RE: trim not working and irreparable errors from btrfsck

[lutz.euler@freenet.de (Lutz Euler)]

Hi Christian, Paul and Austin,

Christian wrote:
> However, fstrim still gives me "0 B (0 bytes) trimmed, so that may be
> another problem. Is there a way to check if trim works?

Paul wrote:
> I've got the same problem. I've got 2 SSDs with 2 partitions in RAID1,
> fstrim always works on the 2nd partition but not the first. There are
> no errors on either filesystem that I know of, but the first one is
> root so I can't take it offline to run btrfs check.

Austin wrote:
> I'm seeing the same issue here, but with a Crucial brand SSD.
> Somewhat interestingly, I don't see any issues like this with BTRFS on
> top of LVM's thin-provisioning volumes, or with any other filesystems,
> so I think it has something to do with how BTRFS is reporting unused
> space or how it is submitting the discard requests.

Probably you all suffer from the same problem I had a few years ago.
It is a bug in how btrfs implements fstrim.

To check whether you are a victim of this bug simply run:

# btrfs-debug-tree /dev/whatever | grep 'FIRST_CHUNK_TREE CHUNK_ITEM'

where /dev/whatever is a device of your filesystem, and interrupt after
the first several output lines with C-c. (Officially the filesystem
should be unmounted when running btrfs-debug-tree, but that is not
necessary as we only read from it and the relevant data doesn't change
very often.)

You get something like:

  item 2 key (FIRST_CHUNK_TREE CHUNK_ITEM 0)
  item 3 key (FIRST_CHUNK_TREE CHUNK_ITEM 12947816448)
  item 4 key (FIRST_CHUNK_TREE CHUNK_ITEM 14021558272)
  ...

(This output is from an old version of btrfs-progs. I understand newer
version are more verbose, but you should nevertheless easily be able to
interpret the output).

If the first number different from 0 (here, the 12947816448) is larger
than the sum of the sizes of the devices the filesystem consists of,
bingo.

This has been discussed already in the past and there is a patch.

Please see for the patch:
http://www.mail-archive.com/linux-btrfs@vger.kernel.org/msg40618.html

and for the background:
http://comments.gmane.org/gmane.comp.file-systems.btrfs/15597

Kind regards,

Lutz Euler
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in