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

/* SPDX-License-Identifier: GPL-2.0 */
/* Copyright(c) 2015 Intel Corporation. All rights reserved. */
#include <linux/device.h>
#include <linux/io.h>
#include <linux/kasan.h>
#include <linux/memory_hotplug.h>
#include <linux/mm.h>
#include <linux/pfn_t.h>
#include <linux/swap.h>
#include <linux/swapops.h>
#include <linux/types.h>
#include <linux/wait_bit.h>
#include <linux/xarray.h>
#include <linux/hmm.h>

static DEFINE_XARRAY(pgmap_array);
#define SECTION_MASK ~((1UL << PA_SECTION_SHIFT) - 1)
#define SECTION_SIZE (1UL << PA_SECTION_SHIFT)

#if IS_ENABLED(CONFIG_DEVICE_PRIVATE)
vm_fault_t device_private_entry_fault(struct vm_area_struct *vma,
		       unsigned long addr,
		       swp_entry_t entry,
		       unsigned int flags,
		       pmd_t *pmdp)
{
	struct page *page = device_private_entry_to_page(entry);
	struct hmm_devmem *devmem;

	devmem = container_of(page->pgmap, typeof(*devmem), pagemap);

	/*
	 * The page_fault() callback must migrate page back to system memory
	 * so that CPU can access it. This might fail for various reasons
	 * (device issue, device was unsafely unplugged, ...). When such
	 * error conditions happen, the callback must return VM_FAULT_SIGBUS.
	 *
	 * Note that because memory cgroup charges are accounted to the device
	 * memory, this should never fail because of memory restrictions (but
	 * allocation of regular system page might still fail because we are
	 * out of memory).
	 *
	 * There is a more in-depth description of what that callback can and
	 * cannot do, in include/linux/memremap.h
	 */
	return devmem->page_fault(vma, addr, page, flags, pmdp);
}
#endif /* CONFIG_DEVICE_PRIVATE */

static void pgmap_array_delete(struct resource *res)
{
	xa_store_range(&pgmap_array, PHYS_PFN(res->start), PHYS_PFN(res->end),
			NULL, GFP_KERNEL);
	synchronize_rcu();
}

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

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

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

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

static unsigned long pfn_next(unsigned long pfn)
{
	if (pfn % 1024 == 0)
		cond_resched();
	return pfn + 1;
}

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

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

	pgmap->kill(pgmap->ref);
	for_each_device_pfn(pfn, pgmap)
		put_page(pfn_to_page(pfn));
	pgmap->cleanup(pgmap->ref);

	/* pages are dead and unused, undo the arch mapping */
	align_start = res->start & ~(SECTION_SIZE - 1);
	align_size = ALIGN(res->start + resource_size(res), SECTION_SIZE)
		- align_start;

	nid = page_to_nid(pfn_to_page(align_start >> PAGE_SHIFT));

	mem_hotplug_begin();
	if (pgmap->type == MEMORY_DEVICE_PRIVATE) {
		pfn = align_start >> PAGE_SHIFT;
		__remove_pages(page_zone(pfn_to_page(pfn)), pfn,
				align_size >> PAGE_SHIFT, NULL);
	} else {
		arch_remove_memory(nid, align_start, align_size,
				pgmap->altmap_valid ? &pgmap->altmap : NULL);
		kasan_remove_zero_shadow(__va(align_start), align_size);
	}
	mem_hotplug_done();

	untrack_pfn(NULL, PHYS_PFN(align_start), align_size);
	pgmap_array_delete(res);
	dev_WARN_ONCE(dev, pgmap->altmap.alloc,
		      "%s: failed to free all reserved pages\n", __func__);
}

/**
 * devm_memremap_pages - remap and provide memmap backing for the given resource
 * @dev: hosting device for @res
 * @pgmap: pointer to a struct dev_pagemap
 *
 * Notes:
 * 1/ At a minimum the res, ref and type members of @pgmap must be initialized
 *    by the caller before passing it to this function
 *
 * 2/ The altmap field may optionally be initialized, in which case altmap_valid
 *    must be set to true
 *
 * 3/ pgmap->ref must be 'live' on entry and will be killed and reaped
 *    at devm_memremap_pages_release() time, or if this routine fails.
 *
 * 4/ 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 dev_pagemap *pgmap)
{
	resource_size_t align_start, align_size, align_end;
	struct vmem_altmap *altmap = pgmap->altmap_valid ?
			&pgmap->altmap : NULL;
	struct resource *res = &pgmap->res;
	struct dev_pagemap *conflict_pgmap;
	struct mhp_restrictions restrictions = {
		/*
		 * We do not want any optional features only our own memmap
		*/
		.altmap = altmap,
	};
	pgprot_t pgprot = PAGE_KERNEL;
	int error, nid, is_ram;

	if (!pgmap->ref || !pgmap->kill || !pgmap->cleanup) {
		WARN(1, "Missing reference count teardown definition\n");
		return ERR_PTR(-EINVAL);
	}

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

	conflict_pgmap = get_dev_pagemap(PHYS_PFN(align_start), NULL);
	if (conflict_pgmap) {
		dev_WARN(dev, "Conflicting mapping in same section\n");
		put_dev_pagemap(conflict_pgmap);
		error = -ENOMEM;
		goto err_array;
	}

	conflict_pgmap = get_dev_pagemap(PHYS_PFN(align_end), NULL);
	if (conflict_pgmap) {
		dev_WARN(dev, "Conflicting mapping in same section\n");
		put_dev_pagemap(conflict_pgmap);
		error = -ENOMEM;
		goto err_array;
	}

	is_ram = region_intersects(align_start, align_size,
		IORESOURCE_SYSTEM_RAM, IORES_DESC_NONE);

	if (is_ram != REGION_DISJOINT) {
		WARN_ONCE(1, "%s attempted on %s region %pr\n", __func__,
				is_ram == REGION_MIXED ? "mixed" : "ram", res);
		error = -ENXIO;
		goto err_array;
	}

	pgmap->dev = dev;

	error = xa_err(xa_store_range(&pgmap_array, PHYS_PFN(res->start),
				PHYS_PFN(res->end), pgmap, GFP_KERNEL));
	if (error)
		goto err_array;

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

	error = track_pfn_remap(NULL, &pgprot, PHYS_PFN(align_start), 0,
			align_size);
	if (error)
		goto err_pfn_remap;

	mem_hotplug_begin();

	/*
	 * For device private memory we call add_pages() as we only need to
	 * allocate and initialize struct page for the device memory. More-
	 * over the device memory is un-accessible thus we do not want to
	 * create a linear mapping for the memory like arch_add_memory()
	 * would do.
	 *
	 * For all other device memory types, which are accessible by
	 * the CPU, we do want the linear mapping and thus use
	 * arch_add_memory().
	 */
	if (pgmap->type == MEMORY_DEVICE_PRIVATE) {
		error = add_pages(nid, align_start >> PAGE_SHIFT,
				align_size >> PAGE_SHIFT, &restrictions);
	} else {
		error = kasan_add_zero_shadow(__va(align_start), align_size);
		if (error) {
			mem_hotplug_done();
			goto err_kasan;
		}

		error = arch_add_memory(nid, align_start, align_size,
					&restrictions);
	}

	if (!error) {
		struct zone *zone;

		zone = &NODE_DATA(nid)->node_zones[ZONE_DEVICE];
		move_pfn_range_to_zone(zone, align_start >> PAGE_SHIFT,
				align_size >> PAGE_SHIFT, altmap);
	}

	mem_hotplug_done();
	if (error)
		goto err_add_memory;

	/*
	 * Initialization of the pages has been deferred until now in order
	 * to allow us to do the work while not holding the hotplug lock.
	 */
	memmap_init_zone_device(&NODE_DATA(nid)->node_zones[ZONE_DEVICE],
				align_start >> PAGE_SHIFT,
				align_size >> PAGE_SHIFT, pgmap);
	percpu_ref_get_many(pgmap->ref, pfn_end(pgmap) - pfn_first(pgmap));

	error = devm_add_action_or_reset(dev, devm_memremap_pages_release,
			pgmap);
	if (error)
		return ERR_PTR(error);

	return __va(res->start);

 err_add_memory:
	kasan_remove_zero_shadow(__va(align_start), align_size);
 err_kasan:
	untrack_pfn(NULL, PHYS_PFN(align_start), align_size);
 err_pfn_remap:
	pgmap_array_delete(res);
 err_array:
	pgmap->kill(pgmap->ref);
	pgmap->cleanup(pgmap->ref);

	return ERR_PTR(error);
}
EXPORT_SYMBOL_GPL(devm_memremap_pages);

void devm_memunmap_pages(struct device *dev, struct dev_pagemap *pgmap)
{
	devm_release_action(dev, devm_memremap_pages_release, pgmap);
}
EXPORT_SYMBOL_GPL(devm_memunmap_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;
}

/**
 * get_dev_pagemap() - take a new live reference on the dev_pagemap for @pfn
 * @pfn: page frame number to lookup page_map
 * @pgmap: optional known pgmap that already has a reference
 *
 * If @pgmap is non-NULL and covers @pfn it will be returned as-is.  If @pgmap
 * is non-NULL but does not cover @pfn the reference to it will be released.
 */
struct dev_pagemap *get_dev_pagemap(unsigned long pfn,
		struct dev_pagemap *pgmap)
{
	resource_size_t phys = PFN_PHYS(pfn);

	/*
	 * In the cached case we're already holding a live reference.
	 */
	if (pgmap) {
		if (phys >= pgmap->res.start && phys <= pgmap->res.end)
			return pgmap;
		put_dev_pagemap(pgmap);
	}

	/* fall back to slow path lookup */
	rcu_read_lock();
	pgmap = xa_load(&pgmap_array, PHYS_PFN(phys));
	if (pgmap && !percpu_ref_tryget_live(pgmap->ref))
		pgmap = NULL;
	rcu_read_unlock();

	return pgmap;
}
EXPORT_SYMBOL_GPL(get_dev_pagemap);

#ifdef CONFIG_DEV_PAGEMAP_OPS
DEFINE_STATIC_KEY_FALSE(devmap_managed_key);
EXPORT_SYMBOL(devmap_managed_key);
static atomic_t devmap_enable;

/*
 * Toggle the static key for ->page_free() callbacks when dev_pagemap
 * pages go idle.
 */
void dev_pagemap_get_ops(void)
{
	if (atomic_inc_return(&devmap_enable) == 1)
		static_branch_enable(&devmap_managed_key);
}
EXPORT_SYMBOL_GPL(dev_pagemap_get_ops);

void dev_pagemap_put_ops(void)
{
	if (atomic_dec_and_test(&devmap_enable))
		static_branch_disable(&devmap_managed_key);
}
EXPORT_SYMBOL_GPL(dev_pagemap_put_ops);

void __put_devmap_managed_page(struct page *page)
{
	int count = page_ref_dec_return(page);

	/*
	 * If refcount is 1 then page is freed and refcount is stable as nobody
	 * holds a reference on the page.
	 */
	if (count == 1) {
		/* Clear Active bit in case of parallel mark_page_accessed */
		__ClearPageActive(page);
		__ClearPageWaiters(page);

		mem_cgroup_uncharge(page);

		page->pgmap->page_free(page, page->pgmap->data);
	} else if (!count)
		__put_page(page);
}
EXPORT_SYMBOL(__put_devmap_managed_page);
#endif /* CONFIG_DEV_PAGEMAP_OPS */
Commit	Line	Data
5981690d DW	1	/* SPDX-License-Identifier: GPL-2.0 */
5981690d DW	2	/* Copyright(c) 2015 Intel Corporation. All rights reserved. */
7d3dcf26	3	#include <linux/device.h>
92281dee	4	#include <linux/io.h>
0207df4f	5	#include <linux/kasan.h>
41e94a85	6	#include <linux/memory_hotplug.h>
bcfa4b72 MW	7	#include <linux/mm.h>
bcfa4b72 MW	8	#include <linux/pfn_t.h>
5042db43 JG	9	#include <linux/swap.h>
5042db43 JG	10	#include <linux/swapops.h>
bcfa4b72	11	#include <linux/types.h>
e7638488	12	#include <linux/wait_bit.h>
bcfa4b72	13	#include <linux/xarray.h>
063a7d1d	14	#include <linux/hmm.h>
92281dee	15
bcfa4b72	16	static DEFINE_XARRAY(pgmap_array);
9476df7d DW	17	#define SECTION_MASK ~((1UL << PA_SECTION_SHIFT) - 1)
	18	#define SECTION_SIZE (1UL << PA_SECTION_SHIFT)
	19
5042db43	20	#if IS_ENABLED(CONFIG_DEVICE_PRIVATE)
2b740303	21	vm_fault_t device_private_entry_fault(struct vm_area_struct *vma,
5042db43 JG	22	unsigned long addr,
	23	swp_entry_t entry,
	24	unsigned int flags,
	25	pmd_t *pmdp)
	26	{
	27	struct page *page = device_private_entry_to_page(entry);
063a7d1d DW	28	struct hmm_devmem *devmem;
	29
	30	devmem = container_of(page->pgmap, typeof(*devmem), pagemap);
5042db43 JG	31
	32	/*
	33	* The page_fault() callback must migrate page back to system memory
	34	* so that CPU can access it. This might fail for various reasons
	35	* (device issue, device was unsafely unplugged, ...). When such
	36	* error conditions happen, the callback must return VM_FAULT_SIGBUS.
	37	*
	38	* Note that because memory cgroup charges are accounted to the device
	39	* memory, this should never fail because of memory restrictions (but
	40	* allocation of regular system page might still fail because we are
	41	* out of memory).
	42	*
	43	* There is a more in-depth description of what that callback can and
	44	* cannot do, in include/linux/memremap.h
	45	*/
063a7d1d	46	return devmem->page_fault(vma, addr, page, flags, pmdp);
5042db43	47	}
5042db43 JG	48	#endif /* CONFIG_DEVICE_PRIVATE */
5042db43 JG	49
bcfa4b72	50	static void pgmap_array_delete(struct resource *res)
ab1b597e	51	{
bcfa4b72 MW	52	xa_store_range(&pgmap_array, PHYS_PFN(res->start), PHYS_PFN(res->end),
bcfa4b72 MW	53	NULL, GFP_KERNEL);
ab1b597e	54	synchronize_rcu();
9476df7d DW	55	}
9476df7d DW	56
e7744aa2	57	static unsigned long pfn_first(struct dev_pagemap *pgmap)
5c2c2587	58	{
e7744aa2 LG	59	const struct resource *res = &pgmap->res;
e7744aa2 LG	60	struct vmem_altmap *altmap = &pgmap->altmap;
5c2c2587 DW	61	unsigned long pfn;
	62
	63	pfn = res->start >> PAGE_SHIFT;
e7744aa2	64	if (pgmap->altmap_valid)
5c2c2587 DW	65	pfn += vmem_altmap_offset(altmap);
	66	return pfn;
	67	}
	68
e7744aa2	69	static unsigned long pfn_end(struct dev_pagemap *pgmap)
5c2c2587	70	{
e7744aa2	71	const struct resource *res = &pgmap->res;
5c2c2587 DW	72
	73	return (res->start + resource_size(res)) >> PAGE_SHIFT;
	74	}
	75
949b9325 DW	76	static unsigned long pfn_next(unsigned long pfn)
	77	{
	78	if (pfn % 1024 == 0)
	79	cond_resched();
	80	return pfn + 1;
	81	}
	82
5c2c2587	83	#define for_each_device_pfn(pfn, map) \
949b9325	84	for (pfn = pfn_first(map); pfn < pfn_end(map); pfn = pfn_next(pfn))
5c2c2587	85
e8d51348	86	static void devm_memremap_pages_release(void *data)
41e94a85	87	{
e7744aa2	88	struct dev_pagemap *pgmap = data;
e8d51348	89	struct device *dev = pgmap->dev;
e7744aa2	90	struct resource *res = &pgmap->res;
9476df7d	91	resource_size_t align_start, align_size;
71389703	92	unsigned long pfn;
2c2a5af6	93	int nid;
71389703	94
a95c90f1	95	pgmap->kill(pgmap->ref);
e7744aa2	96	for_each_device_pfn(pfn, pgmap)
71389703	97	put_page(pfn_to_page(pfn));
50f44ee7	98	pgmap->cleanup(pgmap->ref);
9476df7d	99
41e94a85	100	/* pages are dead and unused, undo the arch mapping */
9476df7d	101	align_start = res->start & ~(SECTION_SIZE - 1);
10a0cd6e JS	102	align_size = ALIGN(res->start + resource_size(res), SECTION_SIZE)
10a0cd6e JS	103	- align_start;
b5d24fda	104
2c2a5af6 OS	105	nid = page_to_nid(pfn_to_page(align_start >> PAGE_SHIFT));
2c2a5af6 OS	106
f931ab47	107	mem_hotplug_begin();
69324b8f DW	108	if (pgmap->type == MEMORY_DEVICE_PRIVATE) {
	109	pfn = align_start >> PAGE_SHIFT;
	110	__remove_pages(page_zone(pfn_to_page(pfn)), pfn,
	111	align_size >> PAGE_SHIFT, NULL);
	112	} else {
2c2a5af6	113	arch_remove_memory(nid, align_start, align_size,
69324b8f DW	114	pgmap->altmap_valid ? &pgmap->altmap : NULL);
	115	kasan_remove_zero_shadow(__va(align_start), align_size);
	116	}
f931ab47	117	mem_hotplug_done();
b5d24fda	118
9049771f	119	untrack_pfn(NULL, PHYS_PFN(align_start), align_size);
bcfa4b72	120	pgmap_array_delete(res);
e7744aa2 LG	121	dev_WARN_ONCE(dev, pgmap->altmap.alloc,
e7744aa2 LG	122	"%s: failed to free all reserved pages\n", __func__);
9476df7d DW	123	}
9476df7d DW	124
4b94ffdc DW	125	/**
	126	* devm_memremap_pages - remap and provide memmap backing for the given resource
	127	* @dev: hosting device for @res
a95c90f1	128	* @pgmap: pointer to a struct dev_pagemap
4b94ffdc	129	*
5c2c2587	130	* Notes:
e8d51348 CH	131	* 1/ At a minimum the res, ref and type members of @pgmap must be initialized
	132	* by the caller before passing it to this function
	133	*
	134	* 2/ The altmap field may optionally be initialized, in which case altmap_valid
	135	* must be set to true
	136	*
50f44ee7 DW	137	* 3/ pgmap->ref must be 'live' on entry and will be killed and reaped
50f44ee7 DW	138	* at devm_memremap_pages_release() time, or if this routine fails.
5c2c2587	139	*
e8d51348	140	* 4/ res is expected to be a host memory range that could feasibly be
5c2c2587 DW	141	* treated as a "System RAM" range, i.e. not a device mmio range, but
5c2c2587 DW	142	* this is not enforced.
4b94ffdc	143	*/
e8d51348	144	void devm_memremap_pages(struct device dev, struct dev_pagemap *pgmap)
41e94a85	145	{
ab1b597e	146	resource_size_t align_start, align_size, align_end;
e8d51348 CH	147	struct vmem_altmap *altmap = pgmap->altmap_valid ?
e8d51348 CH	148	&pgmap->altmap : NULL;
949b9325	149	struct resource *res = &pgmap->res;
966cf44f	150	struct dev_pagemap *conflict_pgmap;
940519f0 MH	151	struct mhp_restrictions restrictions = {
	152	/*
	153	* We do not want any optional features only our own memmap
	154	*/
	155	.altmap = altmap,
	156	};
9049771f	157	pgprot_t pgprot = PAGE_KERNEL;
949b9325	158	int error, nid, is_ram;
5f29a77c	159
50f44ee7 DW	160	if (!pgmap->ref \|\| !pgmap->kill \|\| !pgmap->cleanup) {
50f44ee7 DW	161	WARN(1, "Missing reference count teardown definition\n");
a95c90f1	162	return ERR_PTR(-EINVAL);
50f44ee7	163	}
a95c90f1	164
5f29a77c DW	165	align_start = res->start & ~(SECTION_SIZE - 1);
	166	align_size = ALIGN(res->start + resource_size(res), SECTION_SIZE)
	167	- align_start;
15d36fec DJ	168	align_end = align_start + align_size - 1;
	169
	170	conflict_pgmap = get_dev_pagemap(PHYS_PFN(align_start), NULL);
	171	if (conflict_pgmap) {
	172	dev_WARN(dev, "Conflicting mapping in same section\n");
	173	put_dev_pagemap(conflict_pgmap);
50f44ee7 DW	174	error = -ENOMEM;
50f44ee7 DW	175	goto err_array;
15d36fec DJ	176	}
	177
	178	conflict_pgmap = get_dev_pagemap(PHYS_PFN(align_end), NULL);
	179	if (conflict_pgmap) {
	180	dev_WARN(dev, "Conflicting mapping in same section\n");
	181	put_dev_pagemap(conflict_pgmap);
50f44ee7 DW	182	error = -ENOMEM;
50f44ee7 DW	183	goto err_array;
15d36fec DJ	184	}
15d36fec DJ	185
d37a14bb LT	186	is_ram = region_intersects(align_start, align_size,
d37a14bb LT	187	IORESOURCE_SYSTEM_RAM, IORES_DESC_NONE);
41e94a85	188
06489cfb DW	189	if (is_ram != REGION_DISJOINT) {
	190	WARN_ONCE(1, "%s attempted on %s region %pr\n", __func__,
	191	is_ram == REGION_MIXED ? "mixed" : "ram", res);
a95c90f1 DW	192	error = -ENXIO;
a95c90f1 DW	193	goto err_array;
41e94a85 CH	194	}
41e94a85 CH	195
4b94ffdc	196	pgmap->dev = dev;
4b94ffdc	197
bcfa4b72 MW	198	error = xa_err(xa_store_range(&pgmap_array, PHYS_PFN(res->start),
bcfa4b72 MW	199	PHYS_PFN(res->end), pgmap, GFP_KERNEL));
9476df7d	200	if (error)
bcfa4b72	201	goto err_array;
9476df7d	202
41e94a85 CH	203	nid = dev_to_node(dev);
41e94a85 CH	204	if (nid < 0)
7eff93b7	205	nid = numa_mem_id();
41e94a85	206
9049771f DW	207	error = track_pfn_remap(NULL, &pgprot, PHYS_PFN(align_start), 0,
	208	align_size);
	209	if (error)
	210	goto err_pfn_remap;
	211
f931ab47	212	mem_hotplug_begin();
69324b8f DW	213
	214	/*
	215	* For device private memory we call add_pages() as we only need to
	216	* allocate and initialize struct page for the device memory. More-
	217	* over the device memory is un-accessible thus we do not want to
	218	* create a linear mapping for the memory like arch_add_memory()
	219	* would do.
	220	*
	221	* For all other device memory types, which are accessible by
	222	* the CPU, we do want the linear mapping and thus use
	223	* arch_add_memory().
	224	*/
	225	if (pgmap->type == MEMORY_DEVICE_PRIVATE) {
	226	error = add_pages(nid, align_start >> PAGE_SHIFT,
940519f0	227	align_size >> PAGE_SHIFT, &restrictions);
69324b8f DW	228	} else {
	229	error = kasan_add_zero_shadow(__va(align_start), align_size);
	230	if (error) {
	231	mem_hotplug_done();
	232	goto err_kasan;
	233	}
	234
940519f0 MH	235	error = arch_add_memory(nid, align_start, align_size,
940519f0 MH	236	&restrictions);
69324b8f DW	237	}
	238
	239	if (!error) {
	240	struct zone *zone;
	241
	242	zone = &NODE_DATA(nid)->node_zones[ZONE_DEVICE];
	243	move_pfn_range_to_zone(zone, align_start >> PAGE_SHIFT,
	244	align_size >> PAGE_SHIFT, altmap);
0207df4f AR	245	}
0207df4f AR	246
f931ab47	247	mem_hotplug_done();
9476df7d DW	248	if (error)
9476df7d DW	249	goto err_add_memory;
41e94a85	250
966cf44f AD	251	/*
	252	* Initialization of the pages has been deferred until now in order
	253	* to allow us to do the work while not holding the hotplug lock.
	254	*/
	255	memmap_init_zone_device(&NODE_DATA(nid)->node_zones[ZONE_DEVICE],
	256	align_start >> PAGE_SHIFT,
	257	align_size >> PAGE_SHIFT, pgmap);
	258	percpu_ref_get_many(pgmap->ref, pfn_end(pgmap) - pfn_first(pgmap));
e8d51348	259
a95c90f1 DW	260	error = devm_add_action_or_reset(dev, devm_memremap_pages_release,
	261	pgmap);
	262	if (error)
	263	return ERR_PTR(error);
e8d51348	264
41e94a85	265	return __va(res->start);
9476df7d DW	266
9476df7d DW	267	err_add_memory:
0207df4f AR	268	kasan_remove_zero_shadow(__va(align_start), align_size);
0207df4f AR	269	err_kasan:
9049771f DW	270	untrack_pfn(NULL, PHYS_PFN(align_start), align_size);
9049771f DW	271	err_pfn_remap:
bcfa4b72 MW	272	pgmap_array_delete(res);
bcfa4b72 MW	273	err_array:
a95c90f1	274	pgmap->kill(pgmap->ref);
50f44ee7 DW	275	pgmap->cleanup(pgmap->ref);
50f44ee7 DW	276
9476df7d	277	return ERR_PTR(error);
41e94a85	278	}
808153e1	279	EXPORT_SYMBOL_GPL(devm_memremap_pages);
4b94ffdc	280
2e3f139e DW	281	void devm_memunmap_pages(struct device dev, struct dev_pagemap pgmap)
	282	{
	283	devm_release_action(dev, devm_memremap_pages_release, pgmap);
	284	}
	285	EXPORT_SYMBOL_GPL(devm_memunmap_pages);
	286
4b94ffdc DW	287	unsigned long vmem_altmap_offset(struct vmem_altmap *altmap)
	288	{
	289	/* number of pfns from base where pfn_to_page() is valid */
	290	return altmap->reserve + altmap->free;
	291	}
	292
	293	void vmem_altmap_free(struct vmem_altmap *altmap, unsigned long nr_pfns)
	294	{
	295	altmap->alloc -= nr_pfns;
	296	}
	297
0822acb8 CH	298	/**
	299	* get_dev_pagemap() - take a new live reference on the dev_pagemap for @pfn
	300	* @pfn: page frame number to lookup page_map
	301	* @pgmap: optional known pgmap that already has a reference
	302	*
832d7aa0 CH	303	* If @pgmap is non-NULL and covers @pfn it will be returned as-is. If @pgmap
832d7aa0 CH	304	* is non-NULL but does not cover @pfn the reference to it will be released.
0822acb8 CH	305	*/
	306	struct dev_pagemap *get_dev_pagemap(unsigned long pfn,
	307	struct dev_pagemap *pgmap)
	308	{
0822acb8 CH	309	resource_size_t phys = PFN_PHYS(pfn);
	310
	311	/*
832d7aa0	312	* In the cached case we're already holding a live reference.
0822acb8	313	*/
832d7aa0	314	if (pgmap) {
e7744aa2	315	if (phys >= pgmap->res.start && phys <= pgmap->res.end)
832d7aa0 CH	316	return pgmap;
832d7aa0 CH	317	put_dev_pagemap(pgmap);
0822acb8 CH	318	}
	319
	320	/* fall back to slow path lookup */
	321	rcu_read_lock();
bcfa4b72	322	pgmap = xa_load(&pgmap_array, PHYS_PFN(phys));
0822acb8 CH	323	if (pgmap && !percpu_ref_tryget_live(pgmap->ref))
	324	pgmap = NULL;
	325	rcu_read_unlock();
	326
	327	return pgmap;
	328	}
e7638488	329	EXPORT_SYMBOL_GPL(get_dev_pagemap);
7b2d55d2	330
e7638488 DW	331	#ifdef CONFIG_DEV_PAGEMAP_OPS
e7638488 DW	332	DEFINE_STATIC_KEY_FALSE(devmap_managed_key);
31c5bda3	333	EXPORT_SYMBOL(devmap_managed_key);
e7638488 DW	334	static atomic_t devmap_enable;
	335
	336	/*
	337	* Toggle the static key for ->page_free() callbacks when dev_pagemap
	338	* pages go idle.
	339	*/
	340	void dev_pagemap_get_ops(void)
	341	{
	342	if (atomic_inc_return(&devmap_enable) == 1)
	343	static_branch_enable(&devmap_managed_key);
	344	}
	345	EXPORT_SYMBOL_GPL(dev_pagemap_get_ops);
	346
	347	void dev_pagemap_put_ops(void)
	348	{
	349	if (atomic_dec_and_test(&devmap_enable))
	350	static_branch_disable(&devmap_managed_key);
	351	}
	352	EXPORT_SYMBOL_GPL(dev_pagemap_put_ops);
	353
	354	void __put_devmap_managed_page(struct page *page)
7b2d55d2 JG	355	{
	356	int count = page_ref_dec_return(page);
	357
	358	/*
	359	* If refcount is 1 then page is freed and refcount is stable as nobody
	360	* holds a reference on the page.
	361	*/
	362	if (count == 1) {
	363	/* Clear Active bit in case of parallel mark_page_accessed */
	364	__ClearPageActive(page);
	365	__ClearPageWaiters(page);
	366
c733a828	367	mem_cgroup_uncharge(page);
7b2d55d2 JG	368
	369	page->pgmap->page_free(page, page->pgmap->data);
	370	} else if (!count)
	371	__put_page(page);
	372	}
31c5bda3	373	EXPORT_SYMBOL(__put_devmap_managed_page);
e7638488	374	#endif /* CONFIG_DEV_PAGEMAP_OPS */