SUMMARY: New processes get killed

Dear SysAdmins,

First of all, thanks for all people that has taken some of their
precious time for helping me:

LD             lexicondevil at gmail dot com
Drew Skinner   dskinner at blueprint dot org
Jason Santos   jason dot santos at megaslow dot net
francisco      frisco at blackant dot net
John dot Vaughan at VerizonWireless dot com
Rich Kulawiec  rsk at gsp dot org
Ed             ed dot rolison at itc dot alstom dot com
Paul Wilkinson Paul_Wilkinson at utsglobal dot com
Mike           mike dot salehi at kodak dot com
Paul Richards  paul at ultra5 dot co dot uk

First, the question:

Em Seg, 2005-01-31 `s 10:44 -0200, Feng Chin Sian escreveu:
> Hello SysAdmins
>
> Does anyone have seen anything like this?
>
>         user> mtamaint
>         Killed
>         user>
>         user> truss /data/local/lib/m400/mtamaint
>         truss: cannot execute program: /data/local/lib/m400/mtamaint
>         user>
>
> When I try o run the program, it simply gets killed. And trying to
> 'truss' it, it simple does not run. The /data is mounted from a share on
> a NAS. Running the same program from another Solaris 9 box, it does run
> sucessfully.
>
> Does anyone has any idea of what is happening?
>
> More information:
>         user> uname -a
>         SunOS icnetedi03 5.9 Generic_117171-02 sun4u sparc
SUNW,Sun-Fire-V240
>         user> echo $SHELL
>         /bin/csh
>         user> ulimit
>         unlimited

Now, what has been done:
- the file permission (execution) has been checked (ok)
- the user ulimits was checked (ulimit -a)
- the swap space was checked (/usr/ccs/bin/size, swap -l)
- truss'ing the file was not possible because the executable refuses to
start (output:
        truss /data/local/lib/m400/mtamaint
        truss: cannot execute program: /data/local/lib/m400/mtamaint
- the mount did not have 'nosuid' nor 'noexec' options, and the
executable did not need suid permissions.
- the dependant libraries were checked. Running

        ldd /data/local/lib/m400/mtamaint

brough the following output:

        ldd /data/local/lib/m400/mtamaint
        ldd: /data/local/lib/m400/mtamaint: execution failed due to signal 9

We have another V240 box that mounts the very same share and it works:

        icnetedi02>ldd /data/local/lib/m400/mtamaint
                        libm4asn1.so =>  /data/local/lib/m400/libm4asn1.so
                        libsocket.so.1 =>        /usr/lib/libsocket.so.1
                        libnsl.so.1 =>   /usr/lib/libnsl.so.1
                        libdl.so.1 =>    /usr/lib/libdl.so.1
                        libm400.so =>    /data/local/lib/m400/libm400.so
                        libc.so.1 =>     /usr/lib/libc.so.1
                        libmp.so.2 =>    /usr/lib/libmp.so.2
                        libm.so.1 =>     /usr/lib/libm.so.1
                        /usr/platform/SUNW,Sun-Fire-V240/lib/libc_psr.so.1

 Then, with the list, we checked if we were able to access those files.
- the hardware info was compared (both are UltrasparcIIIi)
- the instruction sets was compared (isoinfo -v)

Someone asked then why ssh and bash were able to run. Then wen took out
the mtamaint executable from the NAS and saved in the local disk, and
voil`! it ran.

Ispecting the mount options, we come to find the "culprit": the
forcedirectio option. Asking Google, mother of (nearly) all answers, it
returned the following link:
<http://developers.sun.com/solaris/articles/tuning_for_streaming.html>.

Well, our app does not need such feature, so it was taken down, and then
the mtamaint finally executes. Other executables that were in the NAS
come to work now.

--
Feng Sian (feng dot sian at interchange dot com dot br)
Interchange Servigos SA
+55 11 2123 2382
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers