2 * SPDX-License-Identifier: MIT
4 * Copyright © 2018 Intel Corporation
7 #include <linux/mutex.h>
10 #include "i915_globals.h"
11 #include "i915_request.h"
12 #include "i915_scheduler.h"
14 static struct i915_global_scheduler {
15 struct i915_global base;
16 struct kmem_cache *slab_dependencies;
17 struct kmem_cache *slab_priorities;
20 static DEFINE_SPINLOCK(schedule_lock);
22 static const struct i915_request *
23 node_to_request(const struct i915_sched_node *node)
25 return container_of(node, const struct i915_request, sched);
28 static inline bool node_started(const struct i915_sched_node *node)
30 return i915_request_started(node_to_request(node));
33 static inline bool node_signaled(const struct i915_sched_node *node)
35 return i915_request_completed(node_to_request(node));
38 static inline struct i915_priolist *to_priolist(struct rb_node *rb)
40 return rb_entry(rb, struct i915_priolist, node);
43 static void assert_priolists(struct intel_engine_execlists * const execlists)
48 if (!IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM))
51 GEM_BUG_ON(rb_first_cached(&execlists->queue) !=
52 rb_first(&execlists->queue.rb_root));
54 last_prio = (INT_MAX >> I915_USER_PRIORITY_SHIFT) + 1;
55 for (rb = rb_first_cached(&execlists->queue); rb; rb = rb_next(rb)) {
56 const struct i915_priolist *p = to_priolist(rb);
58 GEM_BUG_ON(p->priority >= last_prio);
59 last_prio = p->priority;
62 for (i = 0; i < ARRAY_SIZE(p->requests); i++) {
63 if (list_empty(&p->requests[i]))
66 GEM_BUG_ON(!(p->used & BIT(i)));
72 i915_sched_lookup_priolist(struct intel_engine_cs *engine, int prio)
74 struct intel_engine_execlists * const execlists = &engine->execlists;
75 struct i915_priolist *p;
76 struct rb_node **parent, *rb;
80 lockdep_assert_held(&engine->active.lock);
81 assert_priolists(execlists);
83 /* buckets sorted from highest [in slot 0] to lowest priority */
84 idx = I915_PRIORITY_COUNT - (prio & I915_PRIORITY_MASK) - 1;
85 prio >>= I915_USER_PRIORITY_SHIFT;
86 if (unlikely(execlists->no_priolist))
87 prio = I915_PRIORITY_NORMAL;
90 /* most positive priority is scheduled first, equal priorities fifo */
92 parent = &execlists->queue.rb_root.rb_node;
96 if (prio > p->priority) {
97 parent = &rb->rb_left;
98 } else if (prio < p->priority) {
99 parent = &rb->rb_right;
106 if (prio == I915_PRIORITY_NORMAL) {
107 p = &execlists->default_priolist;
109 p = kmem_cache_alloc(global.slab_priorities, GFP_ATOMIC);
110 /* Convert an allocation failure to a priority bump */
112 prio = I915_PRIORITY_NORMAL; /* recurses just once */
114 /* To maintain ordering with all rendering, after an
115 * allocation failure we have to disable all scheduling.
116 * Requests will then be executed in fifo, and schedule
117 * will ensure that dependencies are emitted in fifo.
118 * There will be still some reordering with existing
119 * requests, so if userspace lied about their
120 * dependencies that reordering may be visible.
122 execlists->no_priolist = true;
128 for (i = 0; i < ARRAY_SIZE(p->requests); i++)
129 INIT_LIST_HEAD(&p->requests[i]);
130 rb_link_node(&p->node, rb, parent);
131 rb_insert_color_cached(&p->node, &execlists->queue, first);
136 return &p->requests[idx];
139 void __i915_priolist_free(struct i915_priolist *p)
141 kmem_cache_free(global.slab_priorities, p);
145 struct list_head *priolist;
148 static struct intel_engine_cs *
149 sched_lock_engine(const struct i915_sched_node *node,
150 struct intel_engine_cs *locked,
151 struct sched_cache *cache)
153 const struct i915_request *rq = node_to_request(node);
154 struct intel_engine_cs *engine;
159 * Virtual engines complicate acquiring the engine timeline lock,
160 * as their rq->engine pointer is not stable until under that
161 * engine lock. The simple ploy we use is to take the lock then
162 * check that the rq still belongs to the newly locked engine.
164 while (locked != (engine = READ_ONCE(rq->engine))) {
165 spin_unlock(&locked->active.lock);
166 memset(cache, 0, sizeof(*cache));
167 spin_lock(&engine->active.lock);
171 GEM_BUG_ON(locked != engine);
175 static inline int rq_prio(const struct i915_request *rq)
177 return rq->sched.attr.priority | __NO_PREEMPTION;
180 static inline bool need_preempt(int prio, int active)
183 * Allow preemption of low -> normal -> high, but we do
184 * not allow low priority tasks to preempt other low priority
185 * tasks under the impression that latency for low priority
186 * tasks does not matter (as much as background throughput),
189 return prio >= max(I915_PRIORITY_NORMAL, active);
192 static void kick_submission(struct intel_engine_cs *engine,
193 const struct i915_request *rq,
196 const struct i915_request *inflight;
199 * We only need to kick the tasklet once for the high priority
200 * new context we add into the queue.
202 if (prio <= engine->execlists.queue_priority_hint)
207 /* Nothing currently active? We're overdue for a submission! */
208 inflight = execlists_active(&engine->execlists);
213 * If we are already the currently executing context, don't
214 * bother evaluating if we should preempt ourselves, or if
215 * we expect nothing to change as a result of running the
216 * tasklet, i.e. we have not change the priority queue
217 * sufficiently to oust the running context.
219 if (inflight->hw_context == rq->hw_context)
222 engine->execlists.queue_priority_hint = prio;
223 if (need_preempt(prio, rq_prio(inflight)))
224 tasklet_hi_schedule(&engine->execlists.tasklet);
230 static void __i915_schedule(struct i915_sched_node *node,
231 const struct i915_sched_attr *attr)
233 struct intel_engine_cs *engine;
234 struct i915_dependency *dep, *p;
235 struct i915_dependency stack;
236 const int prio = attr->priority;
237 struct sched_cache cache;
240 /* Needed in order to use the temporary link inside i915_dependency */
241 lockdep_assert_held(&schedule_lock);
242 GEM_BUG_ON(prio == I915_PRIORITY_INVALID);
244 if (prio <= READ_ONCE(node->attr.priority))
247 if (node_signaled(node))
250 stack.signaler = node;
251 list_add(&stack.dfs_link, &dfs);
254 * Recursively bump all dependent priorities to match the new request.
256 * A naive approach would be to use recursion:
257 * static void update_priorities(struct i915_sched_node *node, prio) {
258 * list_for_each_entry(dep, &node->signalers_list, signal_link)
259 * update_priorities(dep->signal, prio)
260 * queue_request(node);
262 * but that may have unlimited recursion depth and so runs a very
263 * real risk of overunning the kernel stack. Instead, we build
264 * a flat list of all dependencies starting with the current request.
265 * As we walk the list of dependencies, we add all of its dependencies
266 * to the end of the list (this may include an already visited
267 * request) and continue to walk onwards onto the new dependencies. The
268 * end result is a topological list of requests in reverse order, the
269 * last element in the list is the request we must execute first.
271 list_for_each_entry(dep, &dfs, dfs_link) {
272 struct i915_sched_node *node = dep->signaler;
274 /* If we are already flying, we know we have no signalers */
275 if (node_started(node))
279 * Within an engine, there can be no cycle, but we may
280 * refer to the same dependency chain multiple times
281 * (redundant dependencies are not eliminated) and across
284 list_for_each_entry(p, &node->signalers_list, signal_link) {
285 GEM_BUG_ON(p == dep); /* no cycles! */
287 if (node_signaled(p->signaler))
290 if (prio > READ_ONCE(p->signaler->attr.priority))
291 list_move_tail(&p->dfs_link, &dfs);
296 * If we didn't need to bump any existing priorities, and we haven't
297 * yet submitted this request (i.e. there is no potential race with
298 * execlists_submit_request()), we can set our own priority and skip
299 * acquiring the engine locks.
301 if (node->attr.priority == I915_PRIORITY_INVALID) {
302 GEM_BUG_ON(!list_empty(&node->link));
305 if (stack.dfs_link.next == stack.dfs_link.prev)
308 __list_del_entry(&stack.dfs_link);
311 memset(&cache, 0, sizeof(cache));
312 engine = node_to_request(node)->engine;
313 spin_lock(&engine->active.lock);
315 /* Fifo and depth-first replacement ensure our deps execute before us */
316 engine = sched_lock_engine(node, engine, &cache);
317 list_for_each_entry_safe_reverse(dep, p, &dfs, dfs_link) {
318 INIT_LIST_HEAD(&dep->dfs_link);
320 node = dep->signaler;
321 engine = sched_lock_engine(node, engine, &cache);
322 lockdep_assert_held(&engine->active.lock);
324 /* Recheck after acquiring the engine->timeline.lock */
325 if (prio <= node->attr.priority || node_signaled(node))
328 GEM_BUG_ON(node_to_request(node)->engine != engine);
330 node->attr.priority = prio;
332 if (list_empty(&node->link)) {
334 * If the request is not in the priolist queue because
335 * it is not yet runnable, then it doesn't contribute
336 * to our preemption decisions. On the other hand,
337 * if the request is on the HW, it too is not in the
338 * queue; but in that case we may still need to reorder
339 * the inflight requests.
344 if (!intel_engine_is_virtual(engine) &&
345 !i915_request_is_active(node_to_request(node))) {
348 i915_sched_lookup_priolist(engine,
350 list_move_tail(&node->link, cache.priolist);
353 /* Defer (tasklet) submission until after all of our updates. */
354 kick_submission(engine, node_to_request(node), prio);
357 spin_unlock(&engine->active.lock);
360 void i915_schedule(struct i915_request *rq, const struct i915_sched_attr *attr)
362 spin_lock_irq(&schedule_lock);
363 __i915_schedule(&rq->sched, attr);
364 spin_unlock_irq(&schedule_lock);
367 static void __bump_priority(struct i915_sched_node *node, unsigned int bump)
369 struct i915_sched_attr attr = node->attr;
371 attr.priority |= bump;
372 __i915_schedule(node, &attr);
375 void i915_schedule_bump_priority(struct i915_request *rq, unsigned int bump)
379 GEM_BUG_ON(bump & ~I915_PRIORITY_MASK);
380 if (READ_ONCE(rq->sched.attr.priority) & bump)
383 spin_lock_irqsave(&schedule_lock, flags);
384 __bump_priority(&rq->sched, bump);
385 spin_unlock_irqrestore(&schedule_lock, flags);
388 void i915_sched_node_init(struct i915_sched_node *node)
390 INIT_LIST_HEAD(&node->signalers_list);
391 INIT_LIST_HEAD(&node->waiters_list);
392 INIT_LIST_HEAD(&node->link);
393 node->attr.priority = I915_PRIORITY_INVALID;
394 node->semaphores = 0;
398 static struct i915_dependency *
399 i915_dependency_alloc(void)
401 return kmem_cache_alloc(global.slab_dependencies, GFP_KERNEL);
405 i915_dependency_free(struct i915_dependency *dep)
407 kmem_cache_free(global.slab_dependencies, dep);
410 bool __i915_sched_node_add_dependency(struct i915_sched_node *node,
411 struct i915_sched_node *signal,
412 struct i915_dependency *dep,
417 spin_lock_irq(&schedule_lock);
419 if (!node_signaled(signal)) {
420 INIT_LIST_HEAD(&dep->dfs_link);
421 dep->signaler = signal;
425 /* Keep track of whether anyone on this chain has a semaphore */
426 if (signal->flags & I915_SCHED_HAS_SEMAPHORE_CHAIN &&
427 !node_started(signal))
428 node->flags |= I915_SCHED_HAS_SEMAPHORE_CHAIN;
430 /* All set, now publish. Beware the lockless walkers. */
431 list_add(&dep->signal_link, &node->signalers_list);
432 list_add_rcu(&dep->wait_link, &signal->waiters_list);
435 * As we do not allow WAIT to preempt inflight requests,
436 * once we have executed a request, along with triggering
437 * any execution callbacks, we must preserve its ordering
438 * within the non-preemptible FIFO.
440 BUILD_BUG_ON(__NO_PREEMPTION & ~I915_PRIORITY_MASK);
441 if (flags & I915_DEPENDENCY_EXTERNAL)
442 __bump_priority(signal, __NO_PREEMPTION);
447 spin_unlock_irq(&schedule_lock);
452 int i915_sched_node_add_dependency(struct i915_sched_node *node,
453 struct i915_sched_node *signal)
455 struct i915_dependency *dep;
457 dep = i915_dependency_alloc();
461 if (!__i915_sched_node_add_dependency(node, signal, dep,
462 I915_DEPENDENCY_EXTERNAL |
463 I915_DEPENDENCY_ALLOC))
464 i915_dependency_free(dep);
469 void i915_sched_node_fini(struct i915_sched_node *node)
471 struct i915_dependency *dep, *tmp;
473 spin_lock_irq(&schedule_lock);
476 * Everyone we depended upon (the fences we wait to be signaled)
477 * should retire before us and remove themselves from our list.
478 * However, retirement is run independently on each timeline and
479 * so we may be called out-of-order.
481 list_for_each_entry_safe(dep, tmp, &node->signalers_list, signal_link) {
482 GEM_BUG_ON(!node_signaled(dep->signaler));
483 GEM_BUG_ON(!list_empty(&dep->dfs_link));
485 list_del(&dep->wait_link);
486 if (dep->flags & I915_DEPENDENCY_ALLOC)
487 i915_dependency_free(dep);
490 /* Remove ourselves from everyone who depends upon us */
491 list_for_each_entry_safe(dep, tmp, &node->waiters_list, wait_link) {
492 GEM_BUG_ON(dep->signaler != node);
493 GEM_BUG_ON(!list_empty(&dep->dfs_link));
495 list_del(&dep->signal_link);
496 if (dep->flags & I915_DEPENDENCY_ALLOC)
497 i915_dependency_free(dep);
500 spin_unlock_irq(&schedule_lock);
503 static void i915_global_scheduler_shrink(void)
505 kmem_cache_shrink(global.slab_dependencies);
506 kmem_cache_shrink(global.slab_priorities);
509 static void i915_global_scheduler_exit(void)
511 kmem_cache_destroy(global.slab_dependencies);
512 kmem_cache_destroy(global.slab_priorities);
515 static struct i915_global_scheduler global = { {
516 .shrink = i915_global_scheduler_shrink,
517 .exit = i915_global_scheduler_exit,
520 int __init i915_global_scheduler_init(void)
522 global.slab_dependencies = KMEM_CACHE(i915_dependency,
524 if (!global.slab_dependencies)
527 global.slab_priorities = KMEM_CACHE(i915_priolist,
529 if (!global.slab_priorities)
532 i915_global_register(&global.base);
536 kmem_cache_destroy(global.slab_priorities);