recent notify_kernel.c changes - needed for oplock_*.c?
jra at samba.org
Thu Jun 6 16:58:01 GMT 2002
On Thu, Jun 06, 2002 at 01:23:10PM +1000, Chris Pascoe wrote:
> Would a similar change be needed for the kernel oplocks case? I'm still
> seeing major timeouts with oplocks under heavy oplock load with current
> CVS, particularly in the case where a transaction across NFS (in my case
> it was a cron job set up by someone to back up a directory periodically)
> causes an oplock interrupt on .mdb/.ldb (MS Access) files.
> I've been looking through the oplock.c/oplock_linux.c, and it seems that
> even if we form a small queue of oplocks there may be some other races
> there too - but perhaps I am missing the way it fits together.
Ok - I've now committed a fix for this in SAMBA_2_2 and HEAD. It could
do with some testing (but it seems to pass everything here).
More information about the samba-technical