linux-kernel - Re: [PATCH 2/5] drm/nouveau/uvmm: Allow larger pages

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <CAA+WOBvfStqh+HTXYrrD_=YUXPYNaKLgHPLQ9g07gfQmAXKAWA@mail.gmail.com>
Date: Thu, 23 Oct 2025 13:14:36 +0300
From: Mohamed Ahmed <mohamedahmedegypt2001@...il.com>
To: Mary Guillemard <mary@...y.zone>
Cc: Danilo Krummrich <dakr@...nel.org>, linux-kernel@...r.kernel.org, 
	dri-devel@...ts.freedesktop.org, 
	Faith Ekstrand <faith.ekstrand@...labora.com>, Lyude Paul <lyude@...hat.com>, 
	Maarten Lankhorst <maarten.lankhorst@...ux.intel.com>, Maxime Ripard <mripard@...nel.org>, 
	Thomas Zimmermann <tzimmermann@...e.de>, David Airlie <airlied@...il.com>, Simona Vetter <simona@...ll.ch>, 
	nouveau@...ts.freedesktop.org
Subject: Re: [PATCH 2/5] drm/nouveau/uvmm: Allow larger pages

The other thing making me hesitant of depending on
nouveau_bo_fixup_align() is that VM_BIND is entirely client controlled
and there isn't really (at least as far as I understand) way for the
bo_fixup_align() path to have enough info to e.g. work around the
"client allocates size and binds to address not aligned to that size"
issue (likely the reason for hitting the mismatch case. this didn't
show in the older kernel versions because everything was forced to 4K
anyways).

On Thu, Oct 23, 2025 at 12:39 AM Mary Guillemard <mary@...y.zone> wrote:
>
> On Wed, Oct 22, 2025 at 10:56 PM Danilo Krummrich <dakr@...nel.org> wrote:
> >
> > On 10/22/25 12:16 PM, Mohamed Ahmed wrote:
> > > Pinging again re: review and also was asking if we can revert the
> > > select_page_shift() handling back to v1 behavior with a fall-back
> > > path, as it looks like there are some cases where
> > > nouveau_bo_fixup_align() isn't enough;
> > > https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36450#note_3159199.
> >
> > I don't think we should add a fallback for something that is expected to be
> > sufficient.
> >
> > Instead we should figure out in which exact case the WARN_ON() was hit and why.
>
> The reason I wrote this code initially was to handle addresses
> provided by userspace that aren't aligned to the page size selected
> during BO creation.
> This is something I did trigger when typing this patch initially with
> my distro provided version of mesa (likely 25.0.x but it has been a
> while)
> Thomas Andersen also confirmed on nouveau irc channel that he did hit
> this case with an old version of NVK and this patchset.
>
> I think we could just remove the WARN_ON and properly document that
> this was previously allowed and is there for backward compatibility.
>
> Regards,
> Mary Guillemard