| 1 | |
| 2 | * linux/fs/read_write.c
|
| 3 | *
|
| 4 | * Copyright (C) 1991, 1992 Linus Torvalds
|
| 5 |
|
| 6 |
|
| 7 | #include <linux/slab.h>
|
| 8 | #include <linux/stat.h>
|
| 9 | #include <linux/fcntl.h>
|
| 10 | #include <linux/file.h>
|
| 11 | #include <linux/uio.h>
|
| 12 | #include <linux/aio.h>
|
| 13 | #include <linux/fsnotify.h>
|
| 14 | #include <linux/security.h>
|
| 15 | #include <linux/export.h>
|
| 16 | #include <linux/syscalls.h>
|
| 17 | #include <linux/pagemap.h>
|
| 18 | #include <linux/splice.h>
|
| 19 | #include <linux/compat.h>
|
| 20 | #include "internal.h"
|
| 21 |
|
| 22 | #include <asm/uaccess.h>
|
| 23 | #include <asm/unistd.h>
|
| 24 |
|
| 25 | typedef ssize_t (*io_fn_t)(struct file *, char __user *, size_t, loff_t *);
|
| 26 | typedef ssize_t (*iov_fn_t)(struct kiocb *, const struct iovec *,
|
| 27 | unsigned long, loff_t);
|
| 28 |
|
| 29 | const struct file_operations generic_ro_fops = {
|
| 30 | .llseek = generic_file_llseek,
|
| 31 | .read = do_sync_read,
|
| 32 | .aio_read = generic_file_aio_read,
|
| 33 | .mmap = generic_file_readonly_mmap,
|
| 34 | .splice_read = generic_file_splice_read,
|
| 35 | };
|
| 36 |
|
| 37 | EXPORT_SYMBOL(generic_ro_fops);
|
| 38 |
|
| 39 | static inline int unsigned_offsets(struct file *file)
|
| 40 | {
|
| 41 | return file->f_mode & FMODE_UNSIGNED_OFFSET;
|
| 42 | }
|
| 43 |
|
| 44 | |
| 45 | * vfs_setpos - update the file offset for lseek
|
| 46 | * @file: file structure in question
|
| 47 | * @offset: file offset to seek to
|
| 48 | * @maxsize: maximum file size
|
| 49 | *
|
| 50 | * This is a low-level filesystem helper for updating the file offset to
|
| 51 | * the value specified by @offset if the given offset is valid and it is
|
| 52 | * not equal to the current file offset.
|
| 53 | *
|
| 54 | * Return the specified offset on success and -EINVAL on invalid offset.
|
| 55 |
|
| 56 | loff_t vfs_setpos(struct file *file, loff_t offset, loff_t maxsize)
|
| 57 | {
|
| 58 | if (offset < 0 && !unsigned_offsets(file))
|
| 59 | return -EINVAL;
|
| 60 | if (offset > maxsize)
|
| 61 | return -EINVAL;
|
| 62 |
|
| 63 | if (offset != file->f_pos) {
|
| 64 | file->f_pos = offset;
|
| 65 | file->f_version = 0;
|
| 66 | }
|
| 67 | return offset;
|
| 68 | }
|
| 69 | EXPORT_SYMBOL(vfs_setpos);
|
| 70 |
|
| 71 | |
| 72 | * generic_file_llseek_size - generic llseek implementation for regular files
|
| 73 | * @file: file structure to seek on
|
| 74 | * @offset: file offset to seek to
|
| 75 | * @whence: type of seek
|
| 76 | * @size: max size of this file in file system
|
| 77 | * @eof: offset used for SEEK_END position
|
| 78 | *
|
| 79 | * This is a variant of generic_file_llseek that allows passing in a custom
|
| 80 | * maximum file size and a custom EOF position, for e.g. hashed directories
|
| 81 | *
|
| 82 | * Synchronization:
|
| 83 | * SEEK_SET and SEEK_END are unsynchronized (but atomic on 64bit platforms)
|
| 84 | * SEEK_CUR is synchronized against other SEEK_CURs, but not read/writes.
|
| 85 | * read/writes behave like SEEK_SET against seeks.
|
| 86 |
|
| 87 | loff_t
|
| 88 | generic_file_llseek_size(struct file *file, loff_t offset, int whence,
|
| 89 | loff_t maxsize, loff_t eof)
|
| 90 | {
|
| 91 | switch (whence) {
|
| 92 | case SEEK_END:
|
| 93 | offset += eof;
|
| 94 | break;
|
| 95 | case SEEK_CUR:
|
| 96 | |
| 97 | * Here we special-case the lseek(fd, 0, SEEK_CUR)
|
| 98 | * position-querying operation. Avoid rewriting the "same"
|
| 99 | * f_pos value back to the file because a concurrent read(),
|
| 100 | * write() or lseek() might have altered it
|
| 101 |
|
| 102 | if (offset == 0)
|
| 103 | return file->f_pos;
|
| 104 | |
| 105 | * f_lock protects against read/modify/write race with other
|
| 106 | * SEEK_CURs. Note that parallel writes and reads behave
|
| 107 | * like SEEK_SET.
|
| 108 |
|
| 109 | spin_lock(&file->f_lock);
|
| 110 | offset = vfs_setpos(file, file->f_pos + offset, maxsize);
|
| 111 | spin_unlock(&file->f_lock);
|
| 112 | return offset;
|
| 113 | case SEEK_DATA:
|
| 114 | |
| 115 | * In the generic case the entire file is data, so as long as
|
| 116 | * offset isn't at the end of the file then the offset is data.
|
| 117 |
|
| 118 | if (offset >= eof)
|
| 119 | return -ENXIO;
|
| 120 | break;
|
| 121 | case SEEK_HOLE:
|
| 122 | |
| 123 | * There is a virtual hole at the end of the file, so as long as
|
| 124 | * offset isn't i_size or larger, return i_size.
|
| 125 |
|
| 126 | if (offset >= eof)
|
| 127 | return -ENXIO;
|
| 128 | offset = eof;
|
| 129 | break;
|
| 130 | }
|
| 131 |
|
| 132 | return vfs_setpos(file, offset, maxsize);
|
| 133 | }
|
| 134 | EXPORT_SYMBOL(generic_file_llseek_size);
|
| 135 |
|
| 136 | |
| 137 | * generic_file_llseek - generic llseek implementation for regular files
|
| 138 | * @file: file structure to seek on
|
| 139 | * @offset: file offset to seek to
|
| 140 | * @whence: type of seek
|
| 141 | *
|
| 142 | * This is a generic implemenation of ->llseek useable for all normal local
|
| 143 | * filesystems. It just updates the file offset to the value specified by
|
| 144 | * @offset and @whence.
|
| 145 |
|
| 146 | loff_t generic_file_llseek(struct file *file, loff_t offset, int whence)
|
| 147 | {
|
| 148 | struct inode *inode = file->f_mapping->host;
|
| 149 |
|
| 150 | return generic_file_llseek_size(file, offset, whence,
|
| 151 | inode->i_sb->s_maxbytes,
|
| 152 | i_size_read(inode));
|
| 153 | }
|
| 154 | EXPORT_SYMBOL(generic_file_llseek);
|
| 155 |
|
| 156 | |
| 157 | * fixed_size_llseek - llseek implementation for fixed-sized devices
|
| 158 | * @file: file structure to seek on
|
| 159 | * @offset: file offset to seek to
|
| 160 | * @whence: type of seek
|
| 161 | * @size: size of the file
|
| 162 | *
|
| 163 |
|
| 164 | loff_t fixed_size_llseek(struct file *file, loff_t offset, int whence, loff_t size)
|
| 165 | {
|
| 166 | switch (whence) {
|
| 167 | case SEEK_SET: case SEEK_CUR: case SEEK_END:
|
| 168 | return generic_file_llseek_size(file, offset, whence,
|
| 169 | size, size);
|
| 170 | default:
|
| 171 | return -EINVAL;
|
| 172 | }
|
| 173 | }
|
| 174 | EXPORT_SYMBOL(fixed_size_llseek);
|
| 175 |
|
| 176 | |
| 177 | * noop_llseek - No Operation Performed llseek implementation
|
| 178 | * @file: file structure to seek on
|
| 179 | * @offset: file offset to seek to
|
| 180 | * @whence: type of seek
|
| 181 | *
|
| 182 | * This is an implementation of ->llseek useable for the rare special case when
|
| 183 | * userspace expects the seek to succeed but the (device) file is actually not
|
| 184 | * able to perform the seek. In this case you use noop_llseek() instead of
|
| 185 | * falling back to the default implementation of ->llseek.
|
| 186 |
|
| 187 | loff_t noop_llseek(struct file *file, loff_t offset, int whence)
|
| 188 | {
|
| 189 | return file->f_pos;
|
| 190 | }
|
| 191 | EXPORT_SYMBOL(noop_llseek);
|
| 192 |
|
| 193 | loff_t no_llseek(struct file *file, loff_t offset, int whence)
|
| 194 | {
|
| 195 | return -ESPIPE;
|
| 196 | }
|
| 197 | EXPORT_SYMBOL(no_llseek);
|
| 198 |
|
| 199 | loff_t default_llseek(struct file *file, loff_t offset, int whence)
|
| 200 | {
|
| 201 | struct inode *inode = file_inode(file);
|
| 202 | loff_t retval;
|
| 203 |
|
| 204 | mutex_lock(&inode->i_mutex);
|
| 205 | switch (whence) {
|
| 206 | case SEEK_END:
|
| 207 | offset += i_size_read(inode);
|
| 208 | break;
|
| 209 | case SEEK_CUR:
|
| 210 | if (offset == 0) {
|
| 211 | retval = file->f_pos;
|
| 212 | goto out;
|
| 213 | }
|
| 214 | offset += file->f_pos;
|
| 215 | break;
|
| 216 | case SEEK_DATA:
|
| 217 | |
| 218 | * In the generic case the entire file is data, so as
|
| 219 | * long as offset isn't at the end of the file then the
|
| 220 | * offset is data.
|
| 221 |
|
| 222 | if (offset >= inode->i_size) {
|
| 223 | retval = -ENXIO;
|
| 224 | goto out;
|
| 225 | }
|
| 226 | break;
|
| 227 | case SEEK_HOLE:
|
| 228 | |
| 229 | * There is a virtual hole at the end of the file, so
|
| 230 | * as long as offset isn't i_size or larger, return
|
| 231 | * i_size.
|
| 232 |
|
| 233 | if (offset >= inode->i_size) {
|
| 234 | retval = -ENXIO;
|
| 235 | goto out;
|
| 236 | }
|
| 237 | offset = inode->i_size;
|
| 238 | break;
|
| 239 | }
|
| 240 | retval = -EINVAL;
|
| 241 | if (offset >= 0 || unsigned_offsets(file)) {
|
| 242 | if (offset != file->f_pos) {
|
| 243 | file->f_pos = offset;
|
| 244 | file->f_version = 0;
|
| 245 | }
|
| 246 | retval = offset;
|
| 247 | }
|
| 248 | out:
|
| 249 | mutex_unlock(&inode->i_mutex);
|
| 250 | return retval;
|
| 251 | }
|
| 252 | EXPORT_SYMBOL(default_llseek);
|
| 253 |
|
| 254 | loff_t vfs_llseek(struct file *file, loff_t offset, int whence)
|
| 255 | {
|
| 256 | loff_t (*fn)(struct file *, loff_t, int);
|
| 257 |
|
| 258 | fn = no_llseek;
|
| 259 | if (file->f_mode & FMODE_LSEEK) {
|
| 260 | if (file->f_op->llseek)
|
| 261 | fn = file->f_op->llseek;
|
| 262 | }
|
| 263 | return fn(file, offset, whence);
|
| 264 | }
|
| 265 | EXPORT_SYMBOL(vfs_llseek);
|
| 266 |
|
| 267 | SYSCALL_DEFINE3(lseek, unsigned int, fd, off_t, offset, unsigned int, whence)
|
| 268 | {
|
| 269 | off_t retval;
|
| 270 | struct fd f = fdget(fd);
|
| 271 | if (!f.file)
|
| 272 | return -EBADF;
|
| 273 |
|
| 274 | retval = -EINVAL;
|
| 275 | if (whence <= SEEK_MAX) {
|
| 276 | loff_t res = vfs_llseek(f.file, offset, whence);
|
| 277 | retval = res;
|
| 278 | if (res != (loff_t)retval)
|
| 279 | retval = -EOVERFLOW;
|
| 280 | }
|
| 281 | fdput(f);
|
| 282 | return retval;
|
| 283 | }
|
| 284 |
|
| 285 | #ifdef CONFIG_COMPAT
|
| 286 | COMPAT_SYSCALL_DEFINE3(lseek, unsigned int, fd, compat_off_t, offset, unsigned int, whence)
|
| 287 | {
|
| 288 | return sys_lseek(fd, offset, whence);
|
| 289 | }
|
| 290 | #endif
|
| 291 |
|
| 292 | #ifdef __ARCH_WANT_SYS_LLSEEK
|
| 293 | SYSCALL_DEFINE5(llseek, unsigned int, fd, unsigned long, offset_high,
|
| 294 | unsigned long, offset_low, loff_t __user *, result,
|
| 295 | unsigned int, whence)
|
| 296 | {
|
| 297 | int retval;
|
| 298 | struct fd f = fdget(fd);
|
| 299 | loff_t offset;
|
| 300 |
|
| 301 | if (!f.file)
|
| 302 | return -EBADF;
|
| 303 |
|
| 304 | retval = -EINVAL;
|
| 305 | if (whence > SEEK_MAX)
|
| 306 | goto out_putf;
|
| 307 |
|
| 308 | offset = vfs_llseek(f.file, ((loff_t) offset_high << 32) | offset_low,
|
| 309 | whence);
|
| 310 |
|
| 311 | retval = (int)offset;
|
| 312 | if (offset >= 0) {
|
| 313 | retval = -EFAULT;
|
| 314 | if (!copy_to_user(result, &offset, sizeof(offset)))
|
| 315 | retval = 0;
|
| 316 | }
|
| 317 | out_putf:
|
| 318 | fdput(f);
|
| 319 | return retval;
|
| 320 | }
|
| 321 | #endif
|
| 322 |
|
| 323 | |
| 324 | * rw_verify_area doesn't like huge counts. We limit
|
| 325 | * them to something that fits in "int" so that others
|
| 326 | * won't have to do range checks all the time.
|
| 327 |
|
| 328 | int rw_verify_area(int read_write, struct file *file, const loff_t *ppos, size_t count)
|
| 329 | {
|
| 330 | struct inode *inode;
|
| 331 | loff_t pos;
|
| 332 | int retval = -EINVAL;
|
| 333 |
|
| 334 | inode = file_inode(file);
|
| 335 | if (unlikely((ssize_t) count < 0))
|
| 336 | return retval;
|
| 337 | pos = *ppos;
|
| 338 | if (unlikely(pos < 0)) {
|
| 339 | if (!unsigned_offsets(file))
|
| 340 | return retval;
|
| 341 | if (count >= -pos)
|
| 342 | return -EOVERFLOW;
|
| 343 | } else if (unlikely((loff_t) (pos + count) < 0)) {
|
| 344 | if (!unsigned_offsets(file))
|
| 345 | return retval;
|
| 346 | }
|
| 347 |
|
| 348 | if (unlikely(inode->i_flock && mandatory_lock(inode))) {
|
| 349 | retval = locks_mandatory_area(
|
| 350 | read_write == READ ? FLOCK_VERIFY_READ : FLOCK_VERIFY_WRITE,
|
| 351 | inode, file, pos, count);
|
| 352 | if (retval < 0)
|
| 353 | return retval;
|
| 354 | }
|
| 355 | retval = security_file_permission(file,
|
| 356 | read_write == READ ? MAY_READ : MAY_WRITE);
|
| 357 | if (retval)
|
| 358 | return retval;
|
| 359 | return count > MAX_RW_COUNT ? MAX_RW_COUNT : count;
|
| 360 | }
|
| 361 |
|
| 362 | ssize_t do_sync_read(struct file *filp, char __user *buf, size_t len, loff_t *ppos)
|
| 363 | {
|
| 364 | struct iovec iov = { .iov_base = buf, .iov_len = len };
|
| 365 | struct kiocb kiocb;
|
| 366 | ssize_t ret;
|
| 367 |
|
| 368 | init_sync_kiocb(&kiocb, filp);
|
| 369 | kiocb.ki_pos = *ppos;
|
| 370 | kiocb.ki_nbytes = len;
|
| 371 |
|
| 372 | ret = filp->f_op->aio_read(&kiocb, &iov, 1, kiocb.ki_pos);
|
| 373 | if (-EIOCBQUEUED == ret)
|
| 374 | ret = wait_on_sync_kiocb(&kiocb);
|
| 375 | *ppos = kiocb.ki_pos;
|
| 376 | return ret;
|
| 377 | }
|
| 378 |
|
| 379 | EXPORT_SYMBOL(do_sync_read);
|
| 380 |
|
| 381 | ssize_t vfs_read(struct file *file, char __user *buf, size_t count, loff_t *pos)
|
| 382 | {
|
| 383 | ssize_t ret;
|
| 384 |
|
| 385 | if (!(file->f_mode & FMODE_READ))
|
| 386 | return -EBADF;
|
| 387 | if (!file->f_op->read && !file->f_op->aio_read)
|
| 388 | return -EINVAL;
|
| 389 | if (unlikely(!access_ok(VERIFY_WRITE, buf, count)))
|
| 390 | return -EFAULT;
|
| 391 |
|
| 392 | ret = rw_verify_area(READ, file, pos, count);
|
| 393 | if (ret >= 0) {
|
| 394 | count = ret;
|
| 395 | if (file->f_op->read)
|
| 396 | ret = file->f_op->read(file, buf, count, pos);
|
| 397 | else
|
| 398 | ret = do_sync_read(file, buf, count, pos);
|
| 399 | if (ret > 0) {
|
| 400 | fsnotify_access(file);
|
| 401 | add_rchar(current, ret);
|
| 402 | }
|
| 403 | inc_syscr(current);
|
| 404 | }
|
| 405 |
|
| 406 | return ret;
|
| 407 | }
|
| 408 |
|
| 409 | EXPORT_SYMBOL(vfs_read);
|
| 410 |
|
| 411 | ssize_t do_sync_write(struct file *filp, const char __user *buf, size_t len, loff_t *ppos)
|
| 412 | {
|
| 413 | struct iovec iov = { .iov_base = (void __user *)buf, .iov_len = len };
|
| 414 | struct kiocb kiocb;
|
| 415 | ssize_t ret;
|
| 416 |
|
| 417 | init_sync_kiocb(&kiocb, filp);
|
| 418 | kiocb.ki_pos = *ppos;
|
| 419 | kiocb.ki_nbytes = len;
|
| 420 |
|
| 421 | ret = filp->f_op->aio_write(&kiocb, &iov, 1, kiocb.ki_pos);
|
| 422 | if (-EIOCBQUEUED == ret)
|
| 423 | ret = wait_on_sync_kiocb(&kiocb);
|
| 424 | *ppos = kiocb.ki_pos;
|
| 425 | return ret;
|
| 426 | }
|
| 427 |
|
| 428 | EXPORT_SYMBOL(do_sync_write);
|
| 429 |
|
| 430 | ssize_t __kernel_write(struct file *file, const char *buf, size_t count, loff_t *pos)
|
| 431 | {
|
| 432 | mm_segment_t old_fs;
|
| 433 | const char __user *p;
|
| 434 | ssize_t ret;
|
| 435 |
|
| 436 | if (!file->f_op->write && !file->f_op->aio_write)
|
| 437 | return -EINVAL;
|
| 438 |
|
| 439 | old_fs = get_fs();
|
| 440 | set_fs(get_ds());
|
| 441 | p = (__force const char __user *)buf;
|
| 442 | if (count > MAX_RW_COUNT)
|
| 443 | count = MAX_RW_COUNT;
|
| 444 | if (file->f_op->write)
|
| 445 | ret = file->f_op->write(file, p, count, pos);
|
| 446 | else
|
| 447 | ret = do_sync_write(file, p, count, pos);
|
| 448 | set_fs(old_fs);
|
| 449 | if (ret > 0) {
|
| 450 | fsnotify_modify(file);
|
| 451 | add_wchar(current, ret);
|
| 452 | }
|
| 453 | inc_syscw(current);
|
| 454 | return ret;
|
| 455 | }
|
| 456 |
|
| 457 | ssize_t vfs_write(struct file *file, const char __user *buf, size_t count, loff_t *pos)
|
| 458 | {
|
| 459 | ssize_t ret;
|
| 460 |
|
| 461 | if (!(file->f_mode & FMODE_WRITE))
|
| 462 | return -EBADF;
|
| 463 | if (!file->f_op->write && !file->f_op->aio_write)
|
| 464 | return -EINVAL;
|
| 465 | if (unlikely(!access_ok(VERIFY_READ, buf, count)))
|
| 466 | return -EFAULT;
|
| 467 |
|
| 468 | ret = rw_verify_area(WRITE, file, pos, count);
|
| 469 | if (ret >= 0) {
|
| 470 | count = ret;
|
| 471 | file_start_write(file);
|
| 472 | if (file->f_op->write)
|
| 473 | ret = file->f_op->write(file, buf, count, pos);
|
| 474 | else
|
| 475 | ret = do_sync_write(file, buf, count, pos);
|
| 476 | if (ret > 0) {
|
| 477 | fsnotify_modify(file);
|
| 478 | add_wchar(current, ret);
|
| 479 | }
|
| 480 | inc_syscw(current);
|
| 481 | file_end_write(file);
|
| 482 | }
|
| 483 |
|
| 484 | return ret;
|
| 485 | }
|
| 486 |
|
| 487 | EXPORT_SYMBOL(vfs_write);
|
| 488 |
|
| 489 | static inline loff_t file_pos_read(struct file *file)
|
| 490 | {
|
| 491 | return file->f_pos;
|
| 492 | }
|
| 493 |
|
| 494 | static inline void file_pos_write(struct file *file, loff_t pos)
|
| 495 | {
|
| 496 | file->f_pos = pos;
|
| 497 | }
|
| 498 |
|
| 499 | SYSCALL_DEFINE3(read, unsigned int, fd, char __user *, buf, size_t, count)
|
| 500 | {
|
| 501 | struct fd f = fdget(fd);
|
| 502 | ssize_t ret = -EBADF;
|
| 503 |
|
| 504 | if (f.file) {
|
| 505 | loff_t pos = file_pos_read(f.file);
|
| 506 | ret = vfs_read(f.file, buf, count, &pos);
|
| 507 | if (ret >= 0)
|
| 508 | file_pos_write(f.file, pos);
|
| 509 | fdput(f);
|
| 510 | }
|
| 511 | return ret;
|
| 512 | }
|
| 513 |
|
| 514 | SYSCALL_DEFINE3(write, unsigned int, fd, const char __user *, buf,
|
| 515 | size_t, count)
|
| 516 | {
|
| 517 | struct fd f = fdget(fd);
|
| 518 | ssize_t ret = -EBADF;
|
| 519 |
|
| 520 | if (f.file) {
|
| 521 | loff_t pos = file_pos_read(f.file);
|
| 522 | ret = vfs_write(f.file, buf, count, &pos);
|
| 523 | if (ret >= 0)
|
| 524 | file_pos_write(f.file, pos);
|
| 525 | fdput(f);
|
| 526 | }
|
| 527 |
|
| 528 | return ret;
|
| 529 | }
|
| 530 |
|
| 531 | SYSCALL_DEFINE4(pread64, unsigned int, fd, char __user *, buf,
|
| 532 | size_t, count, loff_t, pos)
|
| 533 | {
|
| 534 | struct fd f;
|
| 535 | ssize_t ret = -EBADF;
|
| 536 |
|
| 537 | if (pos < 0)
|
| 538 | return -EINVAL;
|
| 539 |
|
| 540 | f = fdget(fd);
|
| 541 | if (f.file) {
|
| 542 | ret = -ESPIPE;
|
| 543 | if (f.file->f_mode & FMODE_PREAD)
|
| 544 | ret = vfs_read(f.file, buf, count, &pos);
|
| 545 | fdput(f);
|
| 546 | }
|
| 547 |
|
| 548 | return ret;
|
| 549 | }
|
| 550 |
|
| 551 | SYSCALL_DEFINE4(pwrite64, unsigned int, fd, const char __user *, buf,
|
| 552 | size_t, count, loff_t, pos)
|
| 553 | {
|
| 554 | struct fd f;
|
| 555 | ssize_t ret = -EBADF;
|
| 556 |
|
| 557 | if (pos < 0)
|
| 558 | return -EINVAL;
|
| 559 |
|
| 560 | f = fdget(fd);
|
| 561 | if (f.file) {
|
| 562 | ret = -ESPIPE;
|
| 563 | if (f.file->f_mode & FMODE_PWRITE)
|
| 564 | ret = vfs_write(f.file, buf, count, &pos);
|
| 565 | fdput(f);
|
| 566 | }
|
| 567 |
|
| 568 | return ret;
|
| 569 | }
|
| 570 |
|
| 571 | |
| 572 | * Reduce an iovec's length in-place. Return the resulting number of segments
|
| 573 |
|
| 574 | unsigned long iov_shorten(struct iovec *iov, unsigned long nr_segs, size_t to)
|
| 575 | {
|
| 576 | unsigned long seg = 0;
|
| 577 | size_t len = 0;
|
| 578 |
|
| 579 | while (seg < nr_segs) {
|
| 580 | seg++;
|
| 581 | if (len + iov->iov_len >= to) {
|
| 582 | iov->iov_len = to - len;
|
| 583 | break;
|
| 584 | }
|
| 585 | len += iov->iov_len;
|
| 586 | iov++;
|
| 587 | }
|
| 588 | return seg;
|
| 589 | }
|
| 590 | EXPORT_SYMBOL(iov_shorten);
|
| 591 |
|
| 592 | static ssize_t do_sync_readv_writev(struct file *filp, const struct iovec *iov,
|
| 593 | unsigned long nr_segs, size_t len, loff_t *ppos, iov_fn_t fn)
|
| 594 | {
|
| 595 | struct kiocb kiocb;
|
| 596 | ssize_t ret;
|
| 597 |
|
| 598 | init_sync_kiocb(&kiocb, filp);
|
| 599 | kiocb.ki_pos = *ppos;
|
| 600 | kiocb.ki_nbytes = len;
|
| 601 |
|
| 602 | ret = fn(&kiocb, iov, nr_segs, kiocb.ki_pos);
|
| 603 | if (ret == -EIOCBQUEUED)
|
| 604 | ret = wait_on_sync_kiocb(&kiocb);
|
| 605 | *ppos = kiocb.ki_pos;
|
| 606 | return ret;
|
| 607 | }
|
| 608 |
|
| 609 |
|
| 610 | static ssize_t do_loop_readv_writev(struct file *filp, struct iovec *iov,
|
| 611 | unsigned long nr_segs, loff_t *ppos, io_fn_t fn)
|
| 612 | {
|
| 613 | struct iovec *vector = iov;
|
| 614 | ssize_t ret = 0;
|
| 615 |
|
| 616 | while (nr_segs > 0) {
|
| 617 | void __user *base;
|
| 618 | size_t len;
|
| 619 | ssize_t nr;
|
| 620 |
|
| 621 | base = vector->iov_base;
|
| 622 | len = vector->iov_len;
|
| 623 | vector++;
|
| 624 | nr_segs--;
|
| 625 |
|
| 626 | nr = fn(filp, base, len, ppos);
|
| 627 |
|
| 628 | if (nr < 0) {
|
| 629 | if (!ret)
|
| 630 | ret = nr;
|
| 631 | break;
|
| 632 | }
|
| 633 | ret += nr;
|
| 634 | if (nr != len)
|
| 635 | break;
|
| 636 | }
|
| 637 |
|
| 638 | return ret;
|
| 639 | }
|
| 640 |
|
| 641 |
|
| 642 | #define vrfy_dir(type) ((type) == READ ? VERIFY_WRITE : VERIFY_READ)
|
| 643 |
|
| 644 | ssize_t rw_copy_check_uvector(int type, const struct iovec __user * uvector,
|
| 645 | unsigned long nr_segs, unsigned long fast_segs,
|
| 646 | struct iovec *fast_pointer,
|
| 647 | struct iovec **ret_pointer)
|
| 648 | {
|
| 649 | unsigned long seg;
|
| 650 | ssize_t ret;
|
| 651 | struct iovec *iov = fast_pointer;
|
| 652 |
|
| 653 | |
| 654 | * SuS says "The readv() function *may* fail if the iovcnt argument
|
| 655 | * was less than or equal to 0, or greater than {IOV_MAX}. Linux has
|
| 656 | * traditionally returned zero for zero segments, so...
|
| 657 |
|
| 658 | if (nr_segs == 0) {
|
| 659 | ret = 0;
|
| 660 | goto out;
|
| 661 | }
|
| 662 |
|
| 663 | |
| 664 | * First get the "struct iovec" from user memory and
|
| 665 | * verify all the pointers
|
| 666 |
|
| 667 | if (nr_segs > UIO_MAXIOV) {
|
| 668 | ret = -EINVAL;
|
| 669 | goto out;
|
| 670 | }
|
| 671 | if (nr_segs > fast_segs) {
|
| 672 | iov = kmalloc(nr_segs*sizeof(struct iovec), GFP_KERNEL);
|
| 673 | if (iov == NULL) {
|
| 674 | ret = -ENOMEM;
|
| 675 | goto out;
|
| 676 | }
|
| 677 | }
|
| 678 | if (copy_from_user(iov, uvector, nr_segs*sizeof(*uvector))) {
|
| 679 | ret = -EFAULT;
|
| 680 | goto out;
|
| 681 | }
|
| 682 |
|
| 683 | |
| 684 | * According to the Single Unix Specification we should return EINVAL
|
| 685 | * if an element length is < 0 when cast to ssize_t or if the
|
| 686 | * total length would overflow the ssize_t return value of the
|
| 687 | * system call.
|
| 688 | *
|
| 689 | * Linux caps all read/write calls to MAX_RW_COUNT, and avoids the
|
| 690 | * overflow case.
|
| 691 |
|
| 692 | ret = 0;
|
| 693 | for (seg = 0; seg < nr_segs; seg++) {
|
| 694 | void __user *buf = iov[seg].iov_base;
|
| 695 | ssize_t len = (ssize_t)iov[seg].iov_len;
|
| 696 |
|
| 697 | |
| 698 | * it's about to overflow ssize_t
|
| 699 | if (len < 0) {
|
| 700 | ret = -EINVAL;
|
| 701 | goto out;
|
| 702 | }
|
| 703 | if (type >= 0
|
| 704 | && unlikely(!access_ok(vrfy_dir(type), buf, len))) {
|
| 705 | ret = -EFAULT;
|
| 706 | goto out;
|
| 707 | }
|
| 708 | if (len > MAX_RW_COUNT - ret) {
|
| 709 | len = MAX_RW_COUNT - ret;
|
| 710 | iov[seg].iov_len = len;
|
| 711 | }
|
| 712 | ret += len;
|
| 713 | }
|
| 714 | out:
|
| 715 | *ret_pointer = iov;
|
| 716 | return ret;
|
| 717 | }
|
| 718 |
|
| 719 | static ssize_t do_readv_writev(int type, struct file *file,
|
| 720 | const struct iovec __user * uvector,
|
| 721 | unsigned long nr_segs, loff_t *pos)
|
| 722 | {
|
| 723 | size_t tot_len;
|
| 724 | struct iovec iovstack[UIO_FASTIOV];
|
| 725 | struct iovec *iov = iovstack;
|
| 726 | ssize_t ret;
|
| 727 | io_fn_t fn;
|
| 728 | iov_fn_t fnv;
|
| 729 |
|
| 730 | ret = rw_copy_check_uvector(type, uvector, nr_segs,
|
| 731 | ARRAY_SIZE(iovstack), iovstack, &iov);
|
| 732 | if (ret <= 0)
|
| 733 | goto out;
|
| 734 |
|
| 735 | tot_len = ret;
|
| 736 | ret = rw_verify_area(type, file, pos, tot_len);
|
| 737 | if (ret < 0)
|
| 738 | goto out;
|
| 739 |
|
| 740 | fnv = NULL;
|
| 741 | if (type == READ) {
|
| 742 | fn = file->f_op->read;
|
| 743 | fnv = file->f_op->aio_read;
|
| 744 | } else {
|
| 745 | fn = (io_fn_t)file->f_op->write;
|
| 746 | fnv = file->f_op->aio_write;
|
| 747 | file_start_write(file);
|
| 748 | }
|
| 749 |
|
| 750 | if (fnv)
|
| 751 | ret = do_sync_readv_writev(file, iov, nr_segs, tot_len,
|
| 752 | pos, fnv);
|
| 753 | else
|
| 754 | ret = do_loop_readv_writev(file, iov, nr_segs, pos, fn);
|
| 755 |
|
| 756 | if (type != READ)
|
| 757 | file_end_write(file);
|
| 758 |
|
| 759 | out:
|
| 760 | if (iov != iovstack)
|
| 761 | kfree(iov);
|
| 762 | if ((ret + (type == READ)) > 0) {
|
| 763 | if (type == READ)
|
| 764 | fsnotify_access(file);
|
| 765 | else
|
| 766 | fsnotify_modify(file);
|
| 767 | }
|
| 768 | return ret;
|
| 769 | }
|
| 770 |
|
| 771 | ssize_t vfs_readv(struct file *file, const struct iovec __user *vec,
|
| 772 | unsigned long vlen, loff_t *pos)
|
| 773 | {
|
| 774 | if (!(file->f_mode & FMODE_READ))
|
| 775 | return -EBADF;
|
| 776 | if (!file->f_op->aio_read && !file->f_op->read)
|
| 777 | return -EINVAL;
|
| 778 |
|
| 779 | return do_readv_writev(READ, file, vec, vlen, pos);
|
| 780 | }
|
| 781 |
|
| 782 | EXPORT_SYMBOL(vfs_readv);
|
| 783 |
|
| 784 | ssize_t vfs_writev(struct file *file, const struct iovec __user *vec,
|
| 785 | unsigned long vlen, loff_t *pos)
|
| 786 | {
|
| 787 | if (!(file->f_mode & FMODE_WRITE))
|
| 788 | return -EBADF;
|
| 789 | if (!file->f_op->aio_write && !file->f_op->write)
|
| 790 | return -EINVAL;
|
| 791 |
|
| 792 | return do_readv_writev(WRITE, file, vec, vlen, pos);
|
| 793 | }
|
| 794 |
|
| 795 | EXPORT_SYMBOL(vfs_writev);
|
| 796 |
|
| 797 | SYSCALL_DEFINE3(readv, unsigned long, fd, const struct iovec __user *, vec,
|
| 798 | unsigned long, vlen)
|
| 799 | {
|
| 800 | struct fd f = fdget(fd);
|
| 801 | ssize_t ret = -EBADF;
|
| 802 |
|
| 803 | if (f.file) {
|
| 804 | loff_t pos = file_pos_read(f.file);
|
| 805 | ret = vfs_readv(f.file, vec, vlen, &pos);
|
| 806 | if (ret >= 0)
|
| 807 | file_pos_write(f.file, pos);
|
| 808 | fdput(f);
|
| 809 | }
|
| 810 |
|
| 811 | if (ret > 0)
|
| 812 | add_rchar(current, ret);
|
| 813 | inc_syscr(current);
|
| 814 | return ret;
|
| 815 | }
|
| 816 |
|
| 817 | SYSCALL_DEFINE3(writev, unsigned long, fd, const struct iovec __user *, vec,
|
| 818 | unsigned long, vlen)
|
| 819 | {
|
| 820 | struct fd f = fdget(fd);
|
| 821 | ssize_t ret = -EBADF;
|
| 822 |
|
| 823 | if (f.file) {
|
| 824 | loff_t pos = file_pos_read(f.file);
|
| 825 | ret = vfs_writev(f.file, vec, vlen, &pos);
|
| 826 | if (ret >= 0)
|
| 827 | file_pos_write(f.file, pos);
|
| 828 | fdput(f);
|
| 829 | }
|
| 830 |
|
| 831 | if (ret > 0)
|
| 832 | add_wchar(current, ret);
|
| 833 | inc_syscw(current);
|
| 834 | return ret;
|
| 835 | }
|
| 836 |
|
| 837 | static inline loff_t pos_from_hilo(unsigned long high, unsigned long low)
|
| 838 | {
|
| 839 | #define HALF_LONG_BITS (BITS_PER_LONG / 2)
|
| 840 | return (((loff_t)high << HALF_LONG_BITS) << HALF_LONG_BITS) | low;
|
| 841 | }
|
| 842 |
|
| 843 | SYSCALL_DEFINE5(preadv, unsigned long, fd, const struct iovec __user *, vec,
|
| 844 | unsigned long, vlen, unsigned long, pos_l, unsigned long, pos_h)
|
| 845 | {
|
| 846 | loff_t pos = pos_from_hilo(pos_h, pos_l);
|
| 847 | struct fd f;
|
| 848 | ssize_t ret = -EBADF;
|
| 849 |
|
| 850 | if (pos < 0)
|
| 851 | return -EINVAL;
|
| 852 |
|
| 853 | f = fdget(fd);
|
| 854 | if (f.file) {
|
| 855 | ret = -ESPIPE;
|
| 856 | if (f.file->f_mode & FMODE_PREAD)
|
| 857 | ret = vfs_readv(f.file, vec, vlen, &pos);
|
| 858 | fdput(f);
|
| 859 | }
|
| 860 |
|
| 861 | if (ret > 0)
|
| 862 | add_rchar(current, ret);
|
| 863 | inc_syscr(current);
|
| 864 | return ret;
|
| 865 | }
|
| 866 |
|
| 867 | SYSCALL_DEFINE5(pwritev, unsigned long, fd, const struct iovec __user *, vec,
|
| 868 | unsigned long, vlen, unsigned long, pos_l, unsigned long, pos_h)
|
| 869 | {
|
| 870 | loff_t pos = pos_from_hilo(pos_h, pos_l);
|
| 871 | struct fd f;
|
| 872 | ssize_t ret = -EBADF;
|
| 873 |
|
| 874 | if (pos < 0)
|
| 875 | return -EINVAL;
|
| 876 |
|
| 877 | f = fdget(fd);
|
| 878 | if (f.file) {
|
| 879 | ret = -ESPIPE;
|
| 880 | if (f.file->f_mode & FMODE_PWRITE)
|
| 881 | ret = vfs_writev(f.file, vec, vlen, &pos);
|
| 882 | fdput(f);
|
| 883 | }
|
| 884 |
|
| 885 | if (ret > 0)
|
| 886 | add_wchar(current, ret);
|
| 887 | inc_syscw(current);
|
| 888 | return ret;
|
| 889 | }
|
| 890 |
|
| 891 | #ifdef CONFIG_COMPAT
|
| 892 |
|
| 893 | static ssize_t compat_do_readv_writev(int type, struct file *file,
|
| 894 | const struct compat_iovec __user *uvector,
|
| 895 | unsigned long nr_segs, loff_t *pos)
|
| 896 | {
|
| 897 | compat_ssize_t tot_len;
|
| 898 | struct iovec iovstack[UIO_FASTIOV];
|
| 899 | struct iovec *iov = iovstack;
|
| 900 | ssize_t ret;
|
| 901 | io_fn_t fn;
|
| 902 | iov_fn_t fnv;
|
| 903 |
|
| 904 | ret = -EFAULT;
|
| 905 | if (!access_ok(VERIFY_READ, uvector, nr_segs*sizeof(*uvector)))
|
| 906 | goto out;
|
| 907 |
|
| 908 | ret = compat_rw_copy_check_uvector(type, uvector, nr_segs,
|
| 909 | UIO_FASTIOV, iovstack, &iov);
|
| 910 | if (ret <= 0)
|
| 911 | goto out;
|
| 912 |
|
| 913 | tot_len = ret;
|
| 914 | ret = rw_verify_area(type, file, pos, tot_len);
|
| 915 | if (ret < 0)
|
| 916 | goto out;
|
| 917 |
|
| 918 | fnv = NULL;
|
| 919 | if (type == READ) {
|
| 920 | fn = file->f_op->read;
|
| 921 | fnv = file->f_op->aio_read;
|
| 922 | } else {
|
| 923 | fn = (io_fn_t)file->f_op->write;
|
| 924 | fnv = file->f_op->aio_write;
|
| 925 | file_start_write(file);
|
| 926 | }
|
| 927 |
|
| 928 | if (fnv)
|
| 929 | ret = do_sync_readv_writev(file, iov, nr_segs, tot_len,
|
| 930 | pos, fnv);
|
| 931 | else
|
| 932 | ret = do_loop_readv_writev(file, iov, nr_segs, pos, fn);
|
| 933 |
|
| 934 | if (type != READ)
|
| 935 | file_end_write(file);
|
| 936 |
|
| 937 | out:
|
| 938 | if (iov != iovstack)
|
| 939 | kfree(iov);
|
| 940 | if ((ret + (type == READ)) > 0) {
|
| 941 | if (type == READ)
|
| 942 | fsnotify_access(file);
|
| 943 | else
|
| 944 | fsnotify_modify(file);
|
| 945 | }
|
| 946 | return ret;
|
| 947 | }
|
| 948 |
|
| 949 | static size_t compat_readv(struct file *file,
|
| 950 | const struct compat_iovec __user *vec,
|
| 951 | unsigned long vlen, loff_t *pos)
|
| 952 | {
|
| 953 | ssize_t ret = -EBADF;
|
| 954 |
|
| 955 | if (!(file->f_mode & FMODE_READ))
|
| 956 | goto out;
|
| 957 |
|
| 958 | ret = -EINVAL;
|
| 959 | if (!file->f_op->aio_read && !file->f_op->read)
|
| 960 | goto out;
|
| 961 |
|
| 962 | ret = compat_do_readv_writev(READ, file, vec, vlen, pos);
|
| 963 |
|
| 964 | out:
|
| 965 | if (ret > 0)
|
| 966 | add_rchar(current, ret);
|
| 967 | inc_syscr(current);
|
| 968 | return ret;
|
| 969 | }
|
| 970 |
|
| 971 | COMPAT_SYSCALL_DEFINE3(readv, unsigned long, fd,
|
| 972 | const struct compat_iovec __user *,vec,
|
| 973 | unsigned long, vlen)
|
| 974 | {
|
| 975 | struct fd f = fdget(fd);
|
| 976 | ssize_t ret;
|
| 977 | loff_t pos;
|
| 978 |
|
| 979 | if (!f.file)
|
| 980 | return -EBADF;
|
| 981 | pos = f.file->f_pos;
|
| 982 | ret = compat_readv(f.file, vec, vlen, &pos);
|
| 983 | if (ret >= 0)
|
| 984 | f.file->f_pos = pos;
|
| 985 | fdput(f);
|
| 986 | return ret;
|
| 987 | }
|
| 988 |
|
| 989 | COMPAT_SYSCALL_DEFINE4(preadv64, unsigned long, fd,
|
| 990 | const struct compat_iovec __user *,vec,
|
| 991 | unsigned long, vlen, loff_t, pos)
|
| 992 | {
|
| 993 | struct fd f;
|
| 994 | ssize_t ret;
|
| 995 |
|
| 996 | if (pos < 0)
|
| 997 | return -EINVAL;
|
| 998 | f = fdget(fd);
|
| 999 | if (!f.file)
|
| 1000 | return -EBADF;
|
| 1001 | ret = -ESPIPE;
|
| 1002 | if (f.file->f_mode & FMODE_PREAD)
|
| 1003 | ret = compat_readv(f.file, vec, vlen, &pos);
|
| 1004 | fdput(f);
|
| 1005 | return ret;
|
| 1006 | }
|
| 1007 |
|
| 1008 | COMPAT_SYSCALL_DEFINE5(preadv, unsigned long, fd,
|
| 1009 | const struct compat_iovec __user *,vec,
|
| 1010 | unsigned long, vlen, u32, pos_low, u32, pos_high)
|
| 1011 | {
|
| 1012 | loff_t pos = ((loff_t)pos_high << 32) | pos_low;
|
| 1013 | return compat_sys_preadv64(fd, vec, vlen, pos);
|
| 1014 | }
|
| 1015 |
|
| 1016 | static size_t compat_writev(struct file *file,
|
| 1017 | const struct compat_iovec __user *vec,
|
| 1018 | unsigned long vlen, loff_t *pos)
|
| 1019 | {
|
| 1020 | ssize_t ret = -EBADF;
|
| 1021 |
|
| 1022 | if (!(file->f_mode & FMODE_WRITE))
|
| 1023 | goto out;
|
| 1024 |
|
| 1025 | ret = -EINVAL;
|
| 1026 | if (!file->f_op->aio_write && !file->f_op->write)
|
| 1027 | goto out;
|
| 1028 |
|
| 1029 | ret = compat_do_readv_writev(WRITE, file, vec, vlen, pos);
|
| 1030 |
|
| 1031 | out:
|
| 1032 | if (ret > 0)
|
| 1033 | add_wchar(current, ret);
|
| 1034 | inc_syscw(current);
|
| 1035 | return ret;
|
| 1036 | }
|
| 1037 |
|
| 1038 | COMPAT_SYSCALL_DEFINE3(writev, unsigned long, fd,
|
| 1039 | const struct compat_iovec __user *, vec,
|
| 1040 | unsigned long, vlen)
|
| 1041 | {
|
| 1042 | struct fd f = fdget(fd);
|
| 1043 | ssize_t ret;
|
| 1044 | loff_t pos;
|
| 1045 |
|
| 1046 | if (!f.file)
|
| 1047 | return -EBADF;
|
| 1048 | pos = f.file->f_pos;
|
| 1049 | ret = compat_writev(f.file, vec, vlen, &pos);
|
| 1050 | if (ret >= 0)
|
| 1051 | f.file->f_pos = pos;
|
| 1052 | fdput(f);
|
| 1053 | return ret;
|
| 1054 | }
|
| 1055 |
|
| 1056 | COMPAT_SYSCALL_DEFINE4(pwritev64, unsigned long, fd,
|
| 1057 | const struct compat_iovec __user *,vec,
|
| 1058 | unsigned long, vlen, loff_t, pos)
|
| 1059 | {
|
| 1060 | struct fd f;
|
| 1061 | ssize_t ret;
|
| 1062 |
|
| 1063 | if (pos < 0)
|
| 1064 | return -EINVAL;
|
| 1065 | f = fdget(fd);
|
| 1066 | if (!f.file)
|
| 1067 | return -EBADF;
|
| 1068 | ret = -ESPIPE;
|
| 1069 | if (f.file->f_mode & FMODE_PWRITE)
|
| 1070 | ret = compat_writev(f.file, vec, vlen, &pos);
|
| 1071 | fdput(f);
|
| 1072 | return ret;
|
| 1073 | }
|
| 1074 |
|
| 1075 | COMPAT_SYSCALL_DEFINE5(pwritev, unsigned long, fd,
|
| 1076 | const struct compat_iovec __user *,vec,
|
| 1077 | unsigned long, vlen, u32, pos_low, u32, pos_high)
|
| 1078 | {
|
| 1079 | loff_t pos = ((loff_t)pos_high << 32) | pos_low;
|
| 1080 | return compat_sys_pwritev64(fd, vec, vlen, pos);
|
| 1081 | }
|
| 1082 | #endif
|
| 1083 |
|
| 1084 | static ssize_t do_sendfile(int out_fd, int in_fd, loff_t *ppos,
|
| 1085 | size_t count, loff_t max)
|
| 1086 | {
|
| 1087 | struct fd in, out;
|
| 1088 | struct inode *in_inode, *out_inode;
|
| 1089 | loff_t pos;
|
| 1090 | loff_t out_pos;
|
| 1091 | ssize_t retval;
|
| 1092 | int fl;
|
| 1093 |
|
| 1094 | |
| 1095 | * Get input file, and verify that it is ok..
|
| 1096 |
|
| 1097 | retval = -EBADF;
|
| 1098 | in = fdget(in_fd);
|
| 1099 | if (!in.file)
|
| 1100 | goto out;
|
| 1101 | if (!(in.file->f_mode & FMODE_READ))
|
| 1102 | goto fput_in;
|
| 1103 | retval = -ESPIPE;
|
| 1104 | if (!ppos) {
|
| 1105 | pos = in.file->f_pos;
|
| 1106 | } else {
|
| 1107 | pos = *ppos;
|
| 1108 | if (!(in.file->f_mode & FMODE_PREAD))
|
| 1109 | goto fput_in;
|
| 1110 | }
|
| 1111 | retval = rw_verify_area(READ, in.file, &pos, count);
|
| 1112 | if (retval < 0)
|
| 1113 | goto fput_in;
|
| 1114 | count = retval;
|
| 1115 |
|
| 1116 | |
| 1117 | * Get output file, and verify that it is ok..
|
| 1118 |
|
| 1119 | retval = -EBADF;
|
| 1120 | out = fdget(out_fd);
|
| 1121 | if (!out.file)
|
| 1122 | goto fput_in;
|
| 1123 | if (!(out.file->f_mode & FMODE_WRITE))
|
| 1124 | goto fput_out;
|
| 1125 | retval = -EINVAL;
|
| 1126 | in_inode = file_inode(in.file);
|
| 1127 | out_inode = file_inode(out.file);
|
| 1128 | out_pos = out.file->f_pos;
|
| 1129 | retval = rw_verify_area(WRITE, out.file, &out_pos, count);
|
| 1130 | if (retval < 0)
|
| 1131 | goto fput_out;
|
| 1132 | count = retval;
|
| 1133 |
|
| 1134 | if (!max)
|
| 1135 | max = min(in_inode->i_sb->s_maxbytes, out_inode->i_sb->s_maxbytes);
|
| 1136 |
|
| 1137 | if (unlikely(pos + count > max)) {
|
| 1138 | retval = -EOVERFLOW;
|
| 1139 | if (pos >= max)
|
| 1140 | goto fput_out;
|
| 1141 | count = max - pos;
|
| 1142 | }
|
| 1143 |
|
| 1144 | fl = 0;
|
| 1145 | #if 0 |
| 1146 | /*
|
| 1147 | * We need to debate whether we can enable this or not. The
|
| 1148 | * man page documents EAGAIN return for the output at least,
|
| 1149 | * and the application is arguably buggy if it doesn't expect
|
| 1150 | * EAGAIN on a non-blocking file descriptor.
|
| 1151 | */
|
| 1152 | if (in.file->f_flags & O_NONBLOCK)
|
| 1153 | fl = SPLICE_F_NONBLOCK;
|
| 1154 | #endif
|
| 1155 | file_start_write(out.file);
|
| 1156 | retval = do_splice_direct(in.file, &pos, out.file, &out_pos, count, fl);
|
| 1157 | file_end_write(out.file);
|
| 1158 |
|
| 1159 | if (retval > 0) {
|
| 1160 | add_rchar(current, retval);
|
| 1161 | add_wchar(current, retval);
|
| 1162 | fsnotify_access(in.file);
|
| 1163 | fsnotify_modify(out.file);
|
| 1164 | out.file->f_pos = out_pos;
|
| 1165 | if (ppos)
|
| 1166 | *ppos = pos;
|
| 1167 | else
|
| 1168 | in.file->f_pos = pos;
|
| 1169 | }
|
| 1170 |
|
| 1171 | inc_syscr(current);
|
| 1172 | inc_syscw(current);
|
| 1173 | if (pos > max)
|
| 1174 | retval = -EOVERFLOW;
|
| 1175 |
|
| 1176 | fput_out:
|
| 1177 | fdput(out);
|
| 1178 | fput_in:
|
| 1179 | fdput(in);
|
| 1180 | out:
|
| 1181 | return retval;
|
| 1182 | }
|
| 1183 |
|
| 1184 | SYSCALL_DEFINE4(sendfile, int, out_fd, int, in_fd, off_t __user *, offset, size_t, count)
|
| 1185 | {
|
| 1186 | loff_t pos;
|
| 1187 | off_t off;
|
| 1188 | ssize_t ret;
|
| 1189 |
|
| 1190 | if (offset) {
|
| 1191 | if (unlikely(get_user(off, offset)))
|
| 1192 | return -EFAULT;
|
| 1193 | pos = off;
|
| 1194 | ret = do_sendfile(out_fd, in_fd, &pos, count, MAX_NON_LFS);
|
| 1195 | if (unlikely(put_user(pos, offset)))
|
| 1196 | return -EFAULT;
|
| 1197 | return ret;
|
| 1198 | }
|
| 1199 |
|
| 1200 | return do_sendfile(out_fd, in_fd, NULL, count, 0);
|
| 1201 | }
|
| 1202 |
|
| 1203 | SYSCALL_DEFINE4(sendfile64, int, out_fd, int, in_fd, loff_t __user *, offset, size_t, count)
|
| 1204 | {
|
| 1205 | loff_t pos;
|
| 1206 | ssize_t ret;
|
| 1207 |
|
| 1208 | if (offset) {
|
| 1209 | if (unlikely(copy_from_user(&pos, offset, sizeof(loff_t))))
|
| 1210 | return -EFAULT;
|
| 1211 | ret = do_sendfile(out_fd, in_fd, &pos, count, 0);
|
| 1212 | if (unlikely(put_user(pos, offset)))
|
| 1213 | return -EFAULT;
|
| 1214 | return ret;
|
| 1215 | }
|
| 1216 |
|
| 1217 | return do_sendfile(out_fd, in_fd, NULL, count, 0);
|
| 1218 | }
|
| 1219 |
|
| 1220 | #ifdef CONFIG_COMPAT
|
| 1221 | COMPAT_SYSCALL_DEFINE4(sendfile, int, out_fd, int, in_fd,
|
| 1222 | compat_off_t __user *, offset, compat_size_t, count)
|
| 1223 | {
|
| 1224 | loff_t pos;
|
| 1225 | off_t off;
|
| 1226 | ssize_t ret;
|
| 1227 |
|
| 1228 | if (offset) {
|
| 1229 | if (unlikely(get_user(off, offset)))
|
| 1230 | return -EFAULT;
|
| 1231 | pos = off;
|
| 1232 | ret = do_sendfile(out_fd, in_fd, &pos, count, MAX_NON_LFS);
|
| 1233 | if (unlikely(put_user(pos, offset)))
|
| 1234 | return -EFAULT;
|
| 1235 | return ret;
|
| 1236 | }
|
| 1237 |
|
| 1238 | return do_sendfile(out_fd, in_fd, NULL, count, 0);
|
| 1239 | }
|
| 1240 |
|
| 1241 | COMPAT_SYSCALL_DEFINE4(sendfile64, int, out_fd, int, in_fd,
|
| 1242 | compat_loff_t __user *, offset, compat_size_t, count)
|
| 1243 | {
|
| 1244 | loff_t pos;
|
| 1245 | ssize_t ret;
|
| 1246 |
|
| 1247 | if (offset) {
|
| 1248 | if (unlikely(copy_from_user(&pos, offset, sizeof(loff_t))))
|
| 1249 | return -EFAULT;
|
| 1250 | ret = do_sendfile(out_fd, in_fd, &pos, count, 0);
|
| 1251 | if (unlikely(put_user(pos, offset)))
|
| 1252 | return -EFAULT;
|
| 1253 | return ret;
|
| 1254 | }
|
| 1255 |
|
| 1256 | return do_sendfile(out_fd, in_fd, NULL, count, 0);
|
| 1257 | }
|
| 1258 | #endif
|