1 // SPDX-License-Identifier: GPL-2.0
3 * linux/fs/read_write.c
5 * Copyright (C) 1991, 1992 Linus Torvalds
8 #include <linux/slab.h>
9 #include <linux/stat.h>
10 #include <linux/sched/xacct.h>
11 #include <linux/fcntl.h>
12 #include <linux/file.h>
13 #include <linux/uio.h>
14 #include <linux/fsnotify.h>
15 #include <linux/security.h>
16 #include <linux/export.h>
17 #include <linux/syscalls.h>
18 #include <linux/pagemap.h>
19 #include <linux/splice.h>
20 #include <linux/compat.h>
21 #include <linux/mount.h>
25 #include <linux/uaccess.h>
26 #include <asm/unistd.h>
28 const struct file_operations generic_ro_fops = {
29 .llseek = generic_file_llseek,
30 .read_iter = generic_file_read_iter,
31 .mmap = generic_file_readonly_mmap,
32 .splice_read = generic_file_splice_read,
35 EXPORT_SYMBOL(generic_ro_fops);
37 static inline bool unsigned_offsets(struct file *file)
39 return file->f_mode & FMODE_UNSIGNED_OFFSET;
43 * vfs_setpos - update the file offset for lseek
44 * @file: file structure in question
45 * @offset: file offset to seek to
46 * @maxsize: maximum file size
48 * This is a low-level filesystem helper for updating the file offset to
49 * the value specified by @offset if the given offset is valid and it is
50 * not equal to the current file offset.
52 * Return the specified offset on success and -EINVAL on invalid offset.
54 loff_t vfs_setpos(struct file *file, loff_t offset, loff_t maxsize)
56 if (offset < 0 && !unsigned_offsets(file))
61 if (offset != file->f_pos) {
67 EXPORT_SYMBOL(vfs_setpos);
70 * generic_file_llseek_size - generic llseek implementation for regular files
71 * @file: file structure to seek on
72 * @offset: file offset to seek to
73 * @whence: type of seek
74 * @size: max size of this file in file system
75 * @eof: offset used for SEEK_END position
77 * This is a variant of generic_file_llseek that allows passing in a custom
78 * maximum file size and a custom EOF position, for e.g. hashed directories
81 * SEEK_SET and SEEK_END are unsynchronized (but atomic on 64bit platforms)
82 * SEEK_CUR is synchronized against other SEEK_CURs, but not read/writes.
83 * read/writes behave like SEEK_SET against seeks.
86 generic_file_llseek_size(struct file *file, loff_t offset, int whence,
87 loff_t maxsize, loff_t eof)
95 * Here we special-case the lseek(fd, 0, SEEK_CUR)
96 * position-querying operation. Avoid rewriting the "same"
97 * f_pos value back to the file because a concurrent read(),
98 * write() or lseek() might have altered it
103 * f_lock protects against read/modify/write race with other
104 * SEEK_CURs. Note that parallel writes and reads behave
107 spin_lock(&file->f_lock);
108 offset = vfs_setpos(file, file->f_pos + offset, maxsize);
109 spin_unlock(&file->f_lock);
113 * In the generic case the entire file is data, so as long as
114 * offset isn't at the end of the file then the offset is data.
116 if ((unsigned long long)offset >= eof)
121 * There is a virtual hole at the end of the file, so as long as
122 * offset isn't i_size or larger, return i_size.
124 if ((unsigned long long)offset >= eof)
130 return vfs_setpos(file, offset, maxsize);
132 EXPORT_SYMBOL(generic_file_llseek_size);
135 * generic_file_llseek - generic llseek implementation for regular files
136 * @file: file structure to seek on
137 * @offset: file offset to seek to
138 * @whence: type of seek
140 * This is a generic implemenation of ->llseek useable for all normal local
141 * filesystems. It just updates the file offset to the value specified by
142 * @offset and @whence.
144 loff_t generic_file_llseek(struct file *file, loff_t offset, int whence)
146 struct inode *inode = file->f_mapping->host;
148 return generic_file_llseek_size(file, offset, whence,
149 inode->i_sb->s_maxbytes,
152 EXPORT_SYMBOL(generic_file_llseek);
155 * fixed_size_llseek - llseek implementation for fixed-sized devices
156 * @file: file structure to seek on
157 * @offset: file offset to seek to
158 * @whence: type of seek
159 * @size: size of the file
162 loff_t fixed_size_llseek(struct file *file, loff_t offset, int whence, loff_t size)
165 case SEEK_SET: case SEEK_CUR: case SEEK_END:
166 return generic_file_llseek_size(file, offset, whence,
172 EXPORT_SYMBOL(fixed_size_llseek);
175 * no_seek_end_llseek - llseek implementation for fixed-sized devices
176 * @file: file structure to seek on
177 * @offset: file offset to seek to
178 * @whence: type of seek
181 loff_t no_seek_end_llseek(struct file *file, loff_t offset, int whence)
184 case SEEK_SET: case SEEK_CUR:
185 return generic_file_llseek_size(file, offset, whence,
191 EXPORT_SYMBOL(no_seek_end_llseek);
194 * no_seek_end_llseek_size - llseek implementation for fixed-sized devices
195 * @file: file structure to seek on
196 * @offset: file offset to seek to
197 * @whence: type of seek
198 * @size: maximal offset allowed
201 loff_t no_seek_end_llseek_size(struct file *file, loff_t offset, int whence, loff_t size)
204 case SEEK_SET: case SEEK_CUR:
205 return generic_file_llseek_size(file, offset, whence,
211 EXPORT_SYMBOL(no_seek_end_llseek_size);
214 * noop_llseek - No Operation Performed llseek implementation
215 * @file: file structure to seek on
216 * @offset: file offset to seek to
217 * @whence: type of seek
219 * This is an implementation of ->llseek useable for the rare special case when
220 * userspace expects the seek to succeed but the (device) file is actually not
221 * able to perform the seek. In this case you use noop_llseek() instead of
222 * falling back to the default implementation of ->llseek.
224 loff_t noop_llseek(struct file *file, loff_t offset, int whence)
228 EXPORT_SYMBOL(noop_llseek);
230 loff_t no_llseek(struct file *file, loff_t offset, int whence)
234 EXPORT_SYMBOL(no_llseek);
236 loff_t default_llseek(struct file *file, loff_t offset, int whence)
238 struct inode *inode = file_inode(file);
244 offset += i_size_read(inode);
248 retval = file->f_pos;
251 offset += file->f_pos;
255 * In the generic case the entire file is data, so as
256 * long as offset isn't at the end of the file then the
259 if (offset >= inode->i_size) {
266 * There is a virtual hole at the end of the file, so
267 * as long as offset isn't i_size or larger, return
270 if (offset >= inode->i_size) {
274 offset = inode->i_size;
278 if (offset >= 0 || unsigned_offsets(file)) {
279 if (offset != file->f_pos) {
280 file->f_pos = offset;
289 EXPORT_SYMBOL(default_llseek);
291 loff_t vfs_llseek(struct file *file, loff_t offset, int whence)
293 loff_t (*fn)(struct file *, loff_t, int);
296 if (file->f_mode & FMODE_LSEEK) {
297 if (file->f_op->llseek)
298 fn = file->f_op->llseek;
300 return fn(file, offset, whence);
302 EXPORT_SYMBOL(vfs_llseek);
304 static off_t ksys_lseek(unsigned int fd, off_t offset, unsigned int whence)
307 struct fd f = fdget_pos(fd);
312 if (whence <= SEEK_MAX) {
313 loff_t res = vfs_llseek(f.file, offset, whence);
315 if (res != (loff_t)retval)
316 retval = -EOVERFLOW; /* LFS: should only happen on 32 bit platforms */
322 SYSCALL_DEFINE3(lseek, unsigned int, fd, off_t, offset, unsigned int, whence)
324 return ksys_lseek(fd, offset, whence);
328 COMPAT_SYSCALL_DEFINE3(lseek, unsigned int, fd, compat_off_t, offset, unsigned int, whence)
330 return ksys_lseek(fd, offset, whence);
334 #if !defined(CONFIG_64BIT) || defined(CONFIG_COMPAT) || \
335 defined(__ARCH_WANT_SYS_LLSEEK)
336 SYSCALL_DEFINE5(llseek, unsigned int, fd, unsigned long, offset_high,
337 unsigned long, offset_low, loff_t __user *, result,
338 unsigned int, whence)
341 struct fd f = fdget_pos(fd);
348 if (whence > SEEK_MAX)
351 offset = vfs_llseek(f.file, ((loff_t) offset_high << 32) | offset_low,
354 retval = (int)offset;
357 if (!copy_to_user(result, &offset, sizeof(offset)))
366 int rw_verify_area(int read_write, struct file *file, const loff_t *ppos, size_t count)
368 if (unlikely((ssize_t) count < 0))
374 if (unlikely(pos < 0)) {
375 if (!unsigned_offsets(file))
377 if (count >= -pos) /* both values are in 0..LLONG_MAX */
379 } else if (unlikely((loff_t) (pos + count) < 0)) {
380 if (!unsigned_offsets(file))
385 return security_file_permission(file,
386 read_write == READ ? MAY_READ : MAY_WRITE);
388 EXPORT_SYMBOL(rw_verify_area);
390 static ssize_t new_sync_read(struct file *filp, char __user *buf, size_t len, loff_t *ppos)
392 struct iovec iov = { .iov_base = buf, .iov_len = len };
394 struct iov_iter iter;
397 init_sync_kiocb(&kiocb, filp);
398 kiocb.ki_pos = (ppos ? *ppos : 0);
399 iov_iter_init(&iter, READ, &iov, 1, len);
401 ret = call_read_iter(filp, &kiocb, &iter);
402 BUG_ON(ret == -EIOCBQUEUED);
404 *ppos = kiocb.ki_pos;
408 static int warn_unsupported(struct file *file, const char *op)
411 "kernel %s not supported for file %pD4 (pid: %d comm: %.20s)\n",
412 op, file, current->pid, current->comm);
416 ssize_t __kernel_read(struct file *file, void *buf, size_t count, loff_t *pos)
420 .iov_len = min_t(size_t, count, MAX_RW_COUNT),
423 struct iov_iter iter;
426 if (WARN_ON_ONCE(!(file->f_mode & FMODE_READ)))
428 if (!(file->f_mode & FMODE_CAN_READ))
431 * Also fail if ->read_iter and ->read are both wired up as that
432 * implies very convoluted semantics.
434 if (unlikely(!file->f_op->read_iter || file->f_op->read))
435 return warn_unsupported(file, "read");
437 init_sync_kiocb(&kiocb, file);
438 kiocb.ki_pos = pos ? *pos : 0;
439 iov_iter_kvec(&iter, READ, &iov, 1, iov.iov_len);
440 ret = file->f_op->read_iter(&kiocb, &iter);
444 fsnotify_access(file);
445 add_rchar(current, ret);
451 ssize_t kernel_read(struct file *file, void *buf, size_t count, loff_t *pos)
455 ret = rw_verify_area(READ, file, pos, count);
458 return __kernel_read(file, buf, count, pos);
460 EXPORT_SYMBOL(kernel_read);
462 ssize_t vfs_read(struct file *file, char __user *buf, size_t count, loff_t *pos)
466 if (!(file->f_mode & FMODE_READ))
468 if (!(file->f_mode & FMODE_CAN_READ))
470 if (unlikely(!access_ok(buf, count)))
473 ret = rw_verify_area(READ, file, pos, count);
476 if (count > MAX_RW_COUNT)
477 count = MAX_RW_COUNT;
479 if (file->f_op->read)
480 ret = file->f_op->read(file, buf, count, pos);
481 else if (file->f_op->read_iter)
482 ret = new_sync_read(file, buf, count, pos);
486 fsnotify_access(file);
487 add_rchar(current, ret);
493 static ssize_t new_sync_write(struct file *filp, const char __user *buf, size_t len, loff_t *ppos)
495 struct iovec iov = { .iov_base = (void __user *)buf, .iov_len = len };
497 struct iov_iter iter;
500 init_sync_kiocb(&kiocb, filp);
501 kiocb.ki_pos = (ppos ? *ppos : 0);
502 iov_iter_init(&iter, WRITE, &iov, 1, len);
504 ret = call_write_iter(filp, &kiocb, &iter);
505 BUG_ON(ret == -EIOCBQUEUED);
507 *ppos = kiocb.ki_pos;
511 /* caller is responsible for file_start_write/file_end_write */
512 ssize_t __kernel_write(struct file *file, const void *buf, size_t count, loff_t *pos)
515 .iov_base = (void *)buf,
516 .iov_len = min_t(size_t, count, MAX_RW_COUNT),
519 struct iov_iter iter;
522 if (WARN_ON_ONCE(!(file->f_mode & FMODE_WRITE)))
524 if (!(file->f_mode & FMODE_CAN_WRITE))
527 * Also fail if ->write_iter and ->write are both wired up as that
528 * implies very convoluted semantics.
530 if (unlikely(!file->f_op->write_iter || file->f_op->write))
531 return warn_unsupported(file, "write");
533 init_sync_kiocb(&kiocb, file);
534 kiocb.ki_pos = pos ? *pos : 0;
535 iov_iter_kvec(&iter, WRITE, &iov, 1, iov.iov_len);
536 ret = file->f_op->write_iter(&kiocb, &iter);
540 fsnotify_modify(file);
541 add_wchar(current, ret);
547 * This "EXPORT_SYMBOL_GPL()" is more of a "EXPORT_SYMBOL_DONTUSE()",
548 * but autofs is one of the few internal kernel users that actually
549 * wants this _and_ can be built as a module. So we need to export
550 * this symbol for autofs, even though it really isn't appropriate
551 * for any other kernel modules.
553 EXPORT_SYMBOL_GPL(__kernel_write);
555 ssize_t kernel_write(struct file *file, const void *buf, size_t count,
560 ret = rw_verify_area(WRITE, file, pos, count);
564 file_start_write(file);
565 ret = __kernel_write(file, buf, count, pos);
566 file_end_write(file);
569 EXPORT_SYMBOL(kernel_write);
571 ssize_t vfs_write(struct file *file, const char __user *buf, size_t count, loff_t *pos)
575 if (!(file->f_mode & FMODE_WRITE))
577 if (!(file->f_mode & FMODE_CAN_WRITE))
579 if (unlikely(!access_ok(buf, count)))
582 ret = rw_verify_area(WRITE, file, pos, count);
585 if (count > MAX_RW_COUNT)
586 count = MAX_RW_COUNT;
587 file_start_write(file);
588 if (file->f_op->write)
589 ret = file->f_op->write(file, buf, count, pos);
590 else if (file->f_op->write_iter)
591 ret = new_sync_write(file, buf, count, pos);
595 fsnotify_modify(file);
596 add_wchar(current, ret);
599 file_end_write(file);
603 /* file_ppos returns &file->f_pos or NULL if file is stream */
604 static inline loff_t *file_ppos(struct file *file)
606 return file->f_mode & FMODE_STREAM ? NULL : &file->f_pos;
609 ssize_t ksys_read(unsigned int fd, char __user *buf, size_t count)
611 struct fd f = fdget_pos(fd);
612 ssize_t ret = -EBADF;
615 loff_t pos, *ppos = file_ppos(f.file);
620 ret = vfs_read(f.file, buf, count, ppos);
621 if (ret >= 0 && ppos)
628 SYSCALL_DEFINE3(read, unsigned int, fd, char __user *, buf, size_t, count)
630 return ksys_read(fd, buf, count);
633 ssize_t ksys_write(unsigned int fd, const char __user *buf, size_t count)
635 struct fd f = fdget_pos(fd);
636 ssize_t ret = -EBADF;
639 loff_t pos, *ppos = file_ppos(f.file);
644 ret = vfs_write(f.file, buf, count, ppos);
645 if (ret >= 0 && ppos)
653 SYSCALL_DEFINE3(write, unsigned int, fd, const char __user *, buf,
656 return ksys_write(fd, buf, count);
659 ssize_t ksys_pread64(unsigned int fd, char __user *buf, size_t count,
663 ssize_t ret = -EBADF;
671 if (f.file->f_mode & FMODE_PREAD)
672 ret = vfs_read(f.file, buf, count, &pos);
679 SYSCALL_DEFINE4(pread64, unsigned int, fd, char __user *, buf,
680 size_t, count, loff_t, pos)
682 return ksys_pread64(fd, buf, count, pos);
685 #if defined(CONFIG_COMPAT) && defined(__ARCH_WANT_COMPAT_PREAD64)
686 COMPAT_SYSCALL_DEFINE5(pread64, unsigned int, fd, char __user *, buf,
687 size_t, count, compat_arg_u64_dual(pos))
689 return ksys_pread64(fd, buf, count, compat_arg_u64_glue(pos));
693 ssize_t ksys_pwrite64(unsigned int fd, const char __user *buf,
694 size_t count, loff_t pos)
697 ssize_t ret = -EBADF;
705 if (f.file->f_mode & FMODE_PWRITE)
706 ret = vfs_write(f.file, buf, count, &pos);
713 SYSCALL_DEFINE4(pwrite64, unsigned int, fd, const char __user *, buf,
714 size_t, count, loff_t, pos)
716 return ksys_pwrite64(fd, buf, count, pos);
719 #if defined(CONFIG_COMPAT) && defined(__ARCH_WANT_COMPAT_PWRITE64)
720 COMPAT_SYSCALL_DEFINE5(pwrite64, unsigned int, fd, const char __user *, buf,
721 size_t, count, compat_arg_u64_dual(pos))
723 return ksys_pwrite64(fd, buf, count, compat_arg_u64_glue(pos));
727 static ssize_t do_iter_readv_writev(struct file *filp, struct iov_iter *iter,
728 loff_t *ppos, int type, rwf_t flags)
733 init_sync_kiocb(&kiocb, filp);
734 ret = kiocb_set_rw_flags(&kiocb, flags);
737 kiocb.ki_pos = (ppos ? *ppos : 0);
740 ret = call_read_iter(filp, &kiocb, iter);
742 ret = call_write_iter(filp, &kiocb, iter);
743 BUG_ON(ret == -EIOCBQUEUED);
745 *ppos = kiocb.ki_pos;
749 /* Do it by hand, with file-ops */
750 static ssize_t do_loop_readv_writev(struct file *filp, struct iov_iter *iter,
751 loff_t *ppos, int type, rwf_t flags)
755 if (flags & ~RWF_HIPRI)
758 while (iov_iter_count(iter)) {
759 struct iovec iovec = iov_iter_iovec(iter);
763 nr = filp->f_op->read(filp, iovec.iov_base,
764 iovec.iov_len, ppos);
766 nr = filp->f_op->write(filp, iovec.iov_base,
767 iovec.iov_len, ppos);
776 if (nr != iovec.iov_len)
778 iov_iter_advance(iter, nr);
784 static ssize_t do_iter_read(struct file *file, struct iov_iter *iter,
785 loff_t *pos, rwf_t flags)
790 if (!(file->f_mode & FMODE_READ))
792 if (!(file->f_mode & FMODE_CAN_READ))
795 tot_len = iov_iter_count(iter);
798 ret = rw_verify_area(READ, file, pos, tot_len);
802 if (file->f_op->read_iter)
803 ret = do_iter_readv_writev(file, iter, pos, READ, flags);
805 ret = do_loop_readv_writev(file, iter, pos, READ, flags);
808 fsnotify_access(file);
812 ssize_t vfs_iocb_iter_read(struct file *file, struct kiocb *iocb,
813 struct iov_iter *iter)
818 if (!file->f_op->read_iter)
820 if (!(file->f_mode & FMODE_READ))
822 if (!(file->f_mode & FMODE_CAN_READ))
825 tot_len = iov_iter_count(iter);
828 ret = rw_verify_area(READ, file, &iocb->ki_pos, tot_len);
832 ret = call_read_iter(file, iocb, iter);
835 fsnotify_access(file);
838 EXPORT_SYMBOL(vfs_iocb_iter_read);
840 ssize_t vfs_iter_read(struct file *file, struct iov_iter *iter, loff_t *ppos,
843 if (!file->f_op->read_iter)
845 return do_iter_read(file, iter, ppos, flags);
847 EXPORT_SYMBOL(vfs_iter_read);
849 static ssize_t do_iter_write(struct file *file, struct iov_iter *iter,
850 loff_t *pos, rwf_t flags)
855 if (!(file->f_mode & FMODE_WRITE))
857 if (!(file->f_mode & FMODE_CAN_WRITE))
860 tot_len = iov_iter_count(iter);
863 ret = rw_verify_area(WRITE, file, pos, tot_len);
867 if (file->f_op->write_iter)
868 ret = do_iter_readv_writev(file, iter, pos, WRITE, flags);
870 ret = do_loop_readv_writev(file, iter, pos, WRITE, flags);
872 fsnotify_modify(file);
876 ssize_t vfs_iocb_iter_write(struct file *file, struct kiocb *iocb,
877 struct iov_iter *iter)
882 if (!file->f_op->write_iter)
884 if (!(file->f_mode & FMODE_WRITE))
886 if (!(file->f_mode & FMODE_CAN_WRITE))
889 tot_len = iov_iter_count(iter);
892 ret = rw_verify_area(WRITE, file, &iocb->ki_pos, tot_len);
896 ret = call_write_iter(file, iocb, iter);
898 fsnotify_modify(file);
902 EXPORT_SYMBOL(vfs_iocb_iter_write);
904 ssize_t vfs_iter_write(struct file *file, struct iov_iter *iter, loff_t *ppos,
907 if (!file->f_op->write_iter)
909 return do_iter_write(file, iter, ppos, flags);
911 EXPORT_SYMBOL(vfs_iter_write);
913 static ssize_t vfs_readv(struct file *file, const struct iovec __user *vec,
914 unsigned long vlen, loff_t *pos, rwf_t flags)
916 struct iovec iovstack[UIO_FASTIOV];
917 struct iovec *iov = iovstack;
918 struct iov_iter iter;
921 ret = import_iovec(READ, vec, vlen, ARRAY_SIZE(iovstack), &iov, &iter);
923 ret = do_iter_read(file, &iter, pos, flags);
930 static ssize_t vfs_writev(struct file *file, const struct iovec __user *vec,
931 unsigned long vlen, loff_t *pos, rwf_t flags)
933 struct iovec iovstack[UIO_FASTIOV];
934 struct iovec *iov = iovstack;
935 struct iov_iter iter;
938 ret = import_iovec(WRITE, vec, vlen, ARRAY_SIZE(iovstack), &iov, &iter);
940 file_start_write(file);
941 ret = do_iter_write(file, &iter, pos, flags);
942 file_end_write(file);
948 static ssize_t do_readv(unsigned long fd, const struct iovec __user *vec,
949 unsigned long vlen, rwf_t flags)
951 struct fd f = fdget_pos(fd);
952 ssize_t ret = -EBADF;
955 loff_t pos, *ppos = file_ppos(f.file);
960 ret = vfs_readv(f.file, vec, vlen, ppos, flags);
961 if (ret >= 0 && ppos)
967 add_rchar(current, ret);
972 static ssize_t do_writev(unsigned long fd, const struct iovec __user *vec,
973 unsigned long vlen, rwf_t flags)
975 struct fd f = fdget_pos(fd);
976 ssize_t ret = -EBADF;
979 loff_t pos, *ppos = file_ppos(f.file);
984 ret = vfs_writev(f.file, vec, vlen, ppos, flags);
985 if (ret >= 0 && ppos)
991 add_wchar(current, ret);
996 static inline loff_t pos_from_hilo(unsigned long high, unsigned long low)
998 #define HALF_LONG_BITS (BITS_PER_LONG / 2)
999 return (((loff_t)high << HALF_LONG_BITS) << HALF_LONG_BITS) | low;
1002 static ssize_t do_preadv(unsigned long fd, const struct iovec __user *vec,
1003 unsigned long vlen, loff_t pos, rwf_t flags)
1006 ssize_t ret = -EBADF;
1014 if (f.file->f_mode & FMODE_PREAD)
1015 ret = vfs_readv(f.file, vec, vlen, &pos, flags);
1020 add_rchar(current, ret);
1025 static ssize_t do_pwritev(unsigned long fd, const struct iovec __user *vec,
1026 unsigned long vlen, loff_t pos, rwf_t flags)
1029 ssize_t ret = -EBADF;
1037 if (f.file->f_mode & FMODE_PWRITE)
1038 ret = vfs_writev(f.file, vec, vlen, &pos, flags);
1043 add_wchar(current, ret);
1048 SYSCALL_DEFINE3(readv, unsigned long, fd, const struct iovec __user *, vec,
1049 unsigned long, vlen)
1051 return do_readv(fd, vec, vlen, 0);
1054 SYSCALL_DEFINE3(writev, unsigned long, fd, const struct iovec __user *, vec,
1055 unsigned long, vlen)
1057 return do_writev(fd, vec, vlen, 0);
1060 SYSCALL_DEFINE5(preadv, unsigned long, fd, const struct iovec __user *, vec,
1061 unsigned long, vlen, unsigned long, pos_l, unsigned long, pos_h)
1063 loff_t pos = pos_from_hilo(pos_h, pos_l);
1065 return do_preadv(fd, vec, vlen, pos, 0);
1068 SYSCALL_DEFINE6(preadv2, unsigned long, fd, const struct iovec __user *, vec,
1069 unsigned long, vlen, unsigned long, pos_l, unsigned long, pos_h,
1072 loff_t pos = pos_from_hilo(pos_h, pos_l);
1075 return do_readv(fd, vec, vlen, flags);
1077 return do_preadv(fd, vec, vlen, pos, flags);
1080 SYSCALL_DEFINE5(pwritev, unsigned long, fd, const struct iovec __user *, vec,
1081 unsigned long, vlen, unsigned long, pos_l, unsigned long, pos_h)
1083 loff_t pos = pos_from_hilo(pos_h, pos_l);
1085 return do_pwritev(fd, vec, vlen, pos, 0);
1088 SYSCALL_DEFINE6(pwritev2, unsigned long, fd, const struct iovec __user *, vec,
1089 unsigned long, vlen, unsigned long, pos_l, unsigned long, pos_h,
1092 loff_t pos = pos_from_hilo(pos_h, pos_l);
1095 return do_writev(fd, vec, vlen, flags);
1097 return do_pwritev(fd, vec, vlen, pos, flags);
1101 * Various compat syscalls. Note that they all pretend to take a native
1102 * iovec - import_iovec will properly treat those as compat_iovecs based on
1103 * in_compat_syscall().
1105 #ifdef CONFIG_COMPAT
1106 #ifdef __ARCH_WANT_COMPAT_SYS_PREADV64
1107 COMPAT_SYSCALL_DEFINE4(preadv64, unsigned long, fd,
1108 const struct iovec __user *, vec,
1109 unsigned long, vlen, loff_t, pos)
1111 return do_preadv(fd, vec, vlen, pos, 0);
1115 COMPAT_SYSCALL_DEFINE5(preadv, compat_ulong_t, fd,
1116 const struct iovec __user *, vec,
1117 compat_ulong_t, vlen, u32, pos_low, u32, pos_high)
1119 loff_t pos = ((loff_t)pos_high << 32) | pos_low;
1121 return do_preadv(fd, vec, vlen, pos, 0);
1124 #ifdef __ARCH_WANT_COMPAT_SYS_PREADV64V2
1125 COMPAT_SYSCALL_DEFINE5(preadv64v2, unsigned long, fd,
1126 const struct iovec __user *, vec,
1127 unsigned long, vlen, loff_t, pos, rwf_t, flags)
1130 return do_readv(fd, vec, vlen, flags);
1131 return do_preadv(fd, vec, vlen, pos, flags);
1135 COMPAT_SYSCALL_DEFINE6(preadv2, compat_ulong_t, fd,
1136 const struct iovec __user *, vec,
1137 compat_ulong_t, vlen, u32, pos_low, u32, pos_high,
1140 loff_t pos = ((loff_t)pos_high << 32) | pos_low;
1143 return do_readv(fd, vec, vlen, flags);
1144 return do_preadv(fd, vec, vlen, pos, flags);
1147 #ifdef __ARCH_WANT_COMPAT_SYS_PWRITEV64
1148 COMPAT_SYSCALL_DEFINE4(pwritev64, unsigned long, fd,
1149 const struct iovec __user *, vec,
1150 unsigned long, vlen, loff_t, pos)
1152 return do_pwritev(fd, vec, vlen, pos, 0);
1156 COMPAT_SYSCALL_DEFINE5(pwritev, compat_ulong_t, fd,
1157 const struct iovec __user *,vec,
1158 compat_ulong_t, vlen, u32, pos_low, u32, pos_high)
1160 loff_t pos = ((loff_t)pos_high << 32) | pos_low;
1162 return do_pwritev(fd, vec, vlen, pos, 0);
1165 #ifdef __ARCH_WANT_COMPAT_SYS_PWRITEV64V2
1166 COMPAT_SYSCALL_DEFINE5(pwritev64v2, unsigned long, fd,
1167 const struct iovec __user *, vec,
1168 unsigned long, vlen, loff_t, pos, rwf_t, flags)
1171 return do_writev(fd, vec, vlen, flags);
1172 return do_pwritev(fd, vec, vlen, pos, flags);
1176 COMPAT_SYSCALL_DEFINE6(pwritev2, compat_ulong_t, fd,
1177 const struct iovec __user *,vec,
1178 compat_ulong_t, vlen, u32, pos_low, u32, pos_high, rwf_t, flags)
1180 loff_t pos = ((loff_t)pos_high << 32) | pos_low;
1183 return do_writev(fd, vec, vlen, flags);
1184 return do_pwritev(fd, vec, vlen, pos, flags);
1186 #endif /* CONFIG_COMPAT */
1188 static ssize_t do_sendfile(int out_fd, int in_fd, loff_t *ppos,
1189 size_t count, loff_t max)
1192 struct inode *in_inode, *out_inode;
1193 struct pipe_inode_info *opipe;
1200 * Get input file, and verify that it is ok..
1206 if (!(in.file->f_mode & FMODE_READ))
1210 pos = in.file->f_pos;
1213 if (!(in.file->f_mode & FMODE_PREAD))
1216 retval = rw_verify_area(READ, in.file, &pos, count);
1219 if (count > MAX_RW_COUNT)
1220 count = MAX_RW_COUNT;
1223 * Get output file, and verify that it is ok..
1226 out = fdget(out_fd);
1229 if (!(out.file->f_mode & FMODE_WRITE))
1231 in_inode = file_inode(in.file);
1232 out_inode = file_inode(out.file);
1233 out_pos = out.file->f_pos;
1236 max = min(in_inode->i_sb->s_maxbytes, out_inode->i_sb->s_maxbytes);
1238 if (unlikely(pos + count > max)) {
1239 retval = -EOVERFLOW;
1248 * We need to debate whether we can enable this or not. The
1249 * man page documents EAGAIN return for the output at least,
1250 * and the application is arguably buggy if it doesn't expect
1251 * EAGAIN on a non-blocking file descriptor.
1253 if (in.file->f_flags & O_NONBLOCK)
1254 fl = SPLICE_F_NONBLOCK;
1256 opipe = get_pipe_info(out.file, true);
1258 retval = rw_verify_area(WRITE, out.file, &out_pos, count);
1261 file_start_write(out.file);
1262 retval = do_splice_direct(in.file, &pos, out.file, &out_pos,
1264 file_end_write(out.file);
1266 retval = splice_file_to_pipe(in.file, opipe, &pos, count, fl);
1270 add_rchar(current, retval);
1271 add_wchar(current, retval);
1272 fsnotify_access(in.file);
1273 fsnotify_modify(out.file);
1274 out.file->f_pos = out_pos;
1278 in.file->f_pos = pos;
1284 retval = -EOVERFLOW;
1294 SYSCALL_DEFINE4(sendfile, int, out_fd, int, in_fd, off_t __user *, offset, size_t, count)
1301 if (unlikely(get_user(off, offset)))
1304 ret = do_sendfile(out_fd, in_fd, &pos, count, MAX_NON_LFS);
1305 if (unlikely(put_user(pos, offset)))
1310 return do_sendfile(out_fd, in_fd, NULL, count, 0);
1313 SYSCALL_DEFINE4(sendfile64, int, out_fd, int, in_fd, loff_t __user *, offset, size_t, count)
1319 if (unlikely(copy_from_user(&pos, offset, sizeof(loff_t))))
1321 ret = do_sendfile(out_fd, in_fd, &pos, count, 0);
1322 if (unlikely(put_user(pos, offset)))
1327 return do_sendfile(out_fd, in_fd, NULL, count, 0);
1330 #ifdef CONFIG_COMPAT
1331 COMPAT_SYSCALL_DEFINE4(sendfile, int, out_fd, int, in_fd,
1332 compat_off_t __user *, offset, compat_size_t, count)
1339 if (unlikely(get_user(off, offset)))
1342 ret = do_sendfile(out_fd, in_fd, &pos, count, MAX_NON_LFS);
1343 if (unlikely(put_user(pos, offset)))
1348 return do_sendfile(out_fd, in_fd, NULL, count, 0);
1351 COMPAT_SYSCALL_DEFINE4(sendfile64, int, out_fd, int, in_fd,
1352 compat_loff_t __user *, offset, compat_size_t, count)
1358 if (unlikely(copy_from_user(&pos, offset, sizeof(loff_t))))
1360 ret = do_sendfile(out_fd, in_fd, &pos, count, 0);
1361 if (unlikely(put_user(pos, offset)))
1366 return do_sendfile(out_fd, in_fd, NULL, count, 0);
1371 * generic_copy_file_range - copy data between two files
1372 * @file_in: file structure to read from
1373 * @pos_in: file offset to read from
1374 * @file_out: file structure to write data to
1375 * @pos_out: file offset to write data to
1376 * @len: amount of data to copy
1377 * @flags: copy flags
1379 * This is a generic filesystem helper to copy data from one file to another.
1380 * It has no constraints on the source or destination file owners - the files
1381 * can belong to different superblocks and different filesystem types. Short
1382 * copies are allowed.
1384 * This should be called from the @file_out filesystem, as per the
1385 * ->copy_file_range() method.
1387 * Returns the number of bytes copied or a negative error indicating the
1391 ssize_t generic_copy_file_range(struct file *file_in, loff_t pos_in,
1392 struct file *file_out, loff_t pos_out,
1393 size_t len, unsigned int flags)
1395 return do_splice_direct(file_in, &pos_in, file_out, &pos_out,
1396 len > MAX_RW_COUNT ? MAX_RW_COUNT : len, 0);
1398 EXPORT_SYMBOL(generic_copy_file_range);
1401 * Performs necessary checks before doing a file copy
1403 * Can adjust amount of bytes to copy via @req_count argument.
1404 * Returns appropriate error code that caller should return or
1405 * zero in case the copy should be allowed.
1407 static int generic_copy_file_checks(struct file *file_in, loff_t pos_in,
1408 struct file *file_out, loff_t pos_out,
1409 size_t *req_count, unsigned int flags)
1411 struct inode *inode_in = file_inode(file_in);
1412 struct inode *inode_out = file_inode(file_out);
1413 uint64_t count = *req_count;
1417 ret = generic_file_rw_checks(file_in, file_out);
1422 * We allow some filesystems to handle cross sb copy, but passing
1423 * a file of the wrong filesystem type to filesystem driver can result
1424 * in an attempt to dereference the wrong type of ->private_data, so
1425 * avoid doing that until we really have a good reason.
1427 * nfs and cifs define several different file_system_type structures
1428 * and several different sets of file_operations, but they all end up
1429 * using the same ->copy_file_range() function pointer.
1431 if (file_out->f_op->copy_file_range) {
1432 if (file_in->f_op->copy_file_range !=
1433 file_out->f_op->copy_file_range)
1435 } else if (file_inode(file_in)->i_sb != file_inode(file_out)->i_sb) {
1439 /* Don't touch certain kinds of inodes */
1440 if (IS_IMMUTABLE(inode_out))
1443 if (IS_SWAPFILE(inode_in) || IS_SWAPFILE(inode_out))
1446 /* Ensure offsets don't wrap. */
1447 if (pos_in + count < pos_in || pos_out + count < pos_out)
1450 /* Shorten the copy to EOF */
1451 size_in = i_size_read(inode_in);
1452 if (pos_in >= size_in)
1455 count = min(count, size_in - (uint64_t)pos_in);
1457 ret = generic_write_check_limits(file_out, pos_out, &count);
1461 /* Don't allow overlapped copying within the same file. */
1462 if (inode_in == inode_out &&
1463 pos_out + count > pos_in &&
1464 pos_out < pos_in + count)
1472 * copy_file_range() differs from regular file read and write in that it
1473 * specifically allows return partial success. When it does so is up to
1474 * the copy_file_range method.
1476 ssize_t vfs_copy_file_range(struct file *file_in, loff_t pos_in,
1477 struct file *file_out, loff_t pos_out,
1478 size_t len, unsigned int flags)
1485 ret = generic_copy_file_checks(file_in, pos_in, file_out, pos_out, &len,
1490 ret = rw_verify_area(READ, file_in, &pos_in, len);
1494 ret = rw_verify_area(WRITE, file_out, &pos_out, len);
1501 file_start_write(file_out);
1504 * Cloning is supported by more file systems, so we implement copy on
1505 * same sb using clone, but for filesystems where both clone and copy
1506 * are supported (e.g. nfs,cifs), we only call the copy method.
1508 if (file_out->f_op->copy_file_range) {
1509 ret = file_out->f_op->copy_file_range(file_in, pos_in,
1515 if (file_in->f_op->remap_file_range &&
1516 file_inode(file_in)->i_sb == file_inode(file_out)->i_sb) {
1517 ret = file_in->f_op->remap_file_range(file_in, pos_in,
1519 min_t(loff_t, MAX_RW_COUNT, len),
1520 REMAP_FILE_CAN_SHORTEN);
1526 * We can get here for same sb copy of filesystems that do not implement
1527 * ->copy_file_range() in case filesystem does not support clone or in
1528 * case filesystem supports clone but rejected the clone request (e.g.
1529 * because it was not block aligned).
1531 * In both cases, fall back to kernel copy so we are able to maintain a
1532 * consistent story about which filesystems support copy_file_range()
1533 * and which filesystems do not, that will allow userspace tools to
1534 * make consistent desicions w.r.t using copy_file_range().
1536 ret = generic_copy_file_range(file_in, pos_in, file_out, pos_out, len,
1541 fsnotify_access(file_in);
1542 add_rchar(current, ret);
1543 fsnotify_modify(file_out);
1544 add_wchar(current, ret);
1550 file_end_write(file_out);
1554 EXPORT_SYMBOL(vfs_copy_file_range);
1556 SYSCALL_DEFINE6(copy_file_range, int, fd_in, loff_t __user *, off_in,
1557 int, fd_out, loff_t __user *, off_out,
1558 size_t, len, unsigned int, flags)
1564 ssize_t ret = -EBADF;
1566 f_in = fdget(fd_in);
1570 f_out = fdget(fd_out);
1576 if (copy_from_user(&pos_in, off_in, sizeof(loff_t)))
1579 pos_in = f_in.file->f_pos;
1583 if (copy_from_user(&pos_out, off_out, sizeof(loff_t)))
1586 pos_out = f_out.file->f_pos;
1589 ret = vfs_copy_file_range(f_in.file, pos_in, f_out.file, pos_out, len,
1596 if (copy_to_user(off_in, &pos_in, sizeof(loff_t)))
1599 f_in.file->f_pos = pos_in;
1603 if (copy_to_user(off_out, &pos_out, sizeof(loff_t)))
1606 f_out.file->f_pos = pos_out;
1619 * Don't operate on ranges the page cache doesn't support, and don't exceed the
1620 * LFS limits. If pos is under the limit it becomes a short access. If it
1621 * exceeds the limit we return -EFBIG.
1623 int generic_write_check_limits(struct file *file, loff_t pos, loff_t *count)
1625 struct inode *inode = file->f_mapping->host;
1626 loff_t max_size = inode->i_sb->s_maxbytes;
1627 loff_t limit = rlimit(RLIMIT_FSIZE);
1629 if (limit != RLIM_INFINITY) {
1631 send_sig(SIGXFSZ, current, 0);
1634 *count = min(*count, limit - pos);
1637 if (!(file->f_flags & O_LARGEFILE))
1638 max_size = MAX_NON_LFS;
1640 if (unlikely(pos >= max_size))
1643 *count = min(*count, max_size - pos);
1648 /* Like generic_write_checks(), but takes size of write instead of iter. */
1649 int generic_write_checks_count(struct kiocb *iocb, loff_t *count)
1651 struct file *file = iocb->ki_filp;
1652 struct inode *inode = file->f_mapping->host;
1654 if (IS_SWAPFILE(inode))
1660 if (iocb->ki_flags & IOCB_APPEND)
1661 iocb->ki_pos = i_size_read(inode);
1663 if ((iocb->ki_flags & IOCB_NOWAIT) && !(iocb->ki_flags & IOCB_DIRECT))
1666 return generic_write_check_limits(iocb->ki_filp, iocb->ki_pos, count);
1668 EXPORT_SYMBOL(generic_write_checks_count);
1671 * Performs necessary checks before doing a write
1673 * Can adjust writing position or amount of bytes to write.
1674 * Returns appropriate error code that caller should return or
1675 * zero in case that write should be allowed.
1677 ssize_t generic_write_checks(struct kiocb *iocb, struct iov_iter *from)
1679 loff_t count = iov_iter_count(from);
1682 ret = generic_write_checks_count(iocb, &count);
1686 iov_iter_truncate(from, count);
1687 return iov_iter_count(from);
1689 EXPORT_SYMBOL(generic_write_checks);
1692 * Performs common checks before doing a file copy/clone
1693 * from @file_in to @file_out.
1695 int generic_file_rw_checks(struct file *file_in, struct file *file_out)
1697 struct inode *inode_in = file_inode(file_in);
1698 struct inode *inode_out = file_inode(file_out);
1700 /* Don't copy dirs, pipes, sockets... */
1701 if (S_ISDIR(inode_in->i_mode) || S_ISDIR(inode_out->i_mode))
1703 if (!S_ISREG(inode_in->i_mode) || !S_ISREG(inode_out->i_mode))
1706 if (!(file_in->f_mode & FMODE_READ) ||
1707 !(file_out->f_mode & FMODE_WRITE) ||
1708 (file_out->f_flags & O_APPEND))