]>
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
;
124 vers
= md_get_version(mdfd
);
126 fprintf(stderr
, Name
": %s appears not to be an md device.\n", mddev
);
130 fprintf(stderr
, Name
": Assemble requires driver version 0.90.0 or later.\n"
131 " Upgrade your kernel or try --build\n");
134 if (get_linux_version() < 2004000)
137 if (ioctl(mdfd
, GET_ARRAY_INFO
, &info
.array
)>=0) {
138 fprintf(stderr
, Name
": device %s already active - cannot assemble it\n",
142 ioctl(mdfd
, STOP_ARRAY
, NULL
); /* just incase it was started but has no content */
145 * If any subdevs are listed, then any that don't
146 * match ident are discarded. Remainder must all match and
148 * If no subdevs, then we scan all devices in the config file, but
149 * there must be something in the identity
153 ident
->uuid_set
== 0 &&
154 ident
->super_minor
< 0 &&
155 ident
->devices
== NULL
) {
156 fprintf(stderr
, Name
": No identity information available for %s - cannot assemble.\n",
161 devlist
= conf_get_devs(conffile
);
164 tmpdev
= devlist
; num_devs
= 0;
167 tmpdev
= tmpdev
->next
;
169 devices
= malloc(num_devs
* sizeof(*devices
));
171 if (!st
&& ident
->st
) st
= ident
->st
;
174 fprintf(stderr
, Name
": looking for devices for %s\n",
181 struct supertype
*tst
= st
;
183 devname
= devlist
->devname
;
184 devlist
= devlist
->next
;
186 if (ident
->devices
&&
187 !match_oneof(ident
->devices
, devname
)) {
188 if ((inargv
&& verbose
>=0) || verbose
> 0)
189 fprintf(stderr
, Name
": %s is not one of %s\n", devname
, ident
->devices
);
198 dfd
= dev_open(devname
, O_RDONLY
|O_EXCL
);
200 if ((inargv
&& verbose
>= 0) || verbose
> 0)
201 fprintf(stderr
, Name
": cannot open device %s: %s\n",
202 devname
, strerror(errno
));
203 } else if (fstat(dfd
, &stb
)< 0) {
205 fprintf(stderr
, Name
": fstat failed for %s: %s\n",
206 devname
, strerror(errno
));
208 } else if ((stb
.st_mode
& S_IFMT
) != S_IFBLK
) {
209 fprintf(stderr
, Name
": %s is not a block device.\n",
212 } else if (!tst
&& (tst
= guess_super(dfd
)) == NULL
) {
213 if ((inargv
&& verbose
>= 0) || verbose
> 0)
214 fprintf(stderr
, Name
": no recogniseable superblock\n");
215 } else if (tst
->ss
->load_super(tst
,dfd
, &super
, NULL
)) {
216 if ((inargv
&& verbose
>= 0) || verbose
> 0)
217 fprintf( stderr
, Name
": no RAID superblock on %s\n",
221 tst
->ss
->getinfo_super(&info
, &ident2
, super
);
225 if (ident
->uuid_set
&&
226 (!super
|| same_uuid(info
.uuid
, ident
->uuid
, tst
->ss
->swapuuid
)==0)) {
227 if ((inargv
&& verbose
>= 0) || verbose
> 0)
228 fprintf(stderr
, Name
": %s has wrong uuid.\n",
232 if (ident
->name
[0] &&
233 (!super
|| strncmp(ident2
.name
, ident
->name
, 32)!=0)) {
234 if ((inargv
&& verbose
>= 0) || verbose
> 0)
235 fprintf(stderr
, Name
": %s has wrong name.\n",
239 if (ident
->super_minor
!= UnSet
&&
240 (!super
|| ident
->super_minor
!= info
.array
.md_minor
)) {
241 if ((inargv
&& verbose
>= 0) || verbose
> 0)
242 fprintf(stderr
, Name
": %s has wrong super-minor.\n",
246 if (ident
->level
!= UnSet
&&
247 (!super
|| ident
->level
!= info
.array
.level
)) {
248 if ((inargv
&& verbose
>= 0) || verbose
> 0)
249 fprintf(stderr
, Name
": %s has wrong raid level.\n",
253 if (ident
->raid_disks
!= UnSet
&&
254 (!super
|| ident
->raid_disks
!= info
.array
.raid_disks
)) {
255 if ((inargv
&& verbose
>= 0) || verbose
> 0)
256 fprintf(stderr
, Name
": %s requires wrong number of drives.\n",
261 /* If we are this far, then we are commited to this device.
262 * If the super_block doesn't exist, or doesn't match others,
263 * then we cannot continue
267 fprintf(stderr
, Name
": %s has no superblock - assembly aborted\n",
272 st
= tst
; /* commit to this format, if haven't already */
273 if (st
->ss
->compare_super(&first_super
, super
)) {
274 fprintf(stderr
, Name
": superblock on %s doesn't match others - assembly aborted\n",
281 /* looks like a good enough match to update the super block if needed */
283 /* prepare useful information in info structures */
286 info
.array
.md_minor
= minor(stb2
.st_rdev
);
288 st
->ss
->update_super(&info
, super
, update
, devname
, verbose
);
290 dfd
= dev_open(devname
, O_RDWR
|O_EXCL
);
292 fprintf(stderr
, Name
": Cannot open %s for superblock update\n",
294 else if (st
->ss
->store_super(st
, dfd
, super
))
295 fprintf(stderr
, Name
": Could not re-write superblock on %s.\n",
302 fprintf(stderr
, Name
": %s is identified as a member of %s, slot %d.\n",
303 devname
, mddev
, info
.disk
.raid_disk
);
304 devices
[devcnt
].devname
= devname
;
305 devices
[devcnt
].major
= major(stb
.st_rdev
);
306 devices
[devcnt
].minor
= minor(stb
.st_rdev
);
307 devices
[devcnt
].oldmajor
= info
.disk
.major
;
308 devices
[devcnt
].oldminor
= info
.disk
.minor
;
309 devices
[devcnt
].events
= info
.events
;
310 devices
[devcnt
].raid_disk
= info
.disk
.raid_disk
;
311 devices
[devcnt
].disk_nr
= info
.disk
.number
;
312 devices
[devcnt
].uptodate
= 0;
313 devices
[devcnt
].state
= info
.disk
.state
;
314 if (most_recent
< devcnt
) {
315 if (devices
[devcnt
].events
316 > devices
[most_recent
].events
)
317 most_recent
= devcnt
;
319 if (info
.array
.level
== -4)
320 /* with multipath, the raid_disk from the superblock is meaningless */
323 i
= devices
[devcnt
].raid_disk
;
325 if (nextspare
< info
.array
.raid_disks
)
326 nextspare
= info
.array
.raid_disks
;
331 unsigned int newbestcnt
= i
+10;
332 int *newbest
= malloc(sizeof(int)*newbestcnt
);
334 for (c
=0; c
< newbestcnt
; c
++)
336 newbest
[c
] = best
[c
];
341 bestcnt
= newbestcnt
;
344 || devices
[best
[i
]].events
< devices
[devcnt
].events
)
354 if (update
&& strcmp(update
, "byteorder")==0)
355 st
->minor_version
= 90;
358 fprintf(stderr
, Name
": no devices found for %s\n",
364 st
->ss
->getinfo_super(&info
, &ident2
, first_super
);
366 /* now we have some devices that might be suitable.
369 avail
= malloc(info
.array
.raid_disks
);
370 memset(avail
, 0, info
.array
.raid_disks
);
373 for (i
=0; i
< bestcnt
;i
++) {
375 int event_margin
= !force
;
377 /* note: we ignore error flags in multipath arrays
378 * as they don't make sense
380 if (info
.array
.level
!= -4)
381 if (!(devices
[j
].state
& (1<<MD_DISK_SYNC
))) {
382 if (!(devices
[j
].state
& (1<<MD_DISK_FAULTY
)))
386 if (devices
[j
].events
+event_margin
>=
387 devices
[most_recent
].events
) {
388 devices
[j
].uptodate
= 1;
389 if (i
< info
.array
.raid_disks
) {
396 while (force
&& !enough(info
.array
.level
, info
.array
.raid_disks
,
399 /* Choose the newest best drive which is
400 * not up-to-date, update the superblock
405 for (i
=0; i
<info
.array
.raid_disks
&& i
< bestcnt
; i
++) {
408 !devices
[j
].uptodate
&&
409 devices
[j
].events
> 0 &&
411 devices
[j
].events
> devices
[chosen_drive
].events
))
414 if (chosen_drive
< 0)
417 fprintf(stderr
, Name
": forcing event count in %s(%d) from %d upto %d\n",
418 devices
[chosen_drive
].devname
, devices
[chosen_drive
].raid_disk
,
419 (int)(devices
[chosen_drive
].events
),
420 (int)(devices
[most_recent
].events
));
421 fd
= dev_open(devices
[chosen_drive
].devname
, O_RDWR
|O_EXCL
);
423 fprintf(stderr
, Name
": Couldn't open %s for write - not updating\n",
424 devices
[chosen_drive
].devname
);
425 devices
[chosen_drive
].events
= 0;
428 if (st
->ss
->load_super(st
,fd
, &super
, NULL
)) {
430 fprintf(stderr
, Name
": RAID superblock disappeared from %s - not updating.\n",
431 devices
[chosen_drive
].devname
);
432 devices
[chosen_drive
].events
= 0;
435 info
.events
= devices
[most_recent
].events
;
436 st
->ss
->update_super(&info
, super
, "force", devices
[chosen_drive
].devname
, verbose
);
438 if (st
->ss
->store_super(st
, fd
, super
)) {
440 fprintf(stderr
, Name
": Could not re-write superblock on %s\n",
441 devices
[chosen_drive
].devname
);
442 devices
[chosen_drive
].events
= 0;
447 devices
[chosen_drive
].events
= devices
[most_recent
].events
;
448 devices
[chosen_drive
].uptodate
= 1;
449 avail
[chosen_drive
] = 1;
454 /* Now we want to look at the superblock which the kernel will base things on
455 * and compare the devices that we think are working with the devices that the
456 * superblock thinks are working.
457 * If there are differences and --force is given, then update this chosen
462 for (i
=0; chosen_drive
< 0 && i
<bestcnt
; i
++) {
468 if (!devices
[j
].uptodate
)
471 if ((fd
=dev_open(devices
[j
].devname
, O_RDONLY
|O_EXCL
))< 0) {
472 fprintf(stderr
, Name
": Cannot open %s: %s\n",
473 devices
[j
].devname
, strerror(errno
));
476 if (st
->ss
->load_super(st
,fd
, &super
, NULL
)) {
478 fprintf(stderr
, Name
": RAID superblock has disappeared from %s\n",
485 fprintf(stderr
, Name
": No suitable drives found for %s\n", mddev
);
488 st
->ss
->getinfo_super(&info
, &ident2
, super
);
489 for (i
=0; i
<bestcnt
; i
++) {
491 unsigned int desired_state
;
493 if (i
< info
.array
.raid_disks
)
494 desired_state
= (1<<MD_DISK_ACTIVE
) | (1<<MD_DISK_SYNC
);
500 if (!devices
[j
].uptodate
)
502 info
.disk
.number
= devices
[j
].disk_nr
;
503 info
.disk
.raid_disk
= i
;
504 info
.disk
.state
= desired_state
;
506 if (devices
[j
].uptodate
&&
507 st
->ss
->update_super(&info
, super
, "assemble", NULL
, verbose
)) {
510 fprintf(stderr
, Name
": "
511 "clearing FAULTY flag for device %d in %s for %s\n",
512 j
, mddev
, devices
[j
].devname
);
516 fprintf(stderr
, Name
": "
517 "device %d in %s has wrong state in superblock, but %s seems ok\n",
518 i
, mddev
, devices
[j
].devname
);
522 if (!devices
[j
].uptodate
&&
523 !(super
.disks
[i
].state
& (1 << MD_DISK_FAULTY
))) {
524 fprintf(stderr
, Name
": devices %d of %s is not marked FAULTY in superblock, but cannot be found\n",
529 if (force
&& okcnt
== info
.array
.raid_disks
-1) {
530 /* FIXME check event count */
531 change
+= st
->ss
->update_super(&info
, super
, "force",
532 devices
[chosen_drive
].devname
, verbose
);
537 fd
= dev_open(devices
[chosen_drive
].devname
, O_RDWR
|O_EXCL
);
539 fprintf(stderr
, Name
": Could open %s for write - cannot Assemble array.\n",
540 devices
[chosen_drive
].devname
);
543 if (st
->ss
->store_super(st
, fd
, super
)) {
545 fprintf(stderr
, Name
": Could not re-write superblock on %s\n",
546 devices
[chosen_drive
].devname
);
552 /* count number of in-sync devices according to the superblock.
553 * We must have this number to start the array without -s or -R
555 req_cnt
= info
.array
.working_disks
;
557 /* Almost ready to actually *do* something */
560 if ((vers
% 100) >= 1) { /* can use different versions */
561 mdu_array_info_t inf
;
562 memset(&inf
, 0, sizeof(inf
));
563 inf
.major_version
= st
->ss
->major
;
564 inf
.minor_version
= st
->minor_version
;
565 rv
= ioctl(mdfd
, SET_ARRAY_INFO
, &inf
);
567 rv
= ioctl(mdfd
, SET_ARRAY_INFO
, NULL
);
570 fprintf(stderr
, Name
": SET_ARRAY_INFO failed for %s: %s\n",
571 mddev
, strerror(errno
));
574 if (ident
->bitmap_fd
>= 0) {
575 if (ioctl(mdfd
, SET_BITMAP_FILE
, ident
->bitmap_fd
) != 0) {
576 fprintf(stderr
, Name
": SET_BITMAP_FILE failed.\n");
581 /* First, add the raid disks, but add the chosen one last */
582 for (i
=0; i
<= bestcnt
; i
++) {
586 if (j
== chosen_drive
)
591 if (j
>= 0 /* && devices[j].uptodate */) {
592 mdu_disk_info_t disk
;
593 memset(&disk
, 0, sizeof(disk
));
594 disk
.major
= devices
[j
].major
;
595 disk
.minor
= devices
[j
].minor
;
596 if (ioctl(mdfd
, ADD_NEW_DISK
, &disk
)!=0) {
597 fprintf(stderr
, Name
": failed to add %s to %s: %s\n",
601 if (i
< info
.array
.raid_disks
|| i
== bestcnt
)
605 } else if (verbose
> 0)
606 fprintf(stderr
, Name
": added %s to %s as %d\n",
607 devices
[j
].devname
, mddev
, devices
[j
].raid_disk
);
608 } else if (verbose
> 0 && i
< info
.array
.raid_disks
)
609 fprintf(stderr
, Name
": no uptodate device for slot %d of %s\n",
615 ( enough(info
.array
.level
, info
.array
.raid_disks
, info
.array
.layout
, avail
, okcnt
) &&
616 (okcnt
>= req_cnt
|| start_partial_ok
)
618 if (ioctl(mdfd
, RUN_ARRAY
, NULL
)==0) {
620 fprintf(stderr
, Name
": %s has been started with %d drive%s",
621 mddev
, okcnt
, okcnt
==1?"":"s");
622 if (okcnt
< info
.array
.raid_disks
)
623 fprintf(stderr
, " (out of %d)", info
.array
.raid_disks
);
625 fprintf(stderr
, " and %d spare%s", sparecnt
, sparecnt
==1?"":"s");
626 fprintf(stderr
, ".\n");
630 fprintf(stderr
, Name
": failed to RUN_ARRAY %s: %s\n",
631 mddev
, strerror(errno
));
635 fprintf(stderr
, Name
": %s assembled from %d drive%s, but not started.\n",
636 mddev
, okcnt
, okcnt
==1?"":"s");
640 fprintf(stderr
, Name
": %s assembled from %d drive%s", mddev
, okcnt
, okcnt
==1?"":"s");
642 fprintf(stderr
, " and %d spare%s", sparecnt
, sparecnt
==1?"":"s");
643 if (!enough(info
.array
.level
, info
.array
.raid_disks
, info
.array
.layout
, avail
, okcnt
))
644 fprintf(stderr
, " - not enough to start the array.\n");
646 if (req_cnt
== info
.array
.raid_disks
)
647 fprintf(stderr
, " - need all %d to start it", req_cnt
);
649 fprintf(stderr
, " - need %d of %d to start", req_cnt
, info
.array
.raid_disks
);
650 fprintf(stderr
, " (use --run to insist).\n");
655 /* The "chosen_drive" is a good choice, and if necessary, the superblock has
656 * been updated to point to the current locations of devices.
657 * so we can just start the array
660 dev
= makedev(devices
[chosen_drive
].major
,
661 devices
[chosen_drive
].minor
);
662 if (ioctl(mdfd
, START_ARRAY
, dev
)) {
663 fprintf(stderr
, Name
": Cannot start array: %s\n",