[Why]
drm_irq_uninstall is called in irq_fini_hw so that irq is disabled in sw
stage. SMU (and maybe other IP blocks) fini_hw will call irq_put for
cleanup and the whole cleanup process will be skipped because of
drm->irq_enable = false.
[How]
Move ip_fini_early before irq_fini_hw.
Signed-off-by: YuBiao Wang <YuBiao.Wang@amd.com>
Reviewed-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Missed a few asics.
v2: update comment
Fixes: 82d05736c4 ("drm/amdgpu/amdgpu_psp: convert to IP version checking")
Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
When creating unregistered new svm range to recover retry fault, avoid
new svm range to overlap with ranges or userptr ranges managed by TTM,
otherwise svm migration will trigger TTM or userptr eviction, to evict
user queues unexpectedly.
Change helper amdgpu_ttm_tt_affect_userptr to return userptr which is
inside the range. Add helper svm_range_check_vm_userptr to scan all
userptr of the vm, and return overlap userptr bo start, last.
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
During mode2 reset, the GPU is temporarily removed from the
mgpu_info list. As a result, page retirement fails because it
cannot find the GPU in the GPU list.
To fix this, create our own list of GPUs that support MCE notifier
based page retirement and use that list to check if the UMC error
occurred on a GPU that supports MCE notifier based page retirement.
Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
It's used internally by firmware. Using it in the driver
could conflict with firmware.
v2: squash in fix for navi1x (Alex)
Reviewed-by: James Zhu <James.Zhu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
It's used internally by firmware. Using it in the driver
could conflict with firmware.
Reviewed-by: James Zhu <James.Zhu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
On Aldebaran, GPU driver will handle bad page retirement
for GPU memory even though UMC is host managed. As a result,
register a bad page retirement handler on the mce notifier
chain to retire bad pages on Aldebaran.
Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Yazen Ghannam <yazen.ghannam@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
ind_block_64b_no_128bcl means INDEP_64B && INDEP_128B &&
MAX_COMPRESSED_BLOCK_SIZE == 64B. Only used by gfx10.3.
ind_block_64b means INDEP_64B && !INDEP_128B &&
MAX_COMPRESSED_BLOCK_SIZE == 64B. Only used by gfx9 and gfx10.
Signed-off-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
In current code, when a PCI error state pci_channel_io_normal is detectd,
it will report PCI_ERS_RESULT_CAN_RECOVER status to PCI driver, and PCI
driver will continue the execution of PCI resume callback report_resume by
pci_walk_bridge, and the callback will go into amdgpu_pci_resume
finally, where write lock is releasd unconditionally without acquiring
such lock first. In this case, a deadlock will happen when other threads
start to acquire the read lock.
To fix this, add a member in amdgpu_device strucutre to cache
pci_channel_state, and only continue the execution in amdgpu_pci_resume
when it's pci_channel_io_frozen.
Fixes: c9a6b82f45 ("drm/amdgpu: Implement DPC recovery")
Suggested-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Signed-off-by: Guchun Chen <guchun.chen@amd.com>
Reviewed-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
In current code, when a PCI error state pci_channel_io_normal is detectd,
it will report PCI_ERS_RESULT_CAN_RECOVER status to PCI driver, and PCI
driver will continue the execution of PCI resume callback report_resume by
pci_walk_bridge, and the callback will go into amdgpu_pci_resume
finally, where write lock is releasd unconditionally without acquiring
such lock first. In this case, a deadlock will happen when other threads
start to acquire the read lock.
To fix this, add a member in amdgpu_device strucutre to cache
pci_channel_state, and only continue the execution in amdgpu_pci_resume
when it's pci_channel_io_frozen.
Fixes: c9a6b82f45 ("drm/amdgpu: Implement DPC recovery")
Suggested-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Signed-off-by: Guchun Chen <guchun.chen@amd.com>
Reviewed-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This reverts commit 728e7e0cd6.
Further discussion reveals that this feature is severely broken
and needs to be reverted ASAP.
GPU reset can never be delayed by userspace even for debugging or
otherwise we can run into in kernel deadlocks.
Signed-off-by: Christian König <christian.koenig@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Acked-by: Nirmoy Das <nirmoy.das@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Memory is allocated for ttm->sg by kmalloc in kfd_mem_dmamap_userptr,
but isn't freed by kfree in kfd_mem_dmaunmap_userptr. Free it!
Fixes: 264fb4d332 ("drm/amdgpu: Add multi-GPU DMA mapping helpers")
Signed-off-by: Lang Yu <lang.yu@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
In the rare event when GFX IP suspend coincides with a s0ix entry, don't
schedule a delayed work, instead signal PMFW immediately to allow GFXOFF
entry. GFXOFF is a prerequisite for s0ix entry. PMFW needs to be
signaled about GFXOFF status before amd-pmc module passes OS HINT
to PMFW telling that everything is ready for a safe s0ix entry.
Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/1712
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Mario Limonciello <mario.limonciell@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
IP_VERSION(11, 0, 13) does the exact same thing as
IP_VERSION(11, 0, 12) so squash them together.
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Memory is allocated for ttm->sg by kmalloc in kfd_mem_dmamap_userptr,
but isn't freed by kfree in kfd_mem_dmaunmap_userptr. Free it!
Fixes: 264fb4d332 ("drm/amdgpu: Add multi-GPU DMA mapping helpers")
Signed-off-by: Lang Yu <lang.yu@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If you set amdgpu.discovery=2 you can force the the driver to
fetch the IP discovery table from a file rather than from the
table shipped on the device. This is useful for debugging and
for device bring up and emulation when the tables may be in flux.
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
We can get the pdev and asic type from the adev. No need
to pass them explicitly.
v2: squash in build fix for !CONFIG_HSA_AMD from Anson
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Handle SRIOV requirements when adding IP blocks.
v2: add comment about UVD/VCE support on vega20 SR-IOV
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Split into several smaller per IP functions to make it
easier to handle ordering issues for things like
SR-IOV in a follow up patch.
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Allow us to query instances versions more cleanly.
Instancing support is not consistent unfortunately. SDMA is a
good example. Sienna cichlid has 4 total SDMA instances, each
enumerated separately (HWIDs 42, 43, 68, 69). Arcturus has 8
total SDMA instances, but they are enumerated as multiple
instances of the same HWIDs (4x HWID 42, 4x HWID 43). UMC
is another example. On most chips there are multiple
instances with the same HWID. This allows us to support both
forms.
v2: rebase
v3: clarify instancing support
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add a new asic type for asics where we don't have an
explicit entry in the PCI ID list. We don't need
an asic type for these asics, other than something higher
than the existing ones, so just use this for all new
asics.
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Use the table rather than asic specific harvest registers.
v2: remove harvesting register checking
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Rather than hardcoding it. We already have the number of VCN
instances from a previous patch, so just update the VCN
instances for chips with static tables.
v2: squash in checks for SDMA3,4 (Guchun)
v3: clarify VCN changes
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
