Received-SPF: pass (google.com: domain of linux-nfs+bounces-11-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45e3:2400::1 as permitted sender) client-ip=2604:1380:45e3:2400::1;
Subject: [PATCH v2 0/6] Was: "Switch NFS server CQs to use soft IRQ"
From: Chuck Lever <cel@kernel.org>
To: linux-nfs@vger.kernel.org, linux-rdma@vger.kernel.org
Cc: Chuck Lever <chuck.lever@oracle.com>, tom@talpey.com
Date: Tue, 21 Nov 2023 11:40:07 -0500
Message-ID: 
 <170058462629.4504.17663192195815644972.stgit@bazille.1015granger.net>
User-Agent: StGit/1.5
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: 7bit

After testing v1 "Switch NFS server CQs to use soft IRQ", I found
that using soft IRQ for completion handling actually reduced
throughput on a 70/30 8KB read/write test. So I've tossed those
patches out of this series.

Also, Jason reminded me that DMA unmapping is pretty costly, so that
has been moved out of Send completion handling on the NFS server.

---

Chuck Lever (6):
      svcrdma: Eliminate allocation of recv_ctxt objects in backchannel
      svcrdma: Pre-allocate svc_rdma_recv_ctxt objects
      svcrdma: Add a utility workqueue to svcrdma
      svcrdma: Add an async version of svc_rdma_send_ctxt_put()
      svcrdma: Add an async version of svc_rdma_write_info_free()
      svcrdma: Clean up locking


 include/linux/sunrpc/svc_rdma.h            |  6 +-
 net/sunrpc/xprtrdma/svc_rdma.c             | 32 +++++++---
 net/sunrpc/xprtrdma/svc_rdma_backchannel.c | 11 ++--
 net/sunrpc/xprtrdma/svc_rdma_recvfrom.c    | 32 ++++++----
 net/sunrpc/xprtrdma/svc_rdma_rw.c          | 12 +++-
 net/sunrpc/xprtrdma/svc_rdma_sendto.c      | 69 ++++++++++++++--------
 net/sunrpc/xprtrdma/svc_rdma_transport.c   |  1 +
 7 files changed, 110 insertions(+), 53 deletions(-)

--
Chuck Lever