drm/msm/a6xx: Skip snapshotting unused GMU buffers

Some buffers are unused on certain sub-generations of a6xx.  So just
skip them.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Reviewed-by: Chia-I Wu <olvaffe@gmail.com>
Reviewed-by: Akhil P Oommen <quic_akhilpo@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/507013/
Link: https://lore.kernel.org/r/20221013225520.371226-3-robdclark@gmail.com

drm/msm/a6xx: Fix kvzalloc vs state_kcalloc usage

adreno_show_object() is a trap!  It will re-allocate the pointer it is
passed on first call, when the data is ascii85 encoded, using kvmalloc/
kvfree().  Which means the data *passed* to it must be kvmalloc'd, ie.
we cannot use the state_kcalloc() helper.

This partially reverts commit ec8f1813bf8d ("drm/msm/a6xx: Replace
kcalloc() with kvzalloc()"), but adds the missing kvfree() to fix the
memory leak that was present previously.  And adds a warning comment.

Fixes: ec8f1813bf8d ("drm/msm/a6xx: Replace kcalloc() with kvzalloc()")
Closes: https://gitlab.freedesktop.org/drm/msm/-/issues/20
Signed-off-by: Rob Clark <robdclark@chromium.org>
Reviewed-by: Chia-I Wu <olvaffe@gmail.com>
Reviewed-by: Akhil P Oommen <quic_akhilpo@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/507014/
Link: https://lore.kernel.org/r/20221013225520.371226-2-robdclark@gmail.com

drm/msm/dp: cleared DP_DOWNSPREAD_CTRL register before start link training

DOWNSPREAD_CTRL (0x107) shall be cleared to 0 upon power-on reset or an
upstream device disconnect. This patch will enforce this rule by always
cleared DOWNSPREAD_CTRL register to 0 before start link training. At rare
case that DP MSA timing parameters may be mis-interpreted by the sink
which causes audio sampling rate be calculated wrongly and cause audio
did not work at sink if DOWNSPREAD_CTRL register is not cleared to 0.

Changes in v2:
1) fix spelling at commit text
2) merge ssc variable into encoding[0]

Changes in v3:
-- correct spelling of DOWNSPREAD_CTRL
-- replace err with len of ssize_t

Changes in v4:
-- split into 2 patches

Signed-off-by: Kuogee Hsieh <quic_khsieh@quicinc.com>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Fixes: c943b4948b58 ("drm/msm/dp: add displayPort driver support")
Patchwork: https://patchwork.freedesktop.org/patch/502532/
Link: https://lore.kernel.org/r/1662999830-13916-2-git-send-email-quic_khsieh@quicinc.com
Signed-off-by: Abhinav Kumar <quic_abhinavk@quicinc.com>

drm/msm/hdmi: fix IRQ lifetime

Device-managed resources allocated post component bind must be tied to
the lifetime of the aggregate DRM device or they will not necessarily be
released when binding of the aggregate device is deferred.

This is specifically true for the HDMI IRQ, which will otherwise remain
requested so that the next bind attempt fails when requesting the IRQ a
second time.

Fix this by tying the device-managed lifetime of the HDMI IRQ to the DRM
device so that it is released when bind fails.

Fixes: 067fef372c73 ("drm/msm/hdmi: refactor bind/init")
Cc: stable@vger.kernel.org # 3.19
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Johan Hovold <johan+linaro@kernel.org>
Tested-by: Kuogee Hsieh <quic_khsieh@quicinc.com>
Reviewed-by: Kuogee Hsieh <quic_khsieh@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/502666/
Link: https://lore.kernel.org/r/20220913085320.8577-9-johan+linaro@kernel.org
Signed-off-by: Abhinav Kumar <quic_abhinavk@quicinc.com>

drm/msm/dp: fix bridge lifetime

Device-managed resources allocated post component bind must be tied to
the lifetime of the aggregate DRM device or they will not necessarily be
released when binding of the aggregate device is deferred.

This can lead resource leaks or failure to bind the aggregate device
when binding is later retried and a second attempt to allocate the
resources is made.

For the DP bridges, previously allocated bridges will leak on probe
deferral.

Fix this by amending the DP parser interface and tying the lifetime of
the bridge device to the DRM device rather than DP platform device.

Fixes: c3bf8e21b38a ("drm/msm/dp: Add eDP support via aux_bus")
Cc: stable@vger.kernel.org # 5.19
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Johan Hovold <johan+linaro@kernel.org>
Tested-by: Kuogee Hsieh <quic_khsieh@quicinc.com>
Reviewed-by: Kuogee Hsieh <quic_khsieh@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/502667/
Link: https://lore.kernel.org/r/20220913085320.8577-8-johan+linaro@kernel.org
Signed-off-by: Abhinav Kumar <quic_abhinavk@quicinc.com>

drm/msm/dp: fix aux-bus EP lifetime

Device-managed resources allocated post component bind must be tied to
the lifetime of the aggregate DRM device or they will not necessarily be
released when binding of the aggregate device is deferred.

This can lead resource leaks or failure to bind the aggregate device
when binding is later retried and a second attempt to allocate the
resources is made.

For the DP aux-bus, an attempt to populate the bus a second time will
simply fail ("DP AUX EP device already populated").

Fix this by tying the lifetime of the EP device to the DRM device rather
than DP controller platform device.

Fixes: c3bf8e21b38a ("drm/msm/dp: Add eDP support via aux_bus")
Cc: stable@vger.kernel.org # 5.19
Signed-off-by: Johan Hovold <johan+linaro@kernel.org>
Reviewed-by: Douglas Anderson <dianders@chromium.org>
Tested-by: Kuogee Hsieh <quic_khsieh@quicinc.com>
Reviewed-by: Kuogee Hsieh <quic_khsieh@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/502672/
Link: https://lore.kernel.org/r/20220913085320.8577-7-johan+linaro@kernel.org
Signed-off-by: Abhinav Kumar <quic_abhinavk@quicinc.com>

drm/msm/dp: fix IRQ lifetime

Device-managed resources allocated post component bind must be tied to
the lifetime of the aggregate DRM device or they will not necessarily be
released when binding of the aggregate device is deferred.

This is specifically true for the DP IRQ, which will otherwise remain
requested so that the next bind attempt fails when requesting the IRQ a
second time.

Since commit c3bf8e21b38a ("drm/msm/dp: Add eDP support via aux_bus")
this can happen when the aux-bus panel driver has not yet been loaded so
that probe is deferred.

Fix this by tying the device-managed lifetime of the DP IRQ to the DRM
device so that it is released when bind fails.

Fixes: c943b4948b58 ("drm/msm/dp: add displayPort driver support")
Cc: stable@vger.kernel.org # 5.10
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Johan Hovold <johan+linaro@kernel.org>
Tested-by: Kuogee Hsieh <quic_khsieh@quicinc.com>
Reviewed-by: Kuogee Hsieh <quic_khsieh@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/502679/
Link: https://lore.kernel.org/r/20220913085320.8577-6-johan+linaro@kernel.org
Signed-off-by: Abhinav Kumar <quic_abhinavk@quicinc.com>

drm/msm/hdmi: fix memory corruption with too many bridges

Add the missing sanity check on the bridge counter to avoid corrupting
data beyond the fixed-sized bridge array in case there are ever more
than eight bridges.

Fixes: a3376e3ec81c ("drm/msm: convert to drm_bridge")
Cc: stable@vger.kernel.org # 3.12
Signed-off-by: Johan Hovold <johan+linaro@kernel.org>
Tested-by: Kuogee Hsieh <quic_khsieh@quicinc.com>
Reviewed-by: Kuogee Hsieh <quic_khsieh@quicinc.com>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/502670/
Link: https://lore.kernel.org/r/20220913085320.8577-5-johan+linaro@kernel.org
Signed-off-by: Abhinav Kumar <quic_abhinavk@quicinc.com>

drm/msm/dsi: fix memory corruption with too many bridges

Add the missing sanity check on the bridge counter to avoid corrupting
data beyond the fixed-sized bridge array in case there are ever more
than eight bridges.

Fixes: a689554ba6ed ("drm/msm: Initial add DSI connector support")
Cc: stable@vger.kernel.org # 4.1
Signed-off-by: Johan Hovold <johan+linaro@kernel.org>
Tested-by: Kuogee Hsieh <quic_khsieh@quicinc.com>
Reviewed-by: Kuogee Hsieh <quic_khsieh@quicinc.com>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/502668/
Link: https://lore.kernel.org/r/20220913085320.8577-4-johan+linaro@kernel.org
Signed-off-by: Abhinav Kumar <quic_abhinavk@quicinc.com>

drm/msm/dp: fix memory corruption with too many bridges

Add the missing sanity check on the bridge counter to avoid corrupting
data beyond the fixed-sized bridge array in case there are ever more
than eight bridges.

Fixes: 8a3b4c17f863 ("drm/msm/dp: employ bridge mechanism for display enable and disable")
Cc: stable@vger.kernel.org # 5.17
Signed-off-by: Johan Hovold <johan+linaro@kernel.org>
Tested-by: Kuogee Hsieh <quic_khsieh@quicinc.com>
Reviewed-by: Kuogee Hsieh <quic_khsieh@quicinc.com>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/502664/
Link: https://lore.kernel.org/r/20220913085320.8577-3-johan+linaro@kernel.org
Signed-off-by: Abhinav Kumar <quic_abhinavk@quicinc.com>

drm/msm: fix use-after-free on probe deferral

The bridge counter was never reset when tearing down the DRM device so
that stale pointers to deallocated structures would be accessed on the
next tear down (e.g. after a second late bind deferral).

Given enough bridges and a few probe deferrals this could currently also
lead to data beyond the bridge array being corrupted.

Fixes: d28ea556267c ("drm/msm: properly add and remove internal bridges")
Fixes: a3376e3ec81c ("drm/msm: convert to drm_bridge")
Cc: stable@vger.kernel.org # 3.12
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Johan Hovold <johan+linaro@kernel.org>
Tested-by: Kuogee Hsieh <quic_khsieh@quicinc.com>
Reviewed-by: Kuogee Hsieh <quic_khsieh@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/502665/
Link: https://lore.kernel.org/r/20220913085320.8577-2-johan+linaro@kernel.org
Signed-off-by: Abhinav Kumar <quic_abhinavk@quicinc.com>

drm/msm: Fix return type of mdp4_lvds_connector_mode_valid

The mode_valid field in drm_connector_helper_funcs is expected to be of
type:
enum drm_mode_status (* mode_valid) (struct drm_connector *connector,
                                     struct drm_display_mode *mode);

The mismatched return type breaks forward edge kCFI since the underlying
function definition does not match the function hook definition.

The return type of mdp4_lvds_connector_mode_valid should be changed from
int to enum drm_mode_status.

Reported-by: Dan Carpenter <error27@gmail.com>
Link: https://github.com/ClangBuiltLinux/linux/issues/1703
Cc: llvm@lists.linux.dev
Signed-off-by: Nathan Huckleberry <nhuck@google.com>
Fixes: 3e87599b68e7 ("drm/msm/mdp4: add LVDS panel support")
Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Patchwork: https://patchwork.freedesktop.org/patch/502878/
Link: https://lore.kernel.org/r/20220913205551.155128-1-nhuck@google.com
Signed-off-by: Abhinav Kumar <quic_abhinavk@quicinc.com>

drm/msm/dp: add atomic_check to bridge ops

DRM commit_tails() will disable downstream crtc/encoder/bridge if
both disable crtc is required and crtc->active is set before pushing
a new frame downstream.

There is a rare case that user space display manager issue an extra
screen update immediately followed by close DRM device while down
stream display interface is disabled. This extra screen update will
timeout due to the downstream interface is disabled but will cause
crtc->active be set. Hence the followed commit_tails() called by
drm_release() will pass the disable downstream crtc/encoder/bridge
conditions checking even downstream interface is disabled.
This cause the crash to happen at dp_bridge_disable() due to it trying
to access the main link register to push the idle pattern out while main
link clocks is disabled.

This patch adds atomic_check to prevent the extra frame will not
be pushed down if display interface is down so that crtc->active
will not be set neither. This will fail the conditions checking
of disabling down stream crtc/encoder/bridge which prevent
drm_release() from calling dp_bridge_disable() so that crash
at dp_bridge_disable() prevented.

There is no protection in the DRM framework to check if the display
pipeline has been already disabled before trying again. The only
check is the crtc_state->active but this is controlled by usermode
using UAPI. Hence if the usermode sets this and then crashes, the
driver needs to protect against double disable.

SError Interrupt on CPU7, code 0x00000000be000411 -- SError
CPU: 7 PID: 3878 Comm: Xorg Not tainted 5.19.0-stb-cbq #19
Hardware name: Google Lazor (rev3 - 8) (DT)
pstate: a04000c9 (NzCv daIF +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
pc : __cmpxchg_case_acq_32+0x14/0x2c
lr : do_raw_spin_lock+0xa4/0xdc
sp : ffffffc01092b6a0
x29: ffffffc01092b6a0 x28: 0000000000000028 x27: 0000000000000038
x26: 0000000000000004 x25: ffffffd2973dce48 x24: 0000000000000000
x23: 00000000ffffffff x22: 00000000ffffffff x21: ffffffd2978d0008
x20: ffffffd2978d0008 x19: ffffff80ff759fc0 x18: 0000000000000000
x17: 004800a501260460 x16: 0441043b04600438 x15: 04380000089807d0
x14: 07b0089807800780 x13: 0000000000000000 x12: 0000000000000000
x11: 0000000000000438 x10: 00000000000007d0 x9 : ffffffd2973e09e4
x8 : ffffff8092d53300 x7 : ffffff808902e8b8 x6 : 0000000000000001
x5 : ffffff808902e880 x4 : 0000000000000000 x3 : ffffff80ff759fc0
x2 : 0000000000000001 x1 : 0000000000000000 x0 : ffffff80ff759fc0
Kernel panic - not syncing: Asynchronous SError Interrupt
CPU: 7 PID: 3878 Comm: Xorg Not tainted 5.19.0-stb-cbq #19
Hardware name: Google Lazor (rev3 - 8) (DT)
Call trace:
 dump_backtrace.part.0+0xbc/0xe4
 show_stack+0x24/0x70
 dump_stack_lvl+0x68/0x84
 dump_stack+0x18/0x34
 panic+0x14c/0x32c
 nmi_panic+0x58/0x7c
 arm64_serror_panic+0x78/0x84
 do_serror+0x40/0x64
 el1h_64_error_handler+0x30/0x48
 el1h_64_error+0x68/0x6c
 __cmpxchg_case_acq_32+0x14/0x2c
 _raw_spin_lock_irqsave+0x38/0x4c
 lock_timer_base+0x40/0x78
 __mod_timer+0xf4/0x25c
 schedule_timeout+0xd4/0xfc
 __wait_for_common+0xac/0x140
 wait_for_completion_timeout+0x2c/0x54
 dp_ctrl_push_idle+0x40/0x88
 dp_bridge_disable+0x24/0x30
 drm_atomic_bridge_chain_disable+0x90/0xbc
 drm_atomic_helper_commit_modeset_disables+0x198/0x444
 msm_atomic_commit_tail+0x1d0/0x374
 commit_tail+0x80/0x108
 drm_atomic_helper_commit+0x118/0x11c
 drm_atomic_commit+0xb4/0xe0
 drm_client_modeset_commit_atomic+0x184/0x224
 drm_client_modeset_commit_locked+0x58/0x160
 drm_client_modeset_commit+0x3c/0x64
 __drm_fb_helper_restore_fbdev_mode_unlocked+0x98/0xac
 drm_fb_helper_set_par+0x74/0x80
 drm_fb_helper_hotplug_event+0xdc/0xe0
 __drm_fb_helper_restore_fbdev_mode_unlocked+0x7c/0xac
 drm_fb_helper_restore_fbdev_mode_unlocked+0x20/0x2c
 drm_fb_helper_lastclose+0x20/0x2c
 drm_lastclose+0x44/0x6c
 drm_release+0x88/0xd4
 __fput+0x104/0x220
 ____fput+0x1c/0x28
 task_work_run+0x8c/0x100
 do_exit+0x450/0x8d0
 do_group_exit+0x40/0xac
 __wake_up_parent+0x0/0x38
 invoke_syscall+0x84/0x11c
 el0_svc_common.constprop.0+0xb8/0xe4
 do_el0_svc+0x8c/0xb8
 el0_svc+0x2c/0x54
 el0t_64_sync_handler+0x120/0x1c0
 el0t_64_sync+0x190/0x194
SMP: stopping secondary CPUs
Kernel Offset: 0x128e800000 from 0xffffffc008000000
PHYS_OFFSET: 0x80000000
CPU features: 0x800,00c2a015,19801c82
Memory Limit: none

Changes in v2:
-- add more commit text

Changes in v3:
-- add comments into dp_bridge_atomic_check()

Changes in v4:
-- rewording the comment into dp_bridge_atomic_check()

Changes in v5:
-- removed quote x at end of commit text

Changes in v6:
-- removed quote x at end of comment in dp_bridge_atomic_check()

Fixes: 8a3b4c17f863 ("drm/msm/dp: employ bridge mechanism for display enable and disable")
Reported-by: Leonard Lausen <leonard@lausen.nl>
Suggested-by: Rob Clark <robdclark@gmail.com>
Closes: https://gitlab.freedesktop.org/drm/msm/-/issues/17
Signed-off-by: Kuogee Hsieh <quic_khsieh@quicinc.com>
Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/505331/
Link: https://lore.kernel.org/r/1664408211-25314-1-git-send-email-quic_khsieh@quicinc.com
Signed-off-by: Abhinav Kumar <quic_abhinavk@quicinc.com>

drm/msm/gpu: Fix crash during system suspend after unbind

In adreno_unbind, we should clean up gpu device's drvdata to avoid
accessing a stale pointer during system suspend. Also, check for NULL
ptr in both system suspend/resume callbacks.

Signed-off-by: Akhil P Oommen <quic_akhilpo@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/505075/
Link: https://lore.kernel.org/r/20220928124830.2.I5ee0ac073ccdeb81961e5ec0cce5f741a7207a71@changeid
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/a6xx: Replace kcalloc() with kvzalloc()

In order to reduce chance of allocation failure while capturing a6xx
gpu state, use kvzalloc() instead of kcalloc() in state_kcalloc().

Indirectly, this patch helps to fix leaking memory allocated for
gmu_debug object.

Fixes: b859f9b009b (drm/msm/gpu: Snapshot GMU debug buffer)
Signed-off-by: Akhil P Oommen <quic_akhilpo@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/505074/
Link: https://lore.kernel.org/r/20220928124830.1.I8ea24a8d586b4978823b848adde000f92f74d5c2@changeid
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/gem: Unpin objects slightly later

The introduction of "drm/msm/gem: Evict active GEM objects when necessary"
exposes a problem with "drm/msm/gem: Unpin buffers earlier", in that we
need to keep the object pinned in the time the submit is queued up in the
gpu scheduler.  Otherwise the shrinker will see it as a thing that can be
evicted if we wait for it to be signaled.  But if the shrinker path is
waiting on it with the obj lock held, the job cannot be scheduled, as that
also requires briefly grabbing the obj lock, leading to deadlock.  (Not to
mention, we don't want the shrinker to evict an obj queued up in gpu
scheduler.)

Fixes: f371bcc0c2ac ("drm/msm/gem: Unpin buffers earlier")
Fixes: 025d27239a2f ("drm/msm/gem: Evict active GEM objects when necessary")
Closes: https://gitlab.freedesktop.org/drm/msm/-/issues/19
Signed-off-by: Rob Clark <robdclark@chromium.org>
Tested-by: Chia-I Wu <olvaffe@gmail.com>
Patchwork: https://patchwork.freedesktop.org/patch/504528/
Link: https://lore.kernel.org/r/20220923224043.2449152-1-robdclark@gmail.com

drm/msm/hdmi: make hdmi_phy_8996 OF clk provider

On MSM8996 the HDMI PHY provides the PLL clock to the MMCC. As we are
preparing to convert the MSM8996 to use DT clocks properties (rather
than global clock names), register the OF clock provider.

While we are at it, also change the driver to use clk_parent_data rather
parent_names to setup a link to the XO clock.

Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/492318/
Link: https://lore.kernel.org/r/20220704161148.814510-3-dmitry.baryshkov@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

dt-bindings: phy: qcom, hdmi-phy-qmp: add clock-cells and XO clock

As the QMP HDMI PHY is a clock provider, add constant #clock-cells
property. For the compatibility with older DTs the property is not
marked as required. Also add the XO clock to the list of the clocks used
by the driver.

Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Acked-by: Vinod Koul <vkoul@kernel.org>
Patchwork: https://patchwork.freedesktop.org/patch/492316/
Link: https://lore.kernel.org/r/20220704161148.814510-2-dmitry.baryshkov@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dsi: Remove use of device_node in dsi_host_parse_dt()

Clang warns:

  drivers/gpu/drm/msm/dsi/dsi_host.c:1903:14: error: variable 'device_node' is uninitialized when used here [-Werror,-Wuninitialized]
          of_node_put(device_node);
                      ^~~~~~~~~~~
  drivers/gpu/drm/msm/dsi/dsi_host.c:1870:44: note: initialize the variable 'device_node' to silence this warning
          struct device_node *endpoint, *device_node;
                                                    ^
                                                    = NULL
  1 error generated.

device_node's assignment was removed but not all of its uses. Remove the
call to of_node_put() and the variable declaration to clean up the
warning.

Fixes: 36246dd50225 ("drm/msm/dsi: switch to DRM_PANEL_BRIDGE")
Link: https://github.com/ClangBuiltLinux/linux/issues/1700
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/500182/
Link: https://lore.kernel.org/r/20220829165450.217628-1-nathan@kernel.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dsi: switch to DRM_PANEL_BRIDGE

Currently the DSI driver has two separate paths: one if the next device
in a chain is a bridge and another one if the panel is connected
directly to the DSI host. Simplify the code path by using panel-bridge
driver (already selected in Kconfig) and dropping support for
handling the panel directly.

Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/493608/
Link: https://lore.kernel.org/r/20220712132258.671263-1-dmitry.baryshkov@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/panel: drop DSC pps pointer

Complete the move of DSC data pointer from struct drm_panel to struct
mipi_dsi_device.

Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/493309/
Link: https://lore.kernel.org/r/20220711094320.368062-4-dmitry.baryshkov@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dsi: fetch DSC pps payload from struct mipi_dsi_device

Now that struct mipi_dsi_device provides DSC data, fetch it from the
mentioned struct rather than from the struct drm_panel itself. This
would allow supporting MIPI DSI bridges handling DSC on their input
side.

Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/493307/
Link: https://lore.kernel.org/r/20220711094320.368062-3-dmitry.baryshkov@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/mipi-dsi: pass DSC data through the struct mipi_dsi_device

The commit 0f40ba48de3b ("drm/msm/dsi: Pass DSC params to drm_panel")
added a pointer to the DSC data to the struct drm_panel. However DSC
support is not limited to the DSI panels. MIPI DSI bridges can also
consume DSC command streams. Thus add struct drm_dsc_config pointer to
the struct mipi_dsi_device.

Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/493306/
Link: https://lore.kernel.org/r/20220711094320.368062-2-dmitry.baryshkov@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dsi: Remove the unneeded result variable

Return the value msm_dsi_phy_enable() directly instead of storing it in
another redundant variable.

Reported-by: Zeal Robot <zealci@zte.com.cn>
Signed-off-by: ye xingchen <ye.xingchen@zte.com.cn>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/499621/
Link: https://lore.kernel.org/r/20220826072821.253150-1-ye.xingchen@zte.com.cn
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dsi: drop the hpd worker

It makes no sense to have the HPD worker in the MSM DSI driver anymore.
It is only queued from the dsi_host_attach/detach() callbacks, where
it plays no useful role. Either way the panel or next bridge will be
present and will report it's status directly.

Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/498740/
Link: https://lore.kernel.org/r/20220822174417.292926-1-dmitry.baryshkov@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dsi: Improve dsi_phy_driver_probe() probe error handling

The dsi_phy_driver_probe() function has a "goto fail" for no
reason. Change it to just always return directly when it sees an
error. Make this simpler by leveraging dev_err_probe() which is
designed to make code like this shorter / simpler.

NOTE: as part of this, we now pass through error codes directly from
msm_ioremap_size() rather than translating to -ENOMEM. This changed
mostly because it's much more convenient when using dev_err_probe()
and also it's usually encouraged not to hide error codes like the old
code was doing unless there is a good reason. I can't see any reason
why we'd need to return -ENOMEM instead of -EINVAL from the probe
function.

Signed-off-by: Douglas Anderson <dianders@chromium.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/496324/
Link: https://lore.kernel.org/r/20220804073608.v4.6.I969118a35934a0e5007fe4f80e3e28e9c0b7602a@changeid
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dsi: Take advantage of devm_regulator_bulk_get_const()

As of the commit 1de452a0edda ("regulator: core: Allow drivers to
define their init data as const") we no longer need to do copying of
regulator bulk data from initdata to something dynamic. Let's take
advantage of that.

In addition to saving some code, this also moves us to using
ARRAY_SIZE() to specify how many regulators we have which is less
error prone.

This gets rid of some layers of wrappers which makes it obvious that
we can get rid of an extra error print.
devm_regulator_bulk_get_const() prints errors for you so you don't
need an extra layer of printing.

In all cases here I have preserved the old settings without any
investigation about whether the loads being set are sensible. In the
cases of some of the PHYs if several PHYs in the same file used
exactly the same settings I had them point to the same data structure.

NOTE: Though I haven't done the math, this is likely an overall
savings in terms of "static const" data. We previously always
allocated space for 8 supplies. Each of these supplies took up 36
bytes of data (32 for name, 4 for an int).

Signed-off-by: Douglas Anderson <dianders@chromium.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/496325/
Link: https://lore.kernel.org/r/20220804073608.v4.5.I55a9e65cb1c22221316629e98768ff473f47a067@changeid
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dsi: Use the new regulator bulk feature to specify the load

As of commit 6eabfc018e8d ("regulator: core: Allow specifying an
initial load w/ the bulk API") we can now specify the initial load in
the bulk data rather than having to manually call regulator_set_load()
on each regulator. Let's use it.

Signed-off-by: Douglas Anderson <dianders@chromium.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/496319/
Link: https://lore.kernel.org/r/20220804073608.v4.4.I7b3c72949883846badb073cfeae985c55239da1d@changeid
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dsi: Don't set a load before disabling a regulator

As of commit 5451781dadf8 ("regulator: core: Only count load for
enabled consumers"), a load isn't counted for a disabled
regulator. That means all the code in the DSI driver to specify and
set loads before disabling a regulator is not actually doing anything
useful. Let's remove it.

It should be noted that all of the loads set that were being specified
were pointless noise anyway. The only use for this number is to pick
between low power and high power modes of regulators. Regulators
appear to do this changeover at loads on the order of 10000 uA. You
would need a lot of clients of the same rail for that 100 uA number to
count for anything.

Note that now that we get rid of the setting of the load at disable
time, we can just set the load once when we first get the regulator
and then forget it.

It should also be noted that the regulator functions
regulator_bulk_enable() and regulator_set_load() already print error
messages when they encounter problems so while moving things around we
get rid of some extra error prints.

Signed-off-by: Douglas Anderson <dianders@chromium.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/496320/
Link: https://lore.kernel.org/r/20220804073608.v4.3.If1f94fbbdb7c1d0fb3961de61483a851ad1971a7@changeid
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dsi_phy_7nm: Replace parent names with clk_hw pointers

parent_hw pointers are easier to manage and cheaper to use than
repeatedly formatting the parent name and subsequently leaving the clk
framework to perform lookups based on that name.

Signed-off-by: Marijn Suijten <marijn.suijten@somainline.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/491927/
Link: https://lore.kernel.org/r/20220629225331.357308-12-marijn.suijten@somainline.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dsi_phy_10nm: Replace parent names with clk_hw pointers

parent_hw pointers are easier to manage and cheaper to use than
repeatedly formatting the parent name and subsequently leaving the clk
framework to perform lookups based on that name.

Signed-off-by: Marijn Suijten <marijn.suijten@somainline.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/491923/
Link: https://lore.kernel.org/r/20220629225331.357308-11-marijn.suijten@somainline.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dsi_phy_14nm: Replace parent names with clk_hw pointers

parent_hw pointers are easier to manage and cheaper to use than
repeatedly formatting the parent name and subsequently leaving the clk
framework to perform lookups based on that name.

Signed-off-by: Marijn Suijten <marijn.suijten@somainline.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/491921/
Link: https://lore.kernel.org/r/20220629225331.357308-10-marijn.suijten@somainline.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dsi_phy_28nm: Replace parent names with clk_hw pointers

parent_hw pointers are easier to manage and cheaper to use than
repeatedly formatting the parent name and subsequently leaving the clk
framework to perform lookups based on that name.

Signed-off-by: Marijn Suijten <marijn.suijten@somainline.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/491925/
Link: https://lore.kernel.org/r/20220629225331.357308-9-marijn.suijten@somainline.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dsi_phy_28nm_8960: Replace parent names with clk_hw pointers

parent_hw pointers are easier to manage and cheaper to use than
repeatedly formatting the parent name and subsequently leaving the clk
framework to perform lookups based on that name.

Signed-off-by: Marijn Suijten <marijn.suijten@somainline.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/491929/
Link: https://lore.kernel.org/r/20220629225331.357308-8-marijn.suijten@somainline.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dsi/phy: Replace hardcoded char-array length with sizeof()

Now that the last DSI PHY PLL driver (dsi_phy_28nm_8960) has been
converted to use a simple stack-local char-array instead of a
devm_kzalloc heap allocation we can safely call sizeof() on every string
variable (that's now a sized array instead of a pointer) passed into
snprintf instead of hardcoding the size.

Signed-off-by: Marijn Suijten <marijn.suijten@somainline.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/491920/
Link: https://lore.kernel.org/r/20220629225331.357308-7-marijn.suijten@somainline.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dsi_phy_28nm_8960: Use stack memory for temporary clock names

The clock names formatted into the hw_clk's init structure are only used
for the duration of the registration function where they are kstrdup'ed,
making it unnecessary to keep the allocations alive for the duration of
the device (through devm).

Just like the other DSI PHY PLL clock trees, use a stack-local char
array and save on memory outside of the pll_28nm_register function.

Signed-off-by: Marijn Suijten <marijn.suijten@somainline.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/491917/
Link: https://lore.kernel.org/r/20220629225331.357308-6-marijn.suijten@somainline.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dsi/phy: Reindent and reflow multiline function calls

The commit 613cbd1da3c9 ("drm/msm/dsi: use devm_clk_*register to registe
DSI PHY clocks") introduced the devm_ prefix to clk_hw registration
calls, without updating the indentation of the arguments on the
following lines.

Similarly commit e55b3fbbbbc8 ("drm/msm/dsi: drop PLL accessor
functions") moved from pll_write to dsi_phy_write without updating the
indentation of followup arguments either.

Preparing for a series that heavily touches the clk calls, reflow and
reindent function calls that are adhering to an 80-char column limit by
spanning multiple lines.  Where function names are very long the
arguments are indented with a fixed number of two tab characters instead
of aligning with the opening parenthesis of the function call.

Signed-off-by: Marijn Suijten <marijn.suijten@somainline.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/491931/
Link: https://lore.kernel.org/r/20220629225331.357308-5-marijn.suijten@somainline.org
[DB: adjusted commit message to make checkpatch happy]
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dp: correct 1.62G link rate at dp_catalog_ctrl_config_msa()

At current implementation there is an extra 0 at 1.62G link rate which
cause no correct pixel_div selected for 1.62G link rate to calculate
mvid and nvid. This patch delete the extra 0 to have mvid and nvid be
calculated correctly.

Changes in v2:
-- fix Fixes tag's text

Changes in v3:
-- fix misspelling of "Reviewed-by"

Fixes: 937f941ca06f  ("drm/msm/dp: Use qmp phy for DP PLL and PHY")
Signed-off-by: Kuogee Hsieh <quic_khsieh@quicinc.com>
Reviewed-by: Stephen Boyd <swboyd@chromium.org>
Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/499328/
Link: https://lore.kernel.org/r/1661372150-3764-1-git-send-email-quic_khsieh@quicinc.com
[DB: rewrapped commit message]
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dp: Silence inconsistent indent warning

Build robots complain

 smatch warnings:
 drivers/gpu/drm/msm/dp/dp_link.c:969 dp_link_process_link_status_update() warn: inconsistent indenting

Fix it along with a trailing space from the same commit.

Cc: Kuogee Hsieh <quic_khsieh@quicinc.com>
Fixes: ea530388e64b ("drm/msm/dp: skip checking LINK_STATUS_UPDATED bit")
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/498954/
Link: https://lore.kernel.org/r/20220823212302.1744145-1-swboyd@chromium.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

dt-bindings: msm/dp: handle DP vs eDP difference

The #sound-dai-cells property should be used only for DP controllers. It
doesn't make sense for eDP, there is no support for audio output. The
aux-bus should not be used for DP controllers. Also p1 MMIO region
should be used only for DP controllers.

Take care of these differences.

Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Stephen Boyd <swboyd@chromium.org>
Patchwork: https://patchwork.freedesktop.org/patch/498751/
Link: https://lore.kernel.org/r/20220822184900.307160-4-dmitry.baryshkov@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

dt-bindings: msm/dp: add missing properties

Document missing definitions for opp-table (DP controller OPPs), aux-bus
(DP AUX BUS) and data-lanes (DP/eDP lanes mapping) properties.

Reviewed-by: Stephen Boyd <swboyd@chromium.org>
Acked-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/498750/
Link: https://lore.kernel.org/r/20220822184900.307160-3-dmitry.baryshkov@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

dt-bindings: msm/dp: mark vdda supplies as deprecated

The commit 85936d4f3815 ("phy: qcom-qmp: add regulator_set_load to dp
phy") moved setting regulator load to the DP PHY driver (QMP). Then, the
commit 7516351bebc1 ("drm/msm/dp: delete vdda regulator related
functions from eDP/DP controller") removed support for VDDA supplies
from the DP controller driver.
Mark these properties as deprecated and drop them from the example.

Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Stephen Boyd <swboyd@chromium.org>
Patchwork: https://patchwork.freedesktop.org/patch/498748/
Link: https://lore.kernel.org/r/20220822184900.307160-2-dmitry.baryshkov@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dp: Get rid of dp_ctrl_on_stream_phy_test_report()

This API isn't really more than a couple lines now that we don't store
the pixel_rate to the struct member. Inline it into the caller.

Cc: Kuogee Hsieh <quic_khsieh@quicinc.com>
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/490775/
Link: https://lore.kernel.org/r/20220623002540.871994-4-swboyd@chromium.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dp: Remove pixel_rate from struct dp_ctrl

This struct member is stored to in the function that calls the function
which uses it. That's possible with a function argument instead of
storing to a struct member. Pass the pixel_rate as an argument instead
to simplify the code. Note that dp_ctrl_link_maintenance() was storing
the pixel_rate but never using it so we just remove the assignment from
there.

Cc: Kuogee Hsieh <quic_khsieh@quicinc.com>
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/490772/
Link: https://lore.kernel.org/r/20220623002540.871994-3-swboyd@chromium.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dp: Reorganize code to avoid forward declaration

Let's move these functions around to avoid having to forward declare
dp_ctrl_on_stream_phy_test_report(). Also remove
dp_ctrl_reinitialize_mainlink() forward declaration because we're doing
that sort of task.

Reviewed-by: Kuogee Hsieh <quic_khsieh@quicinc.com>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
Patchwork: https://patchwork.freedesktop.org/patch/490771/
Link: https://lore.kernel.org/r/20220623002540.871994-2-swboyd@chromium.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dpu: get rid of cached flush_mask

Instead of querying the CTL for the flush mask (for SSPP, LM or DSPP),
storing the mask in the mixer configuration and then pushing the mask to
the CTL, tell CTL to cache the flush in place.

Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/473159/
Link: https://lore.kernel.org/r/20220209172520.3719906-9-dmitry.baryshkov@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dpu: inline dpu_plane_get_ctl_flush

There is no need to keep a separate function for calling into the ctl if
we already know all the details. Inline this function in the dpu_crtc.c

Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/473152/
Link: https://lore.kernel.org/r/20220209172520.3719906-7-dmitry.baryshkov@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dpu: do not limit the zpos property

Stop limiting zpos property values, we use normalized_zpos anyway. And
nothing stops userspace from assigning several planes to a single zpos
(it is a userspace bug, but the kernel is forgiving about it).

Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/473147/
Link: https://lore.kernel.org/r/20220209172520.3719906-3-dmitry.baryshkov@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dpu: rip out master planes support

Master/virtual planes were used for multirect support. In preparation to
reworking DPU planes, drop support for master planes (which was not used
anyway).

Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/473146/
Link: https://lore.kernel.org/r/20220209172520.3719906-2-dmitry.baryshkov@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dpu: drop unused variable from dpu_kms_mdp_snapshot()

Follow up the merge of address fields and drop the variable that became
unused after the commit 9403f9a42c88 ("drm/msm/dpu: merge base_off with
blk_off in struct dpu_hw_blk_reg_map").

Fixes: 9403f9a42c88 ("drm/msm/dpu: merge base_off with blk_off in struct dpu_hw_blk_reg_map")
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Stephen Boyd <swboyd@chromium.org>
Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/498736/
Link: https://lore.kernel.org/r/20220822172204.281045-1-dmitry.baryshkov@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm: Make .remove and .shutdown HW shutdown consistent

Drivers' .remove and .shutdown callbacks are executed on different code
paths. The former is called when a device is removed from the bus, while
the latter is called at system shutdown time to quiesce the device.

This means that some overlap exists between the two, because both have to
take care of properly shutting down the hardware. But currently the logic
used in these two callbacks isn't consistent in msm drivers, which could
lead to kernel panic.

For example, on .remove the component is deleted and its .unbind callback
leads to the hardware being shutdown but only if the DRM device has been
marked as registered.

That check doesn't exist in the .shutdown logic and this can lead to the
driver calling drm_atomic_helper_shutdown() for a DRM device that hasn't
been properly initialized.

A situation like this can happen if drivers for expected sub-devices fail
to probe, since the .bind callback will never be executed. If that is the
case, drm_atomic_helper_shutdown() will attempt to take mutexes that are
only initialized if drm_mode_config_init() is called during a device bind.

This bug was attempted to be fixed in commit 623f279c7781 ("drm/msm: fix
shutdown hook in case GPU components failed to bind"), but unfortunately
it still happens in some cases as the one mentioned above, i.e:

  systemd-shutdown[1]: Powering off.
  kvm: exiting hardware virtualization
  platform wifi-firmware.0: Removing from iommu group 12
  platform video-firmware.0: Removing from iommu group 10
  ------------[ cut here ]------------
  WARNING: CPU: 6 PID: 1 at drivers/gpu/drm/drm_modeset_lock.c:317 drm_modeset_lock_all_ctx+0x3c4/0x3d0
  ...
  Hardware name: Google CoachZ (rev3+) (DT)
  pstate: a0400009 (NzCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
  pc : drm_modeset_lock_all_ctx+0x3c4/0x3d0
  lr : drm_modeset_lock_all_ctx+0x48/0x3d0
  sp : ffff80000805bb80
  x29: ffff80000805bb80 x28: ffff327c00128000 x27: 0000000000000000
  x26: 0000000000000000 x25: 0000000000000001 x24: ffffc95d820ec030
  x23: ffff327c00bbd090 x22: ffffc95d8215eca0 x21: ffff327c039c5800
  x20: ffff327c039c5988 x19: ffff80000805bbe8 x18: 0000000000000034
  x17: 000000040044ffff x16: ffffc95d80cac920 x15: 0000000000000000
  x14: 0000000000000315 x13: 0000000000000315 x12: 0000000000000000
  x11: 0000000000000000 x10: 0000000000000000 x9 : 0000000000000000
  x8 : ffff80000805bc28 x7 : 0000000000000000 x6 : 0000000000000000
  x5 : 0000000000000000 x4 : 0000000000000000 x3 : 0000000000000000
  x2 : ffff327c00128000 x1 : 0000000000000000 x0 : ffff327c039c59b0
  Call trace:
   drm_modeset_lock_all_ctx+0x3c4/0x3d0
   drm_atomic_helper_shutdown+0x70/0x134
   msm_drv_shutdown+0x30/0x40
   platform_shutdown+0x28/0x40
   device_shutdown+0x148/0x350
   kernel_power_off+0x38/0x80
   __do_sys_reboot+0x288/0x2c0
   __arm64_sys_reboot+0x28/0x34
   invoke_syscall+0x48/0x114
   el0_svc_common.constprop.0+0x44/0xec
   do_el0_svc+0x2c/0xc0
   el0_svc+0x2c/0x84
   el0t_64_sync_handler+0x11c/0x150
   el0t_64_sync+0x18c/0x190
  ---[ end trace 0000000000000000 ]---
  Unable to handle kernel NULL pointer dereference at virtual address 0000000000000018
  Mem abort info:
    ESR = 0x0000000096000004
    EC = 0x25: DABT (current EL), IL = 32 bits
    SET = 0, FnV = 0
    EA = 0, S1PTW = 0
    FSC = 0x04: level 0 translation fault
  Data abort info:
    ISV = 0, ISS = 0x00000004
    CM = 0, WnR = 0
  user pgtable: 4k pages, 48-bit VAs, pgdp=000000010eab1000
  [0000000000000018] pgd=0000000000000000, p4d=0000000000000000
  Internal error: Oops: 96000004 [#1] PREEMPT SMP
  ...
  Hardware name: Google CoachZ (rev3+) (DT)
  pstate: a0400009 (NzCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
  pc : ww_mutex_lock+0x28/0x32c
  lr : drm_modeset_lock_all_ctx+0x1b0/0x3d0
  sp : ffff80000805bb50
  x29: ffff80000805bb50 x28: ffff327c00128000 x27: 0000000000000000
  x26: 0000000000000000 x25: 0000000000000001 x24: 0000000000000018
  x23: ffff80000805bc10 x22: ffff327c039c5ad8 x21: ffff327c039c5800
  x20: ffff80000805bbe8 x19: 0000000000000018 x18: 0000000000000034
  x17: 000000040044ffff x16: ffffc95d80cac920 x15: 0000000000000000
  x14: 0000000000000315 x13: 0000000000000315 x12: 0000000000000000
  x11: 0000000000000000 x10: 0000000000000000 x9 : 0000000000000000
  x8 : ffff80000805bc28 x7 : 0000000000000000 x6 : 0000000000000000
  x5 : 0000000000000000 x4 : 0000000000000000 x3 : 0000000000000000
  x2 : ffff327c00128000 x1 : 0000000000000000 x0 : 0000000000000018
  Call trace:
   ww_mutex_lock+0x28/0x32c
   drm_modeset_lock_all_ctx+0x1b0/0x3d0
   drm_atomic_helper_shutdown+0x70/0x134
   msm_drv_shutdown+0x30/0x40
   platform_shutdown+0x28/0x40
   device_shutdown+0x148/0x350
   kernel_power_off+0x38/0x80
   __do_sys_reboot+0x288/0x2c0
   __arm64_sys_reboot+0x28/0x34
   invoke_syscall+0x48/0x114
   el0_svc_common.constprop.0+0x44/0xec
   do_el0_svc+0x2c/0xc0
   el0_svc+0x2c/0x84
   el0t_64_sync_handler+0x11c/0x150
   el0t_64_sync+0x18c/0x190
  Code: aa0103f4 d503201f d2800001 aa0103e3 (c8e37c02)
  ---[ end trace 0000000000000000 ]---
  Kernel panic - not syncing: Attempted to kill init! exitcode=0x0000000b
  Kernel Offset: 0x495d77c00000 from 0xffff800008000000
  PHYS_OFFSET: 0xffffcd8500000000
  CPU features: 0x800,00c2a015,19801c82
  Memory Limit: none
  ---[ end Kernel panic - not syncing: Attempted to kill init! exitcode=0x0000000b ]---

Fixes: 9d5cbf5fe46e ("drm/msm: add shutdown support for display platform_driver")
Signed-off-by: Javier Martinez Canillas <javierm@redhat.com>
Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/497842/
Link: https://lore.kernel.org/r/20220816134612.916527-1-javierm@redhat.com
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dpu: Fix comment typo

The double `be' is duplicated in the comment, remove one.

Signed-off-by: Jason Wang <wangborong@cdjrlc.com>
Fixes: 25fdd5933e4c ("drm/msm: Add SDM845 DPU support")
Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/497853/
Link: https://lore.kernel.org/r/20220724204242.4107-1-wangborong@cdjrlc.com
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dpu: drop unused memory allocation

Drop the dpu_cfg variable and corresponding kzalloc, which became unused
after changing hw catalog to static configuration.

Fixes: de7d480f5e8c ("drm/msm/dpu: make dpu hardware catalog static const")
Reported-by: kernel test robot <lkp@intel.com>
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/498738/
Link: https://lore.kernel.org/r/20220822172455.282923-1-dmitry.baryshkov@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dpu: drop VBIF indices

We do not expect to have other VBIFs. Drop VBIF_n indices and always use
VBIF_RT and VBIF_NRT.

Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/489571/
Link: https://lore.kernel.org/r/20220615125703.24647-3-dmitry.baryshkov@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dpu: fix error handling around dpu_hw_vbif_init

Using IS_ERR_OR_NULL() together with PTR_ERR() is a typical mistake. If
the value is NULL, then the function will return 0 instead of a proper
return code. Moreover dpu_hw_vbif_init() function can not return NULL.
So, replace corresponding IS_ERR_OR_NULL() call with IS_ERR().

Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/489570/
Link: https://lore.kernel.org/r/20220615125703.24647-2-dmitry.baryshkov@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dpu: index dpu_kms->hw_vbif using vbif_idx

Remove loops over hw_vbif. Instead always VBIF's idx as an index in the
array. This fixes an error in dpu_kms_hw_init(), where we fill
dpu_kms->hw_vbif[i], but check for an error pointer at
dpu_kms->hw_vbif[vbif_idx].

Fixes: 25fdd5933e4c ("drm/msm: Add SDM845 DPU support")
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/489569/
Link: https://lore.kernel.org/r/20220615125703.24647-1-dmitry.baryshkov@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm: lookup the ICC paths in both mdp5/dpu and mdss devices

The commit 6874f48bb8b0 ("drm/msm: make mdp5/dpu devices master
components") changed the MDP5 driver to look for the interconnect paths
in the MDSS device rather than in the MDP5 device itself. This was left
unnoticed since on my testing devices the interconnects probably didn't
reach the sync state.

Rather than just using the MDP5 device for ICC path lookups for the MDP5
devices, introduce an additional helper to check both MDP5/DPU and MDSS
nodes. This will be helpful for the MDP5->DPU conversion, since the
driver will have to check both nodes.

Fixes: 6874f48bb8b0 ("drm/msm: make mdp5/dpu devices master components")
Reported-by: Marijn Suijten <marijn.suijten@somainline.org>
Reported-by: Yassine Oudjana <y.oudjana@protonmail.com>
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Tested-by: Marijn Suijten <marijn.suijten@somainline.org> # On sdm630
Tested-by: Yassine Oudjana <y.oudjana@protonmail.com> # msm8996
Patchwork: https://patchwork.freedesktop.org/patch/496488/
Link: https://lore.kernel.org/r/20220805115630.506391-1-dmitry.baryshkov@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

dt-bindings: display/msm: dpu-sdm845: add missing DPU opp-table

The 'display-controller' child (DPU) of Display SubSystem (MDSS) uses
opp-table, so reference it which allows restricting DPU schema to fixed
list of properties.

Fixes: 3d7a0dd8f39b ("dt-bindings: msm: disp: add yaml schemas for DPU bindings")
Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/499212/
Link: https://lore.kernel.org/r/20220817062059.18640-6-krzysztof.kozlowski@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

dt-bindings: display/msm: dpu-sc7280: add missing DPU opp-table

The 'display-controller' child (DPU) of Display SubSystem (MDSS) uses
opp-table, so reference it which allows restricting DPU schema to fixed
list of properties.

Fixes: 57fd4f34ddac ("dt-bindings: msm: add DT bindings for sc7280")
Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/497898/
Link: https://lore.kernel.org/r/20220817062059.18640-5-krzysztof.kozlowski@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

dt-bindings: display/msm: dpu-sc7180: add missing DPU opp-table

The 'display-controller' child (DPU) of Display SubSystem (MDSS) uses
opp-table, so reference it which allows restricting DPU schema to fixed
list of properties.

Fixes: 3d7a0dd8f39b ("dt-bindings: msm: disp: add yaml schemas for DPU bindings")
Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/497894/
Link: https://lore.kernel.org/r/20220817062059.18640-4-krzysztof.kozlowski@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

dt-bindings: display/msm: dpu-qcm2290: add missing DPU opp-table

The 'display-controller' child (DPU) of Display SubSystem (MDSS) uses
opp-table, so reference it which allows restricting DPU schema to fixed
list of properties.

Fixes: 164f69d9d45a ("dt-bindings: msm: disp: add yaml schemas for QCM2290 DPU bindings")
Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/497893/
Link: https://lore.kernel.org/r/20220817062059.18640-3-krzysztof.kozlowski@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

dt-bindings: display/msm: dpu-msm8998: add missing DPU opp-table

The 'display-controller' child (DPU) of Display SubSystem (MDSS) uses
opp-table, so reference it which allows restricting DPU schema to fixed
list of properties.

Fixes: 6e986a8f1cf1 ("dt-bindings: display: msm: Add binding for msm8998 dpu")
Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/497896/
Link: https://lore.kernel.org/r/20220817062059.18640-2-krzysztof.kozlowski@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

dt-bindings: display/msm/mdp4: require 4 IOMMUs

APQ8064, the only supported mdp4 platform, uses 4 IOMMUs for the MDP4
device. Update schema accordingly. When we have other MDP4 platforms,
this spec can be updated accordingly.

Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Acked-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/492541/
Link: https://lore.kernel.org/r/20220706145222.1565238-5-dmitry.baryshkov@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

dt-bindings: display/msm/gmu: account for different GMU variants

Make display/msm/gmu.yaml describe all existing GMU variants rather than
just the 630.2 (SDM845) version of it.

Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/492539/
Link: https://lore.kernel.org/r/20220706145222.1565238-4-dmitry.baryshkov@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

dt-bindings: display/msm/gpu: allow specifying several IOMMU nodes

Different platforms require different amounts of IOMMUs. The APQ8064
uses 64 IOMMU instances for GPU, adjust the schema accordingly.

Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Acked-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/492536/
Link: https://lore.kernel.org/r/20220706145222.1565238-2-dmitry.baryshkov@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dsi: use drm_dsc_config instead of msm_display_dsc_config

There is no need to use the struct msm_display_dsc_config wrapper inside
the dsi driver, use the struct drm_dsc_config directly to pass pps data.

Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Marijn Suijten <marijn.suijten@somainline.org>
Patchwork: https://patchwork.freedesktop.org/patch/493341/
Link: https://lore.kernel.org/r/20220711100432.455268-3-dmitry.baryshkov@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dpu: use drm_dsc_config instead of msm_display_dsc_config

There is no need to use the struct msm_display_dsc_config wrapper inside
the dpu driver, use the struct drm_dsc_config directly to pass pps data.

Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Reviewed-by: Marijn Suijten <marijn.suijten@somainline.org>
Patchwork: https://patchwork.freedesktop.org/patch/493340/
Link: https://lore.kernel.org/r/20220711100432.455268-2-dmitry.baryshkov@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/rd: Fix FIFO-full deadlock

If the previous thing cat'ing $debugfs/rd left the FIFO full, then
subsequent open could deadlock in rd_write() (because open is blocked,
not giving a chance for read() to consume any data in the FIFO).  Also
it is generally a good idea to clear out old data from the FIFO.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Patchwork: https://patchwork.freedesktop.org/patch/496706/
Link: https://lore.kernel.org/r/20220807160901.2353471-2-robdclark@gmail.com
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/gpu: Drop qos request if devm_devfreq_add_device() fails

In the event that devm_devfreq_add_device() fails the device's qos freq
list is left referencing df->idle_freq and df->boost_freq. Attempting to
initialize devfreq again after a probe deferral will then cause invalid
memory accesses in dev_pm_qos_add_request().

Fix this by dropping the requests in the error path.

Fixes: 7c0ffcd40b16 ("drm/msm/gpu: Respect PM QoS constraints")
Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/493001/
Link: https://lore.kernel.org/r/20220708162632.3529864-1-bjorn.andersson@linaro.org
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dsi: Fix number of regulators for SDM660

1 regulator is listed but the number 2 is specified. This presumably
means we try to get a regulator with no name. Fix it.

Fixes: 462f7017a691 ("drm/msm/dsi: Fix DSI and DSI PHY regulator config from SDM660")
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Marijn Suijten <marijn.suijten@somainline.org>
Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/496323/
Link: https://lore.kernel.org/r/20220804073608.v4.2.I94b3c3e412b7c208061349f05659e126483171b1@changeid
Signed-off-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dsi: Fix number of regulators for msm8996_dsi_cfg

3 regulators are listed but the number 2 is specified. Fix it.

Fixes: 3a3ff88a0fc1 ("drm/msm/dsi: Add 8x96 info in dsi_cfg")
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/496318/
Link: https://lore.kernel.org/r/20220804073608.v4.1.I1056ee3f77f71287f333279efe4c85f88d403f65@changeid
Signed-off-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dp: delete DP_RECOVERED_CLOCK_OUT_EN to fix tps4

Data Symbols scrambled is required for tps4 at link training 2.
Therefore SCRAMBLING_DISABLE bit should not be set for tps4 to
work.

RECOVERED_CLOCK_OUT_EN is for enable simple EYE test for jitter
measurement with minimal equipment for embedded applications purpose
and is not required to be set during normal operation. Current
implementation always have RECOVERED_CLOCK_OUT_EN bit set which
cause SCRAMBLING_DISABLE bit wrongly set at tps4 which prevent
tps4 from working.

This patch delete setting RECOVERED_CLOCK_OUT_EN to fix
SCRAMBLING_DISABLE be wrongly set at tps4.

Changes in v2:
-- fix Fixes tag

Changes in v3:
-- revise commit text

Changes in v4:
-- fix commit text newline

Changes in v5:
-- fix commit text line over 75 chars

Fixes: c943b4948b58 ("drm/msm/dp: add displayPort driver support")
Signed-off-by: Kuogee Hsieh <quic_khsieh@quicinc.com>
Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Reviewed-by: Stephen Boyd <swboyd@chromium.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/497194/
Link: https://lore.kernel.org/r/1660258670-4200-1-git-send-email-quic_khsieh@quicinc.com
Signed-off-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dpu: populate wb or intf before reset_intf_cfg

dpu_encoder_helper_phys_cleanup() was not populating neither
wb or intf to the intf_cfg before calling the reset_intf_cfg().

This causes the reset of the active bits of wb/intf to be
skipped which is incorrect.

Fix this by populating the relevant wb or intf indices correctly.

Fixes: ae4d721ce100 ("drm/msm/dpu: add an API to reset the encoder related hw blocks")
Signed-off-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Reviewed-by: Jessica Zhang <quic_jesszhan@quicinc.com>
Tested-by: Jessica Zhang <quic_jesszhan@quicinc.com> # Trogdor (SC8170)
Patchwork: https://patchwork.freedesktop.org/patch/494298/
Link: https://lore.kernel.org/r/1657912468-17254-1-git-send-email-quic_abhinavk@quicinc.com
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dsi: fix the inconsistent indenting

Fix the inconsistent indenting in function msm_dsi_dphy_timing_calc_v3().

Fix the following smatch warnings:

drivers/gpu/drm/msm/dsi/phy/dsi_phy.c:350 msm_dsi_dphy_timing_calc_v3() warn: inconsistent indenting

Fixes: f1fa7ff44056 ("drm/msm/dsi: implement auto PHY timing calculator for 10nm PHY")
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: sunliming <sunliming@kylinos.cn>
Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/494662/
Link: https://lore.kernel.org/r/20220719015622.646718-1-sunliming@kylinos.cn
Signed-off-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/dp: make eDP panel as the first connected connector

Some userspace presumes that the first connected connector is the main
display, where it's supposed to display e.g. the login screen. For
laptops, this should be the main panel.

This patch call drm_helper_move_panel_connectors_to_head() after
drm_bridge_connector_init() to make sure eDP stay at head of
connected connector list. This fixes unexpected corruption happen
at eDP panel if eDP is not placed at head of connected connector
list.

Changes in v2:
-- move drm_helper_move_panel_connectors_to_head() to
dpu_kms_drm_obj_init()

Changes in v4:
-- move drm_helper_move_panel_connectors_to_head() to msm_drm_init()

Signed-off-by: Kuogee Hsieh <quic_khsieh@quicinc.com>
Reviewed-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Reviewed-by: Douglas Anderson <dianders@chromium.org>
Tested-by: Douglas Anderson <dianders@chromium.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Stephen Boyd <swboyd@chromium.org>
Fixes: ef7837ff091c ("drm/msm/dp: Add DP controllers for sc7280")
Patchwork: https://patchwork.freedesktop.org/patch/492581/
Link: https://lore.kernel.org/r/1657135928-31195-1-git-send-email-quic_khsieh@quicinc.com
Signed-off-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/a6xx: Handle GMU prepare-slumber hfi failure

When prepare-slumber hfi fails, we should follow a6xx_gmu_force_off()
sequence.

Signed-off-by: Akhil P Oommen <quic_akhilpo@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/498401/
Link: https://lore.kernel.org/r/20220819015030.v5.7.I54815c7c36b80d4725cd054e536365250454452f@changeid
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/a6xx: Improve gpu recovery sequence

We can do a few more things to improve our chance at a successful gpu
recovery, especially during a hangcheck timeout:
1. Halt CP and GMU core
2. Do RBBM GBIF HALT sequence
3. Do a soft reset of GPU core

Signed-off-by: Akhil P Oommen <quic_akhilpo@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/498400/
Link: https://lore.kernel.org/r/20220819015030.v5.6.Idf2ba51078e87ae7ceb75cc77a5bd4ff2bd31eab@changeid
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm/a6xx: Ensure CX collapse during gpu recovery

Because there could be transient votes from other drivers/tz/hyp which
may keep the cx gdsc enabled, we should poll until cx gdsc collapses.
We can use the reset framework to poll for cx gdsc collapse from gpucc
clk driver.

This feature requires support from the platform's gpucc driver.

Signed-off-by: Akhil P Oommen <quic_akhilpo@quicinc.com>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
Patchwork: https://patchwork.freedesktop.org/patch/498397/
Link: https://lore.kernel.org/r/20220819015030.v5.5.I176567525af2b9439a7e485d0ca130528666a55c@changeid
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm: Fix cx collapse issue during recovery

There are some hardware logic under CX domain. For a successful
recovery, we should ensure cx headswitch collapses to ensure all the
stale states are cleard out. This is especially true to for a6xx family
where we can GMU co-processor.

Currently, cx doesn't collapse due to a devlink between gpu and its
smmu. So the *struct gpu device* needs to be runtime suspended to ensure
that the iommu driver removes its vote on cx gdsc.

Signed-off-by: Akhil P Oommen <quic_akhilpo@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/498398/
Link: https://lore.kernel.org/r/20220819015030.v5.4.I4ac27a0b34ea796ce0f938bb509e257516bc6f57@changeid
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm: Correct pm_runtime votes in recover worker

In the scenario where there is one a single submit which is hung, gpu is
power collapsed when it is retired. Because of this, by the time we call
reover(), gpu state would be already clear. Fix this by correctly
managing the pm runtime votes.

Signed-off-by: Akhil P Oommen <quic_akhilpo@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/498391/
Link: https://lore.kernel.org/r/20220819015030.v5.3.Ib07ecec3d5c17cb0e1efa6fcddaaa019ec2fb556@changeid
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm: Take single rpm refcount on behalf of all submits

Instead of separate refcount for each submit, take single rpm refcount
on behalf of all the submits. This makes it easier to drop the rpm
refcount during recovery in an upcoming patch.

Signed-off-by: Akhil P Oommen <quic_akhilpo@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/498392/
Link: https://lore.kernel.org/r/20220819015030.v5.2.Ifee853f6d8217a0fdacc459092bbc9e81a8a7ac7@changeid
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm: Remove unnecessary pm_runtime_get/put

We already enable gpu power from msm_gpu_submit(), so avoid a duplicate
pm_runtime_get/put from msm_job_run().

Signed-off-by: Akhil P Oommen <quic_akhilpo@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/498390/
Link: https://lore.kernel.org/r/20220819015030.v5.1.Icf1e8f0c9b3e7e9933c3b48c70477d0582f3243f@changeid
Signed-off-by: Rob Clark <robdclark@chromium.org>

drm/msm: De-open-code some CP_EVENT_WRITE

Replace some open coding to improve readability.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/499272/
Link: https://lore.kernel.org/r/20220821155441.1092134-1-robdclark@gmail.com

drm/msm/iommu: optimize map/unmap

Using map_pages/unmap_pages cuts down on the # of pgtable walks needed
in the process of finding where to insert/remove an entry.  The end
result is ~5-10x faster than mapping a single page at a time.

v2: Rename iommu_pgsize(), drop obsolete comments, fix error handling
    in msm_iommu_pagetable_map()

Signed-off-by: Rob Clark <robdclark@chromium.org>
Reviewed-by: Sai Prakash Ranjan <quic_saipraka@quicinc.com>
Patchwork: https://patchwork.freedesktop.org/patch/498892/
Link: https://lore.kernel.org/r/20220823163719.90399-1-robdclark@gmail.com

drm/msm: Add fault-injection support

Intended as a way to trigger error paths in mesa.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/496710/
Link: https://lore.kernel.org/r/20220807172848.2432845-1-robdclark@gmail.com

drm/msm/gem: Convert to lockdep assert

Utilize the power of lockdep for our GEM locking related sanity
checking.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Patchwork: https://patchwork.freedesktop.org/patch/496139/
Link: https://lore.kernel.org/r/20220802155152.1727594-16-robdclark@gmail.com

drm/msm/gem: Add msm_gem_assert_locked()

All use of msm_gem_is_locked() is just for WARN_ON()s, so extract out
into an msm_gem_assert_locked() patch.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Patchwork: https://patchwork.freedesktop.org/patch/496136/
Link: https://lore.kernel.org/r/20220802155152.1727594-15-robdclark@gmail.com

drm/msm/gem: Evict active GEM objects when necessary

If we are under enough memory pressure, we should stall waiting for
active buffers to become idle in order to evict.

v2: Check for __GFP_ATOMIC before blocking

Signed-off-by: Rob Clark <robdclark@chromium.org>
Patchwork: https://patchwork.freedesktop.org/patch/496135/
Link: https://lore.kernel.org/r/20220802155152.1727594-14-robdclark@gmail.com

drm/msm/gem: Consolidate shrinker trace

Combine separate trace events for purge vs evict into one.  When we add
support for purging/evicting active buffers we'll just add more info
into this one trace event, rather than adding a bunch more events.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Patchwork: https://patchwork.freedesktop.org/patch/496133/
Link: https://lore.kernel.org/r/20220802155152.1727594-13-robdclark@gmail.com

drm/msm/gem: Unpin buffers earlier

We've already attached the fences, so obj->resv (which shrinker checks)
tells us whether they are still active.  So we can unpin sooner, before
we drop the queue lock.

This also avoids the need to grab the obj lock in the retire path,
avoiding potential for lock contention between submit and retire.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Patchwork: https://patchwork.freedesktop.org/patch/496132/
Link: https://lore.kernel.org/r/20220802155152.1727594-12-robdclark@gmail.com

drm/msm/gem: Convert to using drm_gem_lru

This converts over to use the shared GEM LRU/shrinker helpers.  Note
that it means we are no longer tracking purgeable or willneed buffers
that are active separately.  But the most recently pinned buffers should
be at the tail of the various LRUs, and the shrinker is already prepared
to encounter objects which are still active.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Patchwork: https://patchwork.freedesktop.org/patch/496131/
Link: https://lore.kernel.org/r/20220802155152.1727594-11-robdclark@gmail.com

drm/gem: Add LRU/shrinker helper

Add a simple LRU helper to assist with driver's shrinker implementation.
It handles tracking the number of backing pages associated with a given
LRU, and provides a helper to implement shrinker_scan.

A driver can use multiple LRU instances to track objects in various
states, for example a dontneed LRU for purgeable objects, a willneed LRU
for evictable objects, and an unpinned LRU for objects without backing
pages.

All LRUs that the object can be moved between must share a single lock.

v2: lockdep_assert_held() instead of WARN_ON(!mutex_is_locked())
v3: make drm_gem_lru_move_tail_locked() static until there is a user

Cc: Daniel Vetter <daniel@ffwll.ch>
Cc: Thomas Zimmermann <tzimmermann@suse.de>
Cc: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Signed-off-by: Rob Clark <robdclark@chromium.org>
Reviewed-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Patchwork: https://patchwork.freedesktop.org/patch/496128/
Link: https://lore.kernel.org/r/20220802155152.1727594-10-robdclark@gmail.com

drm/msm/gem: Remove active refcnt

At this point the pinned refcnt is sufficient, and the shrinker is
already prepared to encounter objects which are still active according
to fences attached to the resv.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Patchwork: https://patchwork.freedesktop.org/patch/496122/
Link: https://lore.kernel.org/r/20220802155152.1727594-9-robdclark@gmail.com

drm/msm/gem: Consolidate pin/unpin paths

Avoid having multiple spots where we increment/decrement pin_count (and
associated LRU updating)

Signed-off-by: Rob Clark <robdclark@chromium.org>
Patchwork: https://patchwork.freedesktop.org/patch/496130/
Link: https://lore.kernel.org/r/20220802155152.1727594-8-robdclark@gmail.com

drm/msm/gem: Rename to pin/unpin_pages

Since that is what these fxns actually do.. they are getting *pinned*
pages (as opposed to cases where we need pages, but don't need them
pinned, like CPU mappings).

Signed-off-by: Rob Clark <robdclark@chromium.org>
Patchwork: https://patchwork.freedesktop.org/patch/496121/
Link: https://lore.kernel.org/r/20220802155152.1727594-7-robdclark@gmail.com

drm/msm/gem: Rename update_inactive

Really what this is doing is updating various LRU lists.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Patchwork: https://patchwork.freedesktop.org/patch/496115/
Link: https://lore.kernel.org/r/20220802155152.1727594-6-robdclark@gmail.com

drm/msm/gem: Check for active in shrinker path

Currently in our shrinker path we shouldn't be encountering anything
that is active, but this will change in subsequent patches.  So check
if there are unsignaled fences.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Patchwork: https://patchwork.freedesktop.org/patch/496117/
Link: https://lore.kernel.org/r/20220802155152.1727594-5-robdclark@gmail.com

drm/msm: Split out idr_lock

Otherwise if we hit reclaim pinning objects in the submit path, we'll be
blocking retire_worker trying to free a submit.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Patchwork: https://patchwork.freedesktop.org/patch/496116/
Link: https://lore.kernel.org/r/20220802155152.1727594-4-robdclark@gmail.com

drm/msm: Small submit cleanup

Move more initialization into submit_create().

Signed-off-by: Rob Clark <robdclark@chromium.org>
Patchwork: https://patchwork.freedesktop.org/patch/496120/
Link: https://lore.kernel.org/r/20220802155152.1727594-3-robdclark@gmail.com

drm/msm: Reorder lock vs submit alloc

This lets us drop the NORETRY.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Patchwork: https://patchwork.freedesktop.org/patch/496114/
Link: https://lore.kernel.org/r/20220802155152.1727594-2-robdclark@gmail.com