[PATCH 0/2] Two fixes for NFS RCU-walk support

[PATCH 0/2] Two fixes for NFS RCU-walk support

[NeilBrown]

Thanks to the Feng Wu and the kbuild test robot I see
that some uses for rcu_dereference are wrong.
This two patches resolve the issue and also fix an
embarrassing but harmless bug.

---

NeilBrown (2):
      NFS: fix two problems in lookup_revalidate in RCU-walk
      SUNRPC: remove all refcounting of groupinfo from rpcauth_lookupcred


 fs/nfs/dir.c      |    8 ++++----
 net/sunrpc/auth.c |    8 +-------
 2 files changed, 5 insertions(+), 11 deletions(-)

-- 
Signature

[PATCH 2/2] SUNRPC: remove all refcounting of groupinfo from rpcauth_lookupcred

[NeilBrown]

current_cred() can only be changed by 'current', and
cred->group_info is never changed.  If a new group_info is
needed, a new 'cred' is created.

Consequently it is always safe to access
   current_cred()->group_info

without taking any further references.
So drop the refcounting and the incorrect rcu_dereference().

Signed-off-by: NeilBrown <neilb@suse.de>
---
 net/sunrpc/auth.c |    8 +-------
 1 file changed, 1 insertion(+), 7 deletions(-)

diff --git a/net/sunrpc/auth.c b/net/sunrpc/auth.c
index 794fc0f4fc4c..df2bd96f129d 100644
--- a/net/sunrpc/auth.c
+++ b/net/sunrpc/auth.c
@@ -595,14 +595,8 @@ rpcauth_lookupcred(struct rpc_auth *auth, int flags)
 	memset(&acred, 0, sizeof(acred));
 	acred.uid = cred->fsuid;
 	acred.gid = cred->fsgid;
-	if (flags & RPCAUTH_LOOKUP_RCU)
-		acred.group_info = rcu_dereference(cred->group_info);
-	else
-		acred.group_info = get_group_info(((struct cred *)cred)->group_info);
-
+	acred.group_info = cred->group_info;
 	ret = auth->au_ops->lookup_cred(auth, &acred, flags);
-	if (!(flags & RPCAUTH_LOOKUP_RCU))
-		put_group_info(acred.group_info);
 	return ret;
 }
 EXPORT_SYMBOL_GPL(rpcauth_lookupcred);

[PATCH 1/2] NFS: fix two problems in lookup_revalidate in RCU-walk

[NeilBrown]

1/ rcu_dereference isn't correct: that field isn't
   RCU protected.   It could potentially change at any time
   so ACCESS_ONCE might be justified.

   changes to ->d_parent are protected by ->d_seq.  However
   that isn't always checked after ->d_revalidate is called,
   so it is safest to keep the double-check that ->d_parent
   hasn't changed at the end of these functions.

2/ in nfs4_lookup_revalidate, "->d_parent" was forgotten.
   So 'parent' was not the parent of 'dentry'.
   This fails safe is the context is that dentry->d_inode is
   NULL, and the result of parent->d_inode being NULL is
   that ECHILD is returned, which is always safe.

Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: NeilBrown <neilb@suse.de>
---
 fs/nfs/dir.c |    8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
index e754d205ea54..0295f78f2976 100644
--- a/fs/nfs/dir.c
+++ b/fs/nfs/dir.c
@@ -1102,7 +1102,7 @@ static int nfs_lookup_revalidate(struct dentry *dentry, unsigned int flags)
 	int error;
 
 	if (flags & LOOKUP_RCU) {
-		parent = rcu_dereference(dentry->d_parent);
+		parent = ACCESS_ONCE(dentry->d_parent);
 		dir = ACCESS_ONCE(parent->d_inode);
 		if (!dir)
 			return -ECHILD;
@@ -1184,7 +1184,7 @@ out_set_verifier:
 	nfs_advise_use_readdirplus(dir);
  out_valid_noent:
 	if (flags & LOOKUP_RCU) {
-		if (parent != rcu_dereference(dentry->d_parent))
+		if (parent != ACCESS_ONCE(dentry->d_parent))
 			return -ECHILD;
 	} else
 		dput(parent);
@@ -1585,7 +1585,7 @@ static int nfs4_lookup_revalidate(struct dentry *dentry, unsigned int flags)
 		struct inode *dir;
 
 		if (flags & LOOKUP_RCU) {
-			parent = rcu_dereference(dentry);
+			parent = ACCESS_ONCE(dentry->d_parent);
 			dir = ACCESS_ONCE(parent->d_inode);
 			if (!dir)
 				return -ECHILD;
@@ -1599,7 +1599,7 @@ static int nfs4_lookup_revalidate(struct dentry *dentry, unsigned int flags)
 			ret = -ECHILD;
 		if (!(flags & LOOKUP_RCU))
 			dput(parent);
-		else if (parent != rcu_dereference(dentry))
+		else if (parent != ACCESS_ONCE(dentry->d_parent))
 			return -ECHILD;
 		goto out;
 	}