From:   Trond Myklebust <trondmy@gmail.com>
To:     Chuck Lever <chuck.lever@oracle.com>
Cc:     linux-nfs@vger.kernel.org
Subject: [PATCH 2/4] SUNRPC: Fix the RPCSEC_GSS sequence semantics after request re-encoding
Date:   Wed,  2 Jan 2019 17:53:11 -0500
Message-Id: <20190102225313.2097-3-trond.myklebust@hammerspace.com>
In-Reply-To: <20190102225313.2097-2-trond.myklebust@hammerspace.com>
References: <20190102225313.2097-1-trond.myklebust@hammerspace.com>
 <20190102225313.2097-2-trond.myklebust@hammerspace.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Sender: linux-nfs-owner@vger.kernel.org
Precedence: bulk

When we have to re-encode the RPCSEC_GSS request because it has fallen
outside the GSS sequence window, we want to avoid moving it to the front
of the transmission queue, since that moves the entire sequence window
forward, and may result in more requests falling out.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
 net/sunrpc/xprt.c | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/net/sunrpc/xprt.c b/net/sunrpc/xprt.c
index 9075ae150ae5..040c45c68fb3 100644
--- a/net/sunrpc/xprt.c
+++ b/net/sunrpc/xprt.c
@@ -1322,6 +1322,12 @@ xprt_request_transmit(struct rpc_rqst *req, struct rpc_task *snd_task)
 		}
 		/* Verify that our message lies in the RPCSEC_GSS window */
 		if (rpcauth_xmit_need_reencode(task)) {
+			if (xprt->ops->release_request &&
+			    xprt_request_retransmit_after_disconnect(task)) {
+				spin_lock_bh(&xprt->transport_lock);
+				xprt->ops->release_request(task);
+				spin_unlock_bh(&xprt->transport_lock);
+			}
 			status = -EBADMSG;
 			goto out_dequeue;
 		}
-- 
2.20.1