From 86dbd08b32838dc8940ccab2eebd47cf80224494 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 22 Jul 2019 09:44:04 +0100 Subject: NFSv4: Print an error in the syslog when state is marked as irrecoverable When error recovery fails due to a fatal error on the server, ensure we log it in the syslog. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4state.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'fs/nfs/nfs4state.c') diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c index 9afd051a4876..a71a61e5fe2c 100644 --- a/fs/nfs/nfs4state.c +++ b/fs/nfs/nfs4state.c @@ -1463,7 +1463,7 @@ void nfs_inode_find_state_and_recover(struct inode *inode, nfs4_schedule_state_manager(clp); } -static void nfs4_state_mark_open_context_bad(struct nfs4_state *state) +static void nfs4_state_mark_open_context_bad(struct nfs4_state *state, int err) { struct inode *inode = state->inode; struct nfs_inode *nfsi = NFS_I(inode); @@ -1474,6 +1474,8 @@ static void nfs4_state_mark_open_context_bad(struct nfs4_state *state) if (ctx->state != state) continue; set_bit(NFS_CONTEXT_BAD, &ctx->flags); + pr_warn("NFSv4: state recovery failed for open file %pd2, " + "error = %d\n", ctx->dentry, err); } rcu_read_unlock(); } @@ -1481,7 +1483,7 @@ static void nfs4_state_mark_open_context_bad(struct nfs4_state *state) static void nfs4_state_mark_recovery_failed(struct nfs4_state *state, int error) { set_bit(NFS_STATE_RECOVERY_FAILED, &state->flags); - nfs4_state_mark_open_context_bad(state); + nfs4_state_mark_open_context_bad(state, error); } -- cgit v1.2.1 From c34fae003c79570b6c930b425fea3f0b7b1e7056 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 22 Jul 2019 09:54:29 +0100 Subject: NFSv4: When recovering state fails with EAGAIN, retry the same recovery If the server returns with EAGAIN when we're trying to recover from a server reboot, we currently delay for 1 second, but then mark the stateid as needing recovery after the grace period has expired. Instead, we should just retry the same recovery process immediately after the 1 second delay. Break out of the loop after 10 retries. Fixes: 35a61606a612 ("NFS: Reduce indentation of the switch statement...") Signed-off-by: Trond Myklebust --- fs/nfs/nfs4state.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) (limited to 'fs/nfs/nfs4state.c') diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c index a71a61e5fe2c..d03b9cf42bd0 100644 --- a/fs/nfs/nfs4state.c +++ b/fs/nfs/nfs4state.c @@ -1607,6 +1607,7 @@ static int __nfs4_reclaim_open_state(struct nfs4_state_owner *sp, struct nfs4_st static int nfs4_reclaim_open_state(struct nfs4_state_owner *sp, const struct nfs4_state_recovery_ops *ops) { struct nfs4_state *state; + unsigned int loop = 0; int status = 0; /* Note: we rely on the sp->so_states list being ordered @@ -1633,8 +1634,10 @@ restart: switch (status) { default: - if (status >= 0) + if (status >= 0) { + loop = 0; break; + } printk(KERN_ERR "NFS: %s: unhandled error %d\n", __func__, status); /* Fall through */ case -ENOENT: @@ -1648,6 +1651,10 @@ restart: break; case -EAGAIN: ssleep(1); + if (loop++ < 10) { + set_bit(ops->state_flag_bit, &state->flags); + break; + } /* Fall through */ case -NFS4ERR_ADMIN_REVOKED: case -NFS4ERR_STALE_STATEID: -- cgit v1.2.1 From c77e22834ae9a11891cb613bd9a551be1b94f2bc Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 3 Aug 2019 10:11:27 -0400 Subject: NFSv4: Fix a potential sleep while atomic in nfs4_do_reclaim() John Hubbard reports seeing the following stack trace: nfs4_do_reclaim rcu_read_lock /* we are now in_atomic() and must not sleep */ nfs4_purge_state_owners nfs4_free_state_owner nfs4_destroy_seqid_counter rpc_destroy_wait_queue cancel_delayed_work_sync __cancel_work_timer __flush_work start_flush_work might_sleep: (kernel/workqueue.c:2975: BUG) The solution is to separate out the freeing of the state owners from nfs4_purge_state_owners(), and perform that outside the atomic context. Reported-by: John Hubbard Fixes: 0aaaf5c424c7f ("NFS: Cache state owners after files are closed") Signed-off-by: Trond Myklebust --- fs/nfs/nfs4state.c | 27 ++++++++++++++++++++++----- 1 file changed, 22 insertions(+), 5 deletions(-) (limited to 'fs/nfs/nfs4state.c') diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c index d03b9cf42bd0..a4e866b2b43b 100644 --- a/fs/nfs/nfs4state.c +++ b/fs/nfs/nfs4state.c @@ -624,24 +624,39 @@ void nfs4_put_state_owner(struct nfs4_state_owner *sp) /** * nfs4_purge_state_owners - Release all cached state owners * @server: nfs_server with cached state owners to release + * @head: resulting list of state owners * * Called at umount time. Remaining state owners will be on * the LRU with ref count of zero. + * Note that the state owners are not freed, but are added + * to the list @head, which can later be used as an argument + * to nfs4_free_state_owners. */ -void nfs4_purge_state_owners(struct nfs_server *server) +void nfs4_purge_state_owners(struct nfs_server *server, struct list_head *head) { struct nfs_client *clp = server->nfs_client; struct nfs4_state_owner *sp, *tmp; - LIST_HEAD(doomed); spin_lock(&clp->cl_lock); list_for_each_entry_safe(sp, tmp, &server->state_owners_lru, so_lru) { - list_move(&sp->so_lru, &doomed); + list_move(&sp->so_lru, head); nfs4_remove_state_owner_locked(sp); } spin_unlock(&clp->cl_lock); +} - list_for_each_entry_safe(sp, tmp, &doomed, so_lru) { +/** + * nfs4_purge_state_owners - Release all cached state owners + * @head: resulting list of state owners + * + * Frees a list of state owners that was generated by + * nfs4_purge_state_owners + */ +void nfs4_free_state_owners(struct list_head *head) +{ + struct nfs4_state_owner *sp, *tmp; + + list_for_each_entry_safe(sp, tmp, head, so_lru) { list_del(&sp->so_lru); nfs4_free_state_owner(sp); } @@ -1865,12 +1880,13 @@ static int nfs4_do_reclaim(struct nfs_client *clp, const struct nfs4_state_recov struct nfs4_state_owner *sp; struct nfs_server *server; struct rb_node *pos; + LIST_HEAD(freeme); int status = 0; restart: rcu_read_lock(); list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) { - nfs4_purge_state_owners(server); + nfs4_purge_state_owners(server, &freeme); spin_lock(&clp->cl_lock); for (pos = rb_first(&server->state_owners); pos != NULL; @@ -1899,6 +1915,7 @@ restart: spin_unlock(&clp->cl_lock); } rcu_read_unlock(); + nfs4_free_state_owners(&freeme); return 0; } -- cgit v1.2.1 From 67e7b52d44e3d539dfbfcd866c3d3d69da23a909 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 7 Aug 2019 07:31:27 -0400 Subject: NFSv4: Ensure state recovery handles ETIMEDOUT correctly Ensure that the state recovery code handles ETIMEDOUT correctly, and also that we set RPC_TASK_TIMEOUT when recovering open state. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4state.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) (limited to 'fs/nfs/nfs4state.c') diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c index a4e866b2b43b..cad4e064b328 100644 --- a/fs/nfs/nfs4state.c +++ b/fs/nfs/nfs4state.c @@ -1529,6 +1529,7 @@ restart: switch (status) { case 0: break; + case -ETIMEDOUT: case -ESTALE: case -NFS4ERR_ADMIN_REVOKED: case -NFS4ERR_STALE_STATEID: @@ -1682,11 +1683,13 @@ restart: case -NFS4ERR_EXPIRED: case -NFS4ERR_NO_GRACE: nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state); + /* Fall through */ case -NFS4ERR_STALE_CLIENTID: case -NFS4ERR_BADSESSION: case -NFS4ERR_BADSLOT: case -NFS4ERR_BAD_HIGH_SLOT: case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION: + case -ETIMEDOUT: goto out_err; } nfs4_put_open_state(state); @@ -1971,7 +1974,6 @@ static int nfs4_handle_reclaim_lease_error(struct nfs_client *clp, int status) return -EPERM; case -EACCES: case -NFS4ERR_DELAY: - case -ETIMEDOUT: case -EAGAIN: ssleep(1); break; @@ -2600,7 +2602,7 @@ static void nfs4_state_manager(struct nfs_client *clp) } /* Now recover expired state... */ - if (test_and_clear_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state)) { + if (test_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state)) { section = "reclaim nograce"; status = nfs4_do_reclaim(clp, clp->cl_mvops->nograce_recovery_ops); @@ -2608,6 +2610,7 @@ static void nfs4_state_manager(struct nfs_client *clp) continue; if (status < 0) goto out_error; + clear_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state); } nfs4_end_drain_session(clp); -- cgit v1.2.1