Return-Path: Received: from mx1.redhat.com ([209.132.183.28]:48668 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755156AbdKGO32 (ORCPT ); Tue, 7 Nov 2017 09:29:28 -0500 From: Scott Mayhew To: trond.myklebust@primarydata.com, anna.schumaker@netapp.com Cc: linux-nfs@vger.kernel.org Subject: [PATCH] nfs: fix a deadlock in nfs v4.1 client initialization Date: Tue, 7 Nov 2017 09:29:27 -0500 Message-Id: <20171107142927.9468-1-smayhew@redhat.com> Sender: linux-nfs-owner@vger.kernel.org List-ID: The following deadlock can occur between a process waiting for a client to initialize in while walking the client list and another process waiting for the nfs_clid_init_mutex so it can initialize that client: Process 1 Process 2 --------- --------- spin_lock(&nn->nfs_client_lock); list_add_tail(&CLIENTA->cl_share_link, &nn->nfs_client_list); spin_unlock(&nn->nfs_client_lock); spin_lock(&nn->nfs_client_lock); list_add_tail(&CLIENTB->cl_share_link, &nn->nfs_client_list); spin_unlock(&nn->nfs_client_lock); mutex_lock(&nfs_clid_init_mutex); nfs41_walk_client_list(clp, result, cred); nfs_wait_client_init_complete(CLIENTA); (waiting for nfs_clid_init_mutex) Make sure nfs_match_client() only evaluates clients that have completed initialization in order to prevent that deadlock. Signed-off-by: Scott Mayhew --- fs/nfs/client.c | 9 +++++++++ 1 file changed, 9 insertions(+) diff --git a/fs/nfs/client.c b/fs/nfs/client.c index 22880ef..8b093994 100644 --- a/fs/nfs/client.c +++ b/fs/nfs/client.c @@ -291,12 +291,21 @@ static struct nfs_client *nfs_match_client(const struct nfs_client_initdata *dat const struct sockaddr *sap = data->addr; struct nfs_net *nn = net_generic(data->net, nfs_net_id); +again: list_for_each_entry(clp, &nn->nfs_client_list, cl_share_link) { const struct sockaddr *clap = (struct sockaddr *)&clp->cl_addr; /* Don't match clients that failed to initialise properly */ if (clp->cl_cons_state < 0) continue; + if (clp->cl_minorversion > 0 && + clp->cl_cons_state > NFS_CS_READY) { + spin_unlock(&nn->nfs_client_lock); + nfs_wait_client_init_complete(clp); + spin_lock(&nn->nfs_client_lock); + goto again; + } + /* Different NFS versions cannot share the same nfs_client */ if (clp->rpc_ops != data->nfs_mod->rpc_ops) continue; -- 2.9.5