Received-SPF: pass (google.com: domain of linux-kernel+bounces-111985-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45d1:ec00::1 as permitted sender) client-ip=2604:1380:45d1:ec00::1;
Date: Fri, 22 Mar 2024 12:19:12 -0700
From: Andrew Morton <akpm@linux-foundation.org>
To: Dev Jain <dev.jain@arm.com>
Cc: shuah@kernel.org, linux-mm@kvack.org, linux-kselftest@vger.kernel.org,
 linux-kernel@vger.kernel.org, Anshuman.Khandual@arm.com
Subject: Re: [PATCH] selftests/mm: Confirm VA exhaustion without reliance on
 correctness of mmap()
Message-Id: <20240322121912.3bc9e58cfda1b7c92aac9c4f@linux-foundation.org>
In-Reply-To: <d91c8c36-b01d-48f9-b0f6-8faa58edaed0@arm.com>
References: <20240321103522.516097-1-dev.jain@arm.com>
	<20240321145146.a3ce8a1e247371e33a437978@linux-foundation.org>
	<d91c8c36-b01d-48f9-b0f6-8faa58edaed0@arm.com>
Precedence: bulk
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable

On Fri, 22 Mar 2024 10:42:47 +0530 Dev Jain <dev.jain@arm.com> wrote:

>=20
> On 3/22/24 03:21, Andrew Morton wrote:
> > On Thu, 21 Mar 2024 16:05:22 +0530 Dev Jain <dev.jain@arm.com> wrote:
> >
> >> Currently, VA exhaustion is being checked by passing a hint to mmap() =
and
> >> expecting it to fail. This patch makes a stricter test by successful w=
rite()
> >> calls from /proc/self/maps to a dump file, confirming that a free chun=
k is
> >> indeed not available.
> > What's wrong with the current approach?
> While populating the lower VA space, mmap() fails because we have=20
> exhausted the space.
>=20
> Then, in validate_lower_address_hint(), because mmap() fails, we confirm=
=20
> that we have
>=20
> indeed exhausted the space. There is a circular logic involved here.
>=20
> Assume that there is a bug in mmap(), also assume that it exists=20
> independent of whether
>=20
> you pass a hint address or not; that for some reason it is not able to=20
> find a 1GB chunk.
>=20
> My idea is to assert the exhaustion against some other method.

Thanks.  I added the above to the changelog.

>=20
> Also, in the following line in validate_complete_va_space():
>=20
>  =A0=A0=A0 =A0=A0=A0 if (start_addr - prev_end_addr >=3D SZ_1GB)
>=20
> I made a small error, I forgot to use MAP_CHUNK_SIZE instead of SZ_1GB.

And the preceding comment might need an edit.  Please send an updated
patch?