On Wed, 2022-04-20 at 14:21 -0400, Olga Kornievskaia wrote:
> From: Olga Kornievskaia <kolga@xxxxxxxxxx>
>
> A relocated task must release its previous transport.
>
> Fixes: 82ee41b85cef1 ("SUNRPC don't resend a task on an offlined
> transport")
> Signed-off-by: Olga Kornievskaia <kolga@xxxxxxxxxx>
> ---
> net/sunrpc/clnt.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c
> index af0174d7ce5a..95de454a858b 100644
> --- a/net/sunrpc/clnt.c
> +++ b/net/sunrpc/clnt.c
> @@ -1067,8 +1067,10 @@ void rpc_task_set_transport(struct rpc_task
> *task, struct rpc_clnt *clnt)
> {
> if (task->tk_xprt &&
> !(test_bit(XPRT_OFFLINE, &task->tk_xprt-
> >state) &&
> - (task->tk_flags & RPC_TASK_MOVEABLE)))
> + (task->tk_flags & RPC_TASK_MOVEABLE))) {
> + xprt_release_write(task->tk_xprt, task);
> return;
> + }
> if (task->tk_flags & RPC_TASK_NO_ROUND_ROBIN)
> task->tk_xprt = rpc_task_get_first_xprt(clnt);
> else
I don't think this patch is correct. It is changing the case where
we're not changing the value of task->tk_xprt, which was never a
problem before commit 82ee41b85cef.
What needs to change is the case where task->tk_xprt is set, but we're
going to change its value anyway because of the tests for XPRT_OFFLINE
and RPC_TASK_MOVEABLE.
In that case we need to call xprt_release(), followed by xprt_put().
Otherwise we're leaking slots, locks and references to the old task-
>tk_xprt.
--
Trond Myklebust
Linux NFS client maintainer, Hammerspace
trond.myklebust@xxxxxxxxxxxxxxx
