linux-kernel - Re: [PATCH] x86/pci: don't use crs for root if we only have one root bus

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 for Android: free password hash cracker in your pocket

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

Date:	Wed, 24 Jun 2009 17:01:14 -0700
From:	Jesse Barnes <jbarnes@...tuousgeek.org>
To:	Linus Torvalds <torvalds@...ux-foundation.org>
Cc:	Yinghai Lu <yinghai@...nel.org>, Ingo Molnar <mingo@...e.hu>,
	Gary Hade <garyhade@...ibm.com>,
	Matthew Wilcox <matthew@....cx>,
	Larry Finger <Larry.Finger@...inger.net>,
	Andrew Morton <akpm@...ux-foundation.org>,
	linux-pci@...r.kernel.org,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
	Jaswinder Singh Rajput <jaswinder@...nel.org>
Subject: Re: [PATCH] x86/pci: don't use crs for root if we only have one
 root bus

On Wed, 24 Jun 2009 16:54:08 -0700 (PDT)
Linus Torvalds <torvalds@...ux-foundation.org> wrote:
> On Wed, 24 Jun 2009, Jesse Barnes wrote:
> > 
> > Yeah, I think it's reasonable to revert, especially given how we do
> > _CRS handling currently.  I'm hoping at some point we can use the
> > _CRS data to at least augment the configuration we get from
> > hardware, since on some machines it seems to be necessary.
> 
> Agreed. I do think we should take _CRS into account - possibly just
> as a minimal hint to determine which root buses to try to scan (maybe
> we do this already, I really didn't check). Or maybe we could use it
> to extend on our scan information.
> 
> But when it seems to have things like "this bus can forward VGA
> cycles" kind of "resources" (which seems to be the main reason Larry
> Finger has so many of them), then that's just worthless knowledge
> that we're much better off just determining on our own.

Yeah, some of those bits don't seem useful; but OTOH if a given bridge
decodes a certain range in a non-configurable way how else will we get
that info w/o having huge tables of per-chip info?  Those are the sorts
of things I worry about with our current resource handling.  Maybe not
directly _CRS related, but still...

> Anyway, I may feel pretty strongly about things like this, but I'm
> also open to being convinced otherwise for 2.6.32. I wanted to do
> -rc1 today (it's been more than two weeks), and while I don't expect
> -rc1 to be flawless, I also hate to release it with _known_ bugs.

Sure, we'll keep at it and see if we can get something better in place
for .32.

> So partly due to timing, I'd rather revert it, and we can revisit it
> for the next release - whatever the actual end result then will be.
> 
> [ There's a difference between "we're supposed to find and fix bugs
> in the -rc series", and "I release known-buggy -rc1's since we're
> supposed to fix it later". For similar reasons, I hate pulling
> known-buggy stuff during the merge window - it's ok if it shows
> itself to be buggy _later_, but if people send me stuff that they
> know is buggy as they send it to me, then that's a problem. ]

Yeah, 100% agreed.  I didn't hear any reports until after people
started using your tree, so I think this case was handled correctly:
push something that *seems* ok upstream, but with eyes wide open for
the possibility we'd need to revert.

Thanks,
-- 
Jesse Barnes, Intel Open Source Technology Center
--
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/