> FAIL: ld-aarch64/tls-relax-gdesc-le-now

> regexp_diff match failure
> regexp "^ 0x.+ \(BIND_NOW\) \s+$"
> line   " 0x000000000000001e (FLAGS)              BIND_NOW"
> FAIL: ld-aarch64/tls-relax-gdesc-le-now

This should be harmless (but why it causes a test failure is a mystery to me).

As far as I understand it right now, marking something with DT_BIND_NOW (which the regexp checks for) is an older way, and nowadays binaries are marked with DF_BIND_NOW (which is the value that fails the regexp).

Did the test never work? Or did the code at some point change (i.e. something was unified between arches and aarch switched to the common variant)?

> spawn [open ...]
> regexp_diff match failure
> regexp "^.*:    f9421611        ldr     x17, \[x16, #1064\]$"
> line   "  40028c:       f941fe11        ldr     x17, [x16, #1016]"
> regexp_diff match failure
> regexp "^.*:    9110a210        add     x16, x16, #0x428$"
> line   "  400290:       910fe210        add     x16, x16, #0x3f8"
> regexp_diff match failure
> regexp "^.*:    f9421a11        ldr     x17, \[x16, #1072\]$"
> line   "  4002a8:       f9420211        ldr     x17, [x16, #1024]"
> regexp_diff match failure
> regexp "^.*:    9110c210        add     x16, x16, #0x430$"
> line   "  4002ac:       91100210        add     x16, x16, #0x400"
> FAIL: BTI PLT with only GNU PROP

CC'ing test author

> > FAIL: BTI PLT with only GNU PROP
> 
Tried to find a commit where it started to fail (by building from git), but could not find one (fails all the way back to the introduction of the test).

Assuming it worked in the beginning, the cause for the failure must lie outside binutils.

Hi Andreas,

  These tests appear to be passing now - at least they are for me on the
  2.40 branch and the current mainline.  Do you get the same results ?
  If so, then I think that we can close this PR.

Cheers
  Nick

I just tried with current master (5aea5eca6c873334deb41f996dec255786a6f84d) and still get the same failures. Full build log upload following.

We apply a small patchset but I dont immediately see how it could be relevant.
https://gitweb.gentoo.org/proj/toolchain/binutils-patches.git/tree/9999

FAIL: ld-aarch64/tls-relax-gdesc-le-now
FAIL: BTI PLT with only GNU PROP


PASS: ld-aarch64/tls-relax-gdesc-le
/var/tmp/portage/sys-devel/binutils-9999/work/build/ld/../gas/as-new     -o tmpdir/tls-relax-gdesc-le.o /var/tmp/portage/sys-devel/binutils-9999/work/binutils/ld/testsuite/ld-aarch64/tls-relax-
gdesc-le.s
Executing on host: sh -c {/var/tmp/portage/sys-devel/binutils-9999/work/build/ld/../gas/as-new     -o tmpdir/tls-relax-gdesc-le.o /var/tmp/portage/sys-devel/binutils-9999/work/binutils/ld/tests
uite/ld-aarch64/tls-relax-gdesc-le.s 2>&1}  /dev/null dump.tmp (timeout = 300)
spawn [open ...]
./ld-new --hash-style=sysv -z norelro  -L/var/tmp/portage/sys-devel/binutils-9999/work/binutils/ld/testsuite/ld-aarch64  -shared -z now -o tmpdir/dump tmpdir/tls-relax-gdesc-le.o 
Executing on host: sh -c {./ld-new --hash-style=sysv -z norelro  -L/var/tmp/portage/sys-devel/binutils-9999/work/binutils/ld/testsuite/ld-aarch64  -shared -z now -o tmpdir/dump tmpdir/tls-relax
-gdesc-le.o  2>&1}  /dev/null dump.tmp (timeout = 300)
spawn [open ...]
/var/tmp/portage/sys-devel/binutils-9999/work/build/ld/../binutils/readelf  -dr tmpdir/dump > tmpdir/dump.out
Executing on host: sh -c {/var/tmp/portage/sys-devel/binutils-9999/work/build/ld/../binutils/readelf  -dr tmpdir/dump > tmpdir/dump.out 2>dump.tmp}  /dev/null  (timeout = 300)
spawn [open ...]
regexp_diff match failure
regexp "^ 0x.+ \(BIND_NOW\) \s+$"
line   " 0x000000000000001e (FLAGS)              BIND_NOW"
FAIL: ld-aarch64/tls-relax-gdesc-le-now


PASS: Check -z force-bti emits BTI feature (exec)
/var/tmp/portage/sys-devel/binutils-9999/work/build/ld/../gas/as-new  -mabi=lp64 -defsym __property_bti__=1  -o tmpdir/property-bti-pac1.o /var/tmp/portage/sys-devel/binutils-9999/work/binutils
/ld/testsuite/ld-aarch64/property-bti-pac1.s
Executing on host: sh -c {/var/tmp/portage/sys-devel/binutils-9999/work/build/ld/../gas/as-new  -mabi=lp64 -defsym __property_bti__=1  -o tmpdir/property-bti-pac1.o /var/tmp/portage/sys-devel/b
inutils-9999/work/binutils/ld/testsuite/ld-aarch64/property-bti-pac1.s 2>&1}  /dev/null dump.tmp (timeout = 300)
spawn [open ...]
./ld-new --hash-style=sysv -z norelro  -L/var/tmp/portage/sys-devel/binutils-9999/work/binutils/ld/testsuite/ld-aarch64  -e _start -L./tmpdir -lbti-plt-so -o tmpdir/dump tmpdir/property-bti-pac
1.o 
Executing on host: sh -c {./ld-new --hash-style=sysv -z norelro  -L/var/tmp/portage/sys-devel/binutils-9999/work/binutils/ld/testsuite/ld-aarch64  -e _start -L./tmpdir -lbti-plt-so -o tmpdir/du
mp tmpdir/property-bti-pac1.o  2>&1}  /dev/null dump.tmp (timeout = 300)
spawn [open ...]
/var/tmp/portage/sys-devel/binutils-9999/work/build/ld/../binutils/objdump  -dr -j .plt tmpdir/dump > tmpdir/dump.out
Executing on host: sh -c {/var/tmp/portage/sys-devel/binutils-9999/work/build/ld/../binutils/objdump  -dr -j .plt tmpdir/dump > tmpdir/dump.out 2>dump.tmp}  /dev/null  (timeout = 300)
spawn [open ...]
regexp_diff match failure
regexp "^.*:    f9421611        ldr     x17, \[x16, #1064\]$"
line   "  40028c:       f941fe11        ldr     x17, [x16, #1016]"
regexp_diff match failure
regexp "^.*:    9110a210        add     x16, x16, #0x428$"
line   "  400290:       910fe210        add     x16, x16, #0x3f8"
regexp_diff match failure
regexp "^.*:    f9421a11        ldr     x17, \[x16, #1072\]$"
line   "  4002a8:       f9420211        ldr     x17, [x16, #1024]"
regexp_diff match failure
regexp "^.*:    9110c210        add     x16, x16, #0x430$"
line   "  4002ac:       91100210        add     x16, x16, #0x400"
FAIL: BTI PLT with only GNU PROP

Created attachment 14550 [details]
build log

(In reply to Andreas K. Huettel from comment #5)
> I just tried with current master (5aea5eca6c873334deb41f996dec255786a6f84d)
> and still get the same failures. Full build log upload following.
> 
> We apply a small patchset but I dont immediately see how it could be
> relevant.
> https://gitweb.gentoo.org/proj/toolchain/binutils-patches.git/tree/9999
> 
> FAIL: ld-aarch64/tls-relax-gdesc-le-now
> FAIL: BTI PLT with only GNU PROP
 
I agree that at first glance that patch set does not look like it would cause
these problems.  But it would be nice to have it confirmed.  ie, if you build 
a toolchain without any patches applied, do the failures still occur ?

I suspect that the issue might be cross-builds vs native.  I was running the
tests on an x86_64 Fedora box.  Are you running the tests on native hardware ?

Cheers
  Nick

> > We apply a small patchset but I dont immediately see how it could be
> > relevant.
> > https://gitweb.gentoo.org/proj/toolchain/binutils-patches.git/tree/9999
> > 
> > FAIL: ld-aarch64/tls-relax-gdesc-le-now
> > FAIL: BTI PLT with only GNU PROP
>  
> I agree that at first glance that patch set does not look like it would cause
> these problems.  But it would be nice to have it confirmed.  ie, if you
> build 
> a toolchain without any patches applied, do the failures still occur ?

Same result with the patchset disabled.

> I suspect that the issue might be cross-builds vs native.  I was running the
> tests on an x86_64 Fedora box.  Are you running the tests on native hardware
> ?

Yes, this is on native hardware. 

My main suspicion lies on our gcc hardening patchset, I'm setting up a chroot with a less-hardened gcc right now.

jiji /var/db/repos/gentoo/sys-devel/binutils # uname -a
Linux jiji 5.15.64-gentoo-dist #1 SMP Wed Aug 31 17:57:10 -00 2022 aarch64 GNU/Linux
jiji /var/db/repos/gentoo/sys-devel/binutils # ld.so --version
ld.so (Gentoo 2.36-r5 p5) stable release version 2.36.
Copyright (C) 2022 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.
There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A
PARTICULAR PURPOSE.
jiji /var/db/repos/gentoo/sys-devel/binutils # ld --version
GNU ld (Gentoo 2.39 p5) 2.39.0
Copyright (C) 2022 Free Software Foundation, Inc.
This program is free software; you may redistribute it under the terms of
the GNU General Public License version 3 or (at your option) a later version.
This program has absolutely no warranty.
jiji /var/db/repos/gentoo/sys-devel/binutils # gcc --version
gcc (Gentoo Hardened 11.3.1_p20221209 p3) 11.3.1 20221209
Copyright (C) 2021 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

jiji /var/db/repos/gentoo/sys-devel/binutils # lscpu
Architecture:           aarch64
  CPU op-mode(s):       32-bit, 64-bit
  Byte Order:           Little Endian
CPU(s):                 80
  On-line CPU(s) list:  0-79
Vendor ID:              ARM
  BIOS Vendor ID:       Ampere(R)
  Model name:           Neoverse-N1
    BIOS Model name:    Ampere(R) Altra(R) Processor Q80-30 CPU @ 2.8GHz
    BIOS CPU family:    257
    Model:              1
    Thread(s) per core: 1
    Core(s) per socket: 80
    Socket(s):          1
    Stepping:           r3p1
    Frequency boost:    disabled
    CPU(s) scaling MHz: 54%
    CPU max MHz:        3000.0000
    CPU min MHz:        1000.0000
    BogoMIPS:           50.08
    Flags:              fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm lrcpc dcpop asimddp ssbs
Caches (sum of all):    
  L1d:                  5 MiB (80 instances)
  L1i:                  5 MiB (80 instances)
  L2:                   80 MiB (80 instances)
NUMA:                   
  NUMA node(s):         1
  NUMA node0 CPU(s):    0-79
Vulnerabilities:        
  Itlb multihit:        Not affected
  L1tf:                 Not affected
  Mds:                  Not affected
  Meltdown:             Not affected
  Mmio stale data:      Not affected
  Retbleed:             Not affected
  Spec store bypass:    Mitigation; Speculative Store Bypass disabled via prctl
  Spectre v1:           Mitigation; __user pointer sanitization
  Spectre v2:           Mitigation; CSV2, BHB
  Srbds:                Not affected
  Tsx async abort:      Not affected
jiji /var/db/repos/gentoo/sys-devel/binutils #

( (In reply to Andreas K. Huettel from comment #8)
 
> Same result with the patchset disabled.
 
OK, good.

> Yes, this is on native hardware. 

Either way, it is obvious that the tests need to relax their regexps slightly, so would you like to propose a patch to do this ?  If not, then I will make one myself.

Cheers
  Nick

> 
> My main suspicion lies on our gcc hardening patchset, I'm setting up a
> chroot with a less-hardened gcc right now.
> 

OK, so it's *not* the gcc hardening patchset. Also, same fails with gcc-12. 

>
> Either way, it is obvious that the tests need to relax their regexps
> slightly, so would you like to propose a patch to do this ?  If not, then I
> will make one myself.
> 

On it.

Created attachment 14553 [details]
git patch relaxing the regexps

Here's a git patch that makes the two tests pass here.
It should also work elsewhere (but of course double-checking is better).

The master branch has been updated by Nick Clifton <nickc@sourceware.org>:

https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;h=502c7218da3c486e58d6ce039b36fb4fa62d3a92

commit 502c7218da3c486e58d6ce039b36fb4fa62d3a92
Author: Andreas K. Huettel <dilfridge@gentoo.org>
Date:   Wed Jan 4 09:30:14 2023 +0000

    Fix AArch64 linker testsuite failures trigeered by differences in build environments.
    
            PR 29843
            * testsuite/ld-aarch64/bti-plt-5.d: Relax regxps slightly to allow
            for differences in build environments.
            * testsuite/ld-aarch64/tls-relax-gdesc-le-now.d: Likewise.

The binutils-2_40-branch branch has been updated by Nick Clifton <nickc@sourceware.org>:

https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;h=f4e184b2747542b123cd06c7f9a4fbaaedc1bbca

commit f4e184b2747542b123cd06c7f9a4fbaaedc1bbca
Author: Andreas K. Huettel <dilfridge@gentoo.org>
Date:   Wed Jan 4 09:31:12 2023 +0000

    Fix AArch64 linker testsuite failures triggered by differences in build environments.
    
            PR 29843
            * testsuite/ld-aarch64/bti-plt-5.d: Relax regxps slightly to allow
            for differences in build environments.
            * testsuite/ld-aarch64/tls-relax-gdesc-le-now.d: Likewise.

(In reply to Andreas K. Huettel from comment #11)
> Here's a git patch that makes the two tests pass here.
> It should also work elsewhere (but of course double-checking is better).

It works fine and did not trigger any regressions in my build farm, so I have gone ahead and applied the patch.  (To the mainline and the 2.40 branch).

Thanks for taking the time to work on this issue.

Cheers
  Nick