Skip to contentSkip to conversation

btrfs: disk-io: Kill the BUG_ON()s in write_and_map_eb()

PATCH Superseded Qu Wenruo 3 participants 3 messages Mar 2, 2020 Btrfs file system

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore

Qu Wenruo

author
#0 lore raw

PATCH 1/1 btrfs: disk-io: Kill the BUG_ON()s in write_and_map_eb()

All callers of write_and_map_eb(), except btrfs-corrupt-block, have handled error, but inside write_and_map_eb() itself, the only error handling is BUG_ON().

This patch will kill all the BUG_ON()s inside write_and_map_eb(), and enhance the the caller in btrfs-corrupt-block() to handle the error.

---
 btrfs-corrupt-block.c       |  9 ++++++++-
 cmds/rescue-super-recover.c |  3 ++-
 disk-io.c                   | 26 +++++++++++++++++++++++---
 3 files changed, 33 insertions(+), 5 deletions(-)
btrfs-corrupt-block.c +8 -1
index 95df871a7822..3c236e146176 100644
@@ -771,8 +771,15 @@ static int corrupt_metadata_block(struct btrfs_fs_info *fs_info, u64 block,
 		u64 bogus = generate_u64(orig);
 
 		btrfs_set_header_generation(eb, bogus);
-		write_and_map_eb(fs_info, eb);
+		ret = write_and_map_eb(fs_info, eb);
 		free_extent_buffer(eb);
+		if (ret < 0) {
+			errno = -ret;
+			fprintf(stderr,
+				"failed to write extent buffer at %llu: %m",
+				eb->start);
+			return ret;
+		}
 		break;
 		}
 	case BTRFS_METADATA_BLOCK_SHIFT_ITEMS:
cmds/rescue-super-recover.c +2 -1
index 5d6bea836c8b..62f4f7754720 100644
@@ -276,7 +276,8 @@ int btrfs_recover_superblocks(const char *dname,
 				  struct super_block_record, list);
 
 	root = open_ctree(record->device_name, record->bytenr,
-			  OPEN_CTREE_RECOVER_SUPER | OPEN_CTREE_WRITES);
+			  OPEN_CTREE_RECOVER_SUPER | OPEN_CTREE_WRITES |
+			  OPEN_CTREE_EXCLUSIVE);
 	if (!root) {
 		ret = 3;
 		goto no_recover;
disk-io.c +23 -3
index e8a2e4afa93a..9ff62fcd54d1 100644
@@ -487,20 +487,40 @@ int write_and_map_eb(struct btrfs_fs_info *fs_info, struct extent_buffer *eb)
 	length = eb->len;
 	ret = btrfs_map_block(fs_info, WRITE, eb->start, &length,
 			      &multi, 0, &raid_map);
+	if (ret < 0) {
+		errno = -ret;
+		error("failed to map bytenr %llu length %u: %m",
+			eb->start, eb->len);
+		goto out;
+	}
 
 	if (raid_map) {
 		ret = write_raid56_with_parity(fs_info, eb, multi,
 					       length, raid_map);
-		BUG_ON(ret);
+		if (ret < 0) {
+			errno = -ret;
+			error(
+		"failed to write raid56 stripe for bytenr %llu length %llu: %m",
+				eb->start, length);
+			goto out;
+		}
 	} else while (dev_nr < multi->num_stripes) {
-		BUG_ON(ret);
 		eb->fd = multi->stripes[dev_nr].dev->fd;
 		eb->dev_bytenr = multi->stripes[dev_nr].physical;
 		multi->stripes[dev_nr].dev->total_ios++;
 		dev_nr++;
 		ret = write_extent_to_disk(eb);
-		BUG_ON(ret);
+		if (ret < 0) {
+			errno = -ret;
+			error(
+"failed to write bytenr %llu length %u devid %llu dev_bytenr %llu: %m",
+				eb->start, eb->len,
+				multi->stripes[dev_nr].dev->devid,
+				eb->dev_bytenr);
+			goto out;
+		}
 	}
+out:
 	kfree(raid_map);
 	kfree(multi);
 	return 0;
Signature · 1 line
2.25.1
#1 lore raw
Quoted 11 lines · Qu Wenruo

All callers of write_and_map_eb(), except btrfs-corrupt-block, have handled error, but inside write_and_map_eb() itself, the only error handling is BUG_ON().

This patch will kill all the BUG_ON()s inside
write_and_map_eb(), and enhance the the caller in
btrfs-corrupt-block() to handle the error.

 btrfs-corrupt-block.c       |  9 ++++++++-
 cmds/rescue-super-recover.c |  3 ++-

So may the subject be prefixed with "btrfs-progs"?

Quoted 25 lines

disk-io.c | 26 +++++++++++++++++++++++--- 3
files changed, 33 insertions(+), 5 deletions(-)

btrfs-corrupt-block.c index renamed from btrfs-corrupt-block.c +0 -0

95df871a7822..3c236e146176 100644 --- a/btrfs-corrupt-block.c +++ b/btrfs-corrupt-block.c @@ -771,8 +771,15 @@ static int corrupt_metadata_block(struct btrfs_fs_info *fs_info, u64 block,

btrfs-corrupt-block.c index renamed from btrfs-corrupt-block.c +0 -1
 		u64 bogus = generate_u64(orig);
 btrfs_set_header_generation(eb, bogus);
-		write_and_map_eb(fs_info, eb); +		ret =

write_and_map_eb(fs_info, eb);

btrfs-corrupt-block.c index renamed from btrfs-corrupt-block.c +1 -0
 		free_extent_buffer(eb);
+		if (ret < 0) { +			errno = -ret; +

fprintf(stderr, + "failed to write extent buffer at
%llu: %m", + eb->start); + return ret; +
}

btrfs-corrupt-block.c index renamed from btrfs-corrupt-block.c +0 -0
 		break; } case BTRFS_METADATA_BLOCK_SHIFT_ITEMS:
(quoted diff) +0 -0

b/cmds/rescue-super-recover.c index 5d6bea836c8b..62f4f7754720 100644 --- a/cmds/rescue-super-recover.c +++ b/cmds/rescue-super-recover.c @@ -276,7 +276,8 @@ int btrfs_recover_superblocks(const char *dname,

(quoted diff) +0 -1
 				  struct super_block_record, list);  root =
 open_ctree(record->device_name, record->bytenr,
-			  OPEN_CTREE_RECOVER_SUPER | OPEN_CTREE_WRITES); +
Quoted 2 lines

OPEN_CTREE_RECOVER_SUPER | OPEN_CTREE_WRITES | +
OPEN_CTREE_EXCLUSIVE);

Any explanation about this change?

Thanks

Quoted 51 lines
cmds/rescue-super-recover.c +0 -0
 	if (!root) {
 		ret = 3;
 		goto no_recover;
disk-io.c +23 -3
index e8a2e4afa93a..9ff62fcd54d1 100644
@@ -487,20 +487,40 @@ int write_and_map_eb(struct btrfs_fs_info *fs_info, struct extent_buffer *eb)
 	length = eb->len;
 	ret = btrfs_map_block(fs_info, WRITE, eb->start, &length,
 			      &multi, 0, &raid_map);
+	if (ret < 0) {
+		errno = -ret;
+		error("failed to map bytenr %llu length %u: %m",
+			eb->start, eb->len);
+		goto out;
+	}
 
 	if (raid_map) {
 		ret = write_raid56_with_parity(fs_info, eb, multi,
 					       length, raid_map);
-		BUG_ON(ret);
+		if (ret < 0) {
+			errno = -ret;
+			error(
+		"failed to write raid56 stripe for bytenr %llu length %llu: %m",
+				eb->start, length);
+			goto out;
+		}
 	} else while (dev_nr < multi->num_stripes) {
-		BUG_ON(ret);
 		eb->fd = multi->stripes[dev_nr].dev->fd;
 		eb->dev_bytenr = multi->stripes[dev_nr].physical;
 		multi->stripes[dev_nr].dev->total_ios++;
 		dev_nr++;
 		ret = write_extent_to_disk(eb);
-		BUG_ON(ret);
+		if (ret < 0) {
+			errno = -ret;
+			error(
+"failed to write bytenr %llu length %u devid %llu dev_bytenr %llu: %m",
+				eb->start, eb->len,
+				multi->stripes[dev_nr].dev->devid,
+				eb->dev_bytenr);
+			goto out;
+		}
 	}
+out:
 	kfree(raid_map);
 	kfree(multi);
 	return 0;
#2 lore raw
Quoted 38 lines · Su Yue
Quoted 11 lines · Qu Wenruo

All callers of write_and_map_eb(), except btrfs-corrupt-block, have handled error, but inside write_and_map_eb() itself, the only error handling is BUG_ON().

This patch will kill all the BUG_ON()s inside
write_and_map_eb(), and enhance the the caller in
btrfs-corrupt-block() to handle the error.

 btrfs-corrupt-block.c       |  9 ++++++++-
 cmds/rescue-super-recover.c |  3 ++-

So may the subject be prefixed with "btrfs-progs"?

Quoted 21 lines

 disk-io.c                   | 26 +++++++++++++++++++++++--- 3 files changed, 33 insertions(+), 5 deletions(-)

btrfs-corrupt-block.c index renamed from btrfs-corrupt-block.c +0 -0

95df871a7822..3c236e146176 100644 --- a/btrfs-corrupt-block.c +++ b/btrfs-corrupt-block.c @@ -771,8 +771,15 @@ static int corrupt_metadata_block(struct btrfs_fs_info *fs_info, u64 block, u64 bogus = generate_u64(orig); btrfs_set_header_generation(eb, bogus);
-        write_and_map_eb(fs_info, eb); +        ret = write_and_map_eb(fs_info, eb);
         free_extent_buffer(eb);
+        if (ret < 0) { +            errno = -ret; + fprintf(stderr, +                "failed to write extent buffer at %llu: %m", +                eb->start); +            return ret; + }
         break; } case BTRFS_METADATA_BLOCK_SHIFT_ITEMS:

(quoted diff) +0 -0

b/cmds/rescue-super-recover.c index 5d6bea836c8b..62f4f7754720 100644 --- a/cmds/rescue-super-recover.c +++ b/cmds/rescue-super-recover.c @@ -276,7 +276,8 @@ int

Quoted 8 lines
Quoted 6 lines

btrfs_recover_superblocks(const char *dname,
                   struct super_block_record, list);  root = open_ctree(record->device_name, record->bytenr,
-              OPEN_CTREE_RECOVER_SUPER | OPEN_CTREE_WRITES); + OPEN_CTREE_RECOVER_SUPER | OPEN_CTREE_WRITES | + OPEN_CTREE_EXCLUSIVE);

Any explanation about this change?

My bad, unclean base.

I'll resend the patch.

Thanks,
Qu

Quoted 56 lines

Thanks

Quoted 52 lines

     if (!root) {
         ret = 3;
         goto no_recover;

disk-io.c +23 -3
index e8a2e4afa93a..9ff62fcd54d1 100644
@@ -487,20 +487,40 @@ int write_and_map_eb(struct btrfs_fs_info
 *fs_info, struct extent_buffer *eb)
      length = eb->len;
      ret = btrfs_map_block(fs_info, WRITE, eb->start, &length,
                    &multi, 0, &raid_map);
+    if (ret < 0) {
+        errno = -ret;
+        error("failed to map bytenr %llu length %u: %m",
+            eb->start, eb->len);
+        goto out;
+    }
 
      if (raid_map) {
          ret = write_raid56_with_parity(fs_info, eb, multi,
                             length, raid_map);
-        BUG_ON(ret);
+        if (ret < 0) {
+            errno = -ret;
+            error(
+        "failed to write raid56 stripe for bytenr %llu length %llu: %m",
+                eb->start, length);
+            goto out;
+        }
      } else while (dev_nr < multi->num_stripes) {
-        BUG_ON(ret);
          eb->fd = multi->stripes[dev_nr].dev->fd;
          eb->dev_bytenr = multi->stripes[dev_nr].physical;
          multi->stripes[dev_nr].dev->total_ios++;
          dev_nr++;
          ret = write_extent_to_disk(eb);
-        BUG_ON(ret);
+        if (ret < 0) {
+            errno = -ret;
+            error(
+"failed to write bytenr %llu length %u devid %llu dev_bytenr %llu: %m",
+                eb->start, eb->len,
+                multi->stripes[dev_nr].dev->devid,
+                eb->dev_bytenr);
+            goto out;
+        }
      }
+out:
      kfree(raid_map);
      kfree(multi);

     return 0;