| Message ID | f301a88d04e1929a313c458bd8ce1218903b648a.1743183579.git.trond.myklebust@hammerspace.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | [1/2] SUNRPC: Don't allow waiting for exiting tasks | expand |
On Fri, 2025-03-28 at 13:40 -0400, trondmy@kernel.org wrote: > From: Trond Myklebust <trond.myklebust@hammerspace.com> > > Once a task calls exit_signals() it can no longer be signalled. So do > not allow it to do killable waits. > > Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com> > --- > net/sunrpc/sched.c | 2 ++ > 1 file changed, 2 insertions(+) > > diff --git a/net/sunrpc/sched.c b/net/sunrpc/sched.c > index 9b45fbdc90ca..73bc39281ef5 100644 > --- a/net/sunrpc/sched.c > +++ b/net/sunrpc/sched.c > @@ -276,6 +276,8 @@ EXPORT_SYMBOL_GPL(rpc_destroy_wait_queue); > > static int rpc_wait_bit_killable(struct wait_bit_key *key, int mode) > { > + if (unlikely(current->flags & PF_EXITING)) > + return -EINTR; > schedule(); > if (signal_pending_state(mode, current)) > return -ERESTARTSYS; Won't this mean that if a task is signalled and does a final fput, that a CLOSE sent in task_work will never get sent?
On Fri, 2025-03-28 at 13:53 -0400, Jeff Layton wrote: > On Fri, 2025-03-28 at 13:40 -0400, trondmy@kernel.org wrote: > > From: Trond Myklebust <trond.myklebust@hammerspace.com> > > > > Once a task calls exit_signals() it can no longer be signalled. So > > do > > not allow it to do killable waits. > > > > Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com> > > --- > > net/sunrpc/sched.c | 2 ++ > > 1 file changed, 2 insertions(+) > > > > diff --git a/net/sunrpc/sched.c b/net/sunrpc/sched.c > > index 9b45fbdc90ca..73bc39281ef5 100644 > > --- a/net/sunrpc/sched.c > > +++ b/net/sunrpc/sched.c > > @@ -276,6 +276,8 @@ EXPORT_SYMBOL_GPL(rpc_destroy_wait_queue); > > > > static int rpc_wait_bit_killable(struct wait_bit_key *key, int > > mode) > > { > > + if (unlikely(current->flags & PF_EXITING)) > > + return -EINTR; > > schedule(); > > if (signal_pending_state(mode, current)) > > return -ERESTARTSYS; > > Won't this mean that if a task is signalled and does a final fput, > that > a CLOSE sent in task_work will never get sent? It should mean that the close gets sent, but the task won't wait for completion.
On Fri, 2025-03-28 at 18:00 +0000, Trond Myklebust wrote: > On Fri, 2025-03-28 at 13:53 -0400, Jeff Layton wrote: > > On Fri, 2025-03-28 at 13:40 -0400, trondmy@kernel.org wrote: > > > From: Trond Myklebust <trond.myklebust@hammerspace.com> > > > > > > Once a task calls exit_signals() it can no longer be signalled. So > > > do > > > not allow it to do killable waits. > > > > > > Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com> > > > --- > > > net/sunrpc/sched.c | 2 ++ > > > 1 file changed, 2 insertions(+) > > > > > > diff --git a/net/sunrpc/sched.c b/net/sunrpc/sched.c > > > index 9b45fbdc90ca..73bc39281ef5 100644 > > > --- a/net/sunrpc/sched.c > > > +++ b/net/sunrpc/sched.c > > > @@ -276,6 +276,8 @@ EXPORT_SYMBOL_GPL(rpc_destroy_wait_queue); > > > > > > static int rpc_wait_bit_killable(struct wait_bit_key *key, int > > > mode) > > > { > > > + if (unlikely(current->flags & PF_EXITING)) > > > + return -EINTR; > > > schedule(); > > > if (signal_pending_state(mode, current)) > > > return -ERESTARTSYS; > > > > Won't this mean that if a task is signalled and does a final fput, > > that > > a CLOSE sent in task_work will never get sent? > > It should mean that the close gets sent, but the task won't wait for > completion. > Good enough, I guess. Reviewed-by: Jeff Layton <jlayton@kernel.org>
On Fri, 2025-03-28 at 14:09 -0400, Jeff Layton wrote: > On Fri, 2025-03-28 at 18:00 +0000, Trond Myklebust wrote: > > On Fri, 2025-03-28 at 13:53 -0400, Jeff Layton wrote: > > > On Fri, 2025-03-28 at 13:40 -0400, trondmy@kernel.org wrote: > > > > From: Trond Myklebust <trond.myklebust@hammerspace.com> > > > > > > > > Once a task calls exit_signals() it can no longer be signalled. > > > > So > > > > do > > > > not allow it to do killable waits. > > > > > > > > Signed-off-by: Trond Myklebust > > > > <trond.myklebust@hammerspace.com> > > > > --- > > > > net/sunrpc/sched.c | 2 ++ > > > > 1 file changed, 2 insertions(+) > > > > > > > > diff --git a/net/sunrpc/sched.c b/net/sunrpc/sched.c > > > > index 9b45fbdc90ca..73bc39281ef5 100644 > > > > --- a/net/sunrpc/sched.c > > > > +++ b/net/sunrpc/sched.c > > > > @@ -276,6 +276,8 @@ EXPORT_SYMBOL_GPL(rpc_destroy_wait_queue); > > > > > > > > static int rpc_wait_bit_killable(struct wait_bit_key *key, int > > > > mode) > > > > { > > > > + if (unlikely(current->flags & PF_EXITING)) > > > > + return -EINTR; > > > > schedule(); > > > > if (signal_pending_state(mode, current)) > > > > return -ERESTARTSYS; > > > > > > Won't this mean that if a task is signalled and does a final > > > fput, > > > that > > > a CLOSE sent in task_work will never get sent? > > > > It should mean that the close gets sent, but the task won't wait > > for > > completion. > > > > Good enough, I guess. > > Reviewed-by: Jeff Layton <jlayton@kernel.org> > It ensures that the task can die, and with the actual close RPC call being asynchronous, it will keep going until either the server comes back and processes it, or someone force umounts the partition and kills the call, etc.
diff --git a/net/sunrpc/sched.c b/net/sunrpc/sched.c index 9b45fbdc90ca..73bc39281ef5 100644 --- a/net/sunrpc/sched.c +++ b/net/sunrpc/sched.c @@ -276,6 +276,8 @@ EXPORT_SYMBOL_GPL(rpc_destroy_wait_queue); static int rpc_wait_bit_killable(struct wait_bit_key *key, int mode) { + if (unlikely(current->flags & PF_EXITING)) + return -EINTR; schedule(); if (signal_pending_state(mode, current)) return -ERESTARTSYS;