]> git.ipfire.org Git - thirdparty/mdadm.git/blob - super-intel.c
projects / thirdparty / mdadm.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
imsm: fix raid1 creation
[thirdparty/mdadm.git] / super-intel.c
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 struct imsm_super *mpb;
131 void *buf;
132 };
133 int updates_pending; /* count of pending updates for mdmon */
134 int creating_imsm; /* flag to indicate container creation */
135 int current_vol; /* index of raid device undergoing creation */
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
146 struct extent {
147 unsigned long long start, size;
148 };
149
150 static struct supertype *match_metadata_desc_imsm(char *arg)
151 {
152 struct supertype *st;
153
154 if (strcmp(arg, "imsm") != 0 &&
155 strcmp(arg, "default") != 0
156 )
157 return NULL;
158
159 st = malloc(sizeof(*st));
160 memset(st, 0, sizeof(*st));
161 st->ss = &super_imsm;
162 st->max_devs = IMSM_MAX_DEVICES;
163 st->minor_version = 0;
164 st->sb = NULL;
165 return st;
166 }
167
168 static __u8 *get_imsm_version(struct imsm_super *mpb)
169 {
170 return &mpb->sig[MPB_SIG_LEN];
171 }
172
173 static struct imsm_disk *get_imsm_disk(struct imsm_super *mpb, __u8 index)
174 {
175 if (index > mpb->num_disks - 1)
176 return NULL;
177 return &mpb->disk[index];
178 }
179
180 static __u32 gen_imsm_checksum(struct imsm_super *mpb)
181 {
182 __u32 end = mpb->mpb_size / sizeof(end);
183 __u32 *p = (__u32 *) mpb;
184 __u32 sum = 0;
185
186 while (end--)
187 sum += __le32_to_cpu(*p++);
188
189 return sum - __le32_to_cpu(mpb->check_sum);
190 }
191
192 static size_t sizeof_imsm_dev(struct imsm_dev *dev)
193 {
194 size_t size = sizeof(*dev);
195
196 /* each map has disk_ord_tbl[num_members - 1] additional space */
197 size += sizeof(__u32) * (dev->vol.map[0].num_members - 1);
198
199 /* migrating means an additional map */
200 if (dev->vol.migr_state) {
201 size += sizeof(struct imsm_map);
202 size += sizeof(__u32) * (dev->vol.map[1].num_members - 1);
203 }
204
205 return size;
206 }
207
208 static struct imsm_dev *get_imsm_dev(struct imsm_super *mpb, __u8 index)
209 {
210 int offset;
211 int i;
212 void *_mpb = mpb;
213
214 if (index > mpb->num_raid_devs - 1)
215 return NULL;
216
217 /* devices start after all disks */
218 offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb;
219
220 for (i = 0; i <= index; i++)
221 if (i == index)
222 return _mpb + offset;
223 else
224 offset += sizeof_imsm_dev(_mpb + offset);
225
226 return NULL;
227 }
228
229 static __u32 get_imsm_disk_idx(struct imsm_map *map, int slot)
230 {
231 __u32 *ord_tbl = &map->disk_ord_tbl[slot];
232
233 /* top byte is 'special' */
234 return __le32_to_cpu(*ord_tbl & ~(0xff << 24));
235 }
236
237 static int get_imsm_raid_level(struct imsm_map *map)
238 {
239 if (map->raid_level == 1) {
240 if (map->num_members == 2)
241 return 1;
242 else
243 return 10;
244 }
245
246 return map->raid_level;
247 }
248
249 static int cmp_extent(const void *av, const void *bv)
250 {
251 const struct extent *a = av;
252 const struct extent *b = bv;
253 if (a->start < b->start)
254 return -1;
255 if (a->start > b->start)
256 return 1;
257 return 0;
258 }
259
260 static struct extent *get_extents(struct intel_super *super, struct dl *dl)
261 {
262 /* find a list of used extents on the given physical device */
263 struct imsm_super *mpb = super->mpb;
264 struct imsm_disk *disk;
265 struct extent *rv, *e;
266 int i, j;
267 int memberships = 0;
268
269 disk = get_imsm_disk(mpb, dl->index);
270 if (!disk)
271 return NULL;
272
273 for (i = 0; i < mpb->num_raid_devs; i++) {
274 struct imsm_dev *dev = get_imsm_dev(mpb, i);
275 struct imsm_map *map = dev->vol.map;
276
277 for (j = 0; j < map->num_members; j++) {
278 __u32 index = get_imsm_disk_idx(map, j);
279
280 if (index == dl->index)
281 memberships++;
282 }
283 }
284 rv = malloc(sizeof(struct extent) * (memberships + 1));
285 if (!rv)
286 return NULL;
287 e = rv;
288
289 for (i = 0; i < mpb->num_raid_devs; i++) {
290 struct imsm_dev *dev = get_imsm_dev(mpb, i);
291 struct imsm_map *map = dev->vol.map;
292
293 for (j = 0; j < map->num_members; j++) {
294 __u32 index = get_imsm_disk_idx(map, j);
295
296 if (index == dl->index) {
297 e->start = __le32_to_cpu(map->pba_of_lba0);
298 e->size = __le32_to_cpu(map->blocks_per_member);
299 e++;
300 }
301 }
302 }
303 qsort(rv, memberships, sizeof(*rv), cmp_extent);
304
305 e->start = __le32_to_cpu(disk->total_blocks) -
306 (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS);
307 e->size = 0;
308 return rv;
309 }
310
311 #ifndef MDASSEMBLE
312 static void print_imsm_dev(struct imsm_dev *dev, int index)
313 {
314 __u64 sz;
315 int slot;
316 struct imsm_map *map = dev->vol.map;
317
318 printf("\n");
319 printf("[%s]:\n", dev->volume);
320 printf(" RAID Level : %d\n", get_imsm_raid_level(map));
321 printf(" Members : %d\n", map->num_members);
322 for (slot = 0; slot < map->num_members; slot++)
323 if (index == get_imsm_disk_idx(map, slot))
324 break;
325 if (slot < map->num_members)
326 printf(" This Slot : %d\n", slot);
327 else
328 printf(" This Slot : ?\n");
329 sz = __le32_to_cpu(dev->size_high);
330 sz <<= 32;
331 sz += __le32_to_cpu(dev->size_low);
332 printf(" Array Size : %llu%s\n", (unsigned long long)sz,
333 human_size(sz * 512));
334 sz = __le32_to_cpu(map->blocks_per_member);
335 printf(" Per Dev Size : %llu%s\n", (unsigned long long)sz,
336 human_size(sz * 512));
337 printf(" Sector Offset : %u\n",
338 __le32_to_cpu(map->pba_of_lba0));
339 printf(" Num Stripes : %u\n",
340 __le32_to_cpu(map->num_data_stripes));
341 printf(" Chunk Size : %u KiB\n",
342 __le16_to_cpu(map->blocks_per_strip) / 2);
343 printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks));
344 printf(" Migrate State : %s\n", dev->vol.migr_state ? "migrating" : "idle");
345 printf(" Dirty State : %s\n", dev->vol.dirty ? "dirty" : "clean");
346 printf(" Map State : %s\n", map_state_str[map->map_state]);
347 }
348
349 static void print_imsm_disk(struct imsm_super *mpb, int index)
350 {
351 struct imsm_disk *disk = get_imsm_disk(mpb, index);
352 char str[MAX_RAID_SERIAL_LEN];
353 __u32 s;
354 __u64 sz;
355
356 printf("\n");
357 snprintf(str, MAX_RAID_SERIAL_LEN, "%s", disk->serial);
358 printf(" Disk%02d Serial : %s\n", index, str);
359 s = __le32_to_cpu(disk->status);
360 printf(" State :%s%s%s%s\n", s&SPARE_DISK ? " spare" : "",
361 s&CONFIGURED_DISK ? " active" : "",
362 s&FAILED_DISK ? " failed" : "",
363 s&USABLE_DISK ? " usable" : "");
364 printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id));
365 sz = __le32_to_cpu(disk->total_blocks) -
366 (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS * mpb->num_raid_devs);
367 printf(" Usable Size : %llu%s\n", (unsigned long long)sz,
368 human_size(sz * 512));
369 }
370
371 static void examine_super_imsm(struct supertype *st, char *homehost)
372 {
373 struct intel_super *super = st->sb;
374 struct imsm_super *mpb = super->mpb;
375 char str[MAX_SIGNATURE_LENGTH];
376 int i;
377 __u32 sum;
378
379 snprintf(str, MPB_SIG_LEN, "%s", mpb->sig);
380 printf(" Magic : %s\n", str);
381 snprintf(str, strlen(MPB_VERSION_RAID0), "%s", get_imsm_version(mpb));
382 printf(" Version : %s\n", get_imsm_version(mpb));
383 printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num));
384 printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num));
385 sum = __le32_to_cpu(mpb->check_sum);
386 printf(" Checksum : %08x %s\n", sum,
387 gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect");
388 printf(" MPB Sectors : %d\n", mpb_sectors(mpb));
389 printf(" Disks : %d\n", mpb->num_disks);
390 printf(" RAID Devices : %d\n", mpb->num_raid_devs);
391 print_imsm_disk(mpb, super->disks->index);
392 for (i = 0; i < mpb->num_raid_devs; i++)
393 print_imsm_dev(get_imsm_dev(mpb, i), super->disks->index);
394 for (i = 0; i < mpb->num_disks; i++) {
395 if (i == super->disks->index)
396 continue;
397 print_imsm_disk(mpb, i);
398 }
399 }
400
401 static void brief_examine_super_imsm(struct supertype *st)
402 {
403 struct intel_super *super = st->sb;
404 struct imsm_super *mpb = super->mpb;
405
406 printf("ARRAY /dev/imsm family=%08x metadata=external:imsm\n",
407 __le32_to_cpu(mpb->family_num));
408 }
409
410 static void detail_super_imsm(struct supertype *st, char *homehost)
411 {
412 printf("%s\n", __FUNCTION__);
413 }
414
415 static void brief_detail_super_imsm(struct supertype *st)
416 {
417 printf("%s\n", __FUNCTION__);
418 }
419 #endif
420
421 static int match_home_imsm(struct supertype *st, char *homehost)
422 {
423 printf("%s\n", __FUNCTION__);
424
425 return 0;
426 }
427
428 static void uuid_from_super_imsm(struct supertype *st, int uuid[4])
429 {
430 printf("%s\n", __FUNCTION__);
431 }
432
433 #if 0
434 static void
435 get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p)
436 {
437 __u8 *v = get_imsm_version(mpb);
438 __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH;
439 char major[] = { 0, 0, 0 };
440 char minor[] = { 0 ,0, 0 };
441 char patch[] = { 0, 0, 0 };
442 char *ver_parse[] = { major, minor, patch };
443 int i, j;
444
445 i = j = 0;
446 while (*v != '\0' && v < end) {
447 if (*v != '.' && j < 2)
448 ver_parse[i][j++] = *v;
449 else {
450 i++;
451 j = 0;
452 }
453 v++;
454 }
455
456 *m = strtol(minor, NULL, 0);
457 *p = strtol(patch, NULL, 0);
458 }
459 #endif
460
461 static int imsm_level_to_layout(int level)
462 {
463 switch (level) {
464 case 0:
465 case 1:
466 return 0;
467 case 5:
468 case 6:
469 return ALGORITHM_LEFT_SYMMETRIC;
470 case 10:
471 return 0x102; //FIXME is this correct?
472 }
473 return -1;
474 }
475
476 static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info)
477 {
478 struct intel_super *super = st->sb;
479 struct imsm_super *mpb = super->mpb;
480 struct imsm_dev *dev = get_imsm_dev(mpb, super->current_vol);
481 struct imsm_map *map = &dev->vol.map[0];
482
483 info->container_member = super->current_vol;
484 info->array.raid_disks = map->num_members;
485 info->array.level = get_imsm_raid_level(map);
486 info->array.layout = imsm_level_to_layout(info->array.level);
487 info->array.md_minor = -1;
488 info->array.ctime = 0;
489 info->array.utime = 0;
490 info->array.chunk_size = __le16_to_cpu(map->blocks_per_strip * 512);
491
492 info->data_offset = __le32_to_cpu(map->pba_of_lba0);
493 info->component_size = __le32_to_cpu(map->blocks_per_member);
494
495 info->disk.major = 0;
496 info->disk.minor = 0;
497
498 sprintf(info->text_version, "/%s/%d",
499 devnum2devname(st->container_dev),
500 info->container_member);
501 }
502
503
504 static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info)
505 {
506 struct intel_super *super = st->sb;
507 struct imsm_super *mpb = super->mpb;
508 struct imsm_disk *disk;
509 __u32 s;
510
511 if (super->current_vol >= 0) {
512 getinfo_super_imsm_volume(st, info);
513 return;
514 }
515 info->array.raid_disks = mpb->num_disks;
516 info->array.level = LEVEL_CONTAINER;
517 info->array.layout = 0;
518 info->array.md_minor = -1;
519 info->array.ctime = 0; /* N/A for imsm */
520 info->array.utime = 0;
521 info->array.chunk_size = 0;
522
523 info->disk.major = 0;
524 info->disk.minor = 0;
525 info->disk.raid_disk = -1;
526 info->reshape_active = 0;
527 strcpy(info->text_version, "imsm");
528 info->disk.number = -1;
529 info->disk.state = 0;
530
531 if (super->disks) {
532 disk = get_imsm_disk(mpb, super->disks->index);
533 info->disk.number = super->disks->index;
534 info->disk.raid_disk = super->disks->index;
535 info->data_offset = __le32_to_cpu(disk->total_blocks) -
536 (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS);
537 info->component_size = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
538 s = __le32_to_cpu(disk->status);
539 info->disk.state = s & CONFIGURED_DISK ? (1 << MD_DISK_ACTIVE) : 0;
540 info->disk.state |= s & FAILED_DISK ? (1 << MD_DISK_FAULTY) : 0;
541 info->disk.state |= s & USABLE_DISK ? (1 << MD_DISK_SYNC) : 0;
542 }
543 }
544
545 static int update_super_imsm(struct supertype *st, struct mdinfo *info,
546 char *update, char *devname, int verbose,
547 int uuid_set, char *homehost)
548 {
549 /* FIXME */
550
551 /* For 'assemble' and 'force' we need to return non-zero if any
552 * change was made. For others, the return value is ignored.
553 * Update options are:
554 * force-one : This device looks a bit old but needs to be included,
555 * update age info appropriately.
556 * assemble: clear any 'faulty' flag to allow this device to
557 * be assembled.
558 * force-array: Array is degraded but being forced, mark it clean
559 * if that will be needed to assemble it.
560 *
561 * newdev: not used ????
562 * grow: Array has gained a new device - this is currently for
563 * linear only
564 * resync: mark as dirty so a resync will happen.
565 * name: update the name - preserving the homehost
566 *
567 * Following are not relevant for this imsm:
568 * sparc2.2 : update from old dodgey metadata
569 * super-minor: change the preferred_minor number
570 * summaries: update redundant counters.
571 * uuid: Change the uuid of the array to match watch is given
572 * homehost: update the recorded homehost
573 * _reshape_progress: record new reshape_progress position.
574 */
575 int rv = 0;
576 //struct intel_super *super = st->sb;
577 //struct imsm_super *mpb = super->mpb;
578
579 if (strcmp(update, "grow") == 0) {
580 }
581 if (strcmp(update, "resync") == 0) {
582 /* dev->vol.dirty = 1; */
583 }
584
585 /* IMSM has no concept of UUID or homehost */
586
587 return rv;
588 }
589
590 static size_t disks_to_mpb_size(int disks)
591 {
592 size_t size;
593
594 size = sizeof(struct imsm_super);
595 size += (disks - 1) * sizeof(struct imsm_disk);
596 size += 2 * sizeof(struct imsm_dev);
597 /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */
598 size += (4 - 2) * sizeof(struct imsm_map);
599 /* 4 possible disk_ord_tbl's */
600 size += 4 * (disks - 1) * sizeof(__u32);
601
602 return size;
603 }
604
605 static __u64 avail_size_imsm(struct supertype *st, __u64 devsize)
606 {
607 if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS))
608 return 0;
609
610 return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS);
611 }
612
613 static int compare_super_imsm(struct supertype *st, struct supertype *tst)
614 {
615 /*
616 * return:
617 * 0 same, or first was empty, and second was copied
618 * 1 second had wrong number
619 * 2 wrong uuid
620 * 3 wrong other info
621 */
622 struct intel_super *first = st->sb;
623 struct intel_super *sec = tst->sb;
624
625 if (!first) {
626 st->sb = tst->sb;
627 tst->sb = NULL;
628 return 0;
629 }
630
631 if (memcmp(first->mpb->sig, sec->mpb->sig, MAX_SIGNATURE_LENGTH) != 0)
632 return 3;
633 if (first->mpb->family_num != sec->mpb->family_num)
634 return 3;
635 if (first->mpb->mpb_size != sec->mpb->mpb_size)
636 return 3;
637 if (first->mpb->check_sum != sec->mpb->check_sum)
638 return 3;
639
640 return 0;
641 }
642
643 extern int scsi_get_serial(int fd, void *buf, size_t buf_len);
644
645 static int imsm_read_serial(int fd, char *devname,
646 __u8 serial[MAX_RAID_SERIAL_LEN])
647 {
648 unsigned char scsi_serial[255];
649 int sg_fd;
650 int rv;
651 int rsp_len;
652 int i, cnt;
653
654 memset(scsi_serial, 0, sizeof(scsi_serial));
655
656 sg_fd = sysfs_disk_to_sg(fd);
657 if (sg_fd < 0) {
658 if (devname)
659 fprintf(stderr,
660 Name ": Failed to open sg interface for %s: %s\n",
661 devname, strerror(errno));
662 return 1;
663 }
664
665 rv = scsi_get_serial(sg_fd, scsi_serial, sizeof(scsi_serial));
666 close(sg_fd);
667
668 if (rv != 0) {
669 if (devname)
670 fprintf(stderr,
671 Name ": Failed to retrieve serial for %s\n",
672 devname);
673 return rv;
674 }
675
676 rsp_len = scsi_serial[3];
677 for (i = 0, cnt = 0; i < rsp_len; i++) {
678 if (!isspace(scsi_serial[4 + i]))
679 serial[cnt++] = scsi_serial[4 + i];
680 if (cnt == MAX_RAID_SERIAL_LEN)
681 break;
682 }
683
684 serial[MAX_RAID_SERIAL_LEN - 1] = '\0';
685
686 return 0;
687 }
688
689 static int
690 load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd)
691 {
692 struct imsm_super *mpb = super->mpb;
693 struct dl *dl;
694 struct stat stb;
695 struct imsm_disk *disk;
696 int rv;
697 int i;
698
699 dl = malloc(sizeof(*dl));
700 if (!dl) {
701 if (devname)
702 fprintf(stderr,
703 Name ": failed to allocate disk buffer for %s\n",
704 devname);
705 return 2;
706 }
707 memset(dl, 0, sizeof(*dl));
708
709 fstat(fd, &stb);
710 dl->major = major(stb.st_rdev);
711 dl->minor = minor(stb.st_rdev);
712 dl->next = super->disks;
713 dl->fd = keep_fd ? fd : -1;
714 dl->devname = devname ? strdup(devname) : NULL;
715 dl->index = -1;
716 super->disks = dl;
717 rv = imsm_read_serial(fd, devname, dl->serial);
718
719 if (rv != 0)
720 return 2;
721
722 /* look up this disk's index */
723 for (i = 0; i < mpb->num_disks; i++) {
724 disk = get_imsm_disk(mpb, i);
725
726 if (memcmp(disk->serial, dl->serial, MAX_RAID_SERIAL_LEN) == 0)
727 break;
728 }
729
730 if (i > mpb->num_disks)
731 return 2;
732
733 dl->index = i;
734
735 return 0;
736 }
737
738 /* load_imsm_mpb - read matrix metadata
739 * allocates super->mpb to be freed by free_super
740 */
741 static int load_imsm_mpb(int fd, struct intel_super *super, char *devname)
742 {
743 unsigned long long dsize;
744 size_t len, mpb_size;
745 unsigned long long sectors;
746 struct stat;
747 struct imsm_super *anchor;
748 __u32 check_sum;
749
750 get_dev_size(fd, NULL, &dsize);
751
752 if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) {
753 if (devname)
754 fprintf(stderr,
755 Name ": Cannot seek to anchor block on %s: %s\n",
756 devname, strerror(errno));
757 return 1;
758 }
759
760 len = 512;
761 if (posix_memalign((void**)&anchor, 512, len) != 0) {
762 if (devname)
763 fprintf(stderr,
764 Name ": Failed to allocate imsm anchor buffer"
765 " on %s\n", devname);
766 return 1;
767 }
768 if (read(fd, anchor, len) != len) {
769 if (devname)
770 fprintf(stderr,
771 Name ": Cannot read anchor block on %s: %s\n",
772 devname, strerror(errno));
773 free(anchor);
774 return 1;
775 }
776
777 if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) {
778 if (devname)
779 fprintf(stderr,
780 Name ": no IMSM anchor on %s\n", devname);
781 free(anchor);
782 return 2;
783 }
784
785 mpb_size = __le32_to_cpu(anchor->mpb_size);
786 mpb_size = ROUND_UP(mpb_size, 512);
787 if (posix_memalign((void**)&super->mpb, 512, mpb_size) != 0) {
788 if (devname)
789 fprintf(stderr,
790 Name ": unable to allocate %zu byte mpb buffer\n",
791 mpb_size);
792 free(anchor);
793 return 2;
794 }
795 memcpy(super->buf, anchor, len);
796
797 sectors = mpb_sectors(anchor) - 1;
798 free(anchor);
799 if (!sectors)
800 return load_imsm_disk(fd, super, devname, 0);
801
802 /* read the extended mpb */
803 if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) {
804 if (devname)
805 fprintf(stderr,
806 Name ": Cannot seek to extended mpb on %s: %s\n",
807 devname, strerror(errno));
808 return 1;
809 }
810
811 len = mpb_size - 512;
812 if (read(fd, super->buf + 512, len) != len) {
813 if (devname)
814 fprintf(stderr,
815 Name ": Cannot read extended mpb on %s: %s\n",
816 devname, strerror(errno));
817 return 2;
818 }
819
820 check_sum = gen_imsm_checksum(super->mpb);
821 if (check_sum != __le32_to_cpu(super->mpb->check_sum)) {
822 if (devname)
823 fprintf(stderr,
824 Name ": IMSM checksum %x != %x on %s\n",
825 check_sum, __le32_to_cpu(super->mpb->check_sum),
826 devname);
827 return 2;
828 }
829
830 return load_imsm_disk(fd, super, devname, 0);
831 }
832
833 static void free_imsm_disks(struct intel_super *super)
834 {
835 while (super->disks) {
836 struct dl *d = super->disks;
837
838 super->disks = d->next;
839 if (d->fd >= 0)
840 close(d->fd);
841 if (d->devname)
842 free(d->devname);
843 free(d);
844 }
845 }
846
847 static void free_imsm(struct intel_super *super)
848 {
849 if (super->mpb)
850 free(super->mpb);
851 free_imsm_disks(super);
852 free(super);
853 }
854
855
856 static void free_super_imsm(struct supertype *st)
857 {
858 struct intel_super *super = st->sb;
859
860 if (!super)
861 return;
862
863 free_imsm(super);
864 st->sb = NULL;
865 }
866
867 static struct intel_super *alloc_super(int creating_imsm)
868 {
869 struct intel_super *super = malloc(sizeof(*super));
870
871 if (super) {
872 memset(super, 0, sizeof(*super));
873 super->creating_imsm = creating_imsm;
874 super->current_vol = -1;
875 }
876
877 return super;
878 }
879
880 #ifndef MDASSEMBLE
881 static int load_super_imsm_all(struct supertype *st, int fd, void **sbp,
882 char *devname, int keep_fd)
883 {
884 struct mdinfo *sra;
885 struct intel_super *super;
886 struct mdinfo *sd, *best = NULL;
887 __u32 bestgen = 0;
888 __u32 gen;
889 char nm[20];
890 int dfd;
891 int rv;
892
893 /* check if this disk is a member of an active array */
894 sra = sysfs_read(fd, 0, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE);
895 if (!sra)
896 return 1;
897
898 if (sra->array.major_version != -1 ||
899 sra->array.minor_version != -2 ||
900 strcmp(sra->text_version, "imsm") != 0)
901 return 1;
902
903 super = alloc_super(0);
904 if (!super)
905 return 1;
906
907 /* find the most up to date disk in this array */
908 for (sd = sra->devs; sd; sd = sd->next) {
909 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
910 dfd = dev_open(nm, keep_fd ? O_RDWR : O_RDONLY);
911 if (!dfd) {
912 free_imsm(super);
913 return 2;
914 }
915 rv = load_imsm_mpb(dfd, super, NULL);
916 if (!keep_fd)
917 close(dfd);
918 if (rv == 0) {
919 gen = __le32_to_cpu(super->mpb->generation_num);
920 if (!best || gen > bestgen) {
921 bestgen = gen;
922 best = sd;
923 }
924 } else {
925 free_imsm(super);
926 return 2;
927 }
928 }
929
930 if (!best) {
931 free_imsm(super);
932 return 1;
933 }
934
935 /* load the most up to date anchor */
936 sprintf(nm, "%d:%d", best->disk.major, best->disk.minor);
937 dfd = dev_open(nm, O_RDONLY);
938 if (!dfd) {
939 free_imsm(super);
940 return 1;
941 }
942 rv = load_imsm_mpb(dfd, super, NULL);
943 close(dfd);
944 if (rv != 0) {
945 free_imsm(super);
946 return 2;
947 }
948
949 /* reset the disk list */
950 free_imsm_disks(super);
951
952 /* populate disk list */
953 for (sd = sra->devs ; sd ; sd = sd->next) {
954 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
955 dfd = dev_open(nm, keep_fd? O_RDWR : O_RDONLY);
956 if (!dfd) {
957 free_imsm(super);
958 return 2;
959 }
960 load_imsm_disk(dfd, super, NULL, keep_fd);
961 if (!keep_fd)
962 close(dfd);
963 }
964
965 if (st->subarray[0]) {
966 if (atoi(st->subarray) <= super->mpb->num_raid_devs)
967 super->current_vol = atoi(st->subarray);
968 else
969 return 1;
970 }
971
972 *sbp = super;
973 if (st->ss == NULL) {
974 st->ss = &super_imsm;
975 st->minor_version = 0;
976 st->max_devs = IMSM_MAX_DEVICES;
977 st->container_dev = fd2devnum(fd);
978 }
979
980 return 0;
981 }
982 #endif
983
984 static int load_super_imsm(struct supertype *st, int fd, char *devname)
985 {
986 struct intel_super *super;
987 int rv;
988
989 #ifndef MDASSEMBLE
990 if (load_super_imsm_all(st, fd, &st->sb, devname, 1) == 0)
991 return 0;
992 #endif
993 if (st->subarray[0])
994 return 1; /* FIXME */
995
996 super = alloc_super(0);
997 if (!super) {
998 fprintf(stderr,
999 Name ": malloc of %zu failed.\n",
1000 sizeof(*super));
1001 return 1;
1002 }
1003
1004 rv = load_imsm_mpb(fd, super, devname);
1005
1006 if (rv) {
1007 if (devname)
1008 fprintf(stderr,
1009 Name ": Failed to load all information "
1010 "sections on %s\n", devname);
1011 free_imsm(super);
1012 return rv;
1013 }
1014
1015 st->sb = super;
1016 if (st->ss == NULL) {
1017 st->ss = &super_imsm;
1018 st->minor_version = 0;
1019 st->max_devs = IMSM_MAX_DEVICES;
1020 }
1021
1022 return 0;
1023 }
1024
1025 static __u16 info_to_blocks_per_strip(mdu_array_info_t *info)
1026 {
1027 if (info->level == 1)
1028 return 128;
1029 return info->chunk_size >> 9;
1030 }
1031
1032 static __u32 info_to_num_data_stripes(mdu_array_info_t *info)
1033 {
1034 __u32 num_stripes;
1035
1036 num_stripes = (info->size * 2) / info_to_blocks_per_strip(info);
1037 if (info->level == 1)
1038 num_stripes /= 2;
1039
1040 return num_stripes;
1041 }
1042
1043 static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info,
1044 unsigned long long size, char *name,
1045 char *homehost, int *uuid)
1046 {
1047 /* We are creating a volume inside a pre-existing container.
1048 * so st->sb is already set.
1049 */
1050 struct intel_super *super = st->sb;
1051 struct imsm_super *mpb = super->mpb;
1052 struct imsm_dev *dev;
1053 struct imsm_vol *vol;
1054 struct imsm_map *map;
1055 int idx = mpb->num_raid_devs;
1056 int i;
1057 unsigned long long array_blocks;
1058 __u32 offset = 0;
1059 size_t size_old, size_new;
1060
1061 if (mpb->num_raid_devs >= 2) {
1062 fprintf(stderr, Name": This imsm-container already has the "
1063 "maximum of 2 volumes\n");
1064 return 0;
1065 }
1066
1067 /* ensure the mpb is large enough for the new data */
1068 size_old = __le32_to_cpu(mpb->mpb_size);
1069 size_new = disks_to_mpb_size(info->nr_disks);
1070 if (size_new > size_old) {
1071 void *mpb_new;
1072 size_t size_round = ROUND_UP(size_new, 512);
1073
1074 if (posix_memalign(&mpb_new, 512, size_round) != 0) {
1075 fprintf(stderr, Name": could not allocate new mpb\n");
1076 return 0;
1077 }
1078 memcpy(mpb_new, mpb, size_old);
1079 free(mpb);
1080 mpb = mpb_new;
1081 super->mpb = mpb_new;
1082 mpb->mpb_size = __cpu_to_le32(size_new);
1083 memset(mpb_new + size_old, 0, size_round - size_old);
1084 }
1085 super->current_vol = idx;
1086 sprintf(st->subarray, "%d", idx);
1087 mpb->num_raid_devs++;
1088 dev = get_imsm_dev(mpb, idx);
1089 strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN);
1090 array_blocks = calc_array_size(info->level, info->raid_disks,
1091 info->layout, info->chunk_size,
1092 info->size*2);
1093 dev->size_low = __cpu_to_le32((__u32) array_blocks);
1094 dev->size_high = __cpu_to_le32((__u32) (array_blocks >> 32));
1095 dev->status = __cpu_to_le32(0);
1096 dev->reserved_blocks = __cpu_to_le32(0);
1097 vol = &dev->vol;
1098 vol->migr_state = 0;
1099 vol->migr_type = 0;
1100 vol->dirty = 0;
1101 for (i = 0; i < idx; i++) {
1102 struct imsm_dev *prev = get_imsm_dev(mpb, i);
1103 struct imsm_map *pmap = &prev->vol.map[0];
1104
1105 offset += __le32_to_cpu(pmap->blocks_per_member);
1106 offset += IMSM_RESERVED_SECTORS;
1107 }
1108 map = &vol->map[0];
1109 map->pba_of_lba0 = __cpu_to_le32(offset);
1110 map->blocks_per_member = __cpu_to_le32(info->size * 2);
1111 map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info));
1112 map->num_data_stripes = __cpu_to_le32(info_to_num_data_stripes(info));
1113 map->map_state = info->level ? IMSM_T_STATE_UNINITIALIZED :
1114 IMSM_T_STATE_NORMAL;
1115
1116 if (info->level == 1 && info->raid_disks > 2) {
1117 fprintf(stderr, Name": imsm does not support more than 2 disks"
1118 "in a raid1 volume\n");
1119 return 0;
1120 }
1121 if (info->level == 10)
1122 map->raid_level = 1;
1123 else
1124 map->raid_level = info->level;
1125
1126 map->num_members = info->raid_disks;
1127 for (i = 0; i < map->num_members; i++) {
1128 /* initialized in add_to_super */
1129 map->disk_ord_tbl[i] = __cpu_to_le32(0);
1130 }
1131
1132 return 1;
1133 }
1134
1135 static int init_super_imsm(struct supertype *st, mdu_array_info_t *info,
1136 unsigned long long size, char *name,
1137 char *homehost, int *uuid)
1138 {
1139 /* This is primarily called by Create when creating a new array.
1140 * We will then get add_to_super called for each component, and then
1141 * write_init_super called to write it out to each device.
1142 * For IMSM, Create can create on fresh devices or on a pre-existing
1143 * array.
1144 * To create on a pre-existing array a different method will be called.
1145 * This one is just for fresh drives.
1146 */
1147 struct intel_super *super;
1148 struct imsm_super *mpb;
1149 size_t mpb_size;
1150
1151 if (!info) {
1152 st->sb = NULL;
1153 return 0;
1154 }
1155 if (st->sb)
1156 return init_super_imsm_volume(st, info, size, name, homehost,
1157 uuid);
1158
1159 super = alloc_super(1);
1160 if (!super)
1161 return 0;
1162 mpb_size = disks_to_mpb_size(info->nr_disks);
1163 if (posix_memalign((void**)&mpb, 512, mpb_size) != 0) {
1164 free(super);
1165 return 0;
1166 }
1167 memset(mpb, 0, mpb_size);
1168
1169 memcpy(mpb->sig, MPB_SIGNATURE, strlen(MPB_SIGNATURE));
1170 memcpy(mpb->sig + strlen(MPB_SIGNATURE), MPB_VERSION_RAID5,
1171 strlen(MPB_VERSION_RAID5));
1172 mpb->mpb_size = mpb_size;
1173
1174 super->mpb = mpb;
1175 st->sb = super;
1176 return 1;
1177 }
1178
1179 static void add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk,
1180 int fd, char *devname)
1181 {
1182 struct intel_super *super = st->sb;
1183 struct imsm_super *mpb = super->mpb;
1184 struct dl *dl;
1185 struct imsm_dev *dev;
1186 struct imsm_map *map;
1187 struct imsm_disk *disk;
1188 __u32 status;
1189
1190 dev = get_imsm_dev(mpb, super->current_vol);
1191 map = &dev->vol.map[0];
1192
1193 for (dl = super->disks; dl ; dl = dl->next)
1194 if (dl->major == dk->major &&
1195 dl->minor == dk->minor)
1196 break;
1197 if (!dl || ! (dk->state & (1<<MD_DISK_SYNC)))
1198 return;
1199
1200 map->disk_ord_tbl[dk->number] = __cpu_to_le32(dl->index);
1201
1202 disk = get_imsm_disk(mpb, dl->index);
1203 status = CONFIGURED_DISK | USABLE_DISK;
1204 disk->status = __cpu_to_le32(status);
1205 }
1206
1207 static void add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk,
1208 int fd, char *devname)
1209 {
1210 struct intel_super *super = st->sb;
1211 struct imsm_super *mpb = super->mpb;
1212 struct imsm_disk *disk;
1213 struct dl *dd;
1214 unsigned long long size;
1215 __u32 status, id;
1216 int rv;
1217 struct stat stb;
1218
1219 if (super->current_vol >= 0) {
1220 add_to_super_imsm_volume(st, dk, fd, devname);
1221 return;
1222 }
1223
1224 fstat(fd, &stb);
1225 dd = malloc(sizeof(*dd));
1226 if (!dd) {
1227 fprintf(stderr,
1228 Name ": malloc failed %s:%d.\n", __func__, __LINE__);
1229 abort();
1230 }
1231 memset(dd, 0, sizeof(*dd));
1232 dd->major = major(stb.st_rdev);
1233 dd->minor = minor(stb.st_rdev);
1234 dd->index = dk->number;
1235 dd->devname = devname ? strdup(devname) : NULL;
1236 dd->next = super->disks;
1237 dd->fd = fd;
1238 rv = imsm_read_serial(fd, devname, dd->serial);
1239 if (rv) {
1240 fprintf(stderr,
1241 Name ": failed to retrieve scsi serial "
1242 "using \'%s\' instead\n", devname);
1243 strcpy((char *) dd->serial, devname);
1244 }
1245
1246 if (mpb->num_disks <= dk->number)
1247 mpb->num_disks = dk->number + 1;
1248
1249 disk = get_imsm_disk(mpb, dk->number);
1250 get_dev_size(fd, NULL, &size);
1251 size /= 512;
1252 status = USABLE_DISK | SPARE_DISK;
1253 strcpy((char *) disk->serial, (char *) dd->serial);
1254 disk->total_blocks = __cpu_to_le32(size);
1255 disk->status = __cpu_to_le32(status);
1256 if (sysfs_disk_to_scsi_id(fd, &id) == 0)
1257 disk->scsi_id = __cpu_to_le32(id);
1258 else
1259 disk->scsi_id = __cpu_to_le32(0);
1260
1261 /* update the family number if we are creating a container */
1262 if (super->creating_imsm)
1263 mpb->family_num = __cpu_to_le32(gen_imsm_checksum(mpb));
1264
1265 super->disks = dd;
1266 }
1267
1268 static int store_imsm_mpb(int fd, struct intel_super *super);
1269
1270 static int write_super_imsm(struct intel_super *super, int doclose)
1271 {
1272 struct imsm_super *mpb = super->mpb;
1273 struct dl *d;
1274 __u32 generation;
1275 __u32 sum;
1276
1277 /* 'generation' is incremented everytime the metadata is written */
1278 generation = __le32_to_cpu(mpb->generation_num);
1279 generation++;
1280 mpb->generation_num = __cpu_to_le32(generation);
1281
1282 /* recalculate checksum */
1283 sum = gen_imsm_checksum(mpb);
1284 mpb->check_sum = __cpu_to_le32(sum);
1285
1286 for (d = super->disks; d ; d = d->next) {
1287 if (store_imsm_mpb(d->fd, super)) {
1288 fprintf(stderr, "%s: failed for device %d:%d %s\n",
1289 __func__, d->major, d->minor, strerror(errno));
1290 return 0;
1291 }
1292 if (doclose) {
1293 close(d->fd);
1294 d->fd = -1;
1295 }
1296 }
1297
1298 return 1;
1299 }
1300
1301 static int write_init_super_imsm(struct supertype *st)
1302 {
1303 return write_super_imsm(st->sb, 1);
1304 }
1305
1306 static int store_zero_imsm(struct supertype *st, int fd)
1307 {
1308 unsigned long long dsize;
1309 void *buf;
1310
1311 get_dev_size(fd, NULL, &dsize);
1312
1313 /* first block is stored on second to last sector of the disk */
1314 if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0)
1315 return 1;
1316
1317 if (posix_memalign(&buf, 512, 512) != 0)
1318 return 1;
1319
1320 memset(buf, 0, sizeof(buf));
1321 if (write(fd, buf, sizeof(buf)) != sizeof(buf))
1322 return 1;
1323 return 0;
1324 }
1325
1326 static int validate_geometry_imsm_container(struct supertype *st, int level,
1327 int layout, int raiddisks, int chunk,
1328 unsigned long long size, char *dev,
1329 unsigned long long *freesize,
1330 int verbose)
1331 {
1332 int fd;
1333 unsigned long long ldsize;
1334
1335 if (level != LEVEL_CONTAINER)
1336 return 0;
1337 if (!dev)
1338 return 1;
1339
1340 fd = open(dev, O_RDONLY|O_EXCL, 0);
1341 if (fd < 0) {
1342 if (verbose)
1343 fprintf(stderr, Name ": imsm: Cannot open %s: %s\n",
1344 dev, strerror(errno));
1345 return 0;
1346 }
1347 if (!get_dev_size(fd, dev, &ldsize)) {
1348 close(fd);
1349 return 0;
1350 }
1351 close(fd);
1352
1353 *freesize = avail_size_imsm(st, ldsize >> 9);
1354
1355 return 1;
1356 }
1357
1358 /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd
1359 * FIX ME add ahci details
1360 */
1361 static int validate_geometry_imsm_volume(struct supertype *st, int level,
1362 int layout, int raiddisks, int chunk,
1363 unsigned long long size, char *dev,
1364 unsigned long long *freesize,
1365 int verbose)
1366 {
1367 struct stat stb;
1368 struct intel_super *super = st->sb;
1369 struct dl *dl;
1370 unsigned long long pos = 0;
1371 unsigned long long maxsize;
1372 struct extent *e;
1373 int i;
1374
1375 if (level == LEVEL_CONTAINER)
1376 return 0;
1377
1378 if (level == 1 && raiddisks > 2) {
1379 if (verbose)
1380 fprintf(stderr, Name ": imsm does not support more "
1381 "than 2 in a raid1 configuration\n");
1382 return 0;
1383 }
1384
1385 /* We must have the container info already read in. */
1386 if (!super)
1387 return 0;
1388
1389 if (!dev) {
1390 /* General test: make sure there is space for
1391 * 'raiddisks' device extents of size 'size'.
1392 */
1393 unsigned long long minsize = size*2 /* convert to blocks */;
1394 int dcnt = 0;
1395 if (minsize == 0)
1396 minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
1397 for (dl = super->disks; dl ; dl = dl->next) {
1398 int found = 0;
1399
1400 pos = 0;
1401 i = 0;
1402 e = get_extents(super, dl);
1403 if (!e) continue;
1404 do {
1405 unsigned long long esize;
1406 esize = e[i].start - pos;
1407 if (esize >= minsize)
1408 found = 1;
1409 pos = e[i].start + e[i].size;
1410 i++;
1411 } while (e[i-1].size);
1412 if (found)
1413 dcnt++;
1414 free(e);
1415 }
1416 if (dcnt < raiddisks) {
1417 if (verbose)
1418 fprintf(stderr, Name ": imsm: Not enough "
1419 "devices with space for this array "
1420 "(%d < %d)\n",
1421 dcnt, raiddisks);
1422 return 0;
1423 }
1424 return 1;
1425 }
1426 /* This device must be a member of the set */
1427 if (stat(dev, &stb) < 0)
1428 return 0;
1429 if ((S_IFMT & stb.st_mode) != S_IFBLK)
1430 return 0;
1431 for (dl = super->disks ; dl ; dl = dl->next) {
1432 if (dl->major == major(stb.st_rdev) &&
1433 dl->minor == minor(stb.st_rdev))
1434 break;
1435 }
1436 if (!dl) {
1437 if (verbose)
1438 fprintf(stderr, Name ": %s is not in the "
1439 "same imsm set\n", dev);
1440 return 0;
1441 }
1442 e = get_extents(super, dl);
1443 maxsize = 0;
1444 i = 0;
1445 if (e) do {
1446 unsigned long long esize;
1447 esize = e[i].start - pos;
1448 if (esize >= maxsize)
1449 maxsize = esize;
1450 pos = e[i].start + e[i].size;
1451 i++;
1452 } while (e[i-1].size);
1453 *freesize = maxsize;
1454
1455 return 1;
1456 }
1457
1458 static int validate_geometry_imsm(struct supertype *st, int level, int layout,
1459 int raiddisks, int chunk, unsigned long long size,
1460 char *dev, unsigned long long *freesize,
1461 int verbose)
1462 {
1463 int fd, cfd;
1464 struct mdinfo *sra;
1465
1466 /* if given unused devices create a container
1467 * if given given devices in a container create a member volume
1468 */
1469 if (level == LEVEL_CONTAINER) {
1470 /* Must be a fresh device to add to a container */
1471 return validate_geometry_imsm_container(st, level, layout,
1472 raiddisks, chunk, size,
1473 dev, freesize,
1474 verbose);
1475 }
1476
1477 if (st->sb) {
1478 /* creating in a given container */
1479 return validate_geometry_imsm_volume(st, level, layout,
1480 raiddisks, chunk, size,
1481 dev, freesize, verbose);
1482 }
1483
1484 /* limit creation to the following levels */
1485 if (!dev)
1486 switch (level) {
1487 case 0:
1488 case 1:
1489 case 10:
1490 case 5:
1491 break;
1492 default:
1493 return 1;
1494 }
1495
1496 /* This device needs to be a device in an 'imsm' container */
1497 fd = open(dev, O_RDONLY|O_EXCL, 0);
1498 if (fd >= 0) {
1499 if (verbose)
1500 fprintf(stderr,
1501 Name ": Cannot create this array on device %s\n",
1502 dev);
1503 close(fd);
1504 return 0;
1505 }
1506 if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) {
1507 if (verbose)
1508 fprintf(stderr, Name ": Cannot open %s: %s\n",
1509 dev, strerror(errno));
1510 return 0;
1511 }
1512 /* Well, it is in use by someone, maybe an 'imsm' container. */
1513 cfd = open_container(fd);
1514 if (cfd < 0) {
1515 close(fd);
1516 if (verbose)
1517 fprintf(stderr, Name ": Cannot use %s: It is busy\n",
1518 dev);
1519 return 0;
1520 }
1521 sra = sysfs_read(cfd, 0, GET_VERSION);
1522 close(fd);
1523 if (sra && sra->array.major_version == -1 &&
1524 strcmp(sra->text_version, "imsm") == 0) {
1525 /* This is a member of a imsm container. Load the container
1526 * and try to create a volume
1527 */
1528 struct intel_super *super;
1529
1530 if (load_super_imsm_all(st, cfd, (void **) &super, NULL, 1) == 0) {
1531 st->sb = super;
1532 st->container_dev = fd2devnum(cfd);
1533 close(cfd);
1534 return validate_geometry_imsm_volume(st, level, layout,
1535 raiddisks, chunk,
1536 size, dev,
1537 freesize, verbose);
1538 }
1539 close(cfd);
1540 } else /* may belong to another container */
1541 return 0;
1542
1543 return 1;
1544 }
1545
1546 static struct mdinfo *container_content_imsm(struct supertype *st)
1547 {
1548 /* Given a container loaded by load_super_imsm_all,
1549 * extract information about all the arrays into
1550 * an mdinfo tree.
1551 *
1552 * For each imsm_dev create an mdinfo, fill it in,
1553 * then look for matching devices in super->disks
1554 * and create appropriate device mdinfo.
1555 */
1556 struct intel_super *super = st->sb;
1557 struct imsm_super *mpb = super->mpb;
1558 struct mdinfo *rest = NULL;
1559 int i;
1560
1561 for (i = 0; i < mpb->num_raid_devs; i++) {
1562 struct imsm_dev *dev = get_imsm_dev(mpb, i);
1563 struct imsm_vol *vol = &dev->vol;
1564 struct imsm_map *map = vol->map;
1565 struct mdinfo *this;
1566 __u64 sz;
1567 int slot;
1568
1569 this = malloc(sizeof(*this));
1570 memset(this, 0, sizeof(*this));
1571 this->next = rest;
1572 rest = this;
1573
1574 this->array.level = get_imsm_raid_level(map);
1575 this->array.raid_disks = map->num_members;
1576 this->array.layout = imsm_level_to_layout(this->array.level);
1577 this->array.md_minor = -1;
1578 this->array.ctime = 0;
1579 this->array.utime = 0;
1580 this->array.chunk_size = __le16_to_cpu(map->blocks_per_strip) << 9;
1581 this->array.state = !vol->dirty;
1582 this->container_member = i;
1583 if (map->map_state == IMSM_T_STATE_UNINITIALIZED || dev->vol.dirty)
1584 this->resync_start = 0;
1585 else
1586 this->resync_start = ~0ULL;
1587
1588 strncpy(this->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN);
1589 this->name[MAX_RAID_SERIAL_LEN] = 0;
1590
1591 sprintf(this->text_version, "/%s/%d",
1592 devnum2devname(st->container_dev),
1593 this->container_member);
1594
1595 memset(this->uuid, 0, sizeof(this->uuid));
1596
1597 sz = __le32_to_cpu(dev->size_high);
1598 sz <<= 32;
1599 sz += __le32_to_cpu(dev->size_low);
1600 this->component_size = sz;
1601 this->array.size = this->component_size / 2;
1602
1603 for (slot = 0 ; slot < map->num_members; slot++) {
1604 struct imsm_disk *disk;
1605 struct mdinfo *info_d;
1606 struct dl *d;
1607 int idx;
1608 __u32 s;
1609
1610 idx = __le32_to_cpu(map->disk_ord_tbl[slot] & ~(0xff << 24));
1611 for (d = super->disks; d ; d = d->next)
1612 if (d->index == idx)
1613 break;
1614
1615 if (d == NULL)
1616 break; /* shouldn't this be continue ?? */
1617
1618 info_d = malloc(sizeof(*info_d));
1619 if (!info_d)
1620 break; /* ditto ?? */
1621 memset(info_d, 0, sizeof(*info_d));
1622 info_d->next = this->devs;
1623 this->devs = info_d;
1624
1625 disk = get_imsm_disk(mpb, idx);
1626 s = __le32_to_cpu(disk->status);
1627
1628 info_d->disk.number = d->index;
1629 info_d->disk.major = d->major;
1630 info_d->disk.minor = d->minor;
1631 info_d->disk.raid_disk = slot;
1632 info_d->disk.state = s & CONFIGURED_DISK ? (1 << MD_DISK_ACTIVE) : 0;
1633 info_d->disk.state |= s & FAILED_DISK ? (1 << MD_DISK_FAULTY) : 0;
1634 info_d->disk.state |= s & USABLE_DISK ? (1 << MD_DISK_SYNC) : 0;
1635
1636 this->array.working_disks++;
1637
1638 info_d->events = __le32_to_cpu(mpb->generation_num);
1639 info_d->data_offset = __le32_to_cpu(map->pba_of_lba0);
1640 info_d->component_size = __le32_to_cpu(map->blocks_per_member);
1641 if (d->devname)
1642 strcpy(info_d->name, d->devname);
1643 }
1644 }
1645
1646 return rest;
1647 }
1648
1649
1650 static int imsm_open_new(struct supertype *c, struct active_array *a,
1651 char *inst)
1652 {
1653 struct intel_super *super = c->sb;
1654 struct imsm_super *mpb = super->mpb;
1655
1656 if (atoi(inst) + 1 > mpb->num_raid_devs) {
1657 fprintf(stderr, "%s: subarry index %d, out of range\n",
1658 __func__, atoi(inst));
1659 return -ENODEV;
1660 }
1661
1662 dprintf("imsm: open_new %s\n", inst);
1663 a->info.container_member = atoi(inst);
1664 return 0;
1665 }
1666
1667 static __u8 imsm_check_degraded(struct imsm_super *mpb, int n, int failed)
1668 {
1669 struct imsm_dev *dev = get_imsm_dev(mpb, n);
1670 struct imsm_map *map = dev->vol.map;
1671
1672 if (!failed)
1673 return map->map_state;
1674
1675 switch (get_imsm_raid_level(map)) {
1676 case 0:
1677 return IMSM_T_STATE_FAILED;
1678 break;
1679 case 1:
1680 if (failed < map->num_members)
1681 return IMSM_T_STATE_DEGRADED;
1682 else
1683 return IMSM_T_STATE_FAILED;
1684 break;
1685 case 10:
1686 {
1687 /**
1688 * check to see if any mirrors have failed,
1689 * otherwise we are degraded
1690 */
1691 int device_per_mirror = 2; /* FIXME is this always the case?
1692 * and are they always adjacent?
1693 */
1694 int failed = 0;
1695 int i;
1696
1697 for (i = 0; i < map->num_members; i++) {
1698 int idx = get_imsm_disk_idx(map, i);
1699 struct imsm_disk *disk = get_imsm_disk(mpb, idx);
1700
1701 if (__le32_to_cpu(disk->status) & FAILED_DISK)
1702 failed++;
1703
1704 if (failed >= device_per_mirror)
1705 return IMSM_T_STATE_FAILED;
1706
1707 /* reset 'failed' for next mirror set */
1708 if (!((i + 1) % device_per_mirror))
1709 failed = 0;
1710 }
1711
1712 return IMSM_T_STATE_DEGRADED;
1713 }
1714 case 5:
1715 if (failed < 2)
1716 return IMSM_T_STATE_DEGRADED;
1717 else
1718 return IMSM_T_STATE_FAILED;
1719 break;
1720 default:
1721 break;
1722 }
1723
1724 return map->map_state;
1725 }
1726
1727 static int imsm_count_failed(struct imsm_super *mpb, struct imsm_map *map)
1728 {
1729 int i;
1730 int failed = 0;
1731 struct imsm_disk *disk;
1732
1733 for (i = 0; i < map->num_members; i++) {
1734 int idx = get_imsm_disk_idx(map, i);
1735
1736 disk = get_imsm_disk(mpb, idx);
1737 if (__le32_to_cpu(disk->status) & FAILED_DISK)
1738 failed++;
1739 }
1740
1741 return failed;
1742 }
1743
1744 static void imsm_set_array_state(struct active_array *a, int consistent)
1745 {
1746 int inst = a->info.container_member;
1747 struct intel_super *super = a->container->sb;
1748 struct imsm_dev *dev = get_imsm_dev(super->mpb, inst);
1749 struct imsm_map *map = &dev->vol.map[0];
1750 int dirty = !consistent;
1751 int failed;
1752 __u8 map_state;
1753
1754 if (a->resync_start == ~0ULL) {
1755 failed = imsm_count_failed(super->mpb, map);
1756 map_state = imsm_check_degraded(super->mpb, inst, failed);
1757 if (!failed)
1758 map_state = IMSM_T_STATE_NORMAL;
1759 if (map->map_state != map_state) {
1760 dprintf("imsm: map_state %d: %d\n",
1761 inst, map_state);
1762 map->map_state = map_state;
1763 super->updates_pending++;
1764 }
1765 }
1766
1767 if (dev->vol.dirty != dirty) {
1768 dprintf("imsm: mark '%s' (%llu)\n",
1769 dirty?"dirty":"clean", a->resync_start);
1770
1771 dev->vol.dirty = dirty;
1772 super->updates_pending++;
1773 }
1774 }
1775
1776 static void imsm_set_disk(struct active_array *a, int n, int state)
1777 {
1778 int inst = a->info.container_member;
1779 struct intel_super *super = a->container->sb;
1780 struct imsm_dev *dev = get_imsm_dev(super->mpb, inst);
1781 struct imsm_map *map = dev->vol.map;
1782 struct imsm_disk *disk;
1783 __u32 status;
1784 int failed = 0;
1785 int new_failure = 0;
1786
1787 if (n > map->num_members)
1788 fprintf(stderr, "imsm: set_disk %d out of range 0..%d\n",
1789 n, map->num_members - 1);
1790
1791 if (n < 0)
1792 return;
1793
1794 dprintf("imsm: set_disk %d:%x\n", n, state);
1795
1796 disk = get_imsm_disk(super->mpb, get_imsm_disk_idx(map, n));
1797
1798 /* check if we have seen this failure before */
1799 status = __le32_to_cpu(disk->status);
1800 if ((state & DS_FAULTY) && !(status & FAILED_DISK)) {
1801 status |= FAILED_DISK;
1802 disk->status = __cpu_to_le32(status);
1803 new_failure = 1;
1804 }
1805
1806 /**
1807 * the number of failures have changed, count up 'failed' to determine
1808 * degraded / failed status
1809 */
1810 if (new_failure && map->map_state != IMSM_T_STATE_FAILED)
1811 failed = imsm_count_failed(super->mpb, map);
1812
1813 if (failed)
1814 map->map_state = imsm_check_degraded(super->mpb, inst, failed);
1815
1816 if (new_failure)
1817 super->updates_pending++;
1818 }
1819
1820 static int store_imsm_mpb(int fd, struct intel_super *super)
1821 {
1822 struct imsm_super *mpb = super->mpb;
1823 __u32 mpb_size = __le32_to_cpu(mpb->mpb_size);
1824 unsigned long long dsize;
1825 unsigned long long sectors;
1826
1827 get_dev_size(fd, NULL, &dsize);
1828
1829 if (mpb_size > 512) {
1830 /* -1 to account for anchor */
1831 sectors = mpb_sectors(mpb) - 1;
1832
1833 /* write the extended mpb to the sectors preceeding the anchor */
1834 if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0)
1835 return 1;
1836
1837 if (write(fd, super->buf + 512, 512 * sectors) != 512 * sectors)
1838 return 1;
1839 }
1840
1841 /* first block is stored on second to last sector of the disk */
1842 if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0)
1843 return 1;
1844
1845 if (write(fd, super->buf, 512) != 512)
1846 return 1;
1847
1848 return 0;
1849 }
1850
1851 static void imsm_sync_metadata(struct supertype *container)
1852 {
1853 struct intel_super *super = container->sb;
1854
1855 if (!super->updates_pending)
1856 return;
1857
1858 write_super_imsm(super, 0);
1859
1860 super->updates_pending = 0;
1861 }
1862
1863 struct superswitch super_imsm = {
1864 #ifndef MDASSEMBLE
1865 .examine_super = examine_super_imsm,
1866 .brief_examine_super = brief_examine_super_imsm,
1867 .detail_super = detail_super_imsm,
1868 .brief_detail_super = brief_detail_super_imsm,
1869 .write_init_super = write_init_super_imsm,
1870 #endif
1871 .match_home = match_home_imsm,
1872 .uuid_from_super= uuid_from_super_imsm,
1873 .getinfo_super = getinfo_super_imsm,
1874 .update_super = update_super_imsm,
1875
1876 .avail_size = avail_size_imsm,
1877
1878 .compare_super = compare_super_imsm,
1879
1880 .load_super = load_super_imsm,
1881 .init_super = init_super_imsm,
1882 .add_to_super = add_to_super_imsm,
1883 .store_super = store_zero_imsm,
1884 .free_super = free_super_imsm,
1885 .match_metadata_desc = match_metadata_desc_imsm,
1886 .container_content = container_content_imsm,
1887
1888 .validate_geometry = validate_geometry_imsm,
1889 .external = 1,
1890
1891 /* for mdmon */
1892 .open_new = imsm_open_new,
1893 .load_super = load_super_imsm,
1894 .set_array_state= imsm_set_array_state,
1895 .set_disk = imsm_set_disk,
1896 .sync_metadata = imsm_sync_metadata,
1897 };
Unnamed repository; edit this file 'description' to name the repository.