From 672c0218e3e22ccaeb2911da8d3b784d3b6cc1d8 Mon Sep 17 00:00:00 2001 From: Victor Zhao Date: Thu, 28 Jul 2022 10:44:47 +0800 Subject: drm/amdgpu: add mode2 reset for sienna_cichlid To meet the requirement for multi container usecase which needs a quicker reset and not causing VRAM lost, adding the Mode2 reset handler for sienna_cichlid. v2: move skip mode2 flag part separately v3: remove the use of asic_reset_res Signed-off-by: Victor Zhao Acked-by: Andrey Grodzovsky Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c index 32c86a0b145c..f778466bb9db 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c @@ -23,6 +23,7 @@ #include "amdgpu_reset.h" #include "aldebaran.h" +#include "sienna_cichlid.h" int amdgpu_reset_add_handler(struct amdgpu_reset_control *reset_ctl, struct amdgpu_reset_handler *handler) @@ -40,6 +41,9 @@ int amdgpu_reset_init(struct amdgpu_device *adev) case IP_VERSION(13, 0, 2): ret = aldebaran_reset_init(adev); break; + case IP_VERSION(11, 0, 7): + ret = sienna_cichlid_reset_init(adev); + break; default: break; } @@ -55,6 +59,9 @@ int amdgpu_reset_fini(struct amdgpu_device *adev) case IP_VERSION(13, 0, 2): ret = aldebaran_reset_fini(adev); break; + case IP_VERSION(11, 0, 7): + ret = sienna_cichlid_reset_fini(adev); + break; default: break; } -- cgit v1.2.3-59-g8ed1b From dac6b80818ac2353631c5a33d140d8d5508e2957 Mon Sep 17 00:00:00 2001 From: Victor Zhao Date: Thu, 28 Jul 2022 10:39:23 +0800 Subject: 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 Acked-by: Andrey Grodzovsky Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 7 ++++++- drivers/gpu/drm/amd/amdgpu/amdgpu_job.c | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c | 6 ++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_reset.h | 1 + drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c | 1 + drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c | 1 + drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c | 1 + 9 files changed, 19 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c index 5e53a5293935..091415a4abf0 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c @@ -135,6 +135,7 @@ static void amdgpu_amdkfd_reset_work(struct work_struct *work) reset_context.method = AMD_RESET_METHOD_NONE; reset_context.reset_req_dev = adev; clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags); + clear_bit(AMDGPU_SKIP_MODE2_RESET, &reset_context.flags); amdgpu_device_gpu_recover(adev, NULL, &reset_context); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index e8a0b19b7398..c1ec4b653ca4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -5148,6 +5148,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, reset_context->job = job; reset_context->hive = hive; + /* * Build list of devices to reset. * In case we are in XGMI hive mode, resort the device list @@ -5267,8 +5268,11 @@ retry: /* Rest of adevs pre asic reset from XGMI hive. */ amdgpu_ras_resume(adev); } else { r = amdgpu_do_asic_reset(device_list_handle, reset_context); - if (r && r == -EAGAIN) + if (r && r == -EAGAIN) { + set_bit(AMDGPU_SKIP_MODE2_RESET, &reset_context->flags); + adev->asic_reset_res = 0; goto retry; + } } skip_hw_reset: @@ -5699,6 +5703,7 @@ pci_ers_result_t amdgpu_pci_slot_reset(struct pci_dev *pdev) reset_context.reset_req_dev = adev; set_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags); set_bit(AMDGPU_SKIP_HW_RESET, &reset_context.flags); + set_bit(AMDGPU_SKIP_MODE2_RESET, &reset_context.flags); adev->no_hw_access = true; r = amdgpu_device_pre_asic_reset(adev, &reset_context); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c index b1099ee79c50..36f49da15800 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c @@ -71,6 +71,7 @@ static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job) reset_context.method = AMD_RESET_METHOD_NONE; reset_context.reset_req_dev = adev; clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags); + clear_bit(AMDGPU_SKIP_MODE2_RESET, &reset_context.flags); r = amdgpu_device_gpu_recover(ring->adev, job, &reset_context); if (r) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index ff5361f5c2d4..ab9ba5a9c33d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -1949,6 +1949,7 @@ static void amdgpu_ras_do_recovery(struct work_struct *work) reset_context.method = AMD_RESET_METHOD_NONE; reset_context.reset_req_dev = adev; clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags); + clear_bit(AMDGPU_SKIP_MODE2_RESET, &reset_context.flags); amdgpu_device_gpu_recover(ras->adev, NULL, &reset_context); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c index f778466bb9db..831fb222139c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c @@ -74,6 +74,9 @@ int amdgpu_reset_prepare_hwcontext(struct amdgpu_device *adev, { struct amdgpu_reset_handler *reset_handler = NULL; + if (test_bit(AMDGPU_SKIP_MODE2_RESET, &reset_context->flags)) + return -ENOSYS; + if (adev->reset_cntl && adev->reset_cntl->get_reset_handler) reset_handler = adev->reset_cntl->get_reset_handler( adev->reset_cntl, reset_context); @@ -90,6 +93,9 @@ int amdgpu_reset_perform_reset(struct amdgpu_device *adev, int ret; struct amdgpu_reset_handler *reset_handler = NULL; + if (test_bit(AMDGPU_SKIP_MODE2_RESET, &reset_context->flags)) + return -ENOSYS; + if (adev->reset_cntl) reset_handler = adev->reset_cntl->get_reset_handler( adev->reset_cntl, reset_context); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.h index ffda1560c648..f71b83c42590 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.h @@ -30,6 +30,7 @@ enum AMDGPU_RESET_FLAGS { AMDGPU_NEED_FULL_RESET = 0, AMDGPU_SKIP_HW_RESET = 1, + AMDGPU_SKIP_MODE2_RESET = 2, }; struct amdgpu_reset_context { diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c index 12906ba74462..a2f04b249132 100644 --- a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c +++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c @@ -290,6 +290,7 @@ flr_done: reset_context.method = AMD_RESET_METHOD_NONE; reset_context.reset_req_dev = adev; clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags); + clear_bit(AMDGPU_SKIP_MODE2_RESET, &reset_context.flags); amdgpu_device_gpu_recover(adev, NULL, &reset_context); } diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c b/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c index e07757eea7ad..a977f0027928 100644 --- a/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c +++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c @@ -317,6 +317,7 @@ flr_done: reset_context.method = AMD_RESET_METHOD_NONE; reset_context.reset_req_dev = adev; clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags); + clear_bit(AMDGPU_SKIP_MODE2_RESET, &reset_context.flags); amdgpu_device_gpu_recover(adev, NULL, &reset_context); } diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c b/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c index 288c414babdf..fd14fa9b9cd7 100644 --- a/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c +++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c @@ -529,6 +529,7 @@ static void xgpu_vi_mailbox_flr_work(struct work_struct *work) reset_context.method = AMD_RESET_METHOD_NONE; reset_context.reset_req_dev = adev; clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags); + clear_bit(AMDGPU_SKIP_MODE2_RESET, &reset_context.flags); amdgpu_device_gpu_recover(adev, NULL, &reset_context); } -- cgit v1.2.3-59-g8ed1b From 5bd8d53f6fa53eab5433698d1362dae2aa53c1cc Mon Sep 17 00:00:00 2001 From: Victor Zhao Date: Tue, 14 Jun 2022 10:48:11 +0800 Subject: drm/amdgpu: add debugfs amdgpu_reset_level Introduce amdgpu_reset_level debugfs in order to help debug and test specific type of reset. Also helps blocking unwanted type of resets. By default, mode2 reset will not be enabled v2: make this debugfs in adev and use debugfs_create_u32 Signed-off-by: Victor Zhao Acked-by: Andrey Grodzovsky Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 5 +++++ drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c | 2 ++ drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c | 8 ++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c | 3 +++ 4 files changed, 18 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index d597e2656c47..66a26f2c51c1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -274,6 +274,9 @@ extern int amdgpu_vcnfw_log; #define AMDGPU_RESET_VCE (1 << 13) #define AMDGPU_RESET_VCE1 (1 << 14) +#define AMDGPU_RESET_LEVEL_SOFT_RECOVERY (1 << 0) +#define AMDGPU_RESET_LEVEL_MODE2 (1 << 1) + /* max cursor sizes (in pixels) */ #define CIK_CURSOR_WIDTH 128 #define CIK_CURSOR_HEIGHT 128 @@ -1060,6 +1063,8 @@ struct amdgpu_device { uint32_t scpm_status; struct work_struct reset_work; + + uint32_t amdgpu_reset_level_mask; }; static inline struct amdgpu_device *drm_to_adev(struct drm_device *ddev) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c index cb00c7d6f50b..4e54569fa223 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c @@ -1786,6 +1786,8 @@ int amdgpu_debugfs_init(struct amdgpu_device *adev) return PTR_ERR(ent); } + debugfs_create_u32("amdgpu_reset_level", 0600, root, &adev->amdgpu_reset_level_mask); + /* Register debugfs entries for amdgpu_ttm */ amdgpu_ttm_debugfs_init(adev); amdgpu_debugfs_pm_init(adev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c index 831fb222139c..9da5ead50c90 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c @@ -37,6 +37,8 @@ int amdgpu_reset_init(struct amdgpu_device *adev) { int ret = 0; + adev->amdgpu_reset_level_mask = 0x1; + switch (adev->ip_versions[MP1_HWIP][0]) { case IP_VERSION(13, 0, 2): ret = aldebaran_reset_init(adev); @@ -74,6 +76,9 @@ int amdgpu_reset_prepare_hwcontext(struct amdgpu_device *adev, { struct amdgpu_reset_handler *reset_handler = NULL; + if (!(adev->amdgpu_reset_level_mask & AMDGPU_RESET_LEVEL_MODE2)) + return -ENOSYS; + if (test_bit(AMDGPU_SKIP_MODE2_RESET, &reset_context->flags)) return -ENOSYS; @@ -93,6 +98,9 @@ int amdgpu_reset_perform_reset(struct amdgpu_device *adev, int ret; struct amdgpu_reset_handler *reset_handler = NULL; + if (!(adev->amdgpu_reset_level_mask & AMDGPU_RESET_LEVEL_MODE2)) + return -ENOSYS; + if (test_bit(AMDGPU_SKIP_MODE2_RESET, &reset_context->flags)) return -ENOSYS; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c index d3558c34d406..3e316b013fd9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c @@ -405,6 +405,9 @@ bool amdgpu_ring_soft_recovery(struct amdgpu_ring *ring, unsigned int vmid, { ktime_t deadline = ktime_add_us(ktime_get(), 10000); + if (!(ring->adev->amdgpu_reset_level_mask & AMDGPU_RESET_LEVEL_SOFT_RECOVERY)) + return false; + if (amdgpu_sriov_vf(ring->adev) || !ring->funcs->soft_recovery || !fence) return false; -- cgit v1.2.3-59-g8ed1b