bug#443: [Fwd: emacs installation - segmentation fault during unexec]

[Sven Joachim <svenjoac@gmx.de>]

On 2008-06-26 09:59 +0200, esf@cfmu.eurocontrol.int wrote:

> Anyway, I tried again with emacs-22.2 and the same problem occurs with
> approximatively the same backtrace, i.e:
>
> #0  0x0026a17c in memcpy () from /lib/tls/libc.so.6
> #1  0x0817d5a6 in unexec (new_name=0x857d414
> "/tmp/esf_emacs-22.2/generated/work/emacs-22.2/src/emacs",
>     old_name=0x857d450
> "/tmp/esf_emacs-22.2/generated/work/emacs-22.2/src/temacs",
> data_start=137491440, bss_start=0, entry_address=0)
>     at unexelf.c:951

Which is your kernel version and what is the value of
/proc/sys/kernel/randomize_va_space?

> Also note that the problem seems to appear randomly so you might have
> to run the installation several times if you need to reproduce the
> problem.

Yes, this is random by nature.  Linux 2.6.25 added a feature called
"brk() randomization" that is enabled if
/proc/sys/kernel/randomize_va_space has the value 2, which is the
default unless you built your kernel with CONFIG_COMPAT_BRK=y.

Setting that value to 1, the default from 2.6.12(?) to 2.6.24, should
make the crashes disappear.  Note that a value of 2 is generally
recommended for security reasons.

See also the section "Dumping" in etc/PROBLEMS for related problems.

Sven