Video Overview

Two Output Paths

Session and display paths. The automatic desktop requires the graphics kernel and an RGB framebuffer; a text-session desktop is started explicitly.

The kernel supports two independent video paths selected at boot based on what the bootloader provides:

Path	Source	Mode	Driver
VESA framebuffer	Multiboot2 framebuffer tag	Linear pixel buffer	video/mode.rs
VGA text mode	Legacy BIOS/VGA	80×25 character cells	video/vga.rs
Framebuffer text console	EFI text entries (console=fb)	80×25 cells drawn as pixels	video/console.rs, video/vga.rs

The kernel console writes boot status to VGA text memory at 0xB8000. When GRUB selects an RGB framebuffer that text memory is not visible; Memento provides the graphics session instead of the kernel shell and clock. See Graphics Session.

EFI text entries select the framebuffer console, which keeps the shell visible on systems without a VGA text display. See Text Mode Writer.

---

Boot-Time Detection

init::check::init() parses the Multiboot2 information structure passed by the bootloader. When a framebuffer tag (type 8) is found, the kernel fills FRAMEBUFFER_PTR:

pub static mut FRAMEBUFFER_PTR: FramebufferTag = FramebufferTag {
    addr:   0,
    pitch:  0,
    width:  0,
    height: 0,
    bpp:    0,
    fb_type: 0,
};

FramebufferTag mirrors the Multiboot2 tag layout:

Field	Type	Description
addr	u64	Physical address of the framebuffer
pitch	u32	Bytes per scan line
width	u32	Width in pixels
height	u32	Height in pixels
bpp	u8	Bits per pixel (16 or 32)
fb_type	u8	1 = RGB, 2 = EGA text

After parsing, VIDEO_MODE is set once and stays fixed for the lifetime of the kernel. For an RGB framebuffer, video::wc::map_write_combining changes PAT entry 1 to write-combining and selects it on the framebuffer's 2 MiB pages. This happens before process page tables are created. See Boot and Initialization.

---

VideoMode Enum (video/mode.rs)

pub enum VideoMode {
    Framebuffer {
        address: u64,
        pitch:   u32,
        width:   u32,
        height:  u32,
        bpp:     u8,
    },
    TextMode,
}

pub static mut VIDEO_MODE: Option<VideoMode> = None;

VIDEO_MODE is read by pixel-writing helpers and video syscalls to dispatch to the correct path.

---

Pixel Writing

put_pixel(x: u32, y: u32, r: u8, g: u8, b: u8) in video/mode.rs writes a single pixel into the VESA framebuffer. The byte offset is:

offset = y * pitch + x * (bpp / 8)

Pixel encoding by bpp:

bpp	Format	Encoding
32	0x00RRGGBB	(r as u32) << 16 \| (g as u32) << 8 \| b as u32
16	RGB565	((r >> 3) << 11) \| ((g >> 2) << 5) \| (b >> 3)

Other bit depths are ignored (no-op).

---

Text-Mode Character Writing

put_char(x: u32, y: u32, ch: u8, color: u8) writes a character cell to the VGA text buffer at 0xB8000. Each cell is 2 bytes: character byte then attribute byte (color). The cell offset is:

offset = (y * 80 + x) * 2

This function is independent of the Writer struct — it is a direct write used by some low-level paths.

---

Boot Status Display (video/sysprint.rs)

sysprint.rs provides a structured boot-status output layer that sits on top of the VGA text Writer. It maintains SYSBUFFER: Mutex<Buffer> — a 24-line display buffer (kept separate from the scrolling writer so status lines stay visible).

pub enum Result {
    Unknown,
    Passed,
    Failed,
    Skipped,
}

Each subsystem check (heap, processes, network, etc.) prints a line using print_result(name, result). Output format:

[  OK  ]  kernel heap
[ FAIL ]  some subsystem
[ SKIP ]  optional feature
[UNKNWN]  unverified

Colors are applied per Result variant using VGA attribute bytes.

---

Video Syscalls

Syscall	Number	Description
put_pixel	0x12	Write single pixel at (x, y) with RGB color
blit_mode13	0x13	Blit a 320×200 byte array into VGA mode 13h VRAM
map_vram	0x14	Map VGA VRAM (0xA00_000) into the calling process's page table
set_video_mode	0x15	Switch VGA hardware mode (0x03 / 0x0D / 0x12 / 0x13)
get_fb_info	0x16	Copy FRAMEBUFFER_PTR fields into userland-supplied buffer
blit_buffer	0x17	Copy userland pixel buffer into VESA framebuffer
get_kernel_font	0x18	Copy the embedded 8×16 CP850 font bitmap into userland buffer
blit_indexed	0x19	Scale palette-indexed frames, update row bands and publish completed presentations
capture	0x1c	Copy framebuffer rows to a packed 32-bit pixel buffer
capture_rgb24	0x1d	Capture and scale to tightly packed RGB24, using a completed snapshot at 640×480 when available

See Video and Audio syscalls for request layouts, presentation transactions and capture retry rules.

Pointer Constraints

Buffers passed to video syscalls must lie wholly inside the program image and initial stacks (0x400_000 – 0xA00_000) or wholly inside the userland heap (0xC00_000 – 0xFFF_FFF) or its allocated extension. See Pointer Arguments. The whole buffer is checked: 64000 bytes for 0x13, width × height × 4 for 0x17.

The exception is map_vram (0x14), which maps 0xA00_000 into the calling process's page table using a P1 (4 KiB) sub-table allocated from PAGE_TABLE_POOL. After map_vram, the process can write to VGA VRAM directly at that virtual address without going through a syscall.

---

Kernel Font (VGA_FONT_8X16)

A complete 8×16 pixel CP850 font (256 glyphs × 16 rows × 1 byte per row = 4096 bytes) is embedded as a static array in video/vga_hw.rs. It is used:

1. By restore_font() when switching back to VGA text mode (mode 0x03) — loaded into VGA plane 2 via planar write mode.
2. Via syscall 0x18 — returned to userland for software font rendering in pixel modes.

---

Related documentation: Kernel Internals, VGA, Video and Audio ABI, Memento.