As Alan Stern pointed out yesterday, cifs queues most of its workqueue jobs to the system_nrt_wq. There are a couple of problems with that: 1/ That workqueue isn't freezable. Some workqueue jobs from cifs (like oplock breaks) can take quite some time to run since they have to kickoff writeback. That can end up making the freezer give up in some cases. 2/ We queue the async write completion handler to this codepath. Since that can be called in memory reclaim codepaths, we really ought to be queueing it to a workqueue with WQ_MEM_RECLAIM set. This patch fixes this by allocating a new workqueue to which we'll queue all the cifs jobs. The first two patches in the series do a little cleanup work in the exit_cifs code, and the last patch adds the new workqueue and has the callers queue their work to it. This patchset is probably best suited for 3.4. Jeff Layton (3): cifs: clean up call to cifs_dfs_release_automount_timer() cifs: clean up ordering in exit_cifs cifs: add new cifsiod_wq workqueue fs/cifs/cifsfs.c | 25 +++++++++++++++++-------- fs/cifs/cifsglob.h | 1 + fs/cifs/cifsproto.h | 6 ++++++ fs/cifs/cifssmb.c | 4 ++-- fs/cifs/connect.c | 8 ++++---- fs/cifs/misc.c | 2 +- 6 files changed, 31 insertions(+), 15 deletions(-) -- 1.7.7.6 -- To unsubscribe from this list: send the line "unsubscribe linux-cifs" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
