On Fri, Mar 29, 2019 at 7:15 AM Christopher Lameter <cl@xxxxxxxxx> wrote:
>
> On Thu, 28 Mar 2019, akpm@xxxxxxxxxxxxxxxxxxxx wrote:
>
> > This implies that calls to kmem_cache_*alloc on a SLAB_CACHE_DMA32
> > kmem_cache must _not_ use GFP_DMA32 (it is anyway redundant and
> > unnecessary).
>
> I hope you verified that this flag cannot be passed to kmem_cache_alloc()
> and friends?
As highlighted in the commit message, just above the block that you
quote, we keep GFP_DMA32 in GFP_SLAB_BUG_MASK, so, yes,
kmem_cache_alloc with GFP_DMA32 will trigger a warning, nothing new
compared to previous behaviour:
We do not create a DMA32 kmalloc cache array, as there are currently
no users of kmalloc(..., GFP_DMA32). These calls will continue to
trigger a warning, as we keep GFP_DMA32 in GFP_SLAB_BUG_MASK.
(I very likely verified this 4 months ago when developing this series,
but now I can't remember for sure, I can test this again next week).
