[PATCH v2 0/6] Correctness fixes for NFIT BLK I/O path

[PATCH v2 0/6] Correctness fixes for NFIT BLK I/O path

[Dan Williams]

Changes since v1 [1]:

1/ Fixup the unit test in response to the 4.2-rc1 merge (handle the new
   usage of ioremap_wt).

2/ Add unit test infrastructure for NFIT "flush" sub-table entries

3/ Fix the mock DSM implementation's usage of ENOTTY

4/ Replace usage of devm_ioremap_resource() with devm_ioremap_nocache()
   for mapping flush table addresses.  Multiple DIMMs are allowed to
   share the same flush address.  Also, the side effect of stray writes
   to a flush address are simply extra memory controller buffer flushes,
   so we don't strictly need exclusive resource ownership.

5/ Fixed the warning "unable to guarantee persistence of writes" to
   consider the case of a platform without PMEM API support, but *with*
   flush address entries.

6/ Fixed the "latch" flag retrieval code to proceed with a sane default
   in the case the "DIMM flags" DSM is not implemented.

[1]: https://lists.01.org/pipermail/linux-nvdimm/2015-July/001499.html


Summary:

This set of fixes to the BLK I/O path is a bit larger than I would have
liked for an -rc update, but they are necessary for correct operation of
a BLK-mode namespace.  Originally posted as part of the initial PMEM API
submission at the end of May [2], they got deferred while we iterated on
the details of memremap_pmem(), memcpy_to_pmem(), and wmb_pmem().  Now
that those APIs are upstream these patches finish off enabling
persistence guarantees for BLK-mode persistent memory.

[2]: https://lists.01.org/pipermail/linux-nvdimm/2015-May/000929.html

---

Dan Williams (3):
      tools/testing/nvdimm: mock ioremap_wt
      tools/testing/nvdimm: fix return code for unimplemented commands
      tools/testing/nvdimm: add mock acpi_nfit_flush_address entries to nfit_test

Ross Zwisler (3):
      pmem: add maintainer for include/linux/pmem.h
      nfit: update block I/O path to use PMEM API
      nfit: add support for NVDIMM "latch" flag


 MAINTAINERS                       |    1 
 drivers/acpi/nfit.c               |  130 ++++++++++++++++++++++++++++++++++---
 drivers/acpi/nfit.h               |   20 +++++-
 tools/testing/nvdimm/Kbuild       |    3 +
 tools/testing/nvdimm/test/iomap.c |   27 ++++++++
 tools/testing/nvdimm/test/nfit.c  |   52 ++++++++++++++-
 6 files changed, 217 insertions(+), 16 deletions(-)

[PATCH v2 1/6] pmem: add maintainer for include/linux/pmem.h

[Dan Williams]

From: Ross Zwisler <ross.zwisler@linux.intel.com>

The file include/linux/pmem.h was recently created to hold the PMEM API,
and is logically part of the PMEM driver.  Add an entry for this file to
MAINTAINERS.

Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 MAINTAINERS |    1 +
 1 file changed, 1 insertion(+)

diff --git a/MAINTAINERS b/MAINTAINERS
index 0e6b09150aad..a8306b25c13c 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -6134,6 +6134,7 @@ L:	linux-nvdimm@lists.01.org
 Q:	https://patchwork.kernel.org/project/linux-nvdimm/list/
 S:	Supported
 F:	drivers/nvdimm/pmem.c
+F:	include/linux/pmem.h
 
 LINUX FOR IBM pSERIES (RS/6000)
 M:	Paul Mackerras <paulus@au.ibm.com>

[PATCH v2 2/6] tools/testing/nvdimm: mock ioremap_wt

[Dan Williams]

In the 4.2-rc1 merge the default_memremap_pmem() implementation switched
from ioremap_nocache() to ioremap_wt().  Add it to the list of mocked
routines to restore the ability to run the unit tests.

Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 tools/testing/nvdimm/Kbuild       |    1 +
 tools/testing/nvdimm/test/iomap.c |    6 ++++++
 2 files changed, 7 insertions(+)

diff --git a/tools/testing/nvdimm/Kbuild b/tools/testing/nvdimm/Kbuild
index 8e9b64520ec1..8e020601c773 100644
--- a/tools/testing/nvdimm/Kbuild
+++ b/tools/testing/nvdimm/Kbuild
@@ -1,3 +1,4 @@
+ldflags-y += --wrap=ioremap_wt
 ldflags-y += --wrap=ioremap_cache
 ldflags-y += --wrap=ioremap_nocache
 ldflags-y += --wrap=iounmap
diff --git a/tools/testing/nvdimm/test/iomap.c b/tools/testing/nvdimm/test/iomap.c
index c85a6f6ba559..9f21b150396b 100644
--- a/tools/testing/nvdimm/test/iomap.c
+++ b/tools/testing/nvdimm/test/iomap.c
@@ -77,6 +77,12 @@ void __iomem *__wrap_ioremap_nocache(resource_size_t offset, unsigned long size)
 }
 EXPORT_SYMBOL(__wrap_ioremap_nocache);
 
+void __iomem *__wrap_ioremap_wt(resource_size_t offset, unsigned long size)
+{
+	return __nfit_test_ioremap(offset, size, ioremap_wt);
+}
+EXPORT_SYMBOL(__wrap_ioremap_wt);
+
 void __wrap_iounmap(volatile void __iomem *addr)
 {
 	struct nfit_test_resource *nfit_res;

[PATCH v2 3/6] tools/testing/nvdimm: fix return code for unimplemented commands

[Dan Williams]

The implementation for the new "DIMM Flags" DSM relies on the -ENOTTY
return code to indicate that the flags are unimplimented and to fall
back to a safe default.  As is the -ENXIO error code erroneoously
indicates to fail enabling a BLK region.

Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 tools/testing/nvdimm/test/nfit.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tools/testing/nvdimm/test/nfit.c b/tools/testing/nvdimm/test/nfit.c
index 4b69b8368de0..092d4724fe16 100644
--- a/tools/testing/nvdimm/test/nfit.c
+++ b/tools/testing/nvdimm/test/nfit.c
@@ -155,7 +155,7 @@ static int nfit_test_ctl(struct nvdimm_bus_descriptor *nd_desc,
 	int i, rc;
 
 	if (!nfit_mem || !test_bit(cmd, &nfit_mem->dsm_mask))
-		return -ENXIO;
+		return -ENOTTY;
 
 	/* lookup label space for the given dimm */
 	for (i = 0; i < ARRAY_SIZE(handle); i++)

[PATCH v2 4/6] tools/testing/nvdimm: add mock acpi_nfit_flush_address entries to nfit_test

[Dan Williams]

In preparation for fixing the BLK path to properly use "directed
pcommit" enable the unit test infrastructure to emit mock "flush"
tables.  Writes to these flush addresses trigger a memory controller to
flush its internal buffers to persistent media, similar to the x86
"pcommit" instruction.

Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 tools/testing/nvdimm/Kbuild       |    2 +
 tools/testing/nvdimm/test/iomap.c |   21 ++++++++++++++++
 tools/testing/nvdimm/test/nfit.c  |   50 ++++++++++++++++++++++++++++++++++++-
 3 files changed, 71 insertions(+), 2 deletions(-)

diff --git a/tools/testing/nvdimm/Kbuild b/tools/testing/nvdimm/Kbuild
index 8e020601c773..f56914c7929b 100644
--- a/tools/testing/nvdimm/Kbuild
+++ b/tools/testing/nvdimm/Kbuild
@@ -1,4 +1,6 @@
 ldflags-y += --wrap=ioremap_wt
+ldflags-y += --wrap=ioremap_wc
+ldflags-y += --wrap=devm_ioremap_nocache
 ldflags-y += --wrap=ioremap_cache
 ldflags-y += --wrap=ioremap_nocache
 ldflags-y += --wrap=iounmap
diff --git a/tools/testing/nvdimm/test/iomap.c b/tools/testing/nvdimm/test/iomap.c
index 9f21b150396b..64bfaa50831c 100644
--- a/tools/testing/nvdimm/test/iomap.c
+++ b/tools/testing/nvdimm/test/iomap.c
@@ -65,6 +65,21 @@ void __iomem *__nfit_test_ioremap(resource_size_t offset, unsigned long size,
 	return fallback_fn(offset, size);
 }
 
+void __iomem *__wrap_devm_ioremap_nocache(struct device *dev,
+		resource_size_t offset, unsigned long size)
+{
+	struct nfit_test_resource *nfit_res;
+
+	rcu_read_lock();
+	nfit_res = get_nfit_res(offset);
+	rcu_read_unlock();
+	if (nfit_res)
+		return (void __iomem *) nfit_res->buf + offset
+			- nfit_res->res->start;
+	return devm_ioremap_nocache(dev, offset, size);
+}
+EXPORT_SYMBOL(__wrap_devm_ioremap_nocache);
+
 void __iomem *__wrap_ioremap_cache(resource_size_t offset, unsigned long size)
 {
 	return __nfit_test_ioremap(offset, size, ioremap_cache);
@@ -83,6 +98,12 @@ void __iomem *__wrap_ioremap_wt(resource_size_t offset, unsigned long size)
 }
 EXPORT_SYMBOL(__wrap_ioremap_wt);
 
+void __iomem *__wrap_ioremap_wc(resource_size_t offset, unsigned long size)
+{
+	return __nfit_test_ioremap(offset, size, ioremap_wc);
+}
+EXPORT_SYMBOL(__wrap_ioremap_wc);
+
 void __wrap_iounmap(volatile void __iomem *addr)
 {
 	struct nfit_test_resource *nfit_res;
diff --git a/tools/testing/nvdimm/test/nfit.c b/tools/testing/nvdimm/test/nfit.c
index 092d4724fe16..d0bdae40ccc9 100644
--- a/tools/testing/nvdimm/test/nfit.c
+++ b/tools/testing/nvdimm/test/nfit.c
@@ -128,6 +128,8 @@ struct nfit_test {
 	int num_pm;
 	void **dimm;
 	dma_addr_t *dimm_dma;
+	void **flush;
+	dma_addr_t *flush_dma;
 	void **label;
 	dma_addr_t *label_dma;
 	void **spa_set;
@@ -331,7 +333,8 @@ static int nfit_test0_alloc(struct nfit_test *t)
 			+ sizeof(struct acpi_nfit_system_address) * NUM_SPA
 			+ sizeof(struct acpi_nfit_memory_map) * NUM_MEM
 			+ sizeof(struct acpi_nfit_control_region) * NUM_DCR
-			+ sizeof(struct acpi_nfit_data_region) * NUM_BDW;
+			+ sizeof(struct acpi_nfit_data_region) * NUM_BDW
+			+ sizeof(struct acpi_nfit_flush_address) * NUM_DCR;
 	int i;
 
 	t->nfit_buf = test_alloc(t, nfit_size, &t->nfit_dma);
@@ -356,6 +359,10 @@ static int nfit_test0_alloc(struct nfit_test *t)
 		if (!t->label[i])
 			return -ENOMEM;
 		sprintf(t->label[i], "label%d", i);
+
+		t->flush[i] = test_alloc(t, 8, &t->flush_dma[i]);
+		if (!t->flush[i])
+			return -ENOMEM;
 	}
 
 	for (i = 0; i < NUM_DCR; i++) {
@@ -408,6 +415,7 @@ static void nfit_test0_setup(struct nfit_test *t)
 	struct acpi_nfit_system_address *spa;
 	struct acpi_nfit_control_region *dcr;
 	struct acpi_nfit_data_region *bdw;
+	struct acpi_nfit_flush_address *flush;
 	unsigned int offset;
 
 	nfit_test_init_header(nfit_buf, size);
@@ -831,6 +839,39 @@ static void nfit_test0_setup(struct nfit_test *t)
 	bdw->capacity = DIMM_SIZE;
 	bdw->start_address = 0;
 
+	offset = offset + sizeof(struct acpi_nfit_data_region) * 4;
+	/* flush0 (dimm0) */
+	flush = nfit_buf + offset;
+	flush->header.type = ACPI_NFIT_TYPE_FLUSH_ADDRESS;
+	flush->header.length = sizeof(struct acpi_nfit_flush_address);
+	flush->device_handle = handle[0];
+	flush->hint_count = 1;
+	flush->hint_address[0] = t->flush_dma[0];
+
+	/* flush1 (dimm1) */
+	flush = nfit_buf + offset + sizeof(struct acpi_nfit_flush_address) * 1;
+	flush->header.type = ACPI_NFIT_TYPE_FLUSH_ADDRESS;
+	flush->header.length = sizeof(struct acpi_nfit_flush_address);
+	flush->device_handle = handle[1];
+	flush->hint_count = 1;
+	flush->hint_address[0] = t->flush_dma[1];
+
+	/* flush2 (dimm2) */
+	flush = nfit_buf + offset + sizeof(struct acpi_nfit_flush_address) * 2;
+	flush->header.type = ACPI_NFIT_TYPE_FLUSH_ADDRESS;
+	flush->header.length = sizeof(struct acpi_nfit_flush_address);
+	flush->device_handle = handle[2];
+	flush->hint_count = 1;
+	flush->hint_address[0] = t->flush_dma[2];
+
+	/* flush3 (dimm3) */
+	flush = nfit_buf + offset + sizeof(struct acpi_nfit_flush_address) * 3;
+	flush->header.type = ACPI_NFIT_TYPE_FLUSH_ADDRESS;
+	flush->header.length = sizeof(struct acpi_nfit_flush_address);
+	flush->device_handle = handle[3];
+	flush->hint_count = 1;
+	flush->hint_address[0] = t->flush_dma[3];
+
 	acpi_desc = &t->acpi_desc;
 	set_bit(ND_CMD_GET_CONFIG_SIZE, &acpi_desc->dimm_dsm_force_en);
 	set_bit(ND_CMD_GET_CONFIG_DATA, &acpi_desc->dimm_dsm_force_en);
@@ -933,6 +974,10 @@ static int nfit_test_probe(struct platform_device *pdev)
 				GFP_KERNEL);
 		nfit_test->dimm_dma = devm_kcalloc(dev, num, sizeof(dma_addr_t),
 				GFP_KERNEL);
+		nfit_test->flush = devm_kcalloc(dev, num, sizeof(void *),
+				GFP_KERNEL);
+		nfit_test->flush_dma = devm_kcalloc(dev, num, sizeof(dma_addr_t),
+				GFP_KERNEL);
 		nfit_test->label = devm_kcalloc(dev, num, sizeof(void *),
 				GFP_KERNEL);
 		nfit_test->label_dma = devm_kcalloc(dev, num,
@@ -943,7 +988,8 @@ static int nfit_test_probe(struct platform_device *pdev)
 				sizeof(dma_addr_t), GFP_KERNEL);
 		if (nfit_test->dimm && nfit_test->dimm_dma && nfit_test->label
 				&& nfit_test->label_dma && nfit_test->dcr
-				&& nfit_test->dcr_dma)
+				&& nfit_test->dcr_dma && nfit_test->flush
+				&& nfit_test->flush_dma)
 			/* pass */;
 		else
 			return -ENOMEM;

[PATCH v2 5/6] nfit: update block I/O path to use PMEM API

[Dan Williams]

From: Ross Zwisler <ross.zwisler@linux.intel.com>

Update the nfit block I/O path to use the new PMEM API and to adhere to
the read/write flows outlined in the "NVDIMM Block Window Driver
Writer's Guide":

http://pmem.io/documents/NVDIMM_Driver_Writers_Guide.pdf

This includes adding support for targeted NVDIMM flushes called "flush
hints" in the ACPI 6.0 specification:

http://www.uefi.org/sites/default/files/resources/ACPI_6.0.pdf

For performance and media durability the mapping for a BLK aperture is
moved to a write-combining mapping which is consistent with
memcpy_to_pmem() and wmb_blk().

Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com>
Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 drivers/acpi/nfit.c |   97 +++++++++++++++++++++++++++++++++++++++++++++------
 drivers/acpi/nfit.h |   15 +++++++-
 2 files changed, 100 insertions(+), 12 deletions(-)

diff --git a/drivers/acpi/nfit.c b/drivers/acpi/nfit.c
index a20b7c883ca0..11aa513f2fda 100644
--- a/drivers/acpi/nfit.c
+++ b/drivers/acpi/nfit.c
@@ -18,6 +18,7 @@
 #include <linux/list.h>
 #include <linux/acpi.h>
 #include <linux/sort.h>
+#include <linux/pmem.h>
 #include <linux/io.h>
 #include "nfit.h"
 
@@ -305,6 +306,23 @@ static bool add_idt(struct acpi_nfit_desc *acpi_desc,
 	return true;
 }
 
+static bool add_flush(struct acpi_nfit_desc *acpi_desc,
+		struct acpi_nfit_flush_address *flush)
+{
+	struct device *dev = acpi_desc->dev;
+	struct nfit_flush *nfit_flush = devm_kzalloc(dev, sizeof(*nfit_flush),
+			GFP_KERNEL);
+
+	if (!nfit_flush)
+		return false;
+	INIT_LIST_HEAD(&nfit_flush->list);
+	nfit_flush->flush = flush;
+	list_add_tail(&nfit_flush->list, &acpi_desc->flushes);
+	dev_dbg(dev, "%s: nfit_flush handle: %d hint_count: %d\n", __func__,
+			flush->device_handle, flush->hint_count);
+	return true;
+}
+
 static void *add_table(struct acpi_nfit_desc *acpi_desc, void *table,
 		const void *end)
 {
@@ -338,7 +356,8 @@ static void *add_table(struct acpi_nfit_desc *acpi_desc, void *table,
 			return err;
 		break;
 	case ACPI_NFIT_TYPE_FLUSH_ADDRESS:
-		dev_dbg(dev, "%s: flush\n", __func__);
+		if (!add_flush(acpi_desc, table))
+			return err;
 		break;
 	case ACPI_NFIT_TYPE_SMBIOS:
 		dev_dbg(dev, "%s: smbios\n", __func__);
@@ -389,6 +408,7 @@ static int nfit_mem_add(struct acpi_nfit_desc *acpi_desc,
 {
 	u16 dcr = __to_nfit_memdev(nfit_mem)->region_index;
 	struct nfit_memdev *nfit_memdev;
+	struct nfit_flush *nfit_flush;
 	struct nfit_dcr *nfit_dcr;
 	struct nfit_bdw *nfit_bdw;
 	struct nfit_idt *nfit_idt;
@@ -442,6 +462,14 @@ static int nfit_mem_add(struct acpi_nfit_desc *acpi_desc,
 			nfit_mem->idt_bdw = nfit_idt->idt;
 			break;
 		}
+
+		list_for_each_entry(nfit_flush, &acpi_desc->flushes, list) {
+			if (nfit_flush->flush->device_handle !=
+					nfit_memdev->memdev->device_handle)
+				continue;
+			nfit_mem->nfit_flush = nfit_flush;
+			break;
+		}
 		break;
 	}
 
@@ -978,6 +1006,24 @@ static u64 to_interleave_offset(u64 offset, struct nfit_blk_mmio *mmio)
 	return mmio->base_offset + line_offset + table_offset + sub_line_offset;
 }
 
+static void wmb_blk(struct nfit_blk *nfit_blk)
+{
+
+	if (nfit_blk->nvdimm_flush) {
+		/*
+		 * 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();
+		writeq(1, nfit_blk->nvdimm_flush);
+		wmb();
+	} else
+		wmb_pmem();
+}
+
 static u64 read_blk_stat(struct nfit_blk *nfit_blk, unsigned int bw)
 {
 	struct nfit_blk_mmio *mmio = &nfit_blk->mmio[DCR];
@@ -1012,6 +1058,7 @@ static void write_blk_ctl(struct nfit_blk *nfit_blk, unsigned int bw,
 		offset = to_interleave_offset(offset, mmio);
 
 	writeq(cmd, mmio->base + offset);
+	wmb_blk(nfit_blk);
 	/* FIXME: conditionally perform read-back if mandated by firmware */
 }
 
@@ -1026,7 +1073,6 @@ static int acpi_nfit_blk_single_io(struct nfit_blk *nfit_blk,
 
 	base_offset = nfit_blk->bdw_offset + dpa % L1_CACHE_BYTES
 		+ lane * mmio->size;
-	/* TODO: non-temporal access, flush hints, cache management etc... */
 	write_blk_ctl(nfit_blk, lane, dpa, len, rw);
 	while (len) {
 		unsigned int c;
@@ -1045,13 +1091,19 @@ static int acpi_nfit_blk_single_io(struct nfit_blk *nfit_blk,
 		}
 
 		if (rw)
-			memcpy(mmio->aperture + offset, iobuf + copied, c);
+			memcpy_to_pmem(mmio->aperture + offset,
+					iobuf + copied, c);
 		else
-			memcpy(iobuf + copied, mmio->aperture + offset, c);
+			memcpy_from_pmem(iobuf + copied,
+					mmio->aperture + offset, c);
 
 		copied += c;
 		len -= c;
 	}
+
+	if (rw)
+		wmb_blk(nfit_blk);
+
 	rc = read_blk_stat(nfit_blk, lane) ? -EIO : 0;
 	return rc;
 }
@@ -1124,7 +1176,7 @@ static void nfit_spa_unmap(struct acpi_nfit_desc *acpi_desc,
 }
 
 static void __iomem *__nfit_spa_map(struct acpi_nfit_desc *acpi_desc,
-		struct acpi_nfit_system_address *spa)
+		struct acpi_nfit_system_address *spa, enum spa_map_type type)
 {
 	resource_size_t start = spa->address;
 	resource_size_t n = spa->length;
@@ -1152,8 +1204,15 @@ static void __iomem *__nfit_spa_map(struct acpi_nfit_desc *acpi_desc,
 	if (!res)
 		goto err_mem;
 
-	/* TODO: cacheability based on the spa type */
-	spa_map->iomem = ioremap_nocache(start, n);
+	if (type == SPA_MAP_APERTURE) {
+		/*
+		 * TODO: memremap_pmem() support, but that requires cache
+		 * flushing when the aperture is moved.
+		 */
+		spa_map->iomem = ioremap_wc(start, n);
+	} else
+		spa_map->iomem = ioremap_nocache(start, n);
+
 	if (!spa_map->iomem)
 		goto err_map;
 
@@ -1171,6 +1230,7 @@ static void __iomem *__nfit_spa_map(struct acpi_nfit_desc *acpi_desc,
  * nfit_spa_map - interleave-aware managed-mappings of acpi_nfit_system_address ranges
  * @nvdimm_bus: NFIT-bus that provided the spa table entry
  * @nfit_spa: spa table to map
+ * @type: aperture or control region
  *
  * In the case where block-data-window apertures and
  * dimm-control-regions are interleaved they will end up sharing a
@@ -1180,12 +1240,12 @@ static void __iomem *__nfit_spa_map(struct acpi_nfit_desc *acpi_desc,
  * unbound.
  */
 static void __iomem *nfit_spa_map(struct acpi_nfit_desc *acpi_desc,
-		struct acpi_nfit_system_address *spa)
+		struct acpi_nfit_system_address *spa, enum spa_map_type type)
 {
 	void __iomem *iomem;
 
 	mutex_lock(&acpi_desc->spa_map_mutex);
-	iomem = __nfit_spa_map(acpi_desc, spa);
+	iomem = __nfit_spa_map(acpi_desc, spa, type);
 	mutex_unlock(&acpi_desc->spa_map_mutex);
 
 	return iomem;
@@ -1212,6 +1272,7 @@ static int acpi_nfit_blk_region_enable(struct nvdimm_bus *nvdimm_bus,
 	struct nvdimm_bus_descriptor *nd_desc = to_nd_desc(nvdimm_bus);
 	struct acpi_nfit_desc *acpi_desc = to_acpi_desc(nd_desc);
 	struct nd_blk_region *ndbr = to_nd_blk_region(dev);
+	struct nfit_flush *nfit_flush;
 	struct nfit_blk_mmio *mmio;
 	struct nfit_blk *nfit_blk;
 	struct nfit_mem *nfit_mem;
@@ -1237,7 +1298,8 @@ static int acpi_nfit_blk_region_enable(struct nvdimm_bus *nvdimm_bus,
 	/* map block aperture memory */
 	nfit_blk->bdw_offset = nfit_mem->bdw->offset;
 	mmio = &nfit_blk->mmio[BDW];
-	mmio->base = nfit_spa_map(acpi_desc, nfit_mem->spa_bdw);
+	mmio->base = nfit_spa_map(acpi_desc, nfit_mem->spa_bdw,
+			SPA_MAP_APERTURE);
 	if (!mmio->base) {
 		dev_dbg(dev, "%s: %s failed to map bdw\n", __func__,
 				nvdimm_name(nvdimm));
@@ -1259,7 +1321,8 @@ static int acpi_nfit_blk_region_enable(struct nvdimm_bus *nvdimm_bus,
 	nfit_blk->cmd_offset = nfit_mem->dcr->command_offset;
 	nfit_blk->stat_offset = nfit_mem->dcr->status_offset;
 	mmio = &nfit_blk->mmio[DCR];
-	mmio->base = nfit_spa_map(acpi_desc, nfit_mem->spa_dcr);
+	mmio->base = nfit_spa_map(acpi_desc, nfit_mem->spa_dcr,
+			SPA_MAP_CONTROL);
 	if (!mmio->base) {
 		dev_dbg(dev, "%s: %s failed to map dcr\n", __func__,
 				nvdimm_name(nvdimm));
@@ -1277,6 +1340,17 @@ static int acpi_nfit_blk_region_enable(struct nvdimm_bus *nvdimm_bus,
 		return rc;
 	}
 
+	nfit_flush = nfit_mem->nfit_flush;
+	if (nfit_flush && nfit_flush->flush->hint_count != 0) {
+		nfit_blk->nvdimm_flush = devm_ioremap_nocache(dev,
+				nfit_flush->flush->hint_address[0], 8);
+		if (!nfit_blk->nvdimm_flush)
+			return -ENOMEM;
+	}
+
+	if (!arch_has_pmem_api() && !nfit_blk->nvdimm_flush)
+		dev_warn(dev, "unable to guarantee persistence of writes\n");
+
 	if (mmio->line_size == 0)
 		return 0;
 
@@ -1459,6 +1533,7 @@ int acpi_nfit_init(struct acpi_nfit_desc *acpi_desc, acpi_size sz)
 	INIT_LIST_HEAD(&acpi_desc->dcrs);
 	INIT_LIST_HEAD(&acpi_desc->bdws);
 	INIT_LIST_HEAD(&acpi_desc->idts);
+	INIT_LIST_HEAD(&acpi_desc->flushes);
 	INIT_LIST_HEAD(&acpi_desc->memdevs);
 	INIT_LIST_HEAD(&acpi_desc->dimms);
 	mutex_init(&acpi_desc->spa_map_mutex);
diff --git a/drivers/acpi/nfit.h b/drivers/acpi/nfit.h
index 81f2e8c5a79c..815cb561cdba 100644
--- a/drivers/acpi/nfit.h
+++ b/drivers/acpi/nfit.h
@@ -60,6 +60,11 @@ struct nfit_idt {
 	struct list_head list;
 };
 
+struct nfit_flush {
+	struct acpi_nfit_flush_address *flush;
+	struct list_head list;
+};
+
 struct nfit_memdev {
 	struct acpi_nfit_memory_map *memdev;
 	struct list_head list;
@@ -77,6 +82,7 @@ struct nfit_mem {
 	struct acpi_nfit_system_address *spa_bdw;
 	struct acpi_nfit_interleave *idt_dcr;
 	struct acpi_nfit_interleave *idt_bdw;
+	struct nfit_flush *nfit_flush;
 	struct list_head list;
 	struct acpi_device *adev;
 	unsigned long dsm_mask;
@@ -88,6 +94,7 @@ struct acpi_nfit_desc {
 	struct mutex spa_map_mutex;
 	struct list_head spa_maps;
 	struct list_head memdevs;
+	struct list_head flushes;
 	struct list_head dimms;
 	struct list_head spas;
 	struct list_head dcrs;
@@ -109,7 +116,7 @@ struct nfit_blk {
 	struct nfit_blk_mmio {
 		union {
 			void __iomem *base;
-			void *aperture;
+			void __pmem  *aperture;
 		};
 		u64 size;
 		u64 base_offset;
@@ -123,6 +130,12 @@ struct nfit_blk {
 	u64 bdw_offset; /* post interleave offset */
 	u64 stat_offset;
 	u64 cmd_offset;
+	void __iomem *nvdimm_flush;
+};
+
+enum spa_map_type {
+	SPA_MAP_CONTROL,
+	SPA_MAP_APERTURE,
 };
 
 struct nfit_spa_mapping {

[PATCH v2 6/6] nfit: add support for NVDIMM "latch" flag

[Dan Williams]

From: Ross Zwisler <ross.zwisler@linux.intel.com>

Add support in the NFIT BLK I/O path for the "latch" flag
defined in the "Get Block NVDIMM Flags" _DSM function:

http://pmem.io/documents/NVDIMM_DSM_Interface_Example.pdf

This flag requires the driver to read back the command register after it
is written in the block I/O path.  This ensures that the hardware has
fully processed the new command and moved the aperture appropriately.

Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com>
Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 drivers/acpi/nfit.c |   33 ++++++++++++++++++++++++++++++++-
 drivers/acpi/nfit.h |    5 +++++
 2 files changed, 37 insertions(+), 1 deletion(-)

diff --git a/drivers/acpi/nfit.c b/drivers/acpi/nfit.c
index 11aa513f2fda..628a42c41ab1 100644
--- a/drivers/acpi/nfit.c
+++ b/drivers/acpi/nfit.c
@@ -1059,7 +1059,9 @@ static void write_blk_ctl(struct nfit_blk *nfit_blk, unsigned int bw,
 
 	writeq(cmd, mmio->base + offset);
 	wmb_blk(nfit_blk);
-	/* FIXME: conditionally perform read-back if mandated by firmware */
+
+	if (nfit_blk->dimm_flags & ND_BLK_DCR_LATCH)
+		readq(mmio->base + offset);
 }
 
 static int acpi_nfit_blk_single_io(struct nfit_blk *nfit_blk,
@@ -1266,6 +1268,28 @@ static int nfit_blk_init_interleave(struct nfit_blk_mmio *mmio,
 	return 0;
 }
 
+static int acpi_nfit_blk_get_flags(struct nvdimm_bus_descriptor *nd_desc,
+		struct nvdimm *nvdimm, struct nfit_blk *nfit_blk)
+{
+	struct nd_cmd_dimm_flags flags;
+	int rc;
+
+	memset(&flags, 0, sizeof(flags));
+	rc = nd_desc->ndctl(nd_desc, nvdimm, ND_CMD_DIMM_FLAGS, &flags,
+			sizeof(flags));
+
+	if (rc >= 0 && flags.status == 0)
+		nfit_blk->dimm_flags = flags.flags;
+	else if (rc == -ENOTTY) {
+		/* fall back to a conservative default */
+		nfit_blk->dimm_flags = ND_BLK_DCR_LATCH;
+		rc = 0;
+	} else
+		rc = -ENXIO;
+
+	return rc;
+}
+
 static int acpi_nfit_blk_region_enable(struct nvdimm_bus *nvdimm_bus,
 		struct device *dev)
 {
@@ -1340,6 +1364,13 @@ static int acpi_nfit_blk_region_enable(struct nvdimm_bus *nvdimm_bus,
 		return rc;
 	}
 
+	rc = acpi_nfit_blk_get_flags(nd_desc, nvdimm, nfit_blk);
+	if (rc < 0) {
+		dev_dbg(dev, "%s: %s failed get DIMM flags\n",
+				__func__, nvdimm_name(nvdimm));
+		return rc;
+	}
+
 	nfit_flush = nfit_mem->nfit_flush;
 	if (nfit_flush && nfit_flush->flush->hint_count != 0) {
 		nfit_blk->nvdimm_flush = devm_ioremap_nocache(dev,
diff --git a/drivers/acpi/nfit.h b/drivers/acpi/nfit.h
index 815cb561cdba..79b6d83875c1 100644
--- a/drivers/acpi/nfit.h
+++ b/drivers/acpi/nfit.h
@@ -40,6 +40,10 @@ enum nfit_uuids {
 	NFIT_UUID_MAX,
 };
 
+enum {
+	ND_BLK_DCR_LATCH = 2,
+};
+
 struct nfit_spa {
 	struct acpi_nfit_system_address *spa;
 	struct list_head list;
@@ -131,6 +135,7 @@ struct nfit_blk {
 	u64 stat_offset;
 	u64 cmd_offset;
 	void __iomem *nvdimm_flush;
+	u32 dimm_flags;
 };
 
 enum spa_map_type {