linux-kernel - Re: PCI MSI breaks when booting with nosmp

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:	Wed, 23 Apr 2008 11:23:45 -0700
From:	Jesse Barnes <jbarnes@...tuousgeek.org>
To:	"Maciej W. Rozycki" <macro@...ux-mips.org>
Cc:	Jean Delvare <jdelvare@...e.de>, Andi Kleen <andi@...stfloor.org>,
	linux-pci@...ey.karlin.mff.cuni.cz,
	LKML <linux-kernel@...r.kernel.org>, Pavel Machek <pavel@....cz>,
	Tejun Heo <htejun@...il.com>,
	Tom Long Nguyen <tom.l.nguyen@...el.com>,
	Randy Dunlap <rdunlap@...otime.net>,
	Jeff Garzik <jgarzik@...ox.com>,
	Thomas Gleixner <tglx@...utronix.de>
Subject: Re: PCI MSI breaks when booting with nosmp

On Wednesday, April 23, 2008 11:13 am Maciej W. Rozycki wrote:
> On Wed, 23 Apr 2008, Jesse Barnes wrote:
> > Yeah I think the patch is reasonable, would be good to get feedback from
> > Thomas/Andi/Ingo though...
>
>  FWIW, the original idea behind "nosmp" or "maxcpus=0" (just as an
> implementation detail) vs "maxcpus=1" was that the two formers would
> disable the APIC circuitry altogether (including resisting from switching
> from the PIC compatibility mode on systems supporting it), while the
> latter would still boot UP, but with interrupts routed through the APICs.
> Essentially SMP implied all the MP circuitry/provisions in this context,
> the APICs being an inherent part of which.  Therefore I think the original
> idea of implying "pci=nomsi" with "nosmp" certainly looks more in the
> spirit of the original setup to me.
>
>  However we have "nolapic" these days as well and with this new proposal
> this option could effectively take over the old meaning of "nosmp" (you
> cannot do SMP without the local APIC, so "nolapic nosmp" is redundant).
> I am not entirely convinced it is the right way though...

Yeah, I'm not particularly attached to either meaning.  It looks like we'll 
setup the local apic on 32 bit if the NMI vector is a local apic one, so in 
that case at least the behavior will be the same.

Anyway, we have two options:
  1) make nosmp/maxcpus=1 imply nolapic (and therefore disable MSI too)
  2) make nosmp enable the lapic (so MSI will work)

Jesse
--
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/