diff mbox series

[2/5] cifs: do not return an invalidated cfid

Message ID 20250502051517.10449-2-sprasad@microsoft.com
State New
Headers show
Series [1/5] cifs: protect cfid accesses with fid_lock | expand

Commit Message

Shyam Prasad N May 2, 2025, 5:13 a.m. UTC
From: Shyam Prasad N <sprasad@microsoft.com>

open_cached_dir should either return an existing valid cfid
or create a new one. Validity of cfid depends on both
cfid->has_lease and cfid->time to be true. However, if has_lease
was invalidated by a worker thread in parallel, we could end up
leaking both a dentry and a server handle.

This change checks if the entry was invalidated and returns
a -ENOENT in such a case.

Signed-off-by: Shyam Prasad N <sprasad@microsoft.com>
---
 fs/smb/client/cached_dir.c | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)
diff mbox series

Patch

diff --git a/fs/smb/client/cached_dir.c b/fs/smb/client/cached_dir.c
index f074675fa6be..d307636c2679 100644
--- a/fs/smb/client/cached_dir.c
+++ b/fs/smb/client/cached_dir.c
@@ -198,9 +198,11 @@  int open_cached_dir(unsigned int xid, struct cifs_tcon *tcon,
 	}
 
 	/*
+	 * check again that the cfid is valid (with mutex held this time).
 	 * Return cached fid if it is valid (has a lease and has a time).
 	 * Otherwise, it is either a new entry or laundromat worker removed it
-	 * from @cfids->entries.  Caller will put last reference if the latter.
+	 * from @cfids->entries.  If the latter, we drop the refcount and return
+	 * an error to the caller.
 	 */
 	spin_lock(&cfid->fid_lock);
 	if (cfid->has_lease && cfid->time) {
@@ -208,6 +210,12 @@  int open_cached_dir(unsigned int xid, struct cifs_tcon *tcon,
 		*ret_cfid = cfid;
 		kfree(utf16_path);
 		return 0;
+	} else if (!cfid->has_lease) {
+		spin_unlock(&cfid->fid_lock);
+		/* drop the ref that we have */
+		kref_put(&cfid->refcount, smb2_close_cached_fid);
+		kfree(utf16_path);
+		return -ENOENT;
 	}
 	spin_unlock(&cfid->fid_lock);