]>
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
;
106 int *best
= NULL
; /* indexed by raid_disk */
107 unsigned int bestcnt
= 0;
109 unsigned int okcnt
, sparecnt
;
110 unsigned int req_cnt
;
116 int start_partial_ok
= force
|| devlist
==NULL
;
117 unsigned int num_devs
;
120 struct mddev_ident_s ident2
;
122 vers
= md_get_version(mdfd
);
124 fprintf(stderr
, Name
": %s appears not to be an md device.\n", mddev
);
128 fprintf(stderr
, Name
": Assemble requires driver version 0.90.0 or later.\n"
129 " Upgrade your kernel or try --build\n");
132 if (get_linux_version() < 2004000)
135 if (ioctl(mdfd
, GET_ARRAY_INFO
, &info
.array
)>=0) {
136 fprintf(stderr
, Name
": device %s already active - cannot assemble it\n",
140 ioctl(mdfd
, STOP_ARRAY
, NULL
); /* just incase it was started but has no content */
143 * If any subdevs are listed, then any that don't
144 * match ident are discarded. Remainder must all match and
146 * If no subdevs, then we scan all devices in the config file, but
147 * there must be something in the identity
151 ident
->uuid_set
== 0 &&
152 ident
->super_minor
< 0 &&
153 ident
->devices
== NULL
) {
154 fprintf(stderr
, Name
": No identity information available for %s - cannot assemble.\n",
159 devlist
= conf_get_devs(conffile
);
162 tmpdev
= devlist
; num_devs
= 0;
165 tmpdev
= tmpdev
->next
;
167 devices
= malloc(num_devs
* sizeof(*devices
));
169 if (!st
&& ident
->st
) st
= ident
->st
;
172 fprintf(stderr
, Name
": looking for devices for %s\n",
179 struct supertype
*tst
= st
;
181 devname
= devlist
->devname
;
182 devlist
= devlist
->next
;
184 if (ident
->devices
&&
185 !match_oneof(ident
->devices
, devname
)) {
186 if ((inargv
&& verbose
>=0) || verbose
> 0)
187 fprintf(stderr
, Name
": %s is not one of %s\n", devname
, ident
->devices
);
196 dfd
= open(devname
, O_RDONLY
|O_EXCL
, 0);
198 if ((inargv
&& verbose
>= 0) || verbose
> 0)
199 fprintf(stderr
, Name
": cannot open device %s: %s\n",
200 devname
, strerror(errno
));
201 } else if (fstat(dfd
, &stb
)< 0) {
203 fprintf(stderr
, Name
": fstat failed for %s: %s\n",
204 devname
, strerror(errno
));
206 } else if ((stb
.st_mode
& S_IFMT
) != S_IFBLK
) {
207 fprintf(stderr
, Name
": %s is not a block device.\n",
210 } else if (!tst
&& (tst
= guess_super(dfd
)) == NULL
) {
211 if ((inargv
&& verbose
>= 0) || verbose
> 0)
212 fprintf(stderr
, Name
": no recogniseable superblock\n");
213 } else if (tst
->ss
->load_super(tst
,dfd
, &super
, NULL
)) {
214 if ((inargv
&& verbose
>= 0) || verbose
> 0)
215 fprintf( stderr
, Name
": no RAID superblock on %s\n",
219 tst
->ss
->getinfo_super(&info
, &ident2
, super
);
223 if (ident
->uuid_set
&&
224 (!super
|| same_uuid(info
.uuid
, ident
->uuid
, tst
->ss
->swapuuid
)==0)) {
225 if ((inargv
&& verbose
>= 0) || verbose
> 0)
226 fprintf(stderr
, Name
": %s has wrong uuid.\n",
230 if (ident
->name
[0] &&
231 (!super
|| strncmp(ident2
.name
, ident
->name
, 32)!=0)) {
232 if ((inargv
&& verbose
>= 0) || verbose
> 0)
233 fprintf(stderr
, Name
": %s has wrong name.\n",
237 if (ident
->super_minor
!= UnSet
&&
238 (!super
|| ident
->super_minor
!= info
.array
.md_minor
)) {
239 if ((inargv
&& verbose
>= 0) || verbose
> 0)
240 fprintf(stderr
, Name
": %s has wrong super-minor.\n",
244 if (ident
->level
!= UnSet
&&
245 (!super
|| ident
->level
!= info
.array
.level
)) {
246 if ((inargv
&& verbose
>= 0) || verbose
> 0)
247 fprintf(stderr
, Name
": %s has wrong raid level.\n",
251 if (ident
->raid_disks
!= UnSet
&&
252 (!super
|| ident
->raid_disks
!= info
.array
.raid_disks
)) {
253 if ((inargv
&& verbose
>= 0) || verbose
> 0)
254 fprintf(stderr
, Name
": %s requires wrong number of drives.\n",
259 /* If we are this far, then we are commited to this device.
260 * If the super_block doesn't exist, or doesn't match others,
261 * then we cannot continue
265 fprintf(stderr
, Name
": %s has no superblock - assembly aborted\n",
270 st
= tst
; /* commit to this format, if haven't already */
271 if (st
->ss
->compare_super(&first_super
, super
)) {
272 fprintf(stderr
, Name
": superblock on %s doesn't match others - assembly aborted\n",
279 /* looks like a good enough match to update the super block if needed */
281 /* prepare useful information in info structures */
284 info
.array
.md_minor
= minor(stb2
.st_rdev
);
286 st
->ss
->update_super(&info
, super
, update
, devname
, verbose
);
288 dfd
= open(devname
, O_RDWR
|O_EXCL
, 0);
290 fprintf(stderr
, Name
": Cannot open %s for superblock update\n",
292 else if (st
->ss
->store_super(st
, dfd
, super
))
293 fprintf(stderr
, Name
": Could not re-write superblock on %s.\n",
300 fprintf(stderr
, Name
": %s is identified as a member of %s, slot %d.\n",
301 devname
, mddev
, info
.disk
.raid_disk
);
302 devices
[devcnt
].devname
= devname
;
303 devices
[devcnt
].major
= major(stb
.st_rdev
);
304 devices
[devcnt
].minor
= minor(stb
.st_rdev
);
305 devices
[devcnt
].oldmajor
= info
.disk
.major
;
306 devices
[devcnt
].oldminor
= info
.disk
.minor
;
307 devices
[devcnt
].events
= info
.events
;
308 devices
[devcnt
].raid_disk
= info
.disk
.raid_disk
;
309 devices
[devcnt
].disk_nr
= info
.disk
.number
;
310 devices
[devcnt
].uptodate
= 0;
311 devices
[devcnt
].state
= info
.disk
.state
;
312 if (most_recent
< devcnt
) {
313 if (devices
[devcnt
].events
314 > devices
[most_recent
].events
)
315 most_recent
= devcnt
;
317 if (info
.array
.level
== -4)
318 /* with multipath, the raid_disk from the superblock is meaningless */
321 i
= devices
[devcnt
].raid_disk
;
324 unsigned int newbestcnt
= i
+10;
325 int *newbest
= malloc(sizeof(int)*newbestcnt
);
327 for (c
=0; c
< newbestcnt
; c
++)
329 newbest
[c
] = best
[c
];
334 bestcnt
= newbestcnt
;
337 || devices
[best
[i
]].events
< devices
[devcnt
].events
)
347 if (update
&& strcmp(update
, "byteorder")==0)
348 st
->minor_version
= 90;
351 fprintf(stderr
, Name
": no devices found for %s\n",
357 st
->ss
->getinfo_super(&info
, &ident2
, first_super
);
359 /* now we have some devices that might be suitable.
364 for (i
=0; i
< bestcnt
;i
++) {
366 int event_margin
= !force
;
368 /* note: we ignore error flags in multipath arrays
369 * as they don't make sense
371 if (info
.array
.level
!= -4)
372 if (!(devices
[j
].state
& (1<<MD_DISK_SYNC
))) {
373 if (!(devices
[j
].state
& (1<<MD_DISK_FAULTY
)))
377 if (devices
[j
].events
+event_margin
>=
378 devices
[most_recent
].events
) {
379 devices
[j
].uptodate
= 1;
380 if (i
< info
.array
.raid_disks
)
386 while (force
&& !enough(info
.array
.level
, info
.array
.raid_disks
, okcnt
)) {
387 /* Choose the newest best drive which is
388 * not up-to-date, update the superblock
393 for (i
=0; i
<info
.array
.raid_disks
&& i
< bestcnt
; i
++) {
396 !devices
[j
].uptodate
&&
397 devices
[j
].events
> 0 &&
399 devices
[j
].events
> devices
[chosen_drive
].events
))
402 if (chosen_drive
< 0)
405 fprintf(stderr
, Name
": forcing event count in %s(%d) from %d upto %d\n",
406 devices
[chosen_drive
].devname
, devices
[chosen_drive
].raid_disk
,
407 (int)(devices
[chosen_drive
].events
),
408 (int)(devices
[most_recent
].events
));
409 fd
= open(devices
[chosen_drive
].devname
, O_RDWR
|O_EXCL
);
411 fprintf(stderr
, Name
": Couldn't open %s for write - not updating\n",
412 devices
[chosen_drive
].devname
);
413 devices
[chosen_drive
].events
= 0;
416 if (st
->ss
->load_super(st
,fd
, &super
, NULL
)) {
418 fprintf(stderr
, Name
": RAID superblock disappeared from %s - not updating.\n",
419 devices
[chosen_drive
].devname
);
420 devices
[chosen_drive
].events
= 0;
423 info
.events
= devices
[most_recent
].events
;
424 st
->ss
->update_super(&info
, super
, "force", devices
[chosen_drive
].devname
, verbose
);
426 if (st
->ss
->store_super(st
, fd
, super
)) {
428 fprintf(stderr
, Name
": Could not re-write superblock on %s\n",
429 devices
[chosen_drive
].devname
);
430 devices
[chosen_drive
].events
= 0;
435 devices
[chosen_drive
].events
= devices
[most_recent
].events
;
436 devices
[chosen_drive
].uptodate
= 1;
441 /* Now we want to look at the superblock which the kernel will base things on
442 * and compare the devices that we think are working with the devices that the
443 * superblock thinks are working.
444 * If there are differences and --force is given, then update this chosen
449 for (i
=0; chosen_drive
< 0 && i
<bestcnt
; i
++) {
455 if (!devices
[j
].uptodate
)
458 if ((fd
=open(devices
[j
].devname
, O_RDONLY
|O_EXCL
))< 0) {
459 fprintf(stderr
, Name
": Cannot open %s: %s\n",
460 devices
[j
].devname
, strerror(errno
));
463 if (st
->ss
->load_super(st
,fd
, &super
, NULL
)) {
465 fprintf(stderr
, Name
": RAID superblock has disappeared from %s\n",
472 fprintf(stderr
, Name
": No suitable drives found for %s\n", mddev
);
475 st
->ss
->getinfo_super(&info
, &ident2
, super
);
476 for (i
=0; i
<bestcnt
; i
++) {
478 unsigned int desired_state
;
480 if (i
< info
.array
.raid_disks
)
481 desired_state
= (1<<MD_DISK_ACTIVE
) | (1<<MD_DISK_SYNC
);
487 if (!devices
[j
].uptodate
)
489 info
.disk
.number
= devices
[j
].disk_nr
;
490 info
.disk
.raid_disk
= i
;
491 info
.disk
.state
= desired_state
;
493 if (devices
[j
].uptodate
&&
494 st
->ss
->update_super(&info
, super
, "assemble", NULL
, verbose
)) {
497 fprintf(stderr
, Name
": "
498 "clearing FAULTY flag for device %d in %s for %s\n",
499 j
, mddev
, devices
[j
].devname
);
503 fprintf(stderr
, Name
": "
504 "device %d in %s has wrong state in superblock, but %s seems ok\n",
505 i
, mddev
, devices
[j
].devname
);
509 if (!devices
[j
].uptodate
&&
510 !(super
.disks
[i
].state
& (1 << MD_DISK_FAULTY
))) {
511 fprintf(stderr
, Name
": devices %d of %s is not marked FAULTY in superblock, but cannot be found\n",
516 if (force
&& okcnt
== info
.array
.raid_disks
-1) {
517 /* FIXME check event count */
518 change
+= st
->ss
->update_super(&info
, super
, "force",
519 devices
[chosen_drive
].devname
, verbose
);
524 fd
= open(devices
[chosen_drive
].devname
, O_RDWR
|O_EXCL
);
526 fprintf(stderr
, Name
": Could open %s for write - cannot Assemble array.\n",
527 devices
[chosen_drive
].devname
);
530 if (st
->ss
->store_super(st
, fd
, super
)) {
532 fprintf(stderr
, Name
": Could not re-write superblock on %s\n",
533 devices
[chosen_drive
].devname
);
539 /* count number of in-sync devices according to the superblock.
540 * We must have this number to start the array without -s or -R
542 req_cnt
= info
.array
.working_disks
;
544 /* Almost ready to actually *do* something */
547 if ((vers
% 100) >= 1) { /* can use different versions */
548 mdu_array_info_t inf
;
549 memset(&inf
, 0, sizeof(inf
));
550 inf
.major_version
= st
->ss
->major
;
551 inf
.minor_version
= st
->minor_version
;
552 rv
= ioctl(mdfd
, SET_ARRAY_INFO
, &inf
);
554 rv
= ioctl(mdfd
, SET_ARRAY_INFO
, NULL
);
557 fprintf(stderr
, Name
": SET_ARRAY_INFO failed for %s: %s\n",
558 mddev
, strerror(errno
));
561 if (ident
->bitmap_fd
>= 0) {
562 if (ioctl(mdfd
, SET_BITMAP_FILE
, ident
->bitmap_fd
) != 0) {
563 fprintf(stderr
, Name
": SET_BITMAP_FILE failed.\n");
568 /* First, add the raid disks, but add the chosen one last */
569 for (i
=0; i
<= bestcnt
; i
++) {
573 if (j
== chosen_drive
)
578 if (j
>= 0 /* && devices[j].uptodate */) {
579 mdu_disk_info_t disk
;
580 memset(&disk
, 0, sizeof(disk
));
581 disk
.major
= devices
[j
].major
;
582 disk
.minor
= devices
[j
].minor
;
583 if (ioctl(mdfd
, ADD_NEW_DISK
, &disk
)!=0) {
584 fprintf(stderr
, Name
": failed to add %s to %s: %s\n",
588 if (i
< info
.array
.raid_disks
|| i
== bestcnt
)
592 } else if (verbose
> 0)
593 fprintf(stderr
, Name
": added %s to %s as %d\n",
594 devices
[j
].devname
, mddev
, devices
[j
].raid_disk
);
595 } else if (verbose
> 0 && i
< info
.array
.raid_disks
)
596 fprintf(stderr
, Name
": no uptodate device for slot %d of %s\n",
602 ( enough(info
.array
.level
, info
.array
.raid_disks
, okcnt
) &&
603 (okcnt
>= req_cnt
|| start_partial_ok
)
605 if (ioctl(mdfd
, RUN_ARRAY
, NULL
)==0) {
607 fprintf(stderr
, Name
": %s has been started with %d drive%s",
608 mddev
, okcnt
, okcnt
==1?"":"s");
609 if (okcnt
< info
.array
.raid_disks
)
610 fprintf(stderr
, " (out of %d)", info
.array
.raid_disks
);
612 fprintf(stderr
, " and %d spare%s", sparecnt
, sparecnt
==1?"":"s");
613 fprintf(stderr
, ".\n");
617 fprintf(stderr
, Name
": failed to RUN_ARRAY %s: %s\n",
618 mddev
, strerror(errno
));
622 fprintf(stderr
, Name
": %s assembled from %d drive%s, but not started.\n",
623 mddev
, okcnt
, okcnt
==1?"":"s");
627 fprintf(stderr
, Name
": %s assembled from %d drive%s", mddev
, okcnt
, okcnt
==1?"":"s");
629 fprintf(stderr
, " and %d spare%s", sparecnt
, sparecnt
==1?"":"s");
630 if (!enough(info
.array
.level
, info
.array
.raid_disks
, okcnt
))
631 fprintf(stderr
, " - not enough to start the array.\n");
633 if (req_cnt
== info
.array
.raid_disks
)
634 fprintf(stderr
, " - need all %d to start it", req_cnt
);
636 fprintf(stderr
, " - need %d of %d to start", req_cnt
, info
.array
.raid_disks
);
637 fprintf(stderr
, " (use --run to insist).\n");
642 /* The "chosen_drive" is a good choice, and if necessary, the superblock has
643 * been updated to point to the current locations of devices.
644 * so we can just start the array
647 dev
= makedev(devices
[chosen_drive
].major
,
648 devices
[chosen_drive
].minor
);
649 if (ioctl(mdfd
, START_ARRAY
, dev
)) {
650 fprintf(stderr
, Name
": Cannot start array: %s\n",