b24413180f560 (Greg Kroah-Hartman 2017-11-01 15:07:57 +0100 1) // SPDX-License-Identifier: GPL-2.0
^1da177e4c3f4 (Linus Torvalds 2005-04-16 15:20:36 -0700 2) /*
^1da177e4c3f4 (Linus Torvalds 2005-04-16 15:20:36 -0700 3) * linux/fs/file.c
^1da177e4c3f4 (Linus Torvalds 2005-04-16 15:20:36 -0700 4) *
^1da177e4c3f4 (Linus Torvalds 2005-04-16 15:20:36 -0700 5) * Copyright (C) 1998-1999, Stephen Tweedie and Bill Hawes
^1da177e4c3f4 (Linus Torvalds 2005-04-16 15:20:36 -0700 6) *
^1da177e4c3f4 (Linus Torvalds 2005-04-16 15:20:36 -0700 7) * Manage the dynamic fd arrays in the process files_struct.
^1da177e4c3f4 (Linus Torvalds 2005-04-16 15:20:36 -0700 8) */
^1da177e4c3f4 (Linus Torvalds 2005-04-16 15:20:36 -0700 9)
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 10) #include <linux/syscalls.h>
630d9c47274aa (Paul Gortmaker 2011-11-16 23:57:37 -0500 11) #include <linux/export.h>
^1da177e4c3f4 (Linus Torvalds 2005-04-16 15:20:36 -0700 12) #include <linux/fs.h>
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 13) #include <linux/kernel.h>
^1da177e4c3f4 (Linus Torvalds 2005-04-16 15:20:36 -0700 14) #include <linux/mm.h>
3f07c0144132e (Ingo Molnar 2017-02-08 18:51:30 +0100 15) #include <linux/sched/signal.h>
^1da177e4c3f4 (Linus Torvalds 2005-04-16 15:20:36 -0700 16) #include <linux/slab.h>
^1da177e4c3f4 (Linus Torvalds 2005-04-16 15:20:36 -0700 17) #include <linux/file.h>
9f3acc3140444 (Al Viro 2008-04-24 07:44:08 -0400 18) #include <linux/fdtable.h>
^1da177e4c3f4 (Linus Torvalds 2005-04-16 15:20:36 -0700 19) #include <linux/bitops.h>
ab2af1f500506 (Dipankar Sarma 2005-09-09 13:04:13 -0700 20) #include <linux/spinlock.h>
ab2af1f500506 (Dipankar Sarma 2005-09-09 13:04:13 -0700 21) #include <linux/rcupdate.h>
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 22) #include <linux/close_range.h>
6659061045cc9 (Kees Cook 2020-06-10 08:20:05 -0700 23) #include <net/sock.h>
ab2af1f500506 (Dipankar Sarma 2005-09-09 13:04:13 -0700 24)
53dec2ea74f2e (Jens Axboe 2021-01-19 15:41:52 -0700 25) #include "internal.h"
53dec2ea74f2e (Jens Axboe 2021-01-19 15:41:52 -0700 26)
9b80a184eaadc (Alexey Dobriyan 2016-09-02 00:38:52 +0300 27) unsigned int sysctl_nr_open __read_mostly = 1024*1024;
9b80a184eaadc (Alexey Dobriyan 2016-09-02 00:38:52 +0300 28) unsigned int sysctl_nr_open_min = BITS_PER_LONG;
752343be63d90 (Rasmus Villemoes 2015-10-29 12:01:41 +0100 29) /* our min() is unusable in constant expressions ;-/ */
752343be63d90 (Rasmus Villemoes 2015-10-29 12:01:41 +0100 30) #define __const_min(x, y) ((x) < (y) ? (x) : (y))
9b80a184eaadc (Alexey Dobriyan 2016-09-02 00:38:52 +0300 31) unsigned int sysctl_nr_open_max =
9b80a184eaadc (Alexey Dobriyan 2016-09-02 00:38:52 +0300 32) __const_min(INT_MAX, ~(size_t)0/sizeof(void *)) & -BITS_PER_LONG;
9cfe015aa424b (Eric Dumazet 2008-02-06 01:37:16 -0800 33)
a892e2d7dcdfa (Changli Gao 2010-08-10 18:01:35 -0700 34) static void __free_fdtable(struct fdtable *fdt)
^1da177e4c3f4 (Linus Torvalds 2005-04-16 15:20:36 -0700 35) {
f6c0a1920e018 (Al Viro 2014-04-23 10:18:46 -0400 36) kvfree(fdt->fd);
f6c0a1920e018 (Al Viro 2014-04-23 10:18:46 -0400 37) kvfree(fdt->open_fds);
a892e2d7dcdfa (Changli Gao 2010-08-10 18:01:35 -0700 38) kfree(fdt);
ab2af1f500506 (Dipankar Sarma 2005-09-09 13:04:13 -0700 39) }
^1da177e4c3f4 (Linus Torvalds 2005-04-16 15:20:36 -0700 40)
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 41) static void free_fdtable_rcu(struct rcu_head *rcu)
ab2af1f500506 (Dipankar Sarma 2005-09-09 13:04:13 -0700 42) {
ac3e3c5b11643 (Al Viro 2013-04-28 21:42:33 -0400 43) __free_fdtable(container_of(rcu, struct fdtable, rcu));
ab2af1f500506 (Dipankar Sarma 2005-09-09 13:04:13 -0700 44) }
ab2af1f500506 (Dipankar Sarma 2005-09-09 13:04:13 -0700 45)
f3f86e33dc3da (Linus Torvalds 2015-10-30 16:53:57 -0700 46) #define BITBIT_NR(nr) BITS_TO_LONGS(BITS_TO_LONGS(nr))
f3f86e33dc3da (Linus Torvalds 2015-10-30 16:53:57 -0700 47) #define BITBIT_SIZE(nr) (BITBIT_NR(nr) * sizeof(long))
f3f86e33dc3da (Linus Torvalds 2015-10-30 16:53:57 -0700 48)
ab2af1f500506 (Dipankar Sarma 2005-09-09 13:04:13 -0700 49) /*
ea5c58e70c3a1 (Eric Biggers 2015-11-06 00:32:04 -0600 50) * Copy 'count' fd bits from the old table to the new table and clear the extra
ea5c58e70c3a1 (Eric Biggers 2015-11-06 00:32:04 -0600 51) * space if any. This does not copy the file pointers. Called with the files
ea5c58e70c3a1 (Eric Biggers 2015-11-06 00:32:04 -0600 52) * spinlock held for write.
ea5c58e70c3a1 (Eric Biggers 2015-11-06 00:32:04 -0600 53) */
ea5c58e70c3a1 (Eric Biggers 2015-11-06 00:32:04 -0600 54) static void copy_fd_bitmaps(struct fdtable *nfdt, struct fdtable *ofdt,
ea5c58e70c3a1 (Eric Biggers 2015-11-06 00:32:04 -0600 55) unsigned int count)
ea5c58e70c3a1 (Eric Biggers 2015-11-06 00:32:04 -0600 56) {
ea5c58e70c3a1 (Eric Biggers 2015-11-06 00:32:04 -0600 57) unsigned int cpy, set;
ea5c58e70c3a1 (Eric Biggers 2015-11-06 00:32:04 -0600 58)
ea5c58e70c3a1 (Eric Biggers 2015-11-06 00:32:04 -0600 59) cpy = count / BITS_PER_BYTE;
ea5c58e70c3a1 (Eric Biggers 2015-11-06 00:32:04 -0600 60) set = (nfdt->max_fds - count) / BITS_PER_BYTE;
ea5c58e70c3a1 (Eric Biggers 2015-11-06 00:32:04 -0600 61) memcpy(nfdt->open_fds, ofdt->open_fds, cpy);
ea5c58e70c3a1 (Eric Biggers 2015-11-06 00:32:04 -0600 62) memset((char *)nfdt->open_fds + cpy, 0, set);
ea5c58e70c3a1 (Eric Biggers 2015-11-06 00:32:04 -0600 63) memcpy(nfdt->close_on_exec, ofdt->close_on_exec, cpy);
ea5c58e70c3a1 (Eric Biggers 2015-11-06 00:32:04 -0600 64) memset((char *)nfdt->close_on_exec + cpy, 0, set);
ea5c58e70c3a1 (Eric Biggers 2015-11-06 00:32:04 -0600 65)
ea5c58e70c3a1 (Eric Biggers 2015-11-06 00:32:04 -0600 66) cpy = BITBIT_SIZE(count);
ea5c58e70c3a1 (Eric Biggers 2015-11-06 00:32:04 -0600 67) set = BITBIT_SIZE(nfdt->max_fds) - cpy;
ea5c58e70c3a1 (Eric Biggers 2015-11-06 00:32:04 -0600 68) memcpy(nfdt->full_fds_bits, ofdt->full_fds_bits, cpy);
ea5c58e70c3a1 (Eric Biggers 2015-11-06 00:32:04 -0600 69) memset((char *)nfdt->full_fds_bits + cpy, 0, set);
ea5c58e70c3a1 (Eric Biggers 2015-11-06 00:32:04 -0600 70) }
ea5c58e70c3a1 (Eric Biggers 2015-11-06 00:32:04 -0600 71)
ea5c58e70c3a1 (Eric Biggers 2015-11-06 00:32:04 -0600 72) /*
ea5c58e70c3a1 (Eric Biggers 2015-11-06 00:32:04 -0600 73) * Copy all file descriptors from the old table to the new, expanded table and
ea5c58e70c3a1 (Eric Biggers 2015-11-06 00:32:04 -0600 74) * clear the extra space. Called with the files spinlock held for write.
ab2af1f500506 (Dipankar Sarma 2005-09-09 13:04:13 -0700 75) */
5466b456ed674 (Vadim Lobanov 2006-12-10 02:21:22 -0800 76) static void copy_fdtable(struct fdtable *nfdt, struct fdtable *ofdt)
ab2af1f500506 (Dipankar Sarma 2005-09-09 13:04:13 -0700 77) {
4e89b7210403f (Al Viro 2020-05-19 17:48:52 -0400 78) size_t cpy, set;
ab2af1f500506 (Dipankar Sarma 2005-09-09 13:04:13 -0700 79)
5466b456ed674 (Vadim Lobanov 2006-12-10 02:21:22 -0800 80) BUG_ON(nfdt->max_fds < ofdt->max_fds);
5466b456ed674 (Vadim Lobanov 2006-12-10 02:21:22 -0800 81)
5466b456ed674 (Vadim Lobanov 2006-12-10 02:21:22 -0800 82) cpy = ofdt->max_fds * sizeof(struct file *);
5466b456ed674 (Vadim Lobanov 2006-12-10 02:21:22 -0800 83) set = (nfdt->max_fds - ofdt->max_fds) * sizeof(struct file *);
5466b456ed674 (Vadim Lobanov 2006-12-10 02:21:22 -0800 84) memcpy(nfdt->fd, ofdt->fd, cpy);
ea5c58e70c3a1 (Eric Biggers 2015-11-06 00:32:04 -0600 85) memset((char *)nfdt->fd + cpy, 0, set);
5466b456ed674 (Vadim Lobanov 2006-12-10 02:21:22 -0800 86)
ea5c58e70c3a1 (Eric Biggers 2015-11-06 00:32:04 -0600 87) copy_fd_bitmaps(nfdt, ofdt, ofdt->max_fds);
^1da177e4c3f4 (Linus Torvalds 2005-04-16 15:20:36 -0700 88) }
^1da177e4c3f4 (Linus Torvalds 2005-04-16 15:20:36 -0700 89)
5466b456ed674 (Vadim Lobanov 2006-12-10 02:21:22 -0800 90) static struct fdtable * alloc_fdtable(unsigned int nr)
^1da177e4c3f4 (Linus Torvalds 2005-04-16 15:20:36 -0700 91) {
5466b456ed674 (Vadim Lobanov 2006-12-10 02:21:22 -0800 92) struct fdtable *fdt;
1fd36adcd98c1 (David Howells 2012-02-16 17:49:54 +0000 93) void *data;
^1da177e4c3f4 (Linus Torvalds 2005-04-16 15:20:36 -0700 94)
ab2af1f500506 (Dipankar Sarma 2005-09-09 13:04:13 -0700 95) /*
5466b456ed674 (Vadim Lobanov 2006-12-10 02:21:22 -0800 96) * Figure out how many fds we actually want to support in this fdtable.
5466b456ed674 (Vadim Lobanov 2006-12-10 02:21:22 -0800 97) * Allocation steps are keyed to the size of the fdarray, since it
5466b456ed674 (Vadim Lobanov 2006-12-10 02:21:22 -0800 98) * grows far faster than any of the other dynamic data. We try to fit
5466b456ed674 (Vadim Lobanov 2006-12-10 02:21:22 -0800 99) * the fdarray into comfortable page-tuned chunks: starting at 1024B
5466b456ed674 (Vadim Lobanov 2006-12-10 02:21:22 -0800 100) * and growing in powers of two from there on.
ab2af1f500506 (Dipankar Sarma 2005-09-09 13:04:13 -0700 101) */
5466b456ed674 (Vadim Lobanov 2006-12-10 02:21:22 -0800 102) nr /= (1024 / sizeof(struct file *));
5466b456ed674 (Vadim Lobanov 2006-12-10 02:21:22 -0800 103) nr = roundup_pow_of_two(nr + 1);
5466b456ed674 (Vadim Lobanov 2006-12-10 02:21:22 -0800 104) nr *= (1024 / sizeof(struct file *));
5c598b3428c37 (Al Viro 2008-04-27 20:04:15 -0400 105) /*
5c598b3428c37 (Al Viro 2008-04-27 20:04:15 -0400 106) * Note that this can drive nr *below* what we had passed if sysctl_nr_open
5c598b3428c37 (Al Viro 2008-04-27 20:04:15 -0400 107) * had been set lower between the check in expand_files() and here. Deal
5c598b3428c37 (Al Viro 2008-04-27 20:04:15 -0400 108) * with that in caller, it's cheaper that way.
5c598b3428c37 (Al Viro 2008-04-27 20:04:15 -0400 109) *
5c598b3428c37 (Al Viro 2008-04-27 20:04:15 -0400 110) * We make sure that nr remains a multiple of BITS_PER_LONG - otherwise
5c598b3428c37 (Al Viro 2008-04-27 20:04:15 -0400 111) * bitmaps handling below becomes unpleasant, to put it mildly...
5c598b3428c37 (Al Viro 2008-04-27 20:04:15 -0400 112) */
5c598b3428c37 (Al Viro 2008-04-27 20:04:15 -0400 113) if (unlikely(nr > sysctl_nr_open))
5c598b3428c37 (Al Viro 2008-04-27 20:04:15 -0400 114) nr = ((sysctl_nr_open - 1) | (BITS_PER_LONG - 1)) + 1;
bbea9f69668a3 (Vadim Lobanov 2006-12-10 02:21:12 -0800 115)
5d097056c9a01 (Vladimir Davydov 2016-01-14 15:18:21 -0800 116) fdt = kmalloc(sizeof(struct fdtable), GFP_KERNEL_ACCOUNT);
5466b456ed674 (Vadim Lobanov 2006-12-10 02:21:22 -0800 117) if (!fdt)
bbea9f69668a3 (Vadim Lobanov 2006-12-10 02:21:12 -0800 118) goto out;
5466b456ed674 (Vadim Lobanov 2006-12-10 02:21:22 -0800 119) fdt->max_fds = nr;
c823bd9244337 (Michal Hocko 2017-07-06 15:36:19 -0700 120) data = kvmalloc_array(nr, sizeof(struct file *), GFP_KERNEL_ACCOUNT);
5466b456ed674 (Vadim Lobanov 2006-12-10 02:21:22 -0800 121) if (!data)
5466b456ed674 (Vadim Lobanov 2006-12-10 02:21:22 -0800 122) goto out_fdt;
1fd36adcd98c1 (David Howells 2012-02-16 17:49:54 +0000 123) fdt->fd = data;
1fd36adcd98c1 (David Howells 2012-02-16 17:49:54 +0000 124)
c823bd9244337 (Michal Hocko 2017-07-06 15:36:19 -0700 125) data = kvmalloc(max_t(size_t,
c823bd9244337 (Michal Hocko 2017-07-06 15:36:19 -0700 126) 2 * nr / BITS_PER_BYTE + BITBIT_SIZE(nr), L1_CACHE_BYTES),
c823bd9244337 (Michal Hocko 2017-07-06 15:36:19 -0700 127) GFP_KERNEL_ACCOUNT);
5466b456ed674 (Vadim Lobanov 2006-12-10 02:21:22 -0800 128) if (!data)
5466b456ed674 (Vadim Lobanov 2006-12-10 02:21:22 -0800 129) goto out_arr;
1fd36adcd98c1 (David Howells 2012-02-16 17:49:54 +0000 130) fdt->open_fds = data;
5466b456ed674 (Vadim Lobanov 2006-12-10 02:21:22 -0800 131) data += nr / BITS_PER_BYTE;
1fd36adcd98c1 (David Howells 2012-02-16 17:49:54 +0000 132) fdt->close_on_exec = data;
f3f86e33dc3da (Linus Torvalds 2015-10-30 16:53:57 -0700 133) data += nr / BITS_PER_BYTE;
f3f86e33dc3da (Linus Torvalds 2015-10-30 16:53:57 -0700 134) fdt->full_fds_bits = data;
5466b456ed674 (Vadim Lobanov 2006-12-10 02:21:22 -0800 135)
ab2af1f500506 (Dipankar Sarma 2005-09-09 13:04:13 -0700 136) return fdt;
5466b456ed674 (Vadim Lobanov 2006-12-10 02:21:22 -0800 137)
5466b456ed674 (Vadim Lobanov 2006-12-10 02:21:22 -0800 138) out_arr:
f6c0a1920e018 (Al Viro 2014-04-23 10:18:46 -0400 139) kvfree(fdt->fd);
5466b456ed674 (Vadim Lobanov 2006-12-10 02:21:22 -0800 140) out_fdt:
ab2af1f500506 (Dipankar Sarma 2005-09-09 13:04:13 -0700 141) kfree(fdt);
5466b456ed674 (Vadim Lobanov 2006-12-10 02:21:22 -0800 142) out:
ab2af1f500506 (Dipankar Sarma 2005-09-09 13:04:13 -0700 143) return NULL;
ab2af1f500506 (Dipankar Sarma 2005-09-09 13:04:13 -0700 144) }
^1da177e4c3f4 (Linus Torvalds 2005-04-16 15:20:36 -0700 145)
ab2af1f500506 (Dipankar Sarma 2005-09-09 13:04:13 -0700 146) /*
74d392aaabfc8 (Vadim Lobanov 2006-09-29 02:01:43 -0700 147) * Expand the file descriptor table.
74d392aaabfc8 (Vadim Lobanov 2006-09-29 02:01:43 -0700 148) * This function will allocate a new fdtable and both fd array and fdset, of
74d392aaabfc8 (Vadim Lobanov 2006-09-29 02:01:43 -0700 149) * the given size.
74d392aaabfc8 (Vadim Lobanov 2006-09-29 02:01:43 -0700 150) * Return <0 error code on error; 1 on successful completion.
74d392aaabfc8 (Vadim Lobanov 2006-09-29 02:01:43 -0700 151) * The files->file_lock should be held on entry, and will be held on exit.
ab2af1f500506 (Dipankar Sarma 2005-09-09 13:04:13 -0700 152) */
9b80a184eaadc (Alexey Dobriyan 2016-09-02 00:38:52 +0300 153) static int expand_fdtable(struct files_struct *files, unsigned int nr)
ab2af1f500506 (Dipankar Sarma 2005-09-09 13:04:13 -0700 154) __releases(files->file_lock)
ab2af1f500506 (Dipankar Sarma 2005-09-09 13:04:13 -0700 155) __acquires(files->file_lock)
ab2af1f500506 (Dipankar Sarma 2005-09-09 13:04:13 -0700 156) {
74d392aaabfc8 (Vadim Lobanov 2006-09-29 02:01:43 -0700 157) struct fdtable *new_fdt, *cur_fdt;
ab2af1f500506 (Dipankar Sarma 2005-09-09 13:04:13 -0700 158)
ab2af1f500506 (Dipankar Sarma 2005-09-09 13:04:13 -0700 159) spin_unlock(&files->file_lock);
74d392aaabfc8 (Vadim Lobanov 2006-09-29 02:01:43 -0700 160) new_fdt = alloc_fdtable(nr);
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 161)
d74ba04d919eb (Eric W. Biederman 2020-11-20 17:14:35 -0600 162) /* make sure all fd_install() have seen resize_in_progress
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 163) * or have finished their rcu_read_lock_sched() section.
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 164) */
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 165) if (atomic_read(&files->count) > 1)
c93ffc15cceb0 (Paul E. McKenney 2018-11-05 17:31:31 -0800 166) synchronize_rcu();
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 167)
ab2af1f500506 (Dipankar Sarma 2005-09-09 13:04:13 -0700 168) spin_lock(&files->file_lock);
74d392aaabfc8 (Vadim Lobanov 2006-09-29 02:01:43 -0700 169) if (!new_fdt)
74d392aaabfc8 (Vadim Lobanov 2006-09-29 02:01:43 -0700 170) return -ENOMEM;
5c598b3428c37 (Al Viro 2008-04-27 20:04:15 -0400 171) /*
5c598b3428c37 (Al Viro 2008-04-27 20:04:15 -0400 172) * extremely unlikely race - sysctl_nr_open decreased between the check in
5c598b3428c37 (Al Viro 2008-04-27 20:04:15 -0400 173) * caller and alloc_fdtable(). Cheaper to catch it here...
5c598b3428c37 (Al Viro 2008-04-27 20:04:15 -0400 174) */
5c598b3428c37 (Al Viro 2008-04-27 20:04:15 -0400 175) if (unlikely(new_fdt->max_fds <= nr)) {
a892e2d7dcdfa (Changli Gao 2010-08-10 18:01:35 -0700 176) __free_fdtable(new_fdt);
5c598b3428c37 (Al Viro 2008-04-27 20:04:15 -0400 177) return -EMFILE;
5c598b3428c37 (Al Viro 2008-04-27 20:04:15 -0400 178) }
74d392aaabfc8 (Vadim Lobanov 2006-09-29 02:01:43 -0700 179) cur_fdt = files_fdtable(files);
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 180) BUG_ON(nr < cur_fdt->max_fds);
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 181) copy_fdtable(new_fdt, cur_fdt);
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 182) rcu_assign_pointer(files->fdt, new_fdt);
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 183) if (cur_fdt != &files->fdtab)
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 184) call_rcu(&cur_fdt->rcu, free_fdtable_rcu);
d74ba04d919eb (Eric W. Biederman 2020-11-20 17:14:35 -0600 185) /* coupled with smp_rmb() in fd_install() */
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 186) smp_wmb();
74d392aaabfc8 (Vadim Lobanov 2006-09-29 02:01:43 -0700 187) return 1;
^1da177e4c3f4 (Linus Torvalds 2005-04-16 15:20:36 -0700 188) }
^1da177e4c3f4 (Linus Torvalds 2005-04-16 15:20:36 -0700 189)
^1da177e4c3f4 (Linus Torvalds 2005-04-16 15:20:36 -0700 190) /*
^1da177e4c3f4 (Linus Torvalds 2005-04-16 15:20:36 -0700 191) * Expand files.
74d392aaabfc8 (Vadim Lobanov 2006-09-29 02:01:43 -0700 192) * This function will expand the file structures, if the requested size exceeds
74d392aaabfc8 (Vadim Lobanov 2006-09-29 02:01:43 -0700 193) * the current capacity and there is room for expansion.
74d392aaabfc8 (Vadim Lobanov 2006-09-29 02:01:43 -0700 194) * Return <0 error code on error; 0 when nothing done; 1 when files were
74d392aaabfc8 (Vadim Lobanov 2006-09-29 02:01:43 -0700 195) * expanded and execution may have blocked.
74d392aaabfc8 (Vadim Lobanov 2006-09-29 02:01:43 -0700 196) * The files->file_lock should be held on entry, and will be held on exit.
^1da177e4c3f4 (Linus Torvalds 2005-04-16 15:20:36 -0700 197) */
9b80a184eaadc (Alexey Dobriyan 2016-09-02 00:38:52 +0300 198) static int expand_files(struct files_struct *files, unsigned int nr)
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 199) __releases(files->file_lock)
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 200) __acquires(files->file_lock)
^1da177e4c3f4 (Linus Torvalds 2005-04-16 15:20:36 -0700 201) {
badf16621c1f9 (Dipankar Sarma 2005-09-09 13:04:10 -0700 202) struct fdtable *fdt;
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 203) int expanded = 0;
^1da177e4c3f4 (Linus Torvalds 2005-04-16 15:20:36 -0700 204)
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 205) repeat:
badf16621c1f9 (Dipankar Sarma 2005-09-09 13:04:10 -0700 206) fdt = files_fdtable(files);
4e1e018ecc6f7 (Al Viro 2008-07-26 16:01:20 -0400 207)
74d392aaabfc8 (Vadim Lobanov 2006-09-29 02:01:43 -0700 208) /* Do we need to expand? */
bbea9f69668a3 (Vadim Lobanov 2006-12-10 02:21:12 -0800 209) if (nr < fdt->max_fds)
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 210) return expanded;
4e1e018ecc6f7 (Al Viro 2008-07-26 16:01:20 -0400 211)
74d392aaabfc8 (Vadim Lobanov 2006-09-29 02:01:43 -0700 212) /* Can we expand? */
9cfe015aa424b (Eric Dumazet 2008-02-06 01:37:16 -0800 213) if (nr >= sysctl_nr_open)
74d392aaabfc8 (Vadim Lobanov 2006-09-29 02:01:43 -0700 214) return -EMFILE;
74d392aaabfc8 (Vadim Lobanov 2006-09-29 02:01:43 -0700 215)
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 216) if (unlikely(files->resize_in_progress)) {
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 217) spin_unlock(&files->file_lock);
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 218) expanded = 1;
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 219) wait_event(files->resize_wait, !files->resize_in_progress);
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 220) spin_lock(&files->file_lock);
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 221) goto repeat;
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 222) }
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 223)
74d392aaabfc8 (Vadim Lobanov 2006-09-29 02:01:43 -0700 224) /* All good, so we try */
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 225) files->resize_in_progress = true;
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 226) expanded = expand_fdtable(files, nr);
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 227) files->resize_in_progress = false;
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 228)
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 229) wake_up_all(&files->resize_wait);
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 230) return expanded;
^1da177e4c3f4 (Linus Torvalds 2005-04-16 15:20:36 -0700 231) }
ab2af1f500506 (Dipankar Sarma 2005-09-09 13:04:13 -0700 232)
9b80a184eaadc (Alexey Dobriyan 2016-09-02 00:38:52 +0300 233) static inline void __set_close_on_exec(unsigned int fd, struct fdtable *fdt)
b8318b01a8f7f (Al Viro 2012-08-21 20:09:42 -0400 234) {
b8318b01a8f7f (Al Viro 2012-08-21 20:09:42 -0400 235) __set_bit(fd, fdt->close_on_exec);
b8318b01a8f7f (Al Viro 2012-08-21 20:09:42 -0400 236) }
b8318b01a8f7f (Al Viro 2012-08-21 20:09:42 -0400 237)
9b80a184eaadc (Alexey Dobriyan 2016-09-02 00:38:52 +0300 238) static inline void __clear_close_on_exec(unsigned int fd, struct fdtable *fdt)
b8318b01a8f7f (Al Viro 2012-08-21 20:09:42 -0400 239) {
fc90888d07b8e (Linus Torvalds 2015-10-31 16:06:40 -0700 240) if (test_bit(fd, fdt->close_on_exec))
fc90888d07b8e (Linus Torvalds 2015-10-31 16:06:40 -0700 241) __clear_bit(fd, fdt->close_on_exec);
b8318b01a8f7f (Al Viro 2012-08-21 20:09:42 -0400 242) }
b8318b01a8f7f (Al Viro 2012-08-21 20:09:42 -0400 243)
f3f86e33dc3da (Linus Torvalds 2015-10-30 16:53:57 -0700 244) static inline void __set_open_fd(unsigned int fd, struct fdtable *fdt)
b8318b01a8f7f (Al Viro 2012-08-21 20:09:42 -0400 245) {
b8318b01a8f7f (Al Viro 2012-08-21 20:09:42 -0400 246) __set_bit(fd, fdt->open_fds);
f3f86e33dc3da (Linus Torvalds 2015-10-30 16:53:57 -0700 247) fd /= BITS_PER_LONG;
f3f86e33dc3da (Linus Torvalds 2015-10-30 16:53:57 -0700 248) if (!~fdt->open_fds[fd])
f3f86e33dc3da (Linus Torvalds 2015-10-30 16:53:57 -0700 249) __set_bit(fd, fdt->full_fds_bits);
b8318b01a8f7f (Al Viro 2012-08-21 20:09:42 -0400 250) }
b8318b01a8f7f (Al Viro 2012-08-21 20:09:42 -0400 251)
f3f86e33dc3da (Linus Torvalds 2015-10-30 16:53:57 -0700 252) static inline void __clear_open_fd(unsigned int fd, struct fdtable *fdt)
b8318b01a8f7f (Al Viro 2012-08-21 20:09:42 -0400 253) {
b8318b01a8f7f (Al Viro 2012-08-21 20:09:42 -0400 254) __clear_bit(fd, fdt->open_fds);
f3f86e33dc3da (Linus Torvalds 2015-10-30 16:53:57 -0700 255) __clear_bit(fd / BITS_PER_LONG, fdt->full_fds_bits);
b8318b01a8f7f (Al Viro 2012-08-21 20:09:42 -0400 256) }
b8318b01a8f7f (Al Viro 2012-08-21 20:09:42 -0400 257)
9b80a184eaadc (Alexey Dobriyan 2016-09-02 00:38:52 +0300 258) static unsigned int count_open_files(struct fdtable *fdt)
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 259) {
9b80a184eaadc (Alexey Dobriyan 2016-09-02 00:38:52 +0300 260) unsigned int size = fdt->max_fds;
9b80a184eaadc (Alexey Dobriyan 2016-09-02 00:38:52 +0300 261) unsigned int i;
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 262)
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 263) /* Find the last open fd */
1fd36adcd98c1 (David Howells 2012-02-16 17:49:54 +0000 264) for (i = size / BITS_PER_LONG; i > 0; ) {
1fd36adcd98c1 (David Howells 2012-02-16 17:49:54 +0000 265) if (fdt->open_fds[--i])
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 266) break;
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 267) }
1fd36adcd98c1 (David Howells 2012-02-16 17:49:54 +0000 268) i = (i + 1) * BITS_PER_LONG;
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 269) return i;
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 270) }
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 271)
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 272) static unsigned int sane_fdtable_size(struct fdtable *fdt, unsigned int max_fds)
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 273) {
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 274) unsigned int count;
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 275)
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 276) count = count_open_files(fdt);
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 277) if (max_fds < NR_OPEN_DEFAULT)
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 278) max_fds = NR_OPEN_DEFAULT;
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 279) return min(count, max_fds);
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 280) }
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 281)
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 282) /*
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 283) * Allocate a new files structure and copy contents from the
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 284) * passed in files structure.
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 285) * errorp will be valid only when the returned files_struct is NULL.
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 286) */
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 287) struct files_struct *dup_fd(struct files_struct *oldf, unsigned int max_fds, int *errorp)
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 288) {
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 289) struct files_struct *newf;
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 290) struct file **old_fds, **new_fds;
9b80a184eaadc (Alexey Dobriyan 2016-09-02 00:38:52 +0300 291) unsigned int open_files, i;
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 292) struct fdtable *old_fdt, *new_fdt;
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 293)
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 294) *errorp = -ENOMEM;
afbec7fff4928 (Al Viro 2008-05-08 21:11:17 -0400 295) newf = kmem_cache_alloc(files_cachep, GFP_KERNEL);
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 296) if (!newf)
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 297) goto out;
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 298)
afbec7fff4928 (Al Viro 2008-05-08 21:11:17 -0400 299) atomic_set(&newf->count, 1);
afbec7fff4928 (Al Viro 2008-05-08 21:11:17 -0400 300)
afbec7fff4928 (Al Viro 2008-05-08 21:11:17 -0400 301) spin_lock_init(&newf->file_lock);
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 302) newf->resize_in_progress = false;
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 303) init_waitqueue_head(&newf->resize_wait);
afbec7fff4928 (Al Viro 2008-05-08 21:11:17 -0400 304) newf->next_fd = 0;
afbec7fff4928 (Al Viro 2008-05-08 21:11:17 -0400 305) new_fdt = &newf->fdtab;
afbec7fff4928 (Al Viro 2008-05-08 21:11:17 -0400 306) new_fdt->max_fds = NR_OPEN_DEFAULT;
1fd36adcd98c1 (David Howells 2012-02-16 17:49:54 +0000 307) new_fdt->close_on_exec = newf->close_on_exec_init;
1fd36adcd98c1 (David Howells 2012-02-16 17:49:54 +0000 308) new_fdt->open_fds = newf->open_fds_init;
f3f86e33dc3da (Linus Torvalds 2015-10-30 16:53:57 -0700 309) new_fdt->full_fds_bits = newf->full_fds_bits_init;
afbec7fff4928 (Al Viro 2008-05-08 21:11:17 -0400 310) new_fdt->fd = &newf->fd_array[0];
afbec7fff4928 (Al Viro 2008-05-08 21:11:17 -0400 311)
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 312) spin_lock(&oldf->file_lock);
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 313) old_fdt = files_fdtable(oldf);
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 314) open_files = sane_fdtable_size(old_fdt, max_fds);
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 315)
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 316) /*
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 317) * Check whether we need to allocate a larger fd array and fd set.
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 318) */
adbecb128cd2c (Al Viro 2008-05-08 21:19:42 -0400 319) while (unlikely(open_files > new_fdt->max_fds)) {
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 320) spin_unlock(&oldf->file_lock);
9dec3c4d306b0 (Al Viro 2008-05-08 21:02:45 -0400 321)
a892e2d7dcdfa (Changli Gao 2010-08-10 18:01:35 -0700 322) if (new_fdt != &newf->fdtab)
a892e2d7dcdfa (Changli Gao 2010-08-10 18:01:35 -0700 323) __free_fdtable(new_fdt);
adbecb128cd2c (Al Viro 2008-05-08 21:19:42 -0400 324)
9dec3c4d306b0 (Al Viro 2008-05-08 21:02:45 -0400 325) new_fdt = alloc_fdtable(open_files - 1);
9dec3c4d306b0 (Al Viro 2008-05-08 21:02:45 -0400 326) if (!new_fdt) {
9dec3c4d306b0 (Al Viro 2008-05-08 21:02:45 -0400 327) *errorp = -ENOMEM;
9dec3c4d306b0 (Al Viro 2008-05-08 21:02:45 -0400 328) goto out_release;
9dec3c4d306b0 (Al Viro 2008-05-08 21:02:45 -0400 329) }
9dec3c4d306b0 (Al Viro 2008-05-08 21:02:45 -0400 330)
9dec3c4d306b0 (Al Viro 2008-05-08 21:02:45 -0400 331) /* beyond sysctl_nr_open; nothing to do */
9dec3c4d306b0 (Al Viro 2008-05-08 21:02:45 -0400 332) if (unlikely(new_fdt->max_fds < open_files)) {
a892e2d7dcdfa (Changli Gao 2010-08-10 18:01:35 -0700 333) __free_fdtable(new_fdt);
9dec3c4d306b0 (Al Viro 2008-05-08 21:02:45 -0400 334) *errorp = -EMFILE;
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 335) goto out_release;
9dec3c4d306b0 (Al Viro 2008-05-08 21:02:45 -0400 336) }
9dec3c4d306b0 (Al Viro 2008-05-08 21:02:45 -0400 337)
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 338) /*
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 339) * Reacquire the oldf lock and a pointer to its fd table
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 340) * who knows it may have a new bigger fd table. We need
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 341) * the latest pointer.
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 342) */
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 343) spin_lock(&oldf->file_lock);
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 344) old_fdt = files_fdtable(oldf);
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 345) open_files = sane_fdtable_size(old_fdt, max_fds);
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 346) }
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 347)
ea5c58e70c3a1 (Eric Biggers 2015-11-06 00:32:04 -0600 348) copy_fd_bitmaps(new_fdt, old_fdt, open_files);
ea5c58e70c3a1 (Eric Biggers 2015-11-06 00:32:04 -0600 349)
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 350) old_fds = old_fdt->fd;
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 351) new_fds = new_fdt->fd;
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 352)
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 353) for (i = open_files; i != 0; i--) {
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 354) struct file *f = *old_fds++;
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 355) if (f) {
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 356) get_file(f);
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 357) } else {
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 358) /*
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 359) * The fd may be claimed in the fd bitmap but not yet
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 360) * instantiated in the files array if a sibling thread
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 361) * is partway through open(). So make sure that this
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 362) * fd is available to the new process.
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 363) */
1dce27c5aa677 (David Howells 2012-02-16 17:49:42 +0000 364) __clear_open_fd(open_files - i, new_fdt);
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 365) }
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 366) rcu_assign_pointer(*new_fds++, f);
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 367) }
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 368) spin_unlock(&oldf->file_lock);
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 369)
ea5c58e70c3a1 (Eric Biggers 2015-11-06 00:32:04 -0600 370) /* clear the remainder */
ea5c58e70c3a1 (Eric Biggers 2015-11-06 00:32:04 -0600 371) memset(new_fds, 0, (new_fdt->max_fds - open_files) * sizeof(struct file *));
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 372)
afbec7fff4928 (Al Viro 2008-05-08 21:11:17 -0400 373) rcu_assign_pointer(newf->fdt, new_fdt);
afbec7fff4928 (Al Viro 2008-05-08 21:11:17 -0400 374)
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 375) return newf;
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 376)
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 377) out_release:
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 378) kmem_cache_free(files_cachep, newf);
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 379) out:
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 380) return NULL;
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 381) }
02afc6267f6d5 (Al Viro 2008-05-08 19:42:56 -0400 382)
ce08b62d18b3f (Oleg Nesterov 2014-01-11 19:19:53 +0100 383) static struct fdtable *close_files(struct files_struct * files)
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 384) {
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 385) /*
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 386) * It is safe to dereference the fd table without RCU or
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 387) * ->file_lock because this is the last reference to the
ce08b62d18b3f (Oleg Nesterov 2014-01-11 19:19:53 +0100 388) * files structure.
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 389) */
ce08b62d18b3f (Oleg Nesterov 2014-01-11 19:19:53 +0100 390) struct fdtable *fdt = rcu_dereference_raw(files->fdt);
9b80a184eaadc (Alexey Dobriyan 2016-09-02 00:38:52 +0300 391) unsigned int i, j = 0;
ce08b62d18b3f (Oleg Nesterov 2014-01-11 19:19:53 +0100 392)
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 393) for (;;) {
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 394) unsigned long set;
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 395) i = j * BITS_PER_LONG;
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 396) if (i >= fdt->max_fds)
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 397) break;
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 398) set = fdt->open_fds[j++];
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 399) while (set) {
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 400) if (set & 1) {
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 401) struct file * file = xchg(&fdt->fd[i], NULL);
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 402) if (file) {
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 403) filp_close(file, files);
388a4c88064e7 (Paul E. McKenney 2017-10-24 08:39:34 -0700 404) cond_resched();
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 405) }
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 406) }
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 407) i++;
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 408) set >>= 1;
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 409) }
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 410) }
ce08b62d18b3f (Oleg Nesterov 2014-01-11 19:19:53 +0100 411)
ce08b62d18b3f (Oleg Nesterov 2014-01-11 19:19:53 +0100 412) return fdt;
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 413) }
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 414)
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 415) void put_files_struct(struct files_struct *files)
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 416) {
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 417) if (atomic_dec_and_test(&files->count)) {
ce08b62d18b3f (Oleg Nesterov 2014-01-11 19:19:53 +0100 418) struct fdtable *fdt = close_files(files);
ce08b62d18b3f (Oleg Nesterov 2014-01-11 19:19:53 +0100 419)
b9e02af0ae078 (Al Viro 2012-08-15 20:00:58 -0400 420) /* free the arrays if they are not embedded */
b9e02af0ae078 (Al Viro 2012-08-15 20:00:58 -0400 421) if (fdt != &files->fdtab)
b9e02af0ae078 (Al Viro 2012-08-15 20:00:58 -0400 422) __free_fdtable(fdt);
b9e02af0ae078 (Al Viro 2012-08-15 20:00:58 -0400 423) kmem_cache_free(files_cachep, files);
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 424) }
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 425) }
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 426)
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 427) void exit_files(struct task_struct *tsk)
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 428) {
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 429) struct files_struct * files = tsk->files;
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 430)
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 431) if (files) {
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 432) task_lock(tsk);
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 433) tsk->files = NULL;
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 434) task_unlock(tsk);
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 435) put_files_struct(files);
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 436) }
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 437) }
7cf4dc3c8dbfd (Al Viro 2012-08-15 19:56:12 -0400 438)
f52111b154694 (Al Viro 2008-05-08 18:19:16 -0400 439) struct files_struct init_files = {
f52111b154694 (Al Viro 2008-05-08 18:19:16 -0400 440) .count = ATOMIC_INIT(1),
f52111b154694 (Al Viro 2008-05-08 18:19:16 -0400 441) .fdt = &init_files.fdtab,
f52111b154694 (Al Viro 2008-05-08 18:19:16 -0400 442) .fdtab = {
f52111b154694 (Al Viro 2008-05-08 18:19:16 -0400 443) .max_fds = NR_OPEN_DEFAULT,
f52111b154694 (Al Viro 2008-05-08 18:19:16 -0400 444) .fd = &init_files.fd_array[0],
1fd36adcd98c1 (David Howells 2012-02-16 17:49:54 +0000 445) .close_on_exec = init_files.close_on_exec_init,
1fd36adcd98c1 (David Howells 2012-02-16 17:49:54 +0000 446) .open_fds = init_files.open_fds_init,
f3f86e33dc3da (Linus Torvalds 2015-10-30 16:53:57 -0700 447) .full_fds_bits = init_files.full_fds_bits_init,
f52111b154694 (Al Viro 2008-05-08 18:19:16 -0400 448) },
eece09ec213e9 (Thomas Gleixner 2011-07-17 21:25:03 +0200 449) .file_lock = __SPIN_LOCK_UNLOCKED(init_files.file_lock),
5704a06810682 (Shuriyc Chu 2019-03-05 15:41:56 -0800 450) .resize_wait = __WAIT_QUEUE_HEAD_INITIALIZER(init_files.resize_wait),
f52111b154694 (Al Viro 2008-05-08 18:19:16 -0400 451) };
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 452)
9b80a184eaadc (Alexey Dobriyan 2016-09-02 00:38:52 +0300 453) static unsigned int find_next_fd(struct fdtable *fdt, unsigned int start)
f3f86e33dc3da (Linus Torvalds 2015-10-30 16:53:57 -0700 454) {
9b80a184eaadc (Alexey Dobriyan 2016-09-02 00:38:52 +0300 455) unsigned int maxfd = fdt->max_fds;
9b80a184eaadc (Alexey Dobriyan 2016-09-02 00:38:52 +0300 456) unsigned int maxbit = maxfd / BITS_PER_LONG;
9b80a184eaadc (Alexey Dobriyan 2016-09-02 00:38:52 +0300 457) unsigned int bitbit = start / BITS_PER_LONG;
f3f86e33dc3da (Linus Torvalds 2015-10-30 16:53:57 -0700 458)
f3f86e33dc3da (Linus Torvalds 2015-10-30 16:53:57 -0700 459) bitbit = find_next_zero_bit(fdt->full_fds_bits, maxbit, bitbit) * BITS_PER_LONG;
f3f86e33dc3da (Linus Torvalds 2015-10-30 16:53:57 -0700 460) if (bitbit > maxfd)
f3f86e33dc3da (Linus Torvalds 2015-10-30 16:53:57 -0700 461) return maxfd;
f3f86e33dc3da (Linus Torvalds 2015-10-30 16:53:57 -0700 462) if (bitbit > start)
f3f86e33dc3da (Linus Torvalds 2015-10-30 16:53:57 -0700 463) start = bitbit;
f3f86e33dc3da (Linus Torvalds 2015-10-30 16:53:57 -0700 464) return find_next_zero_bit(fdt->open_fds, maxfd, start);
f3f86e33dc3da (Linus Torvalds 2015-10-30 16:53:57 -0700 465) }
f3f86e33dc3da (Linus Torvalds 2015-10-30 16:53:57 -0700 466)
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 467) /*
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 468) * allocate a file descriptor, mark it busy.
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 469) */
aa384d10f3d06 (Eric W. Biederman 2020-11-20 17:14:37 -0600 470) static int alloc_fd(unsigned start, unsigned end, unsigned flags)
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 471) {
aa384d10f3d06 (Eric W. Biederman 2020-11-20 17:14:37 -0600 472) struct files_struct *files = current->files;
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 473) unsigned int fd;
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 474) int error;
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 475) struct fdtable *fdt;
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 476)
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 477) spin_lock(&files->file_lock);
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 478) repeat:
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 479) fdt = files_fdtable(files);
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 480) fd = start;
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 481) if (fd < files->next_fd)
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 482) fd = files->next_fd;
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 483)
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 484) if (fd < fdt->max_fds)
f3f86e33dc3da (Linus Torvalds 2015-10-30 16:53:57 -0700 485) fd = find_next_fd(fdt, fd);
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 486)
f33ff9927f420 (Al Viro 2012-08-12 16:17:59 -0400 487) /*
f33ff9927f420 (Al Viro 2012-08-12 16:17:59 -0400 488) * N.B. For clone tasks sharing a files structure, this test
f33ff9927f420 (Al Viro 2012-08-12 16:17:59 -0400 489) * will limit the total number of files that can be opened.
f33ff9927f420 (Al Viro 2012-08-12 16:17:59 -0400 490) */
f33ff9927f420 (Al Viro 2012-08-12 16:17:59 -0400 491) error = -EMFILE;
f33ff9927f420 (Al Viro 2012-08-12 16:17:59 -0400 492) if (fd >= end)
f33ff9927f420 (Al Viro 2012-08-12 16:17:59 -0400 493) goto out;
f33ff9927f420 (Al Viro 2012-08-12 16:17:59 -0400 494)
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 495) error = expand_files(files, fd);
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 496) if (error < 0)
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 497) goto out;
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 498)
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 499) /*
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 500) * If we needed to expand the fs array we
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 501) * might have blocked - try again.
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 502) */
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 503) if (error)
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 504) goto repeat;
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 505)
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 506) if (start <= files->next_fd)
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 507) files->next_fd = fd + 1;
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 508)
1dce27c5aa677 (David Howells 2012-02-16 17:49:42 +0000 509) __set_open_fd(fd, fdt);
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 510) if (flags & O_CLOEXEC)
1dce27c5aa677 (David Howells 2012-02-16 17:49:42 +0000 511) __set_close_on_exec(fd, fdt);
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 512) else
1dce27c5aa677 (David Howells 2012-02-16 17:49:42 +0000 513) __clear_close_on_exec(fd, fdt);
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 514) error = fd;
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 515) #if 1
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 516) /* Sanity check */
add1f09954543 (Paul E. McKenney 2014-02-12 12:51:09 -0800 517) if (rcu_access_pointer(fdt->fd[fd]) != NULL) {
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 518) printk(KERN_WARNING "alloc_fd: slot %d not NULL!\n", fd);
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 519) rcu_assign_pointer(fdt->fd[fd], NULL);
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 520) }
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 521) #endif
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 522)
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 523) out:
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 524) spin_unlock(&files->file_lock);
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 525) return error;
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 526) }
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 527)
4022e7af86be2 (Jens Axboe 2020-03-19 19:23:18 -0600 528) int __get_unused_fd_flags(unsigned flags, unsigned long nofile)
4022e7af86be2 (Jens Axboe 2020-03-19 19:23:18 -0600 529) {
aa384d10f3d06 (Eric W. Biederman 2020-11-20 17:14:37 -0600 530) return alloc_fd(0, nofile, flags);
4022e7af86be2 (Jens Axboe 2020-03-19 19:23:18 -0600 531) }
4022e7af86be2 (Jens Axboe 2020-03-19 19:23:18 -0600 532)
1a7bd2265fc57 (Al Viro 2012-08-12 17:18:05 -0400 533) int get_unused_fd_flags(unsigned flags)
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 534) {
4022e7af86be2 (Jens Axboe 2020-03-19 19:23:18 -0600 535) return __get_unused_fd_flags(flags, rlimit(RLIMIT_NOFILE));
1027abe8827b4 (Al Viro 2008-07-30 04:13:04 -0400 536) }
1a7bd2265fc57 (Al Viro 2012-08-12 17:18:05 -0400 537) EXPORT_SYMBOL(get_unused_fd_flags);
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 538)
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 539) static void __put_unused_fd(struct files_struct *files, unsigned int fd)
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 540) {
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 541) struct fdtable *fdt = files_fdtable(files);
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 542) __clear_open_fd(fd, fdt);
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 543) if (fd < files->next_fd)
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 544) files->next_fd = fd;
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 545) }
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 546)
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 547) void put_unused_fd(unsigned int fd)
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 548) {
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 549) struct files_struct *files = current->files;
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 550) spin_lock(&files->file_lock);
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 551) __put_unused_fd(files, fd);
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 552) spin_unlock(&files->file_lock);
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 553) }
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 554)
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 555) EXPORT_SYMBOL(put_unused_fd);
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 556)
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 557) /*
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 558) * Install a file pointer in the fd array.
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 559) *
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 560) * The VFS is full of places where we drop the files lock between
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 561) * setting the open_fds bitmap and installing the file in the file
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 562) * array. At any such point, we are vulnerable to a dup2() race
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 563) * installing a file in the array before us. We need to detect this and
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 564) * fput() the struct file we are about to overwrite in this case.
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 565) *
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 566) * It should never happen - if we allow dup2() do it, _really_ bad things
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 567) * will follow.
f869e8a7f753e (Al Viro 2012-08-15 21:06:33 -0400 568) *
d74ba04d919eb (Eric W. Biederman 2020-11-20 17:14:35 -0600 569) * This consumes the "file" refcount, so callers should treat it
d74ba04d919eb (Eric W. Biederman 2020-11-20 17:14:35 -0600 570) * as if they had called fput(file).
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 571) */
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 572)
d74ba04d919eb (Eric W. Biederman 2020-11-20 17:14:35 -0600 573) void fd_install(unsigned int fd, struct file *file)
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 574) {
d74ba04d919eb (Eric W. Biederman 2020-11-20 17:14:35 -0600 575) struct files_struct *files = current->files;
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 576) struct fdtable *fdt;
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 577)
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 578) rcu_read_lock_sched();
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 579)
c02b1a9b41c2e (Mateusz Guzik 2017-10-03 12:58:15 +0200 580) if (unlikely(files->resize_in_progress)) {
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 581) rcu_read_unlock_sched();
c02b1a9b41c2e (Mateusz Guzik 2017-10-03 12:58:15 +0200 582) spin_lock(&files->file_lock);
c02b1a9b41c2e (Mateusz Guzik 2017-10-03 12:58:15 +0200 583) fdt = files_fdtable(files);
c02b1a9b41c2e (Mateusz Guzik 2017-10-03 12:58:15 +0200 584) BUG_ON(fdt->fd[fd] != NULL);
c02b1a9b41c2e (Mateusz Guzik 2017-10-03 12:58:15 +0200 585) rcu_assign_pointer(fdt->fd[fd], file);
c02b1a9b41c2e (Mateusz Guzik 2017-10-03 12:58:15 +0200 586) spin_unlock(&files->file_lock);
c02b1a9b41c2e (Mateusz Guzik 2017-10-03 12:58:15 +0200 587) return;
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 588) }
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 589) /* coupled with smp_wmb() in expand_fdtable() */
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 590) smp_rmb();
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 591) fdt = rcu_dereference_sched(files->fdt);
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 592) BUG_ON(fdt->fd[fd] != NULL);
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 593) rcu_assign_pointer(fdt->fd[fd], file);
8a81252b774b5 (Eric Dumazet 2015-06-30 15:54:08 +0200 594) rcu_read_unlock_sched();
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 595) }
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 596)
56007cae94f34 (Al Viro 2012-08-15 21:03:26 -0400 597) EXPORT_SYMBOL(fd_install);
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 598)
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 599) static struct file *pick_file(struct files_struct *files, unsigned fd)
483ce1d4b8c3b (Al Viro 2012-08-19 12:04:24 -0400 600) {
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 601) struct file *file = NULL;
483ce1d4b8c3b (Al Viro 2012-08-19 12:04:24 -0400 602) struct fdtable *fdt;
483ce1d4b8c3b (Al Viro 2012-08-19 12:04:24 -0400 603)
483ce1d4b8c3b (Al Viro 2012-08-19 12:04:24 -0400 604) spin_lock(&files->file_lock);
483ce1d4b8c3b (Al Viro 2012-08-19 12:04:24 -0400 605) fdt = files_fdtable(files);
483ce1d4b8c3b (Al Viro 2012-08-19 12:04:24 -0400 606) if (fd >= fdt->max_fds)
483ce1d4b8c3b (Al Viro 2012-08-19 12:04:24 -0400 607) goto out_unlock;
483ce1d4b8c3b (Al Viro 2012-08-19 12:04:24 -0400 608) file = fdt->fd[fd];
483ce1d4b8c3b (Al Viro 2012-08-19 12:04:24 -0400 609) if (!file)
483ce1d4b8c3b (Al Viro 2012-08-19 12:04:24 -0400 610) goto out_unlock;
483ce1d4b8c3b (Al Viro 2012-08-19 12:04:24 -0400 611) rcu_assign_pointer(fdt->fd[fd], NULL);
483ce1d4b8c3b (Al Viro 2012-08-19 12:04:24 -0400 612) __put_unused_fd(files, fd);
483ce1d4b8c3b (Al Viro 2012-08-19 12:04:24 -0400 613)
483ce1d4b8c3b (Al Viro 2012-08-19 12:04:24 -0400 614) out_unlock:
483ce1d4b8c3b (Al Viro 2012-08-19 12:04:24 -0400 615) spin_unlock(&files->file_lock);
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 616) return file;
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 617) }
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 618)
8760c909f54a8 (Eric W. Biederman 2020-11-20 17:14:38 -0600 619) int close_fd(unsigned fd)
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 620) {
8760c909f54a8 (Eric W. Biederman 2020-11-20 17:14:38 -0600 621) struct files_struct *files = current->files;
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 622) struct file *file;
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 623)
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 624) file = pick_file(files, fd);
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 625) if (!file)
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 626) return -EBADF;
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 627)
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 628) return filp_close(file, files);
483ce1d4b8c3b (Al Viro 2012-08-19 12:04:24 -0400 629) }
8760c909f54a8 (Eric W. Biederman 2020-11-20 17:14:38 -0600 630) EXPORT_SYMBOL(close_fd); /* for ksys_close() */
483ce1d4b8c3b (Al Viro 2012-08-19 12:04:24 -0400 631)
9b5b872215fe6 (Christian Brauner 2021-04-02 10:29:36 +0200 632) /**
9b5b872215fe6 (Christian Brauner 2021-04-02 10:29:36 +0200 633) * last_fd - return last valid index into fd table
9b5b872215fe6 (Christian Brauner 2021-04-02 10:29:36 +0200 634) * @cur_fds: files struct
9b5b872215fe6 (Christian Brauner 2021-04-02 10:29:36 +0200 635) *
9b5b872215fe6 (Christian Brauner 2021-04-02 10:29:36 +0200 636) * Context: Either rcu read lock or files_lock must be held.
9b5b872215fe6 (Christian Brauner 2021-04-02 10:29:36 +0200 637) *
9b5b872215fe6 (Christian Brauner 2021-04-02 10:29:36 +0200 638) * Returns: Last valid index into fdtable.
9b5b872215fe6 (Christian Brauner 2021-04-02 10:29:36 +0200 639) */
9b5b872215fe6 (Christian Brauner 2021-04-02 10:29:36 +0200 640) static inline unsigned last_fd(struct fdtable *fdt)
9b5b872215fe6 (Christian Brauner 2021-04-02 10:29:36 +0200 641) {
9b5b872215fe6 (Christian Brauner 2021-04-02 10:29:36 +0200 642) return fdt->max_fds - 1;
9b5b872215fe6 (Christian Brauner 2021-04-02 10:29:36 +0200 643) }
9b5b872215fe6 (Christian Brauner 2021-04-02 10:29:36 +0200 644)
582f1fb6b721f (Giuseppe Scrivano 2020-11-18 11:47:45 +0100 645) static inline void __range_cloexec(struct files_struct *cur_fds,
582f1fb6b721f (Giuseppe Scrivano 2020-11-18 11:47:45 +0100 646) unsigned int fd, unsigned int max_fd)
582f1fb6b721f (Giuseppe Scrivano 2020-11-18 11:47:45 +0100 647) {
582f1fb6b721f (Giuseppe Scrivano 2020-11-18 11:47:45 +0100 648) struct fdtable *fdt;
582f1fb6b721f (Giuseppe Scrivano 2020-11-18 11:47:45 +0100 649)
9b5b872215fe6 (Christian Brauner 2021-04-02 10:29:36 +0200 650) /* make sure we're using the correct maximum value */
582f1fb6b721f (Giuseppe Scrivano 2020-11-18 11:47:45 +0100 651) spin_lock(&cur_fds->file_lock);
582f1fb6b721f (Giuseppe Scrivano 2020-11-18 11:47:45 +0100 652) fdt = files_fdtable(cur_fds);
9b5b872215fe6 (Christian Brauner 2021-04-02 10:29:36 +0200 653) max_fd = min(last_fd(fdt), max_fd);
9b5b872215fe6 (Christian Brauner 2021-04-02 10:29:36 +0200 654) if (fd <= max_fd)
9b5b872215fe6 (Christian Brauner 2021-04-02 10:29:36 +0200 655) bitmap_set(fdt->close_on_exec, fd, max_fd - fd + 1);
582f1fb6b721f (Giuseppe Scrivano 2020-11-18 11:47:45 +0100 656) spin_unlock(&cur_fds->file_lock);
582f1fb6b721f (Giuseppe Scrivano 2020-11-18 11:47:45 +0100 657) }
582f1fb6b721f (Giuseppe Scrivano 2020-11-18 11:47:45 +0100 658)
582f1fb6b721f (Giuseppe Scrivano 2020-11-18 11:47:45 +0100 659) static inline void __range_close(struct files_struct *cur_fds, unsigned int fd,
582f1fb6b721f (Giuseppe Scrivano 2020-11-18 11:47:45 +0100 660) unsigned int max_fd)
582f1fb6b721f (Giuseppe Scrivano 2020-11-18 11:47:45 +0100 661) {
582f1fb6b721f (Giuseppe Scrivano 2020-11-18 11:47:45 +0100 662) while (fd <= max_fd) {
582f1fb6b721f (Giuseppe Scrivano 2020-11-18 11:47:45 +0100 663) struct file *file;
582f1fb6b721f (Giuseppe Scrivano 2020-11-18 11:47:45 +0100 664)
582f1fb6b721f (Giuseppe Scrivano 2020-11-18 11:47:45 +0100 665) file = pick_file(cur_fds, fd++);
582f1fb6b721f (Giuseppe Scrivano 2020-11-18 11:47:45 +0100 666) if (!file)
582f1fb6b721f (Giuseppe Scrivano 2020-11-18 11:47:45 +0100 667) continue;
582f1fb6b721f (Giuseppe Scrivano 2020-11-18 11:47:45 +0100 668)
582f1fb6b721f (Giuseppe Scrivano 2020-11-18 11:47:45 +0100 669) filp_close(file, cur_fds);
582f1fb6b721f (Giuseppe Scrivano 2020-11-18 11:47:45 +0100 670) cond_resched();
582f1fb6b721f (Giuseppe Scrivano 2020-11-18 11:47:45 +0100 671) }
582f1fb6b721f (Giuseppe Scrivano 2020-11-18 11:47:45 +0100 672) }
582f1fb6b721f (Giuseppe Scrivano 2020-11-18 11:47:45 +0100 673)
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 674) /**
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 675) * __close_range() - Close all file descriptors in a given range.
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 676) *
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 677) * @fd: starting file descriptor to close
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 678) * @max_fd: last file descriptor to close
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 679) *
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 680) * This closes a range of file descriptors. All file descriptors
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 681) * from @fd up to and including @max_fd are closed.
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 682) */
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 683) int __close_range(unsigned fd, unsigned max_fd, unsigned int flags)
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 684) {
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 685) unsigned int cur_max;
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 686) struct task_struct *me = current;
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 687) struct files_struct *cur_fds = me->files, *fds = NULL;
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 688)
582f1fb6b721f (Giuseppe Scrivano 2020-11-18 11:47:45 +0100 689) if (flags & ~(CLOSE_RANGE_UNSHARE | CLOSE_RANGE_CLOEXEC))
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 690) return -EINVAL;
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 691)
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 692) if (fd > max_fd)
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 693) return -EINVAL;
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 694)
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 695) rcu_read_lock();
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 696) cur_max = files_fdtable(cur_fds)->max_fds;
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 697) rcu_read_unlock();
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 698)
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 699) /* cap to last valid index into fdtable */
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 700) cur_max--;
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 701)
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 702) if (flags & CLOSE_RANGE_UNSHARE) {
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 703) int ret;
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 704) unsigned int max_unshare_fds = NR_OPEN_MAX;
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 705)
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 706) /*
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 707) * If the requested range is greater than the current maximum,
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 708) * we're closing everything so only copy all file descriptors
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 709) * beneath the lowest file descriptor.
fec8a6a691033 (Christian Brauner 2020-12-17 22:33:03 +0100 710) * If the caller requested all fds to be made cloexec copy all
fec8a6a691033 (Christian Brauner 2020-12-17 22:33:03 +0100 711) * of the file descriptors since they still want to use them.
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 712) */
fec8a6a691033 (Christian Brauner 2020-12-17 22:33:03 +0100 713) if (!(flags & CLOSE_RANGE_CLOEXEC) && (max_fd >= cur_max))
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 714) max_unshare_fds = fd;
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 715)
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 716) ret = unshare_fd(CLONE_FILES, max_unshare_fds, &fds);
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 717) if (ret)
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 718) return ret;
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 719)
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 720) /*
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 721) * We used to share our file descriptor table, and have now
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 722) * created a private one, make sure we're using it below.
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 723) */
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 724) if (fds)
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 725) swap(cur_fds, fds);
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 726) }
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 727)
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 728) max_fd = min(max_fd, cur_max);
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 729)
582f1fb6b721f (Giuseppe Scrivano 2020-11-18 11:47:45 +0100 730) if (flags & CLOSE_RANGE_CLOEXEC)
582f1fb6b721f (Giuseppe Scrivano 2020-11-18 11:47:45 +0100 731) __range_cloexec(cur_fds, fd, max_fd);
582f1fb6b721f (Giuseppe Scrivano 2020-11-18 11:47:45 +0100 732) else
582f1fb6b721f (Giuseppe Scrivano 2020-11-18 11:47:45 +0100 733) __range_close(cur_fds, fd, max_fd);
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 734)
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 735) if (fds) {
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 736) /*
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 737) * We're done closing the files we were supposed to. Time to install
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 738) * the new file descriptor table and drop the old one.
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 739) */
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 740) task_lock(me);
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 741) me->files = cur_fds;
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 742) task_unlock(me);
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 743) put_files_struct(fds);
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 744) }
60997c3d45d9a (Christian Brauner 2020-06-03 21:48:55 +0200 745)
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 746) return 0;
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 747) }
278a5fbaed89d (Christian Brauner 2019-05-24 11:30:34 +0200 748)
80cd795630d65 (Todd Kjos 2018-12-14 15:58:21 -0800 749) /*
53dec2ea74f2e (Jens Axboe 2021-01-19 15:41:52 -0700 750) * See close_fd_get_file() below, this variant assumes current->files->file_lock
53dec2ea74f2e (Jens Axboe 2021-01-19 15:41:52 -0700 751) * is held.
80cd795630d65 (Todd Kjos 2018-12-14 15:58:21 -0800 752) */
53dec2ea74f2e (Jens Axboe 2021-01-19 15:41:52 -0700 753) int __close_fd_get_file(unsigned int fd, struct file **res)
80cd795630d65 (Todd Kjos 2018-12-14 15:58:21 -0800 754) {
80cd795630d65 (Todd Kjos 2018-12-14 15:58:21 -0800 755) struct files_struct *files = current->files;
80cd795630d65 (Todd Kjos 2018-12-14 15:58:21 -0800 756) struct file *file;
80cd795630d65 (Todd Kjos 2018-12-14 15:58:21 -0800 757) struct fdtable *fdt;
80cd795630d65 (Todd Kjos 2018-12-14 15:58:21 -0800 758)
80cd795630d65 (Todd Kjos 2018-12-14 15:58:21 -0800 759) fdt = files_fdtable(files);
80cd795630d65 (Todd Kjos 2018-12-14 15:58:21 -0800 760) if (fd >= fdt->max_fds)
53dec2ea74f2e (Jens Axboe 2021-01-19 15:41:52 -0700 761) goto out_err;
80cd795630d65 (Todd Kjos 2018-12-14 15:58:21 -0800 762) file = fdt->fd[fd];
80cd795630d65 (Todd Kjos 2018-12-14 15:58:21 -0800 763) if (!file)
53dec2ea74f2e (Jens Axboe 2021-01-19 15:41:52 -0700 764) goto out_err;
80cd795630d65 (Todd Kjos 2018-12-14 15:58:21 -0800 765) rcu_assign_pointer(fdt->fd[fd], NULL);
80cd795630d65 (Todd Kjos 2018-12-14 15:58:21 -0800 766) __put_unused_fd(files, fd);
80cd795630d65 (Todd Kjos 2018-12-14 15:58:21 -0800 767) get_file(file);
80cd795630d65 (Todd Kjos 2018-12-14 15:58:21 -0800 768) *res = file;
6e802a4ba056a (Jens Axboe 2019-12-11 14:10:35 -0700 769) return 0;
53dec2ea74f2e (Jens Axboe 2021-01-19 15:41:52 -0700 770) out_err:
80cd795630d65 (Todd Kjos 2018-12-14 15:58:21 -0800 771) *res = NULL;
80cd795630d65 (Todd Kjos 2018-12-14 15:58:21 -0800 772) return -ENOENT;
80cd795630d65 (Todd Kjos 2018-12-14 15:58:21 -0800 773) }
80cd795630d65 (Todd Kjos 2018-12-14 15:58:21 -0800 774)
53dec2ea74f2e (Jens Axboe 2021-01-19 15:41:52 -0700 775) /*
53dec2ea74f2e (Jens Axboe 2021-01-19 15:41:52 -0700 776) * variant of close_fd that gets a ref on the file for later fput.
53dec2ea74f2e (Jens Axboe 2021-01-19 15:41:52 -0700 777) * The caller must ensure that filp_close() called on the file, and then
53dec2ea74f2e (Jens Axboe 2021-01-19 15:41:52 -0700 778) * an fput().
53dec2ea74f2e (Jens Axboe 2021-01-19 15:41:52 -0700 779) */
53dec2ea74f2e (Jens Axboe 2021-01-19 15:41:52 -0700 780) int close_fd_get_file(unsigned int fd, struct file **res)
53dec2ea74f2e (Jens Axboe 2021-01-19 15:41:52 -0700 781) {
53dec2ea74f2e (Jens Axboe 2021-01-19 15:41:52 -0700 782) struct files_struct *files = current->files;
53dec2ea74f2e (Jens Axboe 2021-01-19 15:41:52 -0700 783) int ret;
53dec2ea74f2e (Jens Axboe 2021-01-19 15:41:52 -0700 784)
53dec2ea74f2e (Jens Axboe 2021-01-19 15:41:52 -0700 785) spin_lock(&files->file_lock);
53dec2ea74f2e (Jens Axboe 2021-01-19 15:41:52 -0700 786) ret = __close_fd_get_file(fd, res);
53dec2ea74f2e (Jens Axboe 2021-01-19 15:41:52 -0700 787) spin_unlock(&files->file_lock);
53dec2ea74f2e (Jens Axboe 2021-01-19 15:41:52 -0700 788)
53dec2ea74f2e (Jens Axboe 2021-01-19 15:41:52 -0700 789) return ret;
53dec2ea74f2e (Jens Axboe 2021-01-19 15:41:52 -0700 790) }
53dec2ea74f2e (Jens Axboe 2021-01-19 15:41:52 -0700 791)
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 792) void do_close_on_exec(struct files_struct *files)
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 793) {
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 794) unsigned i;
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 795) struct fdtable *fdt;
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 796)
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 797) /* exec unshares first */
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 798) spin_lock(&files->file_lock);
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 799) for (i = 0; ; i++) {
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 800) unsigned long set;
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 801) unsigned fd = i * BITS_PER_LONG;
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 802) fdt = files_fdtable(files);
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 803) if (fd >= fdt->max_fds)
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 804) break;
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 805) set = fdt->close_on_exec[i];
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 806) if (!set)
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 807) continue;
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 808) fdt->close_on_exec[i] = 0;
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 809) for ( ; set ; fd++, set >>= 1) {
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 810) struct file *file;
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 811) if (!(set & 1))
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 812) continue;
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 813) file = fdt->fd[fd];
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 814) if (!file)
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 815) continue;
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 816) rcu_assign_pointer(fdt->fd[fd], NULL);
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 817) __put_unused_fd(files, fd);
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 818) spin_unlock(&files->file_lock);
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 819) filp_close(file, files);
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 820) cond_resched();
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 821) spin_lock(&files->file_lock);
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 822) }
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 823)
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 824) }
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 825) spin_unlock(&files->file_lock);
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 826) }
6a6d27de340c8 (Al Viro 2012-08-21 09:56:33 -0400 827)
5e876fb43dbf2 (Sargun Dhillon 2020-01-07 09:59:24 -0800 828) static struct file *__fget_files(struct files_struct *files, unsigned int fd,
5e876fb43dbf2 (Sargun Dhillon 2020-01-07 09:59:24 -0800 829) fmode_t mask, unsigned int refs)
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 830) {
1deb46e256256 (Oleg Nesterov 2014-01-13 16:48:19 +0100 831) struct file *file;
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 832)
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 833) rcu_read_lock();
5ba97d2832f87 (Eric Dumazet 2015-06-29 17:10:30 +0200 834) loop:
f36c294327419 (Eric W. Biederman 2020-11-20 17:14:26 -0600 835) file = files_lookup_fd_rcu(files, fd);
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 836) if (file) {
5ba97d2832f87 (Eric Dumazet 2015-06-29 17:10:30 +0200 837) /* File object ref couldn't be taken.
5ba97d2832f87 (Eric Dumazet 2015-06-29 17:10:30 +0200 838) * dup2() atomicity guarantee is the reason
5ba97d2832f87 (Eric Dumazet 2015-06-29 17:10:30 +0200 839) * we loop to catch the new file (or NULL pointer)
5ba97d2832f87 (Eric Dumazet 2015-06-29 17:10:30 +0200 840) */
5ba97d2832f87 (Eric Dumazet 2015-06-29 17:10:30 +0200 841) if (file->f_mode & mask)
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 842) file = NULL;
091141a42e15f (Jens Axboe 2018-11-21 10:32:39 -0700 843) else if (!get_file_rcu_many(file, refs))
5ba97d2832f87 (Eric Dumazet 2015-06-29 17:10:30 +0200 844) goto loop;
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 845) }
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 846) rcu_read_unlock();
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 847)
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 848) return file;
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 849) }
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 850)
5e876fb43dbf2 (Sargun Dhillon 2020-01-07 09:59:24 -0800 851) static inline struct file *__fget(unsigned int fd, fmode_t mask,
5e876fb43dbf2 (Sargun Dhillon 2020-01-07 09:59:24 -0800 852) unsigned int refs)
5e876fb43dbf2 (Sargun Dhillon 2020-01-07 09:59:24 -0800 853) {
5e876fb43dbf2 (Sargun Dhillon 2020-01-07 09:59:24 -0800 854) return __fget_files(current->files, fd, mask, refs);
5e876fb43dbf2 (Sargun Dhillon 2020-01-07 09:59:24 -0800 855) }
5e876fb43dbf2 (Sargun Dhillon 2020-01-07 09:59:24 -0800 856)
091141a42e15f (Jens Axboe 2018-11-21 10:32:39 -0700 857) struct file *fget_many(unsigned int fd, unsigned int refs)
091141a42e15f (Jens Axboe 2018-11-21 10:32:39 -0700 858) {
091141a42e15f (Jens Axboe 2018-11-21 10:32:39 -0700 859) return __fget(fd, FMODE_PATH, refs);
091141a42e15f (Jens Axboe 2018-11-21 10:32:39 -0700 860) }
091141a42e15f (Jens Axboe 2018-11-21 10:32:39 -0700 861)
1deb46e256256 (Oleg Nesterov 2014-01-13 16:48:19 +0100 862) struct file *fget(unsigned int fd)
1deb46e256256 (Oleg Nesterov 2014-01-13 16:48:19 +0100 863) {
091141a42e15f (Jens Axboe 2018-11-21 10:32:39 -0700 864) return __fget(fd, FMODE_PATH, 1);
1deb46e256256 (Oleg Nesterov 2014-01-13 16:48:19 +0100 865) }
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 866) EXPORT_SYMBOL(fget);
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 867)
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 868) struct file *fget_raw(unsigned int fd)
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 869) {
091141a42e15f (Jens Axboe 2018-11-21 10:32:39 -0700 870) return __fget(fd, 0, 1);
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 871) }
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 872) EXPORT_SYMBOL(fget_raw);
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 873)
5e876fb43dbf2 (Sargun Dhillon 2020-01-07 09:59:24 -0800 874) struct file *fget_task(struct task_struct *task, unsigned int fd)
5e876fb43dbf2 (Sargun Dhillon 2020-01-07 09:59:24 -0800 875) {
5e876fb43dbf2 (Sargun Dhillon 2020-01-07 09:59:24 -0800 876) struct file *file = NULL;
5e876fb43dbf2 (Sargun Dhillon 2020-01-07 09:59:24 -0800 877)
5e876fb43dbf2 (Sargun Dhillon 2020-01-07 09:59:24 -0800 878) task_lock(task);
5e876fb43dbf2 (Sargun Dhillon 2020-01-07 09:59:24 -0800 879) if (task->files)
5e876fb43dbf2 (Sargun Dhillon 2020-01-07 09:59:24 -0800 880) file = __fget_files(task->files, fd, 0, 1);
5e876fb43dbf2 (Sargun Dhillon 2020-01-07 09:59:24 -0800 881) task_unlock(task);
5e876fb43dbf2 (Sargun Dhillon 2020-01-07 09:59:24 -0800 882)
5e876fb43dbf2 (Sargun Dhillon 2020-01-07 09:59:24 -0800 883) return file;
5e876fb43dbf2 (Sargun Dhillon 2020-01-07 09:59:24 -0800 884) }
5e876fb43dbf2 (Sargun Dhillon 2020-01-07 09:59:24 -0800 885)
3a879fb380821 (Eric W. Biederman 2020-11-20 17:14:28 -0600 886) struct file *task_lookup_fd_rcu(struct task_struct *task, unsigned int fd)
3a879fb380821 (Eric W. Biederman 2020-11-20 17:14:28 -0600 887) {
3a879fb380821 (Eric W. Biederman 2020-11-20 17:14:28 -0600 888) /* Must be called with rcu_read_lock held */
3a879fb380821 (Eric W. Biederman 2020-11-20 17:14:28 -0600 889) struct files_struct *files;
3a879fb380821 (Eric W. Biederman 2020-11-20 17:14:28 -0600 890) struct file *file = NULL;
3a879fb380821 (Eric W. Biederman 2020-11-20 17:14:28 -0600 891)
3a879fb380821 (Eric W. Biederman 2020-11-20 17:14:28 -0600 892) task_lock(task);
3a879fb380821 (Eric W. Biederman 2020-11-20 17:14:28 -0600 893) files = task->files;
3a879fb380821 (Eric W. Biederman 2020-11-20 17:14:28 -0600 894) if (files)
3a879fb380821 (Eric W. Biederman 2020-11-20 17:14:28 -0600 895) file = files_lookup_fd_rcu(files, fd);
3a879fb380821 (Eric W. Biederman 2020-11-20 17:14:28 -0600 896) task_unlock(task);
3a879fb380821 (Eric W. Biederman 2020-11-20 17:14:28 -0600 897)
3a879fb380821 (Eric W. Biederman 2020-11-20 17:14:28 -0600 898) return file;
3a879fb380821 (Eric W. Biederman 2020-11-20 17:14:28 -0600 899) }
3a879fb380821 (Eric W. Biederman 2020-11-20 17:14:28 -0600 900)
e9a53aeb5e0a8 (Eric W. Biederman 2020-11-20 17:14:31 -0600 901) struct file *task_lookup_next_fd_rcu(struct task_struct *task, unsigned int *ret_fd)
e9a53aeb5e0a8 (Eric W. Biederman 2020-11-20 17:14:31 -0600 902) {
e9a53aeb5e0a8 (Eric W. Biederman 2020-11-20 17:14:31 -0600 903) /* Must be called with rcu_read_lock held */
e9a53aeb5e0a8 (Eric W. Biederman 2020-11-20 17:14:31 -0600 904) struct files_struct *files;
e9a53aeb5e0a8 (Eric W. Biederman 2020-11-20 17:14:31 -0600 905) unsigned int fd = *ret_fd;
e9a53aeb5e0a8 (Eric W. Biederman 2020-11-20 17:14:31 -0600 906) struct file *file = NULL;
e9a53aeb5e0a8 (Eric W. Biederman 2020-11-20 17:14:31 -0600 907)
e9a53aeb5e0a8 (Eric W. Biederman 2020-11-20 17:14:31 -0600 908) task_lock(task);
e9a53aeb5e0a8 (Eric W. Biederman 2020-11-20 17:14:31 -0600 909) files = task->files;
e9a53aeb5e0a8 (Eric W. Biederman 2020-11-20 17:14:31 -0600 910) if (files) {
e9a53aeb5e0a8 (Eric W. Biederman 2020-11-20 17:14:31 -0600 911) for (; fd < files_fdtable(files)->max_fds; fd++) {
e9a53aeb5e0a8 (Eric W. Biederman 2020-11-20 17:14:31 -0600 912) file = files_lookup_fd_rcu(files, fd);
e9a53aeb5e0a8 (Eric W. Biederman 2020-11-20 17:14:31 -0600 913) if (file)
e9a53aeb5e0a8 (Eric W. Biederman 2020-11-20 17:14:31 -0600 914) break;
e9a53aeb5e0a8 (Eric W. Biederman 2020-11-20 17:14:31 -0600 915) }
e9a53aeb5e0a8 (Eric W. Biederman 2020-11-20 17:14:31 -0600 916) }
e9a53aeb5e0a8 (Eric W. Biederman 2020-11-20 17:14:31 -0600 917) task_unlock(task);
e9a53aeb5e0a8 (Eric W. Biederman 2020-11-20 17:14:31 -0600 918) *ret_fd = fd;
e9a53aeb5e0a8 (Eric W. Biederman 2020-11-20 17:14:31 -0600 919) return file;
e9a53aeb5e0a8 (Eric W. Biederman 2020-11-20 17:14:31 -0600 920) }
e9a53aeb5e0a8 (Eric W. Biederman 2020-11-20 17:14:31 -0600 921)
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 922) /*
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 923) * Lightweight file lookup - no refcnt increment if fd table isn't shared.
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 924) *
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 925) * You can use this instead of fget if you satisfy all of the following
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 926) * conditions:
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 927) * 1) You must call fput_light before exiting the syscall and returning control
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 928) * to userspace (i.e. you cannot remember the returned struct file * after
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 929) * returning to userspace).
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 930) * 2) You must not call filp_close on the returned struct file * in between
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 931) * calls to fget_light and fput_light.
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 932) * 3) You must not clone the current task in between the calls to fget_light
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 933) * and fput_light.
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 934) *
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 935) * The fput_needed flag returned by fget_light should be passed to the
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 936) * corresponding fput_light.
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 937) */
bd2a31d522344 (Al Viro 2014-03-04 14:54:22 -0500 938) static unsigned long __fget_light(unsigned int fd, fmode_t mask)
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 939) {
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 940) struct files_struct *files = current->files;
ad46183445043 (Oleg Nesterov 2014-01-13 16:48:40 +0100 941) struct file *file;
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 942)
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 943) if (atomic_read(&files->count) == 1) {
bebf684bf3309 (Eric W. Biederman 2020-11-20 17:14:24 -0600 944) file = files_lookup_fd_raw(files, fd);
bd2a31d522344 (Al Viro 2014-03-04 14:54:22 -0500 945) if (!file || unlikely(file->f_mode & mask))
bd2a31d522344 (Al Viro 2014-03-04 14:54:22 -0500 946) return 0;
bd2a31d522344 (Al Viro 2014-03-04 14:54:22 -0500 947) return (unsigned long)file;
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 948) } else {
091141a42e15f (Jens Axboe 2018-11-21 10:32:39 -0700 949) file = __fget(fd, mask, 1);
bd2a31d522344 (Al Viro 2014-03-04 14:54:22 -0500 950) if (!file)
bd2a31d522344 (Al Viro 2014-03-04 14:54:22 -0500 951) return 0;
bd2a31d522344 (Al Viro 2014-03-04 14:54:22 -0500 952) return FDPUT_FPUT | (unsigned long)file;
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 953) }
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 954) }
bd2a31d522344 (Al Viro 2014-03-04 14:54:22 -0500 955) unsigned long __fdget(unsigned int fd)
ad46183445043 (Oleg Nesterov 2014-01-13 16:48:40 +0100 956) {
bd2a31d522344 (Al Viro 2014-03-04 14:54:22 -0500 957) return __fget_light(fd, FMODE_PATH);
ad46183445043 (Oleg Nesterov 2014-01-13 16:48:40 +0100 958) }
bd2a31d522344 (Al Viro 2014-03-04 14:54:22 -0500 959) EXPORT_SYMBOL(__fdget);
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 960)
bd2a31d522344 (Al Viro 2014-03-04 14:54:22 -0500 961) unsigned long __fdget_raw(unsigned int fd)
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 962) {
bd2a31d522344 (Al Viro 2014-03-04 14:54:22 -0500 963) return __fget_light(fd, 0);
0ee8cdfe6af05 (Al Viro 2012-08-15 21:12:10 -0400 964) }
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 965)
bd2a31d522344 (Al Viro 2014-03-04 14:54:22 -0500 966) unsigned long __fdget_pos(unsigned int fd)
bd2a31d522344 (Al Viro 2014-03-04 14:54:22 -0500 967) {
99aea68134f3c (Eric Biggers 2014-03-16 15:47:48 -0500 968) unsigned long v = __fdget(fd);
99aea68134f3c (Eric Biggers 2014-03-16 15:47:48 -0500 969) struct file *file = (struct file *)(v & ~3);
bd2a31d522344 (Al Viro 2014-03-04 14:54:22 -0500 970)
2be7d348fe924 (Linus Torvalds 2019-11-26 11:34:06 -0800 971) if (file && (file->f_mode & FMODE_ATOMIC_POS)) {
bd2a31d522344 (Al Viro 2014-03-04 14:54:22 -0500 972) if (file_count(file) > 1) {
bd2a31d522344 (Al Viro 2014-03-04 14:54:22 -0500 973) v |= FDPUT_POS_UNLOCK;
bd2a31d522344 (Al Viro 2014-03-04 14:54:22 -0500 974) mutex_lock(&file->f_pos_lock);
bd2a31d522344 (Al Viro 2014-03-04 14:54:22 -0500 975) }
bd2a31d522344 (Al Viro 2014-03-04 14:54:22 -0500 976) }
99aea68134f3c (Eric Biggers 2014-03-16 15:47:48 -0500 977) return v;
bd2a31d522344 (Al Viro 2014-03-04 14:54:22 -0500 978) }
bd2a31d522344 (Al Viro 2014-03-04 14:54:22 -0500 979)
63b6df14134dd (Al Viro 2016-04-20 17:08:21 -0400 980) void __f_unlock_pos(struct file *f)
63b6df14134dd (Al Viro 2016-04-20 17:08:21 -0400 981) {
63b6df14134dd (Al Viro 2016-04-20 17:08:21 -0400 982) mutex_unlock(&f->f_pos_lock);
63b6df14134dd (Al Viro 2016-04-20 17:08:21 -0400 983) }
63b6df14134dd (Al Viro 2016-04-20 17:08:21 -0400 984)
bd2a31d522344 (Al Viro 2014-03-04 14:54:22 -0500 985) /*
bd2a31d522344 (Al Viro 2014-03-04 14:54:22 -0500 986) * We only lock f_pos if we have threads or if the file might be
bd2a31d522344 (Al Viro 2014-03-04 14:54:22 -0500 987) * shared with another process. In both cases we'll have an elevated
bd2a31d522344 (Al Viro 2014-03-04 14:54:22 -0500 988) * file count (done either by fdget() or by fork()).
bd2a31d522344 (Al Viro 2014-03-04 14:54:22 -0500 989) */
bd2a31d522344 (Al Viro 2014-03-04 14:54:22 -0500 990)
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 991) void set_close_on_exec(unsigned int fd, int flag)
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 992) {
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 993) struct files_struct *files = current->files;
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 994) struct fdtable *fdt;
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 995) spin_lock(&files->file_lock);
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 996) fdt = files_fdtable(files);
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 997) if (flag)
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 998) __set_close_on_exec(fd, fdt);
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 999) else
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1000) __clear_close_on_exec(fd, fdt);
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1001) spin_unlock(&files->file_lock);
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1002) }
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1003)
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1004) bool get_close_on_exec(unsigned int fd)
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1005) {
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1006) struct files_struct *files = current->files;
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1007) struct fdtable *fdt;
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1008) bool res;
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1009) rcu_read_lock();
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1010) fdt = files_fdtable(files);
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1011) res = close_on_exec(fd, fdt);
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1012) rcu_read_unlock();
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1013) return res;
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1014) }
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1015)
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1016) static int do_dup2(struct files_struct *files,
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1017) struct file *file, unsigned fd, unsigned flags)
e983094d6dce5 (Al Viro 2014-08-31 14:12:09 -0400 1018) __releases(&files->file_lock)
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1019) {
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1020) struct file *tofree;
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1021) struct fdtable *fdt;
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1022)
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1023) /*
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1024) * We need to detect attempts to do dup2() over allocated but still
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1025) * not finished descriptor. NB: OpenBSD avoids that at the price of
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1026) * extra work in their equivalent of fget() - they insert struct
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1027) * file immediately after grabbing descriptor, mark it larval if
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1028) * more work (e.g. actual opening) is needed and make sure that
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1029) * fget() treats larval files as absent. Potentially interesting,
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1030) * but while extra work in fget() is trivial, locking implications
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1031) * and amount of surgery on open()-related paths in VFS are not.
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1032) * FreeBSD fails with -EBADF in the same situation, NetBSD "solution"
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1033) * deadlocks in rather amusing ways, AFAICS. All of that is out of
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1034) * scope of POSIX or SUS, since neither considers shared descriptor
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1035) * tables and this condition does not arise without those.
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1036) */
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1037) fdt = files_fdtable(files);
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1038) tofree = fdt->fd[fd];
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1039) if (!tofree && fd_is_open(fd, fdt))
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1040) goto Ebusy;
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1041) get_file(file);
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1042) rcu_assign_pointer(fdt->fd[fd], file);
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1043) __set_open_fd(fd, fdt);
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1044) if (flags & O_CLOEXEC)
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1045) __set_close_on_exec(fd, fdt);
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1046) else
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1047) __clear_close_on_exec(fd, fdt);
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1048) spin_unlock(&files->file_lock);
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1049)
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1050) if (tofree)
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1051) filp_close(tofree, files);
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1052)
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1053) return fd;
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1054)
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1055) Ebusy:
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1056) spin_unlock(&files->file_lock);
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1057) return -EBUSY;
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1058) }
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1059)
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1060) int replace_fd(unsigned fd, struct file *file, unsigned flags)
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1061) {
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1062) int err;
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1063) struct files_struct *files = current->files;
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1064)
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1065) if (!file)
8760c909f54a8 (Eric W. Biederman 2020-11-20 17:14:38 -0600 1066) return close_fd(fd);
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1067)
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1068) if (fd >= rlimit(RLIMIT_NOFILE))
08f05c49749ee (Al Viro 2012-10-31 03:37:48 +0000 1069) return -EBADF;
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1070)
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1071) spin_lock(&files->file_lock);
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1072) err = expand_files(files, fd);
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1073) if (unlikely(err < 0))
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1074) goto out_unlock;
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1075) return do_dup2(files, file, fd, flags);
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1076)
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1077) out_unlock:
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1078) spin_unlock(&files->file_lock);
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1079) return err;
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1080) }
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1081)
6659061045cc9 (Kees Cook 2020-06-10 08:20:05 -0700 1082) /**
6659061045cc9 (Kees Cook 2020-06-10 08:20:05 -0700 1083) * __receive_fd() - Install received file into file descriptor table
6659061045cc9 (Kees Cook 2020-06-10 08:20:05 -0700 1084) * @file: struct file that was received from another process
6659061045cc9 (Kees Cook 2020-06-10 08:20:05 -0700 1085) * @ufd: __user pointer to write new fd number to
6659061045cc9 (Kees Cook 2020-06-10 08:20:05 -0700 1086) * @o_flags: the O_* flags to apply to the new fd entry
6659061045cc9 (Kees Cook 2020-06-10 08:20:05 -0700 1087) *
6659061045cc9 (Kees Cook 2020-06-10 08:20:05 -0700 1088) * Installs a received file into the file descriptor table, with appropriate
deefa7f3505ae (Kees Cook 2020-06-10 20:47:45 -0700 1089) * checks and count updates. Optionally writes the fd number to userspace, if
deefa7f3505ae (Kees Cook 2020-06-10 20:47:45 -0700 1090) * @ufd is non-NULL.
6659061045cc9 (Kees Cook 2020-06-10 08:20:05 -0700 1091) *
6659061045cc9 (Kees Cook 2020-06-10 08:20:05 -0700 1092) * This helper handles its own reference counting of the incoming
6659061045cc9 (Kees Cook 2020-06-10 08:20:05 -0700 1093) * struct file.
6659061045cc9 (Kees Cook 2020-06-10 08:20:05 -0700 1094) *
deefa7f3505ae (Kees Cook 2020-06-10 20:47:45 -0700 1095) * Returns newly install fd or -ve on error.
6659061045cc9 (Kees Cook 2020-06-10 08:20:05 -0700 1096) */
42eb0d54c08a0 (Christoph Hellwig 2021-03-25 09:22:09 +0100 1097) int __receive_fd(struct file *file, int __user *ufd, unsigned int o_flags)
6659061045cc9 (Kees Cook 2020-06-10 08:20:05 -0700 1098) {
6659061045cc9 (Kees Cook 2020-06-10 08:20:05 -0700 1099) int new_fd;
6659061045cc9 (Kees Cook 2020-06-10 08:20:05 -0700 1100) int error;
6659061045cc9 (Kees Cook 2020-06-10 08:20:05 -0700 1101)
6659061045cc9 (Kees Cook 2020-06-10 08:20:05 -0700 1102) error = security_file_receive(file);
6659061045cc9 (Kees Cook 2020-06-10 08:20:05 -0700 1103) if (error)
6659061045cc9 (Kees Cook 2020-06-10 08:20:05 -0700 1104) return error;
6659061045cc9 (Kees Cook 2020-06-10 08:20:05 -0700 1105)
42eb0d54c08a0 (Christoph Hellwig 2021-03-25 09:22:09 +0100 1106) new_fd = get_unused_fd_flags(o_flags);
42eb0d54c08a0 (Christoph Hellwig 2021-03-25 09:22:09 +0100 1107) if (new_fd < 0)
42eb0d54c08a0 (Christoph Hellwig 2021-03-25 09:22:09 +0100 1108) return new_fd;
6659061045cc9 (Kees Cook 2020-06-10 08:20:05 -0700 1109)
deefa7f3505ae (Kees Cook 2020-06-10 20:47:45 -0700 1110) if (ufd) {
deefa7f3505ae (Kees Cook 2020-06-10 20:47:45 -0700 1111) error = put_user(new_fd, ufd);
deefa7f3505ae (Kees Cook 2020-06-10 20:47:45 -0700 1112) if (error) {
42eb0d54c08a0 (Christoph Hellwig 2021-03-25 09:22:09 +0100 1113) put_unused_fd(new_fd);
deefa7f3505ae (Kees Cook 2020-06-10 20:47:45 -0700 1114) return error;
deefa7f3505ae (Kees Cook 2020-06-10 20:47:45 -0700 1115) }
6659061045cc9 (Kees Cook 2020-06-10 08:20:05 -0700 1116) }
6659061045cc9 (Kees Cook 2020-06-10 08:20:05 -0700 1117)
42eb0d54c08a0 (Christoph Hellwig 2021-03-25 09:22:09 +0100 1118) fd_install(new_fd, get_file(file));
42eb0d54c08a0 (Christoph Hellwig 2021-03-25 09:22:09 +0100 1119) __receive_sock(file);
42eb0d54c08a0 (Christoph Hellwig 2021-03-25 09:22:09 +0100 1120) return new_fd;
42eb0d54c08a0 (Christoph Hellwig 2021-03-25 09:22:09 +0100 1121) }
173817151b15d (Kees Cook 2020-06-10 08:46:58 -0700 1122)
42eb0d54c08a0 (Christoph Hellwig 2021-03-25 09:22:09 +0100 1123) int receive_fd_replace(int new_fd, struct file *file, unsigned int o_flags)
42eb0d54c08a0 (Christoph Hellwig 2021-03-25 09:22:09 +0100 1124) {
42eb0d54c08a0 (Christoph Hellwig 2021-03-25 09:22:09 +0100 1125) int error;
42eb0d54c08a0 (Christoph Hellwig 2021-03-25 09:22:09 +0100 1126)
42eb0d54c08a0 (Christoph Hellwig 2021-03-25 09:22:09 +0100 1127) error = security_file_receive(file);
42eb0d54c08a0 (Christoph Hellwig 2021-03-25 09:22:09 +0100 1128) if (error)
42eb0d54c08a0 (Christoph Hellwig 2021-03-25 09:22:09 +0100 1129) return error;
42eb0d54c08a0 (Christoph Hellwig 2021-03-25 09:22:09 +0100 1130) error = replace_fd(new_fd, file, o_flags);
42eb0d54c08a0 (Christoph Hellwig 2021-03-25 09:22:09 +0100 1131) if (error)
42eb0d54c08a0 (Christoph Hellwig 2021-03-25 09:22:09 +0100 1132) return error;
6659061045cc9 (Kees Cook 2020-06-10 08:20:05 -0700 1133) __receive_sock(file);
deefa7f3505ae (Kees Cook 2020-06-10 20:47:45 -0700 1134) return new_fd;
6659061045cc9 (Kees Cook 2020-06-10 08:20:05 -0700 1135) }
6659061045cc9 (Kees Cook 2020-06-10 08:20:05 -0700 1136)
c7248321a3d42 (Dominik Brodowski 2018-03-11 11:34:40 +0100 1137) static int ksys_dup3(unsigned int oldfd, unsigned int newfd, int flags)
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1138) {
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1139) int err = -EBADF;
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1140) struct file *file;
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1141) struct files_struct *files = current->files;
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1142)
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1143) if ((flags & ~O_CLOEXEC) != 0)
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1144) return -EINVAL;
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1145)
aed976475bff9 (Richard W.M. Jones 2012-10-09 15:27:43 +0100 1146) if (unlikely(oldfd == newfd))
aed976475bff9 (Richard W.M. Jones 2012-10-09 15:27:43 +0100 1147) return -EINVAL;
aed976475bff9 (Richard W.M. Jones 2012-10-09 15:27:43 +0100 1148)
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1149) if (newfd >= rlimit(RLIMIT_NOFILE))
08f05c49749ee (Al Viro 2012-10-31 03:37:48 +0000 1150) return -EBADF;
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1151)
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1152) spin_lock(&files->file_lock);
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1153) err = expand_files(files, newfd);
120ce2b0cd52a (Eric W. Biederman 2020-11-20 17:14:25 -0600 1154) file = files_lookup_fd_locked(files, oldfd);
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1155) if (unlikely(!file))
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1156) goto Ebadf;
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1157) if (unlikely(err < 0)) {
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1158) if (err == -EMFILE)
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1159) goto Ebadf;
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1160) goto out_unlock;
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1161) }
8280d16172243 (Al Viro 2012-08-21 12:11:46 -0400 1162) return do_dup2(files, file, newfd, flags);
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1163)
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1164) Ebadf:
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1165) err = -EBADF;
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1166) out_unlock:
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1167) spin_unlock(&files->file_lock);
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1168) return err;
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1169) }
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1170)
c7248321a3d42 (Dominik Brodowski 2018-03-11 11:34:40 +0100 1171) SYSCALL_DEFINE3(dup3, unsigned int, oldfd, unsigned int, newfd, int, flags)
c7248321a3d42 (Dominik Brodowski 2018-03-11 11:34:40 +0100 1172) {
c7248321a3d42 (Dominik Brodowski 2018-03-11 11:34:40 +0100 1173) return ksys_dup3(oldfd, newfd, flags);
c7248321a3d42 (Dominik Brodowski 2018-03-11 11:34:40 +0100 1174) }
c7248321a3d42 (Dominik Brodowski 2018-03-11 11:34:40 +0100 1175)
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1176) SYSCALL_DEFINE2(dup2, unsigned int, oldfd, unsigned int, newfd)
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1177) {
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1178) if (unlikely(newfd == oldfd)) { /* corner case */
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1179) struct files_struct *files = current->files;
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1180) int retval = oldfd;
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1181)
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1182) rcu_read_lock();
f36c294327419 (Eric W. Biederman 2020-11-20 17:14:26 -0600 1183) if (!files_lookup_fd_rcu(files, oldfd))
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1184) retval = -EBADF;
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1185) rcu_read_unlock();
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1186) return retval;
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1187) }
c7248321a3d42 (Dominik Brodowski 2018-03-11 11:34:40 +0100 1188) return ksys_dup3(oldfd, newfd, 0);
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1189) }
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1190)
bc1cd99a9ad7e (Christoph Hellwig 2020-07-14 08:58:49 +0200 1191) SYSCALL_DEFINE1(dup, unsigned int, fildes)
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1192) {
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1193) int ret = -EBADF;
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1194) struct file *file = fget_raw(fildes);
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1195)
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1196) if (file) {
8d10a035829ca (Yann Droneaud 2014-12-10 15:45:44 -0800 1197) ret = get_unused_fd_flags(0);
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1198) if (ret >= 0)
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1199) fd_install(ret, file);
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1200) else
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1201) fput(file);
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1202) }
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1203) return ret;
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1204) }
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1205)
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1206) int f_dupfd(unsigned int from, struct file *file, unsigned flags)
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1207) {
e06b53c22f31e (Eric W. Biederman 2020-11-20 17:14:36 -0600 1208) unsigned long nofile = rlimit(RLIMIT_NOFILE);
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1209) int err;
e06b53c22f31e (Eric W. Biederman 2020-11-20 17:14:36 -0600 1210) if (from >= nofile)
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1211) return -EINVAL;
e06b53c22f31e (Eric W. Biederman 2020-11-20 17:14:36 -0600 1212) err = alloc_fd(from, nofile, flags);
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1213) if (err >= 0) {
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1214) get_file(file);
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1215) fd_install(err, file);
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1216) }
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1217) return err;
fe17f22d7fd0e (Al Viro 2012-08-21 11:48:11 -0400 1218) }
c3c073f808b22 (Al Viro 2012-08-21 22:32:06 -0400 1219)
c3c073f808b22 (Al Viro 2012-08-21 22:32:06 -0400 1220) int iterate_fd(struct files_struct *files, unsigned n,
c3c073f808b22 (Al Viro 2012-08-21 22:32:06 -0400 1221) int (*f)(const void *, struct file *, unsigned),
c3c073f808b22 (Al Viro 2012-08-21 22:32:06 -0400 1222) const void *p)
c3c073f808b22 (Al Viro 2012-08-21 22:32:06 -0400 1223) {
c3c073f808b22 (Al Viro 2012-08-21 22:32:06 -0400 1224) struct fdtable *fdt;
c3c073f808b22 (Al Viro 2012-08-21 22:32:06 -0400 1225) int res = 0;
c3c073f808b22 (Al Viro 2012-08-21 22:32:06 -0400 1226) if (!files)
c3c073f808b22 (Al Viro 2012-08-21 22:32:06 -0400 1227) return 0;
c3c073f808b22 (Al Viro 2012-08-21 22:32:06 -0400 1228) spin_lock(&files->file_lock);
a77cfcb429ed9 (Al Viro 2012-11-29 22:57:33 -0500 1229) for (fdt = files_fdtable(files); n < fdt->max_fds; n++) {
a77cfcb429ed9 (Al Viro 2012-11-29 22:57:33 -0500 1230) struct file *file;
a77cfcb429ed9 (Al Viro 2012-11-29 22:57:33 -0500 1231) file = rcu_dereference_check_fdtable(files, fdt->fd[n]);
a77cfcb429ed9 (Al Viro 2012-11-29 22:57:33 -0500 1232) if (!file)
a77cfcb429ed9 (Al Viro 2012-11-29 22:57:33 -0500 1233) continue;
a77cfcb429ed9 (Al Viro 2012-11-29 22:57:33 -0500 1234) res = f(p, file, n);
a77cfcb429ed9 (Al Viro 2012-11-29 22:57:33 -0500 1235) if (res)
a77cfcb429ed9 (Al Viro 2012-11-29 22:57:33 -0500 1236) break;
c3c073f808b22 (Al Viro 2012-08-21 22:32:06 -0400 1237) }
c3c073f808b22 (Al Viro 2012-08-21 22:32:06 -0400 1238) spin_unlock(&files->file_lock);
c3c073f808b22 (Al Viro 2012-08-21 22:32:06 -0400 1239) return res;
c3c073f808b22 (Al Viro 2012-08-21 22:32:06 -0400 1240) }
c3c073f808b22 (Al Viro 2012-08-21 22:32:06 -0400 1241) EXPORT_SYMBOL(iterate_fd);