vfio/type1: Unpin zero pages
authorAlex Williamson <alex.williamson@redhat.com>
Tue, 30 Aug 2022 03:05:40 +0000 (21:05 -0600)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Thu, 15 Sep 2022 09:30:02 +0000 (11:30 +0200)
commit5321908ef74fb593e0dbc8737d25038fc86c9986
tree4c393bc941d19ed70294d29c66bfd6085a85b2da
parent7da1afa444a19dec5ac49c2f8eae1838562a18f1
vfio/type1: Unpin zero pages

commit 873aefb376bbc0ed1dd2381ea1d6ec88106fdbd4 upstream.

There's currently a reference count leak on the zero page.  We increment
the reference via pin_user_pages_remote(), but the page is later handled
as an invalid/reserved page, therefore it's not accounted against the
user and not unpinned by our put_pfn().

Introducing special zero page handling in put_pfn() would resolve the
leak, but without accounting of the zero page, a single user could
still create enough mappings to generate a reference count overflow.

The zero page is always resident, so for our purposes there's no reason
to keep it pinned.  Therefore, add a loop to walk pages returned from
pin_user_pages_remote() and unpin any zero pages.

Cc: stable@vger.kernel.org
Reported-by: Luboslav Pivarc <lpivarc@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Link: https://lore.kernel.org/r/166182871735.3518559.8884121293045337358.stgit@omen
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/vfio/vfio_iommu_type1.c