]>
git.ipfire.org Git - thirdparty/mdadm.git/blob - Assemble.c
2 * mdadm - manage Linux "md" devices aka RAID arrays.
4 * Copyright (C) 2001-2002 Neil Brown <neilb@cse.unsw.edu.au>
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; either version 2 of the License, or
10 * (at your option) any later version.
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
17 * You should have received a copy of the GNU General Public License
18 * along with this program; if not, write to the Free Software
19 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
22 * Email: <neilb@cse.unsw.edu.au>
24 * School of Computer Science and Engineering
25 * The University of New South Wales
32 int Assemble(struct supertype
*st
, char *mddev
, int mdfd
,
33 mddev_ident_t ident
, char *conffile
,
35 int readonly
, int runstop
,
37 int verbose
, int force
)
40 * The task of Assemble is to find a collection of
41 * devices that should (according to their superblocks)
42 * form an array, and to give this collection to the MD driver.
43 * In Linux-2.4 and later, this involves submitting a
44 * SET_ARRAY_INFO ioctl with no arg - to prepare
45 * the array - and then submit a number of
46 * ADD_NEW_DISK ioctls to add disks into
47 * the array. Finally RUN_ARRAY might
48 * be submitted to start the array.
50 * Much of the work of Assemble is in finding and/or
51 * checking the disks to make sure they look right.
53 * If mddev is not set, then scan must be set and we
54 * read through the config file for dev+uuid mapping
55 * We recurse, setting mddev, for each device that
57 * - has a valid uuid (or any uuid if !uuidset)
59 * If mddev is set, we try to determine state of md.
60 * check version - must be at least 0.90.0
61 * check kernel version. must be at least 2.4.
62 * If not, we can possibly fall back on START_ARRAY
63 * Try to GET_ARRAY_INFO.
64 * If possible, give up
65 * If not, try to STOP_ARRAY just to make sure
67 * If !uuidset and scan, look in conf-file for uuid
68 * If not found, give up
69 * If !devlist and scan and uuidset, get list of devs from conf-file
72 * Check superblock - discard if bad
73 * Check uuid (set if we don't have one) - discard if no match
74 * Check superblock similarity if we have a superblock - discard if different
75 * Record events, devicenum
76 * This should give us a list of devices for the array
77 * We should collect the most recent event number
79 * Count disks with recent enough event count
80 * While force && !enough disks
81 * Choose newest rejected disks, update event count
82 * mark clean and rewrite superblock
85 * foreach device with recent events : ADD_NEW_DISK
86 * if runstop == 1 || "enough" disks and runstop==0 -> RUN_ARRAY
88 * Check the device numbers in superblock are right
89 * update superblock if any changes
95 void *first_super
= NULL
, *super
= NULL
;
98 unsigned int major
, minor
;
99 unsigned int oldmajor
, oldminor
;
105 int *best
= NULL
; /* indexed by raid_disk */
106 unsigned int bestcnt
= 0;
108 unsigned int okcnt
, sparecnt
;
109 unsigned int req_cnt
;
115 int start_partial_ok
= force
|| devlist
==NULL
;
116 unsigned int num_devs
;
120 vers
= md_get_version(mdfd
);
122 fprintf(stderr
, Name
": %s appears not to be an md device.\n", mddev
);
126 fprintf(stderr
, Name
": Assemble requires driver version 0.90.0 or later.\n"
127 " Upgrade your kernel or try --build\n");
130 if (get_linux_version() < 2004000)
133 if (ioctl(mdfd
, GET_ARRAY_INFO
, &info
.array
)>=0) {
134 fprintf(stderr
, Name
": device %s already active - cannot assemble it\n",
138 ioctl(mdfd
, STOP_ARRAY
, NULL
); /* just incase it was started but has no content */
141 * If any subdevs are listed, then any that don't
142 * match ident are discarded. Remainder must all match and
144 * If no subdevs, then we scan all devices in the config file, but
145 * there must be something in the identity
149 ident
->uuid_set
== 0 &&
150 ident
->super_minor
< 0 &&
151 ident
->devices
== NULL
) {
152 fprintf(stderr
, Name
": No identity information available for %s - cannot assemble.\n",
157 devlist
= conf_get_devs(conffile
);
160 tmpdev
= devlist
; num_devs
= 0;
163 tmpdev
= tmpdev
->next
;
165 devices
= malloc(num_devs
* sizeof(*devices
));
167 if (!st
&& ident
->st
) st
= ident
->st
;
170 fprintf(stderr
, Name
": looking for devices for %s\n",
177 struct supertype
*tst
= st
;
179 devname
= devlist
->devname
;
180 devlist
= devlist
->next
;
182 if (ident
->devices
&&
183 !match_oneof(ident
->devices
, devname
)) {
184 if (inargv
|| verbose
)
185 fprintf(stderr
, Name
": %s is not one of %s\n", devname
, ident
->devices
);
194 dfd
= open(devname
, O_RDONLY
|O_EXCL
, 0);
196 if (inargv
|| verbose
)
197 fprintf(stderr
, Name
": cannot open device %s: %s\n",
198 devname
, strerror(errno
));
199 } else if (fstat(dfd
, &stb
)< 0) {
201 fprintf(stderr
, Name
": fstat failed for %s: %s\n",
202 devname
, strerror(errno
));
204 } else if ((stb
.st_mode
& S_IFMT
) != S_IFBLK
) {
205 fprintf(stderr
, Name
": %s is not a block device.\n",
208 } else if (!tst
&& (tst
= guess_super(dfd
)) == NULL
) {
209 if (inargv
|| verbose
)
210 fprintf(stderr
, Name
": no recogniseable superblock\n");
211 } else if (tst
->ss
->load_super(tst
,dfd
, &super
, NULL
)) {
212 if (inargv
|| verbose
)
213 fprintf( stderr
, Name
": no RAID superblock on %s\n",
217 tst
->ss
->getinfo_super(&info
, super
);
221 if (ident
->uuid_set
&&
222 (!super
|| same_uuid(info
.uuid
, ident
->uuid
)==0)) {
223 if (inargv
|| verbose
)
224 fprintf(stderr
, Name
": %s has wrong uuid.\n",
228 if (ident
->super_minor
!= UnSet
&&
229 (!super
|| ident
->super_minor
!= info
.array
.md_minor
)) {
230 if (inargv
|| verbose
)
231 fprintf(stderr
, Name
": %s has wrong super-minor.\n",
235 if (ident
->level
!= UnSet
&&
236 (!super
|| ident
->level
!= info
.array
.level
)) {
237 if (inargv
|| verbose
)
238 fprintf(stderr
, Name
": %s has wrong raid level.\n",
242 if (ident
->raid_disks
!= UnSet
&&
243 (!super
|| ident
->raid_disks
!= info
.array
.raid_disks
)) {
244 if (inargv
|| verbose
)
245 fprintf(stderr
, Name
": %s requires wrong number of drives.\n",
250 /* If we are this far, then we are commited to this device.
251 * If the super_block doesn't exist, or doesn't match others,
252 * then we cannot continue
256 fprintf(stderr
, Name
": %s has no superblock - assembly aborted\n",
261 st
= tst
; /* commit to this format, if haven't already */
262 if (st
->ss
->compare_super(&first_super
, super
)) {
263 fprintf(stderr
, Name
": superblock on %s doesn't match others - assembly aborted\n",
270 /* looks like a good enough match to update the super block if needed */
272 /* prepare useful information in info structures */
275 info
.array
.md_minor
= minor(stb2
.st_rdev
);
277 st
->ss
->update_super(&info
, super
, update
, devname
, verbose
);
279 dfd
= open(devname
, O_RDWR
|O_EXCL
, 0);
281 fprintf(stderr
, Name
": Cannot open %s for superblock update\n",
283 else if (st
->ss
->store_super(st
, dfd
, super
))
284 fprintf(stderr
, Name
": Could not re-write superblock on %s.\n",
291 fprintf(stderr
, Name
": %s is identified as a member of %s, slot %d.\n",
292 devname
, mddev
, info
.disk
.raid_disk
);
293 devices
[devcnt
].devname
= devname
;
294 devices
[devcnt
].major
= major(stb
.st_rdev
);
295 devices
[devcnt
].minor
= minor(stb
.st_rdev
);
296 devices
[devcnt
].oldmajor
= info
.disk
.major
;
297 devices
[devcnt
].oldminor
= info
.disk
.minor
;
298 devices
[devcnt
].events
= info
.events
;
299 devices
[devcnt
].raid_disk
= info
.disk
.raid_disk
;
300 devices
[devcnt
].uptodate
= 0;
301 devices
[devcnt
].state
= info
.disk
.state
;
302 if (most_recent
< devcnt
) {
303 if (devices
[devcnt
].events
304 > devices
[most_recent
].events
)
305 most_recent
= devcnt
;
307 if (info
.array
.level
== -4)
308 /* with multipath, the raid_disk from the superblock is meaningless */
311 i
= devices
[devcnt
].raid_disk
;
314 unsigned int newbestcnt
= i
+10;
315 int *newbest
= malloc(sizeof(int)*newbestcnt
);
317 for (c
=0; c
< newbestcnt
; c
++)
319 newbest
[c
] = best
[c
];
324 bestcnt
= newbestcnt
;
327 || devices
[best
[i
]].events
< devices
[devcnt
].events
)
338 fprintf(stderr
, Name
": no devices found for %s\n",
344 st
->ss
->getinfo_super(&info
, first_super
);
346 /* now we have some devices that might be suitable.
351 for (i
=0; i
< bestcnt
;i
++) {
353 int event_margin
= !force
;
355 /* note: we ignore error flags in multipath arrays
356 * as they don't make sense
358 if (info
.array
.level
!= -4)
359 if (!(devices
[j
].state
& (1<<MD_DISK_SYNC
))) {
360 if (!(devices
[j
].state
& (1<<MD_DISK_FAULTY
)))
364 if (devices
[j
].events
+event_margin
>=
365 devices
[most_recent
].events
) {
366 devices
[j
].uptodate
= 1;
367 if (i
< info
.array
.raid_disks
)
373 while (force
&& !enough(info
.array
.level
, info
.array
.raid_disks
, okcnt
)) {
374 /* Choose the newest best drive which is
375 * not up-to-date, update the superblock
380 for (i
=0; i
<info
.array
.raid_disks
&& i
< bestcnt
; i
++) {
383 !devices
[j
].uptodate
&&
384 devices
[j
].events
> 0 &&
386 devices
[j
].events
> devices
[chosen_drive
].events
))
389 if (chosen_drive
< 0)
391 fprintf(stderr
, Name
": forcing event count in %s(%d) from %d upto %d\n",
392 devices
[chosen_drive
].devname
, devices
[chosen_drive
].raid_disk
,
393 (int)(devices
[chosen_drive
].events
),
394 (int)(devices
[most_recent
].events
));
395 fd
= open(devices
[chosen_drive
].devname
, O_RDWR
|O_EXCL
);
397 fprintf(stderr
, Name
": Couldn't open %s for write - not updating\n",
398 devices
[chosen_drive
].devname
);
399 devices
[chosen_drive
].events
= 0;
402 if (st
->ss
->load_super(st
,fd
, &super
, NULL
)) {
404 fprintf(stderr
, Name
": RAID superblock disappeared from %s - not updating.\n",
405 devices
[chosen_drive
].devname
);
406 devices
[chosen_drive
].events
= 0;
409 info
.events
= devices
[most_recent
].events
;
410 st
->ss
->update_super(&info
, super
, "force", devices
[chosen_drive
].devname
, verbose
);
412 if (st
->ss
->store_super(st
, fd
, super
)) {
414 fprintf(stderr
, Name
": Could not re-write superblock on %s\n",
415 devices
[chosen_drive
].devname
);
416 devices
[chosen_drive
].events
= 0;
421 devices
[chosen_drive
].events
= devices
[most_recent
].events
;
422 devices
[chosen_drive
].uptodate
= 1;
427 /* Now we want to look at the superblock which the kernel will base things on
428 * and compare the devices that we think are working with the devices that the
429 * superblock thinks are working.
430 * If there are differences and --force is given, then update this chosen
435 for (i
=0; chosen_drive
< 0 && i
<bestcnt
; i
++) {
441 if (!devices
[j
].uptodate
)
444 if ((fd
=open(devices
[j
].devname
, O_RDONLY
|O_EXCL
))< 0) {
445 fprintf(stderr
, Name
": Cannot open %s: %s\n",
446 devices
[j
].devname
, strerror(errno
));
449 if (st
->ss
->load_super(st
,fd
, &super
, NULL
)) {
451 fprintf(stderr
, Name
": RAID superblock has disappeared from %s\n",
458 fprintf(stderr
, Name
": No suitable drives found for %s\n", mddev
);
461 st
->ss
->getinfo_super(&info
, super
);
462 for (i
=0; i
<bestcnt
; i
++) {
464 unsigned int desired_state
;
466 if (i
< info
.array
.raid_disks
)
467 desired_state
= (1<<MD_DISK_ACTIVE
) | (1<<MD_DISK_SYNC
);
473 if (!devices
[j
].uptodate
)
475 info
.disk
.number
= i
;
476 info
.disk
.state
= desired_state
;
478 if (devices
[j
].uptodate
&&
479 st
->ss
->update_super(&info
, super
, "assemble", NULL
, 0)) {
481 fprintf(stderr
, Name
": "
482 "clearing FAULTY flag for device %d in %s for %s\n",
483 j
, mddev
, devices
[j
].devname
);
486 fprintf(stderr
, Name
": "
487 "device %d in %s has wrong state in superblock, but %s seems ok\n",
488 i
, mddev
, devices
[j
].devname
);
492 if (!devices
[j
].uptodate
&&
493 !(super
.disks
[i
].state
& (1 << MD_DISK_FAULTY
))) {
494 fprintf(stderr
, Name
": devices %d of %s is not marked FAULTY in superblock, but cannot be found\n",
499 if (force
&& okcnt
== info
.array
.raid_disks
-1) {
500 /* FIXME check event count */
501 change
+= st
->ss
->update_super(&info
, super
, "force",
502 devices
[chosen_drive
].devname
, 0);
507 fd
= open(devices
[chosen_drive
].devname
, O_RDWR
|O_EXCL
);
509 fprintf(stderr
, Name
": Could open %s for write - cannot Assemble array.\n",
510 devices
[chosen_drive
].devname
);
513 if (st
->ss
->store_super(st
, fd
, super
)) {
515 fprintf(stderr
, Name
": Could not re-write superblock on %s\n",
516 devices
[chosen_drive
].devname
);
522 /* count number of in-sync devices according to the superblock.
523 * We must have this number to start the array without -s or -R
525 req_cnt
= info
.array
.working_disks
;
527 /* Almost ready to actually *do* something */
529 if (ioctl(mdfd
, SET_ARRAY_INFO
, NULL
) != 0) {
530 fprintf(stderr
, Name
": SET_ARRAY_INFO failed for %s: %s\n",
531 mddev
, strerror(errno
));
534 if (ident
->bitmap_fd
>= 0) {
535 if (ioctl(mdfd
, SET_BITMAP_FILE
, ident
->bitmap_fd
) != 0) {
536 fprintf(stderr
, Name
": SET_BITMAP_FILE failed.\n");
541 /* First, add the raid disks, but add the chosen one last */
542 for (i
=0; i
<= bestcnt
; i
++) {
546 if (j
== chosen_drive
)
551 if (j
>= 0 /* && devices[j].uptodate */) {
552 mdu_disk_info_t disk
;
553 memset(&disk
, 0, sizeof(disk
));
554 disk
.major
= devices
[j
].major
;
555 disk
.minor
= devices
[j
].minor
;
556 if (ioctl(mdfd
, ADD_NEW_DISK
, &disk
)!=0) {
557 fprintf(stderr
, Name
": failed to add %s to %s: %s\n",
561 if (i
< info
.array
.raid_disks
|| i
== bestcnt
)
566 fprintf(stderr
, Name
": added %s to %s as %d\n",
567 devices
[j
].devname
, mddev
, devices
[j
].raid_disk
);
568 } else if (verbose
&& i
< info
.array
.raid_disks
)
569 fprintf(stderr
, Name
": no uptodate device for slot %d of %s\n",
575 ( enough(info
.array
.level
, info
.array
.raid_disks
, okcnt
) &&
576 (okcnt
>= req_cnt
|| start_partial_ok
)
578 if (ioctl(mdfd
, RUN_ARRAY
, NULL
)==0) {
579 fprintf(stderr
, Name
": %s has been started with %d drive%s",
580 mddev
, okcnt
, okcnt
==1?"":"s");
581 if (okcnt
< info
.array
.raid_disks
)
582 fprintf(stderr
, " (out of %d)", info
.array
.raid_disks
);
584 fprintf(stderr
, " and %d spare%s", sparecnt
, sparecnt
==1?"":"s");
585 fprintf(stderr
, ".\n");
588 fprintf(stderr
, Name
": failed to RUN_ARRAY %s: %s\n",
589 mddev
, strerror(errno
));
593 fprintf(stderr
, Name
": %s assembled from %d drive%s, but not started.\n",
594 mddev
, okcnt
, okcnt
==1?"":"s");
597 fprintf(stderr
, Name
": %s assembled from %d drive%s", mddev
, okcnt
, okcnt
==1?"":"s");
599 fprintf(stderr
, " and %d spare%s", sparecnt
, sparecnt
==1?"":"s");
600 if (!enough(info
.array
.level
, info
.array
.raid_disks
, okcnt
))
601 fprintf(stderr
, " - not enough to start the array.\n");
603 if (req_cnt
== info
.array
.raid_disks
)
604 fprintf(stderr
, " - need all %d to start it", req_cnt
);
606 fprintf(stderr
, " - need %d of %d to start", req_cnt
, info
.array
.raid_disks
);
607 fprintf(stderr
, " (use --run to insist).\n");
611 /* The "chosen_drive" is a good choice, and if necessary, the superblock has
612 * been updated to point to the current locations of devices.
613 * so we can just start the array
616 dev
= makedev(devices
[chosen_drive
].major
,
617 devices
[chosen_drive
].minor
);
618 if (ioctl(mdfd
, START_ARRAY
, dev
)) {
619 fprintf(stderr
, Name
": Cannot start array: %s\n",