From: Thomas Hellström Date: Thu, 18 Sep 2025 14:28:48 +0000 (+0200) Subject: drm/xe/pm: Add lockdep annotation for the pm_block completion X-Git-Url: http://git.ipfire.org/?a=commitdiff_plain;h=f73f6dd312a5616a8d97860663f6c88ffba43ad4;p=thirdparty%2Fkernel%2Flinux.git drm/xe/pm: Add lockdep annotation for the pm_block completion Similar to how we annotate dma-fences, add lockep annotation to the pm_block completion to ensure we don't wait for it while holding locks that are needed in the pm notifier or in the device suspend / resume callbacks. Signed-off-by: Thomas Hellström Reviewed-by: Matthew Auld Link: https://lore.kernel.org/r/20250918142848.21807-3-thomas.hellstrom@linux.intel.com --- diff --git a/drivers/gpu/drm/xe/xe_exec.c b/drivers/gpu/drm/xe/xe_exec.c index 7715e74bb9454..83897950f0da3 100644 --- a/drivers/gpu/drm/xe/xe_exec.c +++ b/drivers/gpu/drm/xe/xe_exec.c @@ -16,6 +16,7 @@ #include "xe_exec_queue.h" #include "xe_hw_engine_group.h" #include "xe_macros.h" +#include "xe_pm.h" #include "xe_ring_ops_types.h" #include "xe_sched_job.h" #include "xe_sync.h" @@ -247,7 +248,7 @@ retry: * 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); + err = xe_pm_block_on_suspend(xe); if (err) goto err_unlock_list; diff --git a/drivers/gpu/drm/xe/xe_pm.c b/drivers/gpu/drm/xe/xe_pm.c index 1ec03ef19d9e8..96afa49f0b4ba 100644 --- a/drivers/gpu/drm/xe/xe_pm.c +++ b/drivers/gpu/drm/xe/xe_pm.c @@ -83,8 +83,58 @@ static struct lockdep_map xe_pm_runtime_d3cold_map = { static struct lockdep_map xe_pm_runtime_nod3cold_map = { .name = "xe_rpm_nod3cold_map" }; + +static struct lockdep_map xe_pm_block_lockdep_map = { + .name = "xe_pm_block_map", +}; #endif +static void xe_pm_block_begin_signalling(void) +{ + lock_acquire_shared_recursive(&xe_pm_block_lockdep_map, 0, 1, NULL, _RET_IP_); +} + +static void xe_pm_block_end_signalling(void) +{ + lock_release(&xe_pm_block_lockdep_map, _RET_IP_); +} + +/** + * xe_pm_might_block_on_suspend() - Annotate that the code might block on suspend + * + * Annotation to use where the code might block or sieze to make + * progress pending resume completion. + */ +void xe_pm_might_block_on_suspend(void) +{ + lock_map_acquire(&xe_pm_block_lockdep_map); + lock_map_release(&xe_pm_block_lockdep_map); +} + +/** + * xe_pm_might_block_on_suspend() - Block pending suspend. + * @xe: The xe device about to be suspended. + * + * Block if the pm notifier has start evicting bos, to avoid + * racing and validating those bos back. The function is + * annotated to ensure no locks are held that are also grabbed + * in the pm notifier or the device suspend / resume. + * This is intended to be used by freezable tasks only. + * (Not freezable workqueues), with the intention that the function + * returns %-ERESTARTSYS when tasks are frozen during suspend, + * and allows the task to freeze. The caller must be able to + * handle the %-ERESTARTSYS. + * + * Return: %0 on success, %-ERESTARTSYS on signal pending or + * if freezing requested. + */ +int xe_pm_block_on_suspend(struct xe_device *xe) +{ + xe_pm_might_block_on_suspend(); + + return wait_for_completion_interruptible(&xe->pm_block); +} + /** * xe_rpm_reclaim_safe() - Whether runtime resume can be done from reclaim context * @xe: The xe device. @@ -124,6 +174,7 @@ int xe_pm_suspend(struct xe_device *xe) int err; drm_dbg(&xe->drm, "Suspending device\n"); + xe_pm_block_begin_signalling(); trace_xe_pm_suspend(xe, __builtin_return_address(0)); err = xe_pxp_pm_suspend(xe->pxp); @@ -155,6 +206,8 @@ int xe_pm_suspend(struct xe_device *xe) xe_i2c_pm_suspend(xe); drm_dbg(&xe->drm, "Device suspended\n"); + xe_pm_block_end_signalling(); + return 0; err_display: @@ -162,6 +215,7 @@ err_display: xe_pxp_pm_resume(xe->pxp); err: drm_dbg(&xe->drm, "Device suspend failed %d\n", err); + xe_pm_block_end_signalling(); return err; } @@ -178,6 +232,7 @@ int xe_pm_resume(struct xe_device *xe) u8 id; int err; + xe_pm_block_begin_signalling(); drm_dbg(&xe->drm, "Resuming device\n"); trace_xe_pm_resume(xe, __builtin_return_address(0)); @@ -222,9 +277,11 @@ int xe_pm_resume(struct xe_device *xe) xe_late_bind_fw_load(&xe->late_bind); drm_dbg(&xe->drm, "Device resumed\n"); + xe_pm_block_end_signalling(); return 0; err: drm_dbg(&xe->drm, "Device resume failed %d\n", err); + xe_pm_block_end_signalling(); return err; } @@ -333,6 +390,7 @@ static int xe_pm_notifier_callback(struct notifier_block *nb, struct xe_validation_ctx ctx; reinit_completion(&xe->pm_block); + xe_pm_block_begin_signalling(); xe_pm_runtime_get(xe); (void)xe_validation_ctx_init(&ctx, &xe->val, NULL, (struct xe_val_flags) {.exclusive = true}); @@ -349,6 +407,7 @@ static int xe_pm_notifier_callback(struct notifier_block *nb, * avoid a runtime suspend interfering with evicted objects or backup * allocations. */ + xe_pm_block_end_signalling(); break; } case PM_POST_HIBERNATION: diff --git a/drivers/gpu/drm/xe/xe_pm.h b/drivers/gpu/drm/xe/xe_pm.h index 59678b310e55f..f7f89a18b6fc7 100644 --- a/drivers/gpu/drm/xe/xe_pm.h +++ b/drivers/gpu/drm/xe/xe_pm.h @@ -33,6 +33,8 @@ int xe_pm_set_vram_threshold(struct xe_device *xe, u32 threshold); void xe_pm_d3cold_allowed_toggle(struct xe_device *xe); bool xe_rpm_reclaim_safe(const struct xe_device *xe); struct task_struct *xe_pm_read_callback_task(struct xe_device *xe); +int xe_pm_block_on_suspend(struct xe_device *xe); +void xe_pm_might_block_on_suspend(void); int xe_pm_module_init(void); #endif diff --git a/drivers/gpu/drm/xe/xe_vm.c b/drivers/gpu/drm/xe/xe_vm.c index 0cacab20ff852..80b7f13ecd804 100644 --- a/drivers/gpu/drm/xe/xe_vm.c +++ b/drivers/gpu/drm/xe/xe_vm.c @@ -466,6 +466,8 @@ 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); + /* We don't actually block but don't make progress. */ + xe_pm_might_block_on_suspend(); return; }