From 796a9f55a8d1d85387b973df9a06cbf4bc2d6327 Mon Sep 17 00:00:00 2001 From: Philipp Stanner Date: Tue, 11 Feb 2025 12:14:23 +0100 Subject: drm/sched: Use struct for drm_sched_init() params MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit drm_sched_init() has a great many parameters and upcoming new functionality for the scheduler might add even more. Generally, the great number of parameters reduces readability and has already caused one missnaming, addressed in: commit 6f1cacf4eba7 ("drm/nouveau: Improve variable name in nouveau_sched_init()"). Introduce a new struct for the scheduler init parameters and port all users. Reviewed-by: Liviu Dudau Acked-by: Matthew Brost # for Xe Reviewed-by: Boris Brezillon # for Panfrost and Panthor Reviewed-by: Christian Gmeiner # for Etnaviv Reviewed-by: Frank Binns # for Imagination Reviewed-by: Tvrtko Ursulin # for Sched Reviewed-by: Maíra Canal # for v3d Reviewed-by: Danilo Krummrich Reviewed-by: Lizhi Hou # for amdxdna Signed-off-by: Philipp Stanner Link: https://patchwork.freedesktop.org/patch/msgid/20250211111422.21235-2-phasta@kernel.org --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 18 ++++++++++++------ 1 file changed, 12 insertions(+), 6 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index d100bb7a137c..6800f72fafd2 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -2823,6 +2823,12 @@ static int amdgpu_device_fw_loading(struct amdgpu_device *adev) static int amdgpu_device_init_schedulers(struct amdgpu_device *adev) { + struct drm_sched_init_args args = { + .ops = &amdgpu_sched_ops, + .num_rqs = DRM_SCHED_PRIORITY_COUNT, + .timeout_wq = adev->reset_domain->wq, + .dev = adev->dev, + }; long timeout; int r, i; @@ -2848,12 +2854,12 @@ static int amdgpu_device_init_schedulers(struct amdgpu_device *adev) break; } - r = drm_sched_init(&ring->sched, &amdgpu_sched_ops, NULL, - DRM_SCHED_PRIORITY_COUNT, - ring->num_hw_submission, 0, - timeout, adev->reset_domain->wq, - ring->sched_score, ring->name, - adev->dev); + args.timeout = timeout; + args.credit_limit = ring->num_hw_submission; + args.score = ring->sched_score; + args.name = ring->name; + + r = drm_sched_init(&ring->sched, &args); if (r) { DRM_ERROR("Failed to create scheduler on ring %s.\n", ring->name); -- cgit v1.2.3 From 1c0b144bf7628a62575a4ec26623f7a6cdf0cb2b Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Thu, 19 Dec 2024 12:50:43 -0500 Subject: drm/amdgpu: rework i2c init and fini No functional change. Rework the code to allow for adding some additional i2c buses in conjunction with DC in the future. Reviewed-by: Harry Wentland Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index d100bb7a137c..4a9580fb0628 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -4465,8 +4465,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, goto failed; } /* init i2c buses */ - if (!amdgpu_device_has_dc_support(adev)) - amdgpu_atombios_i2c_init(adev); + amdgpu_i2c_init(adev); } } @@ -4735,8 +4734,7 @@ void amdgpu_device_fini_sw(struct amdgpu_device *adev) amdgpu_reset_fini(adev); /* free i2c buses */ - if (!amdgpu_device_has_dc_support(adev)) - amdgpu_i2c_fini(adev); + amdgpu_i2c_fini(adev); if (amdgpu_emu_mode != 1) amdgpu_atombios_fini(adev); -- cgit v1.2.3 From e986e89659e18ba986db044df58f165042817dc3 Mon Sep 17 00:00:00 2001 From: Lijo Lazar Date: Wed, 5 Feb 2025 12:24:17 +0530 Subject: drm/amdgpu: Add wrapper for freeing vbios memory Use bios_release wrapper to release memory allocated for vbios image and reset the variables. v2: Use the same wrapper for clean up in sw_fini (Alex Deucher) Signed-off-by: Lijo Lazar Reviewed-by: Alex Deucher Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 4a9580fb0628..947289eb1b2f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -4739,8 +4739,7 @@ void amdgpu_device_fini_sw(struct amdgpu_device *adev) if (amdgpu_emu_mode != 1) amdgpu_atombios_fini(adev); - kfree(adev->bios); - adev->bios = NULL; + amdgpu_bios_release(adev); kfree(adev->fru_info); adev->fru_info = NULL; -- cgit v1.2.3 From 7e0aa706810818d8e3fe865f76bdbd41296e6491 Mon Sep 17 00:00:00 2001 From: Lijo Lazar Date: Wed, 5 Feb 2025 12:02:51 +0530 Subject: drm/amdgpu: Add VBIOS flags Instead of read_bios, use get_bios_flags to get various options around reading VBIOS. Signed-off-by: Lijo Lazar Reviewed-by: Alex Deucher Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 20 +++++++++++++------- 1 file changed, 13 insertions(+), 7 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 947289eb1b2f..e0aea865f76b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -102,6 +102,8 @@ MODULE_FIRMWARE("amdgpu/navi12_gpu_info.bin"); #define AMDGPU_PCIE_INDEX_HI_FALLBACK (0x44 >> 2) #define AMDGPU_PCIE_DATA_FALLBACK (0x3C >> 2) +#define AMDGPU_VBIOS_SKIP (1U << 0) + static const struct drm_driver amdgpu_kms_driver; const char *amdgpu_asic_name[] = { @@ -1698,12 +1700,12 @@ int amdgpu_device_resize_fb_bar(struct amdgpu_device *adev) return 0; } -static bool amdgpu_device_read_bios(struct amdgpu_device *adev) +static uint32_t amdgpu_device_get_vbios_flags(struct amdgpu_device *adev) { if (hweight32(adev->aid_mask) && (adev->flags & AMD_IS_APU)) - return false; + return AMDGPU_VBIOS_SKIP; - return true; + return 0; } /* @@ -1720,12 +1722,13 @@ static bool amdgpu_device_read_bios(struct amdgpu_device *adev) */ bool amdgpu_device_need_post(struct amdgpu_device *adev) { - uint32_t reg; + uint32_t reg, flags; if (amdgpu_sriov_vf(adev)) return false; - if (!amdgpu_device_read_bios(adev)) + flags = amdgpu_device_get_vbios_flags(adev); + if (flags & AMDGPU_VBIOS_SKIP) return false; if (amdgpu_passthrough(adev)) { @@ -2578,8 +2581,9 @@ static int amdgpu_device_ip_early_init(struct amdgpu_device *adev) { struct amdgpu_ip_block *ip_block; struct pci_dev *parent; + bool total, skip_bios; + uint32_t bios_flags; int i, r; - bool total; amdgpu_device_enable_virtual_display(adev); @@ -2692,8 +2696,10 @@ static int amdgpu_device_ip_early_init(struct amdgpu_device *adev) if (r) return r; + bios_flags = amdgpu_device_get_vbios_flags(adev); + skip_bios = !!(bios_flags & AMDGPU_VBIOS_SKIP); /* Read BIOS */ - if (amdgpu_device_read_bios(adev)) { + if (!skip_bios) { if (!amdgpu_get_bios(adev)) return -EINVAL; -- cgit v1.2.3 From 6e8ca38ebc9b130f5be11d4c6e2450f7f838449a Mon Sep 17 00:00:00 2001 From: Lijo Lazar Date: Wed, 5 Feb 2025 14:06:58 +0530 Subject: drm/amdgpu: Add flag to make VBIOS read optional Certain SOCs may not need much data from VBIOS. Some data like VBIOS version used will be missed but it doesn't affect functionality. Add a flag to make VBIOS image optional. Signed-off-by: Lijo Lazar Reviewed-by: Alex Deucher Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 69 +++++++++++++++++++++--------- 1 file changed, 49 insertions(+), 20 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index e0aea865f76b..a3f8f20da7ff 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -103,6 +103,7 @@ MODULE_FIRMWARE("amdgpu/navi12_gpu_info.bin"); #define AMDGPU_PCIE_DATA_FALLBACK (0x3C >> 2) #define AMDGPU_VBIOS_SKIP (1U << 0) +#define AMDGPU_VBIOS_OPTIONAL (1U << 1) static const struct drm_driver amdgpu_kms_driver; @@ -1391,6 +1392,14 @@ static void amdgpu_block_invalid_wreg(struct amdgpu_device *adev, BUG(); } +static uint32_t amdgpu_device_get_vbios_flags(struct amdgpu_device *adev) +{ + if (hweight32(adev->aid_mask) && (adev->flags & AMD_IS_APU)) + return AMDGPU_VBIOS_SKIP; + + return 0; +} + /** * amdgpu_device_asic_init - Wrapper for atom asic_init * @@ -1400,18 +1409,28 @@ static void amdgpu_block_invalid_wreg(struct amdgpu_device *adev, */ static int amdgpu_device_asic_init(struct amdgpu_device *adev) { + uint32_t flags; + bool optional; int ret; amdgpu_asic_pre_asic_init(adev); + flags = amdgpu_device_get_vbios_flags(adev); + optional = !!(flags & (AMDGPU_VBIOS_OPTIONAL | AMDGPU_VBIOS_SKIP)); if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 3) || amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 4) || amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 5, 0) || amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(11, 0, 0)) { amdgpu_psp_wait_for_bootloader(adev); + if (optional && !adev->bios) + return 0; + ret = amdgpu_atomfirmware_asic_init(adev, true); return ret; } else { + if (optional && !adev->bios) + return 0; + return amdgpu_atom_asic_init(adev->mode_info.atom_context); } @@ -1700,14 +1719,6 @@ int amdgpu_device_resize_fb_bar(struct amdgpu_device *adev) return 0; } -static uint32_t amdgpu_device_get_vbios_flags(struct amdgpu_device *adev) -{ - if (hweight32(adev->aid_mask) && (adev->flags & AMD_IS_APU)) - return AMDGPU_VBIOS_SKIP; - - return 0; -} - /* * GPU helpers function. */ @@ -1730,6 +1741,8 @@ bool amdgpu_device_need_post(struct amdgpu_device *adev) flags = amdgpu_device_get_vbios_flags(adev); if (flags & AMDGPU_VBIOS_SKIP) return false; + if ((flags & AMDGPU_VBIOS_OPTIONAL) && !adev->bios) + return false; if (amdgpu_passthrough(adev)) { /* for FIJI: In whole GPU pass-through virtualization case, after VM reboot @@ -2700,14 +2713,27 @@ static int amdgpu_device_ip_early_init(struct amdgpu_device *adev) skip_bios = !!(bios_flags & AMDGPU_VBIOS_SKIP); /* Read BIOS */ if (!skip_bios) { - if (!amdgpu_get_bios(adev)) + bool optional = + !!(bios_flags & AMDGPU_VBIOS_OPTIONAL); + if (!amdgpu_get_bios(adev) && !optional) return -EINVAL; - r = amdgpu_atombios_init(adev); - if (r) { - dev_err(adev->dev, "amdgpu_atombios_init failed\n"); - amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ATOMBIOS_INIT_FAIL, 0, 0); - return r; + if (optional && !adev->bios) + dev_info( + adev->dev, + "VBIOS image optional, proceeding without VBIOS image"); + + if (adev->bios) { + r = amdgpu_atombios_init(adev); + if (r) { + dev_err(adev->dev, + "amdgpu_atombios_init failed\n"); + amdgpu_vf_error_put( + adev, + AMDGIM_ERROR_VF_ATOMBIOS_INIT_FAIL, + 0, 0); + return r; + } } } @@ -4742,10 +4768,11 @@ void amdgpu_device_fini_sw(struct amdgpu_device *adev) /* free i2c buses */ amdgpu_i2c_fini(adev); - if (amdgpu_emu_mode != 1) - amdgpu_atombios_fini(adev); - - amdgpu_bios_release(adev); + if (adev->bios) { + if (amdgpu_emu_mode != 1) + amdgpu_atombios_fini(adev); + amdgpu_bios_release(adev); + } kfree(adev->fru_info); adev->fru_info = NULL; @@ -5371,7 +5398,8 @@ int amdgpu_device_mode1_reset(struct amdgpu_device *adev) u32 i; int ret = 0; - amdgpu_atombios_scratch_regs_engine_hung(adev, true); + if (adev->bios) + amdgpu_atombios_scratch_regs_engine_hung(adev, true); dev_info(adev->dev, "GPU mode1 reset\n"); @@ -5413,7 +5441,8 @@ int amdgpu_device_mode1_reset(struct amdgpu_device *adev) goto mode1_reset_failed; } - amdgpu_atombios_scratch_regs_engine_hung(adev, false); + if (adev->bios) + amdgpu_atombios_scratch_regs_engine_hung(adev, false); return 0; -- cgit v1.2.3 From cc0e91a75533d76def12f2bedb95b00a8f70cebc Mon Sep 17 00:00:00 2001 From: Lijo Lazar Date: Wed, 5 Feb 2025 14:10:48 +0530 Subject: drm/amdgpu: Make VBIOS image read optional Keep VBIOS image read optional for select SOCs in passthrough mode. Signed-off-by: Lijo Lazar Reviewed-by: Alex Deucher Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index a3f8f20da7ff..e1fc020a99bd 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -1397,6 +1397,9 @@ static uint32_t amdgpu_device_get_vbios_flags(struct amdgpu_device *adev) if (hweight32(adev->aid_mask) && (adev->flags & AMD_IS_APU)) return AMDGPU_VBIOS_SKIP; + if (hweight32(adev->aid_mask) && amdgpu_passthrough(adev)) + return AMDGPU_VBIOS_OPTIONAL; + return 0; } -- cgit v1.2.3 From 6fe52b63f569c11a70b737751055afd46c4454b3 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Andr=C3=A9=20Almeida?= Date: Tue, 4 Feb 2025 12:35:28 +0530 Subject: drm/amdgpu: Use device wedged event MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Use DRM's device wedged event to notify userspace that a reset had happened. For now, only use `none` method meant for telemetry capture. In the future we might want to report a recovery method if the reset didn't succeed. Acked-by: Shashank Sharma Signed-off-by: André Almeida Reviewed-by: Christian König Link: https://patchwork.freedesktop.org/patch/msgid/20250204070528.1919158-6-raag.jadav@intel.com Signed-off-by: Rodrigo Vivi --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 6800f72fafd2..24ba52d76045 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -6122,6 +6122,10 @@ end_reset: dev_info(adev->dev, "GPU reset end with ret = %d\n", r); atomic_set(&adev->reset_domain->reset_res, r); + + if (!r) + drm_dev_wedged_event(adev_to_drm(adev), DRM_WEDGE_RECOVERY_NONE); + return r; } -- cgit v1.2.3 From 92d5d2a09de16706fca340b6c1c197e562690da4 Mon Sep 17 00:00:00 2001 From: Hawking Zhang Date: Fri, 24 Jan 2025 23:37:33 +0800 Subject: drm/amdgpu: Introduce funcs for populating CPER Introduce utility functions designed to assist in populating CPER records. v2: call cper_init/fini in device_ip_init/fini. Signed-off-by: Hawking Zhang Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index e1fc020a99bd..0de2476c2ee7 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -3091,6 +3091,8 @@ static int amdgpu_device_ip_init(struct amdgpu_device *adev) amdgpu_fru_get_product_info(adev); + r = amdgpu_cper_init(adev); + init_failed: return r; @@ -3451,6 +3453,8 @@ static int amdgpu_device_ip_fini(struct amdgpu_device *adev) { int i, r; + amdgpu_cper_fini(adev); + if (amdgpu_sriov_vf(adev) && adev->virt.ras_init_done) amdgpu_virt_release_ras_err_handler_data(adev); -- cgit v1.2.3 From 4d614ce8ffd757e4c7944bf9b5598b4a250a8a61 Mon Sep 17 00:00:00 2001 From: Tao Zhou Date: Wed, 22 Jan 2025 16:55:51 +0800 Subject: drm/amdgpu: add RAS CPER ring buffer And initialize it, this is a pure software ring to store RAS CPER data. v2: change ring size to 0x100000 v2: update the initialization of count_dw of cper ring, it's dword variable v3: skip VM inv eng for cper v3: init/fini when aca enabled Signed-off-by: Tao Zhou Signed-off-by: Xiang Liu Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 0de2476c2ee7..1230ab2ba112 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -3091,7 +3091,8 @@ static int amdgpu_device_ip_init(struct amdgpu_device *adev) amdgpu_fru_get_product_info(adev); - r = amdgpu_cper_init(adev); + if (amdgpu_aca_is_enabled(adev)) + r = amdgpu_cper_init(adev); init_failed: @@ -3453,7 +3454,8 @@ static int amdgpu_device_ip_fini(struct amdgpu_device *adev) { int i, r; - amdgpu_cper_fini(adev); + if (amdgpu_aca_is_enabled(adev)) + amdgpu_cper_fini(adev); if (amdgpu_sriov_vf(adev) && adev->virt.ras_init_done) amdgpu_virt_release_ras_err_handler_data(adev); -- cgit v1.2.3 From dc0297f3198bd60108ccbd167ee5d9fa4af31ed0 Mon Sep 17 00:00:00 2001 From: Srinivasan Shanmugam Date: Fri, 14 Feb 2025 14:22:17 +0530 Subject: drm/amdgpu: Replace Mutex with Spinlock for RLCG register access to avoid Priority Inversion in SRIOV MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit RLCG Register Access is a way for virtual functions to safely access GPU registers in a virtualized environment., including TLB flushes and register reads. When multiple threads or VFs try to access the same registers simultaneously, it can lead to race conditions. By using the RLCG interface, the driver can serialize access to the registers. This means that only one thread can access the registers at a time, preventing conflicts and ensuring that operations are performed correctly. Additionally, when a low-priority task holds a mutex that a high-priority task needs, ie., If a thread holding a spinlock tries to acquire a mutex, it can lead to priority inversion. register access in amdgpu_virt_rlcg_reg_rw especially in a fast code path is critical. The call stack shows that the function amdgpu_virt_rlcg_reg_rw is being called, which attempts to acquire the mutex. This function is invoked from amdgpu_sriov_wreg, which in turn is called from gmc_v11_0_flush_gpu_tlb. The [ BUG: Invalid wait context ] indicates that a thread is trying to acquire a mutex while it is in a context that does not allow it to sleep (like holding a spinlock). Fixes the below: [ 253.013423] ============================= [ 253.013434] [ BUG: Invalid wait context ] [ 253.013446] 6.12.0-amdstaging-drm-next-lol-050225 #14 Tainted: G U OE [ 253.013464] ----------------------------- [ 253.013475] kworker/0:1/10 is trying to lock: [ 253.013487] ffff9f30542e3cf8 (&adev->virt.rlcg_reg_lock){+.+.}-{3:3}, at: amdgpu_virt_rlcg_reg_rw+0xf6/0x330 [amdgpu] [ 253.013815] other info that might help us debug this: [ 253.013827] context-{4:4} [ 253.013835] 3 locks held by kworker/0:1/10: [ 253.013847] #0: ffff9f3040050f58 ((wq_completion)events){+.+.}-{0:0}, at: process_one_work+0x3f5/0x680 [ 253.013877] #1: ffffb789c008be40 ((work_completion)(&wfc.work)){+.+.}-{0:0}, at: process_one_work+0x1d6/0x680 [ 253.013905] #2: ffff9f3054281838 (&adev->gmc.invalidate_lock){+.+.}-{2:2}, at: gmc_v11_0_flush_gpu_tlb+0x198/0x4f0 [amdgpu] [ 253.014154] stack backtrace: [ 253.014164] CPU: 0 UID: 0 PID: 10 Comm: kworker/0:1 Tainted: G U OE 6.12.0-amdstaging-drm-next-lol-050225 #14 [ 253.014189] Tainted: [U]=USER, [O]=OOT_MODULE, [E]=UNSIGNED_MODULE [ 253.014203] Hardware name: Microsoft Corporation Virtual Machine/Virtual Machine, BIOS Hyper-V UEFI Release v4.1 11/18/2024 [ 253.014224] Workqueue: events work_for_cpu_fn [ 253.014241] Call Trace: [ 253.014250] [ 253.014260] dump_stack_lvl+0x9b/0xf0 [ 253.014275] dump_stack+0x10/0x20 [ 253.014287] __lock_acquire+0xa47/0x2810 [ 253.014303] ? srso_alias_return_thunk+0x5/0xfbef5 [ 253.014321] lock_acquire+0xd1/0x300 [ 253.014333] ? amdgpu_virt_rlcg_reg_rw+0xf6/0x330 [amdgpu] [ 253.014562] ? __lock_acquire+0xa6b/0x2810 [ 253.014578] __mutex_lock+0x85/0xe20 [ 253.014591] ? amdgpu_virt_rlcg_reg_rw+0xf6/0x330 [amdgpu] [ 253.014782] ? sched_clock_noinstr+0x9/0x10 [ 253.014795] ? srso_alias_return_thunk+0x5/0xfbef5 [ 253.014808] ? local_clock_noinstr+0xe/0xc0 [ 253.014822] ? amdgpu_virt_rlcg_reg_rw+0xf6/0x330 [amdgpu] [ 253.015012] ? srso_alias_return_thunk+0x5/0xfbef5 [ 253.015029] mutex_lock_nested+0x1b/0x30 [ 253.015044] ? mutex_lock_nested+0x1b/0x30 [ 253.015057] amdgpu_virt_rlcg_reg_rw+0xf6/0x330 [amdgpu] [ 253.015249] amdgpu_sriov_wreg+0xc5/0xd0 [amdgpu] [ 253.015435] gmc_v11_0_flush_gpu_tlb+0x44b/0x4f0 [amdgpu] [ 253.015667] gfx_v11_0_hw_init+0x499/0x29c0 [amdgpu] [ 253.015901] ? __pfx_smu_v13_0_update_pcie_parameters+0x10/0x10 [amdgpu] [ 253.016159] ? srso_alias_return_thunk+0x5/0xfbef5 [ 253.016173] ? smu_hw_init+0x18d/0x300 [amdgpu] [ 253.016403] amdgpu_device_init+0x29ad/0x36a0 [amdgpu] [ 253.016614] amdgpu_driver_load_kms+0x1a/0xc0 [amdgpu] [ 253.017057] amdgpu_pci_probe+0x1c2/0x660 [amdgpu] [ 253.017493] local_pci_probe+0x4b/0xb0 [ 253.017746] work_for_cpu_fn+0x1a/0x30 [ 253.017995] process_one_work+0x21e/0x680 [ 253.018248] worker_thread+0x190/0x330 [ 253.018500] ? __pfx_worker_thread+0x10/0x10 [ 253.018746] kthread+0xe7/0x120 [ 253.018988] ? __pfx_kthread+0x10/0x10 [ 253.019231] ret_from_fork+0x3c/0x60 [ 253.019468] ? __pfx_kthread+0x10/0x10 [ 253.019701] ret_from_fork_asm+0x1a/0x30 [ 253.019939] v2: s/spin_trylock/spin_lock_irqsave to be safe (Christian). Fixes: e864180ee49b ("drm/amdgpu: Add lock around VF RLCG interface") Cc: lin cao Cc: Jingwen Chen Cc: Victor Skvortsov Cc: Zhigang Luo Cc: Christian König Cc: Alex Deucher Signed-off-by: Srinivasan Shanmugam Suggested-by: Alex Deucher Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 1230ab2ba112..5aa0b4032291 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -4257,7 +4257,6 @@ int amdgpu_device_init(struct amdgpu_device *adev, mutex_init(&adev->grbm_idx_mutex); mutex_init(&adev->mn_lock); mutex_init(&adev->virt.vf_errors.lock); - mutex_init(&adev->virt.rlcg_reg_lock); hash_init(adev->mn_hash); mutex_init(&adev->psp.mutex); mutex_init(&adev->notifier_lock); @@ -4283,6 +4282,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, spin_lock_init(&adev->se_cac_idx_lock); spin_lock_init(&adev->audio_endpt_idx_lock); spin_lock_init(&adev->mm_stats.lock); + spin_lock_init(&adev->virt.rlcg_reg_lock); spin_lock_init(&adev->wb.lock); INIT_LIST_HEAD(&adev->reset_list); -- cgit v1.2.3 From 663a87763b570d4e92d821b30508bed0025fa285 Mon Sep 17 00:00:00 2001 From: Xiang Liu Date: Wed, 19 Feb 2025 12:21:59 +0800 Subject: drm/amdgpu: Check aca enabled inside cper init/fini func Move code about checking aca enabled to the cper init/fini function to make code clean. Signed-off-by: Xiang Liu Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 5aa0b4032291..38260ffac6ee 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -3091,8 +3091,7 @@ static int amdgpu_device_ip_init(struct amdgpu_device *adev) amdgpu_fru_get_product_info(adev); - if (amdgpu_aca_is_enabled(adev)) - r = amdgpu_cper_init(adev); + r = amdgpu_cper_init(adev); init_failed: @@ -3454,8 +3453,7 @@ static int amdgpu_device_ip_fini(struct amdgpu_device *adev) { int i, r; - if (amdgpu_aca_is_enabled(adev)) - amdgpu_cper_fini(adev); + amdgpu_cper_fini(adev); if (amdgpu_sriov_vf(adev) && adev->virt.ras_init_done) amdgpu_virt_release_ras_err_handler_data(adev); -- cgit v1.2.3 From 3521276ad14fe47ce1c4382749f3c95762629375 Mon Sep 17 00:00:00 2001 From: Sunil Khatri Date: Wed, 19 Feb 2025 20:59:02 +0530 Subject: drm/amdgpu: update the handle ptr in get_clockgating_state Update the *handle to amdgpu_ip_block ptr for all functions pointers of get_clockgating_state. Signed-off-by: Sunil Khatri Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 38260ffac6ee..c8ceea01a221 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -2252,7 +2252,8 @@ void amdgpu_device_ip_get_clockgating_state(struct amdgpu_device *adev, if (!adev->ip_blocks[i].status.valid) continue; if (adev->ip_blocks[i].version->funcs->get_clockgating_state) - adev->ip_blocks[i].version->funcs->get_clockgating_state((void *)adev, flags); + adev->ip_blocks[i].version->funcs->get_clockgating_state( + &adev->ip_blocks[i], flags); } } -- cgit v1.2.3 From 2d0f5001b61c4831d413d12c10caed0e99d73b25 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Thomas=20Wei=C3=9Fschuh?= Date: Mon, 16 Dec 2024 12:34:50 +0100 Subject: drm/amdgpu: Constify 'struct bin_attribute' MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The sysfs core now allows instances of 'struct bin_attribute' to be moved into read-only memory. Make use of that to protect them against accidental or malicious modifications. Signed-off-by: Thomas Weißschuh Reviewed-by: Alex Deucher Link: https://lore.kernel.org/r/20241216-sysfs-const-bin_attr-drm-v1-4-210f2b36b9bf@weissschuh.net Signed-off-by: Greg Kroah-Hartman --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index d100bb7a137c..e6fa63f97687 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -225,7 +225,7 @@ static DEVICE_ATTR(pcie_replay_count, 0444, amdgpu_device_get_pcie_replay_count, NULL); static ssize_t amdgpu_sysfs_reg_state_get(struct file *f, struct kobject *kobj, - struct bin_attribute *attr, char *buf, + const struct bin_attribute *attr, char *buf, loff_t ppos, size_t count) { struct device *dev = kobj_to_dev(kobj); @@ -261,8 +261,8 @@ static ssize_t amdgpu_sysfs_reg_state_get(struct file *f, struct kobject *kobj, return bytes_read; } -BIN_ATTR(reg_state, 0444, amdgpu_sysfs_reg_state_get, NULL, - AMDGPU_SYS_REG_STATE_END); +static const BIN_ATTR(reg_state, 0444, amdgpu_sysfs_reg_state_get, NULL, + AMDGPU_SYS_REG_STATE_END); int amdgpu_reg_state_sysfs_init(struct amdgpu_device *adev) { -- cgit v1.2.3 From 5235053f443cef4210606e5fb71f99b915a9723d Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Mon, 17 Feb 2025 10:55:05 -0500 Subject: drm/amdgpu: disable BAR resize on Dell G5 SE There was a quirk added to add a workaround for a Sapphire RX 5600 XT Pulse that didn't allow BAR resizing. However, the quirk caused a regression with runtime pm on Dell laptops using those chips, rather than narrowing the scope of the resizing quirk, add a quirk to prevent amdgpu from resizing the BAR on those Dell platforms unless runtime pm is disabled. v2: update commit message, add runpm check Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/1707 Fixes: 907830b0fc9e ("PCI: Add a REBAR size quirk for Sapphire RX 5600 XT Pulse") Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index c8ceea01a221..7b60646c6ad2 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -1662,6 +1662,13 @@ int amdgpu_device_resize_fb_bar(struct amdgpu_device *adev) if (amdgpu_sriov_vf(adev)) return 0; + /* resizing on Dell G5 SE platforms causes problems with runtime pm */ + if ((amdgpu_runtime_pm != 0) && + adev->pdev->vendor == PCI_VENDOR_ID_ATI && + adev->pdev->device == 0x731f && + adev->pdev->subsystem_vendor == PCI_VENDOR_ID_DELL) + return 0; + /* PCI_EXT_CAP_ID_VNDR extended capability is located at 0x100 */ if (!pci_find_ext_capability(adev->pdev, PCI_EXT_CAP_ID_VNDR)) DRM_WARN("System can't access extended configuration space, please check!!\n"); -- cgit v1.2.3 From 099bffc7cadff40bfab1517c3461c53a7a38a0d7 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Mon, 17 Feb 2025 10:55:05 -0500 Subject: drm/amdgpu: disable BAR resize on Dell G5 SE There was a quirk added to add a workaround for a Sapphire RX 5600 XT Pulse that didn't allow BAR resizing. However, the quirk caused a regression with runtime pm on Dell laptops using those chips, rather than narrowing the scope of the resizing quirk, add a quirk to prevent amdgpu from resizing the BAR on those Dell platforms unless runtime pm is disabled. v2: update commit message, add runpm check Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/1707 Fixes: 907830b0fc9e ("PCI: Add a REBAR size quirk for Sapphire RX 5600 XT Pulse") Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher (cherry picked from commit 5235053f443cef4210606e5fb71f99b915a9723d) Cc: stable@vger.kernel.org --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index d100bb7a137c..018dfccd771b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -1638,6 +1638,13 @@ int amdgpu_device_resize_fb_bar(struct amdgpu_device *adev) if (amdgpu_sriov_vf(adev)) return 0; + /* resizing on Dell G5 SE platforms causes problems with runtime pm */ + if ((amdgpu_runtime_pm != 0) && + adev->pdev->vendor == PCI_VENDOR_ID_ATI && + adev->pdev->device == 0x731f && + adev->pdev->subsystem_vendor == PCI_VENDOR_ID_DELL) + return 0; + /* PCI_EXT_CAP_ID_VNDR extended capability is located at 0x100 */ if (!pci_find_ext_capability(adev->pdev, PCI_EXT_CAP_ID_VNDR)) DRM_WARN("System can't access extended configuration space, please check!!\n"); -- cgit v1.2.3 From a91d91b6004796b868374394962331a1322da7ab Mon Sep 17 00:00:00 2001 From: Tony Yi Date: Wed, 26 Feb 2025 17:03:10 -0500 Subject: drm/amdgpu: Add support for CPERs on virtualization Add support for CPERs on VFs. VFs do not receive PMFW messages directly; as such, they need to query them from the host. To avoid hitting host event guard, CPER queries need to be rate limited. CPER queries share the same RAS telemetry buffer as error count query, so a mutex protecting the shared buffer was added as well. For readability, the amdgpu_detect_virtualization was refactored into multiple individual functions. Signed-off-by: Tony Yi Reviewed-by: Tao Zhou Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 7b60646c6ad2..a755cc545c91 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -3099,7 +3099,8 @@ static int amdgpu_device_ip_init(struct amdgpu_device *adev) amdgpu_fru_get_product_info(adev); - r = amdgpu_cper_init(adev); + if (!amdgpu_sriov_vf(adev) || amdgpu_sriov_ras_cper_en(adev)) + r = amdgpu_cper_init(adev); init_failed: @@ -4333,10 +4334,8 @@ int amdgpu_device_init(struct amdgpu_device *adev, * for throttling interrupt) = 60 seconds. */ ratelimit_state_init(&adev->throttling_logging_rs, (60 - 1) * HZ, 1); - ratelimit_state_init(&adev->virt.ras_telemetry_rs, 5 * HZ, 1); ratelimit_set_flags(&adev->throttling_logging_rs, RATELIMIT_MSG_ON_RELEASE); - ratelimit_set_flags(&adev->virt.ras_telemetry_rs, RATELIMIT_MSG_ON_RELEASE); /* Registers mapping */ /* TODO: block userspace mapping of io register */ @@ -4368,7 +4367,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, return -ENOMEM; /* detect hw virtualization here */ - amdgpu_detect_virtualization(adev); + amdgpu_virt_init(adev); amdgpu_device_get_pcie_info(adev); -- cgit v1.2.3 From 357506799bacb21ba1ed7d64e02ac9c4a1f45b31 Mon Sep 17 00:00:00 2001 From: Lijo Lazar Date: Thu, 6 Feb 2025 17:40:42 +0530 Subject: drm/amdgpu: Calculate IP specific xgmi bandwidth Use IP version specific xgmi speed/width for bandwidth calculation. Signed-off-by: Lijo Lazar Reviewed-by: Jonathan Kim Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 9e9fec5b52de..17ce1677378e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -2757,6 +2757,9 @@ static int amdgpu_device_ip_early_init(struct amdgpu_device *adev) if (!total) return -ENODEV; + if (adev->gmc.xgmi.supported) + amdgpu_xgmi_early_init(adev); + ip_block = amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_GFX); if (ip_block->status.valid != false) amdgpu_amdkfd_device_probe(adev); -- cgit v1.2.3 From 9deacd6c55f1b31e5ab20db79df2e14ac480203c Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Wed, 25 Sep 2024 10:29:31 -0400 Subject: drm/amdgpu: don't free conflicting apertures for non-display devices PCI_CLASS_ACCELERATOR_PROCESSING devices won't ever be the sysfb, so there is no need to free conflicting apertures. Reviewed-by: Kent Russell Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 15 +++++++++++---- 1 file changed, 11 insertions(+), 4 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 17ce1677378e..cc0b6c70fcb8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -4399,10 +4399,17 @@ int amdgpu_device_init(struct amdgpu_device *adev, if (r) return r; - /* Get rid of things like offb */ - r = aperture_remove_conflicting_pci_devices(adev->pdev, amdgpu_kms_driver.name); - if (r) - return r; + /* + * No need to remove conflicting FBs for non-display class devices. + * This prevents the sysfb from being freed accidently. + */ + if ((pdev->class >> 8) == PCI_CLASS_DISPLAY_VGA || + (pdev->class >> 8) == PCI_CLASS_DISPLAY_OTHER) { + /* Get rid of things like offb */ + r = aperture_remove_conflicting_pci_devices(adev->pdev, amdgpu_kms_driver.name); + if (r) + return r; + } /* Enable TMZ based on IP_VERSION */ amdgpu_gmc_tmz_set(adev); -- cgit v1.2.3 From b5aaa82e2b12feaaa6958f7fa0917ddcc03c24ee Mon Sep 17 00:00:00 2001 From: Flora Cui Date: Fri, 14 Mar 2025 10:27:55 +0800 Subject: drm/amdgpu: release xcp_mgr on exit Free on driver cleanup. Reviewed-by: Lijo Lazar Signed-off-by: Flora Cui Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index cc0b6c70fcb8..e63bcd5edab9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -4807,6 +4807,9 @@ void amdgpu_device_fini_sw(struct amdgpu_device *adev) kfree(adev->fru_info); adev->fru_info = NULL; + kfree(adev->xcp_mgr); + adev->xcp_mgr = NULL; + px = amdgpu_device_supports_px(adev_to_drm(adev)); if (px || (!dev_is_removable(&adev->pdev->dev) && -- cgit v1.2.3 From 0d9a95099dcb05b5f4719c830d15bf4fdcad0dc2 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Christian=20K=C3=B6nig?= Date: Tue, 14 Jan 2025 13:51:39 +0100 Subject: drm/amdgpu: grab an additional reference on the gang fence v2 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We keep the gang submission fence around in adev, make sure that it stays alive. v2: fix memory leak on retry Signed-off-by: Christian König Acked-by: Srinivasan Shanmugam Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index e63bcd5edab9..35ace1368b9c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -6913,18 +6913,26 @@ struct dma_fence *amdgpu_device_switch_gang(struct amdgpu_device *adev, { struct dma_fence *old = NULL; + dma_fence_get(gang); do { dma_fence_put(old); old = amdgpu_device_get_gang(adev); if (old == gang) break; - if (!dma_fence_is_signaled(old)) + if (!dma_fence_is_signaled(old)) { + dma_fence_put(gang); return old; + } } while (cmpxchg((struct dma_fence __force **)&adev->gang_submit, old, gang) != old); + /* + * Drop it once for the exchanged reference in adev and once for the + * thread local reference acquired in amdgpu_device_get_gang(). + */ + dma_fence_put(old); dma_fence_put(old); return NULL; } -- cgit v1.2.3 From 553673a3e1bc0abbb994d9884f772189c739e3a0 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Wed, 12 Mar 2025 09:44:19 -0400 Subject: drm/amdgpu/gfx: fix ref counting for ring based profile handling We need to make sure the workload profile ref counts are balanced. This isn't currently the case because we can increment the count on submissions, but the decrement may be delayed as work comes in. Track when we enable the workload profile so the references are balanced. v2: switch to a mutex and active flag v3: fix mutex init Fixes: 8fdb3958e396 ("drm/amdgpu/gfx: add ring helpers for setting workload profile") Cc: Yang Wang Cc: Kenneth Feng Tested-by: Kenneth Feng Reviewed-by: Kenneth Feng Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 1 + 1 file changed, 1 insertion(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 35ace1368b9c..173048c46fc3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -4282,6 +4282,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, /* Initialize the mutex for cleaner shader isolation between GFX and compute processes */ mutex_init(&adev->enforce_isolation_mutex); mutex_init(&adev->gfx.kfd_sch_mutex); + mutex_init(&adev->gfx.workload_profile_mutex); amdgpu_device_init_apu_flags(adev); -- cgit v1.2.3 From ca6575a32a37c1a3ebab904792bb415249481992 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Wed, 12 Mar 2025 09:48:47 -0400 Subject: drm/amdgpu/vcn: fix ref counting for ring based profile handling We need to make sure the workload profile ref counts are balanced. This isn't currently the case because we can increment the count on submissions, but the decrement may be delayed as work comes in. Track when we enable the workload profile so the references are balanced. v2: switch to a mutex and active flag v3: fix mutex init Fixes: 1443dd3c67f6 ("drm/amd/pm: fix and simplify workload handling") Cc: Yang Wang Cc: Kenneth Feng Reviewed-by: Kenneth Feng Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 1 + 1 file changed, 1 insertion(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 173048c46fc3..44d3afad71d2 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -4283,6 +4283,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, mutex_init(&adev->enforce_isolation_mutex); mutex_init(&adev->gfx.kfd_sch_mutex); mutex_init(&adev->gfx.workload_profile_mutex); + mutex_init(&adev->vcn.workload_profile_mutex); amdgpu_device_init_apu_flags(adev); -- cgit v1.2.3 From 9c05636ca72a2dbf41bf0900380f438a0de47319 Mon Sep 17 00:00:00 2001 From: Victor Skvortsov Date: Mon, 17 Mar 2025 09:32:13 -0400 Subject: drm/amdgpu: Skip pcie_replay_count sysfs creation for VF VFs cannot read the NAK_COUNTER register. This information is only available through PMFW metrics. Signed-off-by: Victor Skvortsov Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 27 ++++++++++++++++++++------- 1 file changed, 20 insertions(+), 7 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 44d3afad71d2..993f2b3ff8a4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -227,6 +227,24 @@ static ssize_t amdgpu_device_get_pcie_replay_count(struct device *dev, static DEVICE_ATTR(pcie_replay_count, 0444, amdgpu_device_get_pcie_replay_count, NULL); +static int amdgpu_device_attr_sysfs_init(struct amdgpu_device *adev) +{ + int ret = 0; + + if (!amdgpu_sriov_vf(adev)) + ret = sysfs_create_file(&adev->dev->kobj, + &dev_attr_pcie_replay_count.attr); + + return ret; +} + +static void amdgpu_device_attr_sysfs_fini(struct amdgpu_device *adev) +{ + if (!amdgpu_sriov_vf(adev)) + sysfs_remove_file(&adev->dev->kobj, + &dev_attr_pcie_replay_count.attr); +} + static ssize_t amdgpu_sysfs_reg_state_get(struct file *f, struct kobject *kobj, struct bin_attribute *attr, char *buf, loff_t ppos, size_t count) @@ -4172,11 +4190,6 @@ static bool amdgpu_device_check_iommu_remap(struct amdgpu_device *adev) } #endif -static const struct attribute *amdgpu_dev_attributes[] = { - &dev_attr_pcie_replay_count.attr, - NULL -}; - static void amdgpu_device_set_mcbp(struct amdgpu_device *adev) { if (amdgpu_mcbp == 1) @@ -4622,7 +4635,7 @@ fence_driver_init: } else adev->ucode_sysfs_en = true; - r = sysfs_create_files(&adev->dev->kobj, amdgpu_dev_attributes); + r = amdgpu_device_attr_sysfs_init(adev); if (r) dev_err(adev->dev, "Could not create amdgpu device attr\n"); @@ -4759,7 +4772,7 @@ void amdgpu_device_fini_hw(struct amdgpu_device *adev) amdgpu_pm_sysfs_fini(adev); if (adev->ucode_sysfs_en) amdgpu_ucode_sysfs_fini(adev); - sysfs_remove_files(&adev->dev->kobj, amdgpu_dev_attributes); + amdgpu_device_attr_sysfs_fini(adev); amdgpu_fru_sysfs_fini(adev); amdgpu_reg_state_sysfs_fini(adev); -- cgit v1.2.3 From d7f5c13e457ba6b0c14a58b12fade55ab7ae5fb0 Mon Sep 17 00:00:00 2001 From: Ellen Pan Date: Tue, 18 Mar 2025 09:45:18 -0400 Subject: drm/amdgpu: Enable amdgpu_ras_resume for gfx 9.5.0 This enables ras to be resumed after gpu recovery on mi350 sriov. Signed-off-by: Ellen Pan Reviewed-by: Ahmad Rehman Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 1 + 1 file changed, 1 insertion(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 993f2b3ff8a4..4ed5e26a13d5 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -5356,6 +5356,7 @@ static int amdgpu_device_reset_sriov(struct amdgpu_device *adev, if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 2) || amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 3) || amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 4) || + amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 5, 0) || amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(11, 0, 3)) amdgpu_ras_resume(adev); -- cgit v1.2.3 From bd22e44ad415ac22e3a4f9a983d2a085f6cb4427 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Christian=20K=C3=B6nig?= Date: Wed, 15 Jan 2025 13:44:26 +0100 Subject: drm/amdgpu: rework how isolation is enforced v2 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Limiting the number of available VMIDs to enforce isolation causes some issues with gang submit and applying certain HW workarounds which require multiple VMIDs to work correctly. So instead start to track all submissions to the relevant engines in a per partition data structure and use the dma_fences of the submissions to enforce isolation similar to what a VMID limit does. v2: use ~0l for jobs without isolation to distinct it from kernel submissions which uses NULL for the owner. Add some warning when we are OOM. Signed-off-by: Christian König Acked-by: Srinivasan Shanmugam Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 98 +++++++++++++++++++++++++++++- 1 file changed, 97 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 4ed5e26a13d5..662c83d72509 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -4294,6 +4294,11 @@ int amdgpu_device_init(struct amdgpu_device *adev, mutex_init(&adev->gfx.reset_sem_mutex); /* Initialize the mutex for cleaner shader isolation between GFX and compute processes */ mutex_init(&adev->enforce_isolation_mutex); + for (i = 0; i < MAX_XCP; ++i) { + adev->isolation[i].spearhead = dma_fence_get_stub(); + amdgpu_sync_create(&adev->isolation[i].active); + amdgpu_sync_create(&adev->isolation[i].prev); + } mutex_init(&adev->gfx.kfd_sch_mutex); mutex_init(&adev->gfx.workload_profile_mutex); mutex_init(&adev->vcn.workload_profile_mutex); @@ -4799,7 +4804,7 @@ void amdgpu_device_fini_hw(struct amdgpu_device *adev) void amdgpu_device_fini_sw(struct amdgpu_device *adev) { - int idx; + int i, idx; bool px; amdgpu_device_ip_fini(adev); @@ -4807,6 +4812,11 @@ void amdgpu_device_fini_sw(struct amdgpu_device *adev) amdgpu_ucode_release(&adev->firmware.gpu_info_fw); adev->accel_working = false; dma_fence_put(rcu_dereference_protected(adev->gang_submit, true)); + for (i = 0; i < MAX_XCP; ++i) { + dma_fence_put(adev->isolation[i].spearhead); + amdgpu_sync_free(&adev->isolation[i].active); + amdgpu_sync_free(&adev->isolation[i].prev); + } amdgpu_reset_fini(adev); @@ -6953,6 +6963,92 @@ struct dma_fence *amdgpu_device_switch_gang(struct amdgpu_device *adev, return NULL; } +/** + * amdgpu_device_enforce_isolation - enforce HW isolation + * @adev: the amdgpu device pointer + * @ring: the HW ring the job is supposed to run on + * @job: the job which is about to be pushed to the HW ring + * + * Makes sure that only one client at a time can use the GFX block. + * Returns: The dependency to wait on before the job can be pushed to the HW. + * The function is called multiple times until NULL is returned. + */ +struct dma_fence *amdgpu_device_enforce_isolation(struct amdgpu_device *adev, + struct amdgpu_ring *ring, + struct amdgpu_job *job) +{ + struct amdgpu_isolation *isolation = &adev->isolation[ring->xcp_id]; + struct drm_sched_fence *f = job->base.s_fence; + struct dma_fence *dep; + void *owner; + int r; + + /* + * For now enforce isolation only for the GFX block since we only need + * the cleaner shader on those rings. + */ + if (ring->funcs->type != AMDGPU_RING_TYPE_GFX && + ring->funcs->type != AMDGPU_RING_TYPE_COMPUTE) + return NULL; + + /* + * All submissions where enforce isolation is false are handled as if + * they come from a single client. Use ~0l as the owner to distinct it + * from kernel submissions where the owner is NULL. + */ + owner = job->enforce_isolation ? f->owner : (void *)~0l; + + mutex_lock(&adev->enforce_isolation_mutex); + + /* + * The "spearhead" submission is the first one which changes the + * ownership to its client. We always need to wait for it to be + * pushed to the HW before proceeding with anything. + */ + if (&f->scheduled != isolation->spearhead && + !dma_fence_is_signaled(isolation->spearhead)) { + dep = isolation->spearhead; + goto out_grab_ref; + } + + if (isolation->owner != owner) { + + /* + * Wait for any gang to be assembled before switching to a + * different owner or otherwise we could deadlock the + * submissions. + */ + if (!job->gang_submit) { + dep = amdgpu_device_get_gang(adev); + if (!dma_fence_is_signaled(dep)) + goto out_return_dep; + dma_fence_put(dep); + } + + dma_fence_put(isolation->spearhead); + isolation->spearhead = dma_fence_get(&f->scheduled); + amdgpu_sync_move(&isolation->active, &isolation->prev); + isolation->owner = owner; + } + + /* + * Specifying the ring here helps to pipeline submissions even when + * isolation is enabled. If that is not desired for testing NULL can be + * used instead of the ring to enforce a CPU round trip while switching + * between clients. + */ + dep = amdgpu_sync_peek_fence(&isolation->prev, ring); + r = amdgpu_sync_fence(&isolation->active, &f->finished, GFP_NOWAIT); + if (r) + DRM_WARN("OOM tracking isolation\n"); + +out_grab_ref: + dma_fence_get(dep); +out_return_dep: + mutex_unlock(&adev->enforce_isolation_mutex); + return dep; +} + bool amdgpu_device_has_display_hardware(struct amdgpu_device *adev) { switch (adev->asic_type) { -- cgit v1.2.3 From 1bb1314d0b15dc6fc2178fe169f5121818094e8a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Christian=20K=C3=B6nig?= Date: Thu, 6 Feb 2025 14:16:06 +0100 Subject: drm/amdgpu: add isolation trace point MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Note when we switch from one isolation owner to another. Signed-off-by: Christian König Acked-by: Srinivasan Shanmugam Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 1 + 1 file changed, 1 insertion(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 662c83d72509..6ebf6179064b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -7028,6 +7028,7 @@ struct dma_fence *amdgpu_device_enforce_isolation(struct amdgpu_device *adev, dma_fence_put(isolation->spearhead); isolation->spearhead = dma_fence_get(&f->scheduled); amdgpu_sync_move(&isolation->active, &isolation->prev); + trace_amdgpu_isolation(isolation->owner, owner); isolation->owner = owner; } -- cgit v1.2.3 From 11bb33766f66d0cfe997601d6d974d0158a0e7b8 Mon Sep 17 00:00:00 2001 From: Ce Sun Date: Thu, 20 Mar 2025 18:12:40 +0800 Subject: drm/amdgpu: refactor amdgpu_device_gpu_recover Split amdgpu_device_gpu_recover into the following stages: halt activities,asic reset,schedule resume and amdgpu resume. The reason is that the subsequent addition of dpc recover code will have a high similarity with gpu reset Signed-off-by: Ce Sun Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 234 ++++++++++++++++++----------- 1 file changed, 144 insertions(+), 90 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 6ebf6179064b..1866d07db133 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -5930,94 +5930,40 @@ static int amdgpu_device_health_check(struct list_head *device_list_handle) return ret; } -/** - * amdgpu_device_gpu_recover - reset the asic and recover scheduler - * - * @adev: amdgpu_device pointer - * @job: which job trigger hang - * @reset_context: amdgpu reset context pointer - * - * Attempt to reset the GPU if it has hung (all asics). - * Attempt to do soft-reset or full-reset and reinitialize Asic - * Returns 0 for success or an error on failure. - */ - -int amdgpu_device_gpu_recover(struct amdgpu_device *adev, +static int amdgpu_device_halt_activities(struct amdgpu_device *adev, struct amdgpu_job *job, - struct amdgpu_reset_context *reset_context) + struct amdgpu_reset_context *reset_context, + struct list_head *device_list, + struct amdgpu_hive_info *hive, + bool need_emergency_restart) { - struct list_head device_list, *device_list_handle = NULL; - bool job_signaled = false; - struct amdgpu_hive_info *hive = NULL; + struct list_head *device_list_handle = NULL; struct amdgpu_device *tmp_adev = NULL; int i, r = 0; - bool need_emergency_restart = false; - bool audio_suspended = false; - int retry_limit = AMDGPU_MAX_RETRY_LIMIT; - /* - * If it reaches here because of hang/timeout and a RAS error is - * detected at the same time, let RAS recovery take care of it. - */ - if (amdgpu_ras_is_err_state(adev, AMDGPU_RAS_BLOCK__ANY) && - !amdgpu_sriov_vf(adev) && - reset_context->src != AMDGPU_RESET_SRC_RAS) { - dev_dbg(adev->dev, - "Gpu recovery from source: %d yielding to RAS error recovery handling", - reset_context->src); - return 0; - } - /* - * Special case: RAS triggered and full reset isn't supported - */ - need_emergency_restart = amdgpu_ras_need_emergency_restart(adev); - - /* - * Flush RAM to disk so that after reboot - * the user can read log and see why the system rebooted. - */ - if (need_emergency_restart && amdgpu_ras_get_context(adev) && - amdgpu_ras_get_context(adev)->reboot) { - DRM_WARN("Emergency reboot."); - - ksys_sync_helper(); - emergency_restart(); - } - - dev_info(adev->dev, "GPU %s begin!\n", - need_emergency_restart ? "jobs stop":"reset"); - - if (!amdgpu_sriov_vf(adev)) - hive = amdgpu_get_xgmi_hive(adev); - if (hive) - mutex_lock(&hive->hive_lock); - - 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 * to put adev in the 1st position. */ - INIT_LIST_HEAD(&device_list); if (!amdgpu_sriov_vf(adev) && (adev->gmc.xgmi.num_physical_nodes > 1) && hive) { list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) { - list_add_tail(&tmp_adev->reset_list, &device_list); + list_add_tail(&tmp_adev->reset_list, device_list); if (adev->shutdown) tmp_adev->shutdown = true; } - if (!list_is_first(&adev->reset_list, &device_list)) - list_rotate_to_front(&adev->reset_list, &device_list); - device_list_handle = &device_list; + if (!list_is_first(&adev->reset_list, device_list)) + list_rotate_to_front(&adev->reset_list, device_list); + device_list_handle = device_list; } else { - list_add_tail(&adev->reset_list, &device_list); - device_list_handle = &device_list; + list_add_tail(&adev->reset_list, device_list); + device_list_handle = device_list; } if (!amdgpu_sriov_vf(adev)) { r = amdgpu_device_health_check(device_list_handle); if (r) - goto end_reset; + return r; } /* We need to lock reset domain only once both for XGMI and single device */ @@ -6041,7 +5987,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, * some audio codec errors. */ if (!amdgpu_device_suspend_display_audio(tmp_adev)) - audio_suspended = true; + tmp_adev->pcie_reset_ctx.audio_suspended = true; amdgpu_ras_set_error_query_ready(tmp_adev, false); @@ -6076,23 +6022,19 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, atomic_inc(&tmp_adev->gpu_reset_counter); } - if (need_emergency_restart) - goto skip_sched_resume; + return r; +} - /* - * Must check guilty signal here since after this point all old - * HW fences are force signaled. - * - * job->base holds a reference to parent fence - */ - if (job && dma_fence_is_signaled(&job->hw_fence)) { - job_signaled = true; - dev_info(adev->dev, "Guilty job already signaled, skipping HW reset"); - goto skip_hw_reset; - } +static int amdgpu_device_asic_reset(struct amdgpu_device *adev, + struct list_head *device_list, + struct amdgpu_reset_context *reset_context) +{ + struct amdgpu_device *tmp_adev = NULL; + int retry_limit = AMDGPU_MAX_RETRY_LIMIT; + int r = 0; retry: /* Rest of adevs pre asic reset from XGMI hive. */ - list_for_each_entry(tmp_adev, device_list_handle, reset_list) { + list_for_each_entry(tmp_adev, device_list, reset_list) { r = amdgpu_device_pre_asic_reset(tmp_adev, reset_context); /*TODO Should we stop ?*/ if (r) { @@ -6119,12 +6061,12 @@ retry: /* Rest of adevs pre asic reset from XGMI hive. */ if (r) adev->asic_reset_res = r; } else { - r = amdgpu_do_asic_reset(device_list_handle, reset_context); + r = amdgpu_do_asic_reset(device_list, reset_context); if (r && r == -EAGAIN) goto retry; } - list_for_each_entry(tmp_adev, device_list_handle, reset_list) { + list_for_each_entry(tmp_adev, device_list, reset_list) { /* * Drop any pending non scheduler resets queued before reset is done. * Any reset scheduled after this point would be valid. Scheduler resets @@ -6134,10 +6076,18 @@ retry: /* Rest of adevs pre asic reset from XGMI hive. */ amdgpu_device_stop_pending_resets(tmp_adev); } -skip_hw_reset: + return r; +} + +static int amdgpu_device_sched_resume(struct list_head *device_list, + struct amdgpu_reset_context *reset_context, + bool job_signaled) +{ + struct amdgpu_device *tmp_adev = NULL; + int i, r = 0; /* Post ASIC reset for all devs .*/ - list_for_each_entry(tmp_adev, device_list_handle, reset_list) { + list_for_each_entry(tmp_adev, device_list, reset_list) { for (i = 0; i < AMDGPU_MAX_RINGS; ++i) { struct amdgpu_ring *ring = tmp_adev->rings[i]; @@ -6173,8 +6123,16 @@ skip_hw_reset: } } -skip_sched_resume: - list_for_each_entry(tmp_adev, device_list_handle, reset_list) { + return r; +} + +static void amdgpu_device_gpu_resume(struct amdgpu_device *adev, + struct list_head *device_list, + bool need_emergency_restart) +{ + struct amdgpu_device *tmp_adev = NULL; + + list_for_each_entry(tmp_adev, device_list, reset_list) { /* unlock kfd: SRIOV would do it separately */ if (!need_emergency_restart && !amdgpu_sriov_vf(tmp_adev)) amdgpu_amdkfd_post_reset(tmp_adev); @@ -6185,18 +6143,114 @@ skip_sched_resume: if (!adev->kfd.init_complete) amdgpu_amdkfd_device_init(adev); - if (audio_suspended) + if (tmp_adev->pcie_reset_ctx.audio_suspended) amdgpu_device_resume_display_audio(tmp_adev); amdgpu_device_unset_mp1_state(tmp_adev); amdgpu_ras_set_error_query_ready(tmp_adev, true); + } - tmp_adev = list_first_entry(device_list_handle, struct amdgpu_device, + tmp_adev = list_first_entry(device_list, struct amdgpu_device, reset_list); amdgpu_device_unlock_reset_domain(tmp_adev->reset_domain); +} + + +/** + * amdgpu_device_gpu_recover - reset the asic and recover scheduler + * + * @adev: amdgpu_device pointer + * @job: which job trigger hang + * @reset_context: amdgpu reset context pointer + * + * Attempt to reset the GPU if it has hung (all asics). + * Attempt to do soft-reset or full-reset and reinitialize Asic + * Returns 0 for success or an error on failure. + */ + +int amdgpu_device_gpu_recover(struct amdgpu_device *adev, + struct amdgpu_job *job, + struct amdgpu_reset_context *reset_context) +{ + struct list_head device_list; + bool job_signaled = false; + struct amdgpu_hive_info *hive = NULL; + int r = 0; + bool need_emergency_restart = false; + + /* + * If it reaches here because of hang/timeout and a RAS error is + * detected at the same time, let RAS recovery take care of it. + */ + if (amdgpu_ras_is_err_state(adev, AMDGPU_RAS_BLOCK__ANY) && + !amdgpu_sriov_vf(adev) && + reset_context->src != AMDGPU_RESET_SRC_RAS) { + dev_dbg(adev->dev, + "Gpu recovery from source: %d yielding to RAS error recovery handling", + reset_context->src); + return 0; + } + + /* + * Special case: RAS triggered and full reset isn't supported + */ + need_emergency_restart = amdgpu_ras_need_emergency_restart(adev); + + /* + * Flush RAM to disk so that after reboot + * the user can read log and see why the system rebooted. + */ + if (need_emergency_restart && amdgpu_ras_get_context(adev) && + amdgpu_ras_get_context(adev)->reboot) { + DRM_WARN("Emergency reboot."); + + ksys_sync_helper(); + emergency_restart(); + } + + dev_info(adev->dev, "GPU %s begin!\n", + need_emergency_restart ? "jobs stop":"reset"); + + if (!amdgpu_sriov_vf(adev)) + hive = amdgpu_get_xgmi_hive(adev); + if (hive) + mutex_lock(&hive->hive_lock); + + reset_context->job = job; + reset_context->hive = hive; + INIT_LIST_HEAD(&device_list); + + r = amdgpu_device_halt_activities(adev, job, reset_context, &device_list, + hive, need_emergency_restart); + if (r) + goto end_reset; + + if (need_emergency_restart) + goto skip_sched_resume; + /* + * Must check guilty signal here since after this point all old + * HW fences are force signaled. + * + * job->base holds a reference to parent fence + */ + if (job && dma_fence_is_signaled(&job->hw_fence)) { + job_signaled = true; + dev_info(adev->dev, "Guilty job already signaled, skipping HW reset"); + goto skip_hw_reset; + } + + r = amdgpu_device_asic_reset(adev, &device_list, reset_context); + if (r) + goto end_reset; +skip_hw_reset: + r = amdgpu_device_sched_resume(&device_list, reset_context, job_signaled); + if (r) + goto end_reset; +skip_sched_resume: + amdgpu_device_gpu_resume(adev, &device_list, need_emergency_restart); end_reset: if (hive) { mutex_unlock(&hive->hive_lock); -- cgit v1.2.3 From 8ba904f54148d2a093650b13a229d2ff18142f7d Mon Sep 17 00:00:00 2001 From: Ce Sun Date: Fri, 21 Mar 2025 10:11:18 +0800 Subject: drm/amdgpu: Multi-GPU DPC recovery support Add support for DPC recover based on refactored code Signed-off-by: Ce Sun Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 172 +++++++++++++++++++---------- 1 file changed, 115 insertions(+), 57 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 1866d07db133..7525128f971f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -3172,6 +3172,7 @@ static bool amdgpu_device_check_vram_lost(struct amdgpu_device *adev) * always assumed to be lost. */ switch (amdgpu_asic_reset_method(adev)) { + case AMD_RESET_METHOD_LINK: case AMD_RESET_METHOD_BACO: case AMD_RESET_METHOD_MODE1: return true; @@ -5510,6 +5511,29 @@ mode1_reset_failed: return ret; } +int amdgpu_device_link_reset(struct amdgpu_device *adev) +{ + int ret = 0; + + dev_info(adev->dev, "GPU link reset\n"); + + if (!adev->pcie_reset_ctx.occurs_dpc) + ret = amdgpu_dpm_link_reset(adev); + + if (ret) + goto link_reset_failed; + + ret = amdgpu_psp_wait_for_bootloader(adev); + if (ret) + goto link_reset_failed; + + return 0; + +link_reset_failed: + dev_err(adev->dev, "GPU link reset failed\n"); + return ret; +} + int amdgpu_device_pre_asic_reset(struct amdgpu_device *adev, struct amdgpu_reset_context *reset_context) { @@ -5814,6 +5838,7 @@ static void amdgpu_device_set_mp1_state(struct amdgpu_device *adev) switch (amdgpu_asic_reset_method(adev)) { case AMD_RESET_METHOD_MODE1: + case AMD_RESET_METHOD_LINK: adev->mp1_state = PP_MP1_STATE_SHUTDOWN; break; case AMD_RESET_METHOD_MODE2: @@ -5951,6 +5976,8 @@ static int amdgpu_device_halt_activities(struct amdgpu_device *adev, list_add_tail(&tmp_adev->reset_list, device_list); if (adev->shutdown) tmp_adev->shutdown = true; + if (adev->pcie_reset_ctx.occurs_dpc) + tmp_adev->pcie_reset_ctx.in_link_reset = true; } if (!list_is_first(&adev->reset_list, device_list)) list_rotate_to_front(&adev->reset_list, device_list); @@ -5960,7 +5987,7 @@ static int amdgpu_device_halt_activities(struct amdgpu_device *adev, device_list_handle = device_list; } - if (!amdgpu_sriov_vf(adev)) { + if (!amdgpu_sriov_vf(adev) && (!adev->pcie_reset_ctx.occurs_dpc)) { r = amdgpu_device_health_check(device_list_handle); if (r) return r; @@ -6005,6 +6032,7 @@ static int amdgpu_device_halt_activities(struct amdgpu_device *adev, /* disable ras on ALL IPs */ if (!need_emergency_restart && + (!adev->pcie_reset_ctx.occurs_dpc) && amdgpu_device_ip_need_full_reset(tmp_adev)) amdgpu_ras_suspend(tmp_adev); @@ -6035,7 +6063,11 @@ static int amdgpu_device_asic_reset(struct amdgpu_device *adev, retry: /* Rest of adevs pre asic reset from XGMI hive. */ list_for_each_entry(tmp_adev, device_list, reset_list) { + if (adev->pcie_reset_ctx.occurs_dpc) + tmp_adev->no_hw_access = true; r = amdgpu_device_pre_asic_reset(tmp_adev, reset_context); + if (adev->pcie_reset_ctx.occurs_dpc) + tmp_adev->no_hw_access = false; /*TODO Should we stop ?*/ if (r) { dev_err(tmp_adev->dev, "GPU pre asic reset failed with err, %d for drm dev, %s ", @@ -6634,12 +6666,15 @@ 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; + struct amdgpu_hive_info *hive = amdgpu_get_xgmi_hive(adev); + struct amdgpu_reset_context reset_context; + struct list_head device_list; + int r = 0; - DRM_INFO("PCI error: detected callback, state(%d)!!\n", state); + dev_info(adev->dev, "PCI error: detected callback!!\n"); - if (adev->gmc.xgmi.num_physical_nodes > 1) { - DRM_WARN("No support for XGMI hive yet..."); + if (!amdgpu_dpm_is_link_reset_supported(adev)) { + dev_warn(adev->dev, "No support for XGMI hive yet...\n"); return PCI_ERS_RESULT_DISCONNECT; } @@ -6647,32 +6682,30 @@ pci_ers_result_t amdgpu_pci_error_detected(struct pci_dev *pdev, pci_channel_sta switch (state) { case pci_channel_io_normal: + dev_info(adev->dev, "pci_channel_io_normal: state(%d)!!\n", state); return PCI_ERS_RESULT_CAN_RECOVER; - /* Fatal error, prepare for slot reset */ case pci_channel_io_frozen: - /* - * Locking adev->reset_domain->sem will prevent any external access - * to GPU during PCI error recovery - */ - amdgpu_device_lock_reset_domain(adev->reset_domain); - amdgpu_device_set_mp1_state(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 (!amdgpu_ring_sched_ready(ring)) - continue; - - drm_sched_stop(&ring->sched, NULL); + /* Fatal error, prepare for slot reset */ + dev_info(adev->dev, "pci_channel_io_frozen: state(%d)!!\n", state); + + if (hive) + mutex_lock(&hive->hive_lock); + adev->pcie_reset_ctx.occurs_dpc = true; + memset(&reset_context, 0, sizeof(reset_context)); + INIT_LIST_HEAD(&device_list); + + r = amdgpu_device_halt_activities(adev, NULL, &reset_context, &device_list, + hive, false); + if (hive) { + mutex_unlock(&hive->hive_lock); + amdgpu_put_xgmi_hive(hive); } - atomic_inc(&adev->gpu_reset_counter); + if (r) + return PCI_ERS_RESULT_DISCONNECT; return PCI_ERS_RESULT_NEED_RESET; case pci_channel_io_perm_failure: /* Permanent error, prepare for device removal */ + dev_info(adev->dev, "pci_channel_io_perm_failure: state(%d)!!\n", state); return PCI_ERS_RESULT_DISCONNECT; } @@ -6685,8 +6718,10 @@ pci_ers_result_t amdgpu_pci_error_detected(struct pci_dev *pdev, pci_channel_sta */ pci_ers_result_t amdgpu_pci_mmio_enabled(struct pci_dev *pdev) { + struct drm_device *dev = pci_get_drvdata(pdev); + struct amdgpu_device *adev = drm_to_adev(dev); - DRM_INFO("PCI error: mmio enabled callback!!\n"); + dev_info(adev->dev, "PCI error: mmio enabled callback!!\n"); /* TODO - dump whatever for debugging purposes */ @@ -6710,10 +6745,12 @@ 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; struct amdgpu_reset_context reset_context; - u32 memsize; + struct amdgpu_device *tmp_adev = NULL; + struct amdgpu_hive_info *hive = NULL; struct list_head device_list; + int r = 0, i; + u32 memsize; /* PCI error slot reset should be skipped During RAS recovery */ if ((amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 3) || @@ -6721,15 +6758,12 @@ pci_ers_result_t amdgpu_pci_slot_reset(struct pci_dev *pdev) amdgpu_ras_in_recovery(adev)) return PCI_ERS_RESULT_RECOVERED; - DRM_INFO("PCI error: slot reset callback!!\n"); + dev_info(adev->dev, "PCI error: slot reset callback!!\n"); memset(&reset_context, 0, sizeof(reset_context)); - INIT_LIST_HEAD(&device_list); - list_add_tail(&adev->reset_list, &device_list); - /* wait for asic to come out of reset */ - msleep(500); + msleep(700); /* Restore PCI confspace */ amdgpu_device_load_pci_state(pdev); @@ -6750,26 +6784,40 @@ pci_ers_result_t amdgpu_pci_slot_reset(struct pci_dev *pdev) reset_context.method = AMD_RESET_METHOD_NONE; reset_context.reset_req_dev = adev; set_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags); - set_bit(AMDGPU_SKIP_HW_RESET, &reset_context.flags); - - adev->no_hw_access = true; - r = amdgpu_device_pre_asic_reset(adev, &reset_context); - adev->no_hw_access = false; - if (r) - goto out; + set_bit(AMDGPU_SKIP_COREDUMP, &reset_context.flags); + INIT_LIST_HEAD(&device_list); - r = amdgpu_do_asic_reset(&device_list, &reset_context); + hive = amdgpu_get_xgmi_hive(adev); + if (hive) { + mutex_lock(&hive->hive_lock); + reset_context.hive = hive; + list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) { + tmp_adev->pcie_reset_ctx.in_link_reset = true; + list_add_tail(&tmp_adev->reset_list, &device_list); + } + } else { + set_bit(AMDGPU_SKIP_HW_RESET, &reset_context.flags); + list_add_tail(&adev->reset_list, &device_list); + } + r = amdgpu_device_asic_reset(adev, &device_list, &reset_context); out: if (!r) { if (amdgpu_device_cache_pci_state(adev->pdev)) pci_restore_state(adev->pdev); - - DRM_INFO("PCIe error recovery succeeded\n"); + dev_info(adev->dev, "PCIe error recovery succeeded\n"); } else { - DRM_ERROR("PCIe error recovery failed, err:%d", r); - amdgpu_device_unset_mp1_state(adev); - amdgpu_device_unlock_reset_domain(adev->reset_domain); + dev_err(adev->dev, "PCIe error recovery failed, err:%d\n", r); + if (tmp_adev) { + list_for_each_entry(tmp_adev, &device_list, reset_list) + amdgpu_device_unset_mp1_state(tmp_adev); + amdgpu_device_unlock_reset_domain(adev->reset_domain); + } + } + + if (hive) { + mutex_unlock(&hive->hive_lock); + amdgpu_put_xgmi_hive(hive); } return r ? PCI_ERS_RESULT_DISCONNECT : PCI_ERS_RESULT_RECOVERED; @@ -6786,26 +6834,36 @@ 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; - + struct list_head device_list; + struct amdgpu_hive_info *hive = NULL; + struct amdgpu_device *tmp_adev = NULL; - DRM_INFO("PCI error: resume callback!!\n"); + dev_info(adev->dev, "PCI error: resume callback!!\n"); /* Only continue execution for the case of pci_channel_io_frozen */ if (adev->pci_channel_state != pci_channel_io_frozen) return; - for (i = 0; i < AMDGPU_MAX_RINGS; ++i) { - struct amdgpu_ring *ring = adev->rings[i]; + INIT_LIST_HEAD(&device_list); - if (!amdgpu_ring_sched_ready(ring)) - continue; + hive = amdgpu_get_xgmi_hive(adev); + if (hive) { + mutex_lock(&hive->hive_lock); + list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) { + tmp_adev->pcie_reset_ctx.in_link_reset = false; + list_add_tail(&tmp_adev->reset_list, &device_list); + } + } else + list_add_tail(&adev->reset_list, &device_list); - drm_sched_start(&ring->sched, 0); - } + amdgpu_device_sched_resume(&device_list, NULL, NULL); + amdgpu_device_gpu_resume(adev, &device_list, false); + adev->pcie_reset_ctx.occurs_dpc = false; - amdgpu_device_unset_mp1_state(adev); - amdgpu_device_unlock_reset_domain(adev->reset_domain); + if (hive) { + mutex_unlock(&hive->hive_lock); + amdgpu_put_xgmi_hive(hive); + } } bool amdgpu_device_cache_pci_state(struct pci_dev *pdev) -- cgit v1.2.3 From 48b733d99b0d8db01fcf55d7ae31f69510fb1a4a Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Thu, 27 Feb 2025 12:31:28 -0500 Subject: drm/amdgpu: add rebar parameter Add a new parameter to disable BAR resizing. Note that this only disables the driver from attempting to resize the BAR, The BIOS may have resized the BAR at boot. Some teams have found this useful in debugging P2P DMA issues on systems where the available MMIO space did not allow for all of the GPUs present to resize their BARs. Reviewed-by: Mario Limonciello Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 7525128f971f..f4a01704effc 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -1680,6 +1680,9 @@ int amdgpu_device_resize_fb_bar(struct amdgpu_device *adev) if (amdgpu_sriov_vf(adev)) return 0; + if (!amdgpu_rebar) + return 0; + /* resizing on Dell G5 SE platforms causes problems with runtime pm */ if ((amdgpu_runtime_pm != 0) && adev->pdev->vendor == PCI_VENDOR_ID_ATI && -- cgit v1.2.3 From daafa303d19f5522e4c24fbf5c1c981a16df2c2f Mon Sep 17 00:00:00 2001 From: Apurv Mishra Date: Mon, 17 Mar 2025 14:00:38 -0400 Subject: drm/amdkfd: Drop workaround for GC v9.4.3 revID 0 Remove workaround code for the early engineering samples GC v9.4.3 SOCs with revID 0 Reviewed-by: Amber Lin Signed-off-by: Apurv Mishra Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index f4a01704effc..1878bbacf0a2 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -2692,6 +2692,13 @@ static int amdgpu_device_ip_early_init(struct amdgpu_device *adev) break; } + /* Check for IP version 9.4.3 with A0 hardware */ + if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 3) && + !amdgpu_device_get_rev_id(adev)) { + dev_err(adev->dev, "Unsupported A0 hardware\n"); + return -ENODEV; /* device unsupported - no device error */ + } + if (amdgpu_has_atpx() && (amdgpu_is_atpx_hybrid() || amdgpu_has_atpx_dgpu_power_cntl()) && @@ -2704,7 +2711,6 @@ static int amdgpu_device_ip_early_init(struct amdgpu_device *adev) adev->has_pr3 = parent ? pci_pr3_present(parent) : false; } - adev->pm.pp_feature = amdgpu_pp_feature_mask; if (amdgpu_sriov_vf(adev) || sched_policy == KFD_SCHED_POLICY_NO_HWS) adev->pm.pp_feature &= ~PP_GFXOFF_MASK; -- cgit v1.2.3 From 906ad451675155380c1dc1881a244ebde8e8df0a Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Sun, 6 Apr 2025 17:27:24 -0400 Subject: drm/amdgpu: cancel gfx idle work in device suspend for s0ix This is normally handled in the gfx IP suspend callbacks, but for S0ix, those are skipped because we don't want to touch gfx. So handle it in device suspend. Fixes: b9467983b774 ("drm/amdgpu: add dynamic workload profile switching for gfx10") Fixes: 963537ca2325 ("drm/amdgpu: add dynamic workload profile switching for gfx11") Fixes: 5f95a1549555 ("drm/amdgpu: add dynamic workload profile switching for gfx12") Reviewed-by: Mario Limonciello Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 1878bbacf0a2..979942f26dda 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -3653,6 +3653,13 @@ static int amdgpu_device_ip_suspend_phase2(struct amdgpu_device *adev) adev, adev->ip_blocks[i].version->type)) continue; + /* Since we skip suspend for S0i3, we need to cancel the delayed + * idle work here as the suspend callback never gets called. + */ + if (adev->in_s0ix && + adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GFX && + amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(10, 0, 0)) + cancel_delayed_work_sync(&adev->gfx.idle_work); /* skip suspend of gfx/mes and psp for S0ix * gfx is in gfxoff state, so on resume it will exit gfxoff just * like at runtime. PSP is also part of the always on hardware -- cgit v1.2.3 From bb00bf17328d80f519df93bebc41f6c9171547d6 Mon Sep 17 00:00:00 2001 From: Kenneth Feng Date: Tue, 1 Apr 2025 15:56:41 +0800 Subject: drm/amd/amdgpu: decouple ASPM with pcie dpm ASPM doesn't need to be disabled if pcie dpm is disabled. So ASPM can be independantly enabled. Signed-off-by: Kenneth Feng Reviewed-by: Yang Wang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 979942f26dda..cfaa5d77b20a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -1897,8 +1897,6 @@ bool amdgpu_device_should_use_aspm(struct amdgpu_device *adev) } if (adev->flags & AMD_IS_APU) return false; - if (!(adev->pm.pp_feature & PP_PCIE_DPM_MASK)) - return false; return pcie_aspm_enabled(adev->pdev); } -- cgit v1.2.3 From 8949843762631d9d0fc526dfb61a272dca29fc6f Mon Sep 17 00:00:00 2001 From: Arunpravin Paneer Selvam Date: Wed, 30 Oct 2024 10:38:46 +0530 Subject: drm/amdgpu: Enable userq fence interrupt support MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Add support to handle the userqueue protected fence signal hardware interrupt. Create a xarray which maps the doorbell index to the fence driver address. This would help to retrieve the fence driver information when an userq fence interrupt is triggered. Firmware sends the doorbell offset value and this info is compared with the queue's mqd doorbell offset value. If they are same, we process the userq fence interrupt. v1:(Christian): - use xa_load to extract the fence driver. - move the amdgpu_userq_fence_driver_process call within the xa_lock as there is a chance that fence_drv might be freed. Signed-off-by: Arunpravin Paneer Selvam Acked-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index cfaa5d77b20a..fdd271edb0b9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -4336,6 +4336,8 @@ int amdgpu_device_init(struct amdgpu_device *adev, spin_lock_init(&adev->virt.rlcg_reg_lock); spin_lock_init(&adev->wb.lock); + xa_init_flags(&adev->userq_xa, XA_FLAGS_LOCK_IRQ); + INIT_LIST_HEAD(&adev->reset_list); INIT_LIST_HEAD(&adev->ras_list); -- cgit v1.2.3 From ac4a1f7f13309f8235a0c4719c34094de3303dfd Mon Sep 17 00:00:00 2001 From: Arunpravin Paneer Selvam Date: Wed, 30 Oct 2024 10:46:49 +0530 Subject: drm/amdgpu: Remove the MES self test MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Remove MES self test as this conflicts the userqueue fence interrupts. v2:(Christian) - remove the amdgpu_mes_self_test() function and any now unused code. Signed-off-by: Arunpravin Paneer Selvam Acked-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 3 --- 1 file changed, 3 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index fdd271edb0b9..4da18ed0beea 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -5144,9 +5144,6 @@ exit: } adev->in_suspend = false; - if (adev->enable_mes) - amdgpu_mes_self_test(adev); - if (amdgpu_acpi_smart_shift_update(dev, AMDGPU_SS_DEV_D0)) DRM_WARN("smart shift update failed\n"); -- cgit v1.2.3 From 4ce60dbada9639e385f2f4be2dacf10d8ba22378 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Thu, 20 Feb 2025 15:56:24 -0500 Subject: drm/amdgpu: store userq_managers in a list in adev So we can iterate across them when we need to manage all user queues. v2: add uq_mgr to adev list in amdgpu_userq_mgr_init Reviewed-by: Arunpravin Paneer Selvam Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 4da18ed0beea..6b38b0511330 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -4317,6 +4317,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, mutex_init(&adev->gfx.kfd_sch_mutex); mutex_init(&adev->gfx.workload_profile_mutex); mutex_init(&adev->vcn.workload_profile_mutex); + mutex_init(&adev->userq_mutex); amdgpu_device_init_apu_flags(adev); @@ -4344,6 +4345,8 @@ int amdgpu_device_init(struct amdgpu_device *adev, INIT_LIST_HEAD(&adev->pm.od_kobj_list); + INIT_LIST_HEAD(&adev->userq_mgr_list); + INIT_DELAYED_WORK(&adev->delayed_init_work, amdgpu_device_delayed_init_work_handler); INIT_DELAYED_WORK(&adev->gfx.gfx_off_delay_work, -- cgit v1.2.3 From 35a5440832b2d6a46841b6bb68855e2622833401 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Sun, 6 Apr 2025 17:27:24 -0400 Subject: drm/amdgpu: cancel gfx idle work in device suspend for s0ix This is normally handled in the gfx IP suspend callbacks, but for S0ix, those are skipped because we don't want to touch gfx. So handle it in device suspend. Fixes: b9467983b774 ("drm/amdgpu: add dynamic workload profile switching for gfx10") Fixes: 963537ca2325 ("drm/amdgpu: add dynamic workload profile switching for gfx11") Fixes: 5f95a1549555 ("drm/amdgpu: add dynamic workload profile switching for gfx12") Reviewed-by: Mario Limonciello Signed-off-by: Alex Deucher (cherry picked from commit 906ad451675155380c1dc1881a244ebde8e8df0a) Cc: stable@vger.kernel.org --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 6ebf6179064b..5a625bdb1a88 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -3643,6 +3643,13 @@ static int amdgpu_device_ip_suspend_phase2(struct amdgpu_device *adev) adev, adev->ip_blocks[i].version->type)) continue; + /* Since we skip suspend for S0i3, we need to cancel the delayed + * idle work here as the suspend callback never gets called. + */ + if (adev->in_s0ix && + adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GFX && + amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(10, 0, 0)) + cancel_delayed_work_sync(&adev->gfx.idle_work); /* skip suspend of gfx/mes and psp for S0ix * gfx is in gfxoff state, so on resume it will exit gfxoff just * like at runtime. PSP is also part of the always on hardware -- cgit v1.2.3 From c770ef19673fb1defcbde2ee2b91c3c89bfcf164 Mon Sep 17 00:00:00 2001 From: Kenneth Feng Date: Tue, 1 Apr 2025 16:04:41 +0800 Subject: drm/amd/amdgpu: disable ASPM in some situations disable ASPM with some ASICs on some specific platforms. required from PCIe controller owner. Signed-off-by: Kenneth Feng Reviewed-by: Yang Wang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 32 ++++++++++++++++++++++++++++++ 1 file changed, 32 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 6b38b0511330..84ddd02579bb 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -85,6 +85,7 @@ #if IS_ENABLED(CONFIG_X86) #include +#include #endif MODULE_FIRMWARE("amdgpu/vega10_gpu_info.bin"); @@ -1873,6 +1874,35 @@ static bool amdgpu_device_pcie_dynamic_switching_supported(struct amdgpu_device return true; } +static bool amdgpu_device_aspm_support_quirk(struct amdgpu_device *adev) +{ +#if IS_ENABLED(CONFIG_X86) + struct cpuinfo_x86 *c = &cpu_data(0); + + if (!(amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(12, 0, 0) || + amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(12, 0, 1))) + return false; + + if (c->x86 == 6 && + adev->pm.pcie_gen_mask & CAIL_PCIE_LINK_SPEED_SUPPORT_GEN5) { + switch (c->x86_model) { + case VFM_MODEL(INTEL_ALDERLAKE): + case VFM_MODEL(INTEL_ALDERLAKE_L): + case VFM_MODEL(INTEL_RAPTORLAKE): + case VFM_MODEL(INTEL_RAPTORLAKE_P): + case VFM_MODEL(INTEL_RAPTORLAKE_S): + return true; + default: + return false; + } + } else { + return false; + } +#else + return false; +#endif +} + /** * amdgpu_device_should_use_aspm - check if the device should program ASPM * @@ -1897,6 +1927,8 @@ bool amdgpu_device_should_use_aspm(struct amdgpu_device *adev) } if (adev->flags & AMD_IS_APU) return false; + if (amdgpu_device_aspm_support_quirk(adev)) + return false; return pcie_aspm_enabled(adev->pdev); } -- cgit v1.2.3 From 39938a8ed979e398faa3791a47e282c82bcc6f04 Mon Sep 17 00:00:00 2001 From: ZhenGuo Yin Date: Tue, 8 Apr 2025 16:18:28 +0800 Subject: drm/amdgpu: fix warning of drm_mm_clean MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Kernel doorbell BOs needs to be freed before ttm_fini. Fixes: 54c30d2a8def ("drm/amdgpu: create kernel doorbell pages") Acked-by: Alex Deucher Reviewed-by: Christian König Signed-off-by: ZhenGuo Yin Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 84ddd02579bb..66ed6a7924ef 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -3550,6 +3550,7 @@ static int amdgpu_device_ip_fini(struct amdgpu_device *adev) amdgpu_device_mem_scratch_fini(adev); amdgpu_ib_pool_fini(adev); amdgpu_seq64_fini(adev); + amdgpu_doorbell_fini(adev); } if (adev->ip_blocks[i].version->funcs->sw_fini) { r = adev->ip_blocks[i].version->funcs->sw_fini(&adev->ip_blocks[i]); @@ -4903,7 +4904,6 @@ void amdgpu_device_fini_sw(struct amdgpu_device *adev) iounmap(adev->rmmio); adev->rmmio = NULL; - amdgpu_doorbell_fini(adev); drm_dev_exit(idx); } -- cgit v1.2.3 From 732c6cefc1ecfc8de5d7a2029480798655d979d8 Mon Sep 17 00:00:00 2001 From: Ce Sun Date: Wed, 9 Apr 2025 19:53:11 +0800 Subject: drm/amdgpu: Replace tmp_adev with hive in amdgpu_pci_slot_reset MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Checking hive is more readable. The following smatch warning: drivers/gpu/drm/amd/amdgpu/amdgpu_device.c:6820 amdgpu_pci_slot_reset() warn: iterator used outside loop: 'tmp_adev' Reported-by: Dan Carpenter Signed-off-by: Ce Sun Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 66ed6a7924ef..8e150e9393c7 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -6794,8 +6794,8 @@ 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); struct amdgpu_reset_context reset_context; - struct amdgpu_device *tmp_adev = NULL; - struct amdgpu_hive_info *hive = NULL; + struct amdgpu_device *tmp_adev; + struct amdgpu_hive_info *hive; struct list_head device_list; int r = 0, i; u32 memsize; @@ -6856,7 +6856,7 @@ out: dev_info(adev->dev, "PCIe error recovery succeeded\n"); } else { dev_err(adev->dev, "PCIe error recovery failed, err:%d\n", r); - if (tmp_adev) { + if (hive) { list_for_each_entry(tmp_adev, &device_list, reset_list) amdgpu_device_unset_mp1_state(tmp_adev); amdgpu_device_unlock_reset_domain(adev->reset_domain); -- cgit v1.2.3 From 2e0454b730648e9349ca54eb4a8142d77e8e7008 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Tue, 8 Apr 2025 10:39:06 -0400 Subject: drm/amdgpu: adjust enforce_isolation handling Switch from a bool to an enum and allow more options for enforce isolation. There are now 3 modes of operation: - Disabled (0) - Enabled (serialization and cleaner shader) (1) - Enabled in legacy mode (no serialization or cleaner shader) (2) This provides better flexibility for more use cases. Acked-by: Srinivasan Shanmugam Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 22 ++++++++++++++++++++-- 1 file changed, 20 insertions(+), 2 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 8e150e9393c7..475bcd2a8a31 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -2145,8 +2145,26 @@ static int amdgpu_device_check_arguments(struct amdgpu_device *adev) adev->firmware.load_type = amdgpu_ucode_get_load_type(adev, amdgpu_fw_load_type); - for (i = 0; i < MAX_XCP; i++) - adev->enforce_isolation[i] = !!enforce_isolation; + for (i = 0; i < MAX_XCP; i++) { + switch (amdgpu_enforce_isolation) { + case -1: + case 0: + default: + /* disable */ + adev->enforce_isolation[i] = AMDGPU_ENFORCE_ISOLATION_DISABLE; + break; + case 1: + /* enable */ + adev->enforce_isolation[i] = + AMDGPU_ENFORCE_ISOLATION_ENABLE; + break; + case 2: + /* enable legacy mode */ + adev->enforce_isolation[i] = + AMDGPU_ENFORCE_ISOLATION_ENABLE_LEGACY; + break; + } + } return 0; } -- cgit v1.2.3 From e7afa85a0d0eba5bf2c0a446ff622ebdbc9812d6 Mon Sep 17 00:00:00 2001 From: ZhenGuo Yin Date: Tue, 8 Apr 2025 16:18:28 +0800 Subject: drm/amdgpu: fix warning of drm_mm_clean MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Kernel doorbell BOs needs to be freed before ttm_fini. Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4145 Fixes: 54c30d2a8def ("drm/amdgpu: create kernel doorbell pages") Acked-by: Alex Deucher Reviewed-by: Christian König Signed-off-by: ZhenGuo Yin Signed-off-by: Alex Deucher (cherry picked from commit 39938a8ed979e398faa3791a47e282c82bcc6f04) Cc: stable@vger.kernel.org --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index b34b915203f2..7f354cd532dc 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -3510,6 +3510,7 @@ static int amdgpu_device_ip_fini(struct amdgpu_device *adev) amdgpu_device_mem_scratch_fini(adev); amdgpu_ib_pool_fini(adev); amdgpu_seq64_fini(adev); + amdgpu_doorbell_fini(adev); } if (adev->ip_blocks[i].version->funcs->sw_fini) { r = adev->ip_blocks[i].version->funcs->sw_fini(&adev->ip_blocks[i]); @@ -4858,7 +4859,6 @@ void amdgpu_device_fini_sw(struct amdgpu_device *adev) iounmap(adev->rmmio); adev->rmmio = NULL; - amdgpu_doorbell_fini(adev); drm_dev_exit(idx); } -- cgit v1.2.3 From c2c722217af4d2d17de8665968975a70a08046ea Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Fri, 21 Feb 2025 12:16:34 -0500 Subject: drm/amdgpu/userq: handle system suspend and resume Unmap user queues on suspend and map them on resume. Reviewed-by: Sunil Khatri Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 14 +++++++++++++- 1 file changed, 13 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 475bcd2a8a31..e966aefc2b0f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -3513,6 +3513,9 @@ static int amdgpu_device_ip_fini_early(struct amdgpu_device *adev) amdgpu_device_set_cg_state(adev, AMD_CG_STATE_UNGATE); amdgpu_amdkfd_suspend(adev, false); +#ifdef CONFIG_DRM_AMDGPU_NAVI3X_USERQ + amdgpu_userq_suspend(adev); +#endif /* Workaround for ASICs need to disable SMC first */ amdgpu_device_smu_fini_early(adev); @@ -5081,8 +5084,12 @@ int amdgpu_device_suspend(struct drm_device *dev, bool notify_clients) amdgpu_device_ip_suspend_phase1(adev); - if (!adev->in_s0ix) + if (!adev->in_s0ix) { amdgpu_amdkfd_suspend(adev, adev->in_runpm); +#ifdef CONFIG_DRM_AMDGPU_NAVI3X_USERQ + amdgpu_userq_suspend(adev); +#endif + } r = amdgpu_device_evict_resources(adev); if (r) @@ -5149,6 +5156,11 @@ int amdgpu_device_resume(struct drm_device *dev, bool notify_clients) r = amdgpu_amdkfd_resume(adev, adev->in_runpm); if (r) goto exit; +#ifdef CONFIG_DRM_AMDGPU_NAVI3X_USERQ + r = amdgpu_userq_resume(adev); + if (r) + goto exit; +#endif } r = amdgpu_device_ip_late_init(adev); -- cgit v1.2.3 From 28fc3172e4204c8cdd8c70226ea02b0ae9930b69 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Fri, 21 Feb 2025 15:20:45 -0500 Subject: drm/amdgpu: rename enforce isolation variables Since they will be used for both KFD and KGD user queues, rename them from kfd to userq. No intended functional change. Acked-by: Sunil Khatri Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index e966aefc2b0f..b96e0613ea7e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -4368,7 +4368,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, amdgpu_sync_create(&adev->isolation[i].active); amdgpu_sync_create(&adev->isolation[i].prev); } - mutex_init(&adev->gfx.kfd_sch_mutex); + mutex_init(&adev->gfx.userq_sch_mutex); mutex_init(&adev->gfx.workload_profile_mutex); mutex_init(&adev->vcn.workload_profile_mutex); mutex_init(&adev->userq_mutex); -- cgit v1.2.3 From 56801cb83c8c95ae23ea576570c75a01c1f07774 Mon Sep 17 00:00:00 2001 From: Arvind Yadav Date: Tue, 22 Apr 2025 19:29:03 +0530 Subject: drm/amdgpu: remove DRM_AMDGPU_NAVI3X_USERQ config for UQ MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit DRM_AMDGPU_NAVI3X_USERQ config support is not required for usermode queue. v2: rebase. Cc: Arunpravin Paneer Selvam Reviewed-by: Sunil Khatri Reviewed-by: Alex Deucher Reviewed-by: Christian König Signed-off-by: Arvind Yadav Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 7 +------ 1 file changed, 1 insertion(+), 6 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index b96e0613ea7e..fe68ba9997ae 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -3513,9 +3513,7 @@ static int amdgpu_device_ip_fini_early(struct amdgpu_device *adev) amdgpu_device_set_cg_state(adev, AMD_CG_STATE_UNGATE); amdgpu_amdkfd_suspend(adev, false); -#ifdef CONFIG_DRM_AMDGPU_NAVI3X_USERQ amdgpu_userq_suspend(adev); -#endif /* Workaround for ASICs need to disable SMC first */ amdgpu_device_smu_fini_early(adev); @@ -5086,9 +5084,7 @@ int amdgpu_device_suspend(struct drm_device *dev, bool notify_clients) if (!adev->in_s0ix) { amdgpu_amdkfd_suspend(adev, adev->in_runpm); -#ifdef CONFIG_DRM_AMDGPU_NAVI3X_USERQ amdgpu_userq_suspend(adev); -#endif } r = amdgpu_device_evict_resources(adev); @@ -5156,11 +5152,10 @@ int amdgpu_device_resume(struct drm_device *dev, bool notify_clients) r = amdgpu_amdkfd_resume(adev, adev->in_runpm); if (r) goto exit; -#ifdef CONFIG_DRM_AMDGPU_NAVI3X_USERQ + r = amdgpu_userq_resume(adev); if (r) goto exit; -#endif } r = amdgpu_device_ip_late_init(adev); -- cgit v1.2.3 From 68071eb0ae64c076ed93897be644ebbd1c89a278 Mon Sep 17 00:00:00 2001 From: Srinivasan Shanmugam Date: Mon, 28 Apr 2025 16:16:34 +0530 Subject: drm/amdgpu: Add Support for enforcing isolation without Cleaner Shader MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Adjusted the enforce isolation setting handling to include the ability to disable the cleaner shader without affecting isolation between tasks. v2: Updated enforce isolation documentation and parameters. (Alex) Cc: Christian König Cc: Alex Deucher Signed-off-by: Srinivasan Shanmugam 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/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index fe68ba9997ae..8330e30f0caf 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -2163,6 +2163,11 @@ static int amdgpu_device_check_arguments(struct amdgpu_device *adev) adev->enforce_isolation[i] = AMDGPU_ENFORCE_ISOLATION_ENABLE_LEGACY; break; + case 3: + /* enable only process isolation without submitting cleaner shader */ + adev->enforce_isolation[i] = + AMDGPU_ENFORCE_ISOLATION_NO_CLEANER_SHADER; + break; } } -- cgit v1.2.3 From ce8f7d95899c2869b47ea6ce0b3e5bf304b2fff4 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Thu, 1 May 2025 13:00:16 -0400 Subject: Revert "drm/amd: Stop evicting resources on APUs in suspend" This reverts commit 3a9626c816db901def438dc2513622e281186d39. This breaks S4 because we end up setting the s3/s0ix flags even when we are entering s4 since prepare is used by both flows. The causes both the S3/s0ix and s4 flags to be set which breaks several checks in the driver which assume they are mutually exclusive. Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/3634 Cc: Mario Limonciello Reviewed-by: Mario Limonciello Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 11 ++--------- 1 file changed, 2 insertions(+), 9 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 8330e30f0caf..ccc6217761bf 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -5018,15 +5018,13 @@ int amdgpu_device_prepare(struct drm_device *dev) struct amdgpu_device *adev = drm_to_adev(dev); int i, r; - amdgpu_choose_low_power_state(adev); - if (dev->switch_power_state == DRM_SWITCH_POWER_OFF) return 0; /* Evict the majority of BOs before starting suspend sequence */ r = amdgpu_device_evict_resources(adev); if (r) - goto unprepare; + return r; flush_delayed_work(&adev->gfx.gfx_off_delay_work); @@ -5037,15 +5035,10 @@ int amdgpu_device_prepare(struct drm_device *dev) continue; r = adev->ip_blocks[i].version->funcs->prepare_suspend(&adev->ip_blocks[i]); if (r) - goto unprepare; + return r; } return 0; - -unprepare: - adev->in_s0ix = adev->in_s3 = adev->in_s4 = false; - - return r; } /** -- cgit v1.2.3 From 086809c82c96116aed78f762e4f1d61a4e0210a7 Mon Sep 17 00:00:00 2001 From: Ellen Pan Date: Tue, 29 Apr 2025 17:18:44 -0400 Subject: drm/amdgpu: Implement unrecoverable error message handling for VFs This notification may arrive in VF mailbox while polling for response from another event. This patches covers the following scenarios: - If VF is already in RMA state, then do not attempt to contact the host. Host will ignore the VF after sending the notification. - If the notification is detected during polling, then set the RMA status, and return error to caller. - If the notification arrives by interrupt, then set the RMA status and queue a reset. This reset will fail and VF will stop runtime services. Reviewed-by: Shravan Kumar Gande Signed-off-by: Victor Skvortsov Signed-off-by: Ellen Pan Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index ccc6217761bf..a2997ededd0e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -6150,6 +6150,11 @@ retry: /* Rest of adevs pre asic reset from XGMI hive. */ /* Actual ASIC resets if needed.*/ /* Host driver will handle XGMI hive reset for SRIOV */ if (amdgpu_sriov_vf(adev)) { + + /* Bail out of reset early */ + if (amdgpu_ras_is_rma(adev)) + return -ENODEV; + if (amdgpu_ras_get_fed_status(adev) || amdgpu_virt_rcvd_ras_interrupt(adev)) { dev_dbg(adev->dev, "Detected RAS error, wait for FLR completion\n"); amdgpu_ras_set_fed(adev, true); -- cgit v1.2.3 From 06f2dcc241e7e5c681f81fbc46cacdf4bfd7d6d7 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Thu, 1 May 2025 13:46:46 -0400 Subject: drm/amdgpu: fix pm notifier handling Set the s3/s0ix and s4 flags in the pm notifier so that we can skip the resource evictions properly in pm prepare based on whether we are suspending or hibernating. Drop the eviction as processes are not frozen at this time, we we can end up getting stuck trying to evict VRAM while applications continue to submit work which causes the buffers to get pulled back into VRAM. v2: Move suspend flags out of pm notifier (Mario) Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4178 Fixes: 2965e6355dcd ("drm/amd: Add Suspend/Hibernate notification callback support") Cc: Mario Limonciello Reviewed-by: Mario Limonciello Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 18 +++++------------- 1 file changed, 5 insertions(+), 13 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_device.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index a2997ededd0e..5d47c36c8280 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -4976,28 +4976,20 @@ static int amdgpu_device_evict_resources(struct amdgpu_device *adev) * @data: data * * This function is called when the system is about to suspend or hibernate. - * It is used to evict resources from the device before the system goes to - * sleep while there is still access to swap. + * It is used to set the appropriate flags so that eviction can be optimized + * in the pm prepare callback. */ static int amdgpu_device_pm_notifier(struct notifier_block *nb, unsigned long mode, void *data) { struct amdgpu_device *adev = container_of(nb, struct amdgpu_device, pm_nb); - int r; switch (mode) { case PM_HIBERNATION_PREPARE: adev->in_s4 = true; - fallthrough; - case PM_SUSPEND_PREPARE: - r = amdgpu_device_evict_resources(adev); - /* - * This is considered non-fatal at this time because - * amdgpu_device_prepare() will also fatally evict resources. - * See https://gitlab.freedesktop.org/drm/amd/-/issues/3781 - */ - if (r) - drm_warn(adev_to_drm(adev), "Failed to evict resources, freeze active processes if problems occur: %d\n", r); + break; + case PM_POST_HIBERNATION: + adev->in_s4 = false; break; } -- cgit v1.2.3