From a6dcf9a7ccfed57abd44c24cc505b559281d44b9 Mon Sep 17 00:00:00 2001 From: Hawking Zhang Date: Sat, 11 Mar 2023 17:36:11 +0800 Subject: drm/amdgpu: Move umc ras block init to gmc ras sw_init Initialize umc ras block only when umc ip block supports ras. Driver queries ras capabilities after early_init, ras block init needs to be moved to sw_init. Signed-off-by: Hawking Zhang Reviewed-by: Stanley Yang Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c index 12a6826caef4..d15790998882 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c @@ -447,8 +447,15 @@ void amdgpu_gmc_filter_faults_remove(struct amdgpu_device *adev, uint64_t addr, } while (fault->timestamp < tmp); } -int amdgpu_gmc_ras_early_init(struct amdgpu_device *adev) +int amdgpu_gmc_ras_sw_init(struct amdgpu_device *adev) { + int r; + + /* umc ras block */ + r = amdgpu_umc_ras_sw_init(adev); + if (r) + return r; + if (!adev->gmc.xgmi.connected_to_cpu) { adev->gmc.xgmi.ras = &xgmi_ras; amdgpu_ras_register_ras_block(adev, &adev->gmc.xgmi.ras->ras_block); -- cgit From fec70a8601a76272b08b7b8077b3c3e3c79bdb72 Mon Sep 17 00:00:00 2001 From: Hawking Zhang Date: Sat, 4 Mar 2023 16:39:48 +0800 Subject: drm/amdgpu: Move mmhub ras block init to ras sw_init Initialize mmhub ras block only when mmhub ip block supports ras features. Driver queries ras capabilities after early_init, ras block init needs to be moved to sw_init. Signed-off-by: Hawking Zhang Reviewed-by: Stanley Yang Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c index d15790998882..8b2a2718a0d1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c @@ -456,6 +456,11 @@ int amdgpu_gmc_ras_sw_init(struct amdgpu_device *adev) if (r) return r; + /* mmhub ras block */ + r = amdgpu_mmhub_ras_sw_init(adev); + if (r) + return r; + if (!adev->gmc.xgmi.connected_to_cpu) { adev->gmc.xgmi.ras = &xgmi_ras; amdgpu_ras_register_ras_block(adev, &adev->gmc.xgmi.ras->ras_block); -- cgit From 474e2d491efe8ce516e743dbce6a6e75bac3b3db Mon Sep 17 00:00:00 2001 From: Hawking Zhang Date: Sat, 4 Mar 2023 17:24:46 +0800 Subject: drm/amdgpu: Move hdp ras block init to ras sw_init Initialize hdp ras block only when mmhub ip block supports ras features. Driver queries ras capabilities after early_init, ras block init needs to be moved to sw_init. Signed-off-by: Hawking Zhang Reviewed-by: Stanley Yang Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c index 8b2a2718a0d1..d5869d121299 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c @@ -461,6 +461,11 @@ int amdgpu_gmc_ras_sw_init(struct amdgpu_device *adev) if (r) return r; + /* hdp ras block */ + r = amdgpu_hdp_ras_sw_init(adev); + if (r) + return r; + if (!adev->gmc.xgmi.connected_to_cpu) { adev->gmc.xgmi.ras = &xgmi_ras; amdgpu_ras_register_ras_block(adev, &adev->gmc.xgmi.ras->ras_block); -- cgit From 7f544c5488cf5bc94b379de750e08fa3e146b6ba Mon Sep 17 00:00:00 2001 From: Hawking Zhang Date: Wed, 15 Mar 2023 08:59:04 +0800 Subject: drm/amdgpu: Rework mca ras sw_init To align with other IP blocks Signed-off-by: Hawking Zhang Reviewed-by: Stanley Yang Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c | 13 +++++++++++++ 1 file changed, 13 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c index d5869d121299..cc8e16194ba1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c @@ -466,6 +466,19 @@ int amdgpu_gmc_ras_sw_init(struct amdgpu_device *adev) if (r) return r; + /* mca.x ras block */ + r = amdgpu_mca_mp0_ras_sw_init(adev); + if (r) + return r; + + r = amdgpu_mca_mp1_ras_sw_init(adev); + if (r) + return r; + + r = amdgpu_mca_mpio_ras_sw_init(adev); + if (r) + return r; + if (!adev->gmc.xgmi.connected_to_cpu) { adev->gmc.xgmi.ras = &xgmi_ras; amdgpu_ras_register_ras_block(adev, &adev->gmc.xgmi.ras->ras_block); -- cgit From da9d669eab152dbd6e9410606a7c8c8a212a7959 Mon Sep 17 00:00:00 2001 From: Hawking Zhang Date: Sat, 4 Mar 2023 19:54:14 +0800 Subject: drm/amdgpu: Rework xgmi_wafl_pcs ras sw_init To align with other IP blocks. Signed-off-by: Hawking Zhang Reviewed-by: Stanley Yang Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c | 9 ++++----- 1 file changed, 4 insertions(+), 5 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c index cc8e16194ba1..655fc8bf936d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c @@ -479,11 +479,10 @@ int amdgpu_gmc_ras_sw_init(struct amdgpu_device *adev) if (r) return r; - if (!adev->gmc.xgmi.connected_to_cpu) { - adev->gmc.xgmi.ras = &xgmi_ras; - amdgpu_ras_register_ras_block(adev, &adev->gmc.xgmi.ras->ras_block); - adev->gmc.xgmi.ras_if = &adev->gmc.xgmi.ras->ras_block.ras_comm; - } + /* xgmi ras block */ + r = amdgpu_xgmi_ras_sw_init(adev); + if (r) + return r; return 0; } -- cgit From cebbfdd5f099a16d57d9221cec76882867839ad3 Mon Sep 17 00:00:00 2001 From: Amber Lin Date: Wed, 9 Jun 2021 12:36:35 -0400 Subject: drm/amdkfd: Set noretry/xnack for GC 9.4.3 For GC 9.4.3, disable retry as default and XNACK can be different modes per process. Signed-off-by: Amber Lin Reviewed-by: Felix Kuehling Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c | 1 + 1 file changed, 1 insertion(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c index 655fc8bf936d..88bc7f5f46e6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c @@ -624,6 +624,7 @@ void amdgpu_gmc_noretry_set(struct amdgpu_device *adev) gc_ver == IP_VERSION(9, 4, 0) || gc_ver == IP_VERSION(9, 4, 1) || gc_ver == IP_VERSION(9, 4, 2) || + gc_ver == IP_VERSION(9, 4, 3) || gc_ver >= IP_VERSION(10, 3, 0)); gmc->noretry = (amdgpu_noretry == -1) ? noretry_default : amdgpu_noretry; -- cgit From dd299441654fd8209056c7985ddf2373ebaba6ed Mon Sep 17 00:00:00 2001 From: Mukul Joshi Date: Tue, 11 Apr 2023 16:32:38 -0400 Subject: drm/amdgpu: Rework retry fault removal Rework retry fault removal from the software filter by storing an expired timestamp for a fault that is being removed. When a new fault comes, and it matches an entry in the sw filter, it will be added as a new fault only when its timestamp is greater than the timestamp expiry of the fault in the sw filter. This helps in avoiding stale faults being added back into the filter and preventing legitimate faults from being handled. Suggested-by: Felix Kuehling Signed-off-by: Mukul Joshi Reviewed-by: Philip Yang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c | 36 ++++++++++++++++++++++++++++++--- 1 file changed, 33 insertions(+), 3 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c index 88bc7f5f46e6..9b0ccb1b84c6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c @@ -395,8 +395,21 @@ bool amdgpu_gmc_filter_faults(struct amdgpu_device *adev, while (fault->timestamp >= stamp) { uint64_t tmp; - if (atomic64_read(&fault->key) == key) - return true; + if (atomic64_read(&fault->key) == key) { + /* + * if we get a fault which is already present in + * the fault_ring and the timestamp of + * the fault is after the expired timestamp, + * then this is a new fault that needs to be added + * into the fault ring. + */ + if (fault->timestamp_expiry != 0 && + amdgpu_ih_ts_after(fault->timestamp_expiry, + timestamp)) + break; + else + return true; + } tmp = fault->timestamp; fault = &gmc->fault_ring[fault->next]; @@ -432,15 +445,32 @@ void amdgpu_gmc_filter_faults_remove(struct amdgpu_device *adev, uint64_t addr, { struct amdgpu_gmc *gmc = &adev->gmc; uint64_t key = amdgpu_gmc_fault_key(addr, pasid); + struct amdgpu_ih_ring *ih; struct amdgpu_gmc_fault *fault; + uint32_t last_wptr; + uint64_t last_ts; uint32_t hash; uint64_t tmp; + ih = adev->irq.retry_cam_enabled ? &adev->irq.ih_soft : &adev->irq.ih1; + /* Get the WPTR of the last entry in IH ring */ + last_wptr = amdgpu_ih_get_wptr(adev, ih); + /* Order wptr with ring data. */ + rmb(); + /* Get the timetamp of the last entry in IH ring */ + last_ts = amdgpu_ih_decode_iv_ts(adev, ih, last_wptr, -1); + hash = hash_64(key, AMDGPU_GMC_FAULT_HASH_ORDER); fault = &gmc->fault_ring[gmc->fault_hash[hash].idx]; do { - if (atomic64_cmpxchg(&fault->key, key, 0) == key) + if (atomic64_read(&fault->key) == key) { + /* + * Update the timestamp when this fault + * expired. + */ + fault->timestamp_expiry = last_ts; break; + } tmp = fault->timestamp; fault = &gmc->fault_ring[fault->next]; -- cgit From 0530553ba842884737a689ae5fac11154dcf3122 Mon Sep 17 00:00:00 2001 From: Le Ma Date: Thu, 19 May 2022 19:49:59 +0800 Subject: drm/amdgpu: move vmhub out of amdgpu_ring_funcs (v4) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit It looks better to place this field in ring structure. Also drop the repeated ring funcs definitions if there's no difference except for vmhub field. v2: rename the field to vm_hub like others (Le) v3: apply the changes to new ip blocks (Hawking) v4: fix vcn sw ring (Alex) Signed-off-by: Le Ma Reviewed-by: Christian König Reviewed-by: Hawking Zhang Signed-off-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c index 9b0ccb1b84c6..4e2531758866 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c @@ -554,7 +554,7 @@ int amdgpu_gmc_allocate_vm_inv_eng(struct amdgpu_device *adev) for (i = 0; i < adev->num_rings; ++i) { ring = adev->rings[i]; - vmhub = ring->funcs->vmhub; + vmhub = ring->vm_hub; if (ring == &adev->mes.ring) continue; @@ -570,7 +570,7 @@ int amdgpu_gmc_allocate_vm_inv_eng(struct amdgpu_device *adev) vm_inv_engs[vmhub] &= ~(1 << ring->vm_inv_eng); dev_info(adev->dev, "ring %s uses VM inv eng %u on hub %u\n", - ring->name, ring->vm_inv_eng, ring->funcs->vmhub); + ring->name, ring->vm_inv_eng, ring->vm_hub); } return 0; -- cgit From 663b930e24842f3d3bb79418bb5cd8d01b40c559 Mon Sep 17 00:00:00 2001 From: Ikshwaku Chauhan Date: Thu, 25 May 2023 10:57:26 +0530 Subject: drm/amdgpu: enable tmz by default for GC 11.0.1 Add IP GC 11.0.1 in the list of target to have tmz enabled by default. Signed-off-by: Ikshwaku Chauhan Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher Cc: stable@vger.kernel.org # 6.1.x --- drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c index 4e2531758866..95b0f984acbf 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c @@ -593,6 +593,8 @@ void amdgpu_gmc_tmz_set(struct amdgpu_device *adev) case IP_VERSION(9, 3, 0): /* GC 10.3.7 */ case IP_VERSION(10, 3, 7): + /* GC 11.0.1 */ + case IP_VERSION(11, 0, 1): if (amdgpu_tmz == 0) { adev->gmc.tmz_enabled = false; dev_info(adev->dev, @@ -616,7 +618,6 @@ void amdgpu_gmc_tmz_set(struct amdgpu_device *adev) case IP_VERSION(10, 3, 1): /* YELLOW_CARP*/ case IP_VERSION(10, 3, 3): - case IP_VERSION(11, 0, 1): case IP_VERSION(11, 0, 4): /* Don't enable it by default yet. */ -- cgit