-
Notifications
You must be signed in to change notification settings - Fork 258
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
eBPF assembler #10
eBPF assembler #10
Changes from 1 commit
23ff50d
4f35aae
0b8e232
0e73d32
5a0f2ef
a6b0d4b
3a169a0
f1ff025
File filter
Filter by extension
Conversations
Jump to
Diff view
Diff view
There are no files selected for viewing
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,225 @@ | ||
// Copyright 2017 Rich Lane <[email protected]> | ||
// | ||
// Licensed under the Apache License, Version 2.0 <http://www.apache.org/licenses/LICENSE-2.0> or | ||
// the MIT license <http://opensource.org/licenses/MIT>, at your option. This file may not be | ||
// copied, modified, or distributed except according to those terms. | ||
|
||
|
||
//! This module translates eBPF assembly language to binary. | ||
|
||
use asm_parser::{Instruction, Operand, parse}; | ||
use ebpf; | ||
use ebpf::Insn; | ||
use std::collections::HashMap; | ||
use self::InstructionType::{AluBinary, AluUnary, LoadImm, LoadReg, StoreImm, StoreReg, | ||
JumpUnconditional, JumpConditional, Call, Endian, NoOperand}; | ||
use asm_parser::Operand::{Integer, Memory, Register, Nil}; | ||
|
||
#[derive(Clone, Copy, Debug, PartialEq)] | ||
enum InstructionType { | ||
AluBinary, | ||
AluUnary, | ||
LoadImm, | ||
LoadReg, | ||
StoreImm, | ||
StoreReg, | ||
JumpUnconditional, | ||
JumpConditional, | ||
Call, | ||
Endian(i64), | ||
NoOperand, | ||
} | ||
|
||
fn make_instruction_map() -> HashMap<String, (InstructionType, u8)> { | ||
let mut result = HashMap::new(); | ||
|
||
let alu_binary_ops = [("add", ebpf::BPF_ADD), | ||
("sub", ebpf::BPF_SUB), | ||
("mul", ebpf::BPF_MUL), | ||
("div", ebpf::BPF_DIV), | ||
("or", ebpf::BPF_OR), | ||
("and", ebpf::BPF_AND), | ||
("lsh", ebpf::BPF_LSH), | ||
("rsh", ebpf::BPF_RSH), | ||
("mod", ebpf::BPF_MOD), | ||
("xor", ebpf::BPF_XOR), | ||
("mov", ebpf::BPF_MOV), | ||
("arsh", ebpf::BPF_ARSH)]; | ||
|
||
let mem_sizes = | ||
[("w", ebpf::BPF_W), ("h", ebpf::BPF_H), ("b", ebpf::BPF_B), ("dw", ebpf::BPF_DW)]; | ||
|
||
let jump_conditions = [("jeq", ebpf::BPF_JEQ), | ||
("jgt", ebpf::BPF_JGT), | ||
("jge", ebpf::BPF_JGE), | ||
("jset", ebpf::BPF_JSET), | ||
("jne", ebpf::BPF_JNE), | ||
("jsgt", ebpf::BPF_JSGT), | ||
("jsge", ebpf::BPF_JSGE)]; | ||
|
||
{ | ||
let mut entry = |name: &str, inst_type: InstructionType, opc: u8| { | ||
result.insert(name.to_string(), (inst_type, opc)) | ||
}; | ||
|
||
// Miscellaneous. | ||
entry("exit", NoOperand, ebpf::EXIT); | ||
entry("ja", JumpUnconditional, ebpf::JA); | ||
entry("call", Call, ebpf::CALL); | ||
entry("lddw", LoadImm, ebpf::LD_DW_IMM); | ||
|
||
// AluUnary. | ||
entry("neg", AluUnary, ebpf::NEG64); | ||
entry("neg32", AluUnary, ebpf::NEG32); | ||
entry("neg64", AluUnary, ebpf::NEG64); | ||
|
||
// AluBinary. | ||
for &(name, opc) in &alu_binary_ops { | ||
entry(name, AluBinary, ebpf::BPF_ALU64 | opc); | ||
entry(&format!("{}32", name), AluBinary, ebpf::BPF_ALU | opc); | ||
entry(&format!("{}64", name), AluBinary, ebpf::BPF_ALU64 | opc); | ||
} | ||
|
||
// Load, StoreImm, and StoreReg. | ||
for &(suffix, size) in &mem_sizes { | ||
entry(&format!("ldx{}", suffix), | ||
LoadReg, | ||
ebpf::BPF_MEM | ebpf::BPF_LDX | size); | ||
entry(&format!("st{}", suffix), | ||
StoreImm, | ||
ebpf::BPF_MEM | ebpf::BPF_ST | size); | ||
entry(&format!("stx{}", suffix), | ||
StoreReg, | ||
ebpf::BPF_MEM | ebpf::BPF_STX | size); | ||
} | ||
|
||
// JumpConditional. | ||
for &(name, condition) in &jump_conditions { | ||
entry(name, JumpConditional, ebpf::BPF_JMP | condition); | ||
} | ||
|
||
// Endian. | ||
for &size in &[16, 32, 64] { | ||
entry(&format!("be{}", size), Endian(size), ebpf::BE); | ||
entry(&format!("le{}", size), Endian(size), ebpf::LE); | ||
} | ||
} | ||
|
||
result | ||
} | ||
|
||
fn insn(opc: u8, dst: i64, src: i64, off: i64, imm: i64) -> Result<Insn, String> { | ||
if dst < 0 || dst >= 16 { | ||
return Err(format!("Invalid destination register {}", dst)); | ||
} | ||
if dst < 0 || src >= 16 { | ||
return Err(format!("Invalid source register {}", src)); | ||
} | ||
if off < -32768 || off >= 32768 { | ||
return Err(format!("Invalid offset {}", off)); | ||
} | ||
if imm < -2147483648 || imm >= 4294967296 { | ||
return Err(format!("Invalid immediate {}", imm)); | ||
} | ||
Ok(Insn { | ||
opc: opc, | ||
dst: dst as u8, | ||
src: src as u8, | ||
off: off as i16, | ||
imm: imm as i32, | ||
}) | ||
} | ||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. This whole function is pretty close to a simplified version of the program building API @alex-diez has been working on in PR #6. Once we have the API, we could use it here instead to avoid re-defining ways to build instructions. |
||
|
||
// TODO Use slice patterns when available and remove this function. | ||
fn operands_tuple(operands: &[Operand]) -> Result<(Operand, Operand, Operand), String> { | ||
match operands.len() { | ||
0 => Ok((Nil, Nil, Nil)), | ||
1 => Ok((operands[0], Nil, Nil)), | ||
2 => Ok((operands[0], operands[1], Nil)), | ||
3 => Ok((operands[0], operands[1], operands[2])), | ||
_ => Err("Too many operands".to_string()), | ||
} | ||
} | ||
|
||
fn encode(inst_type: InstructionType, opc: u8, operands: &[Operand]) -> Result<Insn, String> { | ||
let (a, b, c) = try!(operands_tuple(operands)); | ||
match (inst_type, a, b, c) { | ||
(AluBinary, Register(dst), Register(src), Nil) => insn(opc | ebpf::BPF_X, dst, src, 0, 0), | ||
(AluBinary, Register(dst), Integer(imm), Nil) => insn(opc | ebpf::BPF_K, dst, 0, 0, imm), | ||
(AluUnary, Register(dst), Nil, Nil) => insn(opc, dst, 0, 0, 0), | ||
(LoadReg, Register(dst), Memory(src, off), Nil) | | ||
(StoreReg, Memory(dst, off), Register(src), Nil) => insn(opc, dst, src, off, 0), | ||
(StoreImm, Memory(dst, off), Integer(imm), Nil) => insn(opc, dst, 0, off, imm), | ||
(NoOperand, Nil, Nil, Nil) => insn(opc, 0, 0, 0, 0), | ||
(JumpUnconditional, Integer(off), Nil, Nil) => insn(opc, 0, 0, off, 0), | ||
(JumpConditional, Register(dst), Register(src), Integer(off)) => { | ||
insn(opc | ebpf::BPF_X, dst, src, off, 0) | ||
} | ||
(JumpConditional, Register(dst), Integer(imm), Integer(off)) => { | ||
insn(opc | ebpf::BPF_K, dst, 0, off, imm) | ||
} | ||
(Call, Integer(imm), Nil, Nil) => insn(opc, 0, 0, 0, imm), | ||
(Endian(size), Register(dst), Nil, Nil) => insn(opc, dst, 0, 0, size), | ||
(LoadImm, Register(dst), Integer(imm), Nil) => insn(opc, dst, 0, 0, (imm << 32) >> 32), | ||
_ => Err(format!("Unexpected operands: {:?}", operands)), | ||
} | ||
} | ||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. Probably our major difference in formatting style: I'd definitely align the patterns on something like this :) I find it more readable, but it's not really a good habit since later modifications of a line can force to realign everything (so I'm not asking you to align, that's just for the record). There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. Yeah, I'd normally align it but I've been using rustfmt which would delete the extra spaces. |
||
|
||
fn assemble_internal(parsed: &[Instruction]) -> Result<Vec<Insn>, String> { | ||
let instruction_map = make_instruction_map(); | ||
let mut result: Vec<Insn> = vec![]; | ||
for instruction in parsed { | ||
let name = instruction.name.as_str(); | ||
match instruction_map.get(name) { | ||
Some(&(inst_type, opc)) => { | ||
match encode(inst_type, opc, &instruction.operands) { | ||
Ok(insn) => result.push(insn), | ||
Err(msg) => return Err(format!("Failed to encode {}: {}", name, msg)), | ||
} | ||
// Special case for lddw. | ||
if let LoadImm = inst_type { | ||
if let Integer(imm) = instruction.operands[1] { | ||
result.push(insn(0, 0, 0, 0, imm >> 32).unwrap()); | ||
} | ||
} | ||
} | ||
None => return Err(format!("Invalid instruction {:?}", name)), | ||
} | ||
} | ||
Ok(result) | ||
} | ||
|
||
/// Parse assembly source and translate to binary. | ||
/// | ||
/// # Examples | ||
/// | ||
/// ``` | ||
/// use rbpf::assembler::assemble; | ||
/// let prog = assemble("add64 r1, 0x605 | ||
/// mov64 r2, 0x32 | ||
/// mov64 r1, r0 | ||
/// be16 r0 | ||
/// neg64 r8 | ||
/// exit"); | ||
/// println!("{:?}", prog); | ||
/// ``` | ||
/// | ||
/// This will produce the following output: | ||
/// | ||
/// ```test | ||
/// Ok([0x07, 0x01, 0x00, 0x00, 0x05, 0x06, 0x00, 0x00, | ||
/// 0xb7, 0x02, 0x00, 0x00, 0x32, 0x00, 0x00, 0x00, | ||
/// 0xbf, 0x01, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00, | ||
/// 0xdc, 0x00, 0x10, 0x00, 0x00, 0x00, 0x00, 0x00, | ||
/// 0x87, 0x08, 0x10, 0x00, 0x00, 0x00, 0x00, 0x00, | ||
/// 0x95, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00]) | ||
/// ``` | ||
pub fn assemble(src: &str) -> Result<Vec<u8>, String> { | ||
let parsed = try!(parse(src)); | ||
let insns = try!(assemble_internal(&parsed)); | ||
let mut result: Vec<u8> = vec![]; | ||
for insn in insns { | ||
result.extend_from_slice(&insn.to_array()); | ||
} | ||
Ok(result) | ||
} |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Shouldn't be 2147483648 instead of 4294967296?
For all these values, we could maybe use instead
std::i16::MIN
(-32768),std::i16::MAX
(32767),std::i32::MIN
(-2147483648) andstd::i32::MAX
(2147483647) (orstd::u32::MAX
if you got the right value, but in this case could you explain, please?)There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I'd been assuming we need to support 32-bit unsigned integer constants. Thinking more about it, constants with the high bit set are going to be sign-extended by most instructions and allowing positive-looking constants that are interpreted by the VM as negative would be confusing. I pushed a commit fixing this.
I initially used std:::i32::MIN etc, but found that they were sign-extended when casting to i64. Rather than doing a cast plus a mask, I thought it would be more readable to write out the value.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
… because we have to parse numbers as
i64
… Ok, I see. Well, fine with me, then. Also, forbidding unsigned immediate sounds like a good idea, and we can still revert this later if it appears that it is really needed. Let's merge all this!