struct sigcontext in Hurd/x86_64

[Bruno Haible]

Hi,

While trying to understand the patch submitted at
https://lists.gnu.org/archive/html/bug-gnulib/2023-05/msg00048.html
I'm looking at three files:

* gnumach/x86_64/locore.S lines 512..519

  ENTRY(alltraps)
        pusha                           /* save the general registers */
trap_push_segs:
        movq    %ds,%rax                        /* and the segment registers */
        pushq   %rax
        movq    %es,%rax                        /* and the segment registers */
        pushq   %rax
        PUSH_FSGS

  By the definition of 'pusha' (line 42), this produces on the stack the values 
of the
  registers
          unused
          unused
          es
          ds
          r15
          r14
          r13
          r12
          r11
          r10
          r9
          r8
          rdi
          rsi
          rbp
          unused
          rbx
          rdx
          rcx
          rax

* glibc/sysdeps/mach/hurd/x86_64/bits/sigcontext.h lines 57..79
* glibc/sysdeps/mach/hurd/x86/trampoline.c lines 239..247.
  This code copies the values from the stack into a 'struct sigcontext'.
  But here the order of the registers is
          gs
          fs
          es
          ds
          r8
          r9
          r10
          r11
          r12
          r13
          r14
          r15
          rdi
          rsi
          rbp
          unused
          rbx
          rdx
          rcx
          rax

This makes no sense to me. Either the fields of struct sigcontext should
be renamed
          sc_r8 <--> sc_r15
          sc_r9 <--> sc_r14
          sc_r10 <--> sc_r13
          sc_r11 <--> sc_r12
Or the pusha macro should be changed to

#define pusha pushq %rax ; pushq %rcx ; pushq %rdx ; pushq %rbx ; subq $8,%rsp 
; pushq %rbp ; pushq %rsi ; pushq %rdi ; pushq %r15 ; pushq %r14 ; pushq %r13 ; 
pushq %r12 ; pushq %r11 ; pushq %r10 ; pushq %r9 ; pushq %r8

and the popa macro accordingly.

Bruno