1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * Berkeley style UIO structures - Alan Cox 1994. 4 */ 5 #ifndef __LINUX_UIO_H 6 #define __LINUX_UIO_H 7 8 #include <linux/kernel.h> 9 #include <linux/thread_info.h> 10 #include <linux/mm_types.h> 11 #include <uapi/linux/uio.h> 12 13 struct page; 14 struct pipe_inode_info; 15 16 typedef unsigned int __bitwise iov_iter_extraction_t; 17 18 struct kvec { 19 void *iov_base; /* and that should *never* hold a userland pointer */ 20 size_t iov_len; 21 }; 22 23 enum iter_type { 24 /* iter types */ 25 ITER_IOVEC, 26 ITER_KVEC, 27 ITER_BVEC, 28 ITER_PIPE, 29 ITER_XARRAY, 30 ITER_DISCARD, 31 ITER_UBUF, 32 }; 33 34 #define ITER_SOURCE 1 // == WRITE 35 #define ITER_DEST 0 // == READ 36 37 struct iov_iter_state { 38 size_t iov_offset; 39 size_t count; 40 unsigned long nr_segs; 41 }; 42 43 struct iov_iter { 44 u8 iter_type; 45 bool nofault; 46 bool data_source; 47 bool user_backed; 48 union { 49 size_t iov_offset; 50 int last_offset; 51 }; 52 size_t count; 53 union { 54 const struct iovec *iov; 55 const struct kvec *kvec; 56 const struct bio_vec *bvec; 57 struct xarray *xarray; 58 struct pipe_inode_info *pipe; 59 void __user *ubuf; 60 }; 61 union { 62 unsigned long nr_segs; 63 struct { 64 unsigned int head; 65 unsigned int start_head; 66 }; 67 loff_t xarray_start; 68 }; 69 }; 70 71 static inline enum iter_type iov_iter_type(const struct iov_iter *i) 72 { 73 return i->iter_type; 74 } 75 76 static inline void iov_iter_save_state(struct iov_iter *iter, 77 struct iov_iter_state *state) 78 { 79 state->iov_offset = iter->iov_offset; 80 state->count = iter->count; 81 state->nr_segs = iter->nr_segs; 82 } 83 84 static inline bool iter_is_ubuf(const struct iov_iter *i) 85 { 86 return iov_iter_type(i) == ITER_UBUF; 87 } 88 89 static inline bool iter_is_iovec(const struct iov_iter *i) 90 { 91 return iov_iter_type(i) == ITER_IOVEC; 92 } 93 94 static inline bool iov_iter_is_kvec(const struct iov_iter *i) 95 { 96 return iov_iter_type(i) == ITER_KVEC; 97 } 98 99 static inline bool iov_iter_is_bvec(const struct iov_iter *i) 100 { 101 return iov_iter_type(i) == ITER_BVEC; 102 } 103 104 static inline bool iov_iter_is_pipe(const struct iov_iter *i) 105 { 106 return iov_iter_type(i) == ITER_PIPE; 107 } 108 109 static inline bool iov_iter_is_discard(const struct iov_iter *i) 110 { 111 return iov_iter_type(i) == ITER_DISCARD; 112 } 113 114 static inline bool iov_iter_is_xarray(const struct iov_iter *i) 115 { 116 return iov_iter_type(i) == ITER_XARRAY; 117 } 118 119 static inline unsigned char iov_iter_rw(const struct iov_iter *i) 120 { 121 return i->data_source ? WRITE : READ; 122 } 123 124 static inline bool user_backed_iter(const struct iov_iter *i) 125 { 126 return i->user_backed; 127 } 128 129 /* 130 * Total number of bytes covered by an iovec. 131 * 132 * NOTE that it is not safe to use this function until all the iovec's 133 * segment lengths have been validated. Because the individual lengths can 134 * overflow a size_t when added together. 135 */ 136 static inline size_t iov_length(const struct iovec *iov, unsigned long nr_segs) 137 { 138 unsigned long seg; 139 size_t ret = 0; 140 141 for (seg = 0; seg < nr_segs; seg++) 142 ret += iov[seg].iov_len; 143 return ret; 144 } 145 146 static inline struct iovec iov_iter_iovec(const struct iov_iter *iter) 147 { 148 return (struct iovec) { 149 .iov_base = iter->iov->iov_base + iter->iov_offset, 150 .iov_len = min(iter->count, 151 iter->iov->iov_len - iter->iov_offset), 152 }; 153 } 154 155 size_t copy_page_from_iter_atomic(struct page *page, unsigned offset, 156 size_t bytes, struct iov_iter *i); 157 void iov_iter_advance(struct iov_iter *i, size_t bytes); 158 void iov_iter_revert(struct iov_iter *i, size_t bytes); 159 size_t fault_in_iov_iter_readable(const struct iov_iter *i, size_t bytes); 160 size_t fault_in_iov_iter_writeable(const struct iov_iter *i, size_t bytes); 161 size_t iov_iter_single_seg_count(const struct iov_iter *i); 162 size_t copy_page_to_iter(struct page *page, size_t offset, size_t bytes, 163 struct iov_iter *i); 164 size_t copy_page_from_iter(struct page *page, size_t offset, size_t bytes, 165 struct iov_iter *i); 166 167 size_t _copy_to_iter(const void *addr, size_t bytes, struct iov_iter *i); 168 size_t _copy_from_iter(void *addr, size_t bytes, struct iov_iter *i); 169 size_t _copy_from_iter_nocache(void *addr, size_t bytes, struct iov_iter *i); 170 171 static inline size_t copy_folio_to_iter(struct folio *folio, size_t offset, 172 size_t bytes, struct iov_iter *i) 173 { 174 return copy_page_to_iter(&folio->page, offset, bytes, i); 175 } 176 177 static __always_inline __must_check 178 size_t copy_to_iter(const void *addr, size_t bytes, struct iov_iter *i) 179 { 180 if (check_copy_size(addr, bytes, true)) 181 return _copy_to_iter(addr, bytes, i); 182 return 0; 183 } 184 185 static __always_inline __must_check 186 size_t copy_from_iter(void *addr, size_t bytes, struct iov_iter *i) 187 { 188 if (check_copy_size(addr, bytes, false)) 189 return _copy_from_iter(addr, bytes, i); 190 return 0; 191 } 192 193 static __always_inline __must_check 194 bool copy_from_iter_full(void *addr, size_t bytes, struct iov_iter *i) 195 { 196 size_t copied = copy_from_iter(addr, bytes, i); 197 if (likely(copied == bytes)) 198 return true; 199 iov_iter_revert(i, copied); 200 return false; 201 } 202 203 static __always_inline __must_check 204 size_t copy_from_iter_nocache(void *addr, size_t bytes, struct iov_iter *i) 205 { 206 if (check_copy_size(addr, bytes, false)) 207 return _copy_from_iter_nocache(addr, bytes, i); 208 return 0; 209 } 210 211 static __always_inline __must_check 212 bool copy_from_iter_full_nocache(void *addr, size_t bytes, struct iov_iter *i) 213 { 214 size_t copied = copy_from_iter_nocache(addr, bytes, i); 215 if (likely(copied == bytes)) 216 return true; 217 iov_iter_revert(i, copied); 218 return false; 219 } 220 221 #ifdef CONFIG_ARCH_HAS_UACCESS_FLUSHCACHE 222 /* 223 * Note, users like pmem that depend on the stricter semantics of 224 * _copy_from_iter_flushcache() than _copy_from_iter_nocache() must check for 225 * IS_ENABLED(CONFIG_ARCH_HAS_UACCESS_FLUSHCACHE) before assuming that the 226 * destination is flushed from the cache on return. 227 */ 228 size_t _copy_from_iter_flushcache(void *addr, size_t bytes, struct iov_iter *i); 229 #else 230 #define _copy_from_iter_flushcache _copy_from_iter_nocache 231 #endif 232 233 #ifdef CONFIG_ARCH_HAS_COPY_MC 234 size_t _copy_mc_to_iter(const void *addr, size_t bytes, struct iov_iter *i); 235 #else 236 #define _copy_mc_to_iter _copy_to_iter 237 #endif 238 239 size_t iov_iter_zero(size_t bytes, struct iov_iter *); 240 bool iov_iter_is_aligned(const struct iov_iter *i, unsigned addr_mask, 241 unsigned len_mask); 242 unsigned long iov_iter_alignment(const struct iov_iter *i); 243 unsigned long iov_iter_gap_alignment(const struct iov_iter *i); 244 void iov_iter_init(struct iov_iter *i, unsigned int direction, const struct iovec *iov, 245 unsigned long nr_segs, size_t count); 246 void iov_iter_kvec(struct iov_iter *i, unsigned int direction, const struct kvec *kvec, 247 unsigned long nr_segs, size_t count); 248 void iov_iter_bvec(struct iov_iter *i, unsigned int direction, const struct bio_vec *bvec, 249 unsigned long nr_segs, size_t count); 250 void iov_iter_pipe(struct iov_iter *i, unsigned int direction, struct pipe_inode_info *pipe, 251 size_t count); 252 void iov_iter_discard(struct iov_iter *i, unsigned int direction, size_t count); 253 void iov_iter_xarray(struct iov_iter *i, unsigned int direction, struct xarray *xarray, 254 loff_t start, size_t count); 255 ssize_t iov_iter_get_pages(struct iov_iter *i, struct page **pages, 256 size_t maxsize, unsigned maxpages, size_t *start, 257 iov_iter_extraction_t extraction_flags); 258 ssize_t iov_iter_get_pages2(struct iov_iter *i, struct page **pages, 259 size_t maxsize, unsigned maxpages, size_t *start); 260 ssize_t iov_iter_get_pages_alloc(struct iov_iter *i, 261 struct page ***pages, size_t maxsize, size_t *start, 262 iov_iter_extraction_t extraction_flags); 263 ssize_t iov_iter_get_pages_alloc2(struct iov_iter *i, struct page ***pages, 264 size_t maxsize, size_t *start); 265 int iov_iter_npages(const struct iov_iter *i, int maxpages); 266 void iov_iter_restore(struct iov_iter *i, struct iov_iter_state *state); 267 268 const void *dup_iter(struct iov_iter *new, struct iov_iter *old, gfp_t flags); 269 270 static inline size_t iov_iter_count(const struct iov_iter *i) 271 { 272 return i->count; 273 } 274 275 /* 276 * Cap the iov_iter by given limit; note that the second argument is 277 * *not* the new size - it's upper limit for such. Passing it a value 278 * greater than the amount of data in iov_iter is fine - it'll just do 279 * nothing in that case. 280 */ 281 static inline void iov_iter_truncate(struct iov_iter *i, u64 count) 282 { 283 /* 284 * count doesn't have to fit in size_t - comparison extends both 285 * operands to u64 here and any value that would be truncated by 286 * conversion in assignement is by definition greater than all 287 * values of size_t, including old i->count. 288 */ 289 if (i->count > count) 290 i->count = count; 291 } 292 293 /* 294 * reexpand a previously truncated iterator; count must be no more than how much 295 * we had shrunk it. 296 */ 297 static inline void iov_iter_reexpand(struct iov_iter *i, size_t count) 298 { 299 i->count = count; 300 } 301 302 static inline int 303 iov_iter_npages_cap(struct iov_iter *i, int maxpages, size_t max_bytes) 304 { 305 size_t shorted = 0; 306 int npages; 307 308 if (iov_iter_count(i) > max_bytes) { 309 shorted = iov_iter_count(i) - max_bytes; 310 iov_iter_truncate(i, max_bytes); 311 } 312 npages = iov_iter_npages(i, maxpages); 313 if (shorted) 314 iov_iter_reexpand(i, iov_iter_count(i) + shorted); 315 316 return npages; 317 } 318 319 struct csum_state { 320 __wsum csum; 321 size_t off; 322 }; 323 324 size_t csum_and_copy_to_iter(const void *addr, size_t bytes, void *csstate, struct iov_iter *i); 325 size_t csum_and_copy_from_iter(void *addr, size_t bytes, __wsum *csum, struct iov_iter *i); 326 327 static __always_inline __must_check 328 bool csum_and_copy_from_iter_full(void *addr, size_t bytes, 329 __wsum *csum, struct iov_iter *i) 330 { 331 size_t copied = csum_and_copy_from_iter(addr, bytes, csum, i); 332 if (likely(copied == bytes)) 333 return true; 334 iov_iter_revert(i, copied); 335 return false; 336 } 337 size_t hash_and_copy_to_iter(const void *addr, size_t bytes, void *hashp, 338 struct iov_iter *i); 339 340 struct iovec *iovec_from_user(const struct iovec __user *uvector, 341 unsigned long nr_segs, unsigned long fast_segs, 342 struct iovec *fast_iov, bool compat); 343 ssize_t import_iovec(int type, const struct iovec __user *uvec, 344 unsigned nr_segs, unsigned fast_segs, struct iovec **iovp, 345 struct iov_iter *i); 346 ssize_t __import_iovec(int type, const struct iovec __user *uvec, 347 unsigned nr_segs, unsigned fast_segs, struct iovec **iovp, 348 struct iov_iter *i, bool compat); 349 int import_single_range(int type, void __user *buf, size_t len, 350 struct iovec *iov, struct iov_iter *i); 351 int import_ubuf(int type, void __user *buf, size_t len, struct iov_iter *i); 352 353 static inline void iov_iter_ubuf(struct iov_iter *i, unsigned int direction, 354 void __user *buf, size_t count) 355 { 356 WARN_ON(direction & ~(READ | WRITE)); 357 *i = (struct iov_iter) { 358 .iter_type = ITER_UBUF, 359 .user_backed = true, 360 .data_source = direction, 361 .ubuf = buf, 362 .count = count 363 }; 364 } 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) 368 369 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, 372 size_t *offset0); 373 374 /** 375 * iov_iter_extract_will_pin - Indicate how pages from the iterator will be retained 376 * @iter: The iterator 377 * 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 380 * function. 381 * 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 385 * page). 386 * 387 * %false indicates that no measures are taken and that it's up to the caller 388 * to retain the pages. 389 */ 390 static inline bool iov_iter_extract_will_pin(const struct iov_iter *iter) 391 { 392 return user_backed_iter(iter); 393 } 394 395 #endif 396