GFS2: Add atomic_open support

GFS2: Add atomic_open support

[Steven Whitehouse]

The following patch implements atomic_open for GFS2. This is mostly
straightforward, however there is one corner case which I've had to
deal with, beyond what would normally be expected for a local
filesystem.

The ->atomic_open function can be passed negative dentries, which
in most cases means either ENOENT (->lookup) or a call to d_instantiate
(->create). In the GFS2 case though, we need to actually perform the
look up, since we do not know whether there has been a new inode created
on another node. The look up calls d_splice_alias which then tries to
rehash the dentry - so the solution here is to simply check for that
in d_splice_alias. The same issue is likely to affect any other cluster
filesystem implementing ->atomic_open

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: "J. Bruce Fields" <bfields@fieldses.org>
Cc: Jeff Layton <jlayton@redhat.com>

diff --git a/fs/dcache.c b/fs/dcache.c
index f09b908..5a23073 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -1612,6 +1612,10 @@ EXPORT_SYMBOL(d_obtain_alias);
  * If a dentry was found and moved, then it is returned.  Otherwise NULL
  * is returned.  This matches the expected return value of ->lookup.
  *
+ * Cluster filesystems may call this function with a negative, hashed dentry.
+ * In that case, we know that the inode will be a regular file, and also this
+ * will only occur during atomic_open. So we need to check for the dentry
+ * being already hashed only in the final case.
  */
 struct dentry *d_splice_alias(struct inode *inode, struct dentry *dentry)
 {
@@ -1636,8 +1640,11 @@ struct dentry *d_splice_alias(struct inode *inode, struct dentry *dentry)
 			security_d_instantiate(dentry, inode);
 			d_rehash(dentry);
 		}
-	} else
-		d_add(dentry, inode);
+	} else {
+		d_instantiate(dentry, inode);
+		if (d_unhashed(dentry))
+			d_rehash(dentry);
+	}
 	return new;
 }
 EXPORT_SYMBOL(d_splice_alias);
diff --git a/fs/gfs2/file.c b/fs/gfs2/file.c
index ad0dc38..4ed6a03 100644
--- a/fs/gfs2/file.c
+++ b/fs/gfs2/file.c
@@ -538,21 +538,30 @@ static int gfs2_mmap(struct file *file, struct vm_area_struct *vma)
 }
 
 /**
- * gfs2_open - open a file
- * @inode: the inode to open
- * @file: the struct file for this opening
+ * gfs2_open_common - This is common to open and atomic_open
+ * @inode: The inode being opened
+ * @file: The file being opened
  *
- * Returns: errno
+ * This maybe called under a glock or not depending upon how it has
+ * been called. We must always be called under a glock for regular
+ * files, however. For other file types, it does not matter whether
+ * we hold the glock or not.
+ *
+ * Returns: Error code or 0 for success
  */
 
-static int gfs2_open(struct inode *inode, struct file *file)
+int gfs2_open_common(struct inode *inode, struct file *file)
 {
-	struct gfs2_inode *ip = GFS2_I(inode);
-	struct gfs2_holder i_gh;
 	struct gfs2_file *fp;
-	int error;
+	int ret;
 
-	fp = kzalloc(sizeof(struct gfs2_file), GFP_KERNEL);
+	if (S_ISREG(inode->i_mode)) {
+		ret = generic_file_open(inode, file);
+		if (ret)
+			return ret;
+	}
+
+	fp = kzalloc(sizeof(struct gfs2_file), GFP_NOFS);
 	if (!fp)
 		return -ENOMEM;
 
@@ -560,29 +569,43 @@ static int gfs2_open(struct inode *inode, struct file *file)
 
 	gfs2_assert_warn(GFS2_SB(inode), !file->private_data);
 	file->private_data = fp;
+	return 0;
+}
+
+/**
+ * gfs2_open - open a file
+ * @inode: the inode to open
+ * @file: the struct file for this opening
+ *
+ * After atomic_open, this function is only used for opening files
+ * which are already cached. We must still get the glock for regular
+ * files to ensure that we have the file size uptodate for the large
+ * file check which is in the common code. That is only an issue for
+ * regular files though.
+ *
+ * Returns: errno
+ */
+
+static int gfs2_open(struct inode *inode, struct file *file)
+{
+	struct gfs2_inode *ip = GFS2_I(inode);
+	struct gfs2_holder i_gh;
+	int error;
+	bool need_unlock = false;
 
 	if (S_ISREG(ip->i_inode.i_mode)) {
 		error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_ANY,
 					   &i_gh);
 		if (error)
-			goto fail;
+			return error;
+		need_unlock = true;
+	}
 
-		if (!(file->f_flags & O_LARGEFILE) &&
-		    i_size_read(inode) > MAX_NON_LFS) {
-			error = -EOVERFLOW;
-			goto fail_gunlock;
-		}
+	error = gfs2_open_common(inode, file);
 
+	if (need_unlock)
 		gfs2_glock_dq_uninit(&i_gh);
-	}
-
-	return 0;
 
-fail_gunlock:
-	gfs2_glock_dq_uninit(&i_gh);
-fail:
-	file->private_data = NULL;
-	kfree(fp);
 	return error;
 }
 
diff --git a/fs/gfs2/inode.c b/fs/gfs2/inode.c
index 5fbb8df..065c7fb 100644
--- a/fs/gfs2/inode.c
+++ b/fs/gfs2/inode.c
@@ -546,6 +546,7 @@ static int gfs2_security_init(struct gfs2_inode *dip, struct gfs2_inode *ip,
  * gfs2_create_inode - Create a new inode
  * @dir: The parent directory
  * @dentry: The new dentry
+ * @file: If non-NULL, the file which is being opened
  * @mode: The permissions on the new inode
  * @dev: For device nodes, this is the device number
  * @symname: For symlinks, this is the link destination
@@ -555,8 +556,9 @@ static int gfs2_security_init(struct gfs2_inode *dip, struct gfs2_inode *ip,
  */
 
 static int gfs2_create_inode(struct inode *dir, struct dentry *dentry,
+			     struct file *file,
 			     umode_t mode, dev_t dev, const char *symname,
-			     unsigned int size, int excl)
+			     unsigned int size, int excl, int *opened)
 {
 	const struct qstr *name = &dentry->d_name;
 	struct gfs2_holder ghs[2];
@@ -564,6 +566,7 @@ static int gfs2_create_inode(struct inode *dir, struct dentry *dentry,
 	struct gfs2_inode *dip = GFS2_I(dir), *ip;
 	struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
 	struct gfs2_glock *io_gl;
+	struct dentry *d;
 	int error;
 	u32 aflags = 0;
 	int arq;
@@ -586,9 +589,14 @@ static int gfs2_create_inode(struct inode *dir, struct dentry *dentry,
 	error = create_ok(dip, name, mode);
 	if ((error == -EEXIST) && S_ISREG(mode) && !excl) {
 		inode = gfs2_lookupi(dir, &dentry->d_name, 0);
+		d = d_splice_alias(inode, dentry);
+		error = 0;
+		if (file && !IS_ERR(d))
+			error = finish_open(file, dentry, gfs2_open_common, opened);
 		gfs2_glock_dq_uninit(ghs);
-		d_instantiate(dentry, inode);
-		return PTR_RET(inode);
+		if (IS_ERR(d))
+			return PTR_RET(d);
+		return error;
 	}
 	if (error)
 		goto fail_gunlock;
@@ -686,10 +694,12 @@ static int gfs2_create_inode(struct inode *dir, struct dentry *dentry,
 		goto fail_gunlock3;
 
 	mark_inode_dirty(inode);
+	d_instantiate(dentry, inode);
+	if (file)
+		error = finish_open(file, dentry, gfs2_open_common, opened);
 	gfs2_glock_dq_uninit(ghs);
 	gfs2_glock_dq_uninit(ghs + 1);
-	d_instantiate(dentry, inode);
-	return 0;
+	return error;
 
 fail_gunlock3:
 	gfs2_glock_dq_uninit(ghs + 1);
@@ -729,36 +739,54 @@ fail:
 static int gfs2_create(struct inode *dir, struct dentry *dentry,
 		       umode_t mode, bool excl)
 {
-	return gfs2_create_inode(dir, dentry, S_IFREG | mode, 0, NULL, 0, excl);
+	return gfs2_create_inode(dir, dentry, NULL, S_IFREG | mode, 0, NULL, 0, excl, NULL);
 }
 
 /**
  * gfs2_lookup - Look up a filename in a directory and return its inode
  * @dir: The directory inode
  * @dentry: The dentry of the new inode
- * @nd: passed from Linux VFS, ignored by us
+ * @file: File to be opened
+ * @opened: atomic_open flags
  *
- * Called by the VFS layer. Lock dir and call gfs2_lookupi()
  *
  * Returns: errno
  */
 
-static struct dentry *gfs2_lookup(struct inode *dir, struct dentry *dentry,
-				  unsigned int flags)
+static struct dentry *__gfs2_lookup(struct inode *dir, struct dentry *dentry,
+				    struct file *file, int *opened)
 {
-	struct inode *inode = gfs2_lookupi(dir, &dentry->d_name, 0);
-	if (inode && !IS_ERR(inode)) {
-		struct gfs2_glock *gl = GFS2_I(inode)->i_gl;
-		struct gfs2_holder gh;
-		int error;
-		error = gfs2_glock_nq_init(gl, LM_ST_SHARED, LM_FLAG_ANY, &gh);
-		if (error) {
-			iput(inode);
-			return ERR_PTR(error);
-		}
-		gfs2_glock_dq_uninit(&gh);
+	struct inode *inode;
+	struct dentry *d;
+	struct gfs2_holder gh;
+	struct gfs2_glock *gl;
+	int error;
+
+	inode = gfs2_lookupi(dir, &dentry->d_name, 0);
+	if (!inode || IS_ERR(inode))
+		return d_splice_alias(inode, dentry);
+
+	gl = GFS2_I(inode)->i_gl;
+	error = gfs2_glock_nq_init(gl, LM_ST_SHARED, LM_FLAG_ANY, &gh);
+	if (error) {
+		iput(inode);
+		return ERR_PTR(error);
 	}
-	return d_splice_alias(inode, dentry);
+
+	d = d_splice_alias(inode, dentry);
+	if (file)
+		error = finish_open(file, dentry, gfs2_open_common, opened);
+
+	gfs2_glock_dq_uninit(&gh);
+	if (error)
+		return ERR_PTR(error);
+	return d;
+}
+
+static struct dentry *gfs2_lookup(struct inode *dir, struct dentry *dentry,
+				  unsigned flags)
+{
+	return __gfs2_lookup(dir, dentry, NULL, NULL);
 }
 
 /**
@@ -1076,7 +1104,7 @@ static int gfs2_symlink(struct inode *dir, struct dentry *dentry,
 	if (size > sdp->sd_sb.sb_bsize - sizeof(struct gfs2_dinode) - 1)
 		return -ENAMETOOLONG;
 
-	return gfs2_create_inode(dir, dentry, S_IFLNK | S_IRWXUGO, 0, symname, size, 0);
+	return gfs2_create_inode(dir, dentry, NULL, S_IFLNK | S_IRWXUGO, 0, symname, size, 0, NULL);
 }
 
 /**
@@ -1092,7 +1120,7 @@ static int gfs2_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
 {
 	struct gfs2_sbd *sdp = GFS2_SB(dir);
 	unsigned dsize = sdp->sd_sb.sb_bsize - sizeof(struct gfs2_dinode);
-	return gfs2_create_inode(dir, dentry, S_IFDIR | mode, 0, NULL, dsize, 0);
+	return gfs2_create_inode(dir, dentry, NULL, S_IFDIR | mode, 0, NULL, dsize, 0, NULL);
 }
 
 /**
@@ -1107,7 +1135,38 @@ static int gfs2_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
 static int gfs2_mknod(struct inode *dir, struct dentry *dentry, umode_t mode,
 		      dev_t dev)
 {
-	return gfs2_create_inode(dir, dentry, mode, dev, NULL, 0, 0);
+	return gfs2_create_inode(dir, dentry, NULL, mode, dev, NULL, 0, 0, NULL);
+}
+
+/**
+ * gfs2_atomic_open - Atomically open a file
+ * @dir: The directory
+ * @dentry: The proposed new entry
+ * @file: The proposed new struct file
+ * @flags: open flags
+ * @mode: File mode
+ * @opened: Flag to say whether the file has been opened or not
+ *
+ * Returns: error code or 0 for success
+ */
+
+static int gfs2_atomic_open(struct inode *dir, struct dentry *dentry,
+                            struct file *file, unsigned flags,
+                            umode_t mode, int *opened)
+{
+	struct dentry *d;
+	bool excl = !!(flags & O_EXCL);
+
+	d = __gfs2_lookup(dir, dentry, file, opened);
+	if (IS_ERR(d))
+		return PTR_ERR(d);
+	if (d)
+		return 0;
+
+	if (!(flags & O_CREAT))
+		return -ENOENT;
+
+	return gfs2_create_inode(dir, dentry, file, S_IFREG | mode, 0, NULL, 0, excl, opened);
 }
 
 /*
@@ -1787,6 +1846,7 @@ const struct inode_operations gfs2_dir_iops = {
 	.removexattr = gfs2_removexattr,
 	.fiemap = gfs2_fiemap,
 	.get_acl = gfs2_get_acl,
+	.atomic_open = gfs2_atomic_open,
 };
 
 const struct inode_operations gfs2_symlink_iops = {
diff --git a/fs/gfs2/inode.h b/fs/gfs2/inode.h
index c53c747..ba4d949 100644
--- a/fs/gfs2/inode.h
+++ b/fs/gfs2/inode.h
@@ -109,6 +109,7 @@ extern int gfs2_permission(struct inode *inode, int mask);
 extern int gfs2_setattr_simple(struct inode *inode, struct iattr *attr);
 extern struct inode *gfs2_lookup_simple(struct inode *dip, const char *name);
 extern void gfs2_dinode_out(const struct gfs2_inode *ip, void *buf);
+extern int gfs2_open_common(struct inode *inode, struct file *file);
 
 extern const struct inode_operations gfs2_file_iops;
 extern const struct inode_operations gfs2_dir_iops;

[Cluster-devel] GFS2: Add atomic_open support

[Steven Whitehouse]

The following patch implements atomic_open for GFS2. This is mostly
straightforward, however there is one corner case which I've had to
deal with, beyond what would normally be expected for a local
filesystem.

The ->atomic_open function can be passed negative dentries, which
in most cases means either ENOENT (->lookup) or a call to d_instantiate
(->create). In the GFS2 case though, we need to actually perform the
look up, since we do not know whether there has been a new inode created
on another node. The look up calls d_splice_alias which then tries to
rehash the dentry - so the solution here is to simply check for that
in d_splice_alias. The same issue is likely to affect any other cluster
filesystem implementing ->atomic_open

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: "J. Bruce Fields" <bfields@fieldses.org>
Cc: Jeff Layton <jlayton@redhat.com>

diff --git a/fs/dcache.c b/fs/dcache.c
index f09b908..5a23073 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -1612,6 +1612,10 @@ EXPORT_SYMBOL(d_obtain_alias);
  * If a dentry was found and moved, then it is returned.  Otherwise NULL
  * is returned.  This matches the expected return value of ->lookup.
  *
+ * Cluster filesystems may call this function with a negative, hashed dentry.
+ * In that case, we know that the inode will be a regular file, and also this
+ * will only occur during atomic_open. So we need to check for the dentry
+ * being already hashed only in the final case.
  */
 struct dentry *d_splice_alias(struct inode *inode, struct dentry *dentry)
 {
@@ -1636,8 +1640,11 @@ struct dentry *d_splice_alias(struct inode *inode, struct dentry *dentry)
 			security_d_instantiate(dentry, inode);
 			d_rehash(dentry);
 		}
-	} else
-		d_add(dentry, inode);
+	} else {
+		d_instantiate(dentry, inode);
+		if (d_unhashed(dentry))
+			d_rehash(dentry);
+	}
 	return new;
 }
 EXPORT_SYMBOL(d_splice_alias);
diff --git a/fs/gfs2/file.c b/fs/gfs2/file.c
index ad0dc38..4ed6a03 100644
--- a/fs/gfs2/file.c
+++ b/fs/gfs2/file.c
@@ -538,21 +538,30 @@ static int gfs2_mmap(struct file *file, struct vm_area_struct *vma)
 }
 
 /**
- * gfs2_open - open a file
- * @inode: the inode to open
- * @file: the struct file for this opening
+ * gfs2_open_common - This is common to open and atomic_open
+ * @inode: The inode being opened
+ * @file: The file being opened
  *
- * Returns: errno
+ * This maybe called under a glock or not depending upon how it has
+ * been called. We must always be called under a glock for regular
+ * files, however. For other file types, it does not matter whether
+ * we hold the glock or not.
+ *
+ * Returns: Error code or 0 for success
  */
 
-static int gfs2_open(struct inode *inode, struct file *file)
+int gfs2_open_common(struct inode *inode, struct file *file)
 {
-	struct gfs2_inode *ip = GFS2_I(inode);
-	struct gfs2_holder i_gh;
 	struct gfs2_file *fp;
-	int error;
+	int ret;
 
-	fp = kzalloc(sizeof(struct gfs2_file), GFP_KERNEL);
+	if (S_ISREG(inode->i_mode)) {
+		ret = generic_file_open(inode, file);
+		if (ret)
+			return ret;
+	}
+
+	fp = kzalloc(sizeof(struct gfs2_file), GFP_NOFS);
 	if (!fp)
 		return -ENOMEM;
 
@@ -560,29 +569,43 @@ static int gfs2_open(struct inode *inode, struct file *file)
 
 	gfs2_assert_warn(GFS2_SB(inode), !file->private_data);
 	file->private_data = fp;
+	return 0;
+}
+
+/**
+ * gfs2_open - open a file
+ * @inode: the inode to open
+ * @file: the struct file for this opening
+ *
+ * After atomic_open, this function is only used for opening files
+ * which are already cached. We must still get the glock for regular
+ * files to ensure that we have the file size uptodate for the large
+ * file check which is in the common code. That is only an issue for
+ * regular files though.
+ *
+ * Returns: errno
+ */
+
+static int gfs2_open(struct inode *inode, struct file *file)
+{
+	struct gfs2_inode *ip = GFS2_I(inode);
+	struct gfs2_holder i_gh;
+	int error;
+	bool need_unlock = false;
 
 	if (S_ISREG(ip->i_inode.i_mode)) {
 		error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_ANY,
 					   &i_gh);
 		if (error)
-			goto fail;
+			return error;
+		need_unlock = true;
+	}
 
-		if (!(file->f_flags & O_LARGEFILE) &&
-		    i_size_read(inode) > MAX_NON_LFS) {
-			error = -EOVERFLOW;
-			goto fail_gunlock;
-		}
+	error = gfs2_open_common(inode, file);
 
+	if (need_unlock)
 		gfs2_glock_dq_uninit(&i_gh);
-	}
-
-	return 0;
 
-fail_gunlock:
-	gfs2_glock_dq_uninit(&i_gh);
-fail:
-	file->private_data = NULL;
-	kfree(fp);
 	return error;
 }
 
diff --git a/fs/gfs2/inode.c b/fs/gfs2/inode.c
index 5fbb8df..065c7fb 100644
--- a/fs/gfs2/inode.c
+++ b/fs/gfs2/inode.c
@@ -546,6 +546,7 @@ static int gfs2_security_init(struct gfs2_inode *dip, struct gfs2_inode *ip,
  * gfs2_create_inode - Create a new inode
  * @dir: The parent directory
  * @dentry: The new dentry
+ * @file: If non-NULL, the file which is being opened
  * @mode: The permissions on the new inode
  * @dev: For device nodes, this is the device number
  * @symname: For symlinks, this is the link destination
@@ -555,8 +556,9 @@ static int gfs2_security_init(struct gfs2_inode *dip, struct gfs2_inode *ip,
  */
 
 static int gfs2_create_inode(struct inode *dir, struct dentry *dentry,
+			     struct file *file,
 			     umode_t mode, dev_t dev, const char *symname,
-			     unsigned int size, int excl)
+			     unsigned int size, int excl, int *opened)
 {
 	const struct qstr *name = &dentry->d_name;
 	struct gfs2_holder ghs[2];
@@ -564,6 +566,7 @@ static int gfs2_create_inode(struct inode *dir, struct dentry *dentry,
 	struct gfs2_inode *dip = GFS2_I(dir), *ip;
 	struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
 	struct gfs2_glock *io_gl;
+	struct dentry *d;
 	int error;
 	u32 aflags = 0;
 	int arq;
@@ -586,9 +589,14 @@ static int gfs2_create_inode(struct inode *dir, struct dentry *dentry,
 	error = create_ok(dip, name, mode);
 	if ((error == -EEXIST) && S_ISREG(mode) && !excl) {
 		inode = gfs2_lookupi(dir, &dentry->d_name, 0);
+		d = d_splice_alias(inode, dentry);
+		error = 0;
+		if (file && !IS_ERR(d))
+			error = finish_open(file, dentry, gfs2_open_common, opened);
 		gfs2_glock_dq_uninit(ghs);
-		d_instantiate(dentry, inode);
-		return PTR_RET(inode);
+		if (IS_ERR(d))
+			return PTR_RET(d);
+		return error;
 	}
 	if (error)
 		goto fail_gunlock;
@@ -686,10 +694,12 @@ static int gfs2_create_inode(struct inode *dir, struct dentry *dentry,
 		goto fail_gunlock3;
 
 	mark_inode_dirty(inode);
+	d_instantiate(dentry, inode);
+	if (file)
+		error = finish_open(file, dentry, gfs2_open_common, opened);
 	gfs2_glock_dq_uninit(ghs);
 	gfs2_glock_dq_uninit(ghs + 1);
-	d_instantiate(dentry, inode);
-	return 0;
+	return error;
 
 fail_gunlock3:
 	gfs2_glock_dq_uninit(ghs + 1);
@@ -729,36 +739,54 @@ fail:
 static int gfs2_create(struct inode *dir, struct dentry *dentry,
 		       umode_t mode, bool excl)
 {
-	return gfs2_create_inode(dir, dentry, S_IFREG | mode, 0, NULL, 0, excl);
+	return gfs2_create_inode(dir, dentry, NULL, S_IFREG | mode, 0, NULL, 0, excl, NULL);
 }
 
 /**
  * gfs2_lookup - Look up a filename in a directory and return its inode
  * @dir: The directory inode
  * @dentry: The dentry of the new inode
- * @nd: passed from Linux VFS, ignored by us
+ * @file: File to be opened
+ * @opened: atomic_open flags
  *
- * Called by the VFS layer. Lock dir and call gfs2_lookupi()
  *
  * Returns: errno
  */
 
-static struct dentry *gfs2_lookup(struct inode *dir, struct dentry *dentry,
-				  unsigned int flags)
+static struct dentry *__gfs2_lookup(struct inode *dir, struct dentry *dentry,
+				    struct file *file, int *opened)
 {
-	struct inode *inode = gfs2_lookupi(dir, &dentry->d_name, 0);
-	if (inode && !IS_ERR(inode)) {
-		struct gfs2_glock *gl = GFS2_I(inode)->i_gl;
-		struct gfs2_holder gh;
-		int error;
-		error = gfs2_glock_nq_init(gl, LM_ST_SHARED, LM_FLAG_ANY, &gh);
-		if (error) {
-			iput(inode);
-			return ERR_PTR(error);
-		}
-		gfs2_glock_dq_uninit(&gh);
+	struct inode *inode;
+	struct dentry *d;
+	struct gfs2_holder gh;
+	struct gfs2_glock *gl;
+	int error;
+
+	inode = gfs2_lookupi(dir, &dentry->d_name, 0);
+	if (!inode || IS_ERR(inode))
+		return d_splice_alias(inode, dentry);
+
+	gl = GFS2_I(inode)->i_gl;
+	error = gfs2_glock_nq_init(gl, LM_ST_SHARED, LM_FLAG_ANY, &gh);
+	if (error) {
+		iput(inode);
+		return ERR_PTR(error);
 	}
-	return d_splice_alias(inode, dentry);
+
+	d = d_splice_alias(inode, dentry);
+	if (file)
+		error = finish_open(file, dentry, gfs2_open_common, opened);
+
+	gfs2_glock_dq_uninit(&gh);
+	if (error)
+		return ERR_PTR(error);
+	return d;
+}
+
+static struct dentry *gfs2_lookup(struct inode *dir, struct dentry *dentry,
+				  unsigned flags)
+{
+	return __gfs2_lookup(dir, dentry, NULL, NULL);
 }
 
 /**
@@ -1076,7 +1104,7 @@ static int gfs2_symlink(struct inode *dir, struct dentry *dentry,
 	if (size > sdp->sd_sb.sb_bsize - sizeof(struct gfs2_dinode) - 1)
 		return -ENAMETOOLONG;
 
-	return gfs2_create_inode(dir, dentry, S_IFLNK | S_IRWXUGO, 0, symname, size, 0);
+	return gfs2_create_inode(dir, dentry, NULL, S_IFLNK | S_IRWXUGO, 0, symname, size, 0, NULL);
 }
 
 /**
@@ -1092,7 +1120,7 @@ static int gfs2_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
 {
 	struct gfs2_sbd *sdp = GFS2_SB(dir);
 	unsigned dsize = sdp->sd_sb.sb_bsize - sizeof(struct gfs2_dinode);
-	return gfs2_create_inode(dir, dentry, S_IFDIR | mode, 0, NULL, dsize, 0);
+	return gfs2_create_inode(dir, dentry, NULL, S_IFDIR | mode, 0, NULL, dsize, 0, NULL);
 }
 
 /**
@@ -1107,7 +1135,38 @@ static int gfs2_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
 static int gfs2_mknod(struct inode *dir, struct dentry *dentry, umode_t mode,
 		      dev_t dev)
 {
-	return gfs2_create_inode(dir, dentry, mode, dev, NULL, 0, 0);
+	return gfs2_create_inode(dir, dentry, NULL, mode, dev, NULL, 0, 0, NULL);
+}
+
+/**
+ * gfs2_atomic_open - Atomically open a file
+ * @dir: The directory
+ * @dentry: The proposed new entry
+ * @file: The proposed new struct file
+ * @flags: open flags
+ * @mode: File mode
+ * @opened: Flag to say whether the file has been opened or not
+ *
+ * Returns: error code or 0 for success
+ */
+
+static int gfs2_atomic_open(struct inode *dir, struct dentry *dentry,
+                            struct file *file, unsigned flags,
+                            umode_t mode, int *opened)
+{
+	struct dentry *d;
+	bool excl = !!(flags & O_EXCL);
+
+	d = __gfs2_lookup(dir, dentry, file, opened);
+	if (IS_ERR(d))
+		return PTR_ERR(d);
+	if (d)
+		return 0;
+
+	if (!(flags & O_CREAT))
+		return -ENOENT;
+
+	return gfs2_create_inode(dir, dentry, file, S_IFREG | mode, 0, NULL, 0, excl, opened);
 }
 
 /*
@@ -1787,6 +1846,7 @@ const struct inode_operations gfs2_dir_iops = {
 	.removexattr = gfs2_removexattr,
 	.fiemap = gfs2_fiemap,
 	.get_acl = gfs2_get_acl,
+	.atomic_open = gfs2_atomic_open,
 };
 
 const struct inode_operations gfs2_symlink_iops = {
diff --git a/fs/gfs2/inode.h b/fs/gfs2/inode.h
index c53c747..ba4d949 100644
--- a/fs/gfs2/inode.h
+++ b/fs/gfs2/inode.h
@@ -109,6 +109,7 @@ extern int gfs2_permission(struct inode *inode, int mask);
 extern int gfs2_setattr_simple(struct inode *inode, struct iattr *attr);
 extern struct inode *gfs2_lookup_simple(struct inode *dip, const char *name);
 extern void gfs2_dinode_out(const struct gfs2_inode *ip, void *buf);
+extern int gfs2_open_common(struct inode *inode, struct file *file);
 
 extern const struct inode_operations gfs2_file_iops;
 extern const struct inode_operations gfs2_dir_iops;

Re: GFS2: Add atomic_open support

[Al Viro]

On Thu, Jun 06, 2013 at 03:50:12PM +0100, Steven Whitehouse wrote:
> 
> The following patch implements atomic_open for GFS2. This is mostly
> straightforward, however there is one corner case which I've had to
> deal with, beyond what would normally be expected for a local
> filesystem.

Broken - what will happen if you hit a symlink, for starters?  On everything
handled locally you should just return finish_no_open(file, dentry) and
let the caller deal with that; the only cases that might make sense to
handle in ->atomic_open() are regular files and directories.  For gfs2 it
should be just regular files.  While we are at it, do you even need
->private_data for gfs2 directories?

[Cluster-devel] GFS2: Add atomic_open support

[Al Viro]

On Thu, Jun 06, 2013 at 03:50:12PM +0100, Steven Whitehouse wrote:
> 
> The following patch implements atomic_open for GFS2. This is mostly
> straightforward, however there is one corner case which I've had to
> deal with, beyond what would normally be expected for a local
> filesystem.

Broken - what will happen if you hit a symlink, for starters?  On everything
handled locally you should just return finish_no_open(file, dentry) and
let the caller deal with that; the only cases that might make sense to
handle in ->atomic_open() are regular files and directories.  For gfs2 it
should be just regular files.  While we are at it, do you even need
->private_data for gfs2 directories?

Re: GFS2: Add atomic_open support

[Steven Whitehouse]

Hi,

On Thu, 2013-06-06 at 18:53 +0100, Al Viro wrote:
> On Thu, Jun 06, 2013 at 03:50:12PM +0100, Steven Whitehouse wrote:
> > 
> > The following patch implements atomic_open for GFS2. This is mostly
> > straightforward, however there is one corner case which I've had to
> > deal with, beyond what would normally be expected for a local
> > filesystem.
> 
> Broken - what will happen if you hit a symlink, for starters?  On everything
> handled locally you should just return finish_no_open(file, dentry) and
> let the caller deal with that; the only cases that might make sense to
> handle in ->atomic_open() are regular files and directories.  For gfs2 it
> should be just regular files.  While we are at it, do you even need
> ->private_data for gfs2 directories?

I'm brewing up another version of the patch at the moment, which I will
post shortly. I don't understand why GFS2 specifically would not want to
do this with both files and directories? Why would it be different to
other filesystems in that respect?

We do need private data for gfs2 directories because that is required
for flock which should work equally well on directories as on regular
files. Potentially we might allocate it later, on the first call to
flock for example, but it has been done at open time so that we don't
have to test for it on every flock call. Also I was hoping to put some
info in there to assist with directory readahead at some stage too,

Steve.

[Cluster-devel] GFS2: Add atomic_open support

[Steven Whitehouse]

Hi,

On Thu, 2013-06-06 at 18:53 +0100, Al Viro wrote:
> On Thu, Jun 06, 2013 at 03:50:12PM +0100, Steven Whitehouse wrote:
> > 
> > The following patch implements atomic_open for GFS2. This is mostly
> > straightforward, however there is one corner case which I've had to
> > deal with, beyond what would normally be expected for a local
> > filesystem.
> 
> Broken - what will happen if you hit a symlink, for starters?  On everything
> handled locally you should just return finish_no_open(file, dentry) and
> let the caller deal with that; the only cases that might make sense to
> handle in ->atomic_open() are regular files and directories.  For gfs2 it
> should be just regular files.  While we are at it, do you even need
> ->private_data for gfs2 directories?

I'm brewing up another version of the patch at the moment, which I will
post shortly. I don't understand why GFS2 specifically would not want to
do this with both files and directories? Why would it be different to
other filesystems in that respect?

We do need private data for gfs2 directories because that is required
for flock which should work equally well on directories as on regular
files. Potentially we might allocate it later, on the first call to
flock for example, but it has been done at open time so that we don't
have to test for it on every flock call. Also I was hoping to put some
info in there to assist with directory readahead at some stage too,

Steve.

Re: GFS2: Add atomic_open support

[Steven Whitehouse]

This is the second attempt at an atomic_open patch for GFS2,
now updated in the light of Al's comments on the first patch.

Also, this patch is designed to apply over this one:
https://www.redhat.com/archives/cluster-devel/2013-June/msg00098.html

I've restricted atomic_open to only operate on regular files, although
I still don't understand why atomic_open should not be possible also for
directories on GFS2. That can always be added in later though, if it
makes sense.

The ->atomic_open function can be passed negative dentries, which
in most cases means either ENOENT (->lookup) or a call to d_instantiate
(->create). In the GFS2 case though, we need to actually perform the
look up, since we do not know whether there has been a new inode created
on another node. The look up calls d_splice_alias which then tries to
rehash the dentry - so the solution here is to simply check for that
in d_splice_alias. The same issue is likely to affect any other cluster
filesystem implementing ->atomic_open

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: "J. Bruce Fields" <bfields fieldses org>
Cc: Jeff Layton <jlayton@redhat.com>

diff --git a/fs/dcache.c b/fs/dcache.c
index f09b908..5a23073 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -1612,6 +1612,10 @@ EXPORT_SYMBOL(d_obtain_alias);
  * If a dentry was found and moved, then it is returned.  Otherwise NULL
  * is returned.  This matches the expected return value of ->lookup.
  *
+ * Cluster filesystems may call this function with a negative, hashed dentry.
+ * In that case, we know that the inode will be a regular file, and also this
+ * will only occur during atomic_open. So we need to check for the dentry
+ * being already hashed only in the final case.
  */
 struct dentry *d_splice_alias(struct inode *inode, struct dentry *dentry)
 {
@@ -1636,8 +1640,11 @@ struct dentry *d_splice_alias(struct inode *inode, struct dentry *dentry)
 			security_d_instantiate(dentry, inode);
 			d_rehash(dentry);
 		}
-	} else
-		d_add(dentry, inode);
+	} else {
+		d_instantiate(dentry, inode);
+		if (d_unhashed(dentry))
+			d_rehash(dentry);
+	}
 	return new;
 }
 EXPORT_SYMBOL(d_splice_alias);
diff --git a/fs/gfs2/file.c b/fs/gfs2/file.c
index ad0dc38..4ed6a03 100644
--- a/fs/gfs2/file.c
+++ b/fs/gfs2/file.c
@@ -538,21 +538,30 @@ static int gfs2_mmap(struct file *file, struct vm_area_struct *vma)
 }
 
 /**
- * gfs2_open - open a file
- * @inode: the inode to open
- * @file: the struct file for this opening
+ * gfs2_open_common - This is common to open and atomic_open
+ * @inode: The inode being opened
+ * @file: The file being opened
  *
- * Returns: errno
+ * This maybe called under a glock or not depending upon how it has
+ * been called. We must always be called under a glock for regular
+ * files, however. For other file types, it does not matter whether
+ * we hold the glock or not.
+ *
+ * Returns: Error code or 0 for success
  */
 
-static int gfs2_open(struct inode *inode, struct file *file)
+int gfs2_open_common(struct inode *inode, struct file *file)
 {
-	struct gfs2_inode *ip = GFS2_I(inode);
-	struct gfs2_holder i_gh;
 	struct gfs2_file *fp;
-	int error;
+	int ret;
 
-	fp = kzalloc(sizeof(struct gfs2_file), GFP_KERNEL);
+	if (S_ISREG(inode->i_mode)) {
+		ret = generic_file_open(inode, file);
+		if (ret)
+			return ret;
+	}
+
+	fp = kzalloc(sizeof(struct gfs2_file), GFP_NOFS);
 	if (!fp)
 		return -ENOMEM;
 
@@ -560,29 +569,43 @@ static int gfs2_open(struct inode *inode, struct file *file)
 
 	gfs2_assert_warn(GFS2_SB(inode), !file->private_data);
 	file->private_data = fp;
+	return 0;
+}
+
+/**
+ * gfs2_open - open a file
+ * @inode: the inode to open
+ * @file: the struct file for this opening
+ *
+ * After atomic_open, this function is only used for opening files
+ * which are already cached. We must still get the glock for regular
+ * files to ensure that we have the file size uptodate for the large
+ * file check which is in the common code. That is only an issue for
+ * regular files though.
+ *
+ * Returns: errno
+ */
+
+static int gfs2_open(struct inode *inode, struct file *file)
+{
+	struct gfs2_inode *ip = GFS2_I(inode);
+	struct gfs2_holder i_gh;
+	int error;
+	bool need_unlock = false;
 
 	if (S_ISREG(ip->i_inode.i_mode)) {
 		error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_ANY,
 					   &i_gh);
 		if (error)
-			goto fail;
+			return error;
+		need_unlock = true;
+	}
 
-		if (!(file->f_flags & O_LARGEFILE) &&
-		    i_size_read(inode) > MAX_NON_LFS) {
-			error = -EOVERFLOW;
-			goto fail_gunlock;
-		}
+	error = gfs2_open_common(inode, file);
 
+	if (need_unlock)
 		gfs2_glock_dq_uninit(&i_gh);
-	}
-
-	return 0;
 
-fail_gunlock:
-	gfs2_glock_dq_uninit(&i_gh);
-fail:
-	file->private_data = NULL;
-	kfree(fp);
 	return error;
 }
 
diff --git a/fs/gfs2/inode.c b/fs/gfs2/inode.c
index ede16ae..bbb2715 100644
--- a/fs/gfs2/inode.c
+++ b/fs/gfs2/inode.c
@@ -535,6 +535,7 @@ static int gfs2_security_init(struct gfs2_inode *dip, struct gfs2_inode *ip,
  * gfs2_create_inode - Create a new inode
  * @dir: The parent directory
  * @dentry: The new dentry
+ * @file: If non-NULL, the file which is being opened
  * @mode: The permissions on the new inode
  * @dev: For device nodes, this is the device number
  * @symname: For symlinks, this is the link destination
@@ -544,8 +545,9 @@ static int gfs2_security_init(struct gfs2_inode *dip, struct gfs2_inode *ip,
  */
 
 static int gfs2_create_inode(struct inode *dir, struct dentry *dentry,
+			     struct file *file,
 			     umode_t mode, dev_t dev, const char *symname,
-			     unsigned int size, int excl)
+			     unsigned int size, int excl, int *opened)
 {
 	const struct qstr *name = &dentry->d_name;
 	struct gfs2_holder ghs[2];
@@ -553,6 +555,7 @@ static int gfs2_create_inode(struct inode *dir, struct dentry *dentry,
 	struct gfs2_inode *dip = GFS2_I(dir), *ip;
 	struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
 	struct gfs2_glock *io_gl;
+	struct dentry *d;
 	int error;
 	u32 aflags = 0;
 	int arq;
@@ -579,9 +582,20 @@ static int gfs2_create_inode(struct inode *dir, struct dentry *dentry,
 	inode = gfs2_dir_search(dir, &dentry->d_name, !S_ISREG(mode) || excl);
 	error = PTR_ERR(inode);
 	if (!IS_ERR(inode)) {
+		d = d_splice_alias(inode, dentry);
+		error = 0;
+		if (file && !IS_ERR(d)) {
+			if (d == NULL)
+				d = dentry;
+			if (S_ISREG(inode->i_mode))
+				error = finish_open(file, d, gfs2_open_common, opened);
+			else
+				error = finish_no_open(file, d);
+		}
 		gfs2_glock_dq_uninit(ghs);
-		d_instantiate(dentry, inode);
-		return 0;
+		if (IS_ERR(d))
+			return PTR_RET(d);
+		return error;
 	} else if (error != -ENOENT) {
 		goto fail_gunlock;
 	}
@@ -679,10 +693,12 @@ static int gfs2_create_inode(struct inode *dir, struct dentry *dentry,
 		goto fail_gunlock3;
 
 	mark_inode_dirty(inode);
+	d_instantiate(dentry, inode);
+	if (file)
+		error = finish_open(file, dentry, gfs2_open_common, opened);
 	gfs2_glock_dq_uninit(ghs);
 	gfs2_glock_dq_uninit(ghs + 1);
-	d_instantiate(dentry, inode);
-	return 0;
+	return error;
 
 fail_gunlock3:
 	gfs2_glock_dq_uninit(ghs + 1);
@@ -722,36 +738,56 @@ fail:
 static int gfs2_create(struct inode *dir, struct dentry *dentry,
 		       umode_t mode, bool excl)
 {
-	return gfs2_create_inode(dir, dentry, S_IFREG | mode, 0, NULL, 0, excl);
+	return gfs2_create_inode(dir, dentry, NULL, S_IFREG | mode, 0, NULL, 0, excl, NULL);
 }
 
 /**
- * gfs2_lookup - Look up a filename in a directory and return its inode
+ * __gfs2_lookup - Look up a filename in a directory and return its inode
  * @dir: The directory inode
  * @dentry: The dentry of the new inode
- * @nd: passed from Linux VFS, ignored by us
+ * @file: File to be opened
+ * @opened: atomic_open flags
  *
- * Called by the VFS layer. Lock dir and call gfs2_lookupi()
  *
  * Returns: errno
  */
 
-static struct dentry *gfs2_lookup(struct inode *dir, struct dentry *dentry,
-				  unsigned int flags)
+static struct dentry *__gfs2_lookup(struct inode *dir, struct dentry *dentry,
+				    struct file *file, int *opened)
 {
-	struct inode *inode = gfs2_lookupi(dir, &dentry->d_name, 0);
-	if (inode && !IS_ERR(inode)) {
-		struct gfs2_glock *gl = GFS2_I(inode)->i_gl;
-		struct gfs2_holder gh;
-		int error;
-		error = gfs2_glock_nq_init(gl, LM_ST_SHARED, LM_FLAG_ANY, &gh);
-		if (error) {
-			iput(inode);
-			return ERR_PTR(error);
-		}
-		gfs2_glock_dq_uninit(&gh);
+	struct inode *inode;
+	struct dentry *d;
+	struct gfs2_holder gh;
+	struct gfs2_glock *gl;
+	int error;
+
+	inode = gfs2_lookupi(dir, &dentry->d_name, 0);
+	if (!inode)
+		return NULL;
+	if (IS_ERR(inode))
+		return ERR_CAST(inode);
+
+	gl = GFS2_I(inode)->i_gl;
+	error = gfs2_glock_nq_init(gl, LM_ST_SHARED, LM_FLAG_ANY, &gh);
+	if (error) {
+		iput(inode);
+		return ERR_PTR(error);
 	}
-	return d_splice_alias(inode, dentry);
+
+	d = d_splice_alias(inode, dentry);
+	if (file && S_ISREG(inode->i_mode))
+		error = finish_open(file, dentry, gfs2_open_common, opened);
+
+	gfs2_glock_dq_uninit(&gh);
+	if (error)
+		return ERR_PTR(error);
+	return d;
+}
+
+static struct dentry *gfs2_lookup(struct inode *dir, struct dentry *dentry,
+				  unsigned flags)
+{
+	return __gfs2_lookup(dir, dentry, NULL, NULL);
 }
 
 /**
@@ -1069,7 +1105,7 @@ static int gfs2_symlink(struct inode *dir, struct dentry *dentry,
 	if (size > sdp->sd_sb.sb_bsize - sizeof(struct gfs2_dinode) - 1)
 		return -ENAMETOOLONG;
 
-	return gfs2_create_inode(dir, dentry, S_IFLNK | S_IRWXUGO, 0, symname, size, 0);
+	return gfs2_create_inode(dir, dentry, NULL, S_IFLNK | S_IRWXUGO, 0, symname, size, 0, NULL);
 }
 
 /**
@@ -1085,7 +1121,7 @@ static int gfs2_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
 {
 	struct gfs2_sbd *sdp = GFS2_SB(dir);
 	unsigned dsize = sdp->sd_sb.sb_bsize - sizeof(struct gfs2_dinode);
-	return gfs2_create_inode(dir, dentry, S_IFDIR | mode, 0, NULL, dsize, 0);
+	return gfs2_create_inode(dir, dentry, NULL, S_IFDIR | mode, 0, NULL, dsize, 0, NULL);
 }
 
 /**
@@ -1100,7 +1136,43 @@ static int gfs2_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
 static int gfs2_mknod(struct inode *dir, struct dentry *dentry, umode_t mode,
 		      dev_t dev)
 {
-	return gfs2_create_inode(dir, dentry, mode, dev, NULL, 0, 0);
+	return gfs2_create_inode(dir, dentry, NULL, mode, dev, NULL, 0, 0, NULL);
+}
+
+/**
+ * gfs2_atomic_open - Atomically open a file
+ * @dir: The directory
+ * @dentry: The proposed new entry
+ * @file: The proposed new struct file
+ * @flags: open flags
+ * @mode: File mode
+ * @opened: Flag to say whether the file has been opened or not
+ *
+ * Returns: error code or 0 for success
+ */
+
+static int gfs2_atomic_open(struct inode *dir, struct dentry *dentry,
+                            struct file *file, unsigned flags,
+                            umode_t mode, int *opened)
+{
+	struct dentry *d;
+	bool excl = !!(flags & O_EXCL);
+
+	d = __gfs2_lookup(dir, dentry, file, opened);
+	if (IS_ERR(d))
+		return PTR_ERR(d);
+	if (d == NULL)
+		d = dentry;
+	if (d->d_inode) {
+		if (!(*opened & FILE_OPENED))
+			return finish_no_open(file, d);
+		return 0;
+	}
+
+	if (!(flags & O_CREAT))
+		return -ENOENT;
+
+	return gfs2_create_inode(dir, dentry, file, S_IFREG | mode, 0, NULL, 0, excl, opened);
 }
 
 /*
@@ -1780,6 +1852,7 @@ const struct inode_operations gfs2_dir_iops = {
 	.removexattr = gfs2_removexattr,
 	.fiemap = gfs2_fiemap,
 	.get_acl = gfs2_get_acl,
+	.atomic_open = gfs2_atomic_open,
 };
 
 const struct inode_operations gfs2_symlink_iops = {
diff --git a/fs/gfs2/inode.h b/fs/gfs2/inode.h
index c53c747..ba4d949 100644
--- a/fs/gfs2/inode.h
+++ b/fs/gfs2/inode.h
@@ -109,6 +109,7 @@ extern int gfs2_permission(struct inode *inode, int mask);
 extern int gfs2_setattr_simple(struct inode *inode, struct iattr *attr);
 extern struct inode *gfs2_lookup_simple(struct inode *dip, const char *name);
 extern void gfs2_dinode_out(const struct gfs2_inode *ip, void *buf);
+extern int gfs2_open_common(struct inode *inode, struct file *file);
 
 extern const struct inode_operations gfs2_file_iops;
 extern const struct inode_operations gfs2_dir_iops;

[Cluster-devel] GFS2: Add atomic_open support

[Steven Whitehouse]

This is the second attempt at an atomic_open patch for GFS2,
now updated in the light of Al's comments on the first patch.

Also, this patch is designed to apply over this one:
https://www.redhat.com/archives/cluster-devel/2013-June/msg00098.html

I've restricted atomic_open to only operate on regular files, although
I still don't understand why atomic_open should not be possible also for
directories on GFS2. That can always be added in later though, if it
makes sense.

The ->atomic_open function can be passed negative dentries, which
in most cases means either ENOENT (->lookup) or a call to d_instantiate
(->create). In the GFS2 case though, we need to actually perform the
look up, since we do not know whether there has been a new inode created
on another node. The look up calls d_splice_alias which then tries to
rehash the dentry - so the solution here is to simply check for that
in d_splice_alias. The same issue is likely to affect any other cluster
filesystem implementing ->atomic_open

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: "J. Bruce Fields" <bfields fieldses org>
Cc: Jeff Layton <jlayton@redhat.com>

diff --git a/fs/dcache.c b/fs/dcache.c
index f09b908..5a23073 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -1612,6 +1612,10 @@ EXPORT_SYMBOL(d_obtain_alias);
  * If a dentry was found and moved, then it is returned.  Otherwise NULL
  * is returned.  This matches the expected return value of ->lookup.
  *
+ * Cluster filesystems may call this function with a negative, hashed dentry.
+ * In that case, we know that the inode will be a regular file, and also this
+ * will only occur during atomic_open. So we need to check for the dentry
+ * being already hashed only in the final case.
  */
 struct dentry *d_splice_alias(struct inode *inode, struct dentry *dentry)
 {
@@ -1636,8 +1640,11 @@ struct dentry *d_splice_alias(struct inode *inode, struct dentry *dentry)
 			security_d_instantiate(dentry, inode);
 			d_rehash(dentry);
 		}
-	} else
-		d_add(dentry, inode);
+	} else {
+		d_instantiate(dentry, inode);
+		if (d_unhashed(dentry))
+			d_rehash(dentry);
+	}
 	return new;
 }
 EXPORT_SYMBOL(d_splice_alias);
diff --git a/fs/gfs2/file.c b/fs/gfs2/file.c
index ad0dc38..4ed6a03 100644
--- a/fs/gfs2/file.c
+++ b/fs/gfs2/file.c
@@ -538,21 +538,30 @@ static int gfs2_mmap(struct file *file, struct vm_area_struct *vma)
 }
 
 /**
- * gfs2_open - open a file
- * @inode: the inode to open
- * @file: the struct file for this opening
+ * gfs2_open_common - This is common to open and atomic_open
+ * @inode: The inode being opened
+ * @file: The file being opened
  *
- * Returns: errno
+ * This maybe called under a glock or not depending upon how it has
+ * been called. We must always be called under a glock for regular
+ * files, however. For other file types, it does not matter whether
+ * we hold the glock or not.
+ *
+ * Returns: Error code or 0 for success
  */
 
-static int gfs2_open(struct inode *inode, struct file *file)
+int gfs2_open_common(struct inode *inode, struct file *file)
 {
-	struct gfs2_inode *ip = GFS2_I(inode);
-	struct gfs2_holder i_gh;
 	struct gfs2_file *fp;
-	int error;
+	int ret;
 
-	fp = kzalloc(sizeof(struct gfs2_file), GFP_KERNEL);
+	if (S_ISREG(inode->i_mode)) {
+		ret = generic_file_open(inode, file);
+		if (ret)
+			return ret;
+	}
+
+	fp = kzalloc(sizeof(struct gfs2_file), GFP_NOFS);
 	if (!fp)
 		return -ENOMEM;
 
@@ -560,29 +569,43 @@ static int gfs2_open(struct inode *inode, struct file *file)
 
 	gfs2_assert_warn(GFS2_SB(inode), !file->private_data);
 	file->private_data = fp;
+	return 0;
+}
+
+/**
+ * gfs2_open - open a file
+ * @inode: the inode to open
+ * @file: the struct file for this opening
+ *
+ * After atomic_open, this function is only used for opening files
+ * which are already cached. We must still get the glock for regular
+ * files to ensure that we have the file size uptodate for the large
+ * file check which is in the common code. That is only an issue for
+ * regular files though.
+ *
+ * Returns: errno
+ */
+
+static int gfs2_open(struct inode *inode, struct file *file)
+{
+	struct gfs2_inode *ip = GFS2_I(inode);
+	struct gfs2_holder i_gh;
+	int error;
+	bool need_unlock = false;
 
 	if (S_ISREG(ip->i_inode.i_mode)) {
 		error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_ANY,
 					   &i_gh);
 		if (error)
-			goto fail;
+			return error;
+		need_unlock = true;
+	}
 
-		if (!(file->f_flags & O_LARGEFILE) &&
-		    i_size_read(inode) > MAX_NON_LFS) {
-			error = -EOVERFLOW;
-			goto fail_gunlock;
-		}
+	error = gfs2_open_common(inode, file);
 
+	if (need_unlock)
 		gfs2_glock_dq_uninit(&i_gh);
-	}
-
-	return 0;
 
-fail_gunlock:
-	gfs2_glock_dq_uninit(&i_gh);
-fail:
-	file->private_data = NULL;
-	kfree(fp);
 	return error;
 }
 
diff --git a/fs/gfs2/inode.c b/fs/gfs2/inode.c
index ede16ae..bbb2715 100644
--- a/fs/gfs2/inode.c
+++ b/fs/gfs2/inode.c
@@ -535,6 +535,7 @@ static int gfs2_security_init(struct gfs2_inode *dip, struct gfs2_inode *ip,
  * gfs2_create_inode - Create a new inode
  * @dir: The parent directory
  * @dentry: The new dentry
+ * @file: If non-NULL, the file which is being opened
  * @mode: The permissions on the new inode
  * @dev: For device nodes, this is the device number
  * @symname: For symlinks, this is the link destination
@@ -544,8 +545,9 @@ static int gfs2_security_init(struct gfs2_inode *dip, struct gfs2_inode *ip,
  */
 
 static int gfs2_create_inode(struct inode *dir, struct dentry *dentry,
+			     struct file *file,
 			     umode_t mode, dev_t dev, const char *symname,
-			     unsigned int size, int excl)
+			     unsigned int size, int excl, int *opened)
 {
 	const struct qstr *name = &dentry->d_name;
 	struct gfs2_holder ghs[2];
@@ -553,6 +555,7 @@ static int gfs2_create_inode(struct inode *dir, struct dentry *dentry,
 	struct gfs2_inode *dip = GFS2_I(dir), *ip;
 	struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
 	struct gfs2_glock *io_gl;
+	struct dentry *d;
 	int error;
 	u32 aflags = 0;
 	int arq;
@@ -579,9 +582,20 @@ static int gfs2_create_inode(struct inode *dir, struct dentry *dentry,
 	inode = gfs2_dir_search(dir, &dentry->d_name, !S_ISREG(mode) || excl);
 	error = PTR_ERR(inode);
 	if (!IS_ERR(inode)) {
+		d = d_splice_alias(inode, dentry);
+		error = 0;
+		if (file && !IS_ERR(d)) {
+			if (d == NULL)
+				d = dentry;
+			if (S_ISREG(inode->i_mode))
+				error = finish_open(file, d, gfs2_open_common, opened);
+			else
+				error = finish_no_open(file, d);
+		}
 		gfs2_glock_dq_uninit(ghs);
-		d_instantiate(dentry, inode);
-		return 0;
+		if (IS_ERR(d))
+			return PTR_RET(d);
+		return error;
 	} else if (error != -ENOENT) {
 		goto fail_gunlock;
 	}
@@ -679,10 +693,12 @@ static int gfs2_create_inode(struct inode *dir, struct dentry *dentry,
 		goto fail_gunlock3;
 
 	mark_inode_dirty(inode);
+	d_instantiate(dentry, inode);
+	if (file)
+		error = finish_open(file, dentry, gfs2_open_common, opened);
 	gfs2_glock_dq_uninit(ghs);
 	gfs2_glock_dq_uninit(ghs + 1);
-	d_instantiate(dentry, inode);
-	return 0;
+	return error;
 
 fail_gunlock3:
 	gfs2_glock_dq_uninit(ghs + 1);
@@ -722,36 +738,56 @@ fail:
 static int gfs2_create(struct inode *dir, struct dentry *dentry,
 		       umode_t mode, bool excl)
 {
-	return gfs2_create_inode(dir, dentry, S_IFREG | mode, 0, NULL, 0, excl);
+	return gfs2_create_inode(dir, dentry, NULL, S_IFREG | mode, 0, NULL, 0, excl, NULL);
 }
 
 /**
- * gfs2_lookup - Look up a filename in a directory and return its inode
+ * __gfs2_lookup - Look up a filename in a directory and return its inode
  * @dir: The directory inode
  * @dentry: The dentry of the new inode
- * @nd: passed from Linux VFS, ignored by us
+ * @file: File to be opened
+ * @opened: atomic_open flags
  *
- * Called by the VFS layer. Lock dir and call gfs2_lookupi()
  *
  * Returns: errno
  */
 
-static struct dentry *gfs2_lookup(struct inode *dir, struct dentry *dentry,
-				  unsigned int flags)
+static struct dentry *__gfs2_lookup(struct inode *dir, struct dentry *dentry,
+				    struct file *file, int *opened)
 {
-	struct inode *inode = gfs2_lookupi(dir, &dentry->d_name, 0);
-	if (inode && !IS_ERR(inode)) {
-		struct gfs2_glock *gl = GFS2_I(inode)->i_gl;
-		struct gfs2_holder gh;
-		int error;
-		error = gfs2_glock_nq_init(gl, LM_ST_SHARED, LM_FLAG_ANY, &gh);
-		if (error) {
-			iput(inode);
-			return ERR_PTR(error);
-		}
-		gfs2_glock_dq_uninit(&gh);
+	struct inode *inode;
+	struct dentry *d;
+	struct gfs2_holder gh;
+	struct gfs2_glock *gl;
+	int error;
+
+	inode = gfs2_lookupi(dir, &dentry->d_name, 0);
+	if (!inode)
+		return NULL;
+	if (IS_ERR(inode))
+		return ERR_CAST(inode);
+
+	gl = GFS2_I(inode)->i_gl;
+	error = gfs2_glock_nq_init(gl, LM_ST_SHARED, LM_FLAG_ANY, &gh);
+	if (error) {
+		iput(inode);
+		return ERR_PTR(error);
 	}
-	return d_splice_alias(inode, dentry);
+
+	d = d_splice_alias(inode, dentry);
+	if (file && S_ISREG(inode->i_mode))
+		error = finish_open(file, dentry, gfs2_open_common, opened);
+
+	gfs2_glock_dq_uninit(&gh);
+	if (error)
+		return ERR_PTR(error);
+	return d;
+}
+
+static struct dentry *gfs2_lookup(struct inode *dir, struct dentry *dentry,
+				  unsigned flags)
+{
+	return __gfs2_lookup(dir, dentry, NULL, NULL);
 }
 
 /**
@@ -1069,7 +1105,7 @@ static int gfs2_symlink(struct inode *dir, struct dentry *dentry,
 	if (size > sdp->sd_sb.sb_bsize - sizeof(struct gfs2_dinode) - 1)
 		return -ENAMETOOLONG;
 
-	return gfs2_create_inode(dir, dentry, S_IFLNK | S_IRWXUGO, 0, symname, size, 0);
+	return gfs2_create_inode(dir, dentry, NULL, S_IFLNK | S_IRWXUGO, 0, symname, size, 0, NULL);
 }
 
 /**
@@ -1085,7 +1121,7 @@ static int gfs2_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
 {
 	struct gfs2_sbd *sdp = GFS2_SB(dir);
 	unsigned dsize = sdp->sd_sb.sb_bsize - sizeof(struct gfs2_dinode);
-	return gfs2_create_inode(dir, dentry, S_IFDIR | mode, 0, NULL, dsize, 0);
+	return gfs2_create_inode(dir, dentry, NULL, S_IFDIR | mode, 0, NULL, dsize, 0, NULL);
 }
 
 /**
@@ -1100,7 +1136,43 @@ static int gfs2_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
 static int gfs2_mknod(struct inode *dir, struct dentry *dentry, umode_t mode,
 		      dev_t dev)
 {
-	return gfs2_create_inode(dir, dentry, mode, dev, NULL, 0, 0);
+	return gfs2_create_inode(dir, dentry, NULL, mode, dev, NULL, 0, 0, NULL);
+}
+
+/**
+ * gfs2_atomic_open - Atomically open a file
+ * @dir: The directory
+ * @dentry: The proposed new entry
+ * @file: The proposed new struct file
+ * @flags: open flags
+ * @mode: File mode
+ * @opened: Flag to say whether the file has been opened or not
+ *
+ * Returns: error code or 0 for success
+ */
+
+static int gfs2_atomic_open(struct inode *dir, struct dentry *dentry,
+                            struct file *file, unsigned flags,
+                            umode_t mode, int *opened)
+{
+	struct dentry *d;
+	bool excl = !!(flags & O_EXCL);
+
+	d = __gfs2_lookup(dir, dentry, file, opened);
+	if (IS_ERR(d))
+		return PTR_ERR(d);
+	if (d == NULL)
+		d = dentry;
+	if (d->d_inode) {
+		if (!(*opened & FILE_OPENED))
+			return finish_no_open(file, d);
+		return 0;
+	}
+
+	if (!(flags & O_CREAT))
+		return -ENOENT;
+
+	return gfs2_create_inode(dir, dentry, file, S_IFREG | mode, 0, NULL, 0, excl, opened);
 }
 
 /*
@@ -1780,6 +1852,7 @@ const struct inode_operations gfs2_dir_iops = {
 	.removexattr = gfs2_removexattr,
 	.fiemap = gfs2_fiemap,
 	.get_acl = gfs2_get_acl,
+	.atomic_open = gfs2_atomic_open,
 };
 
 const struct inode_operations gfs2_symlink_iops = {
diff --git a/fs/gfs2/inode.h b/fs/gfs2/inode.h
index c53c747..ba4d949 100644
--- a/fs/gfs2/inode.h
+++ b/fs/gfs2/inode.h
@@ -109,6 +109,7 @@ extern int gfs2_permission(struct inode *inode, int mask);
 extern int gfs2_setattr_simple(struct inode *inode, struct iattr *attr);
 extern struct inode *gfs2_lookup_simple(struct inode *dip, const char *name);
 extern void gfs2_dinode_out(const struct gfs2_inode *ip, void *buf);
+extern int gfs2_open_common(struct inode *inode, struct file *file);
 
 extern const struct inode_operations gfs2_file_iops;
 extern const struct inode_operations gfs2_dir_iops;