imsm: fix num_domains
[thirdparty/mdadm.git] / super-intel.c
1 /*
2 * mdadm - Intel(R) Matrix Storage Manager Support
3 *
4 * Copyright (C) 2002-2008 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 #define HAVE_STDINT_H 1
21 #include "mdadm.h"
22 #include "mdmon.h"
23 #include "sha1.h"
24 #include "platform-intel.h"
25 #include <values.h>
26 #include <scsi/sg.h>
27 #include <ctype.h>
28 #include <dirent.h>
29
30 /* MPB == Metadata Parameter Block */
31 #define MPB_SIGNATURE "Intel Raid ISM Cfg Sig. "
32 #define MPB_SIG_LEN (strlen(MPB_SIGNATURE))
33 #define MPB_VERSION_RAID0 "1.0.00"
34 #define MPB_VERSION_RAID1 "1.1.00"
35 #define MPB_VERSION_MANY_VOLUMES_PER_ARRAY "1.2.00"
36 #define MPB_VERSION_3OR4_DISK_ARRAY "1.2.01"
37 #define MPB_VERSION_RAID5 "1.2.02"
38 #define MPB_VERSION_5OR6_DISK_ARRAY "1.2.04"
39 #define MPB_VERSION_CNG "1.2.06"
40 #define MPB_VERSION_ATTRIBS "1.3.00"
41 #define MAX_SIGNATURE_LENGTH 32
42 #define MAX_RAID_SERIAL_LEN 16
43
44 #define MPB_ATTRIB_CHECKSUM_VERIFY __cpu_to_le32(0x80000000)
45 #define MPB_ATTRIB_PM __cpu_to_le32(0x40000000)
46 #define MPB_ATTRIB_2TB __cpu_to_le32(0x20000000)
47 #define MPB_ATTRIB_RAID0 __cpu_to_le32(0x00000001)
48 #define MPB_ATTRIB_RAID1 __cpu_to_le32(0x00000002)
49 #define MPB_ATTRIB_RAID10 __cpu_to_le32(0x00000004)
50 #define MPB_ATTRIB_RAID1E __cpu_to_le32(0x00000008)
51 #define MPB_ATTRIB_RAID5 __cpu_to_le32(0x00000010)
52 #define MPB_ATTRIB_RAIDCNG __cpu_to_le32(0x00000020)
53
54 #define MPB_SECTOR_CNT 418
55 #define IMSM_RESERVED_SECTORS 4096
56 #define SECT_PER_MB_SHIFT 11
57
58 /* Disk configuration info. */
59 #define IMSM_MAX_DEVICES 255
60 struct imsm_disk {
61 __u8 serial[MAX_RAID_SERIAL_LEN];/* 0xD8 - 0xE7 ascii serial number */
62 __u32 total_blocks; /* 0xE8 - 0xEB total blocks */
63 __u32 scsi_id; /* 0xEC - 0xEF scsi ID */
64 #define SPARE_DISK __cpu_to_le32(0x01) /* Spare */
65 #define CONFIGURED_DISK __cpu_to_le32(0x02) /* Member of some RaidDev */
66 #define FAILED_DISK __cpu_to_le32(0x04) /* Permanent failure */
67 #define USABLE_DISK __cpu_to_le32(0x08) /* Fully usable unless FAILED_DISK is set */
68 __u32 status; /* 0xF0 - 0xF3 */
69 __u32 owner_cfg_num; /* which config 0,1,2... owns this disk */
70 #define IMSM_DISK_FILLERS 4
71 __u32 filler[IMSM_DISK_FILLERS]; /* 0xF4 - 0x107 MPB_DISK_FILLERS for future expansion */
72 };
73
74 /* RAID map configuration infos. */
75 struct imsm_map {
76 __u32 pba_of_lba0; /* start address of partition */
77 __u32 blocks_per_member;/* blocks per member */
78 __u32 num_data_stripes; /* number of data stripes */
79 __u16 blocks_per_strip;
80 __u8 map_state; /* Normal, Uninitialized, Degraded, Failed */
81 #define IMSM_T_STATE_NORMAL 0
82 #define IMSM_T_STATE_UNINITIALIZED 1
83 #define IMSM_T_STATE_DEGRADED 2
84 #define IMSM_T_STATE_FAILED 3
85 __u8 raid_level;
86 #define IMSM_T_RAID0 0
87 #define IMSM_T_RAID1 1
88 #define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */
89 __u8 num_members; /* number of member disks */
90 __u8 num_domains; /* number of parity domains */
91 __u8 failed_disk_num; /* valid only when state is degraded */
92 __u8 ddf;
93 __u32 filler[7]; /* expansion area */
94 #define IMSM_ORD_REBUILD (1 << 24)
95 __u32 disk_ord_tbl[1]; /* disk_ord_tbl[num_members],
96 * top byte contains some flags
97 */
98 } __attribute__ ((packed));
99
100 struct imsm_vol {
101 __u32 curr_migr_unit;
102 __u32 checkpoint_id; /* id to access curr_migr_unit */
103 __u8 migr_state; /* Normal or Migrating */
104 #define MIGR_INIT 0
105 #define MIGR_REBUILD 1
106 #define MIGR_VERIFY 2 /* analagous to echo check > sync_action */
107 #define MIGR_GEN_MIGR 3
108 #define MIGR_STATE_CHANGE 4
109 #define MIGR_REPAIR 5
110 __u8 migr_type; /* Initializing, Rebuilding, ... */
111 __u8 dirty;
112 __u8 fs_state; /* fast-sync state for CnG (0xff == disabled) */
113 __u16 verify_errors; /* number of mismatches */
114 __u16 bad_blocks; /* number of bad blocks during verify */
115 __u32 filler[4];
116 struct imsm_map map[1];
117 /* here comes another one if migr_state */
118 } __attribute__ ((packed));
119
120 struct imsm_dev {
121 __u8 volume[MAX_RAID_SERIAL_LEN];
122 __u32 size_low;
123 __u32 size_high;
124 #define DEV_BOOTABLE __cpu_to_le32(0x01)
125 #define DEV_BOOT_DEVICE __cpu_to_le32(0x02)
126 #define DEV_READ_COALESCING __cpu_to_le32(0x04)
127 #define DEV_WRITE_COALESCING __cpu_to_le32(0x08)
128 #define DEV_LAST_SHUTDOWN_DIRTY __cpu_to_le32(0x10)
129 #define DEV_HIDDEN_AT_BOOT __cpu_to_le32(0x20)
130 #define DEV_CURRENTLY_HIDDEN __cpu_to_le32(0x40)
131 #define DEV_VERIFY_AND_FIX __cpu_to_le32(0x80)
132 #define DEV_MAP_STATE_UNINIT __cpu_to_le32(0x100)
133 #define DEV_NO_AUTO_RECOVERY __cpu_to_le32(0x200)
134 #define DEV_CLONE_N_GO __cpu_to_le32(0x400)
135 #define DEV_CLONE_MAN_SYNC __cpu_to_le32(0x800)
136 #define DEV_CNG_MASTER_DISK_NUM __cpu_to_le32(0x1000)
137 __u32 status; /* Persistent RaidDev status */
138 __u32 reserved_blocks; /* Reserved blocks at beginning of volume */
139 __u8 migr_priority;
140 __u8 num_sub_vols;
141 __u8 tid;
142 __u8 cng_master_disk;
143 __u16 cache_policy;
144 __u8 cng_state;
145 __u8 cng_sub_state;
146 #define IMSM_DEV_FILLERS 10
147 __u32 filler[IMSM_DEV_FILLERS];
148 struct imsm_vol vol;
149 } __attribute__ ((packed));
150
151 struct imsm_super {
152 __u8 sig[MAX_SIGNATURE_LENGTH]; /* 0x00 - 0x1F */
153 __u32 check_sum; /* 0x20 - 0x23 MPB Checksum */
154 __u32 mpb_size; /* 0x24 - 0x27 Size of MPB */
155 __u32 family_num; /* 0x28 - 0x2B Checksum from first time this config was written */
156 __u32 generation_num; /* 0x2C - 0x2F Incremented each time this array's MPB is written */
157 __u32 error_log_size; /* 0x30 - 0x33 in bytes */
158 __u32 attributes; /* 0x34 - 0x37 */
159 __u8 num_disks; /* 0x38 Number of configured disks */
160 __u8 num_raid_devs; /* 0x39 Number of configured volumes */
161 __u8 error_log_pos; /* 0x3A */
162 __u8 fill[1]; /* 0x3B */
163 __u32 cache_size; /* 0x3c - 0x40 in mb */
164 __u32 orig_family_num; /* 0x40 - 0x43 original family num */
165 __u32 pwr_cycle_count; /* 0x44 - 0x47 simulated power cycle count for array */
166 __u32 bbm_log_size; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */
167 #define IMSM_FILLERS 35
168 __u32 filler[IMSM_FILLERS]; /* 0x4C - 0xD7 RAID_MPB_FILLERS */
169 struct imsm_disk disk[1]; /* 0xD8 diskTbl[numDisks] */
170 /* here comes imsm_dev[num_raid_devs] */
171 /* here comes BBM logs */
172 } __attribute__ ((packed));
173
174 #define BBM_LOG_MAX_ENTRIES 254
175
176 struct bbm_log_entry {
177 __u64 defective_block_start;
178 #define UNREADABLE 0xFFFFFFFF
179 __u32 spare_block_offset;
180 __u16 remapped_marked_count;
181 __u16 disk_ordinal;
182 } __attribute__ ((__packed__));
183
184 struct bbm_log {
185 __u32 signature; /* 0xABADB10C */
186 __u32 entry_count;
187 __u32 reserved_spare_block_count; /* 0 */
188 __u32 reserved; /* 0xFFFF */
189 __u64 first_spare_lba;
190 struct bbm_log_entry mapped_block_entries[BBM_LOG_MAX_ENTRIES];
191 } __attribute__ ((__packed__));
192
193
194 #ifndef MDASSEMBLE
195 static char *map_state_str[] = { "normal", "uninitialized", "degraded", "failed" };
196 #endif
197
198 static __u8 migr_type(struct imsm_dev *dev)
199 {
200 if (dev->vol.migr_type == MIGR_VERIFY &&
201 dev->status & DEV_VERIFY_AND_FIX)
202 return MIGR_REPAIR;
203 else
204 return dev->vol.migr_type;
205 }
206
207 static void set_migr_type(struct imsm_dev *dev, __u8 migr_type)
208 {
209 /* for compatibility with older oroms convert MIGR_REPAIR, into
210 * MIGR_VERIFY w/ DEV_VERIFY_AND_FIX status
211 */
212 if (migr_type == MIGR_REPAIR) {
213 dev->vol.migr_type = MIGR_VERIFY;
214 dev->status |= DEV_VERIFY_AND_FIX;
215 } else {
216 dev->vol.migr_type = migr_type;
217 dev->status &= ~DEV_VERIFY_AND_FIX;
218 }
219 }
220
221 static unsigned int sector_count(__u32 bytes)
222 {
223 return ((bytes + (512-1)) & (~(512-1))) / 512;
224 }
225
226 static unsigned int mpb_sectors(struct imsm_super *mpb)
227 {
228 return sector_count(__le32_to_cpu(mpb->mpb_size));
229 }
230
231 struct intel_dev {
232 struct imsm_dev *dev;
233 struct intel_dev *next;
234 int index;
235 };
236
237 /* internal representation of IMSM metadata */
238 struct intel_super {
239 union {
240 void *buf; /* O_DIRECT buffer for reading/writing metadata */
241 struct imsm_super *anchor; /* immovable parameters */
242 };
243 size_t len; /* size of the 'buf' allocation */
244 void *next_buf; /* for realloc'ing buf from the manager */
245 size_t next_len;
246 int updates_pending; /* count of pending updates for mdmon */
247 int creating_imsm; /* flag to indicate container creation */
248 int current_vol; /* index of raid device undergoing creation */
249 __u32 create_offset; /* common start for 'current_vol' */
250 struct intel_dev *devlist;
251 struct dl {
252 struct dl *next;
253 int index;
254 __u8 serial[MAX_RAID_SERIAL_LEN];
255 int major, minor;
256 char *devname;
257 struct imsm_disk disk;
258 int fd;
259 int extent_cnt;
260 struct extent *e; /* for determining freespace @ create */
261 int raiddisk; /* slot to fill in autolayout */
262 } *disks;
263 struct dl *add; /* list of disks to add while mdmon active */
264 struct dl *missing; /* disks removed while we weren't looking */
265 struct bbm_log *bbm_log;
266 const char *hba; /* device path of the raid controller for this metadata */
267 const struct imsm_orom *orom; /* platform firmware support */
268 };
269
270 struct extent {
271 unsigned long long start, size;
272 };
273
274 /* definition of messages passed to imsm_process_update */
275 enum imsm_update_type {
276 update_activate_spare,
277 update_create_array,
278 update_add_disk,
279 };
280
281 struct imsm_update_activate_spare {
282 enum imsm_update_type type;
283 struct dl *dl;
284 int slot;
285 int array;
286 struct imsm_update_activate_spare *next;
287 };
288
289 struct disk_info {
290 __u8 serial[MAX_RAID_SERIAL_LEN];
291 };
292
293 struct imsm_update_create_array {
294 enum imsm_update_type type;
295 int dev_idx;
296 struct imsm_dev dev;
297 };
298
299 struct imsm_update_add_disk {
300 enum imsm_update_type type;
301 };
302
303 static struct supertype *match_metadata_desc_imsm(char *arg)
304 {
305 struct supertype *st;
306
307 if (strcmp(arg, "imsm") != 0 &&
308 strcmp(arg, "default") != 0
309 )
310 return NULL;
311
312 st = malloc(sizeof(*st));
313 memset(st, 0, sizeof(*st));
314 st->ss = &super_imsm;
315 st->max_devs = IMSM_MAX_DEVICES;
316 st->minor_version = 0;
317 st->sb = NULL;
318 return st;
319 }
320
321 #ifndef MDASSEMBLE
322 static __u8 *get_imsm_version(struct imsm_super *mpb)
323 {
324 return &mpb->sig[MPB_SIG_LEN];
325 }
326 #endif
327
328 /* retrieve a disk directly from the anchor when the anchor is known to be
329 * up-to-date, currently only at load time
330 */
331 static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index)
332 {
333 if (index >= mpb->num_disks)
334 return NULL;
335 return &mpb->disk[index];
336 }
337
338 #ifndef MDASSEMBLE
339 /* retrieve a disk from the parsed metadata */
340 static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index)
341 {
342 struct dl *d;
343
344 for (d = super->disks; d; d = d->next)
345 if (d->index == index)
346 return &d->disk;
347
348 return NULL;
349 }
350 #endif
351
352 /* generate a checksum directly from the anchor when the anchor is known to be
353 * up-to-date, currently only at load or write_super after coalescing
354 */
355 static __u32 __gen_imsm_checksum(struct imsm_super *mpb)
356 {
357 __u32 end = mpb->mpb_size / sizeof(end);
358 __u32 *p = (__u32 *) mpb;
359 __u32 sum = 0;
360
361 while (end--) {
362 sum += __le32_to_cpu(*p);
363 p++;
364 }
365
366 return sum - __le32_to_cpu(mpb->check_sum);
367 }
368
369 static size_t sizeof_imsm_map(struct imsm_map *map)
370 {
371 return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1);
372 }
373
374 struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map)
375 {
376 struct imsm_map *map = &dev->vol.map[0];
377
378 if (second_map && !dev->vol.migr_state)
379 return NULL;
380 else if (second_map) {
381 void *ptr = map;
382
383 return ptr + sizeof_imsm_map(map);
384 } else
385 return map;
386
387 }
388
389 /* return the size of the device.
390 * migr_state increases the returned size if map[0] were to be duplicated
391 */
392 static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state)
393 {
394 size_t size = sizeof(*dev) - sizeof(struct imsm_map) +
395 sizeof_imsm_map(get_imsm_map(dev, 0));
396
397 /* migrating means an additional map */
398 if (dev->vol.migr_state)
399 size += sizeof_imsm_map(get_imsm_map(dev, 1));
400 else if (migr_state)
401 size += sizeof_imsm_map(get_imsm_map(dev, 0));
402
403 return size;
404 }
405
406 #ifndef MDASSEMBLE
407 /* retrieve disk serial number list from a metadata update */
408 static struct disk_info *get_disk_info(struct imsm_update_create_array *update)
409 {
410 void *u = update;
411 struct disk_info *inf;
412
413 inf = u + sizeof(*update) - sizeof(struct imsm_dev) +
414 sizeof_imsm_dev(&update->dev, 0);
415
416 return inf;
417 }
418 #endif
419
420 static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index)
421 {
422 int offset;
423 int i;
424 void *_mpb = mpb;
425
426 if (index >= mpb->num_raid_devs)
427 return NULL;
428
429 /* devices start after all disks */
430 offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb;
431
432 for (i = 0; i <= index; i++)
433 if (i == index)
434 return _mpb + offset;
435 else
436 offset += sizeof_imsm_dev(_mpb + offset, 0);
437
438 return NULL;
439 }
440
441 static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index)
442 {
443 struct intel_dev *dv;
444
445 if (index >= super->anchor->num_raid_devs)
446 return NULL;
447 for (dv = super->devlist; dv; dv = dv->next)
448 if (dv->index == index)
449 return dv->dev;
450 return NULL;
451 }
452
453 static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev, int slot)
454 {
455 struct imsm_map *map;
456
457 if (dev->vol.migr_state)
458 map = get_imsm_map(dev, 1);
459 else
460 map = get_imsm_map(dev, 0);
461
462 /* top byte identifies disk under rebuild */
463 return __le32_to_cpu(map->disk_ord_tbl[slot]);
464 }
465
466 #define ord_to_idx(ord) (((ord) << 8) >> 8)
467 static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot)
468 {
469 __u32 ord = get_imsm_ord_tbl_ent(dev, slot);
470
471 return ord_to_idx(ord);
472 }
473
474 static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord)
475 {
476 map->disk_ord_tbl[slot] = __cpu_to_le32(ord);
477 }
478
479 static int get_imsm_disk_slot(struct imsm_map *map, int idx)
480 {
481 int slot;
482 __u32 ord;
483
484 for (slot = 0; slot < map->num_members; slot++) {
485 ord = __le32_to_cpu(map->disk_ord_tbl[slot]);
486 if (ord_to_idx(ord) == idx)
487 return slot;
488 }
489
490 return -1;
491 }
492
493 static int get_imsm_raid_level(struct imsm_map *map)
494 {
495 if (map->raid_level == 1) {
496 if (map->num_members == 2)
497 return 1;
498 else
499 return 10;
500 }
501
502 return map->raid_level;
503 }
504
505 static int cmp_extent(const void *av, const void *bv)
506 {
507 const struct extent *a = av;
508 const struct extent *b = bv;
509 if (a->start < b->start)
510 return -1;
511 if (a->start > b->start)
512 return 1;
513 return 0;
514 }
515
516 static int count_memberships(struct dl *dl, struct intel_super *super)
517 {
518 int memberships = 0;
519 int i;
520
521 for (i = 0; i < super->anchor->num_raid_devs; i++) {
522 struct imsm_dev *dev = get_imsm_dev(super, i);
523 struct imsm_map *map = get_imsm_map(dev, 0);
524
525 if (get_imsm_disk_slot(map, dl->index) >= 0)
526 memberships++;
527 }
528
529 return memberships;
530 }
531
532 static struct extent *get_extents(struct intel_super *super, struct dl *dl)
533 {
534 /* find a list of used extents on the given physical device */
535 struct extent *rv, *e;
536 int i;
537 int memberships = count_memberships(dl, super);
538 __u32 reservation = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
539
540 rv = malloc(sizeof(struct extent) * (memberships + 1));
541 if (!rv)
542 return NULL;
543 e = rv;
544
545 for (i = 0; i < super->anchor->num_raid_devs; i++) {
546 struct imsm_dev *dev = get_imsm_dev(super, i);
547 struct imsm_map *map = get_imsm_map(dev, 0);
548
549 if (get_imsm_disk_slot(map, dl->index) >= 0) {
550 e->start = __le32_to_cpu(map->pba_of_lba0);
551 e->size = __le32_to_cpu(map->blocks_per_member);
552 e++;
553 }
554 }
555 qsort(rv, memberships, sizeof(*rv), cmp_extent);
556
557 /* determine the start of the metadata
558 * when no raid devices are defined use the default
559 * ...otherwise allow the metadata to truncate the value
560 * as is the case with older versions of imsm
561 */
562 if (memberships) {
563 struct extent *last = &rv[memberships - 1];
564 __u32 remainder;
565
566 remainder = __le32_to_cpu(dl->disk.total_blocks) -
567 (last->start + last->size);
568 /* round down to 1k block to satisfy precision of the kernel
569 * 'size' interface
570 */
571 remainder &= ~1UL;
572 /* make sure remainder is still sane */
573 if (remainder < ROUND_UP(super->len, 512) >> 9)
574 remainder = ROUND_UP(super->len, 512) >> 9;
575 if (reservation > remainder)
576 reservation = remainder;
577 }
578 e->start = __le32_to_cpu(dl->disk.total_blocks) - reservation;
579 e->size = 0;
580 return rv;
581 }
582
583 /* try to determine how much space is reserved for metadata from
584 * the last get_extents() entry, otherwise fallback to the
585 * default
586 */
587 static __u32 imsm_reserved_sectors(struct intel_super *super, struct dl *dl)
588 {
589 struct extent *e;
590 int i;
591 __u32 rv;
592
593 /* for spares just return a minimal reservation which will grow
594 * once the spare is picked up by an array
595 */
596 if (dl->index == -1)
597 return MPB_SECTOR_CNT;
598
599 e = get_extents(super, dl);
600 if (!e)
601 return MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
602
603 /* scroll to last entry */
604 for (i = 0; e[i].size; i++)
605 continue;
606
607 rv = __le32_to_cpu(dl->disk.total_blocks) - e[i].start;
608
609 free(e);
610
611 return rv;
612 }
613
614 #ifndef MDASSEMBLE
615 static void print_imsm_dev(struct imsm_dev *dev, char *uuid, int disk_idx)
616 {
617 __u64 sz;
618 int slot;
619 struct imsm_map *map = get_imsm_map(dev, 0);
620 __u32 ord;
621
622 printf("\n");
623 printf("[%.16s]:\n", dev->volume);
624 printf(" UUID : %s\n", uuid);
625 printf(" RAID Level : %d\n", get_imsm_raid_level(map));
626 printf(" Members : %d\n", map->num_members);
627 slot = get_imsm_disk_slot(map, disk_idx);
628 if (slot >= 0) {
629 ord = get_imsm_ord_tbl_ent(dev, slot);
630 printf(" This Slot : %d%s\n", slot,
631 ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : "");
632 } else
633 printf(" This Slot : ?\n");
634 sz = __le32_to_cpu(dev->size_high);
635 sz <<= 32;
636 sz += __le32_to_cpu(dev->size_low);
637 printf(" Array Size : %llu%s\n", (unsigned long long)sz,
638 human_size(sz * 512));
639 sz = __le32_to_cpu(map->blocks_per_member);
640 printf(" Per Dev Size : %llu%s\n", (unsigned long long)sz,
641 human_size(sz * 512));
642 printf(" Sector Offset : %u\n",
643 __le32_to_cpu(map->pba_of_lba0));
644 printf(" Num Stripes : %u\n",
645 __le32_to_cpu(map->num_data_stripes));
646 printf(" Chunk Size : %u KiB\n",
647 __le16_to_cpu(map->blocks_per_strip) / 2);
648 printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks));
649 printf(" Migrate State : %s", dev->vol.migr_state ? "migrating" : "idle\n");
650 if (dev->vol.migr_state) {
651 if (migr_type(dev) == MIGR_INIT)
652 printf(": initializing\n");
653 else if (migr_type(dev) == MIGR_REBUILD)
654 printf(": rebuilding\n");
655 else if (migr_type(dev) == MIGR_VERIFY)
656 printf(": check\n");
657 else if (migr_type(dev) == MIGR_GEN_MIGR)
658 printf(": general migration\n");
659 else if (migr_type(dev) == MIGR_STATE_CHANGE)
660 printf(": state change\n");
661 else if (migr_type(dev) == MIGR_REPAIR)
662 printf(": repair\n");
663 else
664 printf(": <unknown:%d>\n", migr_type(dev));
665 }
666 printf(" Map State : %s", map_state_str[map->map_state]);
667 if (dev->vol.migr_state) {
668 struct imsm_map *map = get_imsm_map(dev, 1);
669 printf(" <-- %s", map_state_str[map->map_state]);
670 }
671 printf("\n");
672 printf(" Dirty State : %s\n", dev->vol.dirty ? "dirty" : "clean");
673 }
674
675 static void print_imsm_disk(struct imsm_super *mpb, int index, __u32 reserved)
676 {
677 struct imsm_disk *disk = __get_imsm_disk(mpb, index);
678 char str[MAX_RAID_SERIAL_LEN + 1];
679 __u32 s;
680 __u64 sz;
681
682 if (index < 0)
683 return;
684
685 printf("\n");
686 snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial);
687 printf(" Disk%02d Serial : %s\n", index, str);
688 s = disk->status;
689 printf(" State :%s%s%s%s\n", s&SPARE_DISK ? " spare" : "",
690 s&CONFIGURED_DISK ? " active" : "",
691 s&FAILED_DISK ? " failed" : "",
692 s&USABLE_DISK ? " usable" : "");
693 printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id));
694 sz = __le32_to_cpu(disk->total_blocks) - reserved;
695 printf(" Usable Size : %llu%s\n", (unsigned long long)sz,
696 human_size(sz * 512));
697 }
698
699 static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info);
700
701 static void examine_super_imsm(struct supertype *st, char *homehost)
702 {
703 struct intel_super *super = st->sb;
704 struct imsm_super *mpb = super->anchor;
705 char str[MAX_SIGNATURE_LENGTH];
706 int i;
707 struct mdinfo info;
708 char nbuf[64];
709 __u32 sum;
710 __u32 reserved = imsm_reserved_sectors(super, super->disks);
711
712
713 snprintf(str, MPB_SIG_LEN, "%s", mpb->sig);
714 printf(" Magic : %s\n", str);
715 snprintf(str, strlen(MPB_VERSION_RAID0), "%s", get_imsm_version(mpb));
716 printf(" Version : %s\n", get_imsm_version(mpb));
717 printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num));
718 printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num));
719 getinfo_super_imsm(st, &info);
720 fname_from_uuid(st, &info, nbuf, ':');
721 printf(" UUID : %s\n", nbuf + 5);
722 sum = __le32_to_cpu(mpb->check_sum);
723 printf(" Checksum : %08x %s\n", sum,
724 __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect");
725 printf(" MPB Sectors : %d\n", mpb_sectors(mpb));
726 printf(" Disks : %d\n", mpb->num_disks);
727 printf(" RAID Devices : %d\n", mpb->num_raid_devs);
728 print_imsm_disk(mpb, super->disks->index, reserved);
729 if (super->bbm_log) {
730 struct bbm_log *log = super->bbm_log;
731
732 printf("\n");
733 printf("Bad Block Management Log:\n");
734 printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size));
735 printf(" Signature : %x\n", __le32_to_cpu(log->signature));
736 printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count));
737 printf(" Spare Blocks : %d\n", __le32_to_cpu(log->reserved_spare_block_count));
738 printf(" First Spare : %llx\n",
739 (unsigned long long) __le64_to_cpu(log->first_spare_lba));
740 }
741 for (i = 0; i < mpb->num_raid_devs; i++) {
742 struct mdinfo info;
743 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
744
745 super->current_vol = i;
746 getinfo_super_imsm(st, &info);
747 fname_from_uuid(st, &info, nbuf, ':');
748 print_imsm_dev(dev, nbuf + 5, super->disks->index);
749 }
750 for (i = 0; i < mpb->num_disks; i++) {
751 if (i == super->disks->index)
752 continue;
753 print_imsm_disk(mpb, i, reserved);
754 }
755 }
756
757 static void brief_examine_super_imsm(struct supertype *st, int verbose)
758 {
759 /* We just write a generic IMSM ARRAY entry */
760 struct mdinfo info;
761 char nbuf[64];
762 char nbuf1[64];
763 struct intel_super *super = st->sb;
764 int i;
765
766 if (!super->anchor->num_raid_devs)
767 return;
768
769 getinfo_super_imsm(st, &info);
770 fname_from_uuid(st, &info, nbuf, ':');
771 printf("ARRAY metadata=imsm auto=md UUID=%s\n", nbuf + 5);
772 for (i = 0; i < super->anchor->num_raid_devs; i++) {
773 struct imsm_dev *dev = get_imsm_dev(super, i);
774
775 super->current_vol = i;
776 getinfo_super_imsm(st, &info);
777 fname_from_uuid(st, &info, nbuf1, ':');
778 printf("ARRAY /dev/md/%.16s container=%s\n"
779 " member=%d auto=mdp UUID=%s\n",
780 dev->volume, nbuf + 5, i, nbuf1 + 5);
781 }
782 }
783
784 static void export_examine_super_imsm(struct supertype *st)
785 {
786 struct intel_super *super = st->sb;
787 struct imsm_super *mpb = super->anchor;
788 struct mdinfo info;
789 char nbuf[64];
790
791 getinfo_super_imsm(st, &info);
792 fname_from_uuid(st, &info, nbuf, ':');
793 printf("MD_METADATA=imsm\n");
794 printf("MD_LEVEL=container\n");
795 printf("MD_UUID=%s\n", nbuf+5);
796 printf("MD_DEVICES=%u\n", mpb->num_disks);
797 }
798
799 static void detail_super_imsm(struct supertype *st, char *homehost)
800 {
801 struct mdinfo info;
802 char nbuf[64];
803
804 getinfo_super_imsm(st, &info);
805 fname_from_uuid(st, &info, nbuf, ':');
806 printf("\n UUID : %s\n", nbuf + 5);
807 }
808
809 static void brief_detail_super_imsm(struct supertype *st)
810 {
811 struct mdinfo info;
812 char nbuf[64];
813 getinfo_super_imsm(st, &info);
814 fname_from_uuid(st, &info, nbuf, ':');
815 printf(" UUID=%s", nbuf + 5);
816 }
817
818 static int imsm_read_serial(int fd, char *devname, __u8 *serial);
819 static void fd2devname(int fd, char *name);
820
821 static int imsm_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose)
822 {
823 /* dump an unsorted list of devices attached to ahci, as well as
824 * non-connected ports
825 */
826 int hba_len = strlen(hba_path) + 1;
827 struct dirent *ent;
828 DIR *dir;
829 char *path = NULL;
830 int err = 0;
831 unsigned long port_mask = (1 << port_count) - 1;
832
833 if (port_count > sizeof(port_mask) * 8) {
834 if (verbose)
835 fprintf(stderr, Name ": port_count %d out of range\n", port_count);
836 return 2;
837 }
838
839 /* scroll through /sys/dev/block looking for devices attached to
840 * this hba
841 */
842 dir = opendir("/sys/dev/block");
843 for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) {
844 int fd;
845 char model[64];
846 char vendor[64];
847 char buf[1024];
848 int major, minor;
849 char *device;
850 char *c;
851 int port;
852 int type;
853
854 if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2)
855 continue;
856 path = devt_to_devpath(makedev(major, minor));
857 if (!path)
858 continue;
859 if (!path_attached_to_hba(path, hba_path)) {
860 free(path);
861 path = NULL;
862 continue;
863 }
864
865 /* retrieve the scsi device type */
866 if (asprintf(&device, "/sys/dev/block/%d:%d/device/xxxxxxx", major, minor) < 0) {
867 if (verbose)
868 fprintf(stderr, Name ": failed to allocate 'device'\n");
869 err = 2;
870 break;
871 }
872 sprintf(device, "/sys/dev/block/%d:%d/device/type", major, minor);
873 if (load_sys(device, buf) != 0) {
874 if (verbose)
875 fprintf(stderr, Name ": failed to read device type for %s\n",
876 path);
877 err = 2;
878 free(device);
879 break;
880 }
881 type = strtoul(buf, NULL, 10);
882
883 /* if it's not a disk print the vendor and model */
884 if (!(type == 0 || type == 7 || type == 14)) {
885 vendor[0] = '\0';
886 model[0] = '\0';
887 sprintf(device, "/sys/dev/block/%d:%d/device/vendor", major, minor);
888 if (load_sys(device, buf) == 0) {
889 strncpy(vendor, buf, sizeof(vendor));
890 vendor[sizeof(vendor) - 1] = '\0';
891 c = (char *) &vendor[sizeof(vendor) - 1];
892 while (isspace(*c) || *c == '\0')
893 *c-- = '\0';
894
895 }
896 sprintf(device, "/sys/dev/block/%d:%d/device/model", major, minor);
897 if (load_sys(device, buf) == 0) {
898 strncpy(model, buf, sizeof(model));
899 model[sizeof(model) - 1] = '\0';
900 c = (char *) &model[sizeof(model) - 1];
901 while (isspace(*c) || *c == '\0')
902 *c-- = '\0';
903 }
904
905 if (vendor[0] && model[0])
906 sprintf(buf, "%.64s %.64s", vendor, model);
907 else
908 switch (type) { /* numbers from hald/linux/device.c */
909 case 1: sprintf(buf, "tape"); break;
910 case 2: sprintf(buf, "printer"); break;
911 case 3: sprintf(buf, "processor"); break;
912 case 4:
913 case 5: sprintf(buf, "cdrom"); break;
914 case 6: sprintf(buf, "scanner"); break;
915 case 8: sprintf(buf, "media_changer"); break;
916 case 9: sprintf(buf, "comm"); break;
917 case 12: sprintf(buf, "raid"); break;
918 default: sprintf(buf, "unknown");
919 }
920 } else
921 buf[0] = '\0';
922 free(device);
923
924 /* chop device path to 'host%d' and calculate the port number */
925 c = strchr(&path[hba_len], '/');
926 *c = '\0';
927 if (sscanf(&path[hba_len], "host%d", &port) == 1)
928 port -= host_base;
929 else {
930 if (verbose) {
931 *c = '/'; /* repair the full string */
932 fprintf(stderr, Name ": failed to determine port number for %s\n",
933 path);
934 }
935 err = 2;
936 break;
937 }
938
939 /* mark this port as used */
940 port_mask &= ~(1 << port);
941
942 /* print out the device information */
943 if (buf[0]) {
944 printf(" Port%d : - non-disk device (%s) -\n", port, buf);
945 continue;
946 }
947
948 fd = dev_open(ent->d_name, O_RDONLY);
949 if (fd < 0)
950 printf(" Port%d : - disk info unavailable -\n", port);
951 else {
952 fd2devname(fd, buf);
953 printf(" Port%d : %s", port, buf);
954 if (imsm_read_serial(fd, NULL, (__u8 *) buf) == 0)
955 printf(" (%s)\n", buf);
956 else
957 printf("()\n");
958 }
959 close(fd);
960 free(path);
961 path = NULL;
962 }
963 if (path)
964 free(path);
965 if (dir)
966 closedir(dir);
967 if (err == 0) {
968 int i;
969
970 for (i = 0; i < port_count; i++)
971 if (port_mask & (1 << i))
972 printf(" Port%d : - no device attached -\n", i);
973 }
974
975 return err;
976 }
977
978 static int detail_platform_imsm(int verbose, int enumerate_only)
979 {
980 /* There are two components to imsm platform support, the ahci SATA
981 * controller and the option-rom. To find the SATA controller we
982 * simply look in /sys/bus/pci/drivers/ahci to see if an ahci
983 * controller with the Intel vendor id is present. This approach
984 * allows mdadm to leverage the kernel's ahci detection logic, with the
985 * caveat that if ahci.ko is not loaded mdadm will not be able to
986 * detect platform raid capabilities. The option-rom resides in a
987 * platform "Adapter ROM". We scan for its signature to retrieve the
988 * platform capabilities. If raid support is disabled in the BIOS the
989 * option-rom capability structure will not be available.
990 */
991 const struct imsm_orom *orom;
992 struct sys_dev *list, *hba;
993 DIR *dir;
994 struct dirent *ent;
995 const char *hba_path;
996 int host_base = 0;
997 int port_count = 0;
998
999 if (enumerate_only) {
1000 if (check_env("IMSM_NO_PLATFORM") || find_imsm_orom())
1001 return 0;
1002 return 2;
1003 }
1004
1005 list = find_driver_devices("pci", "ahci");
1006 for (hba = list; hba; hba = hba->next)
1007 if (devpath_to_vendor(hba->path) == 0x8086)
1008 break;
1009
1010 if (!hba) {
1011 if (verbose)
1012 fprintf(stderr, Name ": unable to find active ahci controller\n");
1013 free_sys_dev(&list);
1014 return 2;
1015 } else if (verbose)
1016 fprintf(stderr, Name ": found Intel SATA AHCI Controller\n");
1017 hba_path = hba->path;
1018 hba->path = NULL;
1019 free_sys_dev(&list);
1020
1021 orom = find_imsm_orom();
1022 if (!orom) {
1023 if (verbose)
1024 fprintf(stderr, Name ": imsm option-rom not found\n");
1025 return 2;
1026 }
1027
1028 printf(" Platform : Intel(R) Matrix Storage Manager\n");
1029 printf(" Version : %d.%d.%d.%d\n", orom->major_ver, orom->minor_ver,
1030 orom->hotfix_ver, orom->build);
1031 printf(" RAID Levels :%s%s%s%s%s\n",
1032 imsm_orom_has_raid0(orom) ? " raid0" : "",
1033 imsm_orom_has_raid1(orom) ? " raid1" : "",
1034 imsm_orom_has_raid1e(orom) ? " raid1e" : "",
1035 imsm_orom_has_raid10(orom) ? " raid10" : "",
1036 imsm_orom_has_raid5(orom) ? " raid5" : "");
1037 printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
1038 imsm_orom_has_chunk(orom, 2) ? " 2k" : "",
1039 imsm_orom_has_chunk(orom, 4) ? " 4k" : "",
1040 imsm_orom_has_chunk(orom, 8) ? " 8k" : "",
1041 imsm_orom_has_chunk(orom, 16) ? " 16k" : "",
1042 imsm_orom_has_chunk(orom, 32) ? " 32k" : "",
1043 imsm_orom_has_chunk(orom, 64) ? " 64k" : "",
1044 imsm_orom_has_chunk(orom, 128) ? " 128k" : "",
1045 imsm_orom_has_chunk(orom, 256) ? " 256k" : "",
1046 imsm_orom_has_chunk(orom, 512) ? " 512k" : "",
1047 imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "",
1048 imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "",
1049 imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "",
1050 imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "",
1051 imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "",
1052 imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "",
1053 imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : "");
1054 printf(" Max Disks : %d\n", orom->tds);
1055 printf(" Max Volumes : %d\n", orom->vpa);
1056 printf(" I/O Controller : %s\n", hba_path);
1057
1058 /* find the smallest scsi host number to determine a port number base */
1059 dir = opendir(hba_path);
1060 for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) {
1061 int host;
1062
1063 if (sscanf(ent->d_name, "host%d", &host) != 1)
1064 continue;
1065 if (port_count == 0)
1066 host_base = host;
1067 else if (host < host_base)
1068 host_base = host;
1069
1070 if (host + 1 > port_count + host_base)
1071 port_count = host + 1 - host_base;
1072
1073 }
1074 if (dir)
1075 closedir(dir);
1076
1077 if (!port_count || imsm_enumerate_ports(hba_path, port_count,
1078 host_base, verbose) != 0) {
1079 if (verbose)
1080 fprintf(stderr, Name ": failed to enumerate ports\n");
1081 return 2;
1082 }
1083
1084 return 0;
1085 }
1086 #endif
1087
1088 static int match_home_imsm(struct supertype *st, char *homehost)
1089 {
1090 /* the imsm metadata format does not specify any host
1091 * identification information. We return -1 since we can never
1092 * confirm nor deny whether a given array is "meant" for this
1093 * host. We rely on compare_super and the 'family_num' field to
1094 * exclude member disks that do not belong, and we rely on
1095 * mdadm.conf to specify the arrays that should be assembled.
1096 * Auto-assembly may still pick up "foreign" arrays.
1097 */
1098
1099 return -1;
1100 }
1101
1102 static void uuid_from_super_imsm(struct supertype *st, int uuid[4])
1103 {
1104 /* The uuid returned here is used for:
1105 * uuid to put into bitmap file (Create, Grow)
1106 * uuid for backup header when saving critical section (Grow)
1107 * comparing uuids when re-adding a device into an array
1108 * In these cases the uuid required is that of the data-array,
1109 * not the device-set.
1110 * uuid to recognise same set when adding a missing device back
1111 * to an array. This is a uuid for the device-set.
1112 *
1113 * For each of these we can make do with a truncated
1114 * or hashed uuid rather than the original, as long as
1115 * everyone agrees.
1116 * In each case the uuid required is that of the data-array,
1117 * not the device-set.
1118 */
1119 /* imsm does not track uuid's so we synthesis one using sha1 on
1120 * - The signature (Which is constant for all imsm array, but no matter)
1121 * - the family_num of the container
1122 * - the index number of the volume
1123 * - the 'serial' number of the volume.
1124 * Hopefully these are all constant.
1125 */
1126 struct intel_super *super = st->sb;
1127
1128 char buf[20];
1129 struct sha1_ctx ctx;
1130 struct imsm_dev *dev = NULL;
1131
1132 sha1_init_ctx(&ctx);
1133 sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx);
1134 sha1_process_bytes(&super->anchor->family_num, sizeof(__u32), &ctx);
1135 if (super->current_vol >= 0)
1136 dev = get_imsm_dev(super, super->current_vol);
1137 if (dev) {
1138 __u32 vol = super->current_vol;
1139 sha1_process_bytes(&vol, sizeof(vol), &ctx);
1140 sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx);
1141 }
1142 sha1_finish_ctx(&ctx, buf);
1143 memcpy(uuid, buf, 4*4);
1144 }
1145
1146 #if 0
1147 static void
1148 get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p)
1149 {
1150 __u8 *v = get_imsm_version(mpb);
1151 __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH;
1152 char major[] = { 0, 0, 0 };
1153 char minor[] = { 0 ,0, 0 };
1154 char patch[] = { 0, 0, 0 };
1155 char *ver_parse[] = { major, minor, patch };
1156 int i, j;
1157
1158 i = j = 0;
1159 while (*v != '\0' && v < end) {
1160 if (*v != '.' && j < 2)
1161 ver_parse[i][j++] = *v;
1162 else {
1163 i++;
1164 j = 0;
1165 }
1166 v++;
1167 }
1168
1169 *m = strtol(minor, NULL, 0);
1170 *p = strtol(patch, NULL, 0);
1171 }
1172 #endif
1173
1174 static int imsm_level_to_layout(int level)
1175 {
1176 switch (level) {
1177 case 0:
1178 case 1:
1179 return 0;
1180 case 5:
1181 case 6:
1182 return ALGORITHM_LEFT_ASYMMETRIC;
1183 case 10:
1184 return 0x102;
1185 }
1186 return UnSet;
1187 }
1188
1189 static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info)
1190 {
1191 struct intel_super *super = st->sb;
1192 struct imsm_dev *dev = get_imsm_dev(super, super->current_vol);
1193 struct imsm_map *map = get_imsm_map(dev, 0);
1194 struct dl *dl;
1195
1196 for (dl = super->disks; dl; dl = dl->next)
1197 if (dl->raiddisk == info->disk.raid_disk)
1198 break;
1199 info->container_member = super->current_vol;
1200 info->array.raid_disks = map->num_members;
1201 info->array.level = get_imsm_raid_level(map);
1202 info->array.layout = imsm_level_to_layout(info->array.level);
1203 info->array.md_minor = -1;
1204 info->array.ctime = 0;
1205 info->array.utime = 0;
1206 info->array.chunk_size = __le16_to_cpu(map->blocks_per_strip) << 9;
1207 info->array.state = !dev->vol.dirty;
1208 info->custom_array_size = __le32_to_cpu(dev->size_high);
1209 info->custom_array_size <<= 32;
1210 info->custom_array_size |= __le32_to_cpu(dev->size_low);
1211
1212 info->disk.major = 0;
1213 info->disk.minor = 0;
1214 if (dl) {
1215 info->disk.major = dl->major;
1216 info->disk.minor = dl->minor;
1217 }
1218
1219 info->data_offset = __le32_to_cpu(map->pba_of_lba0);
1220 info->component_size = __le32_to_cpu(map->blocks_per_member);
1221 memset(info->uuid, 0, sizeof(info->uuid));
1222
1223 if (map->map_state == IMSM_T_STATE_UNINITIALIZED || dev->vol.dirty)
1224 info->resync_start = 0;
1225 else if (dev->vol.migr_state)
1226 /* FIXME add curr_migr_unit to resync_start conversion */
1227 info->resync_start = 0;
1228 else
1229 info->resync_start = ~0ULL;
1230
1231 strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN);
1232 info->name[MAX_RAID_SERIAL_LEN] = 0;
1233
1234 info->array.major_version = -1;
1235 info->array.minor_version = -2;
1236 sprintf(info->text_version, "/%s/%d",
1237 devnum2devname(st->container_dev),
1238 info->container_member);
1239 info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */
1240 uuid_from_super_imsm(st, info->uuid);
1241 }
1242
1243 /* check the config file to see if we can return a real uuid for this spare */
1244 static void fixup_container_spare_uuid(struct mdinfo *inf)
1245 {
1246 struct mddev_ident_s *array_list;
1247
1248 if (inf->array.level != LEVEL_CONTAINER ||
1249 memcmp(inf->uuid, uuid_match_any, sizeof(int[4])) != 0)
1250 return;
1251
1252 array_list = conf_get_ident(NULL);
1253
1254 for (; array_list; array_list = array_list->next) {
1255 if (array_list->uuid_set) {
1256 struct supertype *_sst; /* spare supertype */
1257 struct supertype *_cst; /* container supertype */
1258
1259 _cst = array_list->st;
1260 _sst = _cst->ss->match_metadata_desc(inf->text_version);
1261 if (_sst) {
1262 memcpy(inf->uuid, array_list->uuid, sizeof(int[4]));
1263 free(_sst);
1264 break;
1265 }
1266 }
1267 }
1268 }
1269
1270 static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info)
1271 {
1272 struct intel_super *super = st->sb;
1273 struct imsm_disk *disk;
1274 __u32 s;
1275
1276 if (super->current_vol >= 0) {
1277 getinfo_super_imsm_volume(st, info);
1278 return;
1279 }
1280
1281 /* Set raid_disks to zero so that Assemble will always pull in valid
1282 * spares
1283 */
1284 info->array.raid_disks = 0;
1285 info->array.level = LEVEL_CONTAINER;
1286 info->array.layout = 0;
1287 info->array.md_minor = -1;
1288 info->array.ctime = 0; /* N/A for imsm */
1289 info->array.utime = 0;
1290 info->array.chunk_size = 0;
1291
1292 info->disk.major = 0;
1293 info->disk.minor = 0;
1294 info->disk.raid_disk = -1;
1295 info->reshape_active = 0;
1296 info->array.major_version = -1;
1297 info->array.minor_version = -2;
1298 strcpy(info->text_version, "imsm");
1299 info->safe_mode_delay = 0;
1300 info->disk.number = -1;
1301 info->disk.state = 0;
1302 info->name[0] = 0;
1303
1304 if (super->disks) {
1305 __u32 reserved = imsm_reserved_sectors(super, super->disks);
1306
1307 disk = &super->disks->disk;
1308 info->data_offset = __le32_to_cpu(disk->total_blocks) - reserved;
1309 info->component_size = reserved;
1310 s = disk->status;
1311 info->disk.state = s & CONFIGURED_DISK ? (1 << MD_DISK_ACTIVE) : 0;
1312 /* we don't change info->disk.raid_disk here because
1313 * this state will be finalized in mdmon after we have
1314 * found the 'most fresh' version of the metadata
1315 */
1316 info->disk.state |= s & FAILED_DISK ? (1 << MD_DISK_FAULTY) : 0;
1317 info->disk.state |= s & SPARE_DISK ? 0 : (1 << MD_DISK_SYNC);
1318 }
1319
1320 /* only call uuid_from_super_imsm when this disk is part of a populated container,
1321 * ->compare_super may have updated the 'num_raid_devs' field for spares
1322 */
1323 if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs)
1324 uuid_from_super_imsm(st, info->uuid);
1325 else {
1326 memcpy(info->uuid, uuid_match_any, sizeof(int[4]));
1327 fixup_container_spare_uuid(info);
1328 }
1329 }
1330
1331 static int update_super_imsm(struct supertype *st, struct mdinfo *info,
1332 char *update, char *devname, int verbose,
1333 int uuid_set, char *homehost)
1334 {
1335 /* FIXME */
1336
1337 /* For 'assemble' and 'force' we need to return non-zero if any
1338 * change was made. For others, the return value is ignored.
1339 * Update options are:
1340 * force-one : This device looks a bit old but needs to be included,
1341 * update age info appropriately.
1342 * assemble: clear any 'faulty' flag to allow this device to
1343 * be assembled.
1344 * force-array: Array is degraded but being forced, mark it clean
1345 * if that will be needed to assemble it.
1346 *
1347 * newdev: not used ????
1348 * grow: Array has gained a new device - this is currently for
1349 * linear only
1350 * resync: mark as dirty so a resync will happen.
1351 * name: update the name - preserving the homehost
1352 *
1353 * Following are not relevant for this imsm:
1354 * sparc2.2 : update from old dodgey metadata
1355 * super-minor: change the preferred_minor number
1356 * summaries: update redundant counters.
1357 * uuid: Change the uuid of the array to match watch is given
1358 * homehost: update the recorded homehost
1359 * _reshape_progress: record new reshape_progress position.
1360 */
1361 int rv = 0;
1362 //struct intel_super *super = st->sb;
1363 //struct imsm_super *mpb = super->mpb;
1364
1365 if (strcmp(update, "grow") == 0) {
1366 }
1367 if (strcmp(update, "resync") == 0) {
1368 /* dev->vol.dirty = 1; */
1369 }
1370
1371 /* IMSM has no concept of UUID or homehost */
1372
1373 return rv;
1374 }
1375
1376 static size_t disks_to_mpb_size(int disks)
1377 {
1378 size_t size;
1379
1380 size = sizeof(struct imsm_super);
1381 size += (disks - 1) * sizeof(struct imsm_disk);
1382 size += 2 * sizeof(struct imsm_dev);
1383 /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */
1384 size += (4 - 2) * sizeof(struct imsm_map);
1385 /* 4 possible disk_ord_tbl's */
1386 size += 4 * (disks - 1) * sizeof(__u32);
1387
1388 return size;
1389 }
1390
1391 static __u64 avail_size_imsm(struct supertype *st, __u64 devsize)
1392 {
1393 if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS))
1394 return 0;
1395
1396 return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS);
1397 }
1398
1399 static void free_devlist(struct intel_super *super)
1400 {
1401 struct intel_dev *dv;
1402
1403 while (super->devlist) {
1404 dv = super->devlist->next;
1405 free(super->devlist->dev);
1406 free(super->devlist);
1407 super->devlist = dv;
1408 }
1409 }
1410
1411 static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src)
1412 {
1413 memcpy(dest, src, sizeof_imsm_dev(src, 0));
1414 }
1415
1416 static int compare_super_imsm(struct supertype *st, struct supertype *tst)
1417 {
1418 /*
1419 * return:
1420 * 0 same, or first was empty, and second was copied
1421 * 1 second had wrong number
1422 * 2 wrong uuid
1423 * 3 wrong other info
1424 */
1425 struct intel_super *first = st->sb;
1426 struct intel_super *sec = tst->sb;
1427
1428 if (!first) {
1429 st->sb = tst->sb;
1430 tst->sb = NULL;
1431 return 0;
1432 }
1433
1434 if (memcmp(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH) != 0)
1435 return 3;
1436
1437 /* if an anchor does not have num_raid_devs set then it is a free
1438 * floating spare
1439 */
1440 if (first->anchor->num_raid_devs > 0 &&
1441 sec->anchor->num_raid_devs > 0) {
1442 if (first->anchor->family_num != sec->anchor->family_num)
1443 return 3;
1444 }
1445
1446 /* if 'first' is a spare promote it to a populated mpb with sec's
1447 * family number
1448 */
1449 if (first->anchor->num_raid_devs == 0 &&
1450 sec->anchor->num_raid_devs > 0) {
1451 int i;
1452 struct intel_dev *dv;
1453 struct imsm_dev *dev;
1454
1455 /* we need to copy raid device info from sec if an allocation
1456 * fails here we don't associate the spare
1457 */
1458 for (i = 0; i < sec->anchor->num_raid_devs; i++) {
1459 dv = malloc(sizeof(*dv));
1460 if (!dv)
1461 break;
1462 dev = malloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1));
1463 if (!dev) {
1464 free(dv);
1465 break;
1466 }
1467 dv->dev = dev;
1468 dv->index = i;
1469 dv->next = first->devlist;
1470 first->devlist = dv;
1471 }
1472 if (i <= sec->anchor->num_raid_devs) {
1473 /* allocation failure */
1474 free_devlist(first);
1475 fprintf(stderr, "imsm: failed to associate spare\n");
1476 return 3;
1477 }
1478 for (i = 0; i < sec->anchor->num_raid_devs; i++)
1479 imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i));
1480
1481 first->anchor->num_raid_devs = sec->anchor->num_raid_devs;
1482 first->anchor->family_num = sec->anchor->family_num;
1483 }
1484
1485 return 0;
1486 }
1487
1488 static void fd2devname(int fd, char *name)
1489 {
1490 struct stat st;
1491 char path[256];
1492 char dname[100];
1493 char *nm;
1494 int rv;
1495
1496 name[0] = '\0';
1497 if (fstat(fd, &st) != 0)
1498 return;
1499 sprintf(path, "/sys/dev/block/%d:%d",
1500 major(st.st_rdev), minor(st.st_rdev));
1501
1502 rv = readlink(path, dname, sizeof(dname));
1503 if (rv <= 0)
1504 return;
1505
1506 dname[rv] = '\0';
1507 nm = strrchr(dname, '/');
1508 nm++;
1509 snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm);
1510 }
1511
1512
1513 extern int scsi_get_serial(int fd, void *buf, size_t buf_len);
1514
1515 static int imsm_read_serial(int fd, char *devname,
1516 __u8 serial[MAX_RAID_SERIAL_LEN])
1517 {
1518 unsigned char scsi_serial[255];
1519 int rv;
1520 int rsp_len;
1521 int len;
1522 char *dest;
1523 char *src;
1524 char *rsp_buf;
1525 int i;
1526
1527 memset(scsi_serial, 0, sizeof(scsi_serial));
1528
1529 rv = scsi_get_serial(fd, scsi_serial, sizeof(scsi_serial));
1530
1531 if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) {
1532 memset(serial, 0, MAX_RAID_SERIAL_LEN);
1533 fd2devname(fd, (char *) serial);
1534 return 0;
1535 }
1536
1537 if (rv != 0) {
1538 if (devname)
1539 fprintf(stderr,
1540 Name ": Failed to retrieve serial for %s\n",
1541 devname);
1542 return rv;
1543 }
1544
1545 rsp_len = scsi_serial[3];
1546 if (!rsp_len) {
1547 if (devname)
1548 fprintf(stderr,
1549 Name ": Failed to retrieve serial for %s\n",
1550 devname);
1551 return 2;
1552 }
1553 rsp_buf = (char *) &scsi_serial[4];
1554
1555 /* trim all whitespace and non-printable characters and convert
1556 * ':' to ';'
1557 */
1558 for (i = 0, dest = rsp_buf; i < rsp_len; i++) {
1559 src = &rsp_buf[i];
1560 if (*src > 0x20) {
1561 /* ':' is reserved for use in placeholder serial
1562 * numbers for missing disks
1563 */
1564 if (*src == ':')
1565 *dest++ = ';';
1566 else
1567 *dest++ = *src;
1568 }
1569 }
1570 len = dest - rsp_buf;
1571 dest = rsp_buf;
1572
1573 /* truncate leading characters */
1574 if (len > MAX_RAID_SERIAL_LEN) {
1575 dest += len - MAX_RAID_SERIAL_LEN;
1576 len = MAX_RAID_SERIAL_LEN;
1577 }
1578
1579 memset(serial, 0, MAX_RAID_SERIAL_LEN);
1580 memcpy(serial, dest, len);
1581
1582 return 0;
1583 }
1584
1585 static int serialcmp(__u8 *s1, __u8 *s2)
1586 {
1587 return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN);
1588 }
1589
1590 static void serialcpy(__u8 *dest, __u8 *src)
1591 {
1592 strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN);
1593 }
1594
1595 static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super)
1596 {
1597 struct dl *dl;
1598
1599 for (dl = super->disks; dl; dl = dl->next)
1600 if (serialcmp(dl->serial, serial) == 0)
1601 break;
1602
1603 return dl;
1604 }
1605
1606 static int
1607 load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd)
1608 {
1609 struct dl *dl;
1610 struct stat stb;
1611 int rv;
1612 int i;
1613 int alloc = 1;
1614 __u8 serial[MAX_RAID_SERIAL_LEN];
1615
1616 rv = imsm_read_serial(fd, devname, serial);
1617
1618 if (rv != 0)
1619 return 2;
1620
1621 /* check if this is a disk we have seen before. it may be a spare in
1622 * super->disks while the current anchor believes it is a raid member,
1623 * check if we need to update dl->index
1624 */
1625 dl = serial_to_dl(serial, super);
1626 if (!dl)
1627 dl = malloc(sizeof(*dl));
1628 else
1629 alloc = 0;
1630
1631 if (!dl) {
1632 if (devname)
1633 fprintf(stderr,
1634 Name ": failed to allocate disk buffer for %s\n",
1635 devname);
1636 return 2;
1637 }
1638
1639 if (alloc) {
1640 fstat(fd, &stb);
1641 dl->major = major(stb.st_rdev);
1642 dl->minor = minor(stb.st_rdev);
1643 dl->next = super->disks;
1644 dl->fd = keep_fd ? fd : -1;
1645 dl->devname = devname ? strdup(devname) : NULL;
1646 serialcpy(dl->serial, serial);
1647 dl->index = -2;
1648 dl->e = NULL;
1649 } else if (keep_fd) {
1650 close(dl->fd);
1651 dl->fd = fd;
1652 }
1653
1654 /* look up this disk's index in the current anchor */
1655 for (i = 0; i < super->anchor->num_disks; i++) {
1656 struct imsm_disk *disk_iter;
1657
1658 disk_iter = __get_imsm_disk(super->anchor, i);
1659
1660 if (serialcmp(disk_iter->serial, dl->serial) == 0) {
1661 dl->disk = *disk_iter;
1662 /* only set index on disks that are a member of a
1663 * populated contianer, i.e. one with raid_devs
1664 */
1665 if (dl->disk.status & FAILED_DISK)
1666 dl->index = -2;
1667 else if (dl->disk.status & SPARE_DISK)
1668 dl->index = -1;
1669 else
1670 dl->index = i;
1671
1672 break;
1673 }
1674 }
1675
1676 /* no match, maybe a stale failed drive */
1677 if (i == super->anchor->num_disks && dl->index >= 0) {
1678 dl->disk = *__get_imsm_disk(super->anchor, dl->index);
1679 if (dl->disk.status & FAILED_DISK)
1680 dl->index = -2;
1681 }
1682
1683 if (alloc)
1684 super->disks = dl;
1685
1686 return 0;
1687 }
1688
1689 #ifndef MDASSEMBLE
1690 /* When migrating map0 contains the 'destination' state while map1
1691 * contains the current state. When not migrating map0 contains the
1692 * current state. This routine assumes that map[0].map_state is set to
1693 * the current array state before being called.
1694 *
1695 * Migration is indicated by one of the following states
1696 * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed)
1697 * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal
1698 * map1state=unitialized)
1699 * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal
1700 * map1state=normal)
1701 * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal
1702 * map1state=degraded)
1703 */
1704 static void migrate(struct imsm_dev *dev, __u8 to_state, int migr_type)
1705 {
1706 struct imsm_map *dest;
1707 struct imsm_map *src = get_imsm_map(dev, 0);
1708
1709 dev->vol.migr_state = 1;
1710 set_migr_type(dev, migr_type);
1711 dev->vol.curr_migr_unit = 0;
1712 dest = get_imsm_map(dev, 1);
1713
1714 /* duplicate and then set the target end state in map[0] */
1715 memcpy(dest, src, sizeof_imsm_map(src));
1716 if (migr_type == MIGR_REBUILD) {
1717 __u32 ord;
1718 int i;
1719
1720 for (i = 0; i < src->num_members; i++) {
1721 ord = __le32_to_cpu(src->disk_ord_tbl[i]);
1722 set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord));
1723 }
1724 }
1725
1726 src->map_state = to_state;
1727 }
1728
1729 static void end_migration(struct imsm_dev *dev, __u8 map_state)
1730 {
1731 struct imsm_map *map = get_imsm_map(dev, 0);
1732 struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state);
1733 int i;
1734
1735 /* merge any IMSM_ORD_REBUILD bits that were not successfully
1736 * completed in the last migration.
1737 *
1738 * FIXME add support for online capacity expansion and
1739 * raid-level-migration
1740 */
1741 for (i = 0; i < prev->num_members; i++)
1742 map->disk_ord_tbl[i] |= prev->disk_ord_tbl[i];
1743
1744 dev->vol.migr_state = 0;
1745 dev->vol.curr_migr_unit = 0;
1746 map->map_state = map_state;
1747 }
1748 #endif
1749
1750 static int parse_raid_devices(struct intel_super *super)
1751 {
1752 int i;
1753 struct imsm_dev *dev_new;
1754 size_t len, len_migr;
1755 size_t space_needed = 0;
1756 struct imsm_super *mpb = super->anchor;
1757
1758 for (i = 0; i < super->anchor->num_raid_devs; i++) {
1759 struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i);
1760 struct intel_dev *dv;
1761
1762 len = sizeof_imsm_dev(dev_iter, 0);
1763 len_migr = sizeof_imsm_dev(dev_iter, 1);
1764 if (len_migr > len)
1765 space_needed += len_migr - len;
1766
1767 dv = malloc(sizeof(*dv));
1768 if (!dv)
1769 return 1;
1770 dev_new = malloc(len_migr);
1771 if (!dev_new) {
1772 free(dv);
1773 return 1;
1774 }
1775 imsm_copy_dev(dev_new, dev_iter);
1776 dv->dev = dev_new;
1777 dv->index = i;
1778 dv->next = super->devlist;
1779 super->devlist = dv;
1780 }
1781
1782 /* ensure that super->buf is large enough when all raid devices
1783 * are migrating
1784 */
1785 if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) {
1786 void *buf;
1787
1788 len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed, 512);
1789 if (posix_memalign(&buf, 512, len) != 0)
1790 return 1;
1791
1792 memcpy(buf, super->buf, super->len);
1793 memset(buf + super->len, 0, len - super->len);
1794 free(super->buf);
1795 super->buf = buf;
1796 super->len = len;
1797 }
1798
1799 return 0;
1800 }
1801
1802 /* retrieve a pointer to the bbm log which starts after all raid devices */
1803 struct bbm_log *__get_imsm_bbm_log(struct imsm_super *mpb)
1804 {
1805 void *ptr = NULL;
1806
1807 if (__le32_to_cpu(mpb->bbm_log_size)) {
1808 ptr = mpb;
1809 ptr += mpb->mpb_size - __le32_to_cpu(mpb->bbm_log_size);
1810 }
1811
1812 return ptr;
1813 }
1814
1815 static void __free_imsm(struct intel_super *super, int free_disks);
1816
1817 /* load_imsm_mpb - read matrix metadata
1818 * allocates super->mpb to be freed by free_super
1819 */
1820 static int load_imsm_mpb(int fd, struct intel_super *super, char *devname)
1821 {
1822 unsigned long long dsize;
1823 unsigned long long sectors;
1824 struct stat;
1825 struct imsm_super *anchor;
1826 __u32 check_sum;
1827 int rc;
1828
1829 get_dev_size(fd, NULL, &dsize);
1830
1831 if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) {
1832 if (devname)
1833 fprintf(stderr,
1834 Name ": Cannot seek to anchor block on %s: %s\n",
1835 devname, strerror(errno));
1836 return 1;
1837 }
1838
1839 if (posix_memalign((void**)&anchor, 512, 512) != 0) {
1840 if (devname)
1841 fprintf(stderr,
1842 Name ": Failed to allocate imsm anchor buffer"
1843 " on %s\n", devname);
1844 return 1;
1845 }
1846 if (read(fd, anchor, 512) != 512) {
1847 if (devname)
1848 fprintf(stderr,
1849 Name ": Cannot read anchor block on %s: %s\n",
1850 devname, strerror(errno));
1851 free(anchor);
1852 return 1;
1853 }
1854
1855 if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) {
1856 if (devname)
1857 fprintf(stderr,
1858 Name ": no IMSM anchor on %s\n", devname);
1859 free(anchor);
1860 return 2;
1861 }
1862
1863 __free_imsm(super, 0);
1864 super->len = ROUND_UP(anchor->mpb_size, 512);
1865 if (posix_memalign(&super->buf, 512, super->len) != 0) {
1866 if (devname)
1867 fprintf(stderr,
1868 Name ": unable to allocate %zu byte mpb buffer\n",
1869 super->len);
1870 free(anchor);
1871 return 2;
1872 }
1873 memcpy(super->buf, anchor, 512);
1874
1875 sectors = mpb_sectors(anchor) - 1;
1876 free(anchor);
1877 if (!sectors) {
1878 check_sum = __gen_imsm_checksum(super->anchor);
1879 if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
1880 if (devname)
1881 fprintf(stderr,
1882 Name ": IMSM checksum %x != %x on %s\n",
1883 check_sum,
1884 __le32_to_cpu(super->anchor->check_sum),
1885 devname);
1886 return 2;
1887 }
1888
1889 rc = load_imsm_disk(fd, super, devname, 0);
1890 if (rc == 0)
1891 rc = parse_raid_devices(super);
1892 return rc;
1893 }
1894
1895 /* read the extended mpb */
1896 if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) {
1897 if (devname)
1898 fprintf(stderr,
1899 Name ": Cannot seek to extended mpb on %s: %s\n",
1900 devname, strerror(errno));
1901 return 1;
1902 }
1903
1904 if (read(fd, super->buf + 512, super->len - 512) != super->len - 512) {
1905 if (devname)
1906 fprintf(stderr,
1907 Name ": Cannot read extended mpb on %s: %s\n",
1908 devname, strerror(errno));
1909 return 2;
1910 }
1911
1912 check_sum = __gen_imsm_checksum(super->anchor);
1913 if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
1914 if (devname)
1915 fprintf(stderr,
1916 Name ": IMSM checksum %x != %x on %s\n",
1917 check_sum, __le32_to_cpu(super->anchor->check_sum),
1918 devname);
1919 return 3;
1920 }
1921
1922 /* FIXME the BBM log is disk specific so we cannot use this global
1923 * buffer for all disks. Ok for now since we only look at the global
1924 * bbm_log_size parameter to gate assembly
1925 */
1926 super->bbm_log = __get_imsm_bbm_log(super->anchor);
1927
1928 rc = load_imsm_disk(fd, super, devname, 0);
1929 if (rc == 0)
1930 rc = parse_raid_devices(super);
1931
1932 return rc;
1933 }
1934
1935 static void __free_imsm_disk(struct dl *d)
1936 {
1937 if (d->fd >= 0)
1938 close(d->fd);
1939 if (d->devname)
1940 free(d->devname);
1941 if (d->e)
1942 free(d->e);
1943 free(d);
1944
1945 }
1946 static void free_imsm_disks(struct intel_super *super)
1947 {
1948 struct dl *d;
1949
1950 while (super->disks) {
1951 d = super->disks;
1952 super->disks = d->next;
1953 __free_imsm_disk(d);
1954 }
1955 while (super->missing) {
1956 d = super->missing;
1957 super->missing = d->next;
1958 __free_imsm_disk(d);
1959 }
1960
1961 }
1962
1963 /* free all the pieces hanging off of a super pointer */
1964 static void __free_imsm(struct intel_super *super, int free_disks)
1965 {
1966 if (super->buf) {
1967 free(super->buf);
1968 super->buf = NULL;
1969 }
1970 if (free_disks)
1971 free_imsm_disks(super);
1972 free_devlist(super);
1973 if (super->hba) {
1974 free((void *) super->hba);
1975 super->hba = NULL;
1976 }
1977 }
1978
1979 static void free_imsm(struct intel_super *super)
1980 {
1981 __free_imsm(super, 1);
1982 free(super);
1983 }
1984
1985 static void free_super_imsm(struct supertype *st)
1986 {
1987 struct intel_super *super = st->sb;
1988
1989 if (!super)
1990 return;
1991
1992 free_imsm(super);
1993 st->sb = NULL;
1994 }
1995
1996 static struct intel_super *alloc_super(int creating_imsm)
1997 {
1998 struct intel_super *super = malloc(sizeof(*super));
1999
2000 if (super) {
2001 memset(super, 0, sizeof(*super));
2002 super->creating_imsm = creating_imsm;
2003 super->current_vol = -1;
2004 super->create_offset = ~((__u32 ) 0);
2005 if (!check_env("IMSM_NO_PLATFORM"))
2006 super->orom = find_imsm_orom();
2007 if (super->orom && !check_env("IMSM_TEST_OROM")) {
2008 struct sys_dev *list, *ent;
2009
2010 /* find the first intel ahci controller */
2011 list = find_driver_devices("pci", "ahci");
2012 for (ent = list; ent; ent = ent->next)
2013 if (devpath_to_vendor(ent->path) == 0x8086)
2014 break;
2015 if (ent) {
2016 super->hba = ent->path;
2017 ent->path = NULL;
2018 }
2019 free_sys_dev(&list);
2020 }
2021 }
2022
2023 return super;
2024 }
2025
2026 #ifndef MDASSEMBLE
2027 /* find_missing - helper routine for load_super_imsm_all that identifies
2028 * disks that have disappeared from the system. This routine relies on
2029 * the mpb being uptodate, which it is at load time.
2030 */
2031 static int find_missing(struct intel_super *super)
2032 {
2033 int i;
2034 struct imsm_super *mpb = super->anchor;
2035 struct dl *dl;
2036 struct imsm_disk *disk;
2037
2038 for (i = 0; i < mpb->num_disks; i++) {
2039 disk = __get_imsm_disk(mpb, i);
2040 dl = serial_to_dl(disk->serial, super);
2041 if (dl)
2042 continue;
2043
2044 dl = malloc(sizeof(*dl));
2045 if (!dl)
2046 return 1;
2047 dl->major = 0;
2048 dl->minor = 0;
2049 dl->fd = -1;
2050 dl->devname = strdup("missing");
2051 dl->index = i;
2052 serialcpy(dl->serial, disk->serial);
2053 dl->disk = *disk;
2054 dl->e = NULL;
2055 dl->next = super->missing;
2056 super->missing = dl;
2057 }
2058
2059 return 0;
2060 }
2061
2062 static int load_super_imsm_all(struct supertype *st, int fd, void **sbp,
2063 char *devname, int keep_fd)
2064 {
2065 struct mdinfo *sra;
2066 struct intel_super *super;
2067 struct mdinfo *sd, *best = NULL;
2068 __u32 bestgen = 0;
2069 __u32 gen;
2070 char nm[20];
2071 int dfd;
2072 int rv;
2073 int devnum = fd2devnum(fd);
2074 int retry;
2075 enum sysfs_read_flags flags;
2076
2077 flags = GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE;
2078 if (mdmon_running(devnum))
2079 flags |= SKIP_GONE_DEVS;
2080
2081 /* check if 'fd' an opened container */
2082 sra = sysfs_read(fd, 0, flags);
2083 if (!sra)
2084 return 1;
2085
2086 if (sra->array.major_version != -1 ||
2087 sra->array.minor_version != -2 ||
2088 strcmp(sra->text_version, "imsm") != 0)
2089 return 1;
2090
2091 super = alloc_super(0);
2092 if (!super)
2093 return 1;
2094
2095 /* find the most up to date disk in this array, skipping spares */
2096 for (sd = sra->devs; sd; sd = sd->next) {
2097 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
2098 dfd = dev_open(nm, keep_fd ? O_RDWR : O_RDONLY);
2099 if (dfd < 0) {
2100 free_imsm(super);
2101 return 2;
2102 }
2103 rv = load_imsm_mpb(dfd, super, NULL);
2104
2105 /* retry the load if we might have raced against mdmon */
2106 if (rv == 3 && mdmon_running(devnum))
2107 for (retry = 0; retry < 3; retry++) {
2108 usleep(3000);
2109 rv = load_imsm_mpb(dfd, super, NULL);
2110 if (rv != 3)
2111 break;
2112 }
2113 if (!keep_fd)
2114 close(dfd);
2115 if (rv == 0) {
2116 if (super->anchor->num_raid_devs == 0)
2117 gen = 0;
2118 else
2119 gen = __le32_to_cpu(super->anchor->generation_num);
2120 if (!best || gen > bestgen) {
2121 bestgen = gen;
2122 best = sd;
2123 }
2124 } else {
2125 free_imsm(super);
2126 return rv;
2127 }
2128 }
2129
2130 if (!best) {
2131 free_imsm(super);
2132 return 1;
2133 }
2134
2135 /* load the most up to date anchor */
2136 sprintf(nm, "%d:%d", best->disk.major, best->disk.minor);
2137 dfd = dev_open(nm, O_RDONLY);
2138 if (dfd < 0) {
2139 free_imsm(super);
2140 return 1;
2141 }
2142 rv = load_imsm_mpb(dfd, super, NULL);
2143 close(dfd);
2144 if (rv != 0) {
2145 free_imsm(super);
2146 return 2;
2147 }
2148
2149 /* re-parse the disk list with the current anchor */
2150 for (sd = sra->devs ; sd ; sd = sd->next) {
2151 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
2152 dfd = dev_open(nm, keep_fd? O_RDWR : O_RDONLY);
2153 if (dfd < 0) {
2154 free_imsm(super);
2155 return 2;
2156 }
2157 load_imsm_disk(dfd, super, NULL, keep_fd);
2158 if (!keep_fd)
2159 close(dfd);
2160 }
2161
2162
2163 if (find_missing(super) != 0) {
2164 free_imsm(super);
2165 return 2;
2166 }
2167
2168 if (st->subarray[0]) {
2169 if (atoi(st->subarray) <= super->anchor->num_raid_devs)
2170 super->current_vol = atoi(st->subarray);
2171 else
2172 return 1;
2173 }
2174
2175 *sbp = super;
2176 st->container_dev = devnum;
2177 if (st->ss == NULL) {
2178 st->ss = &super_imsm;
2179 st->minor_version = 0;
2180 st->max_devs = IMSM_MAX_DEVICES;
2181 }
2182 st->loaded_container = 1;
2183
2184 return 0;
2185 }
2186 #endif
2187
2188 static int load_super_imsm(struct supertype *st, int fd, char *devname)
2189 {
2190 struct intel_super *super;
2191 int rv;
2192
2193 #ifndef MDASSEMBLE
2194 if (load_super_imsm_all(st, fd, &st->sb, devname, 1) == 0)
2195 return 0;
2196 #endif
2197 if (st->subarray[0])
2198 return 1; /* FIXME */
2199
2200 super = alloc_super(0);
2201 if (!super) {
2202 fprintf(stderr,
2203 Name ": malloc of %zu failed.\n",
2204 sizeof(*super));
2205 return 1;
2206 }
2207
2208 rv = load_imsm_mpb(fd, super, devname);
2209
2210 if (rv) {
2211 if (devname)
2212 fprintf(stderr,
2213 Name ": Failed to load all information "
2214 "sections on %s\n", devname);
2215 free_imsm(super);
2216 return rv;
2217 }
2218
2219 st->sb = super;
2220 if (st->ss == NULL) {
2221 st->ss = &super_imsm;
2222 st->minor_version = 0;
2223 st->max_devs = IMSM_MAX_DEVICES;
2224 }
2225 st->loaded_container = 0;
2226
2227 return 0;
2228 }
2229
2230 static __u16 info_to_blocks_per_strip(mdu_array_info_t *info)
2231 {
2232 if (info->level == 1)
2233 return 128;
2234 return info->chunk_size >> 9;
2235 }
2236
2237 static __u32 info_to_num_data_stripes(mdu_array_info_t *info, int num_domains)
2238 {
2239 __u32 num_stripes;
2240
2241 num_stripes = (info->size * 2) / info_to_blocks_per_strip(info);
2242 num_stripes /= num_domains;
2243
2244 return num_stripes;
2245 }
2246
2247 static __u32 info_to_blocks_per_member(mdu_array_info_t *info)
2248 {
2249 if (info->level == 1)
2250 return info->size * 2;
2251 else
2252 return (info->size * 2) & ~(info_to_blocks_per_strip(info) - 1);
2253 }
2254
2255 static void imsm_update_version_info(struct intel_super *super)
2256 {
2257 /* update the version and attributes */
2258 struct imsm_super *mpb = super->anchor;
2259 char *version;
2260 struct imsm_dev *dev;
2261 struct imsm_map *map;
2262 int i;
2263
2264 for (i = 0; i < mpb->num_raid_devs; i++) {
2265 dev = get_imsm_dev(super, i);
2266 map = get_imsm_map(dev, 0);
2267 if (__le32_to_cpu(dev->size_high) > 0)
2268 mpb->attributes |= MPB_ATTRIB_2TB;
2269
2270 /* FIXME detect when an array spans a port multiplier */
2271 #if 0
2272 mpb->attributes |= MPB_ATTRIB_PM;
2273 #endif
2274
2275 if (mpb->num_raid_devs > 1 ||
2276 mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) {
2277 version = MPB_VERSION_ATTRIBS;
2278 switch (get_imsm_raid_level(map)) {
2279 case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break;
2280 case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break;
2281 case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break;
2282 case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break;
2283 }
2284 } else {
2285 if (map->num_members >= 5)
2286 version = MPB_VERSION_5OR6_DISK_ARRAY;
2287 else if (dev->status == DEV_CLONE_N_GO)
2288 version = MPB_VERSION_CNG;
2289 else if (get_imsm_raid_level(map) == 5)
2290 version = MPB_VERSION_RAID5;
2291 else if (map->num_members >= 3)
2292 version = MPB_VERSION_3OR4_DISK_ARRAY;
2293 else if (get_imsm_raid_level(map) == 1)
2294 version = MPB_VERSION_RAID1;
2295 else
2296 version = MPB_VERSION_RAID0;
2297 }
2298 strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version);
2299 }
2300 }
2301
2302 static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info,
2303 unsigned long long size, char *name,
2304 char *homehost, int *uuid)
2305 {
2306 /* We are creating a volume inside a pre-existing container.
2307 * so st->sb is already set.
2308 */
2309 struct intel_super *super = st->sb;
2310 struct imsm_super *mpb = super->anchor;
2311 struct intel_dev *dv;
2312 struct imsm_dev *dev;
2313 struct imsm_vol *vol;
2314 struct imsm_map *map;
2315 int idx = mpb->num_raid_devs;
2316 int i;
2317 unsigned long long array_blocks;
2318 size_t size_old, size_new;
2319 __u32 num_data_stripes;
2320
2321 if (super->orom && mpb->num_raid_devs >= super->orom->vpa) {
2322 fprintf(stderr, Name": This imsm-container already has the "
2323 "maximum of %d volumes\n", super->orom->vpa);
2324 return 0;
2325 }
2326
2327 /* ensure the mpb is large enough for the new data */
2328 size_old = __le32_to_cpu(mpb->mpb_size);
2329 size_new = disks_to_mpb_size(info->nr_disks);
2330 if (size_new > size_old) {
2331 void *mpb_new;
2332 size_t size_round = ROUND_UP(size_new, 512);
2333
2334 if (posix_memalign(&mpb_new, 512, size_round) != 0) {
2335 fprintf(stderr, Name": could not allocate new mpb\n");
2336 return 0;
2337 }
2338 memcpy(mpb_new, mpb, size_old);
2339 free(mpb);
2340 mpb = mpb_new;
2341 super->anchor = mpb_new;
2342 mpb->mpb_size = __cpu_to_le32(size_new);
2343 memset(mpb_new + size_old, 0, size_round - size_old);
2344 }
2345 super->current_vol = idx;
2346 /* when creating the first raid device in this container set num_disks
2347 * to zero, i.e. delete this spare and add raid member devices in
2348 * add_to_super_imsm_volume()
2349 */
2350 if (super->current_vol == 0)
2351 mpb->num_disks = 0;
2352
2353 for (i = 0; i < super->current_vol; i++) {
2354 dev = get_imsm_dev(super, i);
2355 if (strncmp((char *) dev->volume, name,
2356 MAX_RAID_SERIAL_LEN) == 0) {
2357 fprintf(stderr, Name": '%s' is already defined for this container\n",
2358 name);
2359 return 0;
2360 }
2361 }
2362
2363 sprintf(st->subarray, "%d", idx);
2364 dv = malloc(sizeof(*dv));
2365 if (!dv) {
2366 fprintf(stderr, Name ": failed to allocate device list entry\n");
2367 return 0;
2368 }
2369 dev = malloc(sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1));
2370 if (!dev) {
2371 free(dv);
2372 fprintf(stderr, Name": could not allocate raid device\n");
2373 return 0;
2374 }
2375 strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN);
2376 if (info->level == 1)
2377 array_blocks = info_to_blocks_per_member(info);
2378 else
2379 array_blocks = calc_array_size(info->level, info->raid_disks,
2380 info->layout, info->chunk_size,
2381 info->size*2);
2382 /* round array size down to closest MB */
2383 array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT;
2384
2385 dev->size_low = __cpu_to_le32((__u32) array_blocks);
2386 dev->size_high = __cpu_to_le32((__u32) (array_blocks >> 32));
2387 dev->status = __cpu_to_le32(0);
2388 dev->reserved_blocks = __cpu_to_le32(0);
2389 vol = &dev->vol;
2390 vol->migr_state = 0;
2391 set_migr_type(dev, MIGR_INIT);
2392 vol->dirty = 0;
2393 vol->curr_migr_unit = 0;
2394 map = get_imsm_map(dev, 0);
2395 map->pba_of_lba0 = __cpu_to_le32(super->create_offset);
2396 map->blocks_per_member = __cpu_to_le32(info_to_blocks_per_member(info));
2397 map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info));
2398 map->failed_disk_num = ~0;
2399 map->map_state = info->level ? IMSM_T_STATE_UNINITIALIZED :
2400 IMSM_T_STATE_NORMAL;
2401 map->ddf = 1;
2402
2403 if (info->level == 1 && info->raid_disks > 2) {
2404 fprintf(stderr, Name": imsm does not support more than 2 disks"
2405 "in a raid1 volume\n");
2406 return 0;
2407 }
2408
2409 map->raid_level = info->level;
2410 if (info->level == 10) {
2411 map->raid_level = 1;
2412 map->num_domains = info->raid_disks / 2;
2413 } else if (info->level == 1)
2414 map->num_domains = info->raid_disks;
2415 else
2416 map->num_domains = 1;
2417
2418 num_data_stripes = info_to_num_data_stripes(info, map->num_domains);
2419 map->num_data_stripes = __cpu_to_le32(num_data_stripes);
2420
2421 map->num_members = info->raid_disks;
2422 for (i = 0; i < map->num_members; i++) {
2423 /* initialized in add_to_super */
2424 set_imsm_ord_tbl_ent(map, i, 0);
2425 }
2426 mpb->num_raid_devs++;
2427
2428 dv->dev = dev;
2429 dv->index = super->current_vol;
2430 dv->next = super->devlist;
2431 super->devlist = dv;
2432
2433 imsm_update_version_info(super);
2434
2435 return 1;
2436 }
2437
2438 static int init_super_imsm(struct supertype *st, mdu_array_info_t *info,
2439 unsigned long long size, char *name,
2440 char *homehost, int *uuid)
2441 {
2442 /* This is primarily called by Create when creating a new array.
2443 * We will then get add_to_super called for each component, and then
2444 * write_init_super called to write it out to each device.
2445 * For IMSM, Create can create on fresh devices or on a pre-existing
2446 * array.
2447 * To create on a pre-existing array a different method will be called.
2448 * This one is just for fresh drives.
2449 */
2450 struct intel_super *super;
2451 struct imsm_super *mpb;
2452 size_t mpb_size;
2453 char *version;
2454
2455 if (!info) {
2456 st->sb = NULL;
2457 return 0;
2458 }
2459 if (st->sb)
2460 return init_super_imsm_volume(st, info, size, name, homehost,
2461 uuid);
2462
2463 super = alloc_super(1);
2464 if (!super)
2465 return 0;
2466 mpb_size = disks_to_mpb_size(info->nr_disks);
2467 if (posix_memalign(&super->buf, 512, mpb_size) != 0) {
2468 free(super);
2469 return 0;
2470 }
2471 mpb = super->buf;
2472 memset(mpb, 0, mpb_size);
2473
2474 mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
2475
2476 version = (char *) mpb->sig;
2477 strcpy(version, MPB_SIGNATURE);
2478 version += strlen(MPB_SIGNATURE);
2479 strcpy(version, MPB_VERSION_RAID0);
2480 mpb->mpb_size = mpb_size;
2481
2482 st->sb = super;
2483 return 1;
2484 }
2485
2486 #ifndef MDASSEMBLE
2487 static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk,
2488 int fd, char *devname)
2489 {
2490 struct intel_super *super = st->sb;
2491 struct imsm_super *mpb = super->anchor;
2492 struct dl *dl;
2493 struct imsm_dev *dev;
2494 struct imsm_map *map;
2495
2496 dev = get_imsm_dev(super, super->current_vol);
2497 map = get_imsm_map(dev, 0);
2498
2499 if (! (dk->state & (1<<MD_DISK_SYNC))) {
2500 fprintf(stderr, Name ": %s: Cannot add spare devices to IMSM volume\n",
2501 devname);
2502 return 1;
2503 }
2504
2505 if (fd == -1) {
2506 /* we're doing autolayout so grab the pre-marked (in
2507 * validate_geometry) raid_disk
2508 */
2509 for (dl = super->disks; dl; dl = dl->next)
2510 if (dl->raiddisk == dk->raid_disk)
2511 break;
2512 } else {
2513 for (dl = super->disks; dl ; dl = dl->next)
2514 if (dl->major == dk->major &&
2515 dl->minor == dk->minor)
2516 break;
2517 }
2518
2519 if (!dl) {
2520 fprintf(stderr, Name ": %s is not a member of the same container\n", devname);
2521 return 1;
2522 }
2523
2524 /* add a pristine spare to the metadata */
2525 if (dl->index < 0) {
2526 dl->index = super->anchor->num_disks;
2527 super->anchor->num_disks++;
2528 }
2529 set_imsm_ord_tbl_ent(map, dk->number, dl->index);
2530 dl->disk.status = CONFIGURED_DISK | USABLE_DISK;
2531
2532 /* if we are creating the first raid device update the family number */
2533 if (super->current_vol == 0) {
2534 __u32 sum;
2535 struct imsm_dev *_dev = __get_imsm_dev(mpb, 0);
2536 struct imsm_disk *_disk = __get_imsm_disk(mpb, dl->index);
2537
2538 *_dev = *dev;
2539 *_disk = dl->disk;
2540 sum = __gen_imsm_checksum(mpb);
2541 mpb->family_num = __cpu_to_le32(sum);
2542 }
2543
2544 return 0;
2545 }
2546
2547 static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk,
2548 int fd, char *devname)
2549 {
2550 struct intel_super *super = st->sb;
2551 struct dl *dd;
2552 unsigned long long size;
2553 __u32 id;
2554 int rv;
2555 struct stat stb;
2556
2557 /* if we are on an RAID enabled platform check that the disk is
2558 * attached to the raid controller
2559 */
2560 if (super->hba && !disk_attached_to_hba(fd, super->hba)) {
2561 fprintf(stderr,
2562 Name ": %s is not attached to the raid controller: %s\n",
2563 devname ? : "disk", super->hba);
2564 return 1;
2565 }
2566
2567 if (super->current_vol >= 0)
2568 return add_to_super_imsm_volume(st, dk, fd, devname);
2569
2570 fstat(fd, &stb);
2571 dd = malloc(sizeof(*dd));
2572 if (!dd) {
2573 fprintf(stderr,
2574 Name ": malloc failed %s:%d.\n", __func__, __LINE__);
2575 return 1;
2576 }
2577 memset(dd, 0, sizeof(*dd));
2578 dd->major = major(stb.st_rdev);
2579 dd->minor = minor(stb.st_rdev);
2580 dd->index = -1;
2581 dd->devname = devname ? strdup(devname) : NULL;
2582 dd->fd = fd;
2583 dd->e = NULL;
2584 rv = imsm_read_serial(fd, devname, dd->serial);
2585 if (rv) {
2586 fprintf(stderr,
2587 Name ": failed to retrieve scsi serial, aborting\n");
2588 free(dd);
2589 abort();
2590 }
2591
2592 get_dev_size(fd, NULL, &size);
2593 size /= 512;
2594 serialcpy(dd->disk.serial, dd->serial);
2595 dd->disk.total_blocks = __cpu_to_le32(size);
2596 dd->disk.status = USABLE_DISK | SPARE_DISK;
2597 if (sysfs_disk_to_scsi_id(fd, &id) == 0)
2598 dd->disk.scsi_id = __cpu_to_le32(id);
2599 else
2600 dd->disk.scsi_id = __cpu_to_le32(0);
2601
2602 if (st->update_tail) {
2603 dd->next = super->add;
2604 super->add = dd;
2605 } else {
2606 dd->next = super->disks;
2607 super->disks = dd;
2608 }
2609
2610 return 0;
2611 }
2612
2613 static int store_imsm_mpb(int fd, struct intel_super *super);
2614
2615 /* spare records have their own family number and do not have any defined raid
2616 * devices
2617 */
2618 static int write_super_imsm_spares(struct intel_super *super, int doclose)
2619 {
2620 struct imsm_super mpb_save;
2621 struct imsm_super *mpb = super->anchor;
2622 __u32 sum;
2623 struct dl *d;
2624
2625 mpb_save = *mpb;
2626 mpb->num_raid_devs = 0;
2627 mpb->num_disks = 1;
2628 mpb->mpb_size = sizeof(struct imsm_super);
2629 mpb->generation_num = __cpu_to_le32(1UL);
2630
2631 for (d = super->disks; d; d = d->next) {
2632 if (d->index != -1)
2633 continue;
2634
2635 mpb->disk[0] = d->disk;
2636 sum = __gen_imsm_checksum(mpb);
2637 mpb->family_num = __cpu_to_le32(sum);
2638 sum = __gen_imsm_checksum(mpb);
2639 mpb->check_sum = __cpu_to_le32(sum);
2640
2641 if (store_imsm_mpb(d->fd, super)) {
2642 fprintf(stderr, "%s: failed for device %d:%d %s\n",
2643 __func__, d->major, d->minor, strerror(errno));
2644 *mpb = mpb_save;
2645 return 1;
2646 }
2647 if (doclose) {
2648 close(d->fd);
2649 d->fd = -1;
2650 }
2651 }
2652
2653 *mpb = mpb_save;
2654 return 0;
2655 }
2656
2657 static int write_super_imsm(struct intel_super *super, int doclose)
2658 {
2659 struct imsm_super *mpb = super->anchor;
2660 struct dl *d;
2661 __u32 generation;
2662 __u32 sum;
2663 int spares = 0;
2664 int i;
2665 __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk);
2666
2667 /* 'generation' is incremented everytime the metadata is written */
2668 generation = __le32_to_cpu(mpb->generation_num);
2669 generation++;
2670 mpb->generation_num = __cpu_to_le32(generation);
2671
2672 mpb_size += sizeof(struct imsm_disk) * mpb->num_disks;
2673 for (d = super->disks; d; d = d->next) {
2674 if (d->index == -1)
2675 spares++;
2676 else
2677 mpb->disk[d->index] = d->disk;
2678 }
2679 for (d = super->missing; d; d = d->next)
2680 mpb->disk[d->index] = d->disk;
2681
2682 for (i = 0; i < mpb->num_raid_devs; i++) {
2683 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
2684
2685 imsm_copy_dev(dev, get_imsm_dev(super, i));
2686 mpb_size += sizeof_imsm_dev(dev, 0);
2687 }
2688 mpb_size += __le32_to_cpu(mpb->bbm_log_size);
2689 mpb->mpb_size = __cpu_to_le32(mpb_size);
2690
2691 /* recalculate checksum */
2692 sum = __gen_imsm_checksum(mpb);
2693 mpb->check_sum = __cpu_to_le32(sum);
2694
2695 /* write the mpb for disks that compose raid devices */
2696 for (d = super->disks; d ; d = d->next) {
2697 if (d->index < 0)
2698 continue;
2699 if (store_imsm_mpb(d->fd, super))
2700 fprintf(stderr, "%s: failed for device %d:%d %s\n",
2701 __func__, d->major, d->minor, strerror(errno));
2702 if (doclose) {
2703 close(d->fd);
2704 d->fd = -1;
2705 }
2706 }
2707
2708 if (spares)
2709 return write_super_imsm_spares(super, doclose);
2710
2711 return 0;
2712 }
2713
2714
2715 static int create_array(struct supertype *st)
2716 {
2717 size_t len;
2718 struct imsm_update_create_array *u;
2719 struct intel_super *super = st->sb;
2720 struct imsm_dev *dev = get_imsm_dev(super, super->current_vol);
2721 struct imsm_map *map = get_imsm_map(dev, 0);
2722 struct disk_info *inf;
2723 struct imsm_disk *disk;
2724 int i;
2725 int idx;
2726
2727 len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) +
2728 sizeof(*inf) * map->num_members;
2729 u = malloc(len);
2730 if (!u) {
2731 fprintf(stderr, "%s: failed to allocate update buffer\n",
2732 __func__);
2733 return 1;
2734 }
2735
2736 u->type = update_create_array;
2737 u->dev_idx = super->current_vol;
2738 imsm_copy_dev(&u->dev, dev);
2739 inf = get_disk_info(u);
2740 for (i = 0; i < map->num_members; i++) {
2741 idx = get_imsm_disk_idx(dev, i);
2742 disk = get_imsm_disk(super, idx);
2743 serialcpy(inf[i].serial, disk->serial);
2744 }
2745 append_metadata_update(st, u, len);
2746
2747 return 0;
2748 }
2749
2750 static int _add_disk(struct supertype *st)
2751 {
2752 struct intel_super *super = st->sb;
2753 size_t len;
2754 struct imsm_update_add_disk *u;
2755
2756 if (!super->add)
2757 return 0;
2758
2759 len = sizeof(*u);
2760 u = malloc(len);
2761 if (!u) {
2762 fprintf(stderr, "%s: failed to allocate update buffer\n",
2763 __func__);
2764 return 1;
2765 }
2766
2767 u->type = update_add_disk;
2768 append_metadata_update(st, u, len);
2769
2770 return 0;
2771 }
2772
2773 static int write_init_super_imsm(struct supertype *st)
2774 {
2775 if (st->update_tail) {
2776 /* queue the recently created array / added disk
2777 * as a metadata update */
2778 struct intel_super *super = st->sb;
2779 struct dl *d;
2780 int rv;
2781
2782 /* determine if we are creating a volume or adding a disk */
2783 if (super->current_vol < 0) {
2784 /* in the add disk case we are running in mdmon
2785 * context, so don't close fd's
2786 */
2787 return _add_disk(st);
2788 } else
2789 rv = create_array(st);
2790
2791 for (d = super->disks; d ; d = d->next) {
2792 close(d->fd);
2793 d->fd = -1;
2794 }
2795
2796 return rv;
2797 } else
2798 return write_super_imsm(st->sb, 1);
2799 }
2800 #endif
2801
2802 static int store_zero_imsm(struct supertype *st, int fd)
2803 {
2804 unsigned long long dsize;
2805 void *buf;
2806
2807 get_dev_size(fd, NULL, &dsize);
2808
2809 /* first block is stored on second to last sector of the disk */
2810 if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0)
2811 return 1;
2812
2813 if (posix_memalign(&buf, 512, 512) != 0)
2814 return 1;
2815
2816 memset(buf, 0, 512);
2817 if (write(fd, buf, 512) != 512)
2818 return 1;
2819 return 0;
2820 }
2821
2822 static int imsm_bbm_log_size(struct imsm_super *mpb)
2823 {
2824 return __le32_to_cpu(mpb->bbm_log_size);
2825 }
2826
2827 #ifndef MDASSEMBLE
2828 static int validate_geometry_imsm_container(struct supertype *st, int level,
2829 int layout, int raiddisks, int chunk,
2830 unsigned long long size, char *dev,
2831 unsigned long long *freesize,
2832 int verbose)
2833 {
2834 int fd;
2835 unsigned long long ldsize;
2836 const struct imsm_orom *orom;
2837
2838 if (level != LEVEL_CONTAINER)
2839 return 0;
2840 if (!dev)
2841 return 1;
2842
2843 if (check_env("IMSM_NO_PLATFORM"))
2844 orom = NULL;
2845 else
2846 orom = find_imsm_orom();
2847 if (orom && raiddisks > orom->tds) {
2848 if (verbose)
2849 fprintf(stderr, Name ": %d exceeds maximum number of"
2850 " platform supported disks: %d\n",
2851 raiddisks, orom->tds);
2852 return 0;
2853 }
2854
2855 fd = open(dev, O_RDONLY|O_EXCL, 0);
2856 if (fd < 0) {
2857 if (verbose)
2858 fprintf(stderr, Name ": imsm: Cannot open %s: %s\n",
2859 dev, strerror(errno));
2860 return 0;
2861 }
2862 if (!get_dev_size(fd, dev, &ldsize)) {
2863 close(fd);
2864 return 0;
2865 }
2866 close(fd);
2867
2868 *freesize = avail_size_imsm(st, ldsize >> 9);
2869
2870 return 1;
2871 }
2872
2873 static unsigned long long find_size(struct extent *e, int *idx, int num_extents)
2874 {
2875 const unsigned long long base_start = e[*idx].start;
2876 unsigned long long end = base_start + e[*idx].size;
2877 int i;
2878
2879 if (base_start == end)
2880 return 0;
2881
2882 *idx = *idx + 1;
2883 for (i = *idx; i < num_extents; i++) {
2884 /* extend overlapping extents */
2885 if (e[i].start >= base_start &&
2886 e[i].start <= end) {
2887 if (e[i].size == 0)
2888 return 0;
2889 if (e[i].start + e[i].size > end)
2890 end = e[i].start + e[i].size;
2891 } else if (e[i].start > end) {
2892 *idx = i;
2893 break;
2894 }
2895 }
2896
2897 return end - base_start;
2898 }
2899
2900 static unsigned long long merge_extents(struct intel_super *super, int sum_extents)
2901 {
2902 /* build a composite disk with all known extents and generate a new
2903 * 'maxsize' given the "all disks in an array must share a common start
2904 * offset" constraint
2905 */
2906 struct extent *e = calloc(sum_extents, sizeof(*e));
2907 struct dl *dl;
2908 int i, j;
2909 int start_extent;
2910 unsigned long long pos;
2911 unsigned long long start = 0;
2912 unsigned long long maxsize;
2913 unsigned long reserve;
2914
2915 if (!e)
2916 return ~0ULL; /* error */
2917
2918 /* coalesce and sort all extents. also, check to see if we need to
2919 * reserve space between member arrays
2920 */
2921 j = 0;
2922 for (dl = super->disks; dl; dl = dl->next) {
2923 if (!dl->e)
2924 continue;
2925 for (i = 0; i < dl->extent_cnt; i++)
2926 e[j++] = dl->e[i];
2927 }
2928 qsort(e, sum_extents, sizeof(*e), cmp_extent);
2929
2930 /* merge extents */
2931 i = 0;
2932 j = 0;
2933 while (i < sum_extents) {
2934 e[j].start = e[i].start;
2935 e[j].size = find_size(e, &i, sum_extents);
2936 j++;
2937 if (e[j-1].size == 0)
2938 break;
2939 }
2940
2941 pos = 0;
2942 maxsize = 0;
2943 start_extent = 0;
2944 i = 0;
2945 do {
2946 unsigned long long esize;
2947
2948 esize = e[i].start - pos;
2949 if (esize >= maxsize) {
2950 maxsize = esize;
2951 start = pos;
2952 start_extent = i;
2953 }
2954 pos = e[i].start + e[i].size;
2955 i++;
2956 } while (e[i-1].size);
2957 free(e);
2958
2959 if (start_extent > 0)
2960 reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */
2961 else
2962 reserve = 0;
2963
2964 if (maxsize < reserve)
2965 return ~0ULL;
2966
2967 super->create_offset = ~((__u32) 0);
2968 if (start + reserve > super->create_offset)
2969 return ~0ULL; /* start overflows create_offset */
2970 super->create_offset = start + reserve;
2971
2972 return maxsize - reserve;
2973 }
2974
2975 static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks)
2976 {
2977 if (level < 0 || level == 6 || level == 4)
2978 return 0;
2979
2980 /* if we have an orom prevent invalid raid levels */
2981 if (orom)
2982 switch (level) {
2983 case 0: return imsm_orom_has_raid0(orom);
2984 case 1:
2985 if (raiddisks > 2)
2986 return imsm_orom_has_raid1e(orom);
2987 return imsm_orom_has_raid1(orom) && raiddisks == 2;
2988 case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4;
2989 case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2;
2990 }
2991 else
2992 return 1; /* not on an Intel RAID platform so anything goes */
2993
2994 return 0;
2995 }
2996
2997 #define pr_vrb(fmt, arg...) (void) (verbose && fprintf(stderr, Name fmt, ##arg))
2998 /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd
2999 * FIX ME add ahci details
3000 */
3001 static int validate_geometry_imsm_volume(struct supertype *st, int level,
3002 int layout, int raiddisks, int chunk,
3003 unsigned long long size, char *dev,
3004 unsigned long long *freesize,
3005 int verbose)
3006 {
3007 struct stat stb;
3008 struct intel_super *super = st->sb;
3009 struct imsm_super *mpb = super->anchor;
3010 struct dl *dl;
3011 unsigned long long pos = 0;
3012 unsigned long long maxsize;
3013 struct extent *e;
3014 int i;
3015
3016 /* We must have the container info already read in. */
3017 if (!super)
3018 return 0;
3019
3020 if (!is_raid_level_supported(super->orom, level, raiddisks)) {
3021 pr_vrb(": platform does not support raid%d with %d disk%s\n",
3022 level, raiddisks, raiddisks > 1 ? "s" : "");
3023 return 0;
3024 }
3025 if (super->orom && level != 1 &&
3026 !imsm_orom_has_chunk(super->orom, chunk)) {
3027 pr_vrb(": platform does not support a chunk size of: %d\n", chunk);
3028 return 0;
3029 }
3030 if (layout != imsm_level_to_layout(level)) {
3031 if (level == 5)
3032 pr_vrb(": imsm raid 5 only supports the left-asymmetric layout\n");
3033 else if (level == 10)
3034 pr_vrb(": imsm raid 10 only supports the n2 layout\n");
3035 else
3036 pr_vrb(": imsm unknown layout %#x for this raid level %d\n",
3037 layout, level);
3038 return 0;
3039 }
3040
3041 if (!dev) {
3042 /* General test: make sure there is space for
3043 * 'raiddisks' device extents of size 'size' at a given
3044 * offset
3045 */
3046 unsigned long long minsize = size;
3047 unsigned long long start_offset = ~0ULL;
3048 int dcnt = 0;
3049 if (minsize == 0)
3050 minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
3051 for (dl = super->disks; dl ; dl = dl->next) {
3052 int found = 0;
3053
3054 pos = 0;
3055 i = 0;
3056 e = get_extents(super, dl);
3057 if (!e) continue;
3058 do {
3059 unsigned long long esize;
3060 esize = e[i].start - pos;
3061 if (esize >= minsize)
3062 found = 1;
3063 if (found && start_offset == ~0ULL) {
3064 start_offset = pos;
3065 break;
3066 } else if (found && pos != start_offset) {
3067 found = 0;
3068 break;
3069 }
3070 pos = e[i].start + e[i].size;
3071 i++;
3072 } while (e[i-1].size);
3073 if (found)
3074 dcnt++;
3075 free(e);
3076 }
3077 if (dcnt < raiddisks) {
3078 if (verbose)
3079 fprintf(stderr, Name ": imsm: Not enough "
3080 "devices with space for this array "
3081 "(%d < %d)\n",
3082 dcnt, raiddisks);
3083 return 0;
3084 }
3085 return 1;
3086 }
3087
3088 /* This device must be a member of the set */
3089 if (stat(dev, &stb) < 0)
3090 return 0;
3091 if ((S_IFMT & stb.st_mode) != S_IFBLK)
3092 return 0;
3093 for (dl = super->disks ; dl ; dl = dl->next) {
3094 if (dl->major == major(stb.st_rdev) &&
3095 dl->minor == minor(stb.st_rdev))
3096 break;
3097 }
3098 if (!dl) {
3099 if (verbose)
3100 fprintf(stderr, Name ": %s is not in the "
3101 "same imsm set\n", dev);