linux-kernel - Re: [PATCH 1/3] drm: add gem ttm helpers

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: Windows password security audit tool. GUI, reports in PDF.

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

Date:   Wed, 7 Aug 2019 13:40:15 +0200
From:   Daniel Vetter <daniel@...ll.ch>
To:     Gerd Hoffmann <kraxel@...hat.com>
Cc:     dri-devel <dri-devel@...ts.freedesktop.org>,
        Thomas Zimmermann <tzimmermann@...e.de>,
        Maarten Lankhorst <maarten.lankhorst@...ux.intel.com>,
        Maxime Ripard <maxime.ripard@...tlin.com>,
        Sean Paul <sean@...rly.run>, David Airlie <airlied@...ux.ie>,
        open list <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH 1/3] drm: add gem ttm helpers

On Wed, Aug 7, 2019 at 12:36 PM Gerd Hoffmann <kraxel@...hat.com> wrote:
>
>   Hi,
>
> > > > Same for this, you're just upcasting to ttm_bo and then downcasting to
> > > > gem_bo again ... I think just a series to roll out the existing gem
> > > > helpers everywhere should work?
> > >
> > > I don't think so.  drm_gem_dumb_map_offset() calls
> > > drm_gem_create_mmap_offset(), which I think is not correct for ttm
> > > objects because ttm_bo_init() handles vma_node initialization.
> >
> > More code to unify first? This should work exactly the same way for
> > all gem based drivers I think ... Only tricky bit is making sure
> > vmwgfx keeps working correctly.
>
> Yea.  Unifying on the gem way of doing things isn't going to work very
> well.  We would have to keep the current way of doing things in the ttm
> code, wrapped into "if (ttm_bo_uses_embedded_gem_object()) { ... }", to
> not break vmwgfx.
>
> So adding gem ttm helpers (where gem+ttm drivers can opt-in) looked like
> the better way of handling this to me ...

Ok I looked again, and your ttm version seems to exactly match
drm_gem_dumb_map_offset(), which we almost called
drm_gem_map_offset(). And could do that again by undoing that revert.
So I'm not seeing how a generic version for this stuff here wouldn't
also work for ttm ... Ofc if vmwgfx does something else they can keep
their own specific dumb map_offset implementation.

What am I missing?
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch