[linux-cifs-client] [PATCH 0/4] [CIFS] clean up cifsd thread
startup and shutdown
Jeff Layton
jlayton at redhat.com
Fri May 2 19:25:16 GMT 2008
Hi Steve,
Shirish ran across a problem when testing a RHEL4.7 beta kernel where
kthread_stop was called on a cifsd that had already exited. Looking at
the code, it looks like this is actually a problem in mainline. There is
no guarantee that cifsd will actually be up when we call kthread_stop on
it.
This patchset is intended to try and close this race and also to do some
other small cleanups to the startup/shutdown of cifs_demultiplex_thread.
For now, I've dropped the patch that makes cifsd use non-blocking I/O.
It works and I still would like to see us do it, but I think we should
probably treat that piece independently of the rest of this set.
While this is being sent as a set, the patches are actually relatively
independent of one another. The first 3 patches are cleanup patches.
The 4th patch is what actually closes the race.
I've tested this primarily by running iozone tests on it with the -U
option to mount and unmount the filesystems between each test, and
it seems to be OK.
As always, comments and suggestions are appreciated...
Signed-off-by: Jeff Layton <jlayton at redhat.com>
More information about the linux-cifs-client
mailing list