drm/sun4i: backend: Support YUV planes

Now that we have the guarantee that we will have only a single YUV plane,
actually support them. The way it works is not really straightforward,
since we first need to enable the YUV mode in the plane that we want to
setup, and then we have a few registers to setup the YUV buffer and
parameters.

We also need to setup the color correction to actually have something
displayed.

Reviewed-by: Chen-Yu Tsai <[email protected]>
Signed-off-by: Maxime Ripard <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/66088c1398bd3189123f28a89a7ccc669fe9f296.1519931807.git-series.maxime.ripard@bootlin.com

drm/sun4i: backend: Check that we only have a single YUV plane

Just like for the frontend, a single plane can use a YUV format. Make sure
we have that constraint covered in our atomic_check.

This is preliminary to the actual YUV support to make sure we don't end up
in an impossible to support situation.

Reviewed-by: Chen-Yu Tsai <[email protected]>
Signed-off-by: Maxime Ripard <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/2f8586493d9139b12efe7e94f65e9a149f818e0e.1519931807.git-series.maxime.ripard@bootlin.com

drm/sun4i: Add driver support for A80 display pipeline

This patch adds support for the compatible strings of the A80 display
pipeline.

Signed-off-by: Chen-Yu Tsai <[email protected]>
Signed-off-by: Maxime Ripard <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/sun4i: Add compatible strings for the A80 display pipeline

This patch adds compatible strings for the remaining documented
components of the Allwinner A80 display pipeline.

Reviewed-by: Rob Herring <[email protected]>
Signed-off-by: Chen-Yu Tsai <[email protected]>
Signed-off-by: Maxime Ripard <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/sun4i: Add support for A80 TCONs

The Allwinner A80 SoC has 2 documented TCONs. The display pipeline
diagram from the user manual shows a third TCON, but it's missing
an interrupt line, and its registers are not explained either.
It's also not used in Allwinner's vendor BSP.

The first TCON only has channel 0, for LCD panel output. The TCON
hardware setup is peculiar in that the eDP reset must also be
deasserted to allow access to the TCON. How the eDP module is wired
in the SoC itself is never explained.

The second TCON only has channel 1, and its output is connected to
the HDMI encoder block.

This patch adds a "needs_edp_reset" field to the tcon quirks structure,
and adds quirks and compatible strings for the 2 documented TCONs.

Signed-off-by: Chen-Yu Tsai <[email protected]>
Signed-off-by: Maxime Ripard <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/sun4i: Add DT binding for Detail Enhancement Unit in Allwinner A80 SoC

The display pipeline on the A80 SoC has what is called the Detail
Enhancement Unit, or DEU for short, block in between the display
frontend and backend. This unit can sharpen images in both luma
and chroma channels. It seems to also do colorspace conversion.

This patch adds the device tree binding for this hardware block.

Reviewed-by: Rob Herring <[email protected]>
Signed-off-by: Chen-Yu Tsai <[email protected]>
Signed-off-by: Maxime Ripard <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/sun4i: Add compatible strings for A80 TCONs

The A80 has 2 or 3 TCONs. The documentation and vendor kernel are very
vague about the third TCON, to the point that it might not exist.

In the documentation, the first TCON is missing channel 1, and the
second is missing channel 0. However the vendor kernel seems to be
able to use them regardless. Here we model them like the old TCONs.

An oddity is that TCON0 requires the reset control for the eDP block
to be deasserted, for any register access to stick.

This patch adds compatible strings for TCON0 and TCON1, with TCON0
requiring an extra "edp" reset control.

Reviewed-by: Rob Herring <[email protected]>
Signed-off-by: Chen-Yu Tsai <[email protected]>
Signed-off-by: Maxime Ripard <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm: Reduce object size of DRM_DEV_<LEVEL> uses

These macros are similar to the DRM_<LEVEL> with the addition
of a struct device * to the arguments.

Convert the single drm_dev_printk function into 2 separate functions.
drm_dev_printk with a KERN_<LEVEL> * for generic use and drm_dev_dbg
for conditional masked use.

Remove the __func__ argument and use __builtin_return_address(0) to be
similar to the DRM_<LEVEL> macros uses.

Convert the DRM_DEV_<LEVEL> macros to remove now unnecessary arguments
and use a consistent style.

These macros are rarely used in the generic gpu/drm code so the code
size does not change much for a defconfig, but when more drivers are
enabled, there is ~4k savings.

Many of these macros have no existing use at all.

$ size -t drivers/gpu/drm/built-in.a | tail -1
1877530   44651     995 1923176  1d5868 (TOTALS)

$ size -t drivers/gpu/drm/built-in.a | tail -1
1877527   44651     995 1923173  1d5865 (TOTALS)

$ size -t drivers/gpu/drm/built-in.a | tail -1
17166750 2689238  108352 19964340 130a1b4 (TOTALS)

$ size -t drivers/gpu/drm/built-in.a | tail -1
17168888 2691734  108352 19968974 130b3ce (TOTALS)

Signed-off-by: Joe Perches <[email protected]>
Signed-off-by: Daniel Vetter <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/e5c164946e15375ac71b69b75f296efdf0b76e6d.1521233717.git.joe@perches.com

drm/doc: Put all driver docs into a separate chapter

We have quite a few driver docs now, which is great, but having them
all in the top-level gpu documentation chapter makes it harder to spot
the core/shared bits.

Stuff them into a separate chapter and ecourage people to add even
more!

Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Daniel Vetter <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm: dma_bufs: Fixed checkpatch issues

Fix a couple of checkpatch issues

Signed-off-by: Paul McQuade <[email protected]>
[seanpaul squashed series of 4 into one patch, and changed commit msg]
Signed-off-by: Sean Paul <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm: remove drm_mode_object_{un/reference} aliases

This patch remove the compatibility aliases
drm_mode_object_{reference/unreference} of drm_mode_object_{get/put}
since all callers have been converted to the prefered _{get/put}.

Remove the helpers from the semantic patch drm-get-put-cocci.

Signed-off-by: Haneen Mohammed <[email protected]>
Signed-off-by: Sean Paul <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/20180319055820.GA17502@haneen-VirtualBox

drm: Add PSR version 3 macro

eDP 1.4a specification defines PSR version 3, it PSR2 with the
addition of Y-coordinate support when doing selective update.

Signed-off-by: José Roberto de Souza <[email protected]>
Signed-off-by: Jani Nikula <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/vc4_validate: Remove VLA usage

In preparation to enabling -Wvla, remove VLA. In this particular
case use macro ARRAY_SIZE so the length of array _bo_ can be
computed at preprocessing time.

The use of stack Variable Length Arrays needs to be avoided, as they
can be a vector for stack exhaustion, which can be both a runtime bug
or a security flaw. Also, in general, as code evolves it is easy to
lose track of how big a VLA can get. Thus, we can end up having runtime
failures that are hard to debug.

Also, fixed as part of the directive to remove all VLAs from
the kernel: https://lkml.org/lkml/2018/3/7/621

Signed-off-by: Gustavo A. R. Silva <[email protected]>
Signed-off-by: Eric Anholt <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/20180313143151.GA27486@embeddedgus

drm: Make drm_mode_vrefresh() a bit more accurate

Do the refresh rate calculation with a single division. This gives
us slightly more accurate results, especially for interlaced since
we don't just double the final truncated result.

We do lose one bit compared to the old way, so with an interlaced
mode the new code can only handle ~2GHz instead of the ~4GHz the
old code handeled.

Signed-off-by: Ville Syrjälä <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]
Reviewed-by: Daniel Vetter <[email protected]>

drm: Nuke the useless 'ret' variable from drm_mode_convert_umode()

No need to store the return value in a variable since we don't have to
do any unwinding.

Signed-off-by: Ville Syrjälä <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]
Reviewed-by: Daniel Vetter <[email protected]>

drm/i915: Use drm_color_lut_size()

Avoid all the sizeof(drm_color_lut) business by using
drm_color_lut_size() to convert the blob length into
number of LUT entries.

Signed-off-by: Ville Syrjälä <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]
Reviewed-by: Daniel Vetter <[email protected]>

drm/i915: Remove the blob->data casts

Now that blob->data is void* again we don't need to cast it.

Signed-off-by: Ville Syrjälä <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]
Reviewed-by: Daniel Vetter <[email protected]>

drm: Introduce drm_color_lut_size()

Provide a small helper to convert the blob length in bytes
to the number of LUT entries.

v2: Add kerneldoc (Daniel)

Cc: Daniel Vetter <[email protected]>
Signed-off-by: Ville Syrjälä <[email protected]>
Reviewed-by: Daniel Vetter <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm: Verify gamma/degamma LUT size

While we want to potentially support multiple different gamma/degamma
LUT sizes we can (and should) at least check that the blob length
is a multiple of the LUT entry size.

v2: s/expected_size_mod/expected_elem_size/ (Daniel)
    Add kernel doc (Daniel)
v3: s/we/were/ typo in the docs

Cc: Daniel Vetter <[email protected]>
Signed-off-by: Ville Syrjälä <[email protected]>
Reviewed-by: Daniel Vetter <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm: Remove now pointelss blob->data casts

Now that blob->data is void* again we don't need the casts anymore.

Signed-off-by: Ville Syrjälä <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]
Reviewed-by: Shashank Sharma <[email protected]>
Reviewed-by: Daniel Vetter <[email protected]>

Revert "drm: Use a flexible array member for blob property data"

Using a flexible array for the blob data was a mistake by me. It
forces all users of the blob data to cast blob->data to something
else. void* is clearly superior so let's go back to the original
scheme.

Not a clean revert as the code has moved.

This reverts commit d63f5e6bf6f2a1573ea39c9937cdf5ab0b3a4b77.

Signed-off-by: Ville Syrjälä <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]
Reviewed-by: Shashank Sharma <[email protected]>
Reviewed-by: Daniel Vetter <[email protected]>

drm/rockchip: cdn-dp: remove the DP phy switch

There are 2 Type-c PHYs in RK3399, but only one DP controller. Hence
only one PHY can connect to DP controller at one time, the other should
be disconnected. The GRF_SOC_CON26 register has a switch bit to do it,
set this bit means enable PHY 1, clear this bit means enable PHY 0.

If the board has 2 Type-C ports, the DP driver get the phy id from
devm_of_phy_get_by_index, and then control this switch according to
this id. But some others board only has one Type-C port, it may be PHY 0
or PHY 1. The dts node id can not tell us the correct PHY id. Hence move
this switch to PHY driver, the PHY driver can distinguish between PHY 0
and PHY 1, and then write the correct register bit.

Signed-off-by: Chris Zhong <[email protected]>
Signed-off-by: Enric Balletbo i Serra <[email protected]>
[The phy-changes are in the phy-tree now and the cdn-dp wasn't
 enabled at all so far, so this change can go through drm-misc
 alone without causing issues when testing drm-misc]
Signed-off-by: Heiko Stuebner <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm: Reduce object size of DRM_ERROR and DRM_DEBUG uses

drm_printk is used for both DRM_ERROR and DRM_DEBUG with unnecessary
arguments that can be removed by creating separate functins.

Create specific functions for these calls to reduce x86/64 defconfig
size by ~20k.

Modify the existing macros to use the specific calls.

new:
$ size -t drivers/gpu/drm/built-in.a | tail -1
1876562   44542     995 1922099  1d5433 (TOTALS)

old:
$ size -t drivers/gpu/drm/built-in.a | tail -1
1897565   44542     995 1943102  1da63e (TOTALS)

Miscellanea:

o intel_display requires a change to use the specific calls.

Signed-off-by: Joe Perches <[email protected]>
Signed-off-by: Daniel Vetter <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/016b5cb84cede20fd0f91ed6965421d99fd5f2ce.1520978414.git.joe@perches.com

drm/bridge: dw-hdmi: Remove unused hdmi_enable_overflow_interrupts()

The cable_plugin member never receives an assignment, so it is always
false, which causes hdmi_enable_overflow_interrupts() to never
be called as per the logic below:

if (hdmi->cable_plugin && hdmi->sink_is_hdmi)
hdmi_enable_overflow_interrupts(hdmi);

This has been the case since the driver was originally introduced
in commit 9aaf880ed4ee ("imx-drm: Add mx6 hdmi transmitter support").

Remove the cable_plugin element and the hdmi_enable_overflow_interrupts()
function that is never called.

Signed-off-by: Fabio Estevam <[email protected]>
Tested-by: Laurent Pinchart <[email protected]> # On R-Car H3
Reviewed-by: Laurent Pinchart <[email protected]>
Signed-off-by: Andrzej Hajda <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/vmwgfx: replace drm_*_unreference with drm_*_put

This patch replace instances of drm_framebuffer_unreference with _put()
suffix, because it is shorter and consistent with the kernel use of
*_get/put() suffixes.
This was done with the following Coccinelle script:

@r@
expression e;
@@

(
-drm_framebuffer_reference(e);
+drm_framebuffer_get(e);
|
-drm_framebuffer_unreference(e);
+drm_framebuffer_put(e);
)

Signed-off-by: Haneen Mohammed <[email protected]>
Acked-by: Sinclair Yeh <[email protected]>
Signed-off-by: Daniel Vetter <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/20180311233313.GA19721@Haneen

drm/rockchip: Don't use spin_lock_irqsave in interrupt context

The rockchip DRM driver is quite careful to disable interrupts
when taking a lock that is also taken in interrupt context,
which is a good thing.

What is a bit over the top is to use spin_lock_irqsave when
already in interrupt context, as you cannot take another
interrupt again, and disabling interrupt is just pure
overhead.

Switching to the non _irqsave version in interrupt context is
more logical, and less heavy handed.

Signed-off-by: Marc Zyngier <[email protected]>
Signed-off-by: Heiko Stuebner <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/rockchip: Do not use memcpy for MMIO addresses

memcpy is only meant to be used for memory, and only that.
MMIO accessors should be used to access MMIO regions, preferably
the ones that correspond to the size of the register accessed.

Let's convert the bulk register copy to writel/readl_relaxed,
which is the correct API.

Signed-off-by: Marc Zyngier <[email protected]>
Signed-off-by: Heiko Stuebner <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/rockchip: Clear all interrupts before requesting the IRQ

Calling request_irq() followed by disable_irq() is usually a bad idea,
specially if the interrupt can be pending, and you're not yet in a
position to handle it.

This is exactly what happens on my kevin system when rebooting in a
second kernel using kexec: Some interrupt is left pending from
the previous kernel, and we take it too early, before disable_irq()
could do anything.

Let's clear the pending interrupts as we initialize the HW, and move
the interrupt request after that point. This ensures that we're in
a sane state when the interrupt is requested.

Cc: [email protected]
Signed-off-by: Marc Zyngier <[email protected]>
[adapted to recent rockchip-drm changes]
Signed-off-by: Heiko Stuebner <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/bridge: analogix_dp: Only wait for panel ACK on PSR entry

We currently wait for the panel to mirror our intended PSR state
before continuing on both PSR enter and PSR exit. This is really
only important to do when we're entering PSR, since we want to
be sure the last frame we pushed is being served from the panel's
internal fb before shutting down the soc blocks (vop/analogix).

This patch changes the behavior such that we only wait for the
panel to complete the PSR transition when we're entering PSR, and
to skip verification when we're exiting.

Cc: Stéphane Marchesin <[email protected]>
Cc: Sonny Rao <[email protected]>
Signed-off-by: zain wang <[email protected]>
Signed-off-by: Sean Paul <[email protected]>
Signed-off-by: Thierry Escande <[email protected]>
Signed-off-by: Enric Balletbo i Serra <[email protected]>
Tested-by: Marek Szyprowski <[email protected]>
Reviewed-by: Archit Taneja <[email protected]>
Signed-off-by: Heiko Stuebner <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/panel: rm68200: Add backlight dependency

Like many other panel drivers, this one fails to build when backlight
support is disabled:

drivers/gpu/drm/panel/panel-raydium-rm68200.o: In function `rm68200_probe':
panel-raydium-rm68200.c:(.text+0x14a): undefined reference to `devm_of_find_backlight'

This adds the appropriate dependency.

Note that while include/linux/backlight.h provides a stub inline when
backlight support is not enabled, this isn't enough to deal with the
case where backlight support is built as a module but the panel driver
is built-in, in which case linking will still fail as above.

One way to avoid this is to add a dependency such as this:

        depends on BACKLIGHT_CLASS_DEVICE || BACKLIGHT_CLASS_DEVICE=n

but that is rather complex and misses the point that the panel support
is mostly useless without backlight support.

Fixes: 2b7ed18bed1a ("drm/panel: Add support for Raydium RM68200 panel driver")
Signed-off-by: Arnd Bergmann <[email protected]>
[[email protected]: clarify the need for the dependency]
Signed-off-by: Thierry Reding <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/rockchip: add mutex vop lock

Add a lock to vop to avoid disabling the crtc while waiting for a line
flag while enabling psr. If we disable in the middle of waiting for the
line flag, we'll end up timing out or worse.

Signed-off-by: zain wang <[email protected]>
Signed-off-by: Sean Paul <[email protected]>
Signed-off-by: Thierry Escande <[email protected]>
Signed-off-by: Enric Balletbo i Serra <[email protected]>
Signed-off-by: Heiko Stuebner <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/bridge: analogix_dp: add fast link train for eDP

We would meet a short black screen when exit PSR with the full link
training, In this case, we should use fast link train instead of full
link training.

Signed-off-by: zain wang <[email protected]>
Signed-off-by: Sean Paul <[email protected]>
Signed-off-by: Thierry Escande <[email protected]>
Signed-off-by: Enric Balletbo i Serra <[email protected]>
Tested-by: Marek Szyprowski <[email protected]>
Reviewed-by: Archit Taneja <[email protected]>
[dropped header reordering]
Signed-off-by: Heiko Stuebner <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/bridge: analogix_dp: Don't change psr while bridge is disabled

There is a race between AUX CH bring-up and enabling bridge which will
cause link training to fail. To avoid hitting it, don't change psr state
while enabling the bridge.

Cc: Tomeu Vizoso <[email protected]>
Cc: Sean Paul <[email protected]>
Signed-off-by: zain wang <[email protected]>
Signed-off-by: Caesar Wang <[email protected]>
[seanpaul fixed up the commit message a bit and renamed *_supported to *_enabled]
Signed-off-by: Sean Paul <[email protected]>
Signed-off-by: Thierry Escande <[email protected]>
Signed-off-by: Enric Balletbo i Serra <[email protected]>
Tested-by: Marek Szyprowski <[email protected]>
Reviewed-by: Archit Taneja <[email protected]>
Signed-off-by: Heiko Stuebner <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/rockchip: Remove analogix psr worker

Now that the spinlocks and timers are gone, we can remove the psr
worker located in rockchip's analogix driver and do the enable/disable
directly. This should simplify the code and remove races on disable.

Cc: 征增 王 <[email protected]>
Cc: Stéphane Marchesin <[email protected]>
Signed-off-by: Sean Paul <[email protected]>
Signed-off-by: Thierry Escande <[email protected]>
Signed-off-by: Enric Balletbo i Serra <[email protected]>
Signed-off-by: Heiko Stuebner <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/bridge: analogix_dp: detect Sink PSR state after configuring the PSR

Make sure the request PSR state takes effect in analogix_dp_send_psr_spd()
function, or print the sink PSR error state if we failed to apply the
requested PSR setting.

Cc: 征增 王 <[email protected]>
Cc: Stéphane Marchesin <[email protected]>
Signed-off-by: Yakir Yang <[email protected]>
[seanpaul changed timeout loop to a readx poll]
Signed-off-by: Sean Paul <[email protected]>
Signed-off-by: Thierry Escande <[email protected]>
Signed-off-by: Enric Balletbo i Serra <[email protected]>
Tested-by: Marek Szyprowski <[email protected]>
Reviewed-by: Archit Taneja <[email protected]>
Signed-off-by: Heiko Stuebner <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/sun4i: Handle DRM_BUS_FLAG_PIXDATA_*EDGE

Handle both positive and negative dclk polarity,
according to bus_flags, taking care of this:

On A20 and similar SoCs, the only way to achieve Positive Edge
(Rising Edge), is setting dclk clock phase to 2/3(240°).
By default TCON works in Negative Edge(Falling Edge), this is why phase
is set to 0 in that case.
Unfortunately there's no way to logically invert dclk through IO_POL
register.
The only acceptable way to work, triple checked with scope,
is using clock phase set to 0° for Negative Edge and set to 240° for
Positive Edge.
On A33 and similar SoCs there would be a 90° phase option, but it divides
also dclk by 2.
This patch is a way to avoid quirks all around TCON and DOTCLOCK drivers
for using A33 90° phase divided by 2 and consequently increase code
complexity.

Signed-off-by: Giulio Benetti <[email protected]>
Signed-off-by: Maxime Ripard <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/1520963677-124239-1-git-send-email-giulio.benetti@micronovasrl.com

drm/sun4i: add lvds mode_valid function

mode_valid function is missing for lvds.

Add it making it pointed by encoder helper functions.

Signed-off-by: Giulio Benetti <[email protected]>
Signed-off-by: Maxime Ripard <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/1520940019-68977-1-git-send-email-giulio.benetti@micronovasrl.com

drm/sun4i: move rgb mode_valid from connector to encoder

mode_valid function must be connected to encoder.
Otherwise it could get not be called by drm in the case there's a
bridge connected to encoder instead of a panel.

Move mode_valid function pointer to encoder helper functions,
changing its prototype according to encoder helper function pointer.

Signed-off-by: Giulio Benetti <[email protected]>
Signed-off-by: Maxime Ripard <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/1520941017-81177-1-git-send-email-giulio.benetti@micronovasrl.com

drm/nouveau: Runtime suspend despite HDA being unbound

Commit 5addcf0a5f0f ("nouveau: add runtime PM support (v0.9)") prevents
runtime suspend of the GPU if its integrated HDA controller is not bound
to a driver.  The rationale appears to be that probing the HDA fails if
the GPU is in D3cold.

However we now use a device link to ensure that the GPU is runtime
resumed while the HDA controller is probed, rendering this safety
measure obsolete.  Remove it.

Cc: Dave Airlie <[email protected]>
Cc: Ben Skeggs <[email protected]>
Cc: Takashi Iwai <[email protected]>
Cc: Alex Deucher <[email protected]>
Cc: Rafael J. Wysocki <[email protected]>
Reviewed-by: Peter Wu <[email protected]>
Tested-by: Denis Lisov <[email protected]> # Nvidia Optimus
Tested-by: Peter Wu <[email protected]> # Nvidia Optimus
Tested-by: Lukas Wunner <[email protected]> # MacBook Pro
Signed-off-by: Lukas Wunner <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/77e0ab74f3377ea9b6acf8fab624acfb4f7dbeca.1520068884.git.lukas@wunner.de

vga_switcheroo: Let HDA autosuspend on mux change

When switching the display on muxed machines, we currently force the HDA
controller into runtime suspend on the previously used GPU and into
runtime active state on the newly used GPU.

That's unnecessary if the GPU uses driver power control, we can just let
the audio device autosuspend or autoresume as it sees fit.

Cc: Dave Airlie <[email protected]>
Cc: Ben Skeggs <[email protected]>
Cc: Takashi Iwai <[email protected]>
Cc: Alex Deucher <[email protected]>
Reviewed-by: Peter Wu <[email protected]>
Signed-off-by: Lukas Wunner <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/098ed883460eb4976a899eac6f5192fefc877c0f.1520068884.git.lukas@wunner.de

vga_switcheroo: Use device link for HDA controller

Back in 2013, runtime PM for GPUs with integrated HDA controller was
introduced with commits 0d69704ae348 ("gpu/vga_switcheroo: add driver
control power feature. (v3)") and 246efa4a072f ("snd/hda: add runtime
suspend/resume on optimus support (v4)").

Briefly, the idea was that the HDA controller is forced on and off in
unison with the GPU.

The original code is mostly still in place even though it was never a
100% perfect solution:  E.g. on access to the HDA controller, the GPU
is powered up via vga_switcheroo_runtime_resume_hdmi_audio() but there
are no provisions to keep it resumed until access to the HDA controller
has ceased:  The GPU autosuspends after 5 seconds, rendering the HDA
controller inaccessible.

Additionally, a kludge is required when hda_intel.c probes:  It has to
check whether the GPU is powered down (check_hdmi_disabled()) and defer
probing if so.

However in the meantime (in v4.10) the driver core has gained a feature
called device links which promises to solve such issues in a clean way:
It allows us to declare a dependency from the HDA controller (consumer)
to the GPU (supplier).  The PM core then automagically ensures that the
GPU is runtime resumed as long as the HDA controller's ->probe hook is
executed and whenever the HDA controller is accessed.

By default, the HDA controller has a dependency on its parent, a PCIe
Root Port.  Adding a device link creates another dependency on its
sibling:

                            PCIe Root Port
                             ^          ^
                             |          |
                             |          |
                            HDA  ===>  GPU

The device link is not only used for runtime PM, it also guarantees that
on system sleep, the HDA controller suspends before the GPU and resumes
after the GPU, and on system shutdown the HDA controller's ->shutdown
hook is executed before the one of the GPU.  It is a complete solution.

Using this functionality is as simple as calling device_link_add(),
which results in a dmesg entry like this:

        pci 0000:01:00.1: Linked as a consumer to 0000:01:00.0

The code for the GPU-governed audio power management can thus be removed
(except where it's still needed for legacy manual power control).

The device link is added in a PCI quirk rather than in hda_intel.c.
It is therefore legal for the GPU to runtime suspend to D3cold even if
the HDA controller is not bound to a driver or if CONFIG_SND_HDA_INTEL
is not enabled, for accesses to the HDA controller will cause the GPU to
wake up regardless if they're occurring outside of hda_intel.c (think
config space readout via sysfs).

Contrary to the previous implementation, the HDA controller's power
state is now self-governed, rather than GPU-governed, whereas the GPU's
power state is no longer fully self-governed.  (The HDA controller needs
to runtime suspend before the GPU can.)

It is thus crucial that runtime PM is always activated on the HDA
controller even if CONFIG_SND_HDA_POWER_SAVE_DEFAULT is set to 0 (which
is the default), lest the GPU stays awake.  This is achieved by setting
the auto_runtime_pm flag on every codec and the AZX_DCAPS_PM_RUNTIME
flag on the HDA controller.

A side effect is that power consumption might be reduced if the GPU is
in use but the HDA controller is not, because the HDA controller is now
allowed to go to D3hot.  Before, it was forced to stay in D0 as long as
the GPU was in use.  (There is no reduction in power consumption on my
Nvidia GK107, but there might be on other chips.)

The code paths for legacy manual power control are adjusted such that
runtime PM is disabled during power off, thereby preventing the PM core
from resuming the HDA controller.

Note that the device link is not only added on vga_switcheroo capable
systems, but for *any* GPU with integrated HDA controller.  The idea is
that the HDA controller streams audio via connectors located on the GPU,
so the GPU needs to be on for the HDA controller to do anything useful.

This commit implicitly fixes an unbalanced runtime PM ref upon unbind of
hda_intel.c:  On ->probe, a runtime PM ref was previously released under
the condition "azx_has_pm_runtime(chip) || hda->use_vga_switcheroo", but
on ->remove a runtime PM ref was only acquired under the first of those
conditions.  Thus, binding and unbinding the driver twice on a
vga_switcheroo capable system caused the runtime PM refcount to drop
below zero.  The issue is resolved because the AZX_DCAPS_PM_RUNTIME flag
is now always set if use_vga_switcheroo is true.

For more information on device links please refer to:
https://www.kernel.org/doc/html/latest/driver-api/device_link.html
Documentation/driver-api/device_link.rst

Cc: Dave Airlie <[email protected]>
Cc: Ben Skeggs <[email protected]>
Cc: Alex Deucher <[email protected]>
Cc: Rafael J. Wysocki <[email protected]>
Acked-by: Bjorn Helgaas <[email protected]>
Reviewed-by: Takashi Iwai <[email protected]>
Reviewed-by: Peter Wu <[email protected]>
Tested-by: Kai Heng Feng <[email protected]> # AMD PowerXpress
Tested-by: Mike Lothian <[email protected]> # AMD PowerXpress
Tested-by: Denis Lisov <[email protected]> # Nvidia Optimus
Tested-by: Peter Wu <[email protected]> # Nvidia Optimus
Tested-by: Lukas Wunner <[email protected]> # MacBook Pro
Signed-off-by: Lukas Wunner <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/51bd38360ff502a8c42b1ebf4405ee1d3f27118d.1520068884.git.lukas@wunner.de

vga_switcheroo: Deduplicate power state tracking

If DRM drivers use runtime PM, they currently notify vga_switcheroo
whenever they ->runtime_suspend or ->runtime_resume to update
vga_switcheroo's internal power state tracking.

That's essentially a duplication of a functionality performed by the
PM core as it already tracks the GPU's power state and vga_switcheroo
can always query it.

Introduce a new internal helper vga_switcheroo_pwr_state() which does
just that if runtime PM is used, or falls back to vga_switcheroo's
internal power state tracking if manual power control is used.
Drop a redundant power state check in set_audio_state() while at it.

This removes one of the two purposes of the notification mechanism
implemented by vga_switcheroo_set_dynamic_switch().  The other one is
power management of the audio device and we'll remove that next.

Cc: Dave Airlie <[email protected]>
Cc: Ben Skeggs <[email protected]>
Cc: Takashi Iwai <[email protected]>
Cc: Alex Deucher <[email protected]>
Cc: Rafael J. Wysocki <[email protected]>
Reviewed-by: Peter Wu <[email protected]>
Tested-by: Kai Heng Feng <[email protected]> # AMD PowerXpress
Tested-by: Mike Lothian <[email protected]> # AMD PowerXpress
Tested-by: Denis Lisov <[email protected]> # Nvidia Optimus
Tested-by: Peter Wu <[email protected]> # Nvidia Optimus
Tested-by: Lukas Wunner <[email protected]> # MacBook Pro
Signed-off-by: Lukas Wunner <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/0aa49d735b988aa04524a8dc339582ace33f0f94.1520068884.git.lukas@wunner.de

vga_switcheroo: Update PCI current_state on power change

When cutting power to a GPU and its integrated HDA controller, their
cached current_state should be updated to D3cold to reflect reality.

We currently rely on the DRM and HDA drivers to do that, however:

- The HDA driver updates the current_state in azx_vs_set_state(), which
  will no longer be called with driver power control once we migrate to
  device links.  (It will still be called with manual power control.)

- If the HDA device is not bound, its current_state remains at D0 even
  though the GPU driver may decide to go to D3cold.

- The DRM drivers update the current_state using pci_set_power_state()
  which can't put the device into a deeper power state than D3hot if the
  GPU is not deemed power-manageable by the platform (even though it
  *is* power-manageable by some nonstandard means, such as a _DSM).

Centralize updating the current_state of the GPU and HDA controller in
vga_switcheroo's ->runtime_suspend hook to overcome these deficiencies.

The GPU and HDA controller are two functions of the same PCI device
(VGA class device on function 0 and audio device on function 1) and
no other PCI devices reside on the same bus since this is a PCIe
point-to-point link, so we can just walk the bus and update the
current_state of all devices.

On ->runtime_resume, the HDA controller is in D0uninitialized state.
Resume to D0active and then let it autosuspend as it sees fit.

Note that vga_switcheroo_init_domain_pm_ops() is not supposed to be
called by hybrid graphics laptops which power down the GPU via its root
port's _PR3 resources and consequently vga_switcheroo_runtime_suspend()
is not used.  On those laptops, the root port is power-manageable by the
platform (instead of by a nonstandard means) and the current_state is
therefore updated by the PCI core through the following call chain:

  pci_set_power_state()
    __pci_complete_power_transition()
      pci_bus_set_current_state()

Resuming to D0active happens through:

  pci_set_power_state()
    __pci_start_power_transition()
      pci_wakeup_bus()

Cc: Dave Airlie <[email protected]>
Cc: Ben Skeggs <[email protected]>
Cc: Takashi Iwai <[email protected]>
Cc: Alex Deucher <[email protected]>
Cc: Bjorn Helgaas <[email protected]>
Cc: Rafael J. Wysocki <[email protected]>
Reviewed-by: Peter Wu <[email protected]>
Tested-by: Kai Heng Feng <[email protected]> # AMD PowerXpress
Tested-by: Mike Lothian <[email protected]> # AMD PowerXpress
Tested-by: Denis Lisov <[email protected]> # Nvidia Optimus
Tested-by: Peter Wu <[email protected]> # Nvidia Optimus
Tested-by: Lukas Wunner <[email protected]> # MacBook Pro
Signed-off-by: Lukas Wunner <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/8416958482c8c42d6f311ea5c1e5a65ccf21f5db.1520068884.git.lukas@wunner.de

PCI: Make pci_wakeup_bus() & pci_bus_set_current_state() public

There are PCI devices which are power-manageable by a nonstandard means,
such as a custom ACPI method.  One example are discrete GPUs in hybrid
graphics laptops, another are Thunderbolt controllers in Macs.

Such devices can't be put into D3cold with pci_set_power_state() because
pci_platform_power_transition() fails with -ENODEV.  Instead they're put
into D3hot by pci_set_power_state() and subsequently into D3cold by
invoking the nonstandard means.  However as a consequence the cached
current_state is incorrectly left at D3hot.

What we need to do is walk the hierarchy below such a PCI device on
powerdown and update the current_state to D3cold.  On powerup the PCI
device itself and the hierarchy below it is in D0uninitialized, so we
need to walk the hierarchy again and wake all devices, causing them to
be put into D0active and then letting them autosuspend as they see fit.

To this end make pci_wakeup_bus() & pci_bus_set_current_state() public
so PCI drivers don't have to reinvent the wheel.

Cc: Rafael J. Wysocki <[email protected]>
Acked-by: Bjorn Helgaas <[email protected]>
Signed-off-by: Lukas Wunner <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/2962443259e7faec577274b4ef8c54aad66f9a94.1520068884.git.lukas@wunner.de

PCI: Restore config space on runtime resume despite being unbound

We leave PCI devices not bound to a driver in D0 during runtime suspend.
But they may have a parent which is bound and can be transitioned to
D3cold at runtime.  Once the parent goes to D3cold, the unbound child
may go to D3cold as well.  When the child goes to D3cold, its internal
state, including configuration of BARs, MSI, ASPM, MPS, etc., is lost.

One example are recent hybrid graphics laptops which cut power to the
discrete GPU when the root port above it goes to ACPI power state D3.
Users may provoke this by unbinding the GPU driver and allowing runtime
PM on the GPU via sysfs:  The PM core will then treat the GPU as
"suspended", which in turn allows the root port to runtime suspend,
causing the power resources listed in its _PR3 object to be powered off.
The GPU's BARs will be uninitialized when a driver later probes it.

Another example are hybrid graphics laptops where the GPU itself (rather
than the root port) is capable of runtime suspending to D3cold.  If the
GPU's integrated HDA controller is not bound and the GPU's driver
decides to runtime suspend to D3cold, the HDA controller's BARs will be
uninitialized when a driver later probes it.

Fix by saving and restoring config space over a runtime suspend cycle
even if the device is not bound.

Acked-by: Bjorn Helgaas <[email protected]>
Tested-by: Peter Wu <[email protected]> # Nvidia Optimus
Tested-by: Lukas Wunner <[email protected]> # MacBook Pro
Signed-off-by: Rafael J. Wysocki <[email protected]>
[lukas: add commit message, bikeshed code comments for clarity]
Signed-off-by: Lukas Wunner <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/92fb6e6ae2730915eb733c08e2f76c6a313e3860.1520068884.git.lukas@wunner.de

drm/meson: Add support for DMT modes on HDMI

This patch adds support for DMT display modes over HDMI.
The modes timings configurations are from the Amlogic Vendor linux tree
and tested over multiples monitors.
Previously only a selected number of CEA modes were supported.

Only these following modes are supported with these changes:
- 640x480@60Hz
- 800x600@60Hz
- 1024x768@60Hz
- 1152x864@75Hz
- 1280x1024@60Hz
- 1600x1200@60Hz
- 1920x1080@60Hz

The associated code to handle the clock rates is also added.

Acked-by: Jerome Brunet <[email protected]>
Signed-off-by: Neil Armstrong <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/meson: Fix some error handling paths in 'meson_drv_bind_master()'

If one of these functions fail, we whould free 'drm', as alreadry done in
the other error handling paths, below and above.

Fixes: bbbe775ec5b5 ("drm: Add support for Amlogic Meson Graphic Controller")
Signed-off-by: Christophe JAILLET <[email protected]>
Acked-by: Neil Armstrong <[email protected]>
Signed-off-by: Neil Armstrong <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/df47e03d36c2cf7bc37ec3105fc47c16555bd946.1520885192.git.christophe.jaillet@wanadoo.fr

drm/meson: Use drm_dev_put() instead of drm_dev_unref()

According to 'drivers/gpu/drm/drm_drv.c', 'drm_dev_unref()' is just a
compatibility alias for 'drm_dev_put()'. So use the latter instead.

Signed-off-by: Christophe JAILLET <[email protected]>
Acked-by: Neil Armstrong <[email protected]>
Signed-off-by: Neil Armstrong <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/d4ce46077d3ada7149c4555acfdda8ade4a6cf50.1520885192.git.christophe.jaillet@wanadoo.fr

drm/meson: Fix an un-handled error path in 'meson_drv_bind_master()'

'drm_vblank_init()' can fail. So handle this (unlikely) error.

Fixes: bbbe775ec5b5 ("drm: Add support for Amlogic Meson Graphic Controller")
Signed-off-by: Christophe JAILLET <[email protected]>
Acked-by: Neil Armstrong <[email protected]>
Signed-off-by: Neil Armstrong <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/6cbf3d70ac3904489c7194c895225c4103aebb96.1520885192.git.christophe.jaillet@wanadoo.fr

drm: Fix some coding style issues

Put an empty line between the variable declarations and the code, and
use tabs for alignment.

Signed-off-by: Ville Syrjälä <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]
Acked-by: Daniel Vetter <[email protected]>

drm/panel: Add support for AUO G104SN02 V2 panel

This patch adds support for AUO G104SN02 V2 800x600 10.4" panel to DRM
simple panel driver.

Signed-off-by: Christoph Fritz <[email protected]>
Signed-off-by: Stefan Riedmueller <[email protected]>
Reviewed-by: Rob Herring <[email protected]>
Signed-off-by: Thierry Reding <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/panel: ili9322: Make local symbols static

Fixes the following sparse warnings:

drivers/gpu/drm/panel/panel-ilitek-ili9322.c:182:12: warning:
 symbol 'ili9322_inputs' was not declared. Should it be static?
drivers/gpu/drm/panel/panel-ilitek-ili9322.c:343:28: warning:
 symbol 'ili9322_regmap_config' was not declared. Should it be static?

Also change ili9322_inputs to 'const char * const' to avoid
chackpatch warning.

Signed-off-by: Wei Yongjun <[email protected]>
Signed-off-by: Thierry Reding <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

dt-bindings: Add vendor prefix for emtrion GmbH

emtrion is a system integrator and manufacturer of embedded systems.

Website: https://www.emtrion.de

Signed-off-by: Jan Tuerk <[email protected]>
Reviewed-by: Andreas Färber <[email protected]>
Acked-by: Rob Herring <[email protected]>
Signed-off-by: Thierry Reding <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/panel: otm8009a: Adopt SPDX identifiers

Add SPDX identifiers to OriseTech OTM8009a panel driver.

Signed-off-by: Philippe Cornu <[email protected]>
Signed-off-by: Thierry Reding <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/panel: lvds: Fix driver description heading

Cc: Laurent Pinchart <[email protected]>
Signed-off-by: Baruch Siach <[email protected]>
Reviewed-by: Laurent Pinchart <[email protected]>
Signed-off-by: Thierry Reding <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/f80a5dd350b057d925d36f5857b5724c88da50df.1516636902.git.baruch@tkos.co.il

drm/panel: simple: Add support for KEO TX31D200VM0BAA

This adds support for the Kaohsiung Opto-Electronics.,
TX31D200VM0BAA 12.3" HSXGA LVDS panel, which can be
supported by the simple panel driver.

Signed-off-by: Jagan Teki <[email protected]>
Reviewed-by: Rob Herring <[email protected]>
Signed-off-by: Thierry Reding <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/panel: otm8009a: Add support for the optional power-supply

Add support for the optional power-supply.

Note: A "dummy regulator" is returned by devm_regulator_get()
if the optional regulator is not present in the device tree,
simplifying the source code when enabling/disabling the regulator.

Signed-off-by: Philippe Cornu <[email protected]>
Reviewed-by: Yannick Fertré <[email protected]>
Signed-off-by: Thierry Reding <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

dt-bindings/display/panel: otm8009a: Add optional power-supply property

Some boards use a dedicated voltage regulator for this panel.
Add & document this related optional power-supply property.

Signed-off-by: Philippe Cornu <[email protected]>
Reviewed-by: Yannick Fertré <[email protected]>
Reviewed-by: Rob Herring <[email protected]>
Signed-off-by: Thierry Reding <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/panel: simple: Use display_timing for lq123p1jx31

Convert the sharp lq123p1jx31 from using a fixed mode to specifying a
display timing with min/typ/max values. This allows us to capture the
timings set forth in the datasheet as well as the additional values that
we've cleared with the display vendor to avoid interference with the
digitizer on the Samsung Chromebook Plus (kevin).

A follow-on patch will specify the override mode for kevin devices.

Changes in v2:
 - None
Changes in v3:
 - None

Cc: Doug Anderson <[email protected]>
Cc: Eric Anholt <[email protected]>
Cc: Heiko Stuebner <[email protected]>
Cc: Jeffy Chen <[email protected]>
Cc: Rob Herring <[email protected]>
Cc: Stéphane Marchesin <[email protected]>
Cc: Thierry Reding <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Signed-off-by: Sean Paul <[email protected]>
Tested-by: Enric Balletbo i Serra <[email protected]>
Signed-off-by: Thierry Reding <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

dt-bindings: Add headings to simple-panel bindings

In preparation for a new subnode section in a follow-on patch, add
explicit headings to the existings sections for simple-panel.

Changes in v2:
 - Added
Changes in v3:
 - None

Cc: Doug Anderson <[email protected]>
Cc: Eric Anholt <[email protected]>
Cc: Heiko Stuebner <[email protected]>
Cc: Jeffy Chen <[email protected]>
Cc: Rob Herring <[email protected]>
Cc: Stéphane Marchesin <[email protected]>
Cc: Thierry Reding <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Signed-off-by: Sean Paul <[email protected]>
Reviewed-by: Rob Herring <[email protected]>
Signed-off-by: Thierry Reding <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

dt-bindings: Clarify timing subnode use as panel-timing

Add a note in the documentation explaining when it's appropriate to use
the display-timings subnode on its own, as well as the preferred name to
use (panel-timing).

Changes in v3:
 - Added

Cc: Doug Anderson <[email protected]>
Cc: Eric Anholt <[email protected]>
Cc: Heiko Stuebner <[email protected]>
Cc: Jeffy Chen <[email protected]>
Cc: Rob Herring <[email protected]>
Cc: Stéphane Marchesin <[email protected]>
Cc: Thierry Reding <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Signed-off-by: Sean Paul <[email protected]>
Reviewed-by: Rob Herring <[email protected]>
Acked-by: Thierry Reding <[email protected]>
Signed-off-by: Thierry Reding <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/panel: Add support for Raydium RM68200 panel driver

This patch adds Raydium Semiconductor Corporation RM68200 5.5" 720x1280
TFT LCD panel driver (MIPI-DSI video mode).

Signed-off-by: Philippe Cornu <[email protected]>
Signed-off-by: Thierry Reding <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

dt-bindings: display: panel: Add support for Raydium RM68200 DSI panel

The Raydium Semiconductor Corporation RM68200 is a 5.5" 720x1280
TFT LCD panel connected using a MIPI-DSI video interface.

Reviewed-by: Rob Herring <[email protected]>
Signed-off-by: Philippe Cornu <[email protected]>
Signed-off-by: Thierry Reding <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/panel: simple: Fix the bus format for the Ontat panel

This fixes bad color output.  When I was first testing the device I
had the DPI hardware set to 666 mode, but apparently in the refactor
to use the bus_format information from the panel driver, I failed to
actually update the panel.

Signed-off-by: Eric Anholt <[email protected]>
Fixes: e8b6f561b2ee ("drm/panel: simple: Add the 7" DPI panel from Adafruit")
Cc: Thierry Reding <[email protected]>
Signed-off-by: Thierry Reding <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/vc4: Enable background color fill when necessary

Using the hint from the plane state, we turn on the background color
to avoid display corruption from planes blending with the background.

Changes from v1:
 - Use needs_bg_fill from plane state

Signed-off-by: Stefan Schake <[email protected]>
Signed-off-by: Eric Anholt <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/vc4: Move plane state to header

We need to reference it from the CRTC to make a decision for enabling
background color fill.

Signed-off-by: Stefan Schake <[email protected]>
Signed-off-by: Eric Anholt <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/vc4: Check if plane requires background fill

Considering a single plane only, we have to enable background color
when the plane has an alpha format and could be blending from the
background or when it doesn't cover the entire screen.

Changes from v1:
 - Drop unrelated change
 - Move needs_bg_fill to plane state

Signed-off-by: Stefan Schake <[email protected]>
Signed-off-by: Eric Anholt <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/vc4: Set premultiplied for alpha formats

Alpha formats in DRM are assumed to be premultiplied, so we should be
setting the PREMULT bit in the plane configuration for HVS.

Changes from v1:
 - Use correct has_alpha

Signed-off-by: Stefan Schake <[email protected]>
Signed-off-by: Eric Anholt <[email protected]>
Reviewed-by: Eric Anholt <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/rockchip: Don't use atomic constructs for psr

Instead of using timer and spinlocks, use delayed_work and
mutexes for rockchip psr. This allows us to make blocking
calls when enabling/disabling psr (which is sort of important
given we're talking over dpcd to the display).

Cc: Caesar Wang <[email protected]>
Cc: 征增 王 <[email protected]>
Cc: Stéphane Marchesin <[email protected]>
Signed-off-by: Sean Paul <[email protected]>
Signed-off-by: Thierry Escande <[email protected]>
Signed-off-by: Enric Balletbo i Serra <[email protected]>
Signed-off-by: Heiko Stuebner <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/rockchip: analogix_dp: set psr activate/deactivate when enable/disable bridge

There's a race between when bridge_disable and when vop_crtc_disable
are called. If the flush timer triggers a new psr work between these,
we will operate eDP without power shutdowned by bridge_disable. In this
case, moving activate/deactivate to enable/disable bridge to avoid it.

Cc: Stéphane Marchesin <[email protected]>
Signed-off-by: zain wang <[email protected]>
Signed-off-by: Caesar Wang <[email protected]>
Signed-off-by: Sean Paul <[email protected]>
Signed-off-by: Thierry Escande <[email protected]>
Signed-off-by: Enric Balletbo i Serra <[email protected]>
Signed-off-by: Heiko Stuebner <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/rockchip: dw_hdmi: Move HDMI vpll clock enable to bind()

The HDMI vpll clock should be enabled when bind() is called. So move the
clk_prepare_enable of that clock to bind() function and add the missing
clk_disable_unprepare() required in error handling path and unbind().

Fixes: 12b9f204e804 ("drm: bridge/dw_hdmi: add rockchip rk3288 support")
Signed-off-by: Jeffy Chen <[email protected]>
Signed-off-by: Thierry Escande <[email protected]>
Signed-off-by: Enric Balletbo i Serra <[email protected]>
Signed-off-by: Heiko Stuebner <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/rockchip: inno_hdmi: reorder clk_disable_unprepare call in unbind

In bind the clk_prepare_enable of the HDMI pclk is called before adding the
i2c_adapter. So it should be the other way around in unbind, first remove
the i2c_adapter and then call the clk_disable_unprepare.

Fixes: 412d4ae6b7a5 ("drm/rockchip: hdmi: add Innosilicon HDMI support")
Signed-off-by: Jeffy Chen <[email protected]>
Signed-off-by: Thierry Escande <[email protected]>
Signed-off-by: Enric Balletbo i Serra <[email protected]>
Signed-off-by: Heiko Stuebner <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/rockchip: inno_hdmi: Fix error handling path.

Add missing error handling in bind().

Fixes: 412d4ae6b7a5 ("drm/rockchip: hdmi: add Innosilicon HDMI support")
Signed-off-by: Jeffy Chen <[email protected]>
Signed-off-by: Thierry Escande <[email protected]>
[moved clk_disable_unprepare reordering in unbind to separate patch]
Signed-off-by: Enric Balletbo i Serra <[email protected]>
Signed-off-by: Heiko Stuebner <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/rockchip: dw-mipi-dsi: Fix connector and encoder cleanup.

In bind()'s error handling path call destroy functions instead of
cleanup functions for encoder and connector and reorder to match how is
called in bind().

In unbind() call the connector and encoder destroy functions.

Signed-off-by: Jeffy Chen <[email protected]>
Signed-off-by: Thierry Escande <[email protected]>
Signed-off-by: Enric Balletbo i Serra <[email protected]>
Signed-off-by: Heiko Stuebner <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/nouveau: Replace the iturbt_709 prop with the standard COLOR_ENCODING prop

Replace the ad-hoc iturbt_709 property with the new standard
COLOR_ENCODING property. Compiles, but not tested.

v2: Fix typos (Ilia)

Cc: Daniel Vetter <[email protected]>
Cc: [email protected]
Cc: Ben Skeggs <[email protected]>
Cc: Ilia Mirkin <[email protected]>
Signed-off-by: Ville Syrjälä <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]
Reviewed-by: Ilia Mirkin <[email protected]>
Acked-by: Ben Skeggs <[email protected]> #irc

drm/pl111: Use max memory bandwidth for resolution

We were previously selecting 1024x768 and 32BPP as the default
set-up for the PL111 consumers.

This does not work on elder systems: the device tree bindings
support a property "max-memory-bandwidth" in bytes/second that
states that if you exceed this the memory bus will saturate.
The result is flickering and unstable images.

Parse the "max-memory-bandwidth" and respect it when
intializing the driver. On the RealView PB11MP, Versatile and
Integrator/CP we get a nice console as default with this code.

Reviewed-by: Eric Anholt <[email protected]>
Signed-off-by: Linus Walleij <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/bridge: sii902x: Retry status read after DDI I2C

The following happens when connection a DVI output driven
from the SiI9022 using a DVI-to-VGA adapter plug:

i2c i2c-0: sendbytes: NAK bailout.
i2c i2c-0: sendbytes: NAK bailout.

Then no picture. Apparently the I2C engine inside the SiI9022
is not smart enough to try to fall back to DDC I2C. Or the
vendor have not integrated the electronics properly. I don't
know which one it is.

After this, the I2C bus seems stalled and the first attempt to
read the status register fails, and the code returns with
negative return value, and the display fails to initialized.

Instead, retry status readout five times and continue even
if this fails.

Tested on the ARM Versatile Express with a DVI-to-VGA
connector, it now gives picture.

Introduce a helper struct device *dev variable to make
the code more readable.

Cc: Ville Syrjälä <[email protected]>
Reviewed-by: Liviu Dudau <[email protected]>
Signed-off-by: Linus Walleij <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/pl111: Handle the RealView variant separately

We want to cut down the default bpp to 16 on the RealView so
we can have a 1024x768 framebuffer console by default. The
memory bandwidth limitations makes this not work with the
PL111 default of 32bpp.

This builds on top of the earlier patches making the
framebuffer default bpp a per-variant variable.

Reviewed-by: Eric Anholt <[email protected]>
Signed-off-by: Linus Walleij <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/pl111: Make the default BPP a per-variant variable

The PL110, Integrator and Versatile boards strongly prefer to
use 16 BPP even if other modes are supported, both to keep down
memory consumption and also to easier find a good match to
supported resolutions with consideration taken to the memory
bandwidth of the platforms.

Reviewed-by: Eric Anholt <[email protected]>
Signed-off-by: Linus Walleij <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm: simple_kms_helper: Fix .mode_valid() documentation

This fixes up the .mode_valid() vtable entry documentation
by copyediting the documentation from the .mode_valid()
documentation in the drm_modeset_helper_vtables.h file.

Fixes: 40275dc4edb4 ("drm: simple_kms_helper: Add mode_valid() callback support")
Suggested-by: Daniel Vetter <[email protected]>
Reviewed-by: Daniel Vetter <[email protected]>
Signed-off-by: Linus Walleij <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

bridge: Elaborate a bit on dumb VGA bridges in Kconfig

It's better if we explain a bit that this pertains to
non-programmable VGA DAC bridges.

Reviewed-by: Laurent Pinchart <[email protected]>
Signed-off-by: Linus Walleij <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/atomic: Add new reverse iterator over all plane state (V2)

Add reverse iterator for_each_oldnew_plane_in_state_reverse to
compliment the for_each_oldnew_plane_in_state way or reading plane
states.

The plane states are required to be read in reverse order for
amd drivers, cause the z order convention followed in linux is
opposite to how the planes are supposed to be presented to DC
engine, which is in common to both windows and linux.

V2: fix compile time errors due to -Werror flag.

Signed-off-by: Shirish S <[email protected]>
Signed-off-by: Pratik Vishwakarma <[email protected]>
Reviewed-by: Daniel Vetter <[email protected]>
Signed-off-by: Harry Wentland <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm: Reject bad property flag combinations

Pimp drm_property_type_valid() to check for more fails with the
property flags. Also make the check before adding the property,
and bail out if things look bad.

Since we're now chekcing for more than the type let's also
change the function name to drm_property_flags_valid().

Signed-off-by: Ville Syrjälä <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]
Reviewed-by: Daniel Vetter <[email protected]>

drm: Make property flags u32

The property flags are part of the uabi and we have 32 bits for them.
Pass them around as u32 internally as well, instead of a signed int.

Signed-off-by: Ville Syrjälä <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]
Reviewed-by: Daniel Vetter <[email protected]>

drm/uapi: Deprecate DRM_MODE_PROP_PENDING

DRM_MODE_PROP_PENDING is not used anywhere (except printed out
by libdrm proptest/modetest).

This seems to be yet another thing blindly copied from xrandr.
Quoting from the protocol spec:
"If 'pending' is TRUE, changes made to property values with
 RRChangeOutputProperty will be saved in the pending property value
 and be automatically copied to the current value on the next
 RRSetCrtcConfig request involving the named output. If 'pending' is
 FALSE, changes are copied immediately."

So it was some kind of early idea for atomic property updates.

Signed-off-by: Ville Syrjälä <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]
Reviewed-by: Daniel Vetter <[email protected]>

drm: WARN when trying to add enum value > 63 to a bitmask property

Enum values >63 with a bitmask property is a programmer error. WARN
when someone is attempting this.

Signed-off-by: Ville Syrjälä <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]
Reviewed-by: Daniel Vetter <[email protected]>

drm: WARN when trying add enum values to non-enum/bitmask properties

Trying to add enum values to non-enum/bitmask properties is a
programmer mistake. WARN to make sure the developers notice
their mistake.

Signed-off-by: Ville Syrjälä <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]
Reviewed-by: Daniel Vetter <[email protected]>

drm: Reject replacing property enum values

If the property already has the enum value WARN and bail.
Replacing enum values doesn't make sense to me.

Throw out the pointless list_empty() while at it.

Cc: Daniel Vetter <[email protected]>
Suggested-by: Daniel Vetter <[email protected]>
Signed-off-by: Ville Syrjälä <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]
Reviewed-by: Daniel Vetter <[email protected]>

drm/bridge: analogix_dp: Don't create useless connectors

If there is another bridge after analogix_dp, then the connector object
should not be created. This fixes following timeouts on Exynos5420-based
Chromebook2 Peach-PIT board during boot:

exynos-dp 145b0000.dp-controller: AUX CH cmd reply timeout!
exynos-dp 145b0000.dp-controller: AUX CH enable timeout!
exynos-dp 145b0000.dp-controller: AUX CH enable timeout!
exynos-dp 145b0000.dp-controller: AUX CH enable timeout!
exynos-dp 145b0000.dp-controller: AUX CH enable timeout!

Signed-off-by: Marek Szyprowski <[email protected]>
Tested-by: Heiko Stuebner <[email protected]>
Signed-off-by: Andrzej Hajda <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/bridge: analogix_dp: Postpone enabling runtime power management

Enabling runtime power management early in analogix_dp_bind() causes following
kernel NULL pointer dereference:

Unable to handle kernel NULL pointer dereference at virtual address 000007d8
pgd = 28ffa2e4
[000007d8] *pgd=00000000
Internal error: Oops: 5 [#1] PREEMPT SMP ARM
Modules linked in:
CPU: 6 PID: 69 Comm: kworker/6:1 Not tainted 4.16.0-rc1-00062-ge25751974ba8 #3622
Hardware name: SAMSUNG EXYNOS (Flattened Device Tree)
Workqueue: events deferred_probe_work_func
PC is at analogix_dp_resume+0x8/0xc0
LR is at pm_generic_runtime_resume+0x2c/0x38
pc : [<c0531b98>]    lr : [<c0543fec>] psr: a0000113
sp : ee13fbd8  ip : 0000001a  fp : 00000001
r10: ee0eb080  r9 : c0552bd8  r8 : c0fb1d98
r7 : eebb1010  r6 : eeae9808  r5 : 00000000  r4 : d4850415
r3 : ee0ed010  r2 : b2d05e00  r1 : 00000000  r0 : 00000000
Flags: NzCv  IRQs on  FIQs on  Mode SVC_32  ISA ARM Segment none
Control: 10c5387d  Table: 2000406a  DAC: 00000051
Process kworker/6:1 (pid: 69, stack limit = 0x913205b4)
Stack: (0xee13fbd8 to 0xee140000)
...
[<c0531b98>] (analogix_dp_resume) from [<c0543fec>] (pm_generic_runtime_resume+0x2c/0x38)
[<c0543fec>] (pm_generic_runtime_resume) from [<c054ffb4>] (__genpd_runtime_resume+0x2c/0x8c)
[<c054ffb4>] (__genpd_runtime_resume) from [<c0552d24>] (genpd_runtime_resume+0x14c/0x258)
[<c0552d24>] (genpd_runtime_resume) from [<c0547798>] (__rpm_callback+0x134/0x214)
[<c0547798>] (__rpm_callback) from [<c0547898>] (rpm_callback+0x20/0x80)
[<c0547898>] (rpm_callback) from [<c0546ff4>] (rpm_resume+0x3a0/0x734)
[<c0546ff4>] (rpm_resume) from [<c05475ec>] (__pm_runtime_resume+0x64/0x9c)
[<c05475ec>] (__pm_runtime_resume) from [<c053b95c>] (__device_attach+0x8c/0x134)
[<c053b95c>] (__device_attach) from [<c053ad08>] (bus_probe_device+0x88/0x90)
[<c053ad08>] (bus_probe_device) from [<c05390d0>] (device_add+0x3a8/0x580)
[<c05390d0>] (device_add) from [<c06764c4>] (i2c_register_adapter+0xd4/0x3ec)
[<c06764c4>] (i2c_register_adapter) from [<c05321c8>] (analogix_dp_bind+0x2a0/0x410)
[<c05321c8>] (analogix_dp_bind) from [<c0528e90>] (exynos_dp_bind+0x9c/0x12c)
[<c0528e90>] (exynos_dp_bind) from [<c0535bc4>] (component_bind_all+0xfc/0x258)
[<c0535bc4>] (component_bind_all) from [<c0522ee8>] (exynos_drm_bind+0x15c/0x28c)
[<c0522ee8>] (exynos_drm_bind) from [<c0536378>] (try_to_bring_up_master+0x1b8/0x29c)
[<c0536378>] (try_to_bring_up_master) from [<c05364fc>] (component_add+0xa0/0x170)
[<c05364fc>] (component_add) from [<c0528fe4>] (exynos_dp_probe+0x64/0xb8)
[<c0528fe4>] (exynos_dp_probe) from [<c053debc>] (platform_drv_probe+0x50/0xb0)
[<c053debc>] (platform_drv_probe) from [<c053bd18>] (driver_probe_device+0x2b8/0x4a0)
[<c053bd18>] (driver_probe_device) from [<c0539e4c>] (bus_for_each_drv+0x44/0x8c)
[<c0539e4c>] (bus_for_each_drv) from [<c053b970>] (__device_attach+0xa0/0x134)
[<c053b970>] (__device_attach) from [<c053ad08>] (bus_probe_device+0x88/0x90)
[<c053ad08>] (bus_probe_device) from [<c053b258>] (deferred_probe_work_func+0x3c/0x168)
[<c053b258>] (deferred_probe_work_func) from [<c014352c>] (process_one_work+0x1d0/0x7bc)
[<c014352c>] (process_one_work) from [<c0143b84>] (worker_thread+0x34/0x4dc)
[<c0143b84>] (worker_thread) from [<c014a30c>] (kthread+0x128/0x164)
[<c014a30c>] (kthread) from [<c01010b4>] (ret_from_fork+0x14/0x20)
Exception stack(0xee13ffb0 to 0xee13fff8)
ffa0:                                     00000000 00000000 00000000 00000000
ffc0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
ffe0: 00000000 00000000 00000000 00000000 00000013 00000000
Code: e2800e37 eafee601 e92d4070 e1a05000 (e59067d8)
---[ end trace bf6046013df7cab2 ]---

This oops happens, because analogix_dp_bind() calls drm_dp_aux_register()
which registers i2c adapter. I2C core tries to runtime get i2c host
device during registration. This ends in analogix_dp_resume(), but dp
context is NULL there. dp context is set in exynos_dp_bind() after
executing analogix_dp_bind(). Fix this issue by postponing enabling runtime
power management after drm_dp_aux_register().

Signed-off-by: Marek Szyprowski <[email protected]>
Tested-by: Heiko Stuebner <[email protected]>
Reviewed-by: Heiko Stuebner <[email protected]>
Signed-off-by: Andrzej Hajda <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/bridge: analogix_dp: Don't power bridge in analogix_dp_bind

The bridge does not need to be powered in analogix_dp_bind(), so
remove the calls to pm_runtime_get()/phy_power_on()/analogix_dp_init_dp()
as well as their power-off counterparts.

Cc: Stéphane Marchesin <[email protected]>
Signed-off-by: zain wang <[email protected]>
Signed-off-by: Caesar Wang <[email protected]>
[the patch originally just removed the power_on portion, seanpaul removed
the power off code as well as improved the commit message]
Signed-off-by: Sean Paul <[email protected]>
Signed-off-by: Thierry Escande <[email protected]>
Signed-off-by: Marek Szyprowski <[email protected]>
Tested-by: Heiko Stuebner <[email protected]>
Reviewed-by: Heiko Stuebner <[email protected]>
Signed-off-by: Andrzej Hajda <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/prime: make the pages array optional for drm_prime_sg_to_page_addr_arrays

Most of the time we only need the dma addresses.

Signed-off-by: Christian König <[email protected]>
Reviewed-by: Roger He <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]
Link: https://patchwork.freedesktop.org/patch/msgid/BN6PR12MB18262C0DE9B5F07B9A42EAE7F2C60@BN6PR12MB1826.namprd12.prod.outlook.com

drm/prime: fix potential race in drm_gem_map_detach

Unpin the GEM object only after freeing the sg table.

Signed-off-by: Christian König <[email protected]>
Reviewed-by: Daniel Vetter <[email protected]>
Acked-by: Roger He <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm: Check property/enum name length

Reject requests to add properties/enums with an overly long name.
Previously we would have just silently truncated the string and exposed
it userspace.

v2: drm_property_create() returns a pointer

Signed-off-by: Ville Syrjälä <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]
Reviewed-by: Daniel Vetter <[email protected]>

drm: Don't create properties without names

Creating a property that doesn't have a name makes no sense to me. Don't
allow it.

Signed-off-by: Ville Syrjälä <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]
Reviewed-by: Daniel Vetter <[email protected]>

drm: bridge: dw-hdmi: Fix overflow workaround for Amlogic Meson GX SoCs

The Amlogic Meson GX SoCs, embedded the v2.01a controller, has been also
identified needing this workaround.
This patch adds the corresponding version to enable a single iteration for
this specific version.

Fixes: be41fc55f1aa ("drm: bridge: dw-hdmi: Handle overflow workaround based on device version")
Acked-by: Archit Taneja <[email protected]>
[narmstrong: s/identifies/identified and rebased against Jernej's change]
Signed-off-by: Neil Armstrong <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm: of: simplify component probe code

Use positive logic for better readability. This also eliminates one
of_node_put() call, making the code shorter.

Signed-off-by: Baruch Siach <[email protected]>
Reviewed-by: Philipp Zabel <[email protected]>
Tested-by: Philipp Zabel <[email protected]>
Signed-off-by: Archit Taneja <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/61c56895d44117d80e7d82f04e729e29c60fadbd.1519327370.git.baruch@tkos.co.il

drm: NULL pointer dereference [null-pointer-deref] (CWE 476) problem

The Parfait (version 2.1.0) static code analysis tool found the
following NULL pointer dereference problem.

- drivers/gpu/drm/drm_drv.c
Any calls to drm_minor_get_slot() could result in the return of a NULL
pointer when an invalid DRM device type is encountered.  The
return of NULL was removed with BUG() from drm_minor_get_slot().

Signed-off-by: Joe Moriarty <[email protected]>
Reviewed-by: Steven Sistare <[email protected]>
Signed-off-by: Daniel Vetter <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm: NULL pointer dereference [null-pointer-deref] (CWE 476) problem

The Parfait (version 2.1.0) static code analysis tool found the
following NULL pointer derefernce problem.

- drivers/gpu/drm/drm_vblank.c
Null pointer checks were added to return values from calls to
drm_crtc_from_index().  There is a possibility, however minute, that
crtc->index may not be found when trying to find the struct crtc
from it's assigned index given in drm_crtc_init_with_planes().
3 return checks for NULL where added with a call to
WARN_ON(!crtc).

Signed-off-by: Joe Moriarty <[email protected]>
Reviewed-by: Steven Sistare <[email protected]>
Signed-off-by: Daniel Vetter <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm: Don't pass clip to drm_atomic_helper_check_plane_state()

Move the plane clip rectangle handling into
drm_atomic_helper_check_plane_state(). Drivers no longer
have to worry about such mundane details.

v2: Convert armada, rcar, and sun4i as well
v3: Resolve simple_kms_helper conflict

Cc: Liviu Dudau <[email protected]>
Cc: Brian Starkey <[email protected]>
Cc: Mali DP Maintainers <[email protected]>
Cc: Daniel Vetter <[email protected]>
Cc: Gustavo Padovan <[email protected]>
Cc: Sean Paul <[email protected]>
Cc: Philipp Zabel <[email protected]>
Cc: CK Hu <[email protected]>
Cc: Neil Armstrong <[email protected]>
Cc: Rob Clark <[email protected]>
Cc: Ben Skeggs <[email protected]>
Cc: Laurent Pinchart <[email protected]>
Cc: Sandy Huang <[email protected]>
Cc: "Heiko Stübner" <[email protected]>
Cc: Maxime Ripard <[email protected]>
Cc: Thierry Reding <[email protected]>
Cc: VMware Graphics <[email protected]>
Cc: Sinclair Yeh <[email protected]>
Cc: Thomas Hellstrom <[email protected]>
Cc: Shawn Guo <[email protected]>
Cc: Archit Taneja <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: Russell King <[email protected]>
Suggested-by: Daniel Vetter <[email protected]>
Signed-off-by: Ville Syrjälä <[email protected]>
Reviewed-by: Daniel Vetter <[email protected]>
Reviewed-by: Thierry Reding <[email protected]>
Reviewed-by: Archit Taneja <[email protected]> #msm
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]
Acked-by: Liviu Dudau <[email protected]> #hdlcd,malidp
Acked-by: Philipp Zabel <[email protected]> #imx,mtk
Reviewed-by: Laurent Pinchart <[email protected]>
Reviewed-by: Sinclair Yeh <[email protected]> #vmwgfx
Acked-by: Neil Armstrong <[email protected]> #meson
Acked-by: Shawn Guo <[email protected]> #zte