Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Thu, 14 Mar 2013 22:08:15 +0100
From: magnum <john.magnum@...hmail.com>
To: john-dev@...ts.openwall.com
Subject: Re: clock err in timers (alarm, status ...) and fix

On 14 Mar, 2013, at 19:37 , Solar Designer <solar@...nwall.com> wrote:
> On Thu, Mar 14, 2013 at 05:40:53PM +0100, magnum wrote:
>> I applied Costin's changes almost verbatim. Is that not good enough?
> 
> The patch looked buggy to me at first glance (where are the timer_*
> variables initialized in accordance with their new meaning? nowhere?
> is the re-initialization via the first += good enough?), but it might
> actually be OK.  What type are those variables, though?  Anyhow, if
> they're at least 32-bit and we're talking seconds, the wraparound won't
> occur before 2038.

I was puzzled too at first glance of it but the used function returns number of wall clock seconds since start. The existing initializations etc. need no changes. The variables are unsigned int. I think that simple patch was all there is to it, possibly except this: If a --max-runtime session runs to end and then is resumed, I bet it will immediately abort again since total run time is restored from the session file. I will check that and fix it if needed.

> Overall, I'd be happier if jumbo only touched the aspects related to its
> added features, not changed core features for no or little reason.

The OpenCL AMD problem with OS_TIMER increases the need for Jumbo. The save interval is not super critical but as long as status_get_time() is called anyway, I think we should use it.

> Calling status_get_time() every second is a bit wasteful.  Perhaps only
> do that when auto-status or auto-abort is requested?
> 
> Perhaps not bother with that for auto-save?  Or only bother with it for
> auto-save when OS_TIMER is 0 (that is, when the signal handler
> invocation frequency is expected to be very rather than just slightly
> imprecise)?

Those are all valid points. I will benchmark the impact and have look at the alternatives. 

magnum

Powered by blists - more mailing lists

Your e-mail address:

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