From mboxrd@z Thu Jan 1 00:00:00 1970 From: Chuck Lever Subject: [PATCH v1 00/12] NFS/RDMA client-side patches proposed for v4.13 Date: Tue, 23 May 2017 10:53:37 -0400 Message-ID: <20170523142629.961.81233.stgit@manet.1015granger.net> Mime-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit Return-path: Sender: linux-rdma-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org To: linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, linux-nfs-u79uwXL29TY76Z2rM5mHXA@public.gmane.org List-Id: linux-rdma@vger.kernel.org This series contains a major fix for client-side NFS/RDMA. When a signal fires, it's possible for the server's RPC reply to race with the client code that terminates the RPC after a signal. The result is both of these code paths try to invalidate the same MR at the same time. Because FRWR invalidation is typically fast, it's nearly impossible to hit the race with FRWR. However, FMR invalidation happens at about the same speed as the NFS server responds, so it's more likely to hit this window when using FMR. FMR is also more sensitive to concurrent operations on the same MR, which can result in a kernel crash or an HCA firmware reset. As part of closing the signal race window, the reply handler is restructured and several error recovery paths in the invalidation code are fixed. In addition to this fix, there is a small but important change to make NFSv4.1 Transparent State Migration work. This enables basic test cases to pass successfully. However, NFSv4.1 TSM is still under test, so the default setting of CONFIG_NFS_V4_1_MIGRATION remains unchanged for the moment. Available in the "nfs-rdma-for-4.13" topic branch of this git repo: git://git.linux-nfs.org/projects/cel/cel-2.6.git Or for browsing: http://git.linux-nfs.org/?p=cel/cel-2.6.git;a=log;h=refs/heads/nfs-rdma-for-4.13 Chuck Lever (12): xprtrdma: Don't defer MR recovery if ro_map fails xprtrdma: On invalidation failure, remove MWs from rl_registered xprtrdma: Fix FRWR invalidation error recovery xprtrdma: Pre-mark remotely invalidated MRs xprtrdma: Pass only the list of registered MRs to ro_unmap_sync xprtrdma: Rename rpcrdma_req::rl_free xprtrdma: Fix client lock-up after application signal fires NFSv4.1: Handle EXCHGID4_FLAG_CONFIRMED_R during NFSv4.1 migration xprtrdma: Demote "connect" log messages xprtrdma: FMR does not need list_del_init() xprtrdma: Replace PAGE_MASK with offset_in_page() xprtrdma: Fix documenting comments in frwr_ops.c fs/nfs/nfs4proc.c | 2 - fs/nfs/nfs4state.c | 11 +++ net/sunrpc/xprtrdma/fmr_ops.c | 47 ++++++++------- net/sunrpc/xprtrdma/frwr_ops.c | 69 ++++++++++------------ net/sunrpc/xprtrdma/rpc_rdma.c | 125 +++++++++++++++++++++++++-------------- net/sunrpc/xprtrdma/transport.c | 3 + net/sunrpc/xprtrdma/verbs.c | 55 ++++------------- net/sunrpc/xprtrdma/xprt_rdma.h | 40 ++++++++++++ 8 files changed, 204 insertions(+), 148 deletions(-) -- Chuck Lever -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org More majordomo info at http://vger.kernel.org/majordomo-info.html From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-io0-f169.google.com ([209.85.223.169]:33414 "EHLO mail-io0-f169.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S967074AbdEWOxm (ORCPT ); Tue, 23 May 2017 10:53:42 -0400 Subject: [PATCH v1 00/12] NFS/RDMA client-side patches proposed for v4.13 From: Chuck Lever To: linux-rdma@vger.kernel.org, linux-nfs@vger.kernel.org Date: Tue, 23 May 2017 10:53:37 -0400 Message-ID: <20170523142629.961.81233.stgit@manet.1015granger.net> MIME-Version: 1.0 Content-Type: text/plain; charset="utf-8" Sender: linux-nfs-owner@vger.kernel.org List-ID: This series contains a major fix for client-side NFS/RDMA. When a signal fires, it's possible for the server's RPC reply to race with the client code that terminates the RPC after a signal. The result is both of these code paths try to invalidate the same MR at the same time. Because FRWR invalidation is typically fast, it's nearly impossible to hit the race with FRWR. However, FMR invalidation happens at about the same speed as the NFS server responds, so it's more likely to hit this window when using FMR. FMR is also more sensitive to concurrent operations on the same MR, which can result in a kernel crash or an HCA firmware reset. As part of closing the signal race window, the reply handler is restructured and several error recovery paths in the invalidation code are fixed. In addition to this fix, there is a small but important change to make NFSv4.1 Transparent State Migration work. This enables basic test cases to pass successfully. However, NFSv4.1 TSM is still under test, so the default setting of CONFIG_NFS_V4_1_MIGRATION remains unchanged for the moment. Available in the "nfs-rdma-for-4.13" topic branch of this git repo: git://git.linux-nfs.org/projects/cel/cel-2.6.git Or for browsing: http://git.linux-nfs.org/?p=cel/cel-2.6.git;a=log;h=refs/heads/nfs-rdma-for-4.13 Chuck Lever (12): xprtrdma: Don't defer MR recovery if ro_map fails xprtrdma: On invalidation failure, remove MWs from rl_registered xprtrdma: Fix FRWR invalidation error recovery xprtrdma: Pre-mark remotely invalidated MRs xprtrdma: Pass only the list of registered MRs to ro_unmap_sync xprtrdma: Rename rpcrdma_req::rl_free xprtrdma: Fix client lock-up after application signal fires NFSv4.1: Handle EXCHGID4_FLAG_CONFIRMED_R during NFSv4.1 migration xprtrdma: Demote "connect" log messages xprtrdma: FMR does not need list_del_init() xprtrdma: Replace PAGE_MASK with offset_in_page() xprtrdma: Fix documenting comments in frwr_ops.c fs/nfs/nfs4proc.c | 2 - fs/nfs/nfs4state.c | 11 +++ net/sunrpc/xprtrdma/fmr_ops.c | 47 ++++++++------- net/sunrpc/xprtrdma/frwr_ops.c | 69 ++++++++++------------ net/sunrpc/xprtrdma/rpc_rdma.c | 125 +++++++++++++++++++++++++-------------- net/sunrpc/xprtrdma/transport.c | 3 + net/sunrpc/xprtrdma/verbs.c | 55 ++++------------- net/sunrpc/xprtrdma/xprt_rdma.h | 40 ++++++++++++ 8 files changed, 204 insertions(+), 148 deletions(-) -- Chuck Lever