/* Recover two failed blocks. */
void raid6_dual_recov(int disks, size_t bytes, int faila, int failb, void **ptrs)
{
if ( faila > failb ) {
int tmp = faila;
faila = failb;
failb = tmp;
}
if ( failb == disks-1 ) {
if ( faila == disks-2 ) {
/* P+Q failure. Just rebuild the syndrome. */
raid6_call.gen_syndrome(disks, bytes, ptrs);
} else {
/* data+Q failure. Reconstruct data from P,
then rebuild syndrome. */
/* NOT IMPLEMENTED - equivalent to RAID-5 */
}
} else {
if ( failb == disks-2 ) {
/* data+P failure. */
raid6_datap_recov(disks, bytes, faila, ptrs);
} else {
/* data+data failure. */
raid6_2data_recov(disks, bytes, faila, failb, ptrs);
}
}
}
/**
* async_raid6_datap_recov - asynchronously calculate a data and the 'p' block
* @disks: number of disks in the RAID-6 array
* @bytes: block size
* @faila: failed drive index
* @blocks: array of source pointers where the last two entries are p and q
* @submit: submission/completion modifiers
*/
struct dma_async_tx_descriptor *
async_raid6_datap_recov(int disks, size_t bytes, int faila,
struct page **blocks, struct async_submit_ctl *submit)
{
struct dma_async_tx_descriptor *tx = NULL;
struct page *p, *q, *dq;
u8 coef;
enum async_tx_flags flags = submit->flags;
dma_async_tx_callback cb_fn = submit->cb_fn;
void *cb_param = submit->cb_param;
void *scribble = submit->scribble;
int good_srcs, good, i;
struct page *srcs[2];
pr_debug("%s: disks: %d len: %zu\n", __func__, disks, bytes);
/* we need to preserve the contents of 'blocks' for the async
* case, so punt to synchronous if a scribble buffer is not available
*/
if (!scribble) {
void **ptrs = (void **) blocks;
async_tx_quiesce(&submit->depend_tx);
for (i = 0; i < disks; i++)
if (blocks[i] == NULL)
ptrs[i] = (void*)raid6_empty_zero_page;
else
ptrs[i] = page_address(blocks[i]);
raid6_datap_recov(disks, bytes, faila, ptrs);
async_tx_sync_epilog(submit);
return NULL;
}
good_srcs = 0;
good = -1;
for (i = 0; i < disks-2; i++) {
if (i == faila)
continue;
if (blocks[i]) {
good = i;
good_srcs++;
if (good_srcs > 1)
break;
}
}
BUG_ON(good_srcs == 0);
p = blocks[disks-2];
q = blocks[disks-1];
/* Compute syndrome with zero for the missing data page
* Use the dead data page as temporary storage for delta q
*/
dq = blocks[faila];
blocks[faila] = NULL;
blocks[disks-1] = dq;
/* in the 4-disk case we only need to perform a single source
* multiplication with the one good data block.
*/
if (good_srcs == 1) {
struct page *g = blocks[good];
init_async_submit(submit, ASYNC_TX_FENCE, tx, NULL, NULL,
scribble);
tx = async_memcpy(p, g, 0, 0, bytes, submit);
init_async_submit(submit, ASYNC_TX_FENCE, tx, NULL, NULL,
scribble);
tx = async_mult(dq, g, raid6_gfexp[good], bytes, submit);
} else {
init_async_submit(submit, ASYNC_TX_FENCE, tx, NULL, NULL,
scribble);
tx = async_gen_syndrome(blocks, 0, disks, bytes, submit);
}
/* Restore pointer table */
blocks[faila] = dq;
blocks[disks-1] = q;
/* calculate g^{-faila} */
coef = raid6_gfinv[raid6_gfexp[faila]];
srcs[0] = dq;
srcs[1] = q;
init_async_submit(submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_DROP_DST, tx,
NULL, NULL, scribble);
tx = async_xor(dq, srcs, 0, 2, bytes, submit);
init_async_submit(submit, ASYNC_TX_FENCE, tx, NULL, NULL, scribble);
tx = async_mult(dq, dq, coef, bytes, submit);
srcs[0] = p;
srcs[1] = dq;
init_async_submit(submit, flags | ASYNC_TX_XOR_DROP_DST, tx, cb_fn,
cb_param, scribble);
tx = async_xor(p, srcs, 0, 2, bytes, submit);
return tx;
}
/*******************************************************************************
* Function: save_stripes
* Description:
* Function reads data (only data without P and Q) from array and writes
* it to buf and opcjonaly to backup files
* Parameters:
* source : A list of 'fds' of the active disks.
* Some may be absent
* offsets : A list of offsets on disk belonging
* to the array [bytes]
* raid_disks : geometry: number of disks in the array
* chunk_size : geometry: chunk size [bytes]
* level : geometry: RAID level
* layout : geometry: layout
* nwrites : number of backup files
* dest : A list of 'fds' for mirrored targets
* (e.g. backup files). They are already seeked to right
* (write) location. If NULL, data will be wrote
* to the buf only
* start : start address of data to read (must be stripe-aligned)
* [bytes]
* length - : length of data to read (must be stripe-aligned)
* [bytes]
* buf : buffer for data. It is large enough to hold
* one stripe. It is stripe aligned
* Returns:
* 0 : success
* -1 : fail
******************************************************************************/
int save_stripes(int *source, unsigned long long *offsets,
int raid_disks, int chunk_size, int level, int layout,
int nwrites, int *dest,
unsigned long long start, unsigned long long length,
char *buf)
{
int len;
int data_disks = raid_disks - (level == 0 ? 0 : level <=5 ? 1 : 2);
int disk;
int i;
unsigned long long length_test;
if (!tables_ready)
make_tables();
ensure_zero_has_size(chunk_size);
len = data_disks * chunk_size;
length_test = length / len;
length_test *= len;
if (length != length_test) {
dprintf("Error: save_stripes(): Data are not alligned. EXIT\n");
dprintf("\tArea for saving stripes (length) = %llu\n", length);
dprintf("\tWork step (len) = %i\n", len);
dprintf("\tExpected save area (length_test) = %llu\n",
length_test);
abort();
}
while (length > 0) {
int failed = 0;
int fdisk[3], fblock[3];
for (disk = 0; disk < raid_disks ; disk++) {
unsigned long long offset;
int dnum;
offset = (start/chunk_size/data_disks)*chunk_size;
dnum = geo_map(disk < data_disks ? disk : data_disks - disk - 1,
start/chunk_size/data_disks,
raid_disks, level, layout);
if (dnum < 0) abort();
if (source[dnum] < 0 ||
lseek64(source[dnum], offsets[dnum]+offset, 0) < 0 ||
read(source[dnum], buf+disk * chunk_size, chunk_size)
!= chunk_size)
if (failed <= 2) {
fdisk[failed] = dnum;
fblock[failed] = disk;
failed++;
}
}
if (failed == 0 || fblock[0] >= data_disks)
/* all data disks are good */
;
else if (failed == 1 || fblock[1] >= data_disks+1) {
/* one failed data disk and good parity */
char *bufs[data_disks];
for (i=0; i < data_disks; i++)
if (fblock[0] == i)
bufs[i] = buf + data_disks*chunk_size;
else
bufs[i] = buf + i*chunk_size;
xor_blocks(buf + fblock[0]*chunk_size,
bufs, data_disks, chunk_size);
} else if (failed > 2 || level != 6)
/* too much failure */
return -1;
else {
/* RAID6 computations needed. */
uint8_t *bufs[data_disks+4];
int qdisk;
int syndrome_disks;
disk = geo_map(-1, start/chunk_size/data_disks,
raid_disks, level, layout);
qdisk = geo_map(-2, start/chunk_size/data_disks,
raid_disks, level, layout);
if (is_ddf(layout)) {
/* q over 'raid_disks' blocks, in device order.
* 'p' and 'q' get to be all zero
*/
for (i = 0; i < raid_disks; i++)
bufs[i] = zero;
for (i = 0; i < data_disks; i++) {
int dnum = geo_map(i,
start/chunk_size/data_disks,
raid_disks, level, layout);
int snum;
/* i is the logical block number, so is index to 'buf'.
* dnum is physical disk number
* and thus the syndrome number.
*/
snum = dnum;
bufs[snum] = (uint8_t*)buf + chunk_size * i;
}
syndrome_disks = raid_disks;
} else {
/* for md, q is over 'data_disks' blocks,
* starting immediately after 'q'
* Note that for the '_6' variety, the p block
* makes a hole that we need to be careful of.
*/
int j;
int snum = 0;
for (j = 0; j < raid_disks; j++) {
int dnum = (qdisk + 1 + j) % raid_disks;
if (dnum == disk || dnum == qdisk)
continue;
for (i = 0; i < data_disks; i++)
if (geo_map(i,
start/chunk_size/data_disks,
raid_disks, level, layout) == dnum)
break;
/* i is the logical block number, so is index to 'buf'.
* dnum is physical disk number
* snum is syndrome disk for which 0 is immediately after Q
*/
bufs[snum] = (uint8_t*)buf + chunk_size * i;
if (fblock[0] == i)
fdisk[0] = snum;
if (fblock[1] == i)
fdisk[1] = snum;
snum++;
}
syndrome_disks = data_disks;
}
/* Place P and Q blocks at end of bufs */
bufs[syndrome_disks] = (uint8_t*)buf + chunk_size * data_disks;
bufs[syndrome_disks+1] = (uint8_t*)buf + chunk_size * (data_disks+1);
if (fblock[1] == data_disks)
/* One data failed, and parity failed */
raid6_datap_recov(syndrome_disks+2, chunk_size,
fdisk[0], bufs, 0);
else {
/* Two data blocks failed, P,Q OK */
raid6_2data_recov(syndrome_disks+2, chunk_size,
fdisk[0], fdisk[1], bufs, 0);
}
}
if (dest) {
for (i = 0; i < nwrites; i++)
if (write(dest[i], buf, len) != len)
return -1;
} else {
/* build next stripe in buffer */
buf += len;
}
length -= len;
start += len;
}
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)
{
/* read the data and p and q blocks, and check we got them right */
char *stripe_buf = xmalloc(raid_disks * chunk_size);
char **stripes = xmalloc(raid_disks * sizeof(char*));
char **blocks = xmalloc(raid_disks * sizeof(char*));
int *block_index_for_slot = xmalloc(raid_disks * sizeof(int));
uint8_t *p = xmalloc(chunk_size);
uint8_t *q = 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 err = 0;
extern int tables_ready;
if (!tables_ready)
make_tables();
for ( i = 0 ; i < raid_disks ; i++)
stripes[i] = stripe_buf + i * chunk_size;
while (length > 0) {
int disk;
printf("pos --> %llu\n", start);
err = lock_stripe(info, start, chunk_size, data_disks, sig);
if(err != 0) {
if (err != 2)
unlock_all_stripes(info, sig);
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);
}
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);
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);
} 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);
}
}
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);
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);
}
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[disk], offsets[disk] + start * chunk_size, 0);
write(source[disk], stripes[disk], chunk_size);
err = unlock_all_stripes(info, sig);
if(err != 0)
goto exitCheck;
}
length--;
start++;
}
exitCheck:
free(stripe_buf);
free(stripes);
free(blocks);
free(p);
free(q);
free(results);
return err;
}
