[Bug dynamic-link/27061] New: Incorrect values in _rtld_global_ro._dl_x86_cpu_features

[Bug dynamic-link/27061] New: Incorrect values in _rtld_global_ro._dl_x86_cpu_features

[hjl.tools at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=27061

            Bug ID: 27061
           Summary: Incorrect values in
                    _rtld_global_ro._dl_x86_cpu_features
           Product: glibc
           Version: 2.33
            Status: NEW
          Severity: normal
          Priority: P2
         Component: dynamic-link
          Assignee: unassigned at sourceware dot org
          Reporter: hjl.tools at gmail dot com
  Target Milestone: ---
            Target: i386,x86-64

On AVX2 machine, I got

Breakpoint 1, main () at test.c:40
40        size_t page_size = sysconf(_SC_PAGESIZE);
(gdb) p _rtld_global_ro._dl_x86_cpu_features 
$1 = {basic = {kind = arch_kind_intel, max_cpuid = 22, family = 6, 
    model = 142, stepping = 10}, features = {{cpuid = {eax = 526058, 
        ebx = 34605056, ecx = 2147154879, edx = 3219913727}, usable = {
        eax = 0, ebx = 0, ecx = 2128097795, edx = 394821904}}, {cpuid = {
        eax = 0, ebx = 43804591, ecx = 0, edx = 2617255424}, usable = {
        eax = 0, ebx = 9175848, ecx = 0, edx = 0}}, {cpuid = {eax = 0, 
        ebx = 0, ecx = 289, edx = 739248128}, usable = {eax = 0, ebx = 0, 
        ecx = 289, edx = 134217728}}, {cpuid = {eax = 15, ebx = 960, 
        ecx = 256, edx = 0}, usable = {eax = 7, ebx = 0, ecx = 0, edx = 0}}, {
      cpuid = {eax = 0, ebx = 0, ecx = 0, edx = 256}, usable = {eax = 0, 
        ebx = 0, ecx = 0, edx = 0}}, {cpuid = {eax = 12327, ebx = 0, ecx = 0, 
        edx = 0}, usable = {eax = 0, ebx = 0, ecx = 0, edx = 0}}, {cpuid = {
        eax = 0, ebx = 0, ecx = 0, edx = 0}, usable = {eax = 0, ebx = 0, 
        ecx = 0, edx = 0}}, {cpuid = {eax = 0, ebx = 0, ecx = 0, edx = 0}, 
      usable = {eax = 0, ebx = 0, ecx = 0, edx = 0}}}, preferred = {17719}, 
  xsave_state_size = 960, xsave_state_full_size = 1152, data_cache_size = 0, 
  shared_cache_size = 0, non_temporal_threshold = 0, 
  rep_movsb_threshold = 2048, rep_stosb_threshold = 2048}
(gdb) 

data_cache_size, shared_cache_size, non_temporal_threshold shouldn't be
zero and rep_movsb_threshold should be 4096.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug dynamic-link/27061] Incorrect values in _rtld_global_ro._dl_x86_cpu_features

[hjl.tools at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=27061

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Version|2.33                        |2.32

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug dynamic-link/27061] Incorrect values in _rtld_global_ro._dl_x86_cpu_features

[hjl.tools at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=27061

--- Comment #1 from H.J. Lu <hjl.tools at gmail dot com> ---
(gdb) p __x86_rep_movsb_threshold
$1 = 2048
(gdb) 

It should be 4096.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug dynamic-link/27061] Incorrect values in _rtld_global_ro._dl_x86_cpu_features

[fweimer at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=27061

Florian Weimer <fweimer at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |fweimer at redhat dot com

--- Comment #2 from Florian Weimer <fweimer at redhat dot com> ---
How do you run the test program?

One effect of putting CPU feature data (whose size changes frequently) is
sensitivity to the internal GLIBC_PRIVATE ABI.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug dynamic-link/27061] Incorrect values in __x86_rep_movsb_threshold

[hjl.tools at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=27061

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|Incorrect values in         |Incorrect values in
                   |_rtld_global_ro._dl_x86_cpu |__x86_rep_movsb_threshold
                   |_features                   |

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug dynamic-link/27061] Incorrect values in __x86_rep_movsb_threshold

[hjl.tools at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=27061

--- Comment #3 from H.J. Lu <hjl.tools at gmail dot com> ---
(In reply to Florian Weimer from comment #2)
> How do you run the test program?
> 
> One effect of putting CPU feature data (whose size changes frequently) is
> sensitivity to the internal GLIBC_PRIVATE ABI.

My program is linked against the newly built glibc:

[hjl@gnu-cfl-2 movsb-1]$ ldd test-dynamic 
        linux-vdso.so.1 (0x00007fff9f3ef000)
        libpthread.so.0 =>
/export/build/gnu/tools-build/glibc-cet-gitlab/build-x86_64-linux/nptl/libpthread.so.0
(0x00007efe0312b000)
        libc.so.6 =>
/export/build/gnu/tools-build/glibc-cet-gitlab/build-x86_64-linux/libc.so.6
(0x00007efe02f65000)
       
/export/build/gnu/tools-build/glibc-cet-gitlab/build-x86_64-linux/elf/ld-linux-x86-64.so.2
=> /lib64/ld-linux-x86-64.so.2 (0x00007efe0314e000)
[hjl@gnu-cfl-2 movsb-1]$ 

I have full access to glibc debug info.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug dynamic-link/27061] Incorrect values in __x86_rep_movsb_threshold

[hjl.tools at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=27061

--- Comment #4 from H.J. Lu <hjl.tools at gmail dot com> ---
Fixed on master so far with

commit 2ee7711bdd7de9dd30073b223ce29d5cd50320f6
Author: H.J. Lu <hjl.tools@gmail.com>
Date:   Sun Dec 13 04:56:41 2020 -0800

    x86: Remove the default REP MOVSB threshold tunable value [BZ #27061]

    Since we can't tell if the tunable value is set by user or not:

    https://sourceware.org/bugzilla/show_bug.cgi?id=27069

    remove the default REP MOVSB threshold tunable value so that the correct
    default value will be set correctly by init_cacheinfo ().

    Reviewed-by: Carlos O'Donell <carlos@redhat.com>

-- 
You are receiving this mail because:
You are on the CC list for the bug.