]>
Commit | Line | Data |
---|---|---|
cdddbdbc DW |
1 | /* |
2 | * mdadm - Intel(R) Matrix Storage Manager Support | |
3 | * | |
4 | * Copyright (C) 2002-2007 Intel Corporation | |
5 | * | |
6 | * This program is free software; you can redistribute it and/or modify it | |
7 | * under the terms and conditions of the GNU General Public License, | |
8 | * version 2, as published by the Free Software Foundation. | |
9 | * | |
10 | * This program is distributed in the hope it will be useful, but WITHOUT | |
11 | * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or | |
12 | * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for | |
13 | * more details. | |
14 | * | |
15 | * You should have received a copy of the GNU General Public License along with | |
16 | * this program; if not, write to the Free Software Foundation, Inc., | |
17 | * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA. | |
18 | */ | |
19 | ||
20 | #include "mdadm.h" | |
c2a1e7da | 21 | #include "mdmon.h" |
cdddbdbc DW |
22 | #include <values.h> |
23 | #include <scsi/sg.h> | |
24 | #include <ctype.h> | |
25 | ||
26 | /* MPB == Metadata Parameter Block */ | |
27 | #define MPB_SIGNATURE "Intel Raid ISM Cfg Sig. " | |
28 | #define MPB_SIG_LEN (strlen(MPB_SIGNATURE)) | |
29 | #define MPB_VERSION_RAID0 "1.0.00" | |
30 | #define MPB_VERSION_RAID1 "1.1.00" | |
31 | #define MPB_VERSION_RAID5 "1.2.02" | |
32 | #define MAX_SIGNATURE_LENGTH 32 | |
33 | #define MAX_RAID_SERIAL_LEN 16 | |
c2c087e6 DW |
34 | #define MPB_SECTOR_CNT 418 |
35 | #define IMSM_RESERVED_SECTORS 4096 | |
cdddbdbc DW |
36 | |
37 | /* Disk configuration info. */ | |
38 | #define IMSM_MAX_DEVICES 255 | |
39 | struct imsm_disk { | |
40 | __u8 serial[MAX_RAID_SERIAL_LEN];/* 0xD8 - 0xE7 ascii serial number */ | |
41 | __u32 total_blocks; /* 0xE8 - 0xEB total blocks */ | |
42 | __u32 scsi_id; /* 0xEC - 0xEF scsi ID */ | |
43 | __u32 status; /* 0xF0 - 0xF3 */ | |
44 | #define SPARE_DISK 0x01 /* Spare */ | |
45 | #define CONFIGURED_DISK 0x02 /* Member of some RaidDev */ | |
46 | #define FAILED_DISK 0x04 /* Permanent failure */ | |
47 | #define USABLE_DISK 0x08 /* Fully usable unless FAILED_DISK is set */ | |
48 | ||
49 | #define IMSM_DISK_FILLERS 5 | |
50 | __u32 filler[IMSM_DISK_FILLERS]; /* 0xF4 - 0x107 MPB_DISK_FILLERS for future expansion */ | |
51 | }; | |
52 | ||
53 | /* RAID map configuration infos. */ | |
54 | struct imsm_map { | |
55 | __u32 pba_of_lba0; /* start address of partition */ | |
56 | __u32 blocks_per_member;/* blocks per member */ | |
57 | __u32 num_data_stripes; /* number of data stripes */ | |
58 | __u16 blocks_per_strip; | |
59 | __u8 map_state; /* Normal, Uninitialized, Degraded, Failed */ | |
60 | #define IMSM_T_STATE_NORMAL 0 | |
61 | #define IMSM_T_STATE_UNINITIALIZED 1 | |
62 | #define IMSM_T_STATE_DEGRADED 2 /* FIXME: is this correct? */ | |
63 | #define IMSM_T_STATE_FAILED 3 /* FIXME: is this correct? */ | |
64 | __u8 raid_level; | |
65 | #define IMSM_T_RAID0 0 | |
66 | #define IMSM_T_RAID1 1 | |
67 | #define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */ | |
68 | __u8 num_members; /* number of member disks */ | |
69 | __u8 reserved[3]; | |
70 | __u32 filler[7]; /* expansion area */ | |
7eef0453 | 71 | #define IMSM_ORD_REBUILD (1 << 24) |
cdddbdbc | 72 | __u32 disk_ord_tbl[1]; /* disk_ord_tbl[num_members], |
7eef0453 DW |
73 | * top byte contains some flags |
74 | */ | |
cdddbdbc DW |
75 | } __attribute__ ((packed)); |
76 | ||
77 | struct imsm_vol { | |
78 | __u32 reserved[2]; | |
79 | __u8 migr_state; /* Normal or Migrating */ | |
80 | __u8 migr_type; /* Initializing, Rebuilding, ... */ | |
81 | __u8 dirty; | |
82 | __u8 fill[1]; | |
83 | __u32 filler[5]; | |
84 | struct imsm_map map[1]; | |
85 | /* here comes another one if migr_state */ | |
86 | } __attribute__ ((packed)); | |
87 | ||
88 | struct imsm_dev { | |
89 | __u8 volume[MAX_RAID_SERIAL_LEN]; | |
90 | __u32 size_low; | |
91 | __u32 size_high; | |
92 | __u32 status; /* Persistent RaidDev status */ | |
93 | __u32 reserved_blocks; /* Reserved blocks at beginning of volume */ | |
94 | #define IMSM_DEV_FILLERS 12 | |
95 | __u32 filler[IMSM_DEV_FILLERS]; | |
96 | struct imsm_vol vol; | |
97 | } __attribute__ ((packed)); | |
98 | ||
99 | struct imsm_super { | |
100 | __u8 sig[MAX_SIGNATURE_LENGTH]; /* 0x00 - 0x1F */ | |
101 | __u32 check_sum; /* 0x20 - 0x23 MPB Checksum */ | |
102 | __u32 mpb_size; /* 0x24 - 0x27 Size of MPB */ | |
103 | __u32 family_num; /* 0x28 - 0x2B Checksum from first time this config was written */ | |
104 | __u32 generation_num; /* 0x2C - 0x2F Incremented each time this array's MPB is written */ | |
604b746f JD |
105 | __u32 error_log_size; /* 0x30 - 0x33 in bytes */ |
106 | __u32 attributes; /* 0x34 - 0x37 */ | |
cdddbdbc DW |
107 | __u8 num_disks; /* 0x38 Number of configured disks */ |
108 | __u8 num_raid_devs; /* 0x39 Number of configured volumes */ | |
604b746f JD |
109 | __u8 error_log_pos; /* 0x3A */ |
110 | __u8 fill[1]; /* 0x3B */ | |
111 | __u32 cache_size; /* 0x3c - 0x40 in mb */ | |
112 | __u32 orig_family_num; /* 0x40 - 0x43 original family num */ | |
113 | __u32 pwr_cycle_count; /* 0x44 - 0x47 simulated power cycle count for array */ | |
114 | __u32 bbm_log_size; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */ | |
115 | #define IMSM_FILLERS 35 | |
116 | __u32 filler[IMSM_FILLERS]; /* 0x4C - 0xD7 RAID_MPB_FILLERS */ | |
cdddbdbc DW |
117 | struct imsm_disk disk[1]; /* 0xD8 diskTbl[numDisks] */ |
118 | /* here comes imsm_dev[num_raid_devs] */ | |
604b746f | 119 | /* here comes BBM logs */ |
cdddbdbc DW |
120 | } __attribute__ ((packed)); |
121 | ||
604b746f JD |
122 | #define BBM_LOG_MAX_ENTRIES 254 |
123 | ||
124 | struct bbm_log_entry { | |
125 | __u64 defective_block_start; | |
126 | #define UNREADABLE 0xFFFFFFFF | |
127 | __u32 spare_block_offset; | |
128 | __u16 remapped_marked_count; | |
129 | __u16 disk_ordinal; | |
130 | } __attribute__ ((__packed__)); | |
131 | ||
132 | struct bbm_log { | |
133 | __u32 signature; /* 0xABADB10C */ | |
134 | __u32 entry_count; | |
135 | __u32 reserved_spare_block_count; /* 0 */ | |
136 | __u32 reserved; /* 0xFFFF */ | |
137 | __u64 first_spare_lba; | |
138 | struct bbm_log_entry mapped_block_entries[BBM_LOG_MAX_ENTRIES]; | |
139 | } __attribute__ ((__packed__)); | |
140 | ||
141 | ||
cdddbdbc DW |
142 | #ifndef MDASSEMBLE |
143 | static char *map_state_str[] = { "normal", "uninitialized", "degraded", "failed" }; | |
144 | #endif | |
145 | ||
87eb16df | 146 | static unsigned int sector_count(__u32 bytes) |
cdddbdbc | 147 | { |
87eb16df DW |
148 | return ((bytes + (512-1)) & (~(512-1))) / 512; |
149 | } | |
cdddbdbc | 150 | |
87eb16df DW |
151 | static unsigned int mpb_sectors(struct imsm_super *mpb) |
152 | { | |
153 | return sector_count(__le32_to_cpu(mpb->mpb_size)); | |
cdddbdbc DW |
154 | } |
155 | ||
156 | /* internal representation of IMSM metadata */ | |
157 | struct intel_super { | |
158 | union { | |
949c47a0 DW |
159 | void *buf; /* O_DIRECT buffer for reading/writing metadata */ |
160 | struct imsm_super *anchor; /* immovable parameters */ | |
cdddbdbc | 161 | }; |
949c47a0 | 162 | size_t len; /* size of the 'buf' allocation */ |
c2c087e6 DW |
163 | int updates_pending; /* count of pending updates for mdmon */ |
164 | int creating_imsm; /* flag to indicate container creation */ | |
bf5a934a | 165 | int current_vol; /* index of raid device undergoing creation */ |
949c47a0 DW |
166 | #define IMSM_MAX_RAID_DEVS 2 |
167 | struct imsm_dev *dev_tbl[IMSM_MAX_RAID_DEVS]; | |
cdddbdbc DW |
168 | struct dl { |
169 | struct dl *next; | |
170 | int index; | |
171 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
172 | int major, minor; | |
173 | char *devname; | |
b9f594fe | 174 | struct imsm_disk disk; |
cdddbdbc DW |
175 | int fd; |
176 | } *disks; | |
43dad3d6 DW |
177 | struct dl *add; /* list of disks to add while mdmon active */ |
178 | struct bbm_log *bbm_log; | |
cdddbdbc DW |
179 | }; |
180 | ||
c2c087e6 DW |
181 | struct extent { |
182 | unsigned long long start, size; | |
183 | }; | |
184 | ||
88758e9d DW |
185 | /* definition of messages passed to imsm_process_update */ |
186 | enum imsm_update_type { | |
187 | update_activate_spare, | |
8273f55e | 188 | update_create_array, |
43dad3d6 | 189 | update_add_disk, |
88758e9d DW |
190 | }; |
191 | ||
192 | struct imsm_update_activate_spare { | |
193 | enum imsm_update_type type; | |
d23fe947 | 194 | struct dl *dl; |
88758e9d DW |
195 | int slot; |
196 | int array; | |
197 | struct imsm_update_activate_spare *next; | |
198 | }; | |
199 | ||
8273f55e DW |
200 | struct imsm_update_create_array { |
201 | enum imsm_update_type type; | |
8273f55e | 202 | int dev_idx; |
6a3e913e | 203 | struct imsm_dev dev; |
8273f55e DW |
204 | }; |
205 | ||
43dad3d6 DW |
206 | struct imsm_update_add_disk { |
207 | enum imsm_update_type type; | |
208 | }; | |
209 | ||
0030e8d6 DW |
210 | static int imsm_env_devname_as_serial(void) |
211 | { | |
212 | char *val = getenv("IMSM_DEVNAME_AS_SERIAL"); | |
213 | ||
214 | if (val && atoi(val) == 1) | |
215 | return 1; | |
216 | ||
217 | return 0; | |
218 | } | |
219 | ||
220 | ||
cdddbdbc DW |
221 | static struct supertype *match_metadata_desc_imsm(char *arg) |
222 | { | |
223 | struct supertype *st; | |
224 | ||
225 | if (strcmp(arg, "imsm") != 0 && | |
226 | strcmp(arg, "default") != 0 | |
227 | ) | |
228 | return NULL; | |
229 | ||
230 | st = malloc(sizeof(*st)); | |
ef609477 | 231 | memset(st, 0, sizeof(*st)); |
cdddbdbc DW |
232 | st->ss = &super_imsm; |
233 | st->max_devs = IMSM_MAX_DEVICES; | |
234 | st->minor_version = 0; | |
235 | st->sb = NULL; | |
236 | return st; | |
237 | } | |
238 | ||
cdddbdbc DW |
239 | static __u8 *get_imsm_version(struct imsm_super *mpb) |
240 | { | |
241 | return &mpb->sig[MPB_SIG_LEN]; | |
242 | } | |
243 | ||
949c47a0 DW |
244 | /* retrieve a disk directly from the anchor when the anchor is known to be |
245 | * up-to-date, currently only at load time | |
246 | */ | |
247 | static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index) | |
cdddbdbc | 248 | { |
949c47a0 | 249 | if (index >= mpb->num_disks) |
cdddbdbc DW |
250 | return NULL; |
251 | return &mpb->disk[index]; | |
252 | } | |
253 | ||
b9f594fe | 254 | /* retrieve a disk from the parsed metadata */ |
949c47a0 DW |
255 | static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index) |
256 | { | |
b9f594fe DW |
257 | struct dl *d; |
258 | ||
259 | for (d = super->disks; d; d = d->next) | |
260 | if (d->index == index) | |
261 | return &d->disk; | |
262 | ||
263 | return NULL; | |
949c47a0 DW |
264 | } |
265 | ||
266 | /* generate a checksum directly from the anchor when the anchor is known to be | |
267 | * up-to-date, currently only at load or write_super after coalescing | |
268 | */ | |
269 | static __u32 __gen_imsm_checksum(struct imsm_super *mpb) | |
cdddbdbc DW |
270 | { |
271 | __u32 end = mpb->mpb_size / sizeof(end); | |
272 | __u32 *p = (__u32 *) mpb; | |
273 | __u32 sum = 0; | |
274 | ||
275 | while (end--) | |
276 | sum += __le32_to_cpu(*p++); | |
277 | ||
278 | return sum - __le32_to_cpu(mpb->check_sum); | |
279 | } | |
280 | ||
a965f303 DW |
281 | static size_t sizeof_imsm_map(struct imsm_map *map) |
282 | { | |
283 | return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1); | |
284 | } | |
285 | ||
286 | struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map) | |
cdddbdbc | 287 | { |
a965f303 DW |
288 | struct imsm_map *map = &dev->vol.map[0]; |
289 | ||
290 | if (second_map && !dev->vol.migr_state) | |
291 | return NULL; | |
292 | else if (second_map) { | |
293 | void *ptr = map; | |
294 | ||
295 | return ptr + sizeof_imsm_map(map); | |
296 | } else | |
297 | return map; | |
298 | ||
299 | } | |
cdddbdbc | 300 | |
3393c6af DW |
301 | /* return the size of the device. |
302 | * migr_state increases the returned size if map[0] were to be duplicated | |
303 | */ | |
304 | static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state) | |
a965f303 DW |
305 | { |
306 | size_t size = sizeof(*dev) - sizeof(struct imsm_map) + | |
307 | sizeof_imsm_map(get_imsm_map(dev, 0)); | |
cdddbdbc DW |
308 | |
309 | /* migrating means an additional map */ | |
a965f303 DW |
310 | if (dev->vol.migr_state) |
311 | size += sizeof_imsm_map(get_imsm_map(dev, 1)); | |
3393c6af DW |
312 | else if (migr_state) |
313 | size += sizeof_imsm_map(get_imsm_map(dev, 0)); | |
cdddbdbc DW |
314 | |
315 | return size; | |
316 | } | |
317 | ||
949c47a0 | 318 | static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index) |
cdddbdbc DW |
319 | { |
320 | int offset; | |
321 | int i; | |
322 | void *_mpb = mpb; | |
323 | ||
949c47a0 | 324 | if (index >= mpb->num_raid_devs) |
cdddbdbc DW |
325 | return NULL; |
326 | ||
327 | /* devices start after all disks */ | |
328 | offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb; | |
329 | ||
330 | for (i = 0; i <= index; i++) | |
331 | if (i == index) | |
332 | return _mpb + offset; | |
333 | else | |
3393c6af | 334 | offset += sizeof_imsm_dev(_mpb + offset, 0); |
cdddbdbc DW |
335 | |
336 | return NULL; | |
337 | } | |
338 | ||
949c47a0 DW |
339 | static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index) |
340 | { | |
341 | if (index >= super->anchor->num_raid_devs) | |
342 | return NULL; | |
343 | return super->dev_tbl[index]; | |
344 | } | |
345 | ||
cdddbdbc DW |
346 | static __u32 get_imsm_disk_idx(struct imsm_map *map, int slot) |
347 | { | |
348 | __u32 *ord_tbl = &map->disk_ord_tbl[slot]; | |
349 | ||
7eef0453 DW |
350 | /* top byte identifies disk under rebuild |
351 | * why not just use the USABLE bit... oh well. | |
352 | */ | |
cdddbdbc DW |
353 | return __le32_to_cpu(*ord_tbl & ~(0xff << 24)); |
354 | } | |
355 | ||
7eef0453 DW |
356 | static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev, int slot) |
357 | { | |
358 | struct imsm_map *map; | |
359 | ||
360 | if (dev->vol.migr_state) | |
361 | map = get_imsm_map(dev, 0); | |
362 | else | |
363 | map = get_imsm_map(dev, 1); | |
364 | ||
365 | return map->disk_ord_tbl[slot]; | |
366 | } | |
367 | ||
cdddbdbc DW |
368 | static int get_imsm_raid_level(struct imsm_map *map) |
369 | { | |
370 | if (map->raid_level == 1) { | |
371 | if (map->num_members == 2) | |
372 | return 1; | |
373 | else | |
374 | return 10; | |
375 | } | |
376 | ||
377 | return map->raid_level; | |
378 | } | |
379 | ||
c2c087e6 DW |
380 | static int cmp_extent(const void *av, const void *bv) |
381 | { | |
382 | const struct extent *a = av; | |
383 | const struct extent *b = bv; | |
384 | if (a->start < b->start) | |
385 | return -1; | |
386 | if (a->start > b->start) | |
387 | return 1; | |
388 | return 0; | |
389 | } | |
390 | ||
391 | static struct extent *get_extents(struct intel_super *super, struct dl *dl) | |
392 | { | |
393 | /* find a list of used extents on the given physical device */ | |
c2c087e6 DW |
394 | struct extent *rv, *e; |
395 | int i, j; | |
396 | int memberships = 0; | |
397 | ||
949c47a0 DW |
398 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
399 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
a965f303 | 400 | struct imsm_map *map = get_imsm_map(dev, 0); |
c2c087e6 DW |
401 | |
402 | for (j = 0; j < map->num_members; j++) { | |
403 | __u32 index = get_imsm_disk_idx(map, j); | |
404 | ||
405 | if (index == dl->index) | |
406 | memberships++; | |
407 | } | |
408 | } | |
409 | rv = malloc(sizeof(struct extent) * (memberships + 1)); | |
410 | if (!rv) | |
411 | return NULL; | |
412 | e = rv; | |
413 | ||
949c47a0 DW |
414 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
415 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
a965f303 | 416 | struct imsm_map *map = get_imsm_map(dev, 0); |
c2c087e6 DW |
417 | |
418 | for (j = 0; j < map->num_members; j++) { | |
419 | __u32 index = get_imsm_disk_idx(map, j); | |
420 | ||
421 | if (index == dl->index) { | |
422 | e->start = __le32_to_cpu(map->pba_of_lba0); | |
423 | e->size = __le32_to_cpu(map->blocks_per_member); | |
424 | e++; | |
425 | } | |
426 | } | |
427 | } | |
428 | qsort(rv, memberships, sizeof(*rv), cmp_extent); | |
429 | ||
b9f594fe | 430 | e->start = __le32_to_cpu(dl->disk.total_blocks) - |
c2c087e6 DW |
431 | (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS); |
432 | e->size = 0; | |
433 | return rv; | |
434 | } | |
435 | ||
4f5bc454 | 436 | #ifndef MDASSEMBLE |
cdddbdbc DW |
437 | static void print_imsm_dev(struct imsm_dev *dev, int index) |
438 | { | |
439 | __u64 sz; | |
440 | int slot; | |
a965f303 | 441 | struct imsm_map *map = get_imsm_map(dev, 0); |
cdddbdbc DW |
442 | |
443 | printf("\n"); | |
444 | printf("[%s]:\n", dev->volume); | |
445 | printf(" RAID Level : %d\n", get_imsm_raid_level(map)); | |
446 | printf(" Members : %d\n", map->num_members); | |
447 | for (slot = 0; slot < map->num_members; slot++) | |
448 | if (index == get_imsm_disk_idx(map, slot)) | |
449 | break; | |
450 | if (slot < map->num_members) | |
451 | printf(" This Slot : %d\n", slot); | |
452 | else | |
453 | printf(" This Slot : ?\n"); | |
454 | sz = __le32_to_cpu(dev->size_high); | |
455 | sz <<= 32; | |
456 | sz += __le32_to_cpu(dev->size_low); | |
457 | printf(" Array Size : %llu%s\n", (unsigned long long)sz, | |
458 | human_size(sz * 512)); | |
459 | sz = __le32_to_cpu(map->blocks_per_member); | |
460 | printf(" Per Dev Size : %llu%s\n", (unsigned long long)sz, | |
461 | human_size(sz * 512)); | |
462 | printf(" Sector Offset : %u\n", | |
463 | __le32_to_cpu(map->pba_of_lba0)); | |
464 | printf(" Num Stripes : %u\n", | |
465 | __le32_to_cpu(map->num_data_stripes)); | |
466 | printf(" Chunk Size : %u KiB\n", | |
467 | __le16_to_cpu(map->blocks_per_strip) / 2); | |
468 | printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks)); | |
3393c6af DW |
469 | printf(" Migrate State : %s", dev->vol.migr_state ? "migrating" : "idle"); |
470 | if (dev->vol.migr_state) | |
471 | printf(": %s", dev->vol.migr_type ? "rebuilding" : "initializing"); | |
472 | printf("\n"); | |
473 | printf(" Map State : %s", map_state_str[map->map_state]); | |
474 | if (dev->vol.migr_state) { | |
475 | struct imsm_map *map = get_imsm_map(dev, 1); | |
476 | printf(", %s", map_state_str[map->map_state]); | |
477 | } | |
478 | printf("\n"); | |
cdddbdbc | 479 | printf(" Dirty State : %s\n", dev->vol.dirty ? "dirty" : "clean"); |
cdddbdbc DW |
480 | } |
481 | ||
482 | static void print_imsm_disk(struct imsm_super *mpb, int index) | |
483 | { | |
949c47a0 | 484 | struct imsm_disk *disk = __get_imsm_disk(mpb, index); |
cdddbdbc DW |
485 | char str[MAX_RAID_SERIAL_LEN]; |
486 | __u32 s; | |
487 | __u64 sz; | |
488 | ||
e9d82038 DW |
489 | if (index < 0) |
490 | return; | |
491 | ||
cdddbdbc DW |
492 | printf("\n"); |
493 | snprintf(str, MAX_RAID_SERIAL_LEN, "%s", disk->serial); | |
494 | printf(" Disk%02d Serial : %s\n", index, str); | |
495 | s = __le32_to_cpu(disk->status); | |
496 | printf(" State :%s%s%s%s\n", s&SPARE_DISK ? " spare" : "", | |
497 | s&CONFIGURED_DISK ? " active" : "", | |
498 | s&FAILED_DISK ? " failed" : "", | |
499 | s&USABLE_DISK ? " usable" : ""); | |
500 | printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id)); | |
c2c087e6 DW |
501 | sz = __le32_to_cpu(disk->total_blocks) - |
502 | (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS * mpb->num_raid_devs); | |
cdddbdbc DW |
503 | printf(" Usable Size : %llu%s\n", (unsigned long long)sz, |
504 | human_size(sz * 512)); | |
505 | } | |
506 | ||
507 | static void examine_super_imsm(struct supertype *st, char *homehost) | |
508 | { | |
509 | struct intel_super *super = st->sb; | |
949c47a0 | 510 | struct imsm_super *mpb = super->anchor; |
cdddbdbc DW |
511 | char str[MAX_SIGNATURE_LENGTH]; |
512 | int i; | |
513 | __u32 sum; | |
514 | ||
515 | snprintf(str, MPB_SIG_LEN, "%s", mpb->sig); | |
516 | printf(" Magic : %s\n", str); | |
517 | snprintf(str, strlen(MPB_VERSION_RAID0), "%s", get_imsm_version(mpb)); | |
518 | printf(" Version : %s\n", get_imsm_version(mpb)); | |
519 | printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num)); | |
520 | printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num)); | |
521 | sum = __le32_to_cpu(mpb->check_sum); | |
522 | printf(" Checksum : %08x %s\n", sum, | |
949c47a0 | 523 | __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect"); |
87eb16df | 524 | printf(" MPB Sectors : %d\n", mpb_sectors(mpb)); |
cdddbdbc DW |
525 | printf(" Disks : %d\n", mpb->num_disks); |
526 | printf(" RAID Devices : %d\n", mpb->num_raid_devs); | |
527 | print_imsm_disk(mpb, super->disks->index); | |
604b746f JD |
528 | if (super->bbm_log) { |
529 | struct bbm_log *log = super->bbm_log; | |
530 | ||
531 | printf("\n"); | |
532 | printf("Bad Block Management Log:\n"); | |
533 | printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size)); | |
534 | printf(" Signature : %x\n", __le32_to_cpu(log->signature)); | |
535 | printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count)); | |
536 | printf(" Spare Blocks : %d\n", __le32_to_cpu(log->reserved_spare_block_count)); | |
537 | printf(" First Spare : %llx\n", __le64_to_cpu(log->first_spare_lba)); | |
538 | } | |
cdddbdbc | 539 | for (i = 0; i < mpb->num_raid_devs; i++) |
949c47a0 | 540 | print_imsm_dev(__get_imsm_dev(mpb, i), super->disks->index); |
cdddbdbc DW |
541 | for (i = 0; i < mpb->num_disks; i++) { |
542 | if (i == super->disks->index) | |
543 | continue; | |
544 | print_imsm_disk(mpb, i); | |
545 | } | |
546 | } | |
547 | ||
548 | static void brief_examine_super_imsm(struct supertype *st) | |
549 | { | |
828408eb | 550 | printf("ARRAY /dev/imsm metadata=imsm\n"); |
cdddbdbc DW |
551 | } |
552 | ||
553 | static void detail_super_imsm(struct supertype *st, char *homehost) | |
554 | { | |
555 | printf("%s\n", __FUNCTION__); | |
556 | } | |
557 | ||
558 | static void brief_detail_super_imsm(struct supertype *st) | |
559 | { | |
560 | printf("%s\n", __FUNCTION__); | |
561 | } | |
562 | #endif | |
563 | ||
564 | static int match_home_imsm(struct supertype *st, char *homehost) | |
565 | { | |
566 | printf("%s\n", __FUNCTION__); | |
567 | ||
568 | return 0; | |
569 | } | |
570 | ||
571 | static void uuid_from_super_imsm(struct supertype *st, int uuid[4]) | |
572 | { | |
43dad3d6 DW |
573 | /* imsm does not track uuid's so just make sure we never return |
574 | * the same value twice to break uuid matching in Manage_subdevs | |
575 | * FIXME what about the use of uuid's with bitmap's? | |
576 | */ | |
577 | static int dummy_id = 0; | |
578 | ||
579 | uuid[0] = dummy_id++; | |
cdddbdbc DW |
580 | } |
581 | ||
0d481d37 | 582 | #if 0 |
4f5bc454 DW |
583 | static void |
584 | get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p) | |
cdddbdbc | 585 | { |
cdddbdbc DW |
586 | __u8 *v = get_imsm_version(mpb); |
587 | __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH; | |
588 | char major[] = { 0, 0, 0 }; | |
589 | char minor[] = { 0 ,0, 0 }; | |
590 | char patch[] = { 0, 0, 0 }; | |
591 | char *ver_parse[] = { major, minor, patch }; | |
592 | int i, j; | |
593 | ||
594 | i = j = 0; | |
595 | while (*v != '\0' && v < end) { | |
596 | if (*v != '.' && j < 2) | |
597 | ver_parse[i][j++] = *v; | |
598 | else { | |
599 | i++; | |
600 | j = 0; | |
601 | } | |
602 | v++; | |
603 | } | |
604 | ||
4f5bc454 DW |
605 | *m = strtol(minor, NULL, 0); |
606 | *p = strtol(patch, NULL, 0); | |
607 | } | |
0d481d37 | 608 | #endif |
4f5bc454 | 609 | |
c2c087e6 DW |
610 | static int imsm_level_to_layout(int level) |
611 | { | |
612 | switch (level) { | |
613 | case 0: | |
614 | case 1: | |
615 | return 0; | |
616 | case 5: | |
617 | case 6: | |
a380c027 | 618 | return ALGORITHM_LEFT_ASYMMETRIC; |
c2c087e6 DW |
619 | case 10: |
620 | return 0x102; //FIXME is this correct? | |
621 | } | |
622 | return -1; | |
623 | } | |
624 | ||
bf5a934a DW |
625 | static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info) |
626 | { | |
627 | struct intel_super *super = st->sb; | |
949c47a0 | 628 | struct imsm_dev *dev = get_imsm_dev(super, super->current_vol); |
a965f303 | 629 | struct imsm_map *map = get_imsm_map(dev, 0); |
bf5a934a DW |
630 | |
631 | info->container_member = super->current_vol; | |
632 | info->array.raid_disks = map->num_members; | |
633 | info->array.level = get_imsm_raid_level(map); | |
634 | info->array.layout = imsm_level_to_layout(info->array.level); | |
635 | info->array.md_minor = -1; | |
636 | info->array.ctime = 0; | |
637 | info->array.utime = 0; | |
638 | info->array.chunk_size = __le16_to_cpu(map->blocks_per_strip * 512); | |
639 | ||
640 | info->data_offset = __le32_to_cpu(map->pba_of_lba0); | |
641 | info->component_size = __le32_to_cpu(map->blocks_per_member); | |
642 | ||
643 | info->disk.major = 0; | |
644 | info->disk.minor = 0; | |
645 | ||
646 | sprintf(info->text_version, "/%s/%d", | |
647 | devnum2devname(st->container_dev), | |
648 | info->container_member); | |
649 | } | |
650 | ||
651 | ||
4f5bc454 DW |
652 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info) |
653 | { | |
654 | struct intel_super *super = st->sb; | |
4f5bc454 DW |
655 | struct imsm_disk *disk; |
656 | __u32 s; | |
4f5bc454 | 657 | |
bf5a934a DW |
658 | if (super->current_vol >= 0) { |
659 | getinfo_super_imsm_volume(st, info); | |
660 | return; | |
661 | } | |
d23fe947 DW |
662 | |
663 | /* Set raid_disks to zero so that Assemble will always pull in valid | |
664 | * spares | |
665 | */ | |
666 | info->array.raid_disks = 0; | |
cdddbdbc DW |
667 | info->array.level = LEVEL_CONTAINER; |
668 | info->array.layout = 0; | |
669 | info->array.md_minor = -1; | |
c2c087e6 | 670 | info->array.ctime = 0; /* N/A for imsm */ |
cdddbdbc DW |
671 | info->array.utime = 0; |
672 | info->array.chunk_size = 0; | |
673 | ||
674 | info->disk.major = 0; | |
675 | info->disk.minor = 0; | |
cdddbdbc | 676 | info->disk.raid_disk = -1; |
c2c087e6 DW |
677 | info->reshape_active = 0; |
678 | strcpy(info->text_version, "imsm"); | |
679 | info->disk.number = -1; | |
680 | info->disk.state = 0; | |
681 | ||
4a04ec6c | 682 | if (super->disks) { |
b9f594fe | 683 | disk = &super->disks->disk; |
4a04ec6c DW |
684 | info->disk.number = super->disks->index; |
685 | info->disk.raid_disk = super->disks->index; | |
bf5a934a DW |
686 | info->data_offset = __le32_to_cpu(disk->total_blocks) - |
687 | (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS); | |
688 | info->component_size = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
4a04ec6c DW |
689 | s = __le32_to_cpu(disk->status); |
690 | info->disk.state = s & CONFIGURED_DISK ? (1 << MD_DISK_ACTIVE) : 0; | |
691 | info->disk.state |= s & FAILED_DISK ? (1 << MD_DISK_FAULTY) : 0; | |
692 | info->disk.state |= s & USABLE_DISK ? (1 << MD_DISK_SYNC) : 0; | |
cdddbdbc | 693 | } |
cdddbdbc DW |
694 | } |
695 | ||
cdddbdbc DW |
696 | static int update_super_imsm(struct supertype *st, struct mdinfo *info, |
697 | char *update, char *devname, int verbose, | |
698 | int uuid_set, char *homehost) | |
699 | { | |
f352c545 DW |
700 | /* FIXME */ |
701 | ||
702 | /* For 'assemble' and 'force' we need to return non-zero if any | |
703 | * change was made. For others, the return value is ignored. | |
704 | * Update options are: | |
705 | * force-one : This device looks a bit old but needs to be included, | |
706 | * update age info appropriately. | |
707 | * assemble: clear any 'faulty' flag to allow this device to | |
708 | * be assembled. | |
709 | * force-array: Array is degraded but being forced, mark it clean | |
710 | * if that will be needed to assemble it. | |
711 | * | |
712 | * newdev: not used ???? | |
713 | * grow: Array has gained a new device - this is currently for | |
714 | * linear only | |
715 | * resync: mark as dirty so a resync will happen. | |
716 | * name: update the name - preserving the homehost | |
717 | * | |
718 | * Following are not relevant for this imsm: | |
719 | * sparc2.2 : update from old dodgey metadata | |
720 | * super-minor: change the preferred_minor number | |
721 | * summaries: update redundant counters. | |
722 | * uuid: Change the uuid of the array to match watch is given | |
723 | * homehost: update the recorded homehost | |
724 | * _reshape_progress: record new reshape_progress position. | |
725 | */ | |
726 | int rv = 0; | |
727 | //struct intel_super *super = st->sb; | |
728 | //struct imsm_super *mpb = super->mpb; | |
729 | ||
730 | if (strcmp(update, "grow") == 0) { | |
731 | } | |
732 | if (strcmp(update, "resync") == 0) { | |
733 | /* dev->vol.dirty = 1; */ | |
734 | } | |
735 | ||
736 | /* IMSM has no concept of UUID or homehost */ | |
737 | ||
738 | return rv; | |
cdddbdbc DW |
739 | } |
740 | ||
c2c087e6 | 741 | static size_t disks_to_mpb_size(int disks) |
cdddbdbc | 742 | { |
c2c087e6 | 743 | size_t size; |
cdddbdbc | 744 | |
c2c087e6 DW |
745 | size = sizeof(struct imsm_super); |
746 | size += (disks - 1) * sizeof(struct imsm_disk); | |
747 | size += 2 * sizeof(struct imsm_dev); | |
748 | /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */ | |
749 | size += (4 - 2) * sizeof(struct imsm_map); | |
750 | /* 4 possible disk_ord_tbl's */ | |
751 | size += 4 * (disks - 1) * sizeof(__u32); | |
752 | ||
753 | return size; | |
754 | } | |
755 | ||
756 | static __u64 avail_size_imsm(struct supertype *st, __u64 devsize) | |
757 | { | |
758 | if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS)) | |
759 | return 0; | |
760 | ||
761 | return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS); | |
cdddbdbc DW |
762 | } |
763 | ||
764 | static int compare_super_imsm(struct supertype *st, struct supertype *tst) | |
765 | { | |
766 | /* | |
767 | * return: | |
768 | * 0 same, or first was empty, and second was copied | |
769 | * 1 second had wrong number | |
770 | * 2 wrong uuid | |
771 | * 3 wrong other info | |
772 | */ | |
773 | struct intel_super *first = st->sb; | |
774 | struct intel_super *sec = tst->sb; | |
775 | ||
776 | if (!first) { | |
777 | st->sb = tst->sb; | |
778 | tst->sb = NULL; | |
779 | return 0; | |
780 | } | |
781 | ||
949c47a0 | 782 | if (memcmp(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH) != 0) |
cdddbdbc | 783 | return 3; |
d23fe947 DW |
784 | |
785 | /* if an anchor does not have num_raid_devs set then it is a free | |
786 | * floating spare | |
787 | */ | |
788 | if (first->anchor->num_raid_devs > 0 && | |
789 | sec->anchor->num_raid_devs > 0) { | |
790 | if (first->anchor->family_num != sec->anchor->family_num) | |
791 | return 3; | |
792 | if (first->anchor->mpb_size != sec->anchor->mpb_size) | |
793 | return 3; | |
794 | if (first->anchor->check_sum != sec->anchor->check_sum) | |
795 | return 3; | |
796 | } | |
cdddbdbc DW |
797 | |
798 | return 0; | |
799 | } | |
800 | ||
0030e8d6 DW |
801 | static void fd2devname(int fd, char *name) |
802 | { | |
803 | struct stat st; | |
804 | char path[256]; | |
805 | char dname[100]; | |
806 | char *nm; | |
807 | int rv; | |
808 | ||
809 | name[0] = '\0'; | |
810 | if (fstat(fd, &st) != 0) | |
811 | return; | |
812 | sprintf(path, "/sys/dev/block/%d:%d", | |
813 | major(st.st_rdev), minor(st.st_rdev)); | |
814 | ||
815 | rv = readlink(path, dname, sizeof(dname)); | |
816 | if (rv <= 0) | |
817 | return; | |
818 | ||
819 | dname[rv] = '\0'; | |
820 | nm = strrchr(dname, '/'); | |
821 | nm++; | |
822 | snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm); | |
823 | } | |
824 | ||
825 | ||
cdddbdbc DW |
826 | extern int scsi_get_serial(int fd, void *buf, size_t buf_len); |
827 | ||
828 | static int imsm_read_serial(int fd, char *devname, | |
829 | __u8 serial[MAX_RAID_SERIAL_LEN]) | |
830 | { | |
831 | unsigned char scsi_serial[255]; | |
cdddbdbc DW |
832 | int rv; |
833 | int rsp_len; | |
834 | int i, cnt; | |
835 | ||
836 | memset(scsi_serial, 0, sizeof(scsi_serial)); | |
cdddbdbc | 837 | |
0030e8d6 DW |
838 | if (imsm_env_devname_as_serial()) { |
839 | char name[MAX_RAID_SERIAL_LEN]; | |
840 | ||
841 | fd2devname(fd, name); | |
842 | strcpy((char *) serial, name); | |
843 | return 0; | |
844 | } | |
845 | ||
755c99fa | 846 | rv = scsi_get_serial(fd, scsi_serial, sizeof(scsi_serial)); |
cdddbdbc DW |
847 | |
848 | if (rv != 0) { | |
849 | if (devname) | |
850 | fprintf(stderr, | |
851 | Name ": Failed to retrieve serial for %s\n", | |
852 | devname); | |
853 | return rv; | |
854 | } | |
855 | ||
856 | rsp_len = scsi_serial[3]; | |
857 | for (i = 0, cnt = 0; i < rsp_len; i++) { | |
858 | if (!isspace(scsi_serial[4 + i])) | |
859 | serial[cnt++] = scsi_serial[4 + i]; | |
860 | if (cnt == MAX_RAID_SERIAL_LEN) | |
861 | break; | |
862 | } | |
863 | ||
864 | serial[MAX_RAID_SERIAL_LEN - 1] = '\0'; | |
865 | ||
866 | return 0; | |
867 | } | |
868 | ||
869 | static int | |
870 | load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd) | |
871 | { | |
cdddbdbc DW |
872 | struct dl *dl; |
873 | struct stat stb; | |
cdddbdbc DW |
874 | int rv; |
875 | int i; | |
d23fe947 DW |
876 | int alloc = 1; |
877 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
878 | ||
879 | rv = imsm_read_serial(fd, devname, serial); | |
880 | ||
881 | if (rv != 0) | |
882 | return 2; | |
883 | ||
884 | /* check if this is a disk we have seen before. it may be a spare in | |
885 | * super->disks while the current anchor believes it is a raid member, | |
886 | * check if we need to update dl->index | |
887 | */ | |
888 | for (dl = super->disks; dl; dl = dl->next) | |
889 | if (memcmp(dl->serial, serial, MAX_RAID_SERIAL_LEN) == 0) | |
890 | break; | |
891 | ||
892 | if (!dl) | |
893 | dl = malloc(sizeof(*dl)); | |
894 | else | |
895 | alloc = 0; | |
cdddbdbc | 896 | |
b9f594fe | 897 | if (!dl) { |
cdddbdbc DW |
898 | if (devname) |
899 | fprintf(stderr, | |
900 | Name ": failed to allocate disk buffer for %s\n", | |
901 | devname); | |
902 | return 2; | |
903 | } | |
cdddbdbc | 904 | |
d23fe947 DW |
905 | if (alloc) { |
906 | fstat(fd, &stb); | |
907 | dl->major = major(stb.st_rdev); | |
908 | dl->minor = minor(stb.st_rdev); | |
909 | dl->next = super->disks; | |
910 | dl->fd = keep_fd ? fd : -1; | |
911 | dl->devname = devname ? strdup(devname) : NULL; | |
912 | strncpy((char *) dl->serial, (char *) serial, MAX_RAID_SERIAL_LEN); | |
913 | } else if (keep_fd) { | |
914 | close(dl->fd); | |
915 | dl->fd = fd; | |
916 | } | |
cdddbdbc | 917 | |
d23fe947 | 918 | /* look up this disk's index in the current anchor */ |
949c47a0 DW |
919 | for (i = 0; i < super->anchor->num_disks; i++) { |
920 | struct imsm_disk *disk_iter; | |
921 | ||
922 | disk_iter = __get_imsm_disk(super->anchor, i); | |
cdddbdbc | 923 | |
949c47a0 DW |
924 | if (memcmp(disk_iter->serial, dl->serial, |
925 | MAX_RAID_SERIAL_LEN) == 0) { | |
d23fe947 DW |
926 | __u32 status; |
927 | ||
b9f594fe | 928 | dl->disk = *disk_iter; |
d23fe947 DW |
929 | status = __le32_to_cpu(dl->disk.status); |
930 | /* only set index on disks that are a member of a | |
931 | * populated contianer, i.e. one with raid_devs | |
932 | */ | |
933 | if (status & SPARE_DISK) | |
934 | dl->index = -1; | |
935 | else | |
936 | dl->index = i; | |
cdddbdbc | 937 | break; |
949c47a0 | 938 | } |
cdddbdbc DW |
939 | } |
940 | ||
d23fe947 | 941 | if (i == super->anchor->num_disks && alloc) { |
e9d82038 DW |
942 | if (devname) |
943 | fprintf(stderr, | |
d23fe947 | 944 | Name ": failed to load disk with serial \'%s\' for %s\n", |
e9d82038 | 945 | dl->serial, devname); |
d23fe947 DW |
946 | free(dl); |
947 | return 1; | |
948 | } | |
949 | if (i == super->anchor->num_disks && dl->index >= 0) { | |
950 | if (devname) | |
951 | fprintf(stderr, | |
952 | Name ": confused... disk %d with serial \'%s\' " | |
953 | "is not listed in the current anchor\n", | |
954 | dl->index, dl->serial); | |
955 | return 1; | |
e9d82038 | 956 | } |
cdddbdbc | 957 | |
d23fe947 DW |
958 | if (alloc) |
959 | super->disks = dl; | |
960 | ||
949c47a0 DW |
961 | return 0; |
962 | } | |
963 | ||
964 | static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src) | |
965 | { | |
3393c6af DW |
966 | memcpy(dest, src, sizeof_imsm_dev(src, 0)); |
967 | } | |
968 | ||
969 | static void dup_map(struct imsm_dev *dev) | |
970 | { | |
971 | struct imsm_map *dest = get_imsm_map(dev, 1); | |
972 | struct imsm_map *src = get_imsm_map(dev, 0); | |
973 | ||
974 | memcpy(dest, src, sizeof_imsm_map(src)); | |
949c47a0 DW |
975 | } |
976 | ||
977 | static int parse_raid_devices(struct intel_super *super) | |
978 | { | |
979 | int i; | |
980 | struct imsm_dev *dev_new; | |
981 | size_t len; | |
982 | ||
983 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
984 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
985 | ||
3393c6af | 986 | len = sizeof_imsm_dev(dev_iter, 1); |
949c47a0 DW |
987 | dev_new = malloc(len); |
988 | if (!dev_new) | |
989 | return 1; | |
990 | imsm_copy_dev(dev_new, dev_iter); | |
991 | super->dev_tbl[i] = dev_new; | |
992 | } | |
cdddbdbc DW |
993 | |
994 | return 0; | |
995 | } | |
996 | ||
604b746f JD |
997 | /* retrieve a pointer to the bbm log which starts after all raid devices */ |
998 | struct bbm_log *__get_imsm_bbm_log(struct imsm_super *mpb) | |
999 | { | |
1000 | void *ptr = NULL; | |
1001 | ||
1002 | if (__le32_to_cpu(mpb->bbm_log_size)) { | |
1003 | ptr = mpb; | |
1004 | ptr += mpb->mpb_size - __le32_to_cpu(mpb->bbm_log_size); | |
1005 | } | |
1006 | ||
1007 | return ptr; | |
1008 | } | |
1009 | ||
d23fe947 | 1010 | static void __free_imsm(struct intel_super *super, int free_disks); |
9ca2c81c | 1011 | |
cdddbdbc DW |
1012 | /* load_imsm_mpb - read matrix metadata |
1013 | * allocates super->mpb to be freed by free_super | |
1014 | */ | |
1015 | static int load_imsm_mpb(int fd, struct intel_super *super, char *devname) | |
1016 | { | |
1017 | unsigned long long dsize; | |
cdddbdbc DW |
1018 | unsigned long long sectors; |
1019 | struct stat; | |
6416d527 | 1020 | struct imsm_super *anchor; |
cdddbdbc | 1021 | __u32 check_sum; |
949c47a0 | 1022 | int rc; |
cdddbdbc | 1023 | |
cdddbdbc DW |
1024 | get_dev_size(fd, NULL, &dsize); |
1025 | ||
1026 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) { | |
1027 | if (devname) | |
1028 | fprintf(stderr, | |
1029 | Name ": Cannot seek to anchor block on %s: %s\n", | |
1030 | devname, strerror(errno)); | |
1031 | return 1; | |
1032 | } | |
1033 | ||
949c47a0 | 1034 | if (posix_memalign((void**)&anchor, 512, 512) != 0) { |
ad97895e DW |
1035 | if (devname) |
1036 | fprintf(stderr, | |
1037 | Name ": Failed to allocate imsm anchor buffer" | |
1038 | " on %s\n", devname); | |
1039 | return 1; | |
1040 | } | |
949c47a0 | 1041 | if (read(fd, anchor, 512) != 512) { |
cdddbdbc DW |
1042 | if (devname) |
1043 | fprintf(stderr, | |
1044 | Name ": Cannot read anchor block on %s: %s\n", | |
1045 | devname, strerror(errno)); | |
6416d527 | 1046 | free(anchor); |
cdddbdbc DW |
1047 | return 1; |
1048 | } | |
1049 | ||
6416d527 | 1050 | if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) { |
cdddbdbc DW |
1051 | if (devname) |
1052 | fprintf(stderr, | |
1053 | Name ": no IMSM anchor on %s\n", devname); | |
6416d527 | 1054 | free(anchor); |
cdddbdbc DW |
1055 | return 2; |
1056 | } | |
1057 | ||
d23fe947 | 1058 | __free_imsm(super, 0); |
949c47a0 DW |
1059 | super->len = __le32_to_cpu(anchor->mpb_size); |
1060 | super->len = ROUND_UP(anchor->mpb_size, 512); | |
1061 | if (posix_memalign(&super->buf, 512, super->len) != 0) { | |
cdddbdbc DW |
1062 | if (devname) |
1063 | fprintf(stderr, | |
1064 | Name ": unable to allocate %zu byte mpb buffer\n", | |
949c47a0 | 1065 | super->len); |
6416d527 | 1066 | free(anchor); |
cdddbdbc DW |
1067 | return 2; |
1068 | } | |
949c47a0 | 1069 | memcpy(super->buf, anchor, 512); |
cdddbdbc | 1070 | |
6416d527 NB |
1071 | sectors = mpb_sectors(anchor) - 1; |
1072 | free(anchor); | |
949c47a0 DW |
1073 | if (!sectors) { |
1074 | rc = load_imsm_disk(fd, super, devname, 0); | |
1075 | if (rc == 0) | |
1076 | rc = parse_raid_devices(super); | |
1077 | return rc; | |
1078 | } | |
cdddbdbc DW |
1079 | |
1080 | /* read the extended mpb */ | |
1081 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) { | |
1082 | if (devname) | |
1083 | fprintf(stderr, | |
1084 | Name ": Cannot seek to extended mpb on %s: %s\n", | |
1085 | devname, strerror(errno)); | |
1086 | return 1; | |
1087 | } | |
1088 | ||
949c47a0 | 1089 | if (read(fd, super->buf + 512, super->len - 512) != super->len - 512) { |
cdddbdbc DW |
1090 | if (devname) |
1091 | fprintf(stderr, | |
1092 | Name ": Cannot read extended mpb on %s: %s\n", | |
1093 | devname, strerror(errno)); | |
1094 | return 2; | |
1095 | } | |
1096 | ||
949c47a0 DW |
1097 | check_sum = __gen_imsm_checksum(super->anchor); |
1098 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
cdddbdbc DW |
1099 | if (devname) |
1100 | fprintf(stderr, | |
1101 | Name ": IMSM checksum %x != %x on %s\n", | |
949c47a0 | 1102 | check_sum, __le32_to_cpu(super->anchor->check_sum), |
cdddbdbc DW |
1103 | devname); |
1104 | return 2; | |
1105 | } | |
1106 | ||
604b746f JD |
1107 | /* FIXME the BBM log is disk specific so we cannot use this global |
1108 | * buffer for all disks. Ok for now since we only look at the global | |
1109 | * bbm_log_size parameter to gate assembly | |
1110 | */ | |
1111 | super->bbm_log = __get_imsm_bbm_log(super->anchor); | |
1112 | ||
949c47a0 DW |
1113 | rc = load_imsm_disk(fd, super, devname, 0); |
1114 | if (rc == 0) | |
1115 | rc = parse_raid_devices(super); | |
1116 | return rc; | |
cdddbdbc DW |
1117 | } |
1118 | ||
ae6aad82 DW |
1119 | static void __free_imsm_disk(struct dl *d) |
1120 | { | |
1121 | if (d->fd >= 0) | |
1122 | close(d->fd); | |
1123 | if (d->devname) | |
1124 | free(d->devname); | |
1125 | free(d); | |
1126 | ||
1127 | } | |
cdddbdbc DW |
1128 | static void free_imsm_disks(struct intel_super *super) |
1129 | { | |
1130 | while (super->disks) { | |
1131 | struct dl *d = super->disks; | |
1132 | ||
1133 | super->disks = d->next; | |
ae6aad82 | 1134 | __free_imsm_disk(d); |
cdddbdbc DW |
1135 | } |
1136 | } | |
1137 | ||
9ca2c81c | 1138 | /* free all the pieces hanging off of a super pointer */ |
d23fe947 | 1139 | static void __free_imsm(struct intel_super *super, int free_disks) |
cdddbdbc | 1140 | { |
949c47a0 DW |
1141 | int i; |
1142 | ||
9ca2c81c | 1143 | if (super->buf) { |
949c47a0 | 1144 | free(super->buf); |
9ca2c81c DW |
1145 | super->buf = NULL; |
1146 | } | |
d23fe947 DW |
1147 | if (free_disks) |
1148 | free_imsm_disks(super); | |
949c47a0 | 1149 | for (i = 0; i < IMSM_MAX_RAID_DEVS; i++) |
9ca2c81c | 1150 | if (super->dev_tbl[i]) { |
949c47a0 | 1151 | free(super->dev_tbl[i]); |
9ca2c81c DW |
1152 | super->dev_tbl[i] = NULL; |
1153 | } | |
cdddbdbc DW |
1154 | } |
1155 | ||
9ca2c81c DW |
1156 | static void free_imsm(struct intel_super *super) |
1157 | { | |
d23fe947 | 1158 | __free_imsm(super, 1); |
9ca2c81c DW |
1159 | free(super); |
1160 | } | |
cdddbdbc DW |
1161 | |
1162 | static void free_super_imsm(struct supertype *st) | |
1163 | { | |
1164 | struct intel_super *super = st->sb; | |
1165 | ||
1166 | if (!super) | |
1167 | return; | |
1168 | ||
1169 | free_imsm(super); | |
1170 | st->sb = NULL; | |
1171 | } | |
1172 | ||
c2c087e6 DW |
1173 | static struct intel_super *alloc_super(int creating_imsm) |
1174 | { | |
1175 | struct intel_super *super = malloc(sizeof(*super)); | |
1176 | ||
1177 | if (super) { | |
1178 | memset(super, 0, sizeof(*super)); | |
1179 | super->creating_imsm = creating_imsm; | |
bf5a934a | 1180 | super->current_vol = -1; |
c2c087e6 DW |
1181 | } |
1182 | ||
1183 | return super; | |
1184 | } | |
1185 | ||
cdddbdbc DW |
1186 | #ifndef MDASSEMBLE |
1187 | static int load_super_imsm_all(struct supertype *st, int fd, void **sbp, | |
1188 | char *devname, int keep_fd) | |
1189 | { | |
1190 | struct mdinfo *sra; | |
1191 | struct intel_super *super; | |
1192 | struct mdinfo *sd, *best = NULL; | |
1193 | __u32 bestgen = 0; | |
1194 | __u32 gen; | |
1195 | char nm[20]; | |
1196 | int dfd; | |
1197 | int rv; | |
1198 | ||
1199 | /* check if this disk is a member of an active array */ | |
1200 | sra = sysfs_read(fd, 0, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE); | |
1201 | if (!sra) | |
1202 | return 1; | |
1203 | ||
1204 | if (sra->array.major_version != -1 || | |
1205 | sra->array.minor_version != -2 || | |
1206 | strcmp(sra->text_version, "imsm") != 0) | |
1207 | return 1; | |
1208 | ||
c2c087e6 | 1209 | super = alloc_super(0); |
cdddbdbc DW |
1210 | if (!super) |
1211 | return 1; | |
1212 | ||
d23fe947 | 1213 | /* find the most up to date disk in this array, skipping spares */ |
cdddbdbc DW |
1214 | for (sd = sra->devs; sd; sd = sd->next) { |
1215 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); | |
1216 | dfd = dev_open(nm, keep_fd ? O_RDWR : O_RDONLY); | |
1217 | if (!dfd) { | |
1218 | free_imsm(super); | |
1219 | return 2; | |
1220 | } | |
1221 | rv = load_imsm_mpb(dfd, super, NULL); | |
1222 | if (!keep_fd) | |
1223 | close(dfd); | |
1224 | if (rv == 0) { | |
d23fe947 DW |
1225 | if (super->anchor->num_raid_devs == 0) |
1226 | gen = 0; | |
1227 | else | |
1228 | gen = __le32_to_cpu(super->anchor->generation_num); | |
cdddbdbc DW |
1229 | if (!best || gen > bestgen) { |
1230 | bestgen = gen; | |
1231 | best = sd; | |
1232 | } | |
1233 | } else { | |
1234 | free_imsm(super); | |
1235 | return 2; | |
1236 | } | |
1237 | } | |
1238 | ||
1239 | if (!best) { | |
1240 | free_imsm(super); | |
1241 | return 1; | |
1242 | } | |
1243 | ||
1244 | /* load the most up to date anchor */ | |
1245 | sprintf(nm, "%d:%d", best->disk.major, best->disk.minor); | |
1246 | dfd = dev_open(nm, O_RDONLY); | |
1247 | if (!dfd) { | |
1248 | free_imsm(super); | |
1249 | return 1; | |
1250 | } | |
1251 | rv = load_imsm_mpb(dfd, super, NULL); | |
1252 | close(dfd); | |
1253 | if (rv != 0) { | |
1254 | free_imsm(super); | |
1255 | return 2; | |
1256 | } | |
1257 | ||
d23fe947 | 1258 | /* re-parse the disk list with the current anchor */ |
cdddbdbc DW |
1259 | for (sd = sra->devs ; sd ; sd = sd->next) { |
1260 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); | |
1261 | dfd = dev_open(nm, keep_fd? O_RDWR : O_RDONLY); | |
1262 | if (!dfd) { | |
1263 | free_imsm(super); | |
1264 | return 2; | |
1265 | } | |
1266 | load_imsm_disk(dfd, super, NULL, keep_fd); | |
1267 | if (!keep_fd) | |
1268 | close(dfd); | |
1269 | } | |
1270 | ||
f7e7067b | 1271 | if (st->subarray[0]) { |
949c47a0 | 1272 | if (atoi(st->subarray) <= super->anchor->num_raid_devs) |
bf5a934a DW |
1273 | super->current_vol = atoi(st->subarray); |
1274 | else | |
1275 | return 1; | |
f7e7067b NB |
1276 | } |
1277 | ||
cdddbdbc | 1278 | *sbp = super; |
43dad3d6 | 1279 | st->container_dev = fd2devnum(fd); |
cdddbdbc | 1280 | if (st->ss == NULL) { |
bf5a934a | 1281 | st->ss = &super_imsm; |
cdddbdbc DW |
1282 | st->minor_version = 0; |
1283 | st->max_devs = IMSM_MAX_DEVICES; | |
1284 | } | |
1285 | ||
1286 | return 0; | |
1287 | } | |
1288 | #endif | |
1289 | ||
1290 | static int load_super_imsm(struct supertype *st, int fd, char *devname) | |
1291 | { | |
1292 | struct intel_super *super; | |
1293 | int rv; | |
1294 | ||
1295 | #ifndef MDASSEMBLE | |
3dbccbcf | 1296 | if (load_super_imsm_all(st, fd, &st->sb, devname, 1) == 0) |
cdddbdbc DW |
1297 | return 0; |
1298 | #endif | |
f7e7067b NB |
1299 | if (st->subarray[0]) |
1300 | return 1; /* FIXME */ | |
cdddbdbc | 1301 | |
c2c087e6 | 1302 | super = alloc_super(0); |
cdddbdbc DW |
1303 | if (!super) { |
1304 | fprintf(stderr, | |
1305 | Name ": malloc of %zu failed.\n", | |
1306 | sizeof(*super)); | |
1307 | return 1; | |
1308 | } | |
1309 | ||
1310 | rv = load_imsm_mpb(fd, super, devname); | |
1311 | ||
1312 | if (rv) { | |
1313 | if (devname) | |
1314 | fprintf(stderr, | |
1315 | Name ": Failed to load all information " | |
1316 | "sections on %s\n", devname); | |
1317 | free_imsm(super); | |
1318 | return rv; | |
1319 | } | |
1320 | ||
1321 | st->sb = super; | |
1322 | if (st->ss == NULL) { | |
1323 | st->ss = &super_imsm; | |
1324 | st->minor_version = 0; | |
1325 | st->max_devs = IMSM_MAX_DEVICES; | |
1326 | } | |
1327 | ||
1328 | return 0; | |
1329 | } | |
1330 | ||
ef6ffade DW |
1331 | static __u16 info_to_blocks_per_strip(mdu_array_info_t *info) |
1332 | { | |
1333 | if (info->level == 1) | |
1334 | return 128; | |
1335 | return info->chunk_size >> 9; | |
1336 | } | |
1337 | ||
1338 | static __u32 info_to_num_data_stripes(mdu_array_info_t *info) | |
1339 | { | |
1340 | __u32 num_stripes; | |
1341 | ||
1342 | num_stripes = (info->size * 2) / info_to_blocks_per_strip(info); | |
1343 | if (info->level == 1) | |
1344 | num_stripes /= 2; | |
1345 | ||
1346 | return num_stripes; | |
1347 | } | |
1348 | ||
fcfd9599 DW |
1349 | static __u32 info_to_blocks_per_member(mdu_array_info_t *info) |
1350 | { | |
1351 | return (info->size * 2) & ~(info_to_blocks_per_strip(info) - 1); | |
1352 | } | |
1353 | ||
8b353278 DW |
1354 | static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info, |
1355 | unsigned long long size, char *name, | |
1356 | char *homehost, int *uuid) | |
cdddbdbc | 1357 | { |
c2c087e6 DW |
1358 | /* We are creating a volume inside a pre-existing container. |
1359 | * so st->sb is already set. | |
1360 | */ | |
1361 | struct intel_super *super = st->sb; | |
949c47a0 | 1362 | struct imsm_super *mpb = super->anchor; |
c2c087e6 DW |
1363 | struct imsm_dev *dev; |
1364 | struct imsm_vol *vol; | |
1365 | struct imsm_map *map; | |
1366 | int idx = mpb->num_raid_devs; | |
1367 | int i; | |
1368 | unsigned long long array_blocks; | |
c2c087e6 | 1369 | __u32 offset = 0; |
2c092cad | 1370 | size_t size_old, size_new; |
cdddbdbc | 1371 | |
c2c087e6 DW |
1372 | if (mpb->num_raid_devs >= 2) { |
1373 | fprintf(stderr, Name": This imsm-container already has the " | |
1374 | "maximum of 2 volumes\n"); | |
1375 | return 0; | |
1376 | } | |
1377 | ||
2c092cad DW |
1378 | /* ensure the mpb is large enough for the new data */ |
1379 | size_old = __le32_to_cpu(mpb->mpb_size); | |
1380 | size_new = disks_to_mpb_size(info->nr_disks); | |
1381 | if (size_new > size_old) { | |
1382 | void *mpb_new; | |
1383 | size_t size_round = ROUND_UP(size_new, 512); | |
1384 | ||
1385 | if (posix_memalign(&mpb_new, 512, size_round) != 0) { | |
1386 | fprintf(stderr, Name": could not allocate new mpb\n"); | |
1387 | return 0; | |
1388 | } | |
1389 | memcpy(mpb_new, mpb, size_old); | |
1390 | free(mpb); | |
1391 | mpb = mpb_new; | |
949c47a0 | 1392 | super->anchor = mpb_new; |
2c092cad DW |
1393 | mpb->mpb_size = __cpu_to_le32(size_new); |
1394 | memset(mpb_new + size_old, 0, size_round - size_old); | |
1395 | } | |
bf5a934a | 1396 | super->current_vol = idx; |
d23fe947 DW |
1397 | /* when creating the first raid device in this container set num_disks |
1398 | * to zero, i.e. delete this spare and add raid member devices in | |
1399 | * add_to_super_imsm_volume() | |
1400 | */ | |
1401 | if (super->current_vol == 0) | |
1402 | mpb->num_disks = 0; | |
bf5a934a | 1403 | sprintf(st->subarray, "%d", idx); |
949c47a0 DW |
1404 | dev = malloc(sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1)); |
1405 | if (!dev) { | |
1406 | fprintf(stderr, Name": could not allocate raid device\n"); | |
1407 | return 0; | |
1408 | } | |
c2c087e6 DW |
1409 | strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN); |
1410 | array_blocks = calc_array_size(info->level, info->raid_disks, | |
1411 | info->layout, info->chunk_size, | |
1412 | info->size*2); | |
1413 | dev->size_low = __cpu_to_le32((__u32) array_blocks); | |
1414 | dev->size_high = __cpu_to_le32((__u32) (array_blocks >> 32)); | |
1415 | dev->status = __cpu_to_le32(0); | |
1416 | dev->reserved_blocks = __cpu_to_le32(0); | |
1417 | vol = &dev->vol; | |
1418 | vol->migr_state = 0; | |
1419 | vol->migr_type = 0; | |
1420 | vol->dirty = 0; | |
1421 | for (i = 0; i < idx; i++) { | |
949c47a0 | 1422 | struct imsm_dev *prev = get_imsm_dev(super, i); |
a965f303 | 1423 | struct imsm_map *pmap = get_imsm_map(prev, 0); |
c2c087e6 DW |
1424 | |
1425 | offset += __le32_to_cpu(pmap->blocks_per_member); | |
1426 | offset += IMSM_RESERVED_SECTORS; | |
1427 | } | |
a965f303 | 1428 | map = get_imsm_map(dev, 0); |
c2c087e6 | 1429 | map->pba_of_lba0 = __cpu_to_le32(offset); |
fcfd9599 | 1430 | map->blocks_per_member = __cpu_to_le32(info_to_blocks_per_member(info)); |
ef6ffade DW |
1431 | map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info)); |
1432 | map->num_data_stripes = __cpu_to_le32(info_to_num_data_stripes(info)); | |
c2c087e6 DW |
1433 | map->map_state = info->level ? IMSM_T_STATE_UNINITIALIZED : |
1434 | IMSM_T_STATE_NORMAL; | |
ef6ffade DW |
1435 | |
1436 | if (info->level == 1 && info->raid_disks > 2) { | |
1437 | fprintf(stderr, Name": imsm does not support more than 2 disks" | |
1438 | "in a raid1 volume\n"); | |
1439 | return 0; | |
1440 | } | |
c2c087e6 DW |
1441 | if (info->level == 10) |
1442 | map->raid_level = 1; | |
1443 | else | |
1444 | map->raid_level = info->level; | |
ef6ffade | 1445 | |
c2c087e6 DW |
1446 | map->num_members = info->raid_disks; |
1447 | for (i = 0; i < map->num_members; i++) { | |
1448 | /* initialized in add_to_super */ | |
1449 | map->disk_ord_tbl[i] = __cpu_to_le32(0); | |
1450 | } | |
949c47a0 DW |
1451 | mpb->num_raid_devs++; |
1452 | super->dev_tbl[super->current_vol] = dev; | |
c2c087e6 DW |
1453 | |
1454 | return 1; | |
cdddbdbc DW |
1455 | } |
1456 | ||
bf5a934a DW |
1457 | static int init_super_imsm(struct supertype *st, mdu_array_info_t *info, |
1458 | unsigned long long size, char *name, | |
1459 | char *homehost, int *uuid) | |
1460 | { | |
1461 | /* This is primarily called by Create when creating a new array. | |
1462 | * We will then get add_to_super called for each component, and then | |
1463 | * write_init_super called to write it out to each device. | |
1464 | * For IMSM, Create can create on fresh devices or on a pre-existing | |
1465 | * array. | |
1466 | * To create on a pre-existing array a different method will be called. | |
1467 | * This one is just for fresh drives. | |
1468 | */ | |
1469 | struct intel_super *super; | |
1470 | struct imsm_super *mpb; | |
1471 | size_t mpb_size; | |
1472 | ||
1473 | if (!info) { | |
1474 | st->sb = NULL; | |
1475 | return 0; | |
1476 | } | |
1477 | if (st->sb) | |
1478 | return init_super_imsm_volume(st, info, size, name, homehost, | |
1479 | uuid); | |
1480 | ||
1481 | super = alloc_super(1); | |
1482 | if (!super) | |
1483 | return 0; | |
1484 | mpb_size = disks_to_mpb_size(info->nr_disks); | |
ef649044 | 1485 | if (posix_memalign(&super->buf, 512, mpb_size) != 0) { |
bf5a934a DW |
1486 | free(super); |
1487 | return 0; | |
1488 | } | |
ef649044 | 1489 | mpb = super->buf; |
bf5a934a DW |
1490 | memset(mpb, 0, mpb_size); |
1491 | ||
1492 | memcpy(mpb->sig, MPB_SIGNATURE, strlen(MPB_SIGNATURE)); | |
1493 | memcpy(mpb->sig + strlen(MPB_SIGNATURE), MPB_VERSION_RAID5, | |
1494 | strlen(MPB_VERSION_RAID5)); | |
1495 | mpb->mpb_size = mpb_size; | |
1496 | ||
bf5a934a DW |
1497 | st->sb = super; |
1498 | return 1; | |
1499 | } | |
1500 | ||
1501 | static void add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk, | |
1502 | int fd, char *devname) | |
1503 | { | |
1504 | struct intel_super *super = st->sb; | |
d23fe947 | 1505 | struct imsm_super *mpb = super->anchor; |
bf5a934a DW |
1506 | struct dl *dl; |
1507 | struct imsm_dev *dev; | |
1508 | struct imsm_map *map; | |
bf5a934a DW |
1509 | __u32 status; |
1510 | ||
949c47a0 | 1511 | dev = get_imsm_dev(super, super->current_vol); |
a965f303 | 1512 | map = get_imsm_map(dev, 0); |
bf5a934a DW |
1513 | |
1514 | for (dl = super->disks; dl ; dl = dl->next) | |
1515 | if (dl->major == dk->major && | |
1516 | dl->minor == dk->minor) | |
1517 | break; | |
d23fe947 | 1518 | |
bf5a934a DW |
1519 | if (!dl || ! (dk->state & (1<<MD_DISK_SYNC))) |
1520 | return; | |
1521 | ||
d23fe947 DW |
1522 | /* add a pristine spare to the metadata */ |
1523 | if (dl->index < 0) { | |
1524 | dl->index = super->anchor->num_disks; | |
1525 | super->anchor->num_disks++; | |
1526 | } | |
bf5a934a | 1527 | map->disk_ord_tbl[dk->number] = __cpu_to_le32(dl->index); |
bf5a934a | 1528 | status = CONFIGURED_DISK | USABLE_DISK; |
d23fe947 DW |
1529 | dl->disk.status = __cpu_to_le32(status); |
1530 | ||
1531 | /* if we are creating the first raid device update the family number */ | |
1532 | if (super->current_vol == 0) { | |
1533 | __u32 sum; | |
1534 | struct imsm_dev *_dev = __get_imsm_dev(mpb, 0); | |
1535 | struct imsm_disk *_disk = __get_imsm_disk(mpb, dl->index); | |
1536 | ||
1537 | *_dev = *dev; | |
1538 | *_disk = dl->disk; | |
1539 | sum = __gen_imsm_checksum(mpb); | |
1540 | mpb->family_num = __cpu_to_le32(sum); | |
1541 | } | |
bf5a934a DW |
1542 | } |
1543 | ||
c2c087e6 | 1544 | static void add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk, |
cdddbdbc DW |
1545 | int fd, char *devname) |
1546 | { | |
c2c087e6 | 1547 | struct intel_super *super = st->sb; |
c2c087e6 DW |
1548 | struct dl *dd; |
1549 | unsigned long long size; | |
1550 | __u32 status, id; | |
1551 | int rv; | |
1552 | struct stat stb; | |
1553 | ||
bf5a934a DW |
1554 | if (super->current_vol >= 0) { |
1555 | add_to_super_imsm_volume(st, dk, fd, devname); | |
1556 | return; | |
1557 | } | |
1558 | ||
c2c087e6 DW |
1559 | fstat(fd, &stb); |
1560 | dd = malloc(sizeof(*dd)); | |
b9f594fe | 1561 | if (!dd) { |
c2c087e6 DW |
1562 | fprintf(stderr, |
1563 | Name ": malloc failed %s:%d.\n", __func__, __LINE__); | |
1564 | abort(); | |
1565 | } | |
1566 | memset(dd, 0, sizeof(*dd)); | |
1567 | dd->major = major(stb.st_rdev); | |
1568 | dd->minor = minor(stb.st_rdev); | |
b9f594fe | 1569 | dd->index = -1; |
c2c087e6 | 1570 | dd->devname = devname ? strdup(devname) : NULL; |
c2c087e6 DW |
1571 | dd->fd = fd; |
1572 | rv = imsm_read_serial(fd, devname, dd->serial); | |
1573 | if (rv) { | |
1574 | fprintf(stderr, | |
0030e8d6 | 1575 | Name ": failed to retrieve scsi serial, aborting\n"); |
949c47a0 | 1576 | free(dd); |
0030e8d6 | 1577 | abort(); |
c2c087e6 DW |
1578 | } |
1579 | ||
c2c087e6 DW |
1580 | get_dev_size(fd, NULL, &size); |
1581 | size /= 512; | |
1582 | status = USABLE_DISK | SPARE_DISK; | |
b9f594fe DW |
1583 | strcpy((char *) dd->disk.serial, (char *) dd->serial); |
1584 | dd->disk.total_blocks = __cpu_to_le32(size); | |
1585 | dd->disk.status = __cpu_to_le32(status); | |
c2c087e6 | 1586 | if (sysfs_disk_to_scsi_id(fd, &id) == 0) |
b9f594fe | 1587 | dd->disk.scsi_id = __cpu_to_le32(id); |
c2c087e6 | 1588 | else |
b9f594fe | 1589 | dd->disk.scsi_id = __cpu_to_le32(0); |
43dad3d6 DW |
1590 | |
1591 | if (st->update_tail) { | |
1592 | dd->next = super->add; | |
1593 | super->add = dd; | |
1594 | } else { | |
1595 | dd->next = super->disks; | |
1596 | super->disks = dd; | |
1597 | } | |
cdddbdbc DW |
1598 | } |
1599 | ||
c2c087e6 DW |
1600 | static int store_imsm_mpb(int fd, struct intel_super *super); |
1601 | ||
d23fe947 DW |
1602 | /* spare records have their own family number and do not have any defined raid |
1603 | * devices | |
1604 | */ | |
1605 | static int write_super_imsm_spares(struct intel_super *super, int doclose) | |
1606 | { | |
1607 | struct imsm_super mpb_save; | |
1608 | struct imsm_super *mpb = super->anchor; | |
1609 | __u32 sum; | |
1610 | struct dl *d; | |
1611 | ||
1612 | mpb_save = *mpb; | |
1613 | mpb->num_raid_devs = 0; | |
1614 | mpb->num_disks = 1; | |
1615 | mpb->mpb_size = sizeof(struct imsm_super); | |
1616 | mpb->generation_num = __cpu_to_le32(1UL); | |
1617 | ||
1618 | for (d = super->disks; d; d = d->next) { | |
1619 | if (d->index >= 0) | |
1620 | continue; | |
1621 | ||
1622 | mpb->disk[0] = d->disk; | |
1623 | sum = __gen_imsm_checksum(mpb); | |
1624 | mpb->family_num = __cpu_to_le32(sum); | |
1625 | sum = __gen_imsm_checksum(mpb); | |
1626 | mpb->check_sum = __cpu_to_le32(sum); | |
1627 | ||
1628 | if (store_imsm_mpb(d->fd, super)) { | |
1629 | fprintf(stderr, "%s: failed for device %d:%d %s\n", | |
1630 | __func__, d->major, d->minor, strerror(errno)); | |
1631 | *mpb = mpb_save; | |
e74255d9 | 1632 | return 1; |
d23fe947 DW |
1633 | } |
1634 | if (doclose) { | |
1635 | close(d->fd); | |
1636 | d->fd = -1; | |
1637 | } | |
1638 | } | |
1639 | ||
1640 | *mpb = mpb_save; | |
e74255d9 | 1641 | return 0; |
d23fe947 DW |
1642 | } |
1643 | ||
c2c087e6 | 1644 | static int write_super_imsm(struct intel_super *super, int doclose) |
cdddbdbc | 1645 | { |
949c47a0 | 1646 | struct imsm_super *mpb = super->anchor; |
c2c087e6 DW |
1647 | struct dl *d; |
1648 | __u32 generation; | |
1649 | __u32 sum; | |
d23fe947 DW |
1650 | int spares = 0; |
1651 | int raid_disks = 0; | |
949c47a0 | 1652 | int i; |
a48ac0a8 | 1653 | __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk); |
cdddbdbc | 1654 | |
c2c087e6 DW |
1655 | /* 'generation' is incremented everytime the metadata is written */ |
1656 | generation = __le32_to_cpu(mpb->generation_num); | |
1657 | generation++; | |
1658 | mpb->generation_num = __cpu_to_le32(generation); | |
1659 | ||
d23fe947 DW |
1660 | for (d = super->disks; d; d = d->next) { |
1661 | if (d->index < 0) | |
1662 | spares++; | |
1663 | else { | |
1664 | raid_disks++; | |
1665 | mpb->disk[d->index] = d->disk; | |
a48ac0a8 | 1666 | mpb_size += sizeof(struct imsm_disk); |
d23fe947 DW |
1667 | } |
1668 | } | |
1669 | if (raid_disks != mpb->num_disks) { | |
1670 | fprintf(stderr, "%s: expected %d disks only found %d\n", | |
1671 | __func__, mpb->num_disks, raid_disks); | |
e74255d9 | 1672 | return 1; |
d23fe947 | 1673 | } |
b9f594fe | 1674 | |
949c47a0 DW |
1675 | for (i = 0; i < mpb->num_raid_devs; i++) { |
1676 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
1677 | ||
1678 | imsm_copy_dev(dev, super->dev_tbl[i]); | |
a48ac0a8 | 1679 | mpb_size += sizeof_imsm_dev(dev, 0); |
949c47a0 | 1680 | } |
a48ac0a8 DW |
1681 | mpb_size += __le32_to_cpu(mpb->bbm_log_size); |
1682 | mpb->mpb_size = __cpu_to_le32(mpb_size); | |
949c47a0 | 1683 | |
c2c087e6 | 1684 | /* recalculate checksum */ |
949c47a0 | 1685 | sum = __gen_imsm_checksum(mpb); |
c2c087e6 DW |
1686 | mpb->check_sum = __cpu_to_le32(sum); |
1687 | ||
d23fe947 | 1688 | /* write the mpb for disks that compose raid devices */ |
c2c087e6 | 1689 | for (d = super->disks; d ; d = d->next) { |
d23fe947 DW |
1690 | if (d->index < 0) |
1691 | continue; | |
c2c087e6 DW |
1692 | if (store_imsm_mpb(d->fd, super)) { |
1693 | fprintf(stderr, "%s: failed for device %d:%d %s\n", | |
1694 | __func__, d->major, d->minor, strerror(errno)); | |
e74255d9 | 1695 | return 1; |
c2c087e6 DW |
1696 | } |
1697 | if (doclose) { | |
1698 | close(d->fd); | |
1699 | d->fd = -1; | |
1700 | } | |
1701 | } | |
1702 | ||
d23fe947 DW |
1703 | if (spares) |
1704 | return write_super_imsm_spares(super, doclose); | |
1705 | ||
e74255d9 | 1706 | return 0; |
c2c087e6 DW |
1707 | } |
1708 | ||
43dad3d6 DW |
1709 | static int create_array(struct supertype *st) |
1710 | { | |
1711 | size_t len; | |
1712 | struct imsm_update_create_array *u; | |
1713 | struct intel_super *super = st->sb; | |
1714 | struct imsm_dev *dev = get_imsm_dev(super, super->current_vol); | |
1715 | ||
1716 | len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0); | |
1717 | u = malloc(len); | |
1718 | if (!u) { | |
1719 | fprintf(stderr, "%s: failed to allocate update buffer\n", | |
1720 | __func__); | |
1721 | return 1; | |
1722 | } | |
1723 | ||
1724 | u->type = update_create_array; | |
1725 | u->dev_idx = super->current_vol; | |
1726 | imsm_copy_dev(&u->dev, dev); | |
1727 | append_metadata_update(st, u, len); | |
1728 | ||
1729 | return 0; | |
1730 | } | |
1731 | ||
1732 | static int add_disk(struct supertype *st) | |
1733 | { | |
1734 | struct intel_super *super = st->sb; | |
1735 | size_t len; | |
1736 | struct imsm_update_add_disk *u; | |
1737 | ||
1738 | if (!super->add) | |
1739 | return 0; | |
1740 | ||
1741 | len = sizeof(*u); | |
1742 | u = malloc(len); | |
1743 | if (!u) { | |
1744 | fprintf(stderr, "%s: failed to allocate update buffer\n", | |
1745 | __func__); | |
1746 | return 1; | |
1747 | } | |
1748 | ||
1749 | u->type = update_add_disk; | |
1750 | append_metadata_update(st, u, len); | |
1751 | ||
1752 | return 0; | |
1753 | } | |
1754 | ||
c2c087e6 DW |
1755 | static int write_init_super_imsm(struct supertype *st) |
1756 | { | |
8273f55e | 1757 | if (st->update_tail) { |
43dad3d6 DW |
1758 | /* queue the recently created array / added disk |
1759 | * as a metadata update */ | |
8273f55e | 1760 | struct intel_super *super = st->sb; |
8273f55e | 1761 | struct dl *d; |
43dad3d6 | 1762 | int rv; |
8273f55e | 1763 | |
43dad3d6 DW |
1764 | /* determine if we are creating a volume or adding a disk */ |
1765 | if (super->current_vol < 0) { | |
1766 | /* in the add disk case we are running in mdmon | |
1767 | * context, so don't close fd's | |
1768 | */ | |
1769 | return add_disk(st); | |
1770 | } else | |
1771 | rv = create_array(st); | |
8273f55e DW |
1772 | |
1773 | for (d = super->disks; d ; d = d->next) { | |
1774 | close(d->fd); | |
1775 | d->fd = -1; | |
1776 | } | |
1777 | ||
43dad3d6 | 1778 | return rv; |
8273f55e DW |
1779 | } else |
1780 | return write_super_imsm(st->sb, 1); | |
cdddbdbc DW |
1781 | } |
1782 | ||
1783 | static int store_zero_imsm(struct supertype *st, int fd) | |
1784 | { | |
551c80c1 | 1785 | unsigned long long dsize; |
6416d527 | 1786 | void *buf; |
551c80c1 DW |
1787 | |
1788 | get_dev_size(fd, NULL, &dsize); | |
1789 | ||
1790 | /* first block is stored on second to last sector of the disk */ | |
1791 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) | |
1792 | return 1; | |
1793 | ||
ad97895e DW |
1794 | if (posix_memalign(&buf, 512, 512) != 0) |
1795 | return 1; | |
1796 | ||
eb7ea463 DW |
1797 | memset(buf, 0, 512); |
1798 | if (write(fd, buf, 512) != 512) | |
551c80c1 | 1799 | return 1; |
cdddbdbc DW |
1800 | return 0; |
1801 | } | |
1802 | ||
cdddbdbc DW |
1803 | static int validate_geometry_imsm_container(struct supertype *st, int level, |
1804 | int layout, int raiddisks, int chunk, | |
c2c087e6 | 1805 | unsigned long long size, char *dev, |
2c514b71 NB |
1806 | unsigned long long *freesize, |
1807 | int verbose) | |
cdddbdbc | 1808 | { |
c2c087e6 DW |
1809 | int fd; |
1810 | unsigned long long ldsize; | |
cdddbdbc | 1811 | |
c2c087e6 DW |
1812 | if (level != LEVEL_CONTAINER) |
1813 | return 0; | |
1814 | if (!dev) | |
1815 | return 1; | |
1816 | ||
1817 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
1818 | if (fd < 0) { | |
2c514b71 NB |
1819 | if (verbose) |
1820 | fprintf(stderr, Name ": imsm: Cannot open %s: %s\n", | |
1821 | dev, strerror(errno)); | |
c2c087e6 DW |
1822 | return 0; |
1823 | } | |
1824 | if (!get_dev_size(fd, dev, &ldsize)) { | |
1825 | close(fd); | |
1826 | return 0; | |
1827 | } | |
1828 | close(fd); | |
1829 | ||
1830 | *freesize = avail_size_imsm(st, ldsize >> 9); | |
1831 | ||
1832 | return 1; | |
cdddbdbc DW |
1833 | } |
1834 | ||
c2c087e6 DW |
1835 | /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd |
1836 | * FIX ME add ahci details | |
1837 | */ | |
8b353278 DW |
1838 | static int validate_geometry_imsm_volume(struct supertype *st, int level, |
1839 | int layout, int raiddisks, int chunk, | |
c2c087e6 | 1840 | unsigned long long size, char *dev, |
2c514b71 NB |
1841 | unsigned long long *freesize, |
1842 | int verbose) | |
cdddbdbc | 1843 | { |
c2c087e6 DW |
1844 | struct stat stb; |
1845 | struct intel_super *super = st->sb; | |
1846 | struct dl *dl; | |
1847 | unsigned long long pos = 0; | |
1848 | unsigned long long maxsize; | |
1849 | struct extent *e; | |
1850 | int i; | |
cdddbdbc | 1851 | |
c2c087e6 DW |
1852 | if (level == LEVEL_CONTAINER) |
1853 | return 0; | |
1854 | ||
1855 | if (level == 1 && raiddisks > 2) { | |
2c514b71 NB |
1856 | if (verbose) |
1857 | fprintf(stderr, Name ": imsm does not support more " | |
1858 | "than 2 in a raid1 configuration\n"); | |
c2c087e6 DW |
1859 | return 0; |
1860 | } | |
1861 | ||
1862 | /* We must have the container info already read in. */ | |
1863 | if (!super) | |
1864 | return 0; | |
1865 | ||
1866 | if (!dev) { | |
1867 | /* General test: make sure there is space for | |
2da8544a DW |
1868 | * 'raiddisks' device extents of size 'size' at a given |
1869 | * offset | |
c2c087e6 DW |
1870 | */ |
1871 | unsigned long long minsize = size*2 /* convert to blocks */; | |
2da8544a | 1872 | unsigned long long start_offset = ~0ULL; |
c2c087e6 DW |
1873 | int dcnt = 0; |
1874 | if (minsize == 0) | |
1875 | minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
1876 | for (dl = super->disks; dl ; dl = dl->next) { | |
1877 | int found = 0; | |
1878 | ||
bf5a934a | 1879 | pos = 0; |
c2c087e6 DW |
1880 | i = 0; |
1881 | e = get_extents(super, dl); | |
1882 | if (!e) continue; | |
1883 | do { | |
1884 | unsigned long long esize; | |
1885 | esize = e[i].start - pos; | |
1886 | if (esize >= minsize) | |
1887 | found = 1; | |
2da8544a DW |
1888 | if (found && start_offset == ~0ULL) { |
1889 | start_offset = pos; | |
1890 | break; | |
1891 | } else if (found && pos != start_offset) { | |
1892 | found = 0; | |
1893 | break; | |
1894 | } | |
c2c087e6 DW |
1895 | pos = e[i].start + e[i].size; |
1896 | i++; | |
1897 | } while (e[i-1].size); | |
1898 | if (found) | |
1899 | dcnt++; | |
1900 | free(e); | |
1901 | } | |
1902 | if (dcnt < raiddisks) { | |
2c514b71 NB |
1903 | if (verbose) |
1904 | fprintf(stderr, Name ": imsm: Not enough " | |
1905 | "devices with space for this array " | |
1906 | "(%d < %d)\n", | |
1907 | dcnt, raiddisks); | |
c2c087e6 DW |
1908 | return 0; |
1909 | } | |
1910 | return 1; | |
1911 | } | |
1912 | /* This device must be a member of the set */ | |
1913 | if (stat(dev, &stb) < 0) | |
1914 | return 0; | |
1915 | if ((S_IFMT & stb.st_mode) != S_IFBLK) | |
1916 | return 0; | |
1917 | for (dl = super->disks ; dl ; dl = dl->next) { | |
1918 | if (dl->major == major(stb.st_rdev) && | |
1919 | dl->minor == minor(stb.st_rdev)) | |
1920 | break; | |
1921 | } | |
1922 | if (!dl) { | |
2c514b71 NB |
1923 | if (verbose) |
1924 | fprintf(stderr, Name ": %s is not in the " | |
1925 | "same imsm set\n", dev); | |
c2c087e6 DW |
1926 | return 0; |
1927 | } | |
1928 | e = get_extents(super, dl); | |
1929 | maxsize = 0; | |
1930 | i = 0; | |
1931 | if (e) do { | |
1932 | unsigned long long esize; | |
1933 | esize = e[i].start - pos; | |
1934 | if (esize >= maxsize) | |
1935 | maxsize = esize; | |
1936 | pos = e[i].start + e[i].size; | |
1937 | i++; | |
1938 | } while (e[i-1].size); | |
1939 | *freesize = maxsize; | |
1940 | ||
1941 | return 1; | |
cdddbdbc DW |
1942 | } |
1943 | ||
604b746f JD |
1944 | int imsm_bbm_log_size(struct imsm_super *mpb) |
1945 | { | |
1946 | return __le32_to_cpu(mpb->bbm_log_size); | |
1947 | } | |
1948 | ||
bf5a934a DW |
1949 | static int validate_geometry_imsm(struct supertype *st, int level, int layout, |
1950 | int raiddisks, int chunk, unsigned long long size, | |
1951 | char *dev, unsigned long long *freesize, | |
1952 | int verbose) | |
1953 | { | |
1954 | int fd, cfd; | |
1955 | struct mdinfo *sra; | |
1956 | ||
1957 | /* if given unused devices create a container | |
1958 | * if given given devices in a container create a member volume | |
1959 | */ | |
1960 | if (level == LEVEL_CONTAINER) { | |
1961 | /* Must be a fresh device to add to a container */ | |
1962 | return validate_geometry_imsm_container(st, level, layout, | |
1963 | raiddisks, chunk, size, | |
1964 | dev, freesize, | |
1965 | verbose); | |
1966 | } | |
1967 | ||
1968 | if (st->sb) { | |
1969 | /* creating in a given container */ | |
1970 | return validate_geometry_imsm_volume(st, level, layout, | |
1971 | raiddisks, chunk, size, | |
1972 | dev, freesize, verbose); | |
1973 | } | |
1974 | ||
1975 | /* limit creation to the following levels */ | |
1976 | if (!dev) | |
1977 | switch (level) { | |
1978 | case 0: | |
1979 | case 1: | |
1980 | case 10: | |
1981 | case 5: | |
1982 | break; | |
1983 | default: | |
1984 | return 1; | |
1985 | } | |
1986 | ||
1987 | /* This device needs to be a device in an 'imsm' container */ | |
1988 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
1989 | if (fd >= 0) { | |
1990 | if (verbose) | |
1991 | fprintf(stderr, | |
1992 | Name ": Cannot create this array on device %s\n", | |
1993 | dev); | |
1994 | close(fd); | |
1995 | return 0; | |
1996 | } | |
1997 | if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) { | |
1998 | if (verbose) | |
1999 | fprintf(stderr, Name ": Cannot open %s: %s\n", | |
2000 | dev, strerror(errno)); | |
2001 | return 0; | |
2002 | } | |
2003 | /* Well, it is in use by someone, maybe an 'imsm' container. */ | |
2004 | cfd = open_container(fd); | |
2005 | if (cfd < 0) { | |
2006 | close(fd); | |
2007 | if (verbose) | |
2008 | fprintf(stderr, Name ": Cannot use %s: It is busy\n", | |
2009 | dev); | |
2010 | return 0; | |
2011 | } | |
2012 | sra = sysfs_read(cfd, 0, GET_VERSION); | |
2013 | close(fd); | |
2014 | if (sra && sra->array.major_version == -1 && | |
2015 | strcmp(sra->text_version, "imsm") == 0) { | |
2016 | /* This is a member of a imsm container. Load the container | |
2017 | * and try to create a volume | |
2018 | */ | |
2019 | struct intel_super *super; | |
2020 | ||
2021 | if (load_super_imsm_all(st, cfd, (void **) &super, NULL, 1) == 0) { | |
2022 | st->sb = super; | |
2023 | st->container_dev = fd2devnum(cfd); | |
2024 | close(cfd); | |
2025 | return validate_geometry_imsm_volume(st, level, layout, | |
2026 | raiddisks, chunk, | |
2027 | size, dev, | |
2028 | freesize, verbose); | |
2029 | } | |
2030 | close(cfd); | |
2031 | } else /* may belong to another container */ | |
2032 | return 0; | |
2033 | ||
2034 | return 1; | |
2035 | } | |
2036 | ||
cdddbdbc DW |
2037 | static struct mdinfo *container_content_imsm(struct supertype *st) |
2038 | { | |
4f5bc454 DW |
2039 | /* Given a container loaded by load_super_imsm_all, |
2040 | * extract information about all the arrays into | |
2041 | * an mdinfo tree. | |
2042 | * | |
2043 | * For each imsm_dev create an mdinfo, fill it in, | |
2044 | * then look for matching devices in super->disks | |
2045 | * and create appropriate device mdinfo. | |
2046 | */ | |
2047 | struct intel_super *super = st->sb; | |
949c47a0 | 2048 | struct imsm_super *mpb = super->anchor; |
4f5bc454 DW |
2049 | struct mdinfo *rest = NULL; |
2050 | int i; | |
cdddbdbc | 2051 | |
604b746f JD |
2052 | /* do not assemble arrays that might have bad blocks */ |
2053 | if (imsm_bbm_log_size(super->anchor)) { | |
2054 | fprintf(stderr, Name ": BBM log found in metadata. " | |
2055 | "Cannot activate array(s).\n"); | |
2056 | return NULL; | |
2057 | } | |
2058 | ||
4f5bc454 | 2059 | for (i = 0; i < mpb->num_raid_devs; i++) { |
949c47a0 | 2060 | struct imsm_dev *dev = get_imsm_dev(super, i); |
4f5bc454 | 2061 | struct imsm_vol *vol = &dev->vol; |
a965f303 | 2062 | struct imsm_map *map = get_imsm_map(dev, 0); |
4f5bc454 | 2063 | struct mdinfo *this; |
4f5bc454 DW |
2064 | int slot; |
2065 | ||
2066 | this = malloc(sizeof(*this)); | |
2067 | memset(this, 0, sizeof(*this)); | |
2068 | this->next = rest; | |
4f5bc454 | 2069 | |
4f5bc454 DW |
2070 | this->array.level = get_imsm_raid_level(map); |
2071 | this->array.raid_disks = map->num_members; | |
c2c087e6 | 2072 | this->array.layout = imsm_level_to_layout(this->array.level); |
4f5bc454 DW |
2073 | this->array.md_minor = -1; |
2074 | this->array.ctime = 0; | |
2075 | this->array.utime = 0; | |
2076 | this->array.chunk_size = __le16_to_cpu(map->blocks_per_strip) << 9; | |
2077 | this->array.state = !vol->dirty; | |
2078 | this->container_member = i; | |
9a1608e5 DW |
2079 | if (map->map_state == IMSM_T_STATE_UNINITIALIZED || |
2080 | dev->vol.dirty || dev->vol.migr_state) | |
0fd5c350 DW |
2081 | this->resync_start = 0; |
2082 | else | |
2083 | this->resync_start = ~0ULL; | |
2084 | ||
4f5bc454 DW |
2085 | strncpy(this->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN); |
2086 | this->name[MAX_RAID_SERIAL_LEN] = 0; | |
2087 | ||
159c3a1a NB |
2088 | sprintf(this->text_version, "/%s/%d", |
2089 | devnum2devname(st->container_dev), | |
2090 | this->container_member); | |
2091 | ||
4f5bc454 DW |
2092 | memset(this->uuid, 0, sizeof(this->uuid)); |
2093 | ||
f54e6321 | 2094 | this->component_size = __le32_to_cpu(map->blocks_per_member); |
4f5bc454 DW |
2095 | |
2096 | for (slot = 0 ; slot < map->num_members; slot++) { | |
4f5bc454 DW |
2097 | struct mdinfo *info_d; |
2098 | struct dl *d; | |
2099 | int idx; | |
9a1608e5 | 2100 | int skip; |
4f5bc454 | 2101 | __u32 s; |
7eef0453 | 2102 | __u32 ord; |
4f5bc454 | 2103 | |
9a1608e5 | 2104 | skip = 0; |
d23fe947 | 2105 | idx = get_imsm_disk_idx(map, slot); |
7eef0453 | 2106 | ord = get_imsm_ord_tbl_ent(dev, slot); |
4f5bc454 DW |
2107 | for (d = super->disks; d ; d = d->next) |
2108 | if (d->index == idx) | |
2109 | break; | |
2110 | ||
2111 | if (d == NULL) | |
9a1608e5 DW |
2112 | skip = 1; |
2113 | ||
2114 | s = d ? __le32_to_cpu(d->disk.status) : 0; | |
2115 | if (s & FAILED_DISK) | |
2116 | skip = 1; | |
2117 | if (!(s & USABLE_DISK)) | |
2118 | skip = 1; | |
7eef0453 DW |
2119 | if (ord & IMSM_ORD_REBUILD) |
2120 | skip = 1; | |
9a1608e5 DW |
2121 | |
2122 | /* | |
2123 | * if we skip some disks the array will be assmebled degraded; | |
2124 | * reset resync start to avoid a dirty-degraded situation | |
2125 | * | |
2126 | * FIXME handle dirty degraded | |
2127 | */ | |
2128 | if (skip && !dev->vol.dirty) | |
2129 | this->resync_start = ~0ULL; | |
2130 | if (skip) | |
2131 | continue; | |
4f5bc454 DW |
2132 | |
2133 | info_d = malloc(sizeof(*info_d)); | |
9a1608e5 DW |
2134 | if (!info_d) { |
2135 | fprintf(stderr, Name ": failed to allocate disk" | |
2136 | " for volume %s\n", (char *) dev->volume); | |
2137 | free(this); | |
2138 | this = rest; | |
2139 | break; | |
2140 | } | |
4f5bc454 DW |
2141 | memset(info_d, 0, sizeof(*info_d)); |
2142 | info_d->next = this->devs; | |
2143 | this->devs = info_d; | |
2144 | ||
4f5bc454 DW |
2145 | info_d->disk.number = d->index; |
2146 | info_d->disk.major = d->major; | |
2147 | info_d->disk.minor = d->minor; | |
2148 | info_d->disk.raid_disk = slot; | |
4f5bc454 DW |
2149 | |
2150 | this->array.working_disks++; | |
2151 | ||
2152 | info_d->events = __le32_to_cpu(mpb->generation_num); | |
2153 | info_d->data_offset = __le32_to_cpu(map->pba_of_lba0); | |
2154 | info_d->component_size = __le32_to_cpu(map->blocks_per_member); | |
2155 | if (d->devname) | |
2156 | strcpy(info_d->name, d->devname); | |
2157 | } | |
9a1608e5 | 2158 | rest = this; |
4f5bc454 DW |
2159 | } |
2160 | ||
2161 | return rest; | |
cdddbdbc DW |
2162 | } |
2163 | ||
845dea95 | 2164 | |
cba0191b NB |
2165 | static int imsm_open_new(struct supertype *c, struct active_array *a, |
2166 | char *inst) | |
845dea95 | 2167 | { |
0372d5a2 | 2168 | struct intel_super *super = c->sb; |
949c47a0 | 2169 | struct imsm_super *mpb = super->anchor; |
0372d5a2 | 2170 | |
949c47a0 | 2171 | if (atoi(inst) >= mpb->num_raid_devs) { |
0372d5a2 DW |
2172 | fprintf(stderr, "%s: subarry index %d, out of range\n", |
2173 | __func__, atoi(inst)); | |
2174 | return -ENODEV; | |
2175 | } | |
2176 | ||
4e6e574a | 2177 | dprintf("imsm: open_new %s\n", inst); |
cba0191b | 2178 | a->info.container_member = atoi(inst); |
845dea95 NB |
2179 | return 0; |
2180 | } | |
2181 | ||
949c47a0 | 2182 | static __u8 imsm_check_degraded(struct intel_super *super, int n, int failed) |
c2a1e7da | 2183 | { |
949c47a0 | 2184 | struct imsm_dev *dev = get_imsm_dev(super, n); |
a965f303 | 2185 | struct imsm_map *map = get_imsm_map(dev, 0); |
c2a1e7da DW |
2186 | |
2187 | if (!failed) | |
3393c6af DW |
2188 | return map->map_state == IMSM_T_STATE_UNINITIALIZED ? |
2189 | IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL; | |
c2a1e7da DW |
2190 | |
2191 | switch (get_imsm_raid_level(map)) { | |
2192 | case 0: | |
2193 | return IMSM_T_STATE_FAILED; | |
2194 | break; | |
2195 | case 1: | |
2196 | if (failed < map->num_members) | |
2197 | return IMSM_T_STATE_DEGRADED; | |
2198 | else | |
2199 | return IMSM_T_STATE_FAILED; | |
2200 | break; | |
2201 | case 10: | |
2202 | { | |
2203 | /** | |
2204 | * check to see if any mirrors have failed, | |
2205 | * otherwise we are degraded | |
2206 | */ | |
2207 | int device_per_mirror = 2; /* FIXME is this always the case? | |
2208 | * and are they always adjacent? | |
2209 | */ | |
2210 | int failed = 0; | |
2211 | int i; | |
2212 | ||
2213 | for (i = 0; i < map->num_members; i++) { | |
2214 | int idx = get_imsm_disk_idx(map, i); | |
949c47a0 | 2215 | struct imsm_disk *disk = get_imsm_disk(super, idx); |
c2a1e7da DW |
2216 | |
2217 | if (__le32_to_cpu(disk->status) & FAILED_DISK) | |
2218 | failed++; | |
2219 | ||
2220 | if (failed >= device_per_mirror) | |
2221 | return IMSM_T_STATE_FAILED; | |
2222 | ||
2223 | /* reset 'failed' for next mirror set */ | |
2224 | if (!((i + 1) % device_per_mirror)) | |
2225 | failed = 0; | |
2226 | } | |
2227 | ||
2228 | return IMSM_T_STATE_DEGRADED; | |
2229 | } | |
2230 | case 5: | |
2231 | if (failed < 2) | |
2232 | return IMSM_T_STATE_DEGRADED; | |
2233 | else | |
2234 | return IMSM_T_STATE_FAILED; | |
2235 | break; | |
2236 | default: | |
2237 | break; | |
2238 | } | |
2239 | ||
2240 | return map->map_state; | |
2241 | } | |
2242 | ||
949c47a0 | 2243 | static int imsm_count_failed(struct intel_super *super, struct imsm_map *map) |
c2a1e7da DW |
2244 | { |
2245 | int i; | |
2246 | int failed = 0; | |
2247 | struct imsm_disk *disk; | |
2248 | ||
2249 | for (i = 0; i < map->num_members; i++) { | |
2250 | int idx = get_imsm_disk_idx(map, i); | |
2251 | ||
949c47a0 | 2252 | disk = get_imsm_disk(super, idx); |
c2a1e7da DW |
2253 | if (__le32_to_cpu(disk->status) & FAILED_DISK) |
2254 | failed++; | |
fcb84475 DW |
2255 | else if (!(__le32_to_cpu(disk->status) & USABLE_DISK)) |
2256 | failed++; | |
c2a1e7da DW |
2257 | } |
2258 | ||
2259 | return failed; | |
845dea95 NB |
2260 | } |
2261 | ||
01f157d7 | 2262 | static int imsm_set_array_state(struct active_array *a, int consistent) |
a862209d DW |
2263 | { |
2264 | int inst = a->info.container_member; | |
2265 | struct intel_super *super = a->container->sb; | |
949c47a0 | 2266 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
a965f303 | 2267 | struct imsm_map *map = get_imsm_map(dev, 0); |
a862209d DW |
2268 | int dirty = !consistent; |
2269 | int failed; | |
2270 | __u8 map_state; | |
2271 | ||
3393c6af DW |
2272 | failed = imsm_count_failed(super, map); |
2273 | map_state = imsm_check_degraded(super, inst, failed); | |
2274 | ||
272906ef DW |
2275 | if (consistent && !dev->vol.dirty && |
2276 | (dev->vol.migr_state || map_state != IMSM_T_STATE_NORMAL)) | |
2277 | a->resync_start = 0ULL; | |
01f157d7 N |
2278 | if (consistent == 2 && a->resync_start != ~0ULL) |
2279 | consistent = 0; | |
272906ef | 2280 | |
a862209d | 2281 | if (a->resync_start == ~0ULL) { |
115c3803 | 2282 | /* complete recovery or initial resync */ |
a862209d | 2283 | if (map->map_state != map_state) { |
4e6e574a | 2284 | dprintf("imsm: map_state %d: %d\n", |
a862209d DW |
2285 | inst, map_state); |
2286 | map->map_state = map_state; | |
2287 | super->updates_pending++; | |
2288 | } | |
3393c6af DW |
2289 | if (dev->vol.migr_state) { |
2290 | dprintf("imsm: mark resync complete\n"); | |
2291 | dev->vol.migr_state = 0; | |
2292 | dev->vol.migr_type = 0; | |
115c3803 | 2293 | super->updates_pending++; |
115c3803 | 2294 | } |
3393c6af DW |
2295 | } else if (!dev->vol.migr_state) { |
2296 | dprintf("imsm: mark '%s' (%llu)\n", | |
2297 | failed ? "rebuild" : "initializing", a->resync_start); | |
2298 | /* mark that we are rebuilding */ | |
2299 | map->map_state = failed ? map_state : IMSM_T_STATE_NORMAL; | |
2300 | dev->vol.migr_state = 1; | |
2301 | dev->vol.migr_type = failed ? 1 : 0; | |
2302 | dup_map(dev); | |
272906ef | 2303 | a->check_degraded = 1; |
3393c6af | 2304 | super->updates_pending++; |
115c3803 | 2305 | } |
a862209d | 2306 | |
3393c6af DW |
2307 | /* mark dirty / clean */ |
2308 | if (dirty != dev->vol.dirty) { | |
2309 | dprintf("imsm: mark '%s' (%llu)\n", | |
2310 | dirty ? "dirty" : "clean", a->resync_start); | |
2311 | dev->vol.dirty = dirty; | |
a862209d DW |
2312 | super->updates_pending++; |
2313 | } | |
01f157d7 | 2314 | return consistent; |
a862209d DW |
2315 | } |
2316 | ||
8d45d196 | 2317 | static void imsm_set_disk(struct active_array *a, int n, int state) |
845dea95 | 2318 | { |
8d45d196 DW |
2319 | int inst = a->info.container_member; |
2320 | struct intel_super *super = a->container->sb; | |
949c47a0 | 2321 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
a965f303 | 2322 | struct imsm_map *map = get_imsm_map(dev, 0); |
8d45d196 DW |
2323 | struct imsm_disk *disk; |
2324 | __u32 status; | |
2325 | int failed = 0; | |
2326 | int new_failure = 0; | |
2327 | ||
2328 | if (n > map->num_members) | |
2329 | fprintf(stderr, "imsm: set_disk %d out of range 0..%d\n", | |
2330 | n, map->num_members - 1); | |
2331 | ||
2332 | if (n < 0) | |
2333 | return; | |
2334 | ||
4e6e574a | 2335 | dprintf("imsm: set_disk %d:%x\n", n, state); |
8d45d196 | 2336 | |
949c47a0 | 2337 | disk = get_imsm_disk(super, get_imsm_disk_idx(map, n)); |
8d45d196 | 2338 | |
5802a811 | 2339 | /* check for new failures */ |
8d45d196 DW |
2340 | status = __le32_to_cpu(disk->status); |
2341 | if ((state & DS_FAULTY) && !(status & FAILED_DISK)) { | |
2342 | status |= FAILED_DISK; | |
2343 | disk->status = __cpu_to_le32(status); | |
2344 | new_failure = 1; | |
5802a811 | 2345 | super->updates_pending++; |
8d45d196 | 2346 | } |
19859edc DW |
2347 | /* check if in_sync */ |
2348 | if ((state & DS_INSYNC) && !(status & USABLE_DISK)) { | |
2349 | status |= USABLE_DISK; | |
2350 | disk->status = __cpu_to_le32(status); | |
2351 | super->updates_pending++; | |
2352 | } | |
8d45d196 | 2353 | |
5802a811 | 2354 | /* the number of failures have changed, count up 'failed' to determine |
8d45d196 DW |
2355 | * degraded / failed status |
2356 | */ | |
2357 | if (new_failure && map->map_state != IMSM_T_STATE_FAILED) | |
949c47a0 | 2358 | failed = imsm_count_failed(super, map); |
8d45d196 | 2359 | |
5802a811 | 2360 | /* determine map_state based on failed or in_sync count */ |
8d45d196 | 2361 | if (failed) |
949c47a0 | 2362 | map->map_state = imsm_check_degraded(super, inst, failed); |
5802a811 DW |
2363 | else if (map->map_state == IMSM_T_STATE_DEGRADED) { |
2364 | struct mdinfo *d; | |
2365 | int working = 0; | |
8d45d196 | 2366 | |
5802a811 DW |
2367 | for (d = a->info.devs ; d ; d = d->next) |
2368 | if (d->curr_state & DS_INSYNC) | |
2369 | working++; | |
2370 | ||
2371 | if (working == a->info.array.raid_disks) { | |
2372 | map->map_state = IMSM_T_STATE_NORMAL; | |
3393c6af DW |
2373 | dev->vol.migr_state = 0; |
2374 | dev->vol.migr_type = 0; | |
5802a811 DW |
2375 | super->updates_pending++; |
2376 | } | |
2377 | } | |
845dea95 NB |
2378 | } |
2379 | ||
c2a1e7da DW |
2380 | static int store_imsm_mpb(int fd, struct intel_super *super) |
2381 | { | |
949c47a0 | 2382 | struct imsm_super *mpb = super->anchor; |
c2a1e7da DW |
2383 | __u32 mpb_size = __le32_to_cpu(mpb->mpb_size); |
2384 | unsigned long long dsize; | |
2385 | unsigned long long sectors; | |
2386 | ||
2387 | get_dev_size(fd, NULL, &dsize); | |
2388 | ||
272f648f DW |
2389 | if (mpb_size > 512) { |
2390 | /* -1 to account for anchor */ | |
2391 | sectors = mpb_sectors(mpb) - 1; | |
c2a1e7da | 2392 | |
272f648f DW |
2393 | /* write the extended mpb to the sectors preceeding the anchor */ |
2394 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) | |
2395 | return 1; | |
c2a1e7da | 2396 | |
99e29264 | 2397 | if (write(fd, super->buf + 512, 512 * sectors) != 512 * sectors) |
272f648f DW |
2398 | return 1; |
2399 | } | |
c2a1e7da | 2400 | |
272f648f DW |
2401 | /* first block is stored on second to last sector of the disk */ |
2402 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) | |
c2a1e7da DW |
2403 | return 1; |
2404 | ||
272f648f | 2405 | if (write(fd, super->buf, 512) != 512) |
c2a1e7da DW |
2406 | return 1; |
2407 | ||
c2a1e7da DW |
2408 | return 0; |
2409 | } | |
2410 | ||
2e735d19 | 2411 | static void imsm_sync_metadata(struct supertype *container) |
845dea95 | 2412 | { |
2e735d19 | 2413 | struct intel_super *super = container->sb; |
c2a1e7da DW |
2414 | |
2415 | if (!super->updates_pending) | |
2416 | return; | |
2417 | ||
c2c087e6 | 2418 | write_super_imsm(super, 0); |
c2a1e7da DW |
2419 | |
2420 | super->updates_pending = 0; | |
845dea95 NB |
2421 | } |
2422 | ||
272906ef DW |
2423 | static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a) |
2424 | { | |
2425 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
2426 | struct imsm_map *map = get_imsm_map(dev, 0); | |
2427 | int i = get_imsm_disk_idx(map, idx); | |
2428 | struct dl *dl; | |
2429 | ||
2430 | for (dl = super->disks; dl; dl = dl->next) | |
2431 | if (dl->index == i) | |
2432 | break; | |
2433 | ||
2434 | if (__le32_to_cpu(dl->disk.status) & FAILED_DISK) | |
2435 | dl = NULL; | |
2436 | ||
2437 | if (dl) | |
2438 | dprintf("%s: found %x:%x\n", __func__, dl->major, dl->minor); | |
2439 | ||
2440 | return dl; | |
2441 | } | |
2442 | ||
2443 | static struct dl *imsm_add_spare(struct intel_super *super, int idx, struct active_array *a) | |
2444 | { | |
2445 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
2446 | struct imsm_map *map = get_imsm_map(dev, 0); | |
2447 | unsigned long long esize; | |
2448 | unsigned long long pos; | |
2449 | struct mdinfo *d; | |
2450 | struct extent *ex; | |
2451 | int j; | |
2452 | int found; | |
2453 | __u32 array_start; | |
9a1608e5 | 2454 | __u32 status; |
272906ef DW |
2455 | struct dl *dl; |
2456 | ||
2457 | for (dl = super->disks; dl; dl = dl->next) { | |
2458 | /* If in this array, skip */ | |
2459 | for (d = a->info.devs ; d ; d = d->next) | |
2460 | if (d->disk.major == dl->major && | |
2461 | d->disk.minor == dl->minor) { | |
2462 | dprintf("%x:%x already in array\n", dl->major, dl->minor); | |
2463 | break; | |
2464 | } | |
2465 | if (d) | |
2466 | continue; | |
2467 | ||
9a1608e5 DW |
2468 | /* skip marked in use or failed drives */ |
2469 | status = __le32_to_cpu(dl->disk.status); | |
2470 | if (status & FAILED_DISK || status & CONFIGURED_DISK) { | |
2471 | dprintf("%x:%x status ( %s%s)\n", | |
2472 | dl->major, dl->minor, | |
2473 | status & FAILED_DISK ? "failed " : "", | |
2474 | status & CONFIGURED_DISK ? "configured " : ""); | |
2475 | continue; | |
2476 | } | |
2477 | ||
272906ef DW |
2478 | /* Does this unused device have the requisite free space? |
2479 | * We need a->info.component_size sectors | |
2480 | */ | |
2481 | ex = get_extents(super, dl); | |
2482 | if (!ex) { | |
2483 | dprintf("cannot get extents\n"); | |
2484 | continue; | |
2485 | } | |
2486 | found = 0; | |
2487 | j = 0; | |
2488 | pos = 0; | |
2489 | array_start = __le32_to_cpu(map->pba_of_lba0); | |
2490 | ||
2491 | do { | |
2492 | /* check that we can start at pba_of_lba0 with | |
2493 | * a->info.component_size of space | |
2494 | */ | |
2495 | esize = ex[j].start - pos; | |
2496 | if (array_start >= pos && | |
2497 | array_start + a->info.component_size < ex[j].start) { | |
2498 | found = 1; | |
2499 | break; | |
2500 | } | |
2501 | pos = ex[j].start + ex[j].size; | |
2502 | j++; | |
2503 | ||
2504 | } while (ex[j-1].size); | |
2505 | ||
2506 | free(ex); | |
2507 | if (!found) { | |
2508 | dprintf("%x:%x does not have %llu at %d\n", | |
2509 | dl->major, dl->minor, | |
2510 | a->info.component_size, | |
2511 | __le32_to_cpu(map->pba_of_lba0)); | |
2512 | /* No room */ | |
2513 | continue; | |
2514 | } else | |
2515 | break; | |
2516 | } | |
2517 | ||
2518 | return dl; | |
2519 | } | |
2520 | ||
88758e9d DW |
2521 | static struct mdinfo *imsm_activate_spare(struct active_array *a, |
2522 | struct metadata_update **updates) | |
2523 | { | |
2524 | /** | |
d23fe947 DW |
2525 | * Find a device with unused free space and use it to replace a |
2526 | * failed/vacant region in an array. We replace failed regions one a | |
2527 | * array at a time. The result is that a new spare disk will be added | |
2528 | * to the first failed array and after the monitor has finished | |
2529 | * propagating failures the remainder will be consumed. | |
88758e9d | 2530 | * |
d23fe947 DW |
2531 | * FIXME add a capability for mdmon to request spares from another |
2532 | * container. | |
88758e9d DW |
2533 | */ |
2534 | ||
2535 | struct intel_super *super = a->container->sb; | |
88758e9d | 2536 | int inst = a->info.container_member; |
949c47a0 | 2537 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
a965f303 | 2538 | struct imsm_map *map = get_imsm_map(dev, 0); |
88758e9d DW |
2539 | int failed = a->info.array.raid_disks; |
2540 | struct mdinfo *rv = NULL; | |
2541 | struct mdinfo *d; | |
2542 | struct mdinfo *di; | |
2543 | struct metadata_update *mu; | |
2544 | struct dl *dl; | |
2545 | struct imsm_update_activate_spare *u; | |
2546 | int num_spares = 0; | |
2547 | int i; | |
2548 | ||
2549 | for (d = a->info.devs ; d ; d = d->next) { | |
2550 | if ((d->curr_state & DS_FAULTY) && | |
2551 | d->state_fd >= 0) | |
2552 | /* wait for Removal to happen */ | |
2553 | return NULL; | |
2554 | if (d->state_fd >= 0) | |
2555 | failed--; | |
2556 | } | |
2557 | ||
2558 | dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n", | |
2559 | inst, failed, a->info.array.raid_disks, a->info.array.level); | |
949c47a0 | 2560 | if (imsm_check_degraded(super, inst, failed) != IMSM_T_STATE_DEGRADED) |
88758e9d DW |
2561 | return NULL; |
2562 | ||
2563 | /* For each slot, if it is not working, find a spare */ | |
88758e9d DW |
2564 | for (i = 0; i < a->info.array.raid_disks; i++) { |
2565 | for (d = a->info.devs ; d ; d = d->next) | |
2566 | if (d->disk.raid_disk == i) | |
2567 | break; | |
2568 | dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0); | |
2569 | if (d && (d->state_fd >= 0)) | |
2570 | continue; | |
2571 | ||
272906ef DW |
2572 | /* |
2573 | * OK, this device needs recovery. Try to re-add the previous | |
2574 | * occupant of this slot, if this fails add a new spare | |
2575 | */ | |
2576 | dl = imsm_readd(super, i, a); | |
2577 | if (!dl) | |
2578 | dl = imsm_add_spare(super, i, a); | |
2579 | if (!dl) | |
2580 | continue; | |
2581 | ||
2582 | /* found a usable disk with enough space */ | |
2583 | di = malloc(sizeof(*di)); | |
2584 | memset(di, 0, sizeof(*di)); | |
2585 | ||
2586 | /* dl->index will be -1 in the case we are activating a | |
2587 | * pristine spare. imsm_process_update() will create a | |
2588 | * new index in this case. Once a disk is found to be | |
2589 | * failed in all member arrays it is kicked from the | |
2590 | * metadata | |
2591 | */ | |
2592 | di->disk.number = dl->index; | |
d23fe947 | 2593 | |
272906ef DW |
2594 | /* (ab)use di->devs to store a pointer to the device |
2595 | * we chose | |
2596 | */ | |
2597 | di->devs = (struct mdinfo *) dl; | |
2598 | ||
2599 | di->disk.raid_disk = i; | |
2600 | di->disk.major = dl->major; | |
2601 | di->disk.minor = dl->minor; | |
2602 | di->disk.state = 0; | |
2603 | di->data_offset = __le32_to_cpu(map->pba_of_lba0); | |
2604 | di->component_size = a->info.component_size; | |
2605 | di->container_member = inst; | |
2606 | di->next = rv; | |
2607 | rv = di; | |
2608 | num_spares++; | |
2609 | dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor, | |
2610 | i, di->data_offset); | |
88758e9d | 2611 | |
272906ef | 2612 | break; |
88758e9d DW |
2613 | } |
2614 | ||
2615 | if (!rv) | |
2616 | /* No spares found */ | |
2617 | return rv; | |
2618 | /* Now 'rv' has a list of devices to return. | |
2619 | * Create a metadata_update record to update the | |
2620 | * disk_ord_tbl for the array | |
2621 | */ | |
2622 | mu = malloc(sizeof(*mu)); | |
2623 | mu->buf = malloc(sizeof(struct imsm_update_activate_spare) * num_spares); | |
2624 | mu->space = NULL; | |
2625 | mu->len = sizeof(struct imsm_update_activate_spare) * num_spares; | |
2626 | mu->next = *updates; | |
2627 | u = (struct imsm_update_activate_spare *) mu->buf; | |
2628 | ||
2629 | for (di = rv ; di ; di = di->next) { | |
2630 | u->type = update_activate_spare; | |
d23fe947 DW |
2631 | u->dl = (struct dl *) di->devs; |
2632 | di->devs = NULL; | |
88758e9d DW |
2633 | u->slot = di->disk.raid_disk; |
2634 | u->array = inst; | |
2635 | u->next = u + 1; | |
2636 | u++; | |
2637 | } | |
2638 | (u-1)->next = NULL; | |
2639 | *updates = mu; | |
2640 | ||
2641 | return rv; | |
2642 | } | |
2643 | ||
8273f55e DW |
2644 | static int disks_overlap(struct imsm_map *m1, struct imsm_map *m2) |
2645 | { | |
2646 | int i; | |
2647 | int j; | |
2648 | int idx; | |
2649 | ||
2650 | for (i = 0; i < m1->num_members; i++) { | |
2651 | idx = get_imsm_disk_idx(m1, i); | |
2652 | for (j = 0; j < m2->num_members; j++) | |
2653 | if (idx == get_imsm_disk_idx(m2, j)) | |
2654 | return 1; | |
2655 | } | |
2656 | ||
2657 | return 0; | |
2658 | } | |
2659 | ||
ae6aad82 DW |
2660 | static void imsm_delete(struct intel_super *super, struct dl **dlp); |
2661 | ||
e8319a19 DW |
2662 | static void imsm_process_update(struct supertype *st, |
2663 | struct metadata_update *update) | |
2664 | { | |
2665 | /** | |
2666 | * crack open the metadata_update envelope to find the update record | |
2667 | * update can be one of: | |
2668 | * update_activate_spare - a spare device has replaced a failed | |
2669 | * device in an array, update the disk_ord_tbl. If this disk is | |
2670 | * present in all member arrays then also clear the SPARE_DISK | |
2671 | * flag | |
2672 | */ | |
2673 | struct intel_super *super = st->sb; | |
949c47a0 | 2674 | struct imsm_super *mpb = super->anchor; |
e8319a19 DW |
2675 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
2676 | ||
2677 | switch (type) { | |
2678 | case update_activate_spare: { | |
2679 | struct imsm_update_activate_spare *u = (void *) update->buf; | |
949c47a0 | 2680 | struct imsm_dev *dev = get_imsm_dev(super, u->array); |
a965f303 | 2681 | struct imsm_map *map = get_imsm_map(dev, 0); |
e8319a19 DW |
2682 | struct active_array *a; |
2683 | struct imsm_disk *disk; | |
2684 | __u32 status; | |
2685 | struct dl *dl; | |
e8319a19 DW |
2686 | unsigned int found; |
2687 | int victim; | |
2688 | int i; | |
2689 | ||
2690 | for (dl = super->disks; dl; dl = dl->next) | |
d23fe947 | 2691 | if (dl == u->dl) |
e8319a19 DW |
2692 | break; |
2693 | ||
2694 | if (!dl) { | |
2695 | fprintf(stderr, "error: imsm_activate_spare passed " | |
d23fe947 DW |
2696 | "an unknown disk (index: %d serial: %s)\n", |
2697 | u->dl->index, u->dl->serial); | |
e8319a19 DW |
2698 | return; |
2699 | } | |
2700 | ||
2701 | super->updates_pending++; | |
2702 | ||
d23fe947 DW |
2703 | /* adding a pristine spare, assign a new index */ |
2704 | if (dl->index < 0) { | |
2705 | dl->index = super->anchor->num_disks; | |
2706 | super->anchor->num_disks++; | |
2707 | } | |
e8319a19 | 2708 | victim = get_imsm_disk_idx(map, u->slot); |
d23fe947 DW |
2709 | map->disk_ord_tbl[u->slot] = __cpu_to_le32(dl->index); |
2710 | disk = &dl->disk; | |
e8319a19 DW |
2711 | status = __le32_to_cpu(disk->status); |
2712 | status |= CONFIGURED_DISK; | |
19859edc | 2713 | status &= ~(SPARE_DISK | USABLE_DISK); |
e8319a19 DW |
2714 | disk->status = __cpu_to_le32(status); |
2715 | ||
e8319a19 DW |
2716 | /* count arrays using the victim in the metadata */ |
2717 | found = 0; | |
2718 | for (a = st->arrays; a ; a = a->next) { | |
949c47a0 | 2719 | dev = get_imsm_dev(super, a->info.container_member); |
a965f303 | 2720 | map = get_imsm_map(dev, 0); |
e8319a19 DW |
2721 | for (i = 0; i < map->num_members; i++) |
2722 | if (victim == get_imsm_disk_idx(map, i)) | |
2723 | found++; | |
2724 | } | |
2725 | ||
2726 | /* clear some flags if the victim is no longer being | |
2727 | * utilized anywhere | |
2728 | */ | |
e8319a19 | 2729 | if (!found) { |
ae6aad82 DW |
2730 | struct dl **dlp; |
2731 | for (dlp = &super->disks; *dlp; ) | |
2732 | if ((*dlp)->index == victim) | |
2733 | break; | |
2734 | disk = &(*dlp)->disk; | |
e8319a19 DW |
2735 | status = __le32_to_cpu(disk->status); |
2736 | status &= ~(CONFIGURED_DISK | USABLE_DISK); | |
2737 | disk->status = __cpu_to_le32(status); | |
ae6aad82 DW |
2738 | /* We know that 'manager' isn't touching anything, |
2739 | * so it is safe to: | |
2740 | */ | |
2741 | imsm_delete(super, dlp); | |
e8319a19 | 2742 | } |
8273f55e DW |
2743 | break; |
2744 | } | |
2745 | case update_create_array: { | |
2746 | /* someone wants to create a new array, we need to be aware of | |
2747 | * a few races/collisions: | |
2748 | * 1/ 'Create' called by two separate instances of mdadm | |
2749 | * 2/ 'Create' versus 'activate_spare': mdadm has chosen | |
2750 | * devices that have since been assimilated via | |
2751 | * activate_spare. | |
2752 | * In the event this update can not be carried out mdadm will | |
2753 | * (FIX ME) notice that its update did not take hold. | |
2754 | */ | |
2755 | struct imsm_update_create_array *u = (void *) update->buf; | |
2756 | struct imsm_dev *dev; | |
2757 | struct imsm_map *map, *new_map; | |
2758 | unsigned long long start, end; | |
2759 | unsigned long long new_start, new_end; | |
2760 | int i; | |
2761 | int overlap = 0; | |
2762 | ||
2763 | /* handle racing creates: first come first serve */ | |
2764 | if (u->dev_idx < mpb->num_raid_devs) { | |
2765 | dprintf("%s: subarray %d already defined\n", | |
2766 | __func__, u->dev_idx); | |
2767 | return; | |
2768 | } | |
2769 | ||
2770 | /* check update is next in sequence */ | |
2771 | if (u->dev_idx != mpb->num_raid_devs) { | |
6a3e913e DW |
2772 | dprintf("%s: can not create array %d expected index %d\n", |
2773 | __func__, u->dev_idx, mpb->num_raid_devs); | |
8273f55e DW |
2774 | return; |
2775 | } | |
2776 | ||
a965f303 | 2777 | new_map = get_imsm_map(&u->dev, 0); |
8273f55e DW |
2778 | new_start = __le32_to_cpu(new_map->pba_of_lba0); |
2779 | new_end = new_start + __le32_to_cpu(new_map->blocks_per_member); | |
2780 | ||
2781 | /* handle activate_spare versus create race: | |
2782 | * check to make sure that overlapping arrays do not include | |
2783 | * overalpping disks | |
2784 | */ | |
2785 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
949c47a0 | 2786 | dev = get_imsm_dev(super, i); |
a965f303 | 2787 | map = get_imsm_map(dev, 0); |
8273f55e DW |
2788 | start = __le32_to_cpu(map->pba_of_lba0); |
2789 | end = start + __le32_to_cpu(map->blocks_per_member); | |
2790 | if ((new_start >= start && new_start <= end) || | |
2791 | (start >= new_start && start <= new_end)) | |
2792 | overlap = 1; | |
2793 | if (overlap && disks_overlap(map, new_map)) { | |
2794 | dprintf("%s: arrays overlap\n", __func__); | |
2795 | return; | |
2796 | } | |
2797 | } | |
2798 | /* check num_members sanity */ | |
2799 | if (new_map->num_members > mpb->num_disks) { | |
2800 | dprintf("%s: num_disks out of range\n", __func__); | |
2801 | return; | |
2802 | } | |
2803 | ||
949c47a0 DW |
2804 | /* check that prepare update was successful */ |
2805 | if (!update->space) { | |
2806 | dprintf("%s: prepare update failed\n", __func__); | |
2807 | return; | |
2808 | } | |
2809 | ||
8273f55e | 2810 | super->updates_pending++; |
949c47a0 DW |
2811 | dev = update->space; |
2812 | update->space = NULL; | |
2813 | imsm_copy_dev(dev, &u->dev); | |
2814 | super->dev_tbl[u->dev_idx] = dev; | |
8273f55e | 2815 | mpb->num_raid_devs++; |
8273f55e DW |
2816 | |
2817 | /* fix up flags, if arrays overlap then the drives can not be | |
2818 | * spares | |
2819 | */ | |
2820 | for (i = 0; i < map->num_members; i++) { | |
2821 | struct imsm_disk *disk; | |
2822 | __u32 status; | |
2823 | ||
949c47a0 | 2824 | disk = get_imsm_disk(super, get_imsm_disk_idx(map, i)); |
8273f55e DW |
2825 | status = __le32_to_cpu(disk->status); |
2826 | status |= CONFIGURED_DISK; | |
2827 | if (overlap) | |
2828 | status &= ~SPARE_DISK; | |
2829 | disk->status = __cpu_to_le32(status); | |
2830 | } | |
2831 | break; | |
e8319a19 | 2832 | } |
43dad3d6 DW |
2833 | case update_add_disk: |
2834 | ||
2835 | /* we may be able to repair some arrays if disks are | |
2836 | * being added */ | |
2837 | if (super->add) { | |
2838 | struct active_array *a; | |
2839 | for (a = st->arrays; a; a = a->next) | |
2840 | a->check_degraded = 1; | |
2841 | } | |
2842 | /* check if we can add / replace some disks in the | |
2843 | * metadata */ | |
2844 | while (super->add) { | |
2845 | struct dl **dlp, *dl, *al; | |
2846 | al = super->add; | |
2847 | super->add = al->next; | |
2848 | for (dlp = &super->disks; *dlp ; ) { | |
2849 | if (memcmp(al->serial, (*dlp)->serial, | |
2850 | MAX_RAID_SERIAL_LEN) == 0) { | |
2851 | dl = *dlp; | |
2852 | *dlp = (*dlp)->next; | |
2853 | __free_imsm_disk(dl); | |
2854 | break; | |
2855 | } else | |
2856 | dlp = &(*dlp)->next; | |
2857 | } | |
2858 | al->next = super->disks; | |
2859 | super->disks = al; | |
2860 | } | |
2861 | ||
2862 | break; | |
e8319a19 DW |
2863 | } |
2864 | } | |
88758e9d | 2865 | |
8273f55e DW |
2866 | static void imsm_prepare_update(struct supertype *st, |
2867 | struct metadata_update *update) | |
2868 | { | |
949c47a0 DW |
2869 | /** |
2870 | * Allocate space to hold new disk entries, raid-device entries or a | |
2871 | * new mpb if necessary. We currently maintain an mpb large enough to | |
2872 | * hold 2 subarrays for the given number of disks. This may not be | |
2873 | * sufficient when reshaping. | |
8273f55e DW |
2874 | * |
2875 | * FIX ME handle the reshape case. | |
2876 | * | |
2877 | * The monitor will be able to safely change super->mpb by arranging | |
2878 | * for it to be freed in check_update_queue(). I.e. the monitor thread | |
2879 | * will start using the new pointer and the manager can continue to use | |
2880 | * the old value until check_update_queue() runs. | |
2881 | */ | |
949c47a0 DW |
2882 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
2883 | ||
2884 | switch (type) { | |
2885 | case update_create_array: { | |
2886 | struct imsm_update_create_array *u = (void *) update->buf; | |
3393c6af | 2887 | size_t len = sizeof_imsm_dev(&u->dev, 1); |
949c47a0 DW |
2888 | |
2889 | update->space = malloc(len); | |
2890 | break; | |
2891 | default: | |
2892 | break; | |
2893 | } | |
2894 | } | |
8273f55e DW |
2895 | |
2896 | return; | |
2897 | } | |
2898 | ||
ae6aad82 DW |
2899 | /* must be called while manager is quiesced */ |
2900 | static void imsm_delete(struct intel_super *super, struct dl **dlp) | |
2901 | { | |
2902 | struct imsm_super *mpb = super->anchor; | |
2903 | struct dl *dl = *dlp; | |
2904 | struct dl *iter; | |
2905 | struct imsm_dev *dev; | |
2906 | struct imsm_map *map; | |
2907 | int i, j; | |
2908 | ||
2909 | dprintf("%s: deleting device %x:%x from imsm_super\n", | |
2910 | __func__, dl->major, dl->minor); | |
2911 | ||
2912 | /* shift all indexes down one */ | |
2913 | for (iter = super->disks; iter; iter = iter->next) | |
2914 | if (iter->index > dl->index) | |
2915 | iter->index--; | |
2916 | ||
2917 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
2918 | dev = get_imsm_dev(super, i); | |
2919 | map = get_imsm_map(dev, 0); | |
2920 | ||
2921 | for (j = 0; j < map->num_members; j++) { | |
2922 | int idx = get_imsm_disk_idx(map, j); | |
2923 | ||
2924 | if (idx > dl->index) | |
2925 | map->disk_ord_tbl[j] = __cpu_to_le32(idx - 1); | |
2926 | } | |
2927 | } | |
2928 | ||
2929 | mpb->num_disks--; | |
2930 | super->updates_pending++; | |
2931 | *dlp = (*dlp)->next; | |
2932 | __free_imsm_disk(dl); | |
2933 | } | |
2934 | ||
cdddbdbc DW |
2935 | struct superswitch super_imsm = { |
2936 | #ifndef MDASSEMBLE | |
2937 | .examine_super = examine_super_imsm, | |
2938 | .brief_examine_super = brief_examine_super_imsm, | |
2939 | .detail_super = detail_super_imsm, | |
2940 | .brief_detail_super = brief_detail_super_imsm, | |
bf5a934a | 2941 | .write_init_super = write_init_super_imsm, |
cdddbdbc DW |
2942 | #endif |
2943 | .match_home = match_home_imsm, | |
2944 | .uuid_from_super= uuid_from_super_imsm, | |
2945 | .getinfo_super = getinfo_super_imsm, | |
2946 | .update_super = update_super_imsm, | |
2947 | ||
2948 | .avail_size = avail_size_imsm, | |
2949 | ||
2950 | .compare_super = compare_super_imsm, | |
2951 | ||
2952 | .load_super = load_super_imsm, | |
bf5a934a DW |
2953 | .init_super = init_super_imsm, |
2954 | .add_to_super = add_to_super_imsm, | |
cdddbdbc DW |
2955 | .store_super = store_zero_imsm, |
2956 | .free_super = free_super_imsm, | |
2957 | .match_metadata_desc = match_metadata_desc_imsm, | |
bf5a934a | 2958 | .container_content = container_content_imsm, |
cdddbdbc DW |
2959 | |
2960 | .validate_geometry = validate_geometry_imsm, | |
cdddbdbc | 2961 | .external = 1, |
845dea95 NB |
2962 | |
2963 | /* for mdmon */ | |
2964 | .open_new = imsm_open_new, | |
2965 | .load_super = load_super_imsm, | |
ed9d66aa | 2966 | .set_array_state= imsm_set_array_state, |
845dea95 NB |
2967 | .set_disk = imsm_set_disk, |
2968 | .sync_metadata = imsm_sync_metadata, | |
88758e9d | 2969 | .activate_spare = imsm_activate_spare, |
e8319a19 | 2970 | .process_update = imsm_process_update, |
8273f55e | 2971 | .prepare_update = imsm_prepare_update, |
cdddbdbc | 2972 | }; |