If O_EXCL is *not* specified, then linkat() can be used to link the temporary file into the filesystem. If O_EXCL is specified then linkat() should fail (-1).
After commit 863f144f12ad ("vfs: open inside ->tmpfile()") the O_EXCL flag is no longer honored by the vfs layer for tmpfile, which means the file can be linked even if O_EXCL flag is specified, which is a change in behaviour for userspace!
The open flags was previously passed as a parameter, so it was uneffected by the changes to file->f_flags caused by finish_open(). This patch fixes the issue by storing file->f_flags in a local variable so the O_EXCL test logic is restored.
This regression was detected by Android CTS Bionic fcntl() tests running on android-mainline [1].
[1] https://android.googlesource.com/platform/bionic/+/ refs/heads/master/tests/fcntl_test.cpp#352
Fixes: 863f144f12ad ("vfs: open inside ->tmpfile()") To: lkml linux-kernel@vger.kernel.org Cc: Alexander Viro viro@zeniv.linux.org.uk Cc: Miklos Szeredi mszeredi@redhat.com Cc: stable@vger.kernel.org Cc: linux-fsdevel@vger.kernel.org Cc: linux-kernel@vger.kernel.org Cc: Will McVicker willmcvicker@google.com Cc: Peter Griffin gpeter@google.com Signed-off-by: Peter Griffin peter.griffin@linaro.org --- fs/namei.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/fs/namei.c b/fs/namei.c index 578c2110df02..9155ecb547ce 100644 --- a/fs/namei.c +++ b/fs/namei.c @@ -3591,6 +3591,7 @@ static int vfs_tmpfile(struct user_namespace *mnt_userns, struct inode *dir = d_inode(parentpath->dentry); struct inode *inode; int error; + int open_flag = file->f_flags;
/* we want directory to be writable */ error = inode_permission(mnt_userns, dir, MAY_WRITE | MAY_EXEC); @@ -3613,7 +3614,7 @@ static int vfs_tmpfile(struct user_namespace *mnt_userns, if (error) return error; inode = file_inode(file); - if (!(file->f_flags & O_EXCL)) { + if (!(open_flag & O_EXCL)) { spin_lock(&inode->i_lock); inode->i_state |= I_LINKABLE; spin_unlock(&inode->i_lock);