From patchwork Sat Apr 24 19:02:28 2021 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Chuck Lever III X-Patchwork-Id: 12222747 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-13.8 required=3.0 tests=BAYES_00, HEADER_FROM_DIFFERENT_DOMAINS,INCLUDES_CR_TRAILER,INCLUDES_PATCH, MAILING_LIST_MULTI,SPF_HELO_NONE,SPF_PASS autolearn=ham autolearn_force=no version=3.4.0 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id DB084C433B4 for ; Sat, 24 Apr 2021 19:02:36 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by mail.kernel.org (Postfix) with ESMTP id A9A8B611CE for ; Sat, 24 Apr 2021 19:02:36 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S232690AbhDXTDN (ORCPT ); Sat, 24 Apr 2021 15:03:13 -0400 Received: from mail.kernel.org ([198.145.29.99]:59408 "EHLO mail.kernel.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S231814AbhDXTDM (ORCPT ); Sat, 24 Apr 2021 15:03:12 -0400 Received: by mail.kernel.org (Postfix) with ESMTPSA id 5D0EF610FA; Sat, 24 Apr 2021 19:02:30 +0000 (UTC) Subject: [PATCH v4] xprtrdma: Improve locking around rpcrdma_rep destruction From: Chuck Lever To: trondmy@hammerspace.com Cc: linux-nfs@vger.kernel.org, linux-rdma@vger.kernel.org Date: Sat, 24 Apr 2021 15:02:28 -0400 Message-ID: <161929084476.3664.10196536550424097179.stgit@manet.1015granger.net> User-Agent: StGit/0.23-29-ga622f1 MIME-Version: 1.0 Precedence: bulk List-ID: X-Mailing-List: linux-rdma@vger.kernel.org Currently rpcrdma_reps_destroy() assumes that, at transport tear-down, the content of the rb_free_reps list is the same as the content of the rb_all_reps list. Although that is usually true, using the rb_all_reps list should be more reliable because of the way it's managed. And, rpcrdma_reps_unmap() uses rb_all_reps; these two functions should both traverse the "all" list. Ensure that all rpcrdma_reps are always destroyed whether they are on the rep free list or not. Signed-off-by: Chuck Lever --- net/sunrpc/xprtrdma/verbs.c | 31 ++++++++++++++++++++++++------- 1 file changed, 24 insertions(+), 7 deletions(-) Changes since v3: - Rename rpcrdma_rep_destroy_locked() as rpcrdma_rep_free() As far as I can tell, none of the subsequent patches in v3 of the series need modification, manual or otherwise. diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c index 1b599a623eea..2bc4589c37ce 100644 --- a/net/sunrpc/xprtrdma/verbs.c +++ b/net/sunrpc/xprtrdma/verbs.c @@ -1007,16 +1007,23 @@ struct rpcrdma_rep *rpcrdma_rep_create(struct rpcrdma_xprt *r_xprt, return NULL; } -/* No locking needed here. This function is invoked only by the - * Receive completion handler, or during transport shutdown. - */ -static void rpcrdma_rep_destroy(struct rpcrdma_rep *rep) +static void rpcrdma_rep_free(struct rpcrdma_rep *rep) { - list_del(&rep->rr_all); rpcrdma_regbuf_free(rep->rr_rdmabuf); kfree(rep); } +static void rpcrdma_rep_destroy(struct rpcrdma_rep *rep) +{ + struct rpcrdma_buffer *buf = &rep->rr_rxprt->rx_buf; + + spin_lock(&buf->rb_lock); + list_del(&rep->rr_all); + spin_unlock(&buf->rb_lock); + + rpcrdma_rep_free(rep); +} + static struct rpcrdma_rep *rpcrdma_rep_get_locked(struct rpcrdma_buffer *buf) { struct llist_node *node; @@ -1049,8 +1056,18 @@ static void rpcrdma_reps_destroy(struct rpcrdma_buffer *buf) { struct rpcrdma_rep *rep; - while ((rep = rpcrdma_rep_get_locked(buf)) != NULL) - rpcrdma_rep_destroy(rep); + spin_lock(&buf->rb_lock); + while ((rep = list_first_entry_or_null(&buf->rb_all_reps, + struct rpcrdma_rep, + rr_all)) != NULL) { + list_del(&rep->rr_all); + spin_unlock(&buf->rb_lock); + + rpcrdma_rep_free(rep); + + spin_lock(&buf->rb_lock); + } + spin_unlock(&buf->rb_lock); } /**