On Mon, 2026-02-09 at 20:15 -0800, H. Peter Anvin wrote: > On February 9, 2026 7:11:25 PM PST, Xi Ruoyao <[email protected]> wrote: > > On Mon, 2026-02-02 at 19:57 -0800, H. Peter Anvin wrote: > > > That hack dates back from before the signal frame extension. It is no > > > longer necessary. > > > > Unfortunately at least it seems libgcc unwinder does not handle the > > signal frame extension properly. The code reads: > > > > fde = _Unwind_Find_FDE (context->ra + _Unwind_IsSignalFrame (context) - 1, > > &context->bases); > > if (fde == NULL) > > { > > #ifdef MD_FALLBACK_FRAME_STATE_FOR > > /* Couldn't find frame unwind info for this function. Try a > > target-specific fallback mechanism. This will necessarily > > not provide a personality routine or LSDA. */ > > return MD_FALLBACK_FRAME_STATE_FOR (context, fs); > > #else > > return _URC_END_OF_STACK; > > #endif > > } > > > > fs->pc = context->bases.func; > > > > cie = get_cie (fde); > > insn = extract_cie_info (cie, context, fs); > > > > Thus, it indeed attempts to avoid subtracting 1 for a signal frame, but > > ... _Unwind_IsSignalFrame (context) actually extracts a flag in context > > which will only be raised up by extract_cie_info. > > > > Or am I missing something here? > > > > Oh, good grief... > > How does this possibly work on non-x86 platforms?
On ARM64 the vdso does not have eh_frame_hdr at all, on LoongArch eh_frame_hdr is empty (note that an ampty en_frame_hdr is actually buggy and I'm trying to fix it), so _Unwind_Find_FDE returns NULL and libgcc falls back to MD_FALLBACK_FRAME_STATE_FOR, which handles the sigreturn trampoline using some machine-dependant logic. On RISC-V things are more theatrical: the sigreturn trampoline happens to be at the beginning of the vdso .text section, so after subtracting 1 from the PC, the result is out of the .text section and so not in any FDE. Thus _Unwind_Find_FDE returns NULL and libgcc again falls back to MD_FALLBACK_FRAME_STATE_FOR. If the RISC-V sigreturn trampoline was not the first in .text, subtracting 1 would cause the PC to be in the FDE of the previous function and then _Unwind_Find_FDE would return that FDE, then RISC-V would have some big trouble. I've not taken a serious look at other architectures yet. -- Xi Ruoyao <[email protected]>
