]> git.ipfire.org Git - thirdparty/mdadm.git/blame - super-intel.c
projects / thirdparty / mdadm.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Add some comments to explain some of the bits of superswitch.
[thirdparty/mdadm.git] / super-intel.c
CommitLineData
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
39struct 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. */
54struct 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
75struct 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
86struct 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
97struct 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
114static char *map_state_str[] = { "normal", "uninitialized", "degraded", "failed" };
115#endif
116
87eb16df 117static unsigned int sector_count(__u32 bytes)
cdddbdbc 118{
87eb16df
DW		 119 return ((bytes + (512-1)) & (~(512-1))) / 512;
120}
cdddbdbc 121
87eb16df
DW		 122static unsigned int mpb_sectors(struct imsm_super *mpb)
123{
124 return sector_count(__le32_to_cpu(mpb->mpb_size));
cdddbdbc
DW		 125}
126
0063ecba
NB		 127static struct superswitch super_imsm_volume;
128
cdddbdbc
DW		 129/* internal representation of IMSM metadata */
130struct intel_super {
131 union {
132 struct imsm_super *mpb;
133 void *buf;
134 };
c2c087e6
DW		 135 int updates_pending; /* count of pending updates for mdmon */
136 int creating_imsm; /* flag to indicate container creation */
137 int creating_dev; /* index of raid device undergoing creation */
cdddbdbc
DW		 138 struct dl {
139 struct dl *next;
140 int index;
141 __u8 serial[MAX_RAID_SERIAL_LEN];
142 int major, minor;
143 char *devname;
144 int fd;
145 } *disks;
146};
147
c2c087e6
DW		 148struct extent {
149 unsigned long long start, size;
150};
151
cdddbdbc
DW		 152static struct supertype *match_metadata_desc_imsm(char *arg)
153{
154 struct supertype *st;
155
156 if (strcmp(arg, "imsm") != 0 &&
157 strcmp(arg, "default") != 0
158 )
159 return NULL;
160
161 st = malloc(sizeof(*st));
ef609477 162 memset(st, 0, sizeof(*st));
cdddbdbc
DW		 163 st->ss = &super_imsm;
164 st->max_devs = IMSM_MAX_DEVICES;
165 st->minor_version = 0;
166 st->sb = NULL;
167 return st;
168}
169
8b353278 170static struct supertype *match_metadata_desc_imsm_volume(char *arg)
cdddbdbc
DW		 171{
172 struct supertype *st;
173
8b353278 174 if (strcmp(arg, "imsm/volume") != 0 &&
cdddbdbc
DW		 175 strcmp(arg, "raid") != 0 &&
176 strcmp(arg, "default") != 0
177 )
178 return NULL;
179
180 st = malloc(sizeof(*st));
ef609477 181 memset(st, 0, sizeof(*st));
8b353278 182 st->ss = &super_imsm_volume;
cdddbdbc
DW		 183 st->max_devs = IMSM_MAX_DEVICES;
184 st->minor_version = 0;
185 st->sb = NULL;
186 return st;
187}
188
189static __u8 *get_imsm_version(struct imsm_super *mpb)
190{
191 return &mpb->sig[MPB_SIG_LEN];
192}
193
194static struct imsm_disk *get_imsm_disk(struct imsm_super *mpb, __u8 index)
195{
196 if (index > mpb->num_disks - 1)
197 return NULL;
198 return &mpb->disk[index];
199}
200
201static __u32 gen_imsm_checksum(struct imsm_super *mpb)
202{
203 __u32 end = mpb->mpb_size / sizeof(end);
204 __u32 *p = (__u32 *) mpb;
205 __u32 sum = 0;
206
207 while (end--)
208 sum += __le32_to_cpu(*p++);
209
210 return sum - __le32_to_cpu(mpb->check_sum);
211}
212
213static size_t sizeof_imsm_dev(struct imsm_dev *dev)
214{
215 size_t size = sizeof(*dev);
216
217 /* each map has disk_ord_tbl[num_members - 1] additional space */
218 size += sizeof(__u32) * (dev->vol.map[0].num_members - 1);
219
220 /* migrating means an additional map */
221 if (dev->vol.migr_state) {
222 size += sizeof(struct imsm_map);
223 size += sizeof(__u32) * (dev->vol.map[1].num_members - 1);
224 }
225
226 return size;
227}
228
229static struct imsm_dev *get_imsm_dev(struct imsm_super *mpb, __u8 index)
230{
231 int offset;
232 int i;
233 void *_mpb = mpb;
234
235 if (index > mpb->num_raid_devs - 1)
236 return NULL;
237
238 /* devices start after all disks */
239 offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb;
240
241 for (i = 0; i <= index; i++)
242 if (i == index)
243 return _mpb + offset;
244 else
245 offset += sizeof_imsm_dev(_mpb + offset);
246
247 return NULL;
248}
249
250static __u32 get_imsm_disk_idx(struct imsm_map *map, int slot)
251{
252 __u32 *ord_tbl = &map->disk_ord_tbl[slot];
253
254 /* top byte is 'special' */
255 return __le32_to_cpu(*ord_tbl & ~(0xff << 24));
256}
257
cdddbdbc
DW		 258static int get_imsm_raid_level(struct imsm_map *map)
259{
260 if (map->raid_level == 1) {
261 if (map->num_members == 2)
262 return 1;
263 else
264 return 10;
265 }
266
267 return map->raid_level;
268}
269
c2c087e6
DW		 270static int cmp_extent(const void *av, const void *bv)
271{
272 const struct extent *a = av;
273 const struct extent *b = bv;
274 if (a->start < b->start)
275 return -1;
276 if (a->start > b->start)
277 return 1;
278 return 0;
279}
280
281static struct extent *get_extents(struct intel_super *super, struct dl *dl)
282{
283 /* find a list of used extents on the given physical device */
284 struct imsm_super *mpb = super->mpb;
285 struct imsm_disk *disk;
286 struct extent *rv, *e;
287 int i, j;
288 int memberships = 0;
289
290 disk = get_imsm_disk(mpb, dl->index);
291 if (!disk)
292 return NULL;
293
294 for (i = 0; i < mpb->num_raid_devs; i++) {
295 struct imsm_dev *dev = get_imsm_dev(mpb, i);
296 struct imsm_map *map = dev->vol.map;
297
298 for (j = 0; j < map->num_members; j++) {
299 __u32 index = get_imsm_disk_idx(map, j);
300
301 if (index == dl->index)
302 memberships++;
303 }
304 }
305 rv = malloc(sizeof(struct extent) * (memberships + 1));
306 if (!rv)
307 return NULL;
308 e = rv;
309
310 for (i = 0; i < mpb->num_raid_devs; i++) {
311 struct imsm_dev *dev = get_imsm_dev(mpb, i);
312 struct imsm_map *map = dev->vol.map;
313
314 for (j = 0; j < map->num_members; j++) {
315 __u32 index = get_imsm_disk_idx(map, j);
316
317 if (index == dl->index) {
318 e->start = __le32_to_cpu(map->pba_of_lba0);
319 e->size = __le32_to_cpu(map->blocks_per_member);
320 e++;
321 }
322 }
323 }
324 qsort(rv, memberships, sizeof(*rv), cmp_extent);
325
326 e->start = __le32_to_cpu(disk->total_blocks) -
327 (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS);
328 e->size = 0;
329 return rv;
330}
331
4f5bc454 332#ifndef MDASSEMBLE
cdddbdbc
DW		 333static void print_imsm_dev(struct imsm_dev *dev, int index)
334{
335 __u64 sz;
336 int slot;
337 struct imsm_map *map = dev->vol.map;
338
339 printf("\n");
340 printf("[%s]:\n", dev->volume);
341 printf(" RAID Level : %d\n", get_imsm_raid_level(map));
342 printf(" Members : %d\n", map->num_members);
343 for (slot = 0; slot < map->num_members; slot++)
344 if (index == get_imsm_disk_idx(map, slot))
345 break;
346 if (slot < map->num_members)
347 printf(" This Slot : %d\n", slot);
348 else
349 printf(" This Slot : ?\n");
350 sz = __le32_to_cpu(dev->size_high);
351 sz <<= 32;
352 sz += __le32_to_cpu(dev->size_low);
353 printf(" Array Size : %llu%s\n", (unsigned long long)sz,
354 human_size(sz * 512));
355 sz = __le32_to_cpu(map->blocks_per_member);
356 printf(" Per Dev Size : %llu%s\n", (unsigned long long)sz,
357 human_size(sz * 512));
358 printf(" Sector Offset : %u\n",
359 __le32_to_cpu(map->pba_of_lba0));
360 printf(" Num Stripes : %u\n",
361 __le32_to_cpu(map->num_data_stripes));
362 printf(" Chunk Size : %u KiB\n",
363 __le16_to_cpu(map->blocks_per_strip) / 2);
364 printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks));
365 printf(" Migrate State : %s\n", dev->vol.migr_state ? "migrating" : "idle");
366 printf(" Dirty State : %s\n", dev->vol.dirty ? "dirty" : "clean");
367 printf(" Map State : %s\n", map_state_str[map->map_state]);
368}
369
370static void print_imsm_disk(struct imsm_super *mpb, int index)
371{
372 struct imsm_disk *disk = get_imsm_disk(mpb, index);
373 char str[MAX_RAID_SERIAL_LEN];
374 __u32 s;
375 __u64 sz;
376
377 printf("\n");
378 snprintf(str, MAX_RAID_SERIAL_LEN, "%s", disk->serial);
379 printf(" Disk%02d Serial : %s\n", index, str);
380 s = __le32_to_cpu(disk->status);
381 printf(" State :%s%s%s%s\n", s&SPARE_DISK ? " spare" : "",
382 s&CONFIGURED_DISK ? " active" : "",
383 s&FAILED_DISK ? " failed" : "",
384 s&USABLE_DISK ? " usable" : "");
385 printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id));
c2c087e6
DW		 386 sz = __le32_to_cpu(disk->total_blocks) -
387 (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS * mpb->num_raid_devs);
cdddbdbc
DW		 388 printf(" Usable Size : %llu%s\n", (unsigned long long)sz,
389 human_size(sz * 512));
390}
391
392static void examine_super_imsm(struct supertype *st, char *homehost)
393{
394 struct intel_super *super = st->sb;
395 struct imsm_super *mpb = super->mpb;
396 char str[MAX_SIGNATURE_LENGTH];
397 int i;
398 __u32 sum;
399
400 snprintf(str, MPB_SIG_LEN, "%s", mpb->sig);
401 printf(" Magic : %s\n", str);
402 snprintf(str, strlen(MPB_VERSION_RAID0), "%s", get_imsm_version(mpb));
403 printf(" Version : %s\n", get_imsm_version(mpb));
404 printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num));
405 printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num));
406 sum = __le32_to_cpu(mpb->check_sum);
407 printf(" Checksum : %08x %s\n", sum,
408 gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect");
87eb16df 409 printf(" MPB Sectors : %d\n", mpb_sectors(mpb));
cdddbdbc
DW		 410 printf(" Disks : %d\n", mpb->num_disks);
411 printf(" RAID Devices : %d\n", mpb->num_raid_devs);
412 print_imsm_disk(mpb, super->disks->index);
413 for (i = 0; i < mpb->num_raid_devs; i++)
414 print_imsm_dev(get_imsm_dev(mpb, i), super->disks->index);
415 for (i = 0; i < mpb->num_disks; i++) {
416 if (i == super->disks->index)
417 continue;
418 print_imsm_disk(mpb, i);
419 }
420}
421
422static void brief_examine_super_imsm(struct supertype *st)
423{
424 struct intel_super *super = st->sb;
425 struct imsm_super *mpb = super->mpb;
426
427 printf("ARRAY /dev/imsm family=%08x metadata=external:imsm\n",
428 __le32_to_cpu(mpb->family_num));
429}
430
431static void detail_super_imsm(struct supertype *st, char *homehost)
432{
433 printf("%s\n", __FUNCTION__);
434}
435
436static void brief_detail_super_imsm(struct supertype *st)
437{
438 printf("%s\n", __FUNCTION__);
439}
440#endif
441
442static int match_home_imsm(struct supertype *st, char *homehost)
443{
444 printf("%s\n", __FUNCTION__);
445
446 return 0;
447}
448
449static void uuid_from_super_imsm(struct supertype *st, int uuid[4])
450{
451 printf("%s\n", __FUNCTION__);
452}
453
4f5bc454
DW		 454static void
455get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p)
cdddbdbc 456{
cdddbdbc
DW		 457 __u8 *v = get_imsm_version(mpb);
458 __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH;
459 char major[] = { 0, 0, 0 };
460 char minor[] = { 0 ,0, 0 };
461 char patch[] = { 0, 0, 0 };
462 char *ver_parse[] = { major, minor, patch };
463 int i, j;
464
465 i = j = 0;
466 while (*v != '\0' && v < end) {
467 if (*v != '.' && j < 2)
468 ver_parse[i][j++] = *v;
469 else {
470 i++;
471 j = 0;
472 }
473 v++;
474 }
475
4f5bc454
DW		 476 *m = strtol(minor, NULL, 0);
477 *p = strtol(patch, NULL, 0);
478}
479
c2c087e6
DW		 480static int imsm_level_to_layout(int level)
481{
482 switch (level) {
483 case 0:
484 case 1:
485 return 0;
486 case 5:
487 case 6:
488 return ALGORITHM_LEFT_SYMMETRIC;
489 case 10:
490 return 0x102; //FIXME is this correct?
491 }
492 return -1;
493}
494
4f5bc454
DW		 495static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info)
496{
497 struct intel_super *super = st->sb;
498 struct imsm_super *mpb = super->mpb;
499 struct imsm_disk *disk;
500 __u32 s;
4f5bc454 501
cdddbdbc 502 info->array.major_version = 2000;
4f5bc454
DW		 503 get_imsm_numerical_version(mpb, &info->array.minor_version,
504 &info->array.patch_version);
cdddbdbc
DW		 505 info->array.raid_disks = mpb->num_disks;
506 info->array.level = LEVEL_CONTAINER;
507 info->array.layout = 0;
508 info->array.md_minor = -1;
c2c087e6 509 info->array.ctime = 0; /* N/A for imsm */
cdddbdbc
DW		 510 info->array.utime = 0;
511 info->array.chunk_size = 0;
512
513 info->disk.major = 0;
514 info->disk.minor = 0;
cdddbdbc 515 info->disk.raid_disk = -1;
c2c087e6
DW		 516 info->reshape_active = 0;
517 strcpy(info->text_version, "imsm");
518 info->disk.number = -1;
519 info->disk.state = 0;
520
4a04ec6c
DW		 521 if (super->disks) {
522 info->disk.number = super->disks->index;
523 info->disk.raid_disk = super->disks->index;
524 disk = get_imsm_disk(mpb, super->disks->index);
525 s = __le32_to_cpu(disk->status);
526 info->disk.state = s & CONFIGURED_DISK ? (1 << MD_DISK_ACTIVE) : 0;
527 info->disk.state |= s & FAILED_DISK ? (1 << MD_DISK_FAULTY) : 0;
528 info->disk.state |= s & USABLE_DISK ? (1 << MD_DISK_SYNC) : 0;
cdddbdbc 529 }
cdddbdbc
DW		 530}
531
8b353278 532static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info)
cdddbdbc 533{
c2c087e6
DW		 534 struct intel_super *super = st->sb;
535 struct imsm_super *mpb = super->mpb;
536 struct imsm_dev *dev = get_imsm_dev(mpb, info->container_member);
537 struct imsm_map *map = &dev->vol.map[0];
538
539 info->array.major_version = 2000;
540 get_imsm_numerical_version(mpb, &info->array.minor_version,
541 &info->array.patch_version);
542 info->array.raid_disks = map->num_members;
543 info->array.level = get_imsm_raid_level(map);
544 info->array.layout = imsm_level_to_layout(info->array.level);
545 info->array.md_minor = -1;
546 info->array.ctime = 0;
547 info->array.utime = 0;
548 info->array.chunk_size = __le16_to_cpu(map->blocks_per_strip * 512);
549
550 info->data_offset = __le32_to_cpu(map->pba_of_lba0);
551 info->component_size = __le32_to_cpu(map->blocks_per_member);
552
553 info->disk.major = 0;
554 info->disk.minor = 0;
159c3a1a
NB		 555
556 sprintf(info->text_version, "/%s/%d",
557 devnum2devname(st->container_dev),
c2c087e6 558 info->container_member);
cdddbdbc
DW		 559}
560
561static int update_super_imsm(struct supertype *st, struct mdinfo *info,
562 char *update, char *devname, int verbose,
563 int uuid_set, char *homehost)
564{
f352c545
DW		 565 /* FIXME */
566
567 /* For 'assemble' and 'force' we need to return non-zero if any
568 * change was made. For others, the return value is ignored.
569 * Update options are:
570 * force-one : This device looks a bit old but needs to be included,
571 * update age info appropriately.
572 * assemble: clear any 'faulty' flag to allow this device to
573 * be assembled.
574 * force-array: Array is degraded but being forced, mark it clean
575 * if that will be needed to assemble it.
576 *
577 * newdev: not used ????
578 * grow: Array has gained a new device - this is currently for
579 * linear only
580 * resync: mark as dirty so a resync will happen.
581 * name: update the name - preserving the homehost
582 *
583 * Following are not relevant for this imsm:
584 * sparc2.2 : update from old dodgey metadata
585 * super-minor: change the preferred_minor number
586 * summaries: update redundant counters.
587 * uuid: Change the uuid of the array to match watch is given
588 * homehost: update the recorded homehost
589 * _reshape_progress: record new reshape_progress position.
590 */
591 int rv = 0;
592 //struct intel_super *super = st->sb;
593 //struct imsm_super *mpb = super->mpb;
594
595 if (strcmp(update, "grow") == 0) {
596 }
597 if (strcmp(update, "resync") == 0) {
598 /* dev->vol.dirty = 1; */
599 }
600
601 /* IMSM has no concept of UUID or homehost */
602
603 return rv;
cdddbdbc
DW		 604}
605
c2c087e6 606static size_t disks_to_mpb_size(int disks)
cdddbdbc 607{
c2c087e6 608 size_t size;
cdddbdbc 609
c2c087e6
DW		 610 size = sizeof(struct imsm_super);
611 size += (disks - 1) * sizeof(struct imsm_disk);
612 size += 2 * sizeof(struct imsm_dev);
613 /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */
614 size += (4 - 2) * sizeof(struct imsm_map);
615 /* 4 possible disk_ord_tbl's */
616 size += 4 * (disks - 1) * sizeof(__u32);
617
618 return size;
619}
620
621static __u64 avail_size_imsm(struct supertype *st, __u64 devsize)
622{
623 if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS))
624 return 0;
625
626 return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS);
cdddbdbc
DW		 627}
628
629static int compare_super_imsm(struct supertype *st, struct supertype *tst)
630{
631 /*
632 * return:
633 * 0 same, or first was empty, and second was copied
634 * 1 second had wrong number
635 * 2 wrong uuid
636 * 3 wrong other info
637 */
638 struct intel_super *first = st->sb;
639 struct intel_super *sec = tst->sb;
640
641 if (!first) {
642 st->sb = tst->sb;
643 tst->sb = NULL;
644 return 0;
645 }
646
647 if (memcmp(first->mpb->sig, sec->mpb->sig, MAX_SIGNATURE_LENGTH) != 0)
648 return 3;
649 if (first->mpb->family_num != sec->mpb->family_num)
650 return 3;
651 if (first->mpb->mpb_size != sec->mpb->mpb_size)
652 return 3;
653 if (first->mpb->check_sum != sec->mpb->check_sum)
654 return 3;
655
656 return 0;
657}
658
659extern int scsi_get_serial(int fd, void *buf, size_t buf_len);
660
661static int imsm_read_serial(int fd, char *devname,
662 __u8 serial[MAX_RAID_SERIAL_LEN])
663{
664 unsigned char scsi_serial[255];
cdddbdbc 665 int sg_fd;
cdddbdbc
DW		 666 int rv;
667 int rsp_len;
668 int i, cnt;
669
670 memset(scsi_serial, 0, sizeof(scsi_serial));
cdddbdbc 671
90c8b707 672 sg_fd = sysfs_disk_to_sg(fd);
cdddbdbc
DW		 673 if (sg_fd < 0) {
674 if (devname)
675 fprintf(stderr,
90c8b707
DW		 676 Name ": Failed to open sg interface for %s: %s\n",
677 devname, strerror(errno));
cdddbdbc
DW		 678 return 1;
679 }
680
681 rv = scsi_get_serial(sg_fd, scsi_serial, sizeof(scsi_serial));
682 close(sg_fd);
683
684 if (rv != 0) {
685 if (devname)
686 fprintf(stderr,
687 Name ": Failed to retrieve serial for %s\n",
688 devname);
689 return rv;
690 }
691
692 rsp_len = scsi_serial[3];
693 for (i = 0, cnt = 0; i < rsp_len; i++) {
694 if (!isspace(scsi_serial[4 + i]))
695 serial[cnt++] = scsi_serial[4 + i];
696 if (cnt == MAX_RAID_SERIAL_LEN)
697 break;
698 }
699
700 serial[MAX_RAID_SERIAL_LEN - 1] = '\0';
701
702 return 0;
703}
704
705static int
706load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd)
707{
708 struct imsm_super *mpb = super->mpb;
709 struct dl *dl;
710 struct stat stb;
711 struct imsm_disk *disk;
712 int rv;
713 int i;
714
715 dl = malloc(sizeof(*dl));
716 if (!dl) {
717 if (devname)
718 fprintf(stderr,
719 Name ": failed to allocate disk buffer for %s\n",
720 devname);
721 return 2;
722 }
723 memset(dl, 0, sizeof(*dl));
724
725 fstat(fd, &stb);
726 dl->major = major(stb.st_rdev);
727 dl->minor = minor(stb.st_rdev);
728 dl->next = super->disks;
729 dl->fd = keep_fd ? fd : -1;
730 dl->devname = devname ? strdup(devname) : NULL;
731 dl->index = -1;
732 super->disks = dl;
733 rv = imsm_read_serial(fd, devname, dl->serial);
734
735 if (rv != 0)
736 return 2;
737
738 /* look up this disk's index */
739 for (i = 0; i < mpb->num_disks; i++) {
740 disk = get_imsm_disk(mpb, i);
741
742 if (memcmp(disk->serial, dl->serial, MAX_RAID_SERIAL_LEN) == 0)
743 break;
744 }
745
746 if (i > mpb->num_disks)
747 return 2;
748
749 dl->index = i;
750
751 return 0;
752}
753
754/* load_imsm_mpb - read matrix metadata
755 * allocates super->mpb to be freed by free_super
756 */
757static int load_imsm_mpb(int fd, struct intel_super *super, char *devname)
758{
759 unsigned long long dsize;
760 size_t len, mpb_size;
761 unsigned long long sectors;
762 struct stat;
763 struct imsm_super anchor;
764 __u32 check_sum;
765
766 memset(super, 0, sizeof(*super));
767 get_dev_size(fd, NULL, &dsize);
768
769 if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) {
770 if (devname)
771 fprintf(stderr,
772 Name ": Cannot seek to anchor block on %s: %s\n",
773 devname, strerror(errno));
774 return 1;
775 }
776
777 len = sizeof(anchor);
778 if (read(fd, &anchor, len) != len) {
779 if (devname)
780 fprintf(stderr,
781 Name ": Cannot read anchor block on %s: %s\n",
782 devname, strerror(errno));
783 return 1;
784 }
785
786 if (strncmp((char *) anchor.sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) {
787 if (devname)
788 fprintf(stderr,
789 Name ": no IMSM anchor on %s\n", devname);
790 return 2;
791 }
792
793 mpb_size = __le32_to_cpu(anchor.mpb_size);
794 super->mpb = malloc(mpb_size < 512 ? 512 : mpb_size);
795 if (!super->mpb) {
796 if (devname)
797 fprintf(stderr,
798 Name ": unable to allocate %zu byte mpb buffer\n",
799 mpb_size);
800 return 2;
801 }
802 memcpy(super->buf, &anchor, sizeof(anchor));
803
804 /* read the rest of the first block */
805 len = 512 - sizeof(anchor);
806 if (read(fd, super->buf + sizeof(anchor), len) != len) {
807 if (devname)
808 fprintf(stderr,
809 Name ": Cannot read anchor remainder on %s: %s\n",
810 devname, strerror(errno));
811 return 2;
812 }
813
814 sectors = mpb_sectors(&anchor) - 1;
815 if (!sectors)
816 return load_imsm_disk(fd, super, devname, 0);
817
818 /* read the extended mpb */
819 if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) {
820 if (devname)
821 fprintf(stderr,
822 Name ": Cannot seek to extended mpb on %s: %s\n",
823 devname, strerror(errno));
824 return 1;
825 }
826
827 len = mpb_size - 512;
828 if (read(fd, super->buf + 512, len) != len) {
829 if (devname)
830 fprintf(stderr,
831 Name ": Cannot read extended mpb on %s: %s\n",
832 devname, strerror(errno));
833 return 2;
834 }
835
836 check_sum = gen_imsm_checksum(super->mpb);
837 if (check_sum != __le32_to_cpu(super->mpb->check_sum)) {
838 if (devname)
839 fprintf(stderr,
840 Name ": IMSM checksum %x != %x on %s\n",
841 check_sum, __le32_to_cpu(super->mpb->check_sum),
842 devname);
843 return 2;
844 }
845
846 return load_imsm_disk(fd, super, devname, 0);
847}
848
849struct superswitch super_imsm_container;
850
851static void free_imsm_disks(struct intel_super *super)
852{
853 while (super->disks) {
854 struct dl *d = super->disks;
855
856 super->disks = d->next;
857 if (d->fd >= 0)
858 close(d->fd);
859 if (d->devname)
860 free(d->devname);
861 free(d);
862 }
863}
864
865static void free_imsm(struct intel_super *super)
866{
867 if (super->mpb)
868 free(super->mpb);
869 free_imsm_disks(super);
870 free(super);
871}
872
873
874static void free_super_imsm(struct supertype *st)
875{
876 struct intel_super *super = st->sb;
877
878 if (!super)
879 return;
880
881 free_imsm(super);
882 st->sb = NULL;
883}
884
c2c087e6
DW		 885static struct intel_super *alloc_super(int creating_imsm)
886{
887 struct intel_super *super = malloc(sizeof(*super));
888
889 if (super) {
890 memset(super, 0, sizeof(*super));
891 super->creating_imsm = creating_imsm;
892 super->creating_dev = -1;
893 }
894
895 return super;
896}
897
cdddbdbc
DW		 898#ifndef MDASSEMBLE
899static int load_super_imsm_all(struct supertype *st, int fd, void **sbp,
900 char *devname, int keep_fd)
901{
902 struct mdinfo *sra;
903 struct intel_super *super;
904 struct mdinfo *sd, *best = NULL;
905 __u32 bestgen = 0;
906 __u32 gen;
907 char nm[20];
908 int dfd;
909 int rv;
910
911 /* check if this disk is a member of an active array */
912 sra = sysfs_read(fd, 0, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE);
913 if (!sra)
914 return 1;
915
916 if (sra->array.major_version != -1 ||
917 sra->array.minor_version != -2 ||
918 strcmp(sra->text_version, "imsm") != 0)
919 return 1;
920
c2c087e6 921 super = alloc_super(0);
cdddbdbc
DW		 922 if (!super)
923 return 1;
924
925 /* find the most up to date disk in this array */
926 for (sd = sra->devs; sd; sd = sd->next) {
927 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
928 dfd = dev_open(nm, keep_fd ? O_RDWR : O_RDONLY);
929 if (!dfd) {
930 free_imsm(super);
931 return 2;
932 }
933 rv = load_imsm_mpb(dfd, super, NULL);
934 if (!keep_fd)
935 close(dfd);
936 if (rv == 0) {
937 gen = __le32_to_cpu(super->mpb->generation_num);
938 if (!best || gen > bestgen) {
939 bestgen = gen;
940 best = sd;
941 }
942 } else {
943 free_imsm(super);
944 return 2;
945 }
946 }
947
948 if (!best) {
949 free_imsm(super);
950 return 1;
951 }
952
953 /* load the most up to date anchor */
954 sprintf(nm, "%d:%d", best->disk.major, best->disk.minor);
955 dfd = dev_open(nm, O_RDONLY);
956 if (!dfd) {
957 free_imsm(super);
958 return 1;
959 }
960 rv = load_imsm_mpb(dfd, super, NULL);
961 close(dfd);
962 if (rv != 0) {
963 free_imsm(super);
964 return 2;
965 }
966
967 /* reset the disk list */
968 free_imsm_disks(super);
969
970 /* populate disk list */
971 for (sd = sra->devs ; sd ; sd = sd->next) {
972 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
973 dfd = dev_open(nm, keep_fd? O_RDWR : O_RDONLY);
974 if (!dfd) {
975 free_imsm(super);
976 return 2;
977 }
978 load_imsm_disk(dfd, super, NULL, keep_fd);
979 if (!keep_fd)
980 close(dfd);
981 }
982
983 *sbp = super;
984 if (st->ss == NULL) {
985 st->ss = &super_imsm_container;
986 st->minor_version = 0;
987 st->max_devs = IMSM_MAX_DEVICES;
f4d11639 988 st->container_dev = fd2devnum(fd);
cdddbdbc
DW		 989 }
990
991 return 0;
992}
993#endif
994
995static int load_super_imsm(struct supertype *st, int fd, char *devname)
996{
997 struct intel_super *super;
998 int rv;
999
1000#ifndef MDASSEMBLE
3dbccbcf 1001 if (load_super_imsm_all(st, fd, &st->sb, devname, 1) == 0)
cdddbdbc
DW		 1002 return 0;
1003#endif
1004
c2c087e6 1005 super = alloc_super(0);
cdddbdbc
DW		 1006 if (!super) {
1007 fprintf(stderr,
1008 Name ": malloc of %zu failed.\n",
1009 sizeof(*super));
1010 return 1;
1011 }
1012
1013 rv = load_imsm_mpb(fd, super, devname);
1014
1015 if (rv) {
1016 if (devname)
1017 fprintf(stderr,
1018 Name ": Failed to load all information "
1019 "sections on %s\n", devname);
1020 free_imsm(super);
1021 return rv;
1022 }
1023
1024 st->sb = super;
1025 if (st->ss == NULL) {
1026 st->ss = &super_imsm;
1027 st->minor_version = 0;
1028 st->max_devs = IMSM_MAX_DEVICES;
1029 }
1030
1031 return 0;
1032}
1033
1034static int init_zero_imsm(struct supertype *st, mdu_array_info_t *info,
1035 unsigned long long size, char *name,
1036 char *homehost, int *uuid)
1037{
551c80c1 1038 st->sb = NULL;
cdddbdbc
DW		 1039 return 0;
1040}
1041
1042static int init_super_imsm(struct supertype *st, mdu_array_info_t *info,
1043 unsigned long long size, char *name,
1044 char *homehost, int *uuid)
1045{
c2c087e6
DW		 1046 /* This is primarily called by Create when creating a new array.
1047 * We will then get add_to_super called for each component, and then
1048 * write_init_super called to write it out to each device.
1049 * For IMSM, Create can create on fresh devices or on a pre-existing
1050 * array.
1051 * To create on a pre-existing array a different method will be called.
1052 * This one is just for fresh drives.
1053 */
1054 struct intel_super *super;
1055 struct imsm_super *mpb;
1056 size_t mpb_size;
cdddbdbc 1057
c2c087e6
DW		 1058 super = alloc_super(1);
1059 if (!super)
1060 return 0;
1061 mpb_size = disks_to_mpb_size(info->nr_disks);
1062 mpb = malloc(mpb_size);
1063 if (!mpb) {
1064 free(super);
1065 return 0;
1066 }
1067 memset(mpb, 0, mpb_size);
1068
1069 memcpy(mpb->sig, MPB_SIGNATURE, strlen(MPB_SIGNATURE));
1070 memcpy(mpb->sig + strlen(MPB_SIGNATURE), MPB_VERSION_RAID5,
1071 strlen(MPB_VERSION_RAID5));
1072 mpb->mpb_size = mpb_size;
1073
1074 super->mpb = mpb;
1075 st->sb = super;
1076 return 1;
cdddbdbc
DW		 1077}
1078
8b353278
DW		 1079static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info,
1080 unsigned long long size, char *name,
1081 char *homehost, int *uuid)
cdddbdbc 1082{
c2c087e6
DW		 1083 /* We are creating a volume inside a pre-existing container.
1084 * so st->sb is already set.
1085 */
1086 struct intel_super *super = st->sb;
1087 struct imsm_super *mpb = super->mpb;
1088 struct imsm_dev *dev;
1089 struct imsm_vol *vol;
1090 struct imsm_map *map;
1091 int idx = mpb->num_raid_devs;
1092 int i;
1093 unsigned long long array_blocks;
1094 unsigned long long sz;
1095 __u32 offset = 0;
cdddbdbc 1096
c2c087e6
DW		 1097 if (mpb->num_raid_devs >= 2) {
1098 fprintf(stderr, Name": This imsm-container already has the "
1099 "maximum of 2 volumes\n");
1100 return 0;
1101 }
1102
1103 super->creating_dev = idx;
1104 mpb->num_raid_devs++;
1105 dev = get_imsm_dev(mpb, idx);
1106 strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN);
1107 array_blocks = calc_array_size(info->level, info->raid_disks,
1108 info->layout, info->chunk_size,
1109 info->size*2);
1110 dev->size_low = __cpu_to_le32((__u32) array_blocks);
1111 dev->size_high = __cpu_to_le32((__u32) (array_blocks >> 32));
1112 dev->status = __cpu_to_le32(0);
1113 dev->reserved_blocks = __cpu_to_le32(0);
1114 vol = &dev->vol;
1115 vol->migr_state = 0;
1116 vol->migr_type = 0;
1117 vol->dirty = 0;
1118 for (i = 0; i < idx; i++) {
1119 struct imsm_dev *prev = get_imsm_dev(mpb, i);
1120 struct imsm_map *pmap = &prev->vol.map[0];
1121
1122 offset += __le32_to_cpu(pmap->blocks_per_member);
1123 offset += IMSM_RESERVED_SECTORS;
1124 }
1125 map = &vol->map[0];
1126 map->pba_of_lba0 = __cpu_to_le32(offset);
1127 sz = info->size * 2;
1128 map->blocks_per_member = __cpu_to_le32(sz);
1129 map->blocks_per_strip = __cpu_to_le16(info->chunk_size >> 9);
1130 map->num_data_stripes = __cpu_to_le32(sz / (info->chunk_size >> 9));
1131 map->map_state = info->level ? IMSM_T_STATE_UNINITIALIZED :
1132 IMSM_T_STATE_NORMAL;
1133 if (info->level == 10)
1134 map->raid_level = 1;
1135 else
1136 map->raid_level = info->level;
1137 map->num_members = info->raid_disks;
1138 for (i = 0; i < map->num_members; i++) {
1139 /* initialized in add_to_super */
1140 map->disk_ord_tbl[i] = __cpu_to_le32(0);
1141 }
1142
1143 return 1;
cdddbdbc
DW		 1144}
1145
c2c087e6 1146static void add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk,
cdddbdbc
DW		 1147 int fd, char *devname)
1148{
c2c087e6
DW		 1149 struct intel_super *super = st->sb;
1150 struct imsm_super *mpb = super->mpb;
1151 struct imsm_disk *disk;
1152 struct dl *dd;
1153 unsigned long long size;
1154 __u32 status, id;
1155 int rv;
1156 struct stat stb;
1157
1158 fstat(fd, &stb);
1159 dd = malloc(sizeof(*dd));
1160 if (!dd) {
1161 fprintf(stderr,
1162 Name ": malloc failed %s:%d.\n", __func__, __LINE__);
1163 abort();
1164 }
1165 memset(dd, 0, sizeof(*dd));
1166 dd->major = major(stb.st_rdev);
1167 dd->minor = minor(stb.st_rdev);
1168 dd->index = dk->number;
1169 dd->devname = devname ? strdup(devname) : NULL;
1170 dd->next = super->disks;
1171 dd->fd = fd;
1172 rv = imsm_read_serial(fd, devname, dd->serial);
1173 if (rv) {
1174 fprintf(stderr,
1175 Name ": failed to retrieve scsi serial "
1176 "using \'%s\' instead\n", devname);
1177 strcpy((char *) dd->serial, devname);
1178 }
1179
1180 if (mpb->num_disks <= dk->number)
1181 mpb->num_disks = dk->number + 1;
1182
1183 disk = get_imsm_disk(mpb, dk->number);
1184 get_dev_size(fd, NULL, &size);
1185 size /= 512;
1186 status = USABLE_DISK | SPARE_DISK;
1187 strcpy((char *) disk->serial, (char *) dd->serial);
1188 disk->total_blocks = __cpu_to_le32(size);
1189 disk->status = __cpu_to_le32(status);
1190 if (sysfs_disk_to_scsi_id(fd, &id) == 0)
1191 disk->scsi_id = __cpu_to_le32(id);
1192 else
1193 disk->scsi_id = __cpu_to_le32(0);
1194
1195 /* update the family number if we are creating a container */
1196 if (super->creating_imsm)
1197 mpb->family_num = __cpu_to_le32(gen_imsm_checksum(mpb));
1198
1199 super->disks = dd;
cdddbdbc
DW		 1200}
1201
c2c087e6 1202static void add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk,
8b353278 1203 int fd, char *devname)
cdddbdbc 1204{
c2c087e6
DW		 1205 struct intel_super *super = st->sb;
1206 struct imsm_super *mpb = super->mpb;
1207 struct dl *dl;
1208 struct imsm_dev *dev;
1209 struct imsm_map *map;
1210 struct imsm_disk *disk;
1211 __u32 status;
1212
1213 if (super->creating_dev == -1) {
1214 fprintf(stderr, Name ": no active raid device\n");
1215 abort();
1216 }
1217
1218 dev = get_imsm_dev(mpb, super->creating_dev);
1219 map = &dev->vol.map[0];
1220
1221 for (dl = super->disks; dl ; dl = dl->next)
1222 if (dl->major == dk->major &&
1223 dl->minor == dk->minor)
1224 break;
1225 if (!dl || ! (dk->state & (1<<MD_DISK_SYNC)))
1226 return;
1227
1228 map->disk_ord_tbl[dk->number] = __cpu_to_le32(dl->index);
1229
1230 disk = get_imsm_disk(mpb, dl->index);
1231 status = CONFIGURED_DISK | USABLE_DISK;
1232 disk->status = __cpu_to_le32(status);
cdddbdbc
DW		 1233}
1234
c2c087e6
DW		 1235static int store_imsm_mpb(int fd, struct intel_super *super);
1236
1237static int write_super_imsm(struct intel_super *super, int doclose)
cdddbdbc 1238{
c2c087e6
DW		 1239 struct imsm_super *mpb = super->mpb;
1240 struct dl *d;
1241 __u32 generation;
1242 __u32 sum;
cdddbdbc 1243
c2c087e6
DW		 1244 /* 'generation' is incremented everytime the metadata is written */
1245 generation = __le32_to_cpu(mpb->generation_num);
1246 generation++;
1247 mpb->generation_num = __cpu_to_le32(generation);
1248
1249 /* recalculate checksum */
1250 sum = gen_imsm_checksum(mpb);
1251 mpb->check_sum = __cpu_to_le32(sum);
1252
1253 for (d = super->disks; d ; d = d->next) {
1254 if (store_imsm_mpb(d->fd, super)) {
1255 fprintf(stderr, "%s: failed for device %d:%d %s\n",
1256 __func__, d->major, d->minor, strerror(errno));
1257 return 0;
1258 }
1259 if (doclose) {
1260 close(d->fd);
1261 d->fd = -1;
1262 }
1263 }
1264
1265 return 1;
1266}
1267
1268static int write_init_super_imsm(struct supertype *st)
1269{
1270 return write_super_imsm(st->sb, 1);
cdddbdbc
DW		 1271}
1272
1273static int store_zero_imsm(struct supertype *st, int fd)
1274{
551c80c1
DW		 1275 unsigned long long dsize;
1276 char buf[512];
1277
1278 get_dev_size(fd, NULL, &dsize);
1279
1280 /* first block is stored on second to last sector of the disk */
1281 if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0)
1282 return 1;
1283
1284 memset(buf, 0, sizeof(buf));
1285 if (write(fd, buf, sizeof(buf)) != sizeof(buf))
1286 return 1;
cdddbdbc
DW		 1287
1288 return 0;
1289}
1290
f352c545 1291static void getinfo_super_n_imsm_container(struct supertype *st, struct mdinfo *info)
cdddbdbc 1292{
f352c545 1293 /* just need offset and size...
c2c087e6 1294 * of the metadata
f352c545
DW		 1295 */
1296 struct intel_super *super = st->sb;
1297 struct imsm_super *mpb = super->mpb;
1298 struct imsm_disk *disk = get_imsm_disk(mpb, info->disk.number);
f352c545 1299
c2c087e6
DW		 1300 info->data_offset = __le32_to_cpu(disk->total_blocks) -
1301 (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS);
1302 info->component_size = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
cdddbdbc
DW		 1303}
1304
8b353278 1305static void getinfo_super_n_imsm_volume(struct supertype *st, struct mdinfo *info)
cdddbdbc 1306{
c2c087e6
DW		 1307 /* Find the particular details for info->disk.raid_disk.
1308 * This includes data_offset, component_size,
1309 */
1310 struct intel_super *super = st->sb;
1311 struct imsm_super *mpb = super->mpb;
1312 struct imsm_dev *dev = get_imsm_dev(mpb, super->creating_dev);
1313 struct imsm_map *map = &dev->vol.map[0];
1314
1315 info->data_offset = __le32_to_cpu(map->pba_of_lba0);
1316 info->component_size = __le32_to_cpu(map->blocks_per_member);
cdddbdbc
DW		 1317}
1318
1319static int validate_geometry_imsm(struct supertype *st, int level, int layout,
1320 int raiddisks, int chunk, unsigned long long size,
c2c087e6 1321 char *dev, unsigned long long *freesize)
cdddbdbc 1322{
c2c087e6
DW		 1323 int fd, cfd;
1324 struct mdinfo *sra;
cdddbdbc 1325
c2c087e6
DW		 1326 /* if given unused devices create a container
1327 * if given given devices in a container create a member volume
1328 */
1329 if (level == LEVEL_CONTAINER) {
1330 st->ss = &super_imsm_container;
1331 if (dev) {
1332 /* validate the container, dev == NULL */
1333 int rv = st->ss->validate_geometry(st, level, layout,
1334 raiddisks, chunk,
1335 size,
1336 NULL, freesize);
1337 if (rv)
1338 return rv;
1339 }
1340 return st->ss->validate_geometry(st, level, layout, raiddisks,
1341 chunk, size, dev, freesize);
1342 }
1343
1344 if (st->sb) {
1345 /* creating in a given container */
1346 st->ss = &super_imsm_volume;
1347 if (dev) {
1348 int rv = st->ss->validate_geometry(st, level, layout,
1349 raiddisks, chunk,
1350 size,
1351 NULL, freesize);
1352 if (rv)
1353 return rv;
1354 }
1355 return st->ss->validate_geometry(st, level, layout, raiddisks,
1356 chunk, size, dev, freesize);
1357 }
1358
1359 /* limit creation to the following levels */
1360 if (!dev)
1361 switch (level) {
1362 case 0:
1363 case 1:
1364 case 10:
1365 case 5:
1366 break;
1367 default:
1368 return 1;
1369 }
1370
1371 /* This device needs to be a device in an 'imsm' container */
1372 fd = open(dev, O_RDONLY|O_EXCL, 0);
1373 if (fd >= 0) {
1374 fprintf(stderr,
1375 Name ": Cannot create this array on device %s\n",
1376 dev);
1377 close(fd);
1378 return 0;
1379 }
1380 if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) {
1381 fprintf(stderr, Name ": Cannot open %s: %s\n",
1382 dev, strerror(errno));
1383 return 0;
1384 }
1385 /* Well, it is in use by someone, maybe an 'imsm' container. */
1386 cfd = open_container(fd);
1387 if (cfd < 0) {
1388 close(fd);
1389 fprintf(stderr, Name ": Cannot use %s: It is busy\n",
1390 dev);
1391 return 0;
1392 }
1393 sra = sysfs_read(cfd, 0, GET_VERSION);
1394 close(fd);
1395 if (sra && sra->array.major_version == -1 &&
1396 strcmp(sra->text_version, "imsm") == 0) {
1397 /* This is a member of a imsm container. Load the container
1398 * and try to create a volume
1399 */
1400 struct intel_super *super;
1401 st->ss = &super_imsm_volume;
1402 if (load_super_imsm_all(st, cfd, (void **) &super, NULL, 1) == 0) {
1403 st->sb = super;
1404 st->container_dev = fd2devnum(cfd);
1405 close(cfd);
1406 return st->ss->validate_geometry(st, level, layout,
1407 raiddisks, chunk, size,
1408 dev, freesize);
1409 }
1410 close(cfd);
1411 } else /* may belong to another container */
1412 return 0;
1413
1414 return 1;
cdddbdbc
DW		 1415}
1416
1417static int validate_geometry_imsm_container(struct supertype *st, int level,
1418 int layout, int raiddisks, int chunk,
c2c087e6 1419 unsigned long long size, char *dev,
cdddbdbc
DW		 1420 unsigned long long *freesize)
1421{
c2c087e6
DW		 1422 int fd;
1423 unsigned long long ldsize;
cdddbdbc 1424
c2c087e6
DW		 1425 if (level != LEVEL_CONTAINER)
1426 return 0;
1427 if (!dev)
1428 return 1;
1429
1430 fd = open(dev, O_RDONLY|O_EXCL, 0);
1431 if (fd < 0) {
1432 fprintf(stderr, Name ": Cannot open %s: %s\n",
1433 dev, strerror(errno));
1434 return 0;
1435 }
1436 if (!get_dev_size(fd, dev, &ldsize)) {
1437 close(fd);
1438 return 0;
1439 }
1440 close(fd);
1441
1442 *freesize = avail_size_imsm(st, ldsize >> 9);
1443
1444 return 1;
cdddbdbc
DW		 1445}
1446
c2c087e6
DW		 1447/* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd
1448 * FIX ME add ahci details
1449 */
8b353278
DW		 1450static int validate_geometry_imsm_volume(struct supertype *st, int level,
1451 int layout, int raiddisks, int chunk,
c2c087e6 1452 unsigned long long size, char *dev,
8b353278 1453 unsigned long long *freesize)
cdddbdbc 1454{
c2c087e6
DW		 1455 struct stat stb;
1456 struct intel_super *super = st->sb;
1457 struct dl *dl;
1458 unsigned long long pos = 0;
1459 unsigned long long maxsize;
1460 struct extent *e;
1461 int i;
cdddbdbc 1462
c2c087e6
DW		 1463 if (level == LEVEL_CONTAINER)
1464 return 0;
1465
1466 if (level == 1 && raiddisks > 2) {
1467 fprintf(stderr, Name ": imsm does not support more than 2 "
1468 "in a raid1 configuration\n");
1469 return 0;
1470 }
1471
1472 /* We must have the container info already read in. */
1473 if (!super)
1474 return 0;
1475
1476 if (!dev) {
1477 /* General test: make sure there is space for
1478 * 'raiddisks' device extents of size 'size'.
1479 */
1480 unsigned long long minsize = size*2 /* convert to blocks */;
1481 int dcnt = 0;
1482 if (minsize == 0)
1483 minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
1484 for (dl = super->disks; dl ; dl = dl->next) {
1485 int found = 0;
1486
1487 i = 0;
1488 e = get_extents(super, dl);
1489 if (!e) continue;
1490 do {
1491 unsigned long long esize;
1492 esize = e[i].start - pos;
1493 if (esize >= minsize)
1494 found = 1;
1495 pos = e[i].start + e[i].size;
1496 i++;
1497 } while (e[i-1].size);
1498 if (found)
1499 dcnt++;
1500 free(e);
1501 }
1502 if (dcnt < raiddisks) {
1503 fprintf(stderr, Name ": Not enough devices with space "
1504 "for this array (%d < %d)\n",
1505 dcnt, raiddisks);
1506 return 0;
1507 }
1508 return 1;
1509 }
1510 /* This device must be a member of the set */
1511 if (stat(dev, &stb) < 0)
1512 return 0;
1513 if ((S_IFMT & stb.st_mode) != S_IFBLK)
1514 return 0;
1515 for (dl = super->disks ; dl ; dl = dl->next) {
1516 if (dl->major == major(stb.st_rdev) &&
1517 dl->minor == minor(stb.st_rdev))
1518 break;
1519 }
1520 if (!dl) {
1521 fprintf(stderr, Name ": %s is not in the same imsm set\n",
1522 dev);
1523 return 0;
1524 }
1525 e = get_extents(super, dl);
1526 maxsize = 0;
1527 i = 0;
1528 if (e) do {
1529 unsigned long long esize;
1530 esize = e[i].start - pos;
1531 if (esize >= maxsize)
1532 maxsize = esize;
1533 pos = e[i].start + e[i].size;
1534 i++;
1535 } while (e[i-1].size);
1536 *freesize = maxsize;
1537
1538 return 1;
cdddbdbc
DW		 1539}
1540
cdddbdbc
DW		 1541static struct mdinfo *container_content_imsm(struct supertype *st)
1542{
4f5bc454
DW		 1543 /* Given a container loaded by load_super_imsm_all,
1544 * extract information about all the arrays into
1545 * an mdinfo tree.
1546 *
1547 * For each imsm_dev create an mdinfo, fill it in,
1548 * then look for matching devices in super->disks
1549 * and create appropriate device mdinfo.
1550 */
1551 struct intel_super *super = st->sb;
1552 struct imsm_super *mpb = super->mpb;
1553 struct mdinfo *rest = NULL;
1554 int i;
cdddbdbc 1555
4f5bc454
DW		 1556 for (i = 0; i < mpb->num_raid_devs; i++) {
1557 struct imsm_dev *dev = get_imsm_dev(mpb, i);
1558 struct imsm_vol *vol = &dev->vol;
1559 struct imsm_map *map = vol->map;
1560 struct mdinfo *this;
1561 __u64 sz;
1562 int slot;
1563
1564 this = malloc(sizeof(*this));
1565 memset(this, 0, sizeof(*this));
1566 this->next = rest;
1567 rest = this;
1568
1569 this->array.major_version = 2000;
1570 get_imsm_numerical_version(mpb, &this->array.minor_version,
1571 &this->array.patch_version);
1572 this->array.level = get_imsm_raid_level(map);
1573 this->array.raid_disks = map->num_members;
c2c087e6 1574 this->array.layout = imsm_level_to_layout(this->array.level);
4f5bc454
DW		 1575 this->array.md_minor = -1;
1576 this->array.ctime = 0;
1577 this->array.utime = 0;
1578 this->array.chunk_size = __le16_to_cpu(map->blocks_per_strip) << 9;
1579 this->array.state = !vol->dirty;
1580 this->container_member = i;
0fd5c350
DW		 1581 if (map->map_state == IMSM_T_STATE_UNINITIALIZED || dev->vol.dirty)
1582 this->resync_start = 0;
1583 else
1584 this->resync_start = ~0ULL;
1585
4f5bc454
DW		 1586 strncpy(this->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN);
1587 this->name[MAX_RAID_SERIAL_LEN] = 0;
1588
159c3a1a
NB		 1589 sprintf(this->text_version, "/%s/%d",
1590 devnum2devname(st->container_dev),
1591 this->container_member);
1592
4f5bc454
DW		 1593 memset(this->uuid, 0, sizeof(this->uuid));
1594
1595 sz = __le32_to_cpu(dev->size_high);
1596 sz <<= 32;
1597 sz += __le32_to_cpu(dev->size_low);
1598 this->component_size = sz;
1599 this->array.size = this->component_size / 2;
1600
1601 for (slot = 0 ; slot < map->num_members; slot++) {
1602 struct imsm_disk *disk;
1603 struct mdinfo *info_d;
1604 struct dl *d;
1605 int idx;
1606 __u32 s;
1607
1608 idx = __le32_to_cpu(map->disk_ord_tbl[slot] & ~(0xff << 24));
1609 for (d = super->disks; d ; d = d->next)
1610 if (d->index == idx)
1611 break;
1612
1613 if (d == NULL)
1614 break; /* shouldn't this be continue ?? */
1615
1616 info_d = malloc(sizeof(*info_d));
1617 if (!info_d)
1618 break; /* ditto ?? */
1619 memset(info_d, 0, sizeof(*info_d));
1620 info_d->next = this->devs;
1621 this->devs = info_d;
1622
1623 disk = get_imsm_disk(mpb, idx);
1624 s = __le32_to_cpu(disk->status);
1625
1626 info_d->disk.number = d->index;
1627 info_d->disk.major = d->major;
1628 info_d->disk.minor = d->minor;
1629 info_d->disk.raid_disk = slot;
1630 info_d->disk.state = s & CONFIGURED_DISK ? (1 << MD_DISK_ACTIVE) : 0;
1631 info_d->disk.state |= s & FAILED_DISK ? (1 << MD_DISK_FAULTY) : 0;
1632 info_d->disk.state |= s & USABLE_DISK ? (1 << MD_DISK_SYNC) : 0;
1633
1634 this->array.working_disks++;
1635
1636 info_d->events = __le32_to_cpu(mpb->generation_num);
1637 info_d->data_offset = __le32_to_cpu(map->pba_of_lba0);
1638 info_d->component_size = __le32_to_cpu(map->blocks_per_member);
1639 if (d->devname)
1640 strcpy(info_d->name, d->devname);
1641 }
1642 }
1643
1644 return rest;
cdddbdbc
DW		 1645}
1646
845dea95 1647
cba0191b
NB		 1648static int imsm_open_new(struct supertype *c, struct active_array *a,
1649 char *inst)
845dea95 1650{
4e6e574a 1651 dprintf("imsm: open_new %s\n", inst);
cba0191b 1652 a->info.container_member = atoi(inst);
845dea95
NB		 1653 return 0;
1654}
1655
c2a1e7da
DW		 1656static __u8 imsm_check_degraded(struct imsm_super *mpb, int n, int failed)
1657{
1658 struct imsm_dev *dev = get_imsm_dev(mpb, n);
1659 struct imsm_map *map = dev->vol.map;
1660
1661 if (!failed)
1662 return map->map_state;
1663
1664 switch (get_imsm_raid_level(map)) {
1665 case 0:
1666 return IMSM_T_STATE_FAILED;
1667 break;
1668 case 1:
1669 if (failed < map->num_members)
1670 return IMSM_T_STATE_DEGRADED;
1671 else
1672 return IMSM_T_STATE_FAILED;
1673 break;
1674 case 10:
1675 {
1676 /**
1677 * check to see if any mirrors have failed,
1678 * otherwise we are degraded
1679 */
1680 int device_per_mirror = 2; /* FIXME is this always the case?
1681 * and are they always adjacent?
1682 */
1683 int failed = 0;
1684 int i;
1685
1686 for (i = 0; i < map->num_members; i++) {
1687 int idx = get_imsm_disk_idx(map, i);
1688 struct imsm_disk *disk = get_imsm_disk(mpb, idx);
1689
1690 if (__le32_to_cpu(disk->status) & FAILED_DISK)
1691 failed++;
1692
1693 if (failed >= device_per_mirror)
1694 return IMSM_T_STATE_FAILED;
1695
1696 /* reset 'failed' for next mirror set */
1697 if (!((i + 1) % device_per_mirror))
1698 failed = 0;
1699 }
1700
1701 return IMSM_T_STATE_DEGRADED;
1702 }
1703 case 5:
1704 if (failed < 2)
1705 return IMSM_T_STATE_DEGRADED;
1706 else
1707 return IMSM_T_STATE_FAILED;
1708 break;
1709 default:
1710 break;
1711 }
1712
1713 return map->map_state;
1714}
1715
1716static int imsm_count_failed(struct imsm_super *mpb, struct imsm_map *map)
1717{
1718 int i;
1719 int failed = 0;
1720 struct imsm_disk *disk;
1721
1722 for (i = 0; i < map->num_members; i++) {
1723 int idx = get_imsm_disk_idx(map, i);
1724
1725 disk = get_imsm_disk(mpb, idx);
1726 if (__le32_to_cpu(disk->status) & FAILED_DISK)
1727 failed++;
1728 }
1729
1730 return failed;
845dea95
NB		 1731}
1732
a862209d
DW		 1733static void imsm_set_array_state(struct active_array *a, int consistent)
1734{
1735 int inst = a->info.container_member;
1736 struct intel_super *super = a->container->sb;
1737 struct imsm_dev *dev = get_imsm_dev(super->mpb, inst);
1738 struct imsm_map *map = &dev->vol.map[0];
1739 int dirty = !consistent;
1740 int failed;
1741 __u8 map_state;
1742
1743 if (a->resync_start == ~0ULL) {
1744 failed = imsm_count_failed(super->mpb, map);
1745 map_state = imsm_check_degraded(super->mpb, inst, failed);
1746 if (!failed)
1747 map_state = IMSM_T_STATE_NORMAL;
1748 if (map->map_state != map_state) {
4e6e574a 1749 dprintf("imsm: map_state %d: %d\n",
a862209d
DW		 1750 inst, map_state);
1751 map->map_state = map_state;
1752 super->updates_pending++;
1753 }
1754 }
1755
1756 if (dev->vol.dirty != dirty) {
4e6e574a 1757 dprintf("imsm: mark '%s' (%llu)\n",
a862209d
DW		 1758 dirty?"dirty":"clean", a->resync_start);
1759
1760 dev->vol.dirty = dirty;
1761 super->updates_pending++;
1762 }
1763}
1764
8d45d196 1765static void imsm_set_disk(struct active_array *a, int n, int state)
845dea95 1766{
8d45d196
DW		 1767 int inst = a->info.container_member;
1768 struct intel_super *super = a->container->sb;
1769 struct imsm_dev *dev = get_imsm_dev(super->mpb, inst);
1770 struct imsm_map *map = dev->vol.map;
1771 struct imsm_disk *disk;
1772 __u32 status;
1773 int failed = 0;
1774 int new_failure = 0;
1775
1776 if (n > map->num_members)
1777 fprintf(stderr, "imsm: set_disk %d out of range 0..%d\n",
1778 n, map->num_members - 1);
1779
1780 if (n < 0)
1781 return;
1782
4e6e574a 1783 dprintf("imsm: set_disk %d:%x\n", n, state);
8d45d196
DW		 1784
1785 disk = get_imsm_disk(super->mpb, get_imsm_disk_idx(map, n));
1786
1787 /* check if we have seen this failure before */
1788 status = __le32_to_cpu(disk->status);
1789 if ((state & DS_FAULTY) && !(status & FAILED_DISK)) {
1790 status |= FAILED_DISK;
1791 disk->status = __cpu_to_le32(status);
1792 new_failure = 1;
1793 }
1794
1795 /**
1796 * the number of failures have changed, count up 'failed' to determine
1797 * degraded / failed status
1798 */
1799 if (new_failure && map->map_state != IMSM_T_STATE_FAILED)
1800 failed = imsm_count_failed(super->mpb, map);
1801
1802 if (failed)
1803 map->map_state = imsm_check_degraded(super->mpb, inst, failed);
1804
1805 if (new_failure)
1806 super->updates_pending++;
845dea95
NB		 1807}
1808
c2a1e7da
DW		 1809static int store_imsm_mpb(int fd, struct intel_super *super)
1810{
1811 struct imsm_super *mpb = super->mpb;
1812 __u32 mpb_size = __le32_to_cpu(mpb->mpb_size);
1813 unsigned long long dsize;
1814 unsigned long long sectors;
1815
1816 get_dev_size(fd, NULL, &dsize);
1817
272f648f
DW		 1818 if (mpb_size > 512) {
1819 /* -1 to account for anchor */
1820 sectors = mpb_sectors(mpb) - 1;
c2a1e7da 1821
272f648f
DW		 1822 /* write the extended mpb to the sectors preceeding the anchor */
1823 if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0)
1824 return 1;
c2a1e7da 1825
272f648f
DW		 1826 if (write(fd, super->buf + 512, mpb_size - 512) != mpb_size - 512)
1827 return 1;
1828 }
c2a1e7da 1829
272f648f
DW		 1830 /* first block is stored on second to last sector of the disk */
1831 if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0)
c2a1e7da
DW		 1832 return 1;
1833
272f648f 1834 if (write(fd, super->buf, 512) != 512)
c2a1e7da
DW		 1835 return 1;
1836
1837 fsync(fd);
1838
1839 return 0;
1840}
1841
2e735d19 1842static void imsm_sync_metadata(struct supertype *container)
845dea95 1843{
2e735d19 1844 struct intel_super *super = container->sb;
c2a1e7da
DW		 1845
1846 if (!super->updates_pending)
1847 return;
1848
c2c087e6 1849 write_super_imsm(super, 0);
c2a1e7da
DW		 1850
1851 super->updates_pending = 0;
845dea95
NB		 1852}
1853
cdddbdbc
DW		 1854struct superswitch super_imsm = {
1855#ifndef MDASSEMBLE
1856 .examine_super = examine_super_imsm,
1857 .brief_examine_super = brief_examine_super_imsm,
1858 .detail_super = detail_super_imsm,
1859 .brief_detail_super = brief_detail_super_imsm,
1860#endif
1861 .match_home = match_home_imsm,
1862 .uuid_from_super= uuid_from_super_imsm,
1863 .getinfo_super = getinfo_super_imsm,
1864 .update_super = update_super_imsm,
1865
1866 .avail_size = avail_size_imsm,
1867
1868 .compare_super = compare_super_imsm,
1869
1870 .load_super = load_super_imsm,
1871 .init_super = init_zero_imsm,
1872 .store_super = store_zero_imsm,
1873 .free_super = free_super_imsm,
1874 .match_metadata_desc = match_metadata_desc_imsm,
f352c545 1875 .getinfo_super_n = getinfo_super_n_imsm_container,
cdddbdbc
DW		 1876
1877 .validate_geometry = validate_geometry_imsm,
cdddbdbc
DW		 1878 .swapuuid = 0,
1879 .external = 1,
845dea95
NB		 1880
1881/* for mdmon */
1882 .open_new = imsm_open_new,
1883 .load_super = load_super_imsm,
ed9d66aa 1884 .set_array_state= imsm_set_array_state,
845dea95
NB		 1885 .set_disk = imsm_set_disk,
1886 .sync_metadata = imsm_sync_metadata,
cdddbdbc
DW		 1887};
1888
1889/* super_imsm_container is set by validate_geometry_imsm when given a
1890 * device that is not part of any array
1891 */
1892struct superswitch super_imsm_container = {
1893
1894 .validate_geometry = validate_geometry_imsm_container,
1895 .init_super = init_super_imsm,
1896 .add_to_super = add_to_super_imsm,
1897 .write_init_super = write_init_super_imsm,
1898 .getinfo_super = getinfo_super_imsm,
f352c545 1899 .getinfo_super_n = getinfo_super_n_imsm_container,
cdddbdbc
DW		 1900 .load_super = load_super_imsm,
1901
1902#ifndef MDASSEMBLE
1903 .examine_super = examine_super_imsm,
1904 .brief_examine_super = brief_examine_super_imsm,
1905 .detail_super = detail_super_imsm,
1906 .brief_detail_super = brief_detail_super_imsm,
1907#endif
1908
1909 .free_super = free_super_imsm,
1910
1911 .container_content = container_content_imsm,
1912
cdddbdbc
DW		 1913 .swapuuid = 0,
1914 .external = 1,
cdddbdbc
DW		 1915};
1916
0063ecba 1917static struct superswitch super_imsm_volume = {
cdddbdbc 1918 .update_super = update_super_imsm,
8b353278
DW		 1919 .init_super = init_super_imsm_volume,
1920 .add_to_super = add_to_super_imsm_volume,
1921 .getinfo_super = getinfo_super_imsm_volume,
1922 .getinfo_super_n = getinfo_super_n_imsm_volume,
cdddbdbc
DW		 1923 .write_init_super = write_init_super_imsm,
1924
1925 .load_super = load_super_imsm,
1926 .free_super = free_super_imsm,
8b353278 1927 .match_metadata_desc = match_metadata_desc_imsm_volume,
cdddbdbc
DW		 1928
1929
8b353278 1930 .validate_geometry = validate_geometry_imsm_volume,
cdddbdbc
DW		 1931 .swapuuid = 0,
1932 .external = 2,
cdddbdbc 1933};
Unnamed repository; edit this file 'description' to name the repository.