Am 2023-03-01 um 18:46 schrieb Daniel Phillips:
Our assumptions about how much KFD memory is currently available for
allocation may be violated by various complexities so we define the
reported value as advisory, however we should never report negative
availability.
Signed-off-by: Daniel Phillips <daniel.phill...@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
index 6f236ded5f12..293f9b2292fa 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
@@ -1583,6 +1583,9 @@ size_t amdgpu_amdkfd_get_available_memory(struct
amdgpu_device *adev)
- reserved_for_pt;
spin_unlock(&kfd_mem_limit.mem_limit_lock);
+ if ((int64_t)available < 0)
+ available = 0;
Instead of casting avaliable to a signed type here, you could define it
with a signed type in the first place. ssize_t would be appropriate.
Regards,
Felix
+
return ALIGN_DOWN(available, VRAM_AVAILABLITY_ALIGN);
}
