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;
52 * Hack alert: overlay ubuf_iovec with iovec + count, so
53 * that the members resolve correctly regardless of the type
54 * of iterator used. This means that you can use:
56 * &iter->__ubuf_iovec or iter->__iov
58 * interchangably for the user_backed cases, hence simplifying
59 * some of the cases that need to deal with both.
63 * This really should be a const, but we cannot do that without
64 * also modifying any of the zero-filling iter init functions.
65 * Leave it non-const for now, but it should be treated as such.
67 struct iovec __ubuf_iovec;
70 /* use iter_iov() to get the current vec */
71 const struct iovec *__iov;
72 const struct kvec *kvec;
73 const struct bio_vec *bvec;
74 struct xarray *xarray;
81 unsigned long nr_segs;
86 static inline const struct iovec *iter_iov(const struct iov_iter *iter)
88 if (iter->iter_type == ITER_UBUF)
89 return (const struct iovec *) &iter->__ubuf_iovec;
93 #define iter_iov_addr(iter) (iter_iov(iter)->iov_base + (iter)->iov_offset)
94 #define iter_iov_len(iter) (iter_iov(iter)->iov_len - (iter)->iov_offset)
96 static inline enum iter_type iov_iter_type(const struct iov_iter *i)
101 static inline void iov_iter_save_state(struct iov_iter *iter,
102 struct iov_iter_state *state)
104 state->iov_offset = iter->iov_offset;
105 state->count = iter->count;
106 state->nr_segs = iter->nr_segs;
109 static inline bool iter_is_ubuf(const struct iov_iter *i)
111 return iov_iter_type(i) == ITER_UBUF;
114 static inline bool iter_is_iovec(const struct iov_iter *i)
116 return iov_iter_type(i) == ITER_IOVEC;
119 static inline bool iov_iter_is_kvec(const struct iov_iter *i)
121 return iov_iter_type(i) == ITER_KVEC;
124 static inline bool iov_iter_is_bvec(const struct iov_iter *i)
126 return iov_iter_type(i) == ITER_BVEC;
129 static inline bool iov_iter_is_discard(const struct iov_iter *i)
131 return iov_iter_type(i) == ITER_DISCARD;
134 static inline bool iov_iter_is_xarray(const struct iov_iter *i)
136 return iov_iter_type(i) == ITER_XARRAY;
139 static inline unsigned char iov_iter_rw(const struct iov_iter *i)
141 return i->data_source ? WRITE : READ;
144 static inline bool user_backed_iter(const struct iov_iter *i)
146 return i->user_backed;
150 * Total number of bytes covered by an iovec.
152 * NOTE that it is not safe to use this function until all the iovec's
153 * segment lengths have been validated. Because the individual lengths can
154 * overflow a size_t when added together.
156 static inline size_t iov_length(const struct iovec *iov, unsigned long nr_segs)
161 for (seg = 0; seg < nr_segs; seg++)
162 ret += iov[seg].iov_len;
166 size_t copy_page_from_iter_atomic(struct page *page, unsigned offset,
167 size_t bytes, struct iov_iter *i);
168 void iov_iter_advance(struct iov_iter *i, size_t bytes);
169 void iov_iter_revert(struct iov_iter *i, size_t bytes);
170 size_t fault_in_iov_iter_readable(const struct iov_iter *i, size_t bytes);
171 size_t fault_in_iov_iter_writeable(const struct iov_iter *i, size_t bytes);
172 size_t iov_iter_single_seg_count(const struct iov_iter *i);
173 size_t copy_page_to_iter(struct page *page, size_t offset, size_t bytes,
175 size_t copy_page_from_iter(struct page *page, size_t offset, size_t bytes,
178 size_t _copy_to_iter(const void *addr, size_t bytes, struct iov_iter *i);
179 size_t _copy_from_iter(void *addr, size_t bytes, struct iov_iter *i);
180 size_t _copy_from_iter_nocache(void *addr, size_t bytes, struct iov_iter *i);
182 static inline size_t copy_folio_to_iter(struct folio *folio, size_t offset,
183 size_t bytes, struct iov_iter *i)
185 return copy_page_to_iter(&folio->page, offset, bytes, i);
187 size_t copy_page_to_iter_nofault(struct page *page, unsigned offset,
188 size_t bytes, struct iov_iter *i);
190 static __always_inline __must_check
191 size_t copy_to_iter(const void *addr, size_t bytes, struct iov_iter *i)
193 if (check_copy_size(addr, bytes, true))
194 return _copy_to_iter(addr, bytes, i);
198 static __always_inline __must_check
199 size_t copy_from_iter(void *addr, size_t bytes, struct iov_iter *i)
201 if (check_copy_size(addr, bytes, false))
202 return _copy_from_iter(addr, bytes, i);
206 static __always_inline __must_check
207 bool copy_from_iter_full(void *addr, size_t bytes, struct iov_iter *i)
209 size_t copied = copy_from_iter(addr, bytes, i);
210 if (likely(copied == bytes))
212 iov_iter_revert(i, copied);
216 static __always_inline __must_check
217 size_t copy_from_iter_nocache(void *addr, size_t bytes, struct iov_iter *i)
219 if (check_copy_size(addr, bytes, false))
220 return _copy_from_iter_nocache(addr, bytes, i);
224 static __always_inline __must_check
225 bool copy_from_iter_full_nocache(void *addr, size_t bytes, struct iov_iter *i)
227 size_t copied = copy_from_iter_nocache(addr, bytes, i);
228 if (likely(copied == bytes))
230 iov_iter_revert(i, copied);
234 #ifdef CONFIG_ARCH_HAS_UACCESS_FLUSHCACHE
236 * Note, users like pmem that depend on the stricter semantics of
237 * _copy_from_iter_flushcache() than _copy_from_iter_nocache() must check for
238 * IS_ENABLED(CONFIG_ARCH_HAS_UACCESS_FLUSHCACHE) before assuming that the
239 * destination is flushed from the cache on return.
241 size_t _copy_from_iter_flushcache(void *addr, size_t bytes, struct iov_iter *i);
243 #define _copy_from_iter_flushcache _copy_from_iter_nocache
246 #ifdef CONFIG_ARCH_HAS_COPY_MC
247 size_t _copy_mc_to_iter(const void *addr, size_t bytes, struct iov_iter *i);
248 static inline void iov_iter_set_copy_mc(struct iov_iter *i)
253 static inline bool iov_iter_is_copy_mc(const struct iov_iter *i)
258 #define _copy_mc_to_iter _copy_to_iter
259 static inline void iov_iter_set_copy_mc(struct iov_iter *i) { }
260 static inline bool iov_iter_is_copy_mc(const struct iov_iter *i)
266 size_t iov_iter_zero(size_t bytes, struct iov_iter *);
267 bool iov_iter_is_aligned(const struct iov_iter *i, unsigned addr_mask,
269 unsigned long iov_iter_alignment(const struct iov_iter *i);
270 unsigned long iov_iter_gap_alignment(const struct iov_iter *i);
271 void iov_iter_init(struct iov_iter *i, unsigned int direction, const struct iovec *iov,
272 unsigned long nr_segs, size_t count);
273 void iov_iter_kvec(struct iov_iter *i, unsigned int direction, const struct kvec *kvec,
274 unsigned long nr_segs, size_t count);
275 void iov_iter_bvec(struct iov_iter *i, unsigned int direction, const struct bio_vec *bvec,
276 unsigned long nr_segs, size_t count);
277 void iov_iter_discard(struct iov_iter *i, unsigned int direction, size_t count);
278 void iov_iter_xarray(struct iov_iter *i, unsigned int direction, struct xarray *xarray,
279 loff_t start, size_t count);
280 ssize_t iov_iter_get_pages2(struct iov_iter *i, struct page **pages,
281 size_t maxsize, unsigned maxpages, size_t *start);
282 ssize_t iov_iter_get_pages_alloc2(struct iov_iter *i, struct page ***pages,
283 size_t maxsize, size_t *start);
284 int iov_iter_npages(const struct iov_iter *i, int maxpages);
285 void iov_iter_restore(struct iov_iter *i, struct iov_iter_state *state);
287 const void *dup_iter(struct iov_iter *new, struct iov_iter *old, gfp_t flags);
289 static inline size_t iov_iter_count(const struct iov_iter *i)
295 * Cap the iov_iter by given limit; note that the second argument is
296 * *not* the new size - it's upper limit for such. Passing it a value
297 * greater than the amount of data in iov_iter is fine - it'll just do
298 * nothing in that case.
300 static inline void iov_iter_truncate(struct iov_iter *i, u64 count)
303 * count doesn't have to fit in size_t - comparison extends both
304 * operands to u64 here and any value that would be truncated by
305 * conversion in assignement is by definition greater than all
306 * values of size_t, including old i->count.
308 if (i->count > count)
313 * reexpand a previously truncated iterator; count must be no more than how much
316 static inline void iov_iter_reexpand(struct iov_iter *i, size_t count)
322 iov_iter_npages_cap(struct iov_iter *i, int maxpages, size_t max_bytes)
327 if (iov_iter_count(i) > max_bytes) {
328 shorted = iov_iter_count(i) - max_bytes;
329 iov_iter_truncate(i, max_bytes);
331 npages = iov_iter_npages(i, maxpages);
333 iov_iter_reexpand(i, iov_iter_count(i) + shorted);
343 size_t csum_and_copy_to_iter(const void *addr, size_t bytes, void *csstate, struct iov_iter *i);
344 size_t csum_and_copy_from_iter(void *addr, size_t bytes, __wsum *csum, struct iov_iter *i);
346 static __always_inline __must_check
347 bool csum_and_copy_from_iter_full(void *addr, size_t bytes,
348 __wsum *csum, struct iov_iter *i)
350 size_t copied = csum_and_copy_from_iter(addr, bytes, csum, i);
351 if (likely(copied == bytes))
353 iov_iter_revert(i, copied);
356 size_t hash_and_copy_to_iter(const void *addr, size_t bytes, void *hashp,
359 struct iovec *iovec_from_user(const struct iovec __user *uvector,
360 unsigned long nr_segs, unsigned long fast_segs,
361 struct iovec *fast_iov, bool compat);
362 ssize_t import_iovec(int type, const struct iovec __user *uvec,
363 unsigned nr_segs, unsigned fast_segs, struct iovec **iovp,
365 ssize_t __import_iovec(int type, const struct iovec __user *uvec,
366 unsigned nr_segs, unsigned fast_segs, struct iovec **iovp,
367 struct iov_iter *i, bool compat);
368 int import_single_range(int type, void __user *buf, size_t len,
369 struct iovec *iov, struct iov_iter *i);
370 int import_ubuf(int type, void __user *buf, size_t len, struct iov_iter *i);
372 static inline void iov_iter_ubuf(struct iov_iter *i, unsigned int direction,
373 void __user *buf, size_t count)
375 WARN_ON(direction & ~(READ | WRITE));
376 *i = (struct iov_iter) {
377 .iter_type = ITER_UBUF,
380 .data_source = direction,
386 /* Flags for iov_iter_get/extract_pages*() */
387 /* Allow P2PDMA on the extracted pages */
388 #define ITER_ALLOW_P2PDMA ((__force iov_iter_extraction_t)0x01)
390 ssize_t iov_iter_extract_pages(struct iov_iter *i, struct page ***pages,
391 size_t maxsize, unsigned int maxpages,
392 iov_iter_extraction_t extraction_flags,
396 * iov_iter_extract_will_pin - Indicate how pages from the iterator will be retained
397 * @iter: The iterator
399 * Examine the iterator and indicate by returning true or false as to how, if
400 * at all, pages extracted from the iterator will be retained by the extraction
403 * %true indicates that the pages will have a pin placed in them that the
404 * caller must unpin. This is must be done for DMA/async DIO to force fork()
405 * to forcibly copy a page for the child (the parent must retain the original
408 * %false indicates that no measures are taken and that it's up to the caller
409 * to retain the pages.
411 static inline bool iov_iter_extract_will_pin(const struct iov_iter *iter)
413 return user_backed_iter(iter);
417 ssize_t extract_iter_to_sg(struct iov_iter *iter, size_t len,
418 struct sg_table *sgtable, unsigned int sg_max,
419 iov_iter_extraction_t extraction_flags);