amdgpu_uvd.c

/*
 * Copyright 2011 Advanced Micro Devices, Inc.
 * 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, sub license, 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 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 NON-INFRINGEMENT. IN NO EVENT SHALL
 * THE COPYRIGHT HOLDERS, AUTHORS 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.
 *
 * The above copyright notice and this permission notice (including the
 * next paragraph) shall be included in all copies or substantial portions
 * of the Software.
 *
 */
/*
 * Authors:
 *    Christian König <deathsimple@vodafone.de>
 */

#include <linux/firmware.h>
#include <linux/module.h>
#include <drm/drmP.h>
#include <drm/drm.h>

#include "amdgpu.h"
#include "amdgpu_pm.h"
#include "amdgpu_uvd.h"
#include "cikd.h"
#include "uvd/uvd_4_2_d.h"

/* 1 second timeout */
#define UVD_IDLE_TIMEOUT_MS	1000

/* Firmware Names */
#ifdef CONFIG_DRM_AMDGPU_CIK
#define FIRMWARE_BONAIRE	"radeon/bonaire_uvd.bin"
#define FIRMWARE_KABINI 	"radeon/kabini_uvd.bin"
#define FIRMWARE_KAVERI 	"radeon/kaveri_uvd.bin"
#define FIRMWARE_HAWAII 	"radeon/hawaii_uvd.bin"
#define FIRMWARE_MULLINS	"radeon/mullins_uvd.bin"
#endif
#define FIRMWARE_TONGA		"amdgpu/tonga_uvd.bin"
#define FIRMWARE_CARRIZO	"amdgpu/carrizo_uvd.bin"
#define FIRMWARE_FIJI		"amdgpu/fiji_uvd.bin"
#define FIRMWARE_STONEY		"amdgpu/stoney_uvd.bin"

/**
 * amdgpu_uvd_cs_ctx - Command submission parser context
 *
 * Used for emulating virtual memory support on UVD 4.2.
 */
struct amdgpu_uvd_cs_ctx {
	struct amdgpu_cs_parser *parser;
	unsigned reg, count;
	unsigned data0, data1;
	unsigned idx;
	unsigned ib_idx;

	/* does the IB has a msg command */
	bool has_msg_cmd;

	/* minimum buffer sizes */
	unsigned *buf_sizes;
};

#ifdef CONFIG_DRM_AMDGPU_CIK
MODULE_FIRMWARE(FIRMWARE_BONAIRE);
MODULE_FIRMWARE(FIRMWARE_KABINI);
MODULE_FIRMWARE(FIRMWARE_KAVERI);
MODULE_FIRMWARE(FIRMWARE_HAWAII);
MODULE_FIRMWARE(FIRMWARE_MULLINS);
#endif
MODULE_FIRMWARE(FIRMWARE_TONGA);
MODULE_FIRMWARE(FIRMWARE_CARRIZO);
MODULE_FIRMWARE(FIRMWARE_FIJI);
MODULE_FIRMWARE(FIRMWARE_STONEY);

static void amdgpu_uvd_note_usage(struct amdgpu_device *adev);
static void amdgpu_uvd_idle_work_handler(struct work_struct *work);

int amdgpu_uvd_sw_init(struct amdgpu_device *adev)
{
	struct amdgpu_ring *ring;
	struct amd_sched_rq *rq;
	unsigned long bo_size;
	const char *fw_name;
	const struct common_firmware_header *hdr;
	unsigned version_major, version_minor, family_id;
	int i, r;

	INIT_DELAYED_WORK(&adev->uvd.idle_work, amdgpu_uvd_idle_work_handler);

	switch (adev->asic_type) {
#ifdef CONFIG_DRM_AMDGPU_CIK
	case CHIP_BONAIRE:
		fw_name = FIRMWARE_BONAIRE;
		break;
	case CHIP_KABINI:
		fw_name = FIRMWARE_KABINI;
		break;
	case CHIP_KAVERI:
		fw_name = FIRMWARE_KAVERI;
		break;
	case CHIP_HAWAII:
		fw_name = FIRMWARE_HAWAII;
		break;
	case CHIP_MULLINS:
		fw_name = FIRMWARE_MULLINS;
		break;
#endif
	case CHIP_TONGA:
		fw_name = FIRMWARE_TONGA;
		break;
	case CHIP_FIJI:
		fw_name = FIRMWARE_FIJI;
		break;
	case CHIP_CARRIZO:
		fw_name = FIRMWARE_CARRIZO;
		break;
	case CHIP_STONEY:
		fw_name = FIRMWARE_STONEY;
		break;
	default:
		return -EINVAL;
	}

	r = request_firmware(&adev->uvd.fw, fw_name, adev->dev);
	if (r) {
		dev_err(adev->dev, "amdgpu_uvd: Can't load firmware \"%s\"\n",
			fw_name);
		return r;
	}

	r = amdgpu_ucode_validate(adev->uvd.fw);
	if (r) {
		dev_err(adev->dev, "amdgpu_uvd: Can't validate firmware \"%s\"\n",
			fw_name);
		release_firmware(adev->uvd.fw);
		adev->uvd.fw = NULL;
		return r;
	}

	hdr = (const struct common_firmware_header *)adev->uvd.fw->data;
	family_id = le32_to_cpu(hdr->ucode_version) & 0xff;
	version_major = (le32_to_cpu(hdr->ucode_version) >> 24) & 0xff;
	version_minor = (le32_to_cpu(hdr->ucode_version) >> 8) & 0xff;
	DRM_INFO("Found UVD firmware Version: %hu.%hu Family ID: %hu\n",
		version_major, version_minor, family_id);

	adev->uvd.fw_version = ((version_major << 24) | (version_minor << 16) |
				(family_id << 8));

	bo_size = AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8)
		 +  AMDGPU_UVD_STACK_SIZE + AMDGPU_UVD_HEAP_SIZE;
	r = amdgpu_bo_create(adev, bo_size, PAGE_SIZE, true,
			     AMDGPU_GEM_DOMAIN_VRAM,
			     AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED,
			     NULL, NULL, &adev->uvd.vcpu_bo);
	if (r) {
		dev_err(adev->dev, "(%d) failed to allocate UVD bo\n", r);
		return r;
	}

	r = amdgpu_bo_reserve(adev->uvd.vcpu_bo, false);
	if (r) {
		amdgpu_bo_unref(&adev->uvd.vcpu_bo);
		dev_err(adev->dev, "(%d) failed to reserve UVD bo\n", r);
		return r;
	}

	r = amdgpu_bo_pin(adev->uvd.vcpu_bo, AMDGPU_GEM_DOMAIN_VRAM,
			  &adev->uvd.gpu_addr);
	if (r) {
		amdgpu_bo_unreserve(adev->uvd.vcpu_bo);
		amdgpu_bo_unref(&adev->uvd.vcpu_bo);
		dev_err(adev->dev, "(%d) UVD bo pin failed\n", r);
		return r;
	}

	r = amdgpu_bo_kmap(adev->uvd.vcpu_bo, &adev->uvd.cpu_addr);
	if (r) {
		dev_err(adev->dev, "(%d) UVD map failed\n", r);
		return r;
	}

	amdgpu_bo_unreserve(adev->uvd.vcpu_bo);

	ring = &adev->uvd.ring;
	rq = &ring->sched.sched_rq[AMD_SCHED_PRIORITY_NORMAL];
	r = amd_sched_entity_init(&ring->sched, &adev->uvd.entity,
				  rq, amdgpu_sched_jobs);
	if (r != 0) {
		DRM_ERROR("Failed setting up UVD run queue.\n");
		return r;
	}

	for (i = 0; i < AMDGPU_MAX_UVD_HANDLES; ++i) {
		atomic_set(&adev->uvd.handles[i], 0);
		adev->uvd.filp[i] = NULL;
	}

	/* from uvd v5.0 HW addressing capacity increased to 64 bits */
	if (!amdgpu_ip_block_version_cmp(adev, AMD_IP_BLOCK_TYPE_UVD, 5, 0))
		adev->uvd.address_64_bit = true;

	return 0;
}

int amdgpu_uvd_sw_fini(struct amdgpu_device *adev)
{
	int r;

	if (adev->uvd.vcpu_bo == NULL)
		return 0;

	amd_sched_entity_fini(&adev->uvd.ring.sched, &adev->uvd.entity);

	r = amdgpu_bo_reserve(adev->uvd.vcpu_bo, false);
	if (!r) {
		amdgpu_bo_kunmap(adev->uvd.vcpu_bo);
		amdgpu_bo_unpin(adev->uvd.vcpu_bo);
		amdgpu_bo_unreserve(adev->uvd.vcpu_bo);
	}

	amdgpu_bo_unref(&adev->uvd.vcpu_bo);

	amdgpu_ring_fini(&adev->uvd.ring);

	release_firmware(adev->uvd.fw);

	return 0;
}

int amdgpu_uvd_suspend(struct amdgpu_device *adev)
{
	unsigned size;
	void *ptr;
	int i;

	if (adev->uvd.vcpu_bo == NULL)
		return 0;

	for (i = 0; i < AMDGPU_MAX_UVD_HANDLES; ++i)
		if (atomic_read(&adev->uvd.handles[i]))
			break;

	if (i == AMDGPU_MAX_UVD_HANDLES)
		return 0;

	cancel_delayed_work_sync(&adev->uvd.idle_work);

	size = amdgpu_bo_size(adev->uvd.vcpu_bo);
	ptr = adev->uvd.cpu_addr;

	adev->uvd.saved_bo = kmalloc(size, GFP_KERNEL);
	if (!adev->uvd.saved_bo)
		return -ENOMEM;

	memcpy(adev->uvd.saved_bo, ptr, size);

	return 0;
}

int amdgpu_uvd_resume(struct amdgpu_device *adev)
{
	unsigned size;
	void *ptr;

	if (adev->uvd.vcpu_bo == NULL)
		return -EINVAL;

	size = amdgpu_bo_size(adev->uvd.vcpu_bo);
	ptr = adev->uvd.cpu_addr;

	if (adev->uvd.saved_bo != NULL) {
		memcpy(ptr, adev->uvd.saved_bo, size);
		kfree(adev->uvd.saved_bo);
		adev->uvd.saved_bo = NULL;
	} else {
		const struct common_firmware_header *hdr;
		unsigned offset;

		hdr = (const struct common_firmware_header *)adev->uvd.fw->data;
		offset = le32_to_cpu(hdr->ucode_array_offset_bytes);
		memcpy(adev->uvd.cpu_addr, (adev->uvd.fw->data) + offset,
			(adev->uvd.fw->size) - offset);
		size -= le32_to_cpu(hdr->ucode_size_bytes);
		ptr += le32_to_cpu(hdr->ucode_size_bytes);
		memset(ptr, 0, size);
	}

	return 0;
}

void amdgpu_uvd_free_handles(struct amdgpu_device *adev, struct drm_file *filp)
{
	struct amdgpu_ring *ring = &adev->uvd.ring;
	int i, r;

	for (i = 0; i < AMDGPU_MAX_UVD_HANDLES; ++i) {
		uint32_t handle = atomic_read(&adev->uvd.handles[i]);
		if (handle != 0 && adev->uvd.filp[i] == filp) {
			struct fence *fence;

			amdgpu_uvd_note_usage(adev);

			r = amdgpu_uvd_get_destroy_msg(ring, handle,
						       false, &fence);
			if (r) {
				DRM_ERROR("Error destroying UVD (%d)!\n", r);
				continue;
			}

			fence_wait(fence, false);
			fence_put(fence);

			adev->uvd.filp[i] = NULL;
			atomic_set(&adev->uvd.handles[i], 0);
		}
	}
}

static void amdgpu_uvd_force_into_uvd_segment(struct amdgpu_bo *rbo)
{
	int i;
	for (i = 0; i < rbo->placement.num_placement; ++i) {
		rbo->placements[i].fpfn = 0 >> PAGE_SHIFT;
		rbo->placements[i].lpfn = (256 * 1024 * 1024) >> PAGE_SHIFT;
	}
}

/**
 * amdgpu_uvd_cs_pass1 - first parsing round
 *
 * @ctx: UVD parser context
 *
 * Make sure UVD message and feedback buffers are in VRAM and
 * nobody is violating an 256MB boundary.
 */
static int amdgpu_uvd_cs_pass1(struct amdgpu_uvd_cs_ctx *ctx)
{
	struct amdgpu_bo_va_mapping *mapping;
	struct amdgpu_bo *bo;
	uint32_t cmd, lo, hi;
	uint64_t addr;
	int r = 0;

	lo = amdgpu_get_ib_value(ctx->parser, ctx->ib_idx, ctx->data0);
	hi = amdgpu_get_ib_value(ctx->parser, ctx->ib_idx, ctx->data1);
	addr = ((uint64_t)lo) | (((uint64_t)hi) << 32);

	mapping = amdgpu_cs_find_mapping(ctx->parser, addr, &bo);
	if (mapping == NULL) {
		DRM_ERROR("Can't find BO for addr 0x%08Lx\n", addr);
		return -EINVAL;
	}

	if (!ctx->parser->adev->uvd.address_64_bit) {
		/* check if it's a message or feedback command */
		cmd = amdgpu_get_ib_value(ctx->parser, ctx->ib_idx, ctx->idx) >> 1;
		if (cmd == 0x0 || cmd == 0x3) {
			/* yes, force it into VRAM */
			uint32_t domain = AMDGPU_GEM_DOMAIN_VRAM;
			amdgpu_ttm_placement_from_domain(bo, domain);
		}
		amdgpu_uvd_force_into_uvd_segment(bo);

		r = ttm_bo_validate(&bo->tbo, &bo->placement, false, false);
	}

	return r;
}

/**
 * amdgpu_uvd_cs_msg_decode - handle UVD decode message
 *
 * @msg: pointer to message structure
 * @buf_sizes: returned buffer sizes
 *
 * Peek into the decode message and calculate the necessary buffer sizes.
 */
static int amdgpu_uvd_cs_msg_decode(uint32_t *msg, unsigned buf_sizes[])
{
	unsigned stream_type = msg[4];
	unsigned width = msg[6];
	unsigned height = msg[7];
	unsigned dpb_size = msg[9];
	unsigned pitch = msg[28];
	unsigned level = msg[57];

	unsigned width_in_mb = width / 16;
	unsigned height_in_mb = ALIGN(height / 16, 2);
	unsigned fs_in_mb = width_in_mb * height_in_mb;

	unsigned image_size, tmp, min_dpb_size, num_dpb_buffer;
	unsigned min_ctx_size = 0;

	image_size = width * height;
	image_size += image_size / 2;
	image_size = ALIGN(image_size, 1024);

	switch (stream_type) {
	case 0: /* H264 */
	case 7: /* H264 Perf */
		switch(level) {
		case 30:
			num_dpb_buffer = 8100 / fs_in_mb;
			break;
		case 31:
			num_dpb_buffer = 18000 / fs_in_mb;
			break;
		case 32:
			num_dpb_buffer = 20480 / fs_in_mb;
			break;
		case 41:
			num_dpb_buffer = 32768 / fs_in_mb;
			break;
		case 42:
			num_dpb_buffer = 34816 / fs_in_mb;
			break;
		case 50:
			num_dpb_buffer = 110400 / fs_in_mb;
			break;
		case 51:
			num_dpb_buffer = 184320 / fs_in_mb;
			break;
		default:
			num_dpb_buffer = 184320 / fs_in_mb;
			break;
		}
		num_dpb_buffer++;
		if (num_dpb_buffer > 17)
			num_dpb_buffer = 17;

		/* reference picture buffer */
		min_dpb_size = image_size * num_dpb_buffer;

		/* macroblock context buffer */
		min_dpb_size += width_in_mb * height_in_mb * num_dpb_buffer * 192;

		/* IT surface buffer */
		min_dpb_size += width_in_mb * height_in_mb * 32;
		break;

	case 1: /* VC1 */

		/* reference picture buffer */
		min_dpb_size = image_size * 3;

		/* CONTEXT_BUFFER */
		min_dpb_size += width_in_mb * height_in_mb * 128;

		/* IT surface buffer */
		min_dpb_size += width_in_mb * 64;

		/* DB surface buffer */
		min_dpb_size += width_in_mb * 128;

		/* BP */
		tmp = max(width_in_mb, height_in_mb);
		min_dpb_size += ALIGN(tmp * 7 * 16, 64);
		break;

	case 3: /* MPEG2 */

		/* reference picture buffer */
		min_dpb_size = image_size * 3;
		break;

	case 4: /* MPEG4 */

		/* reference picture buffer */
		min_dpb_size = image_size * 3;

		/* CM */
		min_dpb_size += width_in_mb * height_in_mb * 64;

		/* IT surface buffer */
		min_dpb_size += ALIGN(width_in_mb * height_in_mb * 32, 64);
		break;

	case 16: /* H265 */
		image_size = (ALIGN(width, 16) * ALIGN(height, 16) * 3) / 2;
		image_size = ALIGN(image_size, 256);

		num_dpb_buffer = (le32_to_cpu(msg[59]) & 0xff) + 2;
		min_dpb_size = image_size * num_dpb_buffer;
		min_ctx_size = ((width + 255) / 16) * ((height + 255) / 16)
					   * 16 * num_dpb_buffer + 52 * 1024;
		break;

	default:
		DRM_ERROR("UVD codec not handled %d!\n", stream_type);
		return -EINVAL;
	}

	if (width > pitch) {
		DRM_ERROR("Invalid UVD decoding target pitch!\n");
		return -EINVAL;
	}

	if (dpb_size < min_dpb_size) {
		DRM_ERROR("Invalid dpb_size in UVD message (%d / %d)!\n",
			  dpb_size, min_dpb_size);
		return -EINVAL;
	}

	buf_sizes[0x1] = dpb_size;
	buf_sizes[0x2] = image_size;
	buf_sizes[0x4] = min_ctx_size;
	return 0;
}

/**
 * amdgpu_uvd_cs_msg - handle UVD message
 *
 * @ctx: UVD parser context
 * @bo: buffer object containing the message
 * @offset: offset into the buffer object
 *
 * Peek into the UVD message and extract the session id.
 * Make sure that we don't open up to many sessions.
 */
static int amdgpu_uvd_cs_msg(struct amdgpu_uvd_cs_ctx *ctx,
			     struct amdgpu_bo *bo, unsigned offset)
{
	struct amdgpu_device *adev = ctx->parser->adev;
	int32_t *msg, msg_type, handle;
	void *ptr;
	long r;
	int i;

	if (offset & 0x3F) {
		DRM_ERROR("UVD messages must be 64 byte aligned!\n");
		return -EINVAL;
	}

	r = amdgpu_bo_kmap(bo, &ptr);
	if (r) {
		DRM_ERROR("Failed mapping the UVD message (%ld)!\n", r);
		return r;
	}

	msg = ptr + offset;

	msg_type = msg[1];
	handle = msg[2];

	if (handle == 0) {
		DRM_ERROR("Invalid UVD handle!\n");
		return -EINVAL;
	}

	switch (msg_type) {
	case 0:
		/* it's a create msg, calc image size (width * height) */
		amdgpu_bo_kunmap(bo);

		/* try to alloc a new handle */
		for (i = 0; i < AMDGPU_MAX_UVD_HANDLES; ++i) {
			if (atomic_read(&adev->uvd.handles[i]) == handle) {
				DRM_ERROR("Handle 0x%x already in use!\n", handle);
				return -EINVAL;
			}

			if (!atomic_cmpxchg(&adev->uvd.handles[i], 0, handle)) {
				adev->uvd.filp[i] = ctx->parser->filp;
				return 0;
			}
		}

		DRM_ERROR("No more free UVD handles!\n");
		return -EINVAL;

	case 1:
		/* it's a decode msg, calc buffer sizes */
		r = amdgpu_uvd_cs_msg_decode(msg, ctx->buf_sizes);
		amdgpu_bo_kunmap(bo);
		if (r)
			return r;

		/* validate the handle */
		for (i = 0; i < AMDGPU_MAX_UVD_HANDLES; ++i) {
			if (atomic_read(&adev->uvd.handles[i]) == handle) {
				if (adev->uvd.filp[i] != ctx->parser->filp) {
					DRM_ERROR("UVD handle collision detected!\n");
					return -EINVAL;
				}
				return 0;
			}
		}

		DRM_ERROR("Invalid UVD handle 0x%x!\n", handle);
		return -ENOENT;

	case 2:
		/* it's a destroy msg, free the handle */
		for (i = 0; i < AMDGPU_MAX_UVD_HANDLES; ++i)
			atomic_cmpxchg(&adev->uvd.handles[i], handle, 0);
		amdgpu_bo_kunmap(bo);
		return 0;

	default:
		DRM_ERROR("Illegal UVD message type (%d)!\n", msg_type);
		return -EINVAL;
	}
	BUG();
	return -EINVAL;
}

/**
 * amdgpu_uvd_cs_pass2 - second parsing round
 *
 * @ctx: UVD parser context
 *
 * Patch buffer addresses, make sure buffer sizes are correct.
 */
static int amdgpu_uvd_cs_pass2(struct amdgpu_uvd_cs_ctx *ctx)
{
	struct amdgpu_bo_va_mapping *mapping;
	struct amdgpu_bo *bo;
	uint32_t cmd, lo, hi;
	uint64_t start, end;
	uint64_t addr;
	int r;

	lo = amdgpu_get_ib_value(ctx->parser, ctx->ib_idx, ctx->data0);
	hi = amdgpu_get_ib_value(ctx->parser, ctx->ib_idx, ctx->data1);
	addr = ((uint64_t)lo) | (((uint64_t)hi) << 32);

	mapping = amdgpu_cs_find_mapping(ctx->parser, addr, &bo);
	if (mapping == NULL)
		return -EINVAL;

	start = amdgpu_bo_gpu_offset(bo);

	end = (mapping->it.last + 1 - mapping->it.start);
	end = end * AMDGPU_GPU_PAGE_SIZE + start;

	addr -= ((uint64_t)mapping->it.start) * AMDGPU_GPU_PAGE_SIZE;
	start += addr;

	amdgpu_set_ib_value(ctx->parser, ctx->ib_idx, ctx->data0,
			    lower_32_bits(start));
	amdgpu_set_ib_value(ctx->parser, ctx->ib_idx, ctx->data1,
			    upper_32_bits(start));

	cmd = amdgpu_get_ib_value(ctx->parser, ctx->ib_idx, ctx->idx) >> 1;
	if (cmd < 0x4) {
		if ((end - start) < ctx->buf_sizes[cmd]) {
			DRM_ERROR("buffer (%d) to small (%d / %d)!\n", cmd,
				  (unsigned)(end - start),
				  ctx->buf_sizes[cmd]);
			return -EINVAL;
		}

	} else if (cmd == 0x206) {
		if ((end - start) < ctx->buf_sizes[4]) {
			DRM_ERROR("buffer (%d) to small (%d / %d)!\n", cmd,
					  (unsigned)(end - start),
					  ctx->buf_sizes[4]);
			return -EINVAL;
		}
	} else if ((cmd != 0x100) && (cmd != 0x204)) {
		DRM_ERROR("invalid UVD command %X!\n", cmd);
		return -EINVAL;
	}

	if (!ctx->parser->adev->uvd.address_64_bit) {
		if ((start >> 28) != ((end - 1) >> 28)) {
			DRM_ERROR("reloc %LX-%LX crossing 256MB boundary!\n",
				  start, end);
			return -EINVAL;
		}

		if ((cmd == 0 || cmd == 0x3) &&
		    (start >> 28) != (ctx->parser->adev->uvd.gpu_addr >> 28)) {
			DRM_ERROR("msg/fb buffer %LX-%LX out of 256MB segment!\n",
				  start, end);
			return -EINVAL;
		}
	}

	if (cmd == 0) {
		ctx->has_msg_cmd = true;
		r = amdgpu_uvd_cs_msg(ctx, bo, addr);
		if (r)
			return r;
	} else if (!ctx->has_msg_cmd) {
		DRM_ERROR("Message needed before other commands are send!\n");
		return -EINVAL;
	}

	return 0;
}

/**
 * amdgpu_uvd_cs_reg - parse register writes
 *
 * @ctx: UVD parser context
 * @cb: callback function
 *
 * Parse the register writes, call cb on each complete command.
 */
static int amdgpu_uvd_cs_reg(struct amdgpu_uvd_cs_ctx *ctx,
			     int (*cb)(struct amdgpu_uvd_cs_ctx *ctx))
{
	struct amdgpu_ib *ib = &ctx->parser->job->ibs[ctx->ib_idx];
	int i, r;

	ctx->idx++;
	for (i = 0; i <= ctx->count; ++i) {
		unsigned reg = ctx->reg + i;

		if (ctx->idx >= ib->length_dw) {
			DRM_ERROR("Register command after end of CS!\n");
			return -EINVAL;
		}

		switch (reg) {
		case mmUVD_GPCOM_VCPU_DATA0:
			ctx->data0 = ctx->idx;
			break;
		case mmUVD_GPCOM_VCPU_DATA1:
			ctx->data1 = ctx->idx;
			break;
		case mmUVD_GPCOM_VCPU_CMD:
			r = cb(ctx);
			if (r)
				return r;
			break;
		case mmUVD_ENGINE_CNTL:
			break;
		default:
			DRM_ERROR("Invalid reg 0x%X!\n", reg);
			return -EINVAL;
		}
		ctx->idx++;
	}
	return 0;
}

/**
 * amdgpu_uvd_cs_packets - parse UVD packets
 *
 * @ctx: UVD parser context
 * @cb: callback function
 *
 * Parse the command stream packets.
 */
static int amdgpu_uvd_cs_packets(struct amdgpu_uvd_cs_ctx *ctx,
				 int (*cb)(struct amdgpu_uvd_cs_ctx *ctx))
{
	struct amdgpu_ib *ib = &ctx->parser->job->ibs[ctx->ib_idx];
	int r;

	for (ctx->idx = 0 ; ctx->idx < ib->length_dw; ) {
		uint32_t cmd = amdgpu_get_ib_value(ctx->parser, ctx->ib_idx, ctx->idx);
		unsigned type = CP_PACKET_GET_TYPE(cmd);
		switch (type) {
		case PACKET_TYPE0:
			ctx->reg = CP_PACKET0_GET_REG(cmd);
			ctx->count = CP_PACKET_GET_COUNT(cmd);
			r = amdgpu_uvd_cs_reg(ctx, cb);
			if (r)
				return r;
			break;
		case PACKET_TYPE2:
			++ctx->idx;
			break;
		default:
			DRM_ERROR("Unknown packet type %d !\n", type);
			return -EINVAL;
		}
	}
	return 0;
}

/**
 * amdgpu_uvd_ring_parse_cs - UVD command submission parser
 *
 * @parser: Command submission parser context
 *
 * Parse the command stream, patch in addresses as necessary.
 */
int amdgpu_uvd_ring_parse_cs(struct amdgpu_cs_parser *parser, uint32_t ib_idx)
{
	struct amdgpu_uvd_cs_ctx ctx = {};
	unsigned buf_sizes[] = {
		[0x00000000]	=	2048,
		[0x00000001]	=	0xFFFFFFFF,
		[0x00000002]	=	0xFFFFFFFF,
		[0x00000003]	=	2048,
		[0x00000004]	=	0xFFFFFFFF,
	};
	struct amdgpu_ib *ib = &parser->job->ibs[ib_idx];
	int r;

	if (ib->length_dw % 16) {
		DRM_ERROR("UVD IB length (%d) not 16 dwords aligned!\n",
			  ib->length_dw);
		return -EINVAL;
	}

	ctx.parser = parser;
	ctx.buf_sizes = buf_sizes;
	ctx.ib_idx = ib_idx;

	/* first round, make sure the buffers are actually in the UVD segment */
	r = amdgpu_uvd_cs_packets(&ctx, amdgpu_uvd_cs_pass1);
	if (r)
		return r;

	/* second round, patch buffer addresses into the command stream */
	r = amdgpu_uvd_cs_packets(&ctx, amdgpu_uvd_cs_pass2);
	if (r)
		return r;

	if (!ctx.has_msg_cmd) {
		DRM_ERROR("UVD-IBs need a msg command!\n");
		return -EINVAL;
	}

	amdgpu_uvd_note_usage(ctx.parser->adev);

	return 0;
}

static int amdgpu_uvd_send_msg(struct amdgpu_ring *ring, struct amdgpu_bo *bo,
			       bool direct, struct fence **fence)
{
	struct ttm_validate_buffer tv;
	struct ww_acquire_ctx ticket;
	struct list_head head;
	struct amdgpu_job *job;
	struct amdgpu_ib *ib;
	struct fence *f = NULL;
	struct amdgpu_device *adev = ring->adev;
	uint64_t addr;
	int i, r;

	memset(&tv, 0, sizeof(tv));
	tv.bo = &bo->tbo;

	INIT_LIST_HEAD(&head);
	list_add(&tv.head, &head);

	r = ttm_eu_reserve_buffers(&ticket, &head, true, NULL);
	if (r)
		return r;

	if (!bo->adev->uvd.address_64_bit) {
		amdgpu_ttm_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_VRAM);
		amdgpu_uvd_force_into_uvd_segment(bo);
	}

	r = ttm_bo_validate(&bo->tbo, &bo->placement, true, false);
	if (r)
		goto err;

	r = amdgpu_job_alloc_with_ib(adev, 64, &job);
	if (r)
		goto err;

	ib = &job->ibs[0];
	addr = amdgpu_bo_gpu_offset(bo);
	ib->ptr[0] = PACKET0(mmUVD_GPCOM_VCPU_DATA0, 0);
	ib->ptr[1] = addr;
	ib->ptr[2] = PACKET0(mmUVD_GPCOM_VCPU_DATA1, 0);
	ib->ptr[3] = addr >> 32;
	ib->ptr[4] = PACKET0(mmUVD_GPCOM_VCPU_CMD, 0);
	ib->ptr[5] = 0;
	for (i = 6; i < 16; ++i)
		ib->ptr[i] = PACKET2(0);
	ib->length_dw = 16;

	if (direct) {
		r = amdgpu_ib_schedule(ring, 1, ib, NULL, &f);
		job->fence = f;
		if (r)
			goto err_free;

		amdgpu_job_free(job);
	} else {
		r = amdgpu_job_submit(job, ring, &adev->uvd.entity,
				      AMDGPU_FENCE_OWNER_UNDEFINED, &f);
		if (r)
			goto err_free;
	}

	ttm_eu_fence_buffer_objects(&ticket, &head, f);

	if (fence)
		*fence = fence_get(f);
	amdgpu_bo_unref(&bo);
	fence_put(f);

	return 0;

err_free:
	amdgpu_job_free(job);

err:
	ttm_eu_backoff_reservation(&ticket, &head);
	return r;
}

/* multiple fence commands without any stream commands in between can
   crash the vcpu so just try to emmit a dummy create/destroy msg to
   avoid this */
int amdgpu_uvd_get_create_msg(struct amdgpu_ring *ring, uint32_t handle,
			      struct fence **fence)
{
	struct amdgpu_device *adev = ring->adev;
	struct amdgpu_bo *bo;
	uint32_t *msg;
	int r, i;

	r = amdgpu_bo_create(adev, 1024, PAGE_SIZE, true,
			     AMDGPU_GEM_DOMAIN_VRAM,
			     AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED,
			     NULL, NULL, &bo);
	if (r)
		return r;

	r = amdgpu_bo_reserve(bo, false);
	if (r) {
		amdgpu_bo_unref(&bo);
		return r;
	}

	r = amdgpu_bo_kmap(bo, (void **)&msg);
	if (r) {
		amdgpu_bo_unreserve(bo);
		amdgpu_bo_unref(&bo);
		return r;
	}

	/* stitch together an UVD create msg */
	msg[0] = cpu_to_le32(0x00000de4);
	msg[1] = cpu_to_le32(0x00000000);
	msg[2] = cpu_to_le32(handle);
	msg[3] = cpu_to_le32(0x00000000);
	msg[4] = cpu_to_le32(0x00000000);
	msg[5] = cpu_to_le32(0x00000000);
	msg[6] = cpu_to_le32(0x00000000);
	msg[7] = cpu_to_le32(0x00000780);
	msg[8] = cpu_to_le32(0x00000440);
	msg[9] = cpu_to_le32(0x00000000);
	msg[10] = cpu_to_le32(0x01b37000);
	for (i = 11; i < 1024; ++i)
		msg[i] = cpu_to_le32(0x0);

	amdgpu_bo_kunmap(bo);
	amdgpu_bo_unreserve(bo);

	return amdgpu_uvd_send_msg(ring, bo, true, fence);
}

int amdgpu_uvd_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle,
			       bool direct, struct fence **fence)
{
	struct amdgpu_device *adev = ring->adev;
	struct amdgpu_bo *bo;
	uint32_t *msg;
	int r, i;

	r = amdgpu_bo_create(adev, 1024, PAGE_SIZE, true,
			     AMDGPU_GEM_DOMAIN_VRAM,
			     AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED,
			     NULL, NULL, &bo);
	if (r)
		return r;

	r = amdgpu_bo_reserve(bo, false);
	if (r) {
		amdgpu_bo_unref(&bo);
		return r;
	}

	r = amdgpu_bo_kmap(bo, (void **)&msg);
	if (r) {
		amdgpu_bo_unreserve(bo);
		amdgpu_bo_unref(&bo);
		return r;
	}

	/* stitch together an UVD destroy msg */
	msg[0] = cpu_to_le32(0x00000de4);
	msg[1] = cpu_to_le32(0x00000002);
	msg[2] = cpu_to_le32(handle);
	msg[3] = cpu_to_le32(0x00000000);
	for (i = 4; i < 1024; ++i)
		msg[i] = cpu_to_le32(0x0);

	amdgpu_bo_kunmap(bo);
	amdgpu_bo_unreserve(bo);

	return amdgpu_uvd_send_msg(ring, bo, direct, fence);
}

static void amdgpu_uvd_idle_work_handler(struct work_struct *work)
{
	struct amdgpu_device *adev =
		container_of(work, struct amdgpu_device, uvd.idle_work.work);
	unsigned i, fences, handles = 0;

	fences = amdgpu_fence_count_emitted(&adev->uvd.ring);

	for (i = 0; i < AMDGPU_MAX_UVD_HANDLES; ++i)
		if (atomic_read(&adev->uvd.handles[i]))
			++handles;

	if (fences == 0 && handles == 0) {
		if (adev->pm.dpm_enabled) {
			amdgpu_dpm_enable_uvd(adev, false);
		} else {
			amdgpu_asic_set_uvd_clocks(adev, 0, 0);
		}
	} else {
		schedule_delayed_work(&adev->uvd.idle_work,
				      msecs_to_jiffies(UVD_IDLE_TIMEOUT_MS));
	}
}

static void amdgpu_uvd_note_usage(struct amdgpu_device *adev)
{
	bool set_clocks = !cancel_delayed_work_sync(&adev->uvd.idle_work);
	set_clocks &= schedule_delayed_work(&adev->uvd.idle_work,
					    msecs_to_jiffies(UVD_IDLE_TIMEOUT_MS));

	if (set_clocks) {
		if (adev->pm.dpm_enabled) {
			amdgpu_dpm_enable_uvd(adev, true);
		} else {
			amdgpu_asic_set_uvd_clocks(adev, 53300, 40000);
		}
	}
}