[linux-2.6-block.git] / kernel / memremap.c

/*
 * Copyright(c) 2015 Intel Corporation. All rights reserved.
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of version 2 of the GNU General Public License as
 * published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful, but
 * WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 */
#include <linux/radix-tree.h>
#include <linux/memremap.h>
#include <linux/device.h>
#include <linux/types.h>
#include <linux/pfn_t.h>
#include <linux/io.h>
#include <linux/mm.h>
#include <linux/memory_hotplug.h>

#ifndef ioremap_cache
/* temporary while we convert existing ioremap_cache users to memremap */
__weak void __iomem *ioremap_cache(resource_size_t offset, unsigned long size)
{
	return ioremap(offset, size);
}
#endif

static void *try_ram_remap(resource_size_t offset, size_t size)
{
	unsigned long pfn = PHYS_PFN(offset);

	/* In the simple case just return the existing linear address */
	if (pfn_valid(pfn) && !PageHighMem(pfn_to_page(pfn)))
		return __va(offset);
	return NULL; /* fallback to ioremap_cache */
}

/**
 * memremap() - remap an iomem_resource as cacheable memory
 * @offset: iomem resource start address
 * @size: size of remap
 * @flags: any of MEMREMAP_WB, MEMREMAP_WT and MEMREMAP_WC
 *
 * memremap() is "ioremap" for cases where it is known that the resource
 * being mapped does not have i/o side effects and the __iomem
 * annotation is not applicable. In the case of multiple flags, the different
 * mapping types will be attempted in the order listed below until one of
 * them succeeds.
 *
 * MEMREMAP_WB - matches the default mapping for System RAM on
 * the architecture.  This is usually a read-allocate write-back cache.
 * Morever, if MEMREMAP_WB is specified and the requested remap region is RAM
 * memremap() will bypass establishing a new mapping and instead return
 * a pointer into the direct map.
 *
 * MEMREMAP_WT - establish a mapping whereby writes either bypass the
 * cache or are written through to memory and never exist in a
 * cache-dirty state with respect to program visibility.  Attempts to
 * map System RAM with this mapping type will fail.
 *
 * MEMREMAP_WC - establish a writecombine mapping, whereby writes may
 * be coalesced together (e.g. in the CPU's write buffers), but is otherwise
 * uncached. Attempts to map System RAM with this mapping type will fail.
 */
void *memremap(resource_size_t offset, size_t size, unsigned long flags)
{
	int is_ram = region_intersects(offset, size,
				       IORESOURCE_SYSTEM_RAM, IORES_DESC_NONE);
	void *addr = NULL;

	if (!flags)
		return NULL;

	if (is_ram == REGION_MIXED) {
		WARN_ONCE(1, "memremap attempted on mixed range %pa size: %#lx\n",
				&offset, (unsigned long) size);
		return NULL;
	}

	/* Try all mapping types requested until one returns non-NULL */
	if (flags & MEMREMAP_WB) {
		/*
		 * MEMREMAP_WB is special in that it can be satisifed
		 * from the direct map.  Some archs depend on the
		 * capability of memremap() to autodetect cases where
		 * the requested range is potentially in System RAM.
		 */
		if (is_ram == REGION_INTERSECTS)
			addr = try_ram_remap(offset, size);
		if (!addr)
			addr = ioremap_cache(offset, size);
	}

	/*
	 * If we don't have a mapping yet and other request flags are
	 * present then we will be attempting to establish a new virtual
	 * address mapping.  Enforce that this mapping is not aliasing
	 * System RAM.
	 */
	if (!addr && is_ram == REGION_INTERSECTS && flags != MEMREMAP_WB) {
		WARN_ONCE(1, "memremap attempted on ram %pa size: %#lx\n",
				&offset, (unsigned long) size);
		return NULL;
	}

	if (!addr && (flags & MEMREMAP_WT))
		addr = ioremap_wt(offset, size);

	if (!addr && (flags & MEMREMAP_WC))
		addr = ioremap_wc(offset, size);

	return addr;
}
EXPORT_SYMBOL(memremap);

void memunmap(void *addr)
{
	if (is_vmalloc_addr(addr))
		iounmap((void __iomem *) addr);
}
EXPORT_SYMBOL(memunmap);

static void devm_memremap_release(struct device *dev, void *res)
{
	memunmap(*(void **)res);
}

static int devm_memremap_match(struct device *dev, void *res, void *match_data)
{
	return *(void **)res == match_data;
}

void *devm_memremap(struct device *dev, resource_size_t offset,
		size_t size, unsigned long flags)
{
	void **ptr, *addr;

	ptr = devres_alloc_node(devm_memremap_release, sizeof(*ptr), GFP_KERNEL,
			dev_to_node(dev));
	if (!ptr)
		return ERR_PTR(-ENOMEM);

	addr = memremap(offset, size, flags);
	if (addr) {
		*ptr = addr;
		devres_add(dev, ptr);
	} else {
		devres_free(ptr);
		return ERR_PTR(-ENXIO);
	}

	return addr;
}
EXPORT_SYMBOL(devm_memremap);

void devm_memunmap(struct device *dev, void *addr)
{
	WARN_ON(devres_release(dev, devm_memremap_release,
				devm_memremap_match, addr));
}
EXPORT_SYMBOL(devm_memunmap);

pfn_t phys_to_pfn_t(phys_addr_t addr, u64 flags)
{
	return __pfn_to_pfn_t(addr >> PAGE_SHIFT, flags);
}
EXPORT_SYMBOL(phys_to_pfn_t);

#ifdef CONFIG_ZONE_DEVICE
static DEFINE_MUTEX(pgmap_lock);
static RADIX_TREE(pgmap_radix, GFP_KERNEL);
#define SECTION_MASK ~((1UL << PA_SECTION_SHIFT) - 1)
#define SECTION_SIZE (1UL << PA_SECTION_SHIFT)

struct page_map {
	struct resource res;
	struct percpu_ref *ref;
	struct dev_pagemap pgmap;
	struct vmem_altmap altmap;
};

void get_zone_device_page(struct page *page)
{
	percpu_ref_get(page->pgmap->ref);
}
EXPORT_SYMBOL(get_zone_device_page);

void put_zone_device_page(struct page *page)
{
	put_dev_pagemap(page->pgmap);
}
EXPORT_SYMBOL(put_zone_device_page);

static void pgmap_radix_release(struct resource *res)
{
	resource_size_t key, align_start, align_size, align_end;

	align_start = res->start & ~(SECTION_SIZE - 1);
	align_size = ALIGN(resource_size(res), SECTION_SIZE);
	align_end = align_start + align_size - 1;

	mutex_lock(&pgmap_lock);
	for (key = res->start; key <= res->end; key += SECTION_SIZE)
		radix_tree_delete(&pgmap_radix, key >> PA_SECTION_SHIFT);
	mutex_unlock(&pgmap_lock);
}

static unsigned long pfn_first(struct page_map *page_map)
{
	struct dev_pagemap *pgmap = &page_map->pgmap;
	const struct resource *res = &page_map->res;
	struct vmem_altmap *altmap = pgmap->altmap;
	unsigned long pfn;

	pfn = res->start >> PAGE_SHIFT;
	if (altmap)
		pfn += vmem_altmap_offset(altmap);
	return pfn;
}

static unsigned long pfn_end(struct page_map *page_map)
{
	const struct resource *res = &page_map->res;

	return (res->start + resource_size(res)) >> PAGE_SHIFT;
}

#define for_each_device_pfn(pfn, map) \
	for (pfn = pfn_first(map); pfn < pfn_end(map); pfn++)

static void devm_memremap_pages_release(struct device *dev, void *data)
{
	struct page_map *page_map = data;
	struct resource *res = &page_map->res;
	resource_size_t align_start, align_size;
	struct dev_pagemap *pgmap = &page_map->pgmap;

	if (percpu_ref_tryget_live(pgmap->ref)) {
		dev_WARN(dev, "%s: page mapping is still live!\n", __func__);
		percpu_ref_put(pgmap->ref);
	}

	/* pages are dead and unused, undo the arch mapping */
	align_start = res->start & ~(SECTION_SIZE - 1);
	align_size = ALIGN(resource_size(res), SECTION_SIZE);
	arch_remove_memory(align_start, align_size);
	pgmap_radix_release(res);
	dev_WARN_ONCE(dev, pgmap->altmap && pgmap->altmap->alloc,
			"%s: failed to free all reserved pages\n", __func__);
}

/* assumes rcu_read_lock() held at entry */
struct dev_pagemap *find_dev_pagemap(resource_size_t phys)
{
	struct page_map *page_map;

	WARN_ON_ONCE(!rcu_read_lock_held());

	page_map = radix_tree_lookup(&pgmap_radix, phys >> PA_SECTION_SHIFT);
	return page_map ? &page_map->pgmap : NULL;
}

/**
 * devm_memremap_pages - remap and provide memmap backing for the given resource
 * @dev: hosting device for @res
 * @res: "host memory" address range
 * @ref: a live per-cpu reference count
 * @altmap: optional descriptor for allocating the memmap from @res
 *
 * Notes:
 * 1/ @ref must be 'live' on entry and 'dead' before devm_memunmap_pages() time
 *    (or devm release event).
 *
 * 2/ @res is expected to be a host memory range that could feasibly be
 *    treated as a "System RAM" range, i.e. not a device mmio range, but
 *    this is not enforced.
 */
void *devm_memremap_pages(struct device *dev, struct resource *res,
		struct percpu_ref *ref, struct vmem_altmap *altmap)
{
	resource_size_t key, align_start, align_size, align_end;
	struct dev_pagemap *pgmap;
	struct page_map *page_map;
	int error, nid, is_ram;
	unsigned long pfn;

	align_start = res->start & ~(SECTION_SIZE - 1);
	align_size = ALIGN(res->start + resource_size(res), SECTION_SIZE)
		- align_start;
	is_ram = region_intersects(align_start, align_size,
		IORESOURCE_SYSTEM_RAM, IORES_DESC_NONE);

	if (is_ram == REGION_MIXED) {
		WARN_ONCE(1, "%s attempted on mixed region %pr\n",
				__func__, res);
		return ERR_PTR(-ENXIO);
	}

	if (is_ram == REGION_INTERSECTS)
		return __va(res->start);

	if (altmap && !IS_ENABLED(CONFIG_SPARSEMEM_VMEMMAP)) {
		dev_err(dev, "%s: altmap requires CONFIG_SPARSEMEM_VMEMMAP=y\n",
				__func__);
		return ERR_PTR(-ENXIO);
	}

	if (!ref)
		return ERR_PTR(-EINVAL);

	page_map = devres_alloc_node(devm_memremap_pages_release,
			sizeof(*page_map), GFP_KERNEL, dev_to_node(dev));
	if (!page_map)
		return ERR_PTR(-ENOMEM);
	pgmap = &page_map->pgmap;

	memcpy(&page_map->res, res, sizeof(*res));

	pgmap->dev = dev;
	if (altmap) {
		memcpy(&page_map->altmap, altmap, sizeof(*altmap));
		pgmap->altmap = &page_map->altmap;
	}
	pgmap->ref = ref;
	pgmap->res = &page_map->res;

	mutex_lock(&pgmap_lock);
	error = 0;
	align_end = align_start + align_size - 1;
	for (key = align_start; key <= align_end; key += SECTION_SIZE) {
		struct dev_pagemap *dup;

		rcu_read_lock();
		dup = find_dev_pagemap(key);
		rcu_read_unlock();
		if (dup) {
			dev_err(dev, "%s: %pr collides with mapping for %s\n",
					__func__, res, dev_name(dup->dev));
			error = -EBUSY;
			break;
		}
		error = radix_tree_insert(&pgmap_radix, key >> PA_SECTION_SHIFT,
				page_map);
		if (error) {
			dev_err(dev, "%s: failed: %d\n", __func__, error);
			break;
		}
	}
	mutex_unlock(&pgmap_lock);
	if (error)
		goto err_radix;

	nid = dev_to_node(dev);
	if (nid < 0)
		nid = numa_mem_id();

	error = arch_add_memory(nid, align_start, align_size, true);
	if (error)
		goto err_add_memory;

	for_each_device_pfn(pfn, page_map) {
		struct page *page = pfn_to_page(pfn);

		/*
		 * ZONE_DEVICE pages union ->lru with a ->pgmap back
		 * pointer.  It is a bug if a ZONE_DEVICE page is ever
		 * freed or placed on a driver-private list.  Seed the
		 * storage with LIST_POISON* values.
		 */
		list_del(&page->lru);
		page->pgmap = pgmap;
	}
	devres_add(dev, page_map);
	return __va(res->start);

 err_add_memory:
 err_radix:
	pgmap_radix_release(res);
	devres_free(page_map);
	return ERR_PTR(error);
}
EXPORT_SYMBOL(devm_memremap_pages);

unsigned long vmem_altmap_offset(struct vmem_altmap *altmap)
{
	/* number of pfns from base where pfn_to_page() is valid */
	return altmap->reserve + altmap->free;
}

void vmem_altmap_free(struct vmem_altmap *altmap, unsigned long nr_pfns)
{
	altmap->alloc -= nr_pfns;
}

#ifdef CONFIG_SPARSEMEM_VMEMMAP
struct vmem_altmap *to_vmem_altmap(unsigned long memmap_start)
{
	/*
	 * 'memmap_start' is the virtual address for the first "struct
	 * page" in this range of the vmemmap array.  In the case of
	 * CONFIG_SPARSEMEM_VMEMMAP a page_to_pfn conversion is simple
	 * pointer arithmetic, so we can perform this to_vmem_altmap()
	 * conversion without concern for the initialization state of
	 * the struct page fields.
	 */
	struct page *page = (struct page *) memmap_start;
	struct dev_pagemap *pgmap;

	/*
	 * Unconditionally retrieve a dev_pagemap associated with the
	 * given physical address, this is only for use in the
	 * arch_{add|remove}_memory() for setting up and tearing down
	 * the memmap.
	 */
	rcu_read_lock();
	pgmap = find_dev_pagemap(__pfn_to_phys(page_to_pfn(page)));
	rcu_read_unlock();

	return pgmap ? pgmap->altmap : NULL;
}
#endif /* CONFIG_SPARSEMEM_VMEMMAP */
#endif /* CONFIG_ZONE_DEVICE */
Commit	Line	Data
92281dee DW	1	/*
	2	* Copyright(c) 2015 Intel Corporation. All rights reserved.
	3	*
	4	* This program is free software; you can redistribute it and/or modify
	5	* it under the terms of version 2 of the GNU General Public License as
	6	* published by the Free Software Foundation.
	7	*
	8	* This program is distributed in the hope that it will be useful, but
	9	* WITHOUT ANY WARRANTY; without even the implied warranty of
	10	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	11	* General Public License for more details.
	12	*/
9476df7d DW	13	#include <linux/radix-tree.h>
9476df7d DW	14	#include <linux/memremap.h>
7d3dcf26	15	#include <linux/device.h>
92281dee	16	#include <linux/types.h>
34c0fd54	17	#include <linux/pfn_t.h>
92281dee DW	18	#include <linux/io.h>
92281dee DW	19	#include <linux/mm.h>
41e94a85	20	#include <linux/memory_hotplug.h>
92281dee DW	21
	22	#ifndef ioremap_cache
	23	/* temporary while we convert existing ioremap_cache users to memremap */
	24	__weak void __iomem *ioremap_cache(resource_size_t offset, unsigned long size)
	25	{
	26	return ioremap(offset, size);
	27	}
	28	#endif
	29
182475b7 DW	30	static void *try_ram_remap(resource_size_t offset, size_t size)
182475b7 DW	31	{
ac343e88	32	unsigned long pfn = PHYS_PFN(offset);
182475b7 DW	33
182475b7 DW	34	/* In the simple case just return the existing linear address */
ac343e88	35	if (pfn_valid(pfn) && !PageHighMem(pfn_to_page(pfn)))
182475b7 DW	36	return __va(offset);
	37	return NULL; /* fallback to ioremap_cache */
	38	}
	39
92281dee DW	40	/**
	41	* memremap() - remap an iomem_resource as cacheable memory
	42	* @offset: iomem resource start address
	43	* @size: size of remap
c907e0eb	44	* @flags: any of MEMREMAP_WB, MEMREMAP_WT and MEMREMAP_WC
92281dee DW	45	*
	46	* memremap() is "ioremap" for cases where it is known that the resource
	47	* being mapped does not have i/o side effects and the __iomem
c907e0eb BS	48	* annotation is not applicable. In the case of multiple flags, the different
	49	* mapping types will be attempted in the order listed below until one of
	50	* them succeeds.
92281dee	51	*
1c29f25b	52	* MEMREMAP_WB - matches the default mapping for System RAM on
92281dee DW	53	* the architecture. This is usually a read-allocate write-back cache.
	54	* Morever, if MEMREMAP_WB is specified and the requested remap region is RAM
	55	* memremap() will bypass establishing a new mapping and instead return
	56	* a pointer into the direct map.
	57	*
	58	* MEMREMAP_WT - establish a mapping whereby writes either bypass the
	59	* cache or are written through to memory and never exist in a
	60	* cache-dirty state with respect to program visibility. Attempts to
1c29f25b	61	* map System RAM with this mapping type will fail.
c907e0eb BS	62	*
	63	* MEMREMAP_WC - establish a writecombine mapping, whereby writes may
	64	* be coalesced together (e.g. in the CPU's write buffers), but is otherwise
	65	* uncached. Attempts to map System RAM with this mapping type will fail.
92281dee DW	66	*/
	67	void *memremap(resource_size_t offset, size_t size, unsigned long flags)
	68	{
1c29f25b TK	69	int is_ram = region_intersects(offset, size,
1c29f25b TK	70	IORESOURCE_SYSTEM_RAM, IORES_DESC_NONE);
92281dee DW	71	void *addr = NULL;
92281dee DW	72
cf61e2a1 BS	73	if (!flags)
	74	return NULL;
	75
92281dee DW	76	if (is_ram == REGION_MIXED) {
	77	WARN_ONCE(1, "memremap attempted on mixed range %pa size: %#lx\n",
	78	&offset, (unsigned long) size);
	79	return NULL;
	80	}
	81
	82	/* Try all mapping types requested until one returns non-NULL */
	83	if (flags & MEMREMAP_WB) {
92281dee DW	84	/*
	85	* MEMREMAP_WB is special in that it can be satisifed
	86	* from the direct map. Some archs depend on the
	87	* capability of memremap() to autodetect cases where
1c29f25b	88	* the requested range is potentially in System RAM.
92281dee DW	89	*/
92281dee DW	90	if (is_ram == REGION_INTERSECTS)
182475b7 DW	91	addr = try_ram_remap(offset, size);
182475b7 DW	92	if (!addr)
92281dee DW	93	addr = ioremap_cache(offset, size);
	94	}
	95
	96	/*
cf61e2a1 BS	97	* If we don't have a mapping yet and other request flags are
cf61e2a1 BS	98	* present then we will be attempting to establish a new virtual
92281dee	99	* address mapping. Enforce that this mapping is not aliasing
1c29f25b	100	* System RAM.
92281dee	101	*/
cf61e2a1	102	if (!addr && is_ram == REGION_INTERSECTS && flags != MEMREMAP_WB) {
92281dee DW	103	WARN_ONCE(1, "memremap attempted on ram %pa size: %#lx\n",
	104	&offset, (unsigned long) size);
	105	return NULL;
	106	}
	107
cf61e2a1	108	if (!addr && (flags & MEMREMAP_WT))
92281dee	109	addr = ioremap_wt(offset, size);
c907e0eb BS	110
	111	if (!addr && (flags & MEMREMAP_WC))
	112	addr = ioremap_wc(offset, size);
92281dee DW	113
	114	return addr;
	115	}
	116	EXPORT_SYMBOL(memremap);
	117
	118	void memunmap(void *addr)
	119	{
	120	if (is_vmalloc_addr(addr))
	121	iounmap((void __iomem *) addr);
	122	}
	123	EXPORT_SYMBOL(memunmap);
7d3dcf26 CH	124
	125	static void devm_memremap_release(struct device dev, void res)
	126	{
9273a8bb	127	memunmap((void *)res);
7d3dcf26 CH	128	}
	129
	130	static int devm_memremap_match(struct device dev, void res, void *match_data)
	131	{
	132	return (void *)res == match_data;
	133	}
	134
	135	void devm_memremap(struct device dev, resource_size_t offset,
	136	size_t size, unsigned long flags)
	137	{
	138	void *ptr, addr;
	139
538ea4aa DW	140	ptr = devres_alloc_node(devm_memremap_release, sizeof(*ptr), GFP_KERNEL,
538ea4aa DW	141	dev_to_node(dev));
7d3dcf26	142	if (!ptr)
b36f4761	143	return ERR_PTR(-ENOMEM);
7d3dcf26 CH	144
	145	addr = memremap(offset, size, flags);
	146	if (addr) {
	147	*ptr = addr;
	148	devres_add(dev, ptr);
93f834df	149	} else {
7d3dcf26	150	devres_free(ptr);
93f834df TK	151	return ERR_PTR(-ENXIO);
93f834df TK	152	}
7d3dcf26 CH	153
	154	return addr;
	155	}
	156	EXPORT_SYMBOL(devm_memremap);
	157
	158	void devm_memunmap(struct device dev, void addr)
	159	{
d741314f DW	160	WARN_ON(devres_release(dev, devm_memremap_release,
d741314f DW	161	devm_memremap_match, addr));
7d3dcf26 CH	162	}
7d3dcf26 CH	163	EXPORT_SYMBOL(devm_memunmap);
41e94a85	164
db78c222	165	pfn_t phys_to_pfn_t(phys_addr_t addr, u64 flags)
34c0fd54 DW	166	{
	167	return __pfn_to_pfn_t(addr >> PAGE_SHIFT, flags);
	168	}
	169	EXPORT_SYMBOL(phys_to_pfn_t);
	170
41e94a85	171	#ifdef CONFIG_ZONE_DEVICE
9476df7d DW	172	static DEFINE_MUTEX(pgmap_lock);
	173	static RADIX_TREE(pgmap_radix, GFP_KERNEL);
	174	#define SECTION_MASK ~((1UL << PA_SECTION_SHIFT) - 1)
	175	#define SECTION_SIZE (1UL << PA_SECTION_SHIFT)
	176
41e94a85 CH	177	struct page_map {
41e94a85 CH	178	struct resource res;
9476df7d DW	179	struct percpu_ref *ref;
9476df7d DW	180	struct dev_pagemap pgmap;
4b94ffdc	181	struct vmem_altmap altmap;
41e94a85 CH	182	};
41e94a85 CH	183
3565fce3 DW	184	void get_zone_device_page(struct page *page)
	185	{
	186	percpu_ref_get(page->pgmap->ref);
	187	}
	188	EXPORT_SYMBOL(get_zone_device_page);
	189
	190	void put_zone_device_page(struct page *page)
	191	{
	192	put_dev_pagemap(page->pgmap);
	193	}
	194	EXPORT_SYMBOL(put_zone_device_page);
	195
9476df7d DW	196	static void pgmap_radix_release(struct resource *res)
9476df7d DW	197	{
eb7d78c9 DW	198	resource_size_t key, align_start, align_size, align_end;
	199
	200	align_start = res->start & ~(SECTION_SIZE - 1);
	201	align_size = ALIGN(resource_size(res), SECTION_SIZE);
	202	align_end = align_start + align_size - 1;
9476df7d DW	203
	204	mutex_lock(&pgmap_lock);
	205	for (key = res->start; key <= res->end; key += SECTION_SIZE)
	206	radix_tree_delete(&pgmap_radix, key >> PA_SECTION_SHIFT);
	207	mutex_unlock(&pgmap_lock);
	208	}
	209
5c2c2587 DW	210	static unsigned long pfn_first(struct page_map *page_map)
	211	{
	212	struct dev_pagemap *pgmap = &page_map->pgmap;
	213	const struct resource *res = &page_map->res;
	214	struct vmem_altmap *altmap = pgmap->altmap;
	215	unsigned long pfn;
	216
	217	pfn = res->start >> PAGE_SHIFT;
	218	if (altmap)
	219	pfn += vmem_altmap_offset(altmap);
	220	return pfn;
	221	}
	222
	223	static unsigned long pfn_end(struct page_map *page_map)
	224	{
	225	const struct resource *res = &page_map->res;
	226
	227	return (res->start + resource_size(res)) >> PAGE_SHIFT;
	228	}
	229
	230	#define for_each_device_pfn(pfn, map) \
	231	for (pfn = pfn_first(map); pfn < pfn_end(map); pfn++)
	232
9476df7d	233	static void devm_memremap_pages_release(struct device dev, void data)
41e94a85	234	{
9476df7d DW	235	struct page_map *page_map = data;
	236	struct resource *res = &page_map->res;
	237	resource_size_t align_start, align_size;
4b94ffdc	238	struct dev_pagemap *pgmap = &page_map->pgmap;
9476df7d	239
5c2c2587 DW	240	if (percpu_ref_tryget_live(pgmap->ref)) {
	241	dev_WARN(dev, "%s: page mapping is still live!\n", __func__);
	242	percpu_ref_put(pgmap->ref);
	243	}
	244
41e94a85	245	/* pages are dead and unused, undo the arch mapping */
9476df7d DW	246	align_start = res->start & ~(SECTION_SIZE - 1);
	247	align_size = ALIGN(resource_size(res), SECTION_SIZE);
	248	arch_remove_memory(align_start, align_size);
eb7d78c9	249	pgmap_radix_release(res);
4b94ffdc DW	250	dev_WARN_ONCE(dev, pgmap->altmap && pgmap->altmap->alloc,
4b94ffdc DW	251	"%s: failed to free all reserved pages\n", __func__);
9476df7d DW	252	}
	253
	254	/* assumes rcu_read_lock() held at entry */
	255	struct dev_pagemap *find_dev_pagemap(resource_size_t phys)
	256	{
	257	struct page_map *page_map;
	258
	259	WARN_ON_ONCE(!rcu_read_lock_held());
	260
	261	page_map = radix_tree_lookup(&pgmap_radix, phys >> PA_SECTION_SHIFT);
	262	return page_map ? &page_map->pgmap : NULL;
41e94a85 CH	263	}
41e94a85 CH	264
4b94ffdc DW	265	/**
	266	* devm_memremap_pages - remap and provide memmap backing for the given resource
	267	* @dev: hosting device for @res
	268	* @res: "host memory" address range
5c2c2587	269	* @ref: a live per-cpu reference count
4b94ffdc DW	270	* @altmap: optional descriptor for allocating the memmap from @res
4b94ffdc DW	271	*
5c2c2587 DW	272	* Notes:
	273	* 1/ @ref must be 'live' on entry and 'dead' before devm_memunmap_pages() time
	274	* (or devm release event).
	275	*
	276	* 2/ @res is expected to be a host memory range that could feasibly be
	277	* treated as a "System RAM" range, i.e. not a device mmio range, but
	278	* this is not enforced.
4b94ffdc DW	279	*/
4b94ffdc DW	280	void devm_memremap_pages(struct device dev, struct resource *res,
5c2c2587	281	struct percpu_ref ref, struct vmem_altmap altmap)
41e94a85	282	{
eb7d78c9	283	resource_size_t key, align_start, align_size, align_end;
4b94ffdc	284	struct dev_pagemap *pgmap;
41e94a85	285	struct page_map *page_map;
5f29a77c	286	int error, nid, is_ram;
5c2c2587	287	unsigned long pfn;
5f29a77c DW	288
	289	align_start = res->start & ~(SECTION_SIZE - 1);
	290	align_size = ALIGN(res->start + resource_size(res), SECTION_SIZE)
	291	- align_start;
d37a14bb LT	292	is_ram = region_intersects(align_start, align_size,
d37a14bb LT	293	IORESOURCE_SYSTEM_RAM, IORES_DESC_NONE);
41e94a85 CH	294
	295	if (is_ram == REGION_MIXED) {
	296	WARN_ONCE(1, "%s attempted on mixed region %pr\n",
	297	__func__, res);
	298	return ERR_PTR(-ENXIO);
	299	}
	300
	301	if (is_ram == REGION_INTERSECTS)
	302	return __va(res->start);
	303
4b94ffdc DW	304	if (altmap && !IS_ENABLED(CONFIG_SPARSEMEM_VMEMMAP)) {
	305	dev_err(dev, "%s: altmap requires CONFIG_SPARSEMEM_VMEMMAP=y\n",
	306	__func__);
	307	return ERR_PTR(-ENXIO);
	308	}
	309
5c2c2587 DW	310	if (!ref)
	311	return ERR_PTR(-EINVAL);
	312
538ea4aa DW	313	page_map = devres_alloc_node(devm_memremap_pages_release,
538ea4aa DW	314	sizeof(*page_map), GFP_KERNEL, dev_to_node(dev));
41e94a85 CH	315	if (!page_map)
41e94a85 CH	316	return ERR_PTR(-ENOMEM);
4b94ffdc	317	pgmap = &page_map->pgmap;
41e94a85 CH	318
	319	memcpy(&page_map->res, res, sizeof(*res));
	320
4b94ffdc DW	321	pgmap->dev = dev;
	322	if (altmap) {
	323	memcpy(&page_map->altmap, altmap, sizeof(*altmap));
	324	pgmap->altmap = &page_map->altmap;
	325	}
5c2c2587	326	pgmap->ref = ref;
4b94ffdc DW	327	pgmap->res = &page_map->res;
4b94ffdc DW	328
9476df7d DW	329	mutex_lock(&pgmap_lock);
9476df7d DW	330	error = 0;
eb7d78c9 DW	331	align_end = align_start + align_size - 1;
eb7d78c9 DW	332	for (key = align_start; key <= align_end; key += SECTION_SIZE) {
9476df7d DW	333	struct dev_pagemap *dup;
	334
	335	rcu_read_lock();
	336	dup = find_dev_pagemap(key);
	337	rcu_read_unlock();
	338	if (dup) {
	339	dev_err(dev, "%s: %pr collides with mapping for %s\n",
	340	__func__, res, dev_name(dup->dev));
	341	error = -EBUSY;
	342	break;
	343	}
	344	error = radix_tree_insert(&pgmap_radix, key >> PA_SECTION_SHIFT,
	345	page_map);
	346	if (error) {
	347	dev_err(dev, "%s: failed: %d\n", __func__, error);
	348	break;
	349	}
	350	}
	351	mutex_unlock(&pgmap_lock);
	352	if (error)
	353	goto err_radix;
	354
41e94a85 CH	355	nid = dev_to_node(dev);
41e94a85 CH	356	if (nid < 0)
7eff93b7	357	nid = numa_mem_id();
41e94a85	358
9476df7d DW	359	error = arch_add_memory(nid, align_start, align_size, true);
	360	if (error)
	361	goto err_add_memory;
41e94a85	362
5c2c2587 DW	363	for_each_device_pfn(pfn, page_map) {
	364	struct page *page = pfn_to_page(pfn);
	365
d77a117e DW	366	/*
	367	* ZONE_DEVICE pages union ->lru with a ->pgmap back
	368	* pointer. It is a bug if a ZONE_DEVICE page is ever
	369	* freed or placed on a driver-private list. Seed the
	370	* storage with LIST_POISON* values.
	371	*/
	372	list_del(&page->lru);
5c2c2587 DW	373	page->pgmap = pgmap;
5c2c2587 DW	374	}
41e94a85 CH	375	devres_add(dev, page_map);
41e94a85 CH	376	return __va(res->start);
9476df7d DW	377
	378	err_add_memory:
	379	err_radix:
	380	pgmap_radix_release(res);
	381	devres_free(page_map);
	382	return ERR_PTR(error);
41e94a85 CH	383	}
41e94a85 CH	384	EXPORT_SYMBOL(devm_memremap_pages);
4b94ffdc DW	385
	386	unsigned long vmem_altmap_offset(struct vmem_altmap *altmap)
	387	{
	388	/* number of pfns from base where pfn_to_page() is valid */
	389	return altmap->reserve + altmap->free;
	390	}
	391
	392	void vmem_altmap_free(struct vmem_altmap *altmap, unsigned long nr_pfns)
	393	{
	394	altmap->alloc -= nr_pfns;
	395	}
	396
	397	#ifdef CONFIG_SPARSEMEM_VMEMMAP
	398	struct vmem_altmap *to_vmem_altmap(unsigned long memmap_start)
	399	{
	400	/*
	401	* 'memmap_start' is the virtual address for the first "struct
	402	* page" in this range of the vmemmap array. In the case of
07061aab	403	* CONFIG_SPARSEMEM_VMEMMAP a page_to_pfn conversion is simple
4b94ffdc DW	404	* pointer arithmetic, so we can perform this to_vmem_altmap()
	405	* conversion without concern for the initialization state of
	406	* the struct page fields.
	407	*/
	408	struct page page = (struct page ) memmap_start;
	409	struct dev_pagemap *pgmap;
	410
	411	/*
07061aab	412	* Unconditionally retrieve a dev_pagemap associated with the
4b94ffdc DW	413	* given physical address, this is only for use in the
	414	* arch_{add\|remove}_memory() for setting up and tearing down
	415	* the memmap.
	416	*/
	417	rcu_read_lock();
	418	pgmap = find_dev_pagemap(__pfn_to_phys(page_to_pfn(page)));
	419	rcu_read_unlock();
	420
	421	return pgmap ? pgmap->altmap : NULL;
	422	}
	423	#endif /* CONFIG_SPARSEMEM_VMEMMAP */
41e94a85	424	#endif /* CONFIG_ZONE_DEVICE */