Re: [PATCH 2/3] mm, bootmem: panic in bootmem alloc functions even if slab is available

[Pekka Enberg <penberg@kernel.org>]

On Fri, Dec 28, 2012 at 12:31 AM, David Rientjes <rientjes@google.com> wrote:
> On Fri, 28 Dec 2012, Pekka Enberg wrote:
>
>> On Sun, 23 Dec 2012, Sasha Levin wrote:
>> >> diff --git a/mm/bootmem.c b/mm/bootmem.c
>> >> index 1324cd7..198a92f 100644
>> >> --- a/mm/bootmem.c
>> >> +++ b/mm/bootmem.c
>> >> @@ -763,9 +763,6 @@ void * __init ___alloc_bootmem_node(pg_data_t *pgdat, unsigned long size,
>> >>  void * __init __alloc_bootmem_node(pg_data_t *pgdat, unsigned long size,
>> >>                                  unsigned long align, unsigned long goal)
>> >>  {
>> >> -     if (WARN_ON_ONCE(slab_is_available()))
>> >> -             return kzalloc_node(size, GFP_NOWAIT, pgdat->node_id);
>> >> -
>> >>       return  ___alloc_bootmem_node(pgdat, size, align, goal, 0);
>> >>  }
>>
>> I'm not sure what Sasha's patch is trying to do here but the fall-back
>> is there simply to let the caller know it's calling the bootmem
>> allocator *too late*. That is, the slab allocator is already up and
>> running so you're expected to use that.
>>
>
> The __alloc_bootmem_node() variant is intended to panic rather than return
> NULL so there are callers that do not check the return value.  I'm
> suggesting rather than removing the fallback to the slab allocator to
> check the return value and panic() here if kzalloc_node() returns NULL.
> The __alloc_bootmem_node_nopanic() variant needs not be changed.

Makes sense. Dropping the fallback completely just makes it more
difficult to find early boot bugs where the bootmem allocator is
called too late.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>