linux-kernel - Re: Why do we still have 32 bit counters? Interrupt counters overflow within 50 days

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Mon, 6 Oct 2014 17:28:47 +0200 (CEST)
From:	Thomas Gleixner <tglx@...utronix.de>
To:	Christoph Lameter <cl@...ux.com>
cc:	Richard Cochran <richardcochran@...il.com>,
	linux-kernel@...r.kernel.org
Subject: Re: Why do we still have 32 bit counters? Interrupt counters overflow
 within 50 days

On Mon, 6 Oct 2014, Christoph Lameter wrote:

> On Mon, 6 Oct 2014, Thomas Gleixner wrote:
> 
> > So if you want to fix that as well, you really need to think about the
> > 32 bit case because there is no serialization for the interrupts which
> > are delivered directly from their own vector. And no, we should not
> > diverge 32 and 64 bit artificially here simply because the same 50
> > days wrap applies to both.
> 
> Is it a divergence if both 64bit and 32 bit are unsing unsigned long?

Sigh, yes. Because unsigned long is 32bit on a 32bit architecture. So
the change would be NOP for 32bit and 32bit would still suffer from
the wrap arounds etc.
 
> >
> > I really start to wonder whether all this is worth the trouble. It has
> > been this way forever and 1k timer interrupts per second is not really
> > a new thing either. So we did not change anything which suddenly makes
> > tools confused.
> 
> Tools expect the number of interrupt to increase linearly and not jump by
> 2^32 once in awhile. There are functions in the kernel (/proc/stat) that
> sum up various interrupt counters and that are types unsigned long. These
> larger numbers can suddenly jump by 2^32. Its pretty unusual for a 64 bit
> conter to do that and it requires some head scratching until we figured
> that one out.

I understand that, I just wonder why nobody noticed before. It's been
that way forever :)

Thanks,

	tglx

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/