1 /* SPDX-License-Identifier: GPL-2.0-or-later */
3 * Berkeley style UIO structures - Alan Cox 1994.
8 #include <linux/kernel.h>
9 #include <linux/thread_info.h>
10 #include <linux/mm_types.h>
11 #include <uapi/linux/uio.h>
15 typedef unsigned int __bitwise iov_iter_extraction_t;
18 void *iov_base; /* and that should *never* hold a userland pointer */
32 #define ITER_SOURCE 1 // == WRITE
33 #define ITER_DEST 0 // == READ
35 struct iov_iter_state {
38 unsigned long nr_segs;
48 * Hack alert: overlay ubuf_iovec with iovec + count, so
49 * that the members resolve correctly regardless of the type
50 * of iterator used. This means that you can use:
52 * &iter->__ubuf_iovec or iter->__iov
54 * interchangably for the user_backed cases, hence simplifying
55 * some of the cases that need to deal with both.
59 * This really should be a const, but we cannot do that without
60 * also modifying any of the zero-filling iter init functions.
61 * Leave it non-const for now, but it should be treated as such.
63 struct iovec __ubuf_iovec;
66 /* use iter_iov() to get the current vec */
67 const struct iovec *__iov;
68 const struct kvec *kvec;
69 const struct bio_vec *bvec;
70 struct xarray *xarray;
77 unsigned long nr_segs;
82 static inline const struct iovec *iter_iov(const struct iov_iter *iter)
84 if (iter->iter_type == ITER_UBUF)
85 return (const struct iovec *) &iter->__ubuf_iovec;
89 #define iter_iov_addr(iter) (iter_iov(iter)->iov_base + (iter)->iov_offset)
90 #define iter_iov_len(iter) (iter_iov(iter)->iov_len - (iter)->iov_offset)
92 static inline enum iter_type iov_iter_type(const struct iov_iter *i)
97 static inline void iov_iter_save_state(struct iov_iter *iter,
98 struct iov_iter_state *state)
100 state->iov_offset = iter->iov_offset;
101 state->count = iter->count;
102 state->nr_segs = iter->nr_segs;
105 static inline bool iter_is_ubuf(const struct iov_iter *i)
107 return iov_iter_type(i) == ITER_UBUF;
110 static inline bool iter_is_iovec(const struct iov_iter *i)
112 return iov_iter_type(i) == ITER_IOVEC;
115 static inline bool iov_iter_is_kvec(const struct iov_iter *i)
117 return iov_iter_type(i) == ITER_KVEC;
120 static inline bool iov_iter_is_bvec(const struct iov_iter *i)
122 return iov_iter_type(i) == ITER_BVEC;
125 static inline bool iov_iter_is_discard(const struct iov_iter *i)
127 return iov_iter_type(i) == ITER_DISCARD;
130 static inline bool iov_iter_is_xarray(const struct iov_iter *i)
132 return iov_iter_type(i) == ITER_XARRAY;
135 static inline unsigned char iov_iter_rw(const struct iov_iter *i)
137 return i->data_source ? WRITE : READ;
140 static inline bool user_backed_iter(const struct iov_iter *i)
142 return iter_is_ubuf(i) || iter_is_iovec(i);
146 * Total number of bytes covered by an iovec.
148 * NOTE that it is not safe to use this function until all the iovec's
149 * segment lengths have been validated. Because the individual lengths can
150 * overflow a size_t when added together.
152 static inline size_t iov_length(const struct iovec *iov, unsigned long nr_segs)
157 for (seg = 0; seg < nr_segs; seg++)
158 ret += iov[seg].iov_len;
162 size_t copy_page_from_iter_atomic(struct page *page, size_t offset,
163 size_t bytes, struct iov_iter *i);
164 void iov_iter_advance(struct iov_iter *i, size_t bytes);
165 void iov_iter_revert(struct iov_iter *i, size_t bytes);
166 size_t fault_in_iov_iter_readable(const struct iov_iter *i, size_t bytes);
167 size_t fault_in_iov_iter_writeable(const struct iov_iter *i, size_t bytes);
168 size_t iov_iter_single_seg_count(const struct iov_iter *i);
169 size_t copy_page_to_iter(struct page *page, size_t offset, size_t bytes,
171 size_t copy_page_from_iter(struct page *page, size_t offset, size_t bytes,
174 size_t _copy_to_iter(const void *addr, size_t bytes, struct iov_iter *i);
175 size_t _copy_from_iter(void *addr, size_t bytes, struct iov_iter *i);
176 size_t _copy_from_iter_nocache(void *addr, size_t bytes, struct iov_iter *i);
178 static inline size_t copy_folio_to_iter(struct folio *folio, size_t offset,
179 size_t bytes, struct iov_iter *i)
181 return copy_page_to_iter(&folio->page, offset, bytes, i);
184 static inline size_t copy_folio_from_iter_atomic(struct folio *folio,
185 size_t offset, size_t bytes, struct iov_iter *i)
187 return copy_page_from_iter_atomic(&folio->page, offset, bytes, i);
190 size_t copy_page_to_iter_nofault(struct page *page, unsigned offset,
191 size_t bytes, struct iov_iter *i);
193 static __always_inline __must_check
194 size_t copy_to_iter(const void *addr, size_t bytes, struct iov_iter *i)
196 if (check_copy_size(addr, bytes, true))
197 return _copy_to_iter(addr, bytes, i);
201 static __always_inline __must_check
202 size_t copy_from_iter(void *addr, size_t bytes, struct iov_iter *i)
204 if (check_copy_size(addr, bytes, false))
205 return _copy_from_iter(addr, bytes, i);
209 static __always_inline __must_check
210 bool copy_from_iter_full(void *addr, size_t bytes, struct iov_iter *i)
212 size_t copied = copy_from_iter(addr, bytes, i);
213 if (likely(copied == bytes))
215 iov_iter_revert(i, copied);
219 static __always_inline __must_check
220 size_t copy_from_iter_nocache(void *addr, size_t bytes, struct iov_iter *i)
222 if (check_copy_size(addr, bytes, false))
223 return _copy_from_iter_nocache(addr, bytes, i);
227 static __always_inline __must_check
228 bool copy_from_iter_full_nocache(void *addr, size_t bytes, struct iov_iter *i)
230 size_t copied = copy_from_iter_nocache(addr, bytes, i);
231 if (likely(copied == bytes))
233 iov_iter_revert(i, copied);
237 #ifdef CONFIG_ARCH_HAS_UACCESS_FLUSHCACHE
239 * Note, users like pmem that depend on the stricter semantics of
240 * _copy_from_iter_flushcache() than _copy_from_iter_nocache() must check for
241 * IS_ENABLED(CONFIG_ARCH_HAS_UACCESS_FLUSHCACHE) before assuming that the
242 * destination is flushed from the cache on return.
244 size_t _copy_from_iter_flushcache(void *addr, size_t bytes, struct iov_iter *i);
246 #define _copy_from_iter_flushcache _copy_from_iter_nocache
249 #ifdef CONFIG_ARCH_HAS_COPY_MC
250 size_t _copy_mc_to_iter(const void *addr, size_t bytes, struct iov_iter *i);
251 static inline void iov_iter_set_copy_mc(struct iov_iter *i)
256 static inline bool iov_iter_is_copy_mc(const struct iov_iter *i)
261 #define _copy_mc_to_iter _copy_to_iter
262 static inline void iov_iter_set_copy_mc(struct iov_iter *i) { }
263 static inline bool iov_iter_is_copy_mc(const struct iov_iter *i)
269 size_t iov_iter_zero(size_t bytes, struct iov_iter *);
270 bool iov_iter_is_aligned(const struct iov_iter *i, unsigned addr_mask,
272 unsigned long iov_iter_alignment(const struct iov_iter *i);
273 unsigned long iov_iter_gap_alignment(const struct iov_iter *i);
274 void iov_iter_init(struct iov_iter *i, unsigned int direction, const struct iovec *iov,
275 unsigned long nr_segs, size_t count);
276 void iov_iter_kvec(struct iov_iter *i, unsigned int direction, const struct kvec *kvec,
277 unsigned long nr_segs, size_t count);
278 void iov_iter_bvec(struct iov_iter *i, unsigned int direction, const struct bio_vec *bvec,
279 unsigned long nr_segs, size_t count);
280 void iov_iter_discard(struct iov_iter *i, unsigned int direction, size_t count);
281 void iov_iter_xarray(struct iov_iter *i, unsigned int direction, struct xarray *xarray,
282 loff_t start, size_t count);
283 ssize_t iov_iter_get_pages2(struct iov_iter *i, struct page **pages,
284 size_t maxsize, unsigned maxpages, size_t *start);
285 ssize_t iov_iter_get_pages_alloc2(struct iov_iter *i, struct page ***pages,
286 size_t maxsize, size_t *start);
287 int iov_iter_npages(const struct iov_iter *i, int maxpages);
288 void iov_iter_restore(struct iov_iter *i, struct iov_iter_state *state);
290 const void *dup_iter(struct iov_iter *new, struct iov_iter *old, gfp_t flags);
292 static inline size_t iov_iter_count(const struct iov_iter *i)
298 * Cap the iov_iter by given limit; note that the second argument is
299 * *not* the new size - it's upper limit for such. Passing it a value
300 * greater than the amount of data in iov_iter is fine - it'll just do
301 * nothing in that case.
303 static inline void iov_iter_truncate(struct iov_iter *i, u64 count)
306 * count doesn't have to fit in size_t - comparison extends both
307 * operands to u64 here and any value that would be truncated by
308 * conversion in assignement is by definition greater than all
309 * values of size_t, including old i->count.
311 if (i->count > count)
316 * reexpand a previously truncated iterator; count must be no more than how much
319 static inline void iov_iter_reexpand(struct iov_iter *i, size_t count)
325 iov_iter_npages_cap(struct iov_iter *i, int maxpages, size_t max_bytes)
330 if (iov_iter_count(i) > max_bytes) {
331 shorted = iov_iter_count(i) - max_bytes;
332 iov_iter_truncate(i, max_bytes);
334 npages = iov_iter_npages(i, maxpages);
336 iov_iter_reexpand(i, iov_iter_count(i) + shorted);
341 struct iovec *iovec_from_user(const struct iovec __user *uvector,
342 unsigned long nr_segs, unsigned long fast_segs,
343 struct iovec *fast_iov, bool compat);
344 ssize_t import_iovec(int type, const struct iovec __user *uvec,
345 unsigned nr_segs, unsigned fast_segs, struct iovec **iovp,
347 ssize_t __import_iovec(int type, const struct iovec __user *uvec,
348 unsigned nr_segs, unsigned fast_segs, struct iovec **iovp,
349 struct iov_iter *i, bool compat);
350 int import_ubuf(int type, void __user *buf, size_t len, struct iov_iter *i);
352 static inline void iov_iter_ubuf(struct iov_iter *i, unsigned int direction,
353 void __user *buf, size_t count)
355 WARN_ON(direction & ~(READ | WRITE));
356 *i = (struct iov_iter) {
357 .iter_type = ITER_UBUF,
359 .data_source = direction,
365 /* Flags for iov_iter_get/extract_pages*() */
366 /* Allow P2PDMA on the extracted pages */
367 #define ITER_ALLOW_P2PDMA ((__force iov_iter_extraction_t)0x01)
369 ssize_t iov_iter_extract_pages(struct iov_iter *i, struct page ***pages,
370 size_t maxsize, unsigned int maxpages,
371 iov_iter_extraction_t extraction_flags,
375 * iov_iter_extract_will_pin - Indicate how pages from the iterator will be retained
376 * @iter: The iterator
378 * Examine the iterator and indicate by returning true or false as to how, if
379 * at all, pages extracted from the iterator will be retained by the extraction
382 * %true indicates that the pages will have a pin placed in them that the
383 * caller must unpin. This is must be done for DMA/async DIO to force fork()
384 * to forcibly copy a page for the child (the parent must retain the original
387 * %false indicates that no measures are taken and that it's up to the caller
388 * to retain the pages.
390 static inline bool iov_iter_extract_will_pin(const struct iov_iter *iter)
392 return user_backed_iter(iter);
396 ssize_t extract_iter_to_sg(struct iov_iter *iter, size_t len,
397 struct sg_table *sgtable, unsigned int sg_max,
398 iov_iter_extraction_t extraction_flags);