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 13struct page; 14 15typedef unsigned int __bitwise iov_iter_extraction_t; 16 17struct kvec { 18 void *iov_base; /* and that should *never* hold a userland pointer */ 19 size_t iov_len; 20}; 21 22enum iter_type { 23 /* iter types */ 24 ITER_UBUF, 25 ITER_IOVEC, 26 ITER_BVEC, 27 ITER_KVEC, 28 ITER_XARRAY, 29 ITER_DISCARD, 30}; 31 32#define ITER_SOURCE 1 // == WRITE 33#define ITER_DEST 0 // == READ 34 35struct iov_iter_state { 36 size_t iov_offset; 37 size_t count; 38 unsigned long nr_segs; 39}; 40 41struct iov_iter { 42 u8 iter_type; 43 bool nofault; 44 bool data_source; 45 size_t iov_offset; 46 /* 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: 50 * 51 * &iter->__ubuf_iovec or iter->__iov 52 * 53 * interchangably for the user_backed cases, hence simplifying 54 * some of the cases that need to deal with both. 55 */ 56 union { 57 /* 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. 61 */ 62 struct iovec __ubuf_iovec; 63 struct { 64 union { 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; 70 void __user *ubuf; 71 }; 72 size_t count; 73 }; 74 }; 75 union { 76 unsigned long nr_segs; 77 loff_t xarray_start; 78 }; 79}; 80 81static inline const struct iovec *iter_iov(const struct iov_iter *iter) 82{ 83 if (iter->iter_type == ITER_UBUF) 84 return (const struct iovec *) &iter->__ubuf_iovec; 85 return iter->__iov; 86} 87 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) 90 91static inline enum iter_type iov_iter_type(const struct iov_iter *i) 92{ 93 return i->iter_type; 94} 95 96static inline void iov_iter_save_state(struct iov_iter *iter, 97 struct iov_iter_state *state) 98{ 99 state->iov_offset = iter->iov_offset; 100 state->count = iter->count; 101 state->nr_segs = iter->nr_segs; 102} 103 104static inline bool iter_is_ubuf(const struct iov_iter *i) 105{ 106 return iov_iter_type(i) == ITER_UBUF; 107} 108 109static inline bool iter_is_iovec(const struct iov_iter *i) 110{ 111 return iov_iter_type(i) == ITER_IOVEC; 112} 113 114static inline bool iov_iter_is_kvec(const struct iov_iter *i) 115{ 116 return iov_iter_type(i) == ITER_KVEC; 117} 118 119static inline bool iov_iter_is_bvec(const struct iov_iter *i) 120{ 121 return iov_iter_type(i) == ITER_BVEC; 122} 123 124static inline bool iov_iter_is_discard(const struct iov_iter *i) 125{ 126 return iov_iter_type(i) == ITER_DISCARD; 127} 128 129static inline bool iov_iter_is_xarray(const struct iov_iter *i) 130{ 131 return iov_iter_type(i) == ITER_XARRAY; 132} 133 134static inline unsigned char iov_iter_rw(const struct iov_iter *i) 135{ 136 return i->data_source ? WRITE : READ; 137} 138 139static inline bool user_backed_iter(const struct iov_iter *i) 140{ 141 return iter_is_ubuf(i) || iter_is_iovec(i); 142} 143 144/* 145 * Total number of bytes covered by an iovec. 146 * 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. 150 */ 151static inline size_t iov_length(const struct iovec *iov, unsigned long nr_segs) 152{ 153 unsigned long seg; 154 size_t ret = 0; 155 156 for (seg = 0; seg < nr_segs; seg++) 157 ret += iov[seg].iov_len; 158 return ret; 159} 160 161size_t copy_page_from_iter_atomic(struct page *page, size_t offset, 162 size_t bytes, struct iov_iter *i); 163void iov_iter_advance(struct iov_iter *i, size_t bytes); 164void iov_iter_revert(struct iov_iter *i, size_t bytes); 165size_t fault_in_iov_iter_readable(const struct iov_iter *i, size_t bytes); 166size_t fault_in_iov_iter_writeable(const struct iov_iter *i, size_t bytes); 167size_t iov_iter_single_seg_count(const struct iov_iter *i); 168size_t copy_page_to_iter(struct page *page, size_t offset, size_t bytes, 169 struct iov_iter *i); 170size_t copy_page_from_iter(struct page *page, size_t offset, size_t bytes, 171 struct iov_iter *i); 172 173size_t _copy_to_iter(const void *addr, size_t bytes, struct iov_iter *i); 174size_t _copy_from_iter(void *addr, size_t bytes, struct iov_iter *i); 175size_t _copy_from_iter_nocache(void *addr, size_t bytes, struct iov_iter *i); 176 177static inline size_t copy_folio_to_iter(struct folio *folio, size_t offset, 178 size_t bytes, struct iov_iter *i) 179{ 180 return copy_page_to_iter(&folio->page, offset, bytes, i); 181} 182 183static inline size_t copy_folio_from_iter_atomic(struct folio *folio, 184 size_t offset, size_t bytes, struct iov_iter *i) 185{ 186 return copy_page_from_iter_atomic(&folio->page, offset, bytes, i); 187} 188 189size_t copy_page_to_iter_nofault(struct page *page, unsigned offset, 190 size_t bytes, struct iov_iter *i); 191 192static __always_inline __must_check 193size_t copy_to_iter(const void *addr, size_t bytes, struct iov_iter *i) 194{ 195 if (check_copy_size(addr, bytes, true)) 196 return _copy_to_iter(addr, bytes, i); 197 return 0; 198} 199 200static __always_inline __must_check 201size_t copy_from_iter(void *addr, size_t bytes, struct iov_iter *i) 202{ 203 if (check_copy_size(addr, bytes, false)) 204 return _copy_from_iter(addr, bytes, i); 205 return 0; 206} 207 208static __always_inline __must_check 209bool copy_from_iter_full(void *addr, size_t bytes, struct iov_iter *i) 210{ 211 size_t copied = copy_from_iter(addr, bytes, i); 212 if (likely(copied == bytes)) 213 return true; 214 iov_iter_revert(i, copied); 215 return false; 216} 217 218static __always_inline __must_check 219size_t copy_from_iter_nocache(void *addr, size_t bytes, struct iov_iter *i) 220{ 221 if (check_copy_size(addr, bytes, false)) 222 return _copy_from_iter_nocache(addr, bytes, i); 223 return 0; 224} 225 226static __always_inline __must_check 227bool copy_from_iter_full_nocache(void *addr, size_t bytes, struct iov_iter *i) 228{ 229 size_t copied = copy_from_iter_nocache(addr, bytes, i); 230 if (likely(copied == bytes)) 231 return true; 232 iov_iter_revert(i, copied); 233 return false; 234} 235 236#ifdef CONFIG_ARCH_HAS_UACCESS_FLUSHCACHE 237/* 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. 242 */ 243size_t _copy_from_iter_flushcache(void *addr, size_t bytes, struct iov_iter *i); 244#else 245#define _copy_from_iter_flushcache _copy_from_iter_nocache 246#endif 247 248#ifdef CONFIG_ARCH_HAS_COPY_MC 249size_t _copy_mc_to_iter(const void *addr, size_t bytes, struct iov_iter *i); 250#else 251#define _copy_mc_to_iter _copy_to_iter 252#endif 253 254size_t iov_iter_zero(size_t bytes, struct iov_iter *); 255bool iov_iter_is_aligned(const struct iov_iter *i, unsigned addr_mask, 256 unsigned len_mask); 257unsigned long iov_iter_alignment(const struct iov_iter *i); 258unsigned long iov_iter_gap_alignment(const struct iov_iter *i); 259void iov_iter_init(struct iov_iter *i, unsigned int direction, const struct iovec *iov, 260 unsigned long nr_segs, size_t count); 261void iov_iter_kvec(struct iov_iter *i, unsigned int direction, const struct kvec *kvec, 262 unsigned long nr_segs, size_t count); 263void iov_iter_bvec(struct iov_iter *i, unsigned int direction, const struct bio_vec *bvec, 264 unsigned long nr_segs, size_t count); 265void iov_iter_discard(struct iov_iter *i, unsigned int direction, size_t count); 266void iov_iter_xarray(struct iov_iter *i, unsigned int direction, struct xarray *xarray, 267 loff_t start, size_t count); 268ssize_t iov_iter_get_pages2(struct iov_iter *i, struct page **pages, 269 size_t maxsize, unsigned maxpages, size_t *start); 270ssize_t iov_iter_get_pages_alloc2(struct iov_iter *i, struct page ***pages, 271 size_t maxsize, size_t *start); 272int iov_iter_npages(const struct iov_iter *i, int maxpages); 273void iov_iter_restore(struct iov_iter *i, struct iov_iter_state *state); 274 275const void *dup_iter(struct iov_iter *new, struct iov_iter *old, gfp_t flags); 276 277static inline size_t iov_iter_count(const struct iov_iter *i) 278{ 279 return i->count; 280} 281 282/* 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. 287 */ 288static inline void iov_iter_truncate(struct iov_iter *i, u64 count) 289{ 290 /* 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. 295 */ 296 if (i->count > count) 297 i->count = count; 298} 299 300/* 301 * reexpand a previously truncated iterator; count must be no more than how much 302 * we had shrunk it. 303 */ 304static inline void iov_iter_reexpand(struct iov_iter *i, size_t count) 305{ 306 i->count = count; 307} 308 309static inline int 310iov_iter_npages_cap(struct iov_iter *i, int maxpages, size_t max_bytes) 311{ 312 size_t shorted = 0; 313 int npages; 314 315 if (iov_iter_count(i) > max_bytes) { 316 shorted = iov_iter_count(i) - max_bytes; 317 iov_iter_truncate(i, max_bytes); 318 } 319 npages = iov_iter_npages(i, maxpages); 320 if (shorted) 321 iov_iter_reexpand(i, iov_iter_count(i) + shorted); 322 323 return npages; 324} 325 326struct 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); 329ssize_t import_iovec(int type, const struct iovec __user *uvec, 330 unsigned nr_segs, unsigned fast_segs, struct iovec **iovp, 331 struct iov_iter *i); 332ssize_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); 335int import_ubuf(int type, void __user *buf, size_t len, struct iov_iter *i); 336 337static inline void iov_iter_ubuf(struct iov_iter *i, unsigned int direction, 338 void __user *buf, size_t count) 339{ 340 WARN_ON(direction & ~(READ | WRITE)); 341 *i = (struct iov_iter) { 342 .iter_type = ITER_UBUF, 343 .data_source = direction, 344 .ubuf = buf, 345 .count = count, 346 .nr_segs = 1 347 }; 348} 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) 352 353ssize_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, 356 size_t *offset0); 357 358/** 359 * iov_iter_extract_will_pin - Indicate how pages from the iterator will be retained 360 * @iter: The iterator 361 * 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 364 * function. 365 * 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 369 * page). 370 * 371 * %false indicates that no measures are taken and that it's up to the caller 372 * to retain the pages. 373 */ 374static inline bool iov_iter_extract_will_pin(const struct iov_iter *iter) 375{ 376 return user_backed_iter(iter); 377} 378 379struct sg_table; 380ssize_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); 383 384#endif 385