]>
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 #define CHECK_PAGE_BITS (12)
31 #define CHECK_PAGE_SIZE (1 << CHECK_PAGE_BITS)
39 int geo_map(int block
, unsigned long long stripe
, int raid_disks
,
40 int level
, int layout
);
41 void qsyndrome(uint8_t *p
, uint8_t *q
, uint8_t **sources
, int disks
, int size
);
42 void make_tables(void);
43 void ensure_zero_has_size(int chunk_size
);
44 void raid6_datap_recov(int disks
, size_t bytes
, int faila
, uint8_t **ptrs
);
45 void raid6_2data_recov(int disks
, size_t bytes
, int faila
, int failb
,
47 void xor_blocks(char *target
, char **sources
, int disks
, int size
);
49 /* Collect per stripe consistency information */
50 void raid6_collect(int chunk_size
, uint8_t *p
, uint8_t *q
,
51 char *chunkP
, char *chunkQ
, int *results
)
56 extern uint8_t raid6_gflog
[];
58 for(i
= 0; i
< chunk_size
; i
++) {
59 Px
= (uint8_t)chunkP
[i
] ^ (uint8_t)p
[i
];
60 Qx
= (uint8_t)chunkQ
[i
] ^ (uint8_t)q
[i
];
62 if((Px
!= 0) && (Qx
== 0))
65 if((Px
== 0) && (Qx
!= 0))
68 if((Px
!= 0) && (Qx
!= 0)) {
69 data_id
= (raid6_gflog
[Qx
] - raid6_gflog
[Px
]);
70 if(data_id
< 0) data_id
+= 255;
74 if((Px
== 0) && (Qx
== 0))
79 /* Try to find out if a specific disk has problems in a CHECK_PAGE_SIZE page size */
80 int raid6_stats_blk(int *results
, int raid_disks
)
83 int curr_broken_disk
= -255;
84 int prev_broken_disk
= -255;
85 int broken_status
= 0;
87 for(i
= 0; i
< CHECK_PAGE_SIZE
; i
++) {
89 if(results
[i
] != -255)
90 curr_broken_disk
= results
[i
];
92 if(curr_broken_disk
>= raid_disks
)
95 switch(broken_status
) {
97 if(curr_broken_disk
!= -255) {
98 prev_broken_disk
= curr_broken_disk
;
104 if(curr_broken_disk
!= prev_broken_disk
)
110 curr_broken_disk
= prev_broken_disk
= -65535;
115 return curr_broken_disk
;
118 /* Collect disks status for a strip in CHECK_PAGE_SIZE page size blocks */
119 void raid6_stats(int *disk
, int *results
, int raid_disks
, int chunk_size
)
123 for(i
= 0, j
= 0; i
< chunk_size
; i
+= CHECK_PAGE_SIZE
, j
++) {
124 disk
[j
] = raid6_stats_blk(&results
[i
], raid_disks
);
128 int lock_stripe(struct mdinfo
*info
, unsigned long long start
,
129 int chunk_size
, int data_disks
, sighandler_t
*sig
) {
131 if(mlockall(MCL_CURRENT
| MCL_FUTURE
) != 0) {
135 sig
[0] = signal(SIGTERM
, SIG_IGN
);
136 sig
[1] = signal(SIGINT
, SIG_IGN
);
137 sig
[2] = signal(SIGQUIT
, SIG_IGN
);
139 rv
= sysfs_set_num(info
, NULL
, "suspend_lo", start
* chunk_size
* data_disks
);
140 rv
|= sysfs_set_num(info
, NULL
, "suspend_hi", (start
+ 1) * chunk_size
* data_disks
);
144 int unlock_all_stripes(struct mdinfo
*info
, sighandler_t
*sig
) {
146 rv
= sysfs_set_num(info
, NULL
, "suspend_lo", 0x7FFFFFFFFFFFFFFFULL
);
147 rv
|= sysfs_set_num(info
, NULL
, "suspend_hi", 0);
148 rv
|= sysfs_set_num(info
, NULL
, "suspend_lo", 0);
150 signal(SIGQUIT
, sig
[2]);
151 signal(SIGINT
, sig
[1]);
152 signal(SIGTERM
, sig
[0]);
154 if(munlockall() != 0)
160 int autorepair(int *disk
, int diskP
, int diskQ
, unsigned long long start
, int chunk_size
,
161 char *name
[], int raid_disks
, int data_disks
, char **blocks_page
,
162 char **blocks
, uint8_t *p
, char **stripes
, int *block_index_for_slot
,
163 int *source
, unsigned long long *offsets
)
166 int pages_to_write_count
= 0;
167 int page_to_write
[chunk_size
>> CHECK_PAGE_BITS
];
168 for(j
= 0; j
< (chunk_size
>> CHECK_PAGE_BITS
); j
++) {
170 printf("Auto-repairing slot %d (%s)\n", disk
[j
], name
[disk
[j
]]);
171 pages_to_write_count
++;
172 page_to_write
[j
] = 1;
173 for(i
= 0; i
< raid_disks
; i
++) {
174 blocks_page
[i
] = blocks
[i
] + j
* CHECK_PAGE_SIZE
;
176 if (disk
[j
] == diskQ
) {
177 qsyndrome(p
, (uint8_t*)stripes
[diskQ
] + j
* CHECK_PAGE_SIZE
, (uint8_t**)blocks_page
, data_disks
, CHECK_PAGE_SIZE
);
180 char *all_but_failed_blocks
[data_disks
];
181 int failed_block_index
= block_index_for_slot
[disk
[j
]];
182 for(i
= 0; i
< data_disks
; i
++) {
183 if (failed_block_index
== i
) {
184 all_but_failed_blocks
[i
] = stripes
[diskP
] + j
* CHECK_PAGE_SIZE
;
187 all_but_failed_blocks
[i
] = blocks_page
[i
];
190 xor_blocks(stripes
[disk
[j
]] + j
* CHECK_PAGE_SIZE
,
191 all_but_failed_blocks
, data_disks
, CHECK_PAGE_SIZE
);
195 page_to_write
[j
] = 0;
199 if(pages_to_write_count
> 0) {
201 for(j
= 0; j
< (chunk_size
>> CHECK_PAGE_BITS
); j
++) {
202 if(page_to_write
[j
] == 1) {
203 lseek64(source
[disk
[j
]], offsets
[disk
[j
]] + start
* chunk_size
+ j
* CHECK_PAGE_SIZE
, SEEK_SET
);
204 write_res
+= write(source
[disk
[j
]], stripes
[disk
[j
]] + j
* CHECK_PAGE_SIZE
, CHECK_PAGE_SIZE
);
208 if (write_res
!= (CHECK_PAGE_SIZE
* pages_to_write_count
)) {
209 fprintf(stderr
, "Failed to write a full chunk.\n");
218 int manual_repair(int diskP
, int diskQ
, int chunk_size
, int raid_disks
, int data_disks
,
219 int failed_disk1
, int failed_disk2
, unsigned long long start
, int *block_index_for_slot
,
220 char *name
[], char **stripes
, char **blocks
, uint8_t *p
, struct mdinfo
*info
, sighandler_t
*sig
,
221 int *source
, unsigned long long *offsets
)
225 printf("Repairing stripe %llu\n", start
);
226 printf("Assuming slots %d (%s) and %d (%s) are incorrect\n",
227 failed_disk1
, name
[failed_disk1
],
228 failed_disk2
, name
[failed_disk2
]);
230 if (failed_disk1
== diskQ
|| failed_disk2
== diskQ
) {
231 char *all_but_failed_blocks
[data_disks
];
232 int failed_data_or_p
;
233 int failed_block_index
;
235 if (failed_disk1
== diskQ
) {
236 failed_data_or_p
= failed_disk2
;
239 failed_data_or_p
= failed_disk1
;
241 printf("Repairing D/P(%d) and Q\n", failed_data_or_p
);
242 failed_block_index
= block_index_for_slot
[failed_data_or_p
];
243 for (i
= 0; i
< data_disks
; i
++) {
244 if (failed_block_index
== i
) {
245 all_but_failed_blocks
[i
] = stripes
[diskP
];
248 all_but_failed_blocks
[i
] = blocks
[i
];
251 xor_blocks(stripes
[failed_data_or_p
],
252 all_but_failed_blocks
, data_disks
, chunk_size
);
253 qsyndrome(p
, (uint8_t*)stripes
[diskQ
], (uint8_t**)blocks
, data_disks
, chunk_size
);
256 ensure_zero_has_size(chunk_size
);
257 if (failed_disk1
== diskP
|| failed_disk2
== diskP
) {
258 int failed_data
, failed_block_index
;
259 if (failed_disk1
== diskP
) {
260 failed_data
= failed_disk2
;
263 failed_data
= failed_disk1
;
265 failed_block_index
= block_index_for_slot
[failed_data
];
266 printf("Repairing D(%d) and P\n", failed_data
);
267 raid6_datap_recov(raid_disks
, chunk_size
, failed_block_index
, (uint8_t**)blocks
);
270 printf("Repairing D and D\n");
271 int failed_block_index1
= block_index_for_slot
[failed_disk1
];
272 int failed_block_index2
= block_index_for_slot
[failed_disk2
];
273 if (failed_block_index1
> failed_block_index2
) {
274 int t
= failed_block_index1
;
275 failed_block_index1
= failed_block_index2
;
276 failed_block_index2
= t
;
278 raid6_2data_recov(raid_disks
, chunk_size
, failed_block_index1
, failed_block_index2
, (uint8_t**)blocks
);
282 err
= lock_stripe(info
, start
, chunk_size
, data_disks
, sig
);
290 int write_res1
, write_res2
;
293 seek_res
= lseek64(source
[failed_disk1
],
294 offsets
[failed_disk1
] + start
* chunk_size
, SEEK_SET
);
296 fprintf(stderr
, "lseek failed for failed_disk1\n");
299 write_res1
= write(source
[failed_disk1
], stripes
[failed_disk1
], chunk_size
);
301 seek_res
= lseek64(source
[failed_disk2
],
302 offsets
[failed_disk2
] + start
* chunk_size
, SEEK_SET
);
304 fprintf(stderr
, "lseek failed for failed_disk1\n");
307 write_res2
= write(source
[failed_disk2
], stripes
[failed_disk2
], chunk_size
);
309 err
= unlock_all_stripes(info
, sig
);
314 if (write_res1
!= chunk_size
|| write_res2
!= chunk_size
) {
315 fprintf(stderr
, "Failed to write a complete chunk.\n");
322 int check_stripes(struct mdinfo
*info
, int *source
, unsigned long long *offsets
,
323 int raid_disks
, int chunk_size
, int level
, int layout
,
324 unsigned long long start
, unsigned long long length
, char *name
[],
325 enum repair repair
, int failed_disk1
, int failed_disk2
)
327 /* read the data and p and q blocks, and check we got them right */
328 char *stripe_buf
= xmalloc(raid_disks
* chunk_size
);
329 char **stripes
= xmalloc(raid_disks
* sizeof(char*));
330 char **blocks
= xmalloc(raid_disks
* sizeof(char*));
331 char **blocks_page
= xmalloc(raid_disks
* sizeof(char*));
332 int *block_index_for_slot
= xmalloc(raid_disks
* sizeof(int));
333 uint8_t *p
= xmalloc(chunk_size
);
334 uint8_t *q
= xmalloc(chunk_size
);
335 int *results
= xmalloc(chunk_size
* sizeof(int));
336 sighandler_t
*sig
= xmalloc(3 * sizeof(sighandler_t
));
340 int data_disks
= raid_disks
- 2;
343 extern int tables_ready
;
348 for ( i
= 0 ; i
< raid_disks
; i
++)
349 stripes
[i
] = stripe_buf
+ i
* chunk_size
;
352 int disk
[chunk_size
>> CHECK_PAGE_BITS
];
354 err
= lock_stripe(info
, start
, chunk_size
, data_disks
, sig
);
357 unlock_all_stripes(info
, sig
);
360 for (i
= 0 ; i
< raid_disks
; i
++) {
361 off64_t seek_res
= lseek64(source
[i
], offsets
[i
] + start
* chunk_size
,
364 fprintf(stderr
, "lseek to source %d failed\n", i
);
365 unlock_all_stripes(info
, sig
);
369 int read_res
= read(source
[i
], stripes
[i
], chunk_size
);
370 if (read_res
< chunk_size
) {
371 fprintf(stderr
, "Failed to read complete chunk disk %d, aborting\n", i
);
372 unlock_all_stripes(info
, sig
);
378 for (i
= 0 ; i
< data_disks
; i
++) {
379 int disk
= geo_map(i
, start
, raid_disks
, level
, layout
);
380 blocks
[i
] = stripes
[disk
];
381 block_index_for_slot
[disk
] = i
;
384 qsyndrome(p
, q
, (uint8_t**)blocks
, data_disks
, chunk_size
);
385 diskP
= geo_map(-1, start
, raid_disks
, level
, layout
);
386 diskQ
= geo_map(-2, start
, raid_disks
, level
, layout
);
387 blocks
[data_disks
] = stripes
[diskP
];
388 block_index_for_slot
[diskP
] = data_disks
;
389 blocks
[data_disks
+1] = stripes
[diskQ
];
390 block_index_for_slot
[diskQ
] = data_disks
+1;
392 raid6_collect(chunk_size
, p
, q
, stripes
[diskP
], stripes
[diskQ
], results
);
393 raid6_stats(disk
, results
, raid_disks
, chunk_size
);
395 for(j
= 0; j
< (chunk_size
>> CHECK_PAGE_BITS
); j
++) {
397 disk
[j
] = geo_map(disk
[j
], start
, raid_disks
, level
, layout
);
400 printf("Error detected at stripe %llu, page %d: possible failed disk slot: %d --> %s\n",
401 start
, j
, disk
[j
], name
[disk
[j
]]);
403 if(disk
[j
] == -65535) {
404 printf("Error detected at stripe %llu, page %d: disk slot unknown\n", start
, j
);
408 if(repair
== AUTO_REPAIR
) {
409 err
= autorepair(disk
, diskP
, diskQ
, start
, chunk_size
,
410 name
, raid_disks
, data_disks
, blocks_page
,
411 blocks
, p
, stripes
, block_index_for_slot
,
414 unlock_all_stripes(info
, sig
);
419 err
= unlock_all_stripes(info
, sig
);
424 if(repair
== MANUAL_REPAIR
) {
425 err
= manual_repair(diskP
, diskQ
, chunk_size
, raid_disks
, data_disks
,
426 failed_disk1
, failed_disk2
, start
, block_index_for_slot
,
427 name
, stripes
, blocks
, p
, info
, sig
,
430 unlock_all_stripes(info
, sig
);
445 free(block_index_for_slot
);
454 unsigned long long getnum(char *str
, char **err
)
457 unsigned long long rv
= strtoull(str
, &e
, 10);
465 int main(int argc
, char *argv
[])
467 /* md_device start length */
470 char **disk_name
= NULL
;
471 unsigned long long *offsets
= NULL
;
477 enum repair repair
= NO_REPAIR
;
478 int failed_disk1
= -1;
479 int failed_disk2
= -1;
480 unsigned long long start
, length
;
483 struct mdinfo
*info
= NULL
, *comp
= NULL
;
487 char *prg
= strrchr(argv
[0], '/');
495 fprintf(stderr
, "Usage: %s md_device start_stripe length_stripes [autorepair]\n", prg
);
496 fprintf(stderr
, " or: %s md_device repair stripe failed_slot_1 failed_slot_2\n", prg
);
501 mdfd
= open(argv
[1], O_RDONLY
);
504 fprintf(stderr
, "%s: cannot open %s\n", prg
, argv
[1]);
509 info
= sysfs_read(mdfd
, NULL
,
521 fprintf(stderr
, "%s: Error reading sysfs information of %s\n", prg
, argv
[1]);
526 if(info
->array
.level
!= level
) {
527 fprintf(stderr
, "%s: %s not a RAID-6\n", prg
, argv
[1]);
532 if(info
->array
.failed_disks
> 0) {
533 fprintf(stderr
, "%s: %s degraded array\n", prg
, argv
[1]);
538 printf("layout: %d\n", info
->array
.layout
);
539 printf("disks: %d\n", info
->array
.raid_disks
);
540 printf("component size: %llu\n", info
->component_size
* 512);
541 printf("total stripes: %llu\n", (info
->component_size
* 512) / info
->array
.chunk_size
);
542 printf("chunk size: %d\n", info
->array
.chunk_size
);
546 for(i
= 0, active_disks
= 0; active_disks
< info
->array
.raid_disks
; i
++) {
547 printf("disk: %d - offset: %llu - size: %llu - name: %s - slot: %d\n",
548 i
, comp
->data_offset
* 512, comp
->component_size
* 512,
549 map_dev(comp
->disk
.major
, comp
->disk
.minor
, 0),
550 comp
->disk
.raid_disk
);
551 if(comp
->disk
.raid_disk
>= 0)
559 raid_disks
= info
->array
.raid_disks
;
560 chunk_size
= info
->array
.chunk_size
;
561 layout
= info
->array
.layout
;
562 if (strcmp(argv
[2], "repair")==0) {
564 fprintf(stderr
, "For repair mode, call %s md_device repair stripe failed_slot_1 failed_slot_2\n", prg
);
568 repair
= MANUAL_REPAIR
;
569 start
= getnum(argv
[3], &err
);
571 failed_disk1
= getnum(argv
[4], &err
);
572 failed_disk2
= getnum(argv
[5], &err
);
574 if(failed_disk1
>= info
->array
.raid_disks
) {
575 fprintf(stderr
, "%s: failed_slot_1 index is higher than number of devices in raid\n", prg
);
579 if(failed_disk2
>= info
->array
.raid_disks
) {
580 fprintf(stderr
, "%s: failed_slot_2 index is higher than number of devices in raid\n", prg
);
584 if(failed_disk1
== failed_disk2
) {
585 fprintf(stderr
, "%s: failed_slot_1 and failed_slot_2 are the same\n", prg
);
591 start
= getnum(argv
[2], &err
);
592 length
= getnum(argv
[3], &err
);
593 if (argc
>= 5 && strcmp(argv
[4], "autorepair")==0)
594 repair
= AUTO_REPAIR
;
598 fprintf(stderr
, "%s: Bad number: %s\n", prg
, err
);
603 if(start
> ((info
->component_size
* 512) / chunk_size
)) {
604 start
= (info
->component_size
* 512) / chunk_size
;
605 fprintf(stderr
, "%s: start beyond disks size\n", prg
);
609 ((length
+ start
) > ((info
->component_size
* 512) / chunk_size
))) {
610 length
= (info
->component_size
* 512) / chunk_size
- start
;
613 disk_name
= xmalloc(raid_disks
* sizeof(*disk_name
));
614 fds
= xmalloc(raid_disks
* sizeof(*fds
));
615 offsets
= xcalloc(raid_disks
, sizeof(*offsets
));
616 buf
= xmalloc(raid_disks
* chunk_size
);
618 for(i
=0; i
<raid_disks
; i
++) {
624 for (i
=0, active_disks
=0; active_disks
<raid_disks
; i
++) {
625 int disk_slot
= comp
->disk
.raid_disk
;
627 disk_name
[disk_slot
] = map_dev(comp
->disk
.major
, comp
->disk
.minor
, 0);
628 offsets
[disk_slot
] = comp
->data_offset
* 512;
629 fds
[disk_slot
] = open(disk_name
[disk_slot
], O_RDWR
| O_SYNC
);
630 if (fds
[disk_slot
] < 0) {
631 perror(disk_name
[disk_slot
]);
632 fprintf(stderr
,"%s: cannot open %s\n", prg
, disk_name
[disk_slot
]);
641 int rv
= check_stripes(info
, fds
, offsets
,
642 raid_disks
, chunk_size
, level
, layout
,
643 start
, length
, disk_name
, repair
, failed_disk1
, failed_disk2
);
646 "%s: check_stripes returned %d\n", prg
, rv
);
654 for(i
= 0; i
< raid_disks
; i
++)