From e1ee2111ca48169a9fdc5075f7863f5d4d591e2f Mon Sep 17 00:00:00 2001 From: Lijo Lazar Date: Thu, 24 Oct 2024 11:01:57 +0530 Subject: drm/amdgpu: Prefer RAS recovery for scheduler hang Before scheduling a recovery due to scheduler/job hang, check if a RAS error is detected. If so, choose RAS recovery to handle the situation. A scheduler/job hang could be the side effect of a RAS error. In such cases, it is required to go through the RAS error recovery process. A RAS error recovery process in certains cases also could avoid a full device device reset. An error state is maintained in RAS context to detect the block affected. Fatal Error state uses unused block id. Set the block id when error is detected. If the interrupt handler detected a poison error, it's not required to look for a fatal error. Skip fatal error checking in such cases. Signed-off-by: Lijo Lazar Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 15 +++++++++++++-- 1 file changed, 13 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 d272d95dd5b2..97d3e5f29638 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -5181,7 +5181,7 @@ static int amdgpu_device_reset_sriov(struct amdgpu_device *adev, if (r) return r; - amdgpu_ras_set_fed(adev, false); + amdgpu_ras_clear_err_state(adev); amdgpu_irq_gpu_reset_resume_helper(adev); /* some sw clean up VF needs to do before recover */ @@ -5484,7 +5484,7 @@ int amdgpu_device_reinit_after_reset(struct amdgpu_reset_context *reset_context) amdgpu_set_init_level(tmp_adev, init_level); if (full_reset) { /* post card */ - amdgpu_ras_set_fed(tmp_adev, false); + amdgpu_ras_clear_err_state(tmp_adev); r = amdgpu_device_asic_init(tmp_adev); if (r) { dev_warn(tmp_adev->dev, "asic atom init failed!"); @@ -5817,6 +5817,17 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, 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) && + 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 */ -- cgit v1.2.3 From ff69bba05f085cd6d4277c27ac7600160167b384 Mon Sep 17 00:00:00 2001 From: Boyuan Zhang Date: Wed, 2 Oct 2024 23:52:01 -0400 Subject: drm/amd/pm: add inst to dpm_set_powergating_by_smu MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Add an instance parameter to amdgpu_dpm_set_powergating_by_smu() function, and use the instance to call set_powergating_by_smu(). v2: remove duplicated functions. remove for-loop in amdgpu_dpm_set_powergating_by_smu(), and temporarily move it to amdgpu_dpm_enable_vcn(), in order to keep the exact same logic as before, until further separation in next patch. v3: drop SI logic in amdgpu_dpm_enable_vcn(). Signed-off-by: Boyuan Zhang Acked-by: Christian König Reviewed-by: Alex Deucher 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 97d3e5f29638..e41182645f94 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -3478,7 +3478,7 @@ static void amdgpu_device_delay_enable_gfx_off(struct work_struct *work) WARN_ON_ONCE(adev->gfx.gfx_off_state); WARN_ON_ONCE(adev->gfx.gfx_off_req_count); - if (!amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_GFX, true)) + if (!amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_GFX, true, 0)) adev->gfx.gfx_off_state = true; } -- cgit v1.2.3 From 80d80511241cd3f532704dc4abbb604425552396 Mon Sep 17 00:00:00 2001 From: Boyuan Zhang Date: Sun, 29 Sep 2024 15:17:51 -0400 Subject: drm/amdgpu: pass ip_block in set_powergating_state MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Pass ip_block instead of adev in set_powergating_state callback function. Modify set_powergating_state ip functions for all correspoding ip blocks. v2: fix a ip block index error. v3: remove type casting Signed-off-by: Boyuan Zhang Suggested-by: Christian König Acked-by: Christian König Acked-by: Alex Deucher Reviewed-by: Sunil Khatri Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 4 ++-- 1 file changed, 2 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 e41182645f94..643a9a44e614 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -2199,7 +2199,7 @@ int amdgpu_device_ip_set_powergating_state(void *dev, if (!adev->ip_blocks[i].version->funcs->set_powergating_state) continue; r = adev->ip_blocks[i].version->funcs->set_powergating_state( - (void *)adev, state); + &adev->ip_blocks[i], state); if (r) DRM_ERROR("set_powergating_state of IP block <%s> failed %d\n", adev->ip_blocks[i].version->funcs->name, r); @@ -3174,7 +3174,7 @@ int amdgpu_device_set_pg_state(struct amdgpu_device *adev, adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_JPEG && adev->ip_blocks[i].version->funcs->set_powergating_state) { /* enable powergating to save power */ - r = adev->ip_blocks[i].version->funcs->set_powergating_state((void *)adev, + r = adev->ip_blocks[i].version->funcs->set_powergating_state(&adev->ip_blocks[i], state); if (r) { DRM_ERROR("set_powergating_state(gate) of IP block <%s> failed %d\n", -- cgit v1.2.3 From f2ba8c3d510695c68ede786b3a37a9aa170ddfae Mon Sep 17 00:00:00 2001 From: Boyuan Zhang Date: Mon, 7 Oct 2024 19:43:31 -0400 Subject: drm/amdgpu: pass ip_block in set_clockgating_state MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Pass ip_block instead of adev in set_clockgating_state() callback functions. Modify set_clockgating_state()for all correspoding ip blocks. v2: remove all changes for is_idle(), remove type casting Signed-off-by: Boyuan Zhang Acked-by: Christian König Reviewed-by: Sunil Khatri Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 4 ++-- 1 file changed, 2 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 643a9a44e614..312c507ef5f9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -2165,7 +2165,7 @@ int amdgpu_device_ip_set_clockgating_state(void *dev, if (!adev->ip_blocks[i].version->funcs->set_clockgating_state) continue; r = adev->ip_blocks[i].version->funcs->set_clockgating_state( - (void *)adev, state); + &adev->ip_blocks[i], state); if (r) DRM_ERROR("set_clockgating_state of IP block <%s> failed %d\n", adev->ip_blocks[i].version->funcs->name, r); @@ -3137,7 +3137,7 @@ int amdgpu_device_set_cg_state(struct amdgpu_device *adev, adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_JPEG && adev->ip_blocks[i].version->funcs->set_clockgating_state) { /* enable clockgating to save power */ - r = adev->ip_blocks[i].version->funcs->set_clockgating_state((void *)adev, + r = adev->ip_blocks[i].version->funcs->set_clockgating_state(&adev->ip_blocks[i], state); if (r) { DRM_ERROR("set_clockgating_state(gate) of IP block <%s> failed %d\n", -- cgit v1.2.3 From 0859eb540f1412cced6234922626c8b1e6072126 Mon Sep 17 00:00:00 2001 From: Shikang Fan Date: Thu, 21 Nov 2024 17:06:30 +0800 Subject: drm/amdgpu: Check fence emitted count to identify bad jobs MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit In SRIOV, when host driver performs MODE 1 reset and notifies FLR to guest driver, there is a small chance that there is no job running on hw but the driver has not updated the pending list yet, causing the driver not respond the FLR request. Modify the has_job_running function to make sure if there is still running job. v2: Use amdgpu_fence_count_emitted to determine job running status. v3: Remove the timeout wait in has_job_running Signed-off-by: Emily Deng Signed-off-by: Shikang Fan Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 14 ++++++-------- 1 file changed, 6 insertions(+), 8 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 312c507ef5f9..07c84e4cef5a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -5238,16 +5238,18 @@ static int amdgpu_device_reset_sriov(struct amdgpu_device *adev, } /** - * amdgpu_device_has_job_running - check if there is any job in mirror list + * amdgpu_device_has_job_running - check if there is any unfinished job * * @adev: amdgpu_device pointer * - * check if there is any job in mirror list + * check if there is any job running on the device when guest driver receives + * FLR notification from host driver. If there are still jobs running, then + * the guest driver will not respond the FLR reset. Instead, let the job hit + * the timeout and guest driver then issue the reset request. */ bool amdgpu_device_has_job_running(struct amdgpu_device *adev) { int i; - struct drm_sched_job *job; for (i = 0; i < AMDGPU_MAX_RINGS; ++i) { struct amdgpu_ring *ring = adev->rings[i]; @@ -5255,11 +5257,7 @@ bool amdgpu_device_has_job_running(struct amdgpu_device *adev) if (!amdgpu_ring_sched_ready(ring)) continue; - spin_lock(&ring->sched.job_list_lock); - job = list_first_entry_or_null(&ring->sched.pending_list, - struct drm_sched_job, list); - spin_unlock(&ring->sched.job_list_lock); - if (job) + if (amdgpu_fence_count_emitted(ring)) return true; } return false; -- cgit v1.2.3 From c2ee5c2f0ea7808f74500ed2eb7a92a87ae8595c Mon Sep 17 00:00:00 2001 From: Mario Limonciello Date: Wed, 27 Nov 2024 21:26:55 -0600 Subject: drm/amd: Invert APU check for amdgpu_device_evict_resources() Resource eviction isn't needed for s3 or s2idle on APUs, but should be run for S4. As amdgpu_device_evict_resources() will be called by prepare notifier adjust logic so that APUs only cover S4. Suggested-by: Lijo Lazar Reviewed-by: Lijo Lazar Link: https://lore.kernel.org/r/20241128032656.2090059-1-superm1@kernel.org Signed-off-by: Mario Limonciello Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 4 ++-- 1 file changed, 2 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 07c84e4cef5a..69109f5c2c8c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -4778,8 +4778,8 @@ static int amdgpu_device_evict_resources(struct amdgpu_device *adev) { int ret; - /* No need to evict vram on APUs for suspend to ram or s2idle */ - if ((adev->in_s3 || adev->in_s0ix) && (adev->flags & AMD_IS_APU)) + /* No need to evict vram on APUs unless going to S4 */ + if (!adev->in_s4 && (adev->flags & AMD_IS_APU)) return 0; ret = amdgpu_ttm_evict_resources(adev, TTM_PL_VRAM); -- cgit v1.2.3 From 2965e6355dcdf157b5fafa25a2715f00064da8bf Mon Sep 17 00:00:00 2001 From: Mario Limonciello Date: Wed, 27 Nov 2024 21:26:56 -0600 Subject: drm/amd: Add Suspend/Hibernate notification callback support As part of the suspend sequence VRAM needs to be evicted on dGPUs. In order to make suspend/resume more reliable we moved this into the pmops prepare() callback so that the suspend sequence would fail but the system could remain operational under high memory usage suspend. Another class of issues exist though where due to memory fragementation there isn't a large enough contiguous space and swap isn't accessible. Add support for a suspend/hibernate notification callback that could evict VRAM before tasks are frozen. This should allow paging out to swap if necessary. Link: https://github.com/ROCm/ROCK-Kernel-Driver/issues/174 Link: https://gitlab.freedesktop.org/drm/amd/-/issues/3476 Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/2362 Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/3781 Reviewed-by: Lijo Lazar Link: https://lore.kernel.org/r/20241128032656.2090059-2-superm1@kernel.org Signed-off-by: Mario Limonciello Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 46 +++++++++++++++++++++++++++++- 1 file changed, 45 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 69109f5c2c8c..d44123df6599 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -199,6 +199,8 @@ void amdgpu_set_init_level(struct amdgpu_device *adev, } static inline void amdgpu_device_stop_pending_resets(struct amdgpu_device *adev); +static int amdgpu_device_pm_notifier(struct notifier_block *nb, unsigned long mode, + void *data); /** * DOC: pcie_replay_count @@ -4596,6 +4598,11 @@ fence_driver_init: amdgpu_device_check_iommu_direct_map(adev); + adev->pm_nb.notifier_call = amdgpu_device_pm_notifier; + r = register_pm_notifier(&adev->pm_nb); + if (r) + goto failed; + return 0; release_ras_con: @@ -4660,6 +4667,8 @@ void amdgpu_device_fini_hw(struct amdgpu_device *adev) drain_workqueue(adev->mman.bdev.wq); adev->shutdown = true; + unregister_pm_notifier(&adev->pm_nb); + /* make sure IB test finished before entering exclusive mode * to avoid preemption on IB test */ @@ -4791,6 +4800,41 @@ static int amdgpu_device_evict_resources(struct amdgpu_device *adev) /* * Suspend & resume. */ +/** + * amdgpu_device_pm_notifier - Notification block for Suspend/Hibernate events + * @nb: notifier block + * @mode: suspend mode + * @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. + */ +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; + } + + return NOTIFY_DONE; +} + /** * amdgpu_device_prepare - prepare for device suspend * @@ -4830,7 +4874,7 @@ int amdgpu_device_prepare(struct drm_device *dev) return 0; unprepare: - adev->in_s0ix = adev->in_s3 = false; + adev->in_s0ix = adev->in_s3 = adev->in_s4 = false; return r; } -- cgit v1.2.3 From a567db808ec9482d8b4ce5f9e5fcbc9e6c5a6da9 Mon Sep 17 00:00:00 2001 From: Randy Dunlap Date: Wed, 27 Nov 2024 19:20:53 -0800 Subject: drm/amdgpu: device: fix spellos and punctuation MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Make spelling and punctuation changes to ease reading of the comments. Signed-off-by: Randy Dunlap Cc: Alex Deucher Cc: Christian König Cc: Xinhui Pan Cc: amd-gfx@lists.freedesktop.org Cc: David Airlie Cc: Simona Vetter Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 30 +++++++++++++++--------------- 1 file changed, 15 insertions(+), 15 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 d44123df6599..bc92bc6b866f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -206,9 +206,9 @@ static int amdgpu_device_pm_notifier(struct notifier_block *nb, unsigned long mo * DOC: pcie_replay_count * * The amdgpu driver provides a sysfs API for reporting the total number - * of PCIe replays (NAKs) + * of PCIe replays (NAKs). * The file pcie_replay_count is used for this and returns the total - * number of replays as a sum of the NAKs generated and NAKs received + * number of replays as a sum of the NAKs generated and NAKs received. */ static ssize_t amdgpu_device_get_pcie_replay_count(struct device *dev, @@ -431,8 +431,8 @@ bool amdgpu_device_supports_boco(struct drm_device *dev) * @dev: drm_device pointer * * Return: - * 1 if the device supporte BACO; - * 3 if the device support MACO (only works if BACO is supported) + * 1 if the device supports BACO; + * 3 if the device supports MACO (only works if BACO is supported) * otherwise return 0. */ int amdgpu_device_supports_baco(struct drm_device *dev) @@ -579,7 +579,7 @@ void amdgpu_device_mm_access(struct amdgpu_device *adev, loff_t pos, } /** - * amdgpu_device_aper_access - access vram by vram aperature + * amdgpu_device_aper_access - access vram by vram aperture * * @adev: amdgpu_device pointer * @pos: offset of the buffer in vram @@ -670,7 +670,7 @@ bool amdgpu_device_skip_hw_access(struct amdgpu_device *adev) * here is that the GPU reset is not running on another thread in parallel. * * For this we trylock the read side of the reset semaphore, if that succeeds - * we know that the reset is not running in paralell. + * we know that the reset is not running in parallel. * * If the trylock fails we assert that we are either already holding the read * side of the lock or are the reset thread itself and hold the write side of @@ -1735,7 +1735,7 @@ bool amdgpu_device_need_post(struct amdgpu_device *adev) uint32_t fw_ver; err = request_firmware(&adev->pm.fw, "amdgpu/fiji_smc.bin", adev->dev); - /* force vPost if error occured */ + /* force vPost if error occurred */ if (err) return true; @@ -2380,7 +2380,7 @@ int amdgpu_device_ip_block_add(struct amdgpu_device *adev, * the module parameter virtual_display. This feature provides a virtual * display hardware on headless boards or in virtualized environments. * This function parses and validates the configuration string specified by - * the user and configues the virtual display configuration (number of + * the user and configures the virtual display configuration (number of * virtual connectors, crtcs, etc.) specified. */ static void amdgpu_device_enable_virtual_display(struct amdgpu_device *adev) @@ -2443,7 +2443,7 @@ void amdgpu_device_set_sriov_virtual_display(struct amdgpu_device *adev) * @adev: amdgpu_device pointer * * Parses the asic configuration parameters specified in the gpu info - * firmware and makes them availale to the driver for use in configuring + * firmware and makes them available to the driver for use in configuring * the asic. * Returns 0 on success, -EINVAL on failure. */ @@ -2503,7 +2503,7 @@ static int amdgpu_device_parse_gpu_info_fw(struct amdgpu_device *adev) le32_to_cpu(hdr->header.ucode_array_offset_bytes)); /* - * Should be droped when DAL no longer needs it. + * Should be dropped when DAL no longer needs it. */ if (adev->asic_type == CHIP_NAVI12) goto parse_soc_bounding_box; @@ -3063,7 +3063,7 @@ init_failed: * * Writes a reset magic value to the gart pointer in VRAM. The driver calls * this function before a GPU reset. If the value is retained after a - * GPU reset, VRAM has not been lost. Some GPU resets may destry VRAM contents. + * GPU reset, VRAM has not been lost. Some GPU resets may destroy VRAM contents. */ static void amdgpu_device_fill_reset_magic(struct amdgpu_device *adev) { @@ -3378,7 +3378,7 @@ static int amdgpu_device_ip_fini_early(struct amdgpu_device *adev) amdgpu_amdkfd_suspend(adev, false); - /* Workaroud for ASICs need to disable SMC first */ + /* Workaround for ASICs need to disable SMC first */ amdgpu_device_smu_fini_early(adev); for (i = adev->num_ip_blocks - 1; i >= 0; i--) { @@ -4308,7 +4308,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, /* * Reset domain needs to be present early, before XGMI hive discovered - * (if any) and intitialized to use reset sem and in_gpu reset flag + * (if any) and initialized to use reset sem and in_gpu reset flag * early on during init and before calling to RREG32. */ adev->reset_domain = amdgpu_reset_create_reset_domain(SINGLE_DEVICE, "amdgpu-reset-dev"); @@ -5953,7 +5953,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, amdgpu_amdkfd_pre_reset(tmp_adev, reset_context); /* - * Mark these ASICs to be reseted as untracked first + * Mark these ASICs to be reset as untracked first * And add them back after reset completed */ amdgpu_unregister_gpu_instance(tmp_adev); @@ -6156,7 +6156,7 @@ static void amdgpu_device_partner_bandwidth(struct amdgpu_device *adev, * * @adev: amdgpu_device pointer * - * Fetchs and stores in the driver the PCIE capabilities (gen speed + * Fetches and stores in the driver the PCIE capabilities (gen speed * and lanes) of the slot the device is in. Handles APUs and * virtualized environments where PCIE config space may not be available. */ -- cgit v1.2.3 From 0b58a55af5d48ce1a4c045853e7d5cd49a4b4e6b Mon Sep 17 00:00:00 2001 From: Le Ma Date: Fri, 8 Nov 2024 14:40:22 -0500 Subject: drm/amdgpu: add initial support for gfx950 add gfx950 basic support Signed-off-by: Le Ma Reviewed-by: Hawking Zhang 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 bc92bc6b866f..63b1f1c0eb34 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -1401,6 +1401,7 @@ static int amdgpu_device_asic_init(struct amdgpu_device *adev) 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); ret = amdgpu_atomfirmware_asic_init(adev, true); -- cgit v1.2.3 From ea5d49349894a7a74ce8dba242e3a487d24b6c0e Mon Sep 17 00:00:00 2001 From: Mario Limonciello Date: Tue, 3 Dec 2024 11:28:54 -0600 Subject: drm/amd: Add the capability to mark certain firmware as "required" Some of the firmware that is loaded by amdgpu is not actually required. For example the ISP firmware on some SoCs is optional, and if it's not present the ISP IP block just won't be initialized. The firmware loader core however will show a warning when this happens like this: ``` Direct firmware load for amdgpu/isp_4_1_0.bin failed with error -2 ``` To avoid confusion for non-required firmware, adjust the amd-ucode helper to take an extra argument indicating if the firmware is required or optional. On optional firmware use firmware_request_nowarn() instead of request_firmware() to avoid the warnings. Reviewed-by: Alex Deucher Link: https://lore.kernel.org/amd-gfx/df71d375-7abd-4b32-97ce-15e57846eed8@amd.com/T/#t Signed-off-by: Mario Limonciello 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 63b1f1c0eb34..144295da9e4c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -2485,6 +2485,7 @@ static int amdgpu_device_parse_gpu_info_fw(struct amdgpu_device *adev) } err = amdgpu_ucode_request(adev, &adev->firmware.gpu_info_fw, + AMDGPU_UCODE_OPTIONAL, "amdgpu/%s_gpu_info.bin", chip_name); if (err) { dev_err(adev->dev, -- cgit v1.2.3 From fccb446f82b9155c05758d1fa30af4a06494e0ec Mon Sep 17 00:00:00 2001 From: Lijo Lazar Date: Mon, 9 Dec 2024 09:14:53 +0530 Subject: drm/amdgpu: Avoid VF for RAS recovery source check VF device sets the RAS flag when mailbox data can't be read properly. There is no conclusive way to tell if the real source is RAS error. Therefore VF schedules a KFD based reset which doesn't set RAS source. SKip checking RAS source for any VF scheduled recovery. Signed-off-by: Lijo Lazar Reported-by: Vojislav Tomasevic Reviewed-by: Yiqing Yao Tested-by: Yiqing Yao Fixes: e1ee2111ca48 ("drm/amdgpu: Prefer RAS recovery for scheduler hang") 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 144295da9e4c..e22fc7a8101f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -5866,6 +5866,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, * 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", -- cgit v1.2.3 From 1ad5bdc28bafa66db0f041cc6cdd278a80426aae Mon Sep 17 00:00:00 2001 From: Mario Limonciello Date: Wed, 11 Dec 2024 09:56:01 -0600 Subject: drm/amd: Require CONFIG_HOTPLUG_PCI_PCIE for BOCO If the kernel hasn't been compiled with PCIe hotplug support this can lead to problems with dGPUs that use BOCO because they effectively drop off the bus. To prevent issues, disable BOCO support when compiled without PCIe hotplug. Reported-by: Gabriel Marcano Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/1707#note_2696862 Acked-by: Alex Deucher Link: https://lore.kernel.org/r/20241211155601.3585256-1-superm1@kernel.org Signed-off-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 e22fc7a8101f..36053b3d48b3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -419,6 +419,9 @@ bool amdgpu_device_supports_boco(struct drm_device *dev) { struct amdgpu_device *adev = drm_to_adev(dev); + if (!IS_ENABLED(CONFIG_HOTPLUG_PCI_PCIE)) + return false; + if (adev->has_pr3 || ((adev->flags & AMD_IS_PX) && amdgpu_is_atpx_hybrid())) return true; -- cgit v1.2.3