Re: btrfsck crashes

[Anand Jain <Anand.Jain@oracle.com>]

> What I have seen: buf is "0", after read_tree_block.

  Yes since we not checking extent_buffer_uptodate for the csum_root_tree,
  that will pass the null buf, The following patch will avoid sending null
  buffer
    https://patchwork.kernel.org/patch/1148831/
  
  However whether --init-csum-tree will build the good csum I think that
  will still depends on the corruption IMO.
  
-Anand


On 09/07/12 00:08, Christian Volkmann wrote:
> Hi there,
>
> I have a corrupted filesystem. This filesystem crashes btrfsck.
>
> A gdb anaylsis showed me:
> (gdb) bt
> #0 0x0000000000402379 in btrfs_header_nritems (eb=0x0) at ctree.h:1426
> #1 0x0000000000408c14 in run_next_block (root=0x73fb40, bits=0x740d50, bits_nr=1024, last=0x7fffffffd948, pending=0x7fffffffda40,
> seen=0x7fffffffda50, reada=0x7fffffffda30, nodes=0x7fffffffda20, extent_cache=0x7fffffffda60) at btrfsck.c:2512
> #2 0x00000000004099e2 in check_extents (root=0x73fb40) at btrfsck.c:2792
> #3 0x0000000000409bec in main (ac=1, av=0x7fffffffdbe8) at btrfsck.c:2853
>
> What I have seen: buf is "0", after read_tree_block.
>
> btrfsck.c:2511 buf = read_tree_block(root, bytenr, size, 0);
> 2512 nritems = btrfs_header_nritems(buf);
>
> So ctree.h crashes here with btrfs_header_nritems(buf)
> ...
> static inline u##bits btrfs_##name(struct extent_buffer *eb) \
> { \
> struct btrfs_header *h = (struct btrfs_header *)eb->data; \
> return le##bits##_to_cpu(h->member); \
> } \
> ...
>
> I expect an error "eb == 0" is not covered by ctree.h.
> May be another fix is required. E.g. harden btrfsck against "0".
>
> The file system crashes the kernel on some access. I did not follow up this,
> cause the file system is corrupt.( Using openSUSE Tumbleweed 3.4.4-31-desktop)
> May be the kernel code requires also checks for this?
>
> Please contact me, if I should do some further tests with this file system
> or use some tools for a fix test. (developer knowledge given)
>
> Another minor issue: btrfsck uses much memory. But this might be normal.
> ( > 800MB)
>
> Best regards,
> Christian
>
>
>
> PS: Just if anyone is interested:
> - History + tried: openSUSE btrfsck showed the messages below in the first step.
> - /sbin/btrfsck /dev/md3 --repair removed some messages, except checksum.
> - File system is mounted with:
> /backup btrfs defaults,compress=zlib,noatime 1 2
> - filesystem is used to back up some unix system with heavy usage of:
> rsync -aH .... --link-dest=...
> So each file should have regular multiple hard links.
>
> ===
> Is there anybody interested in fixing this file system with me,
> to check btrfsck speedy:/home/cv # /sbin/btrfsck /dev/md3
> checking extents
> checksum verify failed on 2327654400 wanted 73CDE79C found 72
> checksum verify failed on 2327654400 wanted 73CDE79C found 72
> checksum verify failed on 2327654400 wanted 73CDE79C found 72
> checksum verify failed on 2327654400 wanted 73CDE79C found 72
> Csum didn't match
> owner ref check failed [2327654400 4096]
> ref mismatch on [101138354176 98304] extent item 1, found 0
> Incorrect local backref count on 101138354176 root 5 owner 1867898 offset 0 found 0 wanted 1 back 0x1f076d0
> backpointer mismatch on [101138354176 98304]
> owner ref check failed [101138354176 98304]
> ref mismatch on [101138452480 106496] extent item 1, found 0
> Incorrect local backref count on 101138452480 root 5 owner 1867899 offset 0 found 0 wanted 1 back 0x6aa85d0
> backpointer mismatch on [101138452480 106496]
> owner ref check failed [101138452480 106496]
> ref mismatch on [101138558976 8192] extent item 1, found 0
> Incorrect local backref count on 101138558976 root 5 owner 1867901 offset 0 found 0 wanted 1 back 0x6aa8610
> backpointer mismatch on [101138558976 8192]
> owner ref check failed [101138558976 8192]
> ref mismatch on [101138567168 16384] extent item 1, found 0
> Incorrect local backref count on 101138567168 root 5 owner 1867902 offset 0 found 0 wanted 1 back 0x1f8fa80
> backpointer mismatch on [101138567168 16384]
> owner ref check failed [101138567168 16384]
> ref mismatch on [101138583552 16384] extent item 1, found 0
> Incorrect local backref count on 101138583552 root 5 owner 1867903 offset 0 found 0 wanted 1 back 0x1f8fac0
> backpointer mismatch on [101138583552 16384]
> owner ref check failed [101138583552 16384]
> Errors found in extent allocation tree
> checking fs roots
> checksum verify failed on 2327654400 wanted 73CDE79C found 72
> checksum verify failed on 2327654400 wanted 73CDE79C found 72
> checksum verify failed on 2327654400 wanted 73CDE79C found 72
> checksum verify failed on 2327654400 wanted 73CDE79C found 72
> Csum didn't match
> Speicherzugriffsfehler
> --
> To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html