Using the ring_muxer without preemption adds overhead for no
reason since mcbp cannot be triggered.
Moving back to a single queue in this case also helps when
high priority app are used: in this case the gpu_scheduler
priority handling will work as expected - much better than
ring_muxer with its 2 independant schedulers competing for
the same hardware queue.
This change requires moving amdgpu_device_set_mcbp above
amdgpu_device_ip_early_init because we use adev->gfx.mcbp.
Signed-off-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Acked-by: Jiadong Zhu <Jiadong.Zhu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Remove the unused function - amdgpu_vm_pt_is_root_clean
and remove the impossible condition
v1: entries == 0 is not possible any more,
so this condition could probably be removed (Felix)
Signed-off-by: Jesse Zhang <jesse.zhang@amd.com>
Suggested-by：Felix Kuehling <felix.kuehling@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Third argument of i915_request_wait() accepts a timeout value in jiffies.
Most users pass either a simple HZ based expression, or a result of
msecs_to_jiffies(), or MAX_SCHEDULE_TIMEOUT, or a very small number not
exceeding 4 if applicable as that value. However, there is one user --
intel_selftest_wait_for_rq() -- that passes a WAIT_FOR_RESET_TIME symbol,
defined as a large constant value that most probably represents a desired
timeout in ms. While that usage results in the intended value of timeout
on usual x86_64 kernel configurations, it is not portable across different
architectures and custom kernel configs.
Rename the symbol to clearly indicate intended units and convert it to
jiffies before use.
Fixes: 3a4bfa091c ("drm/i915/selftest: Fix workarounds selftest for GuC submission")
Signed-off-by: Janusz Krzysztofik <janusz.krzysztofik@linux.intel.com>
Cc: Rahul Kumar Singh <rahul.kumar.singh@intel.com>
Cc: John Harrison <John.C.Harrison@Intel.com>
Cc: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Andi Shyti <andi.shyti@linux.intel.com>
Signed-off-by: Andi Shyti <andi.shyti@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240222113347.648945-2-janusz.krzysztofik@linux.intel.com
(cherry picked from commit 6ee3f54b88)
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
The icl+ power well code currently assumes that every AUX power
well maps to an encoder which is using said power well. That is
by no menas guaranteed as we:
- only register encoders for ports declared in the VBT
- combo PHY HDMI-only encoder no longer get an AUX CH since
commit 9856308c94 ("drm/i915: Only populate aux_ch if really needed")
However we have places such as intel_power_domains_sanitize_state()
that blindly traverse all the possible power wells. So these bits
of code may very well encounbter an aux power well with no associated
encoder.
In this particular case the BIOS seems to have left one AUX power
well enabled even though we're dealing with a HDMI only encoder
on a combo PHY. We then proceed to turn off said power well and
explode when we can't find a matching encoder. As a short term fix
we should be able to just skip the PHY related parts of the power
well programming since we know this situation can only happen with
combo PHYs.
Another option might be to go back to always picking an AUX CH for
all encoders. However I'm a bit wary about that since we might in
theory end up conflicting with the VBT AUX CH assignment. Also
that wouldn't help with encoders not declared in the VBT, should
we ever need to poke the corresponding power wells.
Longer term we need to figure out what the actual relationship
is between the PHY vs. AUX CH vs. AUX power well. Currently this
is entirely unclear.
Cc: stable@vger.kernel.org
Fixes: 9856308c94 ("drm/i915: Only populate aux_ch if really needed")
Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/10184
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240223203216.15210-1-ville.syrjala@linux.intel.com
Reviewed-by: Imre Deak <imre.deak@intel.com>
(cherry picked from commit 6a8c66bf0e)
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
It turned out that executing the SET_Q_MODE packet on every submission
creates to much overhead.
Implement a workaround which allows skipping the SET_Q_MODE packet if
subsequent submissions all use the same parameters.
v2: add a NULL check for ring_obj
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
First of all calculating the number of dw to patch into a
conditional execution is not something HW generation specific.
This is just standard ring buffer calculations. While at it also
reduce the BUG_ON() into WARN_ON().
Then instead of a random bit pattern use 0 as default value for
the number of dw skipped, this way it's not mandatory any more
to patch the conditional execution.
And last make the address to check a parameter of the
conditional execution instead of getting this from the ring.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Because the rpm_mode flag is already set when the driver
is initialized, we use it directly for runtime suspend/resume
instead of checking it again
Signed-off-by: Ma Jun <Jun.Ma2@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This patch changes the handling and lifecycle of vm->task_info object.
The major changes are:
- vm->task_info is a dynamically allocated ptr now, and its uasge is
reference counted.
- introducing two new helper funcs for task_info lifecycle management
- amdgpu_vm_get_task_info: reference counts up task_info before
returning this info
- amdgpu_vm_put_task_info: reference counts down task_info
- last put to task_info() frees task_info from the vm.
This patch also does logistical changes required for existing usage
of vm->task_info.
V2: Do not block all the prints when task_info not found (Felix)
V3: Fixed review comments from Felix
- Fix wrong indentation
- No debug message for -ENOMEM
- Add NULL check for task_info
- Do not duplicate the debug messages (ti vs no ti)
- Get first reference of task_info in vm_init(), put last
in vm_fini()
V4: Fixed review comments from Felix
- fix double reference increment in create_task_info
- change amdgpu_vm_get_task_info_pasid
- additional changes in amdgpu_gem.c while porting
Cc: Christian Koenig <christian.koenig@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Shashank Sharma <shashank.sharma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
fix the issue:
"amdgpu: Failed to create process VM object".
[Why]when amdgpu initialized, seq64 do mampping and update bo mapping in vm page table.
But when clifo run. It also initializes a vm for a process device through the function kfd_process_device_init_vm and ensure the root PD is clean through the function amdgpu_vm_pt_is_root_clean.
So they have a conflict, and clinfo always failed.
v1:
- remove all the pte_supports_ats stuff from the amdgpu_vm code (Felix)
Signed-off-by: Jesse Zhang <Jesse.Zhang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This function does not build on 32-bit targets when the compiler
fails to reduce DIV_ROUND_UP() into a shift:
ld.lld: error: undefined symbol: __aeabi_uldivmod
>>> referenced by xe_migrate.c
>>> drivers/gpu/drm/xe/xe_migrate.o:(pte_update_size) in archive vmlinux.a
There are two instances in this function. Change the first to
use an open-coded shift with the same behavior, and the second
one to a 32-bit calculation, which is sufficient here as the size
is never more than 2^32 pages (16TB).
Fixes: 237412e453 ("drm/xe: Enable 32bits build")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240226124736.1272949-3-arnd@kernel.org
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
(cherry picked from commit 1408784b59)
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
When the driver is built-in but the tests are in loadable modules,
the helpers don't actually get put into the driver:
ERROR: modpost: "xe_kunit_helper_alloc_xe_device" [drivers/gpu/drm/xe/tests/xe_test.ko] undefined!
Change the Makefile to ensure they are always part of the driver
even when the rest of the kunit tests are in loadable modules.
Fixes: 5095d13d75 ("drm/xe/kunit: Define helper functions to allocate fake xe device")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240226124736.1272949-1-arnd@kernel.org
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
(cherry picked from commit 0e6fec6da2)
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
If drm_kms_helper_poll=n the output poll work will only get scheduled
from drm_helper_probe_single_connector_modes() to handle a delayed
hotplug event. Since polling is disabled the work in this case should
just call drm_kms_helper_hotplug_event() w/o detecting the state of
connectors and rescheduling the work.
After commit d33a54e399 after a delayed hotplug event above the
connectors did get re-detected in the poll work and the work got
re-scheduled periodically (since poll_running is also false if
drm_kms_helper_poll=n), in effect ignoring the drm_kms_helper_poll=n
kernel param.
Fix the above by calling only drm_kms_helper_hotplug_event() for a
delayed hotplug event if drm_kms_helper_hotplug_event=n, as was done
before d33a54e399.
Cc: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reported-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Fixes: d33a54e399 ("drm/probe_helper: sort out poll_running vs poll_enabled")
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240301152243.1670573-1-imre.deak@intel.com
All the components of YUV420 over DP are added. Therefore, let's mark the
connector property as true for DP connector when the DP type is not eDP
and when there is a CDM block available.
Changes in v3:
- Move setting the connector's ycbcr_420_allowed parameter so
that it is not dependent on if the dp_display is not eDP
Changes in v2:
- Check for if dp_catalog has a CDM block available instead of
checking if VSC SDP is allowed when setting the dp connector's
ycbcr_420_allowed parameter
Signed-off-by: Paloma Arellano <quic_parellan@quicinc.com>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/579628/
Link: https://lore.kernel.org/r/20240222194025.25329-20-quic_parellan@quicinc.com
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Adjust the encoder format programming in the case of video mode for DP
to accommodate CDM related changes.
Changes in v4:
- Remove hw_cdm check in dpu_encoder_needs_periph_flush()
- Remove hw_cdm check when getting the fmt_fourcc in
dpu_encoder_phys_vid_enable()
Changes in v2:
- Move timing engine programming to a separate patch from this
one
- Move update_pending_flush_periph() invocation completely to
this patch
- Change the logic of dpu_encoder_get_drm_fmt() so that it only
calls drm_mode_is_420_only() instead of doing additional
unnecessary checks
- Create new functions msm_dp_needs_periph_flush() and it's
supporting function dpu_encoder_needs_periph_flush() to check
if the mode is YUV420 and VSC SDP is enabled before doing a
peripheral flush
Signed-off-by: Paloma Arellano <quic_parellan@quicinc.com>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/579641/
Link: https://lore.kernel.org/r/20240222194025.25329-17-quic_parellan@quicinc.com
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
DP controller can be setup to operate in either SDP update flush mode or
peripheral flush mode based on the DP controller hardware version.
Starting in DP v1.2, the hardware documents require the use of
peripheral flush mode for SDP packets such as PPS OR VSC SDP packets.
In-line with this guidance, lets program the DP controller to use
peripheral flush mode starting DP v1.2
Changes in v4:
- Clear up that DP_MAINLINK_CTRL_FLUSH_MODE register requires
the use of bits [24:23]
- Modify macros DP_MAINLINK_FLUSH_MODE_UPDATE_SDP and
DP_MAINLINK_FLUSH_MODE_SDP_PERIPH_UPDATE to explicitly set
their values in the bits of DP_MAINLINK_CTRL_FLUSH_MODE_MASK
Changes in v3:
- Clear up that the DP_MAINLINK_FLUSH_MODE_SDE_PERIPH_UPDATE
macro is setting bits [24:23] to a value of 3
Changes in v2:
- Use the original dp_catalog_hw_revision() function to
correctly check the DP HW version
Signed-off-by: Paloma Arellano <quic_parellan@quicinc.com>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/579621/
Link: https://lore.kernel.org/r/20240222194025.25329-16-quic_parellan@quicinc.com
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Add support to pack and send the VSC SDP packet for DP. This therefore
allows the transmision of format information to the sinks which is
needed for YUV420 support over DP.
Changes in v5:
- Slightly modify use of drm_dp_vsc_sdp_pack()
- Remove dp_catalog NULL checks
- Modify dp_utils_pack_sdp_header() to more clearly pack the
header buffer
- Move dp_utils_pack_sdp_header() inside of
dp_catalog_panel_send_vsc_sdp to clearly show the relationship
between the header buffer and the vsc_sdp struct
- Due to the last point, remove the dp_utils_pack_vsc_sdp()
function and only call drm_dp_vsc_sdp_pack() in
dp_panel_setup_vsc_sdp_yuv_420()
Changes in v4:
- Remove struct msm_dp_sdp_with_parity
- Use dp_utils_pack_sdp_header() to pack the SDP header and
parity bytes into a buffer
- Use this buffer when writing the VSC SDP data in
dp_catalog_panel_send_vsc_sdp()
- Write to all of the MMSS_DP_GENERIC0 registers instead of just
the ones with non-zero values
Changes in v3:
- Create a new struct, msm_dp_sdp_with_parity, which holds the
packing information for VSC SDP
- Use drm_dp_vsc_sdp_pack() to pack the data into the new
msm_dp_sdp_with_parity struct instead of specifically packing
for YUV420 format
- Modify dp_catalog_panel_send_vsc_sdp() to send the VSC SDP
data using the new msm_dp_sdp_with_parity struct
Changes in v2:
- Rename GENERIC0_SDPSIZE macro to GENERIC0_SDPSIZE_VALID
- Remove dp_sdp from the dp_catalog struct since this data is
being allocated at the point used
- Create a new function in dp_utils to pack the VSC SDP data
into a buffer
- Create a new function that packs the SDP header bytes into a
buffer. This function is made generic so that it can be
utilized by dp_audio
header bytes into a buffer
- Create a new function in dp_utils that takes the packed buffer
and writes to the DP_GENERIC0_* registers
- Split the dp_catalog_panel_config_vsc_sdp() function into two
to disable/enable sending VSC SDP packets
- Check the DP HW version using the original useage of
dp_catalog_hw_revision() and correct the version checking
logic
- Rename dp_panel_setup_vsc_sdp() to
dp_panel_setup_vsc_sdp_yuv_420() to explicitly state that
currently VSC SDP is only being set up to support YUV420 modes
Signed-off-by: Paloma Arellano <quic_parellan@quicinc.com>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/579636/
Link: https://lore.kernel.org/r/20240222194025.25329-14-quic_parellan@quicinc.com
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
