| 1 | //! ULEB128 / SLEB128 codec. |
| 2 | //! |
| 3 | //! dyld uses LEB128 pervasively: the export trie (Sprint 5), function-starts |
| 4 | //! deltas (Sprint 16), rebase/bind/lazy-bind opcode streams (Sprint 15), and |
| 5 | //! chained-fixups imports (Sprint 15.5) all encode variable-width integers |
| 6 | //! this way. One codec, reused across all of them. |
| 7 | |
| 8 | use crate::macho::reader::ReadError; |
| 9 | |
| 10 | /// Read a ULEB128 from `bytes`. On success returns `(value, consumed)`. |
| 11 | /// Errors on overrun beyond the buffer or on encodings that exceed 10 bytes |
| 12 | /// (maximum for a `u64`). |
| 13 | pub fn read_uleb(bytes: &[u8]) -> Result<(u64, usize), ReadError> { |
| 14 | let mut value: u64 = 0; |
| 15 | let mut shift = 0u32; |
| 16 | for (i, &b) in bytes.iter().enumerate() { |
| 17 | // The top bit of each byte is the continuation flag; low 7 bits are |
| 18 | // value bits, concatenated little-end first. |
| 19 | if shift >= 64 { |
| 20 | return Err(ReadError::BadRelocation { |
| 21 | at_offset: 0, |
| 22 | reason: "ULEB128 overflows u64", |
| 23 | }); |
| 24 | } |
| 25 | value |= ((b & 0x7f) as u64) << shift; |
| 26 | if b & 0x80 == 0 { |
| 27 | return Ok((value, i + 1)); |
| 28 | } |
| 29 | shift += 7; |
| 30 | } |
| 31 | Err(ReadError::Truncated { |
| 32 | need: bytes.len() + 1, |
| 33 | have: bytes.len(), |
| 34 | context: "ULEB128 (unterminated)", |
| 35 | }) |
| 36 | } |
| 37 | |
| 38 | /// Read an SLEB128 from `bytes`. On success returns `(value, consumed)`. |
| 39 | pub fn read_sleb(bytes: &[u8]) -> Result<(i64, usize), ReadError> { |
| 40 | let mut value: i64 = 0; |
| 41 | let mut shift = 0u32; |
| 42 | for (i, &b) in bytes.iter().enumerate() { |
| 43 | if shift >= 64 { |
| 44 | return Err(ReadError::BadRelocation { |
| 45 | at_offset: 0, |
| 46 | reason: "SLEB128 overflows i64", |
| 47 | }); |
| 48 | } |
| 49 | value |= ((b & 0x7f) as i64) << shift; |
| 50 | shift += 7; |
| 51 | if b & 0x80 == 0 { |
| 52 | // Sign-extend: if the value's sign bit (bit 6 of the last byte) |
| 53 | // is set and we have spare high bits, fill them with 1s. |
| 54 | if shift < 64 && (b & 0x40) != 0 { |
| 55 | value |= !0i64 << shift; |
| 56 | } |
| 57 | return Ok((value, i + 1)); |
| 58 | } |
| 59 | } |
| 60 | Err(ReadError::Truncated { |
| 61 | need: bytes.len() + 1, |
| 62 | have: bytes.len(), |
| 63 | context: "SLEB128 (unterminated)", |
| 64 | }) |
| 65 | } |
| 66 | |
| 67 | /// Append a ULEB128 encoding of `value` to `out`. |
| 68 | pub fn write_uleb(mut value: u64, out: &mut Vec<u8>) { |
| 69 | loop { |
| 70 | let byte = (value & 0x7f) as u8; |
| 71 | value >>= 7; |
| 72 | if value == 0 { |
| 73 | out.push(byte); |
| 74 | return; |
| 75 | } |
| 76 | out.push(byte | 0x80); |
| 77 | } |
| 78 | } |
| 79 | |
| 80 | /// Append an SLEB128 encoding of `value` to `out`. |
| 81 | pub fn write_sleb(mut value: i64, out: &mut Vec<u8>) { |
| 82 | loop { |
| 83 | let byte = (value & 0x7f) as u8; |
| 84 | // Arithmetic shift preserves the sign. |
| 85 | let next = value >> 7; |
| 86 | let sign_bit = byte & 0x40 != 0; |
| 87 | let done = (next == 0 && !sign_bit) || (next == -1 && sign_bit); |
| 88 | value = next; |
| 89 | if done { |
| 90 | out.push(byte); |
| 91 | return; |
| 92 | } |
| 93 | out.push(byte | 0x80); |
| 94 | } |
| 95 | } |
| 96 | |
| 97 | #[cfg(test)] |
| 98 | mod tests { |
| 99 | use super::*; |
| 100 | |
| 101 | #[test] |
| 102 | fn uleb_small_one_byte() { |
| 103 | let mut buf = Vec::new(); |
| 104 | write_uleb(0, &mut buf); |
| 105 | assert_eq!(buf, vec![0x00]); |
| 106 | write_uleb(0x7f, &mut buf); |
| 107 | assert_eq!(buf[1], 0x7f); |
| 108 | } |
| 109 | |
| 110 | #[test] |
| 111 | fn uleb_round_trips_many_values() { |
| 112 | for v in [ |
| 113 | 0u64, |
| 114 | 1, |
| 115 | 127, |
| 116 | 128, |
| 117 | 129, |
| 118 | 16383, |
| 119 | 16384, |
| 120 | 0xdead_beef, |
| 121 | 0xffff_ffff, |
| 122 | u64::MAX, |
| 123 | ] { |
| 124 | let mut buf = Vec::new(); |
| 125 | write_uleb(v, &mut buf); |
| 126 | let (back, consumed) = read_uleb(&buf).unwrap(); |
| 127 | assert_eq!(back, v, "round-trip failed for {v:#x}"); |
| 128 | assert_eq!(consumed, buf.len()); |
| 129 | } |
| 130 | } |
| 131 | |
| 132 | #[test] |
| 133 | fn sleb_round_trips_many_values() { |
| 134 | for v in [ |
| 135 | 0i64, |
| 136 | 1, |
| 137 | -1, |
| 138 | 63, |
| 139 | -63, |
| 140 | 64, |
| 141 | -64, |
| 142 | 65, |
| 143 | -65, |
| 144 | 127, |
| 145 | -128, |
| 146 | 8192, |
| 147 | -8192, |
| 148 | 0x0001_0000, |
| 149 | -0x0001_0000, |
| 150 | i64::MIN, |
| 151 | i64::MAX, |
| 152 | ] { |
| 153 | let mut buf = Vec::new(); |
| 154 | write_sleb(v, &mut buf); |
| 155 | let (back, consumed) = read_sleb(&buf).unwrap(); |
| 156 | assert_eq!(back, v, "sleb round-trip failed for {v}"); |
| 157 | assert_eq!(consumed, buf.len()); |
| 158 | } |
| 159 | } |
| 160 | |
| 161 | #[test] |
| 162 | fn uleb_known_multibyte_encoding() { |
| 163 | // 624485 = 0b100110000111011100101 — canonical LEB example. |
| 164 | let mut buf = Vec::new(); |
| 165 | write_uleb(624485, &mut buf); |
| 166 | assert_eq!(buf, vec![0xe5, 0x8e, 0x26]); |
| 167 | let (v, n) = read_uleb(&buf).unwrap(); |
| 168 | assert_eq!(v, 624485); |
| 169 | assert_eq!(n, 3); |
| 170 | } |
| 171 | |
| 172 | #[test] |
| 173 | fn sleb_known_multibyte_encoding() { |
| 174 | // -12345 in SLEB: widely quoted example. |
| 175 | let mut buf = Vec::new(); |
| 176 | write_sleb(-12345, &mut buf); |
| 177 | let (v, n) = read_sleb(&buf).unwrap(); |
| 178 | assert_eq!(v, -12345); |
| 179 | assert_eq!(n, buf.len()); |
| 180 | } |
| 181 | |
| 182 | #[test] |
| 183 | fn uleb_unterminated_errors() { |
| 184 | // Continuation bit set on every byte but no terminator. |
| 185 | let buf = vec![0x80, 0x80, 0x80]; |
| 186 | assert!(matches!( |
| 187 | read_uleb(&buf).unwrap_err(), |
| 188 | ReadError::Truncated { .. } |
| 189 | )); |
| 190 | } |
| 191 | |
| 192 | #[test] |
| 193 | fn sleb_unterminated_errors() { |
| 194 | let buf = vec![0x80, 0x80]; |
| 195 | assert!(matches!( |
| 196 | read_sleb(&buf).unwrap_err(), |
| 197 | ReadError::Truncated { .. } |
| 198 | )); |
| 199 | } |
| 200 | |
| 201 | #[test] |
| 202 | fn uleb_consumes_exactly_first_encoding() { |
| 203 | let mut buf = Vec::new(); |
| 204 | write_uleb(100, &mut buf); |
| 205 | buf.extend_from_slice(&[0xff, 0xee]); // trailing unrelated bytes |
| 206 | let (v, n) = read_uleb(&buf).unwrap(); |
| 207 | assert_eq!(v, 100); |
| 208 | assert_eq!(n, 1); |
| 209 | } |
| 210 | } |
| 211 |