Debug VRAM access through SDMA has several broken parts resulting in
silent MMIO fallback.
BO kernel creation takes the location of the cpu addr pointer, not
the pointer itself for address kmap.
drm_dev_enter return true on success so change access check.
The source BO is reserved but not pinned so find the address using the
cursor offset relative to its memory domain start.
Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
That's just a leftover from old radeon days and was preventing CS and GART
bindings before the hardware was initialized. But nowdays that is
perfectly valid.
The only thing we need to warn about are GART binding before the table
is even allocated.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Guchun Chen <guchun.chen@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This patch will modify a pair of functions for pcie port wreg/rreg.
AMD GPU have had an independent NBIO block from SOC15 arch.
If the dirver wants to read/write the address space of the pcie devices,
it has to go through the NBIO block.
This patch will move the pcie port wreg/rreg functions to
"amdgpu_device.c", so that to reuse the functions on the
future GPU ASICs.
Signed-off-by: Xiaojian Du <Xiaojian.Du@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This patch will add vram check function for GMC block.
It will write pattern data to the vram and then read back from the vram,
so that to verify the work status of vram.
This patch will cover gmc v6/7/8/9/10.
Signed-off-by: Xiaojian Du <Xiaojian.Du@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Lots of machines these days seem to have a crappy type1 DP dual
mode adaptor chip slapped onto the motherboard. Based on the
DP dual mode spec we currently limit those to 165MHz max TMDS
clock.
Windows OTOH ignores DP dual mode adaptors when the VBT
indicates that the port is not actually DP++, so we can
perhaps assume that the vendors did intend that the 165MHz
clock limit doesn't apply here. Though it would be much
nicer if they actually declared an explicit limit through
VBT, but that doesn't seem to be happening either.
So in order to match Windows behaviour let's ignore the
DP dual mode adaptor's TMDS clock limit for ports that
don't look like DP++ in VBT.
Unfortunately many older VBTs misdelcare their DP++ ports
as just HDMI (eg. ILK Dell Latitude E5410) or DP (eg. SNB
Lenovo ThinkPad X220). So we can't really do this universally
without risking black screens. I suppose a sensible cutoff
is HSW+ since that's when 4k became a thing and one might
assume that the machines have been tested to work with higher
TMDS clock rates.
v2: s/IS_BROADWELL/IS_HASWELL/
Acked-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211222161738.12478-1-ville.syrjala@linux.intel.com
Replace the DEVICE_TYPE_DP_DUAL_MODE_BITS stuff with just
a DP+HDMI check. The rest of the bits shouldn't really
matter anyway.
The slight change in behaviour here is that now we do look at
the DEVICE_TYPE_NOT_HDMI_OUTPUT bit (via
intel_bios_encoder_supports_hdmi()) when we previously ignored it.
The one platform we know that has problems with that bit is VLV.
But IIRC the problem was always that buggy VBTs basically never
set that bit. So that should be OK since all it would do is make
all DVI ports look like HDMI ports instead. Also can't imagine
there are many VLV machines with actual DVI ports in existence.
We still keep the rest of the dvo_port/aux_ch checks as we
can't trust that DP+HDMI device type equals DP++ due to
buggy VBTs.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211217155403.31477-6-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Now that we parse the DDI port info from the VBT on all g4x+ platforms
we can throw out all the old codepaths in intel_bios_is_port_present(),
intel_bios_is_port_edp() and intel_bios_is_port_dp_dual_mode(). None
of these should be called on pre-g4x platforms.
For good measure throw in a WARN into intel_bios_is_port_present()
should someone get the urge to call it on older platforms. The
other two functions are specific to HDMI and DP so should not need
any protection as those encoder types don't even exist on older
platforms.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211217155403.31477-5-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Extend the vbt.ports[] stuff for all g4x+ platforms. We do need
to drop the version check as some elk/ctg machines may have VBTs
older than that. The oldest I know is an elk with version 142.
But the child device stuff has had the correct size since at
least version 125 (observed on my sdg), so from that angle this
should be totally safe.
This does couple of things:
- Start using the aux_ch/ddc_pin from VBT instead of just the
hardcoded defaults. Hopefully there are no VBTs with entirely
bogus information here.
- Start using i915->vbt.ports[] for intel_bios_is_port_dp_dual_mode().
Should be fine as the logic doesn't actually change.
- Start using i915->vbt.ports[] for intel_bios_is_port_edp().
The old codepath only looks at the DP DVO ports, the new codepath
looks at both DP and HDMI DVO ports. In principle that should not
matter. We also stop looking at some of the other device type bits
(eg. LVDS,MIPI,ANALOG,etc.). Hopefully no VBT is broken enough that
it sets up totally conflicting device type bits (eg. LVDS+eDP at the
same time). We also lose the "g4x->no eDP ever" hardcoding (shouldn't
be hard to re-introduce that into eg. sanitize_device_type() if needed).
Lightly smoke tested on a set of machines (one of ctg,ilk,snb,ivb each)
with both DP and HDMI (DP++). Everything still worked as it should.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211217155403.31477-4-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
CHV is currently straddling the divide by using parse_ddi_ports() stuff
for aux_ch/ddc_pin but going through all old codepaths for the rest
(intel_bios_is_port_present(), intel_bios_is_port_edp(),
intel_bios_is_port_dp_dual_mode()). Let's switch over full and use
i915->vbt.ports[] for the rest of the stuff.
dvo_port_to_port() doesn't know about DSI so we won't get into
any kind of "is port B HDMI or DSI or both?" conundrum, which
could otherwise happen on VLV/CHV due to DSI ports living in a
separate world from the other digital ports.
Including Jani's detailed analysis here for posterity:
"We stop checking for port A for CHV in intel_bios_is_port_present(), but
it's a warn and I don't recall any bug reports, so probably fine. We
could add a check in parse_ddi_port(), but meh.
Ditto for intel_bios_is_port_dp_dual_mode(), except it doesn't have a
warn.
The eDP check in intel_bios_is_port_edp() becomes slightly more
relaxed. Both the old and new check require these to be set:
- DEVICE_TYPE_DISPLAYPORT_OUTPUT
- DEVICE_TYPE_INTERNAL_CONNECTOR.
The old code also required these to be unset:
- DEVICE_TYPE_MIPI_OUTPUT
- DEVICE_TYPE_COMPOSITE_OUTPUT
- DEVICE_TYPE_DUAL_CHANNEL
- DEVICE_TYPE_LVDS_SIGNALING
- DEVICE_TYPE_TMDS_DVI_SIGNALING
- DEVICE_TYPE_VIDEO_SIGNALING
- DEVICE_TYPE_ANALOG_OUTPUT
It's possible we've added these just as a sanity check for broken VBTs
more than anything. I guess I'd see if actual problems arise.
Bottom line, I think the functional changes matter only for VBTs with
bogus data."
I agree that it should work assuming the VBT isn't totally insane.
Modern windows drivers also don't seem to check any of those
additional device type bits, which may or may not matter for older
devices (no idea what some old driver versions are checking).
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211217155403.31477-3-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
On some servers with MGA G200_SE_A (rev 42), booting with Legacy BIOS,
the hardware hangs when using kdump and kexec into the kdump kernel.
This happens when the uncompress code tries to write "Decompressing Linux"
to the VGA Console.
It can be reproduced by writing to the VGA console (0xB8000) after
booting to graphic mode, it generates the following error:
kernel:NMI: PCI system error (SERR) for reason a0 on CPU 0.
kernel:Dazed and confused, but trying to continue
The root cause is the configuration of the MGA GCTL6 register
According to the GCTL6 register documentation:
bit 0 is gcgrmode:
0: Enables alpha mode, and the character generator addressing system is
activated.
1: Enables graphics mode, and the character addressing system is not
used.
bit 1 is chainodd even:
0: The A0 signal of the memory address bus is used during system memory
addressing.
1: Allows A0 to be replaced by either the A16 signal of the system
address (ifmemmapsl is ‘00’), or by the hpgoddev (MISC<5>, odd/even
page select) field, described on page 3-294).
bit 3-2 are memmapsl:
Memory map select bits 1 and 0. VGA.
These bits select where the video memory is mapped, as shown below:
00 => A0000h - BFFFFh
01 => A0000h - AFFFFh
10 => B0000h - B7FFFh
11 => B8000h - BFFFFh
bit 7-4 are reserved.
Current code set it to 0x05 => memmapsl to b01 => 0xa0000 (graphic mode)
But on x86, the VGA console is at 0xb8000 (text mode)
In arch/x86/boot/compressed/misc.c debug strings are written to 0xb8000
As the driver doesn't use this mapping at 0xa0000, it is safe to set it to
0xb8000 instead, to avoid kernel hang on G200_SE_A rev42, with kexec/kdump.
Thus changing the value 0x05 to 0x0d
Signed-off-by: Jocelyn Falempe <jfalempe@redhat.com>
Reviewed-by: Javier Martinez Canillas <javierm@redhat.com>
Acked-by: Lyude Paul <lyude@redhat.com>
Cc: stable@vger.kernel.org
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Link: https://patchwork.freedesktop.org/patch/msgid/20220119102905.1194787-1-jfalempe@redhat.com
Move the base i915 buddy allocator code into drm
- Move i915_buddy.h to include/drm
- Move i915_buddy.c to drm root folder
- Rename "i915" string with "drm" string wherever applicable
- Rename "I915" string with "DRM" string wherever applicable
- Fix header file dependencies
- Fix alignment issues
- add Makefile support for drm buddy
- export functions and write kerneldoc description
- Remove i915 selftest config check condition as buddy selftest
will be moved to drm selftest folder
cleanup i915 buddy references in i915 driver module
and replace with drm buddy
v2:
- include header file in alphabetical order(Thomas)
- merged changes listed in the body section into a single patch
to keep the build intact(Christian, Jani)
v3:
- make drm buddy a separate module(Thomas, Christian)
v4:
- Fix build error reported by kernel test robot <lkp@intel.com>
- removed i915 buddy selftest from i915_mock_selftests.h to
avoid build error
- removed selftests/i915_buddy.c file as we create a new set of
buddy test cases in drm/selftests folder
v5:
- Fix merge conflict issue
v6:
- replace drm_buddy_mm structure name as drm_buddy(Thomas, Christian)
- replace drm_buddy_alloc() function name as drm_buddy_alloc_blocks()
(Thomas)
- replace drm_buddy_free() function name as drm_buddy_free_block()
(Thomas)
- export drm_buddy_free_block() function
- fix multiple instances of KMEM_CACHE() entry
v7:
- fix warnings reported by kernel test robot <lkp@intel.com>
- modify the license(Christian)
v8:
- fix warnings reported by kernel test robot <lkp@intel.com>
Signed-off-by: Arunpravin <Arunpravin.PaneerSelvam@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220118104504.2349-1-Arunpravin.PaneerSelvam@amd.com
Signed-off-by: Christian König <christian.koenig@amd.com>
Add a cancel request selftest that results in an engine reset to cancel
the request as it is non-preemptable. Also insert a NOP request after
the cancelled request and confirm that it completes successfully.
v2:
(Tvrtko)
- Skip test if preemption timeout compiled out
- Skip test if engine reset isn't supported
- Update debug prints to be more descriptive
v3:
- Add comment explaining test
v4:
(John Harrison)
- Fix typos in comment explaining test
- goto out_rq is NOP creation fails
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: John Harrison <John.C.Harrison@Intel.com>
Signed-off-by: John Harrison <John.C.Harrison@Intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220113181351.21296-2-matthew.brost@intel.com
The return value was never initialized so the cleanup code executed when
it isn't even necessary.
Just add proper error handling.
Fixes: ab50cb9df8 ("drm/radeon/radeon_kms: Fix a NULL pointer dereference in radeon_driver_open_kms()")
Signed-off-by: Christian König <christian.koenig@amd.com>
Tested-by: Jan Stancek <jstancek@redhat.com>
Tested-by: Borislav Petkov <bp@suse.de>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Why]
This fixes 892deb4826 ("drm/amdgpu: Separate vf2pf work item init from virt data exchange").
we should read pf2vf data based at mman.fw_vram_usage_va after gmc
sw_init. commit 892deb4826 breaks this logic.
[How]
calling amdgpu_virt_exchange_data in amdgpu_virt_init_data_exchange to
set the right base in the right sequence.
v2:
call amdgpu_virt_init_data_exchange after gmc sw_init to make data
exchange workqueue run
v3:
clean up the code logic
v4:
add some comment and make the code more readable
Fixes: 892deb4826 ("drm/amdgpu: Separate vf2pf work item init from virt data exchange")
Signed-off-by: Jingwen Chen <Jingwen.Chen2@amd.com>
Reviewed-by: Horace Chen <horace.chen@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
I met a bug recently and the kernel log:
[ 330.171875] radeon 0000:03:00.0: couldn't schedule ib
[ 330.175781] [drm:radeon_uvd_suspend [radeon]] *ERROR* Error destroying UVD (-22)!
In radeon drivers, using UVD suspend is as follows:
if (rdev->has_uvd) {
uvd_v1_0_fini(rdev);
radeon_uvd_suspend(rdev);
}
In radeon_ib_schedule function, we check the 'ring->ready' state,
but in uvd_v1_0_fini funciton, we've cleared the ready state.
So, just modify the suspend code flow to fix error.
Reviewed-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Qiang Ma <maqianga@uniontech.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
pm_runtime_get_sync() increments the runtime PM usage counter even
when it returns an error code, thus a matching decrement is needed
on the error handling path to keep the counter balanced.
Signed-off-by: Yongzhi Liu <lyz_cs@pku.edu.cn>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The calcs folder has FPU code on it, which should be isolated inside the
DML folder as per https://patchwork.freedesktop.org/series/93042/.
This commit aims single-handedly to correct the location of such FPU
code and does not refactor any functions.
Changes since v2:
- Corrected problems to compile when DCN was disabled.
Reviewed-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Isabella Basso <isabbasso@riseup.net>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This version brings along following fixes:
- Organize FPU associated code to DML
- Modify SMU_TIMEOUT macro
- Organize dcn201 code
- Address DS stays disabled problem under specific scenario
- Fix black screen issue
- Update DML to rev.99
- Address problem of eDP hot-plug feature
Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Aric Cyr <aric.cyr@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Why]
For allow eDP hot-plug feature, the stream signal may change to VIRTUAL
when plug-out and back to eDP when plug-in. OS will still setPathMode
with same timing for each plugging, but eDP gets no stream update as we
don't check signal type changing back as keeping it VIRTUAL. It's also
unsafe for future cases that stream signal is switched with same timing.
[How]
Check stream signal type change include previous HDMI signal case.
Reviewed-by: Aric Cyr <Aric.Cyr@amd.com>
Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Dale Zhao <dale.zhao@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[WHY]
With some monitors when multi plane overlay is enabled the memory
clock switching mechanism has to change and, due to an error in the
initialization sequence, it may cause a black screen.
[HOW]
Change the firmware assisted memory clock switch initialization and
tear-down sequence utilizing the prepare_bandwidth and
optimize_bandwidth contexts.
Reviewed-by: Aric Cyr <Aric.Cyr@amd.com>
Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Felipe Clark <feclark@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Why]
On HG APU + dGPU scenario with no display to dGPU,
DS stays disabled due to no display present.
This problem can be worked around by DAL calling
DCEFCLK DS message to SMU on clk init.
[How]
Call DCEFCLK DS message to SMU on clk init.
Reviewed-by: Jun Lei <Jun.Lei@amd.com>
Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Chris Park <Chris.Park@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[WHY]
If some SMU features are not enabled, SMU will return fail to that
message.
[HOW]
SMU_TIMEOUT macro will treat "return fail" as timeout also.
Correct the macro to only report timeout case.
Reviewed-by: Jun Lei <Jun.Lei@amd.com>
Acked-by: Wayne Lin <wayne.lin@amd.com>
Signed-off-by: Ian Chen <ian.chen@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
