diff options
authorChuck Lever <>2017-12-04 14:04:04 -0500
committerAnna Schumaker <>2017-12-15 14:31:50 -0500
commitccede7598588ae344143f82fb763912535648d58 (patch)
parentc156618e15101a9cc8c815108fec0300a0ec6637 (diff)
xprtrdma: Spread reply processing over more CPUs
Commit d8f532d20ee4 ("xprtrdma: Invoke rpcrdma_reply_handler directly from RECV completion") introduced a performance regression for NFS I/O small enough to not need memory registration. In multi- threaded benchmarks that generate primarily small I/O requests, IOPS throughput is reduced by nearly a third. This patch restores the previous level of throughput. Because workqueues are typically BOUND (in particular ib_comp_wq, nfsiod_workqueue, and rpciod_workqueue), NFS/RDMA workloads tend to aggregate on the CPU that is handling Receive completions. The usual approach to addressing this problem is to create a QP and CQ for each CPU, and then schedule transactions on the QP for the CPU where you want the transaction to complete. The transaction then does not require an extra context switch during completion to end up on the same CPU where the transaction was started. This approach doesn't work for the Linux NFS/RDMA client because currently the Linux NFS client does not support multiple connections per client-server pair, and the RDMA core API does not make it straightforward for ULPs to determine which CPU is responsible for handling Receive completions for a CQ. So for the moment, record the CPU number in the rpcrdma_req before the transport sends each RPC Call. Then during Receive completion, queue the RPC completion on that same CPU. Additionally, move all RPC completion processing to the deferred handler so that even RPCs with simple small replies complete on the CPU that sent the corresponding RPC Call. Fixes: d8f532d20ee4 ("xprtrdma: Invoke rpcrdma_reply_handler ...") Signed-off-by: Chuck Lever <> Signed-off-by: Anna Schumaker <>
4 files changed, 5 insertions, 6 deletions
diff --git a/net/sunrpc/xprtrdma/rpc_rdma.c b/net/sunrpc/xprtrdma/rpc_rdma.c
index ed34dc0f144c..a3f2ab283aeb 100644
--- a/net/sunrpc/xprtrdma/rpc_rdma.c
+++ b/net/sunrpc/xprtrdma/rpc_rdma.c
@@ -1408,11 +1408,7 @@ void rpcrdma_reply_handler(struct rpcrdma_rep *rep)
dprintk("RPC: %s: reply %p completes request %p (xid 0x%08x)\n",
__func__, rep, req, be32_to_cpu(rep->rr_xid));
- if (list_empty(&req->rl_registered) &&
- !test_bit(RPCRDMA_REQ_F_TX_RESOURCES, &req->rl_flags))
- rpcrdma_complete_rqst(rep);
- else
- queue_work(rpcrdma_receive_wq, &rep->rr_work);
+ queue_work_on(req->rl_cpu, rpcrdma_receive_wq, &rep->rr_work);
diff --git a/net/sunrpc/xprtrdma/transport.c b/net/sunrpc/xprtrdma/transport.c
index 646c24494ea7..6ee1ad8978f3 100644
--- a/net/sunrpc/xprtrdma/transport.c
+++ b/net/sunrpc/xprtrdma/transport.c
@@ -52,6 +52,7 @@
#include <linux/slab.h>
#include <linux/seq_file.h>
#include <linux/sunrpc/addr.h>
+#include <linux/smp.h>
#include "xprt_rdma.h"
@@ -656,6 +657,7 @@ xprt_rdma_allocate(struct rpc_task *task)
task->tk_pid, __func__, rqst->rq_callsize,
rqst->rq_rcvsize, req);
+ req->rl_cpu = smp_processor_id();
req->rl_connect_cookie = 0; /* our reserved value */
rpcrdma_set_xprtdata(rqst, req);
rqst->rq_buffer = req->rl_sendbuf->rg_base;
diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c
index 710b3f77db82..8607c029c0dd 100644
--- a/net/sunrpc/xprtrdma/verbs.c
+++ b/net/sunrpc/xprtrdma/verbs.c
@@ -83,7 +83,7 @@ rpcrdma_alloc_wq(void)
struct workqueue_struct *recv_wq;
recv_wq = alloc_workqueue("xprtrdma_receive",
if (!recv_wq)
return -ENOMEM;
diff --git a/net/sunrpc/xprtrdma/xprt_rdma.h b/net/sunrpc/xprtrdma/xprt_rdma.h
index 51686d9eac5f..1342f743f1c4 100644
--- a/net/sunrpc/xprtrdma/xprt_rdma.h
+++ b/net/sunrpc/xprtrdma/xprt_rdma.h
@@ -342,6 +342,7 @@ enum {
struct rpcrdma_buffer;
struct rpcrdma_req {
struct list_head rl_list;
+ int rl_cpu;
unsigned int rl_connect_cookie;
struct rpcrdma_buffer *rl_buffer;
struct rpcrdma_rep *rl_reply;