Re: tmpfs race conditions

[Joerg Sonnenberger <joerg%britannica.bec.de@localhost>]

On Mon, Oct 12, 2015 at 09:40:09PM +0200, Joerg Sonnenberger wrote:
> I have a build machine building pkgsrc where I see some very strange
> issues. The individual chroots share a read-only null-mounted base
> directory with a number of tmpfs instances on top for most of the
> writeable parts. Symptoms of the some data corruption or race
> conditions:

I've been trying to use MFS for the WRKOBJDIR a bit this week. I had one
run that was relatively stable. During that build, I've seen the
checksum issue and LZMA integrity errors during pkg_add. That set of
problems is therefore not specific to tmpfs. Another issue and the
reason why using MFS is not really an option is that it tends to dead
lock in umount. The umount is kind of needed to keep the memory use in
check -- discarded blocks don't result in discarded pages :(

Joerg