On 2015-08-28 11:10, J - wrote:
> I have noticed a BTRFS error mentioned in two consecutive identical entries  in my kernel log:
> "BTRFS error (device sda2): bad extent! em: [0 0] passed [0 4096]"
>
> sda2 contains a btrfs with skinny extents has been created a few days ago and contains a few subvolumes (mounted as root and home) and subsubvolumes a few gigabites of data in smallish files. It is mounted with 'rw,relatime,compress-force=lzo,ssd,space_cache',
>
> Shortly after noticing the errors, which cannot have been long after their occurrence, I rebooted and ran 'btrfs check' on the filesystem, with output as follows:
>
> 'Checking filesystem on /dev/sda2
> UUID: c2c63f52-a322-413f-84f6-b4106fdff8c1
> found 16882602001 bytes used err is 0
> total csum bytes: 15775044
> total tree bytes: 610959360
> total fs tree bytes: 563019776
> total extent tree bytes: 28459008
> btree space waste bytes: 85192798
> file data blocks allocated: 19913826304
>   referenced 24478576640
> btrfs-progs v4.1.2'
>
> After this I ran 'btrfs scrub' which reported no errors.
>
> The filesystem has been created (and also checked) with Arch Linux, kernel 4.1.6, btrfs-progs 4.1.2.
> The computer has a AMD FX-8350CPU, 32GB ECC ram, a 990fx chipset and the drive is a OCZ Vector 180 (GPT, 2 partition, ESP from 2048s to 1050623s, .linux partition from 1050624s to 468860175s)
I would suggest some serious testing of that OCZ SSD, they are known to 
have data corruption issues under heavy load or across unclean 
shutdowns, and these issues will cause problems with any filesystem you 
put on it.  Personally I would not trust an OCZ SSD as anything other 
than scratch space for testing, and even then would only do so cautiously.