Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.lang.c > #142219

Re: heap memory indexed as an array

Show all headers | View raw

On Thu, 21 Mar 2019 17:23:36 GMT, scott@slp53.sl.home (Scott Lurndal)
wrote:

>nobody@example.org (Scott) writes:
>>On Wed, 20 Mar 2019 16:43:53 GMT, scott@slp53.sl.home (Scott Lurndal)
>>wrote:
>>>Correction, no "standard C" method.   One may always catch SIGSEGV.
>>
>>Agreed, one may do so, *if* the signal is sent. Seg fault happens when
>>you hit memory that isn't yours, but in the case of overcommit, the
>>page *is* yours, it just doesn't have any memory behind it.
>
>Some unix will send SIGBUS in this case, but most will send SIGSEGV.
>
>> My
>>experiences here are brief and not conclusive at all, but when I wrote
>>a test program to really push the issue, the result was crashing the
>>entire OS more often than not, with no (catchable) signal ever sent to
>>my process.
>
>You'd have to point out which OS.   The feature originated, IIRC, with
>IBM's AIX in the 80's.

I previously checked on, IIRC, FreeBSD 8.4-R. I began re-running some
of my tests today on a fresh install of FreeBSD 12.0-R on an i5 PC
with 8Gb RAM and no swap. The test program is just a loop to allocate
a 1Gb block and memset it to a random value, repeatedly, until malloc
returns null or any signal is caught.

My results today were consistent over several runs. When memory was
exhausted, the system logged errors and began killing unrelated jobs,
ending with an uncrashed but unusable system. It was running, had a
login prompt, but logging in (even as root) just kicked you back to
the login prompt without comment. Notably, while my program was among
those killed, it never received a signal, and never got a null return
from malloc.

Curiously, 12.0 comes out of the box with overcommit disabled, and
with that I would expect malloc to eventually return null, but no.
I'll have to look harder to find out why. Probably there are kernel
tunables that need tweaking to prevent non-root jobs from doing that
much damage.

Tomorrow, time permitting, I'll repeat my tests with Ubuntu on the
same hardware, and I think another set of tests that more aggressively
harasses the overcommit mechanism. To be continued....

Back to comp.lang.c | Previous | Next — Previous in thread | Next in thread | Find similar | Unroll thread

Thread

heap memory indexed as an array sinbad <sinbad.sinbad@gmail.com> - 2019-03-19 20:05 -0700
  Re: heap memory indexed as an array Joe Pfeiffer <pfeiffer@cs.nmsu.edu> - 2019-03-19 21:37 -0600
    Re: heap memory indexed as an array Spiros Bousbouras <spibou@gmail.com> - 2019-03-20 04:40 +0000
      Re: heap memory indexed as an array nobody@example.org (Scott) - 2019-03-20 15:56 +0000
        Re: heap memory indexed as an array scott@slp53.sl.home (Scott Lurndal) - 2019-03-20 16:43 +0000
          Re: heap memory indexed as an array fir <profesor.fir@gmail.com> - 2019-03-20 10:00 -0700
          Re: heap memory indexed as an array William Ahern <william@25thandClement.com> - 2019-03-20 15:07 -0700
            Re: heap memory indexed as an array fir is a sexy shemale <cdalten@gmail.com> - 2019-03-20 23:52 -0700
          Re: heap memory indexed as an array nobody@example.org (Scott) - 2019-03-21 16:50 +0000
            Re: heap memory indexed as an array scott@slp53.sl.home (Scott Lurndal) - 2019-03-21 17:23 +0000
              Re: heap memory indexed as an array nobody@example.org (Scott) - 2019-03-22 04:29 +0000
                OT: crashing malloc [was Re: heap memory indexed as an array] nobody@example.org (Scott) - 2019-03-23 15:27 +0000
                Re: heap memory indexed as an array Spiros Bousbouras <spibou@gmail.com> - 2019-03-23 19:24 +0000
                Re: heap memory indexed as an array Melzzzzz <Melzzzzz@zzzzz.com> - 2019-03-23 20:20 +0000
      Re: heap memory indexed as an array Tim Rentsch <tr.17687@z991.linuxsc.com> - 2019-03-21 22:35 -0700
  Re: heap memory indexed as an array Jorgen Grahn <grahn+nntp@snipabacken.se> - 2019-03-20 07:05 +0000
  Re: heap memory indexed as an array Ben Bacarisse <ben.usenet@bsb.me.uk> - 2019-03-20 12:12 +0000

csiph-web