NULL; | |
2482 | struct mdinfo *d; | |
2483 | struct mdinfo *di; | |
2484 | struct metadata_update *mu; | |
2485 | struct dl *dl; | |
2486 | struct imsm_update_activate_spare *u; | |
2487 | int num_spares = 0; | |
2488 | int i; | |
2489 | ||
2490 | for (d = a->info.devs ; d ; d = d->next) { | |
2491 | if ((d->curr_state & DS_FAULTY) && | |
2492 | d->state_fd >= 0) | |
2493 | /* wait for Removal to happen */ | |
2494 | return NULL; | |
2495 | if (d->state_fd >= 0) | |
2496 | failed--; | |
2497 | } | |
2498 | ||
2499 | dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n", | |
2500 | inst, failed, a->info.array.raid_disks, a->info.array.level); | |
949c47a0 | 2501 | if (imsm_check_degraded(super, inst, failed) != IMSM_T_STATE_DEGRADED) |
88758e9d DW |
2502 | return NULL; |
2503 | ||
2504 | /* For each slot, if it is not working, find a spare */ | |
88758e9d DW |
2505 | for (i = 0; i < a->info.array.raid_disks; i++) { |
2506 | for (d = a->info.devs ; d ; d = d->next) | |
2507 | if (d->disk.raid_disk == i) | |
2508 | break; | |
2509 | dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0); | |
2510 | if (d && (d->state_fd >= 0)) | |
2511 | continue; | |
2512 | ||
272906ef DW |
2513 | /* |
2514 | * OK, this device needs recovery. Try to re-add the previous | |
2515 | * occupant of this slot, if this fails add a new spare | |
2516 | */ | |
2517 | dl = imsm_readd(super, i, a); | |
2518 | if (!dl) | |
2519 | dl = imsm_add_spare(super, i, a); | |
2520 | if (!dl) | |
2521 | continue; | |
2522 | ||
2523 | /* found a usable disk with enough space */ | |
2524 | di = malloc(sizeof(*di)); | |
2525 | memset(di, 0, sizeof(*di)); | |
2526 | ||
2527 | /* dl->index will be -1 in the case we are activating a | |
2528 | * pristine spare. imsm_process_update() will create a | |
2529 | * new index in this case. Once a disk is found to be | |
2530 | * failed in all member arrays it is kicked from the | |
2531 | * metadata | |
2532 | */ | |
2533 | di->disk.number = dl->index; | |
d23fe947 | 2534 | |
272906ef DW |
2535 | /* (ab)use di->devs to store a pointer to the device |
2536 | * we chose | |
2537 | */ | |
2538 | di->devs = (struct mdinfo *) dl; | |
2539 | ||
2540 | di->disk.raid_disk = i; | |
2541 | di->disk.major = dl->major; | |
2542 | di->disk.minor = dl->minor; | |
2543 | di->disk.state = 0; | |
2544 | di->data_offset = __le32_to_cpu(map->pba_of_lba0); | |
2545 | di->component_size = a->info.component_size; | |
2546 | di->container_member = inst; | |
2547 | di->next = rv; | |
2548 | rv = di; | |
2549 | num_spares++; | |
2550 | dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor, | |
2551 | i, di->data_offset); | |
88758e9d | 2552 | |
272906ef | 2553 | break; |
88758e9d DW |
2554 | } |
2555 | ||
2556 | if (!rv) | |
2557 | /* No spares found */ | |
2558 | return rv; | |
2559 | /* Now 'rv' has a list of devices to return. | |
2560 | * Create a metadata_update record to update the | |
2561 | * disk_ord_tbl for the array | |
2562 | */ | |
2563 | mu = malloc(sizeof(*mu)); | |
2564 | mu->buf = malloc(sizeof(struct imsm_update_activate_spare) * num_spares); | |
2565 | mu->space = NULL; | |
2566 | mu->len = sizeof(struct imsm_update_activate_spare) * num_spares; | |
2567 | mu->next = *updates; | |
2568 | u = (struct imsm_update_activate_spare *) mu->buf; | |
2569 | ||
2570 | for (di = rv ; di ; di = di->next) { | |
2571 | u->type = update_activate_spare; | |
d23fe947 DW |
2572 | u->dl = (struct dl *) di->devs; |
2573 | di->devs = NULL; | |
88758e9d DW |
2574 | u->slot = di->disk.raid_disk; |
2575 | u->array = inst; | |
2576 | u->next = u + 1; | |
2577 | u++; | |
2578 | } | |
2579 | (u-1)->next = NULL; | |
2580 | *updates = mu; | |
2581 | ||
2582 | return rv; | |
2583 | } | |
2584 | ||
8273f55e DW |
2585 | static int disks_overlap(struct imsm_map *m1, struct imsm_map *m2) |
2586 | { | |
2587 | int i; | |
2588 | int j; | |
2589 | int idx; | |
2590 | ||
2591 | for (i = 0; i < m1->num_members; i++) { | |
2592 | idx = get_imsm_disk_idx(m1, i); | |
2593 | for (j = 0; j < m2->num_members; j++) | |
2594 | if (idx == get_imsm_disk_idx(m2, j)) | |
2595 | return 1; | |
2596 | } | |
2597 | ||
2598 | return 0; | |
2599 | } | |
2600 | ||
e8319a19 DW |
2601 | static void imsm_process_update(struct supertype *st, |
2602 | struct metadata_update *update) | |
2603 | { | |
2604 | /** | |
2605 | * crack open the metadata_update envelope to find the update record | |
2606 | * update can be one of: | |
2607 | * update_activate_spare - a spare device has replaced a failed | |
2608 | * device in an array, update the disk_ord_tbl. If this disk is | |
2609 | * present in all member arrays then also clear the SPARE_DISK | |
2610 | * flag | |
2611 | */ | |
2612 | struct intel_super *super = st->sb; | |
949c47a0 | 2613 | struct imsm_super *mpb = super->anchor; |
e8319a19 DW |
2614 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
2615 | ||
2616 | switch (type) { | |
2617 | case update_activate_spare: { | |
2618 | struct imsm_update_activate_spare *u = (void *) update->buf; | |
949c47a0 | 2619 | struct imsm_dev *dev = get_imsm_dev(super, u->array); |
a965f303 | 2620 | struct imsm_map *map = get_imsm_map(dev, 0); |
e8319a19 DW |
2621 | struct active_array *a; |
2622 | struct imsm_disk *disk; | |
2623 | __u32 status; | |
2624 | struct dl *dl; | |
e8319a19 DW |
2625 | unsigned int found; |
2626 | int victim; | |
2627 | int i; | |
2628 | ||
2629 | for (dl = super->disks; dl; dl = dl->next) | |
d23fe947 | 2630 | if (dl == u->dl) |
e8319a19 DW |
2631 | break; |
2632 | ||
2633 | if (!dl) { | |
2634 | fprintf(stderr, "error: imsm_activate_spare passed " | |
d23fe947 DW |
2635 | "an unknown disk (index: %d serial: %s)\n", |
2636 | u->dl->index, u->dl->serial); | |
e8319a19 DW |
2637 | return; |
2638 | } | |
2639 | ||
2640 | super->updates_pending++; | |
2641 | ||
d23fe947 DW |
2642 | /* adding a pristine spare, assign a new index */ |
2643 | if (dl->index < 0) { | |
2644 | dl->index = super->anchor->num_disks; | |
2645 | super->anchor->num_disks++; | |
2646 | } | |
e8319a19 | 2647 | victim = get_imsm_disk_idx(map, u->slot); |
d23fe947 DW |
2648 | map->disk_ord_tbl[u->slot] = __cpu_to_le32(dl->index); |
2649 | disk = &dl->disk; | |
e8319a19 DW |
2650 | status = __le32_to_cpu(disk->status); |
2651 | status |= CONFIGURED_DISK; | |
19859edc | 2652 | status &= ~(SPARE_DISK | USABLE_DISK); |
e8319a19 DW |
2653 | disk->status = __cpu_to_le32(status); |
2654 | ||
e8319a19 DW |
2655 | /* count arrays using the victim in the metadata */ |
2656 | found = 0; | |
2657 | for (a = st->arrays; a ; a = a->next) { | |
949c47a0 | 2658 | dev = get_imsm_dev(super, a->info.container_member); |
a965f303 | 2659 | map = get_imsm_map(dev, 0); |
e8319a19 DW |
2660 | for (i = 0; i < map->num_members; i++) |
2661 | if (victim == get_imsm_disk_idx(map, i)) | |
2662 | found++; | |
2663 | } | |
2664 | ||
2665 | /* clear some flags if the victim is no longer being | |
2666 | * utilized anywhere | |
2667 | */ | |
e8319a19 | 2668 | if (!found) { |
d23fe947 | 2669 | disk = get_imsm_disk(super, victim); |
e8319a19 DW |
2670 | status = __le32_to_cpu(disk->status); |
2671 | status &= ~(CONFIGURED_DISK | USABLE_DISK); | |
2672 | disk->status = __cpu_to_le32(status); | |
d23fe947 DW |
2673 | /* at this point the disk can be removed from the |
2674 | * metadata, however we need to guarantee that we do | |
2675 | * not race with any manager thread routine that relies | |
2676 | * on dl->index or map->disk_ord_tbl | |
2677 | */ | |
e8319a19 | 2678 | } |
8273f55e DW |
2679 | break; |
2680 | } | |
2681 | case update_create_array: { | |
2682 | /* someone wants to create a new array, we need to be aware of | |
2683 | * a few races/collisions: | |
2684 | * 1/ 'Create' called by two separate instances of mdadm | |
2685 | * 2/ 'Create' versus 'activate_spare': mdadm has chosen | |
2686 | * devices that have since been assimilated via | |
2687 | * activate_spare. | |
2688 | * In the event this update can not be carried out mdadm will | |
2689 | * (FIX ME) notice that its update did not take hold. | |
2690 | */ | |
2691 | struct imsm_update_create_array *u = (void *) update->buf; | |
2692 | struct imsm_dev *dev; | |
2693 | struct imsm_map *map, *new_map; | |
2694 | unsigned long long start, end; | |
2695 | unsigned long long new_start, new_end; | |
2696 | int i; | |
2697 | int overlap = 0; | |
2698 | ||
2699 | /* handle racing creates: first come first serve */ | |
2700 | if (u->dev_idx < mpb->num_raid_devs) { | |
2701 | dprintf("%s: subarray %d already defined\n", | |
2702 | __func__, u->dev_idx); | |
2703 | return; | |
2704 | } | |
2705 | ||
2706 | /* check update is next in sequence */ | |
2707 | if (u->dev_idx != mpb->num_raid_devs) { | |
6a3e913e DW |
2708 | dprintf("%s: can not create array %d expected index %d\n", |
2709 | __func__, u->dev_idx, mpb->num_raid_devs); | |
8273f55e DW |
2710 | return; |
2711 | } | |
2712 | ||
a965f303 | 2713 | new_map = get_imsm_map(&u->dev, 0); |
8273f55e DW |
2714 | new_start = __le32_to_cpu(new_map->pba_of_lba0); |
2715 | new_end = new_start + __le32_to_cpu(new_map->blocks_per_member); | |
2716 | ||
2717 | /* handle activate_spare versus create race: | |
2718 | * check to make sure that overlapping arrays do not include | |
2719 | * overalpping disks | |
2720 | */ | |
2721 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
949c47a0 | 2722 | dev = get_imsm_dev(super, i); |
a965f303 | 2723 | map = get_imsm_map(dev, 0); |
8273f55e DW |
2724 | start = __le32_to_cpu(map->pba_of_lba0); |
2725 | end = start + __le32_to_cpu(map->blocks_per_member); | |
2726 | if ((new_start >= start && new_start <= end) || | |
2727 | (start >= new_start && start <= new_end)) | |
2728 | overlap = 1; | |
2729 | if (overlap && disks_overlap(map, new_map)) { | |
2730 | dprintf("%s: arrays overlap\n", __func__); | |
2731 | return; | |
2732 | } | |
2733 | } | |
2734 | /* check num_members sanity */ | |
2735 | if (new_map->num_members > mpb->num_disks) { | |
2736 | dprintf("%s: num_disks out of range\n", __func__); | |
2737 | return; | |
2738 | } | |
2739 | ||
949c47a0 DW |
2740 | /* check that prepare update was successful */ |
2741 | if (!update->space) { | |
2742 | dprintf("%s: prepare update failed\n", __func__); | |
2743 | return; | |
2744 | } | |
2745 | ||
8273f55e | 2746 | super->updates_pending++; |
949c47a0 DW |
2747 | dev = update->space; |
2748 | update->space = NULL; | |
2749 | imsm_copy_dev(dev, &u->dev); | |
2750 | super->dev_tbl[u->dev_idx] = dev; | |
8273f55e | 2751 | mpb->num_raid_devs++; |
8273f55e DW |
2752 | |
2753 | /* fix up flags, if arrays overlap then the drives can not be | |
2754 | * spares | |
2755 | */ | |
2756 | for (i = 0; i < map->num_members; i++) { | |
2757 | struct imsm_disk *disk; | |
2758 | __u32 status; | |
2759 | ||
949c47a0 | 2760 | disk = get_imsm_disk(super, get_imsm_disk_idx(map, i)); |
8273f55e DW |
2761 | status = __le32_to_cpu(disk->status); |
2762 | status |= CONFIGURED_DISK; | |
2763 | if (overlap) | |
2764 | status &= ~SPARE_DISK; | |
2765 | disk->status = __cpu_to_le32(status); | |
2766 | } | |
2767 | break; | |
e8319a19 DW |
2768 | } |
2769 | } | |
2770 | } | |
88758e9d | 2771 | |
8273f55e DW |
2772 | static void imsm_prepare_update(struct supertype *st, |
2773 | struct metadata_update *update) | |
2774 | { | |
949c47a0 DW |
2775 | /** |
2776 | * Allocate space to hold new disk entries, raid-device entries or a | |
2777 | * new mpb if necessary. We currently maintain an mpb large enough to | |
2778 | * hold 2 subarrays for the given number of disks. This may not be | |
2779 | * sufficient when reshaping. | |
8273f55e DW |
2780 | * |
2781 | * FIX ME handle the reshape case. | |
2782 | * | |
2783 | * The monitor will be able to safely change super->mpb by arranging | |
2784 | * for it to be freed in check_update_queue(). I.e. the monitor thread | |
2785 | * will start using the new pointer and the manager can continue to use | |
2786 | * the old value until check_update_queue() runs. | |
2787 | */ | |
949c47a0 DW |
2788 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
2789 | ||
2790 | switch (type) { | |
2791 | case update_create_array: { | |
2792 | struct imsm_update_create_array *u = (void *) update->buf; | |
3393c6af | 2793 | size_t len = sizeof_imsm_dev(&u->dev, 1); |
949c47a0 DW |
2794 | |
2795 | update->space = malloc(len); | |
2796 | break; | |
2797 | default: | |
2798 | break; | |
2799 | } | |
2800 | } | |
8273f55e DW |
2801 | |
2802 | return; | |
2803 | } | |
2804 | ||
cdddbdbc DW |
2805 | struct superswitch super_imsm = { |
2806 | #ifndef MDASSEMBLE | |
2807 | .examine_super = examine_super_imsm, | |
2808 | .brief_examine_super = brief_examine_super_imsm, | |
2809 | .detail_super = detail_super_imsm, | |
2810 | .brief_detail_super = brief_detail_super_imsm, | |
bf5a934a | 2811 | .write_init_super = write_init_super_imsm, |
cdddbdbc DW |
2812 | #endif |
2813 | .match_home = match_home_imsm, | |
2814 | .uuid_from_super= uuid_from_super_imsm, | |
2815 | .getinfo_super = getinfo_super_imsm, | |
2816 | .update_super = update_super_imsm, | |
2817 | ||
2818 | .avail_size = avail_size_imsm, | |
2819 | ||
2820 | .compare_super = compare_super_imsm, | |
2821 | ||
2822 | .load_super = load_super_imsm, | |
bf5a934a DW |
2823 | .init_super = init_super_imsm, |
2824 | .add_to_super = add_to_super_imsm, | |
cdddbdbc DW |
2825 | .store_super = store_zero_imsm, |
2826 | .free_super = free_super_imsm, | |
2827 | .match_metadata_desc = match_metadata_desc_imsm, | |
bf5a934a | 2828 | .container_content = container_content_imsm, |
cdddbdbc DW |
2829 | |
2830 | .validate_geometry = validate_geometry_imsm, | |
cdddbdbc | 2831 | .external = 1, |
845dea95 NB |
2832 | |
2833 | /* for mdmon */ | |
2834 | .open_new = imsm_open_new, | |
2835 | .load_super = load_super_imsm, | |
ed9d66aa | 2836 | .set_array_state= imsm_set_array_state, |
845dea95 NB |
2837 | .set_disk = imsm_set_disk, |
2838 | .sync_metadata = imsm_sync_metadata, | |
88758e9d | 2839 | .activate_spare = imsm_activate_spare, |
e8319a19 | 2840 | .process_update = imsm_process_update, |
8273f55e | 2841 | .prepare_update = imsm_prepare_update, |
cdddbdbc | 2842 | }; |