add ras post init function.
Do some initialization after all IP have finished their late init.
Add new member flags which will control the ras work flow.
For now, vbios enable ras for us on boot. That might change in the
future.
So there should be a flag from vbios to tell us if ras is enabled or not
on boot. Looks like there is no such info now.
Other bits of the flags are reserved to control other parts of ras.
Signed-off-by: xinhui pan <xinhui.pan@amd.com>
Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
add obj management.
add feature control.
add debugfs infrastructure.
add sysfs infrastructure.
add IH infrastructure.
add recovery infrastructure.
It is a framework. Other IPs need call amdgpu_ras_xxx function instead of
psp_ras_xxx functions.
v2: squash in warning fixes
Signed-off-by: xinhui pan <xinhui.pan@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Problem:
Using SDMA for TLB invalidation in certain ASICs exposed a problem
of IB pool not being ready while SDMA already up on Init and already
shutt down while SDMA still running on Fini. This caused
IB allocation failure. Temproary fix was commited into a
bringup branch but this is the generic fix.
Fix:
Init IB pool rigth after GMC is ready but before SDMA is ready.
Do th opposite for Fini.
v2: Remove restriction on SDMA early init and move amdgpu_ib_pool_fini
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Driver vote low to high pstate switch whenever there is an outstanding
XGMI mapping request. Driver vote high to low pstate when all the
outstanding XGMI mapping is terminated.
Signed-off-by: shaoyunl <shaoyun.liu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drm-misc-next for 5.1:
UAPI Changes:
Cross-subsystem Changes:
Core Changes:
- Split out some part of drm_crtc_helper.h into drm_probe_helper.h
- DRIVER_* flags improvements
- New tasks on the TODO-list
- Improvements to the documentation
Driver Changes:
- Continual of drmP.h removal in multiple drivers
- Removal of FBINFO_(FLAG_)DEFAULT in multiple drivers
- sun4i: Addition of the A23 support, multiple fixes for the tiled
formats
- atmel-hlcdc: Fix of clipping and rotation properties
- qxl: various BO-related improvements, prime and generic fbdev emulation
support
- dw-hdmi: Support for HDMI2.0 2160p modes and YUV420 output
- New Sitronix ST7701 panel driver
- New Kingdisplay KD097D04 panel driver
- New LeMaker BL035-RGB-002 panel driver
- New PDA 91-00156-A0 panel driver
Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Maxime Ripard <maxime.ripard@bootlin.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190201144749.t3abxvguhstu6bcl@flea
New stuff for 5.1.
amdgpu:
- DC bandwidth formula updates
- Support for DCC on scanout surfaces
- Support for multiple IH rings on soc15 asics
- Fix xgmi locking
- Add sysfs interface to get pcie usage stats
- Simplify DC i2c/aux code
- Initial support for BACO on vega10/20
- New runtime SMU feature debug interface
- Expand existing sysfs power interfaces to new clock domains
- Handle kexec properly
- Simplify IH programming
- Rework doorbell handling across asics
- Drop old CI DPM implementation
- DC page flipping fixes
- Misc SR-IOV fixes
amdkfd:
- Simplify the interfaces between amdkfd and amdgpu
ttm:
- Add a callback to notify the driver when the lru changes
sched:
- Refactor mirror list handling
- Rework hw fence processing
Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexdeucher@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190125231517.26268-1-alexander.deucher@amd.com
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.
v7:
Wait only for the latest job's fence.
Suggested-by: Christian Koenig <Christian.Koenig@amd.com>
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
sriov would meet guest driver load failure,
if calling amdgpu_asic_reset in amdgpu_device_init.
sriov should skip asic_reset in device_init.
Signed-off-by: Wentao Lou <Wentao.Lou@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
v2: Move locks around in other functions so that this
function can stand on its own. Also only hold the hive
specific lock for add/remove device instead of the driver
global lock so you can't add/remove devices in parallel from
one hive.
v3: add reset_lock
Acked-by: Shaoyun.liu < Shaoyun.liu@amd.com>
Signed-off-by: Tom St Denis <tom.stdenis@amd.com>
Reviewed-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
XGMI hive put kfd_pre_reset into amdgpu_device_lock_adev,
but outside req_full_gpu of sriov.
It would make sriov hang during reset.
Signed-off-by: Wentao Lou <Wentao.Lou@amd.com>
Reviewed-by: Shaoyun Liu <Shaoyun.Liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Use per hive wq to concurrently send reset commands to all nodes
in the hive.
v2:
Switch to system_highpri_wq after dropping dedicated queue.
Fix non XGMI code path KASAN error.
Stop the hive reset for each node loop if there
is a reset failure on any of the nodes.
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
XGMI hive has some resources allocted on device init which
needs to be deallocated when the device is unregistered.
v2: Remove creation of dedicated wq for XGMI hive reset.
v3: Use the gmc.xgmi.supported flag
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
When paging queue is enabled, it use the second page of doorbell.
The AMDGPU_DOORBELL64_MAX_ASSIGNMENT definition assumes all the
kernel doorbells are in the first page. So with paging queue enabled,
the total kernel doorbell range should be original num_doorbell plus
one page (0x400 in dword), not *2.
Signed-off-by: Oak Zeng <ozeng@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
For XGMI hive case do reset in steps where each step iterates over
all devs in hive. This especially important for asic reset
since all PSP FW in hive must come up within a limited time
(around 1 sec) to properply negotiate the link.
Do this by refactoring amdgpu_device_gpu_recover and amdgpu_device_reset
into pre_asic_reset, asic_reset and post_asic_reset functions where is part
is exectued for all the GPUs in the hive before going to the next step.
v2: Update names for amdgpu_device_lock/unlock functions.
v3: Introduce per hive locking to avoid multiple resets for GPUs
in same hive.
v4:
Remove delayed_workqueue()/ttm_bo_unlock_delayed_workqueue() - they
are copy & pasted over from radeon and on amdgpu there isn't
any reason for that any more.
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This is prep work for updating each PSP FW in hive after
GPU reset.
Split into build topology SW state and update each PSP FW in the hive.
Save topology and count of XGMI devices for reuse.
v2: Create seperate header for XGMI.
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Because increase SDMA_DOORBELL_RANGE to add new SDMA doorbell for paging queue will
break SRIOV, instead we can reserve and map two doorbell pages for amdgpu, paging
queues doorbell index use same index as SDMA gfx queues index but on second page.
For Vega20, after we change doorbell layout to increase SDMA doorbell for 8 SDMA RLC
queues later, we could use new doorbell index for paging queue.
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Setup and tear down xgmi as part of psp.
v2:
- make psp_xgmi_terminate static
- squash in:
drm/amdgpu: only issue xgmi cmd when it is enabled
drm/amdgpu/psp: terminate xgmi ta in suspend and hw_fini phase
Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
There is no functional changes,
Use function arguments for SRIOV special variables which
is hardcode in those functions.
so we can share those functions in baremetal.
Reviewed-by: Monk Liu <Monk.Liu@amd.com>
Signed-off-by: Rex Zhu <Rex.Zhu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
After testing looks like these subset of ASICs has GPU reset
working for the most part. Enable reset due to job timeout.
v2: Switch from GFX version to ASIC type.
v3: Fix identation
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Problem:
During GPU recover DAL would hang in
amdgpu_pm_compute_clocks->amdgpu_fence_wait_empty
Fix:
Turns out there was a typo introduced by
3320b8d drm/amdgpu: remove job->ring which caused skipping
amdgpu_fence_driver_force_completion and so the hangged job
was never force signaled and this would cause the hang later in DAL.
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
We need to do some IPs earlier to deal with ordering issues
similar to how resume is split into two phases.
Will do fw loading via smu/psp between the two phases.
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Rex Zhu <Rex.Zhu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
1. one is for create/free bo when init/fini
2. one is for fill the bo before fw loading
the ucode bo only need to be created when load driver
and free when driver unload.
when resume/reset, driver only need to re-fill the bo
if the bo is allocated in vram.
Suggested by Christian.
v2: Return error when bo create failed.
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Rex Zhu <Rex.Zhu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
1. only call late_init when hw_init successful,
so check status.hw instand of status.valid in late_init.
2. set status.late_initialized true if late_init was not implemented.
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Rex Zhu <Rex.Zhu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Move in_suspend flag to adev from gfx, so
can be used in other ip blocks, also keep
consistent with gpu_in_reset flag.
Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Rex Zhu <Rex.Zhu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Treat them all as Raven rather than adding a new picasso
asic type. This simplifies a lot of code and also handles the
case of rv2 chips with the 0x15d8 pci id. It also fixes dmcu
fw handling for picasso.
Acked-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
