From a9a83a92d0da38a78c0afda096c0a4a5f39f9480 Mon Sep 17 00:00:00 2001 From: Dennis Li Date: Wed, 2 Sep 2020 12:57:59 +0800 Subject: drm/kfd: fix a system crash issue during GPU recovery The crash log as the below: [Thu Aug 20 23:18:14 2020] general protection fault: 0000 [#1] SMP NOPTI [Thu Aug 20 23:18:14 2020] CPU: 152 PID: 1837 Comm: kworker/152:1 Tainted: G OE 5.4.0-42-generic #46~18.04.1-Ubuntu [Thu Aug 20 23:18:14 2020] Hardware name: GIGABYTE G482-Z53-YF/MZ52-G40-00, BIOS R12 05/13/2020 [Thu Aug 20 23:18:14 2020] Workqueue: events amdgpu_ras_do_recovery [amdgpu] [Thu Aug 20 23:18:14 2020] RIP: 0010:evict_process_queues_cpsch+0xc9/0x130 [amdgpu] [Thu Aug 20 23:18:14 2020] Code: 49 8d 4d 10 48 39 c8 75 21 eb 44 83 fa 03 74 36 80 78 72 00 74 0c 83 ab 68 01 00 00 01 41 c6 45 41 00 48 8b 00 48 39 c8 74 25 <80> 78 70 00 c6 40 6d 01 74 ee 8b 50 28 c6 40 70 00 83 ab 60 01 00 [Thu Aug 20 23:18:14 2020] RSP: 0018:ffffb29b52f6fc90 EFLAGS: 00010213 [Thu Aug 20 23:18:14 2020] RAX: 1c884edb0a118914 RBX: ffff8a0d45ff3c00 RCX: ffff8a2d83e41038 [Thu Aug 20 23:18:14 2020] RDX: 0000000000000000 RSI: 0000000000000082 RDI: ffff8a0e2e4178c0 [Thu Aug 20 23:18:14 2020] RBP: ffffb29b52f6fcb0 R08: 0000000000001b64 R09: 0000000000000004 [Thu Aug 20 23:18:14 2020] R10: ffffb29b52f6fb78 R11: 0000000000000001 R12: ffff8a0d45ff3d28 [Thu Aug 20 23:18:14 2020] R13: ffff8a2d83e41028 R14: 0000000000000000 R15: 0000000000000000 [Thu Aug 20 23:18:14 2020] FS: 0000000000000000(0000) GS:ffff8a0e2e400000(0000) knlGS:0000000000000000 [Thu Aug 20 23:18:14 2020] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [Thu Aug 20 23:18:14 2020] CR2: 000055c783c0e6a8 CR3: 00000034a1284000 CR4: 0000000000340ee0 [Thu Aug 20 23:18:14 2020] Call Trace: [Thu Aug 20 23:18:14 2020] kfd_process_evict_queues+0x43/0xd0 [amdgpu] [Thu Aug 20 23:18:14 2020] kfd_suspend_all_processes+0x60/0xf0 [amdgpu] [Thu Aug 20 23:18:14 2020] kgd2kfd_suspend.part.7+0x43/0x50 [amdgpu] [Thu Aug 20 23:18:14 2020] kgd2kfd_pre_reset+0x46/0x60 [amdgpu] [Thu Aug 20 23:18:14 2020] amdgpu_amdkfd_pre_reset+0x1a/0x20 [amdgpu] [Thu Aug 20 23:18:14 2020] amdgpu_device_gpu_recover+0x377/0xf90 [amdgpu] [Thu Aug 20 23:18:14 2020] ? amdgpu_ras_error_query+0x1b8/0x2a0 [amdgpu] [Thu Aug 20 23:18:14 2020] amdgpu_ras_do_recovery+0x159/0x190 [amdgpu] [Thu Aug 20 23:18:14 2020] process_one_work+0x20f/0x400 [Thu Aug 20 23:18:14 2020] worker_thread+0x34/0x410 When GPU hang, user process will fail to create a compute queue whose struct object will be freed later, but driver wrongly add this queue to queue list of the proccess. And then kfd_process_evict_queues will access a freed memory, which cause a system crash. v2: The failure to execute_queues should probably not be reported to the caller of create_queue, because the queue was already created. Therefore change to ignore the return value from execute_queues. Reviewed-by: Felix Kuehling Signed-off-by: Dennis Li Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers') diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c index 560adc57a050..069ba4be1e8f 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c @@ -1302,7 +1302,7 @@ static int create_queue_cpsch(struct device_queue_manager *dqm, struct queue *q, if (q->properties.is_active) { increment_queue_count(dqm, q->properties.type); - retval = execute_queues_cpsch(dqm, + execute_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0); } -- cgit v1.2.3 From edb084f487a3b7c86efce81bfa1a7aa666dd4047 Mon Sep 17 00:00:00 2001 From: Dennis Li Date: Wed, 2 Sep 2020 17:11:09 +0800 Subject: drm/amdkfd: fix a memory leak issue In the resume stage of GPU recovery, start_cpsch will call pm_init which set pm->allocated as false, cause the next pm_release_ib has no chance to release ib memory. Add pm_release_ib in stop_cpsch which will be called in the suspend stage of GPU recovery. Reviewed-by: Felix Kuehling Signed-off-by: Dennis Li Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'drivers') diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c index 069ba4be1e8f..20ef048d6a03 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c @@ -1192,6 +1192,8 @@ static int stop_cpsch(struct device_queue_manager *dqm) dqm->sched_running = false; dqm_unlock(dqm); + pm_release_ib(&dqm->packets); + kfd_gtt_sa_free(dqm->dev, dqm->fence_mem); pm_uninit(&dqm->packets, hanging); -- cgit v1.2.3 From 2a9787dcf537d0e4f1fa41cbd883abe9d70b9fcb Mon Sep 17 00:00:00 2001 From: Liu ChengZhe Date: Wed, 9 Sep 2020 16:00:12 +0800 Subject: drm/amdgpu: Do gpu recovery when no job is running MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit In function flr_work, we should do gpu recovery when no job is running. Fix the logic by inverting it. v2: modify the description Reviewed-by: Christian König Signed-off-by: Liu ChengZhe Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c | 3 ++- drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c | 2 +- 2 files changed, 3 insertions(+), 2 deletions(-) (limited to 'drivers') diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c index 9c07014d9bd6..f5ce9a9f4cf5 100644 --- a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c +++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c @@ -262,7 +262,8 @@ flr_done: /* Trigger recovery for world switch failure if no TDR */ if (amdgpu_device_should_recover_gpu(adev) - && (amdgpu_device_has_job_running(adev) || adev->sdma_timeout == MAX_SCHEDULE_TIMEOUT)) + && (!amdgpu_device_has_job_running(adev) || + adev->sdma_timeout == MAX_SCHEDULE_TIMEOUT)) amdgpu_device_gpu_recover(adev, NULL); } diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c b/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c index 9c23abf9b140..666ed99cc14b 100644 --- a/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c +++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c @@ -283,7 +283,7 @@ flr_done: /* Trigger recovery for world switch failure if no TDR */ if (amdgpu_device_should_recover_gpu(adev) - && (amdgpu_device_has_job_running(adev) || + && (!amdgpu_device_has_job_running(adev) || adev->sdma_timeout == MAX_SCHEDULE_TIMEOUT || adev->gfx_timeout == MAX_SCHEDULE_TIMEOUT || adev->compute_timeout == MAX_SCHEDULE_TIMEOUT || -- cgit v1.2.3 From c9a6b82f45e261d247b980a7949aaa6a9bfffe01 Mon Sep 17 00:00:00 2001 From: Andrey Grodzovsky Date: Wed, 29 Jul 2020 12:59:45 -0400 Subject: drm/amdgpu: Implement DPC recovery Add PCI Downstream Port Containment (DPC) with basic recovery functionality v2: remove pci_save_state to avoid breaking suspend/resume v3: Fix style comments v4: Improve description. Signed-off-by: Andrey Grodzovsky Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 8 ++ drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 161 +++++++++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 9 +- 3 files changed, 177 insertions(+), 1 deletion(-) (limited to 'drivers') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 4009d2e30727..29c0b4c87231 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -49,6 +49,8 @@ #include #include #include +#include +#include #include #include @@ -1260,6 +1262,12 @@ static inline int amdgpu_dm_display_resume(struct amdgpu_device *adev) { return void amdgpu_register_gpu_instance(struct amdgpu_device *adev); void amdgpu_unregister_gpu_instance(struct amdgpu_device *adev); +pci_ers_result_t amdgpu_pci_error_detected(struct pci_dev *pdev, + pci_channel_state_t state); +pci_ers_result_t amdgpu_pci_mmio_enabled(struct pci_dev *pdev); +pci_ers_result_t amdgpu_pci_slot_reset(struct pci_dev *pdev); +void amdgpu_pci_resume(struct pci_dev *pdev); + #include "amdgpu_object.h" /* used by df_v3_6.c and amdgpu_pmu.c */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index f7307af76452..99c0e6e53e84 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -2999,6 +2999,7 @@ static const struct attribute *amdgpu_dev_attributes[] = { NULL }; + /** * amdgpu_device_init - initialize the driver * @@ -3217,6 +3218,8 @@ int amdgpu_device_init(struct amdgpu_device *adev, } } + pci_enable_pcie_error_reporting(adev->ddev.pdev); + /* Post card if necessary */ if (amdgpu_device_need_post(adev)) { if (!adev->bios) { @@ -4705,3 +4708,161 @@ int amdgpu_device_baco_exit(struct drm_device *dev) return 0; } + +/** + * amdgpu_pci_error_detected - Called when a PCI error is detected. + * @pdev: PCI device struct + * @state: PCI channel state + * + * Description: Called when a PCI error is detected. + * + * Return: PCI_ERS_RESULT_NEED_RESET or PCI_ERS_RESULT_DISCONNECT. + */ +pci_ers_result_t amdgpu_pci_error_detected(struct pci_dev *pdev, pci_channel_state_t state) +{ + struct drm_device *dev = pci_get_drvdata(pdev); + struct amdgpu_device *adev = drm_to_adev(dev); + + DRM_INFO("PCI error: detected callback, state(%d)!!\n", state); + + switch (state) { + case pci_channel_io_normal: + return PCI_ERS_RESULT_CAN_RECOVER; + case pci_channel_io_frozen: + /* Fatal error, prepare for slot reset */ + amdgpu_device_lock_adev(adev); + return PCI_ERS_RESULT_NEED_RESET; + case pci_channel_io_perm_failure: + /* Permanent error, prepare for device removal */ + return PCI_ERS_RESULT_DISCONNECT; + } + + return PCI_ERS_RESULT_NEED_RESET; +} + +/** + * amdgpu_pci_mmio_enabled - Enable MMIO and dump debug registers + * @pdev: pointer to PCI device + */ +pci_ers_result_t amdgpu_pci_mmio_enabled(struct pci_dev *pdev) +{ + + DRM_INFO("PCI error: mmio enabled callback!!\n"); + + /* TODO - dump whatever for debugging purposes */ + + /* This called only if amdgpu_pci_error_detected returns + * PCI_ERS_RESULT_CAN_RECOVER. Read/write to the device still + * works, no need to reset slot. + */ + + return PCI_ERS_RESULT_RECOVERED; +} + +/** + * amdgpu_pci_slot_reset - Called when PCI slot has been reset. + * @pdev: PCI device struct + * + * Description: This routine is called by the pci error recovery + * code after the PCI slot has been reset, just before we + * should resume normal operations. + */ +pci_ers_result_t amdgpu_pci_slot_reset(struct pci_dev *pdev) +{ + struct drm_device *dev = pci_get_drvdata(pdev); + struct amdgpu_device *adev = drm_to_adev(dev); + int r; + bool vram_lost; + + DRM_INFO("PCI error: slot reset callback!!\n"); + + pci_restore_state(pdev); + + r = amdgpu_device_ip_suspend(adev); + if (r) + goto out; + + + /* post card */ + r = amdgpu_atom_asic_init(adev->mode_info.atom_context); + if (r) + goto out; + + r = amdgpu_device_ip_resume_phase1(adev); + if (r) + goto out; + + vram_lost = amdgpu_device_check_vram_lost(adev); + if (vram_lost) { + DRM_INFO("VRAM is lost due to GPU reset!\n"); + amdgpu_inc_vram_lost(adev); + } + + r = amdgpu_gtt_mgr_recover( + &adev->mman.bdev.man[TTM_PL_TT]); + if (r) + goto out; + + r = amdgpu_device_fw_loading(adev); + if (r) + return r; + + r = amdgpu_device_ip_resume_phase2(adev); + if (r) + goto out; + + if (vram_lost) + amdgpu_device_fill_reset_magic(adev); + + /* + * Add this ASIC as tracked as reset was already + * complete successfully. + */ + amdgpu_register_gpu_instance(adev); + + r = amdgpu_device_ip_late_init(adev); + if (r) + goto out; + + amdgpu_fbdev_set_suspend(adev, 0); + + /* must succeed. */ + amdgpu_ras_resume(adev); + + + amdgpu_irq_gpu_reset_resume_helper(adev); + r = amdgpu_ib_ring_tests(adev); + if (r) + goto out; + + r = amdgpu_device_recover_vram(adev); + +out: + + if (!r) { + DRM_INFO("PCIe error recovery succeeded\n"); + } else { + DRM_ERROR("PCIe error recovery failed, err:%d", r); + amdgpu_device_unlock_adev(adev); + } + + return r ? PCI_ERS_RESULT_DISCONNECT : PCI_ERS_RESULT_RECOVERED; +} + +/** + * amdgpu_pci_resume() - resume normal ops after PCI reset + * @pdev: pointer to PCI device + * + * Called when the error recovery driver tells us that its + * OK to resume normal operation. Use completion to allow + * halted scsi ops to resume. + */ +void amdgpu_pci_resume(struct pci_dev *pdev) +{ + struct drm_device *dev = pci_get_drvdata(pdev); + struct amdgpu_device *adev = drm_to_adev(dev); + + amdgpu_device_unlock_adev(adev); + + DRM_INFO("PCI error: resume callback!!\n"); +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c index 6edde2b9e402..e9daab175d17 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c @@ -32,7 +32,6 @@ #include #include #include -#include #include #include #include @@ -1528,6 +1527,13 @@ static struct drm_driver kms_driver = { .patchlevel = KMS_DRIVER_PATCHLEVEL, }; +static struct pci_error_handlers amdgpu_pci_err_handler = { + .error_detected = amdgpu_pci_error_detected, + .mmio_enabled = amdgpu_pci_mmio_enabled, + .slot_reset = amdgpu_pci_slot_reset, + .resume = amdgpu_pci_resume, +}; + static struct pci_driver amdgpu_kms_pci_driver = { .name = DRIVER_NAME, .id_table = pciidlist, @@ -1535,6 +1541,7 @@ static struct pci_driver amdgpu_kms_pci_driver = { .remove = amdgpu_pci_remove, .shutdown = amdgpu_pci_shutdown, .driver.pm = &amdgpu_pm_ops, + .err_handler = &amdgpu_pci_err_handler, }; static int __init amdgpu_init(void) -- cgit v1.2.3 From bf36b52e781d7412c3fce826f74ba6a73b9be4d0 Mon Sep 17 00:00:00 2001 From: Andrey Grodzovsky Date: Wed, 29 Jul 2020 13:10:29 -0400 Subject: drm/amdgpu: Avoid accessing HW when suspending SW state At this point the ASIC is already post reset by the HW/PSP so the HW not in proper state to be configured for suspension, some blocks might be even gated and so best is to avoid touching it. v2: Rename in_dpc to more meaningful name Signed-off-by: Andrey Grodzovsky Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 38 ++++++++++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c | 6 +++++ drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c | 6 +++++ drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c | 18 ++++++++------ drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c | 3 +++ 6 files changed, 65 insertions(+), 7 deletions(-) (limited to 'drivers') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 29c0b4c87231..207eba012029 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -989,6 +989,7 @@ struct amdgpu_device { atomic_t throttling_logging_enabled; struct ratelimit_state throttling_logging_rs; uint32_t ras_features; + bool in_pci_err_recovery; }; static inline struct amdgpu_device *drm_to_adev(struct drm_device *ddev) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 99c0e6e53e84..4d9a8734ea20 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -319,6 +319,9 @@ uint32_t amdgpu_mm_rreg(struct amdgpu_device *adev, uint32_t reg, { uint32_t ret; + if (adev->in_pci_err_recovery) + return 0; + if (!(acc_flags & AMDGPU_REGS_NO_KIQ) && amdgpu_sriov_runtime(adev) && down_read_trylock(&adev->reset_sem)) { ret = amdgpu_kiq_rreg(adev, reg); @@ -356,6 +359,9 @@ uint32_t amdgpu_mm_rreg(struct amdgpu_device *adev, uint32_t reg, * Returns the 8 bit value from the offset specified. */ uint8_t amdgpu_mm_rreg8(struct amdgpu_device *adev, uint32_t offset) { + if (adev->in_pci_err_recovery) + return 0; + if (offset < adev->rmmio_size) return (readb(adev->rmmio + offset)); BUG(); @@ -377,6 +383,9 @@ uint8_t amdgpu_mm_rreg8(struct amdgpu_device *adev, uint32_t offset) { * Writes the value specified to the offset specified. */ void amdgpu_mm_wreg8(struct amdgpu_device *adev, uint32_t offset, uint8_t value) { + if (adev->in_pci_err_recovery) + return; + if (offset < adev->rmmio_size) writeb(value, adev->rmmio + offset); else @@ -387,6 +396,9 @@ static inline void amdgpu_mm_wreg_mmio(struct amdgpu_device *adev, uint32_t reg, uint32_t v, uint32_t acc_flags) { + if (adev->in_pci_err_recovery) + return; + trace_amdgpu_mm_wreg(adev->pdev->device, reg, v); if ((reg * 4) < adev->rmmio_size) @@ -414,6 +426,9 @@ static inline void amdgpu_mm_wreg_mmio(struct amdgpu_device *adev, void amdgpu_mm_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v, uint32_t acc_flags) { + if (adev->in_pci_err_recovery) + return; + if (!(acc_flags & AMDGPU_REGS_NO_KIQ) && amdgpu_sriov_runtime(adev) && down_read_trylock(&adev->reset_sem)) { amdgpu_kiq_wreg(adev, reg, v); @@ -432,6 +447,9 @@ void amdgpu_mm_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v, void amdgpu_mm_wreg_mmio_rlc(struct amdgpu_device *adev, uint32_t reg, uint32_t v, uint32_t acc_flags) { + if (adev->in_pci_err_recovery) + return; + if (amdgpu_sriov_fullaccess(adev) && adev->gfx.rlc.funcs && adev->gfx.rlc.funcs->is_rlcg_access_range) { @@ -453,6 +471,9 @@ void amdgpu_mm_wreg_mmio_rlc(struct amdgpu_device *adev, uint32_t reg, uint32_t */ u32 amdgpu_io_rreg(struct amdgpu_device *adev, u32 reg) { + if (adev->in_pci_err_recovery) + return 0; + if ((reg * 4) < adev->rio_mem_size) return ioread32(adev->rio_mem + (reg * 4)); else { @@ -472,6 +493,9 @@ u32 amdgpu_io_rreg(struct amdgpu_device *adev, u32 reg) */ void amdgpu_io_wreg(struct amdgpu_device *adev, u32 reg, u32 v) { + if (adev->in_pci_err_recovery) + return; + if ((reg * 4) < adev->rio_mem_size) iowrite32(v, adev->rio_mem + (reg * 4)); else { @@ -491,6 +515,9 @@ void amdgpu_io_wreg(struct amdgpu_device *adev, u32 reg, u32 v) */ u32 amdgpu_mm_rdoorbell(struct amdgpu_device *adev, u32 index) { + if (adev->in_pci_err_recovery) + return 0; + if (index < adev->doorbell.num_doorbells) { return readl(adev->doorbell.ptr + index); } else { @@ -511,6 +538,9 @@ u32 amdgpu_mm_rdoorbell(struct amdgpu_device *adev, u32 index) */ void amdgpu_mm_wdoorbell(struct amdgpu_device *adev, u32 index, u32 v) { + if (adev->in_pci_err_recovery) + return; + if (index < adev->doorbell.num_doorbells) { writel(v, adev->doorbell.ptr + index); } else { @@ -529,6 +559,9 @@ void amdgpu_mm_wdoorbell(struct amdgpu_device *adev, u32 index, u32 v) */ u64 amdgpu_mm_rdoorbell64(struct amdgpu_device *adev, u32 index) { + if (adev->in_pci_err_recovery) + return 0; + if (index < adev->doorbell.num_doorbells) { return atomic64_read((atomic64_t *)(adev->doorbell.ptr + index)); } else { @@ -549,6 +582,9 @@ u64 amdgpu_mm_rdoorbell64(struct amdgpu_device *adev, u32 index) */ void amdgpu_mm_wdoorbell64(struct amdgpu_device *adev, u32 index, u64 v) { + if (adev->in_pci_err_recovery) + return; + if (index < adev->doorbell.num_doorbells) { atomic64_set((atomic64_t *)(adev->doorbell.ptr + index), v); } else { @@ -4778,7 +4814,9 @@ pci_ers_result_t amdgpu_pci_slot_reset(struct pci_dev *pdev) pci_restore_state(pdev); + adev->in_pci_err_recovery = true; r = amdgpu_device_ip_suspend(adev); + adev->in_pci_err_recovery = false; if (r) goto out; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c index d6981425ec51..8c9bacfdbc30 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c @@ -693,6 +693,9 @@ uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg) struct amdgpu_kiq *kiq = &adev->gfx.kiq; struct amdgpu_ring *ring = &kiq->ring; + if (adev->in_pci_err_recovery) + return 0; + BUG_ON(!ring->funcs->emit_rreg); spin_lock_irqsave(&kiq->ring_lock, flags); @@ -757,6 +760,9 @@ void amdgpu_kiq_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v) BUG_ON(!ring->funcs->emit_wreg); + if (adev->in_pci_err_recovery) + return; + spin_lock_irqsave(&kiq->ring_lock, flags); amdgpu_ring_alloc(ring, 32); amdgpu_ring_emit_wreg(ring, reg, v); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c index d6c38e24f130..a7771aaa1c51 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c @@ -219,6 +219,9 @@ int psp_wait_for(struct psp_context *psp, uint32_t reg_index, int i; struct amdgpu_device *adev = psp->adev; + if (psp->adev->in_pci_err_recovery) + return 0; + for (i = 0; i < adev->usec_timeout; i++) { val = RREG32(reg_index); if (check_changed) { @@ -245,6 +248,9 @@ psp_cmd_submit_buf(struct psp_context *psp, bool ras_intr = false; bool skip_unsupport = false; + if (psp->adev->in_pci_err_recovery) + return 0; + mutex_lock(&psp->mutex); memset(psp->cmd_buf_mem, 0, PSP_CMD_BUFFER_SIZE); diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c index d502e30f67d9..9801bb6daa01 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c @@ -6980,15 +6980,19 @@ static int gfx_v10_0_hw_fini(void *handle) amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0); amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0); + + if (!adev->in_pci_err_recovery) { #ifndef BRING_UP_DEBUG - if (amdgpu_async_gfx_ring) { - r = gfx_v10_0_kiq_disable_kgq(adev); - if (r) - DRM_ERROR("KGQ disable failed\n"); - } + if (amdgpu_async_gfx_ring) { + r = gfx_v10_0_kiq_disable_kgq(adev); + if (r) + DRM_ERROR("KGQ disable failed\n"); + } #endif - if (amdgpu_gfx_disable_kcq(adev)) - DRM_ERROR("KCQ disable failed\n"); + if (amdgpu_gfx_disable_kcq(adev)) + DRM_ERROR("KCQ disable failed\n"); + } + if (amdgpu_sriov_vf(adev)) { gfx_v10_0_cp_gfx_enable(adev, false); /* Program KIQ position of RLC_CP_SCHEDULERS during destroy */ diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c b/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c index a58ea08cd115..97aa72a94631 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c @@ -112,6 +112,9 @@ int smu_cmn_send_smc_msg_with_param(struct smu_context *smu, struct amdgpu_device *adev = smu->adev; int ret = 0, index = 0; + if (smu->adev->in_pci_err_recovery) + return 0; + index = smu_cmn_to_asic_specific_index(smu, CMN2ASIC_MAPPING_MSG, msg); -- cgit v1.2.3 From acd89fca672dd2a7c3a83038cce7601498f21105 Mon Sep 17 00:00:00 2001 From: Andrey Grodzovsky Date: Wed, 29 Jul 2020 14:14:36 -0400 Subject: drm/amdgpu: Block all job scheduling activity during DPC recovery DPC recovery involves ASIC reset just as normal GPU recovery so block SW GPU schedulers and wait on all concurrent GPU resets. Signed-off-by: Andrey Grodzovsky Acked-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 57 +++++++++++++++++++++++++++--- 1 file changed, 53 insertions(+), 4 deletions(-) (limited to 'drivers') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 4d9a8734ea20..2d155355a05f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -4745,6 +4745,20 @@ int amdgpu_device_baco_exit(struct drm_device *dev) return 0; } +static void amdgpu_cancel_all_tdr(struct amdgpu_device *adev) +{ + int i; + + for (i = 0; i < AMDGPU_MAX_RINGS; ++i) { + struct amdgpu_ring *ring = adev->rings[i]; + + if (!ring || !ring->sched.thread) + continue; + + cancel_delayed_work_sync(&ring->sched.work_tdr); + } +} + /** * amdgpu_pci_error_detected - Called when a PCI error is detected. * @pdev: PCI device struct @@ -4758,15 +4772,37 @@ pci_ers_result_t amdgpu_pci_error_detected(struct pci_dev *pdev, pci_channel_sta { struct drm_device *dev = pci_get_drvdata(pdev); struct amdgpu_device *adev = drm_to_adev(dev); + int i; DRM_INFO("PCI error: detected callback, state(%d)!!\n", state); switch (state) { case pci_channel_io_normal: return PCI_ERS_RESULT_CAN_RECOVER; - case pci_channel_io_frozen: - /* Fatal error, prepare for slot reset */ - amdgpu_device_lock_adev(adev); + /* Fatal error, prepare for slot reset */ + case pci_channel_io_frozen: + /* + * Cancel and wait for all TDRs in progress if failing to + * set adev->in_gpu_reset in amdgpu_device_lock_adev + * + * Locking adev->reset_sem will prevent any external access + * to GPU during PCI error recovery + */ + while (!amdgpu_device_lock_adev(adev, NULL)) + amdgpu_cancel_all_tdr(adev); + + /* + * Block any work scheduling as we do for regular GPU reset + * for the duration of the recovery + */ + for (i = 0; i < AMDGPU_MAX_RINGS; ++i) { + struct amdgpu_ring *ring = adev->rings[i]; + + if (!ring || !ring->sched.thread) + continue; + + drm_sched_stop(&ring->sched, NULL); + } return PCI_ERS_RESULT_NEED_RESET; case pci_channel_io_perm_failure: /* Permanent error, prepare for device removal */ @@ -4899,8 +4935,21 @@ void amdgpu_pci_resume(struct pci_dev *pdev) { struct drm_device *dev = pci_get_drvdata(pdev); struct amdgpu_device *adev = drm_to_adev(dev); + int i; - amdgpu_device_unlock_adev(adev); DRM_INFO("PCI error: resume callback!!\n"); + + for (i = 0; i < AMDGPU_MAX_RINGS; ++i) { + struct amdgpu_ring *ring = adev->rings[i]; + + if (!ring || !ring->sched.thread) + continue; + + + drm_sched_resubmit_jobs(&ring->sched); + drm_sched_start(&ring->sched, true); + } + + amdgpu_device_unlock_adev(adev); } -- cgit v1.2.3 From 362c7b91c1b00e31ebd219af1e493e7ef50ccfb4 Mon Sep 17 00:00:00 2001 From: Andrey Grodzovsky Date: Mon, 24 Aug 2020 11:02:18 -0400 Subject: drm/amdgpu: Fix SMU error failure Wait for HW/PSP initiated ASIC reset to complete before starting the recovery operations. v2: Remove typo Signed-off-by: Andrey Grodzovsky Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 22 ++++++++++++++++++++-- 1 file changed, 20 insertions(+), 2 deletions(-) (limited to 'drivers') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 2d155355a05f..36a8f128fb6b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -4843,14 +4843,32 @@ pci_ers_result_t amdgpu_pci_slot_reset(struct pci_dev *pdev) { struct drm_device *dev = pci_get_drvdata(pdev); struct amdgpu_device *adev = drm_to_adev(dev); - int r; + int r, i; bool vram_lost; + u32 memsize; DRM_INFO("PCI error: slot reset callback!!\n"); + /* wait for asic to come out of reset */ + msleep(500); + pci_restore_state(pdev); - adev->in_pci_err_recovery = true; + /* confirm ASIC came out of reset */ + for (i = 0; i < adev->usec_timeout; i++) { + memsize = amdgpu_asic_get_config_memsize(adev); + + if (memsize != 0xffffffff) + break; + udelay(1); + } + if (memsize == 0xffffffff) { + r = -ETIME; + goto out; + } + + /* TODO Call amdgpu_pre_asic_reset instead */ + adev->in_pci_err_recovery = true; r = amdgpu_device_ip_suspend(adev); adev->in_pci_err_recovery = false; if (r) -- cgit v1.2.3 From c1dd4aa624076cb6d4724fad2d9e9e71e46bbc9f Mon Sep 17 00:00:00 2001 From: Andrey Grodzovsky Date: Mon, 24 Aug 2020 12:30:47 -0400 Subject: drm/amdgpu: Fix consecutive DPC recovery failures. Cache the PCI state on boot and before each case where we might loose it. v2: Add pci_restore_state while caching the PCI state to avoid breaking PCI core logic for stuff like suspend/resume. v3: Extract pci_restore_state from amdgpu_device_cache_pci_state to avoid superflous restores during GPU resets and suspend/resumes. v4: Style fixes. Signed-off-by: Andrey Grodzovsky Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 5 +++ drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 62 ++++++++++++++++++++++++++++-- drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 4 +- drivers/gpu/drm/amd/amdgpu/nv.c | 4 +- drivers/gpu/drm/amd/amdgpu/soc15.c | 4 +- 5 files changed, 70 insertions(+), 9 deletions(-) (limited to 'drivers') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 207eba012029..6125ba905faf 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -989,7 +989,9 @@ struct amdgpu_device { atomic_t throttling_logging_enabled; struct ratelimit_state throttling_logging_rs; uint32_t ras_features; + bool in_pci_err_recovery; + struct pci_saved_state *pci_state; }; static inline struct amdgpu_device *drm_to_adev(struct drm_device *ddev) @@ -1269,6 +1271,9 @@ pci_ers_result_t amdgpu_pci_mmio_enabled(struct pci_dev *pdev); pci_ers_result_t amdgpu_pci_slot_reset(struct pci_dev *pdev); void amdgpu_pci_resume(struct pci_dev *pdev); +bool amdgpu_device_cache_pci_state(struct pci_dev *pdev); +bool amdgpu_device_load_pci_state(struct pci_dev *pdev); + #include "amdgpu_object.h" /* used by df_v3_6.c and amdgpu_pmu.c */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 36a8f128fb6b..9be9355bb0e0 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -1292,7 +1292,7 @@ static void amdgpu_switcheroo_set_state(struct pci_dev *pdev, dev->switch_power_state = DRM_SWITCH_POWER_CHANGING; pci_set_power_state(dev->pdev, PCI_D0); - pci_restore_state(dev->pdev); + amdgpu_device_load_pci_state(dev->pdev); r = pci_enable_device(dev->pdev); if (r) DRM_WARN("pci_enable_device failed (%d)\n", r); @@ -1305,7 +1305,7 @@ static void amdgpu_switcheroo_set_state(struct pci_dev *pdev, drm_kms_helper_poll_disable(dev); dev->switch_power_state = DRM_SWITCH_POWER_CHANGING; amdgpu_device_suspend(dev, true); - pci_save_state(dev->pdev); + amdgpu_device_cache_pci_state(dev->pdev); /* Shut down the device */ pci_disable_device(dev->pdev); pci_set_power_state(dev->pdev, PCI_D3cold); @@ -3408,6 +3408,10 @@ fence_driver_init: if (r) dev_err(adev->dev, "amdgpu_pmu_init failed\n"); + /* Have stored pci confspace at hand for restore in sudden PCI error */ + if (amdgpu_device_cache_pci_state(adev->pdev)) + pci_restore_state(pdev); + return 0; failed: @@ -3432,6 +3436,8 @@ void amdgpu_device_fini(struct amdgpu_device *adev) flush_delayed_work(&adev->delayed_init_work); adev->shutdown = true; + kfree(adev->pci_state); + /* make sure IB test finished before entering exclusive mode * to avoid preemption on IB test * */ @@ -4852,7 +4858,7 @@ pci_ers_result_t amdgpu_pci_slot_reset(struct pci_dev *pdev) /* wait for asic to come out of reset */ msleep(500); - pci_restore_state(pdev); + amdgpu_device_load_pci_state(pdev); /* confirm ASIC came out of reset */ for (i = 0; i < adev->usec_timeout; i++) { @@ -4932,6 +4938,9 @@ pci_ers_result_t amdgpu_pci_slot_reset(struct pci_dev *pdev) out: if (!r) { + if (amdgpu_device_cache_pci_state(adev->pdev)) + pci_restore_state(adev->pdev); + DRM_INFO("PCIe error recovery succeeded\n"); } else { DRM_ERROR("PCIe error recovery failed, err:%d", r); @@ -4971,3 +4980,50 @@ void amdgpu_pci_resume(struct pci_dev *pdev) amdgpu_device_unlock_adev(adev); } + +bool amdgpu_device_cache_pci_state(struct pci_dev *pdev) +{ + struct drm_device *dev = pci_get_drvdata(pdev); + struct amdgpu_device *adev = drm_to_adev(dev); + int r; + + r = pci_save_state(pdev); + if (!r) { + kfree(adev->pci_state); + + adev->pci_state = pci_store_saved_state(pdev); + + if (!adev->pci_state) { + DRM_ERROR("Failed to store PCI saved state"); + return false; + } + } else { + DRM_WARN("Failed to save PCI state, err:%d\n", r); + return false; + } + + return true; +} + +bool amdgpu_device_load_pci_state(struct pci_dev *pdev) +{ + struct drm_device *dev = pci_get_drvdata(pdev); + struct amdgpu_device *adev = drm_to_adev(dev); + int r; + + if (!adev->pci_state) + return false; + + r = pci_load_saved_state(pdev, adev->pci_state); + + if (!r) { + pci_restore_state(pdev); + } else { + DRM_WARN("Failed to load PCI state, err:%d\n", r); + return false; + } + + return true; +} + + diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c index e9daab175d17..5fafb3dff1a0 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c @@ -1315,7 +1315,7 @@ static int amdgpu_pmops_runtime_suspend(struct device *dev) if (amdgpu_is_atpx_hybrid()) { pci_ignore_hotplug(pdev); } else { - pci_save_state(pdev); + amdgpu_device_cache_pci_state(pdev); pci_disable_device(pdev); pci_ignore_hotplug(pdev); pci_set_power_state(pdev, PCI_D3cold); @@ -1348,7 +1348,7 @@ static int amdgpu_pmops_runtime_resume(struct device *dev) pci_set_master(pdev); } else { pci_set_power_state(pdev, PCI_D0); - pci_restore_state(pdev); + amdgpu_device_load_pci_state(pdev); ret = pci_enable_device(pdev); if (ret) return ret; diff --git a/drivers/gpu/drm/amd/amdgpu/nv.c b/drivers/gpu/drm/amd/amdgpu/nv.c index 4d1402356262..0ec66030bd11 100644 --- a/drivers/gpu/drm/amd/amdgpu/nv.c +++ b/drivers/gpu/drm/amd/amdgpu/nv.c @@ -311,7 +311,7 @@ static int nv_asic_mode1_reset(struct amdgpu_device *adev) /* disable BM */ pci_clear_master(adev->pdev); - pci_save_state(adev->pdev); + amdgpu_device_cache_pci_state(adev->pdev); if (amdgpu_dpm_is_mode1_reset_supported(adev)) { dev_info(adev->dev, "GPU smu mode1 reset\n"); @@ -323,7 +323,7 @@ static int nv_asic_mode1_reset(struct amdgpu_device *adev) if (ret) dev_err(adev->dev, "GPU mode1 reset failed\n"); - pci_restore_state(adev->pdev); + amdgpu_device_load_pci_state(adev->pdev); /* wait for asic to come out of reset */ for (i = 0; i < adev->usec_timeout; i++) { diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.c b/drivers/gpu/drm/amd/amdgpu/soc15.c index 2f93c475d6d8..ddd55e3176c4 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc15.c +++ b/drivers/gpu/drm/amd/amdgpu/soc15.c @@ -484,13 +484,13 @@ static int soc15_asic_mode1_reset(struct amdgpu_device *adev) /* disable BM */ pci_clear_master(adev->pdev); - pci_save_state(adev->pdev); + amdgpu_device_cache_pci_state(adev->pdev); ret = psp_gpu_reset(adev); if (ret) dev_err(adev->dev, "GPU mode1 reset failed\n"); - pci_restore_state(adev->pdev); + amdgpu_device_load_pci_state(adev->pdev); /* wait for asic to come out of reset */ for (i = 0; i < adev->usec_timeout; i++) { -- cgit v1.2.3 From 7ac71382e90aba1f58e304e386acdd02af963027 Mon Sep 17 00:00:00 2001 From: Andrey Grodzovsky Date: Mon, 24 Aug 2020 14:27:40 -0400 Subject: drm/amdgpu: Trim amdgpu_pci_slot_reset by reusing code. Reuse exsisting functions from GPU recovery to avoid code duplications. Signed-off-by: Andrey Grodzovsky Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 73 +++++------------------------- 1 file changed, 12 insertions(+), 61 deletions(-) (limited to 'drivers') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 9be9355bb0e0..ab42d0088f57 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -4117,7 +4117,8 @@ static int amdgpu_device_pre_asic_reset(struct amdgpu_device *adev, static int amdgpu_do_asic_reset(struct amdgpu_hive_info *hive, struct list_head *device_list_handle, - bool *need_full_reset_arg) + bool *need_full_reset_arg, + bool skip_hw_reset) { struct amdgpu_device *tmp_adev = NULL; bool need_full_reset = *need_full_reset_arg, vram_lost = false; @@ -4127,7 +4128,7 @@ static int amdgpu_do_asic_reset(struct amdgpu_hive_info *hive, * ASIC reset has to be done on all HGMI hive nodes ASAP * to allow proper links negotiation in FW (within 1 sec) */ - if (need_full_reset) { + if (!skip_hw_reset && need_full_reset) { list_for_each_entry(tmp_adev, device_list_handle, gmc.xgmi.head) { /* For XGMI run all resets in parallel to speed up the process */ if (tmp_adev->gmc.xgmi.num_physical_nodes > 1) { @@ -4522,7 +4523,7 @@ retry: /* Rest of adevs pre asic reset from XGMI hive. */ if (r) adev->asic_reset_res = r; } else { - r = amdgpu_do_asic_reset(hive, device_list_handle, &need_full_reset); + r = amdgpu_do_asic_reset(hive, device_list_handle, &need_full_reset, false); if (r && r == -EAGAIN) goto retry; } @@ -4850,14 +4851,19 @@ pci_ers_result_t amdgpu_pci_slot_reset(struct pci_dev *pdev) struct drm_device *dev = pci_get_drvdata(pdev); struct amdgpu_device *adev = drm_to_adev(dev); int r, i; - bool vram_lost; + bool need_full_reset = true; u32 memsize; + struct list_head device_list; DRM_INFO("PCI error: slot reset callback!!\n"); + INIT_LIST_HEAD(&device_list); + list_add_tail(&adev->gmc.xgmi.head, &device_list); + /* wait for asic to come out of reset */ msleep(500); + /* Restore PCI confspace */ amdgpu_device_load_pci_state(pdev); /* confirm ASIC came out of reset */ @@ -4873,70 +4879,15 @@ pci_ers_result_t amdgpu_pci_slot_reset(struct pci_dev *pdev) goto out; } - /* TODO Call amdgpu_pre_asic_reset instead */ adev->in_pci_err_recovery = true; - r = amdgpu_device_ip_suspend(adev); + r = amdgpu_device_pre_asic_reset(adev, NULL, &need_full_reset); adev->in_pci_err_recovery = false; if (r) goto out; - - /* post card */ - r = amdgpu_atom_asic_init(adev->mode_info.atom_context); - if (r) - goto out; - - r = amdgpu_device_ip_resume_phase1(adev); - if (r) - goto out; - - vram_lost = amdgpu_device_check_vram_lost(adev); - if (vram_lost) { - DRM_INFO("VRAM is lost due to GPU reset!\n"); - amdgpu_inc_vram_lost(adev); - } - - r = amdgpu_gtt_mgr_recover( - &adev->mman.bdev.man[TTM_PL_TT]); - if (r) - goto out; - - r = amdgpu_device_fw_loading(adev); - if (r) - return r; - - r = amdgpu_device_ip_resume_phase2(adev); - if (r) - goto out; - - if (vram_lost) - amdgpu_device_fill_reset_magic(adev); - - /* - * Add this ASIC as tracked as reset was already - * complete successfully. - */ - amdgpu_register_gpu_instance(adev); - - r = amdgpu_device_ip_late_init(adev); - if (r) - goto out; - - amdgpu_fbdev_set_suspend(adev, 0); - - /* must succeed. */ - amdgpu_ras_resume(adev); - - - amdgpu_irq_gpu_reset_resume_helper(adev); - r = amdgpu_ib_ring_tests(adev); - if (r) - goto out; - - r = amdgpu_device_recover_vram(adev); + r = amdgpu_do_asic_reset(NULL, &device_list, &need_full_reset, true); out: - if (!r) { if (amdgpu_device_cache_pci_state(adev->pdev)) pci_restore_state(adev->pdev); -- cgit v1.2.3 From 6894305c97ceb56c0caa6f701516546a732af95f Mon Sep 17 00:00:00 2001 From: Andrey Grodzovsky Date: Mon, 24 Aug 2020 14:41:56 -0400 Subject: drm/amdgpu: Disable DPC for XGMI for now. XGMI support is more complicated than single device support as questions of synchronization between the device recovering from PCI error and other members of the hive are required. Leaving this for next round. Signed-off-by: Andrey Grodzovsky Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'drivers') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index ab42d0088f57..36231fe9e4e5 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -4783,6 +4783,11 @@ pci_ers_result_t amdgpu_pci_error_detected(struct pci_dev *pdev, pci_channel_sta DRM_INFO("PCI error: detected callback, state(%d)!!\n", state); + if (adev->gmc.xgmi.num_physical_nodes > 1) { + DRM_WARN("No support for XGMI hive yet..."); + return PCI_ERS_RESULT_DISCONNECT; + } + switch (state) { case pci_channel_io_normal: return PCI_ERS_RESULT_CAN_RECOVER; -- cgit v1.2.3 From 7cbbc745dc11a75e11df532ca1566949709d6695 Mon Sep 17 00:00:00 2001 From: Andrey Grodzovsky Date: Mon, 31 Aug 2020 11:34:33 -0400 Subject: drm/amdgpu: Minor checkpatch fix Signed-off-by: Andrey Grodzovsky Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'drivers') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 36231fe9e4e5..d633e5448476 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -358,7 +358,8 @@ uint32_t amdgpu_mm_rreg(struct amdgpu_device *adev, uint32_t reg, * * Returns the 8 bit value from the offset specified. */ -uint8_t amdgpu_mm_rreg8(struct amdgpu_device *adev, uint32_t offset) { +uint8_t amdgpu_mm_rreg8(struct amdgpu_device *adev, uint32_t offset) +{ if (adev->in_pci_err_recovery) return 0; @@ -382,7 +383,8 @@ uint8_t amdgpu_mm_rreg8(struct amdgpu_device *adev, uint32_t offset) { * * Writes the value specified to the offset specified. */ -void amdgpu_mm_wreg8(struct amdgpu_device *adev, uint32_t offset, uint8_t value) { +void amdgpu_mm_wreg8(struct amdgpu_device *adev, uint32_t offset, uint8_t value) +{ if (adev->in_pci_err_recovery) return; -- cgit v1.2.3 From 0e4b291bb76efbda9f542e514d16f9f3956fce5b Mon Sep 17 00:00:00 2001 From: Likun Gao Date: Thu, 10 Sep 2020 16:03:08 +0800 Subject: drm/amd/pm: update driver if file for sienna cichlid Update drive if file for sienna_cichlid. Signed-off-by: Likun Gao Reviewed-by: Jiansong Chen Signed-off-by: Alex Deucher --- .../drm/amd/pm/inc/smu11_driver_if_sienna_cichlid.h | 20 +++++++++++++++----- drivers/gpu/drm/amd/pm/inc/smu_v11_0.h | 2 +- .../gpu/drm/amd/pm/swsmu/smu11/sienna_cichlid_ppt.c | 5 ----- 3 files changed, 16 insertions(+), 11 deletions(-) (limited to 'drivers') diff --git a/drivers/gpu/drm/amd/pm/inc/smu11_driver_if_sienna_cichlid.h b/drivers/gpu/drm/amd/pm/inc/smu11_driver_if_sienna_cichlid.h index 5ef9c92f57c4..11a6cf96fe0c 100644 --- a/drivers/gpu/drm/amd/pm/inc/smu11_driver_if_sienna_cichlid.h +++ b/drivers/gpu/drm/amd/pm/inc/smu11_driver_if_sienna_cichlid.h @@ -27,9 +27,9 @@ // *** IMPORTANT *** // SMU TEAM: Always increment the interface version if // any structure is changed in this file -#define SMU11_DRIVER_IF_VERSION 0x35 +#define SMU11_DRIVER_IF_VERSION 0x37 -#define PPTABLE_Sienna_Cichlid_SMU_VERSION 5 +#define PPTABLE_Sienna_Cichlid_SMU_VERSION 6 #define NUM_GFXCLK_DPM_LEVELS 16 #define NUM_SMNCLK_DPM_LEVELS 2 @@ -169,7 +169,7 @@ typedef enum { #define DPM_OVERRIDE_DISABLE_DFLL_PLL_SHUTDOWN 0x00000200 #define DPM_OVERRIDE_DISABLE_MEMORY_TEMPERATURE_READ 0x00000400 #define DPM_OVERRIDE_DISABLE_VOLT_LINK_VCN_DCEFCLK 0x00000800 -#define DPM_OVERRIDE_ENABLE_FAST_FCLK_TIMER 0x00001000 +#define DPM_OVERRIDE_DISABLE_FAST_FCLK_TIMER 0x00001000 #define DPM_OVERRIDE_DISABLE_VCN_PG 0x00002000 #define DPM_OVERRIDE_DISABLE_FMAX_VMAX 0x00004000 @@ -793,8 +793,18 @@ typedef struct { // SECTION: Sku Reserved uint8_t CustomerVariant; - uint8_t Spare[3]; - uint32_t SkuReserved[14]; + + //VC BTC parameters are only applicable to VDD_GFX domain + uint8_t VcBtcEnabled; + uint16_t VcBtcVminT0; // T0_VMIN + uint16_t VcBtcFixedVminAgingOffset; // FIXED_VMIN_AGING_OFFSET + uint16_t VcBtcVmin2PsmDegrationGb; // VMIN_TO_PSM_DEGRADATION_GB + uint32_t VcBtcPsmA; // A_PSM + uint32_t VcBtcPsmB; // B_PSM + uint32_t VcBtcVminA; // A_VMIN + uint32_t VcBtcVminB; // B_VMIN + + uint32_t SkuReserved[9]; // MAJOR SECTION: BOARD PARAMETERS diff --git a/drivers/gpu/drm/amd/pm/inc/smu_v11_0.h b/drivers/gpu/drm/amd/pm/inc/smu_v11_0.h index 1f9575a4dfe7..99f545d3ccb8 100644 --- a/drivers/gpu/drm/amd/pm/inc/smu_v11_0.h +++ b/drivers/gpu/drm/amd/pm/inc/smu_v11_0.h @@ -30,7 +30,7 @@ #define SMU11_DRIVER_IF_VERSION_NV10 0x36 #define SMU11_DRIVER_IF_VERSION_NV12 0x36 #define SMU11_DRIVER_IF_VERSION_NV14 0x36 -#define SMU11_DRIVER_IF_VERSION_Sienna_Cichlid 0x35 +#define SMU11_DRIVER_IF_VERSION_Sienna_Cichlid 0x37 #define SMU11_DRIVER_IF_VERSION_Navy_Flounder 0x4 /* MP Apertures */ diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu11/sienna_cichlid_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu11/sienna_cichlid_ppt.c index 5c22611d5878..9cf4e9aef117 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu11/sienna_cichlid_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu11/sienna_cichlid_ppt.c @@ -2291,11 +2291,6 @@ static void sienna_cichlid_dump_pptable(struct smu_context *smu) dev_info(smu->adev->dev, "SkuReserved[6] = 0x%x\n", pptable->SkuReserved[6]); dev_info(smu->adev->dev, "SkuReserved[7] = 0x%x\n", pptable->SkuReserved[7]); dev_info(smu->adev->dev, "SkuReserved[8] = 0x%x\n", pptable->SkuReserved[8]); - dev_info(smu->adev->dev, "SkuReserved[9] = 0x%x\n", pptable->SkuReserved[9]); - dev_info(smu->adev->dev, "SkuReserved[10] = 0x%x\n", pptable->SkuReserved[10]); - dev_info(smu->adev->dev, "SkuReserved[11] = 0x%x\n", pptable->SkuReserved[11]); - dev_info(smu->adev->dev, "SkuReserved[12] = 0x%x\n", pptable->SkuReserved[12]); - dev_info(smu->adev->dev, "SkuReserved[13] = 0x%x\n", pptable->SkuReserved[13]); dev_info(smu->adev->dev, "GamingClk[0] = 0x%x\n", pptable->GamingClk[0]); dev_info(smu->adev->dev, "GamingClk[1] = 0x%x\n", pptable->GamingClk[1]); -- cgit v1.2.3 From 9c1615be19490ebf589696224cfedc0abddc9234 Mon Sep 17 00:00:00 2001 From: Jiansong Chen Date: Thu, 10 Sep 2020 17:08:47 +0800 Subject: drm/amd/pm: update driver if version for navy_flounder It's in accordance with pmfw 65.8.0 for navy_flounder. Signed-off-by: Jiansong Chen Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/pm/inc/smu_v11_0.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers') diff --git a/drivers/gpu/drm/amd/pm/inc/smu_v11_0.h b/drivers/gpu/drm/amd/pm/inc/smu_v11_0.h index 99f545d3ccb8..7ae83df83edb 100644 --- a/drivers/gpu/drm/amd/pm/inc/smu_v11_0.h +++ b/drivers/gpu/drm/amd/pm/inc/smu_v11_0.h @@ -31,7 +31,7 @@ #define SMU11_DRIVER_IF_VERSION_NV12 0x36 #define SMU11_DRIVER_IF_VERSION_NV14 0x36 #define SMU11_DRIVER_IF_VERSION_Sienna_Cichlid 0x37 -#define SMU11_DRIVER_IF_VERSION_Navy_Flounder 0x4 +#define SMU11_DRIVER_IF_VERSION_Navy_Flounder 0x5 /* MP Apertures */ #define MP0_Public 0x03800000 -- cgit v1.2.3 From f399d4de2d6a2e329ef3bceaa0e6a74c9d36c3cd Mon Sep 17 00:00:00 2001 From: Changfeng Date: Tue, 8 Sep 2020 16:12:42 +0800 Subject: drm/amdgpu: add ta DTM/HDCP print in amdgpu_firmware_info for apu It needs to add ta DTM/HDCP print to get HDCP/DTM version info when cat amdgpu_firmware_info Signed-off-by: Changfeng Reviewed-by: Huang Rui Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c | 43 +++++++++++++++++++++++++++------ 1 file changed, 36 insertions(+), 7 deletions(-) (limited to 'drivers') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c index 456a4a93b337..bccaf4f77647 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c @@ -282,14 +282,25 @@ static int amdgpu_firmware_info(struct drm_amdgpu_info_firmware *fw_info, fw_info->feature = 0; break; case AMDGPU_INFO_FW_TA: - if (query_fw->index > 1) - return -EINVAL; - if (query_fw->index == 0) { + switch (query_fw->index) { + case 0: fw_info->ver = adev->psp.ta_fw_version; fw_info->feature = adev->psp.ta_xgmi_ucode_version; - } else { + break; + case 1: fw_info->ver = adev->psp.ta_fw_version; fw_info->feature = adev->psp.ta_ras_ucode_version; + break; + case 2: + fw_info->ver = adev->psp.ta_fw_version; + fw_info->feature = adev->psp.ta_hdcp_ucode_version; + break; + case 3: + fw_info->ver = adev->psp.ta_fw_version; + fw_info->feature = adev->psp.ta_dtm_ucode_version; + break; + default: + return -EINVAL; } break; case AMDGPU_INFO_FW_SDMA: @@ -1385,13 +1396,31 @@ static int amdgpu_debugfs_firmware_info(struct seq_file *m, void *data) fw_info.feature, fw_info.ver); query_fw.fw_type = AMDGPU_INFO_FW_TA; - for (i = 0; i < 2; i++) { + for (i = 0; i < 4; i++) { query_fw.index = i; ret = amdgpu_firmware_info(&fw_info, &query_fw, adev); if (ret) continue; - seq_printf(m, "TA %s feature version: %u, firmware version: 0x%08x\n", - i ? "RAS" : "XGMI", fw_info.feature, fw_info.ver); + switch (query_fw.index) { + case 0: + seq_printf(m, "TA %s feature version: 0x%08x, firmware version: 0x%08x\n", + "RAS", fw_info.feature, fw_info.ver); + break; + case 1: + seq_printf(m, "TA %s feature version: 0x%08x, firmware version: 0x%08x\n", + "XGMI", fw_info.feature, fw_info.ver); + break; + case 2: + seq_printf(m, "TA %s feature version: 0x%08x, firmware version: 0x%08x\n", + "HDCP", fw_info.feature, fw_info.ver); + break; + case 3: + seq_printf(m, "TA %s feature version: 0x%08x, firmware version: 0x%08x\n", + "DTM", fw_info.feature, fw_info.ver); + break; + default: + return -EINVAL; + } } /* SMC */ -- cgit v1.2.3 From 2b3bbf23549f186520005cecc6795327eeb29a35 Mon Sep 17 00:00:00 2001 From: YueHaibing Date: Thu, 10 Sep 2020 15:50:06 +0800 Subject: drm/amdkfd: Fix -Wunused-const-variable warning MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit If KFD_SUPPORT_IOMMU_V2 is not set, gcc warns: drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_device.c:121:37: warning: ‘raven_device_info’ defined but not used [-Wunused-const-variable=] static const struct kfd_device_info raven_device_info = { ^~~~~~~~~~~~~~~~~ As Huang Rui suggested, Raven already has the fallback path, so it should be out of IOMMU v2 flag. Suggested-by: Huang Rui Signed-off-by: YueHaibing Acked-by: Huang Rui Reviewed-by: Felix Kuehling Signed-off-by: Felix Kuehling Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdkfd/kfd_device.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers') diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c index 0e71a0543f98..e3fc6ed7b79c 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c @@ -503,8 +503,8 @@ static const struct kfd_device_info *kfd_supported_devices[][2] = { #ifdef KFD_SUPPORT_IOMMU_V2 [CHIP_KAVERI] = {&kaveri_device_info, NULL}, [CHIP_CARRIZO] = {&carrizo_device_info, NULL}, - [CHIP_RAVEN] = {&raven_device_info, NULL}, #endif + [CHIP_RAVEN] = {&raven_device_info, NULL}, [CHIP_HAWAII] = {&hawaii_device_info, NULL}, [CHIP_TONGA] = {&tonga_device_info, NULL}, [CHIP_FIJI] = {&fiji_device_info, &fiji_vf_device_info}, -- cgit v1.2.3 From 9c7e2ceb1db064da10d8386590ce23afdc7dc6f0 Mon Sep 17 00:00:00 2001 From: John Clements Date: Fri, 11 Sep 2020 14:26:13 +0800 Subject: drm/amdgpu: Update RAS init handling Output RAS init status If RAS init fails, teardown RAS context Reviewed-by: Hawking Zhang Signed-off-by: John Clements Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c | 12 +++++++++++- 1 file changed, 11 insertions(+), 1 deletion(-) (limited to 'drivers') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c index a7771aaa1c51..9782fbc000fe 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c @@ -935,6 +935,7 @@ static int psp_ras_load(struct psp_context *psp) { int ret; struct psp_gfx_cmd_resp *cmd; + struct ta_ras_shared_memory *ras_cmd; /* * TODO: bypass the loading in sriov for now @@ -958,11 +959,20 @@ static int psp_ras_load(struct psp_context *psp) ret = psp_cmd_submit_buf(psp, NULL, cmd, psp->fence_buf_mc_addr); + ras_cmd = (struct ta_ras_shared_memory*)psp->ras.ras_shared_buf; + if (!ret) { - psp->ras.ras_initialized = true; psp->ras.session_id = cmd->resp.session_id; + + if (!ras_cmd->ras_status) + psp->ras.ras_initialized = true; + else + dev_warn(psp->adev->dev, "RAS Init Status: 0x%X\n", ras_cmd->ras_status); } + if (ret || ras_cmd->ras_status) + amdgpu_ras_fini(psp->adev); + kfree(cmd); return ret; -- cgit v1.2.3 From 5d1c59c47997e9955be3a52e261c08e8b40b3902 Mon Sep 17 00:00:00 2001 From: Aurabindo Pillai Date: Wed, 12 Aug 2020 12:40:34 -0400 Subject: drm/amdgpu: Move existing pflip fields into separate struct [Why&How] To refactor DM IRQ management, all fields used by IRQ is best moved to a separate struct so that main amdgpu_crtc struct need not be changed Location of the new struct shall be in DM Reviewed-by: Nicholas Kazlauskas Signed-off-by: Aurabindo Pillai Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h | 4 ++- drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 4 +-- .../amd/display/amdgpu_dm/amdgpu_dm_irq_params.h | 33 ++++++++++++++++++++++ 3 files changed, 38 insertions(+), 3 deletions(-) create mode 100644 drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_irq_params.h (limited to 'drivers') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h index 04a430e0e2e1..a04decb934b0 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h @@ -46,6 +46,7 @@ #include #include "modules/inc/mod_freesync.h" +#include "amdgpu_dm_irq_params.h" struct amdgpu_bo; struct amdgpu_device; @@ -404,7 +405,8 @@ struct amdgpu_crtc { struct amdgpu_flip_work *pflip_works; enum amdgpu_flip_status pflip_status; int deferred_flip_completion; - u32 last_flip_vblank; + /* parameters access from DM IRQ handler */ + struct dm_irq_params dm_irq_params; /* pll sharing */ struct amdgpu_atom_ss ss; bool ss_enabled; diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c index ec29376667df..1f05ec92556a 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c @@ -425,7 +425,7 @@ static void dm_pflip_high_irq(void *interrupt_params) * of pageflip completion, so last_flip_vblank is the forbidden count * for queueing new pageflips if vsync + VRR is enabled. */ - amdgpu_crtc->last_flip_vblank = + amdgpu_crtc->dm_irq_params.last_flip_vblank = amdgpu_get_vblank_counter_kms(&amdgpu_crtc->base); amdgpu_crtc->pflip_status = AMDGPU_FLIP_NONE; @@ -7197,7 +7197,7 @@ static void amdgpu_dm_commit_planes(struct drm_atomic_state *state, * on late submission of flips. */ spin_lock_irqsave(&pcrtc->dev->event_lock, flags); - last_flip_vblank = acrtc_attach->last_flip_vblank; + last_flip_vblank = acrtc_attach->dm_irq_params.last_flip_vblank; spin_unlock_irqrestore(&pcrtc->dev->event_lock, flags); } diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_irq_params.h b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_irq_params.h new file mode 100644 index 000000000000..55ef237eed8b --- /dev/null +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_irq_params.h @@ -0,0 +1,33 @@ +/* + * Copyright 2020 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: AMD + * + */ + +#ifndef __AMDGPU_DM_IRQ_PARAMS_H__ +#define __AMDGPU_DM_IRQ_PARAMS_H__ + +struct dm_irq_params { + u32 last_flip_vblank; +}; + +#endif /* __AMDGPU_DM_IRQ_PARAMS_H__ */ -- cgit v1.2.3 From 585d450c76d1d5807a50cb6c22487ae5aed0906b Mon Sep 17 00:00:00 2001 From: Aurabindo Pillai Date: Wed, 12 Aug 2020 18:56:14 -0400 Subject: drm/amd/display: Refactor to prevent crtc state access in DM IRQ handler [Why&How] Currently commit_tail holds global locks and wait for dependencies which is against the DRM API contracts. Inorder to fix this, IRQ handler should be able to run without having to access crtc state. Required parameters are copied over so that they can be directly accessed from the interrupt handler Reviewed-by: Nicholas Kazlauskas Signed-off-by: Aurabindo Pillai Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 117 ++++++++++++--------- drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h | 1 - .../amd/display/amdgpu_dm/amdgpu_dm_irq_params.h | 4 + 3 files changed, 69 insertions(+), 53 deletions(-) (limited to 'drivers') diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c index 1f05ec92556a..b044fab05a2f 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c @@ -228,17 +228,14 @@ static u32 dm_vblank_get_counter(struct amdgpu_device *adev, int crtc) return 0; else { struct amdgpu_crtc *acrtc = adev->mode_info.crtcs[crtc]; - struct dm_crtc_state *acrtc_state = to_dm_crtc_state( - acrtc->base.state); - - if (acrtc_state->stream == NULL) { + if (acrtc->dm_irq_params.stream == NULL) { DRM_ERROR("dc_stream_state is NULL for crtc '%d'!\n", crtc); return 0; } - return dc_stream_get_vblank_counter(acrtc_state->stream); + return dc_stream_get_vblank_counter(acrtc->dm_irq_params.stream); } } @@ -251,10 +248,8 @@ static int dm_crtc_get_scanoutpos(struct amdgpu_device *adev, int crtc, return -EINVAL; else { struct amdgpu_crtc *acrtc = adev->mode_info.crtcs[crtc]; - struct dm_crtc_state *acrtc_state = to_dm_crtc_state( - acrtc->base.state); - if (acrtc_state->stream == NULL) { + if (acrtc->dm_irq_params.stream == NULL) { DRM_ERROR("dc_stream_state is NULL for crtc '%d'!\n", crtc); return 0; @@ -264,7 +259,7 @@ static int dm_crtc_get_scanoutpos(struct amdgpu_device *adev, int crtc, * TODO rework base driver to use values directly. * for now parse it back into reg-format */ - dc_stream_get_scanoutpos(acrtc_state->stream, + dc_stream_get_scanoutpos(acrtc->dm_irq_params.stream, &v_blank_start, &v_blank_end, &h_position, @@ -323,6 +318,14 @@ get_crtc_by_otg_inst(struct amdgpu_device *adev, return NULL; } +static inline bool amdgpu_dm_vrr_active_irq(struct amdgpu_crtc *acrtc) +{ + return acrtc->dm_irq_params.freesync_config.state == + VRR_STATE_ACTIVE_VARIABLE || + acrtc->dm_irq_params.freesync_config.state == + VRR_STATE_ACTIVE_FIXED; +} + static inline bool amdgpu_dm_vrr_active(struct dm_crtc_state *dm_state) { return dm_state->freesync_config.state == VRR_STATE_ACTIVE_VARIABLE || @@ -343,7 +346,6 @@ static void dm_pflip_high_irq(void *interrupt_params) struct amdgpu_device *adev = irq_params->adev; unsigned long flags; struct drm_pending_vblank_event *e; - struct dm_crtc_state *acrtc_state; uint32_t vpos, hpos, v_blank_start, v_blank_end; bool vrr_active; @@ -375,12 +377,11 @@ static void dm_pflip_high_irq(void *interrupt_params) if (!e) WARN_ON(1); - acrtc_state = to_dm_crtc_state(amdgpu_crtc->base.state); - vrr_active = amdgpu_dm_vrr_active(acrtc_state); + vrr_active = amdgpu_dm_vrr_active_irq(amdgpu_crtc); /* Fixed refresh rate, or VRR scanout position outside front-porch? */ if (!vrr_active || - !dc_stream_get_scanoutpos(acrtc_state->stream, &v_blank_start, + !dc_stream_get_scanoutpos(amdgpu_crtc->dm_irq_params.stream, &v_blank_start, &v_blank_end, &hpos, &vpos) || (vpos < v_blank_start)) { /* Update to correct count and vblank timestamp if racing with @@ -441,17 +442,17 @@ static void dm_vupdate_high_irq(void *interrupt_params) struct common_irq_params *irq_params = interrupt_params; struct amdgpu_device *adev = irq_params->adev; struct amdgpu_crtc *acrtc; - struct dm_crtc_state *acrtc_state; unsigned long flags; + int vrr_active; acrtc = get_crtc_by_otg_inst(adev, irq_params->irq_src - IRQ_TYPE_VUPDATE); if (acrtc) { - acrtc_state = to_dm_crtc_state(acrtc->base.state); + vrr_active = amdgpu_dm_vrr_active_irq(acrtc); DRM_DEBUG_VBL("crtc:%d, vupdate-vrr:%d\n", acrtc->crtc_id, - amdgpu_dm_vrr_active(acrtc_state)); + vrr_active); /* Core vblank handling is done here after end of front-porch in * vrr mode, as vblank timestamping will give valid results @@ -459,22 +460,22 @@ static void dm_vupdate_high_irq(void *interrupt_params) * page-flip completion events that have been queued to us * if a pageflip happened inside front-porch. */ - if (amdgpu_dm_vrr_active(acrtc_state)) { + if (vrr_active) { drm_crtc_handle_vblank(&acrtc->base); /* BTR processing for pre-DCE12 ASICs */ - if (acrtc_state->stream && + if (acrtc->dm_irq_params.stream && adev->family < AMDGPU_FAMILY_AI) { spin_lock_irqsave(&adev_to_drm(adev)->event_lock, flags); mod_freesync_handle_v_update( adev->dm.freesync_module, - acrtc_state->stream, - &acrtc_state->vrr_params); + acrtc->dm_irq_params.stream, + &acrtc->dm_irq