On 6/28/16 11:39 PM, Tyler Hicks wrote:
> Cherry-picking mainline commit 2f36db71009304b3f0b95afacd8eba1f9f046b87
> introduces a regression in eCryptfs when mainline commit
> 6a480a7842545ec520a91730209ec0bae41694c1 (4.6+) is not present. The
> regression causes all attempts at opening directory files to fail with
> EMEDIUMTYPE when the lower filesystem's file_operations for directory
> files do not implement mmap.
>
> This is a simple fix that allows the check for the lower file's mmap
> implementation to be ignored if the lower file is a directory.
I have a different fix that I believe is more correct for this. I would
have posted it in response to the original fix if it were ever actually
posted for public discussion.
Denying open is the wrong place to fix this. It's too heavy a hammer
and, as we see here, a bit fragile.
The right fix is to deny the mmap call instead.
-Jeff
> Signed-off-by: Tyler Hicks <tyhicks@xxxxxxxxxxxxx>
> Tested-by: Tyler Hicks <tyhicks@xxxxxxxxxxxxx> # 4.4.y, 3.18.y
> Cc: <stable@xxxxxxxxxxxxxxx> # 4.5-
> ---
> fs/ecryptfs/kthread.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/fs/ecryptfs/kthread.c b/fs/ecryptfs/kthread.c
> index e818f5a..b9faeab 100644
> --- a/fs/ecryptfs/kthread.c
> +++ b/fs/ecryptfs/kthread.c
> @@ -171,7 +171,7 @@ int ecryptfs_privileged_open(struct file **lower_file,
> goto out;
> }
> have_file:
> - if ((*lower_file)->f_op->mmap == NULL) {
> + if ((*lower_file)->f_op->mmap == NULL && !d_is_dir(lower_dentry)) {
> fput(*lower_file);
> *lower_file = NULL;
> rc = -EMEDIUMTYPE;
>
--
Jeff Mahoney
SUSE Labs
Attachment:
signature.asc
Description: OpenPGP digital signature