[linux.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 void devmap_managed_enable_put(struct dev_pagemap *pgmap)
{
	if (pgmap->type == MEMORY_DEVICE_PRIVATE ||
	    pgmap->type == MEMORY_DEVICE_FS_DAX)
		static_branch_dec(&devmap_managed_key);
}

static void devmap_managed_enable_get(struct dev_pagemap *pgmap)
{
	if (pgmap->type == MEMORY_DEVICE_PRIVATE ||
	    pgmap->type == MEMORY_DEVICE_FS_DAX)
		static_branch_inc(&devmap_managed_key);
}
#else
static void devmap_managed_enable_get(struct dev_pagemap *pgmap)
{
}
static void devmap_managed_enable_put(struct dev_pagemap *pgmap)
{
}
#endif /* CONFIG_DEV_PAGEMAP_OPS */

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

static unsigned long pfn_first(struct dev_pagemap *pgmap, int range_id)
{
	struct range *range = &pgmap->ranges[range_id];
	unsigned long pfn = PHYS_PFN(range->start);

	if (range_id)
		return pfn;
	return pfn + vmem_altmap_offset(pgmap_altmap(pgmap));
}

bool pgmap_pfn_valid(struct dev_pagemap *pgmap, unsigned long pfn)
{
	int i;

	for (i = 0; i < pgmap->nr_range; i++) {
		struct range *range = &pgmap->ranges[i];

		if (pfn >= PHYS_PFN(range->start) &&
		    pfn <= PHYS_PFN(range->end))
			return pfn >= pfn_first(pgmap, i);
	}

	return false;
}

static unsigned long pfn_end(struct dev_pagemap *pgmap, int range_id)
{
	const struct range *range = &pgmap->ranges[range_id];

	return (range->start + range_len(range)) >> 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, i) \
	for (pfn = pfn_first(map, i); pfn < pfn_end(map, i); 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;
}

static void pageunmap_range(struct dev_pagemap *pgmap, int range_id)
{
	struct range *range = &pgmap->ranges[range_id];
	struct page *first_page;
	int nid;

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

	/* 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(range->start),
				   PHYS_PFN(range_len(range)));
	if (pgmap->type == MEMORY_DEVICE_PRIVATE) {
		__remove_pages(PHYS_PFN(range->start),
			       PHYS_PFN(range_len(range)), NULL);
	} else {
		arch_remove_memory(nid, range->start, range_len(range),
				pgmap_altmap(pgmap));
		kasan_remove_zero_shadow(__va(range->start), range_len(range));
	}
	mem_hotplug_done();

	untrack_pfn(NULL, PHYS_PFN(range->start), range_len(range));
	pgmap_array_delete(range);
}

void memunmap_pages(struct dev_pagemap *pgmap)
{
	unsigned long pfn;
	int i;

	dev_pagemap_kill(pgmap);
	for (i = 0; i < pgmap->nr_range; i++)
		for_each_device_pfn(pfn, pgmap, i)
			put_page(pfn_to_page(pfn));
	dev_pagemap_cleanup(pgmap);

	for (i = 0; i < pgmap->nr_range; i++)
		pageunmap_range(pgmap, i);

	WARN_ONCE(pgmap->altmap.alloc, "failed to free all reserved pages\n");
	devmap_managed_enable_put(pgmap);
}
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);
}

static int pagemap_range(struct dev_pagemap *pgmap, struct mhp_params *params,
		int range_id, int nid)
{
	const bool is_private = pgmap->type == MEMORY_DEVICE_PRIVATE;
	struct range *range = &pgmap->ranges[range_id];
	struct dev_pagemap *conflict_pgmap;
	int error, is_ram;

	if (WARN_ONCE(pgmap_altmap(pgmap) && range_id > 0,
				"altmap not supported for multiple ranges\n"))
		return -EINVAL;

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

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

	is_ram = region_intersects(range->start, range_len(range),
		IORESOURCE_SYSTEM_RAM, IORES_DESC_NONE);

	if (is_ram != REGION_DISJOINT) {
		WARN_ONCE(1, "attempted on %s region %#llx-%#llx\n",
				is_ram == REGION_MIXED ? "mixed" : "ram",
				range->start, range->end);
		return -ENXIO;
	}

	error = xa_err(xa_store_range(&pgmap_array, PHYS_PFN(range->start),
				PHYS_PFN(range->end), pgmap, GFP_KERNEL));
	if (error)
		return error;

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

	error = track_pfn_remap(NULL, &params->pgprot, PHYS_PFN(range->start), 0,
			range_len(range));
	if (error)
		goto err_pfn_remap;

	if (!mhp_range_allowed(range->start, range_len(range), !is_private)) {
		error = -EINVAL;
		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 (is_private) {
		error = add_pages(nid, PHYS_PFN(range->start),
				PHYS_PFN(range_len(range)), params);
	} else {
		error = kasan_add_zero_shadow(__va(range->start), range_len(range));
		if (error) {
			mem_hotplug_done();
			goto err_kasan;
		}

		error = arch_add_memory(nid, range->start, range_len(range),
					params);
	}

	if (!error) {
		struct zone *zone;

		zone = &NODE_DATA(nid)->node_zones[ZONE_DEVICE];
		move_pfn_range_to_zone(zone, PHYS_PFN(range->start),
				PHYS_PFN(range_len(range)), params->altmap,
				MIGRATE_MOVABLE);
	}

	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(range->start),
				PHYS_PFN(range_len(range)), pgmap);
	percpu_ref_get_many(pgmap->ref, pfn_end(pgmap, range_id)
			- pfn_first(pgmap, range_id));
	return 0;

err_add_memory:
	kasan_remove_zero_shadow(__va(range->start), range_len(range));
err_kasan:
	untrack_pfn(NULL, PHYS_PFN(range->start), range_len(range));
err_pfn_remap:
	pgmap_array_delete(range);
	return error;
}


/*
 * 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 mhp_params params = {
		.altmap = pgmap_altmap(pgmap),
		.pgprot = PAGE_KERNEL,
	};
	const int nr_range = pgmap->nr_range;
	int error, i;

	if (WARN_ONCE(!nr_range, "nr_range must be specified\n"))
		return ERR_PTR(-EINVAL);

	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->ops->page_free) {
			WARN(1, "Missing page_free 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_GENERIC:
		break;
	case MEMORY_DEVICE_PCI_P2PDMA:
		params.pgprot = pgprot_noncached(params.pgprot);
		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);
		}
	}

	devmap_managed_enable_get(pgmap);

	/*
	 * Clear the pgmap nr_range as it will be incremented for each
	 * successfully processed range. This communicates how many
	 * regions to unwind in the abort case.
	 */
	pgmap->nr_range = 0;
	error = 0;
	for (i = 0; i < nr_range; i++) {
		error = pagemap_range(pgmap, &params, i, nid);
		if (error)
			break;
		pgmap->nr_range++;
	}

	if (i < nr_range) {
		memunmap_pages(pgmap);
		pgmap->nr_range = nr_range;
		return ERR_PTR(error);
	}

	return __va(pgmap->ranges[0].start);
}
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/ range 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->range.start && phys <= pgmap->range.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;
	}

	__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);
f6a55e1a	43
46b1ee38	44	static void devmap_managed_enable_put(struct dev_pagemap *pgmap)
f6a55e1a	45	{
46b1ee38 RC	46	if (pgmap->type == MEMORY_DEVICE_PRIVATE \|\|
	47	pgmap->type == MEMORY_DEVICE_FS_DAX)
	48	static_branch_dec(&devmap_managed_key);
f6a55e1a CH	49	}
f6a55e1a CH	50
46b1ee38	51	static void devmap_managed_enable_get(struct dev_pagemap *pgmap)
f6a55e1a	52	{
46b1ee38 RC	53	if (pgmap->type == MEMORY_DEVICE_PRIVATE \|\|
	54	pgmap->type == MEMORY_DEVICE_FS_DAX)
	55	static_branch_inc(&devmap_managed_key);
f6a55e1a CH	56	}
f6a55e1a CH	57	#else
46b1ee38	58	static void devmap_managed_enable_get(struct dev_pagemap *pgmap)
f6a55e1a	59	{
f6a55e1a	60	}
46b1ee38	61	static void devmap_managed_enable_put(struct dev_pagemap *pgmap)
6f42193f CH	62	{
6f42193f CH	63	}
f6a55e1a CH	64	#endif /* CONFIG_DEV_PAGEMAP_OPS */
f6a55e1a CH	65
a4574f63	66	static void pgmap_array_delete(struct range *range)
ab1b597e	67	{
a4574f63	68	xa_store_range(&pgmap_array, PHYS_PFN(range->start), PHYS_PFN(range->end),
bcfa4b72	69	NULL, GFP_KERNEL);
ab1b597e	70	synchronize_rcu();
9476df7d DW	71	}
9476df7d DW	72
b7b3c01b	73	static unsigned long pfn_first(struct dev_pagemap *pgmap, int range_id)
5c2c2587	74	{
b7b3c01b DW	75	struct range *range = &pgmap->ranges[range_id];
	76	unsigned long pfn = PHYS_PFN(range->start);
	77
	78	if (range_id)
	79	return pfn;
	80	return pfn + vmem_altmap_offset(pgmap_altmap(pgmap));
5c2c2587 DW	81	}
5c2c2587 DW	82
34dc45be DW	83	bool pgmap_pfn_valid(struct dev_pagemap *pgmap, unsigned long pfn)
	84	{
	85	int i;
	86
	87	for (i = 0; i < pgmap->nr_range; i++) {
	88	struct range *range = &pgmap->ranges[i];
	89
	90	if (pfn >= PHYS_PFN(range->start) &&
	91	pfn <= PHYS_PFN(range->end))
	92	return pfn >= pfn_first(pgmap, i);
	93	}
	94
	95	return false;
	96	}
	97
b7b3c01b	98	static unsigned long pfn_end(struct dev_pagemap *pgmap, int range_id)
5c2c2587	99	{
b7b3c01b	100	const struct range *range = &pgmap->ranges[range_id];
5c2c2587	101
a4574f63	102	return (range->start + range_len(range)) >> PAGE_SHIFT;
5c2c2587 DW	103	}
5c2c2587 DW	104
949b9325 DW	105	static unsigned long pfn_next(unsigned long pfn)
	106	{
	107	if (pfn % 1024 == 0)
	108	cond_resched();
	109	return pfn + 1;
	110	}
	111
b7b3c01b DW	112	#define for_each_device_pfn(pfn, map, i) \
b7b3c01b DW	113	for (pfn = pfn_first(map, i); pfn < pfn_end(map, i); pfn = pfn_next(pfn))
5c2c2587	114
24917f6b CH	115	static void dev_pagemap_kill(struct dev_pagemap *pgmap)
	116	{
	117	if (pgmap->ops && pgmap->ops->kill)
	118	pgmap->ops->kill(pgmap);
	119	else
	120	percpu_ref_kill(pgmap->ref);
	121	}
	122
	123	static void dev_pagemap_cleanup(struct dev_pagemap *pgmap)
	124	{
	125	if (pgmap->ops && pgmap->ops->cleanup) {
	126	pgmap->ops->cleanup(pgmap);
	127	} else {
	128	wait_for_completion(&pgmap->done);
	129	percpu_ref_exit(pgmap->ref);
	130	}
06282373 DW	131	/*
	132	* Undo the pgmap ref assignment for the internal case as the
	133	* caller may re-enable the same pgmap.
	134	*/
	135	if (pgmap->ref == &pgmap->internal_ref)
	136	pgmap->ref = NULL;
24917f6b CH	137	}
24917f6b CH	138
b7b3c01b	139	static void pageunmap_range(struct dev_pagemap *pgmap, int range_id)
41e94a85	140	{
b7b3c01b	141	struct range *range = &pgmap->ranges[range_id];
77e080e7	142	struct page *first_page;
2c2a5af6	143	int nid;
71389703	144
77e080e7	145	/* make sure to access a memmap that was actually initialized */
b7b3c01b	146	first_page = pfn_to_page(pfn_first(pgmap, range_id));
77e080e7	147
41e94a85	148	/* pages are dead and unused, undo the arch mapping */
77e080e7	149	nid = page_to_nid(first_page);
2c2a5af6	150
f931ab47	151	mem_hotplug_begin();
a4574f63 DW	152	remove_pfn_range_from_zone(page_zone(first_page), PHYS_PFN(range->start),
a4574f63 DW	153	PHYS_PFN(range_len(range)));
69324b8f	154	if (pgmap->type == MEMORY_DEVICE_PRIVATE) {
a4574f63 DW	155	__remove_pages(PHYS_PFN(range->start),
a4574f63 DW	156	PHYS_PFN(range_len(range)), NULL);
69324b8f	157	} else {
a4574f63	158	arch_remove_memory(nid, range->start, range_len(range),
514caf23	159	pgmap_altmap(pgmap));
a4574f63	160	kasan_remove_zero_shadow(__va(range->start), range_len(range));
69324b8f	161	}
f931ab47	162	mem_hotplug_done();
b5d24fda	163
a4574f63 DW	164	untrack_pfn(NULL, PHYS_PFN(range->start), range_len(range));
a4574f63 DW	165	pgmap_array_delete(range);
b7b3c01b DW	166	}
	167
	168	void memunmap_pages(struct dev_pagemap *pgmap)
	169	{
	170	unsigned long pfn;
	171	int i;
	172
	173	dev_pagemap_kill(pgmap);
	174	for (i = 0; i < pgmap->nr_range; i++)
	175	for_each_device_pfn(pfn, pgmap, i)
	176	put_page(pfn_to_page(pfn));
	177	dev_pagemap_cleanup(pgmap);
	178
	179	for (i = 0; i < pgmap->nr_range; i++)
	180	pageunmap_range(pgmap, i);
	181
fdc029b1	182	WARN_ONCE(pgmap->altmap.alloc, "failed to free all reserved pages\n");
46b1ee38	183	devmap_managed_enable_put(pgmap);
9476df7d	184	}
6869b7b2 CH	185	EXPORT_SYMBOL_GPL(memunmap_pages);
	186
	187	static void devm_memremap_pages_release(void *data)
	188	{
	189	memunmap_pages(data);
	190	}
9476df7d	191
24917f6b CH	192	static void dev_pagemap_percpu_release(struct percpu_ref *ref)
	193	{
	194	struct dev_pagemap *pgmap =
	195	container_of(ref, struct dev_pagemap, internal_ref);
	196
	197	complete(&pgmap->done);
	198	}
	199
b7b3c01b DW	200	static int pagemap_range(struct dev_pagemap pgmap, struct mhp_params params,
b7b3c01b DW	201	int range_id, int nid)
41e94a85	202	{
bca3feaa	203	const bool is_private = pgmap->type == MEMORY_DEVICE_PRIVATE;
b7b3c01b	204	struct range *range = &pgmap->ranges[range_id];
966cf44f	205	struct dev_pagemap *conflict_pgmap;
6869b7b2	206	int error, is_ram;
5f29a77c	207
b7b3c01b DW	208	if (WARN_ONCE(pgmap_altmap(pgmap) && range_id > 0,
	209	"altmap not supported for multiple ranges\n"))
	210	return -EINVAL;
f6a55e1a	211
a4574f63	212	conflict_pgmap = get_dev_pagemap(PHYS_PFN(range->start), NULL);
15d36fec	213	if (conflict_pgmap) {
6869b7b2	214	WARN(1, "Conflicting mapping in same section\n");
15d36fec	215	put_dev_pagemap(conflict_pgmap);
b7b3c01b	216	return -ENOMEM;
15d36fec DJ	217	}
15d36fec DJ	218
a4574f63	219	conflict_pgmap = get_dev_pagemap(PHYS_PFN(range->end), NULL);
15d36fec	220	if (conflict_pgmap) {
6869b7b2	221	WARN(1, "Conflicting mapping in same section\n");
15d36fec	222	put_dev_pagemap(conflict_pgmap);
b7b3c01b	223	return -ENOMEM;
15d36fec DJ	224	}
15d36fec DJ	225
a4574f63	226	is_ram = region_intersects(range->start, range_len(range),
d37a14bb	227	IORESOURCE_SYSTEM_RAM, IORES_DESC_NONE);
41e94a85	228
06489cfb	229	if (is_ram != REGION_DISJOINT) {
a4574f63 DW	230	WARN_ONCE(1, "attempted on %s region %#llx-%#llx\n",
	231	is_ram == REGION_MIXED ? "mixed" : "ram",
	232	range->start, range->end);
b7b3c01b	233	return -ENXIO;
41e94a85 CH	234	}
41e94a85 CH	235
a4574f63 DW	236	error = xa_err(xa_store_range(&pgmap_array, PHYS_PFN(range->start),
a4574f63 DW	237	PHYS_PFN(range->end), pgmap, GFP_KERNEL));
9476df7d	238	if (error)
b7b3c01b	239	return error;
9476df7d	240
41e94a85	241	if (nid < 0)
7eff93b7	242	nid = numa_mem_id();
41e94a85	243
b7b3c01b	244	error = track_pfn_remap(NULL, &params->pgprot, PHYS_PFN(range->start), 0,
a4574f63	245	range_len(range));
9049771f DW	246	if (error)
	247	goto err_pfn_remap;
	248
bca3feaa AK	249	if (!mhp_range_allowed(range->start, range_len(range), !is_private)) {
	250	error = -EINVAL;
	251	goto err_pfn_remap;
	252	}
	253
f931ab47	254	mem_hotplug_begin();
69324b8f DW	255
	256	/*
	257	* For device private memory we call add_pages() as we only need to
	258	* allocate and initialize struct page for the device memory. More-
	259	* over the device memory is un-accessible thus we do not want to
	260	* create a linear mapping for the memory like arch_add_memory()
	261	* would do.
	262	*
	263	* For all other device memory types, which are accessible by
	264	* the CPU, we do want the linear mapping and thus use
	265	* arch_add_memory().
	266	*/
bca3feaa	267	if (is_private) {
a4574f63	268	error = add_pages(nid, PHYS_PFN(range->start),
b7b3c01b	269	PHYS_PFN(range_len(range)), params);
69324b8f	270	} else {
a4574f63	271	error = kasan_add_zero_shadow(__va(range->start), range_len(range));
69324b8f DW	272	if (error) {
	273	mem_hotplug_done();
	274	goto err_kasan;
	275	}
	276
a4574f63	277	error = arch_add_memory(nid, range->start, range_len(range),
b7b3c01b	278	params);
69324b8f DW	279	}
	280
	281	if (!error) {
	282	struct zone *zone;
	283
	284	zone = &NODE_DATA(nid)->node_zones[ZONE_DEVICE];
a4574f63	285	move_pfn_range_to_zone(zone, PHYS_PFN(range->start),
d882c006 DH	286	PHYS_PFN(range_len(range)), params->altmap,
d882c006 DH	287	MIGRATE_MOVABLE);
0207df4f AR	288	}
0207df4f AR	289
f931ab47	290	mem_hotplug_done();
9476df7d DW	291	if (error)
9476df7d DW	292	goto err_add_memory;
41e94a85	293
966cf44f AD	294	/*
	295	* Initialization of the pages has been deferred until now in order
	296	* to allow us to do the work while not holding the hotplug lock.
	297	*/
	298	memmap_init_zone_device(&NODE_DATA(nid)->node_zones[ZONE_DEVICE],
a4574f63 DW	299	PHYS_PFN(range->start),
a4574f63 DW	300	PHYS_PFN(range_len(range)), pgmap);
b7b3c01b DW	301	percpu_ref_get_many(pgmap->ref, pfn_end(pgmap, range_id)
	302	- pfn_first(pgmap, range_id));
	303	return 0;
9476df7d	304
b7b3c01b	305	err_add_memory:
a4574f63	306	kasan_remove_zero_shadow(__va(range->start), range_len(range));
b7b3c01b	307	err_kasan:
a4574f63	308	untrack_pfn(NULL, PHYS_PFN(range->start), range_len(range));
b7b3c01b	309	err_pfn_remap:
a4574f63	310	pgmap_array_delete(range);
b7b3c01b DW	311	return error;
	312	}
	313
	314
	315	/*
	316	* Not device managed version of dev_memremap_pages, undone by
	317	* memunmap_pages(). Please use dev_memremap_pages if you have a struct
	318	* device available.
	319	*/
	320	void memremap_pages(struct dev_pagemap pgmap, int nid)
	321	{
	322	struct mhp_params params = {
	323	.altmap = pgmap_altmap(pgmap),
	324	.pgprot = PAGE_KERNEL,
	325	};
	326	const int nr_range = pgmap->nr_range;
b7b3c01b DW	327	int error, i;
	328
	329	if (WARN_ONCE(!nr_range, "nr_range must be specified\n"))
	330	return ERR_PTR(-EINVAL);
	331
	332	switch (pgmap->type) {
	333	case MEMORY_DEVICE_PRIVATE:
	334	if (!IS_ENABLED(CONFIG_DEVICE_PRIVATE)) {
	335	WARN(1, "Device private memory not supported\n");
	336	return ERR_PTR(-EINVAL);
	337	}
	338	if (!pgmap->ops \|\| !pgmap->ops->migrate_to_ram) {
	339	WARN(1, "Missing migrate_to_ram method\n");
	340	return ERR_PTR(-EINVAL);
	341	}
46b1ee38 RC	342	if (!pgmap->ops->page_free) {
	343	WARN(1, "Missing page_free method\n");
	344	return ERR_PTR(-EINVAL);
	345	}
b7b3c01b DW	346	if (!pgmap->owner) {
	347	WARN(1, "Missing owner\n");
	348	return ERR_PTR(-EINVAL);
	349	}
	350	break;
	351	case MEMORY_DEVICE_FS_DAX:
	352	if (!IS_ENABLED(CONFIG_ZONE_DEVICE) \|\|
	353	IS_ENABLED(CONFIG_FS_DAX_LIMITED)) {
	354	WARN(1, "File system DAX not supported\n");
	355	return ERR_PTR(-EINVAL);
	356	}
	357	break;
	358	case MEMORY_DEVICE_GENERIC:
b7b3c01b DW	359	break;
	360	case MEMORY_DEVICE_PCI_P2PDMA:
	361	params.pgprot = pgprot_noncached(params.pgprot);
b7b3c01b DW	362	break;
	363	default:
	364	WARN(1, "Invalid pgmap type %d\n", pgmap->type);
	365	break;
	366	}
	367
	368	if (!pgmap->ref) {
	369	if (pgmap->ops && (pgmap->ops->kill \|\| pgmap->ops->cleanup))
	370	return ERR_PTR(-EINVAL);
	371
	372	init_completion(&pgmap->done);
	373	error = percpu_ref_init(&pgmap->internal_ref,
	374	dev_pagemap_percpu_release, 0, GFP_KERNEL);
	375	if (error)
	376	return ERR_PTR(error);
	377	pgmap->ref = &pgmap->internal_ref;
	378	} else {
	379	if (!pgmap->ops \|\| !pgmap->ops->kill \|\| !pgmap->ops->cleanup) {
	380	WARN(1, "Missing reference count teardown definition\n");
	381	return ERR_PTR(-EINVAL);
	382	}
	383	}
	384
46b1ee38	385	devmap_managed_enable_get(pgmap);
b7b3c01b DW	386
	387	/*
	388	* Clear the pgmap nr_range as it will be incremented for each
	389	* successfully processed range. This communicates how many
	390	* regions to unwind in the abort case.
	391	*/
	392	pgmap->nr_range = 0;
	393	error = 0;
	394	for (i = 0; i < nr_range; i++) {
	395	error = pagemap_range(pgmap, &params, i, nid);
	396	if (error)
	397	break;
	398	pgmap->nr_range++;
	399	}
	400
	401	if (i < nr_range) {
	402	memunmap_pages(pgmap);
	403	pgmap->nr_range = nr_range;
	404	return ERR_PTR(error);
	405	}
	406
	407	return __va(pgmap->ranges[0].start);
41e94a85	408	}
6869b7b2 CH	409	EXPORT_SYMBOL_GPL(memremap_pages);
	410
	411	/**
	412	* devm_memremap_pages - remap and provide memmap backing for the given resource
	413	* @dev: hosting device for @res
	414	* @pgmap: pointer to a struct dev_pagemap
	415	*
	416	* Notes:
	417	* 1/ At a minimum the res and type members of @pgmap must be initialized
	418	* by the caller before passing it to this function
	419	*
	420	* 2/ The altmap field may optionally be initialized, in which case
	421	* PGMAP_ALTMAP_VALID must be set in pgmap->flags.
	422	*
	423	* 3/ The ref field may optionally be provided, in which pgmap->ref must be
	424	* 'live' on entry and will be killed and reaped at
	425	* devm_memremap_pages_release() time, or if this routine fails.
	426	*
a4574f63	427	* 4/ range is expected to be a host memory range that could feasibly be
6869b7b2 CH	428	* treated as a "System RAM" range, i.e. not a device mmio range, but
	429	* this is not enforced.
	430	*/
	431	void devm_memremap_pages(struct device dev, struct dev_pagemap *pgmap)
	432	{
	433	int error;
	434	void *ret;
	435
	436	ret = memremap_pages(pgmap, dev_to_node(dev));
	437	if (IS_ERR(ret))
	438	return ret;
	439
	440	error = devm_add_action_or_reset(dev, devm_memremap_pages_release,
	441	pgmap);
	442	if (error)
	443	return ERR_PTR(error);
	444	return ret;
	445	}
808153e1	446	EXPORT_SYMBOL_GPL(devm_memremap_pages);
4b94ffdc	447
2e3f139e DW	448	void devm_memunmap_pages(struct device dev, struct dev_pagemap pgmap)
	449	{
	450	devm_release_action(dev, devm_memremap_pages_release, pgmap);
	451	}
	452	EXPORT_SYMBOL_GPL(devm_memunmap_pages);
	453
4b94ffdc DW	454	unsigned long vmem_altmap_offset(struct vmem_altmap *altmap)
	455	{
	456	/* number of pfns from base where pfn_to_page() is valid */
514caf23 CH	457	if (altmap)
	458	return altmap->reserve + altmap->free;
	459	return 0;
4b94ffdc DW	460	}
	461
	462	void vmem_altmap_free(struct vmem_altmap *altmap, unsigned long nr_pfns)
	463	{
	464	altmap->alloc -= nr_pfns;
	465	}
	466
0822acb8 CH	467	/**
	468	* get_dev_pagemap() - take a new live reference on the dev_pagemap for @pfn
	469	* @pfn: page frame number to lookup page_map
	470	* @pgmap: optional known pgmap that already has a reference
	471	*
832d7aa0 CH	472	* If @pgmap is non-NULL and covers @pfn it will be returned as-is. If @pgmap
832d7aa0 CH	473	* is non-NULL but does not cover @pfn the reference to it will be released.
0822acb8 CH	474	*/
	475	struct dev_pagemap *get_dev_pagemap(unsigned long pfn,
	476	struct dev_pagemap *pgmap)
	477	{
0822acb8 CH	478	resource_size_t phys = PFN_PHYS(pfn);
	479
	480	/*
832d7aa0	481	* In the cached case we're already holding a live reference.
0822acb8	482	*/
832d7aa0	483	if (pgmap) {
a4574f63	484	if (phys >= pgmap->range.start && phys <= pgmap->range.end)
832d7aa0 CH	485	return pgmap;
832d7aa0 CH	486	put_dev_pagemap(pgmap);
0822acb8 CH	487	}
	488
	489	/* fall back to slow path lookup */
	490	rcu_read_lock();
bcfa4b72	491	pgmap = xa_load(&pgmap_array, PHYS_PFN(phys));
0822acb8 CH	492	if (pgmap && !percpu_ref_tryget_live(pgmap->ref))
	493	pgmap = NULL;
	494	rcu_read_unlock();
	495
	496	return pgmap;
	497	}
e7638488	498	EXPORT_SYMBOL_GPL(get_dev_pagemap);
7b2d55d2	499
e7638488	500	#ifdef CONFIG_DEV_PAGEMAP_OPS
07d80269	501	void free_devmap_managed_page(struct page *page)
7b2d55d2	502	{
429589d6 DW	503	/* notify page idle for dax */
	504	if (!is_device_private_page(page)) {
	505	wake_up_var(&page->_refcount);
	506	return;
	507	}
7ab0ad0e	508
429589d6 DW	509	__ClearPageWaiters(page);
	510
	511	mem_cgroup_uncharge(page);
	512
	513	/*
	514	* When a device_private page is freed, the page->mapping field
	515	* may still contain a (stale) mapping value. For example, the
	516	* lower bits of page->mapping may still identify the page as an
	517	* anonymous page. Ultimately, this entire field is just stale
	518	* and wrong, and it will cause errors if not cleared. One
	519	* example is:
	520	*
	521	* migrate_vma_pages()
	522	* migrate_vma_insert_page()
	523	* page_add_new_anon_rmap()
	524	* __page_set_anon_rmap()
	525	* ...checks page->mapping, via PageAnon(page) call,
	526	* and incorrectly concludes that the page is an
	527	* anonymous page. Therefore, it incorrectly,
	528	* silently fails to set up the new anon rmap.
	529	*
	530	* For other types of ZONE_DEVICE pages, migration is either
	531	* handled differently or not done at all, so there is no need
	532	* to clear page->mapping.
	533	*/
	534	page->mapping = NULL;
	535	page->pgmap->ops->page_free(page);
7b2d55d2	536	}
e7638488	537	#endif /* CONFIG_DEV_PAGEMAP_OPS */