[mirror_ubuntu-jammy-kernel.git] / mm / 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/mmzone.h>
#include <linux/swapops.h>
#include <linux/types.h>
#include <linux/wait_bit.h>
#include <linux/xarray.h>

static DEFINE_XARRAY(pgmap_array);

/*
 * The memremap() and memremap_pages() interfaces are alternately used
 * to map persistent memory namespaces. These interfaces place different
 * constraints on the alignment and size of the mapping (namespace).
 * memremap() can map individual PAGE_SIZE pages. memremap_pages() can
 * only map subsections (2MB), and at least one architecture (PowerPC)
 * the minimum mapping granularity of memremap_pages() is 16MB.
 *
 * The role of memremap_compat_align() is to communicate the minimum
 * arch supported alignment of a namespace such that it can freely
 * switch modes without violating the arch constraint. Namely, do not
 * allow a namespace to be PAGE_SIZE aligned since that namespace may be
 * reconfigured into a mode that requires SUBSECTION_SIZE alignment.
 */
#ifndef CONFIG_ARCH_HAS_MEMREMAP_COMPAT_ALIGN
unsigned long memremap_compat_align(void)
{
	return SUBSECTION_SIZE;
}
EXPORT_SYMBOL_GPL(memremap_compat_align);
#endif

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

static void devmap_managed_enable_put(void)
{
	if (atomic_dec_and_test(&devmap_managed_enable))
		static_branch_disable(&devmap_managed_key);
}

static int devmap_managed_enable_get(struct dev_pagemap *pgmap)
{
	if (pgmap->type == MEMORY_DEVICE_PRIVATE &&
	    (!pgmap->ops || !pgmap->ops->page_free)) {
		WARN(1, "Missing page_free method\n");
		return -EINVAL;
	}

	if (atomic_inc_return(&devmap_managed_enable) == 1)
		static_branch_enable(&devmap_managed_key);
	return 0;
}
#else
static int devmap_managed_enable_get(struct dev_pagemap *pgmap)
{
	return -EINVAL;
}
static void devmap_managed_enable_put(void)
{
}
#endif /* CONFIG_DEV_PAGEMAP_OPS */

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)
{
	return PHYS_PFN(pgmap->res.start) +
		vmem_altmap_offset(pgmap_altmap(pgmap));
}

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 dev_pagemap_kill(struct dev_pagemap *pgmap)
{
	if (pgmap->ops && pgmap->ops->kill)
		pgmap->ops->kill(pgmap);
	else
		percpu_ref_kill(pgmap->ref);
}

static void dev_pagemap_cleanup(struct dev_pagemap *pgmap)
{
	if (pgmap->ops && pgmap->ops->cleanup) {
		pgmap->ops->cleanup(pgmap);
	} else {
		wait_for_completion(&pgmap->done);
		percpu_ref_exit(pgmap->ref);
	}
	/*
	 * Undo the pgmap ref assignment for the internal case as the
	 * caller may re-enable the same pgmap.
	 */
	if (pgmap->ref == &pgmap->internal_ref)
		pgmap->ref = NULL;
}

void memunmap_pages(struct dev_pagemap *pgmap)
{
	struct resource *res = &pgmap->res;
	struct page *first_page;
	unsigned long pfn;
	int nid;

	dev_pagemap_kill(pgmap);
	for_each_device_pfn(pfn, pgmap)
		put_page(pfn_to_page(pfn));
	dev_pagemap_cleanup(pgmap);

	/* make sure to access a memmap that was actually initialized */
	first_page = pfn_to_page(pfn_first(pgmap));

	/* pages are dead and unused, undo the arch mapping */
	nid = page_to_nid(first_page);

	mem_hotplug_begin();
	remove_pfn_range_from_zone(page_zone(first_page), PHYS_PFN(res->start),
				   PHYS_PFN(resource_size(res)));
	if (pgmap->type == MEMORY_DEVICE_PRIVATE) {
		__remove_pages(PHYS_PFN(res->start),
			       PHYS_PFN(resource_size(res)), NULL);
	} else {
		arch_remove_memory(nid, res->start, resource_size(res),
				pgmap_altmap(pgmap));
		kasan_remove_zero_shadow(__va(res->start), resource_size(res));
	}
	mem_hotplug_done();

	untrack_pfn(NULL, PHYS_PFN(res->start), resource_size(res));
	pgmap_array_delete(res);
	WARN_ONCE(pgmap->altmap.alloc, "failed to free all reserved pages\n");
	devmap_managed_enable_put();
}
EXPORT_SYMBOL_GPL(memunmap_pages);

static void devm_memremap_pages_release(void *data)
{
	memunmap_pages(data);
}

static void dev_pagemap_percpu_release(struct percpu_ref *ref)
{
	struct dev_pagemap *pgmap =
		container_of(ref, struct dev_pagemap, internal_ref);

	complete(&pgmap->done);
}

/*
 * Not device managed version of dev_memremap_pages, undone by
 * memunmap_pages().  Please use dev_memremap_pages if you have a struct
 * device available.
 */
void *memremap_pages(struct dev_pagemap *pgmap, int nid)
{
	struct resource *res = &pgmap->res;
	struct dev_pagemap *conflict_pgmap;
	struct mhp_params params = {
		/*
		 * We do not want any optional features only our own memmap
		 */
		.altmap = pgmap_altmap(pgmap),
		.pgprot = PAGE_KERNEL,
	};
	int error, is_ram;
	bool need_devmap_managed = true;

	switch (pgmap->type) {
	case MEMORY_DEVICE_PRIVATE:
		if (!IS_ENABLED(CONFIG_DEVICE_PRIVATE)) {
			WARN(1, "Device private memory not supported\n");
			return ERR_PTR(-EINVAL);
		}
		if (!pgmap->ops || !pgmap->ops->migrate_to_ram) {
			WARN(1, "Missing migrate_to_ram method\n");
			return ERR_PTR(-EINVAL);
		}
		if (!pgmap->owner) {
			WARN(1, "Missing owner\n");
			return ERR_PTR(-EINVAL);
		}
		break;
	case MEMORY_DEVICE_FS_DAX:
		if (!IS_ENABLED(CONFIG_ZONE_DEVICE) ||
		    IS_ENABLED(CONFIG_FS_DAX_LIMITED)) {
			WARN(1, "File system DAX not supported\n");
			return ERR_PTR(-EINVAL);
		}
		break;
	case MEMORY_DEVICE_DEVDAX:
	case MEMORY_DEVICE_PCI_P2PDMA:
		need_devmap_managed = false;
		break;
	default:
		WARN(1, "Invalid pgmap type %d\n", pgmap->type);
		break;
	}

	if (!pgmap->ref) {
		if (pgmap->ops && (pgmap->ops->kill || pgmap->ops->cleanup))
			return ERR_PTR(-EINVAL);

		init_completion(&pgmap->done);
		error = percpu_ref_init(&pgmap->internal_ref,
				dev_pagemap_percpu_release, 0, GFP_KERNEL);
		if (error)
			return ERR_PTR(error);
		pgmap->ref = &pgmap->internal_ref;
	} else {
		if (!pgmap->ops || !pgmap->ops->kill || !pgmap->ops->cleanup) {
			WARN(1, "Missing reference count teardown definition\n");
			return ERR_PTR(-EINVAL);
		}
	}

	if (need_devmap_managed) {
		error = devmap_managed_enable_get(pgmap);
		if (error)
			return ERR_PTR(error);
	}

	conflict_pgmap = get_dev_pagemap(PHYS_PFN(res->start), NULL);
	if (conflict_pgmap) {
		WARN(1, "Conflicting mapping in same section\n");
		put_dev_pagemap(conflict_pgmap);
		error = -ENOMEM;
		goto err_array;
	}

	conflict_pgmap = get_dev_pagemap(PHYS_PFN(res->end), NULL);
	if (conflict_pgmap) {
		WARN(1, "Conflicting mapping in same section\n");
		put_dev_pagemap(conflict_pgmap);
		error = -ENOMEM;
		goto err_array;
	}

	is_ram = region_intersects(res->start, resource_size(res),
		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;
	}

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

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

	error = track_pfn_remap(NULL, &params.pgprot, PHYS_PFN(res->start),
				0, resource_size(res));
	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, PHYS_PFN(res->start),
				PHYS_PFN(resource_size(res)), &params);
	} else {
		error = kasan_add_zero_shadow(__va(res->start), resource_size(res));
		if (error) {
			mem_hotplug_done();
			goto err_kasan;
		}

		error = arch_add_memory(nid, res->start, resource_size(res),
					&params);
	}

	if (!error) {
		struct zone *zone;

		zone = &NODE_DATA(nid)->node_zones[ZONE_DEVICE];
		move_pfn_range_to_zone(zone, PHYS_PFN(res->start),
				PHYS_PFN(resource_size(res)), params.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],
				PHYS_PFN(res->start),
				PHYS_PFN(resource_size(res)), pgmap);
	percpu_ref_get_many(pgmap->ref, pfn_end(pgmap) - pfn_first(pgmap));
	return __va(res->start);

 err_add_memory:
	kasan_remove_zero_shadow(__va(res->start), resource_size(res));
 err_kasan:
	untrack_pfn(NULL, PHYS_PFN(res->start), resource_size(res));
 err_pfn_remap:
	pgmap_array_delete(res);
 err_array:
	dev_pagemap_kill(pgmap);
	dev_pagemap_cleanup(pgmap);
	devmap_managed_enable_put();
	return ERR_PTR(error);
}
EXPORT_SYMBOL_GPL(memremap_pages);

/**
 * 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 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
 *    PGMAP_ALTMAP_VALID must be set in pgmap->flags.
 *
 * 3/ The ref field may optionally be provided, in which 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)
{
	int error;
	void *ret;

	ret = memremap_pages(pgmap, dev_to_node(dev));
	if (IS_ERR(ret))
		return ret;

	error = devm_add_action_or_reset(dev, devm_memremap_pages_release,
			pgmap);
	if (error)
		return ERR_PTR(error);
	return ret;
}
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 */
	if (altmap)
		return altmap->reserve + altmap->free;
	return 0;
}

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
void free_devmap_managed_page(struct page *page)
{
	/* notify page idle for dax */
	if (!is_device_private_page(page)) {
		wake_up_var(&page->_refcount);
		return;
	}

	/* Clear Active bit in case of parallel mark_page_accessed */
	__ClearPageActive(page);
	__ClearPageWaiters(page);

	mem_cgroup_uncharge(page);

	/*
	 * When a device_private page is freed, the page->mapping field
	 * may still contain a (stale) mapping value. For example, the
	 * lower bits of page->mapping may still identify the page as an
	 * anonymous page. Ultimately, this entire field is just stale
	 * and wrong, and it will cause errors if not cleared.  One
	 * example is:
	 *
	 *  migrate_vma_pages()
	 *    migrate_vma_insert_page()
	 *      page_add_new_anon_rmap()
	 *        __page_set_anon_rmap()
	 *          ...checks page->mapping, via PageAnon(page) call,
	 *            and incorrectly concludes that the page is an
	 *            anonymous page. Therefore, it incorrectly,
	 *            silently fails to set up the new anon rmap.
	 *
	 * For other types of ZONE_DEVICE pages, migration is either
	 * handled differently or not done at all, so there is no need
	 * to clear page->mapping.
	 */
	page->mapping = NULL;
	page->pgmap->ops->page_free(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	9	#include <linux/swap.h>
9ffc1d19	10	#include <linux/mmzone.h>
5042db43	11	#include <linux/swapops.h>
bcfa4b72	12	#include <linux/types.h>
e7638488	13	#include <linux/wait_bit.h>
bcfa4b72	14	#include <linux/xarray.h>
92281dee	15
bcfa4b72	16	static DEFINE_XARRAY(pgmap_array);
9476df7d	17
9ffc1d19 DW	18	/*
	19	* The memremap() and memremap_pages() interfaces are alternately used
	20	* to map persistent memory namespaces. These interfaces place different
	21	* constraints on the alignment and size of the mapping (namespace).
	22	* memremap() can map individual PAGE_SIZE pages. memremap_pages() can
	23	* only map subsections (2MB), and at least one architecture (PowerPC)
	24	* the minimum mapping granularity of memremap_pages() is 16MB.
	25	*
	26	* The role of memremap_compat_align() is to communicate the minimum
	27	* arch supported alignment of a namespace such that it can freely
	28	* switch modes without violating the arch constraint. Namely, do not
	29	* allow a namespace to be PAGE_SIZE aligned since that namespace may be
	30	* reconfigured into a mode that requires SUBSECTION_SIZE alignment.
	31	*/
	32	#ifndef CONFIG_ARCH_HAS_MEMREMAP_COMPAT_ALIGN
	33	unsigned long memremap_compat_align(void)
	34	{
	35	return SUBSECTION_SIZE;
	36	}
	37	EXPORT_SYMBOL_GPL(memremap_compat_align);
	38	#endif
	39
f6a55e1a CH	40	#ifdef CONFIG_DEV_PAGEMAP_OPS
	41	DEFINE_STATIC_KEY_FALSE(devmap_managed_key);
	42	EXPORT_SYMBOL(devmap_managed_key);
	43	static atomic_t devmap_managed_enable;
	44
6f42193f	45	static void devmap_managed_enable_put(void)
f6a55e1a CH	46	{
	47	if (atomic_dec_and_test(&devmap_managed_enable))
	48	static_branch_disable(&devmap_managed_key);
	49	}
	50
6f42193f	51	static int devmap_managed_enable_get(struct dev_pagemap *pgmap)
f6a55e1a	52	{
429589d6 DW	53	if (pgmap->type == MEMORY_DEVICE_PRIVATE &&
429589d6 DW	54	(!pgmap->ops \|\| !pgmap->ops->page_free)) {
f6a55e1a CH	55	WARN(1, "Missing page_free method\n");
	56	return -EINVAL;
	57	}
	58
	59	if (atomic_inc_return(&devmap_managed_enable) == 1)
	60	static_branch_enable(&devmap_managed_key);
6f42193f	61	return 0;
f6a55e1a CH	62	}
f6a55e1a CH	63	#else
6f42193f	64	static int devmap_managed_enable_get(struct dev_pagemap *pgmap)
f6a55e1a CH	65	{
	66	return -EINVAL;
	67	}
6f42193f CH	68	static void devmap_managed_enable_put(void)
	69	{
	70	}
f6a55e1a CH	71	#endif /* CONFIG_DEV_PAGEMAP_OPS */
f6a55e1a CH	72
bcfa4b72	73	static void pgmap_array_delete(struct resource *res)
ab1b597e	74	{
bcfa4b72 MW	75	xa_store_range(&pgmap_array, PHYS_PFN(res->start), PHYS_PFN(res->end),
bcfa4b72 MW	76	NULL, GFP_KERNEL);
ab1b597e	77	synchronize_rcu();
9476df7d DW	78	}
9476df7d DW	79
e7744aa2	80	static unsigned long pfn_first(struct dev_pagemap *pgmap)
5c2c2587	81	{
7cc7867f	82	return PHYS_PFN(pgmap->res.start) +
514caf23	83	vmem_altmap_offset(pgmap_altmap(pgmap));
5c2c2587 DW	84	}
5c2c2587 DW	85
e7744aa2	86	static unsigned long pfn_end(struct dev_pagemap *pgmap)
5c2c2587	87	{
e7744aa2	88	const struct resource *res = &pgmap->res;
5c2c2587 DW	89
	90	return (res->start + resource_size(res)) >> PAGE_SHIFT;
	91	}
	92
949b9325 DW	93	static unsigned long pfn_next(unsigned long pfn)
	94	{
	95	if (pfn % 1024 == 0)
	96	cond_resched();
	97	return pfn + 1;
	98	}
	99
5c2c2587	100	#define for_each_device_pfn(pfn, map) \
949b9325	101	for (pfn = pfn_first(map); pfn < pfn_end(map); pfn = pfn_next(pfn))
5c2c2587	102
24917f6b CH	103	static void dev_pagemap_kill(struct dev_pagemap *pgmap)
	104	{
	105	if (pgmap->ops && pgmap->ops->kill)
	106	pgmap->ops->kill(pgmap);
	107	else
	108	percpu_ref_kill(pgmap->ref);
	109	}
	110
	111	static void dev_pagemap_cleanup(struct dev_pagemap *pgmap)
	112	{
	113	if (pgmap->ops && pgmap->ops->cleanup) {
	114	pgmap->ops->cleanup(pgmap);
	115	} else {
	116	wait_for_completion(&pgmap->done);
	117	percpu_ref_exit(pgmap->ref);
	118	}
06282373 DW	119	/*
	120	* Undo the pgmap ref assignment for the internal case as the
	121	* caller may re-enable the same pgmap.
	122	*/
	123	if (pgmap->ref == &pgmap->internal_ref)
	124	pgmap->ref = NULL;
24917f6b CH	125	}
24917f6b CH	126
6869b7b2	127	void memunmap_pages(struct dev_pagemap *pgmap)
41e94a85	128	{
e7744aa2	129	struct resource *res = &pgmap->res;
77e080e7	130	struct page *first_page;
71389703	131	unsigned long pfn;
2c2a5af6	132	int nid;
71389703	133
24917f6b	134	dev_pagemap_kill(pgmap);
e7744aa2	135	for_each_device_pfn(pfn, pgmap)
71389703	136	put_page(pfn_to_page(pfn));
24917f6b	137	dev_pagemap_cleanup(pgmap);
9476df7d	138
77e080e7 AK	139	/* make sure to access a memmap that was actually initialized */
	140	first_page = pfn_to_page(pfn_first(pgmap));
	141
41e94a85	142	/* pages are dead and unused, undo the arch mapping */
77e080e7	143	nid = page_to_nid(first_page);
2c2a5af6	144
f931ab47	145	mem_hotplug_begin();
d33695b1 DH	146	remove_pfn_range_from_zone(page_zone(first_page), PHYS_PFN(res->start),
d33695b1 DH	147	PHYS_PFN(resource_size(res)));
69324b8f	148	if (pgmap->type == MEMORY_DEVICE_PRIVATE) {
feee6b29	149	__remove_pages(PHYS_PFN(res->start),
77e080e7	150	PHYS_PFN(resource_size(res)), NULL);
69324b8f	151	} else {
7cc7867f	152	arch_remove_memory(nid, res->start, resource_size(res),
514caf23	153	pgmap_altmap(pgmap));
7cc7867f	154	kasan_remove_zero_shadow(__va(res->start), resource_size(res));
69324b8f	155	}
f931ab47	156	mem_hotplug_done();
b5d24fda	157
7cc7867f	158	untrack_pfn(NULL, PHYS_PFN(res->start), resource_size(res));
bcfa4b72	159	pgmap_array_delete(res);
fdc029b1	160	WARN_ONCE(pgmap->altmap.alloc, "failed to free all reserved pages\n");
6f42193f	161	devmap_managed_enable_put();
9476df7d	162	}
6869b7b2 CH	163	EXPORT_SYMBOL_GPL(memunmap_pages);
	164
	165	static void devm_memremap_pages_release(void *data)
	166	{
	167	memunmap_pages(data);
	168	}
9476df7d	169
24917f6b CH	170	static void dev_pagemap_percpu_release(struct percpu_ref *ref)
	171	{
	172	struct dev_pagemap *pgmap =
	173	container_of(ref, struct dev_pagemap, internal_ref);
	174
	175	complete(&pgmap->done);
	176	}
	177
6869b7b2 CH	178	/*
	179	* Not device managed version of dev_memremap_pages, undone by
	180	* memunmap_pages(). Please use dev_memremap_pages if you have a struct
	181	* device available.
4b94ffdc	182	*/
6869b7b2	183	void memremap_pages(struct dev_pagemap pgmap, int nid)
41e94a85	184	{
949b9325	185	struct resource *res = &pgmap->res;
966cf44f	186	struct dev_pagemap *conflict_pgmap;
f5637d3b	187	struct mhp_params params = {
940519f0 MH	188	/*
940519f0 MH	189	* We do not want any optional features only our own memmap
7cc7867f	190	*/
514caf23	191	.altmap = pgmap_altmap(pgmap),
bfeb022f	192	.pgprot = PAGE_KERNEL,
940519f0	193	};
6869b7b2	194	int error, is_ram;
f6a55e1a	195	bool need_devmap_managed = true;
5f29a77c	196
3ed2dcdf CH	197	switch (pgmap->type) {
	198	case MEMORY_DEVICE_PRIVATE:
	199	if (!IS_ENABLED(CONFIG_DEVICE_PRIVATE)) {
	200	WARN(1, "Device private memory not supported\n");
	201	return ERR_PTR(-EINVAL);
	202	}
897e6365 CH	203	if (!pgmap->ops \|\| !pgmap->ops->migrate_to_ram) {
	204	WARN(1, "Missing migrate_to_ram method\n");
	205	return ERR_PTR(-EINVAL);
	206	}
f894ddd5 CH	207	if (!pgmap->owner) {
	208	WARN(1, "Missing owner\n");
	209	return ERR_PTR(-EINVAL);
	210	}
3ed2dcdf CH	211	break;
	212	case MEMORY_DEVICE_FS_DAX:
	213	if (!IS_ENABLED(CONFIG_ZONE_DEVICE) \|\|
	214	IS_ENABLED(CONFIG_FS_DAX_LIMITED)) {
	215	WARN(1, "File system DAX not supported\n");
	216	return ERR_PTR(-EINVAL);
	217	}
	218	break;
	219	case MEMORY_DEVICE_DEVDAX:
	220	case MEMORY_DEVICE_PCI_P2PDMA:
f6a55e1a	221	need_devmap_managed = false;
3ed2dcdf CH	222	break;
	223	default:
	224	WARN(1, "Invalid pgmap type %d\n", pgmap->type);
	225	break;
	226	}
	227
24917f6b CH	228	if (!pgmap->ref) {
	229	if (pgmap->ops && (pgmap->ops->kill \|\| pgmap->ops->cleanup))
	230	return ERR_PTR(-EINVAL);
	231
	232	init_completion(&pgmap->done);
	233	error = percpu_ref_init(&pgmap->internal_ref,
	234	dev_pagemap_percpu_release, 0, GFP_KERNEL);
	235	if (error)
	236	return ERR_PTR(error);
	237	pgmap->ref = &pgmap->internal_ref;
	238	} else {
	239	if (!pgmap->ops \|\| !pgmap->ops->kill \|\| !pgmap->ops->cleanup) {
	240	WARN(1, "Missing reference count teardown definition\n");
	241	return ERR_PTR(-EINVAL);
	242	}
50f44ee7	243	}
a95c90f1	244
f6a55e1a	245	if (need_devmap_managed) {
6f42193f	246	error = devmap_managed_enable_get(pgmap);
f6a55e1a CH	247	if (error)
	248	return ERR_PTR(error);
	249	}
	250
7cc7867f	251	conflict_pgmap = get_dev_pagemap(PHYS_PFN(res->start), NULL);
15d36fec	252	if (conflict_pgmap) {
6869b7b2	253	WARN(1, "Conflicting mapping in same section\n");
15d36fec	254	put_dev_pagemap(conflict_pgmap);
50f44ee7 DW	255	error = -ENOMEM;
50f44ee7 DW	256	goto err_array;
15d36fec DJ	257	}
15d36fec DJ	258
7cc7867f	259	conflict_pgmap = get_dev_pagemap(PHYS_PFN(res->end), NULL);
15d36fec	260	if (conflict_pgmap) {
6869b7b2	261	WARN(1, "Conflicting mapping in same section\n");
15d36fec	262	put_dev_pagemap(conflict_pgmap);
50f44ee7 DW	263	error = -ENOMEM;
50f44ee7 DW	264	goto err_array;
15d36fec DJ	265	}
15d36fec DJ	266
7cc7867f	267	is_ram = region_intersects(res->start, resource_size(res),
d37a14bb	268	IORESOURCE_SYSTEM_RAM, IORES_DESC_NONE);
41e94a85	269
06489cfb DW	270	if (is_ram != REGION_DISJOINT) {
	271	WARN_ONCE(1, "%s attempted on %s region %pr\n", __func__,
	272	is_ram == REGION_MIXED ? "mixed" : "ram", res);
a95c90f1 DW	273	error = -ENXIO;
a95c90f1 DW	274	goto err_array;
41e94a85 CH	275	}
41e94a85 CH	276
bcfa4b72 MW	277	error = xa_err(xa_store_range(&pgmap_array, PHYS_PFN(res->start),
bcfa4b72 MW	278	PHYS_PFN(res->end), pgmap, GFP_KERNEL));
9476df7d	279	if (error)
bcfa4b72	280	goto err_array;
9476df7d	281
41e94a85	282	if (nid < 0)
7eff93b7	283	nid = numa_mem_id();
41e94a85	284
bfeb022f LG	285	error = track_pfn_remap(NULL, &params.pgprot, PHYS_PFN(res->start),
bfeb022f LG	286	0, resource_size(res));
9049771f DW	287	if (error)
	288	goto err_pfn_remap;
	289
f931ab47	290	mem_hotplug_begin();
69324b8f DW	291
	292	/*
	293	* For device private memory we call add_pages() as we only need to
	294	* allocate and initialize struct page for the device memory. More-
	295	* over the device memory is un-accessible thus we do not want to
	296	* create a linear mapping for the memory like arch_add_memory()
	297	* would do.
	298	*
	299	* For all other device memory types, which are accessible by
	300	* the CPU, we do want the linear mapping and thus use
	301	* arch_add_memory().
	302	*/
	303	if (pgmap->type == MEMORY_DEVICE_PRIVATE) {
7cc7867f	304	error = add_pages(nid, PHYS_PFN(res->start),
f5637d3b	305	PHYS_PFN(resource_size(res)), &params);
69324b8f	306	} else {
7cc7867f	307	error = kasan_add_zero_shadow(__va(res->start), resource_size(res));
69324b8f DW	308	if (error) {
	309	mem_hotplug_done();
	310	goto err_kasan;
	311	}
	312
7cc7867f	313	error = arch_add_memory(nid, res->start, resource_size(res),
f5637d3b	314	&params);
69324b8f DW	315	}
	316
	317	if (!error) {
	318	struct zone *zone;
	319
	320	zone = &NODE_DATA(nid)->node_zones[ZONE_DEVICE];
7cc7867f	321	move_pfn_range_to_zone(zone, PHYS_PFN(res->start),
f5637d3b	322	PHYS_PFN(resource_size(res)), params.altmap);
0207df4f AR	323	}
0207df4f AR	324
f931ab47	325	mem_hotplug_done();
9476df7d DW	326	if (error)
9476df7d DW	327	goto err_add_memory;
41e94a85	328
966cf44f AD	329	/*
	330	* Initialization of the pages has been deferred until now in order
	331	* to allow us to do the work while not holding the hotplug lock.
	332	*/
	333	memmap_init_zone_device(&NODE_DATA(nid)->node_zones[ZONE_DEVICE],
7cc7867f DW	334	PHYS_PFN(res->start),
7cc7867f DW	335	PHYS_PFN(resource_size(res)), pgmap);
966cf44f	336	percpu_ref_get_many(pgmap->ref, pfn_end(pgmap) - pfn_first(pgmap));
41e94a85	337	return __va(res->start);
9476df7d DW	338
9476df7d DW	339	err_add_memory:
7cc7867f	340	kasan_remove_zero_shadow(__va(res->start), resource_size(res));
0207df4f	341	err_kasan:
7cc7867f	342	untrack_pfn(NULL, PHYS_PFN(res->start), resource_size(res));
9049771f	343	err_pfn_remap:
bcfa4b72 MW	344	pgmap_array_delete(res);
bcfa4b72 MW	345	err_array:
24917f6b CH	346	dev_pagemap_kill(pgmap);
24917f6b CH	347	dev_pagemap_cleanup(pgmap);
6f42193f	348	devmap_managed_enable_put();
9476df7d	349	return ERR_PTR(error);
41e94a85	350	}
6869b7b2 CH	351	EXPORT_SYMBOL_GPL(memremap_pages);
	352
	353	/**
	354	* devm_memremap_pages - remap and provide memmap backing for the given resource
	355	* @dev: hosting device for @res
	356	* @pgmap: pointer to a struct dev_pagemap
	357	*
	358	* Notes:
	359	* 1/ At a minimum the res and type members of @pgmap must be initialized
	360	* by the caller before passing it to this function
	361	*
	362	* 2/ The altmap field may optionally be initialized, in which case
	363	* PGMAP_ALTMAP_VALID must be set in pgmap->flags.
	364	*
	365	* 3/ The ref field may optionally be provided, in which pgmap->ref must be
	366	* 'live' on entry and will be killed and reaped at
	367	* devm_memremap_pages_release() time, or if this routine fails.
	368	*
	369	* 4/ res is expected to be a host memory range that could feasibly be
	370	* treated as a "System RAM" range, i.e. not a device mmio range, but
	371	* this is not enforced.
	372	*/
	373	void devm_memremap_pages(struct device dev, struct dev_pagemap *pgmap)
	374	{
	375	int error;
	376	void *ret;
	377
	378	ret = memremap_pages(pgmap, dev_to_node(dev));
	379	if (IS_ERR(ret))
	380	return ret;
	381
	382	error = devm_add_action_or_reset(dev, devm_memremap_pages_release,
	383	pgmap);
	384	if (error)
	385	return ERR_PTR(error);
	386	return ret;
	387	}
808153e1	388	EXPORT_SYMBOL_GPL(devm_memremap_pages);
4b94ffdc	389
2e3f139e DW	390	void devm_memunmap_pages(struct device dev, struct dev_pagemap pgmap)
	391	{
	392	devm_release_action(dev, devm_memremap_pages_release, pgmap);
	393	}
	394	EXPORT_SYMBOL_GPL(devm_memunmap_pages);
	395
4b94ffdc DW	396	unsigned long vmem_altmap_offset(struct vmem_altmap *altmap)
	397	{
	398	/* number of pfns from base where pfn_to_page() is valid */
514caf23 CH	399	if (altmap)
	400	return altmap->reserve + altmap->free;
	401	return 0;
4b94ffdc DW	402	}
	403
	404	void vmem_altmap_free(struct vmem_altmap *altmap, unsigned long nr_pfns)
	405	{
	406	altmap->alloc -= nr_pfns;
	407	}
	408
0822acb8 CH	409	/**
	410	* get_dev_pagemap() - take a new live reference on the dev_pagemap for @pfn
	411	* @pfn: page frame number to lookup page_map
	412	* @pgmap: optional known pgmap that already has a reference
	413	*
832d7aa0 CH	414	* If @pgmap is non-NULL and covers @pfn it will be returned as-is. If @pgmap
832d7aa0 CH	415	* is non-NULL but does not cover @pfn the reference to it will be released.
0822acb8 CH	416	*/
	417	struct dev_pagemap *get_dev_pagemap(unsigned long pfn,
	418	struct dev_pagemap *pgmap)
	419	{
0822acb8 CH	420	resource_size_t phys = PFN_PHYS(pfn);
	421
	422	/*
832d7aa0	423	* In the cached case we're already holding a live reference.
0822acb8	424	*/
832d7aa0	425	if (pgmap) {
e7744aa2	426	if (phys >= pgmap->res.start && phys <= pgmap->res.end)
832d7aa0 CH	427	return pgmap;
832d7aa0 CH	428	put_dev_pagemap(pgmap);
0822acb8 CH	429	}
	430
	431	/* fall back to slow path lookup */
	432	rcu_read_lock();
bcfa4b72	433	pgmap = xa_load(&pgmap_array, PHYS_PFN(phys));
0822acb8 CH	434	if (pgmap && !percpu_ref_tryget_live(pgmap->ref))
	435	pgmap = NULL;
	436	rcu_read_unlock();
	437
	438	return pgmap;
	439	}
e7638488	440	EXPORT_SYMBOL_GPL(get_dev_pagemap);
7b2d55d2	441
e7638488	442	#ifdef CONFIG_DEV_PAGEMAP_OPS
07d80269	443	void free_devmap_managed_page(struct page *page)
7b2d55d2	444	{
429589d6 DW	445	/* notify page idle for dax */
	446	if (!is_device_private_page(page)) {
	447	wake_up_var(&page->_refcount);
	448	return;
	449	}
7ab0ad0e	450
429589d6 DW	451	/* Clear Active bit in case of parallel mark_page_accessed */
	452	__ClearPageActive(page);
	453	__ClearPageWaiters(page);
	454
	455	mem_cgroup_uncharge(page);
	456
	457	/*
	458	* When a device_private page is freed, the page->mapping field
	459	* may still contain a (stale) mapping value. For example, the
	460	* lower bits of page->mapping may still identify the page as an
	461	* anonymous page. Ultimately, this entire field is just stale
	462	* and wrong, and it will cause errors if not cleared. One
	463	* example is:
	464	*
	465	* migrate_vma_pages()
	466	* migrate_vma_insert_page()
	467	* page_add_new_anon_rmap()
	468	* __page_set_anon_rmap()
	469	* ...checks page->mapping, via PageAnon(page) call,
	470	* and incorrectly concludes that the page is an
	471	* anonymous page. Therefore, it incorrectly,
	472	* silently fails to set up the new anon rmap.
	473	*
	474	* For other types of ZONE_DEVICE pages, migration is either
	475	* handled differently or not done at all, so there is no need
	476	* to clear page->mapping.
	477	*/
	478	page->mapping = NULL;
	479	page->pgmap->ops->page_free(page);
7b2d55d2	480	}
e7638488	481	#endif /* CONFIG_DEV_PAGEMAP_OPS */