1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
#![allow(clippy::cast_ptr_alignment)]
use anyhow::{bail, ensure, Error};
use object::{RelocationEncoding, RelocationKind};
use std::collections::HashMap;
pub use crate::write_debuginfo::{emit_dwarf, DwarfSection, DwarfSectionRelocTarget};
mod gc;
mod transform;
mod write_debuginfo;
pub fn create_gdbjit_image(
mut bytes: Vec<u8>,
code_region: (*const u8, usize),
defined_funcs_offset: usize,
funcs: &[*const u8],
) -> Result<Vec<u8>, Error> {
ensure_supported_elf_format(&mut bytes)?;
relocate_dwarf_sections(&mut bytes, defined_funcs_offset, funcs)?;
convert_object_elf_to_loadable_file(&mut bytes, code_region);
Ok(bytes)
}
fn relocate_dwarf_sections(
bytes: &mut [u8],
defined_funcs_offset: usize,
funcs: &[*const u8],
) -> Result<(), Error> {
use object::read::{File, Object, ObjectSection, ObjectSymbol, RelocationTarget};
let obj = File::parse(bytes)?;
let mut func_symbols = HashMap::new();
for sym in obj.symbols() {
match (sym.name(), sym.section_index()) {
(Ok(name), Some(_section_index)) if name.starts_with("_wasm_function_") => {
let index = name["_wasm_function_".len()..].parse::<usize>()?;
let data = funcs[index - defined_funcs_offset];
func_symbols.insert(sym.index(), data);
}
_ => (),
}
}
for section in obj.sections() {
for (off, r) in section.relocations() {
if r.kind() != RelocationKind::Absolute
|| r.encoding() != RelocationEncoding::Generic
|| r.size() != 64
{
continue;
}
let data = match r.target() {
RelocationTarget::Symbol(ref index) => func_symbols.get(index),
_ => None,
};
let data: *const u8 = match data {
Some(data) => *data,
None => {
continue;
}
};
let target = (data as u64).wrapping_add(r.addend() as u64);
let entry_ptr = section.data_range(off, 8).unwrap().unwrap().as_ptr();
unsafe {
std::ptr::write(entry_ptr as *mut u64, target);
}
}
}
Ok(())
}
fn ensure_supported_elf_format(bytes: &mut Vec<u8>) -> Result<(), Error> {
use object::elf::*;
use object::endian::LittleEndian;
use std::mem::size_of;
let e = LittleEndian;
let header: &FileHeader64<LittleEndian> =
unsafe { &*(bytes.as_mut_ptr() as *const FileHeader64<_>) };
ensure!(
header.e_ident.class == ELFCLASS64 && header.e_ident.data == ELFDATA2LSB,
"bits and endianess in .ELF",
);
match header.e_machine.get(e) {
EM_X86_64 => (),
machine => {
bail!("Unsupported ELF target machine: {:x}", machine);
}
}
ensure!(
header.e_phoff.get(e) == 0 && header.e_phnum.get(e) == 0,
"program header table is empty"
);
let e_shentsize = header.e_shentsize.get(e);
ensure!(
e_shentsize as usize == size_of::<SectionHeader64<LittleEndian>>(),
"size of sh"
);
Ok(())
}
fn convert_object_elf_to_loadable_file(bytes: &mut Vec<u8>, code_region: (*const u8, usize)) {
use object::elf::*;
use object::endian::LittleEndian;
use std::ffi::CStr;
use std::mem::size_of;
use std::os::raw::c_char;
let e = LittleEndian;
let header: &FileHeader64<LittleEndian> =
unsafe { &*(bytes.as_mut_ptr() as *const FileHeader64<_>) };
let e_shentsize = header.e_shentsize.get(e);
let e_shoff = header.e_shoff.get(e);
let e_shnum = header.e_shnum.get(e);
let mut shstrtab_off = 0;
for i in 0..e_shnum {
let off = e_shoff as isize + i as isize * e_shentsize as isize;
let section: &SectionHeader64<LittleEndian> =
unsafe { &*(bytes.as_ptr().offset(off) as *const SectionHeader64<_>) };
if section.sh_type.get(e) != SHT_STRTAB {
continue;
}
shstrtab_off = section.sh_offset.get(e);
}
let mut segment: Option<_> = None;
for i in 0..e_shnum {
let off = e_shoff as isize + i as isize * e_shentsize as isize;
let section: &mut SectionHeader64<LittleEndian> =
unsafe { &mut *(bytes.as_mut_ptr().offset(off) as *mut SectionHeader64<_>) };
if section.sh_type.get(e) != SHT_PROGBITS {
continue;
}
let sh_name_off = section.sh_name.get(e);
let sh_name = unsafe {
CStr::from_ptr(
bytes
.as_ptr()
.offset((shstrtab_off + sh_name_off as u64) as isize)
as *const c_char,
)
.to_str()
.expect("name")
};
if sh_name != ".text" {
continue;
}
assert!(segment.is_none());
section.sh_addr.set(e, code_region.0 as u64);
let sh_offset = section.sh_offset.get(e);
let sh_size = section.sh_size.get(e);
segment = Some((sh_offset, sh_size));
}
let ph_off = bytes.len();
let e_phentsize = size_of::<ProgramHeader64<LittleEndian>>();
let e_phnum = 1;
bytes.resize(ph_off + e_phentsize * e_phnum, 0);
if let Some((sh_offset, sh_size)) = segment {
let (v_offset, size) = code_region;
let program: &mut ProgramHeader64<LittleEndian> =
unsafe { &mut *(bytes.as_ptr().add(ph_off) as *mut ProgramHeader64<_>) };
program.p_type.set(e, PT_LOAD);
program.p_offset.set(e, sh_offset);
program.p_vaddr.set(e, v_offset as u64);
program.p_paddr.set(e, v_offset as u64);
program.p_filesz.set(e, sh_size as u64);
program.p_memsz.set(e, size as u64);
} else {
unreachable!();
}
let header: &mut FileHeader64<LittleEndian> =
unsafe { &mut *(bytes.as_mut_ptr() as *mut FileHeader64<_>) };
header.e_type.set(e, ET_DYN);
header.e_phoff.set(e, ph_off as u64);
header.e_phentsize.set(e, e_phentsize as u16);
header.e_phnum.set(e, e_phnum as u16);
}