Jeff Layton writes via Kernel.org Bugzilla:
I looked at this with Chuck the other day. As far as wait_var_event() getting stuck, I think all that would take is for nfsd4_cb_sequence_done() to continually set cb_need_restart on every call. That would cause the callback to not be destroyed and to never call nfsd41_cb_inflight_end().
That happens in the need_restart: label in nfsd4_cb_sequence_done. These two cases goto that:
if (!clp->cl_minorversion) {
/*
* If the backchannel connection was shut down while this
* task was queued, we need to resubmit it after setting up
* a new backchannel connection.
*
* Note that if we lost our callback connection permanently
* the submission code will error out, so we don't need to
* handle that case here.
*/
if (RPC_SIGNALLED(task))
goto need_restart;
return true;
}
if (cb->cb_held_slot < 0)
goto need_restart;
It doesn't seem likely that it somehow lost the slot, so my guess is that the RPC task is continually returning with RPC_SIGNALLED() set.
Question for Baptiste -- what NFS versions are your clients using?
View: https://bugzilla.kernel.org/show_bug.cgi?id=219710#c6
You can reply to this message to join the discussion.
--
Deet-doot-dot, I am a bot.
Kernel.org Bugzilla (bugspray 0.1-dev)
