boot/elf/src/main.rs

use core::fmt;
use std::{char, env, fs};

fn main() {
  let mut args = env::args().skip(1);
  let path = args.next().unwrap();

  let bytes = fs::read(path).unwrap();

  let elf = Elf::parse(&bytes).unwrap();
}

#[derive(Clone)]
struct InputBytes<'a> {
  bytes: &'a [u8],
}

impl<'a> InputBytes<'a> {
  fn new(bytes: &'a [u8]) -> Self {
    Self { bytes }
  }

  fn new_offset(bytes: &'a [u8], offset: usize) -> Self {
    Self::new(bytes).advanced(offset)
  }

  fn prefix_len(&self, n: usize) -> usize {
    self.bytes.len().min(n)
  }

  fn advanced(&self, n: usize) -> Self {
    Self {
      bytes: &self.bytes[self.prefix_len(n)..],
    }
  }

  fn advance(&mut self, n: usize) -> &mut Self {
    self.bytes = &self.bytes[self.prefix_len(n)..];
    self
  }

  fn peek(&self, n: usize) -> Result<&'a [u8], ParseError> {
    self.bytes.get(..n).ok_or(ParseError::UnexpectedEof)
  }

  fn pop(&mut self, n: usize) -> Result<&'a [u8], ParseError> {
    let res = self.peek(n);
    if res.is_ok() {
      self.advance(n);
    }
    res
  }

  fn pop_arr<const N: usize>(&mut self) -> Result<[u8; N], ParseError> {
    self.pop(N).map(|bs| <[u8; N]>::try_from(bs).unwrap())
  }

  fn pop_u8(&mut self) -> Result<u8, ParseError> {
    self.pop_arr::<1>().map(|[x]| x)
  }

  fn pop_u16(&mut self, endianness: Endianness) -> Result<u16, ParseError> {
    let bytes = self.pop_arr::<2>()?;
    Ok(endianness.read_u16(bytes))
  }

  fn pop_u32(&mut self, endianness: Endianness) -> Result<u32, ParseError> {
    let bytes = self.pop_arr::<4>()?;
    Ok(endianness.read_u32(bytes))
  }

  fn pop_u64(&mut self, endianness: Endianness) -> Result<u64, ParseError> {
    let bytes = self.pop_arr::<8>()?;
    Ok(endianness.read_u64(bytes))
  }

  fn pop_word(&mut self, endianness: Endianness, width: Width) -> Result<u64, ParseError> {
    match width {
      Width::Bits32 => self.pop_u32(endianness).map(u64::from),
      Width::Bits64 => self.pop_u64(endianness),
    }
  }

  fn pop_word_usize(&mut self, endianness: Endianness, width: Width) -> Result<usize, ParseError> {
    self.pop_word(endianness, width)
      .and_then(|x| usize::try_from(x).map_err(|_| ParseError::OffsetOutOfBounds))
  }

  fn hexdump(&self, n: usize) {
    print!("{: <8}", "");
    for i in 0..16 {
      print!("  {:x}", i);
    }
    println!();
    for (i, chunk) in self.bytes[.. n.min(self.bytes.len())].chunks(16).enumerate() {
      print!("{:08x}", i * 16);
      for b in chunk {
        print!(" {:02x}", b);
      }
      print!("{: <1$}", "", 4 + 3 * (16 - chunk.len()));
      for b in chunk {
        let c = match char::from(*b) {
          c if c.is_ascii_graphic() => c,
          _ => '.',
        };
        print!("{}", c);
      }
      println!();
    }
  }
}

struct Elf<'a> {
  elf_bytes: &'a [u8],
  bl: ByteLayout,
  header_version: u8,
  abi: u8,
  ty: ElfType,
  instruction_set: Option<InstructionSet>,
  elf_version: u32,
}

impl<'a> Elf<'a> {
  fn parse(bytes: &'a [u8]) -> Result<Self, ElfError> {
    let mut hdr = InputBytes::new(bytes);
    
    hdr.hexdump(64);

    let magic = hdr.pop(4)?;
    if magic != &[0x7f, b'E', b'L', b'F'] {
      return Err(ElfError::BadMagic);
    }
  
    let width = match hdr.pop_u8()? {
      1 => Width::Bits32,
      2 => Width::Bits64,
      _ => return Err(ElfError::UnsupportedValue),
    };
  
    let endianness = match hdr.pop_u8()? {
      1 => Endianness::Little,
      2 => Endianness::Big,
      _ => return Err(ElfError::UnsupportedValue),
    };

    let bl = ByteLayout {
      endianness,
      width,
    };
  
    let header_version = hdr.pop_u8()?;

    let abi = hdr.pop_u8()?;

    hdr.pop(8)?;

    let ty = match hdr.pop_u16(endianness)? {
      1 => ElfType::Relocatable,
      2 => ElfType::Executable,
      3 => ElfType::Shared,
      4 => ElfType::Core,
      _ => return Err(ElfError::UnsupportedValue),
    };
    
    let instruction_set = match hdr.pop_u16(endianness)? {
      0 => None,
      0x02 => Some(InstructionSet::Sparc),
      0x03 => Some(InstructionSet::I386),
      0x08 => Some(InstructionSet::Mips),
      0x14 => Some(InstructionSet::PowerPc),
      0x28 => Some(InstructionSet::Arm),
      0x2a => Some(InstructionSet::SuperH),
      0x32 => Some(InstructionSet::Ia64),
      0x3e => Some(InstructionSet::Amd64),
      0xb7 => Some(InstructionSet::Aarch64),
      0xf3 => Some(InstructionSet::RiscV),
      _ => return Err(ElfError::UnsupportedValue),
    };

    let elf_version = hdr.pop_u32(endianness)?;
    let entry_offset = hdr.pop_word(endianness, width)?;
    let pht_offset = hdr.pop_word_usize(endianness, width)?;
    let sht_offset = hdr.pop_word_usize(endianness, width)?;
    let flags = hdr.pop_u32(endianness)?;
    let header_size = hdr.pop_u16(endianness)?;
    let pht_entry_size = hdr.pop_u16(endianness)?;
    let pht_len = hdr.pop_u16(endianness)?;
    let sht_entry_size = hdr.pop_u16(endianness)?;
    let sht_len = hdr.pop_u16(endianness)?;
    let shstr_idx = hdr.pop_u16(endianness)?;

    dbg!(entry_offset);
    dbg!(pht_offset);
    dbg!(sht_offset);
    dbg!(flags);
    dbg!(header_size);
    dbg!(pht_entry_size);
    dbg!(pht_len);
    dbg!(sht_entry_size);
    dbg!(sht_len);
    dbg!(shstr_idx);

    // TODO

    let shstr = SectionHeader::parse(bl,
      InputBytes::new_offset(bytes, sht_offset + usize::from(sht_entry_size * shstr_idx)))?;

    let str_table = InputBytes::new_offset(bytes, shstr.offset);
    str_table.hexdump(shstr.size);

    Ok(Self {
      elf_bytes: bytes,
      bl,
      header_version,
      abi,
      ty,
      instruction_set,
      elf_version,
    })
  }
}

struct SectionHeader {
  name: u32,
  // FIXME: parse type
  ty: u32,
  flags: u64,
  addr: u64,
  offset: usize,
  size: usize,
  link: u32,
  info: u32,
  align: u64,
  ent_size: usize,
}

impl SectionHeader {
  fn parse(bl: ByteLayout, mut bytes: InputBytes) -> Result<Self, ElfError> {
    let name = bytes.pop_u32(bl.endianness)?;
    let ty = bytes.pop_u32(bl.endianness)?;
    let flags = bytes.pop_word(bl.endianness, bl.width)?;
    let addr = bytes.pop_word(bl.endianness, bl.width)?;
    let offset = bytes.pop_word_usize(bl.endianness, bl.width)?;
    let size = bytes.pop_word_usize(bl.endianness, bl.width)?;
    let link = bytes.pop_u32(bl.endianness)?;
    let info = bytes.pop_u32(bl.endianness)?;
    let align = bytes.pop_word(bl.endianness, bl.width)?;
    let ent_size = bytes.pop_word_usize(bl.endianness, bl.width)?;

    Ok(Self {
      name,
      ty,
      flags,
      addr,
      offset,
      size,
      link,
      info,
      align,
      ent_size,
    })
  }
}

#[derive(Clone, Copy, PartialEq, Eq, Debug)]
struct ByteLayout {
  endianness: Endianness,
  width: Width,
}

#[derive(Clone, Copy, PartialEq, Eq, Debug)]
enum Width {
  Bits32,
  Bits64,
}

#[derive(Clone, Copy, PartialEq, Eq, Debug)]
enum Endianness {
  Little,
  Big,
}

impl Endianness {
  fn read_u16(self, bytes: [u8; 2]) -> u16 {
    match self {
      Self::Little => u16::from_le_bytes(bytes),
      Self::Big => u16::from_be_bytes(bytes),
    }
  }

  fn read_u32(self, bytes: [u8; 4]) -> u32 {
    match self {
      Self::Little => u32::from_le_bytes(bytes),
      Self::Big => u32::from_be_bytes(bytes),
    }
  }

  fn read_u64(self, bytes: [u8; 8]) -> u64 {
    match self {
      Self::Little => u64::from_le_bytes(bytes),
      Self::Big => u64::from_be_bytes(bytes),
    }
  }
}

#[derive(Clone, Copy, PartialEq, Eq, Debug)]
enum ElfType {
  Relocatable,
  Executable,
  Shared,
  Core,
}

#[derive(Clone, Copy, PartialEq, Eq, Debug)]
enum InstructionSet {
  Sparc,
  I386,
  Mips,
  PowerPc,
  Arm,
  SuperH,
  Ia64,
  Amd64,
  Aarch64,
  RiscV,
}

#[derive(Debug)]
enum ElfError {
  BadMagic,
  UnexpectedEof,
  OffsetOutOfBounds,
  UnsupportedValue,
}

impl From<ParseError> for ElfError {
  fn from(value: ParseError) -> Self {
    match value {
      ParseError::UnexpectedEof => Self::UnexpectedEof,
      ParseError::OffsetOutOfBounds => Self::OffsetOutOfBounds,
    }
  }
}

#[derive(Debug)]
enum ParseError {
  UnexpectedEof,
  OffsetOutOfBounds,
}