Create: fixup 'insert_point', dependent on 'subdevs', for auto-layout
[thirdparty/mdadm.git] / super-intel.c
CommitLineData
cdddbdbc
DW
1/*
2 * mdadm - Intel(R) Matrix Storage Manager Support
3 *
a54d5262 4 * Copyright (C) 2002-2008 Intel Corporation
cdddbdbc
DW
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
51006d85 20#define HAVE_STDINT_H 1
cdddbdbc 21#include "mdadm.h"
c2a1e7da 22#include "mdmon.h"
51006d85 23#include "sha1.h"
88c32bb1 24#include "platform-intel.h"
cdddbdbc
DW
25#include <values.h>
26#include <scsi/sg.h>
27#include <ctype.h>
d665cc31 28#include <dirent.h>
cdddbdbc
DW
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"
fe7ed8cb
DW
35#define MPB_VERSION_MANY_VOLUMES_PER_ARRAY "1.2.00"
36#define MPB_VERSION_3OR4_DISK_ARRAY "1.2.01"
cdddbdbc 37#define MPB_VERSION_RAID5 "1.2.02"
fe7ed8cb
DW
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"
cdddbdbc
DW
41#define MAX_SIGNATURE_LENGTH 32
42#define MAX_RAID_SERIAL_LEN 16
fe7ed8cb
DW
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
c2c087e6
DW
54#define MPB_SECTOR_CNT 418
55#define IMSM_RESERVED_SECTORS 4096
cdddbdbc
DW
56
57/* Disk configuration info. */
58#define IMSM_MAX_DEVICES 255
59struct imsm_disk {
60 __u8 serial[MAX_RAID_SERIAL_LEN];/* 0xD8 - 0xE7 ascii serial number */
61 __u32 total_blocks; /* 0xE8 - 0xEB total blocks */
62 __u32 scsi_id; /* 0xEC - 0xEF scsi ID */
f2f27e63
DW
63#define SPARE_DISK __cpu_to_le32(0x01) /* Spare */
64#define CONFIGURED_DISK __cpu_to_le32(0x02) /* Member of some RaidDev */
65#define FAILED_DISK __cpu_to_le32(0x04) /* Permanent failure */
66#define USABLE_DISK __cpu_to_le32(0x08) /* Fully usable unless FAILED_DISK is set */
cdddbdbc 67 __u32 status; /* 0xF0 - 0xF3 */
fe7ed8cb
DW
68 __u32 owner_cfg_num; /* which config 0,1,2... owns this disk */
69#define IMSM_DISK_FILLERS 4
cdddbdbc
DW
70 __u32 filler[IMSM_DISK_FILLERS]; /* 0xF4 - 0x107 MPB_DISK_FILLERS for future expansion */
71};
72
73/* RAID map configuration infos. */
74struct imsm_map {
75 __u32 pba_of_lba0; /* start address of partition */
76 __u32 blocks_per_member;/* blocks per member */
77 __u32 num_data_stripes; /* number of data stripes */
78 __u16 blocks_per_strip;
79 __u8 map_state; /* Normal, Uninitialized, Degraded, Failed */
80#define IMSM_T_STATE_NORMAL 0
81#define IMSM_T_STATE_UNINITIALIZED 1
e3bba0e0
DW
82#define IMSM_T_STATE_DEGRADED 2
83#define IMSM_T_STATE_FAILED 3
cdddbdbc
DW
84 __u8 raid_level;
85#define IMSM_T_RAID0 0
86#define IMSM_T_RAID1 1
87#define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */
88 __u8 num_members; /* number of member disks */
fe7ed8cb
DW
89 __u8 num_domains; /* number of parity domains */
90 __u8 failed_disk_num; /* valid only when state is degraded */
91 __u8 reserved[1];
cdddbdbc 92 __u32 filler[7]; /* expansion area */
7eef0453 93#define IMSM_ORD_REBUILD (1 << 24)
cdddbdbc 94 __u32 disk_ord_tbl[1]; /* disk_ord_tbl[num_members],
7eef0453
DW
95 * top byte contains some flags
96 */
cdddbdbc
DW
97} __attribute__ ((packed));
98
99struct imsm_vol {
f8f603f1 100 __u32 curr_migr_unit;
fe7ed8cb 101 __u32 checkpoint_id; /* id to access curr_migr_unit */
cdddbdbc 102 __u8 migr_state; /* Normal or Migrating */
e3bba0e0
DW
103#define MIGR_INIT 0
104#define MIGR_REBUILD 1
105#define MIGR_VERIFY 2 /* analagous to echo check > sync_action */
106#define MIGR_GEN_MIGR 3
107#define MIGR_STATE_CHANGE 4
cdddbdbc
DW
108 __u8 migr_type; /* Initializing, Rebuilding, ... */
109 __u8 dirty;
fe7ed8cb
DW
110 __u8 fs_state; /* fast-sync state for CnG (0xff == disabled) */
111 __u16 verify_errors; /* number of mismatches */
112 __u16 bad_blocks; /* number of bad blocks during verify */
113 __u32 filler[4];
cdddbdbc
DW
114 struct imsm_map map[1];
115 /* here comes another one if migr_state */
116} __attribute__ ((packed));
117
118struct imsm_dev {
fe7ed8cb 119 __u8 volume[MAX_RAID_SERIAL_LEN];
cdddbdbc
DW
120 __u32 size_low;
121 __u32 size_high;
fe7ed8cb
DW
122#define DEV_BOOTABLE __cpu_to_le32(0x01)
123#define DEV_BOOT_DEVICE __cpu_to_le32(0x02)
124#define DEV_READ_COALESCING __cpu_to_le32(0x04)
125#define DEV_WRITE_COALESCING __cpu_to_le32(0x08)
126#define DEV_LAST_SHUTDOWN_DIRTY __cpu_to_le32(0x10)
127#define DEV_HIDDEN_AT_BOOT __cpu_to_le32(0x20)
128#define DEV_CURRENTLY_HIDDEN __cpu_to_le32(0x40)
129#define DEV_VERIFY_AND_FIX __cpu_to_le32(0x80)
130#define DEV_MAP_STATE_UNINIT __cpu_to_le32(0x100)
131#define DEV_NO_AUTO_RECOVERY __cpu_to_le32(0x200)
132#define DEV_CLONE_N_GO __cpu_to_le32(0x400)
133#define DEV_CLONE_MAN_SYNC __cpu_to_le32(0x800)
134#define DEV_CNG_MASTER_DISK_NUM __cpu_to_le32(0x1000)
cdddbdbc
DW
135 __u32 status; /* Persistent RaidDev status */
136 __u32 reserved_blocks; /* Reserved blocks at beginning of volume */
fe7ed8cb
DW
137 __u8 migr_priority;
138 __u8 num_sub_vols;
139 __u8 tid;
140 __u8 cng_master_disk;
141 __u16 cache_policy;
142 __u8 cng_state;
143 __u8 cng_sub_state;
144#define IMSM_DEV_FILLERS 10
cdddbdbc
DW
145 __u32 filler[IMSM_DEV_FILLERS];
146 struct imsm_vol vol;
147} __attribute__ ((packed));
148
149struct imsm_super {
150 __u8 sig[MAX_SIGNATURE_LENGTH]; /* 0x00 - 0x1F */
151 __u32 check_sum; /* 0x20 - 0x23 MPB Checksum */
152 __u32 mpb_size; /* 0x24 - 0x27 Size of MPB */
153 __u32 family_num; /* 0x28 - 0x2B Checksum from first time this config was written */
154 __u32 generation_num; /* 0x2C - 0x2F Incremented each time this array's MPB is written */
604b746f
JD
155 __u32 error_log_size; /* 0x30 - 0x33 in bytes */
156 __u32 attributes; /* 0x34 - 0x37 */
cdddbdbc
DW
157 __u8 num_disks; /* 0x38 Number of configured disks */
158 __u8 num_raid_devs; /* 0x39 Number of configured volumes */
604b746f
JD
159 __u8 error_log_pos; /* 0x3A */
160 __u8 fill[1]; /* 0x3B */
161 __u32 cache_size; /* 0x3c - 0x40 in mb */
162 __u32 orig_family_num; /* 0x40 - 0x43 original family num */
163 __u32 pwr_cycle_count; /* 0x44 - 0x47 simulated power cycle count for array */
164 __u32 bbm_log_size; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */
165#define IMSM_FILLERS 35
166 __u32 filler[IMSM_FILLERS]; /* 0x4C - 0xD7 RAID_MPB_FILLERS */
cdddbdbc
DW
167 struct imsm_disk disk[1]; /* 0xD8 diskTbl[numDisks] */
168 /* here comes imsm_dev[num_raid_devs] */
604b746f 169 /* here comes BBM logs */
cdddbdbc
DW
170} __attribute__ ((packed));
171
604b746f
JD
172#define BBM_LOG_MAX_ENTRIES 254
173
174struct bbm_log_entry {
175 __u64 defective_block_start;
176#define UNREADABLE 0xFFFFFFFF
177 __u32 spare_block_offset;
178 __u16 remapped_marked_count;
179 __u16 disk_ordinal;
180} __attribute__ ((__packed__));
181
182struct bbm_log {
183 __u32 signature; /* 0xABADB10C */
184 __u32 entry_count;
185 __u32 reserved_spare_block_count; /* 0 */
186 __u32 reserved; /* 0xFFFF */
187 __u64 first_spare_lba;
188 struct bbm_log_entry mapped_block_entries[BBM_LOG_MAX_ENTRIES];
189} __attribute__ ((__packed__));
190
191
cdddbdbc
DW
192#ifndef MDASSEMBLE
193static char *map_state_str[] = { "normal", "uninitialized", "degraded", "failed" };
194#endif
195
87eb16df 196static unsigned int sector_count(__u32 bytes)
cdddbdbc 197{
87eb16df
DW
198 return ((bytes + (512-1)) & (~(512-1))) / 512;
199}
cdddbdbc 200
87eb16df
DW
201static unsigned int mpb_sectors(struct imsm_super *mpb)
202{
203 return sector_count(__le32_to_cpu(mpb->mpb_size));
cdddbdbc
DW
204}
205
ba2de7ba
DW
206struct intel_dev {
207 struct imsm_dev *dev;
208 struct intel_dev *next;
209 int index;
210};
211
cdddbdbc
DW
212/* internal representation of IMSM metadata */
213struct intel_super {
214 union {
949c47a0
DW
215 void *buf; /* O_DIRECT buffer for reading/writing metadata */
216 struct imsm_super *anchor; /* immovable parameters */
cdddbdbc 217 };
949c47a0 218 size_t len; /* size of the 'buf' allocation */
4d7b1503
DW
219 void *next_buf; /* for realloc'ing buf from the manager */
220 size_t next_len;
c2c087e6
DW
221 int updates_pending; /* count of pending updates for mdmon */
222 int creating_imsm; /* flag to indicate container creation */
bf5a934a 223 int current_vol; /* index of raid device undergoing creation */
0dcecb2e 224 __u32 create_offset; /* common start for 'current_vol' */
ba2de7ba 225 struct intel_dev *devlist;
cdddbdbc
DW
226 struct dl {
227 struct dl *next;
228 int index;
229 __u8 serial[MAX_RAID_SERIAL_LEN];
230 int major, minor;
231 char *devname;
b9f594fe 232 struct imsm_disk disk;
cdddbdbc 233 int fd;
0dcecb2e
DW
234 int extent_cnt;
235 struct extent *e; /* for determining freespace @ create */
cdddbdbc 236 } *disks;
43dad3d6 237 struct dl *add; /* list of disks to add while mdmon active */
47ee5a45 238 struct dl *missing; /* disks removed while we weren't looking */
43dad3d6 239 struct bbm_log *bbm_log;
88c32bb1
DW
240 const char *hba; /* device path of the raid controller for this metadata */
241 const struct imsm_orom *orom; /* platform firmware support */
cdddbdbc
DW
242};
243
c2c087e6
DW
244struct extent {
245 unsigned long long start, size;
246};
247
88758e9d
DW
248/* definition of messages passed to imsm_process_update */
249enum imsm_update_type {
250 update_activate_spare,
8273f55e 251 update_create_array,
43dad3d6 252 update_add_disk,
88758e9d
DW
253};
254
255struct imsm_update_activate_spare {
256 enum imsm_update_type type;
d23fe947 257 struct dl *dl;
88758e9d
DW
258 int slot;
259 int array;
260 struct imsm_update_activate_spare *next;
261};
262
54c2c1ea
DW
263struct disk_info {
264 __u8 serial[MAX_RAID_SERIAL_LEN];
265};
266
8273f55e
DW
267struct imsm_update_create_array {
268 enum imsm_update_type type;
8273f55e 269 int dev_idx;
6a3e913e 270 struct imsm_dev dev;
8273f55e
DW
271};
272
43dad3d6
DW
273struct imsm_update_add_disk {
274 enum imsm_update_type type;
275};
276
cdddbdbc
DW
277static struct supertype *match_metadata_desc_imsm(char *arg)
278{
279 struct supertype *st;
280
281 if (strcmp(arg, "imsm") != 0 &&
282 strcmp(arg, "default") != 0
283 )
284 return NULL;
285
286 st = malloc(sizeof(*st));
ef609477 287 memset(st, 0, sizeof(*st));
cdddbdbc
DW
288 st->ss = &super_imsm;
289 st->max_devs = IMSM_MAX_DEVICES;
290 st->minor_version = 0;
291 st->sb = NULL;
292 return st;
293}
294
0e600426 295#ifndef MDASSEMBLE
cdddbdbc
DW
296static __u8 *get_imsm_version(struct imsm_super *mpb)
297{
298 return &mpb->sig[MPB_SIG_LEN];
299}
0e600426 300#endif
cdddbdbc 301
949c47a0
DW
302/* retrieve a disk directly from the anchor when the anchor is known to be
303 * up-to-date, currently only at load time
304 */
305static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index)
cdddbdbc 306{
949c47a0 307 if (index >= mpb->num_disks)
cdddbdbc
DW
308 return NULL;
309 return &mpb->disk[index];
310}
311
0e600426 312#ifndef MDASSEMBLE
b9f594fe 313/* retrieve a disk from the parsed metadata */
949c47a0
DW
314static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index)
315{
b9f594fe
DW
316 struct dl *d;
317
318 for (d = super->disks; d; d = d->next)
319 if (d->index == index)
320 return &d->disk;
321
322 return NULL;
949c47a0 323}
0e600426 324#endif
949c47a0
DW
325
326/* generate a checksum directly from the anchor when the anchor is known to be
327 * up-to-date, currently only at load or write_super after coalescing
328 */
329static __u32 __gen_imsm_checksum(struct imsm_super *mpb)
cdddbdbc
DW
330{
331 __u32 end = mpb->mpb_size / sizeof(end);
332 __u32 *p = (__u32 *) mpb;
333 __u32 sum = 0;
334
97f734fd
N
335 while (end--) {
336 sum += __le32_to_cpu(*p);
337 p++;
338 }
cdddbdbc
DW
339
340 return sum - __le32_to_cpu(mpb->check_sum);
341}
342
a965f303
DW
343static size_t sizeof_imsm_map(struct imsm_map *map)
344{
345 return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1);
346}
347
348struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map)
cdddbdbc 349{
a965f303
DW
350 struct imsm_map *map = &dev->vol.map[0];
351
352 if (second_map && !dev->vol.migr_state)
353 return NULL;
354 else if (second_map) {
355 void *ptr = map;
356
357 return ptr + sizeof_imsm_map(map);
358 } else
359 return map;
360
361}
cdddbdbc 362
3393c6af
DW
363/* return the size of the device.
364 * migr_state increases the returned size if map[0] were to be duplicated
365 */
366static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state)
a965f303
DW
367{
368 size_t size = sizeof(*dev) - sizeof(struct imsm_map) +
369 sizeof_imsm_map(get_imsm_map(dev, 0));
cdddbdbc
DW
370
371 /* migrating means an additional map */
a965f303
DW
372 if (dev->vol.migr_state)
373 size += sizeof_imsm_map(get_imsm_map(dev, 1));
3393c6af
DW
374 else if (migr_state)
375 size += sizeof_imsm_map(get_imsm_map(dev, 0));
cdddbdbc
DW
376
377 return size;
378}
379
54c2c1ea
DW
380#ifndef MDASSEMBLE
381/* retrieve disk serial number list from a metadata update */
382static struct disk_info *get_disk_info(struct imsm_update_create_array *update)
383{
384 void *u = update;
385 struct disk_info *inf;
386
387 inf = u + sizeof(*update) - sizeof(struct imsm_dev) +
388 sizeof_imsm_dev(&update->dev, 0);
389
390 return inf;
391}
392#endif
393
949c47a0 394static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index)
cdddbdbc
DW
395{
396 int offset;
397 int i;
398 void *_mpb = mpb;
399
949c47a0 400 if (index >= mpb->num_raid_devs)
cdddbdbc
DW
401 return NULL;
402
403 /* devices start after all disks */
404 offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb;
405
406 for (i = 0; i <= index; i++)
407 if (i == index)
408 return _mpb + offset;
409 else
3393c6af 410 offset += sizeof_imsm_dev(_mpb + offset, 0);
cdddbdbc
DW
411
412 return NULL;
413}
414
949c47a0
DW
415static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index)
416{
ba2de7ba
DW
417 struct intel_dev *dv;
418
949c47a0
DW
419 if (index >= super->anchor->num_raid_devs)
420 return NULL;
ba2de7ba
DW
421 for (dv = super->devlist; dv; dv = dv->next)
422 if (dv->index == index)
423 return dv->dev;
424 return NULL;
949c47a0
DW
425}
426
7eef0453
DW
427static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev, int slot)
428{
429 struct imsm_map *map;
430
431 if (dev->vol.migr_state)
7eef0453 432 map = get_imsm_map(dev, 1);
fb9bf0d3
DW
433 else
434 map = get_imsm_map(dev, 0);
7eef0453 435
ff077194
DW
436 /* top byte identifies disk under rebuild */
437 return __le32_to_cpu(map->disk_ord_tbl[slot]);
438}
439
440#define ord_to_idx(ord) (((ord) << 8) >> 8)
441static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot)
442{
443 __u32 ord = get_imsm_ord_tbl_ent(dev, slot);
444
445 return ord_to_idx(ord);
7eef0453
DW
446}
447
be73972f
DW
448static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord)
449{
450 map->disk_ord_tbl[slot] = __cpu_to_le32(ord);
451}
452
620b1713
DW
453static int get_imsm_disk_slot(struct imsm_map *map, int idx)
454{
455 int slot;
456 __u32 ord;
457
458 for (slot = 0; slot < map->num_members; slot++) {
459 ord = __le32_to_cpu(map->disk_ord_tbl[slot]);
460 if (ord_to_idx(ord) == idx)
461 return slot;
462 }
463
464 return -1;
465}
466
cdddbdbc
DW
467static int get_imsm_raid_level(struct imsm_map *map)
468{
469 if (map->raid_level == 1) {
470 if (map->num_members == 2)
471 return 1;
472 else
473 return 10;
474 }
475
476 return map->raid_level;
477}
478
c2c087e6
DW
479static int cmp_extent(const void *av, const void *bv)
480{
481 const struct extent *a = av;
482 const struct extent *b = bv;
483 if (a->start < b->start)
484 return -1;
485 if (a->start > b->start)
486 return 1;
487 return 0;
488}
489
0dcecb2e 490static int count_memberships(struct dl *dl, struct intel_super *super)
c2c087e6 491{
c2c087e6 492 int memberships = 0;
620b1713 493 int i;
c2c087e6 494
949c47a0
DW
495 for (i = 0; i < super->anchor->num_raid_devs; i++) {
496 struct imsm_dev *dev = get_imsm_dev(super, i);
a965f303 497 struct imsm_map *map = get_imsm_map(dev, 0);
c2c087e6 498
620b1713
DW
499 if (get_imsm_disk_slot(map, dl->index) >= 0)
500 memberships++;
c2c087e6 501 }
0dcecb2e
DW
502
503 return memberships;
504}
505
506static struct extent *get_extents(struct intel_super *super, struct dl *dl)
507{
508 /* find a list of used extents on the given physical device */
509 struct extent *rv, *e;
620b1713 510 int i;
0dcecb2e
DW
511 int memberships = count_memberships(dl, super);
512 __u32 reservation = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
513
c2c087e6
DW
514 rv = malloc(sizeof(struct extent) * (memberships + 1));
515 if (!rv)
516 return NULL;
517 e = rv;
518
949c47a0
DW
519 for (i = 0; i < super->anchor->num_raid_devs; i++) {
520 struct imsm_dev *dev = get_imsm_dev(super, i);
a965f303 521 struct imsm_map *map = get_imsm_map(dev, 0);
c2c087e6 522
620b1713
DW
523 if (get_imsm_disk_slot(map, dl->index) >= 0) {
524 e->start = __le32_to_cpu(map->pba_of_lba0);
525 e->size = __le32_to_cpu(map->blocks_per_member);
526 e++;
c2c087e6
DW
527 }
528 }
529 qsort(rv, memberships, sizeof(*rv), cmp_extent);
530
14e8215b
DW
531 /* determine the start of the metadata
532 * when no raid devices are defined use the default
533 * ...otherwise allow the metadata to truncate the value
534 * as is the case with older versions of imsm
535 */
536 if (memberships) {
537 struct extent *last = &rv[memberships - 1];
538 __u32 remainder;
539
540 remainder = __le32_to_cpu(dl->disk.total_blocks) -
541 (last->start + last->size);
dda5855f
DW
542 /* round down to 1k block to satisfy precision of the kernel
543 * 'size' interface
544 */
545 remainder &= ~1UL;
546 /* make sure remainder is still sane */
547 if (remainder < ROUND_UP(super->len, 512) >> 9)
548 remainder = ROUND_UP(super->len, 512) >> 9;
14e8215b
DW
549 if (reservation > remainder)
550 reservation = remainder;
551 }
552 e->start = __le32_to_cpu(dl->disk.total_blocks) - reservation;
c2c087e6
DW
553 e->size = 0;
554 return rv;
555}
556
14e8215b
DW
557/* try to determine how much space is reserved for metadata from
558 * the last get_extents() entry, otherwise fallback to the
559 * default
560 */
561static __u32 imsm_reserved_sectors(struct intel_super *super, struct dl *dl)
562{
563 struct extent *e;
564 int i;
565 __u32 rv;
566
567 /* for spares just return a minimal reservation which will grow
568 * once the spare is picked up by an array
569 */
570 if (dl->index == -1)
571 return MPB_SECTOR_CNT;
572
573 e = get_extents(super, dl);
574 if (!e)
575 return MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
576
577 /* scroll to last entry */
578 for (i = 0; e[i].size; i++)
579 continue;
580
581 rv = __le32_to_cpu(dl->disk.total_blocks) - e[i].start;
582
583 free(e);
584
585 return rv;
586}
587
588#ifndef MDASSEMBLE
44470971 589static void print_imsm_dev(struct imsm_dev *dev, char *uuid, int disk_idx)
cdddbdbc
DW
590{
591 __u64 sz;
592 int slot;
a965f303 593 struct imsm_map *map = get_imsm_map(dev, 0);
b10b37b8 594 __u32 ord;
cdddbdbc
DW
595
596 printf("\n");
1e7bc0ed 597 printf("[%.16s]:\n", dev->volume);
44470971 598 printf(" UUID : %s\n", uuid);
cdddbdbc
DW
599 printf(" RAID Level : %d\n", get_imsm_raid_level(map));
600 printf(" Members : %d\n", map->num_members);
620b1713
DW
601 slot = get_imsm_disk_slot(map, disk_idx);
602 if (slot >= 0) {
b10b37b8
DW
603 ord = get_imsm_ord_tbl_ent(dev, slot);
604 printf(" This Slot : %d%s\n", slot,
605 ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : "");
606 } else
cdddbdbc
DW
607 printf(" This Slot : ?\n");
608 sz = __le32_to_cpu(dev->size_high);
609 sz <<= 32;
610 sz += __le32_to_cpu(dev->size_low);
611 printf(" Array Size : %llu%s\n", (unsigned long long)sz,
612 human_size(sz * 512));
613 sz = __le32_to_cpu(map->blocks_per_member);
614 printf(" Per Dev Size : %llu%s\n", (unsigned long long)sz,
615 human_size(sz * 512));
616 printf(" Sector Offset : %u\n",
617 __le32_to_cpu(map->pba_of_lba0));
618 printf(" Num Stripes : %u\n",
619 __le32_to_cpu(map->num_data_stripes));
620 printf(" Chunk Size : %u KiB\n",
621 __le16_to_cpu(map->blocks_per_strip) / 2);
622 printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks));
3393c6af
DW
623 printf(" Migrate State : %s", dev->vol.migr_state ? "migrating" : "idle");
624 if (dev->vol.migr_state)
625 printf(": %s", dev->vol.migr_type ? "rebuilding" : "initializing");
626 printf("\n");
627 printf(" Map State : %s", map_state_str[map->map_state]);
628 if (dev->vol.migr_state) {
629 struct imsm_map *map = get_imsm_map(dev, 1);
b10b37b8 630 printf(" <-- %s", map_state_str[map->map_state]);
3393c6af
DW
631 }
632 printf("\n");
cdddbdbc 633 printf(" Dirty State : %s\n", dev->vol.dirty ? "dirty" : "clean");
cdddbdbc
DW
634}
635
14e8215b 636static void print_imsm_disk(struct imsm_super *mpb, int index, __u32 reserved)
cdddbdbc 637{
949c47a0 638 struct imsm_disk *disk = __get_imsm_disk(mpb, index);
1f24f035 639 char str[MAX_RAID_SERIAL_LEN + 1];
cdddbdbc
DW
640 __u32 s;
641 __u64 sz;
642
e9d82038
DW
643 if (index < 0)
644 return;
645
cdddbdbc 646 printf("\n");
1f24f035 647 snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial);
cdddbdbc 648 printf(" Disk%02d Serial : %s\n", index, str);
f2f27e63 649 s = disk->status;
cdddbdbc
DW
650 printf(" State :%s%s%s%s\n", s&SPARE_DISK ? " spare" : "",
651 s&CONFIGURED_DISK ? " active" : "",
652 s&FAILED_DISK ? " failed" : "",
653 s&USABLE_DISK ? " usable" : "");
654 printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id));
14e8215b 655 sz = __le32_to_cpu(disk->total_blocks) - reserved;
cdddbdbc
DW
656 printf(" Usable Size : %llu%s\n", (unsigned long long)sz,
657 human_size(sz * 512));
658}
659
44470971
DW
660static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info);
661
cdddbdbc
DW
662static void examine_super_imsm(struct supertype *st, char *homehost)
663{
664 struct intel_super *super = st->sb;
949c47a0 665 struct imsm_super *mpb = super->anchor;
cdddbdbc
DW
666 char str[MAX_SIGNATURE_LENGTH];
667 int i;
27fd6274
DW
668 struct mdinfo info;
669 char nbuf[64];
cdddbdbc 670 __u32 sum;
14e8215b 671 __u32 reserved = imsm_reserved_sectors(super, super->disks);
cdddbdbc 672
27fd6274 673
cdddbdbc
DW
674 snprintf(str, MPB_SIG_LEN, "%s", mpb->sig);
675 printf(" Magic : %s\n", str);
676 snprintf(str, strlen(MPB_VERSION_RAID0), "%s", get_imsm_version(mpb));
677 printf(" Version : %s\n", get_imsm_version(mpb));
678 printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num));
679 printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num));
27fd6274
DW
680 getinfo_super_imsm(st, &info);
681 fname_from_uuid(st, &info, nbuf,'-');
682 printf(" UUID : %s\n", nbuf + 5);
cdddbdbc
DW
683 sum = __le32_to_cpu(mpb->check_sum);
684 printf(" Checksum : %08x %s\n", sum,
949c47a0 685 __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect");
87eb16df 686 printf(" MPB Sectors : %d\n", mpb_sectors(mpb));
cdddbdbc
DW
687 printf(" Disks : %d\n", mpb->num_disks);
688 printf(" RAID Devices : %d\n", mpb->num_raid_devs);
14e8215b 689 print_imsm_disk(mpb, super->disks->index, reserved);
604b746f
JD
690 if (super->bbm_log) {
691 struct bbm_log *log = super->bbm_log;
692
693 printf("\n");
694 printf("Bad Block Management Log:\n");
695 printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size));
696 printf(" Signature : %x\n", __le32_to_cpu(log->signature));
697 printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count));
698 printf(" Spare Blocks : %d\n", __le32_to_cpu(log->reserved_spare_block_count));
699 printf(" First Spare : %llx\n", __le64_to_cpu(log->first_spare_lba));
700 }
44470971
DW
701 for (i = 0; i < mpb->num_raid_devs; i++) {
702 struct mdinfo info;
703 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
704
705 super->current_vol = i;
706 getinfo_super_imsm(st, &info);
707 fname_from_uuid(st, &info, nbuf, '-');
708 print_imsm_dev(dev, nbuf + 5, super->disks->index);
709 }
cdddbdbc
DW
710 for (i = 0; i < mpb->num_disks; i++) {
711 if (i == super->disks->index)
712 continue;
14e8215b 713 print_imsm_disk(mpb, i, reserved);
cdddbdbc
DW
714 }
715}
716
717static void brief_examine_super_imsm(struct supertype *st)
718{
27fd6274 719 /* We just write a generic IMSM ARRAY entry */
ff54de6e
N
720 struct mdinfo info;
721 char nbuf[64];
cf8de691 722 char nbuf1[64];
1e7bc0ed
DW
723 struct intel_super *super = st->sb;
724 int i;
725
726 if (!super->anchor->num_raid_devs)
727 return;
ff54de6e
N
728
729 getinfo_super_imsm(st, &info);
730 fname_from_uuid(st, &info, nbuf,'-');
cf8de691 731 printf("ARRAY metadata=imsm auto=md UUID=%s\n", nbuf + 5);
1e7bc0ed
DW
732 for (i = 0; i < super->anchor->num_raid_devs; i++) {
733 struct imsm_dev *dev = get_imsm_dev(super, i);
734
735 super->current_vol = i;
736 getinfo_super_imsm(st, &info);
cf8de691 737 fname_from_uuid(st, &info, nbuf1,'-');
44470971
DW
738 printf("ARRAY /dev/md/%.16s container=%s\n"
739 " member=%d auto=mdp UUID=%s\n",
cf8de691 740 dev->volume, nbuf + 5, i, nbuf1 + 5);
1e7bc0ed 741 }
cdddbdbc
DW
742}
743
744static void detail_super_imsm(struct supertype *st, char *homehost)
745{
3ebe00a1
DW
746 struct mdinfo info;
747 char nbuf[64];
748
749 getinfo_super_imsm(st, &info);
750 fname_from_uuid(st, &info, nbuf,'-');
751 printf("\n UUID : %s\n", nbuf + 5);
cdddbdbc
DW
752}
753
754static void brief_detail_super_imsm(struct supertype *st)
755{
ff54de6e
N
756 struct mdinfo info;
757 char nbuf[64];
758 getinfo_super_imsm(st, &info);
759 fname_from_uuid(st, &info, nbuf,'-');
760 printf(" UUID=%s", nbuf + 5);
cdddbdbc 761}
d665cc31
DW
762
763static int imsm_read_serial(int fd, char *devname, __u8 *serial);
764static void fd2devname(int fd, char *name);
765
766static int imsm_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose)
767{
768 /* dump an unsorted list of devices attached to ahci, as well as
769 * non-connected ports
770 */
771 int hba_len = strlen(hba_path) + 1;
772 struct dirent *ent;
773 DIR *dir;
774 char *path = NULL;
775 int err = 0;
776 unsigned long port_mask = (1 << port_count) - 1;
777
778 if (port_count > sizeof(port_mask) * 8) {
779 if (verbose)
780 fprintf(stderr, Name ": port_count %d out of range\n", port_count);
781 return 2;
782 }
783
784 /* scroll through /sys/dev/block looking for devices attached to
785 * this hba
786 */
787 dir = opendir("/sys/dev/block");
788 for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) {
789 int fd;
790 char model[64];
791 char vendor[64];
792 char buf[1024];
793 int major, minor;
794 char *device;
795 char *c;
796 int port;
797 int type;
798
799 if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2)
800 continue;
801 path = devt_to_devpath(makedev(major, minor));
802 if (!path)
803 continue;
804 if (!path_attached_to_hba(path, hba_path)) {
805 free(path);
806 path = NULL;
807 continue;
808 }
809
810 /* retrieve the scsi device type */
811 if (asprintf(&device, "/sys/dev/block/%d:%d/device/xxxxxxx", major, minor) < 0) {
812 if (verbose)
813 fprintf(stderr, Name ": failed to allocate 'device'\n");
814 err = 2;
815 break;
816 }
817 sprintf(device, "/sys/dev/block/%d:%d/device/type", major, minor);
818 if (load_sys(device, buf) != 0) {
819 if (verbose)
820 fprintf(stderr, Name ": failed to read device type for %s\n",
821 path);
822 err = 2;
823 free(device);
824 break;
825 }
826 type = strtoul(buf, NULL, 10);
827
828 /* if it's not a disk print the vendor and model */
829 if (!(type == 0 || type == 7 || type == 14)) {
830 vendor[0] = '\0';
831 model[0] = '\0';
832 sprintf(device, "/sys/dev/block/%d:%d/device/vendor", major, minor);
833 if (load_sys(device, buf) == 0) {
834 strncpy(vendor, buf, sizeof(vendor));
835 vendor[sizeof(vendor) - 1] = '\0';
836 c = (char *) &vendor[sizeof(vendor) - 1];
837 while (isspace(*c) || *c == '\0')
838 *c-- = '\0';
839
840 }
841 sprintf(device, "/sys/dev/block/%d:%d/device/model", major, minor);
842 if (load_sys(device, buf) == 0) {
843 strncpy(model, buf, sizeof(model));
844 model[sizeof(model) - 1] = '\0';
845 c = (char *) &model[sizeof(model) - 1];
846 while (isspace(*c) || *c == '\0')
847 *c-- = '\0';
848 }
849
850 if (vendor[0] && model[0])
851 sprintf(buf, "%.64s %.64s", vendor, model);
852 else
853 switch (type) { /* numbers from hald/linux/device.c */
854 case 1: sprintf(buf, "tape"); break;
855 case 2: sprintf(buf, "printer"); break;
856 case 3: sprintf(buf, "processor"); break;
857 case 4:
858 case 5: sprintf(buf, "cdrom"); break;
859 case 6: sprintf(buf, "scanner"); break;
860 case 8: sprintf(buf, "media_changer"); break;
861 case 9: sprintf(buf, "comm"); break;
862 case 12: sprintf(buf, "raid"); break;
863 default: sprintf(buf, "unknown");
864 }
865 } else
866 buf[0] = '\0';
867 free(device);
868
869 /* chop device path to 'host%d' and calculate the port number */
870 c = strchr(&path[hba_len], '/');
871 *c = '\0';
872 if (sscanf(&path[hba_len], "host%d", &port) == 1)
873 port -= host_base;
874 else {
875 if (verbose) {
876 *c = '/'; /* repair the full string */
877 fprintf(stderr, Name ": failed to determine port number for %s\n",
878 path);
879 }
880 err = 2;
881 break;
882 }
883
884 /* mark this port as used */
885 port_mask &= ~(1 << port);
886
887 /* print out the device information */
888 if (buf[0]) {
889 printf(" Port%d : - non-disk device (%s) -\n", port, buf);
890 continue;
891 }
892
893 fd = dev_open(ent->d_name, O_RDONLY);
894 if (fd < 0)
895 printf(" Port%d : - disk info unavailable -\n", port);
896 else {
897 fd2devname(fd, buf);
898 printf(" Port%d : %s", port, buf);
899 if (imsm_read_serial(fd, NULL, (__u8 *) buf) == 0)
900 printf(" (%s)\n", buf);
901 else
902 printf("()\n");
903 }
904 close(fd);
905 free(path);
906 path = NULL;
907 }
908 if (path)
909 free(path);
910 if (dir)
911 closedir(dir);
912 if (err == 0) {
913 int i;
914
915 for (i = 0; i < port_count; i++)
916 if (port_mask & (1 << i))
917 printf(" Port%d : - no device attached -\n", i);
918 }
919
920 return err;
921}
922
5615172f 923static int detail_platform_imsm(int verbose, int enumerate_only)
d665cc31
DW
924{
925 /* There are two components to imsm platform support, the ahci SATA
926 * controller and the option-rom. To find the SATA controller we
927 * simply look in /sys/bus/pci/drivers/ahci to see if an ahci
928 * controller with the Intel vendor id is present. This approach
929 * allows mdadm to leverage the kernel's ahci detection logic, with the
930 * caveat that if ahci.ko is not loaded mdadm will not be able to
931 * detect platform raid capabilities. The option-rom resides in a
932 * platform "Adapter ROM". We scan for its signature to retrieve the
933 * platform capabilities. If raid support is disabled in the BIOS the
934 * option-rom capability structure will not be available.
935 */
936 const struct imsm_orom *orom;
937 struct sys_dev *list, *hba;
938 DIR *dir;
939 struct dirent *ent;
940 const char *hba_path;
941 int host_base = 0;
942 int port_count = 0;
943
5615172f
DW
944 if (enumerate_only) {
945 if (check_env("IMSM_NO_PLATFORM") || find_imsm_orom())
946 return 0;
947 return 2;
948 }
949
d665cc31
DW
950 list = find_driver_devices("pci", "ahci");
951 for (hba = list; hba; hba = hba->next)
952 if (devpath_to_vendor(hba->path) == 0x8086)
953 break;
954
955 if (!hba) {
956 if (verbose)
957 fprintf(stderr, Name ": unable to find active ahci controller\n");
958 free_sys_dev(&list);
959 return 2;
960 } else if (verbose)
961 fprintf(stderr, Name ": found Intel SATA AHCI Controller\n");
962 hba_path = hba->path;
963 hba->path = NULL;
964 free_sys_dev(&list);
965
966 orom = find_imsm_orom();
967 if (!orom) {
968 if (verbose)
969 fprintf(stderr, Name ": imsm option-rom not found\n");
970 return 2;
971 }
972
973 printf(" Platform : Intel(R) Matrix Storage Manager\n");
974 printf(" Version : %d.%d.%d.%d\n", orom->major_ver, orom->minor_ver,
975 orom->hotfix_ver, orom->build);
976 printf(" RAID Levels :%s%s%s%s%s\n",
977 imsm_orom_has_raid0(orom) ? " raid0" : "",
978 imsm_orom_has_raid1(orom) ? " raid1" : "",
979 imsm_orom_has_raid1e(orom) ? " raid1e" : "",
980 imsm_orom_has_raid10(orom) ? " raid10" : "",
981 imsm_orom_has_raid5(orom) ? " raid5" : "");
982 printf(" Max Disks : %d\n", orom->tds);
983 printf(" Max Volumes : %d\n", orom->vpa);
984 printf(" I/O Controller : %s\n", hba_path);
985
986 /* find the smallest scsi host number to determine a port number base */
987 dir = opendir(hba_path);
988 for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) {
989 int host;
990
991 if (sscanf(ent->d_name, "host%d", &host) != 1)
992 continue;
993 if (port_count == 0)
994 host_base = host;
995 else if (host < host_base)
996 host_base = host;
997
998 if (host + 1 > port_count + host_base)
999 port_count = host + 1 - host_base;
1000
1001 }
1002 if (dir)
1003 closedir(dir);
1004
1005 if (!port_count || imsm_enumerate_ports(hba_path, port_count,
1006 host_base, verbose) != 0) {
1007 if (verbose)
1008 fprintf(stderr, Name ": failed to enumerate ports\n");
1009 return 2;
1010 }
1011
1012 return 0;
1013}
cdddbdbc
DW
1014#endif
1015
1016static int match_home_imsm(struct supertype *st, char *homehost)
1017{
5115ca67
DW
1018 /* the imsm metadata format does not specify any host
1019 * identification information. We return -1 since we can never
1020 * confirm nor deny whether a given array is "meant" for this
1021 * host. We rely on compare_super and the 'family_num' field to
1022 * exclude member disks that do not belong, and we rely on
1023 * mdadm.conf to specify the arrays that should be assembled.
1024 * Auto-assembly may still pick up "foreign" arrays.
1025 */
cdddbdbc 1026
9362c1c8 1027 return -1;
cdddbdbc
DW
1028}
1029
1030static void uuid_from_super_imsm(struct supertype *st, int uuid[4])
1031{
51006d85
N
1032 /* The uuid returned here is used for:
1033 * uuid to put into bitmap file (Create, Grow)
1034 * uuid for backup header when saving critical section (Grow)
1035 * comparing uuids when re-adding a device into an array
1036 * In these cases the uuid required is that of the data-array,
1037 * not the device-set.
1038 * uuid to recognise same set when adding a missing device back
1039 * to an array. This is a uuid for the device-set.
1040 *
1041 * For each of these we can make do with a truncated
1042 * or hashed uuid rather than the original, as long as
1043 * everyone agrees.
1044 * In each case the uuid required is that of the data-array,
1045 * not the device-set.
43dad3d6 1046 */
51006d85
N
1047 /* imsm does not track uuid's so we synthesis one using sha1 on
1048 * - The signature (Which is constant for all imsm array, but no matter)
1049 * - the family_num of the container
1050 * - the index number of the volume
1051 * - the 'serial' number of the volume.
1052 * Hopefully these are all constant.
1053 */
1054 struct intel_super *super = st->sb;
43dad3d6 1055
51006d85
N
1056 char buf[20];
1057 struct sha1_ctx ctx;
1058 struct imsm_dev *dev = NULL;
1059
1060 sha1_init_ctx(&ctx);
92bd8f8d 1061 sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx);
51006d85
N
1062 sha1_process_bytes(&super->anchor->family_num, sizeof(__u32), &ctx);
1063 if (super->current_vol >= 0)
1064 dev = get_imsm_dev(super, super->current_vol);
1065 if (dev) {
1066 __u32 vol = super->current_vol;
1067 sha1_process_bytes(&vol, sizeof(vol), &ctx);
1068 sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx);
1069 }
1070 sha1_finish_ctx(&ctx, buf);
1071 memcpy(uuid, buf, 4*4);
cdddbdbc
DW
1072}
1073
0d481d37 1074#if 0
4f5bc454
DW
1075static void
1076get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p)
cdddbdbc 1077{
cdddbdbc
DW
1078 __u8 *v = get_imsm_version(mpb);
1079 __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH;
1080 char major[] = { 0, 0, 0 };
1081 char minor[] = { 0 ,0, 0 };
1082 char patch[] = { 0, 0, 0 };
1083 char *ver_parse[] = { major, minor, patch };
1084 int i, j;
1085
1086 i = j = 0;
1087 while (*v != '\0' && v < end) {
1088 if (*v != '.' && j < 2)
1089 ver_parse[i][j++] = *v;
1090 else {
1091 i++;
1092 j = 0;
1093 }
1094 v++;
1095 }
1096
4f5bc454
DW
1097 *m = strtol(minor, NULL, 0);
1098 *p = strtol(patch, NULL, 0);
1099}
0d481d37 1100#endif
4f5bc454 1101
c2c087e6
DW
1102static int imsm_level_to_layout(int level)
1103{
1104 switch (level) {
1105 case 0:
1106 case 1:
1107 return 0;
1108 case 5:
1109 case 6:
a380c027 1110 return ALGORITHM_LEFT_ASYMMETRIC;
c2c087e6 1111 case 10:
c92a2527 1112 return 0x102;
c2c087e6 1113 }
a18a888e 1114 return UnSet;
c2c087e6
DW
1115}
1116
bf5a934a
DW
1117static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info)
1118{
1119 struct intel_super *super = st->sb;
949c47a0 1120 struct imsm_dev *dev = get_imsm_dev(super, super->current_vol);
a965f303 1121 struct imsm_map *map = get_imsm_map(dev, 0);
bf5a934a
DW
1122
1123 info->container_member = super->current_vol;
1124 info->array.raid_disks = map->num_members;
1125 info->array.level = get_imsm_raid_level(map);
1126 info->array.layout = imsm_level_to_layout(info->array.level);
1127 info->array.md_minor = -1;
1128 info->array.ctime = 0;
1129 info->array.utime = 0;
301406c9
DW
1130 info->array.chunk_size = __le16_to_cpu(map->blocks_per_strip) << 9;
1131 info->array.state = !dev->vol.dirty;
1132
1133 info->disk.major = 0;
1134 info->disk.minor = 0;
bf5a934a
DW
1135
1136 info->data_offset = __le32_to_cpu(map->pba_of_lba0);
1137 info->component_size = __le32_to_cpu(map->blocks_per_member);
301406c9 1138 memset(info->uuid, 0, sizeof(info->uuid));
bf5a934a 1139
f8f603f1 1140 if (map->map_state == IMSM_T_STATE_UNINITIALIZED || dev->vol.dirty)
301406c9 1141 info->resync_start = 0;
f8f603f1
DW
1142 else if (dev->vol.migr_state)
1143 info->resync_start = __le32_to_cpu(dev->vol.curr_migr_unit);
301406c9
DW
1144 else
1145 info->resync_start = ~0ULL;
1146
1147 strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN);
1148 info->name[MAX_RAID_SERIAL_LEN] = 0;
bf5a934a 1149
f35f2525
N
1150 info->array.major_version = -1;
1151 info->array.minor_version = -2;
bf5a934a
DW
1152 sprintf(info->text_version, "/%s/%d",
1153 devnum2devname(st->container_dev),
1154 info->container_member);
a67dd8cc 1155 info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */
51006d85 1156 uuid_from_super_imsm(st, info->uuid);
bf5a934a
DW
1157}
1158
7a70e8aa
DW
1159/* check the config file to see if we can return a real uuid for this spare */
1160static void fixup_container_spare_uuid(struct mdinfo *inf)
1161{
1162 struct mddev_ident_s *array_list;
1163
1164 if (inf->array.level != LEVEL_CONTAINER ||
1165 memcmp(inf->uuid, uuid_match_any, sizeof(int[4])) != 0)
1166 return;
1167
1168 array_list = conf_get_ident(NULL);
1169
1170 for (; array_list; array_list = array_list->next) {
1171 if (array_list->uuid_set) {
1172 struct supertype *_sst; /* spare supertype */
1173 struct supertype *_cst; /* container supertype */
1174
1175 _cst = array_list->st;
1176 _sst = _cst->ss->match_metadata_desc(inf->text_version);
1177 if (_sst) {
1178 memcpy(inf->uuid, array_list->uuid, sizeof(int[4]));
1179 free(_sst);
1180 break;
1181 }
1182 }
1183 }
1184}
bf5a934a 1185
4f5bc454
DW
1186static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info)
1187{
1188 struct intel_super *super = st->sb;
4f5bc454
DW
1189 struct imsm_disk *disk;
1190 __u32 s;
4f5bc454 1191
bf5a934a
DW
1192 if (super->current_vol >= 0) {
1193 getinfo_super_imsm_volume(st, info);
1194 return;
1195 }
d23fe947
DW
1196
1197 /* Set raid_disks to zero so that Assemble will always pull in valid
1198 * spares
1199 */
1200 info->array.raid_disks = 0;
cdddbdbc
DW
1201 info->array.level = LEVEL_CONTAINER;
1202 info->array.layout = 0;
1203 info->array.md_minor = -1;
c2c087e6 1204 info->array.ctime = 0; /* N/A for imsm */
cdddbdbc
DW
1205 info->array.utime = 0;
1206 info->array.chunk_size = 0;
1207
1208 info->disk.major = 0;
1209 info->disk.minor = 0;
cdddbdbc 1210 info->disk.raid_disk = -1;
c2c087e6 1211 info->reshape_active = 0;
f35f2525
N
1212 info->array.major_version = -1;
1213 info->array.minor_version = -2;
c2c087e6 1214 strcpy(info->text_version, "imsm");
a67dd8cc 1215 info->safe_mode_delay = 0;
c2c087e6
DW
1216 info->disk.number = -1;
1217 info->disk.state = 0;
c5afc314 1218 info->name[0] = 0;
c2c087e6 1219
4a04ec6c 1220 if (super->disks) {
14e8215b
DW
1221 __u32 reserved = imsm_reserved_sectors(super, super->disks);
1222
b9f594fe 1223 disk = &super->disks->disk;
14e8215b
DW
1224 info->data_offset = __le32_to_cpu(disk->total_blocks) - reserved;
1225 info->component_size = reserved;
f2f27e63 1226 s = disk->status;
4a04ec6c 1227 info->disk.state = s & CONFIGURED_DISK ? (1 << MD_DISK_ACTIVE) : 0;
df474657
DW
1228 /* we don't change info->disk.raid_disk here because
1229 * this state will be finalized in mdmon after we have
1230 * found the 'most fresh' version of the metadata
1231 */
1232 info->disk.state |= s & FAILED_DISK ? (1 << MD_DISK_FAULTY) : 0;
032e9e29 1233 info->disk.state |= s & SPARE_DISK ? 0 : (1 << MD_DISK_SYNC);
cdddbdbc 1234 }
a575e2a7
DW
1235
1236 /* only call uuid_from_super_imsm when this disk is part of a populated container,
1237 * ->compare_super may have updated the 'num_raid_devs' field for spares
1238 */
1239 if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs)
36ba7d48 1240 uuid_from_super_imsm(st, info->uuid);
7a70e8aa 1241 else {
032e9e29 1242 memcpy(info->uuid, uuid_match_any, sizeof(int[4]));
7a70e8aa
DW
1243 fixup_container_spare_uuid(info);
1244 }
cdddbdbc
DW
1245}
1246
cdddbdbc
DW
1247static int update_super_imsm(struct supertype *st, struct mdinfo *info,
1248 char *update, char *devname, int verbose,
1249 int uuid_set, char *homehost)
1250{
f352c545
DW
1251 /* FIXME */
1252
1253 /* For 'assemble' and 'force' we need to return non-zero if any
1254 * change was made. For others, the return value is ignored.
1255 * Update options are:
1256 * force-one : This device looks a bit old but needs to be included,
1257 * update age info appropriately.
1258 * assemble: clear any 'faulty' flag to allow this device to
1259 * be assembled.
1260 * force-array: Array is degraded but being forced, mark it clean
1261 * if that will be needed to assemble it.
1262 *
1263 * newdev: not used ????
1264 * grow: Array has gained a new device - this is currently for
1265 * linear only
1266 * resync: mark as dirty so a resync will happen.
1267 * name: update the name - preserving the homehost
1268 *
1269 * Following are not relevant for this imsm:
1270 * sparc2.2 : update from old dodgey metadata
1271 * super-minor: change the preferred_minor number
1272 * summaries: update redundant counters.
1273 * uuid: Change the uuid of the array to match watch is given
1274 * homehost: update the recorded homehost
1275 * _reshape_progress: record new reshape_progress position.
1276 */
1277 int rv = 0;
1278 //struct intel_super *super = st->sb;
1279 //struct imsm_super *mpb = super->mpb;
1280
1281 if (strcmp(update, "grow") == 0) {
1282 }
1283 if (strcmp(update, "resync") == 0) {
1284 /* dev->vol.dirty = 1; */
1285 }
1286
1287 /* IMSM has no concept of UUID or homehost */
1288
1289 return rv;
cdddbdbc
DW
1290}
1291
c2c087e6 1292static size_t disks_to_mpb_size(int disks)
cdddbdbc 1293{
c2c087e6 1294 size_t size;
cdddbdbc 1295
c2c087e6
DW
1296 size = sizeof(struct imsm_super);
1297 size += (disks - 1) * sizeof(struct imsm_disk);
1298 size += 2 * sizeof(struct imsm_dev);
1299 /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */
1300 size += (4 - 2) * sizeof(struct imsm_map);
1301 /* 4 possible disk_ord_tbl's */
1302 size += 4 * (disks - 1) * sizeof(__u32);
1303
1304 return size;
1305}
1306
1307static __u64 avail_size_imsm(struct supertype *st, __u64 devsize)
1308{
1309 if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS))
1310 return 0;
1311
1312 return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS);
cdddbdbc
DW
1313}
1314
ba2de7ba
DW
1315static void free_devlist(struct intel_super *super)
1316{
1317 struct intel_dev *dv;
1318
1319 while (super->devlist) {
1320 dv = super->devlist->next;
1321 free(super->devlist->dev);
1322 free(super->devlist);
1323 super->devlist = dv;
1324 }
1325}
1326
1327static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src)
1328{
1329 memcpy(dest, src, sizeof_imsm_dev(src, 0));
1330}
1331
cdddbdbc
DW
1332static int compare_super_imsm(struct supertype *st, struct supertype *tst)
1333{
1334 /*
1335 * return:
1336 * 0 same, or first was empty, and second was copied
1337 * 1 second had wrong number
1338 * 2 wrong uuid
1339 * 3 wrong other info
1340 */
1341 struct intel_super *first = st->sb;
1342 struct intel_super *sec = tst->sb;
1343
1344 if (!first) {
1345 st->sb = tst->sb;
1346 tst->sb = NULL;
1347 return 0;
1348 }
1349
949c47a0 1350 if (memcmp(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH) != 0)
cdddbdbc 1351 return 3;
d23fe947
DW
1352
1353 /* if an anchor does not have num_raid_devs set then it is a free
1354 * floating spare
1355 */
1356 if (first->anchor->num_raid_devs > 0 &&
1357 sec->anchor->num_raid_devs > 0) {
1358 if (first->anchor->family_num != sec->anchor->family_num)
1359 return 3;
d23fe947 1360 }
cdddbdbc 1361
3e372e5a
DW
1362 /* if 'first' is a spare promote it to a populated mpb with sec's
1363 * family number
1364 */
1365 if (first->anchor->num_raid_devs == 0 &&
1366 sec->anchor->num_raid_devs > 0) {
78d30f94 1367 int i;
ba2de7ba
DW
1368 struct intel_dev *dv;
1369 struct imsm_dev *dev;
78d30f94
DW
1370
1371 /* we need to copy raid device info from sec if an allocation
1372 * fails here we don't associate the spare
1373 */
1374 for (i = 0; i < sec->anchor->num_raid_devs; i++) {
ba2de7ba
DW
1375 dv = malloc(sizeof(*dv));
1376 if (!dv)
1377 break;
1378 dev = malloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1));
1379 if (!dev) {
1380 free(dv);
1381 break;
78d30f94 1382 }
ba2de7ba
DW
1383 dv->dev = dev;
1384 dv->index = i;
1385 dv->next = first->devlist;
1386 first->devlist = dv;
78d30f94 1387 }
ba2de7ba
DW
1388 if (i <= sec->anchor->num_raid_devs) {
1389 /* allocation failure */
1390 free_devlist(first);
1391 fprintf(stderr, "imsm: failed to associate spare\n");
1392 return 3;
78d30f94 1393 }
ba2de7ba
DW
1394 for (i = 0; i < sec->anchor->num_raid_devs; i++)
1395 imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i));
78d30f94 1396
3e372e5a
DW
1397 first->anchor->num_raid_devs = sec->anchor->num_raid_devs;
1398 first->anchor->family_num = sec->anchor->family_num;
1399 }
1400
cdddbdbc
DW
1401 return 0;
1402}
1403
0030e8d6
DW
1404static void fd2devname(int fd, char *name)
1405{
1406 struct stat st;
1407 char path[256];
1408 char dname[100];
1409 char *nm;
1410 int rv;
1411
1412 name[0] = '\0';
1413 if (fstat(fd, &st) != 0)
1414 return;
1415 sprintf(path, "/sys/dev/block/%d:%d",
1416 major(st.st_rdev), minor(st.st_rdev));
1417
1418 rv = readlink(path, dname, sizeof(dname));
1419 if (rv <= 0)
1420 return;
1421
1422 dname[rv] = '\0';
1423 nm = strrchr(dname, '/');
1424 nm++;
1425 snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm);
1426}
1427
1428
cdddbdbc
DW
1429extern int scsi_get_serial(int fd, void *buf, size_t buf_len);
1430
1431static int imsm_read_serial(int fd, char *devname,
1432 __u8 serial[MAX_RAID_SERIAL_LEN])
1433{
1434 unsigned char scsi_serial[255];
cdddbdbc
DW
1435 int rv;
1436 int rsp_len;
1f24f035
DW
1437 int len;
1438 char *c, *rsp_buf;
cdddbdbc
DW
1439
1440 memset(scsi_serial, 0, sizeof(scsi_serial));
cdddbdbc 1441
f9ba0ff1
DW
1442 rv = scsi_get_serial(fd, scsi_serial, sizeof(scsi_serial));
1443
40ebbb9c 1444 if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) {
f9ba0ff1
DW
1445 memset(serial, 0, MAX_RAID_SERIAL_LEN);
1446 fd2devname(fd, (char *) serial);
0030e8d6
DW
1447 return 0;
1448 }
1449
cdddbdbc
DW
1450 if (rv != 0) {
1451 if (devname)
1452 fprintf(stderr,
1453 Name ": Failed to retrieve serial for %s\n",
1454 devname);
1455 return rv;
1456 }
1457
5c3db629 1458 /* trim leading whitespace */
cdddbdbc 1459 rsp_len = scsi_serial[3];
03cd4cc8
DW
1460 if (!rsp_len) {
1461 if (devname)
1462 fprintf(stderr,
1463 Name ": Failed to retrieve serial for %s\n",
1464 devname);
1465 return 2;
1466 }
1f24f035
DW
1467 rsp_buf = (char *) &scsi_serial[4];
1468 c = rsp_buf;
1469 while (isspace(*c))
1470 c++;
5c3db629
DW
1471
1472 /* truncate len to the end of rsp_buf if necessary */
1f24f035
DW
1473 if (c + MAX_RAID_SERIAL_LEN > rsp_buf + rsp_len)
1474 len = rsp_len - (c - rsp_buf);
1475 else
1476 len = MAX_RAID_SERIAL_LEN;
5c3db629
DW
1477
1478 /* initialize the buffer and copy rsp_buf characters */
1479 memset(serial, 0, MAX_RAID_SERIAL_LEN);
1f24f035 1480 memcpy(serial, c, len);
5c3db629
DW
1481
1482 /* trim trailing whitespace starting with the last character copied */
1f24f035
DW
1483 c = (char *) &serial[len - 1];
1484 while (isspace(*c) || *c == '\0')
1485 *c-- = '\0';
cdddbdbc
DW
1486
1487 return 0;
1488}
1489
1f24f035
DW
1490static int serialcmp(__u8 *s1, __u8 *s2)
1491{
1492 return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN);
1493}
1494
1495static void serialcpy(__u8 *dest, __u8 *src)
1496{
1497 strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN);
1498}
1499
54c2c1ea
DW
1500static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super)
1501{
1502 struct dl *dl;
1503
1504 for (dl = super->disks; dl; dl = dl->next)
1505 if (serialcmp(dl->serial, serial) == 0)
1506 break;
1507
1508 return dl;
1509}
1510
cdddbdbc
DW
1511static int
1512load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd)
1513{
cdddbdbc
DW
1514 struct dl *dl;
1515 struct stat stb;
cdddbdbc
DW
1516 int rv;
1517 int i;
d23fe947
DW
1518 int alloc = 1;
1519 __u8 serial[MAX_RAID_SERIAL_LEN];
1520
1521 rv = imsm_read_serial(fd, devname, serial);
1522
1523 if (rv != 0)
1524 return 2;
1525
1526 /* check if this is a disk we have seen before. it may be a spare in
1527 * super->disks while the current anchor believes it is a raid member,
1528 * check if we need to update dl->index
1529 */
54c2c1ea 1530 dl = serial_to_dl(serial, super);
d23fe947
DW
1531 if (!dl)
1532 dl = malloc(sizeof(*dl));
1533 else
1534 alloc = 0;
cdddbdbc 1535
b9f594fe 1536 if (!dl) {
cdddbdbc
DW
1537 if (devname)
1538 fprintf(stderr,
1539 Name ": failed to allocate disk buffer for %s\n",
1540 devname);
1541 return 2;
1542 }
cdddbdbc 1543
d23fe947
DW
1544 if (alloc) {
1545 fstat(fd, &stb);
1546 dl->major = major(stb.st_rdev);
1547 dl->minor = minor(stb.st_rdev);
1548 dl->next = super->disks;
1549 dl->fd = keep_fd ? fd : -1;
1550 dl->devname = devname ? strdup(devname) : NULL;
1f24f035 1551 serialcpy(dl->serial, serial);
8796fdc4 1552 dl->index = -2;
0dcecb2e 1553 dl->e = NULL;
d23fe947
DW
1554 } else if (keep_fd) {
1555 close(dl->fd);
1556 dl->fd = fd;
1557 }
cdddbdbc 1558
d23fe947 1559 /* look up this disk's index in the current anchor */
949c47a0
DW
1560 for (i = 0; i < super->anchor->num_disks; i++) {
1561 struct imsm_disk *disk_iter;
1562
1563 disk_iter = __get_imsm_disk(super->anchor, i);
cdddbdbc 1564
1f24f035 1565 if (serialcmp(disk_iter->serial, dl->serial) == 0) {
b9f594fe 1566 dl->disk = *disk_iter;
d23fe947
DW
1567 /* only set index on disks that are a member of a
1568 * populated contianer, i.e. one with raid_devs
1569 */
f2f27e63 1570 if (dl->disk.status & FAILED_DISK)
6c386dd3 1571 dl->index = -2;
f2f27e63 1572 else if (dl->disk.status & SPARE_DISK)
d23fe947
DW
1573 dl->index = -1;
1574 else
1575 dl->index = i;
8796fdc4 1576
cdddbdbc 1577 break;
949c47a0 1578 }
cdddbdbc
DW
1579 }
1580
3f6efecc
DW
1581 /* no match, maybe a stale failed drive */
1582 if (i == super->anchor->num_disks && dl->index >= 0) {
1583 dl->disk = *__get_imsm_disk(super->anchor, dl->index);
f2f27e63 1584 if (dl->disk.status & FAILED_DISK)
3f6efecc
DW
1585 dl->index = -2;
1586 }
1587
d23fe947
DW
1588 if (alloc)
1589 super->disks = dl;
6c386dd3 1590
949c47a0
DW
1591 return 0;
1592}
1593
0e600426 1594#ifndef MDASSEMBLE
0c046afd
DW
1595/* When migrating map0 contains the 'destination' state while map1
1596 * contains the current state. When not migrating map0 contains the
1597 * current state. This routine assumes that map[0].map_state is set to
1598 * the current array state before being called.
1599 *
1600 * Migration is indicated by one of the following states
1601 * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed)
e3bba0e0 1602 * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal
0c046afd 1603 * map1state=unitialized)
e3bba0e0 1604 * 3/ Verify (Resync) (migr_state=1 migr_type=MIGR_REBUILD map0state=normal
0c046afd 1605 * map1state=normal)
e3bba0e0 1606 * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal
0c046afd
DW
1607 * map1state=degraded)
1608 */
0556e1a2 1609static void migrate(struct imsm_dev *dev, __u8 to_state, int migr_type)
3393c6af 1610{
0c046afd 1611 struct imsm_map *dest;
3393c6af
DW
1612 struct imsm_map *src = get_imsm_map(dev, 0);
1613
0c046afd 1614 dev->vol.migr_state = 1;
0556e1a2 1615 dev->vol.migr_type = migr_type;
f8f603f1 1616 dev->vol.curr_migr_unit = 0;
0c046afd
DW
1617 dest = get_imsm_map(dev, 1);
1618
0556e1a2 1619 /* duplicate and then set the target end state in map[0] */
3393c6af 1620 memcpy(dest, src, sizeof_imsm_map(src));
0556e1a2
DW
1621 if (migr_type == MIGR_REBUILD) {
1622 __u32 ord;
1623 int i;
1624
1625 for (i = 0; i < src->num_members; i++) {
1626 ord = __le32_to_cpu(src->disk_ord_tbl[i]);
1627 set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord));
1628 }
1629 }
1630
0c046afd 1631 src->map_state = to_state;
949c47a0 1632}
f8f603f1
DW
1633
1634static void end_migration(struct imsm_dev *dev, __u8 map_state)
1635{
1636 struct imsm_map *map = get_imsm_map(dev, 0);
0556e1a2
DW
1637 struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state);
1638 int i;
1639
1640 /* merge any IMSM_ORD_REBUILD bits that were not successfully
1641 * completed in the last migration.
1642 *
1643 * FIXME add support for online capacity expansion and
1644 * raid-level-migration
1645 */
1646 for (i = 0; i < prev->num_members; i++)
1647 map->disk_ord_tbl[i] |= prev->disk_ord_tbl[i];
f8f603f1
DW
1648
1649 dev->vol.migr_state = 0;
1650 dev->vol.curr_migr_unit = 0;
1651 map->map_state = map_state;
1652}
0e600426 1653#endif
949c47a0
DW
1654
1655static int parse_raid_devices(struct intel_super *super)
1656{
1657 int i;
1658 struct imsm_dev *dev_new;
4d7b1503
DW
1659 size_t len, len_migr;
1660 size_t space_needed = 0;
1661 struct imsm_super *mpb = super->anchor;
949c47a0
DW
1662
1663 for (i = 0; i < super->anchor->num_raid_devs; i++) {
1664 struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i);
ba2de7ba 1665 struct intel_dev *dv;
949c47a0 1666
4d7b1503
DW
1667 len = sizeof_imsm_dev(dev_iter, 0);
1668 len_migr = sizeof_imsm_dev(dev_iter, 1);
1669 if (len_migr > len)
1670 space_needed += len_migr - len;
1671
ba2de7ba
DW
1672 dv = malloc(sizeof(*dv));
1673 if (!dv)
1674 return 1;
4d7b1503 1675 dev_new = malloc(len_migr);
ba2de7ba
DW
1676 if (!dev_new) {
1677 free(dv);
949c47a0 1678 return 1;
ba2de7ba 1679 }
949c47a0 1680 imsm_copy_dev(dev_new, dev_iter);
ba2de7ba
DW
1681 dv->dev = dev_new;
1682 dv->index = i;
1683 dv->next = super->devlist;
1684 super->devlist = dv;
949c47a0 1685 }
cdddbdbc 1686
4d7b1503
DW
1687 /* ensure that super->buf is large enough when all raid devices
1688 * are migrating
1689 */
1690 if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) {
1691 void *buf;
1692
1693 len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed, 512);
1694 if (posix_memalign(&buf, 512, len) != 0)
1695 return 1;
1696
1697 memcpy(buf, super->buf, len);
1698 free(super->buf);
1699 super->buf = buf;
1700 super->len = len;
1701 }
1702
cdddbdbc
DW
1703 return 0;
1704}
1705
604b746f
JD
1706/* retrieve a pointer to the bbm log which starts after all raid devices */
1707struct bbm_log *__get_imsm_bbm_log(struct imsm_super *mpb)
1708{
1709 void *ptr = NULL;
1710
1711 if (__le32_to_cpu(mpb->bbm_log_size)) {
1712 ptr = mpb;
1713 ptr += mpb->mpb_size - __le32_to_cpu(mpb->bbm_log_size);
1714 }
1715
1716 return ptr;
1717}
1718
d23fe947 1719static void __free_imsm(struct intel_super *super, int free_disks);
9ca2c81c 1720
cdddbdbc
DW
1721/* load_imsm_mpb - read matrix metadata
1722 * allocates super->mpb to be freed by free_super
1723 */
1724static int load_imsm_mpb(int fd, struct intel_super *super, char *devname)
1725{
1726 unsigned long long dsize;
cdddbdbc
DW
1727 unsigned long long sectors;
1728 struct stat;
6416d527 1729 struct imsm_super *anchor;
cdddbdbc 1730 __u32 check_sum;
949c47a0 1731 int rc;
cdddbdbc 1732
cdddbdbc
DW
1733 get_dev_size(fd, NULL, &dsize);
1734
1735 if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) {
1736 if (devname)
1737 fprintf(stderr,
1738 Name ": Cannot seek to anchor block on %s: %s\n",
1739 devname, strerror(errno));
1740 return 1;
1741 }
1742
949c47a0 1743 if (posix_memalign((void**)&anchor, 512, 512) != 0) {
ad97895e
DW
1744 if (devname)
1745 fprintf(stderr,
1746 Name ": Failed to allocate imsm anchor buffer"
1747 " on %s\n", devname);
1748 return 1;
1749 }
949c47a0 1750 if (read(fd, anchor, 512) != 512) {
cdddbdbc
DW
1751 if (devname)
1752 fprintf(stderr,
1753 Name ": Cannot read anchor block on %s: %s\n",
1754 devname, strerror(errno));
6416d527 1755 free(anchor);
cdddbdbc
DW
1756 return 1;
1757 }
1758
6416d527 1759 if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) {
cdddbdbc
DW
1760 if (devname)
1761 fprintf(stderr,
1762 Name ": no IMSM anchor on %s\n", devname);
6416d527 1763 free(anchor);
cdddbdbc
DW
1764 return 2;
1765 }
1766
d23fe947 1767 __free_imsm(super, 0);
949c47a0
DW
1768 super->len = ROUND_UP(anchor->mpb_size, 512);
1769 if (posix_memalign(&super->buf, 512, super->len) != 0) {
cdddbdbc
DW
1770 if (devname)
1771 fprintf(stderr,
1772 Name ": unable to allocate %zu byte mpb buffer\n",
949c47a0 1773 super->len);
6416d527 1774 free(anchor);
cdddbdbc
DW
1775 return 2;
1776 }
949c47a0 1777 memcpy(super->buf, anchor, 512);
cdddbdbc 1778
6416d527
NB
1779 sectors = mpb_sectors(anchor) - 1;
1780 free(anchor);
949c47a0 1781 if (!sectors) {
ecf45690
DW
1782 check_sum = __gen_imsm_checksum(super->anchor);
1783 if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
1784 if (devname)
1785 fprintf(stderr,
1786 Name ": IMSM checksum %x != %x on %s\n",
1787 check_sum,
1788 __le32_to_cpu(super->anchor->check_sum),
1789 devname);
1790 return 2;
1791 }
1792
949c47a0
DW
1793 rc = load_imsm_disk(fd, super, devname, 0);
1794 if (rc == 0)
1795 rc = parse_raid_devices(super);
1796 return rc;
1797 }
cdddbdbc
DW
1798
1799 /* read the extended mpb */
1800 if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) {
1801 if (devname)
1802 fprintf(stderr,
1803 Name ": Cannot seek to extended mpb on %s: %s\n",
1804 devname, strerror(errno));
1805 return 1;
1806 }
1807
949c47a0 1808 if (read(fd, super->buf + 512, super->len - 512) != super->len - 512) {
cdddbdbc
DW
1809 if (devname)
1810 fprintf(stderr,
1811 Name ": Cannot read extended mpb on %s: %s\n",
1812 devname, strerror(errno));
1813 return 2;
1814 }
1815
949c47a0
DW
1816 check_sum = __gen_imsm_checksum(super->anchor);
1817 if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
cdddbdbc
DW
1818 if (devname)
1819 fprintf(stderr,
1820 Name ": IMSM checksum %x != %x on %s\n",
949c47a0 1821 check_sum, __le32_to_cpu(super->anchor->check_sum),
cdddbdbc 1822 devname);
db575f3b 1823 return 3;
cdddbdbc
DW
1824 }
1825
604b746f
JD
1826 /* FIXME the BBM log is disk specific so we cannot use this global
1827 * buffer for all disks. Ok for now since we only look at the global
1828 * bbm_log_size parameter to gate assembly
1829 */
1830 super->bbm_log = __get_imsm_bbm_log(super->anchor);
1831
949c47a0
DW
1832 rc = load_imsm_disk(fd, super, devname, 0);
1833 if (rc == 0)
1834 rc = parse_raid_devices(super);
4d7b1503 1835
949c47a0 1836 return rc;
cdddbdbc
DW
1837}
1838
ae6aad82
DW
1839static void __free_imsm_disk(struct dl *d)
1840{
1841 if (d->fd >= 0)
1842 close(d->fd);
1843 if (d->devname)
1844 free(d->devname);
0dcecb2e
DW
1845 if (d->e)
1846 free(d->e);
ae6aad82
DW
1847 free(d);
1848
1849}
cdddbdbc
DW
1850static void free_imsm_disks(struct intel_super *super)
1851{
47ee5a45 1852 struct dl *d;
cdddbdbc 1853
47ee5a45
DW
1854 while (super->disks) {
1855 d = super->disks;
cdddbdbc 1856 super->disks = d->next;
ae6aad82 1857 __free_imsm_disk(d);
cdddbdbc 1858 }
47ee5a45
DW
1859 while (super->missing) {
1860 d = super->missing;
1861 super->missing = d->next;
1862 __free_imsm_disk(d);
1863 }
1864
cdddbdbc
DW
1865}
1866
9ca2c81c 1867/* free all the pieces hanging off of a super pointer */
d23fe947 1868static void __free_imsm(struct intel_super *super, int free_disks)
cdddbdbc 1869{
9ca2c81c 1870 if (super->buf) {
949c47a0 1871 free(super->buf);
9ca2c81c
DW
1872 super->buf = NULL;
1873 }
d23fe947
DW
1874 if (free_disks)
1875 free_imsm_disks(super);
ba2de7ba 1876 free_devlist(super);
88c32bb1
DW
1877 if (super->hba) {
1878 free((void *) super->hba);
1879 super->hba = NULL;
1880 }
cdddbdbc
DW
1881}
1882
9ca2c81c
DW
1883static void free_imsm(struct intel_super *super)
1884{
d23fe947 1885 __free_imsm(super, 1);
9ca2c81c
DW
1886 free(super);
1887}
cdddbdbc
DW
1888
1889static void free_super_imsm(struct supertype *st)
1890{
1891 struct intel_super *super = st->sb;
1892
1893 if (!super)
1894 return;
1895
1896 free_imsm(super);
1897 st->sb = NULL;
1898}
1899
c2c087e6
DW
1900static struct intel_super *alloc_super(int creating_imsm)
1901{
1902 struct intel_super *super = malloc(sizeof(*super));
1903
1904 if (super) {
1905 memset(super, 0, sizeof(*super));
1906 super->creating_imsm = creating_imsm;
bf5a934a 1907 super->current_vol = -1;
0dcecb2e 1908 super->create_offset = ~((__u32 ) 0);
88c32bb1
DW
1909 if (!check_env("IMSM_NO_PLATFORM"))
1910 super->orom = find_imsm_orom();
cceebc67 1911 if (super->orom && !check_env("IMSM_TEST_OROM")) {
88c32bb1
DW
1912 struct sys_dev *list, *ent;
1913
1914 /* find the first intel ahci controller */
1915 list = find_driver_devices("pci", "ahci");
1916 for (ent = list; ent; ent = ent->next)
1917 if (devpath_to_vendor(ent->path) == 0x8086)
1918 break;
1919 if (ent) {
1920 super->hba = ent->path;
1921 ent->path = NULL;
1922 }
1923 free_sys_dev(&list);
1924 }
c2c087e6
DW
1925 }
1926
1927 return super;
1928}
1929
cdddbdbc 1930#ifndef MDASSEMBLE
47ee5a45
DW
1931/* find_missing - helper routine for load_super_imsm_all that identifies
1932 * disks that have disappeared from the system. This routine relies on
1933 * the mpb being uptodate, which it is at load time.
1934 */
1935static int find_missing(struct intel_super *super)
1936{
1937 int i;
1938 struct imsm_super *mpb = super->anchor;
1939 struct dl *dl;
1940 struct imsm_disk *disk;
47ee5a45
DW
1941
1942 for (i = 0; i < mpb->num_disks; i++) {
1943 disk = __get_imsm_disk(mpb, i);
54c2c1ea 1944 dl = serial_to_dl(disk->serial, super);
47ee5a45
DW
1945 if (dl)
1946 continue;
47ee5a45
DW
1947
1948 dl = malloc(sizeof(*dl));
1949 if (!dl)
1950 return 1;
1951 dl->major = 0;
1952 dl->minor = 0;
1953 dl->fd = -1;
1954 dl->devname = strdup("missing");
1955 dl->index = i;
1956 serialcpy(dl->serial, disk->serial);
1957 dl->disk = *disk;
689c9bf3 1958 dl->e = NULL;
47ee5a45
DW
1959 dl->next = super->missing;
1960 super->missing = dl;
1961 }
1962
1963 return 0;
1964}
1965
cdddbdbc
DW
1966static int load_super_imsm_all(struct supertype *st, int fd, void **sbp,
1967 char *devname, int keep_fd)
1968{
1969 struct mdinfo *sra;
1970 struct intel_super *super;
1971 struct mdinfo *sd, *best = NULL;
1972 __u32 bestgen = 0;
1973 __u32 gen;
1974 char nm[20];
1975 int dfd;
1976 int rv;
db575f3b
DW
1977 int devnum = fd2devnum(fd);
1978 int retry;
dab4a513 1979 enum sysfs_read_flags flags;
cdddbdbc 1980
dab4a513
DW
1981 flags = GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE;
1982 if (mdmon_running(devnum))
1983 flags |= SKIP_GONE_DEVS;
1984
1985 /* check if 'fd' an opened container */
1986 sra = sysfs_read(fd, 0, flags);
cdddbdbc
DW
1987 if (!sra)
1988 return 1;
1989
1990 if (sra->array.major_version != -1 ||
1991 sra->array.minor_version != -2 ||
1992 strcmp(sra->text_version, "imsm") != 0)
1993 return 1;
1994
c2c087e6 1995 super = alloc_super(0);
cdddbdbc
DW
1996 if (!super)
1997 return 1;
1998
d23fe947 1999 /* find the most up to date disk in this array, skipping spares */
cdddbdbc
DW
2000 for (sd = sra->devs; sd; sd = sd->next) {
2001 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
2002 dfd = dev_open(nm, keep_fd ? O_RDWR : O_RDONLY);
be2c0e38 2003 if (dfd < 0) {
cdddbdbc
DW
2004 free_imsm(super);
2005 return 2;
2006 }
2007 rv = load_imsm_mpb(dfd, super, NULL);
db575f3b
DW
2008
2009 /* retry the load if we might have raced against mdmon */
2010 if (rv == 3 && mdmon_running(devnum))
2011 for (retry = 0; retry < 3; retry++) {
2012 usleep(3000);
2013 rv = load_imsm_mpb(dfd, super, NULL);
2014 if (rv != 3)
2015 break;
2016 }
cdddbdbc
DW
2017 if (!keep_fd)
2018 close(dfd);
2019 if (rv == 0) {
d23fe947
DW
2020 if (super->anchor->num_raid_devs == 0)
2021 gen = 0;
2022 else
2023 gen = __le32_to_cpu(super->anchor->generation_num);
cdddbdbc
DW
2024 if (!best || gen > bestgen) {
2025 bestgen = gen;
2026 best = sd;
2027 }
2028 } else {
2029 free_imsm(super);
db575f3b 2030 return rv;
cdddbdbc
DW
2031 }
2032 }
2033
2034 if (!best) {
2035 free_imsm(super);
2036 return 1;
2037 }
2038
2039 /* load the most up to date anchor */
2040 sprintf(nm, "%d:%d", best->disk.major, best->disk.minor);
2041 dfd = dev_open(nm, O_RDONLY);
be2c0e38 2042 if (dfd < 0) {
cdddbdbc
DW
2043 free_imsm(super);
2044 return 1;
2045 }
2046 rv = load_imsm_mpb(dfd, super, NULL);
2047 close(dfd);
2048 if (rv != 0) {
2049 free_imsm(super);
2050 return 2;
2051 }
2052
d23fe947 2053 /* re-parse the disk list with the current anchor */
cdddbdbc
DW
2054 for (sd = sra->devs ; sd ; sd = sd->next) {
2055 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
2056 dfd = dev_open(nm, keep_fd? O_RDWR : O_RDONLY);
be2c0e38 2057 if (dfd < 0) {
cdddbdbc
DW
2058 free_imsm(super);
2059 return 2;
2060 }
2061 load_imsm_disk(dfd, super, NULL, keep_fd);
2062 if (!keep_fd)
2063 close(dfd);
2064 }
2065
47ee5a45
DW
2066
2067 if (find_missing(super) != 0) {
2068 free_imsm(super);
2069 return 2;
2070 }
2071
f7e7067b 2072 if (st->subarray[0]) {
949c47a0 2073 if (atoi(st->subarray) <= super->anchor->num_raid_devs)
bf5a934a
DW
2074 super->current_vol = atoi(st->subarray);
2075 else
2076 return 1;
f7e7067b
NB
2077 }
2078
cdddbdbc 2079 *sbp = super;
db575f3b 2080 st->container_dev = devnum;
cdddbdbc 2081 if (st->ss == NULL) {
bf5a934a 2082 st->ss = &super_imsm;
cdddbdbc
DW
2083 st->minor_version = 0;
2084 st->max_devs = IMSM_MAX_DEVICES;
2085 }
352452c3 2086 st->loaded_container = 1;
cdddbdbc
DW
2087
2088 return 0;
2089}
2090#endif
2091
2092static int load_super_imsm(struct supertype *st, int fd, char *devname)
2093{
2094 struct intel_super *super;
2095 int rv;
2096
2097#ifndef MDASSEMBLE
3dbccbcf 2098 if (load_super_imsm_all(st, fd, &st->sb, devname, 1) == 0)
cdddbdbc
DW
2099 return 0;
2100#endif
f7e7067b
NB
2101 if (st->subarray[0])
2102 return 1; /* FIXME */
cdddbdbc 2103
c2c087e6 2104 super = alloc_super(0);
cdddbdbc
DW
2105 if (!super) {
2106 fprintf(stderr,
2107 Name ": malloc of %zu failed.\n",
2108 sizeof(*super));
2109 return 1;
2110 }
2111
2112 rv = load_imsm_mpb(fd, super, devname);
2113
2114 if (rv) {
2115 if (devname)
2116 fprintf(stderr,
2117 Name ": Failed to load all information "
2118 "sections on %s\n", devname);
2119 free_imsm(super);
2120 return rv;
2121 }
2122
2123 st->sb = super;
2124 if (st->ss == NULL) {
2125 st->ss = &super_imsm;
2126 st->minor_version = 0;
2127 st->max_devs = IMSM_MAX_DEVICES;
2128 }
352452c3 2129 st->loaded_container = 0;
cdddbdbc
DW
2130
2131 return 0;
2132}
2133
ef6ffade
DW
2134static __u16 info_to_blocks_per_strip(mdu_array_info_t *info)
2135{
2136 if (info->level == 1)
2137 return 128;
2138 return info->chunk_size >> 9;
2139}
2140
2141static __u32 info_to_num_data_stripes(mdu_array_info_t *info)
2142{
2143 __u32 num_stripes;
2144
2145 num_stripes = (info->size * 2) / info_to_blocks_per_strip(info);
2146 if (info->level == 1)
2147 num_stripes /= 2;
2148
2149 return num_stripes;
2150}
2151
fcfd9599
DW
2152static __u32 info_to_blocks_per_member(mdu_array_info_t *info)
2153{
4025c288
DW
2154 if (info->level == 1)
2155 return info->size * 2;
2156 else
2157 return (info->size * 2) & ~(info_to_blocks_per_strip(info) - 1);
fcfd9599
DW
2158}
2159
4d1313e9
DW
2160static void imsm_update_version_info(struct intel_super *super)
2161{
2162 /* update the version and attributes */
2163 struct imsm_super *mpb = super->anchor;
2164 char *version;
2165 struct imsm_dev *dev;
2166 struct imsm_map *map;
2167 int i;
2168
2169 for (i = 0; i < mpb->num_raid_devs; i++) {
2170 dev = get_imsm_dev(super, i);
2171 map = get_imsm_map(dev, 0);
2172 if (__le32_to_cpu(dev->size_high) > 0)
2173 mpb->attributes |= MPB_ATTRIB_2TB;
2174
2175 /* FIXME detect when an array spans a port multiplier */
2176 #if 0
2177 mpb->attributes |= MPB_ATTRIB_PM;
2178 #endif
2179
2180 if (mpb->num_raid_devs > 1 ||
2181 mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) {
2182 version = MPB_VERSION_ATTRIBS;
2183 switch (get_imsm_raid_level(map)) {
2184 case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break;
2185 case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break;
2186 case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break;
2187 case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break;
2188 }
2189 } else {
2190 if (map->num_members >= 5)
2191 version = MPB_VERSION_5OR6_DISK_ARRAY;
2192 else if (dev->status == DEV_CLONE_N_GO)
2193 version = MPB_VERSION_CNG;
2194 else if (get_imsm_raid_level(map) == 5)
2195 version = MPB_VERSION_RAID5;
2196 else if (map->num_members >= 3)
2197 version = MPB_VERSION_3OR4_DISK_ARRAY;
2198 else if (get_imsm_raid_level(map) == 1)
2199 version = MPB_VERSION_RAID1;
2200 else
2201 version = MPB_VERSION_RAID0;
2202 }
2203 strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version);
2204 }
2205}
2206
8b353278
DW
2207static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info,
2208 unsigned long long size, char *name,
2209 char *homehost, int *uuid)
cdddbdbc 2210{
c2c087e6
DW
2211 /* We are creating a volume inside a pre-existing container.
2212 * so st->sb is already set.
2213 */
2214 struct intel_super *super = st->sb;
949c47a0 2215 struct imsm_super *mpb = super->anchor;
ba2de7ba 2216 struct intel_dev *dv;
c2c087e6
DW
2217 struct imsm_dev *dev;
2218 struct imsm_vol *vol;
2219 struct imsm_map *map;
2220 int idx = mpb->num_raid_devs;
2221 int i;
2222 unsigned long long array_blocks;
2c092cad 2223 size_t size_old, size_new;
cdddbdbc 2224
88c32bb1 2225 if (super->orom && mpb->num_raid_devs >= super->orom->vpa) {
c2c087e6 2226 fprintf(stderr, Name": This imsm-container already has the "
88c32bb1 2227 "maximum of %d volumes\n", super->orom->vpa);
c2c087e6
DW
2228 return 0;
2229 }
2230
2c092cad
DW
2231 /* ensure the mpb is large enough for the new data */
2232 size_old = __le32_to_cpu(mpb->mpb_size);
2233 size_new = disks_to_mpb_size(info->nr_disks);
2234 if (size_new > size_old) {
2235 void *mpb_new;
2236 size_t size_round = ROUND_UP(size_new, 512);
2237
2238 if (posix_memalign(&mpb_new, 512, size_round) != 0) {
2239 fprintf(stderr, Name": could not allocate new mpb\n");
2240 return 0;
2241 }
2242 memcpy(mpb_new, mpb, size_old);
2243 free(mpb);
2244 mpb = mpb_new;
949c47a0 2245 super->anchor = mpb_new;
2c092cad
DW
2246 mpb->mpb_size = __cpu_to_le32(size_new);
2247 memset(mpb_new + size_old, 0, size_round - size_old);
2248 }
bf5a934a 2249 super->current_vol = idx;
d23fe947
DW
2250 /* when creating the first raid device in this container set num_disks
2251 * to zero, i.e. delete this spare and add raid member devices in
2252 * add_to_super_imsm_volume()
2253 */
2254 if (super->current_vol == 0)
2255 mpb->num_disks = 0;
5a038140
DW
2256
2257 for (i = 0; i < super->current_vol; i++) {
2258 dev = get_imsm_dev(super, i);
2259 if (strncmp((char *) dev->volume, name,
2260 MAX_RAID_SERIAL_LEN) == 0) {
2261 fprintf(stderr, Name": '%s' is already defined for this container\n",
2262 name);
2263 return 0;
2264 }
2265 }
2266
bf5a934a 2267 sprintf(st->subarray, "%d", idx);
ba2de7ba
DW
2268 dv = malloc(sizeof(*dv));
2269 if (!dv) {
2270 fprintf(stderr, Name ": failed to allocate device list entry\n");
2271 return 0;
2272 }
949c47a0
DW
2273 dev = malloc(sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1));
2274 if (!dev) {
ba2de7ba 2275 free(dv);
949c47a0
DW
2276 fprintf(stderr, Name": could not allocate raid device\n");
2277 return 0;
2278 }
c2c087e6 2279 strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN);
03bcbc65
DW
2280 if (info->level == 1)
2281 array_blocks = info_to_blocks_per_member(info);
2282 else
2283 array_blocks = calc_array_size(info->level, info->raid_disks,
2284 info->layout, info->chunk_size,
2285 info->size*2);
c2c087e6
DW
2286 dev->size_low = __cpu_to_le32((__u32) array_blocks);
2287 dev->size_high = __cpu_to_le32((__u32) (array_blocks >> 32));
2288 dev->status = __cpu_to_le32(0);
2289 dev->reserved_blocks = __cpu_to_le32(0);
2290 vol = &dev->vol;
2291 vol->migr_state = 0;
e3bba0e0 2292 vol->migr_type = MIGR_INIT;
c2c087e6 2293 vol->dirty = 0;
f8f603f1 2294 vol->curr_migr_unit = 0;
a965f303 2295 map = get_imsm_map(dev, 0);
0dcecb2e 2296 map->pba_of_lba0 = __cpu_to_le32(super->create_offset);
fcfd9599 2297 map->blocks_per_member = __cpu_to_le32(info_to_blocks_per_member(info));
ef6ffade
DW
2298 map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info));
2299 map->num_data_stripes = __cpu_to_le32(info_to_num_data_stripes(info));
0556e1a2 2300 map->failed_disk_num = ~0;
c2c087e6
DW
2301 map->map_state = info->level ? IMSM_T_STATE_UNINITIALIZED :
2302 IMSM_T_STATE_NORMAL;
ef6ffade
DW
2303
2304 if (info->level == 1 && info->raid_disks > 2) {
2305 fprintf(stderr, Name": imsm does not support more than 2 disks"
2306 "in a raid1 volume\n");
2307 return 0;
2308 }
4d1313e9 2309 if (info->level == 10) {
c2c087e6 2310 map->raid_level = 1;
4d1313e9
DW
2311 map->num_domains = info->raid_disks / 2;
2312 } else {
c2c087e6 2313 map->raid_level = info->level;
4d1313e9
DW
2314 map->num_domains = !!map->raid_level;
2315 }
ef6ffade 2316
c2c087e6
DW
2317 map->num_members = info->raid_disks;
2318 for (i = 0; i < map->num_members; i++) {
2319 /* initialized in add_to_super */
be73972f 2320 set_imsm_ord_tbl_ent(map, i, 0);
c2c087e6 2321 }
949c47a0 2322 mpb->num_raid_devs++;
ba2de7ba
DW
2323
2324 dv->dev = dev;
2325 dv->index = super->current_vol;
2326 dv->next = super->devlist;
2327 super->devlist = dv;
c2c087e6 2328
4d1313e9
DW
2329 imsm_update_version_info(super);
2330
c2c087e6 2331 return 1;
cdddbdbc
DW
2332}
2333
bf5a934a
DW
2334static int init_super_imsm(struct supertype *st, mdu_array_info_t *info,
2335 unsigned long long size, char *name,
2336 char *homehost, int *uuid)
2337{
2338 /* This is primarily called by Create when creating a new array.
2339 * We will then get add_to_super called for each component, and then
2340 * write_init_super called to write it out to each device.
2341 * For IMSM, Create can create on fresh devices or on a pre-existing
2342 * array.
2343 * To create on a pre-existing array a different method will be called.
2344 * This one is just for fresh drives.
2345 */
2346 struct intel_super *super;
2347 struct imsm_super *mpb;
2348 size_t mpb_size;
4d1313e9 2349 char *version;
bf5a934a
DW
2350
2351 if (!info) {
2352 st->sb = NULL;
2353 return 0;
2354 }
2355 if (st->sb)
2356 return init_super_imsm_volume(st, info, size, name, homehost,
2357 uuid);
2358
2359 super = alloc_super(1);
2360 if (!super)
2361 return 0;
2362 mpb_size = disks_to_mpb_size(info->nr_disks);
ef649044 2363 if (posix_memalign(&super->buf, 512, mpb_size) != 0) {
bf5a934a
DW
2364 free(super);
2365 return 0;
2366 }
ef649044 2367 mpb = super->buf;
bf5a934a
DW
2368 memset(mpb, 0, mpb_size);
2369
4d1313e9
DW
2370 mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
2371
2372 version = (char *) mpb->sig;
2373 strcpy(version, MPB_SIGNATURE);
2374 version += strlen(MPB_SIGNATURE);
2375 strcpy(version, MPB_VERSION_RAID0);
bf5a934a
DW
2376 mpb->mpb_size = mpb_size;
2377
bf5a934a
DW
2378 st->sb = super;
2379 return 1;
2380}
2381
0e600426 2382#ifndef MDASSEMBLE
f20c3968 2383static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk,
bf5a934a
DW
2384 int fd, char *devname)
2385{
2386 struct intel_super *super = st->sb;
d23fe947 2387 struct imsm_super *mpb = super->anchor;
bf5a934a
DW
2388 struct dl *dl;
2389 struct imsm_dev *dev;
2390 struct imsm_map *map;
bf5a934a 2391
949c47a0 2392 dev = get_imsm_dev(super, super->current_vol);
a965f303 2393 map = get_imsm_map(dev, 0);
bf5a934a 2394
208933a7
N
2395 if (! (dk->state & (1<<MD_DISK_SYNC))) {
2396 fprintf(stderr, Name ": %s: Cannot add spare devices to IMSM volume\n",
2397 devname);
2398 return 1;
2399 }
2400
bf5a934a
DW
2401 for (dl = super->disks; dl ; dl = dl->next)
2402 if (dl->major == dk->major &&
2403 dl->minor == dk->minor)
2404 break;
d23fe947 2405
208933a7
N
2406 if (!dl) {
2407 fprintf(stderr, Name ": %s is not a member of the same container\n", devname);
f20c3968 2408 return 1;
208933a7 2409 }
bf5a934a 2410
d23fe947
DW
2411 /* add a pristine spare to the metadata */
2412 if (dl->index < 0) {
2413 dl->index = super->anchor->num_disks;
2414 super->anchor->num_disks++;
2415 }
be73972f 2416 set_imsm_ord_tbl_ent(map, dk->number, dl->index);
f2f27e63 2417 dl->disk.status = CONFIGURED_DISK | USABLE_DISK;
d23fe947
DW
2418
2419 /* if we are creating the first raid device update the family number */
2420 if (super->current_vol == 0) {
2421 __u32 sum;
2422 struct imsm_dev *_dev = __get_imsm_dev(mpb, 0);
2423 struct imsm_disk *_disk = __get_imsm_disk(mpb, dl->index);
2424
2425 *_dev = *dev;
2426 *_disk = dl->disk;
2427 sum = __gen_imsm_checksum(mpb);
2428 mpb->family_num = __cpu_to_le32(sum);
2429 }
f20c3968
DW
2430
2431 return 0;
bf5a934a
DW
2432}
2433
f20c3968 2434static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk,
cdddbdbc
DW
2435 int fd, char *devname)
2436{
c2c087e6 2437 struct intel_super *super = st->sb;
c2c087e6
DW
2438 struct dl *dd;
2439 unsigned long long size;
f2f27e63 2440 __u32 id;
c2c087e6
DW
2441 int rv;
2442 struct stat stb;
2443
88c32bb1
DW
2444 /* if we are on an RAID enabled platform check that the disk is
2445 * attached to the raid controller
2446 */
2447 if (super->hba && !disk_attached_to_hba(fd, super->hba)) {
2448 fprintf(stderr,
2449 Name ": %s is not attached to the raid controller: %s\n",
2450 devname ? : "disk", super->hba);
2451 return 1;
2452 }
2453
f20c3968
DW
2454 if (super->current_vol >= 0)
2455 return add_to_super_imsm_volume(st, dk, fd, devname);
bf5a934a 2456
c2c087e6
DW
2457 fstat(fd, &stb);
2458 dd = malloc(sizeof(*dd));
b9f594fe 2459 if (!dd) {
c2c087e6
DW
2460 fprintf(stderr,
2461 Name ": malloc failed %s:%d.\n", __func__, __LINE__);
f20c3968 2462 return 1;
c2c087e6
DW
2463 }
2464 memset(dd, 0, sizeof(*dd));
2465 dd->major = major(stb.st_rdev);
2466 dd->minor = minor(stb.st_rdev);
b9f594fe 2467 dd->index = -1;
c2c087e6 2468 dd->devname = devname ? strdup(devname) : NULL;
c2c087e6 2469 dd->fd = fd;
689c9bf3 2470 dd->e = NULL;
c2c087e6
DW
2471 rv = imsm_read_serial(fd, devname, dd->serial);
2472 if (rv) {
2473 fprintf(stderr,
0030e8d6 2474 Name ": failed to retrieve scsi serial, aborting\n");
949c47a0 2475 free(dd);
0030e8d6 2476 abort();
c2c087e6
DW
2477 }
2478
c2c087e6
DW
2479 get_dev_size(fd, NULL, &size);
2480 size /= 512;
1f24f035 2481 serialcpy(dd->disk.serial, dd->serial);
b9f594fe 2482 dd->disk.total_blocks = __cpu_to_le32(size);
f2f27e63 2483 dd->disk.status = USABLE_DISK | SPARE_DISK;
c2c087e6 2484 if (sysfs_disk_to_scsi_id(fd, &id) == 0)
b9f594fe 2485 dd->disk.scsi_id = __cpu_to_le32(id);
c2c087e6 2486 else
b9f594fe 2487 dd->disk.scsi_id = __cpu_to_le32(0);
43dad3d6
DW
2488
2489 if (st->update_tail) {
2490 dd->next = super->add;
2491 super->add = dd;
2492 } else {
2493 dd->next = super->disks;
2494 super->disks = dd;
2495 }
f20c3968
DW
2496
2497 return 0;
cdddbdbc
DW
2498}
2499
c2c087e6
DW
2500static int store_imsm_mpb(int fd, struct intel_super *super);
2501
d23fe947
DW
2502/* spare records have their own family number and do not have any defined raid
2503 * devices
2504 */
2505static int write_super_imsm_spares(struct intel_super *super, int doclose)
2506{
2507 struct imsm_super mpb_save;
2508 struct imsm_super *mpb = super->anchor;
2509 __u32 sum;
2510 struct dl *d;
2511
2512 mpb_save = *mpb;
2513 mpb->num_raid_devs = 0;
2514 mpb->num_disks = 1;
2515 mpb->mpb_size = sizeof(struct imsm_super);
2516 mpb->generation_num = __cpu_to_le32(1UL);
2517
2518 for (d = super->disks; d; d = d->next) {
8796fdc4 2519 if (d->index != -1)
d23fe947
DW
2520 continue;
2521
2522 mpb->disk[0] = d->disk;
2523 sum = __gen_imsm_checksum(mpb);
2524 mpb->family_num = __cpu_to_le32(sum);
2525 sum = __gen_imsm_checksum(mpb);
2526 mpb->check_sum = __cpu_to_le32(sum);
2527
2528 if (store_imsm_mpb(d->fd, super)) {
2529 fprintf(stderr, "%s: failed for device %d:%d %s\n",
2530 __func__, d->major, d->minor, strerror(errno));
2531 *mpb = mpb_save;
e74255d9 2532 return 1;
d23fe947
DW
2533 }
2534 if (doclose) {
2535 close(d->fd);
2536 d->fd = -1;
2537 }
2538 }
2539
2540 *mpb = mpb_save;
e74255d9 2541 return 0;
d23fe947
DW
2542}
2543
c2c087e6 2544static int write_super_imsm(struct intel_super *super, int doclose)
cdddbdbc 2545{
949c47a0 2546 struct imsm_super *mpb = super->anchor;
c2c087e6
DW
2547 struct dl *d;
2548 __u32 generation;
2549 __u32 sum;
d23fe947 2550 int spares = 0;
949c47a0 2551 int i;
a48ac0a8 2552 __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk);
cdddbdbc 2553
c2c087e6
DW
2554 /* 'generation' is incremented everytime the metadata is written */
2555 generation = __le32_to_cpu(mpb->generation_num);
2556 generation++;
2557 mpb->generation_num = __cpu_to_le32(generation);
2558
1ee1e9fc 2559 mpb_size += sizeof(struct imsm_disk) * mpb->num_disks;
d23fe947 2560 for (d = super->disks; d; d = d->next) {
8796fdc4 2561 if (d->index == -1)
d23fe947 2562 spares++;
1ee1e9fc 2563 else
d23fe947 2564 mpb->disk[d->index] = d->disk;
d23fe947 2565 }
47ee5a45
DW
2566 for (d = super->missing; d; d = d->next)
2567 mpb->disk[d->index] = d->disk;
b9f594fe 2568
949c47a0
DW
2569 for (i = 0; i < mpb->num_raid_devs; i++) {
2570 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
2571
ba2de7ba 2572 imsm_copy_dev(dev, get_imsm_dev(super, i));
a48ac0a8 2573 mpb_size += sizeof_imsm_dev(dev, 0);
949c47a0 2574 }
a48ac0a8
DW
2575 mpb_size += __le32_to_cpu(mpb->bbm_log_size);
2576 mpb->mpb_size = __cpu_to_le32(mpb_size);
949c47a0 2577
c2c087e6 2578 /* recalculate checksum */
949c47a0 2579 sum = __gen_imsm_checksum(mpb);
c2c087e6
DW
2580 mpb->check_sum = __cpu_to_le32(sum);
2581
d23fe947 2582 /* write the mpb for disks that compose raid devices */
c2c087e6 2583 for (d = super->disks; d ; d = d->next) {
d23fe947
DW
2584 if (d->index < 0)
2585 continue;
8796fdc4 2586 if (store_imsm_mpb(d->fd, super))
c2c087e6
DW
2587 fprintf(stderr, "%s: failed for device %d:%d %s\n",
2588 __func__, d->major, d->minor, strerror(errno));
c2c087e6
DW
2589 if (doclose) {
2590 close(d->fd);
2591 d->fd = -1;
2592 }
2593 }
2594
d23fe947
DW
2595 if (spares)
2596 return write_super_imsm_spares(super, doclose);
2597
e74255d9 2598 return 0;
c2c087e6
DW
2599}
2600
0e600426 2601
43dad3d6
DW
2602static int create_array(struct supertype *st)
2603{
2604 size_t len;
2605 struct imsm_update_create_array *u;
2606 struct intel_super *super = st->sb;
2607 struct imsm_dev *dev = get_imsm_dev(super, super->current_vol);
54c2c1ea
DW
2608 struct imsm_map *map = get_imsm_map(dev, 0);
2609 struct disk_info *inf;
2610 struct imsm_disk *disk;
2611 int i;
2612 int idx;
43dad3d6 2613
54c2c1ea
DW
2614 len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) +
2615 sizeof(*inf) * map->num_members;
43dad3d6
DW
2616 u = malloc(len);
2617 if (!u) {
2618 fprintf(stderr, "%s: failed to allocate update buffer\n",
2619 __func__);
2620 return 1;
2621 }
2622
2623 u->type = update_create_array;
2624 u->dev_idx = super->current_vol;
2625 imsm_copy_dev(&u->dev, dev);
54c2c1ea
DW
2626 inf = get_disk_info(u);
2627 for (i = 0; i < map->num_members; i++) {
2628 idx = get_imsm_disk_idx(dev, i);
2629 disk = get_imsm_disk(super, idx);
2630 serialcpy(inf[i].serial, disk->serial);
2631 }
43dad3d6
DW
2632 append_metadata_update(st, u, len);
2633
2634 return 0;
2635}
2636
7801ac20 2637static int _add_disk(struct supertype *st)
43dad3d6
DW
2638{
2639 struct intel_super *super = st->sb;
2640 size_t len;
2641 struct imsm_update_add_disk *u;
2642
2643 if (!super->add)
2644 return 0;
2645
2646 len = sizeof(*u);
2647 u = malloc(len);
2648 if (!u) {
2649 fprintf(stderr, "%s: failed to allocate update buffer\n",
2650 __func__);
2651 return 1;
2652 }
2653
2654 u->type = update_add_disk;
2655 append_metadata_update(st, u, len);
2656
2657 return 0;
2658}
2659
c2c087e6
DW
2660static int write_init_super_imsm(struct supertype *st)
2661{
8273f55e 2662 if (st->update_tail) {
43dad3d6
DW
2663 /* queue the recently created array / added disk
2664 * as a metadata update */
8273f55e 2665 struct intel_super *super = st->sb;
8273f55e 2666 struct dl *d;
43dad3d6 2667 int rv;
8273f55e 2668
43dad3d6
DW
2669 /* determine if we are creating a volume or adding a disk */
2670 if (super->current_vol < 0) {
2671 /* in the add disk case we are running in mdmon
2672 * context, so don't close fd's
2673 */
7801ac20 2674 return _add_disk(st);
43dad3d6
DW
2675 } else
2676 rv = create_array(st);
8273f55e
DW
2677
2678 for (d = super->disks; d ; d = d->next) {
2679 close(d->fd);
2680 d->fd = -1;
2681 }
2682
43dad3d6 2683 return rv;
8273f55e
DW
2684 } else
2685 return write_super_imsm(st->sb, 1);
cdddbdbc 2686}
0e600426 2687#endif
cdddbdbc
DW
2688
2689static int store_zero_imsm(struct supertype *st, int fd)
2690{
551c80c1 2691 unsigned long long dsize;
6416d527 2692 void *buf;
551c80c1
DW
2693
2694 get_dev_size(fd, NULL, &dsize);
2695
2696 /* first block is stored on second to last sector of the disk */
2697 if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0)
2698 return 1;
2699
ad97895e
DW
2700 if (posix_memalign(&buf, 512, 512) != 0)
2701 return 1;
2702
eb7ea463
DW
2703 memset(buf, 0, 512);
2704 if (write(fd, buf, 512) != 512)
551c80c1 2705 return 1;
cdddbdbc
DW
2706 return 0;
2707}
2708
0e600426
N
2709static int imsm_bbm_log_size(struct imsm_super *mpb)
2710{
2711 return __le32_to_cpu(mpb->bbm_log_size);
2712}
2713
2714#ifndef MDASSEMBLE
cdddbdbc
DW
2715static int validate_geometry_imsm_container(struct supertype *st, int level,
2716 int layout, int raiddisks, int chunk,
c2c087e6 2717 unsigned long long size, char *dev,
2c514b71
NB
2718 unsigned long long *freesize,
2719 int verbose)
cdddbdbc 2720{
c2c087e6
DW
2721 int fd;
2722 unsigned long long ldsize;
88c32bb1 2723 const struct imsm_orom *orom;
cdddbdbc 2724
c2c087e6
DW
2725 if (level != LEVEL_CONTAINER)
2726 return 0;
2727 if (!dev)
2728 return 1;
2729
88c32bb1
DW
2730 if (check_env("IMSM_NO_PLATFORM"))
2731 orom = NULL;
2732 else
2733 orom = find_imsm_orom();
2734 if (orom && raiddisks > orom->tds) {
2735 if (verbose)
2736 fprintf(stderr, Name ": %d exceeds maximum number of"
2737 " platform supported disks: %d\n",
2738 raiddisks, orom->tds);
2739 return 0;
2740 }
2741
c2c087e6
DW
2742 fd = open(dev, O_RDONLY|O_EXCL, 0);
2743 if (fd < 0) {
2c514b71
NB
2744 if (verbose)
2745 fprintf(stderr, Name ": imsm: Cannot open %s: %s\n",
2746 dev, strerror(errno));
c2c087e6
DW
2747 return 0;
2748 }
2749 if (!get_dev_size(fd, dev, &ldsize)) {
2750 close(fd);
2751 return 0;
2752 }
2753 close(fd);
2754
2755 *freesize = avail_size_imsm(st, ldsize >> 9);
2756
2757 return 1;
cdddbdbc
DW
2758}
2759
0dcecb2e
DW
2760static unsigned long long find_size(struct extent *e, int *idx, int num_extents)
2761{
2762 const unsigned long long base_start = e[*idx].start;
2763 unsigned long long end = base_start + e[*idx].size;
2764 int i;
2765
2766 if (base_start == end)
2767 return 0;
2768
2769 *idx = *idx + 1;
2770 for (i = *idx; i < num_extents; i++) {
2771 /* extend overlapping extents */
2772 if (e[i].start >= base_start &&
2773 e[i].start <= end) {
2774 if (e[i].size == 0)
2775 return 0;
2776 if (e[i].start + e[i].size > end)
2777 end = e[i].start + e[i].size;
2778 } else if (e[i].start > end) {
2779 *idx = i;
2780 break;
2781 }
2782 }
2783
2784 return end - base_start;
2785}
2786
2787static unsigned long long merge_extents(struct intel_super *super, int sum_extents)
2788{
2789 /* build a composite disk with all known extents and generate a new
2790 * 'maxsize' given the "all disks in an array must share a common start
2791 * offset" constraint
2792 */
2793 struct extent *e = calloc(sum_extents, sizeof(*e));
2794 struct dl *dl;
2795 int i, j;
2796 int start_extent;
2797 unsigned long long pos;
2798 unsigned long long start;
2799 unsigned long long maxsize;
2800 unsigned long reserve;
2801
2802 if (!e)
2803 return ~0ULL; /* error */
2804
2805 /* coalesce and sort all extents. also, check to see if we need to
2806 * reserve space between member arrays
2807 */
2808 j = 0;
2809 for (dl = super->disks; dl; dl = dl->next) {
2810 if (!dl->e)
2811 continue;
2812 for (i = 0; i < dl->extent_cnt; i++)
2813 e[j++] = dl->e[i];
2814 }
2815 qsort(e, sum_extents, sizeof(*e), cmp_extent);
2816
2817 /* merge extents */
2818 i = 0;
2819 j = 0;
2820 while (i < sum_extents) {
2821 e[j].start = e[i].start;
2822 e[j].size = find_size(e, &i, sum_extents);
2823 j++;
2824 if (e[j-1].size == 0)
2825 break;
2826 }
2827
2828 pos = 0;
2829 maxsize = 0;
2830 start_extent = 0;
2831 i = 0;
2832 do {
2833 unsigned long long esize;
2834
2835 esize = e[i].start - pos;
2836 if (esize >= maxsize) {
2837 maxsize = esize;
2838 start = pos;
2839 start_extent = i;
2840 }
2841 pos = e[i].start + e[i].size;
2842 i++;
2843 } while (e[i-1].size);
2844 free(e);
2845
2846 if (start_extent > 0)
2847 reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */
2848 else
2849 reserve = 0;
2850
2851 if (maxsize < reserve)
2852 return ~0ULL;
2853
2854 super->create_offset = ~((__u32) 0);
2855 if (start + reserve > super->create_offset)
2856 return ~0ULL; /* start overflows create_offset */
2857 super->create_offset = start + reserve;
2858
2859 return maxsize - reserve;
2860}
2861
88c32bb1
DW
2862static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks)
2863{
2864 if (level < 0 || level == 6 || level == 4)
2865 return 0;
2866
2867 /* if we have an orom prevent invalid raid levels */
2868 if (orom)
2869 switch (level) {
2870 case 0: return imsm_orom_has_raid0(orom);
2871 case 1:
2872 if (raiddisks > 2)
2873 return imsm_orom_has_raid1e(orom);
1c556e92
DW
2874 return imsm_orom_has_raid1(orom) && raiddisks == 2;
2875 case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4;
2876 case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2;
88c32bb1
DW
2877 }
2878 else
2879 return 1; /* not on an Intel RAID platform so anything goes */
2880
2881 return 0;
2882}
2883
35f81cbb 2884#define pr_vrb(fmt, arg...) (void) (verbose && fprintf(stderr, Name fmt, ##arg))
c2c087e6
DW
2885/* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd
2886 * FIX ME add ahci details
2887 */
8b353278
DW
2888static int validate_geometry_imsm_volume(struct supertype *st, int level,
2889 int layout, int raiddisks, int chunk,
c2c087e6 2890 unsigned long long size, char *dev,
2c514b71
NB
2891 unsigned long long *freesize,
2892 int verbose)
cdddbdbc 2893{
c2c087e6
DW
2894 struct stat stb;
2895 struct intel_super *super = st->sb;
a20d2ba5 2896 struct imsm_super *mpb = super->anchor;
c2c087e6
DW
2897 struct dl *dl;
2898 unsigned long long pos = 0;
2899 unsigned long long maxsize;
2900 struct extent *e;
2901 int i;
cdddbdbc 2902
88c32bb1
DW
2903 /* We must have the container info already read in. */
2904 if (!super)
c2c087e6
DW
2905 return 0;
2906
88c32bb1 2907 if (!is_raid_level_supported(super->orom, level, raiddisks)) {
1c556e92
DW
2908 pr_vrb(": platform does not support raid%d with %d disk%s\n",
2909 level, raiddisks, raiddisks > 1 ? "s" : "");
c2c087e6
DW
2910 return 0;
2911 }
78757ce8
DW
2912 if (super->orom && level != 1 &&
2913 !imsm_orom_has_chunk(super->orom, chunk)) {
35f81cbb 2914 pr_vrb(": platform does not support a chunk size of: %d\n", chunk);
c2c087e6 2915 return 0;
88c32bb1
DW
2916 }
2917 if (layout != imsm_level_to_layout(level)) {
2918 if (level == 5)
35f81cbb 2919 pr_vrb(": imsm raid 5 only supports the left-asymmetric layout\n");
88c32bb1 2920 else if (level == 10)
35f81cbb 2921 pr_vrb(": imsm raid 10 only supports the n2 layout\n");
88c32bb1 2922 else
35f81cbb 2923 pr_vrb(": imsm unknown layout %#x for this raid level %d\n",
88c32bb1 2924 layout, level);
c2c087e6 2925 return 0;
88c32bb1 2926 }
c2c087e6
DW
2927
2928 if (!dev) {
2929 /* General test: make sure there is space for
2da8544a
DW
2930 * 'raiddisks' device extents of size 'size' at a given
2931 * offset
c2c087e6 2932 */
e46273eb 2933 unsigned long long minsize = size;
2da8544a 2934 unsigned long long start_offset = ~0ULL;
c2c087e6
DW
2935 int dcnt = 0;
2936 if (minsize == 0)
2937 minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
2938 for (dl = super->disks; dl ; dl = dl->next) {
2939 int found = 0;
2940
bf5a934a 2941 pos = 0;
c2c087e6
DW
2942 i = 0;
2943 e = get_extents(super, dl);
2944 if (!e) continue;
2945 do {
2946 unsigned long long esize;
2947 esize = e[i].start - pos;
2948 if (esize >= minsize)
2949 found = 1;
2da8544a
DW
2950 if (found && start_offset == ~0ULL) {
2951 start_offset = pos;
2952 break;
2953 } else if (found && pos != start_offset) {
2954 found = 0;
2955 break;
2956 }
c2c087e6
DW
2957 pos = e[i].start + e[i].size;
2958 i++;
2959 } while (e[i-1].size);
2960 if (found)
2961 dcnt++;
2962 free(e);
2963 }
2964 if (dcnt < raiddisks) {
2c514b71
NB
2965 if (verbose)
2966 fprintf(stderr, Name ": imsm: Not enough "
2967 "devices with space for this array "
2968 "(%d < %d)\n",
2969 dcnt, raiddisks);
c2c087e6
DW
2970 return 0;
2971 }
2972 return 1;
2973 }
0dcecb2e 2974
c2c087e6
DW
2975 /* This device must be a member of the set */
2976 if (stat(dev, &stb) < 0)
2977 return 0;
2978 if ((S_IFMT & stb.st_mode) != S_IFBLK)
2979 return 0;
2980 for (dl = super->disks ; dl ; dl = dl->next) {
2981 if (dl->major == major(stb.st_rdev) &&
2982 dl->minor == minor(stb.st_rdev))
2983 break;
2984 }
2985 if (!dl) {
2c514b71
NB
2986 if (verbose)
2987 fprintf(stderr, Name ": %s is not in the "
2988 "same imsm set\n", dev);
c2c087e6 2989 return 0;
a20d2ba5
DW
2990 } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) {
2991 /* If a volume is present then the current creation attempt
2992 * cannot incorporate new spares because the orom may not
2993 * understand this configuration (all member disks must be
2994 * members of each array in the container).
2995 */
2996 fprintf(stderr, Name ": %s is a spare and a volume"
2997 " is already defined for this container\n", dev);
2998 fprintf(stderr, Name ": The option-rom requires all member"
2999 " disks to be a member of all volumes\n");
3000 return 0;
c2c087e6 3001 }
0dcecb2e
DW
3002
3003 /* retrieve the largest free space block */
c2c087e6
DW
3004 e = get_extents(super, dl);
3005 maxsize = 0;
3006 i = 0;
0dcecb2e
DW
3007 if (e) {
3008 do {
3009 unsigned long long esize;
3010
3011 esize = e[i].start - pos;
3012 if (esize >= maxsize)
3013 maxsize = esize;
3014 pos = e[i].start + e[i].size;
3015 i++;
3016 } while (e[i-1].size);
3017 dl->e = e;
3018 dl->extent_cnt = i;
3019 } else {
3020 if (verbose)
3021 fprintf(stderr, Name ": unable to determine free space for: %s\n",
3022 dev);
3023 return 0;
3024 }
3025 if (maxsize < size) {
3026 if (verbose)
3027 fprintf(stderr, Name ": %s not enough space (%llu < %llu)\n",
3028 dev, maxsize, size);
3029 return 0;
3030 }
3031
3032 /* count total number of extents for merge */
3033 i = 0;
3034 for (dl = super->disks; dl; dl = dl->next)
3035 if (dl->e)
3036 i += dl->extent_cnt;
3037
3038 maxsize = merge_extents(super, i);
3039 if (maxsize < size) {
3040 if (verbose)
3041 fprintf(stderr, Name ": not enough space after merge (%llu < %llu)\n",
3042 maxsize, size);
3043 return 0;
3044 } else if (maxsize == ~0ULL) {
3045 if (verbose)
3046 fprintf(stderr, Name ": failed to merge %d extents\n", i);
3047 return 0;
3048 }
3049
c2c087e6
DW
3050 *freesize = maxsize;
3051
3052 return 1;
cdddbdbc
DW
3053}
3054
bf5a934a
DW
3055static int validate_geometry_imsm(struct supertype *st, int level, int layout,
3056 int raiddisks, int chunk, unsigned long long size,
3057 char *dev, unsigned long long *freesize,
3058 int verbose)
3059{
3060 int fd, cfd;
3061 struct mdinfo *sra;
3062
3063 /* if given unused devices create a container
3064 * if given given devices in a container create a member volume
3065 */
3066 if (level == LEVEL_CONTAINER) {
3067 /* Must be a fresh device to add to a container */
3068 return validate_geometry_imsm_container(st, level, layout,
3069 raiddisks, chunk, size,
3070 dev, freesize,
3071 verbose);
3072 }
3073
8592f29d
N
3074 if (!dev) {
3075 if (st->sb && freesize) {
3076 /* Should do auto-layout here */
3077 fprintf(stderr, Name ": IMSM does not support auto-layout yet\n");
3078 return 0;
3079 }
3080 return 1;
3081 }
bf5a934a
DW
3082 if (st->sb) {
3083 /* creating in a given container */
3084 return validate_geometry_imsm_volume(st, level, layout,
3085 raiddisks, chunk, size,
3086 dev, freesize, verbose);
3087 }
3088
3089 /* limit creation to the following levels */
3090 if (!dev)
3091 switch (level) {
3092 case 0:
3093 case 1:
3094 case 10:
3095 case 5:
3096 break;
3097 default:
3098 return 1;
3099 }
3100
3101 /* This device needs to be a device in an 'imsm' container */
3102 fd = open(dev, O_RDONLY|O_EXCL, 0);
3103 if (fd >= 0) {
3104 if (verbose)
3105 fprintf(stderr,
3106 Name ": Cannot create this array on device %s\n",
3107 dev);
3108 close(fd);
3109 return 0;
3110 }
3111 if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) {
3112 if (verbose)
3113 fprintf(stderr, Name ": Cannot open %s: %s\n",
3114 dev, strerror(errno));
3115 return 0;
3116 }
3117 /* Well, it is in use by someone, maybe an 'imsm' container. */
3118 cfd = open_container(fd);
3119 if (cfd < 0) {
3120 close(fd);
3121 if (verbose)
3122 fprintf(stderr, Name ": Cannot use %s: It is busy\n",
3123 dev);
3124 return 0;
3125 }
3126 sra = sysfs_read(cfd, 0, GET_VERSION);
3127 close(fd);
3128 if (sra && sra->array.major_version == -1 &&
3129 strcmp(sra->text_version, "imsm") == 0) {
3130 /* This is a member of a imsm container. Load the container
3131 * and try to create a volume
3132 */
3133 struct intel_super *super;
3134
3135 if (load_super_imsm_all(st, cfd, (void **) &super, NULL, 1) == 0) {
3136 st->sb = super;
3137 st->container_dev = fd2devnum(cfd);
3138 close(cfd);
3139 return validate_geometry_imsm_volume(st, level, layout,
3140 raiddisks, chunk,
3141 size, dev,
3142 freesize, verbose);
3143 }
3144 close(cfd);
3145 } else /* may belong to another container */
3146 return 0;
3147
3148 return 1;
3149}
0e600426 3150#endif /* MDASSEMBLE */
bf5a934a 3151
cdddbdbc
DW
3152static struct mdinfo *container_content_imsm(struct supertype *st)
3153{
4f5bc454
DW
3154 /* Given a container loaded by load_super_imsm_all,
3155 * extract information about all the arrays into
3156 * an mdinfo tree.
3157 *
3158 * For each imsm_dev create an mdinfo, fill it in,
3159 * then look for matching devices in super->disks
3160 * and create appropriate device mdinfo.
3161 */
3162 struct intel_super *super = st->sb;
949c47a0 3163 struct imsm_super *mpb = super->anchor;
4f5bc454
DW
3164 struct mdinfo *rest = NULL;
3165 int i;
cdddbdbc 3166
604b746f
JD
3167 /* do not assemble arrays that might have bad blocks */
3168 if (imsm_bbm_log_size(super->anchor)) {
3169 fprintf(stderr, Name ": BBM log found in metadata. "
3170 "Cannot activate array(s).\n");
3171 return NULL;
3172 }
3173
4f5bc454 3174 for (i = 0; i < mpb->num_raid_devs; i++) {
949c47a0 3175 struct imsm_dev *dev = get_imsm_dev(super, i);
a965f303 3176 struct imsm_map *map = get_imsm_map(dev, 0);
4f5bc454 3177 struct mdinfo *this;
4f5bc454
DW
3178 int slot;
3179
3180 this = malloc(sizeof(*this));
3181 memset(this, 0, sizeof(*this));
3182 this->next = rest;
4f5bc454 3183
301406c9
DW
3184 super->current_vol = i;
3185 getinfo_super_imsm_volume(st, this);
4f5bc454 3186 for (slot = 0 ; slot < map->num_members; slot++) {
4f5bc454
DW
3187 struct mdinfo *info_d;
3188 struct dl *d;
3189 int idx;
9a1608e5 3190 int skip;
4f5bc454 3191 __u32 s;
7eef0453 3192 __u32 ord;
4f5bc454 3193
9a1608e5 3194 skip = 0;
ff077194 3195 idx = get_imsm_disk_idx(dev, slot);
7eef0453 3196 ord = get_imsm_ord_tbl_ent(dev, slot);
4f5bc454
DW
3197 for (d = super->disks; d ; d = d->next)
3198 if (d->index == idx)
3199 break;
3200
3201 if (d == NULL)
9a1608e5
DW
3202 skip = 1;
3203
f2f27e63 3204 s = d ? d->disk.status : 0;
9a1608e5
DW
3205 if (s & FAILED_DISK)
3206 skip = 1;
3207 if (!(s & USABLE_DISK))
3208 skip = 1;
7eef0453
DW
3209 if (ord & IMSM_ORD_REBUILD)
3210 skip = 1;
9a1608e5
DW
3211
3212 /*
3213 * if we skip some disks the array will be assmebled degraded;
3214 * reset resync start to avoid a dirty-degraded situation
3215 *
3216 * FIXME handle dirty degraded
3217 */
3218 if (skip && !dev->vol.dirty)
3219 this->resync_start = ~0ULL;
3220 if (skip)
3221 continue;
4f5bc454
DW
3222
3223 info_d = malloc(sizeof(*info_d));
9a1608e5
DW
3224 if (!info_d) {
3225 fprintf(stderr, Name ": failed to allocate disk"
3226 " for volume %s\n", (char *) dev->volume);
3227 free(this);
3228 this = rest;
3229 break;
3230 }
4f5bc454
DW
3231 memset(info_d, 0, sizeof(*info_d));
3232 info_d->next = this->devs;
3233 this->devs = info_d;
3234
4f5bc454
DW
3235 info_d->disk.number = d->index;
3236 info_d->disk.major = d->major;
3237 info_d->disk.minor = d->minor;
3238 info_d->disk.raid_disk = slot;
4f5bc454
DW
3239
3240 this->array.working_disks++;
3241
3242 info_d->events = __le32_to_cpu(mpb->generation_num);
3243 info_d->data_offset = __le32_to_cpu(map->pba_of_lba0);
3244 info_d->component_size = __le32_to_cpu(map->blocks_per_member);
3245 if (d->devname)
3246 strcpy(info_d->name, d->devname);
3247 }
9a1608e5 3248 rest = this;
4f5bc454
DW
3249 }
3250
3251 return rest;
cdddbdbc
DW
3252}
3253
845dea95 3254
0e600426 3255#ifndef MDASSEMBLE
cba0191b
NB
3256static int imsm_open_new(struct supertype *c, struct active_array *a,
3257 char *inst)
845dea95 3258{
0372d5a2 3259 struct intel_super *super = c->sb;
949c47a0 3260 struct imsm_super *mpb = super->anchor;
0372d5a2 3261
949c47a0 3262 if (atoi(inst) >= mpb->num_raid_devs) {
0372d5a2
DW
3263 fprintf(stderr, "%s: subarry index %d, out of range\n",
3264 __func__, atoi(inst));
3265 return -ENODEV;
3266 }
3267
4e6e574a 3268 dprintf("imsm: open_new %s\n", inst);
cba0191b 3269 a->info.container_member = atoi(inst);
845dea95
NB
3270 return 0;
3271}
3272
fb49eef2 3273static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, int failed)
c2a1e7da 3274{
a965f303 3275 struct imsm_map *map = get_imsm_map(dev, 0);
c2a1e7da
DW
3276
3277 if (!failed)
3393c6af
DW
3278 return map->map_state == IMSM_T_STATE_UNINITIALIZED ?
3279 IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL;
c2a1e7da
DW
3280
3281 switch (get_imsm_raid_level(map)) {
3282 case 0:
3283 return IMSM_T_STATE_FAILED;
3284 break;
3285 case 1:
3286 if (failed < map->num_members)
3287 return IMSM_T_STATE_DEGRADED;
3288 else
3289 return IMSM_T_STATE_FAILED;
3290 break;
3291 case 10:
3292 {
3293 /**
c92a2527
DW
3294 * check to see if any mirrors have failed, otherwise we
3295 * are degraded. Even numbered slots are mirrored on
3296 * slot+1
c2a1e7da 3297 */
c2a1e7da 3298 int i;
d9b420a5
N
3299 /* gcc -Os complains that this is unused */
3300 int insync = insync;
c2a1e7da
DW
3301
3302 for (i = 0; i < map->num_members; i++) {
c92a2527
DW
3303 __u32 ord = get_imsm_ord_tbl_ent(dev, i);
3304 int idx = ord_to_idx(ord);
3305 struct imsm_disk *disk;
c2a1e7da 3306
c92a2527
DW
3307 /* reset the potential in-sync count on even-numbered
3308 * slots. num_copies is always 2 for imsm raid10
3309 */
3310 if ((i & 1) == 0)
3311 insync = 2;
c2a1e7da 3312
c92a2527 3313 disk = get_imsm_disk(super, idx);
f2f27e63 3314 if (!disk || disk->status & FAILED_DISK ||
c92a2527
DW
3315 ord & IMSM_ORD_REBUILD)
3316 insync--;
c2a1e7da 3317
c92a2527
DW
3318 /* no in-sync disks left in this mirror the
3319 * array has failed
3320 */
3321 if (insync == 0)
3322 return IMSM_T_STATE_FAILED;
c2a1e7da
DW
3323 }
3324
3325 return IMSM_T_STATE_DEGRADED;
3326 }
3327 case 5:
3328 if (failed < 2)
3329 return IMSM_T_STATE_DEGRADED;
3330 else
3331 return IMSM_T_STATE_FAILED;
3332 break;
3333 default:
3334 break;
3335 }
3336
3337 return map->map_state;
3338}
3339
ff077194 3340static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev)
c2a1e7da
DW
3341{
3342 int i;
3343 int failed = 0;
3344 struct imsm_disk *disk;
ff077194 3345 struct imsm_map *map = get_imsm_map(dev, 0);
0556e1a2
DW
3346 struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state);
3347 __u32 ord;
3348 int idx;
c2a1e7da 3349
0556e1a2
DW
3350 /* at the beginning of migration we set IMSM_ORD_REBUILD on
3351 * disks that are being rebuilt. New failures are recorded to
3352 * map[0]. So we look through all the disks we started with and
3353 * see if any failures are still present, or if any new ones
3354 * have arrived
3355 *
3356 * FIXME add support for online capacity expansion and
3357 * raid-level-migration
3358 */
3359 for (i = 0; i < prev->num_members; i++) {
3360 ord = __le32_to_cpu(prev->disk_ord_tbl[i]);
3361 ord |= __le32_to_cpu(map->disk_ord_tbl[i]);
3362 idx = ord_to_idx(ord);
c2a1e7da 3363
949c47a0 3364 disk = get_imsm_disk(super, idx);
f2f27e63 3365 if (!disk || disk->status & FAILED_DISK ||
b10b37b8 3366 ord & IMSM_ORD_REBUILD)
fcb84475 3367 failed++;
c2a1e7da
DW
3368 }
3369
3370 return failed;
845dea95
NB
3371}
3372
0c046afd
DW
3373static int is_resyncing(struct imsm_dev *dev)
3374{
3375 struct imsm_map *migr_map;
3376
3377 if (!dev->vol.migr_state)
3378 return 0;
3379
e3bba0e0 3380 if (dev->vol.migr_type == MIGR_INIT)
0c046afd
DW
3381 return 1;
3382
3383 migr_map = get_imsm_map(dev, 1);
3384
3385 if (migr_map->map_state == IMSM_T_STATE_NORMAL)
3386 return 1;
3387 else
3388 return 0;
3389}
3390
3391static int is_rebuilding(struct imsm_dev *dev)
3392{
3393 struct imsm_map *migr_map;
3394
3395 if (!dev->vol.migr_state)
3396 return 0;
3397
e3bba0e0 3398 if (dev->vol.migr_type != MIGR_REBUILD)
0c046afd
DW
3399 return 0;
3400
3401 migr_map = get_imsm_map(dev, 1);
3402
3403 if (migr_map->map_state == IMSM_T_STATE_DEGRADED)
3404 return 1;
3405 else
3406 return 0;
3407}
3408
0556e1a2
DW
3409/* return true if we recorded new information */
3410static int mark_failure(struct imsm_dev *dev, struct imsm_disk *disk, int idx)
47ee5a45 3411{
0556e1a2
DW
3412 __u32 ord;
3413 int slot;
3414 struct imsm_map *map;
3415
3416 /* new failures are always set in map[0] */
3417 map = get_imsm_map(dev, 0);
3418
3419 slot = get_imsm_disk_slot(map, idx);
3420 if (slot < 0)
3421 return 0;
3422
3423 ord = __le32_to_cpu(map->disk_ord_tbl[slot]);
3424 if ((disk->status & FAILED_DISK) && (ord & IMSM_ORD_REBUILD))
3425 return 0;
3426
f2f27e63 3427 disk->status |= FAILED_DISK;
0556e1a2
DW
3428 set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD);
3429 if (map->failed_disk_num == ~0)
3430 map->failed_disk_num = slot;
3431 return 1;
3432}
3433
3434static void mark_missing(struct imsm_dev *dev, struct imsm_disk *disk, int idx)
3435{
3436 mark_failure(dev, disk, idx);
3437
3438 if (disk->scsi_id == __cpu_to_le32(~(__u32)0))
3439 return;
3440
47ee5a45
DW
3441 disk->scsi_id = __cpu_to_le32(~(__u32)0);
3442 memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1);
3443}
3444
0c046afd
DW
3445/* Handle dirty -> clean transititions and resync. Degraded and rebuild
3446 * states are handled in imsm_set_disk() with one exception, when a
3447 * resync is stopped due to a new failure this routine will set the
3448 * 'degraded' state for the array.
3449 */
01f157d7 3450static int imsm_set_array_state(struct active_array *a, int consistent)
a862209d
DW
3451{
3452 int inst = a->info.container_member;
3453 struct intel_super *super = a->container->sb;
949c47a0 3454 struct imsm_dev *dev = get_imsm_dev(super, inst);
a965f303 3455 struct imsm_map *map = get_imsm_map(dev, 0);
0c046afd
DW
3456 int failed = imsm_count_failed(super, dev);
3457 __u8 map_state = imsm_check_degraded(super, dev, failed);
a862209d 3458
47ee5a45
DW
3459 /* before we activate this array handle any missing disks */
3460 if (consistent == 2 && super->missing) {
3461 struct dl *dl;
3462
3463 dprintf("imsm: mark missing\n");
3464 end_migration(dev, map_state);
3465 for (dl = super->missing; dl; dl = dl->next)
0556e1a2 3466 mark_missing(dev, &dl->disk, dl->index);
47ee5a45
DW
3467 super->updates_pending++;
3468 }
3469
0c046afd 3470 if (consistent == 2 &&
593add1b 3471 (!is_resync_complete(a) ||
0c046afd
DW
3472 map_state != IMSM_T_STATE_NORMAL ||
3473 dev->vol.migr_state))
01f157d7 3474 consistent = 0;
272906ef 3475
593add1b 3476 if (is_resync_complete(a)) {
0c046afd 3477 /* complete intialization / resync,
0556e1a2
DW
3478 * recovery and interrupted recovery is completed in
3479 * ->set_disk
0c046afd
DW
3480 */
3481 if (is_resyncing(dev)) {
3482 dprintf("imsm: mark resync done\n");
f8f603f1 3483 end_migration(dev, map_state);
115c3803 3484 super->updates_pending++;
115c3803 3485 }
0c046afd
DW
3486 } else if (!is_resyncing(dev) && !failed) {
3487 /* mark the start of the init process if nothing is failed */
3488 dprintf("imsm: mark resync start (%llu)\n", a->resync_start);
e3bba0e0
DW
3489 if (map->map_state == IMSM_T_STATE_NORMAL)
3490 migrate(dev, IMSM_T_STATE_NORMAL, MIGR_REBUILD);
3491 else
3492 migrate(dev, IMSM_T_STATE_NORMAL, MIGR_INIT);
3393c6af 3493 super->updates_pending++;
115c3803 3494 }
a862209d 3495
f8f603f1
DW
3496 /* check if we can update the migration checkpoint */
3497 if (dev->vol.migr_state &&
3498 __le32_to_cpu(dev->vol.curr_migr_unit) != a->resync_start) {
3499 dprintf("imsm: checkpoint migration (%llu)\n", a->resync_start);
3500 dev->vol.curr_migr_unit = __cpu_to_le32(a->resync_start);
3501 super->updates_pending++;
3502 }
3503
3393c6af 3504 /* mark dirty / clean */
0c046afd 3505 if (dev->vol.dirty != !consistent) {
3393c6af 3506 dprintf("imsm: mark '%s' (%llu)\n",
0c046afd
DW
3507 consistent ? "clean" : "dirty", a->resync_start);
3508 if (consistent)
3509 dev->vol.dirty = 0;
3510 else
3511 dev->vol.dirty = 1;
a862209d
DW
3512 super->updates_pending++;
3513 }
01f157d7 3514 return consistent;
a862209d
DW
3515}
3516
8d45d196 3517static void imsm_set_disk(struct active_array *a, int n, int state)
845dea95 3518{
8d45d196
DW
3519 int inst = a->info.container_member;
3520 struct intel_super *super = a->container->sb;
949c47a0 3521 struct imsm_dev *dev = get_imsm_dev(super, inst);
a965f303 3522 struct imsm_map *map = get_imsm_map(dev, 0);
8d45d196 3523 struct imsm_disk *disk;
0c046afd 3524 int failed;
b10b37b8 3525 __u32 ord;
0c046afd 3526 __u8 map_state;