drm: fb-helper: Validate requested pixel format against bpp

Message ID 20181003110457.9318-1-Eugeniy.Paltsev@synopsys.com
State New
Headers show
Series
  • drm: fb-helper: Validate requested pixel format against bpp
Related show

Commit Message

Eugeniy Paltsev Oct. 3, 2018, 11:04 a.m.
Validate requested pixel format against bits_per_pixel to reject
invalid formats with subcomponents length sum is greater than requested
bits_per_pixel.

weston 5.0.0 with fbdev backend tries to set up an ARGB x8r8g8b8 pixel
format without bits_per_pixel updating. So it can request
x8r8g8b8 with 16 bpp which is obviously incorrect and should be
rejected.

Cc: stable@vger.kernel.org
Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
---
 drivers/gpu/drm/drm_fb_helper.c | 7 +++++++
 1 file changed, 7 insertions(+)

Comments

Daniel Vetter Oct. 3, 2018, 11:36 a.m. | #1
On Wed, Oct 3, 2018 at 1:05 PM Eugeniy Paltsev
<Eugeniy.Paltsev@synopsys.com> wrote:
>
> Validate requested pixel format against bits_per_pixel to reject
> invalid formats with subcomponents length sum is greater than requested
> bits_per_pixel.
>
> weston 5.0.0 with fbdev backend tries to set up an ARGB x8r8g8b8 pixel
> format without bits_per_pixel updating. So it can request
> x8r8g8b8 with 16 bpp which is obviously incorrect and should be
> rejected.
>
> Cc: stable@vger.kernel.org
> Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>

drm fbdev emulation doesn't support changing the pixel format at all.
I think we should reject all such request, not just the invalid ones.
Can you pls respin?

Thanks, Daniel

> ---
>  drivers/gpu/drm/drm_fb_helper.c | 7 +++++++
>  1 file changed, 7 insertions(+)
>
> diff --git a/drivers/gpu/drm/drm_fb_helper.c b/drivers/gpu/drm/drm_fb_helper.c
> index 16ec93b75dbf..4f39da07f053 100644
> --- a/drivers/gpu/drm/drm_fb_helper.c
> +++ b/drivers/gpu/drm/drm_fb_helper.c
> @@ -1610,6 +1610,13 @@ int drm_fb_helper_check_var(struct fb_var_screeninfo *var,
>                 return -EINVAL;
>         }
>
> +       if ((var->green.length + var->blue.length + var->red.length +
> +           var->transp.length) > var->bits_per_pixel) {
> +               DRM_DEBUG("fb requested pixel format can't fit in %d bpp\n",
> +                         var->bits_per_pixel);
> +               return -EINVAL;
> +       }
> +
>         switch (var->bits_per_pixel) {
>         case 16:
>                 depth = (var->green.length == 6) ? 16 : 15;
> --
> 2.14.4
>
> _______________________________________________
> dri-devel mailing list
> dri-devel@lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/dri-devel
Ville Syrjälä Oct. 3, 2018, 12:30 p.m. | #2
On Wed, Oct 03, 2018 at 01:36:00PM +0200, Daniel Vetter wrote:
> On Wed, Oct 3, 2018 at 1:05 PM Eugeniy Paltsev
> <Eugeniy.Paltsev@synopsys.com> wrote:
> >
> > Validate requested pixel format against bits_per_pixel to reject
> > invalid formats with subcomponents length sum is greater than requested
> > bits_per_pixel.
> >
> > weston 5.0.0 with fbdev backend tries to set up an ARGB x8r8g8b8 pixel
> > format without bits_per_pixel updating. So it can request
> > x8r8g8b8 with 16 bpp which is obviously incorrect and should be
> > rejected.
> >
> > Cc: stable@vger.kernel.org
> > Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
> 
> drm fbdev emulation doesn't support changing the pixel format at all.
> I think we should reject all such request, not just the invalid ones.
> Can you pls respin?

FYI I once posted a patch to tighten up the fb-helper pixel format
stuff:
https://patchwork.freedesktop.org/patch/203189/

> 
> Thanks, Daniel
> 
> > ---
> >  drivers/gpu/drm/drm_fb_helper.c | 7 +++++++
> >  1 file changed, 7 insertions(+)
> >
> > diff --git a/drivers/gpu/drm/drm_fb_helper.c b/drivers/gpu/drm/drm_fb_helper.c
> > index 16ec93b75dbf..4f39da07f053 100644
> > --- a/drivers/gpu/drm/drm_fb_helper.c
> > +++ b/drivers/gpu/drm/drm_fb_helper.c
> > @@ -1610,6 +1610,13 @@ int drm_fb_helper_check_var(struct fb_var_screeninfo *var,
> >                 return -EINVAL;
> >         }
> >
> > +       if ((var->green.length + var->blue.length + var->red.length +
> > +           var->transp.length) > var->bits_per_pixel) {
> > +               DRM_DEBUG("fb requested pixel format can't fit in %d bpp\n",
> > +                         var->bits_per_pixel);
> > +               return -EINVAL;
> > +       }
> > +
> >         switch (var->bits_per_pixel) {
> >         case 16:
> >                 depth = (var->green.length == 6) ? 16 : 15;
> > --
> > 2.14.4
> >
> > _______________________________________________
> > dri-devel mailing list
> > dri-devel@lists.freedesktop.org
> > https://lists.freedesktop.org/mailman/listinfo/dri-devel
> 
> 
> 
> -- 
> Daniel Vetter
> Software Engineer, Intel Corporation
> +41 (0) 79 365 57 48 - http://blog.ffwll.ch
> _______________________________________________
> dri-devel mailing list
> dri-devel@lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/dri-devel
Eugeniy Paltsev Oct. 3, 2018, 2:30 p.m. | #3
On Wed, 2018-10-03 at 15:30 +0300, Ville Syrjälä wrote:
> On Wed, Oct 03, 2018 at 01:36:00PM +0200, Daniel Vetter wrote:
> > On Wed, Oct 3, 2018 at 1:05 PM Eugeniy Paltsev
> > <Eugeniy.Paltsev@synopsys.com> wrote:
> > > 
> > > Validate requested pixel format against bits_per_pixel to reject
> > > invalid formats with subcomponents length sum is greater than requested
> > > bits_per_pixel.
> > > 
> > > weston 5.0.0 with fbdev backend tries to set up an ARGB x8r8g8b8 pixel
> > > format without bits_per_pixel updating. So it can request
> > > x8r8g8b8 with 16 bpp which is obviously incorrect and should be
> > > rejected.
> > > 
> > > Cc: stable@vger.kernel.org
> > > Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
> > 
> > drm fbdev emulation doesn't support changing the pixel format at all.
> > I think we should reject all such request, not just the invalid ones.
> > Can you pls respin?
> 
> FYI I once posted a patch to tighten up the fb-helper pixel format
> stuff:
> https://patchwork.freedesktop.org/patch/203189/


Hi Daniel,

will you take Ville's patch or should I create the new one which is only related
to new pixel format validation in drm_fb_helper_check_var() ?


> > Thanks, Daniel
> > 
> > > ---
> > >  drivers/gpu/drm/drm_fb_helper.c | 7 +++++++
> > >  1 file changed, 7 insertions(+)
> > > 
> > > diff --git a/drivers/gpu/drm/drm_fb_helper.c b/drivers/gpu/drm/drm_fb_helper.c
> > > index 16ec93b75dbf..4f39da07f053 100644
> > > --- a/drivers/gpu/drm/drm_fb_helper.c
> > > +++ b/drivers/gpu/drm/drm_fb_helper.c
> > > @@ -1610,6 +1610,13 @@ int drm_fb_helper_check_var(struct fb_var_screeninfo *var,
> > >                 return -EINVAL;
> > >         }
> > > 
> > > +       if ((var->green.length + var->blue.length + var->red.length +
> > > +           var->transp.length) > var->bits_per_pixel) {
> > > +               DRM_DEBUG("fb requested pixel format can't fit in %d bpp\n",
> > > +                         var->bits_per_pixel);
> > > +               return -EINVAL;
> > > +       }
> > > +
> > >         switch (var->bits_per_pixel) {
> > >         case 16:
> > >                 depth = (var->green.length == 6) ? 16 : 15;
> > > --
> > > 2.14.4
> > > 
> > > _______________________________________________
> > > dri-devel mailing list
> > > dri-devel@lists.freedesktop.org
> > > https://urldefense.proofpoint.com/v2/url?u=https-3A__lists.freedesktop.org_mailman_listinfo_dri-2Ddevel&d=DwIDAw&c=DPL6_X_6JkXFx7AXWqB0tg&r=ZlJN
> > > 1MriPUTkBKCrPSx67GmaplEUGcAEk9yPtCLdUXI&m=f12ZyEESIeavtqCUKutiZ9F6xtRFC2UUvdqnM4ywBx8&s=CPS8taMiYbIgXo-fxhqErOJXvO6PMTzmr-BNnGJIoy0&e=
> > 
> > 
> > 
> > -- 
> > Daniel Vetter
> > Software Engineer, Intel Corporation
> > +41 (0) 79 365 57 48 - https://urldefense.proofpoint.com/v2/url?u=http-3A__blog.ffwll.ch&d=DwIDAw&c=DPL6_X_6JkXFx7AXWqB0tg&r=ZlJN1MriPUTkBKCrPSx67
> > GmaplEUGcAEk9yPtCLdUXI&m=f12ZyEESIeavtqCUKutiZ9F6xtRFC2UUvdqnM4ywBx8&s=Vt8OX9s9ljSK6GDgbnwsF-Yd35fbBUfe8SBV2jPnVaQ&e=
> > _______________________________________________
> > dri-devel mailing list
> > dri-devel@lists.freedesktop.org
> > https://urldefense.proofpoint.com/v2/url?u=https-3A__lists.freedesktop.org_mailman_listinfo_dri-2Ddevel&d=DwIDAw&c=DPL6_X_6JkXFx7AXWqB0tg&r=ZlJN1M
> > riPUTkBKCrPSx67GmaplEUGcAEk9yPtCLdUXI&m=f12ZyEESIeavtqCUKutiZ9F6xtRFC2UUvdqnM4ywBx8&s=CPS8taMiYbIgXo-fxhqErOJXvO6PMTzmr-BNnGJIoy0&e=
> 
>
Daniel Vetter Oct. 3, 2018, 3:29 p.m. | #4
On Wed, Oct 3, 2018 at 4:30 PM Eugeniy Paltsev
<eugeniy.paltsev@synopsys.com> wrote:
>
> On Wed, 2018-10-03 at 15:30 +0300, Ville Syrjälä wrote:
> > On Wed, Oct 03, 2018 at 01:36:00PM +0200, Daniel Vetter wrote:
> > > On Wed, Oct 3, 2018 at 1:05 PM Eugeniy Paltsev
> > > <Eugeniy.Paltsev@synopsys.com> wrote:
> > > >
> > > > Validate requested pixel format against bits_per_pixel to reject
> > > > invalid formats with subcomponents length sum is greater than requested
> > > > bits_per_pixel.
> > > >
> > > > weston 5.0.0 with fbdev backend tries to set up an ARGB x8r8g8b8 pixel
> > > > format without bits_per_pixel updating. So it can request
> > > > x8r8g8b8 with 16 bpp which is obviously incorrect and should be
> > > > rejected.
> > > >
> > > > Cc: stable@vger.kernel.org
> > > > Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
> > >
> > > drm fbdev emulation doesn't support changing the pixel format at all.
> > > I think we should reject all such request, not just the invalid ones.
> > > Can you pls respin?
> >
> > FYI I once posted a patch to tighten up the fb-helper pixel format
> > stuff:
> > https://patchwork.freedesktop.org/patch/203189/
>
>
> Hi Daniel,
>
> will you take Ville's patch or should I create the new one which is only related
> to new pixel format validation in drm_fb_helper_check_var() ?

Ville's patch isn't the bugfix we're looking for, but a draft version
of what adding proper format support in drm's fbdev emulation could
look like. With lots&lots of open questions. Not anywhere near ready
for merging, and definitely not stable backport material.

So yes, still want the minimal bugfix.
-Daniel

>
>
> > > Thanks, Daniel
> > >
> > > > ---
> > > >  drivers/gpu/drm/drm_fb_helper.c | 7 +++++++
> > > >  1 file changed, 7 insertions(+)
> > > >
> > > > diff --git a/drivers/gpu/drm/drm_fb_helper.c b/drivers/gpu/drm/drm_fb_helper.c
> > > > index 16ec93b75dbf..4f39da07f053 100644
> > > > --- a/drivers/gpu/drm/drm_fb_helper.c
> > > > +++ b/drivers/gpu/drm/drm_fb_helper.c
> > > > @@ -1610,6 +1610,13 @@ int drm_fb_helper_check_var(struct fb_var_screeninfo *var,
> > > >                 return -EINVAL;
> > > >         }
> > > >
> > > > +       if ((var->green.length + var->blue.length + var->red.length +
> > > > +           var->transp.length) > var->bits_per_pixel) {
> > > > +               DRM_DEBUG("fb requested pixel format can't fit in %d bpp\n",
> > > > +                         var->bits_per_pixel);
> > > > +               return -EINVAL;
> > > > +       }
> > > > +
> > > >         switch (var->bits_per_pixel) {
> > > >         case 16:
> > > >                 depth = (var->green.length == 6) ? 16 : 15;
> > > > --
> > > > 2.14.4
> > > >
> > > > _______________________________________________
> > > > dri-devel mailing list
> > > > dri-devel@lists.freedesktop.org
> > > > https://urldefense.proofpoint.com/v2/url?u=https-3A__lists.freedesktop.org_mailman_listinfo_dri-2Ddevel&d=DwIDAw&c=DPL6_X_6JkXFx7AXWqB0tg&r=ZlJN
> > > > 1MriPUTkBKCrPSx67GmaplEUGcAEk9yPtCLdUXI&m=f12ZyEESIeavtqCUKutiZ9F6xtRFC2UUvdqnM4ywBx8&s=CPS8taMiYbIgXo-fxhqErOJXvO6PMTzmr-BNnGJIoy0&e=
> > >
> > >
> > >
> > > --
> > > Daniel Vetter
> > > Software Engineer, Intel Corporation
> > > +41 (0) 79 365 57 48 - https://urldefense.proofpoint.com/v2/url?u=http-3A__blog.ffwll.ch&d=DwIDAw&c=DPL6_X_6JkXFx7AXWqB0tg&r=ZlJN1MriPUTkBKCrPSx67
> > > GmaplEUGcAEk9yPtCLdUXI&m=f12ZyEESIeavtqCUKutiZ9F6xtRFC2UUvdqnM4ywBx8&s=Vt8OX9s9ljSK6GDgbnwsF-Yd35fbBUfe8SBV2jPnVaQ&e=
> > > _______________________________________________
> > > dri-devel mailing list
> > > dri-devel@lists.freedesktop.org
> > > https://urldefense.proofpoint.com/v2/url?u=https-3A__lists.freedesktop.org_mailman_listinfo_dri-2Ddevel&d=DwIDAw&c=DPL6_X_6JkXFx7AXWqB0tg&r=ZlJN1M
> > > riPUTkBKCrPSx67GmaplEUGcAEk9yPtCLdUXI&m=f12ZyEESIeavtqCUKutiZ9F6xtRFC2UUvdqnM4ywBx8&s=CPS8taMiYbIgXo-fxhqErOJXvO6PMTzmr-BNnGJIoy0&e=
> >
> >
> --
>  Eugeniy Paltsev
Ville Syrjälä Oct. 3, 2018, 3:51 p.m. | #5
On Wed, Oct 03, 2018 at 05:29:34PM +0200, Daniel Vetter wrote:
> On Wed, Oct 3, 2018 at 4:30 PM Eugeniy Paltsev
> <eugeniy.paltsev@synopsys.com> wrote:
> >
> > On Wed, 2018-10-03 at 15:30 +0300, Ville Syrjälä wrote:
> > > On Wed, Oct 03, 2018 at 01:36:00PM +0200, Daniel Vetter wrote:
> > > > On Wed, Oct 3, 2018 at 1:05 PM Eugeniy Paltsev
> > > > <Eugeniy.Paltsev@synopsys.com> wrote:
> > > > >
> > > > > Validate requested pixel format against bits_per_pixel to reject
> > > > > invalid formats with subcomponents length sum is greater than requested
> > > > > bits_per_pixel.
> > > > >
> > > > > weston 5.0.0 with fbdev backend tries to set up an ARGB x8r8g8b8 pixel
> > > > > format without bits_per_pixel updating. So it can request
> > > > > x8r8g8b8 with 16 bpp which is obviously incorrect and should be
> > > > > rejected.
> > > > >
> > > > > Cc: stable@vger.kernel.org
> > > > > Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
> > > >
> > > > drm fbdev emulation doesn't support changing the pixel format at all.
> > > > I think we should reject all such request, not just the invalid ones.
> > > > Can you pls respin?
> > >
> > > FYI I once posted a patch to tighten up the fb-helper pixel format
> > > stuff:
> > > https://patchwork.freedesktop.org/patch/203189/
> >
> >
> > Hi Daniel,
> >
> > will you take Ville's patch or should I create the new one which is only related
> > to new pixel format validation in drm_fb_helper_check_var() ?
> 
> Ville's patch isn't the bugfix we're looking for, but a draft version
> of what adding proper format support in drm's fbdev emulation could
> look like. With lots&lots of open questions.

Actually it does pretty much what you seem to be asking for.
Ie. reject any attempt to change the pixel format. Not really
sure how to do it much more minimally.

Hmm. Oh there is a more minimal way actually. I mistakenly remembered
that fbdev clobbers info->var already before check_var(), but actually
it only does that before set_par(). So I guess all we really need is
to compare the fb_bitfields/bits_per_pixel/etc. between info->var
and the passed in var.

> Not anywhere near ready
> for merging, and definitely not stable backport material.
> 
> So yes, still want the minimal bugfix.
> -Daniel
> 
> >
> >
> > > > Thanks, Daniel
> > > >
> > > > > ---
> > > > >  drivers/gpu/drm/drm_fb_helper.c | 7 +++++++
> > > > >  1 file changed, 7 insertions(+)
> > > > >
> > > > > diff --git a/drivers/gpu/drm/drm_fb_helper.c b/drivers/gpu/drm/drm_fb_helper.c
> > > > > index 16ec93b75dbf..4f39da07f053 100644
> > > > > --- a/drivers/gpu/drm/drm_fb_helper.c
> > > > > +++ b/drivers/gpu/drm/drm_fb_helper.c
> > > > > @@ -1610,6 +1610,13 @@ int drm_fb_helper_check_var(struct fb_var_screeninfo *var,
> > > > >                 return -EINVAL;
> > > > >         }
> > > > >
> > > > > +       if ((var->green.length + var->blue.length + var->red.length +
> > > > > +           var->transp.length) > var->bits_per_pixel) {
> > > > > +               DRM_DEBUG("fb requested pixel format can't fit in %d bpp\n",
> > > > > +                         var->bits_per_pixel);
> > > > > +               return -EINVAL;
> > > > > +       }
> > > > > +
> > > > >         switch (var->bits_per_pixel) {
> > > > >         case 16:
> > > > >                 depth = (var->green.length == 6) ? 16 : 15;
> > > > > --
> > > > > 2.14.4
> > > > >
> > > > > _______________________________________________
> > > > > dri-devel mailing list
> > > > > dri-devel@lists.freedesktop.org
> > > > > https://urldefense.proofpoint.com/v2/url?u=https-3A__lists.freedesktop.org_mailman_listinfo_dri-2Ddevel&d=DwIDAw&c=DPL6_X_6JkXFx7AXWqB0tg&r=ZlJN
> > > > > 1MriPUTkBKCrPSx67GmaplEUGcAEk9yPtCLdUXI&m=f12ZyEESIeavtqCUKutiZ9F6xtRFC2UUvdqnM4ywBx8&s=CPS8taMiYbIgXo-fxhqErOJXvO6PMTzmr-BNnGJIoy0&e=
> > > >
> > > >
> > > >
> > > > --
> > > > Daniel Vetter
> > > > Software Engineer, Intel Corporation
> > > > +41 (0) 79 365 57 48 - https://urldefense.proofpoint.com/v2/url?u=http-3A__blog.ffwll.ch&d=DwIDAw&c=DPL6_X_6JkXFx7AXWqB0tg&r=ZlJN1MriPUTkBKCrPSx67
> > > > GmaplEUGcAEk9yPtCLdUXI&m=f12ZyEESIeavtqCUKutiZ9F6xtRFC2UUvdqnM4ywBx8&s=Vt8OX9s9ljSK6GDgbnwsF-Yd35fbBUfe8SBV2jPnVaQ&e=
> > > > _______________________________________________
> > > > dri-devel mailing list
> > > > dri-devel@lists.freedesktop.org
> > > > https://urldefense.proofpoint.com/v2/url?u=https-3A__lists.freedesktop.org_mailman_listinfo_dri-2Ddevel&d=DwIDAw&c=DPL6_X_6JkXFx7AXWqB0tg&r=ZlJN1M
> > > > riPUTkBKCrPSx67GmaplEUGcAEk9yPtCLdUXI&m=f12ZyEESIeavtqCUKutiZ9F6xtRFC2UUvdqnM4ywBx8&s=CPS8taMiYbIgXo-fxhqErOJXvO6PMTzmr-BNnGJIoy0&e=
> > >
> > >
> > --
> >  Eugeniy Paltsev
> 
> 
> 
> -- 
> Daniel Vetter
> Software Engineer, Intel Corporation
> +41 (0) 79 365 57 48 - http://blog.ffwll.ch

Patch

diff --git a/drivers/gpu/drm/drm_fb_helper.c b/drivers/gpu/drm/drm_fb_helper.c
index 16ec93b75dbf..4f39da07f053 100644
--- a/drivers/gpu/drm/drm_fb_helper.c
+++ b/drivers/gpu/drm/drm_fb_helper.c
@@ -1610,6 +1610,13 @@  int drm_fb_helper_check_var(struct fb_var_screeninfo *var,
 		return -EINVAL;
 	}
 
+	if ((var->green.length + var->blue.length + var->red.length +
+	    var->transp.length) > var->bits_per_pixel) {
+		DRM_DEBUG("fb requested pixel format can't fit in %d bpp\n",
+			  var->bits_per_pixel);
+		return -EINVAL;
+	}
+
 	switch (var->bits_per_pixel) {
 	case 16:
 		depth = (var->green.length == 6) ? 16 : 15;