[PATCH] drm/amdgpu: introduce vram lost for reset

Monk Liu Monk.Liu at amd.com
Tue Aug 27 08:34:33 UTC 2019


for SOC15/vega10 the BACO reset & mode1 would introduce vram lost
in high end address range, current kmd's vram lost checking cannot
catch it since it only check very ahead visible frame buffer

Signed-off-by: Monk Liu <Monk.Liu at amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu.h        | 1 +
 drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 4 ++--
 drivers/gpu/drm/amd/amdgpu/soc15.c         | 2 ++
 3 files changed, 5 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
index f6ae565..11e0fc0 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
@@ -1155,6 +1155,7 @@ int emu_soc_asic_init(struct amdgpu_device *adev);
 #define amdgpu_asic_get_pcie_usage(adev, cnt0, cnt1) ((adev)->asic_funcs->get_pcie_usage((adev), (cnt0), (cnt1)))
 #define amdgpu_asic_need_reset_on_init(adev) (adev)->asic_funcs->need_reset_on_init((adev))
 #define amdgpu_asic_get_pcie_replay_count(adev) ((adev)->asic_funcs->get_pcie_replay_count((adev)))
+#define amdgpu_inc_vram_lost(adev) atomic_inc(&((adev)->vram_lost_counter));
 
 /* Common functions */
 bool amdgpu_device_should_recover_gpu(struct amdgpu_device *adev);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index 02b3e7d..31690e9 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -3482,7 +3482,7 @@ static int amdgpu_device_reset_sriov(struct amdgpu_device *adev,
 	amdgpu_virt_init_data_exchange(adev);
 	amdgpu_virt_release_full_gpu(adev, true);
 	if (!r && adev->virt.gim_feature & AMDGIM_FEATURE_GIM_FLR_VRAMLOST) {
-		atomic_inc(&adev->vram_lost_counter);
+		amdgpu_inc_vram_lost(adev);
 		r = amdgpu_device_recover_vram(adev);
 	}
 
@@ -3648,7 +3648,7 @@ static int amdgpu_do_asic_reset(struct amdgpu_hive_info *hive,
 				vram_lost = amdgpu_device_check_vram_lost(tmp_adev);
 				if (vram_lost) {
 					DRM_INFO("VRAM is lost due to GPU reset!\n");
-					atomic_inc(&tmp_adev->vram_lost_counter);
+					amdgpu_inc_vram_lost(tmp_adev);
 				}
 
 				r = amdgpu_gtt_mgr_recover(
diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.c b/drivers/gpu/drm/amd/amdgpu/soc15.c
index fe2212df..8af7501 100644
--- a/drivers/gpu/drm/amd/amdgpu/soc15.c
+++ b/drivers/gpu/drm/amd/amdgpu/soc15.c
@@ -557,10 +557,12 @@ static int soc15_asic_reset(struct amdgpu_device *adev)
 {
 	switch (soc15_asic_reset_method(adev)) {
 		case AMD_RESET_METHOD_BACO:
+			amdgpu_inc_vram_lost(adev);
 			return soc15_asic_baco_reset(adev);
 		case AMD_RESET_METHOD_MODE2:
 			return soc15_mode2_reset(adev);
 		default:
+			amdgpu_inc_vram_lost(adev);
 			return soc15_asic_mode1_reset(adev);
 	}
 }
-- 
2.7.4



More information about the amd-gfx mailing list