Return-Path: linux-nfs-owner@vger.kernel.org Received: from mx2.netapp.com ([216.240.18.37]:57765 "EHLO mx2.netapp.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1760650Ab2EWRcv (ORCPT ); Wed, 23 May 2012 13:32:51 -0400 From: Trond Myklebust To: linux-nfs@vger.kernel.org Cc: Stanislav Kinsbursky Subject: [PATCH 2/3] NFSv4: Fix a race in the net namespace mount notification Date: Wed, 23 May 2012 13:32:42 -0400 Message-Id: <1337794363-11905-2-git-send-email-Trond.Myklebust@netapp.com> In-Reply-To: <1337794363-11905-1-git-send-email-Trond.Myklebust@netapp.com> References: <1337794363-11905-1-git-send-email-Trond.Myklebust@netapp.com> Sender: linux-nfs-owner@vger.kernel.org List-ID: Since the struct nfs_client gets added to the global nfs_client_list before it is initialised, it is possible that rpc_pipefs_event can end up trying to create idmapper entries on such a thing. The solution is to have the mount notification wait for the initialisation of each nfs_client to complete, and then to skip any entries for which the it failed. Reported-by: Stanislav Kinsbursky Signed-off-by: Trond Myklebust Cc: Stanislav Kinsbursky --- fs/nfs/client.c | 16 +++++++++++++--- fs/nfs/idmap.c | 15 +++++++++++++++ fs/nfs/internal.h | 1 + 3 files changed, 29 insertions(+), 3 deletions(-) diff --git a/fs/nfs/client.c b/fs/nfs/client.c index 78970a1..e6070ea 100644 --- a/fs/nfs/client.c +++ b/fs/nfs/client.c @@ -504,6 +504,17 @@ static struct nfs_client *nfs_match_client(const struct nfs_client_initdata *dat return NULL; } +static bool nfs_client_init_is_complete(const struct nfs_client *clp) +{ + return clp->cl_cons_state != NFS_CS_INITING; +} + +int nfs_wait_client_init_complete(const struct nfs_client *clp) +{ + return wait_event_killable(nfs_client_active_wq, + nfs_client_init_is_complete(clp)); +} + /* * Look up a client by IP address and protocol version * - creates a new record if one doesn't yet exist @@ -564,8 +575,7 @@ found_client: if (new) nfs_free_client(new); - error = wait_event_killable(nfs_client_active_wq, - clp->cl_cons_state < NFS_CS_INITING); + error = nfs_wait_client_init_complete(clp); if (error < 0) { nfs_put_client(clp); return ERR_PTR(-ERESTARTSYS); @@ -1317,7 +1327,7 @@ static int nfs4_init_client_minor_version(struct nfs_client *clp) * so that the client back channel can find the * nfs_client struct */ - clp->cl_cons_state = NFS_CS_SESSION_INITING; + nfs_mark_client_ready(clp, NFS_CS_SESSION_INITING); } #endif /* CONFIG_NFS_V4_1 */ diff --git a/fs/nfs/idmap.c b/fs/nfs/idmap.c index 3e8edbe..6ca949b 100644 --- a/fs/nfs/idmap.c +++ b/fs/nfs/idmap.c @@ -530,9 +530,24 @@ static struct nfs_client *nfs_get_client_for_event(struct net *net, int event) struct nfs_net *nn = net_generic(net, nfs_net_id); struct dentry *cl_dentry; struct nfs_client *clp; + int err; +restart: spin_lock(&nn->nfs_client_lock); list_for_each_entry(clp, &nn->nfs_client_list, cl_share_link) { + /* Wait for initialisation to finish */ + if (clp->cl_cons_state == NFS_CS_INITING) { + atomic_inc(&clp->cl_count); + spin_unlock(&nn->nfs_client_lock); + err = nfs_wait_client_init_complete(clp); + nfs_put_client(clp); + if (err) + return NULL; + goto restart; + } + /* Skip nfs_clients that failed to initialise */ + if (clp->cl_cons_state < 0) + continue; if (clp->rpc_ops != &nfs_v4_clientops) continue; cl_dentry = clp->cl_idmap->idmap_pipe->dentry; diff --git a/fs/nfs/internal.h b/fs/nfs/internal.h index 00b66de..a23daa9 100644 --- a/fs/nfs/internal.h +++ b/fs/nfs/internal.h @@ -167,6 +167,7 @@ extern struct nfs_server *nfs_clone_server(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, rpc_authflavor_t); +extern int nfs_wait_client_init_complete(const struct nfs_client *clp); extern void nfs_mark_client_ready(struct nfs_client *clp, int state); extern struct nfs_client *nfs4_set_ds_client(struct nfs_client* mds_clp, const struct sockaddr *ds_addr, -- 1.7.7.6