We've had a number of reports recently of people with NFS and CIFS
mounts that were unable to suspend or hibernate their machines. Here
are a couple of Fedora bugs that illustrate the problem:
https://bugzilla.redhat.com/show_bug.cgi?id=712088
https://bugzilla.redhat.com/show_bug.cgi?id=717735
When it occurs the problem is pretty clear. We have a task that's
sleeping in the kernel in TASK_KILLABLE sleep, generally waiting
for a reply to come in. Often though, userspace has already taken
down the interface so that reply will never come. The process then
fails to freeze and the suspend fails.
This patch fixes this by allowing the TASK_KILLABLE sleeps in NFS and
CIFS to be awoken by the freezer and then to try to freeze. If a freeze
event does occur, then the code will treat it as if a schedule() has
already occured.
I've had a number of people test versions of this set and the results
have been positive. Part of the problem though is that there are
sometimes still unrelated problems with suspend/resume that seem to
cloud the results. As best we can tell though, this patchset does work
as advertised.
Jeff Layton (4):
freezer: make fake_signal_wake_up wake TASK_KILLABLE tasks too
cifs, freezer: add wait_event_freezekillable and have cifs use it
sunrpc: make rpc_wait_bit_killable handle freeze events
nfs: make TASK_KILLABLE sleeps attempt to freeze
fs/cifs/transport.c | 3 ++-
fs/nfs/inode.c | 4 +++-
fs/nfs/nfs3proc.c | 3 ++-
fs/nfs/nfs4proc.c | 12 ++++++++----
fs/nfs/proc.c | 3 ++-
include/linux/freezer.h | 19 +++++++++++++++++--
kernel/freezer.c | 2 +-
net/sunrpc/sched.c | 4 +++-
8 files changed, 38 insertions(+), 12 deletions(-)
--
1.7.6.2
--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
