linux-kernel - Re: [PATCH] Use delayed disable mode of ioapic edge triggered interrupts

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:	Tue, 14 Nov 2006 17:17:53 -0800 (PST)
From:	Linus Torvalds <torvalds@...l.org>
To:	"Eric W. Biederman" <ebiederm@...ssion.com>
cc:	Ingo Molnar <mingo@...hat.com>, Komuro <komurojun-mbn@...ty.com>,
	tglx@...utronix.de, Adrian Bunk <bunk@...sta.de>,
	Andrew Morton <akpm@...l.org>,
	Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] Use delayed disable mode of ioapic edge triggered
 interrupts

On Tue, 14 Nov 2006, Eric W. Biederman wrote:
> 
> Hopefully this is the trivial patch that solves the problem.

Ok, having looked more at this, I have to say that the whole 
"IRQ_DELAYED_DISABLE" thing seems very fragile indeed.

It looks like we should do it not only for APIC edge-triggered interrupts, 
but for HT and MSI interrupts too, as far as I can tell (at least they 
also use the "handle_edge_irq" routine)

So I'm wondering how many other cases there are that are missing this.

In that sense, Ingo's patch was a lot safer, although I still dislike it 
for all the other reasons I mentioned - it's simply wrong to re-send a 
level-triggered irq.

I don't know MSI and HT interrupts well enough to tell whether they will 
re-trigger on their own when we unmask them, but the point is, this 
_looks_ like it might be incomplete.

I think part of the problem is a bad interface. We should simply never set 
the IRQ handler on its own. It should be a field in the "irq_chip" 
structure, and we should use _different_ irq chip structures for level and 
edge-triggered. Then we should also add the "flags" thing there, and you 
could do something like

	static struct irq_chip level_ioapic_chip = {
		..

instead of making the insane decision to use the "same" chip for all 
ioapic things.

Ingo? Eric? Comments?

		Linus

-
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/