/*
 * Copyright (c) 2026 The NetBSD Foundation, Inc.
 * All rights reserved.
 *
 * This code is derived from software contributed to The NetBSD Foundation
 * by Radoslaw Kujawa.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 *
 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
 * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
 * POSSIBILITY OF SUCH DAMAGE.
 */

/* libv3d - thin userland layer over /dev/verite3d. */
#include <sys/types.h>
#include <sys/ioctl.h>
#include <sys/mman.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <fcntl.h>
#include <time.h>
#include <unistd.h>

#include "libv3d.h"

uint64_t
v3d_now_ns(void)
{
	struct timespec ts;

	clock_gettime(CLOCK_MONOTONIC, &ts);
	return (uint64_t)ts.tv_sec * 1000000000ULL + (uint64_t)ts.tv_nsec;
}

/* 0 when timing is off: the accumulators just stay zero */
static uint64_t
v3d_t0(const struct v3d *v)
{

	return v->opt_stats ? v3d_now_ns() : 0;
}

int
v3d_open(struct v3d *v, const char *devpath, const char *ucpath)
{
	struct v3d_init vi;
	struct v3d_mode vm;
	static uint8_t uc[1024 * 1024];
	FILE *f;
	size_t sz;
	uint32_t w;
	int s;

	memset(v, 0, sizeof(*v));
	if (getenv("RLGL_STATS") != NULL)
		v->opt_stats = atoi(getenv("RLGL_STATS"));
	v->opt_nulldraw = getenv("RLGL_NULLDRAW") != NULL;
	v->opt_nosync = getenv("RLGL_NOSYNC") != NULL;
	v->opt_noflip = getenv("RLGL_NOFLIP") != NULL;
	v->opt_tearflip = getenv("RLGL_TEARFLIP") != NULL;
	if (v->opt_nulldraw || v->opt_nosync || v->opt_noflip ||
	    v->opt_tearflip)
		fprintf(stderr, "libv3d: EXPERIMENT MODE nulldraw=%d "
		    "nosync=%d noflip=%d tearflip=%d\n", v->opt_nulldraw,
		    v->opt_nosync, v->opt_noflip, v->opt_tearflip);
	v->fd = open(devpath, O_RDWR);
	if (v->fd == -1) {
		perror(devpath);
		return -1;
	}

	f = fopen(ucpath, "rb");
	if (f == NULL) {
		perror(ucpath);
		goto fail;
	}
	sz = fread(uc, 1, sizeof(uc), f);
	fclose(f);

	vi.vi_ucode = (uint64_t)(uintptr_t)uc;
	vi.vi_size = (uint32_t)sz;
	if (ioctl(v->fd, V3D_INIT, &vi) == -1) {
		perror("V3D_INIT");
		goto fail;
	}
	v->ctx_base = vi.vi_ctx_base;
	v->memsize = vi.vi_memsize;
	v->pool_base = vi.vi_pool_base;

	for (s = 0; s < V3D_RING_SLOTS; s++) {
		v->slot[s] = mmap(NULL, V3D_SLOT_SIZE,
		    PROT_READ | PROT_WRITE, MAP_SHARED, v->fd,
		    (off_t)s * V3D_SLOT_SIZE);
		if (v->slot[s] == MAP_FAILED) {
			perror("mmap slot");
			goto fail;
		}
	}

	/* GL context up; fence it */
	usleep(10000);
	v3d_emit1(v, V3D_GL_CTX_INIT);
	if (v3d_sync(v) != 0)
		goto fail;

	/* mode query fills geometry */
	memset(&vm, 0, sizeof(vm));
	if (ioctl(v->fd, V3D_MODE, &vm) == -1)
		goto fail;
	v->width = vm.vm_width;
	v->height = vm.vm_height;
	v->stride = vm.vm_stride;
	v->pe_stride = vm.vm_pe_stride;
	(void)w;
	return 0;
fail:
	close(v->fd);
	v->fd = -1;
	return -1;
}

void
v3d_close(struct v3d *v)
{

	if (v->fd == -1)
		return;
	/* park politely */
	v3d_emit1(v, V3D_GL_EXIT);
	(void)v3d_flush(v);
	if (v->vram != NULL) {
		munmap(v->vram, v->memsize);
		v->vram = NULL;
	}
	close(v->fd);
	v->fd = -1;
}

const uint8_t *
v3d_vram(struct v3d *v)
{

	if (v->vram == NULL) {
		/* RW for span writes (Mesa swrast), RO fallback */
		v->vram = mmap(NULL, v->memsize, PROT_READ | PROT_WRITE,
		    MAP_SHARED, v->fd, V3D_VRAM_MMAP_OFF);
		if (v->vram == MAP_FAILED)
			v->vram = mmap(NULL, v->memsize, PROT_READ,
			    MAP_SHARED, v->fd, V3D_VRAM_MMAP_OFF);
		if (v->vram == MAP_FAILED) {
			perror("mmap vram");
			v->vram = NULL;
		}
	}
	return v->vram;
}

void
v3d_emit(struct v3d *v, const uint32_t *w, uint32_t n)
{

	if (v->n + n > V3D_SLOT_SIZE / 4) {
		if (v3d_flush(v) != 0)
			return;
	}
	memcpy(v->slot[v->cur] + v->n, w, n * 4);
	v->n += n;
	v->st.words += n;
}

void
v3d_emit1(struct v3d *v, uint32_t w)
{

	v3d_emit(v, &w, 1);
}

uint32_t *
v3d_reserve(struct v3d *v, uint32_t nwords)
{

	if (nwords > V3D_SLOT_SIZE / 4)
		return NULL;
	if (v->n + nwords > V3D_SLOT_SIZE / 4) {
		if (v3d_flush(v) != 0)
			return NULL;
	}
	return v->slot[v->cur] + v->n;
}

void
v3d_commit(struct v3d *v, uint32_t nwords)
{

	v->n += nwords;
	v->st.words += nwords;
}

void
v3d_emitf(struct v3d *v, float f)
{
	union { float f; uint32_t u; } u;

	u.f = f;
	v3d_emit1(v, u.u);
}

int
v3d_flush(struct v3d *v)
{
	struct v3d_submit vs;
	uint64_t t0;

	if (v->n == 0)
		return 0;
	vs.vs_slot = (uint32_t)v->cur;
	vs.vs_len = v->n * 4;
	vs.vs_swap = 1;			/* XXX: host-native words */
	v->n = 0;
	v->cur = (v->cur + 1) % V3D_RING_SLOTS;
	v->st.submits++;
	v->st.submit_bytes += vs.vs_len;
	if (v->opt_nulldraw)
		return 0;
	t0 = v3d_t0(v);
	if (ioctl(v->fd, V3D_SUBMIT, &vs) == -1) {
		perror("V3D_SUBMIT");
		return -1;
	}
	if (t0 != 0)
		v->st.submit_ns += v3d_now_ns() - t0;
	return 0;
}

int
v3d_sync(struct v3d *v)
{
	uint32_t word = V3D_GL_SYNC;
	uint64_t t0;

	if (v3d_flush(v) != 0)
		return -1;
	v->st.syncs++;
	if (v->opt_nulldraw)
		return 0;
	t0 = v3d_t0(v);
	if (ioctl(v->fd, V3D_SYNC, &word) == -1) {
		perror("V3D_SYNC");
		return -1;
	}
	if (t0 != 0)
		v->st.sync_ns += v3d_now_ns() - t0;
	return 0;
}

int
v3d_alloc(struct v3d *v, uint32_t size, uint32_t align, uint32_t *addr)
{
	struct v3d_alloc va;

	va.va_size = size;
	va.va_align = align;
	if (ioctl(v->fd, V3D_ALLOC, &va) == -1)
		return -1;
	*addr = va.va_addr;
	return 0;
}

int
v3d_mode(struct v3d *v, uint32_t depth, uint32_t frame_base)
{
	struct v3d_mode vm;

	memset(&vm, 0, sizeof(vm));
	vm.vm_depth = depth;
	vm.vm_frame_base = frame_base;
	if (ioctl(v->fd, V3D_MODE, &vm) == -1)
		return -1;
	v->width = vm.vm_width;
	v->height = vm.vm_height;
	v->stride = vm.vm_stride;
	v->pe_stride = vm.vm_pe_stride;
	return 0;
}

int
v3d_flip(struct v3d *v, uint32_t frame_base)
{
	uint64_t t0;
	int r;

	if (v3d_flush(v) != 0)
		return -1;
	v->st.flips++;
	if (v->opt_nulldraw || v->opt_noflip)
		return 0;
	if (v->opt_tearflip)
		frame_base |= V3D_FLIP_NOWAIT;
	t0 = v3d_t0(v);
	r = ioctl(v->fd, V3D_FLIP, &frame_base);
	if (t0 != 0)
		v->st.flip_ns += v3d_now_ns() - t0;
	return r;
}

void
v3d_st(struct v3d *v, uint32_t reg, uint32_t val)
{
	uint32_t w[2];

	w[0] = (1 << 12) | reg;
	w[1] = val;
	v->st.st_calls++;
	v3d_emit(v, w, 2);
}

void
v3d_texbind(struct v3d *v, uint32_t addr, uint32_t srcstride,
    uint32_t w, uint32_t h, uint32_t uscale, uint32_t vscale)
{
	uint32_t c[6];

	c[0] = (4 << 12) | V3D_PE_SRCBASE;
	c[1] = addr;
	c[2] = srcstride;
	c[3] = ((h - 1) << 16) | (w - 1);
	c[4] = uscale;
	c[5] = vscale;
	v->st.texbinds++;
	v3d_emit(v, c, 6);
}

/* below this many payload bytes of slot room, flush for a fresh slot */
#define V3D_UPLOAD_MIN	4096

int
v3d_upload(struct v3d *v, uint32_t dst, const void *data, uint32_t bytes)
{
	const uint32_t *p = data;
	uint32_t chunk, room, hdr[3];

	v->st.uploads++;
	v->st.upload_bytes += bytes;
	while (bytes > 0) {
		/*
		 * Batch into the current slot alongside whatever is
		 * already queued
		 */
		room = (V3D_SLOT_SIZE / 4 - v->n) * 4;
		if (room < V3D_UPLOAD_MIN + 3 * 4) {
			if (v3d_flush(v) != 0)
				return -1;
			room = V3D_SLOT_SIZE;
		}
		chunk = bytes;
		if (chunk > room - 3 * 4)
			chunk = room - 3 * 4;
		hdr[0] = V3D_GL_MEM_WRITE;
		hdr[1] = dst;
		hdr[2] = chunk;
		v3d_emit(v, hdr, 3);
		v3d_emit(v, p, chunk / 4);
		p += chunk / 4;
		dst += chunk;
		bytes -= chunk;
	}
	return 0;
}

void
v3d_clear(struct v3d *v, uint32_t addr, uint32_t stride, uint32_t wbytes,
    uint32_t h, uint32_t val)
{
	uint32_t c[6];

	c[0] = V3D_GL_CLEAR_RECT;
	c[1] = addr;
	c[2] = stride;
	c[3] = wbytes;
	c[4] = h;
	c[5] = val;
	v3d_emit(v, c, 6);
}