Support per-queue reset for GFX9. The recommendation is for the driver
to target reset the HW queue via a SPI MMIO register write.
Since this requires pipe and HW queue info and MEC FW is limited to
doorbell reports of hung queues after an unmap failure, scan the HW
queue slots defined by SET_RESOURCES first to identify the user queue
candidates to reset.
Only signal reset events to processes that have had a queue reset.
If queue reset fails, fall back to GPU reset.
Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
debugfs register list for dump is cleaned as it have
some issues related to proper power state of the IP
before register read.
Since the above mentioned is removed we no longer want
this to be dumped part of the devcoredump and hence
removed.
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
There are some problem with existing amdgpu_reset_dump_register_list
debugfs node. It is supposed to read a list of registers but there
could be cases when the IP is not in correct power state. Register
read in such cases could lead to more problems.
We are taking care of all such power states in devcoredump and
dumping the registers of need for debugging. So cleaning this code
and we dont need this functionality via debugfs anymore.
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Instead of manually passing around 'struct edid *' and its size,
use 'struct drm_edid', which encapsulates a validated combination of
both.
As the drm_edid_ can handle NULL gracefully, the explicit checks can be
dropped.
Also save a few characters by transforming '&array[0]' to the equivalent
'array' and using 'max_t(int, ...)' instead of manual casts.
Signed-off-by: Thomas Weißschuh <linux@weissschuh.net>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Problem:
IP dump right now is done post suspend of all
IP's which for some IP's could change power
state and software state too which we do not want
to reflect in the dump as it might not be same at
the time of hang.
Solution:
IP should be dumped as close to the HW state when
the GPU was in hung state without trying to reinitialize
any resource.
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
ISP I2C bus device can't be enumerated via ACPI mechanism
since it shares the memory map with the AMDGPU.
So use the MFD mechanism for registering the ISP I2C device
and add the required resources.
Signed-off-by: Venkata Narendra Kumar Gutta <vengutta@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add macro definition which calculate offset of the
register with index override.
This is useful in case when there is an array of
registers which is common for all instances.
To read registers in that case it is easy to define
registers once and the index value is manually passed
to calculate proper offset of register for each instance.
Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The comment in the vbios structure says:
// = 128 means EDID length is 128 bytes, otherwise the EDID length = ucFakeEDIDLength*128
This fake edid struct has not been used in a long time, so I'm
not sure if there were actually any boards out there with a non-128 byte
EDID, but align the code with the comment.
Reviewed-by: Thomas Weißschuh <linux@weissschuh.net>
Reported-by: Thomas Weißschuh <linux@weissschuh.net>
Link: https://lists.freedesktop.org/archives/amd-gfx/2024-June/109964.html
Fixes: d38ceaf99e ("drm/amdgpu: add core driver (v4)")
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Why]
Page table of compute VM in the VRAM will lost after gpu reset.
VRAM won't be restored since compute VM has no shadows.
[How]
Use higher 32-bit of vm->generation to record a vram_lost_counter.
Reset the VM state machine when vm->genertaion is not equal to
the new generation token.
v2: Check vm->generation instead of calling drm_sched_entity_error
in amdgpu_vm_validate.
v3: Use new generation token instead of vram_lost_counter for check.
Signed-off-by: ZhenGuo Yin <zhenguo.yin@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
(cherry picked from commit 47c0388b05)
The eeprom table is empty before initializing,
set eeprom table version first before initializing.
Changed from V1:
Reuse amdgpu_ras_set_eeprom_table_version function
Signed-off-by: Stanley.Yang <Stanley.Yang@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 015b8a2fdf)
The ras command shared memory is allocated from
VRAM and the response status of the command
buffer will not be zero due to gpu being in
fatal error state after ras UE error injection.
Signed-off-by: YiPeng Chai <YiPeng.Chai@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 8284951a6e)
For VCN/JPEG 4.0.3, use only the local addressing scheme.
- Mask bit higher than AID0 range
v2
remain the case for mmhub use master XCC
Signed-off-by: Jane Jian <Jane.Jian@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit caaf576292)
VCN 4.0.3 does not HDP flush with RRMT enabled. Instead, mmsch
will do the HDP flush.
This change is necessary for VCN v4.0.3, no need for backward compatibility
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Jane Jian <Jane.Jian@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 49cfaebe48)
JPEG v4.0.3 doesn't support HDP flush when RRMT is enabled. Instead,
mmsch fw will do the flush.
This change is necessary for JPEG v4.0.3, no need for backward compatibility
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Jane Jian <Jane.Jian@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 585e3fdb36)
If PCIe supports atomics, configure register to prevent DF from
breaking atomics in separate load/store operations.
Signed-off-by: David Belanger <david.belanger@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 666f14cab2)
We seem to have a case where SDMA will sometimes miss a doorbell
if GFX is entering the powergating state when the doorbell comes in.
To workaround this, we can update the wptr via MMIO, however,
this is only safe because we disallow gfxoff in begin_ring() for
SDMA 5.2 and then allow it again in end_ring().
Enable this workaround while we are root causing the issue with
the HW team.
Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/3440
Tested-by: Friedrich Vock <friedrich.vock@gmx.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
(cherry picked from commit f2ac526349)
[Why]
Page table of compute VM in the VRAM will lost after gpu reset.
VRAM won't be restored since compute VM has no shadows.
[How]
Use higher 32-bit of vm->generation to record a vram_lost_counter.
Reset the VM state machine when vm->genertaion is not equal to
the new generation token.
v2: Check vm->generation instead of calling drm_sched_entity_error
in amdgpu_vm_validate.
v3: Use new generation token instead of vram_lost_counter for check.
Signed-off-by: ZhenGuo Yin <zhenguo.yin@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
For the bad opcode case, it will cause CP/ME hang.
The firmware will prevent the ME side from hanging by raising a bad opcode interrupt.
And the driver needs to perform a vmid reset when receiving the interrupt.
Acked-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
For the bad opcode case, it will cause CP/ME hang.
The firmware will prevent the ME side from hanging by raising a bad opcode interrupt.
And the driver needs to perform a vmid reset when receiving the interrupt.
Acked-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
For the bad opcode case, it will cause CP/ME hang.
The firmware will prevent the ME side from hanging by raising a bad opcode interrupt.
And the driver needs to perform a vmid reset when receiving the interrupt.
v2: update irq naming (drop priv) (Alex)
Acked-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Jesse Zhang <jesse.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
For the bad opcode case, it will cause CP/ME hang.
The firmware will prevent the ME side from hanging by raising a bad opcode interrupt.
And the driver needs to perform a vmid reset when receiving the interrupt.
v2: update irq naming (drop priv) (Alex)
Acked-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Jesse Zhang <jesse.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
For the bad opcode case, it will cause CP/ME hang.
The firmware will prevent the ME side from hanging by raising a bad opcode interrupt.
And the driver needs to perform a vmid reset when receiving the interrupt.
v2: update irq naming (drop priv) (Alex)
Acked-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Jesse Zhang <Jesse.Zhang@amd.com>
Reviewed-by: Prike Liang <Prike.Liang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The eeprom table is empty before initializing,
set eeprom table version first before initializing.
Changed from V1:
Reuse amdgpu_ras_set_eeprom_table_version function
Signed-off-by: Stanley.Yang <Stanley.Yang@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The ras command shared memory is allocated from
VRAM and the response status of the command
buffer will not be zero due to gpu being in
fatal error state after ras UE error injection.
Signed-off-by: YiPeng Chai <YiPeng.Chai@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add atomic queue_refcount to struct bo_va, return -EBUSY to fail unmap
BO from the GPU if the bo_va queue_refcount is not zero.
Create queue to increase the bo_va queue_refcount, destroy queue to
decrease the bo_va queue_refcount, to ensure the queue buffers mapped on
the GPU when queue is active.
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
