[Bug 1843394] Re: FTBFS in Eoan - Error: operand type mismatch for `push' - gcc 9.2.1 / binutils 2.32.51.20190905-0ubuntu1

[Christian Ehrhardt ]

Quoting a post of a ipxe related developer that did a nice check
realizing that e.g. disassembly always lists pushq (and opcode stays the
same) - thanks to Valentine for these checks!:


AFAIU, segment registers
are 16-bit long, however, "pushl" and "pushq" should also be valid and
produce the same opcode.

Quite interesting that obdump -S always shows "pushq" even when
"push" or "pushl" is used in the original source.

Please, take a look at the following log:

user at ubuntu19.10:/tmp$ cat push.S
 push %gs
 push %fs
user at ubuntu19.10:/tmp$ as --64 push.S -o push.out
user at ubuntu19.10:/tmp$ objdump -S push.out

push.out:     file format elf64-x86-64


Disassembly of section .text:

0000000000000000 <.text>:
   0:   0f a8                   pushq  %gs
   2:   0f a0                   pushq  %fs
user at ubuntu19.10:/tmp$ cat pushl.S
 pushl %gs
 pushl %fs
user at ubuntu19.10:/tmp$ as --64 pushl.S -o pushl.out
user at ubuntu19.10:/tmp$ objdump -S pushl.out

pushl.out:     file format elf64-x86-64


Disassembly of section .text:

0000000000000000 <.text>:
   0:   0f a8                   pushq  %gs
   2:   0f a0                   pushq  %fs
user at ubuntu19.10:/tmp$ cat pushq.S
 pushq %gs
 pushq %fs
user at ubuntu19.10:/tmp$ as --64 pushq.S -o pushq.out
pushq.S: Assembler messages:
pushq.S:1: Error: operand type mismatch for `push'
pushq.S:2: Error: operand type mismatch for `push'

-- 
You received this bug notification because you are a member of Ubuntu
Foundations Bugs, which is subscribed to binutils in Ubuntu.
https://bugs.launchpad.net/bugs/1843394

Title:
  FTBFS in Eoan - Error: operand type mismatch for `push' - gcc 9.2.1 /
  binutils 2.32.51.20190905-0ubuntu1

Status in binutils package in Ubuntu:
  New
Status in ipxe package in Ubuntu:
  New

Bug description:
  This might be due to new gcc-9 being more strict, but the build that
  worked before now fails with:

  arch/x86_64/core/gdbidt.S: Assembler messages:
  arch/x86_64/core/gdbidt.S:109: Error: operand type mismatch for `push'
  arch/x86_64/core/gdbidt.S:110: Error: operand type mismatch for `push'
  arch/x86_64/core/gdbidt.S:161: Error: operand type mismatch for `pop'
  arch/x86_64/core/gdbidt.S:162: Error: operand type mismatch for `pop'
  make[2]: *** [Makefile.housekeeping:937: bin-x86_64-efi/gdbidt.o] Error 1

  Full log at: https://launchpadlibrarian.net/441262285/buildlog_ubuntu-
  eoan-amd64.ipxe_1.0.0+git-20190109.133f4c4-0ubuntu2_BUILDING.txt.gz

  Now all of this is about push/pop of %fs and %gs.
  That needs to match the size of the registers which depend on the current running mode.

  In this particular case in ./src/arch/x86_64/core/gdbidt.S
  The failing file is in ".code64" mode.
  In that I'd expect %gs/%fs to be 64 bit.

  
  Usually we see push/pop "w" in .code16 (word), l in .code32 (long) but in that sense here q (quad word) seems right at first (should be what correctly matches the .code64).
  That matches what I see throughout the ipxe code but also throughout the archive https://codesearch.debian.net/search?q=pop%5Ba-z%5D.*%25fs&literal=0&page=2

  Maybe I misread the mode it is in, or it is actually a false positives.
  Or the sizes of FS/GS do not change - haven't touched them in a loooong time.
  Was it that segment registers didn't change size?
  I'll need to do a few checks to first see what the compiler would expect there and from there need to understand this.

  The command used also points to AS being in 64 bit mode when this happens:
  gcc -E  -DARCH=x86_64 -DPLATFORM=efi -DSECUREBOOT=0 -fstrength-reduce -fomit-frame-pointer -falign-jumps=1 -falign-loops=1 -falign-functions=1 -m64 -mno-mmx -mno-sse -fshort-wchar -Ui386 -Ulinux -DNVALGRIND -fpie -mno-red-zone -Iinclude -I. -Iarch/x86/include -Iarch/x86_64/include -Iarch/x86_64/include/efi -Os -g -ffreestanding -Wall -W -Wformat-nonliteral  -fno-stack-protector -fno-dwarf2-cfi-asm -fno-exceptions  -fno-unwind-tables -fno-asynchronous-unwind-tables -Wno-address -Wno-stringop-truncation   -ffunction-sections -fdata-sections -include include/compiler.h -DASM_TCHAR='@' -DASM_TCHAR_OPS='@' -DASSEMBLY  -DOBJECT=gdbidt arch/x86_64/core/gdbidt.S | as  --64    -o bin-x86_64-efi/gdbidt.o

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/binutils/+bug/1843394/+subscriptions