[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: either MEMREMAP_WB or MEMREMAP_WT
 *
 * 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.
 *
 * 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.
 */
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 (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) {
		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 more request flags are
	 * pending 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) {
		WARN_ONCE(1, "memremap attempted on ram %pa size: %#lx\n",
				&offset, (unsigned long) size);
		return NULL;
	}

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