[Bug target/101023] New: [11/12 Regression] wrong code with -mstackrealign

[Bug target/101023] New: [11/12 Regression] wrong code with -mstackrealign

[zsojka at seznam dot cz]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101023

            Bug ID: 101023
           Summary: [11/12 Regression] wrong code with -mstackrealign
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Keywords: wrong-code
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: zsojka at seznam dot cz
  Target Milestone: ---
              Host: x86_64-pc-linux-gnu
            Target: x86_64-pc-linux-gnu

Created attachment 50981
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=50981&action=edit
auto-reduced testcase (from OpenTTD sources)

Compiler output:
$ x86_64-pc-linux-gnu-g++ -O2 -mtune=opteron -mstackrealign
--param=hot-bb-frequency-fraction=1 testcase.C -S

The offending code is:
.L8:
        pushq   %rbp
        .cfi_def_cfa_offset 16
        .cfi_offset 6, -16
        movq    %rsp, %rbp
        .cfi_def_cfa_register 6
        movq    %r12, -8(%rbp)
        pushq   %rax

the "push" overwrites r12 stored just one instruction above; when returning:
        movq    -8(%rbp), %r12
        xorl    %eax, %eax
        leave
        .cfi_def_cfa 7, 8
        ret

the wrong data is restored to r12

I wasn't able yet to generate an executable testcase, but I can try to if
needed. (it shouldn't be that hard, just r12 needs to be used by the caller
during the call)

$ x86_64-pc-linux-gnu-g++ -v
Using built-in specs.
COLLECT_GCC=/repo/gcc-trunk/binary-latest-amd64/bin/x86_64-pc-linux-gnu-g++
COLLECT_LTO_WRAPPER=/repo/gcc-trunk/binary-trunk-r12-1366-20210610095751-g6fcba9ef23e-checking-yes-rtl-df-extra-nobootstrap-amd64/bin/../libexec/gcc/x86_64-pc-linux-gnu/12.0.0/lto-wrapper
Target: x86_64-pc-linux-gnu
Configured with: /repo/gcc-trunk//configure --enable-languages=c,c++
--enable-valgrind-annotations --disable-nls --enable-checking=yes,rtl,df,extra
--disable-bootstrap --with-cloog --with-ppl --with-isl
--build=x86_64-pc-linux-gnu --host=x86_64-pc-linux-gnu
--target=x86_64-pc-linux-gnu --with-ld=/usr/bin/x86_64-pc-linux-gnu-ld
--with-as=/usr/bin/x86_64-pc-linux-gnu-as --disable-libstdcxx-pch
--prefix=/repo/gcc-trunk//binary-trunk-r12-1366-20210610095751-g6fcba9ef23e-checking-yes-rtl-df-extra-nobootstrap-amd64
Thread model: posix
Supported LTO compression algorithms: zlib zstd
gcc version 12.0.0 20210610 (experimental) (GCC)

[Bug target/101023] [11/12 Regression] wrong code with -mstackrealign

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101023

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |11.2

[Bug target/101023] [11/12 Regression] wrong code with -mstackrealign

[ubizjak at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101023

Uroš Bizjak <ubizjak at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |hjl.tools at gmail dot com

--- Comment #1 from Uroš Bizjak <ubizjak at gmail dot com> ---
CC added.

[Bug target/101023] [11/12 Regression] wrong code with -mstackrealign

[hjl.tools at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101023

H.J. Lu <hjl.tools at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2021-06-11
             Status|UNCONFIRMED                 |WAITING

--- Comment #2 from H.J. Lu <hjl.tools at gmail dot com> ---
(In reply to Zdenek Sojka from comment #0)
> Created attachment 50981 [details]
> auto-reduced testcase (from OpenTTD sources)
> 
> Compiler output:
> $ x86_64-pc-linux-gnu-g++ -O2 -mtune=opteron -mstackrealign
> --param=hot-bb-frequency-fraction=1 testcase.C -S
> 
> The offending code is:
> .L8:
> 	pushq	%rbp
> 	.cfi_def_cfa_offset 16
> 	.cfi_offset 6, -16
> 	movq	%rsp, %rbp
> 	.cfi_def_cfa_register 6
> 	movq	%r12, -8(%rbp)
> 	pushq	%rax
> 
> the "push" overwrites r12 stored just one instruction above; when returning:
> 	movq	-8(%rbp), %r12
> 	xorl	%eax, %eax
> 	leave
> 	.cfi_def_cfa 7, 8
> 	ret
> 
> the wrong data is restored to r12
> 
> I wasn't able yet to generate an executable testcase, but I can try to if
> needed. (it shouldn't be that hard, just r12 needs to be used by the caller
> during the call)

Please create a run-time testcase.

[Bug target/101023] [11/12 Regression] wrong code with -mstackrealign

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101023

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|WAITING                     |NEW
                 CC|                            |jakub at gcc dot gnu.org

--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Here it is:

// PR target/101023
// { dg-do run }
// { dg-options "-O2 -mtune=opteron -mstackrealign
--param=hot-bb-frequency-fraction=1" }

struct S {
  __attribute__((noipa)) int m1 ();
  __attribute__((noipa)) void m2 ();
};
struct T {
  __attribute__((noipa)) virtual S m3 ();
};
struct U : T {
  int u;
  __attribute__((noipa)) U (int);
};
int *a;
S *b;
int c;

int
S::m1 ()
{
  return 0;
}

void
S::m2 ()
{
}

S
T::m3 ()
{
  return S ();
}

U::U (int) : u (4)
{
}

__attribute__((noipa)) int
foo ()
{
  if (a)
    return 0;
  U d(c);
  S *e = b;
  e->m2 ();
  return e->m1();
}

int
main ()
{
  register int r12 __asm ("r12") = 1;
  register int rax __asm ("rax") = 2;
  asm volatile ("" : "+r" (r12), "+r" (rax));
  foo ();
  asm volatile ("" : "+r" (r12));
  if (r12 != 1)
    __builtin_abort ();
}

[Bug target/101023] [11/12 Regression] wrong code with -mstackrealign

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101023

--- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
// { dg-do run { target lp64 } }
(or ! ia32 if you want).

[Bug target/101023] [11/12 Regression] wrong code with -mstackrealign

[zsojka at seznam dot cz]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101023

--- Comment #5 from Zdenek Sojka <zsojka at seznam dot cz> ---
(In reply to Jakub Jelinek from comment #3)
> Here it is:
> 

Thank you, I probably wouldn't be able to get back to this until next week.

[Bug target/101023] [11/12 Regression] wrong code with -mstackrealign

[hjl.tools at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101023

H.J. Lu <hjl.tools at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Assignee|unassigned at gcc dot gnu.org      |hjl.tools at gmail dot com

--- Comment #6 from H.J. Lu <hjl.tools at gmail dot com> ---
Created attachment 50986
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=50986&action=edit
A patch

[Bug target/101023] [11/12 Regression] wrong code with -mstackrealign

[hjl.tools at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101023

H.J. Lu <hjl.tools at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |patch

--- Comment #7 from H.J. Lu <hjl.tools at gmail dot com> ---
A patch is posted at

https://gcc.gnu.org/pipermail/gcc-patches/2021-June/572596.html

[Bug target/101023] [11/12 Regression] wrong code with -mstackrealign

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101023

--- Comment #8 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by H.J. Lu <hjl@gcc.gnu.org>:

https://gcc.gnu.org/g:3f04e3782536ad2f9cfbb8cfe6630e9f9dd8af4c

commit r12-1409-g3f04e3782536ad2f9cfbb8cfe6630e9f9dd8af4c
Author: H.J. Lu <hjl.tools@gmail.com>
Date:   Fri Jun 11 07:31:29 2021 -0700

    x86: Replace ix86_red_zone_size with ix86_red_zone_used

    Add red_zone_used to machine_function to track if red zone is used.
    When expanding function prologue, set red_zone_used to true if red
    zone is used.

    gcc/

            PR target/101023
            * config/i386/i386.c (ix86_expand_prologue): Set red_zone_used
            to true if red zone is used.
            (ix86_output_indirect_jmp): Replace ix86_red_zone_size with
            ix86_red_zone_used.
            * config/i386/i386.h (machine_function): Add red_zone_used.
            (ix86_red_zone_size): Removed.
            (ix86_red_zone_used): New.
            * config/i386/i386.md (peephole2 patterns): Replace
            ix86_red_zone_size with ix86_red_zone_used.

    gcc/testsuite/

            PR target/101023
            * g++.target/i386/pr101023a.C: New test.
            * g++.target/i386/pr101023b.C: Likewise.

[Bug target/101023] [11/12 Regression] wrong code with -mstackrealign

[hjl.tools at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101023

--- Comment #9 from H.J. Lu <hjl.tools at gmail dot com> ---
This bug is exposed by r11-508.

[Bug target/101023] [11/12 Regression] wrong code with -mstackrealign

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101023

--- Comment #10 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-11 branch has been updated by H.J. Lu <hjl@gcc.gnu.org>:

https://gcc.gnu.org/g:ba3b30cf70990f4aadb393b64b76ff490bb0bd0f

commit r11-8755-gba3b30cf70990f4aadb393b64b76ff490bb0bd0f
Author: H.J. Lu <hjl.tools@gmail.com>
Date:   Fri Jun 11 07:31:29 2021 -0700

    x86: Replace ix86_red_zone_size with ix86_red_zone_used

    Add red_zone_used to machine_function to track if red zone is used.
    When expanding function prologue, set red_zone_used to true if red
    zone is used.

    gcc/

            PR target/101023
            * config/i386/i386.c (ix86_expand_prologue): Set red_zone_used
            to true if red zone is used.
            (ix86_output_indirect_jmp): Replace ix86_red_zone_size with
            ix86_red_zone_used.
            * config/i386/i386.h (machine_function): Add red_zone_used.
            (ix86_red_zone_size): Removed.
            (ix86_red_zone_used): New.
            * config/i386/i386.md (peephole2 patterns): Replace
            ix86_red_zone_size with ix86_red_zone_used.

    gcc/testsuite/

            PR target/101023
            * g++.target/i386/pr101023a.C: New test.
            * g++.target/i386/pr101023b.C: Likewise.

    (cherry picked from commit 3f04e3782536ad2f9cfbb8cfe6630e9f9dd8af4c)

[Bug target/101023] [11/12 Regression] wrong code with -mstackrealign

[hjl.tools at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101023

H.J. Lu <hjl.tools at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |FIXED
             Status|NEW                         |RESOLVED

--- Comment #11 from H.J. Lu <hjl.tools at gmail dot com> ---
Fixed for GCC 12 and 11.2.