On Tue, Sep 24, 2019 at 10:28:48AM -0700, Jeykumar Sankaran wrote:
Reviving this thread from the context of the below conversion:
https://lore.kernel.org/linux-arm-msm/db26145b-3f64-a334-f698-76f972332881
@baylibre.com/T/#u
On 2018-10-05 01:19, Neil Armstrong wrote:
> On 05/10/2018 09:58, Daniel Vetter wrote:
> > On Fri, Oct 5, 2018 at 9:39 AM Neil Armstrong
> > <narmstrong@xxxxxxxxxxxx> wrote:
> > >
>
> [...]
>
> > > OK, won't this be enough ?
> > > --- a/include/drm/drm_mode_config.h
> > > +++ b/include/drm/drm_mode_config.h
> > > @@ -333,6 +333,8 @@ struct drm_mode_config_funcs {
> > > * @min_height: minimum fb pixel height on this device
> > > * @max_width: maximum fb pixel width on this device
> > > * @max_height: maximum fb pixel height on this device
> > > + * @max_fb_width: maximum fb buffer width if differs from
max_width
> > > + * @max_fb_height: maximum fb buffer height if differs from
> > > max_height
> > > * @funcs: core driver provided mode setting functions
> > > * @fb_base: base address of the framebuffer
> > > * @poll_enabled: track polling support for this device
> > > @@ -508,6 +510,7 @@ struct drm_mode_config {
> > >
> > > int min_width, min_height;
> > > int max_width, max_height;
> > > + int max_fb_width, max_fb_height;
> > > const struct drm_mode_config_funcs *funcs;
> > > resource_size_t fb_base;
> > >
> > > --- a/drivers/gpu/drm/drm_framebuffer.c
> > > +++ b/drivers/gpu/drm/drm_framebuffer.c
> > > @@ -283,14 +283,20 @@ drm_internal_framebuffer_create(struct
> > > drm_device *dev,
> > > return ERR_PTR(-EINVAL);
> > > }
> > >
> > > - if ((config->min_width > r->width) || (r->width >
> > > config->max_width)) {
> > > + if ((config->min_width > r->width) ||
> > > + (!config->max_fb_width && r->width >
> > > config->max_width) ||
> > > + (config->max_fb_width && r->width >
> > > config->max_fb_width)) {
> > > DRM_DEBUG_KMS("bad framebuffer width %d, should
> > > be >= %d && <= %d\n",
> > > - r->width, config->min_width,
> > > config->max_width);
> > > + r->width, config->min_width,
> > > config->max_fb_width ?
> > > + config->max_fb_width :
config->max_width);
> > > return ERR_PTR(-EINVAL);
> > > }
> > > - if ((config->min_height > r->height) || (r->height >
> > > config->max_height)) {
> > > + if ((config->min_height > r->height) ||
> > > + (!config->max_fb_height && r->height >
> > > config->max_height) ||
> > > + (config->max_fb_height && r->height >
> > > config->max_fb_height)) {
> > > DRM_DEBUG_KMS("bad framebuffer height %d, should
> > > be >= %d && <= %d\n",
> > > - r->height, config->min_height,
> > > config->max_height);
> > > + r->height, config->min_height,
> > > config->max_fb_height ?
> > > + config->max_fb_height :
> > > config->max_height);
> > > return ERR_PTR(-EINVAL);
> > > }
> > >
> > > and in the driver :
> > >
> > > + drm->mode_config.max_width = 4096;
> > > + drm->mode_config.max_height = 3840;
> > > + drm->mode_config.max_fb_width = 16384;
> > > + drm->mode_config.max_fb_height = 8192;
> > >
> > > With this I leave the mode filtering intact.
> >
> > Not enough. See
> >
https://dri.freedesktop.org/docs/drm/gpu/drm-kms-helpers.html#c.drm_connec
tor_helper_funcs
> > and scroll down to mode_valid. You need to filter modes both in the
> > detect paths, and the atomic_check paths.
> >
> > Detect is explicitly filtered out, but atomic_check was only
> > implicitly filtered, through the max fb size checks. Ok, you could
> > light up a mode that's bigger than max fb, but in practice, no
> > userspace ever did that.
Daniel, MSM and few other vendor hardware have upscale blocks where
the
driver can expose fb sizes smaller than
the mode resolution and use h/w upscaling to fill the screen. This
would
optimize the fetch bandwidth.
But with your code we're missing crucial
> > validation now, and userspace could fall over that. What I think we
> > need is to add mode filter against mode_config.max_width/height in
> > drm_atomic_helper_check_modeset(). Probably best to stuff that into
> > the mode_valid() function.
>
Agreed! Since the above patch from Niel is taking care of cases where
max/min fb values
are not set, by checking against the original max/min values, can we
separate out this
core change from the driver level mode_valid changes? If Niel couldn't
find
the time, I can
repost the above change.
Sure, I think Neil wouldn't mind if you take this over and get it ready
for merging. Just need to make sure we're not leaving any validation
gaps
in core/helper code.
-Daniel
