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
