broken superblocks

broken superblocks

[Christian Switaiski]

During a system deadlock I was forced to hard shutdown my system while a  
`btrfs balance` was ongoing and ctrl+c & `btrfs balance cancel` failed
Since the restart the partition is broken, the kernel fails to mounts it  
and btrfs' repair tools exit cause of errors w/o doing anything. The  
partition is 20GB, and so got 2 superblocks: both are broken :/



kernel is 3.11.5, btrfs-progs are git ones & distro is gentoo
------------ btrfs output start ------------

# mount /dev/sda2 /mnt/repair
mount: wrong fs type, bad option, bad superblock on /dev/sda2,
        missing codepage or helper program, or other error
        In some cases useful info is found in syslog - try
        dmesg | tail or so

# dmesg | tail
[279861.340796] device label root-x64 devid 1 transid 535462 /dev/sda2
[279861.341416] btrfs: disk space caching is enabled
[279861.342223] attempt to access beyond end of device     <------
[279861.342226] sda2: rw=32, want=44040448, limit=41943040 <------
[279861.342236] attempt to access beyond end of device
[279861.342237] sda2: rw=32, want=44302592, limit=41943040
[279861.347811] btrfs: open_ctree failed

# btrfs-show-super /dev/sda2
superblock: bytenr=65536, device=/dev/sda2
---------------------------------------------------------
csum                    0xbd53d63e [match]
bytenr                  65536
...
total_bytes             21474836480
bytes_used              18661335040
...
dev_item.total_bytes    21474836480
dev_item.bytes_used     22032678912 <--- beyond part size!
...

# btrfs fi show
Label: 'root-x64'  uuid: f87210e5-47a9-44af-9797-6afea2cdaae8
         Total devices 1 FS bytes used 17.38GB
         devid    1 size 20.00GB used 20.52GB path /dev/sda2   <--- again  
beyond partsize
Btrfs v0.20-rc1-358-g194aa4a

# btrfsck /dev/sda2
Check tree block failed, want=39317532672, have=0
read block failed check_tree_block
Couldn't setup extent tree
Checking filesystem on /dev/sda2
UUID: f87210e5-47a9-44af-9797-6afea2cdaae8
Critical roots corrupted, unable to fsck the FS

# btrfs restore /dev/sda2 /tmp/fix
Check tree block failed, want=39317532672, have=0
read block failed check_tree_block
Couldn't setup extent tree
Check tree block failed, want=39317532672, have=0
read block failed check_tree_block
Couldn't read fs root: -5

# btrfs-find-root /dev/sda2
Super think's the tree root is at 29556736, chunk root 40659582976
Found tree root at 29556736 gen 535462 level 1

btrfs-debug-tree & `btrfs chunk-recover -v` output a lot stuff but at the  
end both fail:
# btrfs-debug-tree /dev/sda2
...
Check tree block failed, want=39317532672, have=0
read block failed check_tree_block
btrfs-debug-tree: btrfs-debug-tree.c:237: main: Assertion `!(ret < 0)'  
failed.
Aborted

# btrfs chunk-recover -v /dev/sda2
...
Total Chunks:   31
   Heathy:       29
   Bad:  2
Orphan Block Groups:
   Block Group: start = 37710987264, len = 268435456, flag = 1
   Block Group: start = 37979422720, len = 230686720, flag = 1
   Block Group: start = 38210109440, len = 33554432, flag = 1
Orphan Device Extents:
   Device extent: devid = 1, start = 3527409664, len = 268435456, chunk  
offset = 37710987264
Check tree block failed, want=39317532672, have=0
read block failed check_tree_block
Couldn't setup extent tree
open with broken chunk error
Fail to recover the chunk tree.

------------ btrfs output end ------------



Cause btrfs-debug-tree & `btrfs chunk-recover` say there are still chunks,  
I have the hope most data is still intact, only btrfs & tools stumble on  
some invalid data. Is there any hope to e.g. force btrfs to clamp all data?

Thanks,
Christian Switaiski


PS: I already tried to patch btrfsprogs to not exit after the errors  
occurred (it helped with a problem in the past I had) but it just causes  
segfaults :/
Also I am bit confused that none of the btrfs progs show a 'beyond  
anything' msg like the kernel.