Commit | Line | Data |
---|---|---|
457c8996 | 1 | // SPDX-License-Identifier: GPL-2.0-only |
7999096f | 2 | #include <crypto/hash.h> |
4f18cd31 | 3 | #include <linux/export.h> |
2f8b5444 | 4 | #include <linux/bvec.h> |
4d0e9df5 | 5 | #include <linux/fault-inject-usercopy.h> |
4f18cd31 AV |
6 | #include <linux/uio.h> |
7 | #include <linux/pagemap.h> | |
28961998 | 8 | #include <linux/highmem.h> |
91f79c43 AV |
9 | #include <linux/slab.h> |
10 | #include <linux/vmalloc.h> | |
241699cd | 11 | #include <linux/splice.h> |
bfdc5970 | 12 | #include <linux/compat.h> |
a604ec7e | 13 | #include <net/checksum.h> |
d05f4435 | 14 | #include <linux/scatterlist.h> |
d0ef4c36 | 15 | #include <linux/instrumented.h> |
4f18cd31 | 16 | |
fcb14cb1 AV |
17 | /* covers ubuf and kbuf alike */ |
18 | #define iterate_buf(i, n, base, len, off, __p, STEP) { \ | |
19 | size_t __maybe_unused off = 0; \ | |
20 | len = n; \ | |
21 | base = __p + i->iov_offset; \ | |
22 | len -= (STEP); \ | |
23 | i->iov_offset += len; \ | |
24 | n = len; \ | |
25 | } | |
26 | ||
5c67aa90 | 27 | /* covers iovec and kvec alike */ |
a6e4ec7b | 28 | #define iterate_iovec(i, n, base, len, off, __p, STEP) { \ |
7baa5099 | 29 | size_t off = 0; \ |
a6e4ec7b | 30 | size_t skip = i->iov_offset; \ |
7a1bcb5d | 31 | do { \ |
7baa5099 AV |
32 | len = min(n, __p->iov_len - skip); \ |
33 | if (likely(len)) { \ | |
34 | base = __p->iov_base + skip; \ | |
35 | len -= (STEP); \ | |
36 | off += len; \ | |
37 | skip += len; \ | |
38 | n -= len; \ | |
7a1bcb5d AV |
39 | if (skip < __p->iov_len) \ |
40 | break; \ | |
41 | } \ | |
42 | __p++; \ | |
43 | skip = 0; \ | |
44 | } while (n); \ | |
a6e4ec7b | 45 | i->iov_offset = skip; \ |
7baa5099 | 46 | n = off; \ |
04a31165 AV |
47 | } |
48 | ||
a6e4ec7b | 49 | #define iterate_bvec(i, n, base, len, off, p, STEP) { \ |
7baa5099 | 50 | size_t off = 0; \ |
a6e4ec7b | 51 | unsigned skip = i->iov_offset; \ |
7491a2bf AV |
52 | while (n) { \ |
53 | unsigned offset = p->bv_offset + skip; \ | |
1b4fb5ff | 54 | unsigned left; \ |
21b56c84 AV |
55 | void *kaddr = kmap_local_page(p->bv_page + \ |
56 | offset / PAGE_SIZE); \ | |
7baa5099 | 57 | base = kaddr + offset % PAGE_SIZE; \ |
a6e4ec7b | 58 | len = min(min(n, (size_t)(p->bv_len - skip)), \ |
7491a2bf | 59 | (size_t)(PAGE_SIZE - offset % PAGE_SIZE)); \ |
1b4fb5ff | 60 | left = (STEP); \ |
21b56c84 | 61 | kunmap_local(kaddr); \ |
7baa5099 AV |
62 | len -= left; \ |
63 | off += len; \ | |
64 | skip += len; \ | |
7491a2bf AV |
65 | if (skip == p->bv_len) { \ |
66 | skip = 0; \ | |
67 | p++; \ | |
68 | } \ | |
7baa5099 | 69 | n -= len; \ |
1b4fb5ff AV |
70 | if (left) \ |
71 | break; \ | |
7491a2bf | 72 | } \ |
a6e4ec7b | 73 | i->iov_offset = skip; \ |
7baa5099 | 74 | n = off; \ |
04a31165 AV |
75 | } |
76 | ||
a6e4ec7b | 77 | #define iterate_xarray(i, n, base, len, __off, STEP) { \ |
1b4fb5ff | 78 | __label__ __out; \ |
622838f3 | 79 | size_t __off = 0; \ |
821979f5 | 80 | struct folio *folio; \ |
a6e4ec7b | 81 | loff_t start = i->xarray_start + i->iov_offset; \ |
4b179e9a | 82 | pgoff_t index = start / PAGE_SIZE; \ |
7ff50620 DH |
83 | XA_STATE(xas, i->xarray, index); \ |
84 | \ | |
821979f5 | 85 | len = PAGE_SIZE - offset_in_page(start); \ |
7baa5099 | 86 | rcu_read_lock(); \ |
821979f5 | 87 | xas_for_each(&xas, folio, ULONG_MAX) { \ |
7baa5099 | 88 | unsigned left; \ |
821979f5 MWO |
89 | size_t offset; \ |
90 | if (xas_retry(&xas, folio)) \ | |
7baa5099 | 91 | continue; \ |
821979f5 | 92 | if (WARN_ON(xa_is_value(folio))) \ |
7baa5099 | 93 | break; \ |
821979f5 | 94 | if (WARN_ON(folio_test_hugetlb(folio))) \ |
7baa5099 | 95 | break; \ |
821979f5 MWO |
96 | offset = offset_in_folio(folio, start + __off); \ |
97 | while (offset < folio_size(folio)) { \ | |
98 | base = kmap_local_folio(folio, offset); \ | |
7baa5099 AV |
99 | len = min(n, len); \ |
100 | left = (STEP); \ | |
821979f5 | 101 | kunmap_local(base); \ |
7baa5099 AV |
102 | len -= left; \ |
103 | __off += len; \ | |
104 | n -= len; \ | |
105 | if (left || n == 0) \ | |
106 | goto __out; \ | |
821979f5 MWO |
107 | offset += len; \ |
108 | len = PAGE_SIZE; \ | |
7baa5099 | 109 | } \ |
7ff50620 | 110 | } \ |
1b4fb5ff | 111 | __out: \ |
7ff50620 | 112 | rcu_read_unlock(); \ |
821979f5 | 113 | i->iov_offset += __off; \ |
622838f3 | 114 | n = __off; \ |
7ff50620 DH |
115 | } |
116 | ||
7baa5099 | 117 | #define __iterate_and_advance(i, n, base, len, off, I, K) { \ |
dd254f5a AV |
118 | if (unlikely(i->count < n)) \ |
119 | n = i->count; \ | |
f5da8354 | 120 | if (likely(n)) { \ |
fcb14cb1 AV |
121 | if (likely(iter_is_ubuf(i))) { \ |
122 | void __user *base; \ | |
123 | size_t len; \ | |
124 | iterate_buf(i, n, base, len, off, \ | |
125 | i->ubuf, (I)) \ | |
126 | } else if (likely(iter_is_iovec(i))) { \ | |
de4f5fed | 127 | const struct iovec *iov = iter_iov(i); \ |
7baa5099 AV |
128 | void __user *base; \ |
129 | size_t len; \ | |
130 | iterate_iovec(i, n, base, len, off, \ | |
a6e4ec7b | 131 | iov, (I)) \ |
de4f5fed JA |
132 | i->nr_segs -= iov - iter_iov(i); \ |
133 | i->__iov = iov; \ | |
28f38db7 | 134 | } else if (iov_iter_is_bvec(i)) { \ |
1bdc76ae | 135 | const struct bio_vec *bvec = i->bvec; \ |
7baa5099 AV |
136 | void *base; \ |
137 | size_t len; \ | |
138 | iterate_bvec(i, n, base, len, off, \ | |
a6e4ec7b | 139 | bvec, (K)) \ |
7491a2bf AV |
140 | i->nr_segs -= bvec - i->bvec; \ |
141 | i->bvec = bvec; \ | |
28f38db7 | 142 | } else if (iov_iter_is_kvec(i)) { \ |
5c67aa90 | 143 | const struct kvec *kvec = i->kvec; \ |
7baa5099 AV |
144 | void *base; \ |
145 | size_t len; \ | |
146 | iterate_iovec(i, n, base, len, off, \ | |
a6e4ec7b | 147 | kvec, (K)) \ |
dd254f5a AV |
148 | i->nr_segs -= kvec - i->kvec; \ |
149 | i->kvec = kvec; \ | |
28f38db7 | 150 | } else if (iov_iter_is_xarray(i)) { \ |
7baa5099 AV |
151 | void *base; \ |
152 | size_t len; \ | |
153 | iterate_xarray(i, n, base, len, off, \ | |
a6e4ec7b | 154 | (K)) \ |
7ce2a91e | 155 | } \ |
dd254f5a | 156 | i->count -= n; \ |
7ce2a91e | 157 | } \ |
7ce2a91e | 158 | } |
7baa5099 AV |
159 | #define iterate_and_advance(i, n, base, len, off, I, K) \ |
160 | __iterate_and_advance(i, n, base, len, off, I, ((void)(K),0)) | |
7ce2a91e | 161 | |
09fc68dc AV |
162 | static int copyout(void __user *to, const void *from, size_t n) |
163 | { | |
4d0e9df5 AL |
164 | if (should_fail_usercopy()) |
165 | return n; | |
96d4f267 | 166 | if (access_ok(to, n)) { |
d0ef4c36 | 167 | instrument_copy_to_user(to, from, n); |
09fc68dc AV |
168 | n = raw_copy_to_user(to, from, n); |
169 | } | |
170 | return n; | |
171 | } | |
172 | ||
4f80818b LS |
173 | static int copyout_nofault(void __user *to, const void *from, size_t n) |
174 | { | |
175 | long res; | |
176 | ||
177 | if (should_fail_usercopy()) | |
178 | return n; | |
179 | ||
180 | res = copy_to_user_nofault(to, from, n); | |
181 | ||
182 | return res < 0 ? n : res; | |
183 | } | |
184 | ||
09fc68dc AV |
185 | static int copyin(void *to, const void __user *from, size_t n) |
186 | { | |
33b75c1d AP |
187 | size_t res = n; |
188 | ||
4d0e9df5 AL |
189 | if (should_fail_usercopy()) |
190 | return n; | |
96d4f267 | 191 | if (access_ok(from, n)) { |
33b75c1d AP |
192 | instrument_copy_from_user_before(to, from, n); |
193 | res = raw_copy_from_user(to, from, n); | |
194 | instrument_copy_from_user_after(to, from, n, res); | |
09fc68dc | 195 | } |
33b75c1d | 196 | return res; |
09fc68dc AV |
197 | } |
198 | ||
171a0203 | 199 | /* |
a6294593 AG |
200 | * fault_in_iov_iter_readable - fault in iov iterator for reading |
201 | * @i: iterator | |
202 | * @size: maximum length | |
203 | * | |
171a0203 | 204 | * Fault in one or more iovecs of the given iov_iter, to a maximum length of |
a6294593 AG |
205 | * @size. For each iovec, fault in each page that constitutes the iovec. |
206 | * | |
207 | * Returns the number of bytes not faulted in (like copy_to_user() and | |
208 | * copy_from_user()). | |
171a0203 | 209 | * |
a6294593 | 210 | * Always returns 0 for non-userspace iterators. |
171a0203 | 211 | */ |
a6294593 | 212 | size_t fault_in_iov_iter_readable(const struct iov_iter *i, size_t size) |
171a0203 | 213 | { |
fcb14cb1 AV |
214 | if (iter_is_ubuf(i)) { |
215 | size_t n = min(size, iov_iter_count(i)); | |
216 | n -= fault_in_readable(i->ubuf + i->iov_offset, n); | |
217 | return size - n; | |
218 | } else if (iter_is_iovec(i)) { | |
a6294593 | 219 | size_t count = min(size, iov_iter_count(i)); |
8409a0d2 AV |
220 | const struct iovec *p; |
221 | size_t skip; | |
222 | ||
a6294593 | 223 | size -= count; |
de4f5fed | 224 | for (p = iter_iov(i), skip = i->iov_offset; count; p++, skip = 0) { |
a6294593 AG |
225 | size_t len = min(count, p->iov_len - skip); |
226 | size_t ret; | |
8409a0d2 AV |
227 | |
228 | if (unlikely(!len)) | |
229 | continue; | |
a6294593 AG |
230 | ret = fault_in_readable(p->iov_base + skip, len); |
231 | count -= len - ret; | |
232 | if (ret) | |
233 | break; | |
8409a0d2 | 234 | } |
a6294593 | 235 | return count + size; |
171a0203 AA |
236 | } |
237 | return 0; | |
238 | } | |
a6294593 | 239 | EXPORT_SYMBOL(fault_in_iov_iter_readable); |
171a0203 | 240 | |
cdd591fc AG |
241 | /* |
242 | * fault_in_iov_iter_writeable - fault in iov iterator for writing | |
243 | * @i: iterator | |
244 | * @size: maximum length | |
245 | * | |
246 | * Faults in the iterator using get_user_pages(), i.e., without triggering | |
247 | * hardware page faults. This is primarily useful when we already know that | |
248 | * some or all of the pages in @i aren't in memory. | |
249 | * | |
250 | * Returns the number of bytes not faulted in, like copy_to_user() and | |
251 | * copy_from_user(). | |
252 | * | |
253 | * Always returns 0 for non-user-space iterators. | |
254 | */ | |
255 | size_t fault_in_iov_iter_writeable(const struct iov_iter *i, size_t size) | |
256 | { | |
fcb14cb1 AV |
257 | if (iter_is_ubuf(i)) { |
258 | size_t n = min(size, iov_iter_count(i)); | |
259 | n -= fault_in_safe_writeable(i->ubuf + i->iov_offset, n); | |
260 | return size - n; | |
261 | } else if (iter_is_iovec(i)) { | |
cdd591fc AG |
262 | size_t count = min(size, iov_iter_count(i)); |
263 | const struct iovec *p; | |
264 | size_t skip; | |
265 | ||
266 | size -= count; | |
de4f5fed | 267 | for (p = iter_iov(i), skip = i->iov_offset; count; p++, skip = 0) { |
cdd591fc AG |
268 | size_t len = min(count, p->iov_len - skip); |
269 | size_t ret; | |
270 | ||
271 | if (unlikely(!len)) | |
272 | continue; | |
273 | ret = fault_in_safe_writeable(p->iov_base + skip, len); | |
274 | count -= len - ret; | |
275 | if (ret) | |
276 | break; | |
277 | } | |
278 | return count + size; | |
279 | } | |
280 | return 0; | |
281 | } | |
282 | EXPORT_SYMBOL(fault_in_iov_iter_writeable); | |
283 | ||
aa563d7b | 284 | void iov_iter_init(struct iov_iter *i, unsigned int direction, |
71d8e532 AV |
285 | const struct iovec *iov, unsigned long nr_segs, |
286 | size_t count) | |
287 | { | |
aa563d7b | 288 | WARN_ON(direction & ~(READ | WRITE)); |
8cd54c1c AV |
289 | *i = (struct iov_iter) { |
290 | .iter_type = ITER_IOVEC, | |
245f0922 | 291 | .copy_mc = false, |
3337ab08 | 292 | .nofault = false, |
fcb14cb1 | 293 | .user_backed = true, |
8cd54c1c | 294 | .data_source = direction, |
de4f5fed | 295 | .__iov = iov, |
8cd54c1c AV |
296 | .nr_segs = nr_segs, |
297 | .iov_offset = 0, | |
298 | .count = count | |
299 | }; | |
71d8e532 AV |
300 | } |
301 | EXPORT_SYMBOL(iov_iter_init); | |
7b2c99d1 | 302 | |
f9152895 AV |
303 | static __wsum csum_and_memcpy(void *to, const void *from, size_t len, |
304 | __wsum sum, size_t off) | |
305 | { | |
cc44c17b | 306 | __wsum next = csum_partial_copy_nocheck(from, to, len); |
f9152895 AV |
307 | return csum_block_add(sum, next, off); |
308 | } | |
309 | ||
aa28de27 | 310 | size_t _copy_to_iter(const void *addr, size_t bytes, struct iov_iter *i) |
62a8067a | 311 | { |
a41dad90 AV |
312 | if (WARN_ON_ONCE(i->data_source)) |
313 | return 0; | |
fcb14cb1 | 314 | if (user_backed_iter(i)) |
09fc68dc | 315 | might_fault(); |
7baa5099 AV |
316 | iterate_and_advance(i, bytes, base, len, off, |
317 | copyout(base, addr + off, len), | |
318 | memcpy(base, addr + off, len) | |
3d4d3e48 | 319 | ) |
62a8067a | 320 | |
3d4d3e48 | 321 | return bytes; |
c35e0248 | 322 | } |
aa28de27 | 323 | EXPORT_SYMBOL(_copy_to_iter); |
c35e0248 | 324 | |
ec6347bb DW |
325 | #ifdef CONFIG_ARCH_HAS_COPY_MC |
326 | static int copyout_mc(void __user *to, const void *from, size_t n) | |
8780356e | 327 | { |
96d4f267 | 328 | if (access_ok(to, n)) { |
d0ef4c36 | 329 | instrument_copy_to_user(to, from, n); |
ec6347bb | 330 | n = copy_mc_to_user((__force void *) to, from, n); |
8780356e DW |
331 | } |
332 | return n; | |
333 | } | |
334 | ||
bf3eeb9b | 335 | /** |
ec6347bb | 336 | * _copy_mc_to_iter - copy to iter with source memory error exception handling |
bf3eeb9b DW |
337 | * @addr: source kernel address |
338 | * @bytes: total transfer length | |
44e55997 | 339 | * @i: destination iterator |
bf3eeb9b | 340 | * |
ec6347bb DW |
341 | * The pmem driver deploys this for the dax operation |
342 | * (dax_copy_to_iter()) for dax reads (bypass page-cache and the | |
343 | * block-layer). Upon #MC read(2) aborts and returns EIO or the bytes | |
344 | * successfully copied. | |
bf3eeb9b | 345 | * |
ec6347bb | 346 | * The main differences between this and typical _copy_to_iter(). |
bf3eeb9b DW |
347 | * |
348 | * * Typical tail/residue handling after a fault retries the copy | |
349 | * byte-by-byte until the fault happens again. Re-triggering machine | |
350 | * checks is potentially fatal so the implementation uses source | |
351 | * alignment and poison alignment assumptions to avoid re-triggering | |
352 | * hardware exceptions. | |
353 | * | |
3fc40265 DH |
354 | * * ITER_KVEC and ITER_BVEC can return short copies. Compare to |
355 | * copy_to_iter() where only ITER_IOVEC attempts might return a short copy. | |
44e55997 RD |
356 | * |
357 | * Return: number of bytes copied (may be %0) | |
bf3eeb9b | 358 | */ |
ec6347bb | 359 | size_t _copy_mc_to_iter(const void *addr, size_t bytes, struct iov_iter *i) |
8780356e | 360 | { |
a41dad90 AV |
361 | if (WARN_ON_ONCE(i->data_source)) |
362 | return 0; | |
fcb14cb1 | 363 | if (user_backed_iter(i)) |
8780356e | 364 | might_fault(); |
7baa5099 AV |
365 | __iterate_and_advance(i, bytes, base, len, off, |
366 | copyout_mc(base, addr + off, len), | |
367 | copy_mc_to_kernel(base, addr + off, len) | |
8780356e DW |
368 | ) |
369 | ||
370 | return bytes; | |
371 | } | |
ec6347bb DW |
372 | EXPORT_SYMBOL_GPL(_copy_mc_to_iter); |
373 | #endif /* CONFIG_ARCH_HAS_COPY_MC */ | |
8780356e | 374 | |
245f0922 KW |
375 | static void *memcpy_from_iter(struct iov_iter *i, void *to, const void *from, |
376 | size_t size) | |
377 | { | |
378 | if (iov_iter_is_copy_mc(i)) | |
379 | return (void *)copy_mc_to_kernel(to, from, size); | |
380 | return memcpy(to, from, size); | |
381 | } | |
382 | ||
aa28de27 | 383 | size_t _copy_from_iter(void *addr, size_t bytes, struct iov_iter *i) |
c35e0248 | 384 | { |
a41dad90 | 385 | if (WARN_ON_ONCE(!i->data_source)) |
241699cd | 386 | return 0; |
a41dad90 | 387 | |
fcb14cb1 | 388 | if (user_backed_iter(i)) |
09fc68dc | 389 | might_fault(); |
7baa5099 AV |
390 | iterate_and_advance(i, bytes, base, len, off, |
391 | copyin(addr + off, base, len), | |
245f0922 | 392 | memcpy_from_iter(i, addr + off, base, len) |
0dbca9a4 AV |
393 | ) |
394 | ||
395 | return bytes; | |
c35e0248 | 396 | } |
aa28de27 | 397 | EXPORT_SYMBOL(_copy_from_iter); |
c35e0248 | 398 | |
aa28de27 | 399 | size_t _copy_from_iter_nocache(void *addr, size_t bytes, struct iov_iter *i) |
aa583096 | 400 | { |
a41dad90 | 401 | if (WARN_ON_ONCE(!i->data_source)) |
241699cd | 402 | return 0; |
a41dad90 | 403 | |
7baa5099 AV |
404 | iterate_and_advance(i, bytes, base, len, off, |
405 | __copy_from_user_inatomic_nocache(addr + off, base, len), | |
406 | memcpy(addr + off, base, len) | |
aa583096 AV |
407 | ) |
408 | ||
409 | return bytes; | |
410 | } | |
aa28de27 | 411 | EXPORT_SYMBOL(_copy_from_iter_nocache); |
aa583096 | 412 | |
0aed55af | 413 | #ifdef CONFIG_ARCH_HAS_UACCESS_FLUSHCACHE |
abd08d7d DW |
414 | /** |
415 | * _copy_from_iter_flushcache - write destination through cpu cache | |
416 | * @addr: destination kernel address | |
417 | * @bytes: total transfer length | |
44e55997 | 418 | * @i: source iterator |
abd08d7d DW |
419 | * |
420 | * The pmem driver arranges for filesystem-dax to use this facility via | |
421 | * dax_copy_from_iter() for ensuring that writes to persistent memory | |
422 | * are flushed through the CPU cache. It is differentiated from | |
423 | * _copy_from_iter_nocache() in that guarantees all data is flushed for | |
424 | * all iterator types. The _copy_from_iter_nocache() only attempts to | |
425 | * bypass the cache for the ITER_IOVEC case, and on some archs may use | |
426 | * instructions that strand dirty-data in the cache. | |
44e55997 RD |
427 | * |
428 | * Return: number of bytes copied (may be %0) | |
abd08d7d | 429 | */ |
6a37e940 | 430 | size_t _copy_from_iter_flushcache(void *addr, size_t bytes, struct iov_iter *i) |
0aed55af | 431 | { |
a41dad90 | 432 | if (WARN_ON_ONCE(!i->data_source)) |
0aed55af | 433 | return 0; |
a41dad90 | 434 | |
7baa5099 AV |
435 | iterate_and_advance(i, bytes, base, len, off, |
436 | __copy_from_user_flushcache(addr + off, base, len), | |
437 | memcpy_flushcache(addr + off, base, len) | |
0aed55af DW |
438 | ) |
439 | ||
440 | return bytes; | |
441 | } | |
6a37e940 | 442 | EXPORT_SYMBOL_GPL(_copy_from_iter_flushcache); |
0aed55af DW |
443 | #endif |
444 | ||
72e809ed AV |
445 | static inline bool page_copy_sane(struct page *page, size_t offset, size_t n) |
446 | { | |
6daef95b ED |
447 | struct page *head; |
448 | size_t v = n + offset; | |
449 | ||
450 | /* | |
451 | * The general case needs to access the page order in order | |
452 | * to compute the page size. | |
453 | * However, we mostly deal with order-0 pages and thus can | |
454 | * avoid a possible cache line miss for requests that fit all | |
455 | * page orders. | |
456 | */ | |
457 | if (n <= v && v <= PAGE_SIZE) | |
458 | return true; | |
459 | ||
460 | head = compound_head(page); | |
461 | v += (page - head) << PAGE_SHIFT; | |
a90bcb86 | 462 | |
40a86061 AV |
463 | if (WARN_ON(n > v || v > page_size(head))) |
464 | return false; | |
465 | return true; | |
72e809ed | 466 | } |
cbbd26b8 | 467 | |
08aa6479 AV |
468 | size_t copy_page_to_iter(struct page *page, size_t offset, size_t bytes, |
469 | struct iov_iter *i) | |
470 | { | |
471 | size_t res = 0; | |
40a86061 | 472 | if (!page_copy_sane(page, offset, bytes)) |
08aa6479 | 473 | return 0; |
a41dad90 AV |
474 | if (WARN_ON_ONCE(i->data_source)) |
475 | return 0; | |
08aa6479 AV |
476 | page += offset / PAGE_SIZE; // first subpage |
477 | offset %= PAGE_SIZE; | |
478 | while (1) { | |
f0f6b614 AV |
479 | void *kaddr = kmap_local_page(page); |
480 | size_t n = min(bytes, (size_t)PAGE_SIZE - offset); | |
481 | n = _copy_to_iter(kaddr + offset, n, i); | |
482 | kunmap_local(kaddr); | |
08aa6479 AV |
483 | res += n; |
484 | bytes -= n; | |
485 | if (!bytes || !n) | |
486 | break; | |
487 | offset += n; | |
488 | if (offset == PAGE_SIZE) { | |
489 | page++; | |
490 | offset = 0; | |
491 | } | |
492 | } | |
493 | return res; | |
494 | } | |
62a8067a AV |
495 | EXPORT_SYMBOL(copy_page_to_iter); |
496 | ||
4f80818b LS |
497 | size_t copy_page_to_iter_nofault(struct page *page, unsigned offset, size_t bytes, |
498 | struct iov_iter *i) | |
499 | { | |
500 | size_t res = 0; | |
501 | ||
502 | if (!page_copy_sane(page, offset, bytes)) | |
503 | return 0; | |
504 | if (WARN_ON_ONCE(i->data_source)) | |
505 | return 0; | |
4f80818b LS |
506 | page += offset / PAGE_SIZE; // first subpage |
507 | offset %= PAGE_SIZE; | |
508 | while (1) { | |
509 | void *kaddr = kmap_local_page(page); | |
510 | size_t n = min(bytes, (size_t)PAGE_SIZE - offset); | |
511 | ||
512 | iterate_and_advance(i, n, base, len, off, | |
513 | copyout_nofault(base, kaddr + offset + off, len), | |
514 | memcpy(base, kaddr + offset + off, len) | |
515 | ) | |
516 | kunmap_local(kaddr); | |
517 | res += n; | |
518 | bytes -= n; | |
519 | if (!bytes || !n) | |
520 | break; | |
521 | offset += n; | |
522 | if (offset == PAGE_SIZE) { | |
523 | page++; | |
524 | offset = 0; | |
525 | } | |
526 | } | |
527 | return res; | |
528 | } | |
529 | EXPORT_SYMBOL(copy_page_to_iter_nofault); | |
530 | ||
62a8067a AV |
531 | size_t copy_page_from_iter(struct page *page, size_t offset, size_t bytes, |
532 | struct iov_iter *i) | |
533 | { | |
c03f05f1 AV |
534 | size_t res = 0; |
535 | if (!page_copy_sane(page, offset, bytes)) | |
536 | return 0; | |
537 | page += offset / PAGE_SIZE; // first subpage | |
538 | offset %= PAGE_SIZE; | |
539 | while (1) { | |
55ca375c | 540 | void *kaddr = kmap_local_page(page); |
c03f05f1 AV |
541 | size_t n = min(bytes, (size_t)PAGE_SIZE - offset); |
542 | n = _copy_from_iter(kaddr + offset, n, i); | |
55ca375c | 543 | kunmap_local(kaddr); |
c03f05f1 AV |
544 | res += n; |
545 | bytes -= n; | |
546 | if (!bytes || !n) | |
547 | break; | |
548 | offset += n; | |
549 | if (offset == PAGE_SIZE) { | |
550 | page++; | |
551 | offset = 0; | |
552 | } | |
28f38db7 | 553 | } |
c03f05f1 | 554 | return res; |
62a8067a AV |
555 | } |
556 | EXPORT_SYMBOL(copy_page_from_iter); | |
557 | ||
c35e0248 MW |
558 | size_t iov_iter_zero(size_t bytes, struct iov_iter *i) |
559 | { | |
7baa5099 AV |
560 | iterate_and_advance(i, bytes, base, len, count, |
561 | clear_user(base, len), | |
562 | memset(base, 0, len) | |
8442fa46 AV |
563 | ) |
564 | ||
565 | return bytes; | |
c35e0248 MW |
566 | } |
567 | EXPORT_SYMBOL(iov_iter_zero); | |
568 | ||
908a1ad8 MWO |
569 | size_t copy_page_from_iter_atomic(struct page *page, unsigned offset, |
570 | size_t bytes, struct iov_iter *i) | |
62a8067a | 571 | { |
908a1ad8 | 572 | size_t n, copied = 0; |
f7f9a0c8 MWO |
573 | |
574 | if (!page_copy_sane(page, offset, bytes)) | |
72e809ed | 575 | return 0; |
f7f9a0c8 | 576 | if (WARN_ON_ONCE(!i->data_source)) |
241699cd | 577 | return 0; |
f7f9a0c8 | 578 | |
908a1ad8 MWO |
579 | do { |
580 | char *p; | |
f7f9a0c8 | 581 | |
908a1ad8 MWO |
582 | n = bytes - copied; |
583 | if (PageHighMem(page)) { | |
584 | page += offset / PAGE_SIZE; | |
585 | offset %= PAGE_SIZE; | |
586 | n = min_t(size_t, n, PAGE_SIZE - offset); | |
587 | } | |
588 | ||
589 | p = kmap_atomic(page) + offset; | |
590 | iterate_and_advance(i, n, base, len, off, | |
591 | copyin(p + off, base, len), | |
592 | memcpy_from_iter(i, p + off, base, len) | |
593 | ) | |
594 | kunmap_atomic(p); | |
595 | copied += n; | |
596 | offset += n; | |
597 | } while (PageHighMem(page) && copied != bytes && n > 0); | |
598 | ||
599 | return copied; | |
62a8067a | 600 | } |
f0b65f39 | 601 | EXPORT_SYMBOL(copy_page_from_iter_atomic); |
62a8067a | 602 | |
54c8195b PB |
603 | static void iov_iter_bvec_advance(struct iov_iter *i, size_t size) |
604 | { | |
18fa9af7 | 605 | const struct bio_vec *bvec, *end; |
54c8195b | 606 | |
18fa9af7 AV |
607 | if (!i->count) |
608 | return; | |
609 | i->count -= size; | |
610 | ||
611 | size += i->iov_offset; | |
54c8195b | 612 | |
18fa9af7 AV |
613 | for (bvec = i->bvec, end = bvec + i->nr_segs; bvec < end; bvec++) { |
614 | if (likely(size < bvec->bv_len)) | |
615 | break; | |
616 | size -= bvec->bv_len; | |
617 | } | |
618 | i->iov_offset = size; | |
619 | i->nr_segs -= bvec - i->bvec; | |
620 | i->bvec = bvec; | |
54c8195b PB |
621 | } |
622 | ||
185ac4d4 AV |
623 | static void iov_iter_iovec_advance(struct iov_iter *i, size_t size) |
624 | { | |
625 | const struct iovec *iov, *end; | |
626 | ||
627 | if (!i->count) | |
628 | return; | |
629 | i->count -= size; | |
630 | ||
631 | size += i->iov_offset; // from beginning of current segment | |
de4f5fed | 632 | for (iov = iter_iov(i), end = iov + i->nr_segs; iov < end; iov++) { |
185ac4d4 AV |
633 | if (likely(size < iov->iov_len)) |
634 | break; | |
635 | size -= iov->iov_len; | |
636 | } | |
637 | i->iov_offset = size; | |
de4f5fed JA |
638 | i->nr_segs -= iov - iter_iov(i); |
639 | i->__iov = iov; | |
185ac4d4 AV |
640 | } |
641 | ||
62a8067a AV |
642 | void iov_iter_advance(struct iov_iter *i, size_t size) |
643 | { | |
3b3fc051 AV |
644 | if (unlikely(i->count < size)) |
645 | size = i->count; | |
fcb14cb1 AV |
646 | if (likely(iter_is_ubuf(i)) || unlikely(iov_iter_is_xarray(i))) { |
647 | i->iov_offset += size; | |
648 | i->count -= size; | |
649 | } else if (likely(iter_is_iovec(i) || iov_iter_is_kvec(i))) { | |
185ac4d4 AV |
650 | /* iovec and kvec have identical layouts */ |
651 | iov_iter_iovec_advance(i, size); | |
652 | } else if (iov_iter_is_bvec(i)) { | |
653 | iov_iter_bvec_advance(i, size); | |
185ac4d4 AV |
654 | } else if (iov_iter_is_discard(i)) { |
655 | i->count -= size; | |
54c8195b | 656 | } |
62a8067a AV |
657 | } |
658 | EXPORT_SYMBOL(iov_iter_advance); | |
659 | ||
27c0e374 AV |
660 | void iov_iter_revert(struct iov_iter *i, size_t unroll) |
661 | { | |
662 | if (!unroll) | |
663 | return; | |
5b47d59a AV |
664 | if (WARN_ON(unroll > MAX_RW_COUNT)) |
665 | return; | |
27c0e374 | 666 | i->count += unroll; |
9ea9ce04 DH |
667 | if (unlikely(iov_iter_is_discard(i))) |
668 | return; | |
27c0e374 AV |
669 | if (unroll <= i->iov_offset) { |
670 | i->iov_offset -= unroll; | |
671 | return; | |
672 | } | |
673 | unroll -= i->iov_offset; | |
fcb14cb1 | 674 | if (iov_iter_is_xarray(i) || iter_is_ubuf(i)) { |
7ff50620 DH |
675 | BUG(); /* We should never go beyond the start of the specified |
676 | * range since we might then be straying into pages that | |
677 | * aren't pinned. | |
678 | */ | |
679 | } else if (iov_iter_is_bvec(i)) { | |
27c0e374 AV |
680 | const struct bio_vec *bvec = i->bvec; |
681 | while (1) { | |
682 | size_t n = (--bvec)->bv_len; | |
683 | i->nr_segs++; | |
684 | if (unroll <= n) { | |
685 | i->bvec = bvec; | |
686 | i->iov_offset = n - unroll; | |
687 | return; | |
688 | } | |
689 | unroll -= n; | |
690 | } | |
691 | } else { /* same logics for iovec and kvec */ | |
de4f5fed | 692 | const struct iovec *iov = iter_iov(i); |
27c0e374 AV |
693 | while (1) { |
694 | size_t n = (--iov)->iov_len; | |
695 | i->nr_segs++; | |
696 | if (unroll <= n) { | |
de4f5fed | 697 | i->__iov = iov; |
27c0e374 AV |
698 | i->iov_offset = n - unroll; |
699 | return; | |
700 | } | |
701 | unroll -= n; | |
702 | } | |
703 | } | |
704 | } | |
705 | EXPORT_SYMBOL(iov_iter_revert); | |
706 | ||
62a8067a AV |
707 | /* |
708 | * Return the count of just the current iov_iter segment. | |
709 | */ | |
710 | size_t iov_iter_single_seg_count(const struct iov_iter *i) | |
711 | { | |
28f38db7 AV |
712 | if (i->nr_segs > 1) { |
713 | if (likely(iter_is_iovec(i) || iov_iter_is_kvec(i))) | |
de4f5fed | 714 | return min(i->count, iter_iov(i)->iov_len - i->iov_offset); |
28f38db7 AV |
715 | if (iov_iter_is_bvec(i)) |
716 | return min(i->count, i->bvec->bv_len - i->iov_offset); | |
717 | } | |
718 | return i->count; | |
62a8067a AV |
719 | } |
720 | EXPORT_SYMBOL(iov_iter_single_seg_count); | |
721 | ||
aa563d7b | 722 | void iov_iter_kvec(struct iov_iter *i, unsigned int direction, |
05afcb77 | 723 | const struct kvec *kvec, unsigned long nr_segs, |
abb78f87 AV |
724 | size_t count) |
725 | { | |
aa563d7b | 726 | WARN_ON(direction & ~(READ | WRITE)); |
8cd54c1c AV |
727 | *i = (struct iov_iter){ |
728 | .iter_type = ITER_KVEC, | |
245f0922 | 729 | .copy_mc = false, |
8cd54c1c AV |
730 | .data_source = direction, |
731 | .kvec = kvec, | |
732 | .nr_segs = nr_segs, | |
733 | .iov_offset = 0, | |
734 | .count = count | |
735 | }; | |
abb78f87 AV |
736 | } |
737 | EXPORT_SYMBOL(iov_iter_kvec); | |
738 | ||
aa563d7b | 739 | void iov_iter_bvec(struct iov_iter *i, unsigned int direction, |
05afcb77 AV |
740 | const struct bio_vec *bvec, unsigned long nr_segs, |
741 | size_t count) | |
742 | { | |
aa563d7b | 743 | WARN_ON(direction & ~(READ | WRITE)); |
8cd54c1c AV |
744 | *i = (struct iov_iter){ |
745 | .iter_type = ITER_BVEC, | |
245f0922 | 746 | .copy_mc = false, |
8cd54c1c AV |
747 | .data_source = direction, |
748 | .bvec = bvec, | |
749 | .nr_segs = nr_segs, | |
750 | .iov_offset = 0, | |
751 | .count = count | |
752 | }; | |
05afcb77 AV |
753 | } |
754 | EXPORT_SYMBOL(iov_iter_bvec); | |
755 | ||
7ff50620 DH |
756 | /** |
757 | * iov_iter_xarray - Initialise an I/O iterator to use the pages in an xarray | |
758 | * @i: The iterator to initialise. | |
759 | * @direction: The direction of the transfer. | |
760 | * @xarray: The xarray to access. | |
761 | * @start: The start file position. | |
762 | * @count: The size of the I/O buffer in bytes. | |
763 | * | |
764 | * Set up an I/O iterator to either draw data out of the pages attached to an | |
765 | * inode or to inject data into those pages. The pages *must* be prevented | |
766 | * from evaporation, either by taking a ref on them or locking them by the | |
767 | * caller. | |
768 | */ | |
769 | void iov_iter_xarray(struct iov_iter *i, unsigned int direction, | |
770 | struct xarray *xarray, loff_t start, size_t count) | |
771 | { | |
772 | BUG_ON(direction & ~1); | |
8cd54c1c AV |
773 | *i = (struct iov_iter) { |
774 | .iter_type = ITER_XARRAY, | |
245f0922 | 775 | .copy_mc = false, |
8cd54c1c AV |
776 | .data_source = direction, |
777 | .xarray = xarray, | |
778 | .xarray_start = start, | |
779 | .count = count, | |
780 | .iov_offset = 0 | |
781 | }; | |
7ff50620 DH |
782 | } |
783 | EXPORT_SYMBOL(iov_iter_xarray); | |
784 | ||
9ea9ce04 DH |
785 | /** |
786 | * iov_iter_discard - Initialise an I/O iterator that discards data | |
787 | * @i: The iterator to initialise. | |
788 | * @direction: The direction of the transfer. | |
789 | * @count: The size of the I/O buffer in bytes. | |
790 | * | |
791 | * Set up an I/O iterator that just discards everything that's written to it. | |
792 | * It's only available as a READ iterator. | |
793 | */ | |
794 | void iov_iter_discard(struct iov_iter *i, unsigned int direction, size_t count) | |
795 | { | |
796 | BUG_ON(direction != READ); | |
8cd54c1c AV |
797 | *i = (struct iov_iter){ |
798 | .iter_type = ITER_DISCARD, | |
245f0922 | 799 | .copy_mc = false, |
8cd54c1c AV |
800 | .data_source = false, |
801 | .count = count, | |
802 | .iov_offset = 0 | |
803 | }; | |
9ea9ce04 DH |
804 | } |
805 | EXPORT_SYMBOL(iov_iter_discard); | |
806 | ||
cfa320f7 KB |
807 | static bool iov_iter_aligned_iovec(const struct iov_iter *i, unsigned addr_mask, |
808 | unsigned len_mask) | |
809 | { | |
810 | size_t size = i->count; | |
811 | size_t skip = i->iov_offset; | |
812 | unsigned k; | |
813 | ||
814 | for (k = 0; k < i->nr_segs; k++, skip = 0) { | |
de4f5fed JA |
815 | const struct iovec *iov = iter_iov(i) + k; |
816 | size_t len = iov->iov_len - skip; | |
cfa320f7 KB |
817 | |
818 | if (len > size) | |
819 | len = size; | |
820 | if (len & len_mask) | |
821 | return false; | |
de4f5fed | 822 | if ((unsigned long)(iov->iov_base + skip) & addr_mask) |
cfa320f7 KB |
823 | return false; |
824 | ||
825 | size -= len; | |
826 | if (!size) | |
827 | break; | |
828 | } | |
829 | return true; | |
830 | } | |
831 | ||
832 | static bool iov_iter_aligned_bvec(const struct iov_iter *i, unsigned addr_mask, | |
833 | unsigned len_mask) | |
834 | { | |
835 | size_t size = i->count; | |
836 | unsigned skip = i->iov_offset; | |
837 | unsigned k; | |
838 | ||
839 | for (k = 0; k < i->nr_segs; k++, skip = 0) { | |
840 | size_t len = i->bvec[k].bv_len - skip; | |
841 | ||
842 | if (len > size) | |
843 | len = size; | |
844 | if (len & len_mask) | |
845 | return false; | |
846 | if ((unsigned long)(i->bvec[k].bv_offset + skip) & addr_mask) | |
847 | return false; | |
848 | ||
849 | size -= len; | |
850 | if (!size) | |
851 | break; | |
852 | } | |
853 | return true; | |
854 | } | |
855 | ||
856 | /** | |
857 | * iov_iter_is_aligned() - Check if the addresses and lengths of each segments | |
858 | * are aligned to the parameters. | |
859 | * | |
860 | * @i: &struct iov_iter to restore | |
861 | * @addr_mask: bit mask to check against the iov element's addresses | |
862 | * @len_mask: bit mask to check against the iov element's lengths | |
863 | * | |
864 | * Return: false if any addresses or lengths intersect with the provided masks | |
865 | */ | |
866 | bool iov_iter_is_aligned(const struct iov_iter *i, unsigned addr_mask, | |
867 | unsigned len_mask) | |
868 | { | |
fcb14cb1 AV |
869 | if (likely(iter_is_ubuf(i))) { |
870 | if (i->count & len_mask) | |
871 | return false; | |
872 | if ((unsigned long)(i->ubuf + i->iov_offset) & addr_mask) | |
873 | return false; | |
874 | return true; | |
875 | } | |
876 | ||
cfa320f7 KB |
877 | if (likely(iter_is_iovec(i) || iov_iter_is_kvec(i))) |
878 | return iov_iter_aligned_iovec(i, addr_mask, len_mask); | |
879 | ||
880 | if (iov_iter_is_bvec(i)) | |
881 | return iov_iter_aligned_bvec(i, addr_mask, len_mask); | |
882 | ||
cfa320f7 KB |
883 | if (iov_iter_is_xarray(i)) { |
884 | if (i->count & len_mask) | |
885 | return false; | |
886 | if ((i->xarray_start + i->iov_offset) & addr_mask) | |
887 | return false; | |
888 | } | |
889 | ||
890 | return true; | |
891 | } | |
892 | EXPORT_SYMBOL_GPL(iov_iter_is_aligned); | |
893 | ||
9221d2e3 | 894 | static unsigned long iov_iter_alignment_iovec(const struct iov_iter *i) |
62a8067a | 895 | { |
04a31165 AV |
896 | unsigned long res = 0; |
897 | size_t size = i->count; | |
9221d2e3 AV |
898 | size_t skip = i->iov_offset; |
899 | unsigned k; | |
900 | ||
901 | for (k = 0; k < i->nr_segs; k++, skip = 0) { | |
de4f5fed JA |
902 | const struct iovec *iov = iter_iov(i) + k; |
903 | size_t len = iov->iov_len - skip; | |
9221d2e3 | 904 | if (len) { |
de4f5fed | 905 | res |= (unsigned long)iov->iov_base + skip; |
9221d2e3 AV |
906 | if (len > size) |
907 | len = size; | |
908 | res |= len; | |
909 | size -= len; | |
910 | if (!size) | |
911 | break; | |
912 | } | |
913 | } | |
914 | return res; | |
915 | } | |
04a31165 | 916 | |
9221d2e3 AV |
917 | static unsigned long iov_iter_alignment_bvec(const struct iov_iter *i) |
918 | { | |
919 | unsigned res = 0; | |
920 | size_t size = i->count; | |
921 | unsigned skip = i->iov_offset; | |
922 | unsigned k; | |
923 | ||
924 | for (k = 0; k < i->nr_segs; k++, skip = 0) { | |
925 | size_t len = i->bvec[k].bv_len - skip; | |
926 | res |= (unsigned long)i->bvec[k].bv_offset + skip; | |
927 | if (len > size) | |
928 | len = size; | |
929 | res |= len; | |
930 | size -= len; | |
931 | if (!size) | |
932 | break; | |
933 | } | |
934 | return res; | |
935 | } | |
936 | ||
937 | unsigned long iov_iter_alignment(const struct iov_iter *i) | |
938 | { | |
fcb14cb1 AV |
939 | if (likely(iter_is_ubuf(i))) { |
940 | size_t size = i->count; | |
941 | if (size) | |
942 | return ((unsigned long)i->ubuf + i->iov_offset) | size; | |
943 | return 0; | |
944 | } | |
945 | ||
9221d2e3 AV |
946 | /* iovec and kvec have identical layouts */ |
947 | if (likely(iter_is_iovec(i) || iov_iter_is_kvec(i))) | |
948 | return iov_iter_alignment_iovec(i); | |
949 | ||
950 | if (iov_iter_is_bvec(i)) | |
951 | return iov_iter_alignment_bvec(i); | |
952 | ||
9221d2e3 | 953 | if (iov_iter_is_xarray(i)) |
3d14ec1f | 954 | return (i->xarray_start + i->iov_offset) | i->count; |
9221d2e3 AV |
955 | |
956 | return 0; | |
62a8067a AV |
957 | } |
958 | EXPORT_SYMBOL(iov_iter_alignment); | |
959 | ||
357f435d AV |
960 | unsigned long iov_iter_gap_alignment(const struct iov_iter *i) |
961 | { | |
33844e66 | 962 | unsigned long res = 0; |
610c7a71 | 963 | unsigned long v = 0; |
357f435d | 964 | size_t size = i->count; |
610c7a71 | 965 | unsigned k; |
357f435d | 966 | |
fcb14cb1 AV |
967 | if (iter_is_ubuf(i)) |
968 | return 0; | |
969 | ||
610c7a71 | 970 | if (WARN_ON(!iter_is_iovec(i))) |
241699cd | 971 | return ~0U; |
241699cd | 972 | |
610c7a71 | 973 | for (k = 0; k < i->nr_segs; k++) { |
de4f5fed JA |
974 | const struct iovec *iov = iter_iov(i) + k; |
975 | if (iov->iov_len) { | |
976 | unsigned long base = (unsigned long)iov->iov_base; | |
610c7a71 AV |
977 | if (v) // if not the first one |
978 | res |= base | v; // this start | previous end | |
de4f5fed JA |
979 | v = base + iov->iov_len; |
980 | if (size <= iov->iov_len) | |
610c7a71 | 981 | break; |
de4f5fed | 982 | size -= iov->iov_len; |
610c7a71 AV |
983 | } |
984 | } | |
33844e66 | 985 | return res; |
357f435d AV |
986 | } |
987 | EXPORT_SYMBOL(iov_iter_gap_alignment); | |
988 | ||
3cf42da3 AV |
989 | static int want_pages_array(struct page ***res, size_t size, |
990 | size_t start, unsigned int maxpages) | |
acbdeb83 | 991 | { |
3cf42da3 AV |
992 | unsigned int count = DIV_ROUND_UP(size + start, PAGE_SIZE); |
993 | ||
994 | if (count > maxpages) | |
995 | count = maxpages; | |
996 | WARN_ON(!count); // caller should've prevented that | |
997 | if (!*res) { | |
998 | *res = kvmalloc_array(count, sizeof(struct page *), GFP_KERNEL); | |
999 | if (!*res) | |
1000 | return 0; | |
1001 | } | |
1002 | return count; | |
acbdeb83 AV |
1003 | } |
1004 | ||
7ff50620 DH |
1005 | static ssize_t iter_xarray_populate_pages(struct page **pages, struct xarray *xa, |
1006 | pgoff_t index, unsigned int nr_pages) | |
1007 | { | |
1008 | XA_STATE(xas, xa, index); | |
1009 | struct page *page; | |
1010 | unsigned int ret = 0; | |
1011 | ||
1012 | rcu_read_lock(); | |
1013 | for (page = xas_load(&xas); page; page = xas_next(&xas)) { | |
1014 | if (xas_retry(&xas, page)) | |
1015 | continue; | |
1016 | ||
1017 | /* Has the page moved or been split? */ | |
1018 | if (unlikely(page != xas_reload(&xas))) { | |
1019 | xas_reset(&xas); | |
1020 | continue; | |
1021 | } | |
1022 | ||
1023 | pages[ret] = find_subpage(page, xas.xa_index); | |
1024 | get_page(pages[ret]); | |
1025 | if (++ret == nr_pages) | |
1026 | break; | |
1027 | } | |
1028 | rcu_read_unlock(); | |
1029 | return ret; | |
1030 | } | |
1031 | ||
1032 | static ssize_t iter_xarray_get_pages(struct iov_iter *i, | |
68fe506f | 1033 | struct page ***pages, size_t maxsize, |
7ff50620 DH |
1034 | unsigned maxpages, size_t *_start_offset) |
1035 | { | |
3cf42da3 AV |
1036 | unsigned nr, offset, count; |
1037 | pgoff_t index; | |
7ff50620 DH |
1038 | loff_t pos; |
1039 | ||
7ff50620 DH |
1040 | pos = i->xarray_start + i->iov_offset; |
1041 | index = pos >> PAGE_SHIFT; | |
1042 | offset = pos & ~PAGE_MASK; | |
1043 | *_start_offset = offset; | |
1044 | ||
3cf42da3 AV |
1045 | count = want_pages_array(pages, maxsize, offset, maxpages); |
1046 | if (!count) | |
1047 | return -ENOMEM; | |
68fe506f | 1048 | nr = iter_xarray_populate_pages(*pages, i->xarray, index, count); |
7ff50620 DH |
1049 | if (nr == 0) |
1050 | return 0; | |
1051 | ||
eba2d3d7 | 1052 | maxsize = min_t(size_t, nr * PAGE_SIZE - offset, maxsize); |
310d9d5a AV |
1053 | i->iov_offset += maxsize; |
1054 | i->count -= maxsize; | |
eba2d3d7 | 1055 | return maxsize; |
7ff50620 DH |
1056 | } |
1057 | ||
fcb14cb1 | 1058 | /* must be done on non-empty ITER_UBUF or ITER_IOVEC one */ |
dd45ab9d | 1059 | static unsigned long first_iovec_segment(const struct iov_iter *i, size_t *size) |
3d671ca6 AV |
1060 | { |
1061 | size_t skip; | |
1062 | long k; | |
1063 | ||
fcb14cb1 AV |
1064 | if (iter_is_ubuf(i)) |
1065 | return (unsigned long)i->ubuf + i->iov_offset; | |
1066 | ||
3d671ca6 | 1067 | for (k = 0, skip = i->iov_offset; k < i->nr_segs; k++, skip = 0) { |
de4f5fed JA |
1068 | const struct iovec *iov = iter_iov(i) + k; |
1069 | size_t len = iov->iov_len - skip; | |
3d671ca6 AV |
1070 | |
1071 | if (unlikely(!len)) | |
1072 | continue; | |
59dbd7d0 AV |
1073 | if (*size > len) |
1074 | *size = len; | |
de4f5fed | 1075 | return (unsigned long)iov->iov_base + skip; |
3d671ca6 AV |
1076 | } |
1077 | BUG(); // if it had been empty, we wouldn't get called | |
1078 | } | |
1079 | ||
1080 | /* must be done on non-empty ITER_BVEC one */ | |
1081 | static struct page *first_bvec_segment(const struct iov_iter *i, | |
59dbd7d0 | 1082 | size_t *size, size_t *start) |
3d671ca6 AV |
1083 | { |
1084 | struct page *page; | |
1085 | size_t skip = i->iov_offset, len; | |
1086 | ||
1087 | len = i->bvec->bv_len - skip; | |
59dbd7d0 AV |
1088 | if (*size > len) |
1089 | *size = len; | |
3d671ca6 AV |
1090 | skip += i->bvec->bv_offset; |
1091 | page = i->bvec->bv_page + skip / PAGE_SIZE; | |
dda8e5d1 | 1092 | *start = skip % PAGE_SIZE; |
3d671ca6 AV |
1093 | return page; |
1094 | } | |
1095 | ||
451c0ba9 AV |
1096 | static ssize_t __iov_iter_get_pages_alloc(struct iov_iter *i, |
1097 | struct page ***pages, size_t maxsize, | |
84bd06c6 | 1098 | unsigned int maxpages, size_t *start) |
62a8067a | 1099 | { |
f62e52d1 | 1100 | unsigned int n, gup_flags = 0; |
3d671ca6 | 1101 | |
e5393fae AV |
1102 | if (maxsize > i->count) |
1103 | maxsize = i->count; | |
451c0ba9 | 1104 | if (!maxsize) |
3d671ca6 | 1105 | return 0; |
7392ed17 AV |
1106 | if (maxsize > MAX_RW_COUNT) |
1107 | maxsize = MAX_RW_COUNT; | |
e5393fae | 1108 | |
fcb14cb1 | 1109 | if (likely(user_backed_iter(i))) { |
3d671ca6 | 1110 | unsigned long addr; |
3cf42da3 | 1111 | int res; |
e5393fae | 1112 | |
3337ab08 AG |
1113 | if (iov_iter_rw(i) != WRITE) |
1114 | gup_flags |= FOLL_WRITE; | |
1115 | if (i->nofault) | |
1116 | gup_flags |= FOLL_NOFAULT; | |
1117 | ||
dd45ab9d AV |
1118 | addr = first_iovec_segment(i, &maxsize); |
1119 | *start = addr % PAGE_SIZE; | |
1120 | addr &= PAGE_MASK; | |
3cf42da3 AV |
1121 | n = want_pages_array(pages, maxsize, *start, maxpages); |
1122 | if (!n) | |
1123 | return -ENOMEM; | |
451c0ba9 | 1124 | res = get_user_pages_fast(addr, n, gup_flags, *pages); |
814a6674 | 1125 | if (unlikely(res <= 0)) |
e5393fae | 1126 | return res; |
eba2d3d7 AV |
1127 | maxsize = min_t(size_t, maxsize, res * PAGE_SIZE - *start); |
1128 | iov_iter_advance(i, maxsize); | |
1129 | return maxsize; | |
3d671ca6 AV |
1130 | } |
1131 | if (iov_iter_is_bvec(i)) { | |
451c0ba9 | 1132 | struct page **p; |
3d671ca6 AV |
1133 | struct page *page; |
1134 | ||
59dbd7d0 | 1135 | page = first_bvec_segment(i, &maxsize, start); |
3cf42da3 AV |
1136 | n = want_pages_array(pages, maxsize, *start, maxpages); |
1137 | if (!n) | |
1138 | return -ENOMEM; | |
451c0ba9 | 1139 | p = *pages; |
dda8e5d1 | 1140 | for (int k = 0; k < n; k++) |
eba2d3d7 AV |
1141 | get_page(p[k] = page + k); |
1142 | maxsize = min_t(size_t, maxsize, n * PAGE_SIZE - *start); | |
310d9d5a AV |
1143 | i->count -= maxsize; |
1144 | i->iov_offset += maxsize; | |
1145 | if (i->iov_offset == i->bvec->bv_len) { | |
1146 | i->iov_offset = 0; | |
1147 | i->bvec++; | |
1148 | i->nr_segs--; | |
1149 | } | |
eba2d3d7 | 1150 | return maxsize; |
3d671ca6 | 1151 | } |
3d671ca6 | 1152 | if (iov_iter_is_xarray(i)) |
451c0ba9 | 1153 | return iter_xarray_get_pages(i, pages, maxsize, maxpages, start); |
3d671ca6 | 1154 | return -EFAULT; |
62a8067a | 1155 | } |
62a8067a | 1156 | |
84bd06c6 CH |
1157 | ssize_t iov_iter_get_pages2(struct iov_iter *i, struct page **pages, |
1158 | size_t maxsize, unsigned maxpages, size_t *start) | |
62a8067a | 1159 | { |
451c0ba9 | 1160 | if (!maxpages) |
3d671ca6 | 1161 | return 0; |
451c0ba9 | 1162 | BUG_ON(!pages); |
3d671ca6 | 1163 | |
84bd06c6 | 1164 | return __iov_iter_get_pages_alloc(i, &pages, maxsize, maxpages, start); |
62a8067a | 1165 | } |
eba2d3d7 | 1166 | EXPORT_SYMBOL(iov_iter_get_pages2); |
91329559 | 1167 | |
84bd06c6 CH |
1168 | ssize_t iov_iter_get_pages_alloc2(struct iov_iter *i, |
1169 | struct page ***pages, size_t maxsize, size_t *start) | |
91329559 AV |
1170 | { |
1171 | ssize_t len; | |
1172 | ||
1173 | *pages = NULL; | |
1174 | ||
84bd06c6 | 1175 | len = __iov_iter_get_pages_alloc(i, pages, maxsize, ~0U, start); |
91329559 AV |
1176 | if (len <= 0) { |
1177 | kvfree(*pages); | |
1178 | *pages = NULL; | |
1179 | } | |
1180 | return len; | |
1181 | } | |
eba2d3d7 | 1182 | EXPORT_SYMBOL(iov_iter_get_pages_alloc2); |
62a8067a | 1183 | |
a604ec7e AV |
1184 | size_t csum_and_copy_from_iter(void *addr, size_t bytes, __wsum *csum, |
1185 | struct iov_iter *i) | |
1186 | { | |
a604ec7e | 1187 | __wsum sum, next; |
a604ec7e | 1188 | sum = *csum; |
a41dad90 | 1189 | if (WARN_ON_ONCE(!i->data_source)) |
241699cd | 1190 | return 0; |
a41dad90 | 1191 | |
7baa5099 AV |
1192 | iterate_and_advance(i, bytes, base, len, off, ({ |
1193 | next = csum_and_copy_from_user(base, addr + off, len); | |
2495bdcc | 1194 | sum = csum_block_add(sum, next, off); |
7baa5099 | 1195 | next ? 0 : len; |
a604ec7e | 1196 | }), ({ |
7baa5099 | 1197 | sum = csum_and_memcpy(addr + off, base, len, sum, off); |
a604ec7e AV |
1198 | }) |
1199 | ) | |
1200 | *csum = sum; | |
1201 | return bytes; | |
1202 | } | |
1203 | EXPORT_SYMBOL(csum_and_copy_from_iter); | |
1204 | ||
52cbd23a | 1205 | size_t csum_and_copy_to_iter(const void *addr, size_t bytes, void *_csstate, |
a604ec7e AV |
1206 | struct iov_iter *i) |
1207 | { | |
52cbd23a | 1208 | struct csum_state *csstate = _csstate; |
a604ec7e | 1209 | __wsum sum, next; |
78e1f386 | 1210 | |
a41dad90 AV |
1211 | if (WARN_ON_ONCE(i->data_source)) |
1212 | return 0; | |
78e1f386 | 1213 | if (unlikely(iov_iter_is_discard(i))) { |
c67f1fd2 AV |
1214 | // can't use csum_memcpy() for that one - data is not copied |
1215 | csstate->csum = csum_block_add(csstate->csum, | |
1216 | csum_partial(addr, bytes, 0), | |
1217 | csstate->off); | |
1218 | csstate->off += bytes; | |
1219 | return bytes; | |
241699cd | 1220 | } |
6852df12 AV |
1221 | |
1222 | sum = csum_shift(csstate->csum, csstate->off); | |
3fc40265 | 1223 | iterate_and_advance(i, bytes, base, len, off, ({ |
7baa5099 | 1224 | next = csum_and_copy_to_user(addr + off, base, len); |
2495bdcc | 1225 | sum = csum_block_add(sum, next, off); |
7baa5099 | 1226 | next ? 0 : len; |
a604ec7e | 1227 | }), ({ |
7baa5099 | 1228 | sum = csum_and_memcpy(base, addr + off, len, sum, off); |
a604ec7e AV |
1229 | }) |
1230 | ) | |
594e450b AV |
1231 | csstate->csum = csum_shift(sum, csstate->off); |
1232 | csstate->off += bytes; | |
a604ec7e AV |
1233 | return bytes; |
1234 | } | |
1235 | EXPORT_SYMBOL(csum_and_copy_to_iter); | |
1236 | ||
d05f4435 SG |
1237 | size_t hash_and_copy_to_iter(const void *addr, size_t bytes, void *hashp, |
1238 | struct iov_iter *i) | |
1239 | { | |
7999096f | 1240 | #ifdef CONFIG_CRYPTO_HASH |
d05f4435 SG |
1241 | struct ahash_request *hash = hashp; |
1242 | struct scatterlist sg; | |
1243 | size_t copied; | |
1244 | ||
1245 | copied = copy_to_iter(addr, bytes, i); | |
1246 | sg_init_one(&sg, addr, copied); | |
1247 | ahash_request_set_crypt(hash, &sg, NULL, copied); | |
1248 | crypto_ahash_update(hash); | |
1249 | return copied; | |
27fad74a Y |
1250 | #else |
1251 | return 0; | |
1252 | #endif | |
d05f4435 SG |
1253 | } |
1254 | EXPORT_SYMBOL(hash_and_copy_to_iter); | |
1255 | ||
66531c65 | 1256 | static int iov_npages(const struct iov_iter *i, int maxpages) |
62a8067a | 1257 | { |
66531c65 AV |
1258 | size_t skip = i->iov_offset, size = i->count; |
1259 | const struct iovec *p; | |
e0f2dc40 AV |
1260 | int npages = 0; |
1261 | ||
de4f5fed | 1262 | for (p = iter_iov(i); size; skip = 0, p++) { |
66531c65 AV |
1263 | unsigned offs = offset_in_page(p->iov_base + skip); |
1264 | size_t len = min(p->iov_len - skip, size); | |
e0f2dc40 | 1265 | |
66531c65 AV |
1266 | if (len) { |
1267 | size -= len; | |
1268 | npages += DIV_ROUND_UP(offs + len, PAGE_SIZE); | |
1269 | if (unlikely(npages > maxpages)) | |
1270 | return maxpages; | |
1271 | } | |
1272 | } | |
1273 | return npages; | |
1274 | } | |
1275 | ||
1276 | static int bvec_npages(const struct iov_iter *i, int maxpages) | |
1277 | { | |
1278 | size_t skip = i->iov_offset, size = i->count; | |
1279 | const struct bio_vec *p; | |
1280 | int npages = 0; | |
1281 | ||
1282 | for (p = i->bvec; size; skip = 0, p++) { | |
1283 | unsigned offs = (p->bv_offset + skip) % PAGE_SIZE; | |
1284 | size_t len = min(p->bv_len - skip, size); | |
1285 | ||
1286 | size -= len; | |
1287 | npages += DIV_ROUND_UP(offs + len, PAGE_SIZE); | |
1288 | if (unlikely(npages > maxpages)) | |
1289 | return maxpages; | |
1290 | } | |
1291 | return npages; | |
1292 | } | |
1293 | ||
1294 | int iov_iter_npages(const struct iov_iter *i, int maxpages) | |
1295 | { | |
1296 | if (unlikely(!i->count)) | |
1297 | return 0; | |
fcb14cb1 AV |
1298 | if (likely(iter_is_ubuf(i))) { |
1299 | unsigned offs = offset_in_page(i->ubuf + i->iov_offset); | |
1300 | int npages = DIV_ROUND_UP(offs + i->count, PAGE_SIZE); | |
1301 | return min(npages, maxpages); | |
1302 | } | |
66531c65 AV |
1303 | /* iovec and kvec have identical layouts */ |
1304 | if (likely(iter_is_iovec(i) || iov_iter_is_kvec(i))) | |
1305 | return iov_npages(i, maxpages); | |
1306 | if (iov_iter_is_bvec(i)) | |
1307 | return bvec_npages(i, maxpages); | |
66531c65 | 1308 | if (iov_iter_is_xarray(i)) { |
e4f8df86 AV |
1309 | unsigned offset = (i->xarray_start + i->iov_offset) % PAGE_SIZE; |
1310 | int npages = DIV_ROUND_UP(offset + i->count, PAGE_SIZE); | |
66531c65 AV |
1311 | return min(npages, maxpages); |
1312 | } | |
1313 | return 0; | |
62a8067a | 1314 | } |
f67da30c | 1315 | EXPORT_SYMBOL(iov_iter_npages); |
4b8164b9 AV |
1316 | |
1317 | const void *dup_iter(struct iov_iter *new, struct iov_iter *old, gfp_t flags) | |
1318 | { | |
1319 | *new = *old; | |
00e23707 | 1320 | if (iov_iter_is_bvec(new)) |
4b8164b9 AV |
1321 | return new->bvec = kmemdup(new->bvec, |
1322 | new->nr_segs * sizeof(struct bio_vec), | |
1323 | flags); | |
fcb14cb1 | 1324 | else if (iov_iter_is_kvec(new) || iter_is_iovec(new)) |
4b8164b9 | 1325 | /* iovec and kvec have identical layout */ |
de4f5fed | 1326 | return new->__iov = kmemdup(new->__iov, |
4b8164b9 AV |
1327 | new->nr_segs * sizeof(struct iovec), |
1328 | flags); | |
fcb14cb1 | 1329 | return NULL; |
4b8164b9 AV |
1330 | } |
1331 | EXPORT_SYMBOL(dup_iter); | |
bc917be8 | 1332 | |
50f9a76e | 1333 | static __noclone int copy_compat_iovec_from_user(struct iovec *iov, |
bfdc5970 CH |
1334 | const struct iovec __user *uvec, unsigned long nr_segs) |
1335 | { | |
1336 | const struct compat_iovec __user *uiov = | |
1337 | (const struct compat_iovec __user *)uvec; | |
1338 | int ret = -EFAULT, i; | |
1339 | ||
a959a978 | 1340 | if (!user_access_begin(uiov, nr_segs * sizeof(*uiov))) |
bfdc5970 CH |
1341 | return -EFAULT; |
1342 | ||
1343 | for (i = 0; i < nr_segs; i++) { | |
1344 | compat_uptr_t buf; | |
1345 | compat_ssize_t len; | |
1346 | ||
1347 | unsafe_get_user(len, &uiov[i].iov_len, uaccess_end); | |
1348 | unsafe_get_user(buf, &uiov[i].iov_base, uaccess_end); | |
1349 | ||
1350 | /* check for compat_size_t not fitting in compat_ssize_t .. */ | |
1351 | if (len < 0) { | |
1352 | ret = -EINVAL; | |
1353 | goto uaccess_end; | |
1354 | } | |
1355 | iov[i].iov_base = compat_ptr(buf); | |
1356 | iov[i].iov_len = len; | |
1357 | } | |
1358 | ||
1359 | ret = 0; | |
1360 | uaccess_end: | |
1361 | user_access_end(); | |
1362 | return ret; | |
1363 | } | |
1364 | ||
719a937b | 1365 | static __noclone int copy_iovec_from_user(struct iovec *iov, |
487c20b0 | 1366 | const struct iovec __user *uiov, unsigned long nr_segs) |
fb041b59 | 1367 | { |
487c20b0 | 1368 | int ret = -EFAULT; |
fb041b59 | 1369 | |
487c20b0 | 1370 | if (!user_access_begin(uiov, nr_segs * sizeof(*uiov))) |
bfdc5970 | 1371 | return -EFAULT; |
fb041b59 | 1372 | |
487c20b0 LT |
1373 | do { |
1374 | void __user *buf; | |
1375 | ssize_t len; | |
1376 | ||
1377 | unsafe_get_user(len, &uiov->iov_len, uaccess_end); | |
1378 | unsafe_get_user(buf, &uiov->iov_base, uaccess_end); | |
1379 | ||
1380 | /* check for size_t not fitting in ssize_t .. */ | |
1381 | if (unlikely(len < 0)) { | |
1382 | ret = -EINVAL; | |
1383 | goto uaccess_end; | |
1384 | } | |
1385 | iov->iov_base = buf; | |
1386 | iov->iov_len = len; | |
1387 | ||
1388 | uiov++; iov++; | |
1389 | } while (--nr_segs); | |
1390 | ||
1391 | ret = 0; | |
1392 | uaccess_end: | |
1393 | user_access_end(); | |
1394 | return ret; | |
bfdc5970 CH |
1395 | } |
1396 | ||
1397 | struct iovec *iovec_from_user(const struct iovec __user *uvec, | |
1398 | unsigned long nr_segs, unsigned long fast_segs, | |
1399 | struct iovec *fast_iov, bool compat) | |
1400 | { | |
1401 | struct iovec *iov = fast_iov; | |
1402 | int ret; | |
1403 | ||
fb041b59 | 1404 | /* |
bfdc5970 CH |
1405 | * SuS says "The readv() function *may* fail if the iovcnt argument was |
1406 | * less than or equal to 0, or greater than {IOV_MAX}. Linux has | |
1407 | * traditionally returned zero for zero segments, so... | |
fb041b59 | 1408 | */ |
bfdc5970 CH |
1409 | if (nr_segs == 0) |
1410 | return iov; | |
1411 | if (nr_segs > UIO_MAXIOV) | |
1412 | return ERR_PTR(-EINVAL); | |
fb041b59 DL |
1413 | if (nr_segs > fast_segs) { |
1414 | iov = kmalloc_array(nr_segs, sizeof(struct iovec), GFP_KERNEL); | |
bfdc5970 CH |
1415 | if (!iov) |
1416 | return ERR_PTR(-ENOMEM); | |
fb041b59 | 1417 | } |
bfdc5970 | 1418 | |
487c20b0 | 1419 | if (unlikely(compat)) |
bfdc5970 CH |
1420 | ret = copy_compat_iovec_from_user(iov, uvec, nr_segs); |
1421 | else | |
1422 | ret = copy_iovec_from_user(iov, uvec, nr_segs); | |
1423 | if (ret) { | |
1424 | if (iov != fast_iov) | |
1425 | kfree(iov); | |
1426 | return ERR_PTR(ret); | |
1427 | } | |
1428 | ||
1429 | return iov; | |
1430 | } | |
1431 | ||
3b2deb0e JA |
1432 | /* |
1433 | * Single segment iovec supplied by the user, import it as ITER_UBUF. | |
1434 | */ | |
1435 | static ssize_t __import_iovec_ubuf(int type, const struct iovec __user *uvec, | |
1436 | struct iovec **iovp, struct iov_iter *i, | |
1437 | bool compat) | |
1438 | { | |
1439 | struct iovec *iov = *iovp; | |
1440 | ssize_t ret; | |
1441 | ||
1442 | if (compat) | |
1443 | ret = copy_compat_iovec_from_user(iov, uvec, 1); | |
1444 | else | |
1445 | ret = copy_iovec_from_user(iov, uvec, 1); | |
1446 | if (unlikely(ret)) | |
1447 | return ret; | |
1448 | ||
1449 | ret = import_ubuf(type, iov->iov_base, iov->iov_len, i); | |
1450 | if (unlikely(ret)) | |
1451 | return ret; | |
1452 | *iovp = NULL; | |
1453 | return i->count; | |
1454 | } | |
1455 | ||
bfdc5970 CH |
1456 | ssize_t __import_iovec(int type, const struct iovec __user *uvec, |
1457 | unsigned nr_segs, unsigned fast_segs, struct iovec **iovp, | |
1458 | struct iov_iter *i, bool compat) | |
1459 | { | |
1460 | ssize_t total_len = 0; | |
1461 | unsigned long seg; | |
1462 | struct iovec *iov; | |
1463 | ||
3b2deb0e JA |
1464 | if (nr_segs == 1) |
1465 | return __import_iovec_ubuf(type, uvec, iovp, i, compat); | |
1466 | ||
bfdc5970 CH |
1467 | iov = iovec_from_user(uvec, nr_segs, fast_segs, *iovp, compat); |
1468 | if (IS_ERR(iov)) { | |
1469 | *iovp = NULL; | |
1470 | return PTR_ERR(iov); | |
fb041b59 DL |
1471 | } |
1472 | ||
1473 | /* | |
bfdc5970 CH |
1474 | * According to the Single Unix Specification we should return EINVAL if |
1475 | * an element length is < 0 when cast to ssize_t or if the total length | |
1476 | * would overflow the ssize_t return value of the system call. | |
fb041b59 DL |
1477 | * |
1478 | * Linux caps all read/write calls to MAX_RW_COUNT, and avoids the | |
1479 | * overflow case. | |
1480 | */ | |
fb041b59 | 1481 | for (seg = 0; seg < nr_segs; seg++) { |
fb041b59 DL |
1482 | ssize_t len = (ssize_t)iov[seg].iov_len; |
1483 | ||
bfdc5970 CH |
1484 | if (!access_ok(iov[seg].iov_base, len)) { |
1485 | if (iov != *iovp) | |
1486 | kfree(iov); | |
1487 | *iovp = NULL; | |
1488 | return -EFAULT; | |
fb041b59 | 1489 | } |
bfdc5970 CH |
1490 | |
1491 | if (len > MAX_RW_COUNT - total_len) { | |
1492 | len = MAX_RW_COUNT - total_len; | |
fb041b59 DL |
1493 | iov[seg].iov_len = len; |
1494 | } | |
bfdc5970 | 1495 | total_len += len; |
fb041b59 | 1496 | } |
bfdc5970 CH |
1497 | |
1498 | iov_iter_init(i, type, iov, nr_segs, total_len); | |
1499 | if (iov == *iovp) | |
1500 | *iovp = NULL; | |
1501 | else | |
1502 | *iovp = iov; | |
1503 | return total_len; | |
fb041b59 DL |
1504 | } |
1505 | ||
ffecee4f VN |
1506 | /** |
1507 | * import_iovec() - Copy an array of &struct iovec from userspace | |
1508 | * into the kernel, check that it is valid, and initialize a new | |
1509 | * &struct iov_iter iterator to access it. | |
1510 | * | |
1511 | * @type: One of %READ or %WRITE. | |
bfdc5970 | 1512 | * @uvec: Pointer to the userspace array. |
ffecee4f VN |
1513 | * @nr_segs: Number of elements in userspace array. |
1514 | * @fast_segs: Number of elements in @iov. | |
bfdc5970 | 1515 | * @iovp: (input and output parameter) Pointer to pointer to (usually small |
ffecee4f VN |
1516 | * on-stack) kernel array. |
1517 | * @i: Pointer to iterator that will be initialized on success. | |
1518 | * | |
1519 | * If the array pointed to by *@iov is large enough to hold all @nr_segs, | |
1520 | * then this function places %NULL in *@iov on return. Otherwise, a new | |
1521 | * array will be allocated and the result placed in *@iov. This means that | |
1522 | * the caller may call kfree() on *@iov regardless of whether the small | |
1523 | * on-stack array was used or not (and regardless of whether this function | |
1524 | * returns an error or not). | |
1525 | * | |
87e5e6da | 1526 | * Return: Negative error code on error, bytes imported on success |
ffecee4f | 1527 | */ |
bfdc5970 | 1528 | ssize_t import_iovec(int type, const struct iovec __user *uvec, |
bc917be8 | 1529 | unsigned nr_segs, unsigned fast_segs, |
bfdc5970 | 1530 | struct iovec **iovp, struct iov_iter *i) |
bc917be8 | 1531 | { |
89cd35c5 CH |
1532 | return __import_iovec(type, uvec, nr_segs, fast_segs, iovp, i, |
1533 | in_compat_syscall()); | |
bc917be8 AV |
1534 | } |
1535 | EXPORT_SYMBOL(import_iovec); | |
1536 | ||
bc917be8 AV |
1537 | int import_single_range(int rw, void __user *buf, size_t len, |
1538 | struct iovec *iov, struct iov_iter *i) | |
1539 | { | |
1540 | if (len > MAX_RW_COUNT) | |
1541 | len = MAX_RW_COUNT; | |
96d4f267 | 1542 | if (unlikely(!access_ok(buf, len))) |
bc917be8 AV |
1543 | return -EFAULT; |
1544 | ||
e03ad4ee | 1545 | iov_iter_ubuf(i, rw, buf, len); |
bc917be8 AV |
1546 | return 0; |
1547 | } | |
e1267585 | 1548 | EXPORT_SYMBOL(import_single_range); |
8fb0f47a | 1549 | |
2ad9bd83 JA |
1550 | int import_ubuf(int rw, void __user *buf, size_t len, struct iov_iter *i) |
1551 | { | |
1552 | if (len > MAX_RW_COUNT) | |
1553 | len = MAX_RW_COUNT; | |
1554 | if (unlikely(!access_ok(buf, len))) | |
1555 | return -EFAULT; | |
1556 | ||
1557 | iov_iter_ubuf(i, rw, buf, len); | |
1558 | return 0; | |
1559 | } | |
1560 | ||
8fb0f47a JA |
1561 | /** |
1562 | * iov_iter_restore() - Restore a &struct iov_iter to the same state as when | |
1563 | * iov_iter_save_state() was called. | |
1564 | * | |
1565 | * @i: &struct iov_iter to restore | |
1566 | * @state: state to restore from | |
1567 | * | |
1568 | * Used after iov_iter_save_state() to bring restore @i, if operations may | |
1569 | * have advanced it. | |
1570 | * | |
1571 | * Note: only works on ITER_IOVEC, ITER_BVEC, and ITER_KVEC | |
1572 | */ | |
1573 | void iov_iter_restore(struct iov_iter *i, struct iov_iter_state *state) | |
1574 | { | |
4397a17c KB |
1575 | if (WARN_ON_ONCE(!iov_iter_is_bvec(i) && !iter_is_iovec(i) && |
1576 | !iter_is_ubuf(i)) && !iov_iter_is_kvec(i)) | |
8fb0f47a JA |
1577 | return; |
1578 | i->iov_offset = state->iov_offset; | |
1579 | i->count = state->count; | |
fcb14cb1 AV |
1580 | if (iter_is_ubuf(i)) |
1581 | return; | |
8fb0f47a JA |
1582 | /* |
1583 | * For the *vec iters, nr_segs + iov is constant - if we increment | |
1584 | * the vec, then we also decrement the nr_segs count. Hence we don't | |
1585 | * need to track both of these, just one is enough and we can deduct | |
1586 | * the other from that. ITER_KVEC and ITER_IOVEC are the same struct | |
1587 | * size, so we can just increment the iov pointer as they are unionzed. | |
1588 | * ITER_BVEC _may_ be the same size on some archs, but on others it is | |
1589 | * not. Be safe and handle it separately. | |
1590 | */ | |
1591 | BUILD_BUG_ON(sizeof(struct iovec) != sizeof(struct kvec)); | |
1592 | if (iov_iter_is_bvec(i)) | |
1593 | i->bvec -= state->nr_segs - i->nr_segs; | |
1594 | else | |
de4f5fed | 1595 | i->__iov -= state->nr_segs - i->nr_segs; |
8fb0f47a JA |
1596 | i->nr_segs = state->nr_segs; |
1597 | } | |
7d58fe73 DH |
1598 | |
1599 | /* | |
1600 | * Extract a list of contiguous pages from an ITER_XARRAY iterator. This does not | |
1601 | * get references on the pages, nor does it get a pin on them. | |
1602 | */ | |
1603 | static ssize_t iov_iter_extract_xarray_pages(struct iov_iter *i, | |
1604 | struct page ***pages, size_t maxsize, | |
1605 | unsigned int maxpages, | |
1606 | iov_iter_extraction_t extraction_flags, | |
1607 | size_t *offset0) | |
1608 | { | |
1609 | struct page *page, **p; | |
1610 | unsigned int nr = 0, offset; | |
1611 | loff_t pos = i->xarray_start + i->iov_offset; | |
1612 | pgoff_t index = pos >> PAGE_SHIFT; | |
1613 | XA_STATE(xas, i->xarray, index); | |
1614 | ||
1615 | offset = pos & ~PAGE_MASK; | |
1616 | *offset0 = offset; | |
1617 | ||
1618 | maxpages = want_pages_array(pages, maxsize, offset, maxpages); | |
1619 | if (!maxpages) | |
1620 | return -ENOMEM; | |
1621 | p = *pages; | |
1622 | ||
1623 | rcu_read_lock(); | |
1624 | for (page = xas_load(&xas); page; page = xas_next(&xas)) { | |
1625 | if (xas_retry(&xas, page)) | |
1626 | continue; | |
1627 | ||
1628 | /* Has the page moved or been split? */ | |
1629 | if (unlikely(page != xas_reload(&xas))) { | |
1630 | xas_reset(&xas); | |
1631 | continue; | |
1632 | } | |
1633 | ||
1634 | p[nr++] = find_subpage(page, xas.xa_index); | |
1635 | if (nr == maxpages) | |
1636 | break; | |
1637 | } | |
1638 | rcu_read_unlock(); | |
1639 | ||
1640 | maxsize = min_t(size_t, nr * PAGE_SIZE - offset, maxsize); | |
1641 | iov_iter_advance(i, maxsize); | |
1642 | return maxsize; | |
1643 | } | |
1644 | ||
1645 | /* | |
1646 | * Extract a list of contiguous pages from an ITER_BVEC iterator. This does | |
1647 | * not get references on the pages, nor does it get a pin on them. | |
1648 | */ | |
1649 | static ssize_t iov_iter_extract_bvec_pages(struct iov_iter *i, | |
1650 | struct page ***pages, size_t maxsize, | |
1651 | unsigned int maxpages, | |
1652 | iov_iter_extraction_t extraction_flags, | |
1653 | size_t *offset0) | |
1654 | { | |
1655 | struct page **p, *page; | |
1656 | size_t skip = i->iov_offset, offset; | |
1657 | int k; | |
1658 | ||
1659 | for (;;) { | |
1660 | if (i->nr_segs == 0) | |
1661 | return 0; | |
1662 | maxsize = min(maxsize, i->bvec->bv_len - skip); | |
1663 | if (maxsize) | |
1664 | break; | |
1665 | i->iov_offset = 0; | |
1666 | i->nr_segs--; | |
1667 | i->bvec++; | |
1668 | skip = 0; | |
1669 | } | |
1670 | ||
1671 | skip += i->bvec->bv_offset; | |
1672 | page = i->bvec->bv_page + skip / PAGE_SIZE; | |
1673 | offset = skip % PAGE_SIZE; | |
1674 | *offset0 = offset; | |
1675 | ||
1676 | maxpages = want_pages_array(pages, maxsize, offset, maxpages); | |
1677 | if (!maxpages) | |
1678 | return -ENOMEM; | |
1679 | p = *pages; | |
1680 | for (k = 0; k < maxpages; k++) | |
1681 | p[k] = page + k; | |
1682 | ||
1683 | maxsize = min_t(size_t, maxsize, maxpages * PAGE_SIZE - offset); | |
1684 | iov_iter_advance(i, maxsize); | |
1685 | return maxsize; | |
1686 | } | |
1687 | ||
1688 | /* | |
1689 | * Extract a list of virtually contiguous pages from an ITER_KVEC iterator. | |
1690 | * This does not get references on the pages, nor does it get a pin on them. | |
1691 | */ | |
1692 | static ssize_t iov_iter_extract_kvec_pages(struct iov_iter *i, | |
1693 | struct page ***pages, size_t maxsize, | |
1694 | unsigned int maxpages, | |
1695 | iov_iter_extraction_t extraction_flags, | |
1696 | size_t *offset0) | |
1697 | { | |
1698 | struct page **p, *page; | |
1699 | const void *kaddr; | |
1700 | size_t skip = i->iov_offset, offset, len; | |
1701 | int k; | |
1702 | ||
1703 | for (;;) { | |
1704 | if (i->nr_segs == 0) | |
1705 | return 0; | |
1706 | maxsize = min(maxsize, i->kvec->iov_len - skip); | |
1707 | if (maxsize) | |
1708 | break; | |
1709 | i->iov_offset = 0; | |
1710 | i->nr_segs--; | |
1711 | i->kvec++; | |
1712 | skip = 0; | |
1713 | } | |
1714 | ||
1715 | kaddr = i->kvec->iov_base + skip; | |
1716 | offset = (unsigned long)kaddr & ~PAGE_MASK; | |
1717 | *offset0 = offset; | |
1718 | ||
1719 | maxpages = want_pages_array(pages, maxsize, offset, maxpages); | |
1720 | if (!maxpages) | |
1721 | return -ENOMEM; | |
1722 | p = *pages; | |
1723 | ||
1724 | kaddr -= offset; | |
1725 | len = offset + maxsize; | |
1726 | for (k = 0; k < maxpages; k++) { | |
1727 | size_t seg = min_t(size_t, len, PAGE_SIZE); | |
1728 | ||
1729 | if (is_vmalloc_or_module_addr(kaddr)) | |
1730 | page = vmalloc_to_page(kaddr); | |
1731 | else | |
1732 | page = virt_to_page(kaddr); | |
1733 | ||
1734 | p[k] = page; | |
1735 | len -= seg; | |
1736 | kaddr += PAGE_SIZE; | |
1737 | } | |
1738 | ||
1739 | maxsize = min_t(size_t, maxsize, maxpages * PAGE_SIZE - offset); | |
1740 | iov_iter_advance(i, maxsize); | |
1741 | return maxsize; | |
1742 | } | |
1743 | ||
1744 | /* | |
1745 | * Extract a list of contiguous pages from a user iterator and get a pin on | |
1746 | * each of them. This should only be used if the iterator is user-backed | |
1747 | * (IOBUF/UBUF). | |
1748 | * | |
1749 | * It does not get refs on the pages, but the pages must be unpinned by the | |
1750 | * caller once the transfer is complete. | |
1751 | * | |
1752 | * This is safe to be used where background IO/DMA *is* going to be modifying | |
1753 | * the buffer; using a pin rather than a ref makes forces fork() to give the | |
1754 | * child a copy of the page. | |
1755 | */ | |
1756 | static ssize_t iov_iter_extract_user_pages(struct iov_iter *i, | |
1757 | struct page ***pages, | |
1758 | size_t maxsize, | |
1759 | unsigned int maxpages, | |
1760 | iov_iter_extraction_t extraction_flags, | |
1761 | size_t *offset0) | |
1762 | { | |
1763 | unsigned long addr; | |
1764 | unsigned int gup_flags = 0; | |
1765 | size_t offset; | |
1766 | int res; | |
1767 | ||
1768 | if (i->data_source == ITER_DEST) | |
1769 | gup_flags |= FOLL_WRITE; | |
1770 | if (extraction_flags & ITER_ALLOW_P2PDMA) | |
1771 | gup_flags |= FOLL_PCI_P2PDMA; | |
1772 | if (i->nofault) | |
1773 | gup_flags |= FOLL_NOFAULT; | |
1774 | ||
1775 | addr = first_iovec_segment(i, &maxsize); | |
1776 | *offset0 = offset = addr % PAGE_SIZE; | |
1777 | addr &= PAGE_MASK; | |
1778 | maxpages = want_pages_array(pages, maxsize, offset, maxpages); | |
1779 | if (!maxpages) | |
1780 | return -ENOMEM; | |
1781 | res = pin_user_pages_fast(addr, maxpages, gup_flags, *pages); | |
1782 | if (unlikely(res <= 0)) | |
1783 | return res; | |
1784 | maxsize = min_t(size_t, maxsize, res * PAGE_SIZE - offset); | |
1785 | iov_iter_advance(i, maxsize); | |
1786 | return maxsize; | |
1787 | } | |
1788 | ||
1789 | /** | |
1790 | * iov_iter_extract_pages - Extract a list of contiguous pages from an iterator | |
1791 | * @i: The iterator to extract from | |
1792 | * @pages: Where to return the list of pages | |
1793 | * @maxsize: The maximum amount of iterator to extract | |
1794 | * @maxpages: The maximum size of the list of pages | |
1795 | * @extraction_flags: Flags to qualify request | |
1796 | * @offset0: Where to return the starting offset into (*@pages)[0] | |
1797 | * | |
1798 | * Extract a list of contiguous pages from the current point of the iterator, | |
1799 | * advancing the iterator. The maximum number of pages and the maximum amount | |
1800 | * of page contents can be set. | |
1801 | * | |
1802 | * If *@pages is NULL, a page list will be allocated to the required size and | |
1803 | * *@pages will be set to its base. If *@pages is not NULL, it will be assumed | |
1804 | * that the caller allocated a page list at least @maxpages in size and this | |
1805 | * will be filled in. | |
1806 | * | |
1807 | * @extraction_flags can have ITER_ALLOW_P2PDMA set to request peer-to-peer DMA | |
1808 | * be allowed on the pages extracted. | |
1809 | * | |
1810 | * The iov_iter_extract_will_pin() function can be used to query how cleanup | |
1811 | * should be performed. | |
1812 | * | |
1813 | * Extra refs or pins on the pages may be obtained as follows: | |
1814 | * | |
1815 | * (*) If the iterator is user-backed (ITER_IOVEC/ITER_UBUF), pins will be | |
1816 | * added to the pages, but refs will not be taken. | |
1817 | * iov_iter_extract_will_pin() will return true. | |
1818 | * | |
1819 | * (*) If the iterator is ITER_KVEC, ITER_BVEC or ITER_XARRAY, the pages are | |
1820 | * merely listed; no extra refs or pins are obtained. | |
1821 | * iov_iter_extract_will_pin() will return 0. | |
1822 | * | |
1823 | * Note also: | |
1824 | * | |
1825 | * (*) Use with ITER_DISCARD is not supported as that has no content. | |
1826 | * | |
1827 | * On success, the function sets *@pages to the new pagelist, if allocated, and | |
1828 | * sets *offset0 to the offset into the first page. | |
1829 | * | |
1830 | * It may also return -ENOMEM and -EFAULT. | |
1831 | */ | |
1832 | ssize_t iov_iter_extract_pages(struct iov_iter *i, | |
1833 | struct page ***pages, | |
1834 | size_t maxsize, | |
1835 | unsigned int maxpages, | |
1836 | iov_iter_extraction_t extraction_flags, | |
1837 | size_t *offset0) | |
1838 | { | |
1839 | maxsize = min_t(size_t, min_t(size_t, maxsize, i->count), MAX_RW_COUNT); | |
1840 | if (!maxsize) | |
1841 | return 0; | |
1842 | ||
1843 | if (likely(user_backed_iter(i))) | |
1844 | return iov_iter_extract_user_pages(i, pages, maxsize, | |
1845 | maxpages, extraction_flags, | |
1846 | offset0); | |
1847 | if (iov_iter_is_kvec(i)) | |
1848 | return iov_iter_extract_kvec_pages(i, pages, maxsize, | |
1849 | maxpages, extraction_flags, | |
1850 | offset0); | |
1851 | if (iov_iter_is_bvec(i)) | |
1852 | return iov_iter_extract_bvec_pages(i, pages, maxsize, | |
1853 | maxpages, extraction_flags, | |
1854 | offset0); | |
1855 | if (iov_iter_is_xarray(i)) | |
1856 | return iov_iter_extract_xarray_pages(i, pages, maxsize, | |
1857 | maxpages, extraction_flags, | |
1858 | offset0); | |
1859 | return -EFAULT; | |
1860 | } | |
1861 | EXPORT_SYMBOL_GPL(iov_iter_extract_pages); |