[PATCH] NFS: Fix a race in nfs_call_unlink()
From: trondmy@kernel.org
Date: 2022-11-17 13:53:40
Subsystem:
filesystems (vfs and infrastructure), nfs, sunrpc, and lockd clients, the rest · Maintainers:
Alexander Viro, Christian Brauner, Trond Myklebust, Anna Schumaker, Linus Torvalds
From: Trond Myklebust <redacted> We should check that the filehandles match before transferring the sillyrename data to the newly looked-up dentry in case the name was reused on the server. Signed-off-by: Trond Myklebust <redacted> --- fs/nfs/unlink.c | 1 + 1 file changed, 1 insertion(+)
diff --git a/fs/nfs/unlink.c b/fs/nfs/unlink.c
index 9697cd5d2561..150a953a8be9 100644
--- a/fs/nfs/unlink.c
+++ b/fs/nfs/unlink.c@@ -139,6 +139,7 @@ static int nfs_call_unlink(struct dentry *dentry, struct inode *inode, struct nf */ spin_lock(&alias->d_lock); if (d_really_is_positive(alias) && + !nfs_compare_fh(NFS_FH(inode), NFS_FH(d_inode(alias))) && !(alias->d_flags & DCACHE_NFSFS_RENAMED)) { devname_garbage = alias->d_fsdata; alias->d_fsdata = data;
--
2.38.1