[PATCH v2 0/4] add support for platform persistence capabilities

[PATCH v2 0/4] add support for platform persistence capabilities

[Dave Jiang]

ACPI 6.2a provides an NFIT sub-table that informs if the platform has
auto CPU flush and memory flush on unexpected power loss events. This series
propogates those attributes to nd_region and add sysfs attribute to show
those capabilities.

---

v2:
Per Dan's comments
- Added ADR cap flags propogation
- Added sysfs attribute

Dave Jiang (4):
      acpi: nfit: Add support for detect platform CPU cache flush on power loss
      acpi: nfit: add persistent memory control flag for nd_region
      libnvdimm: expose platform persistence attribute for nd_region
      nfit-test: Add platform cap support from ACPI 6.2a to test


 drivers/acpi/nfit/core.c         |   23 +++++++++++++++++++++++
 drivers/acpi/nfit/nfit.h         |    1 +
 drivers/nvdimm/pmem.c            |    4 +++-
 drivers/nvdimm/region_devs.c     |   14 ++++++++++++++
 include/linux/libnvdimm.h        |   11 +++++++++++
 tools/testing/nvdimm/test/nfit.c |   11 ++++++++++-
 6 files changed, 62 insertions(+), 2 deletions(-)

--
_______________________________________________
Linux-nvdimm mailing list
Linux-nvdimm@lists.01.org
https://lists.01.org/mailman/listinfo/linux-nvdimm

[PATCH v2 1/4] acpi: nfit: Add support for detect platform CPU cache flush on power loss

[Dave Jiang]

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
to set the property on per region from the NFTI tables.

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@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 = pcap->highest_capability - 1 + pcap->highest_capability;
+	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)
 {
@@ -883,6 +895,10 @@ static void *add_table(struct acpi_nfit_desc *acpi_desc,
 	case ACPI_NFIT_TYPE_SMBIOS:
 		dev_dbg(dev, "%s: smbios\n", __func__);
 		break;
+	case ACPI_NFIT_TYPE_CAPABILITIES:
+		if (!add_platform_cap(acpi_desc, table))
+			return err;
+		break;
 	default:
 		dev_err(dev, "unknown table '%d' parsing nfit\n", hdr->type);
 		break;
@@ -2649,6 +2665,9 @@ static int acpi_nfit_register_region(struct acpi_nfit_desc *acpi_desc,
 	else
 		ndr_desc->numa_node = NUMA_NO_NODE;
 
+	if(acpi_desc->platform_cap & ACPI_NFIT_CAPABILITY_CACHE_FLUSH)
+		set_bit(ND_REGION_PERSIST_CACHE, &ndr_desc->flags);
+
 	list_for_each_entry(nfit_memdev, &acpi_desc->memdevs, list) {
 		struct acpi_nfit_memory_map *memdev = nfit_memdev->memdev;
 		struct nd_mapping_desc *mapping;
@@ -3457,6 +3476,7 @@ static __init int nfit_init(void)
 	BUILD_BUG_ON(sizeof(struct acpi_nfit_smbios) != 9);
 	BUILD_BUG_ON(sizeof(struct acpi_nfit_control_region) != 80);
 	BUILD_BUG_ON(sizeof(struct acpi_nfit_data_region) != 40);
+	BUILD_BUG_ON(sizeof(struct acpi_nfit_capabilities) != 16);
 
 	guid_parse(UUID_VOLATILE_MEMORY, &nfit_uuid[NFIT_SPA_VOLATILE]);
 	guid_parse(UUID_PERSISTENT_MEMORY, &nfit_uuid[NFIT_SPA_PM]);
diff --git a/drivers/acpi/nfit/nfit.h b/drivers/acpi/nfit/nfit.h
index f0cf18b2da8b..50d36e166d70 100644
--- a/drivers/acpi/nfit/nfit.h
+++ b/drivers/acpi/nfit/nfit.h
@@ -202,6 +202,7 @@ struct acpi_nfit_desc {
 	unsigned long dimm_cmd_force_en;
 	unsigned long bus_cmd_force_en;
 	unsigned long bus_nfit_cmd_force_en;
+	unsigned int platform_cap;
 	int (*blk_do_io)(struct nd_blk_region *ndbr, resource_size_t dpa,
 			void *iobuf, u64 len, int rw);
 };
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) &
+		!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;
+
 	nd_device_register(dev);
 
 	return nd_region;
diff --git a/include/linux/libnvdimm.h b/include/linux/libnvdimm.h
index f8109ddb5ef1..f2fc0da4da04 100644
--- a/include/linux/libnvdimm.h
+++ b/include/linux/libnvdimm.h
@@ -47,6 +47,11 @@ enum {
 
 	/* region flag indicating to direct-map persistent memory by default */
 	ND_REGION_PAGEMAP = 0,
+	/*
+	 * Platform ensures entire CPU store data path is flushed to pmem on
+	 * system power loss.
+	 */
+	ND_REGION_PERSIST_CACHE = 1,
 
 	/* mark newly adjusted resources as requiring a label update */
 	DPA_RESOURCE_ADJUSTED = 1 << 0,

_______________________________________________
Linux-nvdimm mailing list
Linux-nvdimm@lists.01.org
https://lists.01.org/mailman/listinfo/linux-nvdimm

Re: [PATCH v2 1/4] acpi: nfit: Add support for detect platform CPU cache flush on power loss

[Ross Zwisler]

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@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;
_______________________________________________
Linux-nvdimm mailing list
Linux-nvdimm@lists.01.org
https://lists.01.org/mailman/listinfo/linux-nvdimm

[PATCH v2 2/4] acpi: nfit: add persistent memory control flag for nd_region

[Dave Jiang]

Propogate the ADR attribute flag from the NFIT platform capabilities
sub-table to nd_region.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
---
 drivers/acpi/nfit/core.c  |    3 +++
 include/linux/libnvdimm.h |    6 ++++++
 2 files changed, 9 insertions(+)

diff --git a/drivers/acpi/nfit/core.c b/drivers/acpi/nfit/core.c
index c08b3da61b93..435222944950 100644
--- a/drivers/acpi/nfit/core.c
+++ b/drivers/acpi/nfit/core.c
@@ -2668,6 +2668,9 @@ static int acpi_nfit_register_region(struct acpi_nfit_desc *acpi_desc,
 	if(acpi_desc->platform_cap & ACPI_NFIT_CAPABILITY_CACHE_FLUSH)
 		set_bit(ND_REGION_PERSIST_CACHE, &ndr_desc->flags);
 
+	if (acpi_desc->platform_cap & ACPI_NFIT_CAPABILITY_MEM_FLUSH)
+		set_bit(ND_REGION_PERSIST_MEMCTRL, &ndr_desc->flags);
+
 	list_for_each_entry(nfit_memdev, &acpi_desc->memdevs, list) {
 		struct acpi_nfit_memory_map *memdev = nfit_memdev->memdev;
 		struct nd_mapping_desc *mapping;
diff --git a/include/linux/libnvdimm.h b/include/linux/libnvdimm.h
index f2fc0da4da04..ff855ed965fb 100644
--- a/include/linux/libnvdimm.h
+++ b/include/linux/libnvdimm.h
@@ -52,6 +52,12 @@ enum {
 	 * system power loss.
 	 */
 	ND_REGION_PERSIST_CACHE = 1,
+	/*
+	 * Platform provides mechanisms to automatically flush outstanding
+	 * write data from memory controler to pmem on system power loss.
+	 * (ADR)
+	 */
+	ND_REGION_PERSIST_MEMCTRL = 2,
 
 	/* mark newly adjusted resources as requiring a label update */
 	DPA_RESOURCE_ADJUSTED = 1 << 0,

_______________________________________________
Linux-nvdimm mailing list
Linux-nvdimm@lists.01.org
https://lists.01.org/mailman/listinfo/linux-nvdimm

Re: [PATCH v2 2/4] acpi: nfit: add persistent memory control flag for nd_region

[Ross Zwisler]

On Fri, Dec 08, 2017 at 02:00:14PM -0700, Dave Jiang wrote:
> Propogate the ADR attribute flag from the NFIT platform capabilities
> sub-table to nd_region.
> 
> Signed-off-by: Dave Jiang <dave.jiang@intel.com>

Sure, looks fine.

Reviewed-by: Ross Zwisler <ross.zwisler@linux.intel.com>
_______________________________________________
Linux-nvdimm mailing list
Linux-nvdimm@lists.01.org
https://lists.01.org/mailman/listinfo/linux-nvdimm

[PATCH v2 3/4] libnvdimm: expose platform persistence attribute for nd_region

[Dave Jiang]

Providing a sysfs attribute for nd_region that shows the persistence
capabilities for the platform.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
---
 drivers/nvdimm/region_devs.c |   13 +++++++++++++
 1 file changed, 13 insertions(+)

diff --git a/drivers/nvdimm/region_devs.c b/drivers/nvdimm/region_devs.c
index 87e8cd10de05..211639d836c8 100644
--- a/drivers/nvdimm/region_devs.c
+++ b/drivers/nvdimm/region_devs.c
@@ -528,6 +528,18 @@ static ssize_t resource_show(struct device *dev,
 }
 static DEVICE_ATTR_RO(resource);
 
+static ssize_t persistence_domain_show(struct device *dev,
+		struct device_attribute *attr, char *buf)
+{
+	struct nd_region *nd_region = to_nd_region(dev);
+	unsigned long flags = nd_region->flags;
+
+	return sprintf(buf, "%s%s\n",
+			flags & BIT(ND_REGION_PERSIST_CACHE) ? "cpu_cache " : "",
+			flags & BIT(ND_REGION_PERSIST_MEMCTRL) ? "memory_controller " : "");
+}
+static DEVICE_ATTR_RO(persistence_domain);
+
 static struct attribute *nd_region_attributes[] = {
 	&dev_attr_size.attr,
 	&dev_attr_nstype.attr,
@@ -543,6 +555,7 @@ static struct attribute *nd_region_attributes[] = {
 	&dev_attr_init_namespaces.attr,
 	&dev_attr_badblocks.attr,
 	&dev_attr_resource.attr,
+	&dev_attr_persistence_domain.attr,
 	NULL,
 };
 

_______________________________________________
Linux-nvdimm mailing list
Linux-nvdimm@lists.01.org
https://lists.01.org/mailman/listinfo/linux-nvdimm

Re: [PATCH v2 3/4] libnvdimm: expose platform persistence attribute for nd_region

[Ross Zwisler]

On Fri, Dec 08, 2017 at 02:00:20PM -0700, Dave Jiang wrote:
> Providing a sysfs attribute for nd_region that shows the persistence
> capabilities for the platform.
> 
> Signed-off-by: Dave Jiang <dave.jiang@intel.com>

Sure, looks fine.

Reviewed-by: Ross Zwisler <ross.zwisler@linux.intel.com>
_______________________________________________
Linux-nvdimm mailing list
Linux-nvdimm@lists.01.org
https://lists.01.org/mailman/listinfo/linux-nvdimm

[PATCH v2 4/4] nfit-test: Add platform cap support from ACPI 6.2a to test

[Dave Jiang]

Adding NFIT platform capabilities sub table in nfit_test simulated ACPI
NFIT table. Only the first NFIT table is added with the capability
sub-table.

Signed-off-by: Dave Jiang <dave.jiang@intel.com>
---
 tools/testing/nvdimm/test/nfit.c |   11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)

diff --git a/tools/testing/nvdimm/test/nfit.c b/tools/testing/nvdimm/test/nfit.c
index 7217b2b953b5..8a16f90a9573 100644
--- a/tools/testing/nvdimm/test/nfit.c
+++ b/tools/testing/nvdimm/test/nfit.c
@@ -993,6 +993,7 @@ static void nfit_test0_setup(struct nfit_test *t)
 	struct acpi_nfit_control_region *dcr;
 	struct acpi_nfit_data_region *bdw;
 	struct acpi_nfit_flush_address *flush;
+	struct acpi_nfit_capabilities *pcap;
 	unsigned int offset, i;
 
 	/*
@@ -1500,8 +1501,16 @@ static void nfit_test0_setup(struct nfit_test *t)
 	for (i = 0; i < NUM_HINTS; i++)
 		flush->hint_address[i] = t->flush_dma[3] + i * sizeof(u64);
 
+	/* platform capabilities */
+	pcap = nfit_buf + offset + flush_hint_size * 4;
+	pcap->header.type = ACPI_NFIT_TYPE_CAPABILITIES;
+	pcap->header.length = sizeof(*pcap);
+	pcap->highest_capability = BIT(2);
+	pcap->capabilities = ACPI_NFIT_CAPABILITY_CACHE_FLUSH |
+		ACPI_NFIT_CAPABILITY_MEM_FLUSH;
+
 	if (t->setup_hotplug) {
-		offset = offset + flush_hint_size * 4;
+		offset = offset + sizeof(*pcap);
 		/* dcr-descriptor4: blk */
 		dcr = nfit_buf + offset;
 		dcr->header.type = ACPI_NFIT_TYPE_CONTROL_REGION;

_______________________________________________
Linux-nvdimm mailing list
Linux-nvdimm@lists.01.org
https://lists.01.org/mailman/listinfo/linux-nvdimm

Re: [PATCH v2 4/4] nfit-test: Add platform cap support from ACPI 6.2a to test

[Ross Zwisler]

On Fri, Dec 08, 2017 at 02:00:25PM -0700, Dave Jiang wrote:
> Adding NFIT platform capabilities sub table in nfit_test simulated ACPI
> NFIT table. Only the first NFIT table is added with the capability
> sub-table.
> 
> Signed-off-by: Dave Jiang <dave.jiang@intel.com>
> ---
>  tools/testing/nvdimm/test/nfit.c |   11 ++++++++++-
>  1 file changed, 10 insertions(+), 1 deletion(-)
> 
> diff --git a/tools/testing/nvdimm/test/nfit.c b/tools/testing/nvdimm/test/nfit.c
> index 7217b2b953b5..8a16f90a9573 100644
> --- a/tools/testing/nvdimm/test/nfit.c
> +++ b/tools/testing/nvdimm/test/nfit.c
> @@ -993,6 +993,7 @@ static void nfit_test0_setup(struct nfit_test *t)
>  	struct acpi_nfit_control_region *dcr;
>  	struct acpi_nfit_data_region *bdw;
>  	struct acpi_nfit_flush_address *flush;
> +	struct acpi_nfit_capabilities *pcap;
>  	unsigned int offset, i;
>  
>  	/*
> @@ -1500,8 +1501,16 @@ static void nfit_test0_setup(struct nfit_test *t)
>  	for (i = 0; i < NUM_HINTS; i++)
>  		flush->hint_address[i] = t->flush_dma[3] + i * sizeof(u64);
>  
> +	/* platform capabilities */
> +	pcap = nfit_buf + offset + flush_hint_size * 4;
> +	pcap->header.type = ACPI_NFIT_TYPE_CAPABILITIES;
> +	pcap->header.length = sizeof(*pcap);
> +	pcap->highest_capability = BIT(2);
> +	pcap->capabilities = ACPI_NFIT_CAPABILITY_CACHE_FLUSH |
> +		ACPI_NFIT_CAPABILITY_MEM_FLUSH;
> +
>  	if (t->setup_hotplug) {
> -		offset = offset + flush_hint_size * 4;
> +		offset = offset + sizeof(*pcap);

You still need to increase the offset by flush_hint_size*4, in addition to
sizeof(*pcap):

	offset = offset + flush_hint_size * 4 + sizeof(*pcap);

Also, you need to increase the size of the memory allocation for t->nfit_buf
in nfit_test_alloc() by adding in a sizeof(struct acpi_nfit_capabilities) to
the calculation of nfit_size.  Otherwise you'll increment offset off the end
of nfit_buf in nfit_test0_setup().

>  		/* dcr-descriptor4: blk */
>  		dcr = nfit_buf + offset;
>  		dcr->header.type = ACPI_NFIT_TYPE_CONTROL_REGION;
> 
> _______________________________________________
> Linux-nvdimm mailing list
> Linux-nvdimm@lists.01.org
> https://lists.01.org/mailman/listinfo/linux-nvdimm
_______________________________________________
Linux-nvdimm mailing list
Linux-nvdimm@lists.01.org
https://lists.01.org/mailman/listinfo/linux-nvdimm