[PATCH AUTOSEL 4.19 40/68] NFSv4: Fix a potential sleep while atomic in nfs4_do_reclaim()

[PATCH AUTOSEL 4.19 40/68] NFSv4: Fix a potential sleep while atomic in nfs4_do_reclaim()

[Sasha Levin]

From: Trond Myklebust <trond.myklebust@hammerspace.com>

[ Upstream commit c77e22834ae9a11891cb613bd9a551be1b94f2bc ]

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 <jhubbard@nvidia.com>
Fixes: 0aaaf5c424c7f ("NFS: Cache state owners after files are closed")
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 fs/nfs/nfs4_fs.h    |  3 ++-
 fs/nfs/nfs4client.c |  5 ++++-
 fs/nfs/nfs4state.c  | 27 ++++++++++++++++++++++-----
 3 files changed, 28 insertions(+), 7 deletions(-)

diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h
index 63287d911c088..5b61520dce888 100644
--- a/fs/nfs/nfs4_fs.h
+++ b/fs/nfs/nfs4_fs.h
@@ -469,7 +469,8 @@ static inline void nfs4_schedule_session_recovery(struct nfs4_session *session,
 
 extern struct nfs4_state_owner *nfs4_get_state_owner(struct nfs_server *, struct rpc_cred *, gfp_t);
 extern void nfs4_put_state_owner(struct nfs4_state_owner *);
-extern void nfs4_purge_state_owners(struct nfs_server *);
+extern void nfs4_purge_state_owners(struct nfs_server *, struct list_head *);
+extern void nfs4_free_state_owners(struct list_head *head);
 extern struct nfs4_state * nfs4_get_open_state(struct inode *, struct nfs4_state_owner *);
 extern void nfs4_put_open_state(struct nfs4_state *);
 extern void nfs4_close_state(struct nfs4_state *, fmode_t);
diff --git a/fs/nfs/nfs4client.c b/fs/nfs/nfs4client.c
index 8f53455c47653..86991bcfbeb12 100644
--- a/fs/nfs/nfs4client.c
+++ b/fs/nfs/nfs4client.c
@@ -754,9 +754,12 @@ int nfs41_walk_client_list(struct nfs_client *new,
 
 static void nfs4_destroy_server(struct nfs_server *server)
 {
+	LIST_HEAD(freeme);
+
 	nfs_server_return_all_delegations(server);
 	unset_pnfs_layoutdriver(server);
-	nfs4_purge_state_owners(server);
+	nfs4_purge_state_owners(server, &freeme);
+	nfs4_free_state_owners(&freeme);
 }
 
 /*
diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c
index 3ba2087469ac8..c36ef75f2054b 100644
--- a/fs/nfs/nfs4state.c
+++ b/fs/nfs/nfs4state.c
@@ -628,24 +628,39 @@ 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);
 	}
@@ -1843,12 +1858,13 @@ 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;
@@ -1877,6 +1893,7 @@ static int nfs4_do_reclaim(struct nfs_client *clp, const struct nfs4_state_recov
 		spin_unlock(&clp->cl_lock);
 	}
 	rcu_read_unlock();
+	nfs4_free_state_owners(&freeme);
 	return 0;
 }
 
-- 
2.20.1

[PATCH AUTOSEL 4.19 41/68] NFS: Fix regression whereby fscache errors are appearing on 'nofsc' mounts

[Sasha Levin]

From: Trond Myklebust <trond.myklebust@hammerspace.com>

[ Upstream commit dea1bb35c5f35e0577cfc61f79261d80b8715221 ]

People are reporing seeing fscache errors being reported concerning
duplicate cookies even in cases where they are not setting up fscache
at all. The rule needs to be that if fscache is not enabled, then it
should have no side effects at all.

To ensure this is the case, we disable fscache completely on all superblocks
for which the 'fsc' mount option was not set. In order to avoid issues
with '-oremount', we also disable the ability to turn fscache on via
remount.

Fixes: f1fe29b4a02d ("NFS: Use i_writecount to control whether...")
Link: https://bugzilla.kernel.org/show_bug.cgi?id=200145
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Cc: Steve Dickson <steved@redhat.com>
Cc: David Howells <dhowells@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 fs/nfs/fscache.c | 7 ++++++-
 fs/nfs/fscache.h | 2 +-
 fs/nfs/super.c   | 1 +
 3 files changed, 8 insertions(+), 2 deletions(-)

diff --git a/fs/nfs/fscache.c b/fs/nfs/fscache.c
index 4dc887813c71d..a7bc4e0494f92 100644
--- a/fs/nfs/fscache.c
+++ b/fs/nfs/fscache.c
@@ -118,6 +118,10 @@ void nfs_fscache_get_super_cookie(struct super_block *sb, const char *uniq, int
 	struct rb_node **p, *parent;
 	int diff;
 
+	nfss->fscache_key = NULL;
+	nfss->fscache = NULL;
+	if (!(nfss->options & NFS_OPTION_FSCACHE))
+		return;
 	if (!uniq) {
 		uniq = "";
 		ulen = 1;
@@ -230,10 +234,11 @@ void nfs_fscache_release_super_cookie(struct super_block *sb)
 void nfs_fscache_init_inode(struct inode *inode)
 {
 	struct nfs_fscache_inode_auxdata auxdata;
+	struct nfs_server *nfss = NFS_SERVER(inode);
 	struct nfs_inode *nfsi = NFS_I(inode);
 
 	nfsi->fscache = NULL;
-	if (!S_ISREG(inode->i_mode))
+	if (!(nfss->fscache && S_ISREG(inode->i_mode)))
 		return;
 
 	memset(&auxdata, 0, sizeof(auxdata));
diff --git a/fs/nfs/fscache.h b/fs/nfs/fscache.h
index 161ba2edb9d04..6363ea9568581 100644
--- a/fs/nfs/fscache.h
+++ b/fs/nfs/fscache.h
@@ -186,7 +186,7 @@ static inline void nfs_fscache_wait_on_invalidate(struct inode *inode)
  */
 static inline const char *nfs_server_fscache_state(struct nfs_server *server)
 {
-	if (server->fscache && (server->options & NFS_OPTION_FSCACHE))
+	if (server->fscache)
 		return "yes";
 	return "no ";
 }
diff --git a/fs/nfs/super.c b/fs/nfs/super.c
index 6df9b85caf205..d90efdea9fbd6 100644
--- a/fs/nfs/super.c
+++ b/fs/nfs/super.c
@@ -2239,6 +2239,7 @@ nfs_compare_remount_data(struct nfs_server *nfss,
 	    data->acdirmin != nfss->acdirmin / HZ ||
 	    data->acdirmax != nfss->acdirmax / HZ ||
 	    data->timeo != (10U * nfss->client->cl_timeout->to_initval / HZ) ||
+	    (data->options & NFS_OPTION_FSCACHE) != (nfss->options & NFS_OPTION_FSCACHE) ||
 	    data->nfs_server.port != nfss->port ||
 	    data->nfs_server.addrlen != nfss->nfs_client->cl_addrlen ||
 	    !rpc_cmp_addr((struct sockaddr *)&data->nfs_server.address,
-- 
2.20.1