stack and memory problems

stack and memory problems

Post by Frank Slootw » Tue, 24 Oct 1995 04:00:00

  Yes, maxssiz is a kernel parameter. See the "System Parameters"
appendix of the "System Administration Tasks" manual.

Quote:> The machine has beau coup swap space ...

  Make sure that the "tot" line of the "swapinfo -t" output says that
you still have sufficient swap space. The other lines are difficult to
interpret correctly, and incorrect interpretation normally leads to
incorrect conclusions.

stack and memory problems

Post by Bill Hasse » Fri, 27 Oct 1995 04:00:00

: One of our programs here gives the following error when running:

: Pid XXXX received a SIGSEGV for stack growth failure.
: Possible causes: insufficient memory or swap space,
: or stack size exceeded maxssiz.
: Memory fault (coredump)

: We've played with the program to lessen the stack requirements but we're
: really up against a wall now.  I think that I may need to increase the value
: of maxssiz, but is this a kernel parameter or what?  The machine has beau coup
: swap space and 500MB of ram.  Any help is appreciated.

  The kernel parameter maxssize is set to about 64 MB by default.  Use
  SAM to change it.  See also maxtsize and maxdsize.  


          /     \
         / ^   ^ \
        ( (o) (o) )
         \       /
|   Bill Hassell            |  Hewlett-Packard Response Center    |

      .oooO   (   )
       (   )    ) /   "There are two types of computer users in the
        \ (    (_/     world...those that have lost data, and those
         \_)           that are going to."        (blh, circa 1972)


1. Memory leak detection / stack unwind HP-UX

I've been looking into using a number of open source memory leak
detection tools on HP-UX.  I've found a few tools that work great under
Linux and have managed to hack at least one to produce useful output in
HP-UX (we're using HP-UX 11i).  Are there any known open source tools
people are using?

Ideally I want to track calls to malloc/free in an executable.  The
executable is produced by us, but getting hold of the source and
recompiling, say with Purify, is way too difficult (joys of working for
large multi-site company).  It is relatively easy to override libc in
the original exe and produce a new libc library that overides malloc and
links the original libc.  I see LD_PRELOAD is now supported anyway - but
not tried this yet.

I can keep a tally on allocated/freed memory which is useful, but I
could really do with knowing the return address when the debug malloc is
called, i.e. where the memory is being allocated in calling program.  Is
this possible?

Searching back through time only throws up U_STACK_TRACE(), but this
calls malloc (IIRC) and will end up being called recursively!

If I get really stuck I guess I could force the debug malloc code to
SEGV on some sort of detection threshold, e.g. after x bytes leaked and
use gdb on the core dump???


2. Seeking Benchmarks for vxWorks

3. stack growth failure (X11R6 problem)

4. Autoregister Multiple IP-addresses in DNS

5. HP10.10 - Still have problem with stack growth failure!

6. Rebadged Telebits

7. HP10.10 - Problem with stack growth failure

8. Solidworks News Server

9. Stack size problem

10. Stack/HP-UX Problems

11. Stack trace (unwinding) problems.

12. stack size problem!

13. stack size problem