wasmtime_environ/compile/module_artifacts.rs
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329
//! Definitions of runtime structures and metadata which are serialized into ELF
//! with `postcard` as part of a module's compilation process.
use crate::prelude::*;
use crate::{
obj, CompiledFunctionInfo, CompiledModuleInfo, DebugInfoData, DefinedFuncIndex, FunctionLoc,
FunctionName, MemoryInitialization, Metadata, ModuleInternedTypeIndex, ModuleTranslation,
PrimaryMap, Tunables,
};
use anyhow::{bail, Result};
use object::write::{Object, SectionId, StandardSegment, WritableBuffer};
use object::SectionKind;
use std::ops::Range;
/// Helper structure to create an ELF file as a compilation artifact.
///
/// This structure exposes the process which Wasmtime will encode a core wasm
/// module into an ELF file, notably managing data sections and all that good
/// business going into the final file.
pub struct ObjectBuilder<'a> {
/// The `object`-crate-defined ELF file write we're using.
obj: Object<'a>,
/// General compilation configuration.
tunables: &'a Tunables,
/// The section identifier for "rodata" which is where wasm data segments
/// will go.
data: SectionId,
/// The target triple for this compilation.
triple: target_lexicon::Triple,
/// The section identifier for function name information, or otherwise where
/// the `name` custom section of wasm is copied into.
///
/// This is optional and lazily created on demand.
names: Option<SectionId>,
/// The section identifier for dwarf information copied from the original
/// wasm files.
///
/// This is optional and lazily created on demand.
dwarf: Option<SectionId>,
}
impl<'a> ObjectBuilder<'a> {
/// Creates a new builder for the `obj` specified.
pub fn new(
mut obj: Object<'a>,
tunables: &'a Tunables,
triple: target_lexicon::Triple,
) -> ObjectBuilder<'a> {
let data = obj.add_section(
obj.segment_name(StandardSegment::Data).to_vec(),
obj::ELF_WASM_DATA.as_bytes().to_vec(),
SectionKind::ReadOnlyData,
);
ObjectBuilder {
obj,
tunables,
data,
triple,
names: None,
dwarf: None,
}
}
/// Insert the wasm raw wasm-based debuginfo into the output.
/// Note that this is distinct from the native debuginfo
/// possibly generated by the native compiler, hence these sections
/// getting wasm-specific names.
pub fn push_debuginfo(
&mut self,
dwarf: &mut Vec<(u8, Range<u64>)>,
debuginfo: &DebugInfoData<'_>,
) {
self.push_debug(dwarf, &debuginfo.dwarf.debug_abbrev);
self.push_debug(dwarf, &debuginfo.dwarf.debug_addr);
self.push_debug(dwarf, &debuginfo.dwarf.debug_aranges);
self.push_debug(dwarf, &debuginfo.dwarf.debug_info);
self.push_debug(dwarf, &debuginfo.dwarf.debug_line);
self.push_debug(dwarf, &debuginfo.dwarf.debug_line_str);
self.push_debug(dwarf, &debuginfo.dwarf.debug_str);
self.push_debug(dwarf, &debuginfo.dwarf.debug_str_offsets);
self.push_debug(dwarf, &debuginfo.debug_ranges);
self.push_debug(dwarf, &debuginfo.debug_rnglists);
self.push_debug(dwarf, &debuginfo.debug_cu_index);
// Sort this for binary-search-lookup later in `symbolize_context`.
dwarf.sort_by_key(|(id, _)| *id);
}
/// Completes compilation of the `translation` specified, inserting
/// everything necessary into the `Object` being built.
///
/// This function will consume the final results of compiling a wasm module
/// and finish the ELF image in-progress as part of `self.obj` by appending
/// any compiler-agnostic sections.
///
/// The auxiliary `CompiledModuleInfo` structure returned here has also been
/// serialized into the object returned, but if the caller will quickly
/// turn-around and invoke `CompiledModule::from_artifacts` after this then
/// the information can be passed to that method to avoid extra
/// deserialization. This is done to avoid a serialize-then-deserialize for
/// API calls like `Module::new` where the compiled module is immediately
/// going to be used.
///
/// The various arguments here are:
///
/// * `translation` - the core wasm translation that's being completed.
///
/// * `funcs` - compilation metadata about functions within the translation
/// as well as where the functions are located in the text section and any
/// associated trampolines.
///
/// * `wasm_to_array_trampolines` - list of all trampolines necessary for
/// Wasm callers calling array callees (e.g. `Func::wrap`). One for each
/// function signature in the module. Must be sorted by `SignatureIndex`.
///
/// Returns the `CompiledModuleInfo` corresponding to this core Wasm module
/// as a result of this append operation. This is then serialized into the
/// final artifact by the caller.
pub fn append(
&mut self,
translation: ModuleTranslation<'_>,
funcs: PrimaryMap<DefinedFuncIndex, CompiledFunctionInfo>,
wasm_to_array_trampolines: Vec<(ModuleInternedTypeIndex, FunctionLoc)>,
) -> Result<CompiledModuleInfo> {
let ModuleTranslation {
mut module,
debuginfo,
has_unparsed_debuginfo,
data,
data_align,
passive_data,
..
} = translation;
// Place all data from the wasm module into a section which will the
// source of the data later at runtime. This additionally keeps track of
// the offset of
let mut total_data_len = 0;
let data_offset = self
.obj
.append_section_data(self.data, &[], data_align.unwrap_or(1));
for (i, data) in data.iter().enumerate() {
// The first data segment has its alignment specified as the alignment
// for the entire section, but everything afterwards is adjacent so it
// has alignment of 1.
let align = if i == 0 { data_align.unwrap_or(1) } else { 1 };
self.obj.append_section_data(self.data, data, align);
total_data_len += data.len();
}
for data in passive_data.iter() {
self.obj.append_section_data(self.data, data, 1);
}
// If any names are present in the module then the `ELF_NAME_DATA` section
// is create and appended.
let mut func_names = Vec::new();
if debuginfo.name_section.func_names.len() > 0 {
let name_id = *self.names.get_or_insert_with(|| {
self.obj.add_section(
self.obj.segment_name(StandardSegment::Data).to_vec(),
obj::ELF_NAME_DATA.as_bytes().to_vec(),
SectionKind::ReadOnlyData,
)
});
let mut sorted_names = debuginfo.name_section.func_names.iter().collect::<Vec<_>>();
sorted_names.sort_by_key(|(idx, _name)| *idx);
for (idx, name) in sorted_names {
let offset = self.obj.append_section_data(name_id, name.as_bytes(), 1);
let offset = match u32::try_from(offset) {
Ok(offset) => offset,
Err(_) => bail!("name section too large (> 4gb)"),
};
let len = u32::try_from(name.len()).unwrap();
func_names.push(FunctionName {
idx: *idx,
offset,
len,
});
}
}
// Data offsets in `MemoryInitialization` are offsets within the
// `translation.data` list concatenated which is now present in the data
// segment that's appended to the object. Increase the offsets by
// `self.data_size` to account for any previously added module.
let data_offset = u32::try_from(data_offset).unwrap();
match &mut module.memory_initialization {
MemoryInitialization::Segmented(list) => {
for segment in list {
segment.data.start = segment.data.start.checked_add(data_offset).unwrap();
segment.data.end = segment.data.end.checked_add(data_offset).unwrap();
}
}
MemoryInitialization::Static { map } => {
for (_, segment) in map {
if let Some(segment) = segment {
segment.data.start = segment.data.start.checked_add(data_offset).unwrap();
segment.data.end = segment.data.end.checked_add(data_offset).unwrap();
}
}
}
}
// Data offsets for passive data are relative to the start of
// `translation.passive_data` which was appended to the data segment
// of this object, after active data in `translation.data`. Update the
// offsets to account prior modules added in addition to active data.
let data_offset = data_offset + u32::try_from(total_data_len).unwrap();
for (_, range) in module.passive_data_map.iter_mut() {
range.start = range.start.checked_add(data_offset).unwrap();
range.end = range.end.checked_add(data_offset).unwrap();
}
// Insert the wasm raw wasm-based debuginfo into the output, if
// requested. Note that this is distinct from the native debuginfo
// possibly generated by the native compiler, hence these sections
// getting wasm-specific names.
let mut dwarf = Vec::new();
if self.tunables.parse_wasm_debuginfo {
self.push_debuginfo(&mut dwarf, &debuginfo);
}
let is_pulley = matches!(
self.triple.architecture,
target_lexicon::Architecture::Pulley32 | target_lexicon::Architecture::Pulley64
);
assert!(!is_pulley || wasm_to_array_trampolines.is_empty());
Ok(CompiledModuleInfo {
module,
funcs,
wasm_to_array_trampolines,
func_names,
meta: Metadata {
has_unparsed_debuginfo,
code_section_offset: debuginfo.wasm_file.code_section_offset,
has_wasm_debuginfo: self.tunables.parse_wasm_debuginfo,
is_pulley,
dwarf,
},
})
}
fn push_debug<'b, T>(&mut self, dwarf: &mut Vec<(u8, Range<u64>)>, section: &T)
where
T: gimli::Section<gimli::EndianSlice<'b, gimli::LittleEndian>>,
{
let data = section.reader().slice();
if data.is_empty() {
return;
}
let section_id = *self.dwarf.get_or_insert_with(|| {
self.obj.add_section(
self.obj.segment_name(StandardSegment::Debug).to_vec(),
obj::ELF_WASMTIME_DWARF.as_bytes().to_vec(),
SectionKind::Debug,
)
});
let offset = self.obj.append_section_data(section_id, data, 1);
dwarf.push((T::id() as u8, offset..offset + data.len() as u64));
}
/// Creates the `ELF_WASMTIME_INFO` section from the given serializable data
/// structure.
pub fn serialize_info<T>(&mut self, info: &T)
where
T: serde::Serialize,
{
let section = self.obj.add_section(
self.obj.segment_name(StandardSegment::Data).to_vec(),
obj::ELF_WASMTIME_INFO.as_bytes().to_vec(),
SectionKind::ReadOnlyData,
);
let data = postcard::to_allocvec(info).unwrap();
self.obj.set_section_data(section, data, 1);
}
/// Serializes `self` into a buffer. This can be used for execution as well
/// as serialization.
pub fn finish<T: WritableBuffer>(self, t: &mut T) -> Result<()> {
self.obj.emit(t).map_err(|e| e.into())
}
}
/// A type which can be the result of serializing an object.
pub trait FinishedObject: Sized {
/// Emit the object as `Self`.
fn finish_object(obj: ObjectBuilder<'_>) -> Result<Self>;
}
impl FinishedObject for Vec<u8> {
fn finish_object(obj: ObjectBuilder<'_>) -> Result<Self> {
let mut result = ObjectVec::default();
obj.finish(&mut result)?;
return Ok(result.0);
#[derive(Default)]
struct ObjectVec(Vec<u8>);
impl WritableBuffer for ObjectVec {
fn len(&self) -> usize {
self.0.len()
}
fn reserve(&mut self, additional: usize) -> Result<(), ()> {
assert_eq!(self.0.len(), 0, "cannot reserve twice");
self.0 = Vec::with_capacity(additional);
Ok(())
}
fn resize(&mut self, new_len: usize) {
if new_len <= self.0.len() {
self.0.truncate(new_len)
} else {
self.0.extend(vec![0; new_len - self.0.len()])
}
}
fn write_bytes(&mut self, val: &[u8]) {
self.0.extend(val);
}
}
}
}