LLVM and Clang wanted features & bugfixes

[ojeda]

Features that we would like to see

Required (we almost certainly want them)

• If a new approach for kCFI type IDs is introduced, then a way to query/set the kind being used is needed.

  • The mechanism used in both rustc and Clang need to agree, thus we need to ensure they are the same.
  • If a single approach is supported (rather than supporting several at the same time), then it may be hard to mix and match rustc/Clang versions.
  • PR (introducing a new approach): [X86] Enhance kCFI type IDs with a 3-bit arity indicator. llvm/llvm-project#117121.
  • PR (another approach): [X86] Extend kCFI with a 3-bit arity indicator llvm/llvm-project#121070.

• Clarify that --internalize in llvm-link does what we want it to do for the inline helpers without full LTO approach.

  • PR: [llvm-link] Add more detail to --internalize description llvm/llvm-project#170397.
  • Lore: https://lore.kernel.org/rust-for-linux/CAGSQo01=7YDfYzbjipNtsK8AzXciSpY+v53kAhh+oZ_QisAh4Q@mail.gmail.com/.

• Supporting BTF for Rust with LTO by generating BTF directly instead of going through pahole.

  • PR (--btf-merge for lld): [lld][DebugInfo/BTF] Add BTF section merging and deduplication llvm/llvm-project#183915.
  • PR (-gbtf for Clang, like GCC's): [Clang][CodeGen] Add -gbtf: target-independent BTF emission for ELF targets llvm/llvm-project#183929.
  • Zulip: https://rust-lang.zulipchat.com/#narrow/channel/425075-rust-for-linux/topic/BTF.20generation.20in.20rustc.3F/near/577913877.

• Allow inlining with null_pointer_is_valid mismatch.

  • PR: [Inline] Allow inlining with null_pointer_is_valid mismatch llvm/llvm-project#190510.
  • Rust Zulip: https://rust-lang.zulipchat.com/#narrow/channel/425075-rust-for-linux/topic/Cross.20language.20inlining.20and.20-fno-delete-null-pointer-checks/near/583438139.

Nice to have (not critical, we could workaround if needed, etc.)

• Querying GCC-style inline assembly so that bindgen can translate it into Rust inline assembly.

  • Discourse: https://discourse.llvm.org/t/a-small-proposal-for-extraction-of-inline-assembly-block-information/86658.
  • PR: [clang-c] introduce queries on GCC-style inline assembly statements llvm/llvm-project#143424.
  • Cc: @dingxiangfei2009.

• LLVM emits a 64-bit division even when dividing a u64 by a constant (e.g. 10), which is a problem for e.g. 32-bit arm.

  • Issue: Division of uin64_t by constant on ARM32 should multiply by inverse llvm/llvm-project#37280.
  • Lore: https://lore.kernel.org/rust-for-linux/20250627123854.921704-1-jfalempe@redhat.com/.

Low priority (we will likely not use them in the end)

Done (stabilized, fixed, not needed anymore, etc.)

---

Bugs that we would like to see fixed

Required (we almost certainly want them)

Nice to have (probably not critical, we could workaround if needed, etc.)

Low priority (we will likely not use them in the end)

Done (stabilized, fixed, or not needed anymore, etc.)

• LLVM miscompilation for asm goto with outputs.
  • Issue: asm goto miscompilation llvm/llvm-project#74483.
  • Issue: asm goto wrong output at -O0 llvm/llvm-project#110251.
  • PR: [RegAllocFast] Ensure live-in vregs get reloaded after INLINEASM_BR spills llvm/llvm-project#131350 (21?).