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;
47 * Hack alert: overlay ubuf_iovec with iovec + count, so
48 * that the members resolve correctly regardless of the type
49 * of iterator used. This means that you can use:
51 * &iter->__ubuf_iovec or iter->__iov
53 * interchangably for the user_backed cases, hence simplifying
54 * some of the cases that need to deal with both.
58 * This really should be a const, but we cannot do that without
59 * also modifying any of the zero-filling iter init functions.
60 * Leave it non-const for now, but it should be treated as such.
62 struct iovec __ubuf_iovec;
65 /* use iter_iov() to get the current vec */
66 const struct iovec *__iov;
67 const struct kvec *kvec;
68 const struct bio_vec *bvec;
69 struct xarray *xarray;
76 unsigned long nr_segs;
81 static inline const struct iovec *iter_iov(const struct iov_iter *iter)
83 if (iter->iter_type == ITER_UBUF)
84 return (const struct iovec *) &iter->__ubuf_iovec;
88 #define iter_iov_addr(iter) (iter_iov(iter)->iov_base + (iter)->iov_offset)
89 #define iter_iov_len(iter) (iter_iov(iter)->iov_len - (iter)->iov_offset)
91 static inline enum iter_type iov_iter_type(const struct iov_iter *i)
96 static inline void iov_iter_save_state(struct iov_iter *iter,
97 struct iov_iter_state *state)
99 state->iov_offset = iter->iov_offset;
100 state->count = iter->count;
101 state->nr_segs = iter->nr_segs;
104 static inline bool iter_is_ubuf(const struct iov_iter *i)
106 return iov_iter_type(i) == ITER_UBUF;
109 static inline bool iter_is_iovec(const struct iov_iter *i)
111 return iov_iter_type(i) == ITER_IOVEC;
114 static inline bool iov_iter_is_kvec(const struct iov_iter *i)
116 return iov_iter_type(i) == ITER_KVEC;
119 static inline bool iov_iter_is_bvec(const struct iov_iter *i)
121 return iov_iter_type(i) == ITER_BVEC;
124 static inline bool iov_iter_is_discard(const struct iov_iter *i)
126 return iov_iter_type(i) == ITER_DISCARD;
129 static inline bool iov_iter_is_xarray(const struct iov_iter *i)
131 return iov_iter_type(i) == ITER_XARRAY;
134 static inline unsigned char iov_iter_rw(const struct iov_iter *i)
136 return i->data_source ? WRITE : READ;
139 static inline bool user_backed_iter(const struct iov_iter *i)
141 return iter_is_ubuf(i) || iter_is_iovec(i);
145 * Total number of bytes covered by an iovec.
147 * NOTE that it is not safe to use this function until all the iovec's
148 * segment lengths have been validated. Because the individual lengths can
149 * overflow a size_t when added together.
151 static inline size_t iov_length(const struct iovec *iov, unsigned long nr_segs)
156 for (seg = 0; seg < nr_segs; seg++)
157 ret += iov[seg].iov_len;
161 size_t copy_page_from_iter_atomic(struct page *page, size_t offset,
162 size_t bytes, struct iov_iter *i);
163 void iov_iter_advance(struct iov_iter *i, size_t bytes);
164 void iov_iter_revert(struct iov_iter *i, size_t bytes);
165 size_t fault_in_iov_iter_readable(const struct iov_iter *i, size_t bytes);
166 size_t fault_in_iov_iter_writeable(const struct iov_iter *i, size_t bytes);
167 size_t iov_iter_single_seg_count(const struct iov_iter *i);
168 size_t copy_page_to_iter(struct page *page, size_t offset, size_t bytes,
170 size_t copy_page_from_iter(struct page *page, size_t offset, size_t bytes,
173 size_t _copy_to_iter(const void *addr, size_t bytes, struct iov_iter *i);
174 size_t _copy_from_iter(void *addr, size_t bytes, struct iov_iter *i);
175 size_t _copy_from_iter_nocache(void *addr, size_t bytes, struct iov_iter *i);
177 static inline size_t copy_folio_to_iter(struct folio *folio, size_t offset,
178 size_t bytes, struct iov_iter *i)
180 return copy_page_to_iter(&folio->page, offset, bytes, i);
183 static inline size_t copy_folio_from_iter_atomic(struct folio *folio,
184 size_t offset, size_t bytes, struct iov_iter *i)
186 return copy_page_from_iter_atomic(&folio->page, offset, bytes, i);
189 size_t copy_page_to_iter_nofault(struct page *page, unsigned offset,
190 size_t bytes, struct iov_iter *i);
192 static __always_inline __must_check
193 size_t copy_to_iter(const void *addr, size_t bytes, struct iov_iter *i)
195 if (check_copy_size(addr, bytes, true))
196 return _copy_to_iter(addr, bytes, i);
200 static __always_inline __must_check
201 size_t copy_from_iter(void *addr, size_t bytes, struct iov_iter *i)
203 if (check_copy_size(addr, bytes, false))
204 return _copy_from_iter(addr, bytes, i);
208 static __always_inline __must_check
209 bool copy_from_iter_full(void *addr, size_t bytes, struct iov_iter *i)
211 size_t copied = copy_from_iter(addr, bytes, i);
212 if (likely(copied == bytes))
214 iov_iter_revert(i, copied);
218 static __always_inline __must_check
219 size_t copy_from_iter_nocache(void *addr, size_t bytes, struct iov_iter *i)
221 if (check_copy_size(addr, bytes, false))
222 return _copy_from_iter_nocache(addr, bytes, i);
226 static __always_inline __must_check
227 bool copy_from_iter_full_nocache(void *addr, size_t bytes, struct iov_iter *i)
229 size_t copied = copy_from_iter_nocache(addr, bytes, i);
230 if (likely(copied == bytes))
232 iov_iter_revert(i, copied);
236 #ifdef CONFIG_ARCH_HAS_UACCESS_FLUSHCACHE
238 * Note, users like pmem that depend on the stricter semantics of
239 * _copy_from_iter_flushcache() than _copy_from_iter_nocache() must check for
240 * IS_ENABLED(CONFIG_ARCH_HAS_UACCESS_FLUSHCACHE) before assuming that the
241 * destination is flushed from the cache on return.
243 size_t _copy_from_iter_flushcache(void *addr, size_t bytes, struct iov_iter *i);
245 #define _copy_from_iter_flushcache _copy_from_iter_nocache
248 #ifdef CONFIG_ARCH_HAS_COPY_MC
249 size_t _copy_mc_to_iter(const void *addr, size_t bytes, struct iov_iter *i);
251 #define _copy_mc_to_iter _copy_to_iter
254 size_t iov_iter_zero(size_t bytes, struct iov_iter *);
255 bool iov_iter_is_aligned(const struct iov_iter *i, unsigned addr_mask,
257 unsigned long iov_iter_alignment(const struct iov_iter *i);
258 unsigned long iov_iter_gap_alignment(const struct iov_iter *i);
259 void iov_iter_init(struct iov_iter *i, unsigned int direction, const struct iovec *iov,
260 unsigned long nr_segs, size_t count);
261 void iov_iter_kvec(struct iov_iter *i, unsigned int direction, const struct kvec *kvec,
262 unsigned long nr_segs, size_t count);
263 void iov_iter_bvec(struct iov_iter *i, unsigned int direction, const struct bio_vec *bvec,
264 unsigned long nr_segs, size_t count);
265 void iov_iter_discard(struct iov_iter *i, unsigned int direction, size_t count);
266 void iov_iter_xarray(struct iov_iter *i, unsigned int direction, struct xarray *xarray,
267 loff_t start, size_t count);
268 ssize_t iov_iter_get_pages2(struct iov_iter *i, struct page **pages,
269 size_t maxsize, unsigned maxpages, size_t *start);
270 ssize_t iov_iter_get_pages_alloc2(struct iov_iter *i, struct page ***pages,
271 size_t maxsize, size_t *start);
272 int iov_iter_npages(const struct iov_iter *i, int maxpages);
273 void iov_iter_restore(struct iov_iter *i, struct iov_iter_state *state);
275 const void *dup_iter(struct iov_iter *new, struct iov_iter *old, gfp_t flags);
277 static inline size_t iov_iter_count(const struct iov_iter *i)
283 * Cap the iov_iter by given limit; note that the second argument is
284 * *not* the new size - it's upper limit for such. Passing it a value
285 * greater than the amount of data in iov_iter is fine - it'll just do
286 * nothing in that case.
288 static inline void iov_iter_truncate(struct iov_iter *i, u64 count)
291 * count doesn't have to fit in size_t - comparison extends both
292 * operands to u64 here and any value that would be truncated by
293 * conversion in assignement is by definition greater than all
294 * values of size_t, including old i->count.
296 if (i->count > count)
301 * reexpand a previously truncated iterator; count must be no more than how much
304 static inline void iov_iter_reexpand(struct iov_iter *i, size_t count)
310 iov_iter_npages_cap(struct iov_iter *i, int maxpages, size_t max_bytes)
315 if (iov_iter_count(i) > max_bytes) {
316 shorted = iov_iter_count(i) - max_bytes;
317 iov_iter_truncate(i, max_bytes);
319 npages = iov_iter_npages(i, maxpages);
321 iov_iter_reexpand(i, iov_iter_count(i) + shorted);
326 struct iovec *iovec_from_user(const struct iovec __user *uvector,
327 unsigned long nr_segs, unsigned long fast_segs,
328 struct iovec *fast_iov, bool compat);
329 ssize_t import_iovec(int type, const struct iovec __user *uvec,
330 unsigned nr_segs, unsigned fast_segs, struct iovec **iovp,
332 ssize_t __import_iovec(int type, const struct iovec __user *uvec,
333 unsigned nr_segs, unsigned fast_segs, struct iovec **iovp,
334 struct iov_iter *i, bool compat);
335 int import_ubuf(int type, void __user *buf, size_t len, struct iov_iter *i);
337 static inline void iov_iter_ubuf(struct iov_iter *i, unsigned int direction,
338 void __user *buf, size_t count)
340 WARN_ON(direction & ~(READ | WRITE));
341 *i = (struct iov_iter) {
342 .iter_type = ITER_UBUF,
343 .data_source = direction,
349 /* Flags for iov_iter_get/extract_pages*() */
350 /* Allow P2PDMA on the extracted pages */
351 #define ITER_ALLOW_P2PDMA ((__force iov_iter_extraction_t)0x01)
353 ssize_t iov_iter_extract_pages(struct iov_iter *i, struct page ***pages,
354 size_t maxsize, unsigned int maxpages,
355 iov_iter_extraction_t extraction_flags,
359 * iov_iter_extract_will_pin - Indicate how pages from the iterator will be retained
360 * @iter: The iterator
362 * Examine the iterator and indicate by returning true or false as to how, if
363 * at all, pages extracted from the iterator will be retained by the extraction
366 * %true indicates that the pages will have a pin placed in them that the
367 * caller must unpin. This is must be done for DMA/async DIO to force fork()
368 * to forcibly copy a page for the child (the parent must retain the original
371 * %false indicates that no measures are taken and that it's up to the caller
372 * to retain the pages.
374 static inline bool iov_iter_extract_will_pin(const struct iov_iter *iter)
376 return user_backed_iter(iter);
380 ssize_t extract_iter_to_sg(struct iov_iter *iter, size_t len,
381 struct sg_table *sgtable, unsigned int sg_max,
382 iov_iter_extraction_t extraction_flags);