[openfabrics-ewg] Re: [openib-general] OpenSM segmentation fault on RC5

Hal Rosenstock halr at voltaire.com
Sat May 27 03:08:35 PDT 2006


Hi Paul,

On Sat, 2006-05-27 at 02:26, Paul wrote:
> Hi Hal,
>       My lab is undergoing maitanence this weekend so I wont be able
> to get you any results til tuesday, however the results are readily
> reproducable. Everything is 64bit.

Unfortunately I don't have access to a PPC64 machine on which to do this
myself. I wish I did. So can you help next week ? Thanks.

-- Hal

> Regards.
> 
> On 26 May 2006 12:46:01 -0400, Hal Rosenstock <halr at voltaire.com>
> wrote:
>         Hi again Paul,
>         
>         On Fri, 2006-05-26 at 12:14, Paul wrote:
>         > No, I figured all of that out, ppc64 was not
>         supported/working in RC4.
>         > Either way, here is what I see with opensm:
>         >
>         > [root at something ~]# /etc/init.d/opensmd start
>         > *** glibc detected *** realloc(): invalid next size:
>         > 0x00000000100ab1e0 ***
>         > /etc/init.d/opensmd: line 330: 7854 Done echo $PORT_FLAG
>         > 7855 Aborted | $prog $START_FLAGS >/dev/null 2>&1 
>         > opensm start [FAILED]
>         > [root at something ~]#
>         
>         OK; that's a totally different problem than Don's. I would
>         like to get
>         to the bottom of this.
>         
>         0x100ab1e0 is a pretty big size. Is this reproducible ? 
>         
>         I'm not sure how realloc gets called as I do not believe
>         OpenSM calls it
>         directly (or any of its libraries).
>         
>         Are you using 32 or 64 bit libraries for this ?
>         
>         Would you rebuild OpenSM with debug:
>         ./configure --enable-debug && make clean && make && make
>         install
>         
>         and then run opensm under gdb and provide the backtrace after
>         the
>         failure?
>         
>         Thanks.
>         
>         -- Hal
>         
> 




More information about the general mailing list