[dm-devel] refrigerator in dm-daemon

Christophe Saout christophe at saout.de
Mon Dec 22 12:09:02 UTC 2003


Hi Joe,

> #if 0
>         /* FIXME: not convinced by this */
>         if (current->flags & PF_FREEZE)
>             refrigerator(PF_IOTHREAD);
> #endif

I've got a notebook some time ago and just checked:

Without this refrigerator thing suspend to ram fails:

Dec 22 18:13:11 leto PM: Preparing system for suspend
Dec 22 18:13:18 leto Stopping tasks: ===================================
Dec 22 18:13:18 leto stopping tasks failed (1 tasks remaining)
Dec 22 18:13:18 leto Restarting tasks...<6> Strange, kcopyd not stopped
Dec 22 18:13:18 leto done

After that kcopyd is eating 100% CPU time. If I remove the #if 0
everything just works fine.

BTW, I'm using the snapshot code in 2.6 on a production system for some
time now (backups using rsync every night) and it's working great. I
don't think I'm running into corner cases too much though. I know that
without the vfs locking it's not too reliable but I'm running sync
before creating the snapshots and there's nobody on the machine in the
night so it's ok for me. Since the snapshots are writeable the fs
journal can be replayed on mounting.

When do you think some of the code is ready?

Another thing: There still seems to be trouble with the loop driver, the
block device backend support seems to be broken. There are some patches
but they in turn seem to introduce other problems... I've still my
dm-crypt target floating around. With it there would be a working
alternative. What did Andrew tell you when you asked him? I talked with
one of the people that are working on the cryptoloop support and he
agreed that device-mapper would be a better place for this.

--
Christophe Saout <christophe at saout.de>
Please avoid sending me Word or PowerPoint attachments.
See http://www.fsf.org/philosophy/no-word-attachments.html





More information about the dm-devel mailing list