Messages in this thread | | | Subject | Re: [RFC PATCH 0/3] Fix SLQB on memoryless configurations V2 | From | Benjamin Herrenschmidt <> | Date | Tue, 22 Sep 2009 10:00:03 +1000 |
| |
On Mon, 2009-09-21 at 17:10 +0100, Mel Gorman wrote: > > It needs signed-off from the powerpc side because it's now allocating > more > memory potentially (Ben?). An alternative to this patch is in V1 that > statically declares the per-node structures but this is potentially > sub-optimal but from a performance and memory utilisation perspective.
So if I understand correctly, we have a problem with both cpu-less and memory-less nodes. Interesting setups :-)
I have no strong objection on the allocating of the per-cpu data for the cpu-less nodes. However, I wonder if we should do that a bit more nicely, maybe with some kind of "adjusted" cpu_possible_mask() (could be something like cpu_node_valid_mask or similar) to be used by percpu.
Mostly because it would be nice to have built-in debug features in per-cpu and in that case, it would need some way to know a valid number from an invalid one). Either that or just keep track of the mask of cpus that had percpu data allocated to them
Cheers, Ben.
| |