After commit 8020f0f931 ("drm/amd/amdgpu: enable W=1 for amdgpu"),
there is an instance of -Wunused-const-variable when CONFIG_DEBUG_FS is
disabled:
drivers/gpu/drm/amd/amdgpu/../pm/amdgpu_pm.c:38:34: error: unused variable 'clocks' [-Werror,-Wunused-const-variable]
38 | static const struct cg_flag_name clocks[] = {
| ^
1 error generated.
clocks is only used when CONFIG_DEBUG_FS is set, so move the definition
into the CONFIG_DEBUG_FS block right above its only usage to clear up
the warning.
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Also, correct the comment for NAVI10_PCIE__LC_L1_INACTIVITY_TBT_DEFAULT
as 0x0000000E stands for 400ms instead of 4ms.
Signed-off-by: Evan Quan <evan.quan@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Per VM BOs must be marked as moved or otherwise their ranges are not
updated on use which might be necessary when the replace operation
splits mappings.
This fixes random GPU hangs when replacing sparse mappings from the
userspace, while OP_MAP/OP_UNMAP works fine because always valid BOs
are correctly handled there.
Cc: stable@vger.kernel.org
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Not all devices carry VBIOS version information. Add the device
attribute only if supported.
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Le Ma <le.ma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
After commit 8020f0f931 ("drm/amd/amdgpu: enable W=1 for amdgpu"),
there is an instance of -Wunused-const-variable when CONFIG_DEBUG_FS is
disabled:
drivers/gpu/drm/amd/amdgpu/amdgpu_rap.c:110:37: error: unused variable 'amdgpu_rap_debugfs_ops' [-Werror,-Wunused-const-variable]
110 | static const struct file_operations amdgpu_rap_debugfs_ops = {
| ^
1 error generated.
There is no reason for the body of this function to be guarded when
CONFIG_DEBUG_FS is disabled, as debugfs_create_file() is a stub that
just returns an error pointer in that situation. Remove the preprocessor
guards so that the variable never appears unused, while not changing
anything at run time.
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
fix the following coccicheck warning:
drivers/gpu/drm/amd/pm/swsmu/smu11/navi10_ppt.c:1657:14-18: Unneeded variable: "size".
Signed-off-by: Mingtong Bao <baomingtong001@208suo.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Identify issues that arise by using the tests/doubletest.cocci
semantic patch. Need to remove duplicate expression in if statement.
Signed-off-by: Wang Ming <machel@vivo.com>
Reviewed-by: Ammar Faizi <ammarfaizi2@gnuweeb.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
In order to read the EDID from an eDP panel, you not only need to
power on the bridge chip itself but also the panel. In the ps8640
driver, this was made to work by having the bridge chip manually power
the panel on by calling pre_enable() on everything connectorward on
the bridge chain. This worked OK, but...
...when trying to do the same thing on ti-sn65dsi86, feedback was that
this wasn't a great idea. As a result, we designed the "DP AUX"
bus. With the design we ended up with the panel driver itself was in
charge of reading the EDID. The panel driver could power itself on and
the bridge chip was able to power itself on because it implemented the
DP AUX bus.
Despite the fact that we came up with a new scheme, implemented in on
ti-sn65dsi86, and even implemented it on parade-ps8640, we still kept
the old code around. This was because the new scheme required a DT
change. Previously the panel was a simple "platform_device" and in DT
at the top level. With the new design the panel needs to be listed in
DT under the DP controller node. The old code allowed us to properly
fetch EDIDs with ps8640 with the old DTs.
Unfortunately, the old code stopped working as of commit 102e80d1fa
("drm/bridge: ps8640: Use atomic variants of drm_bridge_funcs"). There
are cases at bootup where connector->state->state is NULL and the
kernel crashed at:
* drm_atomic_bridge_chain_pre_enable
* drm_atomic_get_old_bridge_state
* drm_atomic_get_old_private_obj_state
The crash went away at commit 4fb912e5e1 ("drm/bridge: Introduce
pre_enable_prev_first to alter bridge init order") which added a NULL
check. However, even though we were no longer crashing the end result
was that we weren't actually powering the panel on when we thought we
were. Things could end up working (despite warning splats) if
userspace was persistent and tried to get the mode again, but it
wasn't great.
A bit of digging was done to see if there was an easy fix but there
was nothing obvious. Instead, the only device using ps8640 the "old"
way had its DT updated so that the panel was no longer a simple
"platform_deice". See commit c2d94f7214 ("arm64: dts: mediatek:
mt8173-elm: Move display to ps8640 auxiliary bus") and commit
113b5cc06f44 ("arm64: dts: mediatek: mt8173-elm: remove panel model
number in DT").
Let's delete the old broken code so nobody gets tempted to copy it or
figure out how it works (since it doesn't).
NOTE: from a device tree "purist" point of view, we're supposed to
keep old device trees working and this patch is technically "against
policy". Reasons I'm still proposing it anyway:
1. Officially, old mt8173-elm device trees worked via the "little
white lie" approach. The DT would list an arbitrary/representative
panel that would be used for power sequencing. The mode information
in the panel driver would then be ignored / overridden by the EDID
reading code in ps8640. I don't feel too terrible breaking DTs that
contained the wrong "compatible" string to begin with. NOTE that
any old device trees that _didn't_ lie about their compatible will
still work because the mode information will come from the
hardcoded panels in panel-edp.
2. The only users of the old code were Chromebooks and Chromebooks
don't bake their DTs into the BIOS (they are bundled with the
kernel). Thus we don't need to worry about breaking someone using
an old DT with a new kernel.
3. The old code was broken anyway. If someone wants to fix the old
code instead of deleting it then they have my blessing, but without
a proper fix the old code isn't useful.
Reviewed-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Signed-off-by: Robert Foss <rfoss@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20230616165517.v2.1.I7b8f60b3fbfda068f9bf452d584dc934494bfbfa@changeid
Commit 5d844091f2 ("drm/scdc-helper: Pimp SCDC debugs") changed the scdc
interface to pick up an i2c adapter from a connector instead. However, in
the case of dw-hdmi, the wrong connector was being used to pass i2c adapter
information, since dw-hdmi's embedded connector structure is only populated
when the bridge attachment callback explicitly asks for it.
drm-meson is handling connector creation, so this won't happen, leading to
a NULL pointer dereference.
Fix it by having scdc functions access dw-hdmi's current connector pointer
instead, which is assigned during the bridge enablement stage.
Fixes: 5d844091f2 ("drm/scdc-helper: Pimp SCDC debugs")
Signed-off-by: Adrián Larumbe <adrian.larumbe@collabora.com>
Reported-by: Lukas F. Hartmann <lukas@mntre.com>
Acked-by: Neil Armstrong <neil.armstrong@linaro.org>
[narmstrong: moved Fixes tag before first S-o-b and added Reported-by tag]
Signed-off-by: Neil Armstrong <neil.armstrong@linaro.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20230601123153.196867-1-adrian.larumbe@collabora.com
Before this patch, booting to Linux VT and doing a simple:
echo 2 > /sys/class/graphics/fb0/blank
echo 0 > /sys/class/graphics/fb0/blank
would result in failures to re-enable the panel. Mode set callback is
called only once during boot in this scenario, while calls to
enable/disable callbacks are balanced afterwards. The driver doesn't
work unless userspace calls modeset before enabling the CRTC/connector.
This patch moves enabling of the DSI host from mode_set into pre_enable
callback, and removes some old hacks where this bridge driver is
directly calling into other bridge driver's callbacks.
pre_enable_prev_first flag is set on the panel's bridge so that panel
drivers will get their prepare function called between DSI host's
pre_enable and enable callbacks, so that they get a chance to
perform panel setup while DSI host is already enabled in command
mode. Otherwise panel's prepare would be called before DSI host
is enabled, and any DSI communication used in prepare callback
would fail.
With all these changes, the enable/disable sequence is now well
balanced, and host's and panel's callbacks are called in proper order
documented in the drm_panel API documentation without needing the old
hacks. (Mainly that panel->prepare is called when DSI host is ready to
allow the panel driver to send DSI commands and vice versa during
disable.)
Tested on Pinephone Pro. Trace of the callbacks follows.
Before:
[ 1.253882] dw-mipi-dsi-rockchip ff960000.dsi: mode_set
[ 1.290732] panel-himax-hx8394 ff960000.dsi.0: prepare
[ 1.475576] dw-mipi-dsi-rockchip ff960000.dsi: enable
[ 1.475593] panel-himax-hx8394 ff960000.dsi.0: enable
echo 2 > /sys/class/graphics/fb0/blank
[ 13.722799] panel-himax-hx8394 ff960000.dsi.0: disable
[ 13.774502] dw-mipi-dsi-rockchip ff960000.dsi: post_disable
[ 13.774526] panel-himax-hx8394 ff960000.dsi.0: unprepare
echo 0 > /sys/class/graphics/fb0/blank
[ 17.735796] panel-himax-hx8394 ff960000.dsi.0: prepare
[ 17.923522] dw-mipi-dsi-rockchip ff960000.dsi: enable
[ 17.923540] panel-himax-hx8394 ff960000.dsi.0: enable
[ 17.944330] dw-mipi-dsi-rockchip ff960000.dsi: failed to write command FIFO
[ 17.944335] panel-himax-hx8394 ff960000.dsi.0: sending command 0xb9 failed: -110
[ 17.944340] panel-himax-hx8394 ff960000.dsi.0: Panel init sequence failed: -110
echo 2 > /sys/class/graphics/fb0/blank
[ 431.148583] panel-himax-hx8394 ff960000.dsi.0: disable
[ 431.169259] dw-mipi-dsi-rockchip ff960000.dsi: failed to write command FIFO
[ 431.169268] panel-himax-hx8394 ff960000.dsi.0: Failed to enter sleep mode: -110
[ 431.169282] dw-mipi-dsi-rockchip ff960000.dsi: post_disable
[ 431.169316] panel-himax-hx8394 ff960000.dsi.0: unprepare
[ 431.169357] pclk_mipi_dsi0 already disabled
echo 0 > /sys/class/graphics/fb0/blank
[ 432.796851] panel-himax-hx8394 ff960000.dsi.0: prepare
[ 432.981537] dw-mipi-dsi-rockchip ff960000.dsi: enable
[ 432.981568] panel-himax-hx8394 ff960000.dsi.0: enable
[ 433.002290] dw-mipi-dsi-rockchip ff960000.dsi: failed to write command FIFO
[ 433.002299] panel-himax-hx8394 ff960000.dsi.0: sending command 0xb9 failed: -110
[ 433.002312] panel-himax-hx8394 ff960000.dsi.0: Panel init sequence failed: -110
-----------------------------------------------------------------------
After:
[ 1.248372] dw-mipi-dsi-rockchip ff960000.dsi: mode_set
[ 1.248704] dw-mipi-dsi-rockchip ff960000.dsi: pre_enable
[ 1.285377] panel-himax-hx8394 ff960000.dsi.0: prepare
[ 1.468392] dw-mipi-dsi-rockchip ff960000.dsi: enable
[ 1.468421] panel-himax-hx8394 ff960000.dsi.0: enable
echo 2 > /sys/class/graphics/fb0/blank
[ 16.210357] panel-himax-hx8394 ff960000.dsi.0: disable
[ 16.261315] dw-mipi-dsi-rockchip ff960000.dsi: post_disable
[ 16.261339] panel-himax-hx8394 ff960000.dsi.0: unprepare
echo 0 > /sys/class/graphics/fb0/blank
[ 19.161453] dw-mipi-dsi-rockchip ff960000.dsi: pre_enable
[ 19.197869] panel-himax-hx8394 ff960000.dsi.0: prepare
[ 19.382141] dw-mipi-dsi-rockchip ff960000.dsi: enable
[ 19.382158] panel-himax-hx8394 ff960000.dsi.0: enable
(But depends on functionality intorduced in Linux 6.3, so this patch will
not build on older kernels when applied to older stable branches.)
Fixes: 46fc51546d ("drm/bridge/synopsys: Add MIPI DSI host controller bridge")
Signed-off-by: Ondrej Jirman <megi@xff.cz>
Reviewed-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Robert Foss <rfoss@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20230617224915.1923630-1-megi@xff.cz
drm_sched_entity_kill_jobs_cb() logic is omitting the last fence popped
from the dependency array that was waited upon before
drm_sched_entity_kill() was called (drm_sched_entity::dependency field),
so we're basically waiting for all dependencies except one.
In theory, this wait shouldn't be needed because resources should have
their users registered to the dma_resv object, thus guaranteeing that
future jobs wanting to access these resources wait on all the previous
users (depending on the access type, of course). But we want to keep
these explicit waits in the kill entity path just in case.
Let's make sure we keep all dependencies in the array in
drm_sched_job_dependency(), so we can iterate over the array and wait
in drm_sched_entity_kill_jobs_cb().
We also make sure we wait on drm_sched_fence::finished if we were
originally asked to wait on drm_sched_fence::scheduled. In that case,
we assume the intent was to delegate the wait to the firmware/GPU or
rely on the pipelining done at the entity/scheduler level, but when
killing jobs, we really want to wait for completion not just scheduling.
v2:
- Don't evict deps in drm_sched_job_dependency()
v3:
- Always wait for drm_sched_fence::finished fences in
drm_sched_entity_kill_jobs_cb() when we see a sched_fence
v4:
- Fix commit message
- Fix a use-after-free bug
v5:
- Flag deps on which we should only wait for the scheduled event
at insertion time
v6:
- Back to v4 implementation
- Add Christian's R-b
Cc: Frank Binns <frank.binns@imgtec.com>
Cc: Sarah Walker <sarah.walker@imgtec.com>
Cc: Donald Robson <donald.robson@imgtec.com>
Cc: Luben Tuikov <luben.tuikov@amd.com>
Cc: David Airlie <airlied@gmail.com>
Cc: Daniel Vetter <daniel@ffwll.ch>
Cc: Sumit Semwal <sumit.semwal@linaro.org>
Cc: "Christian König" <christian.koenig@amd.com>
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Suggested-by: "Christian König" <christian.koenig@amd.com>
Reviewed-by: "Christian König" <christian.koenig@amd.com>
Acked-by: Luben Tuikov <luben.tuikov@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230619071921.3465992-1-boris.brezillon@collabora.com
Move the register programming part, which actually enables the bridge and
makes it push data out of its DPI side, into the enable callback. The DSI
host like DSIM may not be able to transmit commands in pre_enable, moving
the register programming into enable assures it can transmit commands.
Signed-off-by: Marek Vasut <marex@denx.de>
Reviewed-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Robert Foss <rfoss@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20230615201902.566182-1-marex@denx.de
The scenario being fixed here is depicted in the following sequence-
modprobe i915
echo 1 > /sys/class/drm/card0/gt/gt0/slpc_ignore_eff_freq
echo 300 > /sys/class/drm/card0/gt_min_freq_mhz (RPn)
cat /sys/class/drm/card0/gt_cur_freq_mhz --> cur == RPn as expected
echo 1 > /sys/kernel/debug/dri/0/gt0/reset --> reset
cat /sys/class/drm/card0/gt_min_freq_mhz --> cached freq is RPn
cat /sys/class/drm/card0/gt_cur_freq_mhz --> it's not RPn, but RPe!!
When SLPC reinitializes, it sets SLPC min freq to efficient frequency.
Even if we disable efficient freq post that, we should restore the cached
min freq (via H2G) for it to take effect.
v2: Clarify commit message (Ashutosh)
Fixes: 95ccf312a1 ("drm/i915/guc/slpc: Allow SLPC to use efficient frequency")
Reviewed-by: Ashutosh Dixit <ashutosh.dixit@intel.com>
Signed-off-by: Vinay Belgaumkar <vinay.belgaumkar@intel.com>
Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230621014257.1769564-1-vinay.belgaumkar@intel.com
Currently the FE is spinning way too fast when polling for new work in
the FE idleloop. As each poll fetches 16 bytes from memory, a GPU running
at 1GHz with the current setting of 200 wait cycle between fetches causes
80 MB/s of memory traffic just to check for new work when the GPU is
otherwise idle, which is more FE traffic than in some GPU loaded cases.
Significantly increase the number of wait cycles to slow down the poll
interval to ~30µs, limiting the FE idle memory traffic to 512 KB/s, while
providing a max latency which should not hurt most use-cases. The FE WAIT
command seems to have some unknown discrete steps in the wait cycles so
we may over/undershoot the target a bit, but that should be harmless.
If the GPU core base frequency is unknown keep the 200 wait cycles as
a sane default.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Sui Jingfeng <suijingfeng@loongson.cn>
Tested-by: Sui Jingfeng <suijingfeng@loongson.cn>
Reviewed-by: Christian Gmeiner <cgmeiner@igalia.com>
Etnaviv doesn't use anything from of_platform.h, but depends on
of.h, of_device.h, and platform_device.h which are all implicitly
included, but that is going to be removed soon.
Signed-off-by: Rob Herring <robh@kernel.org>
Tested-by: Sui Jingfeng <suijingfeng@loongson.cn>
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
