[mirror_ubuntu-artful-kernel.git] / mm / iov_iter.c

#include <linux/export.h>
#include <linux/uio.h>
#include <linux/pagemap.h>
#include <linux/slab.h>
#include <linux/vmalloc.h>

#define iterate_iovec(i, n, __v, __p, skip, STEP) {	\
	size_t left;					\
	size_t wanted = n;				\
	__p = i->iov;					\
	__v.iov_len = min(n, __p->iov_len - skip);	\
	if (likely(__v.iov_len)) {			\
		__v.iov_base = __p->iov_base + skip;	\
		left = (STEP);				\
		__v.iov_len -= left;			\
		skip += __v.iov_len;			\
		n -= __v.iov_len;			\
	} else {					\
		left = 0;				\
	}						\
	while (unlikely(!left && n)) {			\
		__p++;					\
		__v.iov_len = min(n, __p->iov_len);	\
		if (unlikely(!__v.iov_len))		\
			continue;			\
		__v.iov_base = __p->iov_base;		\
		left = (STEP);				\
		__v.iov_len -= left;			\
		skip = __v.iov_len;			\
		n -= __v.iov_len;			\
	}						\
	n = wanted - n;					\
}

#define iterate_bvec(i, n, __v, __p, skip, STEP) {	\
	size_t wanted = n;				\
	__p = i->bvec;					\
	__v.bv_len = min_t(size_t, n, __p->bv_len - skip);	\
	if (likely(__v.bv_len)) {			\
		__v.bv_page = __p->bv_page;		\
		__v.bv_offset = __p->bv_offset + skip; 	\
		(void)(STEP);				\
		skip += __v.bv_len;			\
		n -= __v.bv_len;			\
	}						\
	while (unlikely(n)) {				\
		__p++;					\
		__v.bv_len = min_t(size_t, n, __p->bv_len);	\
		if (unlikely(!__v.bv_len))		\
			continue;			\
		__v.bv_page = __p->bv_page;		\
		__v.bv_offset = __p->bv_offset;		\
		(void)(STEP);				\
		skip = __v.bv_len;			\
		n -= __v.bv_len;			\
	}						\
	n = wanted;					\
}

#define iterate_all_kinds(i, n, v, I, B) {			\
	size_t skip = i->iov_offset;				\
	if (unlikely(i->type & ITER_BVEC)) {			\
		const struct bio_vec *bvec;			\
		struct bio_vec v;				\
		iterate_bvec(i, n, v, bvec, skip, (B))		\
	} else {						\
		const struct iovec *iov;			\
		struct iovec v;					\
		iterate_iovec(i, n, v, iov, skip, (I))		\
	}							\
}

#define iterate_and_advance(i, n, v, I, B) {			\
	size_t skip = i->iov_offset;				\
	if (unlikely(i->type & ITER_BVEC)) {			\
		const struct bio_vec *bvec;			\
		struct bio_vec v;				\
		iterate_bvec(i, n, v, bvec, skip, (B))		\
		if (skip == bvec->bv_len) {			\
			bvec++;					\
			skip = 0;				\
		}						\
		i->nr_segs -= bvec - i->bvec;			\
		i->bvec = bvec;					\
	} else {						\
		const struct iovec *iov;			\
		struct iovec v;					\
		iterate_iovec(i, n, v, iov, skip, (I))		\
		if (skip == iov->iov_len) {			\
			iov++;					\
			skip = 0;				\
		}						\
		i->nr_segs -= iov - i->iov;			\
		i->iov = iov;					\
	}							\
	i->count -= n;						\
	i->iov_offset = skip;					\
}

static size_t copy_to_iter_iovec(void *from, size_t bytes, struct iov_iter *i)
{
	size_t skip, copy, left, wanted;
	const struct iovec *iov;
	char __user *buf;

	if (unlikely(bytes > i->count))
		bytes = i->count;

	if (unlikely(!bytes))
		return 0;

	wanted = bytes;
	iov = i->iov;
	skip = i->iov_offset;
	buf = iov->iov_base + skip;
	copy = min(bytes, iov->iov_len - skip);

	left = __copy_to_user(buf, from, copy);
	copy -= left;
	skip += copy;
	from += copy;
	bytes -= copy;
	while (unlikely(!left && bytes)) {
		iov++;
		buf = iov->iov_base;
		copy = min(bytes, iov->iov_len);
		left = __copy_to_user(buf, from, copy);
		copy -= left;
		skip = copy;
		from += copy;
		bytes -= copy;
	}

	if (skip == iov->iov_len) {
		iov++;
		skip = 0;
	}
	i->count -= wanted - bytes;
	i->nr_segs -= iov - i->iov;
	i->iov = iov;
	i->iov_offset = skip;
	return wanted - bytes;
}

static size_t copy_page_to_iter_iovec(struct page *page, size_t offset, size_t bytes,
			 struct iov_iter *i)
{
	size_t skip, copy, left, wanted;
	const struct iovec *iov;
	char __user *buf;
	void *kaddr, *from;

	if (unlikely(bytes > i->count))
		bytes = i->count;

	if (unlikely(!bytes))
		return 0;

	wanted = bytes;
	iov = i->iov;
	skip = i->iov_offset;
	buf = iov->iov_base + skip;
	copy = min(bytes, iov->iov_len - skip);

	if (!fault_in_pages_writeable(buf, copy)) {
		kaddr = kmap_atomic(page);
		from = kaddr + offset;

		/* first chunk, usually the only one */
		left = __copy_to_user_inatomic(buf, from, copy);
		copy -= left;
		skip += copy;
		from += copy;
		bytes -= copy;

		while (unlikely(!left && bytes)) {
			iov++;
			buf = iov->iov_base;
			copy = min(bytes, iov->iov_len);
			left = __copy_to_user_inatomic(buf, from, copy);
			copy -= left;
			skip = copy;
			from += copy;
			bytes -= copy;
		}
		if (likely(!bytes)) {
			kunmap_atomic(kaddr);
			goto done;
		}
		offset = from - kaddr;
		buf += copy;
		kunmap_atomic(kaddr);
		copy = min(bytes, iov->iov_len - skip);
	}
	/* Too bad - revert to non-atomic kmap */
	kaddr = kmap(page);
	from = kaddr + offset;
	left = __copy_to_user(buf, from, copy);
	copy -= left;
	skip += copy;
	from += copy;
	bytes -= copy;
	while (unlikely(!left && bytes)) {
		iov++;
		buf = iov->iov_base;
		copy = min(bytes, iov->iov_len);
		left = __copy_to_user(buf, from, copy);
		copy -= left;
		skip = copy;
		from += copy;
		bytes -= copy;
	}
	kunmap(page);
done:
	if (skip == iov->iov_len) {
		iov++;
		skip = 0;
	}
	i->count -= wanted - bytes;
	i->nr_segs -= iov - i->iov;
	i->iov = iov;
	i->iov_offset = skip;
	return wanted - bytes;
}

static size_t copy_page_from_iter_iovec(struct page *page, size_t offset, size_t bytes,
			 struct iov_iter *i)
{
	size_t skip, copy, left, wanted;
	const struct iovec *iov;
	char __user *buf;
	void *kaddr, *to;

	if (unlikely(bytes > i->count))
		bytes = i->count;

	if (unlikely(!bytes))
		return 0;

	wanted = bytes;
	iov = i->iov;
	skip = i->iov_offset;
	buf = iov->iov_base + skip;
	copy = min(bytes, iov->iov_len - skip);

	if (!fault_in_pages_readable(buf, copy)) {
		kaddr = kmap_atomic(page);
		to = kaddr + offset;

		/* first chunk, usually the only one */
		left = __copy_from_user_inatomic(to, buf, copy);
		copy -= left;
		skip += copy;
		to += copy;
		bytes -= copy;

		while (unlikely(!left && bytes)) {
			iov++;
			buf = iov->iov_base;
			copy = min(bytes, iov->iov_len);
			left = __copy_from_user_inatomic(to, buf, copy);
			copy -= left;
			skip = copy;
			to += copy;
			bytes -= copy;
		}
		if (likely(!bytes)) {
			kunmap_atomic(kaddr);
			goto done;
		}
		offset = to - kaddr;
		buf += copy;
		kunmap_atomic(kaddr);
		copy = min(bytes, iov->iov_len - skip);
	}
	/* Too bad - revert to non-atomic kmap */
	kaddr = kmap(page);
	to = kaddr + offset;
	left = __copy_from_user(to, buf, copy);
	copy -= left;
	skip += copy;
	to += copy;
	bytes -= copy;
	while (unlikely(!left && bytes)) {
		iov++;
		buf = iov->iov_base;
		copy = min(bytes, iov->iov_len);
		left = __copy_from_user(to, buf, copy);
		copy -= left;
		skip = copy;
		to += copy;
		bytes -= copy;
	}
	kunmap(page);
done:
	if (skip == iov->iov_len) {
		iov++;
		skip = 0;
	}
	i->count -= wanted - bytes;
	i->nr_segs -= iov - i->iov;
	i->iov = iov;
	i->iov_offset = skip;
	return wanted - bytes;
}

/*
 * Fault in the first iovec of the given iov_iter, to a maximum length
 * of bytes. Returns 0 on success, or non-zero if the memory could not be
 * accessed (ie. because it is an invalid address).
 *
 * writev-intensive code may want this to prefault several iovecs -- that
 * would be possible (callers must not rely on the fact that _only_ the
 * first iovec will be faulted with the current implementation).
 */
int iov_iter_fault_in_readable(struct iov_iter *i, size_t bytes)
{
	if (!(i->type & ITER_BVEC)) {
		char __user *buf = i->iov->iov_base + i->iov_offset;
		bytes = min(bytes, i->iov->iov_len - i->iov_offset);
		return fault_in_pages_readable(buf, bytes);
	}
	return 0;
}
EXPORT_SYMBOL(iov_iter_fault_in_readable);

void iov_iter_init(struct iov_iter *i, int direction,
			const struct iovec *iov, unsigned long nr_segs,
			size_t count)
{
	/* It will get better.  Eventually... */
	if (segment_eq(get_fs(), KERNEL_DS))
		direction |= ITER_KVEC;
	i->type = direction;
	i->iov = iov;
	i->nr_segs = nr_segs;
	i->iov_offset = 0;
	i->count = count;
}
EXPORT_SYMBOL(iov_iter_init);

static void memcpy_from_page(char *to, struct page *page, size_t offset, size_t len)
{
	char *from = kmap_atomic(page);
	memcpy(to, from + offset, len);
	kunmap_atomic(from);
}

static void memcpy_to_page(struct page *page, size_t offset, char *from, size_t len)
{
	char *to = kmap_atomic(page);
	memcpy(to + offset, from, len);
	kunmap_atomic(to);
}

static void memzero_page(struct page *page, size_t offset, size_t len)
{
	char *addr = kmap_atomic(page);
	memset(addr + offset, 0, len);
	kunmap_atomic(addr);
}

static size_t copy_to_iter_bvec(void *from, size_t bytes, struct iov_iter *i)
{
	size_t skip, copy, wanted;
	const struct bio_vec *bvec;

	if (unlikely(bytes > i->count))
		bytes = i->count;

	if (unlikely(!bytes))
		return 0;

	wanted = bytes;
	bvec = i->bvec;
	skip = i->iov_offset;
	copy = min_t(size_t, bytes, bvec->bv_len - skip);

	memcpy_to_page(bvec->bv_page, skip + bvec->bv_offset, from, copy);
	skip += copy;
	from += copy;
	bytes -= copy;
	while (bytes) {
		bvec++;
		copy = min(bytes, (size_t)bvec->bv_len);
		memcpy_to_page(bvec->bv_page, bvec->bv_offset, from, copy);
		skip = copy;
		from += copy;
		bytes -= copy;
	}
	if (skip == bvec->bv_len) {
		bvec++;
		skip = 0;
	}
	i->count -= wanted - bytes;
	i->nr_segs -= bvec - i->bvec;
	i->bvec = bvec;
	i->iov_offset = skip;
	return wanted - bytes;
}

size_t copy_to_iter(void *addr, size_t bytes, struct iov_iter *i)
{
	if (i->type & ITER_BVEC)
		return copy_to_iter_bvec(addr, bytes, i);
	else
		return copy_to_iter_iovec(addr, bytes, i);
}
EXPORT_SYMBOL(copy_to_iter);

size_t copy_from_iter(void *addr, size_t bytes, struct iov_iter *i)
{
	char *to = addr;
	if (unlikely(bytes > i->count))
		bytes = i->count;

	if (unlikely(!bytes))
		return 0;

	iterate_and_advance(i, bytes, v,
		__copy_from_user((to += v.iov_len) - v.iov_len, v.iov_base,
				 v.iov_len),
		memcpy_from_page((to += v.bv_len) - v.bv_len, v.bv_page,
				 v.bv_offset, v.bv_len)
	)

	return bytes;
}
EXPORT_SYMBOL(copy_from_iter);

size_t copy_page_to_iter(struct page *page, size_t offset, size_t bytes,
			 struct iov_iter *i)
{
	if (i->type & (ITER_BVEC|ITER_KVEC)) {
		void *kaddr = kmap_atomic(page);
		size_t wanted = copy_to_iter(kaddr + offset, bytes, i);
		kunmap_atomic(kaddr);
		return wanted;
	} else
		return copy_page_to_iter_iovec(page, offset, bytes, i);
}
EXPORT_SYMBOL(copy_page_to_iter);

size_t copy_page_from_iter(struct page *page, size_t offset, size_t bytes,
			 struct iov_iter *i)
{
	if (i->type & ITER_BVEC) {
		void *kaddr = kmap_atomic(page);
		size_t wanted = copy_from_iter(kaddr + offset, bytes, i);
		kunmap_atomic(kaddr);
		return wanted;
	} else
		return copy_page_from_iter_iovec(page, offset, bytes, i);
}
EXPORT_SYMBOL(copy_page_from_iter);

size_t iov_iter_zero(size_t bytes, struct iov_iter *i)
{
	if (unlikely(bytes > i->count))
		bytes = i->count;

	if (unlikely(!bytes))
		return 0;

	iterate_and_advance(i, bytes, v,
		__clear_user(v.iov_base, v.iov_len),
		memzero_page(v.bv_page, v.bv_offset, v.bv_len)
	)

	return bytes;
}
EXPORT_SYMBOL(iov_iter_zero);

size_t iov_iter_copy_from_user_atomic(struct page *page,
		struct iov_iter *i, unsigned long offset, size_t bytes)
{
	char *kaddr = kmap_atomic(page), *p = kaddr + offset;
	iterate_all_kinds(i, bytes, v,
		__copy_from_user_inatomic((p += v.iov_len) - v.iov_len,
					  v.iov_base, v.iov_len),
		memcpy_from_page((p += v.bv_len) - v.bv_len, v.bv_page,
				 v.bv_offset, v.bv_len)
	)
	kunmap_atomic(kaddr);
	return bytes;
}
EXPORT_SYMBOL(iov_iter_copy_from_user_atomic);

void iov_iter_advance(struct iov_iter *i, size_t size)
{
	iterate_and_advance(i, size, v, 0, 0)
}
EXPORT_SYMBOL(iov_iter_advance);

/*
 * Return the count of just the current iov_iter segment.
 */
size_t iov_iter_single_seg_count(const struct iov_iter *i)
{
	if (i->nr_segs == 1)
		return i->count;
	else if (i->type & ITER_BVEC)
		return min(i->count, i->bvec->bv_len - i->iov_offset);
	else
		return min(i->count, i->iov->iov_len - i->iov_offset);
}
EXPORT_SYMBOL(iov_iter_single_seg_count);

unsigned long iov_iter_alignment(const struct iov_iter *i)
{
	unsigned long res = 0;
	size_t size = i->count;

	if (!size)
		return 0;

	iterate_all_kinds(i, size, v,
		(res |= (unsigned long)v.iov_base | v.iov_len, 0),
		res |= v.bv_offset | v.bv_len
	)
	return res;
}
EXPORT_SYMBOL(iov_iter_alignment);

ssize_t iov_iter_get_pages(struct iov_iter *i,
		   struct page **pages, size_t maxsize, unsigned maxpages,
		   size_t *start)
{
	if (maxsize > i->count)
		maxsize = i->count;

	if (!maxsize)
		return 0;

	iterate_all_kinds(i, maxsize, v, ({
		unsigned long addr = (unsigned long)v.iov_base;
		size_t len = v.iov_len + (*start = addr & (PAGE_SIZE - 1));
		int n;
		int res;

		if (len > maxpages * PAGE_SIZE)
			len = maxpages * PAGE_SIZE;
		addr &= ~(PAGE_SIZE - 1);
		n = DIV_ROUND_UP(len, PAGE_SIZE);
		res = get_user_pages_fast(addr, n, (i->type & WRITE) != WRITE, pages);
		if (unlikely(res < 0))
			return res;
		return (res == n ? len : res * PAGE_SIZE) - *start;
	0;}),({
		/* can't be more than PAGE_SIZE */
		*start = v.bv_offset;
		get_page(*pages = v.bv_page);
		return v.bv_len;
	})
	)
	return 0;
}
EXPORT_SYMBOL(iov_iter_get_pages);

static struct page **get_pages_array(size_t n)
{
	struct page **p = kmalloc(n * sizeof(struct page *), GFP_KERNEL);
	if (!p)
		p = vmalloc(n * sizeof(struct page *));
	return p;
}

ssize_t iov_iter_get_pages_alloc(struct iov_iter *i,
		   struct page ***pages, size_t maxsize,
		   size_t *start)
{
	struct page **p;

	if (maxsize > i->count)
		maxsize = i->count;

	if (!maxsize)
		return 0;

	iterate_all_kinds(i, maxsize, v, ({
		unsigned long addr = (unsigned long)v.iov_base;
		size_t len = v.iov_len + (*start = addr & (PAGE_SIZE - 1));
		int n;
		int res;

		addr &= ~(PAGE_SIZE - 1);
		n = DIV_ROUND_UP(len, PAGE_SIZE);
		p = get_pages_array(n);
		if (!p)
			return -ENOMEM;
		res = get_user_pages_fast(addr, n, (i->type & WRITE) != WRITE, p);
		if (unlikely(res < 0)) {
			kvfree(p);
			return res;
		}
		*pages = p;
		return (res == n ? len : res * PAGE_SIZE) - *start;
	0;}),({
		/* can't be more than PAGE_SIZE */
		*start = v.bv_offset;
		*pages = p = get_pages_array(1);
		if (!p)
			return -ENOMEM;
		get_page(*p = v.bv_page);
		return v.bv_len;
	})
	)
	return 0;
}
EXPORT_SYMBOL(iov_iter_get_pages_alloc);

int iov_iter_npages(const struct iov_iter *i, int maxpages)
{
	size_t size = i->count;
	int npages = 0;

	if (!size)
		return 0;

	iterate_all_kinds(i, size, v, ({
		unsigned long p = (unsigned long)v.iov_base;
		npages += DIV_ROUND_UP(p + v.iov_len, PAGE_SIZE)
			- p / PAGE_SIZE;
		if (npages >= maxpages)
			return maxpages;
	0;}),({
		npages++;
		if (npages >= maxpages)
			return maxpages;
	})
	)
	return npages;
}
EXPORT_SYMBOL(iov_iter_npages);
Commit	Line	Data
4f18cd31 AV	1	#include <linux/export.h>
	2	#include <linux/uio.h>
	3	#include <linux/pagemap.h>
91f79c43 AV	4	#include <linux/slab.h>
91f79c43 AV	5	#include <linux/vmalloc.h>
4f18cd31	6
04a31165 AV	7	#define iterate_iovec(i, n, __v, __p, skip, STEP) { \
	8	size_t left; \
	9	size_t wanted = n; \
	10	__p = i->iov; \
	11	__v.iov_len = min(n, __p->iov_len - skip); \
	12	if (likely(__v.iov_len)) { \
	13	__v.iov_base = __p->iov_base + skip; \
	14	left = (STEP); \
	15	__v.iov_len -= left; \
	16	skip += __v.iov_len; \
	17	n -= __v.iov_len; \
	18	} else { \
	19	left = 0; \
	20	} \
	21	while (unlikely(!left && n)) { \
	22	__p++; \
	23	__v.iov_len = min(n, __p->iov_len); \
	24	if (unlikely(!__v.iov_len)) \
	25	continue; \
	26	__v.iov_base = __p->iov_base; \
	27	left = (STEP); \
	28	__v.iov_len -= left; \
	29	skip = __v.iov_len; \
	30	n -= __v.iov_len; \
	31	} \
	32	n = wanted - n; \
	33	}
	34
	35	#define iterate_bvec(i, n, __v, __p, skip, STEP) { \
	36	size_t wanted = n; \
	37	__p = i->bvec; \
	38	__v.bv_len = min_t(size_t, n, __p->bv_len - skip); \
	39	if (likely(__v.bv_len)) { \
	40	__v.bv_page = __p->bv_page; \
	41	__v.bv_offset = __p->bv_offset + skip; \
	42	(void)(STEP); \
	43	skip += __v.bv_len; \
	44	n -= __v.bv_len; \
	45	} \
	46	while (unlikely(n)) { \
	47	__p++; \
	48	__v.bv_len = min_t(size_t, n, __p->bv_len); \
	49	if (unlikely(!__v.bv_len)) \
	50	continue; \
	51	__v.bv_page = __p->bv_page; \
	52	__v.bv_offset = __p->bv_offset; \
	53	(void)(STEP); \
	54	skip = __v.bv_len; \
	55	n -= __v.bv_len; \
	56	} \
	57	n = wanted; \
	58	}
	59
	60	#define iterate_all_kinds(i, n, v, I, B) { \
	61	size_t skip = i->iov_offset; \
	62	if (unlikely(i->type & ITER_BVEC)) { \
	63	const struct bio_vec *bvec; \
	64	struct bio_vec v; \
	65	iterate_bvec(i, n, v, bvec, skip, (B)) \
	66	} else { \
	67	const struct iovec *iov; \
	68	struct iovec v; \
	69	iterate_iovec(i, n, v, iov, skip, (I)) \
	70	} \
71	}
72
7ce2a91e AV	73	#define iterate_and_advance(i, n, v, I, B) { \
	74	size_t skip = i->iov_offset; \
	75	if (unlikely(i->type & ITER_BVEC)) { \
	76	const struct bio_vec *bvec; \
	77	struct bio_vec v; \
	78	iterate_bvec(i, n, v, bvec, skip, (B)) \
	79	if (skip == bvec->bv_len) { \
	80	bvec++; \
	81	skip = 0; \
	82	} \
	83	i->nr_segs -= bvec - i->bvec; \
	84	i->bvec = bvec; \
	85	} else { \
	86	const struct iovec *iov; \
	87	struct iovec v; \
	88	iterate_iovec(i, n, v, iov, skip, (I)) \
	89	if (skip == iov->iov_len) { \
	90	iov++; \
	91	skip = 0; \
	92	} \
	93	i->nr_segs -= iov - i->iov; \
	94	i->iov = iov; \
	95	} \
	96	i->count -= n; \
	97	i->iov_offset = skip; \
	98	}
	99
c35e0248 MW	100	static size_t copy_to_iter_iovec(void from, size_t bytes, struct iov_iter i)
	101	{
	102	size_t skip, copy, left, wanted;
	103	const struct iovec *iov;
	104	char __user *buf;
	105
	106	if (unlikely(bytes > i->count))
	107	bytes = i->count;
	108
	109	if (unlikely(!bytes))
	110	return 0;
	111
	112	wanted = bytes;
	113	iov = i->iov;
	114	skip = i->iov_offset;
	115	buf = iov->iov_base + skip;
	116	copy = min(bytes, iov->iov_len - skip);
	117
	118	left = __copy_to_user(buf, from, copy);
	119	copy -= left;
	120	skip += copy;
	121	from += copy;
	122	bytes -= copy;
	123	while (unlikely(!left && bytes)) {
	124	iov++;
	125	buf = iov->iov_base;
	126	copy = min(bytes, iov->iov_len);
	127	left = __copy_to_user(buf, from, copy);
	128	copy -= left;
	129	skip = copy;
	130	from += copy;
	131	bytes -= copy;
	132	}
	133
	134	if (skip == iov->iov_len) {
	135	iov++;
	136	skip = 0;
	137	}
	138	i->count -= wanted - bytes;
	139	i->nr_segs -= iov - i->iov;
	140	i->iov = iov;
	141	i->iov_offset = skip;
	142	return wanted - bytes;
	143	}
	144
62a8067a	145	static size_t copy_page_to_iter_iovec(struct page *page, size_t offset, size_t bytes,
4f18cd31 AV	146	struct iov_iter *i)
	147	{
	148	size_t skip, copy, left, wanted;
	149	const struct iovec *iov;
	150	char __user *buf;
	151	void kaddr, from;
	152
	153	if (unlikely(bytes > i->count))
	154	bytes = i->count;
	155
	156	if (unlikely(!bytes))
	157	return 0;
	158
	159	wanted = bytes;
	160	iov = i->iov;
	161	skip = i->iov_offset;
	162	buf = iov->iov_base + skip;
	163	copy = min(bytes, iov->iov_len - skip);
	164
	165	if (!fault_in_pages_writeable(buf, copy)) {
	166	kaddr = kmap_atomic(page);
	167	from = kaddr + offset;
	168
	169	/* first chunk, usually the only one */
	170	left = __copy_to_user_inatomic(buf, from, copy);
	171	copy -= left;
	172	skip += copy;
	173	from += copy;
	174	bytes -= copy;
	175
	176	while (unlikely(!left && bytes)) {
	177	iov++;
	178	buf = iov->iov_base;
	179	copy = min(bytes, iov->iov_len);
	180	left = __copy_to_user_inatomic(buf, from, copy);
	181	copy -= left;
	182	skip = copy;
	183	from += copy;
	184	bytes -= copy;
	185	}
	186	if (likely(!bytes)) {
	187	kunmap_atomic(kaddr);
	188	goto done;
	189	}
	190	offset = from - kaddr;
	191	buf += copy;
	192	kunmap_atomic(kaddr);
	193	copy = min(bytes, iov->iov_len - skip);
	194	}
	195	/* Too bad - revert to non-atomic kmap */
	196	kaddr = kmap(page);
	197	from = kaddr + offset;
	198	left = __copy_to_user(buf, from, copy);
	199	copy -= left;
	200	skip += copy;
	201	from += copy;
	202	bytes -= copy;
	203	while (unlikely(!left && bytes)) {
	204	iov++;
	205	buf = iov->iov_base;
	206	copy = min(bytes, iov->iov_len);
	207	left = __copy_to_user(buf, from, copy);
	208	copy -= left;
	209	skip = copy;
210	from += copy;
211	bytes -= copy;
212	}
213	kunmap(page);
214	done:
81055e58 AV	215	if (skip == iov->iov_len) {
	216	iov++;
	217	skip = 0;
	218	}
4f18cd31 AV	219	i->count -= wanted - bytes;
	220	i->nr_segs -= iov - i->iov;
	221	i->iov = iov;
	222	i->iov_offset = skip;
	223	return wanted - bytes;
	224	}
4f18cd31	225
62a8067a	226	static size_t copy_page_from_iter_iovec(struct page *page, size_t offset, size_t bytes,
f0d1bec9 AV	227	struct iov_iter *i)
	228	{
	229	size_t skip, copy, left, wanted;
	230	const struct iovec *iov;
	231	char __user *buf;
	232	void kaddr, to;
	233
	234	if (unlikely(bytes > i->count))
	235	bytes = i->count;
	236
	237	if (unlikely(!bytes))
	238	return 0;
	239
	240	wanted = bytes;
	241	iov = i->iov;
	242	skip = i->iov_offset;
	243	buf = iov->iov_base + skip;
	244	copy = min(bytes, iov->iov_len - skip);
	245
	246	if (!fault_in_pages_readable(buf, copy)) {
	247	kaddr = kmap_atomic(page);
	248	to = kaddr + offset;
	249
	250	/* first chunk, usually the only one */
	251	left = __copy_from_user_inatomic(to, buf, copy);
	252	copy -= left;
	253	skip += copy;
	254	to += copy;
	255	bytes -= copy;
	256
	257	while (unlikely(!left && bytes)) {
	258	iov++;
	259	buf = iov->iov_base;
	260	copy = min(bytes, iov->iov_len);
	261	left = __copy_from_user_inatomic(to, buf, copy);
	262	copy -= left;
	263	skip = copy;
	264	to += copy;
	265	bytes -= copy;
	266	}
	267	if (likely(!bytes)) {
	268	kunmap_atomic(kaddr);
	269	goto done;
	270	}
	271	offset = to - kaddr;
	272	buf += copy;
	273	kunmap_atomic(kaddr);
	274	copy = min(bytes, iov->iov_len - skip);
	275	}
	276	/* Too bad - revert to non-atomic kmap */
	277	kaddr = kmap(page);
	278	to = kaddr + offset;
	279	left = __copy_from_user(to, buf, copy);
	280	copy -= left;
	281	skip += copy;
	282	to += copy;
	283	bytes -= copy;
	284	while (unlikely(!left && bytes)) {
	285	iov++;
	286	buf = iov->iov_base;
	287	copy = min(bytes, iov->iov_len);
	288	left = __copy_from_user(to, buf, copy);
	289	copy -= left;
	290	skip = copy;
291	to += copy;
292	bytes -= copy;
293	}
294	kunmap(page);
295	done:
81055e58 AV	296	if (skip == iov->iov_len) {
	297	iov++;
	298	skip = 0;
	299	}
f0d1bec9 AV	300	i->count -= wanted - bytes;
	301	i->nr_segs -= iov - i->iov;
	302	i->iov = iov;
	303	i->iov_offset = skip;
	304	return wanted - bytes;
	305	}
f0d1bec9	306
4f18cd31 AV	307	/*
	308	* Fault in the first iovec of the given iov_iter, to a maximum length
	309	* of bytes. Returns 0 on success, or non-zero if the memory could not be
	310	* accessed (ie. because it is an invalid address).
	311	*
	312	* writev-intensive code may want this to prefault several iovecs -- that
	313	* would be possible (callers must not rely on the fact that _only_ the
	314	* first iovec will be faulted with the current implementation).
	315	*/
	316	int iov_iter_fault_in_readable(struct iov_iter *i, size_t bytes)
	317	{
62a8067a AV	318	if (!(i->type & ITER_BVEC)) {
	319	char __user *buf = i->iov->iov_base + i->iov_offset;
	320	bytes = min(bytes, i->iov->iov_len - i->iov_offset);
	321	return fault_in_pages_readable(buf, bytes);
	322	}
	323	return 0;
4f18cd31 AV	324	}
	325	EXPORT_SYMBOL(iov_iter_fault_in_readable);
	326
71d8e532 AV	327	void iov_iter_init(struct iov_iter *i, int direction,
	328	const struct iovec *iov, unsigned long nr_segs,
	329	size_t count)
	330	{
	331	/* It will get better. Eventually... */
	332	if (segment_eq(get_fs(), KERNEL_DS))
62a8067a	333	direction \|= ITER_KVEC;
71d8e532 AV	334	i->type = direction;
	335	i->iov = iov;
	336	i->nr_segs = nr_segs;
	337	i->iov_offset = 0;
	338	i->count = count;
	339	}
	340	EXPORT_SYMBOL(iov_iter_init);
7b2c99d1	341
62a8067a AV	342	static void memcpy_from_page(char to, struct page page, size_t offset, size_t len)
	343	{
	344	char *from = kmap_atomic(page);
	345	memcpy(to, from + offset, len);
	346	kunmap_atomic(from);
	347	}
	348
	349	static void memcpy_to_page(struct page page, size_t offset, char from, size_t len)
	350	{
	351	char *to = kmap_atomic(page);
	352	memcpy(to + offset, from, len);
	353	kunmap_atomic(to);
	354	}
	355
c35e0248 MW	356	static void memzero_page(struct page *page, size_t offset, size_t len)
	357	{
	358	char *addr = kmap_atomic(page);
	359	memset(addr + offset, 0, len);
	360	kunmap_atomic(addr);
	361	}
	362
	363	static size_t copy_to_iter_bvec(void from, size_t bytes, struct iov_iter i)
62a8067a AV	364	{
	365	size_t skip, copy, wanted;
	366	const struct bio_vec *bvec;
62a8067a AV	367
	368	if (unlikely(bytes > i->count))
	369	bytes = i->count;
	370
	371	if (unlikely(!bytes))
	372	return 0;
	373
	374	wanted = bytes;
	375	bvec = i->bvec;
	376	skip = i->iov_offset;
	377	copy = min_t(size_t, bytes, bvec->bv_len - skip);
	378
62a8067a AV	379	memcpy_to_page(bvec->bv_page, skip + bvec->bv_offset, from, copy);
	380	skip += copy;
	381	from += copy;
	382	bytes -= copy;
	383	while (bytes) {
	384	bvec++;
	385	copy = min(bytes, (size_t)bvec->bv_len);
	386	memcpy_to_page(bvec->bv_page, bvec->bv_offset, from, copy);
	387	skip = copy;
	388	from += copy;
	389	bytes -= copy;
	390	}
62a8067a AV	391	if (skip == bvec->bv_len) {
	392	bvec++;
	393	skip = 0;
	394	}
	395	i->count -= wanted - bytes;
	396	i->nr_segs -= bvec - i->bvec;
	397	i->bvec = bvec;
	398	i->iov_offset = skip;
	399	return wanted - bytes;
	400	}
	401
d271524a	402	size_t copy_to_iter(void addr, size_t bytes, struct iov_iter i)
c35e0248	403	{
d271524a AV	404	if (i->type & ITER_BVEC)
	405	return copy_to_iter_bvec(addr, bytes, i);
	406	else
	407	return copy_to_iter_iovec(addr, bytes, i);
c35e0248	408	}
d271524a	409	EXPORT_SYMBOL(copy_to_iter);
c35e0248	410
d271524a	411	size_t copy_from_iter(void addr, size_t bytes, struct iov_iter i)
c35e0248	412	{
0dbca9a4 AV	413	char *to = addr;
	414	if (unlikely(bytes > i->count))
	415	bytes = i->count;
	416
	417	if (unlikely(!bytes))
	418	return 0;
	419
	420	iterate_and_advance(i, bytes, v,
	421	__copy_from_user((to += v.iov_len) - v.iov_len, v.iov_base,
	422	v.iov_len),
	423	memcpy_from_page((to += v.bv_len) - v.bv_len, v.bv_page,
	424	v.bv_offset, v.bv_len)
	425	)
	426
	427	return bytes;
c35e0248	428	}
d271524a	429	EXPORT_SYMBOL(copy_from_iter);
c35e0248	430
62a8067a AV	431	size_t copy_page_to_iter(struct page *page, size_t offset, size_t bytes,
	432	struct iov_iter *i)
	433	{
d271524a AV	434	if (i->type & (ITER_BVEC\|ITER_KVEC)) {
	435	void *kaddr = kmap_atomic(page);
	436	size_t wanted = copy_to_iter(kaddr + offset, bytes, i);
	437	kunmap_atomic(kaddr);
	438	return wanted;
	439	} else
62a8067a AV	440	return copy_page_to_iter_iovec(page, offset, bytes, i);
	441	}
	442	EXPORT_SYMBOL(copy_page_to_iter);
	443
	444	size_t copy_page_from_iter(struct page *page, size_t offset, size_t bytes,
	445	struct iov_iter *i)
	446	{
d271524a AV	447	if (i->type & ITER_BVEC) {
	448	void *kaddr = kmap_atomic(page);
	449	size_t wanted = copy_from_iter(kaddr + offset, bytes, i);
	450	kunmap_atomic(kaddr);
	451	return wanted;
	452	} else
62a8067a AV	453	return copy_page_from_iter_iovec(page, offset, bytes, i);
	454	}
	455	EXPORT_SYMBOL(copy_page_from_iter);
	456
c35e0248 MW	457	size_t iov_iter_zero(size_t bytes, struct iov_iter *i)
c35e0248 MW	458	{
8442fa46 AV	459	if (unlikely(bytes > i->count))
	460	bytes = i->count;
	461
	462	if (unlikely(!bytes))
	463	return 0;
	464
	465	iterate_and_advance(i, bytes, v,
	466	__clear_user(v.iov_base, v.iov_len),
	467	memzero_page(v.bv_page, v.bv_offset, v.bv_len)
	468	)
	469
	470	return bytes;
c35e0248 MW	471	}
	472	EXPORT_SYMBOL(iov_iter_zero);
	473
62a8067a AV	474	size_t iov_iter_copy_from_user_atomic(struct page *page,
	475	struct iov_iter *i, unsigned long offset, size_t bytes)
	476	{
04a31165 AV	477	char kaddr = kmap_atomic(page), p = kaddr + offset;
	478	iterate_all_kinds(i, bytes, v,
	479	__copy_from_user_inatomic((p += v.iov_len) - v.iov_len,
	480	v.iov_base, v.iov_len),
	481	memcpy_from_page((p += v.bv_len) - v.bv_len, v.bv_page,
	482	v.bv_offset, v.bv_len)
	483	)
	484	kunmap_atomic(kaddr);
	485	return bytes;
62a8067a AV	486	}
	487	EXPORT_SYMBOL(iov_iter_copy_from_user_atomic);
	488
	489	void iov_iter_advance(struct iov_iter *i, size_t size)
	490	{
7ce2a91e	491	iterate_and_advance(i, size, v, 0, 0)
62a8067a AV	492	}
	493	EXPORT_SYMBOL(iov_iter_advance);
	494
	495	/*
	496	* Return the count of just the current iov_iter segment.
	497	*/
	498	size_t iov_iter_single_seg_count(const struct iov_iter *i)
	499	{
	500	if (i->nr_segs == 1)
	501	return i->count;
	502	else if (i->type & ITER_BVEC)
62a8067a	503	return min(i->count, i->bvec->bv_len - i->iov_offset);
ad0eab92 PM	504	else
ad0eab92 PM	505	return min(i->count, i->iov->iov_len - i->iov_offset);
62a8067a AV	506	}
	507	EXPORT_SYMBOL(iov_iter_single_seg_count);
	508
	509	unsigned long iov_iter_alignment(const struct iov_iter *i)
	510	{
04a31165 AV	511	unsigned long res = 0;
	512	size_t size = i->count;
	513
	514	if (!size)
	515	return 0;
	516
	517	iterate_all_kinds(i, size, v,
	518	(res \|= (unsigned long)v.iov_base \| v.iov_len, 0),
	519	res \|= v.bv_offset \| v.bv_len
	520	)
	521	return res;
62a8067a AV	522	}
	523	EXPORT_SYMBOL(iov_iter_alignment);
	524
	525	ssize_t iov_iter_get_pages(struct iov_iter *i,
2c80929c	526	struct page **pages, size_t maxsize, unsigned maxpages,
62a8067a AV	527	size_t *start)
62a8067a AV	528	{
e5393fae AV	529	if (maxsize > i->count)
	530	maxsize = i->count;
	531
	532	if (!maxsize)
	533	return 0;
	534
	535	iterate_all_kinds(i, maxsize, v, ({
	536	unsigned long addr = (unsigned long)v.iov_base;
	537	size_t len = v.iov_len + (*start = addr & (PAGE_SIZE - 1));
	538	int n;
	539	int res;
	540
	541	if (len > maxpages * PAGE_SIZE)
	542	len = maxpages * PAGE_SIZE;
	543	addr &= ~(PAGE_SIZE - 1);
	544	n = DIV_ROUND_UP(len, PAGE_SIZE);
	545	res = get_user_pages_fast(addr, n, (i->type & WRITE) != WRITE, pages);
	546	if (unlikely(res < 0))
	547	return res;
	548	return (res == n ? len : res * PAGE_SIZE) - *start;
	549	0;}),({
	550	/* can't be more than PAGE_SIZE */
	551	*start = v.bv_offset;
	552	get_page(*pages = v.bv_page);
	553	return v.bv_len;
	554	})
	555	)
	556	return 0;
62a8067a AV	557	}
	558	EXPORT_SYMBOL(iov_iter_get_pages);
	559
1b17f1f2 AV	560	static struct page **get_pages_array(size_t n)
	561	{
	562	struct page *p = kmalloc(n sizeof(struct page *), GFP_KERNEL);
	563	if (!p)
	564	p = vmalloc(n * sizeof(struct page *));
	565	return p;
	566	}
	567
62a8067a AV	568	ssize_t iov_iter_get_pages_alloc(struct iov_iter *i,
	569	struct page ***pages, size_t maxsize,
	570	size_t *start)
	571	{
1b17f1f2 AV	572	struct page **p;
	573
	574	if (maxsize > i->count)
	575	maxsize = i->count;
	576
	577	if (!maxsize)
	578	return 0;
	579
	580	iterate_all_kinds(i, maxsize, v, ({
	581	unsigned long addr = (unsigned long)v.iov_base;
	582	size_t len = v.iov_len + (*start = addr & (PAGE_SIZE - 1));
	583	int n;
	584	int res;
	585
	586	addr &= ~(PAGE_SIZE - 1);
	587	n = DIV_ROUND_UP(len, PAGE_SIZE);
	588	p = get_pages_array(n);
	589	if (!p)
	590	return -ENOMEM;
	591	res = get_user_pages_fast(addr, n, (i->type & WRITE) != WRITE, p);
	592	if (unlikely(res < 0)) {
	593	kvfree(p);
	594	return res;
	595	}
	596	*pages = p;
	597	return (res == n ? len : res * PAGE_SIZE) - *start;
	598	0;}),({
	599	/* can't be more than PAGE_SIZE */
	600	*start = v.bv_offset;
	601	*pages = p = get_pages_array(1);
	602	if (!p)
	603	return -ENOMEM;
	604	get_page(*p = v.bv_page);
	605	return v.bv_len;
	606	})
	607	)
	608	return 0;
62a8067a AV	609	}
	610	EXPORT_SYMBOL(iov_iter_get_pages_alloc);
	611
	612	int iov_iter_npages(const struct iov_iter *i, int maxpages)
	613	{
e0f2dc40 AV	614	size_t size = i->count;
	615	int npages = 0;
	616
	617	if (!size)
	618	return 0;
	619
	620	iterate_all_kinds(i, size, v, ({
	621	unsigned long p = (unsigned long)v.iov_base;
	622	npages += DIV_ROUND_UP(p + v.iov_len, PAGE_SIZE)
	623	- p / PAGE_SIZE;
	624	if (npages >= maxpages)
	625	return maxpages;
	626	0;}),({
	627	npages++;
	628	if (npages >= maxpages)
	629	return maxpages;
	630	})
	631	)
	632	return npages;
62a8067a	633	}
f67da30c	634	EXPORT_SYMBOL(iov_iter_npages);