linux-kernel - Re: [origin tree build failure] Re: [PULL] Please pull hwpoison code for 2.6.32

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>] [day] [month] [year] [list]

Date:	Sat, 26 Sep 2009 12:29:43 -0700
From:	"H. Peter Anvin" <hpa@...or.com>
To:	Andi Kleen <andi@...stfloor.org>,
	Linus Torvalds <torvalds@...ux-foundation.org>
CC:	Ingo Molnar <mingo@...e.hu>, fengguang.wu@...el.com,
	linux-kernel@...r.kernel.org, akpm@...ux-foundation.org,
	"H. Peter Anvin" <hpa@...or.com>,
	Thomas Gleixner <tglx@...utronix.de>,
	Peter Zijlstra <a.p.zijlstra@...llo.nl>
Subject: Re: [origin tree build failure] Re: [PULL] Please pull hwpoison
 code for 2.6.32

Okay, I'm missing something (as I said, no access to the code right now), but isn't NUMA irrelevant since NODES_WIDTH == 0 already?  As such, I believe this is a SPARSEMEM issue regardless of NUMA, and tying it to NUMA is meaningless (there are dependencies there, of course, but they are already handled via existing dependency chains.)

    -hpa

Andi Kleen <andi@...stfloor.org> wrote:

>On Sat, Sep 26, 2009 at 09:35:07AM -0700, Linus Torvalds wrote:
>> 
>> 
>> On Sat, 26 Sep 2009, Andi Kleen wrote:
>> > 
>> > Thanks for the report. The issue comes from NODES_SHIFT=4
>> > 
>> > I think I tested the NUMA case, but perhaps not with full NODES_SHIFT.
>> > 
>> > The easy fix would be to limit NODES_SHIFT to 3 for 32bit (8 nodes max). Do you
>> > have any problems with that? I doubt there are any >8 nodes NUMAQs left.
>> > (last time I heard the last machine at IBM was down to < 4)
>> 
>> No, just say "we don't support HWPOISON on 32-bit NUMA". 
>> 
>> Maybe simply using something like this.
>
>That's a reasonable solution, but I would prefer to do it the other way round 
>(MEMORY_FAILURE disabling NUMA) to minimize user confusion. Also it's not
>enough to just do NUMAQ, but needs all of 32bit NUMA, otherwise the NODES_SHIFT
>could be set too high again.
>
>Here's a compile tested patch (don't have a NUMAQ) implementing this:
>
>---
>
>x86: Don't allow 32bit NUMA together with hwpoison to avoid page flags overflow
>
>From: Andi Kleen <ak@...ux.intel.com>
>    
>Otherwise the enlarged sparsemem field in page->flags overflows 32bit,
>which breaks compilation.
>    
>Reported by Ingo Molnar.
>Originally based on a patch from Linus.
>    
>Signed-off-by: Andi Kleen <ak@...ux.intel.com>
>
>diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
>index 9369879..384a897 100644
>--- a/arch/x86/Kconfig
>+++ b/arch/x86/Kconfig
>@@ -423,6 +423,7 @@ config X86_32_NON_STANDARD
> config X86_NUMAQ
> 	bool "NUMAQ (IBM/Sequent)"
> 	depends on X86_32_NON_STANDARD
>+	depends on !MEMORY_FAILURE
> 	select NUMA
> 	select X86_MPPARSE
> 	---help---
>@@ -1108,6 +1109,7 @@ config NUMA
> 	bool "Numa Memory Allocation and Scheduler Support"
> 	depends on SMP
> 	depends on X86_64 || (X86_32 && HIGHMEM64G && (X86_NUMAQ || X86_BIGSMP || X86_SUMMIT && ACPI) && EXPERIMENTAL)
>+	depends on !X86_32 || !MEMORY_FAILURE
> 	default y if (X86_NUMAQ || X86_SUMMIT || X86_BIGSMP)
> 	---help---
> 	  Enable NUMA (Non Uniform Memory Access) support.
>@@ -1125,6 +1127,9 @@ config NUMA
> 
> 	  Otherwise, you should say N.
> 
>+comment "32bit NUMA support requires disabling MEMORY_FAILURE"
>+	depends on X86_32 && MEMORY_FAILURE
>+
> comment "NUMA (Summit) requires SMP, 64GB highmem support, ACPI"
> 	depends on X86_32 && X86_SUMMIT && (!HIGHMEM64G || !ACPI)
>