drm/amd/display: add -msse2 to prevent Clang from emitting libcalls to undefined SW FP routines
arch/x86/Makefile disables SSE and SSE2 for the whole kernel. The
AMDGPU drivers modified in this patch re-enable SSE but not SSE2. Turn
on SSE2 to support emitting double precision floating point instructions
rather than calls to non-existent (usually available from gcc_s or
compiler_rt) floating point helper routines.
hersen wu [Fri, 11 Jan 2019 17:43:20 +0000 (12:43 -0500)]
drm/amd/display: Connect dig_fe to otg directly instead of calling bios
[Why] After call bios table crtc_source_select, dal will program fmt
again. The bios table program dig_source_select and other fmt register
for bios usage which is redundancy and uncessary.
Eric Yang [Fri, 11 Jan 2019 22:09:18 +0000 (17:09 -0500)]
drm/amd/display: take dpms_off into account for edp turn off logic
[why]
Previously we incorrectly skipped backlight control when stream is
present but dpms_off = true. This causes backlight to remain on in
the we boot up or resume into a external display only configuration
where VBIOS posted on the eDP.
[How]
Add dpms_off into the condition for edp need to turn off.
Krunoslav Kovac [Tue, 8 Jan 2019 23:32:34 +0000 (18:32 -0500)]
drm/amd/display: Default to linear output gamma
[Why] Our output TF calculation doesn't work if no user-specified gamma
correction. Normally, user provides this, but driver sohuld just assume
default (linear) gamma otherwise.
[How] Remove output TF dependency on user gamma being provided.
David Francis [Wed, 19 Dec 2018 15:45:16 +0000 (10:45 -0500)]
drm/amd/display: Perform plane updates only when needed
[Why]
Our old logic: if pageflip, update freesync and plane address.
Otherwise, update everything.
This over-updated on non-pageflip cases, and it failed to
update if pageflip and non-pageflip changes occurred on
the same commit
[How]
Update flip_addrs on pageflips.
Update scaling_info when it changes.
Update color fields on color changes.
Updates plane_info always because we don't have a good way of
knowing when it needs to be updated.
Unfortunately, this means that every stream commit involves two
calls into DC. In particular, on pageflips there is a second,
pointless update that changes nothing but costs several
microseconds (about a 50% increase in time taken). The update is
fast, but there are comparisons and some useless programming.
David Francis [Tue, 18 Dec 2018 20:30:19 +0000 (15:30 -0500)]
drm/amd/display: Simplify underscan and ABM commit
[Why]
Underscan and ABM are connector properties but require updates
to DC stream state. Previously, on updates to these properties
the affected stream and all its planes were committed.
This is unnecessary; only a few fields on the stream need
to be changed.
[How]
If scaling or ABM have changed, change the stream and
create a stream update with those changes. Call
DC with only those fields.
David Francis [Thu, 3 Jan 2019 19:38:06 +0000 (14:38 -0500)]
drm/amd/display: Let updates with no scaling changes be fast
[Why]
DC was assuming that any surface_update->scaling_info
meant the update was at least medium. However, if nothing
has changed there is no scaling to program, so there is
no problem with the update being fast
[How]
If every update flag is not set, the update is fast
Eryk Brol [Fri, 7 Dec 2018 15:25:13 +0000 (10:25 -0500)]
drm/amd/display: Introduce DC VM interface
Virtual memory allows display to support flipping to surfaces which
are not allocated contiguously in memory with physical addresses,
instead a 1-4 level page table is used.
This is beneficial because it allows the scattering of large surfaces
to improve memory efficiency and security.
Eryk Brol [Fri, 4 Jan 2019 15:29:00 +0000 (10:29 -0500)]
drm/amd/display: Restructure DCN10 hubbub
[Why]
Change DCN10 hubbub to use hubbub as a base and allow all future DCN
hubbubs to do the same instead of using DCN10_hubbub. This increases
readability and doesn't require future hubbubs to inherit anything
other than the base hubbub struct.
[How]
Create separate DCN10_hubbub struct which uses the hubbub struct as
a base.
hersen wu [Fri, 11 Jan 2019 15:39:30 +0000 (10:39 -0500)]
drm/amd/display: fix eDP fast bootup for pre-raven asic
[Why]
For fastboot, Bios will light up eDP before SW driver is loaded. SW
driver will check if eDP is lit by bios through reading the
BIOS_SCRATCH_3 register. If lit, SW driver will not power down eDP
power and phy to save time.
Definition of BIOS_SCRATCH_3 are missing for pre-raven asic. This
causes eDP fast boot to not work property. For some eDP panels, even
if dp tx sends NoVideoStream_flag =1 and dpcd 0x600=2, eDP rx may not
handle properly. This may cause a short flash on screen.
[How] Add definition of BIOS_SCRATCH_3 for all asic
Expedite job deletion from ring mirror list to the HW fence signal
callback instead from finish_work, together with waiting for all
such fences to signal in drm_sched_stop we garantee that
already signaled job will not be processed twice.
Remove the sched finish fence callback and just submit finish_work
directly from the HW fence callback.
v2: Fix comments.
v3: Attach hw fence cb to sched_job
v5: Rebase
Decauple sched threads stop and start and ring mirror
list handling from the policy of what to do about the
guilty jobs.
When stoppping the sched thread and detaching sched fences
from non signaled HW fenes wait for all signaled HW fences
to complete before rerunning the jobs.
v2: Fix resubmission of guilty job into HW after refactoring.
v4:
Full restart for all the jobs, not only from guilty ring.
Extract karma increase into standalone function.
v5:
Rework waiting for signaled jobs without relying on the job
struct itself as those might already be freed for non 'guilty'
job's schedulers.
Expose karma increase to drivers.
v6:
Use list_for_each_entry_safe_continue and drm_sched_process_job
in case fence already signaled.
Call drm_sched_increase_karma only once for amdgpu and add documentation.
Replace kzalloc() function with its 2-factor argument form, kcalloc().
This patch replaces cases of:
kzalloc(a * b, gfp)
with:
kcalloc(a, b, gfp)
Also, improve the coding style and the use of sizeof during
allocation by changing sizeof(struct dc_surface_update) and
sizeof(struct dc_plane_state) to sizeof(*updates) and
sizeof(*surfaces), correspondingly.
This code was detected with the help of Coccinelle.
Aaron Liu [Mon, 14 Jan 2019 08:08:32 +0000 (16:08 +0800)]
drm/amdgpu/psp: ignore psp response status
In some cases, psp response status is not 0 even there is no
problem while the command is submitted. Some version of PSP FW
doesn't write 0 to that field.
So here we would like to only print a warning instead of an error
during psp initialization to avoid breaking hw_init and it doesn't
return -EINVAL.
Oak Zeng [Mon, 14 Jan 2019 22:32:53 +0000 (16:32 -0600)]
drm/amdgpu: Setting doorbell range registers earlier
HW doorbell writing routing policy: writing to doorbell
not in SDMA/IH/MM/ACV doorbell range will be routed to CP.
So CP doorbell routing depends on doorbell range setting
of above blocks. Setting doorbell range of above blocks
earlier (soc15_common_hw_init) to make sure CP doorbell
writing be routed to CP block.
Oak Zeng [Tue, 18 Dec 2018 02:03:45 +0000 (20:03 -0600)]
drm/amdgpu: Fix sdma doorbell range setting
Different ASIC has different SDMA queue number so
different SDMA doorbell range. Introduce an extra
parameter to sdma_doorbell_range function and set
sdma doorbell range correctly.
drm/amd/display: Skip create new stream if crtc state doesn't change
Need to check if crtc state is changed so that mode set is
required before trying to create new stream.
It deals with the MST hotplug use case when plug back to the
same connector where the failure to create new stream for the
inactive crtc on the old connector.
Krunoslav Kovac [Wed, 2 Jan 2019 19:12:53 +0000 (14:12 -0500)]
drm/amd/display: Check for NULL when creating gamma struct
[Wjy&How] Some stress test is causing unexpected memory allocation
failure. This prevents null dereference but there will likely be problems
later, hard to gracefully handle memalloc fail for critical objects.
Eric Yang [Mon, 17 Dec 2018 23:28:59 +0000 (18:28 -0500)]
drm/amd/display: add workaround for 4k video underflow
[Why]
On DCN1, there is an issue where on high BW config on single channel
systems, underflow will be observed if DCC is disabled. This issue
can be observed on several use cases. For this particular case,
it is observed when playing 4k video on 4k desktop with video downscaled
to a certain size.
[How]
Block MPO for this particular case, this will prevent extra BW consumed
from downscaling, working around the underflow.
hersen wu [Wed, 28 Nov 2018 21:57:56 +0000 (16:57 -0500)]
drm/amd/display: dal-pplib interface refactor dal part
[WHY] clarify dal input parameters to pplib interface, remove
un-used parameters. dal knows exactly which parameters needed
and their effects at pplib and smu sides.
current dal sequence for dcn1_update_clock to pplib:
1.smu10_display_clock_voltage_request for dcefclk
2.smu10_display_clock_voltage_request for fclk
3.phm_store_dal_configuration_data {
set_min_deep_sleep_dcfclk
set_active_display_count
store_cc6_data --- this data never be referenced
new sequence will be:
1. set_display_count --- need add new pplib interface
2. set_min_deep_sleep_dcfclk -- new pplib interface
3. set_hard_min_dcfclk_by_freq
4. set_hard_min_fclk_by_freq
after this code refactor, smu10_display_clock_voltage_request,
phm_store_dal_configuration_data will not be needed for rv.
David Francis [Fri, 30 Nov 2018 15:32:01 +0000 (10:32 -0500)]
drm/amd/display: Change from aux_engine to dce_aux
[Why]
The aux_engine struct is needlessly complex and
is defined multiple times. It contains function pointers
that each have only one version and are called only from
inside dce_aux.
[How]
Replace aux_engine with a new struct called dce_aux.
Remove all function pointers and call functions directly.
Remove unused functions
David Francis [Thu, 29 Nov 2018 19:36:25 +0000 (14:36 -0500)]
drm/amd/display: Switch ddc to new aux interface
[Why]
The old aux interface goes through i2caux and the aux_engine
and engine function pointers. The multiple layers of indirection
make it hard to tell waht is happening. The aux algorithm
does not need to be this complicated: attempt to submit the
request. If you get an ack (reply = 0), stop. Otherwise,
retry, up to 7 times.
[How]
Add a new helper function in dce_aux that performs aux retries
Move the plumbing of the aux calling code into dce_aux
Add functions in ddc that redirect directly to dce_aux
David Francis [Thu, 29 Nov 2018 18:40:03 +0000 (13:40 -0500)]
drm/amd/display: Shift dc link aux to aux_payload
[Why]
aux_payload should be the struct used inside dc to start
aux transactions. This will allow the old aux interface
to be seamlessly replaced.
[How]
Add three fields to aux_payload: reply, mot, defer_delay
This will mean that aux_payload has all data required
to submit a request. Shift dc_link to use this struct
Martin Tsai [Thu, 13 Dec 2018 04:25:19 +0000 (12:25 +0800)]
drm/amd/display: Redefine DMCU_SCRATCH to identify DMCU state
[why]
To resume system before entering S0i3 completely will cause PSP not
reload DMCU FW since there is not HW power state change.
In this case, driver cannot get correct DMCU version from IRAM
since driver override it and DMCU didn't reload to update it.
It makes driver return false in dcn10_dmcu_init().
[how]
1.To redefine DMCU_SCRATCH to identify different DMCU state.
2.To reserve IRAM 0xF0~0xFF write by DMCU only.
3.To remove dcn10_get_dmcu_state
[Why]
1-Test equipment does not reset test automation flag after completing
current test causing the next test to fail.
2.When test request is empty, we shouldn't ack the test.
[How]
1-Driver should always clear test equipment automation test request
when request is completed.
2-Driver should clear test equipement test automation if driver does
not complete the request.
Joshua Aberback [Mon, 10 Dec 2018 18:46:34 +0000 (13:46 -0500)]
drm/amd/display: Fix missing hwss function for dce
[Why]
The driver will crash on dce hardware due to a null function pointer.
[How]
- bring back "program_csc_matrix" functionality as "program_output_csc" for
dce110
- dce110 doesn't use the "opp_id" parameter, so use 0 when there's no hubp
drm/amd/display: Only get the connector state for VRR when toggled
[Why]
This fixes a stuttering issue that occurs when moving a hardware cursor
when VRR is enabled.
Previously when VRR is enabled atomic check will grab the connector
state for every atomic update. This has to lock the connector in order
to do so. The locking is bad enough by itself for performance, but
it gets worse with what we do just below that - add all the planes
for the CRTC to the commit.
This prevents the cursor fast path from working - there's more than one
plane now. With state->allow_modeset = true on top of this, it also
adds and removes all the planes from the DC context triggering a full
(very slow) update in DC.
[How]
We need the connector state to get the VRR min/max capbilities, but we
only need them when there's a CRTC mode change or when VRR is toggled.
The condition has been updated accordingly.
Fixes: 3cc22f281318 ("drm/amdgpu: Set FreeSync state using drm VRR properties") Signed-off-by: Nicholas Kazlauskas <[email protected]> Reviewed-by: Leo Li <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
projects / linux.git / log