memremap.c 15 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12
/*
 * Copyright(c) 2015 Intel Corporation. All rights reserved.
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of version 2 of the GNU General Public License as
 * published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful, but
 * WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 */
13
#include <linux/radix-tree.h>
14
#include <linux/device.h>
15
#include <linux/types.h>
16
#include <linux/pfn_t.h>
17 18
#include <linux/io.h>
#include <linux/mm.h>
19
#include <linux/memory_hotplug.h>
20 21
#include <linux/swap.h>
#include <linux/swapops.h>
22 23 24 25 26 27 28 29 30

#ifndef ioremap_cache
/* temporary while we convert existing ioremap_cache users to memremap */
__weak void __iomem *ioremap_cache(resource_size_t offset, unsigned long size)
{
	return ioremap(offset, size);
}
#endif

31 32 33 34 35 36 37
#ifndef arch_memremap_wb
static void *arch_memremap_wb(resource_size_t offset, unsigned long size)
{
	return (__force void *)ioremap_cache(offset, size);
}
#endif

38 39 40 41 42 43 44 45 46 47
#ifndef arch_memremap_can_ram_remap
static bool arch_memremap_can_ram_remap(resource_size_t offset, size_t size,
					unsigned long flags)
{
	return true;
}
#endif

static void *try_ram_remap(resource_size_t offset, size_t size,
			   unsigned long flags)
48
{
49
	unsigned long pfn = PHYS_PFN(offset);
50 51

	/* In the simple case just return the existing linear address */
52 53
	if (pfn_valid(pfn) && !PageHighMem(pfn_to_page(pfn)) &&
	    arch_memremap_can_ram_remap(offset, size, flags))
54
		return __va(offset);
55

56
	return NULL; /* fallback to arch_memremap_wb */
57 58
}

59 60 61 62
/**
 * memremap() - remap an iomem_resource as cacheable memory
 * @offset: iomem resource start address
 * @size: size of remap
63 64
 * @flags: any of MEMREMAP_WB, MEMREMAP_WT, MEMREMAP_WC,
 *		  MEMREMAP_ENC, MEMREMAP_DEC
65 66 67
 *
 * memremap() is "ioremap" for cases where it is known that the resource
 * being mapped does not have i/o side effects and the __iomem
68 69 70
 * annotation is not applicable. In the case of multiple flags, the different
 * mapping types will be attempted in the order listed below until one of
 * them succeeds.
71
 *
72
 * MEMREMAP_WB - matches the default mapping for System RAM on
73 74 75 76 77 78 79 80
 * the architecture.  This is usually a read-allocate write-back cache.
 * Morever, if MEMREMAP_WB is specified and the requested remap region is RAM
 * memremap() will bypass establishing a new mapping and instead return
 * a pointer into the direct map.
 *
 * MEMREMAP_WT - establish a mapping whereby writes either bypass the
 * cache or are written through to memory and never exist in a
 * cache-dirty state with respect to program visibility.  Attempts to
81
 * map System RAM with this mapping type will fail.
82 83 84 85
 *
 * MEMREMAP_WC - establish a writecombine mapping, whereby writes may
 * be coalesced together (e.g. in the CPU's write buffers), but is otherwise
 * uncached. Attempts to map System RAM with this mapping type will fail.
86 87 88
 */
void *memremap(resource_size_t offset, size_t size, unsigned long flags)
{
89 90
	int is_ram = region_intersects(offset, size,
				       IORESOURCE_SYSTEM_RAM, IORES_DESC_NONE);
91 92
	void *addr = NULL;

93 94 95
	if (!flags)
		return NULL;

96 97 98 99 100 101 102 103 104 105 106 107
	if (is_ram == REGION_MIXED) {
		WARN_ONCE(1, "memremap attempted on mixed range %pa size: %#lx\n",
				&offset, (unsigned long) size);
		return NULL;
	}

	/* Try all mapping types requested until one returns non-NULL */
	if (flags & MEMREMAP_WB) {
		/*
		 * MEMREMAP_WB is special in that it can be satisifed
		 * from the direct map.  Some archs depend on the
		 * capability of memremap() to autodetect cases where
108
		 * the requested range is potentially in System RAM.
109 110
		 */
		if (is_ram == REGION_INTERSECTS)
111
			addr = try_ram_remap(offset, size, flags);
112
		if (!addr)
113
			addr = arch_memremap_wb(offset, size);
114 115 116
	}

	/*
117 118
	 * If we don't have a mapping yet and other request flags are
	 * present then we will be attempting to establish a new virtual
119
	 * address mapping.  Enforce that this mapping is not aliasing
120
	 * System RAM.
121
	 */
122
	if (!addr && is_ram == REGION_INTERSECTS && flags != MEMREMAP_WB) {
123 124 125 126 127
		WARN_ONCE(1, "memremap attempted on ram %pa size: %#lx\n",
				&offset, (unsigned long) size);
		return NULL;
	}

128
	if (!addr && (flags & MEMREMAP_WT))
129
		addr = ioremap_wt(offset, size);
130 131 132

	if (!addr && (flags & MEMREMAP_WC))
		addr = ioremap_wc(offset, size);
133 134 135 136 137 138 139 140 141 142 143

	return addr;
}
EXPORT_SYMBOL(memremap);

void memunmap(void *addr)
{
	if (is_vmalloc_addr(addr))
		iounmap((void __iomem *) addr);
}
EXPORT_SYMBOL(memunmap);
144 145 146

static void devm_memremap_release(struct device *dev, void *res)
{
147
	memunmap(*(void **)res);
148 149 150 151 152 153 154 155 156 157 158 159
}

static int devm_memremap_match(struct device *dev, void *res, void *match_data)
{
	return *(void **)res == match_data;
}

void *devm_memremap(struct device *dev, resource_size_t offset,
		size_t size, unsigned long flags)
{
	void **ptr, *addr;

160 161
	ptr = devres_alloc_node(devm_memremap_release, sizeof(*ptr), GFP_KERNEL,
			dev_to_node(dev));
162
	if (!ptr)
163
		return ERR_PTR(-ENOMEM);
164 165 166 167 168

	addr = memremap(offset, size, flags);
	if (addr) {
		*ptr = addr;
		devres_add(dev, ptr);
169
	} else {
170
		devres_free(ptr);
171 172
		return ERR_PTR(-ENXIO);
	}
173 174 175 176 177 178 179

	return addr;
}
EXPORT_SYMBOL(devm_memremap);

void devm_memunmap(struct device *dev, void *addr)
{
180 181
	WARN_ON(devres_release(dev, devm_memremap_release,
				devm_memremap_match, addr));
182 183
}
EXPORT_SYMBOL(devm_memunmap);
184 185

#ifdef CONFIG_ZONE_DEVICE
186 187 188 189 190
static DEFINE_MUTEX(pgmap_lock);
static RADIX_TREE(pgmap_radix, GFP_KERNEL);
#define SECTION_MASK ~((1UL << PA_SECTION_SHIFT) - 1)
#define SECTION_SIZE (1UL << PA_SECTION_SHIFT)

191 192
struct page_map {
	struct resource res;
193 194
	struct percpu_ref *ref;
	struct dev_pagemap pgmap;
195
	struct vmem_altmap altmap;
196 197
};

198
static unsigned long order_at(struct resource *res, unsigned long pgoff)
199
{
200 201
	unsigned long phys_pgoff = PHYS_PFN(res->start) + pgoff;
	unsigned long nr_pages, mask;
202

203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222
	nr_pages = PHYS_PFN(resource_size(res));
	if (nr_pages == pgoff)
		return ULONG_MAX;

	/*
	 * What is the largest aligned power-of-2 range available from
	 * this resource pgoff to the end of the resource range,
	 * considering the alignment of the current pgoff?
	 */
	mask = phys_pgoff | rounddown_pow_of_two(nr_pages - pgoff);
	if (!mask)
		return ULONG_MAX;

	return find_first_bit(&mask, BITS_PER_LONG);
}

#define foreach_order_pgoff(res, order, pgoff) \
	for (pgoff = 0, order = order_at((res), pgoff); order < ULONG_MAX; \
			pgoff += 1UL << order, order = order_at((res), pgoff))

223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250
#if IS_ENABLED(CONFIG_DEVICE_PRIVATE)
int device_private_entry_fault(struct vm_area_struct *vma,
		       unsigned long addr,
		       swp_entry_t entry,
		       unsigned int flags,
		       pmd_t *pmdp)
{
	struct page *page = device_private_entry_to_page(entry);

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

251 252 253
static void pgmap_radix_release(struct resource *res)
{
	unsigned long pgoff, order;
254 255

	mutex_lock(&pgmap_lock);
256 257
	foreach_order_pgoff(res, order, pgoff)
		radix_tree_delete(&pgmap_radix, PHYS_PFN(res->start) + pgoff);
258
	mutex_unlock(&pgmap_lock);
259 260

	synchronize_rcu();
261 262
}

263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285
static unsigned long pfn_first(struct page_map *page_map)
{
	struct dev_pagemap *pgmap = &page_map->pgmap;
	const struct resource *res = &page_map->res;
	struct vmem_altmap *altmap = pgmap->altmap;
	unsigned long pfn;

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

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

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

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

286
static void devm_memremap_pages_release(struct device *dev, void *data)
287
{
288 289 290
	struct page_map *page_map = data;
	struct resource *res = &page_map->res;
	resource_size_t align_start, align_size;
291
	struct dev_pagemap *pgmap = &page_map->pgmap;
292 293 294 295
	unsigned long pfn;

	for_each_device_pfn(pfn, page_map)
		put_page(pfn_to_page(pfn));
296

297 298 299 300 301
	if (percpu_ref_tryget_live(pgmap->ref)) {
		dev_WARN(dev, "%s: page mapping is still live!\n", __func__);
		percpu_ref_put(pgmap->ref);
	}

302
	/* pages are dead and unused, undo the arch mapping */
303 304
	align_start = res->start & ~(SECTION_SIZE - 1);
	align_size = ALIGN(resource_size(res), SECTION_SIZE);
305

306
	mem_hotplug_begin();
307
	arch_remove_memory(align_start, align_size);
308
	mem_hotplug_done();
309

310
	untrack_pfn(NULL, PHYS_PFN(align_start), align_size);
311
	pgmap_radix_release(res);
312 313
	dev_WARN_ONCE(dev, pgmap->altmap && pgmap->altmap->alloc,
			"%s: failed to free all reserved pages\n", __func__);
314 315 316 317 318 319 320 321 322
}

/* assumes rcu_read_lock() held at entry */
struct dev_pagemap *find_dev_pagemap(resource_size_t phys)
{
	struct page_map *page_map;

	WARN_ON_ONCE(!rcu_read_lock_held());

323
	page_map = radix_tree_lookup(&pgmap_radix, PHYS_PFN(phys));
324
	return page_map ? &page_map->pgmap : NULL;
325 326
}

327 328 329 330
/**
 * devm_memremap_pages - remap and provide memmap backing for the given resource
 * @dev: hosting device for @res
 * @res: "host memory" address range
331
 * @ref: a live per-cpu reference count
332 333
 * @altmap: optional descriptor for allocating the memmap from @res
 *
334 335
 * Notes:
 * 1/ @ref must be 'live' on entry and 'dead' before devm_memunmap_pages() time
336 337 338 339
 *    (or devm release event). The expected order of events is that @ref has
 *    been through percpu_ref_kill() before devm_memremap_pages_release(). The
 *    wait for the completion of all references being dropped and
 *    percpu_ref_exit() must occur after devm_memremap_pages_release().
340 341 342 343
 *
 * 2/ @res is expected to be a host memory range that could feasibly be
 *    treated as a "System RAM" range, i.e. not a device mmio range, but
 *    this is not enforced.
344 345
 */
void *devm_memremap_pages(struct device *dev, struct resource *res,
346
		struct percpu_ref *ref, struct vmem_altmap *altmap)
347
{
348 349
	resource_size_t align_start, align_size, align_end;
	unsigned long pfn, pgoff, order;
350
	pgprot_t pgprot = PAGE_KERNEL;
351
	struct dev_pagemap *pgmap;
352
	struct page_map *page_map;
353
	int error, nid, is_ram, i = 0;
354 355 356 357

	align_start = res->start & ~(SECTION_SIZE - 1);
	align_size = ALIGN(res->start + resource_size(res), SECTION_SIZE)
		- align_start;
358 359
	is_ram = region_intersects(align_start, align_size,
		IORESOURCE_SYSTEM_RAM, IORES_DESC_NONE);
360 361 362 363 364 365 366 367 368 369

	if (is_ram == REGION_MIXED) {
		WARN_ONCE(1, "%s attempted on mixed region %pr\n",
				__func__, res);
		return ERR_PTR(-ENXIO);
	}

	if (is_ram == REGION_INTERSECTS)
		return __va(res->start);

370 371 372
	if (!ref)
		return ERR_PTR(-EINVAL);

373 374
	page_map = devres_alloc_node(devm_memremap_pages_release,
			sizeof(*page_map), GFP_KERNEL, dev_to_node(dev));
375 376
	if (!page_map)
		return ERR_PTR(-ENOMEM);
377
	pgmap = &page_map->pgmap;
378 379 380

	memcpy(&page_map->res, res, sizeof(*res));

381 382 383 384 385
	pgmap->dev = dev;
	if (altmap) {
		memcpy(&page_map->altmap, altmap, sizeof(*altmap));
		pgmap->altmap = &page_map->altmap;
	}
386
	pgmap->ref = ref;
387
	pgmap->res = &page_map->res;
388 389 390 391
	pgmap->type = MEMORY_DEVICE_HOST;
	pgmap->page_fault = NULL;
	pgmap->page_free = NULL;
	pgmap->data = NULL;
392

393 394
	mutex_lock(&pgmap_lock);
	error = 0;
395
	align_end = align_start + align_size - 1;
396 397

	foreach_order_pgoff(res, order, pgoff) {
398 399 400
		struct dev_pagemap *dup;

		rcu_read_lock();
401
		dup = find_dev_pagemap(res->start + PFN_PHYS(pgoff));
402 403 404 405 406 407 408
		rcu_read_unlock();
		if (dup) {
			dev_err(dev, "%s: %pr collides with mapping for %s\n",
					__func__, res, dev_name(dup->dev));
			error = -EBUSY;
			break;
		}
409 410
		error = __radix_tree_insert(&pgmap_radix,
				PHYS_PFN(res->start) + pgoff, order, page_map);
411 412 413 414 415 416 417 418 419
		if (error) {
			dev_err(dev, "%s: failed: %d\n", __func__, error);
			break;
		}
	}
	mutex_unlock(&pgmap_lock);
	if (error)
		goto err_radix;

420 421
	nid = dev_to_node(dev);
	if (nid < 0)
422
		nid = numa_mem_id();
423

424 425 426 427 428
	error = track_pfn_remap(NULL, &pgprot, PHYS_PFN(align_start), 0,
			align_size);
	if (error)
		goto err_pfn_remap;

429
	mem_hotplug_begin();
430
	error = arch_add_memory(nid, align_start, align_size, false);
431 432 433 434
	if (!error)
		move_pfn_range_to_zone(&NODE_DATA(nid)->node_zones[ZONE_DEVICE],
					align_start >> PAGE_SHIFT,
					align_size >> PAGE_SHIFT);
435
	mem_hotplug_done();
436 437
	if (error)
		goto err_add_memory;
438

439 440 441
	for_each_device_pfn(pfn, page_map) {
		struct page *page = pfn_to_page(pfn);

442 443 444 445 446 447 448
		/*
		 * ZONE_DEVICE pages union ->lru with a ->pgmap back
		 * pointer.  It is a bug if a ZONE_DEVICE page is ever
		 * freed or placed on a driver-private list.  Seed the
		 * storage with LIST_POISON* values.
		 */
		list_del(&page->lru);
449
		page->pgmap = pgmap;
450
		percpu_ref_get(ref);
451 452
		if (!(++i % 1024))
			cond_resched();
453
	}
454 455
	devres_add(dev, page_map);
	return __va(res->start);
456 457

 err_add_memory:
458 459
	untrack_pfn(NULL, PHYS_PFN(align_start), align_size);
 err_pfn_remap:
460 461 462 463
 err_radix:
	pgmap_radix_release(res);
	devres_free(page_map);
	return ERR_PTR(error);
464 465
}
EXPORT_SYMBOL(devm_memremap_pages);
466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482

unsigned long vmem_altmap_offset(struct vmem_altmap *altmap)
{
	/* number of pfns from base where pfn_to_page() is valid */
	return altmap->reserve + altmap->free;
}

void vmem_altmap_free(struct vmem_altmap *altmap, unsigned long nr_pfns)
{
	altmap->alloc -= nr_pfns;
}

struct vmem_altmap *to_vmem_altmap(unsigned long memmap_start)
{
	/*
	 * 'memmap_start' is the virtual address for the first "struct
	 * page" in this range of the vmemmap array.  In the case of
483
	 * CONFIG_SPARSEMEM_VMEMMAP a page_to_pfn conversion is simple
484 485 486 487 488 489 490 491
	 * pointer arithmetic, so we can perform this to_vmem_altmap()
	 * conversion without concern for the initialization state of
	 * the struct page fields.
	 */
	struct page *page = (struct page *) memmap_start;
	struct dev_pagemap *pgmap;

	/*
492
	 * Unconditionally retrieve a dev_pagemap associated with the
493 494 495 496 497 498 499 500 501 502
	 * given physical address, this is only for use in the
	 * arch_{add|remove}_memory() for setting up and tearing down
	 * the memmap.
	 */
	rcu_read_lock();
	pgmap = find_dev_pagemap(__pfn_to_phys(page_to_pfn(page)));
	rcu_read_unlock();

	return pgmap ? pgmap->altmap : NULL;
}
503
#endif /* CONFIG_ZONE_DEVICE */
504 505


506 507
#if IS_ENABLED(CONFIG_DEVICE_PRIVATE) ||  IS_ENABLED(CONFIG_DEVICE_PUBLIC)
void put_zone_device_private_or_public_page(struct page *page)
508 509 510 511 512 513 514 515 516 517 518 519 520
{
	int count = page_ref_dec_return(page);

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

		page->mapping = NULL;
521
		mem_cgroup_uncharge(page);
522 523 524 525 526

		page->pgmap->page_free(page, page->pgmap->data);
	} else if (!count)
		__put_page(page);
}
527 528
EXPORT_SYMBOL(put_zone_device_private_or_public_page);
#endif /* CONFIG_DEVICE_PRIVATE || CONFIG_DEVICE_PUBLIC */