Received-SPF: pass (google.com: domain of linux-kernel+bounces-154696-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:4601:e00::3 as permitted sender) client-ip=2604:1380:4601:e00::3;
Precedence: bulk
MIME-Version: 1.0
Date: Tue, 23 Apr 2024 08:35:03 +0000
Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: quoted-printable
From: "Yajun Deng" <yajun.deng@linux.dev>
Message-ID: <b6ea1fb5bc6c06d2855e41b4034656b0a76b58f5@linux.dev>
TLS-Required: No
Subject: Re: [PATCH] mm/rmap: remove unnecessary page_table_lock
To: "David Hildenbrand" <david@redhat.com>, akpm@linux-foundation.org
Cc: linux-mm@kvack.org, linux-kernel@vger.kernel.org
In-Reply-To: <fd7fde90-21ea-4617-be17-ba387b44feaf@redhat.com>
References: <20240422105212.1485788-1-yajun.deng@linux.dev>
 <b848c431-deca-42e4-925c-673b3fa1f251@redhat.com>
 <3c452d5db5b3d5879160ab62a9e0ac4481a6298a@linux.dev>
 <fd7fde90-21ea-4617-be17-ba387b44feaf@redhat.com>

April 23, 2024 at 4:18 PM, "David Hildenbrand" <david@redhat.com> wrote:


>=20
>=20On 23.04.24 09:53, Yajun Deng wrote:
>=20
>=20>=20
>=20> April 22, 2024 at 7:24 PM, "David Hildenbrand" <david@redhat.com> w=
rote:
> >=20
>=20>  > >>
> >=20
>=20> >=20
>=20> > On 22.04.24 12:52, Yajun Deng wrote:
> > >=20
>=20>=20
>=20>  page_table_lock is a lock that for page table, we won't change pag=
e
> >=20
>=20>  table in __anon_vma_prepare(). As we can see, it works well in
> >=20
>=20>  anon_vma_clone(). They do the same operation.
> >=20
>=20> >=20
>=20> > We are reusing mm->page_table_lock to serialize, not the *actual*=
 low-level page table locks that really protect PTEs.
> > >=20
>=20> >  With that locking gone, there would be nothing protection vma->a=
non_vma.
> > >=20
>=20> >  Note that anon_vma_clone() is likely called with the mmap_lock h=
eld in write mode, which is not the case for __anon_vma_prepare() ...
> > >=20
>=20>=20
>=20>  Yes, anon_vma_clone() is called with the mmap_lock held. I added m=
map_assert_write_locked(dst->vm_mm) to prove it.
> >=20
>=20>  I added mmap_assert_write_locked(vma->vm_mm) in __anon_vma_prepare=
() at the same time, it shows __anon_vma_prepare()
> >=20
>=20>  is also called with the mmap_lock held too.
> >=20
>=20
> Make sure you actually have lockdep built in and enabled.
>=20

This=20is my config.
CONFIG_LOCKDEP=3Dn
CONFIG_DEBUG_VM=3Dy

I did another test.
I put mmap_assert_write_locked(mm) before 'set_bit(MMF_OOM_SKIP, &mm->fla=
gs)' in mmap.c, it's outside the lock.
It will crash when on boot. I think mmap_assert_write_locked() works.


> __anon_vma_prepare() is for example called from do_anonymous_page() whe=
re we might only hold the mmap_lock in read mode (or not at all IIRC with=
 VMA in read mode).
>=20
>=20-- Cheers,
>=20
>=20David / dhildenb
>