T0b1/wasmtime
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Rework/simplify unwind infrastructure and implement Windows unwind.

Browse Source 
Our previous implementation of unwind infrastructure was somewhat
complex and brittle: it parsed generated instructions in order to
reverse-engineer unwind info from prologues. It also relied on some
fragile linkage to communicate instruction-layout information that VCode
was not designed to provide.

A much simpler, more reliable, and easier-to-reason-about approach is to
embed unwind directives as pseudo-instructions in the prologue as we
generate it. That way, we can say what we mean and just emit it
directly.

The usual reasoning that leads to the reverse-engineering approach is
that metadata is hard to keep in sync across optimization passes; but
here, (i) prologues are generated at the very end of the pipeline, and
(ii) if we ever do a post-prologue-gen optimization, we can treat unwind
directives as black boxes with unknown side-effects, just as we do for
some other pseudo-instructions today.

It turns out that it was easier to just build this for both x64 and
aarch64 (since they share a factored-out ABI implementation), and wire
up the platform-specific unwind-info generation for Windows and SystemV.
Now we have simpler unwind on all platforms and we can delete the old
unwind infra as soon as we remove the old backend.

There were a few consequences to supporting Fastcall unwind in
particular that led to a refactor of the common ABI. Windows only
supports naming clobbered-register save locations within 240 bytes of
the frame-pointer register, whatever one chooses that to be (RSP or
RBP). We had previously saved clobbers below the fixed frame (and below
nominal-SP). The 240-byte range has to include the old RBP too, so we're
forced to place clobbers at the top of the frame, just below saved
RBP/RIP. This is fine; we always keep a frame pointer anyway because we
use it to refer to stack args. It does mean that offsets of fixed-frame
slots (spillslots, stackslots) from RBP are no longer known before we do
regalloc, so if we ever want to index these off of RBP rather than
nominal-SP because we add support for `alloca` (dynamic frame growth),
then we'll need a "nominal-BP" mode that is resolved after regalloc and
clobber-save code is generated. I added a comment to this effect in
`abi_impl.rs`.

The above refactor touched both x64 and aarch64 because of shared code.
This had a further effect in that the old aarch64 prologue generation
subtracted from `sp` once to allocate space, then used stores to `[sp,
offset]` to save clobbers. Unfortunately the offset only has 7-bit
range, so if there are enough clobbered registers (and there can be --
aarch64 has 384 bytes of registers; at least one unit test hits this)
the stores/loads will be out-of-range. I really don't want to synthesize
large-offset sequences here; better to go back to the simpler
pre-index/post-index `stp r1, r2, [sp, #-16]` form that works just like
a "push". It's likely not much worse microarchitecturally (dependence
chain on SP, but oh well) and it actually saves an instruction if
there's no other frame to allocate. As a further advantage, it's much
simpler to understand; simpler is usually better.

This PR adds the new backend on Windows to CI as well.
This commit is contained in:
Chris Fallin
2021-03-06 15:43:09 -08:00
parent 05688aa8f4
commit 2d5db92a9e
63 changed files with 905 additions and 1088 deletions
Download Patch File Download Diff File Expand all files Collapse all files

144
cranelift/codegen/src/isa/aarch64/inst/unwind/systemv.rs
View File

@@ -1,9 +1,7 @@
//! Unwind information for System V ABI (Aarch64).
use crate::isa::aarch64::inst::regs;
use crate::isa::unwind::input;
use crate::isa::unwind::systemv::{RegisterMappingError, UnwindInfo};
use crate::result::CodegenResult;
use crate::isa::unwind::systemv::RegisterMappingError;
use gimli::{write::CommonInformationEntry, Encoding, Format, Register};
use regalloc::{Reg, RegClass};
@@ -31,128 +29,40 @@ pub fn create_cie() -> CommonInformationEntry {
/// Map Cranelift registers to their corresponding Gimli registers.
pub fn map_reg(reg: Reg) -> Result<Register, RegisterMappingError> {
// For AArch64 DWARF register mappings, see:
//
// https://developer.arm.com/documentation/ihi0057/e/?lang=en#dwarf-register-names
//
// X0--X31 is 0--31; V0--V31 is 64--95.
match reg.get_class() {
RegClass::I64 => Ok(Register(reg.get_hw_encoding().into())),
RegClass::I64 => {
let reg = reg.get_hw_encoding() as u16;
Ok(Register(reg))
}
RegClass::V128 => {
let reg = reg.get_hw_encoding() as u16;
Ok(Register(64 + reg))
}
_ => Err(RegisterMappingError::UnsupportedRegisterBank("class?")),
}
}
pub(crate) fn create_unwind_info(
unwind: input::UnwindInfo<Reg>,
) -> CodegenResult<Option<UnwindInfo>> {
struct RegisterMapper;
impl crate::isa::unwind::systemv::RegisterMapper<Reg> for RegisterMapper {
fn map(&self, reg: Reg) -> Result<u16, RegisterMappingError> {
Ok(map_reg(reg)?.0)
}
fn sp(&self) -> u16 {
regs::stack_reg().get_hw_encoding().into()
}
pub(crate) struct RegisterMapper;
impl crate::isa::unwind::systemv::RegisterMapper<Reg> for RegisterMapper {
fn map(&self, reg: Reg) -> Result<u16, RegisterMappingError> {
Ok(map_reg(reg)?.0)
}
let map = RegisterMapper;
Ok(Some(UnwindInfo::build(unwind, &map)?))
}
#[cfg(test)]
mod tests {
use crate::cursor::{Cursor, FuncCursor};
use crate::ir::{
types, AbiParam, ExternalName, Function, InstBuilder, Signature, StackSlotData,
StackSlotKind,
};
use crate::isa::{lookup, CallConv};
use crate::settings::{builder, Flags};
use crate::Context;
use gimli::write::Address;
use std::str::FromStr;
use target_lexicon::triple;
#[test]
fn test_simple_func() {
let isa = lookup(triple!("aarch64"))
.expect("expect aarch64 ISA")
.finish(Flags::new(builder()));
let mut context = Context::for_function(create_function(
CallConv::SystemV,
Some(StackSlotData::new(StackSlotKind::ExplicitSlot, 64)),
));
context.compile(&*isa).expect("expected compilation");
let fde = match context
.create_unwind_info(isa.as_ref())
.expect("can create unwind info")
{
Some(crate::isa::unwind::UnwindInfo::SystemV(info)) => {
info.to_fde(Address::Constant(1234))
}
_ => panic!("expected unwind information"),
};
assert_eq!(format!("{:?}", fde), "FrameDescriptionEntry { address: Constant(1234), length: 24, lsda: None, instructions: [(4, CfaOffset(16)), (4, Offset(Register(29), -16)), (4, Offset(Register(30), -8)), (8, CfaRegister(Register(29)))] }");
fn sp(&self) -> u16 {
regs::stack_reg().get_hw_encoding().into()
}
fn create_function(call_conv: CallConv, stack_slot: Option<StackSlotData>) -> Function {
let mut func =
Function::with_name_signature(ExternalName::user(0, 0), Signature::new(call_conv));
let block0 = func.dfg.make_block();
let mut pos = FuncCursor::new(&mut func);
pos.insert_block(block0);
pos.ins().return_(&[]);
if let Some(stack_slot) = stack_slot {
func.stack_slots.push(stack_slot);
}
func
fn fp(&self) -> u16 {
regs::fp_reg().get_hw_encoding().into()
}
#[test]
fn test_multi_return_func() {
let isa = lookup(triple!("aarch64"))
.expect("expect aarch64 ISA")
.finish(Flags::new(builder()));
let mut context = Context::for_function(create_multi_return_function(CallConv::SystemV));
context.compile(&*isa).expect("expected compilation");
let fde = match context
.create_unwind_info(isa.as_ref())
.expect("can create unwind info")
{
Some(crate::isa::unwind::UnwindInfo::SystemV(info)) => {
info.to_fde(Address::Constant(4321))
}
_ => panic!("expected unwind information"),
};
assert_eq!(format!("{:?}", fde), "FrameDescriptionEntry { address: Constant(4321), length: 40, lsda: None, instructions: [(4, CfaOffset(16)), (4, Offset(Register(29), -16)), (4, Offset(Register(30), -8)), (8, CfaRegister(Register(29)))] }");
fn lr(&self) -> Option<u16> {
Some(regs::link_reg().get_hw_encoding().into())
}
fn create_multi_return_function(call_conv: CallConv) -> Function {
let mut sig = Signature::new(call_conv);
sig.params.push(AbiParam::new(types::I32));
let mut func = Function::with_name_signature(ExternalName::user(0, 0), sig);
let block0 = func.dfg.make_block();
let v0 = func.dfg.append_block_param(block0, types::I32);
let block1 = func.dfg.make_block();
let block2 = func.dfg.make_block();
let mut pos = FuncCursor::new(&mut func);
pos.insert_block(block0);
pos.ins().brnz(v0, block2, &[]);
pos.ins().jump(block1, &[]);
pos.insert_block(block1);
pos.ins().return_(&[]);
pos.insert_block(block2);
pos.ins().return_(&[]);
func
fn lr_offset(&self) -> Option<u32> {
Some(8)
}
}