Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs

Pull vfs fixes from Al Viro:
 "atomic_open-related fixes (Miklos' series, with EEXIST-related parts
  replaced with fix in fs/namei.c:atomic_open() instead of messing with
  the instances) + race fix in autofs + leak on failure exit in 9p"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
  9p: don't forget to destroy inode cache if fscache registration fails
  atomic_open: take care of EEXIST in no-open case with O_CREAT|O_EXCL in fs/namei.c
  vfs: don't set FILE_CREATED before calling ->atomic_open()
  nfs: set FILE_CREATED
  gfs2: set FILE_CREATED
  cifs: fix filp leak in cifs_atomic_open()
  vfs: improve i_op->atomic_open() documentation
  autofs4: close the races around autofs4_notify_daemon()

9 files changed, 62 insertions, 43 deletions

diff --git a/Documentation/filesystems/vfs.txt b/Documentation/filesystems/vfs.txt
index f93a88250a44..deb48b5fd883 100644
--- a/Documentation/filesystems/vfs.txt
+++ b/Documentation/filesystems/vfs.txt
@@ -359,11 +359,9 @@ struct inode_operations {
         ssize_t (*listxattr) (struct dentry *, char *, size_t);
         int (*removexattr) (struct dentry *, const char *);
         void (*update_time)(struct inode *, struct timespec *, int);
-        int (*atomic_open)(struct inode *, struct dentry *,
+        int (*atomic_open)(struct inode *, struct dentry *, struct file *,
+                        unsigned open_flag, umode_t create_mode, int *opened);
         int (*tmpfile) (struct inode *, struct dentry *, umode_t);
-} ____cacheline_aligned;
-                                struct file *, unsigned open_flag,
-                                umode_t create_mode, int *opened);
 };
 
 Again, all methods are called without any locks being held, unless
@@ -470,9 +468,11 @@ otherwise noted.
         method the filesystem can look up, possibly create and open the file in
         one atomic operation.  If it cannot perform this (e.g. the file type
         turned out to be wrong) it may signal this by returning 1 instead of
-        usual 0 or -ve .  This method is only called if the last
-        component is negative or needs lookup.  Cached positive dentries are
-        still handled by f_op->open().
+        usual 0 or -ve .  This method is only called if the last component is
+        negative or needs lookup.  Cached positive dentries are still handled by
+        f_op->open().  If the file was created, the FILE_CREATED flag should be
+        set in "opened".  In case of O_EXCL the method must only succeed if the
+        file didn't exist and hence FILE_CREATED shall always be set on success.
 
   tmpfile: called in the end of O_TMPFILE open().  Optional, equivalent to
         atomically creating, opening and unlinking a file in given directory.
diff --git a/fs/9p/v9fs.c b/fs/9p/v9fs.c
index 58e6cbce4156..08f2e1e9a7e6 100644
--- a/fs/9p/v9fs.c
+++ b/fs/9p/v9fs.c
@@ -603,10 +603,11 @@ static int v9fs_cache_register(void)
         if (ret < 0)
                 return ret;
 #ifdef CONFIG_9P_FSCACHE
-        return fscache_register_netfs(&v9fs_cache_netfs);
-#else
-        return ret;
+        ret = fscache_register_netfs(&v9fs_cache_netfs);
+        if (ret < 0)
+                v9fs_destroy_inode_cache();
 #endif
+        return ret;
 }
 
 static void v9fs_cache_unregister(void)
diff --git a/fs/9p/vfs_inode_dotl.c b/fs/9p/vfs_inode_dotl.c
index 53687bbf2296..a7c481402c46 100644
--- a/fs/9p/vfs_inode_dotl.c
+++ b/fs/9p/vfs_inode_dotl.c
@@ -267,14 +267,8 @@ v9fs_vfs_atomic_open_dotl(struct inode *dir, struct dentry *dentry,
         }
 
         /* Only creates */
-        if (!(flags & O_CREAT))
+        if (!(flags & O_CREAT) || dentry->d_inode)
                 return  finish_no_open(file, res);
-        else if (dentry->d_inode) {
-                if ((flags & (O_CREAT | O_EXCL)) == (O_CREAT | O_EXCL))
-                        return -EEXIST;
-                else
-                        return finish_no_open(file, res);
-        }
 
         v9ses = v9fs_inode2v9ses(dir);
 
diff --git a/fs/autofs4/waitq.c b/fs/autofs4/waitq.c
index 3db70dae40d3..689e40d983ad 100644
--- a/fs/autofs4/waitq.c
+++ b/fs/autofs4/waitq.c
@@ -109,13 +109,7 @@ static void autofs4_notify_daemon(struct autofs_sb_info *sbi,
 
         pkt.hdr.proto_version = sbi->version;
         pkt.hdr.type = type;
-        mutex_lock(&sbi->wq_mutex);
 
-        /* Check if we have become catatonic */
-        if (sbi->catatonic) {
-                mutex_unlock(&sbi->wq_mutex);
-                return;
-        }
         switch (type) {
         /* Kernel protocol v4 missing and expire packets */
         case autofs_ptype_missing:
@@ -427,7 +421,6 @@ int autofs4_wait(struct autofs_sb_info *sbi, struct dentry *dentry,
                 wq->tgid = current->tgid;
                 wq->status = -EINTR; /* Status return if interrupted */
                 wq->wait_ctr = 2;
-                mutex_unlock(&sbi->wq_mutex);
 
                 if (sbi->version < 5) {
                         if (notify == NFY_MOUNT)
@@ -449,15 +442,15 @@ int autofs4_wait(struct autofs_sb_info *sbi, struct dentry *dentry,
                         (unsigned long) wq->wait_queue_token, wq->name.len,
                         wq->name.name, notify);
 
-                /* autofs4_notify_daemon() may block */
+                /* autofs4_notify_daemon() may block; it will unlock ->wq_mutex */
                 autofs4_notify_daemon(sbi, wq, type);
         } else {
                 wq->wait_ctr++;
-                mutex_unlock(&sbi->wq_mutex);
-                kfree(qstr.name);
                 DPRINTK("existing wait id = 0x%08lx, name = %.*s, nfy=%d",
                         (unsigned long) wq->wait_queue_token, wq->name.len,
                         wq->name.name, notify);
+                mutex_unlock(&sbi->wq_mutex);
+                kfree(qstr.name);
         }
 
         /*
diff --git a/fs/cifs/dir.c b/fs/cifs/dir.c
index d3e2eaa503a6..5384c2a640ca 100644
--- a/fs/cifs/dir.c
+++ b/fs/cifs/dir.c
@@ -500,6 +500,7 @@ cifs_atomic_open(struct inode *inode, struct dentry *direntry,
                 if (server->ops->close)
                         server->ops->close(xid, tcon, &fid);
                 cifs_del_pending_open(&open);
+                fput(file);
                 rc = -ENOMEM;
         }
 
diff --git a/fs/gfs2/inode.c b/fs/gfs2/inode.c
index 64915eeae5a7..ced3257f06e8 100644
--- a/fs/gfs2/inode.c
+++ b/fs/gfs2/inode.c
@@ -694,8 +694,10 @@ static int gfs2_create_inode(struct inode *dir, struct dentry *dentry,
 
         mark_inode_dirty(inode);
         d_instantiate(dentry, inode);
-        if (file)
+        if (file) {
+                *opened |= FILE_CREATED;
                 error = finish_open(file, dentry, gfs2_open_common, opened);
+        }
         gfs2_glock_dq_uninit(ghs);
         gfs2_glock_dq_uninit(ghs + 1);
         return error;
diff --git a/fs/namei.c b/fs/namei.c
index 0dc4cbf21f37..645268f23eb6 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -2656,6 +2656,7 @@ static int atomic_open(struct nameidata *nd, struct dentry *dentry,
         int acc_mode;
         int create_error = 0;
         struct dentry *const DENTRY_NOT_SET = (void *) -1UL;
+        bool excl;
 
         BUG_ON(dentry->d_inode);
 
@@ -2669,10 +2670,9 @@ static int atomic_open(struct nameidata *nd, struct dentry *dentry,
         if ((open_flag & O_CREAT) && !IS_POSIXACL(dir))
                 mode &= ~current_umask();
 
-        if ((open_flag & (O_EXCL | O_CREAT)) == (O_EXCL | O_CREAT)) {
+        excl = (open_flag & (O_EXCL | O_CREAT)) == (O_EXCL | O_CREAT);
+        if (excl)
                 open_flag &= ~O_TRUNC;
-                *opened |= FILE_CREATED;
-        }
 
         /*
          * Checking write permission is tricky, bacuse we don't know if we are
@@ -2725,12 +2725,6 @@ static int atomic_open(struct nameidata *nd, struct dentry *dentry,
                 goto out;
         }
 
-        acc_mode = op->acc_mode;
-        if (*opened & FILE_CREATED) {
-                fsnotify_create(dir, dentry);
-                acc_mode = MAY_OPEN;
-        }
-
         if (error) {    /* returned 1, that is */
                 if (WARN_ON(file->f_path.dentry == DENTRY_NOT_SET)) {
                         error = -EIO;
@@ -2740,9 +2734,19 @@ static int atomic_open(struct nameidata *nd, struct dentry *dentry,
                         dput(dentry);
                         dentry = file->f_path.dentry;
                 }
-                if (create_error && dentry->d_inode == NULL) {
-                        error = create_error;
-                        goto out;
+                if (*opened & FILE_CREATED)
+                        fsnotify_create(dir, dentry);
+                if (!dentry->d_inode) {
+                        WARN_ON(*opened & FILE_CREATED);
+                        if (create_error) {
+                                error = create_error;
+                                goto out;
+                        }
+                } else {
+                        if (excl && !(*opened & FILE_CREATED)) {
+                                error = -EEXIST;
+                                goto out;
+                        }
                 }
                 goto looked_up;
         }
@@ -2751,6 +2755,12 @@ static int atomic_open(struct nameidata *nd, struct dentry *dentry,
          * We didn't have the inode before the open, so check open permission
          * here.
          */
+        acc_mode = op->acc_mode;
+        if (*opened & FILE_CREATED) {
+                WARN_ON(!(open_flag & O_CREAT));
+                fsnotify_create(dir, dentry);
+                acc_mode = MAY_OPEN;
+        }
         error = may_open(&file->f_path, acc_mode, open_flag);
         if (error)
                 fput(file);
diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
index de434f309af0..854a8f05a610 100644
--- a/fs/nfs/dir.c
+++ b/fs/nfs/dir.c
@@ -1392,6 +1392,9 @@ static int nfs_finish_open(struct nfs_open_context *ctx,
 {
         int err;
 
+        if ((open_flags & (O_CREAT | O_EXCL)) == (O_CREAT | O_EXCL))
+                *opened |= FILE_CREATED;
+
         err = finish_open(file, dentry, do_open, opened);
         if (err)
                 goto out;
diff --git a/fs/open.c b/fs/open.c
index 2a731b0d08bc..d420331ca32a 100644
--- a/fs/open.c
+++ b/fs/open.c
@@ -744,14 +744,24 @@ cleanup_file:
 
 /**
  * finish_open - finish opening a file
- * @od: opaque open data
+ * @file: file pointer
  * @dentry: pointer to dentry
  * @open: open callback
+ * @opened: state of open
  *
  * This can be used to finish opening a file passed to i_op->atomic_open().
  *
  * If the open callback is set to NULL, then the standard f_op->open()
  * filesystem callback is substituted.
+ *
+ * NB: the dentry reference is _not_ consumed.  If, for example, the dentry is
+ * the return value of d_splice_alias(), then the caller needs to perform dput()
+ * on it after finish_open().
+ *
+ * On successful return @file is a fully instantiated open file.  After this, if
+ * an error occurs in ->atomic_open(), it needs to clean up with fput().
+ *
+ * Returns zero on success or -errno if the open failed.
  */
 int finish_open(struct file *file, struct dentry *dentry,
                 int (*open)(struct inode *, struct file *),
@@ -772,11 +782,16 @@ EXPORT_SYMBOL(finish_open);
 /**
  * finish_no_open - finish ->atomic_open() without opening the file
  *
- * @od: opaque open data
+ * @file: file pointer
  * @dentry: dentry or NULL (as returned from ->lookup())
  *
  * This can be used to set the result of a successful lookup in ->atomic_open().
- * The filesystem's atomic_open() method shall return NULL after calling this.
+ *
+ * NB: unlike finish_open() this function does consume the dentry reference and
+ * the caller need not dput() it.
+ *
+ * Returns "1" which must be the return value of ->atomic_open() after having
+ * called this function.
  */
 int finish_no_open(struct file *file, struct dentry *dentry)
 {