]>
git.ipfire.org Git - thirdparty/mdadm.git/blob - raid6check.c
2 * raid6check - extended consistency check for RAID-6
4 * Copyright (C) 2011 Piergiorgio Sartor
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
21 * Author: Piergiorgio Sartor
22 * Based on "restripe.c" from "mdadm" codebase
30 int geo_map(int block
, unsigned long long stripe
, int raid_disks
,
31 int level
, int layout
);
32 void qsyndrome(uint8_t *p
, uint8_t *q
, uint8_t **sources
, int disks
, int size
);
33 void make_tables(void);
34 void ensure_zero_has_size(int chunk_size
);
35 void raid6_datap_recov(int disks
, size_t bytes
, int faila
, uint8_t **ptrs
);
36 void raid6_2data_recov(int disks
, size_t bytes
, int faila
, int failb
,
38 void xor_blocks(char *target
, char **sources
, int disks
, int size
);
41 /* Collect per stripe consistency information */
42 void raid6_collect(int chunk_size
, uint8_t *p
, uint8_t *q
,
43 char *chunkP
, char *chunkQ
, int *results
)
48 extern uint8_t raid6_gflog
[];
50 for(i
= 0; i
< chunk_size
; i
++) {
51 Px
= (uint8_t)chunkP
[i
] ^ (uint8_t)p
[i
];
52 Qx
= (uint8_t)chunkQ
[i
] ^ (uint8_t)q
[i
];
54 if((Px
!= 0) && (Qx
== 0))
57 if((Px
== 0) && (Qx
!= 0))
60 if((Px
!= 0) && (Qx
!= 0)) {
61 data_id
= (raid6_gflog
[Qx
] - raid6_gflog
[Px
]);
62 if(data_id
< 0) data_id
+= 255;
66 if((Px
== 0) && (Qx
== 0))
71 /* Try to find out if a specific disk has problems */
72 int raid6_stats(int *results
, int raid_disks
, int chunk_size
)
75 int curr_broken_disk
= -255;
76 int prev_broken_disk
= -255;
77 int broken_status
= 0;
79 for(i
= 0; i
< chunk_size
; i
++) {
81 if(results
[i
] != -255)
82 curr_broken_disk
= results
[i
];
84 if(curr_broken_disk
>= raid_disks
)
87 switch(broken_status
) {
89 if(curr_broken_disk
!= -255) {
90 prev_broken_disk
= curr_broken_disk
;
96 if(curr_broken_disk
!= prev_broken_disk
)
102 curr_broken_disk
= prev_broken_disk
= -65535;
107 return curr_broken_disk
;
110 int lock_stripe(struct mdinfo
*info
, unsigned long long start
,
111 int chunk_size
, int data_disks
, sighandler_t
*sig
) {
113 if(mlockall(MCL_CURRENT
| MCL_FUTURE
) != 0) {
117 sig
[0] = signal(SIGTERM
, SIG_IGN
);
118 sig
[1] = signal(SIGINT
, SIG_IGN
);
119 sig
[2] = signal(SIGQUIT
, SIG_IGN
);
121 rv
= sysfs_set_num(info
, NULL
, "suspend_lo", start
* chunk_size
* data_disks
);
122 rv
|= sysfs_set_num(info
, NULL
, "suspend_hi", (start
+ 1) * chunk_size
* data_disks
);
126 int unlock_all_stripes(struct mdinfo
*info
, sighandler_t
*sig
) {
128 rv
= sysfs_set_num(info
, NULL
, "suspend_lo", 0x7FFFFFFFFFFFFFFFULL
);
129 rv
|= sysfs_set_num(info
, NULL
, "suspend_hi", 0);
130 rv
|= sysfs_set_num(info
, NULL
, "suspend_lo", 0);
132 signal(SIGQUIT
, sig
[2]);
133 signal(SIGINT
, sig
[1]);
134 signal(SIGTERM
, sig
[0]);
136 if(munlockall() != 0)
142 int check_stripes(struct mdinfo
*info
, int *source
, unsigned long long *offsets
,
143 int raid_disks
, int chunk_size
, int level
, int layout
,
144 unsigned long long start
, unsigned long long length
, char *name
[],
145 int repair
, int failed_disk1
, int failed_disk2
)
147 /* read the data and p and q blocks, and check we got them right */
148 char *stripe_buf
= xmalloc(raid_disks
* chunk_size
);
149 char **stripes
= xmalloc(raid_disks
* sizeof(char*));
150 char **blocks
= xmalloc(raid_disks
* sizeof(char*));
151 int *block_index_for_slot
= xmalloc(raid_disks
* sizeof(int));
152 uint8_t *p
= xmalloc(chunk_size
);
153 uint8_t *q
= xmalloc(chunk_size
);
154 int *results
= xmalloc(chunk_size
* sizeof(int));
155 sighandler_t
*sig
= xmalloc(3 * sizeof(sighandler_t
));
159 int data_disks
= raid_disks
- 2;
162 extern int tables_ready
;
167 for ( i
= 0 ; i
< raid_disks
; i
++)
168 stripes
[i
] = stripe_buf
+ i
* chunk_size
;
173 printf("pos --> %llu\n", start
);
175 err
= lock_stripe(info
, start
, chunk_size
, data_disks
, sig
);
178 unlock_all_stripes(info
, sig
);
181 for (i
= 0 ; i
< raid_disks
; i
++) {
182 lseek64(source
[i
], offsets
[i
] + start
* chunk_size
, 0);
183 read(source
[i
], stripes
[i
], chunk_size
);
185 err
= unlock_all_stripes(info
, sig
);
189 for (i
= 0 ; i
< data_disks
; i
++) {
190 int disk
= geo_map(i
, start
, raid_disks
, level
, layout
);
191 blocks
[i
] = stripes
[disk
];
192 block_index_for_slot
[disk
] = i
;
193 printf("%d->%d\n", i
, disk
);
196 qsyndrome(p
, q
, (uint8_t**)blocks
, data_disks
, chunk_size
);
197 diskP
= geo_map(-1, start
, raid_disks
, level
, layout
);
198 diskQ
= geo_map(-2, start
, raid_disks
, level
, layout
);
199 blocks
[data_disks
] = stripes
[diskP
];
200 block_index_for_slot
[diskP
] = data_disks
;
201 blocks
[data_disks
+1] = stripes
[diskQ
];
202 block_index_for_slot
[diskQ
] = data_disks
+1;
204 if (memcmp(p
, stripes
[diskP
], chunk_size
) != 0) {
205 printf("P(%d) wrong at %llu\n", diskP
, start
);
207 if (memcmp(q
, stripes
[diskQ
], chunk_size
) != 0) {
208 printf("Q(%d) wrong at %llu\n", diskQ
, start
);
210 raid6_collect(chunk_size
, p
, q
, stripes
[diskP
], stripes
[diskQ
], results
);
211 disk
= raid6_stats(results
, raid_disks
, chunk_size
);
214 disk
= geo_map(disk
, start
, raid_disks
, level
, layout
);
217 printf("Error detected at %llu: possible failed disk slot: %d --> %s\n",
218 start
, disk
, name
[disk
]);
221 printf("Error detected at %llu: disk slot unknown\n", start
);
224 printf("Repairing stripe %llu\n", start
);
225 printf("Assuming slots %d (%s) and %d (%s) are incorrect\n",
226 failed_disk1
, name
[failed_disk1
],
227 failed_disk2
, name
[failed_disk2
]);
229 if (failed_disk1
== diskQ
|| failed_disk2
== diskQ
) {
230 char *all_but_failed_blocks
[data_disks
];
231 int failed_data_or_p
;
232 int failed_block_index
;
234 if (failed_disk1
== diskQ
)
235 failed_data_or_p
= failed_disk2
;
237 failed_data_or_p
= failed_disk1
;
238 printf("Repairing D/P(%d) and Q\n", failed_data_or_p
);
239 failed_block_index
= block_index_for_slot
[failed_data_or_p
];
240 for (i
=0; i
< data_disks
; i
++)
241 if (failed_block_index
== i
)
242 all_but_failed_blocks
[i
] = stripes
[diskP
];
244 all_but_failed_blocks
[i
] = blocks
[i
];
245 xor_blocks(stripes
[failed_data_or_p
],
246 all_but_failed_blocks
, data_disks
, chunk_size
);
247 qsyndrome(p
, (uint8_t*)stripes
[diskQ
], (uint8_t**)blocks
, data_disks
, chunk_size
);
249 ensure_zero_has_size(chunk_size
);
250 if (failed_disk1
== diskP
|| failed_disk2
== diskP
) {
251 int failed_data
, failed_block_index
;
252 if (failed_disk1
== diskP
)
253 failed_data
= failed_disk2
;
255 failed_data
= failed_disk1
;
256 failed_block_index
= block_index_for_slot
[failed_data
];
257 printf("Repairing D(%d) and P\n", failed_data
);
258 raid6_datap_recov(raid_disks
, chunk_size
, failed_block_index
, (uint8_t**)blocks
);
260 printf("Repairing D and D\n");
261 int failed_block_index1
= block_index_for_slot
[failed_disk1
];
262 int failed_block_index2
= block_index_for_slot
[failed_disk2
];
263 if (failed_block_index1
> failed_block_index2
) {
264 int t
= failed_block_index1
;
265 failed_block_index1
= failed_block_index2
;
266 failed_block_index2
= t
;
268 raid6_2data_recov(raid_disks
, chunk_size
, failed_block_index1
, failed_block_index2
, (uint8_t**)blocks
);
272 err
= lock_stripe(info
, start
, chunk_size
, data_disks
, sig
);
275 unlock_all_stripes(info
, sig
);
279 lseek64(source
[failed_disk1
], offsets
[failed_disk1
] + start
* chunk_size
, 0);
280 write(source
[failed_disk1
], stripes
[failed_disk1
], chunk_size
);
281 lseek64(source
[failed_disk2
], offsets
[failed_disk2
] + start
* chunk_size
, 0);
282 write(source
[failed_disk2
], stripes
[failed_disk2
], chunk_size
);
284 err
= unlock_all_stripes(info
, sig
);
287 } else if (disk
>= 0 && repair
== 2) {
288 printf("Auto-repairing slot %d (%s)\n", disk
, name
[disk
]);
290 qsyndrome(p
, (uint8_t*)stripes
[diskQ
], (uint8_t**)blocks
, data_disks
, chunk_size
);
292 char *all_but_failed_blocks
[data_disks
];
293 int failed_block_index
= block_index_for_slot
[disk
];
294 for (i
=0; i
< data_disks
; i
++)
295 if (failed_block_index
== i
)
296 all_but_failed_blocks
[i
] = stripes
[diskP
];
298 all_but_failed_blocks
[i
] = blocks
[i
];
299 xor_blocks(stripes
[disk
],
300 all_but_failed_blocks
, data_disks
, chunk_size
);
303 err
= lock_stripe(info
, start
, chunk_size
, data_disks
, sig
);
306 unlock_all_stripes(info
, sig
);
310 lseek64(source
[disk
], offsets
[disk
] + start
* chunk_size
, 0);
311 write(source
[disk
], stripes
[disk
], chunk_size
);
313 err
= unlock_all_stripes(info
, sig
);
335 unsigned long long getnum(char *str
, char **err
)
338 unsigned long long rv
= strtoull(str
, &e
, 10);
346 int main(int argc
, char *argv
[])
348 /* md_device start length */
351 char **disk_name
= NULL
;
352 unsigned long long *offsets
= NULL
;
359 int failed_disk1
, failed_disk2
;
360 unsigned long long start
, length
;
363 struct mdinfo
*info
= NULL
, *comp
= NULL
;
367 char *prg
= strrchr(argv
[0], '/');
375 fprintf(stderr
, "Usage: %s md_device start_stripe length_stripes [autorepair]\n", prg
);
376 fprintf(stderr
, " or: %s md_device repair stripe failed_slot_1 failed_slot_2\n", prg
);
381 mdfd
= open(argv
[1], O_RDONLY
);
384 fprintf(stderr
, "%s: cannot open %s\n", prg
, argv
[1]);
389 info
= sysfs_read(mdfd
, -1,
401 fprintf(stderr
, "%s: Error reading sysfs information of %s\n", prg
, argv
[1]);
406 if(info
->array
.level
!= level
) {
407 fprintf(stderr
, "%s: %s not a RAID-6\n", prg
, argv
[1]);
412 if(info
->array
.failed_disks
> 0) {
413 fprintf(stderr
, "%s: %s degraded array\n", prg
, argv
[1]);
418 printf("layout: %d\n", info
->array
.layout
);
419 printf("disks: %d\n", info
->array
.raid_disks
);
420 printf("component size: %llu\n", info
->component_size
* 512);
421 printf("total stripes: %llu\n", (info
->component_size
* 512) / info
->array
.chunk_size
);
422 printf("chunk size: %d\n", info
->array
.chunk_size
);
426 for(i
= 0, active_disks
= 0; active_disks
< info
->array
.raid_disks
; i
++) {
427 printf("disk: %d - offset: %llu - size: %llu - name: %s - slot: %d\n",
428 i
, comp
->data_offset
* 512, comp
->component_size
* 512,
429 map_dev(comp
->disk
.major
, comp
->disk
.minor
, 0),
430 comp
->disk
.raid_disk
);
431 if(comp
->disk
.raid_disk
>= 0)
439 raid_disks
= info
->array
.raid_disks
;
440 chunk_size
= info
->array
.chunk_size
;
441 layout
= info
->array
.layout
;
442 if (strcmp(argv
[2], "repair")==0) {
444 fprintf(stderr
, "For repair mode, call %s md_device repair stripe failed_slot_1 failed_slot_2\n", prg
);
449 start
= getnum(argv
[3], &err
);
451 failed_disk1
= getnum(argv
[4], &err
);
452 failed_disk2
= getnum(argv
[5], &err
);
454 if(failed_disk1
>= info
->array
.raid_disks
) {
455 fprintf(stderr
, "%s: failed_slot_1 index is higher than number of devices in raid\n", prg
);
459 if(failed_disk2
>= info
->array
.raid_disks
) {
460 fprintf(stderr
, "%s: failed_slot_2 index is higher than number of devices in raid\n", prg
);
464 if(failed_disk1
== failed_disk2
) {
465 fprintf(stderr
, "%s: failed_slot_1 and failed_slot_2 are the same\n", prg
);
471 start
= getnum(argv
[2], &err
);
472 length
= getnum(argv
[3], &err
);
473 if (argc
>= 5 && strcmp(argv
[4], "autorepair")==0)
478 fprintf(stderr
, "%s: Bad number: %s\n", prg
, err
);
483 if(start
> ((info
->component_size
* 512) / chunk_size
)) {
484 start
= (info
->component_size
* 512) / chunk_size
;
485 fprintf(stderr
, "%s: start beyond disks size\n", prg
);
489 ((length
+ start
) > ((info
->component_size
* 512) / chunk_size
))) {
490 length
= (info
->component_size
* 512) / chunk_size
- start
;
493 disk_name
= xmalloc(raid_disks
* sizeof(*disk_name
));
494 fds
= xmalloc(raid_disks
* sizeof(*fds
));
495 offsets
= xcalloc(raid_disks
, sizeof(*offsets
));
496 buf
= xmalloc(raid_disks
* chunk_size
);
498 for(i
=0; i
<raid_disks
; i
++) {
504 for (i
=0, active_disks
=0; active_disks
<raid_disks
; i
++) {
505 int disk_slot
= comp
->disk
.raid_disk
;
507 disk_name
[disk_slot
] = map_dev(comp
->disk
.major
, comp
->disk
.minor
, 0);
508 offsets
[disk_slot
] = comp
->data_offset
* 512;
509 fds
[disk_slot
] = open(disk_name
[disk_slot
], O_RDWR
);
510 if (fds
[disk_slot
] < 0) {
511 perror(disk_name
[disk_slot
]);
512 fprintf(stderr
,"%s: cannot open %s\n", prg
, disk_name
[disk_slot
]);
521 int rv
= check_stripes(info
, fds
, offsets
,
522 raid_disks
, chunk_size
, level
, layout
,
523 start
, length
, disk_name
, repair
, failed_disk1
, failed_disk2
);
526 "%s: check_stripes returned %d\n", prg
, rv
);
534 for(i
= 0; i
< raid_disks
; i
++)