linux-kernel - Re: [PATCH v2 34/43] drm/sun4i: Add buffer stride and offset configuration for tiling mode

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] [day] [month] [year] [list]

Date:   Tue, 04 Dec 2018 14:30:59 +0100
From:   Paul Kocialkowski <paul.kocialkowski@...tlin.com>
To:     Maxime Ripard <maxime.ripard@...tlin.com>
Cc:     linux-kernel@...r.kernel.org, dri-devel@...ts.freedesktop.org,
        linux-arm-kernel@...ts.infradead.org,
        Maarten Lankhorst <maarten.lankhorst@...ux.intel.com>,
        Sean Paul <sean@...rly.run>, David Airlie <airlied@...ux.ie>,
        Chen-Yu Tsai <wens@...e.org>,
        Thomas Petazzoni <thomas.petazzoni@...tlin.com>,
        linux-sunxi@...glegroups.com, Daniel Vetter <daniel@...ll.ch>
Subject: Re: [PATCH v2 34/43] drm/sun4i: Add buffer stride and offset
 configuration for tiling mode

Hi,

On Tue, 2018-11-27 at 10:24 +0100, Maxime Ripard wrote:
> On Fri, Nov 23, 2018 at 10:25:06AM +0100, Paul Kocialkowski wrote:
> > This introduces stride and offset configuration for the VPU tiling mode.
> > Stride is calculated differently than it is for linear formats and an
> > offset is calculated, for which new register definitions are introduced.
> > 
> > Signed-off-by: Paul Kocialkowski <paul.kocialkowski@...tlin.com>
> > ---
> >  drivers/gpu/drm/sun4i/sun4i_frontend.c | 54 ++++++++++++++++++++++++--
> >  drivers/gpu/drm/sun4i/sun4i_frontend.h |  7 ++++
> >  2 files changed, 58 insertions(+), 3 deletions(-)
> > 
> > diff --git a/drivers/gpu/drm/sun4i/sun4i_frontend.c b/drivers/gpu/drm/sun4i/sun4i_frontend.c
> > index efa1ff0802bd..3f76a5572449 100644
> > --- a/drivers/gpu/drm/sun4i/sun4i_frontend.c
> > +++ b/drivers/gpu/drm/sun4i/sun4i_frontend.c
> > @@ -125,21 +125,69 @@ void sun4i_frontend_update_buffer(struct sun4i_frontend *frontend,
> >  {
> >  	struct drm_plane_state *state = plane->state;
> >  	struct drm_framebuffer *fb = state->fb;
> > +	unsigned int strides[3] = {};
> > +
> >  	dma_addr_t paddr;
> >  	bool swap;
> >  
> > +	if (fb->modifier == DRM_FORMAT_MOD_ALLWINNER_TILED) {
> > +		unsigned int width = state->src_w >> 16;
> > +		unsigned int offset;
> > +
> > +		/*
> > +		 * In MB32 tiled mode, the stride is defined as the distance
> > +		 * between the start of the end line of the current tile and
> > +		 * the start of the first line in the next vertical tile.
> > +		 *
> > +		 * Tiles are represented in row-major order, thus the end line
> > +		 * of current tile starts at: 31 * 32 (31 lines of 32 cols),
> > +		 * the next vertical tile starts at: 32-bit-aligned-width * 32
> > +		 * and the distance is: 32 * (32-bit-aligned-width - 31).
> > +		 */
> > +
> > +		strides[0] = (fb->pitches[0] - 31) * 32;
> > +
> > +		/* Offset of the bottom-right point in the end tile. */
> > +		offset = (width + (32 - 1)) & (32 - 1);
> 
> Those computations are a bit obscure. I guess adding a bunch of
> defines, and using the round_up / _down and ALIGN macros would help

Fair enough, I will add a more explicit comment and a macro in v3.

> > +		regmap_write(frontend->regs, SUN4I_FRONTEND_TB_OFF0_REG,
> > +			     SUN4I_FRONTEND_TB_OFF_X1(offset));
> > +
> > +		if (fb->format->num_planes > 1) {
> > +			strides[1] = (fb->pitches[1] - 31) * 32;
> > +
> > +			regmap_write(frontend->regs, SUN4I_FRONTEND_TB_OFF1_REG,
> > +				     SUN4I_FRONTEND_TB_OFF_X1(offset));
> > +		}
> > +
> > +		if (fb->format->num_planes > 2) {
> > +			strides[2] = (fb->pitches[2] - 31) * 32;
> > +
> > +			regmap_write(frontend->regs, SUN4I_FRONTEND_TB_OFF2_REG,
> > +				     SUN4I_FRONTEND_TB_OFF_X1(offset));
> > +		}
> 
> I guess we could fall in a situation where this is not cleared when
> moving from a format with 3 planes to one with 2 for example. Would
> that break anything?

I have tested this case and nothing breaks so we're good!

Cheers,

Paul

-- 
Paul Kocialkowski, Bootlin (formerly Free Electrons)
Embedded Linux and kernel engineering
https://bootlin.com