Received-SPF: pass (google.com: domain of linux-kernel+bounces-83974-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:40f1:3f00::1 as permitted sender) client-ip=2604:1380:40f1:3f00::1;
Date: Tue, 27 Feb 2024 13:25:56 -0700
From: Alex Williamson <alex.williamson@redhat.com>
To: David Hildenbrand <david@redhat.com>
Cc: Yisheng Xie <ethan.xys@linux.alibaba.com>, akpm@linux-foundation.org,
 kvm@vger.kernel.org, linux-kernel@vger.kernel.org, linux-mm@kvack.org
Subject: Re: [PATCH] vfio/type1: unpin PageReserved page
Message-ID: <20240227132556.17e87767.alex.williamson@redhat.com>
In-Reply-To: <abb00aef-378c-481a-a885-327a99aa7b09@redhat.com>
References: <20240226160106.24222-1-ethan.xys@linux.alibaba.com>
	<20240226091438.1fc37957.alex.williamson@redhat.com>
	<e10ace3f-78d3-4843-8028-a0e1cd107c15@linux.alibaba.com>
	<20240226103238.75ad4b24.alex.williamson@redhat.com>
	<abb00aef-378c-481a-a885-327a99aa7b09@redhat.com>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable

On Tue, 27 Feb 2024 11:27:08 +0100
David Hildenbrand <david@redhat.com> wrote:

> On 26.02.24 18:32, Alex Williamson wrote:
> > On Tue, 27 Feb 2024 01:14:54 +0800
> > Yisheng Xie <ethan.xys@linux.alibaba.com> wrote:
> >  =20
> >> =E5=9C=A8 2024/2/27 00:14, Alex Williamson =E5=86=99=E9=81=93: =20
> >>> On Tue, 27 Feb 2024 00:01:06 +0800
> >>> Yisheng Xie<ethan.xys@linux.alibaba.com>  wrote:
> >>>    =20
> >>>> We meet a warning as following:
> >>>>    WARNING: CPU: 99 PID: 1766859 at mm/gup.c:209 try_grab_page.part.=
0+0xe8/0x1b0
> >>>>    CPU: 99 PID: 1766859 Comm: qemu-kvm Kdump: loaded Tainted: GOE  5=
10.134-008.2.x86_64 #1 =20
> >>>                                                                      =
^^^^^^^^
> >>>
> >>> Does this issue reproduce on mainline?  Thanks, =20
> >>
> >> I have check the code of mainline, the logical seems the same as my
> >> version.
> >>
> >> so I think it can reproduce if i understand correctly. =20
> >=20
> > I obviously can't speak to what's in your 5.10.134-008.2 kernel, but I
> > do know there's a very similar issue resolved in v6.0 mainline and
> > included in v5.10.146 of the stable tree.  Please test.  Thanks, =20
>=20
> This commit, to be precise:
>=20
> commit 873aefb376bbc0ed1dd2381ea1d6ec88106fdbd4
> Author: Alex Williamson <alex.williamson@redhat.com>
> Date:   Mon Aug 29 21:05:40 2022 -0600
>=20
>      vfio/type1: Unpin zero pages
>     =20
>      There's currently a reference count leak on the zero page.  We incre=
ment
>      the reference via pin_user_pages_remote(), but the page is later han=
dled
>      as an invalid/reserved page, therefore it's not accounted against the
>      user and not unpinned by our put_pfn().
>     =20
>      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.
>     =20
>      The zero page is always resident, so for our purposes there's no rea=
son
>      to keep it pinned.  Therefore, add a loop to walk pages returned from
>      pin_user_pages_remote() and unpin any zero pages.
>=20
>=20
> BUT
>=20
> in the meantime, we also have
>=20
> commit c8070b78751955e59b42457b974bea4a4fe00187
> Author: David Howells <dhowells@redhat.com>
> Date:   Fri May 26 22:41:40 2023 +0100
>=20
>      mm: Don't pin ZERO_PAGE in pin_user_pages()
>     =20
>      Make pin_user_pages*() leave a ZERO_PAGE unpinned if it extracts a p=
ointer
>      to it from the page tables and make unpin_user_page*() corresponding=
ly
>      ignore a ZERO_PAGE when unpinning.  We don't want to risk overrunnin=
g a
>      zero page's refcount as we're only allowed ~2 million pins on it -
>      something that userspace can conceivably trigger.
>     =20
>      Add a pair of functions to test whether a page or a folio is a ZERO_=
PAGE.
>=20
>=20
> So the unpin_user_page_* won't do anything with the shared zeropage.
>=20
> (likely, we could revert 873aefb376bbc0ed1dd2381ea1d6ec88106fdbd4)


Yes, according to the commit log it seems like the unpin is now just
wasted work since v6.5.  Thanks!

Alex