Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-Id: <BADA5DB8-F088-4D6A-9140-84381C6BDD77@patpro.net>
Date: Wed, 14 Sep 2016 08:45:08 +0200
From: patpro@...pro.net
To: john-users@...ts.openwall.com
Subject: Re: possible memory leak on FreeBSD?

Hello,

On 10 sept. 2016, at 10:09, Solar Designer <solar@...nwall.com> wrote:

> On Sat, Sep 10, 2016 at 08:45:14AM +0200, Patrick Proniewski wrote:
>> I'll try the latest bleeding jumbo with --enable-memdbg as soon as possible (in the next few day probably) and keep you updated with the result.
> 
> If you're getting a lot of passwords cracked as I had suspected, then it
> is fully expected that memory usage will grow a lot when running with
> "--fork".  So there's no need for you to debug this, we're aware of it.
> OTOH, if you see similar behavior without "--fork" or/and with no or few
> successful cracks, then that could be worth looking into.


You are absolutely right. I've managed to reproduce the problem even in wordlist mode. As soon as I'm cracking "enough" hashes per seconds, memory consumption raise dramatically.


On 10 sept. 2016, at 16:57, Solar Designer <solar@...nwall.com> wrote:

> On Sat, Sep 10, 2016 at 01:17:53PM +0200, Patrick Proniewski wrote:
>> The strange thing is that `top` does not reflect this increase in memory, or at least this increase doesn't add up to the memory consumption I'm seeing. But may be I'm missing something here.
> 
> Since the (expected) increase in memory usage is through reduction of
> sharing across 4 processes, it is not surprising that top's reporting
> may appear off.

If I understand correctly, here's what happens. If I'm starting with:

  PID USERNAME    THR PRI NICE   SIZE    RES STATE   C   TIME    WCPU COMMAND
 4243 patpro        1 123   20  3514M  3164M CPU4    4  64:10 100.00% john
 5527 patpro        1 123   20  3514M  3156M CPU5    5  60:56 100.00% john
 5528 patpro        1 123   20  3514M  3171M CPU3    3  60:56 100.00% john
 5526 patpro        1 123   20  3514M  3156M CPU7    7  60:56 100.00% john

then john processes are sharing about 3.5 GB RAM -> total RAM used by john ~ 3.5 GB. But during cracking session the content of memory used by each process diverges and even though the SIZE/RES figures stay about the same, I might have only 2 GB shared and 1.5 non-shared per process, resulting in a total memory usage of 8 GB.

I've ordered 16 GB RAM (second hand, because no longer available...). That should help.

Thanks again for taking the time to explain the memory use of John.

patpro

Powered by blists - more mailing lists

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.