diff options
Diffstat (limited to 'mali_kbase/csf/mali_kbase_csf_scheduler.c')
| -rw-r--r-- | mali_kbase/csf/mali_kbase_csf_scheduler.c | 368 |
1 files changed, 264 insertions, 104 deletions
diff --git a/mali_kbase/csf/mali_kbase_csf_scheduler.c b/mali_kbase/csf/mali_kbase_csf_scheduler.c index 22a08a6..f1ff8bf 100644 --- a/mali_kbase/csf/mali_kbase_csf_scheduler.c +++ b/mali_kbase/csf/mali_kbase_csf_scheduler.c @@ -1,7 +1,7 @@ // SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note /* * - * (C) COPYRIGHT 2018-2023 ARM Limited. All rights reserved. + * (C) COPYRIGHT 2018-2024 ARM Limited. All rights reserved. * * This program is free software and is provided to you under the terms of the * GNU General Public License version 2 as published by the Free Software @@ -39,6 +39,7 @@ #include "mali_kbase_csf_tiler_heap_reclaim.h" #include "mali_kbase_csf_mcu_shared_reg.h" #include <linux/version_compat_defs.h> +#include <hwcnt/mali_kbase_hwcnt_context.h> #if IS_ENABLED(CONFIG_MALI_TRACE_POWER_GPU_WORK_PERIOD) #include <mali_kbase_gpu_metrics.h> #include <csf/mali_kbase_csf_trace_buffer.h> @@ -91,7 +92,8 @@ scheduler_get_protm_enter_async_group(struct kbase_device *const kbdev, struct kbase_queue_group *const group); static struct kbase_queue_group *get_tock_top_group(struct kbase_csf_scheduler *const scheduler); static void scheduler_enable_tick_timer_nolock(struct kbase_device *kbdev); -static int suspend_active_queue_groups(struct kbase_device *kbdev, unsigned long *slot_mask); +static int suspend_active_queue_groups(struct kbase_device *kbdev, unsigned long *slot_mask, + bool reset); static int suspend_active_groups_on_powerdown(struct kbase_device *kbdev, bool system_suspend); static void schedule_in_cycle(struct kbase_queue_group *group, bool force); static bool queue_group_scheduled_locked(struct kbase_queue_group *group); @@ -795,7 +797,8 @@ static void update_on_slot_queues_offsets(struct kbase_device *kbdev) static void enqueue_gpu_idle_work(struct kbase_csf_scheduler *const scheduler) { atomic_set(&scheduler->gpu_no_longer_idle, false); - queue_work(scheduler->idle_wq, &scheduler->gpu_idle_work); + atomic_inc(&scheduler->pending_gpu_idle_work); + complete(&scheduler->kthread_signal); } bool kbase_csf_scheduler_process_gpu_idle_event(struct kbase_device *kbdev) @@ -807,7 +810,8 @@ bool kbase_csf_scheduler_process_gpu_idle_event(struct kbase_device *kbdev) lockdep_assert_held(&kbdev->hwaccess_lock); lockdep_assert_held(&scheduler->interrupt_lock); - can_suspend_on_idle = kbase_pm_idle_groups_sched_suspendable(kbdev); + can_suspend_on_idle = kbase_pm_idle_groups_sched_suspendable(kbdev) && + !kbase_pm_is_mcu_inactive(kbdev, kbdev->pm.backend.mcu_state); KBASE_KTRACE_ADD(kbdev, SCHEDULER_GPU_IDLE_EVENT_CAN_SUSPEND, NULL, (((u64)can_suspend_on_idle) << 32)); @@ -2423,6 +2427,11 @@ static void cancel_tock_work(struct kbase_csf_scheduler *const scheduler) atomic_set(&scheduler->pending_tock_work, false); } +static void cancel_gpu_idle_work(struct kbase_csf_scheduler *const scheduler) +{ + atomic_set(&scheduler->pending_gpu_idle_work, false); +} + static void remove_group_from_runnable(struct kbase_csf_scheduler *const scheduler, struct kbase_queue_group *group, enum kbase_csf_group_state run_state) @@ -3148,8 +3157,9 @@ static void sched_evict_group(struct kbase_queue_group *group, bool fault, static int term_group_sync(struct kbase_queue_group *group) { struct kbase_device *kbdev = group->kctx->kbdev; - const unsigned int fw_timeout_ms = kbase_get_timeout_ms(kbdev, CSF_FIRMWARE_TIMEOUT); - long remaining = kbase_csf_timeout_in_jiffies(fw_timeout_ms); + const unsigned int group_term_timeout_ms = + kbase_get_timeout_ms(kbdev, CSF_CSG_TERM_TIMEOUT); + long remaining = kbase_csf_timeout_in_jiffies(group_term_timeout_ms); int err = 0; term_csg_slot(group); @@ -3166,7 +3176,7 @@ static int term_group_sync(struct kbase_queue_group *group) dev_warn( kbdev->dev, "[%llu] term request timeout (%d ms) for group %d of context %d_%d on slot %d", - kbase_backend_get_cycle_cnt(kbdev), fw_timeout_ms, group->handle, + kbase_backend_get_cycle_cnt(kbdev), group_term_timeout_ms, group->handle, group->kctx->tgid, group->kctx->id, group->csg_nr); if (kbase_csf_firmware_ping_wait(kbdev, FW_PING_AFTER_ERROR_TIMEOUT_MS)) error_type = DF_PING_REQUEST_TIMEOUT; @@ -4824,8 +4834,9 @@ static int suspend_active_groups_on_powerdown(struct kbase_device *kbdev, bool s { struct kbase_csf_scheduler *const scheduler = &kbdev->csf.scheduler; DECLARE_BITMAP(slot_mask, MAX_SUPPORTED_CSGS) = { 0 }; + int ret; - int ret = suspend_active_queue_groups(kbdev, slot_mask); + ret = suspend_active_queue_groups(kbdev, slot_mask, false); if (unlikely(ret)) { const int csg_nr = ffs(slot_mask[0]) - 1; @@ -5016,14 +5027,14 @@ static bool scheduler_suspend_on_idle(struct kbase_device *kbdev) return true; } -static void gpu_idle_worker(struct work_struct *work) +static void gpu_idle_worker(struct kbase_device *kbdev) { - struct kbase_device *kbdev = - container_of(work, struct kbase_device, csf.scheduler.gpu_idle_work); struct kbase_csf_scheduler *const scheduler = &kbdev->csf.scheduler; bool scheduler_is_idle_suspendable = false; bool all_groups_suspended = false; + WARN_ON_ONCE(atomic_read(&scheduler->pending_gpu_idle_work) == 0); + KBASE_KTRACE_ADD(kbdev, SCHEDULER_GPU_IDLE_WORKER_START, NULL, 0u); #define __ENCODE_KTRACE_INFO(reset, idle, all_suspend) \ @@ -5033,7 +5044,7 @@ static void gpu_idle_worker(struct work_struct *work) dev_warn(kbdev->dev, "Quit idle for failing to prevent gpu reset.\n"); KBASE_KTRACE_ADD(kbdev, SCHEDULER_GPU_IDLE_WORKER_END, NULL, __ENCODE_KTRACE_INFO(true, false, false)); - return; + goto exit; } kbase_debug_csf_fault_wait_completion(kbdev); rt_mutex_lock(&scheduler->lock); @@ -5042,7 +5053,7 @@ static void gpu_idle_worker(struct work_struct *work) if (unlikely(scheduler->state == SCHED_BUSY)) { rt_mutex_unlock(&scheduler->lock); kbase_reset_gpu_allow(kbdev); - return; + goto exit; } #endif @@ -5067,6 +5078,9 @@ static void gpu_idle_worker(struct work_struct *work) __ENCODE_KTRACE_INFO(false, scheduler_is_idle_suspendable, all_groups_suspended)); #undef __ENCODE_KTRACE_INFO + +exit: + atomic_dec(&scheduler->pending_gpu_idle_work); } static int scheduler_prepare(struct kbase_device *kbdev) @@ -5691,7 +5705,9 @@ exit_no_schedule_unlock: kbase_reset_gpu_allow(kbdev); } -static int suspend_active_queue_groups(struct kbase_device *kbdev, unsigned long *slot_mask) + +static int suspend_active_queue_groups(struct kbase_device *kbdev, unsigned long *slot_mask, + bool reset) { struct kbase_csf_scheduler *const scheduler = &kbdev->csf.scheduler; u32 num_groups = kbdev->csf.global_iface.group_num; @@ -5704,12 +5720,12 @@ static int suspend_active_queue_groups(struct kbase_device *kbdev, unsigned long struct kbase_queue_group *group = scheduler->csg_slots[slot_num].resident_group; if (group) { - suspend_queue_group(group); + suspend_queue_group(group); set_bit(slot_num, slot_mask); } } - ret = wait_csg_slots_suspend(kbdev, slot_mask); + ret = wait_csg_slots_suspend(kbdev, slot_mask); return ret; } @@ -5722,7 +5738,7 @@ static int suspend_active_queue_groups_on_reset(struct kbase_device *kbdev) rt_mutex_lock(&scheduler->lock); - ret = suspend_active_queue_groups(kbdev, slot_mask); + ret = suspend_active_queue_groups(kbdev, slot_mask, true); if (ret) { dev_warn( @@ -5864,9 +5880,9 @@ static void scheduler_inner_reset(struct kbase_device *kbdev) WARN_ON(kbase_csf_scheduler_get_nr_active_csgs(kbdev)); /* Cancel any potential queued delayed work(s) */ - cancel_work_sync(&kbdev->csf.scheduler.gpu_idle_work); cancel_tick_work(scheduler); cancel_tock_work(scheduler); + cancel_gpu_idle_work(scheduler); cancel_delayed_work_sync(&scheduler->ping_work); rt_mutex_lock(&scheduler->lock); @@ -5894,12 +5910,13 @@ static void scheduler_inner_reset(struct kbase_device *kbdev) void kbase_csf_scheduler_reset(struct kbase_device *kbdev) { struct kbase_context *kctx; - WARN_ON(!kbase_reset_gpu_is_active(kbdev)); KBASE_KTRACE_ADD(kbdev, SCHEDULER_RESET_START, NULL, 0u); - kbase_debug_csf_fault_wait_completion(kbdev); + if (kbase_reset_gpu_is_active(kbdev)) + kbase_debug_csf_fault_wait_completion(kbdev); + if (scheduler_handle_reset_in_protected_mode(kbdev) && !suspend_active_queue_groups_on_reset(kbdev)) { @@ -6488,8 +6505,8 @@ static void check_sync_update_in_sleep_mode(struct kbase_device *kbdev) * check_group_sync_update_worker() - Check the sync wait condition for all the * blocked queue groups * - * @work: Pointer to the context-specific work item for evaluating the wait - * condition for all the queue groups in idle_wait_groups list. + * @kctx: The context to evaluate the wait condition for all the queue groups + * in idle_wait_groups list. * * This function checks the gpu queues of all the groups present in both * idle_wait_groups list of a context and all on slot idle groups (if GPU @@ -6499,25 +6516,14 @@ static void check_sync_update_in_sleep_mode(struct kbase_device *kbdev) * runnable groups so that Scheduler can consider scheduling the group * in next tick or exit protected mode. */ -static void check_group_sync_update_worker(struct kthread_work *work) +static void check_group_sync_update_worker(struct kbase_context *kctx) { - struct kbase_context *const kctx = - container_of(work, struct kbase_context, csf.sched.sync_update_work); struct kbase_device *const kbdev = kctx->kbdev; struct kbase_csf_scheduler *const scheduler = &kbdev->csf.scheduler; bool sync_updated = false; rt_mutex_lock(&scheduler->lock); -#if IS_ENABLED(CONFIG_DEBUG_FS) - if (unlikely(scheduler->state == SCHED_BUSY)) { - kthread_queue_work(&kctx->csf.sched.sync_update_worker, - &kctx->csf.sched.sync_update_work); - rt_mutex_unlock(&scheduler->lock); - return; - } -#endif - KBASE_KTRACE_ADD(kbdev, SCHEDULER_GROUP_SYNC_UPDATE_WORKER_START, kctx, 0u); if (kctx->csf.sched.num_idle_wait_grps != 0) { struct kbase_queue_group *group, *temp; @@ -6558,8 +6564,7 @@ static enum kbase_csf_event_callback_action check_group_sync_update_cb(void *par KBASE_KTRACE_ADD(kctx->kbdev, SCHEDULER_GROUP_SYNC_UPDATE_EVENT, kctx, 0u); - kthread_queue_work(&kctx->csf.sched.sync_update_worker, - &kctx->csf.sched.sync_update_work); + kbase_csf_scheduler_enqueue_sync_update_work(kctx); return KBASE_CSF_EVENT_CALLBACK_KEEP; } @@ -6570,6 +6575,8 @@ int kbase_csf_scheduler_context_init(struct kbase_context *kctx) int err; struct kbase_device *kbdev = kctx->kbdev; + WARN_ON_ONCE(!kbdev->csf.scheduler.kthread_running); + #if IS_ENABLED(CONFIG_MALI_TRACE_POWER_GPU_WORK_PERIOD) err = gpu_metrics_ctx_init(kctx); if (err) @@ -6582,16 +6589,7 @@ int kbase_csf_scheduler_context_init(struct kbase_context *kctx) INIT_LIST_HEAD(&kctx->csf.sched.idle_wait_groups); - err = kbase_kthread_run_worker_rt(kctx->kbdev, &kctx->csf.sched.sync_update_worker, "csf_sync_update"); - if (err) { - dev_err(kctx->kbdev->dev, - "Failed to initialize scheduler context workqueue"); - err = -ENOMEM; - goto alloc_wq_failed; - } - - kthread_init_work(&kctx->csf.sched.sync_update_work, - check_group_sync_update_worker); + INIT_LIST_HEAD(&kctx->csf.sched.sync_update_work); kbase_csf_tiler_heap_reclaim_ctx_init(kctx); @@ -6605,8 +6603,6 @@ int kbase_csf_scheduler_context_init(struct kbase_context *kctx) return err; event_wait_add_failed: - kbase_destroy_kworker_stack(&kctx->csf.sched.sync_update_worker); -alloc_wq_failed: kbase_ctx_sched_remove_ctx(kctx); #if IS_ENABLED(CONFIG_MALI_TRACE_POWER_GPU_WORK_PERIOD) gpu_metrics_ctx_term(kctx); @@ -6617,8 +6613,10 @@ alloc_wq_failed: void kbase_csf_scheduler_context_term(struct kbase_context *kctx) { kbase_csf_event_wait_remove(kctx, check_group_sync_update_cb, kctx); - kthread_cancel_work_sync(&kctx->csf.sched.sync_update_work); - kbase_destroy_kworker_stack(&kctx->csf.sched.sync_update_worker); + + /* Drain a pending SYNC_UPDATE work if any */ + kbase_csf_scheduler_wait_for_kthread_pending_work(kctx->kbdev, + &kctx->csf.pending_sync_update); kbase_ctx_sched_remove_ctx(kctx); #if IS_ENABLED(CONFIG_MALI_TRACE_POWER_GPU_WORK_PERIOD) @@ -6626,53 +6624,157 @@ void kbase_csf_scheduler_context_term(struct kbase_context *kctx) #endif /* CONFIG_MALI_TRACE_POWER_GPU_WORK_PERIOD */ } +static void handle_pending_sync_update_works(struct kbase_csf_scheduler *scheduler) +{ + struct kbase_context *sync_update_ctx; + + if (atomic_cmpxchg(&scheduler->pending_sync_update_works, true, false) == false) + return; + + do { + unsigned long flags; + + spin_lock_irqsave(&scheduler->sync_update_work_ctxs_lock, flags); + sync_update_ctx = NULL; + if (!list_empty(&scheduler->sync_update_work_ctxs)) { + sync_update_ctx = list_first_entry(&scheduler->sync_update_work_ctxs, + struct kbase_context, + csf.sched.sync_update_work); + list_del_init(&sync_update_ctx->csf.sched.sync_update_work); + } + spin_unlock_irqrestore(&scheduler->sync_update_work_ctxs_lock, flags); + + if (sync_update_ctx != NULL) { + WARN_ON_ONCE(atomic_read(&sync_update_ctx->csf.pending_sync_update) == 0); + check_group_sync_update_worker(sync_update_ctx); + atomic_dec(&sync_update_ctx->csf.pending_sync_update); + } + } while (sync_update_ctx != NULL); +} + +static void handle_pending_protm_requests(struct kbase_csf_scheduler *scheduler) +{ + struct kbase_queue_group *protm_grp; + + if (atomic_cmpxchg(&scheduler->pending_protm_event_works, true, false) == false) + return; + + do { + unsigned long flags; + + spin_lock_irqsave(&scheduler->protm_event_work_grps_lock, flags); + protm_grp = NULL; + if (!list_empty(&scheduler->protm_event_work_grps)) { + protm_grp = list_first_entry(&scheduler->protm_event_work_grps, + struct kbase_queue_group, protm_event_work); + list_del_init(&protm_grp->protm_event_work); + } + spin_unlock_irqrestore(&scheduler->protm_event_work_grps_lock, flags); + + if (protm_grp != NULL) { + WARN_ON_ONCE(atomic_read(&protm_grp->pending_protm_event_work) == 0); + kbase_csf_process_protm_event_request(protm_grp); + atomic_dec(&protm_grp->pending_protm_event_work); + } + } while (protm_grp != NULL); +} + +static void handle_pending_kcpuq_commands(struct kbase_csf_scheduler *scheduler) +{ + struct kbase_kcpu_command_queue *kcpuq; + + if (atomic_cmpxchg(&scheduler->pending_kcpuq_works, true, false) == false) + return; + + do { + unsigned long flags; + + spin_lock_irqsave(&scheduler->kcpuq_work_queues_lock, flags); + kcpuq = NULL; + if (!list_empty(&scheduler->kcpuq_work_queues)) { + kcpuq = list_first_entry(&scheduler->kcpuq_work_queues, + struct kbase_kcpu_command_queue, high_prio_work); + list_del_init(&kcpuq->high_prio_work); + } + spin_unlock_irqrestore(&scheduler->kcpuq_work_queues_lock, flags); + + if (kcpuq != NULL) { + WARN_ON_ONCE(atomic_read(&kcpuq->pending_kick) == 0); + + mutex_lock(&kcpuq->lock); + kbase_csf_kcpu_queue_process(kcpuq, false); + mutex_unlock(&kcpuq->lock); + + atomic_dec(&kcpuq->pending_kick); + } + } while (kcpuq != NULL); +} + +static void handle_pending_queue_kicks(struct kbase_device *kbdev) +{ + struct kbase_csf_scheduler *const scheduler = &kbdev->csf.scheduler; + struct kbase_queue *queue; + + if (atomic_cmpxchg(&kbdev->csf.pending_gpuq_kicks, true, false) == false) + return; + + do { + u8 prio; + + spin_lock(&kbdev->csf.pending_gpuq_kick_queues_lock); + queue = NULL; + for (prio = 0; prio != KBASE_QUEUE_GROUP_PRIORITY_COUNT; ++prio) { + if (!list_empty(&kbdev->csf.pending_gpuq_kick_queues[prio])) { + queue = list_first_entry(&kbdev->csf.pending_gpuq_kick_queues[prio], + struct kbase_queue, pending_kick_link); + list_del_init(&queue->pending_kick_link); + break; + } + } + spin_unlock(&kbdev->csf.pending_gpuq_kick_queues_lock); + + if (queue != NULL) { + WARN_ONCE( + prio != queue->group_priority, + "Queue %pK has priority %u but instead its kick was handled at priority %u", + (void *)queue, queue->group_priority, prio); + WARN_ON_ONCE(atomic_read(&queue->pending_kick) == 0); + + kbase_csf_process_queue_kick(queue); + + /* Perform a scheduling tock for high-priority queue groups if + * required. + */ + BUILD_BUG_ON(KBASE_QUEUE_GROUP_PRIORITY_REALTIME != 0); + BUILD_BUG_ON(KBASE_QUEUE_GROUP_PRIORITY_HIGH != 1); + if ((prio <= KBASE_QUEUE_GROUP_PRIORITY_HIGH) && + atomic_read(&scheduler->pending_tock_work)) + schedule_on_tock(kbdev); + } + } while (queue != NULL); +} + static int kbase_csf_scheduler_kthread(void *data) { struct kbase_device *const kbdev = data; struct kbase_csf_scheduler *const scheduler = &kbdev->csf.scheduler; while (scheduler->kthread_running) { - struct kbase_queue *queue; - if (wait_for_completion_interruptible(&scheduler->kthread_signal) != 0) continue; reinit_completion(&scheduler->kthread_signal); - /* Iterate through queues with pending kicks */ - do { - u8 prio; - - spin_lock(&kbdev->csf.pending_gpuq_kicks_lock); - queue = NULL; - for (prio = 0; prio != KBASE_QUEUE_GROUP_PRIORITY_COUNT; ++prio) { - if (!list_empty(&kbdev->csf.pending_gpuq_kicks[prio])) { - queue = list_first_entry( - &kbdev->csf.pending_gpuq_kicks[prio], - struct kbase_queue, pending_kick_link); - list_del_init(&queue->pending_kick_link); - break; - } - } - spin_unlock(&kbdev->csf.pending_gpuq_kicks_lock); - - if (queue != NULL) { - WARN_ONCE( - prio != queue->group_priority, - "Queue %pK has priority %hhu but instead its kick was handled at priority %hhu", - (void *)queue, queue->group_priority, prio); - - kbase_csf_process_queue_kick(queue); + /* + * The order in which these requests are handled is based on + * how they would influence each other's decisions. As a + * result, the tick & tock requests must be handled after all + * other requests, but before the GPU IDLE work. + */ - /* Perform a scheduling tock for high-priority queue groups if - * required. - */ - BUILD_BUG_ON(KBASE_QUEUE_GROUP_PRIORITY_REALTIME != 0); - BUILD_BUG_ON(KBASE_QUEUE_GROUP_PRIORITY_HIGH != 1); - if ((prio <= KBASE_QUEUE_GROUP_PRIORITY_HIGH) && - atomic_read(&scheduler->pending_tock_work)) - schedule_on_tock(kbdev); - } - } while (queue != NULL); + handle_pending_sync_update_works(scheduler); + handle_pending_protm_requests(scheduler); + handle_pending_kcpuq_commands(scheduler); + handle_pending_queue_kicks(kbdev); /* Check if we need to perform a scheduling tick/tock. A tick * event shall override a tock event but not vice-versa. @@ -6684,6 +6786,10 @@ static int kbase_csf_scheduler_kthread(void *data) schedule_on_tock(kbdev); } + /* Drain pending GPU idle works */ + while (atomic_read(&scheduler->pending_gpu_idle_work) > 0) + gpu_idle_worker(kbdev); + dev_dbg(kbdev->dev, "Waking up for event after a scheduling iteration."); wake_up_all(&kbdev->csf.event_wait); @@ -6756,12 +6862,6 @@ int kbase_csf_scheduler_early_init(struct kbase_device *kbdev) atomic_set(&scheduler->timer_enabled, true); - scheduler->idle_wq = alloc_ordered_workqueue("csf_scheduler_gpu_idle_wq", WQ_HIGHPRI); - if (!scheduler->idle_wq) { - dev_err(kbdev->dev, "Failed to allocate GPU idle scheduler workqueue\n"); - return -ENOMEM; - } - INIT_DEFERRABLE_WORK(&scheduler->ping_work, firmware_aliveness_monitor); rt_mutex_init(&scheduler->lock); @@ -6779,18 +6879,23 @@ int kbase_csf_scheduler_early_init(struct kbase_device *kbdev) KBASE_KTRACE_ADD(kbdev, SCHED_SUSPENDED, NULL, scheduler->state); scheduler->csg_scheduling_period_ms = CSF_SCHEDULER_TIME_TICK_MS; scheduler_doorbell_init(kbdev); - - INIT_WORK(&scheduler->gpu_idle_work, gpu_idle_worker); - scheduler->fast_gpu_idle_handling = false; - atomic_set(&scheduler->gpu_no_longer_idle, false); - atomic_set(&scheduler->non_idle_offslot_grps, 0); - hrtimer_init(&scheduler->tick_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); scheduler->tick_timer.function = tick_timer_callback; - kbase_csf_tiler_heap_reclaim_mgr_init(kbdev); + atomic_set(&scheduler->pending_sync_update_works, false); + spin_lock_init(&scheduler->sync_update_work_ctxs_lock); + INIT_LIST_HEAD(&scheduler->sync_update_work_ctxs); + atomic_set(&scheduler->pending_protm_event_works, false); + spin_lock_init(&scheduler->protm_event_work_grps_lock); + INIT_LIST_HEAD(&scheduler->protm_event_work_grps); + atomic_set(&scheduler->pending_kcpuq_works, false); + spin_lock_init(&scheduler->kcpuq_work_queues_lock); + INIT_LIST_HEAD(&scheduler->kcpuq_work_queues); + atomic_set(&scheduler->pending_tick_work, false); + atomic_set(&scheduler->pending_tock_work, false); + atomic_set(&scheduler->pending_gpu_idle_work, 0); - return 0; + return kbase_csf_tiler_heap_reclaim_mgr_init(kbdev); } void kbase_csf_scheduler_term(struct kbase_device *kbdev) @@ -6811,7 +6916,6 @@ void kbase_csf_scheduler_term(struct kbase_device *kbdev) * to be active at the time of Driver unload. */ WARN_ON(kbase_csf_scheduler_get_nr_active_csgs(kbdev)); - flush_work(&kbdev->csf.scheduler.gpu_idle_work); rt_mutex_lock(&kbdev->csf.scheduler.lock); if (kbdev->csf.scheduler.state != SCHED_SUSPENDED) { @@ -6838,9 +6942,6 @@ void kbase_csf_scheduler_term(struct kbase_device *kbdev) void kbase_csf_scheduler_early_term(struct kbase_device *kbdev) { - if (kbdev->csf.scheduler.idle_wq) - destroy_workqueue(kbdev->csf.scheduler.idle_wq); - kbase_csf_tiler_heap_reclaim_mgr_term(kbdev); } @@ -7121,6 +7222,65 @@ int kbase_csf_scheduler_handle_runtime_suspend(struct kbase_device *kbdev) return 0; } +void kbase_csf_scheduler_enqueue_sync_update_work(struct kbase_context *kctx) +{ + struct kbase_csf_scheduler *const scheduler = &kctx->kbdev->csf.scheduler; + unsigned long flags; + + spin_lock_irqsave(&scheduler->sync_update_work_ctxs_lock, flags); + if (list_empty(&kctx->csf.sched.sync_update_work)) { + list_add_tail(&kctx->csf.sched.sync_update_work, &scheduler->sync_update_work_ctxs); + atomic_inc(&kctx->csf.pending_sync_update); + if (atomic_cmpxchg(&scheduler->pending_sync_update_works, false, true) == false) + complete(&scheduler->kthread_signal); + } + spin_unlock_irqrestore(&scheduler->sync_update_work_ctxs_lock, flags); +} + +void kbase_csf_scheduler_enqueue_protm_event_work(struct kbase_queue_group *group) +{ + struct kbase_context *const kctx = group->kctx; + struct kbase_csf_scheduler *const scheduler = &kctx->kbdev->csf.scheduler; + unsigned long flags; + + spin_lock_irqsave(&scheduler->protm_event_work_grps_lock, flags); + if (list_empty(&group->protm_event_work)) { + list_add_tail(&group->protm_event_work, &scheduler->protm_event_work_grps); + atomic_inc(&group->pending_protm_event_work); + if (atomic_cmpxchg(&scheduler->pending_protm_event_works, false, true) == false) + complete(&scheduler->kthread_signal); + } + spin_unlock_irqrestore(&scheduler->protm_event_work_grps_lock, flags); +} + +void kbase_csf_scheduler_enqueue_kcpuq_work(struct kbase_kcpu_command_queue *queue) +{ + struct kbase_csf_scheduler *const scheduler = &queue->kctx->kbdev->csf.scheduler; + unsigned long flags; + + spin_lock_irqsave(&scheduler->kcpuq_work_queues_lock, flags); + if (list_empty(&queue->high_prio_work)) { + list_add_tail(&queue->high_prio_work, &scheduler->kcpuq_work_queues); + atomic_inc(&queue->pending_kick); + if (atomic_cmpxchg(&scheduler->pending_kcpuq_works, false, true) == false) + complete(&scheduler->kthread_signal); + } + spin_unlock_irqrestore(&scheduler->kcpuq_work_queues_lock, flags); +} + +void kbase_csf_scheduler_wait_for_kthread_pending_work(struct kbase_device *kbdev, + atomic_t *pending) +{ + /* + * Signal kbase_csf_scheduler_kthread() to allow for the + * eventual completion of the current iteration. Once the work is + * done, the event_wait wait queue shall be signalled. + */ + + complete(&kbdev->csf.scheduler.kthread_signal); + wait_event(kbdev->csf.event_wait, atomic_read(pending) == 0); +} + void kbase_csf_scheduler_reval_idleness_post_sleep(struct kbase_device *kbdev) { u32 csg_nr; |