dma-mapping.c 20.3 KB
Newer Older
L
Linus Torvalds 已提交
1
/*
2
 *  linux/arch/arm/mm/dma-mapping.c
L
Linus Torvalds 已提交
3 4 5 6 7 8 9 10 11 12 13
 *
 *  Copyright (C) 2000-2004 Russell King
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License version 2 as
 * published by the Free Software Foundation.
 *
 *  DMA uncached mapping support.
 */
#include <linux/module.h>
#include <linux/mm.h>
14
#include <linux/gfp.h>
L
Linus Torvalds 已提交
15 16 17 18 19
#include <linux/errno.h>
#include <linux/list.h>
#include <linux/init.h>
#include <linux/device.h>
#include <linux/dma-mapping.h>
20
#include <linux/highmem.h>
21
#include <linux/slab.h>
L
Linus Torvalds 已提交
22

23
#include <asm/memory.h>
24
#include <asm/highmem.h>
L
Linus Torvalds 已提交
25 26
#include <asm/cacheflush.h>
#include <asm/tlbflush.h>
27
#include <asm/sizes.h>
28
#include <asm/mach/arch.h>
29

30 31
#include "mm.h"

32 33 34 35 36 37 38 39 40 41 42 43
/*
 * The DMA API is built upon the notion of "buffer ownership".  A buffer
 * is either exclusively owned by the CPU (and therefore may be accessed
 * by it) or exclusively owned by the DMA device.  These helper functions
 * represent the transitions between these two ownership states.
 *
 * Note, however, that on later ARMs, this notion does not work due to
 * speculative prefetches.  We model our approach on the assumption that
 * the CPU does do speculative prefetches, which means we clean caches
 * before transfers and delay cache invalidation until transfer completion.
 *
 */
44
static void __dma_page_cpu_to_dev(struct page *, unsigned long,
45
		size_t, enum dma_data_direction);
46
static void __dma_page_dev_to_cpu(struct page *, unsigned long,
47 48
		size_t, enum dma_data_direction);

49 50 51 52 53 54 55 56 57 58 59 60 61 62
/**
 * arm_dma_map_page - map a portion of a page for streaming DMA
 * @dev: valid struct device pointer, or NULL for ISA and EISA-like devices
 * @page: page that buffer resides in
 * @offset: offset into page for start of buffer
 * @size: size of buffer to map
 * @dir: DMA transfer direction
 *
 * Ensure that any data held in the cache is appropriately discarded
 * or written back.
 *
 * The device owns this memory once this call has completed.  The CPU
 * can regain ownership by calling dma_unmap_page().
 */
63
static dma_addr_t arm_dma_map_page(struct device *dev, struct page *page,
64 65 66
	     unsigned long offset, size_t size, enum dma_data_direction dir,
	     struct dma_attrs *attrs)
{
67 68 69
	if (!arch_is_coherent())
		__dma_page_cpu_to_dev(page, offset, size, dir);
	return pfn_to_dma(dev, page_to_pfn(page)) + offset;
70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85
}

/**
 * arm_dma_unmap_page - unmap a buffer previously mapped through dma_map_page()
 * @dev: valid struct device pointer, or NULL for ISA and EISA-like devices
 * @handle: DMA address of buffer
 * @size: size of buffer (same as passed to dma_map_page)
 * @dir: DMA transfer direction (same as passed to dma_map_page)
 *
 * Unmap a page streaming mode DMA translation.  The handle and size
 * must match what was provided in the previous dma_map_page() call.
 * All other usages are undefined.
 *
 * After this call, reads by the CPU to the buffer are guaranteed to see
 * whatever the device wrote there.
 */
86
static void arm_dma_unmap_page(struct device *dev, dma_addr_t handle,
87 88 89
		size_t size, enum dma_data_direction dir,
		struct dma_attrs *attrs)
{
90 91 92
	if (!arch_is_coherent())
		__dma_page_dev_to_cpu(pfn_to_page(dma_to_pfn(dev, handle)),
				      handle & ~PAGE_MASK, size, dir);
93 94
}

95
static void arm_dma_sync_single_for_cpu(struct device *dev,
96 97 98 99
		dma_addr_t handle, size_t size, enum dma_data_direction dir)
{
	unsigned int offset = handle & (PAGE_SIZE - 1);
	struct page *page = pfn_to_page(dma_to_pfn(dev, handle-offset));
100 101
	if (!arch_is_coherent())
		__dma_page_dev_to_cpu(page, offset, size, dir);
102 103
}

104
static void arm_dma_sync_single_for_device(struct device *dev,
105 106 107 108
		dma_addr_t handle, size_t size, enum dma_data_direction dir)
{
	unsigned int offset = handle & (PAGE_SIZE - 1);
	struct page *page = pfn_to_page(dma_to_pfn(dev, handle-offset));
109 110
	if (!arch_is_coherent())
		__dma_page_cpu_to_dev(page, offset, size, dir);
111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127
}

static int arm_dma_set_mask(struct device *dev, u64 dma_mask);

struct dma_map_ops arm_dma_ops = {
	.map_page		= arm_dma_map_page,
	.unmap_page		= arm_dma_unmap_page,
	.map_sg			= arm_dma_map_sg,
	.unmap_sg		= arm_dma_unmap_sg,
	.sync_single_for_cpu	= arm_dma_sync_single_for_cpu,
	.sync_single_for_device	= arm_dma_sync_single_for_device,
	.sync_sg_for_cpu	= arm_dma_sync_sg_for_cpu,
	.sync_sg_for_device	= arm_dma_sync_sg_for_device,
	.set_dma_mask		= arm_dma_set_mask,
};
EXPORT_SYMBOL(arm_dma_ops);

128 129
static u64 get_coherent_dma_mask(struct device *dev)
{
130
	u64 mask = (u64)arm_dma_limit;
131 132 133 134 135 136 137 138 139 140 141 142 143

	if (dev) {
		mask = dev->coherent_dma_mask;

		/*
		 * Sanity check the DMA mask - it must be non-zero, and
		 * must be able to be satisfied by a DMA allocation.
		 */
		if (mask == 0) {
			dev_warn(dev, "coherent DMA mask is unset\n");
			return 0;
		}

144
		if ((~mask) & (u64)arm_dma_limit) {
145 146
			dev_warn(dev, "coherent DMA mask %#llx is smaller "
				 "than system GFP_DMA mask %#llx\n",
147
				 mask, (u64)arm_dma_limit);
148 149 150
			return 0;
		}
	}
L
Linus Torvalds 已提交
151

152 153 154
	return mask;
}

155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216
/*
 * Allocate a DMA buffer for 'dev' of size 'size' using the
 * specified gfp mask.  Note that 'size' must be page aligned.
 */
static struct page *__dma_alloc_buffer(struct device *dev, size_t size, gfp_t gfp)
{
	unsigned long order = get_order(size);
	struct page *page, *p, *e;
	void *ptr;
	u64 mask = get_coherent_dma_mask(dev);

#ifdef CONFIG_DMA_API_DEBUG
	u64 limit = (mask + 1) & ~mask;
	if (limit && size >= limit) {
		dev_warn(dev, "coherent allocation too big (requested %#x mask %#llx)\n",
			size, mask);
		return NULL;
	}
#endif

	if (!mask)
		return NULL;

	if (mask < 0xffffffffULL)
		gfp |= GFP_DMA;

	page = alloc_pages(gfp, order);
	if (!page)
		return NULL;

	/*
	 * Now split the huge page and free the excess pages
	 */
	split_page(page, order);
	for (p = page + (size >> PAGE_SHIFT), e = page + (1 << order); p < e; p++)
		__free_page(p);

	/*
	 * Ensure that the allocated pages are zeroed, and that any data
	 * lurking in the kernel direct-mapped region is invalidated.
	 */
	ptr = page_address(page);
	memset(ptr, 0, size);
	dmac_flush_range(ptr, ptr + size);
	outer_flush_range(__pa(ptr), __pa(ptr) + size);

	return page;
}

/*
 * Free a DMA buffer.  'size' must be page aligned.
 */
static void __dma_free_buffer(struct page *page, size_t size)
{
	struct page *e = page + (size >> PAGE_SHIFT);

	while (page < e) {
		__free_page(page);
		page++;
	}
}

217
#ifdef CONFIG_MMU
218

219
#define CONSISTENT_OFFSET(x)	(((unsigned long)(x) - consistent_base) >> PAGE_SHIFT)
L
Linus Torvalds 已提交
220
#define CONSISTENT_PTE_INDEX(x) (((unsigned long)(x) - consistent_base) >> PMD_SHIFT)
221

L
Linus Torvalds 已提交
222
/*
223
 * These are the page tables (2MB each) covering uncached, DMA consistent allocations
L
Linus Torvalds 已提交
224
 */
225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241
static pte_t **consistent_pte;

#define DEFAULT_CONSISTENT_DMA_SIZE SZ_2M

unsigned long consistent_base = CONSISTENT_END - DEFAULT_CONSISTENT_DMA_SIZE;

void __init init_consistent_dma_size(unsigned long size)
{
	unsigned long base = CONSISTENT_END - ALIGN(size, SZ_2M);

	BUG_ON(consistent_pte); /* Check we're called before DMA region init */
	BUG_ON(base < VMALLOC_END);

	/* Grow region to accommodate specified size  */
	if (base < consistent_base)
		consistent_base = base;
}
L
Linus Torvalds 已提交
242

243
#include "vmregion.h"
L
Linus Torvalds 已提交
244

245 246
static struct arm_vmregion_head consistent_head = {
	.vm_lock	= __SPIN_LOCK_UNLOCKED(&consistent_head.vm_lock),
L
Linus Torvalds 已提交
247 248 249 250 251 252 253 254
	.vm_list	= LIST_HEAD_INIT(consistent_head.vm_list),
	.vm_end		= CONSISTENT_END,
};

#ifdef CONFIG_HUGETLB_PAGE
#error ARM Coherent DMA allocator does not (yet) support huge TLB
#endif

255 256 257 258 259 260 261
/*
 * Initialise the consistent memory allocation.
 */
static int __init consistent_init(void)
{
	int ret = 0;
	pgd_t *pgd;
R
Russell King 已提交
262
	pud_t *pud;
263 264 265
	pmd_t *pmd;
	pte_t *pte;
	int i = 0;
266
	unsigned long base = consistent_base;
267
	unsigned long num_ptes = (CONSISTENT_END - base) >> PMD_SHIFT;
268 269 270 271 272 273 274 275 276

	consistent_pte = kmalloc(num_ptes * sizeof(pte_t), GFP_KERNEL);
	if (!consistent_pte) {
		pr_err("%s: no memory\n", __func__);
		return -ENOMEM;
	}

	pr_debug("DMA memory: 0x%08lx - 0x%08lx:\n", base, CONSISTENT_END);
	consistent_head.vm_start = base;
277 278 279

	do {
		pgd = pgd_offset(&init_mm, base);
R
Russell King 已提交
280 281 282

		pud = pud_alloc(&init_mm, pgd, base);
		if (!pud) {
283
			pr_err("%s: no pud tables\n", __func__);
R
Russell King 已提交
284 285 286 287 288
			ret = -ENOMEM;
			break;
		}

		pmd = pmd_alloc(&init_mm, pud, base);
289
		if (!pmd) {
290
			pr_err("%s: no pmd tables\n", __func__);
291 292 293 294 295 296 297
			ret = -ENOMEM;
			break;
		}
		WARN_ON(!pmd_none(*pmd));

		pte = pte_alloc_kernel(pmd, base);
		if (!pte) {
298
			pr_err("%s: no pte tables\n", __func__);
299 300 301 302 303
			ret = -ENOMEM;
			break;
		}

		consistent_pte[i++] = pte;
304
		base += PMD_SIZE;
305 306 307 308 309 310 311
	} while (base < CONSISTENT_END);

	return ret;
}

core_initcall(consistent_init);

L
Linus Torvalds 已提交
312
static void *
313 314
__dma_alloc_remap(struct page *page, size_t size, gfp_t gfp, pgprot_t prot,
	const void *caller)
L
Linus Torvalds 已提交
315
{
316
	struct arm_vmregion *c;
317 318
	size_t align;
	int bit;
L
Linus Torvalds 已提交
319

320
	if (!consistent_pte) {
321
		pr_err("%s: not initialised\n", __func__);
322 323 324 325
		dump_stack();
		return NULL;
	}

326 327 328 329 330 331
	/*
	 * Align the virtual region allocation - maximum alignment is
	 * a section size, minimum is a page size.  This helps reduce
	 * fragmentation of the DMA space, and also prevents allocations
	 * smaller than a section from crossing a section boundary.
	 */
332
	bit = fls(size - 1);
333 334 335 336
	if (bit > SECTION_SHIFT)
		bit = SECTION_SHIFT;
	align = 1 << bit;

L
Linus Torvalds 已提交
337 338 339
	/*
	 * Allocate a virtual address in the consistent mapping region.
	 */
340
	c = arm_vmregion_alloc(&consistent_head, align, size,
341
			    gfp & ~(__GFP_DMA | __GFP_HIGHMEM), caller);
L
Linus Torvalds 已提交
342
	if (c) {
343 344 345
		pte_t *pte;
		int idx = CONSISTENT_PTE_INDEX(c->vm_start);
		u32 off = CONSISTENT_OFFSET(c->vm_start) & (PTRS_PER_PTE-1);
L
Linus Torvalds 已提交
346

347
		pte = consistent_pte[idx] + off;
L
Linus Torvalds 已提交
348 349 350 351 352
		c->vm_pages = page;

		do {
			BUG_ON(!pte_none(*pte));

R
Russell King 已提交
353
			set_pte_ext(pte, mk_pte(page, prot), 0);
L
Linus Torvalds 已提交
354 355
			page++;
			pte++;
356 357 358 359 360
			off++;
			if (off >= PTRS_PER_PTE) {
				off = 0;
				pte = consistent_pte[++idx];
			}
L
Linus Torvalds 已提交
361 362
		} while (size -= PAGE_SIZE);

363 364
		dsb();

L
Linus Torvalds 已提交
365 366 367 368
		return (void *)c->vm_start;
	}
	return NULL;
}
369 370 371 372 373 374 375 376 377 378 379

static void __dma_free_remap(void *cpu_addr, size_t size)
{
	struct arm_vmregion *c;
	unsigned long addr;
	pte_t *ptep;
	int idx;
	u32 off;

	c = arm_vmregion_find_remove(&consistent_head, (unsigned long)cpu_addr);
	if (!c) {
380
		pr_err("%s: trying to free invalid coherent area: %p\n",
381 382 383 384 385 386
		       __func__, cpu_addr);
		dump_stack();
		return;
	}

	if ((c->vm_end - c->vm_start) != size) {
387
		pr_err("%s: freeing wrong coherent size (%ld != %d)\n",
388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407
		       __func__, c->vm_end - c->vm_start, size);
		dump_stack();
		size = c->vm_end - c->vm_start;
	}

	idx = CONSISTENT_PTE_INDEX(c->vm_start);
	off = CONSISTENT_OFFSET(c->vm_start) & (PTRS_PER_PTE-1);
	ptep = consistent_pte[idx] + off;
	addr = c->vm_start;
	do {
		pte_t pte = ptep_get_and_clear(&init_mm, addr, ptep);

		ptep++;
		addr += PAGE_SIZE;
		off++;
		if (off >= PTRS_PER_PTE) {
			off = 0;
			ptep = consistent_pte[++idx];
		}

408
		if (pte_none(pte) || !pte_present(pte))
409 410
			pr_crit("%s: bad page in kernel page table\n",
				__func__);
411 412 413 414 415 416 417
	} while (size -= PAGE_SIZE);

	flush_tlb_kernel_range(c->vm_start, c->vm_end);

	arm_vmregion_free(&consistent_head, c);
}

418
#else	/* !CONFIG_MMU */
419

420
#define __dma_alloc_remap(page, size, gfp, prot, c)	page_address(page)
421 422 423 424
#define __dma_free_remap(addr, size)			do { } while (0)

#endif	/* CONFIG_MMU */

425 426
static void *
__dma_alloc(struct device *dev, size_t size, dma_addr_t *handle, gfp_t gfp,
427
	    pgprot_t prot, const void *caller)
428
{
429
	struct page *page;
430
	void *addr;
431

432 433 434 435 436 437 438 439 440
	/*
	 * Following is a work-around (a.k.a. hack) to prevent pages
	 * with __GFP_COMP being passed to split_page() which cannot
	 * handle them.  The real problem is that this flag probably
	 * should be 0 on ARM as it is not supported on this
	 * platform; see CONFIG_HUGETLBFS.
	 */
	gfp &= ~(__GFP_COMP);

441
	*handle = DMA_ERROR_CODE;
442
	size = PAGE_ALIGN(size);
443

444 445 446
	page = __dma_alloc_buffer(dev, size, gfp);
	if (!page)
		return NULL;
447

448
	if (!arch_is_coherent())
449
		addr = __dma_alloc_remap(page, size, gfp, prot, caller);
450 451
	else
		addr = page_address(page);
452

453
	if (addr)
454
		*handle = pfn_to_dma(dev, page_to_pfn(page));
455 456
	else
		__dma_free_buffer(page, size);
457

458 459
	return addr;
}
L
Linus Torvalds 已提交
460 461 462 463 464 465

/*
 * Allocate DMA-coherent memory space and return both the kernel remapped
 * virtual and bus address for that space.
 */
void *
A
Al Viro 已提交
466
dma_alloc_coherent(struct device *dev, size_t size, dma_addr_t *handle, gfp_t gfp)
L
Linus Torvalds 已提交
467
{
468 469 470 471 472
	void *memory;

	if (dma_alloc_from_coherent(dev, size, handle, &memory))
		return memory;

L
Linus Torvalds 已提交
473
	return __dma_alloc(dev, size, handle, gfp,
474 475
			   pgprot_dmacoherent(pgprot_kernel),
			   __builtin_return_address(0));
L
Linus Torvalds 已提交
476 477 478 479 480 481 482 483
}
EXPORT_SYMBOL(dma_alloc_coherent);

/*
 * Allocate a writecombining region, in much the same way as
 * dma_alloc_coherent above.
 */
void *
A
Al Viro 已提交
484
dma_alloc_writecombine(struct device *dev, size_t size, dma_addr_t *handle, gfp_t gfp)
L
Linus Torvalds 已提交
485 486
{
	return __dma_alloc(dev, size, handle, gfp,
487 488
			   pgprot_writecombine(pgprot_kernel),
			   __builtin_return_address(0));
L
Linus Torvalds 已提交
489 490 491 492 493 494
}
EXPORT_SYMBOL(dma_alloc_writecombine);

static int dma_mmap(struct device *dev, struct vm_area_struct *vma,
		    void *cpu_addr, dma_addr_t dma_addr, size_t size)
{
495 496
	int ret = -ENXIO;
#ifdef CONFIG_MMU
497 498
	unsigned long user_size, kern_size;
	struct arm_vmregion *c;
L
Linus Torvalds 已提交
499

500 501 502
	if (dma_mmap_from_coherent(dev, vma, cpu_addr, size, &ret))
		return ret;

L
Linus Torvalds 已提交
503 504
	user_size = (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;

505
	c = arm_vmregion_find(&consistent_head, (unsigned long)cpu_addr);
L
Linus Torvalds 已提交
506 507 508 509 510 511 512 513 514 515 516 517 518
	if (c) {
		unsigned long off = vma->vm_pgoff;

		kern_size = (c->vm_end - c->vm_start) >> PAGE_SHIFT;

		if (off < kern_size &&
		    user_size <= (kern_size - off)) {
			ret = remap_pfn_range(vma, vma->vm_start,
					      page_to_pfn(c->vm_pages) + off,
					      user_size << PAGE_SHIFT,
					      vma->vm_page_prot);
		}
	}
519
#endif	/* CONFIG_MMU */
L
Linus Torvalds 已提交
520 521 522 523 524 525 526

	return ret;
}

int dma_mmap_coherent(struct device *dev, struct vm_area_struct *vma,
		      void *cpu_addr, dma_addr_t dma_addr, size_t size)
{
527
	vma->vm_page_prot = pgprot_dmacoherent(vma->vm_page_prot);
L
Linus Torvalds 已提交
528 529 530 531 532 533 534 535 536 537 538 539 540 541
	return dma_mmap(dev, vma, cpu_addr, dma_addr, size);
}
EXPORT_SYMBOL(dma_mmap_coherent);

int dma_mmap_writecombine(struct device *dev, struct vm_area_struct *vma,
			  void *cpu_addr, dma_addr_t dma_addr, size_t size)
{
	vma->vm_page_prot = pgprot_writecombine(vma->vm_page_prot);
	return dma_mmap(dev, vma, cpu_addr, dma_addr, size);
}
EXPORT_SYMBOL(dma_mmap_writecombine);

/*
 * free a page as defined by the above mapping.
542
 * Must not be called with IRQs disabled.
L
Linus Torvalds 已提交
543 544 545
 */
void dma_free_coherent(struct device *dev, size_t size, void *cpu_addr, dma_addr_t handle)
{
546 547
	WARN_ON(irqs_disabled());

548 549 550
	if (dma_release_from_coherent(dev, get_order(size), cpu_addr))
		return;

551 552
	size = PAGE_ALIGN(size);

553 554
	if (!arch_is_coherent())
		__dma_free_remap(cpu_addr, size);
555

556
	__dma_free_buffer(pfn_to_page(dma_to_pfn(dev, handle)), size);
L
Linus Torvalds 已提交
557 558 559
}
EXPORT_SYMBOL(dma_free_coherent);

560
static void dma_cache_maint_page(struct page *page, unsigned long offset,
561 562
	size_t size, enum dma_data_direction dir,
	void (*op)(const void *, size_t, int))
563 564 565 566 567 568 569 570 571 572
{
	/*
	 * A single sg entry may refer to multiple physically contiguous
	 * pages.  But we still need to process highmem pages individually.
	 * If highmem is not configured then the bulk of this loop gets
	 * optimized out.
	 */
	size_t left = size;
	do {
		size_t len = left;
573 574 575 576 577 578 579 580 581 582 583 584 585
		void *vaddr;

		if (PageHighMem(page)) {
			if (len + offset > PAGE_SIZE) {
				if (offset >= PAGE_SIZE) {
					page += offset / PAGE_SIZE;
					offset %= PAGE_SIZE;
				}
				len = PAGE_SIZE - offset;
			}
			vaddr = kmap_high_get(page);
			if (vaddr) {
				vaddr += offset;
586
				op(vaddr, len, dir);
587
				kunmap_high(page);
588
			} else if (cache_is_vipt()) {
589 590
				/* unmapped pages might still be cached */
				vaddr = kmap_atomic(page);
591
				op(vaddr + offset, len, dir);
592
				kunmap_atomic(vaddr);
593
			}
594 595
		} else {
			vaddr = page_address(page) + offset;
596
			op(vaddr, len, dir);
597 598 599 600 601 602
		}
		offset = 0;
		page++;
		left -= len;
	} while (left);
}
603

604 605 606 607 608 609 610
/*
 * Make an area consistent for devices.
 * Note: Drivers should NOT use this function directly, as it will break
 * platforms with CONFIG_DMABOUNCE.
 * Use the driver DMA support - see dma-mapping.h (dma_sync_*)
 */
static void __dma_page_cpu_to_dev(struct page *page, unsigned long off,
611 612
	size_t size, enum dma_data_direction dir)
{
613 614
	unsigned long paddr;

615
	dma_cache_maint_page(page, off, size, dir, dmac_map_area);
616 617

	paddr = page_to_phys(page) + off;
618 619 620 621 622 623
	if (dir == DMA_FROM_DEVICE) {
		outer_inv_range(paddr, paddr + size);
	} else {
		outer_clean_range(paddr, paddr + size);
	}
	/* FIXME: non-speculating: flush on bidirectional mappings? */
624 625
}

626
static void __dma_page_dev_to_cpu(struct page *page, unsigned long off,
627 628
	size_t size, enum dma_data_direction dir)
{
629 630 631 632 633 634 635
	unsigned long paddr = page_to_phys(page) + off;

	/* FIXME: non-speculating: not required */
	/* don't bother invalidating if DMA to device */
	if (dir != DMA_TO_DEVICE)
		outer_inv_range(paddr, paddr + size);

636
	dma_cache_maint_page(page, off, size, dir, dmac_unmap_area);
637 638 639 640 641 642

	/*
	 * Mark the D-cache clean for this page to avoid extra flushing.
	 */
	if (dir != DMA_TO_DEVICE && off == 0 && size >= PAGE_SIZE)
		set_bit(PG_dcache_clean, &page->flags);
643
}
644

645
/**
646
 * arm_dma_map_sg - map a set of SG buffers for streaming mode DMA
647 648 649 650 651 652 653 654 655 656 657 658 659 660
 * @dev: valid struct device pointer, or NULL for ISA and EISA-like devices
 * @sg: list of buffers
 * @nents: number of buffers to map
 * @dir: DMA transfer direction
 *
 * Map a set of buffers described by scatterlist in streaming mode for DMA.
 * This is the scatter-gather version of the dma_map_single interface.
 * Here the scatter gather list elements are each tagged with the
 * appropriate dma address and length.  They are obtained via
 * sg_dma_{address,length}.
 *
 * Device ownership issues as mentioned for dma_map_single are the same
 * here.
 */
661 662
int arm_dma_map_sg(struct device *dev, struct scatterlist *sg, int nents,
		enum dma_data_direction dir, struct dma_attrs *attrs)
663
{
664
	struct dma_map_ops *ops = get_dma_ops(dev);
665
	struct scatterlist *s;
666
	int i, j;
667 668

	for_each_sg(sg, s, nents, i) {
669 670
		s->dma_address = ops->map_page(dev, sg_page(s), s->offset,
						s->length, dir, attrs);
671 672
		if (dma_mapping_error(dev, s->dma_address))
			goto bad_mapping;
673 674
	}
	return nents;
675 676 677

 bad_mapping:
	for_each_sg(sg, s, i, j)
678
		ops->unmap_page(dev, sg_dma_address(s), sg_dma_len(s), dir, attrs);
679
	return 0;
680 681 682
}

/**
683
 * arm_dma_unmap_sg - unmap a set of SG buffers mapped by dma_map_sg
684 685
 * @dev: valid struct device pointer, or NULL for ISA and EISA-like devices
 * @sg: list of buffers
686
 * @nents: number of buffers to unmap (same as was passed to dma_map_sg)
687 688 689 690 691
 * @dir: DMA transfer direction (same as was passed to dma_map_sg)
 *
 * Unmap a set of streaming mode DMA translations.  Again, CPU access
 * rules concerning calls here are the same as for dma_unmap_single().
 */
692 693
void arm_dma_unmap_sg(struct device *dev, struct scatterlist *sg, int nents,
		enum dma_data_direction dir, struct dma_attrs *attrs)
694
{
695
	struct dma_map_ops *ops = get_dma_ops(dev);
696
	struct scatterlist *s;
697

698 699 700
	int i;

	for_each_sg(sg, s, nents, i)
701
		ops->unmap_page(dev, sg_dma_address(s), sg_dma_len(s), dir, attrs);
702 703 704
}

/**
705
 * arm_dma_sync_sg_for_cpu
706 707 708 709 710
 * @dev: valid struct device pointer, or NULL for ISA and EISA-like devices
 * @sg: list of buffers
 * @nents: number of buffers to map (returned from dma_map_sg)
 * @dir: DMA transfer direction (same as was passed to dma_map_sg)
 */
711
void arm_dma_sync_sg_for_cpu(struct device *dev, struct scatterlist *sg,
712 713
			int nents, enum dma_data_direction dir)
{
714
	struct dma_map_ops *ops = get_dma_ops(dev);
715 716 717
	struct scatterlist *s;
	int i;

718 719 720
	for_each_sg(sg, s, nents, i)
		ops->sync_single_for_cpu(dev, sg_dma_address(s), s->length,
					 dir);
721 722 723
}

/**
724
 * arm_dma_sync_sg_for_device
725 726 727 728 729
 * @dev: valid struct device pointer, or NULL for ISA and EISA-like devices
 * @sg: list of buffers
 * @nents: number of buffers to map (returned from dma_map_sg)
 * @dir: DMA transfer direction (same as was passed to dma_map_sg)
 */
730
void arm_dma_sync_sg_for_device(struct device *dev, struct scatterlist *sg,
731 732
			int nents, enum dma_data_direction dir)
{
733
	struct dma_map_ops *ops = get_dma_ops(dev);
734 735 736
	struct scatterlist *s;
	int i;

737 738 739
	for_each_sg(sg, s, nents, i)
		ops->sync_single_for_device(dev, sg_dma_address(s), s->length,
					    dir);
740
}
741

742 743 744 745 746 747 748 749 750 751 752 753 754 755
/*
 * Return whether the given device DMA address mask can be supported
 * properly.  For example, if your device can only drive the low 24-bits
 * during bus mastering, then you would pass 0x00ffffff as the mask
 * to this function.
 */
int dma_supported(struct device *dev, u64 mask)
{
	if (mask < (u64)arm_dma_limit)
		return 0;
	return 1;
}
EXPORT_SYMBOL(dma_supported);

756
static int arm_dma_set_mask(struct device *dev, u64 dma_mask)
757 758 759 760 761 762 763 764 765
{
	if (!dev->dma_mask || !dma_supported(dev, dma_mask))
		return -EIO;

	*dev->dma_mask = dma_mask;

	return 0;
}

766 767 768 769
#define PREALLOC_DMA_DEBUG_ENTRIES	4096

static int __init dma_debug_do_init(void)
{
770 771 772
#ifdef CONFIG_MMU
	arm_vmregion_create_proc("dma-mappings", &consistent_head);
#endif
773 774 775 776
	dma_debug_init(PREALLOC_DMA_DEBUG_ENTRIES);
	return 0;
}
fs_initcall(dma_debug_do_init);