Always initialise a struct super_type to zero
[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 creating_dev; /* 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 struct supertype *match_metadata_desc_imsm_volume(char *arg)
169 {
170         struct supertype *st;
171
172         if (strcmp(arg, "imsm/volume") != 0 &&
173             strcmp(arg, "raid") != 0 &&
174             strcmp(arg, "default") != 0
175                 )
176                 return NULL;
177
178         st = malloc(sizeof(*st));
179         memset(st, 0, sizeof(*st));
180         st->ss = &super_imsm_volume;
181         st->max_devs = IMSM_MAX_DEVICES;
182         st->minor_version = 0;
183         st->sb = NULL;
184         return st;
185 }
186
187 static __u8 *get_imsm_version(struct imsm_super *mpb)
188 {
189         return &mpb->sig[MPB_SIG_LEN];
190 }
191
192 static struct imsm_disk *get_imsm_disk(struct imsm_super *mpb, __u8 index)
193 {
194         if (index > mpb->num_disks - 1)
195                 return NULL;
196         return &mpb->disk[index];
197 }
198
199 static __u32 gen_imsm_checksum(struct imsm_super *mpb)
200 {
201         __u32 end = mpb->mpb_size / sizeof(end);
202         __u32 *p = (__u32 *) mpb;
203         __u32 sum = 0;
204
205         while (end--)
206                 sum += __le32_to_cpu(*p++);
207
208         return sum - __le32_to_cpu(mpb->check_sum);
209 }
210
211 static size_t sizeof_imsm_dev(struct imsm_dev *dev)
212 {
213         size_t size = sizeof(*dev);
214
215         /* each map has disk_ord_tbl[num_members - 1] additional space */
216         size += sizeof(__u32) * (dev->vol.map[0].num_members - 1);
217
218         /* migrating means an additional map */
219         if (dev->vol.migr_state) {
220                 size += sizeof(struct imsm_map);
221                 size += sizeof(__u32) * (dev->vol.map[1].num_members - 1);
222         }
223
224         return size;
225 }
226
227 static struct imsm_dev *get_imsm_dev(struct imsm_super *mpb, __u8 index)
228 {
229         int offset;
230         int i;
231         void *_mpb = mpb;
232
233         if (index > mpb->num_raid_devs - 1)
234                 return NULL;
235
236         /* devices start after all disks */
237         offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb;
238
239         for (i = 0; i <= index; i++)
240                 if (i == index)
241                         return _mpb + offset;
242                 else
243                         offset += sizeof_imsm_dev(_mpb + offset);
244
245         return NULL;
246 }
247
248 static __u32 get_imsm_disk_idx(struct imsm_map *map, int slot)
249 {
250         __u32 *ord_tbl = &map->disk_ord_tbl[slot];
251
252         /* top byte is 'special' */
253         return __le32_to_cpu(*ord_tbl & ~(0xff << 24));
254 }
255
256 static int get_imsm_raid_level(struct imsm_map *map)
257 {
258         if (map->raid_level == 1) {
259                 if (map->num_members == 2)
260                         return 1;
261                 else
262                         return 10;
263         }
264
265         return map->raid_level;
266 }
267
268 static int cmp_extent(const void *av, const void *bv)
269 {
270         const struct extent *a = av;
271         const struct extent *b = bv;
272         if (a->start < b->start)
273                 return -1;
274         if (a->start > b->start)
275                 return 1;
276         return 0;
277 }
278
279 static struct extent *get_extents(struct intel_super *super, struct dl *dl)
280 {
281         /* find a list of used extents on the given physical device */
282         struct imsm_super *mpb = super->mpb;
283         struct imsm_disk *disk;
284         struct extent *rv, *e;
285         int i, j;
286         int memberships = 0;
287
288         disk = get_imsm_disk(mpb, dl->index);
289         if (!disk)
290                 return NULL;
291
292         for (i = 0; i < mpb->num_raid_devs; i++) {
293                 struct imsm_dev *dev = get_imsm_dev(mpb, i);
294                 struct imsm_map *map = dev->vol.map;
295
296                 for (j = 0; j < map->num_members; j++) {
297                         __u32 index = get_imsm_disk_idx(map, j);
298
299                         if (index == dl->index)
300                                 memberships++;
301                 }
302         }
303         rv = malloc(sizeof(struct extent) * (memberships + 1));
304         if (!rv)
305                 return NULL;
306         e = rv;
307
308         for (i = 0; i < mpb->num_raid_devs; i++) {
309                 struct imsm_dev *dev = get_imsm_dev(mpb, i);
310                 struct imsm_map *map = dev->vol.map;
311
312                 for (j = 0; j < map->num_members; j++) {
313                         __u32 index = get_imsm_disk_idx(map, j);
314
315                         if (index == dl->index) {
316                                 e->start = __le32_to_cpu(map->pba_of_lba0);
317                                 e->size = __le32_to_cpu(map->blocks_per_member);
318                                 e++;
319                         }
320                 }
321         }
322         qsort(rv, memberships, sizeof(*rv), cmp_extent);
323
324         e->start = __le32_to_cpu(disk->total_blocks) -
325                    (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS);
326         e->size = 0;
327         return rv;
328 }
329
330 #ifndef MDASSEMBLE
331 static void print_imsm_dev(struct imsm_dev *dev, int index)
332 {
333         __u64 sz;
334         int slot;
335         struct imsm_map *map = dev->vol.map;
336
337         printf("\n");
338         printf("[%s]:\n", dev->volume);
339         printf("     RAID Level : %d\n", get_imsm_raid_level(map));
340         printf("        Members : %d\n", map->num_members);
341         for (slot = 0; slot < map->num_members; slot++)
342                 if (index == get_imsm_disk_idx(map, slot))
343                         break;
344         if (slot < map->num_members)
345                 printf("      This Slot : %d\n", slot);
346         else
347                 printf("      This Slot : ?\n");
348         sz = __le32_to_cpu(dev->size_high);
349         sz <<= 32;
350         sz += __le32_to_cpu(dev->size_low);
351         printf("     Array Size : %llu%s\n", (unsigned long long)sz,
352                human_size(sz * 512));
353         sz = __le32_to_cpu(map->blocks_per_member);
354         printf("   Per Dev Size : %llu%s\n", (unsigned long long)sz,
355                human_size(sz * 512));
356         printf("  Sector Offset : %u\n",
357                 __le32_to_cpu(map->pba_of_lba0));
358         printf("    Num Stripes : %u\n",
359                 __le32_to_cpu(map->num_data_stripes));
360         printf("     Chunk Size : %u KiB\n",
361                 __le16_to_cpu(map->blocks_per_strip) / 2);
362         printf("       Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks));
363         printf("  Migrate State : %s\n", dev->vol.migr_state ? "migrating" : "idle");
364         printf("    Dirty State : %s\n", dev->vol.dirty ? "dirty" : "clean");
365         printf("      Map State : %s\n", map_state_str[map->map_state]);
366 }
367
368 static void print_imsm_disk(struct imsm_super *mpb, int index)
369 {
370         struct imsm_disk *disk = get_imsm_disk(mpb, index);
371         char str[MAX_RAID_SERIAL_LEN];
372         __u32 s;
373         __u64 sz;
374
375         printf("\n");
376         snprintf(str, MAX_RAID_SERIAL_LEN, "%s", disk->serial);
377         printf("  Disk%02d Serial : %s\n", index, str);
378         s = __le32_to_cpu(disk->status);
379         printf("          State :%s%s%s%s\n", s&SPARE_DISK ? " spare" : "",
380                                               s&CONFIGURED_DISK ? " active" : "",
381                                               s&FAILED_DISK ? " failed" : "",
382                                               s&USABLE_DISK ? " usable" : "");
383         printf("             Id : %08x\n", __le32_to_cpu(disk->scsi_id));
384         sz = __le32_to_cpu(disk->total_blocks) -
385              (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS * mpb->num_raid_devs);
386         printf("    Usable Size : %llu%s\n", (unsigned long long)sz,
387                human_size(sz * 512));
388 }
389
390 static void examine_super_imsm(struct supertype *st, char *homehost)
391 {
392         struct intel_super *super = st->sb;
393         struct imsm_super *mpb = super->mpb;
394         char str[MAX_SIGNATURE_LENGTH];
395         int i;
396         __u32 sum;
397
398         snprintf(str, MPB_SIG_LEN, "%s", mpb->sig);
399         printf("          Magic : %s\n", str);
400         snprintf(str, strlen(MPB_VERSION_RAID0), "%s", get_imsm_version(mpb));
401         printf("        Version : %s\n", get_imsm_version(mpb));
402         printf("         Family : %08x\n", __le32_to_cpu(mpb->family_num));
403         printf("     Generation : %08x\n", __le32_to_cpu(mpb->generation_num));
404         sum = __le32_to_cpu(mpb->check_sum);
405         printf("       Checksum : %08x %s\n", sum,
406                 gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect");
407         printf("    MPB Sectors : %d\n", mpb_sectors(mpb));
408         printf("          Disks : %d\n", mpb->num_disks);
409         printf("   RAID Devices : %d\n", mpb->num_raid_devs);
410         print_imsm_disk(mpb, super->disks->index);
411         for (i = 0; i < mpb->num_raid_devs; i++)
412                 print_imsm_dev(get_imsm_dev(mpb, i), super->disks->index);
413         for (i = 0; i < mpb->num_disks; i++) {
414                 if (i == super->disks->index)
415                         continue;
416                 print_imsm_disk(mpb, i);
417         }
418 }
419
420 static void brief_examine_super_imsm(struct supertype *st)
421 {
422         struct intel_super *super = st->sb;
423         struct imsm_super *mpb = super->mpb;
424
425         printf("ARRAY /dev/imsm family=%08x metadata=external:imsm\n",
426                 __le32_to_cpu(mpb->family_num));
427 }
428
429 static void detail_super_imsm(struct supertype *st, char *homehost)
430 {
431         printf("%s\n", __FUNCTION__);
432 }
433
434 static void brief_detail_super_imsm(struct supertype *st)
435 {
436         printf("%s\n", __FUNCTION__);
437 }
438 #endif
439
440 static int match_home_imsm(struct supertype *st, char *homehost)
441 {
442         printf("%s\n", __FUNCTION__);
443
444         return 0;
445 }
446
447 static void uuid_from_super_imsm(struct supertype *st, int uuid[4])
448 {
449         printf("%s\n", __FUNCTION__);
450 }
451
452 static void
453 get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p)
454 {
455         __u8 *v = get_imsm_version(mpb);
456         __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH;
457         char major[] = { 0, 0, 0 };
458         char minor[] = { 0 ,0, 0 };
459         char patch[] = { 0, 0, 0 };
460         char *ver_parse[] = { major, minor, patch };
461         int i, j;
462
463         i = j = 0;
464         while (*v != '\0' && v < end) {
465                 if (*v != '.' && j < 2)
466                         ver_parse[i][j++] = *v;
467                 else {
468                         i++;
469                         j = 0;
470                 }
471                 v++;
472         }
473
474         *m = strtol(minor, NULL, 0);
475         *p = strtol(patch, NULL, 0);
476 }
477
478 static int imsm_level_to_layout(int level)
479 {
480         switch (level) {
481         case 0:
482         case 1:
483                 return 0;
484         case 5:
485         case 6:
486                 return ALGORITHM_LEFT_SYMMETRIC;
487         case 10:
488                 return 0x102; //FIXME is this correct?
489         }
490         return -1;
491 }
492
493 static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info)
494 {
495         struct intel_super *super = st->sb;
496         struct imsm_super *mpb = super->mpb;
497         struct imsm_disk *disk;
498         __u32 s;
499
500         info->array.major_version = 2000;
501         get_imsm_numerical_version(mpb, &info->array.minor_version,
502                                    &info->array.patch_version);
503         info->array.raid_disks    = mpb->num_disks;
504         info->array.level         = LEVEL_CONTAINER;
505         info->array.layout        = 0;
506         info->array.md_minor      = -1;
507         info->array.ctime         = 0; /* N/A for imsm */ 
508         info->array.utime         = 0;
509         info->array.chunk_size    = 0;
510
511         info->disk.major = 0;
512         info->disk.minor = 0;
513         info->disk.raid_disk = -1;
514         info->reshape_active = 0;
515         strcpy(info->text_version, "imsm");
516         info->disk.number = -1;
517         info->disk.state = 0;
518
519         if (super->disks) {
520                 info->disk.number = super->disks->index;
521                 info->disk.raid_disk = super->disks->index;
522                 disk = get_imsm_disk(mpb, super->disks->index);
523                 s = __le32_to_cpu(disk->status);
524                 info->disk.state  = s & CONFIGURED_DISK ? (1 << MD_DISK_ACTIVE) : 0;
525                 info->disk.state |= s & FAILED_DISK ? (1 << MD_DISK_FAULTY) : 0;
526                 info->disk.state |= s & USABLE_DISK ? (1 << MD_DISK_SYNC) : 0;
527         }
528 }
529
530 static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info)
531 {
532         struct intel_super *super = st->sb;
533         struct imsm_super *mpb = super->mpb;
534         struct imsm_dev *dev = get_imsm_dev(mpb, info->container_member);
535         struct imsm_map *map = &dev->vol.map[0];
536
537         info->array.major_version = 2000;
538         get_imsm_numerical_version(mpb, &info->array.minor_version,
539                                    &info->array.patch_version);
540         info->array.raid_disks    = map->num_members;
541         info->array.level         = get_imsm_raid_level(map);
542         info->array.layout        = imsm_level_to_layout(info->array.level);
543         info->array.md_minor      = -1;
544         info->array.ctime         = 0;
545         info->array.utime         = 0;
546         info->array.chunk_size    = __le16_to_cpu(map->blocks_per_strip * 512);
547
548         info->data_offset         = __le32_to_cpu(map->pba_of_lba0);
549         info->component_size      = __le32_to_cpu(map->blocks_per_member);
550
551         info->disk.major = 0;
552         info->disk.minor = 0;
553
554         sprintf(info->text_version, "/%s/%d",
555                 devnum2devname(st->container_dev),
556                 info->container_member);
557 }
558
559 static int update_super_imsm(struct supertype *st, struct mdinfo *info,
560                              char *update, char *devname, int verbose,
561                              int uuid_set, char *homehost)
562 {
563         /* FIXME */
564
565         /* For 'assemble' and 'force' we need to return non-zero if any
566          * change was made.  For others, the return value is ignored.
567          * Update options are:
568          *  force-one : This device looks a bit old but needs to be included,
569          *        update age info appropriately.
570          *  assemble: clear any 'faulty' flag to allow this device to
571          *              be assembled.
572          *  force-array: Array is degraded but being forced, mark it clean
573          *         if that will be needed to assemble it.
574          *
575          *  newdev:  not used ????
576          *  grow:  Array has gained a new device - this is currently for
577          *              linear only
578          *  resync: mark as dirty so a resync will happen.
579          *  name:  update the name - preserving the homehost
580          *
581          * Following are not relevant for this imsm:
582          *  sparc2.2 : update from old dodgey metadata
583          *  super-minor: change the preferred_minor number
584          *  summaries:  update redundant counters.
585          *  uuid:  Change the uuid of the array to match watch is given
586          *  homehost:  update the recorded homehost
587          *  _reshape_progress: record new reshape_progress position.
588          */
589         int rv = 0;
590         //struct intel_super *super = st->sb;
591         //struct imsm_super *mpb = super->mpb;
592
593         if (strcmp(update, "grow") == 0) {
594         }
595         if (strcmp(update, "resync") == 0) {
596                 /* dev->vol.dirty = 1; */
597         }
598
599         /* IMSM has no concept of UUID or homehost */
600
601         return rv;
602 }
603
604 static size_t disks_to_mpb_size(int disks)
605 {
606         size_t size;
607
608         size = sizeof(struct imsm_super);
609         size += (disks - 1) * sizeof(struct imsm_disk);
610         size += 2 * sizeof(struct imsm_dev);
611         /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */
612         size += (4 - 2) * sizeof(struct imsm_map);
613         /* 4 possible disk_ord_tbl's */
614         size += 4 * (disks - 1) * sizeof(__u32);
615
616         return size;
617 }
618
619 static __u64 avail_size_imsm(struct supertype *st, __u64 devsize)
620 {
621         if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS))
622                 return 0;
623
624         return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS);
625 }
626
627 static int compare_super_imsm(struct supertype *st, struct supertype *tst)
628 {
629         /*
630          * return:
631          *  0 same, or first was empty, and second was copied
632          *  1 second had wrong number
633          *  2 wrong uuid
634          *  3 wrong other info
635          */
636         struct intel_super *first = st->sb;
637         struct intel_super *sec = tst->sb;
638
639         if (!first) {
640                 st->sb = tst->sb;
641                 tst->sb = NULL;
642                 return 0;
643         }
644
645         if (memcmp(first->mpb->sig, sec->mpb->sig, MAX_SIGNATURE_LENGTH) != 0)
646                 return 3;
647         if (first->mpb->family_num != sec->mpb->family_num)
648                 return 3;
649         if (first->mpb->mpb_size != sec->mpb->mpb_size)
650                 return 3;
651         if (first->mpb->check_sum != sec->mpb->check_sum)
652                 return 3;
653
654         return 0;
655 }
656
657 extern int scsi_get_serial(int fd, void *buf, size_t buf_len);
658
659 static int imsm_read_serial(int fd, char *devname,
660                             __u8 serial[MAX_RAID_SERIAL_LEN])
661 {
662         unsigned char scsi_serial[255];
663         int sg_fd;
664         int rv;
665         int rsp_len;
666         int i, cnt;
667
668         memset(scsi_serial, 0, sizeof(scsi_serial));
669
670         sg_fd = sysfs_disk_to_sg(fd);
671         if (sg_fd < 0) {
672                 if (devname)
673                         fprintf(stderr,
674                                 Name ": Failed to open sg interface for %s: %s\n",
675                                 devname, strerror(errno));
676                 return 1;
677         }
678
679         rv = scsi_get_serial(sg_fd, scsi_serial, sizeof(scsi_serial));
680         close(sg_fd);
681
682         if (rv != 0) {
683                 if (devname)
684                         fprintf(stderr,
685                                 Name ": Failed to retrieve serial for %s\n",
686                                 devname);
687                 return rv;
688         }
689
690         rsp_len = scsi_serial[3];
691         for (i = 0, cnt = 0; i < rsp_len; i++) {
692                 if (!isspace(scsi_serial[4 + i]))
693                         serial[cnt++] = scsi_serial[4 + i];
694                 if (cnt == MAX_RAID_SERIAL_LEN)
695                         break;
696         }
697
698         serial[MAX_RAID_SERIAL_LEN - 1] = '\0';
699
700         return 0;
701 }
702
703 static int
704 load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd)
705 {
706         struct imsm_super *mpb = super->mpb;
707         struct dl *dl;
708         struct stat stb;
709         struct imsm_disk *disk;
710         int rv;
711         int i;
712
713         dl = malloc(sizeof(*dl));
714         if (!dl) {
715                 if (devname)
716                         fprintf(stderr,
717                                 Name ": failed to allocate disk buffer for %s\n",
718                                 devname);
719                 return 2;
720         }
721         memset(dl, 0, sizeof(*dl));
722
723         fstat(fd, &stb);
724         dl->major = major(stb.st_rdev);
725         dl->minor = minor(stb.st_rdev);
726         dl->next = super->disks;
727         dl->fd = keep_fd ? fd : -1;
728         dl->devname = devname ? strdup(devname) : NULL;
729         dl->index = -1;
730         super->disks = dl;
731         rv = imsm_read_serial(fd, devname, dl->serial);
732
733         if (rv != 0)
734                 return 2;
735
736         /* look up this disk's index */
737         for (i = 0; i < mpb->num_disks; i++) {
738                 disk = get_imsm_disk(mpb, i);
739
740                 if (memcmp(disk->serial, dl->serial, MAX_RAID_SERIAL_LEN) == 0)
741                         break;
742         }
743
744         if (i > mpb->num_disks)
745                 return 2;
746
747         dl->index = i;
748
749         return 0;
750 }
751
752 /* load_imsm_mpb - read matrix metadata
753  * allocates super->mpb to be freed by free_super
754  */
755 static int load_imsm_mpb(int fd, struct intel_super *super, char *devname)
756 {
757         unsigned long long dsize;
758         size_t len, mpb_size;
759         unsigned long long sectors;
760         struct stat;
761         struct imsm_super anchor;
762         __u32 check_sum;
763
764         memset(super, 0, sizeof(*super));
765         get_dev_size(fd, NULL, &dsize);
766
767         if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) {
768                 if (devname)
769                         fprintf(stderr,
770                                 Name ": Cannot seek to anchor block on %s: %s\n",
771                                 devname, strerror(errno));
772                 return 1;
773         }
774
775         len = sizeof(anchor);
776         if (read(fd, &anchor, len) != len) {
777                 if (devname)
778                         fprintf(stderr,
779                                 Name ": Cannot read anchor block on %s: %s\n",
780                                 devname, strerror(errno));
781                 return 1;
782         }
783
784         if (strncmp((char *) anchor.sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) {
785                 if (devname)
786                         fprintf(stderr,
787                                 Name ": no IMSM anchor on %s\n", devname);
788                 return 2;
789         }
790
791         mpb_size = __le32_to_cpu(anchor.mpb_size);
792         super->mpb = malloc(mpb_size < 512 ? 512 : mpb_size);
793         if (!super->mpb) {
794                 if (devname)
795                         fprintf(stderr,
796                                 Name ": unable to allocate %zu byte mpb buffer\n",
797                                 mpb_size);
798                 return 2;
799         }
800         memcpy(super->buf, &anchor, sizeof(anchor));
801
802         /* read the rest of the first block */
803         len = 512 - sizeof(anchor);
804         if (read(fd, super->buf + sizeof(anchor), len) != len) {
805                 if (devname)
806                         fprintf(stderr,
807                                 Name ": Cannot read anchor remainder on %s: %s\n",
808                                 devname, strerror(errno));
809                 return 2;
810         }
811
812         sectors = mpb_sectors(&anchor) - 1;
813         if (!sectors)
814                 return load_imsm_disk(fd, super, devname, 0);
815
816         /* read the extended mpb */
817         if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) {
818                 if (devname)
819                         fprintf(stderr,
820                                 Name ": Cannot seek to extended mpb on %s: %s\n",
821                                 devname, strerror(errno));
822                 return 1;
823         }
824
825         len = mpb_size - 512;
826         if (read(fd, super->buf + 512, len) != len) {
827                 if (devname)
828                         fprintf(stderr,
829                                 Name ": Cannot read extended mpb on %s: %s\n",
830                                 devname, strerror(errno));
831                 return 2;
832         }
833
834         check_sum = gen_imsm_checksum(super->mpb);
835         if (check_sum != __le32_to_cpu(super->mpb->check_sum)) {
836                 if (devname)
837                         fprintf(stderr,
838                                 Name ": IMSM checksum %x != %x on %s\n",
839                                 check_sum, __le32_to_cpu(super->mpb->check_sum),
840                                 devname);
841                 return 2;
842         }
843
844         return load_imsm_disk(fd, super, devname, 0);
845 }
846
847 struct superswitch super_imsm_container;
848
849 static void free_imsm_disks(struct intel_super *super)
850 {
851         while (super->disks) {
852                 struct dl *d = super->disks;
853
854                 super->disks = d->next;
855                 if (d->fd >= 0)
856                         close(d->fd);
857                 if (d->devname)
858                         free(d->devname);
859                 free(d);
860         }
861 }
862
863 static void free_imsm(struct intel_super *super)
864 {
865         if (super->mpb)
866                 free(super->mpb);
867         free_imsm_disks(super);
868         free(super);
869 }
870
871
872 static void free_super_imsm(struct supertype *st)
873 {
874         struct intel_super *super = st->sb;
875
876         if (!super)
877                 return;
878
879         free_imsm(super);
880         st->sb = NULL;
881 }
882
883 static struct intel_super *alloc_super(int creating_imsm)
884 {
885         struct intel_super *super = malloc(sizeof(*super));
886
887         if (super) {
888                 memset(super, 0, sizeof(*super));
889                 super->creating_imsm = creating_imsm;
890                 super->creating_dev = -1;
891         }
892
893         return super;
894 }
895
896 #ifndef MDASSEMBLE
897 static int load_super_imsm_all(struct supertype *st, int fd, void **sbp,
898                                char *devname, int keep_fd)
899 {
900         struct mdinfo *sra;
901         struct intel_super *super;
902         struct mdinfo *sd, *best = NULL;
903         __u32 bestgen = 0;
904         __u32 gen;
905         char nm[20];
906         int dfd;
907         int rv;
908
909         /* check if this disk is a member of an active array */
910         sra = sysfs_read(fd, 0, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE);
911         if (!sra)
912                 return 1;
913
914         if (sra->array.major_version != -1 ||
915             sra->array.minor_version != -2 ||
916             strcmp(sra->text_version, "imsm") != 0)
917                 return 1;
918
919         super = alloc_super(0);
920         if (!super)
921                 return 1;
922
923         /* find the most up to date disk in this array */
924         for (sd = sra->devs; sd; sd = sd->next) {
925                 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
926                 dfd = dev_open(nm, keep_fd ? O_RDWR : O_RDONLY);
927                 if (!dfd) {
928                         free_imsm(super);
929                         return 2;
930                 }
931                 rv = load_imsm_mpb(dfd, super, NULL);
932                 if (!keep_fd)
933                         close(dfd);
934                 if (rv == 0) {
935                         gen = __le32_to_cpu(super->mpb->generation_num);
936                         if (!best || gen > bestgen) {
937                                 bestgen = gen;
938                                 best = sd;
939                         }
940                 } else {
941                         free_imsm(super);
942                         return 2;
943                 }
944         }
945
946         if (!best) {
947                 free_imsm(super);
948                 return 1;
949         }
950
951         /* load the most up to date anchor */
952         sprintf(nm, "%d:%d", best->disk.major, best->disk.minor);
953         dfd = dev_open(nm, O_RDONLY);
954         if (!dfd) {
955                 free_imsm(super);
956                 return 1;
957         }
958         rv = load_imsm_mpb(dfd, super, NULL);
959         close(dfd);
960         if (rv != 0) {
961                 free_imsm(super);
962                 return 2;
963         }
964
965         /* reset the disk list */
966         free_imsm_disks(super);
967
968         /* populate disk list */
969         for (sd = sra->devs ; sd ; sd = sd->next) {
970                 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
971                 dfd = dev_open(nm, keep_fd? O_RDWR : O_RDONLY);
972                 if (!dfd) {
973                         free_imsm(super);
974                         return 2;
975                 }
976                 load_imsm_disk(dfd, super, NULL, keep_fd);
977                 if (!keep_fd)
978                         close(dfd);
979         }
980
981         *sbp = super;
982         if (st->ss == NULL) {
983                 st->ss = &super_imsm_container;
984                 st->minor_version = 0;
985                 st->max_devs = IMSM_MAX_DEVICES;
986                 st->container_dev = fd2devnum(fd);
987         }
988
989         return 0;
990 }
991 #endif
992
993 static int load_super_imsm(struct supertype *st, int fd, char *devname)
994 {
995         struct intel_super *super;
996         int rv;
997
998 #ifndef MDASSEMBLE
999         if (load_super_imsm_all(st, fd, &st->sb, devname, 1) == 0)
1000                 return 0;
1001 #endif
1002
1003         super = alloc_super(0);
1004         if (!super) {
1005                 fprintf(stderr,
1006                         Name ": malloc of %zu failed.\n",
1007                         sizeof(*super));
1008                 return 1;
1009         }
1010
1011         rv = load_imsm_mpb(fd, super, devname);
1012
1013         if (rv) {
1014                 if (devname)
1015                         fprintf(stderr,
1016                                 Name ": Failed to load all information "
1017                                 "sections on %s\n", devname);
1018                 free_imsm(super);
1019                 return rv;
1020         }
1021
1022         st->sb = super;
1023         if (st->ss == NULL) {
1024                 st->ss = &super_imsm;
1025                 st->minor_version = 0;
1026                 st->max_devs = IMSM_MAX_DEVICES;
1027         }
1028
1029         return 0;
1030 }
1031
1032 static int init_zero_imsm(struct supertype *st, mdu_array_info_t *info,
1033                           unsigned long long size, char *name,
1034                           char *homehost, int *uuid)
1035 {
1036         st->sb = NULL;
1037         return 0;
1038 }
1039
1040 static int init_super_imsm(struct supertype *st, mdu_array_info_t *info,
1041                            unsigned long long size, char *name,
1042                            char *homehost, int *uuid)
1043 {
1044         /* This is primarily called by Create when creating a new array.
1045          * We will then get add_to_super called for each component, and then
1046          * write_init_super called to write it out to each device.
1047          * For IMSM, Create can create on fresh devices or on a pre-existing
1048          * array.
1049          * To create on a pre-existing array a different method will be called.
1050          * This one is just for fresh drives.
1051          */
1052         struct intel_super *super;
1053         struct imsm_super *mpb;
1054         size_t mpb_size;
1055
1056         super = alloc_super(1);
1057         if (!super)
1058                 return 0;
1059         mpb_size = disks_to_mpb_size(info->nr_disks);
1060         mpb = malloc(mpb_size);
1061         if (!mpb) {
1062                 free(super);
1063                 return 0;
1064         }
1065         memset(mpb, 0, mpb_size); 
1066
1067         memcpy(mpb->sig, MPB_SIGNATURE, strlen(MPB_SIGNATURE));
1068         memcpy(mpb->sig + strlen(MPB_SIGNATURE), MPB_VERSION_RAID5,
1069                strlen(MPB_VERSION_RAID5)); 
1070         mpb->mpb_size = mpb_size;
1071
1072         super->mpb = mpb;
1073         st->sb = super;
1074         return 1;
1075 }
1076
1077 static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info,
1078                                   unsigned long long size, char *name,
1079                                   char *homehost, int *uuid)
1080 {
1081         /* We are creating a volume inside a pre-existing container.
1082          * so st->sb is already set.
1083          */
1084         struct intel_super *super = st->sb;
1085         struct imsm_super *mpb = super->mpb;
1086         struct imsm_dev *dev;
1087         struct imsm_vol *vol;
1088         struct imsm_map *map;
1089         int idx = mpb->num_raid_devs;
1090         int i;
1091         unsigned long long array_blocks;
1092         unsigned long long sz;
1093         __u32 offset = 0;
1094
1095         if (mpb->num_raid_devs >= 2) {
1096                 fprintf(stderr, Name": This imsm-container already has the "
1097                         "maximum of 2 volumes\n");
1098                 return 0;
1099         }
1100
1101         super->creating_dev = idx;
1102         mpb->num_raid_devs++;
1103         dev = get_imsm_dev(mpb, idx);
1104         strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN);
1105         array_blocks = calc_array_size(info->level, info->raid_disks,
1106                                        info->layout, info->chunk_size,
1107                                        info->size*2);
1108         dev->size_low = __cpu_to_le32((__u32) array_blocks);
1109         dev->size_high = __cpu_to_le32((__u32) (array_blocks >> 32));
1110         dev->status = __cpu_to_le32(0);
1111         dev->reserved_blocks = __cpu_to_le32(0);
1112         vol = &dev->vol;
1113         vol->migr_state = 0;
1114         vol->migr_type = 0;
1115         vol->dirty = 0;
1116         for (i = 0; i < idx; i++) {
1117                 struct imsm_dev *prev = get_imsm_dev(mpb, i);
1118                 struct imsm_map *pmap = &prev->vol.map[0];
1119
1120                 offset += __le32_to_cpu(pmap->blocks_per_member);
1121                 offset += IMSM_RESERVED_SECTORS;
1122         }
1123         map = &vol->map[0];
1124         map->pba_of_lba0 = __cpu_to_le32(offset);
1125         sz = info->size * 2;
1126         map->blocks_per_member = __cpu_to_le32(sz);
1127         map->blocks_per_strip = __cpu_to_le16(info->chunk_size >> 9);
1128         map->num_data_stripes = __cpu_to_le32(sz / (info->chunk_size >> 9));
1129         map->map_state = info->level ? IMSM_T_STATE_UNINITIALIZED :
1130                                        IMSM_T_STATE_NORMAL;
1131         if (info->level == 10)
1132                 map->raid_level = 1;
1133         else
1134                 map->raid_level = info->level;
1135         map->num_members = info->raid_disks;
1136         for (i = 0; i < map->num_members; i++) {
1137                 /* initialized in add_to_super */
1138                 map->disk_ord_tbl[i] = __cpu_to_le32(0);
1139         }
1140
1141         return 1;
1142 }
1143
1144 static void add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk,
1145                               int fd, char *devname)
1146 {
1147         struct intel_super *super = st->sb;
1148         struct imsm_super *mpb = super->mpb;
1149         struct imsm_disk *disk;
1150         struct dl *dd;
1151         unsigned long long size;
1152         __u32 status, id;
1153         int rv;
1154         struct stat stb;
1155
1156         fstat(fd, &stb);
1157         dd = malloc(sizeof(*dd));
1158         if (!dd) {
1159                 fprintf(stderr,
1160                         Name ": malloc failed %s:%d.\n", __func__, __LINE__);
1161                 abort();
1162         }
1163         memset(dd, 0, sizeof(*dd));
1164         dd->major = major(stb.st_rdev);
1165         dd->minor = minor(stb.st_rdev);
1166         dd->index = dk->number;
1167         dd->devname = devname ? strdup(devname) : NULL;
1168         dd->next = super->disks;
1169         dd->fd = fd;
1170         rv = imsm_read_serial(fd, devname, dd->serial);
1171         if (rv) {
1172                 fprintf(stderr,
1173                         Name ": failed to retrieve scsi serial "
1174                         "using \'%s\' instead\n", devname);
1175                 strcpy((char *) dd->serial, devname);
1176         }
1177
1178         if (mpb->num_disks <= dk->number)
1179                 mpb->num_disks = dk->number + 1;
1180
1181         disk = get_imsm_disk(mpb, dk->number);
1182         get_dev_size(fd, NULL, &size);
1183         size /= 512;
1184         status = USABLE_DISK | SPARE_DISK;
1185         strcpy((char *) disk->serial, (char *) dd->serial);
1186         disk->total_blocks = __cpu_to_le32(size);
1187         disk->status = __cpu_to_le32(status);
1188         if (sysfs_disk_to_scsi_id(fd, &id) == 0)
1189                 disk->scsi_id = __cpu_to_le32(id);
1190         else
1191                 disk->scsi_id = __cpu_to_le32(0);
1192
1193         /* update the family number if we are creating a container */
1194         if (super->creating_imsm)
1195                 mpb->family_num = __cpu_to_le32(gen_imsm_checksum(mpb));
1196         
1197         super->disks = dd;
1198 }
1199
1200 static void add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk,
1201                                      int fd, char *devname)
1202 {
1203         struct intel_super *super = st->sb;
1204         struct imsm_super *mpb = super->mpb;
1205         struct dl *dl;
1206         struct imsm_dev *dev;
1207         struct imsm_map *map;
1208         struct imsm_disk *disk;
1209         __u32 status;
1210
1211         if (super->creating_dev == -1) {
1212                 fprintf(stderr, Name ": no active raid device\n");
1213                 abort();
1214         }
1215
1216         dev = get_imsm_dev(mpb, super->creating_dev);
1217         map = &dev->vol.map[0];
1218
1219         for (dl = super->disks; dl ; dl = dl->next)
1220                 if (dl->major == dk->major &&
1221                     dl->minor == dk->minor)
1222                         break;
1223         if (!dl || ! (dk->state & (1<<MD_DISK_SYNC)))
1224                 return;
1225
1226         map->disk_ord_tbl[dk->number] = __cpu_to_le32(dl->index);
1227
1228         disk = get_imsm_disk(mpb, dl->index);
1229         status = CONFIGURED_DISK | USABLE_DISK;
1230         disk->status = __cpu_to_le32(status);
1231 }
1232
1233 static int store_imsm_mpb(int fd, struct intel_super *super);
1234
1235 static int write_super_imsm(struct intel_super *super, int doclose)
1236 {
1237         struct imsm_super *mpb = super->mpb;
1238         struct dl *d;
1239         __u32 generation;
1240         __u32 sum;
1241
1242         /* 'generation' is incremented everytime the metadata is written */
1243         generation = __le32_to_cpu(mpb->generation_num);
1244         generation++;
1245         mpb->generation_num = __cpu_to_le32(generation);
1246
1247         /* recalculate checksum */
1248         sum = gen_imsm_checksum(mpb);
1249         mpb->check_sum = __cpu_to_le32(sum);
1250
1251         for (d = super->disks; d ; d = d->next) {
1252                 if (store_imsm_mpb(d->fd, super)) {
1253                         fprintf(stderr, "%s: failed for device %d:%d %s\n",
1254                                 __func__, d->major, d->minor, strerror(errno));
1255                         return 0;
1256                 }
1257                 if (doclose) {
1258                         close(d->fd);
1259                         d->fd = -1;
1260                 }
1261         }
1262
1263         return 1;
1264 }
1265
1266 static int write_init_super_imsm(struct supertype *st)
1267 {
1268         return write_super_imsm(st->sb, 1);
1269 }
1270
1271 static int store_zero_imsm(struct supertype *st, int fd)
1272 {
1273         unsigned long long dsize;
1274         char buf[512];
1275
1276         get_dev_size(fd, NULL, &dsize);
1277
1278         /* first block is stored on second to last sector of the disk */
1279         if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0)
1280                 return 1;
1281
1282         memset(buf, 0, sizeof(buf));
1283         if (write(fd, buf, sizeof(buf)) != sizeof(buf))
1284                 return 1;
1285
1286         return 0;
1287 }
1288
1289 static void getinfo_super_n_imsm_container(struct supertype *st, struct mdinfo *info)
1290 {
1291         /* just need offset and size...
1292          * of the metadata
1293          */
1294         struct intel_super *super = st->sb;
1295         struct imsm_super *mpb = super->mpb;
1296         struct imsm_disk *disk = get_imsm_disk(mpb, info->disk.number);
1297
1298         info->data_offset = __le32_to_cpu(disk->total_blocks) -
1299                             (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS);
1300         info->component_size = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
1301 }
1302
1303 static void getinfo_super_n_imsm_volume(struct supertype *st, struct mdinfo *info)
1304 {
1305         /* Find the particular details for info->disk.raid_disk.
1306          * This includes data_offset, component_size,
1307          */
1308         struct intel_super *super = st->sb;
1309         struct imsm_super *mpb = super->mpb;
1310         struct imsm_dev *dev = get_imsm_dev(mpb, super->creating_dev);
1311         struct imsm_map *map = &dev->vol.map[0];
1312
1313         info->data_offset = __le32_to_cpu(map->pba_of_lba0);
1314         info->component_size = __le32_to_cpu(map->blocks_per_member);
1315 }
1316
1317 static int validate_geometry_imsm(struct supertype *st, int level, int layout,
1318                                   int raiddisks, int chunk, unsigned long long size,
1319                                   char *dev, unsigned long long *freesize)
1320 {
1321         int fd, cfd;
1322         struct mdinfo *sra;
1323
1324         /* if given unused devices create a container 
1325          * if given given devices in a container create a member volume
1326          */
1327         if (level == LEVEL_CONTAINER) {
1328                 st->ss = &super_imsm_container;
1329                 if (dev) {
1330                         /* validate the container, dev == NULL */
1331                         int rv = st->ss->validate_geometry(st, level, layout,
1332                                                            raiddisks, chunk,
1333                                                            size,
1334                                                            NULL, freesize);
1335                         if (rv)
1336                                 return rv;
1337                 }
1338                 return st->ss->validate_geometry(st, level, layout, raiddisks,
1339                                                  chunk, size, dev, freesize);
1340         }
1341         
1342         if (st->sb) {
1343                 /* creating in a given container */
1344                 st->ss = &super_imsm_volume;
1345                 if (dev) {
1346                         int rv = st->ss->validate_geometry(st, level, layout,
1347                                                            raiddisks, chunk,
1348                                                            size,
1349                                                            NULL, freesize);
1350                         if (rv)
1351                                 return rv;
1352                 }
1353                 return st->ss->validate_geometry(st, level, layout, raiddisks,
1354                                                  chunk, size, dev, freesize);
1355         }
1356
1357         /* limit creation to the following levels */
1358         if (!dev)
1359                 switch (level) {
1360                 case 0:
1361                 case 1:
1362                 case 10:
1363                 case 5:
1364                         break;
1365                 default:
1366                         return 1;
1367                 }
1368
1369         /* This device needs to be a device in an 'imsm' container */
1370         fd = open(dev, O_RDONLY|O_EXCL, 0);
1371         if (fd >= 0) {
1372                 fprintf(stderr,
1373                         Name ": Cannot create this array on device %s\n",
1374                         dev);
1375                 close(fd);
1376                 return 0;
1377         }
1378         if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) {
1379                 fprintf(stderr, Name ": Cannot open %s: %s\n",
1380                         dev, strerror(errno));
1381                 return 0;
1382         }
1383         /* Well, it is in use by someone, maybe an 'imsm' container. */
1384         cfd = open_container(fd);
1385         if (cfd < 0) {
1386                 close(fd);
1387                 fprintf(stderr, Name ": Cannot use %s: It is busy\n",
1388                         dev);
1389                 return 0;
1390         }
1391         sra = sysfs_read(cfd, 0, GET_VERSION);
1392         close(fd);
1393         if (sra && sra->array.major_version == -1 &&
1394             strcmp(sra->text_version, "imsm") == 0) {
1395                 /* This is a member of a imsm container.  Load the container
1396                  * and try to create a volume
1397                  */
1398                 struct intel_super *super;
1399                 st->ss = &super_imsm_volume;
1400                 if (load_super_imsm_all(st, cfd, (void **) &super, NULL, 1) == 0) {
1401                         st->sb = super;
1402                         st->container_dev = fd2devnum(cfd);
1403                         close(cfd);
1404                         return st->ss->validate_geometry(st, level, layout,
1405                                                          raiddisks, chunk, size,
1406                                                          dev, freesize);
1407                 }
1408                 close(cfd);
1409         } else /* may belong to another container */
1410                 return 0;
1411
1412         return 1;
1413 }
1414
1415 static int validate_geometry_imsm_container(struct supertype *st, int level,
1416                                             int layout, int raiddisks, int chunk,
1417                                             unsigned long long size, char *dev,
1418                                             unsigned long long *freesize)
1419 {
1420         int fd;
1421         unsigned long long ldsize;
1422
1423         if (level != LEVEL_CONTAINER)
1424                 return 0;
1425         if (!dev)
1426                 return 1;
1427
1428         fd = open(dev, O_RDONLY|O_EXCL, 0);
1429         if (fd < 0) {
1430                 fprintf(stderr, Name ": Cannot open %s: %s\n",
1431                         dev, strerror(errno));
1432                 return 0;
1433         }
1434         if (!get_dev_size(fd, dev, &ldsize)) {
1435                 close(fd);
1436                 return 0;
1437         }
1438         close(fd);
1439
1440         *freesize = avail_size_imsm(st, ldsize >> 9);
1441
1442         return 1;
1443 }
1444
1445 /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd 
1446  * FIX ME add ahci details
1447  */
1448 static int validate_geometry_imsm_volume(struct supertype *st, int level,
1449                                          int layout, int raiddisks, int chunk,
1450                                          unsigned long long size, char *dev,
1451                                          unsigned long long *freesize)
1452 {
1453         struct stat stb;
1454         struct intel_super *super = st->sb;
1455         struct dl *dl;
1456         unsigned long long pos = 0;
1457         unsigned long long maxsize;
1458         struct extent *e;
1459         int i;
1460
1461         if (level == LEVEL_CONTAINER)
1462                 return 0;
1463
1464         if (level == 1 && raiddisks > 2) {
1465                 fprintf(stderr, Name ": imsm does not support more than 2 "
1466                         "in a raid1 configuration\n");
1467                 return 0;
1468         }
1469
1470         /* We must have the container info already read in. */
1471         if (!super)
1472                 return 0;
1473
1474         if (!dev) {
1475                 /* General test:  make sure there is space for
1476                  * 'raiddisks' device extents of size 'size'.
1477                  */
1478                 unsigned long long minsize = size*2 /* convert to blocks */;
1479                 int dcnt = 0;
1480                 if (minsize == 0)
1481                         minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
1482                 for (dl = super->disks; dl ; dl = dl->next) {
1483                         int found = 0;
1484
1485                         i = 0;
1486                         e = get_extents(super, dl);
1487                         if (!e) continue;
1488                         do {
1489                                 unsigned long long esize;
1490                                 esize = e[i].start - pos;
1491                                 if (esize >= minsize)
1492                                         found = 1;
1493                                 pos = e[i].start + e[i].size;
1494                                 i++;
1495                         } while (e[i-1].size);
1496                         if (found)
1497                                 dcnt++;
1498                         free(e);
1499                 }
1500                 if (dcnt < raiddisks) {
1501                         fprintf(stderr, Name ": Not enough devices with space "
1502                                 "for this array (%d < %d)\n",
1503                                 dcnt, raiddisks);
1504                         return 0;
1505                 }
1506                 return 1;
1507         }
1508         /* This device must be a member of the set */
1509         if (stat(dev, &stb) < 0)
1510                 return 0;
1511         if ((S_IFMT & stb.st_mode) != S_IFBLK)
1512                 return 0;
1513         for (dl = super->disks ; dl ; dl = dl->next) {
1514                 if (dl->major == major(stb.st_rdev) &&
1515                     dl->minor == minor(stb.st_rdev))
1516                         break;
1517         }
1518         if (!dl) {
1519                 fprintf(stderr, Name ": %s is not in the same imsm set\n",
1520                         dev);
1521                 return 0;
1522         }
1523         e = get_extents(super, dl);
1524         maxsize = 0;
1525         i = 0;
1526         if (e) do {
1527                 unsigned long long esize;
1528                 esize = e[i].start - pos;
1529                 if (esize >= maxsize)
1530                         maxsize = esize;
1531                 pos = e[i].start + e[i].size;
1532                 i++;
1533         } while (e[i-1].size);
1534         *freesize = maxsize;
1535
1536         return 1;
1537 }
1538
1539 static struct mdinfo *container_content_imsm(struct supertype *st)
1540 {
1541         /* Given a container loaded by load_super_imsm_all,
1542          * extract information about all the arrays into
1543          * an mdinfo tree.
1544          *
1545          * For each imsm_dev create an mdinfo, fill it in,
1546          *  then look for matching devices in super->disks
1547          *  and create appropriate device mdinfo.
1548          */
1549         struct intel_super *super = st->sb;
1550         struct imsm_super *mpb = super->mpb;
1551         struct mdinfo *rest = NULL;
1552         int i;
1553
1554         for (i = 0; i < mpb->num_raid_devs; i++) {
1555                 struct imsm_dev *dev = get_imsm_dev(mpb, i);
1556                 struct imsm_vol *vol = &dev->vol;
1557                 struct imsm_map *map = vol->map;
1558                 struct mdinfo *this;
1559                 __u64 sz;
1560                 int slot;
1561
1562                 this = malloc(sizeof(*this));
1563                 memset(this, 0, sizeof(*this));
1564                 this->next = rest;
1565                 rest = this;
1566
1567                 this->array.major_version = 2000;
1568                 get_imsm_numerical_version(mpb, &this->array.minor_version,
1569                                            &this->array.patch_version);
1570                 this->array.level = get_imsm_raid_level(map);
1571                 this->array.raid_disks = map->num_members;
1572                 this->array.layout = imsm_level_to_layout(this->array.level);
1573                 this->array.md_minor = -1;
1574                 this->array.ctime = 0;
1575                 this->array.utime = 0;
1576                 this->array.chunk_size = __le16_to_cpu(map->blocks_per_strip) << 9;
1577                 this->array.state = !vol->dirty;
1578                 this->container_member = i;
1579                 if (map->map_state == IMSM_T_STATE_UNINITIALIZED || dev->vol.dirty)
1580                         this->resync_start = 0;
1581                 else
1582                         this->resync_start = ~0ULL;
1583
1584                 strncpy(this->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN);
1585                 this->name[MAX_RAID_SERIAL_LEN] = 0;
1586
1587                 sprintf(this->text_version, "/%s/%d",
1588                         devnum2devname(st->container_dev),
1589                         this->container_member);
1590
1591                 memset(this->uuid, 0, sizeof(this->uuid));
1592
1593                 sz = __le32_to_cpu(dev->size_high);
1594                 sz <<= 32;
1595                 sz += __le32_to_cpu(dev->size_low);
1596                 this->component_size = sz;
1597                 this->array.size = this->component_size / 2;
1598
1599                 for (slot = 0 ; slot <  map->num_members; slot++) {
1600                         struct imsm_disk *disk;
1601                         struct mdinfo *info_d;
1602                         struct dl *d;
1603                         int idx;
1604                         __u32 s;
1605
1606                         idx = __le32_to_cpu(map->disk_ord_tbl[slot] & ~(0xff << 24));
1607                         for (d = super->disks; d ; d = d->next)
1608                                 if (d->index == idx)
1609                                         break;
1610
1611                         if (d == NULL)
1612                                 break; /* shouldn't this be continue ?? */
1613
1614                         info_d = malloc(sizeof(*info_d));
1615                         if (!info_d)
1616                                 break; /* ditto ?? */
1617                         memset(info_d, 0, sizeof(*info_d));
1618                         info_d->next = this->devs;
1619                         this->devs = info_d;
1620
1621                         disk = get_imsm_disk(mpb, idx);
1622                         s = __le32_to_cpu(disk->status);
1623
1624                         info_d->disk.number = d->index;
1625                         info_d->disk.major = d->major;
1626                         info_d->disk.minor = d->minor;
1627                         info_d->disk.raid_disk = slot;
1628                         info_d->disk.state  = s & CONFIGURED_DISK ? (1 << MD_DISK_ACTIVE) : 0;
1629                         info_d->disk.state |= s & FAILED_DISK ? (1 << MD_DISK_FAULTY) : 0;
1630                         info_d->disk.state |= s & USABLE_DISK ? (1 << MD_DISK_SYNC) : 0;
1631
1632                         this->array.working_disks++;
1633
1634                         info_d->events = __le32_to_cpu(mpb->generation_num);
1635                         info_d->data_offset = __le32_to_cpu(map->pba_of_lba0);
1636                         info_d->component_size = __le32_to_cpu(map->blocks_per_member);
1637                         if (d->devname)
1638                                 strcpy(info_d->name, d->devname);
1639                 }
1640         }
1641
1642         return rest;
1643 }
1644
1645
1646 static int imsm_open_new(struct supertype *c, struct active_array *a,
1647                          char *inst)
1648 {
1649         dprintf("imsm: open_new %s\n", inst);
1650         a->info.container_member = atoi(inst);
1651         return 0;
1652 }
1653
1654 static __u8 imsm_check_degraded(struct imsm_super *mpb, int n, int failed)
1655 {
1656         struct imsm_dev *dev = get_imsm_dev(mpb, n);
1657         struct imsm_map *map = dev->vol.map;
1658
1659         if (!failed)
1660                 return map->map_state;
1661
1662         switch (get_imsm_raid_level(map)) {
1663         case 0:
1664                 return IMSM_T_STATE_FAILED;
1665                 break;
1666         case 1:
1667                 if (failed < map->num_members)
1668                         return IMSM_T_STATE_DEGRADED;
1669                 else
1670                         return IMSM_T_STATE_FAILED;
1671                 break;
1672         case 10:
1673         {
1674                 /**
1675                  * check to see if any mirrors have failed,
1676                  * otherwise we are degraded
1677                  */
1678                 int device_per_mirror = 2; /* FIXME is this always the case?
1679                                             * and are they always adjacent?
1680                                             */
1681                 int failed = 0;
1682                 int i;
1683
1684                 for (i = 0; i < map->num_members; i++) {
1685                         int idx = get_imsm_disk_idx(map, i);
1686                         struct imsm_disk *disk = get_imsm_disk(mpb, idx);
1687
1688                         if (__le32_to_cpu(disk->status) & FAILED_DISK)
1689                                 failed++;
1690
1691                         if (failed >= device_per_mirror)
1692                                 return IMSM_T_STATE_FAILED;
1693
1694                         /* reset 'failed' for next mirror set */
1695                         if (!((i + 1) % device_per_mirror))
1696                                 failed = 0;
1697                 }
1698
1699                 return IMSM_T_STATE_DEGRADED;
1700         }
1701         case 5:
1702                 if (failed < 2)
1703                         return IMSM_T_STATE_DEGRADED;
1704                 else
1705                         return IMSM_T_STATE_FAILED;
1706                 break;
1707         default:
1708                 break;
1709         }
1710
1711         return map->map_state;
1712 }
1713
1714 static int imsm_count_failed(struct imsm_super *mpb, struct imsm_map *map)
1715 {
1716         int i;
1717         int failed = 0;
1718         struct imsm_disk *disk;
1719
1720         for (i = 0; i < map->num_members; i++) {
1721                 int idx = get_imsm_disk_idx(map, i);
1722
1723                 disk = get_imsm_disk(mpb, idx);
1724                 if (__le32_to_cpu(disk->status) & FAILED_DISK)
1725                         failed++;
1726         }
1727
1728         return failed;
1729 }
1730
1731 static void imsm_set_array_state(struct active_array *a, int consistent)
1732 {
1733         int inst = a->info.container_member;
1734         struct intel_super *super = a->container->sb;
1735         struct imsm_dev *dev = get_imsm_dev(super->mpb, inst);
1736         struct imsm_map *map = &dev->vol.map[0];
1737         int dirty = !consistent;
1738         int failed;
1739         __u8 map_state;
1740
1741         if (a->resync_start == ~0ULL) {
1742                 failed = imsm_count_failed(super->mpb, map);
1743                 map_state = imsm_check_degraded(super->mpb, inst, failed);
1744                 if (!failed)
1745                         map_state = IMSM_T_STATE_NORMAL;
1746                 if (map->map_state != map_state) {
1747                         dprintf("imsm: map_state %d: %d\n",
1748                                 inst, map_state);
1749                         map->map_state = map_state;
1750                         super->updates_pending++;
1751                 }
1752         }
1753
1754         if (dev->vol.dirty != dirty) {
1755                 dprintf("imsm: mark '%s' (%llu)\n",
1756                         dirty?"dirty":"clean", a->resync_start);
1757
1758                 dev->vol.dirty = dirty;
1759                 super->updates_pending++;
1760         }
1761 }
1762
1763 static void imsm_set_disk(struct active_array *a, int n, int state)
1764 {
1765         int inst = a->info.container_member;
1766         struct intel_super *super = a->container->sb;
1767         struct imsm_dev *dev = get_imsm_dev(super->mpb, inst);
1768         struct imsm_map *map = dev->vol.map;
1769         struct imsm_disk *disk;
1770         __u32 status;
1771         int failed = 0;
1772         int new_failure = 0;
1773
1774         if (n > map->num_members)
1775                 fprintf(stderr, "imsm: set_disk %d out of range 0..%d\n",
1776                         n, map->num_members - 1);
1777
1778         if (n < 0)
1779                 return;
1780
1781         dprintf("imsm: set_disk %d:%x\n", n, state);
1782
1783         disk = get_imsm_disk(super->mpb, get_imsm_disk_idx(map, n));
1784
1785         /* check if we have seen this failure before */
1786         status = __le32_to_cpu(disk->status);
1787         if ((state & DS_FAULTY) && !(status & FAILED_DISK)) {
1788                 status |= FAILED_DISK;
1789                 disk->status = __cpu_to_le32(status);
1790                 new_failure = 1;
1791         }
1792
1793         /**
1794          * the number of failures have changed, count up 'failed' to determine
1795          * degraded / failed status
1796          */
1797         if (new_failure && map->map_state != IMSM_T_STATE_FAILED)
1798                 failed = imsm_count_failed(super->mpb, map);
1799
1800         if (failed)
1801                 map->map_state = imsm_check_degraded(super->mpb, inst, failed);
1802
1803         if (new_failure)
1804                 super->updates_pending++;
1805 }
1806
1807 static int store_imsm_mpb(int fd, struct intel_super *super)
1808 {
1809         struct imsm_super *mpb = super->mpb;
1810         __u32 mpb_size = __le32_to_cpu(mpb->mpb_size);
1811         unsigned long long dsize;
1812         unsigned long long sectors;
1813
1814         get_dev_size(fd, NULL, &dsize);
1815
1816         if (mpb_size > 512) {
1817                 /* -1 to account for anchor */
1818                 sectors = mpb_sectors(mpb) - 1;
1819
1820                 /* write the extended mpb to the sectors preceeding the anchor */
1821                 if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0)
1822                         return 1;
1823
1824                 if (write(fd, super->buf + 512, mpb_size - 512) != mpb_size - 512)
1825                         return 1;
1826         }
1827
1828         /* first block is stored on second to last sector of the disk */
1829         if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0)
1830                 return 1;
1831
1832         if (write(fd, super->buf, 512) != 512)
1833                 return 1;
1834
1835         fsync(fd);
1836
1837         return 0;
1838 }
1839
1840 static void imsm_sync_metadata(struct supertype *container)
1841 {
1842         struct intel_super *super = container->sb;
1843
1844         if (!super->updates_pending)
1845                 return;
1846
1847         write_super_imsm(super, 0);
1848
1849         super->updates_pending = 0;
1850 }
1851
1852 struct superswitch super_imsm = {
1853 #ifndef MDASSEMBLE
1854         .examine_super  = examine_super_imsm,
1855         .brief_examine_super = brief_examine_super_imsm,
1856         .detail_super   = detail_super_imsm,
1857         .brief_detail_super = brief_detail_super_imsm,
1858 #endif
1859         .match_home     = match_home_imsm,
1860         .uuid_from_super= uuid_from_super_imsm,
1861         .getinfo_super  = getinfo_super_imsm,
1862         .update_super   = update_super_imsm,
1863
1864         .avail_size     = avail_size_imsm,
1865
1866         .compare_super  = compare_super_imsm,
1867
1868         .load_super     = load_super_imsm,
1869         .init_super     = init_zero_imsm,
1870         .store_super    = store_zero_imsm,
1871         .free_super     = free_super_imsm,
1872         .match_metadata_desc = match_metadata_desc_imsm,
1873         .getinfo_super_n  = getinfo_super_n_imsm_container,
1874
1875         .validate_geometry = validate_geometry_imsm,
1876         .major          = 2000,
1877         .swapuuid       = 0,
1878         .external       = 1,
1879
1880 /* for mdmon */
1881         .open_new       = imsm_open_new,
1882         .load_super     = load_super_imsm,
1883         .set_array_state= imsm_set_array_state,
1884         .set_disk       = imsm_set_disk,
1885         .sync_metadata  = imsm_sync_metadata,
1886 };
1887
1888 /* super_imsm_container is set by validate_geometry_imsm when given a
1889  * device that is not part of any array
1890  */
1891 struct superswitch super_imsm_container = {
1892
1893         .validate_geometry = validate_geometry_imsm_container,
1894         .init_super     = init_super_imsm,
1895         .add_to_super   = add_to_super_imsm,
1896         .write_init_super = write_init_super_imsm,
1897         .getinfo_super  = getinfo_super_imsm,
1898         .getinfo_super_n  = getinfo_super_n_imsm_container,
1899         .load_super     = load_super_imsm,
1900
1901 #ifndef MDASSEMBLE
1902         .examine_super  = examine_super_imsm,
1903         .brief_examine_super = brief_examine_super_imsm,
1904         .detail_super   = detail_super_imsm,
1905         .brief_detail_super = brief_detail_super_imsm,
1906 #endif
1907
1908         .free_super     = free_super_imsm,
1909
1910         .container_content = container_content_imsm,
1911
1912         .major          = 2000,
1913         .swapuuid       = 0,
1914         .external       = 1,
1915 };
1916
1917 struct superswitch super_imsm_volume = {
1918         .update_super   = update_super_imsm,
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,
1923         .write_init_super = write_init_super_imsm,
1924
1925         .load_super     = load_super_imsm,
1926         .free_super     = free_super_imsm,
1927         .match_metadata_desc = match_metadata_desc_imsm_volume,
1928
1929
1930         .validate_geometry = validate_geometry_imsm_volume,
1931         .major          = 2001,
1932         .swapuuid       = 0,
1933         .external       = 2,
1934 };