]> git.ipfire.org Git - thirdparty/mdadm.git/blame - raid6check.c
Create: add error checking for 'write_init_super'.
[thirdparty/mdadm.git] / raid6check.c
CommitLineData
979afcb8
PS
1/*
2 * raid6check - extended consistency check for RAID-6
3 *
4 * Copyright (C) 2011 Piergiorgio Sartor
5 *
6 *
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.
11 *
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.
16 *
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
20 *
21 * Author: Piergiorgio Sartor
22 * Based on "restripe.c" from "mdadm" codebase
23 */
24
25#include "mdadm.h"
26#include <stdint.h>
27
28int geo_map(int block, unsigned long long stripe, int raid_disks,
29 int level, int layout);
30void qsyndrome(uint8_t *p, uint8_t *q, uint8_t **sources, int disks, int size);
31void make_tables(void);
32
33/* Collect per stripe consistency information */
34void raid6_collect(int chunk_size, uint8_t *p, uint8_t *q,
35 char *chunkP, char *chunkQ, int *results)
36{
37 int i;
38 int data_id;
39 uint8_t Px, Qx;
40 extern uint8_t raid6_gflog[];
41
42 for(i = 0; i < chunk_size; i++) {
43 Px = (uint8_t)chunkP[i] ^ (uint8_t)p[i];
44 Qx = (uint8_t)chunkQ[i] ^ (uint8_t)q[i];
45
46 if((Px != 0) && (Qx == 0))
47 results[i] = -1;
48
49 if((Px == 0) && (Qx != 0))
50 results[i] = -2;
51
52 if((Px != 0) && (Qx != 0)) {
53 data_id = (raid6_gflog[Qx] - raid6_gflog[Px]);
54 if(data_id < 0) data_id += 255;
55 results[i] = data_id;
56 }
57
58 if((Px == 0) && (Qx == 0))
59 results[i] = -255;
60 }
61}
62
63/* Try to find out if a specific disk has problems */
64int raid6_stats(int *results, int raid_disks, int chunk_size)
65{
66 int i;
67 int curr_broken_disk = -255;
68 int prev_broken_disk = -255;
69 int broken_status = 0;
70
71 for(i = 0; i < chunk_size; i++) {
72
73 if(results[i] != -255)
74 curr_broken_disk = results[i];
75
76 if(curr_broken_disk >= raid_disks)
77 broken_status = 2;
78
79 switch(broken_status) {
80 case 0:
81 if(curr_broken_disk != -255) {
82 prev_broken_disk = curr_broken_disk;
83 broken_status = 1;
84 }
85 break;
86
87 case 1:
88 if(curr_broken_disk != prev_broken_disk)
89 broken_status = 2;
90 break;
91
92 case 2:
93 default:
94 curr_broken_disk = prev_broken_disk = -65535;
95 break;
96 }
97 }
98
99 return curr_broken_disk;
100}
101
102int check_stripes(int *source, unsigned long long *offsets,
103 int raid_disks, int chunk_size, int level, int layout,
104 unsigned long long start, unsigned long long length, char *name[])
105{
106 /* read the data and p and q blocks, and check we got them right */
107 char *stripe_buf = malloc(raid_disks * chunk_size);
108 char **stripes = malloc(raid_disks * sizeof(char*));
109 char **blocks = malloc(raid_disks * sizeof(char*));
110 uint8_t *p = malloc(chunk_size);
111 uint8_t *q = malloc(chunk_size);
112 int *results = malloc(chunk_size * sizeof(int));
113
114 int i;
115 int diskP, diskQ;
116 int data_disks = raid_disks - 2;
af3c3750 117 int err = 0;
979afcb8
PS
118
119 extern int tables_ready;
120
af3c3750
PS
121 if((stripe_buf == NULL) ||
122 (stripes == NULL) ||
123 (blocks == NULL) ||
124 (p == NULL) ||
125 (q == NULL) ||
126 (results == NULL)) {
127 err = 1;
128 goto exitCheck;
129 }
130
979afcb8
PS
131 if (!tables_ready)
132 make_tables();
133
134 for ( i = 0 ; i < raid_disks ; i++)
135 stripes[i] = stripe_buf + i * chunk_size;
136
137 while (length > 0) {
138 int disk;
139
af3c3750
PS
140 printf("pos --> %llu\n", start);
141
979afcb8 142 for (i = 0 ; i < raid_disks ; i++) {
af3c3750 143 lseek64(source[i], offsets[i] + start * chunk_size, 0);
979afcb8
PS
144 read(source[i], stripes[i], chunk_size);
145 }
146 for (i = 0 ; i < data_disks ; i++) {
af3c3750 147 int disk = geo_map(i, start, raid_disks, level, layout);
979afcb8
PS
148 blocks[i] = stripes[disk];
149 printf("%d->%d\n", i, disk);
150 }
151
152 qsyndrome(p, q, (uint8_t**)blocks, data_disks, chunk_size);
af3c3750 153 diskP = geo_map(-1, start, raid_disks, level, layout);
979afcb8 154 if (memcmp(p, stripes[diskP], chunk_size) != 0) {
af3c3750 155 printf("P(%d) wrong at %llu\n", diskP, start);
979afcb8 156 }
af3c3750 157 diskQ = geo_map(-2, start, raid_disks, level, layout);
979afcb8 158 if (memcmp(q, stripes[diskQ], chunk_size) != 0) {
af3c3750 159 printf("Q(%d) wrong at %llu\n", diskQ, start);
979afcb8 160 }
af3c3750 161 raid6_collect(chunk_size, p, q, stripes[diskP], stripes[diskQ], results);
979afcb8
PS
162 disk = raid6_stats(results, raid_disks, chunk_size);
163
164 if(disk >= -2) {
af3c3750 165 disk = geo_map(disk, start, raid_disks, level, layout);
979afcb8
PS
166 }
167 if(disk >= 0) {
af3c3750
PS
168 printf("Error detected at %llu: possible failed disk slot: %d --> %s\n",
169 start, disk, name[disk]);
979afcb8
PS
170 }
171 if(disk == -65535) {
af3c3750 172 printf("Error detected at %llu: disk slot unknown\n", start);
979afcb8
PS
173 }
174
af3c3750
PS
175 length--;
176 start++;
979afcb8
PS
177 }
178
af3c3750
PS
179exitCheck:
180
979afcb8
PS
181 free(stripe_buf);
182 free(stripes);
183 free(blocks);
184 free(p);
185 free(q);
186 free(results);
187
af3c3750 188 return err;
979afcb8
PS
189}
190
191unsigned long long getnum(char *str, char **err)
192{
193 char *e;
194 unsigned long long rv = strtoull(str, &e, 10);
195 if (e==str || *e) {
196 *err = str;
197 return 0;
198 }
199 return rv;
200}
201
202int main(int argc, char *argv[])
203{
a9c2c6c6 204 /* md_device start length */
af3c3750
PS
205 int *fds = NULL;
206 char *buf = NULL;
207 char **disk_name = NULL;
208 unsigned long long *offsets = NULL;
209 int raid_disks = 0;
2cf31121 210 int active_disks;
af3c3750
PS
211 int chunk_size = 0;
212 int layout = -1;
979afcb8
PS
213 int level = 6;
214 unsigned long long start, length;
215 int i;
a9c2c6c6
PS
216 int mdfd;
217 struct mdinfo *info, *comp;
979afcb8 218 char *err = NULL;
af3c3750
PS
219 int exit_err = 0;
220 int close_flag = 0;
221 char *prg = strrchr(argv[0], '/');
222
223 if (prg == NULL)
224 prg = argv[0];
225 else
226 prg++;
227
228 if (argc < 4) {
229 fprintf(stderr, "Usage: %s md_device start_stripe length_stripes\n", prg);
230 exit_err = 1;
231 goto exitHere;
979afcb8
PS
232 }
233
a9c2c6c6
PS
234 mdfd = open(argv[1], O_RDONLY);
235 if(mdfd < 0) {
236 perror(argv[1]);
237 fprintf(stderr,"%s: cannot open %s\n", prg, argv[1]);
af3c3750
PS
238 exit_err = 2;
239 goto exitHere;
a9c2c6c6
PS
240 }
241
242 info = sysfs_read(mdfd, -1,
243 GET_LEVEL|
244 GET_LAYOUT|
245 GET_DISKS|
2cf31121 246 GET_DEGRADED |
a9c2c6c6
PS
247 GET_COMPONENT|
248 GET_CHUNK|
249 GET_DEVS|
250 GET_OFFSET|
251 GET_SIZE);
252
253 if(info->array.level != level) {
254 fprintf(stderr, "%s: %s not a RAID-6\n", prg, argv[1]);
af3c3750
PS
255 exit_err = 3;
256 goto exitHere;
a9c2c6c6
PS
257 }
258
2cf31121
PS
259 if(info->array.failed_disks > 0) {
260 fprintf(stderr, "%s: %s degraded array\n", prg, argv[1]);
261 exit_err = 8;
262 goto exitHere;
263 }
264
a9c2c6c6
PS
265 printf("layout: %d\n", info->array.layout);
266 printf("disks: %d\n", info->array.raid_disks);
af3c3750
PS
267 printf("component size: %llu\n", info->component_size * 512);
268 printf("total stripes: %llu\n", (info->component_size * 512) / info->array.chunk_size);
a9c2c6c6
PS
269 printf("chunk size: %d\n", info->array.chunk_size);
270 printf("\n");
271
272 comp = info->devs;
2cf31121 273 for(i = 0, active_disks = 0; active_disks < info->array.raid_disks; i++) {
a9c2c6c6 274 printf("disk: %d - offset: %llu - size: %llu - name: %s - slot: %d\n",
af3c3750 275 i, comp->data_offset * 512, comp->component_size * 512,
a9c2c6c6
PS
276 map_dev(comp->disk.major, comp->disk.minor, 0),
277 comp->disk.raid_disk);
2cf31121
PS
278 if(comp->disk.raid_disk >= 0)
279 active_disks++;
a9c2c6c6
PS
280 comp = comp->next;
281 }
282 printf("\n");
283
284 close(mdfd);
285
286 raid_disks = info->array.raid_disks;
287 chunk_size = info->array.chunk_size;
288 layout = info->array.layout;
289 start = getnum(argv[2], &err);
290 length = getnum(argv[3], &err);
291
979afcb8 292 if (err) {
a9c2c6c6 293 fprintf(stderr, "%s: Bad number: %s\n", prg, err);
af3c3750
PS
294 exit_err = 4;
295 goto exitHere;
979afcb8 296 }
a9c2c6c6 297
af3c3750
PS
298 if(start > ((info->component_size * 512) / chunk_size)) {
299 start = (info->component_size * 512) / chunk_size;
300 fprintf(stderr, "%s: start beyond disks size\n", prg);
301 }
a9c2c6c6 302
af3c3750
PS
303 if((length == 0) ||
304 ((length + start) > ((info->component_size * 512) / chunk_size))) {
305 length = (info->component_size * 512) / chunk_size - start;
979afcb8 306 }
a9c2c6c6
PS
307
308 disk_name = malloc(raid_disks * sizeof(*disk_name));
979afcb8
PS
309 fds = malloc(raid_disks * sizeof(*fds));
310 offsets = malloc(raid_disks * sizeof(*offsets));
af3c3750
PS
311 buf = malloc(raid_disks * chunk_size);
312
313 if((disk_name == NULL) ||
314 (fds == NULL) ||
315 (offsets == NULL) ||
316 (buf == NULL)) {
317 fprintf(stderr, "%s: allocation fail\n", prg);
318 exit_err = 5;
319 goto exitHere;
320 }
321
979afcb8 322 memset(offsets, 0, raid_disks * sizeof(*offsets));
af3c3750
PS
323 for(i=0; i<raid_disks; i++) {
324 fds[i] = -1;
325 }
326 close_flag = 1;
979afcb8 327
a9c2c6c6 328 comp = info->devs;
2cf31121 329 for (i=0, active_disks=0; active_disks<raid_disks; i++) {
a9c2c6c6 330 int disk_slot = comp->disk.raid_disk;
2cf31121
PS
331 if(disk_slot >= 0) {
332 disk_name[disk_slot] = map_dev(comp->disk.major, comp->disk.minor, 0);
333 offsets[disk_slot] = comp->data_offset * 512;
334 fds[disk_slot] = open(disk_name[disk_slot], O_RDWR);
335 if (fds[disk_slot] < 0) {
336 perror(disk_name[disk_slot]);
337 fprintf(stderr,"%s: cannot open %s\n", prg, disk_name[disk_slot]);
338 exit_err = 6;
339 goto exitHere;
340 }
341 active_disks++;
979afcb8 342 }
a9c2c6c6 343 comp = comp->next;
979afcb8
PS
344 }
345
979afcb8
PS
346 int rv = check_stripes(fds, offsets,
347 raid_disks, chunk_size, level, layout,
a9c2c6c6 348 start, length, disk_name);
979afcb8
PS
349 if (rv != 0) {
350 fprintf(stderr,
a9c2c6c6 351 "%s: check_stripes returned %d\n", prg, rv);
af3c3750
PS
352 exit_err = 7;
353 goto exitHere;
979afcb8
PS
354 }
355
af3c3750
PS
356exitHere:
357
358 if (close_flag)
359 for(i = 0; i < raid_disks; i++)
360 close(fds[i]);
361
a9c2c6c6 362 free(disk_name);
979afcb8
PS
363 free(fds);
364 free(offsets);
365 free(buf);
366
af3c3750 367 exit(exit_err);
979afcb8 368}