1 /*-
2 * Copyright (c) 2010 Alexander Motin <mav@FreeBSD.org>
3 * Copyright (c) 2000 - 2008 Søren Schmidt <sos@FreeBSD.org>
4 * All rights reserved.
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
8 * are met:
9 * 1. Redistributions of source code must retain the above copyright
10 * notice, this list of conditions and the following disclaimer.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
14 *
15 * THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
16 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
17 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
18 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
19 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
20 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
21 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
22 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
23 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
24 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
25 * SUCH DAMAGE.
26 */
27
28 #include <sys/cdefs.h>
29 __FBSDID("$FreeBSD$");
30
31 #include <sys/param.h>
32 #include <sys/bio.h>
33 #include <sys/endian.h>
34 #include <sys/kernel.h>
35 #include <sys/kobj.h>
36 #include <sys/limits.h>
37 #include <sys/lock.h>
38 #include <sys/malloc.h>
39 #include <sys/mutex.h>
40 #include <sys/systm.h>
41 #include <sys/taskqueue.h>
42 #include <geom/geom.h>
43 #include "geom/raid/g_raid.h"
44 #include "g_raid_md_if.h"
45
46 static MALLOC_DEFINE(M_MD_INTEL, "md_intel_data", "GEOM_RAID Intel metadata");
47
48 struct intel_raid_map {
49 uint32_t offset;
50 uint32_t disk_sectors;
51 uint32_t stripe_count;
52 uint16_t strip_sectors;
53 uint8_t status;
54 #define INTEL_S_READY 0x00
55 #define INTEL_S_UNINITIALIZED 0x01
56 #define INTEL_S_DEGRADED 0x02
57 #define INTEL_S_FAILURE 0x03
58
59 uint8_t type;
60 #define INTEL_T_RAID0 0x00
61 #define INTEL_T_RAID1 0x01
62 #define INTEL_T_RAID5 0x05
63
64 uint8_t total_disks;
65 uint8_t total_domains;
66 uint8_t failed_disk_num;
67 uint8_t ddf;
68 uint32_t offset_hi;
69 uint32_t disk_sectors_hi;
70 uint32_t stripe_count_hi;
71 uint32_t filler_2[4];
72 uint32_t disk_idx[1]; /* total_disks entries. */
73 #define INTEL_DI_IDX 0x00ffffff
74 #define INTEL_DI_RBLD 0x01000000
75 } __packed;
76
77 struct intel_raid_vol {
78 uint8_t name[16];
79 u_int64_t total_sectors __packed;
80 uint32_t state;
81 #define INTEL_ST_BOOTABLE 0x00000001
82 #define INTEL_ST_BOOT_DEVICE 0x00000002
83 #define INTEL_ST_READ_COALESCING 0x00000004
84 #define INTEL_ST_WRITE_COALESCING 0x00000008
85 #define INTEL_ST_LAST_SHUTDOWN_DIRTY 0x00000010
86 #define INTEL_ST_HIDDEN_AT_BOOT 0x00000020
87 #define INTEL_ST_CURRENTLY_HIDDEN 0x00000040
88 #define INTEL_ST_VERIFY_AND_FIX 0x00000080
89 #define INTEL_ST_MAP_STATE_UNINIT 0x00000100
90 #define INTEL_ST_NO_AUTO_RECOVERY 0x00000200
91 #define INTEL_ST_CLONE_N_GO 0x00000400
92 #define INTEL_ST_CLONE_MAN_SYNC 0x00000800
93 #define INTEL_ST_CNG_MASTER_DISK_NUM 0x00001000
94 uint32_t reserved;
95 uint8_t migr_priority;
96 uint8_t num_sub_vols;
97 uint8_t tid;
98 uint8_t cng_master_disk;
99 uint16_t cache_policy;
100 uint8_t cng_state;
101 #define INTEL_CNGST_UPDATED 0
102 #define INTEL_CNGST_NEEDS_UPDATE 1
103 #define INTEL_CNGST_MASTER_MISSING 2
104 uint8_t cng_sub_state;
105 uint32_t filler_0[10];
106
107 uint32_t curr_migr_unit;
108 uint32_t checkpoint_id;
109 uint8_t migr_state;
110 uint8_t migr_type;
111 #define INTEL_MT_INIT 0
112 #define INTEL_MT_REBUILD 1
113 #define INTEL_MT_VERIFY 2
114 #define INTEL_MT_GEN_MIGR 3
115 #define INTEL_MT_STATE_CHANGE 4
116 #define INTEL_MT_REPAIR 5
117 uint8_t dirty;
118 uint8_t fs_state;
119 uint16_t verify_errors;
120 uint16_t bad_blocks;
121 uint32_t curr_migr_unit_hi;
122 uint32_t filler_1[3];
123 struct intel_raid_map map[1]; /* 2 entries if migr_state != 0. */
124 } __packed;
125
126 struct intel_raid_disk {
127 #define INTEL_SERIAL_LEN 16
128 uint8_t serial[INTEL_SERIAL_LEN];
129 uint32_t sectors;
130 uint32_t id;
131 uint32_t flags;
132 #define INTEL_F_SPARE 0x01
133 #define INTEL_F_ASSIGNED 0x02
134 #define INTEL_F_FAILED 0x04
135 #define INTEL_F_ONLINE 0x08
136 #define INTEL_F_DISABLED 0x80
137 uint32_t owner_cfg_num;
138 uint32_t sectors_hi;
139 uint32_t filler[3];
140 } __packed;
141
142 struct intel_raid_conf {
143 uint8_t intel_id[24];
144 #define INTEL_MAGIC "Intel Raid ISM Cfg Sig. "
145
146 uint8_t version[6];
147 #define INTEL_VERSION_1000 "1.0.00" /* RAID0 */
148 #define INTEL_VERSION_1100 "1.1.00" /* RAID1 */
149 #define INTEL_VERSION_1200 "1.2.00" /* Many volumes */
150 #define INTEL_VERSION_1201 "1.2.01" /* 3 or 4 disks */
151 #define INTEL_VERSION_1202 "1.2.02" /* RAID5 */
152 #define INTEL_VERSION_1204 "1.2.04" /* 5 or 6 disks */
153 #define INTEL_VERSION_1206 "1.2.06" /* CNG */
154 #define INTEL_VERSION_1300 "1.3.00" /* Attributes */
155
156 uint8_t dummy_0[2];
157 uint32_t checksum;
158 uint32_t config_size;
159 uint32_t config_id;
160 uint32_t generation;
161 uint32_t error_log_size;
162 uint32_t attributes;
163 #define INTEL_ATTR_RAID0 0x00000001
164 #define INTEL_ATTR_RAID1 0x00000002
165 #define INTEL_ATTR_RAID10 0x00000004
166 #define INTEL_ATTR_RAID1E 0x00000008
167 #define INTEL_ATTR_RAID5 0x00000010
168 #define INTEL_ATTR_RAIDCNG 0x00000020
169 #define INTEL_ATTR_EXT_STRIP 0x00000040
170 #define INTEL_ATTR_NVM_CACHE 0x02000000
171 #define INTEL_ATTR_2TB_DISK 0x04000000
172 #define INTEL_ATTR_BBM 0x08000000
173 #define INTEL_ATTR_NVM_CACHE2 0x10000000
174 #define INTEL_ATTR_2TB 0x20000000
175 #define INTEL_ATTR_PM 0x40000000
176 #define INTEL_ATTR_CHECKSUM 0x80000000
177
178 uint8_t total_disks;
179 uint8_t total_volumes;
180 uint8_t error_log_pos;
181 uint8_t dummy_2[1];
182 uint32_t cache_size;
183 uint32_t orig_config_id;
184 uint32_t pwr_cycle_count;
185 uint32_t bbm_log_size;
186 uint32_t filler_0[35];
187 struct intel_raid_disk disk[1]; /* total_disks entries. */
188 /* Here goes total_volumes of struct intel_raid_vol. */
189 } __packed;
190
191 #define INTEL_ATTR_SUPPORTED ( INTEL_ATTR_RAID0 | INTEL_ATTR_RAID1 | \
192 INTEL_ATTR_RAID10 | INTEL_ATTR_RAID1E | INTEL_ATTR_RAID5 | \
193 INTEL_ATTR_RAIDCNG | INTEL_ATTR_EXT_STRIP | INTEL_ATTR_2TB_DISK | \
194 INTEL_ATTR_2TB | INTEL_ATTR_PM | INTEL_ATTR_CHECKSUM )
195
196 #define INTEL_MAX_MD_SIZE(ndisks) \
197 (sizeof(struct intel_raid_conf) + \
198 sizeof(struct intel_raid_disk) * (ndisks - 1) + \
199 sizeof(struct intel_raid_vol) * 2 + \
200 sizeof(struct intel_raid_map) * 2 + \
201 sizeof(uint32_t) * (ndisks - 1) * 4)
202
203 struct g_raid_md_intel_perdisk {
204 struct intel_raid_conf *pd_meta;
205 int pd_disk_pos;
206 struct intel_raid_disk pd_disk_meta;
207 };
208
209 struct g_raid_md_intel_pervolume {
210 int pv_volume_pos;
211 int pv_cng;
212 int pv_cng_man_sync;
213 int pv_cng_master_disk;
214 };
215
216 struct g_raid_md_intel_object {
217 struct g_raid_md_object mdio_base;
218 uint32_t mdio_config_id;
219 uint32_t mdio_orig_config_id;
220 uint32_t mdio_generation;
221 struct intel_raid_conf *mdio_meta;
222 struct callout mdio_start_co; /* STARTING state timer. */
223 int mdio_disks_present;
224 int mdio_started;
225 int mdio_incomplete;
226 struct root_hold_token *mdio_rootmount; /* Root mount delay token. */
227 };
228
229 static g_raid_md_create_t g_raid_md_create_intel;
230 static g_raid_md_taste_t g_raid_md_taste_intel;
231 static g_raid_md_event_t g_raid_md_event_intel;
232 static g_raid_md_ctl_t g_raid_md_ctl_intel;
233 static g_raid_md_write_t g_raid_md_write_intel;
234 static g_raid_md_fail_disk_t g_raid_md_fail_disk_intel;
235 static g_raid_md_free_disk_t g_raid_md_free_disk_intel;
236 static g_raid_md_free_volume_t g_raid_md_free_volume_intel;
237 static g_raid_md_free_t g_raid_md_free_intel;
238
239 static kobj_method_t g_raid_md_intel_methods[] = {
240 KOBJMETHOD(g_raid_md_create, g_raid_md_create_intel),
241 KOBJMETHOD(g_raid_md_taste, g_raid_md_taste_intel),
242 KOBJMETHOD(g_raid_md_event, g_raid_md_event_intel),
243 KOBJMETHOD(g_raid_md_ctl, g_raid_md_ctl_intel),
244 KOBJMETHOD(g_raid_md_write, g_raid_md_write_intel),
245 KOBJMETHOD(g_raid_md_fail_disk, g_raid_md_fail_disk_intel),
246 KOBJMETHOD(g_raid_md_free_disk, g_raid_md_free_disk_intel),
247 KOBJMETHOD(g_raid_md_free_volume, g_raid_md_free_volume_intel),
248 KOBJMETHOD(g_raid_md_free, g_raid_md_free_intel),
249 { 0, 0 }
250 };
251
252 static struct g_raid_md_class g_raid_md_intel_class = {
253 "Intel",
254 g_raid_md_intel_methods,
255 sizeof(struct g_raid_md_intel_object),
256 .mdc_enable = 1,
257 .mdc_priority = 100
258 };
259
260
261 static struct intel_raid_map *
262 intel_get_map(struct intel_raid_vol *mvol, int i)
263 {
264 struct intel_raid_map *mmap;
265
266 if (i > (mvol->migr_state ? 1 : 0))
267 return (NULL);
268 mmap = &mvol->map[0];
269 for (; i > 0; i--) {
270 mmap = (struct intel_raid_map *)
271 &mmap->disk_idx[mmap->total_disks];
272 }
273 return ((struct intel_raid_map *)mmap);
274 }
275
276 static struct intel_raid_vol *
277 intel_get_volume(struct intel_raid_conf *meta, int i)
278 {
279 struct intel_raid_vol *mvol;
280 struct intel_raid_map *mmap;
281
282 if (i > 1)
283 return (NULL);
284 mvol = (struct intel_raid_vol *)&meta->disk[meta->total_disks];
285 for (; i > 0; i--) {
286 mmap = intel_get_map(mvol, mvol->migr_state ? 1 : 0);
287 mvol = (struct intel_raid_vol *)
288 &mmap->disk_idx[mmap->total_disks];
289 }
290 return (mvol);
291 }
292
293 static off_t
294 intel_get_map_offset(struct intel_raid_map *mmap)
295 {
296 off_t offset = (off_t)mmap->offset_hi << 32;
297
298 offset += mmap->offset;
299 return (offset);
300 }
301
302 static void
303 intel_set_map_offset(struct intel_raid_map *mmap, off_t offset)
304 {
305
306 mmap->offset = offset & 0xffffffff;
307 mmap->offset_hi = offset >> 32;
308 }
309
310 static off_t
311 intel_get_map_disk_sectors(struct intel_raid_map *mmap)
312 {
313 off_t disk_sectors = (off_t)mmap->disk_sectors_hi << 32;
314
315 disk_sectors += mmap->disk_sectors;
316 return (disk_sectors);
317 }
318
319 static void
320 intel_set_map_disk_sectors(struct intel_raid_map *mmap, off_t disk_sectors)
321 {
322
323 mmap->disk_sectors = disk_sectors & 0xffffffff;
324 mmap->disk_sectors_hi = disk_sectors >> 32;
325 }
326
327 static void
328 intel_set_map_stripe_count(struct intel_raid_map *mmap, off_t stripe_count)
329 {
330
331 mmap->stripe_count = stripe_count & 0xffffffff;
332 mmap->stripe_count_hi = stripe_count >> 32;
333 }
334
335 static off_t
336 intel_get_disk_sectors(struct intel_raid_disk *disk)
337 {
338 off_t sectors = (off_t)disk->sectors_hi << 32;
339
340 sectors += disk->sectors;
341 return (sectors);
342 }
343
344 static void
345 intel_set_disk_sectors(struct intel_raid_disk *disk, off_t sectors)
346 {
347
348 disk->sectors = sectors & 0xffffffff;
349 disk->sectors_hi = sectors >> 32;
350 }
351
352 static off_t
353 intel_get_vol_curr_migr_unit(struct intel_raid_vol *vol)
354 {
355 off_t curr_migr_unit = (off_t)vol->curr_migr_unit_hi << 32;
356
357 curr_migr_unit += vol->curr_migr_unit;
358 return (curr_migr_unit);
359 }
360
361 static void
362 intel_set_vol_curr_migr_unit(struct intel_raid_vol *vol, off_t curr_migr_unit)
363 {
364
365 vol->curr_migr_unit = curr_migr_unit & 0xffffffff;
366 vol->curr_migr_unit_hi = curr_migr_unit >> 32;
367 }
368
369 static char *
370 intel_status2str(int status)
371 {
372
373 switch (status) {
374 case INTEL_S_READY:
375 return ("READY");
376 case INTEL_S_UNINITIALIZED:
377 return ("UNINITIALIZED");
378 case INTEL_S_DEGRADED:
379 return ("DEGRADED");
380 case INTEL_S_FAILURE:
381 return ("FAILURE");
382 default:
383 return ("UNKNOWN");
384 }
385 }
386
387 static char *
388 intel_type2str(int type)
389 {
390
391 switch (type) {
392 case INTEL_T_RAID0:
393 return ("RAID0");
394 case INTEL_T_RAID1:
395 return ("RAID1");
396 case INTEL_T_RAID5:
397 return ("RAID5");
398 default:
399 return ("UNKNOWN");
400 }
401 }
402
403 static char *
404 intel_cngst2str(int cng_state)
405 {
406
407 switch (cng_state) {
408 case INTEL_CNGST_UPDATED:
409 return ("UPDATED");
410 case INTEL_CNGST_NEEDS_UPDATE:
411 return ("NEEDS_UPDATE");
412 case INTEL_CNGST_MASTER_MISSING:
413 return ("MASTER_MISSING");
414 default:
415 return ("UNKNOWN");
416 }
417 }
418
419 static char *
420 intel_mt2str(int type)
421 {
422
423 switch (type) {
424 case INTEL_MT_INIT:
425 return ("INIT");
426 case INTEL_MT_REBUILD:
427 return ("REBUILD");
428 case INTEL_MT_VERIFY:
429 return ("VERIFY");
430 case INTEL_MT_GEN_MIGR:
431 return ("GEN_MIGR");
432 case INTEL_MT_STATE_CHANGE:
433 return ("STATE_CHANGE");
434 case INTEL_MT_REPAIR:
435 return ("REPAIR");
436 default:
437 return ("UNKNOWN");
438 }
439 }
440
441 static void
442 g_raid_md_intel_print(struct intel_raid_conf *meta)
443 {
444 struct intel_raid_vol *mvol;
445 struct intel_raid_map *mmap;
446 int i, j, k;
447
448 if (g_raid_debug < 1)
449 return;
450
451 printf("********* ATA Intel MatrixRAID Metadata *********\n");
452 printf("intel_id <%.24s>\n", meta->intel_id);
453 printf("version <%.6s>\n", meta->version);
454 printf("checksum 0x%08x\n", meta->checksum);
455 printf("config_size 0x%08x\n", meta->config_size);
456 printf("config_id 0x%08x\n", meta->config_id);
457 printf("generation 0x%08x\n", meta->generation);
458 printf("error_log_size %d\n", meta->error_log_size);
459 printf("attributes 0x%b\n", meta->attributes,
460 "\020"
461 "\001RAID0"
462 "\002RAID1"
463 "\003RAID10"
464 "\004RAID1E"
465 "\005RAID15"
466 "\006RAIDCNG"
467 "\007EXT_STRIP"
468 "\032NVM_CACHE"
469 "\0332TB_DISK"
470 "\034BBM"
471 "\035NVM_CACHE"
472 "\0362TB"
473 "\037PM"
474 "\040CHECKSUM");
475 printf("total_disks %u\n", meta->total_disks);
476 printf("total_volumes %u\n", meta->total_volumes);
477 printf("error_log_pos %u\n", meta->error_log_pos);
478 printf("cache_size %u\n", meta->cache_size);
479 printf("orig_config_id 0x%08x\n", meta->orig_config_id);
480 printf("pwr_cycle_count %u\n", meta->pwr_cycle_count);
481 printf("bbm_log_size %u\n", meta->bbm_log_size);
482 printf("Flags: S - Spare, A - Assigned, F - Failed, O - Online, D - Disabled\n");
483 printf("DISK# serial disk_sectors disk_sectors_hi disk_id flags owner\n");
484 for (i = 0; i < meta->total_disks; i++ ) {
485 printf(" %d <%.16s> %u %u 0x%08x 0x%b %08x\n", i,
486 meta->disk[i].serial, meta->disk[i].sectors,
487 meta->disk[i].sectors_hi, meta->disk[i].id,
488 meta->disk[i].flags, "\2\01S\02A\03F\04O\05D",
489 meta->disk[i].owner_cfg_num);
490 }
491 for (i = 0; i < meta->total_volumes; i++) {
492 mvol = intel_get_volume(meta, i);
493 printf(" ****** Volume %d ******\n", i);
494 printf(" name %.16s\n", mvol->name);
495 printf(" total_sectors %ju\n", mvol->total_sectors);
496 printf(" state 0x%b\n", mvol->state,
497 "\020"
498 "\001BOOTABLE"
499 "\002BOOT_DEVICE"
500 "\003READ_COALESCING"
501 "\004WRITE_COALESCING"
502 "\005LAST_SHUTDOWN_DIRTY"
503 "\006HIDDEN_AT_BOOT"
504 "\007CURRENTLY_HIDDEN"
505 "\010VERIFY_AND_FIX"
506 "\011MAP_STATE_UNINIT"
507 "\012NO_AUTO_RECOVERY"
508 "\013CLONE_N_GO"
509 "\014CLONE_MAN_SYNC"
510 "\015CNG_MASTER_DISK_NUM");
511 printf(" reserved %u\n", mvol->reserved);
512 printf(" migr_priority %u\n", mvol->migr_priority);
513 printf(" num_sub_vols %u\n", mvol->num_sub_vols);
514 printf(" tid %u\n", mvol->tid);
515 printf(" cng_master_disk %u\n", mvol->cng_master_disk);
516 printf(" cache_policy %u\n", mvol->cache_policy);
517 printf(" cng_state %u (%s)\n", mvol->cng_state,
518 intel_cngst2str(mvol->cng_state));
519 printf(" cng_sub_state %u\n", mvol->cng_sub_state);
520 printf(" curr_migr_unit %u\n", mvol->curr_migr_unit);
521 printf(" curr_migr_unit_hi %u\n", mvol->curr_migr_unit_hi);
522 printf(" checkpoint_id %u\n", mvol->checkpoint_id);
523 printf(" migr_state %u\n", mvol->migr_state);
524 printf(" migr_type %u (%s)\n", mvol->migr_type,
525 intel_mt2str(mvol->migr_type));
526 printf(" dirty %u\n", mvol->dirty);
527 printf(" fs_state %u\n", mvol->fs_state);
528 printf(" verify_errors %u\n", mvol->verify_errors);
529 printf(" bad_blocks %u\n", mvol->bad_blocks);
530
531 for (j = 0; j < (mvol->migr_state ? 2 : 1); j++) {
532 printf(" *** Map %d ***\n", j);
533 mmap = intel_get_map(mvol, j);
534 printf(" offset %u\n", mmap->offset);
535 printf(" offset_hi %u\n", mmap->offset_hi);
536 printf(" disk_sectors %u\n", mmap->disk_sectors);
537 printf(" disk_sectors_hi %u\n", mmap->disk_sectors_hi);
538 printf(" stripe_count %u\n", mmap->stripe_count);
539 printf(" stripe_count_hi %u\n", mmap->stripe_count_hi);
540 printf(" strip_sectors %u\n", mmap->strip_sectors);
541 printf(" status %u (%s)\n", mmap->status,
542 intel_status2str(mmap->status));
543 printf(" type %u (%s)\n", mmap->type,
544 intel_type2str(mmap->type));
545 printf(" total_disks %u\n", mmap->total_disks);
546 printf(" total_domains %u\n", mmap->total_domains);
547 printf(" failed_disk_num %u\n", mmap->failed_disk_num);
548 printf(" ddf %u\n", mmap->ddf);
549 printf(" disk_idx ");
550 for (k = 0; k < mmap->total_disks; k++)
551 printf(" 0x%08x", mmap->disk_idx[k]);
552 printf("\n");
553 }
554 }
555 printf("=================================================\n");
556 }
557
558 static struct intel_raid_conf *
559 intel_meta_copy(struct intel_raid_conf *meta)
560 {
561 struct intel_raid_conf *nmeta;
562
563 nmeta = malloc(meta->config_size, M_MD_INTEL, M_WAITOK);
564 memcpy(nmeta, meta, meta->config_size);
565 return (nmeta);
566 }
567
568 static int
569 intel_meta_find_disk(struct intel_raid_conf *meta, char *serial)
570 {
571 int pos;
572
573 for (pos = 0; pos < meta->total_disks; pos++) {
574 if (strncmp(meta->disk[pos].serial,
575 serial, INTEL_SERIAL_LEN) == 0)
576 return (pos);
577 }
578 return (-1);
579 }
580
581 static struct intel_raid_conf *
582 intel_meta_read(struct g_consumer *cp)
583 {
584 struct g_provider *pp;
585 struct intel_raid_conf *meta;
586 struct intel_raid_vol *mvol;
587 struct intel_raid_map *mmap, *mmap1;
588 char *buf;
589 int error, i, j, k, left, size;
590 uint32_t checksum, *ptr;
591
592 pp = cp->provider;
593
594 /* Read the anchor sector. */
595 buf = g_read_data(cp,
596 pp->mediasize - pp->sectorsize * 2, pp->sectorsize, &error);
597 if (buf == NULL) {
598 G_RAID_DEBUG(1, "Cannot read metadata from %s (error=%d).",
599 pp->name, error);
600 return (NULL);
601 }
602 meta = (struct intel_raid_conf *)buf;
603
604 /* Check if this is an Intel RAID struct */
605 if (strncmp(meta->intel_id, INTEL_MAGIC, strlen(INTEL_MAGIC))) {
606 G_RAID_DEBUG(1, "Intel signature check failed on %s", pp->name);
607 g_free(buf);
608 return (NULL);
609 }
610 if (meta->config_size > 65536 ||
611 meta->config_size < sizeof(struct intel_raid_conf)) {
612 G_RAID_DEBUG(1, "Intel metadata size looks wrong: %d",
613 meta->config_size);
614 g_free(buf);
615 return (NULL);
616 }
617 size = meta->config_size;
618 meta = malloc(size, M_MD_INTEL, M_WAITOK);
619 memcpy(meta, buf, min(size, pp->sectorsize));
620 g_free(buf);
621
622 /* Read all the rest, if needed. */
623 if (meta->config_size > pp->sectorsize) {
624 left = (meta->config_size - 1) / pp->sectorsize;
625 buf = g_read_data(cp,
626 pp->mediasize - pp->sectorsize * (2 + left),
627 pp->sectorsize * left, &error);
628 if (buf == NULL) {
629 G_RAID_DEBUG(1, "Cannot read remaining metadata"
630 " part from %s (error=%d).",
631 pp->name, error);
632 free(meta, M_MD_INTEL);
633 return (NULL);
634 }
635 memcpy(((char *)meta) + pp->sectorsize, buf,
636 pp->sectorsize * left);
637 g_free(buf);
638 }
639
640 /* Check metadata checksum. */
641 for (checksum = 0, ptr = (uint32_t *)meta, i = 0;
642 i < (meta->config_size / sizeof(uint32_t)); i++) {
643 checksum += *ptr++;
644 }
645 checksum -= meta->checksum;
646 if (checksum != meta->checksum) {
647 G_RAID_DEBUG(1, "Intel checksum check failed on %s", pp->name);
648 free(meta, M_MD_INTEL);
649 return (NULL);
650 }
651
652 /* Validate metadata size. */
653 size = sizeof(struct intel_raid_conf) +
654 sizeof(struct intel_raid_disk) * (meta->total_disks - 1) +
655 sizeof(struct intel_raid_vol) * meta->total_volumes;
656 if (size > meta->config_size) {
657 badsize:
658 G_RAID_DEBUG(1, "Intel metadata size incorrect %d < %d",
659 meta->config_size, size);
660 free(meta, M_MD_INTEL);
661 return (NULL);
662 }
663 for (i = 0; i < meta->total_volumes; i++) {
664 mvol = intel_get_volume(meta, i);
665 mmap = intel_get_map(mvol, 0);
666 size += 4 * (mmap->total_disks - 1);
667 if (size > meta->config_size)
668 goto badsize;
669 if (mvol->migr_state) {
670 size += sizeof(struct intel_raid_map);
671 if (size > meta->config_size)
672 goto badsize;
673 mmap = intel_get_map(mvol, 1);
674 size += 4 * (mmap->total_disks - 1);
675 if (size > meta->config_size)
676 goto badsize;
677 }
678 }
679
680 g_raid_md_intel_print(meta);
681
682 if (strncmp(meta->version, INTEL_VERSION_1300, 6) > 0) {
683 G_RAID_DEBUG(1, "Intel unsupported version: '%.6s'",
684 meta->version);
685 free(meta, M_MD_INTEL);
686 return (NULL);
687 }
688
689 if (strncmp(meta->version, INTEL_VERSION_1300, 6) >= 0 &&
690 (meta->attributes & ~INTEL_ATTR_SUPPORTED) != 0) {
691 G_RAID_DEBUG(1, "Intel unsupported attributes: 0x%08x",
692 meta->attributes & ~INTEL_ATTR_SUPPORTED);
693 free(meta, M_MD_INTEL);
694 return (NULL);
695 }
696
697 /* Validate disk indexes. */
698 for (i = 0; i < meta->total_volumes; i++) {
699 mvol = intel_get_volume(meta, i);
700 for (j = 0; j < (mvol->migr_state ? 2 : 1); j++) {
701 mmap = intel_get_map(mvol, j);
702 for (k = 0; k < mmap->total_disks; k++) {
703 if ((mmap->disk_idx[k] & INTEL_DI_IDX) >
704 meta->total_disks) {
705 G_RAID_DEBUG(1, "Intel metadata disk"
706 " index %d too big (>%d)",
707 mmap->disk_idx[k] & INTEL_DI_IDX,
708 meta->total_disks);
709 free(meta, M_MD_INTEL);
710 return (NULL);
711 }
712 }
713 }
714 }
715
716 /* Validate migration types. */
717 for (i = 0; i < meta->total_volumes; i++) {
718 mvol = intel_get_volume(meta, i);
719 /* Deny unknown migration types. */
720 if (mvol->migr_state &&
721 mvol->migr_type != INTEL_MT_INIT &&
722 mvol->migr_type != INTEL_MT_REBUILD &&
723 mvol->migr_type != INTEL_MT_VERIFY &&
724 mvol->migr_type != INTEL_MT_GEN_MIGR &&
725 mvol->migr_type != INTEL_MT_REPAIR) {
726 G_RAID_DEBUG(1, "Intel metadata has unsupported"
727 " migration type %d", mvol->migr_type);
728 free(meta, M_MD_INTEL);
729 return (NULL);
730 }
731 /* Deny general migrations except SINGLE->RAID1. */
732 if (mvol->migr_state &&
733 mvol->migr_type == INTEL_MT_GEN_MIGR) {
734 mmap = intel_get_map(mvol, 0);
735 mmap1 = intel_get_map(mvol, 1);
736 if (mmap1->total_disks != 1 ||
737 mmap->type != INTEL_T_RAID1 ||
738 mmap->total_disks != 2 ||
739 mmap->offset != mmap1->offset ||
740 mmap->disk_sectors != mmap1->disk_sectors ||
741 mmap->total_domains != mmap->total_disks ||
742 mmap->offset_hi != mmap1->offset_hi ||
743 mmap->disk_sectors_hi != mmap1->disk_sectors_hi ||
744 (mmap->disk_idx[0] != mmap1->disk_idx[0] &&
745 mmap->disk_idx[0] != mmap1->disk_idx[1])) {
746 G_RAID_DEBUG(1, "Intel metadata has unsupported"
747 " variant of general migration");
748 free(meta, M_MD_INTEL);
749 return (NULL);
750 }
751 }
752 }
753
754 return (meta);
755 }
756
757 static int
758 intel_meta_write(struct g_consumer *cp, struct intel_raid_conf *meta)
759 {
760 struct g_provider *pp;
761 char *buf;
762 int error, i, sectors;
763 uint32_t checksum, *ptr;
764
765 pp = cp->provider;
766
767 /* Recalculate checksum for case if metadata were changed. */
768 meta->checksum = 0;
769 for (checksum = 0, ptr = (uint32_t *)meta, i = 0;
770 i < (meta->config_size / sizeof(uint32_t)); i++) {
771 checksum += *ptr++;
772 }
773 meta->checksum = checksum;
774
775 /* Create and fill buffer. */
776 sectors = (meta->config_size + pp->sectorsize - 1) / pp->sectorsize;
777 buf = malloc(sectors * pp->sectorsize, M_MD_INTEL, M_WAITOK | M_ZERO);
778 if (sectors > 1) {
779 memcpy(buf, ((char *)meta) + pp->sectorsize,
780 (sectors - 1) * pp->sectorsize);
781 }
782 memcpy(buf + (sectors - 1) * pp->sectorsize, meta, pp->sectorsize);
783
784 error = g_write_data(cp,
785 pp->mediasize - pp->sectorsize * (1 + sectors),
786 buf, pp->sectorsize * sectors);
787 if (error != 0) {
788 G_RAID_DEBUG(1, "Cannot write metadata to %s (error=%d).",
789 pp->name, error);
790 }
791
792 free(buf, M_MD_INTEL);
793 return (error);
794 }
795
796 static int
797 intel_meta_erase(struct g_consumer *cp)
798 {
799 struct g_provider *pp;
800 char *buf;
801 int error;
802
803 pp = cp->provider;
804 buf = malloc(pp->sectorsize, M_MD_INTEL, M_WAITOK | M_ZERO);
805 error = g_write_data(cp,
806 pp->mediasize - 2 * pp->sectorsize,
807 buf, pp->sectorsize);
808 if (error != 0) {
809 G_RAID_DEBUG(1, "Cannot erase metadata on %s (error=%d).",
810 pp->name, error);
811 }
812 free(buf, M_MD_INTEL);
813 return (error);
814 }
815
816 static int
817 intel_meta_write_spare(struct g_consumer *cp, struct intel_raid_disk *d)
818 {
819 struct intel_raid_conf *meta;
820 int error;
821
822 /* Fill anchor and single disk. */
823 meta = malloc(INTEL_MAX_MD_SIZE(1), M_MD_INTEL, M_WAITOK | M_ZERO);
824 memcpy(&meta->intel_id[0], INTEL_MAGIC, sizeof(INTEL_MAGIC) - 1);
825 memcpy(&meta->version[0], INTEL_VERSION_1000,
826 sizeof(INTEL_VERSION_1000) - 1);
827 meta->config_size = INTEL_MAX_MD_SIZE(1);
828 meta->config_id = meta->orig_config_id = arc4random();
829 meta->generation = 1;
830 meta->total_disks = 1;
831 meta->disk[0] = *d;
832 error = intel_meta_write(cp, meta);
833 free(meta, M_MD_INTEL);
834 return (error);
835 }
836
837 static struct g_raid_disk *
838 g_raid_md_intel_get_disk(struct g_raid_softc *sc, int id)
839 {
840 struct g_raid_disk *disk;
841 struct g_raid_md_intel_perdisk *pd;
842
843 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
844 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
845 if (pd->pd_disk_pos == id)
846 break;
847 }
848 return (disk);
849 }
850
851 static int
852 g_raid_md_intel_supported(int level, int qual, int disks, int force)
853 {
854
855 switch (level) {
856 case G_RAID_VOLUME_RL_RAID0:
857 if (disks < 1)
858 return (0);
859 if (!force && (disks < 2 || disks > 6))
860 return (0);
861 break;
862 case G_RAID_VOLUME_RL_RAID1:
863 if (disks < 1)
864 return (0);
865 if (!force && (disks != 2))
866 return (0);
867 break;
868 case G_RAID_VOLUME_RL_RAID1E:
869 if (disks < 2)
870 return (0);
871 if (!force && (disks != 4))
872 return (0);
873 break;
874 case G_RAID_VOLUME_RL_RAID5:
875 if (disks < 3)
876 return (0);
877 if (!force && disks > 6)
878 return (0);
879 if (qual != G_RAID_VOLUME_RLQ_R5LA)
880 return (0);
881 break;
882 default:
883 return (0);
884 }
885 if (level != G_RAID_VOLUME_RL_RAID5 && qual != G_RAID_VOLUME_RLQ_NONE)
886 return (0);
887 return (1);
888 }
889
890 static struct g_raid_volume *
891 g_raid_md_intel_get_volume(struct g_raid_softc *sc, int id)
892 {
893 struct g_raid_volume *mvol;
894 struct g_raid_md_intel_pervolume *pv;
895
896 TAILQ_FOREACH(mvol, &sc->sc_volumes, v_next) {
897 pv = mvol->v_md_data;
898 if (pv->pv_volume_pos == id)
899 break;
900 }
901 return (mvol);
902 }
903
904 static int
905 g_raid_md_intel_start_disk(struct g_raid_disk *disk)
906 {
907 struct g_raid_softc *sc;
908 struct g_raid_subdisk *sd, *tmpsd;
909 struct g_raid_disk *olddisk, *tmpdisk;
910 struct g_raid_md_object *md;
911 struct g_raid_md_intel_object *mdi;
912 struct g_raid_md_intel_pervolume *pv;
913 struct g_raid_md_intel_perdisk *pd, *oldpd;
914 struct intel_raid_conf *meta;
915 struct intel_raid_vol *mvol;
916 struct intel_raid_map *mmap0, *mmap1;
917 int disk_pos, resurrection = 0, migr_global, i;
918
919 sc = disk->d_softc;
920 md = sc->sc_md;
921 mdi = (struct g_raid_md_intel_object *)md;
922 meta = mdi->mdio_meta;
923 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
924 olddisk = NULL;
925
926 /* Find disk position in metadata by it's serial. */
927 disk_pos = intel_meta_find_disk(meta, pd->pd_disk_meta.serial);
928 if (disk_pos < 0) {
929 G_RAID_DEBUG1(1, sc, "Unknown, probably new or stale disk");
930 /* Failed stale disk is useless for us. */
931 if ((pd->pd_disk_meta.flags & INTEL_F_FAILED) &&
932 !(pd->pd_disk_meta.flags & INTEL_F_DISABLED)) {
933 g_raid_change_disk_state(disk, G_RAID_DISK_S_STALE_FAILED);
934 return (0);
935 }
936 /* If we are in the start process, that's all for now. */
937 if (!mdi->mdio_started)
938 goto nofit;
939 /*
940 * If we have already started - try to get use of the disk.
941 * Try to replace OFFLINE disks first, then FAILED.
942 */
943 TAILQ_FOREACH(tmpdisk, &sc->sc_disks, d_next) {
944 if (tmpdisk->d_state != G_RAID_DISK_S_OFFLINE &&
945 tmpdisk->d_state != G_RAID_DISK_S_FAILED)
946 continue;
947 /* Make sure this disk is big enough. */
948 TAILQ_FOREACH(sd, &tmpdisk->d_subdisks, sd_next) {
949 off_t disk_sectors =
950 intel_get_disk_sectors(&pd->pd_disk_meta);
951
952 if (sd->sd_offset + sd->sd_size + 4096 >
953 disk_sectors * 512) {
954 G_RAID_DEBUG1(1, sc,
955 "Disk too small (%llu < %llu)",
956 (unsigned long long)
957 disk_sectors * 512,
958 (unsigned long long)
959 sd->sd_offset + sd->sd_size + 4096);
960 break;
961 }
962 }
963 if (sd != NULL)
964 continue;
965 if (tmpdisk->d_state == G_RAID_DISK_S_OFFLINE) {
966 olddisk = tmpdisk;
967 break;
968 } else if (olddisk == NULL)
969 olddisk = tmpdisk;
970 }
971 if (olddisk == NULL) {
972 nofit:
973 if (pd->pd_disk_meta.flags & INTEL_F_SPARE) {
974 g_raid_change_disk_state(disk,
975 G_RAID_DISK_S_SPARE);
976 return (1);
977 } else {
978 g_raid_change_disk_state(disk,
979 G_RAID_DISK_S_STALE);
980 return (0);
981 }
982 }
983 oldpd = (struct g_raid_md_intel_perdisk *)olddisk->d_md_data;
984 disk_pos = oldpd->pd_disk_pos;
985 resurrection = 1;
986 }
987
988 if (olddisk == NULL) {
989 /* Find placeholder by position. */
990 olddisk = g_raid_md_intel_get_disk(sc, disk_pos);
991 if (olddisk == NULL)
992 panic("No disk at position %d!", disk_pos);
993 if (olddisk->d_state != G_RAID_DISK_S_OFFLINE) {
994 G_RAID_DEBUG1(1, sc, "More then one disk for pos %d",
995 disk_pos);
996 g_raid_change_disk_state(disk, G_RAID_DISK_S_STALE);
997 return (0);
998 }
999 oldpd = (struct g_raid_md_intel_perdisk *)olddisk->d_md_data;
1000 }
1001
1002 /* Replace failed disk or placeholder with new disk. */
1003 TAILQ_FOREACH_SAFE(sd, &olddisk->d_subdisks, sd_next, tmpsd) {
1004 TAILQ_REMOVE(&olddisk->d_subdisks, sd, sd_next);
1005 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
1006 sd->sd_disk = disk;
1007 }
1008 oldpd->pd_disk_pos = -2;
1009 pd->pd_disk_pos = disk_pos;
1010
1011 /* If it was placeholder -- destroy it. */
1012 if (olddisk->d_state == G_RAID_DISK_S_OFFLINE) {
1013 g_raid_destroy_disk(olddisk);
1014 } else {
1015 /* Otherwise, make it STALE_FAILED. */
1016 g_raid_change_disk_state(olddisk, G_RAID_DISK_S_STALE_FAILED);
1017 /* Update global metadata just in case. */
1018 memcpy(&meta->disk[disk_pos], &pd->pd_disk_meta,
1019 sizeof(struct intel_raid_disk));
1020 }
1021
1022 /* Welcome the new disk. */
1023 if ((meta->disk[disk_pos].flags & INTEL_F_DISABLED) &&
1024 !(pd->pd_disk_meta.flags & INTEL_F_SPARE))
1025 g_raid_change_disk_state(disk, G_RAID_DISK_S_DISABLED);
1026 else if (resurrection)
1027 g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE);
1028 else if (meta->disk[disk_pos].flags & INTEL_F_FAILED)
1029 g_raid_change_disk_state(disk, G_RAID_DISK_S_FAILED);
1030 else if (meta->disk[disk_pos].flags & INTEL_F_SPARE)
1031 g_raid_change_disk_state(disk, G_RAID_DISK_S_SPARE);
1032 else
1033 g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE);
1034 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
1035 pv = sd->sd_volume->v_md_data;
1036 mvol = intel_get_volume(meta, pv->pv_volume_pos);
1037 mmap0 = intel_get_map(mvol, 0);
1038 if (mvol->migr_state)
1039 mmap1 = intel_get_map(mvol, 1);
1040 else
1041 mmap1 = mmap0;
1042
1043 migr_global = 1;
1044 for (i = 0; i < mmap0->total_disks; i++) {
1045 if ((mmap0->disk_idx[i] & INTEL_DI_RBLD) == 0 &&
1046 (mmap1->disk_idx[i] & INTEL_DI_RBLD) != 0)
1047 migr_global = 0;
1048 }
1049
1050 if ((meta->disk[disk_pos].flags & INTEL_F_DISABLED) &&
1051 !(pd->pd_disk_meta.flags & INTEL_F_SPARE)) {
1052 /* Disabled disk, useless. */
1053 g_raid_change_subdisk_state(sd,
1054 G_RAID_SUBDISK_S_NONE);
1055 } else if (resurrection) {
1056 /* Stale disk, almost same as new. */
1057 g_raid_change_subdisk_state(sd,
1058 G_RAID_SUBDISK_S_NEW);
1059 } else if (meta->disk[disk_pos].flags & INTEL_F_FAILED) {
1060 /* Failed disk, almost useless. */
1061 g_raid_change_subdisk_state(sd,
1062 G_RAID_SUBDISK_S_FAILED);
1063 } else if (mvol->migr_state == 0) {
1064 if (mmap0->status == INTEL_S_UNINITIALIZED &&
1065 (!pv->pv_cng || pv->pv_cng_master_disk != disk_pos)) {
1066 /* Freshly created uninitialized volume. */
1067 g_raid_change_subdisk_state(sd,
1068 G_RAID_SUBDISK_S_UNINITIALIZED);
1069 } else if (mmap0->disk_idx[sd->sd_pos] & INTEL_DI_RBLD) {
1070 /* Freshly inserted disk. */
1071 g_raid_change_subdisk_state(sd,
1072 G_RAID_SUBDISK_S_NEW);
1073 } else if (mvol->dirty && (!pv->pv_cng ||
1074 pv->pv_cng_master_disk != disk_pos)) {
1075 /* Dirty volume (unclean shutdown). */
1076 g_raid_change_subdisk_state(sd,
1077 G_RAID_SUBDISK_S_STALE);
1078 } else {
1079 /* Up to date disk. */
1080 g_raid_change_subdisk_state(sd,
1081 G_RAID_SUBDISK_S_ACTIVE);
1082 }
1083 } else if (mvol->migr_type == INTEL_MT_INIT ||
1084 mvol->migr_type == INTEL_MT_REBUILD) {
1085 if (mmap0->disk_idx[sd->sd_pos] & INTEL_DI_RBLD) {
1086 /* Freshly inserted disk. */
1087 g_raid_change_subdisk_state(sd,
1088 G_RAID_SUBDISK_S_NEW);
1089 } else if (mmap1->disk_idx[sd->sd_pos] & INTEL_DI_RBLD) {
1090 /* Rebuilding disk. */
1091 g_raid_change_subdisk_state(sd,
1092 G_RAID_SUBDISK_S_REBUILD);
1093 if (mvol->dirty) {
1094 sd->sd_rebuild_pos = 0;
1095 } else {
1096 sd->sd_rebuild_pos =
1097 intel_get_vol_curr_migr_unit(mvol) *
1098 sd->sd_volume->v_strip_size *
1099 mmap0->total_domains;
1100 }
1101 } else if (mvol->migr_type == INTEL_MT_INIT &&
1102 migr_global) {
1103 /* Freshly created uninitialized volume. */
1104 g_raid_change_subdisk_state(sd,
1105 G_RAID_SUBDISK_S_UNINITIALIZED);
1106 } else if (mvol->dirty && (!pv->pv_cng ||
1107 pv->pv_cng_master_disk != disk_pos)) {
1108 /* Dirty volume (unclean shutdown). */
1109 g_raid_change_subdisk_state(sd,
1110 G_RAID_SUBDISK_S_STALE);
1111 } else {
1112 /* Up to date disk. */
1113 g_raid_change_subdisk_state(sd,
1114 G_RAID_SUBDISK_S_ACTIVE);
1115 }
1116 } else if (mvol->migr_type == INTEL_MT_VERIFY ||
1117 mvol->migr_type == INTEL_MT_REPAIR) {
1118 if (mmap0->disk_idx[sd->sd_pos] & INTEL_DI_RBLD) {
1119 /* Freshly inserted disk. */
1120 g_raid_change_subdisk_state(sd,
1121 G_RAID_SUBDISK_S_NEW);
1122 } else if ((mmap1->disk_idx[sd->sd_pos] & INTEL_DI_RBLD) ||
1123 migr_global) {
1124 /* Resyncing disk. */
1125 g_raid_change_subdisk_state(sd,
1126 G_RAID_SUBDISK_S_RESYNC);
1127 if (mvol->dirty) {
1128 sd->sd_rebuild_pos = 0;
1129 } else {
1130 sd->sd_rebuild_pos =
1131 intel_get_vol_curr_migr_unit(mvol) *
1132 sd->sd_volume->v_strip_size *
1133 mmap0->total_domains;
1134 }
1135 } else if (mvol->dirty) {
1136 /* Dirty volume (unclean shutdown). */
1137 g_raid_change_subdisk_state(sd,
1138 G_RAID_SUBDISK_S_STALE);
1139 } else {
1140 /* Up to date disk. */
1141 g_raid_change_subdisk_state(sd,
1142 G_RAID_SUBDISK_S_ACTIVE);
1143 }
1144 } else if (mvol->migr_type == INTEL_MT_GEN_MIGR) {
1145 if ((mmap1->disk_idx[0] & INTEL_DI_IDX) != disk_pos) {
1146 /* Freshly inserted disk. */
1147 g_raid_change_subdisk_state(sd,
1148 G_RAID_SUBDISK_S_NEW);
1149 } else {
1150 /* Up to date disk. */
1151 g_raid_change_subdisk_state(sd,
1152 G_RAID_SUBDISK_S_ACTIVE);
1153 }
1154 }
1155 g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW,
1156 G_RAID_EVENT_SUBDISK);
1157 }
1158
1159 /* Update status of our need for spare. */
1160 if (mdi->mdio_started) {
1161 mdi->mdio_incomplete =
1162 (g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) +
1163 g_raid_ndisks(sc, G_RAID_DISK_S_DISABLED) <
1164 meta->total_disks);
1165 }
1166
1167 return (resurrection);
1168 }
1169
1170 static void
1171 g_disk_md_intel_retaste(void *arg, int pending)
1172 {
1173
1174 G_RAID_DEBUG(1, "Array is not complete, trying to retaste.");
1175 g_retaste(&g_raid_class);
1176 free(arg, M_MD_INTEL);
1177 }
1178
1179 static void
1180 g_raid_md_intel_refill(struct g_raid_softc *sc)
1181 {
1182 struct g_raid_md_object *md;
1183 struct g_raid_md_intel_object *mdi;
1184 struct intel_raid_conf *meta;
1185 struct g_raid_disk *disk;
1186 struct task *task;
1187 int update, na;
1188
1189 md = sc->sc_md;
1190 mdi = (struct g_raid_md_intel_object *)md;
1191 meta = mdi->mdio_meta;
1192 update = 0;
1193 do {
1194 /* Make sure we miss anything. */
1195 na = g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) +
1196 g_raid_ndisks(sc, G_RAID_DISK_S_DISABLED);
1197 if (na == meta->total_disks)
1198 break;
1199
1200 G_RAID_DEBUG1(1, md->mdo_softc,
1201 "Array is not complete (%d of %d), "
1202 "trying to refill.", na, meta->total_disks);
1203
1204 /* Try to get use some of STALE disks. */
1205 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1206 if (disk->d_state == G_RAID_DISK_S_STALE) {
1207 update += g_raid_md_intel_start_disk(disk);
1208 if (disk->d_state == G_RAID_DISK_S_ACTIVE ||
1209 disk->d_state == G_RAID_DISK_S_DISABLED)
1210 break;
1211 }
1212 }
1213 if (disk != NULL)
1214 continue;
1215
1216 /* Try to get use some of SPARE disks. */
1217 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1218 if (disk->d_state == G_RAID_DISK_S_SPARE) {
1219 update += g_raid_md_intel_start_disk(disk);
1220 if (disk->d_state == G_RAID_DISK_S_ACTIVE)
1221 break;
1222 }
1223 }
1224 } while (disk != NULL);
1225
1226 /* Write new metadata if we changed something. */
1227 if (update) {
1228 g_raid_md_write_intel(md, NULL, NULL, NULL);
1229 meta = mdi->mdio_meta;
1230 }
1231
1232 /* Update status of our need for spare. */
1233 mdi->mdio_incomplete = (g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) +
1234 g_raid_ndisks(sc, G_RAID_DISK_S_DISABLED) < meta->total_disks);
1235
1236 /* Request retaste hoping to find spare. */
1237 if (mdi->mdio_incomplete) {
1238 task = malloc(sizeof(struct task),
1239 M_MD_INTEL, M_WAITOK | M_ZERO);
1240 TASK_INIT(task, 0, g_disk_md_intel_retaste, task);
1241 taskqueue_enqueue(taskqueue_swi, task);
1242 }
1243 }
1244
1245 static void
1246 g_raid_md_intel_start(struct g_raid_softc *sc)
1247 {
1248 struct g_raid_md_object *md;
1249 struct g_raid_md_intel_object *mdi;
1250 struct g_raid_md_intel_pervolume *pv;
1251 struct g_raid_md_intel_perdisk *pd;
1252 struct intel_raid_conf *meta;
1253 struct intel_raid_vol *mvol;
1254 struct intel_raid_map *mmap;
1255 struct g_raid_volume *vol;
1256 struct g_raid_subdisk *sd;
1257 struct g_raid_disk *disk;
1258 int i, j, disk_pos;
1259
1260 md = sc->sc_md;
1261 mdi = (struct g_raid_md_intel_object *)md;
1262 meta = mdi->mdio_meta;
1263
1264 /* Create volumes and subdisks. */
1265 for (i = 0; i < meta->total_volumes; i++) {
1266 mvol = intel_get_volume(meta, i);
1267 mmap = intel_get_map(mvol, 0);
1268 vol = g_raid_create_volume(sc, mvol->name, mvol->tid - 1);
1269 pv = malloc(sizeof(*pv), M_MD_INTEL, M_WAITOK | M_ZERO);
1270 pv->pv_volume_pos = i;
1271 pv->pv_cng = (mvol->state & INTEL_ST_CLONE_N_GO) != 0;
1272 pv->pv_cng_man_sync = (mvol->state & INTEL_ST_CLONE_MAN_SYNC) != 0;
1273 if (mvol->cng_master_disk < mmap->total_disks)
1274 pv->pv_cng_master_disk = mvol->cng_master_disk;
1275 vol->v_md_data = pv;
1276 vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_NONE;
1277 if (mmap->type == INTEL_T_RAID0)
1278 vol->v_raid_level = G_RAID_VOLUME_RL_RAID0;
1279 else if (mmap->type == INTEL_T_RAID1 &&
1280 mmap->total_domains >= 2 &&
1281 mmap->total_domains <= mmap->total_disks) {
1282 /* Assume total_domains is correct. */
1283 if (mmap->total_domains == mmap->total_disks)
1284 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1;
1285 else
1286 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1E;
1287 } else if (mmap->type == INTEL_T_RAID1) {
1288 /* total_domains looks wrong. */
1289 if (mmap->total_disks <= 2)
1290 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1;
1291 else
1292 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1E;
1293 } else if (mmap->type == INTEL_T_RAID5) {
1294 vol->v_raid_level = G_RAID_VOLUME_RL_RAID5;
1295 vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_R5LA;
1296 } else
1297 vol->v_raid_level = G_RAID_VOLUME_RL_UNKNOWN;
1298 vol->v_strip_size = (u_int)mmap->strip_sectors * 512; //ZZZ
1299 vol->v_disks_count = mmap->total_disks;
1300 vol->v_mediasize = (off_t)mvol->total_sectors * 512; //ZZZ
1301 vol->v_sectorsize = 512; //ZZZ
1302 for (j = 0; j < vol->v_disks_count; j++) {
1303 sd = &vol->v_subdisks[j];
1304 sd->sd_offset = intel_get_map_offset(mmap) * 512; //ZZZ
1305 sd->sd_size = intel_get_map_disk_sectors(mmap) * 512; //ZZZ
1306 }
1307 g_raid_start_volume(vol);
1308 }
1309
1310 /* Create disk placeholders to store data for later writing. */
1311 for (disk_pos = 0; disk_pos < meta->total_disks; disk_pos++) {
1312 pd = malloc(sizeof(*pd), M_MD_INTEL, M_WAITOK | M_ZERO);
1313 pd->pd_disk_pos = disk_pos;
1314 pd->pd_disk_meta = meta->disk[disk_pos];
1315 disk = g_raid_create_disk(sc);
1316 disk->d_md_data = (void *)pd;
1317 disk->d_state = G_RAID_DISK_S_OFFLINE;
1318 for (i = 0; i < meta->total_volumes; i++) {
1319 mvol = intel_get_volume(meta, i);
1320 mmap = intel_get_map(mvol, 0);
1321 for (j = 0; j < mmap->total_disks; j++) {
1322 if ((mmap->disk_idx[j] & INTEL_DI_IDX) == disk_pos)
1323 break;
1324 }
1325 if (j == mmap->total_disks)
1326 continue;
1327 vol = g_raid_md_intel_get_volume(sc, i);
1328 sd = &vol->v_subdisks[j];
1329 sd->sd_disk = disk;
1330 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
1331 }
1332 }
1333
1334 /* Make all disks found till the moment take their places. */
1335 do {
1336 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1337 if (disk->d_state == G_RAID_DISK_S_NONE) {
1338 g_raid_md_intel_start_disk(disk);
1339 break;
1340 }
1341 }
1342 } while (disk != NULL);
1343
1344 mdi->mdio_started = 1;
1345 G_RAID_DEBUG1(0, sc, "Array started.");
1346 g_raid_md_write_intel(md, NULL, NULL, NULL);
1347
1348 /* Pickup any STALE/SPARE disks to refill array if needed. */
1349 g_raid_md_intel_refill(sc);
1350
1351 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
1352 g_raid_event_send(vol, G_RAID_VOLUME_E_START,
1353 G_RAID_EVENT_VOLUME);
1354 }
1355
1356 callout_stop(&mdi->mdio_start_co);
1357 G_RAID_DEBUG1(1, sc, "root_mount_rel %p", mdi->mdio_rootmount);
1358 root_mount_rel(mdi->mdio_rootmount);
1359 mdi->mdio_rootmount = NULL;
1360 }
1361
1362 static void
1363 g_raid_md_intel_new_disk(struct g_raid_disk *disk)
1364 {
1365 struct g_raid_softc *sc;
1366 struct g_raid_md_object *md;
1367 struct g_raid_md_intel_object *mdi;
1368 struct intel_raid_conf *pdmeta;
1369 struct g_raid_md_intel_perdisk *pd;
1370
1371 sc = disk->d_softc;
1372 md = sc->sc_md;
1373 mdi = (struct g_raid_md_intel_object *)md;
1374 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
1375 pdmeta = pd->pd_meta;
1376
1377 if (mdi->mdio_started) {
1378 if (g_raid_md_intel_start_disk(disk))
1379 g_raid_md_write_intel(md, NULL, NULL, NULL);
1380 } else {
1381 /* If we haven't started yet - check metadata freshness. */
1382 if (mdi->mdio_meta == NULL ||
1383 ((int32_t)(pdmeta->generation - mdi->mdio_generation)) > 0) {
1384 G_RAID_DEBUG1(1, sc, "Newer disk");
1385 if (mdi->mdio_meta != NULL)
1386 free(mdi->mdio_meta, M_MD_INTEL);
1387 mdi->mdio_meta = intel_meta_copy(pdmeta);
1388 mdi->mdio_generation = mdi->mdio_meta->generation;
1389 mdi->mdio_disks_present = 1;
1390 } else if (pdmeta->generation == mdi->mdio_generation) {
1391 mdi->mdio_disks_present++;
1392 G_RAID_DEBUG1(1, sc, "Matching disk (%d of %d up)",
1393 mdi->mdio_disks_present,
1394 mdi->mdio_meta->total_disks);
1395 } else {
1396 G_RAID_DEBUG1(1, sc, "Older disk");
1397 }
1398 /* If we collected all needed disks - start array. */
1399 if (mdi->mdio_disks_present == mdi->mdio_meta->total_disks)
1400 g_raid_md_intel_start(sc);
1401 }
1402 }
1403
1404 static void
1405 g_raid_intel_go(void *arg)
1406 {
1407 struct g_raid_softc *sc;
1408 struct g_raid_md_object *md;
1409 struct g_raid_md_intel_object *mdi;
1410
1411 sc = arg;
1412 md = sc->sc_md;
1413 mdi = (struct g_raid_md_intel_object *)md;
1414 if (!mdi->mdio_started) {
1415 G_RAID_DEBUG1(0, sc, "Force array start due to timeout.");
1416 g_raid_event_send(sc, G_RAID_NODE_E_START, 0);
1417 }
1418 }
1419
1420 static int
1421 g_raid_md_create_intel(struct g_raid_md_object *md, struct g_class *mp,
1422 struct g_geom **gp)
1423 {
1424 struct g_raid_softc *sc;
1425 struct g_raid_md_intel_object *mdi;
1426 char name[16];
1427
1428 mdi = (struct g_raid_md_intel_object *)md;
1429 mdi->mdio_config_id = mdi->mdio_orig_config_id = arc4random();
1430 mdi->mdio_generation = 0;
1431 snprintf(name, sizeof(name), "Intel-%08x", mdi->mdio_config_id);
1432 sc = g_raid_create_node(mp, name, md);
1433 if (sc == NULL)
1434 return (G_RAID_MD_TASTE_FAIL);
1435 md->mdo_softc = sc;
1436 *gp = sc->sc_geom;
1437 return (G_RAID_MD_TASTE_NEW);
1438 }
1439
1440 /*
1441 * Return the last N characters of the serial label. The Linux and
1442 * ataraid(7) code always uses the last 16 characters of the label to
1443 * store into the Intel meta format. Generalize this to N characters
1444 * since that's easy. Labels can be up to 20 characters for SATA drives
1445 * and up 251 characters for SAS drives. Since intel controllers don't
1446 * support SAS drives, just stick with the SATA limits for stack friendliness.
1447 */
1448 static int
1449 g_raid_md_get_label(struct g_consumer *cp, char *serial, int serlen)
1450 {
1451 char serial_buffer[24];
1452 int len, error;
1453
1454 len = sizeof(serial_buffer);
1455 error = g_io_getattr("GEOM::ident", cp, &len, serial_buffer);
1456 if (error != 0)
1457 return (error);
1458 len = strlen(serial_buffer);
1459 if (len > serlen)
1460 len -= serlen;
1461 else
1462 len = 0;
1463 strncpy(serial, serial_buffer + len, serlen);
1464 return (0);
1465 }
1466
1467 static int
1468 g_raid_md_taste_intel(struct g_raid_md_object *md, struct g_class *mp,
1469 struct g_consumer *cp, struct g_geom **gp)
1470 {
1471 struct g_consumer *rcp;
1472 struct g_provider *pp;
1473 struct g_raid_md_intel_object *mdi, *mdi1;
1474 struct g_raid_softc *sc;
1475 struct g_raid_disk *disk;
1476 struct intel_raid_conf *meta;
1477 struct g_raid_md_intel_perdisk *pd;
1478 struct g_geom *geom;
1479 int error, disk_pos, result, spare, len;
1480 char serial[INTEL_SERIAL_LEN];
1481 char name[16];
1482 uint16_t vendor;
1483
1484 G_RAID_DEBUG(1, "Tasting Intel on %s", cp->provider->name);
1485 mdi = (struct g_raid_md_intel_object *)md;
1486 pp = cp->provider;
1487
1488 /* Read metadata from device. */
1489 meta = NULL;
1490 vendor = 0xffff;
1491 disk_pos = 0;
1492 g_topology_unlock();
1493 error = g_raid_md_get_label(cp, serial, sizeof(serial));
1494 if (error != 0) {
1495 G_RAID_DEBUG(1, "Cannot get serial number from %s (error=%d).",
1496 pp->name, error);
1497 goto fail2;
1498 }
1499 len = 2;
1500 if (pp->geom->rank == 1)
1501 g_io_getattr("GEOM::hba_vendor", cp, &len, &vendor);
1502 meta = intel_meta_read(cp);
1503 g_topology_lock();
1504 if (meta == NULL) {
1505 if (g_raid_aggressive_spare) {
1506 if (vendor != 0x8086) {
1507 G_RAID_DEBUG(1,
1508 "Intel vendor mismatch 0x%04x != 0x8086",
1509 vendor);
1510 } else {
1511 G_RAID_DEBUG(1,
1512 "No Intel metadata, forcing spare.");
1513 spare = 2;
1514 goto search;
1515 }
1516 }
1517 return (G_RAID_MD_TASTE_FAIL);
1518 }
1519
1520 /* Check this disk position in obtained metadata. */
1521 disk_pos = intel_meta_find_disk(meta, serial);
1522 if (disk_pos < 0) {
1523 G_RAID_DEBUG(1, "Intel serial '%s' not found", serial);
1524 goto fail1;
1525 }
1526 if (intel_get_disk_sectors(&meta->disk[disk_pos]) !=
1527 (pp->mediasize / pp->sectorsize)) {
1528 G_RAID_DEBUG(1, "Intel size mismatch %ju != %ju",
1529 intel_get_disk_sectors(&meta->disk[disk_pos]),
1530 (off_t)(pp->mediasize / pp->sectorsize));
1531 goto fail1;
1532 }
1533
1534 G_RAID_DEBUG(1, "Intel disk position %d", disk_pos);
1535 spare = meta->disk[disk_pos].flags & INTEL_F_SPARE;
1536
1537 search:
1538 /* Search for matching node. */
1539 sc = NULL;
1540 mdi1 = NULL;
1541 LIST_FOREACH(geom, &mp->geom, geom) {
1542 sc = geom->softc;
1543 if (sc == NULL)
1544 continue;
1545 if (sc->sc_stopping != 0)
1546 continue;
1547 if (sc->sc_md->mdo_class != md->mdo_class)
1548 continue;
1549 mdi1 = (struct g_raid_md_intel_object *)sc->sc_md;
1550 if (spare) {
1551 if (mdi1->mdio_incomplete)
1552 break;
1553 } else {
1554 if (mdi1->mdio_config_id == meta->config_id)
1555 break;
1556 }
1557 }
1558
1559 /* Found matching node. */
1560 if (geom != NULL) {
1561 G_RAID_DEBUG(1, "Found matching array %s", sc->sc_name);
1562 result = G_RAID_MD_TASTE_EXISTING;
1563
1564 } else if (spare) { /* Not found needy node -- left for later. */
1565 G_RAID_DEBUG(1, "Spare is not needed at this time");
1566 goto fail1;
1567
1568 } else { /* Not found matching node -- create one. */
1569 result = G_RAID_MD_TASTE_NEW;
1570 mdi->mdio_config_id = meta->config_id;
1571 mdi->mdio_orig_config_id = meta->orig_config_id;
1572 snprintf(name, sizeof(name), "Intel-%08x", meta->config_id);
1573 sc = g_raid_create_node(mp, name, md);
1574 md->mdo_softc = sc;
1575 geom = sc->sc_geom;
1576 callout_init(&mdi->mdio_start_co, 1);
1577 callout_reset(&mdi->mdio_start_co, g_raid_start_timeout * hz,
1578 g_raid_intel_go, sc);
1579 mdi->mdio_rootmount = root_mount_hold("GRAID-Intel");
1580 G_RAID_DEBUG1(1, sc, "root_mount_hold %p", mdi->mdio_rootmount);
1581 }
1582
1583 /* There is no return after this point, so we close passed consumer. */
1584 g_access(cp, -1, 0, 0);
1585
1586 rcp = g_new_consumer(geom);
1587 g_attach(rcp, pp);
1588 if (g_access(rcp, 1, 1, 1) != 0)
1589 ; //goto fail1;
1590
1591 g_topology_unlock();
1592 sx_xlock(&sc->sc_lock);
1593
1594 pd = malloc(sizeof(*pd), M_MD_INTEL, M_WAITOK | M_ZERO);
1595 pd->pd_meta = meta;
1596 pd->pd_disk_pos = -1;
1597 if (spare == 2) {
1598 memcpy(&pd->pd_disk_meta.serial[0], serial, INTEL_SERIAL_LEN);
1599 intel_set_disk_sectors(&pd->pd_disk_meta,
1600 pp->mediasize / pp->sectorsize);
1601 pd->pd_disk_meta.id = 0;
1602 pd->pd_disk_meta.flags = INTEL_F_SPARE;
1603 } else {
1604 pd->pd_disk_meta = meta->disk[disk_pos];
1605 }
1606 disk = g_raid_create_disk(sc);
1607 disk->d_md_data = (void *)pd;
1608 disk->d_consumer = rcp;
1609 rcp->private = disk;
1610
1611 g_raid_get_disk_info(disk);
1612
1613 g_raid_md_intel_new_disk(disk);
1614
1615 sx_xunlock(&sc->sc_lock);
1616 g_topology_lock();
1617 *gp = geom;
1618 return (result);
1619 fail2:
1620 g_topology_lock();
1621 fail1:
1622 free(meta, M_MD_INTEL);
1623 return (G_RAID_MD_TASTE_FAIL);
1624 }
1625
1626 static int
1627 g_raid_md_event_intel(struct g_raid_md_object *md,
1628 struct g_raid_disk *disk, u_int event)
1629 {
1630 struct g_raid_softc *sc;
1631 struct g_raid_subdisk *sd;
1632 struct g_raid_md_intel_object *mdi;
1633 struct g_raid_md_intel_perdisk *pd;
1634
1635 sc = md->mdo_softc;
1636 mdi = (struct g_raid_md_intel_object *)md;
1637 if (disk == NULL) {
1638 switch (event) {
1639 case G_RAID_NODE_E_START:
1640 if (!mdi->mdio_started)
1641 g_raid_md_intel_start(sc);
1642 return (0);
1643 }
1644 return (-1);
1645 }
1646 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
1647 switch (event) {
1648 case G_RAID_DISK_E_DISCONNECTED:
1649 /* If disk was assigned, just update statuses. */
1650 if (pd->pd_disk_pos >= 0) {
1651 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE);
1652 if (disk->d_consumer) {
1653 g_raid_kill_consumer(sc, disk->d_consumer);
1654 disk->d_consumer = NULL;
1655 }
1656 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
1657 g_raid_change_subdisk_state(sd,
1658 G_RAID_SUBDISK_S_NONE);
1659 g_raid_event_send(sd, G_RAID_SUBDISK_E_DISCONNECTED,
1660 G_RAID_EVENT_SUBDISK);
1661 }
1662 } else {
1663 /* Otherwise -- delete. */
1664 g_raid_change_disk_state(disk, G_RAID_DISK_S_NONE);
1665 g_raid_destroy_disk(disk);
1666 }
1667
1668 /* Write updated metadata to all disks. */
1669 g_raid_md_write_intel(md, NULL, NULL, NULL);
1670
1671 /* Check if anything left except placeholders. */
1672 if (g_raid_ndisks(sc, -1) ==
1673 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
1674 g_raid_destroy_node(sc, 0);
1675 else
1676 g_raid_md_intel_refill(sc);
1677 return (0);
1678 }
1679 return (-2);
1680 }
1681
1682 static int
1683 g_raid_md_ctl_intel(struct g_raid_md_object *md,
1684 struct gctl_req *req)
1685 {
1686 struct g_raid_softc *sc;
1687 struct g_raid_volume *vol, *vol1;
1688 struct g_raid_subdisk *sd;
1689 struct g_raid_disk *disk;
1690 struct g_raid_md_intel_object *mdi;
1691 struct g_raid_md_intel_pervolume *pv;
1692 struct g_raid_md_intel_perdisk *pd;
1693 struct g_consumer *cp;
1694 struct g_provider *pp;
1695 char arg[16], serial[INTEL_SERIAL_LEN];
1696 const char *nodename, *verb, *volname, *levelname, *diskname;
1697 char *tmp;
1698 int *nargs, *force;
1699 off_t off, size, sectorsize, strip, disk_sectors;
1700 intmax_t *sizearg, *striparg;
1701 int numdisks, i, len, level, qual, update;
1702 int error;
1703
1704 sc = md->mdo_softc;
1705 mdi = (struct g_raid_md_intel_object *)md;
1706 verb = gctl_get_param(req, "verb", NULL);
1707 nargs = gctl_get_paraml(req, "nargs", sizeof(*nargs));
1708 error = 0;
1709 if (strcmp(verb, "label") == 0) {
1710
1711 if (*nargs < 4) {
1712 gctl_error(req, "Invalid number of arguments.");
1713 return (-1);
1714 }
1715 volname = gctl_get_asciiparam(req, "arg1");
1716 if (volname == NULL) {
1717 gctl_error(req, "No volume name.");
1718 return (-2);
1719 }
1720 levelname = gctl_get_asciiparam(req, "arg2");
1721 if (levelname == NULL) {
1722 gctl_error(req, "No RAID level.");
1723 return (-3);
1724 }
1725 if (strcasecmp(levelname, "RAID5") == 0)
1726 levelname = "RAID5-LA";
1727 if (g_raid_volume_str2level(levelname, &level, &qual)) {
1728 gctl_error(req, "Unknown RAID level '%s'.", levelname);
1729 return (-4);
1730 }
1731 numdisks = *nargs - 3;
1732 force = gctl_get_paraml(req, "force", sizeof(*force));
1733 if (!g_raid_md_intel_supported(level, qual, numdisks,
1734 force ? *force : 0)) {
1735 gctl_error(req, "Unsupported RAID level "
1736 "(0x%02x/0x%02x), or number of disks (%d).",
1737 level, qual, numdisks);
1738 return (-5);
1739 }
1740
1741 /* Search for disks, connect them and probe. */
1742 size = 0x7fffffffffffffffllu;
1743 sectorsize = 0;
1744 for (i = 0; i < numdisks; i++) {
1745 snprintf(arg, sizeof(arg), "arg%d", i + 3);
1746 diskname = gctl_get_asciiparam(req, arg);
1747 if (diskname == NULL) {
1748 gctl_error(req, "No disk name (%s).", arg);
1749 error = -6;
1750 break;
1751 }
1752 if (strcmp(diskname, "NONE") == 0) {
1753 cp = NULL;
1754 pp = NULL;
1755 } else {
1756 g_topology_lock();
1757 cp = g_raid_open_consumer(sc, diskname);
1758 if (cp == NULL) {
1759 gctl_error(req, "Can't open disk '%s'.",
1760 diskname);
1761 g_topology_unlock();
1762 error = -7;
1763 break;
1764 }
1765 pp = cp->provider;
1766 }
1767 pd = malloc(sizeof(*pd), M_MD_INTEL, M_WAITOK | M_ZERO);
1768 pd->pd_disk_pos = i;
1769 disk = g_raid_create_disk(sc);
1770 disk->d_md_data = (void *)pd;
1771 disk->d_consumer = cp;
1772 if (cp == NULL) {
1773 strcpy(&pd->pd_disk_meta.serial[0], "NONE");
1774 pd->pd_disk_meta.id = 0xffffffff;
1775 pd->pd_disk_meta.flags = INTEL_F_ASSIGNED;
1776 continue;
1777 }
1778 cp->private = disk;
1779 g_topology_unlock();
1780
1781 error = g_raid_md_get_label(cp,
1782 &pd->pd_disk_meta.serial[0], INTEL_SERIAL_LEN);
1783 if (error != 0) {
1784 gctl_error(req,
1785 "Can't get serial for provider '%s'.",
1786 diskname);
1787 error = -8;
1788 break;
1789 }
1790
1791 g_raid_get_disk_info(disk);
1792
1793 intel_set_disk_sectors(&pd->pd_disk_meta,
1794 pp->mediasize / pp->sectorsize);
1795 if (size > pp->mediasize)
1796 size = pp->mediasize;
1797 if (sectorsize < pp->sectorsize)
1798 sectorsize = pp->sectorsize;
1799 pd->pd_disk_meta.id = 0;
1800 pd->pd_disk_meta.flags = INTEL_F_ASSIGNED | INTEL_F_ONLINE;
1801 }
1802 if (error != 0)
1803 return (error);
1804
1805 if (sectorsize <= 0) {
1806 gctl_error(req, "Can't get sector size.");
1807 return (-8);
1808 }
1809
1810 /* Reserve some space for metadata. */
1811 size -= ((4096 + sectorsize - 1) / sectorsize) * sectorsize;
1812
1813 /* Handle size argument. */
1814 len = sizeof(*sizearg);
1815 sizearg = gctl_get_param(req, "size", &len);
1816 if (sizearg != NULL && len == sizeof(*sizearg) &&
1817 *sizearg > 0) {
1818 if (*sizearg > size) {
1819 gctl_error(req, "Size too big %lld > %lld.",
1820 (long long)*sizearg, (long long)size);
1821 return (-9);
1822 }
1823 size = *sizearg;
1824 }
1825
1826 /* Handle strip argument. */
1827 strip = 131072;
1828 len = sizeof(*striparg);
1829 striparg = gctl_get_param(req, "strip", &len);
1830 if (striparg != NULL && len == sizeof(*striparg) &&
1831 *striparg > 0) {
1832 if (*striparg < sectorsize) {
1833 gctl_error(req, "Strip size too small.");
1834 return (-10);
1835 }
1836 if (*striparg % sectorsize != 0) {
1837 gctl_error(req, "Incorrect strip size.");
1838 return (-11);
1839 }
1840 if (strip > 65535 * sectorsize) {
1841 gctl_error(req, "Strip size too big.");
1842 return (-12);
1843 }
1844 strip = *striparg;
1845 }
1846
1847 /* Round size down to strip or sector. */
1848 if (level == G_RAID_VOLUME_RL_RAID1)
1849 size -= (size % sectorsize);
1850 else if (level == G_RAID_VOLUME_RL_RAID1E &&
1851 (numdisks & 1) != 0)
1852 size -= (size % (2 * strip));
1853 else
1854 size -= (size % strip);
1855 if (size <= 0) {
1856 gctl_error(req, "Size too small.");
1857 return (-13);
1858 }
1859
1860 /* We have all we need, create things: volume, ... */
1861 mdi->mdio_started = 1;
1862 vol = g_raid_create_volume(sc, volname, -1);
1863 pv = malloc(sizeof(*pv), M_MD_INTEL, M_WAITOK | M_ZERO);
1864 pv->pv_volume_pos = 0;
1865 vol->v_md_data = pv;
1866 vol->v_raid_level = level;
1867 vol->v_raid_level_qualifier = qual;
1868 vol->v_strip_size = strip;
1869 vol->v_disks_count = numdisks;
1870 if (level == G_RAID_VOLUME_RL_RAID0)
1871 vol->v_mediasize = size * numdisks;
1872 else if (level == G_RAID_VOLUME_RL_RAID1)
1873 vol->v_mediasize = size;
1874 else if (level == G_RAID_VOLUME_RL_RAID5)
1875 vol->v_mediasize = size * (numdisks - 1);
1876 else { /* RAID1E */
1877 vol->v_mediasize = ((size * numdisks) / strip / 2) *
1878 strip;
1879 }
1880 vol->v_sectorsize = sectorsize;
1881 g_raid_start_volume(vol);
1882
1883 /* , and subdisks. */
1884 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1885 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
1886 sd = &vol->v_subdisks[pd->pd_disk_pos];
1887 sd->sd_disk = disk;
1888 sd->sd_offset = 0;
1889 sd->sd_size = size;
1890 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
1891 if (sd->sd_disk->d_consumer != NULL) {
1892 g_raid_change_disk_state(disk,
1893 G_RAID_DISK_S_ACTIVE);
1894 if (level == G_RAID_VOLUME_RL_RAID5)
1895 g_raid_change_subdisk_state(sd,
1896 G_RAID_SUBDISK_S_UNINITIALIZED);
1897 else
1898 g_raid_change_subdisk_state(sd,
1899 G_RAID_SUBDISK_S_ACTIVE);
1900 g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW,
1901 G_RAID_EVENT_SUBDISK);
1902 } else {
1903 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE);
1904 }
1905 }
1906
1907 /* Write metadata based on created entities. */
1908 G_RAID_DEBUG1(0, sc, "Array started.");
1909 g_raid_md_write_intel(md, NULL, NULL, NULL);
1910
1911 /* Pickup any STALE/SPARE disks to refill array if needed. */
1912 g_raid_md_intel_refill(sc);
1913
1914 g_raid_event_send(vol, G_RAID_VOLUME_E_START,
1915 G_RAID_EVENT_VOLUME);
1916 return (0);
1917 }
1918 if (strcmp(verb, "add") == 0) {
1919
1920 if (*nargs != 3) {
1921 gctl_error(req, "Invalid number of arguments.");
1922 return (-1);
1923 }
1924 volname = gctl_get_asciiparam(req, "arg1");
1925 if (volname == NULL) {
1926 gctl_error(req, "No volume name.");
1927 return (-2);
1928 }
1929 levelname = gctl_get_asciiparam(req, "arg2");
1930 if (levelname == NULL) {
1931 gctl_error(req, "No RAID level.");
1932 return (-3);
1933 }
1934 if (strcasecmp(levelname, "RAID5") == 0)
1935 levelname = "RAID5-LA";
1936 if (g_raid_volume_str2level(levelname, &level, &qual)) {
1937 gctl_error(req, "Unknown RAID level '%s'.", levelname);
1938 return (-4);
1939 }
1940
1941 /* Look for existing volumes. */
1942 i = 0;
1943 vol1 = NULL;
1944 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
1945 vol1 = vol;
1946 i++;
1947 }
1948 if (i > 1) {
1949 gctl_error(req, "Maximum two volumes supported.");
1950 return (-6);
1951 }
1952 if (vol1 == NULL) {
1953 gctl_error(req, "At least one volume must exist.");
1954 return (-7);
1955 }
1956
1957 numdisks = vol1->v_disks_count;
1958 force = gctl_get_paraml(req, "force", sizeof(*force));
1959 if (!g_raid_md_intel_supported(level, qual, numdisks,
1960 force ? *force : 0)) {
1961 gctl_error(req, "Unsupported RAID level "
1962 "(0x%02x/0x%02x), or number of disks (%d).",
1963 level, qual, numdisks);
1964 return (-5);
1965 }
1966
1967 /* Collect info about present disks. */
1968 size = 0x7fffffffffffffffllu;
1969 sectorsize = 512;
1970 for (i = 0; i < numdisks; i++) {
1971 disk = vol1->v_subdisks[i].sd_disk;
1972 pd = (struct g_raid_md_intel_perdisk *)
1973 disk->d_md_data;
1974 disk_sectors =
1975 intel_get_disk_sectors(&pd->pd_disk_meta);
1976
1977 if (disk_sectors * 512 < size)
1978 size = disk_sectors * 512;
1979 if (disk->d_consumer != NULL &&
1980 disk->d_consumer->provider != NULL &&
1981 disk->d_consumer->provider->sectorsize >
1982 sectorsize) {
1983 sectorsize =
1984 disk->d_consumer->provider->sectorsize;
1985 }
1986 }
1987
1988 /* Reserve some space for metadata. */
1989 size -= ((4096 + sectorsize - 1) / sectorsize) * sectorsize;
1990
1991 /* Decide insert before or after. */
1992 sd = &vol1->v_subdisks[0];
1993 if (sd->sd_offset >
1994 size - (sd->sd_offset + sd->sd_size)) {
1995 off = 0;
1996 size = sd->sd_offset;
1997 } else {
1998 off = sd->sd_offset + sd->sd_size;
1999 size = size - (sd->sd_offset + sd->sd_size);
2000 }
2001
2002 /* Handle strip argument. */
2003 strip = 131072;
2004 len = sizeof(*striparg);
2005 striparg = gctl_get_param(req, "strip", &len);
2006 if (striparg != NULL && len == sizeof(*striparg) &&
2007 *striparg > 0) {
2008 if (*striparg < sectorsize) {
2009 gctl_error(req, "Strip size too small.");
2010 return (-10);
2011 }
2012 if (*striparg % sectorsize != 0) {
2013 gctl_error(req, "Incorrect strip size.");
2014 return (-11);
2015 }
2016 if (strip > 65535 * sectorsize) {
2017 gctl_error(req, "Strip size too big.");
2018 return (-12);
2019 }
2020 strip = *striparg;
2021 }
2022
2023 /* Round offset up to strip. */
2024 if (off % strip != 0) {
2025 size -= strip - off % strip;
2026 off += strip - off % strip;
2027 }
2028
2029 /* Handle size argument. */
2030 len = sizeof(*sizearg);
2031 sizearg = gctl_get_param(req, "size", &len);
2032 if (sizearg != NULL && len == sizeof(*sizearg) &&
2033 *sizearg > 0) {
2034 if (*sizearg > size) {
2035 gctl_error(req, "Size too big %lld > %lld.",
2036 (long long)*sizearg, (long long)size);
2037 return (-9);
2038 }
2039 size = *sizearg;
2040 }
2041
2042 /* Round size down to strip or sector. */
2043 if (level == G_RAID_VOLUME_RL_RAID1)
2044 size -= (size % sectorsize);
2045 else
2046 size -= (size % strip);
2047 if (size <= 0) {
2048 gctl_error(req, "Size too small.");
2049 return (-13);
2050 }
2051 if (size > 0xffffffffllu * sectorsize) {
2052 gctl_error(req, "Size too big.");
2053 return (-14);
2054 }
2055
2056 /* We have all we need, create things: volume, ... */
2057 vol = g_raid_create_volume(sc, volname, -1);
2058 pv = malloc(sizeof(*pv), M_MD_INTEL, M_WAITOK | M_ZERO);
2059 pv->pv_volume_pos = i;
2060 vol->v_md_data = pv;
2061 vol->v_raid_level = level;
2062 vol->v_raid_level_qualifier = qual;
2063 vol->v_strip_size = strip;
2064 vol->v_disks_count = numdisks;
2065 if (level == G_RAID_VOLUME_RL_RAID0)
2066 vol->v_mediasize = size * numdisks;
2067 else if (level == G_RAID_VOLUME_RL_RAID1)
2068 vol->v_mediasize = size;
2069 else if (level == G_RAID_VOLUME_RL_RAID5)
2070 vol->v_mediasize = size * (numdisks - 1);
2071 else { /* RAID1E */
2072 vol->v_mediasize = ((size * numdisks) / strip / 2) *
2073 strip;
2074 }
2075 vol->v_sectorsize = sectorsize;
2076 g_raid_start_volume(vol);
2077
2078 /* , and subdisks. */
2079 for (i = 0; i < numdisks; i++) {
2080 disk = vol1->v_subdisks[i].sd_disk;
2081 sd = &vol->v_subdisks[i];
2082 sd->sd_disk = disk;
2083 sd->sd_offset = off;
2084 sd->sd_size = size;
2085 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
2086 if (disk->d_state == G_RAID_DISK_S_ACTIVE) {
2087 if (level == G_RAID_VOLUME_RL_RAID5)
2088 g_raid_change_subdisk_state(sd,
2089 G_RAID_SUBDISK_S_UNINITIALIZED);
2090 else
2091 g_raid_change_subdisk_state(sd,
2092 G_RAID_SUBDISK_S_ACTIVE);
2093 g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW,
2094 G_RAID_EVENT_SUBDISK);
2095 }
2096 }
2097
2098 /* Write metadata based on created entities. */
2099 g_raid_md_write_intel(md, NULL, NULL, NULL);
2100
2101 g_raid_event_send(vol, G_RAID_VOLUME_E_START,
2102 G_RAID_EVENT_VOLUME);
2103 return (0);
2104 }
2105 if (strcmp(verb, "delete") == 0) {
2106
2107 nodename = gctl_get_asciiparam(req, "arg0");
2108 if (nodename != NULL && strcasecmp(sc->sc_name, nodename) != 0)
2109 nodename = NULL;
2110
2111 /* Full node destruction. */
2112 if (*nargs == 1 && nodename != NULL) {
2113 /* Check if some volume is still open. */
2114 force = gctl_get_paraml(req, "force", sizeof(*force));
2115 if (force != NULL && *force == 0 &&
2116 g_raid_nopens(sc) != 0) {
2117 gctl_error(req, "Some volume is still open.");
2118 return (-4);
2119 }
2120
2121 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2122 if (disk->d_consumer)
2123 intel_meta_erase(disk->d_consumer);
2124 }
2125 g_raid_destroy_node(sc, 0);
2126 return (0);
2127 }
2128
2129 /* Destroy specified volume. If it was last - all node. */
2130 if (*nargs > 2) {
2131 gctl_error(req, "Invalid number of arguments.");
2132 return (-1);
2133 }
2134 volname = gctl_get_asciiparam(req,
2135 nodename != NULL ? "arg1" : "arg0");
2136 if (volname == NULL) {
2137 gctl_error(req, "No volume name.");
2138 return (-2);
2139 }
2140
2141 /* Search for volume. */
2142 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
2143 if (strcmp(vol->v_name, volname) == 0)
2144 break;
2145 pp = vol->v_provider;
2146 if (pp == NULL)
2147 continue;
2148 if (strcmp(pp->name, volname) == 0)
2149 break;
2150 if (strncmp(pp->name, "raid/", 5) == 0 &&
2151 strcmp(pp->name + 5, volname) == 0)
2152 break;
2153 }
2154 if (vol == NULL) {
2155 i = strtol(volname, &tmp, 10);
2156 if (verb != volname && tmp[0] == 0) {
2157 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
2158 if (vol->v_global_id == i)
2159 break;
2160 }
2161 }
2162 }
2163 if (vol == NULL) {
2164 gctl_error(req, "Volume '%s' not found.", volname);
2165 return (-3);
2166 }
2167
2168 /* Check if volume is still open. */
2169 force = gctl_get_paraml(req, "force", sizeof(*force));
2170 if (force != NULL && *force == 0 &&
2171 vol->v_provider_open != 0) {
2172 gctl_error(req, "Volume is still open.");
2173 return (-4);
2174 }
2175
2176 /* Destroy volume and potentially node. */
2177 i = 0;
2178 TAILQ_FOREACH(vol1, &sc->sc_volumes, v_next)
2179 i++;
2180 if (i >= 2) {
2181 g_raid_destroy_volume(vol);
2182 g_raid_md_write_intel(md, NULL, NULL, NULL);
2183 } else {
2184 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2185 if (disk->d_consumer)
2186 intel_meta_erase(disk->d_consumer);
2187 }
2188 g_raid_destroy_node(sc, 0);
2189 }
2190 return (0);
2191 }
2192 if (strcmp(verb, "remove") == 0 ||
2193 strcmp(verb, "fail") == 0) {
2194 if (*nargs < 2) {
2195 gctl_error(req, "Invalid number of arguments.");
2196 return (-1);
2197 }
2198 for (i = 1; i < *nargs; i++) {
2199 snprintf(arg, sizeof(arg), "arg%d", i);
2200 diskname = gctl_get_asciiparam(req, arg);
2201 if (diskname == NULL) {
2202 gctl_error(req, "No disk name (%s).", arg);
2203 error = -2;
2204 break;
2205 }
2206 if (strncmp(diskname, "/dev/", 5) == 0)
2207 diskname += 5;
2208
2209 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2210 if (disk->d_consumer != NULL &&
2211 disk->d_consumer->provider != NULL &&
2212 strcmp(disk->d_consumer->provider->name,
2213 diskname) == 0)
2214 break;
2215 }
2216 if (disk == NULL) {
2217 gctl_error(req, "Disk '%s' not found.",
2218 diskname);
2219 error = -3;
2220 break;
2221 }
2222
2223 if (strcmp(verb, "fail") == 0) {
2224 g_raid_md_fail_disk_intel(md, NULL, disk);
2225 continue;
2226 }
2227
2228 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
2229
2230 /* Erase metadata on deleting disk. */
2231 intel_meta_erase(disk->d_consumer);
2232
2233 /* If disk was assigned, just update statuses. */
2234 if (pd->pd_disk_pos >= 0) {
2235 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE);
2236 g_raid_kill_consumer(sc, disk->d_consumer);
2237 disk->d_consumer = NULL;
2238 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
2239 g_raid_change_subdisk_state(sd,
2240 G_RAID_SUBDISK_S_NONE);
2241 g_raid_event_send(sd, G_RAID_SUBDISK_E_DISCONNECTED,
2242 G_RAID_EVENT_SUBDISK);
2243 }
2244 } else {
2245 /* Otherwise -- delete. */
2246 g_raid_change_disk_state(disk, G_RAID_DISK_S_NONE);
2247 g_raid_destroy_disk(disk);
2248 }
2249 }
2250
2251 /* Write updated metadata to remaining disks. */
2252 g_raid_md_write_intel(md, NULL, NULL, NULL);
2253
2254 /* Check if anything left except placeholders. */
2255 if (g_raid_ndisks(sc, -1) ==
2256 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
2257 g_raid_destroy_node(sc, 0);
2258 else
2259 g_raid_md_intel_refill(sc);
2260 return (error);
2261 }
2262 if (strcmp(verb, "insert") == 0) {
2263 if (*nargs < 2) {
2264 gctl_error(req, "Invalid number of arguments.");
2265 return (-1);
2266 }
2267 update = 0;
2268 for (i = 1; i < *nargs; i++) {
2269 /* Get disk name. */
2270 snprintf(arg, sizeof(arg), "arg%d", i);
2271 diskname = gctl_get_asciiparam(req, arg);
2272 if (diskname == NULL) {
2273 gctl_error(req, "No disk name (%s).", arg);
2274 error = -3;
2275 break;
2276 }
2277
2278 /* Try to find provider with specified name. */
2279 g_topology_lock();
2280 cp = g_raid_open_consumer(sc, diskname);
2281 if (cp == NULL) {
2282 gctl_error(req, "Can't open disk '%s'.",
2283 diskname);
2284 g_topology_unlock();
2285 error = -4;
2286 break;
2287 }
2288 pp = cp->provider;
2289 g_topology_unlock();
2290
2291 /* Read disk serial. */
2292 error = g_raid_md_get_label(cp,
2293 &serial[0], INTEL_SERIAL_LEN);
2294 if (error != 0) {
2295 gctl_error(req,
2296 "Can't get serial for provider '%s'.",
2297 diskname);
2298 g_raid_kill_consumer(sc, cp);
2299 error = -7;
2300 break;
2301 }
2302
2303 pd = malloc(sizeof(*pd), M_MD_INTEL, M_WAITOK | M_ZERO);
2304 pd->pd_disk_pos = -1;
2305
2306 disk = g_raid_create_disk(sc);
2307 disk->d_consumer = cp;
2308 disk->d_md_data = (void *)pd;
2309 cp->private = disk;
2310
2311 g_raid_get_disk_info(disk);
2312
2313 memcpy(&pd->pd_disk_meta.serial[0], &serial[0],
2314 INTEL_SERIAL_LEN);
2315 intel_set_disk_sectors(&pd->pd_disk_meta,
2316 pp->mediasize / pp->sectorsize);
2317 pd->pd_disk_meta.id = 0;
2318 pd->pd_disk_meta.flags = INTEL_F_SPARE;
2319
2320 /* Welcome the "new" disk. */
2321 update += g_raid_md_intel_start_disk(disk);
2322 if (disk->d_state == G_RAID_DISK_S_SPARE) {
2323 intel_meta_write_spare(cp, &pd->pd_disk_meta);
2324 g_raid_destroy_disk(disk);
2325 } else if (disk->d_state != G_RAID_DISK_S_ACTIVE) {
2326 gctl_error(req, "Disk '%s' doesn't fit.",
2327 diskname);
2328 g_raid_destroy_disk(disk);
2329 error = -8;
2330 break;
2331 }
2332 }
2333
2334 /* Write new metadata if we changed something. */
2335 if (update)
2336 g_raid_md_write_intel(md, NULL, NULL, NULL);
2337 return (error);
2338 }
2339 return (-100);
2340 }
2341
2342 static int
2343 g_raid_md_write_intel(struct g_raid_md_object *md, struct g_raid_volume *tvol,
2344 struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk)
2345 {
2346 struct g_raid_softc *sc;
2347 struct g_raid_volume *vol;
2348 struct g_raid_subdisk *sd;
2349 struct g_raid_disk *disk;
2350 struct g_raid_md_intel_object *mdi;
2351 struct g_raid_md_intel_pervolume *pv;
2352 struct g_raid_md_intel_perdisk *pd;
2353 struct intel_raid_conf *meta;
2354 struct intel_raid_vol *mvol;
2355 struct intel_raid_map *mmap0, *mmap1;
2356 off_t sectorsize = 512, pos;
2357 const char *version, *cv;
2358 int vi, sdi, numdisks, len, state, stale;
2359
2360 sc = md->mdo_softc;
2361 mdi = (struct g_raid_md_intel_object *)md;
2362
2363 if (sc->sc_stopping == G_RAID_DESTROY_HARD)
2364 return (0);
2365
2366 /* Bump generation. Newly written metadata may differ from previous. */
2367 mdi->mdio_generation++;
2368
2369 /* Count number of disks. */
2370 numdisks = 0;
2371 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2372 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
2373 if (pd->pd_disk_pos < 0)
2374 continue;
2375 numdisks++;
2376 if (disk->d_state == G_RAID_DISK_S_ACTIVE) {
2377 pd->pd_disk_meta.flags =
2378 INTEL_F_ONLINE | INTEL_F_ASSIGNED;
2379 } else if (disk->d_state == G_RAID_DISK_S_FAILED) {
2380 pd->pd_disk_meta.flags = INTEL_F_FAILED |
2381 INTEL_F_ASSIGNED;
2382 } else if (disk->d_state == G_RAID_DISK_S_DISABLED) {
2383 pd->pd_disk_meta.flags = INTEL_F_FAILED |
2384 INTEL_F_ASSIGNED | INTEL_F_DISABLED;
2385 } else {
2386 if (!(pd->pd_disk_meta.flags & INTEL_F_DISABLED))
2387 pd->pd_disk_meta.flags = INTEL_F_ASSIGNED;
2388 if (pd->pd_disk_meta.id != 0xffffffff) {
2389 pd->pd_disk_meta.id = 0xffffffff;
2390 len = strlen(pd->pd_disk_meta.serial);
2391 len = min(len, INTEL_SERIAL_LEN - 3);
2392 strcpy(pd->pd_disk_meta.serial + len, ":0");
2393 }
2394 }
2395 }
2396
2397 /* Fill anchor and disks. */
2398 meta = malloc(INTEL_MAX_MD_SIZE(numdisks),
2399 M_MD_INTEL, M_WAITOK | M_ZERO);
2400 memcpy(&meta->intel_id[0], INTEL_MAGIC, sizeof(INTEL_MAGIC) - 1);
2401 meta->config_size = INTEL_MAX_MD_SIZE(numdisks);
2402 meta->config_id = mdi->mdio_config_id;
2403 meta->orig_config_id = mdi->mdio_orig_config_id;
2404 meta->generation = mdi->mdio_generation;
2405 meta->attributes = INTEL_ATTR_CHECKSUM;
2406 meta->total_disks = numdisks;
2407 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2408 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
2409 if (pd->pd_disk_pos < 0)
2410 continue;
2411 meta->disk[pd->pd_disk_pos] = pd->pd_disk_meta;
2412 if (pd->pd_disk_meta.sectors_hi != 0)
2413 meta->attributes |= INTEL_ATTR_2TB_DISK;
2414 }
2415
2416 /* Fill volumes and maps. */
2417 vi = 0;
2418 version = INTEL_VERSION_1000;
2419 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
2420 pv = vol->v_md_data;
2421 if (vol->v_stopping)
2422 continue;
2423 mvol = intel_get_volume(meta, vi);
2424
2425 /* New metadata may have different volumes order. */
2426 pv->pv_volume_pos = vi;
2427
2428 for (sdi = 0; sdi < vol->v_disks_count; sdi++) {
2429 sd = &vol->v_subdisks[sdi];
2430 if (sd->sd_disk != NULL)
2431 break;
2432 }
2433 if (sdi >= vol->v_disks_count)
2434 panic("No any filled subdisk in volume");
2435 if (vol->v_mediasize >= 0x20000000000llu)
2436 meta->attributes |= INTEL_ATTR_2TB;
2437 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID0)
2438 meta->attributes |= INTEL_ATTR_RAID0;
2439 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1)
2440 meta->attributes |= INTEL_ATTR_RAID1;
2441 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID5)
2442 meta->attributes |= INTEL_ATTR_RAID5;
2443 else if ((vol->v_disks_count & 1) == 0)
2444 meta->attributes |= INTEL_ATTR_RAID10;
2445 else
2446 meta->attributes |= INTEL_ATTR_RAID1E;
2447 if (pv->pv_cng)
2448 meta->attributes |= INTEL_ATTR_RAIDCNG;
2449 if (vol->v_strip_size > 131072)
2450 meta->attributes |= INTEL_ATTR_EXT_STRIP;
2451
2452 if (pv->pv_cng)
2453 cv = INTEL_VERSION_1206;
2454 else if (vol->v_disks_count > 4)
2455 cv = INTEL_VERSION_1204;
2456 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID5)
2457 cv = INTEL_VERSION_1202;
2458 else if (vol->v_disks_count > 2)
2459 cv = INTEL_VERSION_1201;
2460 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1)
2461 cv = INTEL_VERSION_1100;
2462 else
2463 cv = INTEL_VERSION_1000;
2464 if (strcmp(cv, version) > 0)
2465 version = cv;
2466
2467 strlcpy(&mvol->name[0], vol->v_name, sizeof(mvol->name));
2468 mvol->total_sectors = vol->v_mediasize / sectorsize;
2469 mvol->state = (INTEL_ST_READ_COALESCING |
2470 INTEL_ST_WRITE_COALESCING);
2471 mvol->tid = vol->v_global_id + 1;
2472 if (pv->pv_cng) {
2473 mvol->state |= INTEL_ST_CLONE_N_GO;
2474 if (pv->pv_cng_man_sync)
2475 mvol->state |= INTEL_ST_CLONE_MAN_SYNC;
2476 mvol->cng_master_disk = pv->pv_cng_master_disk;
2477 if (vol->v_subdisks[pv->pv_cng_master_disk].sd_state ==
2478 G_RAID_SUBDISK_S_NONE)
2479 mvol->cng_state = INTEL_CNGST_MASTER_MISSING;
2480 else if (vol->v_state != G_RAID_VOLUME_S_OPTIMAL)
2481 mvol->cng_state = INTEL_CNGST_NEEDS_UPDATE;
2482 else
2483 mvol->cng_state = INTEL_CNGST_UPDATED;
2484 }
2485
2486 /* Check for any recovery in progress. */
2487 state = G_RAID_SUBDISK_S_ACTIVE;
2488 pos = 0x7fffffffffffffffllu;
2489 stale = 0;
2490 for (sdi = 0; sdi < vol->v_disks_count; sdi++) {
2491 sd = &vol->v_subdisks[sdi];
2492 if (sd->sd_state == G_RAID_SUBDISK_S_REBUILD)
2493 state = G_RAID_SUBDISK_S_REBUILD;
2494 else if (sd->sd_state == G_RAID_SUBDISK_S_RESYNC &&
2495 state != G_RAID_SUBDISK_S_REBUILD)
2496 state = G_RAID_SUBDISK_S_RESYNC;
2497 else if (sd->sd_state == G_RAID_SUBDISK_S_STALE)
2498 stale = 1;
2499 if ((sd->sd_state == G_RAID_SUBDISK_S_REBUILD ||
2500 sd->sd_state == G_RAID_SUBDISK_S_RESYNC) &&
2501 sd->sd_rebuild_pos < pos)
2502 pos = sd->sd_rebuild_pos;
2503 }
2504 if (state == G_RAID_SUBDISK_S_REBUILD) {
2505 mvol->migr_state = 1;
2506 mvol->migr_type = INTEL_MT_REBUILD;
2507 } else if (state == G_RAID_SUBDISK_S_RESYNC) {
2508 mvol->migr_state = 1;
2509 /* mvol->migr_type = INTEL_MT_REPAIR; */
2510 mvol->migr_type = INTEL_MT_VERIFY;
2511 mvol->state |= INTEL_ST_VERIFY_AND_FIX;
2512 } else
2513 mvol->migr_state = 0;
2514 mvol->dirty = (vol->v_dirty || stale);
2515
2516 mmap0 = intel_get_map(mvol, 0);
2517
2518 /* Write map / common part of two maps. */
2519 intel_set_map_offset(mmap0, sd->sd_offset / sectorsize);
2520 intel_set_map_disk_sectors(mmap0, sd->sd_size / sectorsize);
2521 mmap0->strip_sectors = vol->v_strip_size / sectorsize;
2522 if (vol->v_state == G_RAID_VOLUME_S_BROKEN)
2523 mmap0->status = INTEL_S_FAILURE;
2524 else if (vol->v_state == G_RAID_VOLUME_S_DEGRADED)
2525 mmap0->status = INTEL_S_DEGRADED;
2526 else if (g_raid_nsubdisks(vol, G_RAID_SUBDISK_S_UNINITIALIZED)
2527 == g_raid_nsubdisks(vol, -1))
2528 mmap0->status = INTEL_S_UNINITIALIZED;
2529 else
2530 mmap0->status = INTEL_S_READY;
2531 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID0)
2532 mmap0->type = INTEL_T_RAID0;
2533 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1 ||
2534 vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E)
2535 mmap0->type = INTEL_T_RAID1;
2536 else
2537 mmap0->type = INTEL_T_RAID5;
2538 mmap0->total_disks = vol->v_disks_count;
2539 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1)
2540 mmap0->total_domains = vol->v_disks_count;
2541 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E)
2542 mmap0->total_domains = 2;
2543 else
2544 mmap0->total_domains = 1;
2545 intel_set_map_stripe_count(mmap0,
2546 sd->sd_size / vol->v_strip_size / mmap0->total_domains);
2547 mmap0->failed_disk_num = 0xff;
2548 mmap0->ddf = 1;
2549
2550 /* If there are two maps - copy common and update. */
2551 if (mvol->migr_state) {
2552 intel_set_vol_curr_migr_unit(mvol,
2553 pos / vol->v_strip_size / mmap0->total_domains);
2554 mmap1 = intel_get_map(mvol, 1);
2555 memcpy(mmap1, mmap0, sizeof(struct intel_raid_map));
2556 mmap0->status = INTEL_S_READY;
2557 } else
2558 mmap1 = NULL;
2559
2560 /* Write disk indexes and put rebuild flags. */
2561 for (sdi = 0; sdi < vol->v_disks_count; sdi++) {
2562 sd = &vol->v_subdisks[sdi];
2563 pd = (struct g_raid_md_intel_perdisk *)
2564 sd->sd_disk->d_md_data;
2565 mmap0->disk_idx[sdi] = pd->pd_disk_pos;
2566 if (mvol->migr_state)
2567 mmap1->disk_idx[sdi] = pd->pd_disk_pos;
2568 if (sd->sd_state == G_RAID_SUBDISK_S_REBUILD ||
2569 sd->sd_state == G_RAID_SUBDISK_S_RESYNC) {
2570 mmap1->disk_idx[sdi] |= INTEL_DI_RBLD;
2571 } else if (sd->sd_state != G_RAID_SUBDISK_S_ACTIVE &&
2572 sd->sd_state != G_RAID_SUBDISK_S_STALE &&
2573 sd->sd_state != G_RAID_SUBDISK_S_UNINITIALIZED) {
2574 mmap0->disk_idx[sdi] |= INTEL_DI_RBLD;
2575 if (mvol->migr_state)
2576 mmap1->disk_idx[sdi] |= INTEL_DI_RBLD;
2577 }
2578 if ((sd->sd_state == G_RAID_SUBDISK_S_NONE ||
2579 sd->sd_state == G_RAID_SUBDISK_S_FAILED ||
2580 sd->sd_state == G_RAID_SUBDISK_S_REBUILD) &&
2581 mmap0->failed_disk_num == 0xff) {
2582 mmap0->failed_disk_num = sdi;
2583 if (mvol->migr_state)
2584 mmap1->failed_disk_num = sdi;
2585 }
2586 }
2587 vi++;
2588 }
2589 meta->total_volumes = vi;
2590 if (vi > 1 || meta->attributes &
2591 (INTEL_ATTR_EXT_STRIP | INTEL_ATTR_2TB_DISK | INTEL_ATTR_2TB))
2592 version = INTEL_VERSION_1300;
2593 if (strcmp(version, INTEL_VERSION_1300) < 0)
2594 meta->attributes &= INTEL_ATTR_CHECKSUM;
2595 memcpy(&meta->version[0], version, sizeof(INTEL_VERSION_1000) - 1);
2596
2597 /* We are done. Print meta data and store them to disks. */
2598 g_raid_md_intel_print(meta);
2599 if (mdi->mdio_meta != NULL)
2600 free(mdi->mdio_meta, M_MD_INTEL);
2601 mdi->mdio_meta = meta;
2602 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2603 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
2604 if (disk->d_state != G_RAID_DISK_S_ACTIVE)
2605 continue;
2606 if (pd->pd_meta != NULL) {
2607 free(pd->pd_meta, M_MD_INTEL);
2608 pd->pd_meta = NULL;
2609 }
2610 pd->pd_meta = intel_meta_copy(meta);
2611 intel_meta_write(disk->d_consumer, meta);
2612 }
2613 return (0);
2614 }
2615
2616 static int
2617 g_raid_md_fail_disk_intel(struct g_raid_md_object *md,
2618 struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk)
2619 {
2620 struct g_raid_softc *sc;
2621 struct g_raid_md_intel_object *mdi;
2622 struct g_raid_md_intel_perdisk *pd;
2623 struct g_raid_subdisk *sd;
2624
2625 sc = md->mdo_softc;
2626 mdi = (struct g_raid_md_intel_object *)md;
2627 pd = (struct g_raid_md_intel_perdisk *)tdisk->d_md_data;
2628
2629 /* We can't fail disk that is not a part of array now. */
2630 if (pd->pd_disk_pos < 0)
2631 return (-1);
2632
2633 /*
2634 * Mark disk as failed in metadata and try to write that metadata
2635 * to the disk itself to prevent it's later resurrection as STALE.
2636 */
2637 mdi->mdio_meta->disk[pd->pd_disk_pos].flags = INTEL_F_FAILED;
2638 pd->pd_disk_meta.flags = INTEL_F_FAILED;
2639 g_raid_md_intel_print(mdi->mdio_meta);
2640 if (tdisk->d_consumer)
2641 intel_meta_write(tdisk->d_consumer, mdi->mdio_meta);
2642
2643 /* Change states. */
2644 g_raid_change_disk_state(tdisk, G_RAID_DISK_S_FAILED);
2645 TAILQ_FOREACH(sd, &tdisk->d_subdisks, sd_next) {
2646 g_raid_change_subdisk_state(sd,
2647 G_RAID_SUBDISK_S_FAILED);
2648 g_raid_event_send(sd, G_RAID_SUBDISK_E_FAILED,
2649 G_RAID_EVENT_SUBDISK);
2650 }
2651
2652 /* Write updated metadata to remaining disks. */
2653 g_raid_md_write_intel(md, NULL, NULL, tdisk);
2654
2655 /* Check if anything left except placeholders. */
2656 if (g_raid_ndisks(sc, -1) ==
2657 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
2658 g_raid_destroy_node(sc, 0);
2659 else
2660 g_raid_md_intel_refill(sc);
2661 return (0);
2662 }
2663
2664 static int
2665 g_raid_md_free_disk_intel(struct g_raid_md_object *md,
2666 struct g_raid_disk *disk)
2667 {
2668 struct g_raid_md_intel_perdisk *pd;
2669
2670 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
2671 if (pd->pd_meta != NULL) {
2672 free(pd->pd_meta, M_MD_INTEL);
2673 pd->pd_meta = NULL;
2674 }
2675 free(pd, M_MD_INTEL);
2676 disk->d_md_data = NULL;
2677 return (0);
2678 }
2679
2680 static int
2681 g_raid_md_free_volume_intel(struct g_raid_md_object *md,
2682 struct g_raid_volume *vol)
2683 {
2684 struct g_raid_md_intel_pervolume *pv;
2685
2686 pv = (struct g_raid_md_intel_pervolume *)vol->v_md_data;
2687 free(pv, M_MD_INTEL);
2688 vol->v_md_data = NULL;
2689 return (0);
2690 }
2691
2692 static int
2693 g_raid_md_free_intel(struct g_raid_md_object *md)
2694 {
2695 struct g_raid_md_intel_object *mdi;
2696
2697 mdi = (struct g_raid_md_intel_object *)md;
2698 if (!mdi->mdio_started) {
2699 mdi->mdio_started = 0;
2700 callout_stop(&mdi->mdio_start_co);
2701 G_RAID_DEBUG1(1, md->mdo_softc,
2702 "root_mount_rel %p", mdi->mdio_rootmount);
2703 root_mount_rel(mdi->mdio_rootmount);
2704 mdi->mdio_rootmount = NULL;
2705 }
2706 if (mdi->mdio_meta != NULL) {
2707 free(mdi->mdio_meta, M_MD_INTEL);
2708 mdi->mdio_meta = NULL;
2709 }
2710 return (0);
2711 }
2712
2713 G_RAID_MD_DECLARE(intel, "Intel");
Cache object: 7f29622c779d78007c1afffac6f612ca
|