On Fri, Dec 08, 2017 at 02:00:09PM -0700, Dave Jiang wrote:
In ACPI 6.2a the platform capability structure has been added to the
NFIT
tables. That provides software the ability to determine whether a system
supports the auto flushing of CPU caches on power loss. If the capability
is supported, we do not need to do dax_write_cache(). Plumbing the path
dax_flush() ?
dax_write_cache() just sets or clears DAXDEV_WRITE_CACHE. dax_flush() is the
place where we check DAXDEV_WRITE_CACHE and avoid calling arch_wb_cache_pmem()
if the platform supports a flush-on-fail CPU cache.
to set the property on per region from the NFTI tables.
NFIT
This patch depends on the ACPI NFIT 6.2a platform capabilities support code
in include/acpi/actbl1.h.
Signed-off-by: Dave Jiang <dave.jiang(a)intel.com>
---
drivers/acpi/nfit/core.c | 20 ++++++++++++++++++++
drivers/acpi/nfit/nfit.h | 1 +
drivers/nvdimm/pmem.c | 4 +++-
drivers/nvdimm/region_devs.c | 1 +
include/linux/libnvdimm.h | 5 +++++
5 files changed, 30 insertions(+), 1 deletion(-)
diff --git a/drivers/acpi/nfit/core.c b/drivers/acpi/nfit/core.c
index ff2580e7611d..c08b3da61b93 100644
--- a/drivers/acpi/nfit/core.c
+++ b/drivers/acpi/nfit/core.c
@@ -838,6 +838,18 @@ static bool add_flush(struct acpi_nfit_desc *acpi_desc,
return true;
}
+static bool add_platform_cap(struct acpi_nfit_desc *acpi_desc,
+ struct acpi_nfit_capabilities *pcap)
+{
+ struct device *dev = acpi_desc->dev;
+ u8 mask;
'mask' cannot be a u8, else you'll lose all the upper bits when you compute
it. It needs to be a u32.
+
+ mask = pcap->highest_capability - 1 + pcap->highest_capability;
This calculation is broken. Take the simplest case, highest_capability = 0.
This should translate into a mask of 0x1, so you only look at bit 0, but:
0 - 1 + 0 == -1
giving you a mask that includes all the bits. I think the mask you're looking
for is:
mask = (1 << (pcap->highest_capability + 1)) - 1;
+ acpi_desc->platform_cap = pcap->capabilities &
(u32)mask;
+ dev_dbg(dev, "%s: cap: %#x\n", __func__, acpi_desc->platform_cap);
+ return true;
+}
+
static void *add_table(struct acpi_nfit_desc *acpi_desc,
struct nfit_table_prev *prev, void *table, const void *end)
{
<>
diff --git a/drivers/nvdimm/pmem.c b/drivers/nvdimm/pmem.c
index 7fbc5c5dc8e1..13f2ed80899e 100644
--- a/drivers/nvdimm/pmem.c
+++ b/drivers/nvdimm/pmem.c
@@ -35,6 +35,7 @@
#include "pmem.h"
#include "pfn.h"
#include "nd.h"
+#include "nd-core.h"
static struct device *to_dev(struct pmem_device *pmem)
{
@@ -334,7 +335,8 @@ static int pmem_attach_disk(struct device *dev,
dev_warn(dev, "unable to guarantee persistence of writes\n");
fua = 0;
}
- wbc = nvdimm_has_cache(nd_region);
+ wbc = nvdimm_has_cache(nd_region) &
I'm guessing you meant:
&&
+ !test_bit(ND_REGION_PERSIST_CACHE, &nd_region->flags);
if (!devm_request_mem_region(dev, res->start, resource_size(res),
dev_name(&ndns->dev))) {
diff --git a/drivers/nvdimm/region_devs.c b/drivers/nvdimm/region_devs.c
index abaf38c61220..87e8cd10de05 100644
--- a/drivers/nvdimm/region_devs.c
+++ b/drivers/nvdimm/region_devs.c
@@ -994,6 +994,7 @@ static struct nd_region *nd_region_create(struct nvdimm_bus
*nvdimm_bus,
dev->groups = ndr_desc->attr_groups;
nd_region->ndr_size = resource_size(ndr_desc->res);
nd_region->ndr_start = ndr_desc->res->start;
+
Unrelated whitespace change.
nd_device_register(dev);
return nd_region;