ceph: introduce a new inode flag indicating if cached dentries are ordered

After creating/deleting/renaming file, offsets of sibling dentries may
change. So we can not use cached dentries to satisfy readdir. But we can
still use the cached dentries to conclude -ENOENT for lookup.

This patch introduces a new inode flag indicating if child dentries are
ordered. The flag is set at the same time marking a directory complete.
After creating/deleting/renaming file, we clear the flag on directory
inode. This prevents ceph_readdir() from using cached dentries to satisfy
readdir syscall.

Signed-off-by: Yan, Zheng <zyan@redhat.com>

1 files changed, 15 insertions, 8 deletions

diff --git a/fs/ceph/dir.c b/fs/ceph/dir.c
index e6d63f8f98c0..652619950fa9 100644
--- a/fs/ceph/dir.c
+++ b/fs/ceph/dir.c
@@ -183,7 +183,7 @@ more:
 	spin_unlock(&parent->d_lock);
 
 	/* make sure a dentry wasn't dropped while we didn't have parent lock */
-	if (!ceph_dir_is_complete(dir)) {
+	if (!ceph_dir_is_complete_ordered(dir)) {
 		dout(" lost dir complete on %p; falling back to mds\n", dir);
 		dput(dentry);
 		err = -EAGAIN;
@@ -261,10 +261,6 @@ static int ceph_readdir(struct file *file, struct dir_context *ctx)
 
 	/* always start with . and .. */
 	if (ctx->pos == 0) {
-		/* note dir version at start of readdir so we can tell
-		 * if any dentries get dropped */
-		fi->dir_release_count = atomic_read(&ci->i_release_count);
-
 		dout("readdir off 0 -> '.'\n");
 		if (!dir_emit(ctx, ".", 1, 
 			    ceph_translate_ino(inode->i_sb, inode->i_ino),
@@ -289,7 +285,7 @@ static int ceph_readdir(struct file *file, struct dir_context *ctx)
 	if ((ctx->pos == 2 || fi->dentry) &&
 	    !ceph_test_mount_opt(fsc, NOASYNCREADDIR) &&
 	    ceph_snap(inode) != CEPH_SNAPDIR &&
-	    __ceph_dir_is_complete(ci) &&
+	    __ceph_dir_is_complete_ordered(ci) &&
 	    __ceph_caps_issued_mask(ci, CEPH_CAP_FILE_SHARED, 1)) {
 		u32 shared_gen = ci->i_shared_gen;
 		spin_unlock(&ci->i_ceph_lock);
@@ -312,6 +308,13 @@ static int ceph_readdir(struct file *file, struct dir_context *ctx)
 
 	/* proceed with a normal readdir */
 
+	if (ctx->pos == 2) {
+		/* note dir version at start of readdir so we can tell
+		 * if any dentries get dropped */
+		fi->dir_release_count = atomic_read(&ci->i_release_count);
+		fi->dir_ordered_count = ci->i_ordered_count;
+	}
+
 more:
 	/* do we have the correct frag content buffered? */
 	if (fi->frag != frag || fi->last_readdir == NULL) {
@@ -446,8 +449,12 @@ more:
 	 */
 	spin_lock(&ci->i_ceph_lock);
 	if (atomic_read(&ci->i_release_count) == fi->dir_release_count) {
-		dout(" marking %p complete\n", inode);
-		__ceph_dir_set_complete(ci, fi->dir_release_count);
+		if (ci->i_ordered_count == fi->dir_ordered_count)
+			dout(" marking %p complete and ordered\n", inode);
+		else
+			dout(" marking %p complete\n", inode);
+		__ceph_dir_set_complete(ci, fi->dir_release_count,
+					fi->dir_ordered_count);
 	}
 	spin_unlock(&ci->i_ceph_lock);