libnvdimm/namespace: Differentiate between probe mapping and runtime mapping

The nvdimm core currently maps the full namespace to an ioremap range
while probing the namespace mode. This can result in probe failures on
architectures that have limited ioremap space.

For example, with a large btt namespace that consumes most of I/O remap
range, depending on the sequence of namespace initialization, the user
can find a pfn namespace initialization failure due to unavailable I/O
remap space which nvdimm core uses for temporary mapping.

nvdimm core can avoid this failure by only mapping the reserved info
block area to check for pfn superblock type and map the full namespace
resource only before using the namespace.

Given that personalities like BTT can be layered on top of any namespace
type create a generic form of devm_nsio_enable (devm_namespace_enable)
and use it inside the per-personality attach routines. Now
devm_namespace_enable() is always paired with disable unless the mapping
is going to be used for long term runtime access.

Signed-off-by: Aneesh Kumar K.V <[e-mail]>
Link: https://lore.kernel.org/r/[e-mail]
[djbw: reworks to move devm_namespace_{en,dis}able into *attach helpers]
Reported-by: kbuild test robot <[e-mail]>
Link: https://lore.kernel.org/r/[e-mail]
Signed-off-by: Dan Williams <[e-mail]>

drivers/dax/pmem/core.c

···
 	ndns = nvdimm_namespace_common_probe(dev);
 	if (IS_ERR(ndns))
 		return ERR_CAST(ndns);
-	nsio = to_nd_namespace_io(&ndns->dev);
 
 	/* parse the 'pfn' info block via ->rw_bytes */
-	rc = devm_nsio_enable(dev, nsio);
+	rc = devm_namespace_enable(dev, ndns, nd_info_block_reserve());
 	if (rc)
 		return ERR_PTR(rc);
 	rc = nvdimm_setup_pfn(nd_pfn, &pgmap);
 	if (rc)
 		return ERR_PTR(rc);
-	devm_nsio_disable(dev, nsio);
+	devm_namespace_disable(dev, ndns);
 
 	/* reserve the metadata area, device-dax will reserve the data */
 	pfn_sb = nd_pfn->pfn_sb;
 	offset = le64_to_cpu(pfn_sb->dataoff);
+	nsio = to_nd_namespace_io(&ndns->dev);
 	if (!devm_request_mem_region(dev, nsio->res.start, offset,
 				dev_name(&ndns->dev))) {
 		dev_warn(dev, "could not reserve metadata\n");

drivers/nvdimm/btt.c

···
 	struct nd_region *nd_region;
 	struct btt_sb *btt_sb;
 	struct btt *btt;
-	size_t rawsize;
+	size_t size, rawsize;
+	int rc;
 
 	if (!nd_btt->uuid || !nd_btt->ndns || !nd_btt->lbasize) {
 		dev_dbg(&nd_btt->dev, "incomplete btt configuration\n");
···
 	if (!btt_sb)
 		return -ENOMEM;
 
+	size = nvdimm_namespace_capacity(ndns);
+	rc = devm_namespace_enable(&nd_btt->dev, ndns, size);
+	if (rc)
+		return rc;
+
 	/*
 	 * If this returns < 0, that is ok as it just means there wasn't
 	 * an existing BTT, and we're creating a new one. We still need to
···
 	 */
 	nd_btt_version(nd_btt, ndns, btt_sb);
 
-	rawsize = nvdimm_namespace_capacity(ndns) - nd_btt->initial_offset;
+	rawsize = size - nd_btt->initial_offset;
 	if (rawsize < ARENA_MIN_SIZE) {
 		dev_dbg(&nd_btt->dev, "%s must be at least %ld bytes\n",
 				dev_name(&ndns->dev),

drivers/nvdimm/claim.c

···
 	return rc;
 }
 
-int devm_nsio_enable(struct device *dev, struct nd_namespace_io *nsio)
+int devm_nsio_enable(struct device *dev, struct nd_namespace_io *nsio,
+		resource_size_t size)
 {
 	struct resource *res = &nsio->res;
 	struct nd_namespace_common *ndns = &nsio->common;
 
-	nsio->size = resource_size(res);
-	if (!devm_request_mem_region(dev, res->start, resource_size(res),
+	nsio->size = size;
+	if (!devm_request_mem_region(dev, res->start, size,
 				dev_name(&ndns->dev))) {
 		dev_warn(dev, "could not reserve region %pR\n", res);
 		return -EBUSY;
···
 	nvdimm_badblocks_populate(to_nd_region(ndns->dev.parent), &nsio->bb,
 			&nsio->res);
 
-	nsio->addr = devm_memremap(dev, res->start, resource_size(res),
-			ARCH_MEMREMAP_PMEM);
+	nsio->addr = devm_memremap(dev, res->start, size, ARCH_MEMREMAP_PMEM);
 
 	return PTR_ERR_OR_ZERO(nsio->addr);
 }
-EXPORT_SYMBOL_GPL(devm_nsio_enable);
 
 void devm_nsio_disable(struct device *dev, struct nd_namespace_io *nsio)
 {
···
 
 	devm_memunmap(dev, nsio->addr);
 	devm_exit_badblocks(dev, &nsio->bb);
-	devm_release_mem_region(dev, res->start, resource_size(res));
+	devm_release_mem_region(dev, res->start, nsio->size);
 }
-EXPORT_SYMBOL_GPL(devm_nsio_disable);

drivers/nvdimm/namespace_devs.c

···
 }
 EXPORT_SYMBOL(nvdimm_namespace_common_probe);
 
+int devm_namespace_enable(struct device *dev, struct nd_namespace_common *ndns,
+		resource_size_t size)
+{
+	if (is_namespace_blk(&ndns->dev))
+		return 0;
+	return devm_nsio_enable(dev, to_nd_namespace_io(&ndns->dev), size);
+}
+EXPORT_SYMBOL_GPL(devm_namespace_enable);
+
+void devm_namespace_disable(struct device *dev, struct nd_namespace_common *ndns)
+{
+	if (is_namespace_blk(&ndns->dev))
+		return;
+	devm_nsio_disable(dev, to_nd_namespace_io(&ndns->dev));
+}
+EXPORT_SYMBOL_GPL(devm_namespace_disable);
+
 static struct device **create_namespace_io(struct nd_region *nd_region)
 {
 	struct nd_namespace_io *nsio;

drivers/nvdimm/nd-core.h

···
 struct nd_pfn *to_nd_pfn_safe(struct device *dev);
 bool is_nvdimm_bus(struct device *dev);
 
+#if IS_ENABLED(CONFIG_ND_CLAIM)
+int devm_nsio_enable(struct device *dev, struct nd_namespace_io *nsio,
+		resource_size_t size);
+void devm_nsio_disable(struct device *dev, struct nd_namespace_io *nsio);
+#else
+static inline int devm_nsio_enable(struct device *dev,
+		struct nd_namespace_io *nsio, resource_size_t size)
+{
+	return -ENXIO;
+}
+
+static inline void devm_nsio_disable(struct device *dev,
+		struct nd_namespace_io *nsio)
+{
+}
+#endif
+
 #ifdef CONFIG_PROVE_LOCKING
 extern struct class *nd_class;
 

drivers/nvdimm/nd.h

···
 	struct nd_pfn nd_pfn;
 };
 
+static inline u32 nd_info_block_reserve(void)
+{
+	return ALIGN(SZ_8K, PAGE_SIZE);
+}
+
 enum nd_async_mode {
 	ND_SYNC,
 	ND_ASYNC,
···
 unsigned int pmem_sector_size(struct nd_namespace_common *ndns);
 void nvdimm_badblocks_populate(struct nd_region *nd_region,
 		struct badblocks *bb, const struct resource *res);
+int devm_namespace_enable(struct device *dev, struct nd_namespace_common *ndns,
+		resource_size_t size);
+void devm_namespace_disable(struct device *dev,
+		struct nd_namespace_common *ndns);
 #if IS_ENABLED(CONFIG_ND_CLAIM)
-
 /* max struct page size independent of kernel config */
 #define MAX_STRUCT_PAGE_SIZE 64
-
 int nvdimm_setup_pfn(struct nd_pfn *nd_pfn, struct dev_pagemap *pgmap);
-int devm_nsio_enable(struct device *dev, struct nd_namespace_io *nsio);
-void devm_nsio_disable(struct device *dev, struct nd_namespace_io *nsio);
 #else
 static inline int nvdimm_setup_pfn(struct nd_pfn *nd_pfn,
 				   struct dev_pagemap *pgmap)
 {
 	return -ENXIO;
-}
-static inline int devm_nsio_enable(struct device *dev,
-		struct nd_namespace_io *nsio)
-{
-	return -ENXIO;
-}
-static inline void devm_nsio_disable(struct device *dev,
-		struct nd_namespace_io *nsio)
-{
 }
 #endif
 int nd_blk_region_init(struct nd_region *nd_region);

drivers/nvdimm/pfn_devs.c

···
 	meta_start = (SZ_4K + sizeof(*pfn_sb)) >> 9;
 	meta_num = (le64_to_cpu(pfn_sb->dataoff) >> 9) - meta_start;
 
+	/*
+	 * re-enable the namespace with correct size so that we can access
+	 * the device memmap area.
+	 */
+	devm_namespace_disable(&nd_pfn->dev, ndns);
+	rc = devm_namespace_enable(&nd_pfn->dev, ndns, le64_to_cpu(pfn_sb->dataoff));
+	if (rc)
+		return rc;
+
 	do {
 		unsigned long zero_len;
 		u64 nsoff;
···
 }
 EXPORT_SYMBOL(nd_pfn_probe);
 
-static u32 info_block_reserve(void)
-{
-	return ALIGN(SZ_8K, PAGE_SIZE);
-}
-
 /*
  * We hotplug memory at sub-section granularity, pad the reserved area
  * from the previous section base to the namespace base address.
···
 
 static unsigned long init_altmap_reserve(resource_size_t base)
 {
-	unsigned long reserve = info_block_reserve() >> PAGE_SHIFT;
+	unsigned long reserve = nd_info_block_reserve() >> PAGE_SHIFT;
 	unsigned long base_pfn = PHYS_PFN(base);
 
 	reserve += base_pfn - SUBSECTION_ALIGN_DOWN(base_pfn);
···
 	u64 offset = le64_to_cpu(pfn_sb->dataoff);
 	u32 start_pad = __le32_to_cpu(pfn_sb->start_pad);
 	u32 end_trunc = __le32_to_cpu(pfn_sb->end_trunc);
-	u32 reserve = info_block_reserve();
+	u32 reserve = nd_info_block_reserve();
 	struct nd_namespace_common *ndns = nd_pfn->ndns;
 	struct nd_namespace_io *nsio = to_nd_namespace_io(&ndns->dev);
 	resource_size_t base = nsio->res.start + start_pad;

drivers/nvdimm/pmem.c

···
 	if (!pmem)
 		return -ENOMEM;
 
+	rc = devm_namespace_enable(dev, ndns, nd_info_block_reserve());
+	if (rc)
+		return rc;
+
 	/* while nsio_rw_bytes is active, parse a pfn info block if present */
 	if (is_nd_pfn(dev)) {
 		nd_pfn = to_nd_pfn(dev);
···
 	}
 
 	/* we're attaching a block device, disable raw namespace access */
-	devm_nsio_disable(dev, nsio);
+	devm_namespace_disable(dev, ndns);
 
 	dev_set_drvdata(dev, pmem);
 	pmem->phys_addr = res->start;
···
 	if (IS_ERR(ndns))
 		return PTR_ERR(ndns);
 
-	if (devm_nsio_enable(dev, to_nd_namespace_io(&ndns->dev)))
-		return -ENXIO;
-
 	if (is_nd_btt(dev))
 		return nvdimm_namespace_attach_btt(ndns);
 
 	if (is_nd_pfn(dev))
 		return pmem_attach_disk(dev, ndns);
+
+	ret = devm_namespace_enable(dev, ndns, nd_info_block_reserve());
+	if (ret)
+		return ret;
 
 	ret = nd_btt_probe(dev, ndns);
 	if (ret == 0)
···
 		return -ENXIO;
 	else if (ret == -EOPNOTSUPP)
 		return ret;
+
+	/* probe complete, attach handles namespace enabling */
+	devm_namespace_disable(dev, ndns);
+
 	return pmem_attach_disk(dev, ndns);
 }