#include <signal.h>
#include <sys/mman.h>
+#define CHECK_PAGE_BITS (12)
+#define CHECK_PAGE_SIZE (1 << CHECK_PAGE_BITS)
+
+char const Name[] = "raid6check";
+
+enum repair {
+ NO_REPAIR = 0,
+ MANUAL_REPAIR,
+ AUTO_REPAIR
+};
+
int geo_map(int block, unsigned long long stripe, int raid_disks,
int level, int layout);
+int is_ddf(int layout);
void qsyndrome(uint8_t *p, uint8_t *q, uint8_t **sources, int disks, int size);
void make_tables(void);
void ensure_zero_has_size(int chunk_size);
-void raid6_datap_recov(int disks, size_t bytes, int faila, uint8_t **ptrs);
+void raid6_datap_recov(int disks, size_t bytes, int faila, uint8_t **ptrs,
+ int neg_offset);
void raid6_2data_recov(int disks, size_t bytes, int faila, int failb,
- uint8_t **ptrs);
+ uint8_t **ptrs, int neg_offset);
void xor_blocks(char *target, char **sources, int disks, int size);
-
/* Collect per stripe consistency information */
void raid6_collect(int chunk_size, uint8_t *p, uint8_t *q,
char *chunkP, char *chunkQ, int *results)
}
}
-/* Try to find out if a specific disk has problems */
-int raid6_stats(int *results, int raid_disks, int chunk_size)
+/* Try to find out if a specific disk has problems in a CHECK_PAGE_SIZE page size */
+int raid6_stats_blk(int *results, int raid_disks)
{
int i;
int curr_broken_disk = -255;
int prev_broken_disk = -255;
int broken_status = 0;
- for(i = 0; i < chunk_size; i++) {
+ for(i = 0; i < CHECK_PAGE_SIZE; i++) {
if(results[i] != -255)
curr_broken_disk = results[i];
return curr_broken_disk;
}
+/* Collect disks status for a strip in CHECK_PAGE_SIZE page size blocks */
+void raid6_stats(int *disk, int *results, int raid_disks, int chunk_size)
+{
+ int i, j;
+
+ for(i = 0, j = 0; i < chunk_size; i += CHECK_PAGE_SIZE, j++) {
+ disk[j] = raid6_stats_blk(&results[i], raid_disks);
+ }
+}
+
int lock_stripe(struct mdinfo *info, unsigned long long start,
int chunk_size, int data_disks, sighandler_t *sig) {
int rv;
return rv * 256;
}
+/* Autorepair */
+int autorepair(int *disk, unsigned long long start, int chunk_size,
+ char *name[], int raid_disks, int syndrome_disks, char **blocks_page,
+ char **blocks, uint8_t *p, int *block_index_for_slot,
+ int *source, unsigned long long *offsets)
+{
+ int i, j;
+ int pages_to_write_count = 0;
+ int page_to_write[chunk_size >> CHECK_PAGE_BITS];
+ for(j = 0; j < (chunk_size >> CHECK_PAGE_BITS); j++) {
+ if (disk[j] >= -2 && block_index_for_slot[disk[j]] >= 0) {
+ int slot = block_index_for_slot[disk[j]];
+ printf("Auto-repairing slot %d (%s)\n", slot, name[slot]);
+ pages_to_write_count++;
+ page_to_write[j] = 1;
+ for(i = -2; i < syndrome_disks; i++) {
+ blocks_page[i] = blocks[i] + j * CHECK_PAGE_SIZE;
+ }
+ if (disk[j] == -2) {
+ qsyndrome(p, (uint8_t*)blocks_page[-2],
+ (uint8_t**)blocks_page,
+ syndrome_disks, CHECK_PAGE_SIZE);
+ }
+ else {
+ char *all_but_failed_blocks[syndrome_disks];
+ for(i = 0; i < syndrome_disks; i++) {
+ if (i == disk[j])
+ all_but_failed_blocks[i] = blocks_page[-1];
+ else
+ all_but_failed_blocks[i] = blocks_page[i];
+ }
+ xor_blocks(blocks_page[disk[j]],
+ all_but_failed_blocks, syndrome_disks,
+ CHECK_PAGE_SIZE);
+ }
+ }
+ else {
+ page_to_write[j] = 0;
+ }
+ }
+
+ if(pages_to_write_count > 0) {
+ int write_res = 0;
+ for(j = 0; j < (chunk_size >> CHECK_PAGE_BITS); j++) {
+ if(page_to_write[j] == 1) {
+ int slot = block_index_for_slot[disk[j]];
+ lseek64(source[slot], offsets[slot] + start * chunk_size + j * CHECK_PAGE_SIZE, SEEK_SET);
+ write_res += write(source[slot],
+ blocks[disk[j]] + j * CHECK_PAGE_SIZE,
+ CHECK_PAGE_SIZE);
+ }
+ }
+
+ if (write_res != (CHECK_PAGE_SIZE * pages_to_write_count)) {
+ fprintf(stderr, "Failed to write a full chunk.\n");
+ return -1;
+ }
+ }
+
+ return 0;
+}
+
+/* Manual repair */
+int manual_repair(int chunk_size, int syndrome_disks,
+ int failed_slot1, int failed_slot2,
+ unsigned long long start, int *block_index_for_slot,
+ char *name[], char **stripes, char **blocks, uint8_t *p,
+ int *source, unsigned long long *offsets)
+{
+ int i;
+ int fd1 = block_index_for_slot[failed_slot1];
+ int fd2 = block_index_for_slot[failed_slot2];
+ printf("Repairing stripe %llu\n", start);
+ printf("Assuming slots %d (%s) and %d (%s) are incorrect\n",
+ fd1, name[fd1],
+ fd2, name[fd2]);
+
+ if (failed_slot1 == -2 || failed_slot2 == -2) {
+ char *all_but_failed_blocks[syndrome_disks];
+ int failed_data_or_p;
+
+ if (failed_slot1 == -2)
+ failed_data_or_p = failed_slot2;
+ else
+ failed_data_or_p = failed_slot1;
+
+ printf("Repairing D/P(%d) and Q\n", failed_data_or_p);
+
+ for (i = 0; i < syndrome_disks; i++) {
+ if (i == failed_data_or_p)
+ all_but_failed_blocks[i] = blocks[-1];
+ else
+ all_but_failed_blocks[i] = blocks[i];
+ }
+ xor_blocks(blocks[failed_data_or_p],
+ all_but_failed_blocks, syndrome_disks, chunk_size);
+ qsyndrome(p, (uint8_t*)blocks[-2], (uint8_t**)blocks,
+ syndrome_disks, chunk_size);
+ } else {
+ ensure_zero_has_size(chunk_size);
+ if (failed_slot1 == -1 || failed_slot2 == -1) {
+ int failed_data;
+ if (failed_slot1 == -1)
+ failed_data = failed_slot2;
+ else
+ failed_data = failed_slot1;
+
+ printf("Repairing D(%d) and P\n", failed_data);
+ raid6_datap_recov(syndrome_disks+2, chunk_size,
+ failed_data, (uint8_t**)blocks, 1);
+ } else {
+ printf("Repairing D and D\n");
+ raid6_2data_recov(syndrome_disks+2, chunk_size,
+ failed_slot1, failed_slot2,
+ (uint8_t**)blocks, 1);
+ }
+ }
+
+ int write_res1, write_res2;
+ off64_t seek_res;
+
+ seek_res = lseek64(source[fd1],
+ offsets[fd1] + start * chunk_size, SEEK_SET);
+ if (seek_res < 0) {
+ fprintf(stderr, "lseek failed for failed_disk1\n");
+ return -1;
+ }
+ write_res1 = write(source[fd1], blocks[failed_slot1], chunk_size);
+
+ seek_res = lseek64(source[fd2],
+ offsets[fd2] + start * chunk_size, SEEK_SET);
+ if (seek_res < 0) {
+ fprintf(stderr, "lseek failed for failed_disk2\n");
+ return -1;
+ }
+ write_res2 = write(source[fd2], blocks[failed_slot2], chunk_size);
+
+ if (write_res1 != chunk_size || write_res2 != chunk_size) {
+ fprintf(stderr, "Failed to write a complete chunk.\n");
+ return -2;
+ }
+
+ return 0;
+}
int check_stripes(struct mdinfo *info, int *source, unsigned long long *offsets,
int raid_disks, int chunk_size, int level, int layout,
unsigned long long start, unsigned long long length, char *name[],
- int repair, int failed_disk1, int failed_disk2)
+ enum repair repair, int failed_disk1, int failed_disk2)
{
/* read the data and p and q blocks, and check we got them right */
- char *stripe_buf = xmalloc(raid_disks * chunk_size);
+ int data_disks = raid_disks - 2;
+ int syndrome_disks = data_disks + is_ddf(layout) * 2;
+ char *stripe_buf;
+
+ /* stripes[] is indexed by raid_disk and holds chunks from each device */
char **stripes = xmalloc(raid_disks * sizeof(char*));
- char **blocks = xmalloc(raid_disks * sizeof(char*));
- int *block_index_for_slot = xmalloc(raid_disks * sizeof(int));
+
+ /* blocks[] is indexed by syndrome number and points to either one of the
+ * chunks from 'stripes[]', or to a chunk of zeros. -1 and -2 are
+ * P and Q */
+ char **blocks = xmalloc((syndrome_disks + 2) * sizeof(char*));
+
+ /* blocks_page[] is a temporary index to just one page of the chunks
+ * that blocks[] points to. */
+ char **blocks_page = xmalloc((syndrome_disks + 2) * sizeof(char*));
+
+ /* block_index_for_slot[] provides the reverse mapping from blocks to stripes.
+ * The index is a syndrome position, the content is a raid_disk number.
+ * indicies -1 and -2 work, and are P and Q disks */
+ int *block_index_for_slot = xmalloc((syndrome_disks+2) * sizeof(int));
+
+ /* 'p' and 'q' contain calcualted P and Q, to be compared with
+ * blocks[-1] and blocks[-2];
+ */
uint8_t *p = xmalloc(chunk_size);
uint8_t *q = xmalloc(chunk_size);
+ char *zero = xmalloc(chunk_size);
int *results = xmalloc(chunk_size * sizeof(int));
sighandler_t *sig = xmalloc(3 * sizeof(sighandler_t));
- int i;
- int diskP, diskQ;
- int data_disks = raid_disks - 2;
+ int i, j;
+ int diskP, diskQ, diskD;
int err = 0;
extern int tables_ready;
if (!tables_ready)
make_tables();
+ if (posix_memalign((void**)&stripe_buf, 4096, raid_disks * chunk_size) != 0)
+ exit(4);
+ block_index_for_slot += 2;
+ blocks += 2;
+ blocks_page += 2;
+
+ memset(zero, 0, chunk_size);
for ( i = 0 ; i < raid_disks ; i++)
stripes[i] = stripe_buf + i * chunk_size;
while (length > 0) {
- int disk;
-
- printf("pos --> %llu\n", start);
+ /* The syndrome number of the broken disk is recorded
+ * in 'disk[]' which allows a different broken disk for
+ * each page.
+ */
+ int disk[chunk_size >> CHECK_PAGE_BITS];
err = lock_stripe(info, start, chunk_size, data_disks, sig);
if(err != 0) {
goto exitCheck;
}
for (i = 0 ; i < raid_disks ; i++) {
- lseek64(source[i], offsets[i] + start * chunk_size, 0);
- read(source[i], stripes[i], chunk_size);
- }
- err = unlock_all_stripes(info, sig);
- if(err != 0)
- goto exitCheck;
-
- for (i = 0 ; i < data_disks ; i++) {
- int disk = geo_map(i, start, raid_disks, level, layout);
- blocks[i] = stripes[disk];
- block_index_for_slot[disk] = i;
- printf("%d->%d\n", i, disk);
+ off64_t seek_res = lseek64(source[i], offsets[i] + start * chunk_size,
+ SEEK_SET);
+ if (seek_res < 0) {
+ fprintf(stderr, "lseek to source %d failed\n", i);
+ unlock_all_stripes(info, sig);
+ err = -1;
+ goto exitCheck;
+ }
+ int read_res = read(source[i], stripes[i], chunk_size);
+ if (read_res < chunk_size) {
+ fprintf(stderr, "Failed to read complete chunk disk %d, aborting\n", i);
+ unlock_all_stripes(info, sig);
+ err = -1;
+ goto exitCheck;
+ }
}
- qsyndrome(p, q, (uint8_t**)blocks, data_disks, chunk_size);
diskP = geo_map(-1, start, raid_disks, level, layout);
- diskQ = geo_map(-2, start, raid_disks, level, layout);
- blocks[data_disks] = stripes[diskP];
- block_index_for_slot[diskP] = data_disks;
- blocks[data_disks+1] = stripes[diskQ];
- block_index_for_slot[diskQ] = data_disks+1;
-
- if (memcmp(p, stripes[diskP], chunk_size) != 0) {
- printf("P(%d) wrong at %llu\n", diskP, start);
- }
- if (memcmp(q, stripes[diskQ], chunk_size) != 0) {
- printf("Q(%d) wrong at %llu\n", diskQ, start);
- }
- raid6_collect(chunk_size, p, q, stripes[diskP], stripes[diskQ], results);
- disk = raid6_stats(results, raid_disks, chunk_size);
+ block_index_for_slot[-1] = diskP;
+ blocks[-1] = stripes[diskP];
- if(disk >= -2) {
- disk = geo_map(disk, start, raid_disks, level, layout);
- }
- if(disk >= 0) {
- printf("Error detected at %llu: possible failed disk slot: %d --> %s\n",
- start, disk, name[disk]);
- }
- if(disk == -65535) {
- printf("Error detected at %llu: disk slot unknown\n", start);
- }
- if(repair == 1) {
- printf("Repairing stripe %llu\n", start);
- printf("Assuming slots %d (%s) and %d (%s) are incorrect\n",
- failed_disk1, name[failed_disk1],
- failed_disk2, name[failed_disk2]);
-
- if (failed_disk1 == diskQ || failed_disk2 == diskQ) {
- char *all_but_failed_blocks[data_disks];
- int failed_data_or_p;
- int failed_block_index;
-
- if (failed_disk1 == diskQ)
- failed_data_or_p = failed_disk2;
- else
- failed_data_or_p = failed_disk1;
- printf("Repairing D/P(%d) and Q\n", failed_data_or_p);
- failed_block_index = block_index_for_slot[failed_data_or_p];
- for (i=0; i < data_disks; i++)
- if (failed_block_index == i)
- all_but_failed_blocks[i] = stripes[diskP];
- else
- all_but_failed_blocks[i] = blocks[i];
- xor_blocks(stripes[failed_data_or_p],
- all_but_failed_blocks, data_disks, chunk_size);
- qsyndrome(p, (uint8_t*)stripes[diskQ], (uint8_t**)blocks, data_disks, chunk_size);
- } else {
- ensure_zero_has_size(chunk_size);
- if (failed_disk1 == diskP || failed_disk2 == diskP) {
- int failed_data, failed_block_index;
- if (failed_disk1 == diskP)
- failed_data = failed_disk2;
- else
- failed_data = failed_disk1;
- failed_block_index = block_index_for_slot[failed_data];
- printf("Repairing D(%d) and P\n", failed_data);
- raid6_datap_recov(raid_disks, chunk_size, failed_block_index, (uint8_t**)blocks);
+ diskQ = geo_map(-2, start, raid_disks, level, layout);
+ block_index_for_slot[-2] = diskQ;
+ blocks[-2] = stripes[diskQ];
+
+ if (!is_ddf(layout)) {
+ /* The syndrome-order of disks starts immediately after 'Q',
+ * but skips P */
+ diskD = diskQ;
+ for (i = 0 ; i < data_disks ; i++) {
+ diskD = diskD + 1;
+ if (diskD >= raid_disks)
+ diskD = 0;
+ if (diskD == diskP)
+ diskD += 1;
+ if (diskD >= raid_disks)
+ diskD = 0;
+ blocks[i] = stripes[diskD];
+ block_index_for_slot[i] = diskD;
+ }
+ } else {
+ /* The syndrome-order exactly follows raid-disk
+ * numbers, with ZERO in place of P and Q
+ */
+ for (i = 0 ; i < raid_disks; i++) {
+ if (i == diskP || i == diskQ) {
+ blocks[i] = zero;
+ block_index_for_slot[i] = -1;
} else {
- printf("Repairing D and D\n");
- int failed_block_index1 = block_index_for_slot[failed_disk1];
- int failed_block_index2 = block_index_for_slot[failed_disk2];
- if (failed_block_index1 > failed_block_index2) {
- int t = failed_block_index1;
- failed_block_index1 = failed_block_index2;
- failed_block_index2 = t;
- }
- raid6_2data_recov(raid_disks, chunk_size, failed_block_index1, failed_block_index2, (uint8_t**)blocks);
+ blocks[i] = stripes[i];
+ block_index_for_slot[i] = i;
}
}
+ }
- err = lock_stripe(info, start, chunk_size, data_disks, sig);
- if(err != 0) {
- if (err != 2)
- unlock_all_stripes(info, sig);
- goto exitCheck;
- }
-
- lseek64(source[failed_disk1], offsets[failed_disk1] + start * chunk_size, 0);
- write(source[failed_disk1], stripes[failed_disk1], chunk_size);
- lseek64(source[failed_disk2], offsets[failed_disk2] + start * chunk_size, 0);
- write(source[failed_disk2], stripes[failed_disk2], chunk_size);
+ qsyndrome(p, q, (uint8_t**)blocks, syndrome_disks, chunk_size);
- err = unlock_all_stripes(info, sig);
- if(err != 0)
- goto exitCheck;
- } else if (disk >= 0 && repair == 2) {
- printf("Auto-repairing slot %d (%s)\n", disk, name[disk]);
- if (disk == diskQ) {
- qsyndrome(p, (uint8_t*)stripes[diskQ], (uint8_t**)blocks, data_disks, chunk_size);
- } else {
- char *all_but_failed_blocks[data_disks];
- int failed_block_index = block_index_for_slot[disk];
- for (i=0; i < data_disks; i++)
- if (failed_block_index == i)
- all_but_failed_blocks[i] = stripes[diskP];
- else
- all_but_failed_blocks[i] = blocks[i];
- xor_blocks(stripes[disk],
- all_but_failed_blocks, data_disks, chunk_size);
+ raid6_collect(chunk_size, p, q, stripes[diskP], stripes[diskQ], results);
+ raid6_stats(disk, results, raid_disks, chunk_size);
+
+ for(j = 0; j < (chunk_size >> CHECK_PAGE_BITS); j++) {
+ int role = disk[j];
+ if (role >= -2) {
+ int slot = block_index_for_slot[role];
+ if (slot >= 0)
+ printf("Error detected at stripe %llu, page %d: possible failed disk slot %d: %d --> %s\n",
+ start, j, role, slot, name[slot]);
+ else
+ printf("Error detected at stripe %llu, page %d: failed slot %d should be zeros\n",
+ start, j, role);
+ } else if(disk[j] == -65535) {
+ printf("Error detected at stripe %llu, page %d: disk slot unknown\n", start, j);
}
+ }
- err = lock_stripe(info, start, chunk_size, data_disks, sig);
+ if(repair == AUTO_REPAIR) {
+ err = autorepair(disk, start, chunk_size,
+ name, raid_disks, syndrome_disks, blocks_page,
+ blocks, p, block_index_for_slot,
+ source, offsets);
if(err != 0) {
- if (err != 2)
- unlock_all_stripes(info, sig);
+ unlock_all_stripes(info, sig);
goto exitCheck;
}
+ }
- lseek64(source[disk], offsets[disk] + start * chunk_size, 0);
- write(source[disk], stripes[disk], chunk_size);
-
- err = unlock_all_stripes(info, sig);
- if(err != 0)
+ if(repair == MANUAL_REPAIR) {
+ int failed_slot1 = -1, failed_slot2 = -1;
+ for (i = -2; i < syndrome_disks; i++) {
+ if (block_index_for_slot[i] == failed_disk1)
+ failed_slot1 = i;
+ if (block_index_for_slot[i] == failed_disk2)
+ failed_slot2 = i;
+ }
+ err = manual_repair(chunk_size, syndrome_disks,
+ failed_slot1, failed_slot2,
+ start, block_index_for_slot,
+ name, stripes, blocks, p,
+ source, offsets);
+ if(err == -1) {
+ unlock_all_stripes(info, sig);
goto exitCheck;
+ }
}
+ err = unlock_all_stripes(info, sig);
+ if(err != 0) {
+ goto exitCheck;
+ }
length--;
start++;
free(stripe_buf);
free(stripes);
- free(blocks);
+ free(blocks-2);
+ free(blocks_page-2);
+ free(block_index_for_slot-2);
free(p);
free(q);
free(results);
+ free(sig);
return err;
}
int chunk_size = 0;
int layout = -1;
int level = 6;
- int repair = 0;
- int failed_disk1, failed_disk2;
+ enum repair repair = NO_REPAIR;
+ int failed_disk1 = -1;
+ int failed_disk2 = -1;
unsigned long long start, length;
int i;
int mdfd;
goto exitHere;
}
- info = sysfs_read(mdfd, -1,
+ info = sysfs_read(mdfd, NULL,
GET_LEVEL|
GET_LAYOUT|
GET_DISKS|
- GET_DEGRADED |
+ GET_STATE |
GET_COMPONENT|
GET_CHUNK|
GET_DEVS|
exit_err = 1;
goto exitHere;
}
- repair = 1;
+ repair = MANUAL_REPAIR;
start = getnum(argv[3], &err);
length = 1;
failed_disk1 = getnum(argv[4], &err);
start = getnum(argv[2], &err);
length = getnum(argv[3], &err);
if (argc >= 5 && strcmp(argv[4], "autorepair")==0)
- repair = 2;
+ repair = AUTO_REPAIR;
}
if (err) {
if(disk_slot >= 0) {
disk_name[disk_slot] = map_dev(comp->disk.major, comp->disk.minor, 0);
offsets[disk_slot] = comp->data_offset * 512;
- fds[disk_slot] = open(disk_name[disk_slot], O_RDWR);
+ fds[disk_slot] = open(disk_name[disk_slot], O_RDWR | O_DIRECT);
if (fds[disk_slot] < 0) {
perror(disk_name[disk_slot]);
fprintf(stderr,"%s: cannot open %s\n", prg, disk_name[disk_slot]);
raid_disks, chunk_size, level, layout,
start, length, disk_name, repair, failed_disk1, failed_disk2);
if (rv != 0) {
- fprintf(stderr,
- "%s: check_stripes returned %d\n", prg, rv);
+ fprintf(stderr, "%s: check_stripes returned %d\n", prg, rv);
exit_err = 7;
goto exitHere;
}
projects / thirdparty / mdadm.git / blobdiff