#include "types.h"
#include "x86.h"
#include "memlayout.h"
#include "defs.h"
#include "theme.h"
#include "font.h"
#include "logo.h"
#include "device.h"
#include "display.h"
#include "dma.h"

#define BACKSPACE 0x100
#define CHAR_HEIGHT 16
#define CHAR_WIDTH  8
#define COLOR_BG SCH_BG

volatile uchar *vga_mem = 0;
// Physical base of the linear framebuffer, as reported by VBE. Every other
// part of the kernel that needs to map the framebuffer -- fork's VMA copy,
// mmap of /dev/fb0 -- asks for this rather than assuming a value. Three
// places used to hardcode 0xfd000000, which is only where QEMU's stdvga
// happens to put it.
static uint64 vga_phys = 0;
static uint32 vga_stride = 0;
static uint32 vga_bpp = 0;
static uint32 vga_width = 800;
static uint32 vga_height = 600;
static int cursor_x = 0;
static int cursor_y = 0;
// Current fill color for glyph backgrounds, backspace erase, and scroll —
// normally COLOR_BG, but panic() switches this to paint the whole console
// a solid color (e.g. PSOGM's purple) instead of leaving cream-colored boxes
// around each character.
static uint32 vga_bg = COLOR_BG;

void
vga_set_bg(uint32 color)
{
  vga_bg = color;
}

// Written by boot/bios/stage2.S's VBE call and read by boot/bios/stage2main.c under
// the same name (VBE_INFO_ADDR) — not 0x9000, since stage2's own code now
// runs up to nearly 0x9000 once the boot-menu/progress-bar logic is
// included, which would otherwise clobber a mode info block placed there.
#define VBE_INFO_ADDR 0x30000

void
vga_init(uint64 phys_addr) {
  if (phys_addr == 0) {
    phys_addr = *(uint32*)(P2V(VBE_INFO_ADDR) + 40);
  }
  vga_mem = (volatile uchar*)P2V(phys_addr);
  vga_phys = phys_addr;

  vga_stride = *(ushort*)(P2V(VBE_INFO_ADDR) + 16);
  vga_width  = *(ushort*)(P2V(VBE_INFO_ADDR) + 18);
  vga_height = *(ushort*)(P2V(VBE_INFO_ADDR) + 20);
  vga_bpp    = *(uchar*)(P2V(VBE_INFO_ADDR) + 25);

  if (vga_width == 0) vga_width = 800;
  if (vga_height == 0) vga_height = 600;
  if (vga_stride == 0) vga_stride = vga_width * (vga_bpp / 8);

  cprintf("VGA: %dx%d, %d bpp, stride %d, LFB: 0x%x\n", vga_width, vga_height, vga_bpp, vga_stride, (uint32)phys_addr);

  // Deliberately does NOT clear the screen.
  //
  // The bootloader has already drawn the splash, and wiping it here is what
  // produced the old two-act boot: a purple splash for an instant, then an
  // abrupt cream screen filling with kernel log lines. The splash now stays up
  // for the whole of boot, and console_takeover() clears it exactly once --
  // when userspace starts and the display changes owner.
}

// The mode that was actually set. Exported because the resolution is chosen at
// boot now (press R -- see boot/bios/stage2.S), so nothing above may assume it.
uint32
vga_get_width(void)
{
  return vga_width;
}

uint32
vga_get_height(void)
{
  return vga_height;
}

void
vga_draw_pixel(int x, int y, uint32 color)
{
  if (!vga_mem || x < 0 || x >= vga_width || y < 0 || y >= vga_height)
    return;

  uint32 offset = (y * vga_stride) + (x * (vga_bpp / 8));

  if (vga_bpp == 32) {
    *(volatile uint32*)(vga_mem + offset) = color;
  } else if (vga_bpp == 24) {
    vga_mem[offset]     = color & 0xFF;
    vga_mem[offset + 1] = (color >> 8) & 0xFF;
    vga_mem[offset + 2] = (color >> 16) & 0xFF;
  } else if (vga_bpp == 16) {
    uint16 c16 = (((color >> 19) & 0x1F) << 11) | (((color >> 10) & 0x3F) << 5) | ((color >> 3) & 0x1F);
    *(volatile uint16*)(vga_mem + offset) = c16;
  }
}

void
vga_clear(uint32 color)
{
  for (int y = 0; y < vga_height; y++) {
    for (int x = 0; x < vga_width; x++) {
      vga_draw_pixel(x, y, color);
    }
  }
}

void
vga_draw_char(int x, int y, uchar c, uint32 color)
{
  // No `c > 127` rejection any more: font_data holds 256 Latin-1 glyphs, and
  // a uchar cannot index past them.
  if (!vga_mem) return;

  for (int row = 0; row < CHAR_HEIGHT; row++) {
    uchar row_data = font_data[c][row];
    for (int col = 0; col < CHAR_WIDTH; col++) {
      if (row_data & (0x80 >> col)) {
        vga_draw_pixel(x + col, y + row, color);
      } else {
        vga_draw_pixel(x + col, y + row, vga_bg);
      }
    }
  }
}

void
vga_puts(int x, int y, char *s, uint32 color)
{
  while (*s) {
    vga_putc((uchar)*s, color);
    s++;
  }
}

void
vga_putc(int c, uint32 color)
{
  if (!vga_mem) return;

  if (c == BACKSPACE || c == 0x08 || c == '\b' || c == 127) {
    if (cursor_x >= CHAR_WIDTH) {
      cursor_x -= CHAR_WIDTH;
      for (int i = 0; i < CHAR_HEIGHT; i++) {
        for (int j = 0; j < CHAR_WIDTH; j++) {
          vga_draw_pixel(cursor_x + j, cursor_y + i, vga_bg);
        }
      }
    }
    return;
  }

  if (c == '\n') {
    cursor_x = 0;
    cursor_y += CHAR_HEIGHT;
  } else if (c == '\r') {
    cursor_x = 0;
  } else if (c == '\t') {
    // Advance to the next multiple of eight columns, erasing as it goes so a
    // tab overwrites rather than skipping over stale characters.
    //
    // There was no tab case at all, so '\t' fell through to vga_draw_char and
    // was drawn as glyph 9 -- which in CP437 is a hollow circle. Pressing Tab
    // printed a circle. Same for every other control character: the font has
    // a picture for each of them, and none of those pictures is what the
    // character means.
    int col = cursor_x / CHAR_WIDTH;
    int stop = (col + 8) & ~7;
    while (col < stop && cursor_x < (int)vga_width) {
      for (int i = 0; i < CHAR_HEIGHT; i++)
        for (int j = 0; j < CHAR_WIDTH; j++)
          vga_draw_pixel(cursor_x + j, cursor_y + i, vga_bg);
      cursor_x += CHAR_WIDTH;
      col++;
    }
  } else if (c < 32) {
    return;                 // any other control character: not a glyph
  } else {
    vga_draw_char(cursor_x, cursor_y, (uchar)c, color);
    cursor_x += CHAR_WIDTH;
  }

  if (cursor_x >= vga_width) {
    cursor_x = 0;
    cursor_y += CHAR_HEIGHT;
  }

  if (cursor_y > (vga_height - CHAR_HEIGHT)) {
    uint32 line_size = vga_stride * CHAR_HEIGHT;
    uint32 total_size = vga_stride * vga_height;
    uint32 copy_size = total_size - line_size;

    memmove((void*)vga_mem, (void*)(vga_mem + line_size), copy_size);

    for (int y = vga_height - CHAR_HEIGHT; y < vga_height; y++) {
      for (int x = 0; x < vga_width; x++) {
        vga_draw_pixel(x, y, vga_bg);
      }
    }
    cursor_y = vga_height - CHAR_HEIGHT;
  }
}

void
vga_screen_reset(void)
{
  vga_bg = COLOR_BG;
  vga_clear(COLOR_BG);
  cursor_x = 0;
  cursor_y = 0;
}

/* ------------------------------------------------------------------ */
/* Drawing primitives for the crash screen                             */
/*                                                                     */
/* The crash screen lays itself out rather than printing, which needs  */
/* two things the console never did: filled shapes, and text at an     */
/* absolute position that leaves the cursor alone.                     */
/* ------------------------------------------------------------------ */

int
vga_ready(void)
{
  return vga_mem != 0;
}

void
vga_dims(int *w, int *h)
{
  if (w) *w = (int)vga_width;
  if (h) *h = (int)vga_height;
}

int vga_char_w(void) { return CHAR_WIDTH; }
int vga_char_h(void) { return CHAR_HEIGHT; }

/* ------------------------------------------------------------------ */
/* Framebuffer geometry                                                */
/*                                                                     */
/* vga.c is the only place that reads the VBE mode info, so it is the  */
/* only place that knows where the framebuffer actually is. Everything */
/* else asks. See kernel/dev/fb.c for the device these back.           */
/* ------------------------------------------------------------------ */

uint64 vga_fb_phys(void) { return vga_phys; }

// Bytes of framebuffer, measured in scanlines rather than as
// width * height * bpp/8 -- the two differ whenever the card pads its rows,
// and it is the stride that says how far apart two rows really are.
uint64 vga_fb_size(void)
{
  if (!vga_mem)
    return 0;
  return (uint64)vga_stride * (uint64)vga_height;
}

uint32 vga_fb_stride(void) { return vga_stride; }
uint32 vga_fb_bpp(void)    { return vga_bpp; }

volatile uchar *vga_fb_ptr(void) { return vga_mem; }

/* ------------------------------------------------------------------ */
/* Cursor addressing and erasing                                       */
/*                                                                     */
/* What an ANSI escape parser needs and vga_putc alone cannot give it: */
/* the ability to put the cursor somewhere specific and to erase a     */
/* region without writing spaces over it. See kernel/core/console.c.   */
/* ------------------------------------------------------------------ */

int vga_cols(void) { return vga_width  ? (int)(vga_width  / CHAR_WIDTH)  : 0; }
int vga_rows(void) { return vga_height ? (int)(vga_height / CHAR_HEIGHT) : 0; }

void
vga_cursor_get(int *col, int *row)
{
  if (col) *col = cursor_x / CHAR_WIDTH;
  if (row) *row = cursor_y / CHAR_HEIGHT;
}

// Clamped rather than validated: an escape sequence is untrusted input that
// arrives from any program that can write to the terminal, and "CUP to row
// 9999" should put the cursor on the last row, not off the end of the
// framebuffer.
void
vga_cursor_set(int col, int row)
{
  int maxc = vga_cols(), maxr = vga_rows();

  if (maxc <= 0 || maxr <= 0)
    return;
  if (col < 0) col = 0;
  if (row < 0) row = 0;
  if (col >= maxc) col = maxc - 1;
  if (row >= maxr) row = maxr - 1;

  cursor_x = col * CHAR_WIDTH;
  cursor_y = row * CHAR_HEIGHT;
}

// Fill a run of character cells with the background colour.
static void
erase_cells(int col, int row, int count)
{
  int maxc = vga_cols();

  while (count-- > 0) {
    int x = col * CHAR_WIDTH, y = row * CHAR_HEIGHT;
    for (int i = 0; i < CHAR_HEIGHT; i++)
      for (int j = 0; j < CHAR_WIDTH; j++)
        vga_draw_pixel(x + j, y + i, vga_bg);
    if (++col >= maxc) { col = 0; row++; }
  }
}

// ED -- erase in display. 0: cursor to end, 1: start to cursor, 2: all.
void
vga_erase_display(int mode)
{
  int col, row, maxc = vga_cols(), maxr = vga_rows();

  vga_cursor_get(&col, &row);
  switch (mode) {
  case 0: erase_cells(col, row, (maxr - row) * maxc - col); break;
  case 1: erase_cells(0, 0, row * maxc + col + 1);          break;
  default: vga_clear(vga_bg);                               break;
  }
}

// EL -- erase in line. Same three modes, bounded to the current row.
void
vga_erase_line(int mode)
{
  int col, row, maxc = vga_cols();

  vga_cursor_get(&col, &row);
  switch (mode) {
  case 0: erase_cells(col, row, maxc - col); break;
  case 1: erase_cells(0, row, col + 1);      break;
  default: erase_cells(0, row, maxc);        break;
  }
}

// Row pointer computed once per row rather than per pixel. The crash screen
// paints a solid card behind the QR symbol, and at 8 pixels per module that is
// tens of thousands of writes going through vga_draw_pixel's bounds check and
// multiply.
void
vga_fill_rect(int x0, int y0, int w, int h, uint32 color)
{
  if (!vga_mem)
    return;
  if (x0 < 0) { w += x0; x0 = 0; }
  if (y0 < 0) { h += y0; y0 = 0; }
  if (x0 + w > (int)vga_width)  w = (int)vga_width  - x0;
  if (y0 + h > (int)vga_height) h = (int)vga_height - y0;
  if (w <= 0 || h <= 0)
    return;

  // Only the 32bpp case gets the fast path; anything else is rare enough that
  // correctness through vga_draw_pixel is the better trade.
  if (vga_bpp != 32) {
    for (int y = y0; y < y0 + h; y++)
      for (int x = x0; x < x0 + w; x++)
        vga_draw_pixel(x, y, color);
    return;
  }
  for (int y = y0; y < y0 + h; y++) {
    volatile uint32 *row = (volatile uint32 *)(vga_mem + y * vga_stride) + x0;
    for (int x = 0; x < w; x++)
      row[x] = color;
  }
}

void
vga_fill_circle(int cx, int cy, int r, uint32 color)
{
  for (int dy = -r; dy <= r; dy++)
    for (int dx = -r; dx <= r; dx++)
      if (dx * dx + dy * dy <= r * r)
        vga_draw_pixel(cx + dx, cy + dy, color);
}

// Text at an absolute position, leaving the console cursor alone. The crash
// screen lays itself out, so vga_putc's cursor tracking and scrolling are
// exactly what it does not want.
void
vga_text_at(int x, int y, const char *s, uint32 color)
{
  for (int i = 0; s[i]; i++)
    vga_draw_char(x + i * CHAR_WIDTH, y, (uchar)s[i], color);
}

// --- Logo ------------------------------------------------------------------
//
// The boot splash's logo is drawn by the *loader*, not here -- see splash() in
// boot/bios/stage2main.c. It has to be, because the kernel does not reach vga_init()
// until roughly the last 300ms of a 2.8s boot: drawing it here put the logo on
// screen for one frame before console_takeover() wiped it, which is worse than
// not having one.
//
// What lives here is the same drawing code against the same mask, for the
// things that want the mark *after* boot -- the crash screen, and eventually
// whatever schinit shows. Keeping one copy of the compositing logic is why this
// is not simply deleted.

// Blend `fg` over `bg` by coverage a (0..255). Per-channel, integer only: this
// runs before any floating point is safe to touch.
static uint32
blend(uint32 bg, uint32 fg, uint a)
{
  uint32 r = (((fg >> 16) & 0xFF) * a + ((bg >> 16) & 0xFF) * (255 - a)) / 255;
  uint32 g = (((fg >>  8) & 0xFF) * a + ((bg >>  8) & 0xFF) * (255 - a)) / 255;
  uint32 b = (((fg      ) & 0xFF) * a + ((bg      ) & 0xFF) * (255 - a)) / 255;
  return (r << 16) | (g << 8) | b;
}

void
vga_draw_logo(int x, int y, uint32 color, uint32 bg)
{
  if (!vga_mem)
    return;
  for (int dy = 0; dy < LOGO_H; dy++) {
    for (int dx = 0; dx < LOGO_W; dx++) {
      uint a = LOGO_INK(dy * LOGO_W + dx);
      // Skip the background rather than painting it. The mask is a square but
      // the mark is a circle, so most of it is empty -- and drawing that empty
      // area in bg would stamp a visible box over whatever is underneath.
      if (a == 0)
        continue;
      vga_draw_pixel(x + dx, y + dy, a == 255 ? color : blend(bg, color, a));
    }
  }
}

// Centred, for callers that own the whole screen.
void
vga_logo_centered(int cy, uint32 color, uint32 bg)
{
  if (!vga_mem || vga_width < LOGO_W)
    return;
  vga_draw_logo(((int)vga_width - LOGO_W) / 2, cy - LOGO_H / 2, color, bg);
}

// --- display class implementation ------------------------------------------
//
// See include/display.h. This is the VBE linear framebuffer presenting itself
// as a generic display, so a compositor can be written against the class and
// keep working when virtio-gpu arrives.

static uchar *back;          // shadow buffer in RAM, or 0 if never acquired
static uint64 back_bytes;

static void
vga_disp_mode(struct device *d, struct display_mode *m)
{
  m->width  = (int)vga_width;
  m->height = (int)vga_height;
  m->bpp    = (int)vga_bpp;
  m->stride = vga_stride;
}

static void *
vga_disp_acquire(struct device *d, uint32 *stride_out)
{
  if (stride_out)
    *stride_out = vga_stride;
  if (back)
    return back;
  if (!vga_mem || vga_height == 0 || vga_stride == 0)
    return 0;

  back_bytes = (uint64)vga_stride * vga_height;

  // From the DMA arena, which is the wrong pool and is used here because it is
  // the only *physically contiguous* allocator in the kernel -- kalloc hands
  // out single pages whose neighbours are adjacent only by chance, and a
  // framebuffer wants one flat span.
  //
  // At 800x600x32 this is 1.9 MB of a 4 MB arena meant for device rings, which
  // is tolerable only because nothing else large lives there yet. The real fix
  // is a contiguous page allocator, or reserving this at boot from the memory
  // map -- which is work that belongs with the E820 change in docs/roadmap.md
  // §3b, since that is what stops the kernel pretending 224 MB exists.
  //
  // Allocated on first use rather than at boot on purpose: a machine that never
  // starts a compositor should not pay two megabytes for one.
  back = dma_alloc(back_bytes, 4096);
  if (!back) {
    cprintf("display: no room for a %d KB back buffer; drawing will tear\n",
            (int)(back_bytes / 1024));
    return 0;
  }
  memset(back, 0, back_bytes);
  cprintf("display: %d KB back buffer, double buffered\n",
          (int)(back_bytes / 1024));
  return back;
}

// Blit a rectangle from the shadow buffer to the framebuffer.
//
// Row at a time rather than one memmove for the whole rectangle: only a
// full-width rectangle is contiguous in either buffer, and a compositor
// updating one window is the case worth being right about.
static void
vga_disp_present(struct device *d, int x, int y, int w, int h)
{
  if (!back || !vga_mem)
    return;

  if (x < 0) { w += x; x = 0; }
  if (y < 0) { h += y; y = 0; }
  if (x + w > (int)vga_width)  w = (int)vga_width - x;
  if (y + h > (int)vga_height) h = (int)vga_height - y;
  if (w <= 0 || h <= 0)
    return;

  int bytes = w * (int)(vga_bpp / 8);
  uint32 off = (uint32)x * (vga_bpp / 8);
  for (int row = y; row < y + h; row++)
    memmove((void *)(vga_mem + row * vga_stride + off),
            back + row * vga_stride + off, bytes);
}

// Wait for vertical retrace, by polling the VGA input status register.
//
// Bit 3 of 0x3DA is set during vertical retrace. This is the original 1987 VGA
// interface and it still works on everything that emulates a VGA, QEMU's
// stdvga included -- VBE's own protected-mode interface for this is optional
// and frequently absent, so the old port is the portable answer.
//
// Two loops, not one: waiting only for the flag to *be* set would return
// immediately if a retrace is already in progress, having just missed the
// start of it. Waiting for the current one to end and the next to begin is
// what actually lands the blit between frames.
//
// The counters are a deadlock guard. On hardware that never reports retrace
// this returns after a bounded spin instead of hanging the caller forever,
// which matters because this can be called with the console locked.
static void
vga_disp_wait_vblank(struct device *d)
{
  int guard;

  for (guard = 0; guard < 1000000; guard++)
    if (!(inb(0x3DA) & 0x08))
      break;
  for (guard = 0; guard < 1000000; guard++)
    if (inb(0x3DA) & 0x08)
      break;
}

const struct display_ops vga_display_ops = {
  .mode        = vga_disp_mode,
  .acquire     = vga_disp_acquire,
  .present     = vga_disp_present,
  .wait_vblank = vga_disp_wait_vblank,
};