drm/amdgpu: Get atomicOps info from Host for sriov setup
The AtomicOp Requester Enable bit is reserved in VFs and the PF value applies to all
associated VFs. so guest driver can not directly enable the atomicOps for VF, it
depends on PF to enable it. In current design, amdgpu driver will get the enabled
atomicOps bits through private pf2vf data
drm/amd/display: Link training retry fix for abort case
[Why]
If link training is aborted, it shall be retried if sink is present.
[How]
Check hpd status to find out whether sink is present or not. If sink is
present, then link training shall be tried again with same settings.
Otherwise, link training shall be aborted.
Qingqing Zhuo [Fri, 27 Aug 2021 10:58:38 +0000 (06:58 -0400)]
drm/amd/display: Fix unstable HPCP compliance on Chrome Barcelo
[Why]
Intermittently, there presents two occurrences of 0 stream
commits in a single HPD event. Current HDCP sequence does
not consider such scenerio, and will thus disable HDCP.
[How]
Add condition check to include stream remove and re-enable
case for HDCP enable.
Aric Cyr [Sun, 29 Aug 2021 22:44:22 +0000 (18:44 -0400)]
drm/amd/display: 3.2.152
* Correct degamma coefficients
* Optimize bandwidth on following fast update
* Fix multiple memory leaks reported by coverity
* Get backlight from PWM if DMCU is not initialized
Hersen Wu [Wed, 25 Aug 2021 20:27:47 +0000 (16:27 -0400)]
drm/amd/display: dsc mst 2 4K displays go dark with 2 lane HBR3
[Why]
call stack of amdgpu dsc mst pbn, slot num calculation is as below:
-compute_bpp_x16_from_target_bandwidth
-decide_dsc_target_bpp_x16
-setup_dsc_config
-dc_dsc_compute_bandwidth_range
-compute_mst_dsc_configs_for_link
-compute_mst_dsc_configs_for_state
from pbn -> dsc target bpp_x16
bpp_x16 is calulated by compute_bpp_x16_from_target_bandwidth.
Beside pixel clock and bpp, num_slices_h and bpp_increment_div
will also affect bpp_x16.
from dsc target bpp_x16 -> pbn
within dm_update_mst_vcpi_slots_for_dsc,
pbn = drm_dp_calc_pbn_mode(clock, bpp_x16, true);
bpp / 16 trunc digits after decimal point. This will cause calculation
delta. drm_dp_calc_pbn_mode does not have other informations,
like num_slices_h, bpp_increment_div. therefore, it does not do revese
calcuation properly from bpp_x16 to pbn.
pbn from drm_dp_calc_pbn_mode is less than pbn from
compute_mst_dsc_configs_for_state. This cause not enough mst slot
allocated to display. display could not visually light up.
[How]
pass pbn from compute_mst_dsc_configs_for_state to
dm_update_mst_vcpi_slots_for_dsc
Dale Zhao [Mon, 23 Aug 2021 08:44:37 +0000 (16:44 +0800)]
drm/amd/display: Refine condition of cursor visibility for pipe-split
[Why]
In some scenarios like fullscreen game, major plane is scaled. Then
if a upper layer owns the cursor, cursor is invisiable in the
majority of the screen.
[How]
Instead assuming upper plane handles cursor, summing up upper
split planes on the same layer. If whole upper plane covers current
half/whole pipe plane, disable cursor.
Qingqing Zhuo [Tue, 17 Aug 2021 20:29:07 +0000 (16:29 -0400)]
drm/amd/display: Apply w/a for hard hang on HPD
[Why]
HPD disable and enable sequences are not mutually exclusive
on Linux. For HPDs that spans under 1s (i.e. HPD low = 1s),
part of the disable sequence (specifically, a request to SMU
to lower refclk) could come right before the call to PHY
enablement, causing DMUB to access an irresponsive PHY
and thus a hard hang on the system.
[How]
Disable 48mhz refclk off when there is any HPD status in
connected state.
Jaehyun Chung [Tue, 24 Aug 2021 18:05:48 +0000 (14:05 -0400)]
drm/amd/display: Add regamma/degamma coefficients and set sRGB when TF is BT709
[Why]
In YUV case, need to set the input TF to sRGB instead of BT709,
even though the input TF type is distributed. SRGB was not
being used because pixel format was not being set in the
surface update sequence.
Also, we were using the same coefficients for degamma and
regamma formula, causing the cutoff point of the linear
section of the curve to be incorrect.
[How]
Set pixel format in the surface update sequence. Add separate
coefficient arrays for regamma and degamma.
Harry Wentland [Mon, 16 Aug 2021 19:57:12 +0000 (15:57 -0400)]
drm/amd/display: Get backlight from PWM if DMCU is not initialized
On Carrizo/Stoney systems we set backlight through panel_cntl, i.e.
directly via the PWM registers, if DMCU is not initialized. We
always read it back through ABM registers which leads to a
mismatch and forces atomic_commit to program the backlight
each time.
Instead make sure we use the same logic for backlight readback,
i.e. read it from panel_cntl if DMCU is not initialized.
We also need to remove some extraneous and incorrect calculations
at the end of dce_get_16_bit_backlight_from_pwm.
Jimmy Kizito [Tue, 10 Aug 2021 23:07:43 +0000 (19:07 -0400)]
drm/amd/display: Fix dynamic link encoder access.
[Why]
Assuming DIG link encoders are statically mapped to links can cause
system instability due to null pointer accesses.
[How]
- Add checks for non-null link encoder pointers before trying to access
them.
- When a hardware platform uses dynamic DIG assignment (i.e. resource
function 'link_encs_assign' defined) and a link supports flexible
mapping to DIGs, use the link_enc_cfg API to access the DIG assigned to
a link or stream.
Aurabindo Pillai [Fri, 13 Aug 2021 19:15:03 +0000 (15:15 -0400)]
drm/amd/display: Add flag to detect dpms force off during HPD
[Why] When a connector is unplugged, dpms is forced off so that some
connector allocations are cleared off. This is done outside the commit
sequence from the userspace. This causes HUBP blank. Due to the blank
hubp, a non blocking commit which queues flip will encounter a timeout
waiting for the flip_done because prior to writing the surface flip
address, hubp was in blank.
[How] Add a marker to DM's crtc state and use this field to indicate
whether dpms was forced off during an HPD. Check for this marker before
queuing the flip.
Josip Pavic [Thu, 19 Aug 2021 13:25:22 +0000 (09:25 -0400)]
drm/amd/display: unblock abm when odm is enabled only on configs that support it
[Why]
When ODM is enabled, ABM is blocked on dcn31 but unblocked on dcn30.
Since the dcn31 firmware is now able to handle ABM interop with ODM, it
is no longer necessary to block ABM when ODM is enabled.
Since the dcn30 firmware does not handle ABM interop with ODM, leaving
that combination unblocked can lead to one side of the screen appearing
brighter than the other.
[How]
When ODM is enabled, unblock abm on dcn31 and block it on dcn30
Wayne Lin [Fri, 23 Jul 2021 03:50:28 +0000 (11:50 +0800)]
drm/amd/display: Fork thread to offload work of hpd_rx_irq
[Why]
Currently, we will try to get dm.dc_lock in handle_hpd_rx_irq() when
link lost happened, which is risky and could cause deadlock.
e.g. If we are under procedure to enable MST streams and then monitor
happens to toggle short hpd to notify link lost, then
handle_hpd_rx_irq() will get blocked due to stream enabling flow has
dc_lock. However, under MST, enabling streams involves communication
with remote sinks which need to use handle_hpd_rx_irq() to handle
sideband messages. Thus, we have deadlock here.
[How]
Target is to have handle_hpd_rx_irq() finished as soon as possilble.
Hence we can react to interrupt quickly. Besides, we should avoid to
grabe dm.dc_lock within handle_hpd_rx_irq() to avoid deadlock situation.
Firstly, revert patches which introduced to use dm.dc_lock in
handle_hpd_rx_irq():
* commit ("drm/amd/display: NULL pointer error during ")
* commit ("drm/amd/display: Only one display lights up while using MST")
* commit ("drm/amd/display: take dc_lock in short pulse handler only")
Instead, create work to handle irq events which needs dm.dc_lock.
Besides:
* Create struct hpd_rx_irq_offload_work_queue for each link to handle
its short hpd events
* Avoid to handle link lost/ automated test if the link is disconnected
* Defer dc_lock needed works in dc_link_handle_hpd_rx_irq(). This
function should just handle simple stuff for us (e.g. DPCD R/W).
However, deferred works should still be handled by the order that
dc_link_handle_hpd_rx_irq() used to be.
* Change function name dm_handle_hpd_rx_irq() to
dm_handle_mst_sideband_msg() to be more specific
Wayne Lin [Fri, 13 Aug 2021 07:54:47 +0000 (15:54 +0800)]
drm/amd/display: Add option to defer works of hpd_rx_irq
[Why & How]
Due to some code flow constraints, we need to defer dc_lock needed works
from dc_link_handle_hpd_rx_irq(). Thus, do following changes:
* Change allow_hpd_rx_irq() from static to public
* Change handle_automated_test() from static to public
* Extract link lost handling flow out from dc_link_handle_hpd_rx_irq()
and put those into a new function dc_link_dp_handle_link_loss()
* Add one option parameter to decide whether defer works within
dc_link_handle_hpd_rx_irq()
Wenjing Liu [Fri, 13 Aug 2021 23:49:43 +0000 (19:49 -0400)]
drm/amd/display: update conditions to do dfp cap ext validation
[why]
According to DP specs dfp cap ext validation is only for branch
device withou 128b/132b channel coding support and
downstream of the branch device doesn't have compression.
Therefore we are adding conditions to only do dfp cap
extension validation for branch devcie supporting 8b/10b
channel coding only and it has no DSC passthrough capability.
Wenjing Liu [Mon, 9 Aug 2021 21:00:12 +0000 (17:00 -0400)]
drm/amd/display: move bpp range decision in decide dsc bw range function
[why]
Before get dsc bw range is used to compute DSC bw range
based on the given fixed bpp min/max input.
The new change will merge any specs, signal, timing specific
bpp range decision into this function. So the function needs to make
a decision with all aspects considered.
Felix Kuehling [Tue, 31 Aug 2021 21:42:15 +0000 (17:42 -0400)]
drm/amdkfd: make needs_pcie_atomics FW-version dependent
On some GPUs the PCIe atomic requirement for KFD depends on the MEC
firmware version. Add a firmware version check for this. The minimum
firmware version that works without atomics can be updated in the
device_info structure for each GPU type.
Move PCIe atomic detection from kgd2kfd_probe into kgd2kfd_device_init
because the MEC firmware is not loaded yet at the probe stage.
Lang Yu [Fri, 27 Aug 2021 07:20:51 +0000 (15:20 +0800)]
drm/amdgpu: add manual sclk/vddc setting support for cyan skilfish(v3)
Add manual sclk/vddc setting supoort via pp_od_clk_voltage sysfs
to maintain consistency with other asics. As cyan skillfish doesn't
support DPM, there is only a single frequency and voltage to adjust.
v2: maintain consistency and add command guide.
v3: adjust user settings storage and coding style.
Command guide:
echo vc point sclk vddc > pp_od_clk_voltage
"vc" - sclk voltage curve
"point" - must be 0
"sclk" - target value of sclk(MHz), should be in safe range
"vddc" - target value of vddc(mV), a 6.25(mV) stepping is
recommended and should be in safe range (the real
vddc is an approximation of target value)
echo c > pp_od_clk_voltage
"c" - commit the changes of sclk and vddc, only after
the commit command, the target values set by "vc"
command will take effect
echo r > pp_od_clk_voltage
"r" - reset sclk and vddc to default value, a subsequent
commit command is needed to take effect
Example:
1) Check default sclk and vddc
$ cat pp_od_clk_voltage
OD_SCLK:
0: 1800Mhz *
OD_VDDC:
0: 862mV *
OD_RANGE:
SCLK: 1000Mhz 2000Mhz
VDDC: 700mV 1129mV
2) Set sclk to 1500MHz and vddc to 700mV
$ echo vc 0 1500 700 > pp_od_clk_voltage
$ echo c > pp_od_clk_voltage
$ cat pp_od_clk_voltage
OD_SCLK:
0: 1500Mhz *
OD_VDDC:
0: 693mV *
OD_RANGE:
SCLK: 1000Mhz 2000Mhz
VDDC: 700mV 1129mV
3) Reset sclk and vddc to default
$ echo r > pp_od_clk_voltage
$ echo c > pp_od_clk_voltage
$ cat pp_od_clk_voltage
OD_SCLK:
0: 1800Mhz *
OD_VDDC:
0: 874mV *
OD_RANGE:
SCLK: 1000Mhz 2000Mhz
VDDC: 700mV 1129mV
NOTE:
We don't specify an explicit safe range, you can set any values
between min and max at your own risk. Enjoy!
Nirmoy Das [Thu, 2 Sep 2021 13:31:07 +0000 (15:31 +0200)]
drm/amdgpu: cleanup debugfs for amdgpu rings
Use debugfs_create_file_size API for creating ring debugfs, and as its a
NULL returning API, change the return type for amdgpu_debugfs_ring_init
API as well. Also cleanup surrounding code.
drm/amd/display: make configure_lttpr_mode_transparent and configure_lttpr_mode_non_transparent static
This symbols is not used outside of dc_link_dp.c, so marks it static.
Fix the following sparse warning:
drivers/gpu/drm/amd/amdgpu/../display/dc/core/dc_link_dp.c:1766:16:
warning: symbol 'configure_lttpr_mode_non_transparent' was not declared.
Should it be static?
drivers/gpu/drm/amd/amdgpu/../display/dc/core/dc_link_dp.c:1755:16:
warning: symbol 'configure_lttpr_mode_transparent' was not declared.
Should it be static?
Colin Ian King [Tue, 7 Sep 2021 11:09:13 +0000 (12:09 +0100)]
drm/radeon/ci_dpm: Remove redundant initialization of variables hi_sidd, lo_sidd
The variables hi_sidd and lo_sidd are being initialized with a values
that are never read, they are being updated later on. The assignments
are redundant and can be removed.
In amdgpu_dm_atomic_check, dc_validate_global_state is called. On
failure this logs a warning to the kernel journal. However warnings
shouldn't be used for atomic test-only commit failures: user-space
might be perfoming a lot of atomic test-only commits to find the
best hardware configuration.
Downgrade the log to a regular DRM atomic message. While at it, use
the new device-aware logging infrastructure.
This fixes error messages in the kernel when running gamescope [1].
Len Baker [Sat, 4 Sep 2021 15:41:06 +0000 (17:41 +0200)]
drm/radeon: Prefer kcalloc over open coded arithmetic
As noted in the "Deprecated Interfaces, Language Features, Attributes,
and Conventions" documentation [1], size calculations (especially
multiplication) should not be performed in memory allocator (or similar)
function arguments due to the risk of them overflowing. This could lead
to values wrapping around and a smaller allocation being made than the
caller was expecting. Using those allocations could lead to linear
overflows of heap memory and other misbehaviors.
So, refactor the code a bit to use the purpose specific kcalloc()
function instead of the calculated size argument in the kzalloc()
function.
Christian König [Tue, 7 Sep 2021 07:37:52 +0000 (09:37 +0200)]
drm/amdgpu: fix use after free during BO move
The memory backing old_mem is already freed at that point, move the
check a bit more up.
Signed-off-by: Christian König <[email protected]> Fixes: bfa3357ef9ab ("drm/ttm: allocate resource object instead of embedding it v2")
Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/1699 Acked-by: Nirmoy Das <[email protected]> Reviewed-by: Michel Dänzer <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
Ernst Sjöstrand [Thu, 2 Sep 2021 07:50:27 +0000 (09:50 +0200)]
drm/amd/amdgpu: Increase HWIP_MAX_INSTANCE to 10
Seems like newer cards can have even more instances now.
Found by UBSAN: array-index-out-of-bounds in
drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c:318:29
index 8 is out of range for type 'uint32_t *[8]'
Tom St Denis [Fri, 20 Aug 2021 18:03:12 +0000 (14:03 -0400)]
drm/amd/amdgpu: New debugfs interface for MMIO registers (v5)
This new debugfs interface uses an IOCTL interface in order to pass
along state information like SRBM and GRBM bank switching. This
new interface also allows a full 32-bit MMIO address range which
the previous didn't. With this new design we have room to grow
the flexibility of the file as need be.
(v2): Move read/write to .read/.write, fix style, add comment
for IOCTL data structure
(v3): C style comments
(v4): use u32 in struct and remove offset variable
(v5): Drop flag clearing in op function, use 0xFFFFFFFF for broadcast
instead of 0x3FF, use mutex for op/ioctl.
Nirmoy Das [Wed, 25 Aug 2021 15:11:31 +0000 (17:11 +0200)]
drm/amdgpu: detach ring priority from gfx priority
Currently AMDGPU_RING_PRIO_MAX is redefinition of a
max gfx hwip priority, this won't work well when we will
have a hwip with different set of priorities than gfx.
Also, HW ring priorities are different from ring priorities.
Create a global enum for ring priority levels which each
HWIP can use to define its own priority levels.
Nirmoy Das [Tue, 24 Aug 2021 18:39:18 +0000 (20:39 +0200)]
drm/amdgpu: rework context priority handling
To get a hardware queue priority for a context, we are currently
mapping AMDGPU_CTX_PRIORITY_* to DRM_SCHED_PRIORITY_* and then
to hardware queue priority, which is not the right way to do that
as DRM_SCHED_PRIORITY_* is software scheduler's priority and it is
independent from a hardware queue priority.
Use userspace provided context priority, AMDGPU_CTX_PRIORITY_* to
map a context to proper hardware queue priority.
Aric Cyr [Sun, 15 Aug 2021 00:28:55 +0000 (20:28 -0400)]
drm/amd/display: 3.2.150
This version brings along following fixes:
- FW promotion 0.0.80
- Add missing ABM register offsets
- Fix in swizzle mode mapping
- Emulated sink support for freesync
- Improvoments in max target bpp
Aurabindo Pillai [Wed, 11 Aug 2021 18:38:44 +0000 (14:38 -0400)]
drm/amd/display: Add emulated sink support for updating FS
[Why]
When forced modes are used during certain IGT tests,
without a real connector, dc_sink would be null when
standard modes are added by the driver. Calling the
function to update freesync capabilities at this
point will result in an error being printed
[How]
Use emulated sink when available. If both the normal
and emulated sink are not available, set all freesync
parameters to 0.
Josip Pavic [Mon, 9 Aug 2021 19:13:49 +0000 (15:13 -0400)]
drm/amd/display: add missing ABM register offsets
[Why]
Some ABM registers don't exist on DCN 3.01, so are
missing from its register offset list. However,
this list was copied to later versions of DCN that
do have these registers. As a result, they're
inaccessible from the driver on those DCN versions
even though they exist.
[How]
Add the missing ABM register offsets to DCN 3.02+
Guchun Chen [Fri, 27 Aug 2021 10:31:41 +0000 (18:31 +0800)]
drm/amdgpu: stop scheduler when calling hw_fini (v2)
This gurantees no more work on the ring can be submitted
to hardware in suspend/resume case, otherwise a potential
race will occur and the ring will get no chance to stay
empty before suspend.
v2: Call drm_sched_resubmit_job before drm_sched_start to
restart jobs from the pending list.
Michael Strauss [Mon, 16 Aug 2021 20:15:36 +0000 (16:15 -0400)]
drm/amd/display: Initialize lt_settings on instantiation
[WHY]
lt_settings' pointers remain uninitialized but nonzero if display fails
to light up with no DPCD/EDID info populated, leading to a hang on access
projects / linux.git / log