[linux-2.6-block.git] / drivers / gpu / drm / i915 / i915_active.c

/*
 * SPDX-License-Identifier: MIT
 *
 * Copyright © 2019 Intel Corporation
 */

#include <linux/debugobjects.h>

#include "gt/intel_engine_pm.h"

#include "i915_drv.h"
#include "i915_active.h"
#include "i915_globals.h"

#define BKL(ref) (&(ref)->i915->drm.struct_mutex)

/*
 * Active refs memory management
 *
 * To be more economical with memory, we reap all the i915_active trees as
 * they idle (when we know the active requests are inactive) and allocate the
 * nodes from a local slab cache to hopefully reduce the fragmentation.
 */
static struct i915_global_active {
	struct i915_global base;
	struct kmem_cache *slab_cache;
} global;

struct active_node {
	struct i915_active_request base;
	struct i915_active *ref;
	struct rb_node node;
	u64 timeline;
};

#if IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM) && IS_ENABLED(CONFIG_DEBUG_OBJECTS)

static void *active_debug_hint(void *addr)
{
	struct i915_active *ref = addr;

	return (void *)ref->active ?: (void *)ref->retire ?: (void *)ref;
}

static struct debug_obj_descr active_debug_desc = {
	.name = "i915_active",
	.debug_hint = active_debug_hint,
};

static void debug_active_init(struct i915_active *ref)
{
	debug_object_init(ref, &active_debug_desc);
}

static void debug_active_activate(struct i915_active *ref)
{
	debug_object_activate(ref, &active_debug_desc);
}

static void debug_active_deactivate(struct i915_active *ref)
{
	debug_object_deactivate(ref, &active_debug_desc);
}

static void debug_active_fini(struct i915_active *ref)
{
	debug_object_free(ref, &active_debug_desc);
}

static void debug_active_assert(struct i915_active *ref)
{
	debug_object_assert_init(ref, &active_debug_desc);
}

#else

static inline void debug_active_init(struct i915_active *ref) { }
static inline void debug_active_activate(struct i915_active *ref) { }
static inline void debug_active_deactivate(struct i915_active *ref) { }
static inline void debug_active_fini(struct i915_active *ref) { }
static inline void debug_active_assert(struct i915_active *ref) { }

#endif

static void
__active_retire(struct i915_active *ref)
{
	struct active_node *it, *n;
	struct rb_root root;
	bool retire = false;

	lockdep_assert_held(&ref->mutex);

	/* return the unused nodes to our slabcache -- flushing the allocator */
	if (atomic_dec_and_test(&ref->count)) {
		debug_active_deactivate(ref);
		root = ref->tree;
		ref->tree = RB_ROOT;
		ref->cache = NULL;
		retire = true;
	}

	mutex_unlock(&ref->mutex);
	if (!retire)
		return;

	ref->retire(ref);

	rbtree_postorder_for_each_entry_safe(it, n, &root, node) {
		GEM_BUG_ON(i915_active_request_isset(&it->base));
		kmem_cache_free(global.slab_cache, it);
	}
}

static void
active_retire(struct i915_active *ref)
{
	GEM_BUG_ON(!atomic_read(&ref->count));
	if (atomic_add_unless(&ref->count, -1, 1))
		return;

	/* One active may be flushed from inside the acquire of another */
	mutex_lock_nested(&ref->mutex, SINGLE_DEPTH_NESTING);
	__active_retire(ref);
}

static void
node_retire(struct i915_active_request *base, struct i915_request *rq)
{
	active_retire(container_of(base, struct active_node, base)->ref);
}

static struct i915_active_request *
active_instance(struct i915_active *ref, u64 idx)
{
	struct active_node *node, *prealloc;
	struct rb_node **p, *parent;

	/*
	 * We track the most recently used timeline to skip a rbtree search
	 * for the common case, under typical loads we never need the rbtree
	 * at all. We can reuse the last slot if it is empty, that is
	 * after the previous activity has been retired, or if it matches the
	 * current timeline.
	 */
	node = READ_ONCE(ref->cache);
	if (node && node->timeline == idx)
		return &node->base;

	/* Preallocate a replacement, just in case */
	prealloc = kmem_cache_alloc(global.slab_cache, GFP_KERNEL);
	if (!prealloc)
		return NULL;

	mutex_lock(&ref->mutex);
	GEM_BUG_ON(i915_active_is_idle(ref));

	parent = NULL;
	p = &ref->tree.rb_node;
	while (*p) {
		parent = *p;

		node = rb_entry(parent, struct active_node, node);
		if (node->timeline == idx) {
			kmem_cache_free(global.slab_cache, prealloc);
			goto out;
		}

		if (node->timeline < idx)
			p = &parent->rb_right;
		else
			p = &parent->rb_left;
	}

	node = prealloc;
	i915_active_request_init(&node->base, NULL, node_retire);
	node->ref = ref;
	node->timeline = idx;

	rb_link_node(&node->node, parent, p);
	rb_insert_color(&node->node, &ref->tree);

out:
	ref->cache = node;
	mutex_unlock(&ref->mutex);

	return &node->base;
}

void __i915_active_init(struct drm_i915_private *i915,
			struct i915_active *ref,
			int (*active)(struct i915_active *ref),
			void (*retire)(struct i915_active *ref),
			struct lock_class_key *key)
{
	debug_active_init(ref);

	ref->i915 = i915;
	ref->active = active;
	ref->retire = retire;
	ref->tree = RB_ROOT;
	ref->cache = NULL;
	init_llist_head(&ref->barriers);
	atomic_set(&ref->count, 0);
	__mutex_init(&ref->mutex, "i915_active", key);
}

int i915_active_ref(struct i915_active *ref,
		    u64 timeline,
		    struct i915_request *rq)
{
	struct i915_active_request *active;
	int err;

	/* Prevent reaping in case we malloc/wait while building the tree */
	err = i915_active_acquire(ref);
	if (err)
		return err;

	active = active_instance(ref, timeline);
	if (!active) {
		err = -ENOMEM;
		goto out;
	}

	if (!i915_active_request_isset(active))
		atomic_inc(&ref->count);
	__i915_active_request_set(active, rq);

out:
	i915_active_release(ref);
	return err;
}

int i915_active_acquire(struct i915_active *ref)
{
	int err;

	debug_active_assert(ref);
	if (atomic_add_unless(&ref->count, 1, 0))
		return 0;

	err = mutex_lock_interruptible(&ref->mutex);
	if (err)
		return err;

	if (!atomic_read(&ref->count) && ref->active)
		err = ref->active(ref);
	if (!err) {
		debug_active_activate(ref);
		atomic_inc(&ref->count);
	}

	mutex_unlock(&ref->mutex);

	return err;
}

void i915_active_release(struct i915_active *ref)
{
	debug_active_assert(ref);
	active_retire(ref);
}

int i915_active_wait(struct i915_active *ref)
{
	struct active_node *it, *n;
	int err;

	might_sleep();
	if (RB_EMPTY_ROOT(&ref->tree))
		return 0;

	err = mutex_lock_interruptible(&ref->mutex);
	if (err)
		return err;

	if (!atomic_add_unless(&ref->count, 1, 0)) {
		mutex_unlock(&ref->mutex);
		return 0;
	}

	rbtree_postorder_for_each_entry_safe(it, n, &ref->tree, node) {
		err = i915_active_request_retire(&it->base, BKL(ref));
		if (err)
			break;
	}

	__active_retire(ref);
	if (err)
		return err;

	if (!i915_active_is_idle(ref))
		return -EBUSY;

	return 0;
}

int i915_request_await_active_request(struct i915_request *rq,
				      struct i915_active_request *active)
{
	struct i915_request *barrier =
		i915_active_request_raw(active, &rq->i915->drm.struct_mutex);

	return barrier ? i915_request_await_dma_fence(rq, &barrier->fence) : 0;
}

int i915_request_await_active(struct i915_request *rq, struct i915_active *ref)
{
	struct active_node *it, *n;
	int err;

	if (RB_EMPTY_ROOT(&ref->tree))
		return 0;

	/* await allocates and so we need to avoid hitting the shrinker */
	err = i915_active_acquire(ref);
	if (err)
		return err;

	mutex_lock(&ref->mutex);
	rbtree_postorder_for_each_entry_safe(it, n, &ref->tree, node) {
		err = i915_request_await_active_request(rq, &it->base);
		if (err)
			break;
	}
	mutex_unlock(&ref->mutex);

	i915_active_release(ref);
	return err;
}

#if IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM)
void i915_active_fini(struct i915_active *ref)
{
	debug_active_fini(ref);
	GEM_BUG_ON(!RB_EMPTY_ROOT(&ref->tree));
	GEM_BUG_ON(atomic_read(&ref->count));
	mutex_destroy(&ref->mutex);
}
#endif

int i915_active_acquire_preallocate_barrier(struct i915_active *ref,
					    struct intel_engine_cs *engine)
{
	struct drm_i915_private *i915 = engine->i915;
	struct llist_node *pos, *next;
	unsigned long tmp;
	int err;

	GEM_BUG_ON(!engine->mask);
	for_each_engine_masked(engine, i915, engine->mask, tmp) {
		struct intel_context *kctx = engine->kernel_context;
		struct active_node *node;

		node = kmem_cache_alloc(global.slab_cache, GFP_KERNEL);
		if (unlikely(!node)) {
			err = -ENOMEM;
			goto unwind;
		}

		i915_active_request_init(&node->base,
					 (void *)engine, node_retire);
		node->timeline = kctx->ring->timeline->fence_context;
		node->ref = ref;
		atomic_inc(&ref->count);

		intel_engine_pm_get(engine);
		llist_add((struct llist_node *)&node->base.link,
			  &ref->barriers);
	}

	return 0;

unwind:
	llist_for_each_safe(pos, next, llist_del_all(&ref->barriers)) {
		struct active_node *node;

		node = container_of((struct list_head *)pos,
				    typeof(*node), base.link);
		engine = (void *)rcu_access_pointer(node->base.request);

		intel_engine_pm_put(engine);
		kmem_cache_free(global.slab_cache, node);
	}
	return err;
}

void i915_active_acquire_barrier(struct i915_active *ref)
{
	struct llist_node *pos, *next;

	GEM_BUG_ON(i915_active_is_idle(ref));

	mutex_lock_nested(&ref->mutex, SINGLE_DEPTH_NESTING);
	llist_for_each_safe(pos, next, llist_del_all(&ref->barriers)) {
		struct intel_engine_cs *engine;
		struct active_node *node;
		struct rb_node **p, *parent;

		node = container_of((struct list_head *)pos,
				    typeof(*node), base.link);

		engine = (void *)rcu_access_pointer(node->base.request);
		RCU_INIT_POINTER(node->base.request, ERR_PTR(-EAGAIN));

		parent = NULL;
		p = &ref->tree.rb_node;
		while (*p) {
			parent = *p;
			if (rb_entry(parent,
				     struct active_node,
				     node)->timeline < node->timeline)
				p = &parent->rb_right;
			else
				p = &parent->rb_left;
		}
		rb_link_node(&node->node, parent, p);
		rb_insert_color(&node->node, &ref->tree);

		llist_add((struct llist_node *)&node->base.link,
			  &engine->barrier_tasks);
		intel_engine_pm_put(engine);
	}
	mutex_unlock(&ref->mutex);
}

void i915_request_add_barriers(struct i915_request *rq)
{
	struct intel_engine_cs *engine = rq->engine;
	struct llist_node *node, *next;

	llist_for_each_safe(node, next, llist_del_all(&engine->barrier_tasks))
		list_add_tail((struct list_head *)node, &rq->active_list);
}

int i915_active_request_set(struct i915_active_request *active,
			    struct i915_request *rq)
{
	int err;

	/* Must maintain ordering wrt previous active requests */
	err = i915_request_await_active_request(rq, active);
	if (err)
		return err;

	__i915_active_request_set(active, rq);
	return 0;
}

void i915_active_retire_noop(struct i915_active_request *active,
			     struct i915_request *request)
{
	/* Space left intentionally blank */
}

#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
#include "selftests/i915_active.c"
#endif

static void i915_global_active_shrink(void)
{
	kmem_cache_shrink(global.slab_cache);
}

static void i915_global_active_exit(void)
{
	kmem_cache_destroy(global.slab_cache);
}

static struct i915_global_active global = { {
	.shrink = i915_global_active_shrink,
	.exit = i915_global_active_exit,
} };

int __init i915_global_active_init(void)
{
	global.slab_cache = KMEM_CACHE(active_node, SLAB_HWCACHE_ALIGN);
	if (!global.slab_cache)
		return -ENOMEM;

	i915_global_register(&global.base);
	return 0;
}
Commit	Line	Data
64d6c500 CW	1	/*
	2	* SPDX-License-Identifier: MIT
	3	*
	4	* Copyright © 2019 Intel Corporation
	5	*/
	6
5361db1a CW	7	#include <linux/debugobjects.h>
5361db1a CW	8
7009db14 CW	9	#include "gt/intel_engine_pm.h"
7009db14 CW	10
64d6c500 CW	11	#include "i915_drv.h"
64d6c500 CW	12	#include "i915_active.h"
103b76ee	13	#include "i915_globals.h"
64d6c500 CW	14
	15	#define BKL(ref) (&(ref)->i915->drm.struct_mutex)
	16
5f5c139d CW	17	/*
	18	* Active refs memory management
	19	*
	20	* To be more economical with memory, we reap all the i915_active trees as
	21	* they idle (when we know the active requests are inactive) and allocate the
	22	* nodes from a local slab cache to hopefully reduce the fragmentation.
	23	*/
	24	static struct i915_global_active {
103b76ee	25	struct i915_global base;
5f5c139d CW	26	struct kmem_cache *slab_cache;
	27	} global;
	28
64d6c500	29	struct active_node {
21950ee7	30	struct i915_active_request base;
64d6c500 CW	31	struct i915_active *ref;
	32	struct rb_node node;
	33	u64 timeline;
	34	};
	35
5361db1a CW	36	#if IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM) && IS_ENABLED(CONFIG_DEBUG_OBJECTS)
	37
	38	static void active_debug_hint(void addr)
	39	{
	40	struct i915_active *ref = addr;
	41
12c255b5	42	return (void )ref->active ?: (void )ref->retire ?: (void *)ref;
5361db1a CW	43	}
	44
	45	static struct debug_obj_descr active_debug_desc = {
	46	.name = "i915_active",
	47	.debug_hint = active_debug_hint,
	48	};
	49
	50	static void debug_active_init(struct i915_active *ref)
	51	{
	52	debug_object_init(ref, &active_debug_desc);
	53	}
	54
	55	static void debug_active_activate(struct i915_active *ref)
	56	{
	57	debug_object_activate(ref, &active_debug_desc);
	58	}
	59
	60	static void debug_active_deactivate(struct i915_active *ref)
	61	{
	62	debug_object_deactivate(ref, &active_debug_desc);
	63	}
	64
	65	static void debug_active_fini(struct i915_active *ref)
	66	{
	67	debug_object_free(ref, &active_debug_desc);
	68	}
	69
	70	static void debug_active_assert(struct i915_active *ref)
	71	{
	72	debug_object_assert_init(ref, &active_debug_desc);
	73	}
	74
	75	#else
	76
	77	static inline void debug_active_init(struct i915_active *ref) { }
	78	static inline void debug_active_activate(struct i915_active *ref) { }
	79	static inline void debug_active_deactivate(struct i915_active *ref) { }
	80	static inline void debug_active_fini(struct i915_active *ref) { }
	81	static inline void debug_active_assert(struct i915_active *ref) { }
	82
	83	#endif
	84
a42375af	85	static void
12c255b5	86	__active_retire(struct i915_active *ref)
a42375af CW	87	{
a42375af CW	88	struct active_node it, n;
12c255b5 CW	89	struct rb_root root;
	90	bool retire = false;
	91
	92	lockdep_assert_held(&ref->mutex);
	93
	94	/* return the unused nodes to our slabcache -- flushing the allocator */
	95	if (atomic_dec_and_test(&ref->count)) {
	96	debug_active_deactivate(ref);
	97	root = ref->tree;
	98	ref->tree = RB_ROOT;
	99	ref->cache = NULL;
	100	retire = true;
	101	}
a42375af	102
12c255b5 CW	103	mutex_unlock(&ref->mutex);
	104	if (!retire)
	105	return;
	106
	107	ref->retire(ref);
	108
	109	rbtree_postorder_for_each_entry_safe(it, n, &root, node) {
21950ee7	110	GEM_BUG_ON(i915_active_request_isset(&it->base));
5f5c139d	111	kmem_cache_free(global.slab_cache, it);
a42375af	112	}
a42375af CW	113	}
a42375af CW	114
64d6c500	115	static void
12c255b5	116	active_retire(struct i915_active *ref)
64d6c500	117	{
12c255b5 CW	118	GEM_BUG_ON(!atomic_read(&ref->count));
12c255b5 CW	119	if (atomic_add_unless(&ref->count, -1, 1))
a42375af CW	120	return;
a42375af CW	121
12c255b5 CW	122	/* One active may be flushed from inside the acquire of another */
	123	mutex_lock_nested(&ref->mutex, SINGLE_DEPTH_NESTING);
	124	__active_retire(ref);
64d6c500 CW	125	}
	126
	127	static void
21950ee7	128	node_retire(struct i915_active_request base, struct i915_request rq)
64d6c500	129	{
12c255b5	130	active_retire(container_of(base, struct active_node, base)->ref);
64d6c500 CW	131	}
64d6c500 CW	132
21950ee7	133	static struct i915_active_request *
64d6c500 CW	134	active_instance(struct i915_active *ref, u64 idx)
64d6c500 CW	135	{
12c255b5	136	struct active_node node, prealloc;
64d6c500	137	struct rb_node *p, parent;
64d6c500 CW	138
	139	/*
	140	* We track the most recently used timeline to skip a rbtree search
	141	* for the common case, under typical loads we never need the rbtree
	142	* at all. We can reuse the last slot if it is empty, that is
	143	* after the previous activity has been retired, or if it matches the
	144	* current timeline.
64d6c500	145	*/
12c255b5 CW	146	node = READ_ONCE(ref->cache);
	147	if (node && node->timeline == idx)
	148	return &node->base;
	149
	150	/* Preallocate a replacement, just in case */
	151	prealloc = kmem_cache_alloc(global.slab_cache, GFP_KERNEL);
	152	if (!prealloc)
	153	return NULL;
64d6c500	154
12c255b5 CW	155	mutex_lock(&ref->mutex);
12c255b5 CW	156	GEM_BUG_ON(i915_active_is_idle(ref));
64d6c500 CW	157
	158	parent = NULL;
	159	p = &ref->tree.rb_node;
	160	while (*p) {
	161	parent = *p;
	162
	163	node = rb_entry(parent, struct active_node, node);
12c255b5 CW	164	if (node->timeline == idx) {
	165	kmem_cache_free(global.slab_cache, prealloc);
	166	goto out;
	167	}
64d6c500 CW	168
	169	if (node->timeline < idx)
	170	p = &parent->rb_right;
	171	else
	172	p = &parent->rb_left;
	173	}
	174
12c255b5	175	node = prealloc;
21950ee7	176	i915_active_request_init(&node->base, NULL, node_retire);
64d6c500 CW	177	node->ref = ref;
	178	node->timeline = idx;
	179
	180	rb_link_node(&node->node, parent, p);
	181	rb_insert_color(&node->node, &ref->tree);
	182
64d6c500	183	out:
12c255b5 CW	184	ref->cache = node;
	185	mutex_unlock(&ref->mutex);
	186
	187	return &node->base;
64d6c500 CW	188	}
64d6c500 CW	189
12c255b5 CW	190	void __i915_active_init(struct drm_i915_private *i915,
	191	struct i915_active *ref,
	192	int (active)(struct i915_active ref),
	193	void (retire)(struct i915_active ref),
	194	struct lock_class_key *key)
64d6c500	195	{
5361db1a CW	196	debug_active_init(ref);
5361db1a CW	197
64d6c500	198	ref->i915 = i915;
12c255b5	199	ref->active = active;
64d6c500 CW	200	ref->retire = retire;
64d6c500 CW	201	ref->tree = RB_ROOT;
12c255b5	202	ref->cache = NULL;
ce476c80	203	init_llist_head(&ref->barriers);
12c255b5 CW	204	atomic_set(&ref->count, 0);
12c255b5 CW	205	__mutex_init(&ref->mutex, "i915_active", key);
64d6c500 CW	206	}
	207
	208	int i915_active_ref(struct i915_active *ref,
	209	u64 timeline,
	210	struct i915_request *rq)
	211	{
21950ee7	212	struct i915_active_request *active;
12c255b5	213	int err;
312c4ba1 CW	214
312c4ba1 CW	215	/* Prevent reaping in case we malloc/wait while building the tree */
12c255b5 CW	216	err = i915_active_acquire(ref);
	217	if (err)
	218	return err;
64d6c500 CW	219
64d6c500 CW	220	active = active_instance(ref, timeline);
12c255b5 CW	221	if (!active) {
12c255b5 CW	222	err = -ENOMEM;
312c4ba1 CW	223	goto out;
312c4ba1 CW	224	}
64d6c500	225
21950ee7	226	if (!i915_active_request_isset(active))
12c255b5	227	atomic_inc(&ref->count);
21950ee7	228	__i915_active_request_set(active, rq);
64d6c500	229
312c4ba1 CW	230	out:
	231	i915_active_release(ref);
	232	return err;
64d6c500 CW	233	}
64d6c500 CW	234
12c255b5	235	int i915_active_acquire(struct i915_active *ref)
64d6c500	236	{
12c255b5 CW	237	int err;
12c255b5 CW	238
5361db1a	239	debug_active_assert(ref);
12c255b5 CW	240	if (atomic_add_unless(&ref->count, 1, 0))
12c255b5 CW	241	return 0;
5361db1a	242
12c255b5 CW	243	err = mutex_lock_interruptible(&ref->mutex);
	244	if (err)
	245	return err;
5361db1a	246
12c255b5 CW	247	if (!atomic_read(&ref->count) && ref->active)
	248	err = ref->active(ref);
	249	if (!err) {
	250	debug_active_activate(ref);
	251	atomic_inc(&ref->count);
	252	}
	253
	254	mutex_unlock(&ref->mutex);
	255
	256	return err;
64d6c500 CW	257	}
	258
	259	void i915_active_release(struct i915_active *ref)
	260	{
5361db1a	261	debug_active_assert(ref);
12c255b5	262	active_retire(ref);
64d6c500 CW	263	}
	264
	265	int i915_active_wait(struct i915_active *ref)
	266	{
	267	struct active_node it, n;
12c255b5	268	int err;
64d6c500	269
12c255b5 CW	270	might_sleep();
	271	if (RB_EMPTY_ROOT(&ref->tree))
	272	return 0;
64d6c500	273
12c255b5 CW	274	err = mutex_lock_interruptible(&ref->mutex);
	275	if (err)
	276	return err;
	277
	278	if (!atomic_add_unless(&ref->count, 1, 0)) {
	279	mutex_unlock(&ref->mutex);
	280	return 0;
	281	}
64d6c500 CW	282
64d6c500 CW	283	rbtree_postorder_for_each_entry_safe(it, n, &ref->tree, node) {
12c255b5 CW	284	err = i915_active_request_retire(&it->base, BKL(ref));
12c255b5 CW	285	if (err)
64d6c500 CW	286	break;
	287	}
	288
12c255b5	289	__active_retire(ref);
afd1bcd4 CW	290	if (err)
	291	return err;
	292
	293	if (!i915_active_is_idle(ref))
	294	return -EBUSY;
	295
	296	return 0;
64d6c500 CW	297	}
64d6c500 CW	298
21950ee7 CW	299	int i915_request_await_active_request(struct i915_request *rq,
21950ee7 CW	300	struct i915_active_request *active)
64d6c500 CW	301	{
64d6c500 CW	302	struct i915_request *barrier =
21950ee7	303	i915_active_request_raw(active, &rq->i915->drm.struct_mutex);
64d6c500 CW	304
	305	return barrier ? i915_request_await_dma_fence(rq, &barrier->fence) : 0;
	306	}
	307
	308	int i915_request_await_active(struct i915_request rq, struct i915_active ref)
	309	{
	310	struct active_node it, n;
12c255b5	311	int err;
64d6c500	312
12c255b5 CW	313	if (RB_EMPTY_ROOT(&ref->tree))
12c255b5 CW	314	return 0;
312c4ba1	315
12c255b5 CW	316	/* await allocates and so we need to avoid hitting the shrinker */
12c255b5 CW	317	err = i915_active_acquire(ref);
312c4ba1	318	if (err)
12c255b5	319	return err;
64d6c500	320
12c255b5	321	mutex_lock(&ref->mutex);
64d6c500	322	rbtree_postorder_for_each_entry_safe(it, n, &ref->tree, node) {
312c4ba1 CW	323	err = i915_request_await_active_request(rq, &it->base);
312c4ba1 CW	324	if (err)
12c255b5	325	break;
64d6c500	326	}
12c255b5	327	mutex_unlock(&ref->mutex);
64d6c500	328
312c4ba1 CW	329	i915_active_release(ref);
312c4ba1 CW	330	return err;
64d6c500 CW	331	}
64d6c500 CW	332
a42375af	333	#if IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM)
64d6c500 CW	334	void i915_active_fini(struct i915_active *ref)
64d6c500 CW	335	{
5361db1a	336	debug_active_fini(ref);
a42375af	337	GEM_BUG_ON(!RB_EMPTY_ROOT(&ref->tree));
12c255b5 CW	338	GEM_BUG_ON(atomic_read(&ref->count));
12c255b5 CW	339	mutex_destroy(&ref->mutex);
64d6c500	340	}
a42375af	341	#endif
64d6c500	342
ce476c80 CW	343	int i915_active_acquire_preallocate_barrier(struct i915_active *ref,
	344	struct intel_engine_cs *engine)
	345	{
	346	struct drm_i915_private *i915 = engine->i915;
7009db14	347	struct llist_node pos, next;
ce476c80	348	unsigned long tmp;
7009db14	349	int err;
ce476c80 CW	350
	351	GEM_BUG_ON(!engine->mask);
	352	for_each_engine_masked(engine, i915, engine->mask, tmp) {
	353	struct intel_context *kctx = engine->kernel_context;
	354	struct active_node *node;
	355
	356	node = kmem_cache_alloc(global.slab_cache, GFP_KERNEL);
	357	if (unlikely(!node)) {
	358	err = -ENOMEM;
7009db14	359	goto unwind;
ce476c80 CW	360	}
	361
	362	i915_active_request_init(&node->base,
	363	(void *)engine, node_retire);
	364	node->timeline = kctx->ring->timeline->fence_context;
	365	node->ref = ref;
12c255b5	366	atomic_inc(&ref->count);
ce476c80	367
7009db14	368	intel_engine_pm_get(engine);
ce476c80 CW	369	llist_add((struct llist_node *)&node->base.link,
	370	&ref->barriers);
	371	}
	372
7009db14 CW	373	return 0;
	374
	375	unwind:
	376	llist_for_each_safe(pos, next, llist_del_all(&ref->barriers)) {
	377	struct active_node *node;
	378
	379	node = container_of((struct list_head *)pos,
	380	typeof(*node), base.link);
	381	engine = (void *)rcu_access_pointer(node->base.request);
	382
	383	intel_engine_pm_put(engine);
	384	kmem_cache_free(global.slab_cache, node);
	385	}
ce476c80 CW	386	return err;
	387	}
	388
	389	void i915_active_acquire_barrier(struct i915_active *ref)
	390	{
	391	struct llist_node pos, next;
	392
12c255b5	393	GEM_BUG_ON(i915_active_is_idle(ref));
ce476c80	394
12c255b5	395	mutex_lock_nested(&ref->mutex, SINGLE_DEPTH_NESTING);
ce476c80 CW	396	llist_for_each_safe(pos, next, llist_del_all(&ref->barriers)) {
	397	struct intel_engine_cs *engine;
	398	struct active_node *node;
	399	struct rb_node *p, parent;
	400
	401	node = container_of((struct list_head *)pos,
	402	typeof(*node), base.link);
	403
	404	engine = (void *)rcu_access_pointer(node->base.request);
	405	RCU_INIT_POINTER(node->base.request, ERR_PTR(-EAGAIN));
	406
	407	parent = NULL;
	408	p = &ref->tree.rb_node;
	409	while (*p) {
	410	parent = *p;
	411	if (rb_entry(parent,
	412	struct active_node,
	413	node)->timeline < node->timeline)
	414	p = &parent->rb_right;
	415	else
	416	p = &parent->rb_left;
	417	}
	418	rb_link_node(&node->node, parent, p);
	419	rb_insert_color(&node->node, &ref->tree);
	420
	421	llist_add((struct llist_node *)&node->base.link,
	422	&engine->barrier_tasks);
7009db14	423	intel_engine_pm_put(engine);
ce476c80	424	}
12c255b5	425	mutex_unlock(&ref->mutex);
ce476c80 CW	426	}
	427
	428	void i915_request_add_barriers(struct i915_request *rq)
	429	{
	430	struct intel_engine_cs *engine = rq->engine;
	431	struct llist_node node, next;
	432
	433	llist_for_each_safe(node, next, llist_del_all(&engine->barrier_tasks))
	434	list_add_tail((struct list_head *)node, &rq->active_list);
	435	}
	436
21950ee7 CW	437	int i915_active_request_set(struct i915_active_request *active,
	438	struct i915_request *rq)
	439	{
	440	int err;
	441
	442	/* Must maintain ordering wrt previous active requests */
	443	err = i915_request_await_active_request(rq, active);
	444	if (err)
	445	return err;
	446
	447	__i915_active_request_set(active, rq);
	448	return 0;
	449	}
	450
	451	void i915_active_retire_noop(struct i915_active_request *active,
	452	struct i915_request *request)
	453	{
	454	/* Space left intentionally blank */
	455	}
	456
64d6c500 CW	457	#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
	458	#include "selftests/i915_active.c"
	459	#endif
5f5c139d	460
103b76ee	461	static void i915_global_active_shrink(void)
5f5c139d	462	{
103b76ee	463	kmem_cache_shrink(global.slab_cache);
5f5c139d CW	464	}
5f5c139d CW	465
103b76ee	466	static void i915_global_active_exit(void)
32eb6bcf	467	{
103b76ee	468	kmem_cache_destroy(global.slab_cache);
32eb6bcf CW	469	}
32eb6bcf CW	470
103b76ee CW	471	static struct i915_global_active global = { {
	472	.shrink = i915_global_active_shrink,
	473	.exit = i915_global_active_exit,
	474	} };
	475
	476	int __init i915_global_active_init(void)
5f5c139d	477	{
103b76ee CW	478	global.slab_cache = KMEM_CACHE(active_node, SLAB_HWCACHE_ALIGN);
	479	if (!global.slab_cache)
	480	return -ENOMEM;
	481
	482	i915_global_register(&global.base);
	483	return 0;
5f5c139d	484	}