Bug 386795 – app-editors/emacs-23.4-r1 fails to dump on s390 systems

• Status: RESOLVED FIXED
• Product: Gentoo Linux

[Description SpanKY 2011-10-11 15:05:40 UTC]

annoyingly, this seems to only happen on the s390 system.  the s390x system builds emacs-23.3-r2 just fine.

Dumping under the name emacs
**************************************************
Warning: Your system has a gap between BSS and the
heap (5507560 bytes).  This usually means that exec-shield
or something similar is in effect.  The dump may
fail because of this.  See the section about
exec-shield in etc/PROBLEMS for more information.
**************************************************
make[1]: *** [bootstrap-emacs] Segmentation fault (core dumped)

[Comment 1 SpanKY 2011-10-11 15:08:29 UTC]

Created attachment 289573 [details]
build.log

[Comment 2 SpanKY 2011-10-11 15:08:54 UTC]

Created attachment 289575 [details]
emerge --info app-editors/emacs

[Comment 3 Ulrich Müller 2011-10-11 15:57:19 UTC]

Does it build with the recipe from bug 345001 comment #5?

[Comment 4 SpanKY 2011-10-11 16:20:00 UTC]

Created attachment 289603 [details]
config.log

i ran the following ... it did not help ... still crashed:
# echo 0 > /proc/sys/kernel/randomize_va_space
# make

wrt the code, HAVE_PERSONALITY_LINUX32 is defined, so that should have executed.  see attached config.log for more info.

[Comment 5 Ulrich Müller 2011-10-12 06:50:06 UTC]

I guess this needs to be debugged on the platform where it fails, so the Emacs team can't help much.

You could try again with emacs-vcs-24.0.90; the machine dependent configuration has considerably changed between versions 23 and 24. If this also fails then please report the bug upstream.


(In reply to comment #0)
> **************************************************
> Warning: Your system has a gap between BSS and the
> heap (5507560 bytes).

Is this number the same every time?

[Comment 6 SpanKY 2011-10-12 15:58:39 UTC]

hrm, when i tried to write 0 to randomize_va_space, i failed at life.  once i fixed that, the dump worked fine.

i added a call to personality() in the process that got exec-ed and it seems that its flags are back to not having ADD_NO_RANDOMIZE set.

[Comment 7 Raúl Porcel (RETIRED) 2012-08-27 17:33:03 UTC]

FYI happens the same with emacs-24

[Comment 8 SpanKY 2012-08-27 20:45:05 UTC]

i wonder if the patch merged to the s390 kernel (in linux-3.1) fixes this:

    [S390] Do not clobber personality flags on exec
    
    Analog to git commit 59e4c3a2fe9cb1681bb2cff508ff79466f7585ba
    do not clear the additional personality flags on exec. We
    need to inherit the personality bits in PER_MASK across exec.

we're running 2.6.38.x on those vms, and with 3.4.x being the latest LTS, i'll see about updating to that ...

[Comment 9 SpanKY 2012-08-27 23:14:04 UTC]

(In reply to comment #8)

doesn't seem to have helped.  upgraded to 3.4.10 and 23.4-r1 fails same way.

[Comment 10 SpanKY 2012-08-28 22:31:18 UTC]

yeah, turned out to be a kernel bug.  i'll update our kernel again.

[Comment 11 SpanKY 2012-08-29 20:20:09 UTC]

i've applied that patch, rebooted, and emacs looks good now