linux-toradex.git/drivers/gpu/drm/exynos, branch v4.14-rc7

drm/exynos: Clear drvdata after component unbind

2017-10-15T22:44:49+00:00

When components are unbound, DRM driver is unregistered and freed,
so clear drvdata to avoid potential use-after-free issue in
suspend/resume paths.

Signed-off-by: Marek Szyprowski 
Signed-off-by: Inki Dae

drm/exynos: Fix potential NULL pointer dereference in suspend/resume paths

2017-10-15T22:44:48+00:00

The patch 6e8edf8a7d8d: "drm/exynos: Fix suspend/resume support" introduced
a new code in suspend/resume paths. However it unconditionally dereference
drm_dev pointer, which might be NULL if suspend/resume happens before
Exynos DRM driver components bind. This patch fixes this issue.

Reported-by: Dan Carpenter 
Fixes: 6e8edf8a7d8d "drm/exynos: Fix suspend/resume support"
Signed-off-by: Marek Szyprowski 
Signed-off-by: Inki Dae

drm/exynos/hdmi: Fix unsafe list iteration

2017-09-20T03:05:23+00:00

Function hdmi_mode_fixup() used bare list_for_each entry, which was
unsafe and caused memory corruption detected by kasan.
It now uses drm_for_each_connector_iter macro, which is now recommended
by the documentation and safe.

Signed-off-by: Maciej Purski 
Signed-off-by: Inki Dae

drm: exynos: include linux/irq.h

2017-09-19T10:50:47+00:00

I ran into a build error on x86:

drivers/gpu/drm/exynos/exynos5433_drm_decon.c: In function 'decon_conf_irq':
drivers/gpu/drm/exynos/exynos5433_drm_decon.c:706:2: error: implicit declaration of function 'irq_set_status_flags'; did you mean 'dquot_state_flag'? [-Werror=implicit-function-declaration]
  irq_set_status_flags(irq, IRQ_NOAUTOEN);

Adding the missing include fixes the error.

Fixes: b37d53a0382c ("drm/exynos/decon5433: move TE handling to DECON")
Signed-off-by: Arnd Bergmann 
Signed-off-by: Inki Dae

drm/exynos: Fix suspend/resume support

2017-09-19T10:50:44+00:00

Commit 7d902c05b480 ("drm: Nuke drm_atomic_helper_connector_dpms")
removed drm_atomic_helper_connector_dpms() helper saying that it was a dead
code. It was however indirectly used by Exynos DRM driver for implementing
suspend/resume support. To fix this regression (after that patch Exynos DRM
suspend/resume functions became no-ops and hardware fails to suspend),
this patch rewrites them with drm_atomic_helper_suspend/resume() helpers.

Fixes: 7d902c05b480 ("drm: Nuke drm_atomic_helper_connector_dpms")
Signed-off-by: Marek Szyprowski 
Acked-by: Krzysztof Kozlowski 
Signed-off-by: Inki Dae

drm/exynos: Fix locking in the suspend/resume paths

2017-09-19T10:50:43+00:00

Commit 48a92916729b ("drm/exynos: use drm_for_each_connector_iter()")
replaced unsafe drm_for_each_connector() with drm_for_each_connector_iter()
and removed surrounding drm_modeset_lock calls. However, that lock was
there not only to protect unsafe drm_for_each_connector(), but it was also
required to be held by the dpms code which was called from the loop body.
This patch restores those drm_modeset_lock calls to fix broken suspend
and resume of Exynos DRM subsystem in v4.13 kernel.

Fixes: 48a92916729b ("drm/exynos: use drm_for_each_connector_iter()")
CC: stable@vger.kernel.org # v4.13
Signed-off-by: Marek Szyprowski 
Acked-by: Krzysztof Kozlowski 
Signed-off-by: Inki Dae

drm/exynos: simplify set_pixfmt() in DECON and FIMD drivers

2017-08-25T05:30:27+00:00

DRM core already checks the validity of the pixelformat.

Signed-off-by: Tobias Jakobi 
Signed-off-by: Inki Dae

drm/exynos: consistent use of cpp

2017-08-25T05:06:48+00:00

A recent commit (272725c7db4da1fd3229d944fc76d2e98e3a144e) has removed
the use of 'bits_per_pixel' in DRM. However the corresponding Exynos
driver code still uses the ambiguous 'bpp', even though it is now
initialized from fb->cpp[0].

Consistenly use 'cpp' in FIMD, DECON7 and DECON5433 drivers.

Signed-off-by: Tobias Jakobi

drm/exynos: mixer: remove src offset from mixer_graph_buffer()

2017-08-25T03:10:33+00:00

We always translate the dma address such that the offsets of
the source image are zero. Hence we can remove manipulation of
the MXR_GRAPHIC_SXY(win) register and just zero them once
in mixer_win_reset().

Signed-off-by: Tobias Jakobi 
Signed-off-by: Inki Dae

drm/exynos: mixer: simplify mixer_graph_buffer()

2017-08-25T03:07:11+00:00

DRM core already checks in drm_atomic_plane_check() if the
pixelformat is valid. Hence we can collapse the default case
of the switch statement with the XRGB8888 case.

No functional change.

Signed-off-by: Tobias Jakobi 
Signed-off-by: Inki Dae