lkml.org 
[lkml]   [2011]   [Mar]   [4]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH x86/mm UPDATED] x86-64, NUMA: Fix distance table handling
Hello,

On Thu, Mar 03, 2011 at 12:07:10PM -0800, David Rientjes wrote:
> The new emulation code needs to remove the assumption that node 0 always
> is online and become more robust since there's no such restriction in the
> ACPI spec.

Yeah, indeed, that's possible. ACPI spec isn't really the deciding
factor here. The node IDs are purely software and a system is
required to have at least one node, so it seems reasonable to require
node 0 to be always online - or even to require all online node IDs to
be consecutive starting from 0. The only reason we may have holes in
node IDs now is because PXM -> nid mapping isn't released for empty
memory-only nodes.

The biggest downside of deferring checks for this type of rare
conditions to upper layers when not strictly necessary is that it
tends to lead to one-off bugs which trigger very infrequently.

That said, maybe we're already too deep toward that direction and it's
much easier to weed out the assumptions that node0 is always online.

Thanks.

--
tejun


\
 
 \ /
  Last update: 2011-03-06 11:19    [W:0.077 / U:19.744 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site