blob: 829999dafd937800afc605a477cf649b60eb3440 [file] [log] [blame]
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
19#include <linux/kernel.h>
20#include <linux/bio.h>
21#include <linux/buffer_head.h>
22#include <linux/file.h>
23#include <linux/fs.h>
Christoph Hellwigcb8e7092008-10-09 13:39:39 -040024#include <linux/fsnotify.h>
Christoph Hellwigf46b5a62008-06-11 21:53:53 -040025#include <linux/pagemap.h>
26#include <linux/highmem.h>
27#include <linux/time.h>
28#include <linux/init.h>
29#include <linux/string.h>
Christoph Hellwigf46b5a62008-06-11 21:53:53 -040030#include <linux/backing-dev.h>
Christoph Hellwigcb8e7092008-10-09 13:39:39 -040031#include <linux/mount.h>
Christoph Hellwigf46b5a62008-06-11 21:53:53 -040032#include <linux/mpage.h>
Christoph Hellwigcb8e7092008-10-09 13:39:39 -040033#include <linux/namei.h>
Christoph Hellwigf46b5a62008-06-11 21:53:53 -040034#include <linux/swap.h>
35#include <linux/writeback.h>
36#include <linux/statfs.h>
37#include <linux/compat.h>
38#include <linux/bit_spinlock.h>
Christoph Hellwigcb8e7092008-10-09 13:39:39 -040039#include <linux/security.h>
Christoph Hellwigf46b5a62008-06-11 21:53:53 -040040#include <linux/xattr.h>
Yan Zheng7ea394f2008-08-05 13:05:02 -040041#include <linux/vmalloc.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090042#include <linux/slab.h>
Li Dongyangf7039b12011-03-24 10:24:28 +000043#include <linux/blkdev.h>
Alexander Block8ea05e32012-07-25 17:35:53 +020044#include <linux/uuid.h>
Filipe Brandenburger55e301f2013-01-29 06:04:50 +000045#include <linux/btrfs.h>
Mark Fasheh416161d2013-08-06 11:42:51 -070046#include <linux/uaccess.h>
Christoph Hellwigf46b5a62008-06-11 21:53:53 -040047#include "ctree.h"
48#include "disk-io.h"
49#include "transaction.h"
50#include "btrfs_inode.h"
Christoph Hellwigf46b5a62008-06-11 21:53:53 -040051#include "print-tree.h"
52#include "volumes.h"
Chris Mason925baed2008-06-25 16:01:30 -040053#include "locking.h"
Li Zefan581bb052011-04-20 10:06:11 +080054#include "inode-map.h"
Jan Schmidtd7728c92011-07-07 16:48:38 +020055#include "backref.h"
Josef Bacik606686e2012-06-04 14:03:51 -040056#include "rcu-string.h"
Alexander Block31db9f72012-07-25 23:19:24 +020057#include "send.h"
Stefan Behrens3f6bcfb2012-11-06 15:08:53 +010058#include "dev-replace.h"
Filipe David Borba Manana63541922014-01-07 11:47:46 +000059#include "props.h"
Jeff Mahoney3b02a682013-11-01 13:07:02 -040060#include "sysfs.h"
Christoph Hellwigf46b5a62008-06-11 21:53:53 -040061
Mark Fasheh416161d2013-08-06 11:42:51 -070062static int btrfs_clone(struct inode *src, struct inode *inode,
63 u64 off, u64 olen, u64 olen_aligned, u64 destoff);
64
Christoph Hellwig6cbff002009-04-17 10:37:41 +020065/* Mask out flags that are inappropriate for the given type of inode. */
66static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
67{
68 if (S_ISDIR(mode))
69 return flags;
70 else if (S_ISREG(mode))
71 return flags & ~FS_DIRSYNC_FL;
72 else
73 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
74}
Christoph Hellwigf46b5a62008-06-11 21:53:53 -040075
Christoph Hellwig6cbff002009-04-17 10:37:41 +020076/*
77 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
78 */
79static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
80{
81 unsigned int iflags = 0;
82
83 if (flags & BTRFS_INODE_SYNC)
84 iflags |= FS_SYNC_FL;
85 if (flags & BTRFS_INODE_IMMUTABLE)
86 iflags |= FS_IMMUTABLE_FL;
87 if (flags & BTRFS_INODE_APPEND)
88 iflags |= FS_APPEND_FL;
89 if (flags & BTRFS_INODE_NODUMP)
90 iflags |= FS_NODUMP_FL;
91 if (flags & BTRFS_INODE_NOATIME)
92 iflags |= FS_NOATIME_FL;
93 if (flags & BTRFS_INODE_DIRSYNC)
94 iflags |= FS_DIRSYNC_FL;
Li Zefand0092bd2011-04-15 03:03:06 +000095 if (flags & BTRFS_INODE_NODATACOW)
96 iflags |= FS_NOCOW_FL;
97
98 if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
99 iflags |= FS_COMPR_FL;
100 else if (flags & BTRFS_INODE_NOCOMPRESS)
101 iflags |= FS_NOCOMP_FL;
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200102
103 return iflags;
104}
105
106/*
107 * Update inode->i_flags based on the btrfs internal flags.
108 */
109void btrfs_update_iflags(struct inode *inode)
110{
111 struct btrfs_inode *ip = BTRFS_I(inode);
112
113 inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
114
115 if (ip->flags & BTRFS_INODE_SYNC)
116 inode->i_flags |= S_SYNC;
117 if (ip->flags & BTRFS_INODE_IMMUTABLE)
118 inode->i_flags |= S_IMMUTABLE;
119 if (ip->flags & BTRFS_INODE_APPEND)
120 inode->i_flags |= S_APPEND;
121 if (ip->flags & BTRFS_INODE_NOATIME)
122 inode->i_flags |= S_NOATIME;
123 if (ip->flags & BTRFS_INODE_DIRSYNC)
124 inode->i_flags |= S_DIRSYNC;
125}
126
127/*
128 * Inherit flags from the parent inode.
129 *
Josef Bacike27425d2011-09-27 11:01:30 -0400130 * Currently only the compression flags and the cow flags are inherited.
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200131 */
132void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
133{
Chris Mason0b4dcea2009-06-11 11:13:35 -0400134 unsigned int flags;
135
136 if (!dir)
137 return;
138
139 flags = BTRFS_I(dir)->flags;
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200140
Josef Bacike27425d2011-09-27 11:01:30 -0400141 if (flags & BTRFS_INODE_NOCOMPRESS) {
142 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
143 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
144 } else if (flags & BTRFS_INODE_COMPRESS) {
145 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
146 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
147 }
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200148
Liu Bo213490b2012-09-11 08:33:50 -0600149 if (flags & BTRFS_INODE_NODATACOW) {
Josef Bacike27425d2011-09-27 11:01:30 -0400150 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
Liu Bo213490b2012-09-11 08:33:50 -0600151 if (S_ISREG(inode->i_mode))
152 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
153 }
Josef Bacike27425d2011-09-27 11:01:30 -0400154
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200155 btrfs_update_iflags(inode);
156}
157
158static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
159{
Al Viro496ad9a2013-01-23 17:07:38 -0500160 struct btrfs_inode *ip = BTRFS_I(file_inode(file));
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200161 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
162
163 if (copy_to_user(arg, &flags, sizeof(flags)))
164 return -EFAULT;
165 return 0;
166}
167
Liu Bo75e7cb72011-03-22 10:12:20 +0000168static int check_flags(unsigned int flags)
169{
170 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
171 FS_NOATIME_FL | FS_NODUMP_FL | \
172 FS_SYNC_FL | FS_DIRSYNC_FL | \
Li Zefane1e8fb62011-04-15 03:02:49 +0000173 FS_NOCOMP_FL | FS_COMPR_FL |
174 FS_NOCOW_FL))
Liu Bo75e7cb72011-03-22 10:12:20 +0000175 return -EOPNOTSUPP;
176
177 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
178 return -EINVAL;
179
Liu Bo75e7cb72011-03-22 10:12:20 +0000180 return 0;
181}
182
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200183static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
184{
Al Viro496ad9a2013-01-23 17:07:38 -0500185 struct inode *inode = file_inode(file);
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200186 struct btrfs_inode *ip = BTRFS_I(inode);
187 struct btrfs_root *root = ip->root;
188 struct btrfs_trans_handle *trans;
189 unsigned int flags, oldflags;
190 int ret;
Li Zefanf062abf2011-12-29 13:36:45 +0800191 u64 ip_oldflags;
192 unsigned int i_oldflags;
David Sterba7e97b8d2012-09-07 05:56:55 -0600193 umode_t mode;
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200194
Li Zefanb83cc962010-12-20 16:04:08 +0800195 if (btrfs_root_readonly(root))
196 return -EROFS;
197
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200198 if (copy_from_user(&flags, arg, sizeof(flags)))
199 return -EFAULT;
200
Liu Bo75e7cb72011-03-22 10:12:20 +0000201 ret = check_flags(flags);
202 if (ret)
203 return ret;
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200204
Serge E. Hallyn2e149672011-03-23 16:43:26 -0700205 if (!inode_owner_or_capable(inode))
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200206 return -EACCES;
207
Jan Karae7848682012-06-12 16:20:32 +0200208 ret = mnt_want_write_file(file);
209 if (ret)
210 return ret;
211
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200212 mutex_lock(&inode->i_mutex);
213
Li Zefanf062abf2011-12-29 13:36:45 +0800214 ip_oldflags = ip->flags;
215 i_oldflags = inode->i_flags;
David Sterba7e97b8d2012-09-07 05:56:55 -0600216 mode = inode->i_mode;
Li Zefanf062abf2011-12-29 13:36:45 +0800217
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200218 flags = btrfs_mask_flags(inode->i_mode, flags);
219 oldflags = btrfs_flags_to_ioctl(ip->flags);
220 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
221 if (!capable(CAP_LINUX_IMMUTABLE)) {
222 ret = -EPERM;
223 goto out_unlock;
224 }
225 }
226
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200227 if (flags & FS_SYNC_FL)
228 ip->flags |= BTRFS_INODE_SYNC;
229 else
230 ip->flags &= ~BTRFS_INODE_SYNC;
231 if (flags & FS_IMMUTABLE_FL)
232 ip->flags |= BTRFS_INODE_IMMUTABLE;
233 else
234 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
235 if (flags & FS_APPEND_FL)
236 ip->flags |= BTRFS_INODE_APPEND;
237 else
238 ip->flags &= ~BTRFS_INODE_APPEND;
239 if (flags & FS_NODUMP_FL)
240 ip->flags |= BTRFS_INODE_NODUMP;
241 else
242 ip->flags &= ~BTRFS_INODE_NODUMP;
243 if (flags & FS_NOATIME_FL)
244 ip->flags |= BTRFS_INODE_NOATIME;
245 else
246 ip->flags &= ~BTRFS_INODE_NOATIME;
247 if (flags & FS_DIRSYNC_FL)
248 ip->flags |= BTRFS_INODE_DIRSYNC;
249 else
250 ip->flags &= ~BTRFS_INODE_DIRSYNC;
David Sterba7e97b8d2012-09-07 05:56:55 -0600251 if (flags & FS_NOCOW_FL) {
252 if (S_ISREG(mode)) {
253 /*
254 * It's safe to turn csums off here, no extents exist.
255 * Otherwise we want the flag to reflect the real COW
256 * status of the file and will not set it.
257 */
258 if (inode->i_size == 0)
259 ip->flags |= BTRFS_INODE_NODATACOW
260 | BTRFS_INODE_NODATASUM;
261 } else {
262 ip->flags |= BTRFS_INODE_NODATACOW;
263 }
264 } else {
265 /*
266 * Revert back under same assuptions as above
267 */
268 if (S_ISREG(mode)) {
269 if (inode->i_size == 0)
270 ip->flags &= ~(BTRFS_INODE_NODATACOW
271 | BTRFS_INODE_NODATASUM);
272 } else {
273 ip->flags &= ~BTRFS_INODE_NODATACOW;
274 }
275 }
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200276
Liu Bo75e7cb72011-03-22 10:12:20 +0000277 /*
278 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
279 * flag may be changed automatically if compression code won't make
280 * things smaller.
281 */
282 if (flags & FS_NOCOMP_FL) {
283 ip->flags &= ~BTRFS_INODE_COMPRESS;
284 ip->flags |= BTRFS_INODE_NOCOMPRESS;
Filipe David Borba Manana63541922014-01-07 11:47:46 +0000285
286 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
287 if (ret && ret != -ENODATA)
288 goto out_drop;
Liu Bo75e7cb72011-03-22 10:12:20 +0000289 } else if (flags & FS_COMPR_FL) {
Filipe David Borba Manana63541922014-01-07 11:47:46 +0000290 const char *comp;
291
Liu Bo75e7cb72011-03-22 10:12:20 +0000292 ip->flags |= BTRFS_INODE_COMPRESS;
293 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
Filipe David Borba Manana63541922014-01-07 11:47:46 +0000294
295 if (root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
296 comp = "lzo";
297 else
298 comp = "zlib";
299 ret = btrfs_set_prop(inode, "btrfs.compression",
300 comp, strlen(comp), 0);
301 if (ret)
302 goto out_drop;
303
Li Zefanebcb9042011-04-15 03:03:17 +0000304 } else {
305 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
Liu Bo75e7cb72011-03-22 10:12:20 +0000306 }
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200307
Li Zefan4da6f1a2011-12-29 13:39:50 +0800308 trans = btrfs_start_transaction(root, 1);
Li Zefanf062abf2011-12-29 13:36:45 +0800309 if (IS_ERR(trans)) {
310 ret = PTR_ERR(trans);
311 goto out_drop;
312 }
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200313
Li Zefan306424cc2011-12-14 20:12:02 -0500314 btrfs_update_iflags(inode);
Josef Bacik0c4d2d92012-04-05 15:03:02 -0400315 inode_inc_iversion(inode);
Li Zefan306424cc2011-12-14 20:12:02 -0500316 inode->i_ctime = CURRENT_TIME;
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200317 ret = btrfs_update_inode(trans, root, inode);
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200318
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200319 btrfs_end_transaction(trans, root);
Li Zefanf062abf2011-12-29 13:36:45 +0800320 out_drop:
321 if (ret) {
322 ip->flags = ip_oldflags;
323 inode->i_flags = i_oldflags;
324 }
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200325
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200326 out_unlock:
327 mutex_unlock(&inode->i_mutex);
Jan Karae7848682012-06-12 16:20:32 +0200328 mnt_drop_write_file(file);
liubo2d4e6f62011-02-24 09:38:16 +0000329 return ret;
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200330}
331
332static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
333{
Al Viro496ad9a2013-01-23 17:07:38 -0500334 struct inode *inode = file_inode(file);
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200335
336 return put_user(inode->i_generation, arg);
337}
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400338
Li Dongyangf7039b12011-03-24 10:24:28 +0000339static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
340{
Al Viro54563d42013-09-01 15:57:51 -0400341 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
Li Dongyangf7039b12011-03-24 10:24:28 +0000342 struct btrfs_device *device;
343 struct request_queue *q;
344 struct fstrim_range range;
345 u64 minlen = ULLONG_MAX;
346 u64 num_devices = 0;
Al Viro815745c2011-11-17 15:40:49 -0500347 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
Li Dongyangf7039b12011-03-24 10:24:28 +0000348 int ret;
349
350 if (!capable(CAP_SYS_ADMIN))
351 return -EPERM;
352
Xiao Guangrong1f781602011-04-20 10:09:16 +0000353 rcu_read_lock();
354 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
355 dev_list) {
Li Dongyangf7039b12011-03-24 10:24:28 +0000356 if (!device->bdev)
357 continue;
358 q = bdev_get_queue(device->bdev);
359 if (blk_queue_discard(q)) {
360 num_devices++;
361 minlen = min((u64)q->limits.discard_granularity,
362 minlen);
363 }
364 }
Xiao Guangrong1f781602011-04-20 10:09:16 +0000365 rcu_read_unlock();
Lukas Czernerf4c697e2011-09-05 16:34:54 +0200366
Li Dongyangf7039b12011-03-24 10:24:28 +0000367 if (!num_devices)
368 return -EOPNOTSUPP;
Li Dongyangf7039b12011-03-24 10:24:28 +0000369 if (copy_from_user(&range, arg, sizeof(range)))
370 return -EFAULT;
Lukas Czernere515c182012-10-16 09:34:36 +0000371 if (range.start > total_bytes ||
372 range.len < fs_info->sb->s_blocksize)
Lukas Czernerf4c697e2011-09-05 16:34:54 +0200373 return -EINVAL;
Li Dongyangf7039b12011-03-24 10:24:28 +0000374
Lukas Czernerf4c697e2011-09-05 16:34:54 +0200375 range.len = min(range.len, total_bytes - range.start);
Li Dongyangf7039b12011-03-24 10:24:28 +0000376 range.minlen = max(range.minlen, minlen);
Al Viro815745c2011-11-17 15:40:49 -0500377 ret = btrfs_trim_fs(fs_info->tree_root, &range);
Li Dongyangf7039b12011-03-24 10:24:28 +0000378 if (ret < 0)
379 return ret;
380
381 if (copy_to_user(arg, &range, sizeof(range)))
382 return -EFAULT;
383
384 return 0;
385}
386
Stefan Behrensdd5f9612013-08-15 17:11:20 +0200387int btrfs_is_empty_uuid(u8 *uuid)
388{
Chris Mason46e0f662013-11-15 12:14:55 +0100389 int i;
390
391 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
392 if (uuid[i])
393 return 0;
394 }
395 return 1;
Stefan Behrensdd5f9612013-08-15 17:11:20 +0200396}
397
Miao Xied5c12072013-02-28 10:04:33 +0000398static noinline int create_subvol(struct inode *dir,
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400399 struct dentry *dentry,
Sage Weil72fd0322010-10-29 15:41:32 -0400400 char *name, int namelen,
Arne Jansen6f72c7e2011-09-14 15:58:21 +0200401 u64 *async_transid,
Miao Xie8696c532013-02-07 06:02:44 +0000402 struct btrfs_qgroup_inherit *inherit)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400403{
404 struct btrfs_trans_handle *trans;
405 struct btrfs_key key;
406 struct btrfs_root_item root_item;
407 struct btrfs_inode_item *inode_item;
408 struct extent_buffer *leaf;
Miao Xied5c12072013-02-28 10:04:33 +0000409 struct btrfs_root *root = BTRFS_I(dir)->root;
Yan, Zheng76dda932009-09-21 16:00:26 -0400410 struct btrfs_root *new_root;
Miao Xied5c12072013-02-28 10:04:33 +0000411 struct btrfs_block_rsv block_rsv;
Alexander Block8ea05e32012-07-25 17:35:53 +0200412 struct timespec cur_time = CURRENT_TIME;
Tsutomu Itoh5662344b32013-12-13 09:51:42 +0900413 struct inode *inode;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400414 int ret;
415 int err;
416 u64 objectid;
417 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
Chris Mason3de45862008-11-17 21:02:50 -0500418 u64 index = 0;
Miao Xied5c12072013-02-28 10:04:33 +0000419 u64 qgroup_reserved;
Alexander Block8ea05e32012-07-25 17:35:53 +0200420 uuid_le new_uuid;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400421
Li Zefan581bb052011-04-20 10:06:11 +0800422 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
Al Viro2fbe8c82011-07-16 21:38:06 -0400423 if (ret)
Yan, Zhenga22285a2010-05-16 10:48:46 -0400424 return ret;
Josef Bacik6a912212010-11-20 09:48:00 +0000425
Miao Xied5c12072013-02-28 10:04:33 +0000426 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
Josef Bacik9ed74f22009-09-11 16:12:44 -0400427 /*
Miao Xied5c12072013-02-28 10:04:33 +0000428 * The same as the snapshot creation, please see the comment
429 * of create_snapshot().
Josef Bacik9ed74f22009-09-11 16:12:44 -0400430 */
Miao Xied5c12072013-02-28 10:04:33 +0000431 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
Stefan Behrensdd5f9612013-08-15 17:11:20 +0200432 8, &qgroup_reserved, false);
Miao Xied5c12072013-02-28 10:04:33 +0000433 if (ret)
434 return ret;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400435
Miao Xied5c12072013-02-28 10:04:33 +0000436 trans = btrfs_start_transaction(root, 0);
437 if (IS_ERR(trans)) {
438 ret = PTR_ERR(trans);
Liu Bode6e8202014-01-09 14:57:06 +0800439 btrfs_subvolume_release_metadata(root, &block_rsv,
440 qgroup_reserved);
441 return ret;
Miao Xied5c12072013-02-28 10:04:33 +0000442 }
443 trans->block_rsv = &block_rsv;
444 trans->bytes_reserved = block_rsv.size;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400445
Miao Xie8696c532013-02-07 06:02:44 +0000446 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
Arne Jansen6f72c7e2011-09-14 15:58:21 +0200447 if (ret)
448 goto fail;
449
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400450 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
Jan Schmidt5581a512012-05-16 17:04:52 +0200451 0, objectid, NULL, 0, 0, 0);
Josef Bacik8e8a1e32008-07-24 12:17:14 -0400452 if (IS_ERR(leaf)) {
453 ret = PTR_ERR(leaf);
454 goto fail;
455 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400456
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400457 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400458 btrfs_set_header_bytenr(leaf, leaf->start);
459 btrfs_set_header_generation(leaf, trans->transid);
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400460 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400461 btrfs_set_header_owner(leaf, objectid);
462
Ross Kirk0a4e5582013-09-24 10:12:38 +0100463 write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(),
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400464 BTRFS_FSID_SIZE);
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400465 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
Geert Uytterhoevenb308bc22013-08-20 13:20:15 +0200466 btrfs_header_chunk_tree_uuid(leaf),
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400467 BTRFS_UUID_SIZE);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400468 btrfs_mark_buffer_dirty(leaf);
469
Alexander Block8ea05e32012-07-25 17:35:53 +0200470 memset(&root_item, 0, sizeof(root_item));
471
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400472 inode_item = &root_item.inode;
Qu Wenruo3cae2102013-07-16 11:19:18 +0800473 btrfs_set_stack_inode_generation(inode_item, 1);
474 btrfs_set_stack_inode_size(inode_item, 3);
475 btrfs_set_stack_inode_nlink(inode_item, 1);
476 btrfs_set_stack_inode_nbytes(inode_item, root->leafsize);
477 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400478
Qu Wenruo3cae2102013-07-16 11:19:18 +0800479 btrfs_set_root_flags(&root_item, 0);
480 btrfs_set_root_limit(&root_item, 0);
481 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
Li Zefan08fe4db2011-03-28 02:01:25 +0000482
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400483 btrfs_set_root_bytenr(&root_item, leaf->start);
Yan Zheng84234f32008-10-29 14:49:05 -0400484 btrfs_set_root_generation(&root_item, trans->transid);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400485 btrfs_set_root_level(&root_item, 0);
486 btrfs_set_root_refs(&root_item, 1);
Yan, Zheng86b9f2e2009-11-12 09:36:50 +0000487 btrfs_set_root_used(&root_item, leaf->len);
Yan Zheng80ff3852008-10-30 14:20:02 -0400488 btrfs_set_root_last_snapshot(&root_item, 0);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400489
Alexander Block8ea05e32012-07-25 17:35:53 +0200490 btrfs_set_root_generation_v2(&root_item,
491 btrfs_root_generation(&root_item));
492 uuid_le_gen(&new_uuid);
493 memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
Qu Wenruo3cae2102013-07-16 11:19:18 +0800494 btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec);
495 btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec);
Alexander Block8ea05e32012-07-25 17:35:53 +0200496 root_item.ctime = root_item.otime;
497 btrfs_set_root_ctransid(&root_item, trans->transid);
498 btrfs_set_root_otransid(&root_item, trans->transid);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400499
Chris Mason925baed2008-06-25 16:01:30 -0400500 btrfs_tree_unlock(leaf);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400501 free_extent_buffer(leaf);
502 leaf = NULL;
503
504 btrfs_set_root_dirid(&root_item, new_dirid);
505
506 key.objectid = objectid;
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400507 key.offset = 0;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400508 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
509 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
510 &root_item);
511 if (ret)
512 goto fail;
513
Yan, Zheng76dda932009-09-21 16:00:26 -0400514 key.offset = (u64)-1;
515 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100516 if (IS_ERR(new_root)) {
517 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
518 ret = PTR_ERR(new_root);
519 goto fail;
520 }
Yan, Zheng76dda932009-09-21 16:00:26 -0400521
522 btrfs_record_root_in_trans(trans, new_root);
523
Filipe David Borba Manana63541922014-01-07 11:47:46 +0000524 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
Mark Fashehce598972011-07-26 11:32:23 -0700525 if (ret) {
526 /* We potentially lose an unused inode item here */
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100527 btrfs_abort_transaction(trans, root, ret);
Mark Fashehce598972011-07-26 11:32:23 -0700528 goto fail;
529 }
530
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400531 /*
532 * insert the directory item
533 */
Chris Mason3de45862008-11-17 21:02:50 -0500534 ret = btrfs_set_inode_index(dir, &index);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100535 if (ret) {
536 btrfs_abort_transaction(trans, root, ret);
537 goto fail;
538 }
Chris Mason3de45862008-11-17 21:02:50 -0500539
540 ret = btrfs_insert_dir_item(trans, root,
Miao Xie16cdcec2011-04-22 18:12:22 +0800541 name, namelen, dir, &key,
Chris Mason3de45862008-11-17 21:02:50 -0500542 BTRFS_FT_DIR, index);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100543 if (ret) {
544 btrfs_abort_transaction(trans, root, ret);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400545 goto fail;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100546 }
Chris Mason0660b5a2008-11-17 20:37:39 -0500547
Yan Zheng52c26172009-01-05 15:43:43 -0500548 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
549 ret = btrfs_update_inode(trans, root, dir);
550 BUG_ON(ret);
551
Chris Mason0660b5a2008-11-17 20:37:39 -0500552 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
Yan, Zheng4df27c4d2009-09-21 15:56:00 -0400553 objectid, root->root_key.objectid,
Li Zefan33345d012011-04-20 10:31:50 +0800554 btrfs_ino(dir), index, name, namelen);
Chris Mason0660b5a2008-11-17 20:37:39 -0500555 BUG_ON(ret);
556
Stefan Behrensdd5f9612013-08-15 17:11:20 +0200557 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
558 root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
559 objectid);
560 if (ret)
561 btrfs_abort_transaction(trans, root, ret);
562
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400563fail:
Miao Xied5c12072013-02-28 10:04:33 +0000564 trans->block_rsv = NULL;
565 trans->bytes_reserved = 0;
Liu Bode6e8202014-01-09 14:57:06 +0800566 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
567
Sage Weil72fd0322010-10-29 15:41:32 -0400568 if (async_transid) {
569 *async_transid = trans->transid;
570 err = btrfs_commit_transaction_async(trans, root, 1);
Miao Xie00d71c92013-03-04 09:45:06 +0000571 if (err)
572 err = btrfs_commit_transaction(trans, root);
Sage Weil72fd0322010-10-29 15:41:32 -0400573 } else {
574 err = btrfs_commit_transaction(trans, root);
575 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400576 if (err && !ret)
577 ret = err;
Chris Mason1a65e242013-02-06 12:06:02 -0500578
Tsutomu Itoh5662344b32013-12-13 09:51:42 +0900579 if (!ret) {
580 inode = btrfs_lookup_dentry(dir, dentry);
Liu Bode6e8202014-01-09 14:57:06 +0800581 if (IS_ERR(inode))
582 return PTR_ERR(inode);
Tsutomu Itoh5662344b32013-12-13 09:51:42 +0900583 d_instantiate(dentry, inode);
584 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400585 return ret;
586}
587
Miao Xiee9662f72013-02-28 10:01:15 +0000588static int create_snapshot(struct btrfs_root *root, struct inode *dir,
589 struct dentry *dentry, char *name, int namelen,
590 u64 *async_transid, bool readonly,
591 struct btrfs_qgroup_inherit *inherit)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400592{
Yan, Zheng2e4bfab2009-11-12 09:37:02 +0000593 struct inode *inode;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400594 struct btrfs_pending_snapshot *pending_snapshot;
595 struct btrfs_trans_handle *trans;
Yan, Zheng2e4bfab2009-11-12 09:37:02 +0000596 int ret;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400597
598 if (!root->ref_cows)
599 return -EINVAL;
600
Miao Xie6a038432013-05-15 07:48:24 +0000601 ret = btrfs_start_delalloc_inodes(root, 0);
602 if (ret)
603 return ret;
604
Miao Xieb0244192013-11-04 23:13:25 +0800605 btrfs_wait_ordered_extents(root, -1);
Miao Xie6a038432013-05-15 07:48:24 +0000606
Yan, Zhenga22285a2010-05-16 10:48:46 -0400607 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
608 if (!pending_snapshot)
609 return -ENOMEM;
610
Miao Xie66d8f3d2012-09-06 04:02:28 -0600611 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
612 BTRFS_BLOCK_RSV_TEMP);
Miao Xied5c12072013-02-28 10:04:33 +0000613 /*
614 * 1 - parent dir inode
615 * 2 - dir entries
616 * 1 - root item
617 * 2 - root ref/backref
618 * 1 - root of snapshot
Stefan Behrensdd5f9612013-08-15 17:11:20 +0200619 * 1 - UUID item
Miao Xied5c12072013-02-28 10:04:33 +0000620 */
621 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
Stefan Behrensdd5f9612013-08-15 17:11:20 +0200622 &pending_snapshot->block_rsv, 8,
Jeff Mahoneyee3441b2013-07-09 16:37:21 -0400623 &pending_snapshot->qgroup_reserved,
624 false);
Miao Xied5c12072013-02-28 10:04:33 +0000625 if (ret)
626 goto out;
627
Yan, Zhenga22285a2010-05-16 10:48:46 -0400628 pending_snapshot->dentry = dentry;
629 pending_snapshot->root = root;
Li Zefanb83cc962010-12-20 16:04:08 +0800630 pending_snapshot->readonly = readonly;
Miao Xiee9662f72013-02-28 10:01:15 +0000631 pending_snapshot->dir = dir;
Miao Xie8696c532013-02-07 06:02:44 +0000632 pending_snapshot->inherit = inherit;
Yan, Zhenga22285a2010-05-16 10:48:46 -0400633
Miao Xied5c12072013-02-28 10:04:33 +0000634 trans = btrfs_start_transaction(root, 0);
Yan, Zhenga22285a2010-05-16 10:48:46 -0400635 if (IS_ERR(trans)) {
636 ret = PTR_ERR(trans);
637 goto fail;
638 }
639
Josef Bacik83515832011-06-14 15:16:14 -0400640 spin_lock(&root->fs_info->trans_lock);
Yan, Zhenga22285a2010-05-16 10:48:46 -0400641 list_add(&pending_snapshot->list,
642 &trans->transaction->pending_snapshots);
Josef Bacik83515832011-06-14 15:16:14 -0400643 spin_unlock(&root->fs_info->trans_lock);
Sage Weil72fd0322010-10-29 15:41:32 -0400644 if (async_transid) {
645 *async_transid = trans->transid;
646 ret = btrfs_commit_transaction_async(trans,
647 root->fs_info->extent_root, 1);
Miao Xie00d71c92013-03-04 09:45:06 +0000648 if (ret)
649 ret = btrfs_commit_transaction(trans, root);
Sage Weil72fd0322010-10-29 15:41:32 -0400650 } else {
651 ret = btrfs_commit_transaction(trans,
652 root->fs_info->extent_root);
653 }
Miao Xieaec80302013-03-04 09:44:29 +0000654 if (ret)
Josef Bacikc37b2b62012-10-22 15:51:44 -0400655 goto fail;
Yan, Zhenga22285a2010-05-16 10:48:46 -0400656
657 ret = pending_snapshot->error;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400658 if (ret)
Yan, Zheng2e4bfab2009-11-12 09:37:02 +0000659 goto fail;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400660
Josef Bacik66b4ffd2011-01-31 16:22:42 -0500661 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
662 if (ret)
663 goto fail;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400664
Al Viro2fbe8c82011-07-16 21:38:06 -0400665 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
Yan, Zheng2e4bfab2009-11-12 09:37:02 +0000666 if (IS_ERR(inode)) {
667 ret = PTR_ERR(inode);
668 goto fail;
669 }
Tsutomu Itoh5662344b32013-12-13 09:51:42 +0900670
Yan, Zheng2e4bfab2009-11-12 09:37:02 +0000671 d_instantiate(dentry, inode);
672 ret = 0;
673fail:
Miao Xied5c12072013-02-28 10:04:33 +0000674 btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
675 &pending_snapshot->block_rsv,
676 pending_snapshot->qgroup_reserved);
677out:
Yan, Zhenga22285a2010-05-16 10:48:46 -0400678 kfree(pending_snapshot);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400679 return ret;
680}
681
Sage Weil4260f7c2010-10-29 15:46:43 -0400682/* copy of check_sticky in fs/namei.c()
683* It's inline, so penalty for filesystems that don't use sticky bit is
684* minimal.
685*/
686static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
687{
Eric W. Biederman2f2f43d2012-02-10 11:05:07 -0800688 kuid_t fsuid = current_fsuid();
Sage Weil4260f7c2010-10-29 15:46:43 -0400689
690 if (!(dir->i_mode & S_ISVTX))
691 return 0;
Eric W. Biederman2f2f43d2012-02-10 11:05:07 -0800692 if (uid_eq(inode->i_uid, fsuid))
Sage Weil4260f7c2010-10-29 15:46:43 -0400693 return 0;
Eric W. Biederman2f2f43d2012-02-10 11:05:07 -0800694 if (uid_eq(dir->i_uid, fsuid))
Sage Weil4260f7c2010-10-29 15:46:43 -0400695 return 0;
696 return !capable(CAP_FOWNER);
697}
698
699/* copy of may_delete in fs/namei.c()
700 * Check whether we can remove a link victim from directory dir, check
701 * whether the type of victim is right.
702 * 1. We can't do it if dir is read-only (done in permission())
703 * 2. We should have write and exec permissions on dir
704 * 3. We can't remove anything from append-only dir
705 * 4. We can't do anything with immutable dir (done in permission())
706 * 5. If the sticky bit on dir is set we should either
707 * a. be owner of dir, or
708 * b. be owner of victim, or
709 * c. have CAP_FOWNER capability
710 * 6. If the victim is append-only or immutable we can't do antyhing with
711 * links pointing to it.
712 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
713 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
714 * 9. We can't remove a root or mountpoint.
715 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
716 * nfs_async_unlink().
717 */
718
Dulshani Gunawardhana67871252013-10-31 10:33:04 +0530719static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
Sage Weil4260f7c2010-10-29 15:46:43 -0400720{
721 int error;
722
723 if (!victim->d_inode)
724 return -ENOENT;
725
726 BUG_ON(victim->d_parent->d_inode != dir);
Jeff Layton4fa6b5e2012-10-10 15:25:25 -0400727 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
Sage Weil4260f7c2010-10-29 15:46:43 -0400728
729 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
730 if (error)
731 return error;
732 if (IS_APPEND(dir))
733 return -EPERM;
734 if (btrfs_check_sticky(dir, victim->d_inode)||
735 IS_APPEND(victim->d_inode)||
736 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
737 return -EPERM;
738 if (isdir) {
739 if (!S_ISDIR(victim->d_inode->i_mode))
740 return -ENOTDIR;
741 if (IS_ROOT(victim))
742 return -EBUSY;
743 } else if (S_ISDIR(victim->d_inode->i_mode))
744 return -EISDIR;
745 if (IS_DEADDIR(dir))
746 return -ENOENT;
747 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
748 return -EBUSY;
749 return 0;
750}
751
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400752/* copy of may_create in fs/namei.c() */
753static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
754{
755 if (child->d_inode)
756 return -EEXIST;
757 if (IS_DEADDIR(dir))
758 return -ENOENT;
759 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
760}
761
762/*
763 * Create a new subvolume below @parent. This is largely modeled after
764 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
765 * inside this filesystem so it's quite a bit simpler.
766 */
Yan, Zheng76dda932009-09-21 16:00:26 -0400767static noinline int btrfs_mksubvol(struct path *parent,
768 char *name, int namelen,
Sage Weil72fd0322010-10-29 15:41:32 -0400769 struct btrfs_root *snap_src,
Arne Jansen6f72c7e2011-09-14 15:58:21 +0200770 u64 *async_transid, bool readonly,
Miao Xie8696c532013-02-07 06:02:44 +0000771 struct btrfs_qgroup_inherit *inherit)
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400772{
Yan, Zheng76dda932009-09-21 16:00:26 -0400773 struct inode *dir = parent->dentry->d_inode;
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400774 struct dentry *dentry;
775 int error;
776
David Sterba5c50c9b2013-03-22 18:12:51 +0000777 error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
778 if (error == -EINTR)
779 return error;
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400780
781 dentry = lookup_one_len(name, parent->dentry, namelen);
782 error = PTR_ERR(dentry);
783 if (IS_ERR(dentry))
784 goto out_unlock;
785
786 error = -EEXIST;
787 if (dentry->d_inode)
788 goto out_dput;
789
Yan, Zheng76dda932009-09-21 16:00:26 -0400790 error = btrfs_may_create(dir, dentry);
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400791 if (error)
Liu Boa874a632012-06-29 03:58:46 -0600792 goto out_dput;
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400793
Chris Mason9c520572012-12-17 14:26:57 -0500794 /*
795 * even if this name doesn't exist, we may get hash collisions.
796 * check for them now when we can safely fail
797 */
798 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
799 dir->i_ino, name,
800 namelen);
801 if (error)
802 goto out_dput;
803
Yan, Zheng76dda932009-09-21 16:00:26 -0400804 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
805
806 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
807 goto out_up_read;
808
Chris Mason3de45862008-11-17 21:02:50 -0500809 if (snap_src) {
Miao Xiee9662f72013-02-28 10:01:15 +0000810 error = create_snapshot(snap_src, dir, dentry, name, namelen,
Arne Jansen6f72c7e2011-09-14 15:58:21 +0200811 async_transid, readonly, inherit);
Chris Mason3de45862008-11-17 21:02:50 -0500812 } else {
Miao Xied5c12072013-02-28 10:04:33 +0000813 error = create_subvol(dir, dentry, name, namelen,
814 async_transid, inherit);
Chris Mason3de45862008-11-17 21:02:50 -0500815 }
Yan, Zheng76dda932009-09-21 16:00:26 -0400816 if (!error)
817 fsnotify_mkdir(dir, dentry);
818out_up_read:
819 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400820out_dput:
821 dput(dentry);
822out_unlock:
Yan, Zheng76dda932009-09-21 16:00:26 -0400823 mutex_unlock(&dir->i_mutex);
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400824 return error;
825}
826
Chris Mason4cb53002011-05-24 15:35:30 -0400827/*
828 * When we're defragging a range, we don't want to kick it off again
829 * if it is really just waiting for delalloc to send it down.
830 * If we find a nice big extent or delalloc range for the bytes in the
831 * file you want to defrag, we return 0 to let you know to skip this
832 * part of the file
833 */
834static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
835{
836 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
837 struct extent_map *em = NULL;
838 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
839 u64 end;
840
841 read_lock(&em_tree->lock);
842 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
843 read_unlock(&em_tree->lock);
844
845 if (em) {
846 end = extent_map_end(em);
847 free_extent_map(em);
848 if (end - offset > thresh)
849 return 0;
850 }
851 /* if we already have a nice delalloc here, just stop */
852 thresh /= 2;
853 end = count_range_bits(io_tree, &offset, offset + thresh,
854 thresh, EXTENT_DELALLOC, 1);
855 if (end >= thresh)
856 return 0;
857 return 1;
858}
859
860/*
861 * helper function to walk through a file and find extents
862 * newer than a specific transid, and smaller than thresh.
863 *
864 * This is used by the defragging code to find new and small
865 * extents
866 */
867static int find_new_extents(struct btrfs_root *root,
868 struct inode *inode, u64 newer_than,
869 u64 *off, int thresh)
870{
871 struct btrfs_path *path;
872 struct btrfs_key min_key;
Chris Mason4cb53002011-05-24 15:35:30 -0400873 struct extent_buffer *leaf;
874 struct btrfs_file_extent_item *extent;
875 int type;
876 int ret;
David Sterbaa4689d22011-05-31 17:08:14 +0000877 u64 ino = btrfs_ino(inode);
Chris Mason4cb53002011-05-24 15:35:30 -0400878
879 path = btrfs_alloc_path();
880 if (!path)
881 return -ENOMEM;
882
David Sterbaa4689d22011-05-31 17:08:14 +0000883 min_key.objectid = ino;
Chris Mason4cb53002011-05-24 15:35:30 -0400884 min_key.type = BTRFS_EXTENT_DATA_KEY;
885 min_key.offset = *off;
886
Chris Mason4cb53002011-05-24 15:35:30 -0400887 path->keep_locks = 1;
888
Dulshani Gunawardhana67871252013-10-31 10:33:04 +0530889 while (1) {
Filipe David Borba Manana6174d3c2013-10-01 16:13:42 +0100890 ret = btrfs_search_forward(root, &min_key, path, newer_than);
Chris Mason4cb53002011-05-24 15:35:30 -0400891 if (ret != 0)
892 goto none;
David Sterbaa4689d22011-05-31 17:08:14 +0000893 if (min_key.objectid != ino)
Chris Mason4cb53002011-05-24 15:35:30 -0400894 goto none;
895 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
896 goto none;
897
898 leaf = path->nodes[0];
899 extent = btrfs_item_ptr(leaf, path->slots[0],
900 struct btrfs_file_extent_item);
901
902 type = btrfs_file_extent_type(leaf, extent);
903 if (type == BTRFS_FILE_EXTENT_REG &&
904 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
905 check_defrag_in_cache(inode, min_key.offset, thresh)) {
906 *off = min_key.offset;
907 btrfs_free_path(path);
908 return 0;
909 }
910
911 if (min_key.offset == (u64)-1)
912 goto none;
913
914 min_key.offset++;
915 btrfs_release_path(path);
916 }
917none:
918 btrfs_free_path(path);
919 return -ENOENT;
920}
921
Li Zefan6c282eb2012-06-11 16:03:35 +0800922static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
Liu Bo17ce6ef2012-03-29 09:57:45 -0400923{
924 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
Chris Mason940100a2010-03-10 10:52:59 -0500925 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
Li Zefan6c282eb2012-06-11 16:03:35 +0800926 struct extent_map *em;
927 u64 len = PAGE_CACHE_SIZE;
Chris Mason940100a2010-03-10 10:52:59 -0500928
929 /*
930 * hopefully we have this extent in the tree already, try without
931 * the full extent lock
932 */
933 read_lock(&em_tree->lock);
934 em = lookup_extent_mapping(em_tree, start, len);
935 read_unlock(&em_tree->lock);
936
937 if (!em) {
938 /* get the big lock and read metadata off disk */
Jeff Mahoneyd0082372012-03-01 14:57:19 +0100939 lock_extent(io_tree, start, start + len - 1);
Chris Mason940100a2010-03-10 10:52:59 -0500940 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
Jeff Mahoneyd0082372012-03-01 14:57:19 +0100941 unlock_extent(io_tree, start, start + len - 1);
Chris Mason940100a2010-03-10 10:52:59 -0500942
Dan Carpenter6cf8bfb2010-03-20 11:22:10 +0000943 if (IS_ERR(em))
Li Zefan6c282eb2012-06-11 16:03:35 +0800944 return NULL;
Chris Mason940100a2010-03-10 10:52:59 -0500945 }
946
Li Zefan6c282eb2012-06-11 16:03:35 +0800947 return em;
948}
949
950static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
951{
952 struct extent_map *next;
953 bool ret = true;
954
955 /* this is the last extent */
956 if (em->start + em->len >= i_size_read(inode))
957 return false;
958
959 next = defrag_lookup_extent(inode, em->start + em->len);
960 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
961 ret = false;
962
963 free_extent_map(next);
964 return ret;
965}
966
967static int should_defrag_range(struct inode *inode, u64 start, int thresh,
Andrew Mahonea43a2112012-06-19 21:08:32 -0400968 u64 *last_len, u64 *skip, u64 *defrag_end,
969 int compress)
Li Zefan6c282eb2012-06-11 16:03:35 +0800970{
971 struct extent_map *em;
972 int ret = 1;
973 bool next_mergeable = true;
974
975 /*
976 * make sure that once we start defragging an extent, we keep on
977 * defragging it
978 */
979 if (start < *defrag_end)
980 return 1;
981
982 *skip = 0;
983
984 em = defrag_lookup_extent(inode, start);
985 if (!em)
986 return 0;
987
Chris Mason940100a2010-03-10 10:52:59 -0500988 /* this will cover holes, and inline extents */
Liu Bo17ce6ef2012-03-29 09:57:45 -0400989 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
Chris Mason940100a2010-03-10 10:52:59 -0500990 ret = 0;
Liu Bo17ce6ef2012-03-29 09:57:45 -0400991 goto out;
992 }
993
Li Zefan6c282eb2012-06-11 16:03:35 +0800994 next_mergeable = defrag_check_next_extent(inode, em);
Chris Mason940100a2010-03-10 10:52:59 -0500995
996 /*
Li Zefan6c282eb2012-06-11 16:03:35 +0800997 * we hit a real extent, if it is big or the next extent is not a
998 * real extent, don't bother defragging it
Chris Mason940100a2010-03-10 10:52:59 -0500999 */
Andrew Mahonea43a2112012-06-19 21:08:32 -04001000 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
Li Zefan6c282eb2012-06-11 16:03:35 +08001001 (em->len >= thresh || !next_mergeable))
Chris Mason940100a2010-03-10 10:52:59 -05001002 ret = 0;
Liu Bo17ce6ef2012-03-29 09:57:45 -04001003out:
Chris Mason940100a2010-03-10 10:52:59 -05001004 /*
1005 * last_len ends up being a counter of how many bytes we've defragged.
1006 * every time we choose not to defrag an extent, we reset *last_len
1007 * so that the next tiny extent will force a defrag.
1008 *
1009 * The end result of this is that tiny extents before a single big
1010 * extent will force at least part of that big extent to be defragged.
1011 */
1012 if (ret) {
Chris Mason940100a2010-03-10 10:52:59 -05001013 *defrag_end = extent_map_end(em);
1014 } else {
1015 *last_len = 0;
1016 *skip = extent_map_end(em);
1017 *defrag_end = 0;
1018 }
1019
1020 free_extent_map(em);
1021 return ret;
1022}
1023
Chris Mason4cb53002011-05-24 15:35:30 -04001024/*
1025 * it doesn't do much good to defrag one or two pages
1026 * at a time. This pulls in a nice chunk of pages
1027 * to COW and defrag.
1028 *
1029 * It also makes sure the delalloc code has enough
1030 * dirty data to avoid making new small extents as part
1031 * of the defrag
1032 *
1033 * It's a good idea to start RA on this range
1034 * before calling this.
1035 */
1036static int cluster_pages_for_defrag(struct inode *inode,
1037 struct page **pages,
1038 unsigned long start_index,
1039 int num_pages)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001040{
Chris Mason4cb53002011-05-24 15:35:30 -04001041 unsigned long file_end;
1042 u64 isize = i_size_read(inode);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001043 u64 page_start;
1044 u64 page_end;
Liu Bo1f12bd02012-03-29 09:57:44 -04001045 u64 page_cnt;
Chris Mason4cb53002011-05-24 15:35:30 -04001046 int ret;
1047 int i;
1048 int i_done;
1049 struct btrfs_ordered_extent *ordered;
1050 struct extent_state *cached_state = NULL;
Miao Xie600a45e2012-02-16 15:01:24 +08001051 struct extent_io_tree *tree;
Josef Bacik3b16a4e2011-09-21 15:05:58 -04001052 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
Chris Mason4cb53002011-05-24 15:35:30 -04001053
Chris Mason4cb53002011-05-24 15:35:30 -04001054 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
Liu Bo1f12bd02012-03-29 09:57:44 -04001055 if (!isize || start_index > file_end)
1056 return 0;
1057
1058 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
Chris Mason4cb53002011-05-24 15:35:30 -04001059
1060 ret = btrfs_delalloc_reserve_space(inode,
Liu Bo1f12bd02012-03-29 09:57:44 -04001061 page_cnt << PAGE_CACHE_SHIFT);
Chris Mason4cb53002011-05-24 15:35:30 -04001062 if (ret)
1063 return ret;
Chris Mason4cb53002011-05-24 15:35:30 -04001064 i_done = 0;
Miao Xie600a45e2012-02-16 15:01:24 +08001065 tree = &BTRFS_I(inode)->io_tree;
Chris Mason4cb53002011-05-24 15:35:30 -04001066
1067 /* step one, lock all the pages */
Liu Bo1f12bd02012-03-29 09:57:44 -04001068 for (i = 0; i < page_cnt; i++) {
Chris Mason4cb53002011-05-24 15:35:30 -04001069 struct page *page;
Miao Xie600a45e2012-02-16 15:01:24 +08001070again:
Josef Bacika94733d2011-07-11 10:47:06 -04001071 page = find_or_create_page(inode->i_mapping,
Miao Xie600a45e2012-02-16 15:01:24 +08001072 start_index + i, mask);
Chris Mason4cb53002011-05-24 15:35:30 -04001073 if (!page)
1074 break;
1075
Miao Xie600a45e2012-02-16 15:01:24 +08001076 page_start = page_offset(page);
1077 page_end = page_start + PAGE_CACHE_SIZE - 1;
1078 while (1) {
Jeff Mahoneyd0082372012-03-01 14:57:19 +01001079 lock_extent(tree, page_start, page_end);
Miao Xie600a45e2012-02-16 15:01:24 +08001080 ordered = btrfs_lookup_ordered_extent(inode,
1081 page_start);
Jeff Mahoneyd0082372012-03-01 14:57:19 +01001082 unlock_extent(tree, page_start, page_end);
Miao Xie600a45e2012-02-16 15:01:24 +08001083 if (!ordered)
1084 break;
1085
1086 unlock_page(page);
1087 btrfs_start_ordered_extent(inode, ordered, 1);
1088 btrfs_put_ordered_extent(ordered);
1089 lock_page(page);
Liu Bo1f12bd02012-03-29 09:57:44 -04001090 /*
1091 * we unlocked the page above, so we need check if
1092 * it was released or not.
1093 */
1094 if (page->mapping != inode->i_mapping) {
1095 unlock_page(page);
1096 page_cache_release(page);
1097 goto again;
1098 }
Miao Xie600a45e2012-02-16 15:01:24 +08001099 }
1100
Chris Mason4cb53002011-05-24 15:35:30 -04001101 if (!PageUptodate(page)) {
1102 btrfs_readpage(NULL, page);
1103 lock_page(page);
1104 if (!PageUptodate(page)) {
1105 unlock_page(page);
1106 page_cache_release(page);
1107 ret = -EIO;
1108 break;
1109 }
1110 }
Miao Xie600a45e2012-02-16 15:01:24 +08001111
Miao Xie600a45e2012-02-16 15:01:24 +08001112 if (page->mapping != inode->i_mapping) {
1113 unlock_page(page);
1114 page_cache_release(page);
1115 goto again;
1116 }
1117
Chris Mason4cb53002011-05-24 15:35:30 -04001118 pages[i] = page;
1119 i_done++;
1120 }
1121 if (!i_done || ret)
1122 goto out;
1123
1124 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1125 goto out;
1126
1127 /*
1128 * so now we have a nice long stream of locked
1129 * and up to date pages, lets wait on them
1130 */
1131 for (i = 0; i < i_done; i++)
1132 wait_on_page_writeback(pages[i]);
1133
1134 page_start = page_offset(pages[0]);
1135 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1136
1137 lock_extent_bits(&BTRFS_I(inode)->io_tree,
Jeff Mahoneyd0082372012-03-01 14:57:19 +01001138 page_start, page_end - 1, 0, &cached_state);
Chris Mason4cb53002011-05-24 15:35:30 -04001139 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1140 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
Liu Bo9e8a4a82012-09-05 19:10:51 -06001141 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1142 &cached_state, GFP_NOFS);
Chris Mason4cb53002011-05-24 15:35:30 -04001143
Liu Bo1f12bd02012-03-29 09:57:44 -04001144 if (i_done != page_cnt) {
Josef Bacik9e0baf62011-07-15 15:16:44 +00001145 spin_lock(&BTRFS_I(inode)->lock);
1146 BTRFS_I(inode)->outstanding_extents++;
1147 spin_unlock(&BTRFS_I(inode)->lock);
Chris Mason4cb53002011-05-24 15:35:30 -04001148 btrfs_delalloc_release_space(inode,
Liu Bo1f12bd02012-03-29 09:57:44 -04001149 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
Chris Mason4cb53002011-05-24 15:35:30 -04001150 }
1151
1152
Liu Bo9e8a4a82012-09-05 19:10:51 -06001153 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1154 &cached_state, GFP_NOFS);
Chris Mason4cb53002011-05-24 15:35:30 -04001155
1156 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1157 page_start, page_end - 1, &cached_state,
1158 GFP_NOFS);
1159
1160 for (i = 0; i < i_done; i++) {
1161 clear_page_dirty_for_io(pages[i]);
1162 ClearPageChecked(pages[i]);
1163 set_page_extent_mapped(pages[i]);
1164 set_page_dirty(pages[i]);
1165 unlock_page(pages[i]);
1166 page_cache_release(pages[i]);
1167 }
1168 return i_done;
1169out:
1170 for (i = 0; i < i_done; i++) {
1171 unlock_page(pages[i]);
1172 page_cache_release(pages[i]);
1173 }
Liu Bo1f12bd02012-03-29 09:57:44 -04001174 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
Chris Mason4cb53002011-05-24 15:35:30 -04001175 return ret;
1176
1177}
1178
1179int btrfs_defrag_file(struct inode *inode, struct file *file,
1180 struct btrfs_ioctl_defrag_range_args *range,
1181 u64 newer_than, unsigned long max_to_defrag)
1182{
1183 struct btrfs_root *root = BTRFS_I(inode)->root;
Chris Mason4cb53002011-05-24 15:35:30 -04001184 struct file_ra_state *ra = NULL;
1185 unsigned long last_index;
Li Zefan151a31b2011-09-02 15:56:39 +08001186 u64 isize = i_size_read(inode);
Chris Mason940100a2010-03-10 10:52:59 -05001187 u64 last_len = 0;
1188 u64 skip = 0;
1189 u64 defrag_end = 0;
Chris Mason4cb53002011-05-24 15:35:30 -04001190 u64 newer_off = range->start;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001191 unsigned long i;
Li Zefan008873e2011-09-02 15:57:07 +08001192 unsigned long ra_index = 0;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001193 int ret;
Chris Mason4cb53002011-05-24 15:35:30 -04001194 int defrag_count = 0;
Li Zefan1a419d82010-10-25 15:12:50 +08001195 int compress_type = BTRFS_COMPRESS_ZLIB;
Chris Mason4cb53002011-05-24 15:35:30 -04001196 int extent_thresh = range->extent_thresh;
Li Zefan008873e2011-09-02 15:57:07 +08001197 int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1198 int cluster = max_cluster;
Chris Mason4cb53002011-05-24 15:35:30 -04001199 u64 new_align = ~((u64)128 * 1024 - 1);
1200 struct page **pages = NULL;
1201
Liu Bo0abd5b12013-04-16 09:20:28 +00001202 if (isize == 0)
1203 return 0;
1204
1205 if (range->start >= isize)
1206 return -EINVAL;
Li Zefan1a419d82010-10-25 15:12:50 +08001207
1208 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1209 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1210 return -EINVAL;
1211 if (range->compress_type)
1212 compress_type = range->compress_type;
1213 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001214
Liu Bo0abd5b12013-04-16 09:20:28 +00001215 if (extent_thresh == 0)
1216 extent_thresh = 256 * 1024;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001217
Chris Mason4cb53002011-05-24 15:35:30 -04001218 /*
1219 * if we were not given a file, allocate a readahead
1220 * context
1221 */
1222 if (!file) {
1223 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1224 if (!ra)
1225 return -ENOMEM;
1226 file_ra_state_init(ra, inode->i_mapping);
1227 } else {
1228 ra = &file->f_ra;
1229 }
1230
Dulshani Gunawardhanad9b0d9b2013-10-31 10:32:18 +05301231 pages = kmalloc_array(max_cluster, sizeof(struct page *),
Chris Mason4cb53002011-05-24 15:35:30 -04001232 GFP_NOFS);
1233 if (!pages) {
1234 ret = -ENOMEM;
1235 goto out_ra;
1236 }
1237
1238 /* find the last page to defrag */
Chris Mason1e701a32010-03-11 09:42:04 -05001239 if (range->start + range->len > range->start) {
Li Zefan151a31b2011-09-02 15:56:39 +08001240 last_index = min_t(u64, isize - 1,
Chris Mason1e701a32010-03-11 09:42:04 -05001241 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1242 } else {
Li Zefan151a31b2011-09-02 15:56:39 +08001243 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
Chris Mason1e701a32010-03-11 09:42:04 -05001244 }
1245
Chris Mason4cb53002011-05-24 15:35:30 -04001246 if (newer_than) {
1247 ret = find_new_extents(root, inode, newer_than,
1248 &newer_off, 64 * 1024);
1249 if (!ret) {
1250 range->start = newer_off;
1251 /*
1252 * we always align our defrag to help keep
1253 * the extents in the file evenly spaced
1254 */
1255 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
Chris Mason4cb53002011-05-24 15:35:30 -04001256 } else
1257 goto out_ra;
1258 } else {
1259 i = range->start >> PAGE_CACHE_SHIFT;
1260 }
1261 if (!max_to_defrag)
Liu Bo7ec31b52012-01-26 15:01:12 -05001262 max_to_defrag = last_index + 1;
Chris Mason4cb53002011-05-24 15:35:30 -04001263
Li Zefan2a0f7f52011-10-10 15:43:34 -04001264 /*
1265 * make writeback starts from i, so the defrag range can be
1266 * written sequentially.
1267 */
1268 if (i < inode->i_mapping->writeback_index)
1269 inode->i_mapping->writeback_index = i;
1270
Chris Masonf7f43cc2011-10-11 11:41:40 -04001271 while (i <= last_index && defrag_count < max_to_defrag &&
1272 (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1273 PAGE_CACHE_SHIFT)) {
Chris Mason4cb53002011-05-24 15:35:30 -04001274 /*
1275 * make sure we stop running if someone unmounts
1276 * the FS
1277 */
1278 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1279 break;
1280
David Sterba210549e2013-02-09 23:38:06 +00001281 if (btrfs_defrag_cancelled(root->fs_info)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05001282 printk(KERN_DEBUG "BTRFS: defrag_file cancelled\n");
David Sterba210549e2013-02-09 23:38:06 +00001283 ret = -EAGAIN;
1284 break;
1285 }
1286
Liu Bo66c26892012-03-29 09:57:45 -04001287 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
Li Zefan6c282eb2012-06-11 16:03:35 +08001288 extent_thresh, &last_len, &skip,
Andrew Mahonea43a2112012-06-19 21:08:32 -04001289 &defrag_end, range->flags &
1290 BTRFS_DEFRAG_RANGE_COMPRESS)) {
Chris Mason940100a2010-03-10 10:52:59 -05001291 unsigned long next;
1292 /*
1293 * the should_defrag function tells us how much to skip
1294 * bump our counter by the suggested amount
1295 */
1296 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1297 i = max(i + 1, next);
1298 continue;
1299 }
Li Zefan008873e2011-09-02 15:57:07 +08001300
1301 if (!newer_than) {
1302 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1303 PAGE_CACHE_SHIFT) - i;
1304 cluster = min(cluster, max_cluster);
1305 } else {
1306 cluster = max_cluster;
1307 }
1308
Li Zefan008873e2011-09-02 15:57:07 +08001309 if (i + cluster > ra_index) {
1310 ra_index = max(i, ra_index);
1311 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1312 cluster);
1313 ra_index += max_cluster;
1314 }
Chris Mason940100a2010-03-10 10:52:59 -05001315
Liu Boecb8bea2012-03-29 09:57:44 -04001316 mutex_lock(&inode->i_mutex);
Filipe David Borba Manana633085c2013-08-16 15:23:33 +01001317 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1318 BTRFS_I(inode)->force_compress = compress_type;
Li Zefan008873e2011-09-02 15:57:07 +08001319 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
Liu Boecb8bea2012-03-29 09:57:44 -04001320 if (ret < 0) {
1321 mutex_unlock(&inode->i_mutex);
Chris Mason4cb53002011-05-24 15:35:30 -04001322 goto out_ra;
Liu Boecb8bea2012-03-29 09:57:44 -04001323 }
Chris Mason940100a2010-03-10 10:52:59 -05001324
Chris Mason4cb53002011-05-24 15:35:30 -04001325 defrag_count += ret;
Namjae Jeond0e1d662012-12-11 16:00:21 -08001326 balance_dirty_pages_ratelimited(inode->i_mapping);
Liu Boecb8bea2012-03-29 09:57:44 -04001327 mutex_unlock(&inode->i_mutex);
Chris Mason4cb53002011-05-24 15:35:30 -04001328
1329 if (newer_than) {
1330 if (newer_off == (u64)-1)
1331 break;
1332
Liu Boe1f041e2012-03-29 09:57:45 -04001333 if (ret > 0)
1334 i += ret;
1335
Chris Mason4cb53002011-05-24 15:35:30 -04001336 newer_off = max(newer_off + 1,
1337 (u64)i << PAGE_CACHE_SHIFT);
1338
1339 ret = find_new_extents(root, inode,
1340 newer_than, &newer_off,
1341 64 * 1024);
1342 if (!ret) {
1343 range->start = newer_off;
1344 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
Chris Mason4cb53002011-05-24 15:35:30 -04001345 } else {
1346 break;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001347 }
Chris Mason4cb53002011-05-24 15:35:30 -04001348 } else {
Li Zefan008873e2011-09-02 15:57:07 +08001349 if (ret > 0) {
Li Zefancbcc8322011-09-02 15:56:25 +08001350 i += ret;
Li Zefan008873e2011-09-02 15:57:07 +08001351 last_len += ret << PAGE_CACHE_SHIFT;
1352 } else {
Li Zefancbcc8322011-09-02 15:56:25 +08001353 i++;
Li Zefan008873e2011-09-02 15:57:07 +08001354 last_len = 0;
1355 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001356 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001357 }
1358
Chris Mason1e701a32010-03-11 09:42:04 -05001359 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1360 filemap_flush(inode->i_mapping);
1361
1362 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1363 /* the filemap_flush will queue IO into the worker threads, but
1364 * we have to make sure the IO is actually started and that
1365 * ordered extents get created before we return
1366 */
1367 atomic_inc(&root->fs_info->async_submit_draining);
1368 while (atomic_read(&root->fs_info->nr_async_submits) ||
1369 atomic_read(&root->fs_info->async_delalloc_pages)) {
1370 wait_event(root->fs_info->async_submit_wait,
1371 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1372 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1373 }
1374 atomic_dec(&root->fs_info->async_submit_draining);
Chris Mason1e701a32010-03-11 09:42:04 -05001375 }
1376
Li Zefan1a419d82010-10-25 15:12:50 +08001377 if (range->compress_type == BTRFS_COMPRESS_LZO) {
Mitch Harder2b0ce2c2012-07-24 11:58:43 -06001378 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
Li Zefan1a419d82010-10-25 15:12:50 +08001379 }
1380
Diego Calleja60ccf822011-09-01 16:33:57 +02001381 ret = defrag_count;
Chris Mason940100a2010-03-10 10:52:59 -05001382
Chris Mason4cb53002011-05-24 15:35:30 -04001383out_ra:
Filipe David Borba Manana633085c2013-08-16 15:23:33 +01001384 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1385 mutex_lock(&inode->i_mutex);
1386 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1387 mutex_unlock(&inode->i_mutex);
1388 }
Chris Mason4cb53002011-05-24 15:35:30 -04001389 if (!file)
1390 kfree(ra);
1391 kfree(pages);
Chris Mason940100a2010-03-10 10:52:59 -05001392 return ret;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001393}
1394
Miao Xie198605a2012-11-26 08:43:45 +00001395static noinline int btrfs_ioctl_resize(struct file *file,
Yan, Zheng76dda932009-09-21 16:00:26 -04001396 void __user *arg)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001397{
1398 u64 new_size;
1399 u64 old_size;
1400 u64 devid = 1;
Al Viro496ad9a2013-01-23 17:07:38 -05001401 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001402 struct btrfs_ioctl_vol_args *vol_args;
1403 struct btrfs_trans_handle *trans;
1404 struct btrfs_device *device = NULL;
1405 char *sizestr;
1406 char *devstr = NULL;
1407 int ret = 0;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001408 int mod = 0;
1409
Chris Masone441d542009-01-05 16:57:23 -05001410 if (!capable(CAP_SYS_ADMIN))
1411 return -EPERM;
1412
Miao Xie198605a2012-11-26 08:43:45 +00001413 ret = mnt_want_write_file(file);
1414 if (ret)
1415 return ret;
1416
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01001417 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1418 1)) {
Miao Xie97547672012-12-21 10:38:50 +00001419 mnt_drop_write_file(file);
Anand Jaine57138b2013-08-21 11:44:48 +08001420 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001421 }
1422
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01001423 mutex_lock(&root->fs_info->volume_mutex);
Li Zefandae7b662009-04-08 15:06:54 +08001424 vol_args = memdup_user(arg, sizeof(*vol_args));
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001425 if (IS_ERR(vol_args)) {
1426 ret = PTR_ERR(vol_args);
1427 goto out;
1428 }
Mark Fasheh5516e592008-07-24 12:20:14 -04001429
1430 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001431
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001432 sizestr = vol_args->name;
1433 devstr = strchr(sizestr, ':');
1434 if (devstr) {
1435 char *end;
1436 sizestr = devstr + 1;
1437 *devstr = '\0';
1438 devstr = vol_args->name;
1439 devid = simple_strtoull(devstr, &end, 10);
Miao Xiedfd79822012-12-21 09:21:30 +00001440 if (!devid) {
1441 ret = -EINVAL;
1442 goto out_free;
1443 }
Frank Holtonefe120a2013-12-20 11:37:06 -05001444 btrfs_info(root->fs_info, "resizing devid %llu", devid);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001445 }
Miao Xiedba60f32012-12-21 09:19:51 +00001446
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01001447 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001448 if (!device) {
Frank Holtonefe120a2013-12-20 11:37:06 -05001449 btrfs_info(root->fs_info, "resizer unable to find device %llu",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02001450 devid);
Miao Xiedfd79822012-12-21 09:21:30 +00001451 ret = -ENODEV;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001452 goto out_free;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001453 }
Miao Xiedba60f32012-12-21 09:19:51 +00001454
1455 if (!device->writeable) {
Frank Holtonefe120a2013-12-20 11:37:06 -05001456 btrfs_info(root->fs_info,
1457 "resizer unable to apply on readonly device %llu",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02001458 devid);
Miao Xiedfd79822012-12-21 09:21:30 +00001459 ret = -EPERM;
Liu Bo4e42ae12012-06-14 02:23:19 -06001460 goto out_free;
1461 }
1462
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001463 if (!strcmp(sizestr, "max"))
1464 new_size = device->bdev->bd_inode->i_size;
1465 else {
1466 if (sizestr[0] == '-') {
1467 mod = -1;
1468 sizestr++;
1469 } else if (sizestr[0] == '+') {
1470 mod = 1;
1471 sizestr++;
1472 }
Akinobu Mita91748462010-02-28 10:59:11 +00001473 new_size = memparse(sizestr, NULL);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001474 if (new_size == 0) {
1475 ret = -EINVAL;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001476 goto out_free;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001477 }
1478 }
1479
Stefan Behrens63a212a2012-11-05 18:29:28 +01001480 if (device->is_tgtdev_for_dev_replace) {
Miao Xiedfd79822012-12-21 09:21:30 +00001481 ret = -EPERM;
Stefan Behrens63a212a2012-11-05 18:29:28 +01001482 goto out_free;
1483 }
1484
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001485 old_size = device->total_bytes;
1486
1487 if (mod < 0) {
1488 if (new_size > old_size) {
1489 ret = -EINVAL;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001490 goto out_free;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001491 }
1492 new_size = old_size - new_size;
1493 } else if (mod > 0) {
Wenliang Faneb8052e2013-12-20 15:28:56 +08001494 if (new_size > ULLONG_MAX - old_size) {
1495 ret = -EINVAL;
1496 goto out_free;
1497 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001498 new_size = old_size + new_size;
1499 }
1500
1501 if (new_size < 256 * 1024 * 1024) {
1502 ret = -EINVAL;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001503 goto out_free;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001504 }
1505 if (new_size > device->bdev->bd_inode->i_size) {
1506 ret = -EFBIG;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001507 goto out_free;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001508 }
1509
1510 do_div(new_size, root->sectorsize);
1511 new_size *= root->sectorsize;
1512
Frank Holtonefe120a2013-12-20 11:37:06 -05001513 printk_in_rcu(KERN_INFO "BTRFS: new size for %s is %llu\n",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02001514 rcu_str_deref(device->name), new_size);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001515
1516 if (new_size > old_size) {
Yan, Zhenga22285a2010-05-16 10:48:46 -04001517 trans = btrfs_start_transaction(root, 0);
Tsutomu Itoh98d5dc12011-01-20 06:19:37 +00001518 if (IS_ERR(trans)) {
1519 ret = PTR_ERR(trans);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001520 goto out_free;
Tsutomu Itoh98d5dc12011-01-20 06:19:37 +00001521 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001522 ret = btrfs_grow_device(trans, device, new_size);
1523 btrfs_commit_transaction(trans, root);
Mike Fleetwoodece7d202011-11-18 18:55:01 +00001524 } else if (new_size < old_size) {
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001525 ret = btrfs_shrink_device(device, new_size);
jeff.liu0253f402012-10-27 12:06:39 +00001526 } /* equal, nothing need to do */
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001527
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001528out_free:
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001529 kfree(vol_args);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001530out:
1531 mutex_unlock(&root->fs_info->volume_mutex);
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01001532 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
Ilya Dryomov18f39c42013-01-20 15:57:57 +02001533 mnt_drop_write_file(file);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001534 return ret;
1535}
1536
Sage Weil72fd0322010-10-29 15:41:32 -04001537static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001538 char *name, unsigned long fd, int subvol,
1539 u64 *transid, bool readonly,
Miao Xie8696c532013-02-07 06:02:44 +00001540 struct btrfs_qgroup_inherit *inherit)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001541{
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001542 int namelen;
Chris Mason3de45862008-11-17 21:02:50 -05001543 int ret = 0;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001544
Liu Boa874a632012-06-29 03:58:46 -06001545 ret = mnt_want_write_file(file);
1546 if (ret)
1547 goto out;
1548
Sage Weil72fd0322010-10-29 15:41:32 -04001549 namelen = strlen(name);
1550 if (strchr(name, '/')) {
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001551 ret = -EINVAL;
Liu Boa874a632012-06-29 03:58:46 -06001552 goto out_drop_write;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001553 }
1554
Chris Mason16780ca2012-02-20 22:14:55 -05001555 if (name[0] == '.' &&
1556 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1557 ret = -EEXIST;
Liu Boa874a632012-06-29 03:58:46 -06001558 goto out_drop_write;
Chris Mason16780ca2012-02-20 22:14:55 -05001559 }
1560
Chris Mason3de45862008-11-17 21:02:50 -05001561 if (subvol) {
Sage Weil72fd0322010-10-29 15:41:32 -04001562 ret = btrfs_mksubvol(&file->f_path, name, namelen,
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001563 NULL, transid, readonly, inherit);
Christoph Hellwigcb8e7092008-10-09 13:39:39 -04001564 } else {
Al Viro2903ff02012-08-28 12:52:22 -04001565 struct fd src = fdget(fd);
Chris Mason3de45862008-11-17 21:02:50 -05001566 struct inode *src_inode;
Al Viro2903ff02012-08-28 12:52:22 -04001567 if (!src.file) {
Chris Mason3de45862008-11-17 21:02:50 -05001568 ret = -EINVAL;
Liu Boa874a632012-06-29 03:58:46 -06001569 goto out_drop_write;
Chris Mason3de45862008-11-17 21:02:50 -05001570 }
1571
Al Viro496ad9a2013-01-23 17:07:38 -05001572 src_inode = file_inode(src.file);
1573 if (src_inode->i_sb != file_inode(file)->i_sb) {
Frank Holtonefe120a2013-12-20 11:37:06 -05001574 btrfs_info(BTRFS_I(src_inode)->root->fs_info,
1575 "Snapshot src from another FS");
Chris Mason3de45862008-11-17 21:02:50 -05001576 ret = -EINVAL;
Al Viroecd18812012-08-26 21:20:24 -04001577 } else {
1578 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1579 BTRFS_I(src_inode)->root,
1580 transid, readonly, inherit);
Chris Mason3de45862008-11-17 21:02:50 -05001581 }
Al Viro2903ff02012-08-28 12:52:22 -04001582 fdput(src);
Christoph Hellwigcb8e7092008-10-09 13:39:39 -04001583 }
Liu Boa874a632012-06-29 03:58:46 -06001584out_drop_write:
1585 mnt_drop_write_file(file);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001586out:
Sage Weil72fd0322010-10-29 15:41:32 -04001587 return ret;
1588}
1589
1590static noinline int btrfs_ioctl_snap_create(struct file *file,
Li Zefanfa0d2b92010-12-20 15:53:28 +08001591 void __user *arg, int subvol)
Sage Weil72fd0322010-10-29 15:41:32 -04001592{
Li Zefanfa0d2b92010-12-20 15:53:28 +08001593 struct btrfs_ioctl_vol_args *vol_args;
Sage Weil72fd0322010-10-29 15:41:32 -04001594 int ret;
1595
Li Zefanfa0d2b92010-12-20 15:53:28 +08001596 vol_args = memdup_user(arg, sizeof(*vol_args));
1597 if (IS_ERR(vol_args))
1598 return PTR_ERR(vol_args);
1599 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
Sage Weil72fd0322010-10-29 15:41:32 -04001600
Li Zefanfa0d2b92010-12-20 15:53:28 +08001601 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
Li Zefanb83cc962010-12-20 16:04:08 +08001602 vol_args->fd, subvol,
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001603 NULL, false, NULL);
Li Zefanfdfb1e42010-12-10 06:41:56 +00001604
Li Zefanfa0d2b92010-12-20 15:53:28 +08001605 kfree(vol_args);
1606 return ret;
1607}
Li Zefanfdfb1e42010-12-10 06:41:56 +00001608
Li Zefanfa0d2b92010-12-20 15:53:28 +08001609static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1610 void __user *arg, int subvol)
1611{
1612 struct btrfs_ioctl_vol_args_v2 *vol_args;
1613 int ret;
1614 u64 transid = 0;
1615 u64 *ptr = NULL;
Li Zefanb83cc962010-12-20 16:04:08 +08001616 bool readonly = false;
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001617 struct btrfs_qgroup_inherit *inherit = NULL;
Li Zefanfdfb1e42010-12-10 06:41:56 +00001618
Li Zefanfa0d2b92010-12-20 15:53:28 +08001619 vol_args = memdup_user(arg, sizeof(*vol_args));
1620 if (IS_ERR(vol_args))
1621 return PTR_ERR(vol_args);
1622 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
Sage Weil75eaa0e2010-12-10 00:36:28 +00001623
Li Zefanb83cc962010-12-20 16:04:08 +08001624 if (vol_args->flags &
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001625 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1626 BTRFS_SUBVOL_QGROUP_INHERIT)) {
Li Zefanb83cc962010-12-20 16:04:08 +08001627 ret = -EOPNOTSUPP;
Li Zefanfa0d2b92010-12-20 15:53:28 +08001628 goto out;
Sage Weil72fd0322010-10-29 15:41:32 -04001629 }
Li Zefanfa0d2b92010-12-20 15:53:28 +08001630
1631 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1632 ptr = &transid;
Li Zefanb83cc962010-12-20 16:04:08 +08001633 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1634 readonly = true;
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001635 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1636 if (vol_args->size > PAGE_CACHE_SIZE) {
1637 ret = -EINVAL;
1638 goto out;
1639 }
1640 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1641 if (IS_ERR(inherit)) {
1642 ret = PTR_ERR(inherit);
1643 goto out;
1644 }
1645 }
Li Zefanfa0d2b92010-12-20 15:53:28 +08001646
1647 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001648 vol_args->fd, subvol, ptr,
Miao Xie8696c532013-02-07 06:02:44 +00001649 readonly, inherit);
Li Zefanfa0d2b92010-12-20 15:53:28 +08001650
1651 if (ret == 0 && ptr &&
1652 copy_to_user(arg +
1653 offsetof(struct btrfs_ioctl_vol_args_v2,
1654 transid), ptr, sizeof(*ptr)))
1655 ret = -EFAULT;
Li Zefanfdfb1e42010-12-10 06:41:56 +00001656out:
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001657 kfree(vol_args);
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001658 kfree(inherit);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001659 return ret;
1660}
1661
Li Zefan0caa1022010-12-20 16:30:25 +08001662static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1663 void __user *arg)
1664{
Al Viro496ad9a2013-01-23 17:07:38 -05001665 struct inode *inode = file_inode(file);
Li Zefan0caa1022010-12-20 16:30:25 +08001666 struct btrfs_root *root = BTRFS_I(inode)->root;
1667 int ret = 0;
1668 u64 flags = 0;
1669
Li Zefan33345d012011-04-20 10:31:50 +08001670 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
Li Zefan0caa1022010-12-20 16:30:25 +08001671 return -EINVAL;
1672
1673 down_read(&root->fs_info->subvol_sem);
1674 if (btrfs_root_readonly(root))
1675 flags |= BTRFS_SUBVOL_RDONLY;
1676 up_read(&root->fs_info->subvol_sem);
1677
1678 if (copy_to_user(arg, &flags, sizeof(flags)))
1679 ret = -EFAULT;
1680
1681 return ret;
1682}
1683
1684static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1685 void __user *arg)
1686{
Al Viro496ad9a2013-01-23 17:07:38 -05001687 struct inode *inode = file_inode(file);
Li Zefan0caa1022010-12-20 16:30:25 +08001688 struct btrfs_root *root = BTRFS_I(inode)->root;
1689 struct btrfs_trans_handle *trans;
1690 u64 root_flags;
1691 u64 flags;
1692 int ret = 0;
1693
Liu Bob9ca0662012-06-29 03:58:49 -06001694 ret = mnt_want_write_file(file);
1695 if (ret)
1696 goto out;
Li Zefan0caa1022010-12-20 16:30:25 +08001697
Liu Bob9ca0662012-06-29 03:58:49 -06001698 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1699 ret = -EINVAL;
1700 goto out_drop_write;
1701 }
Li Zefan0caa1022010-12-20 16:30:25 +08001702
Liu Bob9ca0662012-06-29 03:58:49 -06001703 if (copy_from_user(&flags, arg, sizeof(flags))) {
1704 ret = -EFAULT;
1705 goto out_drop_write;
1706 }
Li Zefan0caa1022010-12-20 16:30:25 +08001707
Liu Bob9ca0662012-06-29 03:58:49 -06001708 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1709 ret = -EINVAL;
1710 goto out_drop_write;
1711 }
Li Zefan0caa1022010-12-20 16:30:25 +08001712
Liu Bob9ca0662012-06-29 03:58:49 -06001713 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1714 ret = -EOPNOTSUPP;
1715 goto out_drop_write;
1716 }
Li Zefan0caa1022010-12-20 16:30:25 +08001717
Liu Bob9ca0662012-06-29 03:58:49 -06001718 if (!inode_owner_or_capable(inode)) {
1719 ret = -EACCES;
1720 goto out_drop_write;
1721 }
Li Zefanb4dc2b82011-02-16 06:06:34 +00001722
Li Zefan0caa1022010-12-20 16:30:25 +08001723 down_write(&root->fs_info->subvol_sem);
1724
1725 /* nothing to do */
1726 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
Liu Bob9ca0662012-06-29 03:58:49 -06001727 goto out_drop_sem;
Li Zefan0caa1022010-12-20 16:30:25 +08001728
1729 root_flags = btrfs_root_flags(&root->root_item);
David Sterba2c686532013-12-16 17:34:17 +01001730 if (flags & BTRFS_SUBVOL_RDONLY) {
Li Zefan0caa1022010-12-20 16:30:25 +08001731 btrfs_set_root_flags(&root->root_item,
1732 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
David Sterba2c686532013-12-16 17:34:17 +01001733 } else {
1734 /*
1735 * Block RO -> RW transition if this subvolume is involved in
1736 * send
1737 */
1738 spin_lock(&root->root_item_lock);
1739 if (root->send_in_progress == 0) {
1740 btrfs_set_root_flags(&root->root_item,
Li Zefan0caa1022010-12-20 16:30:25 +08001741 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
David Sterba2c686532013-12-16 17:34:17 +01001742 spin_unlock(&root->root_item_lock);
1743 } else {
1744 spin_unlock(&root->root_item_lock);
1745 btrfs_warn(root->fs_info,
1746 "Attempt to set subvolume %llu read-write during send",
1747 root->root_key.objectid);
1748 ret = -EPERM;
1749 goto out_drop_sem;
1750 }
1751 }
Li Zefan0caa1022010-12-20 16:30:25 +08001752
1753 trans = btrfs_start_transaction(root, 1);
1754 if (IS_ERR(trans)) {
1755 ret = PTR_ERR(trans);
1756 goto out_reset;
1757 }
1758
Li Zefanb4dc2b82011-02-16 06:06:34 +00001759 ret = btrfs_update_root(trans, root->fs_info->tree_root,
Li Zefan0caa1022010-12-20 16:30:25 +08001760 &root->root_key, &root->root_item);
1761
1762 btrfs_commit_transaction(trans, root);
1763out_reset:
1764 if (ret)
1765 btrfs_set_root_flags(&root->root_item, root_flags);
Liu Bob9ca0662012-06-29 03:58:49 -06001766out_drop_sem:
Li Zefan0caa1022010-12-20 16:30:25 +08001767 up_write(&root->fs_info->subvol_sem);
Liu Bob9ca0662012-06-29 03:58:49 -06001768out_drop_write:
1769 mnt_drop_write_file(file);
1770out:
Li Zefan0caa1022010-12-20 16:30:25 +08001771 return ret;
1772}
1773
Yan, Zheng76dda932009-09-21 16:00:26 -04001774/*
1775 * helper to check if the subvolume references other subvolumes
1776 */
1777static noinline int may_destroy_subvol(struct btrfs_root *root)
1778{
1779 struct btrfs_path *path;
Josef Bacik175a2b82013-08-12 15:36:44 -04001780 struct btrfs_dir_item *di;
Yan, Zheng76dda932009-09-21 16:00:26 -04001781 struct btrfs_key key;
Josef Bacik175a2b82013-08-12 15:36:44 -04001782 u64 dir_id;
Yan, Zheng76dda932009-09-21 16:00:26 -04001783 int ret;
1784
1785 path = btrfs_alloc_path();
1786 if (!path)
1787 return -ENOMEM;
1788
Josef Bacik175a2b82013-08-12 15:36:44 -04001789 /* Make sure this root isn't set as the default subvol */
1790 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1791 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1792 dir_id, "default", 7, 0);
1793 if (di && !IS_ERR(di)) {
1794 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1795 if (key.objectid == root->root_key.objectid) {
1796 ret = -ENOTEMPTY;
1797 goto out;
1798 }
1799 btrfs_release_path(path);
1800 }
1801
Yan, Zheng76dda932009-09-21 16:00:26 -04001802 key.objectid = root->root_key.objectid;
1803 key.type = BTRFS_ROOT_REF_KEY;
1804 key.offset = (u64)-1;
1805
1806 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1807 &key, path, 0, 0);
1808 if (ret < 0)
1809 goto out;
1810 BUG_ON(ret == 0);
1811
1812 ret = 0;
1813 if (path->slots[0] > 0) {
1814 path->slots[0]--;
1815 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1816 if (key.objectid == root->root_key.objectid &&
1817 key.type == BTRFS_ROOT_REF_KEY)
1818 ret = -ENOTEMPTY;
1819 }
1820out:
1821 btrfs_free_path(path);
1822 return ret;
1823}
1824
Chris Masonac8e9812010-02-28 15:39:26 -05001825static noinline int key_in_sk(struct btrfs_key *key,
1826 struct btrfs_ioctl_search_key *sk)
1827{
Chris Masonabc6e132010-03-18 12:10:08 -04001828 struct btrfs_key test;
1829 int ret;
1830
1831 test.objectid = sk->min_objectid;
1832 test.type = sk->min_type;
1833 test.offset = sk->min_offset;
1834
1835 ret = btrfs_comp_cpu_keys(key, &test);
1836 if (ret < 0)
Chris Masonac8e9812010-02-28 15:39:26 -05001837 return 0;
Chris Masonabc6e132010-03-18 12:10:08 -04001838
1839 test.objectid = sk->max_objectid;
1840 test.type = sk->max_type;
1841 test.offset = sk->max_offset;
1842
1843 ret = btrfs_comp_cpu_keys(key, &test);
1844 if (ret > 0)
Chris Masonac8e9812010-02-28 15:39:26 -05001845 return 0;
1846 return 1;
1847}
1848
1849static noinline int copy_to_sk(struct btrfs_root *root,
1850 struct btrfs_path *path,
1851 struct btrfs_key *key,
1852 struct btrfs_ioctl_search_key *sk,
1853 char *buf,
1854 unsigned long *sk_offset,
1855 int *num_found)
1856{
1857 u64 found_transid;
1858 struct extent_buffer *leaf;
1859 struct btrfs_ioctl_search_header sh;
1860 unsigned long item_off;
1861 unsigned long item_len;
1862 int nritems;
1863 int i;
1864 int slot;
Chris Masonac8e9812010-02-28 15:39:26 -05001865 int ret = 0;
1866
1867 leaf = path->nodes[0];
1868 slot = path->slots[0];
1869 nritems = btrfs_header_nritems(leaf);
1870
1871 if (btrfs_header_generation(leaf) > sk->max_transid) {
1872 i = nritems;
1873 goto advance_key;
1874 }
1875 found_transid = btrfs_header_generation(leaf);
1876
1877 for (i = slot; i < nritems; i++) {
1878 item_off = btrfs_item_ptr_offset(leaf, i);
1879 item_len = btrfs_item_size_nr(leaf, i);
1880
Gabriel de Perthuis03b71c62013-05-06 17:40:18 +00001881 btrfs_item_key_to_cpu(leaf, key, i);
1882 if (!key_in_sk(key, sk))
1883 continue;
1884
1885 if (sizeof(sh) + item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
Chris Masonac8e9812010-02-28 15:39:26 -05001886 item_len = 0;
1887
1888 if (sizeof(sh) + item_len + *sk_offset >
1889 BTRFS_SEARCH_ARGS_BUFSIZE) {
1890 ret = 1;
1891 goto overflow;
1892 }
1893
Chris Masonac8e9812010-02-28 15:39:26 -05001894 sh.objectid = key->objectid;
1895 sh.offset = key->offset;
1896 sh.type = key->type;
1897 sh.len = item_len;
1898 sh.transid = found_transid;
1899
1900 /* copy search result header */
1901 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1902 *sk_offset += sizeof(sh);
1903
1904 if (item_len) {
1905 char *p = buf + *sk_offset;
1906 /* copy the item */
1907 read_extent_buffer(leaf, p,
1908 item_off, item_len);
1909 *sk_offset += item_len;
Chris Masonac8e9812010-02-28 15:39:26 -05001910 }
Hugo Millse2156862011-05-14 17:43:41 +00001911 (*num_found)++;
Chris Masonac8e9812010-02-28 15:39:26 -05001912
1913 if (*num_found >= sk->nr_items)
1914 break;
1915 }
1916advance_key:
Chris Masonac8e9812010-02-28 15:39:26 -05001917 ret = 0;
Chris Masonabc6e132010-03-18 12:10:08 -04001918 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1919 key->offset++;
1920 else if (key->type < (u8)-1 && key->type < sk->max_type) {
1921 key->offset = 0;
1922 key->type++;
1923 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1924 key->offset = 0;
1925 key->type = 0;
1926 key->objectid++;
1927 } else
1928 ret = 1;
Chris Masonac8e9812010-02-28 15:39:26 -05001929overflow:
Chris Masonac8e9812010-02-28 15:39:26 -05001930 return ret;
1931}
1932
1933static noinline int search_ioctl(struct inode *inode,
1934 struct btrfs_ioctl_search_args *args)
1935{
1936 struct btrfs_root *root;
1937 struct btrfs_key key;
Chris Masonac8e9812010-02-28 15:39:26 -05001938 struct btrfs_path *path;
1939 struct btrfs_ioctl_search_key *sk = &args->key;
1940 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1941 int ret;
1942 int num_found = 0;
1943 unsigned long sk_offset = 0;
1944
1945 path = btrfs_alloc_path();
1946 if (!path)
1947 return -ENOMEM;
1948
1949 if (sk->tree_id == 0) {
1950 /* search the root of the inode that was passed */
1951 root = BTRFS_I(inode)->root;
1952 } else {
1953 key.objectid = sk->tree_id;
1954 key.type = BTRFS_ROOT_ITEM_KEY;
1955 key.offset = (u64)-1;
1956 root = btrfs_read_fs_root_no_name(info, &key);
1957 if (IS_ERR(root)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05001958 printk(KERN_ERR "BTRFS: could not find root %llu\n",
Chris Masonac8e9812010-02-28 15:39:26 -05001959 sk->tree_id);
1960 btrfs_free_path(path);
1961 return -ENOENT;
1962 }
1963 }
1964
1965 key.objectid = sk->min_objectid;
1966 key.type = sk->min_type;
1967 key.offset = sk->min_offset;
1968
Chris Masonac8e9812010-02-28 15:39:26 -05001969 path->keep_locks = 1;
1970
Dulshani Gunawardhana67871252013-10-31 10:33:04 +05301971 while (1) {
Filipe David Borba Manana6174d3c2013-10-01 16:13:42 +01001972 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
Chris Masonac8e9812010-02-28 15:39:26 -05001973 if (ret != 0) {
1974 if (ret > 0)
1975 ret = 0;
1976 goto err;
1977 }
1978 ret = copy_to_sk(root, path, &key, sk, args->buf,
1979 &sk_offset, &num_found);
David Sterbab3b4aa72011-04-21 01:20:15 +02001980 btrfs_release_path(path);
Chris Masonac8e9812010-02-28 15:39:26 -05001981 if (ret || num_found >= sk->nr_items)
1982 break;
1983
1984 }
1985 ret = 0;
1986err:
1987 sk->nr_items = num_found;
1988 btrfs_free_path(path);
1989 return ret;
1990}
1991
1992static noinline int btrfs_ioctl_tree_search(struct file *file,
1993 void __user *argp)
1994{
1995 struct btrfs_ioctl_search_args *args;
1996 struct inode *inode;
1997 int ret;
1998
1999 if (!capable(CAP_SYS_ADMIN))
2000 return -EPERM;
2001
Julia Lawall2354d082010-10-29 15:14:18 -04002002 args = memdup_user(argp, sizeof(*args));
2003 if (IS_ERR(args))
2004 return PTR_ERR(args);
Chris Masonac8e9812010-02-28 15:39:26 -05002005
Al Viro496ad9a2013-01-23 17:07:38 -05002006 inode = file_inode(file);
Chris Masonac8e9812010-02-28 15:39:26 -05002007 ret = search_ioctl(inode, args);
2008 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2009 ret = -EFAULT;
2010 kfree(args);
2011 return ret;
2012}
2013
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002014/*
Chris Masonac8e9812010-02-28 15:39:26 -05002015 * Search INODE_REFs to identify path name of 'dirid' directory
2016 * in a 'tree_id' tree. and sets path name to 'name'.
2017 */
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002018static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2019 u64 tree_id, u64 dirid, char *name)
2020{
2021 struct btrfs_root *root;
2022 struct btrfs_key key;
Chris Masonac8e9812010-02-28 15:39:26 -05002023 char *ptr;
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002024 int ret = -1;
2025 int slot;
2026 int len;
2027 int total_len = 0;
2028 struct btrfs_inode_ref *iref;
2029 struct extent_buffer *l;
2030 struct btrfs_path *path;
2031
2032 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2033 name[0]='\0';
2034 return 0;
2035 }
2036
2037 path = btrfs_alloc_path();
2038 if (!path)
2039 return -ENOMEM;
2040
Chris Masonac8e9812010-02-28 15:39:26 -05002041 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002042
2043 key.objectid = tree_id;
2044 key.type = BTRFS_ROOT_ITEM_KEY;
2045 key.offset = (u64)-1;
2046 root = btrfs_read_fs_root_no_name(info, &key);
2047 if (IS_ERR(root)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002048 printk(KERN_ERR "BTRFS: could not find root %llu\n", tree_id);
Chris Mason8ad6fca2010-03-18 12:23:10 -04002049 ret = -ENOENT;
2050 goto out;
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002051 }
2052
2053 key.objectid = dirid;
2054 key.type = BTRFS_INODE_REF_KEY;
Chris Mason8ad6fca2010-03-18 12:23:10 -04002055 key.offset = (u64)-1;
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002056
Dulshani Gunawardhana67871252013-10-31 10:33:04 +05302057 while (1) {
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002058 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2059 if (ret < 0)
2060 goto out;
Filipe David Borba Manana18674c62013-08-14 03:00:21 +01002061 else if (ret > 0) {
2062 ret = btrfs_previous_item(root, path, dirid,
2063 BTRFS_INODE_REF_KEY);
2064 if (ret < 0)
2065 goto out;
2066 else if (ret > 0) {
2067 ret = -ENOENT;
2068 goto out;
2069 }
2070 }
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002071
2072 l = path->nodes[0];
2073 slot = path->slots[0];
2074 btrfs_item_key_to_cpu(l, &key, slot);
2075
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002076 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2077 len = btrfs_inode_ref_name_len(l, iref);
2078 ptr -= len + 1;
2079 total_len += len + 1;
Filipe David Borba Mananaa696cf32013-08-14 03:00:20 +01002080 if (ptr < name) {
2081 ret = -ENAMETOOLONG;
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002082 goto out;
Filipe David Borba Mananaa696cf32013-08-14 03:00:20 +01002083 }
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002084
2085 *(ptr + len) = '/';
Dulshani Gunawardhana67871252013-10-31 10:33:04 +05302086 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002087
2088 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2089 break;
2090
David Sterbab3b4aa72011-04-21 01:20:15 +02002091 btrfs_release_path(path);
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002092 key.objectid = key.offset;
Chris Mason8ad6fca2010-03-18 12:23:10 -04002093 key.offset = (u64)-1;
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002094 dirid = key.objectid;
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002095 }
Li Zefan77906a502011-07-14 03:16:00 +00002096 memmove(name, ptr, total_len);
Dulshani Gunawardhana67871252013-10-31 10:33:04 +05302097 name[total_len] = '\0';
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002098 ret = 0;
2099out:
2100 btrfs_free_path(path);
Chris Masonac8e9812010-02-28 15:39:26 -05002101 return ret;
2102}
2103
2104static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2105 void __user *argp)
2106{
2107 struct btrfs_ioctl_ino_lookup_args *args;
2108 struct inode *inode;
2109 int ret;
2110
2111 if (!capable(CAP_SYS_ADMIN))
2112 return -EPERM;
2113
Julia Lawall2354d082010-10-29 15:14:18 -04002114 args = memdup_user(argp, sizeof(*args));
2115 if (IS_ERR(args))
2116 return PTR_ERR(args);
Dan Carpenterc2b96922010-03-20 11:24:15 +00002117
Al Viro496ad9a2013-01-23 17:07:38 -05002118 inode = file_inode(file);
Chris Masonac8e9812010-02-28 15:39:26 -05002119
Chris Mason1b53ac42010-03-18 12:17:05 -04002120 if (args->treeid == 0)
2121 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2122
Chris Masonac8e9812010-02-28 15:39:26 -05002123 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2124 args->treeid, args->objectid,
2125 args->name);
2126
2127 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2128 ret = -EFAULT;
2129
2130 kfree(args);
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002131 return ret;
2132}
2133
Yan, Zheng76dda932009-09-21 16:00:26 -04002134static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2135 void __user *arg)
2136{
Al Viro54563d42013-09-01 15:57:51 -04002137 struct dentry *parent = file->f_path.dentry;
Yan, Zheng76dda932009-09-21 16:00:26 -04002138 struct dentry *dentry;
2139 struct inode *dir = parent->d_inode;
2140 struct inode *inode;
2141 struct btrfs_root *root = BTRFS_I(dir)->root;
2142 struct btrfs_root *dest = NULL;
2143 struct btrfs_ioctl_vol_args *vol_args;
2144 struct btrfs_trans_handle *trans;
Miao Xiec58aaad2013-02-28 10:05:36 +00002145 struct btrfs_block_rsv block_rsv;
2146 u64 qgroup_reserved;
Yan, Zheng76dda932009-09-21 16:00:26 -04002147 int namelen;
2148 int ret;
2149 int err = 0;
2150
Yan, Zheng76dda932009-09-21 16:00:26 -04002151 vol_args = memdup_user(arg, sizeof(*vol_args));
2152 if (IS_ERR(vol_args))
2153 return PTR_ERR(vol_args);
2154
2155 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2156 namelen = strlen(vol_args->name);
2157 if (strchr(vol_args->name, '/') ||
2158 strncmp(vol_args->name, "..", namelen) == 0) {
2159 err = -EINVAL;
2160 goto out;
2161 }
2162
Al Viroa561be72011-11-23 11:57:51 -05002163 err = mnt_want_write_file(file);
Yan, Zheng76dda932009-09-21 16:00:26 -04002164 if (err)
2165 goto out;
2166
David Sterba5c50c9b2013-03-22 18:12:51 +00002167 err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2168 if (err == -EINTR)
David Sterbae43f9982013-12-06 17:51:32 +01002169 goto out_drop_write;
Yan, Zheng76dda932009-09-21 16:00:26 -04002170 dentry = lookup_one_len(vol_args->name, parent, namelen);
2171 if (IS_ERR(dentry)) {
2172 err = PTR_ERR(dentry);
2173 goto out_unlock_dir;
2174 }
2175
2176 if (!dentry->d_inode) {
2177 err = -ENOENT;
2178 goto out_dput;
2179 }
2180
2181 inode = dentry->d_inode;
Sage Weil4260f7c2010-10-29 15:46:43 -04002182 dest = BTRFS_I(inode)->root;
Dulshani Gunawardhana67871252013-10-31 10:33:04 +05302183 if (!capable(CAP_SYS_ADMIN)) {
Sage Weil4260f7c2010-10-29 15:46:43 -04002184 /*
2185 * Regular user. Only allow this with a special mount
2186 * option, when the user has write+exec access to the
2187 * subvol root, and when rmdir(2) would have been
2188 * allowed.
2189 *
2190 * Note that this is _not_ check that the subvol is
2191 * empty or doesn't contain data that we wouldn't
2192 * otherwise be able to delete.
2193 *
2194 * Users who want to delete empty subvols should try
2195 * rmdir(2).
2196 */
2197 err = -EPERM;
2198 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2199 goto out_dput;
2200
2201 /*
2202 * Do not allow deletion if the parent dir is the same
2203 * as the dir to be deleted. That means the ioctl
2204 * must be called on the dentry referencing the root
2205 * of the subvol, not a random directory contained
2206 * within it.
2207 */
2208 err = -EINVAL;
2209 if (root == dest)
2210 goto out_dput;
2211
2212 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2213 if (err)
2214 goto out_dput;
Sage Weil4260f7c2010-10-29 15:46:43 -04002215 }
2216
Miao Xie5c39da52012-10-22 11:39:53 +00002217 /* check if subvolume may be deleted by a user */
2218 err = btrfs_may_delete(dir, dentry, 1);
2219 if (err)
2220 goto out_dput;
2221
Li Zefan33345d012011-04-20 10:31:50 +08002222 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
Yan, Zheng76dda932009-09-21 16:00:26 -04002223 err = -EINVAL;
2224 goto out_dput;
2225 }
2226
Yan, Zheng76dda932009-09-21 16:00:26 -04002227 mutex_lock(&inode->i_mutex);
2228 err = d_invalidate(dentry);
2229 if (err)
2230 goto out_unlock;
2231
2232 down_write(&root->fs_info->subvol_sem);
2233
2234 err = may_destroy_subvol(dest);
2235 if (err)
2236 goto out_up_write;
2237
Miao Xiec58aaad2013-02-28 10:05:36 +00002238 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2239 /*
2240 * One for dir inode, two for dir entries, two for root
2241 * ref/backref.
2242 */
2243 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
Jeff Mahoneyee3441b2013-07-09 16:37:21 -04002244 5, &qgroup_reserved, true);
Miao Xiec58aaad2013-02-28 10:05:36 +00002245 if (err)
2246 goto out_up_write;
2247
Yan, Zhenga22285a2010-05-16 10:48:46 -04002248 trans = btrfs_start_transaction(root, 0);
2249 if (IS_ERR(trans)) {
2250 err = PTR_ERR(trans);
Miao Xiec58aaad2013-02-28 10:05:36 +00002251 goto out_release;
Yan, Zhenga22285a2010-05-16 10:48:46 -04002252 }
Miao Xiec58aaad2013-02-28 10:05:36 +00002253 trans->block_rsv = &block_rsv;
2254 trans->bytes_reserved = block_rsv.size;
Yan, Zhenga22285a2010-05-16 10:48:46 -04002255
Yan, Zheng76dda932009-09-21 16:00:26 -04002256 ret = btrfs_unlink_subvol(trans, root, dir,
2257 dest->root_key.objectid,
2258 dentry->d_name.name,
2259 dentry->d_name.len);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002260 if (ret) {
2261 err = ret;
2262 btrfs_abort_transaction(trans, root, ret);
2263 goto out_end_trans;
2264 }
Yan, Zheng76dda932009-09-21 16:00:26 -04002265
2266 btrfs_record_root_in_trans(trans, dest);
2267
2268 memset(&dest->root_item.drop_progress, 0,
2269 sizeof(dest->root_item.drop_progress));
2270 dest->root_item.drop_level = 0;
2271 btrfs_set_root_refs(&dest->root_item, 0);
2272
Yan, Zhengd68fc572010-05-16 10:49:58 -04002273 if (!xchg(&dest->orphan_item_inserted, 1)) {
2274 ret = btrfs_insert_orphan_item(trans,
2275 root->fs_info->tree_root,
2276 dest->root_key.objectid);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002277 if (ret) {
2278 btrfs_abort_transaction(trans, root, ret);
2279 err = ret;
2280 goto out_end_trans;
2281 }
Yan, Zhengd68fc572010-05-16 10:49:58 -04002282 }
Stefan Behrensdd5f9612013-08-15 17:11:20 +02002283
2284 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2285 dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
2286 dest->root_key.objectid);
2287 if (ret && ret != -ENOENT) {
2288 btrfs_abort_transaction(trans, root, ret);
2289 err = ret;
2290 goto out_end_trans;
2291 }
2292 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2293 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2294 dest->root_item.received_uuid,
2295 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2296 dest->root_key.objectid);
2297 if (ret && ret != -ENOENT) {
2298 btrfs_abort_transaction(trans, root, ret);
2299 err = ret;
2300 goto out_end_trans;
2301 }
2302 }
2303
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002304out_end_trans:
Miao Xiec58aaad2013-02-28 10:05:36 +00002305 trans->block_rsv = NULL;
2306 trans->bytes_reserved = 0;
Sage Weil531cb132010-10-29 15:41:32 -04002307 ret = btrfs_end_transaction(trans, root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002308 if (ret && !err)
2309 err = ret;
Yan, Zheng76dda932009-09-21 16:00:26 -04002310 inode->i_flags |= S_DEAD;
Miao Xiec58aaad2013-02-28 10:05:36 +00002311out_release:
2312 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
Yan, Zheng76dda932009-09-21 16:00:26 -04002313out_up_write:
2314 up_write(&root->fs_info->subvol_sem);
2315out_unlock:
2316 mutex_unlock(&inode->i_mutex);
2317 if (!err) {
Yan, Zhengefefb142009-10-09 09:25:16 -04002318 shrink_dcache_sb(root->fs_info->sb);
Yan, Zheng76dda932009-09-21 16:00:26 -04002319 btrfs_invalidate_inodes(dest);
2320 d_delete(dentry);
Liu Bofa6ac872013-02-20 14:10:23 +00002321
2322 /* the last ref */
2323 if (dest->cache_inode) {
2324 iput(dest->cache_inode);
2325 dest->cache_inode = NULL;
2326 }
Yan, Zheng76dda932009-09-21 16:00:26 -04002327 }
2328out_dput:
2329 dput(dentry);
2330out_unlock_dir:
2331 mutex_unlock(&dir->i_mutex);
David Sterbae43f9982013-12-06 17:51:32 +01002332out_drop_write:
Al Viro2a79f172011-12-09 08:06:57 -05002333 mnt_drop_write_file(file);
Yan, Zheng76dda932009-09-21 16:00:26 -04002334out:
2335 kfree(vol_args);
2336 return err;
2337}
2338
Chris Mason1e701a32010-03-11 09:42:04 -05002339static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002340{
Al Viro496ad9a2013-01-23 17:07:38 -05002341 struct inode *inode = file_inode(file);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002342 struct btrfs_root *root = BTRFS_I(inode)->root;
Chris Mason1e701a32010-03-11 09:42:04 -05002343 struct btrfs_ioctl_defrag_range_args *range;
Yan Zhengc146afa2008-11-12 14:34:12 -05002344 int ret;
2345
Ilya Dryomov25122d12013-01-20 15:57:57 +02002346 ret = mnt_want_write_file(file);
2347 if (ret)
2348 return ret;
Li Zefanb83cc962010-12-20 16:04:08 +08002349
Ilya Dryomov25122d12013-01-20 15:57:57 +02002350 if (btrfs_root_readonly(root)) {
2351 ret = -EROFS;
2352 goto out;
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01002353 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002354
2355 switch (inode->i_mode & S_IFMT) {
2356 case S_IFDIR:
Chris Masone441d542009-01-05 16:57:23 -05002357 if (!capable(CAP_SYS_ADMIN)) {
2358 ret = -EPERM;
2359 goto out;
2360 }
Eric Sandeende78b512013-01-31 18:21:12 +00002361 ret = btrfs_defrag_root(root);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002362 if (ret)
2363 goto out;
Eric Sandeende78b512013-01-31 18:21:12 +00002364 ret = btrfs_defrag_root(root->fs_info->extent_root);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002365 break;
2366 case S_IFREG:
Chris Masone441d542009-01-05 16:57:23 -05002367 if (!(file->f_mode & FMODE_WRITE)) {
2368 ret = -EINVAL;
2369 goto out;
2370 }
Chris Mason1e701a32010-03-11 09:42:04 -05002371
2372 range = kzalloc(sizeof(*range), GFP_KERNEL);
2373 if (!range) {
2374 ret = -ENOMEM;
2375 goto out;
2376 }
2377
2378 if (argp) {
2379 if (copy_from_user(range, argp,
2380 sizeof(*range))) {
2381 ret = -EFAULT;
2382 kfree(range);
Dan Carpenter683be162010-03-20 11:24:48 +00002383 goto out;
Chris Mason1e701a32010-03-11 09:42:04 -05002384 }
2385 /* compression requires us to start the IO */
2386 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2387 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2388 range->extent_thresh = (u32)-1;
2389 }
2390 } else {
2391 /* the rest are all set to zero by kzalloc */
2392 range->len = (u64)-1;
2393 }
Al Viro496ad9a2013-01-23 17:07:38 -05002394 ret = btrfs_defrag_file(file_inode(file), file,
Chris Mason4cb53002011-05-24 15:35:30 -04002395 range, 0, 0);
2396 if (ret > 0)
2397 ret = 0;
Chris Mason1e701a32010-03-11 09:42:04 -05002398 kfree(range);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002399 break;
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002400 default:
2401 ret = -EINVAL;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002402 }
Chris Masone441d542009-01-05 16:57:23 -05002403out:
Ilya Dryomov25122d12013-01-20 15:57:57 +02002404 mnt_drop_write_file(file);
Chris Masone441d542009-01-05 16:57:23 -05002405 return ret;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002406}
2407
Christoph Hellwigb2950862008-12-02 09:54:17 -05002408static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002409{
2410 struct btrfs_ioctl_vol_args *vol_args;
2411 int ret;
2412
Chris Masone441d542009-01-05 16:57:23 -05002413 if (!capable(CAP_SYS_ADMIN))
2414 return -EPERM;
2415
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01002416 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2417 1)) {
Anand Jaine57138b2013-08-21 11:44:48 +08002418 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002419 }
2420
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01002421 mutex_lock(&root->fs_info->volume_mutex);
Li Zefandae7b662009-04-08 15:06:54 +08002422 vol_args = memdup_user(arg, sizeof(*vol_args));
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002423 if (IS_ERR(vol_args)) {
2424 ret = PTR_ERR(vol_args);
2425 goto out;
2426 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002427
Mark Fasheh5516e592008-07-24 12:20:14 -04002428 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002429 ret = btrfs_init_new_device(root, vol_args->name);
2430
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002431 kfree(vol_args);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002432out:
2433 mutex_unlock(&root->fs_info->volume_mutex);
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01002434 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002435 return ret;
2436}
2437
Miao Xieda249272012-11-26 08:44:50 +00002438static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002439{
Al Viro496ad9a2013-01-23 17:07:38 -05002440 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002441 struct btrfs_ioctl_vol_args *vol_args;
2442 int ret;
2443
Chris Masone441d542009-01-05 16:57:23 -05002444 if (!capable(CAP_SYS_ADMIN))
2445 return -EPERM;
2446
Miao Xieda249272012-11-26 08:44:50 +00002447 ret = mnt_want_write_file(file);
2448 if (ret)
2449 return ret;
Yan Zhengc146afa2008-11-12 14:34:12 -05002450
Li Zefandae7b662009-04-08 15:06:54 +08002451 vol_args = memdup_user(arg, sizeof(*vol_args));
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002452 if (IS_ERR(vol_args)) {
2453 ret = PTR_ERR(vol_args);
2454 goto out;
2455 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002456
Mark Fasheh5516e592008-07-24 12:20:14 -04002457 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002458
Anand Jain183860f2013-05-17 10:52:45 +00002459 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2460 1)) {
2461 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2462 goto out;
2463 }
2464
2465 mutex_lock(&root->fs_info->volume_mutex);
2466 ret = btrfs_rm_device(root, vol_args->name);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002467 mutex_unlock(&root->fs_info->volume_mutex);
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01002468 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
Anand Jain183860f2013-05-17 10:52:45 +00002469
2470out:
2471 kfree(vol_args);
Ilya Dryomov4ac20c72013-01-20 15:57:57 +02002472 mnt_drop_write_file(file);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002473 return ret;
2474}
2475
Jan Schmidt475f6382011-03-11 15:41:01 +01002476static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2477{
Li Zefan027ed2f2011-06-08 08:27:56 +00002478 struct btrfs_ioctl_fs_info_args *fi_args;
Jan Schmidt475f6382011-03-11 15:41:01 +01002479 struct btrfs_device *device;
2480 struct btrfs_device *next;
2481 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
Li Zefan027ed2f2011-06-08 08:27:56 +00002482 int ret = 0;
Jan Schmidt475f6382011-03-11 15:41:01 +01002483
2484 if (!capable(CAP_SYS_ADMIN))
2485 return -EPERM;
2486
Li Zefan027ed2f2011-06-08 08:27:56 +00002487 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2488 if (!fi_args)
2489 return -ENOMEM;
2490
Filipe David Borba Mananaf7171752013-08-12 20:56:58 +01002491 mutex_lock(&fs_devices->device_list_mutex);
Li Zefan027ed2f2011-06-08 08:27:56 +00002492 fi_args->num_devices = fs_devices->num_devices;
2493 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
Jan Schmidt475f6382011-03-11 15:41:01 +01002494
Jan Schmidt475f6382011-03-11 15:41:01 +01002495 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
Li Zefan027ed2f2011-06-08 08:27:56 +00002496 if (device->devid > fi_args->max_id)
2497 fi_args->max_id = device->devid;
Jan Schmidt475f6382011-03-11 15:41:01 +01002498 }
2499 mutex_unlock(&fs_devices->device_list_mutex);
2500
Li Zefan027ed2f2011-06-08 08:27:56 +00002501 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2502 ret = -EFAULT;
Jan Schmidt475f6382011-03-11 15:41:01 +01002503
Li Zefan027ed2f2011-06-08 08:27:56 +00002504 kfree(fi_args);
2505 return ret;
Jan Schmidt475f6382011-03-11 15:41:01 +01002506}
2507
2508static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2509{
2510 struct btrfs_ioctl_dev_info_args *di_args;
2511 struct btrfs_device *dev;
2512 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2513 int ret = 0;
2514 char *s_uuid = NULL;
Jan Schmidt475f6382011-03-11 15:41:01 +01002515
2516 if (!capable(CAP_SYS_ADMIN))
2517 return -EPERM;
2518
2519 di_args = memdup_user(arg, sizeof(*di_args));
2520 if (IS_ERR(di_args))
2521 return PTR_ERR(di_args);
2522
Stefan Behrensdd5f9612013-08-15 17:11:20 +02002523 if (!btrfs_is_empty_uuid(di_args->uuid))
Jan Schmidt475f6382011-03-11 15:41:01 +01002524 s_uuid = di_args->uuid;
2525
2526 mutex_lock(&fs_devices->device_list_mutex);
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01002527 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
Jan Schmidt475f6382011-03-11 15:41:01 +01002528
2529 if (!dev) {
2530 ret = -ENODEV;
2531 goto out;
2532 }
2533
2534 di_args->devid = dev->devid;
2535 di_args->bytes_used = dev->bytes_used;
2536 di_args->total_bytes = dev->total_bytes;
2537 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
Jim Meyeringa27202f2012-04-26 18:36:56 +02002538 if (dev->name) {
Josef Bacik606686e2012-06-04 14:03:51 -04002539 struct rcu_string *name;
2540
2541 rcu_read_lock();
2542 name = rcu_dereference(dev->name);
2543 strncpy(di_args->path, name->str, sizeof(di_args->path));
2544 rcu_read_unlock();
Jim Meyeringa27202f2012-04-26 18:36:56 +02002545 di_args->path[sizeof(di_args->path) - 1] = 0;
2546 } else {
Stefan Behrens99ba55a2012-03-19 16:17:22 +01002547 di_args->path[0] = '\0';
Jim Meyeringa27202f2012-04-26 18:36:56 +02002548 }
Jan Schmidt475f6382011-03-11 15:41:01 +01002549
2550out:
David Sterba55793c02013-04-26 15:20:23 +00002551 mutex_unlock(&fs_devices->device_list_mutex);
Jan Schmidt475f6382011-03-11 15:41:01 +01002552 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2553 ret = -EFAULT;
2554
2555 kfree(di_args);
2556 return ret;
2557}
2558
Mark Fasheh416161d2013-08-06 11:42:51 -07002559static struct page *extent_same_get_page(struct inode *inode, u64 off)
2560{
2561 struct page *page;
2562 pgoff_t index;
2563 struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
2564
2565 index = off >> PAGE_CACHE_SHIFT;
2566
2567 page = grab_cache_page(inode->i_mapping, index);
2568 if (!page)
2569 return NULL;
2570
2571 if (!PageUptodate(page)) {
2572 if (extent_read_full_page_nolock(tree, page, btrfs_get_extent,
2573 0))
2574 return NULL;
2575 lock_page(page);
2576 if (!PageUptodate(page)) {
2577 unlock_page(page);
2578 page_cache_release(page);
2579 return NULL;
2580 }
2581 }
2582 unlock_page(page);
2583
2584 return page;
2585}
2586
Mark Fasheh77fe20dc2013-08-06 11:42:48 -07002587static inline void lock_extent_range(struct inode *inode, u64 off, u64 len)
2588{
2589 /* do any pending delalloc/csum calc on src, one way or
2590 another, and lock file content */
2591 while (1) {
2592 struct btrfs_ordered_extent *ordered;
2593 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2594 ordered = btrfs_lookup_first_ordered_extent(inode,
2595 off + len - 1);
2596 if (!ordered &&
2597 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2598 off + len - 1, EXTENT_DELALLOC, 0, NULL))
2599 break;
2600 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2601 if (ordered)
2602 btrfs_put_ordered_extent(ordered);
2603 btrfs_wait_ordered_range(inode, off, len);
2604 }
2605}
2606
Mark Fasheh416161d2013-08-06 11:42:51 -07002607static void btrfs_double_unlock(struct inode *inode1, u64 loff1,
2608 struct inode *inode2, u64 loff2, u64 len)
2609{
2610 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2611 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2612
2613 mutex_unlock(&inode1->i_mutex);
2614 mutex_unlock(&inode2->i_mutex);
2615}
2616
2617static void btrfs_double_lock(struct inode *inode1, u64 loff1,
2618 struct inode *inode2, u64 loff2, u64 len)
2619{
2620 if (inode1 < inode2) {
2621 swap(inode1, inode2);
2622 swap(loff1, loff2);
2623 }
2624
2625 mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT);
2626 lock_extent_range(inode1, loff1, len);
2627 if (inode1 != inode2) {
2628 mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD);
2629 lock_extent_range(inode2, loff2, len);
2630 }
2631}
2632
2633static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
2634 u64 dst_loff, u64 len)
2635{
2636 int ret = 0;
2637 struct page *src_page, *dst_page;
2638 unsigned int cmp_len = PAGE_CACHE_SIZE;
2639 void *addr, *dst_addr;
2640
2641 while (len) {
2642 if (len < PAGE_CACHE_SIZE)
2643 cmp_len = len;
2644
2645 src_page = extent_same_get_page(src, loff);
2646 if (!src_page)
2647 return -EINVAL;
2648 dst_page = extent_same_get_page(dst, dst_loff);
2649 if (!dst_page) {
2650 page_cache_release(src_page);
2651 return -EINVAL;
2652 }
2653 addr = kmap_atomic(src_page);
2654 dst_addr = kmap_atomic(dst_page);
2655
2656 flush_dcache_page(src_page);
2657 flush_dcache_page(dst_page);
2658
2659 if (memcmp(addr, dst_addr, cmp_len))
2660 ret = BTRFS_SAME_DATA_DIFFERS;
2661
2662 kunmap_atomic(addr);
2663 kunmap_atomic(dst_addr);
2664 page_cache_release(src_page);
2665 page_cache_release(dst_page);
2666
2667 if (ret)
2668 break;
2669
2670 loff += cmp_len;
2671 dst_loff += cmp_len;
2672 len -= cmp_len;
2673 }
2674
2675 return ret;
2676}
2677
2678static int extent_same_check_offsets(struct inode *inode, u64 off, u64 len)
2679{
2680 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
2681
2682 if (off + len > inode->i_size || off + len < off)
2683 return -EINVAL;
2684 /* Check that we are block aligned - btrfs_clone() requires this */
2685 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
2686 return -EINVAL;
2687
2688 return 0;
2689}
2690
2691static int btrfs_extent_same(struct inode *src, u64 loff, u64 len,
2692 struct inode *dst, u64 dst_loff)
2693{
2694 int ret;
2695
2696 /*
2697 * btrfs_clone() can't handle extents in the same file
2698 * yet. Once that works, we can drop this check and replace it
2699 * with a check for the same inode, but overlapping extents.
2700 */
2701 if (src == dst)
2702 return -EINVAL;
2703
2704 btrfs_double_lock(src, loff, dst, dst_loff, len);
2705
2706 ret = extent_same_check_offsets(src, loff, len);
2707 if (ret)
2708 goto out_unlock;
2709
2710 ret = extent_same_check_offsets(dst, dst_loff, len);
2711 if (ret)
2712 goto out_unlock;
2713
2714 /* don't make the dst file partly checksummed */
2715 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2716 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
2717 ret = -EINVAL;
2718 goto out_unlock;
2719 }
2720
2721 ret = btrfs_cmp_data(src, loff, dst, dst_loff, len);
2722 if (ret == 0)
2723 ret = btrfs_clone(src, dst, loff, len, len, dst_loff);
2724
2725out_unlock:
2726 btrfs_double_unlock(src, loff, dst, dst_loff, len);
2727
2728 return ret;
2729}
2730
2731#define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024)
2732
2733static long btrfs_ioctl_file_extent_same(struct file *file,
2734 void __user *argp)
2735{
Mark Fashehcbf8b8c2013-09-17 15:43:54 -07002736 struct btrfs_ioctl_same_args tmp;
2737 struct btrfs_ioctl_same_args *same;
2738 struct btrfs_ioctl_same_extent_info *info;
Mark Fasheh416161d2013-08-06 11:42:51 -07002739 struct inode *src = file->f_dentry->d_inode;
2740 struct file *dst_file = NULL;
2741 struct inode *dst;
2742 u64 off;
2743 u64 len;
2744 int i;
2745 int ret;
Mark Fashehcbf8b8c2013-09-17 15:43:54 -07002746 unsigned long size;
Mark Fasheh416161d2013-08-06 11:42:51 -07002747 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
2748 bool is_admin = capable(CAP_SYS_ADMIN);
2749
2750 if (!(file->f_mode & FMODE_READ))
2751 return -EINVAL;
2752
2753 ret = mnt_want_write_file(file);
2754 if (ret)
2755 return ret;
2756
Mark Fashehcbf8b8c2013-09-17 15:43:54 -07002757 if (copy_from_user(&tmp,
Mark Fasheh416161d2013-08-06 11:42:51 -07002758 (struct btrfs_ioctl_same_args __user *)argp,
Mark Fashehcbf8b8c2013-09-17 15:43:54 -07002759 sizeof(tmp))) {
Mark Fasheh416161d2013-08-06 11:42:51 -07002760 ret = -EFAULT;
2761 goto out;
2762 }
2763
Mark Fashehcbf8b8c2013-09-17 15:43:54 -07002764 size = sizeof(tmp) +
2765 tmp.dest_count * sizeof(struct btrfs_ioctl_same_extent_info);
2766
Geyslan G. Bem229eed42013-10-14 12:18:25 -03002767 same = memdup_user((struct btrfs_ioctl_same_args __user *)argp, size);
Mark Fashehcbf8b8c2013-09-17 15:43:54 -07002768
Geyslan G. Bem229eed42013-10-14 12:18:25 -03002769 if (IS_ERR(same)) {
2770 ret = PTR_ERR(same);
Mark Fashehcbf8b8c2013-09-17 15:43:54 -07002771 goto out;
2772 }
2773
2774 off = same->logical_offset;
2775 len = same->length;
Mark Fasheh416161d2013-08-06 11:42:51 -07002776
2777 /*
2778 * Limit the total length we will dedupe for each operation.
2779 * This is intended to bound the total time spent in this
2780 * ioctl to something sane.
2781 */
2782 if (len > BTRFS_MAX_DEDUPE_LEN)
2783 len = BTRFS_MAX_DEDUPE_LEN;
2784
2785 if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) {
2786 /*
2787 * Btrfs does not support blocksize < page_size. As a
2788 * result, btrfs_cmp_data() won't correctly handle
2789 * this situation without an update.
2790 */
2791 ret = -EINVAL;
2792 goto out;
2793 }
2794
2795 ret = -EISDIR;
2796 if (S_ISDIR(src->i_mode))
2797 goto out;
2798
2799 ret = -EACCES;
2800 if (!S_ISREG(src->i_mode))
2801 goto out;
2802
Mark Fashehcbf8b8c2013-09-17 15:43:54 -07002803 /* pre-format output fields to sane values */
2804 for (i = 0; i < same->dest_count; i++) {
2805 same->info[i].bytes_deduped = 0ULL;
2806 same->info[i].status = 0;
2807 }
2808
Mark Fasheh416161d2013-08-06 11:42:51 -07002809 ret = 0;
Mark Fashehcbf8b8c2013-09-17 15:43:54 -07002810 for (i = 0; i < same->dest_count; i++) {
2811 info = &same->info[i];
Mark Fasheh416161d2013-08-06 11:42:51 -07002812
Mark Fashehcbf8b8c2013-09-17 15:43:54 -07002813 dst_file = fget(info->fd);
Mark Fasheh416161d2013-08-06 11:42:51 -07002814 if (!dst_file) {
Mark Fashehcbf8b8c2013-09-17 15:43:54 -07002815 info->status = -EBADF;
Mark Fasheh416161d2013-08-06 11:42:51 -07002816 goto next;
2817 }
2818
2819 if (!(is_admin || (dst_file->f_mode & FMODE_WRITE))) {
Mark Fashehcbf8b8c2013-09-17 15:43:54 -07002820 info->status = -EINVAL;
Mark Fasheh416161d2013-08-06 11:42:51 -07002821 goto next;
2822 }
2823
Mark Fashehcbf8b8c2013-09-17 15:43:54 -07002824 info->status = -EXDEV;
Mark Fasheh416161d2013-08-06 11:42:51 -07002825 if (file->f_path.mnt != dst_file->f_path.mnt)
2826 goto next;
2827
2828 dst = dst_file->f_dentry->d_inode;
2829 if (src->i_sb != dst->i_sb)
2830 goto next;
2831
2832 if (S_ISDIR(dst->i_mode)) {
Mark Fashehcbf8b8c2013-09-17 15:43:54 -07002833 info->status = -EISDIR;
Mark Fasheh416161d2013-08-06 11:42:51 -07002834 goto next;
2835 }
2836
2837 if (!S_ISREG(dst->i_mode)) {
Mark Fashehcbf8b8c2013-09-17 15:43:54 -07002838 info->status = -EACCES;
Mark Fasheh416161d2013-08-06 11:42:51 -07002839 goto next;
2840 }
2841
Mark Fashehcbf8b8c2013-09-17 15:43:54 -07002842 info->status = btrfs_extent_same(src, off, len, dst,
2843 info->logical_offset);
2844 if (info->status == 0)
2845 info->bytes_deduped += len;
Mark Fasheh416161d2013-08-06 11:42:51 -07002846
2847next:
2848 if (dst_file)
2849 fput(dst_file);
Mark Fasheh416161d2013-08-06 11:42:51 -07002850 }
2851
Mark Fashehcbf8b8c2013-09-17 15:43:54 -07002852 ret = copy_to_user(argp, same, size);
2853 if (ret)
2854 ret = -EFAULT;
2855
Mark Fasheh416161d2013-08-06 11:42:51 -07002856out:
2857 mnt_drop_write_file(file);
2858 return ret;
2859}
2860
Mark Fasheh32b7c682013-08-06 11:42:49 -07002861/**
2862 * btrfs_clone() - clone a range from inode file to another
2863 *
2864 * @src: Inode to clone from
2865 * @inode: Inode to clone to
2866 * @off: Offset within source to start clone from
2867 * @olen: Original length, passed by user, of range to clone
2868 * @olen_aligned: Block-aligned value of olen, extent_same uses
2869 * identical values here
2870 * @destoff: Offset within @inode to start clone
2871 */
2872static int btrfs_clone(struct inode *src, struct inode *inode,
2873 u64 off, u64 olen, u64 olen_aligned, u64 destoff)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002874{
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002875 struct btrfs_root *root = BTRFS_I(inode)->root;
Mark Fasheh32b7c682013-08-06 11:42:49 -07002876 struct btrfs_path *path = NULL;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002877 struct extent_buffer *leaf;
Mark Fasheh32b7c682013-08-06 11:42:49 -07002878 struct btrfs_trans_handle *trans;
2879 char *buf = NULL;
Yan Zhengae01a0a2008-08-04 23:23:47 -04002880 struct btrfs_key key;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002881 u32 nritems;
2882 int slot;
Yan Zhengae01a0a2008-08-04 23:23:47 -04002883 int ret;
Mark Fasheh32b7c682013-08-06 11:42:49 -07002884 u64 len = olen_aligned;
Yan Zhengae01a0a2008-08-04 23:23:47 -04002885
2886 ret = -ENOMEM;
2887 buf = vmalloc(btrfs_level_size(root, 0));
2888 if (!buf)
Mark Fasheh32b7c682013-08-06 11:42:49 -07002889 return ret;
Yan Zhengae01a0a2008-08-04 23:23:47 -04002890
2891 path = btrfs_alloc_path();
2892 if (!path) {
2893 vfree(buf);
Mark Fasheh32b7c682013-08-06 11:42:49 -07002894 return ret;
Yan Zhengae01a0a2008-08-04 23:23:47 -04002895 }
Mark Fasheh32b7c682013-08-06 11:42:49 -07002896
Yan Zhengae01a0a2008-08-04 23:23:47 -04002897 path->reada = 2;
Sage Weilc5c9cd42008-11-12 14:32:25 -05002898 /* clone data */
Li Zefan33345d012011-04-20 10:31:50 +08002899 key.objectid = btrfs_ino(src);
Yan Zhengae01a0a2008-08-04 23:23:47 -04002900 key.type = BTRFS_EXTENT_DATA_KEY;
2901 key.offset = 0;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002902
2903 while (1) {
2904 /*
2905 * note the key will change type as we walk through the
2906 * tree.
2907 */
Filipe David Borba Mananae4355f32014-01-13 19:35:01 +00002908 path->leave_spinning = 1;
David Sterba362a20c2011-08-01 18:11:57 +02002909 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
2910 0, 0);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002911 if (ret < 0)
2912 goto out;
2913
Yan Zhengae01a0a2008-08-04 23:23:47 -04002914 nritems = btrfs_header_nritems(path->nodes[0]);
Filipe David Borba Mananae4355f32014-01-13 19:35:01 +00002915process_slot:
Yan Zhengae01a0a2008-08-04 23:23:47 -04002916 if (path->slots[0] >= nritems) {
David Sterba362a20c2011-08-01 18:11:57 +02002917 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002918 if (ret < 0)
2919 goto out;
2920 if (ret > 0)
2921 break;
Yan Zhengae01a0a2008-08-04 23:23:47 -04002922 nritems = btrfs_header_nritems(path->nodes[0]);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002923 }
2924 leaf = path->nodes[0];
2925 slot = path->slots[0];
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002926
Yan Zhengae01a0a2008-08-04 23:23:47 -04002927 btrfs_item_key_to_cpu(leaf, &key, slot);
Chris Masond20f7042008-12-08 16:58:54 -05002928 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
Li Zefan33345d012011-04-20 10:31:50 +08002929 key.objectid != btrfs_ino(src))
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002930 break;
2931
Sage Weilc5c9cd42008-11-12 14:32:25 -05002932 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2933 struct btrfs_file_extent_item *extent;
2934 int type;
Zheng Yan31840ae2008-09-23 13:14:14 -04002935 u32 size;
2936 struct btrfs_key new_key;
Sage Weilc5c9cd42008-11-12 14:32:25 -05002937 u64 disko = 0, diskl = 0;
2938 u64 datao = 0, datal = 0;
2939 u8 comp;
Sage Weilb5384d42010-06-12 22:31:14 +00002940 u64 endoff;
Zheng Yan31840ae2008-09-23 13:14:14 -04002941
Sage Weilc5c9cd42008-11-12 14:32:25 -05002942 extent = btrfs_item_ptr(leaf, slot,
2943 struct btrfs_file_extent_item);
2944 comp = btrfs_file_extent_compression(leaf, extent);
2945 type = btrfs_file_extent_type(leaf, extent);
Chris Masonc8a894d2009-06-27 21:07:03 -04002946 if (type == BTRFS_FILE_EXTENT_REG ||
2947 type == BTRFS_FILE_EXTENT_PREALLOC) {
Chris Masond3977122009-01-05 21:25:51 -05002948 disko = btrfs_file_extent_disk_bytenr(leaf,
2949 extent);
2950 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2951 extent);
Sage Weilc5c9cd42008-11-12 14:32:25 -05002952 datao = btrfs_file_extent_offset(leaf, extent);
Chris Masond3977122009-01-05 21:25:51 -05002953 datal = btrfs_file_extent_num_bytes(leaf,
2954 extent);
Sage Weilc5c9cd42008-11-12 14:32:25 -05002955 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2956 /* take upper bound, may be compressed */
2957 datal = btrfs_file_extent_ram_bytes(leaf,
2958 extent);
2959 }
Zheng Yan31840ae2008-09-23 13:14:14 -04002960
Sage Weil050006a2010-10-29 15:37:33 -04002961 if (key.offset + datal <= off ||
Filipe David Borba Mananae4355f32014-01-13 19:35:01 +00002962 key.offset >= off + len - 1) {
2963 path->slots[0]++;
2964 goto process_slot;
2965 }
2966
2967 size = btrfs_item_size_nr(leaf, slot);
2968 read_extent_buffer(leaf, buf,
2969 btrfs_item_ptr_offset(leaf, slot),
2970 size);
2971
2972 btrfs_release_path(path);
2973 path->leave_spinning = 0;
Sage Weilc5c9cd42008-11-12 14:32:25 -05002974
Zheng Yan31840ae2008-09-23 13:14:14 -04002975 memcpy(&new_key, &key, sizeof(new_key));
Li Zefan33345d012011-04-20 10:31:50 +08002976 new_key.objectid = btrfs_ino(inode);
Li Zefan4d728ec2011-01-26 14:10:43 +08002977 if (off <= key.offset)
2978 new_key.offset = key.offset + destoff - off;
2979 else
2980 new_key.offset = destoff;
Sage Weilc5c9cd42008-11-12 14:32:25 -05002981
Sage Weilb6f34092011-09-20 14:48:51 -04002982 /*
2983 * 1 - adjusting old extent (we may have to split it)
2984 * 1 - add new extent
2985 * 1 - inode update
2986 */
2987 trans = btrfs_start_transaction(root, 3);
Yan, Zhenga22285a2010-05-16 10:48:46 -04002988 if (IS_ERR(trans)) {
2989 ret = PTR_ERR(trans);
2990 goto out;
2991 }
2992
Chris Masonc8a894d2009-06-27 21:07:03 -04002993 if (type == BTRFS_FILE_EXTENT_REG ||
2994 type == BTRFS_FILE_EXTENT_PREALLOC) {
Li Zefand72c0842011-09-11 10:52:25 -04002995 /*
2996 * a | --- range to clone ---| b
2997 * | ------------- extent ------------- |
2998 */
2999
3000 /* substract range b */
3001 if (key.offset + datal > off + len)
3002 datal = off + len - key.offset;
3003
3004 /* substract range a */
Yan, Zhenga22285a2010-05-16 10:48:46 -04003005 if (off > key.offset) {
3006 datao += off - key.offset;
3007 datal -= off - key.offset;
3008 }
3009
Josef Bacik5dc562c2012-08-17 13:14:17 -04003010 ret = btrfs_drop_extents(trans, root, inode,
Yan, Zhenga22285a2010-05-16 10:48:46 -04003011 new_key.offset,
3012 new_key.offset + datal,
Josef Bacik26714852012-08-29 12:24:27 -04003013 1);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003014 if (ret) {
3015 btrfs_abort_transaction(trans, root,
3016 ret);
3017 btrfs_end_transaction(trans, root);
3018 goto out;
3019 }
Yan, Zhenga22285a2010-05-16 10:48:46 -04003020
Sage Weilc5c9cd42008-11-12 14:32:25 -05003021 ret = btrfs_insert_empty_item(trans, root, path,
3022 &new_key, size);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003023 if (ret) {
3024 btrfs_abort_transaction(trans, root,
3025 ret);
3026 btrfs_end_transaction(trans, root);
3027 goto out;
3028 }
Sage Weilc5c9cd42008-11-12 14:32:25 -05003029
3030 leaf = path->nodes[0];
3031 slot = path->slots[0];
3032 write_extent_buffer(leaf, buf,
3033 btrfs_item_ptr_offset(leaf, slot),
3034 size);
3035
3036 extent = btrfs_item_ptr(leaf, slot,
3037 struct btrfs_file_extent_item);
Sage Weilc5c9cd42008-11-12 14:32:25 -05003038
Sage Weilc5c9cd42008-11-12 14:32:25 -05003039 /* disko == 0 means it's a hole */
3040 if (!disko)
3041 datao = 0;
Sage Weilc5c9cd42008-11-12 14:32:25 -05003042
3043 btrfs_set_file_extent_offset(leaf, extent,
3044 datao);
3045 btrfs_set_file_extent_num_bytes(leaf, extent,
3046 datal);
3047 if (disko) {
3048 inode_add_bytes(inode, datal);
3049 ret = btrfs_inc_extent_ref(trans, root,
Yan Zheng5d4f98a2009-06-10 10:45:14 -04003050 disko, diskl, 0,
3051 root->root_key.objectid,
Li Zefan33345d012011-04-20 10:31:50 +08003052 btrfs_ino(inode),
Arne Jansen66d7e7f2011-09-12 15:26:38 +02003053 new_key.offset - datao,
3054 0);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003055 if (ret) {
3056 btrfs_abort_transaction(trans,
3057 root,
3058 ret);
3059 btrfs_end_transaction(trans,
3060 root);
3061 goto out;
3062
3063 }
Sage Weilc5c9cd42008-11-12 14:32:25 -05003064 }
3065 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3066 u64 skip = 0;
3067 u64 trim = 0;
3068 if (off > key.offset) {
3069 skip = off - key.offset;
3070 new_key.offset += skip;
3071 }
Chris Masond3977122009-01-05 21:25:51 -05003072
Liu Boaa42ffd2012-09-18 03:52:23 -06003073 if (key.offset + datal > off + len)
3074 trim = key.offset + datal - (off + len);
Chris Masond3977122009-01-05 21:25:51 -05003075
Sage Weilc5c9cd42008-11-12 14:32:25 -05003076 if (comp && (skip || trim)) {
Sage Weilc5c9cd42008-11-12 14:32:25 -05003077 ret = -EINVAL;
Yan, Zhenga22285a2010-05-16 10:48:46 -04003078 btrfs_end_transaction(trans, root);
Sage Weilc5c9cd42008-11-12 14:32:25 -05003079 goto out;
3080 }
3081 size -= skip + trim;
3082 datal -= skip + trim;
Yan, Zhenga22285a2010-05-16 10:48:46 -04003083
Josef Bacik5dc562c2012-08-17 13:14:17 -04003084 ret = btrfs_drop_extents(trans, root, inode,
Yan, Zhenga22285a2010-05-16 10:48:46 -04003085 new_key.offset,
3086 new_key.offset + datal,
Josef Bacik26714852012-08-29 12:24:27 -04003087 1);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003088 if (ret) {
3089 btrfs_abort_transaction(trans, root,
3090 ret);
3091 btrfs_end_transaction(trans, root);
3092 goto out;
3093 }
Yan, Zhenga22285a2010-05-16 10:48:46 -04003094
Sage Weilc5c9cd42008-11-12 14:32:25 -05003095 ret = btrfs_insert_empty_item(trans, root, path,
3096 &new_key, size);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003097 if (ret) {
3098 btrfs_abort_transaction(trans, root,
3099 ret);
3100 btrfs_end_transaction(trans, root);
3101 goto out;
3102 }
Sage Weilc5c9cd42008-11-12 14:32:25 -05003103
3104 if (skip) {
Chris Masond3977122009-01-05 21:25:51 -05003105 u32 start =
3106 btrfs_file_extent_calc_inline_size(0);
Sage Weilc5c9cd42008-11-12 14:32:25 -05003107 memmove(buf+start, buf+start+skip,
3108 datal);
3109 }
3110
3111 leaf = path->nodes[0];
3112 slot = path->slots[0];
3113 write_extent_buffer(leaf, buf,
3114 btrfs_item_ptr_offset(leaf, slot),
3115 size);
3116 inode_add_bytes(inode, datal);
3117 }
3118
3119 btrfs_mark_buffer_dirty(leaf);
David Sterbab3b4aa72011-04-21 01:20:15 +02003120 btrfs_release_path(path);
Sage Weilc5c9cd42008-11-12 14:32:25 -05003121
Josef Bacik0c4d2d92012-04-05 15:03:02 -04003122 inode_inc_iversion(inode);
Yan, Zhenga22285a2010-05-16 10:48:46 -04003123 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
Sage Weilb5384d42010-06-12 22:31:14 +00003124
3125 /*
3126 * we round up to the block size at eof when
3127 * determining which extents to clone above,
3128 * but shouldn't round up the file size
3129 */
3130 endoff = new_key.offset + datal;
Li Zefan5f3888f2010-11-19 01:36:34 +00003131 if (endoff > destoff+olen)
3132 endoff = destoff+olen;
Sage Weilb5384d42010-06-12 22:31:14 +00003133 if (endoff > inode->i_size)
3134 btrfs_i_size_write(inode, endoff);
3135
Yan, Zhenga22285a2010-05-16 10:48:46 -04003136 ret = btrfs_update_inode(trans, root, inode);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003137 if (ret) {
3138 btrfs_abort_transaction(trans, root, ret);
3139 btrfs_end_transaction(trans, root);
3140 goto out;
3141 }
3142 ret = btrfs_end_transaction(trans, root);
Yan, Zhenga22285a2010-05-16 10:48:46 -04003143 }
David Sterbab3b4aa72011-04-21 01:20:15 +02003144 btrfs_release_path(path);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003145 key.offset++;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003146 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003147 ret = 0;
Mark Fasheh32b7c682013-08-06 11:42:49 -07003148
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003149out:
David Sterbab3b4aa72011-04-21 01:20:15 +02003150 btrfs_release_path(path);
Mark Fasheh32b7c682013-08-06 11:42:49 -07003151 btrfs_free_path(path);
3152 vfree(buf);
3153 return ret;
3154}
3155
3156static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
3157 u64 off, u64 olen, u64 destoff)
3158{
Al Viro54563d42013-09-01 15:57:51 -04003159 struct inode *inode = file_inode(file);
Mark Fasheh32b7c682013-08-06 11:42:49 -07003160 struct btrfs_root *root = BTRFS_I(inode)->root;
3161 struct fd src_file;
3162 struct inode *src;
3163 int ret;
3164 u64 len = olen;
3165 u64 bs = root->fs_info->sb->s_blocksize;
3166 int same_inode = 0;
3167
3168 /*
3169 * TODO:
3170 * - split compressed inline extents. annoying: we need to
3171 * decompress into destination's address_space (the file offset
3172 * may change, so source mapping won't do), then recompress (or
3173 * otherwise reinsert) a subrange.
3174 * - allow ranges within the same file to be cloned (provided
3175 * they don't overlap)?
3176 */
3177
3178 /* the destination must be opened for writing */
3179 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
3180 return -EINVAL;
3181
3182 if (btrfs_root_readonly(root))
3183 return -EROFS;
3184
3185 ret = mnt_want_write_file(file);
3186 if (ret)
3187 return ret;
3188
3189 src_file = fdget(srcfd);
3190 if (!src_file.file) {
3191 ret = -EBADF;
3192 goto out_drop_write;
3193 }
3194
3195 ret = -EXDEV;
3196 if (src_file.file->f_path.mnt != file->f_path.mnt)
3197 goto out_fput;
3198
3199 src = file_inode(src_file.file);
3200
3201 ret = -EINVAL;
3202 if (src == inode)
3203 same_inode = 1;
3204
3205 /* the src must be open for reading */
3206 if (!(src_file.file->f_mode & FMODE_READ))
3207 goto out_fput;
3208
3209 /* don't make the dst file partly checksummed */
3210 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3211 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3212 goto out_fput;
3213
3214 ret = -EISDIR;
3215 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3216 goto out_fput;
3217
3218 ret = -EXDEV;
3219 if (src->i_sb != inode->i_sb)
3220 goto out_fput;
3221
3222 if (!same_inode) {
3223 if (inode < src) {
3224 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
3225 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
3226 } else {
3227 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
3228 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
3229 }
3230 } else {
3231 mutex_lock(&src->i_mutex);
3232 }
3233
3234 /* determine range to clone */
3235 ret = -EINVAL;
3236 if (off + len > src->i_size || off + len < off)
3237 goto out_unlock;
3238 if (len == 0)
3239 olen = len = src->i_size - off;
3240 /* if we extend to eof, continue to block boundary */
3241 if (off + len == src->i_size)
3242 len = ALIGN(src->i_size, bs) - off;
3243
3244 /* verify the end result is block aligned */
3245 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3246 !IS_ALIGNED(destoff, bs))
3247 goto out_unlock;
3248
3249 /* verify if ranges are overlapped within the same file */
3250 if (same_inode) {
3251 if (destoff + len > off && destoff < off + len)
3252 goto out_unlock;
3253 }
3254
3255 if (destoff > inode->i_size) {
3256 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3257 if (ret)
3258 goto out_unlock;
3259 }
3260
3261 /* truncate page cache pages from target inode range */
3262 truncate_inode_pages_range(&inode->i_data, destoff,
3263 PAGE_CACHE_ALIGN(destoff + len) - 1);
3264
3265 lock_extent_range(src, off, len);
3266
3267 ret = btrfs_clone(src, inode, off, olen, len, destoff);
3268
Liu Boaa42ffd2012-09-18 03:52:23 -06003269 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003270out_unlock:
Filipe David Borba Mananac57c2b32014-01-11 21:31:25 +00003271 if (!same_inode) {
3272 if (inode < src) {
3273 mutex_unlock(&src->i_mutex);
3274 mutex_unlock(&inode->i_mutex);
3275 } else {
3276 mutex_unlock(&inode->i_mutex);
3277 mutex_unlock(&src->i_mutex);
3278 }
3279 } else {
3280 mutex_unlock(&src->i_mutex);
3281 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003282out_fput:
Al Viro2903ff02012-08-28 12:52:22 -04003283 fdput(src_file);
Yan Zhengab67b7c2008-12-19 10:58:39 -05003284out_drop_write:
Al Viro2a79f172011-12-09 08:06:57 -05003285 mnt_drop_write_file(file);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003286 return ret;
3287}
3288
Christoph Hellwig7a865e82008-12-02 09:52:24 -05003289static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
Sage Weilc5c9cd42008-11-12 14:32:25 -05003290{
3291 struct btrfs_ioctl_clone_range_args args;
3292
Christoph Hellwig7a865e82008-12-02 09:52:24 -05003293 if (copy_from_user(&args, argp, sizeof(args)))
Sage Weilc5c9cd42008-11-12 14:32:25 -05003294 return -EFAULT;
3295 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
3296 args.src_length, args.dest_offset);
3297}
3298
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003299/*
3300 * there are many ways the trans_start and trans_end ioctls can lead
3301 * to deadlocks. They should only be used by applications that
3302 * basically own the machine, and have a very in depth understanding
3303 * of all the possible deadlocks and enospc problems.
3304 */
Christoph Hellwigb2950862008-12-02 09:54:17 -05003305static long btrfs_ioctl_trans_start(struct file *file)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003306{
Al Viro496ad9a2013-01-23 17:07:38 -05003307 struct inode *inode = file_inode(file);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003308 struct btrfs_root *root = BTRFS_I(inode)->root;
3309 struct btrfs_trans_handle *trans;
Sage Weil1ab86ae2009-09-29 18:38:44 -04003310 int ret;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003311
Sage Weil1ab86ae2009-09-29 18:38:44 -04003312 ret = -EPERM;
Christoph Hellwigdf5b5522008-06-11 21:53:58 -04003313 if (!capable(CAP_SYS_ADMIN))
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003314 goto out;
Sage Weil1ab86ae2009-09-29 18:38:44 -04003315
3316 ret = -EINPROGRESS;
3317 if (file->private_data)
3318 goto out;
Sage Weil9ca9ee02008-08-04 10:41:27 -04003319
Li Zefanb83cc962010-12-20 16:04:08 +08003320 ret = -EROFS;
3321 if (btrfs_root_readonly(root))
3322 goto out;
3323
Al Viroa561be72011-11-23 11:57:51 -05003324 ret = mnt_want_write_file(file);
Yan Zhengc146afa2008-11-12 14:34:12 -05003325 if (ret)
3326 goto out;
3327
Josef Bacika4abeea2011-04-11 17:25:13 -04003328 atomic_inc(&root->fs_info->open_ioctl_trans);
Sage Weil9ca9ee02008-08-04 10:41:27 -04003329
Sage Weil1ab86ae2009-09-29 18:38:44 -04003330 ret = -ENOMEM;
Josef Bacik7a7eaa42011-04-13 12:54:33 -04003331 trans = btrfs_start_ioctl_transaction(root);
Tsutomu Itohabd30bb2011-01-24 00:57:10 +00003332 if (IS_ERR(trans))
Sage Weil1ab86ae2009-09-29 18:38:44 -04003333 goto out_drop;
3334
3335 file->private_data = trans;
3336 return 0;
3337
3338out_drop:
Josef Bacika4abeea2011-04-11 17:25:13 -04003339 atomic_dec(&root->fs_info->open_ioctl_trans);
Al Viro2a79f172011-12-09 08:06:57 -05003340 mnt_drop_write_file(file);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003341out:
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003342 return ret;
3343}
3344
Josef Bacik6ef5ed02009-12-11 21:11:29 +00003345static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3346{
Al Viro496ad9a2013-01-23 17:07:38 -05003347 struct inode *inode = file_inode(file);
Josef Bacik6ef5ed02009-12-11 21:11:29 +00003348 struct btrfs_root *root = BTRFS_I(inode)->root;
3349 struct btrfs_root *new_root;
3350 struct btrfs_dir_item *di;
3351 struct btrfs_trans_handle *trans;
3352 struct btrfs_path *path;
3353 struct btrfs_key location;
3354 struct btrfs_disk_key disk_key;
Josef Bacik6ef5ed02009-12-11 21:11:29 +00003355 u64 objectid = 0;
3356 u64 dir_id;
Miao Xie3c04ce02012-11-26 08:43:07 +00003357 int ret;
Josef Bacik6ef5ed02009-12-11 21:11:29 +00003358
3359 if (!capable(CAP_SYS_ADMIN))
3360 return -EPERM;
3361
Miao Xie3c04ce02012-11-26 08:43:07 +00003362 ret = mnt_want_write_file(file);
3363 if (ret)
3364 return ret;
3365
3366 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3367 ret = -EFAULT;
3368 goto out;
3369 }
Josef Bacik6ef5ed02009-12-11 21:11:29 +00003370
3371 if (!objectid)
chandan1cecf572013-09-13 19:34:10 +05303372 objectid = BTRFS_FS_TREE_OBJECTID;
Josef Bacik6ef5ed02009-12-11 21:11:29 +00003373
3374 location.objectid = objectid;
3375 location.type = BTRFS_ROOT_ITEM_KEY;
3376 location.offset = (u64)-1;
3377
3378 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
Miao Xie3c04ce02012-11-26 08:43:07 +00003379 if (IS_ERR(new_root)) {
3380 ret = PTR_ERR(new_root);
3381 goto out;
3382 }
Josef Bacik6ef5ed02009-12-11 21:11:29 +00003383
Josef Bacik6ef5ed02009-12-11 21:11:29 +00003384 path = btrfs_alloc_path();
Miao Xie3c04ce02012-11-26 08:43:07 +00003385 if (!path) {
3386 ret = -ENOMEM;
3387 goto out;
3388 }
Josef Bacik6ef5ed02009-12-11 21:11:29 +00003389 path->leave_spinning = 1;
3390
3391 trans = btrfs_start_transaction(root, 1);
Tsutomu Itoh98d5dc12011-01-20 06:19:37 +00003392 if (IS_ERR(trans)) {
Josef Bacik6ef5ed02009-12-11 21:11:29 +00003393 btrfs_free_path(path);
Miao Xie3c04ce02012-11-26 08:43:07 +00003394 ret = PTR_ERR(trans);
3395 goto out;
Josef Bacik6ef5ed02009-12-11 21:11:29 +00003396 }
3397
David Sterba6c417612011-04-13 15:41:04 +02003398 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
Josef Bacik6ef5ed02009-12-11 21:11:29 +00003399 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
3400 dir_id, "default", 7, 1);
Dan Carpentercf1e99a2010-05-29 09:47:24 +00003401 if (IS_ERR_OR_NULL(di)) {
Josef Bacik6ef5ed02009-12-11 21:11:29 +00003402 btrfs_free_path(path);
3403 btrfs_end_transaction(trans, root);
Frank Holtonefe120a2013-12-20 11:37:06 -05003404 btrfs_err(new_root->fs_info, "Umm, you don't have the default dir"
3405 "item, this isn't going to work");
Miao Xie3c04ce02012-11-26 08:43:07 +00003406 ret = -ENOENT;
3407 goto out;
Josef Bacik6ef5ed02009-12-11 21:11:29 +00003408 }
3409
3410 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3411 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3412 btrfs_mark_buffer_dirty(path->nodes[0]);
3413 btrfs_free_path(path);
3414
Mitch Harder2b0ce2c2012-07-24 11:58:43 -06003415 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
Josef Bacik6ef5ed02009-12-11 21:11:29 +00003416 btrfs_end_transaction(trans, root);
Miao Xie3c04ce02012-11-26 08:43:07 +00003417out:
3418 mnt_drop_write_file(file);
3419 return ret;
Josef Bacik6ef5ed02009-12-11 21:11:29 +00003420}
3421
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003422void btrfs_get_block_group_info(struct list_head *groups_list,
3423 struct btrfs_ioctl_space_info *space)
Josef Bacikbf5fc0932010-09-29 11:22:36 -04003424{
3425 struct btrfs_block_group_cache *block_group;
3426
3427 space->total_bytes = 0;
3428 space->used_bytes = 0;
3429 space->flags = 0;
3430 list_for_each_entry(block_group, groups_list, list) {
3431 space->flags = block_group->flags;
3432 space->total_bytes += block_group->key.offset;
3433 space->used_bytes +=
3434 btrfs_block_group_used(&block_group->item);
3435 }
3436}
3437
Eric Sandeen48a3b632013-04-25 20:41:01 +00003438static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
Josef Bacik1406e432010-01-13 18:19:06 +00003439{
3440 struct btrfs_ioctl_space_args space_args;
3441 struct btrfs_ioctl_space_info space;
3442 struct btrfs_ioctl_space_info *dest;
Chris Mason7fde62b2010-03-16 15:40:10 -04003443 struct btrfs_ioctl_space_info *dest_orig;
Daniel J Blueman13f26962011-04-11 15:56:31 +00003444 struct btrfs_ioctl_space_info __user *user_dest;
Josef Bacik1406e432010-01-13 18:19:06 +00003445 struct btrfs_space_info *info;
Josef Bacikbf5fc0932010-09-29 11:22:36 -04003446 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3447 BTRFS_BLOCK_GROUP_SYSTEM,
3448 BTRFS_BLOCK_GROUP_METADATA,
3449 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3450 int num_types = 4;
Chris Mason7fde62b2010-03-16 15:40:10 -04003451 int alloc_size;
Josef Bacik1406e432010-01-13 18:19:06 +00003452 int ret = 0;
Dan Rosenberg51788b12011-02-14 16:04:23 -05003453 u64 slot_count = 0;
Josef Bacikbf5fc0932010-09-29 11:22:36 -04003454 int i, c;
Josef Bacik1406e432010-01-13 18:19:06 +00003455
3456 if (copy_from_user(&space_args,
3457 (struct btrfs_ioctl_space_args __user *)arg,
3458 sizeof(space_args)))
3459 return -EFAULT;
3460
Josef Bacikbf5fc0932010-09-29 11:22:36 -04003461 for (i = 0; i < num_types; i++) {
3462 struct btrfs_space_info *tmp;
3463
3464 info = NULL;
3465 rcu_read_lock();
3466 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3467 list) {
3468 if (tmp->flags == types[i]) {
3469 info = tmp;
3470 break;
3471 }
3472 }
3473 rcu_read_unlock();
3474
3475 if (!info)
3476 continue;
3477
3478 down_read(&info->groups_sem);
3479 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3480 if (!list_empty(&info->block_groups[c]))
3481 slot_count++;
3482 }
3483 up_read(&info->groups_sem);
3484 }
Josef Bacik1406e432010-01-13 18:19:06 +00003485
Chris Mason7fde62b2010-03-16 15:40:10 -04003486 /* space_slots == 0 means they are asking for a count */
3487 if (space_args.space_slots == 0) {
3488 space_args.total_spaces = slot_count;
3489 goto out;
3490 }
Josef Bacikbf5fc0932010-09-29 11:22:36 -04003491
Dan Rosenberg51788b12011-02-14 16:04:23 -05003492 slot_count = min_t(u64, space_args.space_slots, slot_count);
Josef Bacikbf5fc0932010-09-29 11:22:36 -04003493
Chris Mason7fde62b2010-03-16 15:40:10 -04003494 alloc_size = sizeof(*dest) * slot_count;
Josef Bacikbf5fc0932010-09-29 11:22:36 -04003495
Chris Mason7fde62b2010-03-16 15:40:10 -04003496 /* we generally have at most 6 or so space infos, one for each raid
3497 * level. So, a whole page should be more than enough for everyone
3498 */
3499 if (alloc_size > PAGE_CACHE_SIZE)
3500 return -ENOMEM;
3501
3502 space_args.total_spaces = 0;
3503 dest = kmalloc(alloc_size, GFP_NOFS);
3504 if (!dest)
3505 return -ENOMEM;
3506 dest_orig = dest;
3507
3508 /* now we have a buffer to copy into */
Josef Bacikbf5fc0932010-09-29 11:22:36 -04003509 for (i = 0; i < num_types; i++) {
3510 struct btrfs_space_info *tmp;
Chris Mason7fde62b2010-03-16 15:40:10 -04003511
Dan Rosenberg51788b12011-02-14 16:04:23 -05003512 if (!slot_count)
3513 break;
3514
Josef Bacikbf5fc0932010-09-29 11:22:36 -04003515 info = NULL;
3516 rcu_read_lock();
3517 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3518 list) {
3519 if (tmp->flags == types[i]) {
3520 info = tmp;
3521 break;
3522 }
3523 }
3524 rcu_read_unlock();
Chris Mason7fde62b2010-03-16 15:40:10 -04003525
Josef Bacikbf5fc0932010-09-29 11:22:36 -04003526 if (!info)
3527 continue;
3528 down_read(&info->groups_sem);
3529 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3530 if (!list_empty(&info->block_groups[c])) {
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003531 btrfs_get_block_group_info(
3532 &info->block_groups[c], &space);
Josef Bacikbf5fc0932010-09-29 11:22:36 -04003533 memcpy(dest, &space, sizeof(space));
3534 dest++;
3535 space_args.total_spaces++;
Dan Rosenberg51788b12011-02-14 16:04:23 -05003536 slot_count--;
Josef Bacikbf5fc0932010-09-29 11:22:36 -04003537 }
Dan Rosenberg51788b12011-02-14 16:04:23 -05003538 if (!slot_count)
3539 break;
Josef Bacikbf5fc0932010-09-29 11:22:36 -04003540 }
3541 up_read(&info->groups_sem);
Josef Bacik1406e432010-01-13 18:19:06 +00003542 }
Josef Bacik1406e432010-01-13 18:19:06 +00003543
Daniel J Blueman2eec6c82012-04-26 00:37:14 +08003544 user_dest = (struct btrfs_ioctl_space_info __user *)
Chris Mason7fde62b2010-03-16 15:40:10 -04003545 (arg + sizeof(struct btrfs_ioctl_space_args));
3546
3547 if (copy_to_user(user_dest, dest_orig, alloc_size))
3548 ret = -EFAULT;
3549
3550 kfree(dest_orig);
3551out:
3552 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
Josef Bacik1406e432010-01-13 18:19:06 +00003553 ret = -EFAULT;
3554
3555 return ret;
3556}
3557
Jeff Mahoney01e219e2013-11-01 13:07:03 -04003558static long btrfs_ioctl_global_rsv(struct btrfs_root *root, void __user *arg)
3559{
3560 struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv;
3561 u64 reserved;
3562
3563 spin_lock(&block_rsv->lock);
3564 reserved = block_rsv->reserved;
3565 spin_unlock(&block_rsv->lock);
3566
3567 if (arg && copy_to_user(arg, &reserved, sizeof(reserved)))
3568 return -EFAULT;
3569 return 0;
3570}
3571
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003572/*
3573 * there are many ways the trans_start and trans_end ioctls can lead
3574 * to deadlocks. They should only be used by applications that
3575 * basically own the machine, and have a very in depth understanding
3576 * of all the possible deadlocks and enospc problems.
3577 */
3578long btrfs_ioctl_trans_end(struct file *file)
3579{
Al Viro496ad9a2013-01-23 17:07:38 -05003580 struct inode *inode = file_inode(file);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003581 struct btrfs_root *root = BTRFS_I(inode)->root;
3582 struct btrfs_trans_handle *trans;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003583
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003584 trans = file->private_data;
Sage Weil1ab86ae2009-09-29 18:38:44 -04003585 if (!trans)
3586 return -EINVAL;
Christoph Hellwigb2141072008-09-05 16:43:31 -04003587 file->private_data = NULL;
Sage Weil9ca9ee02008-08-04 10:41:27 -04003588
Sage Weil1ab86ae2009-09-29 18:38:44 -04003589 btrfs_end_transaction(trans, root);
3590
Josef Bacika4abeea2011-04-11 17:25:13 -04003591 atomic_dec(&root->fs_info->open_ioctl_trans);
Sage Weil9ca9ee02008-08-04 10:41:27 -04003592
Al Viro2a79f172011-12-09 08:06:57 -05003593 mnt_drop_write_file(file);
Sage Weil1ab86ae2009-09-29 18:38:44 -04003594 return 0;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003595}
3596
Miao Xie9a8c28b2012-11-26 08:40:43 +00003597static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3598 void __user *argp)
Sage Weil46204592010-10-29 15:41:32 -04003599{
Sage Weil46204592010-10-29 15:41:32 -04003600 struct btrfs_trans_handle *trans;
3601 u64 transid;
Tsutomu Itohdb5b4932011-03-23 08:14:16 +00003602 int ret;
Sage Weil46204592010-10-29 15:41:32 -04003603
Miao Xied4edf392013-02-20 09:17:06 +00003604 trans = btrfs_attach_transaction_barrier(root);
Miao Xieff7c1d32012-11-26 08:41:29 +00003605 if (IS_ERR(trans)) {
3606 if (PTR_ERR(trans) != -ENOENT)
3607 return PTR_ERR(trans);
3608
3609 /* No running transaction, don't bother */
3610 transid = root->fs_info->last_trans_committed;
3611 goto out;
3612 }
Sage Weil46204592010-10-29 15:41:32 -04003613 transid = trans->transid;
Tsutomu Itohdb5b4932011-03-23 08:14:16 +00003614 ret = btrfs_commit_transaction_async(trans, root, 0);
Tsutomu Itoh8b2b2d32011-04-04 01:52:13 +00003615 if (ret) {
3616 btrfs_end_transaction(trans, root);
Tsutomu Itohdb5b4932011-03-23 08:14:16 +00003617 return ret;
Tsutomu Itoh8b2b2d32011-04-04 01:52:13 +00003618 }
Miao Xieff7c1d32012-11-26 08:41:29 +00003619out:
Sage Weil46204592010-10-29 15:41:32 -04003620 if (argp)
3621 if (copy_to_user(argp, &transid, sizeof(transid)))
3622 return -EFAULT;
3623 return 0;
3624}
3625
Miao Xie9a8c28b2012-11-26 08:40:43 +00003626static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
3627 void __user *argp)
Sage Weil46204592010-10-29 15:41:32 -04003628{
Sage Weil46204592010-10-29 15:41:32 -04003629 u64 transid;
3630
3631 if (argp) {
3632 if (copy_from_user(&transid, argp, sizeof(transid)))
3633 return -EFAULT;
3634 } else {
3635 transid = 0; /* current trans */
3636 }
3637 return btrfs_wait_for_commit(root, transid);
3638}
3639
Miao Xieb8e95482012-11-26 08:48:01 +00003640static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
Jan Schmidt475f6382011-03-11 15:41:01 +01003641{
Al Viro496ad9a2013-01-23 17:07:38 -05003642 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Jan Schmidt475f6382011-03-11 15:41:01 +01003643 struct btrfs_ioctl_scrub_args *sa;
Miao Xieb8e95482012-11-26 08:48:01 +00003644 int ret;
Jan Schmidt475f6382011-03-11 15:41:01 +01003645
3646 if (!capable(CAP_SYS_ADMIN))
3647 return -EPERM;
3648
3649 sa = memdup_user(arg, sizeof(*sa));
3650 if (IS_ERR(sa))
3651 return PTR_ERR(sa);
3652
Miao Xieb8e95482012-11-26 08:48:01 +00003653 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3654 ret = mnt_want_write_file(file);
3655 if (ret)
3656 goto out;
3657 }
3658
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01003659 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
Stefan Behrens63a212a2012-11-05 18:29:28 +01003660 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3661 0);
Jan Schmidt475f6382011-03-11 15:41:01 +01003662
3663 if (copy_to_user(arg, sa, sizeof(*sa)))
3664 ret = -EFAULT;
3665
Miao Xieb8e95482012-11-26 08:48:01 +00003666 if (!(sa->flags & BTRFS_SCRUB_READONLY))
3667 mnt_drop_write_file(file);
3668out:
Jan Schmidt475f6382011-03-11 15:41:01 +01003669 kfree(sa);
3670 return ret;
3671}
3672
3673static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3674{
3675 if (!capable(CAP_SYS_ADMIN))
3676 return -EPERM;
3677
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01003678 return btrfs_scrub_cancel(root->fs_info);
Jan Schmidt475f6382011-03-11 15:41:01 +01003679}
3680
3681static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3682 void __user *arg)
3683{
3684 struct btrfs_ioctl_scrub_args *sa;
3685 int ret;
3686
3687 if (!capable(CAP_SYS_ADMIN))
3688 return -EPERM;
3689
3690 sa = memdup_user(arg, sizeof(*sa));
3691 if (IS_ERR(sa))
3692 return PTR_ERR(sa);
3693
3694 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3695
3696 if (copy_to_user(arg, sa, sizeof(*sa)))
3697 ret = -EFAULT;
3698
3699 kfree(sa);
3700 return ret;
3701}
3702
Stefan Behrensc11d2c22012-05-25 16:06:09 +02003703static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
David Sterbab27f7c02012-06-22 06:30:39 -06003704 void __user *arg)
Stefan Behrensc11d2c22012-05-25 16:06:09 +02003705{
3706 struct btrfs_ioctl_get_dev_stats *sa;
3707 int ret;
3708
Stefan Behrensc11d2c22012-05-25 16:06:09 +02003709 sa = memdup_user(arg, sizeof(*sa));
3710 if (IS_ERR(sa))
3711 return PTR_ERR(sa);
3712
David Sterbab27f7c02012-06-22 06:30:39 -06003713 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3714 kfree(sa);
3715 return -EPERM;
3716 }
3717
3718 ret = btrfs_get_dev_stats(root, sa);
Stefan Behrensc11d2c22012-05-25 16:06:09 +02003719
3720 if (copy_to_user(arg, sa, sizeof(*sa)))
3721 ret = -EFAULT;
3722
3723 kfree(sa);
3724 return ret;
3725}
3726
Stefan Behrens3f6bcfb2012-11-06 15:08:53 +01003727static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
3728{
3729 struct btrfs_ioctl_dev_replace_args *p;
3730 int ret;
3731
3732 if (!capable(CAP_SYS_ADMIN))
3733 return -EPERM;
3734
3735 p = memdup_user(arg, sizeof(*p));
3736 if (IS_ERR(p))
3737 return PTR_ERR(p);
3738
3739 switch (p->cmd) {
3740 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
Ilya Dryomovadfa97c2013-10-10 20:39:28 +03003741 if (root->fs_info->sb->s_flags & MS_RDONLY) {
3742 ret = -EROFS;
3743 goto out;
3744 }
Stefan Behrens3f6bcfb2012-11-06 15:08:53 +01003745 if (atomic_xchg(
3746 &root->fs_info->mutually_exclusive_operation_running,
3747 1)) {
Anand Jaine57138b2013-08-21 11:44:48 +08003748 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
Stefan Behrens3f6bcfb2012-11-06 15:08:53 +01003749 } else {
3750 ret = btrfs_dev_replace_start(root, p);
3751 atomic_set(
3752 &root->fs_info->mutually_exclusive_operation_running,
3753 0);
3754 }
3755 break;
3756 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3757 btrfs_dev_replace_status(root->fs_info, p);
3758 ret = 0;
3759 break;
3760 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3761 ret = btrfs_dev_replace_cancel(root->fs_info, p);
3762 break;
3763 default:
3764 ret = -EINVAL;
3765 break;
3766 }
3767
3768 if (copy_to_user(arg, p, sizeof(*p)))
3769 ret = -EFAULT;
Ilya Dryomovadfa97c2013-10-10 20:39:28 +03003770out:
Stefan Behrens3f6bcfb2012-11-06 15:08:53 +01003771 kfree(p);
3772 return ret;
3773}
3774
Jan Schmidtd7728c92011-07-07 16:48:38 +02003775static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3776{
3777 int ret = 0;
3778 int i;
Chris Mason740c3d22011-11-02 15:48:34 -04003779 u64 rel_ptr;
Jan Schmidtd7728c92011-07-07 16:48:38 +02003780 int size;
Chris Mason806468f2011-11-06 03:07:10 -05003781 struct btrfs_ioctl_ino_path_args *ipa = NULL;
Jan Schmidtd7728c92011-07-07 16:48:38 +02003782 struct inode_fs_paths *ipath = NULL;
3783 struct btrfs_path *path;
3784
Kusanagi Kouichi82b22ac2013-01-28 11:33:31 +00003785 if (!capable(CAP_DAC_READ_SEARCH))
Jan Schmidtd7728c92011-07-07 16:48:38 +02003786 return -EPERM;
3787
3788 path = btrfs_alloc_path();
3789 if (!path) {
3790 ret = -ENOMEM;
3791 goto out;
3792 }
3793
3794 ipa = memdup_user(arg, sizeof(*ipa));
3795 if (IS_ERR(ipa)) {
3796 ret = PTR_ERR(ipa);
3797 ipa = NULL;
3798 goto out;
3799 }
3800
3801 size = min_t(u32, ipa->size, 4096);
3802 ipath = init_ipath(size, root, path);
3803 if (IS_ERR(ipath)) {
3804 ret = PTR_ERR(ipath);
3805 ipath = NULL;
3806 goto out;
3807 }
3808
3809 ret = paths_from_inode(ipa->inum, ipath);
3810 if (ret < 0)
3811 goto out;
3812
3813 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
Jeff Mahoney745c4d82011-11-20 07:31:57 -05003814 rel_ptr = ipath->fspath->val[i] -
3815 (u64)(unsigned long)ipath->fspath->val;
Chris Mason740c3d22011-11-02 15:48:34 -04003816 ipath->fspath->val[i] = rel_ptr;
Jan Schmidtd7728c92011-07-07 16:48:38 +02003817 }
3818
Jeff Mahoney745c4d82011-11-20 07:31:57 -05003819 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3820 (void *)(unsigned long)ipath->fspath, size);
Jan Schmidtd7728c92011-07-07 16:48:38 +02003821 if (ret) {
3822 ret = -EFAULT;
3823 goto out;
3824 }
3825
3826out:
3827 btrfs_free_path(path);
3828 free_ipath(ipath);
3829 kfree(ipa);
3830
3831 return ret;
3832}
3833
3834static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3835{
3836 struct btrfs_data_container *inodes = ctx;
3837 const size_t c = 3 * sizeof(u64);
3838
3839 if (inodes->bytes_left >= c) {
3840 inodes->bytes_left -= c;
3841 inodes->val[inodes->elem_cnt] = inum;
3842 inodes->val[inodes->elem_cnt + 1] = offset;
3843 inodes->val[inodes->elem_cnt + 2] = root;
3844 inodes->elem_cnt += 3;
3845 } else {
3846 inodes->bytes_missing += c - inodes->bytes_left;
3847 inodes->bytes_left = 0;
3848 inodes->elem_missed += 3;
3849 }
3850
3851 return 0;
3852}
3853
3854static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3855 void __user *arg)
3856{
3857 int ret = 0;
3858 int size;
Jan Schmidtd7728c92011-07-07 16:48:38 +02003859 struct btrfs_ioctl_logical_ino_args *loi;
3860 struct btrfs_data_container *inodes = NULL;
3861 struct btrfs_path *path = NULL;
Jan Schmidtd7728c92011-07-07 16:48:38 +02003862
3863 if (!capable(CAP_SYS_ADMIN))
3864 return -EPERM;
3865
3866 loi = memdup_user(arg, sizeof(*loi));
3867 if (IS_ERR(loi)) {
3868 ret = PTR_ERR(loi);
3869 loi = NULL;
3870 goto out;
3871 }
3872
3873 path = btrfs_alloc_path();
3874 if (!path) {
3875 ret = -ENOMEM;
3876 goto out;
3877 }
3878
Liu Bo425d17a2012-09-07 20:01:30 -06003879 size = min_t(u32, loi->size, 64 * 1024);
Jan Schmidtd7728c92011-07-07 16:48:38 +02003880 inodes = init_data_container(size);
3881 if (IS_ERR(inodes)) {
3882 ret = PTR_ERR(inodes);
3883 inodes = NULL;
3884 goto out;
3885 }
3886
Liu Bodf031f02012-09-07 20:01:29 -06003887 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
3888 build_ino_list, inodes);
3889 if (ret == -EINVAL)
Jan Schmidtd7728c92011-07-07 16:48:38 +02003890 ret = -ENOENT;
3891 if (ret < 0)
3892 goto out;
3893
Jeff Mahoney745c4d82011-11-20 07:31:57 -05003894 ret = copy_to_user((void *)(unsigned long)loi->inodes,
3895 (void *)(unsigned long)inodes, size);
Jan Schmidtd7728c92011-07-07 16:48:38 +02003896 if (ret)
3897 ret = -EFAULT;
3898
3899out:
3900 btrfs_free_path(path);
Liu Bo425d17a2012-09-07 20:01:30 -06003901 vfree(inodes);
Jan Schmidtd7728c92011-07-07 16:48:38 +02003902 kfree(loi);
3903
3904 return ret;
3905}
3906
Ilya Dryomov19a39dc2012-01-16 22:04:49 +02003907void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02003908 struct btrfs_ioctl_balance_args *bargs)
3909{
3910 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3911
3912 bargs->flags = bctl->flags;
3913
Ilya Dryomov837d5b62012-01-16 22:04:49 +02003914 if (atomic_read(&fs_info->balance_running))
3915 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3916 if (atomic_read(&fs_info->balance_pause_req))
3917 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
Ilya Dryomova7e99c62012-01-16 22:04:49 +02003918 if (atomic_read(&fs_info->balance_cancel_req))
3919 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
Ilya Dryomov837d5b62012-01-16 22:04:49 +02003920
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02003921 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3922 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3923 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
Ilya Dryomov19a39dc2012-01-16 22:04:49 +02003924
3925 if (lock) {
3926 spin_lock(&fs_info->balance_lock);
3927 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3928 spin_unlock(&fs_info->balance_lock);
3929 } else {
3930 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3931 }
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02003932}
3933
Liu Bo9ba1f6e2012-05-11 18:11:26 +08003934static long btrfs_ioctl_balance(struct file *file, void __user *arg)
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02003935{
Al Viro496ad9a2013-01-23 17:07:38 -05003936 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02003937 struct btrfs_fs_info *fs_info = root->fs_info;
3938 struct btrfs_ioctl_balance_args *bargs;
3939 struct btrfs_balance_control *bctl;
Ilya Dryomoved0fb782013-01-20 15:57:57 +02003940 bool need_unlock; /* for mut. excl. ops lock */
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02003941 int ret;
3942
3943 if (!capable(CAP_SYS_ADMIN))
3944 return -EPERM;
3945
Liu Boe54bfa32012-06-29 03:58:48 -06003946 ret = mnt_want_write_file(file);
Liu Bo9ba1f6e2012-05-11 18:11:26 +08003947 if (ret)
3948 return ret;
3949
Ilya Dryomoved0fb782013-01-20 15:57:57 +02003950again:
3951 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
3952 mutex_lock(&fs_info->volume_mutex);
3953 mutex_lock(&fs_info->balance_mutex);
3954 need_unlock = true;
3955 goto locked;
3956 }
3957
3958 /*
3959 * mut. excl. ops lock is locked. Three possibilites:
3960 * (1) some other op is running
3961 * (2) balance is running
3962 * (3) balance is paused -- special case (think resume)
3963 */
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02003964 mutex_lock(&fs_info->balance_mutex);
Ilya Dryomoved0fb782013-01-20 15:57:57 +02003965 if (fs_info->balance_ctl) {
3966 /* this is either (2) or (3) */
3967 if (!atomic_read(&fs_info->balance_running)) {
3968 mutex_unlock(&fs_info->balance_mutex);
3969 if (!mutex_trylock(&fs_info->volume_mutex))
3970 goto again;
3971 mutex_lock(&fs_info->balance_mutex);
3972
3973 if (fs_info->balance_ctl &&
3974 !atomic_read(&fs_info->balance_running)) {
3975 /* this is (3) */
3976 need_unlock = false;
3977 goto locked;
3978 }
3979
3980 mutex_unlock(&fs_info->balance_mutex);
3981 mutex_unlock(&fs_info->volume_mutex);
3982 goto again;
3983 } else {
3984 /* this is (2) */
3985 mutex_unlock(&fs_info->balance_mutex);
3986 ret = -EINPROGRESS;
3987 goto out;
3988 }
3989 } else {
3990 /* this is (1) */
3991 mutex_unlock(&fs_info->balance_mutex);
Anand Jaine57138b2013-08-21 11:44:48 +08003992 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
Ilya Dryomoved0fb782013-01-20 15:57:57 +02003993 goto out;
3994 }
3995
3996locked:
3997 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02003998
3999 if (arg) {
4000 bargs = memdup_user(arg, sizeof(*bargs));
4001 if (IS_ERR(bargs)) {
4002 ret = PTR_ERR(bargs);
Ilya Dryomoved0fb782013-01-20 15:57:57 +02004003 goto out_unlock;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004004 }
Ilya Dryomovde322262012-01-16 22:04:49 +02004005
4006 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4007 if (!fs_info->balance_ctl) {
4008 ret = -ENOTCONN;
4009 goto out_bargs;
4010 }
4011
4012 bctl = fs_info->balance_ctl;
4013 spin_lock(&fs_info->balance_lock);
4014 bctl->flags |= BTRFS_BALANCE_RESUME;
4015 spin_unlock(&fs_info->balance_lock);
4016
4017 goto do_balance;
4018 }
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004019 } else {
4020 bargs = NULL;
4021 }
4022
Ilya Dryomoved0fb782013-01-20 15:57:57 +02004023 if (fs_info->balance_ctl) {
Ilya Dryomov837d5b62012-01-16 22:04:49 +02004024 ret = -EINPROGRESS;
4025 goto out_bargs;
4026 }
4027
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004028 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
4029 if (!bctl) {
4030 ret = -ENOMEM;
4031 goto out_bargs;
4032 }
4033
4034 bctl->fs_info = fs_info;
4035 if (arg) {
4036 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4037 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4038 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4039
4040 bctl->flags = bargs->flags;
Ilya Dryomovf43ffb62012-01-16 22:04:47 +02004041 } else {
4042 /* balance everything - no filters */
4043 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004044 }
4045
Ilya Dryomovde322262012-01-16 22:04:49 +02004046do_balance:
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004047 /*
Ilya Dryomoved0fb782013-01-20 15:57:57 +02004048 * Ownership of bctl and mutually_exclusive_operation_running
4049 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4050 * or, if restriper was paused all the way until unmount, in
4051 * free_fs_info. mutually_exclusive_operation_running is
4052 * cleared in __cancel_balance.
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004053 */
Ilya Dryomoved0fb782013-01-20 15:57:57 +02004054 need_unlock = false;
4055
4056 ret = btrfs_balance(bctl, bargs);
4057
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004058 if (arg) {
4059 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4060 ret = -EFAULT;
4061 }
4062
4063out_bargs:
4064 kfree(bargs);
Ilya Dryomoved0fb782013-01-20 15:57:57 +02004065out_unlock:
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004066 mutex_unlock(&fs_info->balance_mutex);
4067 mutex_unlock(&fs_info->volume_mutex);
Ilya Dryomoved0fb782013-01-20 15:57:57 +02004068 if (need_unlock)
4069 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4070out:
Liu Boe54bfa32012-06-29 03:58:48 -06004071 mnt_drop_write_file(file);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004072 return ret;
4073}
4074
Ilya Dryomov837d5b62012-01-16 22:04:49 +02004075static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4076{
4077 if (!capable(CAP_SYS_ADMIN))
4078 return -EPERM;
4079
4080 switch (cmd) {
4081 case BTRFS_BALANCE_CTL_PAUSE:
4082 return btrfs_pause_balance(root->fs_info);
Ilya Dryomova7e99c62012-01-16 22:04:49 +02004083 case BTRFS_BALANCE_CTL_CANCEL:
4084 return btrfs_cancel_balance(root->fs_info);
Ilya Dryomov837d5b62012-01-16 22:04:49 +02004085 }
4086
4087 return -EINVAL;
4088}
4089
Ilya Dryomov19a39dc2012-01-16 22:04:49 +02004090static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4091 void __user *arg)
4092{
4093 struct btrfs_fs_info *fs_info = root->fs_info;
4094 struct btrfs_ioctl_balance_args *bargs;
4095 int ret = 0;
4096
4097 if (!capable(CAP_SYS_ADMIN))
4098 return -EPERM;
4099
4100 mutex_lock(&fs_info->balance_mutex);
4101 if (!fs_info->balance_ctl) {
4102 ret = -ENOTCONN;
4103 goto out;
4104 }
4105
4106 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
4107 if (!bargs) {
4108 ret = -ENOMEM;
4109 goto out;
4110 }
4111
4112 update_ioctl_balance_args(fs_info, 1, bargs);
4113
4114 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4115 ret = -EFAULT;
4116
4117 kfree(bargs);
4118out:
4119 mutex_unlock(&fs_info->balance_mutex);
4120 return ret;
4121}
4122
Miao Xie905b0dd2012-11-26 08:50:11 +00004123static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
Arne Jansen5d13a372011-09-14 15:53:51 +02004124{
Al Viro496ad9a2013-01-23 17:07:38 -05004125 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Arne Jansen5d13a372011-09-14 15:53:51 +02004126 struct btrfs_ioctl_quota_ctl_args *sa;
4127 struct btrfs_trans_handle *trans = NULL;
4128 int ret;
4129 int err;
4130
4131 if (!capable(CAP_SYS_ADMIN))
4132 return -EPERM;
4133
Miao Xie905b0dd2012-11-26 08:50:11 +00004134 ret = mnt_want_write_file(file);
4135 if (ret)
4136 return ret;
Arne Jansen5d13a372011-09-14 15:53:51 +02004137
4138 sa = memdup_user(arg, sizeof(*sa));
Miao Xie905b0dd2012-11-26 08:50:11 +00004139 if (IS_ERR(sa)) {
4140 ret = PTR_ERR(sa);
4141 goto drop_write;
4142 }
Arne Jansen5d13a372011-09-14 15:53:51 +02004143
Wang Shilong7708f022013-04-07 10:24:57 +00004144 down_write(&root->fs_info->subvol_sem);
Jan Schmidt2f232032013-04-25 16:04:51 +00004145 trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4146 if (IS_ERR(trans)) {
4147 ret = PTR_ERR(trans);
4148 goto out;
Arne Jansen5d13a372011-09-14 15:53:51 +02004149 }
4150
4151 switch (sa->cmd) {
4152 case BTRFS_QUOTA_CTL_ENABLE:
4153 ret = btrfs_quota_enable(trans, root->fs_info);
4154 break;
4155 case BTRFS_QUOTA_CTL_DISABLE:
4156 ret = btrfs_quota_disable(trans, root->fs_info);
4157 break;
Arne Jansen5d13a372011-09-14 15:53:51 +02004158 default:
4159 ret = -EINVAL;
4160 break;
4161 }
4162
Jan Schmidt2f232032013-04-25 16:04:51 +00004163 err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4164 if (err && !ret)
4165 ret = err;
Arne Jansen5d13a372011-09-14 15:53:51 +02004166out:
4167 kfree(sa);
Wang Shilong7708f022013-04-07 10:24:57 +00004168 up_write(&root->fs_info->subvol_sem);
Miao Xie905b0dd2012-11-26 08:50:11 +00004169drop_write:
4170 mnt_drop_write_file(file);
Arne Jansen5d13a372011-09-14 15:53:51 +02004171 return ret;
4172}
4173
Miao Xie905b0dd2012-11-26 08:50:11 +00004174static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
Arne Jansen5d13a372011-09-14 15:53:51 +02004175{
Al Viro496ad9a2013-01-23 17:07:38 -05004176 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Arne Jansen5d13a372011-09-14 15:53:51 +02004177 struct btrfs_ioctl_qgroup_assign_args *sa;
4178 struct btrfs_trans_handle *trans;
4179 int ret;
4180 int err;
4181
4182 if (!capable(CAP_SYS_ADMIN))
4183 return -EPERM;
4184
Miao Xie905b0dd2012-11-26 08:50:11 +00004185 ret = mnt_want_write_file(file);
4186 if (ret)
4187 return ret;
Arne Jansen5d13a372011-09-14 15:53:51 +02004188
4189 sa = memdup_user(arg, sizeof(*sa));
Miao Xie905b0dd2012-11-26 08:50:11 +00004190 if (IS_ERR(sa)) {
4191 ret = PTR_ERR(sa);
4192 goto drop_write;
4193 }
Arne Jansen5d13a372011-09-14 15:53:51 +02004194
4195 trans = btrfs_join_transaction(root);
4196 if (IS_ERR(trans)) {
4197 ret = PTR_ERR(trans);
4198 goto out;
4199 }
4200
4201 /* FIXME: check if the IDs really exist */
4202 if (sa->assign) {
4203 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4204 sa->src, sa->dst);
4205 } else {
4206 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4207 sa->src, sa->dst);
4208 }
4209
4210 err = btrfs_end_transaction(trans, root);
4211 if (err && !ret)
4212 ret = err;
4213
4214out:
4215 kfree(sa);
Miao Xie905b0dd2012-11-26 08:50:11 +00004216drop_write:
4217 mnt_drop_write_file(file);
Arne Jansen5d13a372011-09-14 15:53:51 +02004218 return ret;
4219}
4220
Miao Xie905b0dd2012-11-26 08:50:11 +00004221static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
Arne Jansen5d13a372011-09-14 15:53:51 +02004222{
Al Viro496ad9a2013-01-23 17:07:38 -05004223 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Arne Jansen5d13a372011-09-14 15:53:51 +02004224 struct btrfs_ioctl_qgroup_create_args *sa;
4225 struct btrfs_trans_handle *trans;
4226 int ret;
4227 int err;
4228
4229 if (!capable(CAP_SYS_ADMIN))
4230 return -EPERM;
4231
Miao Xie905b0dd2012-11-26 08:50:11 +00004232 ret = mnt_want_write_file(file);
4233 if (ret)
4234 return ret;
Arne Jansen5d13a372011-09-14 15:53:51 +02004235
4236 sa = memdup_user(arg, sizeof(*sa));
Miao Xie905b0dd2012-11-26 08:50:11 +00004237 if (IS_ERR(sa)) {
4238 ret = PTR_ERR(sa);
4239 goto drop_write;
4240 }
Arne Jansen5d13a372011-09-14 15:53:51 +02004241
Miao Xied86e56c2012-11-15 11:35:41 +00004242 if (!sa->qgroupid) {
4243 ret = -EINVAL;
4244 goto out;
4245 }
4246
Arne Jansen5d13a372011-09-14 15:53:51 +02004247 trans = btrfs_join_transaction(root);
4248 if (IS_ERR(trans)) {
4249 ret = PTR_ERR(trans);
4250 goto out;
4251 }
4252
4253 /* FIXME: check if the IDs really exist */
4254 if (sa->create) {
4255 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
4256 NULL);
4257 } else {
4258 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4259 }
4260
4261 err = btrfs_end_transaction(trans, root);
4262 if (err && !ret)
4263 ret = err;
4264
4265out:
4266 kfree(sa);
Miao Xie905b0dd2012-11-26 08:50:11 +00004267drop_write:
4268 mnt_drop_write_file(file);
Arne Jansen5d13a372011-09-14 15:53:51 +02004269 return ret;
4270}
4271
Miao Xie905b0dd2012-11-26 08:50:11 +00004272static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
Arne Jansen5d13a372011-09-14 15:53:51 +02004273{
Al Viro496ad9a2013-01-23 17:07:38 -05004274 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Arne Jansen5d13a372011-09-14 15:53:51 +02004275 struct btrfs_ioctl_qgroup_limit_args *sa;
4276 struct btrfs_trans_handle *trans;
4277 int ret;
4278 int err;
4279 u64 qgroupid;
4280
4281 if (!capable(CAP_SYS_ADMIN))
4282 return -EPERM;
4283
Miao Xie905b0dd2012-11-26 08:50:11 +00004284 ret = mnt_want_write_file(file);
4285 if (ret)
4286 return ret;
Arne Jansen5d13a372011-09-14 15:53:51 +02004287
4288 sa = memdup_user(arg, sizeof(*sa));
Miao Xie905b0dd2012-11-26 08:50:11 +00004289 if (IS_ERR(sa)) {
4290 ret = PTR_ERR(sa);
4291 goto drop_write;
4292 }
Arne Jansen5d13a372011-09-14 15:53:51 +02004293
4294 trans = btrfs_join_transaction(root);
4295 if (IS_ERR(trans)) {
4296 ret = PTR_ERR(trans);
4297 goto out;
4298 }
4299
4300 qgroupid = sa->qgroupid;
4301 if (!qgroupid) {
4302 /* take the current subvol as qgroup */
4303 qgroupid = root->root_key.objectid;
4304 }
4305
4306 /* FIXME: check if the IDs really exist */
4307 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
4308
4309 err = btrfs_end_transaction(trans, root);
4310 if (err && !ret)
4311 ret = err;
4312
4313out:
4314 kfree(sa);
Miao Xie905b0dd2012-11-26 08:50:11 +00004315drop_write:
4316 mnt_drop_write_file(file);
Arne Jansen5d13a372011-09-14 15:53:51 +02004317 return ret;
4318}
4319
Jan Schmidt2f232032013-04-25 16:04:51 +00004320static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4321{
Al Viro6d0379e2013-06-16 19:32:35 +04004322 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Jan Schmidt2f232032013-04-25 16:04:51 +00004323 struct btrfs_ioctl_quota_rescan_args *qsa;
4324 int ret;
4325
4326 if (!capable(CAP_SYS_ADMIN))
4327 return -EPERM;
4328
4329 ret = mnt_want_write_file(file);
4330 if (ret)
4331 return ret;
4332
4333 qsa = memdup_user(arg, sizeof(*qsa));
4334 if (IS_ERR(qsa)) {
4335 ret = PTR_ERR(qsa);
4336 goto drop_write;
4337 }
4338
4339 if (qsa->flags) {
4340 ret = -EINVAL;
4341 goto out;
4342 }
4343
4344 ret = btrfs_qgroup_rescan(root->fs_info);
4345
4346out:
4347 kfree(qsa);
4348drop_write:
4349 mnt_drop_write_file(file);
4350 return ret;
4351}
4352
4353static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4354{
Al Viro6d0379e2013-06-16 19:32:35 +04004355 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Jan Schmidt2f232032013-04-25 16:04:51 +00004356 struct btrfs_ioctl_quota_rescan_args *qsa;
4357 int ret = 0;
4358
4359 if (!capable(CAP_SYS_ADMIN))
4360 return -EPERM;
4361
4362 qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
4363 if (!qsa)
4364 return -ENOMEM;
4365
4366 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4367 qsa->flags = 1;
4368 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
4369 }
4370
4371 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4372 ret = -EFAULT;
4373
4374 kfree(qsa);
4375 return ret;
4376}
4377
Jan Schmidt57254b6e2013-05-06 19:14:17 +00004378static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
4379{
Al Viro54563d42013-09-01 15:57:51 -04004380 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Jan Schmidt57254b6e2013-05-06 19:14:17 +00004381
4382 if (!capable(CAP_SYS_ADMIN))
4383 return -EPERM;
4384
4385 return btrfs_qgroup_wait_for_completion(root->fs_info);
4386}
4387
Alexander Block8ea05e32012-07-25 17:35:53 +02004388static long btrfs_ioctl_set_received_subvol(struct file *file,
4389 void __user *arg)
4390{
4391 struct btrfs_ioctl_received_subvol_args *sa = NULL;
Al Viro496ad9a2013-01-23 17:07:38 -05004392 struct inode *inode = file_inode(file);
Alexander Block8ea05e32012-07-25 17:35:53 +02004393 struct btrfs_root *root = BTRFS_I(inode)->root;
4394 struct btrfs_root_item *root_item = &root->root_item;
4395 struct btrfs_trans_handle *trans;
4396 struct timespec ct = CURRENT_TIME;
4397 int ret = 0;
Stefan Behrensdd5f9612013-08-15 17:11:20 +02004398 int received_uuid_changed;
Alexander Block8ea05e32012-07-25 17:35:53 +02004399
4400 ret = mnt_want_write_file(file);
4401 if (ret < 0)
4402 return ret;
4403
4404 down_write(&root->fs_info->subvol_sem);
4405
4406 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
4407 ret = -EINVAL;
4408 goto out;
4409 }
4410
4411 if (btrfs_root_readonly(root)) {
4412 ret = -EROFS;
4413 goto out;
4414 }
4415
4416 if (!inode_owner_or_capable(inode)) {
4417 ret = -EACCES;
4418 goto out;
4419 }
4420
4421 sa = memdup_user(arg, sizeof(*sa));
4422 if (IS_ERR(sa)) {
4423 ret = PTR_ERR(sa);
4424 sa = NULL;
4425 goto out;
4426 }
4427
Stefan Behrensdd5f9612013-08-15 17:11:20 +02004428 /*
4429 * 1 - root item
4430 * 2 - uuid items (received uuid + subvol uuid)
4431 */
4432 trans = btrfs_start_transaction(root, 3);
Alexander Block8ea05e32012-07-25 17:35:53 +02004433 if (IS_ERR(trans)) {
4434 ret = PTR_ERR(trans);
4435 trans = NULL;
4436 goto out;
4437 }
4438
4439 sa->rtransid = trans->transid;
4440 sa->rtime.sec = ct.tv_sec;
4441 sa->rtime.nsec = ct.tv_nsec;
4442
Stefan Behrensdd5f9612013-08-15 17:11:20 +02004443 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4444 BTRFS_UUID_SIZE);
4445 if (received_uuid_changed &&
4446 !btrfs_is_empty_uuid(root_item->received_uuid))
4447 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
4448 root_item->received_uuid,
4449 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4450 root->root_key.objectid);
Alexander Block8ea05e32012-07-25 17:35:53 +02004451 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4452 btrfs_set_root_stransid(root_item, sa->stransid);
4453 btrfs_set_root_rtransid(root_item, sa->rtransid);
Qu Wenruo3cae2102013-07-16 11:19:18 +08004454 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4455 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4456 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4457 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
Alexander Block8ea05e32012-07-25 17:35:53 +02004458
4459 ret = btrfs_update_root(trans, root->fs_info->tree_root,
4460 &root->root_key, &root->root_item);
4461 if (ret < 0) {
4462 btrfs_end_transaction(trans, root);
Alexander Block8ea05e32012-07-25 17:35:53 +02004463 goto out;
Stefan Behrensdd5f9612013-08-15 17:11:20 +02004464 }
4465 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4466 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
4467 sa->uuid,
4468 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4469 root->root_key.objectid);
4470 if (ret < 0 && ret != -EEXIST) {
4471 btrfs_abort_transaction(trans, root, ret);
Alexander Block8ea05e32012-07-25 17:35:53 +02004472 goto out;
Stefan Behrensdd5f9612013-08-15 17:11:20 +02004473 }
4474 }
4475 ret = btrfs_commit_transaction(trans, root);
4476 if (ret < 0) {
4477 btrfs_abort_transaction(trans, root, ret);
4478 goto out;
Alexander Block8ea05e32012-07-25 17:35:53 +02004479 }
4480
4481 ret = copy_to_user(arg, sa, sizeof(*sa));
4482 if (ret)
4483 ret = -EFAULT;
4484
4485out:
4486 kfree(sa);
4487 up_write(&root->fs_info->subvol_sem);
4488 mnt_drop_write_file(file);
4489 return ret;
4490}
4491
jeff.liu867ab662013-01-05 02:48:01 +00004492static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
4493{
Al Viro6d0379e2013-06-16 19:32:35 +04004494 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Anand Jaina1b83ac2013-07-19 17:39:32 +08004495 size_t len;
jeff.liu867ab662013-01-05 02:48:01 +00004496 int ret;
Anand Jaina1b83ac2013-07-19 17:39:32 +08004497 char label[BTRFS_LABEL_SIZE];
4498
4499 spin_lock(&root->fs_info->super_lock);
4500 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4501 spin_unlock(&root->fs_info->super_lock);
4502
4503 len = strnlen(label, BTRFS_LABEL_SIZE);
jeff.liu867ab662013-01-05 02:48:01 +00004504
4505 if (len == BTRFS_LABEL_SIZE) {
Frank Holtonefe120a2013-12-20 11:37:06 -05004506 btrfs_warn(root->fs_info,
4507 "label is too long, return the first %zu bytes", --len);
jeff.liu867ab662013-01-05 02:48:01 +00004508 }
4509
jeff.liu867ab662013-01-05 02:48:01 +00004510 ret = copy_to_user(arg, label, len);
jeff.liu867ab662013-01-05 02:48:01 +00004511
4512 return ret ? -EFAULT : 0;
4513}
4514
jeff.liua8bfd4a2013-01-05 02:48:08 +00004515static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4516{
Al Viro6d0379e2013-06-16 19:32:35 +04004517 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
jeff.liua8bfd4a2013-01-05 02:48:08 +00004518 struct btrfs_super_block *super_block = root->fs_info->super_copy;
4519 struct btrfs_trans_handle *trans;
4520 char label[BTRFS_LABEL_SIZE];
4521 int ret;
4522
4523 if (!capable(CAP_SYS_ADMIN))
4524 return -EPERM;
4525
4526 if (copy_from_user(label, arg, sizeof(label)))
4527 return -EFAULT;
4528
4529 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
Frank Holtonefe120a2013-12-20 11:37:06 -05004530 btrfs_err(root->fs_info, "unable to set label with more than %d bytes",
jeff.liua8bfd4a2013-01-05 02:48:08 +00004531 BTRFS_LABEL_SIZE - 1);
4532 return -EINVAL;
4533 }
4534
4535 ret = mnt_want_write_file(file);
4536 if (ret)
4537 return ret;
4538
jeff.liua8bfd4a2013-01-05 02:48:08 +00004539 trans = btrfs_start_transaction(root, 0);
4540 if (IS_ERR(trans)) {
4541 ret = PTR_ERR(trans);
4542 goto out_unlock;
4543 }
4544
Anand Jaina1b83ac2013-07-19 17:39:32 +08004545 spin_lock(&root->fs_info->super_lock);
jeff.liua8bfd4a2013-01-05 02:48:08 +00004546 strcpy(super_block->label, label);
Anand Jaina1b83ac2013-07-19 17:39:32 +08004547 spin_unlock(&root->fs_info->super_lock);
jeff.liua8bfd4a2013-01-05 02:48:08 +00004548 ret = btrfs_end_transaction(trans, root);
4549
4550out_unlock:
jeff.liua8bfd4a2013-01-05 02:48:08 +00004551 mnt_drop_write_file(file);
4552 return ret;
4553}
4554
Jeff Mahoney2eaa0552013-11-15 15:33:55 -05004555#define INIT_FEATURE_FLAGS(suffix) \
4556 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
4557 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
4558 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
4559
4560static int btrfs_ioctl_get_supported_features(struct file *file,
4561 void __user *arg)
4562{
4563 static struct btrfs_ioctl_feature_flags features[3] = {
4564 INIT_FEATURE_FLAGS(SUPP),
4565 INIT_FEATURE_FLAGS(SAFE_SET),
4566 INIT_FEATURE_FLAGS(SAFE_CLEAR)
4567 };
4568
4569 if (copy_to_user(arg, &features, sizeof(features)))
4570 return -EFAULT;
4571
4572 return 0;
4573}
4574
4575static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
4576{
4577 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4578 struct btrfs_super_block *super_block = root->fs_info->super_copy;
4579 struct btrfs_ioctl_feature_flags features;
4580
4581 features.compat_flags = btrfs_super_compat_flags(super_block);
4582 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
4583 features.incompat_flags = btrfs_super_incompat_flags(super_block);
4584
4585 if (copy_to_user(arg, &features, sizeof(features)))
4586 return -EFAULT;
4587
4588 return 0;
4589}
4590
Jeff Mahoney3b02a682013-11-01 13:07:02 -04004591static int check_feature_bits(struct btrfs_root *root,
4592 enum btrfs_feature_set set,
Jeff Mahoney2eaa0552013-11-15 15:33:55 -05004593 u64 change_mask, u64 flags, u64 supported_flags,
4594 u64 safe_set, u64 safe_clear)
4595{
Jeff Mahoney3b02a682013-11-01 13:07:02 -04004596 const char *type = btrfs_feature_set_names[set];
4597 char *names;
Jeff Mahoney2eaa0552013-11-15 15:33:55 -05004598 u64 disallowed, unsupported;
4599 u64 set_mask = flags & change_mask;
4600 u64 clear_mask = ~flags & change_mask;
4601
4602 unsupported = set_mask & ~supported_flags;
4603 if (unsupported) {
Jeff Mahoney3b02a682013-11-01 13:07:02 -04004604 names = btrfs_printable_features(set, unsupported);
4605 if (names) {
4606 btrfs_warn(root->fs_info,
4607 "this kernel does not support the %s feature bit%s",
4608 names, strchr(names, ',') ? "s" : "");
4609 kfree(names);
4610 } else
4611 btrfs_warn(root->fs_info,
Jeff Mahoney2eaa0552013-11-15 15:33:55 -05004612 "this kernel does not support %s bits 0x%llx",
4613 type, unsupported);
4614 return -EOPNOTSUPP;
4615 }
4616
4617 disallowed = set_mask & ~safe_set;
4618 if (disallowed) {
Jeff Mahoney3b02a682013-11-01 13:07:02 -04004619 names = btrfs_printable_features(set, disallowed);
4620 if (names) {
4621 btrfs_warn(root->fs_info,
4622 "can't set the %s feature bit%s while mounted",
4623 names, strchr(names, ',') ? "s" : "");
4624 kfree(names);
4625 } else
4626 btrfs_warn(root->fs_info,
Jeff Mahoney2eaa0552013-11-15 15:33:55 -05004627 "can't set %s bits 0x%llx while mounted",
4628 type, disallowed);
4629 return -EPERM;
4630 }
4631
4632 disallowed = clear_mask & ~safe_clear;
4633 if (disallowed) {
Jeff Mahoney3b02a682013-11-01 13:07:02 -04004634 names = btrfs_printable_features(set, disallowed);
4635 if (names) {
4636 btrfs_warn(root->fs_info,
4637 "can't clear the %s feature bit%s while mounted",
4638 names, strchr(names, ',') ? "s" : "");
4639 kfree(names);
4640 } else
4641 btrfs_warn(root->fs_info,
Jeff Mahoney2eaa0552013-11-15 15:33:55 -05004642 "can't clear %s bits 0x%llx while mounted",
4643 type, disallowed);
4644 return -EPERM;
4645 }
4646
4647 return 0;
4648}
4649
4650#define check_feature(root, change_mask, flags, mask_base) \
Jeff Mahoney3b02a682013-11-01 13:07:02 -04004651check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
Jeff Mahoney2eaa0552013-11-15 15:33:55 -05004652 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
4653 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
4654 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
4655
4656static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
4657{
4658 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4659 struct btrfs_super_block *super_block = root->fs_info->super_copy;
4660 struct btrfs_ioctl_feature_flags flags[2];
4661 struct btrfs_trans_handle *trans;
4662 u64 newflags;
4663 int ret;
4664
4665 if (!capable(CAP_SYS_ADMIN))
4666 return -EPERM;
4667
4668 if (copy_from_user(flags, arg, sizeof(flags)))
4669 return -EFAULT;
4670
4671 /* Nothing to do */
4672 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
4673 !flags[0].incompat_flags)
4674 return 0;
4675
4676 ret = check_feature(root, flags[0].compat_flags,
4677 flags[1].compat_flags, COMPAT);
4678 if (ret)
4679 return ret;
4680
4681 ret = check_feature(root, flags[0].compat_ro_flags,
4682 flags[1].compat_ro_flags, COMPAT_RO);
4683 if (ret)
4684 return ret;
4685
4686 ret = check_feature(root, flags[0].incompat_flags,
4687 flags[1].incompat_flags, INCOMPAT);
4688 if (ret)
4689 return ret;
4690
4691 trans = btrfs_start_transaction(root, 1);
4692 if (IS_ERR(trans))
4693 return PTR_ERR(trans);
4694
4695 spin_lock(&root->fs_info->super_lock);
4696 newflags = btrfs_super_compat_flags(super_block);
4697 newflags |= flags[0].compat_flags & flags[1].compat_flags;
4698 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
4699 btrfs_set_super_compat_flags(super_block, newflags);
4700
4701 newflags = btrfs_super_compat_ro_flags(super_block);
4702 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
4703 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
4704 btrfs_set_super_compat_ro_flags(super_block, newflags);
4705
4706 newflags = btrfs_super_incompat_flags(super_block);
4707 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
4708 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
4709 btrfs_set_super_incompat_flags(super_block, newflags);
4710 spin_unlock(&root->fs_info->super_lock);
4711
4712 return btrfs_end_transaction(trans, root);
4713}
4714
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04004715long btrfs_ioctl(struct file *file, unsigned int
4716 cmd, unsigned long arg)
4717{
Al Viro496ad9a2013-01-23 17:07:38 -05004718 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Christoph Hellwig4bcabaa2008-12-02 06:36:08 -05004719 void __user *argp = (void __user *)arg;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04004720
4721 switch (cmd) {
Christoph Hellwig6cbff002009-04-17 10:37:41 +02004722 case FS_IOC_GETFLAGS:
4723 return btrfs_ioctl_getflags(file, argp);
4724 case FS_IOC_SETFLAGS:
4725 return btrfs_ioctl_setflags(file, argp);
4726 case FS_IOC_GETVERSION:
4727 return btrfs_ioctl_getversion(file, argp);
Li Dongyangf7039b12011-03-24 10:24:28 +00004728 case FITRIM:
4729 return btrfs_ioctl_fitrim(file, argp);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04004730 case BTRFS_IOC_SNAP_CREATE:
Li Zefanfa0d2b92010-12-20 15:53:28 +08004731 return btrfs_ioctl_snap_create(file, argp, 0);
Li Zefanfdfb1e42010-12-10 06:41:56 +00004732 case BTRFS_IOC_SNAP_CREATE_V2:
Li Zefanfa0d2b92010-12-20 15:53:28 +08004733 return btrfs_ioctl_snap_create_v2(file, argp, 0);
Chris Mason3de45862008-11-17 21:02:50 -05004734 case BTRFS_IOC_SUBVOL_CREATE:
Li Zefanfa0d2b92010-12-20 15:53:28 +08004735 return btrfs_ioctl_snap_create(file, argp, 1);
Arne Jansen6f72c7e2011-09-14 15:58:21 +02004736 case BTRFS_IOC_SUBVOL_CREATE_V2:
4737 return btrfs_ioctl_snap_create_v2(file, argp, 1);
Yan, Zheng76dda932009-09-21 16:00:26 -04004738 case BTRFS_IOC_SNAP_DESTROY:
4739 return btrfs_ioctl_snap_destroy(file, argp);
Li Zefan0caa1022010-12-20 16:30:25 +08004740 case BTRFS_IOC_SUBVOL_GETFLAGS:
4741 return btrfs_ioctl_subvol_getflags(file, argp);
4742 case BTRFS_IOC_SUBVOL_SETFLAGS:
4743 return btrfs_ioctl_subvol_setflags(file, argp);
Josef Bacik6ef5ed02009-12-11 21:11:29 +00004744 case BTRFS_IOC_DEFAULT_SUBVOL:
4745 return btrfs_ioctl_default_subvol(file, argp);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04004746 case BTRFS_IOC_DEFRAG:
Chris Mason1e701a32010-03-11 09:42:04 -05004747 return btrfs_ioctl_defrag(file, NULL);
4748 case BTRFS_IOC_DEFRAG_RANGE:
4749 return btrfs_ioctl_defrag(file, argp);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04004750 case BTRFS_IOC_RESIZE:
Miao Xie198605a2012-11-26 08:43:45 +00004751 return btrfs_ioctl_resize(file, argp);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04004752 case BTRFS_IOC_ADD_DEV:
Christoph Hellwig4bcabaa2008-12-02 06:36:08 -05004753 return btrfs_ioctl_add_dev(root, argp);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04004754 case BTRFS_IOC_RM_DEV:
Miao Xieda249272012-11-26 08:44:50 +00004755 return btrfs_ioctl_rm_dev(file, argp);
Jan Schmidt475f6382011-03-11 15:41:01 +01004756 case BTRFS_IOC_FS_INFO:
4757 return btrfs_ioctl_fs_info(root, argp);
4758 case BTRFS_IOC_DEV_INFO:
4759 return btrfs_ioctl_dev_info(root, argp);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04004760 case BTRFS_IOC_BALANCE:
Liu Bo9ba1f6e2012-05-11 18:11:26 +08004761 return btrfs_ioctl_balance(file, NULL);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04004762 case BTRFS_IOC_CLONE:
Sage Weilc5c9cd42008-11-12 14:32:25 -05004763 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
4764 case BTRFS_IOC_CLONE_RANGE:
Christoph Hellwig7a865e82008-12-02 09:52:24 -05004765 return btrfs_ioctl_clone_range(file, argp);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04004766 case BTRFS_IOC_TRANS_START:
4767 return btrfs_ioctl_trans_start(file);
4768 case BTRFS_IOC_TRANS_END:
4769 return btrfs_ioctl_trans_end(file);
Chris Masonac8e9812010-02-28 15:39:26 -05004770 case BTRFS_IOC_TREE_SEARCH:
4771 return btrfs_ioctl_tree_search(file, argp);
4772 case BTRFS_IOC_INO_LOOKUP:
4773 return btrfs_ioctl_ino_lookup(file, argp);
Jan Schmidtd7728c92011-07-07 16:48:38 +02004774 case BTRFS_IOC_INO_PATHS:
4775 return btrfs_ioctl_ino_to_path(root, argp);
4776 case BTRFS_IOC_LOGICAL_INO:
4777 return btrfs_ioctl_logical_to_ino(root, argp);
Josef Bacik1406e432010-01-13 18:19:06 +00004778 case BTRFS_IOC_SPACE_INFO:
4779 return btrfs_ioctl_space_info(root, argp);
Jeff Mahoney01e219e2013-11-01 13:07:03 -04004780 case BTRFS_IOC_GLOBAL_RSV:
4781 return btrfs_ioctl_global_rsv(root, argp);
Filipe David Borba Manana9b199852013-09-23 11:35:11 +01004782 case BTRFS_IOC_SYNC: {
4783 int ret;
4784
Miao Xie91aef862013-11-04 23:13:26 +08004785 ret = btrfs_start_delalloc_roots(root->fs_info, 0);
Filipe David Borba Manana9b199852013-09-23 11:35:11 +01004786 if (ret)
4787 return ret;
4788 ret = btrfs_sync_fs(file->f_dentry->d_sb, 1);
4789 return ret;
4790 }
Sage Weil46204592010-10-29 15:41:32 -04004791 case BTRFS_IOC_START_SYNC:
Miao Xie9a8c28b2012-11-26 08:40:43 +00004792 return btrfs_ioctl_start_sync(root, argp);
Sage Weil46204592010-10-29 15:41:32 -04004793 case BTRFS_IOC_WAIT_SYNC:
Miao Xie9a8c28b2012-11-26 08:40:43 +00004794 return btrfs_ioctl_wait_sync(root, argp);
Jan Schmidt475f6382011-03-11 15:41:01 +01004795 case BTRFS_IOC_SCRUB:
Miao Xieb8e95482012-11-26 08:48:01 +00004796 return btrfs_ioctl_scrub(file, argp);
Jan Schmidt475f6382011-03-11 15:41:01 +01004797 case BTRFS_IOC_SCRUB_CANCEL:
4798 return btrfs_ioctl_scrub_cancel(root, argp);
4799 case BTRFS_IOC_SCRUB_PROGRESS:
4800 return btrfs_ioctl_scrub_progress(root, argp);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004801 case BTRFS_IOC_BALANCE_V2:
Liu Bo9ba1f6e2012-05-11 18:11:26 +08004802 return btrfs_ioctl_balance(file, argp);
Ilya Dryomov837d5b62012-01-16 22:04:49 +02004803 case BTRFS_IOC_BALANCE_CTL:
4804 return btrfs_ioctl_balance_ctl(root, arg);
Ilya Dryomov19a39dc2012-01-16 22:04:49 +02004805 case BTRFS_IOC_BALANCE_PROGRESS:
4806 return btrfs_ioctl_balance_progress(root, argp);
Alexander Block8ea05e32012-07-25 17:35:53 +02004807 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4808 return btrfs_ioctl_set_received_subvol(file, argp);
Alexander Block31db9f72012-07-25 23:19:24 +02004809 case BTRFS_IOC_SEND:
4810 return btrfs_ioctl_send(file, argp);
Stefan Behrensc11d2c22012-05-25 16:06:09 +02004811 case BTRFS_IOC_GET_DEV_STATS:
David Sterbab27f7c02012-06-22 06:30:39 -06004812 return btrfs_ioctl_get_dev_stats(root, argp);
Arne Jansen5d13a372011-09-14 15:53:51 +02004813 case BTRFS_IOC_QUOTA_CTL:
Miao Xie905b0dd2012-11-26 08:50:11 +00004814 return btrfs_ioctl_quota_ctl(file, argp);
Arne Jansen5d13a372011-09-14 15:53:51 +02004815 case BTRFS_IOC_QGROUP_ASSIGN:
Miao Xie905b0dd2012-11-26 08:50:11 +00004816 return btrfs_ioctl_qgroup_assign(file, argp);
Arne Jansen5d13a372011-09-14 15:53:51 +02004817 case BTRFS_IOC_QGROUP_CREATE:
Miao Xie905b0dd2012-11-26 08:50:11 +00004818 return btrfs_ioctl_qgroup_create(file, argp);
Arne Jansen5d13a372011-09-14 15:53:51 +02004819 case BTRFS_IOC_QGROUP_LIMIT:
Miao Xie905b0dd2012-11-26 08:50:11 +00004820 return btrfs_ioctl_qgroup_limit(file, argp);
Jan Schmidt2f232032013-04-25 16:04:51 +00004821 case BTRFS_IOC_QUOTA_RESCAN:
4822 return btrfs_ioctl_quota_rescan(file, argp);
4823 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
4824 return btrfs_ioctl_quota_rescan_status(file, argp);
Jan Schmidt57254b6e2013-05-06 19:14:17 +00004825 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
4826 return btrfs_ioctl_quota_rescan_wait(file, argp);
Stefan Behrens3f6bcfb2012-11-06 15:08:53 +01004827 case BTRFS_IOC_DEV_REPLACE:
4828 return btrfs_ioctl_dev_replace(root, argp);
jeff.liu867ab662013-01-05 02:48:01 +00004829 case BTRFS_IOC_GET_FSLABEL:
4830 return btrfs_ioctl_get_fslabel(file, argp);
jeff.liua8bfd4a2013-01-05 02:48:08 +00004831 case BTRFS_IOC_SET_FSLABEL:
4832 return btrfs_ioctl_set_fslabel(file, argp);
Mark Fasheh416161d2013-08-06 11:42:51 -07004833 case BTRFS_IOC_FILE_EXTENT_SAME:
4834 return btrfs_ioctl_file_extent_same(file, argp);
Jeff Mahoney2eaa0552013-11-15 15:33:55 -05004835 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
4836 return btrfs_ioctl_get_supported_features(file, argp);
4837 case BTRFS_IOC_GET_FEATURES:
4838 return btrfs_ioctl_get_features(file, argp);
4839 case BTRFS_IOC_SET_FEATURES:
4840 return btrfs_ioctl_set_features(file, argp);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04004841 }
4842
4843 return -ENOTTY;
4844}