Re: MD5SUMs differ across cluster nodes: Any ideas?
Michael Clark <[email protected]>
| Newsgroups | gmane.user-groups.linux.singapore |
|---|---|
| Message-ID | <[email protected]> |
Hi Junhao,
On Friday 09,April,2010 09:53 AM, Junhao wrote:
> A couple of us dug further. It seems that dynamically linked binaries
> have different md5sums, while static binaries have the same. Still not
> quite sure why though...
>
> Regards,
> Junhao
If the systems had prelinking (popular in fedora a few years ago), all
of the the shared objects get rewritten with non-conflicting link
addresses (to avoid runtime relocation overhead). It is usually set up
in /etc/cron (check cron.daily, cron.weekly, etc)
I'm not sure if more recent linux systems use prelinking or not - it may
conflict with link address randomisation (a security mechanism, so that
exploits can't run code that targets known address offsets). e.g. look
at the link addresses of /bin/ls over two runs (on Debian Sid/2.6.32):
mclark@monty:~$ ldd /bin/ls
libselinux.so.1 => /lib/libselinux.so.1 (0xb777f000)
librt.so.1 => /lib/i686/cmov/librt.so.1 (0xb7776000)
libacl.so.1 => /lib/libacl.so.1 (0xb776e000)
libc.so.6 => /lib/i686/cmov/libc.so.6 (0xb7627000)
libdl.so.2 => /lib/i686/cmov/libdl.so.2 (0xb7623000)
/lib/ld-linux.so.2 (0xb77b6000)
libpthread.so.0 => /lib/i686/cmov/libpthread.so.0 (0xb760a000)
libattr.so.1 => /lib/libattr.so.1 (0xb7605000)
mclark@monty:~$ ldd /bin/ls
libselinux.so.1 => /lib/libselinux.so.1 (0xb788a000)
librt.so.1 => /lib/i686/cmov/librt.so.1 (0xb7881000)
libacl.so.1 => /lib/libacl.so.1 (0xb7879000)
libc.so.6 => /lib/i686/cmov/libc.so.6 (0xb7732000)
libdl.so.2 => /lib/i686/cmov/libdl.so.2 (0xb772e000)
/lib/ld-linux.so.2 (0xb78c1000)
libpthread.so.0 => /lib/i686/cmov/libpthread.so.0 (0xb7715000)
libattr.so.1 => /lib/libattr.so.1 (0xb7710000)
_______________________________________________
LUGS Mailing list - [email protected]
List FAQ: http://wiki.lugs.org.sg/LugsMailingListFaq
Info page: http://www.lugs.org.sg/mailman/listinfo/slugnet
To unsubscribe send an empty email to: [email protected]