* [PATCH] NFS4: Fix race to cl_confirm for v4.0 mount
@ 2019-11-07 21:56 Benjamin Coddington
0 siblings, 0 replies; only message in thread
From: Benjamin Coddington @ 2019-11-07 21:56 UTC (permalink / raw)
To: trond.myklebust, anna.schumaker, zhangxiaoxu; +Cc: linux-nfs
After commit f02f3755dbd1 ("NFS4: Fix v4.0 client state corruption when
mount") I have reports that sometimes all open() calls fail with -EPERM.
It appears that a race between the SETCLIENTID in trunking discovery and
another superflous SETCLIENTID triggered by immediately running the state
manager causes the nfs_client cl_confirm value to be stored in the wrong
order.
It doesn't really make sense to invoke the state manager just to clear the
NFS4CLNT_LEASE_EXPIRED flag. The flag has historically been used to signal
that a new client still needs to perform a SETCLIENTID before creating
state, but now that we have trunking discovery that flag is unnecessary.
Remove the flag, remove the call to schedule the state manager from
trunking discovery.
Signed-off-by: Benjamin Coddington <bcodding@redhat.com>
Fixes: f02f3755dbd1 ("NFS4: Fix v4.0 client state corruption when mount")
Cc: stable@vger.kernel.org #v5.2
---
fs/nfs/nfs4client.c | 1 -
fs/nfs/nfs4state.c | 4 ----
2 files changed, 5 deletions(-)
diff --git a/fs/nfs/nfs4client.c b/fs/nfs/nfs4client.c
index da6204025a2d..56cb3613082d 100644
--- a/fs/nfs/nfs4client.c
+++ b/fs/nfs/nfs4client.c
@@ -215,7 +215,6 @@ struct nfs_client *nfs4_alloc_client(const struct nfs_client_initdata *cl_init)
INIT_DELAYED_WORK(&clp->cl_renewd, nfs4_renew_state);
INIT_LIST_HEAD(&clp->cl_ds_clients);
rpc_init_wait_queue(&clp->cl_rpcwaitq, "NFS client");
- clp->cl_state = 1 << NFS4CLNT_LEASE_EXPIRED;
clp->cl_minorversion = cl_init->minorversion;
clp->cl_mvops = nfs_v4_minor_ops[cl_init->minorversion];
clp->cl_mig_gen = 1;
diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c
index cad4e064b328..e1d5dc4a31d7 100644
--- a/fs/nfs/nfs4state.c
+++ b/fs/nfs/nfs4state.c
@@ -180,10 +180,6 @@ int nfs40_discover_server_trunking(struct nfs_client *clp,
/* Sustain the lease, even if it's empty. If the clientid4
* goes stale it's of no use for trunking discovery. */
nfs4_schedule_state_renewal(*result);
-
- /* If the client state need to recover, do it. */
- if (clp->cl_state)
- nfs4_schedule_state_manager(clp);
}
out:
return status;
--
2.20.1
^ permalink raw reply related [flat|nested] only message in thread
only message in thread, other threads:[~2019-11-07 14:45 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-11-07 21:56 [PATCH] NFS4: Fix race to cl_confirm for v4.0 mount Benjamin Coddington
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).