Re: [PATCH] Avoid mask based num_possible_cpus and num_online_cpus -v5

From: Venki Pallipadi
Date: Thu Feb 02 2012 - 16:00:21 EST

On Thu, Feb 2, 2012 at 12:19 PM, Andrew Morton
<akpm@xxxxxxxxxxxxxxxxxxxx> wrote:
> On Fri, 03 Feb 2012 06:33:02 +1030
> Rusty Russell <rusty@xxxxxxxxxxxxxxx> wrote:
>> On Wed, 1 Feb 2012 14:01:25 -0800, Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> wrote:
>> > > Though most of the callers of these two routines are init time (with few
>> > > exceptions of runtime calls), it is cleaner to use variables
>> > > and not go through this repeated mask based calculation.
>> >
>> > Looks good to me.
>> But, I wonder who's asking num_possible_cpus().  It's not a very useful
>> thing to know, though some arch's "know" it's contiguous, so can cheat.
>> Optimizing it seems particularly foolish.
> We're fools for optimisations!

I would think that if we are giving an API people will abuse it sooner
or later :-).

>> We either audit and wean
>> everyone off who's using it incorrectly, or insist on contiguous CPU
>> numbers and drop the mask altogether.
> drivers/block/nvme.c looks like it's assuming a contiguous map.  Maybe
> also drivers/scsi/bnx2fc (wtf?).  I didn't see anything else outside
> arch code.

Yes. I found a bunch of them which seemed obviously wrong. Doing
things like allocating space based on num_possible_cpus() and
accessing the space with get_cpu() or doing cou_online() check etc.


I have a patch to fix these obvious ones. But, there are other users
which were not very obvious to me and also I am know of code in older
kernels (code which since have been rewritten) which can get benefit
of this API change.

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at
Please read the FAQ at