]>
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 */ | |
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 | ||
87eb16df | 117 | static unsigned int sector_count(__u32 bytes) |
cdddbdbc | 118 | { |
87eb16df DW |
119 | return ((bytes + (512-1)) & (~(512-1))) / 512; |
120 | } | |
cdddbdbc | 121 | |
87eb16df DW |
122 | static unsigned int mpb_sectors(struct imsm_super *mpb) |
123 | { | |
124 | return sector_count(__le32_to_cpu(mpb->mpb_size)); | |
cdddbdbc DW |
125 | } |
126 | ||
127 | /* internal representation of IMSM metadata */ | |
128 | struct intel_super { | |
129 | union { | |
130 | struct imsm_super *mpb; | |
131 | void *buf; | |
132 | }; | |
c2c087e6 DW |
133 | int updates_pending; /* count of pending updates for mdmon */ |
134 | int creating_imsm; /* flag to indicate container creation */ | |
bf5a934a | 135 | int current_vol; /* index of raid device undergoing creation */ |
cdddbdbc DW |
136 | struct dl { |
137 | struct dl *next; | |
138 | int index; | |
139 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
140 | int major, minor; | |
141 | char *devname; | |
142 | int fd; | |
143 | } *disks; | |
144 | }; | |
145 | ||
c2c087e6 DW |
146 | struct extent { |
147 | unsigned long long start, size; | |
148 | }; | |
149 | ||
88758e9d DW |
150 | /* definition of messages passed to imsm_process_update */ |
151 | enum imsm_update_type { | |
152 | update_activate_spare, | |
8273f55e | 153 | update_create_array, |
88758e9d DW |
154 | }; |
155 | ||
156 | struct imsm_update_activate_spare { | |
157 | enum imsm_update_type type; | |
158 | int disk_idx; | |
159 | int slot; | |
160 | int array; | |
161 | struct imsm_update_activate_spare *next; | |
162 | }; | |
163 | ||
8273f55e DW |
164 | struct imsm_update_create_array { |
165 | enum imsm_update_type type; | |
166 | struct imsm_dev dev; | |
167 | int dev_idx; | |
168 | }; | |
169 | ||
0030e8d6 DW |
170 | static int imsm_env_devname_as_serial(void) |
171 | { | |
172 | char *val = getenv("IMSM_DEVNAME_AS_SERIAL"); | |
173 | ||
174 | if (val && atoi(val) == 1) | |
175 | return 1; | |
176 | ||
177 | return 0; | |
178 | } | |
179 | ||
180 | ||
cdddbdbc DW |
181 | static struct supertype *match_metadata_desc_imsm(char *arg) |
182 | { | |
183 | struct supertype *st; | |
184 | ||
185 | if (strcmp(arg, "imsm") != 0 && | |
186 | strcmp(arg, "default") != 0 | |
187 | ) | |
188 | return NULL; | |
189 | ||
190 | st = malloc(sizeof(*st)); | |
ef609477 | 191 | memset(st, 0, sizeof(*st)); |
cdddbdbc DW |
192 | st->ss = &super_imsm; |
193 | st->max_devs = IMSM_MAX_DEVICES; | |
194 | st->minor_version = 0; | |
195 | st->sb = NULL; | |
196 | return st; | |
197 | } | |
198 | ||
cdddbdbc DW |
199 | static __u8 *get_imsm_version(struct imsm_super *mpb) |
200 | { | |
201 | return &mpb->sig[MPB_SIG_LEN]; | |
202 | } | |
203 | ||
204 | static struct imsm_disk *get_imsm_disk(struct imsm_super *mpb, __u8 index) | |
205 | { | |
206 | if (index > mpb->num_disks - 1) | |
207 | return NULL; | |
208 | return &mpb->disk[index]; | |
209 | } | |
210 | ||
211 | static __u32 gen_imsm_checksum(struct imsm_super *mpb) | |
212 | { | |
213 | __u32 end = mpb->mpb_size / sizeof(end); | |
214 | __u32 *p = (__u32 *) mpb; | |
215 | __u32 sum = 0; | |
216 | ||
217 | while (end--) | |
218 | sum += __le32_to_cpu(*p++); | |
219 | ||
220 | return sum - __le32_to_cpu(mpb->check_sum); | |
221 | } | |
222 | ||
223 | static size_t sizeof_imsm_dev(struct imsm_dev *dev) | |
224 | { | |
225 | size_t size = sizeof(*dev); | |
226 | ||
227 | /* each map has disk_ord_tbl[num_members - 1] additional space */ | |
228 | size += sizeof(__u32) * (dev->vol.map[0].num_members - 1); | |
229 | ||
230 | /* migrating means an additional map */ | |
231 | if (dev->vol.migr_state) { | |
232 | size += sizeof(struct imsm_map); | |
233 | size += sizeof(__u32) * (dev->vol.map[1].num_members - 1); | |
234 | } | |
235 | ||
236 | return size; | |
237 | } | |
238 | ||
239 | static struct imsm_dev *get_imsm_dev(struct imsm_super *mpb, __u8 index) | |
240 | { | |
241 | int offset; | |
242 | int i; | |
243 | void *_mpb = mpb; | |
244 | ||
245 | if (index > mpb->num_raid_devs - 1) | |
246 | return NULL; | |
247 | ||
248 | /* devices start after all disks */ | |
249 | offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb; | |
250 | ||
251 | for (i = 0; i <= index; i++) | |
252 | if (i == index) | |
253 | return _mpb + offset; | |
254 | else | |
255 | offset += sizeof_imsm_dev(_mpb + offset); | |
256 | ||
257 | return NULL; | |
258 | } | |
259 | ||
260 | static __u32 get_imsm_disk_idx(struct imsm_map *map, int slot) | |
261 | { | |
262 | __u32 *ord_tbl = &map->disk_ord_tbl[slot]; | |
263 | ||
264 | /* top byte is 'special' */ | |
265 | return __le32_to_cpu(*ord_tbl & ~(0xff << 24)); | |
266 | } | |
267 | ||
cdddbdbc DW |
268 | static int get_imsm_raid_level(struct imsm_map *map) |
269 | { | |
270 | if (map->raid_level == 1) { | |
271 | if (map->num_members == 2) | |
272 | return 1; | |
273 | else | |
274 | return 10; | |
275 | } | |
276 | ||
277 | return map->raid_level; | |
278 | } | |
279 | ||
c2c087e6 DW |
280 | static int cmp_extent(const void *av, const void *bv) |
281 | { | |
282 | const struct extent *a = av; | |
283 | const struct extent *b = bv; | |
284 | if (a->start < b->start) | |
285 | return -1; | |
286 | if (a->start > b->start) | |
287 | return 1; | |
288 | return 0; | |
289 | } | |
290 | ||
291 | static struct extent *get_extents(struct intel_super *super, struct dl *dl) | |
292 | { | |
293 | /* find a list of used extents on the given physical device */ | |
294 | struct imsm_super *mpb = super->mpb; | |
295 | struct imsm_disk *disk; | |
296 | struct extent *rv, *e; | |
297 | int i, j; | |
298 | int memberships = 0; | |
299 | ||
300 | disk = get_imsm_disk(mpb, dl->index); | |
301 | if (!disk) | |
302 | return NULL; | |
303 | ||
304 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
305 | struct imsm_dev *dev = get_imsm_dev(mpb, i); | |
306 | struct imsm_map *map = dev->vol.map; | |
307 | ||
308 | for (j = 0; j < map->num_members; j++) { | |
309 | __u32 index = get_imsm_disk_idx(map, j); | |
310 | ||
311 | if (index == dl->index) | |
312 | memberships++; | |
313 | } | |
314 | } | |
315 | rv = malloc(sizeof(struct extent) * (memberships + 1)); | |
316 | if (!rv) | |
317 | return NULL; | |
318 | e = rv; | |
319 | ||
320 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
321 | struct imsm_dev *dev = get_imsm_dev(mpb, i); | |
322 | struct imsm_map *map = dev->vol.map; | |
323 | ||
324 | for (j = 0; j < map->num_members; j++) { | |
325 | __u32 index = get_imsm_disk_idx(map, j); | |
326 | ||
327 | if (index == dl->index) { | |
328 | e->start = __le32_to_cpu(map->pba_of_lba0); | |
329 | e->size = __le32_to_cpu(map->blocks_per_member); | |
330 | e++; | |
331 | } | |
332 | } | |
333 | } | |
334 | qsort(rv, memberships, sizeof(*rv), cmp_extent); | |
335 | ||
336 | e->start = __le32_to_cpu(disk->total_blocks) - | |
337 | (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS); | |
338 | e->size = 0; | |
339 | return rv; | |
340 | } | |
341 | ||
4f5bc454 | 342 | #ifndef MDASSEMBLE |
cdddbdbc DW |
343 | static void print_imsm_dev(struct imsm_dev *dev, int index) |
344 | { | |
345 | __u64 sz; | |
346 | int slot; | |
347 | struct imsm_map *map = dev->vol.map; | |
348 | ||
349 | printf("\n"); | |
350 | printf("[%s]:\n", dev->volume); | |
351 | printf(" RAID Level : %d\n", get_imsm_raid_level(map)); | |
352 | printf(" Members : %d\n", map->num_members); | |
353 | for (slot = 0; slot < map->num_members; slot++) | |
354 | if (index == get_imsm_disk_idx(map, slot)) | |
355 | break; | |
356 | if (slot < map->num_members) | |
357 | printf(" This Slot : %d\n", slot); | |
358 | else | |
359 | printf(" This Slot : ?\n"); | |
360 | sz = __le32_to_cpu(dev->size_high); | |
361 | sz <<= 32; | |
362 | sz += __le32_to_cpu(dev->size_low); | |
363 | printf(" Array Size : %llu%s\n", (unsigned long long)sz, | |
364 | human_size(sz * 512)); | |
365 | sz = __le32_to_cpu(map->blocks_per_member); | |
366 | printf(" Per Dev Size : %llu%s\n", (unsigned long long)sz, | |
367 | human_size(sz * 512)); | |
368 | printf(" Sector Offset : %u\n", | |
369 | __le32_to_cpu(map->pba_of_lba0)); | |
370 | printf(" Num Stripes : %u\n", | |
371 | __le32_to_cpu(map->num_data_stripes)); | |
372 | printf(" Chunk Size : %u KiB\n", | |
373 | __le16_to_cpu(map->blocks_per_strip) / 2); | |
374 | printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks)); | |
375 | printf(" Migrate State : %s\n", dev->vol.migr_state ? "migrating" : "idle"); | |
376 | printf(" Dirty State : %s\n", dev->vol.dirty ? "dirty" : "clean"); | |
377 | printf(" Map State : %s\n", map_state_str[map->map_state]); | |
378 | } | |
379 | ||
380 | static void print_imsm_disk(struct imsm_super *mpb, int index) | |
381 | { | |
382 | struct imsm_disk *disk = get_imsm_disk(mpb, index); | |
383 | char str[MAX_RAID_SERIAL_LEN]; | |
384 | __u32 s; | |
385 | __u64 sz; | |
386 | ||
e9d82038 DW |
387 | if (index < 0) |
388 | return; | |
389 | ||
cdddbdbc DW |
390 | printf("\n"); |
391 | snprintf(str, MAX_RAID_SERIAL_LEN, "%s", disk->serial); | |
392 | printf(" Disk%02d Serial : %s\n", index, str); | |
393 | s = __le32_to_cpu(disk->status); | |
394 | printf(" State :%s%s%s%s\n", s&SPARE_DISK ? " spare" : "", | |
395 | s&CONFIGURED_DISK ? " active" : "", | |
396 | s&FAILED_DISK ? " failed" : "", | |
397 | s&USABLE_DISK ? " usable" : ""); | |
398 | printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id)); | |
c2c087e6 DW |
399 | sz = __le32_to_cpu(disk->total_blocks) - |
400 | (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS * mpb->num_raid_devs); | |
cdddbdbc DW |
401 | printf(" Usable Size : %llu%s\n", (unsigned long long)sz, |
402 | human_size(sz * 512)); | |
403 | } | |
404 | ||
405 | static void examine_super_imsm(struct supertype *st, char *homehost) | |
406 | { | |
407 | struct intel_super *super = st->sb; | |
408 | struct imsm_super *mpb = super->mpb; | |
409 | char str[MAX_SIGNATURE_LENGTH]; | |
410 | int i; | |
411 | __u32 sum; | |
412 | ||
413 | snprintf(str, MPB_SIG_LEN, "%s", mpb->sig); | |
414 | printf(" Magic : %s\n", str); | |
415 | snprintf(str, strlen(MPB_VERSION_RAID0), "%s", get_imsm_version(mpb)); | |
416 | printf(" Version : %s\n", get_imsm_version(mpb)); | |
417 | printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num)); | |
418 | printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num)); | |
419 | sum = __le32_to_cpu(mpb->check_sum); | |
420 | printf(" Checksum : %08x %s\n", sum, | |
421 | gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect"); | |
87eb16df | 422 | printf(" MPB Sectors : %d\n", mpb_sectors(mpb)); |
cdddbdbc DW |
423 | printf(" Disks : %d\n", mpb->num_disks); |
424 | printf(" RAID Devices : %d\n", mpb->num_raid_devs); | |
425 | print_imsm_disk(mpb, super->disks->index); | |
426 | for (i = 0; i < mpb->num_raid_devs; i++) | |
427 | print_imsm_dev(get_imsm_dev(mpb, i), super->disks->index); | |
428 | for (i = 0; i < mpb->num_disks; i++) { | |
429 | if (i == super->disks->index) | |
430 | continue; | |
431 | print_imsm_disk(mpb, i); | |
432 | } | |
433 | } | |
434 | ||
435 | static void brief_examine_super_imsm(struct supertype *st) | |
436 | { | |
437 | struct intel_super *super = st->sb; | |
438 | struct imsm_super *mpb = super->mpb; | |
439 | ||
440 | printf("ARRAY /dev/imsm family=%08x metadata=external:imsm\n", | |
441 | __le32_to_cpu(mpb->family_num)); | |
442 | } | |
443 | ||
444 | static void detail_super_imsm(struct supertype *st, char *homehost) | |
445 | { | |
446 | printf("%s\n", __FUNCTION__); | |
447 | } | |
448 | ||
449 | static void brief_detail_super_imsm(struct supertype *st) | |
450 | { | |
451 | printf("%s\n", __FUNCTION__); | |
452 | } | |
453 | #endif | |
454 | ||
455 | static int match_home_imsm(struct supertype *st, char *homehost) | |
456 | { | |
457 | printf("%s\n", __FUNCTION__); | |
458 | ||
459 | return 0; | |
460 | } | |
461 | ||
462 | static void uuid_from_super_imsm(struct supertype *st, int uuid[4]) | |
463 | { | |
464 | printf("%s\n", __FUNCTION__); | |
465 | } | |
466 | ||
0d481d37 | 467 | #if 0 |
4f5bc454 DW |
468 | static void |
469 | get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p) | |
cdddbdbc | 470 | { |
cdddbdbc DW |
471 | __u8 *v = get_imsm_version(mpb); |
472 | __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH; | |
473 | char major[] = { 0, 0, 0 }; | |
474 | char minor[] = { 0 ,0, 0 }; | |
475 | char patch[] = { 0, 0, 0 }; | |
476 | char *ver_parse[] = { major, minor, patch }; | |
477 | int i, j; | |
478 | ||
479 | i = j = 0; | |
480 | while (*v != '\0' && v < end) { | |
481 | if (*v != '.' && j < 2) | |
482 | ver_parse[i][j++] = *v; | |
483 | else { | |
484 | i++; | |
485 | j = 0; | |
486 | } | |
487 | v++; | |
488 | } | |
489 | ||
4f5bc454 DW |
490 | *m = strtol(minor, NULL, 0); |
491 | *p = strtol(patch, NULL, 0); | |
492 | } | |
0d481d37 | 493 | #endif |
4f5bc454 | 494 | |
c2c087e6 DW |
495 | static int imsm_level_to_layout(int level) |
496 | { | |
497 | switch (level) { | |
498 | case 0: | |
499 | case 1: | |
500 | return 0; | |
501 | case 5: | |
502 | case 6: | |
503 | return ALGORITHM_LEFT_SYMMETRIC; | |
504 | case 10: | |
505 | return 0x102; //FIXME is this correct? | |
506 | } | |
507 | return -1; | |
508 | } | |
509 | ||
bf5a934a DW |
510 | static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info) |
511 | { | |
512 | struct intel_super *super = st->sb; | |
513 | struct imsm_super *mpb = super->mpb; | |
514 | struct imsm_dev *dev = get_imsm_dev(mpb, super->current_vol); | |
515 | struct imsm_map *map = &dev->vol.map[0]; | |
516 | ||
517 | info->container_member = super->current_vol; | |
518 | info->array.raid_disks = map->num_members; | |
519 | info->array.level = get_imsm_raid_level(map); | |
520 | info->array.layout = imsm_level_to_layout(info->array.level); | |
521 | info->array.md_minor = -1; | |
522 | info->array.ctime = 0; | |
523 | info->array.utime = 0; | |
524 | info->array.chunk_size = __le16_to_cpu(map->blocks_per_strip * 512); | |
525 | ||
526 | info->data_offset = __le32_to_cpu(map->pba_of_lba0); | |
527 | info->component_size = __le32_to_cpu(map->blocks_per_member); | |
528 | ||
529 | info->disk.major = 0; | |
530 | info->disk.minor = 0; | |
531 | ||
532 | sprintf(info->text_version, "/%s/%d", | |
533 | devnum2devname(st->container_dev), | |
534 | info->container_member); | |
535 | } | |
536 | ||
537 | ||
4f5bc454 DW |
538 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info) |
539 | { | |
540 | struct intel_super *super = st->sb; | |
541 | struct imsm_super *mpb = super->mpb; | |
542 | struct imsm_disk *disk; | |
543 | __u32 s; | |
4f5bc454 | 544 | |
bf5a934a DW |
545 | if (super->current_vol >= 0) { |
546 | getinfo_super_imsm_volume(st, info); | |
547 | return; | |
548 | } | |
cdddbdbc DW |
549 | info->array.raid_disks = mpb->num_disks; |
550 | info->array.level = LEVEL_CONTAINER; | |
551 | info->array.layout = 0; | |
552 | info->array.md_minor = -1; | |
c2c087e6 | 553 | info->array.ctime = 0; /* N/A for imsm */ |
cdddbdbc DW |
554 | info->array.utime = 0; |
555 | info->array.chunk_size = 0; | |
556 | ||
557 | info->disk.major = 0; | |
558 | info->disk.minor = 0; | |
cdddbdbc | 559 | info->disk.raid_disk = -1; |
c2c087e6 DW |
560 | info->reshape_active = 0; |
561 | strcpy(info->text_version, "imsm"); | |
562 | info->disk.number = -1; | |
563 | info->disk.state = 0; | |
564 | ||
4a04ec6c | 565 | if (super->disks) { |
bf5a934a | 566 | disk = get_imsm_disk(mpb, super->disks->index); |
e9d82038 DW |
567 | if (!disk) { |
568 | info->disk.number = -1; | |
569 | info->disk.raid_disk = -1; | |
570 | return; | |
571 | } | |
4a04ec6c DW |
572 | info->disk.number = super->disks->index; |
573 | info->disk.raid_disk = super->disks->index; | |
bf5a934a DW |
574 | info->data_offset = __le32_to_cpu(disk->total_blocks) - |
575 | (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS); | |
576 | info->component_size = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
4a04ec6c DW |
577 | s = __le32_to_cpu(disk->status); |
578 | info->disk.state = s & CONFIGURED_DISK ? (1 << MD_DISK_ACTIVE) : 0; | |
579 | info->disk.state |= s & FAILED_DISK ? (1 << MD_DISK_FAULTY) : 0; | |
580 | info->disk.state |= s & USABLE_DISK ? (1 << MD_DISK_SYNC) : 0; | |
cdddbdbc | 581 | } |
cdddbdbc DW |
582 | } |
583 | ||
cdddbdbc DW |
584 | static int update_super_imsm(struct supertype *st, struct mdinfo *info, |
585 | char *update, char *devname, int verbose, | |
586 | int uuid_set, char *homehost) | |
587 | { | |
f352c545 DW |
588 | /* FIXME */ |
589 | ||
590 | /* For 'assemble' and 'force' we need to return non-zero if any | |
591 | * change was made. For others, the return value is ignored. | |
592 | * Update options are: | |
593 | * force-one : This device looks a bit old but needs to be included, | |
594 | * update age info appropriately. | |
595 | * assemble: clear any 'faulty' flag to allow this device to | |
596 | * be assembled. | |
597 | * force-array: Array is degraded but being forced, mark it clean | |
598 | * if that will be needed to assemble it. | |
599 | * | |
600 | * newdev: not used ???? | |
601 | * grow: Array has gained a new device - this is currently for | |
602 | * linear only | |
603 | * resync: mark as dirty so a resync will happen. | |
604 | * name: update the name - preserving the homehost | |
605 | * | |
606 | * Following are not relevant for this imsm: | |
607 | * sparc2.2 : update from old dodgey metadata | |
608 | * super-minor: change the preferred_minor number | |
609 | * summaries: update redundant counters. | |
610 | * uuid: Change the uuid of the array to match watch is given | |
611 | * homehost: update the recorded homehost | |
612 | * _reshape_progress: record new reshape_progress position. | |
613 | */ | |
614 | int rv = 0; | |
615 | //struct intel_super *super = st->sb; | |
616 | //struct imsm_super *mpb = super->mpb; | |
617 | ||
618 | if (strcmp(update, "grow") == 0) { | |
619 | } | |
620 | if (strcmp(update, "resync") == 0) { | |
621 | /* dev->vol.dirty = 1; */ | |
622 | } | |
623 | ||
624 | /* IMSM has no concept of UUID or homehost */ | |
625 | ||
626 | return rv; | |
cdddbdbc DW |
627 | } |
628 | ||
c2c087e6 | 629 | static size_t disks_to_mpb_size(int disks) |
cdddbdbc | 630 | { |
c2c087e6 | 631 | size_t size; |
cdddbdbc | 632 | |
c2c087e6 DW |
633 | size = sizeof(struct imsm_super); |
634 | size += (disks - 1) * sizeof(struct imsm_disk); | |
635 | size += 2 * sizeof(struct imsm_dev); | |
636 | /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */ | |
637 | size += (4 - 2) * sizeof(struct imsm_map); | |
638 | /* 4 possible disk_ord_tbl's */ | |
639 | size += 4 * (disks - 1) * sizeof(__u32); | |
640 | ||
641 | return size; | |
642 | } | |
643 | ||
644 | static __u64 avail_size_imsm(struct supertype *st, __u64 devsize) | |
645 | { | |
646 | if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS)) | |
647 | return 0; | |
648 | ||
649 | return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS); | |
cdddbdbc DW |
650 | } |
651 | ||
652 | static int compare_super_imsm(struct supertype *st, struct supertype *tst) | |
653 | { | |
654 | /* | |
655 | * return: | |
656 | * 0 same, or first was empty, and second was copied | |
657 | * 1 second had wrong number | |
658 | * 2 wrong uuid | |
659 | * 3 wrong other info | |
660 | */ | |
661 | struct intel_super *first = st->sb; | |
662 | struct intel_super *sec = tst->sb; | |
663 | ||
664 | if (!first) { | |
665 | st->sb = tst->sb; | |
666 | tst->sb = NULL; | |
667 | return 0; | |
668 | } | |
669 | ||
670 | if (memcmp(first->mpb->sig, sec->mpb->sig, MAX_SIGNATURE_LENGTH) != 0) | |
671 | return 3; | |
672 | if (first->mpb->family_num != sec->mpb->family_num) | |
673 | return 3; | |
674 | if (first->mpb->mpb_size != sec->mpb->mpb_size) | |
675 | return 3; | |
676 | if (first->mpb->check_sum != sec->mpb->check_sum) | |
677 | return 3; | |
678 | ||
679 | return 0; | |
680 | } | |
681 | ||
0030e8d6 DW |
682 | static void fd2devname(int fd, char *name) |
683 | { | |
684 | struct stat st; | |
685 | char path[256]; | |
686 | char dname[100]; | |
687 | char *nm; | |
688 | int rv; | |
689 | ||
690 | name[0] = '\0'; | |
691 | if (fstat(fd, &st) != 0) | |
692 | return; | |
693 | sprintf(path, "/sys/dev/block/%d:%d", | |
694 | major(st.st_rdev), minor(st.st_rdev)); | |
695 | ||
696 | rv = readlink(path, dname, sizeof(dname)); | |
697 | if (rv <= 0) | |
698 | return; | |
699 | ||
700 | dname[rv] = '\0'; | |
701 | nm = strrchr(dname, '/'); | |
702 | nm++; | |
703 | snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm); | |
704 | } | |
705 | ||
706 | ||
cdddbdbc DW |
707 | extern int scsi_get_serial(int fd, void *buf, size_t buf_len); |
708 | ||
709 | static int imsm_read_serial(int fd, char *devname, | |
710 | __u8 serial[MAX_RAID_SERIAL_LEN]) | |
711 | { | |
712 | unsigned char scsi_serial[255]; | |
cdddbdbc | 713 | int sg_fd; |
cdddbdbc DW |
714 | int rv; |
715 | int rsp_len; | |
716 | int i, cnt; | |
717 | ||
718 | memset(scsi_serial, 0, sizeof(scsi_serial)); | |
cdddbdbc | 719 | |
90c8b707 | 720 | sg_fd = sysfs_disk_to_sg(fd); |
cdddbdbc DW |
721 | if (sg_fd < 0) { |
722 | if (devname) | |
723 | fprintf(stderr, | |
90c8b707 DW |
724 | Name ": Failed to open sg interface for %s: %s\n", |
725 | devname, strerror(errno)); | |
cdddbdbc DW |
726 | return 1; |
727 | } | |
728 | ||
0030e8d6 DW |
729 | if (imsm_env_devname_as_serial()) { |
730 | char name[MAX_RAID_SERIAL_LEN]; | |
731 | ||
732 | fd2devname(fd, name); | |
733 | strcpy((char *) serial, name); | |
734 | return 0; | |
735 | } | |
736 | ||
cdddbdbc DW |
737 | rv = scsi_get_serial(sg_fd, scsi_serial, sizeof(scsi_serial)); |
738 | close(sg_fd); | |
739 | ||
740 | if (rv != 0) { | |
741 | if (devname) | |
742 | fprintf(stderr, | |
743 | Name ": Failed to retrieve serial for %s\n", | |
744 | devname); | |
745 | return rv; | |
746 | } | |
747 | ||
748 | rsp_len = scsi_serial[3]; | |
749 | for (i = 0, cnt = 0; i < rsp_len; i++) { | |
750 | if (!isspace(scsi_serial[4 + i])) | |
751 | serial[cnt++] = scsi_serial[4 + i]; | |
752 | if (cnt == MAX_RAID_SERIAL_LEN) | |
753 | break; | |
754 | } | |
755 | ||
756 | serial[MAX_RAID_SERIAL_LEN - 1] = '\0'; | |
757 | ||
758 | return 0; | |
759 | } | |
760 | ||
761 | static int | |
762 | load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd) | |
763 | { | |
764 | struct imsm_super *mpb = super->mpb; | |
765 | struct dl *dl; | |
766 | struct stat stb; | |
767 | struct imsm_disk *disk; | |
768 | int rv; | |
769 | int i; | |
770 | ||
771 | dl = malloc(sizeof(*dl)); | |
772 | if (!dl) { | |
773 | if (devname) | |
774 | fprintf(stderr, | |
775 | Name ": failed to allocate disk buffer for %s\n", | |
776 | devname); | |
777 | return 2; | |
778 | } | |
779 | memset(dl, 0, sizeof(*dl)); | |
780 | ||
781 | fstat(fd, &stb); | |
782 | dl->major = major(stb.st_rdev); | |
783 | dl->minor = minor(stb.st_rdev); | |
784 | dl->next = super->disks; | |
785 | dl->fd = keep_fd ? fd : -1; | |
786 | dl->devname = devname ? strdup(devname) : NULL; | |
787 | dl->index = -1; | |
788 | super->disks = dl; | |
789 | rv = imsm_read_serial(fd, devname, dl->serial); | |
790 | ||
791 | if (rv != 0) | |
792 | return 2; | |
793 | ||
794 | /* look up this disk's index */ | |
795 | for (i = 0; i < mpb->num_disks; i++) { | |
796 | disk = get_imsm_disk(mpb, i); | |
797 | ||
798 | if (memcmp(disk->serial, dl->serial, MAX_RAID_SERIAL_LEN) == 0) | |
799 | break; | |
800 | } | |
801 | ||
e9d82038 DW |
802 | if (i > mpb->num_disks - 1) { |
803 | if (devname) | |
804 | fprintf(stderr, | |
805 | Name ": failed to match serial \'%s\' for %s\n", | |
806 | dl->serial, devname); | |
807 | return 0; | |
808 | } | |
cdddbdbc DW |
809 | |
810 | dl->index = i; | |
811 | ||
812 | return 0; | |
813 | } | |
814 | ||
815 | /* load_imsm_mpb - read matrix metadata | |
816 | * allocates super->mpb to be freed by free_super | |
817 | */ | |
818 | static int load_imsm_mpb(int fd, struct intel_super *super, char *devname) | |
819 | { | |
820 | unsigned long long dsize; | |
821 | size_t len, mpb_size; | |
822 | unsigned long long sectors; | |
823 | struct stat; | |
6416d527 | 824 | struct imsm_super *anchor; |
cdddbdbc DW |
825 | __u32 check_sum; |
826 | ||
cdddbdbc DW |
827 | get_dev_size(fd, NULL, &dsize); |
828 | ||
829 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) { | |
830 | if (devname) | |
831 | fprintf(stderr, | |
832 | Name ": Cannot seek to anchor block on %s: %s\n", | |
833 | devname, strerror(errno)); | |
834 | return 1; | |
835 | } | |
836 | ||
6416d527 | 837 | len = 512; |
ad97895e DW |
838 | if (posix_memalign((void**)&anchor, 512, len) != 0) { |
839 | if (devname) | |
840 | fprintf(stderr, | |
841 | Name ": Failed to allocate imsm anchor buffer" | |
842 | " on %s\n", devname); | |
843 | return 1; | |
844 | } | |
6416d527 | 845 | if (read(fd, anchor, len) != len) { |
cdddbdbc DW |
846 | if (devname) |
847 | fprintf(stderr, | |
848 | Name ": Cannot read anchor block on %s: %s\n", | |
849 | devname, strerror(errno)); | |
6416d527 | 850 | free(anchor); |
cdddbdbc DW |
851 | return 1; |
852 | } | |
853 | ||
6416d527 | 854 | if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) { |
cdddbdbc DW |
855 | if (devname) |
856 | fprintf(stderr, | |
857 | Name ": no IMSM anchor on %s\n", devname); | |
6416d527 | 858 | free(anchor); |
cdddbdbc DW |
859 | return 2; |
860 | } | |
861 | ||
6416d527 NB |
862 | mpb_size = __le32_to_cpu(anchor->mpb_size); |
863 | mpb_size = ROUND_UP(mpb_size, 512); | |
ef649044 | 864 | if (posix_memalign(&super->buf, 512, mpb_size) != 0) { |
cdddbdbc DW |
865 | if (devname) |
866 | fprintf(stderr, | |
867 | Name ": unable to allocate %zu byte mpb buffer\n", | |
868 | mpb_size); | |
6416d527 | 869 | free(anchor); |
cdddbdbc DW |
870 | return 2; |
871 | } | |
6416d527 | 872 | memcpy(super->buf, anchor, len); |
cdddbdbc | 873 | |
6416d527 NB |
874 | sectors = mpb_sectors(anchor) - 1; |
875 | free(anchor); | |
cdddbdbc DW |
876 | if (!sectors) |
877 | return load_imsm_disk(fd, super, devname, 0); | |
878 | ||
879 | /* read the extended mpb */ | |
880 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) { | |
881 | if (devname) | |
882 | fprintf(stderr, | |
883 | Name ": Cannot seek to extended mpb on %s: %s\n", | |
884 | devname, strerror(errno)); | |
885 | return 1; | |
886 | } | |
887 | ||
888 | len = mpb_size - 512; | |
889 | if (read(fd, super->buf + 512, len) != len) { | |
890 | if (devname) | |
891 | fprintf(stderr, | |
892 | Name ": Cannot read extended mpb on %s: %s\n", | |
893 | devname, strerror(errno)); | |
894 | return 2; | |
895 | } | |
896 | ||
897 | check_sum = gen_imsm_checksum(super->mpb); | |
898 | if (check_sum != __le32_to_cpu(super->mpb->check_sum)) { | |
899 | if (devname) | |
900 | fprintf(stderr, | |
901 | Name ": IMSM checksum %x != %x on %s\n", | |
902 | check_sum, __le32_to_cpu(super->mpb->check_sum), | |
903 | devname); | |
904 | return 2; | |
905 | } | |
906 | ||
907 | return load_imsm_disk(fd, super, devname, 0); | |
908 | } | |
909 | ||
cdddbdbc DW |
910 | static void free_imsm_disks(struct intel_super *super) |
911 | { | |
912 | while (super->disks) { | |
913 | struct dl *d = super->disks; | |
914 | ||
915 | super->disks = d->next; | |
916 | if (d->fd >= 0) | |
917 | close(d->fd); | |
918 | if (d->devname) | |
919 | free(d->devname); | |
920 | free(d); | |
921 | } | |
922 | } | |
923 | ||
924 | static void free_imsm(struct intel_super *super) | |
925 | { | |
926 | if (super->mpb) | |
927 | free(super->mpb); | |
928 | free_imsm_disks(super); | |
929 | free(super); | |
930 | } | |
931 | ||
932 | ||
933 | static void free_super_imsm(struct supertype *st) | |
934 | { | |
935 | struct intel_super *super = st->sb; | |
936 | ||
937 | if (!super) | |
938 | return; | |
939 | ||
940 | free_imsm(super); | |
941 | st->sb = NULL; | |
942 | } | |
943 | ||
c2c087e6 DW |
944 | static struct intel_super *alloc_super(int creating_imsm) |
945 | { | |
946 | struct intel_super *super = malloc(sizeof(*super)); | |
947 | ||
948 | if (super) { | |
949 | memset(super, 0, sizeof(*super)); | |
950 | super->creating_imsm = creating_imsm; | |
bf5a934a | 951 | super->current_vol = -1; |
c2c087e6 DW |
952 | } |
953 | ||
954 | return super; | |
955 | } | |
956 | ||
cdddbdbc DW |
957 | #ifndef MDASSEMBLE |
958 | static int load_super_imsm_all(struct supertype *st, int fd, void **sbp, | |
959 | char *devname, int keep_fd) | |
960 | { | |
961 | struct mdinfo *sra; | |
962 | struct intel_super *super; | |
963 | struct mdinfo *sd, *best = NULL; | |
964 | __u32 bestgen = 0; | |
965 | __u32 gen; | |
966 | char nm[20]; | |
967 | int dfd; | |
968 | int rv; | |
969 | ||
970 | /* check if this disk is a member of an active array */ | |
971 | sra = sysfs_read(fd, 0, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE); | |
972 | if (!sra) | |
973 | return 1; | |
974 | ||
975 | if (sra->array.major_version != -1 || | |
976 | sra->array.minor_version != -2 || | |
977 | strcmp(sra->text_version, "imsm") != 0) | |
978 | return 1; | |
979 | ||
c2c087e6 | 980 | super = alloc_super(0); |
cdddbdbc DW |
981 | if (!super) |
982 | return 1; | |
983 | ||
984 | /* find the most up to date disk in this array */ | |
985 | for (sd = sra->devs; sd; sd = sd->next) { | |
986 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); | |
987 | dfd = dev_open(nm, keep_fd ? O_RDWR : O_RDONLY); | |
988 | if (!dfd) { | |
989 | free_imsm(super); | |
990 | return 2; | |
991 | } | |
992 | rv = load_imsm_mpb(dfd, super, NULL); | |
993 | if (!keep_fd) | |
994 | close(dfd); | |
995 | if (rv == 0) { | |
996 | gen = __le32_to_cpu(super->mpb->generation_num); | |
997 | if (!best || gen > bestgen) { | |
998 | bestgen = gen; | |
999 | best = sd; | |
1000 | } | |
1001 | } else { | |
1002 | free_imsm(super); | |
1003 | return 2; | |
1004 | } | |
1005 | } | |
1006 | ||
1007 | if (!best) { | |
1008 | free_imsm(super); | |
1009 | return 1; | |
1010 | } | |
1011 | ||
1012 | /* load the most up to date anchor */ | |
1013 | sprintf(nm, "%d:%d", best->disk.major, best->disk.minor); | |
1014 | dfd = dev_open(nm, O_RDONLY); | |
1015 | if (!dfd) { | |
1016 | free_imsm(super); | |
1017 | return 1; | |
1018 | } | |
1019 | rv = load_imsm_mpb(dfd, super, NULL); | |
1020 | close(dfd); | |
1021 | if (rv != 0) { | |
1022 | free_imsm(super); | |
1023 | return 2; | |
1024 | } | |
1025 | ||
1026 | /* reset the disk list */ | |
1027 | free_imsm_disks(super); | |
1028 | ||
1029 | /* populate disk list */ | |
1030 | for (sd = sra->devs ; sd ; sd = sd->next) { | |
1031 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); | |
1032 | dfd = dev_open(nm, keep_fd? O_RDWR : O_RDONLY); | |
1033 | if (!dfd) { | |
1034 | free_imsm(super); | |
1035 | return 2; | |
1036 | } | |
1037 | load_imsm_disk(dfd, super, NULL, keep_fd); | |
1038 | if (!keep_fd) | |
1039 | close(dfd); | |
1040 | } | |
1041 | ||
f7e7067b | 1042 | if (st->subarray[0]) { |
bf5a934a DW |
1043 | if (atoi(st->subarray) <= super->mpb->num_raid_devs) |
1044 | super->current_vol = atoi(st->subarray); | |
1045 | else | |
1046 | return 1; | |
f7e7067b NB |
1047 | } |
1048 | ||
cdddbdbc DW |
1049 | *sbp = super; |
1050 | if (st->ss == NULL) { | |
bf5a934a | 1051 | st->ss = &super_imsm; |
cdddbdbc DW |
1052 | st->minor_version = 0; |
1053 | st->max_devs = IMSM_MAX_DEVICES; | |
f4d11639 | 1054 | st->container_dev = fd2devnum(fd); |
cdddbdbc DW |
1055 | } |
1056 | ||
1057 | return 0; | |
1058 | } | |
1059 | #endif | |
1060 | ||
1061 | static int load_super_imsm(struct supertype *st, int fd, char *devname) | |
1062 | { | |
1063 | struct intel_super *super; | |
1064 | int rv; | |
1065 | ||
1066 | #ifndef MDASSEMBLE | |
3dbccbcf | 1067 | if (load_super_imsm_all(st, fd, &st->sb, devname, 1) == 0) |
cdddbdbc DW |
1068 | return 0; |
1069 | #endif | |
f7e7067b NB |
1070 | if (st->subarray[0]) |
1071 | return 1; /* FIXME */ | |
cdddbdbc | 1072 | |
c2c087e6 | 1073 | super = alloc_super(0); |
cdddbdbc DW |
1074 | if (!super) { |
1075 | fprintf(stderr, | |
1076 | Name ": malloc of %zu failed.\n", | |
1077 | sizeof(*super)); | |
1078 | return 1; | |
1079 | } | |
1080 | ||
1081 | rv = load_imsm_mpb(fd, super, devname); | |
1082 | ||
1083 | if (rv) { | |
1084 | if (devname) | |
1085 | fprintf(stderr, | |
1086 | Name ": Failed to load all information " | |
1087 | "sections on %s\n", devname); | |
1088 | free_imsm(super); | |
1089 | return rv; | |
1090 | } | |
1091 | ||
1092 | st->sb = super; | |
1093 | if (st->ss == NULL) { | |
1094 | st->ss = &super_imsm; | |
1095 | st->minor_version = 0; | |
1096 | st->max_devs = IMSM_MAX_DEVICES; | |
1097 | } | |
1098 | ||
1099 | return 0; | |
1100 | } | |
1101 | ||
ef6ffade DW |
1102 | static __u16 info_to_blocks_per_strip(mdu_array_info_t *info) |
1103 | { | |
1104 | if (info->level == 1) | |
1105 | return 128; | |
1106 | return info->chunk_size >> 9; | |
1107 | } | |
1108 | ||
1109 | static __u32 info_to_num_data_stripes(mdu_array_info_t *info) | |
1110 | { | |
1111 | __u32 num_stripes; | |
1112 | ||
1113 | num_stripes = (info->size * 2) / info_to_blocks_per_strip(info); | |
1114 | if (info->level == 1) | |
1115 | num_stripes /= 2; | |
1116 | ||
1117 | return num_stripes; | |
1118 | } | |
1119 | ||
8b353278 DW |
1120 | static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info, |
1121 | unsigned long long size, char *name, | |
1122 | char *homehost, int *uuid) | |
cdddbdbc | 1123 | { |
c2c087e6 DW |
1124 | /* We are creating a volume inside a pre-existing container. |
1125 | * so st->sb is already set. | |
1126 | */ | |
1127 | struct intel_super *super = st->sb; | |
1128 | struct imsm_super *mpb = super->mpb; | |
1129 | struct imsm_dev *dev; | |
1130 | struct imsm_vol *vol; | |
1131 | struct imsm_map *map; | |
1132 | int idx = mpb->num_raid_devs; | |
1133 | int i; | |
1134 | unsigned long long array_blocks; | |
c2c087e6 | 1135 | __u32 offset = 0; |
2c092cad | 1136 | size_t size_old, size_new; |
cdddbdbc | 1137 | |
c2c087e6 DW |
1138 | if (mpb->num_raid_devs >= 2) { |
1139 | fprintf(stderr, Name": This imsm-container already has the " | |
1140 | "maximum of 2 volumes\n"); | |
1141 | return 0; | |
1142 | } | |
1143 | ||
2c092cad DW |
1144 | /* ensure the mpb is large enough for the new data */ |
1145 | size_old = __le32_to_cpu(mpb->mpb_size); | |
1146 | size_new = disks_to_mpb_size(info->nr_disks); | |
1147 | if (size_new > size_old) { | |
1148 | void *mpb_new; | |
1149 | size_t size_round = ROUND_UP(size_new, 512); | |
1150 | ||
1151 | if (posix_memalign(&mpb_new, 512, size_round) != 0) { | |
1152 | fprintf(stderr, Name": could not allocate new mpb\n"); | |
1153 | return 0; | |
1154 | } | |
1155 | memcpy(mpb_new, mpb, size_old); | |
1156 | free(mpb); | |
1157 | mpb = mpb_new; | |
1158 | super->mpb = mpb_new; | |
1159 | mpb->mpb_size = __cpu_to_le32(size_new); | |
1160 | memset(mpb_new + size_old, 0, size_round - size_old); | |
1161 | } | |
bf5a934a DW |
1162 | super->current_vol = idx; |
1163 | sprintf(st->subarray, "%d", idx); | |
c2c087e6 DW |
1164 | mpb->num_raid_devs++; |
1165 | dev = get_imsm_dev(mpb, idx); | |
1166 | strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN); | |
1167 | array_blocks = calc_array_size(info->level, info->raid_disks, | |
1168 | info->layout, info->chunk_size, | |
1169 | info->size*2); | |
1170 | dev->size_low = __cpu_to_le32((__u32) array_blocks); | |
1171 | dev->size_high = __cpu_to_le32((__u32) (array_blocks >> 32)); | |
1172 | dev->status = __cpu_to_le32(0); | |
1173 | dev->reserved_blocks = __cpu_to_le32(0); | |
1174 | vol = &dev->vol; | |
1175 | vol->migr_state = 0; | |
1176 | vol->migr_type = 0; | |
1177 | vol->dirty = 0; | |
1178 | for (i = 0; i < idx; i++) { | |
1179 | struct imsm_dev *prev = get_imsm_dev(mpb, i); | |
1180 | struct imsm_map *pmap = &prev->vol.map[0]; | |
1181 | ||
1182 | offset += __le32_to_cpu(pmap->blocks_per_member); | |
1183 | offset += IMSM_RESERVED_SECTORS; | |
1184 | } | |
1185 | map = &vol->map[0]; | |
1186 | map->pba_of_lba0 = __cpu_to_le32(offset); | |
ef6ffade DW |
1187 | map->blocks_per_member = __cpu_to_le32(info->size * 2); |
1188 | map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info)); | |
1189 | map->num_data_stripes = __cpu_to_le32(info_to_num_data_stripes(info)); | |
c2c087e6 DW |
1190 | map->map_state = info->level ? IMSM_T_STATE_UNINITIALIZED : |
1191 | IMSM_T_STATE_NORMAL; | |
ef6ffade DW |
1192 | |
1193 | if (info->level == 1 && info->raid_disks > 2) { | |
1194 | fprintf(stderr, Name": imsm does not support more than 2 disks" | |
1195 | "in a raid1 volume\n"); | |
1196 | return 0; | |
1197 | } | |
c2c087e6 DW |
1198 | if (info->level == 10) |
1199 | map->raid_level = 1; | |
1200 | else | |
1201 | map->raid_level = info->level; | |
ef6ffade | 1202 | |
c2c087e6 DW |
1203 | map->num_members = info->raid_disks; |
1204 | for (i = 0; i < map->num_members; i++) { | |
1205 | /* initialized in add_to_super */ | |
1206 | map->disk_ord_tbl[i] = __cpu_to_le32(0); | |
1207 | } | |
1208 | ||
1209 | return 1; | |
cdddbdbc DW |
1210 | } |
1211 | ||
bf5a934a DW |
1212 | static int init_super_imsm(struct supertype *st, mdu_array_info_t *info, |
1213 | unsigned long long size, char *name, | |
1214 | char *homehost, int *uuid) | |
1215 | { | |
1216 | /* This is primarily called by Create when creating a new array. | |
1217 | * We will then get add_to_super called for each component, and then | |
1218 | * write_init_super called to write it out to each device. | |
1219 | * For IMSM, Create can create on fresh devices or on a pre-existing | |
1220 | * array. | |
1221 | * To create on a pre-existing array a different method will be called. | |
1222 | * This one is just for fresh drives. | |
1223 | */ | |
1224 | struct intel_super *super; | |
1225 | struct imsm_super *mpb; | |
1226 | size_t mpb_size; | |
1227 | ||
1228 | if (!info) { | |
1229 | st->sb = NULL; | |
1230 | return 0; | |
1231 | } | |
1232 | if (st->sb) | |
1233 | return init_super_imsm_volume(st, info, size, name, homehost, | |
1234 | uuid); | |
1235 | ||
1236 | super = alloc_super(1); | |
1237 | if (!super) | |
1238 | return 0; | |
1239 | mpb_size = disks_to_mpb_size(info->nr_disks); | |
ef649044 | 1240 | if (posix_memalign(&super->buf, 512, mpb_size) != 0) { |
bf5a934a DW |
1241 | free(super); |
1242 | return 0; | |
1243 | } | |
ef649044 | 1244 | mpb = super->buf; |
bf5a934a DW |
1245 | memset(mpb, 0, mpb_size); |
1246 | ||
1247 | memcpy(mpb->sig, MPB_SIGNATURE, strlen(MPB_SIGNATURE)); | |
1248 | memcpy(mpb->sig + strlen(MPB_SIGNATURE), MPB_VERSION_RAID5, | |
1249 | strlen(MPB_VERSION_RAID5)); | |
1250 | mpb->mpb_size = mpb_size; | |
1251 | ||
bf5a934a DW |
1252 | st->sb = super; |
1253 | return 1; | |
1254 | } | |
1255 | ||
1256 | static void add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk, | |
1257 | int fd, char *devname) | |
1258 | { | |
1259 | struct intel_super *super = st->sb; | |
1260 | struct imsm_super *mpb = super->mpb; | |
1261 | struct dl *dl; | |
1262 | struct imsm_dev *dev; | |
1263 | struct imsm_map *map; | |
1264 | struct imsm_disk *disk; | |
1265 | __u32 status; | |
1266 | ||
1267 | dev = get_imsm_dev(mpb, super->current_vol); | |
1268 | map = &dev->vol.map[0]; | |
1269 | ||
1270 | for (dl = super->disks; dl ; dl = dl->next) | |
1271 | if (dl->major == dk->major && | |
1272 | dl->minor == dk->minor) | |
1273 | break; | |
1274 | if (!dl || ! (dk->state & (1<<MD_DISK_SYNC))) | |
1275 | return; | |
1276 | ||
1277 | map->disk_ord_tbl[dk->number] = __cpu_to_le32(dl->index); | |
1278 | ||
1279 | disk = get_imsm_disk(mpb, dl->index); | |
1280 | status = CONFIGURED_DISK | USABLE_DISK; | |
1281 | disk->status = __cpu_to_le32(status); | |
1282 | } | |
1283 | ||
c2c087e6 | 1284 | static void add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk, |
cdddbdbc DW |
1285 | int fd, char *devname) |
1286 | { | |
c2c087e6 DW |
1287 | struct intel_super *super = st->sb; |
1288 | struct imsm_super *mpb = super->mpb; | |
1289 | struct imsm_disk *disk; | |
1290 | struct dl *dd; | |
1291 | unsigned long long size; | |
1292 | __u32 status, id; | |
1293 | int rv; | |
1294 | struct stat stb; | |
1295 | ||
bf5a934a DW |
1296 | if (super->current_vol >= 0) { |
1297 | add_to_super_imsm_volume(st, dk, fd, devname); | |
1298 | return; | |
1299 | } | |
1300 | ||
c2c087e6 DW |
1301 | fstat(fd, &stb); |
1302 | dd = malloc(sizeof(*dd)); | |
1303 | if (!dd) { | |
1304 | fprintf(stderr, | |
1305 | Name ": malloc failed %s:%d.\n", __func__, __LINE__); | |
1306 | abort(); | |
1307 | } | |
1308 | memset(dd, 0, sizeof(*dd)); | |
1309 | dd->major = major(stb.st_rdev); | |
1310 | dd->minor = minor(stb.st_rdev); | |
1311 | dd->index = dk->number; | |
1312 | dd->devname = devname ? strdup(devname) : NULL; | |
1313 | dd->next = super->disks; | |
1314 | dd->fd = fd; | |
1315 | rv = imsm_read_serial(fd, devname, dd->serial); | |
1316 | if (rv) { | |
1317 | fprintf(stderr, | |
0030e8d6 DW |
1318 | Name ": failed to retrieve scsi serial, aborting\n"); |
1319 | abort(); | |
c2c087e6 DW |
1320 | } |
1321 | ||
1322 | if (mpb->num_disks <= dk->number) | |
1323 | mpb->num_disks = dk->number + 1; | |
1324 | ||
1325 | disk = get_imsm_disk(mpb, dk->number); | |
1326 | get_dev_size(fd, NULL, &size); | |
1327 | size /= 512; | |
1328 | status = USABLE_DISK | SPARE_DISK; | |
1329 | strcpy((char *) disk->serial, (char *) dd->serial); | |
1330 | disk->total_blocks = __cpu_to_le32(size); | |
1331 | disk->status = __cpu_to_le32(status); | |
1332 | if (sysfs_disk_to_scsi_id(fd, &id) == 0) | |
1333 | disk->scsi_id = __cpu_to_le32(id); | |
1334 | else | |
1335 | disk->scsi_id = __cpu_to_le32(0); | |
1336 | ||
1337 | /* update the family number if we are creating a container */ | |
1338 | if (super->creating_imsm) | |
1339 | mpb->family_num = __cpu_to_le32(gen_imsm_checksum(mpb)); | |
1340 | ||
1341 | super->disks = dd; | |
cdddbdbc DW |
1342 | } |
1343 | ||
c2c087e6 DW |
1344 | static int store_imsm_mpb(int fd, struct intel_super *super); |
1345 | ||
1346 | static int write_super_imsm(struct intel_super *super, int doclose) | |
cdddbdbc | 1347 | { |
c2c087e6 DW |
1348 | struct imsm_super *mpb = super->mpb; |
1349 | struct dl *d; | |
1350 | __u32 generation; | |
1351 | __u32 sum; | |
cdddbdbc | 1352 | |
c2c087e6 DW |
1353 | /* 'generation' is incremented everytime the metadata is written */ |
1354 | generation = __le32_to_cpu(mpb->generation_num); | |
1355 | generation++; | |
1356 | mpb->generation_num = __cpu_to_le32(generation); | |
1357 | ||
1358 | /* recalculate checksum */ | |
1359 | sum = gen_imsm_checksum(mpb); | |
1360 | mpb->check_sum = __cpu_to_le32(sum); | |
1361 | ||
1362 | for (d = super->disks; d ; d = d->next) { | |
1363 | if (store_imsm_mpb(d->fd, super)) { | |
1364 | fprintf(stderr, "%s: failed for device %d:%d %s\n", | |
1365 | __func__, d->major, d->minor, strerror(errno)); | |
1366 | return 0; | |
1367 | } | |
1368 | if (doclose) { | |
1369 | close(d->fd); | |
1370 | d->fd = -1; | |
1371 | } | |
1372 | } | |
1373 | ||
1374 | return 1; | |
1375 | } | |
1376 | ||
1377 | static int write_init_super_imsm(struct supertype *st) | |
1378 | { | |
8273f55e DW |
1379 | if (st->update_tail) { |
1380 | /* queue the recently created array as a metadata update */ | |
1381 | size_t len; | |
1382 | struct imsm_update_create_array *u; | |
1383 | struct intel_super *super = st->sb; | |
1384 | struct imsm_super *mpb = super->mpb; | |
1385 | struct imsm_dev *dev; | |
1386 | struct imsm_map *map; | |
1387 | struct dl *d; | |
1388 | ||
1389 | if (super->current_vol < 0 || | |
1390 | !(dev = get_imsm_dev(mpb, super->current_vol))) { | |
1391 | fprintf(stderr, "%s: could not determine sub-array\n", | |
1392 | __func__); | |
1393 | return 1; | |
1394 | } | |
1395 | ||
1396 | ||
1397 | map = &dev->vol.map[0]; | |
1398 | len = sizeof(*u) + sizeof(__u32) * (map->num_members - 1); | |
1399 | u = malloc(len); | |
1400 | if (!u) { | |
1401 | fprintf(stderr, "%s: failed to allocate update buffer\n", | |
1402 | __func__); | |
1403 | return 1; | |
1404 | } | |
1405 | ||
1406 | u->type = update_create_array; | |
1407 | u->dev_idx = super->current_vol; | |
1408 | memcpy(&u->dev, dev, sizeof(*dev)); | |
1409 | memcpy(u->dev.vol.map[0].disk_ord_tbl, map->disk_ord_tbl, | |
1410 | sizeof(__u32) * map->num_members); | |
1411 | append_metadata_update(st, u, len); | |
1412 | ||
1413 | for (d = super->disks; d ; d = d->next) { | |
1414 | close(d->fd); | |
1415 | d->fd = -1; | |
1416 | } | |
1417 | ||
1418 | return 0; | |
1419 | } else | |
1420 | return write_super_imsm(st->sb, 1); | |
cdddbdbc DW |
1421 | } |
1422 | ||
1423 | static int store_zero_imsm(struct supertype *st, int fd) | |
1424 | { | |
551c80c1 | 1425 | unsigned long long dsize; |
6416d527 | 1426 | void *buf; |
551c80c1 DW |
1427 | |
1428 | get_dev_size(fd, NULL, &dsize); | |
1429 | ||
1430 | /* first block is stored on second to last sector of the disk */ | |
1431 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) | |
1432 | return 1; | |
1433 | ||
ad97895e DW |
1434 | if (posix_memalign(&buf, 512, 512) != 0) |
1435 | return 1; | |
1436 | ||
eb7ea463 DW |
1437 | memset(buf, 0, 512); |
1438 | if (write(fd, buf, 512) != 512) | |
551c80c1 | 1439 | return 1; |
cdddbdbc DW |
1440 | return 0; |
1441 | } | |
1442 | ||
cdddbdbc DW |
1443 | static int validate_geometry_imsm_container(struct supertype *st, int level, |
1444 | int layout, int raiddisks, int chunk, | |
c2c087e6 | 1445 | unsigned long long size, char *dev, |
2c514b71 NB |
1446 | unsigned long long *freesize, |
1447 | int verbose) | |
cdddbdbc | 1448 | { |
c2c087e6 DW |
1449 | int fd; |
1450 | unsigned long long ldsize; | |
cdddbdbc | 1451 | |
c2c087e6 DW |
1452 | if (level != LEVEL_CONTAINER) |
1453 | return 0; | |
1454 | if (!dev) | |
1455 | return 1; | |
1456 | ||
1457 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
1458 | if (fd < 0) { | |
2c514b71 NB |
1459 | if (verbose) |
1460 | fprintf(stderr, Name ": imsm: Cannot open %s: %s\n", | |
1461 | dev, strerror(errno)); | |
c2c087e6 DW |
1462 | return 0; |
1463 | } | |
1464 | if (!get_dev_size(fd, dev, &ldsize)) { | |
1465 | close(fd); | |
1466 | return 0; | |
1467 | } | |
1468 | close(fd); | |
1469 | ||
1470 | *freesize = avail_size_imsm(st, ldsize >> 9); | |
1471 | ||
1472 | return 1; | |
cdddbdbc DW |
1473 | } |
1474 | ||
c2c087e6 DW |
1475 | /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd |
1476 | * FIX ME add ahci details | |
1477 | */ | |
8b353278 DW |
1478 | static int validate_geometry_imsm_volume(struct supertype *st, int level, |
1479 | int layout, int raiddisks, int chunk, | |
c2c087e6 | 1480 | unsigned long long size, char *dev, |
2c514b71 NB |
1481 | unsigned long long *freesize, |
1482 | int verbose) | |
cdddbdbc | 1483 | { |
c2c087e6 DW |
1484 | struct stat stb; |
1485 | struct intel_super *super = st->sb; | |
1486 | struct dl *dl; | |
1487 | unsigned long long pos = 0; | |
1488 | unsigned long long maxsize; | |
1489 | struct extent *e; | |
1490 | int i; | |
cdddbdbc | 1491 | |
c2c087e6 DW |
1492 | if (level == LEVEL_CONTAINER) |
1493 | return 0; | |
1494 | ||
1495 | if (level == 1 && raiddisks > 2) { | |
2c514b71 NB |
1496 | if (verbose) |
1497 | fprintf(stderr, Name ": imsm does not support more " | |
1498 | "than 2 in a raid1 configuration\n"); | |
c2c087e6 DW |
1499 | return 0; |
1500 | } | |
1501 | ||
1502 | /* We must have the container info already read in. */ | |
1503 | if (!super) | |
1504 | return 0; | |
1505 | ||
1506 | if (!dev) { | |
1507 | /* General test: make sure there is space for | |
2da8544a DW |
1508 | * 'raiddisks' device extents of size 'size' at a given |
1509 | * offset | |
c2c087e6 DW |
1510 | */ |
1511 | unsigned long long minsize = size*2 /* convert to blocks */; | |
2da8544a | 1512 | unsigned long long start_offset = ~0ULL; |
c2c087e6 DW |
1513 | int dcnt = 0; |
1514 | if (minsize == 0) | |
1515 | minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
1516 | for (dl = super->disks; dl ; dl = dl->next) { | |
1517 | int found = 0; | |
1518 | ||
bf5a934a | 1519 | pos = 0; |
c2c087e6 DW |
1520 | i = 0; |
1521 | e = get_extents(super, dl); | |
1522 | if (!e) continue; | |
1523 | do { | |
1524 | unsigned long long esize; | |
1525 | esize = e[i].start - pos; | |
1526 | if (esize >= minsize) | |
1527 | found = 1; | |
2da8544a DW |
1528 | if (found && start_offset == ~0ULL) { |
1529 | start_offset = pos; | |
1530 | break; | |
1531 | } else if (found && pos != start_offset) { | |
1532 | found = 0; | |
1533 | break; | |
1534 | } | |
c2c087e6 DW |
1535 | pos = e[i].start + e[i].size; |
1536 | i++; | |
1537 | } while (e[i-1].size); | |
1538 | if (found) | |
1539 | dcnt++; | |
1540 | free(e); | |
1541 | } | |
1542 | if (dcnt < raiddisks) { | |
2c514b71 NB |
1543 | if (verbose) |
1544 | fprintf(stderr, Name ": imsm: Not enough " | |
1545 | "devices with space for this array " | |
1546 | "(%d < %d)\n", | |
1547 | dcnt, raiddisks); | |
c2c087e6 DW |
1548 | return 0; |
1549 | } | |
1550 | return 1; | |
1551 | } | |
1552 | /* This device must be a member of the set */ | |
1553 | if (stat(dev, &stb) < 0) | |
1554 | return 0; | |
1555 | if ((S_IFMT & stb.st_mode) != S_IFBLK) | |
1556 | return 0; | |
1557 | for (dl = super->disks ; dl ; dl = dl->next) { | |
1558 | if (dl->major == major(stb.st_rdev) && | |
1559 | dl->minor == minor(stb.st_rdev)) | |
1560 | break; | |
1561 | } | |
1562 | if (!dl) { | |
2c514b71 NB |
1563 | if (verbose) |
1564 | fprintf(stderr, Name ": %s is not in the " | |
1565 | "same imsm set\n", dev); | |
c2c087e6 DW |
1566 | return 0; |
1567 | } | |
1568 | e = get_extents(super, dl); | |
1569 | maxsize = 0; | |
1570 | i = 0; | |
1571 | if (e) do { | |
1572 | unsigned long long esize; | |
1573 | esize = e[i].start - pos; | |
1574 | if (esize >= maxsize) | |
1575 | maxsize = esize; | |
1576 | pos = e[i].start + e[i].size; | |
1577 | i++; | |
1578 | } while (e[i-1].size); | |
1579 | *freesize = maxsize; | |
1580 | ||
1581 | return 1; | |
cdddbdbc DW |
1582 | } |
1583 | ||
bf5a934a DW |
1584 | static int validate_geometry_imsm(struct supertype *st, int level, int layout, |
1585 | int raiddisks, int chunk, unsigned long long size, | |
1586 | char *dev, unsigned long long *freesize, | |
1587 | int verbose) | |
1588 | { | |
1589 | int fd, cfd; | |
1590 | struct mdinfo *sra; | |
1591 | ||
1592 | /* if given unused devices create a container | |
1593 | * if given given devices in a container create a member volume | |
1594 | */ | |
1595 | if (level == LEVEL_CONTAINER) { | |
1596 | /* Must be a fresh device to add to a container */ | |
1597 | return validate_geometry_imsm_container(st, level, layout, | |
1598 | raiddisks, chunk, size, | |
1599 | dev, freesize, | |
1600 | verbose); | |
1601 | } | |
1602 | ||
1603 | if (st->sb) { | |
1604 | /* creating in a given container */ | |
1605 | return validate_geometry_imsm_volume(st, level, layout, | |
1606 | raiddisks, chunk, size, | |
1607 | dev, freesize, verbose); | |
1608 | } | |
1609 | ||
1610 | /* limit creation to the following levels */ | |
1611 | if (!dev) | |
1612 | switch (level) { | |
1613 | case 0: | |
1614 | case 1: | |
1615 | case 10: | |
1616 | case 5: | |
1617 | break; | |
1618 | default: | |
1619 | return 1; | |
1620 | } | |
1621 | ||
1622 | /* This device needs to be a device in an 'imsm' container */ | |
1623 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
1624 | if (fd >= 0) { | |
1625 | if (verbose) | |
1626 | fprintf(stderr, | |
1627 | Name ": Cannot create this array on device %s\n", | |
1628 | dev); | |
1629 | close(fd); | |
1630 | return 0; | |
1631 | } | |
1632 | if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) { | |
1633 | if (verbose) | |
1634 | fprintf(stderr, Name ": Cannot open %s: %s\n", | |
1635 | dev, strerror(errno)); | |
1636 | return 0; | |
1637 | } | |
1638 | /* Well, it is in use by someone, maybe an 'imsm' container. */ | |
1639 | cfd = open_container(fd); | |
1640 | if (cfd < 0) { | |
1641 | close(fd); | |
1642 | if (verbose) | |
1643 | fprintf(stderr, Name ": Cannot use %s: It is busy\n", | |
1644 | dev); | |
1645 | return 0; | |
1646 | } | |
1647 | sra = sysfs_read(cfd, 0, GET_VERSION); | |
1648 | close(fd); | |
1649 | if (sra && sra->array.major_version == -1 && | |
1650 | strcmp(sra->text_version, "imsm") == 0) { | |
1651 | /* This is a member of a imsm container. Load the container | |
1652 | * and try to create a volume | |
1653 | */ | |
1654 | struct intel_super *super; | |
1655 | ||
1656 | if (load_super_imsm_all(st, cfd, (void **) &super, NULL, 1) == 0) { | |
1657 | st->sb = super; | |
1658 | st->container_dev = fd2devnum(cfd); | |
1659 | close(cfd); | |
1660 | return validate_geometry_imsm_volume(st, level, layout, | |
1661 | raiddisks, chunk, | |
1662 | size, dev, | |
1663 | freesize, verbose); | |
1664 | } | |
1665 | close(cfd); | |
1666 | } else /* may belong to another container */ | |
1667 | return 0; | |
1668 | ||
1669 | return 1; | |
1670 | } | |
1671 | ||
cdddbdbc DW |
1672 | static struct mdinfo *container_content_imsm(struct supertype *st) |
1673 | { | |
4f5bc454 DW |
1674 | /* Given a container loaded by load_super_imsm_all, |
1675 | * extract information about all the arrays into | |
1676 | * an mdinfo tree. | |
1677 | * | |
1678 | * For each imsm_dev create an mdinfo, fill it in, | |
1679 | * then look for matching devices in super->disks | |
1680 | * and create appropriate device mdinfo. | |
1681 | */ | |
1682 | struct intel_super *super = st->sb; | |
1683 | struct imsm_super *mpb = super->mpb; | |
1684 | struct mdinfo *rest = NULL; | |
1685 | int i; | |
cdddbdbc | 1686 | |
4f5bc454 DW |
1687 | for (i = 0; i < mpb->num_raid_devs; i++) { |
1688 | struct imsm_dev *dev = get_imsm_dev(mpb, i); | |
1689 | struct imsm_vol *vol = &dev->vol; | |
1690 | struct imsm_map *map = vol->map; | |
1691 | struct mdinfo *this; | |
1692 | __u64 sz; | |
1693 | int slot; | |
1694 | ||
1695 | this = malloc(sizeof(*this)); | |
1696 | memset(this, 0, sizeof(*this)); | |
1697 | this->next = rest; | |
1698 | rest = this; | |
1699 | ||
4f5bc454 DW |
1700 | this->array.level = get_imsm_raid_level(map); |
1701 | this->array.raid_disks = map->num_members; | |
c2c087e6 | 1702 | this->array.layout = imsm_level_to_layout(this->array.level); |
4f5bc454 DW |
1703 | this->array.md_minor = -1; |
1704 | this->array.ctime = 0; | |
1705 | this->array.utime = 0; | |
1706 | this->array.chunk_size = __le16_to_cpu(map->blocks_per_strip) << 9; | |
1707 | this->array.state = !vol->dirty; | |
1708 | this->container_member = i; | |
0fd5c350 DW |
1709 | if (map->map_state == IMSM_T_STATE_UNINITIALIZED || dev->vol.dirty) |
1710 | this->resync_start = 0; | |
1711 | else | |
1712 | this->resync_start = ~0ULL; | |
1713 | ||
4f5bc454 DW |
1714 | strncpy(this->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN); |
1715 | this->name[MAX_RAID_SERIAL_LEN] = 0; | |
1716 | ||
159c3a1a NB |
1717 | sprintf(this->text_version, "/%s/%d", |
1718 | devnum2devname(st->container_dev), | |
1719 | this->container_member); | |
1720 | ||
4f5bc454 DW |
1721 | memset(this->uuid, 0, sizeof(this->uuid)); |
1722 | ||
1723 | sz = __le32_to_cpu(dev->size_high); | |
1724 | sz <<= 32; | |
1725 | sz += __le32_to_cpu(dev->size_low); | |
1726 | this->component_size = sz; | |
1727 | this->array.size = this->component_size / 2; | |
1728 | ||
1729 | for (slot = 0 ; slot < map->num_members; slot++) { | |
1730 | struct imsm_disk *disk; | |
1731 | struct mdinfo *info_d; | |
1732 | struct dl *d; | |
1733 | int idx; | |
1734 | __u32 s; | |
1735 | ||
1736 | idx = __le32_to_cpu(map->disk_ord_tbl[slot] & ~(0xff << 24)); | |
1737 | for (d = super->disks; d ; d = d->next) | |
1738 | if (d->index == idx) | |
1739 | break; | |
1740 | ||
1741 | if (d == NULL) | |
1742 | break; /* shouldn't this be continue ?? */ | |
1743 | ||
1744 | info_d = malloc(sizeof(*info_d)); | |
1745 | if (!info_d) | |
1746 | break; /* ditto ?? */ | |
1747 | memset(info_d, 0, sizeof(*info_d)); | |
1748 | info_d->next = this->devs; | |
1749 | this->devs = info_d; | |
1750 | ||
1751 | disk = get_imsm_disk(mpb, idx); | |
1752 | s = __le32_to_cpu(disk->status); | |
1753 | ||
1754 | info_d->disk.number = d->index; | |
1755 | info_d->disk.major = d->major; | |
1756 | info_d->disk.minor = d->minor; | |
1757 | info_d->disk.raid_disk = slot; | |
1758 | info_d->disk.state = s & CONFIGURED_DISK ? (1 << MD_DISK_ACTIVE) : 0; | |
1759 | info_d->disk.state |= s & FAILED_DISK ? (1 << MD_DISK_FAULTY) : 0; | |
1760 | info_d->disk.state |= s & USABLE_DISK ? (1 << MD_DISK_SYNC) : 0; | |
1761 | ||
1762 | this->array.working_disks++; | |
1763 | ||
1764 | info_d->events = __le32_to_cpu(mpb->generation_num); | |
1765 | info_d->data_offset = __le32_to_cpu(map->pba_of_lba0); | |
1766 | info_d->component_size = __le32_to_cpu(map->blocks_per_member); | |
1767 | if (d->devname) | |
1768 | strcpy(info_d->name, d->devname); | |
1769 | } | |
1770 | } | |
1771 | ||
1772 | return rest; | |
cdddbdbc DW |
1773 | } |
1774 | ||
845dea95 | 1775 | |
cba0191b NB |
1776 | static int imsm_open_new(struct supertype *c, struct active_array *a, |
1777 | char *inst) | |
845dea95 | 1778 | { |
0372d5a2 DW |
1779 | struct intel_super *super = c->sb; |
1780 | struct imsm_super *mpb = super->mpb; | |
1781 | ||
1782 | if (atoi(inst) + 1 > mpb->num_raid_devs) { | |
1783 | fprintf(stderr, "%s: subarry index %d, out of range\n", | |
1784 | __func__, atoi(inst)); | |
1785 | return -ENODEV; | |
1786 | } | |
1787 | ||
4e6e574a | 1788 | dprintf("imsm: open_new %s\n", inst); |
cba0191b | 1789 | a->info.container_member = atoi(inst); |
845dea95 NB |
1790 | return 0; |
1791 | } | |
1792 | ||
c2a1e7da DW |
1793 | static __u8 imsm_check_degraded(struct imsm_super *mpb, int n, int failed) |
1794 | { | |
1795 | struct imsm_dev *dev = get_imsm_dev(mpb, n); | |
1796 | struct imsm_map *map = dev->vol.map; | |
1797 | ||
1798 | if (!failed) | |
1799 | return map->map_state; | |
1800 | ||
1801 | switch (get_imsm_raid_level(map)) { | |
1802 | case 0: | |
1803 | return IMSM_T_STATE_FAILED; | |
1804 | break; | |
1805 | case 1: | |
1806 | if (failed < map->num_members) | |
1807 | return IMSM_T_STATE_DEGRADED; | |
1808 | else | |
1809 | return IMSM_T_STATE_FAILED; | |
1810 | break; | |
1811 | case 10: | |
1812 | { | |
1813 | /** | |
1814 | * check to see if any mirrors have failed, | |
1815 | * otherwise we are degraded | |
1816 | */ | |
1817 | int device_per_mirror = 2; /* FIXME is this always the case? | |
1818 | * and are they always adjacent? | |
1819 | */ | |
1820 | int failed = 0; | |
1821 | int i; | |
1822 | ||
1823 | for (i = 0; i < map->num_members; i++) { | |
1824 | int idx = get_imsm_disk_idx(map, i); | |
1825 | struct imsm_disk *disk = get_imsm_disk(mpb, idx); | |
1826 | ||
1827 | if (__le32_to_cpu(disk->status) & FAILED_DISK) | |
1828 | failed++; | |
1829 | ||
1830 | if (failed >= device_per_mirror) | |
1831 | return IMSM_T_STATE_FAILED; | |
1832 | ||
1833 | /* reset 'failed' for next mirror set */ | |
1834 | if (!((i + 1) % device_per_mirror)) | |
1835 | failed = 0; | |
1836 | } | |
1837 | ||
1838 | return IMSM_T_STATE_DEGRADED; | |
1839 | } | |
1840 | case 5: | |
1841 | if (failed < 2) | |
1842 | return IMSM_T_STATE_DEGRADED; | |
1843 | else | |
1844 | return IMSM_T_STATE_FAILED; | |
1845 | break; | |
1846 | default: | |
1847 | break; | |
1848 | } | |
1849 | ||
1850 | return map->map_state; | |
1851 | } | |
1852 | ||
1853 | static int imsm_count_failed(struct imsm_super *mpb, struct imsm_map *map) | |
1854 | { | |
1855 | int i; | |
1856 | int failed = 0; | |
1857 | struct imsm_disk *disk; | |
1858 | ||
1859 | for (i = 0; i < map->num_members; i++) { | |
1860 | int idx = get_imsm_disk_idx(map, i); | |
1861 | ||
1862 | disk = get_imsm_disk(mpb, idx); | |
1863 | if (__le32_to_cpu(disk->status) & FAILED_DISK) | |
1864 | failed++; | |
1865 | } | |
1866 | ||
1867 | return failed; | |
845dea95 NB |
1868 | } |
1869 | ||
a862209d DW |
1870 | static void imsm_set_array_state(struct active_array *a, int consistent) |
1871 | { | |
1872 | int inst = a->info.container_member; | |
1873 | struct intel_super *super = a->container->sb; | |
1874 | struct imsm_dev *dev = get_imsm_dev(super->mpb, inst); | |
1875 | struct imsm_map *map = &dev->vol.map[0]; | |
1876 | int dirty = !consistent; | |
1877 | int failed; | |
1878 | __u8 map_state; | |
1879 | ||
1880 | if (a->resync_start == ~0ULL) { | |
1881 | failed = imsm_count_failed(super->mpb, map); | |
1882 | map_state = imsm_check_degraded(super->mpb, inst, failed); | |
1883 | if (!failed) | |
1884 | map_state = IMSM_T_STATE_NORMAL; | |
1885 | if (map->map_state != map_state) { | |
4e6e574a | 1886 | dprintf("imsm: map_state %d: %d\n", |
a862209d DW |
1887 | inst, map_state); |
1888 | map->map_state = map_state; | |
1889 | super->updates_pending++; | |
1890 | } | |
1891 | } | |
1892 | ||
1893 | if (dev->vol.dirty != dirty) { | |
4e6e574a | 1894 | dprintf("imsm: mark '%s' (%llu)\n", |
a862209d DW |
1895 | dirty?"dirty":"clean", a->resync_start); |
1896 | ||
1897 | dev->vol.dirty = dirty; | |
1898 | super->updates_pending++; | |
1899 | } | |
1900 | } | |
1901 | ||
8d45d196 | 1902 | static void imsm_set_disk(struct active_array *a, int n, int state) |
845dea95 | 1903 | { |
8d45d196 DW |
1904 | int inst = a->info.container_member; |
1905 | struct intel_super *super = a->container->sb; | |
1906 | struct imsm_dev *dev = get_imsm_dev(super->mpb, inst); | |
1907 | struct imsm_map *map = dev->vol.map; | |
1908 | struct imsm_disk *disk; | |
1909 | __u32 status; | |
1910 | int failed = 0; | |
1911 | int new_failure = 0; | |
1912 | ||
1913 | if (n > map->num_members) | |
1914 | fprintf(stderr, "imsm: set_disk %d out of range 0..%d\n", | |
1915 | n, map->num_members - 1); | |
1916 | ||
1917 | if (n < 0) | |
1918 | return; | |
1919 | ||
4e6e574a | 1920 | dprintf("imsm: set_disk %d:%x\n", n, state); |
8d45d196 DW |
1921 | |
1922 | disk = get_imsm_disk(super->mpb, get_imsm_disk_idx(map, n)); | |
1923 | ||
5802a811 | 1924 | /* check for new failures */ |
8d45d196 DW |
1925 | status = __le32_to_cpu(disk->status); |
1926 | if ((state & DS_FAULTY) && !(status & FAILED_DISK)) { | |
1927 | status |= FAILED_DISK; | |
1928 | disk->status = __cpu_to_le32(status); | |
1929 | new_failure = 1; | |
5802a811 | 1930 | super->updates_pending++; |
8d45d196 DW |
1931 | } |
1932 | ||
5802a811 | 1933 | /* the number of failures have changed, count up 'failed' to determine |
8d45d196 DW |
1934 | * degraded / failed status |
1935 | */ | |
1936 | if (new_failure && map->map_state != IMSM_T_STATE_FAILED) | |
1937 | failed = imsm_count_failed(super->mpb, map); | |
1938 | ||
5802a811 | 1939 | /* determine map_state based on failed or in_sync count */ |
8d45d196 DW |
1940 | if (failed) |
1941 | map->map_state = imsm_check_degraded(super->mpb, inst, failed); | |
5802a811 DW |
1942 | else if (map->map_state == IMSM_T_STATE_DEGRADED) { |
1943 | struct mdinfo *d; | |
1944 | int working = 0; | |
8d45d196 | 1945 | |
5802a811 DW |
1946 | for (d = a->info.devs ; d ; d = d->next) |
1947 | if (d->curr_state & DS_INSYNC) | |
1948 | working++; | |
1949 | ||
1950 | if (working == a->info.array.raid_disks) { | |
1951 | map->map_state = IMSM_T_STATE_NORMAL; | |
1952 | super->updates_pending++; | |
1953 | } | |
1954 | } | |
845dea95 NB |
1955 | } |
1956 | ||
c2a1e7da DW |
1957 | static int store_imsm_mpb(int fd, struct intel_super *super) |
1958 | { | |
1959 | struct imsm_super *mpb = super->mpb; | |
1960 | __u32 mpb_size = __le32_to_cpu(mpb->mpb_size); | |
1961 | unsigned long long dsize; | |
1962 | unsigned long long sectors; | |
1963 | ||
1964 | get_dev_size(fd, NULL, &dsize); | |
1965 | ||
272f648f DW |
1966 | if (mpb_size > 512) { |
1967 | /* -1 to account for anchor */ | |
1968 | sectors = mpb_sectors(mpb) - 1; | |
c2a1e7da | 1969 | |
272f648f DW |
1970 | /* write the extended mpb to the sectors preceeding the anchor */ |
1971 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) | |
1972 | return 1; | |
c2a1e7da | 1973 | |
99e29264 | 1974 | if (write(fd, super->buf + 512, 512 * sectors) != 512 * sectors) |
272f648f DW |
1975 | return 1; |
1976 | } | |
c2a1e7da | 1977 | |
272f648f DW |
1978 | /* first block is stored on second to last sector of the disk */ |
1979 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) | |
c2a1e7da DW |
1980 | return 1; |
1981 | ||
272f648f | 1982 | if (write(fd, super->buf, 512) != 512) |
c2a1e7da DW |
1983 | return 1; |
1984 | ||
c2a1e7da DW |
1985 | return 0; |
1986 | } | |
1987 | ||
2e735d19 | 1988 | static void imsm_sync_metadata(struct supertype *container) |
845dea95 | 1989 | { |
2e735d19 | 1990 | struct intel_super *super = container->sb; |
c2a1e7da DW |
1991 | |
1992 | if (!super->updates_pending) | |
1993 | return; | |
1994 | ||
c2c087e6 | 1995 | write_super_imsm(super, 0); |
c2a1e7da DW |
1996 | |
1997 | super->updates_pending = 0; | |
845dea95 NB |
1998 | } |
1999 | ||
88758e9d DW |
2000 | static struct mdinfo *imsm_activate_spare(struct active_array *a, |
2001 | struct metadata_update **updates) | |
2002 | { | |
2003 | /** | |
2004 | * Take a device that is marked spare in the metadata and use it to | |
2005 | * replace a failed/vacant slot in an array. There may be a case where | |
2006 | * a device is failed in one array but active in a second. | |
2007 | * imsm_process_update catches this case and does not clear the SPARE_DISK | |
2008 | * flag, allowing the second array to start using the device on failure. | |
2009 | * SPARE_DISK is cleared when all arrays are using a device. | |
2010 | * | |
2011 | * FIXME: is this a valid use of SPARE_DISK? | |
2012 | */ | |
2013 | ||
2014 | struct intel_super *super = a->container->sb; | |
2015 | struct imsm_super *mpb = super->mpb; | |
2016 | int inst = a->info.container_member; | |
2017 | struct imsm_dev *dev = get_imsm_dev(mpb, inst); | |
2018 | struct imsm_map *map = dev->vol.map; | |
2019 | int failed = a->info.array.raid_disks; | |
2020 | struct mdinfo *rv = NULL; | |
2021 | struct mdinfo *d; | |
2022 | struct mdinfo *di; | |
2023 | struct metadata_update *mu; | |
2024 | struct dl *dl; | |
2025 | struct imsm_update_activate_spare *u; | |
2026 | int num_spares = 0; | |
2027 | int i; | |
2028 | ||
2029 | for (d = a->info.devs ; d ; d = d->next) { | |
2030 | if ((d->curr_state & DS_FAULTY) && | |
2031 | d->state_fd >= 0) | |
2032 | /* wait for Removal to happen */ | |
2033 | return NULL; | |
2034 | if (d->state_fd >= 0) | |
2035 | failed--; | |
2036 | } | |
2037 | ||
2038 | dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n", | |
2039 | inst, failed, a->info.array.raid_disks, a->info.array.level); | |
2040 | if (imsm_check_degraded(mpb, inst, failed) != IMSM_T_STATE_DEGRADED) | |
2041 | return NULL; | |
2042 | ||
2043 | /* For each slot, if it is not working, find a spare */ | |
2044 | dl = super->disks; | |
2045 | for (i = 0; i < a->info.array.raid_disks; i++) { | |
2046 | for (d = a->info.devs ; d ; d = d->next) | |
2047 | if (d->disk.raid_disk == i) | |
2048 | break; | |
2049 | dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0); | |
2050 | if (d && (d->state_fd >= 0)) | |
2051 | continue; | |
2052 | ||
2053 | /* OK, this device needs recovery. Find a spare */ | |
2054 | for ( ; dl ; dl = dl->next) { | |
2055 | unsigned long long esize; | |
2056 | unsigned long long pos; | |
2057 | struct mdinfo *d2; | |
2058 | struct extent *ex; | |
2059 | struct imsm_disk *disk; | |
2060 | int j; | |
2061 | int found; | |
5257ce36 | 2062 | __u32 array_start; |
88758e9d DW |
2063 | |
2064 | /* If in this array, skip */ | |
2065 | for (d2 = a->info.devs ; d2 ; d2 = d2->next) | |
2066 | if (d2->disk.major == dl->major && | |
2067 | d2->disk.minor == dl->minor) { | |
2068 | dprintf("%x:%x already in array\n", dl->major, dl->minor); | |
2069 | break; | |
2070 | } | |
2071 | if (d2) | |
2072 | continue; | |
2073 | ||
2074 | /* is this unused device marked as a spare? */ | |
2075 | disk = get_imsm_disk(mpb, dl->index); | |
2076 | if (!(__le32_to_cpu(disk->status) & SPARE_DISK)) | |
2077 | continue; | |
2078 | ||
2079 | /* We are allowed to use this device - is there space? | |
2080 | * We need a->info.component_size sectors */ | |
2081 | ex = get_extents(super, dl); | |
2082 | if (!ex) { | |
2083 | dprintf("cannot get extents\n"); | |
2084 | continue; | |
2085 | } | |
2086 | found = 0; | |
2087 | j = 0; | |
2088 | pos = 0; | |
5257ce36 | 2089 | array_start = __le32_to_cpu(map->pba_of_lba0); |
88758e9d DW |
2090 | |
2091 | do { | |
5257ce36 DW |
2092 | /* check that we can start at pba_of_lba0 with |
2093 | * a->info.component_size of space | |
2094 | */ | |
88758e9d | 2095 | esize = ex[j].start - pos; |
5257ce36 DW |
2096 | if (array_start >= pos && |
2097 | array_start + a->info.component_size < ex[j].start) { | |
2098 | found = 1; | |
88758e9d DW |
2099 | break; |
2100 | } | |
5257ce36 DW |
2101 | pos = ex[j].start + ex[j].size; |
2102 | j++; | |
2103 | ||
2104 | } while (ex[j-1].size); | |
88758e9d DW |
2105 | |
2106 | free(ex); | |
2107 | if (!found) { | |
2108 | dprintf("%x:%x does not have %llu at %d\n", | |
2109 | dl->major, dl->minor, | |
2110 | a->info.component_size, | |
2111 | __le32_to_cpu(map->pba_of_lba0)); | |
2112 | /* No room */ | |
2113 | continue; | |
2114 | } | |
2115 | ||
5257ce36 | 2116 | /* found a usable disk with enough space */ |
88758e9d DW |
2117 | di = malloc(sizeof(*di)); |
2118 | memset(di, 0, sizeof(*di)); | |
2119 | di->disk.number = dl->index; | |
2120 | di->disk.raid_disk = i; | |
2121 | di->disk.major = dl->major; | |
2122 | di->disk.minor = dl->minor; | |
2123 | di->disk.state = 0; | |
5257ce36 | 2124 | di->data_offset = array_start; |
88758e9d DW |
2125 | di->component_size = a->info.component_size; |
2126 | di->container_member = inst; | |
2127 | di->next = rv; | |
2128 | rv = di; | |
2129 | num_spares++; | |
2130 | dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor, | |
2131 | i, pos); | |
2132 | ||
2133 | break; | |
2134 | } | |
2135 | } | |
2136 | ||
2137 | if (!rv) | |
2138 | /* No spares found */ | |
2139 | return rv; | |
2140 | /* Now 'rv' has a list of devices to return. | |
2141 | * Create a metadata_update record to update the | |
2142 | * disk_ord_tbl for the array | |
2143 | */ | |
2144 | mu = malloc(sizeof(*mu)); | |
2145 | mu->buf = malloc(sizeof(struct imsm_update_activate_spare) * num_spares); | |
2146 | mu->space = NULL; | |
2147 | mu->len = sizeof(struct imsm_update_activate_spare) * num_spares; | |
2148 | mu->next = *updates; | |
2149 | u = (struct imsm_update_activate_spare *) mu->buf; | |
2150 | ||
2151 | for (di = rv ; di ; di = di->next) { | |
2152 | u->type = update_activate_spare; | |
2153 | u->disk_idx = di->disk.number; | |
2154 | u->slot = di->disk.raid_disk; | |
2155 | u->array = inst; | |
2156 | u->next = u + 1; | |
2157 | u++; | |
2158 | } | |
2159 | (u-1)->next = NULL; | |
2160 | *updates = mu; | |
2161 | ||
2162 | return rv; | |
2163 | } | |
2164 | ||
e8319a19 DW |
2165 | static int weight(unsigned int field) |
2166 | { | |
2167 | int weight; | |
2168 | ||
2169 | for (weight = 0; field; weight++) | |
2170 | field &= field - 1; | |
2171 | ||
2172 | return weight; | |
2173 | } | |
2174 | ||
8273f55e DW |
2175 | static int disks_overlap(struct imsm_map *m1, struct imsm_map *m2) |
2176 | { | |
2177 | int i; | |
2178 | int j; | |
2179 | int idx; | |
2180 | ||
2181 | for (i = 0; i < m1->num_members; i++) { | |
2182 | idx = get_imsm_disk_idx(m1, i); | |
2183 | for (j = 0; j < m2->num_members; j++) | |
2184 | if (idx == get_imsm_disk_idx(m2, j)) | |
2185 | return 1; | |
2186 | } | |
2187 | ||
2188 | return 0; | |
2189 | } | |
2190 | ||
e8319a19 DW |
2191 | static void imsm_process_update(struct supertype *st, |
2192 | struct metadata_update *update) | |
2193 | { | |
2194 | /** | |
2195 | * crack open the metadata_update envelope to find the update record | |
2196 | * update can be one of: | |
2197 | * update_activate_spare - a spare device has replaced a failed | |
2198 | * device in an array, update the disk_ord_tbl. If this disk is | |
2199 | * present in all member arrays then also clear the SPARE_DISK | |
2200 | * flag | |
2201 | */ | |
2202 | struct intel_super *super = st->sb; | |
2203 | struct imsm_super *mpb = super->mpb; | |
2204 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; | |
2205 | ||
2206 | switch (type) { | |
2207 | case update_activate_spare: { | |
2208 | struct imsm_update_activate_spare *u = (void *) update->buf; | |
2209 | struct imsm_dev *dev = get_imsm_dev(mpb, u->array); | |
2210 | struct imsm_map *map = &dev->vol.map[0]; | |
2211 | struct active_array *a; | |
2212 | struct imsm_disk *disk; | |
2213 | __u32 status; | |
2214 | struct dl *dl; | |
2215 | struct mdinfo *d; | |
2216 | unsigned int members; | |
2217 | unsigned int found; | |
2218 | int victim; | |
2219 | int i; | |
2220 | ||
2221 | for (dl = super->disks; dl; dl = dl->next) | |
2222 | if (dl->index == u->disk_idx) | |
2223 | break; | |
2224 | ||
2225 | if (!dl) { | |
2226 | fprintf(stderr, "error: imsm_activate_spare passed " | |
2227 | "an unknown disk_idx: %d\n", u->disk_idx); | |
2228 | return; | |
2229 | } | |
2230 | ||
2231 | super->updates_pending++; | |
2232 | ||
2233 | victim = get_imsm_disk_idx(map, u->slot); | |
2234 | map->disk_ord_tbl[u->slot] = __cpu_to_le32(u->disk_idx); | |
2235 | disk = get_imsm_disk(mpb, u->disk_idx); | |
2236 | status = __le32_to_cpu(disk->status); | |
2237 | status |= CONFIGURED_DISK; | |
2238 | disk->status = __cpu_to_le32(status); | |
2239 | ||
2240 | /* map unique/live arrays using the spare */ | |
2241 | members = 0; | |
2242 | found = 0; | |
2243 | for (a = st->arrays; a; a = a->next) { | |
2244 | int inst = a->info.container_member; | |
2245 | ||
2246 | dev = get_imsm_dev(mpb, inst); | |
2247 | map = &dev->vol.map[0]; | |
2248 | if (map->raid_level > 0) | |
2249 | members |= 1 << inst; | |
2250 | for (d = a->info.devs; d; d = d->next) | |
2251 | if (d->disk.major == dl->major && | |
2252 | d->disk.minor == dl->minor) | |
2253 | found |= 1 << inst; | |
2254 | } | |
2255 | ||
2256 | /* until all arrays that can absorb this disk have absorbed | |
2257 | * this disk it can still be considered a spare | |
2258 | */ | |
2259 | if (weight(found) >= weight(members)) { | |
2260 | status = __le32_to_cpu(disk->status); | |
2261 | status &= ~SPARE_DISK; | |
2262 | disk->status = __cpu_to_le32(status); | |
2263 | } | |
2264 | ||
2265 | /* count arrays using the victim in the metadata */ | |
2266 | found = 0; | |
2267 | for (a = st->arrays; a ; a = a->next) { | |
2268 | dev = get_imsm_dev(mpb, a->info.container_member); | |
2269 | map = &dev->vol.map[0]; | |
2270 | for (i = 0; i < map->num_members; i++) | |
2271 | if (victim == get_imsm_disk_idx(map, i)) | |
2272 | found++; | |
2273 | } | |
2274 | ||
2275 | /* clear some flags if the victim is no longer being | |
2276 | * utilized anywhere | |
2277 | */ | |
2278 | disk = get_imsm_disk(mpb, victim); | |
2279 | if (!found) { | |
2280 | status = __le32_to_cpu(disk->status); | |
2281 | status &= ~(CONFIGURED_DISK | USABLE_DISK); | |
2282 | disk->status = __cpu_to_le32(status); | |
2283 | } | |
8273f55e DW |
2284 | break; |
2285 | } | |
2286 | case update_create_array: { | |
2287 | /* someone wants to create a new array, we need to be aware of | |
2288 | * a few races/collisions: | |
2289 | * 1/ 'Create' called by two separate instances of mdadm | |
2290 | * 2/ 'Create' versus 'activate_spare': mdadm has chosen | |
2291 | * devices that have since been assimilated via | |
2292 | * activate_spare. | |
2293 | * In the event this update can not be carried out mdadm will | |
2294 | * (FIX ME) notice that its update did not take hold. | |
2295 | */ | |
2296 | struct imsm_update_create_array *u = (void *) update->buf; | |
2297 | struct imsm_dev *dev; | |
2298 | struct imsm_map *map, *new_map; | |
2299 | unsigned long long start, end; | |
2300 | unsigned long long new_start, new_end; | |
2301 | int i; | |
2302 | int overlap = 0; | |
2303 | ||
2304 | /* handle racing creates: first come first serve */ | |
2305 | if (u->dev_idx < mpb->num_raid_devs) { | |
2306 | dprintf("%s: subarray %d already defined\n", | |
2307 | __func__, u->dev_idx); | |
2308 | return; | |
2309 | } | |
2310 | ||
2311 | /* check update is next in sequence */ | |
2312 | if (u->dev_idx != mpb->num_raid_devs) { | |
2313 | dprintf("%s: can not create arrays out of sequence\n", | |
2314 | __func__); | |
2315 | return; | |
2316 | } | |
2317 | ||
2318 | new_map = &u->dev.vol.map[0]; | |
2319 | new_start = __le32_to_cpu(new_map->pba_of_lba0); | |
2320 | new_end = new_start + __le32_to_cpu(new_map->blocks_per_member); | |
2321 | ||
2322 | /* handle activate_spare versus create race: | |
2323 | * check to make sure that overlapping arrays do not include | |
2324 | * overalpping disks | |
2325 | */ | |
2326 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
2327 | dev = get_imsm_dev(mpb, i); | |
2328 | map = &dev->vol.map[0]; | |
2329 | start = __le32_to_cpu(map->pba_of_lba0); | |
2330 | end = start + __le32_to_cpu(map->blocks_per_member); | |
2331 | if ((new_start >= start && new_start <= end) || | |
2332 | (start >= new_start && start <= new_end)) | |
2333 | overlap = 1; | |
2334 | if (overlap && disks_overlap(map, new_map)) { | |
2335 | dprintf("%s: arrays overlap\n", __func__); | |
2336 | return; | |
2337 | } | |
2338 | } | |
2339 | /* check num_members sanity */ | |
2340 | if (new_map->num_members > mpb->num_disks) { | |
2341 | dprintf("%s: num_disks out of range\n", __func__); | |
2342 | return; | |
2343 | } | |
2344 | ||
2345 | super->updates_pending++; | |
2346 | mpb->num_raid_devs++; | |
2347 | dev = get_imsm_dev(mpb, u->dev_idx); | |
2348 | memcpy(dev, &u->dev, sizeof(*dev)); | |
2349 | map = &dev->vol.map[0]; | |
2350 | memcpy(map->disk_ord_tbl, new_map->disk_ord_tbl, | |
2351 | sizeof(__u32) * new_map->num_members); | |
2352 | ||
2353 | /* fix up flags, if arrays overlap then the drives can not be | |
2354 | * spares | |
2355 | */ | |
2356 | for (i = 0; i < map->num_members; i++) { | |
2357 | struct imsm_disk *disk; | |
2358 | __u32 status; | |
2359 | ||
2360 | disk = get_imsm_disk(mpb, get_imsm_disk_idx(map, i)); | |
2361 | status = __le32_to_cpu(disk->status); | |
2362 | status |= CONFIGURED_DISK; | |
2363 | if (overlap) | |
2364 | status &= ~SPARE_DISK; | |
2365 | disk->status = __cpu_to_le32(status); | |
2366 | } | |
2367 | break; | |
e8319a19 DW |
2368 | } |
2369 | } | |
2370 | } | |
88758e9d | 2371 | |
8273f55e DW |
2372 | static void imsm_prepare_update(struct supertype *st, |
2373 | struct metadata_update *update) | |
2374 | { | |
2375 | /* Allocate space to hold a new mpb if necessary. We currently | |
2376 | * allocate enough to hold 2 subarrays for the given number of disks. | |
2377 | * This may not be sufficient iff reshaping. | |
2378 | * | |
2379 | * FIX ME handle the reshape case. | |
2380 | * | |
2381 | * The monitor will be able to safely change super->mpb by arranging | |
2382 | * for it to be freed in check_update_queue(). I.e. the monitor thread | |
2383 | * will start using the new pointer and the manager can continue to use | |
2384 | * the old value until check_update_queue() runs. | |
2385 | */ | |
2386 | ||
2387 | return; | |
2388 | } | |
2389 | ||
cdddbdbc DW |
2390 | struct superswitch super_imsm = { |
2391 | #ifndef MDASSEMBLE | |
2392 | .examine_super = examine_super_imsm, | |
2393 | .brief_examine_super = brief_examine_super_imsm, | |
2394 | .detail_super = detail_super_imsm, | |
2395 | .brief_detail_super = brief_detail_super_imsm, | |
bf5a934a | 2396 | .write_init_super = write_init_super_imsm, |
cdddbdbc DW |
2397 | #endif |
2398 | .match_home = match_home_imsm, | |
2399 | .uuid_from_super= uuid_from_super_imsm, | |
2400 | .getinfo_super = getinfo_super_imsm, | |
2401 | .update_super = update_super_imsm, | |
2402 | ||
2403 | .avail_size = avail_size_imsm, | |
2404 | ||
2405 | .compare_super = compare_super_imsm, | |
2406 | ||
2407 | .load_super = load_super_imsm, | |
bf5a934a DW |
2408 | .init_super = init_super_imsm, |
2409 | .add_to_super = add_to_super_imsm, | |
cdddbdbc DW |
2410 | .store_super = store_zero_imsm, |
2411 | .free_super = free_super_imsm, | |
2412 | .match_metadata_desc = match_metadata_desc_imsm, | |
bf5a934a | 2413 | .container_content = container_content_imsm, |
cdddbdbc DW |
2414 | |
2415 | .validate_geometry = validate_geometry_imsm, | |
cdddbdbc | 2416 | .external = 1, |
845dea95 NB |
2417 | |
2418 | /* for mdmon */ | |
2419 | .open_new = imsm_open_new, | |
2420 | .load_super = load_super_imsm, | |
ed9d66aa | 2421 | .set_array_state= imsm_set_array_state, |
845dea95 NB |
2422 | .set_disk = imsm_set_disk, |
2423 | .sync_metadata = imsm_sync_metadata, | |
88758e9d | 2424 | .activate_spare = imsm_activate_spare, |
e8319a19 | 2425 | .process_update = imsm_process_update, |
8273f55e | 2426 | .prepare_update = imsm_prepare_update, |
cdddbdbc | 2427 | }; |