blob: 2be49b4c82d681a3f8c74e8a086acd2b1f28cca2 [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>
Chris Mason4b4e25f2008-11-20 10:22:27 -050045#include "compat.h"
Christoph Hellwigf46b5a62008-06-11 21:53:53 -040046#include "ctree.h"
47#include "disk-io.h"
48#include "transaction.h"
49#include "btrfs_inode.h"
50#include "ioctl.h"
51#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"
Christoph Hellwigf46b5a62008-06-11 21:53:53 -040059
Christoph Hellwig6cbff002009-04-17 10:37:41 +020060/* Mask out flags that are inappropriate for the given type of inode. */
61static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
62{
63 if (S_ISDIR(mode))
64 return flags;
65 else if (S_ISREG(mode))
66 return flags & ~FS_DIRSYNC_FL;
67 else
68 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
69}
Christoph Hellwigf46b5a62008-06-11 21:53:53 -040070
Christoph Hellwig6cbff002009-04-17 10:37:41 +020071/*
72 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
73 */
74static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
75{
76 unsigned int iflags = 0;
77
78 if (flags & BTRFS_INODE_SYNC)
79 iflags |= FS_SYNC_FL;
80 if (flags & BTRFS_INODE_IMMUTABLE)
81 iflags |= FS_IMMUTABLE_FL;
82 if (flags & BTRFS_INODE_APPEND)
83 iflags |= FS_APPEND_FL;
84 if (flags & BTRFS_INODE_NODUMP)
85 iflags |= FS_NODUMP_FL;
86 if (flags & BTRFS_INODE_NOATIME)
87 iflags |= FS_NOATIME_FL;
88 if (flags & BTRFS_INODE_DIRSYNC)
89 iflags |= FS_DIRSYNC_FL;
Li Zefand0092bd2011-04-15 03:03:06 +000090 if (flags & BTRFS_INODE_NODATACOW)
91 iflags |= FS_NOCOW_FL;
92
93 if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
94 iflags |= FS_COMPR_FL;
95 else if (flags & BTRFS_INODE_NOCOMPRESS)
96 iflags |= FS_NOCOMP_FL;
Christoph Hellwig6cbff002009-04-17 10:37:41 +020097
98 return iflags;
99}
100
101/*
102 * Update inode->i_flags based on the btrfs internal flags.
103 */
104void btrfs_update_iflags(struct inode *inode)
105{
106 struct btrfs_inode *ip = BTRFS_I(inode);
107
108 inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
109
110 if (ip->flags & BTRFS_INODE_SYNC)
111 inode->i_flags |= S_SYNC;
112 if (ip->flags & BTRFS_INODE_IMMUTABLE)
113 inode->i_flags |= S_IMMUTABLE;
114 if (ip->flags & BTRFS_INODE_APPEND)
115 inode->i_flags |= S_APPEND;
116 if (ip->flags & BTRFS_INODE_NOATIME)
117 inode->i_flags |= S_NOATIME;
118 if (ip->flags & BTRFS_INODE_DIRSYNC)
119 inode->i_flags |= S_DIRSYNC;
120}
121
122/*
123 * Inherit flags from the parent inode.
124 *
Josef Bacike27425d2011-09-27 11:01:30 -0400125 * Currently only the compression flags and the cow flags are inherited.
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200126 */
127void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
128{
Chris Mason0b4dcea2009-06-11 11:13:35 -0400129 unsigned int flags;
130
131 if (!dir)
132 return;
133
134 flags = BTRFS_I(dir)->flags;
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200135
Josef Bacike27425d2011-09-27 11:01:30 -0400136 if (flags & BTRFS_INODE_NOCOMPRESS) {
137 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
138 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
139 } else if (flags & BTRFS_INODE_COMPRESS) {
140 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
141 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
142 }
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200143
Josef Bacike27425d2011-09-27 11:01:30 -0400144 if (flags & BTRFS_INODE_NODATACOW)
145 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
146
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200147 btrfs_update_iflags(inode);
148}
149
150static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
151{
152 struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode);
153 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
154
155 if (copy_to_user(arg, &flags, sizeof(flags)))
156 return -EFAULT;
157 return 0;
158}
159
Liu Bo75e7cb72011-03-22 10:12:20 +0000160static int check_flags(unsigned int flags)
161{
162 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
163 FS_NOATIME_FL | FS_NODUMP_FL | \
164 FS_SYNC_FL | FS_DIRSYNC_FL | \
Li Zefane1e8fb62011-04-15 03:02:49 +0000165 FS_NOCOMP_FL | FS_COMPR_FL |
166 FS_NOCOW_FL))
Liu Bo75e7cb72011-03-22 10:12:20 +0000167 return -EOPNOTSUPP;
168
169 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
170 return -EINVAL;
171
Liu Bo75e7cb72011-03-22 10:12:20 +0000172 return 0;
173}
174
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200175static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
176{
177 struct inode *inode = file->f_path.dentry->d_inode;
178 struct btrfs_inode *ip = BTRFS_I(inode);
179 struct btrfs_root *root = ip->root;
180 struct btrfs_trans_handle *trans;
181 unsigned int flags, oldflags;
182 int ret;
Li Zefanf062abf02011-12-29 13:36:45 +0800183 u64 ip_oldflags;
184 unsigned int i_oldflags;
David Sterba7e97b8d2012-09-07 05:56:55 -0600185 umode_t mode;
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200186
Li Zefanb83cc962010-12-20 16:04:08 +0800187 if (btrfs_root_readonly(root))
188 return -EROFS;
189
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200190 if (copy_from_user(&flags, arg, sizeof(flags)))
191 return -EFAULT;
192
Liu Bo75e7cb72011-03-22 10:12:20 +0000193 ret = check_flags(flags);
194 if (ret)
195 return ret;
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200196
Serge E. Hallyn2e149672011-03-23 16:43:26 -0700197 if (!inode_owner_or_capable(inode))
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200198 return -EACCES;
199
Jan Karae7848682012-06-12 16:20:32 +0200200 ret = mnt_want_write_file(file);
201 if (ret)
202 return ret;
203
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200204 mutex_lock(&inode->i_mutex);
205
Li Zefanf062abf02011-12-29 13:36:45 +0800206 ip_oldflags = ip->flags;
207 i_oldflags = inode->i_flags;
David Sterba7e97b8d2012-09-07 05:56:55 -0600208 mode = inode->i_mode;
Li Zefanf062abf02011-12-29 13:36:45 +0800209
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200210 flags = btrfs_mask_flags(inode->i_mode, flags);
211 oldflags = btrfs_flags_to_ioctl(ip->flags);
212 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
213 if (!capable(CAP_LINUX_IMMUTABLE)) {
214 ret = -EPERM;
215 goto out_unlock;
216 }
217 }
218
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200219 if (flags & FS_SYNC_FL)
220 ip->flags |= BTRFS_INODE_SYNC;
221 else
222 ip->flags &= ~BTRFS_INODE_SYNC;
223 if (flags & FS_IMMUTABLE_FL)
224 ip->flags |= BTRFS_INODE_IMMUTABLE;
225 else
226 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
227 if (flags & FS_APPEND_FL)
228 ip->flags |= BTRFS_INODE_APPEND;
229 else
230 ip->flags &= ~BTRFS_INODE_APPEND;
231 if (flags & FS_NODUMP_FL)
232 ip->flags |= BTRFS_INODE_NODUMP;
233 else
234 ip->flags &= ~BTRFS_INODE_NODUMP;
235 if (flags & FS_NOATIME_FL)
236 ip->flags |= BTRFS_INODE_NOATIME;
237 else
238 ip->flags &= ~BTRFS_INODE_NOATIME;
239 if (flags & FS_DIRSYNC_FL)
240 ip->flags |= BTRFS_INODE_DIRSYNC;
241 else
242 ip->flags &= ~BTRFS_INODE_DIRSYNC;
David Sterba7e97b8d2012-09-07 05:56:55 -0600243 if (flags & FS_NOCOW_FL) {
244 if (S_ISREG(mode)) {
245 /*
246 * It's safe to turn csums off here, no extents exist.
247 * Otherwise we want the flag to reflect the real COW
248 * status of the file and will not set it.
249 */
250 if (inode->i_size == 0)
251 ip->flags |= BTRFS_INODE_NODATACOW
252 | BTRFS_INODE_NODATASUM;
253 } else {
254 ip->flags |= BTRFS_INODE_NODATACOW;
255 }
256 } else {
257 /*
258 * Revert back under same assuptions as above
259 */
260 if (S_ISREG(mode)) {
261 if (inode->i_size == 0)
262 ip->flags &= ~(BTRFS_INODE_NODATACOW
263 | BTRFS_INODE_NODATASUM);
264 } else {
265 ip->flags &= ~BTRFS_INODE_NODATACOW;
266 }
267 }
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200268
Liu Bo75e7cb72011-03-22 10:12:20 +0000269 /*
270 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
271 * flag may be changed automatically if compression code won't make
272 * things smaller.
273 */
274 if (flags & FS_NOCOMP_FL) {
275 ip->flags &= ~BTRFS_INODE_COMPRESS;
276 ip->flags |= BTRFS_INODE_NOCOMPRESS;
277 } else if (flags & FS_COMPR_FL) {
278 ip->flags |= BTRFS_INODE_COMPRESS;
279 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
Li Zefanebcb9042011-04-15 03:03:17 +0000280 } else {
281 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
Liu Bo75e7cb72011-03-22 10:12:20 +0000282 }
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200283
Li Zefan4da6f1a2011-12-29 13:39:50 +0800284 trans = btrfs_start_transaction(root, 1);
Li Zefanf062abf02011-12-29 13:36:45 +0800285 if (IS_ERR(trans)) {
286 ret = PTR_ERR(trans);
287 goto out_drop;
288 }
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200289
Li Zefan306424cc2011-12-14 20:12:02 -0500290 btrfs_update_iflags(inode);
Josef Bacik0c4d2d92012-04-05 15:03:02 -0400291 inode_inc_iversion(inode);
Li Zefan306424cc2011-12-14 20:12:02 -0500292 inode->i_ctime = CURRENT_TIME;
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200293 ret = btrfs_update_inode(trans, root, inode);
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200294
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200295 btrfs_end_transaction(trans, root);
Li Zefanf062abf02011-12-29 13:36:45 +0800296 out_drop:
297 if (ret) {
298 ip->flags = ip_oldflags;
299 inode->i_flags = i_oldflags;
300 }
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200301
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200302 out_unlock:
303 mutex_unlock(&inode->i_mutex);
Jan Karae7848682012-06-12 16:20:32 +0200304 mnt_drop_write_file(file);
liubo2d4e6f6ad2011-02-24 09:38:16 +0000305 return ret;
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200306}
307
308static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
309{
310 struct inode *inode = file->f_path.dentry->d_inode;
311
312 return put_user(inode->i_generation, arg);
313}
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400314
Li Dongyangf7039b12011-03-24 10:24:28 +0000315static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
316{
Al Viro815745c2011-11-17 15:40:49 -0500317 struct btrfs_fs_info *fs_info = btrfs_sb(fdentry(file)->d_sb);
Li Dongyangf7039b12011-03-24 10:24:28 +0000318 struct btrfs_device *device;
319 struct request_queue *q;
320 struct fstrim_range range;
321 u64 minlen = ULLONG_MAX;
322 u64 num_devices = 0;
Al Viro815745c2011-11-17 15:40:49 -0500323 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
Li Dongyangf7039b12011-03-24 10:24:28 +0000324 int ret;
325
326 if (!capable(CAP_SYS_ADMIN))
327 return -EPERM;
328
Xiao Guangrong1f781602011-04-20 10:09:16 +0000329 rcu_read_lock();
330 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
331 dev_list) {
Li Dongyangf7039b12011-03-24 10:24:28 +0000332 if (!device->bdev)
333 continue;
334 q = bdev_get_queue(device->bdev);
335 if (blk_queue_discard(q)) {
336 num_devices++;
337 minlen = min((u64)q->limits.discard_granularity,
338 minlen);
339 }
340 }
Xiao Guangrong1f781602011-04-20 10:09:16 +0000341 rcu_read_unlock();
Lukas Czernerf4c697e2011-09-05 16:34:54 +0200342
Li Dongyangf7039b12011-03-24 10:24:28 +0000343 if (!num_devices)
344 return -EOPNOTSUPP;
Li Dongyangf7039b12011-03-24 10:24:28 +0000345 if (copy_from_user(&range, arg, sizeof(range)))
346 return -EFAULT;
Lukas Czernere515c182012-10-16 09:34:36 +0000347 if (range.start > total_bytes ||
348 range.len < fs_info->sb->s_blocksize)
Lukas Czernerf4c697e2011-09-05 16:34:54 +0200349 return -EINVAL;
Li Dongyangf7039b12011-03-24 10:24:28 +0000350
Lukas Czernerf4c697e2011-09-05 16:34:54 +0200351 range.len = min(range.len, total_bytes - range.start);
Li Dongyangf7039b12011-03-24 10:24:28 +0000352 range.minlen = max(range.minlen, minlen);
Al Viro815745c2011-11-17 15:40:49 -0500353 ret = btrfs_trim_fs(fs_info->tree_root, &range);
Li Dongyangf7039b12011-03-24 10:24:28 +0000354 if (ret < 0)
355 return ret;
356
357 if (copy_to_user(arg, &range, sizeof(range)))
358 return -EFAULT;
359
360 return 0;
361}
362
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400363static noinline int create_subvol(struct btrfs_root *root,
364 struct dentry *dentry,
Sage Weil72fd0322010-10-29 15:41:32 -0400365 char *name, int namelen,
Arne Jansen6f72c7e2011-09-14 15:58:21 +0200366 u64 *async_transid,
367 struct btrfs_qgroup_inherit **inherit)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400368{
369 struct btrfs_trans_handle *trans;
370 struct btrfs_key key;
371 struct btrfs_root_item root_item;
372 struct btrfs_inode_item *inode_item;
373 struct extent_buffer *leaf;
Yan, Zheng76dda932009-09-21 16:00:26 -0400374 struct btrfs_root *new_root;
Al Viro2fbe8c82011-07-16 21:38:06 -0400375 struct dentry *parent = dentry->d_parent;
Josef Bacik6a912212010-11-20 09:48:00 +0000376 struct inode *dir;
Alexander Block8ea05e32012-07-25 17:35:53 +0200377 struct timespec cur_time = CURRENT_TIME;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400378 int ret;
379 int err;
380 u64 objectid;
381 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
Chris Mason3de45862008-11-17 21:02:50 -0500382 u64 index = 0;
Alexander Block8ea05e32012-07-25 17:35:53 +0200383 uuid_le new_uuid;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400384
Li Zefan581bb052011-04-20 10:06:11 +0800385 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
Al Viro2fbe8c82011-07-16 21:38:06 -0400386 if (ret)
Yan, Zhenga22285a2010-05-16 10:48:46 -0400387 return ret;
Josef Bacik6a912212010-11-20 09:48:00 +0000388
389 dir = parent->d_inode;
390
Josef Bacik9ed74f22009-09-11 16:12:44 -0400391 /*
392 * 1 - inode item
393 * 2 - refs
394 * 1 - root item
395 * 2 - dir items
396 */
Yan, Zhenga22285a2010-05-16 10:48:46 -0400397 trans = btrfs_start_transaction(root, 6);
Al Viro2fbe8c82011-07-16 21:38:06 -0400398 if (IS_ERR(trans))
Yan, Zhenga22285a2010-05-16 10:48:46 -0400399 return PTR_ERR(trans);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400400
Arne Jansen6f72c7e2011-09-14 15:58:21 +0200401 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid,
402 inherit ? *inherit : NULL);
403 if (ret)
404 goto fail;
405
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400406 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
Jan Schmidt5581a512012-05-16 17:04:52 +0200407 0, objectid, NULL, 0, 0, 0);
Josef Bacik8e8a1e32008-07-24 12:17:14 -0400408 if (IS_ERR(leaf)) {
409 ret = PTR_ERR(leaf);
410 goto fail;
411 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400412
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400413 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400414 btrfs_set_header_bytenr(leaf, leaf->start);
415 btrfs_set_header_generation(leaf, trans->transid);
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400416 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400417 btrfs_set_header_owner(leaf, objectid);
418
419 write_extent_buffer(leaf, root->fs_info->fsid,
420 (unsigned long)btrfs_header_fsid(leaf),
421 BTRFS_FSID_SIZE);
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400422 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
423 (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
424 BTRFS_UUID_SIZE);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400425 btrfs_mark_buffer_dirty(leaf);
426
Alexander Block8ea05e32012-07-25 17:35:53 +0200427 memset(&root_item, 0, sizeof(root_item));
428
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400429 inode_item = &root_item.inode;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400430 inode_item->generation = cpu_to_le64(1);
431 inode_item->size = cpu_to_le64(3);
432 inode_item->nlink = cpu_to_le32(1);
Yan Zhenga76a3cd2008-10-09 11:46:29 -0400433 inode_item->nbytes = cpu_to_le64(root->leafsize);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400434 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
435
Li Zefan08fe4db2011-03-28 02:01:25 +0000436 root_item.flags = 0;
437 root_item.byte_limit = 0;
438 inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
439
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400440 btrfs_set_root_bytenr(&root_item, leaf->start);
Yan Zheng84234f32008-10-29 14:49:05 -0400441 btrfs_set_root_generation(&root_item, trans->transid);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400442 btrfs_set_root_level(&root_item, 0);
443 btrfs_set_root_refs(&root_item, 1);
Yan, Zheng86b9f2e2009-11-12 09:36:50 +0000444 btrfs_set_root_used(&root_item, leaf->len);
Yan Zheng80ff3852008-10-30 14:20:02 -0400445 btrfs_set_root_last_snapshot(&root_item, 0);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400446
Alexander Block8ea05e32012-07-25 17:35:53 +0200447 btrfs_set_root_generation_v2(&root_item,
448 btrfs_root_generation(&root_item));
449 uuid_le_gen(&new_uuid);
450 memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
451 root_item.otime.sec = cpu_to_le64(cur_time.tv_sec);
Dan Carpenterdadd1102012-07-30 02:10:44 -0600452 root_item.otime.nsec = cpu_to_le32(cur_time.tv_nsec);
Alexander Block8ea05e32012-07-25 17:35:53 +0200453 root_item.ctime = root_item.otime;
454 btrfs_set_root_ctransid(&root_item, trans->transid);
455 btrfs_set_root_otransid(&root_item, trans->transid);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400456
Chris Mason925baed2008-06-25 16:01:30 -0400457 btrfs_tree_unlock(leaf);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400458 free_extent_buffer(leaf);
459 leaf = NULL;
460
461 btrfs_set_root_dirid(&root_item, new_dirid);
462
463 key.objectid = objectid;
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400464 key.offset = 0;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400465 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
466 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
467 &root_item);
468 if (ret)
469 goto fail;
470
Yan, Zheng76dda932009-09-21 16:00:26 -0400471 key.offset = (u64)-1;
472 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100473 if (IS_ERR(new_root)) {
474 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
475 ret = PTR_ERR(new_root);
476 goto fail;
477 }
Yan, Zheng76dda932009-09-21 16:00:26 -0400478
479 btrfs_record_root_in_trans(trans, new_root);
480
Josef Bacikd82a6f1d2011-05-11 15:26:06 -0400481 ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
Mark Fashehce598972011-07-26 11:32:23 -0700482 if (ret) {
483 /* We potentially lose an unused inode item here */
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100484 btrfs_abort_transaction(trans, root, ret);
Mark Fashehce598972011-07-26 11:32:23 -0700485 goto fail;
486 }
487
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400488 /*
489 * insert the directory item
490 */
Chris Mason3de45862008-11-17 21:02:50 -0500491 ret = btrfs_set_inode_index(dir, &index);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100492 if (ret) {
493 btrfs_abort_transaction(trans, root, ret);
494 goto fail;
495 }
Chris Mason3de45862008-11-17 21:02:50 -0500496
497 ret = btrfs_insert_dir_item(trans, root,
Miao Xie16cdcec2011-04-22 18:12:22 +0800498 name, namelen, dir, &key,
Chris Mason3de45862008-11-17 21:02:50 -0500499 BTRFS_FT_DIR, index);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100500 if (ret) {
501 btrfs_abort_transaction(trans, root, ret);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400502 goto fail;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100503 }
Chris Mason0660b5a2008-11-17 20:37:39 -0500504
Yan Zheng52c26172009-01-05 15:43:43 -0500505 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
506 ret = btrfs_update_inode(trans, root, dir);
507 BUG_ON(ret);
508
Chris Mason0660b5a2008-11-17 20:37:39 -0500509 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
Yan, Zheng4df27c4d2009-09-21 15:56:00 -0400510 objectid, root->root_key.objectid,
Li Zefan33345d012011-04-20 10:31:50 +0800511 btrfs_ino(dir), index, name, namelen);
Yan, Zheng76dda932009-09-21 16:00:26 -0400512
Chris Mason0660b5a2008-11-17 20:37:39 -0500513 BUG_ON(ret);
514
Yan, Zheng76dda932009-09-21 16:00:26 -0400515 d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400516fail:
Sage Weil72fd0322010-10-29 15:41:32 -0400517 if (async_transid) {
518 *async_transid = trans->transid;
519 err = btrfs_commit_transaction_async(trans, root, 1);
520 } else {
521 err = btrfs_commit_transaction(trans, root);
522 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400523 if (err && !ret)
524 ret = err;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400525 return ret;
526}
527
Sage Weil72fd0322010-10-29 15:41:32 -0400528static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
Li Zefanb83cc962010-12-20 16:04:08 +0800529 char *name, int namelen, u64 *async_transid,
Arne Jansen6f72c7e2011-09-14 15:58:21 +0200530 bool readonly, struct btrfs_qgroup_inherit **inherit)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400531{
Yan, Zheng2e4bfab2009-11-12 09:37:02 +0000532 struct inode *inode;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400533 struct btrfs_pending_snapshot *pending_snapshot;
534 struct btrfs_trans_handle *trans;
Yan, Zheng2e4bfab2009-11-12 09:37:02 +0000535 int ret;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400536
537 if (!root->ref_cows)
538 return -EINVAL;
539
Yan, Zhenga22285a2010-05-16 10:48:46 -0400540 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
541 if (!pending_snapshot)
542 return -ENOMEM;
543
Miao Xie66d8f3d2012-09-06 04:02:28 -0600544 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
545 BTRFS_BLOCK_RSV_TEMP);
Yan, Zhenga22285a2010-05-16 10:48:46 -0400546 pending_snapshot->dentry = dentry;
547 pending_snapshot->root = root;
Li Zefanb83cc962010-12-20 16:04:08 +0800548 pending_snapshot->readonly = readonly;
Arne Jansen6f72c7e2011-09-14 15:58:21 +0200549 if (inherit) {
550 pending_snapshot->inherit = *inherit;
551 *inherit = NULL; /* take responsibility to free it */
552 }
Yan, Zhenga22285a2010-05-16 10:48:46 -0400553
Miao Xie48c03c42012-09-06 04:03:56 -0600554 trans = btrfs_start_transaction(root->fs_info->extent_root, 6);
Yan, Zhenga22285a2010-05-16 10:48:46 -0400555 if (IS_ERR(trans)) {
556 ret = PTR_ERR(trans);
557 goto fail;
558 }
559
560 ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
561 BUG_ON(ret);
562
Josef Bacik83515832011-06-14 15:16:14 -0400563 spin_lock(&root->fs_info->trans_lock);
Yan, Zhenga22285a2010-05-16 10:48:46 -0400564 list_add(&pending_snapshot->list,
565 &trans->transaction->pending_snapshots);
Josef Bacik83515832011-06-14 15:16:14 -0400566 spin_unlock(&root->fs_info->trans_lock);
Sage Weil72fd0322010-10-29 15:41:32 -0400567 if (async_transid) {
568 *async_transid = trans->transid;
569 ret = btrfs_commit_transaction_async(trans,
570 root->fs_info->extent_root, 1);
571 } else {
572 ret = btrfs_commit_transaction(trans,
573 root->fs_info->extent_root);
574 }
Liu Bo109f2362012-11-05 12:42:09 +0000575 if (ret) {
576 /* cleanup_transaction has freed this for us */
577 if (trans->aborted)
578 pending_snapshot = NULL;
Josef Bacikc37b2b62012-10-22 15:51:44 -0400579 goto fail;
Liu Bo109f2362012-11-05 12:42:09 +0000580 }
Yan, Zhenga22285a2010-05-16 10:48:46 -0400581
582 ret = pending_snapshot->error;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400583 if (ret)
Yan, Zheng2e4bfab2009-11-12 09:37:02 +0000584 goto fail;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400585
Josef Bacik66b4ffd2011-01-31 16:22:42 -0500586 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
587 if (ret)
588 goto fail;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400589
Al Viro2fbe8c82011-07-16 21:38:06 -0400590 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
Yan, Zheng2e4bfab2009-11-12 09:37:02 +0000591 if (IS_ERR(inode)) {
592 ret = PTR_ERR(inode);
593 goto fail;
594 }
595 BUG_ON(!inode);
596 d_instantiate(dentry, inode);
597 ret = 0;
598fail:
Yan, Zhenga22285a2010-05-16 10:48:46 -0400599 kfree(pending_snapshot);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400600 return ret;
601}
602
Sage Weil4260f7c2010-10-29 15:46:43 -0400603/* copy of check_sticky in fs/namei.c()
604* It's inline, so penalty for filesystems that don't use sticky bit is
605* minimal.
606*/
607static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
608{
Eric W. Biederman2f2f43d32012-02-10 11:05:07 -0800609 kuid_t fsuid = current_fsuid();
Sage Weil4260f7c2010-10-29 15:46:43 -0400610
611 if (!(dir->i_mode & S_ISVTX))
612 return 0;
Eric W. Biederman2f2f43d32012-02-10 11:05:07 -0800613 if (uid_eq(inode->i_uid, fsuid))
Sage Weil4260f7c2010-10-29 15:46:43 -0400614 return 0;
Eric W. Biederman2f2f43d32012-02-10 11:05:07 -0800615 if (uid_eq(dir->i_uid, fsuid))
Sage Weil4260f7c2010-10-29 15:46:43 -0400616 return 0;
617 return !capable(CAP_FOWNER);
618}
619
620/* copy of may_delete in fs/namei.c()
621 * Check whether we can remove a link victim from directory dir, check
622 * whether the type of victim is right.
623 * 1. We can't do it if dir is read-only (done in permission())
624 * 2. We should have write and exec permissions on dir
625 * 3. We can't remove anything from append-only dir
626 * 4. We can't do anything with immutable dir (done in permission())
627 * 5. If the sticky bit on dir is set we should either
628 * a. be owner of dir, or
629 * b. be owner of victim, or
630 * c. have CAP_FOWNER capability
631 * 6. If the victim is append-only or immutable we can't do antyhing with
632 * links pointing to it.
633 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
634 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
635 * 9. We can't remove a root or mountpoint.
636 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
637 * nfs_async_unlink().
638 */
639
640static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
641{
642 int error;
643
644 if (!victim->d_inode)
645 return -ENOENT;
646
647 BUG_ON(victim->d_parent->d_inode != dir);
Jeff Layton4fa6b5e2012-10-10 15:25:25 -0400648 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
Sage Weil4260f7c2010-10-29 15:46:43 -0400649
650 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
651 if (error)
652 return error;
653 if (IS_APPEND(dir))
654 return -EPERM;
655 if (btrfs_check_sticky(dir, victim->d_inode)||
656 IS_APPEND(victim->d_inode)||
657 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
658 return -EPERM;
659 if (isdir) {
660 if (!S_ISDIR(victim->d_inode->i_mode))
661 return -ENOTDIR;
662 if (IS_ROOT(victim))
663 return -EBUSY;
664 } else if (S_ISDIR(victim->d_inode->i_mode))
665 return -EISDIR;
666 if (IS_DEADDIR(dir))
667 return -ENOENT;
668 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
669 return -EBUSY;
670 return 0;
671}
672
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400673/* copy of may_create in fs/namei.c() */
674static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
675{
676 if (child->d_inode)
677 return -EEXIST;
678 if (IS_DEADDIR(dir))
679 return -ENOENT;
680 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
681}
682
683/*
684 * Create a new subvolume below @parent. This is largely modeled after
685 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
686 * inside this filesystem so it's quite a bit simpler.
687 */
Yan, Zheng76dda932009-09-21 16:00:26 -0400688static noinline int btrfs_mksubvol(struct path *parent,
689 char *name, int namelen,
Sage Weil72fd0322010-10-29 15:41:32 -0400690 struct btrfs_root *snap_src,
Arne Jansen6f72c7e2011-09-14 15:58:21 +0200691 u64 *async_transid, bool readonly,
692 struct btrfs_qgroup_inherit **inherit)
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400693{
Yan, Zheng76dda932009-09-21 16:00:26 -0400694 struct inode *dir = parent->dentry->d_inode;
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400695 struct dentry *dentry;
696 int error;
697
Yan, Zheng76dda932009-09-21 16:00:26 -0400698 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400699
700 dentry = lookup_one_len(name, parent->dentry, namelen);
701 error = PTR_ERR(dentry);
702 if (IS_ERR(dentry))
703 goto out_unlock;
704
705 error = -EEXIST;
706 if (dentry->d_inode)
707 goto out_dput;
708
Yan, Zheng76dda932009-09-21 16:00:26 -0400709 error = btrfs_may_create(dir, dentry);
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400710 if (error)
Liu Boa874a632012-06-29 03:58:46 -0600711 goto out_dput;
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400712
Yan, Zheng76dda932009-09-21 16:00:26 -0400713 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
714
715 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
716 goto out_up_read;
717
Chris Mason3de45862008-11-17 21:02:50 -0500718 if (snap_src) {
Arne Jansen6f72c7e2011-09-14 15:58:21 +0200719 error = create_snapshot(snap_src, dentry, name, namelen,
720 async_transid, readonly, inherit);
Chris Mason3de45862008-11-17 21:02:50 -0500721 } else {
Yan, Zheng76dda932009-09-21 16:00:26 -0400722 error = create_subvol(BTRFS_I(dir)->root, dentry,
Arne Jansen6f72c7e2011-09-14 15:58:21 +0200723 name, namelen, async_transid, inherit);
Chris Mason3de45862008-11-17 21:02:50 -0500724 }
Yan, Zheng76dda932009-09-21 16:00:26 -0400725 if (!error)
726 fsnotify_mkdir(dir, dentry);
727out_up_read:
728 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400729out_dput:
730 dput(dentry);
731out_unlock:
Yan, Zheng76dda932009-09-21 16:00:26 -0400732 mutex_unlock(&dir->i_mutex);
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400733 return error;
734}
735
Chris Mason4cb53002011-05-24 15:35:30 -0400736/*
737 * When we're defragging a range, we don't want to kick it off again
738 * if it is really just waiting for delalloc to send it down.
739 * If we find a nice big extent or delalloc range for the bytes in the
740 * file you want to defrag, we return 0 to let you know to skip this
741 * part of the file
742 */
743static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
744{
745 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
746 struct extent_map *em = NULL;
747 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
748 u64 end;
749
750 read_lock(&em_tree->lock);
751 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
752 read_unlock(&em_tree->lock);
753
754 if (em) {
755 end = extent_map_end(em);
756 free_extent_map(em);
757 if (end - offset > thresh)
758 return 0;
759 }
760 /* if we already have a nice delalloc here, just stop */
761 thresh /= 2;
762 end = count_range_bits(io_tree, &offset, offset + thresh,
763 thresh, EXTENT_DELALLOC, 1);
764 if (end >= thresh)
765 return 0;
766 return 1;
767}
768
769/*
770 * helper function to walk through a file and find extents
771 * newer than a specific transid, and smaller than thresh.
772 *
773 * This is used by the defragging code to find new and small
774 * extents
775 */
776static int find_new_extents(struct btrfs_root *root,
777 struct inode *inode, u64 newer_than,
778 u64 *off, int thresh)
779{
780 struct btrfs_path *path;
781 struct btrfs_key min_key;
782 struct btrfs_key max_key;
783 struct extent_buffer *leaf;
784 struct btrfs_file_extent_item *extent;
785 int type;
786 int ret;
David Sterbaa4689d22011-05-31 17:08:14 +0000787 u64 ino = btrfs_ino(inode);
Chris Mason4cb53002011-05-24 15:35:30 -0400788
789 path = btrfs_alloc_path();
790 if (!path)
791 return -ENOMEM;
792
David Sterbaa4689d22011-05-31 17:08:14 +0000793 min_key.objectid = ino;
Chris Mason4cb53002011-05-24 15:35:30 -0400794 min_key.type = BTRFS_EXTENT_DATA_KEY;
795 min_key.offset = *off;
796
David Sterbaa4689d22011-05-31 17:08:14 +0000797 max_key.objectid = ino;
Chris Mason4cb53002011-05-24 15:35:30 -0400798 max_key.type = (u8)-1;
799 max_key.offset = (u64)-1;
800
801 path->keep_locks = 1;
802
803 while(1) {
804 ret = btrfs_search_forward(root, &min_key, &max_key,
805 path, 0, newer_than);
806 if (ret != 0)
807 goto none;
David Sterbaa4689d22011-05-31 17:08:14 +0000808 if (min_key.objectid != ino)
Chris Mason4cb53002011-05-24 15:35:30 -0400809 goto none;
810 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
811 goto none;
812
813 leaf = path->nodes[0];
814 extent = btrfs_item_ptr(leaf, path->slots[0],
815 struct btrfs_file_extent_item);
816
817 type = btrfs_file_extent_type(leaf, extent);
818 if (type == BTRFS_FILE_EXTENT_REG &&
819 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
820 check_defrag_in_cache(inode, min_key.offset, thresh)) {
821 *off = min_key.offset;
822 btrfs_free_path(path);
823 return 0;
824 }
825
826 if (min_key.offset == (u64)-1)
827 goto none;
828
829 min_key.offset++;
830 btrfs_release_path(path);
831 }
832none:
833 btrfs_free_path(path);
834 return -ENOENT;
835}
836
Li Zefan6c282eb2012-06-11 16:03:35 +0800837static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
Liu Bo17ce6ef2012-03-29 09:57:45 -0400838{
839 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
Chris Mason940100a2010-03-10 10:52:59 -0500840 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
Li Zefan6c282eb2012-06-11 16:03:35 +0800841 struct extent_map *em;
842 u64 len = PAGE_CACHE_SIZE;
Chris Mason940100a2010-03-10 10:52:59 -0500843
844 /*
845 * hopefully we have this extent in the tree already, try without
846 * the full extent lock
847 */
848 read_lock(&em_tree->lock);
849 em = lookup_extent_mapping(em_tree, start, len);
850 read_unlock(&em_tree->lock);
851
852 if (!em) {
853 /* get the big lock and read metadata off disk */
Jeff Mahoneyd0082372012-03-01 14:57:19 +0100854 lock_extent(io_tree, start, start + len - 1);
Chris Mason940100a2010-03-10 10:52:59 -0500855 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
Jeff Mahoneyd0082372012-03-01 14:57:19 +0100856 unlock_extent(io_tree, start, start + len - 1);
Chris Mason940100a2010-03-10 10:52:59 -0500857
Dan Carpenter6cf8bfb2010-03-20 11:22:10 +0000858 if (IS_ERR(em))
Li Zefan6c282eb2012-06-11 16:03:35 +0800859 return NULL;
Chris Mason940100a2010-03-10 10:52:59 -0500860 }
861
Li Zefan6c282eb2012-06-11 16:03:35 +0800862 return em;
863}
864
865static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
866{
867 struct extent_map *next;
868 bool ret = true;
869
870 /* this is the last extent */
871 if (em->start + em->len >= i_size_read(inode))
872 return false;
873
874 next = defrag_lookup_extent(inode, em->start + em->len);
875 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
876 ret = false;
877
878 free_extent_map(next);
879 return ret;
880}
881
882static int should_defrag_range(struct inode *inode, u64 start, int thresh,
Andrew Mahonea43a2112012-06-19 21:08:32 -0400883 u64 *last_len, u64 *skip, u64 *defrag_end,
884 int compress)
Li Zefan6c282eb2012-06-11 16:03:35 +0800885{
886 struct extent_map *em;
887 int ret = 1;
888 bool next_mergeable = true;
889
890 /*
891 * make sure that once we start defragging an extent, we keep on
892 * defragging it
893 */
894 if (start < *defrag_end)
895 return 1;
896
897 *skip = 0;
898
899 em = defrag_lookup_extent(inode, start);
900 if (!em)
901 return 0;
902
Chris Mason940100a2010-03-10 10:52:59 -0500903 /* this will cover holes, and inline extents */
Liu Bo17ce6ef2012-03-29 09:57:45 -0400904 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
Chris Mason940100a2010-03-10 10:52:59 -0500905 ret = 0;
Liu Bo17ce6ef2012-03-29 09:57:45 -0400906 goto out;
907 }
908
Li Zefan6c282eb2012-06-11 16:03:35 +0800909 next_mergeable = defrag_check_next_extent(inode, em);
Chris Mason940100a2010-03-10 10:52:59 -0500910
911 /*
Li Zefan6c282eb2012-06-11 16:03:35 +0800912 * we hit a real extent, if it is big or the next extent is not a
913 * real extent, don't bother defragging it
Chris Mason940100a2010-03-10 10:52:59 -0500914 */
Andrew Mahonea43a2112012-06-19 21:08:32 -0400915 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
Li Zefan6c282eb2012-06-11 16:03:35 +0800916 (em->len >= thresh || !next_mergeable))
Chris Mason940100a2010-03-10 10:52:59 -0500917 ret = 0;
Liu Bo17ce6ef2012-03-29 09:57:45 -0400918out:
Chris Mason940100a2010-03-10 10:52:59 -0500919 /*
920 * last_len ends up being a counter of how many bytes we've defragged.
921 * every time we choose not to defrag an extent, we reset *last_len
922 * so that the next tiny extent will force a defrag.
923 *
924 * The end result of this is that tiny extents before a single big
925 * extent will force at least part of that big extent to be defragged.
926 */
927 if (ret) {
Chris Mason940100a2010-03-10 10:52:59 -0500928 *defrag_end = extent_map_end(em);
929 } else {
930 *last_len = 0;
931 *skip = extent_map_end(em);
932 *defrag_end = 0;
933 }
934
935 free_extent_map(em);
936 return ret;
937}
938
Chris Mason4cb53002011-05-24 15:35:30 -0400939/*
940 * it doesn't do much good to defrag one or two pages
941 * at a time. This pulls in a nice chunk of pages
942 * to COW and defrag.
943 *
944 * It also makes sure the delalloc code has enough
945 * dirty data to avoid making new small extents as part
946 * of the defrag
947 *
948 * It's a good idea to start RA on this range
949 * before calling this.
950 */
951static int cluster_pages_for_defrag(struct inode *inode,
952 struct page **pages,
953 unsigned long start_index,
954 int num_pages)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400955{
Chris Mason4cb53002011-05-24 15:35:30 -0400956 unsigned long file_end;
957 u64 isize = i_size_read(inode);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400958 u64 page_start;
959 u64 page_end;
Liu Bo1f12bd02012-03-29 09:57:44 -0400960 u64 page_cnt;
Chris Mason4cb53002011-05-24 15:35:30 -0400961 int ret;
962 int i;
963 int i_done;
964 struct btrfs_ordered_extent *ordered;
965 struct extent_state *cached_state = NULL;
Miao Xie600a45e2012-02-16 15:01:24 +0800966 struct extent_io_tree *tree;
Josef Bacik3b16a4e2011-09-21 15:05:58 -0400967 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
Chris Mason4cb53002011-05-24 15:35:30 -0400968
Chris Mason4cb53002011-05-24 15:35:30 -0400969 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
Liu Bo1f12bd02012-03-29 09:57:44 -0400970 if (!isize || start_index > file_end)
971 return 0;
972
973 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
Chris Mason4cb53002011-05-24 15:35:30 -0400974
975 ret = btrfs_delalloc_reserve_space(inode,
Liu Bo1f12bd02012-03-29 09:57:44 -0400976 page_cnt << PAGE_CACHE_SHIFT);
Chris Mason4cb53002011-05-24 15:35:30 -0400977 if (ret)
978 return ret;
Chris Mason4cb53002011-05-24 15:35:30 -0400979 i_done = 0;
Miao Xie600a45e2012-02-16 15:01:24 +0800980 tree = &BTRFS_I(inode)->io_tree;
Chris Mason4cb53002011-05-24 15:35:30 -0400981
982 /* step one, lock all the pages */
Liu Bo1f12bd02012-03-29 09:57:44 -0400983 for (i = 0; i < page_cnt; i++) {
Chris Mason4cb53002011-05-24 15:35:30 -0400984 struct page *page;
Miao Xie600a45e2012-02-16 15:01:24 +0800985again:
Josef Bacika94733d2011-07-11 10:47:06 -0400986 page = find_or_create_page(inode->i_mapping,
Miao Xie600a45e2012-02-16 15:01:24 +0800987 start_index + i, mask);
Chris Mason4cb53002011-05-24 15:35:30 -0400988 if (!page)
989 break;
990
Miao Xie600a45e2012-02-16 15:01:24 +0800991 page_start = page_offset(page);
992 page_end = page_start + PAGE_CACHE_SIZE - 1;
993 while (1) {
Jeff Mahoneyd0082372012-03-01 14:57:19 +0100994 lock_extent(tree, page_start, page_end);
Miao Xie600a45e2012-02-16 15:01:24 +0800995 ordered = btrfs_lookup_ordered_extent(inode,
996 page_start);
Jeff Mahoneyd0082372012-03-01 14:57:19 +0100997 unlock_extent(tree, page_start, page_end);
Miao Xie600a45e2012-02-16 15:01:24 +0800998 if (!ordered)
999 break;
1000
1001 unlock_page(page);
1002 btrfs_start_ordered_extent(inode, ordered, 1);
1003 btrfs_put_ordered_extent(ordered);
1004 lock_page(page);
Liu Bo1f12bd02012-03-29 09:57:44 -04001005 /*
1006 * we unlocked the page above, so we need check if
1007 * it was released or not.
1008 */
1009 if (page->mapping != inode->i_mapping) {
1010 unlock_page(page);
1011 page_cache_release(page);
1012 goto again;
1013 }
Miao Xie600a45e2012-02-16 15:01:24 +08001014 }
1015
Chris Mason4cb53002011-05-24 15:35:30 -04001016 if (!PageUptodate(page)) {
1017 btrfs_readpage(NULL, page);
1018 lock_page(page);
1019 if (!PageUptodate(page)) {
1020 unlock_page(page);
1021 page_cache_release(page);
1022 ret = -EIO;
1023 break;
1024 }
1025 }
Miao Xie600a45e2012-02-16 15:01:24 +08001026
Miao Xie600a45e2012-02-16 15:01:24 +08001027 if (page->mapping != inode->i_mapping) {
1028 unlock_page(page);
1029 page_cache_release(page);
1030 goto again;
1031 }
1032
Chris Mason4cb53002011-05-24 15:35:30 -04001033 pages[i] = page;
1034 i_done++;
1035 }
1036 if (!i_done || ret)
1037 goto out;
1038
1039 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1040 goto out;
1041
1042 /*
1043 * so now we have a nice long stream of locked
1044 * and up to date pages, lets wait on them
1045 */
1046 for (i = 0; i < i_done; i++)
1047 wait_on_page_writeback(pages[i]);
1048
1049 page_start = page_offset(pages[0]);
1050 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1051
1052 lock_extent_bits(&BTRFS_I(inode)->io_tree,
Jeff Mahoneyd0082372012-03-01 14:57:19 +01001053 page_start, page_end - 1, 0, &cached_state);
Chris Mason4cb53002011-05-24 15:35:30 -04001054 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1055 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
Liu Bo9e8a4a82012-09-05 19:10:51 -06001056 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1057 &cached_state, GFP_NOFS);
Chris Mason4cb53002011-05-24 15:35:30 -04001058
Liu Bo1f12bd02012-03-29 09:57:44 -04001059 if (i_done != page_cnt) {
Josef Bacik9e0baf62011-07-15 15:16:44 +00001060 spin_lock(&BTRFS_I(inode)->lock);
1061 BTRFS_I(inode)->outstanding_extents++;
1062 spin_unlock(&BTRFS_I(inode)->lock);
Chris Mason4cb53002011-05-24 15:35:30 -04001063 btrfs_delalloc_release_space(inode,
Liu Bo1f12bd02012-03-29 09:57:44 -04001064 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
Chris Mason4cb53002011-05-24 15:35:30 -04001065 }
1066
1067
Liu Bo9e8a4a82012-09-05 19:10:51 -06001068 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1069 &cached_state, GFP_NOFS);
Chris Mason4cb53002011-05-24 15:35:30 -04001070
1071 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1072 page_start, page_end - 1, &cached_state,
1073 GFP_NOFS);
1074
1075 for (i = 0; i < i_done; i++) {
1076 clear_page_dirty_for_io(pages[i]);
1077 ClearPageChecked(pages[i]);
1078 set_page_extent_mapped(pages[i]);
1079 set_page_dirty(pages[i]);
1080 unlock_page(pages[i]);
1081 page_cache_release(pages[i]);
1082 }
1083 return i_done;
1084out:
1085 for (i = 0; i < i_done; i++) {
1086 unlock_page(pages[i]);
1087 page_cache_release(pages[i]);
1088 }
Liu Bo1f12bd02012-03-29 09:57:44 -04001089 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
Chris Mason4cb53002011-05-24 15:35:30 -04001090 return ret;
1091
1092}
1093
1094int btrfs_defrag_file(struct inode *inode, struct file *file,
1095 struct btrfs_ioctl_defrag_range_args *range,
1096 u64 newer_than, unsigned long max_to_defrag)
1097{
1098 struct btrfs_root *root = BTRFS_I(inode)->root;
Chris Mason4cb53002011-05-24 15:35:30 -04001099 struct file_ra_state *ra = NULL;
1100 unsigned long last_index;
Li Zefan151a31b2011-09-02 15:56:39 +08001101 u64 isize = i_size_read(inode);
Chris Mason940100a2010-03-10 10:52:59 -05001102 u64 last_len = 0;
1103 u64 skip = 0;
1104 u64 defrag_end = 0;
Chris Mason4cb53002011-05-24 15:35:30 -04001105 u64 newer_off = range->start;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001106 unsigned long i;
Li Zefan008873e2011-09-02 15:57:07 +08001107 unsigned long ra_index = 0;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001108 int ret;
Chris Mason4cb53002011-05-24 15:35:30 -04001109 int defrag_count = 0;
Li Zefan1a419d82010-10-25 15:12:50 +08001110 int compress_type = BTRFS_COMPRESS_ZLIB;
Chris Mason4cb53002011-05-24 15:35:30 -04001111 int extent_thresh = range->extent_thresh;
Li Zefan008873e2011-09-02 15:57:07 +08001112 int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1113 int cluster = max_cluster;
Chris Mason4cb53002011-05-24 15:35:30 -04001114 u64 new_align = ~((u64)128 * 1024 - 1);
1115 struct page **pages = NULL;
1116
1117 if (extent_thresh == 0)
1118 extent_thresh = 256 * 1024;
Li Zefan1a419d82010-10-25 15:12:50 +08001119
1120 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1121 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1122 return -EINVAL;
1123 if (range->compress_type)
1124 compress_type = range->compress_type;
1125 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001126
Li Zefan151a31b2011-09-02 15:56:39 +08001127 if (isize == 0)
Chris Mason940100a2010-03-10 10:52:59 -05001128 return 0;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001129
Chris Mason4cb53002011-05-24 15:35:30 -04001130 /*
1131 * if we were not given a file, allocate a readahead
1132 * context
1133 */
1134 if (!file) {
1135 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1136 if (!ra)
1137 return -ENOMEM;
1138 file_ra_state_init(ra, inode->i_mapping);
1139 } else {
1140 ra = &file->f_ra;
1141 }
1142
Li Zefan008873e2011-09-02 15:57:07 +08001143 pages = kmalloc(sizeof(struct page *) * max_cluster,
Chris Mason4cb53002011-05-24 15:35:30 -04001144 GFP_NOFS);
1145 if (!pages) {
1146 ret = -ENOMEM;
1147 goto out_ra;
1148 }
1149
1150 /* find the last page to defrag */
Chris Mason1e701a32010-03-11 09:42:04 -05001151 if (range->start + range->len > range->start) {
Li Zefan151a31b2011-09-02 15:56:39 +08001152 last_index = min_t(u64, isize - 1,
Chris Mason1e701a32010-03-11 09:42:04 -05001153 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1154 } else {
Li Zefan151a31b2011-09-02 15:56:39 +08001155 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
Chris Mason1e701a32010-03-11 09:42:04 -05001156 }
1157
Chris Mason4cb53002011-05-24 15:35:30 -04001158 if (newer_than) {
1159 ret = find_new_extents(root, inode, newer_than,
1160 &newer_off, 64 * 1024);
1161 if (!ret) {
1162 range->start = newer_off;
1163 /*
1164 * we always align our defrag to help keep
1165 * the extents in the file evenly spaced
1166 */
1167 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
Chris Mason4cb53002011-05-24 15:35:30 -04001168 } else
1169 goto out_ra;
1170 } else {
1171 i = range->start >> PAGE_CACHE_SHIFT;
1172 }
1173 if (!max_to_defrag)
Liu Bo7ec31b52012-01-26 15:01:12 -05001174 max_to_defrag = last_index + 1;
Chris Mason4cb53002011-05-24 15:35:30 -04001175
Li Zefan2a0f7f52011-10-10 15:43:34 -04001176 /*
1177 * make writeback starts from i, so the defrag range can be
1178 * written sequentially.
1179 */
1180 if (i < inode->i_mapping->writeback_index)
1181 inode->i_mapping->writeback_index = i;
1182
Chris Masonf7f43cc2011-10-11 11:41:40 -04001183 while (i <= last_index && defrag_count < max_to_defrag &&
1184 (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1185 PAGE_CACHE_SHIFT)) {
Chris Mason4cb53002011-05-24 15:35:30 -04001186 /*
1187 * make sure we stop running if someone unmounts
1188 * the FS
1189 */
1190 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1191 break;
1192
Liu Bo66c26892012-03-29 09:57:45 -04001193 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
Li Zefan6c282eb2012-06-11 16:03:35 +08001194 extent_thresh, &last_len, &skip,
Andrew Mahonea43a2112012-06-19 21:08:32 -04001195 &defrag_end, range->flags &
1196 BTRFS_DEFRAG_RANGE_COMPRESS)) {
Chris Mason940100a2010-03-10 10:52:59 -05001197 unsigned long next;
1198 /*
1199 * the should_defrag function tells us how much to skip
1200 * bump our counter by the suggested amount
1201 */
1202 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1203 i = max(i + 1, next);
1204 continue;
1205 }
Li Zefan008873e2011-09-02 15:57:07 +08001206
1207 if (!newer_than) {
1208 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1209 PAGE_CACHE_SHIFT) - i;
1210 cluster = min(cluster, max_cluster);
1211 } else {
1212 cluster = max_cluster;
1213 }
1214
Chris Mason1e701a32010-03-11 09:42:04 -05001215 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
Li Zefan1a419d82010-10-25 15:12:50 +08001216 BTRFS_I(inode)->force_compress = compress_type;
Chris Mason940100a2010-03-10 10:52:59 -05001217
Li Zefan008873e2011-09-02 15:57:07 +08001218 if (i + cluster > ra_index) {
1219 ra_index = max(i, ra_index);
1220 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1221 cluster);
1222 ra_index += max_cluster;
1223 }
Chris Mason940100a2010-03-10 10:52:59 -05001224
Liu Boecb8bea2012-03-29 09:57:44 -04001225 mutex_lock(&inode->i_mutex);
Li Zefan008873e2011-09-02 15:57:07 +08001226 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
Liu Boecb8bea2012-03-29 09:57:44 -04001227 if (ret < 0) {
1228 mutex_unlock(&inode->i_mutex);
Chris Mason4cb53002011-05-24 15:35:30 -04001229 goto out_ra;
Liu Boecb8bea2012-03-29 09:57:44 -04001230 }
Chris Mason940100a2010-03-10 10:52:59 -05001231
Chris Mason4cb53002011-05-24 15:35:30 -04001232 defrag_count += ret;
1233 balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret);
Liu Boecb8bea2012-03-29 09:57:44 -04001234 mutex_unlock(&inode->i_mutex);
Chris Mason4cb53002011-05-24 15:35:30 -04001235
1236 if (newer_than) {
1237 if (newer_off == (u64)-1)
1238 break;
1239
Liu Boe1f041e2012-03-29 09:57:45 -04001240 if (ret > 0)
1241 i += ret;
1242
Chris Mason4cb53002011-05-24 15:35:30 -04001243 newer_off = max(newer_off + 1,
1244 (u64)i << PAGE_CACHE_SHIFT);
1245
1246 ret = find_new_extents(root, inode,
1247 newer_than, &newer_off,
1248 64 * 1024);
1249 if (!ret) {
1250 range->start = newer_off;
1251 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
Chris Mason4cb53002011-05-24 15:35:30 -04001252 } else {
1253 break;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001254 }
Chris Mason4cb53002011-05-24 15:35:30 -04001255 } else {
Li Zefan008873e2011-09-02 15:57:07 +08001256 if (ret > 0) {
Li Zefancbcc8322011-09-02 15:56:25 +08001257 i += ret;
Li Zefan008873e2011-09-02 15:57:07 +08001258 last_len += ret << PAGE_CACHE_SHIFT;
1259 } else {
Li Zefancbcc8322011-09-02 15:56:25 +08001260 i++;
Li Zefan008873e2011-09-02 15:57:07 +08001261 last_len = 0;
1262 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001263 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001264 }
1265
Chris Mason1e701a32010-03-11 09:42:04 -05001266 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1267 filemap_flush(inode->i_mapping);
1268
1269 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1270 /* the filemap_flush will queue IO into the worker threads, but
1271 * we have to make sure the IO is actually started and that
1272 * ordered extents get created before we return
1273 */
1274 atomic_inc(&root->fs_info->async_submit_draining);
1275 while (atomic_read(&root->fs_info->nr_async_submits) ||
1276 atomic_read(&root->fs_info->async_delalloc_pages)) {
1277 wait_event(root->fs_info->async_submit_wait,
1278 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1279 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1280 }
1281 atomic_dec(&root->fs_info->async_submit_draining);
1282
1283 mutex_lock(&inode->i_mutex);
Li Zefan261507a02010-12-17 14:21:50 +08001284 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
Chris Mason1e701a32010-03-11 09:42:04 -05001285 mutex_unlock(&inode->i_mutex);
1286 }
1287
Li Zefan1a419d82010-10-25 15:12:50 +08001288 if (range->compress_type == BTRFS_COMPRESS_LZO) {
Mitch Harder2b0ce2c2012-07-24 11:58:43 -06001289 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
Li Zefan1a419d82010-10-25 15:12:50 +08001290 }
1291
Diego Calleja60ccf822011-09-01 16:33:57 +02001292 ret = defrag_count;
Chris Mason940100a2010-03-10 10:52:59 -05001293
Chris Mason4cb53002011-05-24 15:35:30 -04001294out_ra:
1295 if (!file)
1296 kfree(ra);
1297 kfree(pages);
Chris Mason940100a2010-03-10 10:52:59 -05001298 return ret;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001299}
1300
Miao Xie198605a2012-11-26 08:43:45 +00001301static noinline int btrfs_ioctl_resize(struct file *file,
Yan, Zheng76dda932009-09-21 16:00:26 -04001302 void __user *arg)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001303{
1304 u64 new_size;
1305 u64 old_size;
1306 u64 devid = 1;
Miao Xie198605a2012-11-26 08:43:45 +00001307 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001308 struct btrfs_ioctl_vol_args *vol_args;
1309 struct btrfs_trans_handle *trans;
1310 struct btrfs_device *device = NULL;
1311 char *sizestr;
1312 char *devstr = NULL;
1313 int ret = 0;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001314 int mod = 0;
1315
Yan Zhengc146afa2008-11-12 14:34:12 -05001316 if (root->fs_info->sb->s_flags & MS_RDONLY)
1317 return -EROFS;
1318
Chris Masone441d542009-01-05 16:57:23 -05001319 if (!capable(CAP_SYS_ADMIN))
1320 return -EPERM;
1321
Miao Xie198605a2012-11-26 08:43:45 +00001322 ret = mnt_want_write_file(file);
1323 if (ret)
1324 return ret;
1325
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01001326 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1327 1)) {
1328 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
1329 return -EINPROGRESS;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001330 }
1331
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01001332 mutex_lock(&root->fs_info->volume_mutex);
Li Zefandae7b662009-04-08 15:06:54 +08001333 vol_args = memdup_user(arg, sizeof(*vol_args));
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001334 if (IS_ERR(vol_args)) {
1335 ret = PTR_ERR(vol_args);
1336 goto out;
1337 }
Mark Fasheh5516e592008-07-24 12:20:14 -04001338
1339 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001340
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001341 sizestr = vol_args->name;
1342 devstr = strchr(sizestr, ':');
1343 if (devstr) {
1344 char *end;
1345 sizestr = devstr + 1;
1346 *devstr = '\0';
1347 devstr = vol_args->name;
1348 devid = simple_strtoull(devstr, &end, 10);
Arnd Hannemann5bb14682011-11-20 07:33:38 -05001349 printk(KERN_INFO "btrfs: resizing devid %llu\n",
Joel Becker21380932009-04-21 12:38:29 -07001350 (unsigned long long)devid);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001351 }
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01001352 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001353 if (!device) {
Arnd Hannemann5bb14682011-11-20 07:33:38 -05001354 printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
Joel Becker21380932009-04-21 12:38:29 -07001355 (unsigned long long)devid);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001356 ret = -EINVAL;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001357 goto out_free;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001358 }
Liu Bo4e42ae12012-06-14 02:23:19 -06001359 if (device->fs_devices && device->fs_devices->seeding) {
1360 printk(KERN_INFO "btrfs: resizer unable to apply on "
Chris Masona8c4a332012-06-15 20:07:17 -04001361 "seeding device %llu\n",
1362 (unsigned long long)devid);
Liu Bo4e42ae12012-06-14 02:23:19 -06001363 ret = -EINVAL;
1364 goto out_free;
1365 }
1366
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001367 if (!strcmp(sizestr, "max"))
1368 new_size = device->bdev->bd_inode->i_size;
1369 else {
1370 if (sizestr[0] == '-') {
1371 mod = -1;
1372 sizestr++;
1373 } else if (sizestr[0] == '+') {
1374 mod = 1;
1375 sizestr++;
1376 }
Akinobu Mita91748462010-02-28 10:59:11 +00001377 new_size = memparse(sizestr, NULL);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001378 if (new_size == 0) {
1379 ret = -EINVAL;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001380 goto out_free;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001381 }
1382 }
1383
Stefan Behrens63a212a2012-11-05 18:29:28 +01001384 if (device->is_tgtdev_for_dev_replace) {
1385 ret = -EINVAL;
1386 goto out_free;
1387 }
1388
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001389 old_size = device->total_bytes;
1390
1391 if (mod < 0) {
1392 if (new_size > old_size) {
1393 ret = -EINVAL;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001394 goto out_free;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001395 }
1396 new_size = old_size - new_size;
1397 } else if (mod > 0) {
1398 new_size = old_size + new_size;
1399 }
1400
1401 if (new_size < 256 * 1024 * 1024) {
1402 ret = -EINVAL;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001403 goto out_free;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001404 }
1405 if (new_size > device->bdev->bd_inode->i_size) {
1406 ret = -EFBIG;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001407 goto out_free;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001408 }
1409
1410 do_div(new_size, root->sectorsize);
1411 new_size *= root->sectorsize;
1412
Josef Bacik606686e2012-06-04 14:03:51 -04001413 printk_in_rcu(KERN_INFO "btrfs: new size for %s is %llu\n",
1414 rcu_str_deref(device->name),
1415 (unsigned long long)new_size);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001416
1417 if (new_size > old_size) {
Yan, Zhenga22285a2010-05-16 10:48:46 -04001418 trans = btrfs_start_transaction(root, 0);
Tsutomu Itoh98d5dc12011-01-20 06:19:37 +00001419 if (IS_ERR(trans)) {
1420 ret = PTR_ERR(trans);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001421 goto out_free;
Tsutomu Itoh98d5dc12011-01-20 06:19:37 +00001422 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001423 ret = btrfs_grow_device(trans, device, new_size);
1424 btrfs_commit_transaction(trans, root);
Mike Fleetwoodece7d202011-11-18 18:55:01 +00001425 } else if (new_size < old_size) {
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001426 ret = btrfs_shrink_device(device, new_size);
jeff.liu0253f402012-10-27 12:06:39 +00001427 } /* equal, nothing need to do */
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001428
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001429out_free:
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001430 kfree(vol_args);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001431out:
1432 mutex_unlock(&root->fs_info->volume_mutex);
Miao Xie198605a2012-11-26 08:43:45 +00001433 mnt_drop_write_file(file);
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01001434 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001435 return ret;
1436}
1437
Sage Weil72fd0322010-10-29 15:41:32 -04001438static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001439 char *name, unsigned long fd, int subvol,
1440 u64 *transid, bool readonly,
1441 struct btrfs_qgroup_inherit **inherit)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001442{
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001443 int namelen;
Chris Mason3de45862008-11-17 21:02:50 -05001444 int ret = 0;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001445
Liu Boa874a632012-06-29 03:58:46 -06001446 ret = mnt_want_write_file(file);
1447 if (ret)
1448 goto out;
1449
Sage Weil72fd0322010-10-29 15:41:32 -04001450 namelen = strlen(name);
1451 if (strchr(name, '/')) {
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001452 ret = -EINVAL;
Liu Boa874a632012-06-29 03:58:46 -06001453 goto out_drop_write;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001454 }
1455
Chris Mason16780ca2012-02-20 22:14:55 -05001456 if (name[0] == '.' &&
1457 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1458 ret = -EEXIST;
Liu Boa874a632012-06-29 03:58:46 -06001459 goto out_drop_write;
Chris Mason16780ca2012-02-20 22:14:55 -05001460 }
1461
Chris Mason3de45862008-11-17 21:02:50 -05001462 if (subvol) {
Sage Weil72fd0322010-10-29 15:41:32 -04001463 ret = btrfs_mksubvol(&file->f_path, name, namelen,
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001464 NULL, transid, readonly, inherit);
Christoph Hellwigcb8e7092008-10-09 13:39:39 -04001465 } else {
Al Viro2903ff02012-08-28 12:52:22 -04001466 struct fd src = fdget(fd);
Chris Mason3de45862008-11-17 21:02:50 -05001467 struct inode *src_inode;
Al Viro2903ff02012-08-28 12:52:22 -04001468 if (!src.file) {
Chris Mason3de45862008-11-17 21:02:50 -05001469 ret = -EINVAL;
Liu Boa874a632012-06-29 03:58:46 -06001470 goto out_drop_write;
Chris Mason3de45862008-11-17 21:02:50 -05001471 }
1472
Al Viro2903ff02012-08-28 12:52:22 -04001473 src_inode = src.file->f_path.dentry->d_inode;
Chris Mason3de45862008-11-17 21:02:50 -05001474 if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
Chris Masond3977122009-01-05 21:25:51 -05001475 printk(KERN_INFO "btrfs: Snapshot src from "
1476 "another FS\n");
Chris Mason3de45862008-11-17 21:02:50 -05001477 ret = -EINVAL;
Al Viroecd18812012-08-26 21:20:24 -04001478 } else {
1479 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1480 BTRFS_I(src_inode)->root,
1481 transid, readonly, inherit);
Chris Mason3de45862008-11-17 21:02:50 -05001482 }
Al Viro2903ff02012-08-28 12:52:22 -04001483 fdput(src);
Christoph Hellwigcb8e7092008-10-09 13:39:39 -04001484 }
Liu Boa874a632012-06-29 03:58:46 -06001485out_drop_write:
1486 mnt_drop_write_file(file);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001487out:
Sage Weil72fd0322010-10-29 15:41:32 -04001488 return ret;
1489}
1490
1491static noinline int btrfs_ioctl_snap_create(struct file *file,
Li Zefanfa0d2b92010-12-20 15:53:28 +08001492 void __user *arg, int subvol)
Sage Weil72fd0322010-10-29 15:41:32 -04001493{
Li Zefanfa0d2b92010-12-20 15:53:28 +08001494 struct btrfs_ioctl_vol_args *vol_args;
Sage Weil72fd0322010-10-29 15:41:32 -04001495 int ret;
1496
Li Zefanfa0d2b92010-12-20 15:53:28 +08001497 vol_args = memdup_user(arg, sizeof(*vol_args));
1498 if (IS_ERR(vol_args))
1499 return PTR_ERR(vol_args);
1500 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
Sage Weil72fd0322010-10-29 15:41:32 -04001501
Li Zefanfa0d2b92010-12-20 15:53:28 +08001502 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
Li Zefanb83cc962010-12-20 16:04:08 +08001503 vol_args->fd, subvol,
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001504 NULL, false, NULL);
Li Zefanfdfb1e42010-12-10 06:41:56 +00001505
Li Zefanfa0d2b92010-12-20 15:53:28 +08001506 kfree(vol_args);
1507 return ret;
1508}
Li Zefanfdfb1e42010-12-10 06:41:56 +00001509
Li Zefanfa0d2b92010-12-20 15:53:28 +08001510static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1511 void __user *arg, int subvol)
1512{
1513 struct btrfs_ioctl_vol_args_v2 *vol_args;
1514 int ret;
1515 u64 transid = 0;
1516 u64 *ptr = NULL;
Li Zefanb83cc962010-12-20 16:04:08 +08001517 bool readonly = false;
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001518 struct btrfs_qgroup_inherit *inherit = NULL;
Li Zefanfdfb1e42010-12-10 06:41:56 +00001519
Li Zefanfa0d2b92010-12-20 15:53:28 +08001520 vol_args = memdup_user(arg, sizeof(*vol_args));
1521 if (IS_ERR(vol_args))
1522 return PTR_ERR(vol_args);
1523 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
Sage Weil75eaa0e2010-12-10 00:36:28 +00001524
Li Zefanb83cc962010-12-20 16:04:08 +08001525 if (vol_args->flags &
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001526 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1527 BTRFS_SUBVOL_QGROUP_INHERIT)) {
Li Zefanb83cc962010-12-20 16:04:08 +08001528 ret = -EOPNOTSUPP;
Li Zefanfa0d2b92010-12-20 15:53:28 +08001529 goto out;
Sage Weil72fd0322010-10-29 15:41:32 -04001530 }
Li Zefanfa0d2b92010-12-20 15:53:28 +08001531
1532 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1533 ptr = &transid;
Li Zefanb83cc962010-12-20 16:04:08 +08001534 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1535 readonly = true;
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001536 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1537 if (vol_args->size > PAGE_CACHE_SIZE) {
1538 ret = -EINVAL;
1539 goto out;
1540 }
1541 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1542 if (IS_ERR(inherit)) {
1543 ret = PTR_ERR(inherit);
1544 goto out;
1545 }
1546 }
Li Zefanfa0d2b92010-12-20 15:53:28 +08001547
1548 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001549 vol_args->fd, subvol, ptr,
1550 readonly, &inherit);
Li Zefanfa0d2b92010-12-20 15:53:28 +08001551
1552 if (ret == 0 && ptr &&
1553 copy_to_user(arg +
1554 offsetof(struct btrfs_ioctl_vol_args_v2,
1555 transid), ptr, sizeof(*ptr)))
1556 ret = -EFAULT;
Li Zefanfdfb1e42010-12-10 06:41:56 +00001557out:
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001558 kfree(vol_args);
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001559 kfree(inherit);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001560 return ret;
1561}
1562
Li Zefan0caa1022010-12-20 16:30:25 +08001563static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1564 void __user *arg)
1565{
1566 struct inode *inode = fdentry(file)->d_inode;
1567 struct btrfs_root *root = BTRFS_I(inode)->root;
1568 int ret = 0;
1569 u64 flags = 0;
1570
Li Zefan33345d012011-04-20 10:31:50 +08001571 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
Li Zefan0caa1022010-12-20 16:30:25 +08001572 return -EINVAL;
1573
1574 down_read(&root->fs_info->subvol_sem);
1575 if (btrfs_root_readonly(root))
1576 flags |= BTRFS_SUBVOL_RDONLY;
1577 up_read(&root->fs_info->subvol_sem);
1578
1579 if (copy_to_user(arg, &flags, sizeof(flags)))
1580 ret = -EFAULT;
1581
1582 return ret;
1583}
1584
1585static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1586 void __user *arg)
1587{
1588 struct inode *inode = fdentry(file)->d_inode;
1589 struct btrfs_root *root = BTRFS_I(inode)->root;
1590 struct btrfs_trans_handle *trans;
1591 u64 root_flags;
1592 u64 flags;
1593 int ret = 0;
1594
Liu Bob9ca0662012-06-29 03:58:49 -06001595 ret = mnt_want_write_file(file);
1596 if (ret)
1597 goto out;
Li Zefan0caa1022010-12-20 16:30:25 +08001598
Liu Bob9ca0662012-06-29 03:58:49 -06001599 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1600 ret = -EINVAL;
1601 goto out_drop_write;
1602 }
Li Zefan0caa1022010-12-20 16:30:25 +08001603
Liu Bob9ca0662012-06-29 03:58:49 -06001604 if (copy_from_user(&flags, arg, sizeof(flags))) {
1605 ret = -EFAULT;
1606 goto out_drop_write;
1607 }
Li Zefan0caa1022010-12-20 16:30:25 +08001608
Liu Bob9ca0662012-06-29 03:58:49 -06001609 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1610 ret = -EINVAL;
1611 goto out_drop_write;
1612 }
Li Zefan0caa1022010-12-20 16:30:25 +08001613
Liu Bob9ca0662012-06-29 03:58:49 -06001614 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1615 ret = -EOPNOTSUPP;
1616 goto out_drop_write;
1617 }
Li Zefan0caa1022010-12-20 16:30:25 +08001618
Liu Bob9ca0662012-06-29 03:58:49 -06001619 if (!inode_owner_or_capable(inode)) {
1620 ret = -EACCES;
1621 goto out_drop_write;
1622 }
Li Zefanb4dc2b82011-02-16 06:06:34 +00001623
Li Zefan0caa1022010-12-20 16:30:25 +08001624 down_write(&root->fs_info->subvol_sem);
1625
1626 /* nothing to do */
1627 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
Liu Bob9ca0662012-06-29 03:58:49 -06001628 goto out_drop_sem;
Li Zefan0caa1022010-12-20 16:30:25 +08001629
1630 root_flags = btrfs_root_flags(&root->root_item);
1631 if (flags & BTRFS_SUBVOL_RDONLY)
1632 btrfs_set_root_flags(&root->root_item,
1633 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1634 else
1635 btrfs_set_root_flags(&root->root_item,
1636 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1637
1638 trans = btrfs_start_transaction(root, 1);
1639 if (IS_ERR(trans)) {
1640 ret = PTR_ERR(trans);
1641 goto out_reset;
1642 }
1643
Li Zefanb4dc2b82011-02-16 06:06:34 +00001644 ret = btrfs_update_root(trans, root->fs_info->tree_root,
Li Zefan0caa1022010-12-20 16:30:25 +08001645 &root->root_key, &root->root_item);
1646
1647 btrfs_commit_transaction(trans, root);
1648out_reset:
1649 if (ret)
1650 btrfs_set_root_flags(&root->root_item, root_flags);
Liu Bob9ca0662012-06-29 03:58:49 -06001651out_drop_sem:
Li Zefan0caa1022010-12-20 16:30:25 +08001652 up_write(&root->fs_info->subvol_sem);
Liu Bob9ca0662012-06-29 03:58:49 -06001653out_drop_write:
1654 mnt_drop_write_file(file);
1655out:
Li Zefan0caa1022010-12-20 16:30:25 +08001656 return ret;
1657}
1658
Yan, Zheng76dda932009-09-21 16:00:26 -04001659/*
1660 * helper to check if the subvolume references other subvolumes
1661 */
1662static noinline int may_destroy_subvol(struct btrfs_root *root)
1663{
1664 struct btrfs_path *path;
1665 struct btrfs_key key;
1666 int ret;
1667
1668 path = btrfs_alloc_path();
1669 if (!path)
1670 return -ENOMEM;
1671
1672 key.objectid = root->root_key.objectid;
1673 key.type = BTRFS_ROOT_REF_KEY;
1674 key.offset = (u64)-1;
1675
1676 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1677 &key, path, 0, 0);
1678 if (ret < 0)
1679 goto out;
1680 BUG_ON(ret == 0);
1681
1682 ret = 0;
1683 if (path->slots[0] > 0) {
1684 path->slots[0]--;
1685 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1686 if (key.objectid == root->root_key.objectid &&
1687 key.type == BTRFS_ROOT_REF_KEY)
1688 ret = -ENOTEMPTY;
1689 }
1690out:
1691 btrfs_free_path(path);
1692 return ret;
1693}
1694
Chris Masonac8e9812010-02-28 15:39:26 -05001695static noinline int key_in_sk(struct btrfs_key *key,
1696 struct btrfs_ioctl_search_key *sk)
1697{
Chris Masonabc6e132010-03-18 12:10:08 -04001698 struct btrfs_key test;
1699 int ret;
1700
1701 test.objectid = sk->min_objectid;
1702 test.type = sk->min_type;
1703 test.offset = sk->min_offset;
1704
1705 ret = btrfs_comp_cpu_keys(key, &test);
1706 if (ret < 0)
Chris Masonac8e9812010-02-28 15:39:26 -05001707 return 0;
Chris Masonabc6e132010-03-18 12:10:08 -04001708
1709 test.objectid = sk->max_objectid;
1710 test.type = sk->max_type;
1711 test.offset = sk->max_offset;
1712
1713 ret = btrfs_comp_cpu_keys(key, &test);
1714 if (ret > 0)
Chris Masonac8e9812010-02-28 15:39:26 -05001715 return 0;
1716 return 1;
1717}
1718
1719static noinline int copy_to_sk(struct btrfs_root *root,
1720 struct btrfs_path *path,
1721 struct btrfs_key *key,
1722 struct btrfs_ioctl_search_key *sk,
1723 char *buf,
1724 unsigned long *sk_offset,
1725 int *num_found)
1726{
1727 u64 found_transid;
1728 struct extent_buffer *leaf;
1729 struct btrfs_ioctl_search_header sh;
1730 unsigned long item_off;
1731 unsigned long item_len;
1732 int nritems;
1733 int i;
1734 int slot;
Chris Masonac8e9812010-02-28 15:39:26 -05001735 int ret = 0;
1736
1737 leaf = path->nodes[0];
1738 slot = path->slots[0];
1739 nritems = btrfs_header_nritems(leaf);
1740
1741 if (btrfs_header_generation(leaf) > sk->max_transid) {
1742 i = nritems;
1743 goto advance_key;
1744 }
1745 found_transid = btrfs_header_generation(leaf);
1746
1747 for (i = slot; i < nritems; i++) {
1748 item_off = btrfs_item_ptr_offset(leaf, i);
1749 item_len = btrfs_item_size_nr(leaf, i);
1750
1751 if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1752 item_len = 0;
1753
1754 if (sizeof(sh) + item_len + *sk_offset >
1755 BTRFS_SEARCH_ARGS_BUFSIZE) {
1756 ret = 1;
1757 goto overflow;
1758 }
1759
1760 btrfs_item_key_to_cpu(leaf, key, i);
1761 if (!key_in_sk(key, sk))
1762 continue;
1763
1764 sh.objectid = key->objectid;
1765 sh.offset = key->offset;
1766 sh.type = key->type;
1767 sh.len = item_len;
1768 sh.transid = found_transid;
1769
1770 /* copy search result header */
1771 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1772 *sk_offset += sizeof(sh);
1773
1774 if (item_len) {
1775 char *p = buf + *sk_offset;
1776 /* copy the item */
1777 read_extent_buffer(leaf, p,
1778 item_off, item_len);
1779 *sk_offset += item_len;
Chris Masonac8e9812010-02-28 15:39:26 -05001780 }
Hugo Millse2156862011-05-14 17:43:41 +00001781 (*num_found)++;
Chris Masonac8e9812010-02-28 15:39:26 -05001782
1783 if (*num_found >= sk->nr_items)
1784 break;
1785 }
1786advance_key:
Chris Masonac8e9812010-02-28 15:39:26 -05001787 ret = 0;
Chris Masonabc6e132010-03-18 12:10:08 -04001788 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1789 key->offset++;
1790 else if (key->type < (u8)-1 && key->type < sk->max_type) {
1791 key->offset = 0;
1792 key->type++;
1793 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1794 key->offset = 0;
1795 key->type = 0;
1796 key->objectid++;
1797 } else
1798 ret = 1;
Chris Masonac8e9812010-02-28 15:39:26 -05001799overflow:
Chris Masonac8e9812010-02-28 15:39:26 -05001800 return ret;
1801}
1802
1803static noinline int search_ioctl(struct inode *inode,
1804 struct btrfs_ioctl_search_args *args)
1805{
1806 struct btrfs_root *root;
1807 struct btrfs_key key;
1808 struct btrfs_key max_key;
1809 struct btrfs_path *path;
1810 struct btrfs_ioctl_search_key *sk = &args->key;
1811 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1812 int ret;
1813 int num_found = 0;
1814 unsigned long sk_offset = 0;
1815
1816 path = btrfs_alloc_path();
1817 if (!path)
1818 return -ENOMEM;
1819
1820 if (sk->tree_id == 0) {
1821 /* search the root of the inode that was passed */
1822 root = BTRFS_I(inode)->root;
1823 } else {
1824 key.objectid = sk->tree_id;
1825 key.type = BTRFS_ROOT_ITEM_KEY;
1826 key.offset = (u64)-1;
1827 root = btrfs_read_fs_root_no_name(info, &key);
1828 if (IS_ERR(root)) {
1829 printk(KERN_ERR "could not find root %llu\n",
1830 sk->tree_id);
1831 btrfs_free_path(path);
1832 return -ENOENT;
1833 }
1834 }
1835
1836 key.objectid = sk->min_objectid;
1837 key.type = sk->min_type;
1838 key.offset = sk->min_offset;
1839
1840 max_key.objectid = sk->max_objectid;
1841 max_key.type = sk->max_type;
1842 max_key.offset = sk->max_offset;
1843
1844 path->keep_locks = 1;
1845
1846 while(1) {
1847 ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1848 sk->min_transid);
1849 if (ret != 0) {
1850 if (ret > 0)
1851 ret = 0;
1852 goto err;
1853 }
1854 ret = copy_to_sk(root, path, &key, sk, args->buf,
1855 &sk_offset, &num_found);
David Sterbab3b4aa72011-04-21 01:20:15 +02001856 btrfs_release_path(path);
Chris Masonac8e9812010-02-28 15:39:26 -05001857 if (ret || num_found >= sk->nr_items)
1858 break;
1859
1860 }
1861 ret = 0;
1862err:
1863 sk->nr_items = num_found;
1864 btrfs_free_path(path);
1865 return ret;
1866}
1867
1868static noinline int btrfs_ioctl_tree_search(struct file *file,
1869 void __user *argp)
1870{
1871 struct btrfs_ioctl_search_args *args;
1872 struct inode *inode;
1873 int ret;
1874
1875 if (!capable(CAP_SYS_ADMIN))
1876 return -EPERM;
1877
Julia Lawall2354d08f2010-10-29 15:14:18 -04001878 args = memdup_user(argp, sizeof(*args));
1879 if (IS_ERR(args))
1880 return PTR_ERR(args);
Chris Masonac8e9812010-02-28 15:39:26 -05001881
Chris Masonac8e9812010-02-28 15:39:26 -05001882 inode = fdentry(file)->d_inode;
1883 ret = search_ioctl(inode, args);
1884 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1885 ret = -EFAULT;
1886 kfree(args);
1887 return ret;
1888}
1889
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00001890/*
Chris Masonac8e9812010-02-28 15:39:26 -05001891 * Search INODE_REFs to identify path name of 'dirid' directory
1892 * in a 'tree_id' tree. and sets path name to 'name'.
1893 */
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00001894static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1895 u64 tree_id, u64 dirid, char *name)
1896{
1897 struct btrfs_root *root;
1898 struct btrfs_key key;
Chris Masonac8e9812010-02-28 15:39:26 -05001899 char *ptr;
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00001900 int ret = -1;
1901 int slot;
1902 int len;
1903 int total_len = 0;
1904 struct btrfs_inode_ref *iref;
1905 struct extent_buffer *l;
1906 struct btrfs_path *path;
1907
1908 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1909 name[0]='\0';
1910 return 0;
1911 }
1912
1913 path = btrfs_alloc_path();
1914 if (!path)
1915 return -ENOMEM;
1916
Chris Masonac8e9812010-02-28 15:39:26 -05001917 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00001918
1919 key.objectid = tree_id;
1920 key.type = BTRFS_ROOT_ITEM_KEY;
1921 key.offset = (u64)-1;
1922 root = btrfs_read_fs_root_no_name(info, &key);
1923 if (IS_ERR(root)) {
1924 printk(KERN_ERR "could not find root %llu\n", tree_id);
Chris Mason8ad6fca2010-03-18 12:23:10 -04001925 ret = -ENOENT;
1926 goto out;
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00001927 }
1928
1929 key.objectid = dirid;
1930 key.type = BTRFS_INODE_REF_KEY;
Chris Mason8ad6fca2010-03-18 12:23:10 -04001931 key.offset = (u64)-1;
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00001932
1933 while(1) {
1934 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1935 if (ret < 0)
1936 goto out;
1937
1938 l = path->nodes[0];
1939 slot = path->slots[0];
Chris Mason8ad6fca2010-03-18 12:23:10 -04001940 if (ret > 0 && slot > 0)
1941 slot--;
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00001942 btrfs_item_key_to_cpu(l, &key, slot);
1943
1944 if (ret > 0 && (key.objectid != dirid ||
Chris Masonac8e9812010-02-28 15:39:26 -05001945 key.type != BTRFS_INODE_REF_KEY)) {
1946 ret = -ENOENT;
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00001947 goto out;
Chris Masonac8e9812010-02-28 15:39:26 -05001948 }
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00001949
1950 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1951 len = btrfs_inode_ref_name_len(l, iref);
1952 ptr -= len + 1;
1953 total_len += len + 1;
Chris Masonac8e9812010-02-28 15:39:26 -05001954 if (ptr < name)
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00001955 goto out;
1956
1957 *(ptr + len) = '/';
1958 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1959
1960 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1961 break;
1962
David Sterbab3b4aa72011-04-21 01:20:15 +02001963 btrfs_release_path(path);
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00001964 key.objectid = key.offset;
Chris Mason8ad6fca2010-03-18 12:23:10 -04001965 key.offset = (u64)-1;
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00001966 dirid = key.objectid;
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00001967 }
Chris Masonac8e9812010-02-28 15:39:26 -05001968 if (ptr < name)
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00001969 goto out;
Li Zefan77906a502011-07-14 03:16:00 +00001970 memmove(name, ptr, total_len);
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00001971 name[total_len]='\0';
1972 ret = 0;
1973out:
1974 btrfs_free_path(path);
Chris Masonac8e9812010-02-28 15:39:26 -05001975 return ret;
1976}
1977
1978static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1979 void __user *argp)
1980{
1981 struct btrfs_ioctl_ino_lookup_args *args;
1982 struct inode *inode;
1983 int ret;
1984
1985 if (!capable(CAP_SYS_ADMIN))
1986 return -EPERM;
1987
Julia Lawall2354d08f2010-10-29 15:14:18 -04001988 args = memdup_user(argp, sizeof(*args));
1989 if (IS_ERR(args))
1990 return PTR_ERR(args);
Dan Carpenterc2b96922010-03-20 11:24:15 +00001991
Chris Masonac8e9812010-02-28 15:39:26 -05001992 inode = fdentry(file)->d_inode;
1993
Chris Mason1b53ac42010-03-18 12:17:05 -04001994 if (args->treeid == 0)
1995 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1996
Chris Masonac8e9812010-02-28 15:39:26 -05001997 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1998 args->treeid, args->objectid,
1999 args->name);
2000
2001 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2002 ret = -EFAULT;
2003
2004 kfree(args);
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002005 return ret;
2006}
2007
Yan, Zheng76dda932009-09-21 16:00:26 -04002008static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2009 void __user *arg)
2010{
2011 struct dentry *parent = fdentry(file);
2012 struct dentry *dentry;
2013 struct inode *dir = parent->d_inode;
2014 struct inode *inode;
2015 struct btrfs_root *root = BTRFS_I(dir)->root;
2016 struct btrfs_root *dest = NULL;
2017 struct btrfs_ioctl_vol_args *vol_args;
2018 struct btrfs_trans_handle *trans;
2019 int namelen;
2020 int ret;
2021 int err = 0;
2022
Yan, Zheng76dda932009-09-21 16:00:26 -04002023 vol_args = memdup_user(arg, sizeof(*vol_args));
2024 if (IS_ERR(vol_args))
2025 return PTR_ERR(vol_args);
2026
2027 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2028 namelen = strlen(vol_args->name);
2029 if (strchr(vol_args->name, '/') ||
2030 strncmp(vol_args->name, "..", namelen) == 0) {
2031 err = -EINVAL;
2032 goto out;
2033 }
2034
Al Viroa561be72011-11-23 11:57:51 -05002035 err = mnt_want_write_file(file);
Yan, Zheng76dda932009-09-21 16:00:26 -04002036 if (err)
2037 goto out;
2038
2039 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
2040 dentry = lookup_one_len(vol_args->name, parent, namelen);
2041 if (IS_ERR(dentry)) {
2042 err = PTR_ERR(dentry);
2043 goto out_unlock_dir;
2044 }
2045
2046 if (!dentry->d_inode) {
2047 err = -ENOENT;
2048 goto out_dput;
2049 }
2050
2051 inode = dentry->d_inode;
Sage Weil4260f7c2010-10-29 15:46:43 -04002052 dest = BTRFS_I(inode)->root;
2053 if (!capable(CAP_SYS_ADMIN)){
2054 /*
2055 * Regular user. Only allow this with a special mount
2056 * option, when the user has write+exec access to the
2057 * subvol root, and when rmdir(2) would have been
2058 * allowed.
2059 *
2060 * Note that this is _not_ check that the subvol is
2061 * empty or doesn't contain data that we wouldn't
2062 * otherwise be able to delete.
2063 *
2064 * Users who want to delete empty subvols should try
2065 * rmdir(2).
2066 */
2067 err = -EPERM;
2068 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2069 goto out_dput;
2070
2071 /*
2072 * Do not allow deletion if the parent dir is the same
2073 * as the dir to be deleted. That means the ioctl
2074 * must be called on the dentry referencing the root
2075 * of the subvol, not a random directory contained
2076 * within it.
2077 */
2078 err = -EINVAL;
2079 if (root == dest)
2080 goto out_dput;
2081
2082 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2083 if (err)
2084 goto out_dput;
2085
2086 /* check if subvolume may be deleted by a non-root user */
2087 err = btrfs_may_delete(dir, dentry, 1);
2088 if (err)
2089 goto out_dput;
2090 }
2091
Li Zefan33345d012011-04-20 10:31:50 +08002092 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
Yan, Zheng76dda932009-09-21 16:00:26 -04002093 err = -EINVAL;
2094 goto out_dput;
2095 }
2096
Yan, Zheng76dda932009-09-21 16:00:26 -04002097 mutex_lock(&inode->i_mutex);
2098 err = d_invalidate(dentry);
2099 if (err)
2100 goto out_unlock;
2101
2102 down_write(&root->fs_info->subvol_sem);
2103
2104 err = may_destroy_subvol(dest);
2105 if (err)
2106 goto out_up_write;
2107
Yan, Zhenga22285a2010-05-16 10:48:46 -04002108 trans = btrfs_start_transaction(root, 0);
2109 if (IS_ERR(trans)) {
2110 err = PTR_ERR(trans);
Dan Carpenterd3270992010-05-29 09:46:47 +00002111 goto out_up_write;
Yan, Zhenga22285a2010-05-16 10:48:46 -04002112 }
2113 trans->block_rsv = &root->fs_info->global_block_rsv;
2114
Yan, Zheng76dda932009-09-21 16:00:26 -04002115 ret = btrfs_unlink_subvol(trans, root, dir,
2116 dest->root_key.objectid,
2117 dentry->d_name.name,
2118 dentry->d_name.len);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002119 if (ret) {
2120 err = ret;
2121 btrfs_abort_transaction(trans, root, ret);
2122 goto out_end_trans;
2123 }
Yan, Zheng76dda932009-09-21 16:00:26 -04002124
2125 btrfs_record_root_in_trans(trans, dest);
2126
2127 memset(&dest->root_item.drop_progress, 0,
2128 sizeof(dest->root_item.drop_progress));
2129 dest->root_item.drop_level = 0;
2130 btrfs_set_root_refs(&dest->root_item, 0);
2131
Yan, Zhengd68fc572010-05-16 10:49:58 -04002132 if (!xchg(&dest->orphan_item_inserted, 1)) {
2133 ret = btrfs_insert_orphan_item(trans,
2134 root->fs_info->tree_root,
2135 dest->root_key.objectid);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002136 if (ret) {
2137 btrfs_abort_transaction(trans, root, ret);
2138 err = ret;
2139 goto out_end_trans;
2140 }
Yan, Zhengd68fc572010-05-16 10:49:58 -04002141 }
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002142out_end_trans:
Sage Weil531cb132010-10-29 15:41:32 -04002143 ret = btrfs_end_transaction(trans, root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002144 if (ret && !err)
2145 err = ret;
Yan, Zheng76dda932009-09-21 16:00:26 -04002146 inode->i_flags |= S_DEAD;
2147out_up_write:
2148 up_write(&root->fs_info->subvol_sem);
2149out_unlock:
2150 mutex_unlock(&inode->i_mutex);
2151 if (!err) {
Yan, Zhengefefb142009-10-09 09:25:16 -04002152 shrink_dcache_sb(root->fs_info->sb);
Yan, Zheng76dda932009-09-21 16:00:26 -04002153 btrfs_invalidate_inodes(dest);
2154 d_delete(dentry);
2155 }
2156out_dput:
2157 dput(dentry);
2158out_unlock_dir:
2159 mutex_unlock(&dir->i_mutex);
Al Viro2a79f172011-12-09 08:06:57 -05002160 mnt_drop_write_file(file);
Yan, Zheng76dda932009-09-21 16:00:26 -04002161out:
2162 kfree(vol_args);
2163 return err;
2164}
2165
Chris Mason1e701a32010-03-11 09:42:04 -05002166static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002167{
2168 struct inode *inode = fdentry(file)->d_inode;
2169 struct btrfs_root *root = BTRFS_I(inode)->root;
Chris Mason1e701a32010-03-11 09:42:04 -05002170 struct btrfs_ioctl_defrag_range_args *range;
Yan Zhengc146afa2008-11-12 14:34:12 -05002171 int ret;
2172
Li Zefanb83cc962010-12-20 16:04:08 +08002173 if (btrfs_root_readonly(root))
2174 return -EROFS;
2175
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01002176 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2177 1)) {
2178 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2179 return -EINPROGRESS;
2180 }
Al Viroa561be72011-11-23 11:57:51 -05002181 ret = mnt_want_write_file(file);
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01002182 if (ret) {
2183 atomic_set(&root->fs_info->mutually_exclusive_operation_running,
2184 0);
Yan Zhengc146afa2008-11-12 14:34:12 -05002185 return ret;
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01002186 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002187
2188 switch (inode->i_mode & S_IFMT) {
2189 case S_IFDIR:
Chris Masone441d542009-01-05 16:57:23 -05002190 if (!capable(CAP_SYS_ADMIN)) {
2191 ret = -EPERM;
2192 goto out;
2193 }
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002194 ret = btrfs_defrag_root(root, 0);
2195 if (ret)
2196 goto out;
2197 ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002198 break;
2199 case S_IFREG:
Chris Masone441d542009-01-05 16:57:23 -05002200 if (!(file->f_mode & FMODE_WRITE)) {
2201 ret = -EINVAL;
2202 goto out;
2203 }
Chris Mason1e701a32010-03-11 09:42:04 -05002204
2205 range = kzalloc(sizeof(*range), GFP_KERNEL);
2206 if (!range) {
2207 ret = -ENOMEM;
2208 goto out;
2209 }
2210
2211 if (argp) {
2212 if (copy_from_user(range, argp,
2213 sizeof(*range))) {
2214 ret = -EFAULT;
2215 kfree(range);
Dan Carpenter683be162010-03-20 11:24:48 +00002216 goto out;
Chris Mason1e701a32010-03-11 09:42:04 -05002217 }
2218 /* compression requires us to start the IO */
2219 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2220 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2221 range->extent_thresh = (u32)-1;
2222 }
2223 } else {
2224 /* the rest are all set to zero by kzalloc */
2225 range->len = (u64)-1;
2226 }
Chris Mason4cb53002011-05-24 15:35:30 -04002227 ret = btrfs_defrag_file(fdentry(file)->d_inode, file,
2228 range, 0, 0);
2229 if (ret > 0)
2230 ret = 0;
Chris Mason1e701a32010-03-11 09:42:04 -05002231 kfree(range);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002232 break;
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002233 default:
2234 ret = -EINVAL;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002235 }
Chris Masone441d542009-01-05 16:57:23 -05002236out:
Al Viro2a79f172011-12-09 08:06:57 -05002237 mnt_drop_write_file(file);
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01002238 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
Chris Masone441d542009-01-05 16:57:23 -05002239 return ret;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002240}
2241
Christoph Hellwigb2950862008-12-02 09:54:17 -05002242static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002243{
2244 struct btrfs_ioctl_vol_args *vol_args;
2245 int ret;
2246
Chris Masone441d542009-01-05 16:57:23 -05002247 if (!capable(CAP_SYS_ADMIN))
2248 return -EPERM;
2249
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01002250 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2251 1)) {
2252 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2253 return -EINPROGRESS;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002254 }
2255
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01002256 mutex_lock(&root->fs_info->volume_mutex);
Li Zefandae7b662009-04-08 15:06:54 +08002257 vol_args = memdup_user(arg, sizeof(*vol_args));
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002258 if (IS_ERR(vol_args)) {
2259 ret = PTR_ERR(vol_args);
2260 goto out;
2261 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002262
Mark Fasheh5516e592008-07-24 12:20:14 -04002263 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002264 ret = btrfs_init_new_device(root, vol_args->name);
2265
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002266 kfree(vol_args);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002267out:
2268 mutex_unlock(&root->fs_info->volume_mutex);
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01002269 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002270 return ret;
2271}
2272
Christoph Hellwigb2950862008-12-02 09:54:17 -05002273static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002274{
2275 struct btrfs_ioctl_vol_args *vol_args;
2276 int ret;
2277
Chris Masone441d542009-01-05 16:57:23 -05002278 if (!capable(CAP_SYS_ADMIN))
2279 return -EPERM;
2280
Yan Zhengc146afa2008-11-12 14:34:12 -05002281 if (root->fs_info->sb->s_flags & MS_RDONLY)
2282 return -EROFS;
2283
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01002284 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2285 1)) {
2286 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2287 return -EINPROGRESS;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002288 }
2289
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01002290 mutex_lock(&root->fs_info->volume_mutex);
Li Zefandae7b662009-04-08 15:06:54 +08002291 vol_args = memdup_user(arg, sizeof(*vol_args));
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002292 if (IS_ERR(vol_args)) {
2293 ret = PTR_ERR(vol_args);
2294 goto out;
2295 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002296
Mark Fasheh5516e592008-07-24 12:20:14 -04002297 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002298 ret = btrfs_rm_device(root, vol_args->name);
2299
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002300 kfree(vol_args);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002301out:
2302 mutex_unlock(&root->fs_info->volume_mutex);
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01002303 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002304 return ret;
2305}
2306
Jan Schmidt475f6382011-03-11 15:41:01 +01002307static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2308{
Li Zefan027ed2f2011-06-08 08:27:56 +00002309 struct btrfs_ioctl_fs_info_args *fi_args;
Jan Schmidt475f6382011-03-11 15:41:01 +01002310 struct btrfs_device *device;
2311 struct btrfs_device *next;
2312 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
Li Zefan027ed2f2011-06-08 08:27:56 +00002313 int ret = 0;
Jan Schmidt475f6382011-03-11 15:41:01 +01002314
2315 if (!capable(CAP_SYS_ADMIN))
2316 return -EPERM;
2317
Li Zefan027ed2f2011-06-08 08:27:56 +00002318 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2319 if (!fi_args)
2320 return -ENOMEM;
2321
2322 fi_args->num_devices = fs_devices->num_devices;
2323 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
Jan Schmidt475f6382011-03-11 15:41:01 +01002324
2325 mutex_lock(&fs_devices->device_list_mutex);
2326 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
Li Zefan027ed2f2011-06-08 08:27:56 +00002327 if (device->devid > fi_args->max_id)
2328 fi_args->max_id = device->devid;
Jan Schmidt475f6382011-03-11 15:41:01 +01002329 }
2330 mutex_unlock(&fs_devices->device_list_mutex);
2331
Li Zefan027ed2f2011-06-08 08:27:56 +00002332 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2333 ret = -EFAULT;
Jan Schmidt475f6382011-03-11 15:41:01 +01002334
Li Zefan027ed2f2011-06-08 08:27:56 +00002335 kfree(fi_args);
2336 return ret;
Jan Schmidt475f6382011-03-11 15:41:01 +01002337}
2338
2339static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2340{
2341 struct btrfs_ioctl_dev_info_args *di_args;
2342 struct btrfs_device *dev;
2343 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2344 int ret = 0;
2345 char *s_uuid = NULL;
2346 char empty_uuid[BTRFS_UUID_SIZE] = {0};
2347
2348 if (!capable(CAP_SYS_ADMIN))
2349 return -EPERM;
2350
2351 di_args = memdup_user(arg, sizeof(*di_args));
2352 if (IS_ERR(di_args))
2353 return PTR_ERR(di_args);
2354
2355 if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2356 s_uuid = di_args->uuid;
2357
2358 mutex_lock(&fs_devices->device_list_mutex);
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01002359 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
Jan Schmidt475f6382011-03-11 15:41:01 +01002360 mutex_unlock(&fs_devices->device_list_mutex);
2361
2362 if (!dev) {
2363 ret = -ENODEV;
2364 goto out;
2365 }
2366
2367 di_args->devid = dev->devid;
2368 di_args->bytes_used = dev->bytes_used;
2369 di_args->total_bytes = dev->total_bytes;
2370 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
Jim Meyeringa27202f2012-04-26 18:36:56 +02002371 if (dev->name) {
Josef Bacik606686e2012-06-04 14:03:51 -04002372 struct rcu_string *name;
2373
2374 rcu_read_lock();
2375 name = rcu_dereference(dev->name);
2376 strncpy(di_args->path, name->str, sizeof(di_args->path));
2377 rcu_read_unlock();
Jim Meyeringa27202f2012-04-26 18:36:56 +02002378 di_args->path[sizeof(di_args->path) - 1] = 0;
2379 } else {
Stefan Behrens99ba55a2012-03-19 16:17:22 +01002380 di_args->path[0] = '\0';
Jim Meyeringa27202f2012-04-26 18:36:56 +02002381 }
Jan Schmidt475f6382011-03-11 15:41:01 +01002382
2383out:
2384 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2385 ret = -EFAULT;
2386
2387 kfree(di_args);
2388 return ret;
2389}
2390
Yan, Zheng76dda932009-09-21 16:00:26 -04002391static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2392 u64 off, u64 olen, u64 destoff)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002393{
2394 struct inode *inode = fdentry(file)->d_inode;
2395 struct btrfs_root *root = BTRFS_I(inode)->root;
Al Viro2903ff02012-08-28 12:52:22 -04002396 struct fd src_file;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002397 struct inode *src;
2398 struct btrfs_trans_handle *trans;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002399 struct btrfs_path *path;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002400 struct extent_buffer *leaf;
Yan Zhengae01a0a2008-08-04 23:23:47 -04002401 char *buf;
2402 struct btrfs_key key;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002403 u32 nritems;
2404 int slot;
Yan Zhengae01a0a2008-08-04 23:23:47 -04002405 int ret;
Sage Weilc5c9cd42008-11-12 14:32:25 -05002406 u64 len = olen;
2407 u64 bs = root->fs_info->sb->s_blocksize;
Chris Masond20f7042008-12-08 16:58:54 -05002408
Sage Weilc5c9cd42008-11-12 14:32:25 -05002409 /*
2410 * TODO:
2411 * - split compressed inline extents. annoying: we need to
2412 * decompress into destination's address_space (the file offset
2413 * may change, so source mapping won't do), then recompress (or
2414 * otherwise reinsert) a subrange.
2415 * - allow ranges within the same file to be cloned (provided
2416 * they don't overlap)?
2417 */
2418
Chris Masone441d542009-01-05 16:57:23 -05002419 /* the destination must be opened for writing */
Dan Rosenberg2ebc3462010-07-19 16:58:20 -04002420 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
Chris Masone441d542009-01-05 16:57:23 -05002421 return -EINVAL;
2422
Li Zefanb83cc962010-12-20 16:04:08 +08002423 if (btrfs_root_readonly(root))
2424 return -EROFS;
2425
Al Viroa561be72011-11-23 11:57:51 -05002426 ret = mnt_want_write_file(file);
Yan Zhengc146afa2008-11-12 14:34:12 -05002427 if (ret)
2428 return ret;
2429
Al Viro2903ff02012-08-28 12:52:22 -04002430 src_file = fdget(srcfd);
2431 if (!src_file.file) {
Yan Zhengab67b7c2008-12-19 10:58:39 -05002432 ret = -EBADF;
2433 goto out_drop_write;
2434 }
Dan Rosenberg5dc64162010-05-15 11:27:37 -04002435
David Sterba362a20c2011-08-01 18:11:57 +02002436 ret = -EXDEV;
Al Viro2903ff02012-08-28 12:52:22 -04002437 if (src_file.file->f_path.mnt != file->f_path.mnt)
David Sterba362a20c2011-08-01 18:11:57 +02002438 goto out_fput;
2439
Al Viro2903ff02012-08-28 12:52:22 -04002440 src = src_file.file->f_dentry->d_inode;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002441
Sage Weilc5c9cd42008-11-12 14:32:25 -05002442 ret = -EINVAL;
2443 if (src == inode)
2444 goto out_fput;
2445
Dan Rosenberg5dc64162010-05-15 11:27:37 -04002446 /* the src must be open for reading */
Al Viro2903ff02012-08-28 12:52:22 -04002447 if (!(src_file.file->f_mode & FMODE_READ))
Dan Rosenberg5dc64162010-05-15 11:27:37 -04002448 goto out_fput;
2449
Li Zefan0e7b8242011-09-18 10:20:46 -04002450 /* don't make the dst file partly checksummed */
2451 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2452 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2453 goto out_fput;
2454
Yan Zhengae01a0a2008-08-04 23:23:47 -04002455 ret = -EISDIR;
2456 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002457 goto out_fput;
2458
Yan Zhengae01a0a2008-08-04 23:23:47 -04002459 ret = -EXDEV;
David Sterba362a20c2011-08-01 18:11:57 +02002460 if (src->i_sb != inode->i_sb)
Yan Zhengae01a0a2008-08-04 23:23:47 -04002461 goto out_fput;
2462
2463 ret = -ENOMEM;
2464 buf = vmalloc(btrfs_level_size(root, 0));
2465 if (!buf)
2466 goto out_fput;
2467
2468 path = btrfs_alloc_path();
2469 if (!path) {
2470 vfree(buf);
2471 goto out_fput;
2472 }
2473 path->reada = 2;
2474
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002475 if (inode < src) {
Sage Weilfccdae42010-10-29 15:37:33 -04002476 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2477 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002478 } else {
Sage Weilfccdae42010-10-29 15:37:33 -04002479 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2480 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002481 }
2482
Sage Weilc5c9cd42008-11-12 14:32:25 -05002483 /* determine range to clone */
2484 ret = -EINVAL;
Dan Rosenberg2ebc3462010-07-19 16:58:20 -04002485 if (off + len > src->i_size || off + len < off)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002486 goto out_unlock;
Sage Weilc5c9cd42008-11-12 14:32:25 -05002487 if (len == 0)
2488 olen = len = src->i_size - off;
2489 /* if we extend to eof, continue to block boundary */
2490 if (off + len == src->i_size)
Li Zefan2a6b8da2010-11-19 01:36:10 +00002491 len = ALIGN(src->i_size, bs) - off;
Sage Weilc5c9cd42008-11-12 14:32:25 -05002492
2493 /* verify the end result is block aligned */
Li Zefan2a6b8da2010-11-19 01:36:10 +00002494 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2495 !IS_ALIGNED(destoff, bs))
Sage Weilc5c9cd42008-11-12 14:32:25 -05002496 goto out_unlock;
2497
Li Zefand525e8a2011-09-11 10:52:25 -04002498 if (destoff > inode->i_size) {
2499 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2500 if (ret)
2501 goto out_unlock;
2502 }
2503
Li Zefan71ef0782011-09-18 10:20:46 -04002504 /* truncate page cache pages from target inode range */
2505 truncate_inode_pages_range(&inode->i_data, destoff,
2506 PAGE_CACHE_ALIGN(destoff + len) - 1);
2507
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002508 /* do any pending delalloc/csum calc on src, one way or
2509 another, and lock file content */
2510 while (1) {
Zheng Yan31840ae2008-09-23 13:14:14 -04002511 struct btrfs_ordered_extent *ordered;
Liu Boaa42ffd2012-09-18 03:52:23 -06002512 lock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
2513 ordered = btrfs_lookup_first_ordered_extent(src, off + len - 1);
Sage Weil9a019192010-10-29 15:37:33 -04002514 if (!ordered &&
Liu Boaa42ffd2012-09-18 03:52:23 -06002515 !test_range_bit(&BTRFS_I(src)->io_tree, off, off + len - 1,
2516 EXTENT_DELALLOC, 0, NULL))
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002517 break;
Liu Boaa42ffd2012-09-18 03:52:23 -06002518 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
Yan Zhengae01a0a2008-08-04 23:23:47 -04002519 if (ordered)
2520 btrfs_put_ordered_extent(ordered);
Sage Weil9a019192010-10-29 15:37:33 -04002521 btrfs_wait_ordered_range(src, off, len);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002522 }
2523
Sage Weilc5c9cd42008-11-12 14:32:25 -05002524 /* clone data */
Li Zefan33345d012011-04-20 10:31:50 +08002525 key.objectid = btrfs_ino(src);
Yan Zhengae01a0a2008-08-04 23:23:47 -04002526 key.type = BTRFS_EXTENT_DATA_KEY;
2527 key.offset = 0;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002528
2529 while (1) {
2530 /*
2531 * note the key will change type as we walk through the
2532 * tree.
2533 */
David Sterba362a20c2011-08-01 18:11:57 +02002534 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
2535 0, 0);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002536 if (ret < 0)
2537 goto out;
2538
Yan Zhengae01a0a2008-08-04 23:23:47 -04002539 nritems = btrfs_header_nritems(path->nodes[0]);
2540 if (path->slots[0] >= nritems) {
David Sterba362a20c2011-08-01 18:11:57 +02002541 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002542 if (ret < 0)
2543 goto out;
2544 if (ret > 0)
2545 break;
Yan Zhengae01a0a2008-08-04 23:23:47 -04002546 nritems = btrfs_header_nritems(path->nodes[0]);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002547 }
2548 leaf = path->nodes[0];
2549 slot = path->slots[0];
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002550
Yan Zhengae01a0a2008-08-04 23:23:47 -04002551 btrfs_item_key_to_cpu(leaf, &key, slot);
Chris Masond20f7042008-12-08 16:58:54 -05002552 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
Li Zefan33345d012011-04-20 10:31:50 +08002553 key.objectid != btrfs_ino(src))
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002554 break;
2555
Sage Weilc5c9cd42008-11-12 14:32:25 -05002556 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2557 struct btrfs_file_extent_item *extent;
2558 int type;
Zheng Yan31840ae2008-09-23 13:14:14 -04002559 u32 size;
2560 struct btrfs_key new_key;
Sage Weilc5c9cd42008-11-12 14:32:25 -05002561 u64 disko = 0, diskl = 0;
2562 u64 datao = 0, datal = 0;
2563 u8 comp;
Sage Weilb5384d42010-06-12 22:31:14 +00002564 u64 endoff;
Zheng Yan31840ae2008-09-23 13:14:14 -04002565
2566 size = btrfs_item_size_nr(leaf, slot);
2567 read_extent_buffer(leaf, buf,
2568 btrfs_item_ptr_offset(leaf, slot),
2569 size);
Sage Weilc5c9cd42008-11-12 14:32:25 -05002570
2571 extent = btrfs_item_ptr(leaf, slot,
2572 struct btrfs_file_extent_item);
2573 comp = btrfs_file_extent_compression(leaf, extent);
2574 type = btrfs_file_extent_type(leaf, extent);
Chris Masonc8a894d2009-06-27 21:07:03 -04002575 if (type == BTRFS_FILE_EXTENT_REG ||
2576 type == BTRFS_FILE_EXTENT_PREALLOC) {
Chris Masond3977122009-01-05 21:25:51 -05002577 disko = btrfs_file_extent_disk_bytenr(leaf,
2578 extent);
2579 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2580 extent);
Sage Weilc5c9cd42008-11-12 14:32:25 -05002581 datao = btrfs_file_extent_offset(leaf, extent);
Chris Masond3977122009-01-05 21:25:51 -05002582 datal = btrfs_file_extent_num_bytes(leaf,
2583 extent);
Sage Weilc5c9cd42008-11-12 14:32:25 -05002584 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2585 /* take upper bound, may be compressed */
2586 datal = btrfs_file_extent_ram_bytes(leaf,
2587 extent);
2588 }
David Sterbab3b4aa72011-04-21 01:20:15 +02002589 btrfs_release_path(path);
Zheng Yan31840ae2008-09-23 13:14:14 -04002590
Sage Weil050006a2010-10-29 15:37:33 -04002591 if (key.offset + datal <= off ||
Liu Boaa42ffd2012-09-18 03:52:23 -06002592 key.offset >= off + len - 1)
Sage Weilc5c9cd42008-11-12 14:32:25 -05002593 goto next;
2594
Zheng Yan31840ae2008-09-23 13:14:14 -04002595 memcpy(&new_key, &key, sizeof(new_key));
Li Zefan33345d012011-04-20 10:31:50 +08002596 new_key.objectid = btrfs_ino(inode);
Li Zefan4d728ec2011-01-26 14:10:43 +08002597 if (off <= key.offset)
2598 new_key.offset = key.offset + destoff - off;
2599 else
2600 new_key.offset = destoff;
Sage Weilc5c9cd42008-11-12 14:32:25 -05002601
Sage Weilb6f34092011-09-20 14:48:51 -04002602 /*
2603 * 1 - adjusting old extent (we may have to split it)
2604 * 1 - add new extent
2605 * 1 - inode update
2606 */
2607 trans = btrfs_start_transaction(root, 3);
Yan, Zhenga22285a2010-05-16 10:48:46 -04002608 if (IS_ERR(trans)) {
2609 ret = PTR_ERR(trans);
2610 goto out;
2611 }
2612
Chris Masonc8a894d2009-06-27 21:07:03 -04002613 if (type == BTRFS_FILE_EXTENT_REG ||
2614 type == BTRFS_FILE_EXTENT_PREALLOC) {
Li Zefand72c0842011-09-11 10:52:25 -04002615 /*
2616 * a | --- range to clone ---| b
2617 * | ------------- extent ------------- |
2618 */
2619
2620 /* substract range b */
2621 if (key.offset + datal > off + len)
2622 datal = off + len - key.offset;
2623
2624 /* substract range a */
Yan, Zhenga22285a2010-05-16 10:48:46 -04002625 if (off > key.offset) {
2626 datao += off - key.offset;
2627 datal -= off - key.offset;
2628 }
2629
Josef Bacik5dc562c2012-08-17 13:14:17 -04002630 ret = btrfs_drop_extents(trans, root, inode,
Yan, Zhenga22285a2010-05-16 10:48:46 -04002631 new_key.offset,
2632 new_key.offset + datal,
Josef Bacik26714852012-08-29 12:24:27 -04002633 1);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002634 if (ret) {
2635 btrfs_abort_transaction(trans, root,
2636 ret);
2637 btrfs_end_transaction(trans, root);
2638 goto out;
2639 }
Yan, Zhenga22285a2010-05-16 10:48:46 -04002640
Sage Weilc5c9cd42008-11-12 14:32:25 -05002641 ret = btrfs_insert_empty_item(trans, root, path,
2642 &new_key, size);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002643 if (ret) {
2644 btrfs_abort_transaction(trans, root,
2645 ret);
2646 btrfs_end_transaction(trans, root);
2647 goto out;
2648 }
Sage Weilc5c9cd42008-11-12 14:32:25 -05002649
2650 leaf = path->nodes[0];
2651 slot = path->slots[0];
2652 write_extent_buffer(leaf, buf,
2653 btrfs_item_ptr_offset(leaf, slot),
2654 size);
2655
2656 extent = btrfs_item_ptr(leaf, slot,
2657 struct btrfs_file_extent_item);
Sage Weilc5c9cd42008-11-12 14:32:25 -05002658
Sage Weilc5c9cd42008-11-12 14:32:25 -05002659 /* disko == 0 means it's a hole */
2660 if (!disko)
2661 datao = 0;
Sage Weilc5c9cd42008-11-12 14:32:25 -05002662
2663 btrfs_set_file_extent_offset(leaf, extent,
2664 datao);
2665 btrfs_set_file_extent_num_bytes(leaf, extent,
2666 datal);
2667 if (disko) {
2668 inode_add_bytes(inode, datal);
2669 ret = btrfs_inc_extent_ref(trans, root,
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002670 disko, diskl, 0,
2671 root->root_key.objectid,
Li Zefan33345d012011-04-20 10:31:50 +08002672 btrfs_ino(inode),
Arne Jansen66d7e7f2011-09-12 15:26:38 +02002673 new_key.offset - datao,
2674 0);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002675 if (ret) {
2676 btrfs_abort_transaction(trans,
2677 root,
2678 ret);
2679 btrfs_end_transaction(trans,
2680 root);
2681 goto out;
2682
2683 }
Sage Weilc5c9cd42008-11-12 14:32:25 -05002684 }
2685 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2686 u64 skip = 0;
2687 u64 trim = 0;
2688 if (off > key.offset) {
2689 skip = off - key.offset;
2690 new_key.offset += skip;
2691 }
Chris Masond3977122009-01-05 21:25:51 -05002692
Liu Boaa42ffd2012-09-18 03:52:23 -06002693 if (key.offset + datal > off + len)
2694 trim = key.offset + datal - (off + len);
Chris Masond3977122009-01-05 21:25:51 -05002695
Sage Weilc5c9cd42008-11-12 14:32:25 -05002696 if (comp && (skip || trim)) {
Sage Weilc5c9cd42008-11-12 14:32:25 -05002697 ret = -EINVAL;
Yan, Zhenga22285a2010-05-16 10:48:46 -04002698 btrfs_end_transaction(trans, root);
Sage Weilc5c9cd42008-11-12 14:32:25 -05002699 goto out;
2700 }
2701 size -= skip + trim;
2702 datal -= skip + trim;
Yan, Zhenga22285a2010-05-16 10:48:46 -04002703
Josef Bacik5dc562c2012-08-17 13:14:17 -04002704 ret = btrfs_drop_extents(trans, root, inode,
Yan, Zhenga22285a2010-05-16 10:48:46 -04002705 new_key.offset,
2706 new_key.offset + datal,
Josef Bacik26714852012-08-29 12:24:27 -04002707 1);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002708 if (ret) {
2709 btrfs_abort_transaction(trans, root,
2710 ret);
2711 btrfs_end_transaction(trans, root);
2712 goto out;
2713 }
Yan, Zhenga22285a2010-05-16 10:48:46 -04002714
Sage Weilc5c9cd42008-11-12 14:32:25 -05002715 ret = btrfs_insert_empty_item(trans, root, path,
2716 &new_key, size);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002717 if (ret) {
2718 btrfs_abort_transaction(trans, root,
2719 ret);
2720 btrfs_end_transaction(trans, root);
2721 goto out;
2722 }
Sage Weilc5c9cd42008-11-12 14:32:25 -05002723
2724 if (skip) {
Chris Masond3977122009-01-05 21:25:51 -05002725 u32 start =
2726 btrfs_file_extent_calc_inline_size(0);
Sage Weilc5c9cd42008-11-12 14:32:25 -05002727 memmove(buf+start, buf+start+skip,
2728 datal);
2729 }
2730
2731 leaf = path->nodes[0];
2732 slot = path->slots[0];
2733 write_extent_buffer(leaf, buf,
2734 btrfs_item_ptr_offset(leaf, slot),
2735 size);
2736 inode_add_bytes(inode, datal);
2737 }
2738
2739 btrfs_mark_buffer_dirty(leaf);
David Sterbab3b4aa72011-04-21 01:20:15 +02002740 btrfs_release_path(path);
Sage Weilc5c9cd42008-11-12 14:32:25 -05002741
Josef Bacik0c4d2d92012-04-05 15:03:02 -04002742 inode_inc_iversion(inode);
Yan, Zhenga22285a2010-05-16 10:48:46 -04002743 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
Sage Weilb5384d42010-06-12 22:31:14 +00002744
2745 /*
2746 * we round up to the block size at eof when
2747 * determining which extents to clone above,
2748 * but shouldn't round up the file size
2749 */
2750 endoff = new_key.offset + datal;
Li Zefan5f3888f2010-11-19 01:36:34 +00002751 if (endoff > destoff+olen)
2752 endoff = destoff+olen;
Sage Weilb5384d42010-06-12 22:31:14 +00002753 if (endoff > inode->i_size)
2754 btrfs_i_size_write(inode, endoff);
2755
Yan, Zhenga22285a2010-05-16 10:48:46 -04002756 ret = btrfs_update_inode(trans, root, inode);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002757 if (ret) {
2758 btrfs_abort_transaction(trans, root, ret);
2759 btrfs_end_transaction(trans, root);
2760 goto out;
2761 }
2762 ret = btrfs_end_transaction(trans, root);
Yan, Zhenga22285a2010-05-16 10:48:46 -04002763 }
Chris Masond3977122009-01-05 21:25:51 -05002764next:
David Sterbab3b4aa72011-04-21 01:20:15 +02002765 btrfs_release_path(path);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002766 key.offset++;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002767 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002768 ret = 0;
2769out:
David Sterbab3b4aa72011-04-21 01:20:15 +02002770 btrfs_release_path(path);
Liu Boaa42ffd2012-09-18 03:52:23 -06002771 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002772out_unlock:
2773 mutex_unlock(&src->i_mutex);
2774 mutex_unlock(&inode->i_mutex);
Yan Zhengae01a0a2008-08-04 23:23:47 -04002775 vfree(buf);
2776 btrfs_free_path(path);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002777out_fput:
Al Viro2903ff02012-08-28 12:52:22 -04002778 fdput(src_file);
Yan Zhengab67b7c2008-12-19 10:58:39 -05002779out_drop_write:
Al Viro2a79f172011-12-09 08:06:57 -05002780 mnt_drop_write_file(file);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002781 return ret;
2782}
2783
Christoph Hellwig7a865e82008-12-02 09:52:24 -05002784static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
Sage Weilc5c9cd42008-11-12 14:32:25 -05002785{
2786 struct btrfs_ioctl_clone_range_args args;
2787
Christoph Hellwig7a865e82008-12-02 09:52:24 -05002788 if (copy_from_user(&args, argp, sizeof(args)))
Sage Weilc5c9cd42008-11-12 14:32:25 -05002789 return -EFAULT;
2790 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2791 args.src_length, args.dest_offset);
2792}
2793
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002794/*
2795 * there are many ways the trans_start and trans_end ioctls can lead
2796 * to deadlocks. They should only be used by applications that
2797 * basically own the machine, and have a very in depth understanding
2798 * of all the possible deadlocks and enospc problems.
2799 */
Christoph Hellwigb2950862008-12-02 09:54:17 -05002800static long btrfs_ioctl_trans_start(struct file *file)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002801{
2802 struct inode *inode = fdentry(file)->d_inode;
2803 struct btrfs_root *root = BTRFS_I(inode)->root;
2804 struct btrfs_trans_handle *trans;
Sage Weil1ab86ae2009-09-29 18:38:44 -04002805 int ret;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002806
Sage Weil1ab86ae2009-09-29 18:38:44 -04002807 ret = -EPERM;
Christoph Hellwigdf5b5522008-06-11 21:53:58 -04002808 if (!capable(CAP_SYS_ADMIN))
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002809 goto out;
Sage Weil1ab86ae2009-09-29 18:38:44 -04002810
2811 ret = -EINPROGRESS;
2812 if (file->private_data)
2813 goto out;
Sage Weil9ca9ee02008-08-04 10:41:27 -04002814
Li Zefanb83cc962010-12-20 16:04:08 +08002815 ret = -EROFS;
2816 if (btrfs_root_readonly(root))
2817 goto out;
2818
Al Viroa561be72011-11-23 11:57:51 -05002819 ret = mnt_want_write_file(file);
Yan Zhengc146afa2008-11-12 14:34:12 -05002820 if (ret)
2821 goto out;
2822
Josef Bacika4abeea2011-04-11 17:25:13 -04002823 atomic_inc(&root->fs_info->open_ioctl_trans);
Sage Weil9ca9ee02008-08-04 10:41:27 -04002824
Sage Weil1ab86ae2009-09-29 18:38:44 -04002825 ret = -ENOMEM;
Josef Bacik7a7eaa42011-04-13 12:54:33 -04002826 trans = btrfs_start_ioctl_transaction(root);
Tsutomu Itohabd30bb2011-01-24 00:57:10 +00002827 if (IS_ERR(trans))
Sage Weil1ab86ae2009-09-29 18:38:44 -04002828 goto out_drop;
2829
2830 file->private_data = trans;
2831 return 0;
2832
2833out_drop:
Josef Bacika4abeea2011-04-11 17:25:13 -04002834 atomic_dec(&root->fs_info->open_ioctl_trans);
Al Viro2a79f172011-12-09 08:06:57 -05002835 mnt_drop_write_file(file);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002836out:
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002837 return ret;
2838}
2839
Josef Bacik6ef5ed02009-12-11 21:11:29 +00002840static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2841{
2842 struct inode *inode = fdentry(file)->d_inode;
2843 struct btrfs_root *root = BTRFS_I(inode)->root;
2844 struct btrfs_root *new_root;
2845 struct btrfs_dir_item *di;
2846 struct btrfs_trans_handle *trans;
2847 struct btrfs_path *path;
2848 struct btrfs_key location;
2849 struct btrfs_disk_key disk_key;
Josef Bacik6ef5ed02009-12-11 21:11:29 +00002850 u64 objectid = 0;
2851 u64 dir_id;
Miao Xie3c04ce02012-11-26 08:43:07 +00002852 int ret;
Josef Bacik6ef5ed02009-12-11 21:11:29 +00002853
2854 if (!capable(CAP_SYS_ADMIN))
2855 return -EPERM;
2856
Miao Xie3c04ce02012-11-26 08:43:07 +00002857 ret = mnt_want_write_file(file);
2858 if (ret)
2859 return ret;
2860
2861 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
2862 ret = -EFAULT;
2863 goto out;
2864 }
Josef Bacik6ef5ed02009-12-11 21:11:29 +00002865
2866 if (!objectid)
2867 objectid = root->root_key.objectid;
2868
2869 location.objectid = objectid;
2870 location.type = BTRFS_ROOT_ITEM_KEY;
2871 location.offset = (u64)-1;
2872
2873 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
Miao Xie3c04ce02012-11-26 08:43:07 +00002874 if (IS_ERR(new_root)) {
2875 ret = PTR_ERR(new_root);
2876 goto out;
2877 }
Josef Bacik6ef5ed02009-12-11 21:11:29 +00002878
Miao Xie3c04ce02012-11-26 08:43:07 +00002879 if (btrfs_root_refs(&new_root->root_item) == 0) {
2880 ret = -ENOENT;
2881 goto out;
2882 }
Josef Bacik6ef5ed02009-12-11 21:11:29 +00002883
2884 path = btrfs_alloc_path();
Miao Xie3c04ce02012-11-26 08:43:07 +00002885 if (!path) {
2886 ret = -ENOMEM;
2887 goto out;
2888 }
Josef Bacik6ef5ed02009-12-11 21:11:29 +00002889 path->leave_spinning = 1;
2890
2891 trans = btrfs_start_transaction(root, 1);
Tsutomu Itoh98d5dc12011-01-20 06:19:37 +00002892 if (IS_ERR(trans)) {
Josef Bacik6ef5ed02009-12-11 21:11:29 +00002893 btrfs_free_path(path);
Miao Xie3c04ce02012-11-26 08:43:07 +00002894 ret = PTR_ERR(trans);
2895 goto out;
Josef Bacik6ef5ed02009-12-11 21:11:29 +00002896 }
2897
David Sterba6c417612011-04-13 15:41:04 +02002898 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
Josef Bacik6ef5ed02009-12-11 21:11:29 +00002899 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2900 dir_id, "default", 7, 1);
Dan Carpentercf1e99a2010-05-29 09:47:24 +00002901 if (IS_ERR_OR_NULL(di)) {
Josef Bacik6ef5ed02009-12-11 21:11:29 +00002902 btrfs_free_path(path);
2903 btrfs_end_transaction(trans, root);
2904 printk(KERN_ERR "Umm, you don't have the default dir item, "
2905 "this isn't going to work\n");
Miao Xie3c04ce02012-11-26 08:43:07 +00002906 ret = -ENOENT;
2907 goto out;
Josef Bacik6ef5ed02009-12-11 21:11:29 +00002908 }
2909
2910 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2911 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2912 btrfs_mark_buffer_dirty(path->nodes[0]);
2913 btrfs_free_path(path);
2914
Mitch Harder2b0ce2c2012-07-24 11:58:43 -06002915 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
Josef Bacik6ef5ed02009-12-11 21:11:29 +00002916 btrfs_end_transaction(trans, root);
Miao Xie3c04ce02012-11-26 08:43:07 +00002917out:
2918 mnt_drop_write_file(file);
2919 return ret;
Josef Bacik6ef5ed02009-12-11 21:11:29 +00002920}
2921
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02002922void btrfs_get_block_group_info(struct list_head *groups_list,
2923 struct btrfs_ioctl_space_info *space)
Josef Bacikbf5fc0932010-09-29 11:22:36 -04002924{
2925 struct btrfs_block_group_cache *block_group;
2926
2927 space->total_bytes = 0;
2928 space->used_bytes = 0;
2929 space->flags = 0;
2930 list_for_each_entry(block_group, groups_list, list) {
2931 space->flags = block_group->flags;
2932 space->total_bytes += block_group->key.offset;
2933 space->used_bytes +=
2934 btrfs_block_group_used(&block_group->item);
2935 }
2936}
2937
Josef Bacik1406e432010-01-13 18:19:06 +00002938long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2939{
2940 struct btrfs_ioctl_space_args space_args;
2941 struct btrfs_ioctl_space_info space;
2942 struct btrfs_ioctl_space_info *dest;
Chris Mason7fde62b2010-03-16 15:40:10 -04002943 struct btrfs_ioctl_space_info *dest_orig;
Daniel J Blueman13f26962011-04-11 15:56:31 +00002944 struct btrfs_ioctl_space_info __user *user_dest;
Josef Bacik1406e432010-01-13 18:19:06 +00002945 struct btrfs_space_info *info;
Josef Bacikbf5fc0932010-09-29 11:22:36 -04002946 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2947 BTRFS_BLOCK_GROUP_SYSTEM,
2948 BTRFS_BLOCK_GROUP_METADATA,
2949 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2950 int num_types = 4;
Chris Mason7fde62b2010-03-16 15:40:10 -04002951 int alloc_size;
Josef Bacik1406e432010-01-13 18:19:06 +00002952 int ret = 0;
Dan Rosenberg51788b12011-02-14 16:04:23 -05002953 u64 slot_count = 0;
Josef Bacikbf5fc0932010-09-29 11:22:36 -04002954 int i, c;
Josef Bacik1406e432010-01-13 18:19:06 +00002955
2956 if (copy_from_user(&space_args,
2957 (struct btrfs_ioctl_space_args __user *)arg,
2958 sizeof(space_args)))
2959 return -EFAULT;
2960
Josef Bacikbf5fc0932010-09-29 11:22:36 -04002961 for (i = 0; i < num_types; i++) {
2962 struct btrfs_space_info *tmp;
2963
2964 info = NULL;
2965 rcu_read_lock();
2966 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2967 list) {
2968 if (tmp->flags == types[i]) {
2969 info = tmp;
2970 break;
2971 }
2972 }
2973 rcu_read_unlock();
2974
2975 if (!info)
2976 continue;
2977
2978 down_read(&info->groups_sem);
2979 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2980 if (!list_empty(&info->block_groups[c]))
2981 slot_count++;
2982 }
2983 up_read(&info->groups_sem);
2984 }
Josef Bacik1406e432010-01-13 18:19:06 +00002985
Chris Mason7fde62b2010-03-16 15:40:10 -04002986 /* space_slots == 0 means they are asking for a count */
2987 if (space_args.space_slots == 0) {
2988 space_args.total_spaces = slot_count;
2989 goto out;
2990 }
Josef Bacikbf5fc0932010-09-29 11:22:36 -04002991
Dan Rosenberg51788b12011-02-14 16:04:23 -05002992 slot_count = min_t(u64, space_args.space_slots, slot_count);
Josef Bacikbf5fc0932010-09-29 11:22:36 -04002993
Chris Mason7fde62b2010-03-16 15:40:10 -04002994 alloc_size = sizeof(*dest) * slot_count;
Josef Bacikbf5fc0932010-09-29 11:22:36 -04002995
Chris Mason7fde62b2010-03-16 15:40:10 -04002996 /* we generally have at most 6 or so space infos, one for each raid
2997 * level. So, a whole page should be more than enough for everyone
2998 */
2999 if (alloc_size > PAGE_CACHE_SIZE)
3000 return -ENOMEM;
3001
3002 space_args.total_spaces = 0;
3003 dest = kmalloc(alloc_size, GFP_NOFS);
3004 if (!dest)
3005 return -ENOMEM;
3006 dest_orig = dest;
3007
3008 /* now we have a buffer to copy into */
Josef Bacikbf5fc0932010-09-29 11:22:36 -04003009 for (i = 0; i < num_types; i++) {
3010 struct btrfs_space_info *tmp;
Chris Mason7fde62b2010-03-16 15:40:10 -04003011
Dan Rosenberg51788b12011-02-14 16:04:23 -05003012 if (!slot_count)
3013 break;
3014
Josef Bacikbf5fc0932010-09-29 11:22:36 -04003015 info = NULL;
3016 rcu_read_lock();
3017 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3018 list) {
3019 if (tmp->flags == types[i]) {
3020 info = tmp;
3021 break;
3022 }
3023 }
3024 rcu_read_unlock();
Chris Mason7fde62b2010-03-16 15:40:10 -04003025
Josef Bacikbf5fc0932010-09-29 11:22:36 -04003026 if (!info)
3027 continue;
3028 down_read(&info->groups_sem);
3029 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3030 if (!list_empty(&info->block_groups[c])) {
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003031 btrfs_get_block_group_info(
3032 &info->block_groups[c], &space);
Josef Bacikbf5fc0932010-09-29 11:22:36 -04003033 memcpy(dest, &space, sizeof(space));
3034 dest++;
3035 space_args.total_spaces++;
Dan Rosenberg51788b12011-02-14 16:04:23 -05003036 slot_count--;
Josef Bacikbf5fc0932010-09-29 11:22:36 -04003037 }
Dan Rosenberg51788b12011-02-14 16:04:23 -05003038 if (!slot_count)
3039 break;
Josef Bacikbf5fc0932010-09-29 11:22:36 -04003040 }
3041 up_read(&info->groups_sem);
Josef Bacik1406e432010-01-13 18:19:06 +00003042 }
Josef Bacik1406e432010-01-13 18:19:06 +00003043
Daniel J Blueman2eec6c82012-04-26 00:37:14 +08003044 user_dest = (struct btrfs_ioctl_space_info __user *)
Chris Mason7fde62b2010-03-16 15:40:10 -04003045 (arg + sizeof(struct btrfs_ioctl_space_args));
3046
3047 if (copy_to_user(user_dest, dest_orig, alloc_size))
3048 ret = -EFAULT;
3049
3050 kfree(dest_orig);
3051out:
3052 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
Josef Bacik1406e432010-01-13 18:19:06 +00003053 ret = -EFAULT;
3054
3055 return ret;
3056}
3057
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003058/*
3059 * there are many ways the trans_start and trans_end ioctls can lead
3060 * to deadlocks. They should only be used by applications that
3061 * basically own the machine, and have a very in depth understanding
3062 * of all the possible deadlocks and enospc problems.
3063 */
3064long btrfs_ioctl_trans_end(struct file *file)
3065{
3066 struct inode *inode = fdentry(file)->d_inode;
3067 struct btrfs_root *root = BTRFS_I(inode)->root;
3068 struct btrfs_trans_handle *trans;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003069
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003070 trans = file->private_data;
Sage Weil1ab86ae2009-09-29 18:38:44 -04003071 if (!trans)
3072 return -EINVAL;
Christoph Hellwigb2141072008-09-05 16:43:31 -04003073 file->private_data = NULL;
Sage Weil9ca9ee02008-08-04 10:41:27 -04003074
Sage Weil1ab86ae2009-09-29 18:38:44 -04003075 btrfs_end_transaction(trans, root);
3076
Josef Bacika4abeea2011-04-11 17:25:13 -04003077 atomic_dec(&root->fs_info->open_ioctl_trans);
Sage Weil9ca9ee02008-08-04 10:41:27 -04003078
Al Viro2a79f172011-12-09 08:06:57 -05003079 mnt_drop_write_file(file);
Sage Weil1ab86ae2009-09-29 18:38:44 -04003080 return 0;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003081}
3082
Miao Xie9a8c28b2012-11-26 08:40:43 +00003083static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3084 void __user *argp)
Sage Weil46204592010-10-29 15:41:32 -04003085{
Sage Weil46204592010-10-29 15:41:32 -04003086 struct btrfs_trans_handle *trans;
3087 u64 transid;
Tsutomu Itohdb5b4932011-03-23 08:14:16 +00003088 int ret;
Sage Weil46204592010-10-29 15:41:32 -04003089
Miao Xieff7c1d32012-11-26 08:41:29 +00003090 trans = btrfs_attach_transaction(root);
3091 if (IS_ERR(trans)) {
3092 if (PTR_ERR(trans) != -ENOENT)
3093 return PTR_ERR(trans);
3094
3095 /* No running transaction, don't bother */
3096 transid = root->fs_info->last_trans_committed;
3097 goto out;
3098 }
Sage Weil46204592010-10-29 15:41:32 -04003099 transid = trans->transid;
Tsutomu Itohdb5b4932011-03-23 08:14:16 +00003100 ret = btrfs_commit_transaction_async(trans, root, 0);
Tsutomu Itoh8b2b2d32011-04-04 01:52:13 +00003101 if (ret) {
3102 btrfs_end_transaction(trans, root);
Tsutomu Itohdb5b4932011-03-23 08:14:16 +00003103 return ret;
Tsutomu Itoh8b2b2d32011-04-04 01:52:13 +00003104 }
Miao Xieff7c1d32012-11-26 08:41:29 +00003105out:
Sage Weil46204592010-10-29 15:41:32 -04003106 if (argp)
3107 if (copy_to_user(argp, &transid, sizeof(transid)))
3108 return -EFAULT;
3109 return 0;
3110}
3111
Miao Xie9a8c28b2012-11-26 08:40:43 +00003112static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
3113 void __user *argp)
Sage Weil46204592010-10-29 15:41:32 -04003114{
Sage Weil46204592010-10-29 15:41:32 -04003115 u64 transid;
3116
3117 if (argp) {
3118 if (copy_from_user(&transid, argp, sizeof(transid)))
3119 return -EFAULT;
3120 } else {
3121 transid = 0; /* current trans */
3122 }
3123 return btrfs_wait_for_commit(root, transid);
3124}
3125
Jan Schmidt475f6382011-03-11 15:41:01 +01003126static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg)
3127{
3128 int ret;
3129 struct btrfs_ioctl_scrub_args *sa;
3130
3131 if (!capable(CAP_SYS_ADMIN))
3132 return -EPERM;
3133
3134 sa = memdup_user(arg, sizeof(*sa));
3135 if (IS_ERR(sa))
3136 return PTR_ERR(sa);
3137
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01003138 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
Stefan Behrens63a212a2012-11-05 18:29:28 +01003139 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3140 0);
Jan Schmidt475f6382011-03-11 15:41:01 +01003141
3142 if (copy_to_user(arg, sa, sizeof(*sa)))
3143 ret = -EFAULT;
3144
3145 kfree(sa);
3146 return ret;
3147}
3148
3149static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3150{
3151 if (!capable(CAP_SYS_ADMIN))
3152 return -EPERM;
3153
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01003154 return btrfs_scrub_cancel(root->fs_info);
Jan Schmidt475f6382011-03-11 15:41:01 +01003155}
3156
3157static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3158 void __user *arg)
3159{
3160 struct btrfs_ioctl_scrub_args *sa;
3161 int ret;
3162
3163 if (!capable(CAP_SYS_ADMIN))
3164 return -EPERM;
3165
3166 sa = memdup_user(arg, sizeof(*sa));
3167 if (IS_ERR(sa))
3168 return PTR_ERR(sa);
3169
3170 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3171
3172 if (copy_to_user(arg, sa, sizeof(*sa)))
3173 ret = -EFAULT;
3174
3175 kfree(sa);
3176 return ret;
3177}
3178
Stefan Behrensc11d2c232012-05-25 16:06:09 +02003179static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
David Sterbab27f7c02012-06-22 06:30:39 -06003180 void __user *arg)
Stefan Behrensc11d2c232012-05-25 16:06:09 +02003181{
3182 struct btrfs_ioctl_get_dev_stats *sa;
3183 int ret;
3184
Stefan Behrensc11d2c232012-05-25 16:06:09 +02003185 sa = memdup_user(arg, sizeof(*sa));
3186 if (IS_ERR(sa))
3187 return PTR_ERR(sa);
3188
David Sterbab27f7c02012-06-22 06:30:39 -06003189 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3190 kfree(sa);
3191 return -EPERM;
3192 }
3193
3194 ret = btrfs_get_dev_stats(root, sa);
Stefan Behrensc11d2c232012-05-25 16:06:09 +02003195
3196 if (copy_to_user(arg, sa, sizeof(*sa)))
3197 ret = -EFAULT;
3198
3199 kfree(sa);
3200 return ret;
3201}
3202
Stefan Behrens3f6bcfb2012-11-06 15:08:53 +01003203static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
3204{
3205 struct btrfs_ioctl_dev_replace_args *p;
3206 int ret;
3207
3208 if (!capable(CAP_SYS_ADMIN))
3209 return -EPERM;
3210
3211 p = memdup_user(arg, sizeof(*p));
3212 if (IS_ERR(p))
3213 return PTR_ERR(p);
3214
3215 switch (p->cmd) {
3216 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3217 if (atomic_xchg(
3218 &root->fs_info->mutually_exclusive_operation_running,
3219 1)) {
3220 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3221 ret = -EINPROGRESS;
3222 } else {
3223 ret = btrfs_dev_replace_start(root, p);
3224 atomic_set(
3225 &root->fs_info->mutually_exclusive_operation_running,
3226 0);
3227 }
3228 break;
3229 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3230 btrfs_dev_replace_status(root->fs_info, p);
3231 ret = 0;
3232 break;
3233 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3234 ret = btrfs_dev_replace_cancel(root->fs_info, p);
3235 break;
3236 default:
3237 ret = -EINVAL;
3238 break;
3239 }
3240
3241 if (copy_to_user(arg, p, sizeof(*p)))
3242 ret = -EFAULT;
3243
3244 kfree(p);
3245 return ret;
3246}
3247
Jan Schmidtd7728c92011-07-07 16:48:38 +02003248static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3249{
3250 int ret = 0;
3251 int i;
Chris Mason740c3d22011-11-02 15:48:34 -04003252 u64 rel_ptr;
Jan Schmidtd7728c92011-07-07 16:48:38 +02003253 int size;
Chris Mason806468f2011-11-06 03:07:10 -05003254 struct btrfs_ioctl_ino_path_args *ipa = NULL;
Jan Schmidtd7728c92011-07-07 16:48:38 +02003255 struct inode_fs_paths *ipath = NULL;
3256 struct btrfs_path *path;
3257
3258 if (!capable(CAP_SYS_ADMIN))
3259 return -EPERM;
3260
3261 path = btrfs_alloc_path();
3262 if (!path) {
3263 ret = -ENOMEM;
3264 goto out;
3265 }
3266
3267 ipa = memdup_user(arg, sizeof(*ipa));
3268 if (IS_ERR(ipa)) {
3269 ret = PTR_ERR(ipa);
3270 ipa = NULL;
3271 goto out;
3272 }
3273
3274 size = min_t(u32, ipa->size, 4096);
3275 ipath = init_ipath(size, root, path);
3276 if (IS_ERR(ipath)) {
3277 ret = PTR_ERR(ipath);
3278 ipath = NULL;
3279 goto out;
3280 }
3281
3282 ret = paths_from_inode(ipa->inum, ipath);
3283 if (ret < 0)
3284 goto out;
3285
3286 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
Jeff Mahoney745c4d82011-11-20 07:31:57 -05003287 rel_ptr = ipath->fspath->val[i] -
3288 (u64)(unsigned long)ipath->fspath->val;
Chris Mason740c3d22011-11-02 15:48:34 -04003289 ipath->fspath->val[i] = rel_ptr;
Jan Schmidtd7728c92011-07-07 16:48:38 +02003290 }
3291
Jeff Mahoney745c4d82011-11-20 07:31:57 -05003292 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3293 (void *)(unsigned long)ipath->fspath, size);
Jan Schmidtd7728c92011-07-07 16:48:38 +02003294 if (ret) {
3295 ret = -EFAULT;
3296 goto out;
3297 }
3298
3299out:
3300 btrfs_free_path(path);
3301 free_ipath(ipath);
3302 kfree(ipa);
3303
3304 return ret;
3305}
3306
3307static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3308{
3309 struct btrfs_data_container *inodes = ctx;
3310 const size_t c = 3 * sizeof(u64);
3311
3312 if (inodes->bytes_left >= c) {
3313 inodes->bytes_left -= c;
3314 inodes->val[inodes->elem_cnt] = inum;
3315 inodes->val[inodes->elem_cnt + 1] = offset;
3316 inodes->val[inodes->elem_cnt + 2] = root;
3317 inodes->elem_cnt += 3;
3318 } else {
3319 inodes->bytes_missing += c - inodes->bytes_left;
3320 inodes->bytes_left = 0;
3321 inodes->elem_missed += 3;
3322 }
3323
3324 return 0;
3325}
3326
3327static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3328 void __user *arg)
3329{
3330 int ret = 0;
3331 int size;
Jan Schmidtd7728c92011-07-07 16:48:38 +02003332 struct btrfs_ioctl_logical_ino_args *loi;
3333 struct btrfs_data_container *inodes = NULL;
3334 struct btrfs_path *path = NULL;
Jan Schmidtd7728c92011-07-07 16:48:38 +02003335
3336 if (!capable(CAP_SYS_ADMIN))
3337 return -EPERM;
3338
3339 loi = memdup_user(arg, sizeof(*loi));
3340 if (IS_ERR(loi)) {
3341 ret = PTR_ERR(loi);
3342 loi = NULL;
3343 goto out;
3344 }
3345
3346 path = btrfs_alloc_path();
3347 if (!path) {
3348 ret = -ENOMEM;
3349 goto out;
3350 }
3351
Liu Bo425d17a2012-09-07 20:01:30 -06003352 size = min_t(u32, loi->size, 64 * 1024);
Jan Schmidtd7728c92011-07-07 16:48:38 +02003353 inodes = init_data_container(size);
3354 if (IS_ERR(inodes)) {
3355 ret = PTR_ERR(inodes);
3356 inodes = NULL;
3357 goto out;
3358 }
3359
Liu Bodf031f02012-09-07 20:01:29 -06003360 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
3361 build_ino_list, inodes);
3362 if (ret == -EINVAL)
Jan Schmidtd7728c92011-07-07 16:48:38 +02003363 ret = -ENOENT;
3364 if (ret < 0)
3365 goto out;
3366
Jeff Mahoney745c4d82011-11-20 07:31:57 -05003367 ret = copy_to_user((void *)(unsigned long)loi->inodes,
3368 (void *)(unsigned long)inodes, size);
Jan Schmidtd7728c92011-07-07 16:48:38 +02003369 if (ret)
3370 ret = -EFAULT;
3371
3372out:
3373 btrfs_free_path(path);
Liu Bo425d17a2012-09-07 20:01:30 -06003374 vfree(inodes);
Jan Schmidtd7728c92011-07-07 16:48:38 +02003375 kfree(loi);
3376
3377 return ret;
3378}
3379
Ilya Dryomov19a39dc2012-01-16 22:04:49 +02003380void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02003381 struct btrfs_ioctl_balance_args *bargs)
3382{
3383 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3384
3385 bargs->flags = bctl->flags;
3386
Ilya Dryomov837d5b62012-01-16 22:04:49 +02003387 if (atomic_read(&fs_info->balance_running))
3388 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3389 if (atomic_read(&fs_info->balance_pause_req))
3390 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
Ilya Dryomova7e99c62012-01-16 22:04:49 +02003391 if (atomic_read(&fs_info->balance_cancel_req))
3392 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
Ilya Dryomov837d5b62012-01-16 22:04:49 +02003393
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02003394 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3395 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3396 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
Ilya Dryomov19a39dc2012-01-16 22:04:49 +02003397
3398 if (lock) {
3399 spin_lock(&fs_info->balance_lock);
3400 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3401 spin_unlock(&fs_info->balance_lock);
3402 } else {
3403 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3404 }
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02003405}
3406
Liu Bo9ba1f6e2012-05-11 18:11:26 +08003407static long btrfs_ioctl_balance(struct file *file, void __user *arg)
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02003408{
Liu Bo9ba1f6e2012-05-11 18:11:26 +08003409 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02003410 struct btrfs_fs_info *fs_info = root->fs_info;
3411 struct btrfs_ioctl_balance_args *bargs;
3412 struct btrfs_balance_control *bctl;
3413 int ret;
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01003414 int need_to_clear_lock = 0;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02003415
3416 if (!capable(CAP_SYS_ADMIN))
3417 return -EPERM;
3418
Liu Boe54bfa312012-06-29 03:58:48 -06003419 ret = mnt_want_write_file(file);
Liu Bo9ba1f6e2012-05-11 18:11:26 +08003420 if (ret)
3421 return ret;
3422
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02003423 mutex_lock(&fs_info->volume_mutex);
3424 mutex_lock(&fs_info->balance_mutex);
3425
3426 if (arg) {
3427 bargs = memdup_user(arg, sizeof(*bargs));
3428 if (IS_ERR(bargs)) {
3429 ret = PTR_ERR(bargs);
3430 goto out;
3431 }
Ilya Dryomovde322262012-01-16 22:04:49 +02003432
3433 if (bargs->flags & BTRFS_BALANCE_RESUME) {
3434 if (!fs_info->balance_ctl) {
3435 ret = -ENOTCONN;
3436 goto out_bargs;
3437 }
3438
3439 bctl = fs_info->balance_ctl;
3440 spin_lock(&fs_info->balance_lock);
3441 bctl->flags |= BTRFS_BALANCE_RESUME;
3442 spin_unlock(&fs_info->balance_lock);
3443
3444 goto do_balance;
3445 }
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02003446 } else {
3447 bargs = NULL;
3448 }
3449
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01003450 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
3451 1)) {
3452 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
Ilya Dryomov837d5b62012-01-16 22:04:49 +02003453 ret = -EINPROGRESS;
3454 goto out_bargs;
3455 }
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01003456 need_to_clear_lock = 1;
Ilya Dryomov837d5b62012-01-16 22:04:49 +02003457
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02003458 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3459 if (!bctl) {
3460 ret = -ENOMEM;
3461 goto out_bargs;
3462 }
3463
3464 bctl->fs_info = fs_info;
3465 if (arg) {
3466 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3467 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3468 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3469
3470 bctl->flags = bargs->flags;
Ilya Dryomovf43ffb62012-01-16 22:04:47 +02003471 } else {
3472 /* balance everything - no filters */
3473 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02003474 }
3475
Ilya Dryomovde322262012-01-16 22:04:49 +02003476do_balance:
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02003477 ret = btrfs_balance(bctl, bargs);
3478 /*
Ilya Dryomov837d5b62012-01-16 22:04:49 +02003479 * bctl is freed in __cancel_balance or in free_fs_info if
3480 * restriper was paused all the way until unmount
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02003481 */
3482 if (arg) {
3483 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3484 ret = -EFAULT;
3485 }
3486
3487out_bargs:
3488 kfree(bargs);
3489out:
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01003490 if (need_to_clear_lock)
3491 atomic_set(&root->fs_info->mutually_exclusive_operation_running,
3492 0);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02003493 mutex_unlock(&fs_info->balance_mutex);
3494 mutex_unlock(&fs_info->volume_mutex);
Liu Boe54bfa312012-06-29 03:58:48 -06003495 mnt_drop_write_file(file);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02003496 return ret;
3497}
3498
Ilya Dryomov837d5b62012-01-16 22:04:49 +02003499static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
3500{
3501 if (!capable(CAP_SYS_ADMIN))
3502 return -EPERM;
3503
3504 switch (cmd) {
3505 case BTRFS_BALANCE_CTL_PAUSE:
3506 return btrfs_pause_balance(root->fs_info);
Ilya Dryomova7e99c62012-01-16 22:04:49 +02003507 case BTRFS_BALANCE_CTL_CANCEL:
3508 return btrfs_cancel_balance(root->fs_info);
Ilya Dryomov837d5b62012-01-16 22:04:49 +02003509 }
3510
3511 return -EINVAL;
3512}
3513
Ilya Dryomov19a39dc2012-01-16 22:04:49 +02003514static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
3515 void __user *arg)
3516{
3517 struct btrfs_fs_info *fs_info = root->fs_info;
3518 struct btrfs_ioctl_balance_args *bargs;
3519 int ret = 0;
3520
3521 if (!capable(CAP_SYS_ADMIN))
3522 return -EPERM;
3523
3524 mutex_lock(&fs_info->balance_mutex);
3525 if (!fs_info->balance_ctl) {
3526 ret = -ENOTCONN;
3527 goto out;
3528 }
3529
3530 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
3531 if (!bargs) {
3532 ret = -ENOMEM;
3533 goto out;
3534 }
3535
3536 update_ioctl_balance_args(fs_info, 1, bargs);
3537
3538 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3539 ret = -EFAULT;
3540
3541 kfree(bargs);
3542out:
3543 mutex_unlock(&fs_info->balance_mutex);
3544 return ret;
3545}
3546
Arne Jansen5d13a372011-09-14 15:53:51 +02003547static long btrfs_ioctl_quota_ctl(struct btrfs_root *root, void __user *arg)
3548{
3549 struct btrfs_ioctl_quota_ctl_args *sa;
3550 struct btrfs_trans_handle *trans = NULL;
3551 int ret;
3552 int err;
3553
3554 if (!capable(CAP_SYS_ADMIN))
3555 return -EPERM;
3556
3557 if (root->fs_info->sb->s_flags & MS_RDONLY)
3558 return -EROFS;
3559
3560 sa = memdup_user(arg, sizeof(*sa));
3561 if (IS_ERR(sa))
3562 return PTR_ERR(sa);
3563
3564 if (sa->cmd != BTRFS_QUOTA_CTL_RESCAN) {
3565 trans = btrfs_start_transaction(root, 2);
3566 if (IS_ERR(trans)) {
3567 ret = PTR_ERR(trans);
3568 goto out;
3569 }
3570 }
3571
3572 switch (sa->cmd) {
3573 case BTRFS_QUOTA_CTL_ENABLE:
3574 ret = btrfs_quota_enable(trans, root->fs_info);
3575 break;
3576 case BTRFS_QUOTA_CTL_DISABLE:
3577 ret = btrfs_quota_disable(trans, root->fs_info);
3578 break;
3579 case BTRFS_QUOTA_CTL_RESCAN:
3580 ret = btrfs_quota_rescan(root->fs_info);
3581 break;
3582 default:
3583 ret = -EINVAL;
3584 break;
3585 }
3586
3587 if (copy_to_user(arg, sa, sizeof(*sa)))
3588 ret = -EFAULT;
3589
3590 if (trans) {
3591 err = btrfs_commit_transaction(trans, root);
3592 if (err && !ret)
3593 ret = err;
3594 }
3595
3596out:
3597 kfree(sa);
3598 return ret;
3599}
3600
3601static long btrfs_ioctl_qgroup_assign(struct btrfs_root *root, void __user *arg)
3602{
3603 struct btrfs_ioctl_qgroup_assign_args *sa;
3604 struct btrfs_trans_handle *trans;
3605 int ret;
3606 int err;
3607
3608 if (!capable(CAP_SYS_ADMIN))
3609 return -EPERM;
3610
3611 if (root->fs_info->sb->s_flags & MS_RDONLY)
3612 return -EROFS;
3613
3614 sa = memdup_user(arg, sizeof(*sa));
3615 if (IS_ERR(sa))
3616 return PTR_ERR(sa);
3617
3618 trans = btrfs_join_transaction(root);
3619 if (IS_ERR(trans)) {
3620 ret = PTR_ERR(trans);
3621 goto out;
3622 }
3623
3624 /* FIXME: check if the IDs really exist */
3625 if (sa->assign) {
3626 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
3627 sa->src, sa->dst);
3628 } else {
3629 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
3630 sa->src, sa->dst);
3631 }
3632
3633 err = btrfs_end_transaction(trans, root);
3634 if (err && !ret)
3635 ret = err;
3636
3637out:
3638 kfree(sa);
3639 return ret;
3640}
3641
3642static long btrfs_ioctl_qgroup_create(struct btrfs_root *root, void __user *arg)
3643{
3644 struct btrfs_ioctl_qgroup_create_args *sa;
3645 struct btrfs_trans_handle *trans;
3646 int ret;
3647 int err;
3648
3649 if (!capable(CAP_SYS_ADMIN))
3650 return -EPERM;
3651
3652 if (root->fs_info->sb->s_flags & MS_RDONLY)
3653 return -EROFS;
3654
3655 sa = memdup_user(arg, sizeof(*sa));
3656 if (IS_ERR(sa))
3657 return PTR_ERR(sa);
3658
3659 trans = btrfs_join_transaction(root);
3660 if (IS_ERR(trans)) {
3661 ret = PTR_ERR(trans);
3662 goto out;
3663 }
3664
3665 /* FIXME: check if the IDs really exist */
3666 if (sa->create) {
3667 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
3668 NULL);
3669 } else {
3670 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
3671 }
3672
3673 err = btrfs_end_transaction(trans, root);
3674 if (err && !ret)
3675 ret = err;
3676
3677out:
3678 kfree(sa);
3679 return ret;
3680}
3681
3682static long btrfs_ioctl_qgroup_limit(struct btrfs_root *root, void __user *arg)
3683{
3684 struct btrfs_ioctl_qgroup_limit_args *sa;
3685 struct btrfs_trans_handle *trans;
3686 int ret;
3687 int err;
3688 u64 qgroupid;
3689
3690 if (!capable(CAP_SYS_ADMIN))
3691 return -EPERM;
3692
3693 if (root->fs_info->sb->s_flags & MS_RDONLY)
3694 return -EROFS;
3695
3696 sa = memdup_user(arg, sizeof(*sa));
3697 if (IS_ERR(sa))
3698 return PTR_ERR(sa);
3699
3700 trans = btrfs_join_transaction(root);
3701 if (IS_ERR(trans)) {
3702 ret = PTR_ERR(trans);
3703 goto out;
3704 }
3705
3706 qgroupid = sa->qgroupid;
3707 if (!qgroupid) {
3708 /* take the current subvol as qgroup */
3709 qgroupid = root->root_key.objectid;
3710 }
3711
3712 /* FIXME: check if the IDs really exist */
3713 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
3714
3715 err = btrfs_end_transaction(trans, root);
3716 if (err && !ret)
3717 ret = err;
3718
3719out:
3720 kfree(sa);
3721 return ret;
3722}
3723
Alexander Block8ea05e32012-07-25 17:35:53 +02003724static long btrfs_ioctl_set_received_subvol(struct file *file,
3725 void __user *arg)
3726{
3727 struct btrfs_ioctl_received_subvol_args *sa = NULL;
3728 struct inode *inode = fdentry(file)->d_inode;
3729 struct btrfs_root *root = BTRFS_I(inode)->root;
3730 struct btrfs_root_item *root_item = &root->root_item;
3731 struct btrfs_trans_handle *trans;
3732 struct timespec ct = CURRENT_TIME;
3733 int ret = 0;
3734
3735 ret = mnt_want_write_file(file);
3736 if (ret < 0)
3737 return ret;
3738
3739 down_write(&root->fs_info->subvol_sem);
3740
3741 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
3742 ret = -EINVAL;
3743 goto out;
3744 }
3745
3746 if (btrfs_root_readonly(root)) {
3747 ret = -EROFS;
3748 goto out;
3749 }
3750
3751 if (!inode_owner_or_capable(inode)) {
3752 ret = -EACCES;
3753 goto out;
3754 }
3755
3756 sa = memdup_user(arg, sizeof(*sa));
3757 if (IS_ERR(sa)) {
3758 ret = PTR_ERR(sa);
3759 sa = NULL;
3760 goto out;
3761 }
3762
3763 trans = btrfs_start_transaction(root, 1);
3764 if (IS_ERR(trans)) {
3765 ret = PTR_ERR(trans);
3766 trans = NULL;
3767 goto out;
3768 }
3769
3770 sa->rtransid = trans->transid;
3771 sa->rtime.sec = ct.tv_sec;
3772 sa->rtime.nsec = ct.tv_nsec;
3773
3774 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
3775 btrfs_set_root_stransid(root_item, sa->stransid);
3776 btrfs_set_root_rtransid(root_item, sa->rtransid);
3777 root_item->stime.sec = cpu_to_le64(sa->stime.sec);
3778 root_item->stime.nsec = cpu_to_le32(sa->stime.nsec);
3779 root_item->rtime.sec = cpu_to_le64(sa->rtime.sec);
3780 root_item->rtime.nsec = cpu_to_le32(sa->rtime.nsec);
3781
3782 ret = btrfs_update_root(trans, root->fs_info->tree_root,
3783 &root->root_key, &root->root_item);
3784 if (ret < 0) {
3785 btrfs_end_transaction(trans, root);
3786 trans = NULL;
3787 goto out;
3788 } else {
3789 ret = btrfs_commit_transaction(trans, root);
3790 if (ret < 0)
3791 goto out;
3792 }
3793
3794 ret = copy_to_user(arg, sa, sizeof(*sa));
3795 if (ret)
3796 ret = -EFAULT;
3797
3798out:
3799 kfree(sa);
3800 up_write(&root->fs_info->subvol_sem);
3801 mnt_drop_write_file(file);
3802 return ret;
3803}
3804
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003805long btrfs_ioctl(struct file *file, unsigned int
3806 cmd, unsigned long arg)
3807{
3808 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
Christoph Hellwig4bcabaa2008-12-02 06:36:08 -05003809 void __user *argp = (void __user *)arg;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003810
3811 switch (cmd) {
Christoph Hellwig6cbff002009-04-17 10:37:41 +02003812 case FS_IOC_GETFLAGS:
3813 return btrfs_ioctl_getflags(file, argp);
3814 case FS_IOC_SETFLAGS:
3815 return btrfs_ioctl_setflags(file, argp);
3816 case FS_IOC_GETVERSION:
3817 return btrfs_ioctl_getversion(file, argp);
Li Dongyangf7039b12011-03-24 10:24:28 +00003818 case FITRIM:
3819 return btrfs_ioctl_fitrim(file, argp);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003820 case BTRFS_IOC_SNAP_CREATE:
Li Zefanfa0d2b92010-12-20 15:53:28 +08003821 return btrfs_ioctl_snap_create(file, argp, 0);
Li Zefanfdfb1e42010-12-10 06:41:56 +00003822 case BTRFS_IOC_SNAP_CREATE_V2:
Li Zefanfa0d2b92010-12-20 15:53:28 +08003823 return btrfs_ioctl_snap_create_v2(file, argp, 0);
Chris Mason3de45862008-11-17 21:02:50 -05003824 case BTRFS_IOC_SUBVOL_CREATE:
Li Zefanfa0d2b92010-12-20 15:53:28 +08003825 return btrfs_ioctl_snap_create(file, argp, 1);
Arne Jansen6f72c7e2011-09-14 15:58:21 +02003826 case BTRFS_IOC_SUBVOL_CREATE_V2:
3827 return btrfs_ioctl_snap_create_v2(file, argp, 1);
Yan, Zheng76dda932009-09-21 16:00:26 -04003828 case BTRFS_IOC_SNAP_DESTROY:
3829 return btrfs_ioctl_snap_destroy(file, argp);
Li Zefan0caa1022010-12-20 16:30:25 +08003830 case BTRFS_IOC_SUBVOL_GETFLAGS:
3831 return btrfs_ioctl_subvol_getflags(file, argp);
3832 case BTRFS_IOC_SUBVOL_SETFLAGS:
3833 return btrfs_ioctl_subvol_setflags(file, argp);
Josef Bacik6ef5ed02009-12-11 21:11:29 +00003834 case BTRFS_IOC_DEFAULT_SUBVOL:
3835 return btrfs_ioctl_default_subvol(file, argp);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003836 case BTRFS_IOC_DEFRAG:
Chris Mason1e701a32010-03-11 09:42:04 -05003837 return btrfs_ioctl_defrag(file, NULL);
3838 case BTRFS_IOC_DEFRAG_RANGE:
3839 return btrfs_ioctl_defrag(file, argp);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003840 case BTRFS_IOC_RESIZE:
Miao Xie198605a2012-11-26 08:43:45 +00003841 return btrfs_ioctl_resize(file, argp);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003842 case BTRFS_IOC_ADD_DEV:
Christoph Hellwig4bcabaa2008-12-02 06:36:08 -05003843 return btrfs_ioctl_add_dev(root, argp);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003844 case BTRFS_IOC_RM_DEV:
Christoph Hellwig4bcabaa2008-12-02 06:36:08 -05003845 return btrfs_ioctl_rm_dev(root, argp);
Jan Schmidt475f6382011-03-11 15:41:01 +01003846 case BTRFS_IOC_FS_INFO:
3847 return btrfs_ioctl_fs_info(root, argp);
3848 case BTRFS_IOC_DEV_INFO:
3849 return btrfs_ioctl_dev_info(root, argp);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003850 case BTRFS_IOC_BALANCE:
Liu Bo9ba1f6e2012-05-11 18:11:26 +08003851 return btrfs_ioctl_balance(file, NULL);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003852 case BTRFS_IOC_CLONE:
Sage Weilc5c9cd42008-11-12 14:32:25 -05003853 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
3854 case BTRFS_IOC_CLONE_RANGE:
Christoph Hellwig7a865e82008-12-02 09:52:24 -05003855 return btrfs_ioctl_clone_range(file, argp);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003856 case BTRFS_IOC_TRANS_START:
3857 return btrfs_ioctl_trans_start(file);
3858 case BTRFS_IOC_TRANS_END:
3859 return btrfs_ioctl_trans_end(file);
Chris Masonac8e9812010-02-28 15:39:26 -05003860 case BTRFS_IOC_TREE_SEARCH:
3861 return btrfs_ioctl_tree_search(file, argp);
3862 case BTRFS_IOC_INO_LOOKUP:
3863 return btrfs_ioctl_ino_lookup(file, argp);
Jan Schmidtd7728c92011-07-07 16:48:38 +02003864 case BTRFS_IOC_INO_PATHS:
3865 return btrfs_ioctl_ino_to_path(root, argp);
3866 case BTRFS_IOC_LOGICAL_INO:
3867 return btrfs_ioctl_logical_to_ino(root, argp);
Josef Bacik1406e432010-01-13 18:19:06 +00003868 case BTRFS_IOC_SPACE_INFO:
3869 return btrfs_ioctl_space_info(root, argp);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003870 case BTRFS_IOC_SYNC:
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003871 btrfs_sync_fs(file->f_dentry->d_sb, 1);
3872 return 0;
Sage Weil46204592010-10-29 15:41:32 -04003873 case BTRFS_IOC_START_SYNC:
Miao Xie9a8c28b2012-11-26 08:40:43 +00003874 return btrfs_ioctl_start_sync(root, argp);
Sage Weil46204592010-10-29 15:41:32 -04003875 case BTRFS_IOC_WAIT_SYNC:
Miao Xie9a8c28b2012-11-26 08:40:43 +00003876 return btrfs_ioctl_wait_sync(root, argp);
Jan Schmidt475f6382011-03-11 15:41:01 +01003877 case BTRFS_IOC_SCRUB:
3878 return btrfs_ioctl_scrub(root, argp);
3879 case BTRFS_IOC_SCRUB_CANCEL:
3880 return btrfs_ioctl_scrub_cancel(root, argp);
3881 case BTRFS_IOC_SCRUB_PROGRESS:
3882 return btrfs_ioctl_scrub_progress(root, argp);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02003883 case BTRFS_IOC_BALANCE_V2:
Liu Bo9ba1f6e2012-05-11 18:11:26 +08003884 return btrfs_ioctl_balance(file, argp);
Ilya Dryomov837d5b62012-01-16 22:04:49 +02003885 case BTRFS_IOC_BALANCE_CTL:
3886 return btrfs_ioctl_balance_ctl(root, arg);
Ilya Dryomov19a39dc2012-01-16 22:04:49 +02003887 case BTRFS_IOC_BALANCE_PROGRESS:
3888 return btrfs_ioctl_balance_progress(root, argp);
Alexander Block8ea05e32012-07-25 17:35:53 +02003889 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
3890 return btrfs_ioctl_set_received_subvol(file, argp);
Alexander Block31db9f72012-07-25 23:19:24 +02003891 case BTRFS_IOC_SEND:
3892 return btrfs_ioctl_send(file, argp);
Stefan Behrensc11d2c232012-05-25 16:06:09 +02003893 case BTRFS_IOC_GET_DEV_STATS:
David Sterbab27f7c02012-06-22 06:30:39 -06003894 return btrfs_ioctl_get_dev_stats(root, argp);
Arne Jansen5d13a372011-09-14 15:53:51 +02003895 case BTRFS_IOC_QUOTA_CTL:
3896 return btrfs_ioctl_quota_ctl(root, argp);
3897 case BTRFS_IOC_QGROUP_ASSIGN:
3898 return btrfs_ioctl_qgroup_assign(root, argp);
3899 case BTRFS_IOC_QGROUP_CREATE:
3900 return btrfs_ioctl_qgroup_create(root, argp);
3901 case BTRFS_IOC_QGROUP_LIMIT:
3902 return btrfs_ioctl_qgroup_limit(root, argp);
Stefan Behrens3f6bcfb2012-11-06 15:08:53 +01003903 case BTRFS_IOC_DEV_REPLACE:
3904 return btrfs_ioctl_dev_replace(root, argp);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003905 }
3906
3907 return -ENOTTY;
3908}