burritos/src/simulator/loader.rs

use crate::Machine;
use std::fs;
use std::io::Read;

/// load a 32-bits binary file into the machine
/// 
/// ### Parameters
/// 
/// - **path** path of the file to load
/// - **machine** the machine where the bin file will be loaded
/// - **start_index** at which index of machine memory you want to start to load the program
/// 
/// Returns in a Result any io error
pub fn load(path: &str, machine: &mut Machine, start_index: usize) -> Result<(), std::io::Error> {
    let mut file = fs::File::open(path)?;
    let mut instructions: Vec<u32> = Default::default();
    loop {
        let mut buf: [u8; 4] = [0; 4];
        let res = file.read(&mut buf)?;
        if res == 0 {
            break; // eof
        } else {
            instructions.push(u32::from_le_bytes(buf));
        }
    }
    for (i, inst)  in instructions.iter().enumerate() {
        machine.write_memory(4, 4 * i + start_index, inst.to_owned() as u64);
    }
    // #[cfg(debug_assertions)]
    // println!("{:04x?}", instructions); // only print loaded program in debug build
    Ok(())
}

/// The elf header defines principes aspects of the binary files, it's place at the start of the file
/// see <https://en.wikipedia.org/wiki/Executable_and_Linkable_Format#File_header> for more informations
pub struct ElfHeader {
    /// Defines whether the file is big or little endian
    /// true correspond to big endian, false otherwise
    /// 
    /// Offset: 0x05, size: 1 byte
    pub endianess: bool,
    /// Defines whether the file is 32 bits or 64 bits
    /// 
    /// Offset: 0x04, size: 1 byte
    pub is_32bits: bool,
    /// Version of the elf file, current version is 1
    /// 
    /// Offset: 0x06, size: 1 byte
    pub version: u8,
    /// Identifies the target ABI.
    /// 
    /// In this implementation: Defines if the target abi is system V compliant
    /// 
    /// Offset: 0x07, size: 1 byte
    pub sys_v_abi: bool,
    /// Identifies target ISA, 0xF3 correspond to RISC-V
    /// 
    /// In this implementatio, true if target isa is RISC-V, false otherwise
    /// 
    /// Offset: 0x12, size: 2 bytes
    pub is_riscv_target: bool,
    /// Memory address of the entry point from w<here the process starts its execution.
    /// If the program doesn't have an entrypoint (i.e. not an executable), the value is 0
    /// 
    /// Offset: 0x18, size: 4 (32 bits) or 8 (64 bits)
    pub entrypoint: u64,
    /// Size of the elf header, 64 bytes for 64 bits and 52 for 32 bits
    /// 
    /// Offset: 0x28(32 bits) or 0x34 (64 bits), size: 2 bytes
    pub elf_header_size: u16,
    /// Position of the first program header entry
    /// 
    /// Offset: 0x1C (32 bits) or 0x20 (64 bits), size: 4 (32 bits) or 8 (64 bits) bytes
    pub program_header_location: u64,
    /// Number of entries in the progream header table
    /// 
    /// Offset: 0x2C (32 bits) or 0x38 (64 bits), size: 2 bytes
    pub program_header_entries: u16,
    /// Size of a program header entry
    /// 
    /// Offset: 0x2A (32 bits) or 0x36 (64 bits), size: 2 bytes
    pub program_header_size: u16,
    /// Position of the first section header entry
    /// 
    /// Offset: 0x20 (32 bits) or 0x28 (64 bits), size: 4 (32 bits) or 8 (64 bits) bytes
    pub section_header_location: u64,
    /// Number of entries in the section header table
    /// 
    /// Offset: 0x30 (32 bits) or 0x3C (64 bits), size: 2 bytes
    pub section_header_entries: u16,
    /// Size of a section header entry
    /// 
    /// Offset: 0x2E (32 bits) or 0x36 (64 bits), size: 2 bytes
    pub section_header_size: u16,
}

impl ElfHeader {
    
    /// return true if the 4 first bytes constitude the elf magic number
    fn is_elf(instructions: &[u8]) -> bool {
        instructions.get(0..4) == Option::Some(&[0x7f, 0x45, 0x4c, 0x46])
    }

    /// return true if big endian, false otherwise
    fn check_endianess(instructions: &[u8]) -> bool {
        instructions.get(5) == Option::Some(&2)
    }

    /// return true if file is 32 bits, false if 64 bits
    fn is_32bits(instructions: &[u8]) -> bool {
        instructions.get(4) == Option::Some(&1)
    }

    /// return the version of the elf file (should be 1)
    /// Can be None if the file is smaller than 7 bytes -> the file is invalid
    fn get_version(instructions: &[u8]) -> Option<u8> {
        instructions.get(6).copied() // work as primitives implements Copy
    }

    /// return true if target abi of the binary file is System V, false otherwise
    fn is_system_v_elf(instructions: &[u8]) -> bool {
        instructions.get(7) == Option::Some(&0)
    }

    /// return true if specified target instruction set architecture is RISCV
    fn is_riscv_isa(instructions: &[u8]) -> bool {
        Self::get_u16_value(instructions, 0x12) == Option::Some(0xf3)
    }

    /// memory address of the entry point from where the process starts its execution
    /// 
    /// ## Arguments:
    /// 
    /// **instructions** List of bytes of the loaded binary file
    /// **is_32bits** defines whether the binary file is 32 bits or 64 bits
    fn get_entrypoint(instructions: &[u8], is_32bits: bool) -> Option<u64> {
        if is_32bits {
            get_address_point(instructions, 0x18, true)
        } else {
            get_address_point(instructions, 0x18, false)
        }
    }

    /// Memory address of the start of the program header table
    /// 
    /// ## Arguments:
    /// 
    /// **instructions** List of bytes of the loaded binary file
    /// **is_32bits** defines whether the binary file is 32 bits or 64 bits
    fn get_program_header_table_location(instructions: &[u8], is_32bits: bool) -> Option<u64> {
        if is_32bits {
            get_address_point(instructions, 0x1c, true)
        } else {
            get_address_point(instructions, 0x20, false)
        }
    }

    /// Memory address of the start of the section header table
    ///
    ///  ## Arguments:
    /// 
    /// **instructions** List of bytes of the loaded binary file
    /// **is_32bits** defines whether the binary file is 32 bits or 64 bits
    fn get_section_header_table_location(instructions: &[u8], is_32bits: bool) -> Option<u64> {
        if is_32bits {
            get_address_point(instructions, 0x20, true)
        } else {
            get_address_point(instructions, 0x28, false)
        }
    }

    /// Return the size of the header, normally, 0x40 for 64 bits bin and 0x34 for 32 bits
    /// 
    /// ## Arguments:
    /// 
    /// **instructions** List of bytes of the loaded binary file
    /// **is_32bits** defines whether the binary file is 32 bits or 64 bits
    fn get_elf_header_size(instructions: &[u8], is_32bits: bool) -> Option<u16> {
        let address = if is_32bits { 0x28 } else { 0x34 };
        Self::get_u16_value(instructions, address)
    }

    /// return the size of a program header table entry
    /// 
    /// ## Arguments:
    /// 
    /// **instructions** List of bytes of the loaded binary file
    /// **is_32bits** defines whether the binary file is 32 bits or 64 bits
    fn get_program_header_size(instructions: &[u8], is_32bits: bool) -> Option<u16> {
        let address = if is_32bits { 0x2a } else { 0x36 };
        Self::get_u16_value(instructions, address)
    }

    /// return the number of entries in the program header
    /// 
    /// ## Arguments:
    /// 
    /// **instructions** List of bytes of the loaded binary file
    /// **is_32bits** defines whether the binary file is 32 bits or 64 bits
    fn get_number_entries_program_header(instructions: &[u8], is_32bits: bool) -> Option<u16> {
        let address = if is_32bits { 0x2c } else { 0x38 };
        Self::get_u16_value(instructions, address)
    }

    /// Return the size of a section header table entry
    /// 
    /// ## Arguments:
    /// 
    /// **instructions** List of bytes of the loaded binary file
    /// **is_32bits** defines whether the binary file is 32 bits or 64 bits
    fn get_section_header_size(instructions: &[u8], is_32bits: bool) -> Option<u16> {
        let address = if is_32bits { 0x2e } else { 0x3a };
        Self::get_u16_value(instructions, address)
    }

    /// Return the number of entries in the section header
    /// 
    /// ## Arguments:
    /// 
    /// **instructions** List of bytes of the loaded binary file
    /// **is_32bits** defines whether the binary file is 32 bits or 64 bits
    fn get_section_header_num_entries(instructions: &[u8], is_32bits: bool) -> Option<u16> {
        let address = if is_32bits { 0x30 } else { 0x3c };
        Self::get_u16_value(instructions, address)
    }

    /// Return a u16 value, usually for the size or the number of entries inside a header
    /// 
    /// This method retrieve 2 bytes and concatenate them assuming the file is little endian
    /// 
    /// ## Arguments:
    /// 
    /// **instructions** List of bytes of the loaded binary file
    /// **address** Position of the first byte
    fn get_u16_value(instructions: &[u8], address: usize) -> Option<u16> {
        let mut bytes: [u8; 2] = [0; 2];
        bytes[0] = instructions.get(address).copied()?;
        bytes[1] = instructions.get(address + 1).copied()?;
        Option::Some(u16::from_le_bytes(bytes))
    }
    
}

impl TryFrom<&Vec<u8>> for ElfHeader {
    type Error = ();

    fn try_from(instructions: &Vec<u8>) -> Result<Self, Self::Error> {
        if Self::is_elf(instructions) {
            let format = Self::is_32bits(instructions);
            let endianess = Self::check_endianess(instructions);
            let version = Self::get_version(instructions).ok_or(())?;
            let is_sys_v_abi = Self::is_system_v_elf(instructions);
            let is_rv_target = Self::is_riscv_isa(instructions);
            let entrypoint = Self::get_entrypoint(instructions, format).ok_or(())?;
            let elf_header_size = Self::get_elf_header_size(instructions, format).ok_or(())?;
            let program_header_location = Self::get_program_header_table_location(instructions, format).ok_or(())?;
            let program_header_entries = Self::get_number_entries_program_header(instructions, format).ok_or(())? ;
            let program_header_size = Self::get_program_header_size(instructions, format).ok_or(())?;
            let section_header_location = Self::get_section_header_table_location(instructions, format).ok_or(())?;
            let section_header_entries = Self::get_section_header_num_entries(instructions, format).ok_or(())?;
            let section_header_size = Self::get_section_header_size(instructions, format).ok_or(())?;
            Ok(ElfHeader { 
                endianess, 
                is_32bits: format,
                version,
                sys_v_abi: is_sys_v_abi,
                is_riscv_target: is_rv_target,
                entrypoint,
                elf_header_size,
                program_header_location,
                program_header_entries,
                program_header_size,
                section_header_location,
                section_header_entries,
                section_header_size
            })
        } else {
            Err(())
        }
    }
}


/// Flag of a section, a section can have multiples flags by adding the values
#[allow(clippy::enum_variant_names)]
#[allow(dead_code)]
pub enum FlagValue {
    /// The section is writable
    ShfWrite = 0x1,
    /// The section need to be allocate/occupe memory during execution
    ShfAlloc = 0x2,
    /// The section need to be executable
    ShfExecinstr = 0x4,
    /// Section might ber merged
    ShfMerge = 0x10,
    /// Contain null-terminated (\0) strings
    ShfStrings = 0x20,
    // There is others but are unrelevant (I think)
}

/// Section header entry, contains useful informations for each sections of the binary file
/// 
/// see <https://en.wikipedia.org/wiki/Executable_and_Linkable_Format#Section_header>
#[derive(Debug)]
pub struct SectionHeader {
    /// Offset to a string in .shstrtab section that represent the name of this section
    /// 
    /// Offset: 0x0, size: 4 bytes
    pub name_offset: u32,
    /// Identify the type of this header
    /// 
    /// Offset: 0x4, size: 4 bytes
    pub header_type: u32,
    /// Identify the atributes of this section
    /// 
    /// see `Self::does_flag_contains_key(self, FlagValue)`
    /// 
    /// Offset: 0x8, size: 4 (32 bits) or 8 (64 bits) bytes
    pub flags: u64,
    /// Virtual address of the section in memory if section is loaded, 0x0 otherwise
    /// 
    /// Offset: 0x0C (32 bits) or 0x10 (64 bits), size: 4 (32 bits) or 8 (64 bits) bytes
    pub virt_addr: u64,
    /// Offset of the section in the file image (binary file)
    /// 
    /// Offset: 0x10 (32 bits) or 0x18 (64 bits), size: 4 (32 bits) or 8 (64 bits) bytes
    pub image_offset: u64,
    /// Size of the section in the file image, may be 0
    /// 
    /// Offset: 0x14 (32 bits) or 0x20 (64 bits), size: 4 (32 bits) or 8 (64 bits) bytes
    pub section_size: u64,
    pub section_link: u32,
    pub section_info: u32,
    /// Contain the required alignment of the section, must be a power of 2
    /// 
    /// Offset: 0x20 (32 bits) or 0x30 (64 bits), size: 4 (32 bits) or 8 (64 bits) bytes
    pub required_align: u64,
    /// Contain the size of each entry, for sections that contain fixed size entries, otherwise 0
    /// 
    /// Offset: 0x24 (32 bits) or 0x38 (64 bits), size: 4 (32 bits) or 8 (64 bits) bytes
    pub entry_size: u64
}

impl SectionHeader {

    /// return true if flag of this section contains / have `key`, false otherwise
    pub fn does_flag_contains_key(&self, key: FlagValue) -> bool {
        self.flags & key as u64 != 0
    }

    /// Return the offset to a string in .shstrtab that represents the name of this section
    fn get_name_offset(instructions: &[u8], address: usize) -> Option<u32> {
        get_address_point(instructions, address, true).map(|v| { v as u32 })
                                                        // set true to return a u32
    }

    /// Return the type of header of the section
    fn get_header_type(instructions: &[u8], address: usize) -> Option<u32> {
        get_address_point(instructions, address + 0x4, true).map(|v| { v as u32 })
    }

    /// Return the flags of the section, can hold multiples values, see [`FlagValue`]
    fn get_flags(instructions: &[u8], address: usize, is_32bits: bool) -> Option<u64> {
        get_address_point(instructions, address + 0x8, is_32bits)
    }


    /// Return the virtual address of the section in memory if the sectino is loaded(see section flag), otherwise 0
    fn get_virtual_address(instructions: &[u8], address: usize, is_32bits: bool) -> Option<u64> {
        get_address_point(instructions, address + if is_32bits { 0x0C } else { 0x10 }, is_32bits)
    }

    fn get_image_offset(instructions: &Vec<u8>, address: usize, is_32bits: bool) -> Option<u64> {
        get_address_point(instructions, address + if is_32bits { 0x10 } else { 0x18 }, is_32bits)
    }

    fn get_section_size(instructions: &Vec<u8>, address: usize, is_32bits: bool) -> Option<u64> {
        get_address_point(instructions, address + if is_32bits { 0x14 } else { 0x20 }, is_32bits)
    }

    fn get_section_link(instructions: &Vec<u8>, address: usize, is_32bits: bool) -> Option<u32> {
        get_address_point(instructions, address + if is_32bits { 0x18 } else { 0x28 }, false).map(|v| { v as u32 })
    }

    fn get_section_info(instructions: &Vec<u8>, address: usize, is_32bits: bool) -> Option<u32> {
        get_address_point(instructions, address + if is_32bits { 0x1C } else { 0x2C }, false).map(|v| { v as u32 })
    }

    fn get_required_align(instructions: &Vec<u8>, address: usize, is_32bits: bool) -> Option<u64> {
        get_address_point(instructions, address + if is_32bits { 0x20 } else { 0x30 }, is_32bits)
    }

    fn get_entry_size(instructions: &Vec<u8>, address: usize, is_32bits: bool) -> Option<u64> {
        get_address_point(instructions, address + if is_32bits { 0x24 } else { 0x38 }, is_32bits)
    }

}

impl TryFrom<(&Vec<u8>, u64, bool)> for SectionHeader {
    type Error = ();

    fn try_from(value: (&Vec<u8>, u64, bool)) -> Result<Self, Self::Error> {
        let instructions = value.0;
        let address = value.1 as usize;
        let is_32bits = value.2;

        let name_offset = Self::get_name_offset(instructions, address).ok_or(())?;
        let header_type = Self::get_header_type(instructions, address).ok_or(())?;
        let attribute = Self::get_flags(instructions, address, is_32bits).ok_or(())?;
        let virt_addr = Self::get_virtual_address(instructions, address, is_32bits).ok_or(())?;
        let image_offset = Self::get_image_offset(instructions, address, is_32bits).ok_or(())?;
        let section_size = Self::get_section_size(instructions, address, is_32bits).ok_or(())?;
        let section_link = Self::get_section_link(instructions, address, is_32bits).ok_or(())?;
        let section_info = Self::get_section_info(instructions, address, is_32bits).ok_or(())?;
        let required_align = Self::get_required_align(instructions, address, is_32bits).ok_or(())?;
        let entry_size = Self::get_entry_size(instructions, address, is_32bits).ok_or(())?;
        Ok(Self { name_offset, 
            header_type, 
            flags: attribute, 
            virt_addr, 
            image_offset,
            section_size, 
            section_link,
            section_info,
            required_align, 
            entry_size  
        })
    }
}

pub struct Loader {
    bytes: Vec<u8>,
    pub elf_header: ElfHeader,
    pub sections: Vec<SectionHeader>
}

#[derive(Debug)]
pub enum LoaderError {
    IOError(std::io::Error),
    ParsingError
}

impl Loader {

    pub fn new(path: &str, machine: &mut Machine, start_index: usize) -> Result<(Self, u64), LoaderError> {
        let loader = Self::load_and_parse(path)?;
        let end_alloc = loader.load_into_machine(machine, start_index)?;
        Ok((loader, end_alloc))
    }

    fn load_into_machine(&self, machine: &mut Machine, start_index: usize) -> Result<u64, LoaderError> {
        let mut end_index = 0;
        for i in 0..self.sections.len() {
            let section = &self.sections[i];
            if section.does_flag_contains_key(FlagValue::ShfAlloc) {
                end_index = section.virt_addr + section.section_size;
                // Can allocate to machine memory
                for j in (0..section.section_size as usize).step_by(4) {
                    let mut buf: [u8; 4] = [0; 4];
                    for k in 0..4 {
                        buf[k] = self.bytes.get(section.image_offset as usize + j + k).copied().ok_or(LoaderError::ParsingError)?;
                    }
                    machine.write_memory(4, start_index + section.virt_addr as usize + j, u32::from_le_bytes(buf) as u64);
                }
            }
        }
        Ok(start_index as u64 + end_index)
    }

    fn load_and_parse(path: &str) -> Result<Self, LoaderError> {
        let file = fs::File::open(path);
        match file {
            Ok(mut file) => {
                let mut instructions: Vec<u8> = Default::default();
                loop {
                    let mut buf: [u8; 1] = [0; 1];
                    let res = file.read(&mut buf);
                    match res {
                        Ok(res) => {
                            if res == 0 {
                                break; // eof
                            } else {
                                instructions.push(buf[0]);
                            }
                        },
                        Err(err) => {
                            return Err(LoaderError::IOError(err))
                        }
                    }
                    
                }
                let elf_header = match ElfHeader::try_from(&instructions) {
                    Ok(header) => {
                        header
                    },
                    Err(_) => {
                        return Err(LoaderError::ParsingError);
                    }
                };
                let section_header = match Self::parse_section_header(&instructions, elf_header.is_32bits, elf_header.section_header_location, elf_header.section_header_entries, elf_header.section_header_size) {
                    Ok(header) => {
                        header
                    },
                    Err(_) => {
                        return Err(LoaderError::ParsingError);
                    }
                };
                // #[cfg(debug_assertions)]
                // println!("{:04x?}", instructions); // only print loaded program in debug build
                return Ok(Self { bytes: instructions, elf_header, sections: section_header });
            },
            Err(err) => {
                return Err(LoaderError::IOError(err));
            }
        };
    }

    fn parse_section_header(instructions: &Vec<u8>, is_32bits: bool, header_location: u64, num_of_entries: u16, entry_size: u16) -> Result<Vec<SectionHeader>, ()> {
        let mut sections: Vec<SectionHeader> = Default::default();
        for i in 0..num_of_entries as u64 {
            sections.push(Self::parse_section_entry(instructions, is_32bits, header_location + i * entry_size as u64)?);
        }
        Ok(sections)
    }

    fn parse_section_entry(instructions: &Vec<u8>, is_32bits: bool, location: u64) -> Result<SectionHeader, ()> {
        SectionHeader::try_from((instructions, location, is_32bits))
    }


}

/// return the memory address of something stored at address
/// Can return None if the file is smaller than adress + 3 (or 7 if 64 bits), in this case, the elf header is incorrect
fn get_address_point(instructions: &[u8], address: usize, is_32bits: bool) -> Option<u64> {
    if is_32bits {
        let mut bytes: [u8; 4] = [0; 4];
        bytes[0] = instructions.get(address).copied()?;
        bytes[1] = instructions.get(address + 1).copied()?;
        bytes[2] = instructions.get(address + 2).copied()?;
        bytes[3] = instructions.get(address + 3).copied()?;
        Option::Some(u32::from_le_bytes(bytes) as u64)
    } else {
        let mut bytes: [u8; 8] = [0; 8];
        bytes[0] = instructions.get(address).copied()?;
        bytes[1] = instructions.get(address + 1).copied()?;
        bytes[2] = instructions.get(address + 2).copied()?;
        bytes[3] = instructions.get(address + 3).copied()?;
        bytes[4] = instructions.get(address + 4).copied()?;
        bytes[5] = instructions.get(address + 5).copied()?;
        bytes[6] = instructions.get(address + 6).copied()?;
        bytes[7] = instructions.get(address + 7).copied()?;
        Option::Some(u64::from_le_bytes(bytes))
    }
}

#[cfg(test)]
mod test {
    use crate::simulator::{loader::Loader, machine::Machine};


    #[test]
    #[ignore = "CI gitlab a modifié"]
    fn test_parse_elf() {
        let mut machine = Machine::init_machine();
        let loader = Loader::load_and_parse("./test/riscv_instructions/simple_arithmetics/unsigned_addition").expect("IO Error");
        loader.load_into_machine(&mut machine, 0).expect("Parsing error");
        assert!(!loader.elf_header.is_32bits);
        assert!(!loader.elf_header.endianess);
        assert!(loader.elf_header.sys_v_abi);
        assert!(loader.elf_header.is_riscv_target);
        assert_eq!(1, loader.elf_header.version);
        assert_eq!(0x4000, loader.elf_header.entrypoint);
        assert_eq!(64, loader.elf_header.elf_header_size);
        assert_eq!(64, loader.elf_header.program_header_location);
        assert_eq!(18984, loader.elf_header.section_header_location);
        assert_eq!(56, loader.elf_header.program_header_size);
        assert_eq!(64, loader.elf_header.section_header_size);
        assert_eq!(4, loader.elf_header.program_header_entries);
        assert_eq!(9, loader.elf_header.section_header_entries);
        println!("{:#x?}", loader.sections);
    }

}