[linux-block.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(void)
{
	static_branch_dec(&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;
	}

	static_branch_inc(&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 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));
}

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();
}
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)
{
	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;

	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(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);
	}

	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;
	bool need_devmap_managed = true;
	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->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:
		need_devmap_managed = false;
		break;
	case MEMORY_DEVICE_PCI_P2PDMA:
		params.pgprot = pgprot_noncached(params.pgprot);
		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);
	}

	/*
	 * 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
6f42193f	44	static void devmap_managed_enable_put(void)
f6a55e1a	45	{
433e7d31	46	static_branch_dec(&devmap_managed_key);
f6a55e1a CH	47	}
f6a55e1a CH	48
6f42193f	49	static int devmap_managed_enable_get(struct dev_pagemap *pgmap)
f6a55e1a	50	{
429589d6 DW	51	if (pgmap->type == MEMORY_DEVICE_PRIVATE &&
429589d6 DW	52	(!pgmap->ops \|\| !pgmap->ops->page_free)) {
f6a55e1a CH	53	WARN(1, "Missing page_free method\n");
	54	return -EINVAL;
	55	}
	56
433e7d31	57	static_branch_inc(&devmap_managed_key);
6f42193f	58	return 0;
f6a55e1a CH	59	}
f6a55e1a CH	60	#else
6f42193f	61	static int devmap_managed_enable_get(struct dev_pagemap *pgmap)
f6a55e1a CH	62	{
	63	return -EINVAL;
	64	}
6f42193f CH	65	static void devmap_managed_enable_put(void)
	66	{
	67	}
f6a55e1a CH	68	#endif /* CONFIG_DEV_PAGEMAP_OPS */
f6a55e1a CH	69
a4574f63	70	static void pgmap_array_delete(struct range *range)
ab1b597e	71	{
a4574f63	72	xa_store_range(&pgmap_array, PHYS_PFN(range->start), PHYS_PFN(range->end),
bcfa4b72	73	NULL, GFP_KERNEL);
ab1b597e	74	synchronize_rcu();
9476df7d DW	75	}
9476df7d DW	76
b7b3c01b	77	static unsigned long pfn_first(struct dev_pagemap *pgmap, int range_id)
5c2c2587	78	{
b7b3c01b DW	79	struct range *range = &pgmap->ranges[range_id];
	80	unsigned long pfn = PHYS_PFN(range->start);
	81
	82	if (range_id)
	83	return pfn;
	84	return pfn + vmem_altmap_offset(pgmap_altmap(pgmap));
5c2c2587 DW	85	}
5c2c2587 DW	86
b7b3c01b	87	static unsigned long pfn_end(struct dev_pagemap *pgmap, int range_id)
5c2c2587	88	{
b7b3c01b	89	const struct range *range = &pgmap->ranges[range_id];
5c2c2587	90
a4574f63	91	return (range->start + range_len(range)) >> PAGE_SHIFT;
5c2c2587 DW	92	}
5c2c2587 DW	93
949b9325 DW	94	static unsigned long pfn_next(unsigned long pfn)
	95	{
	96	if (pfn % 1024 == 0)
	97	cond_resched();
	98	return pfn + 1;
	99	}
	100
b7b3c01b DW	101	#define for_each_device_pfn(pfn, map, i) \
b7b3c01b DW	102	for (pfn = pfn_first(map, i); pfn < pfn_end(map, i); pfn = pfn_next(pfn))
5c2c2587	103
24917f6b CH	104	static void dev_pagemap_kill(struct dev_pagemap *pgmap)
	105	{
	106	if (pgmap->ops && pgmap->ops->kill)
	107	pgmap->ops->kill(pgmap);
	108	else
	109	percpu_ref_kill(pgmap->ref);
	110	}
	111
	112	static void dev_pagemap_cleanup(struct dev_pagemap *pgmap)
	113	{
	114	if (pgmap->ops && pgmap->ops->cleanup) {
	115	pgmap->ops->cleanup(pgmap);
	116	} else {
	117	wait_for_completion(&pgmap->done);
	118	percpu_ref_exit(pgmap->ref);
	119	}
06282373 DW	120	/*
	121	* Undo the pgmap ref assignment for the internal case as the
	122	* caller may re-enable the same pgmap.
	123	*/
	124	if (pgmap->ref == &pgmap->internal_ref)
	125	pgmap->ref = NULL;
24917f6b CH	126	}
24917f6b CH	127
b7b3c01b	128	static void pageunmap_range(struct dev_pagemap *pgmap, int range_id)
41e94a85	129	{
b7b3c01b	130	struct range *range = &pgmap->ranges[range_id];
77e080e7	131	struct page *first_page;
2c2a5af6	132	int nid;
71389703	133
77e080e7	134	/* make sure to access a memmap that was actually initialized */
b7b3c01b	135	first_page = pfn_to_page(pfn_first(pgmap, range_id));
77e080e7	136
41e94a85	137	/* pages are dead and unused, undo the arch mapping */
77e080e7	138	nid = page_to_nid(first_page);
2c2a5af6	139
f931ab47	140	mem_hotplug_begin();
a4574f63 DW	141	remove_pfn_range_from_zone(page_zone(first_page), PHYS_PFN(range->start),
a4574f63 DW	142	PHYS_PFN(range_len(range)));
69324b8f	143	if (pgmap->type == MEMORY_DEVICE_PRIVATE) {
a4574f63 DW	144	__remove_pages(PHYS_PFN(range->start),
a4574f63 DW	145	PHYS_PFN(range_len(range)), NULL);
69324b8f	146	} else {
a4574f63	147	arch_remove_memory(nid, range->start, range_len(range),
514caf23	148	pgmap_altmap(pgmap));
a4574f63	149	kasan_remove_zero_shadow(__va(range->start), range_len(range));
69324b8f	150	}
f931ab47	151	mem_hotplug_done();
b5d24fda	152
a4574f63 DW	153	untrack_pfn(NULL, PHYS_PFN(range->start), range_len(range));
a4574f63 DW	154	pgmap_array_delete(range);
b7b3c01b DW	155	}
	156
	157	void memunmap_pages(struct dev_pagemap *pgmap)
	158	{
	159	unsigned long pfn;
	160	int i;
	161
	162	dev_pagemap_kill(pgmap);
	163	for (i = 0; i < pgmap->nr_range; i++)
	164	for_each_device_pfn(pfn, pgmap, i)
	165	put_page(pfn_to_page(pfn));
	166	dev_pagemap_cleanup(pgmap);
	167
	168	for (i = 0; i < pgmap->nr_range; i++)
	169	pageunmap_range(pgmap, i);
	170
fdc029b1	171	WARN_ONCE(pgmap->altmap.alloc, "failed to free all reserved pages\n");
6f42193f	172	devmap_managed_enable_put();
9476df7d	173	}
6869b7b2 CH	174	EXPORT_SYMBOL_GPL(memunmap_pages);
	175
	176	static void devm_memremap_pages_release(void *data)
	177	{
	178	memunmap_pages(data);
	179	}
9476df7d	180
24917f6b CH	181	static void dev_pagemap_percpu_release(struct percpu_ref *ref)
	182	{
	183	struct dev_pagemap *pgmap =
	184	container_of(ref, struct dev_pagemap, internal_ref);
	185
	186	complete(&pgmap->done);
	187	}
	188
b7b3c01b DW	189	static int pagemap_range(struct dev_pagemap pgmap, struct mhp_params params,
b7b3c01b DW	190	int range_id, int nid)
41e94a85	191	{
b7b3c01b	192	struct range *range = &pgmap->ranges[range_id];
966cf44f	193	struct dev_pagemap *conflict_pgmap;
6869b7b2	194	int error, is_ram;
5f29a77c	195
b7b3c01b DW	196	if (WARN_ONCE(pgmap_altmap(pgmap) && range_id > 0,
	197	"altmap not supported for multiple ranges\n"))
	198	return -EINVAL;
f6a55e1a	199
a4574f63	200	conflict_pgmap = get_dev_pagemap(PHYS_PFN(range->start), NULL);
15d36fec	201	if (conflict_pgmap) {
6869b7b2	202	WARN(1, "Conflicting mapping in same section\n");
15d36fec	203	put_dev_pagemap(conflict_pgmap);
b7b3c01b	204	return -ENOMEM;
15d36fec DJ	205	}
15d36fec DJ	206
a4574f63	207	conflict_pgmap = get_dev_pagemap(PHYS_PFN(range->end), NULL);
15d36fec	208	if (conflict_pgmap) {
6869b7b2	209	WARN(1, "Conflicting mapping in same section\n");
15d36fec	210	put_dev_pagemap(conflict_pgmap);
b7b3c01b	211	return -ENOMEM;
15d36fec DJ	212	}
15d36fec DJ	213
a4574f63	214	is_ram = region_intersects(range->start, range_len(range),
d37a14bb	215	IORESOURCE_SYSTEM_RAM, IORES_DESC_NONE);
41e94a85	216
06489cfb	217	if (is_ram != REGION_DISJOINT) {
a4574f63 DW	218	WARN_ONCE(1, "attempted on %s region %#llx-%#llx\n",
	219	is_ram == REGION_MIXED ? "mixed" : "ram",
	220	range->start, range->end);
b7b3c01b	221	return -ENXIO;
41e94a85 CH	222	}
41e94a85 CH	223
a4574f63 DW	224	error = xa_err(xa_store_range(&pgmap_array, PHYS_PFN(range->start),
a4574f63 DW	225	PHYS_PFN(range->end), pgmap, GFP_KERNEL));
9476df7d	226	if (error)
b7b3c01b	227	return error;
9476df7d	228
41e94a85	229	if (nid < 0)
7eff93b7	230	nid = numa_mem_id();
41e94a85	231
b7b3c01b	232	error = track_pfn_remap(NULL, &params->pgprot, PHYS_PFN(range->start), 0,
a4574f63	233	range_len(range));
9049771f DW	234	if (error)
	235	goto err_pfn_remap;
	236
f931ab47	237	mem_hotplug_begin();
69324b8f DW	238
	239	/*
	240	* For device private memory we call add_pages() as we only need to
	241	* allocate and initialize struct page for the device memory. More-
	242	* over the device memory is un-accessible thus we do not want to
	243	* create a linear mapping for the memory like arch_add_memory()
	244	* would do.
	245	*
	246	* For all other device memory types, which are accessible by
	247	* the CPU, we do want the linear mapping and thus use
	248	* arch_add_memory().
	249	*/
	250	if (pgmap->type == MEMORY_DEVICE_PRIVATE) {
a4574f63	251	error = add_pages(nid, PHYS_PFN(range->start),
b7b3c01b	252	PHYS_PFN(range_len(range)), params);
69324b8f	253	} else {
a4574f63	254	error = kasan_add_zero_shadow(__va(range->start), range_len(range));
69324b8f DW	255	if (error) {
	256	mem_hotplug_done();
	257	goto err_kasan;
	258	}
	259
a4574f63	260	error = arch_add_memory(nid, range->start, range_len(range),
b7b3c01b	261	params);
69324b8f DW	262	}
	263
	264	if (!error) {
	265	struct zone *zone;
	266
	267	zone = &NODE_DATA(nid)->node_zones[ZONE_DEVICE];
a4574f63	268	move_pfn_range_to_zone(zone, PHYS_PFN(range->start),
b7b3c01b	269	PHYS_PFN(range_len(range)), params->altmap);
0207df4f AR	270	}
0207df4f AR	271
f931ab47	272	mem_hotplug_done();
9476df7d DW	273	if (error)
9476df7d DW	274	goto err_add_memory;
41e94a85	275
966cf44f AD	276	/*
	277	* Initialization of the pages has been deferred until now in order
	278	* to allow us to do the work while not holding the hotplug lock.
	279	*/
	280	memmap_init_zone_device(&NODE_DATA(nid)->node_zones[ZONE_DEVICE],
a4574f63 DW	281	PHYS_PFN(range->start),
a4574f63 DW	282	PHYS_PFN(range_len(range)), pgmap);
b7b3c01b DW	283	percpu_ref_get_many(pgmap->ref, pfn_end(pgmap, range_id)
	284	- pfn_first(pgmap, range_id));
	285	return 0;
9476df7d	286
b7b3c01b	287	err_add_memory:
a4574f63	288	kasan_remove_zero_shadow(__va(range->start), range_len(range));
b7b3c01b	289	err_kasan:
a4574f63	290	untrack_pfn(NULL, PHYS_PFN(range->start), range_len(range));
b7b3c01b	291	err_pfn_remap:
a4574f63	292	pgmap_array_delete(range);
b7b3c01b DW	293	return error;
	294	}
	295
	296
	297	/*
	298	* Not device managed version of dev_memremap_pages, undone by
	299	* memunmap_pages(). Please use dev_memremap_pages if you have a struct
	300	* device available.
	301	*/
	302	void memremap_pages(struct dev_pagemap pgmap, int nid)
	303	{
	304	struct mhp_params params = {
	305	.altmap = pgmap_altmap(pgmap),
	306	.pgprot = PAGE_KERNEL,
	307	};
	308	const int nr_range = pgmap->nr_range;
	309	bool need_devmap_managed = true;
	310	int error, i;
	311
	312	if (WARN_ONCE(!nr_range, "nr_range must be specified\n"))
	313	return ERR_PTR(-EINVAL);
	314
	315	switch (pgmap->type) {
	316	case MEMORY_DEVICE_PRIVATE:
	317	if (!IS_ENABLED(CONFIG_DEVICE_PRIVATE)) {
	318	WARN(1, "Device private memory not supported\n");
	319	return ERR_PTR(-EINVAL);
	320	}
	321	if (!pgmap->ops \|\| !pgmap->ops->migrate_to_ram) {
	322	WARN(1, "Missing migrate_to_ram method\n");
	323	return ERR_PTR(-EINVAL);
	324	}
	325	if (!pgmap->owner) {
	326	WARN(1, "Missing owner\n");
	327	return ERR_PTR(-EINVAL);
	328	}
	329	break;
	330	case MEMORY_DEVICE_FS_DAX:
	331	if (!IS_ENABLED(CONFIG_ZONE_DEVICE) \|\|
	332	IS_ENABLED(CONFIG_FS_DAX_LIMITED)) {
	333	WARN(1, "File system DAX not supported\n");
	334	return ERR_PTR(-EINVAL);
	335	}
	336	break;
	337	case MEMORY_DEVICE_GENERIC:
	338	need_devmap_managed = false;
	339	break;
	340	case MEMORY_DEVICE_PCI_P2PDMA:
	341	params.pgprot = pgprot_noncached(params.pgprot);
	342	need_devmap_managed = false;
	343	break;
	344	default:
	345	WARN(1, "Invalid pgmap type %d\n", pgmap->type);
	346	break;
	347	}
	348
	349	if (!pgmap->ref) {
	350	if (pgmap->ops && (pgmap->ops->kill \|\| pgmap->ops->cleanup))
	351	return ERR_PTR(-EINVAL);
	352
	353	init_completion(&pgmap->done);
	354	error = percpu_ref_init(&pgmap->internal_ref,
	355	dev_pagemap_percpu_release, 0, GFP_KERNEL);
	356	if (error)
357	return ERR_PTR(error);
358	pgmap->ref = &pgmap->internal_ref;
359	} else {
360	if (!pgmap->ops \|\| !pgmap->ops->kill \|\| !pgmap->ops->cleanup) {
361	WARN(1, "Missing reference count teardown definition\n");
362	return ERR_PTR(-EINVAL);
363	}
364	}
365
366	if (need_devmap_managed) {
367	error = devmap_managed_enable_get(pgmap);
368	if (error)
369	return ERR_PTR(error);
370	}
371
372	/*
373	* Clear the pgmap nr_range as it will be incremented for each
374	* successfully processed range. This communicates how many
375	* regions to unwind in the abort case.
376	*/
377	pgmap->nr_range = 0;
378	error = 0;
379	for (i = 0; i < nr_range; i++) {
380	error = pagemap_range(pgmap, &params, i, nid);
381	if (error)
382	break;
383	pgmap->nr_range++;
384	}
385
386	if (i < nr_range) {
387	memunmap_pages(pgmap);
388	pgmap->nr_range = nr_range;
389	return ERR_PTR(error);
390	}
391
392	return __va(pgmap->ranges[0].start);
41e94a85	393	}
6869b7b2 CH	394	EXPORT_SYMBOL_GPL(memremap_pages);
	395
	396	/**
	397	* devm_memremap_pages - remap and provide memmap backing for the given resource
	398	* @dev: hosting device for @res
	399	* @pgmap: pointer to a struct dev_pagemap
	400	*
	401	* Notes:
	402	* 1/ At a minimum the res and type members of @pgmap must be initialized
	403	* by the caller before passing it to this function
	404	*
	405	* 2/ The altmap field may optionally be initialized, in which case
	406	* PGMAP_ALTMAP_VALID must be set in pgmap->flags.
	407	*
	408	* 3/ The ref field may optionally be provided, in which pgmap->ref must be
	409	* 'live' on entry and will be killed and reaped at
	410	* devm_memremap_pages_release() time, or if this routine fails.
	411	*
a4574f63	412	* 4/ range is expected to be a host memory range that could feasibly be
6869b7b2 CH	413	* treated as a "System RAM" range, i.e. not a device mmio range, but
	414	* this is not enforced.
	415	*/
	416	void devm_memremap_pages(struct device dev, struct dev_pagemap *pgmap)
	417	{
	418	int error;
	419	void *ret;
	420
	421	ret = memremap_pages(pgmap, dev_to_node(dev));
	422	if (IS_ERR(ret))
	423	return ret;
	424
	425	error = devm_add_action_or_reset(dev, devm_memremap_pages_release,
	426	pgmap);
	427	if (error)
	428	return ERR_PTR(error);
	429	return ret;
	430	}
808153e1	431	EXPORT_SYMBOL_GPL(devm_memremap_pages);
4b94ffdc	432
2e3f139e DW	433	void devm_memunmap_pages(struct device dev, struct dev_pagemap pgmap)
	434	{
	435	devm_release_action(dev, devm_memremap_pages_release, pgmap);
	436	}
	437	EXPORT_SYMBOL_GPL(devm_memunmap_pages);
	438
4b94ffdc DW	439	unsigned long vmem_altmap_offset(struct vmem_altmap *altmap)
	440	{
	441	/* number of pfns from base where pfn_to_page() is valid */
514caf23 CH	442	if (altmap)
	443	return altmap->reserve + altmap->free;
	444	return 0;
4b94ffdc DW	445	}
	446
	447	void vmem_altmap_free(struct vmem_altmap *altmap, unsigned long nr_pfns)
	448	{
	449	altmap->alloc -= nr_pfns;
	450	}
	451
0822acb8 CH	452	/**
	453	* get_dev_pagemap() - take a new live reference on the dev_pagemap for @pfn
	454	* @pfn: page frame number to lookup page_map
	455	* @pgmap: optional known pgmap that already has a reference
	456	*
832d7aa0 CH	457	* If @pgmap is non-NULL and covers @pfn it will be returned as-is. If @pgmap
832d7aa0 CH	458	* is non-NULL but does not cover @pfn the reference to it will be released.
0822acb8 CH	459	*/
	460	struct dev_pagemap *get_dev_pagemap(unsigned long pfn,
	461	struct dev_pagemap *pgmap)
	462	{
0822acb8 CH	463	resource_size_t phys = PFN_PHYS(pfn);
	464
	465	/*
832d7aa0	466	* In the cached case we're already holding a live reference.
0822acb8	467	*/
832d7aa0	468	if (pgmap) {
a4574f63	469	if (phys >= pgmap->range.start && phys <= pgmap->range.end)
832d7aa0 CH	470	return pgmap;
832d7aa0 CH	471	put_dev_pagemap(pgmap);
0822acb8 CH	472	}
	473
	474	/* fall back to slow path lookup */
	475	rcu_read_lock();
bcfa4b72	476	pgmap = xa_load(&pgmap_array, PHYS_PFN(phys));
0822acb8 CH	477	if (pgmap && !percpu_ref_tryget_live(pgmap->ref))
	478	pgmap = NULL;
	479	rcu_read_unlock();
	480
	481	return pgmap;
	482	}
e7638488	483	EXPORT_SYMBOL_GPL(get_dev_pagemap);
7b2d55d2	484
e7638488	485	#ifdef CONFIG_DEV_PAGEMAP_OPS
07d80269	486	void free_devmap_managed_page(struct page *page)
7b2d55d2	487	{
429589d6 DW	488	/* notify page idle for dax */
	489	if (!is_device_private_page(page)) {
	490	wake_up_var(&page->_refcount);
	491	return;
	492	}
7ab0ad0e	493
429589d6 DW	494	__ClearPageWaiters(page);
	495
	496	mem_cgroup_uncharge(page);
	497
	498	/*
	499	* When a device_private page is freed, the page->mapping field
	500	* may still contain a (stale) mapping value. For example, the
	501	* lower bits of page->mapping may still identify the page as an
	502	* anonymous page. Ultimately, this entire field is just stale
	503	* and wrong, and it will cause errors if not cleared. One
	504	* example is:
	505	*
	506	* migrate_vma_pages()
	507	* migrate_vma_insert_page()
	508	* page_add_new_anon_rmap()
	509	* __page_set_anon_rmap()
	510	* ...checks page->mapping, via PageAnon(page) call,
	511	* and incorrectly concludes that the page is an
	512	* anonymous page. Therefore, it incorrectly,
	513	* silently fails to set up the new anon rmap.
	514	*
	515	* For other types of ZONE_DEVICE pages, migration is either
	516	* handled differently or not done at all, so there is no need
	517	* to clear page->mapping.
	518	*/
	519	page->mapping = NULL;
	520	page->pgmap->ops->page_free(page);
7b2d55d2	521	}
e7638488	522	#endif /* CONFIG_DEV_PAGEMAP_OPS */