[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: writing processes are blocking in log_wait_common with data=ordered



Hi,

On Fri, May 03, 2002 at 06:40:18PM +0100, Stephen C. Tweedie wrote:

> So 2.4.18 is broken too, but so much less so that it doesn't show up
> in benchmarks.  The 2.4.9 fix should fix 2.4.18 too, and will let us
> remove the above performance degradation.  The patch I've been testing
> for 2.4.9 is attached.

Bah.  I'm trying to put this into ext3 cvs, and it's down:

ssh: cvs.gkernel.sourceforge.net: Name or service not known

(ssh itself does work, just cvs is down.)

I've got two other fixes I want to get in quickly: one is the tune2fs
fix, because we found a bug in blockdev-in-pagecache which made it
evident that we see locked buffers in ext3 even if we're just
colliding with a user-space read of the block device (ie. dump(8) can
corrupt and panic your fs), and another subtle one I just discovered
very recently where there's a race between __journal_file_buffer and
bdflush.

But cvs is down...

Andrew, would you object greatly to using bk for ext3 if we can't rely
on gkernel?  In the mean time I'll set up a patch ball with the fixes
so we've got an alternative to cvs for testing.  Probably won't be
done until Monday, I'm about to quit for the weekend.

Cheers,
 Stephen





[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]