On 2017年05月16日 18:49, Christian König wrote:
Am 16.05.2017 um 11:25 schrieb Chunming Zhou:
bakup first 64 byte of gart table as reset magic, check if magic is same
after gpu hw reset.
v2: use memcmp instead of manual innovation.
Change-Id: I9a73720da4084ea8677c3031dfb62e8157ee5704
Signed-off-by: Chunming Zhou <david1.z...@amd.com>
Patch #1-#3 are Reviewed-by: Christian König <christian.koe...@amd.com>
pushed.
Patch #4:
You need to add the new enum on line 591 or otherwise you will get an
"unsupported operation" error.
Line 604 should be changed as well or otherwise we need a BO for this
operation.
are you sure you are talking this patch#4? I cannot address what you said.
A libdrm test case to just call this IOCTL would probably be a good idea.
Additional to that I would ping Marek (Mesa) and Michel (DDX) for
their opinion on this. Could be that this is completely superfluous
and the UMDs needs something else.
Michel seems have different opinion/concern, maybe we need more
discussions before we make new interfaces.
Thanks,
David Zhou
Regards,
Christian.
---
drivers/gpu/drm/amd/amdgpu/amdgpu.h | 2 ++
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 20 +++++++++++++++++++-
2 files changed, 21 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
index de08ff0..f9da215 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
@@ -1502,6 +1502,7 @@ struct amdgpu_ssg {
#endif
};
+#define AMDGPU_RESET_MAGIC_NUM 64
struct amdgpu_device {
struct device *dev;
struct drm_device *ddev;
@@ -1705,6 +1706,7 @@ struct amdgpu_device {
/* record hw reset is performed */
bool has_hw_reset;
+ u8 reset_magic[AMDGPU_RESET_MAGIC_NUM];
};
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index 0a31fb1..c56ae4a 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -1685,6 +1685,17 @@ static int amdgpu_init(struct amdgpu_device
*adev)
return 0;
}
+static void amdgpu_fill_reset_magic(struct amdgpu_device *adev)
+{
+ memcpy(adev->reset_magic, adev->gart.ptr, AMDGPU_RESET_MAGIC_NUM);
+}
+
+static bool amdgpu_check_vram_lost(struct amdgpu_device *adev)
+{
+ return !!memcmp(adev->gart.ptr, adev->reset_magic,
+ AMDGPU_RESET_MAGIC_NUM);
+}
+
static int amdgpu_late_init(struct amdgpu_device *adev)
{
int i = 0, r;
@@ -1715,6 +1726,8 @@ static int amdgpu_late_init(struct
amdgpu_device *adev)
}
}
+ amdgpu_fill_reset_magic(adev);
+
return 0;
}
@@ -2830,7 +2843,7 @@ int amdgpu_gpu_reset(struct amdgpu_device *adev)
struct drm_atomic_state *state = NULL;
int i, r;
int resched;
- bool need_full_reset;
+ bool need_full_reset, vram_lost = false;
if (amdgpu_sriov_vf(adev))
return amdgpu_sriov_gpu_reset(adev, true);
@@ -2899,12 +2912,17 @@ int amdgpu_gpu_reset(struct amdgpu_device *adev)
r = amdgpu_resume_phase1(adev);
if (r)
goto out;
+ vram_lost = amdgpu_check_vram_lost(adev);
+ if (vram_lost)
+ DRM_ERROR("VRAM is lost!\n");
r = amdgpu_ttm_recover_gart(adev);
if (r)
goto out;
r = amdgpu_resume_phase2(adev);
if (r)
goto out;
+ if (vram_lost)
+ amdgpu_fill_reset_magic(adev);
}
}
out:
_______________________________________________
amd-gfx mailing list
amd-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/amd-gfx