Received-SPF: pass (google.com: domain of linux-kernel+bounces-43400-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: Mon, 29 Jan 2024 20:26:19 +0100
From: Petr =?UTF-8?B?VGVzYcWZw61r?= <petr@tesarici.cz>
To: Will Deacon <will@kernel.org>
Cc: linux-kernel@vger.kernel.org, kernel-team@android.com,
 iommu@lists.linux.dev, Christoph Hellwig <hch@lst.de>, Marek Szyprowski
 <m.szyprowski@samsung.com>, Robin Murphy <robin.murphy@arm.com>, Petr
 Tesarik <petr.tesarik1@huawei-partners.com>, Dexuan Cui
 <decui@microsoft.com>
Subject: Re: [PATCH 0/2] Fix double allocation in swiotlb_alloc()
Message-ID: <20240129202619.69178dea@meshulam.tesarici.cz>
In-Reply-To: <20240129184255.GA12631@willie-the-truck>
References: <20240126151956.10014-1-will@kernel.org>
	<20240126172059.48ad7b9f@meshulam.tesarici.cz>
	<20240129184255.GA12631@willie-the-truck>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable

On Mon, 29 Jan 2024 18:42:55 +0000
Will Deacon <will@kernel.org> wrote:

> On Fri, Jan 26, 2024 at 05:20:59PM +0100, Petr Tesa=C5=99=C3=ADk wrote:
> > On Fri, 26 Jan 2024 15:19:54 +0000
> > Will Deacon <will@kernel.org> wrote:
> >  =20
> > > Hi folks,
> > >=20
> > > These two patches fix a nasty double allocation problem in swiotlb_al=
loc()
> > > and add a diagnostic to help catch any similar issues in future. This=
 was
> > > a royal pain to track down and I've had to make a bit of a leap at the
> > > correct alignment semantics (i.e. iotlb_align_mask vs alloc_align_mas=
k). =20
> >=20
> > Welcome to the club. I believe you had to re-discover what I described =
here:
> >=20
> >   https://lore.kernel.org/linux-iommu/20231108101347.77cab795@meshulam.=
tesarici.cz/ =20
>=20
> Lucky me...
>=20
> > The relevant part would be this:
> >=20
> >   To sum it up, there are two types of alignment:
> >=20
> >   1. specified by a device's min_align_mask; this says how many low
> >      bits of a buffer's physical address must be preserved,
> >=20
> >   2. specified by allocation size and/or the alignment parameter;
> >      this says how many low bits in the first IO TLB slot's physical
> >      address must be zero. =20
> >=20
> > Fix for that has been sitting on my TODO list for too long. :-( =20
>=20
> FWIW, it did _used_ to work (or appear to work), so it would be good to
> at least get it back to the old behaviour if nothing else.

Yes, now that I look at the code, it was probably misunderstanding on
my side as to how the three different alignment requirements are
supposed to work together.

AFAICT your patch addresses everything that has ever worked. The rest
needs some more thought, and before I touch this loop again, I'll write
a proper test case.

Petr T

> Anyway, cheers for reviewing the patches. I'll go through your
> comments now...
>=20
> Will