]>
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
36 int geo_map(int block
, unsigned long long stripe
, int raid_disks
,
37 int level
, int layout
);
38 void qsyndrome(uint8_t *p
, uint8_t *q
, uint8_t **sources
, int disks
, int size
);
39 void make_tables(void);
40 void ensure_zero_has_size(int chunk_size
);
41 void raid6_datap_recov(int disks
, size_t bytes
, int faila
, uint8_t **ptrs
);
42 void raid6_2data_recov(int disks
, size_t bytes
, int faila
, int failb
,
44 void xor_blocks(char *target
, char **sources
, int disks
, int size
);
46 /* Collect per stripe consistency information */
47 void raid6_collect(int chunk_size
, uint8_t *p
, uint8_t *q
,
48 char *chunkP
, char *chunkQ
, int *results
)
53 extern uint8_t raid6_gflog
[];
55 for(i
= 0; i
< chunk_size
; i
++) {
56 Px
= (uint8_t)chunkP
[i
] ^ (uint8_t)p
[i
];
57 Qx
= (uint8_t)chunkQ
[i
] ^ (uint8_t)q
[i
];
59 if((Px
!= 0) && (Qx
== 0))
62 if((Px
== 0) && (Qx
!= 0))
65 if((Px
!= 0) && (Qx
!= 0)) {
66 data_id
= (raid6_gflog
[Qx
] - raid6_gflog
[Px
]);
67 if(data_id
< 0) data_id
+= 255;
71 if((Px
== 0) && (Qx
== 0))
76 /* Try to find out if a specific disk has problems */
77 int raid6_stats(int *results
, int raid_disks
, int chunk_size
)
80 int curr_broken_disk
= -255;
81 int prev_broken_disk
= -255;
82 int broken_status
= 0;
84 for(i
= 0; i
< chunk_size
; i
++) {
86 if(results
[i
] != -255)
87 curr_broken_disk
= results
[i
];
89 if(curr_broken_disk
>= raid_disks
)
92 switch(broken_status
) {
94 if(curr_broken_disk
!= -255) {
95 prev_broken_disk
= curr_broken_disk
;
101 if(curr_broken_disk
!= prev_broken_disk
)
107 curr_broken_disk
= prev_broken_disk
= -65535;
112 return curr_broken_disk
;
115 int lock_stripe(struct mdinfo
*info
, unsigned long long start
,
116 int chunk_size
, int data_disks
, sighandler_t
*sig
) {
118 if(mlockall(MCL_CURRENT
| MCL_FUTURE
) != 0) {
122 sig
[0] = signal(SIGTERM
, SIG_IGN
);
123 sig
[1] = signal(SIGINT
, SIG_IGN
);
124 sig
[2] = signal(SIGQUIT
, SIG_IGN
);
126 rv
= sysfs_set_num(info
, NULL
, "suspend_lo", start
* chunk_size
* data_disks
);
127 rv
|= sysfs_set_num(info
, NULL
, "suspend_hi", (start
+ 1) * chunk_size
* data_disks
);
131 int unlock_all_stripes(struct mdinfo
*info
, sighandler_t
*sig
) {
133 rv
= sysfs_set_num(info
, NULL
, "suspend_lo", 0x7FFFFFFFFFFFFFFFULL
);
134 rv
|= sysfs_set_num(info
, NULL
, "suspend_hi", 0);
135 rv
|= sysfs_set_num(info
, NULL
, "suspend_lo", 0);
137 signal(SIGQUIT
, sig
[2]);
138 signal(SIGINT
, sig
[1]);
139 signal(SIGTERM
, sig
[0]);
141 if(munlockall() != 0)
146 int check_stripes(struct mdinfo
*info
, int *source
, unsigned long long *offsets
,
147 int raid_disks
, int chunk_size
, int level
, int layout
,
148 unsigned long long start
, unsigned long long length
, char *name
[],
149 enum repair repair
, int failed_disk1
, int failed_disk2
)
151 /* read the data and p and q blocks, and check we got them right */
152 char *stripe_buf
= xmalloc(raid_disks
* chunk_size
);
153 char **stripes
= xmalloc(raid_disks
* sizeof(char*));
154 char **blocks
= xmalloc(raid_disks
* sizeof(char*));
155 int *block_index_for_slot
= xmalloc(raid_disks
* sizeof(int));
156 uint8_t *p
= xmalloc(chunk_size
);
157 uint8_t *q
= xmalloc(chunk_size
);
158 int *results
= xmalloc(chunk_size
* sizeof(int));
159 sighandler_t
*sig
= xmalloc(3 * sizeof(sighandler_t
));
163 int data_disks
= raid_disks
- 2;
166 extern int tables_ready
;
171 for ( i
= 0 ; i
< raid_disks
; i
++)
172 stripes
[i
] = stripe_buf
+ i
* chunk_size
;
177 printf("pos --> %llu\n", start
);
179 err
= lock_stripe(info
, start
, chunk_size
, data_disks
, sig
);
182 unlock_all_stripes(info
, sig
);
185 for (i
= 0 ; i
< raid_disks
; i
++) {
186 off64_t seek_res
= lseek64(source
[i
], offsets
[i
] + start
* chunk_size
,
189 fprintf(stderr
, "lseek to source %d failed\n", i
);
190 unlock_all_stripes(info
, sig
);
194 int read_res
= read(source
[i
], stripes
[i
], chunk_size
);
195 if (read_res
< chunk_size
) {
196 fprintf(stderr
, "Failed to read complete chunk disk %d, aborting\n", i
);
197 unlock_all_stripes(info
, sig
);
202 err
= unlock_all_stripes(info
, sig
);
206 for (i
= 0 ; i
< data_disks
; i
++) {
207 int disk
= geo_map(i
, start
, raid_disks
, level
, layout
);
208 blocks
[i
] = stripes
[disk
];
209 block_index_for_slot
[disk
] = i
;
210 printf("%d->%d\n", i
, disk
);
213 qsyndrome(p
, q
, (uint8_t**)blocks
, data_disks
, chunk_size
);
214 diskP
= geo_map(-1, start
, raid_disks
, level
, layout
);
215 diskQ
= geo_map(-2, start
, raid_disks
, level
, layout
);
216 blocks
[data_disks
] = stripes
[diskP
];
217 block_index_for_slot
[diskP
] = data_disks
;
218 blocks
[data_disks
+1] = stripes
[diskQ
];
219 block_index_for_slot
[diskQ
] = data_disks
+1;
221 if (memcmp(p
, stripes
[diskP
], chunk_size
) != 0) {
222 printf("P(%d) wrong at %llu\n", diskP
, start
);
224 if (memcmp(q
, stripes
[diskQ
], chunk_size
) != 0) {
225 printf("Q(%d) wrong at %llu\n", diskQ
, start
);
227 raid6_collect(chunk_size
, p
, q
, stripes
[diskP
], stripes
[diskQ
], results
);
228 disk
= raid6_stats(results
, raid_disks
, chunk_size
);
231 disk
= geo_map(disk
, start
, raid_disks
, level
, layout
);
234 printf("Error detected at %llu: possible failed disk slot: %d --> %s\n",
235 start
, disk
, name
[disk
]);
238 printf("Error detected at %llu: disk slot unknown\n", start
);
240 if(repair
== MANUAL_REPAIR
) {
241 printf("Repairing stripe %llu\n", start
);
242 printf("Assuming slots %d (%s) and %d (%s) are incorrect\n",
243 failed_disk1
, name
[failed_disk1
],
244 failed_disk2
, name
[failed_disk2
]);
246 if (failed_disk1
== diskQ
|| failed_disk2
== diskQ
) {
247 char *all_but_failed_blocks
[data_disks
];
248 int failed_data_or_p
;
249 int failed_block_index
;
251 if (failed_disk1
== diskQ
)
252 failed_data_or_p
= failed_disk2
;
254 failed_data_or_p
= failed_disk1
;
255 printf("Repairing D/P(%d) and Q\n", failed_data_or_p
);
256 failed_block_index
= block_index_for_slot
[failed_data_or_p
];
257 for (i
=0; i
< data_disks
; i
++)
258 if (failed_block_index
== i
)
259 all_but_failed_blocks
[i
] = stripes
[diskP
];
261 all_but_failed_blocks
[i
] = blocks
[i
];
262 xor_blocks(stripes
[failed_data_or_p
],
263 all_but_failed_blocks
, data_disks
, chunk_size
);
264 qsyndrome(p
, (uint8_t*)stripes
[diskQ
], (uint8_t**)blocks
, data_disks
, chunk_size
);
266 ensure_zero_has_size(chunk_size
);
267 if (failed_disk1
== diskP
|| failed_disk2
== diskP
) {
268 int failed_data
, failed_block_index
;
269 if (failed_disk1
== diskP
)
270 failed_data
= failed_disk2
;
272 failed_data
= failed_disk1
;
273 failed_block_index
= block_index_for_slot
[failed_data
];
274 printf("Repairing D(%d) and P\n", failed_data
);
275 raid6_datap_recov(raid_disks
, chunk_size
, failed_block_index
, (uint8_t**)blocks
);
277 printf("Repairing D and D\n");
278 int failed_block_index1
= block_index_for_slot
[failed_disk1
];
279 int failed_block_index2
= block_index_for_slot
[failed_disk2
];
280 if (failed_block_index1
> failed_block_index2
) {
281 int t
= failed_block_index1
;
282 failed_block_index1
= failed_block_index2
;
283 failed_block_index2
= t
;
285 raid6_2data_recov(raid_disks
, chunk_size
, failed_block_index1
, failed_block_index2
, (uint8_t**)blocks
);
289 err
= lock_stripe(info
, start
, chunk_size
, data_disks
, sig
);
292 unlock_all_stripes(info
, sig
);
296 int write_res1
, write_res2
;
299 seek_res
= lseek64(source
[failed_disk1
],
300 offsets
[failed_disk1
] + start
* chunk_size
, SEEK_SET
);
302 fprintf(stderr
, "lseek failed for failed_disk1\n");
303 unlock_all_stripes(info
, sig
);
307 write_res1
= write(source
[failed_disk1
], stripes
[failed_disk1
], chunk_size
);
309 seek_res
= lseek64(source
[failed_disk2
],
310 offsets
[failed_disk2
] + start
* chunk_size
, SEEK_SET
);
312 fprintf(stderr
, "lseek failed for failed_disk1\n");
313 unlock_all_stripes(info
, sig
);
317 write_res2
= write(source
[failed_disk2
], stripes
[failed_disk2
], chunk_size
);
319 err
= unlock_all_stripes(info
, sig
);
323 if (write_res1
!= chunk_size
|| write_res2
!= chunk_size
) {
324 fprintf(stderr
, "Failed to write a complete chunk.\n");
328 } else if (disk
>= 0 && repair
== AUTO_REPAIR
) {
329 printf("Auto-repairing slot %d (%s)\n", disk
, name
[disk
]);
331 qsyndrome(p
, (uint8_t*)stripes
[diskQ
], (uint8_t**)blocks
, data_disks
, chunk_size
);
333 char *all_but_failed_blocks
[data_disks
];
334 int failed_block_index
= block_index_for_slot
[disk
];
335 for (i
=0; i
< data_disks
; i
++)
336 if (failed_block_index
== i
)
337 all_but_failed_blocks
[i
] = stripes
[diskP
];
339 all_but_failed_blocks
[i
] = blocks
[i
];
340 xor_blocks(stripes
[disk
],
341 all_but_failed_blocks
, data_disks
, chunk_size
);
344 err
= lock_stripe(info
, start
, chunk_size
, data_disks
, sig
);
347 unlock_all_stripes(info
, sig
);
351 lseek64(source
[disk
], offsets
[disk
] + start
* chunk_size
, 0);
352 int write_res
= write(source
[disk
], stripes
[disk
], chunk_size
);
354 err
= unlock_all_stripes(info
, sig
);
355 if(err
!= 0 || write_res
!= chunk_size
)
358 if (write_res
!= chunk_size
) {
359 fprintf(stderr
, "Failed to write a full chunk.\n");
373 free(block_index_for_slot
);
382 unsigned long long getnum(char *str
, char **err
)
385 unsigned long long rv
= strtoull(str
, &e
, 10);
393 int main(int argc
, char *argv
[])
395 /* md_device start length */
398 char **disk_name
= NULL
;
399 unsigned long long *offsets
= NULL
;
405 enum repair repair
= NO_REPAIR
;
406 int failed_disk1
= -1;
407 int failed_disk2
= -1;
408 unsigned long long start
, length
;
411 struct mdinfo
*info
= NULL
, *comp
= NULL
;
415 char *prg
= strrchr(argv
[0], '/');
423 fprintf(stderr
, "Usage: %s md_device start_stripe length_stripes [autorepair]\n", prg
);
424 fprintf(stderr
, " or: %s md_device repair stripe failed_slot_1 failed_slot_2\n", prg
);
429 mdfd
= open(argv
[1], O_RDONLY
);
432 fprintf(stderr
, "%s: cannot open %s\n", prg
, argv
[1]);
437 info
= sysfs_read(mdfd
, NULL
,
449 fprintf(stderr
, "%s: Error reading sysfs information of %s\n", prg
, argv
[1]);
454 if(info
->array
.level
!= level
) {
455 fprintf(stderr
, "%s: %s not a RAID-6\n", prg
, argv
[1]);
460 if(info
->array
.failed_disks
> 0) {
461 fprintf(stderr
, "%s: %s degraded array\n", prg
, argv
[1]);
466 printf("layout: %d\n", info
->array
.layout
);
467 printf("disks: %d\n", info
->array
.raid_disks
);
468 printf("component size: %llu\n", info
->component_size
* 512);
469 printf("total stripes: %llu\n", (info
->component_size
* 512) / info
->array
.chunk_size
);
470 printf("chunk size: %d\n", info
->array
.chunk_size
);
474 for(i
= 0, active_disks
= 0; active_disks
< info
->array
.raid_disks
; i
++) {
475 printf("disk: %d - offset: %llu - size: %llu - name: %s - slot: %d\n",
476 i
, comp
->data_offset
* 512, comp
->component_size
* 512,
477 map_dev(comp
->disk
.major
, comp
->disk
.minor
, 0),
478 comp
->disk
.raid_disk
);
479 if(comp
->disk
.raid_disk
>= 0)
487 raid_disks
= info
->array
.raid_disks
;
488 chunk_size
= info
->array
.chunk_size
;
489 layout
= info
->array
.layout
;
490 if (strcmp(argv
[2], "repair")==0) {
492 fprintf(stderr
, "For repair mode, call %s md_device repair stripe failed_slot_1 failed_slot_2\n", prg
);
496 repair
= MANUAL_REPAIR
;
497 start
= getnum(argv
[3], &err
);
499 failed_disk1
= getnum(argv
[4], &err
);
500 failed_disk2
= getnum(argv
[5], &err
);
502 if(failed_disk1
>= info
->array
.raid_disks
) {
503 fprintf(stderr
, "%s: failed_slot_1 index is higher than number of devices in raid\n", prg
);
507 if(failed_disk2
>= info
->array
.raid_disks
) {
508 fprintf(stderr
, "%s: failed_slot_2 index is higher than number of devices in raid\n", prg
);
512 if(failed_disk1
== failed_disk2
) {
513 fprintf(stderr
, "%s: failed_slot_1 and failed_slot_2 are the same\n", prg
);
519 start
= getnum(argv
[2], &err
);
520 length
= getnum(argv
[3], &err
);
521 if (argc
>= 5 && strcmp(argv
[4], "autorepair")==0)
522 repair
= AUTO_REPAIR
;
526 fprintf(stderr
, "%s: Bad number: %s\n", prg
, err
);
531 if(start
> ((info
->component_size
* 512) / chunk_size
)) {
532 start
= (info
->component_size
* 512) / chunk_size
;
533 fprintf(stderr
, "%s: start beyond disks size\n", prg
);
537 ((length
+ start
) > ((info
->component_size
* 512) / chunk_size
))) {
538 length
= (info
->component_size
* 512) / chunk_size
- start
;
541 disk_name
= xmalloc(raid_disks
* sizeof(*disk_name
));
542 fds
= xmalloc(raid_disks
* sizeof(*fds
));
543 offsets
= xcalloc(raid_disks
, sizeof(*offsets
));
544 buf
= xmalloc(raid_disks
* chunk_size
);
546 for(i
=0; i
<raid_disks
; i
++) {
552 for (i
=0, active_disks
=0; active_disks
<raid_disks
; i
++) {
553 int disk_slot
= comp
->disk
.raid_disk
;
555 disk_name
[disk_slot
] = map_dev(comp
->disk
.major
, comp
->disk
.minor
, 0);
556 offsets
[disk_slot
] = comp
->data_offset
* 512;
557 fds
[disk_slot
] = open(disk_name
[disk_slot
], O_RDWR
);
558 if (fds
[disk_slot
] < 0) {
559 perror(disk_name
[disk_slot
]);
560 fprintf(stderr
,"%s: cannot open %s\n", prg
, disk_name
[disk_slot
]);
569 int rv
= check_stripes(info
, fds
, offsets
,
570 raid_disks
, chunk_size
, level
, layout
,
571 start
, length
, disk_name
, repair
, failed_disk1
, failed_disk2
);
574 "%s: check_stripes returned %d\n", prg
, rv
);
582 for(i
= 0; i
< raid_disks
; i
++)