Re: [PATCH 04/13] libnvdimm, nfit: move flush hint mapping to dimm driver

["Kani, Toshimitsu" <toshi.kani@hpe.com>]

On Sat, 2016-06-04 at 13:52 -0700, Dan Williams wrote:
 :
> diff --git a/drivers/nvdimm/region_devs.c b/drivers/nvdimm/region_devs.c
> index 420e1a5e2250..5b6f85d00bb5 100644
> --- a/drivers/nvdimm/region_devs.c
> +++ b/drivers/nvdimm/region_devs.c
> @@ -803,11 +803,29 @@ EXPORT_SYMBOL_GPL(nvdimm_volatile_region_create);
>   */
>  void nvdimm_flush(struct nd_region *nd_region)
>  {
> +	int i;
> +
>  	/*
> -	 * TODO: replace wmb_pmem() usage with flush hint writes where
> -	 * available.
> +	 * The first wmb() is needed to 'sfence' all previous writes
> +	 * such that they are architecturally visible for the platform
> +	 * buffer flush.  Note that we've already arranged for pmem
> +	 * writes to avoid the cache via arch_memcpy_to_pmem().  The
> +	 * final wmb() ensures ordering for the NVDIMM flush write.
>  	 */
> -	wmb_pmem();
> +	wmb();
> +	for (i = 0; i < nd_region->ndr_mappings; i++) {
> +		struct nd_mapping *nd_mapping = &nd_region->mapping[i];
> +		struct nvdimm_drvdata *ndd = to_ndd_unlocked(nd_mapping);
> +
> +		/*
> +		 * Note, nvdimm_drvdata guaranteed to be live since we
> +		 * arrange for all associated regions to be disabled
> +		 * before the dimm is disabled.
> +		 */
> +		if (ndd->flush_wpq[0])

Hi Dan,

ndd is NULL on our system since nvdimm_probe() always fails
in nvdimm_init_nsarea(), which then calls put_ndd() and
nvdimm_drvdata_release().

Thanks,
-Toshi