Re: [RFC 1/1] memremap: devm_memremap_pages has wrong nid

[Toshi Kani <toshi.kani@hpe.com>]

On Wed, 2015-11-11 at 13:46 -0800, Dan Williams wrote:
> On Wed, Nov 11, 2015 at 1:16 PM, Boaz Harrosh <boaz@plexistor.com> wrote:
> > 
> > The pmem dev as received in devm_memremap_pages() does not have
> > its dev->numa_node properly initialized yet.
> > 
> > What I see is that all pmem devices will call arch_add_memory
> > with nid==0 regardless of the real nid the memory is actually
> > on. Needless to say that after that all the NUMA page and zone
> > members (at page->flags) come out wrong.
> > 
> > If I do the below code it will all work properly.
> > 
> > RFC because probably we want to fix dev->numa_node before the
> > call to devm_memremap_pages.
> 
> Let's just do that instead.  I.e. in the case of NFIT numa node should
> already be set, and in the case of the memmap=ss!nn or e820-type-12 we
> can set the numa node like this:

Agreed.  memory_add_physaddr_to_nid() uses the SRAT info, which does not work
with the NFIT case. 

Thanks,
-Toshi 


> diff --git a/drivers/nvdimm/e820.c b/drivers/nvdimm/e820.c
> index 8282db2ef99e..e40df8fedf73 100644
> --- a/drivers/nvdimm/e820.c
> +++ b/drivers/nvdimm/e820.c
> @@ -48,7 +48,7 @@ static int e820_pmem_probe(struct platform_device *pdev)
>                memset(&ndr_desc, 0, sizeof(ndr_desc));
>                ndr_desc.res = p;
>                ndr_desc.attr_groups = e820_pmem_region_attribute_groups;
> -               ndr_desc.numa_node = NUMA_NO_NODE;
> +               ndr_desc.numa_node = memory_add_physaddr_to_nid(p->start);
>                set_bit(ND_REGION_PAGEMAP, &ndr_desc.flags);
>                if (!nvdimm_pmem_region_create(nvdimm_bus, &ndr_desc))
>                        goto err;
> 
> Thanks for pointing out memory_add_physaddr_to_nid(). I did not know
> that existed.