Re: [PATCH] percpu: fix unit_map[] verification in pcpu_setup_first_chunk()

[Tejun Heo <tj@kernel.org>]

Tejun Heo wrote:
> Christoph Lameter wrote:
>> On Tue, 29 Sep 2009, Tejun Heo wrote:
>>
>>> pcpu_setup_first_chunk() incorrectly used NR_CPUS as the impossible
>>> unit number while unit number can equal and go over NR_CPUS with
>>> sparse unit map.  This triggers BUG_ON() spuriously on machines which
>>> have non-power-of-two number of cpus.  Use UINT_MAX instead.
>> Uhhh. Funky. The assumption nr_cpu_ids < NR_CPUS has been broken. Wonder
>> what other effects that will have. In particular since the slab and page
>> allocators have arrays indexed by the cpu number and those arrays are
>> dimensioned for NR_CPUS.
> 
> Heh.. that's not what's broken.  unit# can legally go over NR_CPUS or
> nr_cpu_ids as cpus can be sparsely mapped to units.  The problem is
> that
      ^
      I'm delusional at the moment.  :-)

Christoph, can you please elaborate why nr_cpu_ids < NR_CPUS is
broken?

Thanks.

-- 
tejun