2.1.37-pre-7-crash-me also crashes final 2.1.37

Wolfgang Wander (wwc@lars.desy.de)
Wed, 14 May 1997 11:17:58 +0200

Wolfgang Wander writes:
> Hi,
>
> the following program reproducibly crashes pre-7 while trying to
> generate a (large) core dump on an ext2 filesystem (53c7,8xx).
>
> pre7-crash.c----------------------------------------
> #include <stdlib.h>
> main()
> {
> char *array = calloc(1,32000000); /* alloc and touch me */
> array[36000000L] = 255; /* but --don't touch me here */
> return 0; /* just pretent to be polite */
> }
> ----------------------------------------------------

[configure and other info deleted]

Yesterday posted crashme also crashes (SMP-)2.1.37-final during core
generation. It is however important to generate a core size in the
order of about half the physical memory. The above program did not
crash any of our 128MB machines but only the 64MB machine.

A modified crashme.c:
----------------------------------------------------------------------
#include <stdlib.h>
main( int argc, char *argv[] )
{
int SIZE = 1000000;
char *array;
if( argc == 2 )
SIZE=atoi(argv[1]);
array = calloc(1,SIZE); /* alloc and touch me */
array[SIZE+32000] = 255; /* don't touch me here */
return 0; /* just pretent to be polite */
}
----------------------------------------------------------------------

which you should start with an argument of incrementing size:
gcc -o crashme crashme.c

crashme 16000000
crashme 24000000
crashme 32000000
crashme 48000000
crashme 64000000
...

At some time the SMP machine either freezes or panics with any panic
from fs/buffer.c

Can someone try to confirm this, please?

Wolfgang

--
   _/  _/ _/  _/ _/_/_/ . Wolfgang Wander HERMES Collaboration 
  _/  _/ _/  _/ _/     . DESY Hamburg    Email: Wolfgang.Wander@desy.de
 _/_/_/ _/_/_/ _/     . Notkestr. 85    Tel: +49 40 8998 4638  Fax: -4034
_/_/_/ _/_/_/ _/_/_/ . 22603 Hamburg   http://www-hermes.desy.de/wander.html