[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20230410122607.GA11253@pendragon.ideasonboard.com>
Date: Mon, 10 Apr 2023 15:26:07 +0300
From: Laurent Pinchart <laurent.pinchart@...asonboard.com>
To: Geert Uytterhoeven <geert@...ux-m68k.org>
Cc: Kieran Bingham <kieran.bingham+renesas@...asonboard.com>,
David Airlie <airlied@...il.com>,
Daniel Vetter <daniel@...ll.ch>,
dri-devel@...ts.freedesktop.org, linux-renesas-soc@...r.kernel.org,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH 3/5] drm: shmobile: Switch to drm_crtc_init_with_planes()
Hi Geert,
On Mon, Apr 10, 2023 at 11:35:56AM +0200, Geert Uytterhoeven wrote:
> On Wed, Apr 5, 2023 at 5:59 AM Laurent Pinchart wrote:
> > On Fri, Mar 31, 2023 at 04:48:09PM +0200, Geert Uytterhoeven wrote:
> > > The SH-Mobile DRM driver uses the legacy drm_crtc_init(), which
> > > advertizes only the formats in safe_modeset_formats[] (XR24 and AR24) as
> > > being supported.
> > >
> > > Switch to drm_crtc_init_with_planes(), and advertize all supported
> > > (A)RGB modes, so we can use RGB565 as the default mode for the console.
> > >
> > > Signed-off-by: Geert Uytterhoeven <geert+renesas@...der.be>
>
> > > --- a/drivers/gpu/drm/shmobile/shmob_drm_crtc.c
> > > +++ b/drivers/gpu/drm/shmobile/shmob_drm_crtc.c
> > > @@ -18,6 +18,7 @@
> > > #include <drm/drm_gem_dma_helper.h>
> > > #include <drm/drm_modeset_helper.h>
> > > #include <drm/drm_modeset_helper_vtables.h>
> > > +#include <drm/drm_plane_helper.h>
> > > #include <drm/drm_probe_helper.h>
> > > #include <drm/drm_simple_kms_helper.h>
> > > #include <drm/drm_vblank.h>
> > > @@ -478,16 +479,41 @@ static const struct drm_crtc_funcs crtc_funcs = {
> > > .disable_vblank = shmob_drm_disable_vblank,
> > > };
> > >
> > > +static const uint32_t modeset_formats[] = {
> > > + DRM_FORMAT_RGB565,
> > > + DRM_FORMAT_RGB888,
> > > + DRM_FORMAT_ARGB8888,
> > > + DRM_FORMAT_XRGB8888,
> > > +};
> > > +
> > > +static const struct drm_plane_funcs primary_plane_funcs = {
> > > + DRM_PLANE_NON_ATOMIC_FUNCS,
> > > +};
> > > +
> > > int shmob_drm_crtc_create(struct shmob_drm_device *sdev)
> > > {
> > > struct drm_crtc *crtc = &sdev->crtc.crtc;
> > > + struct drm_plane *primary;
> > > int ret;
> > >
> > > sdev->crtc.dpms = DRM_MODE_DPMS_OFF;
> > >
> > > - ret = drm_crtc_init(sdev->ddev, crtc, &crtc_funcs);
> > > - if (ret < 0)
> > > + primary = __drm_universal_plane_alloc(sdev->ddev, sizeof(*primary), 0,
> > > + 0, &primary_plane_funcs,
> > > + modeset_formats,
> > > + ARRAY_SIZE(modeset_formats),
> > > + NULL, DRM_PLANE_TYPE_PRIMARY,
> > > + NULL);
> > > + if (IS_ERR(primary))
> > > + return PTR_ERR(primary);
> >
> > This seems like a bit of a hack to me. Why don't you use the planes
>
> I'm following what Thomas did in the nouveau driver....
>
> > created by shmob_drm_plane_create() instead of allocating a new one ?
>
> Is that possible? shmob_drm_plane_create() creates overlay planes,
> while this is for the primary plane.
You're right, for some reason I overlooked that. Sorry for the noise.
It would be good to move handling of the primary plane to
shmob_drm_plane.c, but that's for later, when moving the driver to the
atomic API. For now,
Reviewed-by: Laurent Pinchart <laurent.pinchart+renesas@...asonboard.com>
> > > +
> > > + ret = drm_crtc_init_with_planes(sdev->ddev, crtc, primary, NULL,
> > > + &crtc_funcs, NULL);
> > > + if (ret < 0) {
> > > + drm_plane_cleanup(primary);
> > > + kfree(primary);
> > > return ret;
> > > + }
> > >
> > > drm_crtc_helper_add(crtc, &crtc_helper_funcs);
--
Regards,
Laurent Pinchart
Powered by blists - more mailing lists