raid5-ppl: Recovery support for multiple partial parity logs
authorPawel Baldysiak <pawel.baldysiak@intel.com>
Wed, 16 Aug 2017 15:13:46 +0000 (17:13 +0200)
committerShaohua Li <shli@fb.com>
Mon, 28 Aug 2017 14:45:49 +0000 (07:45 -0700)
Search PPL buffer in order to find out the latest PPL header (the one
with largest generation number) and use it for recovery. The PPL entry
format and recovery algorithm are the same as for single PPL approach.

Signed-off-by: Pawel Baldysiak <pawel.baldysiak@intel.com>
Signed-off-by: Shaohua Li <shli@fb.com>
drivers/md/raid5-ppl.c

index b313f17a62601640817f8ba1bb3629c1a80e29dd..a98ef172f8e8b58cc23db42059d82ba1f995f959 100644 (file)
@@ -831,12 +831,14 @@ out:
        return ret;
 }
 
-static int ppl_recover(struct ppl_log *log, struct ppl_header *pplhdr)
+static int ppl_recover(struct ppl_log *log, struct ppl_header *pplhdr,
+                      sector_t offset)
 {
        struct ppl_conf *ppl_conf = log->ppl_conf;
        struct md_rdev *rdev = log->rdev;
        struct mddev *mddev = rdev->mddev;
-       sector_t ppl_sector = rdev->ppl.sector + (PPL_HEADER_SIZE >> 9);
+       sector_t ppl_sector = rdev->ppl.sector + offset +
+                             (PPL_HEADER_SIZE >> 9);
        struct page *page;
        int i;
        int ret = 0;
@@ -920,6 +922,9 @@ static int ppl_write_empty_header(struct ppl_log *log)
                return -ENOMEM;
 
        pplhdr = page_address(page);
+       /* zero out PPL space to avoid collision with old PPLs */
+       blkdev_issue_zeroout(rdev->bdev, rdev->ppl.sector,
+                           log->rdev->ppl.size, GFP_NOIO, 0);
        memset(pplhdr->reserved, 0xff, PPL_HDR_RESERVED);
        pplhdr->signature = cpu_to_le32(log->ppl_conf->signature);
        pplhdr->checksum = cpu_to_le32(~crc32c_le(~0, pplhdr, PAGE_SIZE));
@@ -940,63 +945,110 @@ static int ppl_load_distributed(struct ppl_log *log)
        struct ppl_conf *ppl_conf = log->ppl_conf;
        struct md_rdev *rdev = log->rdev;
        struct mddev *mddev = rdev->mddev;
-       struct page *page;
-       struct ppl_header *pplhdr;
+       struct page *page, *page2, *tmp;
+       struct ppl_header *pplhdr = NULL, *prev_pplhdr = NULL;
        u32 crc, crc_stored;
        u32 signature;
-       int ret = 0;
+       int ret = 0, i;
+       sector_t pplhdr_offset = 0, prev_pplhdr_offset = 0;
 
        pr_debug("%s: disk: %d\n", __func__, rdev->raid_disk);
-
-       /* read PPL header */
+       /* read PPL headers, find the recent one */
        page = alloc_page(GFP_KERNEL);
        if (!page)
                return -ENOMEM;
 
-       if (!sync_page_io(rdev, rdev->ppl.sector - rdev->data_offset,
-                         PAGE_SIZE, page, REQ_OP_READ, 0, false)) {
-               md_error(mddev, rdev);
-               ret = -EIO;
-               goto out;
+       page2 = alloc_page(GFP_KERNEL);
+       if (!page2) {
+               __free_page(page);
+               return -ENOMEM;
        }
-       pplhdr = page_address(page);
 
-       /* check header validity */
-       crc_stored = le32_to_cpu(pplhdr->checksum);
-       pplhdr->checksum = 0;
-       crc = ~crc32c_le(~0, pplhdr, PAGE_SIZE);
+       /* searching ppl area for latest ppl */
+       while (pplhdr_offset < rdev->ppl.size - (PPL_HEADER_SIZE >> 9)) {
+               if (!sync_page_io(rdev,
+                                 rdev->ppl.sector - rdev->data_offset +
+                                 pplhdr_offset, PAGE_SIZE, page, REQ_OP_READ,
+                                 0, false)) {
+                       md_error(mddev, rdev);
+                       ret = -EIO;
+                       /* if not able to read - don't recover any PPL */
+                       pplhdr = NULL;
+                       break;
+               }
+               pplhdr = page_address(page);
+
+               /* check header validity */
+               crc_stored = le32_to_cpu(pplhdr->checksum);
+               pplhdr->checksum = 0;
+               crc = ~crc32c_le(~0, pplhdr, PAGE_SIZE);
+
+               if (crc_stored != crc) {
+                       pr_debug("%s: ppl header crc does not match: stored: 0x%x calculated: 0x%x (offset: %llu)\n",
+                                __func__, crc_stored, crc,
+                                (unsigned long long)pplhdr_offset);
+                       pplhdr = prev_pplhdr;
+                       pplhdr_offset = prev_pplhdr_offset;
+                       break;
+               }
 
-       if (crc_stored != crc) {
-               pr_debug("%s: ppl header crc does not match: stored: 0x%x calculated: 0x%x\n",
-                        __func__, crc_stored, crc);
-               ppl_conf->mismatch_count++;
-               goto out;
-       }
+               signature = le32_to_cpu(pplhdr->signature);
 
-       signature = le32_to_cpu(pplhdr->signature);
+               if (mddev->external) {
+                       /*
+                        * For external metadata the header signature is set and
+                        * validated in userspace.
+                        */
+                       ppl_conf->signature = signature;
+               } else if (ppl_conf->signature != signature) {
+                       pr_debug("%s: ppl header signature does not match: stored: 0x%x configured: 0x%x (offset: %llu)\n",
+                                __func__, signature, ppl_conf->signature,
+                                (unsigned long long)pplhdr_offset);
+                       pplhdr = prev_pplhdr;
+                       pplhdr_offset = prev_pplhdr_offset;
+                       break;
+               }
 
-       if (mddev->external) {
-               /*
-                * For external metadata the header signature is set and
-                * validated in userspace.
-                */
-               ppl_conf->signature = signature;
-       } else if (ppl_conf->signature != signature) {
-               pr_debug("%s: ppl header signature does not match: stored: 0x%x configured: 0x%x\n",
-                        __func__, signature, ppl_conf->signature);
-               ppl_conf->mismatch_count++;
-               goto out;
+               if (prev_pplhdr && le64_to_cpu(prev_pplhdr->generation) >
+                   le64_to_cpu(pplhdr->generation)) {
+                       /* previous was newest */
+                       pplhdr = prev_pplhdr;
+                       pplhdr_offset = prev_pplhdr_offset;
+                       break;
+               }
+
+               prev_pplhdr_offset = pplhdr_offset;
+               prev_pplhdr = pplhdr;
+
+               tmp = page;
+               page = page2;
+               page2 = tmp;
+
+               /* calculate next potential ppl offset */
+               for (i = 0; i < le32_to_cpu(pplhdr->entries_count); i++)
+                       pplhdr_offset +=
+                           le32_to_cpu(pplhdr->entries[i].pp_size) >> 9;
+               pplhdr_offset += PPL_HEADER_SIZE >> 9;
        }
 
+       /* no valid ppl found */
+       if (!pplhdr)
+               ppl_conf->mismatch_count++;
+       else
+               pr_debug("%s: latest PPL found at offset: %llu, with generation: %llu\n",
+                   __func__, (unsigned long long)pplhdr_offset,
+                   le64_to_cpu(pplhdr->generation));
+
        /* attempt to recover from log if we are starting a dirty array */
-       if (!mddev->pers && mddev->recovery_cp != MaxSector)
-               ret = ppl_recover(log, pplhdr);
-out:
+       if (pplhdr && !mddev->pers && mddev->recovery_cp != MaxSector)
+               ret = ppl_recover(log, pplhdr, pplhdr_offset);
+
        /* write empty header if we are starting the array */
        if (!ret && !mddev->pers)
                ret = ppl_write_empty_header(log);
 
        __free_page(page);
+       __free_page(page2);
 
        pr_debug("%s: return: %d mismatch_count: %d recovered_entries: %d\n",
                 __func__, ret, ppl_conf->mismatch_count,