Move to git

.gitignore

@@ -0,0 +1,8 @@
+bx_enh_dbg.ini
+*.bin
+*.o
+*.elf
+*.mod
+bootfs/*
+partuuid.txt
+*.a

Makefile

@@ -0,0 +1,46 @@
+rwildcard=$(foreach d,$(wildcard $(1:=/*)),$(call rwildcard,$d,$2) $(filter $(subst *,%,$2),$d))
+
+HEADERS := $(call rwildcard,src/kernel,*.h)
+SOURCEC := $(call rwildcard,src/kernel,*.c)
+SOURCEA := $(call rwildcard,src/kernel,*.asm)
+OBJECTS := $(patsubst src/%.c, build/%.o, $(SOURCEC)) $(patsubst src/%.asm, build/%.o, $(SOURCEA))
+
+drive.bin: Makefile partuuid.txt mod.ld elftomod.py bootfs/luma_vid.mod bootfs/luma_hda.mod bootfs/luma_fs.mod bootfs/luma_wm.mod bootfs/luma_uhci.mod bootfs/luma_ps2.mod boot0.bin boot1.bin $(OBJECTS)
+	i386-elf-ld -pie -T mod.ld -no-dynamic-linker -nostdlib -o kernel.elf $(OBJECTS) $(LDFLAGS)
+	./elftomod.py kernel.elf bootfs/krnl.mod ppm_Bitmap:0 vpm_init:1 vpm_map:2 canal_init:3 pci_init:4 scheduler_init:5 scheduler_spawn:6 scheduler_start:7
+
+	dd if=/dev/zero of=drive.bin bs=1024 count=10240
+	/bin/echo -e "mklabel msdos\nmkpart primary 8192b -1s\nquit\n" | parted drive.bin
+	mkdir -p mnt
+	/bin/echo -e "512\n20464\n0\n0\n0\n0\n" | leanfuse -O8192 -I`cat partuuid.txt` -Cdrive.bin
+	./leanfuse mnt drive.bin -O8192
+	cp -r bootfs/* ./mnt/
+	fusermount -u ./mnt
+
+	dd if=boot0.bin of=drive.bin bs=1 count=446 seek=0 conv=notrunc
+	dd if=boot1.bin of=drive.bin bs=1 seek=512 conv=notrunc
+
+	cp drive.bin drive2.bin
+
+bootfs/luma_%.mod:
+	$(MAKE) -C luma/$* mod
+
+boot0.bin: src/boot0.asm boot1.bin
+	nasm -DBOOT1_SIZE=`stat -c %s boot1.bin` -fbin -o $@ $<
+
+boot1.bin: src/boot1.asm partuuid.txt
+	nasm -DPARTUUID=`cat partuuid.txt` -fbin -o $@ $<
+
+build/kernel/%.o: src/kernel/%.c
+	mkdir -p $(@D)
+	i386-elf-gcc -march=i386 -Isrc -Os -pie -nodefaultlibs -nostdlib -nostartfiles -fomit-frame-pointer -c -o $@ -ffreestanding -mgeneral-regs-only -std=gnu99 -Wall $<
+
+build/kernel/%.o: src/kernel/%.asm
+	mkdir -p $(@D)
+	nasm -felf32 -o $@ $<
+
+partuuid.txt:
+	openssl rand -hex 16 > partuuid.txt
+
+qemu:
+	qemu-system-i386 -m 32 -drive if=none,id=driev,format=raw,file=drive.bin -usb -debugcon stdio -soundhw hda -device usb-storage,bus=usb-bus.0,drive=driev

README.md

@@ -0,0 +1,41 @@
+# Luma
+
+This is a PC operating system mostly for my own learning and enjoyment, but also I like the idea of giving new life to old hardware that is slowly being abandoned by the masses. It has been in development since around 2022.
+
+In order to work in the absence of virtual memory, all software on Luma, including the Luma kernel itself, is forced to be at least relocatable if not position-independent. Because of this Luma, unlike other kernels, does not require to be higher-half.
+
+Luma as a microkernel enforces only a few things, such as process scheduling or hardware access.
+
+Supported hardware:
+
+1. Minimum 80386+
+2. VGA graphics, using k-means clustering for maximal color depth
+3. Intel High Definition Audio
+4. Universal Host Controller Interface
+5. USB BBB MSD
+6. PS/2 keyboard
+7. Filesystem driver for MeanFS (based on LeanFS)
+
+Note: though the drivers themselves work, they're limited in functionality API-wise and are mostly useless for user-space. Also, the original 80386 processors have too many bugs for Luma to feasibly work on real hardware.
+
+Incomplete list of planned features:
+
+1. Fix remaining undefined behavior before continuing
+2. Rewrite the window manager to use k3Menu
+3. Create a user interface for audio output
+4. Kernel-level RNG
+5. Support more than 16MB of RAM
+6. Intel GM45 graphics for 2D
+7. Intel GM45 graphics for 3D :)
+
+# Building
+
+Simply run `make` (GNU Make is required).
+
+Building requires an i386-elf-gcc toolchain. Clang *does not work* for reasons I have not figured out.
+
+Also required is the `meanfuse` program, which is a MeanFS user-space driver for Linux.
+
+# Boot logic
+
+In the interest of minimal RAM usage, Luma uses a custom bootloader (src/boot{0,1}.asm) which first initializes Luma to a valid state before handing off execution. This includes setting up the segments, the bitmap of physical memory, the initial page tables, loading the first kernel modules among other things. Slightly breaking the microkernel abstraction, this also includes setting up Mode X graphics through the 16-bit BIOS, because changing modes manually is painful. After this is done, the bootloader jumps to the Luma scheduler.

elftoexe.py

@@ -0,0 +1,107 @@
+#!/usr/bin/env python3
+
+from elftools.elf.elffile import ELFFile
+from elftools.elf.relocation import RelocationSection
+import struct, sys, itertools
+from dataclasses import dataclass
+from collections import namedtuple
+
+def bitsize(i):
+	return max(len(bin(i)[2:]), 1)
+
+@dataclass
+class X:
+	id: int
+	children: "any type"
+
+def dump(x):
+	data = b''.join(dump(c) for c in x.children) if type(x.children) == list else x.children
+	l = len(data)
+	l = l | (1 << ((bitsize(l) + 6) // 7 * 7))
+	return x.id.to_bytes(length = (bitsize(x.id) + 7) // 8) + l.to_bytes(length = (bitsize(l) + 7) // 8) + data
+
+ARGS = {}
+
+for s in sys.argv[3:]:
+	k, v = s.split("=", 1)
+	ARGS[k] = v
+
+SYMTRANSLATION = {}
+
+HEADER = X(0x1A45DFA3, [X(0x4282, b'BDexeF')])
+OUTPUT = X(0x10000000, [])
+
+@dataclass
+class ELFReloc:
+	offset: int
+	reference: int
+
+@dataclass
+class ELFSegment:
+	data: bytearray
+	vaddr: int
+	relocations: list[ELFReloc]
+
+with open(sys.argv[1], "rb") as inp, open(sys.argv[2], "wb") as outp:
+	elf = ELFFile(inp)
+	
+	segments = []
+	
+	sections_base = min([sec["sh_offset"] for sec in elf.iter_sections() if sec.name])
+	
+	base_vaddr = min([seg["p_vaddr"] for seg in elf.iter_segments()])
+	
+	for seg in elf.iter_segments():
+		if seg["p_type"] != "PT_LOAD":
+			continue
+		
+		# If no section is in the segment, then ignore it
+		if seg["p_offset"] < sections_base and seg["p_offset"] + seg["p_filesz"] < sections_base:
+			continue
+		
+		inp.seek(seg["p_offset"])
+		
+		segments.append(ELFSegment(vaddr = seg["p_vaddr"], data = bytearray(inp.read(seg["p_filesz"]) + b'\0' * (seg["p_memsz"] - seg["p_filesz"])), relocations = []))
+	
+	rel_section = elf.get_section_by_name(".rel.dyn")
+	
+	if rel_section:
+		for rel in rel_section.iter_relocations():
+			assert rel["r_info"] % 256 == 8
+			
+			# The segment in which the relocation is found/required
+			seg_in = next((s for s in reversed(segments) if s.vaddr <= rel["r_offset"]), None)
+			
+			offset = rel["r_offset"] - seg_in.vaddr
+			
+			val = struct.unpack("I", seg_in.data[offset : offset + 4])[0]
+			
+			# The segment pointed to by the relocation
+			seg_ref = next((s for s in reversed(segments) if s.vaddr <= val), None)
+			
+			seg_in.data[offset : offset + 4] = struct.pack("I", val - seg_ref.vaddr)
+			
+			seg_in.relocations.append(ELFReloc(offset = rel["r_offset"] - seg_in.vaddr, reference = segments.index(seg_ref)))
+	
+	for seg in segments:
+		OUTPUT.children.append(X(0x6200, [ #Segment
+			X(0x6201, seg.data), #Segment data
+			X(0x6202, b''.join(struct.pack("IB", rel.offset, rel.reference) for rel in seg.relocations)) #Segment relocations
+		]))
+	
+	for pub, sym in [(False, s) for s in elf.get_section_by_name(".symtab").iter_symbols()] + [(True, s) for s in elf.get_section_by_name(".dynsym").iter_symbols()]:
+		if not sym.name:
+			continue
+		
+		seg = next((s for s in reversed(segments) if s.vaddr <= sym["st_value"]), None)
+		
+		if not seg:
+			continue
+		
+		OUTPUT.children.append(X(0xC0, [
+			X(0xC1, sym.name.encode()),
+			X(0xC2, struct.pack("IB", sym["st_value"] - seg.vaddr, segments.index(seg) | (128 if pub else 0)))
+		]))
+	
+	outp.write(dump(HEADER))
+	outp.write(dump(OUTPUT))

elftomod.py

@@ -0,0 +1,48 @@
+#!/usr/bin/env python3
+
+from elftools.elf.elffile import ELFFile
+from elftools.elf.relocation import RelocationSection
+import sys, struct, itertools, re, os
+
+SYMTRANSLATION = {key: int(value) for key, value in [z.split(":") for z in sys.argv[3:]]}
+
+print(f"Converting {os.path.realpath(sys.argv[1])} to {os.path.realpath(sys.argv[2])}...")
+
+with open(sys.argv[1], "rb") as inp, open(sys.argv[2], "wb") as outp:
+	elf = ELFFile(inp)
+	loadhdr = next(elf.get_segment(i) for i in itertools.count() if elf.get_segment(i)["p_type"] == "PT_LOAD")
+	
+	relsec = elf.get_section_by_name(".rel.dyn")
+	datasec = elf.get_section_by_name(".text")
+	
+	data = bytearray(datasec.data())
+	data += bytearray(loadhdr["p_memsz"] - len(data)) #Make space to fit .bss
+	
+	outp.write(b"MOD\0")
+	
+	syms = [s for s in elf.get_section_by_name(".symtab").iter_symbols() if s.name in SYMTRANSLATION]
+	syms.sort(key = lambda s: SYMTRANSLATION[s.name])
+	
+	relocs = [r for r in relsec.iter_relocations()] if relsec else []
+	
+	outp.write(struct.pack("H", len(syms)))
+	outp.write(struct.pack("H", len(relocs)))
+	outp.write(struct.pack("I", len(data)))
+	
+	for i, s in enumerate(syms):
+		outp.write(struct.pack("H", i)) #struct.pack emits 8 bytes when you put HI directly
+		outp.write(struct.pack("I", s.entry["st_value"] - datasec["sh_addr"]))
+	
+	for r in relocs:
+		assert(r["r_info"] % 256 == 8)
+		
+		idx = r["r_offset"] - datasec["sh_addr"]
+		data[idx:idx + 4] = struct.pack("I", struct.unpack("I", data[idx:idx + 4])[0] - datasec["sh_addr"])
+		
+		outp.write(struct.pack("I", idx))
+	
+	# Align to 16
+	if outp.tell() % 16 != 0:
+		outp.write(b"\0" * (16 - outp.tell() % 16))
+	
+	outp.write(data)

exe.ld

@@ -0,0 +1,20 @@
+SECTIONS
+{
+	/*	Linkers like to insert the ELF headers into the segments themselves
+		Page-aligning the base address minimzies the likelihood of this happening */
+	. = 0x10000;
+	.text : ALIGN(0x1000) { *(.text) *(.rodata) }
+	.data : ALIGN(0x1000) { *(.data) *(.bss) }
+	
+	/DISCARD/ : {
+		*(.note.GNU-stack)
+		*(.gnu_debuglink)
+		*(.gnu.lto_*)
+		*(.note.gnu.build-id)
+		*(.eh_frame_hdr)
+		*(.eh_frame_entry .eh_frame_entry.*)
+		*(.eh_frame)
+		*(.eh_frame.*)
+		*(.interp)
+	}
+}

luma/ata/dio.h

@@ -0,0 +1,17 @@
+#ifndef _LUMA_DIO
+#define _LUMA_DIO
+
+typedef enum {
+	LUMA_DIO_COMMAND_READ
+} LumaDIOCommandOp;
+
+typedef union {
+	uint8_t op;
+	struct {
+		uint8_t op;
+		uint32_t lba;
+		uint32_t count;
+	} read;
+} LumaDIOCommand;
+
+#endif

luma/ata/main.c

@@ -0,0 +1,85 @@
+#include<stdint.h>
+#include<stddef.h>
+#include"../sys.h"
+#include"dio.h"
+
+#define IO_BASE 0x1F0
+#define REG_DATA 0
+#define REG_RERROR_WFEATURES 1
+#define REG_SECTOR_COUNT 2
+#define REG_LBA_LOW 3
+#define REG_LBA_MIDDLE 4
+#define REG_LBA_HIGH 5
+#define REG_DRIVE 6
+#define REG_RSTATUS_WCOMMAND 7
+
+char stack[768];
+uint16_t data[256];
+
+static void waitUntilReady() {
+	while((inb(IO_BASE + REG_RSTATUS_WCOMMAND) & 128) == 128) {
+	}
+	while((inb(IO_BASE + REG_RSTATUS_WCOMMAND) & 8) == 0) {
+	}
+}
+
+static void wait400() {
+	inb(0x3F6);
+	inb(0x3F6);
+	inb(0x3F6);
+	inb(0x3F6);
+}
+
+void startc() {
+	outb(IO_BASE + REG_DRIVE, 0xA0);
+	outb(IO_BASE + REG_SECTOR_COUNT, 0);
+	outb(IO_BASE + REG_LBA_LOW, 0);
+	outb(IO_BASE + REG_LBA_MIDDLE, 0);
+	outb(IO_BASE + REG_LBA_HIGH, 0);
+	outb(IO_BASE + REG_RSTATUS_WCOMMAND, 0xEC);
+	
+	waitUntilReady();
+	
+	/* IDENTIFY data, which we don't need (yet). */
+	for(int i = 0; i < 256; i++) {
+		inw(IO_BASE + REG_DATA);
+	}
+	
+	uint32_t canalID = sys_canal_create(0x42444844, 0x00000000);
+	
+	uint16_t *buf = (void*) -1;
+	do {
+		buf = sys_canal_accept(canalID);
+	} while(buf == (void*) -1);
+	
+	while(1) {
+		memory_barrier();
+		sys_signal_wait(SYS_SIGNAL_WAIT_ANY);
+		
+		LumaDIOCommand *cmd = (void*) buf;
+		if(cmd->op == LUMA_DIO_COMMAND_READ) {
+			uint8_t count = cmd->read.count;
+			
+			outb(IO_BASE + REG_DRIVE, 0xE0);
+			outb(IO_BASE + REG_RERROR_WFEATURES, 0x00);
+			outb(IO_BASE + REG_SECTOR_COUNT, count);
+			outb(IO_BASE + REG_LBA_LOW, (cmd->read.lba >> 0) & 0xFF);
+			outb(IO_BASE + REG_LBA_MIDDLE, (cmd->read.lba >> 8) & 0xFF);
+			outb(IO_BASE + REG_LBA_HIGH, (cmd->read.lba >> 16) & 0xFF);
+			outb(IO_BASE + REG_RSTATUS_WCOMMAND, 0x20);
+			
+			int w = 0;
+			while(count--) {
+				waitUntilReady();
+				for(int i = 0; i < 256; i++) {
+					buf[w++] = inw(IO_BASE + REG_DATA);
+				}
+			}
+			
+			wait400();
+		}
+		
+		memory_barrier();
+		sys_signal_send(buf);
+	}
+}

luma/cardistry/main.c

@@ -0,0 +1,21 @@
+#include"../sys.h"
+
+#include"../wm/api.h"
+
+void giveup() {
+	for(int i = 0; i < 15; i++) {
+		asm("outb %%al, $0xE9" :: "a"('A') :);
+		asm("outb %%al, $0xE9" :: "a"('\n') :);
+	}
+	while(1);
+}
+
+__attribute__((noreturn)) void ProgramEntry() {
+	LumaWMClient wmcli;
+	
+	if(luma_wm_client(&wmcli) == 0) {
+		giveup();
+	}
+	
+	while(1);
+}

luma/fs/Makefile

@@ -0,0 +1,13 @@
+rwildcard=$(foreach d,$(wildcard $(1:=/*)),$(call rwildcard,$d,$2) $(filter $(subst *,%,$2),$d))
+
+HEADERS := $(call rwildcard,.,*.h)
+SOURCEC := $(call rwildcard,.,*.c)
+OBJECTS := $(patsubst %.c, ../../build/lean/%.o, $(SOURCEC))
+
+mod: $(OBJECTS)
+	i386-elf-gcc -pie -T ../../mod.ld -Wl,-no-dynamic-linker -nostdlib -o main.elf $(OBJECTS) -lgcc
+	../../elftomod.py main.elf ../../bootfs/luma_fs.mod modentry:0
+
+../../build/lean/%.o: %.c
+	mkdir -p $(@D)
+	i386-elf-gcc -march=i386 -Isrc -Os -pie -nodefaultlibs -nostdlib -nostartfiles -fomit-frame-pointer -c -o $@ -ffreestanding -mgeneral-regs-only -std=gnu99 -Wall $<

luma/fs/fs.h

@@ -0,0 +1,39 @@
+#ifndef _LUMA_FS_H
+#define _LUMA_FS_H
+
+#include"../std.h"
+
+// Paths are not handled as strings, but rather as packed arrays of strings
+// This way there is no need for escaping, separators, etc.
+
+typedef enum {
+	LUMA_FS_COMMAND_READ, LUMA_FS_COMMAND_INFO, LUMA_FS_COMMAND_TRAVERSE
+} LumaFSCommandOp;
+
+typedef union {
+	uint8_t op;
+	struct {
+		uint8_t op;
+		uint64_t offset;
+		uint32_t bytes;
+		uint8_t path[]; //Packed array of Str16
+	} read;
+	struct {
+		uint8_t op;
+		uint64_t fileSize;
+		uint8_t path[]; //Packed array of Str16
+	} info;
+	struct {
+		uint8_t op;
+		uint8_t path[]; // Packed array of Str16
+	} traverse;
+} LumaFSCommand;
+
+#define LUMA_FS_DIR_ENTRY_ISDIR 2
+#define LUMA_FS_DIR_ENTRY_LAST 1
+typedef struct {
+	uint16_t type;
+	Str16 name;
+} LumaFSDirectoryEntry;
+
+#endif

luma/fs/main.c

@@ -0,0 +1,475 @@
+#include"../sys.h"
+
+#include<stdint.h>
+#include<stddef.h>
+#include"../std.h"
+#include"fs.h"
+
+asm("modentry:\n"
+"mov $stack + 2048, %esp\n"
+"call startc");
+
+char stack[3072];
+
+// Should be renamed. This *isn't* LeanFS, however inspired it may be.
+
+#define EXTENTS_PER_INODE 6
+
+#define INODE_ftRegular 1
+#define INODE_ftDirectory 2
+#define INODE_ftMask 7
+
+#define INODE_ATTRIBUTE_iaInlineExtAttr (1 << 19)
+#define INODE_ATTRIBUTE_iaFmtRegular (1 << 29)
+#define INODE_ATTRIBUTE_iaFmtDirectory (2 << 29)
+#define INODE_ATTRIBUTE_iaFmtMask (7 << 29)
+
+
+struct Superblock {
+	uint32_t checksum;
+	uint32_t magic;
+	uint16_t fsVersion;
+	uint8_t preallocCount;
+	uint8_t logBlocksPerBand;
+	uint32_t state;
+	uint8_t uuid[16];
+	char volumeLabel[64];
+	uint64_t blockCount;
+	uint64_t freeBlockCount;
+	uint64_t primarySuper;
+	uint64_t backupSuper;
+	uint64_t bitmapStart;
+	uint64_t rootInode;
+	uint64_t badInode;
+	uint64_t journalInode;
+	uint8_t logBlockSize;
+	uint8_t reserved0[7];
+} __attribute__((packed, aligned(4)));
+
+struct INode {
+	uint32_t checksum;
+	uint32_t magic;
+	uint8_t extentCount;
+	uint8_t reserved[3];
+	uint32_t indirectCount;
+	uint32_t linkCount;
+	uint32_t uid;
+	uint32_t gid;
+	uint32_t attributes;
+	uint64_t fileSize;
+	uint64_t blockCount;
+	int64_t accessTime;
+	int64_t statusChangeTime;
+	int64_t modificationTime;
+	int64_t creationTime;
+	uint64_t firstIndirect;
+	uint64_t lastIndirect;
+	uint64_t fork;
+	uint64_t extentStarts[EXTENTS_PER_INODE];
+	uint32_t extentSizes[EXTENTS_PER_INODE];
+} __attribute__((packed));
+
+struct DirectoryEntry {
+	uint64_t inode;
+	uint8_t type;
+	uint8_t recLen;
+	uint16_t nameLen;
+	uint8_t name[];
+} __attribute__((packed));
+
+uint8_t *diskWindow = (uint8_t*) -1;
+
+size_t volumeStartLBA, superblockLBA;
+struct Superblock superblock;
+
+void read(size_t start, size_t length) {
+	((uint64_t*) diskWindow)[0] = start;
+	((uint32_t*) diskWindow)[2] = length;
+	
+	memory_barrier();
+	sys_signal_send(diskWindow, 0);
+	
+	sys_signal_wait(diskWindow, NULL);
+}
+
+static uint32_t lean_compute_checksum(const uint32_t *buf, size_t dwordLength, size_t dwordPadTo) {
+	uint32_t sum = 0;
+	
+	size_t i = 1;
+	for(; i < dwordLength; i++) {
+		sum = (sum << 31) + (sum >> 1);
+		sum += buf[i];
+	}
+	
+	for(; i < dwordPadTo; i++) {
+		sum = (sum << 31) + (sum >> 1);
+	}
+	
+	return sum;
+}
+
+int parse_inode(size_t id) {
+	read(volumeStartLBA * 512 + (id << superblock.logBlockSize), 1 << superblock.logBlockSize);
+	
+	struct INode *test = (struct INode*) diskWindow;
+	
+	if(test->magic != 0x45444F4E) {
+		return 0;
+	}
+	
+	// Driver has outdated checksums
+#if 0
+	if(lean_compute_checksum((uint32_t*) test, (sizeof(struct INode)) >> 2, (sizeof(struct INode)) >> 2) != test->checksum) {
+		return 0;
+	}
+#endif
+	
+	return 1;
+}
+
+size_t lean_read_inode(const struct INode *inode, size_t off, size_t len, uint8_t *ret) {
+	if(off + len > inode->fileSize) {
+		len = inode->fileSize - off;
+	}
+	
+	if(inode->attributes & INODE_ATTRIBUTE_iaInlineExtAttr) {
+		off += 1 << superblock.logBlockSize;
+	} else {
+		off += sizeof(struct INode);
+	}
+	
+	size_t startBlk = off >> superblock.logBlockSize;
+	size_t endBlk = (off + len - 1) >> superblock.logBlockSize;
+	
+	uint32_t b = 0;
+	uint32_t ei, eb;
+	for(ei = 0; ei < EXTENTS_PER_INODE; ei++) {
+		for(eb = 0; eb < inode->extentSizes[ei]; eb++) {
+			if(b == startBlk) {
+				goto foundStart;
+			}
+			
+			b++;
+		}
+	}
+	
+	// Not found. Corruption?
+	return 0;
+	
+foundStart:
+	
+	while(1) {
+		read(volumeStartLBA * 512 + ((inode->extentStarts[ei] + eb) << superblock.logBlockSize), 1 << superblock.logBlockSize);
+		
+		if(b == startBlk && b == endBlk) {
+			std_copy(ret, diskWindow + off, len);
+		
+			break;
+		} else if(b == startBlk) {
+			uint16_t mod = off % (1 << superblock.logBlockSize);
+			
+			std_copy(ret, diskWindow + mod, (1 << superblock.logBlockSize) - mod);
+			
+			ret += (1 << superblock.logBlockSize) - mod;
+		} else if(b == endBlk) {
+			uint16_t mod = (off + len - 1) % (1 << superblock.logBlockSize);
+			
+			std_copy(ret, diskWindow, mod + 1);
+			
+			break;
+		} else {
+			std_copy(ret, diskWindow, 1 << superblock.logBlockSize);
+			
+			ret += 1 << superblock.logBlockSize;
+		}
+		
+		eb++, b++;
+		
+		if(eb == inode->extentSizes[ei]) {
+			eb = 0;
+			ei++;
+			
+			if(ei == EXTENTS_PER_INODE) {
+				// Stop. Indirect blocks not yet supported.
+				return 0;
+			}
+		}
+	}
+	
+	return len;
+}
+
+static inline void pr(int c) {
+	asm volatile("outb %%al, $0xE9" :: "a"(c) :);
+}
+
+static inline void pri(size_t num) {
+	char buf[16] = {};
+	int i = 16;
+	
+	do {
+		buf[--i] = num % 10;
+		num /= 10;
+	} while(num);
+	
+	for(; i < 16; i++) {
+		pr(buf[i] + '0');
+	}
+}
+
+static inline void prs(const Str16 *str) {
+	for(int i = 0; i < str->len; i++) {
+		pr(str->data[i]);
+	}
+}
+
+static inline void prcs(const char *str) {
+	while(*str) {
+		pr(*str);
+		str++;
+	}
+}
+
+// I don't like this.
+// Biggest ugly by far is how it operates in 512-byte chunks, even though lean_read_inode already handles unaligned access
+uint64_t find_child(const struct INode *node, const Str16 *name) {
+	uint8_t buf[512] __attribute__((aligned(16)));
+	
+	size_t offset = 0;
+	lean_read_inode(node, offset, sizeof(buf), buf);
+	
+	struct DirectoryEntry *e = (struct DirectoryEntry*) buf;
+	while(1) {
+		if(offset + ((uintptr_t) e - (uintptr_t) buf) >= node->fileSize) {
+			// Reached EOF.
+			return 0;
+		} else while((uintptr_t) e >= (uintptr_t) buf + sizeof(buf)) {
+			offset += sizeof(buf);
+			lean_read_inode(node, offset, sizeof(buf), buf);
+			e = (struct DirectoryEntry*) (((uintptr_t) e - (uintptr_t) buf) % sizeof(buf) + (uintptr_t) buf);
+		}
+		
+		if(e->nameLen != name->len) {
+			uintptr_t after = (uintptr_t) e + 16 * e->recLen;
+			
+			e = (struct DirectoryEntry*) after;
+			
+			goto next;
+		}
+		
+		// Must be saved, as `e` becomes invalid in the following for loop
+		uint64_t childinode = e->inode;
+		size_t nameLen = e->nameLen;
+		uint8_t *testname = e->name;
+		for(uint16_t i = 0; i < nameLen; i++) {
+			if(testname - buf == sizeof(buf)) {
+				offset += sizeof(buf);
+				lean_read_inode(node, offset, sizeof(buf), buf);
+				testname = buf;
+			}
+			if(*testname != name->data[i]) {
+				testname += nameLen - i;
+				
+				e = (struct DirectoryEntry*) ((((uintptr_t) testname) + 15) & ~15);
+				
+				goto next;
+			}
+			testname++;
+		}
+		
+		return childinode;
+
+next:
+	}
+}
+
+uint64_t get_inode_from_str16_path(const Str16 *path, struct INode *data) {
+	uint64_t child = superblock.rootInode;
+	if(!parse_inode(child)) {
+		return 0;
+	}
+	std_copy(data, diskWindow, sizeof(*data));
+	
+	while(path->len) {
+		child = find_child(data, path);
+		
+		if(!child) {
+			return 0;
+		}
+		
+		if(!parse_inode(child)) {
+			return 0;
+		}
+		std_copy(data, diskWindow, sizeof(*data));
+		
+		path = (const Str16*) ((uintptr_t) path + sizeof(*path) + path->len);
+	}
+	
+	return child;
+}
+
+void startc() {
+	do {
+		diskWindow = sys_link_create('STRG', 'UD\0\0', 4096);
+	} while(diskWindow == (void*) -1);
+	
+	read(0, 512);
+	volumeStartLBA = *(uint32_t*) &diskWindow[0x1C6];
+	
+	int found = 0;
+	for(size_t superblockLBA = 0; superblockLBA < 256; superblockLBA++) {
+		read((volumeStartLBA + superblockLBA) * 512, 512);
+		
+		struct Superblock *candidate = (struct Superblock*) diskWindow;
+		
+		if(candidate->magic == 0x4E41454C && lean_compute_checksum((uint32_t*) diskWindow, sizeof(struct Superblock) >> 2, (1 << candidate->logBlockSize) >> 2) == candidate->checksum) {
+			superblock = *(struct Superblock*) diskWindow;
+			found = 1;
+			break;
+		}
+	}
+	
+	if(found) {
+		uint32_t canalID = sys_canal_create('FS\0\0', 'ROOT');
+		while(1) {
+			sys_canal_accept(canalID);
+			
+			uint8_t *fsBuf = sys_signal_wait(SYS_SIGNAL_WAIT_ANY, NULL);
+			memory_barrier();
+			
+			LumaFSCommand *cmd = (void*) fsBuf;
+			if(cmd->op == LUMA_FS_COMMAND_READ) {
+				struct INode data;
+				uint64_t inode = get_inode_from_str16_path((Str16*) cmd->read.path, &data);
+				
+				if(inode) {
+					size_t offset = cmd->read.offset;
+					uint32_t left = cmd->read.bytes;
+					
+					if(left) {
+						while(1) {
+							size_t amount = left > 4096 ? 4096 : left;
+							
+							size_t actuallyRead = lean_read_inode(&data, offset, amount, fsBuf);
+							if(actuallyRead < amount) {
+								amount = actuallyRead;
+								left = actuallyRead;
+							}
+							
+							memory_barrier();
+							sys_signal_send(fsBuf, amount);
+							
+							if(left == 4096) {
+								sys_signal_wait(fsBuf, NULL);
+								sys_signal_send(fsBuf, 0);
+								break;
+							} else {
+								offset += amount;
+								left -= amount;
+								
+								if(left == 0) {
+									break;
+								} else {
+									sys_signal_wait(fsBuf, NULL);
+								}
+							}
+						}
+					}
+				} else {
+					memory_barrier();
+					sys_signal_send(fsBuf, 0);
+				}
+			} else if(cmd->op == LUMA_FS_COMMAND_INFO) {
+				struct INode data;
+				uint64_t inode = get_inode_from_str16_path((Str16*) cmd->info.path, &data);
+				
+				prcs("Requesting info ");
+				prs(cmd->info.path);
+				pr('\n');
+				
+				prcs("Got inode ");
+				pri(inode);
+				pr('\n');
+				
+				if(inode) {
+					prcs("Size ");
+					pri(data.fileSize);
+					pr('\n');
+					
+					prcs("magic ");
+					pri(data.magic);
+					pr('\n');
+					
+					cmd->info.fileSize = data.fileSize;
+					
+					memory_barrier();
+					sys_signal_send(fsBuf, 1);
+				} else {
+					memory_barrier();
+					sys_signal_send(fsBuf, 0);
+				}
+			} else if(cmd->op == LUMA_FS_COMMAND_TRAVERSE) {
+				struct INode data;
+				uint64_t inode = get_inode_from_str16_path((Str16*) cmd->traverse.path, &data);
+				
+				if(inode) {
+					size_t offset = 0, left = data.fileSize;
+					uint8_t buf[512] __attribute__((aligned(16)));
+					
+					lean_read_inode(&data, offset, sizeof(buf), buf);
+					
+					struct DirectoryEntry *in = (struct DirectoryEntry*) buf;
+					
+					LumaFSDirectoryEntry *out = (LumaFSDirectoryEntry*) fsBuf;
+					
+					while(1) {
+						size_t len = in->nameLen > 255 ? 255 : in->nameLen;
+						
+						if((uintptr_t) out + sizeof(*buf) + len > (uintptr_t) fsBuf + 4096) {
+							memory_barrier();
+							sys_signal_send(fsBuf, 2);
+							sys_signal_wait(fsBuf, NULL);
+						}
+						
+						out->type = ((in->type & INODE_ftMask) == INODE_ftDirectory) ? LUMA_FS_DIR_ENTRY_ISDIR : 0;
+						out->name.len = len;
+						uint8_t *nameIn = in->name;
+						for(size_t n = 0; n < out->name.len; n++) {
+							out->name.data[n] = *(nameIn++);
+							if((uintptr_t) nameIn == (uintptr_t) buf + sizeof(buf)) {
+								offset += sizeof(buf);
+								left -= sizeof(buf);
+								lean_read_inode(&data, offset, sizeof(buf), buf);
+								nameIn = buf;
+							}
+						}
+						
+						in = (struct DirectoryEntry*) (((uintptr_t) nameIn + 15) & ~15);
+						if(offset + ((uintptr_t) in - (uintptr_t) buf) == data.fileSize) {
+							out->type |= LUMA_FS_DIR_ENTRY_LAST;
+							break;
+						} else if((uintptr_t) in == (uintptr_t) in + sizeof(buf)) {
+							offset += sizeof(buf);
+							left -= sizeof(buf);
+							lean_read_inode(&data, offset, sizeof(buf), buf);
+							in = (struct DirectoryEntry*) buf;
+						}
+						
+						out = (LumaFSDirectoryEntry*) ((uintptr_t) out + sizeof(*out) + len);
+					}
+					
+					memory_barrier();
+					sys_signal_send(fsBuf, 1);
+				} else {
+					memory_barrier();
+					sys_signal_send(fsBuf, 0);
+				}
+			} else {
+				memory_barrier();
+				sys_signal_send(fsBuf, 0);
+			}
+		}
+	} else {
+		while(1) { sys_signal_wait(SYS_SIGNAL_WAIT_ANY, NULL); }
+	}
+}

luma/hda/Makefile

@@ -0,0 +1,13 @@
+rwildcard=$(foreach d,$(wildcard $(1:=/*)),$(call rwildcard,$d,$2) $(filter $(subst *,%,$2),$d))
+
+HEADERS := $(call rwildcard,.,*.h)
+SOURCEC := $(call rwildcard,.,*.c)
+OBJECTS := $(patsubst %.c, ../../build/hda/%.o, $(SOURCEC))
+
+mod: $(OBJECTS)
+	i386-elf-gcc -pie -T ../../mod.ld -Wl,-no-dynamic-linker -nostdlib -o main.elf $(OBJECTS) -lgcc
+	../../elftomod.py main.elf ../../bootfs/luma_hda.mod modentry:0
+
+../../build/hda/%.o: %.c
+	mkdir -p $(@D)
+	i386-elf-gcc -march=i386 -Isrc -Os -pie -nodefaultlibs -nostdlib -nostartfiles -fomit-frame-pointer -c -o $@ -ffreestanding -mgeneral-regs-only -std=gnu99 -Wall $<

luma/hda/main.c

@@ -0,0 +1,304 @@
+#include"../sys.h"
+
+asm("modentry:\n"
+"mov $stack + 1536, %esp\n"
+"call startc");
+
+#include"../std.h"
+
+char stack[2048];
+
+#define FORMAT_CHANNELS(n) ((n) - 1)
+#define FORMAT_16BPS (1 << 4)
+#define FORMAT_RATE_MUL_1 0
+#define FORMAT_RATE_DIV_1 0
+#define FORMAT_PCM (0 << 15)
+#define FORMAT_48KHZ (0 << 14)
+
+#define STANDARD_48KHZ_16BPS (FORMAT_CHANNELS(1) | FORMAT_PCM | FORMAT_48KHZ | FORMAT_RATE_MUL_1 | FORMAT_RATE_DIV_1 | FORMAT_16BPS)
+
+#define STREAM_RUN_BIT 2
+#define STREAM_NO(n) ((n) << 20)
+
+struct BufDescEntry {
+	uint64_t addr;
+	uint32_t length;
+	uint32_t flags;
+};
+static struct BufDescEntry *bufDesc;
+
+volatile uint32_t *HDA;
+
+#define WIDGET_TYPE(w) (((w)->cap >> 20) & 15)
+#define WIDGET_AMP_OVERRIDE(w) (((w)->cap >> 3) & 1)
+struct Widget {
+	uint32_t cap;
+	uint8_t connectionCount;
+	uint8_t queued; //Used for BFS
+	uint8_t parent;
+};
+static uint8_t WidgetCount, WidgetsStart;
+
+static uint8_t TheCodec, ThePin;
+static uint32_t DefaultAFGInAmpCaps;
+static uint32_t DefaultAFGOutAmpCaps;
+
+static struct Widget widgets[64];
+
+static uint32_t cmd(uint8_t codec, uint8_t node, uint32_t c) {
+	HDA[26] = 2;
+	HDA[24] = (codec << 28) | (node << 20) | c;
+	HDA[26] = 1;
+	while((HDA[26] & 3) != 2);
+	return HDA[25];
+}
+static uint32_t verb(uint8_t codec, uint8_t node, uint16_t v, uint8_t payload) {
+	return cmd(codec, node, (v << 8) | payload);
+}
+
+static uint32_t verb16(uint8_t codec, uint8_t node, uint16_t v, uint16_t payload) {
+	return cmd(codec, node, (v << 16) | payload);
+}
+
+static void unmute(uint8_t widget, int output, int from, int to) {
+	uint32_t ampcap;
+	if(WIDGET_AMP_OVERRIDE(widgets + widget - WidgetsStart)) {
+		ampcap = verb(TheCodec, widget, 0xF00, 0xD + 5 * output);
+	} else {
+		ampcap = output ? DefaultAFGOutAmpCaps : DefaultAFGInAmpCaps;
+	}
+	for(int i = from; i <= to; i++) {
+		verb16(TheCodec, widget, 0x3, (ampcap & 127) | (i << 8) | (3 << 12) | ((1 + output) << 14));
+	}
+}
+
+static void klogc(int c) {
+	asm volatile("outb %%al, $0xE9"::"a"(c):);
+}
+static void klogp(void *p) {
+	static char hex[16]="0123456789ABCDEF";
+	
+	uintptr_t i = (uintptr_t) p;
+	klogc(hex[(i >> 28) & 0xF]);
+	klogc(hex[(i >> 24) & 0xF]);
+	klogc(hex[(i >> 20) & 0xF]);
+	klogc(hex[(i >> 16) & 0xF]);
+	klogc(hex[(i >> 12) & 0xF]);
+	klogc(hex[(i >> 8) & 0xF]);
+	klogc(hex[(i >> 4) & 0xF]);
+	klogc(hex[(i >> 0) & 0xF]);
+}
+
+void startc() {
+	uint32_t hcId = sys_pci_claim(0x04030000);
+	if(hcId == (uint32_t) -1) {
+		while(1) sys_signal_wait(SYS_SIGNAL_WAIT_ANY, NULL);
+	}
+	
+	// Clear TCSEL
+	sys_pci_write_1(hcId, 0x44, sys_pci_read(hcId, 0x44) & ~7);
+	
+	// Enable busmastering
+	sys_pci_write_2(hcId, 0x04, sys_pci_read(hcId, 0x04) | 7);
+	
+	HDA = sys_vpm_map(SYS_MAP_VIRT_ANY, (void*) (sys_pci_read(hcId, 0x10) & 0xFFFFFFF0), 4096);
+	
+	// Start reset.
+	HDA[2] = 0;
+	while((HDA[2] & 1) != 0);
+	sys_sleep(50);
+	
+	// Finish reset.
+	HDA[2] = 1;
+	while((HDA[2] & 1) == 0);
+	sys_sleep(10);
+	
+	int16_t *audbuf = sys_vpm_map(SYS_MAP_VIRT_ANY, SYS_MAP_PHYS_ANY, 8192 * sizeof(*audbuf));
+	for(int i = 0; i < 8192; i++) {
+		//// 2 buffers of square wave
+		audbuf[i] = i % 128 < 64 ? 10000 : -10000;
+		//audbuf[i] = 0;
+	}
+	
+	bufDesc = sys_vpm_map(SYS_MAP_VIRT_ANY, SYS_MAP_PHYS_ANY, 4096);
+	
+	bufDesc[0].addr = (uintptr_t) sys_vpm_get_phys(&audbuf[0]);
+	bufDesc[0].length = 8192;
+	bufDesc[0].flags = 1;
+	
+	bufDesc[1].addr = (uintptr_t) sys_vpm_get_phys(&audbuf[4096]);
+	bufDesc[1].length = 8192;
+	bufDesc[1].flags = 1;
+	
+	uint32_t *dmapos = (uint32_t*) &bufDesc[8];
+	HDA[28] = 1 | (uintptr_t) sys_vpm_get_phys(dmapos);
+	HDA[29] = 0;
+	
+	uint16_t gcap = *(volatile uint16_t*) HDA;
+	uint8_t inputStreamCount = (gcap >> 8) & 15;
+	
+	volatile uint32_t *outStream = HDA + 0x20 + 8 * inputStreamCount;
+	outStream[0] &= ~2; // Disable run
+	
+	// Reset.
+	outStream[0] |= 1;
+	int timeout = 500;
+	while((outStream[0] & 1) == 0 && timeout--);
+	
+	// Unreset.
+	outStream[0] &= ~1;
+	timeout = 500;
+	while((outStream[0] & 1) != 0 && timeout--);
+	
+	outStream[0] = STREAM_NO(1);
+	outStream[2] = 16384;
+	((volatile uint16_t*) outStream)[9] = STANDARD_48KHZ_16BPS;
+	((volatile uint16_t*) outStream)[6] = 1;
+	outStream[6] = (uintptr_t) sys_vpm_get_phys(bufDesc);
+	outStream[7] = 0;
+	
+	sys_sleep(10);
+	
+	uint16_t statests = ((volatile uint16_t*) HDA)[7];
+	for(int codec = 0; codec < 15; codec++) {
+		if(statests & (1 << codec)) {
+			TheCodec = codec;
+			break;
+		}
+	}
+	
+	// Get subordinate nodes
+	uint32_t boob = verb(TheCodec, 0, 0xF00, 4);
+	
+	uint8_t startIdx = (boob >> 16) & 0xFF;
+	uint8_t endIdx = startIdx + (boob & 0xFF);
+	for(int idx = startIdx; idx < endIdx; idx++) {
+		// Get function group type
+		uint32_t type = verb(TheCodec, idx, 0xF00, 5);
+		
+		if((type & 127) == 1) {
+			// Is audio function group
+			
+			// Power
+			verb(TheCodec, idx, 0x705, 0);
+			
+			sys_sleep(200);
+			
+			DefaultAFGInAmpCaps = verb(TheCodec, idx, 0xF00, 0xD);
+			DefaultAFGOutAmpCaps = verb(TheCodec, idx, 0xF00, 0x12);
+			
+			// Get subordinate nodes
+			uint32_t tit = verb(TheCodec, idx, 0xF00, 4);
+			
+			// Iterate over widgets
+			uint8_t wIdxStart = (tit >> 16) & 0xFF;
+			uint8_t wIdxEnd = wIdxStart + (tit & 0xFF);
+			
+			int outPinWPriority = 0;
+			int outPinWIdx = -1;
+			
+			WidgetsStart = wIdxStart;
+			
+			// We'll play audio through one pin with highest priority (headphones)
+			for(int w = wIdxStart; w < wIdxEnd; w++) {
+				struct Widget *wstruc = &widgets[WidgetCount];
+				
+				wstruc->cap = verb(TheCodec, w, 0xF00, 9);
+				
+				if(WIDGET_TYPE(wstruc) == 4) { //pin
+					uint32_t confdef = verb(TheCodec, w, 0xF1C, 0);
+					
+					uint8_t dev = (confdef >> 20) & 15;
+					
+					int priority;
+					if(dev == 2) {
+						// headphone
+						priority = 3;
+					} else if(dev == 1) {
+						// speaker
+						priority = 2;
+					} else if(dev == 0) {
+						// line out
+						priority = 1;
+						outPinWIdx = WidgetCount;
+					}
+					
+					if(priority > outPinWPriority) {
+						outPinWPriority = priority;
+						outPinWIdx = WidgetCount;
+					}
+				}
+				
+				wstruc->connectionCount = verb(TheCodec, w, 0xF00, 14);
+				
+				WidgetCount++;
+				
+				// power on
+				verb(TheCodec, w, 0x705, 0);
+			}
+			
+			ThePin = outPinWIdx;
+			
+			break;
+		}
+	}
+	
+	// Hardcoded path
+	unmute(0x11, 0, 0, 0);
+	unmute(0x11, 1, 0, 0);
+	verb(TheCodec, 0x11, 0x70C, 2); //eapd on
+	verb(TheCodec, 0x11, 0x707, (1 << 6) | (1 << 7)); // pin out
+	
+	unmute(0x07, 0, 0, 15);
+	unmute(0x07, 1, 0, 15);
+	
+	unmute(0x22, 1, 0, 0);
+	verb(TheCodec, 0x22, 0x701, 0);
+	
+	unmute(0x03, 1, 0, 0);
+	verb16(TheCodec, 0x03, 0x2, STANDARD_48KHZ_16BPS);
+	verb(TheCodec, 0x03, 0x706, (1 << 4) | (0 << 0));
+	
+	HDA[13] = 0xFFFFFFFF;
+	HDA[14] = 0xFFFFFFFF;
+	
+	sys_sleep(10);
+	
+	memory_barrier();
+	outStream[0] |= STREAM_RUN_BIT;
+	
+	while((outStream[0] & (1 << 29)) == 0) memory_barrier();
+	
+	sys_sleep(10);
+	
+	HDA[13] = 0;
+	HDA[14] = 0;
+	
+	// Create canal
+	uint32_t canalID = sys_canal_create('AUD\0', 'HDA0');
+	uint8_t *AudioLink = (void*) -1;
+	do {
+		AudioLink = sys_canal_accept(canalID);
+	} while(AudioLink == (void*) -1);
+	
+	sys_signal_send(AudioLink, 0);
+	
+	while(1) {
+		sys_signal_wait(AudioLink, NULL);
+		memory_barrier();
+		
+		while((outStream[0] & (1 << 26)) == 0);
+		outStream[0] |= 1 << 26;
+		
+		//int buf = (outStream[1] / 8192) ^ 1;
+		static int buf = 0;
+		
+		std_copy(audbuf + buf * 4096, AudioLink, 8192);
+		buf ^= 1;
+		
+		memory_barrier();
+		sys_signal_send(AudioLink, 0);
+	}
+	
+	while(1) sys_signal_wait(SYS_SIGNAL_WAIT_ANY, NULL);
+}

luma/ps2/Makefile

@@ -0,0 +1,13 @@
+rwildcard=$(foreach d,$(wildcard $(1:=/*)),$(call rwildcard,$d,$2) $(filter $(subst *,%,$2),$d))
+
+HEADERS := $(call rwildcard,.,*.h)
+SOURCEC := $(call rwildcard,.,*.c)
+OBJECTS := $(patsubst %.c, ../../build/ps2/%.o, $(SOURCEC))
+
+mod: $(OBJECTS)
+	i386-elf-gcc -pie -T ../../mod.ld -Wl,-no-dynamic-linker -nostdlib -o main.elf $(OBJECTS) -lgcc
+	../../elftomod.py main.elf ../../bootfs/luma_ps2.mod modentry:0
+
+../../build/ps2/%.o: %.c
+	mkdir -p $(@D)
+	i386-elf-gcc -march=i386 -Isrc -Os -pie -nodefaultlibs -nostdlib -nostartfiles -fomit-frame-pointer -c -o $@ -ffreestanding -mgeneral-regs-only -std=gnu99 -Wall $<

luma/ps2/api.h

@@ -0,0 +1,97 @@
+#ifndef _LUMA_KBD_H
+#define _LUMA_KBD_H
+
+enum LumaKbdScancode {
+	LK_INVALID,
+	LK_0,
+	LK_1,
+	LK_2,
+	LK_3,
+	LK_4,
+	LK_5,
+	LK_6,
+	LK_7,
+	LK_8,
+	LK_9,
+	LK_F1,
+	LK_F2,
+	LK_F3,
+	LK_F4,
+	LK_F5,
+	LK_F6,
+	LK_F7,
+	LK_F8,
+	LK_F9,
+	LK_F10,
+	LK_F11,
+	LK_F12,
+	LK_PAUSE,
+	LK_LEFT_ALT,
+	LK_RIGHT_ALT,
+	LK_LEFT_SHIFT,
+	LK_RIGHT_SHIFT,
+	LK_LEFT_CTRL,
+	LK_RIGHT_CTRL,
+	LK_LEFT_SYS,
+	LK_RIGHT_SYS,
+	LK_BACKSPACE,
+	LK_A,
+	LK_B,
+	LK_C,
+	LK_D,
+	LK_E,
+	LK_F,
+	LK_G,
+	LK_H,
+	LK_I,
+	LK_J,
+	LK_K,
+	LK_L,
+	LK_M,
+	LK_N,
+	LK_O,
+	LK_P,
+	LK_Q,
+	LK_R,
+	LK_S,
+	LK_T,
+	LK_U,
+	LK_V,
+	LK_W,
+	LK_X,
+	LK_Y,
+	LK_Z,
+	LK_SPACE,
+	LK_SLASH,
+	LK_COMMA,
+	LK_PERIOD,
+	LK_SOLIDUS,
+	LK_SEMICOLON,
+	LK_SQUOTE,
+	LK_DQUOTE,
+	LK_S_BRACKET_L,
+	LK_S_BRACKET_R,
+	LK_R_BRACKET_L,
+	LK_R_BRACKET_R,
+	LK_DASH,
+	LK_EQUALS,
+	LK_ENTER,
+	LK_ESCAPE,
+	LK_UP,
+	LK_DOWN,
+	LK_LEFT,
+	LK_RIGHT,
+};
+
+#define KEYEV_FLAG_FREE 0
+#define KEYEV_FLAG_PRESS 1
+#define KEYEV_FLAG_RELEASE 2
+struct LumaKbdEvent {
+	uint8_t scancode;
+	uint8_t flag;
+} __attribute__((packed));
+
+#define KEYBUF_SIZE 4096
+#define KEYBUF_MAX (4096 / sizeof(struct LumaKbdEvent))
+
+#endif

luma/ps2/main.c

@@ -0,0 +1,254 @@
+#include"../sys.h"
+
+asm("modentry:\n"
+"mov $stack + 768, %esp\n"
+"call startc");
+
+#include"api.h"
+#include"../std.h"
+
+char stack[768];
+
+struct LumaKbdEvent *KeyBuf = (void*) -1;
+size_t KeyBufIndex = 0;
+
+void __attribute__((naked)) irq() {
+	asm(
+		"call irqc\n"
+		"movl $15, %eax\n"
+		"movl $1, %edi\n"
+		"int $0xEC"
+	);
+}
+
+static void pushev(struct LumaKbdEvent ev) {
+	if(KeyBuf[KeyBufIndex].flag != KEYEV_FLAG_FREE) {
+		return;
+	} else {
+		KeyBuf[KeyBufIndex++] = ev;
+		if(KeyBufIndex == KEYBUF_MAX) KeyBufIndex = 0;
+		sys_signal_send(KeyBuf, KeyBufIndex);
+	}
+}
+
+void irqc() {
+	static const uint8_t ss21blchars1[] = {LK_Q, LK_1, 0, 0, 0, LK_Z, LK_S, LK_A, LK_W, LK_2, 0, 0, LK_C, LK_X, LK_D, LK_E, LK_4, LK_3, 0, 0, LK_SPACE, LK_V, LK_F, LK_T, LK_R, LK_5, 0, 0, LK_N, LK_B, LK_H, LK_G, LK_Y, LK_6, 0, 0, 0, LK_M, LK_J, LK_U, LK_7, LK_8, 0, 0, LK_COMMA, LK_K, LK_I, LK_O, LK_0, LK_9, 0, 0, LK_PERIOD, LK_SOLIDUS, LK_L, LK_SEMICOLON, LK_P, LK_DASH, 0, 0, 0, 0, LK_SQUOTE, 0, LK_S_BRACKET_L, LK_EQUALS};
+	
+	static int stateLen = 0;
+	static uint8_t state[8]; // max 8 bytes per event
+	
+	state[stateLen++] = inb(0x60);
+	
+	if(stateLen == 8) { // Pause key, and state[0] should be 0xE1, but even if it isn't do this so as to reset state just in case
+		stateLen = 0;
+		pushev((struct LumaKbdEvent) {.scancode = LK_PAUSE, .flag = KEYEV_FLAG_PRESS});
+		pushev((struct LumaKbdEvent) {.scancode = LK_PAUSE, .flag = KEYEV_FLAG_RELEASE});
+	} else if(stateLen == 1 && state[0] != 0xE0 && state[0] != 0xE1 && state[0] != 0xF0) {
+		stateLen = 0;
+		struct LumaKbdEvent ev = {.flag = KEYEV_FLAG_PRESS};
+		if(state[0] >= 0x15 && state[0] <= 0x55) {
+			if((ev.scancode = ss21blchars1[state[0] - 0x15])) {
+				pushev(ev);
+			}
+		} else {
+			if(state[0] == 0x12) {
+				ev.scancode = LK_LEFT_SHIFT;
+			} else if(state[0] == 0x59) {
+				ev.scancode = LK_RIGHT_SHIFT;
+			} else if(state[0] == 0x66) {
+				ev.scancode = LK_BACKSPACE;
+			} else if(state[0] == 0x5A) {
+				ev.scancode = LK_ENTER;
+			} else if(state[0] == 0x11) {
+				ev.scancode = LK_LEFT_ALT;
+			} else if(state[0] == 0x76) {
+				ev.scancode = LK_ESCAPE;
+			}
+			pushev(ev);
+		}
+	} else if(stateLen == 2 && state[0] == 0xE0 && state[1] == 0x1F) {
+		stateLen = 0;
+		struct LumaKbdEvent ev = {.flag = KEYEV_FLAG_PRESS, .scancode = LK_LEFT_SYS};
+		pushev(ev);
+	} else if(stateLen == 2 && state[0] == 0xE0 && state[1] == 0x6B) {
+		stateLen = 0;
+		struct LumaKbdEvent ev = {.flag = KEYEV_FLAG_PRESS, .scancode = LK_LEFT};
+		pushev(ev);
+	} else if(stateLen == 2 && state[0] == 0xE0 && state[1] == 0x74) {
+		stateLen = 0;
+		struct LumaKbdEvent ev = {.flag = KEYEV_FLAG_PRESS, .scancode = LK_RIGHT};
+		pushev(ev);
+	} else if(stateLen == 2 && state[0] == 0xE0 && state[1] == 0x75) {
+		stateLen = 0;
+		struct LumaKbdEvent ev = {.flag = KEYEV_FLAG_PRESS, .scancode = LK_UP};
+		pushev(ev);
+	} else if(stateLen == 2 && state[0] == 0xE0 && state[1] == 0x72) {
+		stateLen = 0;
+		struct LumaKbdEvent ev = {.flag = KEYEV_FLAG_PRESS, .scancode = LK_DOWN};
+		pushev(ev);
+	} else if(stateLen == 2 && state[0] == 0xE0 && state[1] != 0xF0 && state[1] != 0x12) {
+		stateLen = 0;
+		struct LumaKbdEvent ev = {.flag = KEYEV_FLAG_PRESS};
+		pushev(ev);
+	} else if(stateLen == 4 && state[0] == 0xE0 && state[1] == 0x12 && state[2] == 0xE0 && state[3] == 0x7C) { // Print screen pressed
+		stateLen = 0;
+		struct LumaKbdEvent ev = {.flag = KEYEV_FLAG_PRESS};
+		pushev(ev);
+	} else if(stateLen == 6 && state[0] == 0xE0 && state[1] == 0xF0 && state[2] == 0x7C && state[3] == 0xE0 && state[4] == 0xF0 && state[5] == 0x12) { // Print screen released
+		stateLen = 0;
+		struct LumaKbdEvent ev = {.flag = KEYEV_FLAG_RELEASE};
+		pushev(ev);
+	} else if(stateLen == 2 && state[0] == 0xF0) {
+		stateLen = 0;
+		struct LumaKbdEvent ev = {.flag = KEYEV_FLAG_RELEASE};
+		if(state[1] >= 0x15 && state[1] <= 0x55) {
+			if((ev.scancode = ss21blchars1[state[1] - 0x15])) {
+				pushev(ev);
+			}
+		} else {
+			if(state[1] == 0x12) {
+				ev.scancode = LK_LEFT_SHIFT;
+			} else if(state[1] == 0x59) {
+				ev.scancode = LK_RIGHT_SHIFT;
+			} else if(state[1] == 0x66) {
+				ev.scancode = LK_BACKSPACE;
+			} else if(state[0] == 0x5A) {
+				ev.scancode = LK_ENTER;
+			} else if(state[1] == 0x11) {
+				ev.scancode = LK_LEFT_ALT;
+			} else if(state[1] == 0x76) {
+				ev.scancode = LK_ESCAPE;
+			}
+			pushev(ev);
+		}
+	} else if(stateLen == 3 && state[0] == 0xE0 && state[1] == 0xF0 && state[2] == 0x1F) {
+		stateLen = 0;
+		struct LumaKbdEvent ev = {.flag = KEYEV_FLAG_RELEASE, .scancode = LK_LEFT_SYS};
+		pushev(ev);
+	} else if(stateLen == 3 && state[0] == 0xE0 && state[1] == 0xF0 && state[2] == 0x6B) {
+		stateLen = 0;
+		struct LumaKbdEvent ev = {.flag = KEYEV_FLAG_RELEASE, .scancode = LK_LEFT};
+		pushev(ev);
+	} else if(stateLen == 3 && state[0] == 0xE0 && state[1] == 0xF0 && state[2] == 0x74) {
+		stateLen = 0;
+		struct LumaKbdEvent ev = {.flag = KEYEV_FLAG_RELEASE, .scancode = LK_RIGHT};
+		pushev(ev);
+	} else if(stateLen == 3 && state[0] == 0xE0 && state[1] == 0xF0 && state[2] == 0x75) {
+		stateLen = 0;
+		struct LumaKbdEvent ev = {.flag = KEYEV_FLAG_RELEASE, .scancode = LK_UP};
+		pushev(ev);
+	} else if(stateLen == 3 && state[0] == 0xE0 && state[1] == 0xF0 && state[2] == 0x72) {
+		stateLen = 0;
+		struct LumaKbdEvent ev = {.flag = KEYEV_FLAG_RELEASE, .scancode = LK_DOWN};
+		pushev(ev);
+	} else if(stateLen == 3 && state[0] == 0xE0 && state[1] == 0xF0 && state[2] != 0x7C) {
+		stateLen = 0;
+		struct LumaKbdEvent ev = {.flag = KEYEV_FLAG_RELEASE};
+		pushev(ev);
+	};
+}
+
+static volatile uint8_t *Framebuffer;
+static void dbg(int j, uint8_t b) {
+	for(int i = 0; i < 8; i++) {
+		Framebuffer[j * 9 + 7 - i] = (b & (1 << i)) ? 80 : 0;
+	}
+}
+
+void startc() {
+	Framebuffer = sys_vpm_map(SYS_MAP_VIRT_ANY, (void*) 0xA0000, 32000);
+	for(int try = 0; try < 50; try++) {
+		Framebuffer[try] = 50;
+		sys_sleep(500);
+		if(sys_link_create('USB\0', 'UHCI', 4096) != (void*) -1) {
+			goto done;
+		}
+	}
+	while(1);
+done:
+	while(inb(0x64) & 2);
+	outb(0x64, 0xAD);
+	while(inb(0x64) & 2);
+	outb(0x64, 0xA7);
+	inb(0x60);
+	while(inb(0x64) & 2);
+	outb(0x64, 0x20);
+	while((inb(0x64) & 1) == 0);
+	uint8_t c = inb(0x60);
+	while(inb(0x64) & 2);
+	outb(0x64, 0x60);
+	while(inb(0x64) & 2);
+	outb(0x60, c & ~67);
+	
+	/* Enable */
+	while(inb(0x64) & 2);
+	outb(0x64, 0xAE);
+	
+	while(inb(0x64) & 2);
+	outb(0x60, 0xFF);
+	while((inb(0x64) & 1) == 0);
+	if(inb(0x60) != 0xFA) while(1);
+	while((inb(0x64) & 1) == 0);
+	if(inb(0x60) != 0xAA) while(1);
+	
+	do {
+		while(inb(0x64) & 2);
+		outb(0x60, 0xF5);
+		while((inb(0x64) & 1) == 0);
+	} while(inb(0x60) != 0xFA);
+
+	do {
+		while(inb(0x64) & 2);
+		outb(0x60, 0xF2);
+		while((inb(0x64) & 1) == 0);
+	} while(inb(0x60) != 0xFA);
+
+	while((inb(0x64) & 1) == 0);
+
+	uint8_t b1 = inb(0x60);
+
+	sys_sleep(50);
+
+	if(inb(0x64) & 1) {
+		uint8_t b2 = inb(0x60);
+		if(b1 != 0xAB) {
+			for(int i = 0; i < 15; i++) {
+				Framebuffer[i] = 90;
+			}
+			while(1);
+		}
+	} else while(1);
+	
+	do {
+		KeyBuf = sys_link_create('LUMA', 'WM\0\0', 4096);
+	} while(KeyBuf == (void*) -1);
+	memory_barrier();
+	std_w8(KeyBuf, 0, 4096);
+	
+	// Send type "input device"
+	sys_signal_send(KeyBuf, 0);
+	sys_signal_wait(KeyBuf, NULL);
+	
+	sys_irq_claim(1, &irq);
+	
+	while(inb(0x64) & 2);
+	outb(0x64, 0x20);
+	while((inb(0x64) & 1) == 0);
+	c = inb(0x60);
+	while(inb(0x64) & 2);
+	outb(0x64, 0x60);
+	while(inb(0x64) & 2);
+	outb(0x60, c | 1); // enable interrupt
+	
+	inb(0x60);
+	
+	do {
+		while(inb(0x64) & 2);
+		outb(0x60, 0xF4);
+		while((inb(0x64) & 1) == 0);
+	} while(inb(0x60) != 0xFA);
+	
+	sys_wait_for_irq(1);
+	
+	while(1) sys_signal_wait(SYS_SIGNAL_WAIT_ANY, NULL);
+}

luma/std.h

@@ -0,0 +1,79 @@
+#ifndef LUMA_STD_H
+#define LUMA_STD_H
+
+#include<stddef.h>
+#include<stdint.h>
+
+#if defined(__has_builtin) && __has_builtin(__builtin_memcpy)
+#define std_copy __builtin_memcpy
+#else
+static inline void std_copy(void *dst, const void *src, size_t len) {
+	char *d = dst;
+	const char *s = src;
+	while(len--) {
+		*(d++) = *(s++);
+	}
+}
+#endif
+
+static inline void std_move(void *dst, const void *src, size_t len) {
+	if(src > dst) {
+		char *d = dst;
+		const char *s = src;
+		while(len--) {
+			*(d++) = *(s++);
+		}
+	} else {
+		char *d = dst + len;
+		const char *s = src + len;
+		while(len--) {
+			*(--d) = *(--s);
+		}
+	}
+}
+
+static inline void std_w8(uint8_t *dst, uint8_t b, size_t len) {
+	while(len--) {
+		*dst = b;
+		dst++;
+	}
+}
+
+typedef struct {
+	uint16_t len;
+	uint8_t data[];
+} Str16;
+#define WALK_STR16(x) ((Str16*)((uintptr_t)(x)+sizeof(Str16)+((Str16*)(x))->len))
+
+typedef struct {
+	uint16_t cap;
+	uint16_t len;
+	uint8_t data[];
+} DynStr16;
+#define DYN16_TO_ST(x) ((Str16*)((uintptr_t) x + offsetof(DynStr16, len)))
+
+static inline char* std_bytefind(size_t len, const char *data, const char what) {
+	while(len--) {
+		if(*data == what) {
+			return (char*) data;
+		}
+		data++;
+	}
+	return NULL;
+}
+
+static inline unsigned char std_bytecomp(size_t len, const void *one_, const void *two_) {
+	const char *one = one_, *two = two_;
+	while(len--) {
+		unsigned char asdf = *(one++) - *(two++);
+		if(asdf) return asdf;
+	}
+	return 0;
+}
+
+#define std_min(a, b) ({__typeof__(a) _a = (a); __typeof__(b) _b = (b); _a<_b?_a:_b;})
+#define std_max(a, b) ({__typeof__(a) _a = (a); __typeof__(b) _b = (b); _a>_b?_a:_b;})
+
+//#define S16_ARG(str) ((Str16*) &(struct {uint16_t len; char content[sizeof(str) - 1]}) {sizeof(str) - 1, (str)})
+
+#endif

luma/sys.h

@@ -0,0 +1,185 @@
+#ifndef _LUMA_SYS_H
+#define _LUMA_SYS_H
+
+#include<stdint.h>
+#include<stddef.h>
+
+#define memory_barrier() asm volatile("" : : : "memory")
+#define STRINGIFY2(v) #v
+#define STRINGIFY(v) STRINGIFY2(v)
+
+static __attribute__((always_inline)) inline uint32_t syscall1(uint32_t eax, uint32_t edi, uint32_t esi, uint32_t edx) {
+	uint32_t ret;
+	asm volatile("int $0xEC" : "=a"(ret) : "a"(eax), "S"(esi), "D"(edi), "d"(edx));
+	return ret;
+}
+
+static __attribute__((always_inline)) inline uint32_t sys5call1(uint32_t eax, uint32_t edi, uint32_t esi, uint32_t edx, uint32_t ecx, uint32_t ebx) {
+	uint32_t ret;
+	asm volatile("int $0xEC" : "=a"(ret) : "a"(eax), "S"(esi), "D"(edi), "d"(edx), "c"(ecx), "b"(ebx));
+	return ret;
+}
+
+#define SYS_MAP_PHYS_ANY ((void*)-1)
+#define SYS_MAP_PHYS_CONTIGUOUS ((void*)-2)
+#define SYS_MAP_VIRT_ANY ((void*)-1)
+static __attribute__((always_inline)) inline void *sys_vpm_map(void *virt, void *phys, size_t length) {
+	return (void*) syscall1(0, (uint32_t) virt, (uint32_t) phys, length);
+}
+
+static __attribute__((always_inline)) inline uint32_t sys_canal_create(uint32_t nameLow, uint32_t nameHigh) {
+	return syscall1(1, nameLow, nameHigh, 0);
+}
+
+static __attribute__((always_inline)) inline void *sys_link_create(uint32_t nameLow, uint32_t nameHigh, size_t areaSize) {
+	return (void*) syscall1(2, nameLow, nameHigh, areaSize);
+}
+
+#define SYS_CANAL_ACCEPT_ANY ((uint32_t)-1)
+static __attribute__((always_inline)) inline void *sys_canal_accept(uint32_t canalId) {
+	return (void*) syscall1(3, canalId, 0, 0);
+}
+
+static __attribute__((always_inline)) inline void sys_signal_send(void *area, uint32_t meta) {
+	syscall1(4, (uint32_t) area, meta, 0);
+}
+
+#define SYS_SIGNAL_WAIT_ANY ((void*)-1)
+static __attribute__((always_inline)) inline void *sys_signal_wait(void *area, uint32_t *meta) {
+	uint32_t eax, edx;
+	asm volatile("int $0xEC" : "=a"(eax), "=d"(edx) : "a"(5), "D"(area) : "ebx");
+	if(meta) *meta = edx;
+	return (void*) eax;
+}
+static __attribute__((always_inline)) inline void *sys_signal_wait3(void *area, uint32_t *meta, uint32_t *canalId) {
+	uint32_t eax, edx, ebx;
+	asm volatile("int $0xEC" : "=a"(eax), "=d"(edx), "=b"(ebx) : "a"(5), "D"(area));
+	if(meta) *meta = edx;
+	if(canalId) *canalId = ebx;
+	return (void*) eax;
+}
+
+static __attribute__((always_inline)) inline uint32_t sys_pci_claim(uint32_t cscp) {
+	return syscall1(6, cscp, 0, 0);
+}
+
+static __attribute__((always_inline)) inline uint32_t sys_pci_read(uint32_t id, uint8_t offset) {
+	return syscall1(7, id, offset, 0);
+}
+
+static __attribute__((always_inline)) inline void sys_vpm_unmap(void *virt, size_t length) {
+	syscall1(8, (uint32_t) virt, length, 0);
+}
+
+static __attribute__((always_inline)) inline void *sys_vpm_get_phys(void *virt) {
+	return (void*) syscall1(9, (uint32_t) virt, 0, 0);
+}
+
+static __attribute__((always_inline)) inline void sys_sleep(uint32_t millis) {
+	syscall1(10, millis, 0, 0);
+}
+
+static __attribute__((always_inline)) inline void sys_pci_write_1(uint32_t id, uint8_t offset, uint8_t val) {
+	syscall1(11, id, offset, val);
+}
+
+static __attribute__((always_inline)) inline void sys_pci_write_2(uint32_t id, uint8_t offset, uint16_t val) {
+	syscall1(12, id, offset, val);
+}
+
+static __attribute__((always_inline)) inline void sys_pci_write_4(uint32_t id, uint8_t offset, uint32_t val) {
+	syscall1(13, id, offset, val);
+}
+
+static __attribute__((always_inline)) inline void sys_irq_claim(uint32_t irq, void(*handler)()) {
+	syscall1(14, irq, (uintptr_t) handler, 0);
+}
+
+static __attribute__((always_inline)) inline void sys_wait_for_irq(uint32_t i) {
+	syscall1(15, i, 0, 0);
+}
+
+static __attribute__((always_inline)) inline void sys_create_thread(void *cod) {
+	syscall1(16, (uintptr_t) cod, 0, 0);
+}
+
+static __attribute__((always_inline)) inline size_t sys_spawn() {
+	return syscall1(17, 0, 0, 0);
+}
+
+static __attribute__((always_inline)) inline uintptr_t sys_abuse_map(size_t handle, void **virtMy, void **virtTheir, void *phys, size_t length) {
+	memory_barrier();
+	return sys5call1(18, handle, (uintptr_t) virtMy, (uintptr_t) virtTheir, (uintptr_t) phys, length);
+}
+
+static __attribute__((always_inline)) inline int sys_release(size_t handle) {
+	return syscall1(19, handle, 0, 0);
+}
+
+#define SYS_STATE_X86_EIP 0
+static __attribute__((always_inline)) inline int sys_abuse_state(size_t handle, uintmax_t type, uintmax_t val) {
+	return syscall1(20, handle, type, (uintptr_t) val);
+}
+
+static __attribute__((always_inline)) inline int sys_exit() {
+	return syscall1(21, 0, 0, 0);
+}
+
+#define KEV_TYPE_NONE 0
+#define KEV_TYPE_LINK_REQUEST 1
+#define KEV_TYPE_LINK_SIGNAL 2
+#define KEV_TYPE_LINK_CLOSE 3
+typedef struct __attribute__((packed)) {
+	uint8_t type;
+	uint32_t canalId;
+	void *area;
+	uint32_t meta;
+} KEvent;
+
+static __attribute__((always_inline)) inline int sys_event_wait(KEvent *kev) {
+	int ret = syscall1(22, (uintptr_t) (void*) kev, 0, 0);
+	memory_barrier();
+	return ret;
+}
+
+static __attribute__((always_inline)) inline uint32_t ind(uint32_t p) {
+	uint32_t ret;
+	asm volatile("inl %%dx, %%eax" : "=a"(ret) : "d"(p));
+	return ret;
+}
+
+static __attribute__((always_inline)) inline uint16_t inw(uint32_t p) {
+	uint16_t ret;
+	asm volatile("inw %%dx, %%ax" : "=a"(ret) : "d"(p));
+	return ret;
+}
+
+static __attribute__((always_inline)) inline uint8_t inb(uint32_t p) {
+	uint8_t ret;
+	asm volatile("inb %%dx, %%al" : "=a"(ret) : "d"(p));
+	return ret;
+}
+
+static __attribute__((always_inline)) inline void outb(uint32_t p, uint8_t v) {
+	asm volatile("outb %%al, %%dx" : : "d"(p), "a"(v));
+}
+
+static __attribute__((always_inline)) inline void outw(uint32_t p, uint16_t v) {
+	asm volatile("outw %%ax, %%dx" : : "d"(p), "a"(v));
+}
+
+static __attribute__((always_inline)) inline void outd(uint32_t p, uint32_t v) {
+	asm volatile("outl %%eax, %%dx" : : "d"(p), "a"(v));
+}
+
+#if __BYTE_ORDER == __LITTLE_ENDIAN
+#define BIG_TO_HOST_ENDIAN32(x) __builtin_bswap32(x)
+#define BIG_TO_HOST_ENDIAN16(x) __builtin_bswap16(x)
+#elif __BYTE_ORDER == __BIG_ENDIAN
+#define BIG_TO_HOST_ENDIAN32(x) (x)
+#define BIG_TO_HOST_ENDIAN16(x) (x)
+#else
+#error "PDP??"
+#endif
+
+#endif

luma/uhci/Makefile

@@ -0,0 +1,13 @@
+rwildcard=$(foreach d,$(wildcard $(1:=/*)),$(call rwildcard,$d,$2) $(filter $(subst *,%,$2),$d))
+
+HEADERS := $(call rwildcard,.,*.h)
+SOURCEC := $(call rwildcard,.,*.c)
+OBJECTS := $(patsubst %.c, ../../build/uhci/%.o, $(SOURCEC))
+
+mod: $(OBJECTS)
+	i386-elf-gcc -pie -T ../../mod.ld -Wl,-no-dynamic-linker -nostdlib -o main.elf $(OBJECTS) -lgcc
+	../../elftomod.py main.elf ../../bootfs/luma_uhci.mod modentry:0
+
+../../build/uhci/%.o: %.c
+	mkdir -p $(@D)
+	i386-elf-gcc -march=i386 -Isrc -Os -pie -nodefaultlibs -nostdlib -nostartfiles -fomit-frame-pointer -c -o $@ -ffreestanding -mgeneral-regs-only -std=gnu99 -Wall $<

luma/uhci/bbb.c

@@ -0,0 +1,293 @@
+#include"bbb.h"
+
+#include<stddef.h>
+#include"host.h"
+#include"../sys.h"
+#include"../std.h"
+
+#define SCSI_OP_INQUIRY 0x12
+#define SCSI_OP_TEST_UNIT_READY 0x00
+#define SCSI_OP_REQUEST_SENSE 0x03
+#define SCSI_OP_READ_CAPACITY_10 0x25
+#define SCSI_OP_READ_10 0x28
+#define SCSI_OP_MODE_SENSE_6 0x1A
+
+typedef struct __attribute__((packed)) {
+	uint8_t op;
+	union __attribute__((packed)) {
+		struct __attribute__((packed)) {
+			uint8_t lun5;
+			uint8_t page;
+			uint8_t reserved;
+			uint8_t dataSz;
+			uint8_t control;
+		} inquiry;
+		struct __attribute__((packed)) {
+			uint8_t res0;
+			uint8_t res1;
+			uint8_t res2;
+			uint8_t res3;
+			uint8_t control;
+		} testUnitReady;
+		struct __attribute__((packed)) {
+			uint8_t desc;
+			uint8_t res0;
+			uint8_t res1;
+			uint8_t dataSz;
+			uint8_t control;
+		} requestSense;
+		struct __attribute__((packed)) {
+			uint8_t res0;
+			uint8_t res1;
+			uint8_t res2;
+			uint8_t res3;
+			uint8_t res4;
+			uint8_t res5;
+			uint8_t res6;
+			uint8_t res7;
+			uint8_t control;
+		} readCapacity10;
+		struct __attribute__((packed)) {
+			uint8_t flags;
+			uint32_t lba;
+			uint8_t group;
+			uint16_t lbaCount;
+			uint8_t control;
+		} read10;
+		struct __attribute__((packed)) {
+			uint8_t flags;
+			uint8_t pageCode;
+			uint8_t subpageCode;
+			uint8_t allocationSz;
+			uint8_t control;
+		} modeSense6;
+	};
+} SCSICommand;
+
+#define CBW_SIGN 0x43425355
+#define CBW_FLAG_IN 0x80
+#define CBW_FLAG_OUT 0x00
+typedef struct __attribute__((packed)) {
+	uint32_t signature;
+	uint32_t tag;
+	uint32_t dataSz;
+	uint8_t flags;
+	uint8_t lun;
+	uint8_t cmdSz;
+	union {
+		SCSICommand cmd;
+		uint8_t padding[16];
+	};
+} CBW;
+
+#define CSW_SIGN 0x53425355
+typedef struct __attribute__((packed)) {
+	uint32_t signature;
+	uint32_t tag;
+	uint32_t residue;
+	uint8_t status;
+} CSW;
+
+void dbg(int j, uint8_t b);
+extern uint32_t asdfasdf;
+
+static int BBBRequest(USBDevice *dev, CBW cbw, void *buf, size_t len, CSW *ret) {
+	if(HostSendBulk(dev, dev->bbb.maxPacketSize, &dev->bbb.toggleOut, dev->bbb.endpOut, &cbw, sizeof(cbw)) == 0) {
+		if(BBBReset(dev) == 0) {
+			return 0;
+		}
+		if(HostSendBulk(dev, dev->bbb.maxPacketSize, &dev->bbb.toggleOut, dev->bbb.endpOut, &cbw, sizeof(cbw)) == 0) {
+			return 0;
+		}
+	}
+	
+	sys_sleep(10);
+
+	if(len) {
+		HostSendBulk(dev, dev->bbb.maxPacketSize,
+			(cbw.flags & CBW_FLAG_IN) ? &dev->bbb.toggleIn : &dev->bbb.toggleOut,
+			(cbw.flags & CBW_FLAG_IN) ? dev->bbb.endpIn : dev->bbb.endpOut,
+			buf, len);
+	}
+	
+	return HostSendBulk(dev, dev->bbb.maxPacketSize, &dev->bbb.toggleIn, dev->bbb.endpIn, ret, sizeof(*ret));
+}
+
+int BBBReset(USBDevice *dev) {
+	if(!HostSendRequest(dev, NULL, (USBStandardRequest) {
+		.requestType = 0x21,
+		.request = 0xFF, // mass storage reset
+		.value = 0,
+		.index = 0,
+		.length = 0
+	})) return 0;
+	sys_sleep(150);
+	if(!HostSendRequest(dev, NULL, (USBStandardRequest) {
+		.requestType = 2, // endpoint
+		.request = 1, // clear feature
+		.value = 0, // halt
+		.index = dev->bbb.endpIn,
+		.length = 0
+	})) return 0;
+	sys_sleep(150);
+	if(!HostSendRequest(dev, NULL, (USBStandardRequest) {
+		.requestType = 2, // endpoint
+		.request = 1, // clear feature
+		.value = 0, // halt
+		.index = dev->bbb.endpOut,
+		.length = 0
+	})) return 0;
+	sys_sleep(150);
+	dev->bbb.toggleIn = dev->bbb.toggleOut = 0;
+	return 1;
+}
+
+int BBBInquiry(USBDevice *dev) {
+	uint8_t inquiry[36];
+	
+	CSW csw;
+	
+	return BBBRequest(dev, (CBW) {
+		.signature = CBW_SIGN,
+		.tag = 0,
+		.dataSz = 36,
+		.flags = CBW_FLAG_IN,
+		.lun = 0,
+		.cmdSz = 6,
+		.cmd = {
+			.op = SCSI_OP_INQUIRY,
+			.inquiry = {
+				.lun5 = 0,
+				.page = 0,
+				.reserved = 0,
+				.dataSz = 36,
+				.control = 0
+			}
+		}
+	}, inquiry, 36, &csw) && csw.status == 0;
+}
+
+int BBBTestUnitReady(USBDevice *dev) {
+	CSW csw;
+	
+	return BBBRequest(dev, (CBW) {
+		.signature = CBW_SIGN,
+		.tag = 0,
+		.dataSz = 0,
+		.flags = CBW_FLAG_OUT,
+		.lun = 0,
+		.cmdSz = 6,
+		.cmd = {
+			.op = SCSI_OP_TEST_UNIT_READY,
+			.testUnitReady = {}
+		}
+	}, NULL, 0, &csw) && csw.status == 0;
+}
+
+int BBBRequestSense(USBDevice *dev) {
+	uint8_t data[18];
+	
+	CSW csw;
+	
+	return BBBRequest(dev, (CBW) {
+		.signature = CBW_SIGN,
+		.tag = 0,
+		.dataSz = 18,
+		.flags = CBW_FLAG_IN,
+		.lun = 0,
+		.cmdSz = 6,
+		.cmd = {
+			.op = SCSI_OP_REQUEST_SENSE,
+			.requestSense = {.dataSz = 18}
+		}
+	}, data, 18, &csw) && csw.status == 0;
+}
+
+int BBBReadCapacity10(USBDevice *dev) {
+	uint32_t data[2];
+	CSW csw;
+	if(!BBBRequest(dev, (CBW) {
+		.signature = CBW_SIGN,
+		.tag = 0,
+		.dataSz = sizeof(data),
+		.flags = CBW_FLAG_IN,
+		.lun = 0,
+		.cmdSz = 10,
+		.cmd = {
+			.op = SCSI_OP_READ_CAPACITY_10,
+			.readCapacity10 = {}
+		}
+	}, data, sizeof(data), &csw)) return 0;
+	
+	if(csw.status) return 0;
+	
+	dev->bbb.sectorSize = BIG_TO_HOST_ENDIAN32(data[1]);
+	
+	//dbg(asdfasdf + 2, csw.status);
+	//dbg(asdfasdf + 3, dev->bbb.maxPacketSize);
+	//dbg(asdfasdf + 4, dev->bbb.sectorSize >> 9);
+	
+	return 1;
+}
+
+int BBBRead10(USBDevice *dev, uint32_t lbaStart, void *buf, uint16_t lbaCount) {
+	CSW csw;
+	if(!BBBRequest(dev, (CBW) {
+		.signature = CBW_SIGN,
+		.tag = 0,
+		.dataSz = lbaCount * dev->bbb.sectorSize,
+		.flags = CBW_FLAG_IN,
+		.lun = 0,
+		.cmdSz = 10,
+		.cmd = {
+			.op = SCSI_OP_READ_10,
+			.read10 = {.lba = BIG_TO_HOST_ENDIAN32(lbaStart), .lbaCount = BIG_TO_HOST_ENDIAN16(lbaCount)}
+		}
+	}, buf, lbaCount * dev->bbb.sectorSize, &csw)) return 0;
+	
+	return csw.status == 0;
+}
+
+int BBBDeviceResponse(USBDevice *dev, void *area, uint32_t meta) {
+	uint64_t start = ((uint64_t*) area)[0];
+	uint32_t length = ((uint32_t*) area)[2];
+	
+	uint32_t startLBA = start / dev->bbb.sectorSize;
+	uint32_t endLBA = (start + length - 1) / dev->bbb.sectorSize;
+	
+	dbg(asdfasdf + 10, startLBA & 0xFF);
+	dbg(asdfasdf + 11, endLBA & 0xFF);
+	
+	if(!BBBRead10(dev, startLBA, area, endLBA - startLBA + 1)) {
+		dbg(asdfasdf + 12, 0xFF);
+		sys_signal_send(area, 0);
+	} else {
+		dbg(asdfasdf + 12, 0x00);
+		std_move(&area[0], &area[start % dev->bbb.sectorSize], length);
+		sys_signal_send(area, 1);
+	}
+	
+	return 1;
+}
+
+int BBBModeSense(USBDevice *dev) {
+	uint8_t data[192];
+	
+	CSW csw;
+	
+	return BBBRequest(dev, (CBW) {
+		.signature = CBW_SIGN,
+		.tag = 0,
+		.dataSz = 192,
+		.flags = CBW_FLAG_IN,
+		.lun = 0,
+		.cmdSz = 6,
+		.cmd = {
+			.op = SCSI_OP_MODE_SENSE_6,
+			.modeSense6 = {
+				.pageCode = 0x3F,
+				.allocationSz = 192
+			}
+		}
+	}, data, 192, &csw) && csw.status == 0;
+}

luma/uhci/bbb.h

@@ -0,0 +1,12 @@
+#pragma once
+
+#include"usb.h"
+
+int BBBReset(USBDevice *dev);
+int BBBInquiry(USBDevice *dev);
+int BBBTestUnitReady(USBDevice *dev);
+int BBBRequestSense(USBDevice *dev);
+int BBBReadCapacity10(USBDevice *dev);
+int BBBRead10(USBDevice *dev, uint32_t lbaStart, void *buf, uint16_t lbaCount);
+int BBBDeviceResponse(USBDevice *dev, void *area, uint32_t meta);
+int BBBModeSense(USBDevice *dev);

luma/uhci/host.h

@@ -0,0 +1,26 @@
+#pragma once
+
+#include"usb.h"
+#include<stddef.h>
+#include<stdint.h>
+
+typedef struct {
+	uint32_t link;
+	uint32_t control;
+	uint32_t token;
+	uint32_t buffer;
+} TD;
+
+typedef struct {
+	uint32_t link;
+	uint32_t element;
+} QH;
+
+typedef struct {
+	QH masterQueue;
+	uint32_t io;
+	uint32_t *frameList;
+} __attribute__((aligned(16))) HC;
+
+int HostSendBulk(USBDevice *dev, size_t maxPacketSize, int *toggle, int endpoint, void *buf, size_t len);
+int HostSendRequest(USBDevice *dev, void *buf, USBStandardRequest req);

luma/uhci/main.c

@@ -0,0 +1,306 @@
+#include"../sys.h"
+
+// Currently, the driver waits for any requests and performs them
+// This method fails for active USB devices (rather than passive)
+//
+// Should either poll for signals or switch to multithreading, once
+// it becomes a need.
+
+asm("modentry:\n"
+"mov $stack + 4096, %esp\n"
+"call startc");
+
+#include"host.h"
+
+#define TD_ACTIVE (1 << 23)
+#define TD_MAXERRS(x) ((x) << 27)
+#define TD_TOGGLE(x) ((!!(x)) << 19)
+#define TD_MAXLEN(sz) (((sz) - 1) << 21)
+#define TD_MAXLEN_0 (0x7FF << 21)
+#define TD_DEVADDR(i) ((i) << 8)
+#define TD_DEVENDP(i) ((i) << 15)
+#define TD_LSD (1 << 26)
+#define TD_SPD (1 << 29)
+#define TD_IOC (1 << 24)
+#define TD_DEPTH_FIRST (4)
+
+char __attribute__((aligned(16))) stack[4096];
+
+size_t hcCount = 0;
+HC __attribute__((aligned(16))) hcs[4];
+
+size_t deviceCount = 0;
+USBDevice *devices;
+
+// FOR DEBUG PURPOSES {
+static volatile uint8_t *Framebuffer;
+uint32_t asdfasdf = 28;
+void dbg(int j, uint8_t b) {
+	for(int i = 0; i < 8; i++) {
+		Framebuffer[j * 9 + 10 - i] = (b & (1 << i)) ? 85 : 40;
+	}
+	Framebuffer[j * 9 + 11] = 0;
+}
+// }
+
+int HostSendBulk(USBDevice *dev, size_t maxPacketSize, int *toggle, int endpoint, void *buf, size_t len) {
+	size_t sz = len;
+	int num = (sz + maxPacketSize - 1) / maxPacketSize;
+	
+	TD tds[num] __attribute__((aligned(16)));
+	
+	for(int i = 0;; i++, sz -= maxPacketSize) {
+		tds[i] = (TD) {
+			.link = (i == num - 1 ? 1 : ((uintptr_t) sys_vpm_get_phys(&tds[i + 1]))) | TD_DEPTH_FIRST,
+			.control = TD_ACTIVE | TD_MAXERRS(3) | ((endpoint & 128) ? TD_SPD : 0) | (i == num - 1 ? TD_IOC : 0),
+			.token = TD_MAXLEN(sz > maxPacketSize ? maxPacketSize : sz) | TD_TOGGLE(*toggle) | TD_DEVADDR(dev->id) | TD_DEVENDP(endpoint & 0x0F) | ((endpoint & 128) ? IN : OUT),
+			.buffer = (uintptr_t) sys_vpm_get_phys(&buf[i * maxPacketSize])
+		};
+		*toggle ^= 1;
+		if(sz <= maxPacketSize) break;
+	}
+	
+	while(inw(hcs[dev->hc].io + 2) & 3) {
+		outw(hcs[dev->hc].io + 2, 3); // clear usbint
+	}
+	
+	QH *masterQ = &hcs[dev->hc].masterQueue;
+	
+	QH qh __attribute__((aligned(16)));
+	qh.link = 1;
+	
+	memory_barrier();
+	
+	qh.element = (uint32_t) sys_vpm_get_phys(&tds[0]);
+	masterQ->element = ((uint32_t) sys_vpm_get_phys(&qh)) | 2;
+	
+	// Wait until complete.
+	// Polling doesn't seem to work reliably. Sometimes USBINT just isn't set.
+	// From iPXE source, though can't find a relation:
+	/* UHCI defers interrupts (including short packet detection)
+	* until the end of the frame.  This can result in bulk IN
+	* endpoints remaining halted for much of the time, waiting
+	* for software action to reset the data toggles.  We
+	* therefore ignore USBSTS and unconditionally poll all
+	* endpoints for completed transfer descriptors.
+	*
+	* As with EHCI, we trust that completion handlers are minimal
+	* and will not do anything that could plausibly affect the
+	* endpoint list itself.
+	*/
+	for(int i = 0; i < num; i++) {
+		while(tds[i].control & TD_ACTIVE) memory_barrier();
+		
+		if(tds[i].control & 0x00FF0000) {
+			return 0; //fail
+		}
+	}
+	
+	return 1;
+}
+
+/* buf must be physically contiguous */
+int HostSendRequest(USBDevice *dev, void *buf, USBStandardRequest req) {
+	size_t sz = req.length;
+	int num = 2 + (sz + dev->maxPacketSize - 1) / dev->maxPacketSize;
+	
+	TD tds[num] __attribute__((aligned(16)));
+	
+	tds[0] = (TD) {
+		.link = (sz == 0 ? 1 : (uintptr_t) sys_vpm_get_phys(&tds[1])) | TD_DEPTH_FIRST,
+		.control = TD_ACTIVE | TD_MAXERRS(3) | (sz == 0 ? TD_IOC : 0),
+		.token = TD_MAXLEN(8) | TD_TOGGLE(0) | TD_DEVADDR(dev->id) | SETUP,
+		.buffer = (uintptr_t) sys_vpm_get_phys(&req)
+	};
+	
+	for(int i = 1;; i++, sz -= dev->maxPacketSize) {
+		tds[i] = (TD) {
+			.link = (i == num - 2 ? 1 : ((uintptr_t) sys_vpm_get_phys(&tds[i + 1]))) | TD_DEPTH_FIRST,
+			.control = TD_ACTIVE | TD_MAXERRS(3) | ((req.requestType & 128) ? TD_SPD : 0) | (i == num - 2 ? TD_IOC : 0),
+			.token = TD_MAXLEN(sz > dev->maxPacketSize ? dev->maxPacketSize : sz) | TD_TOGGLE(i & 1) | TD_DEVADDR(dev->id) | ((req.requestType & 128) ? IN : OUT),
+			.buffer = (uintptr_t) sys_vpm_get_phys(&buf[(i - 1) * dev->maxPacketSize])
+		};
+		if(sz <= dev->maxPacketSize) break;
+	}
+	
+	tds[num - 1] = (TD) {
+		.link = 1,
+		.control = TD_ACTIVE | TD_MAXERRS(3),
+		.token = TD_MAXLEN_0 | TD_TOGGLE(1) | TD_DEVADDR(dev->id) | ((req.requestType & 128) ? OUT : IN),
+		.buffer = 0
+	};
+	
+	while(inw(hcs[dev->hc].io + 2) & 1) {
+		outw(hcs[dev->hc].io + 2, 3); // clear usbint
+	}
+	
+	QH *masterQ = &hcs[dev->hc].masterQueue;
+	
+	QH qh __attribute__((aligned(16)));
+	qh.link = 1;
+	
+	memory_barrier();
+	
+	qh.element = (uint32_t) sys_vpm_get_phys(&tds[0]);
+	masterQ->element = ((uint32_t) sys_vpm_get_phys(&qh)) | 2;
+	
+	sys_sleep(15);
+	
+	// Wait for usbint
+	memory_barrier();
+	while((inw(hcs[dev->hc].io + 2) & 1) == 0);
+	
+	outw(hcs[dev->hc].io + 2, 3); // clear usbint
+	
+	qh.element = (uint32_t) sys_vpm_get_phys(&tds[num - 1]);
+	masterQ->element = ((uint32_t) sys_vpm_get_phys(&qh)) | 2;
+	
+	// Wait until status packet becomes inactive
+	do {
+		memory_barrier();
+	} while(tds[num - 1].control & TD_ACTIVE);
+	
+	// Return whether success
+	return (tds[num - 1].control & (0xFF0000)) == 0;
+}
+
+static int HCsInit() {
+	// Go through all EHCIs and disable them
+	while(1) {
+		uint32_t hcId = sys_pci_claim(0x0C032000);
+		
+		if(hcId == -1) {
+			break;
+		}
+		
+		volatile uint32_t *usbbase = sys_vpm_map(SYS_MAP_VIRT_ANY, (void*) (sys_pci_read(hcId, 16) & ~15), 4096);
+		volatile uint32_t *opregs = (uint32_t*) ((uintptr_t) usbbase + (usbbase[0] & 0xFF));
+		
+		uint8_t caps = (usbbase[2] & 0xFF00) >> 8;
+		if(caps < 0x40) {
+			return 0;
+		} else {
+			sys_pci_write_1(hcId, caps + 3, 1);
+			while((sys_pci_read(hcId, caps) & (1 << 16)) != 0);
+			while((sys_pci_read(hcId, caps) & (1 << 24)) == 0);
+			sys_pci_write_4(hcId, caps + 4, 0);
+		}
+		
+		opregs[0] &= ~1;
+		
+		do {
+			memory_barrier();
+		} while((opregs[1] & 4096) == 0);
+		
+		opregs[0] |= 2;
+		
+		do {
+			memory_barrier();
+		} while(opregs[0] & 2);
+		
+		opregs[1] = opregs[1];
+		opregs[2] = 0;
+		opregs[4] = 0;
+		opregs[0] = 0x80000;
+		opregs[16] = 0;
+		
+		sys_vpm_unmap(usbbase, 4096);
+	}
+	
+	int success = 1;
+	
+	// Go through all UHCIs
+	while(1) {
+		uint32_t hcId = sys_pci_claim(0x0C030000);
+		
+		if(hcId == -1) {
+			break;
+		}
+		
+		HC *hc = &hcs[hcCount];
+		hc->io = sys_pci_read(hcId, 0x20) & ~3;
+		hc->frameList = sys_vpm_map(SYS_MAP_VIRT_ANY, SYS_MAP_PHYS_ANY, 4096);
+		hc->masterQueue = (QH) {1, 1};
+		for(int i = 0; i < 1024; i++) {
+			hc->frameList[i] = ((uint32_t) sys_vpm_get_phys(&hc->masterQueue)) | 2;
+		}
+		
+		sys_pci_write_2(hcId, 0xC0, 0x8F00);
+		
+		outw(hc->io + 0, 2);
+		while(inw(hc->io + 2) & 2);
+		
+		outw(hc->io + 4, 0);
+		outw(hc->io + 0, 0);
+		outw(hc->io + 2, 0xFFFF);
+		outb(hc->io + 12, 64);
+		outd(hc->io + 8, (uint32_t) sys_vpm_get_phys(hc->frameList));
+		outw(hc->io + 6, 0);
+		outw(hc->io + 0, 1 | 128);
+		
+		for(int p = 0; p < 2; p++) {
+			outw(hc->io + 16 + p+p, 512);
+			sys_sleep(80);
+			outw(hc->io + 16 + p+p, 0);
+			
+			for(int try = 0; try < 10; try++) {
+				sys_sleep(10);
+				uint16_t sc = inw(hc->io + 16 + p+p);
+				//if((sc & 1) == 0) break;
+				if(sc & (2 | 8)) outw(hc->io + 16 + p+p, sc);
+				if(sc & 4) break;
+				outw(hc->io + 16 + p+p, sc | 4);
+			}
+			
+			sys_sleep(100);
+			
+			if(inw(hc->io + 16 + p+p) & 1) {
+				dbg(asdfasdf++, 2 * hcCount + p + 128);
+				
+				USBDevice *dev = &devices[deviceCount];
+				dev->hc = hcCount;
+				dev->id = deviceCount + 1;
+				dev->maxPacketSize = 8;
+				if(USBDeviceDiscover(dev) == 0) { /* The rest of the fields are filled here */
+					success = 0;
+				}
+				deviceCount++;
+			}
+		}
+		
+		hcCount++;
+	}
+	
+	return success;
+}
+
+void startc() {
+	devices = sys_vpm_map(SYS_MAP_VIRT_ANY, SYS_MAP_PHYS_ANY, 4096);
+	
+	Framebuffer = sys_vpm_map(SYS_MAP_VIRT_ANY, (void*) 0xA0000, 32000);
+	
+	if(HCsInit()) {
+		// Let other drivers know we exist
+		sys_canal_create('USB\0', 'UHCI');
+		
+		// Main loop
+		while(1) {
+			sys_canal_accept(SYS_CANAL_ACCEPT_ANY);
+			
+			uint32_t meta, canal;
+			void* area = sys_signal_wait3(SYS_SIGNAL_WAIT_ANY, &meta, &canal);
+			for(int i = 0; i < deviceCount; i++) {
+				if(devices[i].canal == canal) {
+					if(devices[i].respond) {
+						devices[i].respond(&devices[i], area, meta);
+					}
+					break;
+				}
+			}
+		}
+	} else {
+		while(1) {
+			sys_signal_wait(SYS_SIGNAL_WAIT_ANY, NULL);
+		}
+	}
+}

luma/uhci/usb.c

@@ -0,0 +1,153 @@
+#include"usb.h"
+
+#include"host.h"
+#include"bbb.h"
+#include"../sys.h"
+
+void dbg(int j, uint8_t b);
+extern uint32_t asdfasdf;
+
+int USBDeviceDiscover(USBDevice *dev) {
+	uint8_t tmp = dev->id;
+	dev->id = 0; // 0 is the correct device ID as far as USB is concerned
+	if(!HostSendRequest(dev, NULL, (USBStandardRequest) {
+		.requestType = 0x00,
+		.request = 5, // SET_ADDRESS
+		.value = tmp,
+		.index = 0,
+		.length = 0
+	})) {
+		dev->id = tmp;
+		return 0;
+	}
+	dev->id = tmp;
+	
+	USBDeviceDescriptor ddesc;
+	if(!HostSendRequest(dev, &ddesc, (USBStandardRequest) {
+		.requestType = 0x80,
+		.request = 6, // GET_DESCRIPTOR
+		.value = 1 << 8, // device
+		.index = 0,
+		.length = 8
+	})) return 0;
+	
+	dev->maxPacketSize = ddesc.maxPacketSize;
+	
+	if(ddesc.clazz == 0 && ddesc.subclazz == 0 && ddesc.protocol == 0) {
+		uint8_t buf[64];
+		if(HostSendRequest(dev, buf, (USBStandardRequest) {
+			.requestType = 0x80,
+			.request = 6, // GET_DESCRIPTOR
+			.value = 2 << 8, // configuration
+			.index = 0,
+			.length = sizeof(buf)
+		})) {
+			USBConfigurationDescriptor *conf = (void*) buf;
+			
+			if(HostSendRequest(dev, NULL, (USBStandardRequest) {
+				.requestType = 0x00,
+				.request = 9,
+				.value = conf->id,
+				.index = 0,
+				.length = 0
+			})) {
+				if(conf->interfaceCount == 1) {
+					USBInterfaceDescriptor *interf = (USBInterfaceDescriptor*) ((uintptr_t) conf + conf->sz);
+					
+					if(HostSendRequest(dev, NULL, (USBStandardRequest) {
+						.requestType = 0x01,
+						.request = 11,
+						.value = interf->alternate,
+						.index = interf->interface,
+						.length = 0
+					})) {
+						if(interf->clazz == 0x08 && interf->subclazz == 0x06 && interf->protocol == 0x50) {
+							USBEndpointDescriptor *endp0 = (USBEndpointDescriptor*) ((uintptr_t) interf + interf->sz);
+							USBEndpointDescriptor *endp1 = (USBEndpointDescriptor*) ((uintptr_t) endp0 + endp0->sz);
+
+							dev->bbb.maxPacketSize = endp0->maxPacketSize;
+							dev->bbb.toggleIn = dev->bbb.toggleOut = 0;
+							
+							if(endp0->endpoint & 0x80) {
+								dev->bbb.endpIn = endp0->endpoint;
+								dev->bbb.endpOut = endp1->endpoint;
+							} else {
+								dev->bbb.endpIn = endp1->endpoint;
+								dev->bbb.endpOut = endp0->endpoint;
+							}
+							
+							uint8_t lun;
+							if(!HostSendRequest(dev, &lun, (USBStandardRequest) {
+								.requestType = 0xA1,
+								.request = 0xFE, // get max lun
+								.value = 0,
+								.index = 0,
+								.length = sizeof(lun)
+							})) lun = 0;
+							
+							// (Almost) initialization procedure laid out by Doug Brown
+							
+							int attempts;
+							
+							attempts = 100;
+							
+							while(attempts-- > 0) {
+								
+								int success = 1;
+								
+								if(!BBBTestUnitReady(dev)) {
+									success = 0;
+								}
+								
+								if(!BBBInquiry(dev)) {
+									success = 0;
+								}
+								
+								if(success) {
+									break;
+								}
+								
+							}
+							
+							if(attempts > 0) {
+								
+								attempts = 100;
+								
+								while(attempts-- > 0) {
+									if(BBBReadCapacity10(dev)) {
+										break;
+									}
+								}
+								
+								if(attempts > 0) {
+									
+									//BBBModeSense(dev);
+									
+									// One flash drive doesn't work at all without this Read(10)
+									uint8_t lolibuf[dev->bbb.sectorSize];
+									BBBRead10(dev, 0, lolibuf, 1);
+									
+									for(uint32_t try = 0; try <= 0xFFFF; try++) {
+										if((dev->canal = sys_canal_create('STRG', 'UD\0\0' | try)) < (uint32_t) -256) {
+											break;
+										}
+									}
+									
+									if(dev->canal >= (uint32_t) -256) {
+										return 0;
+									}
+									
+									dev->respond = BBBDeviceResponse;
+									
+								}
+								
+							}
+						}
+					}
+				}
+			}
+		}
+	}
+	
+	return 1;
+}

luma/uhci/usb.h

@@ -0,0 +1,86 @@
+#pragma once
+
+#include<stdint.h>
+
+typedef enum {
+	SETUP = 0x2D, IN = 0x69, OUT = 0xE1
+} PacketType;
+
+typedef struct USBDevice {
+	uint8_t hc;
+	uint8_t addr;
+	uint8_t speedy;
+	uint8_t id;
+	uint8_t maxPacketSize;
+	uint32_t canal;
+	union {
+		struct {
+			int endpIn;
+			int endpOut;
+			int maxPacketSize;
+			int toggleIn;
+			int toggleOut;
+			uint16_t sectorSize;
+		} bbb;
+	};
+	int(*respond)(struct USBDevice *dev, void *area, uint32_t meta);
+} USBDevice;
+
+typedef struct __attribute((packed)) {
+	uint8_t requestType;
+	uint8_t request;
+	uint16_t value;
+	uint16_t index;
+	uint16_t length;
+} USBStandardRequest;
+
+typedef struct {
+	uint8_t sz;
+	uint8_t type;
+	uint16_t usbVer;
+	uint8_t clazz;
+	uint8_t subclazz;
+	uint8_t protocol;
+	uint8_t maxPacketSize;
+	uint16_t idVendor;
+	uint16_t idProduct;
+	uint16_t devVer;
+	uint8_t stringManufacturer;
+	uint8_t stringProduct;
+	uint8_t stringSerial;
+	uint8_t configurationCount;
+} USBDeviceDescriptor;
+
+typedef struct {
+	uint8_t sz;
+	uint8_t type;
+	uint16_t szTotal;
+	uint8_t interfaceCount;
+	uint8_t id;
+	uint8_t string;
+	uint8_t attributes;
+	uint8_t maxPower;
+} USBConfigurationDescriptor;
+
+typedef struct {
+	uint8_t sz;
+	uint8_t type;
+	uint8_t interface;
+	uint8_t alternate;
+	uint8_t endpointCount;
+	uint8_t clazz;
+	uint8_t subclazz;
+	uint8_t protocol;
+	uint8_t string;
+} USBInterfaceDescriptor;
+
+typedef struct {
+	uint8_t sz;
+	uint8_t type;
+	uint8_t endpoint;
+	uint8_t attributes;
+	uint16_t maxPacketSize;
+	uint8_t interval;
+} USBEndpointDescriptor;
+
+int USBDeviceDiscover(USBDevice*);

luma/vid/Makefile

@@ -0,0 +1,13 @@
+rwildcard=$(foreach d,$(wildcard $(1:=/*)),$(call rwildcard,$d,$2) $(filter $(subst *,%,$2),$d))
+
+HEADERS := $(call rwildcard,.,*.h)
+SOURCEC := $(call rwildcard,.,*.c)
+OBJECTS := $(patsubst %.c, ../../build/vid/%.o, $(SOURCEC))
+
+mod: $(OBJECTS)
+	i386-elf-gcc -pie -T ../../mod.ld -Wl,-no-dynamic-linker -nostdlib -o main.elf $(OBJECTS) -lgcc
+	../../elftomod.py main.elf ../../bootfs/luma_vid.mod modentry:0
+
+../../build/vid/%.o: %.c
+	mkdir -p $(@D)
+	i386-elf-gcc -march=i386 -Isrc -Os -pie -nodefaultlibs -nostdlib -nostartfiles -fomit-frame-pointer -c -o $@ -ffreestanding -mgeneral-regs-only -std=gnu99 -Wall $<

luma/vid/api.h

@@ -0,0 +1,88 @@
+#ifndef _LUMA_VBE_VBE_H
+#define _LUMA_VBE_VBE_H
+
+#include<stdint.h>
+
+typedef enum {
+	LUMA_VID_COMMAND_END = 0, LUMA_VID_COMMAND_CLEAR = 1, LUMA_VID_COMMAND_RECT = 2, LUMA_VID_COMMAND_IMAGE = 3, LUMA_VID_COMMAND_IMAGE_NEW = 4, LUMA_VID_COMMAND_SET_COLOR_FLAT = 5,
+	LUMA_VID_COMMAND_SET_COLOR_GRAD = 6, LUMA_VID_COMMAND_IMAGE_DEL = 7, LUMA_VID_COMMAND_TEXT = 8, LUMA_VID_COMMAND_NEW_COLOR_FLAT = 9, LUMA_VID_COMMAND_NEW_COLOR_GRAD = 10, LUMA_VID_COMMAND_IMAGE_WRITE = 11,
+} LumaVidCommandOp;
+
+typedef enum {
+	LUMA_VID_IMAGE_FORMAT_RAW = 0, LUMA_VID_IMAGE_FORMAT_R8G8B8 = 1, LUMA_VID_IMAGE_FORMAT_BLALPHA = 2
+} LumaVidImageDescriptionFormat;
+
+typedef enum {
+	LUMA_VID_ALIGN_TOP_LEFT = 0,
+	LUMA_VID_ALIGN_TOP_RIGHT = 1,
+	LUMA_VID_ALIGN_BOTTOM_LEFT = 2,
+	LUMA_VID_ALIGN_BOTTOM_RIGHT = 3,
+} LumaVidAlign;
+
+typedef union {
+	uint8_t op;
+	struct {
+		uint8_t op;
+	} __attribute__((packed)) clear;
+	struct {
+		uint8_t op;
+		uint16_t y2;
+		uint16_t x2;
+		uint16_t y1;
+		uint16_t x1;
+	} __attribute__((packed)) rect;
+	struct {
+		uint8_t op;
+		uint16_t y;
+		uint16_t x;
+		uint16_t subY;
+		uint16_t subX;
+		uint16_t width;
+		uint16_t height;
+		uint32_t name;
+	} __attribute__((packed)) image;
+	struct {
+		uint8_t op;
+		uint32_t name;
+		uint16_t width;
+		uint16_t height;
+		uint8_t format;
+	} __attribute__((packed)) imageNew;
+	struct {
+		uint8_t op;
+		uint8_t r;
+		uint8_t g;
+		uint8_t b;
+	} __attribute__((packed)) colorFlat;
+	struct {
+		uint8_t op;
+		uint8_t r1;
+		uint8_t r2;
+		uint8_t g1;
+		uint8_t g2;
+		uint8_t b1;
+		uint8_t b2;
+	} __attribute__((packed)) colorGrad;
+	struct {
+		uint8_t op;
+		uint32_t name;
+	} __attribute__((packed)) imageDel;
+	struct {
+		uint8_t op;
+		uint16_t y;
+		uint16_t x;
+		uint16_t len;
+		uint16_t wall;
+		uint8_t data[];
+	} __attribute__((packed)) text;
+	struct {
+		uint8_t op;
+		uint32_t name;
+		uint16_t y;
+		uint16_t x;
+		uint16_t len;
+		uint8_t data[];
+	} __attribute__((packed)) imageWrite;
+} __attribute__((packed)) LumaVidCommand;
+
+#endif

luma/vid/kmeans.c

@@ -0,0 +1,101 @@
+#include"kmeans.h"
+
+#include"../sys.h"
+
+#define RED(c) (c >> 11)
+#define GRN(c) ((c >> 5) & 63)
+#define BLU(c) (c & 31)
+
+#define CLUSTERS 256
+struct {
+	MiniCol averages[CLUSTERS];
+} Cluster;
+
+typedef struct {
+	uint8_t cluster;
+	MiniCol color;
+} __attribute__((packed)) Datapoint;
+
+struct {
+	size_t sz;
+	Datapoint *buf;
+} Dataset = {};
+
+__attribute__((optimize("Ofast"))) static uint32_t coldist(MiniCol a, MiniCol b) {
+	uint16_t dr = RED(a) - RED(b);
+	uint16_t dg = GRN(a) - GRN(b);
+	uint16_t db = BLU(a) - BLU(b);
+	return (uint16_t) (dr * dr) + (uint16_t) (dg * dg) + (uint16_t) (db * db);
+}
+
+void q_reavg() {
+	if(Dataset.sz < CLUSTERS) {
+		for(size_t i = 0; i < Dataset.sz; i++) {
+			MiniCol c = Dataset.buf[i].color;
+			Cluster.averages[Dataset.buf[i].cluster = i] = c;
+		}
+	} else {
+		uint16_t counts[CLUSTERS] = {};
+		uint16_t avgRed[CLUSTERS] = {};
+		uint32_t avgGreen[CLUSTERS] = {};
+		uint16_t avgBlue[CLUSTERS] = {};
+		for(size_t i = 0; i < Dataset.sz; i++) {
+			counts[Dataset.buf[i].cluster]++;
+			avgRed[Dataset.buf[i].cluster] += RED(Dataset.buf[i].color);
+			avgGreen[Dataset.buf[i].cluster] += GRN(Dataset.buf[i].color);
+			avgBlue[Dataset.buf[i].cluster] += BLU(Dataset.buf[i].color);
+		}
+		
+		for(int i = 0; i < CLUSTERS; i++) {
+			if(counts[i] == 0) continue;
+			
+			avgRed[i] /= counts[i];
+			avgGreen[i] /= counts[i];
+			avgBlue[i] /= counts[i];
+			
+			Cluster.averages[i] = (avgRed[i] << 11) | (avgGreen[i] << 5) | avgBlue[i];
+		}
+	}
+	
+	for(size_t i = 0; i < Dataset.sz; i++) {
+		Dataset.buf[i].cluster = q_p2c(Dataset.buf[i].color);
+	}
+}
+
+__attribute__((optimize("Ofast"))) int q_p2c(MiniCol col) {
+	int clu = 0;
+	uint32_t dist = coldist(col, Cluster.averages[clu]);
+	
+	for(int i = 1; i < CLUSTERS; i++) {
+		uint32_t d = coldist(col, Cluster.averages[i]);
+		if(dist > d) {
+			dist = d;
+			clu = i;
+		}
+	}
+	
+	return clu;
+}
+
+void q_reassign() {
+	asm("out %%al, %%dx" : : "a"(0), "d"(0x3C8));
+	size_t bound = Dataset.sz < CLUSTERS ? Dataset.sz : CLUSTERS;
+	for(size_t i = 0; i < bound; i++) {
+		MiniCol c = Cluster.averages[i];
+		asm("out %%al, %%dx" : : "a"((c >> 10) & ~1), "d"(0x3C9));
+		asm("out %%al, %%dx" : : "a"((c >> 5) & 63), "d"(0x3C9));
+		asm("out %%al, %%dx" : : "a"((c & 31) << 1), "d"(0x3C9));
+	}
+}
+
+void q_addp(MiniCol col) {
+	if(Dataset.sz == 4096 / sizeof(Datapoint)) {
+		return;
+	}
+	
+	Dataset.buf[Dataset.sz++] = (Datapoint) {.cluster = q_p2c(col), .color = col};
+}
+
+void q_init() {
+	Dataset.buf = sys_vpm_map(SYS_MAP_VIRT_ANY, SYS_MAP_PHYS_ANY, 4096);
+}

luma/vid/kmeans.c.test

@@ -0,0 +1,13 @@
+#include<stdlib.h>
+#include<stdio.h>
+
+#define sys_vpm_map(x,y,z) malloc(z)
+#include"kmeans.c"
+
+int main() {
+	q_init();
+	
+	MiniCol c = 30 | (30 << 5) | (30 << 11);
+	q_addp(c);
+	q_p2c();
+}

luma/vid/kmeans.h

@@ -0,0 +1,15 @@
+#ifndef KMEANS_H
+#define KMEANS_H
+
+#include<stdint.h>
+#include<stddef.h>
+
+typedef uint16_t MiniCol;
+
+void q_init();
+void q_reavg();
+void q_reassign();
+int q_p2c(MiniCol col);
+void q_addp(MiniCol col);
+
+#endif

luma/vid/main.c

@@ -0,0 +1,497 @@
+#include<stddef.h>
+#include<stdint.h>
+#include"../sys.h"
+#include"../std.h"
+#include"api.h"
+#include"kmeans.h"
+
+asm("modentry:\n"
+"mov $stack + 2816, %esp\n"
+"call startc");
+
+static uint8_t *FrameBuffer;
+
+typedef struct {
+	uint32_t name;
+	uint32_t len;
+	uint8_t code[];
+} DL;
+
+static struct {
+	size_t idx;
+	uint8_t *buf;
+} DLs;
+
+typedef struct {
+	uint16_t w, h;
+	uint8_t format;
+	uint8_t data[];
+} Image;
+
+static struct {
+	size_t idx;
+	Image **handles;
+} Images;
+
+char stack[2816];
+
+static uint8_t(*colorFunc)(uint16_t x, uint16_t y);
+
+static uint8_t color_flat[] = {0xB8, 0x78, 0x56, 0x34, 0x12, 0xC3};
+
+static uint8_t gradCols[16];
+__attribute__((optimize("Ofast"))) static uint8_t color_grad(uint16_t x, uint16_t y) {
+	static uint8_t matrix[16][16] = {{0, 64, 16, 80, 4, 68, 20, 84, 1, 65, 17, 81, 5, 69, 21, 85}, {192, 128, 208, 144, 196, 132, 212, 148, 193, 129, 209, 145, 197, 133, 213, 149}, {48, 112, 32, 96, 52, 116, 36, 100, 49, 113, 33, 97, 53, 117, 37, 101}, {240, 176, 224, 160, 244, 180, 228, 164, 241, 177, 225, 161, 245, 181, 229, 165}, {12, 76, 28, 92, 8, 72, 24, 88, 13, 77, 29, 93, 9, 73, 25, 89}, {204, 140, 220, 156, 200, 136, 216, 152, 205, 141, 221, 157, 201, 137, 217, 153}, {60, 124, 44, 108, 56, 120, 40, 104, 61, 125, 45, 109, 57, 121, 41, 105}, {252, 188, 236, 172, 248, 184, 232, 168, 253, 189, 237, 173, 249, 185, 233, 169}, {3, 67, 19, 83, 7, 71, 23, 87, 2, 66, 18, 82, 6, 70, 22, 86}, {195, 131, 211, 147, 199, 135, 215, 151, 194, 130, 210, 146, 198, 134, 214, 150}, {51, 115, 35, 99, 55, 119, 39, 103, 50, 114, 34, 98, 54, 118, 38, 102}, {243, 179, 227, 163, 247, 183, 231, 167, 242, 178, 226, 162, 246, 182, 230, 166}, {15, 79, 31, 95, 11, 75, 27, 91, 14, 78, 30, 94, 10, 74, 26, 90}, {207, 143, 223, 159, 203, 139, 219, 155, 206, 142, 222, 158, 202, 138, 218, 154}, {63, 127, 47, 111, 59, 123, 43, 107, 62, 126, 46, 110, 58, 122, 42, 106}, {255, 191, 239, 175, 251, 187, 235, 171, 254, 190, 238, 174, 250, 186, 234, 170}};
+	
+	int i = x * 256 / 320;
+	i += matrix[x % 16][y % 16] / 16;
+	i /= 16;
+	if(i > 15) i = 15;
+	
+	return gradCols[i];
+}
+
+__attribute__((optimize("Ofast"))) static void select_planes(int planes) {
+	asm("out %%al, %%dx" : : "a"(2), "d"(0x3C4));
+	asm("out %%al, %%dx" : : "a"(planes), "d"(0x3C5));
+}
+
+__attribute__((optimize("Ofast"))) static void clear() {
+	if((uintptr_t) colorFunc == (uintptr_t) color_flat) {
+		uint8_t col = colorFunc(0, 0);
+		
+		select_planes(15);
+		for(int i = 0; i < 4800; i++) {
+			((uint32_t*) FrameBuffer)[i] = col | (col << 8) | (col << 16) | (col << 24);
+		}
+	} else {
+		for(int p = 0; p < 4; p++) {
+			select_planes(1 << p);
+			for(int y = 0; y < 240; y++) {
+				for(int x = 0; x < 80; x++) {
+					FrameBuffer[80 * y + x] = colorFunc(x * 4 + p, y);
+				}
+			}
+		}
+	}
+}
+
+__attribute__((optimize("Ofast"))) static void rect(int x1, int y1, int x2, int y2) {
+	select_planes(1 << ((x1 + 0) & 3));
+	for(int y = y1; y <= y2; y++) {
+		for(int x = x1 + 0; x <= x2; x += 4) {
+			FrameBuffer[80 * y + x / 4] = colorFunc(x, y);
+		}
+	}
+	select_planes(1 << ((x1 + 1) & 3));
+	for(int y = y1; y <= y2; y++) {
+		for(int x = x1 + 1; x <= x2; x += 4) {
+			FrameBuffer[80 * y + x / 4] = colorFunc(x, y);
+		}
+	}
+	select_planes(1 << ((x1 + 2) & 3));
+	for(int y = y1; y <= y2; y++) {
+		for(int x = x1 + 2; x <= x2; x += 4) {
+			FrameBuffer[80 * y + x / 4] = colorFunc(x, y);
+		}
+	}
+	select_planes(1 << ((x1 + 3) & 3));
+	for(int y = y1; y <= y2; y++) {
+		for(int x = x1 + 3; x <= x2; x += 4) {
+			FrameBuffer[80 * y + x / 4] = colorFunc(x, y);
+		}
+	}
+}
+
+__attribute__((optimize("Ofast"))) static void image(Image *img, uint16_t subX, uint16_t subY, uint16_t x, uint16_t y, uint16_t width, uint16_t height) {
+	if(img->format == LUMA_VID_IMAGE_FORMAT_BLALPHA) {
+		uint8_t c = colorFunc(0, 0);
+		for(int ix = 0; ix < width; ix++) {
+			select_planes(1 << ((x + ix) & 3));
+			for(int iy = 0; iy < height; iy++) {
+				uint8_t v = img->data[(iy * img->w + ix) / 8];
+				if(v & (1 << (ix % 8))) {
+					FrameBuffer[80 * (iy + y) + (ix + x) / 4] = c;
+				}
+			}
+		}
+	} else if(img->format == LUMA_VID_IMAGE_FORMAT_R8G8B8) {
+		for(int plane = 0; plane < 4; plane++) {
+			select_planes(1 << ((x + plane) & 3));
+			
+			for(int iy = 0; iy < height; iy++) {
+				uint8_t *dst = FrameBuffer + 80 * (iy + y) + x / 4;
+				for(int ix = plane; ix < width; ix += 4) {
+					*(dst++) = q_p2c(*(MiniCol*) &img->data[((iy + subY) * img->w + ix + subX) * 2]);
+				}
+			}
+		}
+	}
+}
+
+__attribute__((optimize("Ofast"))) static void text(const char *str, size_t len, uint16_t x, uint16_t y, uint16_t wall) {
+	if(wall <= x) wall = 320;
+	
+	static const uint8_t font[] = {
+		0x20, 0x31, 0x95, 0x00, 0x21, 0x17, 0x15, 0x5f, 0x22, 0x32, 0x15, 0x2d,
+		0x23, 0x47, 0x15, 0xaa, 0xaf, 0x5f, 0x05, 0x24, 0x48, 0x15, 0x5e, 0x65,
+		0xcc, 0x47, 0x25, 0x57, 0x15, 0x6b, 0x1d, 0xc2, 0xb5, 0x06, 0x26, 0x57,
+		0x15, 0x44, 0x19, 0x9b, 0x92, 0x05, 0x27, 0x12, 0x15, 0x03, 0x28, 0x28,
+		0x15, 0x56, 0x95, 0x29, 0x28, 0x15, 0xa9, 0x6a, 0x2a, 0x34, 0x15, 0xfa,
+		0x05, 0x2b, 0x55, 0x35, 0x84, 0x7c, 0x42, 0x00, 0x2c, 0x22, 0x75, 0x06,
+		0x2d, 0x31, 0x55, 0x07, 0x2e, 0x11, 0x75, 0x01, 0x2f, 0x37, 0x15, 0xa4,
+		0xa4, 0x04, 0x30, 0x47, 0x15, 0x96, 0xb9, 0x99, 0x06, 0x31, 0x37, 0x15,
+		0x93, 0x24, 0x1d, 0x32, 0x47, 0x15, 0x87, 0x48, 0x12, 0x0f, 0x33, 0x47,
+		0x15, 0x87, 0x68, 0x88, 0x07, 0x34, 0x57, 0x15, 0x8c, 0xa9, 0xf4, 0x11,
+		0x02, 0x35, 0x47, 0x15, 0x1f, 0x71, 0x88, 0x07, 0x36, 0x47, 0x15, 0x1e,
+		0x71, 0x99, 0x06, 0x37, 0x47, 0x15, 0x8f, 0x44, 0x22, 0x01, 0x38, 0x47,
+		0x15, 0x96, 0x69, 0x99, 0x06, 0x39, 0x47, 0x15, 0x96, 0xe9, 0x88, 0x07,
+		0x3a, 0x15, 0x35, 0x11, 0x3b, 0x26, 0x35, 0x02, 0x06, 0x3c, 0x35, 0x25,
+		0x54, 0x44, 0x3d, 0x43, 0x45, 0x0f, 0x0f, 0x3e, 0x35, 0x25, 0x11, 0x15,
+		0x3f, 0x37, 0x15, 0x2a, 0x25, 0x08, 0x40, 0x57, 0x15, 0x2e, 0xf6, 0xde,
+		0x82, 0x03, 0x41, 0x57, 0x15, 0x84, 0x28, 0xe5, 0x62, 0x04, 0x42, 0x47,
+		0x15, 0x97, 0x79, 0x99, 0x07, 0x43, 0x47, 0x15, 0x1e, 0x11, 0x11, 0x0e,
+		0x44, 0x47, 0x15, 0x97, 0x99, 0x99, 0x07, 0x45, 0x47, 0x15, 0x1f, 0xf1,
+		0x11, 0x0f, 0x46, 0x47, 0x15, 0x1f, 0xf1, 0x11, 0x01, 0x47, 0x47, 0x15,
+		0x1e, 0x11, 0x9d, 0x0e, 0x48, 0x47, 0x15, 0x99, 0xf9, 0x99, 0x09, 0x49,
+		0x37, 0x15, 0x97, 0x24, 0x1d, 0x4a, 0x37, 0x15, 0x24, 0x49, 0x0e, 0x4b,
+		0x57, 0x15, 0x31, 0x95, 0x93, 0x52, 0x04, 0x4c, 0x47, 0x15, 0x11, 0x11,
+		0x11, 0x0f, 0x4d, 0x47, 0x15, 0xb9, 0xbf, 0x99, 0x09, 0x4e, 0x47, 0x15,
+		0xb9, 0xbb, 0xdd, 0x09, 0x4f, 0x47, 0x15, 0x96, 0x99, 0x99, 0x06, 0x50,
+		0x47, 0x15, 0x97, 0x79, 0x11, 0x01, 0x51, 0x48, 0x15, 0x96, 0x99, 0x99,
+		0x86, 0x52, 0x57, 0x15, 0x27, 0xa5, 0x53, 0x52, 0x04, 0x53, 0x47, 0x15,
+		0x1e, 0x61, 0x88, 0x07, 0x54, 0x57, 0x15, 0x9f, 0x10, 0x42, 0x08, 0x01,
+		0x55, 0x47, 0x15, 0x99, 0x99, 0x99, 0x06, 0x56, 0x57, 0x15, 0x31, 0x2a,
+		0xa5, 0x08, 0x01, 0x57, 0x57, 0x15, 0x31, 0xd6, 0xba, 0x94, 0x02, 0x58,
+		0x47, 0x15, 0x99, 0x66, 0x96, 0x09, 0x59, 0x57, 0x15, 0x51, 0x29, 0x42,
+		0x08, 0x01, 0x5a, 0x47, 0x15, 0x8f, 0x24, 0x12, 0x0f, 0x5b, 0x28, 0x15,
+		0x57, 0xd5, 0x5c, 0x37, 0x15, 0x89, 0x24, 0x12, 0x5d, 0x28, 0x15, 0xab,
+		0xea, 0x5e, 0x53, 0x25, 0x44, 0x45, 0x5f, 0x51, 0x95, 0x1f, 0x60, 0x22,
+		0x15, 0x09, 0x61, 0x45, 0x35, 0x87, 0x9e, 0x0e, 0x62, 0x47, 0x15, 0x11,
+		0x97, 0x99, 0x07, 0x63, 0x45, 0x35, 0x1e, 0x11, 0x0e, 0x64, 0x47, 0x15,
+		0x88, 0x9e, 0x99, 0x0e, 0x65, 0x45, 0x35, 0x96, 0x1f, 0x0e, 0x66, 0x47,
+		0x15, 0x2c, 0x2f, 0x22, 0x02, 0x67, 0x47, 0x35, 0x5e, 0x12, 0x9f, 0x06,
+		0x68, 0x47, 0x15, 0x11, 0xb5, 0x99, 0x09, 0x69, 0x27, 0x15, 0xb2, 0x2a,
+		0x6a, 0x39, 0x15, 0x84, 0x49, 0x92, 0x03, 0x6b, 0x57, 0x15, 0x21, 0xa4,
+		0x72, 0x52, 0x04, 0x6c, 0x37, 0x15, 0x93, 0x24, 0x19, 0x6d, 0x55, 0x35,
+		0xab, 0xd6, 0x5a, 0x01, 0x6e, 0x45, 0x35, 0xb5, 0x99, 0x09, 0x6f, 0x45,
+		0x35, 0x96, 0x99, 0x06, 0x70, 0x47, 0x35, 0x97, 0x99, 0x17, 0x01, 0x71,
+		0x47, 0x35, 0x9e, 0x99, 0x8e, 0x08, 0x72, 0x35, 0x35, 0x4f, 0x12, 0x73,
+		0x45, 0x35, 0x1f, 0x86, 0x07, 0x74, 0x46, 0x25, 0xf2, 0x22, 0xc2, 0x75,
+		0x45, 0x35, 0x99, 0x99, 0x0e, 0x76, 0x45, 0x35, 0xa9, 0x4a, 0x04, 0x77,
+		0x55, 0x35, 0xb1, 0x56, 0xa5, 0x00, 0x78, 0x45, 0x35, 0x69, 0x66, 0x09,
+		0x79, 0x47, 0x35, 0xa9, 0x4a, 0x44, 0x03, 0x7a, 0x45, 0x35, 0x4f, 0x12,
+		0x0f, 0x7b, 0x39, 0x15, 0x94, 0x14, 0x49, 0x04, 0x7c, 0x18, 0x15, 0xff,
+		0x7d, 0x39, 0x15, 0x91, 0x44, 0x49, 0x01, 0x7e, 0x52, 0x45, 0x36, 0x01
+	};
+	
+	const uint8_t *find(char c) {
+		const uint8_t *a = font;
+		while(a - font < sizeof(font)) {
+			if(*a == c) {
+				return a;
+			} else {
+				a += 3 + ((a[1] & 15) * (a[1] >> 4) + 7) / 8;
+			}
+		}
+		return NULL;
+	}
+	
+	uint16_t startX = x;
+
+	int col = colorFunc(0, 0);
+	while(len--) {
+		if(*str == 10) {
+			str++;
+			x = startX;
+			y += 13;
+			continue;
+		}
+		
+		const uint8_t *data = find(*(str++));
+		
+		if(data == NULL) continue;
+		
+		uint16_t cx = 0, cy = data[2] >> 4;
+		
+		uint8_t width = data[2] & 0x0F;
+		if(x + width >= wall) {
+			x = startX;
+			y += 13;
+		}
+		
+		for(int ix = 0; ix < (data[1] >> 4); ix++) {
+			select_planes(1 << ((x + ix) & 3));
+			for(int iy = 0; iy < (data[1] & 15); iy++) {
+				uint8_t v = data[3 + (iy * (data[1] >> 4) + ix) / 8];
+				if(v & (1 << ((iy * (data[1] >> 4) + ix) % 8))) {
+					FrameBuffer[80 * (iy + y + cy) + (ix + x + cx) / 4] = col;
+				}
+			}
+		}
+		
+		x += width;
+	}
+}
+
+static void compileCommandList(uint8_t *src, uint8_t *dst) {
+	size_t len = 0;
+	while(1) {
+		LumaVidCommand *cmd = (void*) src;
+		
+		if(cmd->op == LUMA_VID_COMMAND_CLEAR) {
+			dst[len++] = 0xE8; /* call rel32 */
+			*((uint32_t*) &dst[len]) = (uint32_t) &clear - (uint32_t) &dst[len + 4];
+			len += 4;
+			src += sizeof(cmd->clear);
+		} else if(cmd->op == LUMA_VID_COMMAND_END) {
+			dst[len++] = 0xC3; /* ret */
+			break;
+		} else if(cmd->op == LUMA_VID_COMMAND_RECT) {
+			dst[len++] = 0x68; /* push imm32 */
+			dst[len++] = cmd->rect.y2 & 0xFF;
+			dst[len++] = cmd->rect.y2 >> 8;
+			dst[len++] = 0;
+			dst[len++] = 0;
+			dst[len++] = 0x68; /* push imm32 */
+			dst[len++] = cmd->rect.x2 & 0xFF;
+			dst[len++] = cmd->rect.x2 >> 8;
+			dst[len++] = 0;
+			dst[len++] = 0;
+			dst[len++] = 0x68; /* push imm32 */
+			dst[len++] = cmd->rect.y1 & 0xFF;
+			dst[len++] = cmd->rect.y1 >> 8;
+			dst[len++] = 0;
+			dst[len++] = 0;
+			dst[len++] = 0x68; /* push imm32 */
+			dst[len++] = cmd->rect.x1 & 0xFF;
+			dst[len++] = cmd->rect.x1 >> 8;
+			dst[len++] = 0;
+			dst[len++] = 0;
+			dst[len++] = 0xE8; /* call rel32 */
+			*((uint32_t*) &dst[len]) = (uint32_t) &rect - (uint32_t) &dst[len + 4];
+			len += 4;
+			dst[len++] = 0x83; /* add */
+			dst[len++] = 0xC4; /* esp */
+			dst[len++] = 16;
+			src += sizeof(cmd->rect);
+		} else if(cmd->op == LUMA_VID_COMMAND_IMAGE) {
+			dst[len++] = 0x68; /* push imm32 */
+			dst[len++] = cmd->image.height & 0xFF;
+			dst[len++] = cmd->image.height >> 8;
+			dst[len++] = 0;
+			dst[len++] = 0;
+			dst[len++] = 0x68; /* push imm32 */
+			dst[len++] = cmd->image.width & 0xFF;
+			dst[len++] = cmd->image.width >> 8;
+			dst[len++] = 0;
+			dst[len++] = 0;
+			dst[len++] = 0x68; /* push imm32 */
+			dst[len++] = cmd->image.y & 0xFF;
+			dst[len++] = cmd->image.y >> 8;
+			dst[len++] = 0;
+			dst[len++] = 0;
+			dst[len++] = 0x68; /* push imm32 */
+			dst[len++] = cmd->image.x & 0xFF;
+			dst[len++] = cmd->image.x >> 8;
+			dst[len++] = 0;
+			dst[len++] = 0;
+			dst[len++] = 0x68; /* push imm32 */
+			dst[len++] = cmd->image.subY & 0xFF;
+			dst[len++] = cmd->image.subY >> 8;
+			dst[len++] = 0;
+			dst[len++] = 0;
+			dst[len++] = 0x68; /* push imm32 */
+			dst[len++] = cmd->image.subX & 0xFF;
+			dst[len++] = cmd->image.subX >> 8;
+			dst[len++] = 0;
+			dst[len++] = 0;
+			dst[len++] = 0x68; /* push imm32 */
+			Image *img = Images.handles[cmd->image.name];
+			dst[len++] = ((uint32_t) img >> 0) & 0xFF;
+			dst[len++] = ((uint32_t) img >> 8) & 0xFF;
+			dst[len++] = ((uint32_t) img >> 16) & 0xFF;
+			dst[len++] = ((uint32_t) img >> 24) & 0xFF;
+			dst[len++] = 0xE8; /* call rel32 */
+			*((uint32_t*) &dst[len]) = (uint32_t) &image - (uint32_t) &dst[len + 4];
+			len += 4;
+			dst[len++] = 0x83; /* add */
+			dst[len++] = 0xC4; /* esp */
+			dst[len++] = 28;
+			src += sizeof(cmd->image);
+		} else if(cmd->op == LUMA_VID_COMMAND_IMAGE_NEW) {
+			Image *img = NULL;
+			if(cmd->imageNew.format == LUMA_VID_IMAGE_FORMAT_BLALPHA) {
+				img = sys_vpm_map(SYS_MAP_VIRT_ANY, SYS_MAP_PHYS_ANY, sizeof(Image) + (cmd->imageNew.width * cmd->imageNew.height + 7) / 8);
+			} else if(cmd->imageNew.format == LUMA_VID_IMAGE_FORMAT_R8G8B8) {
+				img = sys_vpm_map(SYS_MAP_VIRT_ANY, SYS_MAP_PHYS_ANY, sizeof(Image) + cmd->imageNew.width * cmd->imageNew.height * 2);
+			}
+			if(img) {
+				img->w = cmd->imageNew.width;
+				img->h = cmd->imageNew.height;
+				img->format = cmd->imageNew.format;
+			}
+			Images.handles[cmd->imageNew.name] = img;
+			
+			src += sizeof(cmd->imageNew);
+		} else if(cmd->op == LUMA_VID_COMMAND_IMAGE_WRITE) {
+			Image *img = Images.handles[cmd->imageWrite.name];
+			if(img->format == LUMA_VID_IMAGE_FORMAT_BLALPHA) {
+				// Copy data
+				std_copy(img->data + cmd->imageWrite.y * img->w + cmd->imageWrite.x, cmd->imageWrite.data, cmd->imageWrite.len);
+			} else if(img->format == LUMA_VID_IMAGE_FORMAT_R8G8B8) {
+				// Copy data and reformat
+				MiniCol *dst = (MiniCol*) img->data + cmd->imageWrite.y * img->w + cmd->imageWrite.x;
+				for(size_t i = 0; i < cmd->imageWrite.len; i += 3) {
+					*(dst++) = ((cmd->imageWrite.data[i + 0] & ~7) << 8) | ((cmd->imageWrite.data[i + 1] & ~3) << 3) | (cmd->imageWrite.data[i + 2] >> 3);
+				}
+			}
+			src += sizeof(cmd->imageWrite) + cmd->imageWrite.len;
+		} else if(cmd->op == LUMA_VID_COMMAND_SET_COLOR_FLAT) {
+			MiniCol minicol = ((cmd->colorFlat.r & ~7) << 8) | ((cmd->colorFlat.g & ~3) << 3) | (cmd->colorFlat.b >> 3);
+			
+			dst[len++] = 0xC7; /* mov [abs], imm */
+			dst[len++] = 0x05;
+			*((uint32_t*) &dst[len]) = (uint32_t) &colorFunc;
+			len += 4;
+			*((uint32_t*) &dst[len]) = (uint32_t) color_flat;
+			len += 4;
+			dst[len++] = 0x68; /* push imm32 */
+			*((uint32_t*) &dst[len]) = minicol;
+			len += 4;
+			dst[len++] = 0xE8; /* call rel32 */
+			*((uint32_t*) &dst[len]) = (uint32_t) &q_p2c - (uint32_t) &dst[len + 4];
+			len += 4;
+			dst[len++] = 0x83; /* add */
+			dst[len++] = 0xC4; /* esp */
+			dst[len++] = 4;
+			dst[len++] = 0xA3; /* mov [abs], eax */
+			*((uint32_t*) &dst[len]) = (uint32_t) color_flat + 1;
+			len += 4;
+			src += sizeof(cmd->colorFlat);
+		} else if(cmd->op == LUMA_VID_COMMAND_SET_COLOR_GRAD) {
+			MiniCol cols[16];
+			for(int i = 15; i >= 0; i--) {
+				unsigned int r = (cmd->colorGrad.r1 * (15 - i) + cmd->colorGrad.r2 * i) / 15;
+				unsigned int g = (cmd->colorGrad.g1 * (15 - i) + cmd->colorGrad.g2 * i) / 15;
+				unsigned int b = (cmd->colorGrad.b1 * (15 - i) + cmd->colorGrad.b2 * i) / 15;
+				cols[i] = ((r & ~7) << 8) | ((g & ~3) << 3) | (b >> 3);
+			}
+			
+			dst[len++] = 0xC7; /* mov [abs], imm32 */
+			dst[len++] = 0x05;
+			*((uint32_t*) &dst[len]) = (uint32_t) &colorFunc;
+			len += 4;
+			*((uint32_t*) &dst[len]) = (uint32_t) &color_grad;
+			len += 4;
+			dst[len++] = 0x83; /* sub esp, imm8 */
+			dst[len++] = 0xEC;
+			dst[len++] = 0x04;
+			for(int i = 0; i < 16; i++) {
+				dst[len++] = 0x66; /* mov [esp], imm16 */
+				dst[len++] = 0xC7;
+				dst[len++] = 0x04;
+				dst[len++] = 0x24;
+				*((uint16_t*) &dst[len]) = cols[i];
+				len += 2;
+				
+				dst[len++] = 0xE8; /* call rel32 */
+				*((uint32_t*) &dst[len]) = (uint32_t) &q_p2c - (uint32_t) &dst[len + 4];
+				len += 4;
+				
+				dst[len++] = 0xA2; /* mov [abs], al */
+				*((uint32_t*) &dst[len]) = (uint32_t) &gradCols[i];
+				len += 4;
+			}
+			dst[len++] = 0x83; /* add esp, imm8 */
+			dst[len++] = 0xC4;
+			dst[len++] = 0x04;
+			src += sizeof(cmd->colorGrad);
+		} else if(cmd->op == LUMA_VID_COMMAND_IMAGE_DEL) {
+			src += sizeof(cmd->imageDel);
+		} else if(cmd->op == LUMA_VID_COMMAND_TEXT) {
+			dst[len++] = 0x68;
+			*((uint32_t*) &dst[len]) = cmd->text.wall;
+			len += 4;
+			dst[len++] = 0x68;
+			*((uint32_t*) &dst[len]) = cmd->text.y;
+			len += 4;
+			dst[len++] = 0x68;
+			*((uint32_t*) &dst[len]) = cmd->text.x;
+			len += 4;
+			dst[len++] = 0x68;
+			*((uint32_t*) &dst[len]) = cmd->text.len;
+			len += 4;
+			dst[len++] = 0x68;
+			*((uint32_t*) &dst[len]) = (uint32_t) &dst[len + 14];
+			len += 4;
+			dst[len++] = 0x68;
+			*((uint32_t*) &dst[len]) = (uint32_t) &dst[len + 9 + cmd->text.len];
+			len += 4;
+			dst[len++] = 0xE9;
+			*((uint32_t*) &dst[len]) = (uint32_t) &text - (uint32_t) &dst[len + 4];
+			len += 4;
+			std_copy(&dst[len], cmd->text.data, cmd->text.len);
+			len += cmd->text.len;
+			dst[len++] = 0x83; /* add esp, imm8 */
+			dst[len++] = 0xC4;
+			dst[len++] = 0x14;
+			src += sizeof(cmd->text) + cmd->text.len;
+		} else if(cmd->op == LUMA_VID_COMMAND_NEW_COLOR_FLAT) {
+			MiniCol minicol = ((cmd->colorFlat.r & ~7) << 8) | ((cmd->colorFlat.g & ~3) << 3) | (cmd->colorFlat.b >> 3);
+			q_addp(minicol);
+			q_reavg();
+			src += sizeof(cmd->colorFlat);
+		} else if(cmd->op == LUMA_VID_COMMAND_NEW_COLOR_GRAD) {
+			MiniCol cols[16];
+			for(int i = 15; i >= 0; i--) {
+				unsigned int r = (cmd->colorGrad.r1 * (15 - i) + cmd->colorGrad.r2 * i) / 15;
+				unsigned int g = (cmd->colorGrad.g1 * (15 - i) + cmd->colorGrad.g2 * i) / 15;
+				unsigned int b = (cmd->colorGrad.b1 * (15 - i) + cmd->colorGrad.b2 * i) / 15;
+				q_addp(cols[i] = ((r & ~7) << 8) | ((g & ~3) << 3) | (b >> 2));
+			}
+			q_reavg();
+			src += sizeof(cmd->colorGrad);
+		}
+	}
+}
+
+static void removeDL(DL *dl) {
+	/*size_t sz = sizeof(*dl) + dl->len;
+	std_move(dl, (void*) ((uintptr_t) dl + sz), (uintptr_t) dl + DLs.idx - (uintptr_t) DLs.buf - sz);
+	memory_barrier();
+	DLs.idx -= sz;*/
+	/* TODO: relocate */
+}
+
+void startc() {
+	FrameBuffer = sys_vpm_map(SYS_MAP_VIRT_ANY, (void*) 0xA0000, 32000);
+	DLs.buf = sys_vpm_map(SYS_MAP_VIRT_ANY, SYS_MAP_PHYS_ANY, 4096);
+	Images.handles = sys_vpm_map(SYS_MAP_VIRT_ANY, SYS_MAP_PHYS_ANY, 4096);
+	
+	q_init();
+	
+	uint32_t canalID = sys_canal_create('LUMA', 'VID\0');
+	
+	uint8_t *buf = (void*) -1;
+	do {
+		buf = sys_canal_accept(canalID);
+	} while(buf == (void*) -1);
+	
+	while(1) {
+		uint32_t meta;
+		sys_signal_wait(SYS_SIGNAL_WAIT_ANY, &meta);
+		memory_barrier();
+		
+		if(meta == 0xFFFFFFFF) {
+			void *f = sys_vpm_map(SYS_MAP_VIRT_ANY, SYS_MAP_PHYS_ANY, 4096);
+			compileCommandList(buf, f);
+			q_reassign();
+			((void(*)()) f)();
+			sys_vpm_unmap(f, 4096);
+		}
+		
+		memory_barrier();
+		sys_signal_send(buf, 1);
+	}
+}

luma/wm/Makefile

@@ -0,0 +1,13 @@
+rwildcard=$(foreach d,$(wildcard $(1:=/*)),$(call rwildcard,$d,$2) $(filter $(subst *,%,$2),$d))
+
+HEADERS := $(call rwildcard,.,*.h)
+SOURCEC := $(call rwildcard,.,*.c)
+OBJECTS := $(patsubst %.c, ../../build/wm/%.o, $(SOURCEC))
+
+mod: $(OBJECTS)
+	i386-elf-ld -pie -T ../../mod.ld -no-dynamic-linker -nostdlib -o main.elf $(OBJECTS) -L./ -lz
+	../../elftomod.py main.elf ../../bootfs/luma_wm.mod modentry:0
+
+../../build/wm/%.o: %.c
+	mkdir -p $(@D)
+	i386-elf-gcc -march=i386 -Isrc -Os -pie -nodefaultlibs -nostdlib -nostartfiles -DEEBIE_MEMSET=__builtin_memset -fms-extensions -fomit-frame-pointer -c -o $@ -ffreestanding -mgeneral-regs-only -std=gnu11 -Wall $<

luma/wm/api.h

@@ -0,0 +1,49 @@
+#pragma once
+
+#include<stdint.h>
+#include<stddef.h>
+
+typedef uint32_t LumaWMObjID;
+typedef uint16_t LumaWMObjKind;
+typedef uint32_t LumaWMObjProperty;
+
+#define LUMA_WM_KIND_CONTAINER 1
+#define LUMA_WM_KIND_LABEL 2
+
+#define LUMA_WM_PROP_TEXT 1
+#define LUMA_WM_PROP_CONTAINER_ORDER 2
+
+typedef struct {
+	void *area;
+} LumaWMClient;
+
+typedef struct {
+	uint16_t cmd;
+	uint16_t sz;
+} LumaWMCmdHeader;
+
+typedef struct __attribute__((packed)) {
+	LumaWMObjID id;
+	LumaWMObjKind kind;
+} LumaWMObjDesc;
+
+typedef struct __attribute__((packed)) {
+	LumaWMCmdHeader header;
+	
+	LumaWMObjDesc descs[];
+} LumaWMRegisterObjects;
+
+typedef struct __attribute__((packed)) {
+	LumaWMCmdHeader header;
+	
+	LumaWMObjID object;
+	
+	LumaWMObjProperty property;
+	
+	uint8_t data[];
+} LumaWMSetProperty;
+
+int luma_wm_client(LumaWMClient*);
+int luma_wm_window_create(LumaWMClient*);
+int luma_wm_register_objects(LumaWMClient*, LumaWMObjDesc*, size_t count);
+int luma_wm_set_property(LumaWMClient*, LumaWMObjID object, LumaWMObjProperty prop, void *data, size_t length);

luma/wm/chars/trans.py

@@ -0,0 +1,28 @@
+#!/usr/bin/env python3
+
+import PIL.Image, struct, bitstring, xml.etree.ElementTree
+
+o = open("../chars.bin", "wb")
+
+xml = xml.etree.ElementTree.parse("l.fnt")
+img = PIL.Image.open("l_0.tga")
+
+for ch in xml.getroot().find("chars").findall("char"):
+	assert int(ch.attrib["width"]) < 16 or int(ch.attrib["height"]) < 16, "Image too big: " + ch.attrib["id"]
+	o.write(struct.pack("BBB",
+		int(ch.attrib["id"]),
+		(int(ch.attrib["width"]) << 4) | int(ch.attrib["height"]),
+		(int(ch.attrib["yoffset"]) << 4) | int(ch.attrib["xadvance"])
+	))
+	a = []
+	for y in range(0, int(ch.attrib["height"])):
+		for x in range(0, int(ch.attrib["width"])):
+			a += [img.getpixel((int(ch.attrib["x"]) + x, int(ch.attrib["y"]) + y)) // 255]
+	a = [a[i * 8 : (i + 1) * 8] for i in range((len(a) + 7) // 8)]
+	while len(a[-1]) < 8:
+		a[-1] += [0]
+	for l in a: l.reverse()
+	a = sum(a, [])
+	bitstring.BitArray(a).tofile(o)
+
+o.close()

luma/wm/eebie/ebml.h

@@ -0,0 +1,28 @@
+#ifndef EEBIE_H
+#define EEBIE_H
+
+#include<stdint.h>
+
+typedef enum EBMLElementType {
+	EBML_SIGNED_INTEGER,
+	EBML_UNSIGNED_INTEGER,
+	EBML_FLOAT4,
+	EBML_FLOAT8,
+	EBML_STRING,
+	EBML_UTF8,
+	EBML_DATE,
+	EBML_TREE,
+	EBML_BINARY,
+} EBMLElementType;
+
+typedef union EBMLPrimitive {
+	int64_t sInt;
+	uint64_t uInt;
+	float flt4;
+	double flt8;
+	const char *string;
+	const uint8_t *binary;
+	uint64_t date;
+} EBMLPrimitive;
+
+#endif

luma/wm/eebie/reader.c

@@ -0,0 +1,203 @@
+#include<stddef.h>
+#include<stdint.h>
+
+#include"ebml.h"
+
+#include"reader.h"
+
+void ebml_reader_init(EBMLReader *this) {
+	EEBIE_MEMSET(this, 0, sizeof(*this));
+	this->state = EBMLRS_WAITING_FOR_ELEMENT_ID;
+	this->currentDepth = 0;
+}
+
+static uint64_t VARINT_MASKS[] = {0, 0x80L, 0xC000L, 0xE00000L, 0xF0000000L, 0xF800000000L, 0xFC0000000000L, 0xFE000000000000L, 0L};
+
+#include"../../std.h"
+static inline void pr(int c) {
+	asm volatile("outb %%al, $0xE9" :: "a"(c) :);
+}
+
+static inline void pri(size_t num) {
+	char buf[25] = {};
+	int i = 16;
+	
+	do {
+		buf[--i] = num % 10;
+		num /= 10;
+	} while(num);
+	
+	for(; i < 16; i++) {
+		pr(buf[i] + '0');
+	}
+}
+
+static inline void prs(const Str16 *str) {
+	for(int i = 0; i < str->len; i++) {
+		pr(str->data[i]);
+	}
+}
+
+static inline void prcs(const char *str) {
+	while(*str) {
+		pr(*str);
+		str++;
+	}
+}
+
+#define I(x) ((uint64_t)(x))
+static int read_varint(const uint8_t *data, size_t length, uint64_t *result) {
+	if(data[0] & 0x80) {
+		*result = data[0];
+		
+		return 1;
+	} else if(data[0] & 0xC0) {
+		if(length < 2) {
+			return 0;
+		}
+		
+		*result = (I(data[0]) << 8) | data[1];
+		
+		return 2;
+	} else if(data[0] & 0xE0) {
+		if(length < 3) {
+			return 0;
+		}
+		
+		*result = (I(data[0]) << 16) | (I(data[1]) << 8) | data[2];
+		
+		return 3;
+	} else if(data[0] & 0xF0) {
+		if(length < 4) {
+			return 0;
+		}
+		
+		*result = (I(data[0]) << 24) | (I(data[1]) << 16) | (I(data[2]) << 8) | data[3];
+		
+		return 4;
+	} else if(data[0] & 0xF8) {
+		if(length < 5) {
+			return 0;
+		}
+		
+		*result = (I(data[0]) << 32) | (I(data[1]) << 24) | (I(data[2]) << 16) | (I(data[3]) << 8) | data[4];
+		
+		return 5;
+	} else if(data[0] & 0xFC) {
+		if(length < 6) {
+			return 0;
+		}
+		
+		*result = (I(data[0]) << 40) | (I(data[1]) << 32) | (I(data[2]) << 24) | (I(data[3]) << 16) | (I(data[4]) << 8) | data[5];
+		
+		return 6;
+	} else if(data[0] & 0xFE) {
+		if(length < 7) {
+			return 0;
+		}
+		
+		*result = (I(data[0]) << 48) | (I(data[1]) << 40) | (I(data[2]) << 32) | (I(data[3]) << 24) | (I(data[4]) << 16) | (I(data[5]) << 8) | data[6];
+		
+		return 7;
+	} else if(data[0] == 0x01) {
+		if(length < 8) {
+			return 0;
+		}
+		
+		*result = (I(data[1]) << 48) | (I(data[2]) << 40) | (I(data[3]) << 32) | (I(data[4]) << 24) | (I(data[5]) << 16) | (I(data[6]) << 8) | data[7];
+		
+		return 8;
+	}
+	
+	return -1;
+}
+
+static int get_varint(const uint8_t *data, size_t length, uint64_t *result) {
+	int ret = read_varint(data, length, result);
+	
+	if(ret >= 0) {
+		*result &= ~VARINT_MASKS[ret];
+	}
+	
+	return ret;
+}
+
+int ebml_reader_feed(EBMLReader *this, const uint8_t *data, size_t length) {
+	if(length == 0) {
+		return 0;
+	}
+	
+	size_t eaten = -1;
+	
+	if(this->state == EBMLRS_WAITING_FOR_ELEMENT_ID) {
+		
+		uint64_t elId;
+		
+		int status = read_varint(data, length, &elId);
+		
+		if(status <= 0) {
+			return status;
+		}
+		
+		this->state = EBMLRS_WAITING_FOR_ELEMENT_LENGTH;
+		this->inside.id = elId;
+		
+		eaten = status;
+		
+	} else if(this->state == EBMLRS_WAITING_FOR_ELEMENT_LENGTH) {
+		
+		uint64_t elLength;
+		
+		int status = get_varint(data, length, &elLength);
+		
+		if(status <= 0) {
+			return status;
+		}
+		
+		this->inside.type = this->eventEnterElement(this, this->inside.id, elLength);
+		
+		if(this->inside.type == EBML_TREE) {
+			this->state = EBMLRS_WAITING_FOR_ELEMENT_ID;
+		} else {
+			this->state = EBMLRS_WAITING_FOR_ELEMENT_DATA;
+			
+			this->inside.length = elLength;
+		}
+		
+		this->idStack[this->currentDepth] = this->inside.id;
+		this->stack[this->currentDepth] = elLength + status;
+		
+		this->currentDepth++;
+		
+		eaten = status;
+		
+	} else if(this->state == EBMLRS_WAITING_FOR_ELEMENT_DATA) {
+	
+		size_t realLen = this->stack[this->currentDepth - 1] < length ? this->stack[this->currentDepth - 1] : length;
+		
+		if(this->eventDataChunk) {
+			this->eventDataChunk(this, data, realLen);
+		}
+		
+		eaten = realLen;
+		
+	}
+	
+	for(int i = 0; i < this->currentDepth; i++) {
+		this->stack[i] -= eaten;
+	}
+	
+	while(this->currentDepth > 0 && this->stack[this->currentDepth - 1] == 0) {
+		
+		if(this->eventExitElement) {
+			this->eventExitElement(this);
+		}
+		
+		this->currentDepth--;
+		
+		this->state = EBMLRS_WAITING_FOR_ELEMENT_ID;
+		
+	}
+	
+	return eaten;
+}

luma/wm/eebie/reader.h

@@ -0,0 +1,48 @@
+#ifndef EEBIE_READER_H
+#define EEBIE_READER_H
+
+#define EBML_READ_MAXIMUM_DEPTH 128
+
+#include"ebml.h"
+
+#include<stddef.h>
+#include<stdint.h>
+
+enum EBMLReaderState {
+	EBMLRS_WAITING_FOR_ELEMENT_ID,
+	EBMLRS_WAITING_FOR_ELEMENT_LENGTH,
+	EBMLRS_WAITING_FOR_ELEMENT_DATA,
+};
+
+struct EBMLReader;
+
+typedef EBMLElementType(*EBMLEventEnterElement)(struct EBMLReader*, uint64_t id, uint64_t length);
+typedef void(*EBMLEventDataChunk)(struct EBMLReader*, const uint8_t *data, size_t length);
+typedef void(*EBMLEventExitElement)(struct EBMLReader*);
+
+typedef struct EBMLReader {
+	enum EBMLReaderState state;
+	union {
+		struct {
+			uint64_t id;
+			uint64_t length;
+			
+			EBMLElementType type;
+		} inside;
+	};
+	
+	int currentDepth;
+	uint64_t stack[EBML_READ_MAXIMUM_DEPTH];
+	uint64_t idStack[EBML_READ_MAXIMUM_DEPTH];
+	
+	EBMLEventEnterElement eventEnterElement;
+	EBMLEventDataChunk eventDataChunk;
+	EBMLEventExitElement eventExitElement;
+	
+	void *ud;
+} EBMLReader;
+
+void ebml_reader_init(EBMLReader *this);
+int ebml_reader_feed(EBMLReader *this, const uint8_t *data, size_t length);
+
+#endif

luma/wm/fileswindow.c

@@ -0,0 +1,219 @@
+#include"fileswindow.h"
+
+#include"../std.h"
+#include"../sys.h"
+#include"../fs/fs.h"
+#include"wm.h"
+#include"tiny.h"
+
+#define dynstr16_ends_with(ds16, cstr) (((ds16)->len < sizeof(cstr) - 1) ? 0 : !std_bytecomp(sizeof(cstr) - 1, (cstr), (ds16)->data + (ds16)->len - sizeof(cstr) + 1))
+
+struct Ctx {
+	Str16 *cd;
+	size_t cdSize;
+} ;
+
+void __attribute__((naked)) audio_thread() {
+	asm(
+		"movl %eax, %ebp\n"
+		"movl $0, %eax\n"
+		"movl $-1, %edi\n"
+		"movl $-1, %esi\n"
+		"movl $4096, %edx\n"
+		"int $0xEC\n"
+		"leal 4096(%eax), %esp\n"
+		"pushl %ebp\n"
+		"call caudio_thread\n"
+	);
+}
+void caudio_thread(void*) {
+	void *FSLink = sys_link_create('FS\0\0', 'ROOT', 4096);
+	
+	LumaFSCommand *cmd = FSLink;
+	cmd->op = LUMA_FS_COMMAND_READ;
+	cmd->read.offset = 44;
+	cmd->read.bytes = 1024 * 1024;
+	Str16 *path = (Str16*) cmd->read.path;
+	std_copy(path->data, "duane.wav", path->len = 9);
+	WALK_STR16(path)->len = 0; //terminator
+	
+	uint8_t *wavebuffer = sys_vpm_map(SYS_MAP_VIRT_ANY, SYS_MAP_PHYS_ANY, 1024 * 1024);
+	size_t wavebufferoffset = 0;
+	while(1) {
+		memory_barrier();
+		sys_signal_send(FSLink, 1);
+		
+		uint32_t dasize;
+		sys_signal_wait(FSLink, &dasize);
+		
+		std_copy(wavebuffer + wavebufferoffset, FSLink, dasize);
+		wavebufferoffset += dasize;
+		
+		if(dasize != 4096) break;
+	}
+	
+	wavebufferoffset = 0;
+	
+	uint8_t *AudLink = sys_link_create('AUD\0', 'HDA0', 8192);
+	
+	while(1) {
+		if(wavebufferoffset == 1024 * 1024) wavebufferoffset = 0;
+		
+		sys_signal_wait(AudLink, NULL);
+		std_copy(AudLink, wavebuffer + wavebufferoffset, 8192);
+		wavebufferoffset += 8192;
+		sys_signal_send(AudLink, 0);
+	}
+}
+
+static void update_listing(struct GWindow *w) {
+	struct Ctx *ctx = w->ud;
+	
+	struct GList *list = (struct GList*) w->child;
+	
+	glist_clear(list);
+	
+	LumaFSCommand *cmd = FSLink;
+	cmd->op = LUMA_FS_COMMAND_TRAVERSE;
+	
+	std_copy(cmd->traverse.path, ctx->cd, ctx->cdSize);
+	
+	sys_signal_send(FSLink, 0);
+	sys_signal_wait(FSLink, NULL);
+	
+	for(LumaFSDirectoryEntry *e = FSLink;; e = (LumaFSDirectoryEntry*) ((uintptr_t) e + sizeof(*e) + e->name.len)) {
+		glist_add(list, g_dynstr16_new(e->name.len, e->name.len, e->name.data), (void*) (uintptr_t) e->type);
+		
+		if(e->type & LUMA_FS_DIR_ENTRY_LAST) {
+			break;
+		}
+	}
+}
+
+static int files_window_view_cmd(void *ud, const DynStr16 *cmdline) {
+	if(!is_free_space()) return 0;
+	
+	struct GWindow *w = ud;
+	
+	struct GList *list = (struct GList*) w->child;
+	
+	struct Ctx *ctx = w->ud;
+	
+	if(list->cursor >= list->childCount) return 0;
+	
+	if(((uintptr_t) (list->children[list->cursor].ud)) & LUMA_FS_DIR_ENTRY_ISDIR) {
+		Str16 *newud = tiny_realloc(ctx->cd, ctx->cdSize + sizeof(*newud) + list->children[list->cursor].text->len);
+		
+		if(newud) {
+			ctx->cd = newud;
+			
+			Str16 *end = (void*) ((uintptr_t) ctx->cd + ctx->cdSize - sizeof(*end));
+			
+			end->len = list->children[list->cursor].text->len;
+			std_copy(end->data, list->children[list->cursor].text->data, end->len);
+			
+			// Terminator.
+			WALK_STR16(end)->len = 0;
+			
+			ctx->cdSize += sizeof(*newud) + list->children[list->cursor].text->len;
+			
+			update_listing(w);
+		} else {
+			// TODO: log
+		}
+	} else if(dynstr16_ends_with(list->children[list->cursor].text, ".txt")) {
+		size_t amount = 100;
+		
+		Str16 *newterm = DYN16_TO_ST(list->children[list->cursor].text);
+		
+		LumaFSCommand *cmd = FSLink;
+		cmd->op = LUMA_FS_COMMAND_READ;
+		cmd->read.offset = 0;
+		cmd->read.bytes = amount;
+		std_copy(cmd->read.path, ctx->cd, ctx->cdSize - sizeof(*newterm));
+		std_copy(cmd->read.path + ctx->cdSize - sizeof(*newterm), newterm, sizeof(*newterm) + newterm->len);
+		WALK_STR16(cmd->read.path + ctx->cdSize - sizeof(*newterm))->len = 0; //terminator
+
+		memory_barrier();
+		sys_signal_send(FSLink, 0);
+		
+		uint32_t dasize;
+		sys_signal_wait(FSLink, &dasize);
+		
+		memory_barrier();
+		
+		DynStr16 *str = g_dynstr16_new(512, dasize, FSLink);
+		
+		struct GText *bigtxt = gtext_new(0, 0, 0, 0, NULL, str);
+		
+		static const char name[] = "Viewer";
+		struct GWindow *viewwnd = gwindow_new(0, 0, 0, 0, g_dynstr16_new(sizeof(name) - 1, sizeof(name) - 1, name), (struct GObj*) bigtxt, 0);
+		
+		add_window(viewwnd);
+		
+		return 1;
+	} else if(dynstr16_ends_with(list->children[list->cursor].text, ".wav")) {
+		static const char name[] = "Audio Player";
+		struct GWindow *playerwnd = gwindow_new(0, 0, 0, 0, g_dynstr16_new(sizeof(name) - 1, sizeof(name) - 1, name), NULL, 0);
+		
+		add_window(playerwnd);
+		
+		sys_create_thread(audio_thread);
+		
+		return 1;
+	}
+	
+	return 0;
+}
+
+static int files_window_back_cmd(void *ud, const DynStr16 *cmdline) {
+	struct GWindow *w = ud;
+	
+	struct Ctx *ctx = w->ud;
+	
+	Str16 *pathPrev = NULL;
+	Str16 *path = ctx->cd;
+	
+	if(path->len == 0) {
+		return 0;
+	}
+	
+	while(path->len) {
+		pathPrev = path;
+		path = WALK_STR16(path);
+	}
+	
+	// Turn term into terminator.
+	ctx->cdSize -= pathPrev->len;
+	pathPrev->len = 0;
+	
+	update_listing(w);
+	
+	return 1;
+}
+
+void FilesWindowCreate() {
+	if(!is_free_space()) return;
+	
+	struct GList *list = glist_new(0, 0, 0, 0);
+	
+	static const char name[] = "File Explorer";
+	struct GWindow *w = gwindow_new(0, 0, 0, 0, g_dynstr16_new(sizeof(name) - 1, sizeof(name) - 1, name), (struct GObj*) list, 0);
+	
+	struct Ctx *ctx = tiny_malloc(sizeof(*ctx));
+	
+	ctx->cd = tiny_malloc(ctx->cdSize = sizeof(*ctx->cd));
+	ctx->cd->len = 0; // root
+	
+	w->ud = ctx;
+	
+	update_listing(w);
+	
+	static const char viewcmd[] = "open";
+	gwindow_regcmd(w, g_str16_new(sizeof(viewcmd) - 1, viewcmd), w, files_window_view_cmd);
+	
+	static const char backcmd[] = "back";
+	gwindow_regcmd(w, g_str16_new(sizeof(backcmd) - 1, backcmd), w, files_window_back_cmd);
+	
+	add_window(w);
+}

luma/wm/fileswindow.h

@@ -0,0 +1,3 @@
+#pragma once
+
+void FilesWindowCreate();

luma/wm/io.h

@@ -0,0 +1,6 @@
+#ifndef LUMA_WM_IO_H
+#define LUMA_WM_IO_H
+
+uint8_t *get_file_from_path(Str16 *name, size_t *len);
+
+#endif

luma/wm/logotrans.py

@@ -0,0 +1,14 @@
+#!/usr/bin/env python3
+
+import PIL
+import PIL.Image
+import bitstring, os
+
+for
+i = PIL.Image.open("logo.png")
+
+o = bitstring.BitArray()
+for y in range(0, i.height):
+	for x in range(0, i.width):
+		o += [i.getpixel((x // 8 * 8 + 7 - x % 8, y))]
+o.tofile(open("logo.bin", "wb"))

luma/wm/logserver.c

@@ -0,0 +1,27 @@
+#include"logserver.h"
+
+#include"../sys.h"
+
+DynStr16 *the;
+
+void log_init() {
+	the = sys_vpm_map(SYS_MAP_VIRT_ANY, SYS_MAP_PHYS_ANY, 4096);
+	the->len = 0;
+	the->cap = 4096 - sizeof(*the);
+}
+
+void log_update(Str16 *next) {
+	if(the->len + next->len > the->cap) {
+		size_t fit = the->len + next->len - the->cap;
+		std_move(the->data, the->data + fit, the->len - fit);
+		the->len -= fit;
+	}
+	
+	std_copy(the->data + the->len, next->data, next->len);
+	
+	the->len += next->len;
+}
+
+DynStr16 *log_get() {
+	return the;
+}

luma/wm/logserver.h

@@ -0,0 +1,7 @@
+#pragma once
+
+#include"../std.h"
+
+void log_init();
+void log_update(Str16*);
+DynStr16 *log_get();

luma/wm/main.c

@@ -0,0 +1,978 @@
+#include<stddef.h>
+#include<stdint.h>
+#include"../sys.h"
+#include"../std.h"
+#include"../vid/api.h"
+#include"../ps2/api.h"
+#include"ui.h"
+#include"perform.h"
+#include"wm.h"
+#include"../fs/fs.h"
+#include"zlib.h"
+#include"logserver.h"
+#include"progdb.h"
+#include"io.h"
+
+asm("modentry:\n"
+"mov $stack + 2048, %esp\n"
+"call startc");
+
+#include"tiny.h"
+
+char stack[2048];
+
+void *VidLink;
+void *FSLink;
+
+static inline void pr(int c) {
+	asm volatile("outb %%al, $0xE9" :: "a"(c) :);
+}
+
+static inline void pri(size_t num) {
+	char buf[16] = {};
+	int i = 16;
+	
+	do {
+		buf[--i] = num % 10;
+		num /= 10;
+	} while(num);
+	
+	for(; i < 16; i++) {
+		pr(buf[i] + '0');
+	}
+}
+
+static inline void prs(const Str16 *str) {
+	for(int i = 0; i < str->len; i++) {
+		pr(str->data[i]);
+	}
+}
+
+static inline void prcs(const char *str) {
+	while(*str) {
+		pr(*str);
+		str++;
+	}
+}
+
+void initvid(int col) {
+	static uint8_t logo[] = {
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xe0, 0x07, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x80,
+		0x07, 0x00, 0xf0, 0x0f, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x0c, 0x00, 0x00, 0xe0, 0x1f, 0x00, 0xf8, 0x1f,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x1c, 0x00, 0x00, 0xf0, 0x1f, 0x00, 0xf8, 0x3f, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x1c, 0x00, 0x00, 0xf8,
+		0x3c, 0x00, 0x78, 0x3e, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x3c, 0x00, 0x00, 0x78, 0x38, 0x00, 0x78, 0x7c,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x38, 0x00, 0x00, 0x3c, 0x78, 0x00, 0x7c, 0x78, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x78, 0x00, 0x00, 0x1c,
+		0x78, 0x00, 0x7c, 0x78, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x70, 0x00, 0x00, 0x1c, 0x78, 0x00, 0x7c, 0x7c,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0xf0, 0x00, 0x00, 0x0e, 0x70, 0x00, 0x7c, 0x3c, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xf0, 0x00, 0x00, 0x0e,
+		0x70, 0x00, 0x78, 0x3c, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0xe0, 0x01, 0x00, 0x0c, 0x70, 0x00, 0x78, 0x3c,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0xe0, 0x01, 0x00, 0x1c, 0x78, 0x00, 0x78, 0x1e, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xc0, 0x03, 0x00, 0x1c,
+		0x78, 0x00, 0x78, 0x1e, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0xc0, 0x03, 0x00, 0x1c, 0x3c, 0x00, 0x78, 0x1f,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0xc0, 0x03, 0x00, 0x38, 0x3e, 0x00, 0xf8, 0x0f, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x80, 0x07, 0x00, 0x38,
+		0x1e, 0x00, 0xf8, 0x0f, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x80, 0x07, 0x00, 0x38, 0x0f, 0x00, 0xf0, 0x07,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x80, 0x07, 0x00, 0xb0, 0x0f, 0x00, 0xf0, 0x07, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x0f, 0x00, 0xf0,
+		0x07, 0x00, 0xf0, 0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x0f, 0x00, 0xf0, 0x03, 0x00, 0xf0, 0x03,
+		0x00, 0xe0, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x0f, 0x00, 0xf0, 0x03, 0x00, 0xf0, 0x01, 0x00, 0xe0, 0x01, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x0f, 0x00, 0xf0,
+		0x01, 0x00, 0xf0, 0x01, 0x00, 0xe0, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x1e, 0x00, 0xf0, 0x00, 0x00, 0xf0, 0x01,
+		0x00, 0xe0, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x1e, 0x00, 0xf8, 0x00, 0x00, 0xf8, 0x01, 0x00, 0xe0, 0x01, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x1e, 0x00, 0xfc,
+		0x00, 0x00, 0xfc, 0x01, 0x00, 0xe0, 0x01, 0x00, 0x1c, 0x80, 0x83, 0x71,
+		0x78, 0x00, 0xfc, 0x01, 0x00, 0x1e, 0x00, 0xfe, 0x01, 0x00, 0xfc, 0x03,
+		0x00, 0xe0, 0x01, 0x00, 0x1c, 0x80, 0x83, 0xfd, 0xfc, 0x00, 0xff, 0x07,
+		0x00, 0x3e, 0x00, 0xdf, 0x01, 0x00, 0xfe, 0x03, 0x00, 0xe0, 0x01, 0x00,
+		0x1c, 0x80, 0x83, 0xc7, 0xe7, 0x81, 0x07, 0x0f, 0x00, 0x3c, 0x00, 0xcf,
+		0x01, 0x00, 0xdf, 0x03, 0x00, 0xe0, 0x01, 0x00, 0x1c, 0x80, 0x83, 0xc7,
+		0xc3, 0xc1, 0x03, 0x0e, 0x00, 0x3c, 0x80, 0xcf, 0x01, 0x00, 0xcf, 0x03,
+		0x00, 0xe0, 0x01, 0x00, 0x1c, 0x80, 0x83, 0xc3, 0xc3, 0xc1, 0x03, 0x1c,
+		0x00, 0x3c, 0xc0, 0x87, 0x03, 0x80, 0xcf, 0x03, 0x00, 0xe0, 0x01, 0x00,
+		0x1c, 0x80, 0x83, 0xc3, 0xc3, 0x01, 0x00, 0x1c, 0x00, 0x3c, 0xe0, 0x83,
+		0x03, 0xc0, 0x87, 0x07, 0x00, 0xe0, 0x01, 0x00, 0x1c, 0x80, 0x83, 0x83,
+		0xc3, 0x01, 0x00, 0x1c, 0x00, 0x3c, 0xf0, 0x81, 0x03, 0xc0, 0x83, 0x07,
+		0x00, 0xe0, 0x01, 0x00, 0x1c, 0x80, 0x83, 0x83, 0xc1, 0x01, 0xfc, 0x1f,
+		0x00, 0x7c, 0xf8, 0x80, 0x03, 0xe0, 0x81, 0x07, 0x00, 0xe0, 0x01, 0x00,
+		0x1c, 0x80, 0x83, 0x83, 0xc1, 0x01, 0xff, 0x1f, 0x00, 0x78, 0xf8, 0x00,
+		0x03, 0xf0, 0x81, 0x07, 0x00, 0xe0, 0x01, 0x00, 0x1c, 0x80, 0x83, 0x83,
+		0xc1, 0xc1, 0x07, 0x1c, 0x00, 0x78, 0x7c, 0x00, 0x07, 0xf8, 0x80, 0x07,
+		0x00, 0xe0, 0x01, 0x00, 0x1c, 0x80, 0x83, 0x83, 0xc1, 0xc1, 0x03, 0x1c,
+		0x00, 0x78, 0x3e, 0x00, 0x07, 0x78, 0x00, 0x07, 0x00, 0xe0, 0x01, 0x00,
+		0x1c, 0x80, 0x83, 0x83, 0xc1, 0xe1, 0x01, 0x1c, 0x00, 0x78, 0x1f, 0x00,
+		0x07, 0x3c, 0x00, 0x0f, 0x00, 0xe0, 0x01, 0x00, 0x1c, 0xc0, 0x83, 0x83,
+		0xc1, 0xe1, 0x01, 0x1e, 0x00, 0xf8, 0x0f, 0x00, 0x06, 0x3e, 0x00, 0x0f,
+		0x00, 0xe0, 0x01, 0x00, 0x3c, 0xc0, 0x83, 0x83, 0xc1, 0xe1, 0x01, 0x1e,
+		0x00, 0xf8, 0x07, 0x00, 0x0e, 0x1f, 0x00, 0x0f, 0x00, 0xe0, 0x01, 0x00,
+		0x3c, 0xc0, 0x83, 0x83, 0xc1, 0xe1, 0x01, 0x1f, 0x00, 0xf8, 0x03, 0x00,
+		0x8e, 0x0f, 0x00, 0x0f, 0x00, 0xe0, 0x01, 0x00, 0x38, 0xe0, 0x83, 0x83,
+		0xc1, 0xc1, 0x01, 0x1d, 0x00, 0xf8, 0x03, 0x00, 0xce, 0x07, 0x00, 0x0e,
+		0x00, 0xe0, 0x01, 0x00, 0x78, 0xb0, 0x83, 0x83, 0xc1, 0xc1, 0xc3, 0x1d,
+		0x00, 0xf8, 0x01, 0x00, 0xee, 0x03, 0x00, 0x0e, 0x00, 0xe0, 0xff, 0x1f,
+		0xf0, 0x9f, 0x83, 0x83, 0xc1, 0x81, 0xff, 0xfc, 0x00, 0xf8, 0x00, 0x00,
+		0xfc, 0x01, 0x00, 0x0e, 0x00, 0xe0, 0xff, 0x1f, 0xe0, 0x87, 0x83, 0x83,
+		0xc1, 0x01, 0x3f, 0xf8, 0x00, 0x7c, 0x00, 0x00, 0xfc, 0x00, 0x00, 0x1e,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0xfe, 0x00, 0x00, 0x7c, 0x00, 0x00, 0x1e, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xff, 0x00, 0x00,
+		0x3e, 0x00, 0x00, 0x1e, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x80, 0xff, 0x00, 0x80, 0x1f, 0x00, 0x00, 0x1c,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0xc0, 0x7f, 0x00, 0xc0, 0x1f, 0x00, 0x00, 0x1c, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xe0, 0x7f, 0x00, 0xf0,
+		0x1f, 0x00, 0x00, 0x1c, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0xf0, 0x7b, 0x00, 0xf8, 0x19, 0x00, 0x00, 0x1c,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0xf8, 0x79, 0x00, 0x7c, 0x18, 0x00, 0x00, 0x1c, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xfc, 0x78, 0x00, 0x3e,
+		0x38, 0x00, 0x00, 0x18, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x7c, 0x78, 0x00, 0x0f, 0x38, 0x00, 0x00, 0x18,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x3e, 0x78, 0x00, 0x07, 0x38, 0x00, 0x00, 0x38, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x1e, 0x7c, 0x80, 0x07,
+		0x38, 0x00, 0x00, 0x38, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x1f, 0x3c, 0x80, 0x03, 0x38, 0x00, 0x00, 0x38,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x0f, 0x3c, 0x80, 0x03, 0x38, 0x00, 0x00, 0x38, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x0f, 0x3e, 0x80, 0x03,
+		0x18, 0x00, 0x00, 0x30, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x0f, 0x3e, 0x80, 0x03, 0x1c, 0x00, 0x00, 0x10,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x1f, 0x1f, 0x80, 0x07, 0x1c, 0x00, 0x00, 0x10, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xfe, 0x1f, 0x00, 0x0f,
+		0x1e, 0x00, 0x00, 0x10, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0xfe, 0x0f, 0x00, 0x1f, 0x0f, 0x00, 0x00, 0x10,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0xfc, 0x07, 0x00, 0xfe, 0x07, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xf8, 0x03, 0x00, 0xfc,
+		0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x40, 0x00, 0x00, 0xf0, 0x00, 0x00, 0x00, 0x00,
+		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00
+	};
+
+	do {
+		VidLink = sys_link_create('LUMA', 'VID\0', 4096);
+	} while(VidLink == (void*) -1);
+	
+	LumaVidCommand *cmd = VidLink;
+	cmd->op = LUMA_VID_COMMAND_IMAGE_NEW;
+	cmd->imageNew.width = 160;
+	cmd->imageNew.height = 66;
+	cmd->imageNew.format = LUMA_VID_IMAGE_FORMAT_BLALPHA;
+	cmd->imageNew.name = 0;
+	cmd = (LumaVidCommand*) ((uintptr_t) cmd + sizeof(cmd->imageNew));
+	cmd->op = LUMA_VID_COMMAND_IMAGE_WRITE;
+	cmd->imageWrite.name = 0;
+	cmd->imageWrite.x = 0;
+	cmd->imageWrite.y = 0;
+	cmd->imageWrite.len = sizeof(logo);
+	std_copy(cmd->imageWrite.data, logo, cmd->imageWrite.len);
+	cmd = (LumaVidCommand*) ((uintptr_t) cmd + sizeof(cmd->imageWrite) + cmd->imageWrite.len);
+	cmd->op = LUMA_VID_COMMAND_NEW_COLOR_GRAD;
+	cmd->colorGrad.r1 = 135;
+	cmd->colorGrad.g1 = 214;
+	cmd->colorGrad.b1 = 255;
+	cmd->colorGrad.r2 = 255;
+	cmd->colorGrad.g2 = 252;
+	cmd->colorGrad.b2 = 135;
+	cmd = (LumaVidCommand*) ((uintptr_t) cmd + sizeof(cmd->colorGrad));
+	cmd->op = LUMA_VID_COMMAND_NEW_COLOR_FLAT;
+	cmd->colorFlat.r = 0x80;
+	cmd->colorFlat.g = 0x80;
+	cmd->colorFlat.b = 0x80;
+	cmd = (LumaVidCommand*) ((uintptr_t) cmd + sizeof(cmd->colorFlat));
+	cmd->op = LUMA_VID_COMMAND_NEW_COLOR_FLAT;
+	cmd->colorFlat.r = 0xF0;
+	cmd->colorFlat.g = 0xF0;
+	cmd->colorFlat.b = 0xF0;
+	cmd = (LumaVidCommand*) ((uintptr_t) cmd + sizeof(cmd->colorFlat));
+	cmd->op = LUMA_VID_COMMAND_NEW_COLOR_FLAT;
+	cmd->colorFlat.r = 0x00;
+	cmd->colorFlat.g = 0x00;
+	cmd->colorFlat.b = 0x00;
+	cmd = (LumaVidCommand*) ((uintptr_t) cmd + sizeof(cmd->colorFlat));
+	cmd->op = LUMA_VID_COMMAND_NEW_COLOR_FLAT;
+	cmd->colorFlat.r = 0x85;
+	cmd->colorFlat.g = 0x98;
+	cmd->colorFlat.b = 0x17;
+	cmd = (LumaVidCommand*) ((uintptr_t) cmd + sizeof(cmd->colorFlat));
+	cmd->op = LUMA_VID_COMMAND_END;
+	
+	memory_barrier();
+	sys_signal_send(VidLink, -1);
+	sys_signal_wait(VidLink, NULL);
+	
+	cmd = VidLink;
+	cmd->op = LUMA_VID_COMMAND_SET_COLOR_GRAD;
+	cmd->colorGrad.r1 = 135;
+	cmd->colorGrad.g1 = 214;
+	cmd->colorGrad.b1 = 255;
+	cmd->colorGrad.r2 = 255;
+	cmd->colorGrad.g2 = 252;
+	cmd->colorGrad.b2 = 135;
+	cmd = (LumaVidCommand*) ((uintptr_t) cmd + sizeof(cmd->colorGrad));
+	cmd->op = LUMA_VID_COMMAND_CLEAR;
+	cmd = (LumaVidCommand*) ((uintptr_t) cmd + sizeof(cmd->clear));
+	cmd->op = LUMA_VID_COMMAND_IMAGE;
+	cmd->image.name = 0;
+	cmd->image.x = 80;
+	cmd->image.y = 80;
+	cmd->image.width = 160;
+	cmd->image.height = 66;
+	cmd->image.subX = cmd->image.subY = 0;
+	cmd = (LumaVidCommand*) ((uintptr_t) cmd + sizeof(cmd->image));
+	cmd->op = LUMA_VID_COMMAND_SET_COLOR_FLAT;
+	cmd->colorFlat.r = 0xF0;
+	cmd->colorFlat.g = 0xF0;
+	cmd->colorFlat.b = 0xF0;
+	cmd = (LumaVidCommand*) ((uintptr_t) cmd + sizeof(cmd->colorFlat));
+	cmd->op = LUMA_VID_COMMAND_TEXT;
+	static const char a[] = "Press any key to continue...";
+	std_copy(cmd->text.data, a, cmd->text.len = sizeof(a) - 1);
+	cmd->text.x = 0;
+	cmd->text.y = 0;
+	cmd->text.wall = 0;
+	cmd = (LumaVidCommand*) ((uintptr_t) cmd + sizeof(cmd->text) + cmd->text.len);
+	cmd->op = LUMA_VID_COMMAND_END;
+	
+	memory_barrier();
+	sys_signal_send(VidLink, -1);
+	sys_signal_wait(VidLink, NULL);
+}
+
+void drawdesk(int x, int y, int w, int h) {
+	LumaVidCommand *cmd = VidLink;
+	cmd->op = LUMA_VID_COMMAND_IMAGE;
+	cmd->image.name = 1;
+	cmd->image.y = cmd->image.subY = y;
+	cmd->image.x = cmd->image.subX = x;
+	cmd->image.height = h;
+	cmd->image.width = w;
+	cmd = (LumaVidCommand*) ((uintptr_t) cmd + sizeof(cmd->image));
+	cmd->op = LUMA_VID_COMMAND_END;
+	
+	memory_barrier();
+	sys_signal_send(VidLink, -1);
+	sys_signal_wait(VidLink, NULL);
+}
+
+struct GContainer *ui;
+
+struct GObj *windowTree;
+struct WTreeNode *windowTreeRoot;
+struct WTreeNode *windowTreeFocus;
+
+struct GText *cmdline;
+
+static void focus_next() {
+	if(windowTreeFocus->window) {
+		windowTreeFocus->window->focus = 0;
+	}
+	
+	if(windowTreeFocus->parent) {
+		if(windowTreeFocus->parent->first == windowTreeFocus) {
+			windowTreeFocus = windowTreeFocus->parent->second;
+		} else if(windowTreeFocus->parent->second == windowTreeFocus) {
+			if(windowTreeFocus->parent->parent) {
+				windowTreeFocus = windowTreeFocus->parent->parent->second;
+			} else {
+				windowTreeFocus = windowTreeRoot;
+			}
+			while(!windowTreeFocus->window) {
+				windowTreeFocus = windowTreeFocus->first;
+			}
+		}
+	}
+	
+	if(windowTreeFocus->window) {
+		windowTreeFocus->window->focus = 1;
+	}
+}
+
+static KeyReactFunc baseinputonkey;
+static void inputonkey(struct GObj *_, struct GKeyEvent ev) {
+	if(ev.scancode == LK_ENTER) {
+		struct GText *t = (void*) _;
+		
+		perform(t->data);
+		
+		t->data->len = 0;
+		
+		cmdline->focus ^= 1;
+		windowTree->focus ^= 1;
+		
+		dirty_obj(_);
+	} else {
+		baseinputonkey(_, ev);
+	}
+}
+
+static void wtree_render_recur(struct WTreeNode *node) {
+	if(node->window) {
+		if(node->window->onRender) {
+			node->window->onRender((struct GObj*) node->window);
+		}
+	} else {
+		if(node->first) wtree_render_recur(node->first);
+		if(node->second) wtree_render_recur(node->second);
+	}
+}
+static void wtree_render(struct GObj *this) {
+	wtree_render_recur(windowTreeRoot);
+}
+static void wtree_key(struct GObj *this, struct GKeyEvent ev) {
+	if(this->focus == 0) return;
+	
+	if(windowTreeFocus && windowTreeFocus->window && windowTreeFocus->window->onKey) {
+		windowTreeFocus->window->onKey((struct GObj*) windowTreeFocus->window, ev);
+	}
+}
+static void wtree_fitter_recur(struct WTreeNode *node, uint16_t x, uint16_t y, uint16_t w, uint16_t h) {
+	if(node->window) {
+		node->window->x = x;
+		node->window->y = y;
+		node->window->w = w;
+		node->window->h = h;
+		node->window->fitter((struct GObj*) node->window);
+	} else {
+		if(node->vertical) {
+			if(node->first) {
+				wtree_fitter_recur(node->first, x, y, w, h / 2);
+			}
+			if(node->second) {
+				wtree_fitter_recur(node->second, x, y + h / 2, w, h / 2);
+			}
+		} else {
+			if(node->first) {
+				wtree_fitter_recur(node->first, x, y, w / 2, h);
+			}
+			if(node->second) {
+				wtree_fitter_recur(node->second, x + w / 2, y, w / 2, h);
+			}
+		}
+	}
+}
+static void wtree_fitter(struct GObj *_) {
+	wtree_fitter_recur(windowTreeRoot, windowTree->x, windowTree->y, windowTree->w, windowTree->h);
+}
+static void wtree_destroy_recur(struct WTreeNode *node) {
+	if(node->window) node->window->destroy((struct GObj*) node->window);
+	else {
+		if(node->first) wtree_destroy_recur(node->first);
+		if(node->second) wtree_destroy_recur(node->second);
+	}
+}
+static void wtree_destroy(struct GObj *_) {
+	wtree_destroy_recur(windowTreeRoot);
+	tiny_free(_);
+}
+static void init_ui() {
+	// Contains window tree and command bar
+	ui = gcont_new(0, 0, 320, 240, NULL);
+	
+	// Contains windows
+	windowTree = tiny_calloc(sizeof(*windowTree), 1);
+	windowTree->onRender = wtree_render;
+	windowTree->onKey = wtree_key;
+	windowTree->destroy = wtree_destroy;
+	windowTree->fitter = wtree_fitter;
+	windowTree->x = windowTree->y = 0;
+	windowTree->w = 320;
+	windowTree->h = 226;
+	windowTree->focus = 0;
+	
+	windowTreeRoot = tiny_calloc(sizeof(*windowTreeRoot), 1);
+	windowTreeRoot->vertical = 0;
+	windowTreeRoot->window = NULL;
+	windowTreeRoot->parent = NULL;
+	windowTreeRoot->first = NULL;
+	windowTreeRoot->second = NULL;
+	windowTreeFocus = windowTreeRoot;
+	
+	gcont_add(ui, 0, (struct GObj*) windowTree);
+	
+	static const char asdf[] = "Enter command, e.g. help";
+	cmdline = gtext_new(0, 226, 320, 14, g_dynstr16_new(sizeof(asdf) - 1, sizeof(asdf) - 1, asdf), NULL);
+	baseinputonkey = cmdline->onKey;
+	cmdline->onKey = inputonkey;
+	cmdline->focus = 1;
+	
+	gcont_add(ui, 0, (struct GObj*) cmdline);
+	
+	ui->focus = 1;
+}
+
+int add_window(struct GWindow *w) {
+	if(!windowTreeFocus) {
+		return 0;
+	}
+	
+	if(windowTreeFocus && windowTreeFocus->window) {
+		windowTreeFocus->window->focus = 0;
+	}
+	
+	if(!windowTreeFocus->window) {
+		windowTreeFocus->window = w;
+	} else {
+		struct WTreeNode *top = tiny_calloc(sizeof(*top), 1);
+		struct WTreeNode *second = tiny_calloc(sizeof(*second), 1);
+		
+		top->vertical = windowTreeFocus->parent ? !windowTreeFocus->parent->vertical : 0;
+		top->window = NULL;
+		top->parent = windowTreeFocus->parent;
+		top->first = windowTreeFocus;
+		top->second = second;
+		
+		second->window = w;
+		second->parent = top;
+		
+		windowTreeFocus->parent = top;
+		
+		if(top->parent) {
+			if(top->parent->first == windowTreeFocus) {
+				top->parent->first = top;
+			} else if(top->parent->second == windowTreeFocus) {
+				top->parent->second = top;
+			} else {
+				// wtf
+			}
+		} else {
+			windowTreeRoot = top;
+		}
+		
+		windowTreeFocus = second;
+	}
+	
+	windowTreeFocus->window->focus = 1;
+	
+	windowTree->fitter(windowTree);
+	
+	return 1;
+}
+int is_free_space() {
+	return 1;
+}
+
+static void destroy_window() {
+	if(!windowTreeFocus) return;
+	
+	if(windowTreeFocus->window) {
+		windowTreeFocus->window->focus = 0;
+	
+		windowTreeFocus->window->destroy((struct GObj*) windowTreeFocus->window);
+		windowTreeFocus->window = NULL;
+		if(windowTreeFocus->parent) {
+			struct WTreeNode *parent = windowTreeFocus->parent;
+			
+			struct WTreeNode *other = NULL;
+			if(parent->first == windowTreeFocus) {
+				other = parent->second;
+			} else if(parent->second == windowTreeFocus) {
+				other = parent->first;
+			}
+			
+			other->parent = parent->parent;
+			
+			std_copy(parent, other, sizeof(struct WTreeNode));
+			
+			tiny_free(windowTreeFocus);
+			
+			windowTreeFocus = parent;
+		}
+	}
+	
+	if(windowTreeFocus->window) {
+		windowTreeFocus->window->focus = 1;
+	}
+	
+	windowTree->fitter(windowTree);
+}
+
+// Next lexicographical permutation of windows
+static void permute() {
+	// TODO.
+}
+
+static size_t path_size(Str16 *p) {
+	size_t ret = 0;
+	
+	while(p->len) {
+		ret += sizeof(*p) + p->len;
+		p = WALK_STR16(p);
+	}
+	
+	ret += sizeof(*p) + p->len;
+	
+	return ret;
+}
+
+uint8_t *get_file_from_path(Str16 *path, size_t *filelen) {
+	LumaFSCommand *fscmd = FSLink;
+	
+	fscmd->op = LUMA_FS_COMMAND_INFO;
+	std_copy(fscmd->info.path, path, path_size(path));
+	
+	memory_barrier();
+	sys_signal_send(FSLink, 1);
+	sys_signal_wait(FSLink, NULL);
+	
+	prcs("Getting ");
+	prs(path);
+	pr('\n');
+	
+	size_t fileSize = fscmd->info.fileSize;
+	uint8_t *ret = sys_vpm_map(SYS_MAP_VIRT_ANY, SYS_MAP_PHYS_ANY, (fileSize + 4095) & ~4095);
+	
+	prcs("Size: ");
+	pri(fileSize);
+	pr('\n');
+	
+	*filelen = fileSize;
+	
+	fscmd->op = LUMA_FS_COMMAND_READ;
+	fscmd->read.offset = 0;
+	fscmd->read.bytes = fileSize;
+	std_copy(fscmd->read.path, path, path_size(path));
+	
+	for(size_t i = 0; i < fileSize; i += 4096) {
+		memory_barrier();
+		sys_signal_send(FSLink, 1);
+		
+		uint32_t meta;
+		sys_signal_wait(FSLink, &meta);
+		
+		std_copy(ret + i, FSLink, meta);
+		
+		if(meta != 4096) break;
+	}
+	
+	return ret;
+}
+
+static void *thezalloc(void *_, unsigned int items, unsigned int size) {
+	return tiny_calloc(items, size);
+}
+static void thezfree(void *_, void *addr) {
+	tiny_free(addr);
+}
+int load_wallpaper() {
+	Str16 *path = __builtin_alloca(sizeof(*path) + 9 + sizeof(*path));
+	std_copy(path->data, "Wallpaper", 9);
+	path->len = 9;
+	WALK_STR16(path)->len = 0; // Terminator.
+	
+	size_t pngsize;
+	uint8_t *png = get_file_from_path(path, &pngsize);
+	
+	if(png[0] != 0x89 || png[1] != 'P' || png[2] != 'N' || png[3] != 'G') {
+		return 0;
+	}
+	
+	png += 8;
+	
+	uint32_t width = 0, height = 0;
+	
+	uint8_t *palette = NULL;
+	
+	z_stream zstrm;
+	zstrm.zalloc = thezalloc;
+	zstrm.zfree = thezfree;
+	zstrm.opaque = NULL;
+	zstrm.avail_in = 0;
+	zstrm.next_in = NULL;
+	if(inflateInit(&zstrm) != Z_OK) {
+		return 0;
+	}
+	
+	uint8_t *rawimg = NULL;
+	
+	while(1) {
+		uint32_t length = BIG_TO_HOST_ENDIAN32(*(uint32_t*) &png[0]);
+		uint8_t *type = &png[4];
+		uint8_t *data = &png[8];
+		
+		if(type[0] == 'I' && type[1] == 'H' && type[2] == 'D' && type[3] == 'R') {
+			if(length != 13) return 0;
+			
+			uint8_t depth = data[8];
+			uint8_t color = data[9];
+			uint8_t interlace = data[12];
+			
+			if(depth != 8) return 0;
+			if(color != 3) return 0; // Must be palette
+			if(interlace != 0) return 0;
+			
+			width = BIG_TO_HOST_ENDIAN32(*(uint32_t*) &data[0]);
+			height = BIG_TO_HOST_ENDIAN32(*(uint32_t*) &data[4]);
+			
+			rawimg = tiny_malloc((width + 1) * height); // +1 because of filter byte in each row
+			
+			zstrm.avail_out = (width + 1) * height;
+			zstrm.next_out = rawimg;
+			
+			LumaVidCommand *cmd = VidLink;
+			cmd->op = LUMA_VID_COMMAND_IMAGE_NEW;
+			cmd->imageNew.name = 1;
+			cmd->imageNew.width = width;
+			cmd->imageNew.height = height;
+			cmd->imageNew.format = LUMA_VID_IMAGE_FORMAT_R8G8B8;
+			cmd = (LumaVidCommand*) ((uintptr_t) cmd + sizeof(cmd->imageNew));
+			cmd->op = LUMA_VID_COMMAND_END;
+			
+			sys_signal_send(VidLink, -1);
+			sys_signal_wait(VidLink, NULL);
+		} else if(type[0] == 'P' && type[1] == 'L' && type[2] == 'T' && type[3] == 'E') {
+			if(length % 3 != 0) return 0;
+			
+			palette = data;
+			
+			LumaVidCommand *cmd = VidLink;
+			for(size_t i = 0; i < length; i += 3) {
+				cmd->op = LUMA_VID_COMMAND_NEW_COLOR_FLAT;
+				cmd->colorFlat.r = data[i + 0];
+				cmd->colorFlat.g = data[i + 1];
+				cmd->colorFlat.b = data[i + 2];
+				cmd = (LumaVidCommand*) ((uintptr_t) cmd + sizeof(cmd->colorFlat));
+			}
+			cmd->op = LUMA_VID_COMMAND_END;
+			
+			sys_signal_send(VidLink, -1);
+			sys_signal_wait(VidLink, NULL);
+		} else if(type[0] == 'I' && type[1] == 'D' && type[2] == 'A' && type[3] == 'T') {
+			zstrm.avail_in = length;
+			zstrm.next_in = data;
+			inflate(&zstrm, Z_NO_FLUSH);
+		} else if(type[0] == 'I' && type[1] == 'E' && type[2] == 'N' && type[3] == 'D') {
+			break;
+		}
+		
+		png += 8 + length + 4;
+	}
+	
+	inflateEnd(&zstrm);
+	
+	uint8_t *ptr = rawimg;
+	
+	for(size_t y = 0; y < height; y++) {
+		LumaVidCommand *cmd = VidLink;
+		cmd->op = LUMA_VID_COMMAND_IMAGE_WRITE;
+		cmd->imageWrite.y = y;
+		cmd->imageWrite.x = 0;
+		cmd->imageWrite.name = 1;
+		cmd->imageWrite.len = width * 3;
+		
+		uint8_t filter = *(ptr++);
+		
+		if(filter == 0) { // None
+			for(size_t x = 0; x < width; x++) {
+				uint8_t *col = &palette[3 * *ptr];
+				
+				cmd->imageWrite.data[x * 3 + 0] = col[0];
+				cmd->imageWrite.data[x * 3 + 1] = col[1];
+				cmd->imageWrite.data[x * 3 + 2] = col[2];
+				
+				ptr++;
+			}
+		} else if(filter == 1) { // Sub
+			for(size_t x = 0; x < width; x++) {
+				if(x > 0) {
+					ptr[0] += ptr[-1];
+				}
+				
+				uint8_t *col = &palette[3 * *ptr];
+				
+				cmd->imageWrite.data[x * 3 + 0] = col[0];
+				cmd->imageWrite.data[x * 3 + 1] = col[1];
+				cmd->imageWrite.data[x * 3 + 2] = col[2];
+				
+				ptr++;
+			}
+		} else if(filter == 2) { // Up
+			for(size_t x = 0; x < width; x++) {
+				if(y > 0) {
+					ptr[0] += ptr[-width - 1];
+				}
+				
+				uint8_t *col = &palette[3 * *ptr];
+				
+				cmd->imageWrite.data[x * 3 + 0] = col[0];
+				cmd->imageWrite.data[x * 3 + 1] = col[1];
+				cmd->imageWrite.data[x * 3 + 2] = col[2];
+				
+				ptr++;
+			}
+		} else if(filter == 3) { // Average
+			for(size_t x = 0; x < width; x++) {
+				uint16_t off = 0;
+				if(x > 0) {
+					off += ptr[-1];
+				}
+				if(y > 0) {
+					off += ptr[-width - 1];
+				}
+				*ptr += off >> 1;
+				
+				uint8_t *col = &palette[3 * *ptr];
+				
+				cmd->imageWrite.data[x * 3 + 0] = col[0];
+				cmd->imageWrite.data[x * 3 + 1] = col[1];
+				cmd->imageWrite.data[x * 3 + 2] = col[2];
+				
+				ptr++;
+			}
+		} else if(filter == 4) { // Paeth prediction
+			for(size_t x = 0; x < width; x++) {
+				// Predict
+				uint8_t a = x == 0				? 0 : *(ptr - 1);			// left
+				uint8_t b = y == 0				? 0 : *(ptr - width - 1);	// up
+				uint8_t c = (x == 0 || y == 0)	? 0 : *(ptr - width - 2);	// up-left
+				
+				uint16_t p = a + b - c;
+				
+				// Absolute differences
+				uint16_t pa = p > a ? p - a : a - p;
+				uint16_t pb = p > b ? p - b : b - p;
+				uint16_t pc = p > c ? p - c : c - p;
+				
+				if(pa <= pb && pa <= pc) {
+					*ptr = a;
+				} else if(pb <= pc) {
+					*ptr = b;
+				} else {
+					*ptr = c;
+				}
+				
+				uint8_t *col = &palette[3 * *ptr];
+				
+				cmd->imageWrite.data[x * 3 + 0] = col[0];
+				cmd->imageWrite.data[x * 3 + 1] = col[1];
+				cmd->imageWrite.data[x * 3 + 2] = col[2];
+				
+				ptr++;
+			}
+		}
+		
+		sys_signal_send(VidLink, -1);
+		sys_signal_wait(VidLink, NULL);
+	}
+	
+	tiny_free(rawimg);
+	
+	return 1;
+}
+
+struct ProgramDB ProgramDB;
+
+static void load_programs() {
+	Str16 *path = __builtin_alloca(sizeof(*path) + 8 + sizeof(*path));
+	std_copy(path->data, "Programs", 8);
+	path->len = 8;
+	WALK_STR16(path)->len = 0; // Terminator.
+	
+	size_t filelen;
+	uint8_t *png = get_file_from_path(path, &filelen);
+	
+	uint8_t *start = png;
+	while(start < png + filelen) {
+		uint8_t *end = std_bytefind(filelen - (start - png), start, '\n');
+		
+		size_t len = end - png;
+		
+		ProgramDB.names = tiny_realloc(ProgramDB.names, sizeof(*ProgramDB.names) * (ProgramDB.count + 1));
+		
+		Str16 *name = tiny_malloc(sizeof(*name) + len);
+		name->len = len;
+		std_copy(name->data, png, len);
+		
+		ProgramDB.names[ProgramDB.count++] = name;
+		
+		start = end + 1;
+	}
+	
+	tiny_free(png);
+}
+
+static int dirtyX1 = INT32_MAX;
+static int dirtyX2 = INT32_MIN;
+static int dirtyY1 = INT32_MAX;
+static int dirtyY2 = INT32_MIN;
+void dirty(int x1, int y1, int w, int h) {
+	int x2 = x1 + w;
+	int y2 = y1 + h;
+	
+	if(dirtyX1 > x1) {
+		dirtyX1 = x1;
+	}
+	
+	if(dirtyX2 < x2) {
+		dirtyX2 = x2;
+	}
+	
+	if(dirtyY1 > y1) {
+		dirtyY1 = y1;
+	}
+	
+	if(dirtyY2 < y2) {
+		dirtyY2 = y2;
+	}
+}
+
+void dirty_obj(struct GObj *o) {
+	dirty(o->x, o->y, o->w, o->h);
+}
+
+void startc() {
+	int connected = 0;
+	for(int try = 0; try < 50; try++) {
+		sys_sleep(1000);
+		if((FSLink = sys_link_create('FS\0\0', 'ROOT', 4096)) != (void*) -1) {
+			connected = 1;
+			break;
+		}
+	}
+	
+	#define ALLOC_AREA_SZ (128 * 1024)
+	tiny_init(sys_vpm_map(SYS_MAP_VIRT_ANY, SYS_MAP_PHYS_ANY, ALLOC_AREA_SZ), ALLOC_AREA_SZ);
+
+	initvid(connected);
+	
+	if(connected == 0) {
+		while(1) {
+			sys_signal_wait(SYS_SIGNAL_WAIT_ANY, NULL);
+		}
+	}
+	
+	load_wallpaper();
+	
+	load_programs();
+	
+	uint32_t wmCanal = sys_canal_create('LUMA', 'WM\0\0');
+	
+	struct LumaKbdEvent *keyBuf = NULL;
+	size_t keyBufIndex = 1;
+	int shift = 0, alt = 0, sys = 0, sysNothingElseWasPressed;
+	
+	init_ui();
+	
+	log_init();
+	
+	drawdesk(0, 0, 320, 240);
+	
+	ui->onRender((struct GObj*) ui);
+	
+	while(1) {
+		KEvent kev;
+		sys_event_wait(&kev);
+		
+		if(kev.type == KEV_TYPE_LINK_REQUEST && kev.canalId == wmCanal) {
+			void *area = sys_canal_accept(wmCanal);
+			
+			uint32_t type;
+			sys_signal_wait(area, &type);
+			
+			// Acknowledge
+			sys_signal_send(area, 0);
+			
+			if(type == 0) {
+				if(keyBuf) {
+					//sys_canal_close(wmCanal);
+				} else {
+					keyBuf = area;
+					keyBuf[0].flag = KEYEV_FLAG_FREE;
+				}
+			}
+		} else if(kev.type == KEV_TYPE_LINK_SIGNAL) {
+			uint32_t meta = kev.meta;
+			
+			if(kev.area == keyBuf) {
+				int doRedraw = 0;
+				
+				for(; keyBufIndex != meta; keyBufIndex = (keyBufIndex + 1) % KEYBUF_MAX) {
+					if(keyBuf[keyBufIndex].scancode) {
+						doRedraw = 1;
+						
+						if(alt && keyBuf[keyBufIndex].flag == KEYEV_FLAG_PRESS) {
+							if(keyBuf[keyBufIndex].scancode == LK_P) {
+								permute();
+							} else if(keyBuf[keyBufIndex].scancode == LK_ESCAPE) {
+								destroy_window();
+							} else if(keyBuf[keyBufIndex].scancode == LK_RIGHT) {
+								focus_next();
+							}
+						} else {
+							ui->onKey((struct GObj*) ui, (struct GKeyEvent) {.scancode = keyBuf[keyBufIndex].scancode, .press = keyBuf[keyBufIndex].flag == KEYEV_FLAG_PRESS, .shift = shift, .alt = alt});
+						}
+						
+						if(keyBuf[keyBufIndex].scancode == LK_LEFT_SHIFT) {
+							if(keyBuf[keyBufIndex].flag == KEYEV_FLAG_PRESS) {
+								shift |= SHIFT_LEFT;
+							} else {
+								shift &= ~SHIFT_LEFT;
+							}
+						} else if(keyBuf[keyBufIndex].scancode == LK_RIGHT_SHIFT) {
+							if(keyBuf[keyBufIndex].flag == KEYEV_FLAG_PRESS) {
+								shift |= SHIFT_RIGHT;
+							} else {
+								shift &= ~SHIFT_RIGHT;
+							}
+						} else if(keyBuf[keyBufIndex].scancode == LK_LEFT_ALT) {
+							if(keyBuf[keyBufIndex].flag == KEYEV_FLAG_PRESS) {
+								alt |= ALT_LEFT;
+								
+								sysNothingElseWasPressed = 1;
+							} else {
+								alt &= ~ALT_LEFT;
+								
+								if(sysNothingElseWasPressed) {
+									cmdline->focus ^= 1;
+									windowTree->focus ^= 1;
+								}
+							}
+						} else if(keyBuf[keyBufIndex].scancode == LK_LEFT_SYS) {
+							if(keyBuf[keyBufIndex].flag == KEYEV_FLAG_PRESS) {
+								sys |= SYS_LEFT;
+							} else {
+								sys &= ~SYS_LEFT;
+							}
+						}
+					}
+					
+					// Mark as received.
+					keyBuf[keyBufIndex].flag = KEYEV_FLAG_FREE;
+				}
+				
+				if(doRedraw) {
+					if(dirtyX1 < dirtyX2 && dirtyY1 < dirtyY2) {
+						drawdesk(dirtyX1, dirtyY1, dirtyX2 - dirtyX1, dirtyY2 - dirtyY1);
+						dirtyX1 = INT32_MAX;
+						dirtyX2 = INT32_MIN;
+						dirtyY1 = INT32_MAX;
+						dirtyY2 = INT32_MIN;
+					}
+					ui->onRender((struct GObj*) ui);
+				}
+				
+				memory_barrier();
+			}
+		}
+	}
+}

luma/wm/perform.c

@@ -0,0 +1,324 @@
+#include"perform.h"
+
+#include"../fs/fs.h"
+#include"wm.h"
+#include"../sys.h"
+#include"logserver.h"
+#include"progdb.h"
+#include"eebie/reader.h"
+#include"tiny.h"
+#include"io.h"
+#include"fileswindow.h"
+
+static inline void pr(int c) {
+	asm volatile("outb %%al, $0xE9" :: "a"(c) :);
+}
+
+static inline void pri(size_t num) {
+	char buf[16] = {};
+	int i = 16;
+	
+	do {
+		buf[--i] = num % 10;
+		num /= 10;
+	} while(num);
+	
+	for(; i < 16; i++) {
+		pr(buf[i] + '0');
+	}
+}
+
+static inline void prs(const Str16 *str) {
+	for(int i = 0; i < str->len; i++) {
+		pr(str->data[i]);
+	}
+}
+
+static inline void prcs(const char *str) {
+	while(*str) {
+		pr(*str);
+		str++;
+	}
+}
+
+__attribute__((optimize("Ofast"))) static void b2hex(uint8_t b, char *ret) {
+	uint8_t top = b >> 4;
+	uint8_t bottom = b & 0x0F;
+	
+	ret[0] = top < 0xA ? (top + '0' - 0) : (top + 'A' - 0xA);
+	ret[1] = bottom < 0xA ? (bottom + '0' - 0) : (bottom + 'A' - 0xA);
+}
+
+static void show_help_window() {
+	if(!is_free_space()) return;
+	
+	static const char txt[] = "Luma's interface is designed around the keyboard. In fact, you are unlikely to find any use for a mouse except in niche cases.\n\nTurn to the command bar below by pressing the System (Sys) key. Start off by exploring Luma's software by entering \"list\" or your filesystem with \"files\".\n\nSys+H and Sys+V split the desktop to fit more windows. Use Sys+Escape to close windows or remove free space. Sys+P moves windows around.";
+	struct GLabel *l = glabel_new(0, 0, 0, 0, g_dynstr16_new(sizeof(txt) - 1, sizeof(txt) - 1, txt), 1);
+	
+	static const char name[] = "Welcome to Luma";
+	struct GWindow *w = gwindow_new(0, 0, 0, 0, g_dynstr16_new(sizeof(name) - 1, sizeof(name) - 1, name), (struct GObj*) l, 0);
+	
+	add_window(w);
+}
+
+static void show_log_window() {
+	if(!is_free_space()) return;
+	
+	struct GLabel *l = glabel_new(0, 0, 0, 0, log_get(), 0);
+	
+	static const char name[] = "System Logs";
+	struct GWindow *w = gwindow_new(0, 0, 0, 0, g_dynstr16_new(sizeof(name) - 1, sizeof(name) - 1, name), (struct GObj*) l, 0);
+	
+	add_window(w);
+}
+
+static void e9(const char *buf, size_t len) {
+	while(len--) {
+		asm volatile("out %%al, $0xE9" :: "a"(*buf) :);
+		buf++;
+	}
+}
+
+static void create_process_from_file(const uint8_t *exefile, size_t exesize) {
+	size_t childHandle = sys_spawn();
+	
+	EBMLReader *rdr = tiny_malloc(sizeof(*rdr));
+	ebml_reader_init(rdr);
+	
+	struct Section {
+		uint8_t *data;
+		size_t dataSz;
+		
+		uint8_t *their;
+		
+		uint8_t *relocs;
+		size_t relocsSz;
+	};
+	
+	struct Ctx {
+		uint8_t *buf;
+		size_t size;
+		
+		struct Section section[4];
+		int sIdx;
+		
+		int currentSymbolIsEntry;
+		
+		int entrySection;
+		size_t entryOffset;
+	} ctx = {};
+	
+	ctx.entrySection = -1; // i.e. no entry symbol
+	
+	rdr->ud = &ctx;
+	
+	EBMLElementType eventer(EBMLReader *rdr, uint64_t id, uint64_t length) {
+		struct Ctx *ctx = rdr->ud;
+		
+		ctx->buf = NULL;
+		ctx->size = 0;
+		
+		if(id == 0x10000000 || id == 0x6200 || id == 0xC0) {
+			return EBML_TREE;
+		}
+		
+		return EBML_BINARY;
+	}
+	void evchu(EBMLReader *rdr, const uint8_t *data, size_t length) {
+		struct Ctx *ctx = rdr->ud;
+		
+		ctx->buf = tiny_realloc(ctx->buf, ctx->size + length);
+		std_copy(ctx->buf + ctx->size, data, length);
+		ctx->size += length;
+	}
+	void evexit(EBMLReader *rdr) {
+		struct Ctx *ctx = rdr->ud;
+		
+		size_t exitingId = rdr->idStack[rdr->currentDepth - 1];
+		
+		if(exitingId == 0x6201) {
+			ctx->section[ctx->sIdx].data = ctx->buf;
+			ctx->section[ctx->sIdx].dataSz = ctx->size;
+			
+			ctx->buf = NULL;
+		} else if(exitingId == 0x6202) {
+			ctx->section[ctx->sIdx].relocs = ctx->buf;
+			ctx->section[ctx->sIdx].relocsSz = ctx->size;
+			
+			ctx->buf = NULL;
+		} else if(exitingId == 0x6200) {
+			uint8_t *my = SYS_MAP_VIRT_ANY;
+			uint8_t *their = SYS_MAP_VIRT_ANY;
+			
+			sys_abuse_map(childHandle, &my, &their, SYS_MAP_PHYS_ANY, ctx->section[ctx->sIdx].dataSz);
+			
+			std_copy(my, ctx->section[ctx->sIdx].data, ctx->section[ctx->sIdx].dataSz);
+			
+			tiny_free(ctx->section[ctx->sIdx].data);
+			
+			ctx->section[ctx->sIdx].data = my;
+			ctx->section[ctx->sIdx].their = their;
+			
+			ctx->sIdx++;
+		}
+		
+		if(exitingId == 0xC0) {
+			ctx->currentSymbolIsEntry = 0;
+		} else if(exitingId == 0xC1) {
+			if(ctx->size == 12) {
+				if(ctx->buf[0] == 'P' && ctx->buf[1] == 'r' && ctx->buf[2] == 'o' && ctx->buf[3] == 'g' && ctx->buf[4] == 'r' && ctx->buf[5] == 'a' && ctx->buf[6] == 'm' && ctx->buf[7] == 'E' && ctx->buf[8] == 'n' && ctx->buf[9] == 't'
+					&& ctx->buf[10] == 'r' && ctx->buf[11] == 'y') {
+					
+					ctx->currentSymbolIsEntry = 1;
+				}
+			}
+		} else if(exitingId == 0xC2) {
+			if(ctx->currentSymbolIsEntry) {
+				ctx->entryOffset = *(uint32_t*) &ctx->buf[0];
+				ctx->entrySection = *(uint8_t*) &ctx->buf[4];
+			}
+		}
+		
+		if(ctx->buf) {
+			tiny_free(ctx->buf);
+		}
+	}
+	
+	rdr->eventEnterElement = eventer;
+	rdr->eventDataChunk = evchu;
+	rdr->eventExitElement = evexit;
+	
+	uint8_t *exefilestart = exefile;
+	
+	while(1) {
+		int readCount = ebml_reader_feed(rdr, exefile, exesize);
+		
+		if(readCount <= 0) {
+			break;
+		}
+		
+		exefile += readCount;
+		exesize -= readCount;
+	}
+	
+	// Perform relocations
+	for(int s = 0; s < ctx.sIdx; s++) {
+		for(int r = 0; r < ctx.section[s].relocsSz / 5; r++) {
+			size_t offset = *(uint32_t*) &ctx.section[s].relocs[r * 5 + 0];
+			int refSec = *(uint8_t*) &ctx.section[s].relocs[r * 5 + 4];
+			
+			*(uint32_t*) &ctx.section[s].data[offset] += (uint32_t) ctx.section[refSec].their;
+		}
+		
+		tiny_free(ctx.section[s].relocs);
+		ctx.section[s].relocs = NULL;
+		
+		sys_vpm_unmap(ctx.section[s].data, ctx.section[s].dataSz); // I'm scared this will break
+	}
+	
+	sys_abuse_state(childHandle, SYS_STATE_X86_EIP, ctx.section[ctx.entrySection].their + ctx.entryOffset);
+	
+	sys_release(childHandle);
+	
+	tiny_free(rdr);
+}
+
+static void attempt_program_launch(const uint8_t *str, size_t len) {
+	size_t p;
+	
+	for(p = 0; p < ProgramDB.count; p++) {
+		Str16 *name = ProgramDB.names[p];
+		
+		uint8_t *displayname = name->data;
+		uint8_t *codename = std_bytefind(name->len, name->data, ';') + 1;
+		
+		size_t testlen = name->len - (codename - displayname);
+		if(testlen > len) {
+			testlen = len;
+		}
+		
+		if(std_bytecomp(testlen, codename, str) == 0) {
+			break;
+		}
+	}
+	
+	if(p == ProgramDB.count) {
+		return;
+	}
+	
+	size_t exesize;
+	uint8_t *exefile;
+	{
+		Str16 *name = ProgramDB.names[p];
+		
+		uint8_t *displayname = name->data;
+		uint8_t *term = std_bytefind(name->len, name->data, ';');
+		
+		Str16 *directory = __builtin_alloca(sizeof(*directory) + (term - displayname) + sizeof(*directory) + 7 + sizeof(*directory));
+		std_copy(directory->data, displayname, term - displayname);
+		directory->len = term - displayname;
+		
+		Str16 *fn = WALK_STR16(directory);
+		fn->len = 7;
+		fn->data[0] = 'P';
+		fn->data[1] = 'r';
+		fn->data[2] = 'o';
+		fn->data[3] = 'g';
+		fn->data[4] = 'r';
+		fn->data[5] = 'a';
+		fn->data[6] = 'm';
+		
+		// Terminator.
+		WALK_STR16(fn)->len = 0;
+		
+		exefile = get_file_from_path(directory, &exesize);
+	}
+	
+	create_process_from_file(exefile, exesize);
+	
+	tiny_free(exefile);
+	
+	/*size_t handle = sys_spawn();
+	
+	uint8_t *code = sys_abuse_map(handle, SYS_MAP_VIRT_ANY, 0x40000000, SYS_MAP_PHYS_ANY, 4096);
+	code[0] = 0x66;
+	code[1] = 0x87;
+	code[2] = 0xDB;
+	code[3] = 0xEB;
+	code[4] = 0xFB;
+	
+	sys_abuse_state(handle, SYS_STATE_X86_EIP, 0x40000000);
+	
+	sys_release(handle);*/
+}
+
+void perform(DynStr16 *cmd) {
+	if(cmd->len == 4 && cmd->data[0] == 'h' && cmd->data[1] == 'e' && cmd->data[2] == 'l' && cmd->data[3] == 'p') {
+		show_help_window();
+	} else if(cmd->len == 5 && cmd->data[0] == 'f' && cmd->data[1] == 'i' && cmd->data[2] == 'l' && cmd->data[3] == 'e' && cmd->data[4] == 's') {
+		FilesWindowCreate();
+	} else if(cmd->len == 4 && cmd->data[0] == 'l' && cmd->data[1] == 'o' && cmd->data[2] == 'g' && cmd->data[3] == 's') {
+		show_log_window();
+	} else if(cmd->len > 2 && cmd->data[0] == 'g' && cmd->data[1] == 'o' && cmd->data[2] == ' ') {
+		attempt_program_launch(cmd->data + 3, cmd->len - 3);
+	} else if(windowTreeFocus && windowTreeFocus->window) {
+		uint8_t *space = (uint8_t*) std_bytefind(cmd->len, (char*) cmd->data, ' ');
+		
+		size_t end = space ? space - cmd->data : cmd->len;
+		for(size_t i = 0; i < windowTreeFocus->window->cmds.count; i++) {
+			if(windowTreeFocus->window->cmds.array[i].name->len != end) {
+				continue;
+			}
+			
+			if(std_bytecomp(end, cmd->data, windowTreeFocus->window->cmds.array[i].name->data) != 0) {
+				continue;
+			}
+			
+			windowTreeFocus->window->cmds.array[i].run(windowTreeFocus->window->cmds.array[i].ud, cmd);
+			
+			break;
+		}
+	} else {
+		// Failed.
+	}
+}

luma/wm/perform.h

@@ -0,0 +1,5 @@
+#pragma once
+
+#include"../std.h"
+
+void perform(DynStr16*);

luma/wm/progdb.h

@@ -0,0 +1,8 @@
+#pragma once
+
+#include"../std.h"
+
+extern struct ProgramDB {
+	size_t count;
+	Str16 **names;
+} ProgramDB;

luma/wm/tiny.c

@@ -0,0 +1,367 @@
+#include "tiny.h"
+#include <stdint.h>
+#include"../std.h"
+
+// Casts a size_t to its closest aligned size
+#define ALIGN_SIZE(size) ((size + ALIGNMENT - 1) & ~(ALIGNMENT - 1))
+
+// Casts an unsigned char * to its closest aligned pointer
+#define ALIGN_PTR(ptr) (unsigned char *)(((uintptr_t)ptr + ALIGNMENT - 1) & ~(ALIGNMENT - 1))
+
+#ifdef TINY_ALIGNMENT
+// Alignment is provided manually
+struct tiny_max_align{ char c; TINY_ALIGNMENT align; };
+#else
+// Alignment is automatically calculated (requires max_align_t)
+#define TINY_ALIGNMENT max_align_t
+struct tiny_max_align{ char c; max_align_t align; };
+#endif
+enum { ALIGNMENT = offsetof(struct tiny_max_align, align) };
+
+#define EXPAND1(X) #X
+#define EXPAND2(X) EXPAND1(X)
+#define ALIGNED_TYPE EXPAND2(TINY_ALIGNMENT)
+
+// Defines how many blocks a size_t takes 
+enum { HEADER_BLOCKS = ALIGN_SIZE(sizeof(size_t)) / ALIGNMENT };
+
+// Defines the upper bit of the header as a the taken flag
+#define TAKEN_BIT ((size_t)-1 ^ (((size_t)-1)>>1))
+
+// Defines blocks as arrays with ALIGNMENT bytes
+typedef unsigned char tiny_block[ALIGNMENT];
+
+#ifdef TINY_BUFFER
+// Statically declares and initialises a buffer. This allows the library to be
+// used before `tiny_init()` is called.
+
+// Defines an union padded at the end to fit alignment that can hold a size_t
+union tiny_padded_size { size_t size; tiny_block padding[HEADER_BLOCKS]; };
+
+// Defines an union that contains the buffer and is aligned to alignof(TINY_ALIGN)
+static union tiny_aligned_buffer {
+    // Splits the buffer into header, content and footer
+    struct tiny_main_layout {
+        union tiny_padded_size header;
+        union tiny_padded_size content[TINY_BUFFER / ALIGNMENT / HEADER_BLOCKS - 2];
+        union tiny_padded_size  footer;
+    } layout;
+
+    // Taking the address of this member allows to read the layout as an array
+    union tiny_padded_size buffer; 
+
+    // Aligns the buffer to the boundary defined by alignof(TINY_ALIGN)
+    TINY_ALIGNMENT align;
+} tiny_buffer = { {
+    { (size_t)(TINY_BUFFER / ALIGNMENT - 2 * HEADER_BLOCKS) },
+    { { 0 } },
+    { (size_t)TAKEN_BIT }
+} };
+
+// Initialises the library with the statically allocated buffer
+#define TINY_INITIAL {                                  \
+    (tiny_block *)&tiny_buffer.buffer,                  \
+    (TINY_BUFFER / ALIGNMENT - 2 * HEADER_BLOCKS),      \
+    false,                                              \
+    {                                                   \
+        TINY_LOAD,                                      \
+        true,                                           \
+        (TINY_BUFFER / ALIGNMENT - 2 * HEADER_BLOCKS)   \
+    }                                                   \
+}
+#else
+// Initialised the library with no allocated buffer.
+#define TINY_INITIAL { NULL, 0, false, { TINY_LOAD, true, 0 } }
+#endif
+
+
+// Describes a section of the buffer that may or may not be taken
+typedef struct tiny_block_section {
+    bool taken; // Marks if this section is currently in use by the programmer
+    size_t size; // Specifies how many blocks are there in this section
+    tiny_block *header; // The address of the section header
+    void *data; // The address of the section data
+} tiny_block_section;
+
+// The main library context
+static struct tiny {
+    tiny_block *buffer; // The buffer to operate on
+    size_t size;    // The minimum amount of blocks available for allocation
+    bool out_of_memory; // Whether should the library fake an out-of-memory situation
+    tiny_operation last_operation; // Stores the last operation executed
+} tiny = TINY_INITIAL;
+
+// Writes a header size and availability
+static void write_header(tiny_block *header, size_t size, bool taken) {
+    *(size_t *)header = taken ? TAKEN_BIT | size : size;
+}
+
+// Parses a header and returns the parsed information
+static tiny_block_section read_header(tiny_block *header) {
+    size_t header_value = *(size_t *)header;
+    tiny_block_section section = { 
+        header_value & TAKEN_BIT,
+        header_value & ~TAKEN_BIT,
+        header,
+        (void *)(header + HEADER_BLOCKS) 
+    };
+    return section;
+}
+
+// Allocates some blocks of memory in the provided section.
+// If the section is bigger than necessary, it may be split and a new section
+// with the remaining space may be created.
+static void allocate_at(tiny_block_section section, size_t block_count) {
+    size_t remaining_space = 
+        section.size - block_count;
+
+    if(remaining_space <= HEADER_BLOCKS) {
+        write_header(section.header, section.size, true);
+    } else {
+        write_header(section.header, block_count, true);
+        write_header(
+            section.header + block_count + HEADER_BLOCKS, 
+            remaining_space - HEADER_BLOCKS, 
+            false
+        );
+    }
+}
+
+// Returns the address of the next section
+static tiny_block *next_section(tiny_block *header) {
+    tiny_block_section info = read_header(header);
+    return header + info.size + HEADER_BLOCKS;
+}
+
+// Stores the last operation performed by the library into the main context
+static void store_operation(enum tiny_function function, bool success, size_t size) {
+    tiny_operation op = { function, success, size };
+    tiny.last_operation = op;
+}
+
+// Initialises the library with a buffer.
+// This will partition the buffer accordingly and allow allocating and
+// deallocating memory from it.
+void tiny_init(unsigned char *buffer, size_t size) {
+    unsigned char *aligned = ALIGN_PTR(buffer); 
+    size_t lost_alignment = aligned - buffer;
+
+    if(lost_alignment + (2 * HEADER_BLOCKS + 1) * ALIGNMENT >= size) {
+        store_operation(TINY_INIT, false, size);
+        return;
+    } 
+
+    tiny.buffer = (tiny_block *)aligned;
+    tiny.size = (size - lost_alignment) / ALIGNMENT - 2 * HEADER_BLOCKS;
+
+    write_header(&tiny.buffer[0], tiny.size, false);
+    write_header(&tiny.buffer[tiny.size + HEADER_BLOCKS], 0, true);
+    store_operation(TINY_INIT, true, size);
+}
+
+// Clears the library buffer
+void tiny_clear() {
+    tiny.buffer = NULL;
+    tiny.size = 0;
+    store_operation(TINY_CLEAR, true, 0);
+}
+
+// Resets the library buffer to its initial value
+void tiny_reset() {
+    #ifdef TINY_BUFFER
+    tiny.buffer = (tiny_block *)&tiny_buffer.buffer;
+    tiny.size = (TINY_BUFFER / ALIGNMENT - 2 * HEADER_BLOCKS);
+    #else
+    tiny.buffer = NULL;
+    tiny.size = 0;
+    #endif
+    store_operation(TINY_RESET, true, tiny.size);
+}
+
+// Sets the out-of-memory flag status. If set, calls to `malloc` and similar will
+// return NULL always.
+void tiny_out_of_memory(bool out_of_memory) {
+    tiny.out_of_memory = out_of_memory;
+}
+
+// Returns the last operation performed by the library.
+tiny_operation tiny_last_operation() {
+    return tiny.last_operation;
+}
+
+// Gets the size of each block in the buffer (also, the library alignment)
+size_t tiny_block_size() { return ALIGNMENT; }
+
+
+// Prints a summary of the library
+void tiny_print(bool summary, bool last_op, bool heap) {
+}
+
+
+tiny_summary tiny_inspect() {
+    #ifdef TINY_BUFFER 
+    void *static_buffer = (void *)&tiny_buffer;
+    size_t static_buffer_size = TINY_BUFFER;
+    #else
+    void *static_buffer = NULL;
+    size_t static_buffer_size = 0;
+    #endif
+
+    size_t free_blocks = 0, taken_blocks = 0;
+    size_t total_sections = 0, free_sections = 0, taken_sections = 0;
+
+    tiny_block *header = &tiny.buffer[0];
+    if(header) {
+        tiny_block_section section = read_header(header);
+        while(section.size > 0) {
+            total_sections++;
+            if(section.taken) {
+                taken_sections++;
+                taken_blocks += section.size;
+            } else {
+                free_sections++;
+                free_blocks += section.size;
+            }
+            header = next_section(header);
+            section = read_header(header);
+        }
+    }
+
+    tiny_summary summ = {
+        ALIGNMENT,
+        ALIGNED_TYPE,
+        static_buffer,
+        static_buffer_size,
+        tiny.out_of_memory,
+        tiny.buffer,
+        { tiny.size, tiny.size * ALIGNMENT },
+        { free_blocks, free_blocks * ALIGNMENT },
+        { taken_blocks, taken_blocks * ALIGNMENT },
+        { total_sections, free_sections, taken_sections }
+    };
+    return summ;
+}
+
+
+tiny_section tiny_next_section(void *previous_header) {
+    if(tiny.buffer == NULL) {
+        tiny_section info = { false, NULL, NULL, { 0, 0 } };
+        return info;    
+    }
+
+    tiny_block *header = previous_header ? 
+        next_section(previous_header) : 
+        &tiny.buffer[0];
+    tiny_block_section section = read_header(header);
+    tiny_section info = {
+        section.taken,
+        section.size != 0 ? (void *)header : NULL,
+        section.size != 0 ? (void *)(header + HEADER_BLOCKS) : NULL,
+        { section.size, section.size * ALIGNMENT }
+    };
+    return info;
+}
+
+void *tiny_malloc(size_t size) {
+    if(tiny.out_of_memory || tiny.buffer == NULL || size == 0) {
+        store_operation(TINY_MALLOC, false, size);
+        return NULL;
+    }
+    
+    size_t aligned_size = ALIGN_SIZE(size);
+    if(aligned_size < size) {
+        store_operation(TINY_MALLOC, false, size);
+        return NULL;
+    }
+
+    size_t blocks_required = aligned_size / ALIGNMENT;
+    tiny_block *header = &tiny.buffer[0];
+    tiny_block_section section = read_header(header);
+    while(section.size > 0) {
+        if(!section.taken && section.size >= blocks_required) {
+            allocate_at(section, blocks_required);
+            store_operation(TINY_MALLOC, true, size);
+            return section.data;
+        }
+        header = next_section(header);
+        section = read_header(header);
+    } 
+    store_operation(TINY_MALLOC, false, size);
+    return NULL;
+}
+
+void *tiny_realloc(void *ptr, size_t size) {
+    if(ptr == NULL) {
+        void *data = tiny_malloc(size);
+        store_operation(TINY_REALLOC, data != NULL, size);
+        return data;
+    }
+    if(tiny.out_of_memory || tiny.buffer == NULL || size == 0) {
+        store_operation(TINY_REALLOC, false, size);
+        return NULL;
+    }
+
+    size_t blocks_required = ALIGN_SIZE(size) / ALIGNMENT;
+    tiny_block *header = (tiny_block *)ptr - HEADER_BLOCKS;
+    tiny_block_section section = read_header(header);
+    tiny_block *next = next_section(header);
+    tiny_block_section next_section = read_header(next);
+
+    if(!next_section.taken && next_section.size >= blocks_required - section.size + HEADER_BLOCKS) {
+        write_header(header, section.size + next_section.size + HEADER_BLOCKS, true);
+        allocate_at(read_header(header), blocks_required);
+        store_operation(TINY_REALLOC, true, size);
+        return  ptr;
+    } else {
+        void *new_block = tiny_malloc(size);
+        if(new_block) {
+            std_copy(new_block, section.data, section.size * ALIGNMENT);
+            tiny_free(ptr);
+        }
+        store_operation(TINY_REALLOC, new_block != NULL, size);
+        return new_block;
+    }
+}
+
+void *tiny_calloc(size_t num, size_t size) {
+    size_t full_size = num * size;
+    if(size == 0 || num == 0 || full_size / num != size) {
+        store_operation(TINY_CALLOC, false, size);
+        return NULL;
+    }
+    void *data = tiny_malloc(full_size);
+    if(data != NULL) {
+        std_w8(data, 0, full_size);
+    }
+    store_operation(TINY_CALLOC, data != NULL, num * size);
+    return data;
+}
+
+void tiny_free(void *ptr) {
+    if(ptr == NULL || tiny.buffer == NULL) {
+        store_operation(TINY_FREE, false, 0);
+        return;
+    }
+
+    tiny_block *current = (tiny_block *)ptr - HEADER_BLOCKS;
+    tiny_block_section current_section = read_header(current);
+    write_header(current, current_section.size, false);
+
+    tiny_block *header = &tiny.buffer[0];
+    tiny_block_section section = read_header(header);
+    while(section.size > 0) {
+        tiny_block *next = next_section(header);
+        if(!section.taken) {
+            tiny_block_section next_section = read_header(next);
+            if(!next_section.taken) {
+                write_header(header, section.size + next_section.size + HEADER_BLOCKS, false);
+                section = read_header(header);
+                continue;
+            }
+        }
+
+        header = next;
+        section = read_header(header);
+    } 
+    store_operation(TINY_FREE, true, current_section.size);
+}

luma/wm/tiny.h

@@ -0,0 +1,65 @@
+#ifndef TINY_H
+#define TINY_H
+
+#include <stddef.h>
+#include <stdbool.h>
+
+typedef struct tiny_operation {
+    enum tiny_function {
+        TINY_LOAD,
+        TINY_INIT,
+        TINY_CLEAR,
+        TINY_RESET,
+        TINY_MALLOC,
+        TINY_REALLOC,
+        TINY_CALLOC,
+        TINY_FREE
+    } function;
+    bool success;
+    size_t size;
+} tiny_operation;
+
+typedef struct tiny_size {
+    size_t blocks;
+    size_t bytes;
+} tiny_size;
+
+typedef struct tiny_summary {
+    size_t alignment;
+    char *aligned_type;
+    void *static_buffer;
+    size_t static_buffer_size;
+    bool out_of_memory;
+    void *buffer;
+    tiny_size total;
+    tiny_size free;
+    tiny_size taken;
+    struct tiny_sections {
+        size_t total;
+        size_t free;
+        size_t taken;
+    } sections;
+} tiny_summary;
+
+typedef struct tiny_section {
+    bool taken;
+    void *header;
+    void *data;
+    tiny_size size;
+} tiny_section;
+
+void tiny_init(unsigned char *buffer, size_t size);
+void tiny_clear(void);
+void tiny_reset(void);
+void tiny_out_of_memory(bool status);
+tiny_operation tiny_last_operation(void);
+size_t tiny_block_size(void);
+void tiny_print(bool summary, bool last_op, bool heap);
+tiny_summary tiny_inspect(void);
+tiny_section tiny_next_section(void *previous_header);
+void *tiny_malloc(size_t size);
+void *tiny_realloc(void *ptr, size_t size);
+void *tiny_calloc(size_t num, size_t size);
+void tiny_free(void *ptr);
+
+#endif /* end of guard: TINY_H */

luma/wm/ui.c

@@ -0,0 +1,475 @@
+#include"ui.h"
+
+#include"tiny.h"
+#include"../vid/api.h"
+#include"../sys.h"
+#include"wm.h"
+
+#include"logserver.h"
+
+static void gtext_render(struct GObj *_) {
+	struct GText *this = (void*) _;
+	
+	LumaVidCommand *cmd = VidLink;
+	cmd->op = LUMA_VID_COMMAND_SET_COLOR_FLAT;
+	cmd->colorFlat.r = 0xF0;
+	cmd->colorFlat.g = 0xF0;
+	cmd->colorFlat.b = 0xF0;
+	cmd = (LumaVidCommand*) ((uintptr_t) cmd + sizeof(cmd->colorFlat));
+	cmd->op = LUMA_VID_COMMAND_RECT;
+	cmd->rect.x1 = this->x;
+	cmd->rect.x2 = this->x + this->w - 1;
+	cmd->rect.y1 = this->y;
+	cmd->rect.y2 = this->y + this->h - 1;
+	cmd = (LumaVidCommand*) ((uintptr_t) cmd + sizeof(cmd->rect));
+	cmd->op = LUMA_VID_COMMAND_SET_COLOR_FLAT;
+	cmd->colorFlat.r = 0x00;
+	cmd->colorFlat.g = 0x00;
+	cmd->colorFlat.b = 0x00;
+	cmd = (LumaVidCommand*) ((uintptr_t) cmd + sizeof(cmd->colorFlat));
+	DynStr16 *txt = this->data;
+	if(!txt || !txt->len) txt = this->placeholder;
+	if(txt) {
+		cmd->op = LUMA_VID_COMMAND_TEXT;
+		cmd->text.x = this->x + 5;
+		cmd->text.y = this->y + 2;
+		cmd->text.wall = this->x + this->w;
+		std_copy(cmd->text.data, txt->data, cmd->text.len = txt->len);
+		cmd = (LumaVidCommand*) ((uintptr_t) cmd + sizeof(cmd->text) + cmd->text.len);
+	}
+	cmd->op = LUMA_VID_COMMAND_END;
+	
+	memory_barrier();
+	sys_signal_send(VidLink, -1);
+	sys_signal_wait(VidLink, NULL);
+}
+static void gtext_key(struct GObj *_, struct GKeyEvent ev) {
+	if(_->focus == 0) return;
+	
+	struct GText *this = (void*) _;
+	
+	if(ev.press) {
+		if(!this->data) {
+			this->data = g_dynstr16_new(8, 0, NULL);
+		}
+		
+		if(ev.scancode == LK_BACKSPACE) {
+			if(this->data->len) {
+				this->data->len--;
+			}
+		} else {
+			uint8_t newCh = 0;
+			
+			if(ev.scancode >= LK_A && ev.scancode <= LK_Z) {
+				newCh = ev.scancode - LK_A;
+				newCh += ev.shift ? 'A' : 'a';
+			} else if(ev.scancode == LK_SPACE) {
+				newCh = ' ';
+			}
+			
+			if(newCh) this->data = g_dynstr16_append(this->data, 1, &newCh);
+		}
+	}
+}
+static void gtext_destroy(struct GObj *_) {
+	struct GText *this = (void*) _;
+	
+	if(this->data) g_dynstr16_free(this->data);
+	
+	if(this->placeholder) g_dynstr16_free(this->placeholder);
+	
+	tiny_free(this);
+}
+struct GText *gtext_new(Coord x, Coord y, Coord w, Coord h, DynStr16 *placeholder, DynStr16 *str) {
+	struct GText *ret = tiny_calloc(sizeof(*ret), 1);
+	ret->onRender = gtext_render;
+	ret->onKey = gtext_key;
+	ret->destroy = gtext_destroy;
+	ret->x = x;
+	ret->y = y;
+	ret->w = w;
+	ret->h = h;
+	ret->placeholder = placeholder;
+	ret->data = str;
+	return ret;
+}
+
+static void gcont_render(struct GObj *_) {
+	struct GContainer *this = (void*) _;
+	for(size_t i = 0; i < this->childCount; i++) {
+		if(this->children[i].obj && this->children[i].obj->onRender) {
+			this->children[i].obj->onRender(this->children[i].obj);
+		}
+	}
+}
+static void gcont_key(struct GObj *_, struct GKeyEvent ev) {
+	if(_->focus == 0) return;
+	
+	struct GContainer *this = (void*) _;
+	for(size_t i = 0; i < this->childCount; i++) {
+		if(this->children[i].obj && this->children[i].obj->onKey) {
+			this->children[i].obj->onKey(this->children[i].obj, ev);
+		}
+	}
+}
+static void gcont_destroy(struct GObj *_) {
+	struct GContainer *this = (void*) _;
+	
+	for(size_t i = 0; i < this->childCount; i++) {
+		if(this->children[i].obj) {
+			this->children[i].obj->destroy(this->children[i].obj);
+		}
+	}
+	
+	tiny_free(this->children);
+	
+	tiny_free(this);
+}
+struct GContainer *gcont_new(Coord x, Coord y, Coord w, Coord h, FitterFunc fitter) {
+	struct GContainer *ret = tiny_calloc(sizeof(*ret), 1);
+	ret->onRender = gcont_render;
+	ret->onKey = gcont_key;
+	ret->destroy = gcont_destroy;
+	ret->x = x;
+	ret->y = y;
+	ret->w = w;
+	ret->h = h;
+	ret->fitter = fitter;
+	return ret;
+}
+int gcont_add(struct GContainer *this, GContainerChildMeta meta, struct GObj* child) {
+	struct GContainerChild *newchildren = tiny_realloc(this->children, sizeof(*this->children) * (this->childCount + 1));
+	if(newchildren) {
+		this->children = newchildren;
+		this->children[this->childCount].meta = meta;
+		this->children[this->childCount].obj = child;
+		this->childCount++;
+		if(this->fitter) this->fitter((struct GObj*) this);
+		return 1;
+	} else {
+		return 0;
+	}
+}
+void gcont_set(struct GContainer *this, size_t index, GContainerChildMeta meta, struct GObj *obj) {
+	this->children[index] = (struct GContainerChild) {
+		.meta = meta,
+		.obj = obj,
+	};
+	if(this->fitter) this->fitter((struct GObj*) this);
+}
+void gcont_del(struct GContainer *this, size_t index) {
+	std_move(&this->children[index], &this->children[index + 1], sizeof(*this->children) * (this->childCount - index - 1));
+	this->children = tiny_realloc(this->children, sizeof(*this->children) * (--this->childCount));
+}
+
+void gcont_splitter_v(struct GObj *_) {
+	struct GContainer *c = (void*) _;
+	
+	struct GContainerChild *o = c->children;
+	for(size_t i = 0; i < c->childCount; i++, o++) {
+		if(o[0].obj) {
+			o[0].obj->x = c->x;
+			o[0].obj->y = c->y + c->h * i / c->childCount;
+			o[0].obj->w = c->w;
+			o[0].obj->h = c->h / c->childCount;
+			
+			if(o[0].obj->fitter) o[0].obj->fitter(o[0].obj);
+		}
+	}
+}
+
+void gcont_splitter_h(struct GObj *_) {
+	struct GContainer *c = (void*) _;
+	
+	struct GContainerChild *o = c->children;
+	for(size_t i = 0; i < c->childCount; i++, o++) {
+		if(o[0].obj) {
+			o[0].obj->x = c->x + c->w * i / c->childCount;
+			o[0].obj->y = c->y;
+			o[0].obj->w = c->w / c->childCount;
+			o[0].obj->h = c->h;
+			
+			if(o[0].obj->fitter) o[0].obj->fitter(o[0].obj);
+		}
+	}
+}
+
+static void gwindow_render(struct GObj *_) {
+	struct GWindow *this = (void*) _;
+	
+	LumaVidCommand *cmd = VidLink;
+	cmd->op = LUMA_VID_COMMAND_SET_COLOR_FLAT;
+	if(this->focus) {
+		cmd->colorFlat.r = 0xF0;
+		cmd->colorFlat.g = 0xF0;
+		cmd->colorFlat.b = 0xF0;
+	} else {
+		cmd->colorFlat.r = 0x80;
+		cmd->colorFlat.g = 0x80;
+		cmd->colorFlat.b = 0x80;
+	}
+	cmd = (LumaVidCommand*) ((uintptr_t) cmd + sizeof(cmd->colorFlat));
+	cmd->op = LUMA_VID_COMMAND_RECT;
+	cmd->rect.x1 = this->x;
+	cmd->rect.x2 = this->x + this->w - 1;
+	cmd->rect.y1 = this->y;
+	cmd->rect.y2 = this->y + 11;
+	cmd = (LumaVidCommand*) ((uintptr_t) cmd + sizeof(cmd->rect));
+	cmd->op = LUMA_VID_COMMAND_SET_COLOR_FLAT;
+	cmd->colorFlat.r = 0x00;
+	cmd->colorFlat.g = 0x00;
+	cmd->colorFlat.b = 0x00;
+	cmd = (LumaVidCommand*) ((uintptr_t) cmd + sizeof(cmd->colorFlat));
+	cmd->op = LUMA_VID_COMMAND_TEXT;
+	cmd->text.x = this->x + 3;
+	cmd->text.y = this->y + 1;
+	std_copy(cmd->text.data, this->name->data, cmd->text.len = this->name->len);
+	cmd = (LumaVidCommand*) ((uintptr_t) cmd + sizeof(cmd->text) + cmd->text.len);
+	cmd->op = LUMA_VID_COMMAND_END;
+	
+	memory_barrier();
+	sys_signal_send(VidLink, -1);
+	sys_signal_wait(VidLink, NULL);
+	
+	if(this->child) {
+		if(this->child->onRender) this->child->onRender(this->child);
+	}
+}
+static void gwindow_key(struct GObj *_, struct GKeyEvent ev) {
+	if(_->focus == 0) return;
+	
+	struct GWindow *this = (void*) _;
+	if(this->child) {
+		if(this->child->onKey) this->child->onKey(this->child, ev);
+	}
+}
+static void gwindow_fitter(struct GObj *_) {
+	struct GWindow *this = (void*) _;
+	if(this->child) {
+		this->child->x = this->x;
+		this->child->y = this->y + 12;
+		this->child->w = this->w;
+		this->child->h = this->h - 12;
+		if(this->child->fitter) this->child->fitter(this->child);
+	}
+}
+static void gwindow_destroy(struct GObj *_) {
+	struct GWindow *this = (void*) _;
+	if(this->child) {
+		this->child->destroy(this->child);
+	}
+	if(this->name) {
+		g_dynstr16_free(this->name);
+	}
+	tiny_free(this);
+}
+struct GWindow *gwindow_new(Coord x, Coord y, Coord w, Coord h, DynStr16 *name, struct GObj *child, size_t owner) {
+	struct GWindow *ret = tiny_calloc(sizeof(*ret), 1);
+	ret->onRender = gwindow_render;
+	ret->onKey = gwindow_key;
+	ret->fitter = gwindow_fitter;
+	ret->destroy = gwindow_destroy;
+	ret->x = x;
+	ret->y = y;
+	ret->w = w;
+	ret->h = h;
+	ret->name = name;
+	ret->child = child;
+	ret->owner = owner;
+	return ret;
+}
+int gwindow_regcmd(struct GWindow *this, Str16 *name, void *ud, GCMDFunc handler) {
+	struct Command *newarray = tiny_realloc(this->cmds.array, sizeof(*this->cmds.array) * (this->cmds.count + 1));
+	if(!newarray) {
+		return 0;
+	}
+	
+	newarray[this->cmds.count] = (struct Command) {
+		.name = name,
+		.ud = ud,
+		.run = handler
+	};
+	this->cmds.array = newarray;
+	
+	this->cmds.count++;
+	
+	return 1;
+}
+
+static void glabel_render(struct GObj *_) {
+	struct GLabel *this = (void*) _;
+	
+	LumaVidCommand *cmd = VidLink;
+	cmd->op = LUMA_VID_COMMAND_SET_COLOR_FLAT;
+	cmd->colorFlat.r = 0x00;
+	cmd->colorFlat.g = 0x00;
+	cmd->colorFlat.b = 0x00;
+	cmd = (LumaVidCommand*) ((uintptr_t) cmd + sizeof(cmd->colorFlat));
+	if(this->data) {
+		cmd->op = LUMA_VID_COMMAND_TEXT;
+		cmd->text.x = this->x;
+		cmd->text.y = this->y;
+		cmd->text.wall = this->x + this->w;
+		std_copy(cmd->text.data, this->data->data, cmd->text.len = this->data->len);
+		cmd = (LumaVidCommand*) ((uintptr_t) cmd + sizeof(cmd->text) + cmd->text.len);
+	}
+	cmd->op = LUMA_VID_COMMAND_END;
+	
+	memory_barrier();
+	sys_signal_send(VidLink, -1);
+	sys_signal_wait(VidLink, NULL);
+}
+static void glabel_destroy(struct GObj *_) {
+	struct GLabel *this = (void*) _;
+	
+	if(this->data && this->owned) {
+		g_dynstr16_free(this->data);
+	}
+	
+	tiny_free(this);
+}
+struct GLabel *glabel_new(Coord x, Coord y, Coord w, Coord h, DynStr16 *data, int owned) {
+	struct GLabel *ret = tiny_calloc(sizeof(*ret), 1);
+	ret->onRender = glabel_render;
+	ret->onKey = NULL;
+	ret->fitter = NULL;
+	ret->destroy = glabel_destroy;
+	ret->x = x;
+	ret->y = y;
+	ret->w = w;
+	ret->h = h;
+	ret->data = data;
+	ret->owned = owned;
+	return ret;
+}
+
+static void glist_render(struct GObj *_) {
+	struct GList *this = (void*) _;
+	
+	for(size_t i = 0; i < this->childCount; i++) {
+		LumaVidCommand *cmd = VidLink;
+		cmd->op = LUMA_VID_COMMAND_SET_COLOR_FLAT;
+		if(this->children[i].selected) {
+			cmd->colorFlat.r = 0xFF;
+			cmd->colorFlat.g = 0xFF;
+			cmd->colorFlat.b = 0xFF;
+		} else if(i == this->cursor) {
+			cmd->colorFlat.r = 0x80;
+			cmd->colorFlat.g = 0x80;
+			cmd->colorFlat.b = 0x80;
+		} else {
+			cmd->colorFlat.r = 0x00;
+			cmd->colorFlat.g = 0x00;
+			cmd->colorFlat.b = 0x00;
+		}
+		cmd = (LumaVidCommand*) ((uintptr_t) cmd + sizeof(cmd->colorFlat));
+		cmd->op = LUMA_VID_COMMAND_TEXT;
+		cmd->text.x = this->x;
+		cmd->text.y = this->y + 13 * i;
+		cmd->text.wall = this->x + this->w;
+		std_copy(cmd->text.data, this->children[i].text->data, cmd->text.len = this->children[i].text->len);
+		cmd = (LumaVidCommand*) ((uintptr_t) cmd + sizeof(cmd->text) + cmd->text.len);
+		cmd->op = LUMA_VID_COMMAND_END;
+		memory_barrier();
+		sys_signal_send(VidLink, -1);
+		sys_signal_wait(VidLink, NULL);
+	}
+}
+static void glist_key(struct GObj *_, struct GKeyEvent ev) {
+	struct GList *this = (void*) _;
+	if(!ev.press) return;
+	
+	if(ev.scancode == LK_DOWN) {
+		if(this->childCount) {
+			this->cursor = (this->cursor + 1) % this->childCount;
+		} else {
+			this->cursor = 0;
+		}
+		
+		dirty_obj(_);
+	} else if(ev.scancode == LK_UP) {
+		if(this->childCount) {
+			this->cursor = (this->cursor - 1 + this->childCount) % this->childCount;
+		} else {
+			this->cursor = 0;
+		}
+		
+		dirty_obj(_);
+	} else if(ev.scancode == LK_SPACE) {
+		if(this->cursor < this->childCount) {
+			this->children[this->cursor].selected ^= 1;
+		}
+		
+		dirty_obj(_);
+	}
+}
+static void glist_destroy(struct GObj *_) {
+	struct GList *this = (void*) _;
+	
+	for(size_t i = 0; i < this->childCount; i++) {
+		g_dynstr16_free(this->children[i].text);
+	}
+	
+	tiny_free(this->children);
+}
+struct GList *glist_new(Coord x, Coord y, Coord w, Coord h) {
+	struct GList *ret = tiny_calloc(sizeof(*ret), 1);
+	ret->onRender = glist_render;
+	ret->onKey = glist_key;
+	ret->fitter = NULL;
+	ret->destroy = glist_destroy;
+	ret->x = x;
+	ret->y = y;
+	ret->w = w;
+	ret->h = h;
+	ret->childCount = 0;
+	ret->children = NULL;
+	return ret;
+}
+int glist_add(struct GList *this, DynStr16 *str, void *ud) {
+	struct GListChild *newchildren = tiny_realloc(this->children, sizeof(*this->children) * (this->childCount + 1));
+	if(newchildren) {
+		newchildren[this->childCount++] = (struct GListChild) {
+			.text = str,
+			.selected = 0,
+			.ud = ud,
+		};
+		this->children = newchildren;
+		return 1;
+	}
+	return 0;
+}
+void glist_clear(struct GList *this) {
+	this->childCount = 0;
+	tiny_free(this->children);
+	this->children = NULL;
+	this->cursor = 0;
+}
+
+DynStr16 *g_dynstr16_new(uint16_t cap, uint16_t len, const void *data) {
+	DynStr16 *ret = tiny_malloc(sizeof(*ret) + cap);
+	ret->cap = cap;
+	ret->len = len;
+	if(data) {
+		std_copy(ret->data, data, len);
+	}
+	return ret;
+}
+DynStr16 *g_dynstr16_append(DynStr16 *this, uint16_t len, const void *data) {
+	if(this->len + len > this->cap) {
+		this = tiny_realloc(this, sizeof(*this) + (this->cap = this->len + len));
+	}
+	std_copy(this->data + this->len, data, len);
+	this->len += len;
+	return this;
+}
+void g_dynstr16_free(DynStr16 *this) {
+	tiny_free(this);
+}
+
+Str16 *g_str16_new(uint16_t len, const void *data) {
+	Str16 *ret = tiny_malloc(sizeof(*ret) + len);
+	ret->len = len;
+	std_copy(ret->data, data, len);
+	return ret;
+}

luma/wm/ui.h

@@ -0,0 +1,124 @@
+#pragma once
+
+#include"../std.h"
+#include"../ps2/api.h"
+
+typedef uint16_t Coord;
+
+struct GObj;
+struct GContainer;
+
+#define SHIFT_LEFT 1
+#define SHIFT_RIGHT 2
+#define ALT_LEFT 1
+#define ALT_RIGHT 2
+#define SYS_LEFT 1
+#define SYS_RIGHT 2
+struct GKeyEvent {
+	uint8_t scancode;
+	uint8_t press;
+	uint8_t shift;
+	uint8_t alt;
+	uint8_t sys;
+};
+
+typedef void(*RenderFunc)(struct GObj*);
+typedef void(*KeyReactFunc)(struct GObj*, struct GKeyEvent);
+typedef void(*FitterFunc)(struct GObj*);
+typedef void(*DestroyFunc)(struct GObj*);
+
+struct GObj {
+	RenderFunc onRender;
+	KeyReactFunc onKey;
+	DestroyFunc destroy;
+	FitterFunc fitter;
+	Coord x, y;
+	Coord w, h;
+	uint8_t focus;
+	void *ud;
+};
+
+struct GText {
+	struct GObj;
+	
+	DynStr16 *placeholder;
+	DynStr16 *data;
+};
+struct GText *gtext_new(Coord x, Coord y, Coord w, Coord h, DynStr16 *placeholder, DynStr16 *str);
+
+typedef uint8_t GContainerChildMeta;
+struct GContainerChild {
+	GContainerChildMeta meta;
+	struct GObj *obj;
+};
+struct GContainer {
+	struct GObj;
+	
+	size_t childCount;
+	struct GContainerChild *children;
+	
+	uint8_t meta;
+};
+struct GContainer *gcont_new(Coord x, Coord y, Coord w, Coord h, FitterFunc fitter);
+int gcont_add(struct GContainer*, GContainerChildMeta, struct GObj*);
+void gcont_set(struct GContainer*, size_t index, GContainerChildMeta, struct GObj*);
+void gcont_del(struct GContainer*, size_t index);
+
+void gcont_splitter_v(struct GObj*);
+void gcont_splitter_h(struct GObj*);
+
+typedef int(*GCMDFunc)(void*, const DynStr16 *line);
+struct Command {
+	Str16 *name;
+	void *ud;
+	GCMDFunc run;
+};
+struct CommandList {
+	size_t count;
+	struct Command *array;
+};
+
+struct GWindow {
+	struct GObj;
+	
+	DynStr16 *name;
+	
+	struct GObj *child;
+	
+	size_t owner;
+	
+	struct CommandList cmds;
+};
+struct GWindow *gwindow_new(Coord x, Coord y, Coord w, Coord h, DynStr16 *name, struct GObj *child, size_t owner);
+int gwindow_regcmd(struct GWindow*, Str16 *cmd, void *ud, GCMDFunc handler);
+
+struct GLabel {
+	struct GObj;
+	
+	DynStr16 *data;
+	int owned;
+};
+struct GLabel *glabel_new(Coord x, Coord y, Coord w, Coord h, DynStr16 *data, int owned);
+
+struct GListChild {
+	DynStr16 *text;
+	void *ud;
+	uint8_t selected;
+};
+struct GList {
+	struct GObj;
+	
+	size_t childCount;
+	struct GListChild *children;
+	
+	size_t cursor;
+};
+struct GList *glist_new(Coord x, Coord y, Coord w, Coord h);
+int glist_add(struct GList*, DynStr16*, void *ud);
+void glist_clear(struct GList*);
+
+DynStr16 *g_dynstr16_new(uint16_t cap, uint16_t len, const void *data);
+DynStr16 *g_dynstr16_append(DynStr16*, uint16_t len, const void *data);
+void g_dynstr16_free(DynStr16*);
+
+Str16 *g_str16_new(uint16_t len, const void *data);

luma/wm/wm.h

@@ -0,0 +1,21 @@
+#pragma once
+
+#include"ui.h"
+
+extern void *FSLink;
+extern void *VidLink;
+
+struct WTreeNode {
+	char vertical;
+	struct GWindow *window;
+	struct WTreeNode *parent;
+	struct WTreeNode *first;
+	struct WTreeNode *second;
+};
+extern struct WTreeNode *windowTreeFocus;
+
+int add_window(struct GWindow *w);
+int is_free_space();
+
+void dirty(int x, int y, int w, int h);
+void dirty_obj(struct GObj*);

luma/wm/zconf.h

@@ -0,0 +1,548 @@
+/* zconf.h -- configuration of the zlib compression library
+ * Copyright (C) 1995-2016 Jean-loup Gailly, Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* @(#) $Id$ */
+
+#ifndef ZCONF_H
+#define ZCONF_H
+#define Z_SOLO
+
+/*
+ * If you *really* need a unique prefix for all types and library functions,
+ * compile with -DZ_PREFIX. The "standard" zlib should be compiled without it.
+ * Even better than compiling with -DZ_PREFIX would be to use configure to set
+ * this permanently in zconf.h using "./configure --zprefix".
+ */
+#if 1     /* was set to #if 1 by ./configure */
+#  define Z_PREFIX_SET
+
+/* all linked symbols and init macros */
+#  define _dist_code            z__dist_code
+#  define _length_code          z__length_code
+#  define _tr_align             z__tr_align
+#  define _tr_flush_bits        z__tr_flush_bits
+#  define _tr_flush_block       z__tr_flush_block
+#  define _tr_init              z__tr_init
+#  define _tr_stored_block      z__tr_stored_block
+#  define _tr_tally             z__tr_tally
+#  define adler32               z_adler32
+#  define adler32_combine       z_adler32_combine
+#  define adler32_combine64     z_adler32_combine64
+#  define adler32_z             z_adler32_z
+#  ifndef Z_SOLO
+#    define compress              z_compress
+#    define compress2             z_compress2
+#    define compressBound         z_compressBound
+#  endif
+#  define crc32                 z_crc32
+#  define crc32_combine         z_crc32_combine
+#  define crc32_combine64       z_crc32_combine64
+#  define crc32_combine_gen     z_crc32_combine_gen
+#  define crc32_combine_gen64   z_crc32_combine_gen64
+#  define crc32_combine_op      z_crc32_combine_op
+#  define crc32_z               z_crc32_z
+#  define deflate               z_deflate
+#  define deflateBound          z_deflateBound
+#  define deflateCopy           z_deflateCopy
+#  define deflateEnd            z_deflateEnd
+#  define deflateGetDictionary  z_deflateGetDictionary
+#  define deflateInit           z_deflateInit
+#  define deflateInit2          z_deflateInit2
+#  define deflateInit2_         z_deflateInit2_
+#  define deflateInit_          z_deflateInit_
+#  define deflateParams         z_deflateParams
+#  define deflatePending        z_deflatePending
+#  define deflatePrime          z_deflatePrime
+#  define deflateReset          z_deflateReset
+#  define deflateResetKeep      z_deflateResetKeep
+#  define deflateSetDictionary  z_deflateSetDictionary
+#  define deflateSetHeader      z_deflateSetHeader
+#  define deflateTune           z_deflateTune
+#  define deflate_copyright     z_deflate_copyright
+#  define get_crc_table         z_get_crc_table
+#  ifndef Z_SOLO
+#    define gz_error              z_gz_error
+#    define gz_intmax             z_gz_intmax
+#    define gz_strwinerror        z_gz_strwinerror
+#    define gzbuffer              z_gzbuffer
+#    define gzclearerr            z_gzclearerr
+#    define gzclose               z_gzclose
+#    define gzclose_r             z_gzclose_r
+#    define gzclose_w             z_gzclose_w
+#    define gzdirect              z_gzdirect
+#    define gzdopen               z_gzdopen
+#    define gzeof                 z_gzeof
+#    define gzerror               z_gzerror
+#    define gzflush               z_gzflush
+#    define gzfread               z_gzfread
+#    define gzfwrite              z_gzfwrite
+#    define gzgetc                z_gzgetc
+#    define gzgetc_               z_gzgetc_
+#    define gzgets                z_gzgets
+#    define gzoffset              z_gzoffset
+#    define gzoffset64            z_gzoffset64
+#    define gzopen                z_gzopen
+#    define gzopen64              z_gzopen64
+#    ifdef _WIN32
+#      define gzopen_w              z_gzopen_w
+#    endif
+#    define gzprintf              z_gzprintf
+#    define gzputc                z_gzputc
+#    define gzputs                z_gzputs
+#    define gzread                z_gzread
+#    define gzrewind              z_gzrewind
+#    define gzseek                z_gzseek
+#    define gzseek64              z_gzseek64
+#    define gzsetparams           z_gzsetparams
+#    define gztell                z_gztell
+#    define gztell64              z_gztell64
+#    define gzungetc              z_gzungetc
+#    define gzvprintf             z_gzvprintf
+#    define gzwrite               z_gzwrite
+#  endif
+#  define inflate               z_inflate
+#  define inflateBack           z_inflateBack
+#  define inflateBackEnd        z_inflateBackEnd
+#  define inflateBackInit       z_inflateBackInit
+#  define inflateBackInit_      z_inflateBackInit_
+#  define inflateCodesUsed      z_inflateCodesUsed
+#  define inflateCopy           z_inflateCopy
+#  define inflateEnd            z_inflateEnd
+#  define inflateGetDictionary  z_inflateGetDictionary
+#  define inflateGetHeader      z_inflateGetHeader
+#  define inflateInit           z_inflateInit
+#  define inflateInit2          z_inflateInit2
+#  define inflateInit2_         z_inflateInit2_
+#  define inflateInit_          z_inflateInit_
+#  define inflateMark           z_inflateMark
+#  define inflatePrime          z_inflatePrime
+#  define inflateReset          z_inflateReset
+#  define inflateReset2         z_inflateReset2
+#  define inflateResetKeep      z_inflateResetKeep
+#  define inflateSetDictionary  z_inflateSetDictionary
+#  define inflateSync           z_inflateSync
+#  define inflateSyncPoint      z_inflateSyncPoint
+#  define inflateUndermine      z_inflateUndermine
+#  define inflateValidate       z_inflateValidate
+#  define inflate_copyright     z_inflate_copyright
+#  define inflate_fast          z_inflate_fast
+#  define inflate_table         z_inflate_table
+#  ifndef Z_SOLO
+#    define uncompress            z_uncompress
+#    define uncompress2           z_uncompress2
+#  endif
+#  define zError                z_zError
+#  ifndef Z_SOLO
+#    define zcalloc               z_zcalloc
+#    define zcfree                z_zcfree
+#  endif
+#  define zlibCompileFlags      z_zlibCompileFlags
+#  define zlibVersion           z_zlibVersion
+
+/* all zlib typedefs in zlib.h and zconf.h */
+#  define Byte                  z_Byte
+#  define Bytef                 z_Bytef
+#  define alloc_func            z_alloc_func
+#  define charf                 z_charf
+#  define free_func             z_free_func
+#  ifndef Z_SOLO
+#    define gzFile                z_gzFile
+#  endif
+#  define gz_header             z_gz_header
+#  define gz_headerp            z_gz_headerp
+#  define in_func               z_in_func
+#  define intf                  z_intf
+#  define out_func              z_out_func
+#  define uInt                  z_uInt
+#  define uIntf                 z_uIntf
+#  define uLong                 z_uLong
+#  define uLongf                z_uLongf
+#  define voidp                 z_voidp
+#  define voidpc                z_voidpc
+#  define voidpf                z_voidpf
+
+/* all zlib structs in zlib.h and zconf.h */
+#  define gz_header_s           z_gz_header_s
+#  define internal_state        z_internal_state
+
+#endif
+
+#if defined(__MSDOS__) && !defined(MSDOS)
+#  define MSDOS
+#endif
+#if (defined(OS_2) || defined(__OS2__)) && !defined(OS2)
+#  define OS2
+#endif
+#if defined(_WINDOWS) && !defined(WINDOWS)
+#  define WINDOWS
+#endif
+#if defined(_WIN32) || defined(_WIN32_WCE) || defined(__WIN32__)
+#  ifndef WIN32
+#    define WIN32
+#  endif
+#endif
+#if (defined(MSDOS) || defined(OS2) || defined(WINDOWS)) && !defined(WIN32)
+#  if !defined(__GNUC__) && !defined(__FLAT__) && !defined(__386__)
+#    ifndef SYS16BIT
+#      define SYS16BIT
+#    endif
+#  endif
+#endif
+
+/*
+ * Compile with -DMAXSEG_64K if the alloc function cannot allocate more
+ * than 64k bytes at a time (needed on systems with 16-bit int).
+ */
+#ifdef SYS16BIT
+#  define MAXSEG_64K
+#endif
+#ifdef MSDOS
+#  define UNALIGNED_OK
+#endif
+
+#ifdef __STDC_VERSION__
+#  ifndef STDC
+#    define STDC
+#  endif
+#  if __STDC_VERSION__ >= 199901L
+#    ifndef STDC99
+#      define STDC99
+#    endif
+#  endif
+#endif
+#if !defined(STDC) && (defined(__STDC__) || defined(__cplusplus))
+#  define STDC
+#endif
+#if !defined(STDC) && (defined(__GNUC__) || defined(__BORLANDC__))
+#  define STDC
+#endif
+#if !defined(STDC) && (defined(MSDOS) || defined(WINDOWS) || defined(WIN32))
+#  define STDC
+#endif
+#if !defined(STDC) && (defined(OS2) || defined(__HOS_AIX__))
+#  define STDC
+#endif
+
+#if defined(__OS400__) && !defined(STDC)    /* iSeries (formerly AS/400). */
+#  define STDC
+#endif
+
+#ifndef STDC
+#  ifndef const /* cannot use !defined(STDC) && !defined(const) on Mac */
+#    define const       /* note: need a more gentle solution here */
+#  endif
+#endif
+
+#if defined(ZLIB_CONST) && !defined(z_const)
+#  define z_const const
+#else
+#  define z_const
+#endif
+
+#ifdef Z_SOLO
+   typedef unsigned long z_size_t;
+#else
+#  define z_longlong long long
+#  if defined(NO_SIZE_T)
+     typedef unsigned NO_SIZE_T z_size_t;
+#  elif defined(STDC)
+#    include <stddef.h>
+     typedef size_t z_size_t;
+#  else
+     typedef unsigned long z_size_t;
+#  endif
+#  undef z_longlong
+#endif
+
+/* Maximum value for memLevel in deflateInit2 */
+#ifndef MAX_MEM_LEVEL
+#  ifdef MAXSEG_64K
+#    define MAX_MEM_LEVEL 8
+#  else
+#    define MAX_MEM_LEVEL 9
+#  endif
+#endif
+
+/* Maximum value for windowBits in deflateInit2 and inflateInit2.
+ * WARNING: reducing MAX_WBITS makes minigzip unable to extract .gz files
+ * created by gzip. (Files created by minigzip can still be extracted by
+ * gzip.)
+ */
+#ifndef MAX_WBITS
+#  define MAX_WBITS   15 /* 32K LZ77 window */
+#endif
+
+/* The memory requirements for deflate are (in bytes):
+            (1 << (windowBits+2)) +  (1 << (memLevel+9))
+ that is: 128K for windowBits=15  +  128K for memLevel = 8  (default values)
+ plus a few kilobytes for small objects. For example, if you want to reduce
+ the default memory requirements from 256K to 128K, compile with
+     make CFLAGS="-O -DMAX_WBITS=14 -DMAX_MEM_LEVEL=7"
+ Of course this will generally degrade compression (there's no free lunch).
+
+   The memory requirements for inflate are (in bytes) 1 << windowBits
+ that is, 32K for windowBits=15 (default value) plus about 7 kilobytes
+ for small objects.
+*/
+
+                        /* Type declarations */
+
+#ifndef OF /* function prototypes */
+#  ifdef STDC
+#    define OF(args)  args
+#  else
+#    define OF(args)  ()
+#  endif
+#endif
+
+#ifndef Z_ARG /* function prototypes for stdarg */
+#  if defined(STDC) || defined(Z_HAVE_STDARG_H)
+#    define Z_ARG(args)  args
+#  else
+#    define Z_ARG(args)  ()
+#  endif
+#endif
+
+/* The following definitions for FAR are needed only for MSDOS mixed
+ * model programming (small or medium model with some far allocations).
+ * This was tested only with MSC; for other MSDOS compilers you may have
+ * to define NO_MEMCPY in zutil.h.  If you don't need the mixed model,
+ * just define FAR to be empty.
+ */
+#ifdef SYS16BIT
+#  if defined(M_I86SM) || defined(M_I86MM)
+     /* MSC small or medium model */
+#    define SMALL_MEDIUM
+#    ifdef _MSC_VER
+#      define FAR _far
+#    else
+#      define FAR far
+#    endif
+#  endif
+#  if (defined(__SMALL__) || defined(__MEDIUM__))
+     /* Turbo C small or medium model */
+#    define SMALL_MEDIUM
+#    ifdef __BORLANDC__
+#      define FAR _far
+#    else
+#      define FAR far
+#    endif
+#  endif
+#endif
+
+#if defined(WINDOWS) || defined(WIN32)
+   /* If building or using zlib as a DLL, define ZLIB_DLL.
+    * This is not mandatory, but it offers a little performance increase.
+    */
+#  ifdef ZLIB_DLL
+#    if defined(WIN32) && (!defined(__BORLANDC__) || (__BORLANDC__ >= 0x500))
+#      ifdef ZLIB_INTERNAL
+#        define ZEXTERN extern __declspec(dllexport)
+#      else
+#        define ZEXTERN extern __declspec(dllimport)
+#      endif
+#    endif
+#  endif  /* ZLIB_DLL */
+   /* If building or using zlib with the WINAPI/WINAPIV calling convention,
+    * define ZLIB_WINAPI.
+    * Caution: the standard ZLIB1.DLL is NOT compiled using ZLIB_WINAPI.
+    */
+#  ifdef ZLIB_WINAPI
+#    ifdef FAR
+#      undef FAR
+#    endif
+#    ifndef WIN32_LEAN_AND_MEAN
+#      define WIN32_LEAN_AND_MEAN
+#    endif
+#    include <windows.h>
+     /* No need for _export, use ZLIB.DEF instead. */
+     /* For complete Windows compatibility, use WINAPI, not __stdcall. */
+#    define ZEXPORT WINAPI
+#    ifdef WIN32
+#      define ZEXPORTVA WINAPIV
+#    else
+#      define ZEXPORTVA FAR CDECL
+#    endif
+#  endif
+#endif
+
+#if defined (__BEOS__)
+#  ifdef ZLIB_DLL
+#    ifdef ZLIB_INTERNAL
+#      define ZEXPORT   __declspec(dllexport)
+#      define ZEXPORTVA __declspec(dllexport)
+#    else
+#      define ZEXPORT   __declspec(dllimport)
+#      define ZEXPORTVA __declspec(dllimport)
+#    endif
+#  endif
+#endif
+
+#ifndef ZEXTERN
+#  define ZEXTERN extern
+#endif
+#ifndef ZEXPORT
+#  define ZEXPORT
+#endif
+#ifndef ZEXPORTVA
+#  define ZEXPORTVA
+#endif
+
+#ifndef FAR
+#  define FAR
+#endif
+
+#if !defined(__MACTYPES__)
+typedef unsigned char  Byte;  /* 8 bits */
+#endif
+typedef unsigned int   uInt;  /* 16 bits or more */
+typedef unsigned long  uLong; /* 32 bits or more */
+
+#ifdef SMALL_MEDIUM
+   /* Borland C/C++ and some old MSC versions ignore FAR inside typedef */
+#  define Bytef Byte FAR
+#else
+   typedef Byte  FAR Bytef;
+#endif
+typedef char  FAR charf;
+typedef int   FAR intf;
+typedef uInt  FAR uIntf;
+typedef uLong FAR uLongf;
+
+#ifdef STDC
+   typedef void const *voidpc;
+   typedef void FAR   *voidpf;
+   typedef void       *voidp;
+#else
+   typedef Byte const *voidpc;
+   typedef Byte FAR   *voidpf;
+   typedef Byte       *voidp;
+#endif
+
+#if !defined(Z_U4) && !defined(Z_SOLO) && defined(STDC)
+#  include <limits.h>
+#  if (UINT_MAX == 0xffffffffUL)
+#    define Z_U4 unsigned
+#  elif (ULONG_MAX == 0xffffffffUL)
+#    define Z_U4 unsigned long
+#  elif (USHRT_MAX == 0xffffffffUL)
+#    define Z_U4 unsigned short
+#  endif
+#endif
+
+#ifdef Z_U4
+   typedef Z_U4 z_crc_t;
+#else
+   typedef unsigned long z_crc_t;
+#endif
+
+#ifdef HAVE_UNISTD_H    /* may be set to #if 1 by ./configure */
+#  define Z_HAVE_UNISTD_H
+#endif
+
+#if 1    /* was set to #if 1 by ./configure */
+#  define Z_HAVE_STDARG_H
+#endif
+
+#ifdef STDC
+#  ifndef Z_SOLO
+#    include <sys/types.h>      /* for off_t */
+#  endif
+#endif
+
+#if defined(STDC) || defined(Z_HAVE_STDARG_H)
+#  ifndef Z_SOLO
+#    include <stdarg.h>         /* for va_list */
+#  endif
+#endif
+
+#ifdef _WIN32
+#  ifndef Z_SOLO
+#    include <stddef.h>         /* for wchar_t */
+#  endif
+#endif
+
+/* a little trick to accommodate both "#define _LARGEFILE64_SOURCE" and
+ * "#define _LARGEFILE64_SOURCE 1" as requesting 64-bit operations, (even
+ * though the former does not conform to the LFS document), but considering
+ * both "#undef _LARGEFILE64_SOURCE" and "#define _LARGEFILE64_SOURCE 0" as
+ * equivalently requesting no 64-bit operations
+ */
+#if defined(_LARGEFILE64_SOURCE) && -_LARGEFILE64_SOURCE - -1 == 1
+#  undef _LARGEFILE64_SOURCE
+#endif
+
+#ifndef Z_HAVE_UNISTD_H
+#  ifdef __WATCOMC__
+#    define Z_HAVE_UNISTD_H
+#  endif
+#endif
+#ifndef Z_HAVE_UNISTD_H
+#  if defined(_LARGEFILE64_SOURCE) && !defined(_WIN32)
+#    define Z_HAVE_UNISTD_H
+#  endif
+#endif
+#ifndef Z_SOLO
+#  if defined(Z_HAVE_UNISTD_H)
+#    include <unistd.h>         /* for SEEK_*, off_t, and _LFS64_LARGEFILE */
+#    ifdef VMS
+#      include <unixio.h>       /* for off_t */
+#    endif
+#    ifndef z_off_t
+#      define z_off_t off_t
+#    endif
+#  endif
+#endif
+
+#if defined(_LFS64_LARGEFILE) && _LFS64_LARGEFILE-0
+#  define Z_LFS64
+#endif
+
+#if defined(_LARGEFILE64_SOURCE) && defined(Z_LFS64)
+#  define Z_LARGE64
+#endif
+
+#if defined(_FILE_OFFSET_BITS) && _FILE_OFFSET_BITS-0 == 64 && defined(Z_LFS64)
+#  define Z_WANT64
+#endif
+
+#if !defined(SEEK_SET) && !defined(Z_SOLO)
+#  define SEEK_SET        0       /* Seek from beginning of file.  */
+#  define SEEK_CUR        1       /* Seek from current position.  */
+#  define SEEK_END        2       /* Set file pointer to EOF plus "offset" */
+#endif
+
+#ifndef z_off_t
+#  define z_off_t long
+#endif
+
+#if !defined(_WIN32) && defined(Z_LARGE64)
+#  define z_off64_t off64_t
+#else
+#  if defined(_WIN32) && !defined(__GNUC__) && !defined(Z_SOLO)
+#    define z_off64_t __int64
+#  else
+#    define z_off64_t z_off_t
+#  endif
+#endif
+
+/* MVS linker does not support external names larger than 8 bytes */
+#if defined(__MVS__)
+  #pragma map(deflateInit_,"DEIN")
+  #pragma map(deflateInit2_,"DEIN2")
+  #pragma map(deflateEnd,"DEEND")
+  #pragma map(deflateBound,"DEBND")
+  #pragma map(inflateInit_,"ININ")
+  #pragma map(inflateInit2_,"ININ2")
+  #pragma map(inflateEnd,"INEND")
+  #pragma map(inflateSync,"INSY")
+  #pragma map(inflateSetDictionary,"INSEDI")
+  #pragma map(compressBound,"CMBND")
+  #pragma map(inflate_table,"INTABL")
+  #pragma map(inflate_fast,"INFA")
+  #pragma map(inflate_copyright,"INCOPY")
+#endif
+
+#endif /* ZCONF_H */

mod.ld

@@ -0,0 +1,20 @@
+OUTPUT_FORMAT("elf32-i386")
+OUTPUT_ARCH(i386)
+
+ENTRY(kmain)
+
+SECTIONS
+{
+	. = 0;
+	.text ALIGN(16) : {
+		*(.text)
+		. = ALIGN(16);
+		*(.data)
+		. = ALIGN(16);
+		*(.rodata*)
+		. = ALIGN(16);
+		*(.bss)
+	}
+	/*. += 512; /*Something to try if we get spooky memory corruption*/
+	_kernel_end = .;
+}

src/boot0.asm

@@ -0,0 +1,41 @@
+cpu 386
+
+bits 16
+org 0x7C00
+
+start:
+	jmp 0:cseg
+cseg:
+	cld
+	
+	mov ax, 0xB800
+	mov fs, ax
+	
+	mov ax, 0x3000
+	mov es, ax
+	
+	mov ax, 0x0200 | ((BOOT1_SIZE + 511) / 512)
+	mov cx, 0x0002
+	mov dh, 0
+	mov bx, 0
+	int 0x13
+	
+	setc al
+	add al, '0'
+	mov [fs:0], al
+	
+	xor ax, ax
+	mov es, ax
+	
+	mov ax, 0x3000
+	mov ds, ax
+	
+	mov ax, 0x4000 ;Apparently Bochs' BIOS bugs out if below 0x7C00 linear
+	mov ss, ax
+	xor sp, sp
+	
+	jmp 0x3000:0
+	
+db "Yo waddup"
+
+times 446 - ($ - $$) db 0

src/boot1.asm

@@ -0,0 +1,864 @@
+cpu 386
+
+bits 16
+org 0
+
+%define BIT_READWRITE (1 << 1)
+%define BIT_EXECUTABLE (1 << 3)
+%define BIT_ISCODEORDATA (1 << 4)
+%define BIT_PRESENT (1 << 7)
+
+%define PBIT_PRESENT 1
+%define PBIT_WRITABLE 2
+%define PBIT_USERLEVEL 4
+%define PBIT_WRITETHROUGH 8
+%define PBIT_CACHEDISABLE 16
+%define PBIT_ACCESSED 32
+
+MALLOC_MEM equ 0x520
+
+BOOT1_SEGMENT equ 0x3000
+BOOT1_ADDR equ (BOOT1_SEGMENT * 0x10)
+
+start:
+	mov [ld_lba.drive + 1], dl
+	
+	mov ah, 8
+	mov di, 0
+	int 0x13
+	mov dl, dh
+	mov dh, 0
+	inc dx
+	mov [lba2chs.numheads + 1], dx
+	and cl, 0x3F
+	mov ch, 0
+	mov [lba2chs.spt + 1], cx
+	
+	mov eax, [es:0x7C00 + 0x1BE + 8]
+	mov [get_inode.abs1], eax
+	mov [read_inode.abs1], eax
+	
+	mov ecx, 40 * 25
+	mov edi, 0
+.lup:
+	mov [fs:edi + ecx * 4 - 4], dword 0x0F000F00
+	dec ecx
+	jnz .lup
+	
+	mov si, MSG_START
+	mov ax, MSG_START.end - MSG_START
+	call print16
+	
+	mov edi, MALLOC_MEM ;Bump allocator pointer
+	
+	add edi, 4096
+	mov [TSS.esp0], edi ;End of stack ~ beginning of memory map!!
+	
+a20:
+	cli
+	call .waito
+	mov al, 0xAD
+	out 0x64, al
+	call .waito
+	mov al, 0xD0
+	out 0x64, al
+.waiti:
+	in al, 0x64
+	test al, 1
+	jz .waiti
+	in al, 0x60
+	push ax
+	call .waito
+	mov al, 0xD1
+	out 0x64, al
+	call .waito
+	pop ax
+	or al, 2
+	out 0x60, al
+	call .waito
+	mov al, 0xAE
+	out 0x64, al
+	call .waito
+	sti
+	jmp ramdetect
+.waito:
+	in al, 0x64
+	test al, 2
+	jnz .waito
+	ret
+	
+ramdetect: ;Currently only detects RAM below 16MB, which will be enough for a while.
+	clc
+	int 0x12
+	jc .fail
+	mov [es:di + 0], dword 0
+	mov [es:di + 4], dword 0
+	movzx ecx, ax
+	shr cx, 2
+	mov [es:di + 8], ecx
+	mov [es:di + 12], byte 0
+	add di, 13
+	mov al, 0
+	add cx, 7
+	shr cx, 3
+	rep stosb
+	
+	xor cx, cx
+	xor dx, dx
+	mov ax, 0xE801
+	int 0x15
+	jc .fail
+	cmp ah, 0x86
+	je .fail
+	cmp ah, 0x80
+	je .fail
+	jcxz .useax
+	mov ax, cx
+	mov bx, dx
+.useax:
+	mov [es:di + 0], dword 0x100000
+	mov [es:di + 4], dword 0
+	movzx ecx, ax
+	shr ecx, 2
+	mov [es:di + 8], ecx
+	mov [es:di + 12], byte 0
+	add di, 13
+	mov al, 0
+	shr ecx, 3
+	rep stosb
+	jmp .foundMMap
+	
+.fail:
+	mov si, MSG_FAIL
+	mov ax, MSG_FAIL.end - MSG_FAIL
+	call print16
+	
+	jmp $
+.foundMMap:
+	mov si, MSG_DETECTED_MEMORY_MAP
+	mov ax, MSG_DETECTED_MEMORY_MAP.end - MSG_DETECTED_MEMORY_MAP
+	call print16
+	
+pg:
+	; Create page directory
+	add edi, 4095
+	and edi, ~4095
+	mov cr3, edi ;In advance
+	lea eax, [edi + 4096]
+	or eax, PBIT_PRESENT | PBIT_WRITABLE
+	mov [es:edi], eax
+	add edi, 4
+	mov eax, 0
+	mov ecx, 1023
+	rep stosd
+
+	mov ecx, 192
+	mov edx, PBIT_PRESENT | PBIT_WRITABLE
+.lup:
+	mov [es:edi], edx
+	add edi, 4
+	add edx, 4096
+	dec ecx
+	jnz .lup
+	
+	mov ecx, 1024 - 192
+	mov eax, 0
+	rep stosd
+	
+	mov eax, cr3 ;Back to the page directory
+	mov [es:eax + 4092], edi
+	or dword [es:eax + 4092], PBIT_PRESENT | PBIT_WRITABLE
+	mov [es:edi + 4092], edi
+	or dword [es:edi + 4092], PBIT_PRESENT | PBIT_WRITABLE
+	add edi, 4096
+	
+	mov esi, MSG_STEP3s
+	mov eax, MSG_STEP3s.end - MSG_STEP3s
+	call print16
+	
+loadkernel:
+	mov eax, 2 ;root
+	call get_inode
+	mov bx, di
+	call read_inode
+	mov eax, edi
+	mov esi, KERNEL_FILE
+	mov ecx, KERNEL_FILE.end - KERNEL_FILE
+	call find_in_dirnode
+	test eax, eax
+	jnz .found
+	
+	mov si, MSG_STEP5f
+	mov ax, MSG_STEP5f.end - MSG_STEP5f
+	call print16
+	jmp $
+.found:
+	call get_inode
+	mov bx, di
+	call read_inode
+	; Now edi is kernel location, ebx is after kernel.
+	call fix_module_addresses
+	mov [inPE.KERNEL_START], edi
+	mov edi, ebx ;Allocate modules after kernel
+	;~ mov [inPE.MODULES_START], edi
+	
+loadmods:
+	mov eax, 2
+	call get_inode
+	mov esi, [endSector + 32] ;filesize
+	push es
+	push ds
+	pop es
+	mov bx, endSector + 0x200 ;Will break if root directory is larger than 1KB.
+	call read_inode
+	pop es
+	mov eax, endSector + 0x200
+.lup: ;Loop over files, the names of which end with ".mod"
+	movzx ecx, word [eax + 10] ;Length of name
+	lea edx, [eax + 12] ;Get name string
+	cmp ecx, 4 ;Test if shorter than 4 characters
+	jb .cont
+.testext:
+	cmp [eax + ecx + 12 - 4], dword ".mod"
+	jne .cont
+.testkrnl:
+	cmp [eax + 12], dword "krnl"
+	je .cont
+.ismod:
+	inc dword [inPE.MOD_COUNT]
+	push eax
+	mov eax, [eax] ;inode
+	call get_inode
+	call page_align_es_edi
+	push esi
+	mov esi, cr3
+	mov cx, 1024
+	push ds
+	push word 0
+	pop ds
+	rep movsd ;Copy page directory to one local to process
+	pop ds
+	pop esi
+	mov bx, di
+	call read_inode ;edi = start, ebx = end
+	mov di, bx
+	pop eax
+.cont:
+	movzx ecx, byte [eax + 9]
+	shl cx, 4
+	add eax, ecx
+	sub esi, ecx
+	jnz .lup
+.end:
+	mov di, es
+	shl edi, 4
+	add edi, 4095
+	and edi, ~4095
+	mov [inPE.IMPORTANT_MEM_END], edi
+	xor ax, ax
+	mov es, ax
+	
+	add ebx, edi
+fixmmap: ; Make all module memories "used"
+	lea ecx, [es:ebx + 4095]
+	shr ecx, 12
+	mov ebx, ecx
+	and ebx, 7
+	shr ecx, 3
+	mov edi, [TSS.esp0]
+	add edi, 13
+	push edi
+		mov al, -1
+		rep stosb
+		mov ecx, ebx
+		mov al, 1
+		shl al, cl
+		dec al
+		mov [es:edi], al
+	pop edi
+	or [es:edi + (BOOT1_ADDR / 4096 / 8)], byte 3 ; TODO: Let boot1 get overwritten after load
+
+modex:
+	mov ax, 0x13
+	int 0x10
+	
+	mov dx, 0x3C4
+	mov al, 4
+	out dx, al
+	
+	inc dx
+	mov al, 6
+	out dx, al
+	
+	mov dx, 0x3D4
+	mov al, 0x14
+	out dx, al
+	
+	inc dx
+	xor al, al
+	out dx, al
+	
+	dec dx
+	mov al, 0x17
+	out dx, al
+	
+	inc dx
+	mov al, 0xE3
+	out dx, al
+	
+	dec dx
+	mov al, 0x11
+	out dx, al
+	
+	inc dx
+	mov al, 0x2C
+	out dx, al
+	
+	dec dx
+	mov al, 0x06
+	out dx, al
+	
+	inc dx
+	mov al, 0x0D
+	out dx, al
+	
+	dec dx
+	mov al, 0x07
+	out dx, al
+	
+	inc dx
+	mov al, 0x3E
+	out dx, al
+	
+	dec dx
+	mov al, 0x10
+	out dx, al
+	
+	inc dx
+	mov al, 0xEA
+	out dx, al
+	
+	dec dx
+	mov al, 0x11
+	out dx, al
+	
+	inc dx
+	mov al, 0xAC
+	out dx, al
+	
+	dec dx
+	mov al, 0x12
+	out dx, al
+	
+	inc dx
+	mov al, 0xDF
+	out dx, al
+	
+	dec dx
+	mov al, 0x15
+	out dx, al
+	
+	inc dx
+	mov al, 0xE7
+	out dx, al
+	
+	dec dx
+	mov al, 0x16
+	out dx, al
+	
+	inc dx
+	mov al, 0x06
+	out dx, al
+	
+	; ESP should be zero here.
+	push strict dword BOOT1_ADDR + inPE
+	mov esp, BOOT1_ADDR + 0x10000 + 0xFFFC
+	jmp enterPM
+
+enterPM:
+	cli
+	or [GDT.R0CSLIMFLAGS], byte 64 ;Make 32-bit
+	lgdt [GDT]
+	mov eax, cr0
+	or eax, 1 | (1 << 31)
+	mov cr0, eax
+	jmp 8:dword BOOT1_ADDR + .inPM
+.inPM:
+bits 32
+	mov eax, 0x23
+	mov ds, ax
+	mov es, ax
+	mov fs, ax
+	mov gs, ax
+	mov al, 0x10
+	mov ss, ax
+	ret
+
+enterRM:
+	and [GDT.R0CSLIMFLAGS], byte ~64 ;Make 16-bit
+	lgdt [GDT]
+	jmp 8:.in16
+.in16:
+bits 16
+	mov eax, cr0
+	and eax, ~(1 | (1 << 31))
+	mov cr0, eax
+	jmp 0:.inRM
+.inRM:
+	mov ax, 0
+	mov ds, ax
+	mov es, ax
+	mov fs, ax
+	mov gs, ax
+	mov ss, ax
+	sti
+	ret
+
+; Makes es:edi a normalized far pointer
+normalize_es_edi:
+	push eax
+	push edi
+	mov ax, es
+	shr di, 4
+	add ax, di
+	mov es, ax
+	pop edi
+	and edi, 15
+	pop eax
+	ret
+
+page_align_es_edi:
+	push eax
+	xor eax, eax
+	mov ax, es
+	shl eax, 4
+	add eax, edi
+	add eax, 4095
+	and eax, ~4095
+	shr eax, 4
+	mov es, ax
+	pop eax
+	xor edi, edi
+	ret
+
+print16:
+	push di
+	push cx
+	push ax
+.ind:
+		mov di, 0
+		mov cx, ax
+.lup:
+		mov al, [ds:si]
+		mov [fs:di], al
+		inc si
+		inc di
+		inc di
+		dec cx
+		jnz .lup
+		mov [.ind + 1], di ;SMC
+	pop ax
+	pop cx
+	pop di
+	ret
+
+ld_lba:
+	pushad
+	call lba2chs
+	mov ch, al
+	mov dh, dl
+.drive:
+	mov dl, 0
+	mov ah, 2
+	mov al, 1
+	int 0x13
+	popad
+	ret
+
+; In:	AX = lba
+; Out:	CX = sector, AX = cylinder, DX = head
+lba2chs:
+	push bx
+.spt:
+	mov bx, 0xF157
+	xor dx, dx
+	div bx
+	mov cx, dx
+	inc cx
+.numheads:
+	mov bx, 0x9001
+	xor dx, dx
+	div bx
+	pop bx
+	ret
+
+; In:	EAX = inode
+; Out:	*BOOT1_SEGMENT:endSector = inode struct
+get_inode:
+	push eax
+	push ebx
+.abs1 equ ($ + 2)
+	add eax, strict dword 0xABCDEF42
+	push es
+	push ds
+	pop es
+	mov bx, endSector
+	call ld_lba
+	pop es
+	pop ebx
+	pop eax
+	ret
+
+; In:	*BOOT1_SEGMENT:endSector = inode struct
+; Out:	*ES:BX = data
+read_inode:
+	push edx
+	push ecx
+	push eax
+	push esi
+	push edi
+	push ebx
+	
+	mov dl, [endSector + 8] ;extent count
+	mov ecx, endSector + 104 ;extent starts
+	
+.lup:
+	mov eax, [ecx] ;start
+	mov esi, [ecx + 48] ;size of extent
+.abs1 equ ($ + 2)
+	add eax, strict dword 0x12345678
+	
+.lup2:
+	call ld_lba
+	add bx, 512
+	inc eax
+	dec esi
+	jnz .lup2
+	
+	add ecx, 8
+	dec dl
+	jnz .lup
+	
+	mov edi, [esp] ;dest address
+	lea esi, [edi + 176]
+	test dword [endSector + 28], 1 << 19
+	jz .doesntHaveInline
+	add esi, 512 - 176
+.doesntHaveInline:
+	mov ecx, [endSector + 32] ;filesize
+	push ds
+	push es
+	pop ds
+	rep movsb
+	pop ds
+	
+	add esp, 4 ;ignore old ebx, because read_inode should return the addr after
+	pop edi
+	pop esi
+	pop eax
+	pop ecx
+	pop edx
+	ret
+
+; In:	*ES:EAX = directory content buffer, *ESI = name to test, ECX = name length
+; Out:	EAX = inode or 0 if not found
+find_in_dirnode:
+	push ebx
+	push edx
+	push edi
+	
+	mov edi, [endSector + 32] ;filesize
+.lup:
+	movzx edx, word [es:eax + 10] ;record name length
+	cmp dx, cx
+.lup2:
+	dec dx
+	mov bl, [es:eax + 12 + edx]
+	cmp bl, [esi + edx]
+	jne .next
+	test dx, dx
+	jnz .lup2
+	jmp .found
+.next:
+	movzx edx, byte [es:eax + 9] ;record length
+	shl dx, 4
+	add eax, edx
+	sub edi, edx
+	jnz .lup
+	
+	mov eax, 0
+	jmp .end
+.found:
+	mov eax, [es:eax]
+.end:
+	pop edi
+	pop edx
+	pop ebx
+	ret
+
+fix_module_addresses:
+	push esi
+	push ecx
+	push eax
+	push ebx
+	
+	movzx ecx, word [es:edi + 4] ;amount of syms
+	lea esi, [es:edi + 12 + ecx * 4]
+	lea esi, [es:esi + ecx * 2]
+	movzx ecx, word [es:edi + 6] ;amount of relocs
+	mov eax, esi ;ptr to relocs
+	lea esi, [esi + ecx * 4] ;ptr to data
+	add esi, 15
+	and esi, ~15
+	mov [es:edi], esi ;Replace the MOD\0 magic header with the base address
+	test ecx, ecx ;If none, end
+	jz .end
+.lup:
+	; Apply relocation
+	mov ebx, [es:eax]
+	add ebx, esi
+	add [es:ebx], esi
+	
+	add eax, 4
+	dec ecx
+	jnz .lup
+.end:
+	
+	pop ebx
+	pop eax
+	pop ecx
+	pop esi
+	ret
+
+bits 32
+
+inPE:
+	mov [BOOT1_ADDR + TSS.ss0], dword 0x10
+	mov esp, [BOOT1_ADDR + TSS.esp0]
+	
+	mov eax, 40
+	ltr ax
+	
+	mov ax, [BOOT1_ADDR + print16.ind + 1]
+	mov [BOOT1_ADDR + print32.LAST], ax
+	
+	mov esi, BOOT1_ADDR + MSG_STEP2s
+	mov eax, MSG_STEP2s.end - MSG_STEP2s
+	call print32
+	
+.KERNEL_START equ ($ + 1)
+	mov edi, strict dword 0
+	lea ebx, [edi + 14]
+	; Sym 0: ppm_Bitmap
+	mov eax, [ebx]
+	add eax, [edi] ;Base address that used to be magic header
+	push dword [BOOT1_ADDR + TSS.esp0] ;As said before, end of stack ~ beginning of mmap!
+	pop dword [eax]
+	add ebx, 6
+	; Sym 1: vpm_init
+	mov eax, [ebx]
+	add eax, [edi]
+.IMPORTANT_MEM_END equ ($ + 1)
+	push strict dword 0
+	call eax
+	add esp, 4
+	add ebx, 6
+	; Sym 2: vpm_map
+	mov eax, [ebx]
+	add eax, [edi]
+	mov [BOOT1_ADDR + .VPM_MAP_ENTRY], eax
+	add ebx, 6
+	; Sym 3: canal_init
+	mov eax, [ebx]
+	add eax, [edi]
+	call eax
+	add ebx, 6
+	; Sym 4: pci_init
+	mov eax, [ebx]
+	add eax, [edi]
+	call eax
+	add ebx, 6
+	; Sym 5: scheduler_init
+	mov eax, [ebx]
+	add eax, [edi]
+	call eax
+	add ebx, 6
+	; Sym 6: scheduler_spawn
+	mov eax, [ebx]
+	add eax, [edi]
+	mov [BOOT1_ADDR + .SCHEDULER_SPAWN_ENTRY], eax
+	add ebx, 6
+	; Sym 7: scheduler_start
+	mov eax, [ebx]
+	add eax, [edi]
+	mov [BOOT1_ADDR + .SCHEDULER_START_ENTRY], eax
+	
+.MOD_COUNT equ ($ + 1)
+	mov ebp, strict dword 0
+.modspawnlup:
+	; Find next module
+	movzx ecx, word [edi + 4]
+	lea eax, [edi + ecx * 4]
+	lea eax, [eax + ecx * 2]
+	movzx ecx, word [edi + 6]
+	lea eax, [eax + ecx * 4]
+	add eax, 15
+	and eax, ~15
+	mov ecx, [edi + 8]
+	lea eax, [eax + ecx]
+	add eax, 4096 + 4095
+	and eax, ~4095
+	mov edi, eax
+	mov ebx, 0x40000000
+	call rebase_module32
+	sub edx, edi
+	
+	push strict dword edx
+	push strict dword edi
+	push strict dword 0x40000000
+	push strict dword 1 ;user
+	lea eax, [edi - 4096] ; cr3
+	push eax
+.VPM_MAP_ENTRY equ ($ + 1)
+	mov eax, strict dword 0
+	call eax
+	add esp, 20
+	
+	push strict dword 0
+.SCHEDULER_SPAWN_ENTRY equ ($ + 1)
+	mov eax, strict dword 0
+	call eax
+	add esp, 4
+	lea ebx, [edi - 4096]
+	mov [eax + 46], ebx ;cr3
+	mov ebx, [edi + 14] ;get sym 0 (modentry)
+	add ebx, [edi] ;add data offset
+	mov [eax + 34], ebx ;eip
+	
+	dec ebp
+	jnz .modspawnlup
+	
+.SCHEDULER_START_ENTRY equ ($ + 1)
+	mov eax, strict dword 0
+	jmp eax
+
+print32:
+	push ebx
+	push eax
+	push ecx
+	movzx ebx, word [BOOT1_ADDR + .LAST]
+	mov ecx, eax
+.lup:
+	mov al, [esi]
+	mov [0xB8000 + ebx], al
+	inc esi
+	inc ebx
+	inc ebx
+	dec ecx
+	jnz .lup
+	mov [BOOT1_ADDR + .LAST], bx
+	pop ecx
+	pop eax
+	pop ebx
+	ret
+.LAST: dw 0
+
+; In: EDI = module pointer, EBX = rebase address
+; Out: EDX = physical module end
+rebase_module32:
+	push ecx
+	push eax
+	push ebx
+	push ebp
+	movzx ecx, word [edi + 4]
+	lea eax, [edi + 12 + ecx * 2]
+	lea eax, [eax + ecx * 4] ; eax = relocations pointer
+	movzx ecx, word [edi + 6] ;ecx = amount of relocations
+	lea edx, [eax + ecx * 4] ; edx = data pointer
+	add edx, 15
+	and edx, ~15
+	mov [edi], edx ;Store offset of data buffer in
+	sub [edi], edi ;file in place of magic header
+	add [edi], ebx ;
+	mov ebx, [edi]
+	test ecx, ecx
+	jz .end
+.lup:
+	mov ebp, [eax]
+	add [edx + ebp], ebx
+	add eax, 4
+	dec ecx
+	jnz .lup
+.end:
+	add edx, [edi + 8]
+	pop ebp
+	pop ebx
+	pop eax
+	pop ecx
+	ret
+
+MSG_FAIL: db "Failed to detect memory map. "
+.end:
+
+MSG_START: db "Starting eklernel. "
+.end:
+
+MSG_DETECTED_MEMORY_MAP: db "Found memory map. "
+.end:
+
+MSG_STEP2s: db "Entered protected mode. "
+.end:
+
+MSG_STEP3s: db "Generated paging structures. "
+.end:
+
+MSG_STEP4f: db "Boot filesystem must have block size of 1024. "
+.end:
+
+KERNEL_FILE: db "krnl.mod"
+.end:
+
+MSG_STEP5f: db "Kernel not found; aborting boot. "
+.end:
+
+GDT:
+	dw (GDT.END - GDT) - 1 ;GDT descriptor in the null entry
+	dd BOOT1_ADDR + GDT
+	dw 0
+	
+	dw 0xFFFF, 0x0000
+	db 0x00, BIT_READWRITE | BIT_EXECUTABLE | BIT_ISCODEORDATA | BIT_PRESENT | (0 << 5)
+.R0CSLIMFLAGS:
+	db (128 | 64) | 0xF
+	db 0
+	
+	dw 0xFFFF, 0x0000
+	db 0x00, BIT_READWRITE | BIT_ISCODEORDATA | BIT_PRESENT | (0 << 5)
+	db (128 | 64) | 0xF
+	db 0
+	
+	dw 0xFFFF, 0x0000
+	db 0x00, BIT_READWRITE | BIT_EXECUTABLE | BIT_ISCODEORDATA | BIT_PRESENT | (3 << 5)
+	db (128 | 64) | 0xF
+	db 0
+	
+	dw 0xFFFF, 0x0000
+	db 0x00, BIT_READWRITE | BIT_ISCODEORDATA | BIT_PRESENT | (3 << 5)
+	db (128 | 64) | 0xF
+	db 0
+	
+	dw (TSS.END - TSS) - 1
+	dw TSS
+	db BOOT1_ADDR / 0x10000, 1 | 8 | 128
+	db 0, 0
+.END:
+TSS:
+	dd 0
+.esp0: dd 0
+.ss0: dd 0
+	times 23 dd 0
+.END:
+
+end:
+times (end - start + 511) / 512 * 512 - ($ - $$) db 0
+endSector:

src/kernel/arch/x86/canal.c

@@ -0,0 +1,181 @@
+#include"canal.h"
+
+#include"vpm.h"
+#include"mem.h"
+#include"scheduler.h"
+#include"ppm.h"
+
+static inline void pr(int c) {
+	asm volatile("outb %%al, $0xE9" :: "a"(c) :);
+}
+
+static inline void pri(size_t num) {
+	char buf[16] = {};
+	int i = 16;
+	
+	do {
+		buf[--i] = num % 10;
+		num /= 10;
+	} while(num);
+	
+	for(; i < 16; i++) {
+		pr(buf[i] + '0');
+	}
+}
+
+static inline void prcs(const char *str) {
+	while(*str) {
+		pr(*str);
+		str++;
+	}
+}
+
+static Canal *canals = 0;
+static Link *links = 0;
+
+void canal_init() {
+	uint32_t cr3;
+	asm volatile("movl %%cr3, %0" : "=r"(cr3) :);
+	
+	canals = vpm_map(cr3, VPM_KERNEL, -1, -1, 4096);
+	kmemw1(canals, 0, 4096);
+	
+	links = vpm_map(cr3, VPM_KERNEL, -1, -1, 4096);
+	kmemw1(links, 0xFF, 4096);
+}
+
+uint32_t canal_create(uint64_t name, uint16_t hostProcessId) {
+	Process *hostProcess = scheduler_find(hostProcessId);
+	
+	for(int i = 0; i < 4096 / sizeof(Canal); i++) {
+		if(canals[i].name == name) return -2;
+		
+		if(canals[i].name == 0) {
+			canals[i].name = name;
+			canals[i].hostProcess = hostProcess;
+			canals[i].links = NULL;
+			
+			canals[i].ownedNext = hostProcess->ownedCanal;
+			hostProcess->ownedCanal = &canals[i];
+			
+			return i;
+		}
+	}
+	return -1;
+}
+
+uint32_t link_create(uint64_t canalName, uint16_t guestProcessId, uint32_t areaSize) {
+	int canalId = -1;
+	for(int i = 0; i < 4096 / sizeof(Canal); i++) {
+		if(canals[i].name == canalName) {
+			canalId = i;
+			break;
+		}
+	}
+	
+	Process *guest = scheduler_find(guestProcessId);
+	
+	if(canalId == -1) return -1;
+	if(canals[canalId].hostProcess == guest) return -1;
+	
+	uint32_t hostCR3 = canals[canalId].hostProcess->ctx.cr3;
+	uint32_t hostAreaVirtAddr = vpm_find_free(hostCR3, VPM_USER, areaSize);
+	
+	uint32_t guestCR3 = guest->ctx.cr3;
+	uint32_t guestAreaVirtAddr = vpm_find_free(guestCR3, VPM_USER, areaSize);
+	
+	/* TODO: Fail if areas not found. */
+	
+	for(int i = 0; i < 4096 / sizeof(Link); i++) {
+		if(links[i].canalId == -1) {
+			links[i].canalId = canalId;
+			links[i].guestProcess = guest;
+			links[i].hostVirtAddr = hostAreaVirtAddr;
+			links[i].guestVirtAddr = guestAreaVirtAddr;
+			links[i].flags = 0;
+			
+			uint32_t pages = (areaSize + 4095) / 4096;
+			
+			uint32_t phys = ppm_alloc(pages);
+			vpm_map(hostCR3, VPM_USER, hostAreaVirtAddr, phys, areaSize);
+			vpm_map(guestCR3, VPM_USER, guestAreaVirtAddr, phys, areaSize);
+			
+			canals[canalId].hostProcess->eventsRequest++;
+			
+			links[i].nextInCanal = canals[canalId].links;
+			canals[canalId].links = &links[i];
+			
+			links[i].nextGuestLink = guest->guestLinks;
+			guest->guestLinks = &links[i];
+			
+			return guestAreaVirtAddr;
+		}
+	}
+	
+	return -1;
+}
+
+uint32_t canal_accept(uint16_t hostProcessId, uint32_t canalId) {
+	for(int i = 0; i < 4096 / sizeof(Link); i++) {
+		if(links[i].canalId != -1 && canals[links[i].canalId].hostProcess->id == hostProcessId && (canalId == (uint32_t)-1 || links[i].canalId == canalId) && !(links[i].flags & LINK_FLAG_ACCEPTED)) {
+			links[i].flags = links[i].flags | LINK_FLAG_ACCEPTED;
+			prcs("RETURNING ");
+			pri(links[i].hostVirtAddr);
+			pr('\n');
+			return links[i].hostVirtAddr;
+		}
+	}
+	
+	return -1;
+}
+
+void link_send_signal_from(uint16_t processId, uint32_t processAreaAddr, uint32_t meta) {
+	for(int i = 0; i < 4096 / sizeof(Link); i++) {
+		if(links[i].guestProcess->id == processId && links[i].guestVirtAddr == processAreaAddr) {
+			if(!(links[i].flags & LINK_FLAG_G2H)) {
+				canals[links[i].canalId].hostProcess->eventsSignal++;
+				links[i].flags |= LINK_FLAG_G2H;
+			}
+			
+			links[i].metaG2H = meta;
+			
+			return;
+		} else if(canals[links[i].canalId].hostProcess->id == processId && links[i].hostVirtAddr == processAreaAddr) {
+			if(!(links[i].flags & LINK_FLAG_H2G)) {
+				links[i].guestProcess->eventsSignal++;
+				links[i].flags |= LINK_FLAG_H2G;
+			}
+			
+			links[i].metaH2G = meta;
+			
+			return;
+		}
+	}
+}
+
+uint32_t is_signal_awaiting_to_concerning_link(uint16_t processId, uint32_t area, uint32_t *metaOut, uint32_t *canalId) {
+	for(int i = 0; i < 4096 / sizeof(Link); i++) {
+		if((links[i].guestProcess->id == processId && (area == -1 || links[i].guestVirtAddr == area) && (links[i].flags & LINK_FLAG_H2G))
+			|| (canals[links[i].canalId].hostProcess->id == processId && (area == -1 || area == links[i].hostVirtAddr) && (links[i].flags & LINK_FLAG_G2H))) {
+			
+			*canalId = links[i].canalId;
+			
+			return signal_accept(scheduler_find(processId), &links[i], metaOut);
+		}
+	}
+	return -1;
+}
+
+uint32_t signal_accept(Process *p, Link *l, uint32_t *meta) {
+	p->eventsSignal--;
+	
+	if(l->guestProcess == p) {
+		l->flags &= ~LINK_FLAG_H2G;
+		*meta = l->metaH2G;
+		return l->guestVirtAddr;
+	} else {
+		l->flags &= ~LINK_FLAG_G2H;
+		*meta = l->metaG2H;
+		return l->hostVirtAddr;
+	}
+}

src/kernel/arch/x86/canal.h

@@ -0,0 +1,42 @@
+#ifndef _CANAL_H
+#define _CANAL_H
+
+#include<stdint.h>
+
+#define LINK_FLAG_ACCEPTED 1
+
+#define LINK_FLAG_G2H 2
+#define LINK_FLAG_H2G 4
+
+struct Process;
+struct Link;
+
+typedef struct Canal {
+	uint64_t name;
+	struct Process *hostProcess;
+	struct Canal *ownedNext;
+	struct Link *links;
+} __attribute__((packed)) Canal;
+
+typedef struct Link {
+	uint32_t canalId;
+	uint32_t hostVirtAddr;
+	struct Process *guestProcess;
+	uint32_t guestVirtAddr;
+	uint32_t metaG2H;
+	uint32_t metaH2G;
+	uint8_t flags;
+	struct Link *nextInCanal;
+	struct Link *nextGuestLink;
+} __attribute__((packed)) Link;
+
+uint32_t canal_create(uint64_t name, uint16_t hostProcessId);
+uint32_t link_create(uint64_t canal, uint16_t guestProcessId, uint32_t areaSize);
+uint32_t canal_accept(uint16_t hostProcessId, uint32_t canalId);
+void link_send_signal_from(uint16_t processId, uint32_t processAreaAddr, uint32_t meta);
+uint32_t link_get_oldest_signal_to(uint16_t processId);
+uint32_t is_signal_awaiting_to_concerning_link(uint16_t processId, uint32_t area, uint32_t *metaOut, uint32_t *canalId);
+
+uint32_t signal_accept(struct Process *p, Link *l, uint32_t *meta);
+
+#endif

src/kernel/arch/x86/consts.h

@@ -0,0 +1,20 @@
+#ifndef _X86_CONSTS_H
+#define _X86_CONSTS_H
+
+#define EFLAGS_IOPL_0 (0 << 12)
+#define EFLAGS_IOPL_1 (1 << 12)
+#define EFLAGS_IOPL_2 (2 << 12)
+#define EFLAGS_IOPL_3 (3 << 12)
+#define EFLAGS_IF (1 << 9)
+#define EFLAGS_RESET 2
+
+#define RPL_0 0
+#define RPL_1 1
+#define RPL_2 2
+#define RPL_3 3
+
+#define PAGE_BIT_PRESENT 1
+#define PAGE_BIT_WRITABLE 2
+#define PAGE_BIT_USER 4
+
+#endif

src/kernel/arch/x86/interrupt.asm

@@ -0,0 +1,827 @@
+cpu 386
+bits 32
+
+TIMER_FREQUENCY equ 250;Hz
+
+; TODO: perform IO waits
+
+section .text
+
+extern scheduler_switch
+extern scheduler_CurrentProcess
+pit_handler:
+extern scheduler_Tick
+	inc dword [scheduler_Tick]
+	push eax
+	mov al, 0x20
+	out 0x20, al
+	pop eax
+	
+ctx_switch:
+	push eax
+	
+	mov eax, [scheduler_CurrentProcess]
+	test eax, eax
+	jnz .a
+	add esp, 16
+	jmp .b
+.a:
+	pop dword [eax + 2] ;eax
+	pop dword [eax + 34] ;eip
+	pop dword [eax + 42] ;cs
+	pop dword [eax + 38] ;eflags
+	mov [eax + 6], ebx
+	mov [eax + 10], ecx
+	mov [eax + 14], edx
+	mov [eax + 18], esi
+	mov [eax + 22], edi
+	mov [eax + 26], ebp
+	cmp dword [eax + 42], 8 ;CS 8 implies ring 0
+	jne .isR3
+	mov [eax + 30], esp
+	jmp .b
+.isR3:
+	pop dword [eax + 30] ;esp
+	add esp, 4 ;ss
+.b:
+	call scheduler_switch
+	jmp jump_to_process
+
+global jump_to_process
+jump_to_process:
+	mov eax, [scheduler_CurrentProcess]
+	test eax, eax
+	jnz .c
+	; No next process, goto hlt loop.
+	push dword 0x202
+	push dword 8
+	push haltLoop
+	iretd
+.c:
+	mov ebx, [eax + 46]
+	mov cr3, ebx
+	mov ebx, [eax + 6]
+	mov ecx, [eax + 10]
+	mov edx, [eax + 14]
+	mov esi, [eax + 18]
+	mov edi, [eax + 22]
+	mov ebp, [eax + 26]
+	cmp dword [eax + 42], 8
+	je .isR0
+	push dword 32 | 3
+	push dword [eax + 30]
+	jmp .b
+.isR0:
+	mov esp, [eax + 30]
+.b:
+	push dword [eax + 38] ;eflags
+	push dword [eax + 42] ;cs
+	push dword [eax + 34] ;eip
+	mov eax, [eax + 2]
+	iretd
+
+set_bit:
+	push edx
+	push ebx
+	push ecx
+	push eax
+	
+	shl ebx, 1
+	
+	mov dx, 0x3C4
+	mov al, 2
+	out dx, al
+	
+	mov ecx, ebx
+	and ecx, 3
+	mov eax, 1
+	shl eax, cl
+	
+	mov dx, 0x3C5
+	out dx, al
+	
+	mov ecx, ebx
+	sar ecx, 2
+	
+	pop eax
+	
+	mov [0xA0000 + ecx], byte al
+	
+	pop ecx
+	pop ebx
+	pop edx
+	
+	ret
+
+write_dword:
+	push eax
+	push ecx
+	
+	mov ecx, 32
+	
+.lup:
+	
+	rol eax, 1
+	
+	push eax
+	and eax, 1
+	shl eax, 4
+	add eax, 20
+	call set_bit
+	pop eax
+.nz:
+	inc ebx
+	
+	dec ecx
+	jnz .lup
+	
+	pop ecx
+	pop eax
+	
+	ret
+
+clear_screen:
+	push eax
+	push ecx
+	push edx
+	push edi
+	
+	mov dx, 0x3C4
+	mov al, 2
+	out dx, al
+	
+	inc dx
+	mov al, 15
+	out dx, al
+	
+	mov edi, 0xA0000
+	mov ecx, 4800
+	mov eax, 0
+	rep stosd
+	
+	pop edi
+	pop edx
+	pop ecx
+	pop eax
+	
+	ret
+
+pgf_handler:
+	call clear_screen
+	
+	pop eax ;error code
+	mov ebx, 0
+	call write_dword
+	
+	inc ebx
+	
+	mov eax, cr2 ;accessed address
+	call write_dword
+	
+	inc ebx
+	
+	pop eax ;instruction cause of exception
+	call write_dword
+	
+	add ebx, 100
+	
+	mov eax, [eax] ;instruction data
+	call write_dword
+.LOUP:
+	hlt
+	jmp .LOUP
+
+dbf_handler:
+	add esp, 4
+	mov [0xA0000], dword "D_F_"
+.LOUP:
+	hlt
+	jmp .LOUP
+
+gpf_handler:
+	pop eax
+	movzx ebx, al
+	and bl, 0xF
+	mov bl, [.tbl + ebx]
+	mov [0xB800E], bl
+	movzx ebx, al
+	shr bl, 4
+	mov bl, [.tbl + ebx]
+	mov [0xB800C], bl
+	shr eax, 8
+	movzx ebx, al
+	and bl, 0xF
+	mov bl, [.tbl + ebx]
+	mov [0xB800A], bl
+	movzx ebx, al
+	shr bl, 4
+	mov bl, [.tbl + ebx]
+	mov [0xB8008], bl
+	shr eax, 8
+	movzx ebx, al
+	and bl, 0xF
+	mov bl, [.tbl + ebx]
+	mov [0xB8006], bl
+	movzx ebx, al
+	shr bl, 4
+	mov bl, [.tbl + ebx]
+	mov [0xB8004], bl
+	shr eax, 8
+	movzx ebx, al
+	and bl, 0xF
+	mov bl, [.tbl + ebx]
+	mov [0xB8002], bl
+	movzx ebx, al
+	shr bl, 4
+	mov bl, [.tbl + ebx]
+	mov [0xB8000], bl
+	shr eax, 8
+.LOUP:
+	hlt
+	jmp .LOUP
+.tbl: db "0123456789ABCDEF"
+
+spur_handler:
+	iret
+
+syscall_int_entry:
+	cmp eax, (.tblend - .tbl) / 4
+	jae .invalid
+	jmp [.tbl + eax * 4]
+.invalid:
+	mov eax, 0
+	iret
+.tbl:
+	dd .sys_vpm_map, .sys_canal_create, .sys_link_create, .sys_canal_accept, .sys_signal_send, .sys_signal_wait, .sys_pci_claim, .sys_pci_read, .sys_vpm_unmap, .sys_vpm_get_phys, .sys_sleep, .sys_pci_write_1, .sys_pci_write_2, .sys_pci_write_4, .sys_irq_claim, .sys_wait_for_irq, .sys_create_thread, .sys_spawn, .sys_abuse_map, .sys_release, .sys_abuse_state, .sys_exit, .sys_event_wait
+.tblend:
+.sys_vpm_map:
+	push ecx
+	push edx
+	push esi
+	push edi
+	push 1 ;USER mapping
+	mov eax, cr3
+	push eax
+extern vpm_map
+	call vpm_map
+	add esp, 16
+	pop edx
+	pop ecx
+	iret
+.sys_canal_create:
+	push ecx
+	push edx
+	mov eax, [scheduler_CurrentProcess]
+	movzx eax, word [eax] ;process ID
+	push eax
+	push esi
+	push edi
+extern canal_create
+	call canal_create
+	add esp, 12
+	pop edx
+	pop ecx
+	iret
+.sys_link_create:
+	push ecx
+	push edx
+	mov eax, [scheduler_CurrentProcess]
+	movzx eax, word [eax]
+	push eax
+	push esi
+	push edi
+extern link_create
+	call link_create
+	add esp, 12
+	pop edx
+	pop ecx
+	iret
+.sys_canal_accept:
+	push ecx
+	push edx
+	push edi
+	mov eax, [scheduler_CurrentProcess]
+	movzx eax, word [eax] ;id
+	push eax
+extern canal_accept
+	call canal_accept
+	add esp, 8
+	pop edx
+	pop ecx
+	iret
+.sys_signal_send:
+	push ecx
+	push edx
+	push esi
+	push edi
+	mov eax, [scheduler_CurrentProcess]
+	movzx eax, word [eax] ;id
+	push eax
+extern link_send_signal_from
+	call link_send_signal_from
+	pop eax
+	add esp, 8
+	pop edx
+	pop ecx
+	iret
+.sys_signal_wait:
+	push eax
+	mov eax, [scheduler_CurrentProcess]
+	mov [eax + 50], byte 1
+	mov [eax + 51], edi
+	pop eax
+	jmp ctx_switch
+.sys_pci_claim:
+	push ecx
+	push edx
+	push edi
+extern pci_claim
+	call pci_claim
+	pop edx
+	pop edx
+	pop ecx
+	iret
+.sys_pci_read:
+	push ecx
+	push edx
+	push esi
+	push edi
+extern pci_read
+	call pci_read
+	add esp, 8
+	pop edx
+	pop ecx
+	iret
+.sys_vpm_unmap:
+	push eax
+	push ecx
+	push edx
+	push esi
+	push edi
+	mov eax, cr3
+	push eax
+extern vpm_unmap
+	call vpm_unmap
+	add esp, 12
+	pop edx
+	pop ecx
+	pop eax
+	iret
+.sys_vpm_get_phys:
+	push ecx
+	push edx
+	push edi
+	mov eax, cr3
+	push eax
+extern vpm_get_phys
+	call vpm_get_phys
+	add esp, 8
+	pop edx
+	pop ecx
+	iret
+.sys_sleep:
+	push edx
+	push ebx
+	push eax
+	xor edx, edx
+	mov eax, edi
+	mov ebx, 1000 / TIMER_FREQUENCY
+	div ebx
+	add eax, [scheduler_Tick]
+	mov ebx, [scheduler_CurrentProcess]
+	mov byte [ebx + 50], 2
+	mov [ebx + 51], eax
+	pop eax
+	pop ebx
+	pop edx
+	jmp ctx_switch
+.sys_pci_write_1:
+	push eax
+	push ecx
+	push edx
+	push esi
+	push edi
+extern pci_write_1
+	call pci_write_1
+	add esp, 8
+	pop edx
+	pop ecx
+	pop eax
+	iret
+.sys_pci_write_2:
+	push eax
+	push ecx
+	push edx
+	push esi
+	push edi
+extern pci_write_2
+	call pci_write_2
+	add esp, 8
+	pop edx
+	pop ecx
+	pop eax
+	iret
+.sys_pci_write_4:
+	push eax
+	push ecx
+	push edx
+	push esi
+	push edi
+extern pci_write_4
+	call pci_write_4
+	add esp, 8
+	pop edx
+	pop ecx
+	pop eax
+	iret
+.sys_irq_claim:
+	push ecx
+	push edx
+	call irq_claim
+	pop edx
+	pop ecx
+	iret
+.sys_wait_for_irq:
+	push ecx
+	mov ecx, edi
+	; TODO: check privilege
+	call toggle_irq_mask
+	mov ecx, [scheduler_CurrentProcess]
+	mov [ecx + 50], byte 3 ; Wait type IRQ
+	pop ecx
+	jmp ctx_switch
+.sys_create_thread:
+	push eax
+	push ecx
+	push edx
+extern scheduler_spawn
+	call scheduler_spawn
+	mov ecx, cr3
+	mov [eax + 46], ecx ;cr3
+	mov [eax + 34], edi ;eip
+	pop edx
+	pop ecx
+	pop eax
+	mov eax, 1
+	iret
+.sys_spawn:
+	push ecx
+	push edx
+	call scheduler_spawn
+	test eax, eax
+	jnz .sys_spawn_spawned
+	pop edx
+	pop ecx
+	iret
+.sys_spawn_spawned:
+	push eax
+	mov eax, [scheduler_CurrentProcess]
+	movzx eax, word [eax] ;proc ID
+	push eax
+extern scheduler_enslave
+	call scheduler_enslave
+	add esp, 8
+	pop edx
+	pop ecx
+	iret
+.sys_abuse_map:
+	push ecx
+	push edx
+		push edi
+		call scheduler_find
+		add esp, 4
+	pop edx
+	pop ecx
+	
+	test eax, eax
+	jz .notfound
+	
+	push ecx
+	push edx
+		push ebx
+		push ecx
+		push edx
+		push esi
+		push dword [eax + 46] ;cr3
+		mov eax, cr3
+		push eax
+extern vpm_double_map
+		call vpm_double_map
+		add esp, 24
+	pop edx
+	pop ecx
+	
+	iret
+.notfound:
+	mov eax, -1
+	iret
+.sys_release:
+	push ecx
+	push edx
+	push edi
+extern scheduler_unslave
+	call scheduler_unslave
+	pop edi
+	pop edx
+	pop ecx
+	iret
+.sys_abuse_state:
+	push ecx
+	push edx
+		push edi
+		call scheduler_find
+		add esp, 4
+	pop edx
+	pop ecx
+	
+	test eax, eax
+	jz .notfound2
+	
+	push ecx
+	push edx
+		push edx
+		push esi
+		push eax
+extern process_abuse_state
+		call process_abuse_state
+		add esp, 12
+	pop edx
+	pop ecx
+	
+	iret
+.notfound2:
+	mov eax, -1
+	iret
+.sys_exit:
+	iret
+.sys_event_wait:
+	push eax
+	push ecx
+	push edx
+		push edi
+		push dword [scheduler_CurrentProcess]
+extern process_search_event
+		call process_search_event
+		add esp, 8
+	pop edx
+	pop ecx
+	pop eax
+	iret
+	
+
+global scheduler_start:
+scheduler_start:
+.pic_init:
+	mov al, 0x10 | 0x01
+	out 0x20, al
+	out 0xA0, al
+	
+	mov al, 32
+	out 0x21, al
+	mov al, 40
+	out 0xA1, al
+	
+	mov al, 4
+	out 0x21, al
+	mov al, 2
+	out 0xA1, al
+	
+	mov al, 0x01
+	out 0x21, al
+	out 0xA1, al
+	 
+.pic_mask:
+	mov al, ~1 ;PIT
+	out 0x21, al
+	mov al, ~0
+	out 0xA1, al
+	
+.idt:
+	; Page fault handler
+	mov eax, pgf_handler
+	and eax, 0x0000FFFF
+	or eax, 0x00080000
+	mov [idt + 14 * 8], eax
+	
+	mov eax, pgf_handler
+	and eax, 0xFFFF0000
+	or eax, (14 << 8) | (1 << 15)
+	mov [idt + 14 * 8 + 4], eax
+	
+	; Double fault handler
+	mov eax, dbf_handler
+	and eax, 0x0000FFFF
+	or eax, 0x00080000
+	mov [idt + 8 * 8], eax
+	
+	mov eax, dbf_handler
+	and eax, 0xFFFF0000
+	or eax, (14 << 8) | (1 << 15)
+	mov [idt + 8 * 8 + 4], eax
+	
+	; General protection fault handler
+	mov eax, gpf_handler
+	and eax, 0x0000FFFF
+	or eax, 0x00080000
+	mov [idt + 13 * 8], eax
+	
+	mov eax, gpf_handler
+	and eax, 0xFFFF0000
+	or eax, (14 << 8) | (1 << 15)
+	mov [idt + 13 * 8 + 4], eax
+	
+	; Timer scheduler handler
+	mov eax, pit_handler
+	and eax, 0x0000FFFF
+	or eax, 0x00080000
+	mov [idt + 32 * 8], eax
+	
+	mov eax, pit_handler
+	and eax, 0xFFFF0000
+	or eax, (14 << 8) | (1 << 15)
+	mov [idt + 32 * 8 + 4], eax
+	
+	; Spurious interrupt handler
+	mov eax, spur_handler
+	and eax, 0x0000FFFF
+	or eax, 0x00080000
+	mov [idt + 39 * 8], eax
+	
+	mov eax, spur_handler
+	and eax, 0xFFFF0000
+	or eax, (14 << 8) | (1 << 15)
+	mov [idt + 39 * 8 + 4], eax
+	
+	; Software interrupt syscall entry point
+	mov eax, syscall_int_entry
+	and eax, 0x0000FFFF
+	or eax, 0x00080000
+	mov [idt + 0xEC * 8], eax
+	
+	mov eax, syscall_int_entry
+	and eax, 0xFFFF0000
+	or eax, (14 << 8) | (1 << 15) | (3 << 13)
+	mov [idt + 0xEC * 8 + 4], eax
+	
+	push idt
+	push (idt.end - idt - 1) << 16
+	lidt [esp + 2]
+	add esp, 8
+	
+.pit:
+	mov al, (2 << 1) | (2 << 4)
+	out 0x43, al
+	
+	mov al, (1193182 / TIMER_FREQUENCY) % 256
+	out 0x40, al
+	mov al, (1193182 / TIMER_FREQUENCY) / 256
+	out 0x40, al
+	
+	sti
+	
+haltLoop:
+	hlt
+	jmp haltLoop
+
+; ecx = id
+toggle_irq_mask:
+	push ecx
+	push ebx
+	push edx
+	push eax
+	mov dx, 0x21
+	cmp cl, 8
+	jl .lower
+	sub cl, 8
+	mov dx, 0xA1
+.lower:
+	mov bl, 1
+	shl bl, cl
+	in al, dx
+	xor al, bl
+	out dx, al
+	pop eax
+	pop edx
+	pop ebx
+	pop ecx
+	ret
+
+irq_claim:
+	test edi, edi
+	jnz .notTimer
+	xor eax, eax
+	ret
+.notTimer:
+	mov eax, [scheduler_CurrentProcess]
+	mov bx, [eax] ;id
+	lea eax, [edi * 8]
+	lea eax, [eax * 2 + edi] ;17x
+	lea eax, [eax + irqhandlers - 17]
+	mov [eax + 2], esi ;mov edi, handler
+	mov [eax + 7], bx ;mov esi, process id
+	mov ecx, edi
+	mov [eax + 12], cl
+	
+	push eax
+	and eax, 0x0000FFFF
+	or eax, 0x00080000
+	mov [idt + edi * 8 + 256], eax
+	
+	pop eax
+	and eax, 0xFFFF0000
+	or eax, (14 << 8) | (1 << 15)
+	mov [idt + edi * 8 + 256 + 4], eax
+	
+	mov eax, 1
+	ret
+
+irqhandlers:
+	pushad
+	mov edi, strict dword 0
+	mov esi, strict dword 0
+	mov cl, strict byte 0
+	jmp strict word irqhandler
+	pushad
+	mov edi, strict dword 0
+	mov esi, strict dword 0
+	mov cl, strict byte 0
+	jmp strict word irqhandler
+	pushad
+	mov edi, strict dword 0
+	mov esi, strict dword 0
+	mov cl, strict byte 0
+	jmp strict word irqhandler
+	pushad
+	mov edi, strict dword 0
+	mov esi, strict dword 0
+	mov cl, strict byte 0
+	jmp strict word irqhandler
+	pushad
+	mov edi, strict dword 0
+	mov esi, strict dword 0
+	mov cl, strict byte 0
+	jmp strict word irqhandler
+	pushad
+	mov edi, strict dword 0
+	mov esi, strict dword 0
+	mov cl, strict byte 0
+	jmp strict word irqhandler
+	pushad
+	mov edi, strict dword 0
+	mov esi, strict dword 0
+	mov cl, strict byte 0
+	jmp strict word irqhandler
+	pushad
+	mov edi, strict dword 0
+	mov esi, strict dword 0
+	mov cl, strict byte 0
+	jmp strict word irqhandler
+	pushad
+	mov edi, strict dword 0
+	mov esi, strict dword 0
+	mov cl, strict byte 0
+	jmp strict word irqhandler
+	pushad
+	mov edi, strict dword 0
+	mov esi, strict dword 0
+	mov cl, strict byte 0
+	jmp strict word irqhandler
+	pushad
+	mov edi, strict dword 0
+	mov esi, strict dword 0
+	mov cl, strict byte 0
+	jmp strict word irqhandler
+	pushad
+	mov edi, strict dword 0
+	mov esi, strict dword 0
+	mov cl, strict byte 0
+	jmp strict word irqhandler
+	pushad
+	mov edi, strict dword 0
+	mov esi, strict dword 0
+	mov cl, strict byte 0
+	jmp strict word irqhandler
+	pushad
+	mov edi, strict dword 0
+	mov esi, strict dword 0
+	mov cl, strict byte 0
+	jmp strict word irqhandler
+	pushad
+	mov edi, strict dword 0
+	mov esi, strict dword 0
+	mov cl, strict byte 0
+irqhandler:
+	call toggle_irq_mask
+	mov al, 0x20
+	out 0x20, al
+	push esi
+extern scheduler_find
+	call scheduler_find
+	add esp, 4
+	mov [eax + 55], edi
+	popad
+	iret
+
+idt:
+	times 256 dq 0
+.end:

src/kernel/arch/x86/log.asm

@@ -0,0 +1,67 @@
+section .data
+
+section .text
+
+global klogc
+klogc:
+	mov eax, [esp + 4]
+	out 0xE9, al
+	ret
+
+global klogs
+klogs:
+	mov ecx, [esp + 4]
+.lup:
+	mov al, [ecx]
+	test al, al
+	jz .end
+	out 0xE9, al
+	inc ecx
+	jmp .lup
+.end:
+	ret
+
+;~ global klogp
+;~ klogp:
+	;~ push eax
+	;~ mov edx, [esp + 12]
+	;~ mov eax, edx
+	;~ shr eax, 28
+	;~ call .logh
+	;~ mov eax, edx
+	;~ shr eax, 24
+	;~ and al, 0x0F
+	;~ call .logh
+	;~ mov eax, edx
+	;~ shr eax, 20
+	;~ and al, 0x0F
+	;~ call .logh
+	;~ mov eax, edx
+	;~ shr eax, 16
+	;~ and al, 0x0F
+	;~ call .logh
+	;~ mov eax, edx
+	;~ shr eax, 12
+	;~ and al, 0x0F
+	;~ call .logh
+	;~ mov eax, edx
+	;~ shr eax, 8
+	;~ and al, 0x0F
+	;~ call .logh
+	;~ mov eax, edx
+	;~ shr eax, 4
+	;~ and al, 0x0F
+	;~ call .logh
+	;~ mov eax, edx
+	;~ and al, 0x0F
+	;~ call .logh
+	;~ pop eax
+	;~ ret
+;~ .logh:
+	;~ add al, '0'
+	;~ cmp al, '0' + 9
+	;~ jle .p
+	;~ add al, 7
+;~ .p:
+	;~ out 0xE9, al
+	;~ ret

src/kernel/arch/x86/mem.c

@@ -0,0 +1,15 @@
+#include"mem.h"
+
+void kmemw1(void *_ptr, uint8_t val, size_t amount) {
+	uint8_t *ptr = _ptr;
+	for(size_t i = 0; i < amount; i++) {
+		*(ptr++) = val;
+	}
+}
+
+void kmemw4(void *_ptr, uint32_t val, size_t amount) {
+	uint32_t *ptr = _ptr;
+	for(size_t i = 0; i < amount; i++) {
+		*(ptr++) = val;
+	}
+}

src/kernel/arch/x86/mem.h

@@ -0,0 +1,10 @@
+#ifndef _MEM_H
+#define _MEM_H
+
+#include<stdint.h>
+#include<stddef.h>
+
+void kmemw1(void *ptr, uint8_t val, size_t amount);
+void kmemw4(void *ptr, uint32_t val, size_t amount);
+
+#endif

src/kernel/arch/x86/pci.c

@@ -0,0 +1,60 @@
+#include"pci.h"
+
+#include<stddef.h>
+#include"vpm.h"
+
+size_t claimedCount = 0;
+uint32_t *claimed;
+
+void pci_init() {
+	uint32_t cr3;
+	asm volatile("movl %%cr3, %0" : "=r" (cr3) :);
+	
+	claimed = vpm_map(cr3, VPM_KERNEL, -1, -1, 4096);
+}
+
+uint32_t pci_claim(uint32_t cscp) {
+	for(int b = 0; b < 256; b++) {
+		for(int d = 0; d < 32; d++) {
+			for(int f = 0; f < 8; f++) {
+				uint32_t id = (b << 16) | (d << 11) | (f << 8);
+				if((pci_read(id, 8) & 0xFFFFFF00) == cscp) {
+					for(size_t i = 0; i < claimedCount; i++) {
+						if(claimed[i] == id) {
+							goto cont;
+						}
+					}
+					
+					claimed[claimedCount++] = id;
+					return id;
+				}
+cont:;
+			}
+		}
+	}
+	return -1;
+}
+
+uint32_t pci_read(uint32_t id, uint8_t offset) {
+	asm volatile("outl %%eax, %%dx" : : "a"(0x80000000 | id | offset), "d"(0xCF8));
+	
+	uint32_t ret;
+	asm volatile("inl %%dx, %%eax" : "=a"(ret) : "d"(0xCFC));
+	
+	return ret;
+}
+
+void pci_write_1(uint32_t id, uint8_t offset, uint8_t val) {
+	asm volatile("outl %%eax, %%dx" : : "a"(0x80000000 | id | offset), "d"(0xCF8));
+	asm volatile("outb %%al, %%dx" : : "a"(val), "d"(0xCFC + (offset & 3)));
+}
+
+void pci_write_2(uint32_t id, uint8_t offset, uint16_t val) {
+	asm volatile("outl %%eax, %%dx" : : "a"(0x80000000 | id | offset), "d"(0xCF8));
+	asm volatile("outw %%ax, %%dx" : : "a"(val), "d"(0xCFC + (offset & 2)));
+}
+
+void pci_write_4(uint32_t id, uint8_t offset, uint32_t val) {
+	asm volatile("outl %%eax, %%dx" : : "a"(0x80000000 | id | offset), "d"(0xCF8));
+	asm volatile("outl %%eax, %%dx" : : "a"(val), "d"(0xCFC));
+}

src/kernel/arch/x86/pci.h

@@ -0,0 +1,13 @@
+#ifndef _PCI_H
+#define _PCI_H
+
+#include<stdint.h>
+
+void pci_init();
+uint32_t pci_claim(uint32_t);
+uint32_t pci_read(uint32_t, uint8_t);
+void pci_write_1(uint32_t, uint8_t, uint8_t);
+void pci_write_2(uint32_t, uint8_t, uint16_t);
+void pci_write_4(uint32_t, uint8_t, uint32_t);
+
+#endif

src/kernel/arch/x86/ppm.c

@@ -0,0 +1,64 @@
+#include"ppm.h"
+
+#include<stdint.h>
+
+typedef struct __attribute__((packed)) {
+	uint64_t start;
+	uint32_t length;
+	uint8_t flags;
+	uint8_t data[];
+} PPMBitmap;
+
+PPMBitmap *ppm_Bitmap;
+
+static int availability_test(PPMBitmap *bm, size_t pg, size_t len) {
+	if(pg > bm->length - len) {
+		return 0;
+	}
+	
+	while(len--) {
+		if(bm->data[pg / 8] & (1 << (pg % 8))) {
+			return 0;
+		}
+		
+		pg++;
+	}
+	
+	return 1;
+}
+
+static int mark(PPMBitmap *bm, size_t pg, size_t len) {
+	if(pg > bm->length - len) {
+		return 0;
+	}
+	
+	while(len--) {
+		bm->data[pg / 8] |= 1 << (pg % 8);
+		
+		pg++;
+	}
+	
+	return 1;
+}
+
+uint64_t ppm_alloc(size_t len) {
+	PPMBitmap *bm = ppm_Bitmap;
+	
+	for(int bmid = 0; bmid < 2; bmid++) {
+		if(bm->length < len) {
+			continue;
+		}
+		
+		for(size_t pg = 0; pg <= bm->length - len; pg++) {
+			if(availability_test(bm, pg, len)) {
+				if(mark(bm, pg, len)) {
+					return bm->start + pg * 4096;
+				}
+			}
+		}
+		
+		bm = (PPMBitmap*) ((uintptr_t) bm + sizeof(*bm) + (bm->length + 7) / 8);
+	}
+	
+	return 0;
+}

src/kernel/arch/x86/ppm.h

@@ -0,0 +1,9 @@
+#ifndef _PPM_H
+#define _PPM_H
+
+#include<stddef.h>
+#include<stdint.h>
+
+uint64_t ppm_alloc(size_t len);
+
+#endif

src/kernel/arch/x86/process.h

@@ -0,0 +1,46 @@
+#ifndef _PROCESS_H
+#define _PROCESS_H
+
+#include<stdint.h>
+
+struct Canal;
+struct Link;
+
+typedef struct __attribute__((packed)) {
+	uint32_t eax;    //0
+	uint32_t ebx;    //4
+	uint32_t ecx;    //8
+	uint32_t edx;    //12
+	uint32_t esi;    //16
+	uint32_t edi;    //20
+	uint32_t ebp;    //24
+	uint32_t esp;    //28
+	uint32_t eip;    //32
+	uint32_t eflags; //36
+	uint32_t cs;     //40
+	uint32_t cr3;    //44
+} Context;
+
+#define WAIT_TYPE_NONE 0
+#define WAIT_TYPE_SIGNAL 1
+#define WAIT_TYPE_TIME 2
+#define WAIT_TYPE_IRQ 3
+#define WAIT_TYPE_MINOR 4
+
+typedef struct __attribute__((packed)) Process {
+	uint16_t id;
+	Context ctx;
+	uint8_t waitType;
+	uint32_t waitValue;
+	uint32_t irqWaiting;
+	uint8_t minipriority;
+	struct Process *next;
+	uint16_t ownerId;
+	uint8_t eventsRequest;
+	uint8_t eventsSignal;
+	uint8_t eventsClosed;
+	struct Canal *ownedCanal;
+	struct Link *guestLinks;
+} Process;
+
+#endif

src/kernel/arch/x86/scheduler.c

@@ -0,0 +1,301 @@
+#include"scheduler.h"
+
+#include"vpm.h"
+#include"mem.h"
+
+#include<kernel/arch/x86/consts.h>
+
+#include"canal.h"
+
+#define PROCESS_ZONE_SIZE 4096
+#define HANDLE_ZONE_SIZE 4096
+
+Process *scheduler_ProcessZone = 0;
+Process *scheduler_ProcessQueues[TOTAL_PRIORITIES];
+
+static int scheduler_CurrentPriority;
+volatile Process *scheduler_CurrentProcess = 0;
+
+uint32_t scheduler_Tick;
+
+void klogs(const char *str);
+void klogc(int c);
+
+static void klogp(void *p) {
+	static char hex[16]="0123456789ABCDEF";
+	
+	uintptr_t i = (uintptr_t) p;
+	klogc(hex[(i >> 28) & 0xF]);
+	klogc(hex[(i >> 24) & 0xF]);
+	klogc(hex[(i >> 20) & 0xF]);
+	klogc(hex[(i >> 16) & 0xF]);
+	klogc(hex[(i >> 12) & 0xF]);
+	klogc(hex[(i >> 8) & 0xF]);
+	klogc(hex[(i >> 4) & 0xF]);
+	klogc(hex[(i >> 0) & 0xF]);
+}
+
+void scheduler_init() {
+	uint32_t cr3;
+	asm volatile("movl %%cr3, %0" : "=r" (cr3) :);
+	
+	scheduler_ProcessZone = vpm_map(cr3, VPM_KERNEL, -1, -1, PROCESS_ZONE_SIZE);
+	kmemw1(scheduler_ProcessZone, 0, PROCESS_ZONE_SIZE);
+}
+
+static int consider_switch(Process *p) {
+	if(p->waitType == WAIT_TYPE_SIGNAL) {
+		uint32_t meta = 0;
+		uint32_t canalId = -1;
+		uint32_t which = is_signal_awaiting_to_concerning_link(p->id, p->waitValue, &meta, &canalId);
+		if(which != 0xFFFFFFFF) {
+			p->waitType = WAIT_TYPE_NONE;
+			p->ctx.eax = which;
+			p->ctx.edx = meta;
+			p->ctx.ebx = canalId;
+		} else return 0;
+	} else if(p->waitType == WAIT_TYPE_TIME) {
+		if(p->waitValue <= scheduler_Tick) {
+			p->waitType = WAIT_TYPE_NONE;
+		} else return 0;
+	} else if(p->waitType == WAIT_TYPE_IRQ) {
+		if(p->irqWaiting) {
+			p->waitType = WAIT_TYPE_NONE;
+			p->ctx.eip = p->irqWaiting;
+			p->irqWaiting = 0;
+		} else return 0;
+	} else if(p->waitType == WAIT_TYPE_MINOR) {
+		//asm("xchg %bx, %bx");
+		return 0;
+	}
+	
+	/*static char hex[16]="0123456789ABCDEF";
+	klogc('0' + p->id);
+	klogc(' ');
+	klogc(hex[p->priority / PRIORITY_UNIT]);
+	klogc('\n');*/
+	
+	scheduler_CurrentPriority = p->minipriority / PRIORITY_UNIT;
+	scheduler_CurrentProcess = p;
+	
+	return 1;
+}
+
+void scheduler_switch() {
+	if(scheduler_CurrentProcess == 0) {
+		for(int p = 0; p < TOTAL_PRIORITIES; p++) {
+			Process *q = scheduler_ProcessQueues[p];
+			while(q) {
+				if(consider_switch(q)) {
+					return;
+				}
+				q = q->next;
+			}
+		}
+		
+		scheduler_CurrentProcess = 0;
+		
+		return;
+	}
+	
+	for(int p = 0; p < scheduler_CurrentPriority; p++) {
+		Process *q = scheduler_ProcessQueues[p];
+		
+		while(q) {
+			if(consider_switch(q)) {
+				return;
+			}
+			q = q->next;
+		}
+	}
+	
+	Process *p = scheduler_CurrentProcess;
+	do {
+		p = p->next;
+		
+		if(!p) {
+			p = scheduler_ProcessQueues[scheduler_CurrentPriority];
+		}
+		
+		if(consider_switch(p)) {
+			return;
+		}
+	} while(p != scheduler_CurrentProcess);
+	
+	for(int p = scheduler_CurrentPriority + 1; p < TOTAL_PRIORITIES; p++) {
+		Process *q = scheduler_ProcessQueues[p];
+		
+		while(q) {
+			if(consider_switch(q)) {
+				return;
+			}
+			q = q->next;
+		}
+	}
+	
+	scheduler_CurrentProcess = 0;
+}
+
+Process *scheduler_find(uint16_t id) {
+	for(int p = 0; p < TOTAL_PRIORITIES; p++) {
+		Process *q = scheduler_ProcessQueues[p];
+		
+		while(q) {
+			if(q->id == id) {
+				return q;
+			}
+			q = q->next;
+		}
+	}
+	
+	return NULL;
+}
+
+static uint16_t next_free_id() {
+	for(uint32_t id = 1; id < 65536; id++) {
+		if(!scheduler_find(id)) {
+			return id;
+		}
+	}
+	return 0;
+}
+
+Process *scheduler_spawn(int minipriority) {
+	int found = 0;
+	Process *p = scheduler_ProcessZone;
+	
+	for(size_t i = 0; i < PROCESS_ZONE_SIZE / sizeof(Process); i++) {
+		if(p->id == 0) {
+			found = 1;
+			break;
+		}
+		
+		p++;
+	}
+	
+	if(!found) {
+		return NULL;
+	}
+	
+	kmemw1(p, 0, sizeof(*p));
+	p->id = next_free_id();
+	p->ctx.eflags = EFLAGS_RESET | EFLAGS_IF | EFLAGS_IOPL_3;
+	p->ctx.cs = 24 | RPL_3;
+	p->waitType = WAIT_TYPE_NONE;
+	p->irqWaiting = 0;
+	p->minipriority = minipriority;
+	p->next = scheduler_ProcessQueues[minipriority / PRIORITY_UNIT];
+	p->eventsRequest = 0;
+	p->eventsSignal = 0;
+	p->eventsClosed = 0;
+	p->ownedCanal = NULL;
+	p->guestLinks = NULL;
+	
+	scheduler_ProcessQueues[minipriority / PRIORITY_UNIT] = p;
+	
+	return p;
+}
+
+size_t scheduler_enslave(uint16_t ownerId, Process *child) {
+	child->waitType = WAIT_TYPE_MINOR;
+	child->ownerId = ownerId;
+	
+	// Should not be how it works.
+	child->ctx.cr3 = vpm_create_space();
+	
+	return child->id;
+}
+
+int scheduler_unslave(size_t handle) {
+	struct Process *proc = scheduler_find(handle);
+	if(proc && proc->waitType == WAIT_TYPE_MINOR) {
+		proc->waitType = WAIT_TYPE_NONE;
+		return 0;
+	}
+	return -1;
+}
+
+#define SYS_STATE_X86_EIP 0
+int process_abuse_state(struct Process *proc, size_t type, uintmax_t val) {
+	switch(type) {
+	case SYS_STATE_X86_EIP:
+		proc->ctx.eip = val;
+		break;
+	default:
+		return -1;
+	}
+	return 0;
+}
+
+#define KEV_TYPE_NONE 0
+#define KEV_TYPE_LINK_REQUEST 1
+#define KEV_TYPE_LINK_SIGNAL 2
+#define KEV_TYPE_LINK_CLOSE 3
+typedef struct __attribute__((packed)) {
+	uint8_t type;
+	uint32_t canalId;
+	void *area;
+	uint32_t meta;
+} KEvent;
+
+int process_search_event(Process *proc, KEvent *kev) {
+	if(proc->eventsRequest) {
+		kev->type = KEV_TYPE_LINK_REQUEST;
+		
+		Canal *c = proc->ownedCanal;
+		while(c) {
+			Link *l = c->links;
+			
+			while(l) {
+				if(!(l->flags & LINK_FLAG_ACCEPTED)) {
+					kev->canalId = l->canalId;
+					
+					return 1;
+				}
+				
+				l = l->nextInCanal;
+			}
+			
+			c = c->ownedNext;
+		}
+	}
+	
+	if(proc->eventsSignal) {
+		kev->type = KEV_TYPE_LINK_SIGNAL;
+		
+		Link *l = proc->guestLinks;
+		while(l) {
+			if(l->flags & LINK_FLAG_H2G) {
+				kev->area = signal_accept(proc, l, &kev->meta);
+				
+				return 1;
+			}
+			
+			l = l->nextGuestLink;
+		}
+	}
+	
+	if(proc->eventsSignal) {
+		kev->type = KEV_TYPE_LINK_SIGNAL;
+		
+		Canal *c = proc->ownedCanal;
+		while(c) {
+			Link *l = c->links;
+			while(l) {
+				if(l->flags & LINK_FLAG_G2H) {
+					kev->area = signal_accept(proc, l, &kev->meta);
+					
+					return 1;
+				}
+				
+				l = l->nextInCanal;
+			}
+			
+			c = c->ownedNext;
+		}
+	}
+	
+	kev->type = KEV_TYPE_NONE;
+	
+	return 0;
+}

src/kernel/arch/x86/scheduler.h

@@ -0,0 +1,22 @@
+#ifndef _SCHEDULER_H
+#define _SCHEDULER_H
+
+#include"process.h"
+#include<stddef.h>
+
+void scheduler_init();
+void scheduler_switch();
+Process *scheduler_spawn();
+size_t scheduler_create_handle(uint16_t ownerId, Process *child);
+
+Process *scheduler_find(uint16_t procId);
+
+#define MAX_MINIPRIORITY 255
+#define TOTAL_PRIORITIES 16
+#define PRIORITY_UNIT ((MAX_MINIPRIORITY + 1) / TOTAL_PRIORITIES)
+
+//extern size_t scheduler_ProcessCount;
+extern Process *scheduler_ProcessZone;
+extern Process *scheduler_ProcessQueues[TOTAL_PRIORITIES];
+
+#endif

src/kernel/arch/x86/vpm.c

@@ -0,0 +1,268 @@
+#include"vpm.h"
+
+#include"ppm.h"
+#include"mem.h"
+
+extern void *_kernel_end;
+
+void set_temporary_mapping(uint32_t addr) {
+	if((addr & 0xFFF) != 0) asm volatile("xchgw %bx, %bx");
+	
+	*(volatile uint32_t*) 0xFFFFFFF8 = addr | 3;
+	asm volatile("invlpg 0xFFFFE000" : : : "memory");
+}
+
+static uint32_t nextAllocs[2] = {[VPM_KERNEL] = 0, [VPM_USER] = 1 * 1024 * 1024 * 1024};
+
+void vpm_init(uint32_t kernelEnd) {
+	nextAllocs[VPM_KERNEL] = (kernelEnd + 4095) & ~4095;
+}
+
+/* TODO: Make better. */
+uint32_t vpm_find_free(uint32_t addrspace, AllocationOwner owner, size_t lenBytes) {
+	size_t len = (lenBytes + 4095) / 4096;
+	
+	uint32_t ret = nextAllocs[owner];
+	for(size_t i = 0; i < len;) {
+		uint32_t pdeid = (ret + i * 4096) >> 22;
+		uint32_t pteid = ((ret + i * 4096) >> 12) & 1023;
+		
+		set_temporary_mapping(addrspace);
+		
+		volatile uint32_t *pde = &((volatile uint32_t*) 0xFFFFE000)[pdeid];
+		if((*pde & 1) == 0) {
+			i += 4096;
+		} else {
+			set_temporary_mapping(*pde & ~0xFFF);
+			
+			if((((volatile uint32_t*) 0xFFFFE000)[pteid] & 1) == 0) {
+				i++;
+			} else {
+				ret += (i + 1) * 4096;
+				i = 0;
+			}
+		}
+	}
+	
+	return ret;
+}
+
+/* If phys is -1, map to uncontiguous physical pages anywhere, else map contiguous span of physical memory. */
+/* If phys is -2, automatically find contiguous span of physical memory */
+/* If virt is -1, find free space, else overwrite. */
+void *vpm_map(uint32_t addrspace, AllocationOwner owner, uint32_t virt, uint32_t phys, size_t len) {
+	uint32_t thisspace;
+	asm volatile("movl %%cr3, %0" : "=r" (thisspace) :);
+	
+	//if(len == 32768)asm("xchg %%bx, %%bx":::);
+	
+	if(virt == -1) {
+		virt = vpm_find_free(addrspace, owner, len);
+	}
+	
+	len = (len + 4095) / 4096;
+	
+	if(phys == -2) {
+		phys = ppm_alloc(len);
+		
+		if(!phys) {
+			goto error;
+		}
+	}
+	
+	for(size_t i = 0; i < len; i++) {
+		set_temporary_mapping(addrspace);
+		
+		uint32_t pdeid = (virt + i * 4096) >> 22;
+		uint32_t pteid = ((virt + i * 4096) >> 12) & 1023;
+		
+		volatile uint32_t *pde = &((volatile uint32_t*) 0xFFFFE000)[pdeid];
+		
+		if((*pde & 1) == 0) {
+			uint64_t physpag = ppm_alloc(1);
+			
+			if(physpag) {
+				*pde = physpag | 3 | (owner == VPM_USER ? 4 : 0);
+			} else {
+				goto error;
+			}
+		}
+		set_temporary_mapping(*pde & ~0xFFF);
+		
+		volatile uint32_t *pte = &((volatile uint32_t*) 0xFFFFE000)[pteid];
+		if(phys == -1) {
+			*pte = ppm_alloc(1);
+			if(!*pte) goto error;
+		} else {
+			*pte = phys + i * 4096;
+		}
+		*pte |= 3 | (owner == VPM_USER ? 4 : 0);
+		
+		if(thisspace == addrspace) {
+			//~ asm volatile("xchg %bx, %bx");
+			asm volatile("invlpg %0" : : "m"(*(char*) (virt + i * 4096)) : "memory");
+		}
+	}
+	
+	return (void*) virt;
+
+error:
+	// TODO: Deallocate physical pages.
+	
+	return 0;
+}
+
+int vpm_double_map(uint32_t cr3a, uint32_t cr3b, uint32_t *va, uint32_t *vb, uintptr_t phys, size_t len) {
+	uint32_t thisspace;
+	asm volatile("movl %%cr3, %0" : "=r" (thisspace) :);
+	
+	if(*va == -1) {
+		*va = vpm_find_free(cr3a, VPM_USER, len);
+	}
+	if(*vb == -1) {
+		*vb = vpm_find_free(cr3b, VPM_USER, len);
+	}
+	
+	len = (len + 4095) / 4096;
+	
+	if(phys == -2) {
+		phys = ppm_alloc(len);
+	}
+	
+	for(size_t i = 0; i < len; i++) {
+		uint32_t backend;
+		
+		if(phys == -1) {
+			backend = ppm_alloc(1);
+			if(!backend) goto error;
+		} else {
+			backend = phys + i * 4096;
+		}
+		
+		{
+			set_temporary_mapping(cr3a);
+			
+			uint32_t pdeid = (*va + i * 4096) >> 22;
+			uint32_t pteid = ((*va + i * 4096) >> 12) & 1023;
+			
+			volatile uint32_t *pde = &((volatile uint32_t*) 0xFFFFE000)[pdeid];
+			
+			if((*pde & 1) == 0) {
+				uintptr_t physpag = ppm_alloc(1);
+				
+				if(physpag) {
+					*pde = physpag | 3 | 4;
+				} else {
+					goto error;
+				}
+			}
+			
+			set_temporary_mapping(*pde & ~0xFFF);
+			
+			volatile uint32_t *pte = &((volatile uint32_t*) 0xFFFFE000)[pteid];
+			
+			*pte = backend | 3 | 4;
+			
+			if(thisspace == cr3a) {
+				asm volatile("invlpg %0" : : "m"(*(char*) (*va + i * 4096)) : "memory");
+			}
+		}
+		
+		{
+			set_temporary_mapping(cr3b);
+			
+			uint32_t pdeid = (*vb + i * 4096) >> 22;
+			uint32_t pteid = ((*vb + i * 4096) >> 12) & 1023;
+			
+			volatile uint32_t *pde = &((volatile uint32_t*) 0xFFFFE000)[pdeid];
+			
+			if((*pde & 1) == 0) {
+				uintptr_t physpag = ppm_alloc(1);
+				
+				if(physpag) {
+					*pde = physpag | 3 | 4;
+				} else {
+					goto error;
+				}
+			}
+			
+			set_temporary_mapping(*pde & ~0xFFF);
+			
+			volatile uint32_t *pte = &((volatile uint32_t*) 0xFFFFE000)[pteid];
+			
+			*pte = backend | 3 | 4;
+			
+			if(thisspace == cr3b) {
+				asm volatile("invlpg %0" : : "m"(*(char*) (*vb + i * 4096)) : "memory");
+			}
+		}
+	}
+	
+	return 1;
+
+error:
+	return 0;
+}
+
+uint32_t vpm_create_space() {
+	uint32_t curPD;
+	asm volatile("movl %%cr3, %0" : "=r"(curPD) :);
+	set_temporary_mapping(curPD);
+	
+	uint32_t pde0000 = *(volatile uint32_t*) 0xFFFFE000;
+	uint32_t pde1023 = *(volatile uint32_t*) 0xFFFFEFFC;
+	
+	uint32_t newPD = ppm_alloc(1);
+	
+	set_temporary_mapping(newPD);
+	
+	*(volatile uint32_t*) 0xFFFFE000 = pde0000;
+	kmemw4((void*) 0xFFFFE004, 0, 1022);
+	*(volatile uint32_t*) 0xFFFFEFFC = pde1023;
+	
+	return newPD;
+}
+
+void vpm_set_space(uint32_t cr3) {
+	asm volatile("movl %0, %%cr3" : : "r"(cr3));
+}
+
+uint32_t vpm_get_phys(uint32_t cr3, uint32_t v) {
+	uint32_t offset = v & 4095;
+	
+	uint32_t pdeid = v >> 22;
+	uint32_t pteid = (v >> 12) & 1023;
+	
+	set_temporary_mapping(cr3);
+	set_temporary_mapping(((uint32_t*) 0xFFFFE000)[pdeid] & ~0xFFF);
+	
+	return (((uint32_t*) 0xFFFFE000)[pteid] & ~0xFFF) + offset;
+}
+
+void vpm_unmap(uint32_t cr3, uint32_t virt, size_t len) {
+	uint32_t thisspace;
+	asm volatile("movl %%cr3, %0" : "=r"(thisspace) :);
+	
+	len = (len + 4095) / 4096;
+	
+	for(size_t i = 0; i < len; i++) {
+		set_temporary_mapping(cr3);
+		
+		uint32_t pdeid = (virt + i * 4096) >> 22;
+		uint32_t pteid = ((virt + i * 4096) >> 12) & 1023;
+		
+		volatile uint32_t *pde = &((volatile uint32_t*) 0xFFFFE000)[pdeid];
+		
+		if((*pde & 1) == 0) {
+			continue;
+		}
+		
+		set_temporary_mapping(*pde & ~0xFFF);
+		
+		((volatile uint32_t*) 0xFFFFE000)[pteid] = 0;
+		
+		if(thisspace == cr3) {
+			asm volatile("invlpg %0" : : "m"(*(char*) (virt + i * 4096)) : "memory");
+		}
+	}
+}

src/kernel/arch/x86/vpm.h

@@ -0,0 +1,24 @@
+#ifndef _VPM_H
+#define _VPM_H
+
+#include<stdint.h>
+#include<stddef.h>
+
+typedef struct {
+	uint32_t from, to;
+} SharedRange;
+
+typedef enum {
+	VPM_KERNEL, VPM_USER
+} AllocationOwner;
+
+void set_temporary_mapping(uint32_t addr);
+
+uint32_t vpm_find_free(uint32_t addrspace, AllocationOwner owner, size_t len);
+void *vpm_map(uint32_t cr3, AllocationOwner owner, uint32_t virt, uint32_t phys, size_t len);
+uint32_t vpm_create_space();
+void vpm_set_space(uint32_t cr3);
+uint32_t vpm_get_phys(uint32_t cr3, uint32_t);
+void vpm_unmap(uint32_t cr3, uint32_t virt, size_t len);
+
+#endif