[linux-block.git] / drivers / gpu / drm / i915 / gem / i915_gem_client_blt.c

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

#include "i915_drv.h"
#include "gt/intel_context.h"
#include "gt/intel_engine_pm.h"
#include "i915_gem_client_blt.h"
#include "i915_gem_object_blt.h"

struct i915_sleeve {
	struct i915_vma *vma;
	struct drm_i915_gem_object *obj;
	struct sg_table *pages;
	struct i915_page_sizes page_sizes;
};

static int vma_set_pages(struct i915_vma *vma)
{
	struct i915_sleeve *sleeve = vma->private;

	vma->pages = sleeve->pages;
	vma->page_sizes = sleeve->page_sizes;

	return 0;
}

static void vma_clear_pages(struct i915_vma *vma)
{
	GEM_BUG_ON(!vma->pages);
	vma->pages = NULL;
}

static void vma_bind(struct i915_address_space *vm,
		     struct i915_vm_pt_stash *stash,
		     struct i915_vma *vma,
		     enum i915_cache_level cache_level,
		     u32 flags)
{
	vm->vma_ops.bind_vma(vm, stash, vma, cache_level, flags);
}

static void vma_unbind(struct i915_address_space *vm, struct i915_vma *vma)
{
	vm->vma_ops.unbind_vma(vm, vma);
}

static const struct i915_vma_ops proxy_vma_ops = {
	.set_pages = vma_set_pages,
	.clear_pages = vma_clear_pages,
	.bind_vma = vma_bind,
	.unbind_vma = vma_unbind,
};

static struct i915_sleeve *create_sleeve(struct i915_address_space *vm,
					 struct drm_i915_gem_object *obj,
					 struct sg_table *pages,
					 struct i915_page_sizes *page_sizes)
{
	struct i915_sleeve *sleeve;
	struct i915_vma *vma;
	int err;

	sleeve = kzalloc(sizeof(*sleeve), GFP_KERNEL);
	if (!sleeve)
		return ERR_PTR(-ENOMEM);

	vma = i915_vma_instance(obj, vm, NULL);
	if (IS_ERR(vma)) {
		err = PTR_ERR(vma);
		goto err_free;
	}

	vma->private = sleeve;
	vma->ops = &proxy_vma_ops;

	sleeve->vma = vma;
	sleeve->pages = pages;
	sleeve->page_sizes = *page_sizes;

	return sleeve;

err_free:
	kfree(sleeve);
	return ERR_PTR(err);
}

static void destroy_sleeve(struct i915_sleeve *sleeve)
{
	kfree(sleeve);
}

struct clear_pages_work {
	struct dma_fence dma;
	struct dma_fence_cb cb;
	struct i915_sw_fence wait;
	struct work_struct work;
	struct irq_work irq_work;
	struct i915_sleeve *sleeve;
	struct intel_context *ce;
	u32 value;
};

static const char *clear_pages_work_driver_name(struct dma_fence *fence)
{
	return DRIVER_NAME;
}

static const char *clear_pages_work_timeline_name(struct dma_fence *fence)
{
	return "clear";
}

static void clear_pages_work_release(struct dma_fence *fence)
{
	struct clear_pages_work *w = container_of(fence, typeof(*w), dma);

	destroy_sleeve(w->sleeve);

	i915_sw_fence_fini(&w->wait);

	BUILD_BUG_ON(offsetof(typeof(*w), dma));
	dma_fence_free(&w->dma);
}

static const struct dma_fence_ops clear_pages_work_ops = {
	.get_driver_name = clear_pages_work_driver_name,
	.get_timeline_name = clear_pages_work_timeline_name,
	.release = clear_pages_work_release,
};

static void clear_pages_signal_irq_worker(struct irq_work *work)
{
	struct clear_pages_work *w = container_of(work, typeof(*w), irq_work);

	dma_fence_signal(&w->dma);
	dma_fence_put(&w->dma);
}

static void clear_pages_dma_fence_cb(struct dma_fence *fence,
				     struct dma_fence_cb *cb)
{
	struct clear_pages_work *w = container_of(cb, typeof(*w), cb);

	if (fence->error)
		dma_fence_set_error(&w->dma, fence->error);

	/*
	 * Push the signalling of the fence into yet another worker to avoid
	 * the nightmare locking around the fence spinlock.
	 */
	irq_work_queue(&w->irq_work);
}

static void clear_pages_worker(struct work_struct *work)
{
	struct clear_pages_work *w = container_of(work, typeof(*w), work);
	struct drm_i915_gem_object *obj = w->sleeve->vma->obj;
	struct i915_vma *vma = w->sleeve->vma;
	struct i915_request *rq;
	struct i915_vma *batch;
	int err = w->dma.error;

	if (unlikely(err))
		goto out_signal;

	if (obj->cache_dirty) {
		if (i915_gem_object_has_struct_page(obj))
			drm_clflush_sg(w->sleeve->pages);
		obj->cache_dirty = false;
	}
	obj->read_domains = I915_GEM_GPU_DOMAINS;
	obj->write_domain = 0;

	err = i915_vma_pin(vma, 0, 0, PIN_USER);
	if (unlikely(err))
		goto out_signal;

	batch = intel_emit_vma_fill_blt(w->ce, vma, w->value);
	if (IS_ERR(batch)) {
		err = PTR_ERR(batch);
		goto out_unpin;
	}

	rq = intel_context_create_request(w->ce);
	if (IS_ERR(rq)) {
		err = PTR_ERR(rq);
		goto out_batch;
	}

	/* There's no way the fence has signalled */
	if (dma_fence_add_callback(&rq->fence, &w->cb,
				   clear_pages_dma_fence_cb))
		GEM_BUG_ON(1);

	err = intel_emit_vma_mark_active(batch, rq);
	if (unlikely(err))
		goto out_request;

	if (w->ce->engine->emit_init_breadcrumb) {
		err = w->ce->engine->emit_init_breadcrumb(rq);
		if (unlikely(err))
			goto out_request;
	}

	/*
	 * w->dma is already exported via (vma|obj)->resv we need only
	 * keep track of the GPU activity within this vma/request, and
	 * propagate the signal from the request to w->dma.
	 */
	err = __i915_vma_move_to_active(vma, rq);
	if (err)
		goto out_request;

	err = w->ce->engine->emit_bb_start(rq,
					   batch->node.start, batch->node.size,
					   0);
out_request:
	if (unlikely(err)) {
		i915_request_set_error_once(rq, err);
		err = 0;
	}

	i915_request_add(rq);
out_batch:
	intel_emit_vma_release(w->ce, batch);
out_unpin:
	i915_vma_unpin(vma);
out_signal:
	if (unlikely(err)) {
		dma_fence_set_error(&w->dma, err);
		dma_fence_signal(&w->dma);
		dma_fence_put(&w->dma);
	}
}

static int __i915_sw_fence_call
clear_pages_work_notify(struct i915_sw_fence *fence,
			enum i915_sw_fence_notify state)
{
	struct clear_pages_work *w = container_of(fence, typeof(*w), wait);

	switch (state) {
	case FENCE_COMPLETE:
		schedule_work(&w->work);
		break;

	case FENCE_FREE:
		dma_fence_put(&w->dma);
		break;
	}

	return NOTIFY_DONE;
}

static DEFINE_SPINLOCK(fence_lock);

/* XXX: better name please */
int i915_gem_schedule_fill_pages_blt(struct drm_i915_gem_object *obj,
				     struct intel_context *ce,
				     struct sg_table *pages,
				     struct i915_page_sizes *page_sizes,
				     u32 value)
{
	struct clear_pages_work *work;
	struct i915_sleeve *sleeve;
	int err;

	sleeve = create_sleeve(ce->vm, obj, pages, page_sizes);
	if (IS_ERR(sleeve))
		return PTR_ERR(sleeve);

	work = kmalloc(sizeof(*work), GFP_KERNEL);
	if (!work) {
		destroy_sleeve(sleeve);
		return -ENOMEM;
	}

	work->value = value;
	work->sleeve = sleeve;
	work->ce = ce;

	INIT_WORK(&work->work, clear_pages_worker);

	init_irq_work(&work->irq_work, clear_pages_signal_irq_worker);

	dma_fence_init(&work->dma, &clear_pages_work_ops, &fence_lock, 0, 0);
	i915_sw_fence_init(&work->wait, clear_pages_work_notify);

	i915_gem_object_lock(obj);
	err = i915_sw_fence_await_reservation(&work->wait,
					      obj->base.resv, NULL, true, 0,
					      I915_FENCE_GFP);
	if (err < 0) {
		dma_fence_set_error(&work->dma, err);
	} else {
		dma_resv_add_excl_fence(obj->base.resv, &work->dma);
		err = 0;
	}
	i915_gem_object_unlock(obj);

	dma_fence_get(&work->dma);
	i915_sw_fence_commit(&work->wait);

	return err;
}

#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
#include "selftests/i915_gem_client_blt.c"
#endif
Commit	Line	Data
6501aa4e MA	1	// SPDX-License-Identifier: MIT
	2	/*
	3	* Copyright © 2019 Intel Corporation
	4	*/
6501aa4e	5
6da4a2c4	6	#include "i915_drv.h"
963ad128	7	#include "gt/intel_context.h"
554e330c	8	#include "gt/intel_engine_pm.h"
1d455f8d	9	#include "i915_gem_client_blt.h"
6501aa4e	10	#include "i915_gem_object_blt.h"
6501aa4e MA	11
	12	struct i915_sleeve {
	13	struct i915_vma *vma;
	14	struct drm_i915_gem_object *obj;
	15	struct sg_table *pages;
	16	struct i915_page_sizes page_sizes;
	17	};
	18
	19	static int vma_set_pages(struct i915_vma *vma)
	20	{
	21	struct i915_sleeve *sleeve = vma->private;
	22
	23	vma->pages = sleeve->pages;
	24	vma->page_sizes = sleeve->page_sizes;
	25
	26	return 0;
	27	}
	28
	29	static void vma_clear_pages(struct i915_vma *vma)
	30	{
	31	GEM_BUG_ON(!vma->pages);
	32	vma->pages = NULL;
	33	}
	34
cd0452aa CW	35	static void vma_bind(struct i915_address_space *vm,
	36	struct i915_vm_pt_stash *stash,
	37	struct i915_vma *vma,
	38	enum i915_cache_level cache_level,
	39	u32 flags)
6501aa4e	40	{
cd0452aa	41	vm->vma_ops.bind_vma(vm, stash, vma, cache_level, flags);
6501aa4e MA	42	}
6501aa4e MA	43
12b07256	44	static void vma_unbind(struct i915_address_space vm, struct i915_vma vma)
6501aa4e	45	{
12b07256	46	vm->vma_ops.unbind_vma(vm, vma);
6501aa4e MA	47	}
	48
	49	static const struct i915_vma_ops proxy_vma_ops = {
	50	.set_pages = vma_set_pages,
	51	.clear_pages = vma_clear_pages,
	52	.bind_vma = vma_bind,
	53	.unbind_vma = vma_unbind,
	54	};
	55
	56	static struct i915_sleeve create_sleeve(struct i915_address_space vm,
	57	struct drm_i915_gem_object *obj,
	58	struct sg_table *pages,
	59	struct i915_page_sizes *page_sizes)
	60	{
	61	struct i915_sleeve *sleeve;
	62	struct i915_vma *vma;
	63	int err;
	64
	65	sleeve = kzalloc(sizeof(*sleeve), GFP_KERNEL);
	66	if (!sleeve)
	67	return ERR_PTR(-ENOMEM);
	68
	69	vma = i915_vma_instance(obj, vm, NULL);
	70	if (IS_ERR(vma)) {
	71	err = PTR_ERR(vma);
	72	goto err_free;
	73	}
	74
	75	vma->private = sleeve;
	76	vma->ops = &proxy_vma_ops;
	77
	78	sleeve->vma = vma;
6501aa4e MA	79	sleeve->pages = pages;
	80	sleeve->page_sizes = *page_sizes;
	81
	82	return sleeve;
	83
	84	err_free:
	85	kfree(sleeve);
	86	return ERR_PTR(err);
	87	}
	88
	89	static void destroy_sleeve(struct i915_sleeve *sleeve)
	90	{
6501aa4e MA	91	kfree(sleeve);
	92	}
	93
	94	struct clear_pages_work {
	95	struct dma_fence dma;
	96	struct dma_fence_cb cb;
	97	struct i915_sw_fence wait;
	98	struct work_struct work;
	99	struct irq_work irq_work;
	100	struct i915_sleeve *sleeve;
	101	struct intel_context *ce;
	102	u32 value;
	103	};
	104
	105	static const char clear_pages_work_driver_name(struct dma_fence fence)
	106	{
	107	return DRIVER_NAME;
	108	}
	109
	110	static const char clear_pages_work_timeline_name(struct dma_fence fence)
	111	{
	112	return "clear";
	113	}
	114
	115	static void clear_pages_work_release(struct dma_fence *fence)
	116	{
	117	struct clear_pages_work w = container_of(fence, typeof(w), dma);
	118
	119	destroy_sleeve(w->sleeve);
	120
	121	i915_sw_fence_fini(&w->wait);
	122
	123	BUILD_BUG_ON(offsetof(typeof(*w), dma));
	124	dma_fence_free(&w->dma);
	125	}
	126
	127	static const struct dma_fence_ops clear_pages_work_ops = {
	128	.get_driver_name = clear_pages_work_driver_name,
	129	.get_timeline_name = clear_pages_work_timeline_name,
	130	.release = clear_pages_work_release,
	131	};
	132
	133	static void clear_pages_signal_irq_worker(struct irq_work *work)
	134	{
	135	struct clear_pages_work w = container_of(work, typeof(w), irq_work);
	136
	137	dma_fence_signal(&w->dma);
	138	dma_fence_put(&w->dma);
	139	}
	140
	141	static void clear_pages_dma_fence_cb(struct dma_fence *fence,
	142	struct dma_fence_cb *cb)
	143	{
	144	struct clear_pages_work w = container_of(cb, typeof(w), cb);
	145
	146	if (fence->error)
	147	dma_fence_set_error(&w->dma, fence->error);
	148
	149	/*
	150	* Push the signalling of the fence into yet another worker to avoid
	151	* the nightmare locking around the fence spinlock.
	152	*/
	153	irq_work_queue(&w->irq_work);
	154	}
155
156	static void clear_pages_worker(struct work_struct *work)
157	{
158	struct clear_pages_work w = container_of(work, typeof(w), work);
b2dbf8d9	159	struct drm_i915_gem_object *obj = w->sleeve->vma->obj;
6501aa4e MA	160	struct i915_vma *vma = w->sleeve->vma;
6501aa4e MA	161	struct i915_request *rq;
554e330c	162	struct i915_vma *batch;
6501aa4e MA	163	int err = w->dma.error;
	164
	165	if (unlikely(err))
	166	goto out_signal;
	167
	168	if (obj->cache_dirty) {
6501aa4e MA	169	if (i915_gem_object_has_struct_page(obj))
	170	drm_clflush_sg(w->sleeve->pages);
	171	obj->cache_dirty = false;
	172	}
871918df CW	173	obj->read_domains = I915_GEM_GPU_DOMAINS;
871918df CW	174	obj->write_domain = 0;
6501aa4e	175
6501aa4e MA	176	err = i915_vma_pin(vma, 0, 0, PIN_USER);
6501aa4e MA	177	if (unlikely(err))
7e805762	178	goto out_signal;
6501aa4e	179
554e330c MA	180	batch = intel_emit_vma_fill_blt(w->ce, vma, w->value);
	181	if (IS_ERR(batch)) {
	182	err = PTR_ERR(batch);
	183	goto out_unpin;
	184	}
	185
963ad128	186	rq = intel_context_create_request(w->ce);
6501aa4e MA	187	if (IS_ERR(rq)) {
6501aa4e MA	188	err = PTR_ERR(rq);
554e330c	189	goto out_batch;
6501aa4e MA	190	}
	191
	192	/* There's no way the fence has signalled */
	193	if (dma_fence_add_callback(&rq->fence, &w->cb,
	194	clear_pages_dma_fence_cb))
	195	GEM_BUG_ON(1);
	196
554e330c MA	197	err = intel_emit_vma_mark_active(batch, rq);
	198	if (unlikely(err))
	199	goto out_request;
	200
6501aa4e MA	201	if (w->ce->engine->emit_init_breadcrumb) {
	202	err = w->ce->engine->emit_init_breadcrumb(rq);
	203	if (unlikely(err))
	204	goto out_request;
	205	}
	206
b2dbf8d9 CW	207	/*
	208	* w->dma is already exported via (vma\|obj)->resv we need only
	209	* keep track of the GPU activity within this vma/request, and
	210	* propagate the signal from the request to w->dma.
	211	*/
2850748e	212	err = __i915_vma_move_to_active(vma, rq);
6501aa4e MA	213	if (err)
	214	goto out_request;
	215
554e330c MA	216	err = w->ce->engine->emit_bb_start(rq,
	217	batch->node.start, batch->node.size,
	218	0);
6501aa4e MA	219	out_request:
6501aa4e MA	220	if (unlikely(err)) {
36e191f0	221	i915_request_set_error_once(rq, err);
6501aa4e MA	222	err = 0;
	223	}
	224
	225	i915_request_add(rq);
554e330c MA	226	out_batch:
554e330c MA	227	intel_emit_vma_release(w->ce, batch);
6501aa4e MA	228	out_unpin:
6501aa4e MA	229	i915_vma_unpin(vma);
6501aa4e MA	230	out_signal:
	231	if (unlikely(err)) {
	232	dma_fence_set_error(&w->dma, err);
	233	dma_fence_signal(&w->dma);
	234	dma_fence_put(&w->dma);
	235	}
	236	}
	237
	238	static int __i915_sw_fence_call
	239	clear_pages_work_notify(struct i915_sw_fence *fence,
	240	enum i915_sw_fence_notify state)
	241	{
	242	struct clear_pages_work w = container_of(fence, typeof(w), wait);
	243
	244	switch (state) {
	245	case FENCE_COMPLETE:
	246	schedule_work(&w->work);
	247	break;
	248
	249	case FENCE_FREE:
	250	dma_fence_put(&w->dma);
	251	break;
	252	}
	253
	254	return NOTIFY_DONE;
	255	}
	256
	257	static DEFINE_SPINLOCK(fence_lock);
	258
	259	/* XXX: better name please */
	260	int i915_gem_schedule_fill_pages_blt(struct drm_i915_gem_object *obj,
	261	struct intel_context *ce,
	262	struct sg_table *pages,
	263	struct i915_page_sizes *page_sizes,
	264	u32 value)
	265	{
6501aa4e MA	266	struct clear_pages_work *work;
	267	struct i915_sleeve *sleeve;
	268	int err;
	269
f5d974f9	270	sleeve = create_sleeve(ce->vm, obj, pages, page_sizes);
6501aa4e MA	271	if (IS_ERR(sleeve))
	272	return PTR_ERR(sleeve);
	273
	274	work = kmalloc(sizeof(*work), GFP_KERNEL);
	275	if (!work) {
	276	destroy_sleeve(sleeve);
	277	return -ENOMEM;
	278	}
	279
	280	work->value = value;
	281	work->sleeve = sleeve;
	282	work->ce = ce;
	283
	284	INIT_WORK(&work->work, clear_pages_worker);
	285
	286	init_irq_work(&work->irq_work, clear_pages_signal_irq_worker);
	287
cc337560	288	dma_fence_init(&work->dma, &clear_pages_work_ops, &fence_lock, 0, 0);
6501aa4e MA	289	i915_sw_fence_init(&work->wait, clear_pages_work_notify);
	290
	291	i915_gem_object_lock(obj);
	292	err = i915_sw_fence_await_reservation(&work->wait,
16dc224f	293	obj->base.resv, NULL, true, 0,
6501aa4e MA	294	I915_FENCE_GFP);
	295	if (err < 0) {
	296	dma_fence_set_error(&work->dma, err);
	297	} else {
52791eee	298	dma_resv_add_excl_fence(obj->base.resv, &work->dma);
6501aa4e MA	299	err = 0;
	300	}
	301	i915_gem_object_unlock(obj);
	302
	303	dma_fence_get(&work->dma);
	304	i915_sw_fence_commit(&work->wait);
	305
	306	return err;
	307	}
	308
	309	#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
	310	#include "selftests/i915_gem_client_blt.c"
	311	#endif