SIGILL / LLVM ERROR from rustc when compiling to a x86_64 target that lacks SSE2 instructions

[japaric]

Steps to reproduce

$ cargo new --lib repro
$ cd repro

$ echo '#![no_std]' > src/lib.rs
$ cargo add poly1305@0.8.0

$ rustup default 1.73.0
$ rustup target add x86_64-unknown-none
$ cargo b --target x86_64-unknown-none
error: could not compile `poly1305` (lib)

Caused by:
  process didn't exit successfully: `$RUSTUP_TOOLCHAIN/bin/rustc (..)` (signal: 4, SIGILL: illegal instruction)

at some point I was able to get a "LLVM ERROR" message but I can't no longer reproduce that

---

after some digging it seems that the lack of SSE2 instructions in the x86_64-unknown-none is the problem:

$ rustup default nightly-2023-11-15
$ rustc -Z unstable-options --print target-spec-json --target x86_64-unknown-none | grep features
  "features": "-mmx,-sse,-sse2,-sse3,-ssse3,-sse4.1,-sse4.2,-3dnow,-3dnowa,-avx,-avx2,+soft-float",

creating a custom target that adds back the sse2 feature makes the crate compile

$ rustc -Z unstable-options --print target-spec-json --target x86_64-unknown-none > x86_64.json

patch the JSON like this

--- before.json	2023-11-15 14:35:19.565448966 +0100
+++ x86_64.json	2023-11-15 14:35:10.115355396 +0100
@@ -5,8 +5,7 @@
   "crt-objects-fallback": "false",
   "data-layout": "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128",
   "disable-redzone": true,
-  "features": "-mmx,-sse,-sse2,-sse3,-ssse3,-sse4.1,-sse4.2,-3dnow,-3dnowa,-avx,-avx2,+soft-float",
-  "is-builtin": true,
+  "features": "-mmx,-sse3,-ssse3,-sse4.1,-sse4.2,-3dnow,-3dnowa,-avx,-avx2",
   "linker": "rust-lld",
   "linker-flavor": "gnu-lld",
   "llvm-target": "x86_64-unknown-none-elf",

$ rustup component add rust-src
$ cargo b -j1 -Z build-std=core --target ./x86_64.json && echo OK
OK

keeping the -sse2 produces "LLVM ERROR: Access past stack top!"

perhaps the auto_detect module could use #[cfg(not(target_feature = "sse2"))] to skip the auto-detection and directly use the soft implementation? or you could add something like this:

#[cfg(not(target_feature = "sse2"))]
compile_error!("compilation targets without SSE2 instructions are not supported");

which is nicer than the SIGILL

---

for additional context ("why are you even using this target?"): I was looking for a built-in "OS agnostic" / no-std target for a no-std rustls demo (rustls/rustls PR1534) and tried x86_64-unknown-none and run into this and other problems. I ended up using a custom target for the demo but I figured I should still report what I found.

by the way, the sha2 has a similar problem with this target. Let me know if you think I should report a similar issue in the hashes repo.

[newpavlov]

cpufeatures should support freestanding x86 targets, see RustCrypto/utils#821

What do you get for target_os?

Small rant: we are dealing with a clear deficiency in both the Rust language and LLVM. Currently Rust only supports +feature and (effectively) ?feature, while we also need -feature, i.e. a way to indicate that autodetection code should eliminate branches dependent on the feature. While LLVM in theory should be able to compile SIMD-dependent functions for which we explicitly enabled target features dependent on hard floats.

[tarcieri]

@japaric there's a cfg(poly1305_force_soft) you can set to disable CPU feature autodetection.

Unfortunately these x86_64-unknown-none targets are something of a longstanding problem which I thought we had fixed at this point. Apparently not.

I'm a little curious why the existing cpufeatures detection for avx2 wouldn't prevent those codepaths from being taken in the first place: https://github.com/RustCrypto/universal-hashes/blob/master/poly1305/src/backend/autodetect.rs#L9

I would think that would prevent any such codepaths from being taken on x86_64-unknown-none already, so I'm kind of curious what SSE2 instructions are even being executed.

[japaric]

target_os for x86_64-unknown-none is "none"

$ rustc --print cfg --target x86_64-unknown-none | grep target_os
target_os="none"

I don't think target_os="none" is a good proxy for the instructions supported by a compilation target since one can create a custom compilation target that has the same codegen settings as nightly-x86_64-unknown-linux-gnu but set target_os to "none" like I have done in the issue description. target_feature could perhaps be more appropriate but it's lacking in what it currently exposes and I'm no x86_64 expert to know which instruction set / feature should be exposed over there to better support this particular scenario

[tarcieri]

cpufeatures does consult target_feature, but only to bypass runtime checks. Its main goal is to autodetect CPU features at runtime when they haven't been explicitly enabled.

I'm curious how you're running into a situation where the runtime CPUID checks are being bypassed and you're hitting SSE2 code.

[newpavlov]

I don't think target_os="none" is a good proxy for the instructions supported by a compilation target

There are currently no other way. Rust does not expose soft-float "target feature" (let's forget about it's peculiar "negativity"). And even if it did, I think it would be a bad way for handling this situation. We need something like this.

[japaric]

there's a cfg(poly1305_force_soft) you can set to disable CPU feature autodetection.

thanks. I was aware of poly1305_force_soft, curve25519_dalek_backend and sha2's force-soft Cargo feature but using those together with a particular mix of crypto libraries produced a compilation error in curve25519-dalek. that could be a different bug or PEBCAK that I didn't manage to identify -- at that point I gave up on the x86_64-unknown-none target and used a custom one.

[tarcieri]

Is it possible for you to capture this in a debugger and get a backtrace of some sort? I'm curious where the SIGILL is actually occurring and whether or not it's inside a backend which should not be compiled at all on x86_64-unknown-none targets

[newpavlov]

Can you try to compile you code with --release? Maybe LLVM simply fails to eliminate the SIMD branch with opt-level=0 and thus attempts to compile it? The cpufeatures module simply results in false, while the branch is still present in source code.

[japaric]

I'm curious where the SIGILL is actually occurring and whether or not it's inside a backend which should not be compiled at all on x86_64-unknown-none targets

I think the SIGILL is Cargo misreporting the rustc error on stable. if you follow the repro steps with nightly-2023-11-15 you get this LLVM error and no SIGILL

Compiling poly1305 v0.8.0
LLVM ERROR: Do not know how to split the result of this operator!

error: could not compile `poly1305` (lib)

[japaric]

Can you try to compile you code with --release?

interestingly, adding --release produces the SIGILL on both stable and nightly. I don't get the "LLVM ERROR" message using --release

[newpavlov]

Oh... It's SIGILL triggered by the Rust compiler, not "LLVM ERROR: Do not know how to split the result of this operator!". The latter should be handled by the cpufeatures change linked earlier.

I think you should report this error to the Rust lang repository.

[japaric]

the Cargo backtrace does not say much

Thread 2 "cargo" received signal SIGUSR1, User defined signal 1.
[Switching to Thread 0x7ffff7c8e6c0 (LWP 99944)]
0x00007ffff7d9556c in read () from /usr/lib/libc.so.6
(gdb) backtrace
#0  0x00007ffff7d9556c in read () from /usr/lib/libc.so.6
#1  0x0000555556a859d6 in std::sys::unix::fd::FileDesc::read () at library/std/src/sys/unix/fd.rs:90
#2  std::sys::unix::fs::File::read () at library/std/src/sys/unix/fs.rs:1162
#3  std::fs::{impl#5}::read () at library/std/src/fs.rs:749
#4  0x0000555556a535cf in <jobserver[63568e20ee72e952]::imp::Client>::acquire_allow_interrupts ()
#5  0x0000555556a5541d in <jobserver[63568e20ee72e952]::HelperState>::for_each_request::<jobserver[63568e20ee72e952]::imp::spawn_helper::{closure#1}::{closure#0}> ()
#6  0x0000555556a5585b in std[3759e478f3a6c4f2]::sys_common::backtrace::__rust_begin_short_backtrace::<jobserver[63568e20ee72e952]::imp::spawn_helper::{closure#1}, ()> ()
#7  0x0000555556a56a49 in <<std[3759e478f3a6c4f2]::thread::Builder>::spawn_unchecked_<jobserver[63568e20ee72e952]::imp::spawn_helper::{closure#1}, ()>::{closure#1} as core[d28c4e8d9c4eebaa]::ops::function::FnOnce<()>>::call_once::{shim:vtable#0} ()
#8  0x0000555556a9bcd5 in alloc::boxed::{impl#47}::call_once<(), dyn core::ops::function::FnOnce<(), Output=()>, alloc::alloc::Global> () at library/alloc/src/boxed.rs:2007
#9  alloc::boxed::{impl#47}::call_once<(), alloc::boxed::Box<dyn core::ops::function::FnOnce<(), Output=()>, alloc::alloc::Global>, alloc::alloc::Global> () at library/alloc/src/boxed.rs:2007
#10 std::sys::unix::thread::{impl#2}::new::thread_start () at library/std/src/sys/unix/thread.rs:108
#11 0x00007ffff7d1e9eb in ?? () from /usr/lib/libc.so.6
#12 0x00007ffff7da27cc in ?? () from /usr/lib/libc.so.6

running the rustc invocation that Cargo prints through the debugger seems to indicate that something in LLVM triggered the SIGILL

Thread 7 "opt cgu.1" received signal SIGILL, Illegal instruction.
[Switching to Thread 0x7fffe11ff6c0 (LWP 100591)]
0x00007ffff13c939f in llvm::X86TargetLowering::ReplaceNodeResults(llvm::SDNode*, llvm::SmallVectorImpl<llvm::SDValue>&, llvm::SelectionDAG&) const [clone .cold.0] ()
   from /home/japaric/.rustup/toolchains/1.73.0-x86_64-unknown-linux-gnu/bin/../lib/../lib/libLLVM-17-rust-1.73.0-stable.so
(gdb) backtrace
#0  0x00007ffff13c939f in llvm::X86TargetLowering::ReplaceNodeResults(llvm::SDNode*, llvm::SmallVectorImpl<llvm::SDValue>&, llvm::SelectionDAG&) const [clone .cold.0] ()
   from /home/japaric/.rustup/toolchains/1.73.0-x86_64-unknown-linux-gnu/bin/../lib/../lib/libLLVM-17-rust-1.73.0-stable.so
#1  0x00007ffff1130ef5 in llvm::DAGTypeLegalizer::SplitVectorResult(llvm::SDNode*, unsigned int) [clone .cold.0] () from /home/japaric/.rustup/toolchains/1.73.0-x86_64-unknown-linux-gnu/bin/../lib/../lib/libLLVM-17-rust-1.73.0-stable.so
#2  0x00007ffff000d41a in llvm::DAGTypeLegalizer::run() () from /home/japaric/.rustup/toolchains/1.73.0-x86_64-unknown-linux-gnu/bin/../lib/../lib/libLLVM-17-rust-1.73.0-stable.so
#3  0x00007ffff01ff81a in llvm::SelectionDAGISel::SelectBasicBlock(llvm::ilist_iterator<llvm::ilist_detail::node_options<llvm::Instruction, false, false, void>, false, true>, llvm::ilist_iterator<llvm::ilist_detail::node_options<llvm::Instruction, false, false, void>, false, true>, bool&) () from /home/japaric/.rustup/toolchains/1.73.0-x86_64-unknown-linux-gnu/bin/../lib/../lib/libLLVM-17-rust-1.73.0-stable.so
#4  0x00007ffff04ff282 in llvm::SelectionDAGISel::SelectAllBasicBlocks(llvm::Function const&) () from /home/japaric/.rustup/toolchains/1.73.0-x86_64-unknown-linux-gnu/bin/../lib/../lib/libLLVM-17-rust-1.73.0-stable.so
#5  0x00007ffff034fc0a in llvm::SelectionDAGISel::runOnMachineFunction(llvm::MachineFunction&) () from /home/japaric/.rustup/toolchains/1.73.0-x86_64-unknown-linux-gnu/bin/../lib/../lib/libLLVM-17-rust-1.73.0-stable.so
#6  0x00007ffff034f4ee in (anonymous namespace)::X86DAGToDAGISel::runOnMachineFunction(llvm::MachineFunction&) [clone .llvm.6232165262612102610] () from /home/japaric/.rustup/toolchains/1.73.0-x86_64-unknown-linux-gnu/bin/../lib/../lib/libLLVM-17-rust-1.73.0-stable.so
#7  0x00007ffff016d66a in llvm::FPPassManager::runOnModule(llvm::Module&) () from /home/japaric/.rustup/toolchains/1.73.0-x86_64-unknown-linux-gnu/bin/../lib/../lib/libLLVM-17-rust-1.73.0-stable.so
#8  0x00007ffff05dcaac in llvm::legacy::PassManagerImpl::run(llvm::Module&) () from /home/japaric/.rustup/toolchains/1.73.0-x86_64-unknown-linux-gnu/bin/../lib/../lib/libLLVM-17-rust-1.73.0-stable.so
#9  0x00007ffff64d36a6 in LLVMRustWriteOutputFile () from /home/japaric/.rustup/toolchains/1.73.0-x86_64-unknown-linux-gnu/bin/../lib/librustc_driver-453cf35e1dd187fa.so
#10 0x00007ffff64d2558 in rustc_codegen_llvm[13e834ec38ef84a5]::back::write::write_output_file () from /home/japaric/.rustup/toolchains/1.73.0-x86_64-unknown-linux-gnu/bin/../lib/librustc_driver-453cf35e1dd187fa.so
#11 0x00007ffff64cfcd4 in rustc_codegen_llvm[13e834ec38ef84a5]::back::write::codegen () from /home/japaric/.rustup/toolchains/1.73.0-x86_64-unknown-linux-gnu/bin/../lib/librustc_driver-453cf35e1dd187fa.so
#12 0x00007ffff64cd074 in rustc_codegen_ssa[1239057ba2d16fcb]::back::write::finish_intra_module_work::<rustc_codegen_llvm[13e834ec38ef84a5]::LlvmCodegenBackend> () from /home/japaric/.rustup/toolchains/1.73.0-x86_64-unknown-linux-gnu/bin/../lib/librustc_driver-453cf35e1dd187fa.so
#13 0x00007ffff64cc75d in rustc_codegen_ssa[1239057ba2d16fcb]::back::write::execute_optimize_work_item::<rustc_codegen_llvm[13e834ec38ef84a5]::LlvmCodegenBackend> () from /home/japaric/.rustup/toolchains/1.73.0-x86_64-unknown-linux-gnu/bin/../lib/librustc_driver-453cf35e1dd187fa.so
#14 0x00007ffff64ca627 in std[3759e478f3a6c4f2]::sys_common::backtrace::__rust_begin_short_backtrace::<<rustc_codegen_llvm[13e834ec38ef84a5]::LlvmCodegenBackend as rustc_codegen_ssa[1239057ba2d16fcb]::traits::backend::ExtraBackendMethods>::spawn_named_thread<rustc_codegen_ssa[1239057ba2d16fcb]::back::write::spawn_work<rustc_codegen_llvm[13e834ec38ef84a5]::LlvmCodegenBackend>::{closure#0}, ()>::{closure#0}, ()> () from /home/japaric/.rustup/toolchains/1.73.0-x86_64-unknown-linux-gnu/bin/../lib/librustc_driver-453cf35e1dd187fa.so
#15 0x00007ffff6459256 in <<std[3759e478f3a6c4f2]::thread::Builder>::spawn_unchecked_<<rustc_codegen_llvm[13e834ec38ef84a5]::LlvmCodegenBackend as rustc_codegen_ssa[1239057ba2d16fcb]::traits::backend::ExtraBackendMethods>::spawn_named_thread<rustc_codegen_ssa[1239057ba2d16fcb]::back::write::spawn_work<rustc_codegen_llvm[13e834ec38ef84a5]::LlvmCodegenBackend>::{closure#0}, ()>::{closure#0}, ()>::{closure#1} as core[d28c4e8d9c4eebaa]::ops::function::FnOnce<()>>::call_once::{shim:vtable#0} ()
   from /home/japaric/.rustup/toolchains/1.73.0-x86_64-unknown-linux-gnu/bin/../lib/librustc_driver-453cf35e1dd187fa.so
#16 0x00007ffff3d71295 in alloc::boxed::{impl#47}::call_once<(), dyn core::ops::function::FnOnce<(), Output=()>, alloc::alloc::Global> () at library/alloc/src/boxed.rs:2007
#17 alloc::boxed::{impl#47}::call_once<(), alloc::boxed::Box<dyn core::ops::function::FnOnce<(), Output=()>, alloc::alloc::Global>, alloc::alloc::Global> () at library/alloc/src/boxed.rs:2007
#18 std::sys::unix::thread::{impl#2}::new::thread_start () at library/std/src/sys/unix/thread.rs:108
#19 0x00007ffff3ae59eb in ?? () from /usr/lib/libc.so.6
#20 0x00007ffff3b697cc in ?? () from /usr/lib/libc.so.6

[tarcieri]

Aha, we've definitely encountered that "LLVM ERROR: Do not know how to split the result of this operator!" before:

RustCrypto/AEADs#405
RustCrypto/AEADs#498
RustCrypto/hashes#446

I thought we had successfully worked around it in RustCrypto/utils#821 though. Apparently not.

[newpavlov]

@tarcieri
I think our workarounds works fine, but it triggers a Rust/LLVM compiler bug, which results in SIGILL.

[tarcieri]

Yeah, something very strange is happening

[japaric]

reported this in the rust-lang/rust repo: rust-lang/rust#117938