From f957138cc30ae904346f117013fc7b428700c1f7 Mon Sep 17 00:00:00 2001 From: Hawking Zhang Date: Mon, 31 Jul 2023 23:43:12 +0800 Subject: drm/amdgpu: Only create err_count sysfs when hw_op is supported Some IP blocks only support partial ras feature and don't have ras counter and/or ras error status register at all. Driver should not create err_count sysfs node for those IP blocks. Signed-off-by: Hawking Zhang Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 31 ++++++++++++++++++------------- 1 file changed, 18 insertions(+), 13 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 62011a521833..fa53a51db69f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -2777,23 +2777,28 @@ int amdgpu_ras_block_late_init(struct amdgpu_device *adev, goto cleanup; } - r = amdgpu_ras_sysfs_create(adev, ras_block); - if (r) - goto interrupt; + if (ras_obj->hw_ops && + (ras_obj->hw_ops->query_ras_error_count || + ras_obj->hw_ops->query_ras_error_status)) { + r = amdgpu_ras_sysfs_create(adev, ras_block); + if (r) + goto interrupt; - /* Those are the cached values at init. - */ - query_info = kzalloc(sizeof(struct ras_query_if), GFP_KERNEL); - if (!query_info) - return -ENOMEM; - memcpy(&query_info->head, ras_block, sizeof(struct ras_common_if)); + /* Those are the cached values at init. + */ + query_info = kzalloc(sizeof(*query_info), GFP_KERNEL); + if (!query_info) + return -ENOMEM; + memcpy(&query_info->head, ras_block, sizeof(struct ras_common_if)); - if (amdgpu_ras_query_error_count(adev, &ce_count, &ue_count, query_info) == 0) { - atomic_set(&con->ras_ce_count, ce_count); - atomic_set(&con->ras_ue_count, ue_count); + if (amdgpu_ras_query_error_count(adev, &ce_count, &ue_count, query_info) == 0) { + atomic_set(&con->ras_ce_count, ce_count); + atomic_set(&con->ras_ue_count, ue_count); + } + + kfree(query_info); } - kfree(query_info); return 0; interrupt: -- cgit From 62c4b772bdd91ed1745c4a55b681e94aa9e48071 Mon Sep 17 00:00:00 2001 From: Hawking Zhang Date: Mon, 3 Jul 2023 14:38:45 +0800 Subject: drm/amdgpu: Apply poison mode check to GFX IP only For GFX IP that only supports poison consumption, GFX RAS won't be marked as enabled. i.e., hardware doesn't support gfx sram ecc. But driver still needs to issue firmware to enable poison consumption mode for GFX IP. In such case, check poison mode and treat GFX IP as RAS capable IP block. Signed-off-by: Hawking Zhang Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 1 + 1 file changed, 1 insertion(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index fa53a51db69f..664b841f5cec 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -3153,6 +3153,7 @@ int amdgpu_ras_is_supported(struct amdgpu_device *adev, * that the ras block supports ras function. */ if (!ret && + block == AMDGPU_RAS_BLOCK__GFX && amdgpu_ras_is_poison_mode_supported(adev) && amdgpu_ras_get_ras_block(adev, block, 0)) ret = 1; -- cgit From 6fc9d92c3d27f572eb1d884662f199c0d7f90d16 Mon Sep 17 00:00:00 2001 From: Hawking Zhang Date: Mon, 3 Jul 2023 16:17:46 +0800 Subject: drm/amdgpu: Issue ras enable_feature for gfx ip only For non-GFX IP blocks, set up ras obj if ras feature is allowed. For GFX IP blocks, force issue ras enable_feature command to firmware and only set up ras obj if ras feature is allowed Signed-off-by: Hawking Zhang Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 30 ++++++++++-------------------- 1 file changed, 10 insertions(+), 20 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 664b841f5cec..bb29cb57add5 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -757,16 +757,6 @@ static int __amdgpu_ras_feature_enable(struct amdgpu_device *adev, return 0; } -static int amdgpu_ras_check_feature_allowed(struct amdgpu_device *adev, - struct ras_common_if *head) -{ - if (amdgpu_ras_is_feature_allowed(adev, head) || - amdgpu_ras_is_poison_mode_supported(adev)) - return 1; - else - return 0; -} - /* wrapper of psp_ras_enable_features */ int amdgpu_ras_feature_enable(struct amdgpu_device *adev, struct ras_common_if *head, bool enable) @@ -778,7 +768,16 @@ int amdgpu_ras_feature_enable(struct amdgpu_device *adev, if (!con) return -EINVAL; - if (head->block == AMDGPU_RAS_BLOCK__GFX) { + /* Do not enable ras feature if it is not allowed */ + if (enable && + head->block != AMDGPU_RAS_BLOCK__GFX && + !amdgpu_ras_is_feature_allowed(adev, head)) + goto out; + + /* Only enable gfx ras feature from host side */ + if (head->block == AMDGPU_RAS_BLOCK__GFX && + !amdgpu_sriov_vf(adev) && + !amdgpu_ras_intr_triggered()) { info = kzalloc(sizeof(union ta_ras_cmd_input), GFP_KERNEL); if (!info) return -ENOMEM; @@ -794,16 +793,7 @@ int amdgpu_ras_feature_enable(struct amdgpu_device *adev, .error_type = amdgpu_ras_error_to_ta(head->type), }; } - } - /* Do not enable if it is not allowed. */ - if (enable && !amdgpu_ras_check_feature_allowed(adev, head)) - goto out; - - /* Only enable ras feature operation handle on host side */ - if (head->block == AMDGPU_RAS_BLOCK__GFX && - !amdgpu_sriov_vf(adev) && - !amdgpu_ras_intr_triggered()) { ret = psp_ras_enable_features(&adev->psp, info, enable); if (ret) { dev_err(adev->dev, "ras %s %s failed poison:%d ret:%d\n", -- cgit From 7692e1ee2446fd1940b5caa6e09779504a58881a Mon Sep 17 00:00:00 2001 From: Tao Zhou Date: Thu, 8 Jun 2023 16:25:42 +0800 Subject: drm/amdgpu: add RAS fatal error handler for NBIO v7.9 Register RAS fatal error interrupt and add handler. v2: only register NBIO RAS for dGPU platform. change nbio_v7_9_set_ras_controller_irq_state and nbio_v7_9_set_ras_err_event_athub_irq_state to dummy functions. Signed-off-by: Tao Zhou Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index bb29cb57add5..00658c2816dc 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -35,6 +35,7 @@ #include "amdgpu_xgmi.h" #include "ivsrcid/nbio/irqsrcs_nbif_7_4.h" #include "nbio_v4_3.h" +#include "nbio_v7_9.h" #include "atom.h" #include "amdgpu_reset.h" @@ -2644,6 +2645,10 @@ int amdgpu_ras_init(struct amdgpu_device *adev) * check DF RAS */ adev->nbio.ras = &nbio_v4_3_ras; break; + case IP_VERSION(7, 9, 0): + if (!adev->gmc.is_app_apu) + adev->nbio.ras = &nbio_v7_9_ras; + break; default: /* nbio ras is not available */ break; -- cgit From bc0f80802d735950738167e2cc0b51b0dd41e68d Mon Sep 17 00:00:00 2001 From: Candice Li Date: Tue, 8 Aug 2023 16:59:18 +0800 Subject: drm/amdgpu: Extend poison mode check to SDMA/VCN/JPEG Treat SDMA/VCN/JPEG as RAS capable IP blocks in poison mode. Signed-off-by: Candice Li Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 00658c2816dc..886c7c6f1c3b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -3148,7 +3148,10 @@ int amdgpu_ras_is_supported(struct amdgpu_device *adev, * that the ras block supports ras function. */ if (!ret && - block == AMDGPU_RAS_BLOCK__GFX && + (block == AMDGPU_RAS_BLOCK__GFX || + block == AMDGPU_RAS_BLOCK__SDMA || + block == AMDGPU_RAS_BLOCK__VCN || + block == AMDGPU_RAS_BLOCK__JPEG) && amdgpu_ras_is_poison_mode_supported(adev) && amdgpu_ras_get_ras_block(adev, block, 0)) ret = 1; -- cgit