include/linux/uio.h

   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 folio_queue;
  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_UBUF,
  26         ITER_IOVEC,
  27         ITER_BVEC,
  28         ITER_KVEC,
  29         ITER_FOLIOQ,
  30         ITER_XARRAY,
  31         ITER_DISCARD,
  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         size_t iov_offset;
  48         /*
  49          * Hack alert: overlay ubuf_iovec with iovec + count, so
  50          * that the members resolve correctly regardless of the type
  51          * of iterator used. This means that you can use:
  52          *
  53          * &iter->__ubuf_iovec or iter->__iov
  54          *
  55          * interchangably for the user_backed cases, hence simplifying
  56          * some of the cases that need to deal with both.
  57          */
  58         union {
  59                 /*
  60                  * This really should be a const, but we cannot do that without
  61                  * also modifying any of the zero-filling iter init functions.
  62                  * Leave it non-const for now, but it should be treated as such.
  63                  */
  64                 struct iovec __ubuf_iovec;
  65                 struct {
  66                         union {
  67                                 /* use iter_iov() to get the current vec */
  68                                 const struct iovec *__iov;
  69                                 const struct kvec *kvec;
  70                                 const struct bio_vec *bvec;
  71                                 const struct folio_queue *folioq;
  72                                 struct xarray *xarray;
  73                                 void __user *ubuf;
  74                         };
  75                         size_t count;
  76                 };
  77         };
  78         union {
  79                 unsigned long nr_segs;
  80                 u8 folioq_slot;
  81                 loff_t xarray_start;
  82         };
  83 };
  84
  85 static inline const struct iovec *iter_iov(const struct iov_iter *iter)
  86 {
  87         if (iter->iter_type == ITER_UBUF)
  88                 return (const struct iovec *) &iter->__ubuf_iovec;
  89         return iter->__iov;
  90 }
  91
  92 #define iter_iov_addr(iter)     (iter_iov(iter)->iov_base + (iter)->iov_offset)
  93 #define iter_iov_len(iter)      (iter_iov(iter)->iov_len - (iter)->iov_offset)
  94
  95 static inline enum iter_type iov_iter_type(const struct iov_iter *i)
  96 {
  97         return i->iter_type;
  98 }
  99
 100 static inline void iov_iter_save_state(struct iov_iter *iter,
 101                                        struct iov_iter_state *state)
 102 {
 103         state->iov_offset = iter->iov_offset;
 104         state->count = iter->count;
 105         state->nr_segs = iter->nr_segs;
 106 }
 107
 108 static inline bool iter_is_ubuf(const struct iov_iter *i)
 109 {
 110         return iov_iter_type(i) == ITER_UBUF;
 111 }
 112
 113 static inline bool iter_is_iovec(const struct iov_iter *i)
 114 {
 115         return iov_iter_type(i) == ITER_IOVEC;
 116 }
 117
 118 static inline bool iov_iter_is_kvec(const struct iov_iter *i)
 119 {
 120         return iov_iter_type(i) == ITER_KVEC;
 121 }
 122
 123 static inline bool iov_iter_is_bvec(const struct iov_iter *i)
 124 {
 125         return iov_iter_type(i) == ITER_BVEC;
 126 }
 127
 128 static inline bool iov_iter_is_discard(const struct iov_iter *i)
 129 {
 130         return iov_iter_type(i) == ITER_DISCARD;
 131 }
 132
 133 static inline bool iov_iter_is_folioq(const struct iov_iter *i)
 134 {
 135         return iov_iter_type(i) == ITER_FOLIOQ;
 136 }
 137
 138 static inline bool iov_iter_is_xarray(const struct iov_iter *i)
 139 {
 140         return iov_iter_type(i) == ITER_XARRAY;
 141 }
 142
 143 static inline unsigned char iov_iter_rw(const struct iov_iter *i)
 144 {
 145         return i->data_source ? WRITE : READ;
 146 }
 147
 148 static inline bool user_backed_iter(const struct iov_iter *i)
 149 {
 150         return iter_is_ubuf(i) || iter_is_iovec(i);
 151 }
 152
 153 /*
 154  * Total number of bytes covered by an iovec.
 155  *
 156  * NOTE that it is not safe to use this function until all the iovec's
 157  * segment lengths have been validated.  Because the individual lengths can
 158  * overflow a size_t when added together.
 159  */
 160 static inline size_t iov_length(const struct iovec *iov, unsigned long nr_segs)
 161 {
 162         unsigned long seg;
 163         size_t ret = 0;
 164
 165         for (seg = 0; seg < nr_segs; seg++)
 166                 ret += iov[seg].iov_len;
 167         return ret;
 168 }
 169
 170 size_t copy_page_from_iter_atomic(struct page *page, size_t offset,
 171                                   size_t bytes, struct iov_iter *i);
 172 void iov_iter_advance(struct iov_iter *i, size_t bytes);
 173 void iov_iter_revert(struct iov_iter *i, size_t bytes);
 174 size_t fault_in_iov_iter_readable(const struct iov_iter *i, size_t bytes);
 175 size_t fault_in_iov_iter_writeable(const struct iov_iter *i, size_t bytes);
 176 size_t iov_iter_single_seg_count(const struct iov_iter *i);
 177 size_t copy_page_to_iter(struct page *page, size_t offset, size_t bytes,
 178                          struct iov_iter *i);
 179 size_t copy_page_from_iter(struct page *page, size_t offset, size_t bytes,
 180                          struct iov_iter *i);
 181
 182 size_t _copy_to_iter(const void *addr, size_t bytes, struct iov_iter *i);
 183 size_t _copy_from_iter(void *addr, size_t bytes, struct iov_iter *i);
 184 size_t _copy_from_iter_nocache(void *addr, size_t bytes, struct iov_iter *i);
 185
 186 static inline size_t copy_folio_to_iter(struct folio *folio, size_t offset,
 187                 size_t bytes, struct iov_iter *i)
 188 {
 189         return copy_page_to_iter(&folio->page, offset, bytes, i);
 190 }
 191
 192 static inline size_t copy_folio_from_iter(struct folio *folio, size_t offset,
 193                                           size_t bytes, struct iov_iter *i)
 194 {
 195         return copy_page_from_iter(&folio->page, offset, bytes, i);
 196 }
 197
 198 static inline size_t copy_folio_from_iter_atomic(struct folio *folio,
 199                 size_t offset, size_t bytes, struct iov_iter *i)
 200 {
 201         return copy_page_from_iter_atomic(&folio->page, offset, bytes, i);
 202 }
 203
 204 size_t copy_page_to_iter_nofault(struct page *page, unsigned offset,
 205                                  size_t bytes, struct iov_iter *i);
 206
 207 static __always_inline __must_check
 208 size_t copy_to_iter(const void *addr, size_t bytes, struct iov_iter *i)
 209 {
 210         if (check_copy_size(addr, bytes, true))
 211                 return _copy_to_iter(addr, bytes, i);
 212         return 0;
 213 }
 214
 215 static __always_inline __must_check
 216 size_t copy_from_iter(void *addr, size_t bytes, struct iov_iter *i)
 217 {
 218         if (check_copy_size(addr, bytes, false))
 219                 return _copy_from_iter(addr, bytes, i);
 220         return 0;
 221 }
 222
 223 static __always_inline __must_check
 224 bool copy_to_iter_full(const void *addr, size_t bytes, struct iov_iter *i)
 225 {
 226         size_t copied = copy_to_iter(addr, bytes, i);
 227         if (likely(copied == bytes))
 228                 return true;
 229         iov_iter_revert(i, copied);
 230         return false;
 231 }
 232
 233 static __always_inline __must_check
 234 bool copy_from_iter_full(void *addr, size_t bytes, struct iov_iter *i)
 235 {
 236         size_t copied = copy_from_iter(addr, bytes, i);
 237         if (likely(copied == bytes))
 238                 return true;
 239         iov_iter_revert(i, copied);
 240         return false;
 241 }
 242
 243 static __always_inline __must_check
 244 size_t copy_from_iter_nocache(void *addr, size_t bytes, struct iov_iter *i)
 245 {
 246         if (check_copy_size(addr, bytes, false))
 247                 return _copy_from_iter_nocache(addr, bytes, i);
 248         return 0;
 249 }
 250
 251 static __always_inline __must_check
 252 bool copy_from_iter_full_nocache(void *addr, size_t bytes, struct iov_iter *i)
 253 {
 254         size_t copied = copy_from_iter_nocache(addr, bytes, i);
 255         if (likely(copied == bytes))
 256                 return true;
 257         iov_iter_revert(i, copied);
 258         return false;
 259 }
 260
 261 #ifdef CONFIG_ARCH_HAS_UACCESS_FLUSHCACHE
 262 /*
 263  * Note, users like pmem that depend on the stricter semantics of
 264  * _copy_from_iter_flushcache() than _copy_from_iter_nocache() must check for
 265  * IS_ENABLED(CONFIG_ARCH_HAS_UACCESS_FLUSHCACHE) before assuming that the
 266  * destination is flushed from the cache on return.
 267  */
 268 size_t _copy_from_iter_flushcache(void *addr, size_t bytes, struct iov_iter *i);
 269 #else
 270 #define _copy_from_iter_flushcache _copy_from_iter_nocache
 271 #endif
 272
 273 #ifdef CONFIG_ARCH_HAS_COPY_MC
 274 size_t _copy_mc_to_iter(const void *addr, size_t bytes, struct iov_iter *i);
 275 #else
 276 #define _copy_mc_to_iter _copy_to_iter
 277 #endif
 278
 279 size_t iov_iter_zero(size_t bytes, struct iov_iter *);
 280 bool iov_iter_is_aligned(const struct iov_iter *i, unsigned addr_mask,
 281                         unsigned len_mask);
 282 unsigned long iov_iter_alignment(const struct iov_iter *i);
 283 unsigned long iov_iter_gap_alignment(const struct iov_iter *i);
 284 void iov_iter_init(struct iov_iter *i, unsigned int direction, const struct iovec *iov,
 285                         unsigned long nr_segs, size_t count);
 286 void iov_iter_kvec(struct iov_iter *i, unsigned int direction, const struct kvec *kvec,
 287                         unsigned long nr_segs, size_t count);
 288 void iov_iter_bvec(struct iov_iter *i, unsigned int direction, const struct bio_vec *bvec,
 289                         unsigned long nr_segs, size_t count);
 290 void iov_iter_discard(struct iov_iter *i, unsigned int direction, size_t count);
 291 void iov_iter_folio_queue(struct iov_iter *i, unsigned int direction,
 292                           const struct folio_queue *folioq,
 293                           unsigned int first_slot, unsigned int offset, size_t count);
 294 void iov_iter_xarray(struct iov_iter *i, unsigned int direction, struct xarray *xarray,
 295                      loff_t start, size_t count);
 296 ssize_t iov_iter_get_pages2(struct iov_iter *i, struct page **pages,
 297                         size_t maxsize, unsigned maxpages, size_t *start);
 298 ssize_t iov_iter_get_pages_alloc2(struct iov_iter *i, struct page ***pages,
 299                         size_t maxsize, size_t *start);
 300 int iov_iter_npages(const struct iov_iter *i, int maxpages);
 301 void iov_iter_restore(struct iov_iter *i, struct iov_iter_state *state);
 302
 303 const void *dup_iter(struct iov_iter *new, struct iov_iter *old, gfp_t flags);
 304
 305 static inline size_t iov_iter_count(const struct iov_iter *i)
 306 {
 307         return i->count;
 308 }
 309
 310 /*
 311  * Cap the iov_iter by given limit; note that the second argument is
 312  * *not* the new size - it's upper limit for such.  Passing it a value
 313  * greater than the amount of data in iov_iter is fine - it'll just do
 314  * nothing in that case.
 315  */
 316 static inline void iov_iter_truncate(struct iov_iter *i, u64 count)
 317 {
 318         /*
 319          * count doesn't have to fit in size_t - comparison extends both
 320          * operands to u64 here and any value that would be truncated by
 321          * conversion in assignement is by definition greater than all
 322          * values of size_t, including old i->count.
 323          */
 324         if (i->count > count)
 325                 i->count = count;
 326 }
 327
 328 /*
 329  * reexpand a previously truncated iterator; count must be no more than how much
 330  * we had shrunk it.
 331  */
 332 static inline void iov_iter_reexpand(struct iov_iter *i, size_t count)
 333 {
 334         i->count = count;
 335 }
 336
 337 static inline int
 338 iov_iter_npages_cap(struct iov_iter *i, int maxpages, size_t max_bytes)
 339 {
 340         size_t shorted = 0;
 341         int npages;
 342
 343         if (iov_iter_count(i) > max_bytes) {
 344                 shorted = iov_iter_count(i) - max_bytes;
 345                 iov_iter_truncate(i, max_bytes);
 346         }
 347         npages = iov_iter_npages(i, maxpages);
 348         if (shorted)
 349                 iov_iter_reexpand(i, iov_iter_count(i) + shorted);
 350
 351         return npages;
 352 }
 353
 354 struct iovec *iovec_from_user(const struct iovec __user *uvector,
 355                 unsigned long nr_segs, unsigned long fast_segs,
 356                 struct iovec *fast_iov, bool compat);
 357 ssize_t import_iovec(int type, const struct iovec __user *uvec,
 358                  unsigned nr_segs, unsigned fast_segs, struct iovec **iovp,
 359                  struct iov_iter *i);
 360 ssize_t __import_iovec(int type, const struct iovec __user *uvec,
 361                  unsigned nr_segs, unsigned fast_segs, struct iovec **iovp,
 362                  struct iov_iter *i, bool compat);
 363 int import_ubuf(int type, void __user *buf, size_t len, struct iov_iter *i);
 364
 365 static inline void iov_iter_ubuf(struct iov_iter *i, unsigned int direction,
 366                         void __user *buf, size_t count)
 367 {
 368         WARN_ON(direction & ~(READ | WRITE));
 369         *i = (struct iov_iter) {
 370                 .iter_type = ITER_UBUF,
 371                 .data_source = direction,
 372                 .ubuf = buf,
 373                 .count = count,
 374                 .nr_segs = 1
 375         };
 376 }
 377 /* Flags for iov_iter_get/extract_pages*() */
 378 /* Allow P2PDMA on the extracted pages */
 379 #define ITER_ALLOW_P2PDMA       ((__force iov_iter_extraction_t)0x01)
 380
 381 ssize_t iov_iter_extract_pages(struct iov_iter *i, struct page ***pages,
 382                                size_t maxsize, unsigned int maxpages,
 383                                iov_iter_extraction_t extraction_flags,
 384                                size_t *offset0);
 385
 386 /**
 387  * iov_iter_extract_will_pin - Indicate how pages from the iterator will be retained
 388  * @iter: The iterator
 389  *
 390  * Examine the iterator and indicate by returning true or false as to how, if
 391  * at all, pages extracted from the iterator will be retained by the extraction
 392  * function.
 393  *
 394  * %true indicates that the pages will have a pin placed in them that the
 395  * caller must unpin.  This is must be done for DMA/async DIO to force fork()
 396  * to forcibly copy a page for the child (the parent must retain the original
 397  * page).
 398  *
 399  * %false indicates that no measures are taken and that it's up to the caller
 400  * to retain the pages.
 401  */
 402 static inline bool iov_iter_extract_will_pin(const struct iov_iter *iter)
 403 {
 404         return user_backed_iter(iter);
 405 }
 406
 407 struct sg_table;
 408 ssize_t extract_iter_to_sg(struct iov_iter *iter, size_t len,
 409                            struct sg_table *sgtable, unsigned int sg_max,
 410                            iov_iter_extraction_t extraction_flags);
 411
 412 #endif