[ Upstream commit e864180ee4 ]
flush_gpu_tlb may be called from another thread while
device_gpu_recover is running.
Both of these threads access registers through the VF
RLCG interface during VF Full Access. Add a lock around this interface
to prevent race conditions between these threads.
Signed-off-by: Victor Skvortsov <victor.skvortsov@amd.com>
Reviewed-by: Zhigang Luo <zhigang.luo@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit 5659b0c93a upstream.
[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)
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit a03ebf1163 upstream.
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)
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 1446226d32 ]
The following commit updated gmc->noretry from 0 to 1 for GC HW IP
9.3.0:
commit 5f3854f1f4 ("drm/amdgpu: add more cases to noretry=1")
This causes the device to hang when a page fault occurs, until the
device is rebooted. Instead, revert back to gmc->noretry=0 so the device
is still responsive.
Fixes: 5f3854f1f4 ("drm/amdgpu: add more cases to noretry=1")
Signed-off-by: Tim Van Patten <timvp@google.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 88a9a467c5 ]
Initialize the size before calling amdgpu_vce_cs_reloc, such as case 0x03000001.
V2: To really improve the handling we would actually
need to have a separate value of 0xffffffff.(Christian)
Signed-off-by: Jesse Zhang <jesse.zhang@amd.com>
Suggested-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit f6f49dda49 upstream.
v3.x changed the how vram width was encoded. The previous
implementation actually worked correctly for most boards.
Fix the implementation to work correctly everywhere.
This fixes the vram width reported in the kernel log on
some boards.
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 74fa02c4a5 ]
Cache the PCI state before bus master is disabled. The saved state is
later used for other cases like restoring config space after mode-2
reset.
Fixes: 5c03e5843e ("drm/amdgpu:add smu mode1/2 support for aldebaran")
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Feifei Xu <Feifei.Xu@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit acce6479e3 ]
The function gfx_v9_4_3_init_microcode in gfx_v9_4_3.c was generating
about potential truncation of output when using the snprintf function.
The issue was due to the size of the buffer 'ucode_prefix' being too
small to accommodate the maximum possible length of the string being
written into it.
The string being written is "amdgpu/%s_mec.bin" or "amdgpu/%s_rlc.bin",
where %s is replaced by the value of 'chip_name'. The length of this
string without the %s is 16 characters. The warning message indicated
that 'chip_name' could be up to 29 characters long, resulting in a total
of 45 characters, which exceeds the buffer size of 30 characters.
To resolve this issue, the size of the 'ucode_prefix' buffer has been
reduced from 30 to 15. This ensures that the maximum possible length of
the string being written into the buffer will not exceed its size, thus
preventing potential buffer overflow and truncation issues.
Fixes the below with gcc W=1:
drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c: In function ‘gfx_v9_4_3_early_init’:
drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c:379:52: warning: ‘%s’ directive output may be truncated writing up to 29 bytes into a region of size 23 [-Wformat-truncation=]
379 | snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_rlc.bin", chip_name);
| ^~
......
439 | r = gfx_v9_4_3_init_rlc_microcode(adev, ucode_prefix);
| ~~~~~~~~~~~~
drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c:379:9: note: ‘snprintf’ output between 16 and 45 bytes into a destination of size 30
379 | snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_rlc.bin", chip_name);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c:413:52: warning: ‘%s’ directive output may be truncated writing up to 29 bytes into a region of size 23 [-Wformat-truncation=]
413 | snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec.bin", chip_name);
| ^~
......
443 | r = gfx_v9_4_3_init_cp_compute_microcode(adev, ucode_prefix);
| ~~~~~~~~~~~~
drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c:413:9: note: ‘snprintf’ output between 16 and 45 bytes into a destination of size 30
413 | snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec.bin", chip_name);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Fixes: 8630112969 ("drm/amdgpu: split gc v9_4_3 functionality from gc v9_0")
Cc: Hawking Zhang <Hawking.Zhang@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Suggested-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit f06446ef23 ]
Subtract the VRAM pinned memory when checking for available memory
in amdgpu_amdkfd_reserve_mem_limit function since that memory is not
available for use.
Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit fe93b0927b ]
Here are the corrections needed for the queue ring buffer size
calculation for the following cases:
- Remove the KIQ VM flush ring usage.
- Add the invalidate TLBs packet for gfx10 and gfx11 queue.
- There's no VM flush and PFP sync, so remove the gfx9 real
ring and compute ring buffer usage.
Signed-off-by: Prike Liang <Prike.Liang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit b0b13d5321 ]
Make SVM BOs more likely to get evicted than other BOs. These BOs
opportunistically use available VRAM, but can fall back relatively
seamlessly to system memory. It also avoids SVM migrations evicting
other, more important BOs as they will evict other SVM allocations
first.
Signed-off-by: Felix Kuehling <felix.kuehling@amd.com>
Acked-by: Mukul Joshi <mukul.joshi@amd.com>
Tested-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit d3a9331a65 upstream.
This reverts drm/amdgpu: fix ftrace event amdgpu_bo_move always move
on same heap. The basic problem here is that after the move the old
location is simply not available any more.
Some fixes were suggested, but essentially we should call the move
notification before actually moving things because only this way we have
the correct order for DMA-buf and VM move notifications as well.
Also rework the statistic handling so that we don't update the eviction
counter before the move.
v2: add missing NULL check
Signed-off-by: Christian König <christian.koenig@amd.com>
Fixes: 94aeb41173 ("drm/amdgpu: fix ftrace event amdgpu_bo_move always move on same heap")
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/3171
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
CC: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit a6ff969fe9 ]
When we removed the hacky start code check we actually didn't took into
account that *all* VRAM pages needs to be CPU accessible.
Clean up the code and unify the handling into a single helper which
checks if the whole resource is CPU accessible.
The only place where a partial check would make sense is during
eviction, but that is neglitible.
Signed-off-by: Christian König <christian.koenig@amd.com>
Fixes: aed01a6804 ("drm/amdgpu: Remove TTM resource->start visible VRAM condition v2")
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
CC: stable@vger.kernel.org
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit ca7c4507ba upstream.
The majority of those where removed in the commit aed01a6804
("drm/amdgpu: Remove TTM resource->start visible VRAM condition v2")
But this one was missed because it's working on the resource and not the
BO. Since we also no longer use a fake start address for visible BOs
this will now trigger invalid mapping errors.
v2: also remove the unused variable
Signed-off-by: Christian König <christian.koenig@amd.com>
Fixes: aed01a6804 ("drm/amdgpu: Remove TTM resource->start visible VRAM condition v2")
CC: stable@vger.kernel.org
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit eb4f139888 ]
This ensures that the memory mapped by ioremap for adev->rmmio, is
properly handled in amdgpu_device_init(). If the function exits early
due to an error, the memory is unmapped. If the function completes
successfully, the memory remains mapped.
Reported by smatch:
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c:4337 amdgpu_device_init() warn: 'adev->rmmio' from ioremap() not released on lines: 4035,4045,4051,4058,4068,4337
Cc: Christian König <christian.koenig@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
