[linux-2.6-block.git] / drivers / gpu / drm / radeon / radeon_cs.c

/*
 * Copyright 2008 Jerome Glisse.
 * All Rights Reserved.
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the "Software"),
 * to deal in the Software without restriction, including without limitation
 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
 * and/or sell copies of the Software, and to permit persons to whom the
 * Software is furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice (including the next
 * paragraph) shall be included in all copies or substantial portions of the
 * Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
 * PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
 * DEALINGS IN THE SOFTWARE.
 *
 * Authors:
 *    Jerome Glisse <glisse@freedesktop.org>
 */
#include "drmP.h"
#include "radeon_drm.h"
#include "radeon_reg.h"
#include "radeon.h"

void r100_cs_dump_packet(struct radeon_cs_parser *p,
			 struct radeon_cs_packet *pkt);

int radeon_cs_parser_relocs(struct radeon_cs_parser *p)
{
	struct drm_device *ddev = p->rdev->ddev;
	struct radeon_cs_chunk *chunk;
	unsigned i, j;
	bool duplicate;

	if (p->chunk_relocs_idx == -1) {
		return 0;
	}
	chunk = &p->chunks[p->chunk_relocs_idx];
	/* FIXME: we assume that each relocs use 4 dwords */
	p->nrelocs = chunk->length_dw / 4;
	p->relocs_ptr = kcalloc(p->nrelocs, sizeof(void *), GFP_KERNEL);
	if (p->relocs_ptr == NULL) {
		return -ENOMEM;
	}
	p->relocs = kcalloc(p->nrelocs, sizeof(struct radeon_cs_reloc), GFP_KERNEL);
	if (p->relocs == NULL) {
		return -ENOMEM;
	}
	for (i = 0; i < p->nrelocs; i++) {
		struct drm_radeon_cs_reloc *r;

		duplicate = false;
		r = (struct drm_radeon_cs_reloc *)&chunk->kdata[i*4];
		for (j = 0; j < i; j++) {
			if (r->handle == p->relocs[j].handle) {
				p->relocs_ptr[i] = &p->relocs[j];
				duplicate = true;
				break;
			}
		}
		if (!duplicate) {
			p->relocs[i].gobj = drm_gem_object_lookup(ddev,
								  p->filp,
								  r->handle);
			if (p->relocs[i].gobj == NULL) {
				DRM_ERROR("gem object lookup failed 0x%x\n",
					  r->handle);
				return -ENOENT;
			}
			p->relocs_ptr[i] = &p->relocs[i];
			p->relocs[i].robj = gem_to_radeon_bo(p->relocs[i].gobj);
			p->relocs[i].lobj.bo = p->relocs[i].robj;
			p->relocs[i].lobj.wdomain = r->write_domain;
			p->relocs[i].lobj.rdomain = r->read_domains;
			p->relocs[i].lobj.tv.bo = &p->relocs[i].robj->tbo;
			p->relocs[i].handle = r->handle;
			p->relocs[i].flags = r->flags;
			radeon_bo_list_add_object(&p->relocs[i].lobj,
						  &p->validated);

		} else
			p->relocs[i].handle = 0;
	}
	return radeon_bo_list_validate(&p->validated);
}

static int radeon_cs_get_ring(struct radeon_cs_parser *p, u32 ring, s32 priority)
{
	p->priority = priority;

	switch (ring) {
	default:
		DRM_ERROR("unknown ring id: %d\n", ring);
		return -EINVAL;
	case RADEON_CS_RING_GFX:
		p->ring = RADEON_RING_TYPE_GFX_INDEX;
		break;
	case RADEON_CS_RING_COMPUTE:
		/* for now */
		p->ring = RADEON_RING_TYPE_GFX_INDEX;
		break;
	}
	return 0;
}

static int radeon_cs_sync_rings(struct radeon_cs_parser *p)
{
	bool sync_to_ring[RADEON_NUM_RINGS] = { };
	int i, r;

	for (i = 0; i < p->nrelocs; i++) {
		if (!p->relocs[i].robj || !p->relocs[i].robj->tbo.sync_obj)
			continue;

		if (!(p->relocs[i].flags & RADEON_RELOC_DONT_SYNC)) {
			struct radeon_fence *fence = p->relocs[i].robj->tbo.sync_obj;
			if (!radeon_fence_signaled(fence)) {
				sync_to_ring[fence->ring] = true;
			}
		}
	}

	for (i = 0; i < RADEON_NUM_RINGS; ++i) {
		/* no need to sync to our own or unused rings */
		if (i == p->ring || !sync_to_ring[i] || !p->rdev->ring[i].ready)
			continue;

		if (!p->ib->fence->semaphore) {
			r = radeon_semaphore_create(p->rdev, &p->ib->fence->semaphore);
			if (r)
				return r;
		}

		r = radeon_ring_lock(p->rdev, &p->rdev->ring[i], 3);
		if (r)
			return r;
		radeon_semaphore_emit_signal(p->rdev, i, p->ib->fence->semaphore);
		radeon_ring_unlock_commit(p->rdev, &p->rdev->ring[i]);

		r = radeon_ring_lock(p->rdev, &p->rdev->ring[p->ring], 3);
		if (r)
			return r;
		radeon_semaphore_emit_wait(p->rdev, p->ring, p->ib->fence->semaphore);
		radeon_ring_unlock_commit(p->rdev, &p->rdev->ring[p->ring]);
	}
	return 0;
}

int radeon_cs_parser_init(struct radeon_cs_parser *p, void *data)
{
	struct drm_radeon_cs *cs = data;
	uint64_t *chunk_array_ptr;
	unsigned size, i;
	u32 ring = RADEON_CS_RING_GFX;
	s32 priority = 0;

	if (!cs->num_chunks) {
		return 0;
	}
	/* get chunks */
	INIT_LIST_HEAD(&p->validated);
	p->idx = 0;
	p->chunk_ib_idx = -1;
	p->chunk_relocs_idx = -1;
	p->chunk_flags_idx = -1;
	p->chunks_array = kcalloc(cs->num_chunks, sizeof(uint64_t), GFP_KERNEL);
	if (p->chunks_array == NULL) {
		return -ENOMEM;
	}
	chunk_array_ptr = (uint64_t *)(unsigned long)(cs->chunks);
	if (DRM_COPY_FROM_USER(p->chunks_array, chunk_array_ptr,
			       sizeof(uint64_t)*cs->num_chunks)) {
		return -EFAULT;
	}
	p->cs_flags = 0;
	p->nchunks = cs->num_chunks;
	p->chunks = kcalloc(p->nchunks, sizeof(struct radeon_cs_chunk), GFP_KERNEL);
	if (p->chunks == NULL) {
		return -ENOMEM;
	}
	for (i = 0; i < p->nchunks; i++) {
		struct drm_radeon_cs_chunk __user **chunk_ptr = NULL;
		struct drm_radeon_cs_chunk user_chunk;
		uint32_t __user *cdata;

		chunk_ptr = (void __user*)(unsigned long)p->chunks_array[i];
		if (DRM_COPY_FROM_USER(&user_chunk, chunk_ptr,
				       sizeof(struct drm_radeon_cs_chunk))) {
			return -EFAULT;
		}
		p->chunks[i].length_dw = user_chunk.length_dw;
		p->chunks[i].kdata = NULL;
		p->chunks[i].chunk_id = user_chunk.chunk_id;

		if (p->chunks[i].chunk_id == RADEON_CHUNK_ID_RELOCS) {
			p->chunk_relocs_idx = i;
		}
		if (p->chunks[i].chunk_id == RADEON_CHUNK_ID_IB) {
			p->chunk_ib_idx = i;
			/* zero length IB isn't useful */
			if (p->chunks[i].length_dw == 0)
				return -EINVAL;
		}
		if (p->chunks[i].chunk_id == RADEON_CHUNK_ID_FLAGS) {
			p->chunk_flags_idx = i;
			/* zero length flags aren't useful */
			if (p->chunks[i].length_dw == 0)
				return -EINVAL;
		}

		p->chunks[i].length_dw = user_chunk.length_dw;
		p->chunks[i].user_ptr = (void __user *)(unsigned long)user_chunk.chunk_data;

		cdata = (uint32_t *)(unsigned long)user_chunk.chunk_data;
		if ((p->chunks[i].chunk_id == RADEON_CHUNK_ID_RELOCS) ||
		    (p->chunks[i].chunk_id == RADEON_CHUNK_ID_FLAGS)) {
			size = p->chunks[i].length_dw * sizeof(uint32_t);
			p->chunks[i].kdata = kmalloc(size, GFP_KERNEL);
			if (p->chunks[i].kdata == NULL) {
				return -ENOMEM;
			}
			if (DRM_COPY_FROM_USER(p->chunks[i].kdata,
					       p->chunks[i].user_ptr, size)) {
				return -EFAULT;
			}
			if (p->chunks[i].chunk_id == RADEON_CHUNK_ID_FLAGS) {
				p->cs_flags = p->chunks[i].kdata[0];
				if (p->chunks[i].length_dw > 1)
					ring = p->chunks[i].kdata[1];
				if (p->chunks[i].length_dw > 2)
					priority = (s32)p->chunks[i].kdata[2];
			}
		}
	}

	if ((p->cs_flags & RADEON_CS_USE_VM) &&
	    !p->rdev->vm_manager.enabled) {
		DRM_ERROR("VM not active on asic!\n");
		return -EINVAL;
	}

	if (radeon_cs_get_ring(p, ring, priority))
		return -EINVAL;


	/* deal with non-vm */
	if ((p->chunk_ib_idx != -1) &&
	    ((p->cs_flags & RADEON_CS_USE_VM) == 0) &&
	    (p->chunks[p->chunk_ib_idx].chunk_id == RADEON_CHUNK_ID_IB)) {
		if (p->chunks[p->chunk_ib_idx].length_dw > (16 * 1024)) {
			DRM_ERROR("cs IB too big: %d\n",
				  p->chunks[p->chunk_ib_idx].length_dw);
			return -EINVAL;
		}
		p->chunks[p->chunk_ib_idx].kpage[0] = kmalloc(PAGE_SIZE, GFP_KERNEL);
		p->chunks[p->chunk_ib_idx].kpage[1] = kmalloc(PAGE_SIZE, GFP_KERNEL);
		if (p->chunks[p->chunk_ib_idx].kpage[0] == NULL ||
		    p->chunks[p->chunk_ib_idx].kpage[1] == NULL)
			return -ENOMEM;
		p->chunks[p->chunk_ib_idx].kpage_idx[0] = -1;
		p->chunks[p->chunk_ib_idx].kpage_idx[1] = -1;
		p->chunks[p->chunk_ib_idx].last_copied_page = -1;
		p->chunks[p->chunk_ib_idx].last_page_index =
			((p->chunks[p->chunk_ib_idx].length_dw * 4) - 1) / PAGE_SIZE;
	}

	return 0;
}

/**
 * cs_parser_fini() - clean parser states
 * @parser:	parser structure holding parsing context.
 * @error:	error number
 *
 * If error is set than unvalidate buffer, otherwise just free memory
 * used by parsing context.
 **/
static void radeon_cs_parser_fini(struct radeon_cs_parser *parser, int error)
{
	unsigned i;


	if (!error && parser->ib)
		ttm_eu_fence_buffer_objects(&parser->validated,
					    parser->ib->fence);
	else
		ttm_eu_backoff_reservation(&parser->validated);

	if (parser->relocs != NULL) {
		for (i = 0; i < parser->nrelocs; i++) {
			if (parser->relocs[i].gobj)
				drm_gem_object_unreference_unlocked(parser->relocs[i].gobj);
		}
	}
	kfree(parser->track);
	kfree(parser->relocs);
	kfree(parser->relocs_ptr);
	for (i = 0; i < parser->nchunks; i++) {
		kfree(parser->chunks[i].kdata);
		kfree(parser->chunks[i].kpage[0]);
		kfree(parser->chunks[i].kpage[1]);
	}
	kfree(parser->chunks);
	kfree(parser->chunks_array);
	radeon_ib_free(parser->rdev, &parser->ib);
}

static int radeon_cs_ib_chunk(struct radeon_device *rdev,
			      struct radeon_cs_parser *parser)
{
	struct radeon_cs_chunk *ib_chunk;
	int r;

	if (parser->chunk_ib_idx == -1)
		return 0;

	if (parser->cs_flags & RADEON_CS_USE_VM)
		return 0;

	ib_chunk = &parser->chunks[parser->chunk_ib_idx];
	/* Copy the packet into the IB, the parser will read from the
	 * input memory (cached) and write to the IB (which can be
	 * uncached).
	 */
	r =  radeon_ib_get(rdev, parser->ring, &parser->ib,
			   ib_chunk->length_dw * 4);
	if (r) {
		DRM_ERROR("Failed to get ib !\n");
		return r;
	}
	parser->ib->length_dw = ib_chunk->length_dw;
	r = radeon_cs_parse(rdev, parser->ring, parser);
	if (r || parser->parser_error) {
		DRM_ERROR("Invalid command stream !\n");
		return r;
	}
	r = radeon_cs_finish_pages(parser);
	if (r) {
		DRM_ERROR("Invalid command stream !\n");
		return r;
	}
	r = radeon_cs_sync_rings(parser);
	if (r) {
		DRM_ERROR("Failed to synchronize rings !\n");
	}
	parser->ib->vm_id = 0;
	r = radeon_ib_schedule(rdev, parser->ib);
	if (r) {
		DRM_ERROR("Failed to schedule IB !\n");
	}
	return 0;
}

static int radeon_bo_vm_update_pte(struct radeon_cs_parser *parser,
				   struct radeon_vm *vm)
{
	struct radeon_bo_list *lobj;
	struct radeon_bo *bo;
	int r;

	list_for_each_entry(lobj, &parser->validated, tv.head) {
		bo = lobj->bo;
		r = radeon_vm_bo_update_pte(parser->rdev, vm, bo, &bo->tbo.mem);
		if (r) {
			return r;
		}
	}
	return 0;
}

static int radeon_cs_ib_vm_chunk(struct radeon_device *rdev,
				 struct radeon_cs_parser *parser)
{
	struct radeon_cs_chunk *ib_chunk;
	struct radeon_fpriv *fpriv = parser->filp->driver_priv;
	struct radeon_vm *vm = &fpriv->vm;
	int r;

	if (parser->chunk_ib_idx == -1)
		return 0;

	if ((parser->cs_flags & RADEON_CS_USE_VM) == 0)
		return 0;

	ib_chunk = &parser->chunks[parser->chunk_ib_idx];
	if (ib_chunk->length_dw > RADEON_IB_VM_MAX_SIZE) {
		DRM_ERROR("cs IB too big: %d\n", ib_chunk->length_dw);
		return -EINVAL;
	}
	r =  radeon_ib_get(rdev, parser->ring, &parser->ib,
			   ib_chunk->length_dw * 4);
	if (r) {
		DRM_ERROR("Failed to get ib !\n");
		return r;
	}
	parser->ib->length_dw = ib_chunk->length_dw;
	/* Copy the packet into the IB */
	if (DRM_COPY_FROM_USER(parser->ib->ptr, ib_chunk->user_ptr,
			       ib_chunk->length_dw * 4)) {
		return -EFAULT;
	}
	r = radeon_ring_ib_parse(rdev, parser->ring, parser->ib);
	if (r) {
		return r;
	}

	mutex_lock(&vm->mutex);
	r = radeon_vm_bind(rdev, vm);
	if (r) {
		goto out;
	}
	r = radeon_bo_vm_update_pte(parser, vm);
	if (r) {
		goto out;
	}
	r = radeon_cs_sync_rings(parser);
	if (r) {
		DRM_ERROR("Failed to synchronize rings !\n");
	}
	parser->ib->vm_id = vm->id;
	/* ib pool is bind at 0 in virtual address space to gpu_addr is the
	 * offset inside the pool bo
	 */
	parser->ib->gpu_addr = parser->ib->sa_bo.offset;
	r = radeon_ib_schedule(rdev, parser->ib);
out:
	if (!r) {
		if (vm->fence) {
			radeon_fence_unref(&vm->fence);
		}
		vm->fence = radeon_fence_ref(parser->ib->fence);
	}
	mutex_unlock(&fpriv->vm.mutex);
	return r;
}

int radeon_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
{
	struct radeon_device *rdev = dev->dev_private;
	struct radeon_cs_parser parser;
	int r;

	radeon_mutex_lock(&rdev->cs_mutex);
	if (!rdev->accel_working) {
		radeon_mutex_unlock(&rdev->cs_mutex);
		return -EBUSY;
	}
	/* initialize parser */
	memset(&parser, 0, sizeof(struct radeon_cs_parser));
	parser.filp = filp;
	parser.rdev = rdev;
	parser.dev = rdev->dev;
	parser.family = rdev->family;
	r = radeon_cs_parser_init(&parser, data);
	if (r) {
		DRM_ERROR("Failed to initialize parser !\n");
		radeon_cs_parser_fini(&parser, r);
		radeon_mutex_unlock(&rdev->cs_mutex);
		return r;
	}
	r = radeon_cs_parser_relocs(&parser);
	if (r) {
		if (r != -ERESTARTSYS)
			DRM_ERROR("Failed to parse relocation %d!\n", r);
		radeon_cs_parser_fini(&parser, r);
		radeon_mutex_unlock(&rdev->cs_mutex);
		return r;
	}
	r = radeon_cs_ib_chunk(rdev, &parser);
	if (r) {
		goto out;
	}
	r = radeon_cs_ib_vm_chunk(rdev, &parser);
	if (r) {
		goto out;
	}
out:
	radeon_cs_parser_fini(&parser, r);
	radeon_mutex_unlock(&rdev->cs_mutex);
	return r;
}

int radeon_cs_finish_pages(struct radeon_cs_parser *p)
{
	struct radeon_cs_chunk *ibc = &p->chunks[p->chunk_ib_idx];
	int i;
	int size = PAGE_SIZE;

	for (i = ibc->last_copied_page + 1; i <= ibc->last_page_index; i++) {
		if (i == ibc->last_page_index) {
			size = (ibc->length_dw * 4) % PAGE_SIZE;
			if (size == 0)
				size = PAGE_SIZE;
		}
		
		if (DRM_COPY_FROM_USER(p->ib->ptr + (i * (PAGE_SIZE/4)),
				       ibc->user_ptr + (i * PAGE_SIZE),
				       size))
			return -EFAULT;
	}
	return 0;
}

int radeon_cs_update_pages(struct radeon_cs_parser *p, int pg_idx)
{
	int new_page;
	struct radeon_cs_chunk *ibc = &p->chunks[p->chunk_ib_idx];
	int i;
	int size = PAGE_SIZE;

	for (i = ibc->last_copied_page + 1; i < pg_idx; i++) {
		if (DRM_COPY_FROM_USER(p->ib->ptr + (i * (PAGE_SIZE/4)),
				       ibc->user_ptr + (i * PAGE_SIZE),
				       PAGE_SIZE)) {
			p->parser_error = -EFAULT;
			return 0;
		}
	}

	new_page = ibc->kpage_idx[0] < ibc->kpage_idx[1] ? 0 : 1;

	if (pg_idx == ibc->last_page_index) {
		size = (ibc->length_dw * 4) % PAGE_SIZE;
			if (size == 0)
				size = PAGE_SIZE;
	}

	if (DRM_COPY_FROM_USER(ibc->kpage[new_page],
			       ibc->user_ptr + (pg_idx * PAGE_SIZE),
			       size)) {
		p->parser_error = -EFAULT;
		return 0;
	}

	/* copy to IB here */
	memcpy((void *)(p->ib->ptr+(pg_idx*(PAGE_SIZE/4))), ibc->kpage[new_page], size);

	ibc->last_copied_page = pg_idx;
	ibc->kpage_idx[new_page] = pg_idx;

	return new_page;
}
Commit	Line	Data
771fe6b9 JG	1	/*
	2	* Copyright 2008 Jerome Glisse.
	3	* All Rights Reserved.
	4	*
	5	* Permission is hereby granted, free of charge, to any person obtaining a
	6	* copy of this software and associated documentation files (the "Software"),
	7	* to deal in the Software without restriction, including without limitation
	8	* the rights to use, copy, modify, merge, publish, distribute, sublicense,
	9	* and/or sell copies of the Software, and to permit persons to whom the
	10	* Software is furnished to do so, subject to the following conditions:
	11	*
	12	* The above copyright notice and this permission notice (including the next
	13	* paragraph) shall be included in all copies or substantial portions of the
	14	* Software.
	15	*
	16	* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
	17	* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
	18	* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
	19	* PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
	20	* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
	21	* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
	22	* DEALINGS IN THE SOFTWARE.
	23	*
	24	* Authors:
	25	* Jerome Glisse <glisse@freedesktop.org>
	26	*/
	27	#include "drmP.h"
	28	#include "radeon_drm.h"
	29	#include "radeon_reg.h"
	30	#include "radeon.h"
	31
	32	void r100_cs_dump_packet(struct radeon_cs_parser *p,
	33	struct radeon_cs_packet *pkt);
	34
	35	int radeon_cs_parser_relocs(struct radeon_cs_parser *p)
	36	{
	37	struct drm_device *ddev = p->rdev->ddev;
	38	struct radeon_cs_chunk *chunk;
	39	unsigned i, j;
	40	bool duplicate;
	41
	42	if (p->chunk_relocs_idx == -1) {
	43	return 0;
	44	}
	45	chunk = &p->chunks[p->chunk_relocs_idx];
	46	/* FIXME: we assume that each relocs use 4 dwords */
	47	p->nrelocs = chunk->length_dw / 4;
	48	p->relocs_ptr = kcalloc(p->nrelocs, sizeof(void *), GFP_KERNEL);
	49	if (p->relocs_ptr == NULL) {
	50	return -ENOMEM;
	51	}
	52	p->relocs = kcalloc(p->nrelocs, sizeof(struct radeon_cs_reloc), GFP_KERNEL);
	53	if (p->relocs == NULL) {
	54	return -ENOMEM;
	55	}
	56	for (i = 0; i < p->nrelocs; i++) {
	57	struct drm_radeon_cs_reloc *r;
	58
	59	duplicate = false;
	60	r = (struct drm_radeon_cs_reloc )&chunk->kdata[i4];
16557f1e	61	for (j = 0; j < i; j++) {
771fe6b9 JG	62	if (r->handle == p->relocs[j].handle) {
	63	p->relocs_ptr[i] = &p->relocs[j];
	64	duplicate = true;
	65	break;
	66	}
	67	}
	68	if (!duplicate) {
	69	p->relocs[i].gobj = drm_gem_object_lookup(ddev,
	70	p->filp,
	71	r->handle);
	72	if (p->relocs[i].gobj == NULL) {
	73	DRM_ERROR("gem object lookup failed 0x%x\n",
	74	r->handle);
bf79cb91	75	return -ENOENT;
771fe6b9 JG	76	}
771fe6b9 JG	77	p->relocs_ptr[i] = &p->relocs[i];
7e4d15d9	78	p->relocs[i].robj = gem_to_radeon_bo(p->relocs[i].gobj);
4c788679	79	p->relocs[i].lobj.bo = p->relocs[i].robj;
771fe6b9	80	p->relocs[i].lobj.wdomain = r->write_domain;
147666fb TH	81	p->relocs[i].lobj.rdomain = r->read_domains;
147666fb TH	82	p->relocs[i].lobj.tv.bo = &p->relocs[i].robj->tbo;
771fe6b9 JG	83	p->relocs[i].handle = r->handle;
771fe6b9 JG	84	p->relocs[i].flags = r->flags;
4c788679	85	radeon_bo_list_add_object(&p->relocs[i].lobj,
147666fb	86	&p->validated);
93504fce	87
16557f1e CK	88	} else
16557f1e CK	89	p->relocs[i].handle = 0;
771fe6b9	90	}
94429bb6	91	return radeon_bo_list_validate(&p->validated);
771fe6b9 JG	92	}
771fe6b9 JG	93
721604a1 JG	94	static int radeon_cs_get_ring(struct radeon_cs_parser *p, u32 ring, s32 priority)
	95	{
	96	p->priority = priority;
	97
	98	switch (ring) {
	99	default:
	100	DRM_ERROR("unknown ring id: %d\n", ring);
	101	return -EINVAL;
	102	case RADEON_CS_RING_GFX:
	103	p->ring = RADEON_RING_TYPE_GFX_INDEX;
	104	break;
	105	case RADEON_CS_RING_COMPUTE:
	106	/* for now */
	107	p->ring = RADEON_RING_TYPE_GFX_INDEX;
	108	break;
	109	}
	110	return 0;
	111	}
	112
93504fce CK	113	static int radeon_cs_sync_rings(struct radeon_cs_parser *p)
93504fce CK	114	{
cdac5504	115	bool sync_to_ring[RADEON_NUM_RINGS] = { };
93504fce CK	116	int i, r;
93504fce CK	117
cdac5504 CK	118	for (i = 0; i < p->nrelocs; i++) {
	119	if (!p->relocs[i].robj \|\| !p->relocs[i].robj->tbo.sync_obj)
	120	continue;
	121
	122	if (!(p->relocs[i].flags & RADEON_RELOC_DONT_SYNC)) {
	123	struct radeon_fence *fence = p->relocs[i].robj->tbo.sync_obj;
	124	if (!radeon_fence_signaled(fence)) {
	125	sync_to_ring[fence->ring] = true;
	126	}
	127	}
	128	}
	129
93504fce CK	130	for (i = 0; i < RADEON_NUM_RINGS; ++i) {
93504fce CK	131	/* no need to sync to our own or unused rings */
cdac5504	132	if (i == p->ring \|\| !sync_to_ring[i] \|\| !p->rdev->ring[i].ready)
93504fce CK	133	continue;
	134
	135	if (!p->ib->fence->semaphore) {
	136	r = radeon_semaphore_create(p->rdev, &p->ib->fence->semaphore);
	137	if (r)
	138	return r;
	139	}
	140
	141	r = radeon_ring_lock(p->rdev, &p->rdev->ring[i], 3);
	142	if (r)
	143	return r;
	144	radeon_semaphore_emit_signal(p->rdev, i, p->ib->fence->semaphore);
	145	radeon_ring_unlock_commit(p->rdev, &p->rdev->ring[i]);
	146
	147	r = radeon_ring_lock(p->rdev, &p->rdev->ring[p->ring], 3);
	148	if (r)
	149	return r;
	150	radeon_semaphore_emit_wait(p->rdev, p->ring, p->ib->fence->semaphore);
	151	radeon_ring_unlock_commit(p->rdev, &p->rdev->ring[p->ring]);
	152	}
	153	return 0;
	154	}
	155
771fe6b9 JG	156	int radeon_cs_parser_init(struct radeon_cs_parser p, void data)
	157	{
	158	struct drm_radeon_cs *cs = data;
	159	uint64_t *chunk_array_ptr;
721604a1 JG	160	unsigned size, i;
	161	u32 ring = RADEON_CS_RING_GFX;
	162	s32 priority = 0;
771fe6b9 JG	163
	164	if (!cs->num_chunks) {
	165	return 0;
	166	}
	167	/* get chunks */
	168	INIT_LIST_HEAD(&p->validated);
	169	p->idx = 0;
	170	p->chunk_ib_idx = -1;
	171	p->chunk_relocs_idx = -1;
721604a1	172	p->chunk_flags_idx = -1;
771fe6b9 JG	173	p->chunks_array = kcalloc(cs->num_chunks, sizeof(uint64_t), GFP_KERNEL);
	174	if (p->chunks_array == NULL) {
	175	return -ENOMEM;
	176	}
	177	chunk_array_ptr = (uint64_t *)(unsigned long)(cs->chunks);
	178	if (DRM_COPY_FROM_USER(p->chunks_array, chunk_array_ptr,
	179	sizeof(uint64_t)*cs->num_chunks)) {
	180	return -EFAULT;
	181	}
721604a1	182	p->cs_flags = 0;
771fe6b9 JG	183	p->nchunks = cs->num_chunks;
	184	p->chunks = kcalloc(p->nchunks, sizeof(struct radeon_cs_chunk), GFP_KERNEL);
	185	if (p->chunks == NULL) {
	186	return -ENOMEM;
	187	}
	188	for (i = 0; i < p->nchunks; i++) {
	189	struct drm_radeon_cs_chunk __user **chunk_ptr = NULL;
	190	struct drm_radeon_cs_chunk user_chunk;
	191	uint32_t __user *cdata;
	192
	193	chunk_ptr = (void __user*)(unsigned long)p->chunks_array[i];
	194	if (DRM_COPY_FROM_USER(&user_chunk, chunk_ptr,
	195	sizeof(struct drm_radeon_cs_chunk))) {
	196	return -EFAULT;
	197	}
5176fdc4 DA	198	p->chunks[i].length_dw = user_chunk.length_dw;
5176fdc4 DA	199	p->chunks[i].kdata = NULL;
771fe6b9	200	p->chunks[i].chunk_id = user_chunk.chunk_id;
5176fdc4	201
771fe6b9 JG	202	if (p->chunks[i].chunk_id == RADEON_CHUNK_ID_RELOCS) {
	203	p->chunk_relocs_idx = i;
	204	}
	205	if (p->chunks[i].chunk_id == RADEON_CHUNK_ID_IB) {
	206	p->chunk_ib_idx = i;
5176fdc4 DA	207	/* zero length IB isn't useful */
	208	if (p->chunks[i].length_dw == 0)
	209	return -EINVAL;
771fe6b9	210	}
721604a1 JG	211	if (p->chunks[i].chunk_id == RADEON_CHUNK_ID_FLAGS) {
	212	p->chunk_flags_idx = i;
	213	/* zero length flags aren't useful */
	214	if (p->chunks[i].length_dw == 0)
	215	return -EINVAL;
e70f224c	216	}
5176fdc4	217
771fe6b9	218	p->chunks[i].length_dw = user_chunk.length_dw;
513bcb46	219	p->chunks[i].user_ptr = (void __user *)(unsigned long)user_chunk.chunk_data;
771fe6b9	220
513bcb46	221	cdata = (uint32_t *)(unsigned long)user_chunk.chunk_data;
721604a1 JG	222	if ((p->chunks[i].chunk_id == RADEON_CHUNK_ID_RELOCS) \|\|
721604a1 JG	223	(p->chunks[i].chunk_id == RADEON_CHUNK_ID_FLAGS)) {
513bcb46 DA	224	size = p->chunks[i].length_dw * sizeof(uint32_t);
	225	p->chunks[i].kdata = kmalloc(size, GFP_KERNEL);
	226	if (p->chunks[i].kdata == NULL) {
	227	return -ENOMEM;
	228	}
	229	if (DRM_COPY_FROM_USER(p->chunks[i].kdata,
	230	p->chunks[i].user_ptr, size)) {
	231	return -EFAULT;
	232	}
e70f224c	233	if (p->chunks[i].chunk_id == RADEON_CHUNK_ID_FLAGS) {
721604a1 JG	234	p->cs_flags = p->chunks[i].kdata[0];
	235	if (p->chunks[i].length_dw > 1)
	236	ring = p->chunks[i].kdata[1];
	237	if (p->chunks[i].length_dw > 2)
	238	priority = (s32)p->chunks[i].kdata[2];
e70f224c	239	}
771fe6b9 JG	240	}
771fe6b9 JG	241	}
721604a1 JG	242
721604a1 JG	243	if ((p->cs_flags & RADEON_CS_USE_VM) &&
67e915e4 AD	244	!p->rdev->vm_manager.enabled) {
67e915e4 AD	245	DRM_ERROR("VM not active on asic!\n");
771fe6b9 JG	246	return -EINVAL;
771fe6b9 JG	247	}
e70f224c	248
f48bb04a	249	if (radeon_cs_get_ring(p, ring, priority))
721604a1	250	return -EINVAL;
721604a1 JG	251
	252
	253	/* deal with non-vm */
	254	if ((p->chunk_ib_idx != -1) &&
	255	((p->cs_flags & RADEON_CS_USE_VM) == 0) &&
	256	(p->chunks[p->chunk_ib_idx].chunk_id == RADEON_CHUNK_ID_IB)) {
	257	if (p->chunks[p->chunk_ib_idx].length_dw > (16 * 1024)) {
	258	DRM_ERROR("cs IB too big: %d\n",
	259	p->chunks[p->chunk_ib_idx].length_dw);
	260	return -EINVAL;
	261	}
	262	p->chunks[p->chunk_ib_idx].kpage[0] = kmalloc(PAGE_SIZE, GFP_KERNEL);
	263	p->chunks[p->chunk_ib_idx].kpage[1] = kmalloc(PAGE_SIZE, GFP_KERNEL);
	264	if (p->chunks[p->chunk_ib_idx].kpage[0] == NULL \|\|
f48bb04a	265	p->chunks[p->chunk_ib_idx].kpage[1] == NULL)
721604a1	266	return -ENOMEM;
721604a1 JG	267	p->chunks[p->chunk_ib_idx].kpage_idx[0] = -1;
	268	p->chunks[p->chunk_ib_idx].kpage_idx[1] = -1;
	269	p->chunks[p->chunk_ib_idx].last_copied_page = -1;
	270	p->chunks[p->chunk_ib_idx].last_page_index =
	271	((p->chunks[p->chunk_ib_idx].length_dw * 4) - 1) / PAGE_SIZE;
	272	}
	273
771fe6b9 JG	274	return 0;
	275	}
	276
	277	/**
	278	* cs_parser_fini() - clean parser states
	279	* @parser: parser structure holding parsing context.
	280	* @error: error number
	281	*
	282	* If error is set than unvalidate buffer, otherwise just free memory
	283	* used by parsing context.
	284	**/
	285	static void radeon_cs_parser_fini(struct radeon_cs_parser *parser, int error)
	286	{
	287	unsigned i;
	288
147666fb TH	289
	290	if (!error && parser->ib)
	291	ttm_eu_fence_buffer_objects(&parser->validated,
	292	parser->ib->fence);
	293	else
	294	ttm_eu_backoff_reservation(&parser->validated);
	295
fcbc451b PN	296	if (parser->relocs != NULL) {
	297	for (i = 0; i < parser->nrelocs; i++) {
	298	if (parser->relocs[i].gobj)
	299	drm_gem_object_unreference_unlocked(parser->relocs[i].gobj);
	300	}
771fe6b9	301	}
48e113e5	302	kfree(parser->track);
771fe6b9 JG	303	kfree(parser->relocs);
	304	kfree(parser->relocs_ptr);
	305	for (i = 0; i < parser->nchunks; i++) {
	306	kfree(parser->chunks[i].kdata);
513bcb46 DA	307	kfree(parser->chunks[i].kpage[0]);
513bcb46 DA	308	kfree(parser->chunks[i].kpage[1]);
771fe6b9 JG	309	}
	310	kfree(parser->chunks);
	311	kfree(parser->chunks_array);
	312	radeon_ib_free(parser->rdev, &parser->ib);
	313	}
	314
721604a1 JG	315	static int radeon_cs_ib_chunk(struct radeon_device *rdev,
	316	struct radeon_cs_parser *parser)
	317	{
	318	struct radeon_cs_chunk *ib_chunk;
	319	int r;
	320
	321	if (parser->chunk_ib_idx == -1)
	322	return 0;
	323
	324	if (parser->cs_flags & RADEON_CS_USE_VM)
	325	return 0;
	326
	327	ib_chunk = &parser->chunks[parser->chunk_ib_idx];
	328	/* Copy the packet into the IB, the parser will read from the
	329	* input memory (cached) and write to the IB (which can be
	330	* uncached).
	331	*/
	332	r = radeon_ib_get(rdev, parser->ring, &parser->ib,
	333	ib_chunk->length_dw * 4);
	334	if (r) {
	335	DRM_ERROR("Failed to get ib !\n");
	336	return r;
	337	}
	338	parser->ib->length_dw = ib_chunk->length_dw;
eb0c19c5	339	r = radeon_cs_parse(rdev, parser->ring, parser);
721604a1 JG	340	if (r \|\| parser->parser_error) {
	341	DRM_ERROR("Invalid command stream !\n");
	342	return r;
	343	}
	344	r = radeon_cs_finish_pages(parser);
	345	if (r) {
	346	DRM_ERROR("Invalid command stream !\n");
	347	return r;
	348	}
93504fce CK	349	r = radeon_cs_sync_rings(parser);
	350	if (r) {
	351	DRM_ERROR("Failed to synchronize rings !\n");
	352	}
721604a1 JG	353	parser->ib->vm_id = 0;
	354	r = radeon_ib_schedule(rdev, parser->ib);
	355	if (r) {
	356	DRM_ERROR("Failed to schedule IB !\n");
	357	}
	358	return 0;
	359	}
	360
	361	static int radeon_bo_vm_update_pte(struct radeon_cs_parser *parser,
	362	struct radeon_vm *vm)
	363	{
	364	struct radeon_bo_list *lobj;
	365	struct radeon_bo *bo;
	366	int r;
	367
	368	list_for_each_entry(lobj, &parser->validated, tv.head) {
	369	bo = lobj->bo;
	370	r = radeon_vm_bo_update_pte(parser->rdev, vm, bo, &bo->tbo.mem);
	371	if (r) {
	372	return r;
	373	}
	374	}
	375	return 0;
	376	}
	377
	378	static int radeon_cs_ib_vm_chunk(struct radeon_device *rdev,
	379	struct radeon_cs_parser *parser)
	380	{
	381	struct radeon_cs_chunk *ib_chunk;
	382	struct radeon_fpriv *fpriv = parser->filp->driver_priv;
	383	struct radeon_vm *vm = &fpriv->vm;
	384	int r;
	385
	386	if (parser->chunk_ib_idx == -1)
	387	return 0;
	388
	389	if ((parser->cs_flags & RADEON_CS_USE_VM) == 0)
	390	return 0;
	391
	392	ib_chunk = &parser->chunks[parser->chunk_ib_idx];
	393	if (ib_chunk->length_dw > RADEON_IB_VM_MAX_SIZE) {
	394	DRM_ERROR("cs IB too big: %d\n", ib_chunk->length_dw);
	395	return -EINVAL;
	396	}
	397	r = radeon_ib_get(rdev, parser->ring, &parser->ib,
	398	ib_chunk->length_dw * 4);
	399	if (r) {
	400	DRM_ERROR("Failed to get ib !\n");
	401	return r;
	402	}
	403	parser->ib->length_dw = ib_chunk->length_dw;
	404	/* Copy the packet into the IB */
	405	if (DRM_COPY_FROM_USER(parser->ib->ptr, ib_chunk->user_ptr,
	406	ib_chunk->length_dw * 4)) {
	407	return -EFAULT;
	408	}
	409	r = radeon_ring_ib_parse(rdev, parser->ring, parser->ib);
	410	if (r) {
	411	return r;
	412	}
	413
	414	mutex_lock(&vm->mutex);
	415	r = radeon_vm_bind(rdev, vm);
	416	if (r) {
417	goto out;
418	}
419	r = radeon_bo_vm_update_pte(parser, vm);
420	if (r) {
421	goto out;
422	}
93504fce CK	423	r = radeon_cs_sync_rings(parser);
	424	if (r) {
	425	DRM_ERROR("Failed to synchronize rings !\n");
	426	}
721604a1 JG	427	parser->ib->vm_id = vm->id;
	428	/* ib pool is bind at 0 in virtual address space to gpu_addr is the
	429	* offset inside the pool bo
	430	*/
	431	parser->ib->gpu_addr = parser->ib->sa_bo.offset;
	432	r = radeon_ib_schedule(rdev, parser->ib);
	433	out:
	434	if (!r) {
	435	if (vm->fence) {
	436	radeon_fence_unref(&vm->fence);
	437	}
	438	vm->fence = radeon_fence_ref(parser->ib->fence);
	439	}
	440	mutex_unlock(&fpriv->vm.mutex);
	441	return r;
	442	}
	443
771fe6b9 JG	444	int radeon_cs_ioctl(struct drm_device dev, void data, struct drm_file *filp)
	445	{
	446	struct radeon_device *rdev = dev->dev_private;
	447	struct radeon_cs_parser parser;
771fe6b9 JG	448	int r;
771fe6b9 JG	449
7a1619b9	450	radeon_mutex_lock(&rdev->cs_mutex);
6b7746e8 JG	451	if (!rdev->accel_working) {
	452	radeon_mutex_unlock(&rdev->cs_mutex);
	453	return -EBUSY;
	454	}
771fe6b9 JG	455	/* initialize parser */
	456	memset(&parser, 0, sizeof(struct radeon_cs_parser));
	457	parser.filp = filp;
	458	parser.rdev = rdev;
c8c15ff1	459	parser.dev = rdev->dev;
428c6e36	460	parser.family = rdev->family;
771fe6b9 JG	461	r = radeon_cs_parser_init(&parser, data);
	462	if (r) {
	463	DRM_ERROR("Failed to initialize parser !\n");
	464	radeon_cs_parser_fini(&parser, r);
7a1619b9	465	radeon_mutex_unlock(&rdev->cs_mutex);
771fe6b9 JG	466	return r;
771fe6b9 JG	467	}
771fe6b9 JG	468	r = radeon_cs_parser_relocs(&parser);
771fe6b9 JG	469	if (r) {
97f23b3d DA	470	if (r != -ERESTARTSYS)
97f23b3d DA	471	DRM_ERROR("Failed to parse relocation %d!\n", r);
771fe6b9	472	radeon_cs_parser_fini(&parser, r);
7a1619b9	473	radeon_mutex_unlock(&rdev->cs_mutex);
771fe6b9 JG	474	return r;
771fe6b9 JG	475	}
721604a1	476	r = radeon_cs_ib_chunk(rdev, &parser);
771fe6b9	477	if (r) {
721604a1	478	goto out;
771fe6b9	479	}
721604a1	480	r = radeon_cs_ib_vm_chunk(rdev, &parser);
771fe6b9	481	if (r) {
721604a1	482	goto out;
771fe6b9	483	}
721604a1	484	out:
771fe6b9	485	radeon_cs_parser_fini(&parser, r);
7a1619b9	486	radeon_mutex_unlock(&rdev->cs_mutex);
771fe6b9 JG	487	return r;
771fe6b9 JG	488	}
513bcb46 DA	489
	490	int radeon_cs_finish_pages(struct radeon_cs_parser *p)
	491	{
	492	struct radeon_cs_chunk *ibc = &p->chunks[p->chunk_ib_idx];
	493	int i;
	494	int size = PAGE_SIZE;
	495
	496	for (i = ibc->last_copied_page + 1; i <= ibc->last_page_index; i++) {
	497	if (i == ibc->last_page_index) {
	498	size = (ibc->length_dw * 4) % PAGE_SIZE;
	499	if (size == 0)
	500	size = PAGE_SIZE;
	501	}
	502
	503	if (DRM_COPY_FROM_USER(p->ib->ptr + (i * (PAGE_SIZE/4)),
	504	ibc->user_ptr + (i * PAGE_SIZE),
	505	size))
	506	return -EFAULT;
	507	}
	508	return 0;
	509	}
	510
	511	int radeon_cs_update_pages(struct radeon_cs_parser *p, int pg_idx)
	512	{
	513	int new_page;
513bcb46 DA	514	struct radeon_cs_chunk *ibc = &p->chunks[p->chunk_ib_idx];
	515	int i;
	516	int size = PAGE_SIZE;
	517
c5e617e2	518	for (i = ibc->last_copied_page + 1; i < pg_idx; i++) {
513bcb46 DA	519	if (DRM_COPY_FROM_USER(p->ib->ptr + (i * (PAGE_SIZE/4)),
	520	ibc->user_ptr + (i * PAGE_SIZE),
	521	PAGE_SIZE)) {
	522	p->parser_error = -EFAULT;
	523	return 0;
	524	}
	525	}
	526
	527	new_page = ibc->kpage_idx[0] < ibc->kpage_idx[1] ? 0 : 1;
	528
	529	if (pg_idx == ibc->last_page_index) {
	530	size = (ibc->length_dw * 4) % PAGE_SIZE;
	531	if (size == 0)
	532	size = PAGE_SIZE;
	533	}
	534
	535	if (DRM_COPY_FROM_USER(ibc->kpage[new_page],
	536	ibc->user_ptr + (pg_idx * PAGE_SIZE),
	537	size)) {
	538	p->parser_error = -EFAULT;
	539	return 0;
	540	}
	541
	542	/* copy to IB here */
	543	memcpy((void )(p->ib->ptr+(pg_idx(PAGE_SIZE/4))), ibc->kpage[new_page], size);
	544
	545	ibc->last_copied_page = pg_idx;
	546	ibc->kpage_idx[new_page] = pg_idx;
	547
	548	return new_page;
	549	}