drm/amdgpu: Correct amdgpu_amdkfd_total_mem_size calculation
Subsystems (3)
PATCH 1/1 drm/amdgpu: Correct amdgpu_amdkfd_total_mem_size calculation
amdkfd_total_mem_size is the size of total GPUs vram plus system memory to estimate page tables memory usage and leave enough VRAM room for page tables allocation.
Calculate amdkfd_total_mem_size in amdgpu_amdkfd_device_probe is incorrect because adev->gmc.real_vram_size is still 0 called from amdgpu_device_ip_early_init. Move the calculation
to amdgpu_amdkfd_device_init to get the correct VRAM size.
--- drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c +2 -3
@@ -75,9 +75,6 @@ void amdgpu_amdkfd_device_probe(struct amdgpu_device *adev)return ; adev->kfd.dev =kgd2kfd_probe (adev, vf); - -if (adev->kfd.dev) - amdgpu_amdkfd_total_mem_size += adev->gmc.real_vram_size; }/**
@@ -201,6 +198,8 @@ void amdgpu_amdkfd_device_init(struct amdgpu_device *adev) adev->kfd.init_complete =kgd2kfd_device_init (adev->kfd.dev,adev_to_drm (adev), &gpu_resources); + amdgpu_amdkfd_total_mem_size += adev->gmc.real_vram_size; +INIT_WORK (&adev->kfd.reset_work, amdgpu_amdkfd_reset_work); } }
Signature · 1 line
2.35.1
Quoted 10 lines · Philip Yang
amdkfd_total_mem_sizeis the size of total GPUs vram plus system memory to estimate page tables memory usage and leave enough VRAM room for page tables allocation.Calculate
amdkfd_total_mem_sizeinamdgpu_amdkfd_device_probeis incorrect becauseadev->gmc.real_vram_sizeis still 0 called fromamdgpu_device_ip_early_init. Move the calculation
toamdgpu_amdkfd_device_initto get the correct VRAM size.Signed-off-by Philip Yang <[email protected]>
Semi-related to this, there should probably be a reverse calculation in amdgpu_amdkfd_device_fini_sw to support hot-unplugging GPUs.
Regards,
Felix
Quoted 27 lines
--- drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-)drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c +2
-3@@ -75,9 +75,6 @@ void amdgpu_amdkfd_device_probe(struct amdgpu_device *adev)return ; adev->kfd.dev =kgd2kfd_probe (adev, vf); - -if (adev->kfd.dev) - amdgpu_amdkfd_total_mem_size += adev->gmc.real_vram_size; }/** @@ -201,6 +198,8 @@ void amdgpu_amdkfd_device_init(struct amdgpu_device *adev) adev->kfd.init_complete =kgd2kfd_device_init (adev->kfd.dev,adev_to_drm (adev), &gpu_resources); + amdgpu_amdkfd_total_mem_size += adev->gmc.real_vram_size; +INIT_WORK (&adev->kfd.reset_work, amdgpu_amdkfd_reset_work); } }
Quoted 10 lines · Felix Kuehling
Quoted 9 lines · Philip Yang
amdkfd_total_mem_sizeis the size of total GPUs vram plus system memory to estimate page tables memory usage and leave enough VRAM room for page tables allocation.Calculate
amdkfd_total_mem_sizeinamdgpu_amdkfd_device_probeis incorrect becauseadev->gmc.real_vram_sizeis still 0 called fromamdgpu_device_ip_early_init. Move the calculation
toamdgpu_amdkfd_device_initto get the correct VRAM size.
Quoted 5 lines
Quoted 1 line
Signed-off-by Philip Yang <[email protected]>Reviewed-by Felix Kuehling <[email protected]>Semi-related to this, there should probably be a reverse calculation in
amdgpu_amdkfd_device_fini_swto support hot-unplugging GPUs.
I will add the reverse calculation then submit.
Regards,
Philip
Quoted 31 lines
Regards,
FelixQuoted 26 lines
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c| 5 ++---1 file changed, 2 insertions(+), 3 deletions(-)drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c +1
-3@@ -75,9 +75,6 @@ void amdgpu_amdkfd_device_probe(struct amdgpu_device *adev) return; adev->kfd.dev = kgd2kfd_probe(adev, vf); - - if (adev->kfd.dev) - amdgpu_amdkfd_total_mem_size += adev->gmc.real_vram_size; } /**@@ -201,6 +198,8 @@ void amdgpu_amdkfd_device_init(struct amdgpu_device *adev) adev->kfd.init_complete = kgd2kfd_device_init(adev->kfd.dev, adev_to_drm(adev), &gpu_resources); + amdgpu_amdkfd_total_mem_size += adev->gmc.real_vram_size; + INIT_WORK(&adev->kfd.reset_work, amdgpu_amdkfd_reset_work); }}
Felix Kuehling
maintainerI'd prefer a separate patch and code review for the fini-case, because that addresses a different (potential) problem.
Thanks,
Felix
Quoted 55 lines · Philip Yang
Quoted 10 lines · Felix Kuehling
Quoted 9 lines · Philip Yang
amdkfd_total_mem_sizeis the size of total GPUs vram plus system memory to estimate page tables memory usage and leave enough VRAM room for page tables allocation.Calculate
amdkfd_total_mem_sizeinamdgpu_amdkfd_device_probeis incorrect becauseadev->gmc.real_vram_sizeis still 0 called fromamdgpu_device_ip_early_init. Move the calculation
toamdgpu_amdkfd_device_initto get the correct VRAM size.Quoted 5 lines
Quoted 1 line
Signed-off-by Philip Yang <[email protected]>Reviewed-by Felix Kuehling <[email protected]>Semi-related to this, there should probably be a reverse calculation in
amdgpu_amdkfd_device_fini_swto support hot-unplugging GPUs.I will add the reverse calculation then submit.
Regards,
Philip
Quoted 31 lines
Regards,
FelixQuoted 26 lines
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c| 5 ++---1 file changed, 2 insertions(+), 3 deletions(-)drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c +1
-3@@ -75,9 +75,6 @@ void amdgpu_amdkfd_device_probe(struct amdgpu_device *adev) return; adev->kfd.dev = kgd2kfd_probe(adev, vf); - - if (adev->kfd.dev) - amdgpu_amdkfd_total_mem_size += adev->gmc.real_vram_size; } /**@@ -201,6 +198,8 @@ void amdgpu_amdkfd_device_init(struct amdgpu_device *adev) adev->kfd.init_complete = kgd2kfd_device_init(adev->kfd.dev, adev_to_drm(adev), &gpu_resources); + amdgpu_amdkfd_total_mem_size += adev->gmc.real_vram_size; + INIT_WORK(&adev->kfd.reset_work, amdgpu_amdkfd_reset_work); }}