On Tue, Mar 24, 2020 at 8:50 AM Christoph Hellwig <hch@xxxxxx> wrote:
>
> There is no excuse to ever perform actions related to a specific handler
> directly from the generic xattr code as we have handler that understand
> the specific data in given attrs. As a nice sideeffect this removes
> tons of pointless boilerplate code.
>
> Signed-off-by: Christoph Hellwig <hch@xxxxxx>
> ---
>
> Changes since v1:
> - fix up file systems that have their own ACL xattr handlers
>
> fs/9p/acl.c | 4 +--
> fs/overlayfs/dir.c | 2 +-
> fs/overlayfs/super.c | 2 +-
> fs/posix_acl.c | 62 ++-------------------------------
> fs/xattr.c | 8 +----
> include/linux/posix_acl_xattr.h | 12 -------
> 6 files changed, 7 insertions(+), 83 deletions(-)
>
> diff --git a/fs/9p/acl.c b/fs/9p/acl.c
> index 6261719f6f2a..f3455ba2a84d 100644
> --- a/fs/9p/acl.c
> +++ b/fs/9p/acl.c
> @@ -232,7 +232,7 @@ static int v9fs_xattr_get_acl(const struct xattr_handler *handler,
> return PTR_ERR(acl);
> if (acl == NULL)
> return -ENODATA;
> - error = posix_acl_to_xattr(&init_user_ns, acl, buffer, size);
> + error = posix_acl_to_xattr(current_user_ns(), acl, buffer, size);
Okay, but the uncached cache is still broken. It needs the xattr to
be converted to acl (posix_acl_to_xattr(&init_user_ns, ...)) then back
to xattr here.
> posix_acl_release(acl);
>
> return error;
> @@ -262,7 +262,7 @@ static int v9fs_xattr_set_acl(const struct xattr_handler *handler,
> return -EPERM;
> if (value) {
> /* update the cached acl value */
> - acl = posix_acl_from_xattr(&init_user_ns, value, size);
> + acl = posix_acl_from_xattr(current_user_ns(), value, size);
Same in this function.
Thanks,
Miklos
