drm/etnaviv: Use FOLL_FORCE for userptr

Nothing checks userptr.ro except this call to pup_fast, which means
there's nothing actually preventing userspace from writing to this.
Which means you can just read-only mmap any file you want, userptr it
and then write to it with the gpu. Not good.

The right way to handle this is FOLL_WRITE | FOLL_FORCE, which will
break any COW mappings and update tracking for MAY_WRITE mappings so
there's no exploit and the vm isn't confused about what's going on.
For any legit use case there's no difference from what userspace can
observe and do.

Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
Cc: stable@vger.kernel.org
Cc: John Hubbard <jhubbard@nvidia.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Cc: Lucas Stach <l.stach@pengutronix.de>
Cc: Russell King <linux+etnaviv@armlinux.org.uk>
Cc: Christian Gmeiner <christian.gmeiner@gmail.com>
Cc: etnaviv@lists.freedesktop.org
Link: https://patchwork.freedesktop.org/patch/msgid/20210301095254.1946084-1-daniel.vetter@ffwll.ch

diff --git a/drivers/gpu/drm/etnaviv/etnaviv_gem.c b/drivers/gpu/drm/etnaviv/etnaviv_gem.c
index 6d38c5c..a9e696d 100644 (file)
--- a/drivers/gpu/drm/etnaviv/etnaviv_gem.c
+++ b/drivers/gpu/drm/etnaviv/etnaviv_gem.c
@@ -689,7 +689,7 @@ static int etnaviv_gem_userptr_get_pages(struct etnaviv_gem_object *etnaviv_obj)
                struct page **pages = pvec + pinned;
 
                ret = pin_user_pages_fast(ptr, num_pages,
-                                         !userptr->ro ? FOLL_WRITE : 0, pages);
+                                         FOLL_WRITE | FOLL_FORCE, pages);
                if (ret < 0) {
                        unpin_user_pages(pvec, pinned);
                        kvfree(pvec);