fs/btrfs/raid56.c

115  * the stripe hash table is used for locking, and to collect
116  * bios in hopes of making a full stripe
160  * use the page uptodate bit in the stripe cache array
200  * we hash on the first logical address of the stripe
289  * Stealing an rbio means taking all the uptodate pages from the stripe array
314 		 * all data stripe pages present and uptodate.
454  * insert an rbio into the stripe cache.  It
524  * Returns true if the bio list inside this rbio covers an entire stripe (no
544  * the same stripe and if they are both going in the same
576 	 * We've need read the full stripe from the drive.
581 	 * change this stripe needs to do their own rmw.
611 /* Grab a sector inside P stripe */
618 /* Grab a sector inside Q stripe, return NULL if not RAID6 */
628  * The first stripe in the table for a logical address
631  * 1) Nobody has the stripe locked yet.  The rbio is given
635  * 2) Someone has the stripe locked, but we're able to merge
639  * 3) Someone has the stripe locked, but we're not able to merge.
732  * rbios waiting for this stripe, the next one on the list will be started
768 		 * waiting for the chance to lock this stripe.
863  * @sector_nr:		Sector number inside the stripe,
915 	 * Our current stripe len should be fixed to 64k thus stripe_nsectors
998  * Return the total number of errors found in the vertical stripe of @sector_nr.
1000  * @faila and @failb will also be updated to the first and second stripe
1053 	struct btrfs_io_stripe *stripe;
1065 	stripe = &rbio->bioc->stripes[stripe_nr];
1066 	disk_start = stripe->physical + sector_nr * sectorsize;
1068 	/* if the device is missing, just fail this stripe */
1069 	if (!stripe->dev->bdev) {
1093 		    last->bi_bdev == stripe->dev->bdev) {
1102 	bio = bio_alloc(stripe->dev->bdev,
1139  * searching through the bio list as we setup the IO in finish_rmw or stripe
1163 	/* We rely on bio->bi_bdev to find the stripe number. */
1211 /* Generate PQ for one vertical stripe. */
1217 	int stripe;
1220 	/* First collect one sector from each data stripe */
1221 	for (stripe = 0; stripe < rbio->nr_data; stripe++) {
1222 		sector = sector_in_rbio(rbio, stripe, sectornr, 0);
1223 		pointers[stripe] = kmap_local_page(sector->page) +
1227 	/* Then add the parity stripe */
1230 	pointers[stripe++] = kmap_local_page(sector->page) + sector->pgoff;
1239 		pointers[stripe++] = kmap_local_page(sector->page) +
1250 	for (stripe = stripe - 1; stripe >= 0; stripe--)
1251 		kunmap_local(pointers[stripe]);
1257 	/* The total sector number inside the full stripe. */
1260 	int stripe;
1282 		stripe = total_sector_nr / rbio->stripe_nsectors;
1285 		/* This vertical stripe has no data, skip it. */
1289 		if (stripe < rbio->nr_data) {
1290 			sector = sector_in_rbio(rbio, stripe, sectornr, 1);
1294 			sector = rbio_stripe_sector(rbio, stripe, sectornr);
1297 		ret = rbio_add_io_sector(rbio, bio_list, sector, stripe,
1309 	 * Thus the source stripe number (in replace_stripe_src) should be valid.
1317 		stripe = total_sector_nr / rbio->stripe_nsectors;
1322 		 * and replace_stripe_src[0] indicates the stripe number we
1325 		if (stripe != rbio->bioc->replace_stripe_src) {
1327 			 * We can skip the whole stripe completely, note
1335 		/* This vertical stripe has no data, skip it. */
1339 		if (stripe < rbio->nr_data) {
1340 			sector = sector_in_rbio(rbio, stripe, sectornr, 1);
1344 			sector = rbio_stripe_sector(rbio, stripe, sectornr);
1376 	 * and mark the stripe error instead.
1488 	/* No data csum for the whole stripe, no need to verify. */
1571  * Any time we get a partial stripe write while plugged
1616 			/* We have a full stripe, queue it down. */
1699 	 * Either we don't have any existing plug, or we're doing a full stripe,
1741  * Recover a vertical stripe specified by @sector_nr.
1768 	 * No errors in the vertical stripe, skip it.  Can happen for recovery
1769 	 * which only part of a stripe failed csum check.
1778 	 * Setup our array of pointers with sectors from each stripe
1805 				 * Just the P stripe has failed, without
1806 				 * a bad data or Q stripe.
1808 				 * recovery for this stripe.
1819 		 * If the q stripe is failed, do a pstripe reconstruction from
1821 		 * If both the q stripe and the P stripe are failed, we're
1830 				 * can skip this vertical stripe.
1834 			 * Otherwise we have one bad data stripe and
1835 			 * a good P stripe.  raid5!
1850 		/* Rebuild from P stripe here (raid5 or raid6). */
1870 	 * failed sectors repaired in the vertical stripe, thus they are now
1970 		int stripe = total_sector_nr / rbio->stripe_nsectors;
1979 		if (!rbio->bioc->stripes[stripe].dev->bdev ||
1989 		sector = rbio_stripe_sector(rbio, stripe, sectornr);
1990 		ret = rbio_add_io_sector(rbio, &bio_list, sector, stripe,
2037 		/* This vertical stripe doesn't have errors. */
2048 		/* Now select another stripe to mark as error. */
2059 	/* We should found at least one vertical stripe with error.*/
2090 	 * for 'mirror_num > 2', select a stripe to fail on every retry.
2119 	 *   stripe lock, if we trigger a metadata read, and it needs to do
2147 	 * longer safe for this particular sub-stripe write.
2150 "sub-stripe write for full stripe %llu is not safe, failed to get csum: %d",
2180 		int stripe = total_sector_nr / rbio->stripe_nsectors;
2183 		sector = rbio_stripe_sector(rbio, stripe, sectornr);
2185 			       stripe, sectornr, REQ_OP_READ);
2261 	 * needed for both full-stripe and sub-stripe writes.
2268 	 * Either full stripe write, or we have every data sector already
2273 		 * Now we're doing sub-stripe write, also need all data stripes
2289 	 * bio list any more, anyone else that wants to change this stripe
2353  * The following code is used to scrub/replace the parity stripe
2385 	 * stripe.
2434 	int stripe;
2453 	 * Replace is running and our P/Q stripe is being replaced, then we
2475 		/* RAID6, allocate and map temp space for the Q stripe */
2489 	/* Map the parity stripe just once */
2496 		/* first collect one page from each data stripe */
2497 		for (stripe = 0; stripe < nr_data; stripe++) {
2498 			sector = sector_in_rbio(rbio, stripe, sectornr, 0);
2499 			pointers[stripe] = kmap_local_page(sector->page) +
2524 		for (stripe = nr_data - 1; stripe >= 0; stripe--)
2525 			kunmap_local(pointers[stripe]);
2556 	 * Replace is running and our parity stripe needs to be duplicated to
2557 	 * the target device.  Check we have a valid source stripe number.
2580 static inline int is_data_stripe(struct btrfs_raid_bio *rbio, int stripe)
2582 	if (stripe >= 0 && stripe < rbio->nr_data)
2651 		 * Here means we got one corrupted data stripe and one
2654 		 * data, or we can not repair the data stripe.
2681 		int stripe = total_sector_nr / rbio->stripe_nsectors;
2684 		/* No data in the vertical stripe, no need to read. */
2691 		 * in the bio list we don't need to read it off the stripe.
2693 		sector = sector_in_rbio(rbio, stripe, sectornr, 1);
2697 		sector = rbio_stripe_sector(rbio, stripe, sectornr);
2705 		ret = rbio_add_io_sector(rbio, &bio_list, sector, stripe,
2796 	/* data_logical must be at stripe boundary and inside the full stripe. */