#define _POSIX_C_SOURCE 200809L
#include <assert.h>
#include <fcntl.h>
#include <math.h>
#include <poll.h>
#include <stdlib.h>
#include <stdint.h>
#include <sys/types.h>
#include <unistd.h>
#include <drm_fourcc.h>
#include <vulkan/vulkan.h>
#include <wlr/render/interface.h>
#include <wlr/types/wlr_drm.h>
#include <wlr/types/wlr_matrix.h>
#include <wlr/util/box.h>
#include <wlr/util/log.h>
#include <wlr/render/vulkan.h>
#include <wlr/backend/interface.h>
#include <wlr/types/wlr_linux_dmabuf_v1.h>
#include <xf86drm.h>

#include "render/dmabuf.h"
#include "render/pixel_format.h"
#include "render/vulkan.h"
#include "render/vulkan/shaders/common.vert.h"
#include "render/vulkan/shaders/texture.frag.h"
#include "render/vulkan/shaders/quad.frag.h"
#include "render/vulkan/shaders/output.frag.h"
#include "types/wlr_buffer.h"
#include "types/wlr_matrix.h"

// TODO:
// - simplify stage allocation, don't track allocations but use ringbuffer-like
// - use a pipeline cache (not sure when to save though, after every pipeline
//   creation?)
// - create pipelines as derivatives of each other
// - evaluate if creating VkDeviceMemory pools is a good idea.
//   We can expect wayland client images to be fairly large (and shouldn't
//   have more than 4k of those I guess) but pooling memory allocations
//   might still be a good idea.

static const VkDeviceSize min_stage_size = 1024 * 1024; // 1MB
static const VkDeviceSize max_stage_size = 256 * min_stage_size; // 256MB
static const size_t start_descriptor_pool_size = 256u;
static bool default_debug = true;

static const struct wlr_renderer_impl renderer_impl;

bool wlr_renderer_is_vk(struct wlr_renderer *wlr_renderer) {
	return wlr_renderer->impl == &renderer_impl;
}

struct wlr_vk_renderer *vulkan_get_renderer(struct wlr_renderer *wlr_renderer) {
	assert(wlr_renderer_is_vk(wlr_renderer));
	struct wlr_vk_renderer *renderer = wl_container_of(wlr_renderer, renderer, wlr_renderer);
	return renderer;
}

static struct wlr_vk_render_format_setup *find_or_create_render_setup(
		struct wlr_vk_renderer *renderer, const struct wlr_vk_format *format,
		bool has_blending_buffer);

// https://www.w3.org/Graphics/Color/srgb
static float color_to_linear(float non_linear) {
	return (non_linear > 0.04045) ?
		pow((non_linear + 0.055) / 1.055, 2.4) :
		non_linear / 12.92;
}

static void mat3_to_mat4(const float mat3[9], float mat4[4][4]) {
	memset(mat4, 0, sizeof(float) * 16);
	mat4[0][0] = mat3[0];
	mat4[0][1] = mat3[1];
	mat4[0][3] = mat3[2];

	mat4[1][0] = mat3[3];
	mat4[1][1] = mat3[4];
	mat4[1][3] = mat3[5];

	mat4[2][2] = 1.f;
	mat4[3][3] = 1.f;
}

static struct wlr_vk_descriptor_pool *alloc_ds(
		struct wlr_vk_renderer *renderer, VkDescriptorSet *ds,
		VkDescriptorType type, const VkDescriptorSetLayout *layout,
		struct wl_list *pool_list, size_t *last_pool_size) {
	VkResult res;

	bool found = false;
	struct wlr_vk_descriptor_pool *pool;
	wl_list_for_each(pool, pool_list, link) {
		if (pool->free > 0) {
			found = true;
			break;
		}
	}

	if (!found) { // create new pool
		pool = calloc(1, sizeof(*pool));
		if (!pool) {
			wlr_log_errno(WLR_ERROR, "allocation failed");
			return NULL;
		}

		size_t count = 2 * (*last_pool_size);
		if (!count) {
			count = start_descriptor_pool_size;
		}

		pool->free = count;
		VkDescriptorPoolSize pool_size = {
			.descriptorCount = count,
			.type = type,
		};

		VkDescriptorPoolCreateInfo dpool_info = {
			.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_POOL_CREATE_INFO,
			.maxSets = count,
			.poolSizeCount = 1,
			.pPoolSizes = &pool_size,
			.flags = VK_DESCRIPTOR_POOL_CREATE_FREE_DESCRIPTOR_SET_BIT,
		};

		res = vkCreateDescriptorPool(renderer->dev->dev, &dpool_info, NULL,
			&pool->pool);
		if (res != VK_SUCCESS) {
			wlr_vk_error("vkCreateDescriptorPool", res);
			free(pool);
			return NULL;
		}

		*last_pool_size = count;
		wl_list_insert(pool_list, &pool->link);
	}

	VkDescriptorSetAllocateInfo ds_info = {
		.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_ALLOCATE_INFO,
		.descriptorSetCount = 1,
		.pSetLayouts = layout,
		.descriptorPool = pool->pool,
	};
	res = vkAllocateDescriptorSets(renderer->dev->dev, &ds_info, ds);
	if (res != VK_SUCCESS) {
		wlr_vk_error("vkAllocateDescriptorSets", res);
		return NULL;
	}

	--pool->free;
	return pool;
}

struct wlr_vk_descriptor_pool *vulkan_alloc_texture_ds(
		struct wlr_vk_renderer *renderer, VkDescriptorSetLayout ds_layout,
		VkDescriptorSet *ds) {
	return alloc_ds(renderer, ds, VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER,
		&ds_layout, &renderer->descriptor_pools,
		&renderer->last_pool_size);
}

struct wlr_vk_descriptor_pool *vulkan_alloc_blend_ds(
	struct wlr_vk_renderer *renderer, VkDescriptorSet *ds) {
	return alloc_ds(renderer, ds, VK_DESCRIPTOR_TYPE_INPUT_ATTACHMENT,
		&renderer->output_ds_layout, &renderer->output_descriptor_pools,
		&renderer->last_output_pool_size);
}

void vulkan_free_ds(struct wlr_vk_renderer *renderer,
		struct wlr_vk_descriptor_pool *pool, VkDescriptorSet ds) {
	vkFreeDescriptorSets(renderer->dev->dev, pool->pool, 1, &ds);
	++pool->free;
}

static void destroy_render_format_setup(struct wlr_vk_renderer *renderer,
		struct wlr_vk_render_format_setup *setup) {
	if (!setup) {
		return;
	}

	VkDevice dev = renderer->dev->dev;
	vkDestroyRenderPass(dev, setup->render_pass, NULL);
	vkDestroyPipeline(dev, setup->output_pipe, NULL);

	struct wlr_vk_pipeline *pipeline, *tmp_pipeline;
	wl_list_for_each_safe(pipeline, tmp_pipeline, &setup->pipelines, link) {
		vkDestroyPipeline(dev, pipeline->vk, NULL);
		free(pipeline);
	}
}

static void shared_buffer_destroy(struct wlr_vk_renderer *r,
		struct wlr_vk_shared_buffer *buffer) {
	if (!buffer) {
		return;
	}

	if (buffer->allocs.size > 0) {
		wlr_log(WLR_ERROR, "shared_buffer_finish: %zu allocations left",
			buffer->allocs.size / sizeof(struct wlr_vk_allocation));
	}

	wl_array_release(&buffer->allocs);
	if (buffer->buffer) {
		vkDestroyBuffer(r->dev->dev, buffer->buffer, NULL);
	}
	if (buffer->memory) {
		vkFreeMemory(r->dev->dev, buffer->memory, NULL);
	}

	wl_list_remove(&buffer->link);
	free(buffer);
}

struct wlr_vk_buffer_span vulkan_get_stage_span(struct wlr_vk_renderer *r,
		VkDeviceSize size, VkDeviceSize alignment) {
	// try to find free span
	// simple greedy allocation algorithm - should be enough for this usecase
	// since all allocations are freed together after the frame
	struct wlr_vk_shared_buffer *buf;
	wl_list_for_each_reverse(buf, &r->stage.buffers, link) {
		VkDeviceSize start = 0u;
		if (buf->allocs.size > 0) {
			const struct wlr_vk_allocation *allocs = buf->allocs.data;
			size_t allocs_len = buf->allocs.size / sizeof(struct wlr_vk_allocation);
			const struct wlr_vk_allocation *last = &allocs[allocs_len - 1];
			start = last->start + last->size;
		}

		assert(start <= buf->buf_size);

		// ensure the proposed start is a multiple of alignment
		start += alignment - 1 - ((start + alignment - 1) % alignment);

		if (buf->buf_size - start < size) {
			continue;
		}

		struct wlr_vk_allocation *a = wl_array_add(&buf->allocs, sizeof(*a));
		if (a == NULL) {
			wlr_log_errno(WLR_ERROR, "Allocation failed");
			goto error_alloc;
		}

		*a = (struct wlr_vk_allocation){
			.start = start,
			.size = size,
		};
		return (struct wlr_vk_buffer_span) {
			.buffer = buf,
			.alloc = *a,
		};
	}

	if (size > max_stage_size) {
		wlr_log(WLR_ERROR, "cannot vulkan stage buffer: "
			"requested size (%zu bytes) exceeds maximum (%zu bytes)",
			(size_t)size, (size_t)max_stage_size);
		goto error_alloc;
	}

	// we didn't find a free buffer - create one
	// size = clamp(max(size * 2, prev_size * 2), min_size, max_size)
	VkDeviceSize bsize = size * 2;
	bsize = bsize < min_stage_size ? min_stage_size : bsize;
	if (!wl_list_empty(&r->stage.buffers)) {
		struct wl_list *last_link = r->stage.buffers.prev;
		struct wlr_vk_shared_buffer *prev = wl_container_of(
			last_link, prev, link);
		VkDeviceSize last_size = 2 * prev->buf_size;
		bsize = bsize < last_size ? last_size : bsize;
	}

	if (bsize > max_stage_size) {
		wlr_log(WLR_INFO, "vulkan stage buffers have reached max size");
		bsize = max_stage_size;
	}

	// create buffer
	buf = calloc(1, sizeof(*buf));
	if (!buf) {
		wlr_log_errno(WLR_ERROR, "Allocation failed");
		goto error_alloc;
	}

	VkResult res;
	VkBufferCreateInfo buf_info = {
		.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
		.size = bsize,
		.usage = VK_BUFFER_USAGE_TRANSFER_DST_BIT |
			VK_BUFFER_USAGE_TRANSFER_SRC_BIT,
		.sharingMode = VK_SHARING_MODE_EXCLUSIVE,
	};
	res = vkCreateBuffer(r->dev->dev, &buf_info, NULL, &buf->buffer);
	if (res != VK_SUCCESS) {
		wlr_vk_error("vkCreateBuffer", res);
		goto error;
	}

	VkMemoryRequirements mem_reqs;
	vkGetBufferMemoryRequirements(r->dev->dev, buf->buffer, &mem_reqs);

	int mem_type_index = vulkan_find_mem_type(r->dev,
		VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT |
		VK_MEMORY_PROPERTY_HOST_COHERENT_BIT, mem_reqs.memoryTypeBits);
	if (mem_type_index < 0) {
		wlr_log(WLR_ERROR, "Failed to find memory type");
		goto error;
	}

	VkMemoryAllocateInfo mem_info = {
		.sType = VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO,
		.allocationSize = mem_reqs.size,
		.memoryTypeIndex = (uint32_t)mem_type_index,
	};
	res = vkAllocateMemory(r->dev->dev, &mem_info, NULL, &buf->memory);
	if (res != VK_SUCCESS) {
		wlr_vk_error("vkAllocatorMemory", res);
		goto error;
	}

	res = vkBindBufferMemory(r->dev->dev, buf->buffer, buf->memory, 0);
	if (res != VK_SUCCESS) {
		wlr_vk_error("vkBindBufferMemory", res);
		goto error;
	}

	struct wlr_vk_allocation *a = wl_array_add(&buf->allocs, sizeof(*a));
	if (a == NULL) {
		wlr_log_errno(WLR_ERROR, "Allocation failed");
		goto error;
	}

	wlr_log(WLR_DEBUG, "Created new vk staging buffer of size %" PRIu64, bsize);
	buf->buf_size = bsize;
	wl_list_insert(&r->stage.buffers, &buf->link);

	*a = (struct wlr_vk_allocation){
		.start = 0,
		.size = size,
	};
	return (struct wlr_vk_buffer_span) {
		.buffer = buf,
		.alloc = *a,
	};

error:
	shared_buffer_destroy(r, buf);

error_alloc:
	return (struct wlr_vk_buffer_span) {
		.buffer = NULL,
		.alloc = (struct wlr_vk_allocation) {0, 0},
	};
}

VkCommandBuffer vulkan_record_stage_cb(struct wlr_vk_renderer *renderer) {
	if (renderer->stage.cb == NULL) {
		renderer->stage.cb = vulkan_acquire_command_buffer(renderer);
		if (renderer->stage.cb == NULL) {
			return VK_NULL_HANDLE;
		}

		VkCommandBufferBeginInfo begin_info = {
			.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO,
		};
		vkBeginCommandBuffer(renderer->stage.cb->vk, &begin_info);
	}

	return renderer->stage.cb->vk;
}

bool vulkan_submit_stage_wait(struct wlr_vk_renderer *renderer) {
	if (renderer->stage.cb == NULL) {
		return false;
	}

	struct wlr_vk_command_buffer *cb = renderer->stage.cb;
	renderer->stage.cb = NULL;

	uint64_t timeline_point = vulkan_end_command_buffer(cb, renderer);
	if (timeline_point == 0) {
		return false;
	}

	VkTimelineSemaphoreSubmitInfoKHR timeline_submit_info = {
		.sType = VK_STRUCTURE_TYPE_TIMELINE_SEMAPHORE_SUBMIT_INFO_KHR,
		.signalSemaphoreValueCount = 1,
		.pSignalSemaphoreValues = &timeline_point,
	};
	VkSubmitInfo submit_info = {
		.sType = VK_STRUCTURE_TYPE_SUBMIT_INFO,
		.pNext = &timeline_submit_info,
		.commandBufferCount = 1,
		.pCommandBuffers = &cb->vk,
		.signalSemaphoreCount = 1,
		.pSignalSemaphores = &renderer->timeline_semaphore,
	};
	VkResult res = vkQueueSubmit(renderer->dev->queue, 1, &submit_info, VK_NULL_HANDLE);
	if (res != VK_SUCCESS) {
		wlr_vk_error("vkQueueSubmit", res);
		return false;
	}

	// NOTE: don't release stage allocations here since they may still be
	// used for reading. Will be done next frame.

	return vulkan_wait_command_buffer(cb, renderer);
}

struct wlr_vk_format_props *vulkan_format_props_from_drm(
		struct wlr_vk_device *dev, uint32_t drm_fmt) {
	for (size_t i = 0u; i < dev->format_prop_count; ++i) {
		if (dev->format_props[i].format.drm == drm_fmt) {
			return &dev->format_props[i];
		}
	}
	return NULL;
}

static bool init_command_buffer(struct wlr_vk_command_buffer *cb,
		struct wlr_vk_renderer *renderer) {
	VkResult res;

	VkCommandBuffer vk_cb = VK_NULL_HANDLE;
	VkCommandBufferAllocateInfo cmd_buf_info = {
		.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_ALLOCATE_INFO,
		.commandPool = renderer->command_pool,
		.level = VK_COMMAND_BUFFER_LEVEL_PRIMARY,
		.commandBufferCount = 1,
	};
	res = vkAllocateCommandBuffers(renderer->dev->dev, &cmd_buf_info, &vk_cb);
	if (res != VK_SUCCESS) {
		wlr_vk_error("vkAllocateCommandBuffers", res);
		return false;
	}

	*cb = (struct wlr_vk_command_buffer){
		.vk = vk_cb,
	};
	wl_list_init(&cb->destroy_textures);
	wl_list_init(&cb->stage_buffers);
	return true;
}

bool vulkan_wait_command_buffer(struct wlr_vk_command_buffer *cb,
		struct wlr_vk_renderer *renderer) {
	VkResult res;

	assert(cb->vk != VK_NULL_HANDLE && !cb->recording);

	VkSemaphoreWaitInfoKHR wait_info = {
		.sType = VK_STRUCTURE_TYPE_SEMAPHORE_WAIT_INFO_KHR,
		.semaphoreCount = 1,
		.pSemaphores = &renderer->timeline_semaphore,
		.pValues = &cb->timeline_point,
	};
	res = renderer->dev->api.vkWaitSemaphoresKHR(renderer->dev->dev, &wait_info, UINT64_MAX);
	if (res != VK_SUCCESS) {
		wlr_vk_error("vkWaitSemaphoresKHR", res);
		return false;
	}

	return true;
}

static void release_command_buffer_resources(struct wlr_vk_command_buffer *cb,
		struct wlr_vk_renderer *renderer) {
	struct wlr_vk_texture *texture, *texture_tmp;
	wl_list_for_each_safe(texture, texture_tmp, &cb->destroy_textures, destroy_link) {
		wl_list_remove(&texture->destroy_link);
		texture->last_used_cb = NULL;
		wlr_texture_destroy(&texture->wlr_texture);
	}

	struct wlr_vk_shared_buffer *buf, *buf_tmp;
	wl_list_for_each_safe(buf, buf_tmp, &cb->stage_buffers, link) {
		buf->allocs.size = 0;

		wl_list_remove(&buf->link);
		wl_list_insert(&renderer->stage.buffers, &buf->link);
	}
}

static struct wlr_vk_command_buffer *get_command_buffer(
		struct wlr_vk_renderer *renderer) {
	VkResult res;

	uint64_t current_point;
	res = renderer->dev->api.vkGetSemaphoreCounterValueKHR(renderer->dev->dev,
		renderer->timeline_semaphore, &current_point);
	if (res != VK_SUCCESS) {
		wlr_vk_error("vkGetSemaphoreCounterValueKHR", res);
		return NULL;
	}

	// Destroy textures for completed command buffers
	for (size_t i = 0; i < VULKAN_COMMAND_BUFFERS_CAP; i++) {
		struct wlr_vk_command_buffer *cb = &renderer->command_buffers[i];
		if (cb->vk != VK_NULL_HANDLE && !cb->recording &&
				cb->timeline_point <= current_point) {
			release_command_buffer_resources(cb, renderer);
		}
	}

	// First try to find an existing command buffer which isn't busy
	struct wlr_vk_command_buffer *unused = NULL;
	struct wlr_vk_command_buffer *wait = NULL;
	for (size_t i = 0; i < VULKAN_COMMAND_BUFFERS_CAP; i++) {
		struct wlr_vk_command_buffer *cb = &renderer->command_buffers[i];
		if (cb->vk == VK_NULL_HANDLE) {
			unused = cb;
			break;
		}
		if (cb->recording) {
			continue;
		}

		if (cb->timeline_point <= current_point) {
			return cb;
		}
		if (wait == NULL || cb->timeline_point < wait->timeline_point) {
			wait = cb;
		}
	}

	// If there is an unused slot, initialize it
	if (unused != NULL) {
		if (!init_command_buffer(unused, renderer)) {
			return NULL;
		}
		return unused;
	}

	// Block until a busy command buffer becomes available
	if (!vulkan_wait_command_buffer(wait, renderer)) {
		return NULL;
	}
	return wait;
}

struct wlr_vk_command_buffer *vulkan_acquire_command_buffer(
		struct wlr_vk_renderer *renderer) {
	struct wlr_vk_command_buffer *cb = get_command_buffer(renderer);
	if (cb == NULL) {
		return NULL;
	}

	assert(!cb->recording);
	cb->recording = true;

	return cb;
}

uint64_t vulkan_end_command_buffer(struct wlr_vk_command_buffer *cb,
		struct wlr_vk_renderer *renderer) {
	assert(cb->recording);
	cb->recording = false;

	VkResult res = vkEndCommandBuffer(cb->vk);
	if (res != VK_SUCCESS) {
		wlr_vk_error("vkEndCommandBuffer", res);
		return 0;
	}

	renderer->timeline_point++;
	cb->timeline_point = renderer->timeline_point;
	return cb->timeline_point;
}

void vulkan_reset_command_buffer(struct wlr_vk_command_buffer *cb) {
	if (cb == NULL) {
		return;
	}

	cb->recording = false;

	VkResult res = vkResetCommandBuffer(cb->vk, 0);
	if (res != VK_SUCCESS) {
		wlr_vk_error("vkResetCommandBuffer", res);
	}
}

static void destroy_render_buffer(struct wlr_vk_render_buffer *buffer) {
	wl_list_remove(&buffer->link);
	wlr_addon_finish(&buffer->addon);

	assert(buffer->renderer->current_render_buffer != buffer);

	VkDevice dev = buffer->renderer->dev->dev;

	// TODO: asynchronously wait for the command buffers using this render
	// buffer to complete (just like we do for textures)
	VkResult res = vkQueueWaitIdle(buffer->renderer->dev->queue);
	if (res != VK_SUCCESS) {
		wlr_vk_error("vkQueueWaitIdle", res);
	}

	vkDestroyFramebuffer(dev, buffer->framebuffer, NULL);
	vkDestroyImageView(dev, buffer->image_view, NULL);
	vkDestroyImage(dev, buffer->image, NULL);

	for (size_t i = 0u; i < buffer->mem_count; ++i) {
		vkFreeMemory(dev, buffer->memories[i], NULL);
	}

	vkDestroyImage(dev, buffer->blend_image, NULL);
	vkFreeMemory(dev, buffer->blend_memory, NULL);
	vkDestroyImageView(dev, buffer->blend_image_view, NULL);
	if (buffer->blend_attachment_pool) {
		vulkan_free_ds(buffer->renderer, buffer->blend_attachment_pool,
			buffer->blend_descriptor_set);
	}

	free(buffer);
}

static void handle_render_buffer_destroy(struct wlr_addon *addon) {
	struct wlr_vk_render_buffer *buffer = wl_container_of(addon, buffer, addon);
	destroy_render_buffer(buffer);
}

static struct wlr_addon_interface render_buffer_addon_impl = {
	.name = "wlr_vk_render_buffer",
	.destroy = handle_render_buffer_destroy,
};

static bool setup_blend_image(struct wlr_vk_renderer *renderer,
		struct wlr_vk_render_buffer *buffer, int32_t width, int32_t height) {
	VkResult res;
	VkDevice dev = renderer->dev->dev;

	// Set up an extra 16F buffer on which to do linear blending,
	// and afterwards to render onto the target
	VkImageCreateInfo img_info = {
		.sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO,
		.imageType = VK_IMAGE_TYPE_2D,
		.format = VK_FORMAT_R16G16B16A16_SFLOAT,
		.mipLevels = 1,
		.arrayLayers = 1,
		.samples = VK_SAMPLE_COUNT_1_BIT,
		.sharingMode = VK_SHARING_MODE_EXCLUSIVE,
		.tiling = VK_IMAGE_TILING_OPTIMAL,
		.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED,
		.extent = (VkExtent3D) { width, height, 1 },
		.usage = VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT | VK_IMAGE_USAGE_INPUT_ATTACHMENT_BIT,
	};

	res = vkCreateImage(dev, &img_info, NULL, &buffer->blend_image);
	if (res != VK_SUCCESS) {
		wlr_vk_error("vkCreateImage failed", res);
		goto error;
	}

	VkMemoryRequirements mem_reqs;
	vkGetImageMemoryRequirements(dev, buffer->blend_image, &mem_reqs);

	int mem_type_index = vulkan_find_mem_type(renderer->dev,
		VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT, mem_reqs.memoryTypeBits);
	if (mem_type_index == -1) {
		wlr_log(WLR_ERROR, "failed to find suitable vulkan memory type");
		goto error;
	}

	VkMemoryAllocateInfo mem_info = {
		.sType = VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO,
		.allocationSize = mem_reqs.size,
		.memoryTypeIndex = mem_type_index,
	};

	res = vkAllocateMemory(dev, &mem_info, NULL, &buffer->blend_memory);
	if (res != VK_SUCCESS) {
		wlr_vk_error("vkAllocatorMemory failed", res);
		goto error;
	}

	res = vkBindImageMemory(dev, buffer->blend_image, buffer->blend_memory, 0);
	if (res != VK_SUCCESS) {
		wlr_vk_error("vkBindMemory failed", res);
		goto error;
	}

	VkImageViewCreateInfo blend_view_info = {
		.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO,
		.image = buffer->blend_image,
		.viewType = VK_IMAGE_VIEW_TYPE_2D,
		.format = img_info.format,
		.components.r = VK_COMPONENT_SWIZZLE_IDENTITY,
		.components.g = VK_COMPONENT_SWIZZLE_IDENTITY,
		.components.b = VK_COMPONENT_SWIZZLE_IDENTITY,
		.components.a = VK_COMPONENT_SWIZZLE_IDENTITY,
		.subresourceRange = (VkImageSubresourceRange) {
			.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
			.baseMipLevel = 0,
			.levelCount = 1,
			.baseArrayLayer = 0,
			.layerCount = 1,
		},
	};

	res = vkCreateImageView(dev, &blend_view_info, NULL, &buffer->blend_image_view);
	if (res != VK_SUCCESS) {
		wlr_vk_error("vkCreateImageView failed", res);
		goto error;
	}

	buffer->blend_attachment_pool = vulkan_alloc_blend_ds(renderer,
		&buffer->blend_descriptor_set);
	if (!buffer->blend_attachment_pool) {
		wlr_log(WLR_ERROR, "failed to allocate descriptor");
		goto error;
	}

	VkDescriptorImageInfo ds_attach_info = {
		.imageLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL,
		.imageView = buffer->blend_image_view,
		.sampler = VK_NULL_HANDLE,
	};
	VkWriteDescriptorSet ds_write = {
		.sType = VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET,
		.descriptorCount = 1,
		.descriptorType = VK_DESCRIPTOR_TYPE_INPUT_ATTACHMENT,
		.dstSet = buffer->blend_descriptor_set,
		.dstBinding = 0,
		.pImageInfo = &ds_attach_info,
	};
	vkUpdateDescriptorSets(dev, 1, &ds_write, 0, NULL);
	return true;

error:
	// cleaning up blend_attachment_pool, blend_descriptor_set, blend_image,
	// blend_memory, and blend_image_view is the caller's responsibility,
	// since it will need to do this anyway if framebuffer setup fails
	return false;
}

static struct wlr_vk_render_buffer *create_render_buffer(
		struct wlr_vk_renderer *renderer, struct wlr_buffer *wlr_buffer) {
	VkResult res;
	VkDevice dev = renderer->dev->dev;

	struct wlr_vk_render_buffer *buffer = calloc(1, sizeof(*buffer));
	if (buffer == NULL) {
		wlr_log_errno(WLR_ERROR, "Allocation failed");
		return NULL;
	}
	buffer->wlr_buffer = wlr_buffer;
	buffer->renderer = renderer;

	struct wlr_dmabuf_attributes dmabuf = {0};
	if (!wlr_buffer_get_dmabuf(wlr_buffer, &dmabuf)) {
		goto error;
	}

	wlr_log(WLR_DEBUG, "vulkan create_render_buffer: %.4s, %dx%d",
		(const char*) &dmabuf.format, dmabuf.width, dmabuf.height);

	buffer->image = vulkan_import_dmabuf(renderer, &dmabuf,
		buffer->memories, &buffer->mem_count, true);
	if (!buffer->image) {
		goto error;
	}

	const struct wlr_vk_format_props *fmt = vulkan_format_props_from_drm(
		renderer->dev, dmabuf.format);
	if (fmt == NULL) {
		wlr_log(WLR_ERROR, "Unsupported pixel format %"PRIx32 " (%.4s)",
			dmabuf.format, (const char*) &dmabuf.format);
		goto error;
	}

	VkImageViewCreateInfo view_info = {
		.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO,
		.image = buffer->image,
		.viewType = VK_IMAGE_VIEW_TYPE_2D,
		.format = fmt->format.vk,
		.components.r = VK_COMPONENT_SWIZZLE_IDENTITY,
		.components.g = VK_COMPONENT_SWIZZLE_IDENTITY,
		.components.b = VK_COMPONENT_SWIZZLE_IDENTITY,
		.components.a = VK_COMPONENT_SWIZZLE_IDENTITY,
		.subresourceRange = (VkImageSubresourceRange) {
			.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
			.baseMipLevel = 0,
			.levelCount = 1,
			.baseArrayLayer = 0,
			.layerCount = 1,
		},
	};

	res = vkCreateImageView(dev, &view_info, NULL, &buffer->image_view);
	if (res != VK_SUCCESS) {
		wlr_vk_error("vkCreateImageView failed", res);
		goto error;
	}

	bool has_blending_buffer = !fmt->format.is_srgb;

	buffer->render_setup = find_or_create_render_setup(
		renderer, &fmt->format, has_blending_buffer);
	if (!buffer->render_setup) {
		goto error;
	}

	VkImageView attachments[2] = {0};
	uint32_t attachment_count = 0;

	if (has_blending_buffer) {
		if (!setup_blend_image(renderer, buffer, dmabuf.width, dmabuf.height)) {
			goto error;
		}
		attachments[attachment_count++] = buffer->blend_image_view;
	}
	attachments[attachment_count++] = buffer->image_view;

	VkFramebufferCreateInfo fb_info = {
		.sType = VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO,
		.attachmentCount = attachment_count,
		.pAttachments = attachments,
		.flags = 0u,
		.width = dmabuf.width,
		.height = dmabuf.height,
		.layers = 1u,
		.renderPass = buffer->render_setup->render_pass,
	};

	res = vkCreateFramebuffer(dev, &fb_info, NULL, &buffer->framebuffer);
	if (res != VK_SUCCESS) {
		wlr_vk_error("vkCreateFramebuffer", res);
		goto error;
	}


	wlr_addon_init(&buffer->addon, &wlr_buffer->addons, renderer,
		&render_buffer_addon_impl);
	wl_list_insert(&renderer->render_buffers, &buffer->link);

	return buffer;

error:
	if (buffer->blend_attachment_pool) {
		vulkan_free_ds(buffer->renderer, buffer->blend_attachment_pool,
			buffer->blend_descriptor_set);
	}
	vkDestroyImage(dev, buffer->blend_image, NULL);
	vkFreeMemory(dev, buffer->blend_memory, NULL);
	vkDestroyImageView(dev, buffer->blend_image_view, NULL);

	vkDestroyFramebuffer(dev, buffer->framebuffer, NULL);
	vkDestroyImageView(dev, buffer->image_view, NULL);
	vkDestroyImage(dev, buffer->image, NULL);
	for (size_t i = 0u; i < buffer->mem_count; ++i) {
		vkFreeMemory(dev, buffer->memories[i], NULL);
	}

	wlr_dmabuf_attributes_finish(&dmabuf);
	free(buffer);
	return NULL;
}

static struct wlr_vk_render_buffer *get_render_buffer(
		struct wlr_vk_renderer *renderer, struct wlr_buffer *wlr_buffer) {
	struct wlr_addon *addon =
		wlr_addon_find(&wlr_buffer->addons, renderer, &render_buffer_addon_impl);
	if (addon == NULL) {
		return NULL;
	}

	struct wlr_vk_render_buffer *buffer = wl_container_of(addon, buffer, addon);
	return buffer;
}

static bool vulkan_bind_buffer(struct wlr_renderer *wlr_renderer,
		struct wlr_buffer *wlr_buffer) {
	struct wlr_vk_renderer *renderer = vulkan_get_renderer(wlr_renderer);

	if (renderer->current_render_buffer) {
		wlr_buffer_unlock(renderer->current_render_buffer->wlr_buffer);
		renderer->current_render_buffer = NULL;
	}

	if (!wlr_buffer) {
		return true;
	}

	struct wlr_vk_render_buffer *buffer = get_render_buffer(renderer, wlr_buffer);
	if (!buffer) {
		buffer = create_render_buffer(renderer, wlr_buffer);
		if (!buffer) {
			return false;
		}
	}

	wlr_buffer_lock(wlr_buffer);
	renderer->current_render_buffer = buffer;
	return true;
}

static bool vulkan_begin(struct wlr_renderer *wlr_renderer,
		uint32_t width, uint32_t height) {
	struct wlr_vk_renderer *renderer = vulkan_get_renderer(wlr_renderer);
	assert(renderer->current_render_buffer);

	struct wlr_vk_command_buffer *cb = vulkan_acquire_command_buffer(renderer);
	if (cb == NULL) {
		return false;
	}

	assert(renderer->current_command_buffer == NULL);
	renderer->current_command_buffer = cb;

	VkCommandBufferBeginInfo begin_info = {
		.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO,
	};
	VkResult res = vkBeginCommandBuffer(cb->vk, &begin_info);
	if (res != VK_SUCCESS) {
		wlr_vk_error("vkBeginCommandBuffer", res);
		return false;
	}

	// begin render pass
	VkFramebuffer fb = renderer->current_render_buffer->framebuffer;

	VkRect2D rect = {{0, 0}, {width, height}};
	renderer->scissor = rect;

	VkRenderPassBeginInfo rp_info = {
		.sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO,
		.renderArea = rect,
		.renderPass = renderer->current_render_buffer->render_setup->render_pass,
		.framebuffer = fb,
		.clearValueCount = 0,
	};
	vkCmdBeginRenderPass(cb->vk, &rp_info, VK_SUBPASS_CONTENTS_INLINE);

	VkViewport vp = {0.f, 0.f, (float) width, (float) height, 0.f, 1.f};
	vkCmdSetViewport(cb->vk, 0, 1, &vp);
	vkCmdSetScissor(cb->vk, 0, 1, &rect);

	// Refresh projection matrix.
	// matrix_projection() assumes a GL coordinate system so we need
	// to pass WL_OUTPUT_TRANSFORM_FLIPPED_180 to adjust it for vulkan.
	matrix_projection(renderer->projection, width, height,
		WL_OUTPUT_TRANSFORM_FLIPPED_180);

	renderer->render_width = width;
	renderer->render_height = height;
	renderer->bound_pipe = VK_NULL_HANDLE;

	return true;
}

bool vulkan_sync_foreign_texture(struct wlr_vk_texture *texture) {
	struct wlr_vk_renderer *renderer = texture->renderer;
	VkResult res;

	struct wlr_dmabuf_attributes dmabuf = {0};
	if (!wlr_buffer_get_dmabuf(texture->buffer, &dmabuf)) {
		wlr_log(WLR_ERROR, "Failed to get texture DMA-BUF");
		return false;
	}

	if (!renderer->dev->implicit_sync_interop) {
		// We have no choice but to block here sadly

		for (int i = 0; i < dmabuf.n_planes; i++) {
			struct pollfd pollfd = {
				.fd = dmabuf.fd[i],
				.events = POLLIN,
			};
			int timeout_ms = 1000;
			int ret = poll(&pollfd, 1, timeout_ms);
			if (ret < 0) {
				wlr_log_errno(WLR_ERROR, "Failed to wait for DMA-BUF fence");
				return false;
			} else if (ret == 0) {
				wlr_log(WLR_ERROR, "Timed out while waiting for DMA-BUF fence");
				return false;
			}
		}

		return true;
	}

	for (int i = 0; i < dmabuf.n_planes; i++) {
		int sync_file_fd = dmabuf_export_sync_file(dmabuf.fd[i], DMA_BUF_SYNC_READ);
		if (sync_file_fd < 0) {
			wlr_log(WLR_ERROR, "Failed to extract DMA-BUF fence");
			return false;
		}

		if (texture->foreign_semaphores[i] == VK_NULL_HANDLE) {
			VkSemaphoreCreateInfo semaphore_info = {
				.sType = VK_STRUCTURE_TYPE_SEMAPHORE_CREATE_INFO,
			};
			res = vkCreateSemaphore(renderer->dev->dev, &semaphore_info, NULL,
				&texture->foreign_semaphores[i]);
			if (res != VK_SUCCESS) {
				close(sync_file_fd);
				wlr_vk_error("vkCreateSemaphore", res);
				return false;
			}
		}

		VkImportSemaphoreFdInfoKHR import_info = {
			.sType = VK_STRUCTURE_TYPE_IMPORT_SEMAPHORE_FD_INFO_KHR,
			.handleType = VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_SYNC_FD_BIT,
			.flags = VK_SEMAPHORE_IMPORT_TEMPORARY_BIT,
			.semaphore = texture->foreign_semaphores[i],
			.fd = sync_file_fd,
		};
		res = renderer->dev->api.vkImportSemaphoreFdKHR(renderer->dev->dev, &import_info);
		if (res != VK_SUCCESS) {
			close(sync_file_fd);
			wlr_vk_error("vkImportSemaphoreFdKHR", res);
			return false;
		}
	}

	return true;
}

bool vulkan_sync_render_buffer(struct wlr_vk_renderer *renderer,
		struct wlr_vk_render_buffer *render_buffer, struct wlr_vk_command_buffer *cb) {
	VkResult res;

	if (!renderer->dev->implicit_sync_interop) {
		// We have no choice but to block here sadly
		return vulkan_wait_command_buffer(cb, renderer);
	}

	struct wlr_dmabuf_attributes dmabuf = {0};
	if (!wlr_buffer_get_dmabuf(render_buffer->wlr_buffer, &dmabuf)) {
		wlr_log(WLR_ERROR, "wlr_buffer_get_dmabuf failed");
		return false;
	}

	// Note: vkGetSemaphoreFdKHR implicitly resets the semaphore
	const VkSemaphoreGetFdInfoKHR get_fence_fd_info = {
		.sType = VK_STRUCTURE_TYPE_SEMAPHORE_GET_FD_INFO_KHR,
		.semaphore = cb->binary_semaphore,
		.handleType = VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_SYNC_FD_BIT,
	};
	int sync_file_fd = -1;
	res = renderer->dev->api.vkGetSemaphoreFdKHR(renderer->dev->dev,
		&get_fence_fd_info, &sync_file_fd);
	if (res != VK_SUCCESS) {
		wlr_vk_error("vkGetSemaphoreFdKHR", res);
		return false;
	}

	for (int i = 0; i < dmabuf.n_planes; i++) {
		if (!dmabuf_import_sync_file(dmabuf.fd[i], DMA_BUF_SYNC_WRITE,
				sync_file_fd)) {
			close(sync_file_fd);
			return false;
		}
	}

	close(sync_file_fd);

	return true;
}

static void vulkan_end(struct wlr_renderer *wlr_renderer) {
	struct wlr_vk_renderer *renderer = vulkan_get_renderer(wlr_renderer);
	assert(renderer->current_render_buffer);

	struct wlr_vk_command_buffer *render_cb = renderer->current_command_buffer;
	assert(render_cb != NULL);
	renderer->current_command_buffer = NULL;

	if (vulkan_record_stage_cb(renderer) == VK_NULL_HANDLE) {
		return;
	}

	struct wlr_vk_command_buffer *stage_cb = renderer->stage.cb;
	assert(stage_cb != NULL);
	renderer->stage.cb = NULL;

	struct wlr_vk_render_buffer *current_rb = renderer->current_render_buffer;

	if (current_rb->blend_image) {
		// Apply output shader to map blend image to actual output image
		vkCmdNextSubpass(render_cb->vk, VK_SUBPASS_CONTENTS_INLINE);

		VkPipeline pipe = current_rb->render_setup->output_pipe;
		if (pipe != renderer->bound_pipe) {
			vkCmdBindPipeline(render_cb->vk, VK_PIPELINE_BIND_POINT_GRAPHICS, pipe);
			renderer->bound_pipe = pipe;
		}

		float final_matrix[9] = {
			renderer->render_width, 0.f, -1.f,
			0.f, renderer->render_height, -1.f,
			0.f, 0.f, 0.f,
		};
		struct wlr_vk_vert_pcr_data vert_pcr_data;
		mat3_to_mat4(final_matrix, vert_pcr_data.mat4);
		vert_pcr_data.uv_off[0] = 0.f;
		vert_pcr_data.uv_off[1] = 0.f;
		vert_pcr_data.uv_size[0] = 1.f;
		vert_pcr_data.uv_size[1] = 1.f;

		vkCmdPushConstants(render_cb->vk, renderer->output_pipe_layout,
			VK_SHADER_STAGE_VERTEX_BIT, 0, sizeof(vert_pcr_data), &vert_pcr_data);
		vkCmdBindDescriptorSets(render_cb->vk,
			VK_PIPELINE_BIND_POINT_GRAPHICS, renderer->output_pipe_layout,
			0, 1, &current_rb->blend_descriptor_set, 0, NULL);

		vkCmdDraw(render_cb->vk, 4, 1, 0, 0);
	}

	vkCmdEndRenderPass(render_cb->vk);

	renderer->render_width = 0u;
	renderer->render_height = 0u;
	renderer->bound_pipe = VK_NULL_HANDLE;

	// insert acquire and release barriers for dmabuf-images
	unsigned barrier_count = wl_list_length(&renderer->foreign_textures) + 1;
	VkImageMemoryBarrier *acquire_barriers = calloc(barrier_count, sizeof(*acquire_barriers));
	VkImageMemoryBarrier *release_barriers = calloc(barrier_count, sizeof(*release_barriers));
	VkSemaphoreSubmitInfoKHR *render_wait = calloc(barrier_count * WLR_DMABUF_MAX_PLANES, sizeof(*render_wait));
	if (acquire_barriers == NULL || release_barriers == NULL || render_wait == NULL) {
		wlr_log_errno(WLR_ERROR, "Allocation failed");
		free(acquire_barriers);
		free(release_barriers);
		free(render_wait);
		return;
	}

	struct wlr_vk_texture *texture, *tmp_tex;
	unsigned idx = 0;
	uint32_t render_wait_len = 0;
	wl_list_for_each_safe(texture, tmp_tex, &renderer->foreign_textures, foreign_link) {
		VkImageLayout src_layout = VK_IMAGE_LAYOUT_GENERAL;
		if (!texture->transitioned) {
			src_layout = VK_IMAGE_LAYOUT_UNDEFINED;
			texture->transitioned = true;
		}

		acquire_barriers[idx] = (VkImageMemoryBarrier){
			.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
			.srcQueueFamilyIndex = VK_QUEUE_FAMILY_FOREIGN_EXT,
			.dstQueueFamilyIndex = renderer->dev->queue_family,
			.image = texture->image,
			.oldLayout = src_layout,
			.newLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL,
			.srcAccessMask = 0, // ignored anyways
			.dstAccessMask = VK_ACCESS_SHADER_READ_BIT,
			.subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
			.subresourceRange.layerCount = 1,
			.subresourceRange.levelCount = 1,
		};

		release_barriers[idx] = (VkImageMemoryBarrier){
			.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
			.srcQueueFamilyIndex = renderer->dev->queue_family,
			.dstQueueFamilyIndex = VK_QUEUE_FAMILY_FOREIGN_EXT,
			.image = texture->image,
			.oldLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL,
			.newLayout = VK_IMAGE_LAYOUT_GENERAL,
			.srcAccessMask = VK_ACCESS_SHADER_READ_BIT,
			.dstAccessMask = 0, // ignored anyways
			.subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
			.subresourceRange.layerCount = 1,
			.subresourceRange.levelCount = 1,
		};

		++idx;

		if (!vulkan_sync_foreign_texture(texture)) {
			wlr_log(WLR_ERROR, "Failed to wait for foreign texture DMA-BUF fence");
		} else {
			for (size_t i = 0; i < WLR_DMABUF_MAX_PLANES; i++) {
				if (texture->foreign_semaphores[i] != VK_NULL_HANDLE) {
					assert(render_wait_len < barrier_count * WLR_DMABUF_MAX_PLANES);
					render_wait[render_wait_len++] = (VkSemaphoreSubmitInfoKHR){
						.sType = VK_STRUCTURE_TYPE_SEMAPHORE_SUBMIT_INFO_KHR,
						.semaphore = texture->foreign_semaphores[i],
						.stageMask = VK_PIPELINE_STAGE_2_ALL_COMMANDS_BIT_KHR,
					};
				}
			}
		}

		wl_list_remove(&texture->foreign_link);
		texture->owned = false;
	}

	// also add acquire/release barriers for the current render buffer
	VkImageLayout src_layout = VK_IMAGE_LAYOUT_GENERAL;
	if (!current_rb->transitioned) {
		src_layout = VK_IMAGE_LAYOUT_PREINITIALIZED;
		current_rb->transitioned = true;
	}

	// acquire render buffer before rendering
	acquire_barriers[idx] = (VkImageMemoryBarrier){
		.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
		.srcQueueFamilyIndex = VK_QUEUE_FAMILY_FOREIGN_EXT,
		.dstQueueFamilyIndex = renderer->dev->queue_family,
		.image = renderer->current_render_buffer->image,
		.oldLayout = src_layout,
		.newLayout = VK_IMAGE_LAYOUT_GENERAL,
		.srcAccessMask = 0, // ignored anyways
		.dstAccessMask = VK_ACCESS_COLOR_ATTACHMENT_READ_BIT |
			VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT,
		.subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
		.subresourceRange.layerCount = 1,
		.subresourceRange.levelCount = 1,
	};

	// release render buffer after rendering
	release_barriers[idx] = (VkImageMemoryBarrier){
		.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
		.srcQueueFamilyIndex = renderer->dev->queue_family,
		.dstQueueFamilyIndex = VK_QUEUE_FAMILY_FOREIGN_EXT,
		.image = renderer->current_render_buffer->image,
		.oldLayout = VK_IMAGE_LAYOUT_GENERAL,
		.newLayout = VK_IMAGE_LAYOUT_GENERAL,
		.srcAccessMask = VK_ACCESS_COLOR_ATTACHMENT_READ_BIT |
			VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT,
		.dstAccessMask = 0, // ignored anyways
		.subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
		.subresourceRange.layerCount = 1,
		.subresourceRange.levelCount = 1,
	};

	++idx;

	if (current_rb->blend_image) {
		// The render pass changes the blend image layout from
		// color attachment to read only, so on each frame, before
		// the render pass starts, we change it back
		VkImageLayout blend_src_layout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;
		if (!current_rb->blend_transitioned) {
			blend_src_layout = VK_IMAGE_LAYOUT_UNDEFINED;
			current_rb->blend_transitioned = true;
		}

		VkImageMemoryBarrier blend_acq_barrier = {
			.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
			.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
			.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
			.image = current_rb->blend_image,
			.oldLayout = blend_src_layout,
			.newLayout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL,
			.srcAccessMask = VK_ACCESS_SHADER_READ_BIT,
			.dstAccessMask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT,
			.subresourceRange = {
				.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
				.layerCount = 1,
				.levelCount = 1,
			}
		};
		vkCmdPipelineBarrier(stage_cb->vk, VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT,
			VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT,
			0, 0, NULL, 0, NULL, 1, &blend_acq_barrier);
	}

	vkCmdPipelineBarrier(stage_cb->vk, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT,
		VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT,
		0, 0, NULL, 0, NULL, barrier_count, acquire_barriers);

	vkCmdPipelineBarrier(render_cb->vk, VK_PIPELINE_STAGE_ALL_GRAPHICS_BIT,
		VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT, 0, 0, NULL, 0, NULL,
		barrier_count, release_barriers);

	free(acquire_barriers);
	free(release_barriers);

	// No semaphores needed here.
	// We don't need a semaphore from the stage/transfer submission
	// to the render submissions since they are on the same queue
	// and we have a renderpass dependency for that.
	uint64_t stage_timeline_point = vulkan_end_command_buffer(stage_cb, renderer);
	if (stage_timeline_point == 0) {
		return;
	}

	VkCommandBufferSubmitInfoKHR stage_cb_info = {
		.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_SUBMIT_INFO_KHR,
		.commandBuffer = stage_cb->vk,
	};
	VkSemaphoreSubmitInfoKHR stage_signal = {
		.sType = VK_STRUCTURE_TYPE_SEMAPHORE_SUBMIT_INFO_KHR,
		.semaphore = renderer->timeline_semaphore,
		.value = stage_timeline_point,
	};
	VkSubmitInfo2KHR stage_submit = {
		.sType = VK_STRUCTURE_TYPE_SUBMIT_INFO_2_KHR,
		.commandBufferInfoCount = 1,
		.pCommandBufferInfos = &stage_cb_info,
		.signalSemaphoreInfoCount = 1,
		.pSignalSemaphoreInfos = &stage_signal,
	};

	VkSemaphoreSubmitInfoKHR stage_wait;
	if (renderer->stage.last_timeline_point > 0) {
		stage_wait = (VkSemaphoreSubmitInfoKHR){
			.sType = VK_STRUCTURE_TYPE_SEMAPHORE_SUBMIT_INFO_KHR,
			.semaphore = renderer->timeline_semaphore,
			.value = renderer->stage.last_timeline_point,
			.stageMask = VK_PIPELINE_STAGE_2_ALL_COMMANDS_BIT_KHR,
		};

		stage_submit.waitSemaphoreInfoCount = 1;
		stage_submit.pWaitSemaphoreInfos = &stage_wait;
	}

	renderer->stage.last_timeline_point = stage_timeline_point;

	uint64_t render_timeline_point = vulkan_end_command_buffer(render_cb, renderer);
	if (render_timeline_point == 0) {
		return;
	}

	size_t render_signal_len = 1;
	VkSemaphoreSubmitInfoKHR render_signal[2] = {0};
	render_signal[0] = (VkSemaphoreSubmitInfoKHR){
		.sType = VK_STRUCTURE_TYPE_SEMAPHORE_SUBMIT_INFO_KHR,
		.semaphore = renderer->timeline_semaphore,
		.value = render_timeline_point,
	};
	if (renderer->dev->implicit_sync_interop) {
		if (render_cb->binary_semaphore == VK_NULL_HANDLE) {
			VkExportSemaphoreCreateInfo export_info = {
				.sType = VK_STRUCTURE_TYPE_EXPORT_SEMAPHORE_CREATE_INFO,
				.handleTypes = VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_SYNC_FD_BIT,
			};
			VkSemaphoreCreateInfo semaphore_info = {
				.sType = VK_STRUCTURE_TYPE_SEMAPHORE_CREATE_INFO,
				.pNext = &export_info,
			};
			VkResult res = vkCreateSemaphore(renderer->dev->dev, &semaphore_info,
				NULL, &render_cb->binary_semaphore);
			if (res != VK_SUCCESS) {
				wlr_vk_error("vkCreateSemaphore", res);
				return;
			}
		}

		render_signal[render_signal_len++] = (VkSemaphoreSubmitInfoKHR){
			.sType = VK_STRUCTURE_TYPE_SEMAPHORE_SUBMIT_INFO_KHR,
			.semaphore = render_cb->binary_semaphore,
		};
	}

	VkCommandBufferSubmitInfoKHR render_cb_info = {
		.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_SUBMIT_INFO_KHR,
		.commandBuffer = render_cb->vk,
	};
	VkSubmitInfo2KHR render_submit = {
		.sType = VK_STRUCTURE_TYPE_SUBMIT_INFO_2_KHR,
		.waitSemaphoreInfoCount = render_wait_len,
		.pWaitSemaphoreInfos = render_wait,
		.commandBufferInfoCount = 1,
		.pCommandBufferInfos = &render_cb_info,
		.signalSemaphoreInfoCount = render_signal_len,
		.pSignalSemaphoreInfos = render_signal,
	};

	VkSubmitInfo2KHR submit_infos[] = { stage_submit, render_submit };
	VkResult res = renderer->dev->api.vkQueueSubmit2KHR(renderer->dev->queue, 2, submit_infos, VK_NULL_HANDLE);
	if (res == VK_ERROR_DEVICE_LOST) {
		wlr_log(WLR_ERROR, "vkQueueSubmit failed with VK_ERROR_DEVICE_LOST");
		wl_signal_emit_mutable(&wlr_renderer->events.lost, NULL);
		return;
	} else if (res != VK_SUCCESS) {
		wlr_vk_error("vkQueueSubmit", res);
		return;
	}

	free(render_wait);

	struct wlr_vk_shared_buffer *stage_buf, *stage_buf_tmp;
	wl_list_for_each_safe(stage_buf, stage_buf_tmp, &renderer->stage.buffers, link) {
		if (stage_buf->allocs.size == 0) {
			continue;
		}
		wl_list_remove(&stage_buf->link);
		wl_list_insert(&stage_cb->stage_buffers, &stage_buf->link);
	}

	if (!vulkan_sync_render_buffer(renderer, renderer->current_render_buffer, render_cb)) {
		return;
	}
}

static bool vulkan_render_subtexture_with_matrix(struct wlr_renderer *wlr_renderer,
		struct wlr_texture *wlr_texture, const struct wlr_fbox *box,
		const float matrix[static 9], float alpha) {
	struct wlr_vk_renderer *renderer = vulkan_get_renderer(wlr_renderer);
	VkCommandBuffer cb = renderer->current_command_buffer->vk;

	struct wlr_vk_texture *texture = vulkan_get_texture(wlr_texture);
	assert(texture->renderer == renderer);
	if (texture->dmabuf_imported && !texture->owned) {
		// Store this texture in the list of textures that need to be
		// acquired before rendering and released after rendering.
		// We don't do it here immediately since barriers inside
		// a renderpass are suboptimal (would require additional renderpass
		// dependency and potentially multiple barriers) and it's
		// better to issue one barrier for all used textures anyways.
		texture->owned = true;
		assert(texture->foreign_link.prev == NULL);
		assert(texture->foreign_link.next == NULL);
		wl_list_insert(&renderer->foreign_textures, &texture->foreign_link);
	}

	struct wlr_vk_pipeline *pipe = setup_get_or_create_pipeline(
		renderer->current_render_buffer->render_setup,
		&(struct wlr_vk_pipeline_key) {
			.layout = {
				.ycbcr_format = texture->format->is_ycbcr ? texture->format : NULL,
			},
			.texture_transform = texture->transform,
		});
	if (!pipe) {
		return false;
	}

	struct wlr_vk_texture_view *view =
		vulkan_texture_get_or_create_view(texture, pipe->layout);
	if (!view) {
		return false;
	}

	if (pipe->vk != renderer->bound_pipe) {
		vkCmdBindPipeline(cb, VK_PIPELINE_BIND_POINT_GRAPHICS, pipe->vk);
		renderer->bound_pipe = pipe->vk;
	}

	vkCmdBindDescriptorSets(cb, VK_PIPELINE_BIND_POINT_GRAPHICS,
		pipe->layout->vk, 0, 1, &view->ds, 0, NULL);

	float final_matrix[9];
	wlr_matrix_multiply(final_matrix, renderer->projection, matrix);

	struct wlr_vk_vert_pcr_data vert_pcr_data;
	mat3_to_mat4(final_matrix, vert_pcr_data.mat4);

	vert_pcr_data.uv_off[0] = box->x / wlr_texture->width;
	vert_pcr_data.uv_off[1] = box->y / wlr_texture->height;
	vert_pcr_data.uv_size[0] = box->width / wlr_texture->width;
	vert_pcr_data.uv_size[1] = box->height / wlr_texture->height;

	vkCmdPushConstants(cb, pipe->layout->vk,
		VK_SHADER_STAGE_VERTEX_BIT, 0, sizeof(vert_pcr_data), &vert_pcr_data);
	vkCmdPushConstants(cb, pipe->layout->vk,
		VK_SHADER_STAGE_FRAGMENT_BIT, sizeof(vert_pcr_data), sizeof(float),
		&alpha);
	vkCmdDraw(cb, 4, 1, 0, 0);
	texture->last_used_cb = renderer->current_command_buffer;

	return true;
}

static void vulkan_clear(struct wlr_renderer *wlr_renderer,
		const float color[static 4]) {
	struct wlr_vk_renderer *renderer = vulkan_get_renderer(wlr_renderer);
	VkCommandBuffer cb = renderer->current_command_buffer->vk;

	if (renderer->scissor.extent.width == 0 || renderer->scissor.extent.height == 0) {
		return;
	}

	VkClearAttachment att = {
		.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
		.colorAttachment = 0u,
		// Input color values are given in srgb space, vulkan expects
		// them in linear space. We explicitly import argb8 render buffers
		// as srgb, vulkan will convert the input values we give here to
		// srgb first.
		// But in other parts of wlroots we just always assume
		// srgb so that's why we have to convert here.
		.clearValue.color.float32 = {
			color_to_linear(color[0]),
			color_to_linear(color[1]),
			color_to_linear(color[2]),
			color[3], // no conversion for alpha
		},
	};

	VkClearRect rect = {
		.rect = renderer->scissor,
		.layerCount = 1,
	};
	vkCmdClearAttachments(cb, 1, &att, 1, &rect);
}

static void vulkan_scissor(struct wlr_renderer *wlr_renderer,
		struct wlr_box *box) {
	struct wlr_vk_renderer *renderer = vulkan_get_renderer(wlr_renderer);
	VkCommandBuffer cb = renderer->current_command_buffer->vk;

	uint32_t w = renderer->render_width;
	uint32_t h = renderer->render_height;
	struct wlr_box dst = {0, 0, w, h};
	if (box && !wlr_box_intersection(&dst, box, &dst)) {
		dst = (struct wlr_box) {0, 0, 0, 0}; // empty
	}

	VkRect2D rect = (VkRect2D) {{dst.x, dst.y}, {dst.width, dst.height}};
	renderer->scissor = rect;
	vkCmdSetScissor(cb, 0, 1, &rect);
}

static const uint32_t *vulkan_get_shm_texture_formats(
		struct wlr_renderer *wlr_renderer, size_t *len) {
	struct wlr_vk_renderer *renderer = vulkan_get_renderer(wlr_renderer);
	*len = renderer->dev->shm_format_count;
	return renderer->dev->shm_formats;
}

static void vulkan_render_quad_with_matrix(struct wlr_renderer *wlr_renderer,
		const float color[static 4], const float matrix[static 9]) {
	struct wlr_vk_renderer *renderer = vulkan_get_renderer(wlr_renderer);
	VkCommandBuffer cb = renderer->current_command_buffer->vk;

	struct wlr_vk_pipeline *pipe = setup_get_or_create_pipeline(
		renderer->current_render_buffer->render_setup,
		&(struct wlr_vk_pipeline_key) {
			.source = WLR_VK_SHADER_SOURCE_SINGLE_COLOR,
			.layout = {
				.ycbcr_format = NULL,
			},
		});
	if (!pipe) {
		return;
	}

	if (pipe->vk != renderer->bound_pipe) {
		vkCmdBindPipeline(cb, VK_PIPELINE_BIND_POINT_GRAPHICS, pipe->vk);
		renderer->bound_pipe = pipe->vk;
	}

	float final_matrix[9];
	wlr_matrix_multiply(final_matrix, renderer->projection, matrix);

	struct wlr_vk_vert_pcr_data vert_pcr_data;
	mat3_to_mat4(final_matrix, vert_pcr_data.mat4);
	vert_pcr_data.uv_off[0] = 0.f;
	vert_pcr_data.uv_off[1] = 0.f;
	vert_pcr_data.uv_size[0] = 1.f;
	vert_pcr_data.uv_size[1] = 1.f;

	// Input color values are given in srgb space, shader expects
	// them in linear space. The shader does all computation in linear
	// space and expects in inputs in linear space since it outputs
	// colors in linear space as well (and vulkan then automatically
	// does the conversion for out SRGB render targets).
	// But in other parts of wlroots we just always assume
	// srgb so that's why we have to convert here.
	float linear_color[4];
	linear_color[0] = color_to_linear(color[0]);
	linear_color[1] = color_to_linear(color[1]);
	linear_color[2] = color_to_linear(color[2]);
	linear_color[3] = color[3]; // no conversion for alpha

	vkCmdPushConstants(cb, pipe->layout->vk,
		VK_SHADER_STAGE_VERTEX_BIT, 0, sizeof(vert_pcr_data), &vert_pcr_data);
	vkCmdPushConstants(cb, pipe->layout->vk,
		VK_SHADER_STAGE_FRAGMENT_BIT, sizeof(vert_pcr_data), sizeof(float) * 4,
		linear_color);
	vkCmdDraw(cb, 4, 1, 0, 0);
}

static const struct wlr_drm_format_set *vulkan_get_dmabuf_texture_formats(
		struct wlr_renderer *wlr_renderer) {
	struct wlr_vk_renderer *renderer = vulkan_get_renderer(wlr_renderer);
	return &renderer->dev->dmabuf_texture_formats;
}

static const struct wlr_drm_format_set *vulkan_get_render_formats(
		struct wlr_renderer *wlr_renderer) {
	struct wlr_vk_renderer *renderer = vulkan_get_renderer(wlr_renderer);
	return &renderer->dev->dmabuf_render_formats;
}

static uint32_t vulkan_preferred_read_format(
		struct wlr_renderer *wlr_renderer) {
	struct wlr_vk_renderer *renderer = vulkan_get_renderer(wlr_renderer);
	struct wlr_dmabuf_attributes dmabuf = {0};
	if (!wlr_buffer_get_dmabuf(renderer->current_render_buffer->wlr_buffer,
				&dmabuf)) {
		wlr_log(WLR_ERROR, "vulkan_preferred_read_format: Failed to get dmabuf of current render buffer");
		return DRM_FORMAT_INVALID;
	}
	return dmabuf.format;
}

static void vulkan_destroy(struct wlr_renderer *wlr_renderer) {
	struct wlr_vk_renderer *renderer = vulkan_get_renderer(wlr_renderer);
	struct wlr_vk_device *dev = renderer->dev;
	if (!dev) {
		free(renderer);
		return;
	}

	assert(!renderer->current_render_buffer);

	VkResult res = vkDeviceWaitIdle(renderer->dev->dev);
	if (res != VK_SUCCESS) {
		wlr_vk_error("vkDeviceWaitIdle", res);
	}

	for (size_t i = 0; i < VULKAN_COMMAND_BUFFERS_CAP; i++) {
		struct wlr_vk_command_buffer *cb = &renderer->command_buffers[i];
		if (cb->vk == VK_NULL_HANDLE) {
			continue;
		}
		release_command_buffer_resources(cb, renderer);
		if (cb->binary_semaphore != VK_NULL_HANDLE) {
			vkDestroySemaphore(renderer->dev->dev, cb->binary_semaphore, NULL);
		}
	}

	// stage.cb automatically freed with command pool
	struct wlr_vk_shared_buffer *buf, *tmp_buf;
	wl_list_for_each_safe(buf, tmp_buf, &renderer->stage.buffers, link) {
		shared_buffer_destroy(renderer, buf);
	}

	struct wlr_vk_texture *tex, *tex_tmp;
	wl_list_for_each_safe(tex, tex_tmp, &renderer->textures, link) {
		vulkan_texture_destroy(tex);
	}

	struct wlr_vk_render_buffer *render_buffer, *render_buffer_tmp;
	wl_list_for_each_safe(render_buffer, render_buffer_tmp,
			&renderer->render_buffers, link) {
		destroy_render_buffer(render_buffer);
	}

	struct wlr_vk_render_format_setup *setup, *tmp_setup;
	wl_list_for_each_safe(setup, tmp_setup,
			&renderer->render_format_setups, link) {
		destroy_render_format_setup(renderer, setup);
	}

	struct wlr_vk_descriptor_pool *pool, *tmp_pool;
	wl_list_for_each_safe(pool, tmp_pool, &renderer->descriptor_pools, link) {
		vkDestroyDescriptorPool(dev->dev, pool->pool, NULL);
		free(pool);
	}
	wl_list_for_each_safe(pool, tmp_pool, &renderer->output_descriptor_pools, link) {
		vkDestroyDescriptorPool(dev->dev, pool->pool, NULL);
		free(pool);
	}

	vkDestroyShaderModule(dev->dev, renderer->vert_module, NULL);
	vkDestroyShaderModule(dev->dev, renderer->tex_frag_module, NULL);
	vkDestroyShaderModule(dev->dev, renderer->quad_frag_module, NULL);
	vkDestroyShaderModule(dev->dev, renderer->output_module, NULL);

	struct wlr_vk_pipeline_layout *pipeline_layout, *pipeline_layout_tmp;
	wl_list_for_each_safe(pipeline_layout, pipeline_layout_tmp,
			&renderer->pipeline_layouts, link) {
		vkDestroyPipelineLayout(dev->dev, pipeline_layout->vk, NULL);
		vkDestroyDescriptorSetLayout(dev->dev, pipeline_layout->ds, NULL);
		vkDestroySampler(dev->dev, pipeline_layout->sampler, NULL);
		vkDestroySamplerYcbcrConversion(dev->dev, pipeline_layout->ycbcr.conversion, NULL);
	}

	vkDestroySemaphore(dev->dev, renderer->timeline_semaphore, NULL);
	vkDestroyPipelineLayout(dev->dev, renderer->output_pipe_layout, NULL);
	vkDestroyDescriptorSetLayout(dev->dev, renderer->output_ds_layout, NULL);
	vkDestroyCommandPool(dev->dev, renderer->command_pool, NULL);

	if (renderer->read_pixels_cache.initialized) {
		vkFreeMemory(dev->dev, renderer->read_pixels_cache.dst_img_memory, NULL);
		vkDestroyImage(dev->dev, renderer->read_pixels_cache.dst_image, NULL);
	}

	struct wlr_vk_instance *ini = dev->instance;
	vulkan_device_destroy(dev);
	vulkan_instance_destroy(ini);
	free(renderer);
}

static bool vulkan_read_pixels(struct wlr_renderer *wlr_renderer,
		uint32_t drm_format, uint32_t stride,
		uint32_t width, uint32_t height, uint32_t src_x, uint32_t src_y,
		uint32_t dst_x, uint32_t dst_y, void *data) {
	struct wlr_vk_renderer *vk_renderer = vulkan_get_renderer(wlr_renderer);
	VkDevice dev = vk_renderer->dev->dev;
	VkImage src_image = vk_renderer->current_render_buffer->image;

	const struct wlr_pixel_format_info *pixel_format_info = drm_get_pixel_format_info(drm_format);
	if (!pixel_format_info) {
		wlr_log(WLR_ERROR, "vulkan_read_pixels: could not find pixel format info "
				"for DRM format 0x%08x", drm_format);
		return false;
	} else if (pixel_format_info_pixels_per_block(pixel_format_info) != 1) {
		wlr_log(WLR_ERROR, "vulkan_read_pixels: block formats are not supported");
		return false;
	}

	const struct wlr_vk_format *wlr_vk_format = vulkan_get_format_from_drm(drm_format);
	if (!wlr_vk_format) {
		wlr_log(WLR_ERROR, "vulkan_read_pixels: no vulkan format "
				"matching drm format 0x%08x available", drm_format);
		return false;
	}
	VkFormat dst_format = wlr_vk_format->vk;
	VkFormat src_format = vk_renderer->current_render_buffer->render_setup->render_format->vk;
	VkFormatProperties dst_format_props = {0}, src_format_props = {0};
	vkGetPhysicalDeviceFormatProperties(vk_renderer->dev->phdev, dst_format, &dst_format_props);
	vkGetPhysicalDeviceFormatProperties(vk_renderer->dev->phdev, src_format, &src_format_props);

	bool blit_supported = src_format_props.optimalTilingFeatures & VK_FORMAT_FEATURE_BLIT_SRC_BIT &&
		dst_format_props.linearTilingFeatures & VK_FORMAT_FEATURE_BLIT_DST_BIT;
	if (!blit_supported && src_format != dst_format) {
		wlr_log(WLR_ERROR, "vulkan_read_pixels: blit unsupported and no manual "
					"conversion available from src to dst format.");
		return false;
	}

	VkResult res;
	VkImage dst_image;
	VkDeviceMemory dst_img_memory;
	bool use_cached = vk_renderer->read_pixels_cache.initialized &&
		vk_renderer->read_pixels_cache.drm_format == drm_format &&
		vk_renderer->read_pixels_cache.width == width &&
		vk_renderer->read_pixels_cache.height == height;

	if (use_cached) {
		dst_image = vk_renderer->read_pixels_cache.dst_image;
		dst_img_memory = vk_renderer->read_pixels_cache.dst_img_memory;
	} else {
		VkImageCreateInfo image_create_info = {
			.sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO,
			.imageType = VK_IMAGE_TYPE_2D,
			.format = dst_format,
			.extent.width = width,
			.extent.height = height,
			.extent.depth = 1,
			.arrayLayers = 1,
			.mipLevels = 1,
			.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED,
			.samples = VK_SAMPLE_COUNT_1_BIT,
			.tiling = VK_IMAGE_TILING_LINEAR,
			.usage = VK_IMAGE_USAGE_TRANSFER_DST_BIT
		};
		res = vkCreateImage(dev, &image_create_info, NULL, &dst_image);
		if (res != VK_SUCCESS) {
			wlr_vk_error("vkCreateImage", res);
			return false;
		}

		VkMemoryRequirements mem_reqs;
		vkGetImageMemoryRequirements(dev, dst_image, &mem_reqs);

		int mem_type = vulkan_find_mem_type(vk_renderer->dev,
				VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT |
				VK_MEMORY_PROPERTY_HOST_COHERENT_BIT |
				VK_MEMORY_PROPERTY_HOST_CACHED_BIT,
				mem_reqs.memoryTypeBits);
		if (mem_type < 0) {
			wlr_log(WLR_ERROR, "vulkan_read_pixels: could not find adequate memory type");
			goto destroy_image;
		}

		VkMemoryAllocateInfo mem_alloc_info = {
			.sType = VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO,
		};
		mem_alloc_info.allocationSize = mem_reqs.size;
		mem_alloc_info.memoryTypeIndex = mem_type;

		res = vkAllocateMemory(dev, &mem_alloc_info, NULL, &dst_img_memory);
		if (res != VK_SUCCESS) {
			wlr_vk_error("vkAllocateMemory", res);
			goto destroy_image;
		}
		res = vkBindImageMemory(dev, dst_image, dst_img_memory, 0);
		if (res != VK_SUCCESS) {
			wlr_vk_error("vkBindImageMemory", res);
			goto free_memory;
		}

		if (vk_renderer->read_pixels_cache.initialized) {
			vkFreeMemory(dev, vk_renderer->read_pixels_cache.dst_img_memory, NULL);
			vkDestroyImage(dev, vk_renderer->read_pixels_cache.dst_image, NULL);
		}
		vk_renderer->read_pixels_cache.initialized = true;
		vk_renderer->read_pixels_cache.drm_format = drm_format;
		vk_renderer->read_pixels_cache.dst_image = dst_image;
		vk_renderer->read_pixels_cache.dst_img_memory = dst_img_memory;
		vk_renderer->read_pixels_cache.width = width;
		vk_renderer->read_pixels_cache.height = height;
	}

	VkCommandBuffer cb = vulkan_record_stage_cb(vk_renderer);
	if (cb == VK_NULL_HANDLE) {
		return false;
	}

	vulkan_change_layout(cb, dst_image,
			VK_IMAGE_LAYOUT_UNDEFINED,
			VK_PIPELINE_STAGE_TRANSFER_BIT,
			0,
			VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,
			VK_PIPELINE_STAGE_TRANSFER_BIT,
			VK_ACCESS_TRANSFER_WRITE_BIT);
	vulkan_change_layout(cb, src_image,
			VK_IMAGE_LAYOUT_GENERAL,
			VK_PIPELINE_STAGE_TRANSFER_BIT,
			VK_ACCESS_MEMORY_READ_BIT,
			VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL,
			VK_PIPELINE_STAGE_TRANSFER_BIT,
			VK_ACCESS_TRANSFER_READ_BIT);

	if (blit_supported) {
		VkImageBlit image_blit_region = {
			.srcSubresource.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
			.srcSubresource.layerCount = 1,
			.srcOffsets[0] = {
				.x = src_x,
				.y = src_y,
				.z = 0,
			},
			.srcOffsets[1] = {
				.x = src_x + width,
				.y = src_y + height,
				.z = 1,
			},
			.dstSubresource.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
			.dstSubresource.layerCount = 1,
			.dstOffsets[1] = {
				.x = width,
				.y = height,
				.z = 1,
			}
		};
		vkCmdBlitImage(cb, src_image, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL,
				dst_image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,
				1, &image_blit_region, VK_FILTER_NEAREST);
	} else {
		wlr_log(WLR_DEBUG, "vulkan_read_pixels: blit unsupported, falling back to vkCmdCopyImage.");
		VkImageCopy image_region = {
			.srcSubresource.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
			.srcSubresource.layerCount = 1,
			.srcOffset = {
				.x = src_x,
				.y = src_y,
			},
			.dstSubresource.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
			.dstSubresource.layerCount = 1,
			.extent = {
				.width = width,
				.height = height,
				.depth = 1,
			}
		};
		vkCmdCopyImage(cb, src_image, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL,
				dst_image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, 1, &image_region);
	}

	vulkan_change_layout(cb, dst_image,
			VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,
			VK_PIPELINE_STAGE_TRANSFER_BIT,
			VK_ACCESS_TRANSFER_WRITE_BIT,
			VK_IMAGE_LAYOUT_GENERAL,
			VK_PIPELINE_STAGE_TRANSFER_BIT,
			0);
	vulkan_change_layout(cb, src_image,
			VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL,
			VK_PIPELINE_STAGE_TRANSFER_BIT,
			VK_ACCESS_TRANSFER_READ_BIT,
			VK_IMAGE_LAYOUT_GENERAL,
			VK_PIPELINE_STAGE_TRANSFER_BIT,
			VK_ACCESS_MEMORY_READ_BIT);

	if (!vulkan_submit_stage_wait(vk_renderer)) {
		return false;
	}

	VkImageSubresource img_sub_res = {
		.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
		.arrayLayer = 0,
		.mipLevel = 0
	};
	VkSubresourceLayout img_sub_layout;
	vkGetImageSubresourceLayout(dev, dst_image, &img_sub_res, &img_sub_layout);

	void *v;
	res = vkMapMemory(dev, dst_img_memory, 0, VK_WHOLE_SIZE, 0, &v);
	if (res != VK_SUCCESS) {
		wlr_vk_error("vkMapMemory", res);
		return false;
	}

	const char *d = (const char *)v + img_sub_layout.offset;
	unsigned char *p = (unsigned char *)data + dst_y * stride;
	uint32_t bytes_per_pixel = pixel_format_info->bytes_per_block;
	uint32_t pack_stride = img_sub_layout.rowPitch;
	if (pack_stride == stride && dst_x == 0) {
		memcpy(p, d, height * stride);
	} else {
		for (size_t i = 0; i < height; ++i) {
			memcpy(p + i * stride + dst_x * bytes_per_pixel, d + i * pack_stride, width * bytes_per_pixel);
		}
	}

	vkUnmapMemory(dev, dst_img_memory);
	// Don't need to free anything else, since memory and image are cached
	return true;
free_memory:
	vkFreeMemory(dev, dst_img_memory, NULL);
destroy_image:
	vkDestroyImage(dev, dst_image, NULL);

	return false;
}

static int vulkan_get_drm_fd(struct wlr_renderer *wlr_renderer) {
	struct wlr_vk_renderer *renderer = vulkan_get_renderer(wlr_renderer);
	return renderer->dev->drm_fd;
}

static uint32_t vulkan_get_render_buffer_caps(struct wlr_renderer *wlr_renderer) {
	return WLR_BUFFER_CAP_DMABUF;
}

static struct wlr_render_pass *vulkan_begin_buffer_pass(struct wlr_renderer *wlr_renderer,
		struct wlr_buffer *buffer, const struct wlr_buffer_pass_options *options) {
	struct wlr_vk_renderer *renderer = vulkan_get_renderer(wlr_renderer);

	struct wlr_vk_render_buffer *render_buffer = get_render_buffer(renderer, buffer);
	if (!render_buffer) {
		render_buffer = create_render_buffer(renderer, buffer);
		if (!render_buffer) {
			return NULL;
		}
	}

	struct wlr_vk_render_pass *render_pass = vulkan_begin_render_pass(renderer, render_buffer);
	if (render_pass == NULL) {
		return NULL;
	}
	return &render_pass->base;
}

static const struct wlr_renderer_impl renderer_impl = {
	.bind_buffer = vulkan_bind_buffer,
	.begin = vulkan_begin,
	.end = vulkan_end,
	.clear = vulkan_clear,
	.scissor = vulkan_scissor,
	.render_subtexture_with_matrix = vulkan_render_subtexture_with_matrix,
	.render_quad_with_matrix = vulkan_render_quad_with_matrix,
	.get_shm_texture_formats = vulkan_get_shm_texture_formats,
	.get_dmabuf_texture_formats = vulkan_get_dmabuf_texture_formats,
	.get_render_formats = vulkan_get_render_formats,
	.preferred_read_format = vulkan_preferred_read_format,
	.read_pixels = vulkan_read_pixels,
	.destroy = vulkan_destroy,
	.get_drm_fd = vulkan_get_drm_fd,
	.get_render_buffer_caps = vulkan_get_render_buffer_caps,
	.texture_from_buffer = vulkan_texture_from_buffer,
	.begin_buffer_pass = vulkan_begin_buffer_pass,
};

// Initializes the VkDescriptorSetLayout and VkPipelineLayout needed
// for the texture rendering pipeline using the given VkSampler.
static bool init_tex_layouts(struct wlr_vk_renderer *renderer,
		VkSampler tex_sampler, VkDescriptorSetLayout *out_ds_layout,
		VkPipelineLayout *out_pipe_layout) {
	VkResult res;
	VkDevice dev = renderer->dev->dev;

	VkDescriptorSetLayoutBinding ds_binding = {
		.binding = 0,
		.descriptorType = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER,
		.descriptorCount = 1,
		.stageFlags = VK_SHADER_STAGE_FRAGMENT_BIT,
		.pImmutableSamplers = &tex_sampler,
	};

	VkDescriptorSetLayoutCreateInfo ds_info = {
		.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO,
		.bindingCount = 1,
		.pBindings = &ds_binding,
	};

	res = vkCreateDescriptorSetLayout(dev, &ds_info, NULL, out_ds_layout);
	if (res != VK_SUCCESS) {
		wlr_vk_error("vkCreateDescriptorSetLayout", res);
		return false;
	}

	VkPushConstantRange pc_ranges[2] = {
		{
			.size = sizeof(struct wlr_vk_vert_pcr_data),
			.stageFlags = VK_SHADER_STAGE_VERTEX_BIT,
		},
		{
			.offset = pc_ranges[0].size,
			.size = sizeof(float) * 4, // alpha or color
			.stageFlags = VK_SHADER_STAGE_FRAGMENT_BIT,
		},
	};

	VkPipelineLayoutCreateInfo pl_info = {
		.sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO,
		.setLayoutCount = 1,
		.pSetLayouts = out_ds_layout,
		.pushConstantRangeCount = 2,
		.pPushConstantRanges = pc_ranges,
	};

	res = vkCreatePipelineLayout(dev, &pl_info, NULL, out_pipe_layout);
	if (res != VK_SUCCESS) {
		wlr_vk_error("vkCreatePipelineLayout", res);
		return false;
	}

	return true;
}

static bool init_blend_to_output_layouts(struct wlr_vk_renderer *renderer,
		VkDescriptorSetLayout *out_ds_layout,
		VkPipelineLayout *out_pipe_layout) {
	VkResult res;
	VkDevice dev = renderer->dev->dev;

	VkDescriptorSetLayoutBinding ds_binding = {
		.binding = 0,
		.descriptorType = VK_DESCRIPTOR_TYPE_INPUT_ATTACHMENT,
		.descriptorCount = 1,
		.stageFlags = VK_SHADER_STAGE_FRAGMENT_BIT,
		.pImmutableSamplers = NULL,
	};

	VkDescriptorSetLayoutCreateInfo ds_info = {
		.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO,
		.bindingCount = 1,
		.pBindings = &ds_binding,
	};

	res = vkCreateDescriptorSetLayout(dev, &ds_info, NULL, out_ds_layout);
	if (res != VK_SUCCESS) {
		wlr_vk_error("vkCreateDescriptorSetLayout", res);
		return false;
	}

	// pipeline layout -- standard vertex uniforms, no shader uniforms
	VkPushConstantRange pc_ranges[1] = {
		{
			.size = sizeof(struct wlr_vk_vert_pcr_data),
			.stageFlags = VK_SHADER_STAGE_VERTEX_BIT,
		},
	};

	VkPipelineLayoutCreateInfo pl_info = {
		.sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO,
		.setLayoutCount = 1,
		.pSetLayouts = out_ds_layout,
		.pushConstantRangeCount = 1,
		.pPushConstantRanges = pc_ranges,
	};

	res = vkCreatePipelineLayout(dev, &pl_info, NULL, out_pipe_layout);
	if (res != VK_SUCCESS) {
		wlr_vk_error("vkCreatePipelineLayout", res);
		return false;
	}

	return true;
}

static bool pipeline_layout_key_equals(
		const struct wlr_vk_pipeline_layout_key *a,
		const struct wlr_vk_pipeline_layout_key *b) {
	assert(!a->ycbcr_format || a->ycbcr_format->is_ycbcr);
	assert(!b->ycbcr_format || b->ycbcr_format->is_ycbcr);

	if (a->filter_mode != b->filter_mode) {
		return false;
	}

	if (a->ycbcr_format != b->ycbcr_format) {
		return false;
	}

	return true;
}

static bool pipeline_key_equals(const struct wlr_vk_pipeline_key *a,
		const struct wlr_vk_pipeline_key *b) {
	if (!pipeline_layout_key_equals(&a->layout, &b->layout)) {
		return false;
	}

	if (a->blend_mode != b->blend_mode) {
		return false;
	}

	if (a->source != b->source) {
		return false;
	}

	if (a->source == WLR_VK_SHADER_SOURCE_TEXTURE &&
			a->texture_transform != b->texture_transform) {
		return false;
	}

	return true;
}

// Initializes the pipeline for rendering textures and using the given
// VkRenderPass and VkPipelineLayout.
struct wlr_vk_pipeline *setup_get_or_create_pipeline(
		struct wlr_vk_render_format_setup *setup,
		const struct wlr_vk_pipeline_key *key) {
	struct wlr_vk_pipeline *pipeline;
	wl_list_for_each(pipeline, &setup->pipelines, link) {
		if (pipeline_key_equals(&pipeline->key, key)) {
			return pipeline;
		}
	}

	struct wlr_vk_renderer *renderer = setup->renderer;

	struct wlr_vk_pipeline_layout *pipeline_layout = get_or_create_pipeline_layout(
		renderer, &key->layout);
	if (!pipeline_layout) {
		return NULL;
	}

	pipeline = calloc(1, sizeof(*pipeline));
	if (!pipeline) {
		return NULL;
	}

	pipeline->setup = setup;
	pipeline->key = *key;
	pipeline->layout = pipeline_layout;

	VkResult res;
	VkDevice dev = renderer->dev->dev;

	uint32_t color_transform_type = key->texture_transform;

	VkSpecializationMapEntry spec_entry = {
		.constantID = 0,
		.offset = 0,
		.size = sizeof(uint32_t),
	};

	VkSpecializationInfo specialization = {
		.mapEntryCount = 1,
		.pMapEntries = &spec_entry,
		.dataSize = sizeof(uint32_t),
		.pData = &color_transform_type,
	};

	VkPipelineShaderStageCreateInfo stages[2];
	stages[0] = (VkPipelineShaderStageCreateInfo) {
		.sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO,
		.stage = VK_SHADER_STAGE_VERTEX_BIT,
		.module = renderer->vert_module,
		.pName = "main",
	};

	switch (key->source) {
	case WLR_VK_SHADER_SOURCE_SINGLE_COLOR:
		stages[1] = (VkPipelineShaderStageCreateInfo) {
			.sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO,
			.stage = VK_SHADER_STAGE_FRAGMENT_BIT,
			.module = renderer->quad_frag_module,
			.pName = "main",
		};
		break;
	case WLR_VK_SHADER_SOURCE_TEXTURE:
		stages[1] = (VkPipelineShaderStageCreateInfo) {
			.sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO,
			.stage = VK_SHADER_STAGE_FRAGMENT_BIT,
			.module = renderer->tex_frag_module,
			.pName = "main",
			.pSpecializationInfo = &specialization,
		};
		break;
	}

	VkPipelineInputAssemblyStateCreateInfo assembly = {
		.sType = VK_STRUCTURE_TYPE_PIPELINE_INPUT_ASSEMBLY_STATE_CREATE_INFO,
		.topology = VK_PRIMITIVE_TOPOLOGY_TRIANGLE_FAN,
	};

	VkPipelineRasterizationStateCreateInfo rasterization = {
		.sType = VK_STRUCTURE_TYPE_PIPELINE_RASTERIZATION_STATE_CREATE_INFO,
		.polygonMode = VK_POLYGON_MODE_FILL,
		.cullMode = VK_CULL_MODE_NONE,
		.frontFace = VK_FRONT_FACE_COUNTER_CLOCKWISE,
		.lineWidth = 1.f,
	};

	VkPipelineColorBlendAttachmentState blend_attachment = {
		.blendEnable = key->blend_mode == WLR_RENDER_BLEND_MODE_PREMULTIPLIED,
		// we generally work with pre-multiplied alpha
		.srcColorBlendFactor = VK_BLEND_FACTOR_ONE,
		.dstColorBlendFactor = VK_BLEND_FACTOR_ONE_MINUS_SRC_ALPHA,
		.colorBlendOp = VK_BLEND_OP_ADD,
		.srcAlphaBlendFactor = VK_BLEND_FACTOR_ONE,
		.dstAlphaBlendFactor = VK_BLEND_FACTOR_ONE_MINUS_SRC_ALPHA,
		.alphaBlendOp = VK_BLEND_OP_ADD,
		.colorWriteMask =
			VK_COLOR_COMPONENT_R_BIT |
			VK_COLOR_COMPONENT_G_BIT |
			VK_COLOR_COMPONENT_B_BIT |
			VK_COLOR_COMPONENT_A_BIT,
	};

	VkPipelineColorBlendStateCreateInfo blend = {
		.sType = VK_STRUCTURE_TYPE_PIPELINE_COLOR_BLEND_STATE_CREATE_INFO,
		.attachmentCount = 1,
		.pAttachments = &blend_attachment,
	};

	VkPipelineMultisampleStateCreateInfo multisample = {
		.sType = VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO,
		.rasterizationSamples = VK_SAMPLE_COUNT_1_BIT,
	};

	VkPipelineViewportStateCreateInfo viewport = {
		.sType = VK_STRUCTURE_TYPE_PIPELINE_VIEWPORT_STATE_CREATE_INFO,
		.viewportCount = 1,
		.scissorCount = 1,
	};

	VkDynamicState dynStates[2] = {
		VK_DYNAMIC_STATE_VIEWPORT,
		VK_DYNAMIC_STATE_SCISSOR,
	};
	VkPipelineDynamicStateCreateInfo dynamic = {
		.sType = VK_STRUCTURE_TYPE_PIPELINE_DYNAMIC_STATE_CREATE_INFO,
		.pDynamicStates = dynStates,
		.dynamicStateCount = 2,
	};

	VkPipelineVertexInputStateCreateInfo vertex = {
		.sType = VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO,
	};

	VkGraphicsPipelineCreateInfo pinfo = {
		.sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO,
		.layout = pipeline_layout->vk,
		.renderPass = setup->render_pass,
		.subpass = 0,
		.stageCount = 2,
		.pStages = stages,

		.pInputAssemblyState = &assembly,
		.pRasterizationState = &rasterization,
		.pColorBlendState = &blend,
		.pMultisampleState = &multisample,
		.pViewportState = &viewport,
		.pDynamicState = &dynamic,
		.pVertexInputState = &vertex,
	};

	VkPipelineCache cache = VK_NULL_HANDLE;
	res = vkCreateGraphicsPipelines(dev, cache, 1, &pinfo, NULL, &pipeline->vk);
	if (res != VK_SUCCESS) {
		wlr_vk_error("failed to create vulkan pipelines:", res);
		free(pipeline);
		return NULL;
	}

	wl_list_insert(&setup->pipelines, &pipeline->link);
	return pipeline;
}

static bool init_blend_to_output_pipeline(struct wlr_vk_renderer *renderer,
		VkRenderPass rp, VkPipelineLayout pipe_layout, VkPipeline *pipe) {
	VkResult res;
	VkDevice dev = renderer->dev->dev;

	VkPipelineShaderStageCreateInfo tex_stages[2] = {
		{
			.sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO,
			.stage = VK_SHADER_STAGE_VERTEX_BIT,
			.module = renderer->vert_module,
			.pName = "main",
		},
		{
			.sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO,
			.stage = VK_SHADER_STAGE_FRAGMENT_BIT,
			.module = renderer->output_module,
			.pName = "main",
		},
	};

	VkPipelineInputAssemblyStateCreateInfo assembly = {
		.sType = VK_STRUCTURE_TYPE_PIPELINE_INPUT_ASSEMBLY_STATE_CREATE_INFO,
		.topology = VK_PRIMITIVE_TOPOLOGY_TRIANGLE_FAN,
	};

	VkPipelineRasterizationStateCreateInfo rasterization = {
		.sType = VK_STRUCTURE_TYPE_PIPELINE_RASTERIZATION_STATE_CREATE_INFO,
		.polygonMode = VK_POLYGON_MODE_FILL,
		.cullMode = VK_CULL_MODE_NONE,
		.frontFace = VK_FRONT_FACE_COUNTER_CLOCKWISE,
		.lineWidth = 1.f,
	};

	VkPipelineColorBlendAttachmentState blend_attachment = {
		.blendEnable = false,
		.colorWriteMask =
			VK_COLOR_COMPONENT_R_BIT |
			VK_COLOR_COMPONENT_G_BIT |
			VK_COLOR_COMPONENT_B_BIT |
			VK_COLOR_COMPONENT_A_BIT,
	};

	VkPipelineColorBlendStateCreateInfo blend = {
		.sType = VK_STRUCTURE_TYPE_PIPELINE_COLOR_BLEND_STATE_CREATE_INFO,
		.attachmentCount = 1,
		.pAttachments = &blend_attachment,
	};

	VkPipelineMultisampleStateCreateInfo multisample = {
		.sType = VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO,
		.rasterizationSamples = VK_SAMPLE_COUNT_1_BIT,
	};

	VkPipelineViewportStateCreateInfo viewport = {
		.sType = VK_STRUCTURE_TYPE_PIPELINE_VIEWPORT_STATE_CREATE_INFO,
		.viewportCount = 1,
		.scissorCount = 1,
	};

	VkDynamicState dynStates[2] = {
		VK_DYNAMIC_STATE_VIEWPORT,
		VK_DYNAMIC_STATE_SCISSOR,
	};
	VkPipelineDynamicStateCreateInfo dynamic = {
		.sType = VK_STRUCTURE_TYPE_PIPELINE_DYNAMIC_STATE_CREATE_INFO,
		.pDynamicStates = dynStates,
		.dynamicStateCount = 2,
	};

	VkPipelineVertexInputStateCreateInfo vertex = {
		.sType = VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO,
	};

	VkGraphicsPipelineCreateInfo pinfo = {
		.sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO,
		.pNext = NULL,
		.layout = pipe_layout,
		.renderPass = rp,
		.subpass = 1, // second subpass!
		.stageCount = 2,
		.pStages = tex_stages,
		.pInputAssemblyState = &assembly,
		.pRasterizationState = &rasterization,
		.pColorBlendState = &blend,
		.pMultisampleState = &multisample,
		.pViewportState = &viewport,
		.pDynamicState = &dynamic,
		.pVertexInputState = &vertex,
	};

	VkPipelineCache cache = VK_NULL_HANDLE;
	res = vkCreateGraphicsPipelines(dev, cache, 1, &pinfo, NULL, pipe);
	if (res != VK_SUCCESS) {
		wlr_vk_error("failed to create vulkan pipelines:", res);
		return false;
	}

	return true;
}

struct wlr_vk_pipeline_layout *get_or_create_pipeline_layout(
		struct wlr_vk_renderer *renderer,
		const struct wlr_vk_pipeline_layout_key *key) {
	struct wlr_vk_pipeline_layout *pipeline_layout;
	wl_list_for_each(pipeline_layout, &renderer->pipeline_layouts, link) {
		if (pipeline_layout_key_equals(&pipeline_layout->key, key)) {
			return pipeline_layout;
		}
	}

	pipeline_layout = calloc(1, sizeof(*pipeline_layout));
	if (!pipeline_layout) {
		return NULL;
	}

	pipeline_layout->key = *key;

	VkResult res;
	VkFilter filter = VK_FILTER_LINEAR;
	switch (key->filter_mode) {
	case WLR_SCALE_FILTER_BILINEAR:
		filter = VK_FILTER_LINEAR;
		break;
	case WLR_SCALE_FILTER_NEAREST:
		filter = VK_FILTER_NEAREST;
		break;
	}

	VkSamplerYcbcrConversionInfo conversion_info;
	VkSamplerCreateInfo sampler_create_info = {
		.sType = VK_STRUCTURE_TYPE_SAMPLER_CREATE_INFO,
		.magFilter = filter,
		.minFilter = filter,
		.mipmapMode = VK_SAMPLER_MIPMAP_MODE_NEAREST,
		.addressModeU = VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE,
		.addressModeV = VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE,
		.addressModeW = VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE,
		.minLod = 0.f,
		.maxLod = 0.25f,
	};

	if (key->ycbcr_format) {
		VkSamplerYcbcrConversionCreateInfo conversion_create_info = {
			.sType = VK_STRUCTURE_TYPE_SAMPLER_YCBCR_CONVERSION_CREATE_INFO,
			.format = key->ycbcr_format->vk,
			.ycbcrModel = VK_SAMPLER_YCBCR_MODEL_CONVERSION_YCBCR_601,
			.ycbcrRange = VK_SAMPLER_YCBCR_RANGE_ITU_NARROW,
			.xChromaOffset = VK_CHROMA_LOCATION_MIDPOINT,
			.yChromaOffset = VK_CHROMA_LOCATION_MIDPOINT,
			.chromaFilter = VK_FILTER_LINEAR,
		};
		res = vkCreateSamplerYcbcrConversion(renderer->dev->dev,
			&conversion_create_info, NULL, &pipeline_layout->ycbcr.conversion);
		if (res != VK_SUCCESS) {
			wlr_vk_error("vkCreateSamplerYcbcrConversion", res);
			free(pipeline_layout);
			return NULL;
		}

		conversion_info = (VkSamplerYcbcrConversionInfo){
			.sType = VK_STRUCTURE_TYPE_SAMPLER_YCBCR_CONVERSION_INFO,
			.conversion = pipeline_layout->ycbcr.conversion,
		};
		sampler_create_info.pNext = &conversion_info;
	}

	res = vkCreateSampler(renderer->dev->dev, &sampler_create_info, NULL, &pipeline_layout->sampler);
	if (res != VK_SUCCESS) {
		wlr_vk_error("vkCreateSampler", res);
		free(pipeline_layout);
		return false;
	}

	if (!init_tex_layouts(renderer, pipeline_layout->sampler, &pipeline_layout->ds, &pipeline_layout->vk)) {
		free(pipeline_layout);
		return false;
	}

	wl_list_insert(&renderer->pipeline_layouts, &pipeline_layout->link);
	return pipeline_layout;
}

// Creates static render data, such as sampler, layouts and shader modules
// for the given renderer.
// Cleanup is done by destroying the renderer.
static bool init_static_render_data(struct wlr_vk_renderer *renderer) {
	VkResult res;
	VkDevice dev = renderer->dev->dev;

	if (!init_blend_to_output_layouts(renderer, &renderer->output_ds_layout,
			&renderer->output_pipe_layout)) {
		return false;
	}

	// load vert module and tex frag module since they are needed to
	// initialize the tex pipeline
	VkShaderModuleCreateInfo sinfo = {
		.sType = VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO,
		.codeSize = sizeof(common_vert_data),
		.pCode = common_vert_data,
	};
	res = vkCreateShaderModule(dev, &sinfo, NULL, &renderer->vert_module);
	if (res != VK_SUCCESS) {
		wlr_vk_error("Failed to create vertex shader module", res);
		return false;
	}

	sinfo = (VkShaderModuleCreateInfo){
		.sType = VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO,
		.codeSize = sizeof(texture_frag_data),
		.pCode = texture_frag_data,
	};
	res = vkCreateShaderModule(dev, &sinfo, NULL, &renderer->tex_frag_module);
	if (res != VK_SUCCESS) {
		wlr_vk_error("Failed to create tex fragment shader module", res);
		return false;
	}

	sinfo = (VkShaderModuleCreateInfo){
		.sType = VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO,
		.codeSize = sizeof(quad_frag_data),
		.pCode = quad_frag_data,
	};
	res = vkCreateShaderModule(dev, &sinfo, NULL, &renderer->quad_frag_module);
	if (res != VK_SUCCESS) {
		wlr_vk_error("Failed to create quad fragment shader module", res);
		return false;
	}

	sinfo = (VkShaderModuleCreateInfo){
		.sType = VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO,
		.codeSize = sizeof(output_frag_data),
		.pCode = output_frag_data,
	};
	res = vkCreateShaderModule(dev, &sinfo, NULL, &renderer->output_module);
	if (res != VK_SUCCESS) {
		wlr_vk_error("Failed to create blend->output fragment shader module", res);
		return false;
	}

	return true;
}

static struct wlr_vk_render_format_setup *find_or_create_render_setup(
		struct wlr_vk_renderer *renderer, const struct wlr_vk_format *format,
		bool has_blending_buffer) {
	struct wlr_vk_render_format_setup *setup;
	wl_list_for_each(setup, &renderer->render_format_setups, link) {
		if (setup->render_format == format) {
			return setup;
		}
	}

	setup = calloc(1u, sizeof(*setup));
	if (!setup) {
		wlr_log(WLR_ERROR, "Allocation failed");
		return NULL;
	}

	setup->render_format = format;
	setup->renderer = renderer;
	wl_list_init(&setup->pipelines);

	VkDevice dev = renderer->dev->dev;
	VkResult res;

	if (has_blending_buffer) {
		VkAttachmentDescription attachments[2] = {
			{
				.format = VK_FORMAT_R16G16B16A16_SFLOAT,
				.samples = VK_SAMPLE_COUNT_1_BIT,
				.loadOp = VK_ATTACHMENT_LOAD_OP_LOAD,
				.storeOp = VK_ATTACHMENT_STORE_OP_STORE,
				.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE,
				.stencilStoreOp = VK_ATTACHMENT_STORE_OP_DONT_CARE,
				.initialLayout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL,
				.finalLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL,
			},
			{
				.format = format->vk,
				.samples = VK_SAMPLE_COUNT_1_BIT,
				.loadOp = VK_ATTACHMENT_LOAD_OP_LOAD,
				.storeOp = VK_ATTACHMENT_STORE_OP_STORE,
				.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE,
				.stencilStoreOp = VK_ATTACHMENT_STORE_OP_DONT_CARE,
				.initialLayout = VK_IMAGE_LAYOUT_GENERAL,
				.finalLayout = VK_IMAGE_LAYOUT_GENERAL,
			}
		};

		VkAttachmentReference blend_write_ref = {
			.attachment = 0u,
			.layout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL,
		};

		VkAttachmentReference blend_read_ref = {
			.attachment = 0u,
			.layout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL,
		};

		VkAttachmentReference color_ref = {
			.attachment = 1u,
			.layout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL,
		};

		VkSubpassDescription subpasses[2] = {
			{
				.pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS,
				.colorAttachmentCount = 1,
				.pColorAttachments = &blend_write_ref,
			},
			{
				.pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS,
				.inputAttachmentCount = 1,
				.pInputAttachments = &blend_read_ref,
				.colorAttachmentCount = 1,
				.pColorAttachments = &color_ref,
			}
		};

		VkSubpassDependency deps[3] = {
			{
				.srcSubpass = VK_SUBPASS_EXTERNAL,
				.srcStageMask = VK_PIPELINE_STAGE_HOST_BIT |
					VK_PIPELINE_STAGE_TRANSFER_BIT |
					VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT |
					VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT,
				.srcAccessMask = VK_ACCESS_HOST_WRITE_BIT |
					VK_ACCESS_TRANSFER_WRITE_BIT |
					VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT,
				.dstSubpass = 0,
				.dstStageMask = VK_PIPELINE_STAGE_ALL_GRAPHICS_BIT,
				.dstAccessMask = VK_ACCESS_UNIFORM_READ_BIT |
					VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT |
					VK_ACCESS_INDIRECT_COMMAND_READ_BIT |
					VK_ACCESS_SHADER_READ_BIT,
			},
			{
				.srcSubpass = 0,
				.srcStageMask = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT,
				.srcAccessMask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT,
				.dstSubpass = 1,
				.dstStageMask = VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT,
				.dstAccessMask = VK_ACCESS_SHADER_READ_BIT,
				.dependencyFlags = VK_DEPENDENCY_BY_REGION_BIT,
			},
			{
				.srcSubpass = 1,
				.srcStageMask = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT,
				.srcAccessMask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT,
				.dstSubpass = VK_SUBPASS_EXTERNAL,
				.dstStageMask = VK_PIPELINE_STAGE_TRANSFER_BIT |
					VK_PIPELINE_STAGE_HOST_BIT | VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT,
				.dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT |
					VK_ACCESS_MEMORY_READ_BIT,
			},
		};

		VkRenderPassCreateInfo rp_info = {
			.sType = VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO,
			.pNext = NULL,
			.flags = 0,
			.attachmentCount = 2u,
			.pAttachments = attachments,
			.subpassCount = 2u,
			.pSubpasses = subpasses,
			.dependencyCount = 3u,
			.pDependencies = deps,
		};

		res = vkCreateRenderPass(dev, &rp_info, NULL, &setup->render_pass);
		if (res != VK_SUCCESS) {
			wlr_vk_error("Failed to create 2-step render pass", res);
			goto error;
		}

		// this is only well defined if render pass has a 2nd subpass
		if (!init_blend_to_output_pipeline(
				renderer, setup->render_pass, renderer->output_pipe_layout,
				 &setup->output_pipe)) {
			goto error;
		}
	} else {
		VkAttachmentDescription attachment = {
			.format = format->vk,
			.samples = VK_SAMPLE_COUNT_1_BIT,
			.loadOp = VK_ATTACHMENT_LOAD_OP_LOAD,
			.storeOp = VK_ATTACHMENT_STORE_OP_STORE,
			.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE,
			.stencilStoreOp = VK_ATTACHMENT_STORE_OP_DONT_CARE,
			.initialLayout = VK_IMAGE_LAYOUT_GENERAL,
			.finalLayout = VK_IMAGE_LAYOUT_GENERAL,
		};

		VkAttachmentReference color_ref = {
			.attachment = 0u,
			.layout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL,
		};

		VkSubpassDescription subpass = {
			.pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS,
			.colorAttachmentCount = 1,
			.pColorAttachments = &color_ref,
		};

		VkSubpassDependency deps[2] = {
			{
				.srcSubpass = VK_SUBPASS_EXTERNAL,
				.srcStageMask = VK_PIPELINE_STAGE_HOST_BIT |
					VK_PIPELINE_STAGE_TRANSFER_BIT |
					VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT |
					VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT,
				.srcAccessMask = VK_ACCESS_HOST_WRITE_BIT |
					VK_ACCESS_TRANSFER_WRITE_BIT |
					VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT,
				.dstSubpass = 0,
				.dstStageMask = VK_PIPELINE_STAGE_ALL_GRAPHICS_BIT,
				.dstAccessMask = VK_ACCESS_UNIFORM_READ_BIT |
					VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT |
					VK_ACCESS_INDIRECT_COMMAND_READ_BIT |
					VK_ACCESS_SHADER_READ_BIT,
			},
			{
				.srcSubpass = 0,
				.srcStageMask = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT,
				.srcAccessMask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT,
				.dstSubpass = VK_SUBPASS_EXTERNAL,
				.dstStageMask = VK_PIPELINE_STAGE_TRANSFER_BIT |
					VK_PIPELINE_STAGE_HOST_BIT | VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT,
				.dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT |
					VK_ACCESS_MEMORY_READ_BIT,
			},
		};

		VkRenderPassCreateInfo rp_info = {
			.sType = VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO,
			.attachmentCount = 1,
			.pAttachments = &attachment,
			.subpassCount = 1,
			.pSubpasses = &subpass,
			.dependencyCount = 2u,
			.pDependencies = deps,
		};

		res = vkCreateRenderPass(dev, &rp_info, NULL, &setup->render_pass);
		if (res != VK_SUCCESS) {
			wlr_vk_error("Failed to create render pass", res);
			goto error;
		}
	}

	if (!setup_get_or_create_pipeline(setup, &(struct wlr_vk_pipeline_key){
		.source = WLR_VK_SHADER_SOURCE_SINGLE_COLOR,
		.layout = { .ycbcr_format = NULL },
	})) {
		goto error;
	}

	if (!setup_get_or_create_pipeline(setup, &(struct wlr_vk_pipeline_key){
		.source = WLR_VK_SHADER_SOURCE_TEXTURE,
		.texture_transform = WLR_VK_TEXTURE_TRANSFORM_IDENTITY,
		.layout = {.ycbcr_format = NULL },
	})) {
		goto error;
	}

	if (!setup_get_or_create_pipeline(setup, &(struct wlr_vk_pipeline_key){
		.source = WLR_VK_SHADER_SOURCE_TEXTURE,
		.texture_transform = WLR_VK_TEXTURE_TRANSFORM_SRGB,
		.layout = {.ycbcr_format = NULL },
	})) {
		goto error;
	}

	for (size_t i = 0; i < renderer->dev->format_prop_count; i++) {
		const struct wlr_vk_format *format = &renderer->dev->format_props[i].format;
		const struct wlr_vk_pipeline_layout_key layout = {
			.ycbcr_format = format,
		};

		if (format->is_ycbcr) {
			if (!setup_get_or_create_pipeline(setup, &(struct wlr_vk_pipeline_key){
				.texture_transform = WLR_VK_TEXTURE_TRANSFORM_SRGB,
				.layout = layout
			})) {
				goto error;
			}
		}
	}

	wl_list_insert(&renderer->render_format_setups, &setup->link);
	return setup;

error:
	destroy_render_format_setup(renderer, setup);
	return NULL;
}

struct wlr_renderer *vulkan_renderer_create_for_device(struct wlr_vk_device *dev) {
	struct wlr_vk_renderer *renderer;
	VkResult res;
	if (!(renderer = calloc(1, sizeof(*renderer)))) {
		wlr_log_errno(WLR_ERROR, "failed to allocate wlr_vk_renderer");
		return NULL;
	}

	renderer->dev = dev;
	wlr_renderer_init(&renderer->wlr_renderer, &renderer_impl);
	wl_list_init(&renderer->stage.buffers);
	wl_list_init(&renderer->foreign_textures);
	wl_list_init(&renderer->textures);
	wl_list_init(&renderer->descriptor_pools);
	wl_list_init(&renderer->output_descriptor_pools);
	wl_list_init(&renderer->render_format_setups);
	wl_list_init(&renderer->render_buffers);
	wl_list_init(&renderer->pipeline_layouts);

	if (!init_static_render_data(renderer)) {
		goto error;
	}

	VkCommandPoolCreateInfo cpool_info = {
		.sType = VK_STRUCTURE_TYPE_COMMAND_POOL_CREATE_INFO,
		.flags = VK_COMMAND_POOL_CREATE_RESET_COMMAND_BUFFER_BIT,
		.queueFamilyIndex = dev->queue_family,
	};
	res = vkCreateCommandPool(dev->dev, &cpool_info, NULL,
		&renderer->command_pool);
	if (res != VK_SUCCESS) {
		wlr_vk_error("vkCreateCommandPool", res);
		goto error;
	}

	VkSemaphoreTypeCreateInfoKHR semaphore_type_info = {
		.sType = VK_STRUCTURE_TYPE_SEMAPHORE_TYPE_CREATE_INFO_KHR,
		.semaphoreType = VK_SEMAPHORE_TYPE_TIMELINE_KHR,
		.initialValue = 0,
	};
	VkSemaphoreCreateInfo semaphore_info = {
		.sType = VK_STRUCTURE_TYPE_SEMAPHORE_CREATE_INFO,
		.pNext = &semaphore_type_info,
	};
	res = vkCreateSemaphore(dev->dev, &semaphore_info, NULL,
		&renderer->timeline_semaphore);
	if (res != VK_SUCCESS) {
		wlr_vk_error("vkCreateSemaphore", res);
		goto error;
	}

	return &renderer->wlr_renderer;

error:
	vulkan_destroy(&renderer->wlr_renderer);
	return NULL;
}

struct wlr_renderer *wlr_vk_renderer_create_with_drm_fd(int drm_fd) {
	wlr_log(WLR_INFO, "The vulkan renderer is only experimental and "
		"not expected to be ready for daily use");
	wlr_log(WLR_INFO, "Run with VK_INSTANCE_LAYERS=VK_LAYER_KHRONOS_validation "
		"to enable the validation layer");

	struct wlr_vk_instance *ini = vulkan_instance_create(default_debug);
	if (!ini) {
		wlr_log(WLR_ERROR, "creating vulkan instance for renderer failed");
		return NULL;
	}

	VkPhysicalDevice phdev = vulkan_find_drm_phdev(ini, drm_fd);
	if (!phdev) {
		// We rather fail here than doing some guesswork
		wlr_log(WLR_ERROR, "Could not match drm and vulkan device");
		return NULL;
	}

	uint32_t qfam_count;
	vkGetPhysicalDeviceQueueFamilyProperties(phdev, &qfam_count, NULL);
	VkQueueFamilyProperties queue_props[qfam_count];
	vkGetPhysicalDeviceQueueFamilyProperties(phdev, &qfam_count,
		queue_props);

	struct wlr_vk_device *dev = vulkan_device_create(ini, phdev);
	if (!dev) {
		wlr_log(WLR_ERROR, "Failed to create vulkan device");
		vulkan_instance_destroy(ini);
		return NULL;
	}

	// Do not use the drm_fd that was passed in: we should prefer the render
	// node even if a primary node was provided
	dev->drm_fd = vulkan_open_phdev_drm_fd(phdev);
	if (dev->drm_fd < 0) {
		vulkan_device_destroy(dev);
		vulkan_instance_destroy(ini);
		return NULL;
	}

	return vulkan_renderer_create_for_device(dev);
}

VkInstance wlr_vk_renderer_get_instance(struct wlr_renderer *renderer) {
	struct wlr_vk_renderer *vk_renderer = vulkan_get_renderer(renderer);
	return vk_renderer->dev->instance->instance;
}

VkPhysicalDevice wlr_vk_renderer_get_physical_device(struct wlr_renderer *renderer) {
	struct wlr_vk_renderer *vk_renderer = vulkan_get_renderer(renderer);
	return vk_renderer->dev->phdev;
}

VkDevice wlr_vk_renderer_get_device(struct wlr_renderer *renderer) {
	struct wlr_vk_renderer *vk_renderer = vulkan_get_renderer(renderer);
	return vk_renderer->dev->dev;
}

uint32_t wlr_vk_renderer_get_queue_family(struct wlr_renderer *renderer) {
	struct wlr_vk_renderer *vk_renderer = vulkan_get_renderer(renderer);
	return vk_renderer->dev->queue_family;
}

void wlr_vk_renderer_get_current_image_attribs(struct wlr_renderer *renderer,
		struct wlr_vk_image_attribs *attribs) {
	struct wlr_vk_renderer *vk_renderer = vulkan_get_renderer(renderer);
	attribs->image = vk_renderer->current_render_buffer->image;
	attribs->format = vk_renderer->current_render_buffer->render_setup->render_format->vk;
	attribs->layout = VK_IMAGE_LAYOUT_UNDEFINED;
}