]>
Commit | Line | Data |
---|---|---|
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" | |
21 | #include "mdmon.h" | |
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 | |
34 | #define MPB_SECTOR_CNT 418 | |
35 | #define IMSM_RESERVED_SECTORS 4096 | |
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 */ | |
71 | __u32 disk_ord_tbl[1]; /* disk_ord_tbl[num_members], | |
72 | top byte special */ | |
73 | } __attribute__ ((packed)); | |
74 | ||
75 | struct imsm_vol { | |
76 | __u32 reserved[2]; | |
77 | __u8 migr_state; /* Normal or Migrating */ | |
78 | __u8 migr_type; /* Initializing, Rebuilding, ... */ | |
79 | __u8 dirty; | |
80 | __u8 fill[1]; | |
81 | __u32 filler[5]; | |
82 | struct imsm_map map[1]; | |
83 | /* here comes another one if migr_state */ | |
84 | } __attribute__ ((packed)); | |
85 | ||
86 | struct imsm_dev { | |
87 | __u8 volume[MAX_RAID_SERIAL_LEN]; | |
88 | __u32 size_low; | |
89 | __u32 size_high; | |
90 | __u32 status; /* Persistent RaidDev status */ | |
91 | __u32 reserved_blocks; /* Reserved blocks at beginning of volume */ | |
92 | #define IMSM_DEV_FILLERS 12 | |
93 | __u32 filler[IMSM_DEV_FILLERS]; | |
94 | struct imsm_vol vol; | |
95 | } __attribute__ ((packed)); | |
96 | ||
97 | struct imsm_super { | |
98 | __u8 sig[MAX_SIGNATURE_LENGTH]; /* 0x00 - 0x1F */ | |
99 | __u32 check_sum; /* 0x20 - 0x23 MPB Checksum */ | |
100 | __u32 mpb_size; /* 0x24 - 0x27 Size of MPB */ | |
101 | __u32 family_num; /* 0x28 - 0x2B Checksum from first time this config was written */ | |
102 | __u32 generation_num; /* 0x2C - 0x2F Incremented each time this array's MPB is written */ | |
103 | __u32 reserved[2]; /* 0x30 - 0x37 */ | |
104 | __u8 num_disks; /* 0x38 Number of configured disks */ | |
105 | __u8 num_raid_devs; /* 0x39 Number of configured volumes */ | |
106 | __u8 fill[2]; /* 0x3A - 0x3B */ | |
107 | #define IMSM_FILLERS 39 | |
108 | __u32 filler[IMSM_FILLERS]; /* 0x3C - 0xD7 RAID_MPB_FILLERS */ | |
109 | struct imsm_disk disk[1]; /* 0xD8 diskTbl[numDisks] */ | |
110 | /* here comes imsm_dev[num_raid_devs] */ | |
111 | } __attribute__ ((packed)); | |
112 | ||
113 | #ifndef MDASSEMBLE | |
114 | static char *map_state_str[] = { "normal", "uninitialized", "degraded", "failed" }; | |
115 | #endif | |
116 | ||
117 | static unsigned int sector_count(__u32 bytes) | |
118 | { | |
119 | return ((bytes + (512-1)) & (~(512-1))) / 512; | |
120 | } | |
121 | ||
122 | static unsigned int mpb_sectors(struct imsm_super *mpb) | |
123 | { | |
124 | return sector_count(__le32_to_cpu(mpb->mpb_size)); | |
125 | } | |
126 | ||
127 | /* internal representation of IMSM metadata */ | |
128 | struct intel_super { | |
129 | union { | |
130 | void *buf; /* O_DIRECT buffer for reading/writing metadata */ | |
131 | struct imsm_super *anchor; /* immovable parameters */ | |
132 | }; | |
133 | size_t len; /* size of the 'buf' allocation */ | |
134 | int updates_pending; /* count of pending updates for mdmon */ | |
135 | int creating_imsm; /* flag to indicate container creation */ | |
136 | int current_vol; /* index of raid device undergoing creation */ | |
137 | #define IMSM_MAX_DISKS 6 | |
138 | struct imsm_disk *disk_tbl[IMSM_MAX_DISKS]; | |
139 | #define IMSM_MAX_RAID_DEVS 2 | |
140 | struct imsm_dev *dev_tbl[IMSM_MAX_RAID_DEVS]; | |
141 | struct dl { | |
142 | struct dl *next; | |
143 | int index; | |
144 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
145 | int major, minor; | |
146 | char *devname; | |
147 | int fd; | |
148 | } *disks; | |
149 | }; | |
150 | ||
151 | struct extent { | |
152 | unsigned long long start, size; | |
153 | }; | |
154 | ||
155 | /* definition of messages passed to imsm_process_update */ | |
156 | enum imsm_update_type { | |
157 | update_activate_spare, | |
158 | update_create_array, | |
159 | }; | |
160 | ||
161 | struct imsm_update_activate_spare { | |
162 | enum imsm_update_type type; | |
163 | int disk_idx; | |
164 | int slot; | |
165 | int array; | |
166 | struct imsm_update_activate_spare *next; | |
167 | }; | |
168 | ||
169 | struct imsm_update_create_array { | |
170 | enum imsm_update_type type; | |
171 | struct imsm_dev dev; | |
172 | int dev_idx; | |
173 | }; | |
174 | ||
175 | static int imsm_env_devname_as_serial(void) | |
176 | { | |
177 | char *val = getenv("IMSM_DEVNAME_AS_SERIAL"); | |
178 | ||
179 | if (val && atoi(val) == 1) | |
180 | return 1; | |
181 | ||
182 | return 0; | |
183 | } | |
184 | ||
185 | ||
186 | static struct supertype *match_metadata_desc_imsm(char *arg) | |
187 | { | |
188 | struct supertype *st; | |
189 | ||
190 | if (strcmp(arg, "imsm") != 0 && | |
191 | strcmp(arg, "default") != 0 | |
192 | ) | |
193 | return NULL; | |
194 | ||
195 | st = malloc(sizeof(*st)); | |
196 | memset(st, 0, sizeof(*st)); | |
197 | st->ss = &super_imsm; | |
198 | st->max_devs = IMSM_MAX_DEVICES; | |
199 | st->minor_version = 0; | |
200 | st->sb = NULL; | |
201 | return st; | |
202 | } | |
203 | ||
204 | static __u8 *get_imsm_version(struct imsm_super *mpb) | |
205 | { | |
206 | return &mpb->sig[MPB_SIG_LEN]; | |
207 | } | |
208 | ||
209 | /* retrieve a disk directly from the anchor when the anchor is known to be | |
210 | * up-to-date, currently only at load time | |
211 | */ | |
212 | static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index) | |
213 | { | |
214 | if (index >= mpb->num_disks) | |
215 | return NULL; | |
216 | return &mpb->disk[index]; | |
217 | } | |
218 | ||
219 | static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index) | |
220 | { | |
221 | if (index >= super->anchor->num_disks) | |
222 | return NULL; | |
223 | return super->disk_tbl[index]; | |
224 | } | |
225 | ||
226 | /* generate a checksum directly from the anchor when the anchor is known to be | |
227 | * up-to-date, currently only at load or write_super after coalescing | |
228 | */ | |
229 | static __u32 __gen_imsm_checksum(struct imsm_super *mpb) | |
230 | { | |
231 | __u32 end = mpb->mpb_size / sizeof(end); | |
232 | __u32 *p = (__u32 *) mpb; | |
233 | __u32 sum = 0; | |
234 | ||
235 | while (end--) | |
236 | sum += __le32_to_cpu(*p++); | |
237 | ||
238 | return sum - __le32_to_cpu(mpb->check_sum); | |
239 | } | |
240 | ||
241 | static size_t sizeof_imsm_dev(struct imsm_dev *dev) | |
242 | { | |
243 | size_t size = sizeof(*dev); | |
244 | ||
245 | /* each map has disk_ord_tbl[num_members - 1] additional space */ | |
246 | size += sizeof(__u32) * (dev->vol.map[0].num_members - 1); | |
247 | ||
248 | /* migrating means an additional map */ | |
249 | if (dev->vol.migr_state) { | |
250 | size += sizeof(struct imsm_map); | |
251 | size += sizeof(__u32) * (dev->vol.map[1].num_members - 1); | |
252 | } | |
253 | ||
254 | return size; | |
255 | } | |
256 | ||
257 | static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index) | |
258 | { | |
259 | int offset; | |
260 | int i; | |
261 | void *_mpb = mpb; | |
262 | ||
263 | if (index >= mpb->num_raid_devs) | |
264 | return NULL; | |
265 | ||
266 | /* devices start after all disks */ | |
267 | offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb; | |
268 | ||
269 | for (i = 0; i <= index; i++) | |
270 | if (i == index) | |
271 | return _mpb + offset; | |
272 | else | |
273 | offset += sizeof_imsm_dev(_mpb + offset); | |
274 | ||
275 | return NULL; | |
276 | } | |
277 | ||
278 | static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index) | |
279 | { | |
280 | if (index >= super->anchor->num_raid_devs) | |
281 | return NULL; | |
282 | return super->dev_tbl[index]; | |
283 | } | |
284 | ||
285 | static __u32 get_imsm_disk_idx(struct imsm_map *map, int slot) | |
286 | { | |
287 | __u32 *ord_tbl = &map->disk_ord_tbl[slot]; | |
288 | ||
289 | /* top byte is 'special' */ | |
290 | return __le32_to_cpu(*ord_tbl & ~(0xff << 24)); | |
291 | } | |
292 | ||
293 | static int get_imsm_raid_level(struct imsm_map *map) | |
294 | { | |
295 | if (map->raid_level == 1) { | |
296 | if (map->num_members == 2) | |
297 | return 1; | |
298 | else | |
299 | return 10; | |
300 | } | |
301 | ||
302 | return map->raid_level; | |
303 | } | |
304 | ||
305 | static int cmp_extent(const void *av, const void *bv) | |
306 | { | |
307 | const struct extent *a = av; | |
308 | const struct extent *b = bv; | |
309 | if (a->start < b->start) | |
310 | return -1; | |
311 | if (a->start > b->start) | |
312 | return 1; | |
313 | return 0; | |
314 | } | |
315 | ||
316 | static struct extent *get_extents(struct intel_super *super, struct dl *dl) | |
317 | { | |
318 | /* find a list of used extents on the given physical device */ | |
319 | struct imsm_disk *disk; | |
320 | struct extent *rv, *e; | |
321 | int i, j; | |
322 | int memberships = 0; | |
323 | ||
324 | disk = get_imsm_disk(super, dl->index); | |
325 | if (!disk) | |
326 | return NULL; | |
327 | ||
328 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
329 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
330 | struct imsm_map *map = dev->vol.map; | |
331 | ||
332 | for (j = 0; j < map->num_members; j++) { | |
333 | __u32 index = get_imsm_disk_idx(map, j); | |
334 | ||
335 | if (index == dl->index) | |
336 | memberships++; | |
337 | } | |
338 | } | |
339 | rv = malloc(sizeof(struct extent) * (memberships + 1)); | |
340 | if (!rv) | |
341 | return NULL; | |
342 | e = rv; | |
343 | ||
344 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
345 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
346 | struct imsm_map *map = dev->vol.map; | |
347 | ||
348 | for (j = 0; j < map->num_members; j++) { | |
349 | __u32 index = get_imsm_disk_idx(map, j); | |
350 | ||
351 | if (index == dl->index) { | |
352 | e->start = __le32_to_cpu(map->pba_of_lba0); | |
353 | e->size = __le32_to_cpu(map->blocks_per_member); | |
354 | e++; | |
355 | } | |
356 | } | |
357 | } | |
358 | qsort(rv, memberships, sizeof(*rv), cmp_extent); | |
359 | ||
360 | e->start = __le32_to_cpu(disk->total_blocks) - | |
361 | (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS); | |
362 | e->size = 0; | |
363 | return rv; | |
364 | } | |
365 | ||
366 | #ifndef MDASSEMBLE | |
367 | static void print_imsm_dev(struct imsm_dev *dev, int index) | |
368 | { | |
369 | __u64 sz; | |
370 | int slot; | |
371 | struct imsm_map *map = dev->vol.map; | |
372 | ||
373 | printf("\n"); | |
374 | printf("[%s]:\n", dev->volume); | |
375 | printf(" RAID Level : %d\n", get_imsm_raid_level(map)); | |
376 | printf(" Members : %d\n", map->num_members); | |
377 | for (slot = 0; slot < map->num_members; slot++) | |
378 | if (index == get_imsm_disk_idx(map, slot)) | |
379 | break; | |
380 | if (slot < map->num_members) | |
381 | printf(" This Slot : %d\n", slot); | |
382 | else | |
383 | printf(" This Slot : ?\n"); | |
384 | sz = __le32_to_cpu(dev->size_high); | |
385 | sz <<= 32; | |
386 | sz += __le32_to_cpu(dev->size_low); | |
387 | printf(" Array Size : %llu%s\n", (unsigned long long)sz, | |
388 | human_size(sz * 512)); | |
389 | sz = __le32_to_cpu(map->blocks_per_member); | |
390 | printf(" Per Dev Size : %llu%s\n", (unsigned long long)sz, | |
391 | human_size(sz * 512)); | |
392 | printf(" Sector Offset : %u\n", | |
393 | __le32_to_cpu(map->pba_of_lba0)); | |
394 | printf(" Num Stripes : %u\n", | |
395 | __le32_to_cpu(map->num_data_stripes)); | |
396 | printf(" Chunk Size : %u KiB\n", | |
397 | __le16_to_cpu(map->blocks_per_strip) / 2); | |
398 | printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks)); | |
399 | printf(" Migrate State : %s\n", dev->vol.migr_state ? "migrating" : "idle"); | |
400 | printf(" Dirty State : %s\n", dev->vol.dirty ? "dirty" : "clean"); | |
401 | printf(" Map State : %s\n", map_state_str[map->map_state]); | |
402 | } | |
403 | ||
404 | static void print_imsm_disk(struct imsm_super *mpb, int index) | |
405 | { | |
406 | struct imsm_disk *disk = __get_imsm_disk(mpb, index); | |
407 | char str[MAX_RAID_SERIAL_LEN]; | |
408 | __u32 s; | |
409 | __u64 sz; | |
410 | ||
411 | if (index < 0) | |
412 | return; | |
413 | ||
414 | printf("\n"); | |
415 | snprintf(str, MAX_RAID_SERIAL_LEN, "%s", disk->serial); | |
416 | printf(" Disk%02d Serial : %s\n", index, str); | |
417 | s = __le32_to_cpu(disk->status); | |
418 | printf(" State :%s%s%s%s\n", s&SPARE_DISK ? " spare" : "", | |
419 | s&CONFIGURED_DISK ? " active" : "", | |
420 | s&FAILED_DISK ? " failed" : "", | |
421 | s&USABLE_DISK ? " usable" : ""); | |
422 | printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id)); | |
423 | sz = __le32_to_cpu(disk->total_blocks) - | |
424 | (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS * mpb->num_raid_devs); | |
425 | printf(" Usable Size : %llu%s\n", (unsigned long long)sz, | |
426 | human_size(sz * 512)); | |
427 | } | |
428 | ||
429 | static void examine_super_imsm(struct supertype *st, char *homehost) | |
430 | { | |
431 | struct intel_super *super = st->sb; | |
432 | struct imsm_super *mpb = super->anchor; | |
433 | char str[MAX_SIGNATURE_LENGTH]; | |
434 | int i; | |
435 | __u32 sum; | |
436 | ||
437 | snprintf(str, MPB_SIG_LEN, "%s", mpb->sig); | |
438 | printf(" Magic : %s\n", str); | |
439 | snprintf(str, strlen(MPB_VERSION_RAID0), "%s", get_imsm_version(mpb)); | |
440 | printf(" Version : %s\n", get_imsm_version(mpb)); | |
441 | printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num)); | |
442 | printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num)); | |
443 | sum = __le32_to_cpu(mpb->check_sum); | |
444 | printf(" Checksum : %08x %s\n", sum, | |
445 | __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect"); | |
446 | printf(" MPB Sectors : %d\n", mpb_sectors(mpb)); | |
447 | printf(" Disks : %d\n", mpb->num_disks); | |
448 | printf(" RAID Devices : %d\n", mpb->num_raid_devs); | |
449 | print_imsm_disk(mpb, super->disks->index); | |
450 | for (i = 0; i < mpb->num_raid_devs; i++) | |
451 | print_imsm_dev(__get_imsm_dev(mpb, i), super->disks->index); | |
452 | for (i = 0; i < mpb->num_disks; i++) { | |
453 | if (i == super->disks->index) | |
454 | continue; | |
455 | print_imsm_disk(mpb, i); | |
456 | } | |
457 | } | |
458 | ||
459 | static void brief_examine_super_imsm(struct supertype *st) | |
460 | { | |
461 | struct intel_super *super = st->sb; | |
462 | ||
463 | printf("ARRAY /dev/imsm family=%08x metadata=external:imsm\n", | |
464 | __le32_to_cpu(super->anchor->family_num)); | |
465 | } | |
466 | ||
467 | static void detail_super_imsm(struct supertype *st, char *homehost) | |
468 | { | |
469 | printf("%s\n", __FUNCTION__); | |
470 | } | |
471 | ||
472 | static void brief_detail_super_imsm(struct supertype *st) | |
473 | { | |
474 | printf("%s\n", __FUNCTION__); | |
475 | } | |
476 | #endif | |
477 | ||
478 | static int match_home_imsm(struct supertype *st, char *homehost) | |
479 | { | |
480 | printf("%s\n", __FUNCTION__); | |
481 | ||
482 | return 0; | |
483 | } | |
484 | ||
485 | static void uuid_from_super_imsm(struct supertype *st, int uuid[4]) | |
486 | { | |
487 | printf("%s\n", __FUNCTION__); | |
488 | } | |
489 | ||
490 | #if 0 | |
491 | static void | |
492 | get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p) | |
493 | { | |
494 | __u8 *v = get_imsm_version(mpb); | |
495 | __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH; | |
496 | char major[] = { 0, 0, 0 }; | |
497 | char minor[] = { 0 ,0, 0 }; | |
498 | char patch[] = { 0, 0, 0 }; | |
499 | char *ver_parse[] = { major, minor, patch }; | |
500 | int i, j; | |
501 | ||
502 | i = j = 0; | |
503 | while (*v != '\0' && v < end) { | |
504 | if (*v != '.' && j < 2) | |
505 | ver_parse[i][j++] = *v; | |
506 | else { | |
507 | i++; | |
508 | j = 0; | |
509 | } | |
510 | v++; | |
511 | } | |
512 | ||
513 | *m = strtol(minor, NULL, 0); | |
514 | *p = strtol(patch, NULL, 0); | |
515 | } | |
516 | #endif | |
517 | ||
518 | static int imsm_level_to_layout(int level) | |
519 | { | |
520 | switch (level) { | |
521 | case 0: | |
522 | case 1: | |
523 | return 0; | |
524 | case 5: | |
525 | case 6: | |
526 | return ALGORITHM_LEFT_SYMMETRIC; | |
527 | case 10: | |
528 | return 0x102; //FIXME is this correct? | |
529 | } | |
530 | return -1; | |
531 | } | |
532 | ||
533 | static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info) | |
534 | { | |
535 | struct intel_super *super = st->sb; | |
536 | struct imsm_dev *dev = get_imsm_dev(super, super->current_vol); | |
537 | struct imsm_map *map = &dev->vol.map[0]; | |
538 | ||
539 | info->container_member = super->current_vol; | |
540 | info->array.raid_disks = map->num_members; | |
541 | info->array.level = get_imsm_raid_level(map); | |
542 | info->array.layout = imsm_level_to_layout(info->array.level); | |
543 | info->array.md_minor = -1; | |
544 | info->array.ctime = 0; | |
545 | info->array.utime = 0; | |
546 | info->array.chunk_size = __le16_to_cpu(map->blocks_per_strip * 512); | |
547 | ||
548 | info->data_offset = __le32_to_cpu(map->pba_of_lba0); | |
549 | info->component_size = __le32_to_cpu(map->blocks_per_member); | |
550 | ||
551 | info->disk.major = 0; | |
552 | info->disk.minor = 0; | |
553 | ||
554 | sprintf(info->text_version, "/%s/%d", | |
555 | devnum2devname(st->container_dev), | |
556 | info->container_member); | |
557 | } | |
558 | ||
559 | ||
560 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info) | |
561 | { | |
562 | struct intel_super *super = st->sb; | |
563 | struct imsm_disk *disk; | |
564 | __u32 s; | |
565 | ||
566 | if (super->current_vol >= 0) { | |
567 | getinfo_super_imsm_volume(st, info); | |
568 | return; | |
569 | } | |
570 | info->array.raid_disks = super->anchor->num_disks; | |
571 | info->array.level = LEVEL_CONTAINER; | |
572 | info->array.layout = 0; | |
573 | info->array.md_minor = -1; | |
574 | info->array.ctime = 0; /* N/A for imsm */ | |
575 | info->array.utime = 0; | |
576 | info->array.chunk_size = 0; | |
577 | ||
578 | info->disk.major = 0; | |
579 | info->disk.minor = 0; | |
580 | info->disk.raid_disk = -1; | |
581 | info->reshape_active = 0; | |
582 | strcpy(info->text_version, "imsm"); | |
583 | info->disk.number = -1; | |
584 | info->disk.state = 0; | |
585 | ||
586 | if (super->disks) { | |
587 | disk = get_imsm_disk(super, super->disks->index); | |
588 | if (!disk) { | |
589 | info->disk.number = -1; | |
590 | info->disk.raid_disk = -1; | |
591 | return; | |
592 | } | |
593 | info->disk.number = super->disks->index; | |
594 | info->disk.raid_disk = super->disks->index; | |
595 | info->data_offset = __le32_to_cpu(disk->total_blocks) - | |
596 | (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS); | |
597 | info->component_size = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
598 | s = __le32_to_cpu(disk->status); | |
599 | info->disk.state = s & CONFIGURED_DISK ? (1 << MD_DISK_ACTIVE) : 0; | |
600 | info->disk.state |= s & FAILED_DISK ? (1 << MD_DISK_FAULTY) : 0; | |
601 | info->disk.state |= s & USABLE_DISK ? (1 << MD_DISK_SYNC) : 0; | |
602 | } | |
603 | } | |
604 | ||
605 | static int update_super_imsm(struct supertype *st, struct mdinfo *info, | |
606 | char *update, char *devname, int verbose, | |
607 | int uuid_set, char *homehost) | |
608 | { | |
609 | /* FIXME */ | |
610 | ||
611 | /* For 'assemble' and 'force' we need to return non-zero if any | |
612 | * change was made. For others, the return value is ignored. | |
613 | * Update options are: | |
614 | * force-one : This device looks a bit old but needs to be included, | |
615 | * update age info appropriately. | |
616 | * assemble: clear any 'faulty' flag to allow this device to | |
617 | * be assembled. | |
618 | * force-array: Array is degraded but being forced, mark it clean | |
619 | * if that will be needed to assemble it. | |
620 | * | |
621 | * newdev: not used ???? | |
622 | * grow: Array has gained a new device - this is currently for | |
623 | * linear only | |
624 | * resync: mark as dirty so a resync will happen. | |
625 | * name: update the name - preserving the homehost | |
626 | * | |
627 | * Following are not relevant for this imsm: | |
628 | * sparc2.2 : update from old dodgey metadata | |
629 | * super-minor: change the preferred_minor number | |
630 | * summaries: update redundant counters. | |
631 | * uuid: Change the uuid of the array to match watch is given | |
632 | * homehost: update the recorded homehost | |
633 | * _reshape_progress: record new reshape_progress position. | |
634 | */ | |
635 | int rv = 0; | |
636 | //struct intel_super *super = st->sb; | |
637 | //struct imsm_super *mpb = super->mpb; | |
638 | ||
639 | if (strcmp(update, "grow") == 0) { | |
640 | } | |
641 | if (strcmp(update, "resync") == 0) { | |
642 | /* dev->vol.dirty = 1; */ | |
643 | } | |
644 | ||
645 | /* IMSM has no concept of UUID or homehost */ | |
646 | ||
647 | return rv; | |
648 | } | |
649 | ||
650 | static size_t disks_to_mpb_size(int disks) | |
651 | { | |
652 | size_t size; | |
653 | ||
654 | size = sizeof(struct imsm_super); | |
655 | size += (disks - 1) * sizeof(struct imsm_disk); | |
656 | size += 2 * sizeof(struct imsm_dev); | |
657 | /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */ | |
658 | size += (4 - 2) * sizeof(struct imsm_map); | |
659 | /* 4 possible disk_ord_tbl's */ | |
660 | size += 4 * (disks - 1) * sizeof(__u32); | |
661 | ||
662 | return size; | |
663 | } | |
664 | ||
665 | static __u64 avail_size_imsm(struct supertype *st, __u64 devsize) | |
666 | { | |
667 | if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS)) | |
668 | return 0; | |
669 | ||
670 | return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS); | |
671 | } | |
672 | ||
673 | static int compare_super_imsm(struct supertype *st, struct supertype *tst) | |
674 | { | |
675 | /* | |
676 | * return: | |
677 | * 0 same, or first was empty, and second was copied | |
678 | * 1 second had wrong number | |
679 | * 2 wrong uuid | |
680 | * 3 wrong other info | |
681 | */ | |
682 | struct intel_super *first = st->sb; | |
683 | struct intel_super *sec = tst->sb; | |
684 | ||
685 | if (!first) { | |
686 | st->sb = tst->sb; | |
687 | tst->sb = NULL; | |
688 | return 0; | |
689 | } | |
690 | ||
691 | if (memcmp(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH) != 0) | |
692 | return 3; | |
693 | if (first->anchor->family_num != sec->anchor->family_num) | |
694 | return 3; | |
695 | if (first->anchor->mpb_size != sec->anchor->mpb_size) | |
696 | return 3; | |
697 | if (first->anchor->check_sum != sec->anchor->check_sum) | |
698 | return 3; | |
699 | ||
700 | return 0; | |
701 | } | |
702 | ||
703 | static void fd2devname(int fd, char *name) | |
704 | { | |
705 | struct stat st; | |
706 | char path[256]; | |
707 | char dname[100]; | |
708 | char *nm; | |
709 | int rv; | |
710 | ||
711 | name[0] = '\0'; | |
712 | if (fstat(fd, &st) != 0) | |
713 | return; | |
714 | sprintf(path, "/sys/dev/block/%d:%d", | |
715 | major(st.st_rdev), minor(st.st_rdev)); | |
716 | ||
717 | rv = readlink(path, dname, sizeof(dname)); | |
718 | if (rv <= 0) | |
719 | return; | |
720 | ||
721 | dname[rv] = '\0'; | |
722 | nm = strrchr(dname, '/'); | |
723 | nm++; | |
724 | snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm); | |
725 | } | |
726 | ||
727 | ||
728 | extern int scsi_get_serial(int fd, void *buf, size_t buf_len); | |
729 | ||
730 | static int imsm_read_serial(int fd, char *devname, | |
731 | __u8 serial[MAX_RAID_SERIAL_LEN]) | |
732 | { | |
733 | unsigned char scsi_serial[255]; | |
734 | int rv; | |
735 | int rsp_len; | |
736 | int i, cnt; | |
737 | ||
738 | memset(scsi_serial, 0, sizeof(scsi_serial)); | |
739 | ||
740 | if (imsm_env_devname_as_serial()) { | |
741 | char name[MAX_RAID_SERIAL_LEN]; | |
742 | ||
743 | fd2devname(fd, name); | |
744 | strcpy((char *) serial, name); | |
745 | return 0; | |
746 | } | |
747 | ||
748 | rv = scsi_get_serial(fd, scsi_serial, sizeof(scsi_serial)); | |
749 | ||
750 | if (rv != 0) { | |
751 | if (devname) | |
752 | fprintf(stderr, | |
753 | Name ": Failed to retrieve serial for %s\n", | |
754 | devname); | |
755 | return rv; | |
756 | } | |
757 | ||
758 | rsp_len = scsi_serial[3]; | |
759 | for (i = 0, cnt = 0; i < rsp_len; i++) { | |
760 | if (!isspace(scsi_serial[4 + i])) | |
761 | serial[cnt++] = scsi_serial[4 + i]; | |
762 | if (cnt == MAX_RAID_SERIAL_LEN) | |
763 | break; | |
764 | } | |
765 | ||
766 | serial[MAX_RAID_SERIAL_LEN - 1] = '\0'; | |
767 | ||
768 | return 0; | |
769 | } | |
770 | ||
771 | static int | |
772 | load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd) | |
773 | { | |
774 | struct dl *dl; | |
775 | struct stat stb; | |
776 | struct imsm_disk *disk; | |
777 | int rv; | |
778 | int i; | |
779 | ||
780 | dl = malloc(sizeof(*dl)); | |
781 | disk = malloc(sizeof(*disk)); | |
782 | if (!dl || !disk) { | |
783 | if (devname) | |
784 | fprintf(stderr, | |
785 | Name ": failed to allocate disk buffer for %s\n", | |
786 | devname); | |
787 | if (disk) | |
788 | free(disk); | |
789 | if (dl) | |
790 | free(dl); | |
791 | return 2; | |
792 | } | |
793 | memset(dl, 0, sizeof(*dl)); | |
794 | memset(disk, 0, sizeof(*disk)); | |
795 | ||
796 | fstat(fd, &stb); | |
797 | dl->major = major(stb.st_rdev); | |
798 | dl->minor = minor(stb.st_rdev); | |
799 | dl->next = super->disks; | |
800 | dl->fd = keep_fd ? fd : -1; | |
801 | dl->devname = devname ? strdup(devname) : NULL; | |
802 | dl->index = -1; | |
803 | super->disks = dl; | |
804 | rv = imsm_read_serial(fd, devname, dl->serial); | |
805 | ||
806 | if (rv != 0) | |
807 | return 2; | |
808 | ||
809 | /* look up this disk's index */ | |
810 | for (i = 0; i < super->anchor->num_disks; i++) { | |
811 | struct imsm_disk *disk_iter; | |
812 | ||
813 | disk_iter = __get_imsm_disk(super->anchor, i); | |
814 | ||
815 | if (memcmp(disk_iter->serial, dl->serial, | |
816 | MAX_RAID_SERIAL_LEN) == 0) { | |
817 | *disk = *disk_iter; | |
818 | super->disk_tbl[i] = disk; | |
819 | dl->index = i; | |
820 | break; | |
821 | } | |
822 | } | |
823 | ||
824 | if (i == super->anchor->num_disks) { | |
825 | if (devname) | |
826 | fprintf(stderr, | |
827 | Name ": failed to match serial \'%s\' for %s\n", | |
828 | dl->serial, devname); | |
829 | free(disk); | |
830 | return 0; | |
831 | } | |
832 | ||
833 | return 0; | |
834 | } | |
835 | ||
836 | static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src) | |
837 | { | |
838 | int i; | |
839 | ||
840 | *dest = *src; | |
841 | ||
842 | for (i = 0; i < src->vol.map[0].num_members; i++) | |
843 | dest->vol.map[0].disk_ord_tbl[i] = src->vol.map[0].disk_ord_tbl[i]; | |
844 | ||
845 | if (!src->vol.migr_state) | |
846 | return; | |
847 | ||
848 | dest->vol.map[1] = src->vol.map[1]; | |
849 | for (i = 0; i < src->vol.map[1].num_members; i++) | |
850 | dest->vol.map[1].disk_ord_tbl[i] = src->vol.map[1].disk_ord_tbl[i]; | |
851 | } | |
852 | ||
853 | static int parse_raid_devices(struct intel_super *super) | |
854 | { | |
855 | int i; | |
856 | struct imsm_dev *dev_new; | |
857 | size_t len; | |
858 | ||
859 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
860 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
861 | ||
862 | len = sizeof_imsm_dev(dev_iter); | |
863 | dev_new = malloc(len); | |
864 | if (!dev_new) | |
865 | return 1; | |
866 | imsm_copy_dev(dev_new, dev_iter); | |
867 | super->dev_tbl[i] = dev_new; | |
868 | } | |
869 | ||
870 | return 0; | |
871 | } | |
872 | ||
873 | /* load_imsm_mpb - read matrix metadata | |
874 | * allocates super->mpb to be freed by free_super | |
875 | */ | |
876 | static int load_imsm_mpb(int fd, struct intel_super *super, char *devname) | |
877 | { | |
878 | unsigned long long dsize; | |
879 | unsigned long long sectors; | |
880 | struct stat; | |
881 | struct imsm_super *anchor; | |
882 | __u32 check_sum; | |
883 | int rc; | |
884 | ||
885 | get_dev_size(fd, NULL, &dsize); | |
886 | ||
887 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) { | |
888 | if (devname) | |
889 | fprintf(stderr, | |
890 | Name ": Cannot seek to anchor block on %s: %s\n", | |
891 | devname, strerror(errno)); | |
892 | return 1; | |
893 | } | |
894 | ||
895 | if (posix_memalign((void**)&anchor, 512, 512) != 0) { | |
896 | if (devname) | |
897 | fprintf(stderr, | |
898 | Name ": Failed to allocate imsm anchor buffer" | |
899 | " on %s\n", devname); | |
900 | return 1; | |
901 | } | |
902 | if (read(fd, anchor, 512) != 512) { | |
903 | if (devname) | |
904 | fprintf(stderr, | |
905 | Name ": Cannot read anchor block on %s: %s\n", | |
906 | devname, strerror(errno)); | |
907 | free(anchor); | |
908 | return 1; | |
909 | } | |
910 | ||
911 | if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) { | |
912 | if (devname) | |
913 | fprintf(stderr, | |
914 | Name ": no IMSM anchor on %s\n", devname); | |
915 | free(anchor); | |
916 | return 2; | |
917 | } | |
918 | ||
919 | super->len = __le32_to_cpu(anchor->mpb_size); | |
920 | super->len = ROUND_UP(anchor->mpb_size, 512); | |
921 | if (posix_memalign(&super->buf, 512, super->len) != 0) { | |
922 | if (devname) | |
923 | fprintf(stderr, | |
924 | Name ": unable to allocate %zu byte mpb buffer\n", | |
925 | super->len); | |
926 | free(anchor); | |
927 | return 2; | |
928 | } | |
929 | memcpy(super->buf, anchor, 512); | |
930 | ||
931 | sectors = mpb_sectors(anchor) - 1; | |
932 | free(anchor); | |
933 | if (!sectors) { | |
934 | rc = load_imsm_disk(fd, super, devname, 0); | |
935 | if (rc == 0) | |
936 | rc = parse_raid_devices(super); | |
937 | return rc; | |
938 | } | |
939 | ||
940 | /* read the extended mpb */ | |
941 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) { | |
942 | if (devname) | |
943 | fprintf(stderr, | |
944 | Name ": Cannot seek to extended mpb on %s: %s\n", | |
945 | devname, strerror(errno)); | |
946 | return 1; | |
947 | } | |
948 | ||
949 | if (read(fd, super->buf + 512, super->len - 512) != super->len - 512) { | |
950 | if (devname) | |
951 | fprintf(stderr, | |
952 | Name ": Cannot read extended mpb on %s: %s\n", | |
953 | devname, strerror(errno)); | |
954 | return 2; | |
955 | } | |
956 | ||
957 | check_sum = __gen_imsm_checksum(super->anchor); | |
958 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
959 | if (devname) | |
960 | fprintf(stderr, | |
961 | Name ": IMSM checksum %x != %x on %s\n", | |
962 | check_sum, __le32_to_cpu(super->anchor->check_sum), | |
963 | devname); | |
964 | return 2; | |
965 | } | |
966 | ||
967 | rc = load_imsm_disk(fd, super, devname, 0); | |
968 | if (rc == 0) | |
969 | rc = parse_raid_devices(super); | |
970 | return rc; | |
971 | } | |
972 | ||
973 | static void free_imsm_disks(struct intel_super *super) | |
974 | { | |
975 | int i; | |
976 | ||
977 | while (super->disks) { | |
978 | struct dl *d = super->disks; | |
979 | ||
980 | super->disks = d->next; | |
981 | if (d->fd >= 0) | |
982 | close(d->fd); | |
983 | if (d->devname) | |
984 | free(d->devname); | |
985 | free(d); | |
986 | } | |
987 | for (i = 0; i < IMSM_MAX_DISKS; i++) | |
988 | if (super->disk_tbl[i]) { | |
989 | free(super->disk_tbl[i]); | |
990 | super->disk_tbl[i] = NULL; | |
991 | } | |
992 | } | |
993 | ||
994 | static void free_imsm(struct intel_super *super) | |
995 | { | |
996 | int i; | |
997 | ||
998 | if (super->buf) | |
999 | free(super->buf); | |
1000 | free_imsm_disks(super); | |
1001 | for (i = 0; i < IMSM_MAX_RAID_DEVS; i++) | |
1002 | if (super->dev_tbl[i]) | |
1003 | free(super->dev_tbl[i]); | |
1004 | free(super); | |
1005 | } | |
1006 | ||
1007 | ||
1008 | static void free_super_imsm(struct supertype *st) | |
1009 | { | |
1010 | struct intel_super *super = st->sb; | |
1011 | ||
1012 | if (!super) | |
1013 | return; | |
1014 | ||
1015 | free_imsm(super); | |
1016 | st->sb = NULL; | |
1017 | } | |
1018 | ||
1019 | static struct intel_super *alloc_super(int creating_imsm) | |
1020 | { | |
1021 | struct intel_super *super = malloc(sizeof(*super)); | |
1022 | ||
1023 | if (super) { | |
1024 | memset(super, 0, sizeof(*super)); | |
1025 | super->creating_imsm = creating_imsm; | |
1026 | super->current_vol = -1; | |
1027 | } | |
1028 | ||
1029 | return super; | |
1030 | } | |
1031 | ||
1032 | #ifndef MDASSEMBLE | |
1033 | static int load_super_imsm_all(struct supertype *st, int fd, void **sbp, | |
1034 | char *devname, int keep_fd) | |
1035 | { | |
1036 | struct mdinfo *sra; | |
1037 | struct intel_super *super; | |
1038 | struct mdinfo *sd, *best = NULL; | |
1039 | __u32 bestgen = 0; | |
1040 | __u32 gen; | |
1041 | char nm[20]; | |
1042 | int dfd; | |
1043 | int rv; | |
1044 | ||
1045 | /* check if this disk is a member of an active array */ | |
1046 | sra = sysfs_read(fd, 0, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE); | |
1047 | if (!sra) | |
1048 | return 1; | |
1049 | ||
1050 | if (sra->array.major_version != -1 || | |
1051 | sra->array.minor_version != -2 || | |
1052 | strcmp(sra->text_version, "imsm") != 0) | |
1053 | return 1; | |
1054 | ||
1055 | super = alloc_super(0); | |
1056 | if (!super) | |
1057 | return 1; | |
1058 | ||
1059 | /* find the most up to date disk in this array */ | |
1060 | for (sd = sra->devs; sd; sd = sd->next) { | |
1061 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); | |
1062 | dfd = dev_open(nm, keep_fd ? O_RDWR : O_RDONLY); | |
1063 | if (!dfd) { | |
1064 | free_imsm(super); | |
1065 | return 2; | |
1066 | } | |
1067 | rv = load_imsm_mpb(dfd, super, NULL); | |
1068 | if (!keep_fd) | |
1069 | close(dfd); | |
1070 | if (rv == 0) { | |
1071 | gen = __le32_to_cpu(super->anchor->generation_num); | |
1072 | if (!best || gen > bestgen) { | |
1073 | bestgen = gen; | |
1074 | best = sd; | |
1075 | } | |
1076 | } else { | |
1077 | free_imsm(super); | |
1078 | return 2; | |
1079 | } | |
1080 | } | |
1081 | ||
1082 | if (!best) { | |
1083 | free_imsm(super); | |
1084 | return 1; | |
1085 | } | |
1086 | ||
1087 | /* load the most up to date anchor */ | |
1088 | sprintf(nm, "%d:%d", best->disk.major, best->disk.minor); | |
1089 | dfd = dev_open(nm, O_RDONLY); | |
1090 | if (!dfd) { | |
1091 | free_imsm(super); | |
1092 | return 1; | |
1093 | } | |
1094 | rv = load_imsm_mpb(dfd, super, NULL); | |
1095 | close(dfd); | |
1096 | if (rv != 0) { | |
1097 | free_imsm(super); | |
1098 | return 2; | |
1099 | } | |
1100 | ||
1101 | /* reset the disk list */ | |
1102 | free_imsm_disks(super); | |
1103 | ||
1104 | /* populate disk list */ | |
1105 | for (sd = sra->devs ; sd ; sd = sd->next) { | |
1106 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); | |
1107 | dfd = dev_open(nm, keep_fd? O_RDWR : O_RDONLY); | |
1108 | if (!dfd) { | |
1109 | free_imsm(super); | |
1110 | return 2; | |
1111 | } | |
1112 | load_imsm_disk(dfd, super, NULL, keep_fd); | |
1113 | if (!keep_fd) | |
1114 | close(dfd); | |
1115 | } | |
1116 | ||
1117 | if (st->subarray[0]) { | |
1118 | if (atoi(st->subarray) <= super->anchor->num_raid_devs) | |
1119 | super->current_vol = atoi(st->subarray); | |
1120 | else | |
1121 | return 1; | |
1122 | } | |
1123 | ||
1124 | *sbp = super; | |
1125 | if (st->ss == NULL) { | |
1126 | st->ss = &super_imsm; | |
1127 | st->minor_version = 0; | |
1128 | st->max_devs = IMSM_MAX_DEVICES; | |
1129 | st->container_dev = fd2devnum(fd); | |
1130 | } | |
1131 | ||
1132 | return 0; | |
1133 | } | |
1134 | #endif | |
1135 | ||
1136 | static int load_super_imsm(struct supertype *st, int fd, char *devname) | |
1137 | { | |
1138 | struct intel_super *super; | |
1139 | int rv; | |
1140 | ||
1141 | #ifndef MDASSEMBLE | |
1142 | if (load_super_imsm_all(st, fd, &st->sb, devname, 1) == 0) | |
1143 | return 0; | |
1144 | #endif | |
1145 | if (st->subarray[0]) | |
1146 | return 1; /* FIXME */ | |
1147 | ||
1148 | super = alloc_super(0); | |
1149 | if (!super) { | |
1150 | fprintf(stderr, | |
1151 | Name ": malloc of %zu failed.\n", | |
1152 | sizeof(*super)); | |
1153 | return 1; | |
1154 | } | |
1155 | ||
1156 | rv = load_imsm_mpb(fd, super, devname); | |
1157 | ||
1158 | if (rv) { | |
1159 | if (devname) | |
1160 | fprintf(stderr, | |
1161 | Name ": Failed to load all information " | |
1162 | "sections on %s\n", devname); | |
1163 | free_imsm(super); | |
1164 | return rv; | |
1165 | } | |
1166 | ||
1167 | st->sb = super; | |
1168 | if (st->ss == NULL) { | |
1169 | st->ss = &super_imsm; | |
1170 | st->minor_version = 0; | |
1171 | st->max_devs = IMSM_MAX_DEVICES; | |
1172 | } | |
1173 | ||
1174 | return 0; | |
1175 | } | |
1176 | ||
1177 | static __u16 info_to_blocks_per_strip(mdu_array_info_t *info) | |
1178 | { | |
1179 | if (info->level == 1) | |
1180 | return 128; | |
1181 | return info->chunk_size >> 9; | |
1182 | } | |
1183 | ||
1184 | static __u32 info_to_num_data_stripes(mdu_array_info_t *info) | |
1185 | { | |
1186 | __u32 num_stripes; | |
1187 | ||
1188 | num_stripes = (info->size * 2) / info_to_blocks_per_strip(info); | |
1189 | if (info->level == 1) | |
1190 | num_stripes /= 2; | |
1191 | ||
1192 | return num_stripes; | |
1193 | } | |
1194 | ||
1195 | static __u32 info_to_blocks_per_member(mdu_array_info_t *info) | |
1196 | { | |
1197 | return (info->size * 2) & ~(info_to_blocks_per_strip(info) - 1); | |
1198 | } | |
1199 | ||
1200 | static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info, | |
1201 | unsigned long long size, char *name, | |
1202 | char *homehost, int *uuid) | |
1203 | { | |
1204 | /* We are creating a volume inside a pre-existing container. | |
1205 | * so st->sb is already set. | |
1206 | */ | |
1207 | struct intel_super *super = st->sb; | |
1208 | struct imsm_super *mpb = super->anchor; | |
1209 | struct imsm_dev *dev; | |
1210 | struct imsm_vol *vol; | |
1211 | struct imsm_map *map; | |
1212 | int idx = mpb->num_raid_devs; | |
1213 | int i; | |
1214 | unsigned long long array_blocks; | |
1215 | __u32 offset = 0; | |
1216 | size_t size_old, size_new; | |
1217 | ||
1218 | if (mpb->num_raid_devs >= 2) { | |
1219 | fprintf(stderr, Name": This imsm-container already has the " | |
1220 | "maximum of 2 volumes\n"); | |
1221 | return 0; | |
1222 | } | |
1223 | ||
1224 | /* ensure the mpb is large enough for the new data */ | |
1225 | size_old = __le32_to_cpu(mpb->mpb_size); | |
1226 | size_new = disks_to_mpb_size(info->nr_disks); | |
1227 | if (size_new > size_old) { | |
1228 | void *mpb_new; | |
1229 | size_t size_round = ROUND_UP(size_new, 512); | |
1230 | ||
1231 | if (posix_memalign(&mpb_new, 512, size_round) != 0) { | |
1232 | fprintf(stderr, Name": could not allocate new mpb\n"); | |
1233 | return 0; | |
1234 | } | |
1235 | memcpy(mpb_new, mpb, size_old); | |
1236 | free(mpb); | |
1237 | mpb = mpb_new; | |
1238 | super->anchor = mpb_new; | |
1239 | mpb->mpb_size = __cpu_to_le32(size_new); | |
1240 | memset(mpb_new + size_old, 0, size_round - size_old); | |
1241 | } | |
1242 | super->current_vol = idx; | |
1243 | sprintf(st->subarray, "%d", idx); | |
1244 | dev = malloc(sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1)); | |
1245 | if (!dev) { | |
1246 | fprintf(stderr, Name": could not allocate raid device\n"); | |
1247 | return 0; | |
1248 | } | |
1249 | strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN); | |
1250 | array_blocks = calc_array_size(info->level, info->raid_disks, | |
1251 | info->layout, info->chunk_size, | |
1252 | info->size*2); | |
1253 | dev->size_low = __cpu_to_le32((__u32) array_blocks); | |
1254 | dev->size_high = __cpu_to_le32((__u32) (array_blocks >> 32)); | |
1255 | dev->status = __cpu_to_le32(0); | |
1256 | dev->reserved_blocks = __cpu_to_le32(0); | |
1257 | vol = &dev->vol; | |
1258 | vol->migr_state = 0; | |
1259 | vol->migr_type = 0; | |
1260 | vol->dirty = 0; | |
1261 | for (i = 0; i < idx; i++) { | |
1262 | struct imsm_dev *prev = get_imsm_dev(super, i); | |
1263 | struct imsm_map *pmap = &prev->vol.map[0]; | |
1264 | ||
1265 | offset += __le32_to_cpu(pmap->blocks_per_member); | |
1266 | offset += IMSM_RESERVED_SECTORS; | |
1267 | } | |
1268 | map = &vol->map[0]; | |
1269 | map->pba_of_lba0 = __cpu_to_le32(offset); | |
1270 | map->blocks_per_member = __cpu_to_le32(info_to_blocks_per_member(info)); | |
1271 | map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info)); | |
1272 | map->num_data_stripes = __cpu_to_le32(info_to_num_data_stripes(info)); | |
1273 | map->map_state = info->level ? IMSM_T_STATE_UNINITIALIZED : | |
1274 | IMSM_T_STATE_NORMAL; | |
1275 | ||
1276 | if (info->level == 1 && info->raid_disks > 2) { | |
1277 | fprintf(stderr, Name": imsm does not support more than 2 disks" | |
1278 | "in a raid1 volume\n"); | |
1279 | return 0; | |
1280 | } | |
1281 | if (info->level == 10) | |
1282 | map->raid_level = 1; | |
1283 | else | |
1284 | map->raid_level = info->level; | |
1285 | ||
1286 | map->num_members = info->raid_disks; | |
1287 | for (i = 0; i < map->num_members; i++) { | |
1288 | /* initialized in add_to_super */ | |
1289 | map->disk_ord_tbl[i] = __cpu_to_le32(0); | |
1290 | } | |
1291 | mpb->num_raid_devs++; | |
1292 | super->dev_tbl[super->current_vol] = dev; | |
1293 | ||
1294 | return 1; | |
1295 | } | |
1296 | ||
1297 | static int init_super_imsm(struct supertype *st, mdu_array_info_t *info, | |
1298 | unsigned long long size, char *name, | |
1299 | char *homehost, int *uuid) | |
1300 | { | |
1301 | /* This is primarily called by Create when creating a new array. | |
1302 | * We will then get add_to_super called for each component, and then | |
1303 | * write_init_super called to write it out to each device. | |
1304 | * For IMSM, Create can create on fresh devices or on a pre-existing | |
1305 | * array. | |
1306 | * To create on a pre-existing array a different method will be called. | |
1307 | * This one is just for fresh drives. | |
1308 | */ | |
1309 | struct intel_super *super; | |
1310 | struct imsm_super *mpb; | |
1311 | size_t mpb_size; | |
1312 | ||
1313 | if (!info) { | |
1314 | st->sb = NULL; | |
1315 | return 0; | |
1316 | } | |
1317 | if (st->sb) | |
1318 | return init_super_imsm_volume(st, info, size, name, homehost, | |
1319 | uuid); | |
1320 | ||
1321 | super = alloc_super(1); | |
1322 | if (!super) | |
1323 | return 0; | |
1324 | mpb_size = disks_to_mpb_size(info->nr_disks); | |
1325 | if (posix_memalign(&super->buf, 512, mpb_size) != 0) { | |
1326 | free(super); | |
1327 | return 0; | |
1328 | } | |
1329 | mpb = super->buf; | |
1330 | memset(mpb, 0, mpb_size); | |
1331 | ||
1332 | memcpy(mpb->sig, MPB_SIGNATURE, strlen(MPB_SIGNATURE)); | |
1333 | memcpy(mpb->sig + strlen(MPB_SIGNATURE), MPB_VERSION_RAID5, | |
1334 | strlen(MPB_VERSION_RAID5)); | |
1335 | mpb->mpb_size = mpb_size; | |
1336 | ||
1337 | st->sb = super; | |
1338 | return 1; | |
1339 | } | |
1340 | ||
1341 | static void add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk, | |
1342 | int fd, char *devname) | |
1343 | { | |
1344 | struct intel_super *super = st->sb; | |
1345 | struct dl *dl; | |
1346 | struct imsm_dev *dev; | |
1347 | struct imsm_map *map; | |
1348 | struct imsm_disk *disk; | |
1349 | __u32 status; | |
1350 | ||
1351 | dev = get_imsm_dev(super, super->current_vol); | |
1352 | map = &dev->vol.map[0]; | |
1353 | ||
1354 | for (dl = super->disks; dl ; dl = dl->next) | |
1355 | if (dl->major == dk->major && | |
1356 | dl->minor == dk->minor) | |
1357 | break; | |
1358 | if (!dl || ! (dk->state & (1<<MD_DISK_SYNC))) | |
1359 | return; | |
1360 | ||
1361 | map->disk_ord_tbl[dk->number] = __cpu_to_le32(dl->index); | |
1362 | ||
1363 | disk = get_imsm_disk(super, dl->index); | |
1364 | status = CONFIGURED_DISK | USABLE_DISK; | |
1365 | disk->status = __cpu_to_le32(status); | |
1366 | } | |
1367 | ||
1368 | static void add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk, | |
1369 | int fd, char *devname) | |
1370 | { | |
1371 | struct intel_super *super = st->sb; | |
1372 | struct imsm_super *mpb = super->anchor; | |
1373 | struct imsm_disk *disk; | |
1374 | struct dl *dd; | |
1375 | unsigned long long size; | |
1376 | __u32 status, id; | |
1377 | int rv; | |
1378 | struct stat stb; | |
1379 | ||
1380 | if (super->current_vol >= 0) { | |
1381 | add_to_super_imsm_volume(st, dk, fd, devname); | |
1382 | return; | |
1383 | } | |
1384 | ||
1385 | fstat(fd, &stb); | |
1386 | dd = malloc(sizeof(*dd)); | |
1387 | disk = malloc(sizeof(*disk)); | |
1388 | if (!dd || !disk) { | |
1389 | fprintf(stderr, | |
1390 | Name ": malloc failed %s:%d.\n", __func__, __LINE__); | |
1391 | if (!dd) | |
1392 | free(dd); | |
1393 | if (!disk) | |
1394 | free(disk); | |
1395 | abort(); | |
1396 | } | |
1397 | memset(dd, 0, sizeof(*dd)); | |
1398 | memset(disk, 0, sizeof(*disk)); | |
1399 | dd->major = major(stb.st_rdev); | |
1400 | dd->minor = minor(stb.st_rdev); | |
1401 | dd->index = dk->number; | |
1402 | dd->devname = devname ? strdup(devname) : NULL; | |
1403 | dd->next = super->disks; | |
1404 | dd->fd = fd; | |
1405 | rv = imsm_read_serial(fd, devname, dd->serial); | |
1406 | if (rv) { | |
1407 | fprintf(stderr, | |
1408 | Name ": failed to retrieve scsi serial, aborting\n"); | |
1409 | free(dd); | |
1410 | free(disk); | |
1411 | abort(); | |
1412 | } | |
1413 | ||
1414 | if (mpb->num_disks <= dk->number) | |
1415 | mpb->num_disks = dk->number + 1; | |
1416 | ||
1417 | get_dev_size(fd, NULL, &size); | |
1418 | size /= 512; | |
1419 | status = USABLE_DISK | SPARE_DISK; | |
1420 | strcpy((char *) disk->serial, (char *) dd->serial); | |
1421 | disk->total_blocks = __cpu_to_le32(size); | |
1422 | disk->status = __cpu_to_le32(status); | |
1423 | if (sysfs_disk_to_scsi_id(fd, &id) == 0) | |
1424 | disk->scsi_id = __cpu_to_le32(id); | |
1425 | else | |
1426 | disk->scsi_id = __cpu_to_le32(0); | |
1427 | super->disk_tbl[dd->index] = disk; | |
1428 | ||
1429 | /* update the family number if we are creating a container */ | |
1430 | if (super->creating_imsm) { | |
1431 | disk = __get_imsm_disk(mpb, dd->index); | |
1432 | *disk = *super->disk_tbl[dd->index]; /* copy in new disk */ | |
1433 | mpb->family_num = __cpu_to_le32(__gen_imsm_checksum(mpb)); | |
1434 | } | |
1435 | ||
1436 | super->disks = dd; | |
1437 | } | |
1438 | ||
1439 | static int store_imsm_mpb(int fd, struct intel_super *super); | |
1440 | ||
1441 | static int write_super_imsm(struct intel_super *super, int doclose) | |
1442 | { | |
1443 | struct imsm_super *mpb = super->anchor; | |
1444 | struct dl *d; | |
1445 | __u32 generation; | |
1446 | __u32 sum; | |
1447 | int i; | |
1448 | ||
1449 | /* 'generation' is incremented everytime the metadata is written */ | |
1450 | generation = __le32_to_cpu(mpb->generation_num); | |
1451 | generation++; | |
1452 | mpb->generation_num = __cpu_to_le32(generation); | |
1453 | ||
1454 | for (i = 0; i < mpb->num_disks; i++) | |
1455 | mpb->disk[i] = *super->disk_tbl[i]; | |
1456 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
1457 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
1458 | ||
1459 | imsm_copy_dev(dev, super->dev_tbl[i]); | |
1460 | } | |
1461 | ||
1462 | /* recalculate checksum */ | |
1463 | sum = __gen_imsm_checksum(mpb); | |
1464 | mpb->check_sum = __cpu_to_le32(sum); | |
1465 | ||
1466 | for (d = super->disks; d ; d = d->next) { | |
1467 | if (store_imsm_mpb(d->fd, super)) { | |
1468 | fprintf(stderr, "%s: failed for device %d:%d %s\n", | |
1469 | __func__, d->major, d->minor, strerror(errno)); | |
1470 | return 0; | |
1471 | } | |
1472 | if (doclose) { | |
1473 | close(d->fd); | |
1474 | d->fd = -1; | |
1475 | } | |
1476 | } | |
1477 | ||
1478 | return 1; | |
1479 | } | |
1480 | ||
1481 | static int write_init_super_imsm(struct supertype *st) | |
1482 | { | |
1483 | if (st->update_tail) { | |
1484 | /* queue the recently created array as a metadata update */ | |
1485 | size_t len; | |
1486 | struct imsm_update_create_array *u; | |
1487 | struct intel_super *super = st->sb; | |
1488 | struct imsm_dev *dev; | |
1489 | struct dl *d; | |
1490 | ||
1491 | if (super->current_vol < 0 || | |
1492 | !(dev = get_imsm_dev(super, super->current_vol))) { | |
1493 | fprintf(stderr, "%s: could not determine sub-array\n", | |
1494 | __func__); | |
1495 | return 1; | |
1496 | } | |
1497 | ||
1498 | ||
1499 | len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev); | |
1500 | u = malloc(len); | |
1501 | if (!u) { | |
1502 | fprintf(stderr, "%s: failed to allocate update buffer\n", | |
1503 | __func__); | |
1504 | return 1; | |
1505 | } | |
1506 | ||
1507 | u->type = update_create_array; | |
1508 | u->dev_idx = super->current_vol; | |
1509 | imsm_copy_dev(&u->dev, dev); | |
1510 | append_metadata_update(st, u, len); | |
1511 | ||
1512 | for (d = super->disks; d ; d = d->next) { | |
1513 | close(d->fd); | |
1514 | d->fd = -1; | |
1515 | } | |
1516 | ||
1517 | return 0; | |
1518 | } else | |
1519 | return write_super_imsm(st->sb, 1); | |
1520 | } | |
1521 | ||
1522 | static int store_zero_imsm(struct supertype *st, int fd) | |
1523 | { | |
1524 | unsigned long long dsize; | |
1525 | void *buf; | |
1526 | ||
1527 | get_dev_size(fd, NULL, &dsize); | |
1528 | ||
1529 | /* first block is stored on second to last sector of the disk */ | |
1530 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) | |
1531 | return 1; | |
1532 | ||
1533 | if (posix_memalign(&buf, 512, 512) != 0) | |
1534 | return 1; | |
1535 | ||
1536 | memset(buf, 0, 512); | |
1537 | if (write(fd, buf, 512) != 512) | |
1538 | return 1; | |
1539 | return 0; | |
1540 | } | |
1541 | ||
1542 | static int validate_geometry_imsm_container(struct supertype *st, int level, | |
1543 | int layout, int raiddisks, int chunk, | |
1544 | unsigned long long size, char *dev, | |
1545 | unsigned long long *freesize, | |
1546 | int verbose) | |
1547 | { | |
1548 | int fd; | |
1549 | unsigned long long ldsize; | |
1550 | ||
1551 | if (level != LEVEL_CONTAINER) | |
1552 | return 0; | |
1553 | if (!dev) | |
1554 | return 1; | |
1555 | ||
1556 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
1557 | if (fd < 0) { | |
1558 | if (verbose) | |
1559 | fprintf(stderr, Name ": imsm: Cannot open %s: %s\n", | |
1560 | dev, strerror(errno)); | |
1561 | return 0; | |
1562 | } | |
1563 | if (!get_dev_size(fd, dev, &ldsize)) { | |
1564 | close(fd); | |
1565 | return 0; | |
1566 | } | |
1567 | close(fd); | |
1568 | ||
1569 | *freesize = avail_size_imsm(st, ldsize >> 9); | |
1570 | ||
1571 | return 1; | |
1572 | } | |
1573 | ||
1574 | /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd | |
1575 | * FIX ME add ahci details | |
1576 | */ | |
1577 | static int validate_geometry_imsm_volume(struct supertype *st, int level, | |
1578 | int layout, int raiddisks, int chunk, | |
1579 | unsigned long long size, char *dev, | |
1580 | unsigned long long *freesize, | |
1581 | int verbose) | |
1582 | { | |
1583 | struct stat stb; | |
1584 | struct intel_super *super = st->sb; | |
1585 | struct dl *dl; | |
1586 | unsigned long long pos = 0; | |
1587 | unsigned long long maxsize; | |
1588 | struct extent *e; | |
1589 | int i; | |
1590 | ||
1591 | if (level == LEVEL_CONTAINER) | |
1592 | return 0; | |
1593 | ||
1594 | if (level == 1 && raiddisks > 2) { | |
1595 | if (verbose) | |
1596 | fprintf(stderr, Name ": imsm does not support more " | |
1597 | "than 2 in a raid1 configuration\n"); | |
1598 | return 0; | |
1599 | } | |
1600 | ||
1601 | /* We must have the container info already read in. */ | |
1602 | if (!super) | |
1603 | return 0; | |
1604 | ||
1605 | if (!dev) { | |
1606 | /* General test: make sure there is space for | |
1607 | * 'raiddisks' device extents of size 'size' at a given | |
1608 | * offset | |
1609 | */ | |
1610 | unsigned long long minsize = size*2 /* convert to blocks */; | |
1611 | unsigned long long start_offset = ~0ULL; | |
1612 | int dcnt = 0; | |
1613 | if (minsize == 0) | |
1614 | minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
1615 | for (dl = super->disks; dl ; dl = dl->next) { | |
1616 | int found = 0; | |
1617 | ||
1618 | pos = 0; | |
1619 | i = 0; | |
1620 | e = get_extents(super, dl); | |
1621 | if (!e) continue; | |
1622 | do { | |
1623 | unsigned long long esize; | |
1624 | esize = e[i].start - pos; | |
1625 | if (esize >= minsize) | |
1626 | found = 1; | |
1627 | if (found && start_offset == ~0ULL) { | |
1628 | start_offset = pos; | |
1629 | break; | |
1630 | } else if (found && pos != start_offset) { | |
1631 | found = 0; | |
1632 | break; | |
1633 | } | |
1634 | pos = e[i].start + e[i].size; | |
1635 | i++; | |
1636 | } while (e[i-1].size); | |
1637 | if (found) | |
1638 | dcnt++; | |
1639 | free(e); | |
1640 | } | |
1641 | if (dcnt < raiddisks) { | |
1642 | if (verbose) | |
1643 | fprintf(stderr, Name ": imsm: Not enough " | |
1644 | "devices with space for this array " | |
1645 | "(%d < %d)\n", | |
1646 | dcnt, raiddisks); | |
1647 | return 0; | |
1648 | } | |
1649 | return 1; | |
1650 | } | |
1651 | /* This device must be a member of the set */ | |
1652 | if (stat(dev, &stb) < 0) | |
1653 | return 0; | |
1654 | if ((S_IFMT & stb.st_mode) != S_IFBLK) | |
1655 | return 0; | |
1656 | for (dl = super->disks ; dl ; dl = dl->next) { | |
1657 | if (dl->major == major(stb.st_rdev) && | |
1658 | dl->minor == minor(stb.st_rdev)) | |
1659 | break; | |
1660 | } | |
1661 | if (!dl) { | |
1662 | if (verbose) | |
1663 | fprintf(stderr, Name ": %s is not in the " | |
1664 | "same imsm set\n", dev); | |
1665 | return 0; | |
1666 | } | |
1667 | e = get_extents(super, dl); | |
1668 | maxsize = 0; | |
1669 | i = 0; | |
1670 | if (e) do { | |
1671 | unsigned long long esize; | |
1672 | esize = e[i].start - pos; | |
1673 | if (esize >= maxsize) | |
1674 | maxsize = esize; | |
1675 | pos = e[i].start + e[i].size; | |
1676 | i++; | |
1677 | } while (e[i-1].size); | |
1678 | *freesize = maxsize; | |
1679 | ||
1680 | return 1; | |
1681 | } | |
1682 | ||
1683 | static int validate_geometry_imsm(struct supertype *st, int level, int layout, | |
1684 | int raiddisks, int chunk, unsigned long long size, | |
1685 | char *dev, unsigned long long *freesize, | |
1686 | int verbose) | |
1687 | { | |
1688 | int fd, cfd; | |
1689 | struct mdinfo *sra; | |
1690 | ||
1691 | /* if given unused devices create a container | |
1692 | * if given given devices in a container create a member volume | |
1693 | */ | |
1694 | if (level == LEVEL_CONTAINER) { | |
1695 | /* Must be a fresh device to add to a container */ | |
1696 | return validate_geometry_imsm_container(st, level, layout, | |
1697 | raiddisks, chunk, size, | |
1698 | dev, freesize, | |
1699 | verbose); | |
1700 | } | |
1701 | ||
1702 | if (st->sb) { | |
1703 | /* creating in a given container */ | |
1704 | return validate_geometry_imsm_volume(st, level, layout, | |
1705 | raiddisks, chunk, size, | |
1706 | dev, freesize, verbose); | |
1707 | } | |
1708 | ||
1709 | /* limit creation to the following levels */ | |
1710 | if (!dev) | |
1711 | switch (level) { | |
1712 | case 0: | |
1713 | case 1: | |
1714 | case 10: | |
1715 | case 5: | |
1716 | break; | |
1717 | default: | |
1718 | return 1; | |
1719 | } | |
1720 | ||
1721 | /* This device needs to be a device in an 'imsm' container */ | |
1722 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
1723 | if (fd >= 0) { | |
1724 | if (verbose) | |
1725 | fprintf(stderr, | |
1726 | Name ": Cannot create this array on device %s\n", | |
1727 | dev); | |
1728 | close(fd); | |
1729 | return 0; | |
1730 | } | |
1731 | if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) { | |
1732 | if (verbose) | |
1733 | fprintf(stderr, Name ": Cannot open %s: %s\n", | |
1734 | dev, strerror(errno)); | |
1735 | return 0; | |
1736 | } | |
1737 | /* Well, it is in use by someone, maybe an 'imsm' container. */ | |
1738 | cfd = open_container(fd); | |
1739 | if (cfd < 0) { | |
1740 | close(fd); | |
1741 | if (verbose) | |
1742 | fprintf(stderr, Name ": Cannot use %s: It is busy\n", | |
1743 | dev); | |
1744 | return 0; | |
1745 | } | |
1746 | sra = sysfs_read(cfd, 0, GET_VERSION); | |
1747 | close(fd); | |
1748 | if (sra && sra->array.major_version == -1 && | |
1749 | strcmp(sra->text_version, "imsm") == 0) { | |
1750 | /* This is a member of a imsm container. Load the container | |
1751 | * and try to create a volume | |
1752 | */ | |
1753 | struct intel_super *super; | |
1754 | ||
1755 | if (load_super_imsm_all(st, cfd, (void **) &super, NULL, 1) == 0) { | |
1756 | st->sb = super; | |
1757 | st->container_dev = fd2devnum(cfd); | |
1758 | close(cfd); | |
1759 | return validate_geometry_imsm_volume(st, level, layout, | |
1760 | raiddisks, chunk, | |
1761 | size, dev, | |
1762 | freesize, verbose); | |
1763 | } | |
1764 | close(cfd); | |
1765 | } else /* may belong to another container */ | |
1766 | return 0; | |
1767 | ||
1768 | return 1; | |
1769 | } | |
1770 | ||
1771 | static struct mdinfo *container_content_imsm(struct supertype *st) | |
1772 | { | |
1773 | /* Given a container loaded by load_super_imsm_all, | |
1774 | * extract information about all the arrays into | |
1775 | * an mdinfo tree. | |
1776 | * | |
1777 | * For each imsm_dev create an mdinfo, fill it in, | |
1778 | * then look for matching devices in super->disks | |
1779 | * and create appropriate device mdinfo. | |
1780 | */ | |
1781 | struct intel_super *super = st->sb; | |
1782 | struct imsm_super *mpb = super->anchor; | |
1783 | struct mdinfo *rest = NULL; | |
1784 | int i; | |
1785 | ||
1786 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
1787 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
1788 | struct imsm_vol *vol = &dev->vol; | |
1789 | struct imsm_map *map = vol->map; | |
1790 | struct mdinfo *this; | |
1791 | int slot; | |
1792 | ||
1793 | this = malloc(sizeof(*this)); | |
1794 | memset(this, 0, sizeof(*this)); | |
1795 | this->next = rest; | |
1796 | rest = this; | |
1797 | ||
1798 | this->array.level = get_imsm_raid_level(map); | |
1799 | this->array.raid_disks = map->num_members; | |
1800 | this->array.layout = imsm_level_to_layout(this->array.level); | |
1801 | this->array.md_minor = -1; | |
1802 | this->array.ctime = 0; | |
1803 | this->array.utime = 0; | |
1804 | this->array.chunk_size = __le16_to_cpu(map->blocks_per_strip) << 9; | |
1805 | this->array.state = !vol->dirty; | |
1806 | this->container_member = i; | |
1807 | if (map->map_state == IMSM_T_STATE_UNINITIALIZED || dev->vol.dirty) | |
1808 | this->resync_start = 0; | |
1809 | else | |
1810 | this->resync_start = ~0ULL; | |
1811 | ||
1812 | strncpy(this->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN); | |
1813 | this->name[MAX_RAID_SERIAL_LEN] = 0; | |
1814 | ||
1815 | sprintf(this->text_version, "/%s/%d", | |
1816 | devnum2devname(st->container_dev), | |
1817 | this->container_member); | |
1818 | ||
1819 | memset(this->uuid, 0, sizeof(this->uuid)); | |
1820 | ||
1821 | this->component_size = __le32_to_cpu(map->blocks_per_member); | |
1822 | ||
1823 | for (slot = 0 ; slot < map->num_members; slot++) { | |
1824 | struct imsm_disk *disk; | |
1825 | struct mdinfo *info_d; | |
1826 | struct dl *d; | |
1827 | int idx; | |
1828 | __u32 s; | |
1829 | ||
1830 | idx = __le32_to_cpu(map->disk_ord_tbl[slot] & ~(0xff << 24)); | |
1831 | for (d = super->disks; d ; d = d->next) | |
1832 | if (d->index == idx) | |
1833 | break; | |
1834 | ||
1835 | if (d == NULL) | |
1836 | break; /* shouldn't this be continue ?? */ | |
1837 | ||
1838 | info_d = malloc(sizeof(*info_d)); | |
1839 | if (!info_d) | |
1840 | break; /* ditto ?? */ | |
1841 | memset(info_d, 0, sizeof(*info_d)); | |
1842 | info_d->next = this->devs; | |
1843 | this->devs = info_d; | |
1844 | ||
1845 | disk = get_imsm_disk(super, idx); | |
1846 | s = __le32_to_cpu(disk->status); | |
1847 | ||
1848 | info_d->disk.number = d->index; | |
1849 | info_d->disk.major = d->major; | |
1850 | info_d->disk.minor = d->minor; | |
1851 | info_d->disk.raid_disk = slot; | |
1852 | info_d->disk.state = s & CONFIGURED_DISK ? (1 << MD_DISK_ACTIVE) : 0; | |
1853 | info_d->disk.state |= s & FAILED_DISK ? (1 << MD_DISK_FAULTY) : 0; | |
1854 | info_d->disk.state |= s & USABLE_DISK ? (1 << MD_DISK_SYNC) : 0; | |
1855 | ||
1856 | this->array.working_disks++; | |
1857 | ||
1858 | info_d->events = __le32_to_cpu(mpb->generation_num); | |
1859 | info_d->data_offset = __le32_to_cpu(map->pba_of_lba0); | |
1860 | info_d->component_size = __le32_to_cpu(map->blocks_per_member); | |
1861 | if (d->devname) | |
1862 | strcpy(info_d->name, d->devname); | |
1863 | } | |
1864 | } | |
1865 | ||
1866 | return rest; | |
1867 | } | |
1868 | ||
1869 | ||
1870 | static int imsm_open_new(struct supertype *c, struct active_array *a, | |
1871 | char *inst) | |
1872 | { | |
1873 | struct intel_super *super = c->sb; | |
1874 | struct imsm_super *mpb = super->anchor; | |
1875 | ||
1876 | if (atoi(inst) >= mpb->num_raid_devs) { | |
1877 | fprintf(stderr, "%s: subarry index %d, out of range\n", | |
1878 | __func__, atoi(inst)); | |
1879 | return -ENODEV; | |
1880 | } | |
1881 | ||
1882 | dprintf("imsm: open_new %s\n", inst); | |
1883 | a->info.container_member = atoi(inst); | |
1884 | return 0; | |
1885 | } | |
1886 | ||
1887 | static __u8 imsm_check_degraded(struct intel_super *super, int n, int failed) | |
1888 | { | |
1889 | struct imsm_dev *dev = get_imsm_dev(super, n); | |
1890 | struct imsm_map *map = dev->vol.map; | |
1891 | ||
1892 | if (!failed) | |
1893 | return map->map_state; | |
1894 | ||
1895 | switch (get_imsm_raid_level(map)) { | |
1896 | case 0: | |
1897 | return IMSM_T_STATE_FAILED; | |
1898 | break; | |
1899 | case 1: | |
1900 | if (failed < map->num_members) | |
1901 | return IMSM_T_STATE_DEGRADED; | |
1902 | else | |
1903 | return IMSM_T_STATE_FAILED; | |
1904 | break; | |
1905 | case 10: | |
1906 | { | |
1907 | /** | |
1908 | * check to see if any mirrors have failed, | |
1909 | * otherwise we are degraded | |
1910 | */ | |
1911 | int device_per_mirror = 2; /* FIXME is this always the case? | |
1912 | * and are they always adjacent? | |
1913 | */ | |
1914 | int failed = 0; | |
1915 | int i; | |
1916 | ||
1917 | for (i = 0; i < map->num_members; i++) { | |
1918 | int idx = get_imsm_disk_idx(map, i); | |
1919 | struct imsm_disk *disk = get_imsm_disk(super, idx); | |
1920 | ||
1921 | if (__le32_to_cpu(disk->status) & FAILED_DISK) | |
1922 | failed++; | |
1923 | ||
1924 | if (failed >= device_per_mirror) | |
1925 | return IMSM_T_STATE_FAILED; | |
1926 | ||
1927 | /* reset 'failed' for next mirror set */ | |
1928 | if (!((i + 1) % device_per_mirror)) | |
1929 | failed = 0; | |
1930 | } | |
1931 | ||
1932 | return IMSM_T_STATE_DEGRADED; | |
1933 | } | |
1934 | case 5: | |
1935 | if (failed < 2) | |
1936 | return IMSM_T_STATE_DEGRADED; | |
1937 | else | |
1938 | return IMSM_T_STATE_FAILED; | |
1939 | break; | |
1940 | default: | |
1941 | break; | |
1942 | } | |
1943 | ||
1944 | return map->map_state; | |
1945 | } | |
1946 | ||
1947 | static int imsm_count_failed(struct intel_super *super, struct imsm_map *map) | |
1948 | { | |
1949 | int i; | |
1950 | int failed = 0; | |
1951 | struct imsm_disk *disk; | |
1952 | ||
1953 | for (i = 0; i < map->num_members; i++) { | |
1954 | int idx = get_imsm_disk_idx(map, i); | |
1955 | ||
1956 | disk = get_imsm_disk(super, idx); | |
1957 | if (__le32_to_cpu(disk->status) & FAILED_DISK) | |
1958 | failed++; | |
1959 | } | |
1960 | ||
1961 | return failed; | |
1962 | } | |
1963 | ||
1964 | static void imsm_set_array_state(struct active_array *a, int consistent) | |
1965 | { | |
1966 | int inst = a->info.container_member; | |
1967 | struct intel_super *super = a->container->sb; | |
1968 | struct imsm_dev *dev = get_imsm_dev(super, inst); | |
1969 | struct imsm_map *map = &dev->vol.map[0]; | |
1970 | int dirty = !consistent; | |
1971 | int failed; | |
1972 | __u8 map_state; | |
1973 | ||
1974 | if (a->resync_start == ~0ULL) { | |
1975 | failed = imsm_count_failed(super, map); | |
1976 | map_state = imsm_check_degraded(super, inst, failed); | |
1977 | if (!failed) | |
1978 | map_state = IMSM_T_STATE_NORMAL; | |
1979 | if (map->map_state != map_state) { | |
1980 | dprintf("imsm: map_state %d: %d\n", | |
1981 | inst, map_state); | |
1982 | map->map_state = map_state; | |
1983 | super->updates_pending++; | |
1984 | } | |
1985 | } | |
1986 | ||
1987 | if (dev->vol.dirty != dirty) { | |
1988 | dprintf("imsm: mark '%s' (%llu)\n", | |
1989 | dirty?"dirty":"clean", a->resync_start); | |
1990 | ||
1991 | dev->vol.dirty = dirty; | |
1992 | super->updates_pending++; | |
1993 | } | |
1994 | } | |
1995 | ||
1996 | static void imsm_set_disk(struct active_array *a, int n, int state) | |
1997 | { | |
1998 | int inst = a->info.container_member; | |
1999 | struct intel_super *super = a->container->sb; | |
2000 | struct imsm_dev *dev = get_imsm_dev(super, inst); | |
2001 | struct imsm_map *map = dev->vol.map; | |
2002 | struct imsm_disk *disk; | |
2003 | __u32 status; | |
2004 | int failed = 0; | |
2005 | int new_failure = 0; | |
2006 | ||
2007 | if (n > map->num_members) | |
2008 | fprintf(stderr, "imsm: set_disk %d out of range 0..%d\n", | |
2009 | n, map->num_members - 1); | |
2010 | ||
2011 | if (n < 0) | |
2012 | return; | |
2013 | ||
2014 | dprintf("imsm: set_disk %d:%x\n", n, state); | |
2015 | ||
2016 | disk = get_imsm_disk(super, get_imsm_disk_idx(map, n)); | |
2017 | ||
2018 | /* check for new failures */ | |
2019 | status = __le32_to_cpu(disk->status); | |
2020 | if ((state & DS_FAULTY) && !(status & FAILED_DISK)) { | |
2021 | status |= FAILED_DISK; | |
2022 | disk->status = __cpu_to_le32(status); | |
2023 | new_failure = 1; | |
2024 | super->updates_pending++; | |
2025 | } | |
2026 | ||
2027 | /* the number of failures have changed, count up 'failed' to determine | |
2028 | * degraded / failed status | |
2029 | */ | |
2030 | if (new_failure && map->map_state != IMSM_T_STATE_FAILED) | |
2031 | failed = imsm_count_failed(super, map); | |
2032 | ||
2033 | /* determine map_state based on failed or in_sync count */ | |
2034 | if (failed) | |
2035 | map->map_state = imsm_check_degraded(super, inst, failed); | |
2036 | else if (map->map_state == IMSM_T_STATE_DEGRADED) { | |
2037 | struct mdinfo *d; | |
2038 | int working = 0; | |
2039 | ||
2040 | for (d = a->info.devs ; d ; d = d->next) | |
2041 | if (d->curr_state & DS_INSYNC) | |
2042 | working++; | |
2043 | ||
2044 | if (working == a->info.array.raid_disks) { | |
2045 | map->map_state = IMSM_T_STATE_NORMAL; | |
2046 | super->updates_pending++; | |
2047 | } | |
2048 | } | |
2049 | } | |
2050 | ||
2051 | static int store_imsm_mpb(int fd, struct intel_super *super) | |
2052 | { | |
2053 | struct imsm_super *mpb = super->anchor; | |
2054 | __u32 mpb_size = __le32_to_cpu(mpb->mpb_size); | |
2055 | unsigned long long dsize; | |
2056 | unsigned long long sectors; | |
2057 | ||
2058 | get_dev_size(fd, NULL, &dsize); | |
2059 | ||
2060 | if (mpb_size > 512) { | |
2061 | /* -1 to account for anchor */ | |
2062 | sectors = mpb_sectors(mpb) - 1; | |
2063 | ||
2064 | /* write the extended mpb to the sectors preceeding the anchor */ | |
2065 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) | |
2066 | return 1; | |
2067 | ||
2068 | if (write(fd, super->buf + 512, 512 * sectors) != 512 * sectors) | |
2069 | return 1; | |
2070 | } | |
2071 | ||
2072 | /* first block is stored on second to last sector of the disk */ | |
2073 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) | |
2074 | return 1; | |
2075 | ||
2076 | if (write(fd, super->buf, 512) != 512) | |
2077 | return 1; | |
2078 | ||
2079 | return 0; | |
2080 | } | |
2081 | ||
2082 | static void imsm_sync_metadata(struct supertype *container) | |
2083 | { | |
2084 | struct intel_super *super = container->sb; | |
2085 | ||
2086 | if (!super->updates_pending) | |
2087 | return; | |
2088 | ||
2089 | write_super_imsm(super, 0); | |
2090 | ||
2091 | super->updates_pending = 0; | |
2092 | } | |
2093 | ||
2094 | static struct mdinfo *imsm_activate_spare(struct active_array *a, | |
2095 | struct metadata_update **updates) | |
2096 | { | |
2097 | /** | |
2098 | * Take a device that is marked spare in the metadata and use it to | |
2099 | * replace a failed/vacant slot in an array. There may be a case where | |
2100 | * a device is failed in one array but active in a second. | |
2101 | * imsm_process_update catches this case and does not clear the SPARE_DISK | |
2102 | * flag, allowing the second array to start using the device on failure. | |
2103 | * SPARE_DISK is cleared when all arrays are using a device. | |
2104 | * | |
2105 | * FIXME: is this a valid use of SPARE_DISK? | |
2106 | */ | |
2107 | ||
2108 | struct intel_super *super = a->container->sb; | |
2109 | int inst = a->info.container_member; | |
2110 | struct imsm_dev *dev = get_imsm_dev(super, inst); | |
2111 | struct imsm_map *map = dev->vol.map; | |
2112 | int failed = a->info.array.raid_disks; | |
2113 | struct mdinfo *rv = NULL; | |
2114 | struct mdinfo *d; | |
2115 | struct mdinfo *di; | |
2116 | struct metadata_update *mu; | |
2117 | struct dl *dl; | |
2118 | struct imsm_update_activate_spare *u; | |
2119 | int num_spares = 0; | |
2120 | int i; | |
2121 | ||
2122 | for (d = a->info.devs ; d ; d = d->next) { | |
2123 | if ((d->curr_state & DS_FAULTY) && | |
2124 | d->state_fd >= 0) | |
2125 | /* wait for Removal to happen */ | |
2126 | return NULL; | |
2127 | if (d->state_fd >= 0) | |
2128 | failed--; | |
2129 | } | |
2130 | ||
2131 | dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n", | |
2132 | inst, failed, a->info.array.raid_disks, a->info.array.level); | |
2133 | if (imsm_check_degraded(super, inst, failed) != IMSM_T_STATE_DEGRADED) | |
2134 | return NULL; | |
2135 | ||
2136 | /* For each slot, if it is not working, find a spare */ | |
2137 | dl = super->disks; | |
2138 | for (i = 0; i < a->info.array.raid_disks; i++) { | |
2139 | for (d = a->info.devs ; d ; d = d->next) | |
2140 | if (d->disk.raid_disk == i) | |
2141 | break; | |
2142 | dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0); | |
2143 | if (d && (d->state_fd >= 0)) | |
2144 | continue; | |
2145 | ||
2146 | /* OK, this device needs recovery. Find a spare */ | |
2147 | for ( ; dl ; dl = dl->next) { | |
2148 | unsigned long long esize; | |
2149 | unsigned long long pos; | |
2150 | struct mdinfo *d2; | |
2151 | struct extent *ex; | |
2152 | struct imsm_disk *disk; | |
2153 | int j; | |
2154 | int found; | |
2155 | __u32 array_start; | |
2156 | ||
2157 | /* If in this array, skip */ | |
2158 | for (d2 = a->info.devs ; d2 ; d2 = d2->next) | |
2159 | if (d2->disk.major == dl->major && | |
2160 | d2->disk.minor == dl->minor) { | |
2161 | dprintf("%x:%x already in array\n", dl->major, dl->minor); | |
2162 | break; | |
2163 | } | |
2164 | if (d2) | |
2165 | continue; | |
2166 | ||
2167 | /* is this unused device marked as a spare? */ | |
2168 | disk = get_imsm_disk(super, dl->index); | |
2169 | if (!(__le32_to_cpu(disk->status) & SPARE_DISK)) | |
2170 | continue; | |
2171 | ||
2172 | /* We are allowed to use this device - is there space? | |
2173 | * We need a->info.component_size sectors */ | |
2174 | ex = get_extents(super, dl); | |
2175 | if (!ex) { | |
2176 | dprintf("cannot get extents\n"); | |
2177 | continue; | |
2178 | } | |
2179 | found = 0; | |
2180 | j = 0; | |
2181 | pos = 0; | |
2182 | array_start = __le32_to_cpu(map->pba_of_lba0); | |
2183 | ||
2184 | do { | |
2185 | /* check that we can start at pba_of_lba0 with | |
2186 | * a->info.component_size of space | |
2187 | */ | |
2188 | esize = ex[j].start - pos; | |
2189 | if (array_start >= pos && | |
2190 | array_start + a->info.component_size < ex[j].start) { | |
2191 | found = 1; | |
2192 | break; | |
2193 | } | |
2194 | pos = ex[j].start + ex[j].size; | |
2195 | j++; | |
2196 | ||
2197 | } while (ex[j-1].size); | |
2198 | ||
2199 | free(ex); | |
2200 | if (!found) { | |
2201 | dprintf("%x:%x does not have %llu at %d\n", | |
2202 | dl->major, dl->minor, | |
2203 | a->info.component_size, | |
2204 | __le32_to_cpu(map->pba_of_lba0)); | |
2205 | /* No room */ | |
2206 | continue; | |
2207 | } | |
2208 | ||
2209 | /* found a usable disk with enough space */ | |
2210 | di = malloc(sizeof(*di)); | |
2211 | memset(di, 0, sizeof(*di)); | |
2212 | di->disk.number = dl->index; | |
2213 | di->disk.raid_disk = i; | |
2214 | di->disk.major = dl->major; | |
2215 | di->disk.minor = dl->minor; | |
2216 | di->disk.state = 0; | |
2217 | di->data_offset = array_start; | |
2218 | di->component_size = a->info.component_size; | |
2219 | di->container_member = inst; | |
2220 | di->next = rv; | |
2221 | rv = di; | |
2222 | num_spares++; | |
2223 | dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor, | |
2224 | i, pos); | |
2225 | ||
2226 | break; | |
2227 | } | |
2228 | } | |
2229 | ||
2230 | if (!rv) | |
2231 | /* No spares found */ | |
2232 | return rv; | |
2233 | /* Now 'rv' has a list of devices to return. | |
2234 | * Create a metadata_update record to update the | |
2235 | * disk_ord_tbl for the array | |
2236 | */ | |
2237 | mu = malloc(sizeof(*mu)); | |
2238 | mu->buf = malloc(sizeof(struct imsm_update_activate_spare) * num_spares); | |
2239 | mu->space = NULL; | |
2240 | mu->len = sizeof(struct imsm_update_activate_spare) * num_spares; | |
2241 | mu->next = *updates; | |
2242 | u = (struct imsm_update_activate_spare *) mu->buf; | |
2243 | ||
2244 | for (di = rv ; di ; di = di->next) { | |
2245 | u->type = update_activate_spare; | |
2246 | u->disk_idx = di->disk.number; | |
2247 | u->slot = di->disk.raid_disk; | |
2248 | u->array = inst; | |
2249 | u->next = u + 1; | |
2250 | u++; | |
2251 | } | |
2252 | (u-1)->next = NULL; | |
2253 | *updates = mu; | |
2254 | ||
2255 | return rv; | |
2256 | } | |
2257 | ||
2258 | static int weight(unsigned int field) | |
2259 | { | |
2260 | int weight; | |
2261 | ||
2262 | for (weight = 0; field; weight++) | |
2263 | field &= field - 1; | |
2264 | ||
2265 | return weight; | |
2266 | } | |
2267 | ||
2268 | static int disks_overlap(struct imsm_map *m1, struct imsm_map *m2) | |
2269 | { | |
2270 | int i; | |
2271 | int j; | |
2272 | int idx; | |
2273 | ||
2274 | for (i = 0; i < m1->num_members; i++) { | |
2275 | idx = get_imsm_disk_idx(m1, i); | |
2276 | for (j = 0; j < m2->num_members; j++) | |
2277 | if (idx == get_imsm_disk_idx(m2, j)) | |
2278 | return 1; | |
2279 | } | |
2280 | ||
2281 | return 0; | |
2282 | } | |
2283 | ||
2284 | static void imsm_process_update(struct supertype *st, | |
2285 | struct metadata_update *update) | |
2286 | { | |
2287 | /** | |
2288 | * crack open the metadata_update envelope to find the update record | |
2289 | * update can be one of: | |
2290 | * update_activate_spare - a spare device has replaced a failed | |
2291 | * device in an array, update the disk_ord_tbl. If this disk is | |
2292 | * present in all member arrays then also clear the SPARE_DISK | |
2293 | * flag | |
2294 | */ | |
2295 | struct intel_super *super = st->sb; | |
2296 | struct imsm_super *mpb = super->anchor; | |
2297 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; | |
2298 | ||
2299 | switch (type) { | |
2300 | case update_activate_spare: { | |
2301 | struct imsm_update_activate_spare *u = (void *) update->buf; | |
2302 | struct imsm_dev *dev = get_imsm_dev(super, u->array); | |
2303 | struct imsm_map *map = &dev->vol.map[0]; | |
2304 | struct active_array *a; | |
2305 | struct imsm_disk *disk; | |
2306 | __u32 status; | |
2307 | struct dl *dl; | |
2308 | struct mdinfo *d; | |
2309 | unsigned int members; | |
2310 | unsigned int found; | |
2311 | int victim; | |
2312 | int i; | |
2313 | ||
2314 | for (dl = super->disks; dl; dl = dl->next) | |
2315 | if (dl->index == u->disk_idx) | |
2316 | break; | |
2317 | ||
2318 | if (!dl) { | |
2319 | fprintf(stderr, "error: imsm_activate_spare passed " | |
2320 | "an unknown disk_idx: %d\n", u->disk_idx); | |
2321 | return; | |
2322 | } | |
2323 | ||
2324 | super->updates_pending++; | |
2325 | ||
2326 | victim = get_imsm_disk_idx(map, u->slot); | |
2327 | map->disk_ord_tbl[u->slot] = __cpu_to_le32(u->disk_idx); | |
2328 | disk = get_imsm_disk(super, u->disk_idx); | |
2329 | status = __le32_to_cpu(disk->status); | |
2330 | status |= CONFIGURED_DISK; | |
2331 | disk->status = __cpu_to_le32(status); | |
2332 | ||
2333 | /* map unique/live arrays using the spare */ | |
2334 | members = 0; | |
2335 | found = 0; | |
2336 | for (a = st->arrays; a; a = a->next) { | |
2337 | int inst = a->info.container_member; | |
2338 | ||
2339 | dev = get_imsm_dev(super, inst); | |
2340 | map = &dev->vol.map[0]; | |
2341 | if (map->raid_level > 0) | |
2342 | members |= 1 << inst; | |
2343 | for (d = a->info.devs; d; d = d->next) | |
2344 | if (d->disk.major == dl->major && | |
2345 | d->disk.minor == dl->minor) | |
2346 | found |= 1 << inst; | |
2347 | } | |
2348 | ||
2349 | /* until all arrays that can absorb this disk have absorbed | |
2350 | * this disk it can still be considered a spare | |
2351 | */ | |
2352 | if (weight(found) >= weight(members)) { | |
2353 | status = __le32_to_cpu(disk->status); | |
2354 | status &= ~SPARE_DISK; | |
2355 | disk->status = __cpu_to_le32(status); | |
2356 | } | |
2357 | ||
2358 | /* count arrays using the victim in the metadata */ | |
2359 | found = 0; | |
2360 | for (a = st->arrays; a ; a = a->next) { | |
2361 | dev = get_imsm_dev(super, a->info.container_member); | |
2362 | map = &dev->vol.map[0]; | |
2363 | for (i = 0; i < map->num_members; i++) | |
2364 | if (victim == get_imsm_disk_idx(map, i)) | |
2365 | found++; | |
2366 | } | |
2367 | ||
2368 | /* clear some flags if the victim is no longer being | |
2369 | * utilized anywhere | |
2370 | */ | |
2371 | disk = get_imsm_disk(super, victim); | |
2372 | if (!found) { | |
2373 | status = __le32_to_cpu(disk->status); | |
2374 | status &= ~(CONFIGURED_DISK | USABLE_DISK); | |
2375 | disk->status = __cpu_to_le32(status); | |
2376 | } | |
2377 | break; | |
2378 | } | |
2379 | case update_create_array: { | |
2380 | /* someone wants to create a new array, we need to be aware of | |
2381 | * a few races/collisions: | |
2382 | * 1/ 'Create' called by two separate instances of mdadm | |
2383 | * 2/ 'Create' versus 'activate_spare': mdadm has chosen | |
2384 | * devices that have since been assimilated via | |
2385 | * activate_spare. | |
2386 | * In the event this update can not be carried out mdadm will | |
2387 | * (FIX ME) notice that its update did not take hold. | |
2388 | */ | |
2389 | struct imsm_update_create_array *u = (void *) update->buf; | |
2390 | struct imsm_dev *dev; | |
2391 | struct imsm_map *map, *new_map; | |
2392 | unsigned long long start, end; | |
2393 | unsigned long long new_start, new_end; | |
2394 | int i; | |
2395 | int overlap = 0; | |
2396 | ||
2397 | /* handle racing creates: first come first serve */ | |
2398 | if (u->dev_idx < mpb->num_raid_devs) { | |
2399 | dprintf("%s: subarray %d already defined\n", | |
2400 | __func__, u->dev_idx); | |
2401 | return; | |
2402 | } | |
2403 | ||
2404 | /* check update is next in sequence */ | |
2405 | if (u->dev_idx != mpb->num_raid_devs) { | |
2406 | dprintf("%s: can not create arrays out of sequence\n", | |
2407 | __func__); | |
2408 | return; | |
2409 | } | |
2410 | ||
2411 | new_map = &u->dev.vol.map[0]; | |
2412 | new_start = __le32_to_cpu(new_map->pba_of_lba0); | |
2413 | new_end = new_start + __le32_to_cpu(new_map->blocks_per_member); | |
2414 | ||
2415 | /* handle activate_spare versus create race: | |
2416 | * check to make sure that overlapping arrays do not include | |
2417 | * overalpping disks | |
2418 | */ | |
2419 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
2420 | dev = get_imsm_dev(super, i); | |
2421 | map = &dev->vol.map[0]; | |
2422 | start = __le32_to_cpu(map->pba_of_lba0); | |
2423 | end = start + __le32_to_cpu(map->blocks_per_member); | |
2424 | if ((new_start >= start && new_start <= end) || | |
2425 | (start >= new_start && start <= new_end)) | |
2426 | overlap = 1; | |
2427 | if (overlap && disks_overlap(map, new_map)) { | |
2428 | dprintf("%s: arrays overlap\n", __func__); | |
2429 | return; | |
2430 | } | |
2431 | } | |
2432 | /* check num_members sanity */ | |
2433 | if (new_map->num_members > mpb->num_disks) { | |
2434 | dprintf("%s: num_disks out of range\n", __func__); | |
2435 | return; | |
2436 | } | |
2437 | ||
2438 | /* check that prepare update was successful */ | |
2439 | if (!update->space) { | |
2440 | dprintf("%s: prepare update failed\n", __func__); | |
2441 | return; | |
2442 | } | |
2443 | ||
2444 | super->updates_pending++; | |
2445 | dev = update->space; | |
2446 | update->space = NULL; | |
2447 | imsm_copy_dev(dev, &u->dev); | |
2448 | super->dev_tbl[u->dev_idx] = dev; | |
2449 | mpb->num_raid_devs++; | |
2450 | ||
2451 | /* fix up flags, if arrays overlap then the drives can not be | |
2452 | * spares | |
2453 | */ | |
2454 | for (i = 0; i < map->num_members; i++) { | |
2455 | struct imsm_disk *disk; | |
2456 | __u32 status; | |
2457 | ||
2458 | disk = get_imsm_disk(super, get_imsm_disk_idx(map, i)); | |
2459 | status = __le32_to_cpu(disk->status); | |
2460 | status |= CONFIGURED_DISK; | |
2461 | if (overlap) | |
2462 | status &= ~SPARE_DISK; | |
2463 | disk->status = __cpu_to_le32(status); | |
2464 | } | |
2465 | break; | |
2466 | } | |
2467 | } | |
2468 | } | |
2469 | ||
2470 | static void imsm_prepare_update(struct supertype *st, | |
2471 | struct metadata_update *update) | |
2472 | { | |
2473 | /** | |
2474 | * Allocate space to hold new disk entries, raid-device entries or a | |
2475 | * new mpb if necessary. We currently maintain an mpb large enough to | |
2476 | * hold 2 subarrays for the given number of disks. This may not be | |
2477 | * sufficient when reshaping. | |
2478 | * | |
2479 | * FIX ME handle the reshape case. | |
2480 | * | |
2481 | * The monitor will be able to safely change super->mpb by arranging | |
2482 | * for it to be freed in check_update_queue(). I.e. the monitor thread | |
2483 | * will start using the new pointer and the manager can continue to use | |
2484 | * the old value until check_update_queue() runs. | |
2485 | */ | |
2486 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; | |
2487 | ||
2488 | switch (type) { | |
2489 | case update_create_array: { | |
2490 | struct imsm_update_create_array *u = (void *) update->buf; | |
2491 | size_t len = sizeof_imsm_dev(&u->dev); | |
2492 | ||
2493 | update->space = malloc(len); | |
2494 | break; | |
2495 | default: | |
2496 | break; | |
2497 | } | |
2498 | } | |
2499 | ||
2500 | return; | |
2501 | } | |
2502 | ||
2503 | struct superswitch super_imsm = { | |
2504 | #ifndef MDASSEMBLE | |
2505 | .examine_super = examine_super_imsm, | |
2506 | .brief_examine_super = brief_examine_super_imsm, | |
2507 | .detail_super = detail_super_imsm, | |
2508 | .brief_detail_super = brief_detail_super_imsm, | |
2509 | .write_init_super = write_init_super_imsm, | |
2510 | #endif | |
2511 | .match_home = match_home_imsm, | |
2512 | .uuid_from_super= uuid_from_super_imsm, | |
2513 | .getinfo_super = getinfo_super_imsm, | |
2514 | .update_super = update_super_imsm, | |
2515 | ||
2516 | .avail_size = avail_size_imsm, | |
2517 | ||
2518 | .compare_super = compare_super_imsm, | |
2519 | ||
2520 | .load_super = load_super_imsm, | |
2521 | .init_super = init_super_imsm, | |
2522 | .add_to_super = add_to_super_imsm, | |
2523 | .store_super = store_zero_imsm, | |
2524 | .free_super = free_super_imsm, | |
2525 | .match_metadata_desc = match_metadata_desc_imsm, | |
2526 | .container_content = container_content_imsm, | |
2527 | ||
2528 | .validate_geometry = validate_geometry_imsm, | |
2529 | .external = 1, | |
2530 | ||
2531 | /* for mdmon */ | |
2532 | .open_new = imsm_open_new, | |
2533 | .load_super = load_super_imsm, | |
2534 | .set_array_state= imsm_set_array_state, | |
2535 | .set_disk = imsm_set_disk, | |
2536 | .sync_metadata = imsm_sync_metadata, | |
2537 | .activate_spare = imsm_activate_spare, | |
2538 | .process_update = imsm_process_update, | |
2539 | .prepare_update = imsm_prepare_update, | |
2540 | }; |