]>
git.ipfire.org Git - thirdparty/mdadm.git/blob - Assemble.c
786880221a1d24ffe2589c035b132923b8407cd1
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
34 int Assemble(char *mddev
, int mdfd
,
35 mddev_ident_t ident
, char *conffile
,
37 int readonly
, int runstop
,
38 int verbose
, int force
)
41 * The task of Assemble is to find a collection of
42 * devices that should (according to their superblocks)
43 * form an array, and to give this collection to the MD driver.
44 * In Linux-2.4 and later, this involves submitting a
45 * SET_ARRAY_INFO ioctl with no arg - to prepare
46 * the array - and then submit a number of
47 * ADD_NEW_DISK ioctls to add disks into
48 * the array. Finally RUN_ARRAY might
49 * be submitted to start the array.
51 * Much of the work of Assemble is in finding and/or
52 * checking the disks to make sure they look right.
54 * If mddev is not set, then scan must be and we
55 * read through the config file for dev+uuid mapping
56 * We recurse, setting mddev, for each device that
58 * - has a valid uuid (or any uuid if !uuidset
60 * If mddev is set, we try to determine state of md.
61 * check version - must be at least 0.90.0
62 * check kernel version. must be at least 2.4.
63 * If not, we can possibly fall back on START_ARRAY
64 * Try to GET_ARRAY_INFO.
65 * If possible, give up
66 * If not, try to STOP_ARRAY just to make sure
68 * If !uuidset and scan, look in conf-file for uuid
69 * If not found, give up
70 * If !devlist and scan and uuidset, get list of devs from conf-file
73 * Check superblock - discard if bad
74 * Check uuid (set if we don't have one) - discard if no match
75 * Check superblock similarity if we have a superbloc - discard if different
76 * Record events, devicenum, utime
77 * This should give us a list of devices for the array
78 * We should collect the most recent event and utime numbers
80 * Count disks with recent enough event count
81 * While force && !enough disks
82 * Choose newest rejected disks, update event count
83 * mark clean and rewrite superblock
86 * foreach device with recent events : ADD_NEW_DISK
87 * if runstop == 1 || "enough" disks and runstop==0 -> RUN_ARRAY
89 * Check the device numbers in superblock are right
90 * update superblock if any changes
96 mdu_array_info_t array
;
97 mdp_super_t first_super
, super
;
101 int oldmajor
, oldminor
;
106 } devices
[MD_SB_DISKS
];
107 int best
[MD_SB_DISKS
]; /* indexed by raid_disk */
108 int devcnt
= 0, okcnt
, sparecnt
;
114 int start_partial_ok
= force
|| devlist
==NULL
;
116 vers
= md_get_version(mdfd
);
118 fprintf(stderr
, Name
": %s appears not to be an md device.\n", mddev
);
122 fprintf(stderr
, Name
": Assemble requires driver version 0.90.0 or later.\n"
123 " Upgrade your kernel or try --build\n");
126 if (get_linux_version() < 2004000)
129 if (ioctl(mdfd
, GET_ARRAY_INFO
, &array
)>=0) {
130 fprintf(stderr
, Name
": device %s already active - cannot assemble it\n",
134 ioctl(mdfd
, STOP_ARRAY
, NULL
); /* just incase it was started but has no content */
137 * If any subdevs are listed, then any that don't
138 * match ident are discarded. Remainder must all match and
140 * If no subdevs, then we scan all devices in the config file, but
141 * there must be something in the identity
145 ident
->uuid_set
== 0 &&
146 ident
->super_minor
< 0 &&
147 ident
->devices
== NULL
) {
148 fprintf(stderr
, Name
": No identity information available for %s - cannot assemble.\n",
153 devlist
= conf_get_devs(conffile
);
156 first_super
.md_magic
= 0;
157 for (i
=0; i
<MD_SB_DISKS
; i
++)
161 fprintf(stderr
, Name
": looking for devices for %s\n",
171 devname
= devlist
->devname
;
172 devlist
= devlist
->next
;
174 if (ident
->devices
&&
175 !match_oneof(ident
->devices
, devname
))
178 dfd
= open(devname
, O_RDONLY
, 0);
180 if (inargv
|| verbose
)
181 fprintf(stderr
, Name
": cannot open device %s: %s\n",
182 devname
, strerror(errno
));
183 } else if (fstat(dfd
, &stb
)< 0) {
185 fprintf(stderr
, Name
": fstat failed for %s: %s\n",
186 devname
, strerror(errno
));
188 } else if ((stb
.st_mode
& S_IFMT
) != S_IFBLK
) {
189 fprintf(stderr
, Name
": %s is not a block device.\n",
192 } else if (load_super(dfd
, &super
)) {
193 if (inargv
|| verbose
)
194 fprintf( stderr
, Name
": no RAID superblock on %s\n",
199 uuid_from_super(this_uuid
, &super
);
203 if (ident
->uuid_set
&&
204 (!havesuper
|| same_uuid(this_uuid
, ident
->uuid
)==0)) {
205 if (inargv
|| verbose
)
206 fprintf(stderr
, Name
": %s has wrong uuid.\n",
210 if (ident
->super_minor
>= 0 &&
211 (!havesuper
|| ident
->super_minor
!= super
.md_minor
)) {
212 if (inargv
|| verbose
)
213 fprintf(stderr
, Name
": %s has wrong super-minor.\n",
217 if (ident
->level
!= -10 &&
218 (!havesuper
|| ident
->level
!= super
.level
)) {
219 if (inargv
|| verbose
)
220 fprintf(stderr
, Name
": %s has wrong raid level.\n",
224 if (ident
->raid_disks
!= -1 &&
225 (!havesuper
|| ident
->raid_disks
!= super
.raid_disks
)) {
226 if (inargv
|| verbose
)
227 fprintf(stderr
, Name
": %s requires wrong number of drives.\n",
232 /* If we are this far, then we are commited to this device.
233 * If the super_block doesn't exist, or doesn't match others,
234 * then we cannot continue
238 fprintf(stderr
, Name
": %s has no superblock - assembly aborted\n",
242 if (compare_super(&first_super
, &super
)) {
243 fprintf(stderr
, Name
": superblock on %s doesn't match others - assembly aborted\n",
248 if (devcnt
>= MD_SB_DISKS
) {
249 fprintf(stderr
, Name
": ouch - too many devices appear to be in this array. Ignoring %s\n",
254 fprintf(stderr
, Name
": %s is identified as a member of %s, slot %d.\n",
255 devname
, mddev
, super
.this_disk
.raid_disk
);
256 devices
[devcnt
].devname
= devname
;
257 devices
[devcnt
].major
= MAJOR(stb
.st_rdev
);
258 devices
[devcnt
].minor
= MINOR(stb
.st_rdev
);
259 devices
[devcnt
].oldmajor
= super
.this_disk
.major
;
260 devices
[devcnt
].oldminor
= super
.this_disk
.minor
;
261 devices
[devcnt
].events
= md_event(&super
);
262 devices
[devcnt
].utime
= super
.utime
;
263 devices
[devcnt
].raid_disk
= super
.this_disk
.raid_disk
;
264 devices
[devcnt
].uptodate
= 0;
265 if (most_recent
< devcnt
) {
266 if (devices
[devcnt
].events
267 > devices
[most_recent
].events
)
268 most_recent
= devcnt
;
270 i
= devices
[devcnt
].raid_disk
;
271 if (i
>=0 && i
< MD_SB_DISKS
)
273 || devices
[best
[i
]].events
< devices
[devcnt
].events
)
280 fprintf(stderr
, Name
": no devices found for %s\n",
284 /* now we have some devices that might be suitable.
289 for (i
=0; i
< MD_SB_DISKS
;i
++) {
291 int event_margin
= !force
;
293 if (devices
[j
].events
+event_margin
>=
294 devices
[most_recent
].events
) {
295 devices
[j
].uptodate
= 1;
296 if (i
< first_super
.raid_disks
)
302 while (force
&& !enough(first_super
.level
, first_super
.raid_disks
, okcnt
)) {
303 /* Choose the newest best drive which is
304 * not up-to-date, update the superblock
309 for (i
=0; i
<first_super
.raid_disks
; i
++) {
312 !devices
[j
].uptodate
&&
313 devices
[j
].events
> 0 &&
315 devices
[j
].events
> devices
[chosen_drive
].events
))
318 if (chosen_drive
< 0)
320 fprintf(stderr
, Name
": forcing event count in %s(%d) from %d upto %d\n",
321 devices
[chosen_drive
].devname
, devices
[chosen_drive
].raid_disk
,
322 (int)(devices
[chosen_drive
].events
),
323 (int)(devices
[most_recent
].events
));
324 fd
= open(devices
[chosen_drive
].devname
, O_RDWR
);
326 fprintf(stderr
, Name
": Couldn't open %s for write - not updating\n",
327 devices
[chosen_drive
].devname
);
328 devices
[chosen_drive
].events
= 0;
331 if (load_super(fd
, &super
)) {
333 fprintf(stderr
, Name
": RAID superblock disappeared from %s - not updating.\n",
334 devices
[chosen_drive
].devname
);
335 devices
[chosen_drive
].events
= 0;
338 super
.events_hi
= (devices
[most_recent
].events
>>32)&0xFFFFFFFF;
339 super
.events_lo
= (devices
[most_recent
].events
)&0xFFFFFFFF;
340 super
.sb_csum
= calc_sb_csum(&super
);
341 /*DRYRUN*/ if (store_super(fd
, &super
)) {
343 fprintf(stderr
, Name
": Could not re-write superblock on %s\n",
344 devices
[chosen_drive
].devname
);
345 devices
[chosen_drive
].events
= 0;
349 devices
[chosen_drive
].events
= devices
[most_recent
].events
;
350 devices
[chosen_drive
].uptodate
= 1;
354 /* Now we want to look at the superblock which the kernel will base things on
355 * and compare the devices that we think are working with the devices that the
356 * superblock thinks are working.
357 * If there are differences and --force is given, then update this chosen
361 for (i
=0; chosen_drive
< 0 && i
<MD_SB_DISKS
; i
++) {
366 if (!devices
[j
].uptodate
)
369 if ((fd
=open(devices
[j
].devname
, O_RDONLY
))< 0) {
370 fprintf(stderr
, Name
": Cannot open %s: %s\n",
371 devices
[j
].devname
, strerror(errno
));
374 if (load_super(fd
, &super
)) {
376 fprintf(stderr
, Name
": RAID superblock has disappeared from %s\n",
383 for (i
=0; i
<MD_SB_DISKS
; i
++) {
387 if (i
< super
.raid_disks
)
388 desired_state
= (1<<MD_DISK_ACTIVE
) | (1<<MD_DISK_SYNC
);
394 if (!devices
[j
].uptodate
)
398 if (devices
[j
].major
!= super
.disks
[i
].major
||
399 devices
[j
].minor
!= super
.disks
[i
].minor
) {
401 super
.disks
[i
].major
= devices
[j
].major
;
402 super
.disks
[i
].minor
= devices
[j
].minor
;
405 if (devices
[j
].oldmajor
!= super
.disks
[i
].major
||
406 devices
[j
].oldminor
!= super
.disks
[i
].minor
) {
408 super
.disks
[i
].major
= devices
[j
].oldmajor
;
409 super
.disks
[i
].minor
= devices
[j
].oldminor
;
411 if (devices
[j
].uptodate
&&
412 (super
.disks
[i
].state
!= desired_state
)) {
414 fprintf(stderr
, Name
": "
415 "clearing FAULTY flag for device %d in %s for %s\n",
416 j
, mddev
, devices
[j
].devname
);
417 super
.disks
[i
].state
= desired_state
;
420 fprintf(stderr
, Name
": "
421 "device %d in %s has wrong state in superblock, but %s seems ok\n",
422 i
, mddev
, devices
[j
].devname
);
425 if (!devices
[j
].uptodate
&&
426 !(super
.disks
[i
].state
& (1 << MD_DISK_FAULTY
))) {
427 fprintf(stderr
, Name
": devices %d of %s is not marked FAULTY in superblock, but cannot be found\n",
432 if ((force
&& (change
& 2))
433 || (old_linux
&& (change
& 1))) {
435 super
.sb_csum
= calc_sb_csum(&super
);
436 fd
= open(devices
[chosen_drive
].devname
, O_RDWR
);
438 fprintf(stderr
, Name
": Could open %s for write - cannot Assemble array.\n",
439 devices
[chosen_drive
].devname
);
442 if (store_super(fd
, &super
)) {
444 fprintf(stderr
, Name
": Could not re-write superblock on %s\n",
445 devices
[chosen_drive
].devname
);
452 /* Almost ready to actually *do* something */
454 if (ioctl(mdfd
, SET_ARRAY_INFO
, NULL
) != 0) {
455 fprintf(stderr
, Name
": SET_ARRAY_INFO failed for %s: %s\n",
456 mddev
, strerror(errno
));
459 /* First, add the raid disks, but add the chosen one last */
460 for (i
=0; i
<=MD_SB_DISKS
; i
++) {
462 if (i
< MD_SB_DISKS
) {
464 if (j
== chosen_drive
)
469 if (j
>= 0 && devices
[j
].uptodate
) {
470 mdu_disk_info_t disk
;
471 memset(&disk
, 0, sizeof(disk
));
472 disk
.major
= devices
[j
].major
;
473 disk
.minor
= devices
[j
].minor
;
474 if (ioctl(mdfd
, ADD_NEW_DISK
, &disk
)!=0) {
475 fprintf(stderr
, Name
": failed to add %s to %s: %s\n",
479 if (i
< first_super
.raid_disks
)
484 fprintf(stderr
, Name
": added %s to %s as %d\n",
485 devices
[j
].devname
, mddev
, devices
[j
].raid_disk
);
486 } else if (verbose
&& i
< first_super
.raid_disks
)
487 fprintf(stderr
, Name
": no uptodate device for slot %d of %s\n",
493 ( first_super
.raid_disks
== okcnt
494 || (start_partial_ok
&& enough(first_super
.level
, first_super
.raid_disks
, okcnt
)))
496 if (ioctl(mdfd
, RUN_ARRAY
, NULL
)==0) {
497 fprintf(stderr
, Name
": %s has been started with %d drive%s",
498 mddev
, okcnt
, okcnt
==1?"":"s");
500 fprintf(stderr
, " and %d spare%s", sparecnt
, sparecnt
==1?"":"s");
501 fprintf(stderr
, ".\n");
504 fprintf(stderr
, Name
": failed to RUN_ARRAY %s: %s\n",
505 mddev
, strerror(errno
));
509 fprintf(stderr
, Name
": %s assembled from %d drive%s, but not started.\n",
510 mddev
, okcnt
, okcnt
==1?"":"s");
513 fprintf(stderr
, Name
": %s assembled from %d drive%s - not enough to start it (use --run to insist).\n",
514 mddev
, okcnt
, okcnt
==1?"":"s");
517 /* The "chosen_drive" is a good choice, and if necessary, the superblock has
518 * been updated to point to the current locations of devices.
519 * so we can just start the array
522 dev
= MKDEV(devices
[chosen_drive
].major
,
523 devices
[chosen_drive
].minor
);
524 if (ioctl(mdfd
, START_ARRAY
, dev
)) {
525 fprintf(stderr
, Name
": Cannot start array: %s\n",