On Fri, Jan 1, 2021 at 10:12 PM Icenowy Zheng icenowy@aosc.io wrote:
The function ovl_dir_real_file() currently uses the semaphore of the inode to synchronize write to the upperfile cache field.
However, this function will get called by ovl_ioctl_set_flags(), which utilizes the inode semaphore too. In this case ovl_dir_real_file() will try to claim a lock that is owned by a function in its call stack, which won't get released before ovl_dir_real_file() returns.
oops. I wondered why I didn't see any warnings on this from lockdep. Ah! because the xfstest that exercises ovl_ioctl_set_flags() on directory, generic/079, starts with an already upper dir.
And the xfstest that checks chattr+i on lower/upper files, overlay/040, does not check chattr on dirs (ioctl on overlay dirs wasn't supported at the time the test was written).
Would you be able to create a variant of test overlay/040 that also tests chattr +i on lower/upper dirs to test your patch and confirm that the test fails on master with the appropriate Kconfig debug options.
Define a dedicated semaphore for the upperfile cache, so that the deadlock won't happen.
Fixes: 61536bed2149 ("ovl: support [S|G]ETFLAGS and FS[S|G]ETXATTR ioctls for directories") Cc: stable@vger.kernel.org # v5.10 Signed-off-by: Icenowy Zheng icenowy@aosc.io
fs/overlayfs/readdir.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-)
diff --git a/fs/overlayfs/readdir.c b/fs/overlayfs/readdir.c index 01620ebae1bd..f10701aabb71 100644 --- a/fs/overlayfs/readdir.c +++ b/fs/overlayfs/readdir.c @@ -56,6 +56,7 @@ struct ovl_dir_file { struct list_head *cursor; struct file *realfile; struct file *upperfile;
struct semaphore upperfile_sem;
mutex please
};
static struct ovl_cache_entry *ovl_cache_entry_from_node(struct rb_node *n) @@ -883,7 +884,7 @@ struct file *ovl_dir_real_file(const struct file *file, bool want_upper) ovl_path_upper(dentry, &upperpath); realfile = ovl_dir_open_realfile(file, &upperpath);
inode_lock(inode);
down(&od->upperfile_sem); if (!od->upperfile) { if (IS_ERR(realfile)) { inode_unlock(inode);
You missed this unlock
Thanks, Amir.