From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S932364AbbE3M51 (ORCPT ); Sat, 30 May 2015 08:57:27 -0400 Received: from mail-wi0-f182.google.com ([209.85.212.182]:34322 "EHLO mail-wi0-f182.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752733AbbE3M5T (ORCPT ); Sat, 30 May 2015 08:57:19 -0400 Date: Sat, 30 May 2015 08:57:10 -0400 From: Jeff Layton To: trond.myklebust@primarydata.com Cc: linux-nfs@vger.kernel.org, linux-mm@kvack.org, linux-kernel@vger.kernel.org, Mel Gorman , Jerome Marchand Subject: Re: [PATCH 4/4] sunrpc: lock xprt before trying to set memalloc on the sockets Message-ID: <20150530085710.34035d84@tlielax.poochiereds.net> In-Reply-To: <1432987393-15604-5-git-send-email-jeff.layton@primarydata.com> References: <1432987393-15604-1-git-send-email-jeff.layton@primarydata.com> <1432987393-15604-5-git-send-email-jeff.layton@primarydata.com> X-Mailer: Claws Mail 3.11.1 (GTK+ 2.24.28; x86_64-redhat-linux-gnu) MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Sat, 30 May 2015 08:03:13 -0400 Jeff Layton wrote: > It's possible that we could race with a call to xs_reset_transport, in > which case the xprt->inet pointer could be zeroed out while we're > accessing it. Lock the xprt before we try to set memalloc on it. > > Cc: Mel Gorman > Signed-off-by: Jeff Layton > --- > net/sunrpc/xprtsock.c | 17 +++++++++++++++++ > 1 file changed, 17 insertions(+) > > diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c > index 359446442112..91770ccab608 100644 > --- a/net/sunrpc/xprtsock.c > +++ b/net/sunrpc/xprtsock.c > @@ -1964,11 +1964,22 @@ static void xs_local_connect(struct rpc_xprt *xprt, struct rpc_task *task) > } > > #ifdef CONFIG_SUNRPC_SWAP > +/* > + * Note that this should be called with XPRT_LOCKED held (or when we otherwise > + * know that we have exclusive access to the socket), to guard against > + * races with xs_reset_transport. > + */ > static void xs_set_memalloc(struct rpc_xprt *xprt) > { > struct sock_xprt *transport = container_of(xprt, struct sock_xprt, > xprt); > > + /* > + * If there's no sock, then we have nothing to set. The > + * reconnecting process will get it for us. > + */ > + if (!transport->inet) > + return; > if (atomic_read(&xprt->swapper)) > sk_set_memalloc(transport->inet); > } > @@ -1986,8 +1997,11 @@ xs_swapper_enable(struct rpc_xprt *xprt) > struct sock_xprt *transport = container_of(xprt, struct sock_xprt, > xprt); > > + if (wait_on_bit_lock(&xprt->state, XPRT_LOCKED, TASK_KILLABLE)) > + return -ERESTARTSYS; > if (atomic_inc_return(&xprt->swapper) == 1) > sk_set_memalloc(transport->inet); Oh, we need to check that transport->inet is not NULL before we call sk_set/clear_memalloc on it. I'll respin with that fix once I give everyone a chance to comment on the rest... > + xprt_release_xprt(xprt, NULL); > return 0; > } > > @@ -2004,8 +2018,11 @@ xs_swapper_disable(struct rpc_xprt *xprt) > struct sock_xprt *transport = container_of(xprt, struct sock_xprt, > xprt); > > + if (wait_on_bit_lock(&xprt->state, XPRT_LOCKED, TASK_UNINTERRUPTIBLE)) > + return; > if (atomic_dec_and_test(&xprt->swapper)) > sk_clear_memalloc(transport->inet); > + xprt_release_xprt(xprt, NULL); > } > #else > static void xs_set_memalloc(struct rpc_xprt *xprt) -- Jeff Layton