Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
drm/amdgpu: let mode2 reset fallback to default when failure
- introduce AMDGPU_SKIP_MODE2_RESET flag
- let mode2 reset fallback to default reset method if failed
v2: move this part out from the asic specific part
Signed-off-by: Victor Zhao <Victor.Zhao@amd.com>
Acked-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
authored by
Victor Zhao
and committed by
Alex Deucher
dac6b808
672c0218
+19
-1
+1
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
+1
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
···
135
135
reset_context.method = AMD_RESET_METHOD_NONE;
136
136
reset_context.reset_req_dev = adev;
137
137
clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
138
+
clear_bit(AMDGPU_SKIP_MODE2_RESET, &reset_context.flags);
138
139
139
140
amdgpu_device_gpu_recover(adev, NULL, &reset_context);
140
141
}
+6
-1
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
+6
-1
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
···
5148
5148
5149
5149
reset_context->job = job;
5150
5150
reset_context->hive = hive;
5151
+
5151
5152
/*
5152
5153
* Build list of devices to reset.
5153
5154
* In case we are in XGMI hive mode, resort the device list
···
5268
5267
amdgpu_ras_resume(adev);
5269
5268
} else {
5270
5269
r = amdgpu_do_asic_reset(device_list_handle, reset_context);
5271
-
if (r && r == -EAGAIN)
5270
+
if (r && r == -EAGAIN) {
5271
+
set_bit(AMDGPU_SKIP_MODE2_RESET, &reset_context->flags);
5272
+
adev->asic_reset_res = 0;
5272
5273
goto retry;
5274
+
}
5273
5275
}
5274
5276
5275
5277
skip_hw_reset:
···
5703
5699
reset_context.reset_req_dev = adev;
5704
5700
set_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
5705
5701
set_bit(AMDGPU_SKIP_HW_RESET, &reset_context.flags);
5702
+
set_bit(AMDGPU_SKIP_MODE2_RESET, &reset_context.flags);
5706
5703
5707
5704
adev->no_hw_access = true;
5708
5705
r = amdgpu_device_pre_asic_reset(adev, &reset_context);
+1
drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
+1
drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
···
71
71
reset_context.method = AMD_RESET_METHOD_NONE;
72
72
reset_context.reset_req_dev = adev;
73
73
clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
74
+
clear_bit(AMDGPU_SKIP_MODE2_RESET, &reset_context.flags);
74
75
75
76
r = amdgpu_device_gpu_recover(ring->adev, job, &reset_context);
76
77
if (r)
+1
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
+1
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
···
1949
1949
reset_context.method = AMD_RESET_METHOD_NONE;
1950
1950
reset_context.reset_req_dev = adev;
1951
1951
clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
1952
+
clear_bit(AMDGPU_SKIP_MODE2_RESET, &reset_context.flags);
1952
1953
1953
1954
amdgpu_device_gpu_recover(ras->adev, NULL, &reset_context);
1954
1955
}
+6
drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c
+6
drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c
···
74
74
{
75
75
struct amdgpu_reset_handler *reset_handler = NULL;
76
76
77
+
if (test_bit(AMDGPU_SKIP_MODE2_RESET, &reset_context->flags))
78
+
return -ENOSYS;
79
+
77
80
if (adev->reset_cntl && adev->reset_cntl->get_reset_handler)
78
81
reset_handler = adev->reset_cntl->get_reset_handler(
79
82
adev->reset_cntl, reset_context);
···
92
89
{
93
90
int ret;
94
91
struct amdgpu_reset_handler *reset_handler = NULL;
92
+
93
+
if (test_bit(AMDGPU_SKIP_MODE2_RESET, &reset_context->flags))
94
+
return -ENOSYS;
95
95
96
96
if (adev->reset_cntl)
97
97
reset_handler = adev->reset_cntl->get_reset_handler(
+1
drivers/gpu/drm/amd/amdgpu/amdgpu_reset.h
+1
drivers/gpu/drm/amd/amdgpu/amdgpu_reset.h
+1
drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c
+1
drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c
···
290
290
reset_context.method = AMD_RESET_METHOD_NONE;
291
291
reset_context.reset_req_dev = adev;
292
292
clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
293
+
clear_bit(AMDGPU_SKIP_MODE2_RESET, &reset_context.flags);
293
294
294
295
amdgpu_device_gpu_recover(adev, NULL, &reset_context);
295
296
}
+1
drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c
+1
drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c
···
317
317
reset_context.method = AMD_RESET_METHOD_NONE;
318
318
reset_context.reset_req_dev = adev;
319
319
clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
320
+
clear_bit(AMDGPU_SKIP_MODE2_RESET, &reset_context.flags);
320
321
321
322
amdgpu_device_gpu_recover(adev, NULL, &reset_context);
322
323
}
+1
drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c
+1
drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c
···
529
529
reset_context.method = AMD_RESET_METHOD_NONE;
530
530
reset_context.reset_req_dev = adev;
531
531
clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
532
+
clear_bit(AMDGPU_SKIP_MODE2_RESET, &reset_context.flags);
532
533
533
534
amdgpu_device_gpu_recover(adev, NULL, &reset_context);
534
535
}