[linux-cifs-client] Re: [PATCH 00/10] cifs: implement proper hardlink

[Steve French]

On Thu, Jun 25, 2009 at 10:57 AM, Steve French<smfrench at gmail.com> wrote:
> Since getting back from vacation have reviewed and merged about 1/2 of
> the backlog.   These are next, but also would like to run a few more
> tests on them since they may affect performance.

On a loosely related topic to cifs inode handling ...
During the week of testing in Redmond earlier in the month, we had a
long discussion about the best way to handle caching/oplock.  It looks
like requesting "batch" oplock first (which allows you to defer
close operations) may be preferable, but doing this would require changing
the cifs destroy inode routine to send SMB close (would the vfs allow
the free inode routine to block?) and would allow us to use the same
network file handle for multiple opens in some case which would
greatly improve performance.



-- 
Thanks,

Steve