diff options
Diffstat (limited to 'meta-amd-bsp/recipes-kernel/linux/linux-yocto-4.19.8/1912-drm-amdgpu-handle-ras-reset.patch')
| -rw-r--r-- | meta-amd-bsp/recipes-kernel/linux/linux-yocto-4.19.8/1912-drm-amdgpu-handle-ras-reset.patch | 99 |
1 files changed, 99 insertions, 0 deletions
diff --git a/meta-amd-bsp/recipes-kernel/linux/linux-yocto-4.19.8/1912-drm-amdgpu-handle-ras-reset.patch b/meta-amd-bsp/recipes-kernel/linux/linux-yocto-4.19.8/1912-drm-amdgpu-handle-ras-reset.patch new file mode 100644 index 00000000..c318626c --- /dev/null +++ b/meta-amd-bsp/recipes-kernel/linux/linux-yocto-4.19.8/1912-drm-amdgpu-handle-ras-reset.patch @@ -0,0 +1,99 @@ +From 551030b390b6087cf625a648262d5442d46f337a Mon Sep 17 00:00:00 2001 +From: xinhui pan <xinhui.pan@amd.com> +Date: Wed, 8 May 2019 19:12:24 +0800 +Subject: [PATCH 1912/2940] drm/amdgpu: handle ras reset + +add another flag to allow IP do a gpu reset after device init. + +Signed-off-by: xinhui pan <xinhui.pan@amd.com> +Reviewed-by: Alex Deucher <alexander.deucher@amd.com> +--- + drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 36 +++++++++++++++++++++++-- + drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h | 3 +++ + 2 files changed, 37 insertions(+), 2 deletions(-) + +diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +index 5f8e1163a75d..37cb3de08494 100644 +--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c ++++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +@@ -118,7 +118,8 @@ const char *ras_block_string[] = { + #define ras_err_str(i) (ras_error_string[ffs(i)]) + #define ras_block_str(i) (ras_block_string[i]) + +-#define AMDGPU_RAS_FLAG_INIT_BY_VBIOS 1 ++#define AMDGPU_RAS_FLAG_INIT_BY_VBIOS 1 ++#define AMDGPU_RAS_FLAG_INIT_NEED_RESET 2 + #define RAS_DEFAULT_FLAGS (AMDGPU_RAS_FLAG_INIT_BY_VBIOS) + + static void amdgpu_ras_self_test(struct amdgpu_device *adev) +@@ -1358,6 +1359,19 @@ static int amdgpu_ras_recovery_fini(struct amdgpu_device *adev) + } + /* recovery end */ + ++/* return 0 if ras will reset gpu and repost.*/ ++int amdgpu_ras_request_reset_on_boot(struct amdgpu_device *adev, ++ unsigned int block) ++{ ++ struct amdgpu_ras *ras = amdgpu_ras_get_context(adev); ++ ++ if (!ras) ++ return -EINVAL; ++ ++ ras->flags |= AMDGPU_RAS_FLAG_INIT_NEED_RESET; ++ return 0; ++} ++ + /* + * check hardware's ras ability which will be saved in hw_supported. + * if hardware does not support ras, we can skip some ras initializtion and +@@ -1433,7 +1447,12 @@ int amdgpu_ras_init(struct amdgpu_device *adev) + return -EINVAL; + } + +-/* do some init work after IP late init as dependence */ ++/* do some init work after IP late init as dependence. ++ * TODO ++ * gpu reset will re-enable ras, need fint out one way to run it again. ++ * for now, if a gpu reset happened, unless IP enable its ras, the ras state ++ * will be showed as disabled. ++ */ + void amdgpu_ras_post_init(struct amdgpu_device *adev) + { + struct amdgpu_ras *con = amdgpu_ras_get_context(adev); +@@ -1462,6 +1481,19 @@ void amdgpu_ras_post_init(struct amdgpu_device *adev) + } + } + } ++ ++ if (con->flags & AMDGPU_RAS_FLAG_INIT_NEED_RESET) { ++ con->flags &= ~AMDGPU_RAS_FLAG_INIT_NEED_RESET; ++ /* setup ras obj state as disabled. ++ * for init_by_vbios case. ++ * if we want to enable ras, just enable it in a normal way. ++ * If we want do disable it, need setup ras obj as enabled, ++ * then issue another TA disable cmd. ++ * See feature_enable_on_boot ++ */ ++ amdgpu_ras_disable_all_features(adev, 1); ++ amdgpu_ras_reset_gpu(adev, 0); ++ } + } + + /* do some fini work before IP fini as dependence */ +diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +index e60a554656ca..06ef325b61b8 100644 +--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h ++++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +@@ -175,6 +175,9 @@ static inline int amdgpu_ras_is_supported(struct amdgpu_device *adev, + return ras && (ras->supported & (1 << block)); + } + ++int amdgpu_ras_request_reset_on_boot(struct amdgpu_device *adev, ++ unsigned int block); ++ + int amdgpu_ras_query_error_count(struct amdgpu_device *adev, + bool is_ce); + +-- +2.17.1 + |