Clean up: The asynchronous path of ro_unmap_safe is the same for all
registration modes, so just move it into the only function that
calls it.
"Safe" disposal is done because attempting to invalidate an MR with
a memory key from the previous connection epoch can result in a
memory management operation error for some registration modes, and
that will kill the current connection.
Signed-off-by: Chuck Lever <chuck.lever@xxxxxxxxxx>
---
net/sunrpc/xprtrdma/transport.c | 8 ++++++--
1 file changed, 6 insertions(+), 2 deletions(-)
Hi Anna-
This patch replaces 2/3 in the for-4.15 series I sent a few days
ago.
diff --git a/net/sunrpc/xprtrdma/transport.c b/net/sunrpc/xprtrdma/transport.c
index 8cf5ccf..aa5cab2 100644
--- a/net/sunrpc/xprtrdma/transport.c
+++ b/net/sunrpc/xprtrdma/transport.c
@@ -727,8 +727,12 @@
goto drop_connection;
/* On retransmit, remove any previously registered chunks */
- if (unlikely(!list_empty(&req->rl_registered)))
- r_xprt->rx_ia.ri_ops->ro_unmap_safe(r_xprt, req, false);
+ while (unlikely(!list_empty(&req->rl_registered))) {
+ struct rpcrdma_mw *mw;
+
+ mw = rpcrdma_pop_mw(&req->rl_registered);
+ rpcrdma_defer_mr_recovery(mw);
+ }
rc = rpcrdma_marshal_req(r_xprt, rqst);
if (rc < 0)
--
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
