From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from gabe.freedesktop.org (gabe.freedesktop.org [131.252.210.177]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.lore.kernel.org (Postfix) with ESMTPS id 16BEACA1002 for ; Thu, 4 Sep 2025 16:07:45 +0000 (UTC) Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id CD5AF10EA95; Thu, 4 Sep 2025 16:07:44 +0000 (UTC) Authentication-Results: gabe.freedesktop.org; dkim=pass (2048-bit key; unprotected) header.d=intel.com header.i=@intel.com header.b="IH8mXnJ+"; dkim-atps=neutral Received: from mgamail.intel.com (mgamail.intel.com [198.175.65.15]) by gabe.freedesktop.org (Postfix) with ESMTPS id 664BC10EA95 for ; Thu, 4 Sep 2025 16:07:44 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=intel.com; i=@intel.com; q=dns/txt; s=Intel; t=1757002065; x=1788538065; h=from:to:cc:subject:date:message-id:in-reply-to: references:mime-version:content-transfer-encoding; bh=ajqOCOkvvVRdqEWLkpBjkvH1LfXDmyn2cawKyc2sk8k=; b=IH8mXnJ+aQaaqGDexR4fpCezwvRltppTD/UD4/0KErYdToZczMQWxFal DShU7KQWBeKWXtpSoQBzHvpuyJskn1U4rsDTalQAqOC1zwUKzK6c931E2 v3W4kvu2aX7aALQ28m22j+sZHREcryTz4ONX8J0397b79WW1kGdHeXcaB de+UxlitJxG6RuMdsKods3e1Th3oZpGWY9ULH7gVYcLyDpWukSPVqHo34 E4synVgbd7UNO8z0bZogPq9zYVIMQHmNysouWBONESNghQyujJNq0jQZG wvUen2wq8/D+z1BEIexOlEOmBNRzxwExE3hBllwIkI8GGFGQbA3YvWzzW w==; X-CSE-ConnectionGUID: dzed2aC3TKKSdE/b9p/dSg== X-CSE-MsgGUID: 0Pm5FyWMTwSWIjfsDpsOGg== X-IronPort-AV: E=McAfee;i="6800,10657,11543"; a="62988502" X-IronPort-AV: E=Sophos;i="6.18,238,1751266800"; d="scan'208";a="62988502" Received: from orviesa005.jf.intel.com ([10.64.159.145]) by orvoesa107.jf.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 04 Sep 2025 09:07:44 -0700 X-CSE-ConnectionGUID: z0kXa+RCSliapZBjcQUlfA== X-CSE-MsgGUID: RfWJpBkNQamQw29A9PM4uw== X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="6.18,238,1751266800"; d="scan'208";a="177201765" Received: from abityuts-desk.ger.corp.intel.com (HELO fedora) ([10.245.244.98]) by orviesa005-auth.jf.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 04 Sep 2025 09:07:43 -0700 From: =?UTF-8?q?Thomas=20Hellstr=C3=B6m?= To: intel-xe@lists.freedesktop.org Cc: =?UTF-8?q?Thomas=20Hellstr=C3=B6m?= , Matthew Auld , Matthew Brost , Rodrigo vivi , Maarten Lankhorst Subject: [PATCH v4 3/3] drm/xe: Block exec and rebind worker while evicting for suspend / hibernate Date: Thu, 4 Sep 2025 18:07:15 +0200 Message-ID: <20250904160715.2613-4-thomas.hellstrom@linux.intel.com> X-Mailer: git-send-email 2.50.1 In-Reply-To: <20250904160715.2613-1-thomas.hellstrom@linux.intel.com> References: <20250904160715.2613-1-thomas.hellstrom@linux.intel.com> MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit X-BeenThere: intel-xe@lists.freedesktop.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: Intel Xe graphics driver List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: intel-xe-bounces@lists.freedesktop.org Sender: "Intel-xe" When the xe pm_notifier evicts for suspend / hibernate, there might be racing tasks trying to re-validate again. This can lead to suspend taking excessive time or get stuck in a live-lock. This behaviour becomes much worse with the fix that actually makes re-validation bring back bos to VRAM rather than letting them remain in TT. Prevent that by having exec and the rebind worker waiting for a completion that is set to block by the pm_notifier before suspend and is signaled by the pm_notifier after resume / wakeup. It's probably still possible to craft malicious applications that block suspending. More work is pending to fix that. v3: - Avoid wait_for_completion() in the kernel worker since it could potentially cause work item flushes from freezable processes to wait forever. Instead terminate the rebind workers if needed and re-launch at resume. (Matt Auld) v4: - Fix some bad naming and leftover debug printouts. - Fix kerneldoc. - Use drmm_mutex_init() for the xe->rebind_resume_lock (Matt Auld). - Rework the interface of xe_vm_rebind_resume_worker (Matt Auld). Link: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/4288 Fixes: c6a4d46ec1d7 ("drm/xe: evict user memory in PM notifier") Cc: Matthew Auld Cc: Rodrigo Vivi Cc: # v6.16+ Signed-off-by: Thomas Hellström --- drivers/gpu/drm/xe/xe_device_types.h | 6 ++++ drivers/gpu/drm/xe/xe_exec.c | 9 ++++++ drivers/gpu/drm/xe/xe_pm.c | 25 +++++++++++++++++ drivers/gpu/drm/xe/xe_vm.c | 42 +++++++++++++++++++++++++++- drivers/gpu/drm/xe/xe_vm.h | 2 ++ drivers/gpu/drm/xe/xe_vm_types.h | 5 ++++ 6 files changed, 88 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/xe/xe_device_types.h b/drivers/gpu/drm/xe/xe_device_types.h index 092004d14db2..1e780f8a2a8c 100644 --- a/drivers/gpu/drm/xe/xe_device_types.h +++ b/drivers/gpu/drm/xe/xe_device_types.h @@ -507,6 +507,12 @@ struct xe_device { /** @pm_notifier: Our PM notifier to perform actions in response to various PM events. */ struct notifier_block pm_notifier; + /** @pm_block: Completion to block validating tasks on suspend / hibernate prepare */ + struct completion pm_block; + /** @rebind_resume_list: List of wq items to kick on resume. */ + struct list_head rebind_resume_list; + /** @rebind_resume_lock: Lock to protect the rebind_resume_list */ + struct mutex rebind_resume_lock; /** @pmt: Support the PMT driver callback interface */ struct { diff --git a/drivers/gpu/drm/xe/xe_exec.c b/drivers/gpu/drm/xe/xe_exec.c index 44364c042ad7..374c831e691b 100644 --- a/drivers/gpu/drm/xe/xe_exec.c +++ b/drivers/gpu/drm/xe/xe_exec.c @@ -237,6 +237,15 @@ int xe_exec_ioctl(struct drm_device *dev, void *data, struct drm_file *file) goto err_unlock_list; } + /* + * It's OK to block interruptible here with the vm lock held, since + * on task freezing during suspend / hibernate, the call will + * return -ERESTARTSYS and the IOCTL will be rerun. + */ + err = wait_for_completion_interruptible(&xe->pm_block); + if (err) + goto err_unlock_list; + vm_exec.vm = &vm->gpuvm; vm_exec.flags = DRM_EXEC_INTERRUPTIBLE_WAIT; if (xe_vm_in_lr_mode(vm)) { diff --git a/drivers/gpu/drm/xe/xe_pm.c b/drivers/gpu/drm/xe/xe_pm.c index bee9aacd82e7..d9c466177147 100644 --- a/drivers/gpu/drm/xe/xe_pm.c +++ b/drivers/gpu/drm/xe/xe_pm.c @@ -25,6 +25,7 @@ #include "xe_pxp.h" #include "xe_sriov_vf_ccs.h" #include "xe_trace.h" +#include "xe_vm.h" #include "xe_wa.h" /** @@ -297,6 +298,19 @@ static u32 vram_threshold_value(struct xe_device *xe) return DEFAULT_VRAM_THRESHOLD; } +static void xe_pm_wake_rebind_workers(struct xe_device *xe) +{ + struct xe_vm *vm, *next; + + mutex_lock(&xe->rebind_resume_lock); + list_for_each_entry_safe(vm, next, &xe->rebind_resume_list, + preempt.pm_activate_link) { + list_del_init(&vm->preempt.pm_activate_link); + xe_vm_resume_rebind_worker(vm); + } + mutex_unlock(&xe->rebind_resume_lock); +} + static int xe_pm_notifier_callback(struct notifier_block *nb, unsigned long action, void *data) { @@ -306,6 +320,7 @@ static int xe_pm_notifier_callback(struct notifier_block *nb, switch (action) { case PM_HIBERNATION_PREPARE: case PM_SUSPEND_PREPARE: + reinit_completion(&xe->pm_block); xe_pm_runtime_get(xe); err = xe_bo_evict_all_user(xe); if (err) @@ -322,6 +337,8 @@ static int xe_pm_notifier_callback(struct notifier_block *nb, break; case PM_POST_HIBERNATION: case PM_POST_SUSPEND: + complete_all(&xe->pm_block); + xe_pm_wake_rebind_workers(xe); xe_bo_notifier_unprepare_all_pinned(xe); xe_pm_runtime_put(xe); break; @@ -348,6 +365,14 @@ int xe_pm_init(struct xe_device *xe) if (err) return err; + err = drmm_mutex_init(&xe->drm, &xe->rebind_resume_lock); + if (err) + goto err_unregister; + + init_completion(&xe->pm_block); + complete_all(&xe->pm_block); + INIT_LIST_HEAD(&xe->rebind_resume_list); + /* For now suspend/resume is only allowed with GuC */ if (!xe_device_uc_enabled(xe)) return 0; diff --git a/drivers/gpu/drm/xe/xe_vm.c b/drivers/gpu/drm/xe/xe_vm.c index f55f96bb240a..8dc3eb6a265a 100644 --- a/drivers/gpu/drm/xe/xe_vm.c +++ b/drivers/gpu/drm/xe/xe_vm.c @@ -394,6 +394,9 @@ static int xe_gpuvm_validate(struct drm_gpuvm_bo *vm_bo, struct drm_exec *exec) list_move_tail(&gpuva_to_vma(gpuva)->combined_links.rebind, &vm->rebind_list); + if (!try_wait_for_completion(&vm->xe->pm_block)) + return -EAGAIN; + ret = xe_bo_validate(gem_to_xe_bo(vm_bo->obj), vm, false); if (ret) return ret; @@ -480,6 +483,33 @@ static int xe_preempt_work_begin(struct drm_exec *exec, struct xe_vm *vm, return xe_vm_validate_rebind(vm, exec, vm->preempt.num_exec_queues); } +static bool vm_suspend_rebind_worker(struct xe_vm *vm) +{ + struct xe_device *xe = vm->xe; + bool ret = false; + + mutex_lock(&xe->rebind_resume_lock); + if (!try_wait_for_completion(&vm->xe->pm_block)) { + ret = true; + list_move_tail(&vm->preempt.pm_activate_link, &xe->rebind_resume_list); + } + mutex_unlock(&xe->rebind_resume_lock); + + return ret; +} + +/** + * xe_vm_resume_rebind_worker() - Resume the rebind worker. + * @vm: The vm whose preempt worker to resume. + * + * Resume a preempt worker that was previously suspended by + * vm_suspend_rebind_worker(). + */ +void xe_vm_resume_rebind_worker(struct xe_vm *vm) +{ + queue_work(vm->xe->ordered_wq, &vm->preempt.rebind_work); +} + static void preempt_rebind_work_func(struct work_struct *w) { struct xe_vm *vm = container_of(w, struct xe_vm, preempt.rebind_work); @@ -503,6 +533,11 @@ static void preempt_rebind_work_func(struct work_struct *w) } retry: + if (!try_wait_for_completion(&vm->xe->pm_block) && vm_suspend_rebind_worker(vm)) { + up_write(&vm->lock); + return; + } + if (xe_vm_userptr_check_repin(vm)) { err = xe_vm_userptr_pin(vm); if (err) @@ -1741,6 +1776,7 @@ struct xe_vm *xe_vm_create(struct xe_device *xe, u32 flags, struct xe_file *xef) if (flags & XE_VM_FLAG_LR_MODE) { INIT_WORK(&vm->preempt.rebind_work, preempt_rebind_work_func); xe_pm_runtime_get_noresume(xe); + INIT_LIST_HEAD(&vm->preempt.pm_activate_link); } if (flags & XE_VM_FLAG_FAULT_MODE) { @@ -1922,8 +1958,12 @@ void xe_vm_close_and_put(struct xe_vm *vm) xe_assert(xe, !vm->preempt.num_exec_queues); xe_vm_close(vm); - if (xe_vm_in_preempt_fence_mode(vm)) + if (xe_vm_in_preempt_fence_mode(vm)) { + mutex_lock(&xe->rebind_resume_lock); + list_del_init(&vm->preempt.pm_activate_link); + mutex_unlock(&xe->rebind_resume_lock); flush_work(&vm->preempt.rebind_work); + } if (xe_vm_in_fault_mode(vm)) xe_svm_close(vm); diff --git a/drivers/gpu/drm/xe/xe_vm.h b/drivers/gpu/drm/xe/xe_vm.h index b3e5bec0fa58..d631c4b25c51 100644 --- a/drivers/gpu/drm/xe/xe_vm.h +++ b/drivers/gpu/drm/xe/xe_vm.h @@ -281,6 +281,8 @@ struct dma_fence *xe_vm_bind_kernel_bo(struct xe_vm *vm, struct xe_bo *bo, struct xe_exec_queue *q, u64 addr, enum xe_cache_level cache_lvl); +void xe_vm_resume_rebind_worker(struct xe_vm *vm); + /** * xe_vm_resv() - Return's the vm's reservation object * @vm: The vm diff --git a/drivers/gpu/drm/xe/xe_vm_types.h b/drivers/gpu/drm/xe/xe_vm_types.h index b5108d010786..e1a786db5f89 100644 --- a/drivers/gpu/drm/xe/xe_vm_types.h +++ b/drivers/gpu/drm/xe/xe_vm_types.h @@ -338,6 +338,11 @@ struct xe_vm { * BOs */ struct work_struct rebind_work; + /** + * @preempt.pm_activate_link: Link to list of rebind workers to be + * kicked on resume. + */ + struct list_head pm_activate_link; } preempt; /** @um: unified memory state */ -- 2.50.1