linux-kernel - Re: [PATCH] x86: Document the hotplug code is incompatible with x86 irq handling

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]

Message-Id: <200706012206.15794.rjw@sisk.pl>
Date:	Fri, 1 Jun 2007 22:06:14 +0200
From:	"Rafael J. Wysocki" <rjw@...k.pl>
To:	"Eric W. Biederman" <ebiederm@...ssion.com>
Cc:	Robert Hancock <hancockr@...w.ca>,
	Andrew Morton <akpm@...ux-foundation.org>,
	Andi Kleen <ak@...e.de>, linux-kernel@...r.kernel.org,
	Neil Brown <neilb@...e.de>, Ingo Molnar <mingo@...e.hu>,
	Zwane Mwaikambo <zwane@...uxpower.ca>
Subject: Re: [PATCH] x86: Document the hotplug code is incompatible with x86 irq handling

On Friday, 1 June 2007 21:48, Eric W. Biederman wrote:
> "Rafael J. Wysocki" <rjw@...k.pl> writes:
> >> Cool.
> >> 
> >> My patch does not change the functionality of the code just complains
> >> very loudly that it is broken.
> >> 
> >> Further the code is broken at a design level.  The code isn't
> >> problematic the code is impossible.  The cpu hotplug code can not be
> >> fixed on x86 without a redesign of the generic cpu hotplug code.
> >> 
> >> Suspend does not need to use cpu hotplug because it already gets in
> >> deep with the drivers, and can stop interrupts at the source.  I know
> >> there was some talk about this doing this earlier, but I don't know if
> >> anything came of that discussion.
> >> 
> >> Regardless if you care about this being a problem feel free to fix the
> >> relevant code so it attempts to do something that the hardware
> >> actually supports.
> >> 
> >> But if the suspend needs this code for smp support it is also broken.
> >
> > Well, from the functionality point of view, it's not.  We have no problems
> > with it, at least not that I know of.
> 
> Luck, or enough other issues someone hasn't tracked their problems
> down to this.  On the pure cpu hotplug path I just got a bug
> report about it not working:  http://lkml.org/lkml/2007/5/31/419
> 
> So apparently real people can hit this one.  Not just theoretical
> people seen by code reviewers.
> 
> The code is broken by design and cannot be made to support existing
> x86 SMP systems.  Please find a way to use something that works.
> 
> The suspend path is already talking to the drivers and can stop IRQs
> at their source so it should not be a big deal.

Very well, but could you _please_ give us some time to do this?

We know of the problem now and will work to fix it, but it's not _that_ easy.

In fact, we also rely on the CPU hotplug's code that takes tasks away from the
offlined CPUs (and does the opposite with respect to the onlined CPUs), so
we just can't get rid of the CPU hotplug _right_ _now_.

Greetings,
Rafael 


-- 
"Premature optimization is the root of all evil." - Donald Knuth
-
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/