blob: 48aee9846329c526260aa09f5751335c351ea1e3 [file] [log] [blame]
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
19#include <linux/kernel.h>
20#include <linux/bio.h>
21#include <linux/buffer_head.h>
22#include <linux/file.h>
23#include <linux/fs.h>
Christoph Hellwigcb8e7092008-10-09 13:39:39 -040024#include <linux/fsnotify.h>
Christoph Hellwigf46b5a62008-06-11 21:53:53 -040025#include <linux/pagemap.h>
26#include <linux/highmem.h>
27#include <linux/time.h>
28#include <linux/init.h>
29#include <linux/string.h>
Christoph Hellwigf46b5a62008-06-11 21:53:53 -040030#include <linux/backing-dev.h>
Christoph Hellwigcb8e7092008-10-09 13:39:39 -040031#include <linux/mount.h>
Christoph Hellwigf46b5a62008-06-11 21:53:53 -040032#include <linux/mpage.h>
Christoph Hellwigcb8e7092008-10-09 13:39:39 -040033#include <linux/namei.h>
Christoph Hellwigf46b5a62008-06-11 21:53:53 -040034#include <linux/swap.h>
35#include <linux/writeback.h>
36#include <linux/statfs.h>
37#include <linux/compat.h>
38#include <linux/bit_spinlock.h>
Christoph Hellwigcb8e7092008-10-09 13:39:39 -040039#include <linux/security.h>
Christoph Hellwigf46b5a62008-06-11 21:53:53 -040040#include <linux/xattr.h>
Yan Zheng7ea394f2008-08-05 13:05:02 -040041#include <linux/vmalloc.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090042#include <linux/slab.h>
Li Dongyangf7039b12011-03-24 10:24:28 +000043#include <linux/blkdev.h>
Alexander Block8ea05e32012-07-25 17:35:53 +020044#include <linux/uuid.h>
Filipe Brandenburger55e301f2013-01-29 06:04:50 +000045#include <linux/btrfs.h>
Mark Fasheh416161d2013-08-06 11:42:51 -070046#include <linux/uaccess.h>
Christoph Hellwigf46b5a62008-06-11 21:53:53 -040047#include "ctree.h"
48#include "disk-io.h"
49#include "transaction.h"
50#include "btrfs_inode.h"
Christoph Hellwigf46b5a62008-06-11 21:53:53 -040051#include "print-tree.h"
52#include "volumes.h"
Chris Mason925baed2008-06-25 16:01:30 -040053#include "locking.h"
Li Zefan581bb052011-04-20 10:06:11 +080054#include "inode-map.h"
Jan Schmidtd7728c92011-07-07 16:48:38 +020055#include "backref.h"
Josef Bacik606686e2012-06-04 14:03:51 -040056#include "rcu-string.h"
Alexander Block31db9f72012-07-25 23:19:24 +020057#include "send.h"
Stefan Behrens3f6bcfb2012-11-06 15:08:53 +010058#include "dev-replace.h"
Filipe David Borba Manana63541922014-01-07 11:47:46 +000059#include "props.h"
Jeff Mahoney3b02a682013-11-01 13:07:02 -040060#include "sysfs.h"
Josef Bacikfcebe452014-05-13 17:30:47 -070061#include "qgroup.h"
Christoph Hellwigf46b5a62008-06-11 21:53:53 -040062
Hugo Millsabccd002014-01-30 20:17:00 +000063#ifdef CONFIG_64BIT
64/* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
65 * structures are incorrect, as the timespec structure from userspace
66 * is 4 bytes too small. We define these alternatives here to teach
67 * the kernel about the 32-bit struct packing.
68 */
69struct btrfs_ioctl_timespec_32 {
70 __u64 sec;
71 __u32 nsec;
72} __attribute__ ((__packed__));
73
74struct btrfs_ioctl_received_subvol_args_32 {
75 char uuid[BTRFS_UUID_SIZE]; /* in */
76 __u64 stransid; /* in */
77 __u64 rtransid; /* out */
78 struct btrfs_ioctl_timespec_32 stime; /* in */
79 struct btrfs_ioctl_timespec_32 rtime; /* out */
80 __u64 flags; /* in */
81 __u64 reserved[16]; /* in */
82} __attribute__ ((__packed__));
83
84#define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
85 struct btrfs_ioctl_received_subvol_args_32)
86#endif
87
88
Mark Fasheh416161d2013-08-06 11:42:51 -070089static int btrfs_clone(struct inode *src, struct inode *inode,
Mark Fasheh1c919a52015-06-30 14:42:08 -070090 u64 off, u64 olen, u64 olen_aligned, u64 destoff,
91 int no_time_update);
Mark Fasheh416161d2013-08-06 11:42:51 -070092
Christoph Hellwig6cbff002009-04-17 10:37:41 +020093/* Mask out flags that are inappropriate for the given type of inode. */
94static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
95{
96 if (S_ISDIR(mode))
97 return flags;
98 else if (S_ISREG(mode))
99 return flags & ~FS_DIRSYNC_FL;
100 else
101 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
102}
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400103
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200104/*
105 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
106 */
107static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
108{
109 unsigned int iflags = 0;
110
111 if (flags & BTRFS_INODE_SYNC)
112 iflags |= FS_SYNC_FL;
113 if (flags & BTRFS_INODE_IMMUTABLE)
114 iflags |= FS_IMMUTABLE_FL;
115 if (flags & BTRFS_INODE_APPEND)
116 iflags |= FS_APPEND_FL;
117 if (flags & BTRFS_INODE_NODUMP)
118 iflags |= FS_NODUMP_FL;
119 if (flags & BTRFS_INODE_NOATIME)
120 iflags |= FS_NOATIME_FL;
121 if (flags & BTRFS_INODE_DIRSYNC)
122 iflags |= FS_DIRSYNC_FL;
Li Zefand0092bd2011-04-15 03:03:06 +0000123 if (flags & BTRFS_INODE_NODATACOW)
124 iflags |= FS_NOCOW_FL;
125
126 if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
127 iflags |= FS_COMPR_FL;
128 else if (flags & BTRFS_INODE_NOCOMPRESS)
129 iflags |= FS_NOCOMP_FL;
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200130
131 return iflags;
132}
133
134/*
135 * Update inode->i_flags based on the btrfs internal flags.
136 */
137void btrfs_update_iflags(struct inode *inode)
138{
139 struct btrfs_inode *ip = BTRFS_I(inode);
Filipe Manana3cc79392014-06-25 22:36:02 +0100140 unsigned int new_fl = 0;
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200141
142 if (ip->flags & BTRFS_INODE_SYNC)
Filipe Manana3cc79392014-06-25 22:36:02 +0100143 new_fl |= S_SYNC;
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200144 if (ip->flags & BTRFS_INODE_IMMUTABLE)
Filipe Manana3cc79392014-06-25 22:36:02 +0100145 new_fl |= S_IMMUTABLE;
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200146 if (ip->flags & BTRFS_INODE_APPEND)
Filipe Manana3cc79392014-06-25 22:36:02 +0100147 new_fl |= S_APPEND;
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200148 if (ip->flags & BTRFS_INODE_NOATIME)
Filipe Manana3cc79392014-06-25 22:36:02 +0100149 new_fl |= S_NOATIME;
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200150 if (ip->flags & BTRFS_INODE_DIRSYNC)
Filipe Manana3cc79392014-06-25 22:36:02 +0100151 new_fl |= S_DIRSYNC;
152
153 set_mask_bits(&inode->i_flags,
154 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC,
155 new_fl);
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200156}
157
158/*
159 * Inherit flags from the parent inode.
160 *
Josef Bacike27425d2011-09-27 11:01:30 -0400161 * Currently only the compression flags and the cow flags are inherited.
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200162 */
163void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
164{
Chris Mason0b4dcea2009-06-11 11:13:35 -0400165 unsigned int flags;
166
167 if (!dir)
168 return;
169
170 flags = BTRFS_I(dir)->flags;
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200171
Josef Bacike27425d2011-09-27 11:01:30 -0400172 if (flags & BTRFS_INODE_NOCOMPRESS) {
173 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
174 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
175 } else if (flags & BTRFS_INODE_COMPRESS) {
176 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
177 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
178 }
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200179
Liu Bo213490b2012-09-11 08:33:50 -0600180 if (flags & BTRFS_INODE_NODATACOW) {
Josef Bacike27425d2011-09-27 11:01:30 -0400181 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
Liu Bo213490b2012-09-11 08:33:50 -0600182 if (S_ISREG(inode->i_mode))
183 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
184 }
Josef Bacike27425d2011-09-27 11:01:30 -0400185
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200186 btrfs_update_iflags(inode);
187}
188
189static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
190{
Al Viro496ad9a2013-01-23 17:07:38 -0500191 struct btrfs_inode *ip = BTRFS_I(file_inode(file));
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200192 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
193
194 if (copy_to_user(arg, &flags, sizeof(flags)))
195 return -EFAULT;
196 return 0;
197}
198
Liu Bo75e7cb72011-03-22 10:12:20 +0000199static int check_flags(unsigned int flags)
200{
201 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
202 FS_NOATIME_FL | FS_NODUMP_FL | \
203 FS_SYNC_FL | FS_DIRSYNC_FL | \
Li Zefane1e8fb62011-04-15 03:02:49 +0000204 FS_NOCOMP_FL | FS_COMPR_FL |
205 FS_NOCOW_FL))
Liu Bo75e7cb72011-03-22 10:12:20 +0000206 return -EOPNOTSUPP;
207
208 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
209 return -EINVAL;
210
Liu Bo75e7cb72011-03-22 10:12:20 +0000211 return 0;
212}
213
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200214static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
215{
Al Viro496ad9a2013-01-23 17:07:38 -0500216 struct inode *inode = file_inode(file);
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200217 struct btrfs_inode *ip = BTRFS_I(inode);
218 struct btrfs_root *root = ip->root;
219 struct btrfs_trans_handle *trans;
220 unsigned int flags, oldflags;
221 int ret;
Li Zefanf062abf2011-12-29 13:36:45 +0800222 u64 ip_oldflags;
223 unsigned int i_oldflags;
David Sterba7e97b8d2012-09-07 05:56:55 -0600224 umode_t mode;
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200225
David Sterbabd60ea02014-01-16 15:50:22 +0100226 if (!inode_owner_or_capable(inode))
227 return -EPERM;
228
Li Zefanb83cc962010-12-20 16:04:08 +0800229 if (btrfs_root_readonly(root))
230 return -EROFS;
231
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200232 if (copy_from_user(&flags, arg, sizeof(flags)))
233 return -EFAULT;
234
Liu Bo75e7cb72011-03-22 10:12:20 +0000235 ret = check_flags(flags);
236 if (ret)
237 return ret;
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200238
Jan Karae7848682012-06-12 16:20:32 +0200239 ret = mnt_want_write_file(file);
240 if (ret)
241 return ret;
242
Al Viro59551022016-01-22 15:40:57 -0500243 inode_lock(inode);
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200244
Li Zefanf062abf2011-12-29 13:36:45 +0800245 ip_oldflags = ip->flags;
246 i_oldflags = inode->i_flags;
David Sterba7e97b8d2012-09-07 05:56:55 -0600247 mode = inode->i_mode;
Li Zefanf062abf2011-12-29 13:36:45 +0800248
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200249 flags = btrfs_mask_flags(inode->i_mode, flags);
250 oldflags = btrfs_flags_to_ioctl(ip->flags);
251 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
252 if (!capable(CAP_LINUX_IMMUTABLE)) {
253 ret = -EPERM;
254 goto out_unlock;
255 }
256 }
257
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200258 if (flags & FS_SYNC_FL)
259 ip->flags |= BTRFS_INODE_SYNC;
260 else
261 ip->flags &= ~BTRFS_INODE_SYNC;
262 if (flags & FS_IMMUTABLE_FL)
263 ip->flags |= BTRFS_INODE_IMMUTABLE;
264 else
265 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
266 if (flags & FS_APPEND_FL)
267 ip->flags |= BTRFS_INODE_APPEND;
268 else
269 ip->flags &= ~BTRFS_INODE_APPEND;
270 if (flags & FS_NODUMP_FL)
271 ip->flags |= BTRFS_INODE_NODUMP;
272 else
273 ip->flags &= ~BTRFS_INODE_NODUMP;
274 if (flags & FS_NOATIME_FL)
275 ip->flags |= BTRFS_INODE_NOATIME;
276 else
277 ip->flags &= ~BTRFS_INODE_NOATIME;
278 if (flags & FS_DIRSYNC_FL)
279 ip->flags |= BTRFS_INODE_DIRSYNC;
280 else
281 ip->flags &= ~BTRFS_INODE_DIRSYNC;
David Sterba7e97b8d2012-09-07 05:56:55 -0600282 if (flags & FS_NOCOW_FL) {
283 if (S_ISREG(mode)) {
284 /*
285 * It's safe to turn csums off here, no extents exist.
286 * Otherwise we want the flag to reflect the real COW
287 * status of the file and will not set it.
288 */
289 if (inode->i_size == 0)
290 ip->flags |= BTRFS_INODE_NODATACOW
291 | BTRFS_INODE_NODATASUM;
292 } else {
293 ip->flags |= BTRFS_INODE_NODATACOW;
294 }
295 } else {
296 /*
297 * Revert back under same assuptions as above
298 */
299 if (S_ISREG(mode)) {
300 if (inode->i_size == 0)
301 ip->flags &= ~(BTRFS_INODE_NODATACOW
302 | BTRFS_INODE_NODATASUM);
303 } else {
304 ip->flags &= ~BTRFS_INODE_NODATACOW;
305 }
306 }
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200307
Liu Bo75e7cb72011-03-22 10:12:20 +0000308 /*
309 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
310 * flag may be changed automatically if compression code won't make
311 * things smaller.
312 */
313 if (flags & FS_NOCOMP_FL) {
314 ip->flags &= ~BTRFS_INODE_COMPRESS;
315 ip->flags |= BTRFS_INODE_NOCOMPRESS;
Filipe David Borba Manana63541922014-01-07 11:47:46 +0000316
317 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
318 if (ret && ret != -ENODATA)
319 goto out_drop;
Liu Bo75e7cb72011-03-22 10:12:20 +0000320 } else if (flags & FS_COMPR_FL) {
Filipe David Borba Manana63541922014-01-07 11:47:46 +0000321 const char *comp;
322
Liu Bo75e7cb72011-03-22 10:12:20 +0000323 ip->flags |= BTRFS_INODE_COMPRESS;
324 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
Filipe David Borba Manana63541922014-01-07 11:47:46 +0000325
326 if (root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
327 comp = "lzo";
328 else
329 comp = "zlib";
330 ret = btrfs_set_prop(inode, "btrfs.compression",
331 comp, strlen(comp), 0);
332 if (ret)
333 goto out_drop;
334
Li Zefanebcb9042011-04-15 03:03:17 +0000335 } else {
Filipe Manana78a017a2014-09-11 11:44:49 +0100336 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
337 if (ret && ret != -ENODATA)
338 goto out_drop;
Li Zefanebcb9042011-04-15 03:03:17 +0000339 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
Liu Bo75e7cb72011-03-22 10:12:20 +0000340 }
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200341
Li Zefan4da6f1a2011-12-29 13:39:50 +0800342 trans = btrfs_start_transaction(root, 1);
Li Zefanf062abf2011-12-29 13:36:45 +0800343 if (IS_ERR(trans)) {
344 ret = PTR_ERR(trans);
345 goto out_drop;
346 }
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200347
Li Zefan306424cc2011-12-14 20:12:02 -0500348 btrfs_update_iflags(inode);
Josef Bacik0c4d2d92012-04-05 15:03:02 -0400349 inode_inc_iversion(inode);
Li Zefan306424cc2011-12-14 20:12:02 -0500350 inode->i_ctime = CURRENT_TIME;
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200351 ret = btrfs_update_inode(trans, root, inode);
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200352
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200353 btrfs_end_transaction(trans, root);
Li Zefanf062abf2011-12-29 13:36:45 +0800354 out_drop:
355 if (ret) {
356 ip->flags = ip_oldflags;
357 inode->i_flags = i_oldflags;
358 }
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200359
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200360 out_unlock:
Al Viro59551022016-01-22 15:40:57 -0500361 inode_unlock(inode);
Jan Karae7848682012-06-12 16:20:32 +0200362 mnt_drop_write_file(file);
liubo2d4e6f62011-02-24 09:38:16 +0000363 return ret;
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200364}
365
366static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
367{
Al Viro496ad9a2013-01-23 17:07:38 -0500368 struct inode *inode = file_inode(file);
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200369
370 return put_user(inode->i_generation, arg);
371}
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400372
Li Dongyangf7039b12011-03-24 10:24:28 +0000373static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
374{
Al Viro54563d42013-09-01 15:57:51 -0400375 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
Li Dongyangf7039b12011-03-24 10:24:28 +0000376 struct btrfs_device *device;
377 struct request_queue *q;
378 struct fstrim_range range;
379 u64 minlen = ULLONG_MAX;
380 u64 num_devices = 0;
Al Viro815745c2011-11-17 15:40:49 -0500381 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
Li Dongyangf7039b12011-03-24 10:24:28 +0000382 int ret;
383
384 if (!capable(CAP_SYS_ADMIN))
385 return -EPERM;
386
Xiao Guangrong1f781602011-04-20 10:09:16 +0000387 rcu_read_lock();
388 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
389 dev_list) {
Li Dongyangf7039b12011-03-24 10:24:28 +0000390 if (!device->bdev)
391 continue;
392 q = bdev_get_queue(device->bdev);
393 if (blk_queue_discard(q)) {
394 num_devices++;
395 minlen = min((u64)q->limits.discard_granularity,
396 minlen);
397 }
398 }
Xiao Guangrong1f781602011-04-20 10:09:16 +0000399 rcu_read_unlock();
Lukas Czernerf4c697e2011-09-05 16:34:54 +0200400
Li Dongyangf7039b12011-03-24 10:24:28 +0000401 if (!num_devices)
402 return -EOPNOTSUPP;
Li Dongyangf7039b12011-03-24 10:24:28 +0000403 if (copy_from_user(&range, arg, sizeof(range)))
404 return -EFAULT;
Lukas Czernere515c182012-10-16 09:34:36 +0000405 if (range.start > total_bytes ||
406 range.len < fs_info->sb->s_blocksize)
Lukas Czernerf4c697e2011-09-05 16:34:54 +0200407 return -EINVAL;
Li Dongyangf7039b12011-03-24 10:24:28 +0000408
Lukas Czernerf4c697e2011-09-05 16:34:54 +0200409 range.len = min(range.len, total_bytes - range.start);
Li Dongyangf7039b12011-03-24 10:24:28 +0000410 range.minlen = max(range.minlen, minlen);
Al Viro815745c2011-11-17 15:40:49 -0500411 ret = btrfs_trim_fs(fs_info->tree_root, &range);
Li Dongyangf7039b12011-03-24 10:24:28 +0000412 if (ret < 0)
413 return ret;
414
415 if (copy_to_user(arg, &range, sizeof(range)))
416 return -EFAULT;
417
418 return 0;
419}
420
Stefan Behrensdd5f9612013-08-15 17:11:20 +0200421int btrfs_is_empty_uuid(u8 *uuid)
422{
Chris Mason46e0f662013-11-15 12:14:55 +0100423 int i;
424
425 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
426 if (uuid[i])
427 return 0;
428 }
429 return 1;
Stefan Behrensdd5f9612013-08-15 17:11:20 +0200430}
431
Miao Xied5c12072013-02-28 10:04:33 +0000432static noinline int create_subvol(struct inode *dir,
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400433 struct dentry *dentry,
Sage Weil72fd0322010-10-29 15:41:32 -0400434 char *name, int namelen,
Arne Jansen6f72c7e2011-09-14 15:58:21 +0200435 u64 *async_transid,
Miao Xie8696c532013-02-07 06:02:44 +0000436 struct btrfs_qgroup_inherit *inherit)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400437{
438 struct btrfs_trans_handle *trans;
439 struct btrfs_key key;
440 struct btrfs_root_item root_item;
441 struct btrfs_inode_item *inode_item;
442 struct extent_buffer *leaf;
Miao Xied5c12072013-02-28 10:04:33 +0000443 struct btrfs_root *root = BTRFS_I(dir)->root;
Yan, Zheng76dda932009-09-21 16:00:26 -0400444 struct btrfs_root *new_root;
Miao Xied5c12072013-02-28 10:04:33 +0000445 struct btrfs_block_rsv block_rsv;
Alexander Block8ea05e32012-07-25 17:35:53 +0200446 struct timespec cur_time = CURRENT_TIME;
Tsutomu Itoh5662344b32013-12-13 09:51:42 +0900447 struct inode *inode;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400448 int ret;
449 int err;
450 u64 objectid;
451 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
Chris Mason3de45862008-11-17 21:02:50 -0500452 u64 index = 0;
Miao Xied5c12072013-02-28 10:04:33 +0000453 u64 qgroup_reserved;
Alexander Block8ea05e32012-07-25 17:35:53 +0200454 uuid_le new_uuid;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400455
Li Zefan581bb052011-04-20 10:06:11 +0800456 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
Al Viro2fbe8c82011-07-16 21:38:06 -0400457 if (ret)
Yan, Zhenga22285a2010-05-16 10:48:46 -0400458 return ret;
Josef Bacik6a912212010-11-20 09:48:00 +0000459
Qu Wenruoe09fe2d2015-02-27 16:24:23 +0800460 /*
461 * Don't create subvolume whose level is not zero. Or qgroup will be
462 * screwed up since it assume subvolme qgroup's level to be 0.
463 */
464 if (btrfs_qgroup_level(objectid))
465 return -ENOSPC;
466
Miao Xied5c12072013-02-28 10:04:33 +0000467 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
Josef Bacik9ed74f22009-09-11 16:12:44 -0400468 /*
Miao Xied5c12072013-02-28 10:04:33 +0000469 * The same as the snapshot creation, please see the comment
470 * of create_snapshot().
Josef Bacik9ed74f22009-09-11 16:12:44 -0400471 */
Miao Xied5c12072013-02-28 10:04:33 +0000472 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
Stefan Behrensdd5f9612013-08-15 17:11:20 +0200473 8, &qgroup_reserved, false);
Miao Xied5c12072013-02-28 10:04:33 +0000474 if (ret)
475 return ret;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400476
Miao Xied5c12072013-02-28 10:04:33 +0000477 trans = btrfs_start_transaction(root, 0);
478 if (IS_ERR(trans)) {
479 ret = PTR_ERR(trans);
Liu Bode6e8202014-01-09 14:57:06 +0800480 btrfs_subvolume_release_metadata(root, &block_rsv,
481 qgroup_reserved);
482 return ret;
Miao Xied5c12072013-02-28 10:04:33 +0000483 }
484 trans->block_rsv = &block_rsv;
485 trans->bytes_reserved = block_rsv.size;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400486
Miao Xie8696c532013-02-07 06:02:44 +0000487 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
Arne Jansen6f72c7e2011-09-14 15:58:21 +0200488 if (ret)
489 goto fail;
490
David Sterba4d75f8a2014-06-15 01:54:12 +0200491 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
Josef Bacik8e8a1e32008-07-24 12:17:14 -0400492 if (IS_ERR(leaf)) {
493 ret = PTR_ERR(leaf);
494 goto fail;
495 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400496
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400497 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400498 btrfs_set_header_bytenr(leaf, leaf->start);
499 btrfs_set_header_generation(leaf, trans->transid);
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400500 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400501 btrfs_set_header_owner(leaf, objectid);
502
Ross Kirk0a4e5582013-09-24 10:12:38 +0100503 write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(),
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400504 BTRFS_FSID_SIZE);
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400505 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
Geert Uytterhoevenb308bc22013-08-20 13:20:15 +0200506 btrfs_header_chunk_tree_uuid(leaf),
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400507 BTRFS_UUID_SIZE);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400508 btrfs_mark_buffer_dirty(leaf);
509
Alexander Block8ea05e32012-07-25 17:35:53 +0200510 memset(&root_item, 0, sizeof(root_item));
511
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400512 inode_item = &root_item.inode;
Qu Wenruo3cae2102013-07-16 11:19:18 +0800513 btrfs_set_stack_inode_generation(inode_item, 1);
514 btrfs_set_stack_inode_size(inode_item, 3);
515 btrfs_set_stack_inode_nlink(inode_item, 1);
David Sterba707e8a02014-06-04 19:22:26 +0200516 btrfs_set_stack_inode_nbytes(inode_item, root->nodesize);
Qu Wenruo3cae2102013-07-16 11:19:18 +0800517 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400518
Qu Wenruo3cae2102013-07-16 11:19:18 +0800519 btrfs_set_root_flags(&root_item, 0);
520 btrfs_set_root_limit(&root_item, 0);
521 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
Li Zefan08fe4db2011-03-28 02:01:25 +0000522
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400523 btrfs_set_root_bytenr(&root_item, leaf->start);
Yan Zheng84234f32008-10-29 14:49:05 -0400524 btrfs_set_root_generation(&root_item, trans->transid);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400525 btrfs_set_root_level(&root_item, 0);
526 btrfs_set_root_refs(&root_item, 1);
Yan, Zheng86b9f2e2009-11-12 09:36:50 +0000527 btrfs_set_root_used(&root_item, leaf->len);
Yan Zheng80ff3852008-10-30 14:20:02 -0400528 btrfs_set_root_last_snapshot(&root_item, 0);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400529
Alexander Block8ea05e32012-07-25 17:35:53 +0200530 btrfs_set_root_generation_v2(&root_item,
531 btrfs_root_generation(&root_item));
532 uuid_le_gen(&new_uuid);
533 memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
Qu Wenruo3cae2102013-07-16 11:19:18 +0800534 btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec);
535 btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec);
Alexander Block8ea05e32012-07-25 17:35:53 +0200536 root_item.ctime = root_item.otime;
537 btrfs_set_root_ctransid(&root_item, trans->transid);
538 btrfs_set_root_otransid(&root_item, trans->transid);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400539
Chris Mason925baed2008-06-25 16:01:30 -0400540 btrfs_tree_unlock(leaf);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400541 free_extent_buffer(leaf);
542 leaf = NULL;
543
544 btrfs_set_root_dirid(&root_item, new_dirid);
545
546 key.objectid = objectid;
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400547 key.offset = 0;
David Sterba962a2982014-06-04 18:41:45 +0200548 key.type = BTRFS_ROOT_ITEM_KEY;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400549 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
550 &root_item);
551 if (ret)
552 goto fail;
553
Yan, Zheng76dda932009-09-21 16:00:26 -0400554 key.offset = (u64)-1;
555 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100556 if (IS_ERR(new_root)) {
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100557 ret = PTR_ERR(new_root);
David Sterba6d13f542015-04-24 19:12:01 +0200558 btrfs_abort_transaction(trans, root, ret);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100559 goto fail;
560 }
Yan, Zheng76dda932009-09-21 16:00:26 -0400561
562 btrfs_record_root_in_trans(trans, new_root);
563
Filipe David Borba Manana63541922014-01-07 11:47:46 +0000564 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
Mark Fashehce598972011-07-26 11:32:23 -0700565 if (ret) {
566 /* We potentially lose an unused inode item here */
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100567 btrfs_abort_transaction(trans, root, ret);
Mark Fashehce598972011-07-26 11:32:23 -0700568 goto fail;
569 }
570
Chandan Rajendraf32e48e2016-01-07 18:56:59 +0530571 mutex_lock(&new_root->objectid_mutex);
572 new_root->highest_objectid = new_dirid;
573 mutex_unlock(&new_root->objectid_mutex);
574
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400575 /*
576 * insert the directory item
577 */
Chris Mason3de45862008-11-17 21:02:50 -0500578 ret = btrfs_set_inode_index(dir, &index);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100579 if (ret) {
580 btrfs_abort_transaction(trans, root, ret);
581 goto fail;
582 }
Chris Mason3de45862008-11-17 21:02:50 -0500583
584 ret = btrfs_insert_dir_item(trans, root,
Miao Xie16cdcec2011-04-22 18:12:22 +0800585 name, namelen, dir, &key,
Chris Mason3de45862008-11-17 21:02:50 -0500586 BTRFS_FT_DIR, index);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100587 if (ret) {
588 btrfs_abort_transaction(trans, root, ret);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400589 goto fail;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100590 }
Chris Mason0660b5a2008-11-17 20:37:39 -0500591
Yan Zheng52c26172009-01-05 15:43:43 -0500592 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
593 ret = btrfs_update_inode(trans, root, dir);
594 BUG_ON(ret);
595
Chris Mason0660b5a2008-11-17 20:37:39 -0500596 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
Yan, Zheng4df27c4d2009-09-21 15:56:00 -0400597 objectid, root->root_key.objectid,
Li Zefan33345d012011-04-20 10:31:50 +0800598 btrfs_ino(dir), index, name, namelen);
Chris Mason0660b5a2008-11-17 20:37:39 -0500599 BUG_ON(ret);
600
Stefan Behrensdd5f9612013-08-15 17:11:20 +0200601 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
602 root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
603 objectid);
604 if (ret)
605 btrfs_abort_transaction(trans, root, ret);
606
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400607fail:
Miao Xied5c12072013-02-28 10:04:33 +0000608 trans->block_rsv = NULL;
609 trans->bytes_reserved = 0;
Liu Bode6e8202014-01-09 14:57:06 +0800610 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
611
Sage Weil72fd0322010-10-29 15:41:32 -0400612 if (async_transid) {
613 *async_transid = trans->transid;
614 err = btrfs_commit_transaction_async(trans, root, 1);
Miao Xie00d71c92013-03-04 09:45:06 +0000615 if (err)
616 err = btrfs_commit_transaction(trans, root);
Sage Weil72fd0322010-10-29 15:41:32 -0400617 } else {
618 err = btrfs_commit_transaction(trans, root);
619 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400620 if (err && !ret)
621 ret = err;
Chris Mason1a65e242013-02-06 12:06:02 -0500622
Tsutomu Itoh5662344b32013-12-13 09:51:42 +0900623 if (!ret) {
624 inode = btrfs_lookup_dentry(dir, dentry);
Liu Bode6e8202014-01-09 14:57:06 +0800625 if (IS_ERR(inode))
626 return PTR_ERR(inode);
Tsutomu Itoh5662344b32013-12-13 09:51:42 +0900627 d_instantiate(dentry, inode);
628 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400629 return ret;
630}
631
Filipe Manana9ea24bb2014-10-29 11:57:59 +0000632static void btrfs_wait_for_no_snapshoting_writes(struct btrfs_root *root)
Miao Xie8257b2d2014-03-06 13:38:19 +0800633{
634 s64 writers;
635 DEFINE_WAIT(wait);
636
637 do {
638 prepare_to_wait(&root->subv_writers->wait, &wait,
639 TASK_UNINTERRUPTIBLE);
640
641 writers = percpu_counter_sum(&root->subv_writers->counter);
642 if (writers)
643 schedule();
644
645 finish_wait(&root->subv_writers->wait, &wait);
646 } while (writers);
647}
648
Miao Xiee9662f72013-02-28 10:01:15 +0000649static int create_snapshot(struct btrfs_root *root, struct inode *dir,
650 struct dentry *dentry, char *name, int namelen,
651 u64 *async_transid, bool readonly,
652 struct btrfs_qgroup_inherit *inherit)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400653{
Yan, Zheng2e4bfab2009-11-12 09:37:02 +0000654 struct inode *inode;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400655 struct btrfs_pending_snapshot *pending_snapshot;
656 struct btrfs_trans_handle *trans;
Yan, Zheng2e4bfab2009-11-12 09:37:02 +0000657 int ret;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400658
Miao Xie27cdeb72014-04-02 19:51:05 +0800659 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400660 return -EINVAL;
661
David Sterbaa1ee7362015-11-10 18:53:56 +0100662 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
663 if (!pending_snapshot)
664 return -ENOMEM;
665
David Sterbab0c0ea62015-11-10 18:54:00 +0100666 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
667 GFP_NOFS);
David Sterba8546b572015-11-10 18:54:03 +0100668 pending_snapshot->path = btrfs_alloc_path();
669 if (!pending_snapshot->root_item || !pending_snapshot->path) {
David Sterbab0c0ea62015-11-10 18:54:00 +0100670 ret = -ENOMEM;
671 goto free_pending;
672 }
673
Miao Xie8257b2d2014-03-06 13:38:19 +0800674 atomic_inc(&root->will_be_snapshoted);
Peter Zijlstra4e857c52014-03-17 18:06:10 +0100675 smp_mb__after_atomic();
Filipe Manana9ea24bb2014-10-29 11:57:59 +0000676 btrfs_wait_for_no_snapshoting_writes(root);
Miao Xie8257b2d2014-03-06 13:38:19 +0800677
Miao Xie6a038432013-05-15 07:48:24 +0000678 ret = btrfs_start_delalloc_inodes(root, 0);
679 if (ret)
David Sterbaa1ee7362015-11-10 18:53:56 +0100680 goto dec_and_free;
Miao Xie6a038432013-05-15 07:48:24 +0000681
Miao Xieb0244192013-11-04 23:13:25 +0800682 btrfs_wait_ordered_extents(root, -1);
Miao Xie6a038432013-05-15 07:48:24 +0000683
Miao Xie66d8f3d2012-09-06 04:02:28 -0600684 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
685 BTRFS_BLOCK_RSV_TEMP);
Miao Xied5c12072013-02-28 10:04:33 +0000686 /*
687 * 1 - parent dir inode
688 * 2 - dir entries
689 * 1 - root item
690 * 2 - root ref/backref
691 * 1 - root of snapshot
Stefan Behrensdd5f9612013-08-15 17:11:20 +0200692 * 1 - UUID item
Miao Xied5c12072013-02-28 10:04:33 +0000693 */
694 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
Stefan Behrensdd5f9612013-08-15 17:11:20 +0200695 &pending_snapshot->block_rsv, 8,
Jeff Mahoneyee3441b2013-07-09 16:37:21 -0400696 &pending_snapshot->qgroup_reserved,
697 false);
Miao Xied5c12072013-02-28 10:04:33 +0000698 if (ret)
David Sterbaa1ee7362015-11-10 18:53:56 +0100699 goto dec_and_free;
Miao Xied5c12072013-02-28 10:04:33 +0000700
Yan, Zhenga22285a2010-05-16 10:48:46 -0400701 pending_snapshot->dentry = dentry;
702 pending_snapshot->root = root;
Li Zefanb83cc962010-12-20 16:04:08 +0800703 pending_snapshot->readonly = readonly;
Miao Xiee9662f72013-02-28 10:01:15 +0000704 pending_snapshot->dir = dir;
Miao Xie8696c532013-02-07 06:02:44 +0000705 pending_snapshot->inherit = inherit;
Yan, Zhenga22285a2010-05-16 10:48:46 -0400706
Miao Xied5c12072013-02-28 10:04:33 +0000707 trans = btrfs_start_transaction(root, 0);
Yan, Zhenga22285a2010-05-16 10:48:46 -0400708 if (IS_ERR(trans)) {
709 ret = PTR_ERR(trans);
710 goto fail;
711 }
712
Josef Bacik83515832011-06-14 15:16:14 -0400713 spin_lock(&root->fs_info->trans_lock);
Yan, Zhenga22285a2010-05-16 10:48:46 -0400714 list_add(&pending_snapshot->list,
715 &trans->transaction->pending_snapshots);
Josef Bacik83515832011-06-14 15:16:14 -0400716 spin_unlock(&root->fs_info->trans_lock);
Sage Weil72fd0322010-10-29 15:41:32 -0400717 if (async_transid) {
718 *async_transid = trans->transid;
719 ret = btrfs_commit_transaction_async(trans,
720 root->fs_info->extent_root, 1);
Miao Xie00d71c92013-03-04 09:45:06 +0000721 if (ret)
722 ret = btrfs_commit_transaction(trans, root);
Sage Weil72fd0322010-10-29 15:41:32 -0400723 } else {
724 ret = btrfs_commit_transaction(trans,
725 root->fs_info->extent_root);
726 }
Miao Xieaec80302013-03-04 09:44:29 +0000727 if (ret)
Josef Bacikc37b2b62012-10-22 15:51:44 -0400728 goto fail;
Yan, Zhenga22285a2010-05-16 10:48:46 -0400729
730 ret = pending_snapshot->error;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400731 if (ret)
Yan, Zheng2e4bfab2009-11-12 09:37:02 +0000732 goto fail;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400733
Chris Masond3797302014-10-15 13:50:56 -0700734 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
735 if (ret)
736 goto fail;
737
David Howells2b0143b2015-03-17 22:25:59 +0000738 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
Yan, Zheng2e4bfab2009-11-12 09:37:02 +0000739 if (IS_ERR(inode)) {
740 ret = PTR_ERR(inode);
741 goto fail;
742 }
Tsutomu Itoh5662344b32013-12-13 09:51:42 +0900743
Yan, Zheng2e4bfab2009-11-12 09:37:02 +0000744 d_instantiate(dentry, inode);
745 ret = 0;
746fail:
Miao Xied5c12072013-02-28 10:04:33 +0000747 btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
748 &pending_snapshot->block_rsv,
749 pending_snapshot->qgroup_reserved);
David Sterbaa1ee7362015-11-10 18:53:56 +0100750dec_and_free:
Filipe Manana9ea24bb2014-10-29 11:57:59 +0000751 if (atomic_dec_and_test(&root->will_be_snapshoted))
752 wake_up_atomic_t(&root->will_be_snapshoted);
David Sterbab0c0ea62015-11-10 18:54:00 +0100753free_pending:
754 kfree(pending_snapshot->root_item);
David Sterba8546b572015-11-10 18:54:03 +0100755 btrfs_free_path(pending_snapshot->path);
David Sterbaa1ee7362015-11-10 18:53:56 +0100756 kfree(pending_snapshot);
757
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400758 return ret;
759}
760
Sage Weil4260f7c2010-10-29 15:46:43 -0400761/* copy of may_delete in fs/namei.c()
762 * Check whether we can remove a link victim from directory dir, check
763 * whether the type of victim is right.
764 * 1. We can't do it if dir is read-only (done in permission())
765 * 2. We should have write and exec permissions on dir
766 * 3. We can't remove anything from append-only dir
767 * 4. We can't do anything with immutable dir (done in permission())
768 * 5. If the sticky bit on dir is set we should either
769 * a. be owner of dir, or
770 * b. be owner of victim, or
771 * c. have CAP_FOWNER capability
772 * 6. If the victim is append-only or immutable we can't do antyhing with
773 * links pointing to it.
774 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
775 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
776 * 9. We can't remove a root or mountpoint.
777 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
778 * nfs_async_unlink().
779 */
780
Dulshani Gunawardhana67871252013-10-31 10:33:04 +0530781static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
Sage Weil4260f7c2010-10-29 15:46:43 -0400782{
783 int error;
784
David Howells2b0143b2015-03-17 22:25:59 +0000785 if (d_really_is_negative(victim))
Sage Weil4260f7c2010-10-29 15:46:43 -0400786 return -ENOENT;
787
David Howells2b0143b2015-03-17 22:25:59 +0000788 BUG_ON(d_inode(victim->d_parent) != dir);
Jeff Layton4fa6b5e2012-10-10 15:25:25 -0400789 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
Sage Weil4260f7c2010-10-29 15:46:43 -0400790
791 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
792 if (error)
793 return error;
794 if (IS_APPEND(dir))
795 return -EPERM;
David Howells2b0143b2015-03-17 22:25:59 +0000796 if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) ||
797 IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim)))
Sage Weil4260f7c2010-10-29 15:46:43 -0400798 return -EPERM;
799 if (isdir) {
David Howellse36cb0b2015-01-29 12:02:35 +0000800 if (!d_is_dir(victim))
Sage Weil4260f7c2010-10-29 15:46:43 -0400801 return -ENOTDIR;
802 if (IS_ROOT(victim))
803 return -EBUSY;
David Howellse36cb0b2015-01-29 12:02:35 +0000804 } else if (d_is_dir(victim))
Sage Weil4260f7c2010-10-29 15:46:43 -0400805 return -EISDIR;
806 if (IS_DEADDIR(dir))
807 return -ENOENT;
808 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
809 return -EBUSY;
810 return 0;
811}
812
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400813/* copy of may_create in fs/namei.c() */
814static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
815{
David Howells2b0143b2015-03-17 22:25:59 +0000816 if (d_really_is_positive(child))
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400817 return -EEXIST;
818 if (IS_DEADDIR(dir))
819 return -ENOENT;
820 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
821}
822
823/*
824 * Create a new subvolume below @parent. This is largely modeled after
825 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
826 * inside this filesystem so it's quite a bit simpler.
827 */
Yan, Zheng76dda932009-09-21 16:00:26 -0400828static noinline int btrfs_mksubvol(struct path *parent,
829 char *name, int namelen,
Sage Weil72fd0322010-10-29 15:41:32 -0400830 struct btrfs_root *snap_src,
Arne Jansen6f72c7e2011-09-14 15:58:21 +0200831 u64 *async_transid, bool readonly,
Miao Xie8696c532013-02-07 06:02:44 +0000832 struct btrfs_qgroup_inherit *inherit)
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400833{
David Howells2b0143b2015-03-17 22:25:59 +0000834 struct inode *dir = d_inode(parent->dentry);
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400835 struct dentry *dentry;
836 int error;
837
David Sterba5c50c9b2013-03-22 18:12:51 +0000838 error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
839 if (error == -EINTR)
840 return error;
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400841
842 dentry = lookup_one_len(name, parent->dentry, namelen);
843 error = PTR_ERR(dentry);
844 if (IS_ERR(dentry))
845 goto out_unlock;
846
847 error = -EEXIST;
David Howells2b0143b2015-03-17 22:25:59 +0000848 if (d_really_is_positive(dentry))
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400849 goto out_dput;
850
Yan, Zheng76dda932009-09-21 16:00:26 -0400851 error = btrfs_may_create(dir, dentry);
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400852 if (error)
Liu Boa874a632012-06-29 03:58:46 -0600853 goto out_dput;
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400854
Chris Mason9c520572012-12-17 14:26:57 -0500855 /*
856 * even if this name doesn't exist, we may get hash collisions.
857 * check for them now when we can safely fail
858 */
859 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
860 dir->i_ino, name,
861 namelen);
862 if (error)
863 goto out_dput;
864
Yan, Zheng76dda932009-09-21 16:00:26 -0400865 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
866
867 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
868 goto out_up_read;
869
Chris Mason3de45862008-11-17 21:02:50 -0500870 if (snap_src) {
Miao Xiee9662f72013-02-28 10:01:15 +0000871 error = create_snapshot(snap_src, dir, dentry, name, namelen,
Arne Jansen6f72c7e2011-09-14 15:58:21 +0200872 async_transid, readonly, inherit);
Chris Mason3de45862008-11-17 21:02:50 -0500873 } else {
Miao Xied5c12072013-02-28 10:04:33 +0000874 error = create_subvol(dir, dentry, name, namelen,
875 async_transid, inherit);
Chris Mason3de45862008-11-17 21:02:50 -0500876 }
Yan, Zheng76dda932009-09-21 16:00:26 -0400877 if (!error)
878 fsnotify_mkdir(dir, dentry);
879out_up_read:
880 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400881out_dput:
882 dput(dentry);
883out_unlock:
Al Viro59551022016-01-22 15:40:57 -0500884 inode_unlock(dir);
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400885 return error;
886}
887
Chris Mason4cb53002011-05-24 15:35:30 -0400888/*
889 * When we're defragging a range, we don't want to kick it off again
890 * if it is really just waiting for delalloc to send it down.
891 * If we find a nice big extent or delalloc range for the bytes in the
892 * file you want to defrag, we return 0 to let you know to skip this
893 * part of the file
894 */
David Sterbaaab110a2014-07-29 17:32:10 +0200895static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
Chris Mason4cb53002011-05-24 15:35:30 -0400896{
897 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
898 struct extent_map *em = NULL;
899 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
900 u64 end;
901
902 read_lock(&em_tree->lock);
903 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
904 read_unlock(&em_tree->lock);
905
906 if (em) {
907 end = extent_map_end(em);
908 free_extent_map(em);
909 if (end - offset > thresh)
910 return 0;
911 }
912 /* if we already have a nice delalloc here, just stop */
913 thresh /= 2;
914 end = count_range_bits(io_tree, &offset, offset + thresh,
915 thresh, EXTENT_DELALLOC, 1);
916 if (end >= thresh)
917 return 0;
918 return 1;
919}
920
921/*
922 * helper function to walk through a file and find extents
923 * newer than a specific transid, and smaller than thresh.
924 *
925 * This is used by the defragging code to find new and small
926 * extents
927 */
928static int find_new_extents(struct btrfs_root *root,
929 struct inode *inode, u64 newer_than,
David Sterbaaab110a2014-07-29 17:32:10 +0200930 u64 *off, u32 thresh)
Chris Mason4cb53002011-05-24 15:35:30 -0400931{
932 struct btrfs_path *path;
933 struct btrfs_key min_key;
Chris Mason4cb53002011-05-24 15:35:30 -0400934 struct extent_buffer *leaf;
935 struct btrfs_file_extent_item *extent;
936 int type;
937 int ret;
David Sterbaa4689d22011-05-31 17:08:14 +0000938 u64 ino = btrfs_ino(inode);
Chris Mason4cb53002011-05-24 15:35:30 -0400939
940 path = btrfs_alloc_path();
941 if (!path)
942 return -ENOMEM;
943
David Sterbaa4689d22011-05-31 17:08:14 +0000944 min_key.objectid = ino;
Chris Mason4cb53002011-05-24 15:35:30 -0400945 min_key.type = BTRFS_EXTENT_DATA_KEY;
946 min_key.offset = *off;
947
Dulshani Gunawardhana67871252013-10-31 10:33:04 +0530948 while (1) {
Filipe David Borba Manana6174d3c2013-10-01 16:13:42 +0100949 ret = btrfs_search_forward(root, &min_key, path, newer_than);
Chris Mason4cb53002011-05-24 15:35:30 -0400950 if (ret != 0)
951 goto none;
Filipe Mananaf094c9bd2014-03-12 01:28:24 +0000952process_slot:
David Sterbaa4689d22011-05-31 17:08:14 +0000953 if (min_key.objectid != ino)
Chris Mason4cb53002011-05-24 15:35:30 -0400954 goto none;
955 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
956 goto none;
957
958 leaf = path->nodes[0];
959 extent = btrfs_item_ptr(leaf, path->slots[0],
960 struct btrfs_file_extent_item);
961
962 type = btrfs_file_extent_type(leaf, extent);
963 if (type == BTRFS_FILE_EXTENT_REG &&
964 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
965 check_defrag_in_cache(inode, min_key.offset, thresh)) {
966 *off = min_key.offset;
967 btrfs_free_path(path);
968 return 0;
969 }
970
Filipe Mananaf094c9bd2014-03-12 01:28:24 +0000971 path->slots[0]++;
972 if (path->slots[0] < btrfs_header_nritems(leaf)) {
973 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
974 goto process_slot;
975 }
976
Chris Mason4cb53002011-05-24 15:35:30 -0400977 if (min_key.offset == (u64)-1)
978 goto none;
979
980 min_key.offset++;
981 btrfs_release_path(path);
982 }
983none:
984 btrfs_free_path(path);
985 return -ENOENT;
986}
987
Li Zefan6c282eb2012-06-11 16:03:35 +0800988static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
Liu Bo17ce6ef2012-03-29 09:57:45 -0400989{
990 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
Chris Mason940100a2010-03-10 10:52:59 -0500991 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
Li Zefan6c282eb2012-06-11 16:03:35 +0800992 struct extent_map *em;
993 u64 len = PAGE_CACHE_SIZE;
Chris Mason940100a2010-03-10 10:52:59 -0500994
995 /*
996 * hopefully we have this extent in the tree already, try without
997 * the full extent lock
998 */
999 read_lock(&em_tree->lock);
1000 em = lookup_extent_mapping(em_tree, start, len);
1001 read_unlock(&em_tree->lock);
1002
1003 if (!em) {
Filipe Manana308d9802014-03-11 13:56:15 +00001004 struct extent_state *cached = NULL;
1005 u64 end = start + len - 1;
1006
Chris Mason940100a2010-03-10 10:52:59 -05001007 /* get the big lock and read metadata off disk */
David Sterbaff13db42015-12-03 14:30:40 +01001008 lock_extent_bits(io_tree, start, end, &cached);
Chris Mason940100a2010-03-10 10:52:59 -05001009 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
Filipe Manana308d9802014-03-11 13:56:15 +00001010 unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
Chris Mason940100a2010-03-10 10:52:59 -05001011
Dan Carpenter6cf8bfb2010-03-20 11:22:10 +00001012 if (IS_ERR(em))
Li Zefan6c282eb2012-06-11 16:03:35 +08001013 return NULL;
Chris Mason940100a2010-03-10 10:52:59 -05001014 }
1015
Li Zefan6c282eb2012-06-11 16:03:35 +08001016 return em;
1017}
1018
1019static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1020{
1021 struct extent_map *next;
1022 bool ret = true;
1023
1024 /* this is the last extent */
1025 if (em->start + em->len >= i_size_read(inode))
1026 return false;
1027
1028 next = defrag_lookup_extent(inode, em->start + em->len);
Chris Masone9512d72014-08-26 13:55:54 -07001029 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1030 ret = false;
1031 else if ((em->block_start + em->block_len == next->block_start) &&
Byongho Leeee221842015-12-15 01:42:10 +09001032 (em->block_len > SZ_128K && next->block_len > SZ_128K))
Li Zefan6c282eb2012-06-11 16:03:35 +08001033 ret = false;
1034
1035 free_extent_map(next);
1036 return ret;
1037}
1038
David Sterbaaab110a2014-07-29 17:32:10 +02001039static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
Andrew Mahonea43a2112012-06-19 21:08:32 -04001040 u64 *last_len, u64 *skip, u64 *defrag_end,
1041 int compress)
Li Zefan6c282eb2012-06-11 16:03:35 +08001042{
1043 struct extent_map *em;
1044 int ret = 1;
1045 bool next_mergeable = true;
Liu Bo4a3560c2015-08-07 16:48:41 +08001046 bool prev_mergeable = true;
Li Zefan6c282eb2012-06-11 16:03:35 +08001047
1048 /*
1049 * make sure that once we start defragging an extent, we keep on
1050 * defragging it
1051 */
1052 if (start < *defrag_end)
1053 return 1;
1054
1055 *skip = 0;
1056
1057 em = defrag_lookup_extent(inode, start);
1058 if (!em)
1059 return 0;
1060
Chris Mason940100a2010-03-10 10:52:59 -05001061 /* this will cover holes, and inline extents */
Liu Bo17ce6ef2012-03-29 09:57:45 -04001062 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
Chris Mason940100a2010-03-10 10:52:59 -05001063 ret = 0;
Liu Bo17ce6ef2012-03-29 09:57:45 -04001064 goto out;
1065 }
1066
Liu Bo4a3560c2015-08-07 16:48:41 +08001067 if (!*defrag_end)
1068 prev_mergeable = false;
1069
Li Zefan6c282eb2012-06-11 16:03:35 +08001070 next_mergeable = defrag_check_next_extent(inode, em);
Chris Mason940100a2010-03-10 10:52:59 -05001071 /*
Li Zefan6c282eb2012-06-11 16:03:35 +08001072 * we hit a real extent, if it is big or the next extent is not a
1073 * real extent, don't bother defragging it
Chris Mason940100a2010-03-10 10:52:59 -05001074 */
Andrew Mahonea43a2112012-06-19 21:08:32 -04001075 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
Liu Bo4a3560c2015-08-07 16:48:41 +08001076 (em->len >= thresh || (!next_mergeable && !prev_mergeable)))
Chris Mason940100a2010-03-10 10:52:59 -05001077 ret = 0;
Liu Bo17ce6ef2012-03-29 09:57:45 -04001078out:
Chris Mason940100a2010-03-10 10:52:59 -05001079 /*
1080 * last_len ends up being a counter of how many bytes we've defragged.
1081 * every time we choose not to defrag an extent, we reset *last_len
1082 * so that the next tiny extent will force a defrag.
1083 *
1084 * The end result of this is that tiny extents before a single big
1085 * extent will force at least part of that big extent to be defragged.
1086 */
1087 if (ret) {
Chris Mason940100a2010-03-10 10:52:59 -05001088 *defrag_end = extent_map_end(em);
1089 } else {
1090 *last_len = 0;
1091 *skip = extent_map_end(em);
1092 *defrag_end = 0;
1093 }
1094
1095 free_extent_map(em);
1096 return ret;
1097}
1098
Chris Mason4cb53002011-05-24 15:35:30 -04001099/*
1100 * it doesn't do much good to defrag one or two pages
1101 * at a time. This pulls in a nice chunk of pages
1102 * to COW and defrag.
1103 *
1104 * It also makes sure the delalloc code has enough
1105 * dirty data to avoid making new small extents as part
1106 * of the defrag
1107 *
1108 * It's a good idea to start RA on this range
1109 * before calling this.
1110 */
1111static int cluster_pages_for_defrag(struct inode *inode,
1112 struct page **pages,
1113 unsigned long start_index,
Justin Maggardc41570c2014-01-21 11:18:29 -08001114 unsigned long num_pages)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001115{
Chris Mason4cb53002011-05-24 15:35:30 -04001116 unsigned long file_end;
1117 u64 isize = i_size_read(inode);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001118 u64 page_start;
1119 u64 page_end;
Liu Bo1f12bd02012-03-29 09:57:44 -04001120 u64 page_cnt;
Chris Mason4cb53002011-05-24 15:35:30 -04001121 int ret;
1122 int i;
1123 int i_done;
1124 struct btrfs_ordered_extent *ordered;
1125 struct extent_state *cached_state = NULL;
Miao Xie600a45e2012-02-16 15:01:24 +08001126 struct extent_io_tree *tree;
Josef Bacik3b16a4e2011-09-21 15:05:58 -04001127 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
Chris Mason4cb53002011-05-24 15:35:30 -04001128
Chris Mason4cb53002011-05-24 15:35:30 -04001129 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
Liu Bo1f12bd02012-03-29 09:57:44 -04001130 if (!isize || start_index > file_end)
1131 return 0;
1132
1133 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
Chris Mason4cb53002011-05-24 15:35:30 -04001134
1135 ret = btrfs_delalloc_reserve_space(inode,
Qu Wenruodf480632015-09-08 17:25:54 +08001136 start_index << PAGE_CACHE_SHIFT,
1137 page_cnt << PAGE_CACHE_SHIFT);
Chris Mason4cb53002011-05-24 15:35:30 -04001138 if (ret)
1139 return ret;
Chris Mason4cb53002011-05-24 15:35:30 -04001140 i_done = 0;
Miao Xie600a45e2012-02-16 15:01:24 +08001141 tree = &BTRFS_I(inode)->io_tree;
Chris Mason4cb53002011-05-24 15:35:30 -04001142
1143 /* step one, lock all the pages */
Liu Bo1f12bd02012-03-29 09:57:44 -04001144 for (i = 0; i < page_cnt; i++) {
Chris Mason4cb53002011-05-24 15:35:30 -04001145 struct page *page;
Miao Xie600a45e2012-02-16 15:01:24 +08001146again:
Josef Bacika94733d2011-07-11 10:47:06 -04001147 page = find_or_create_page(inode->i_mapping,
Miao Xie600a45e2012-02-16 15:01:24 +08001148 start_index + i, mask);
Chris Mason4cb53002011-05-24 15:35:30 -04001149 if (!page)
1150 break;
1151
Miao Xie600a45e2012-02-16 15:01:24 +08001152 page_start = page_offset(page);
1153 page_end = page_start + PAGE_CACHE_SIZE - 1;
1154 while (1) {
Filipe Manana308d9802014-03-11 13:56:15 +00001155 lock_extent_bits(tree, page_start, page_end,
David Sterbaff13db42015-12-03 14:30:40 +01001156 &cached_state);
Miao Xie600a45e2012-02-16 15:01:24 +08001157 ordered = btrfs_lookup_ordered_extent(inode,
1158 page_start);
Filipe Manana308d9802014-03-11 13:56:15 +00001159 unlock_extent_cached(tree, page_start, page_end,
1160 &cached_state, GFP_NOFS);
Miao Xie600a45e2012-02-16 15:01:24 +08001161 if (!ordered)
1162 break;
1163
1164 unlock_page(page);
1165 btrfs_start_ordered_extent(inode, ordered, 1);
1166 btrfs_put_ordered_extent(ordered);
1167 lock_page(page);
Liu Bo1f12bd02012-03-29 09:57:44 -04001168 /*
1169 * we unlocked the page above, so we need check if
1170 * it was released or not.
1171 */
1172 if (page->mapping != inode->i_mapping) {
1173 unlock_page(page);
1174 page_cache_release(page);
1175 goto again;
1176 }
Miao Xie600a45e2012-02-16 15:01:24 +08001177 }
1178
Chris Mason4cb53002011-05-24 15:35:30 -04001179 if (!PageUptodate(page)) {
1180 btrfs_readpage(NULL, page);
1181 lock_page(page);
1182 if (!PageUptodate(page)) {
1183 unlock_page(page);
1184 page_cache_release(page);
1185 ret = -EIO;
1186 break;
1187 }
1188 }
Miao Xie600a45e2012-02-16 15:01:24 +08001189
Miao Xie600a45e2012-02-16 15:01:24 +08001190 if (page->mapping != inode->i_mapping) {
1191 unlock_page(page);
1192 page_cache_release(page);
1193 goto again;
1194 }
1195
Chris Mason4cb53002011-05-24 15:35:30 -04001196 pages[i] = page;
1197 i_done++;
1198 }
1199 if (!i_done || ret)
1200 goto out;
1201
1202 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1203 goto out;
1204
1205 /*
1206 * so now we have a nice long stream of locked
1207 * and up to date pages, lets wait on them
1208 */
1209 for (i = 0; i < i_done; i++)
1210 wait_on_page_writeback(pages[i]);
1211
1212 page_start = page_offset(pages[0]);
1213 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1214
1215 lock_extent_bits(&BTRFS_I(inode)->io_tree,
David Sterbaff13db42015-12-03 14:30:40 +01001216 page_start, page_end - 1, &cached_state);
Chris Mason4cb53002011-05-24 15:35:30 -04001217 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1218 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
Liu Bo9e8a4a82012-09-05 19:10:51 -06001219 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1220 &cached_state, GFP_NOFS);
Chris Mason4cb53002011-05-24 15:35:30 -04001221
Liu Bo1f12bd02012-03-29 09:57:44 -04001222 if (i_done != page_cnt) {
Josef Bacik9e0baf62011-07-15 15:16:44 +00001223 spin_lock(&BTRFS_I(inode)->lock);
1224 BTRFS_I(inode)->outstanding_extents++;
1225 spin_unlock(&BTRFS_I(inode)->lock);
Chris Mason4cb53002011-05-24 15:35:30 -04001226 btrfs_delalloc_release_space(inode,
Qu Wenruodf480632015-09-08 17:25:54 +08001227 start_index << PAGE_CACHE_SHIFT,
1228 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
Chris Mason4cb53002011-05-24 15:35:30 -04001229 }
1230
1231
Liu Bo9e8a4a82012-09-05 19:10:51 -06001232 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1233 &cached_state, GFP_NOFS);
Chris Mason4cb53002011-05-24 15:35:30 -04001234
1235 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1236 page_start, page_end - 1, &cached_state,
1237 GFP_NOFS);
1238
1239 for (i = 0; i < i_done; i++) {
1240 clear_page_dirty_for_io(pages[i]);
1241 ClearPageChecked(pages[i]);
1242 set_page_extent_mapped(pages[i]);
1243 set_page_dirty(pages[i]);
1244 unlock_page(pages[i]);
1245 page_cache_release(pages[i]);
1246 }
1247 return i_done;
1248out:
1249 for (i = 0; i < i_done; i++) {
1250 unlock_page(pages[i]);
1251 page_cache_release(pages[i]);
1252 }
Qu Wenruo7cf5b972015-09-08 17:25:55 +08001253 btrfs_delalloc_release_space(inode,
Qu Wenruodf480632015-09-08 17:25:54 +08001254 start_index << PAGE_CACHE_SHIFT,
1255 page_cnt << PAGE_CACHE_SHIFT);
Chris Mason4cb53002011-05-24 15:35:30 -04001256 return ret;
1257
1258}
1259
1260int btrfs_defrag_file(struct inode *inode, struct file *file,
1261 struct btrfs_ioctl_defrag_range_args *range,
1262 u64 newer_than, unsigned long max_to_defrag)
1263{
1264 struct btrfs_root *root = BTRFS_I(inode)->root;
Chris Mason4cb53002011-05-24 15:35:30 -04001265 struct file_ra_state *ra = NULL;
1266 unsigned long last_index;
Li Zefan151a31b2011-09-02 15:56:39 +08001267 u64 isize = i_size_read(inode);
Chris Mason940100a2010-03-10 10:52:59 -05001268 u64 last_len = 0;
1269 u64 skip = 0;
1270 u64 defrag_end = 0;
Chris Mason4cb53002011-05-24 15:35:30 -04001271 u64 newer_off = range->start;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001272 unsigned long i;
Li Zefan008873e2011-09-02 15:57:07 +08001273 unsigned long ra_index = 0;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001274 int ret;
Chris Mason4cb53002011-05-24 15:35:30 -04001275 int defrag_count = 0;
Li Zefan1a419d82010-10-25 15:12:50 +08001276 int compress_type = BTRFS_COMPRESS_ZLIB;
David Sterbaaab110a2014-07-29 17:32:10 +02001277 u32 extent_thresh = range->extent_thresh;
Byongho Leeee221842015-12-15 01:42:10 +09001278 unsigned long max_cluster = SZ_256K >> PAGE_CACHE_SHIFT;
Justin Maggardc41570c2014-01-21 11:18:29 -08001279 unsigned long cluster = max_cluster;
Byongho Leeee221842015-12-15 01:42:10 +09001280 u64 new_align = ~((u64)SZ_128K - 1);
Chris Mason4cb53002011-05-24 15:35:30 -04001281 struct page **pages = NULL;
1282
Liu Bo0abd5b12013-04-16 09:20:28 +00001283 if (isize == 0)
1284 return 0;
1285
1286 if (range->start >= isize)
1287 return -EINVAL;
Li Zefan1a419d82010-10-25 15:12:50 +08001288
1289 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1290 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1291 return -EINVAL;
1292 if (range->compress_type)
1293 compress_type = range->compress_type;
1294 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001295
Liu Bo0abd5b12013-04-16 09:20:28 +00001296 if (extent_thresh == 0)
Byongho Leeee221842015-12-15 01:42:10 +09001297 extent_thresh = SZ_256K;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001298
Chris Mason4cb53002011-05-24 15:35:30 -04001299 /*
1300 * if we were not given a file, allocate a readahead
1301 * context
1302 */
1303 if (!file) {
1304 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1305 if (!ra)
1306 return -ENOMEM;
1307 file_ra_state_init(ra, inode->i_mapping);
1308 } else {
1309 ra = &file->f_ra;
1310 }
1311
Dulshani Gunawardhanad9b0d9b2013-10-31 10:32:18 +05301312 pages = kmalloc_array(max_cluster, sizeof(struct page *),
Chris Mason4cb53002011-05-24 15:35:30 -04001313 GFP_NOFS);
1314 if (!pages) {
1315 ret = -ENOMEM;
1316 goto out_ra;
1317 }
1318
1319 /* find the last page to defrag */
Chris Mason1e701a32010-03-11 09:42:04 -05001320 if (range->start + range->len > range->start) {
Li Zefan151a31b2011-09-02 15:56:39 +08001321 last_index = min_t(u64, isize - 1,
Chris Mason1e701a32010-03-11 09:42:04 -05001322 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1323 } else {
Li Zefan151a31b2011-09-02 15:56:39 +08001324 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
Chris Mason1e701a32010-03-11 09:42:04 -05001325 }
1326
Chris Mason4cb53002011-05-24 15:35:30 -04001327 if (newer_than) {
1328 ret = find_new_extents(root, inode, newer_than,
Byongho Leeee221842015-12-15 01:42:10 +09001329 &newer_off, SZ_64K);
Chris Mason4cb53002011-05-24 15:35:30 -04001330 if (!ret) {
1331 range->start = newer_off;
1332 /*
1333 * we always align our defrag to help keep
1334 * the extents in the file evenly spaced
1335 */
1336 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
Chris Mason4cb53002011-05-24 15:35:30 -04001337 } else
1338 goto out_ra;
1339 } else {
1340 i = range->start >> PAGE_CACHE_SHIFT;
1341 }
1342 if (!max_to_defrag)
chandan070034b2015-06-09 10:35:11 +05301343 max_to_defrag = last_index - i + 1;
Chris Mason4cb53002011-05-24 15:35:30 -04001344
Li Zefan2a0f7f52011-10-10 15:43:34 -04001345 /*
1346 * make writeback starts from i, so the defrag range can be
1347 * written sequentially.
1348 */
1349 if (i < inode->i_mapping->writeback_index)
1350 inode->i_mapping->writeback_index = i;
1351
Chris Masonf7f43cc2011-10-11 11:41:40 -04001352 while (i <= last_index && defrag_count < max_to_defrag &&
David Sterbaed6078f2014-06-05 01:59:57 +02001353 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_CACHE_SIZE))) {
Chris Mason4cb53002011-05-24 15:35:30 -04001354 /*
1355 * make sure we stop running if someone unmounts
1356 * the FS
1357 */
1358 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1359 break;
1360
David Sterba210549e2013-02-09 23:38:06 +00001361 if (btrfs_defrag_cancelled(root->fs_info)) {
David Sterbaf14d1042015-10-08 11:37:06 +02001362 btrfs_debug(root->fs_info, "defrag_file cancelled");
David Sterba210549e2013-02-09 23:38:06 +00001363 ret = -EAGAIN;
1364 break;
1365 }
1366
Liu Bo66c26892012-03-29 09:57:45 -04001367 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
Li Zefan6c282eb2012-06-11 16:03:35 +08001368 extent_thresh, &last_len, &skip,
Andrew Mahonea43a2112012-06-19 21:08:32 -04001369 &defrag_end, range->flags &
1370 BTRFS_DEFRAG_RANGE_COMPRESS)) {
Chris Mason940100a2010-03-10 10:52:59 -05001371 unsigned long next;
1372 /*
1373 * the should_defrag function tells us how much to skip
1374 * bump our counter by the suggested amount
1375 */
David Sterbaed6078f2014-06-05 01:59:57 +02001376 next = DIV_ROUND_UP(skip, PAGE_CACHE_SIZE);
Chris Mason940100a2010-03-10 10:52:59 -05001377 i = max(i + 1, next);
1378 continue;
1379 }
Li Zefan008873e2011-09-02 15:57:07 +08001380
1381 if (!newer_than) {
1382 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1383 PAGE_CACHE_SHIFT) - i;
1384 cluster = min(cluster, max_cluster);
1385 } else {
1386 cluster = max_cluster;
1387 }
1388
Li Zefan008873e2011-09-02 15:57:07 +08001389 if (i + cluster > ra_index) {
1390 ra_index = max(i, ra_index);
1391 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1392 cluster);
chandane4826a52015-06-09 17:38:32 +05301393 ra_index += cluster;
Li Zefan008873e2011-09-02 15:57:07 +08001394 }
Chris Mason940100a2010-03-10 10:52:59 -05001395
Al Viro59551022016-01-22 15:40:57 -05001396 inode_lock(inode);
Filipe David Borba Manana633085c2013-08-16 15:23:33 +01001397 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1398 BTRFS_I(inode)->force_compress = compress_type;
Li Zefan008873e2011-09-02 15:57:07 +08001399 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
Liu Boecb8bea2012-03-29 09:57:44 -04001400 if (ret < 0) {
Al Viro59551022016-01-22 15:40:57 -05001401 inode_unlock(inode);
Chris Mason4cb53002011-05-24 15:35:30 -04001402 goto out_ra;
Liu Boecb8bea2012-03-29 09:57:44 -04001403 }
Chris Mason940100a2010-03-10 10:52:59 -05001404
Chris Mason4cb53002011-05-24 15:35:30 -04001405 defrag_count += ret;
Namjae Jeond0e1d662012-12-11 16:00:21 -08001406 balance_dirty_pages_ratelimited(inode->i_mapping);
Al Viro59551022016-01-22 15:40:57 -05001407 inode_unlock(inode);
Chris Mason4cb53002011-05-24 15:35:30 -04001408
1409 if (newer_than) {
1410 if (newer_off == (u64)-1)
1411 break;
1412
Liu Boe1f041e2012-03-29 09:57:45 -04001413 if (ret > 0)
1414 i += ret;
1415
Chris Mason4cb53002011-05-24 15:35:30 -04001416 newer_off = max(newer_off + 1,
1417 (u64)i << PAGE_CACHE_SHIFT);
1418
Byongho Leeee221842015-12-15 01:42:10 +09001419 ret = find_new_extents(root, inode, newer_than,
1420 &newer_off, SZ_64K);
Chris Mason4cb53002011-05-24 15:35:30 -04001421 if (!ret) {
1422 range->start = newer_off;
1423 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
Chris Mason4cb53002011-05-24 15:35:30 -04001424 } else {
1425 break;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001426 }
Chris Mason4cb53002011-05-24 15:35:30 -04001427 } else {
Li Zefan008873e2011-09-02 15:57:07 +08001428 if (ret > 0) {
Li Zefancbcc8322011-09-02 15:56:25 +08001429 i += ret;
Li Zefan008873e2011-09-02 15:57:07 +08001430 last_len += ret << PAGE_CACHE_SHIFT;
1431 } else {
Li Zefancbcc8322011-09-02 15:56:25 +08001432 i++;
Li Zefan008873e2011-09-02 15:57:07 +08001433 last_len = 0;
1434 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001435 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001436 }
1437
Filipe Mananadec8ef92014-03-01 10:55:54 +00001438 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
Chris Mason1e701a32010-03-11 09:42:04 -05001439 filemap_flush(inode->i_mapping);
Filipe Mananadec8ef92014-03-01 10:55:54 +00001440 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1441 &BTRFS_I(inode)->runtime_flags))
1442 filemap_flush(inode->i_mapping);
1443 }
Chris Mason1e701a32010-03-11 09:42:04 -05001444
1445 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1446 /* the filemap_flush will queue IO into the worker threads, but
1447 * we have to make sure the IO is actually started and that
1448 * ordered extents get created before we return
1449 */
1450 atomic_inc(&root->fs_info->async_submit_draining);
1451 while (atomic_read(&root->fs_info->nr_async_submits) ||
1452 atomic_read(&root->fs_info->async_delalloc_pages)) {
1453 wait_event(root->fs_info->async_submit_wait,
1454 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1455 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1456 }
1457 atomic_dec(&root->fs_info->async_submit_draining);
Chris Mason1e701a32010-03-11 09:42:04 -05001458 }
1459
Li Zefan1a419d82010-10-25 15:12:50 +08001460 if (range->compress_type == BTRFS_COMPRESS_LZO) {
Mitch Harder2b0ce2c2012-07-24 11:58:43 -06001461 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
Li Zefan1a419d82010-10-25 15:12:50 +08001462 }
1463
Diego Calleja60ccf822011-09-01 16:33:57 +02001464 ret = defrag_count;
Chris Mason940100a2010-03-10 10:52:59 -05001465
Chris Mason4cb53002011-05-24 15:35:30 -04001466out_ra:
Filipe David Borba Manana633085c2013-08-16 15:23:33 +01001467 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
Al Viro59551022016-01-22 15:40:57 -05001468 inode_lock(inode);
Filipe David Borba Manana633085c2013-08-16 15:23:33 +01001469 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
Al Viro59551022016-01-22 15:40:57 -05001470 inode_unlock(inode);
Filipe David Borba Manana633085c2013-08-16 15:23:33 +01001471 }
Chris Mason4cb53002011-05-24 15:35:30 -04001472 if (!file)
1473 kfree(ra);
1474 kfree(pages);
Chris Mason940100a2010-03-10 10:52:59 -05001475 return ret;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001476}
1477
Miao Xie198605a2012-11-26 08:43:45 +00001478static noinline int btrfs_ioctl_resize(struct file *file,
Yan, Zheng76dda932009-09-21 16:00:26 -04001479 void __user *arg)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001480{
1481 u64 new_size;
1482 u64 old_size;
1483 u64 devid = 1;
Al Viro496ad9a2013-01-23 17:07:38 -05001484 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001485 struct btrfs_ioctl_vol_args *vol_args;
1486 struct btrfs_trans_handle *trans;
1487 struct btrfs_device *device = NULL;
1488 char *sizestr;
Gui Hecheng9a40f122014-03-31 18:03:25 +08001489 char *retptr;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001490 char *devstr = NULL;
1491 int ret = 0;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001492 int mod = 0;
1493
Chris Masone441d542009-01-05 16:57:23 -05001494 if (!capable(CAP_SYS_ADMIN))
1495 return -EPERM;
1496
Miao Xie198605a2012-11-26 08:43:45 +00001497 ret = mnt_want_write_file(file);
1498 if (ret)
1499 return ret;
1500
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01001501 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1502 1)) {
Miao Xie97547672012-12-21 10:38:50 +00001503 mnt_drop_write_file(file);
Anand Jaine57138b2013-08-21 11:44:48 +08001504 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001505 }
1506
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01001507 mutex_lock(&root->fs_info->volume_mutex);
Li Zefandae7b662009-04-08 15:06:54 +08001508 vol_args = memdup_user(arg, sizeof(*vol_args));
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001509 if (IS_ERR(vol_args)) {
1510 ret = PTR_ERR(vol_args);
1511 goto out;
1512 }
Mark Fasheh5516e592008-07-24 12:20:14 -04001513
1514 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001515
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001516 sizestr = vol_args->name;
1517 devstr = strchr(sizestr, ':');
1518 if (devstr) {
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001519 sizestr = devstr + 1;
1520 *devstr = '\0';
1521 devstr = vol_args->name;
ZhangZhen58dfae62014-05-13 16:36:08 +08001522 ret = kstrtoull(devstr, 10, &devid);
1523 if (ret)
1524 goto out_free;
Miao Xiedfd79822012-12-21 09:21:30 +00001525 if (!devid) {
1526 ret = -EINVAL;
1527 goto out_free;
1528 }
Frank Holtonefe120a2013-12-20 11:37:06 -05001529 btrfs_info(root->fs_info, "resizing devid %llu", devid);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001530 }
Miao Xiedba60f32012-12-21 09:19:51 +00001531
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01001532 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001533 if (!device) {
Frank Holtonefe120a2013-12-20 11:37:06 -05001534 btrfs_info(root->fs_info, "resizer unable to find device %llu",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02001535 devid);
Miao Xiedfd79822012-12-21 09:21:30 +00001536 ret = -ENODEV;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001537 goto out_free;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001538 }
Miao Xiedba60f32012-12-21 09:19:51 +00001539
1540 if (!device->writeable) {
Frank Holtonefe120a2013-12-20 11:37:06 -05001541 btrfs_info(root->fs_info,
1542 "resizer unable to apply on readonly device %llu",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02001543 devid);
Miao Xiedfd79822012-12-21 09:21:30 +00001544 ret = -EPERM;
Liu Bo4e42ae12012-06-14 02:23:19 -06001545 goto out_free;
1546 }
1547
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001548 if (!strcmp(sizestr, "max"))
1549 new_size = device->bdev->bd_inode->i_size;
1550 else {
1551 if (sizestr[0] == '-') {
1552 mod = -1;
1553 sizestr++;
1554 } else if (sizestr[0] == '+') {
1555 mod = 1;
1556 sizestr++;
1557 }
Gui Hecheng9a40f122014-03-31 18:03:25 +08001558 new_size = memparse(sizestr, &retptr);
1559 if (*retptr != '\0' || new_size == 0) {
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001560 ret = -EINVAL;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001561 goto out_free;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001562 }
1563 }
1564
Stefan Behrens63a212a2012-11-05 18:29:28 +01001565 if (device->is_tgtdev_for_dev_replace) {
Miao Xiedfd79822012-12-21 09:21:30 +00001566 ret = -EPERM;
Stefan Behrens63a212a2012-11-05 18:29:28 +01001567 goto out_free;
1568 }
1569
Miao Xie7cc8e582014-09-03 21:35:38 +08001570 old_size = btrfs_device_get_total_bytes(device);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001571
1572 if (mod < 0) {
1573 if (new_size > old_size) {
1574 ret = -EINVAL;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001575 goto out_free;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001576 }
1577 new_size = old_size - new_size;
1578 } else if (mod > 0) {
Wenliang Faneb8052e2013-12-20 15:28:56 +08001579 if (new_size > ULLONG_MAX - old_size) {
Gui Hecheng902c68a2014-05-29 09:19:58 +08001580 ret = -ERANGE;
Wenliang Faneb8052e2013-12-20 15:28:56 +08001581 goto out_free;
1582 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001583 new_size = old_size + new_size;
1584 }
1585
Byongho Leeee221842015-12-15 01:42:10 +09001586 if (new_size < SZ_256M) {
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001587 ret = -EINVAL;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001588 goto out_free;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001589 }
1590 if (new_size > device->bdev->bd_inode->i_size) {
1591 ret = -EFBIG;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001592 goto out_free;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001593 }
1594
David Sterbab8b93ad2015-01-16 17:26:13 +01001595 new_size = div_u64(new_size, root->sectorsize);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001596 new_size *= root->sectorsize;
1597
David Sterbaecaeb142015-10-08 09:01:03 +02001598 btrfs_info_in_rcu(root->fs_info, "new size for %s is %llu",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02001599 rcu_str_deref(device->name), new_size);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001600
1601 if (new_size > old_size) {
Yan, Zhenga22285a2010-05-16 10:48:46 -04001602 trans = btrfs_start_transaction(root, 0);
Tsutomu Itoh98d5dc12011-01-20 06:19:37 +00001603 if (IS_ERR(trans)) {
1604 ret = PTR_ERR(trans);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001605 goto out_free;
Tsutomu Itoh98d5dc12011-01-20 06:19:37 +00001606 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001607 ret = btrfs_grow_device(trans, device, new_size);
1608 btrfs_commit_transaction(trans, root);
Mike Fleetwoodece7d202011-11-18 18:55:01 +00001609 } else if (new_size < old_size) {
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001610 ret = btrfs_shrink_device(device, new_size);
jeff.liu0253f402012-10-27 12:06:39 +00001611 } /* equal, nothing need to do */
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001612
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001613out_free:
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001614 kfree(vol_args);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001615out:
1616 mutex_unlock(&root->fs_info->volume_mutex);
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01001617 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
Ilya Dryomov18f39c42013-01-20 15:57:57 +02001618 mnt_drop_write_file(file);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001619 return ret;
1620}
1621
Sage Weil72fd0322010-10-29 15:41:32 -04001622static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001623 char *name, unsigned long fd, int subvol,
1624 u64 *transid, bool readonly,
Miao Xie8696c532013-02-07 06:02:44 +00001625 struct btrfs_qgroup_inherit *inherit)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001626{
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001627 int namelen;
Chris Mason3de45862008-11-17 21:02:50 -05001628 int ret = 0;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001629
Liu Boa874a632012-06-29 03:58:46 -06001630 ret = mnt_want_write_file(file);
1631 if (ret)
1632 goto out;
1633
Sage Weil72fd0322010-10-29 15:41:32 -04001634 namelen = strlen(name);
1635 if (strchr(name, '/')) {
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001636 ret = -EINVAL;
Liu Boa874a632012-06-29 03:58:46 -06001637 goto out_drop_write;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001638 }
1639
Chris Mason16780ca2012-02-20 22:14:55 -05001640 if (name[0] == '.' &&
1641 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1642 ret = -EEXIST;
Liu Boa874a632012-06-29 03:58:46 -06001643 goto out_drop_write;
Chris Mason16780ca2012-02-20 22:14:55 -05001644 }
1645
Chris Mason3de45862008-11-17 21:02:50 -05001646 if (subvol) {
Sage Weil72fd0322010-10-29 15:41:32 -04001647 ret = btrfs_mksubvol(&file->f_path, name, namelen,
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001648 NULL, transid, readonly, inherit);
Christoph Hellwigcb8e7092008-10-09 13:39:39 -04001649 } else {
Al Viro2903ff02012-08-28 12:52:22 -04001650 struct fd src = fdget(fd);
Chris Mason3de45862008-11-17 21:02:50 -05001651 struct inode *src_inode;
Al Viro2903ff02012-08-28 12:52:22 -04001652 if (!src.file) {
Chris Mason3de45862008-11-17 21:02:50 -05001653 ret = -EINVAL;
Liu Boa874a632012-06-29 03:58:46 -06001654 goto out_drop_write;
Chris Mason3de45862008-11-17 21:02:50 -05001655 }
1656
Al Viro496ad9a2013-01-23 17:07:38 -05001657 src_inode = file_inode(src.file);
1658 if (src_inode->i_sb != file_inode(file)->i_sb) {
Frank Holtonefe120a2013-12-20 11:37:06 -05001659 btrfs_info(BTRFS_I(src_inode)->root->fs_info,
1660 "Snapshot src from another FS");
Kusanagi Kouichi23ad5b12014-01-30 16:32:02 +09001661 ret = -EXDEV;
David Sterbad0242062014-01-15 18:15:52 +01001662 } else if (!inode_owner_or_capable(src_inode)) {
1663 /*
1664 * Subvolume creation is not restricted, but snapshots
1665 * are limited to own subvolumes only
1666 */
1667 ret = -EPERM;
Al Viroecd18812012-08-26 21:20:24 -04001668 } else {
1669 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1670 BTRFS_I(src_inode)->root,
1671 transid, readonly, inherit);
Chris Mason3de45862008-11-17 21:02:50 -05001672 }
Al Viro2903ff02012-08-28 12:52:22 -04001673 fdput(src);
Christoph Hellwigcb8e7092008-10-09 13:39:39 -04001674 }
Liu Boa874a632012-06-29 03:58:46 -06001675out_drop_write:
1676 mnt_drop_write_file(file);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001677out:
Sage Weil72fd0322010-10-29 15:41:32 -04001678 return ret;
1679}
1680
1681static noinline int btrfs_ioctl_snap_create(struct file *file,
Li Zefanfa0d2b92010-12-20 15:53:28 +08001682 void __user *arg, int subvol)
Sage Weil72fd0322010-10-29 15:41:32 -04001683{
Li Zefanfa0d2b92010-12-20 15:53:28 +08001684 struct btrfs_ioctl_vol_args *vol_args;
Sage Weil72fd0322010-10-29 15:41:32 -04001685 int ret;
1686
Li Zefanfa0d2b92010-12-20 15:53:28 +08001687 vol_args = memdup_user(arg, sizeof(*vol_args));
1688 if (IS_ERR(vol_args))
1689 return PTR_ERR(vol_args);
1690 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
Sage Weil72fd0322010-10-29 15:41:32 -04001691
Li Zefanfa0d2b92010-12-20 15:53:28 +08001692 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
Li Zefanb83cc962010-12-20 16:04:08 +08001693 vol_args->fd, subvol,
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001694 NULL, false, NULL);
Li Zefanfdfb1e42010-12-10 06:41:56 +00001695
Li Zefanfa0d2b92010-12-20 15:53:28 +08001696 kfree(vol_args);
1697 return ret;
1698}
Li Zefanfdfb1e42010-12-10 06:41:56 +00001699
Li Zefanfa0d2b92010-12-20 15:53:28 +08001700static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1701 void __user *arg, int subvol)
1702{
1703 struct btrfs_ioctl_vol_args_v2 *vol_args;
1704 int ret;
1705 u64 transid = 0;
1706 u64 *ptr = NULL;
Li Zefanb83cc962010-12-20 16:04:08 +08001707 bool readonly = false;
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001708 struct btrfs_qgroup_inherit *inherit = NULL;
Li Zefanfdfb1e42010-12-10 06:41:56 +00001709
Li Zefanfa0d2b92010-12-20 15:53:28 +08001710 vol_args = memdup_user(arg, sizeof(*vol_args));
1711 if (IS_ERR(vol_args))
1712 return PTR_ERR(vol_args);
1713 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
Sage Weil75eaa0e2010-12-10 00:36:28 +00001714
Li Zefanb83cc962010-12-20 16:04:08 +08001715 if (vol_args->flags &
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001716 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1717 BTRFS_SUBVOL_QGROUP_INHERIT)) {
Li Zefanb83cc962010-12-20 16:04:08 +08001718 ret = -EOPNOTSUPP;
Dan Carpenterc47ca322014-09-04 14:09:15 +03001719 goto free_args;
Sage Weil72fd0322010-10-29 15:41:32 -04001720 }
Li Zefanfa0d2b92010-12-20 15:53:28 +08001721
1722 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1723 ptr = &transid;
Li Zefanb83cc962010-12-20 16:04:08 +08001724 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1725 readonly = true;
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001726 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1727 if (vol_args->size > PAGE_CACHE_SIZE) {
1728 ret = -EINVAL;
Dan Carpenterc47ca322014-09-04 14:09:15 +03001729 goto free_args;
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001730 }
1731 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1732 if (IS_ERR(inherit)) {
1733 ret = PTR_ERR(inherit);
Dan Carpenterc47ca322014-09-04 14:09:15 +03001734 goto free_args;
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001735 }
1736 }
Li Zefanfa0d2b92010-12-20 15:53:28 +08001737
1738 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001739 vol_args->fd, subvol, ptr,
Miao Xie8696c532013-02-07 06:02:44 +00001740 readonly, inherit);
Dan Carpenterc47ca322014-09-04 14:09:15 +03001741 if (ret)
1742 goto free_inherit;
Li Zefanfa0d2b92010-12-20 15:53:28 +08001743
Dan Carpenterc47ca322014-09-04 14:09:15 +03001744 if (ptr && copy_to_user(arg +
1745 offsetof(struct btrfs_ioctl_vol_args_v2,
1746 transid),
1747 ptr, sizeof(*ptr)))
Li Zefanfa0d2b92010-12-20 15:53:28 +08001748 ret = -EFAULT;
Dan Carpenterc47ca322014-09-04 14:09:15 +03001749
1750free_inherit:
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001751 kfree(inherit);
Dan Carpenterc47ca322014-09-04 14:09:15 +03001752free_args:
1753 kfree(vol_args);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001754 return ret;
1755}
1756
Li Zefan0caa1022010-12-20 16:30:25 +08001757static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1758 void __user *arg)
1759{
Al Viro496ad9a2013-01-23 17:07:38 -05001760 struct inode *inode = file_inode(file);
Li Zefan0caa1022010-12-20 16:30:25 +08001761 struct btrfs_root *root = BTRFS_I(inode)->root;
1762 int ret = 0;
1763 u64 flags = 0;
1764
Li Zefan33345d012011-04-20 10:31:50 +08001765 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
Li Zefan0caa1022010-12-20 16:30:25 +08001766 return -EINVAL;
1767
1768 down_read(&root->fs_info->subvol_sem);
1769 if (btrfs_root_readonly(root))
1770 flags |= BTRFS_SUBVOL_RDONLY;
1771 up_read(&root->fs_info->subvol_sem);
1772
1773 if (copy_to_user(arg, &flags, sizeof(flags)))
1774 ret = -EFAULT;
1775
1776 return ret;
1777}
1778
1779static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1780 void __user *arg)
1781{
Al Viro496ad9a2013-01-23 17:07:38 -05001782 struct inode *inode = file_inode(file);
Li Zefan0caa1022010-12-20 16:30:25 +08001783 struct btrfs_root *root = BTRFS_I(inode)->root;
1784 struct btrfs_trans_handle *trans;
1785 u64 root_flags;
1786 u64 flags;
1787 int ret = 0;
1788
David Sterbabd60ea02014-01-16 15:50:22 +01001789 if (!inode_owner_or_capable(inode))
1790 return -EPERM;
1791
Liu Bob9ca0662012-06-29 03:58:49 -06001792 ret = mnt_want_write_file(file);
1793 if (ret)
1794 goto out;
Li Zefan0caa1022010-12-20 16:30:25 +08001795
Liu Bob9ca0662012-06-29 03:58:49 -06001796 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1797 ret = -EINVAL;
1798 goto out_drop_write;
1799 }
Li Zefan0caa1022010-12-20 16:30:25 +08001800
Liu Bob9ca0662012-06-29 03:58:49 -06001801 if (copy_from_user(&flags, arg, sizeof(flags))) {
1802 ret = -EFAULT;
1803 goto out_drop_write;
1804 }
Li Zefan0caa1022010-12-20 16:30:25 +08001805
Liu Bob9ca0662012-06-29 03:58:49 -06001806 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1807 ret = -EINVAL;
1808 goto out_drop_write;
1809 }
Li Zefan0caa1022010-12-20 16:30:25 +08001810
Liu Bob9ca0662012-06-29 03:58:49 -06001811 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1812 ret = -EOPNOTSUPP;
1813 goto out_drop_write;
1814 }
Li Zefan0caa1022010-12-20 16:30:25 +08001815
1816 down_write(&root->fs_info->subvol_sem);
1817
1818 /* nothing to do */
1819 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
Liu Bob9ca0662012-06-29 03:58:49 -06001820 goto out_drop_sem;
Li Zefan0caa1022010-12-20 16:30:25 +08001821
1822 root_flags = btrfs_root_flags(&root->root_item);
David Sterba2c686532013-12-16 17:34:17 +01001823 if (flags & BTRFS_SUBVOL_RDONLY) {
Li Zefan0caa1022010-12-20 16:30:25 +08001824 btrfs_set_root_flags(&root->root_item,
1825 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
David Sterba2c686532013-12-16 17:34:17 +01001826 } else {
1827 /*
1828 * Block RO -> RW transition if this subvolume is involved in
1829 * send
1830 */
1831 spin_lock(&root->root_item_lock);
1832 if (root->send_in_progress == 0) {
1833 btrfs_set_root_flags(&root->root_item,
Li Zefan0caa1022010-12-20 16:30:25 +08001834 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
David Sterba2c686532013-12-16 17:34:17 +01001835 spin_unlock(&root->root_item_lock);
1836 } else {
1837 spin_unlock(&root->root_item_lock);
1838 btrfs_warn(root->fs_info,
1839 "Attempt to set subvolume %llu read-write during send",
1840 root->root_key.objectid);
1841 ret = -EPERM;
1842 goto out_drop_sem;
1843 }
1844 }
Li Zefan0caa1022010-12-20 16:30:25 +08001845
1846 trans = btrfs_start_transaction(root, 1);
1847 if (IS_ERR(trans)) {
1848 ret = PTR_ERR(trans);
1849 goto out_reset;
1850 }
1851
Li Zefanb4dc2b82011-02-16 06:06:34 +00001852 ret = btrfs_update_root(trans, root->fs_info->tree_root,
Li Zefan0caa1022010-12-20 16:30:25 +08001853 &root->root_key, &root->root_item);
1854
1855 btrfs_commit_transaction(trans, root);
1856out_reset:
1857 if (ret)
1858 btrfs_set_root_flags(&root->root_item, root_flags);
Liu Bob9ca0662012-06-29 03:58:49 -06001859out_drop_sem:
Li Zefan0caa1022010-12-20 16:30:25 +08001860 up_write(&root->fs_info->subvol_sem);
Liu Bob9ca0662012-06-29 03:58:49 -06001861out_drop_write:
1862 mnt_drop_write_file(file);
1863out:
Li Zefan0caa1022010-12-20 16:30:25 +08001864 return ret;
1865}
1866
Yan, Zheng76dda932009-09-21 16:00:26 -04001867/*
1868 * helper to check if the subvolume references other subvolumes
1869 */
1870static noinline int may_destroy_subvol(struct btrfs_root *root)
1871{
1872 struct btrfs_path *path;
Josef Bacik175a2b82013-08-12 15:36:44 -04001873 struct btrfs_dir_item *di;
Yan, Zheng76dda932009-09-21 16:00:26 -04001874 struct btrfs_key key;
Josef Bacik175a2b82013-08-12 15:36:44 -04001875 u64 dir_id;
Yan, Zheng76dda932009-09-21 16:00:26 -04001876 int ret;
1877
1878 path = btrfs_alloc_path();
1879 if (!path)
1880 return -ENOMEM;
1881
Josef Bacik175a2b82013-08-12 15:36:44 -04001882 /* Make sure this root isn't set as the default subvol */
1883 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1884 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1885 dir_id, "default", 7, 0);
1886 if (di && !IS_ERR(di)) {
1887 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1888 if (key.objectid == root->root_key.objectid) {
Guangyu Sun72de6b52014-03-11 11:24:18 -07001889 ret = -EPERM;
1890 btrfs_err(root->fs_info, "deleting default subvolume "
1891 "%llu is not allowed", key.objectid);
Josef Bacik175a2b82013-08-12 15:36:44 -04001892 goto out;
1893 }
1894 btrfs_release_path(path);
1895 }
1896
Yan, Zheng76dda932009-09-21 16:00:26 -04001897 key.objectid = root->root_key.objectid;
1898 key.type = BTRFS_ROOT_REF_KEY;
1899 key.offset = (u64)-1;
1900
1901 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1902 &key, path, 0, 0);
1903 if (ret < 0)
1904 goto out;
1905 BUG_ON(ret == 0);
1906
1907 ret = 0;
1908 if (path->slots[0] > 0) {
1909 path->slots[0]--;
1910 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1911 if (key.objectid == root->root_key.objectid &&
1912 key.type == BTRFS_ROOT_REF_KEY)
1913 ret = -ENOTEMPTY;
1914 }
1915out:
1916 btrfs_free_path(path);
1917 return ret;
1918}
1919
Chris Masonac8e9812010-02-28 15:39:26 -05001920static noinline int key_in_sk(struct btrfs_key *key,
1921 struct btrfs_ioctl_search_key *sk)
1922{
Chris Masonabc6e132010-03-18 12:10:08 -04001923 struct btrfs_key test;
1924 int ret;
1925
1926 test.objectid = sk->min_objectid;
1927 test.type = sk->min_type;
1928 test.offset = sk->min_offset;
1929
1930 ret = btrfs_comp_cpu_keys(key, &test);
1931 if (ret < 0)
Chris Masonac8e9812010-02-28 15:39:26 -05001932 return 0;
Chris Masonabc6e132010-03-18 12:10:08 -04001933
1934 test.objectid = sk->max_objectid;
1935 test.type = sk->max_type;
1936 test.offset = sk->max_offset;
1937
1938 ret = btrfs_comp_cpu_keys(key, &test);
1939 if (ret > 0)
Chris Masonac8e9812010-02-28 15:39:26 -05001940 return 0;
1941 return 1;
1942}
1943
1944static noinline int copy_to_sk(struct btrfs_root *root,
1945 struct btrfs_path *path,
1946 struct btrfs_key *key,
1947 struct btrfs_ioctl_search_key *sk,
Gerhard Heift9b6e8172014-01-30 16:24:00 +01001948 size_t *buf_size,
Gerhard Heiftba346b32014-01-30 16:24:02 +01001949 char __user *ubuf,
Chris Masonac8e9812010-02-28 15:39:26 -05001950 unsigned long *sk_offset,
1951 int *num_found)
1952{
1953 u64 found_transid;
1954 struct extent_buffer *leaf;
1955 struct btrfs_ioctl_search_header sh;
Naohiro Aotadd81d452015-06-30 11:25:43 +09001956 struct btrfs_key test;
Chris Masonac8e9812010-02-28 15:39:26 -05001957 unsigned long item_off;
1958 unsigned long item_len;
1959 int nritems;
1960 int i;
1961 int slot;
Chris Masonac8e9812010-02-28 15:39:26 -05001962 int ret = 0;
1963
1964 leaf = path->nodes[0];
1965 slot = path->slots[0];
1966 nritems = btrfs_header_nritems(leaf);
1967
1968 if (btrfs_header_generation(leaf) > sk->max_transid) {
1969 i = nritems;
1970 goto advance_key;
1971 }
1972 found_transid = btrfs_header_generation(leaf);
1973
1974 for (i = slot; i < nritems; i++) {
1975 item_off = btrfs_item_ptr_offset(leaf, i);
1976 item_len = btrfs_item_size_nr(leaf, i);
1977
Gabriel de Perthuis03b71c62013-05-06 17:40:18 +00001978 btrfs_item_key_to_cpu(leaf, key, i);
1979 if (!key_in_sk(key, sk))
1980 continue;
1981
Gerhard Heift9b6e8172014-01-30 16:24:00 +01001982 if (sizeof(sh) + item_len > *buf_size) {
Gerhard Heift8f5f6172014-01-30 16:23:59 +01001983 if (*num_found) {
1984 ret = 1;
1985 goto out;
1986 }
Chris Masonac8e9812010-02-28 15:39:26 -05001987
Gerhard Heift8f5f6172014-01-30 16:23:59 +01001988 /*
1989 * return one empty item back for v1, which does not
1990 * handle -EOVERFLOW
1991 */
1992
Gerhard Heift9b6e8172014-01-30 16:24:00 +01001993 *buf_size = sizeof(sh) + item_len;
Chris Masonac8e9812010-02-28 15:39:26 -05001994 item_len = 0;
Gerhard Heift8f5f6172014-01-30 16:23:59 +01001995 ret = -EOVERFLOW;
1996 }
Chris Masonac8e9812010-02-28 15:39:26 -05001997
Gerhard Heift9b6e8172014-01-30 16:24:00 +01001998 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
Chris Masonac8e9812010-02-28 15:39:26 -05001999 ret = 1;
Gerhard Heift25c9bc22014-01-30 16:23:57 +01002000 goto out;
Chris Masonac8e9812010-02-28 15:39:26 -05002001 }
2002
Chris Masonac8e9812010-02-28 15:39:26 -05002003 sh.objectid = key->objectid;
2004 sh.offset = key->offset;
2005 sh.type = key->type;
2006 sh.len = item_len;
2007 sh.transid = found_transid;
2008
2009 /* copy search result header */
Gerhard Heiftba346b32014-01-30 16:24:02 +01002010 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
2011 ret = -EFAULT;
2012 goto out;
2013 }
2014
Chris Masonac8e9812010-02-28 15:39:26 -05002015 *sk_offset += sizeof(sh);
2016
2017 if (item_len) {
Gerhard Heiftba346b32014-01-30 16:24:02 +01002018 char __user *up = ubuf + *sk_offset;
Chris Masonac8e9812010-02-28 15:39:26 -05002019 /* copy the item */
Gerhard Heiftba346b32014-01-30 16:24:02 +01002020 if (read_extent_buffer_to_user(leaf, up,
2021 item_off, item_len)) {
2022 ret = -EFAULT;
2023 goto out;
2024 }
2025
Chris Masonac8e9812010-02-28 15:39:26 -05002026 *sk_offset += item_len;
Chris Masonac8e9812010-02-28 15:39:26 -05002027 }
Hugo Millse2156862011-05-14 17:43:41 +00002028 (*num_found)++;
Chris Masonac8e9812010-02-28 15:39:26 -05002029
Gerhard Heift8f5f6172014-01-30 16:23:59 +01002030 if (ret) /* -EOVERFLOW from above */
2031 goto out;
2032
Gerhard Heift25c9bc22014-01-30 16:23:57 +01002033 if (*num_found >= sk->nr_items) {
2034 ret = 1;
2035 goto out;
2036 }
Chris Masonac8e9812010-02-28 15:39:26 -05002037 }
2038advance_key:
Chris Masonac8e9812010-02-28 15:39:26 -05002039 ret = 0;
Naohiro Aotadd81d452015-06-30 11:25:43 +09002040 test.objectid = sk->max_objectid;
2041 test.type = sk->max_type;
2042 test.offset = sk->max_offset;
2043 if (btrfs_comp_cpu_keys(key, &test) >= 0)
2044 ret = 1;
2045 else if (key->offset < (u64)-1)
Chris Masonabc6e132010-03-18 12:10:08 -04002046 key->offset++;
Naohiro Aotadd81d452015-06-30 11:25:43 +09002047 else if (key->type < (u8)-1) {
Chris Masonabc6e132010-03-18 12:10:08 -04002048 key->offset = 0;
2049 key->type++;
Naohiro Aotadd81d452015-06-30 11:25:43 +09002050 } else if (key->objectid < (u64)-1) {
Chris Masonabc6e132010-03-18 12:10:08 -04002051 key->offset = 0;
2052 key->type = 0;
2053 key->objectid++;
2054 } else
2055 ret = 1;
Gerhard Heift25c9bc22014-01-30 16:23:57 +01002056out:
Gerhard Heiftba346b32014-01-30 16:24:02 +01002057 /*
2058 * 0: all items from this leaf copied, continue with next
2059 * 1: * more items can be copied, but unused buffer is too small
2060 * * all items were found
2061 * Either way, it will stops the loop which iterates to the next
2062 * leaf
2063 * -EOVERFLOW: item was to large for buffer
2064 * -EFAULT: could not copy extent buffer back to userspace
2065 */
Chris Masonac8e9812010-02-28 15:39:26 -05002066 return ret;
2067}
2068
2069static noinline int search_ioctl(struct inode *inode,
Gerhard Heift12544442014-01-30 16:23:58 +01002070 struct btrfs_ioctl_search_key *sk,
Gerhard Heift9b6e8172014-01-30 16:24:00 +01002071 size_t *buf_size,
Gerhard Heiftba346b32014-01-30 16:24:02 +01002072 char __user *ubuf)
Chris Masonac8e9812010-02-28 15:39:26 -05002073{
2074 struct btrfs_root *root;
2075 struct btrfs_key key;
Chris Masonac8e9812010-02-28 15:39:26 -05002076 struct btrfs_path *path;
Chris Masonac8e9812010-02-28 15:39:26 -05002077 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
2078 int ret;
2079 int num_found = 0;
2080 unsigned long sk_offset = 0;
2081
Gerhard Heift9b6e8172014-01-30 16:24:00 +01002082 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2083 *buf_size = sizeof(struct btrfs_ioctl_search_header);
Gerhard Heift12544442014-01-30 16:23:58 +01002084 return -EOVERFLOW;
Gerhard Heift9b6e8172014-01-30 16:24:00 +01002085 }
Gerhard Heift12544442014-01-30 16:23:58 +01002086
Chris Masonac8e9812010-02-28 15:39:26 -05002087 path = btrfs_alloc_path();
2088 if (!path)
2089 return -ENOMEM;
2090
2091 if (sk->tree_id == 0) {
2092 /* search the root of the inode that was passed */
2093 root = BTRFS_I(inode)->root;
2094 } else {
2095 key.objectid = sk->tree_id;
2096 key.type = BTRFS_ROOT_ITEM_KEY;
2097 key.offset = (u64)-1;
2098 root = btrfs_read_fs_root_no_name(info, &key);
2099 if (IS_ERR(root)) {
David Sterbaf14d1042015-10-08 11:37:06 +02002100 btrfs_err(info, "could not find root %llu",
Chris Masonac8e9812010-02-28 15:39:26 -05002101 sk->tree_id);
2102 btrfs_free_path(path);
2103 return -ENOENT;
2104 }
2105 }
2106
2107 key.objectid = sk->min_objectid;
2108 key.type = sk->min_type;
2109 key.offset = sk->min_offset;
2110
Dulshani Gunawardhana67871252013-10-31 10:33:04 +05302111 while (1) {
Filipe David Borba Manana6174d3c2013-10-01 16:13:42 +01002112 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
Chris Masonac8e9812010-02-28 15:39:26 -05002113 if (ret != 0) {
2114 if (ret > 0)
2115 ret = 0;
2116 goto err;
2117 }
Gerhard Heiftba346b32014-01-30 16:24:02 +01002118 ret = copy_to_sk(root, path, &key, sk, buf_size, ubuf,
Chris Masonac8e9812010-02-28 15:39:26 -05002119 &sk_offset, &num_found);
David Sterbab3b4aa72011-04-21 01:20:15 +02002120 btrfs_release_path(path);
Gerhard Heift25c9bc22014-01-30 16:23:57 +01002121 if (ret)
Chris Masonac8e9812010-02-28 15:39:26 -05002122 break;
2123
2124 }
Gerhard Heift8f5f6172014-01-30 16:23:59 +01002125 if (ret > 0)
2126 ret = 0;
Chris Masonac8e9812010-02-28 15:39:26 -05002127err:
2128 sk->nr_items = num_found;
2129 btrfs_free_path(path);
2130 return ret;
2131}
2132
2133static noinline int btrfs_ioctl_tree_search(struct file *file,
2134 void __user *argp)
2135{
Gerhard Heiftba346b32014-01-30 16:24:02 +01002136 struct btrfs_ioctl_search_args __user *uargs;
2137 struct btrfs_ioctl_search_key sk;
Gerhard Heift9b6e8172014-01-30 16:24:00 +01002138 struct inode *inode;
2139 int ret;
2140 size_t buf_size;
Chris Masonac8e9812010-02-28 15:39:26 -05002141
2142 if (!capable(CAP_SYS_ADMIN))
2143 return -EPERM;
2144
Gerhard Heiftba346b32014-01-30 16:24:02 +01002145 uargs = (struct btrfs_ioctl_search_args __user *)argp;
Chris Masonac8e9812010-02-28 15:39:26 -05002146
Gerhard Heiftba346b32014-01-30 16:24:02 +01002147 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2148 return -EFAULT;
2149
2150 buf_size = sizeof(uargs->buf);
Chris Masonac8e9812010-02-28 15:39:26 -05002151
Al Viro496ad9a2013-01-23 17:07:38 -05002152 inode = file_inode(file);
Gerhard Heiftba346b32014-01-30 16:24:02 +01002153 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
Gerhard Heift8f5f6172014-01-30 16:23:59 +01002154
2155 /*
2156 * In the origin implementation an overflow is handled by returning a
2157 * search header with a len of zero, so reset ret.
2158 */
2159 if (ret == -EOVERFLOW)
2160 ret = 0;
2161
Gerhard Heiftba346b32014-01-30 16:24:02 +01002162 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
Chris Masonac8e9812010-02-28 15:39:26 -05002163 ret = -EFAULT;
Chris Masonac8e9812010-02-28 15:39:26 -05002164 return ret;
2165}
2166
Gerhard Heiftcc68a8a2014-01-30 16:24:03 +01002167static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2168 void __user *argp)
2169{
2170 struct btrfs_ioctl_search_args_v2 __user *uarg;
2171 struct btrfs_ioctl_search_args_v2 args;
2172 struct inode *inode;
2173 int ret;
2174 size_t buf_size;
Byongho Leeee221842015-12-15 01:42:10 +09002175 const size_t buf_limit = SZ_16M;
Gerhard Heiftcc68a8a2014-01-30 16:24:03 +01002176
2177 if (!capable(CAP_SYS_ADMIN))
2178 return -EPERM;
2179
2180 /* copy search header and buffer size */
2181 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2182 if (copy_from_user(&args, uarg, sizeof(args)))
2183 return -EFAULT;
2184
2185 buf_size = args.buf_size;
2186
2187 if (buf_size < sizeof(struct btrfs_ioctl_search_header))
2188 return -EOVERFLOW;
2189
2190 /* limit result size to 16MB */
2191 if (buf_size > buf_limit)
2192 buf_size = buf_limit;
2193
2194 inode = file_inode(file);
2195 ret = search_ioctl(inode, &args.key, &buf_size,
2196 (char *)(&uarg->buf[0]));
2197 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2198 ret = -EFAULT;
2199 else if (ret == -EOVERFLOW &&
2200 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2201 ret = -EFAULT;
2202
Yan, Zheng76dda932009-09-21 16:00:26 -04002203 return ret;
2204}
2205
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002206/*
Chris Masonac8e9812010-02-28 15:39:26 -05002207 * Search INODE_REFs to identify path name of 'dirid' directory
2208 * in a 'tree_id' tree. and sets path name to 'name'.
2209 */
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002210static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2211 u64 tree_id, u64 dirid, char *name)
2212{
2213 struct btrfs_root *root;
2214 struct btrfs_key key;
Chris Masonac8e9812010-02-28 15:39:26 -05002215 char *ptr;
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002216 int ret = -1;
2217 int slot;
2218 int len;
2219 int total_len = 0;
2220 struct btrfs_inode_ref *iref;
2221 struct extent_buffer *l;
2222 struct btrfs_path *path;
2223
2224 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2225 name[0]='\0';
2226 return 0;
2227 }
2228
2229 path = btrfs_alloc_path();
2230 if (!path)
2231 return -ENOMEM;
2232
Chris Masonac8e9812010-02-28 15:39:26 -05002233 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002234
2235 key.objectid = tree_id;
2236 key.type = BTRFS_ROOT_ITEM_KEY;
2237 key.offset = (u64)-1;
2238 root = btrfs_read_fs_root_no_name(info, &key);
2239 if (IS_ERR(root)) {
David Sterbaf14d1042015-10-08 11:37:06 +02002240 btrfs_err(info, "could not find root %llu", tree_id);
Chris Mason8ad6fca2010-03-18 12:23:10 -04002241 ret = -ENOENT;
2242 goto out;
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002243 }
2244
2245 key.objectid = dirid;
2246 key.type = BTRFS_INODE_REF_KEY;
Chris Mason8ad6fca2010-03-18 12:23:10 -04002247 key.offset = (u64)-1;
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002248
Dulshani Gunawardhana67871252013-10-31 10:33:04 +05302249 while (1) {
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002250 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2251 if (ret < 0)
2252 goto out;
Filipe David Borba Manana18674c62013-08-14 03:00:21 +01002253 else if (ret > 0) {
2254 ret = btrfs_previous_item(root, path, dirid,
2255 BTRFS_INODE_REF_KEY);
2256 if (ret < 0)
2257 goto out;
2258 else if (ret > 0) {
2259 ret = -ENOENT;
2260 goto out;
2261 }
2262 }
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002263
2264 l = path->nodes[0];
2265 slot = path->slots[0];
2266 btrfs_item_key_to_cpu(l, &key, slot);
2267
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002268 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2269 len = btrfs_inode_ref_name_len(l, iref);
2270 ptr -= len + 1;
2271 total_len += len + 1;
Filipe David Borba Mananaa696cf32013-08-14 03:00:20 +01002272 if (ptr < name) {
2273 ret = -ENAMETOOLONG;
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002274 goto out;
Filipe David Borba Mananaa696cf32013-08-14 03:00:20 +01002275 }
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002276
2277 *(ptr + len) = '/';
Dulshani Gunawardhana67871252013-10-31 10:33:04 +05302278 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002279
2280 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2281 break;
2282
David Sterbab3b4aa72011-04-21 01:20:15 +02002283 btrfs_release_path(path);
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002284 key.objectid = key.offset;
Chris Mason8ad6fca2010-03-18 12:23:10 -04002285 key.offset = (u64)-1;
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002286 dirid = key.objectid;
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002287 }
Li Zefan77906a502011-07-14 03:16:00 +00002288 memmove(name, ptr, total_len);
Dulshani Gunawardhana67871252013-10-31 10:33:04 +05302289 name[total_len] = '\0';
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002290 ret = 0;
2291out:
2292 btrfs_free_path(path);
Chris Masonac8e9812010-02-28 15:39:26 -05002293 return ret;
2294}
2295
2296static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2297 void __user *argp)
2298{
2299 struct btrfs_ioctl_ino_lookup_args *args;
2300 struct inode *inode;
David Sterba01b810b2015-05-12 19:14:49 +02002301 int ret = 0;
Chris Masonac8e9812010-02-28 15:39:26 -05002302
Julia Lawall2354d082010-10-29 15:14:18 -04002303 args = memdup_user(argp, sizeof(*args));
2304 if (IS_ERR(args))
2305 return PTR_ERR(args);
Dan Carpenterc2b96922010-03-20 11:24:15 +00002306
Al Viro496ad9a2013-01-23 17:07:38 -05002307 inode = file_inode(file);
Chris Masonac8e9812010-02-28 15:39:26 -05002308
David Sterba01b810b2015-05-12 19:14:49 +02002309 /*
2310 * Unprivileged query to obtain the containing subvolume root id. The
2311 * path is reset so it's consistent with btrfs_search_path_in_tree.
2312 */
Chris Mason1b53ac42010-03-18 12:17:05 -04002313 if (args->treeid == 0)
2314 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2315
David Sterba01b810b2015-05-12 19:14:49 +02002316 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2317 args->name[0] = 0;
2318 goto out;
2319 }
2320
2321 if (!capable(CAP_SYS_ADMIN)) {
2322 ret = -EPERM;
2323 goto out;
2324 }
2325
Chris Masonac8e9812010-02-28 15:39:26 -05002326 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2327 args->treeid, args->objectid,
2328 args->name);
2329
David Sterba01b810b2015-05-12 19:14:49 +02002330out:
Chris Masonac8e9812010-02-28 15:39:26 -05002331 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2332 ret = -EFAULT;
2333
2334 kfree(args);
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002335 return ret;
2336}
2337
Yan, Zheng76dda932009-09-21 16:00:26 -04002338static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2339 void __user *arg)
2340{
Al Viro54563d42013-09-01 15:57:51 -04002341 struct dentry *parent = file->f_path.dentry;
Yan, Zheng76dda932009-09-21 16:00:26 -04002342 struct dentry *dentry;
David Howells2b0143b2015-03-17 22:25:59 +00002343 struct inode *dir = d_inode(parent);
Yan, Zheng76dda932009-09-21 16:00:26 -04002344 struct inode *inode;
2345 struct btrfs_root *root = BTRFS_I(dir)->root;
2346 struct btrfs_root *dest = NULL;
2347 struct btrfs_ioctl_vol_args *vol_args;
2348 struct btrfs_trans_handle *trans;
Miao Xiec58aaad2013-02-28 10:05:36 +00002349 struct btrfs_block_rsv block_rsv;
David Sterba521e0542014-04-15 16:41:44 +02002350 u64 root_flags;
Miao Xiec58aaad2013-02-28 10:05:36 +00002351 u64 qgroup_reserved;
Yan, Zheng76dda932009-09-21 16:00:26 -04002352 int namelen;
2353 int ret;
2354 int err = 0;
2355
Yan, Zheng76dda932009-09-21 16:00:26 -04002356 vol_args = memdup_user(arg, sizeof(*vol_args));
2357 if (IS_ERR(vol_args))
2358 return PTR_ERR(vol_args);
2359
2360 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2361 namelen = strlen(vol_args->name);
2362 if (strchr(vol_args->name, '/') ||
2363 strncmp(vol_args->name, "..", namelen) == 0) {
2364 err = -EINVAL;
2365 goto out;
2366 }
2367
Al Viroa561be72011-11-23 11:57:51 -05002368 err = mnt_want_write_file(file);
Yan, Zheng76dda932009-09-21 16:00:26 -04002369 if (err)
2370 goto out;
2371
David Sterba521e0542014-04-15 16:41:44 +02002372
David Sterba5c50c9b2013-03-22 18:12:51 +00002373 err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2374 if (err == -EINTR)
David Sterbae43f9982013-12-06 17:51:32 +01002375 goto out_drop_write;
Yan, Zheng76dda932009-09-21 16:00:26 -04002376 dentry = lookup_one_len(vol_args->name, parent, namelen);
2377 if (IS_ERR(dentry)) {
2378 err = PTR_ERR(dentry);
2379 goto out_unlock_dir;
2380 }
2381
David Howells2b0143b2015-03-17 22:25:59 +00002382 if (d_really_is_negative(dentry)) {
Yan, Zheng76dda932009-09-21 16:00:26 -04002383 err = -ENOENT;
2384 goto out_dput;
2385 }
2386
David Howells2b0143b2015-03-17 22:25:59 +00002387 inode = d_inode(dentry);
Sage Weil4260f7c2010-10-29 15:46:43 -04002388 dest = BTRFS_I(inode)->root;
Dulshani Gunawardhana67871252013-10-31 10:33:04 +05302389 if (!capable(CAP_SYS_ADMIN)) {
Sage Weil4260f7c2010-10-29 15:46:43 -04002390 /*
2391 * Regular user. Only allow this with a special mount
2392 * option, when the user has write+exec access to the
2393 * subvol root, and when rmdir(2) would have been
2394 * allowed.
2395 *
2396 * Note that this is _not_ check that the subvol is
2397 * empty or doesn't contain data that we wouldn't
2398 * otherwise be able to delete.
2399 *
2400 * Users who want to delete empty subvols should try
2401 * rmdir(2).
2402 */
2403 err = -EPERM;
2404 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2405 goto out_dput;
2406
2407 /*
2408 * Do not allow deletion if the parent dir is the same
2409 * as the dir to be deleted. That means the ioctl
2410 * must be called on the dentry referencing the root
2411 * of the subvol, not a random directory contained
2412 * within it.
2413 */
2414 err = -EINVAL;
2415 if (root == dest)
2416 goto out_dput;
2417
2418 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2419 if (err)
2420 goto out_dput;
Sage Weil4260f7c2010-10-29 15:46:43 -04002421 }
2422
Miao Xie5c39da52012-10-22 11:39:53 +00002423 /* check if subvolume may be deleted by a user */
2424 err = btrfs_may_delete(dir, dentry, 1);
2425 if (err)
2426 goto out_dput;
2427
Li Zefan33345d012011-04-20 10:31:50 +08002428 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
Yan, Zheng76dda932009-09-21 16:00:26 -04002429 err = -EINVAL;
2430 goto out_dput;
2431 }
2432
Al Viro59551022016-01-22 15:40:57 -05002433 inode_lock(inode);
David Sterba521e0542014-04-15 16:41:44 +02002434
2435 /*
2436 * Don't allow to delete a subvolume with send in progress. This is
2437 * inside the i_mutex so the error handling that has to drop the bit
2438 * again is not run concurrently.
2439 */
2440 spin_lock(&dest->root_item_lock);
Filipe Mananac55bfa62014-05-25 03:55:44 +01002441 root_flags = btrfs_root_flags(&dest->root_item);
2442 if (dest->send_in_progress == 0) {
2443 btrfs_set_root_flags(&dest->root_item,
David Sterba521e0542014-04-15 16:41:44 +02002444 root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2445 spin_unlock(&dest->root_item_lock);
2446 } else {
2447 spin_unlock(&dest->root_item_lock);
2448 btrfs_warn(root->fs_info,
2449 "Attempt to delete subvolume %llu during send",
Filipe Mananac55bfa62014-05-25 03:55:44 +01002450 dest->root_key.objectid);
David Sterba521e0542014-04-15 16:41:44 +02002451 err = -EPERM;
Omar Sandoval909e26d2015-04-10 14:20:40 -07002452 goto out_unlock_inode;
David Sterba521e0542014-04-15 16:41:44 +02002453 }
2454
Yan, Zheng76dda932009-09-21 16:00:26 -04002455 down_write(&root->fs_info->subvol_sem);
2456
2457 err = may_destroy_subvol(dest);
2458 if (err)
2459 goto out_up_write;
2460
Miao Xiec58aaad2013-02-28 10:05:36 +00002461 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2462 /*
2463 * One for dir inode, two for dir entries, two for root
2464 * ref/backref.
2465 */
2466 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
Jeff Mahoneyee3441b2013-07-09 16:37:21 -04002467 5, &qgroup_reserved, true);
Miao Xiec58aaad2013-02-28 10:05:36 +00002468 if (err)
2469 goto out_up_write;
2470
Yan, Zhenga22285a2010-05-16 10:48:46 -04002471 trans = btrfs_start_transaction(root, 0);
2472 if (IS_ERR(trans)) {
2473 err = PTR_ERR(trans);
Miao Xiec58aaad2013-02-28 10:05:36 +00002474 goto out_release;
Yan, Zhenga22285a2010-05-16 10:48:46 -04002475 }
Miao Xiec58aaad2013-02-28 10:05:36 +00002476 trans->block_rsv = &block_rsv;
2477 trans->bytes_reserved = block_rsv.size;
Yan, Zhenga22285a2010-05-16 10:48:46 -04002478
Yan, Zheng76dda932009-09-21 16:00:26 -04002479 ret = btrfs_unlink_subvol(trans, root, dir,
2480 dest->root_key.objectid,
2481 dentry->d_name.name,
2482 dentry->d_name.len);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002483 if (ret) {
2484 err = ret;
2485 btrfs_abort_transaction(trans, root, ret);
2486 goto out_end_trans;
2487 }
Yan, Zheng76dda932009-09-21 16:00:26 -04002488
2489 btrfs_record_root_in_trans(trans, dest);
2490
2491 memset(&dest->root_item.drop_progress, 0,
2492 sizeof(dest->root_item.drop_progress));
2493 dest->root_item.drop_level = 0;
2494 btrfs_set_root_refs(&dest->root_item, 0);
2495
Miao Xie27cdeb72014-04-02 19:51:05 +08002496 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
Yan, Zhengd68fc572010-05-16 10:49:58 -04002497 ret = btrfs_insert_orphan_item(trans,
2498 root->fs_info->tree_root,
2499 dest->root_key.objectid);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002500 if (ret) {
2501 btrfs_abort_transaction(trans, root, ret);
2502 err = ret;
2503 goto out_end_trans;
2504 }
Yan, Zhengd68fc572010-05-16 10:49:58 -04002505 }
Stefan Behrensdd5f9612013-08-15 17:11:20 +02002506
2507 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2508 dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
2509 dest->root_key.objectid);
2510 if (ret && ret != -ENOENT) {
2511 btrfs_abort_transaction(trans, root, ret);
2512 err = ret;
2513 goto out_end_trans;
2514 }
2515 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2516 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2517 dest->root_item.received_uuid,
2518 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2519 dest->root_key.objectid);
2520 if (ret && ret != -ENOENT) {
2521 btrfs_abort_transaction(trans, root, ret);
2522 err = ret;
2523 goto out_end_trans;
2524 }
2525 }
2526
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002527out_end_trans:
Miao Xiec58aaad2013-02-28 10:05:36 +00002528 trans->block_rsv = NULL;
2529 trans->bytes_reserved = 0;
Sage Weil531cb132010-10-29 15:41:32 -04002530 ret = btrfs_end_transaction(trans, root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002531 if (ret && !err)
2532 err = ret;
Yan, Zheng76dda932009-09-21 16:00:26 -04002533 inode->i_flags |= S_DEAD;
Miao Xiec58aaad2013-02-28 10:05:36 +00002534out_release:
2535 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
Yan, Zheng76dda932009-09-21 16:00:26 -04002536out_up_write:
2537 up_write(&root->fs_info->subvol_sem);
David Sterba521e0542014-04-15 16:41:44 +02002538 if (err) {
2539 spin_lock(&dest->root_item_lock);
Filipe Mananac55bfa62014-05-25 03:55:44 +01002540 root_flags = btrfs_root_flags(&dest->root_item);
2541 btrfs_set_root_flags(&dest->root_item,
David Sterba521e0542014-04-15 16:41:44 +02002542 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2543 spin_unlock(&dest->root_item_lock);
2544 }
Omar Sandoval909e26d2015-04-10 14:20:40 -07002545out_unlock_inode:
Al Viro59551022016-01-22 15:40:57 -05002546 inode_unlock(inode);
Yan, Zheng76dda932009-09-21 16:00:26 -04002547 if (!err) {
Omar Sandoval64ad6c42015-06-02 17:31:00 -07002548 d_invalidate(dentry);
Yan, Zheng76dda932009-09-21 16:00:26 -04002549 btrfs_invalidate_inodes(dest);
2550 d_delete(dentry);
David Sterba61155aa2014-04-15 16:42:03 +02002551 ASSERT(dest->send_in_progress == 0);
Liu Bofa6ac872013-02-20 14:10:23 +00002552
2553 /* the last ref */
David Sterba57cdc8d2014-02-05 02:37:48 +01002554 if (dest->ino_cache_inode) {
2555 iput(dest->ino_cache_inode);
2556 dest->ino_cache_inode = NULL;
Liu Bofa6ac872013-02-20 14:10:23 +00002557 }
Yan, Zheng76dda932009-09-21 16:00:26 -04002558 }
2559out_dput:
2560 dput(dentry);
2561out_unlock_dir:
Al Viro59551022016-01-22 15:40:57 -05002562 inode_unlock(dir);
David Sterbae43f9982013-12-06 17:51:32 +01002563out_drop_write:
Al Viro2a79f172011-12-09 08:06:57 -05002564 mnt_drop_write_file(file);
Yan, Zheng76dda932009-09-21 16:00:26 -04002565out:
2566 kfree(vol_args);
2567 return err;
2568}
2569
Chris Mason1e701a32010-03-11 09:42:04 -05002570static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002571{
Al Viro496ad9a2013-01-23 17:07:38 -05002572 struct inode *inode = file_inode(file);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002573 struct btrfs_root *root = BTRFS_I(inode)->root;
Chris Mason1e701a32010-03-11 09:42:04 -05002574 struct btrfs_ioctl_defrag_range_args *range;
Yan Zhengc146afa2008-11-12 14:34:12 -05002575 int ret;
2576
Ilya Dryomov25122d12013-01-20 15:57:57 +02002577 ret = mnt_want_write_file(file);
2578 if (ret)
2579 return ret;
Li Zefanb83cc962010-12-20 16:04:08 +08002580
Ilya Dryomov25122d12013-01-20 15:57:57 +02002581 if (btrfs_root_readonly(root)) {
2582 ret = -EROFS;
2583 goto out;
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01002584 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002585
2586 switch (inode->i_mode & S_IFMT) {
2587 case S_IFDIR:
Chris Masone441d542009-01-05 16:57:23 -05002588 if (!capable(CAP_SYS_ADMIN)) {
2589 ret = -EPERM;
2590 goto out;
2591 }
Eric Sandeende78b512013-01-31 18:21:12 +00002592 ret = btrfs_defrag_root(root);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002593 if (ret)
2594 goto out;
Eric Sandeende78b512013-01-31 18:21:12 +00002595 ret = btrfs_defrag_root(root->fs_info->extent_root);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002596 break;
2597 case S_IFREG:
Chris Masone441d542009-01-05 16:57:23 -05002598 if (!(file->f_mode & FMODE_WRITE)) {
2599 ret = -EINVAL;
2600 goto out;
2601 }
Chris Mason1e701a32010-03-11 09:42:04 -05002602
2603 range = kzalloc(sizeof(*range), GFP_KERNEL);
2604 if (!range) {
2605 ret = -ENOMEM;
2606 goto out;
2607 }
2608
2609 if (argp) {
2610 if (copy_from_user(range, argp,
2611 sizeof(*range))) {
2612 ret = -EFAULT;
2613 kfree(range);
Dan Carpenter683be162010-03-20 11:24:48 +00002614 goto out;
Chris Mason1e701a32010-03-11 09:42:04 -05002615 }
2616 /* compression requires us to start the IO */
2617 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2618 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2619 range->extent_thresh = (u32)-1;
2620 }
2621 } else {
2622 /* the rest are all set to zero by kzalloc */
2623 range->len = (u64)-1;
2624 }
Al Viro496ad9a2013-01-23 17:07:38 -05002625 ret = btrfs_defrag_file(file_inode(file), file,
Chris Mason4cb53002011-05-24 15:35:30 -04002626 range, 0, 0);
2627 if (ret > 0)
2628 ret = 0;
Chris Mason1e701a32010-03-11 09:42:04 -05002629 kfree(range);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002630 break;
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002631 default:
2632 ret = -EINVAL;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002633 }
Chris Masone441d542009-01-05 16:57:23 -05002634out:
Ilya Dryomov25122d12013-01-20 15:57:57 +02002635 mnt_drop_write_file(file);
Chris Masone441d542009-01-05 16:57:23 -05002636 return ret;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002637}
2638
Christoph Hellwigb2950862008-12-02 09:54:17 -05002639static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002640{
2641 struct btrfs_ioctl_vol_args *vol_args;
2642 int ret;
2643
Chris Masone441d542009-01-05 16:57:23 -05002644 if (!capable(CAP_SYS_ADMIN))
2645 return -EPERM;
2646
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01002647 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2648 1)) {
Anand Jaine57138b2013-08-21 11:44:48 +08002649 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002650 }
2651
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01002652 mutex_lock(&root->fs_info->volume_mutex);
Li Zefandae7b662009-04-08 15:06:54 +08002653 vol_args = memdup_user(arg, sizeof(*vol_args));
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002654 if (IS_ERR(vol_args)) {
2655 ret = PTR_ERR(vol_args);
2656 goto out;
2657 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002658
Mark Fasheh5516e592008-07-24 12:20:14 -04002659 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002660 ret = btrfs_init_new_device(root, vol_args->name);
2661
Anand Jain43d20762014-07-01 00:58:56 +08002662 if (!ret)
2663 btrfs_info(root->fs_info, "disk added %s",vol_args->name);
2664
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002665 kfree(vol_args);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002666out:
2667 mutex_unlock(&root->fs_info->volume_mutex);
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01002668 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002669 return ret;
2670}
2671
Miao Xieda249272012-11-26 08:44:50 +00002672static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002673{
Al Viro496ad9a2013-01-23 17:07:38 -05002674 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002675 struct btrfs_ioctl_vol_args *vol_args;
2676 int ret;
2677
Chris Masone441d542009-01-05 16:57:23 -05002678 if (!capable(CAP_SYS_ADMIN))
2679 return -EPERM;
2680
Miao Xieda249272012-11-26 08:44:50 +00002681 ret = mnt_want_write_file(file);
2682 if (ret)
2683 return ret;
Yan Zhengc146afa2008-11-12 14:34:12 -05002684
Li Zefandae7b662009-04-08 15:06:54 +08002685 vol_args = memdup_user(arg, sizeof(*vol_args));
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002686 if (IS_ERR(vol_args)) {
2687 ret = PTR_ERR(vol_args);
Dan Carpenterc47ca322014-09-04 14:09:15 +03002688 goto err_drop;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002689 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002690
Mark Fasheh5516e592008-07-24 12:20:14 -04002691 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002692
Anand Jain183860f2013-05-17 10:52:45 +00002693 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2694 1)) {
2695 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2696 goto out;
2697 }
2698
2699 mutex_lock(&root->fs_info->volume_mutex);
2700 ret = btrfs_rm_device(root, vol_args->name);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002701 mutex_unlock(&root->fs_info->volume_mutex);
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01002702 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
Anand Jain183860f2013-05-17 10:52:45 +00002703
Anand Jainec95d492014-07-01 00:58:57 +08002704 if (!ret)
2705 btrfs_info(root->fs_info, "disk deleted %s",vol_args->name);
2706
Anand Jain183860f2013-05-17 10:52:45 +00002707out:
2708 kfree(vol_args);
Dan Carpenterc47ca322014-09-04 14:09:15 +03002709err_drop:
Ilya Dryomov4ac20c72013-01-20 15:57:57 +02002710 mnt_drop_write_file(file);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002711 return ret;
2712}
2713
Jan Schmidt475f6382011-03-11 15:41:01 +01002714static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2715{
Li Zefan027ed2f2011-06-08 08:27:56 +00002716 struct btrfs_ioctl_fs_info_args *fi_args;
Jan Schmidt475f6382011-03-11 15:41:01 +01002717 struct btrfs_device *device;
Jan Schmidt475f6382011-03-11 15:41:01 +01002718 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
Li Zefan027ed2f2011-06-08 08:27:56 +00002719 int ret = 0;
Jan Schmidt475f6382011-03-11 15:41:01 +01002720
Li Zefan027ed2f2011-06-08 08:27:56 +00002721 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2722 if (!fi_args)
2723 return -ENOMEM;
2724
Filipe David Borba Mananaf7171752013-08-12 20:56:58 +01002725 mutex_lock(&fs_devices->device_list_mutex);
Li Zefan027ed2f2011-06-08 08:27:56 +00002726 fi_args->num_devices = fs_devices->num_devices;
2727 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
Jan Schmidt475f6382011-03-11 15:41:01 +01002728
Byongho Leed7641a42015-09-01 23:10:57 +09002729 list_for_each_entry(device, &fs_devices->devices, dev_list) {
Li Zefan027ed2f2011-06-08 08:27:56 +00002730 if (device->devid > fi_args->max_id)
2731 fi_args->max_id = device->devid;
Jan Schmidt475f6382011-03-11 15:41:01 +01002732 }
2733 mutex_unlock(&fs_devices->device_list_mutex);
2734
David Sterba80a773f2014-05-07 18:17:06 +02002735 fi_args->nodesize = root->fs_info->super_copy->nodesize;
2736 fi_args->sectorsize = root->fs_info->super_copy->sectorsize;
2737 fi_args->clone_alignment = root->fs_info->super_copy->sectorsize;
2738
Li Zefan027ed2f2011-06-08 08:27:56 +00002739 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2740 ret = -EFAULT;
Jan Schmidt475f6382011-03-11 15:41:01 +01002741
Li Zefan027ed2f2011-06-08 08:27:56 +00002742 kfree(fi_args);
2743 return ret;
Jan Schmidt475f6382011-03-11 15:41:01 +01002744}
2745
2746static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2747{
2748 struct btrfs_ioctl_dev_info_args *di_args;
2749 struct btrfs_device *dev;
2750 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2751 int ret = 0;
2752 char *s_uuid = NULL;
Jan Schmidt475f6382011-03-11 15:41:01 +01002753
Jan Schmidt475f6382011-03-11 15:41:01 +01002754 di_args = memdup_user(arg, sizeof(*di_args));
2755 if (IS_ERR(di_args))
2756 return PTR_ERR(di_args);
2757
Stefan Behrensdd5f9612013-08-15 17:11:20 +02002758 if (!btrfs_is_empty_uuid(di_args->uuid))
Jan Schmidt475f6382011-03-11 15:41:01 +01002759 s_uuid = di_args->uuid;
2760
2761 mutex_lock(&fs_devices->device_list_mutex);
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01002762 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
Jan Schmidt475f6382011-03-11 15:41:01 +01002763
2764 if (!dev) {
2765 ret = -ENODEV;
2766 goto out;
2767 }
2768
2769 di_args->devid = dev->devid;
Miao Xie7cc8e582014-09-03 21:35:38 +08002770 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2771 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
Jan Schmidt475f6382011-03-11 15:41:01 +01002772 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
Jim Meyeringa27202f2012-04-26 18:36:56 +02002773 if (dev->name) {
Josef Bacik606686e2012-06-04 14:03:51 -04002774 struct rcu_string *name;
2775
2776 rcu_read_lock();
2777 name = rcu_dereference(dev->name);
2778 strncpy(di_args->path, name->str, sizeof(di_args->path));
2779 rcu_read_unlock();
Jim Meyeringa27202f2012-04-26 18:36:56 +02002780 di_args->path[sizeof(di_args->path) - 1] = 0;
2781 } else {
Stefan Behrens99ba55a2012-03-19 16:17:22 +01002782 di_args->path[0] = '\0';
Jim Meyeringa27202f2012-04-26 18:36:56 +02002783 }
Jan Schmidt475f6382011-03-11 15:41:01 +01002784
2785out:
David Sterba55793c02013-04-26 15:20:23 +00002786 mutex_unlock(&fs_devices->device_list_mutex);
Jan Schmidt475f6382011-03-11 15:41:01 +01002787 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2788 ret = -EFAULT;
2789
2790 kfree(di_args);
2791 return ret;
2792}
2793
Mark Fashehf4414602015-06-30 14:42:05 -07002794static struct page *extent_same_get_page(struct inode *inode, pgoff_t index)
Mark Fasheh416161d2013-08-06 11:42:51 -07002795{
2796 struct page *page;
Mark Fasheh416161d2013-08-06 11:42:51 -07002797
Mark Fasheh416161d2013-08-06 11:42:51 -07002798 page = grab_cache_page(inode->i_mapping, index);
2799 if (!page)
Filipe Manana31314002016-01-27 18:37:47 +00002800 return ERR_PTR(-ENOMEM);
Mark Fasheh416161d2013-08-06 11:42:51 -07002801
2802 if (!PageUptodate(page)) {
Filipe Manana31314002016-01-27 18:37:47 +00002803 int ret;
2804
2805 ret = btrfs_readpage(NULL, page);
2806 if (ret)
2807 return ERR_PTR(ret);
Mark Fasheh416161d2013-08-06 11:42:51 -07002808 lock_page(page);
2809 if (!PageUptodate(page)) {
2810 unlock_page(page);
2811 page_cache_release(page);
Filipe Manana31314002016-01-27 18:37:47 +00002812 return ERR_PTR(-EIO);
2813 }
2814 if (page->mapping != inode->i_mapping) {
2815 unlock_page(page);
2816 page_cache_release(page);
2817 return ERR_PTR(-EAGAIN);
Mark Fasheh416161d2013-08-06 11:42:51 -07002818 }
2819 }
Mark Fasheh416161d2013-08-06 11:42:51 -07002820
2821 return page;
2822}
2823
Mark Fashehf4414602015-06-30 14:42:05 -07002824static int gather_extent_pages(struct inode *inode, struct page **pages,
2825 int num_pages, u64 off)
2826{
2827 int i;
2828 pgoff_t index = off >> PAGE_CACHE_SHIFT;
2829
2830 for (i = 0; i < num_pages; i++) {
Filipe Manana31314002016-01-27 18:37:47 +00002831again:
Mark Fashehf4414602015-06-30 14:42:05 -07002832 pages[i] = extent_same_get_page(inode, index + i);
Filipe Manana31314002016-01-27 18:37:47 +00002833 if (IS_ERR(pages[i])) {
2834 int err = PTR_ERR(pages[i]);
2835
2836 if (err == -EAGAIN)
2837 goto again;
2838 pages[i] = NULL;
2839 return err;
2840 }
Mark Fashehf4414602015-06-30 14:42:05 -07002841 }
2842 return 0;
2843}
2844
Filipe Mananae0bd70c2016-01-27 10:20:58 +00002845static int lock_extent_range(struct inode *inode, u64 off, u64 len,
2846 bool retry_range_locking)
Mark Fasheh77fe20dc2013-08-06 11:42:48 -07002847{
Filipe Mananae0bd70c2016-01-27 10:20:58 +00002848 /*
2849 * Do any pending delalloc/csum calculations on inode, one way or
2850 * another, and lock file content.
2851 * The locking order is:
2852 *
2853 * 1) pages
2854 * 2) range in the inode's io tree
2855 */
Mark Fasheh77fe20dc2013-08-06 11:42:48 -07002856 while (1) {
2857 struct btrfs_ordered_extent *ordered;
2858 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2859 ordered = btrfs_lookup_first_ordered_extent(inode,
2860 off + len - 1);
Filipe Mananaff5df9b2014-05-30 17:56:24 +01002861 if ((!ordered ||
2862 ordered->file_offset + ordered->len <= off ||
2863 ordered->file_offset >= off + len) &&
Mark Fasheh77fe20dc2013-08-06 11:42:48 -07002864 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
Filipe Mananaff5df9b2014-05-30 17:56:24 +01002865 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2866 if (ordered)
2867 btrfs_put_ordered_extent(ordered);
Mark Fasheh77fe20dc2013-08-06 11:42:48 -07002868 break;
Filipe Mananaff5df9b2014-05-30 17:56:24 +01002869 }
Mark Fasheh77fe20dc2013-08-06 11:42:48 -07002870 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2871 if (ordered)
2872 btrfs_put_ordered_extent(ordered);
Filipe Mananae0bd70c2016-01-27 10:20:58 +00002873 if (!retry_range_locking)
2874 return -EAGAIN;
Mark Fasheh77fe20dc2013-08-06 11:42:48 -07002875 btrfs_wait_ordered_range(inode, off, len);
2876 }
Filipe Mananae0bd70c2016-01-27 10:20:58 +00002877 return 0;
Mark Fasheh77fe20dc2013-08-06 11:42:48 -07002878}
2879
Mark Fashehf4414602015-06-30 14:42:05 -07002880static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2)
Mark Fasheh416161d2013-08-06 11:42:51 -07002881{
Al Viro59551022016-01-22 15:40:57 -05002882 inode_unlock(inode1);
2883 inode_unlock(inode2);
Mark Fasheh416161d2013-08-06 11:42:51 -07002884}
2885
Mark Fashehf4414602015-06-30 14:42:05 -07002886static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2)
2887{
2888 if (inode1 < inode2)
2889 swap(inode1, inode2);
2890
Al Viro59551022016-01-22 15:40:57 -05002891 inode_lock_nested(inode1, I_MUTEX_PARENT);
2892 inode_lock_nested(inode2, I_MUTEX_CHILD);
Mark Fashehf4414602015-06-30 14:42:05 -07002893}
2894
2895static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1,
2896 struct inode *inode2, u64 loff2, u64 len)
2897{
2898 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2899 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2900}
2901
Filipe Mananae0bd70c2016-01-27 10:20:58 +00002902static int btrfs_double_extent_lock(struct inode *inode1, u64 loff1,
2903 struct inode *inode2, u64 loff2, u64 len,
2904 bool retry_range_locking)
Mark Fasheh416161d2013-08-06 11:42:51 -07002905{
Filipe Mananae0bd70c2016-01-27 10:20:58 +00002906 int ret;
2907
Mark Fasheh416161d2013-08-06 11:42:51 -07002908 if (inode1 < inode2) {
2909 swap(inode1, inode2);
2910 swap(loff1, loff2);
2911 }
Filipe Mananae0bd70c2016-01-27 10:20:58 +00002912 ret = lock_extent_range(inode1, loff1, len, retry_range_locking);
2913 if (ret)
2914 return ret;
2915 ret = lock_extent_range(inode2, loff2, len, retry_range_locking);
2916 if (ret)
2917 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1,
2918 loff1 + len - 1);
2919 return ret;
Mark Fashehf4414602015-06-30 14:42:05 -07002920}
2921
2922struct cmp_pages {
2923 int num_pages;
2924 struct page **src_pages;
2925 struct page **dst_pages;
2926};
2927
2928static void btrfs_cmp_data_free(struct cmp_pages *cmp)
2929{
2930 int i;
2931 struct page *pg;
2932
2933 for (i = 0; i < cmp->num_pages; i++) {
2934 pg = cmp->src_pages[i];
Filipe Mananae0bd70c2016-01-27 10:20:58 +00002935 if (pg) {
2936 unlock_page(pg);
Mark Fashehf4414602015-06-30 14:42:05 -07002937 page_cache_release(pg);
Filipe Mananae0bd70c2016-01-27 10:20:58 +00002938 }
Mark Fashehf4414602015-06-30 14:42:05 -07002939 pg = cmp->dst_pages[i];
Filipe Mananae0bd70c2016-01-27 10:20:58 +00002940 if (pg) {
2941 unlock_page(pg);
Mark Fashehf4414602015-06-30 14:42:05 -07002942 page_cache_release(pg);
Filipe Mananae0bd70c2016-01-27 10:20:58 +00002943 }
Mark Fasheh416161d2013-08-06 11:42:51 -07002944 }
Mark Fashehf4414602015-06-30 14:42:05 -07002945 kfree(cmp->src_pages);
2946 kfree(cmp->dst_pages);
2947}
2948
2949static int btrfs_cmp_data_prepare(struct inode *src, u64 loff,
2950 struct inode *dst, u64 dst_loff,
2951 u64 len, struct cmp_pages *cmp)
2952{
2953 int ret;
2954 int num_pages = PAGE_CACHE_ALIGN(len) >> PAGE_CACHE_SHIFT;
2955 struct page **src_pgarr, **dst_pgarr;
2956
2957 /*
2958 * We must gather up all the pages before we initiate our
2959 * extent locking. We use an array for the page pointers. Size
2960 * of the array is bounded by len, which is in turn bounded by
2961 * BTRFS_MAX_DEDUPE_LEN.
2962 */
2963 src_pgarr = kzalloc(num_pages * sizeof(struct page *), GFP_NOFS);
2964 dst_pgarr = kzalloc(num_pages * sizeof(struct page *), GFP_NOFS);
2965 if (!src_pgarr || !dst_pgarr) {
2966 kfree(src_pgarr);
2967 kfree(dst_pgarr);
2968 return -ENOMEM;
2969 }
2970 cmp->num_pages = num_pages;
2971 cmp->src_pages = src_pgarr;
2972 cmp->dst_pages = dst_pgarr;
2973
2974 ret = gather_extent_pages(src, cmp->src_pages, cmp->num_pages, loff);
2975 if (ret)
2976 goto out;
2977
2978 ret = gather_extent_pages(dst, cmp->dst_pages, cmp->num_pages, dst_loff);
2979
2980out:
2981 if (ret)
2982 btrfs_cmp_data_free(cmp);
2983 return 0;
Mark Fasheh416161d2013-08-06 11:42:51 -07002984}
2985
2986static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
Mark Fashehf4414602015-06-30 14:42:05 -07002987 u64 dst_loff, u64 len, struct cmp_pages *cmp)
Mark Fasheh416161d2013-08-06 11:42:51 -07002988{
2989 int ret = 0;
Mark Fashehf4414602015-06-30 14:42:05 -07002990 int i;
Mark Fasheh416161d2013-08-06 11:42:51 -07002991 struct page *src_page, *dst_page;
2992 unsigned int cmp_len = PAGE_CACHE_SIZE;
2993 void *addr, *dst_addr;
2994
Mark Fashehf4414602015-06-30 14:42:05 -07002995 i = 0;
Mark Fasheh416161d2013-08-06 11:42:51 -07002996 while (len) {
2997 if (len < PAGE_CACHE_SIZE)
2998 cmp_len = len;
2999
Mark Fashehf4414602015-06-30 14:42:05 -07003000 BUG_ON(i >= cmp->num_pages);
3001
3002 src_page = cmp->src_pages[i];
3003 dst_page = cmp->dst_pages[i];
Filipe Mananae0bd70c2016-01-27 10:20:58 +00003004 ASSERT(PageLocked(src_page));
3005 ASSERT(PageLocked(dst_page));
Mark Fashehf4414602015-06-30 14:42:05 -07003006
Mark Fasheh416161d2013-08-06 11:42:51 -07003007 addr = kmap_atomic(src_page);
3008 dst_addr = kmap_atomic(dst_page);
3009
3010 flush_dcache_page(src_page);
3011 flush_dcache_page(dst_page);
3012
3013 if (memcmp(addr, dst_addr, cmp_len))
Darrick J. Wong2b3909f2015-12-19 00:56:05 -08003014 ret = -EBADE;
Mark Fasheh416161d2013-08-06 11:42:51 -07003015
3016 kunmap_atomic(addr);
3017 kunmap_atomic(dst_addr);
Mark Fasheh416161d2013-08-06 11:42:51 -07003018
3019 if (ret)
3020 break;
3021
Mark Fasheh416161d2013-08-06 11:42:51 -07003022 len -= cmp_len;
Mark Fashehf4414602015-06-30 14:42:05 -07003023 i++;
Mark Fasheh416161d2013-08-06 11:42:51 -07003024 }
3025
3026 return ret;
3027}
3028
Mark Fashehe1d227a2015-06-08 15:05:25 -07003029static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen,
3030 u64 olen)
Mark Fasheh416161d2013-08-06 11:42:51 -07003031{
Mark Fashehe1d227a2015-06-08 15:05:25 -07003032 u64 len = *plen;
Mark Fasheh416161d2013-08-06 11:42:51 -07003033 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
3034
Mark Fashehe1d227a2015-06-08 15:05:25 -07003035 if (off + olen > inode->i_size || off + olen < off)
Mark Fasheh416161d2013-08-06 11:42:51 -07003036 return -EINVAL;
Mark Fashehe1d227a2015-06-08 15:05:25 -07003037
3038 /* if we extend to eof, continue to block boundary */
3039 if (off + len == inode->i_size)
3040 *plen = len = ALIGN(inode->i_size, bs) - off;
3041
Mark Fasheh416161d2013-08-06 11:42:51 -07003042 /* Check that we are block aligned - btrfs_clone() requires this */
3043 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
3044 return -EINVAL;
3045
3046 return 0;
3047}
3048
Mark Fashehe1d227a2015-06-08 15:05:25 -07003049static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen,
Mark Fasheh416161d2013-08-06 11:42:51 -07003050 struct inode *dst, u64 dst_loff)
3051{
3052 int ret;
Mark Fashehe1d227a2015-06-08 15:05:25 -07003053 u64 len = olen;
Mark Fashehf4414602015-06-30 14:42:05 -07003054 struct cmp_pages cmp;
Mark Fasheh0efa9f42015-06-30 14:42:07 -07003055 int same_inode = 0;
3056 u64 same_lock_start = 0;
3057 u64 same_lock_len = 0;
Mark Fasheh416161d2013-08-06 11:42:51 -07003058
Mark Fasheh416161d2013-08-06 11:42:51 -07003059 if (src == dst)
Mark Fasheh0efa9f42015-06-30 14:42:07 -07003060 same_inode = 1;
Mark Fasheh416161d2013-08-06 11:42:51 -07003061
Filipe Manana113e8282015-03-30 18:26:47 +01003062 if (len == 0)
3063 return 0;
3064
Mark Fasheh0efa9f42015-06-30 14:42:07 -07003065 if (same_inode) {
Al Viro59551022016-01-22 15:40:57 -05003066 inode_lock(src);
Mark Fasheh416161d2013-08-06 11:42:51 -07003067
Mark Fasheh0efa9f42015-06-30 14:42:07 -07003068 ret = extent_same_check_offsets(src, loff, &len, olen);
3069 if (ret)
3070 goto out_unlock;
Mark Fasheh416161d2013-08-06 11:42:51 -07003071
Mark Fasheh0efa9f42015-06-30 14:42:07 -07003072 /*
3073 * Single inode case wants the same checks, except we
3074 * don't want our length pushed out past i_size as
3075 * comparing that data range makes no sense.
3076 *
3077 * extent_same_check_offsets() will do this for an
3078 * unaligned length at i_size, so catch it here and
3079 * reject the request.
3080 *
3081 * This effectively means we require aligned extents
3082 * for the single-inode case, whereas the other cases
3083 * allow an unaligned length so long as it ends at
3084 * i_size.
3085 */
3086 if (len != olen) {
3087 ret = -EINVAL;
3088 goto out_unlock;
3089 }
3090
3091 /* Check for overlapping ranges */
3092 if (dst_loff + len > loff && dst_loff < loff + len) {
3093 ret = -EINVAL;
3094 goto out_unlock;
3095 }
3096
3097 same_lock_start = min_t(u64, loff, dst_loff);
3098 same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start;
3099 } else {
3100 btrfs_double_inode_lock(src, dst);
3101
3102 ret = extent_same_check_offsets(src, loff, &len, olen);
3103 if (ret)
3104 goto out_unlock;
3105
3106 ret = extent_same_check_offsets(dst, dst_loff, &len, olen);
3107 if (ret)
3108 goto out_unlock;
3109 }
Mark Fasheh416161d2013-08-06 11:42:51 -07003110
3111 /* don't make the dst file partly checksummed */
3112 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3113 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
3114 ret = -EINVAL;
3115 goto out_unlock;
3116 }
3117
Filipe Mananae0bd70c2016-01-27 10:20:58 +00003118again:
Mark Fashehf4414602015-06-30 14:42:05 -07003119 ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, &cmp);
3120 if (ret)
3121 goto out_unlock;
3122
Mark Fasheh0efa9f42015-06-30 14:42:07 -07003123 if (same_inode)
Filipe Mananae0bd70c2016-01-27 10:20:58 +00003124 ret = lock_extent_range(src, same_lock_start, same_lock_len,
3125 false);
Mark Fasheh0efa9f42015-06-30 14:42:07 -07003126 else
Filipe Mananae0bd70c2016-01-27 10:20:58 +00003127 ret = btrfs_double_extent_lock(src, loff, dst, dst_loff, len,
3128 false);
3129 /*
3130 * If one of the inodes has dirty pages in the respective range or
3131 * ordered extents, we need to flush dellaloc and wait for all ordered
3132 * extents in the range. We must unlock the pages and the ranges in the
3133 * io trees to avoid deadlocks when flushing delalloc (requires locking
3134 * pages) and when waiting for ordered extents to complete (they require
3135 * range locking).
3136 */
3137 if (ret == -EAGAIN) {
3138 /*
3139 * Ranges in the io trees already unlocked. Now unlock all
3140 * pages before waiting for all IO to complete.
3141 */
3142 btrfs_cmp_data_free(&cmp);
3143 if (same_inode) {
3144 btrfs_wait_ordered_range(src, same_lock_start,
3145 same_lock_len);
3146 } else {
3147 btrfs_wait_ordered_range(src, loff, len);
3148 btrfs_wait_ordered_range(dst, dst_loff, len);
3149 }
3150 goto again;
3151 }
3152 ASSERT(ret == 0);
3153 if (WARN_ON(ret)) {
3154 /* ranges in the io trees already unlocked */
3155 btrfs_cmp_data_free(&cmp);
3156 return ret;
3157 }
Mark Fashehf4414602015-06-30 14:42:05 -07003158
Mark Fasheh207910d2015-06-30 14:42:04 -07003159 /* pass original length for comparison so we stay within i_size */
Mark Fashehf4414602015-06-30 14:42:05 -07003160 ret = btrfs_cmp_data(src, loff, dst, dst_loff, olen, &cmp);
Mark Fasheh416161d2013-08-06 11:42:51 -07003161 if (ret == 0)
Mark Fasheh1c919a52015-06-30 14:42:08 -07003162 ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1);
Mark Fasheh416161d2013-08-06 11:42:51 -07003163
Mark Fasheh0efa9f42015-06-30 14:42:07 -07003164 if (same_inode)
3165 unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start,
3166 same_lock_start + same_lock_len - 1);
3167 else
3168 btrfs_double_extent_unlock(src, loff, dst, dst_loff, len);
Mark Fashehf4414602015-06-30 14:42:05 -07003169
3170 btrfs_cmp_data_free(&cmp);
Mark Fasheh416161d2013-08-06 11:42:51 -07003171out_unlock:
Mark Fasheh0efa9f42015-06-30 14:42:07 -07003172 if (same_inode)
Al Viro59551022016-01-22 15:40:57 -05003173 inode_unlock(src);
Mark Fasheh0efa9f42015-06-30 14:42:07 -07003174 else
3175 btrfs_double_inode_unlock(src, dst);
Mark Fasheh416161d2013-08-06 11:42:51 -07003176
3177 return ret;
3178}
3179
Byongho Leeee221842015-12-15 01:42:10 +09003180#define BTRFS_MAX_DEDUPE_LEN SZ_16M
Mark Fasheh416161d2013-08-06 11:42:51 -07003181
Darrick J. Wong2b3909f2015-12-19 00:56:05 -08003182ssize_t btrfs_dedupe_file_range(struct file *src_file, u64 loff, u64 olen,
3183 struct file *dst_file, u64 dst_loff)
Mark Fasheh416161d2013-08-06 11:42:51 -07003184{
Darrick J. Wong2b3909f2015-12-19 00:56:05 -08003185 struct inode *src = file_inode(src_file);
3186 struct inode *dst = file_inode(dst_file);
Mark Fasheh416161d2013-08-06 11:42:51 -07003187 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
Darrick J. Wong2b3909f2015-12-19 00:56:05 -08003188 ssize_t res;
Mark Fasheh416161d2013-08-06 11:42:51 -07003189
Darrick J. Wong2b3909f2015-12-19 00:56:05 -08003190 if (olen > BTRFS_MAX_DEDUPE_LEN)
3191 olen = BTRFS_MAX_DEDUPE_LEN;
Mark Fasheh416161d2013-08-06 11:42:51 -07003192
3193 if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) {
3194 /*
3195 * Btrfs does not support blocksize < page_size. As a
3196 * result, btrfs_cmp_data() won't correctly handle
3197 * this situation without an update.
3198 */
Darrick J. Wong2b3909f2015-12-19 00:56:05 -08003199 return -EINVAL;
Mark Fasheh416161d2013-08-06 11:42:51 -07003200 }
3201
Darrick J. Wong2b3909f2015-12-19 00:56:05 -08003202 res = btrfs_extent_same(src, loff, olen, dst, dst_loff);
3203 if (res)
3204 return res;
3205 return olen;
Mark Fasheh416161d2013-08-06 11:42:51 -07003206}
3207
Filipe Mananaf82a9902014-06-01 01:50:28 +01003208static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3209 struct inode *inode,
3210 u64 endoff,
3211 const u64 destoff,
Mark Fasheh1c919a52015-06-30 14:42:08 -07003212 const u64 olen,
3213 int no_time_update)
Filipe Mananaf82a9902014-06-01 01:50:28 +01003214{
3215 struct btrfs_root *root = BTRFS_I(inode)->root;
3216 int ret;
3217
3218 inode_inc_iversion(inode);
Mark Fasheh1c919a52015-06-30 14:42:08 -07003219 if (!no_time_update)
3220 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
Filipe Mananaf82a9902014-06-01 01:50:28 +01003221 /*
3222 * We round up to the block size at eof when determining which
3223 * extents to clone above, but shouldn't round up the file size.
3224 */
3225 if (endoff > destoff + olen)
3226 endoff = destoff + olen;
3227 if (endoff > inode->i_size)
3228 btrfs_i_size_write(inode, endoff);
3229
3230 ret = btrfs_update_inode(trans, root, inode);
3231 if (ret) {
3232 btrfs_abort_transaction(trans, root, ret);
3233 btrfs_end_transaction(trans, root);
3234 goto out;
3235 }
3236 ret = btrfs_end_transaction(trans, root);
3237out:
3238 return ret;
3239}
3240
Filipe Manana7ffbb592014-06-09 03:48:05 +01003241static void clone_update_extent_map(struct inode *inode,
3242 const struct btrfs_trans_handle *trans,
3243 const struct btrfs_path *path,
Filipe Manana7ffbb592014-06-09 03:48:05 +01003244 const u64 hole_offset,
3245 const u64 hole_len)
3246{
3247 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
3248 struct extent_map *em;
3249 int ret;
3250
3251 em = alloc_extent_map();
3252 if (!em) {
3253 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3254 &BTRFS_I(inode)->runtime_flags);
3255 return;
3256 }
3257
Filipe Manana14f59792014-06-29 21:45:40 +01003258 if (path) {
3259 struct btrfs_file_extent_item *fi;
3260
3261 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3262 struct btrfs_file_extent_item);
Filipe Manana7ffbb592014-06-09 03:48:05 +01003263 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3264 em->generation = -1;
3265 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3266 BTRFS_FILE_EXTENT_INLINE)
3267 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3268 &BTRFS_I(inode)->runtime_flags);
3269 } else {
3270 em->start = hole_offset;
3271 em->len = hole_len;
3272 em->ram_bytes = em->len;
3273 em->orig_start = hole_offset;
3274 em->block_start = EXTENT_MAP_HOLE;
3275 em->block_len = 0;
3276 em->orig_block_len = 0;
3277 em->compress_type = BTRFS_COMPRESS_NONE;
3278 em->generation = trans->transid;
3279 }
3280
3281 while (1) {
3282 write_lock(&em_tree->lock);
3283 ret = add_extent_mapping(em_tree, em, 1);
3284 write_unlock(&em_tree->lock);
3285 if (ret != -EEXIST) {
3286 free_extent_map(em);
3287 break;
3288 }
3289 btrfs_drop_extent_cache(inode, em->start,
3290 em->start + em->len - 1, 0);
3291 }
3292
David Sterbaee39b432014-09-30 01:33:33 +02003293 if (ret)
Filipe Manana7ffbb592014-06-09 03:48:05 +01003294 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3295 &BTRFS_I(inode)->runtime_flags);
3296}
3297
Filipe Manana8039d872015-10-13 15:15:00 +01003298/*
3299 * Make sure we do not end up inserting an inline extent into a file that has
3300 * already other (non-inline) extents. If a file has an inline extent it can
3301 * not have any other extents and the (single) inline extent must start at the
3302 * file offset 0. Failing to respect these rules will lead to file corruption,
3303 * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc
3304 *
3305 * We can have extents that have been already written to disk or we can have
3306 * dirty ranges still in delalloc, in which case the extent maps and items are
3307 * created only when we run delalloc, and the delalloc ranges might fall outside
3308 * the range we are currently locking in the inode's io tree. So we check the
3309 * inode's i_size because of that (i_size updates are done while holding the
3310 * i_mutex, which we are holding here).
3311 * We also check to see if the inode has a size not greater than "datal" but has
3312 * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are
3313 * protected against such concurrent fallocate calls by the i_mutex).
3314 *
3315 * If the file has no extents but a size greater than datal, do not allow the
3316 * copy because we would need turn the inline extent into a non-inline one (even
3317 * with NO_HOLES enabled). If we find our destination inode only has one inline
3318 * extent, just overwrite it with the source inline extent if its size is less
3319 * than the source extent's size, or we could copy the source inline extent's
3320 * data into the destination inode's inline extent if the later is greater then
3321 * the former.
3322 */
3323static int clone_copy_inline_extent(struct inode *src,
3324 struct inode *dst,
3325 struct btrfs_trans_handle *trans,
3326 struct btrfs_path *path,
3327 struct btrfs_key *new_key,
3328 const u64 drop_start,
3329 const u64 datal,
3330 const u64 skip,
3331 const u64 size,
3332 char *inline_data)
3333{
3334 struct btrfs_root *root = BTRFS_I(dst)->root;
3335 const u64 aligned_end = ALIGN(new_key->offset + datal,
3336 root->sectorsize);
3337 int ret;
3338 struct btrfs_key key;
3339
3340 if (new_key->offset > 0)
3341 return -EOPNOTSUPP;
3342
3343 key.objectid = btrfs_ino(dst);
3344 key.type = BTRFS_EXTENT_DATA_KEY;
3345 key.offset = 0;
3346 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3347 if (ret < 0) {
3348 return ret;
3349 } else if (ret > 0) {
3350 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
3351 ret = btrfs_next_leaf(root, path);
3352 if (ret < 0)
3353 return ret;
3354 else if (ret > 0)
3355 goto copy_inline_extent;
3356 }
3357 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
3358 if (key.objectid == btrfs_ino(dst) &&
3359 key.type == BTRFS_EXTENT_DATA_KEY) {
3360 ASSERT(key.offset > 0);
3361 return -EOPNOTSUPP;
3362 }
3363 } else if (i_size_read(dst) <= datal) {
3364 struct btrfs_file_extent_item *ei;
3365 u64 ext_len;
3366
3367 /*
3368 * If the file size is <= datal, make sure there are no other
3369 * extents following (can happen do to an fallocate call with
3370 * the flag FALLOC_FL_KEEP_SIZE).
3371 */
3372 ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
3373 struct btrfs_file_extent_item);
3374 /*
3375 * If it's an inline extent, it can not have other extents
3376 * following it.
3377 */
3378 if (btrfs_file_extent_type(path->nodes[0], ei) ==
3379 BTRFS_FILE_EXTENT_INLINE)
3380 goto copy_inline_extent;
3381
3382 ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei);
3383 if (ext_len > aligned_end)
3384 return -EOPNOTSUPP;
3385
3386 ret = btrfs_next_item(root, path);
3387 if (ret < 0) {
3388 return ret;
3389 } else if (ret == 0) {
3390 btrfs_item_key_to_cpu(path->nodes[0], &key,
3391 path->slots[0]);
3392 if (key.objectid == btrfs_ino(dst) &&
3393 key.type == BTRFS_EXTENT_DATA_KEY)
3394 return -EOPNOTSUPP;
3395 }
3396 }
3397
3398copy_inline_extent:
3399 /*
3400 * We have no extent items, or we have an extent at offset 0 which may
3401 * or may not be inlined. All these cases are dealt the same way.
3402 */
3403 if (i_size_read(dst) > datal) {
3404 /*
3405 * If the destination inode has an inline extent...
3406 * This would require copying the data from the source inline
3407 * extent into the beginning of the destination's inline extent.
3408 * But this is really complex, both extents can be compressed
3409 * or just one of them, which would require decompressing and
3410 * re-compressing data (which could increase the new compressed
3411 * size, not allowing the compressed data to fit anymore in an
3412 * inline extent).
3413 * So just don't support this case for now (it should be rare,
3414 * we are not really saving space when cloning inline extents).
3415 */
3416 return -EOPNOTSUPP;
3417 }
3418
3419 btrfs_release_path(path);
3420 ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1);
3421 if (ret)
3422 return ret;
3423 ret = btrfs_insert_empty_item(trans, root, path, new_key, size);
3424 if (ret)
3425 return ret;
3426
3427 if (skip) {
3428 const u32 start = btrfs_file_extent_calc_inline_size(0);
3429
3430 memmove(inline_data + start, inline_data + start + skip, datal);
3431 }
3432
3433 write_extent_buffer(path->nodes[0], inline_data,
3434 btrfs_item_ptr_offset(path->nodes[0],
3435 path->slots[0]),
3436 size);
3437 inode_add_bytes(dst, datal);
3438
3439 return 0;
3440}
3441
Mark Fasheh32b7c682013-08-06 11:42:49 -07003442/**
3443 * btrfs_clone() - clone a range from inode file to another
3444 *
3445 * @src: Inode to clone from
3446 * @inode: Inode to clone to
3447 * @off: Offset within source to start clone from
3448 * @olen: Original length, passed by user, of range to clone
Mark Fasheh1c919a52015-06-30 14:42:08 -07003449 * @olen_aligned: Block-aligned value of olen
Mark Fasheh32b7c682013-08-06 11:42:49 -07003450 * @destoff: Offset within @inode to start clone
Mark Fasheh1c919a52015-06-30 14:42:08 -07003451 * @no_time_update: Whether to update mtime/ctime on the target inode
Mark Fasheh32b7c682013-08-06 11:42:49 -07003452 */
3453static int btrfs_clone(struct inode *src, struct inode *inode,
Filipe Mananaf82a9902014-06-01 01:50:28 +01003454 const u64 off, const u64 olen, const u64 olen_aligned,
Mark Fasheh1c919a52015-06-30 14:42:08 -07003455 const u64 destoff, int no_time_update)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003456{
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003457 struct btrfs_root *root = BTRFS_I(inode)->root;
Mark Fasheh32b7c682013-08-06 11:42:49 -07003458 struct btrfs_path *path = NULL;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003459 struct extent_buffer *leaf;
Mark Fasheh32b7c682013-08-06 11:42:49 -07003460 struct btrfs_trans_handle *trans;
3461 char *buf = NULL;
Yan Zhengae01a0a2008-08-04 23:23:47 -04003462 struct btrfs_key key;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003463 u32 nritems;
3464 int slot;
Yan Zhengae01a0a2008-08-04 23:23:47 -04003465 int ret;
Filipe Mananaf82a9902014-06-01 01:50:28 +01003466 const u64 len = olen_aligned;
Filipe Mananaf82a9902014-06-01 01:50:28 +01003467 u64 last_dest_end = destoff;
Yan Zhengae01a0a2008-08-04 23:23:47 -04003468
3469 ret = -ENOMEM;
David Sterba707e8a02014-06-04 19:22:26 +02003470 buf = vmalloc(root->nodesize);
Yan Zhengae01a0a2008-08-04 23:23:47 -04003471 if (!buf)
Mark Fasheh32b7c682013-08-06 11:42:49 -07003472 return ret;
Yan Zhengae01a0a2008-08-04 23:23:47 -04003473
3474 path = btrfs_alloc_path();
3475 if (!path) {
3476 vfree(buf);
Mark Fasheh32b7c682013-08-06 11:42:49 -07003477 return ret;
Yan Zhengae01a0a2008-08-04 23:23:47 -04003478 }
Mark Fasheh32b7c682013-08-06 11:42:49 -07003479
David Sterbae4058b52015-11-27 16:31:35 +01003480 path->reada = READA_FORWARD;
Sage Weilc5c9cd42008-11-12 14:32:25 -05003481 /* clone data */
Li Zefan33345d012011-04-20 10:31:50 +08003482 key.objectid = btrfs_ino(src);
Yan Zhengae01a0a2008-08-04 23:23:47 -04003483 key.type = BTRFS_EXTENT_DATA_KEY;
Filipe Manana2c463822014-05-31 02:31:05 +01003484 key.offset = off;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003485
3486 while (1) {
Chris Masonde249e62015-04-11 05:09:06 -07003487 u64 next_key_min_offset = key.offset + 1;
Filipe Mananadf858e72015-03-31 14:56:46 +01003488
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003489 /*
3490 * note the key will change type as we walk through the
3491 * tree.
3492 */
Filipe David Borba Mananae4355f32014-01-13 19:35:01 +00003493 path->leave_spinning = 1;
David Sterba362a20c2011-08-01 18:11:57 +02003494 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3495 0, 0);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003496 if (ret < 0)
3497 goto out;
Filipe Manana2c463822014-05-31 02:31:05 +01003498 /*
3499 * First search, if no extent item that starts at offset off was
3500 * found but the previous item is an extent item, it's possible
3501 * it might overlap our target range, therefore process it.
3502 */
3503 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3504 btrfs_item_key_to_cpu(path->nodes[0], &key,
3505 path->slots[0] - 1);
3506 if (key.type == BTRFS_EXTENT_DATA_KEY)
3507 path->slots[0]--;
3508 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003509
Yan Zhengae01a0a2008-08-04 23:23:47 -04003510 nritems = btrfs_header_nritems(path->nodes[0]);
Filipe David Borba Mananae4355f32014-01-13 19:35:01 +00003511process_slot:
Yan Zhengae01a0a2008-08-04 23:23:47 -04003512 if (path->slots[0] >= nritems) {
David Sterba362a20c2011-08-01 18:11:57 +02003513 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003514 if (ret < 0)
3515 goto out;
3516 if (ret > 0)
3517 break;
Yan Zhengae01a0a2008-08-04 23:23:47 -04003518 nritems = btrfs_header_nritems(path->nodes[0]);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003519 }
3520 leaf = path->nodes[0];
3521 slot = path->slots[0];
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003522
Yan Zhengae01a0a2008-08-04 23:23:47 -04003523 btrfs_item_key_to_cpu(leaf, &key, slot);
David Sterba962a2982014-06-04 18:41:45 +02003524 if (key.type > BTRFS_EXTENT_DATA_KEY ||
Li Zefan33345d012011-04-20 10:31:50 +08003525 key.objectid != btrfs_ino(src))
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003526 break;
3527
David Sterba962a2982014-06-04 18:41:45 +02003528 if (key.type == BTRFS_EXTENT_DATA_KEY) {
Sage Weilc5c9cd42008-11-12 14:32:25 -05003529 struct btrfs_file_extent_item *extent;
3530 int type;
Zheng Yan31840ae2008-09-23 13:14:14 -04003531 u32 size;
3532 struct btrfs_key new_key;
Sage Weilc5c9cd42008-11-12 14:32:25 -05003533 u64 disko = 0, diskl = 0;
3534 u64 datao = 0, datal = 0;
3535 u8 comp;
Filipe Mananaf82a9902014-06-01 01:50:28 +01003536 u64 drop_start;
Zheng Yan31840ae2008-09-23 13:14:14 -04003537
Sage Weilc5c9cd42008-11-12 14:32:25 -05003538 extent = btrfs_item_ptr(leaf, slot,
3539 struct btrfs_file_extent_item);
3540 comp = btrfs_file_extent_compression(leaf, extent);
3541 type = btrfs_file_extent_type(leaf, extent);
Chris Masonc8a894d2009-06-27 21:07:03 -04003542 if (type == BTRFS_FILE_EXTENT_REG ||
3543 type == BTRFS_FILE_EXTENT_PREALLOC) {
Chris Masond3977122009-01-05 21:25:51 -05003544 disko = btrfs_file_extent_disk_bytenr(leaf,
3545 extent);
3546 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3547 extent);
Sage Weilc5c9cd42008-11-12 14:32:25 -05003548 datao = btrfs_file_extent_offset(leaf, extent);
Chris Masond3977122009-01-05 21:25:51 -05003549 datal = btrfs_file_extent_num_bytes(leaf,
3550 extent);
Sage Weilc5c9cd42008-11-12 14:32:25 -05003551 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3552 /* take upper bound, may be compressed */
3553 datal = btrfs_file_extent_ram_bytes(leaf,
3554 extent);
3555 }
Zheng Yan31840ae2008-09-23 13:14:14 -04003556
Filipe Manana2c463822014-05-31 02:31:05 +01003557 /*
3558 * The first search might have left us at an extent
3559 * item that ends before our target range's start, can
3560 * happen if we have holes and NO_HOLES feature enabled.
3561 */
3562 if (key.offset + datal <= off) {
Filipe David Borba Mananae4355f32014-01-13 19:35:01 +00003563 path->slots[0]++;
3564 goto process_slot;
Filipe Manana2c463822014-05-31 02:31:05 +01003565 } else if (key.offset >= off + len) {
3566 break;
Filipe David Borba Mananae4355f32014-01-13 19:35:01 +00003567 }
Filipe Mananadf858e72015-03-31 14:56:46 +01003568 next_key_min_offset = key.offset + datal;
Filipe David Borba Mananae4355f32014-01-13 19:35:01 +00003569 size = btrfs_item_size_nr(leaf, slot);
3570 read_extent_buffer(leaf, buf,
3571 btrfs_item_ptr_offset(leaf, slot),
3572 size);
3573
3574 btrfs_release_path(path);
3575 path->leave_spinning = 0;
Sage Weilc5c9cd42008-11-12 14:32:25 -05003576
Zheng Yan31840ae2008-09-23 13:14:14 -04003577 memcpy(&new_key, &key, sizeof(new_key));
Li Zefan33345d012011-04-20 10:31:50 +08003578 new_key.objectid = btrfs_ino(inode);
Li Zefan4d728ec2011-01-26 14:10:43 +08003579 if (off <= key.offset)
3580 new_key.offset = key.offset + destoff - off;
3581 else
3582 new_key.offset = destoff;
Sage Weilc5c9cd42008-11-12 14:32:25 -05003583
Sage Weilb6f34092011-09-20 14:48:51 -04003584 /*
Filipe Mananaf82a9902014-06-01 01:50:28 +01003585 * Deal with a hole that doesn't have an extent item
3586 * that represents it (NO_HOLES feature enabled).
3587 * This hole is either in the middle of the cloning
3588 * range or at the beginning (fully overlaps it or
3589 * partially overlaps it).
3590 */
3591 if (new_key.offset != last_dest_end)
3592 drop_start = last_dest_end;
3593 else
3594 drop_start = new_key.offset;
3595
3596 /*
Sage Weilb6f34092011-09-20 14:48:51 -04003597 * 1 - adjusting old extent (we may have to split it)
3598 * 1 - add new extent
3599 * 1 - inode update
3600 */
3601 trans = btrfs_start_transaction(root, 3);
Yan, Zhenga22285a2010-05-16 10:48:46 -04003602 if (IS_ERR(trans)) {
3603 ret = PTR_ERR(trans);
3604 goto out;
3605 }
3606
Chris Masonc8a894d2009-06-27 21:07:03 -04003607 if (type == BTRFS_FILE_EXTENT_REG ||
3608 type == BTRFS_FILE_EXTENT_PREALLOC) {
Li Zefand72c0842011-09-11 10:52:25 -04003609 /*
3610 * a | --- range to clone ---| b
3611 * | ------------- extent ------------- |
3612 */
3613
Antonio Ospite93915582014-06-04 14:03:48 +02003614 /* subtract range b */
Li Zefand72c0842011-09-11 10:52:25 -04003615 if (key.offset + datal > off + len)
3616 datal = off + len - key.offset;
3617
Antonio Ospite93915582014-06-04 14:03:48 +02003618 /* subtract range a */
Yan, Zhenga22285a2010-05-16 10:48:46 -04003619 if (off > key.offset) {
3620 datao += off - key.offset;
3621 datal -= off - key.offset;
3622 }
3623
Josef Bacik5dc562c2012-08-17 13:14:17 -04003624 ret = btrfs_drop_extents(trans, root, inode,
Filipe Mananaf82a9902014-06-01 01:50:28 +01003625 drop_start,
Yan, Zhenga22285a2010-05-16 10:48:46 -04003626 new_key.offset + datal,
Josef Bacik26714852012-08-29 12:24:27 -04003627 1);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003628 if (ret) {
David Sterba3f9e3df2014-04-15 18:50:17 +02003629 if (ret != -EOPNOTSUPP)
Liu Bo00fdf132014-03-10 18:56:07 +08003630 btrfs_abort_transaction(trans,
3631 root, ret);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003632 btrfs_end_transaction(trans, root);
3633 goto out;
3634 }
Yan, Zhenga22285a2010-05-16 10:48:46 -04003635
Sage Weilc5c9cd42008-11-12 14:32:25 -05003636 ret = btrfs_insert_empty_item(trans, root, path,
3637 &new_key, size);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003638 if (ret) {
3639 btrfs_abort_transaction(trans, root,
3640 ret);
3641 btrfs_end_transaction(trans, root);
3642 goto out;
3643 }
Sage Weilc5c9cd42008-11-12 14:32:25 -05003644
3645 leaf = path->nodes[0];
3646 slot = path->slots[0];
3647 write_extent_buffer(leaf, buf,
3648 btrfs_item_ptr_offset(leaf, slot),
3649 size);
3650
3651 extent = btrfs_item_ptr(leaf, slot,
3652 struct btrfs_file_extent_item);
Sage Weilc5c9cd42008-11-12 14:32:25 -05003653
Sage Weilc5c9cd42008-11-12 14:32:25 -05003654 /* disko == 0 means it's a hole */
3655 if (!disko)
3656 datao = 0;
Sage Weilc5c9cd42008-11-12 14:32:25 -05003657
3658 btrfs_set_file_extent_offset(leaf, extent,
3659 datao);
3660 btrfs_set_file_extent_num_bytes(leaf, extent,
3661 datal);
Josef Bacikfcebe452014-05-13 17:30:47 -07003662
Sage Weilc5c9cd42008-11-12 14:32:25 -05003663 if (disko) {
3664 inode_add_bytes(inode, datal);
3665 ret = btrfs_inc_extent_ref(trans, root,
Yan Zheng5d4f98a2009-06-10 10:45:14 -04003666 disko, diskl, 0,
3667 root->root_key.objectid,
Li Zefan33345d012011-04-20 10:31:50 +08003668 btrfs_ino(inode),
Filipe Mananab06c4bf2015-10-23 07:52:54 +01003669 new_key.offset - datao);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003670 if (ret) {
3671 btrfs_abort_transaction(trans,
3672 root,
3673 ret);
3674 btrfs_end_transaction(trans,
3675 root);
3676 goto out;
3677
3678 }
Sage Weilc5c9cd42008-11-12 14:32:25 -05003679 }
3680 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3681 u64 skip = 0;
3682 u64 trim = 0;
Filipe Mananaed958762015-07-14 16:09:39 +01003683
Sage Weilc5c9cd42008-11-12 14:32:25 -05003684 if (off > key.offset) {
3685 skip = off - key.offset;
3686 new_key.offset += skip;
3687 }
Chris Masond3977122009-01-05 21:25:51 -05003688
Liu Boaa42ffd2012-09-18 03:52:23 -06003689 if (key.offset + datal > off + len)
3690 trim = key.offset + datal - (off + len);
Chris Masond3977122009-01-05 21:25:51 -05003691
Sage Weilc5c9cd42008-11-12 14:32:25 -05003692 if (comp && (skip || trim)) {
Sage Weilc5c9cd42008-11-12 14:32:25 -05003693 ret = -EINVAL;
Yan, Zhenga22285a2010-05-16 10:48:46 -04003694 btrfs_end_transaction(trans, root);
Sage Weilc5c9cd42008-11-12 14:32:25 -05003695 goto out;
3696 }
3697 size -= skip + trim;
3698 datal -= skip + trim;
Yan, Zhenga22285a2010-05-16 10:48:46 -04003699
Filipe Manana8039d872015-10-13 15:15:00 +01003700 ret = clone_copy_inline_extent(src, inode,
3701 trans, path,
3702 &new_key,
3703 drop_start,
3704 datal,
3705 skip, size, buf);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003706 if (ret) {
David Sterba3f9e3df2014-04-15 18:50:17 +02003707 if (ret != -EOPNOTSUPP)
Chris Mason3a29bc02014-04-07 07:10:40 -07003708 btrfs_abort_transaction(trans,
Filipe Manana8039d872015-10-13 15:15:00 +01003709 root,
3710 ret);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003711 btrfs_end_transaction(trans, root);
3712 goto out;
3713 }
Sage Weilc5c9cd42008-11-12 14:32:25 -05003714 leaf = path->nodes[0];
3715 slot = path->slots[0];
Sage Weilc5c9cd42008-11-12 14:32:25 -05003716 }
3717
Filipe Manana7ffbb592014-06-09 03:48:05 +01003718 /* If we have an implicit hole (NO_HOLES feature). */
3719 if (drop_start < new_key.offset)
3720 clone_update_extent_map(inode, trans,
Filipe Manana14f59792014-06-29 21:45:40 +01003721 NULL, drop_start,
Filipe Manana7ffbb592014-06-09 03:48:05 +01003722 new_key.offset - drop_start);
3723
Filipe Manana14f59792014-06-29 21:45:40 +01003724 clone_update_extent_map(inode, trans, path, 0, 0);
Filipe Manana7ffbb592014-06-09 03:48:05 +01003725
Sage Weilc5c9cd42008-11-12 14:32:25 -05003726 btrfs_mark_buffer_dirty(leaf);
David Sterbab3b4aa72011-04-21 01:20:15 +02003727 btrfs_release_path(path);
Sage Weilc5c9cd42008-11-12 14:32:25 -05003728
Filipe Manana62e23902014-08-08 02:47:06 +01003729 last_dest_end = ALIGN(new_key.offset + datal,
3730 root->sectorsize);
Filipe Mananaf82a9902014-06-01 01:50:28 +01003731 ret = clone_finish_inode_update(trans, inode,
3732 last_dest_end,
Mark Fasheh1c919a52015-06-30 14:42:08 -07003733 destoff, olen,
3734 no_time_update);
Filipe Mananaf82a9902014-06-01 01:50:28 +01003735 if (ret)
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003736 goto out;
Filipe Manana2c463822014-05-31 02:31:05 +01003737 if (new_key.offset + datal >= destoff + len)
3738 break;
Yan, Zhenga22285a2010-05-16 10:48:46 -04003739 }
David Sterbab3b4aa72011-04-21 01:20:15 +02003740 btrfs_release_path(path);
Filipe Mananadf858e72015-03-31 14:56:46 +01003741 key.offset = next_key_min_offset;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003742 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003743 ret = 0;
Mark Fasheh32b7c682013-08-06 11:42:49 -07003744
Filipe Mananaf82a9902014-06-01 01:50:28 +01003745 if (last_dest_end < destoff + len) {
3746 /*
3747 * We have an implicit hole (NO_HOLES feature is enabled) that
3748 * fully or partially overlaps our cloning range at its end.
3749 */
3750 btrfs_release_path(path);
3751
3752 /*
3753 * 1 - remove extent(s)
3754 * 1 - inode update
3755 */
3756 trans = btrfs_start_transaction(root, 2);
3757 if (IS_ERR(trans)) {
3758 ret = PTR_ERR(trans);
3759 goto out;
3760 }
3761 ret = btrfs_drop_extents(trans, root, inode,
3762 last_dest_end, destoff + len, 1);
3763 if (ret) {
3764 if (ret != -EOPNOTSUPP)
3765 btrfs_abort_transaction(trans, root, ret);
3766 btrfs_end_transaction(trans, root);
3767 goto out;
3768 }
Filipe Manana14f59792014-06-29 21:45:40 +01003769 clone_update_extent_map(inode, trans, NULL, last_dest_end,
3770 destoff + len - last_dest_end);
Filipe Mananaf82a9902014-06-01 01:50:28 +01003771 ret = clone_finish_inode_update(trans, inode, destoff + len,
Mark Fasheh1c919a52015-06-30 14:42:08 -07003772 destoff, olen, no_time_update);
Filipe Mananaf82a9902014-06-01 01:50:28 +01003773 }
3774
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003775out:
Mark Fasheh32b7c682013-08-06 11:42:49 -07003776 btrfs_free_path(path);
3777 vfree(buf);
3778 return ret;
3779}
3780
Zach Brown3db11b22015-11-10 16:53:32 -05003781static noinline int btrfs_clone_files(struct file *file, struct file *file_src,
3782 u64 off, u64 olen, u64 destoff)
Mark Fasheh32b7c682013-08-06 11:42:49 -07003783{
Al Viro54563d42013-09-01 15:57:51 -04003784 struct inode *inode = file_inode(file);
Zach Brown3db11b22015-11-10 16:53:32 -05003785 struct inode *src = file_inode(file_src);
Mark Fasheh32b7c682013-08-06 11:42:49 -07003786 struct btrfs_root *root = BTRFS_I(inode)->root;
Mark Fasheh32b7c682013-08-06 11:42:49 -07003787 int ret;
3788 u64 len = olen;
3789 u64 bs = root->fs_info->sb->s_blocksize;
Zach Brown3db11b22015-11-10 16:53:32 -05003790 int same_inode = src == inode;
Mark Fasheh32b7c682013-08-06 11:42:49 -07003791
3792 /*
3793 * TODO:
3794 * - split compressed inline extents. annoying: we need to
3795 * decompress into destination's address_space (the file offset
3796 * may change, so source mapping won't do), then recompress (or
3797 * otherwise reinsert) a subrange.
Liu Bo00fdf132014-03-10 18:56:07 +08003798 *
3799 * - split destination inode's inline extents. The inline extents can
3800 * be either compressed or non-compressed.
Mark Fasheh32b7c682013-08-06 11:42:49 -07003801 */
3802
Mark Fasheh32b7c682013-08-06 11:42:49 -07003803 if (btrfs_root_readonly(root))
3804 return -EROFS;
3805
Zach Brown3db11b22015-11-10 16:53:32 -05003806 if (file_src->f_path.mnt != file->f_path.mnt ||
3807 src->i_sb != inode->i_sb)
3808 return -EXDEV;
Mark Fasheh32b7c682013-08-06 11:42:49 -07003809
3810 /* don't make the dst file partly checksummed */
3811 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3812 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
Zach Brown3db11b22015-11-10 16:53:32 -05003813 return -EINVAL;
Mark Fasheh32b7c682013-08-06 11:42:49 -07003814
Mark Fasheh32b7c682013-08-06 11:42:49 -07003815 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
Zach Brown3db11b22015-11-10 16:53:32 -05003816 return -EISDIR;
Mark Fasheh32b7c682013-08-06 11:42:49 -07003817
3818 if (!same_inode) {
Mark Fasheh293a8482015-06-30 14:42:06 -07003819 btrfs_double_inode_lock(src, inode);
Mark Fasheh32b7c682013-08-06 11:42:49 -07003820 } else {
Al Viro59551022016-01-22 15:40:57 -05003821 inode_lock(src);
Mark Fasheh32b7c682013-08-06 11:42:49 -07003822 }
3823
3824 /* determine range to clone */
3825 ret = -EINVAL;
3826 if (off + len > src->i_size || off + len < off)
3827 goto out_unlock;
3828 if (len == 0)
3829 olen = len = src->i_size - off;
3830 /* if we extend to eof, continue to block boundary */
3831 if (off + len == src->i_size)
3832 len = ALIGN(src->i_size, bs) - off;
3833
Filipe Mananaccccf3d62015-03-30 18:23:59 +01003834 if (len == 0) {
3835 ret = 0;
3836 goto out_unlock;
3837 }
3838
Mark Fasheh32b7c682013-08-06 11:42:49 -07003839 /* verify the end result is block aligned */
3840 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3841 !IS_ALIGNED(destoff, bs))
3842 goto out_unlock;
3843
3844 /* verify if ranges are overlapped within the same file */
3845 if (same_inode) {
3846 if (destoff + len > off && destoff < off + len)
3847 goto out_unlock;
3848 }
3849
3850 if (destoff > inode->i_size) {
3851 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3852 if (ret)
3853 goto out_unlock;
3854 }
3855
Filipe Mananac125b8b2014-05-23 05:03:34 +01003856 /*
3857 * Lock the target range too. Right after we replace the file extent
3858 * items in the fs tree (which now point to the cloned data), we might
3859 * have a worker replace them with extent items relative to a write
3860 * operation that was issued before this clone operation (i.e. confront
3861 * with inode.c:btrfs_finish_ordered_io).
3862 */
3863 if (same_inode) {
3864 u64 lock_start = min_t(u64, off, destoff);
3865 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
Mark Fasheh32b7c682013-08-06 11:42:49 -07003866
Filipe Mananae0bd70c2016-01-27 10:20:58 +00003867 ret = lock_extent_range(src, lock_start, lock_len, true);
Filipe Mananac125b8b2014-05-23 05:03:34 +01003868 } else {
Filipe Mananae0bd70c2016-01-27 10:20:58 +00003869 ret = btrfs_double_extent_lock(src, off, inode, destoff, len,
3870 true);
3871 }
3872 ASSERT(ret == 0);
3873 if (WARN_ON(ret)) {
3874 /* ranges in the io trees already unlocked */
3875 goto out_unlock;
Filipe Mananac125b8b2014-05-23 05:03:34 +01003876 }
Mark Fasheh32b7c682013-08-06 11:42:49 -07003877
Mark Fasheh1c919a52015-06-30 14:42:08 -07003878 ret = btrfs_clone(src, inode, off, olen, len, destoff, 0);
Mark Fasheh32b7c682013-08-06 11:42:49 -07003879
Filipe Mananac125b8b2014-05-23 05:03:34 +01003880 if (same_inode) {
3881 u64 lock_start = min_t(u64, off, destoff);
3882 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3883
3884 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3885 } else {
Mark Fasheh293a8482015-06-30 14:42:06 -07003886 btrfs_double_extent_unlock(src, off, inode, destoff, len);
Filipe Mananac125b8b2014-05-23 05:03:34 +01003887 }
3888 /*
3889 * Truncate page cache pages so that future reads will see the cloned
3890 * data immediately and not the previous data.
3891 */
3892 truncate_inode_pages_range(&inode->i_data, destoff,
3893 PAGE_CACHE_ALIGN(destoff + len) - 1);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003894out_unlock:
Mark Fasheh293a8482015-06-30 14:42:06 -07003895 if (!same_inode)
3896 btrfs_double_inode_unlock(src, inode);
3897 else
Al Viro59551022016-01-22 15:40:57 -05003898 inode_unlock(src);
Zach Brown3db11b22015-11-10 16:53:32 -05003899 return ret;
3900}
3901
3902ssize_t btrfs_copy_file_range(struct file *file_in, loff_t pos_in,
3903 struct file *file_out, loff_t pos_out,
3904 size_t len, unsigned int flags)
3905{
3906 ssize_t ret;
3907
3908 ret = btrfs_clone_files(file_out, file_in, pos_in, len, pos_out);
3909 if (ret == 0)
3910 ret = len;
3911 return ret;
3912}
3913
Christoph Hellwig04b38d62015-12-03 12:59:50 +01003914int btrfs_clone_file_range(struct file *src_file, loff_t off,
3915 struct file *dst_file, loff_t destoff, u64 len)
Zach Brown3db11b22015-11-10 16:53:32 -05003916{
Christoph Hellwig04b38d62015-12-03 12:59:50 +01003917 return btrfs_clone_files(dst_file, src_file, off, len, destoff);
Sage Weilc5c9cd42008-11-12 14:32:25 -05003918}
3919
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003920/*
3921 * there are many ways the trans_start and trans_end ioctls can lead
3922 * to deadlocks. They should only be used by applications that
3923 * basically own the machine, and have a very in depth understanding
3924 * of all the possible deadlocks and enospc problems.
3925 */
Christoph Hellwigb2950862008-12-02 09:54:17 -05003926static long btrfs_ioctl_trans_start(struct file *file)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003927{
Al Viro496ad9a2013-01-23 17:07:38 -05003928 struct inode *inode = file_inode(file);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003929 struct btrfs_root *root = BTRFS_I(inode)->root;
3930 struct btrfs_trans_handle *trans;
Sage Weil1ab86ae2009-09-29 18:38:44 -04003931 int ret;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003932
Sage Weil1ab86ae2009-09-29 18:38:44 -04003933 ret = -EPERM;
Christoph Hellwigdf5b5522008-06-11 21:53:58 -04003934 if (!capable(CAP_SYS_ADMIN))
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003935 goto out;
Sage Weil1ab86ae2009-09-29 18:38:44 -04003936
3937 ret = -EINPROGRESS;
3938 if (file->private_data)
3939 goto out;
Sage Weil9ca9ee02008-08-04 10:41:27 -04003940
Li Zefanb83cc962010-12-20 16:04:08 +08003941 ret = -EROFS;
3942 if (btrfs_root_readonly(root))
3943 goto out;
3944
Al Viroa561be72011-11-23 11:57:51 -05003945 ret = mnt_want_write_file(file);
Yan Zhengc146afa2008-11-12 14:34:12 -05003946 if (ret)
3947 goto out;
3948
Josef Bacika4abeea2011-04-11 17:25:13 -04003949 atomic_inc(&root->fs_info->open_ioctl_trans);
Sage Weil9ca9ee02008-08-04 10:41:27 -04003950
Sage Weil1ab86ae2009-09-29 18:38:44 -04003951 ret = -ENOMEM;
Josef Bacik7a7eaa42011-04-13 12:54:33 -04003952 trans = btrfs_start_ioctl_transaction(root);
Tsutomu Itohabd30bb2011-01-24 00:57:10 +00003953 if (IS_ERR(trans))
Sage Weil1ab86ae2009-09-29 18:38:44 -04003954 goto out_drop;
3955
3956 file->private_data = trans;
3957 return 0;
3958
3959out_drop:
Josef Bacika4abeea2011-04-11 17:25:13 -04003960 atomic_dec(&root->fs_info->open_ioctl_trans);
Al Viro2a79f172011-12-09 08:06:57 -05003961 mnt_drop_write_file(file);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003962out:
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003963 return ret;
3964}
3965
Josef Bacik6ef5ed02009-12-11 21:11:29 +00003966static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3967{
Al Viro496ad9a2013-01-23 17:07:38 -05003968 struct inode *inode = file_inode(file);
Josef Bacik6ef5ed02009-12-11 21:11:29 +00003969 struct btrfs_root *root = BTRFS_I(inode)->root;
3970 struct btrfs_root *new_root;
3971 struct btrfs_dir_item *di;
3972 struct btrfs_trans_handle *trans;
3973 struct btrfs_path *path;
3974 struct btrfs_key location;
3975 struct btrfs_disk_key disk_key;
Josef Bacik6ef5ed02009-12-11 21:11:29 +00003976 u64 objectid = 0;
3977 u64 dir_id;
Miao Xie3c04ce02012-11-26 08:43:07 +00003978 int ret;
Josef Bacik6ef5ed02009-12-11 21:11:29 +00003979
3980 if (!capable(CAP_SYS_ADMIN))
3981 return -EPERM;
3982
Miao Xie3c04ce02012-11-26 08:43:07 +00003983 ret = mnt_want_write_file(file);
3984 if (ret)
3985 return ret;
3986
3987 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3988 ret = -EFAULT;
3989 goto out;
3990 }
Josef Bacik6ef5ed02009-12-11 21:11:29 +00003991
3992 if (!objectid)
chandan1cecf572013-09-13 19:34:10 +05303993 objectid = BTRFS_FS_TREE_OBJECTID;
Josef Bacik6ef5ed02009-12-11 21:11:29 +00003994
3995 location.objectid = objectid;
3996 location.type = BTRFS_ROOT_ITEM_KEY;
3997 location.offset = (u64)-1;
3998
3999 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
Miao Xie3c04ce02012-11-26 08:43:07 +00004000 if (IS_ERR(new_root)) {
4001 ret = PTR_ERR(new_root);
4002 goto out;
4003 }
Josef Bacik6ef5ed02009-12-11 21:11:29 +00004004
Josef Bacik6ef5ed02009-12-11 21:11:29 +00004005 path = btrfs_alloc_path();
Miao Xie3c04ce02012-11-26 08:43:07 +00004006 if (!path) {
4007 ret = -ENOMEM;
4008 goto out;
4009 }
Josef Bacik6ef5ed02009-12-11 21:11:29 +00004010 path->leave_spinning = 1;
4011
4012 trans = btrfs_start_transaction(root, 1);
Tsutomu Itoh98d5dc12011-01-20 06:19:37 +00004013 if (IS_ERR(trans)) {
Josef Bacik6ef5ed02009-12-11 21:11:29 +00004014 btrfs_free_path(path);
Miao Xie3c04ce02012-11-26 08:43:07 +00004015 ret = PTR_ERR(trans);
4016 goto out;
Josef Bacik6ef5ed02009-12-11 21:11:29 +00004017 }
4018
David Sterba6c417612011-04-13 15:41:04 +02004019 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
Josef Bacik6ef5ed02009-12-11 21:11:29 +00004020 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
4021 dir_id, "default", 7, 1);
Dan Carpentercf1e99a2010-05-29 09:47:24 +00004022 if (IS_ERR_OR_NULL(di)) {
Josef Bacik6ef5ed02009-12-11 21:11:29 +00004023 btrfs_free_path(path);
4024 btrfs_end_transaction(trans, root);
Frank Holtonefe120a2013-12-20 11:37:06 -05004025 btrfs_err(new_root->fs_info, "Umm, you don't have the default dir"
4026 "item, this isn't going to work");
Miao Xie3c04ce02012-11-26 08:43:07 +00004027 ret = -ENOENT;
4028 goto out;
Josef Bacik6ef5ed02009-12-11 21:11:29 +00004029 }
4030
4031 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
4032 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
4033 btrfs_mark_buffer_dirty(path->nodes[0]);
4034 btrfs_free_path(path);
4035
Mitch Harder2b0ce2c2012-07-24 11:58:43 -06004036 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
Josef Bacik6ef5ed02009-12-11 21:11:29 +00004037 btrfs_end_transaction(trans, root);
Miao Xie3c04ce02012-11-26 08:43:07 +00004038out:
4039 mnt_drop_write_file(file);
4040 return ret;
Josef Bacik6ef5ed02009-12-11 21:11:29 +00004041}
4042
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02004043void btrfs_get_block_group_info(struct list_head *groups_list,
4044 struct btrfs_ioctl_space_info *space)
Josef Bacikbf5fc0932010-09-29 11:22:36 -04004045{
4046 struct btrfs_block_group_cache *block_group;
4047
4048 space->total_bytes = 0;
4049 space->used_bytes = 0;
4050 space->flags = 0;
4051 list_for_each_entry(block_group, groups_list, list) {
4052 space->flags = block_group->flags;
4053 space->total_bytes += block_group->key.offset;
4054 space->used_bytes +=
4055 btrfs_block_group_used(&block_group->item);
4056 }
4057}
4058
Eric Sandeen48a3b632013-04-25 20:41:01 +00004059static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
Josef Bacik1406e432010-01-13 18:19:06 +00004060{
4061 struct btrfs_ioctl_space_args space_args;
4062 struct btrfs_ioctl_space_info space;
4063 struct btrfs_ioctl_space_info *dest;
Chris Mason7fde62b2010-03-16 15:40:10 -04004064 struct btrfs_ioctl_space_info *dest_orig;
Daniel J Blueman13f26962011-04-11 15:56:31 +00004065 struct btrfs_ioctl_space_info __user *user_dest;
Josef Bacik1406e432010-01-13 18:19:06 +00004066 struct btrfs_space_info *info;
Josef Bacikbf5fc0932010-09-29 11:22:36 -04004067 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
4068 BTRFS_BLOCK_GROUP_SYSTEM,
4069 BTRFS_BLOCK_GROUP_METADATA,
4070 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
4071 int num_types = 4;
Chris Mason7fde62b2010-03-16 15:40:10 -04004072 int alloc_size;
Josef Bacik1406e432010-01-13 18:19:06 +00004073 int ret = 0;
Dan Rosenberg51788b12011-02-14 16:04:23 -05004074 u64 slot_count = 0;
Josef Bacikbf5fc0932010-09-29 11:22:36 -04004075 int i, c;
Josef Bacik1406e432010-01-13 18:19:06 +00004076
4077 if (copy_from_user(&space_args,
4078 (struct btrfs_ioctl_space_args __user *)arg,
4079 sizeof(space_args)))
4080 return -EFAULT;
4081
Josef Bacikbf5fc0932010-09-29 11:22:36 -04004082 for (i = 0; i < num_types; i++) {
4083 struct btrfs_space_info *tmp;
4084
4085 info = NULL;
4086 rcu_read_lock();
4087 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
4088 list) {
4089 if (tmp->flags == types[i]) {
4090 info = tmp;
4091 break;
4092 }
4093 }
4094 rcu_read_unlock();
4095
4096 if (!info)
4097 continue;
4098
4099 down_read(&info->groups_sem);
4100 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4101 if (!list_empty(&info->block_groups[c]))
4102 slot_count++;
4103 }
4104 up_read(&info->groups_sem);
4105 }
Josef Bacik1406e432010-01-13 18:19:06 +00004106
David Sterba36523e952014-02-07 14:34:12 +01004107 /*
4108 * Global block reserve, exported as a space_info
4109 */
4110 slot_count++;
4111
Chris Mason7fde62b2010-03-16 15:40:10 -04004112 /* space_slots == 0 means they are asking for a count */
4113 if (space_args.space_slots == 0) {
4114 space_args.total_spaces = slot_count;
4115 goto out;
4116 }
Josef Bacikbf5fc0932010-09-29 11:22:36 -04004117
Dan Rosenberg51788b12011-02-14 16:04:23 -05004118 slot_count = min_t(u64, space_args.space_slots, slot_count);
Josef Bacikbf5fc0932010-09-29 11:22:36 -04004119
Chris Mason7fde62b2010-03-16 15:40:10 -04004120 alloc_size = sizeof(*dest) * slot_count;
Josef Bacikbf5fc0932010-09-29 11:22:36 -04004121
Chris Mason7fde62b2010-03-16 15:40:10 -04004122 /* we generally have at most 6 or so space infos, one for each raid
4123 * level. So, a whole page should be more than enough for everyone
4124 */
4125 if (alloc_size > PAGE_CACHE_SIZE)
4126 return -ENOMEM;
4127
4128 space_args.total_spaces = 0;
David Sterba8d2db782015-11-04 15:38:29 +01004129 dest = kmalloc(alloc_size, GFP_KERNEL);
Chris Mason7fde62b2010-03-16 15:40:10 -04004130 if (!dest)
4131 return -ENOMEM;
4132 dest_orig = dest;
4133
4134 /* now we have a buffer to copy into */
Josef Bacikbf5fc0932010-09-29 11:22:36 -04004135 for (i = 0; i < num_types; i++) {
4136 struct btrfs_space_info *tmp;
Chris Mason7fde62b2010-03-16 15:40:10 -04004137
Dan Rosenberg51788b12011-02-14 16:04:23 -05004138 if (!slot_count)
4139 break;
4140
Josef Bacikbf5fc0932010-09-29 11:22:36 -04004141 info = NULL;
4142 rcu_read_lock();
4143 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
4144 list) {
4145 if (tmp->flags == types[i]) {
4146 info = tmp;
4147 break;
4148 }
4149 }
4150 rcu_read_unlock();
Chris Mason7fde62b2010-03-16 15:40:10 -04004151
Josef Bacikbf5fc0932010-09-29 11:22:36 -04004152 if (!info)
4153 continue;
4154 down_read(&info->groups_sem);
4155 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4156 if (!list_empty(&info->block_groups[c])) {
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02004157 btrfs_get_block_group_info(
4158 &info->block_groups[c], &space);
Josef Bacikbf5fc0932010-09-29 11:22:36 -04004159 memcpy(dest, &space, sizeof(space));
4160 dest++;
4161 space_args.total_spaces++;
Dan Rosenberg51788b12011-02-14 16:04:23 -05004162 slot_count--;
Josef Bacikbf5fc0932010-09-29 11:22:36 -04004163 }
Dan Rosenberg51788b12011-02-14 16:04:23 -05004164 if (!slot_count)
4165 break;
Josef Bacikbf5fc0932010-09-29 11:22:36 -04004166 }
4167 up_read(&info->groups_sem);
Josef Bacik1406e432010-01-13 18:19:06 +00004168 }
Josef Bacik1406e432010-01-13 18:19:06 +00004169
David Sterba36523e952014-02-07 14:34:12 +01004170 /*
4171 * Add global block reserve
4172 */
4173 if (slot_count) {
4174 struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv;
4175
4176 spin_lock(&block_rsv->lock);
4177 space.total_bytes = block_rsv->size;
4178 space.used_bytes = block_rsv->size - block_rsv->reserved;
4179 spin_unlock(&block_rsv->lock);
4180 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
4181 memcpy(dest, &space, sizeof(space));
4182 space_args.total_spaces++;
4183 }
4184
Daniel J Blueman2eec6c82012-04-26 00:37:14 +08004185 user_dest = (struct btrfs_ioctl_space_info __user *)
Chris Mason7fde62b2010-03-16 15:40:10 -04004186 (arg + sizeof(struct btrfs_ioctl_space_args));
4187
4188 if (copy_to_user(user_dest, dest_orig, alloc_size))
4189 ret = -EFAULT;
4190
4191 kfree(dest_orig);
4192out:
4193 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
Josef Bacik1406e432010-01-13 18:19:06 +00004194 ret = -EFAULT;
4195
4196 return ret;
4197}
4198
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04004199/*
4200 * there are many ways the trans_start and trans_end ioctls can lead
4201 * to deadlocks. They should only be used by applications that
4202 * basically own the machine, and have a very in depth understanding
4203 * of all the possible deadlocks and enospc problems.
4204 */
4205long btrfs_ioctl_trans_end(struct file *file)
4206{
Al Viro496ad9a2013-01-23 17:07:38 -05004207 struct inode *inode = file_inode(file);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04004208 struct btrfs_root *root = BTRFS_I(inode)->root;
4209 struct btrfs_trans_handle *trans;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04004210
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04004211 trans = file->private_data;
Sage Weil1ab86ae2009-09-29 18:38:44 -04004212 if (!trans)
4213 return -EINVAL;
Christoph Hellwigb2141072008-09-05 16:43:31 -04004214 file->private_data = NULL;
Sage Weil9ca9ee02008-08-04 10:41:27 -04004215
Sage Weil1ab86ae2009-09-29 18:38:44 -04004216 btrfs_end_transaction(trans, root);
4217
Josef Bacika4abeea2011-04-11 17:25:13 -04004218 atomic_dec(&root->fs_info->open_ioctl_trans);
Sage Weil9ca9ee02008-08-04 10:41:27 -04004219
Al Viro2a79f172011-12-09 08:06:57 -05004220 mnt_drop_write_file(file);
Sage Weil1ab86ae2009-09-29 18:38:44 -04004221 return 0;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04004222}
4223
Miao Xie9a8c28b2012-11-26 08:40:43 +00004224static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4225 void __user *argp)
Sage Weil46204592010-10-29 15:41:32 -04004226{
Sage Weil46204592010-10-29 15:41:32 -04004227 struct btrfs_trans_handle *trans;
4228 u64 transid;
Tsutomu Itohdb5b4932011-03-23 08:14:16 +00004229 int ret;
Sage Weil46204592010-10-29 15:41:32 -04004230
Miao Xied4edf392013-02-20 09:17:06 +00004231 trans = btrfs_attach_transaction_barrier(root);
Miao Xieff7c1d32012-11-26 08:41:29 +00004232 if (IS_ERR(trans)) {
4233 if (PTR_ERR(trans) != -ENOENT)
4234 return PTR_ERR(trans);
4235
4236 /* No running transaction, don't bother */
4237 transid = root->fs_info->last_trans_committed;
4238 goto out;
4239 }
Sage Weil46204592010-10-29 15:41:32 -04004240 transid = trans->transid;
Tsutomu Itohdb5b4932011-03-23 08:14:16 +00004241 ret = btrfs_commit_transaction_async(trans, root, 0);
Tsutomu Itoh8b2b2d32011-04-04 01:52:13 +00004242 if (ret) {
4243 btrfs_end_transaction(trans, root);
Tsutomu Itohdb5b4932011-03-23 08:14:16 +00004244 return ret;
Tsutomu Itoh8b2b2d32011-04-04 01:52:13 +00004245 }
Miao Xieff7c1d32012-11-26 08:41:29 +00004246out:
Sage Weil46204592010-10-29 15:41:32 -04004247 if (argp)
4248 if (copy_to_user(argp, &transid, sizeof(transid)))
4249 return -EFAULT;
4250 return 0;
4251}
4252
Miao Xie9a8c28b2012-11-26 08:40:43 +00004253static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
4254 void __user *argp)
Sage Weil46204592010-10-29 15:41:32 -04004255{
Sage Weil46204592010-10-29 15:41:32 -04004256 u64 transid;
4257
4258 if (argp) {
4259 if (copy_from_user(&transid, argp, sizeof(transid)))
4260 return -EFAULT;
4261 } else {
4262 transid = 0; /* current trans */
4263 }
4264 return btrfs_wait_for_commit(root, transid);
4265}
4266
Miao Xieb8e95482012-11-26 08:48:01 +00004267static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
Jan Schmidt475f6382011-03-11 15:41:01 +01004268{
Al Viro496ad9a2013-01-23 17:07:38 -05004269 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Jan Schmidt475f6382011-03-11 15:41:01 +01004270 struct btrfs_ioctl_scrub_args *sa;
Miao Xieb8e95482012-11-26 08:48:01 +00004271 int ret;
Jan Schmidt475f6382011-03-11 15:41:01 +01004272
4273 if (!capable(CAP_SYS_ADMIN))
4274 return -EPERM;
4275
4276 sa = memdup_user(arg, sizeof(*sa));
4277 if (IS_ERR(sa))
4278 return PTR_ERR(sa);
4279
Miao Xieb8e95482012-11-26 08:48:01 +00004280 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4281 ret = mnt_want_write_file(file);
4282 if (ret)
4283 goto out;
4284 }
4285
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01004286 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
Stefan Behrens63a212a2012-11-05 18:29:28 +01004287 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4288 0);
Jan Schmidt475f6382011-03-11 15:41:01 +01004289
4290 if (copy_to_user(arg, sa, sizeof(*sa)))
4291 ret = -EFAULT;
4292
Miao Xieb8e95482012-11-26 08:48:01 +00004293 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4294 mnt_drop_write_file(file);
4295out:
Jan Schmidt475f6382011-03-11 15:41:01 +01004296 kfree(sa);
4297 return ret;
4298}
4299
4300static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
4301{
4302 if (!capable(CAP_SYS_ADMIN))
4303 return -EPERM;
4304
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01004305 return btrfs_scrub_cancel(root->fs_info);
Jan Schmidt475f6382011-03-11 15:41:01 +01004306}
4307
4308static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
4309 void __user *arg)
4310{
4311 struct btrfs_ioctl_scrub_args *sa;
4312 int ret;
4313
4314 if (!capable(CAP_SYS_ADMIN))
4315 return -EPERM;
4316
4317 sa = memdup_user(arg, sizeof(*sa));
4318 if (IS_ERR(sa))
4319 return PTR_ERR(sa);
4320
4321 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
4322
4323 if (copy_to_user(arg, sa, sizeof(*sa)))
4324 ret = -EFAULT;
4325
4326 kfree(sa);
4327 return ret;
4328}
4329
Stefan Behrensc11d2c22012-05-25 16:06:09 +02004330static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
David Sterbab27f7c02012-06-22 06:30:39 -06004331 void __user *arg)
Stefan Behrensc11d2c22012-05-25 16:06:09 +02004332{
4333 struct btrfs_ioctl_get_dev_stats *sa;
4334 int ret;
4335
Stefan Behrensc11d2c22012-05-25 16:06:09 +02004336 sa = memdup_user(arg, sizeof(*sa));
4337 if (IS_ERR(sa))
4338 return PTR_ERR(sa);
4339
David Sterbab27f7c02012-06-22 06:30:39 -06004340 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4341 kfree(sa);
4342 return -EPERM;
4343 }
4344
4345 ret = btrfs_get_dev_stats(root, sa);
Stefan Behrensc11d2c22012-05-25 16:06:09 +02004346
4347 if (copy_to_user(arg, sa, sizeof(*sa)))
4348 ret = -EFAULT;
4349
4350 kfree(sa);
4351 return ret;
4352}
4353
Stefan Behrens3f6bcfb2012-11-06 15:08:53 +01004354static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
4355{
4356 struct btrfs_ioctl_dev_replace_args *p;
4357 int ret;
4358
4359 if (!capable(CAP_SYS_ADMIN))
4360 return -EPERM;
4361
4362 p = memdup_user(arg, sizeof(*p));
4363 if (IS_ERR(p))
4364 return PTR_ERR(p);
4365
4366 switch (p->cmd) {
4367 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
Ilya Dryomovadfa97c2013-10-10 20:39:28 +03004368 if (root->fs_info->sb->s_flags & MS_RDONLY) {
4369 ret = -EROFS;
4370 goto out;
4371 }
Stefan Behrens3f6bcfb2012-11-06 15:08:53 +01004372 if (atomic_xchg(
4373 &root->fs_info->mutually_exclusive_operation_running,
4374 1)) {
Anand Jaine57138b2013-08-21 11:44:48 +08004375 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
Stefan Behrens3f6bcfb2012-11-06 15:08:53 +01004376 } else {
4377 ret = btrfs_dev_replace_start(root, p);
4378 atomic_set(
4379 &root->fs_info->mutually_exclusive_operation_running,
4380 0);
4381 }
4382 break;
4383 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4384 btrfs_dev_replace_status(root->fs_info, p);
4385 ret = 0;
4386 break;
4387 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4388 ret = btrfs_dev_replace_cancel(root->fs_info, p);
4389 break;
4390 default:
4391 ret = -EINVAL;
4392 break;
4393 }
4394
4395 if (copy_to_user(arg, p, sizeof(*p)))
4396 ret = -EFAULT;
Ilya Dryomovadfa97c2013-10-10 20:39:28 +03004397out:
Stefan Behrens3f6bcfb2012-11-06 15:08:53 +01004398 kfree(p);
4399 return ret;
4400}
4401
Jan Schmidtd7728c92011-07-07 16:48:38 +02004402static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4403{
4404 int ret = 0;
4405 int i;
Chris Mason740c3d22011-11-02 15:48:34 -04004406 u64 rel_ptr;
Jan Schmidtd7728c92011-07-07 16:48:38 +02004407 int size;
Chris Mason806468f2011-11-06 03:07:10 -05004408 struct btrfs_ioctl_ino_path_args *ipa = NULL;
Jan Schmidtd7728c92011-07-07 16:48:38 +02004409 struct inode_fs_paths *ipath = NULL;
4410 struct btrfs_path *path;
4411
Kusanagi Kouichi82b22ac2013-01-28 11:33:31 +00004412 if (!capable(CAP_DAC_READ_SEARCH))
Jan Schmidtd7728c92011-07-07 16:48:38 +02004413 return -EPERM;
4414
4415 path = btrfs_alloc_path();
4416 if (!path) {
4417 ret = -ENOMEM;
4418 goto out;
4419 }
4420
4421 ipa = memdup_user(arg, sizeof(*ipa));
4422 if (IS_ERR(ipa)) {
4423 ret = PTR_ERR(ipa);
4424 ipa = NULL;
4425 goto out;
4426 }
4427
4428 size = min_t(u32, ipa->size, 4096);
4429 ipath = init_ipath(size, root, path);
4430 if (IS_ERR(ipath)) {
4431 ret = PTR_ERR(ipath);
4432 ipath = NULL;
4433 goto out;
4434 }
4435
4436 ret = paths_from_inode(ipa->inum, ipath);
4437 if (ret < 0)
4438 goto out;
4439
4440 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
Jeff Mahoney745c4d82011-11-20 07:31:57 -05004441 rel_ptr = ipath->fspath->val[i] -
4442 (u64)(unsigned long)ipath->fspath->val;
Chris Mason740c3d22011-11-02 15:48:34 -04004443 ipath->fspath->val[i] = rel_ptr;
Jan Schmidtd7728c92011-07-07 16:48:38 +02004444 }
4445
Jeff Mahoney745c4d82011-11-20 07:31:57 -05004446 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
4447 (void *)(unsigned long)ipath->fspath, size);
Jan Schmidtd7728c92011-07-07 16:48:38 +02004448 if (ret) {
4449 ret = -EFAULT;
4450 goto out;
4451 }
4452
4453out:
4454 btrfs_free_path(path);
4455 free_ipath(ipath);
4456 kfree(ipa);
4457
4458 return ret;
4459}
4460
4461static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4462{
4463 struct btrfs_data_container *inodes = ctx;
4464 const size_t c = 3 * sizeof(u64);
4465
4466 if (inodes->bytes_left >= c) {
4467 inodes->bytes_left -= c;
4468 inodes->val[inodes->elem_cnt] = inum;
4469 inodes->val[inodes->elem_cnt + 1] = offset;
4470 inodes->val[inodes->elem_cnt + 2] = root;
4471 inodes->elem_cnt += 3;
4472 } else {
4473 inodes->bytes_missing += c - inodes->bytes_left;
4474 inodes->bytes_left = 0;
4475 inodes->elem_missed += 3;
4476 }
4477
4478 return 0;
4479}
4480
4481static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
4482 void __user *arg)
4483{
4484 int ret = 0;
4485 int size;
Jan Schmidtd7728c92011-07-07 16:48:38 +02004486 struct btrfs_ioctl_logical_ino_args *loi;
4487 struct btrfs_data_container *inodes = NULL;
4488 struct btrfs_path *path = NULL;
Jan Schmidtd7728c92011-07-07 16:48:38 +02004489
4490 if (!capable(CAP_SYS_ADMIN))
4491 return -EPERM;
4492
4493 loi = memdup_user(arg, sizeof(*loi));
4494 if (IS_ERR(loi)) {
4495 ret = PTR_ERR(loi);
4496 loi = NULL;
4497 goto out;
4498 }
4499
4500 path = btrfs_alloc_path();
4501 if (!path) {
4502 ret = -ENOMEM;
4503 goto out;
4504 }
4505
Byongho Leeee221842015-12-15 01:42:10 +09004506 size = min_t(u32, loi->size, SZ_64K);
Jan Schmidtd7728c92011-07-07 16:48:38 +02004507 inodes = init_data_container(size);
4508 if (IS_ERR(inodes)) {
4509 ret = PTR_ERR(inodes);
4510 inodes = NULL;
4511 goto out;
4512 }
4513
Liu Bodf031f02012-09-07 20:01:29 -06004514 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
4515 build_ino_list, inodes);
4516 if (ret == -EINVAL)
Jan Schmidtd7728c92011-07-07 16:48:38 +02004517 ret = -ENOENT;
4518 if (ret < 0)
4519 goto out;
4520
Jeff Mahoney745c4d82011-11-20 07:31:57 -05004521 ret = copy_to_user((void *)(unsigned long)loi->inodes,
4522 (void *)(unsigned long)inodes, size);
Jan Schmidtd7728c92011-07-07 16:48:38 +02004523 if (ret)
4524 ret = -EFAULT;
4525
4526out:
4527 btrfs_free_path(path);
Liu Bo425d17a2012-09-07 20:01:30 -06004528 vfree(inodes);
Jan Schmidtd7728c92011-07-07 16:48:38 +02004529 kfree(loi);
4530
4531 return ret;
4532}
4533
Ilya Dryomov19a39dc2012-01-16 22:04:49 +02004534void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004535 struct btrfs_ioctl_balance_args *bargs)
4536{
4537 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4538
4539 bargs->flags = bctl->flags;
4540
Ilya Dryomov837d5b62012-01-16 22:04:49 +02004541 if (atomic_read(&fs_info->balance_running))
4542 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4543 if (atomic_read(&fs_info->balance_pause_req))
4544 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
Ilya Dryomova7e99c62012-01-16 22:04:49 +02004545 if (atomic_read(&fs_info->balance_cancel_req))
4546 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
Ilya Dryomov837d5b62012-01-16 22:04:49 +02004547
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004548 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4549 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4550 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
Ilya Dryomov19a39dc2012-01-16 22:04:49 +02004551
4552 if (lock) {
4553 spin_lock(&fs_info->balance_lock);
4554 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4555 spin_unlock(&fs_info->balance_lock);
4556 } else {
4557 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4558 }
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004559}
4560
Liu Bo9ba1f6e2012-05-11 18:11:26 +08004561static long btrfs_ioctl_balance(struct file *file, void __user *arg)
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004562{
Al Viro496ad9a2013-01-23 17:07:38 -05004563 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004564 struct btrfs_fs_info *fs_info = root->fs_info;
4565 struct btrfs_ioctl_balance_args *bargs;
4566 struct btrfs_balance_control *bctl;
Ilya Dryomoved0fb782013-01-20 15:57:57 +02004567 bool need_unlock; /* for mut. excl. ops lock */
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004568 int ret;
4569
4570 if (!capable(CAP_SYS_ADMIN))
4571 return -EPERM;
4572
Liu Boe54bfa32012-06-29 03:58:48 -06004573 ret = mnt_want_write_file(file);
Liu Bo9ba1f6e2012-05-11 18:11:26 +08004574 if (ret)
4575 return ret;
4576
Ilya Dryomoved0fb782013-01-20 15:57:57 +02004577again:
4578 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
4579 mutex_lock(&fs_info->volume_mutex);
4580 mutex_lock(&fs_info->balance_mutex);
4581 need_unlock = true;
4582 goto locked;
4583 }
4584
4585 /*
4586 * mut. excl. ops lock is locked. Three possibilites:
4587 * (1) some other op is running
4588 * (2) balance is running
4589 * (3) balance is paused -- special case (think resume)
4590 */
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004591 mutex_lock(&fs_info->balance_mutex);
Ilya Dryomoved0fb782013-01-20 15:57:57 +02004592 if (fs_info->balance_ctl) {
4593 /* this is either (2) or (3) */
4594 if (!atomic_read(&fs_info->balance_running)) {
4595 mutex_unlock(&fs_info->balance_mutex);
4596 if (!mutex_trylock(&fs_info->volume_mutex))
4597 goto again;
4598 mutex_lock(&fs_info->balance_mutex);
4599
4600 if (fs_info->balance_ctl &&
4601 !atomic_read(&fs_info->balance_running)) {
4602 /* this is (3) */
4603 need_unlock = false;
4604 goto locked;
4605 }
4606
4607 mutex_unlock(&fs_info->balance_mutex);
4608 mutex_unlock(&fs_info->volume_mutex);
4609 goto again;
4610 } else {
4611 /* this is (2) */
4612 mutex_unlock(&fs_info->balance_mutex);
4613 ret = -EINPROGRESS;
4614 goto out;
4615 }
4616 } else {
4617 /* this is (1) */
4618 mutex_unlock(&fs_info->balance_mutex);
Anand Jaine57138b2013-08-21 11:44:48 +08004619 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
Ilya Dryomoved0fb782013-01-20 15:57:57 +02004620 goto out;
4621 }
4622
4623locked:
4624 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004625
4626 if (arg) {
4627 bargs = memdup_user(arg, sizeof(*bargs));
4628 if (IS_ERR(bargs)) {
4629 ret = PTR_ERR(bargs);
Ilya Dryomoved0fb782013-01-20 15:57:57 +02004630 goto out_unlock;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004631 }
Ilya Dryomovde322262012-01-16 22:04:49 +02004632
4633 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4634 if (!fs_info->balance_ctl) {
4635 ret = -ENOTCONN;
4636 goto out_bargs;
4637 }
4638
4639 bctl = fs_info->balance_ctl;
4640 spin_lock(&fs_info->balance_lock);
4641 bctl->flags |= BTRFS_BALANCE_RESUME;
4642 spin_unlock(&fs_info->balance_lock);
4643
4644 goto do_balance;
4645 }
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004646 } else {
4647 bargs = NULL;
4648 }
4649
Ilya Dryomoved0fb782013-01-20 15:57:57 +02004650 if (fs_info->balance_ctl) {
Ilya Dryomov837d5b62012-01-16 22:04:49 +02004651 ret = -EINPROGRESS;
4652 goto out_bargs;
4653 }
4654
David Sterba8d2db782015-11-04 15:38:29 +01004655 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004656 if (!bctl) {
4657 ret = -ENOMEM;
4658 goto out_bargs;
4659 }
4660
4661 bctl->fs_info = fs_info;
4662 if (arg) {
4663 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4664 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4665 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4666
4667 bctl->flags = bargs->flags;
Ilya Dryomovf43ffb62012-01-16 22:04:47 +02004668 } else {
4669 /* balance everything - no filters */
4670 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004671 }
4672
David Sterba8eb93452015-10-12 16:55:54 +02004673 if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
4674 ret = -EINVAL;
Christian Engelmayer0f89abf2015-10-21 00:50:06 +02004675 goto out_bctl;
David Sterba8eb93452015-10-12 16:55:54 +02004676 }
4677
Ilya Dryomovde322262012-01-16 22:04:49 +02004678do_balance:
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004679 /*
Ilya Dryomoved0fb782013-01-20 15:57:57 +02004680 * Ownership of bctl and mutually_exclusive_operation_running
4681 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4682 * or, if restriper was paused all the way until unmount, in
4683 * free_fs_info. mutually_exclusive_operation_running is
4684 * cleared in __cancel_balance.
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004685 */
Ilya Dryomoved0fb782013-01-20 15:57:57 +02004686 need_unlock = false;
4687
4688 ret = btrfs_balance(bctl, bargs);
Christian Engelmayer0f89abf2015-10-21 00:50:06 +02004689 bctl = NULL;
Ilya Dryomoved0fb782013-01-20 15:57:57 +02004690
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004691 if (arg) {
4692 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4693 ret = -EFAULT;
4694 }
4695
Christian Engelmayer0f89abf2015-10-21 00:50:06 +02004696out_bctl:
4697 kfree(bctl);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004698out_bargs:
4699 kfree(bargs);
Ilya Dryomoved0fb782013-01-20 15:57:57 +02004700out_unlock:
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004701 mutex_unlock(&fs_info->balance_mutex);
4702 mutex_unlock(&fs_info->volume_mutex);
Ilya Dryomoved0fb782013-01-20 15:57:57 +02004703 if (need_unlock)
4704 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4705out:
Liu Boe54bfa32012-06-29 03:58:48 -06004706 mnt_drop_write_file(file);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004707 return ret;
4708}
4709
Ilya Dryomov837d5b62012-01-16 22:04:49 +02004710static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4711{
4712 if (!capable(CAP_SYS_ADMIN))
4713 return -EPERM;
4714
4715 switch (cmd) {
4716 case BTRFS_BALANCE_CTL_PAUSE:
4717 return btrfs_pause_balance(root->fs_info);
Ilya Dryomova7e99c62012-01-16 22:04:49 +02004718 case BTRFS_BALANCE_CTL_CANCEL:
4719 return btrfs_cancel_balance(root->fs_info);
Ilya Dryomov837d5b62012-01-16 22:04:49 +02004720 }
4721
4722 return -EINVAL;
4723}
4724
Ilya Dryomov19a39dc2012-01-16 22:04:49 +02004725static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4726 void __user *arg)
4727{
4728 struct btrfs_fs_info *fs_info = root->fs_info;
4729 struct btrfs_ioctl_balance_args *bargs;
4730 int ret = 0;
4731
4732 if (!capable(CAP_SYS_ADMIN))
4733 return -EPERM;
4734
4735 mutex_lock(&fs_info->balance_mutex);
4736 if (!fs_info->balance_ctl) {
4737 ret = -ENOTCONN;
4738 goto out;
4739 }
4740
David Sterba8d2db782015-11-04 15:38:29 +01004741 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
Ilya Dryomov19a39dc2012-01-16 22:04:49 +02004742 if (!bargs) {
4743 ret = -ENOMEM;
4744 goto out;
4745 }
4746
4747 update_ioctl_balance_args(fs_info, 1, bargs);
4748
4749 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4750 ret = -EFAULT;
4751
4752 kfree(bargs);
4753out:
4754 mutex_unlock(&fs_info->balance_mutex);
4755 return ret;
4756}
4757
Miao Xie905b0dd2012-11-26 08:50:11 +00004758static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
Arne Jansen5d13a372011-09-14 15:53:51 +02004759{
Al Viro496ad9a2013-01-23 17:07:38 -05004760 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Arne Jansen5d13a372011-09-14 15:53:51 +02004761 struct btrfs_ioctl_quota_ctl_args *sa;
4762 struct btrfs_trans_handle *trans = NULL;
4763 int ret;
4764 int err;
4765
4766 if (!capable(CAP_SYS_ADMIN))
4767 return -EPERM;
4768
Miao Xie905b0dd2012-11-26 08:50:11 +00004769 ret = mnt_want_write_file(file);
4770 if (ret)
4771 return ret;
Arne Jansen5d13a372011-09-14 15:53:51 +02004772
4773 sa = memdup_user(arg, sizeof(*sa));
Miao Xie905b0dd2012-11-26 08:50:11 +00004774 if (IS_ERR(sa)) {
4775 ret = PTR_ERR(sa);
4776 goto drop_write;
4777 }
Arne Jansen5d13a372011-09-14 15:53:51 +02004778
Wang Shilong7708f022013-04-07 10:24:57 +00004779 down_write(&root->fs_info->subvol_sem);
Jan Schmidt2f232032013-04-25 16:04:51 +00004780 trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4781 if (IS_ERR(trans)) {
4782 ret = PTR_ERR(trans);
4783 goto out;
Arne Jansen5d13a372011-09-14 15:53:51 +02004784 }
4785
4786 switch (sa->cmd) {
4787 case BTRFS_QUOTA_CTL_ENABLE:
4788 ret = btrfs_quota_enable(trans, root->fs_info);
4789 break;
4790 case BTRFS_QUOTA_CTL_DISABLE:
4791 ret = btrfs_quota_disable(trans, root->fs_info);
4792 break;
Arne Jansen5d13a372011-09-14 15:53:51 +02004793 default:
4794 ret = -EINVAL;
4795 break;
4796 }
4797
Jan Schmidt2f232032013-04-25 16:04:51 +00004798 err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4799 if (err && !ret)
4800 ret = err;
Arne Jansen5d13a372011-09-14 15:53:51 +02004801out:
4802 kfree(sa);
Wang Shilong7708f022013-04-07 10:24:57 +00004803 up_write(&root->fs_info->subvol_sem);
Miao Xie905b0dd2012-11-26 08:50:11 +00004804drop_write:
4805 mnt_drop_write_file(file);
Arne Jansen5d13a372011-09-14 15:53:51 +02004806 return ret;
4807}
4808
Miao Xie905b0dd2012-11-26 08:50:11 +00004809static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
Arne Jansen5d13a372011-09-14 15:53:51 +02004810{
Al Viro496ad9a2013-01-23 17:07:38 -05004811 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Arne Jansen5d13a372011-09-14 15:53:51 +02004812 struct btrfs_ioctl_qgroup_assign_args *sa;
4813 struct btrfs_trans_handle *trans;
4814 int ret;
4815 int err;
4816
4817 if (!capable(CAP_SYS_ADMIN))
4818 return -EPERM;
4819
Miao Xie905b0dd2012-11-26 08:50:11 +00004820 ret = mnt_want_write_file(file);
4821 if (ret)
4822 return ret;
Arne Jansen5d13a372011-09-14 15:53:51 +02004823
4824 sa = memdup_user(arg, sizeof(*sa));
Miao Xie905b0dd2012-11-26 08:50:11 +00004825 if (IS_ERR(sa)) {
4826 ret = PTR_ERR(sa);
4827 goto drop_write;
4828 }
Arne Jansen5d13a372011-09-14 15:53:51 +02004829
4830 trans = btrfs_join_transaction(root);
4831 if (IS_ERR(trans)) {
4832 ret = PTR_ERR(trans);
4833 goto out;
4834 }
4835
4836 /* FIXME: check if the IDs really exist */
4837 if (sa->assign) {
4838 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4839 sa->src, sa->dst);
4840 } else {
4841 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4842 sa->src, sa->dst);
4843 }
4844
Qu Wenruoe082f562015-02-27 16:24:28 +08004845 /* update qgroup status and info */
4846 err = btrfs_run_qgroups(trans, root->fs_info);
4847 if (err < 0)
Anand Jaina4553fe2015-09-25 14:43:01 +08004848 btrfs_std_error(root->fs_info, ret,
Qu Wenruoe082f562015-02-27 16:24:28 +08004849 "failed to update qgroup status and info\n");
Arne Jansen5d13a372011-09-14 15:53:51 +02004850 err = btrfs_end_transaction(trans, root);
4851 if (err && !ret)
4852 ret = err;
4853
4854out:
4855 kfree(sa);
Miao Xie905b0dd2012-11-26 08:50:11 +00004856drop_write:
4857 mnt_drop_write_file(file);
Arne Jansen5d13a372011-09-14 15:53:51 +02004858 return ret;
4859}
4860
Miao Xie905b0dd2012-11-26 08:50:11 +00004861static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
Arne Jansen5d13a372011-09-14 15:53:51 +02004862{
Al Viro496ad9a2013-01-23 17:07:38 -05004863 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Arne Jansen5d13a372011-09-14 15:53:51 +02004864 struct btrfs_ioctl_qgroup_create_args *sa;
4865 struct btrfs_trans_handle *trans;
4866 int ret;
4867 int err;
4868
4869 if (!capable(CAP_SYS_ADMIN))
4870 return -EPERM;
4871
Miao Xie905b0dd2012-11-26 08:50:11 +00004872 ret = mnt_want_write_file(file);
4873 if (ret)
4874 return ret;
Arne Jansen5d13a372011-09-14 15:53:51 +02004875
4876 sa = memdup_user(arg, sizeof(*sa));
Miao Xie905b0dd2012-11-26 08:50:11 +00004877 if (IS_ERR(sa)) {
4878 ret = PTR_ERR(sa);
4879 goto drop_write;
4880 }
Arne Jansen5d13a372011-09-14 15:53:51 +02004881
Miao Xied86e56c2012-11-15 11:35:41 +00004882 if (!sa->qgroupid) {
4883 ret = -EINVAL;
4884 goto out;
4885 }
4886
Arne Jansen5d13a372011-09-14 15:53:51 +02004887 trans = btrfs_join_transaction(root);
4888 if (IS_ERR(trans)) {
4889 ret = PTR_ERR(trans);
4890 goto out;
4891 }
4892
4893 /* FIXME: check if the IDs really exist */
4894 if (sa->create) {
Dongsheng Yang4087cf22015-01-18 10:59:23 -05004895 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid);
Arne Jansen5d13a372011-09-14 15:53:51 +02004896 } else {
4897 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4898 }
4899
4900 err = btrfs_end_transaction(trans, root);
4901 if (err && !ret)
4902 ret = err;
4903
4904out:
4905 kfree(sa);
Miao Xie905b0dd2012-11-26 08:50:11 +00004906drop_write:
4907 mnt_drop_write_file(file);
Arne Jansen5d13a372011-09-14 15:53:51 +02004908 return ret;
4909}
4910
Miao Xie905b0dd2012-11-26 08:50:11 +00004911static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
Arne Jansen5d13a372011-09-14 15:53:51 +02004912{
Al Viro496ad9a2013-01-23 17:07:38 -05004913 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Arne Jansen5d13a372011-09-14 15:53:51 +02004914 struct btrfs_ioctl_qgroup_limit_args *sa;
4915 struct btrfs_trans_handle *trans;
4916 int ret;
4917 int err;
4918 u64 qgroupid;
4919
4920 if (!capable(CAP_SYS_ADMIN))
4921 return -EPERM;
4922
Miao Xie905b0dd2012-11-26 08:50:11 +00004923 ret = mnt_want_write_file(file);
4924 if (ret)
4925 return ret;
Arne Jansen5d13a372011-09-14 15:53:51 +02004926
4927 sa = memdup_user(arg, sizeof(*sa));
Miao Xie905b0dd2012-11-26 08:50:11 +00004928 if (IS_ERR(sa)) {
4929 ret = PTR_ERR(sa);
4930 goto drop_write;
4931 }
Arne Jansen5d13a372011-09-14 15:53:51 +02004932
4933 trans = btrfs_join_transaction(root);
4934 if (IS_ERR(trans)) {
4935 ret = PTR_ERR(trans);
4936 goto out;
4937 }
4938
4939 qgroupid = sa->qgroupid;
4940 if (!qgroupid) {
4941 /* take the current subvol as qgroup */
4942 qgroupid = root->root_key.objectid;
4943 }
4944
4945 /* FIXME: check if the IDs really exist */
4946 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
4947
4948 err = btrfs_end_transaction(trans, root);
4949 if (err && !ret)
4950 ret = err;
4951
4952out:
4953 kfree(sa);
Miao Xie905b0dd2012-11-26 08:50:11 +00004954drop_write:
4955 mnt_drop_write_file(file);
Arne Jansen5d13a372011-09-14 15:53:51 +02004956 return ret;
4957}
4958
Jan Schmidt2f232032013-04-25 16:04:51 +00004959static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4960{
Al Viro6d0379e2013-06-16 19:32:35 +04004961 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Jan Schmidt2f232032013-04-25 16:04:51 +00004962 struct btrfs_ioctl_quota_rescan_args *qsa;
4963 int ret;
4964
4965 if (!capable(CAP_SYS_ADMIN))
4966 return -EPERM;
4967
4968 ret = mnt_want_write_file(file);
4969 if (ret)
4970 return ret;
4971
4972 qsa = memdup_user(arg, sizeof(*qsa));
4973 if (IS_ERR(qsa)) {
4974 ret = PTR_ERR(qsa);
4975 goto drop_write;
4976 }
4977
4978 if (qsa->flags) {
4979 ret = -EINVAL;
4980 goto out;
4981 }
4982
4983 ret = btrfs_qgroup_rescan(root->fs_info);
4984
4985out:
4986 kfree(qsa);
4987drop_write:
4988 mnt_drop_write_file(file);
4989 return ret;
4990}
4991
4992static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4993{
Al Viro6d0379e2013-06-16 19:32:35 +04004994 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Jan Schmidt2f232032013-04-25 16:04:51 +00004995 struct btrfs_ioctl_quota_rescan_args *qsa;
4996 int ret = 0;
4997
4998 if (!capable(CAP_SYS_ADMIN))
4999 return -EPERM;
5000
David Sterba8d2db782015-11-04 15:38:29 +01005001 qsa = kzalloc(sizeof(*qsa), GFP_KERNEL);
Jan Schmidt2f232032013-04-25 16:04:51 +00005002 if (!qsa)
5003 return -ENOMEM;
5004
5005 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
5006 qsa->flags = 1;
5007 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
5008 }
5009
5010 if (copy_to_user(arg, qsa, sizeof(*qsa)))
5011 ret = -EFAULT;
5012
5013 kfree(qsa);
5014 return ret;
5015}
5016
Jan Schmidt57254b6e2013-05-06 19:14:17 +00005017static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
5018{
Al Viro54563d42013-09-01 15:57:51 -04005019 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Jan Schmidt57254b6e2013-05-06 19:14:17 +00005020
5021 if (!capable(CAP_SYS_ADMIN))
5022 return -EPERM;
5023
5024 return btrfs_qgroup_wait_for_completion(root->fs_info);
5025}
5026
Hugo Millsabccd002014-01-30 20:17:00 +00005027static long _btrfs_ioctl_set_received_subvol(struct file *file,
5028 struct btrfs_ioctl_received_subvol_args *sa)
Alexander Block8ea05e32012-07-25 17:35:53 +02005029{
Al Viro496ad9a2013-01-23 17:07:38 -05005030 struct inode *inode = file_inode(file);
Alexander Block8ea05e32012-07-25 17:35:53 +02005031 struct btrfs_root *root = BTRFS_I(inode)->root;
5032 struct btrfs_root_item *root_item = &root->root_item;
5033 struct btrfs_trans_handle *trans;
5034 struct timespec ct = CURRENT_TIME;
5035 int ret = 0;
Stefan Behrensdd5f9612013-08-15 17:11:20 +02005036 int received_uuid_changed;
Alexander Block8ea05e32012-07-25 17:35:53 +02005037
David Sterbabd60ea02014-01-16 15:50:22 +01005038 if (!inode_owner_or_capable(inode))
5039 return -EPERM;
5040
Alexander Block8ea05e32012-07-25 17:35:53 +02005041 ret = mnt_want_write_file(file);
5042 if (ret < 0)
5043 return ret;
5044
5045 down_write(&root->fs_info->subvol_sem);
5046
5047 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
5048 ret = -EINVAL;
5049 goto out;
5050 }
5051
5052 if (btrfs_root_readonly(root)) {
5053 ret = -EROFS;
5054 goto out;
5055 }
5056
Stefan Behrensdd5f9612013-08-15 17:11:20 +02005057 /*
5058 * 1 - root item
5059 * 2 - uuid items (received uuid + subvol uuid)
5060 */
5061 trans = btrfs_start_transaction(root, 3);
Alexander Block8ea05e32012-07-25 17:35:53 +02005062 if (IS_ERR(trans)) {
5063 ret = PTR_ERR(trans);
5064 trans = NULL;
5065 goto out;
5066 }
5067
5068 sa->rtransid = trans->transid;
5069 sa->rtime.sec = ct.tv_sec;
5070 sa->rtime.nsec = ct.tv_nsec;
5071
Stefan Behrensdd5f9612013-08-15 17:11:20 +02005072 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
5073 BTRFS_UUID_SIZE);
5074 if (received_uuid_changed &&
5075 !btrfs_is_empty_uuid(root_item->received_uuid))
5076 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
5077 root_item->received_uuid,
5078 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5079 root->root_key.objectid);
Alexander Block8ea05e32012-07-25 17:35:53 +02005080 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
5081 btrfs_set_root_stransid(root_item, sa->stransid);
5082 btrfs_set_root_rtransid(root_item, sa->rtransid);
Qu Wenruo3cae2102013-07-16 11:19:18 +08005083 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
5084 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
5085 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
5086 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
Alexander Block8ea05e32012-07-25 17:35:53 +02005087
5088 ret = btrfs_update_root(trans, root->fs_info->tree_root,
5089 &root->root_key, &root->root_item);
5090 if (ret < 0) {
5091 btrfs_end_transaction(trans, root);
Alexander Block8ea05e32012-07-25 17:35:53 +02005092 goto out;
Stefan Behrensdd5f9612013-08-15 17:11:20 +02005093 }
5094 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
5095 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
5096 sa->uuid,
5097 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5098 root->root_key.objectid);
5099 if (ret < 0 && ret != -EEXIST) {
5100 btrfs_abort_transaction(trans, root, ret);
Alexander Block8ea05e32012-07-25 17:35:53 +02005101 goto out;
Stefan Behrensdd5f9612013-08-15 17:11:20 +02005102 }
5103 }
5104 ret = btrfs_commit_transaction(trans, root);
5105 if (ret < 0) {
5106 btrfs_abort_transaction(trans, root, ret);
5107 goto out;
Alexander Block8ea05e32012-07-25 17:35:53 +02005108 }
5109
Hugo Millsabccd002014-01-30 20:17:00 +00005110out:
5111 up_write(&root->fs_info->subvol_sem);
5112 mnt_drop_write_file(file);
5113 return ret;
5114}
5115
5116#ifdef CONFIG_64BIT
5117static long btrfs_ioctl_set_received_subvol_32(struct file *file,
5118 void __user *arg)
5119{
5120 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
5121 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
5122 int ret = 0;
5123
5124 args32 = memdup_user(arg, sizeof(*args32));
5125 if (IS_ERR(args32)) {
5126 ret = PTR_ERR(args32);
5127 args32 = NULL;
5128 goto out;
5129 }
5130
David Sterba8d2db782015-11-04 15:38:29 +01005131 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
Dan Carpenter84dbeb82014-03-28 11:06:00 +03005132 if (!args64) {
5133 ret = -ENOMEM;
Hugo Millsabccd002014-01-30 20:17:00 +00005134 goto out;
5135 }
5136
5137 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
5138 args64->stransid = args32->stransid;
5139 args64->rtransid = args32->rtransid;
5140 args64->stime.sec = args32->stime.sec;
5141 args64->stime.nsec = args32->stime.nsec;
5142 args64->rtime.sec = args32->rtime.sec;
5143 args64->rtime.nsec = args32->rtime.nsec;
5144 args64->flags = args32->flags;
5145
5146 ret = _btrfs_ioctl_set_received_subvol(file, args64);
5147 if (ret)
5148 goto out;
5149
5150 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
5151 args32->stransid = args64->stransid;
5152 args32->rtransid = args64->rtransid;
5153 args32->stime.sec = args64->stime.sec;
5154 args32->stime.nsec = args64->stime.nsec;
5155 args32->rtime.sec = args64->rtime.sec;
5156 args32->rtime.nsec = args64->rtime.nsec;
5157 args32->flags = args64->flags;
5158
5159 ret = copy_to_user(arg, args32, sizeof(*args32));
5160 if (ret)
5161 ret = -EFAULT;
5162
5163out:
5164 kfree(args32);
5165 kfree(args64);
5166 return ret;
5167}
5168#endif
5169
5170static long btrfs_ioctl_set_received_subvol(struct file *file,
5171 void __user *arg)
5172{
5173 struct btrfs_ioctl_received_subvol_args *sa = NULL;
5174 int ret = 0;
5175
5176 sa = memdup_user(arg, sizeof(*sa));
5177 if (IS_ERR(sa)) {
5178 ret = PTR_ERR(sa);
5179 sa = NULL;
5180 goto out;
5181 }
5182
5183 ret = _btrfs_ioctl_set_received_subvol(file, sa);
5184
5185 if (ret)
5186 goto out;
5187
Alexander Block8ea05e32012-07-25 17:35:53 +02005188 ret = copy_to_user(arg, sa, sizeof(*sa));
5189 if (ret)
5190 ret = -EFAULT;
5191
5192out:
5193 kfree(sa);
Alexander Block8ea05e32012-07-25 17:35:53 +02005194 return ret;
5195}
5196
jeff.liu867ab662013-01-05 02:48:01 +00005197static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
5198{
Al Viro6d0379e2013-06-16 19:32:35 +04005199 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Anand Jaina1b83ac2013-07-19 17:39:32 +08005200 size_t len;
jeff.liu867ab662013-01-05 02:48:01 +00005201 int ret;
Anand Jaina1b83ac2013-07-19 17:39:32 +08005202 char label[BTRFS_LABEL_SIZE];
5203
5204 spin_lock(&root->fs_info->super_lock);
5205 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5206 spin_unlock(&root->fs_info->super_lock);
5207
5208 len = strnlen(label, BTRFS_LABEL_SIZE);
jeff.liu867ab662013-01-05 02:48:01 +00005209
5210 if (len == BTRFS_LABEL_SIZE) {
Frank Holtonefe120a2013-12-20 11:37:06 -05005211 btrfs_warn(root->fs_info,
5212 "label is too long, return the first %zu bytes", --len);
jeff.liu867ab662013-01-05 02:48:01 +00005213 }
5214
jeff.liu867ab662013-01-05 02:48:01 +00005215 ret = copy_to_user(arg, label, len);
jeff.liu867ab662013-01-05 02:48:01 +00005216
5217 return ret ? -EFAULT : 0;
5218}
5219
jeff.liua8bfd4a2013-01-05 02:48:08 +00005220static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5221{
Al Viro6d0379e2013-06-16 19:32:35 +04005222 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
jeff.liua8bfd4a2013-01-05 02:48:08 +00005223 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5224 struct btrfs_trans_handle *trans;
5225 char label[BTRFS_LABEL_SIZE];
5226 int ret;
5227
5228 if (!capable(CAP_SYS_ADMIN))
5229 return -EPERM;
5230
5231 if (copy_from_user(label, arg, sizeof(label)))
5232 return -EFAULT;
5233
5234 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
Frank Holtonefe120a2013-12-20 11:37:06 -05005235 btrfs_err(root->fs_info, "unable to set label with more than %d bytes",
jeff.liua8bfd4a2013-01-05 02:48:08 +00005236 BTRFS_LABEL_SIZE - 1);
5237 return -EINVAL;
5238 }
5239
5240 ret = mnt_want_write_file(file);
5241 if (ret)
5242 return ret;
5243
jeff.liua8bfd4a2013-01-05 02:48:08 +00005244 trans = btrfs_start_transaction(root, 0);
5245 if (IS_ERR(trans)) {
5246 ret = PTR_ERR(trans);
5247 goto out_unlock;
5248 }
5249
Anand Jaina1b83ac2013-07-19 17:39:32 +08005250 spin_lock(&root->fs_info->super_lock);
jeff.liua8bfd4a2013-01-05 02:48:08 +00005251 strcpy(super_block->label, label);
Anand Jaina1b83ac2013-07-19 17:39:32 +08005252 spin_unlock(&root->fs_info->super_lock);
Jeff Mahoneyd0270ac2014-02-07 14:33:57 +01005253 ret = btrfs_commit_transaction(trans, root);
jeff.liua8bfd4a2013-01-05 02:48:08 +00005254
5255out_unlock:
jeff.liua8bfd4a2013-01-05 02:48:08 +00005256 mnt_drop_write_file(file);
5257 return ret;
5258}
5259
Jeff Mahoney2eaa0552013-11-15 15:33:55 -05005260#define INIT_FEATURE_FLAGS(suffix) \
5261 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5262 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5263 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5264
5265static int btrfs_ioctl_get_supported_features(struct file *file,
5266 void __user *arg)
5267{
David Sterba4d4ab6d2015-11-19 11:42:31 +01005268 static const struct btrfs_ioctl_feature_flags features[3] = {
Jeff Mahoney2eaa0552013-11-15 15:33:55 -05005269 INIT_FEATURE_FLAGS(SUPP),
5270 INIT_FEATURE_FLAGS(SAFE_SET),
5271 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5272 };
5273
5274 if (copy_to_user(arg, &features, sizeof(features)))
5275 return -EFAULT;
5276
5277 return 0;
5278}
5279
5280static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5281{
5282 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5283 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5284 struct btrfs_ioctl_feature_flags features;
5285
5286 features.compat_flags = btrfs_super_compat_flags(super_block);
5287 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5288 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5289
5290 if (copy_to_user(arg, &features, sizeof(features)))
5291 return -EFAULT;
5292
5293 return 0;
5294}
5295
Jeff Mahoney3b02a682013-11-01 13:07:02 -04005296static int check_feature_bits(struct btrfs_root *root,
5297 enum btrfs_feature_set set,
Jeff Mahoney2eaa0552013-11-15 15:33:55 -05005298 u64 change_mask, u64 flags, u64 supported_flags,
5299 u64 safe_set, u64 safe_clear)
5300{
Jeff Mahoney3b02a682013-11-01 13:07:02 -04005301 const char *type = btrfs_feature_set_names[set];
5302 char *names;
Jeff Mahoney2eaa0552013-11-15 15:33:55 -05005303 u64 disallowed, unsupported;
5304 u64 set_mask = flags & change_mask;
5305 u64 clear_mask = ~flags & change_mask;
5306
5307 unsupported = set_mask & ~supported_flags;
5308 if (unsupported) {
Jeff Mahoney3b02a682013-11-01 13:07:02 -04005309 names = btrfs_printable_features(set, unsupported);
5310 if (names) {
5311 btrfs_warn(root->fs_info,
5312 "this kernel does not support the %s feature bit%s",
5313 names, strchr(names, ',') ? "s" : "");
5314 kfree(names);
5315 } else
5316 btrfs_warn(root->fs_info,
Jeff Mahoney2eaa0552013-11-15 15:33:55 -05005317 "this kernel does not support %s bits 0x%llx",
5318 type, unsupported);
5319 return -EOPNOTSUPP;
5320 }
5321
5322 disallowed = set_mask & ~safe_set;
5323 if (disallowed) {
Jeff Mahoney3b02a682013-11-01 13:07:02 -04005324 names = btrfs_printable_features(set, disallowed);
5325 if (names) {
5326 btrfs_warn(root->fs_info,
5327 "can't set the %s feature bit%s while mounted",
5328 names, strchr(names, ',') ? "s" : "");
5329 kfree(names);
5330 } else
5331 btrfs_warn(root->fs_info,
Jeff Mahoney2eaa0552013-11-15 15:33:55 -05005332 "can't set %s bits 0x%llx while mounted",
5333 type, disallowed);
5334 return -EPERM;
5335 }
5336
5337 disallowed = clear_mask & ~safe_clear;
5338 if (disallowed) {
Jeff Mahoney3b02a682013-11-01 13:07:02 -04005339 names = btrfs_printable_features(set, disallowed);
5340 if (names) {
5341 btrfs_warn(root->fs_info,
5342 "can't clear the %s feature bit%s while mounted",
5343 names, strchr(names, ',') ? "s" : "");
5344 kfree(names);
5345 } else
5346 btrfs_warn(root->fs_info,
Jeff Mahoney2eaa0552013-11-15 15:33:55 -05005347 "can't clear %s bits 0x%llx while mounted",
5348 type, disallowed);
5349 return -EPERM;
5350 }
5351
5352 return 0;
5353}
5354
5355#define check_feature(root, change_mask, flags, mask_base) \
Jeff Mahoney3b02a682013-11-01 13:07:02 -04005356check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
Jeff Mahoney2eaa0552013-11-15 15:33:55 -05005357 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5358 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5359 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5360
5361static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5362{
5363 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5364 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5365 struct btrfs_ioctl_feature_flags flags[2];
5366 struct btrfs_trans_handle *trans;
5367 u64 newflags;
5368 int ret;
5369
5370 if (!capable(CAP_SYS_ADMIN))
5371 return -EPERM;
5372
5373 if (copy_from_user(flags, arg, sizeof(flags)))
5374 return -EFAULT;
5375
5376 /* Nothing to do */
5377 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5378 !flags[0].incompat_flags)
5379 return 0;
5380
5381 ret = check_feature(root, flags[0].compat_flags,
5382 flags[1].compat_flags, COMPAT);
5383 if (ret)
5384 return ret;
5385
5386 ret = check_feature(root, flags[0].compat_ro_flags,
5387 flags[1].compat_ro_flags, COMPAT_RO);
5388 if (ret)
5389 return ret;
5390
5391 ret = check_feature(root, flags[0].incompat_flags,
5392 flags[1].incompat_flags, INCOMPAT);
5393 if (ret)
5394 return ret;
5395
David Sterba8051aa12014-02-07 14:34:04 +01005396 trans = btrfs_start_transaction(root, 0);
Jeff Mahoney2eaa0552013-11-15 15:33:55 -05005397 if (IS_ERR(trans))
5398 return PTR_ERR(trans);
5399
5400 spin_lock(&root->fs_info->super_lock);
5401 newflags = btrfs_super_compat_flags(super_block);
5402 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5403 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5404 btrfs_set_super_compat_flags(super_block, newflags);
5405
5406 newflags = btrfs_super_compat_ro_flags(super_block);
5407 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5408 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5409 btrfs_set_super_compat_ro_flags(super_block, newflags);
5410
5411 newflags = btrfs_super_incompat_flags(super_block);
5412 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5413 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5414 btrfs_set_super_incompat_flags(super_block, newflags);
5415 spin_unlock(&root->fs_info->super_lock);
5416
Jeff Mahoneyd0270ac2014-02-07 14:33:57 +01005417 return btrfs_commit_transaction(trans, root);
Jeff Mahoney2eaa0552013-11-15 15:33:55 -05005418}
5419
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04005420long btrfs_ioctl(struct file *file, unsigned int
5421 cmd, unsigned long arg)
5422{
Al Viro496ad9a2013-01-23 17:07:38 -05005423 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Christoph Hellwig4bcabaa2008-12-02 06:36:08 -05005424 void __user *argp = (void __user *)arg;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04005425
5426 switch (cmd) {
Christoph Hellwig6cbff002009-04-17 10:37:41 +02005427 case FS_IOC_GETFLAGS:
5428 return btrfs_ioctl_getflags(file, argp);
5429 case FS_IOC_SETFLAGS:
5430 return btrfs_ioctl_setflags(file, argp);
5431 case FS_IOC_GETVERSION:
5432 return btrfs_ioctl_getversion(file, argp);
Li Dongyangf7039b12011-03-24 10:24:28 +00005433 case FITRIM:
5434 return btrfs_ioctl_fitrim(file, argp);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04005435 case BTRFS_IOC_SNAP_CREATE:
Li Zefanfa0d2b92010-12-20 15:53:28 +08005436 return btrfs_ioctl_snap_create(file, argp, 0);
Li Zefanfdfb1e42010-12-10 06:41:56 +00005437 case BTRFS_IOC_SNAP_CREATE_V2:
Li Zefanfa0d2b92010-12-20 15:53:28 +08005438 return btrfs_ioctl_snap_create_v2(file, argp, 0);
Chris Mason3de45862008-11-17 21:02:50 -05005439 case BTRFS_IOC_SUBVOL_CREATE:
Li Zefanfa0d2b92010-12-20 15:53:28 +08005440 return btrfs_ioctl_snap_create(file, argp, 1);
Arne Jansen6f72c7e2011-09-14 15:58:21 +02005441 case BTRFS_IOC_SUBVOL_CREATE_V2:
5442 return btrfs_ioctl_snap_create_v2(file, argp, 1);
Yan, Zheng76dda932009-09-21 16:00:26 -04005443 case BTRFS_IOC_SNAP_DESTROY:
5444 return btrfs_ioctl_snap_destroy(file, argp);
Li Zefan0caa1022010-12-20 16:30:25 +08005445 case BTRFS_IOC_SUBVOL_GETFLAGS:
5446 return btrfs_ioctl_subvol_getflags(file, argp);
5447 case BTRFS_IOC_SUBVOL_SETFLAGS:
5448 return btrfs_ioctl_subvol_setflags(file, argp);
Josef Bacik6ef5ed02009-12-11 21:11:29 +00005449 case BTRFS_IOC_DEFAULT_SUBVOL:
5450 return btrfs_ioctl_default_subvol(file, argp);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04005451 case BTRFS_IOC_DEFRAG:
Chris Mason1e701a32010-03-11 09:42:04 -05005452 return btrfs_ioctl_defrag(file, NULL);
5453 case BTRFS_IOC_DEFRAG_RANGE:
5454 return btrfs_ioctl_defrag(file, argp);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04005455 case BTRFS_IOC_RESIZE:
Miao Xie198605a2012-11-26 08:43:45 +00005456 return btrfs_ioctl_resize(file, argp);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04005457 case BTRFS_IOC_ADD_DEV:
Christoph Hellwig4bcabaa2008-12-02 06:36:08 -05005458 return btrfs_ioctl_add_dev(root, argp);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04005459 case BTRFS_IOC_RM_DEV:
Miao Xieda249272012-11-26 08:44:50 +00005460 return btrfs_ioctl_rm_dev(file, argp);
Jan Schmidt475f6382011-03-11 15:41:01 +01005461 case BTRFS_IOC_FS_INFO:
5462 return btrfs_ioctl_fs_info(root, argp);
5463 case BTRFS_IOC_DEV_INFO:
5464 return btrfs_ioctl_dev_info(root, argp);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04005465 case BTRFS_IOC_BALANCE:
Liu Bo9ba1f6e2012-05-11 18:11:26 +08005466 return btrfs_ioctl_balance(file, NULL);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04005467 case BTRFS_IOC_TRANS_START:
5468 return btrfs_ioctl_trans_start(file);
5469 case BTRFS_IOC_TRANS_END:
5470 return btrfs_ioctl_trans_end(file);
Chris Masonac8e9812010-02-28 15:39:26 -05005471 case BTRFS_IOC_TREE_SEARCH:
5472 return btrfs_ioctl_tree_search(file, argp);
Gerhard Heiftcc68a8a2014-01-30 16:24:03 +01005473 case BTRFS_IOC_TREE_SEARCH_V2:
5474 return btrfs_ioctl_tree_search_v2(file, argp);
Chris Masonac8e9812010-02-28 15:39:26 -05005475 case BTRFS_IOC_INO_LOOKUP:
5476 return btrfs_ioctl_ino_lookup(file, argp);
Jan Schmidtd7728c92011-07-07 16:48:38 +02005477 case BTRFS_IOC_INO_PATHS:
5478 return btrfs_ioctl_ino_to_path(root, argp);
5479 case BTRFS_IOC_LOGICAL_INO:
5480 return btrfs_ioctl_logical_to_ino(root, argp);
Josef Bacik1406e432010-01-13 18:19:06 +00005481 case BTRFS_IOC_SPACE_INFO:
5482 return btrfs_ioctl_space_info(root, argp);
Filipe David Borba Manana9b199852013-09-23 11:35:11 +01005483 case BTRFS_IOC_SYNC: {
5484 int ret;
5485
Miao Xie6c255e62014-03-06 13:55:01 +08005486 ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1);
Filipe David Borba Manana9b199852013-09-23 11:35:11 +01005487 if (ret)
5488 return ret;
Al Viroddb52f42014-10-21 20:21:18 -04005489 ret = btrfs_sync_fs(file_inode(file)->i_sb, 1);
David Sterba2fad4e82014-07-23 14:39:35 +02005490 /*
5491 * The transaction thread may want to do more work,
5492 * namely it pokes the cleaner ktread that will start
5493 * processing uncleaned subvols.
5494 */
5495 wake_up_process(root->fs_info->transaction_kthread);
Filipe David Borba Manana9b199852013-09-23 11:35:11 +01005496 return ret;
5497 }
Sage Weil46204592010-10-29 15:41:32 -04005498 case BTRFS_IOC_START_SYNC:
Miao Xie9a8c28b2012-11-26 08:40:43 +00005499 return btrfs_ioctl_start_sync(root, argp);
Sage Weil46204592010-10-29 15:41:32 -04005500 case BTRFS_IOC_WAIT_SYNC:
Miao Xie9a8c28b2012-11-26 08:40:43 +00005501 return btrfs_ioctl_wait_sync(root, argp);
Jan Schmidt475f6382011-03-11 15:41:01 +01005502 case BTRFS_IOC_SCRUB:
Miao Xieb8e95482012-11-26 08:48:01 +00005503 return btrfs_ioctl_scrub(file, argp);
Jan Schmidt475f6382011-03-11 15:41:01 +01005504 case BTRFS_IOC_SCRUB_CANCEL:
5505 return btrfs_ioctl_scrub_cancel(root, argp);
5506 case BTRFS_IOC_SCRUB_PROGRESS:
5507 return btrfs_ioctl_scrub_progress(root, argp);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02005508 case BTRFS_IOC_BALANCE_V2:
Liu Bo9ba1f6e2012-05-11 18:11:26 +08005509 return btrfs_ioctl_balance(file, argp);
Ilya Dryomov837d5b62012-01-16 22:04:49 +02005510 case BTRFS_IOC_BALANCE_CTL:
5511 return btrfs_ioctl_balance_ctl(root, arg);
Ilya Dryomov19a39dc2012-01-16 22:04:49 +02005512 case BTRFS_IOC_BALANCE_PROGRESS:
5513 return btrfs_ioctl_balance_progress(root, argp);
Alexander Block8ea05e32012-07-25 17:35:53 +02005514 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5515 return btrfs_ioctl_set_received_subvol(file, argp);
Hugo Millsabccd002014-01-30 20:17:00 +00005516#ifdef CONFIG_64BIT
5517 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5518 return btrfs_ioctl_set_received_subvol_32(file, argp);
5519#endif
Alexander Block31db9f72012-07-25 23:19:24 +02005520 case BTRFS_IOC_SEND:
5521 return btrfs_ioctl_send(file, argp);
Stefan Behrensc11d2c22012-05-25 16:06:09 +02005522 case BTRFS_IOC_GET_DEV_STATS:
David Sterbab27f7c02012-06-22 06:30:39 -06005523 return btrfs_ioctl_get_dev_stats(root, argp);
Arne Jansen5d13a372011-09-14 15:53:51 +02005524 case BTRFS_IOC_QUOTA_CTL:
Miao Xie905b0dd2012-11-26 08:50:11 +00005525 return btrfs_ioctl_quota_ctl(file, argp);
Arne Jansen5d13a372011-09-14 15:53:51 +02005526 case BTRFS_IOC_QGROUP_ASSIGN:
Miao Xie905b0dd2012-11-26 08:50:11 +00005527 return btrfs_ioctl_qgroup_assign(file, argp);
Arne Jansen5d13a372011-09-14 15:53:51 +02005528 case BTRFS_IOC_QGROUP_CREATE:
Miao Xie905b0dd2012-11-26 08:50:11 +00005529 return btrfs_ioctl_qgroup_create(file, argp);
Arne Jansen5d13a372011-09-14 15:53:51 +02005530 case BTRFS_IOC_QGROUP_LIMIT:
Miao Xie905b0dd2012-11-26 08:50:11 +00005531 return btrfs_ioctl_qgroup_limit(file, argp);
Jan Schmidt2f232032013-04-25 16:04:51 +00005532 case BTRFS_IOC_QUOTA_RESCAN:
5533 return btrfs_ioctl_quota_rescan(file, argp);
5534 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5535 return btrfs_ioctl_quota_rescan_status(file, argp);
Jan Schmidt57254b6e2013-05-06 19:14:17 +00005536 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5537 return btrfs_ioctl_quota_rescan_wait(file, argp);
Stefan Behrens3f6bcfb2012-11-06 15:08:53 +01005538 case BTRFS_IOC_DEV_REPLACE:
5539 return btrfs_ioctl_dev_replace(root, argp);
jeff.liu867ab662013-01-05 02:48:01 +00005540 case BTRFS_IOC_GET_FSLABEL:
5541 return btrfs_ioctl_get_fslabel(file, argp);
jeff.liua8bfd4a2013-01-05 02:48:08 +00005542 case BTRFS_IOC_SET_FSLABEL:
5543 return btrfs_ioctl_set_fslabel(file, argp);
Jeff Mahoney2eaa0552013-11-15 15:33:55 -05005544 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5545 return btrfs_ioctl_get_supported_features(file, argp);
5546 case BTRFS_IOC_GET_FEATURES:
5547 return btrfs_ioctl_get_features(file, argp);
5548 case BTRFS_IOC_SET_FEATURES:
5549 return btrfs_ioctl_set_features(file, argp);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04005550 }
5551
5552 return -ENOTTY;
5553}