On Mon, Mar 14, 2016 at 01:48:18PM +0100, Holger Hoffstätte wrote:

On 03/14/16 13:07, Marc Haber wrote:

quoted

quoted

And btrfs balance runs into the same ENOSPC issues as the old one:

... with Qu's patch, I now get a reproducible kernel trace:

<snip>

That is interesting and useful. Sorry if this was asked before, but
did you ever try to clear the free-space cache via -o clear_cache
on mount?

This was not asked, and I didn't try. Since this is an encrypted root
filesystem, is it a workable way to add clear_cache to /etc/fstab,
rebuild initramfs and reboot? Or do you recommend using a rescue system?

Give it a try, let it run for a while and then try balancing
again.

Do I need to wait for clear_cache to finish, like until I see disk
usage dropping?

Uncle Occam's razor also suggests that the involvement of dm
doesn't help. Why not just use the device/partition directly?

I need the dm intermediate since I don't want to repartition the
expensive SSD and the entire system is crypted.

_Someone_ is lying to btrfs in terms of device size and/or allocated
chunks, otherwise you wouldn't get the ENOSPC.

Which properties does a block device report other than size?

Greetings
Marc

-- 
-----------------------------------------------------------------------------
Marc Haber         | "I don't trust Computers. They | Mailadresse im Header
Leimen, Germany    |  lose things."    Winona Ryder | Fon: *49 6224 1600402
Nordisch by Nature |  How to make an American Quilt | Fax: *49 6224 1600421