blob: af69129d7e0e0b3bd269d30c4d6a20b583eb8d32 [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"
Filipe Manana1ec9a1a2016-02-10 10:42:25 +000062#include "tree-log.h"
Anand Jainebb87652016-03-10 17:26:59 +080063#include "compression.h"
Christoph Hellwigf46b5a62008-06-11 21:53:53 -040064
Hugo Millsabccd002014-01-30 20:17:00 +000065#ifdef CONFIG_64BIT
66/* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
67 * structures are incorrect, as the timespec structure from userspace
68 * is 4 bytes too small. We define these alternatives here to teach
69 * the kernel about the 32-bit struct packing.
70 */
71struct btrfs_ioctl_timespec_32 {
72 __u64 sec;
73 __u32 nsec;
74} __attribute__ ((__packed__));
75
76struct btrfs_ioctl_received_subvol_args_32 {
77 char uuid[BTRFS_UUID_SIZE]; /* in */
78 __u64 stransid; /* in */
79 __u64 rtransid; /* out */
80 struct btrfs_ioctl_timespec_32 stime; /* in */
81 struct btrfs_ioctl_timespec_32 rtime; /* out */
82 __u64 flags; /* in */
83 __u64 reserved[16]; /* in */
84} __attribute__ ((__packed__));
85
86#define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
87 struct btrfs_ioctl_received_subvol_args_32)
88#endif
89
90
Mark Fasheh416161d2013-08-06 11:42:51 -070091static int btrfs_clone(struct inode *src, struct inode *inode,
Mark Fasheh1c919a52015-06-30 14:42:08 -070092 u64 off, u64 olen, u64 olen_aligned, u64 destoff,
93 int no_time_update);
Mark Fasheh416161d2013-08-06 11:42:51 -070094
Christoph Hellwig6cbff002009-04-17 10:37:41 +020095/* Mask out flags that are inappropriate for the given type of inode. */
96static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
97{
98 if (S_ISDIR(mode))
99 return flags;
100 else if (S_ISREG(mode))
101 return flags & ~FS_DIRSYNC_FL;
102 else
103 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
104}
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400105
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200106/*
107 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
108 */
109static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
110{
111 unsigned int iflags = 0;
112
113 if (flags & BTRFS_INODE_SYNC)
114 iflags |= FS_SYNC_FL;
115 if (flags & BTRFS_INODE_IMMUTABLE)
116 iflags |= FS_IMMUTABLE_FL;
117 if (flags & BTRFS_INODE_APPEND)
118 iflags |= FS_APPEND_FL;
119 if (flags & BTRFS_INODE_NODUMP)
120 iflags |= FS_NODUMP_FL;
121 if (flags & BTRFS_INODE_NOATIME)
122 iflags |= FS_NOATIME_FL;
123 if (flags & BTRFS_INODE_DIRSYNC)
124 iflags |= FS_DIRSYNC_FL;
Li Zefand0092bd2011-04-15 03:03:06 +0000125 if (flags & BTRFS_INODE_NODATACOW)
126 iflags |= FS_NOCOW_FL;
127
Satoru Takeuchi13f48dc2016-03-15 09:09:59 +0900128 if (flags & BTRFS_INODE_NOCOMPRESS)
Li Zefand0092bd2011-04-15 03:03:06 +0000129 iflags |= FS_NOCOMP_FL;
Satoru Takeuchi13f48dc2016-03-15 09:09:59 +0900130 else if (flags & BTRFS_INODE_COMPRESS)
131 iflags |= FS_COMPR_FL;
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200132
133 return iflags;
134}
135
136/*
137 * Update inode->i_flags based on the btrfs internal flags.
138 */
139void btrfs_update_iflags(struct inode *inode)
140{
141 struct btrfs_inode *ip = BTRFS_I(inode);
Filipe Manana3cc79392014-06-25 22:36:02 +0100142 unsigned int new_fl = 0;
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200143
144 if (ip->flags & BTRFS_INODE_SYNC)
Filipe Manana3cc79392014-06-25 22:36:02 +0100145 new_fl |= S_SYNC;
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200146 if (ip->flags & BTRFS_INODE_IMMUTABLE)
Filipe Manana3cc79392014-06-25 22:36:02 +0100147 new_fl |= S_IMMUTABLE;
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200148 if (ip->flags & BTRFS_INODE_APPEND)
Filipe Manana3cc79392014-06-25 22:36:02 +0100149 new_fl |= S_APPEND;
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200150 if (ip->flags & BTRFS_INODE_NOATIME)
Filipe Manana3cc79392014-06-25 22:36:02 +0100151 new_fl |= S_NOATIME;
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200152 if (ip->flags & BTRFS_INODE_DIRSYNC)
Filipe Manana3cc79392014-06-25 22:36:02 +0100153 new_fl |= S_DIRSYNC;
154
155 set_mask_bits(&inode->i_flags,
156 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC,
157 new_fl);
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200158}
159
160/*
161 * Inherit flags from the parent inode.
162 *
Josef Bacike27425d2011-09-27 11:01:30 -0400163 * Currently only the compression flags and the cow flags are inherited.
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200164 */
165void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
166{
Chris Mason0b4dcea2009-06-11 11:13:35 -0400167 unsigned int flags;
168
169 if (!dir)
170 return;
171
172 flags = BTRFS_I(dir)->flags;
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200173
Josef Bacike27425d2011-09-27 11:01:30 -0400174 if (flags & BTRFS_INODE_NOCOMPRESS) {
175 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
176 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
177 } else if (flags & BTRFS_INODE_COMPRESS) {
178 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
179 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
180 }
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200181
Liu Bo213490b2012-09-11 08:33:50 -0600182 if (flags & BTRFS_INODE_NODATACOW) {
Josef Bacike27425d2011-09-27 11:01:30 -0400183 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
Liu Bo213490b2012-09-11 08:33:50 -0600184 if (S_ISREG(inode->i_mode))
185 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
186 }
Josef Bacike27425d2011-09-27 11:01:30 -0400187
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200188 btrfs_update_iflags(inode);
189}
190
191static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
192{
Al Viro496ad9a2013-01-23 17:07:38 -0500193 struct btrfs_inode *ip = BTRFS_I(file_inode(file));
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200194 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
195
196 if (copy_to_user(arg, &flags, sizeof(flags)))
197 return -EFAULT;
198 return 0;
199}
200
Liu Bo75e7cb72011-03-22 10:12:20 +0000201static int check_flags(unsigned int flags)
202{
203 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
204 FS_NOATIME_FL | FS_NODUMP_FL | \
205 FS_SYNC_FL | FS_DIRSYNC_FL | \
Li Zefane1e8fb62011-04-15 03:02:49 +0000206 FS_NOCOMP_FL | FS_COMPR_FL |
207 FS_NOCOW_FL))
Liu Bo75e7cb72011-03-22 10:12:20 +0000208 return -EOPNOTSUPP;
209
210 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
211 return -EINVAL;
212
Liu Bo75e7cb72011-03-22 10:12:20 +0000213 return 0;
214}
215
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200216static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
217{
Al Viro496ad9a2013-01-23 17:07:38 -0500218 struct inode *inode = file_inode(file);
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200219 struct btrfs_inode *ip = BTRFS_I(inode);
220 struct btrfs_root *root = ip->root;
221 struct btrfs_trans_handle *trans;
222 unsigned int flags, oldflags;
223 int ret;
Li Zefanf062abf02011-12-29 13:36:45 +0800224 u64 ip_oldflags;
225 unsigned int i_oldflags;
David Sterba7e97b8d2012-09-07 05:56:55 -0600226 umode_t mode;
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200227
David Sterbabd60ea02014-01-16 15:50:22 +0100228 if (!inode_owner_or_capable(inode))
229 return -EPERM;
230
Li Zefanb83cc962010-12-20 16:04:08 +0800231 if (btrfs_root_readonly(root))
232 return -EROFS;
233
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200234 if (copy_from_user(&flags, arg, sizeof(flags)))
235 return -EFAULT;
236
Liu Bo75e7cb72011-03-22 10:12:20 +0000237 ret = check_flags(flags);
238 if (ret)
239 return ret;
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200240
Jan Karae7848682012-06-12 16:20:32 +0200241 ret = mnt_want_write_file(file);
242 if (ret)
243 return ret;
244
Al Viro59551022016-01-22 15:40:57 -0500245 inode_lock(inode);
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200246
Li Zefanf062abf02011-12-29 13:36:45 +0800247 ip_oldflags = ip->flags;
248 i_oldflags = inode->i_flags;
David Sterba7e97b8d2012-09-07 05:56:55 -0600249 mode = inode->i_mode;
Li Zefanf062abf02011-12-29 13:36:45 +0800250
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200251 flags = btrfs_mask_flags(inode->i_mode, flags);
252 oldflags = btrfs_flags_to_ioctl(ip->flags);
253 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
254 if (!capable(CAP_LINUX_IMMUTABLE)) {
255 ret = -EPERM;
256 goto out_unlock;
257 }
258 }
259
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200260 if (flags & FS_SYNC_FL)
261 ip->flags |= BTRFS_INODE_SYNC;
262 else
263 ip->flags &= ~BTRFS_INODE_SYNC;
264 if (flags & FS_IMMUTABLE_FL)
265 ip->flags |= BTRFS_INODE_IMMUTABLE;
266 else
267 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
268 if (flags & FS_APPEND_FL)
269 ip->flags |= BTRFS_INODE_APPEND;
270 else
271 ip->flags &= ~BTRFS_INODE_APPEND;
272 if (flags & FS_NODUMP_FL)
273 ip->flags |= BTRFS_INODE_NODUMP;
274 else
275 ip->flags &= ~BTRFS_INODE_NODUMP;
276 if (flags & FS_NOATIME_FL)
277 ip->flags |= BTRFS_INODE_NOATIME;
278 else
279 ip->flags &= ~BTRFS_INODE_NOATIME;
280 if (flags & FS_DIRSYNC_FL)
281 ip->flags |= BTRFS_INODE_DIRSYNC;
282 else
283 ip->flags &= ~BTRFS_INODE_DIRSYNC;
David Sterba7e97b8d2012-09-07 05:56:55 -0600284 if (flags & FS_NOCOW_FL) {
285 if (S_ISREG(mode)) {
286 /*
287 * It's safe to turn csums off here, no extents exist.
288 * Otherwise we want the flag to reflect the real COW
289 * status of the file and will not set it.
290 */
291 if (inode->i_size == 0)
292 ip->flags |= BTRFS_INODE_NODATACOW
293 | BTRFS_INODE_NODATASUM;
294 } else {
295 ip->flags |= BTRFS_INODE_NODATACOW;
296 }
297 } else {
298 /*
Nicholas D Steeves01327612016-05-19 21:18:45 -0400299 * Revert back under same assumptions as above
David Sterba7e97b8d2012-09-07 05:56:55 -0600300 */
301 if (S_ISREG(mode)) {
302 if (inode->i_size == 0)
303 ip->flags &= ~(BTRFS_INODE_NODATACOW
304 | BTRFS_INODE_NODATASUM);
305 } else {
306 ip->flags &= ~BTRFS_INODE_NODATACOW;
307 }
308 }
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200309
Liu Bo75e7cb72011-03-22 10:12:20 +0000310 /*
311 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
312 * flag may be changed automatically if compression code won't make
313 * things smaller.
314 */
315 if (flags & FS_NOCOMP_FL) {
316 ip->flags &= ~BTRFS_INODE_COMPRESS;
317 ip->flags |= BTRFS_INODE_NOCOMPRESS;
Filipe David Borba Manana63541922014-01-07 11:47:46 +0000318
319 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
320 if (ret && ret != -ENODATA)
321 goto out_drop;
Liu Bo75e7cb72011-03-22 10:12:20 +0000322 } else if (flags & FS_COMPR_FL) {
Filipe David Borba Manana63541922014-01-07 11:47:46 +0000323 const char *comp;
324
Liu Bo75e7cb72011-03-22 10:12:20 +0000325 ip->flags |= BTRFS_INODE_COMPRESS;
326 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
Filipe David Borba Manana63541922014-01-07 11:47:46 +0000327
328 if (root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
329 comp = "lzo";
330 else
331 comp = "zlib";
332 ret = btrfs_set_prop(inode, "btrfs.compression",
333 comp, strlen(comp), 0);
334 if (ret)
335 goto out_drop;
336
Li Zefanebcb9042011-04-15 03:03:17 +0000337 } else {
Filipe Manana78a017a2014-09-11 11:44:49 +0100338 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
339 if (ret && ret != -ENODATA)
340 goto out_drop;
Li Zefanebcb9042011-04-15 03:03:17 +0000341 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
Liu Bo75e7cb72011-03-22 10:12:20 +0000342 }
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200343
Li Zefan4da6f1a2011-12-29 13:39:50 +0800344 trans = btrfs_start_transaction(root, 1);
Li Zefanf062abf02011-12-29 13:36:45 +0800345 if (IS_ERR(trans)) {
346 ret = PTR_ERR(trans);
347 goto out_drop;
348 }
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200349
Li Zefan306424cc2011-12-14 20:12:02 -0500350 btrfs_update_iflags(inode);
Josef Bacik0c4d2d92012-04-05 15:03:02 -0400351 inode_inc_iversion(inode);
Deepa Dinamani04b285f2016-02-06 23:57:21 -0800352 inode->i_ctime = current_fs_time(inode->i_sb);
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200353 ret = btrfs_update_inode(trans, root, inode);
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200354
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200355 btrfs_end_transaction(trans, root);
Li Zefanf062abf02011-12-29 13:36:45 +0800356 out_drop:
357 if (ret) {
358 ip->flags = ip_oldflags;
359 inode->i_flags = i_oldflags;
360 }
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200361
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200362 out_unlock:
Al Viro59551022016-01-22 15:40:57 -0500363 inode_unlock(inode);
Jan Karae7848682012-06-12 16:20:32 +0200364 mnt_drop_write_file(file);
liubo2d4e6f6ad2011-02-24 09:38:16 +0000365 return ret;
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200366}
367
368static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
369{
Al Viro496ad9a2013-01-23 17:07:38 -0500370 struct inode *inode = file_inode(file);
Christoph Hellwig6cbff002009-04-17 10:37:41 +0200371
372 return put_user(inode->i_generation, arg);
373}
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400374
Li Dongyangf7039b12011-03-24 10:24:28 +0000375static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
376{
Al Viro54563d42013-09-01 15:57:51 -0400377 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
Li Dongyangf7039b12011-03-24 10:24:28 +0000378 struct btrfs_device *device;
379 struct request_queue *q;
380 struct fstrim_range range;
381 u64 minlen = ULLONG_MAX;
382 u64 num_devices = 0;
Al Viro815745c2011-11-17 15:40:49 -0500383 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
Li Dongyangf7039b12011-03-24 10:24:28 +0000384 int ret;
385
386 if (!capable(CAP_SYS_ADMIN))
387 return -EPERM;
388
Xiao Guangrong1f781602011-04-20 10:09:16 +0000389 rcu_read_lock();
390 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
391 dev_list) {
Li Dongyangf7039b12011-03-24 10:24:28 +0000392 if (!device->bdev)
393 continue;
394 q = bdev_get_queue(device->bdev);
395 if (blk_queue_discard(q)) {
396 num_devices++;
397 minlen = min((u64)q->limits.discard_granularity,
398 minlen);
399 }
400 }
Xiao Guangrong1f781602011-04-20 10:09:16 +0000401 rcu_read_unlock();
Lukas Czernerf4c697e2011-09-05 16:34:54 +0200402
Li Dongyangf7039b12011-03-24 10:24:28 +0000403 if (!num_devices)
404 return -EOPNOTSUPP;
Li Dongyangf7039b12011-03-24 10:24:28 +0000405 if (copy_from_user(&range, arg, sizeof(range)))
406 return -EFAULT;
Lukas Czernere515c182012-10-16 09:34:36 +0000407 if (range.start > total_bytes ||
408 range.len < fs_info->sb->s_blocksize)
Lukas Czernerf4c697e2011-09-05 16:34:54 +0200409 return -EINVAL;
Li Dongyangf7039b12011-03-24 10:24:28 +0000410
Lukas Czernerf4c697e2011-09-05 16:34:54 +0200411 range.len = min(range.len, total_bytes - range.start);
Li Dongyangf7039b12011-03-24 10:24:28 +0000412 range.minlen = max(range.minlen, minlen);
Al Viro815745c2011-11-17 15:40:49 -0500413 ret = btrfs_trim_fs(fs_info->tree_root, &range);
Li Dongyangf7039b12011-03-24 10:24:28 +0000414 if (ret < 0)
415 return ret;
416
417 if (copy_to_user(arg, &range, sizeof(range)))
418 return -EFAULT;
419
420 return 0;
421}
422
Stefan Behrensdd5f9612013-08-15 17:11:20 +0200423int btrfs_is_empty_uuid(u8 *uuid)
424{
Chris Mason46e0f662013-11-15 12:14:55 +0100425 int i;
426
427 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
428 if (uuid[i])
429 return 0;
430 }
431 return 1;
Stefan Behrensdd5f9612013-08-15 17:11:20 +0200432}
433
Miao Xied5c12072013-02-28 10:04:33 +0000434static noinline int create_subvol(struct inode *dir,
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400435 struct dentry *dentry,
Sage Weil72fd0322010-10-29 15:41:32 -0400436 char *name, int namelen,
Arne Jansen6f72c7e2011-09-14 15:58:21 +0200437 u64 *async_transid,
Miao Xie8696c532013-02-07 06:02:44 +0000438 struct btrfs_qgroup_inherit *inherit)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400439{
440 struct btrfs_trans_handle *trans;
441 struct btrfs_key key;
David Sterba49a3c4d2016-03-24 17:49:22 +0100442 struct btrfs_root_item *root_item;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400443 struct btrfs_inode_item *inode_item;
444 struct extent_buffer *leaf;
Miao Xied5c12072013-02-28 10:04:33 +0000445 struct btrfs_root *root = BTRFS_I(dir)->root;
Yan, Zheng76dda932009-09-21 16:00:26 -0400446 struct btrfs_root *new_root;
Miao Xied5c12072013-02-28 10:04:33 +0000447 struct btrfs_block_rsv block_rsv;
Deepa Dinamani04b285f2016-02-06 23:57:21 -0800448 struct timespec cur_time = current_fs_time(dir->i_sb);
Tsutomu Itoh5662344b32013-12-13 09:51:42 +0900449 struct inode *inode;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400450 int ret;
451 int err;
452 u64 objectid;
453 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
Chris Mason3de45862008-11-17 21:02:50 -0500454 u64 index = 0;
Miao Xied5c12072013-02-28 10:04:33 +0000455 u64 qgroup_reserved;
Alexander Block8ea05e32012-07-25 17:35:53 +0200456 uuid_le new_uuid;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400457
David Sterba49a3c4d2016-03-24 17:49:22 +0100458 root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
459 if (!root_item)
460 return -ENOMEM;
461
Li Zefan581bb052011-04-20 10:06:11 +0800462 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
Al Viro2fbe8c82011-07-16 21:38:06 -0400463 if (ret)
David Sterba49a3c4d2016-03-24 17:49:22 +0100464 goto fail_free;
Josef Bacik6a912212010-11-20 09:48:00 +0000465
Qu Wenruoe09fe2d2015-02-27 16:24:23 +0800466 /*
467 * Don't create subvolume whose level is not zero. Or qgroup will be
Nicholas D Steeves01327612016-05-19 21:18:45 -0400468 * screwed up since it assumes subvolume qgroup's level to be 0.
Qu Wenruoe09fe2d2015-02-27 16:24:23 +0800469 */
David Sterba49a3c4d2016-03-24 17:49:22 +0100470 if (btrfs_qgroup_level(objectid)) {
471 ret = -ENOSPC;
472 goto fail_free;
473 }
Qu Wenruoe09fe2d2015-02-27 16:24:23 +0800474
Miao Xied5c12072013-02-28 10:04:33 +0000475 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
Josef Bacik9ed74f22009-09-11 16:12:44 -0400476 /*
Miao Xied5c12072013-02-28 10:04:33 +0000477 * The same as the snapshot creation, please see the comment
478 * of create_snapshot().
Josef Bacik9ed74f22009-09-11 16:12:44 -0400479 */
Miao Xied5c12072013-02-28 10:04:33 +0000480 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
Stefan Behrensdd5f9612013-08-15 17:11:20 +0200481 8, &qgroup_reserved, false);
Miao Xied5c12072013-02-28 10:04:33 +0000482 if (ret)
David Sterba49a3c4d2016-03-24 17:49:22 +0100483 goto fail_free;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400484
Miao Xied5c12072013-02-28 10:04:33 +0000485 trans = btrfs_start_transaction(root, 0);
486 if (IS_ERR(trans)) {
487 ret = PTR_ERR(trans);
Liu Bode6e8202014-01-09 14:57:06 +0800488 btrfs_subvolume_release_metadata(root, &block_rsv,
489 qgroup_reserved);
David Sterba49a3c4d2016-03-24 17:49:22 +0100490 goto fail_free;
Miao Xied5c12072013-02-28 10:04:33 +0000491 }
492 trans->block_rsv = &block_rsv;
493 trans->bytes_reserved = block_rsv.size;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400494
Miao Xie8696c532013-02-07 06:02:44 +0000495 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
Arne Jansen6f72c7e2011-09-14 15:58:21 +0200496 if (ret)
497 goto fail;
498
David Sterba4d75f8a2014-06-15 01:54:12 +0200499 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
Josef Bacik8e8a1e32008-07-24 12:17:14 -0400500 if (IS_ERR(leaf)) {
501 ret = PTR_ERR(leaf);
502 goto fail;
503 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400504
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400505 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400506 btrfs_set_header_bytenr(leaf, leaf->start);
507 btrfs_set_header_generation(leaf, trans->transid);
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400508 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400509 btrfs_set_header_owner(leaf, objectid);
510
Ross Kirk0a4e5582013-09-24 10:12:38 +0100511 write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(),
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400512 BTRFS_FSID_SIZE);
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400513 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
Geert Uytterhoevenb308bc22013-08-20 13:20:15 +0200514 btrfs_header_chunk_tree_uuid(leaf),
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400515 BTRFS_UUID_SIZE);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400516 btrfs_mark_buffer_dirty(leaf);
517
David Sterba49a3c4d2016-03-24 17:49:22 +0100518 inode_item = &root_item->inode;
Qu Wenruo3cae2102013-07-16 11:19:18 +0800519 btrfs_set_stack_inode_generation(inode_item, 1);
520 btrfs_set_stack_inode_size(inode_item, 3);
521 btrfs_set_stack_inode_nlink(inode_item, 1);
David Sterba707e8a02014-06-04 19:22:26 +0200522 btrfs_set_stack_inode_nbytes(inode_item, root->nodesize);
Qu Wenruo3cae2102013-07-16 11:19:18 +0800523 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400524
David Sterba49a3c4d2016-03-24 17:49:22 +0100525 btrfs_set_root_flags(root_item, 0);
526 btrfs_set_root_limit(root_item, 0);
Qu Wenruo3cae2102013-07-16 11:19:18 +0800527 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
Li Zefan08fe4db2011-03-28 02:01:25 +0000528
David Sterba49a3c4d2016-03-24 17:49:22 +0100529 btrfs_set_root_bytenr(root_item, leaf->start);
530 btrfs_set_root_generation(root_item, trans->transid);
531 btrfs_set_root_level(root_item, 0);
532 btrfs_set_root_refs(root_item, 1);
533 btrfs_set_root_used(root_item, leaf->len);
534 btrfs_set_root_last_snapshot(root_item, 0);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400535
David Sterba49a3c4d2016-03-24 17:49:22 +0100536 btrfs_set_root_generation_v2(root_item,
537 btrfs_root_generation(root_item));
Alexander Block8ea05e32012-07-25 17:35:53 +0200538 uuid_le_gen(&new_uuid);
David Sterba49a3c4d2016-03-24 17:49:22 +0100539 memcpy(root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE);
540 btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
541 btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
542 root_item->ctime = root_item->otime;
543 btrfs_set_root_ctransid(root_item, trans->transid);
544 btrfs_set_root_otransid(root_item, trans->transid);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400545
Chris Mason925baed2008-06-25 16:01:30 -0400546 btrfs_tree_unlock(leaf);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400547 free_extent_buffer(leaf);
548 leaf = NULL;
549
David Sterba49a3c4d2016-03-24 17:49:22 +0100550 btrfs_set_root_dirid(root_item, new_dirid);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400551
552 key.objectid = objectid;
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400553 key.offset = 0;
David Sterba962a2982014-06-04 18:41:45 +0200554 key.type = BTRFS_ROOT_ITEM_KEY;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400555 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
David Sterba49a3c4d2016-03-24 17:49:22 +0100556 root_item);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400557 if (ret)
558 goto fail;
559
Yan, Zheng76dda932009-09-21 16:00:26 -0400560 key.offset = (u64)-1;
561 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100562 if (IS_ERR(new_root)) {
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100563 ret = PTR_ERR(new_root);
Jeff Mahoney66642832016-06-10 18:19:25 -0400564 btrfs_abort_transaction(trans, ret);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100565 goto fail;
566 }
Yan, Zheng76dda932009-09-21 16:00:26 -0400567
568 btrfs_record_root_in_trans(trans, new_root);
569
Filipe David Borba Manana63541922014-01-07 11:47:46 +0000570 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
Mark Fashehce598972011-07-26 11:32:23 -0700571 if (ret) {
572 /* We potentially lose an unused inode item here */
Jeff Mahoney66642832016-06-10 18:19:25 -0400573 btrfs_abort_transaction(trans, ret);
Mark Fashehce598972011-07-26 11:32:23 -0700574 goto fail;
575 }
576
Chandan Rajendraf32e48e2016-01-07 18:56:59 +0530577 mutex_lock(&new_root->objectid_mutex);
578 new_root->highest_objectid = new_dirid;
579 mutex_unlock(&new_root->objectid_mutex);
580
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400581 /*
582 * insert the directory item
583 */
Chris Mason3de45862008-11-17 21:02:50 -0500584 ret = btrfs_set_inode_index(dir, &index);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100585 if (ret) {
Jeff Mahoney66642832016-06-10 18:19:25 -0400586 btrfs_abort_transaction(trans, ret);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100587 goto fail;
588 }
Chris Mason3de45862008-11-17 21:02:50 -0500589
590 ret = btrfs_insert_dir_item(trans, root,
Miao Xie16cdcec2011-04-22 18:12:22 +0800591 name, namelen, dir, &key,
Chris Mason3de45862008-11-17 21:02:50 -0500592 BTRFS_FT_DIR, index);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100593 if (ret) {
Jeff Mahoney66642832016-06-10 18:19:25 -0400594 btrfs_abort_transaction(trans, ret);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400595 goto fail;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100596 }
Chris Mason0660b5a2008-11-17 20:37:39 -0500597
Yan Zheng52c26172009-01-05 15:43:43 -0500598 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
599 ret = btrfs_update_inode(trans, root, dir);
600 BUG_ON(ret);
601
Chris Mason0660b5a2008-11-17 20:37:39 -0500602 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
Yan, Zheng4df27c4d2009-09-21 15:56:00 -0400603 objectid, root->root_key.objectid,
Li Zefan33345d012011-04-20 10:31:50 +0800604 btrfs_ino(dir), index, name, namelen);
Chris Mason0660b5a2008-11-17 20:37:39 -0500605 BUG_ON(ret);
606
Stefan Behrensdd5f9612013-08-15 17:11:20 +0200607 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
David Sterba49a3c4d2016-03-24 17:49:22 +0100608 root_item->uuid, BTRFS_UUID_KEY_SUBVOL,
Stefan Behrensdd5f9612013-08-15 17:11:20 +0200609 objectid);
610 if (ret)
Jeff Mahoney66642832016-06-10 18:19:25 -0400611 btrfs_abort_transaction(trans, ret);
Stefan Behrensdd5f9612013-08-15 17:11:20 +0200612
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400613fail:
David Sterba49a3c4d2016-03-24 17:49:22 +0100614 kfree(root_item);
Miao Xied5c12072013-02-28 10:04:33 +0000615 trans->block_rsv = NULL;
616 trans->bytes_reserved = 0;
Liu Bode6e8202014-01-09 14:57:06 +0800617 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
618
Sage Weil72fd0322010-10-29 15:41:32 -0400619 if (async_transid) {
620 *async_transid = trans->transid;
621 err = btrfs_commit_transaction_async(trans, root, 1);
Miao Xie00d71c92013-03-04 09:45:06 +0000622 if (err)
623 err = btrfs_commit_transaction(trans, root);
Sage Weil72fd0322010-10-29 15:41:32 -0400624 } else {
625 err = btrfs_commit_transaction(trans, root);
626 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400627 if (err && !ret)
628 ret = err;
Chris Mason1a65e242013-02-06 12:06:02 -0500629
Tsutomu Itoh5662344b32013-12-13 09:51:42 +0900630 if (!ret) {
631 inode = btrfs_lookup_dentry(dir, dentry);
Liu Bode6e8202014-01-09 14:57:06 +0800632 if (IS_ERR(inode))
633 return PTR_ERR(inode);
Tsutomu Itoh5662344b32013-12-13 09:51:42 +0900634 d_instantiate(dentry, inode);
635 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400636 return ret;
David Sterba49a3c4d2016-03-24 17:49:22 +0100637
638fail_free:
639 kfree(root_item);
640 return ret;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400641}
642
Filipe Manana9ea24bb2014-10-29 11:57:59 +0000643static void btrfs_wait_for_no_snapshoting_writes(struct btrfs_root *root)
Miao Xie8257b2d2014-03-06 13:38:19 +0800644{
645 s64 writers;
646 DEFINE_WAIT(wait);
647
648 do {
649 prepare_to_wait(&root->subv_writers->wait, &wait,
650 TASK_UNINTERRUPTIBLE);
651
652 writers = percpu_counter_sum(&root->subv_writers->counter);
653 if (writers)
654 schedule();
655
656 finish_wait(&root->subv_writers->wait, &wait);
657 } while (writers);
658}
659
Miao Xiee9662f72013-02-28 10:01:15 +0000660static int create_snapshot(struct btrfs_root *root, struct inode *dir,
661 struct dentry *dentry, char *name, int namelen,
662 u64 *async_transid, bool readonly,
663 struct btrfs_qgroup_inherit *inherit)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400664{
Yan, Zheng2e4bfab2009-11-12 09:37:02 +0000665 struct inode *inode;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400666 struct btrfs_pending_snapshot *pending_snapshot;
667 struct btrfs_trans_handle *trans;
Yan, Zheng2e4bfab2009-11-12 09:37:02 +0000668 int ret;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400669
Miao Xie27cdeb72014-04-02 19:51:05 +0800670 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400671 return -EINVAL;
672
David Sterbaa1ee7362015-11-10 18:53:56 +0100673 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
674 if (!pending_snapshot)
675 return -ENOMEM;
676
David Sterbab0c0ea62015-11-10 18:54:00 +0100677 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
678 GFP_NOFS);
David Sterba8546b572015-11-10 18:54:03 +0100679 pending_snapshot->path = btrfs_alloc_path();
680 if (!pending_snapshot->root_item || !pending_snapshot->path) {
David Sterbab0c0ea62015-11-10 18:54:00 +0100681 ret = -ENOMEM;
682 goto free_pending;
683 }
684
Miao Xie8257b2d2014-03-06 13:38:19 +0800685 atomic_inc(&root->will_be_snapshoted);
Peter Zijlstra4e857c52014-03-17 18:06:10 +0100686 smp_mb__after_atomic();
Filipe Manana9ea24bb2014-10-29 11:57:59 +0000687 btrfs_wait_for_no_snapshoting_writes(root);
Miao Xie8257b2d2014-03-06 13:38:19 +0800688
Miao Xie6a038432013-05-15 07:48:24 +0000689 ret = btrfs_start_delalloc_inodes(root, 0);
690 if (ret)
David Sterbaa1ee7362015-11-10 18:53:56 +0100691 goto dec_and_free;
Miao Xie6a038432013-05-15 07:48:24 +0000692
Filipe Manana578def72016-04-26 15:36:38 +0100693 btrfs_wait_ordered_extents(root, -1, 0, (u64)-1);
Miao Xie6a038432013-05-15 07:48:24 +0000694
Miao Xie66d8f3d2012-09-06 04:02:28 -0600695 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
696 BTRFS_BLOCK_RSV_TEMP);
Miao Xied5c12072013-02-28 10:04:33 +0000697 /*
698 * 1 - parent dir inode
699 * 2 - dir entries
700 * 1 - root item
701 * 2 - root ref/backref
702 * 1 - root of snapshot
Stefan Behrensdd5f9612013-08-15 17:11:20 +0200703 * 1 - UUID item
Miao Xied5c12072013-02-28 10:04:33 +0000704 */
705 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
Stefan Behrensdd5f9612013-08-15 17:11:20 +0200706 &pending_snapshot->block_rsv, 8,
Jeff Mahoneyee3441b2013-07-09 16:37:21 -0400707 &pending_snapshot->qgroup_reserved,
708 false);
Miao Xied5c12072013-02-28 10:04:33 +0000709 if (ret)
David Sterbaa1ee7362015-11-10 18:53:56 +0100710 goto dec_and_free;
Miao Xied5c12072013-02-28 10:04:33 +0000711
Yan, Zhenga22285a2010-05-16 10:48:46 -0400712 pending_snapshot->dentry = dentry;
713 pending_snapshot->root = root;
Li Zefanb83cc962010-12-20 16:04:08 +0800714 pending_snapshot->readonly = readonly;
Miao Xiee9662f72013-02-28 10:01:15 +0000715 pending_snapshot->dir = dir;
Miao Xie8696c532013-02-07 06:02:44 +0000716 pending_snapshot->inherit = inherit;
Yan, Zhenga22285a2010-05-16 10:48:46 -0400717
Miao Xied5c12072013-02-28 10:04:33 +0000718 trans = btrfs_start_transaction(root, 0);
Yan, Zhenga22285a2010-05-16 10:48:46 -0400719 if (IS_ERR(trans)) {
720 ret = PTR_ERR(trans);
721 goto fail;
722 }
723
Josef Bacik83515832011-06-14 15:16:14 -0400724 spin_lock(&root->fs_info->trans_lock);
Yan, Zhenga22285a2010-05-16 10:48:46 -0400725 list_add(&pending_snapshot->list,
726 &trans->transaction->pending_snapshots);
Josef Bacik83515832011-06-14 15:16:14 -0400727 spin_unlock(&root->fs_info->trans_lock);
Sage Weil72fd0322010-10-29 15:41:32 -0400728 if (async_transid) {
729 *async_transid = trans->transid;
730 ret = btrfs_commit_transaction_async(trans,
731 root->fs_info->extent_root, 1);
Miao Xie00d71c92013-03-04 09:45:06 +0000732 if (ret)
733 ret = btrfs_commit_transaction(trans, root);
Sage Weil72fd0322010-10-29 15:41:32 -0400734 } else {
735 ret = btrfs_commit_transaction(trans,
736 root->fs_info->extent_root);
737 }
Miao Xieaec80302013-03-04 09:44:29 +0000738 if (ret)
Josef Bacikc37b2b62012-10-22 15:51:44 -0400739 goto fail;
Yan, Zhenga22285a2010-05-16 10:48:46 -0400740
741 ret = pending_snapshot->error;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400742 if (ret)
Yan, Zheng2e4bfab2009-11-12 09:37:02 +0000743 goto fail;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400744
Chris Masond3797302014-10-15 13:50:56 -0700745 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
746 if (ret)
747 goto fail;
748
David Howells2b0143b2015-03-17 22:25:59 +0000749 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
Yan, Zheng2e4bfab2009-11-12 09:37:02 +0000750 if (IS_ERR(inode)) {
751 ret = PTR_ERR(inode);
752 goto fail;
753 }
Tsutomu Itoh5662344b32013-12-13 09:51:42 +0900754
Yan, Zheng2e4bfab2009-11-12 09:37:02 +0000755 d_instantiate(dentry, inode);
756 ret = 0;
757fail:
Miao Xied5c12072013-02-28 10:04:33 +0000758 btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
759 &pending_snapshot->block_rsv,
760 pending_snapshot->qgroup_reserved);
David Sterbaa1ee7362015-11-10 18:53:56 +0100761dec_and_free:
Filipe Manana9ea24bb2014-10-29 11:57:59 +0000762 if (atomic_dec_and_test(&root->will_be_snapshoted))
763 wake_up_atomic_t(&root->will_be_snapshoted);
David Sterbab0c0ea62015-11-10 18:54:00 +0100764free_pending:
765 kfree(pending_snapshot->root_item);
David Sterba8546b572015-11-10 18:54:03 +0100766 btrfs_free_path(pending_snapshot->path);
David Sterbaa1ee7362015-11-10 18:53:56 +0100767 kfree(pending_snapshot);
768
Christoph Hellwigf46b5a62008-06-11 21:53:53 -0400769 return ret;
770}
771
Sage Weil4260f7c2010-10-29 15:46:43 -0400772/* copy of may_delete in fs/namei.c()
773 * Check whether we can remove a link victim from directory dir, check
774 * whether the type of victim is right.
775 * 1. We can't do it if dir is read-only (done in permission())
776 * 2. We should have write and exec permissions on dir
777 * 3. We can't remove anything from append-only dir
778 * 4. We can't do anything with immutable dir (done in permission())
779 * 5. If the sticky bit on dir is set we should either
780 * a. be owner of dir, or
781 * b. be owner of victim, or
782 * c. have CAP_FOWNER capability
Nicholas D Steeves01327612016-05-19 21:18:45 -0400783 * 6. If the victim is append-only or immutable we can't do anything with
Sage Weil4260f7c2010-10-29 15:46:43 -0400784 * links pointing to it.
785 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
786 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
787 * 9. We can't remove a root or mountpoint.
788 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
789 * nfs_async_unlink().
790 */
791
Dulshani Gunawardhana67871252013-10-31 10:33:04 +0530792static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
Sage Weil4260f7c2010-10-29 15:46:43 -0400793{
794 int error;
795
David Howells2b0143b2015-03-17 22:25:59 +0000796 if (d_really_is_negative(victim))
Sage Weil4260f7c2010-10-29 15:46:43 -0400797 return -ENOENT;
798
David Howells2b0143b2015-03-17 22:25:59 +0000799 BUG_ON(d_inode(victim->d_parent) != dir);
Jeff Layton4fa6b5e2012-10-10 15:25:25 -0400800 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
Sage Weil4260f7c2010-10-29 15:46:43 -0400801
802 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
803 if (error)
804 return error;
805 if (IS_APPEND(dir))
806 return -EPERM;
David Howells2b0143b2015-03-17 22:25:59 +0000807 if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) ||
808 IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim)))
Sage Weil4260f7c2010-10-29 15:46:43 -0400809 return -EPERM;
810 if (isdir) {
David Howellse36cb0b2015-01-29 12:02:35 +0000811 if (!d_is_dir(victim))
Sage Weil4260f7c2010-10-29 15:46:43 -0400812 return -ENOTDIR;
813 if (IS_ROOT(victim))
814 return -EBUSY;
David Howellse36cb0b2015-01-29 12:02:35 +0000815 } else if (d_is_dir(victim))
Sage Weil4260f7c2010-10-29 15:46:43 -0400816 return -EISDIR;
817 if (IS_DEADDIR(dir))
818 return -ENOENT;
819 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
820 return -EBUSY;
821 return 0;
822}
823
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400824/* copy of may_create in fs/namei.c() */
825static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
826{
David Howells2b0143b2015-03-17 22:25:59 +0000827 if (d_really_is_positive(child))
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400828 return -EEXIST;
829 if (IS_DEADDIR(dir))
830 return -ENOENT;
831 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
832}
833
834/*
835 * Create a new subvolume below @parent. This is largely modeled after
836 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
837 * inside this filesystem so it's quite a bit simpler.
838 */
Yan, Zheng76dda932009-09-21 16:00:26 -0400839static noinline int btrfs_mksubvol(struct path *parent,
840 char *name, int namelen,
Sage Weil72fd0322010-10-29 15:41:32 -0400841 struct btrfs_root *snap_src,
Arne Jansen6f72c7e2011-09-14 15:58:21 +0200842 u64 *async_transid, bool readonly,
Miao Xie8696c532013-02-07 06:02:44 +0000843 struct btrfs_qgroup_inherit *inherit)
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400844{
David Howells2b0143b2015-03-17 22:25:59 +0000845 struct inode *dir = d_inode(parent->dentry);
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400846 struct dentry *dentry;
847 int error;
848
Al Viro00235412016-05-26 00:05:12 -0400849 error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
850 if (error == -EINTR)
851 return error;
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400852
853 dentry = lookup_one_len(name, parent->dentry, namelen);
854 error = PTR_ERR(dentry);
855 if (IS_ERR(dentry))
856 goto out_unlock;
857
Yan, Zheng76dda932009-09-21 16:00:26 -0400858 error = btrfs_may_create(dir, dentry);
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400859 if (error)
Liu Boa874a632012-06-29 03:58:46 -0600860 goto out_dput;
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400861
Chris Mason9c520572012-12-17 14:26:57 -0500862 /*
863 * even if this name doesn't exist, we may get hash collisions.
864 * check for them now when we can safely fail
865 */
866 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
867 dir->i_ino, name,
868 namelen);
869 if (error)
870 goto out_dput;
871
Yan, Zheng76dda932009-09-21 16:00:26 -0400872 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
873
874 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
875 goto out_up_read;
876
Chris Mason3de45862008-11-17 21:02:50 -0500877 if (snap_src) {
Miao Xiee9662f72013-02-28 10:01:15 +0000878 error = create_snapshot(snap_src, dir, dentry, name, namelen,
Arne Jansen6f72c7e2011-09-14 15:58:21 +0200879 async_transid, readonly, inherit);
Chris Mason3de45862008-11-17 21:02:50 -0500880 } else {
Miao Xied5c12072013-02-28 10:04:33 +0000881 error = create_subvol(dir, dentry, name, namelen,
882 async_transid, inherit);
Chris Mason3de45862008-11-17 21:02:50 -0500883 }
Yan, Zheng76dda932009-09-21 16:00:26 -0400884 if (!error)
885 fsnotify_mkdir(dir, dentry);
886out_up_read:
887 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400888out_dput:
889 dput(dentry);
890out_unlock:
Al Viro59551022016-01-22 15:40:57 -0500891 inode_unlock(dir);
Christoph Hellwigcb8e7092008-10-09 13:39:39 -0400892 return error;
893}
894
Chris Mason4cb53002011-05-24 15:35:30 -0400895/*
896 * When we're defragging a range, we don't want to kick it off again
897 * if it is really just waiting for delalloc to send it down.
898 * If we find a nice big extent or delalloc range for the bytes in the
899 * file you want to defrag, we return 0 to let you know to skip this
900 * part of the file
901 */
David Sterbaaab110a2014-07-29 17:32:10 +0200902static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
Chris Mason4cb53002011-05-24 15:35:30 -0400903{
904 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
905 struct extent_map *em = NULL;
906 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
907 u64 end;
908
909 read_lock(&em_tree->lock);
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +0300910 em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE);
Chris Mason4cb53002011-05-24 15:35:30 -0400911 read_unlock(&em_tree->lock);
912
913 if (em) {
914 end = extent_map_end(em);
915 free_extent_map(em);
916 if (end - offset > thresh)
917 return 0;
918 }
919 /* if we already have a nice delalloc here, just stop */
920 thresh /= 2;
921 end = count_range_bits(io_tree, &offset, offset + thresh,
922 thresh, EXTENT_DELALLOC, 1);
923 if (end >= thresh)
924 return 0;
925 return 1;
926}
927
928/*
929 * helper function to walk through a file and find extents
930 * newer than a specific transid, and smaller than thresh.
931 *
932 * This is used by the defragging code to find new and small
933 * extents
934 */
935static int find_new_extents(struct btrfs_root *root,
936 struct inode *inode, u64 newer_than,
David Sterbaaab110a2014-07-29 17:32:10 +0200937 u64 *off, u32 thresh)
Chris Mason4cb53002011-05-24 15:35:30 -0400938{
939 struct btrfs_path *path;
940 struct btrfs_key min_key;
Chris Mason4cb53002011-05-24 15:35:30 -0400941 struct extent_buffer *leaf;
942 struct btrfs_file_extent_item *extent;
943 int type;
944 int ret;
David Sterbaa4689d22011-05-31 17:08:14 +0000945 u64 ino = btrfs_ino(inode);
Chris Mason4cb53002011-05-24 15:35:30 -0400946
947 path = btrfs_alloc_path();
948 if (!path)
949 return -ENOMEM;
950
David Sterbaa4689d22011-05-31 17:08:14 +0000951 min_key.objectid = ino;
Chris Mason4cb53002011-05-24 15:35:30 -0400952 min_key.type = BTRFS_EXTENT_DATA_KEY;
953 min_key.offset = *off;
954
Dulshani Gunawardhana67871252013-10-31 10:33:04 +0530955 while (1) {
Filipe David Borba Manana6174d3c2013-10-01 16:13:42 +0100956 ret = btrfs_search_forward(root, &min_key, path, newer_than);
Chris Mason4cb53002011-05-24 15:35:30 -0400957 if (ret != 0)
958 goto none;
Filipe Mananaf094c9bd2014-03-12 01:28:24 +0000959process_slot:
David Sterbaa4689d22011-05-31 17:08:14 +0000960 if (min_key.objectid != ino)
Chris Mason4cb53002011-05-24 15:35:30 -0400961 goto none;
962 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
963 goto none;
964
965 leaf = path->nodes[0];
966 extent = btrfs_item_ptr(leaf, path->slots[0],
967 struct btrfs_file_extent_item);
968
969 type = btrfs_file_extent_type(leaf, extent);
970 if (type == BTRFS_FILE_EXTENT_REG &&
971 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
972 check_defrag_in_cache(inode, min_key.offset, thresh)) {
973 *off = min_key.offset;
974 btrfs_free_path(path);
975 return 0;
976 }
977
Filipe Mananaf094c9bd2014-03-12 01:28:24 +0000978 path->slots[0]++;
979 if (path->slots[0] < btrfs_header_nritems(leaf)) {
980 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
981 goto process_slot;
982 }
983
Chris Mason4cb53002011-05-24 15:35:30 -0400984 if (min_key.offset == (u64)-1)
985 goto none;
986
987 min_key.offset++;
988 btrfs_release_path(path);
989 }
990none:
991 btrfs_free_path(path);
992 return -ENOENT;
993}
994
Li Zefan6c282eb2012-06-11 16:03:35 +0800995static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
Liu Bo17ce6ef2012-03-29 09:57:45 -0400996{
997 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
Chris Mason940100a2010-03-10 10:52:59 -0500998 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
Li Zefan6c282eb2012-06-11 16:03:35 +0800999 struct extent_map *em;
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001000 u64 len = PAGE_SIZE;
Chris Mason940100a2010-03-10 10:52:59 -05001001
1002 /*
1003 * hopefully we have this extent in the tree already, try without
1004 * the full extent lock
1005 */
1006 read_lock(&em_tree->lock);
1007 em = lookup_extent_mapping(em_tree, start, len);
1008 read_unlock(&em_tree->lock);
1009
1010 if (!em) {
Filipe Manana308d9802014-03-11 13:56:15 +00001011 struct extent_state *cached = NULL;
1012 u64 end = start + len - 1;
1013
Chris Mason940100a2010-03-10 10:52:59 -05001014 /* get the big lock and read metadata off disk */
David Sterbaff13db42015-12-03 14:30:40 +01001015 lock_extent_bits(io_tree, start, end, &cached);
Chris Mason940100a2010-03-10 10:52:59 -05001016 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
Filipe Manana308d9802014-03-11 13:56:15 +00001017 unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
Chris Mason940100a2010-03-10 10:52:59 -05001018
Dan Carpenter6cf8bfb2010-03-20 11:22:10 +00001019 if (IS_ERR(em))
Li Zefan6c282eb2012-06-11 16:03:35 +08001020 return NULL;
Chris Mason940100a2010-03-10 10:52:59 -05001021 }
1022
Li Zefan6c282eb2012-06-11 16:03:35 +08001023 return em;
1024}
1025
1026static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1027{
1028 struct extent_map *next;
1029 bool ret = true;
1030
1031 /* this is the last extent */
1032 if (em->start + em->len >= i_size_read(inode))
1033 return false;
1034
1035 next = defrag_lookup_extent(inode, em->start + em->len);
Chris Masone9512d72014-08-26 13:55:54 -07001036 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1037 ret = false;
1038 else if ((em->block_start + em->block_len == next->block_start) &&
Byongho Leeee221842015-12-15 01:42:10 +09001039 (em->block_len > SZ_128K && next->block_len > SZ_128K))
Li Zefan6c282eb2012-06-11 16:03:35 +08001040 ret = false;
1041
1042 free_extent_map(next);
1043 return ret;
1044}
1045
David Sterbaaab110a2014-07-29 17:32:10 +02001046static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
Andrew Mahonea43a2112012-06-19 21:08:32 -04001047 u64 *last_len, u64 *skip, u64 *defrag_end,
1048 int compress)
Li Zefan6c282eb2012-06-11 16:03:35 +08001049{
1050 struct extent_map *em;
1051 int ret = 1;
1052 bool next_mergeable = true;
Liu Bo4a3560c2015-08-07 16:48:41 +08001053 bool prev_mergeable = true;
Li Zefan6c282eb2012-06-11 16:03:35 +08001054
1055 /*
1056 * make sure that once we start defragging an extent, we keep on
1057 * defragging it
1058 */
1059 if (start < *defrag_end)
1060 return 1;
1061
1062 *skip = 0;
1063
1064 em = defrag_lookup_extent(inode, start);
1065 if (!em)
1066 return 0;
1067
Chris Mason940100a2010-03-10 10:52:59 -05001068 /* this will cover holes, and inline extents */
Liu Bo17ce6ef2012-03-29 09:57:45 -04001069 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
Chris Mason940100a2010-03-10 10:52:59 -05001070 ret = 0;
Liu Bo17ce6ef2012-03-29 09:57:45 -04001071 goto out;
1072 }
1073
Liu Bo4a3560c2015-08-07 16:48:41 +08001074 if (!*defrag_end)
1075 prev_mergeable = false;
1076
Li Zefan6c282eb2012-06-11 16:03:35 +08001077 next_mergeable = defrag_check_next_extent(inode, em);
Chris Mason940100a2010-03-10 10:52:59 -05001078 /*
Li Zefan6c282eb2012-06-11 16:03:35 +08001079 * we hit a real extent, if it is big or the next extent is not a
1080 * real extent, don't bother defragging it
Chris Mason940100a2010-03-10 10:52:59 -05001081 */
Andrew Mahonea43a2112012-06-19 21:08:32 -04001082 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
Liu Bo4a3560c2015-08-07 16:48:41 +08001083 (em->len >= thresh || (!next_mergeable && !prev_mergeable)))
Chris Mason940100a2010-03-10 10:52:59 -05001084 ret = 0;
Liu Bo17ce6ef2012-03-29 09:57:45 -04001085out:
Chris Mason940100a2010-03-10 10:52:59 -05001086 /*
1087 * last_len ends up being a counter of how many bytes we've defragged.
1088 * every time we choose not to defrag an extent, we reset *last_len
1089 * so that the next tiny extent will force a defrag.
1090 *
1091 * The end result of this is that tiny extents before a single big
1092 * extent will force at least part of that big extent to be defragged.
1093 */
1094 if (ret) {
Chris Mason940100a2010-03-10 10:52:59 -05001095 *defrag_end = extent_map_end(em);
1096 } else {
1097 *last_len = 0;
1098 *skip = extent_map_end(em);
1099 *defrag_end = 0;
1100 }
1101
1102 free_extent_map(em);
1103 return ret;
1104}
1105
Chris Mason4cb53002011-05-24 15:35:30 -04001106/*
1107 * it doesn't do much good to defrag one or two pages
1108 * at a time. This pulls in a nice chunk of pages
1109 * to COW and defrag.
1110 *
1111 * It also makes sure the delalloc code has enough
1112 * dirty data to avoid making new small extents as part
1113 * of the defrag
1114 *
1115 * It's a good idea to start RA on this range
1116 * before calling this.
1117 */
1118static int cluster_pages_for_defrag(struct inode *inode,
1119 struct page **pages,
1120 unsigned long start_index,
Justin Maggardc41570c2014-01-21 11:18:29 -08001121 unsigned long num_pages)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001122{
Chris Mason4cb53002011-05-24 15:35:30 -04001123 unsigned long file_end;
1124 u64 isize = i_size_read(inode);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001125 u64 page_start;
1126 u64 page_end;
Liu Bo1f12bd02012-03-29 09:57:44 -04001127 u64 page_cnt;
Chris Mason4cb53002011-05-24 15:35:30 -04001128 int ret;
1129 int i;
1130 int i_done;
1131 struct btrfs_ordered_extent *ordered;
1132 struct extent_state *cached_state = NULL;
Miao Xie600a45e2012-02-16 15:01:24 +08001133 struct extent_io_tree *tree;
Josef Bacik3b16a4e2011-09-21 15:05:58 -04001134 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
Chris Mason4cb53002011-05-24 15:35:30 -04001135
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001136 file_end = (isize - 1) >> PAGE_SHIFT;
Liu Bo1f12bd02012-03-29 09:57:44 -04001137 if (!isize || start_index > file_end)
1138 return 0;
1139
1140 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
Chris Mason4cb53002011-05-24 15:35:30 -04001141
1142 ret = btrfs_delalloc_reserve_space(inode,
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001143 start_index << PAGE_SHIFT,
1144 page_cnt << PAGE_SHIFT);
Chris Mason4cb53002011-05-24 15:35:30 -04001145 if (ret)
1146 return ret;
Chris Mason4cb53002011-05-24 15:35:30 -04001147 i_done = 0;
Miao Xie600a45e2012-02-16 15:01:24 +08001148 tree = &BTRFS_I(inode)->io_tree;
Chris Mason4cb53002011-05-24 15:35:30 -04001149
1150 /* step one, lock all the pages */
Liu Bo1f12bd02012-03-29 09:57:44 -04001151 for (i = 0; i < page_cnt; i++) {
Chris Mason4cb53002011-05-24 15:35:30 -04001152 struct page *page;
Miao Xie600a45e2012-02-16 15:01:24 +08001153again:
Josef Bacika94733d2011-07-11 10:47:06 -04001154 page = find_or_create_page(inode->i_mapping,
Miao Xie600a45e2012-02-16 15:01:24 +08001155 start_index + i, mask);
Chris Mason4cb53002011-05-24 15:35:30 -04001156 if (!page)
1157 break;
1158
Miao Xie600a45e2012-02-16 15:01:24 +08001159 page_start = page_offset(page);
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001160 page_end = page_start + PAGE_SIZE - 1;
Miao Xie600a45e2012-02-16 15:01:24 +08001161 while (1) {
Filipe Manana308d9802014-03-11 13:56:15 +00001162 lock_extent_bits(tree, page_start, page_end,
David Sterbaff13db42015-12-03 14:30:40 +01001163 &cached_state);
Miao Xie600a45e2012-02-16 15:01:24 +08001164 ordered = btrfs_lookup_ordered_extent(inode,
1165 page_start);
Filipe Manana308d9802014-03-11 13:56:15 +00001166 unlock_extent_cached(tree, page_start, page_end,
1167 &cached_state, GFP_NOFS);
Miao Xie600a45e2012-02-16 15:01:24 +08001168 if (!ordered)
1169 break;
1170
1171 unlock_page(page);
1172 btrfs_start_ordered_extent(inode, ordered, 1);
1173 btrfs_put_ordered_extent(ordered);
1174 lock_page(page);
Liu Bo1f12bd02012-03-29 09:57:44 -04001175 /*
1176 * we unlocked the page above, so we need check if
1177 * it was released or not.
1178 */
1179 if (page->mapping != inode->i_mapping) {
1180 unlock_page(page);
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001181 put_page(page);
Liu Bo1f12bd02012-03-29 09:57:44 -04001182 goto again;
1183 }
Miao Xie600a45e2012-02-16 15:01:24 +08001184 }
1185
Chris Mason4cb53002011-05-24 15:35:30 -04001186 if (!PageUptodate(page)) {
1187 btrfs_readpage(NULL, page);
1188 lock_page(page);
1189 if (!PageUptodate(page)) {
1190 unlock_page(page);
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001191 put_page(page);
Chris Mason4cb53002011-05-24 15:35:30 -04001192 ret = -EIO;
1193 break;
1194 }
1195 }
Miao Xie600a45e2012-02-16 15:01:24 +08001196
Miao Xie600a45e2012-02-16 15:01:24 +08001197 if (page->mapping != inode->i_mapping) {
1198 unlock_page(page);
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001199 put_page(page);
Miao Xie600a45e2012-02-16 15:01:24 +08001200 goto again;
1201 }
1202
Chris Mason4cb53002011-05-24 15:35:30 -04001203 pages[i] = page;
1204 i_done++;
1205 }
1206 if (!i_done || ret)
1207 goto out;
1208
1209 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1210 goto out;
1211
1212 /*
1213 * so now we have a nice long stream of locked
1214 * and up to date pages, lets wait on them
1215 */
1216 for (i = 0; i < i_done; i++)
1217 wait_on_page_writeback(pages[i]);
1218
1219 page_start = page_offset(pages[0]);
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001220 page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE;
Chris Mason4cb53002011-05-24 15:35:30 -04001221
1222 lock_extent_bits(&BTRFS_I(inode)->io_tree,
David Sterbaff13db42015-12-03 14:30:40 +01001223 page_start, page_end - 1, &cached_state);
Chris Mason4cb53002011-05-24 15:35:30 -04001224 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1225 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
Liu Bo9e8a4a82012-09-05 19:10:51 -06001226 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1227 &cached_state, GFP_NOFS);
Chris Mason4cb53002011-05-24 15:35:30 -04001228
Liu Bo1f12bd02012-03-29 09:57:44 -04001229 if (i_done != page_cnt) {
Josef Bacik9e0baf62011-07-15 15:16:44 +00001230 spin_lock(&BTRFS_I(inode)->lock);
1231 BTRFS_I(inode)->outstanding_extents++;
1232 spin_unlock(&BTRFS_I(inode)->lock);
Chris Mason4cb53002011-05-24 15:35:30 -04001233 btrfs_delalloc_release_space(inode,
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001234 start_index << PAGE_SHIFT,
1235 (page_cnt - i_done) << PAGE_SHIFT);
Chris Mason4cb53002011-05-24 15:35:30 -04001236 }
1237
1238
Liu Bo9e8a4a82012-09-05 19:10:51 -06001239 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
David Sterba018ed4f2016-04-26 23:54:39 +02001240 &cached_state);
Chris Mason4cb53002011-05-24 15:35:30 -04001241
1242 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1243 page_start, page_end - 1, &cached_state,
1244 GFP_NOFS);
1245
1246 for (i = 0; i < i_done; i++) {
1247 clear_page_dirty_for_io(pages[i]);
1248 ClearPageChecked(pages[i]);
1249 set_page_extent_mapped(pages[i]);
1250 set_page_dirty(pages[i]);
1251 unlock_page(pages[i]);
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001252 put_page(pages[i]);
Chris Mason4cb53002011-05-24 15:35:30 -04001253 }
1254 return i_done;
1255out:
1256 for (i = 0; i < i_done; i++) {
1257 unlock_page(pages[i]);
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001258 put_page(pages[i]);
Chris Mason4cb53002011-05-24 15:35:30 -04001259 }
Qu Wenruo7cf5b972015-09-08 17:25:55 +08001260 btrfs_delalloc_release_space(inode,
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001261 start_index << PAGE_SHIFT,
1262 page_cnt << PAGE_SHIFT);
Chris Mason4cb53002011-05-24 15:35:30 -04001263 return ret;
1264
1265}
1266
1267int btrfs_defrag_file(struct inode *inode, struct file *file,
1268 struct btrfs_ioctl_defrag_range_args *range,
1269 u64 newer_than, unsigned long max_to_defrag)
1270{
1271 struct btrfs_root *root = BTRFS_I(inode)->root;
Chris Mason4cb53002011-05-24 15:35:30 -04001272 struct file_ra_state *ra = NULL;
1273 unsigned long last_index;
Li Zefan151a31b2011-09-02 15:56:39 +08001274 u64 isize = i_size_read(inode);
Chris Mason940100a2010-03-10 10:52:59 -05001275 u64 last_len = 0;
1276 u64 skip = 0;
1277 u64 defrag_end = 0;
Chris Mason4cb53002011-05-24 15:35:30 -04001278 u64 newer_off = range->start;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001279 unsigned long i;
Li Zefan008873e2011-09-02 15:57:07 +08001280 unsigned long ra_index = 0;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001281 int ret;
Chris Mason4cb53002011-05-24 15:35:30 -04001282 int defrag_count = 0;
Li Zefan1a419d82010-10-25 15:12:50 +08001283 int compress_type = BTRFS_COMPRESS_ZLIB;
David Sterbaaab110a2014-07-29 17:32:10 +02001284 u32 extent_thresh = range->extent_thresh;
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001285 unsigned long max_cluster = SZ_256K >> PAGE_SHIFT;
Justin Maggardc41570c2014-01-21 11:18:29 -08001286 unsigned long cluster = max_cluster;
Byongho Leeee221842015-12-15 01:42:10 +09001287 u64 new_align = ~((u64)SZ_128K - 1);
Chris Mason4cb53002011-05-24 15:35:30 -04001288 struct page **pages = NULL;
1289
Liu Bo0abd5b12013-04-16 09:20:28 +00001290 if (isize == 0)
1291 return 0;
1292
1293 if (range->start >= isize)
1294 return -EINVAL;
Li Zefan1a419d82010-10-25 15:12:50 +08001295
1296 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1297 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1298 return -EINVAL;
1299 if (range->compress_type)
1300 compress_type = range->compress_type;
1301 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001302
Liu Bo0abd5b12013-04-16 09:20:28 +00001303 if (extent_thresh == 0)
Byongho Leeee221842015-12-15 01:42:10 +09001304 extent_thresh = SZ_256K;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001305
Chris Mason4cb53002011-05-24 15:35:30 -04001306 /*
1307 * if we were not given a file, allocate a readahead
1308 * context
1309 */
1310 if (!file) {
1311 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1312 if (!ra)
1313 return -ENOMEM;
1314 file_ra_state_init(ra, inode->i_mapping);
1315 } else {
1316 ra = &file->f_ra;
1317 }
1318
Dulshani Gunawardhanad9b0d9b2013-10-31 10:32:18 +05301319 pages = kmalloc_array(max_cluster, sizeof(struct page *),
Chris Mason4cb53002011-05-24 15:35:30 -04001320 GFP_NOFS);
1321 if (!pages) {
1322 ret = -ENOMEM;
1323 goto out_ra;
1324 }
1325
1326 /* find the last page to defrag */
Chris Mason1e701a32010-03-11 09:42:04 -05001327 if (range->start + range->len > range->start) {
Li Zefan151a31b2011-09-02 15:56:39 +08001328 last_index = min_t(u64, isize - 1,
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001329 range->start + range->len - 1) >> PAGE_SHIFT;
Chris Mason1e701a32010-03-11 09:42:04 -05001330 } else {
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001331 last_index = (isize - 1) >> PAGE_SHIFT;
Chris Mason1e701a32010-03-11 09:42:04 -05001332 }
1333
Chris Mason4cb53002011-05-24 15:35:30 -04001334 if (newer_than) {
1335 ret = find_new_extents(root, inode, newer_than,
Byongho Leeee221842015-12-15 01:42:10 +09001336 &newer_off, SZ_64K);
Chris Mason4cb53002011-05-24 15:35:30 -04001337 if (!ret) {
1338 range->start = newer_off;
1339 /*
1340 * we always align our defrag to help keep
1341 * the extents in the file evenly spaced
1342 */
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001343 i = (newer_off & new_align) >> PAGE_SHIFT;
Chris Mason4cb53002011-05-24 15:35:30 -04001344 } else
1345 goto out_ra;
1346 } else {
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001347 i = range->start >> PAGE_SHIFT;
Chris Mason4cb53002011-05-24 15:35:30 -04001348 }
1349 if (!max_to_defrag)
chandan070034b2015-06-09 10:35:11 +05301350 max_to_defrag = last_index - i + 1;
Chris Mason4cb53002011-05-24 15:35:30 -04001351
Li Zefan2a0f7f52011-10-10 15:43:34 -04001352 /*
1353 * make writeback starts from i, so the defrag range can be
1354 * written sequentially.
1355 */
1356 if (i < inode->i_mapping->writeback_index)
1357 inode->i_mapping->writeback_index = i;
1358
Chris Masonf7f43cc2011-10-11 11:41:40 -04001359 while (i <= last_index && defrag_count < max_to_defrag &&
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001360 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) {
Chris Mason4cb53002011-05-24 15:35:30 -04001361 /*
1362 * make sure we stop running if someone unmounts
1363 * the FS
1364 */
1365 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1366 break;
1367
David Sterba210549e2013-02-09 23:38:06 +00001368 if (btrfs_defrag_cancelled(root->fs_info)) {
David Sterbaf14d1042015-10-08 11:37:06 +02001369 btrfs_debug(root->fs_info, "defrag_file cancelled");
David Sterba210549e2013-02-09 23:38:06 +00001370 ret = -EAGAIN;
1371 break;
1372 }
1373
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001374 if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT,
Li Zefan6c282eb2012-06-11 16:03:35 +08001375 extent_thresh, &last_len, &skip,
Andrew Mahonea43a2112012-06-19 21:08:32 -04001376 &defrag_end, range->flags &
1377 BTRFS_DEFRAG_RANGE_COMPRESS)) {
Chris Mason940100a2010-03-10 10:52:59 -05001378 unsigned long next;
1379 /*
1380 * the should_defrag function tells us how much to skip
1381 * bump our counter by the suggested amount
1382 */
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001383 next = DIV_ROUND_UP(skip, PAGE_SIZE);
Chris Mason940100a2010-03-10 10:52:59 -05001384 i = max(i + 1, next);
1385 continue;
1386 }
Li Zefan008873e2011-09-02 15:57:07 +08001387
1388 if (!newer_than) {
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001389 cluster = (PAGE_ALIGN(defrag_end) >>
1390 PAGE_SHIFT) - i;
Li Zefan008873e2011-09-02 15:57:07 +08001391 cluster = min(cluster, max_cluster);
1392 } else {
1393 cluster = max_cluster;
1394 }
1395
Li Zefan008873e2011-09-02 15:57:07 +08001396 if (i + cluster > ra_index) {
1397 ra_index = max(i, ra_index);
1398 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1399 cluster);
chandane4826a52015-06-09 17:38:32 +05301400 ra_index += cluster;
Li Zefan008873e2011-09-02 15:57:07 +08001401 }
Chris Mason940100a2010-03-10 10:52:59 -05001402
Al Viro59551022016-01-22 15:40:57 -05001403 inode_lock(inode);
Filipe David Borba Manana633085c2013-08-16 15:23:33 +01001404 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1405 BTRFS_I(inode)->force_compress = compress_type;
Li Zefan008873e2011-09-02 15:57:07 +08001406 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
Liu Boecb8bea2012-03-29 09:57:44 -04001407 if (ret < 0) {
Al Viro59551022016-01-22 15:40:57 -05001408 inode_unlock(inode);
Chris Mason4cb53002011-05-24 15:35:30 -04001409 goto out_ra;
Liu Boecb8bea2012-03-29 09:57:44 -04001410 }
Chris Mason940100a2010-03-10 10:52:59 -05001411
Chris Mason4cb53002011-05-24 15:35:30 -04001412 defrag_count += ret;
Namjae Jeond0e1d662012-12-11 16:00:21 -08001413 balance_dirty_pages_ratelimited(inode->i_mapping);
Al Viro59551022016-01-22 15:40:57 -05001414 inode_unlock(inode);
Chris Mason4cb53002011-05-24 15:35:30 -04001415
1416 if (newer_than) {
1417 if (newer_off == (u64)-1)
1418 break;
1419
Liu Boe1f041e2012-03-29 09:57:45 -04001420 if (ret > 0)
1421 i += ret;
1422
Chris Mason4cb53002011-05-24 15:35:30 -04001423 newer_off = max(newer_off + 1,
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001424 (u64)i << PAGE_SHIFT);
Chris Mason4cb53002011-05-24 15:35:30 -04001425
Byongho Leeee221842015-12-15 01:42:10 +09001426 ret = find_new_extents(root, inode, newer_than,
1427 &newer_off, SZ_64K);
Chris Mason4cb53002011-05-24 15:35:30 -04001428 if (!ret) {
1429 range->start = newer_off;
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001430 i = (newer_off & new_align) >> PAGE_SHIFT;
Chris Mason4cb53002011-05-24 15:35:30 -04001431 } else {
1432 break;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001433 }
Chris Mason4cb53002011-05-24 15:35:30 -04001434 } else {
Li Zefan008873e2011-09-02 15:57:07 +08001435 if (ret > 0) {
Li Zefancbcc8322011-09-02 15:56:25 +08001436 i += ret;
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001437 last_len += ret << PAGE_SHIFT;
Li Zefan008873e2011-09-02 15:57:07 +08001438 } else {
Li Zefancbcc8322011-09-02 15:56:25 +08001439 i++;
Li Zefan008873e2011-09-02 15:57:07 +08001440 last_len = 0;
1441 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001442 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001443 }
1444
Filipe Mananadec8ef92014-03-01 10:55:54 +00001445 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
Chris Mason1e701a32010-03-11 09:42:04 -05001446 filemap_flush(inode->i_mapping);
Filipe Mananadec8ef92014-03-01 10:55:54 +00001447 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1448 &BTRFS_I(inode)->runtime_flags))
1449 filemap_flush(inode->i_mapping);
1450 }
Chris Mason1e701a32010-03-11 09:42:04 -05001451
1452 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1453 /* the filemap_flush will queue IO into the worker threads, but
1454 * we have to make sure the IO is actually started and that
1455 * ordered extents get created before we return
1456 */
1457 atomic_inc(&root->fs_info->async_submit_draining);
1458 while (atomic_read(&root->fs_info->nr_async_submits) ||
1459 atomic_read(&root->fs_info->async_delalloc_pages)) {
1460 wait_event(root->fs_info->async_submit_wait,
1461 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1462 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1463 }
1464 atomic_dec(&root->fs_info->async_submit_draining);
Chris Mason1e701a32010-03-11 09:42:04 -05001465 }
1466
Li Zefan1a419d82010-10-25 15:12:50 +08001467 if (range->compress_type == BTRFS_COMPRESS_LZO) {
Mitch Harder2b0ce2c2012-07-24 11:58:43 -06001468 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
Li Zefan1a419d82010-10-25 15:12:50 +08001469 }
1470
Diego Calleja60ccf822011-09-01 16:33:57 +02001471 ret = defrag_count;
Chris Mason940100a2010-03-10 10:52:59 -05001472
Chris Mason4cb53002011-05-24 15:35:30 -04001473out_ra:
Filipe David Borba Manana633085c2013-08-16 15:23:33 +01001474 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
Al Viro59551022016-01-22 15:40:57 -05001475 inode_lock(inode);
Filipe David Borba Manana633085c2013-08-16 15:23:33 +01001476 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
Al Viro59551022016-01-22 15:40:57 -05001477 inode_unlock(inode);
Filipe David Borba Manana633085c2013-08-16 15:23:33 +01001478 }
Chris Mason4cb53002011-05-24 15:35:30 -04001479 if (!file)
1480 kfree(ra);
1481 kfree(pages);
Chris Mason940100a2010-03-10 10:52:59 -05001482 return ret;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001483}
1484
Miao Xie198605a2012-11-26 08:43:45 +00001485static noinline int btrfs_ioctl_resize(struct file *file,
Yan, Zheng76dda932009-09-21 16:00:26 -04001486 void __user *arg)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001487{
1488 u64 new_size;
1489 u64 old_size;
1490 u64 devid = 1;
Al Viro496ad9a2013-01-23 17:07:38 -05001491 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001492 struct btrfs_ioctl_vol_args *vol_args;
1493 struct btrfs_trans_handle *trans;
1494 struct btrfs_device *device = NULL;
1495 char *sizestr;
Gui Hecheng9a40f122014-03-31 18:03:25 +08001496 char *retptr;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001497 char *devstr = NULL;
1498 int ret = 0;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001499 int mod = 0;
1500
Chris Masone441d542009-01-05 16:57:23 -05001501 if (!capable(CAP_SYS_ADMIN))
1502 return -EPERM;
1503
Miao Xie198605a2012-11-26 08:43:45 +00001504 ret = mnt_want_write_file(file);
1505 if (ret)
1506 return ret;
1507
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01001508 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1509 1)) {
Miao Xie97547672012-12-21 10:38:50 +00001510 mnt_drop_write_file(file);
Anand Jaine57138b2013-08-21 11:44:48 +08001511 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001512 }
1513
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01001514 mutex_lock(&root->fs_info->volume_mutex);
Li Zefandae7b662009-04-08 15:06:54 +08001515 vol_args = memdup_user(arg, sizeof(*vol_args));
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001516 if (IS_ERR(vol_args)) {
1517 ret = PTR_ERR(vol_args);
1518 goto out;
1519 }
Mark Fasheh5516e592008-07-24 12:20:14 -04001520
1521 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001522
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001523 sizestr = vol_args->name;
1524 devstr = strchr(sizestr, ':');
1525 if (devstr) {
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001526 sizestr = devstr + 1;
1527 *devstr = '\0';
1528 devstr = vol_args->name;
ZhangZhen58dfae62014-05-13 16:36:08 +08001529 ret = kstrtoull(devstr, 10, &devid);
1530 if (ret)
1531 goto out_free;
Miao Xiedfd79822012-12-21 09:21:30 +00001532 if (!devid) {
1533 ret = -EINVAL;
1534 goto out_free;
1535 }
Frank Holtonefe120a2013-12-20 11:37:06 -05001536 btrfs_info(root->fs_info, "resizing devid %llu", devid);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001537 }
Miao Xiedba60f32012-12-21 09:19:51 +00001538
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01001539 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001540 if (!device) {
Frank Holtonefe120a2013-12-20 11:37:06 -05001541 btrfs_info(root->fs_info, "resizer unable to find device %llu",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02001542 devid);
Miao Xiedfd79822012-12-21 09:21:30 +00001543 ret = -ENODEV;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001544 goto out_free;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001545 }
Miao Xiedba60f32012-12-21 09:19:51 +00001546
1547 if (!device->writeable) {
Frank Holtonefe120a2013-12-20 11:37:06 -05001548 btrfs_info(root->fs_info,
1549 "resizer unable to apply on readonly device %llu",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02001550 devid);
Miao Xiedfd79822012-12-21 09:21:30 +00001551 ret = -EPERM;
Liu Bo4e42ae12012-06-14 02:23:19 -06001552 goto out_free;
1553 }
1554
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001555 if (!strcmp(sizestr, "max"))
1556 new_size = device->bdev->bd_inode->i_size;
1557 else {
1558 if (sizestr[0] == '-') {
1559 mod = -1;
1560 sizestr++;
1561 } else if (sizestr[0] == '+') {
1562 mod = 1;
1563 sizestr++;
1564 }
Gui Hecheng9a40f122014-03-31 18:03:25 +08001565 new_size = memparse(sizestr, &retptr);
1566 if (*retptr != '\0' || new_size == 0) {
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001567 ret = -EINVAL;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001568 goto out_free;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001569 }
1570 }
1571
Stefan Behrens63a212a2012-11-05 18:29:28 +01001572 if (device->is_tgtdev_for_dev_replace) {
Miao Xiedfd79822012-12-21 09:21:30 +00001573 ret = -EPERM;
Stefan Behrens63a212a2012-11-05 18:29:28 +01001574 goto out_free;
1575 }
1576
Miao Xie7cc8e582014-09-03 21:35:38 +08001577 old_size = btrfs_device_get_total_bytes(device);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001578
1579 if (mod < 0) {
1580 if (new_size > old_size) {
1581 ret = -EINVAL;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001582 goto out_free;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001583 }
1584 new_size = old_size - new_size;
1585 } else if (mod > 0) {
Wenliang Faneb8052e2013-12-20 15:28:56 +08001586 if (new_size > ULLONG_MAX - old_size) {
Gui Hecheng902c68a2014-05-29 09:19:58 +08001587 ret = -ERANGE;
Wenliang Faneb8052e2013-12-20 15:28:56 +08001588 goto out_free;
1589 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001590 new_size = old_size + new_size;
1591 }
1592
Byongho Leeee221842015-12-15 01:42:10 +09001593 if (new_size < SZ_256M) {
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001594 ret = -EINVAL;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001595 goto out_free;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001596 }
1597 if (new_size > device->bdev->bd_inode->i_size) {
1598 ret = -EFBIG;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001599 goto out_free;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001600 }
1601
David Sterbab8b93ad2015-01-16 17:26:13 +01001602 new_size = div_u64(new_size, root->sectorsize);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001603 new_size *= root->sectorsize;
1604
David Sterbaecaeb142015-10-08 09:01:03 +02001605 btrfs_info_in_rcu(root->fs_info, "new size for %s is %llu",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02001606 rcu_str_deref(device->name), new_size);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001607
1608 if (new_size > old_size) {
Yan, Zhenga22285a2010-05-16 10:48:46 -04001609 trans = btrfs_start_transaction(root, 0);
Tsutomu Itoh98d5dc12011-01-20 06:19:37 +00001610 if (IS_ERR(trans)) {
1611 ret = PTR_ERR(trans);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001612 goto out_free;
Tsutomu Itoh98d5dc12011-01-20 06:19:37 +00001613 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001614 ret = btrfs_grow_device(trans, device, new_size);
1615 btrfs_commit_transaction(trans, root);
Mike Fleetwoodece7d202011-11-18 18:55:01 +00001616 } else if (new_size < old_size) {
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001617 ret = btrfs_shrink_device(device, new_size);
jeff.liu0253f402012-10-27 12:06:39 +00001618 } /* equal, nothing need to do */
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001619
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001620out_free:
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001621 kfree(vol_args);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001622out:
1623 mutex_unlock(&root->fs_info->volume_mutex);
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01001624 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
Ilya Dryomov18f39c42013-01-20 15:57:57 +02001625 mnt_drop_write_file(file);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001626 return ret;
1627}
1628
Sage Weil72fd0322010-10-29 15:41:32 -04001629static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001630 char *name, unsigned long fd, int subvol,
1631 u64 *transid, bool readonly,
Miao Xie8696c532013-02-07 06:02:44 +00001632 struct btrfs_qgroup_inherit *inherit)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001633{
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001634 int namelen;
Chris Mason3de45862008-11-17 21:02:50 -05001635 int ret = 0;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001636
Jeff Mahoney325c50e2016-09-21 08:31:29 -04001637 if (!S_ISDIR(file_inode(file)->i_mode))
1638 return -ENOTDIR;
1639
Liu Boa874a632012-06-29 03:58:46 -06001640 ret = mnt_want_write_file(file);
1641 if (ret)
1642 goto out;
1643
Sage Weil72fd0322010-10-29 15:41:32 -04001644 namelen = strlen(name);
1645 if (strchr(name, '/')) {
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001646 ret = -EINVAL;
Liu Boa874a632012-06-29 03:58:46 -06001647 goto out_drop_write;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001648 }
1649
Chris Mason16780ca2012-02-20 22:14:55 -05001650 if (name[0] == '.' &&
1651 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1652 ret = -EEXIST;
Liu Boa874a632012-06-29 03:58:46 -06001653 goto out_drop_write;
Chris Mason16780ca2012-02-20 22:14:55 -05001654 }
1655
Chris Mason3de45862008-11-17 21:02:50 -05001656 if (subvol) {
Sage Weil72fd0322010-10-29 15:41:32 -04001657 ret = btrfs_mksubvol(&file->f_path, name, namelen,
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001658 NULL, transid, readonly, inherit);
Christoph Hellwigcb8e7092008-10-09 13:39:39 -04001659 } else {
Al Viro2903ff02012-08-28 12:52:22 -04001660 struct fd src = fdget(fd);
Chris Mason3de45862008-11-17 21:02:50 -05001661 struct inode *src_inode;
Al Viro2903ff02012-08-28 12:52:22 -04001662 if (!src.file) {
Chris Mason3de45862008-11-17 21:02:50 -05001663 ret = -EINVAL;
Liu Boa874a632012-06-29 03:58:46 -06001664 goto out_drop_write;
Chris Mason3de45862008-11-17 21:02:50 -05001665 }
1666
Al Viro496ad9a2013-01-23 17:07:38 -05001667 src_inode = file_inode(src.file);
1668 if (src_inode->i_sb != file_inode(file)->i_sb) {
Josef Bacikc79b4712016-03-25 10:02:41 -04001669 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
Frank Holtonefe120a2013-12-20 11:37:06 -05001670 "Snapshot src from another FS");
Kusanagi Kouichi23ad5b12014-01-30 16:32:02 +09001671 ret = -EXDEV;
David Sterbad0242062014-01-15 18:15:52 +01001672 } else if (!inode_owner_or_capable(src_inode)) {
1673 /*
1674 * Subvolume creation is not restricted, but snapshots
1675 * are limited to own subvolumes only
1676 */
1677 ret = -EPERM;
Al Viroecd18812012-08-26 21:20:24 -04001678 } else {
1679 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1680 BTRFS_I(src_inode)->root,
1681 transid, readonly, inherit);
Chris Mason3de45862008-11-17 21:02:50 -05001682 }
Al Viro2903ff02012-08-28 12:52:22 -04001683 fdput(src);
Christoph Hellwigcb8e7092008-10-09 13:39:39 -04001684 }
Liu Boa874a632012-06-29 03:58:46 -06001685out_drop_write:
1686 mnt_drop_write_file(file);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001687out:
Sage Weil72fd0322010-10-29 15:41:32 -04001688 return ret;
1689}
1690
1691static noinline int btrfs_ioctl_snap_create(struct file *file,
Li Zefanfa0d2b92010-12-20 15:53:28 +08001692 void __user *arg, int subvol)
Sage Weil72fd0322010-10-29 15:41:32 -04001693{
Li Zefanfa0d2b92010-12-20 15:53:28 +08001694 struct btrfs_ioctl_vol_args *vol_args;
Sage Weil72fd0322010-10-29 15:41:32 -04001695 int ret;
1696
Jeff Mahoney325c50e2016-09-21 08:31:29 -04001697 if (!S_ISDIR(file_inode(file)->i_mode))
1698 return -ENOTDIR;
1699
Li Zefanfa0d2b92010-12-20 15:53:28 +08001700 vol_args = memdup_user(arg, sizeof(*vol_args));
1701 if (IS_ERR(vol_args))
1702 return PTR_ERR(vol_args);
1703 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
Sage Weil72fd0322010-10-29 15:41:32 -04001704
Li Zefanfa0d2b92010-12-20 15:53:28 +08001705 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
Li Zefanb83cc962010-12-20 16:04:08 +08001706 vol_args->fd, subvol,
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001707 NULL, false, NULL);
Li Zefanfdfb1e42010-12-10 06:41:56 +00001708
Li Zefanfa0d2b92010-12-20 15:53:28 +08001709 kfree(vol_args);
1710 return ret;
1711}
Li Zefanfdfb1e42010-12-10 06:41:56 +00001712
Li Zefanfa0d2b92010-12-20 15:53:28 +08001713static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1714 void __user *arg, int subvol)
1715{
1716 struct btrfs_ioctl_vol_args_v2 *vol_args;
1717 int ret;
1718 u64 transid = 0;
1719 u64 *ptr = NULL;
Li Zefanb83cc962010-12-20 16:04:08 +08001720 bool readonly = false;
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001721 struct btrfs_qgroup_inherit *inherit = NULL;
Li Zefanfdfb1e42010-12-10 06:41:56 +00001722
Jeff Mahoney325c50e2016-09-21 08:31:29 -04001723 if (!S_ISDIR(file_inode(file)->i_mode))
1724 return -ENOTDIR;
1725
Li Zefanfa0d2b92010-12-20 15:53:28 +08001726 vol_args = memdup_user(arg, sizeof(*vol_args));
1727 if (IS_ERR(vol_args))
1728 return PTR_ERR(vol_args);
1729 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
Sage Weil75eaa0e2010-12-10 00:36:28 +00001730
Li Zefanb83cc962010-12-20 16:04:08 +08001731 if (vol_args->flags &
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001732 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1733 BTRFS_SUBVOL_QGROUP_INHERIT)) {
Li Zefanb83cc962010-12-20 16:04:08 +08001734 ret = -EOPNOTSUPP;
Dan Carpenterc47ca322014-09-04 14:09:15 +03001735 goto free_args;
Sage Weil72fd0322010-10-29 15:41:32 -04001736 }
Li Zefanfa0d2b92010-12-20 15:53:28 +08001737
1738 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1739 ptr = &transid;
Li Zefanb83cc962010-12-20 16:04:08 +08001740 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1741 readonly = true;
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001742 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001743 if (vol_args->size > PAGE_SIZE) {
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001744 ret = -EINVAL;
Dan Carpenterc47ca322014-09-04 14:09:15 +03001745 goto free_args;
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001746 }
1747 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1748 if (IS_ERR(inherit)) {
1749 ret = PTR_ERR(inherit);
Dan Carpenterc47ca322014-09-04 14:09:15 +03001750 goto free_args;
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001751 }
1752 }
Li Zefanfa0d2b92010-12-20 15:53:28 +08001753
1754 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001755 vol_args->fd, subvol, ptr,
Miao Xie8696c532013-02-07 06:02:44 +00001756 readonly, inherit);
Dan Carpenterc47ca322014-09-04 14:09:15 +03001757 if (ret)
1758 goto free_inherit;
Li Zefanfa0d2b92010-12-20 15:53:28 +08001759
Dan Carpenterc47ca322014-09-04 14:09:15 +03001760 if (ptr && copy_to_user(arg +
1761 offsetof(struct btrfs_ioctl_vol_args_v2,
1762 transid),
1763 ptr, sizeof(*ptr)))
Li Zefanfa0d2b92010-12-20 15:53:28 +08001764 ret = -EFAULT;
Dan Carpenterc47ca322014-09-04 14:09:15 +03001765
1766free_inherit:
Arne Jansen6f72c7e2011-09-14 15:58:21 +02001767 kfree(inherit);
Dan Carpenterc47ca322014-09-04 14:09:15 +03001768free_args:
1769 kfree(vol_args);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04001770 return ret;
1771}
1772
Li Zefan0caa1022010-12-20 16:30:25 +08001773static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1774 void __user *arg)
1775{
Al Viro496ad9a2013-01-23 17:07:38 -05001776 struct inode *inode = file_inode(file);
Li Zefan0caa1022010-12-20 16:30:25 +08001777 struct btrfs_root *root = BTRFS_I(inode)->root;
1778 int ret = 0;
1779 u64 flags = 0;
1780
Li Zefan33345d012011-04-20 10:31:50 +08001781 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
Li Zefan0caa1022010-12-20 16:30:25 +08001782 return -EINVAL;
1783
1784 down_read(&root->fs_info->subvol_sem);
1785 if (btrfs_root_readonly(root))
1786 flags |= BTRFS_SUBVOL_RDONLY;
1787 up_read(&root->fs_info->subvol_sem);
1788
1789 if (copy_to_user(arg, &flags, sizeof(flags)))
1790 ret = -EFAULT;
1791
1792 return ret;
1793}
1794
1795static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1796 void __user *arg)
1797{
Al Viro496ad9a2013-01-23 17:07:38 -05001798 struct inode *inode = file_inode(file);
Li Zefan0caa1022010-12-20 16:30:25 +08001799 struct btrfs_root *root = BTRFS_I(inode)->root;
1800 struct btrfs_trans_handle *trans;
1801 u64 root_flags;
1802 u64 flags;
1803 int ret = 0;
1804
David Sterbabd60ea02014-01-16 15:50:22 +01001805 if (!inode_owner_or_capable(inode))
1806 return -EPERM;
1807
Liu Bob9ca0662012-06-29 03:58:49 -06001808 ret = mnt_want_write_file(file);
1809 if (ret)
1810 goto out;
Li Zefan0caa1022010-12-20 16:30:25 +08001811
Liu Bob9ca0662012-06-29 03:58:49 -06001812 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1813 ret = -EINVAL;
1814 goto out_drop_write;
1815 }
Li Zefan0caa1022010-12-20 16:30:25 +08001816
Liu Bob9ca0662012-06-29 03:58:49 -06001817 if (copy_from_user(&flags, arg, sizeof(flags))) {
1818 ret = -EFAULT;
1819 goto out_drop_write;
1820 }
Li Zefan0caa1022010-12-20 16:30:25 +08001821
Liu Bob9ca0662012-06-29 03:58:49 -06001822 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1823 ret = -EINVAL;
1824 goto out_drop_write;
1825 }
Li Zefan0caa1022010-12-20 16:30:25 +08001826
Liu Bob9ca0662012-06-29 03:58:49 -06001827 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1828 ret = -EOPNOTSUPP;
1829 goto out_drop_write;
1830 }
Li Zefan0caa1022010-12-20 16:30:25 +08001831
1832 down_write(&root->fs_info->subvol_sem);
1833
1834 /* nothing to do */
1835 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
Liu Bob9ca0662012-06-29 03:58:49 -06001836 goto out_drop_sem;
Li Zefan0caa1022010-12-20 16:30:25 +08001837
1838 root_flags = btrfs_root_flags(&root->root_item);
David Sterba2c686532013-12-16 17:34:17 +01001839 if (flags & BTRFS_SUBVOL_RDONLY) {
Li Zefan0caa1022010-12-20 16:30:25 +08001840 btrfs_set_root_flags(&root->root_item,
1841 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
David Sterba2c686532013-12-16 17:34:17 +01001842 } else {
1843 /*
1844 * Block RO -> RW transition if this subvolume is involved in
1845 * send
1846 */
1847 spin_lock(&root->root_item_lock);
1848 if (root->send_in_progress == 0) {
1849 btrfs_set_root_flags(&root->root_item,
Li Zefan0caa1022010-12-20 16:30:25 +08001850 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
David Sterba2c686532013-12-16 17:34:17 +01001851 spin_unlock(&root->root_item_lock);
1852 } else {
1853 spin_unlock(&root->root_item_lock);
1854 btrfs_warn(root->fs_info,
1855 "Attempt to set subvolume %llu read-write during send",
1856 root->root_key.objectid);
1857 ret = -EPERM;
1858 goto out_drop_sem;
1859 }
1860 }
Li Zefan0caa1022010-12-20 16:30:25 +08001861
1862 trans = btrfs_start_transaction(root, 1);
1863 if (IS_ERR(trans)) {
1864 ret = PTR_ERR(trans);
1865 goto out_reset;
1866 }
1867
Li Zefanb4dc2b82011-02-16 06:06:34 +00001868 ret = btrfs_update_root(trans, root->fs_info->tree_root,
Li Zefan0caa1022010-12-20 16:30:25 +08001869 &root->root_key, &root->root_item);
1870
1871 btrfs_commit_transaction(trans, root);
1872out_reset:
1873 if (ret)
1874 btrfs_set_root_flags(&root->root_item, root_flags);
Liu Bob9ca0662012-06-29 03:58:49 -06001875out_drop_sem:
Li Zefan0caa1022010-12-20 16:30:25 +08001876 up_write(&root->fs_info->subvol_sem);
Liu Bob9ca0662012-06-29 03:58:49 -06001877out_drop_write:
1878 mnt_drop_write_file(file);
1879out:
Li Zefan0caa1022010-12-20 16:30:25 +08001880 return ret;
1881}
1882
Yan, Zheng76dda932009-09-21 16:00:26 -04001883/*
1884 * helper to check if the subvolume references other subvolumes
1885 */
1886static noinline int may_destroy_subvol(struct btrfs_root *root)
1887{
1888 struct btrfs_path *path;
Josef Bacik175a2b82013-08-12 15:36:44 -04001889 struct btrfs_dir_item *di;
Yan, Zheng76dda932009-09-21 16:00:26 -04001890 struct btrfs_key key;
Josef Bacik175a2b82013-08-12 15:36:44 -04001891 u64 dir_id;
Yan, Zheng76dda932009-09-21 16:00:26 -04001892 int ret;
1893
1894 path = btrfs_alloc_path();
1895 if (!path)
1896 return -ENOMEM;
1897
Josef Bacik175a2b82013-08-12 15:36:44 -04001898 /* Make sure this root isn't set as the default subvol */
1899 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1900 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1901 dir_id, "default", 7, 0);
1902 if (di && !IS_ERR(di)) {
1903 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1904 if (key.objectid == root->root_key.objectid) {
Guangyu Sun72de6b52014-03-11 11:24:18 -07001905 ret = -EPERM;
Jeff Mahoney5d163e02016-09-20 10:05:00 -04001906 btrfs_err(root->fs_info,
1907 "deleting default subvolume %llu is not allowed",
1908 key.objectid);
Josef Bacik175a2b82013-08-12 15:36:44 -04001909 goto out;
1910 }
1911 btrfs_release_path(path);
1912 }
1913
Yan, Zheng76dda932009-09-21 16:00:26 -04001914 key.objectid = root->root_key.objectid;
1915 key.type = BTRFS_ROOT_REF_KEY;
1916 key.offset = (u64)-1;
1917
1918 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1919 &key, path, 0, 0);
1920 if (ret < 0)
1921 goto out;
1922 BUG_ON(ret == 0);
1923
1924 ret = 0;
1925 if (path->slots[0] > 0) {
1926 path->slots[0]--;
1927 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1928 if (key.objectid == root->root_key.objectid &&
1929 key.type == BTRFS_ROOT_REF_KEY)
1930 ret = -ENOTEMPTY;
1931 }
1932out:
1933 btrfs_free_path(path);
1934 return ret;
1935}
1936
Chris Masonac8e9812010-02-28 15:39:26 -05001937static noinline int key_in_sk(struct btrfs_key *key,
1938 struct btrfs_ioctl_search_key *sk)
1939{
Chris Masonabc6e132010-03-18 12:10:08 -04001940 struct btrfs_key test;
1941 int ret;
1942
1943 test.objectid = sk->min_objectid;
1944 test.type = sk->min_type;
1945 test.offset = sk->min_offset;
1946
1947 ret = btrfs_comp_cpu_keys(key, &test);
1948 if (ret < 0)
Chris Masonac8e9812010-02-28 15:39:26 -05001949 return 0;
Chris Masonabc6e132010-03-18 12:10:08 -04001950
1951 test.objectid = sk->max_objectid;
1952 test.type = sk->max_type;
1953 test.offset = sk->max_offset;
1954
1955 ret = btrfs_comp_cpu_keys(key, &test);
1956 if (ret > 0)
Chris Masonac8e9812010-02-28 15:39:26 -05001957 return 0;
1958 return 1;
1959}
1960
Jeff Mahoneydf397562016-06-21 20:18:21 -04001961static noinline int copy_to_sk(struct btrfs_path *path,
Chris Masonac8e9812010-02-28 15:39:26 -05001962 struct btrfs_key *key,
1963 struct btrfs_ioctl_search_key *sk,
Gerhard Heift9b6e8172014-01-30 16:24:00 +01001964 size_t *buf_size,
Gerhard Heiftba346b32014-01-30 16:24:02 +01001965 char __user *ubuf,
Chris Masonac8e9812010-02-28 15:39:26 -05001966 unsigned long *sk_offset,
1967 int *num_found)
1968{
1969 u64 found_transid;
1970 struct extent_buffer *leaf;
1971 struct btrfs_ioctl_search_header sh;
Naohiro Aotadd81d452015-06-30 11:25:43 +09001972 struct btrfs_key test;
Chris Masonac8e9812010-02-28 15:39:26 -05001973 unsigned long item_off;
1974 unsigned long item_len;
1975 int nritems;
1976 int i;
1977 int slot;
Chris Masonac8e9812010-02-28 15:39:26 -05001978 int ret = 0;
1979
1980 leaf = path->nodes[0];
1981 slot = path->slots[0];
1982 nritems = btrfs_header_nritems(leaf);
1983
1984 if (btrfs_header_generation(leaf) > sk->max_transid) {
1985 i = nritems;
1986 goto advance_key;
1987 }
1988 found_transid = btrfs_header_generation(leaf);
1989
1990 for (i = slot; i < nritems; i++) {
1991 item_off = btrfs_item_ptr_offset(leaf, i);
1992 item_len = btrfs_item_size_nr(leaf, i);
1993
Gabriel de Perthuis03b71c62013-05-06 17:40:18 +00001994 btrfs_item_key_to_cpu(leaf, key, i);
1995 if (!key_in_sk(key, sk))
1996 continue;
1997
Gerhard Heift9b6e8172014-01-30 16:24:00 +01001998 if (sizeof(sh) + item_len > *buf_size) {
Gerhard Heift8f5f6172014-01-30 16:23:59 +01001999 if (*num_found) {
2000 ret = 1;
2001 goto out;
2002 }
Chris Masonac8e9812010-02-28 15:39:26 -05002003
Gerhard Heift8f5f6172014-01-30 16:23:59 +01002004 /*
2005 * return one empty item back for v1, which does not
2006 * handle -EOVERFLOW
2007 */
2008
Gerhard Heift9b6e8172014-01-30 16:24:00 +01002009 *buf_size = sizeof(sh) + item_len;
Chris Masonac8e9812010-02-28 15:39:26 -05002010 item_len = 0;
Gerhard Heift8f5f6172014-01-30 16:23:59 +01002011 ret = -EOVERFLOW;
2012 }
Chris Masonac8e9812010-02-28 15:39:26 -05002013
Gerhard Heift9b6e8172014-01-30 16:24:00 +01002014 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
Chris Masonac8e9812010-02-28 15:39:26 -05002015 ret = 1;
Gerhard Heift25c9bc22014-01-30 16:23:57 +01002016 goto out;
Chris Masonac8e9812010-02-28 15:39:26 -05002017 }
2018
Chris Masonac8e9812010-02-28 15:39:26 -05002019 sh.objectid = key->objectid;
2020 sh.offset = key->offset;
2021 sh.type = key->type;
2022 sh.len = item_len;
2023 sh.transid = found_transid;
2024
2025 /* copy search result header */
Gerhard Heiftba346b32014-01-30 16:24:02 +01002026 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
2027 ret = -EFAULT;
2028 goto out;
2029 }
2030
Chris Masonac8e9812010-02-28 15:39:26 -05002031 *sk_offset += sizeof(sh);
2032
2033 if (item_len) {
Gerhard Heiftba346b32014-01-30 16:24:02 +01002034 char __user *up = ubuf + *sk_offset;
Chris Masonac8e9812010-02-28 15:39:26 -05002035 /* copy the item */
Gerhard Heiftba346b32014-01-30 16:24:02 +01002036 if (read_extent_buffer_to_user(leaf, up,
2037 item_off, item_len)) {
2038 ret = -EFAULT;
2039 goto out;
2040 }
2041
Chris Masonac8e9812010-02-28 15:39:26 -05002042 *sk_offset += item_len;
Chris Masonac8e9812010-02-28 15:39:26 -05002043 }
Hugo Millse2156862011-05-14 17:43:41 +00002044 (*num_found)++;
Chris Masonac8e9812010-02-28 15:39:26 -05002045
Gerhard Heift8f5f6172014-01-30 16:23:59 +01002046 if (ret) /* -EOVERFLOW from above */
2047 goto out;
2048
Gerhard Heift25c9bc22014-01-30 16:23:57 +01002049 if (*num_found >= sk->nr_items) {
2050 ret = 1;
2051 goto out;
2052 }
Chris Masonac8e9812010-02-28 15:39:26 -05002053 }
2054advance_key:
Chris Masonac8e9812010-02-28 15:39:26 -05002055 ret = 0;
Naohiro Aotadd81d452015-06-30 11:25:43 +09002056 test.objectid = sk->max_objectid;
2057 test.type = sk->max_type;
2058 test.offset = sk->max_offset;
2059 if (btrfs_comp_cpu_keys(key, &test) >= 0)
2060 ret = 1;
2061 else if (key->offset < (u64)-1)
Chris Masonabc6e132010-03-18 12:10:08 -04002062 key->offset++;
Naohiro Aotadd81d452015-06-30 11:25:43 +09002063 else if (key->type < (u8)-1) {
Chris Masonabc6e132010-03-18 12:10:08 -04002064 key->offset = 0;
2065 key->type++;
Naohiro Aotadd81d452015-06-30 11:25:43 +09002066 } else if (key->objectid < (u64)-1) {
Chris Masonabc6e132010-03-18 12:10:08 -04002067 key->offset = 0;
2068 key->type = 0;
2069 key->objectid++;
2070 } else
2071 ret = 1;
Gerhard Heift25c9bc22014-01-30 16:23:57 +01002072out:
Gerhard Heiftba346b32014-01-30 16:24:02 +01002073 /*
2074 * 0: all items from this leaf copied, continue with next
2075 * 1: * more items can be copied, but unused buffer is too small
2076 * * all items were found
2077 * Either way, it will stops the loop which iterates to the next
2078 * leaf
2079 * -EOVERFLOW: item was to large for buffer
2080 * -EFAULT: could not copy extent buffer back to userspace
2081 */
Chris Masonac8e9812010-02-28 15:39:26 -05002082 return ret;
2083}
2084
2085static noinline int search_ioctl(struct inode *inode,
Gerhard Heift12544442014-01-30 16:23:58 +01002086 struct btrfs_ioctl_search_key *sk,
Gerhard Heift9b6e8172014-01-30 16:24:00 +01002087 size_t *buf_size,
Gerhard Heiftba346b32014-01-30 16:24:02 +01002088 char __user *ubuf)
Chris Masonac8e9812010-02-28 15:39:26 -05002089{
2090 struct btrfs_root *root;
2091 struct btrfs_key key;
Chris Masonac8e9812010-02-28 15:39:26 -05002092 struct btrfs_path *path;
Chris Masonac8e9812010-02-28 15:39:26 -05002093 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
2094 int ret;
2095 int num_found = 0;
2096 unsigned long sk_offset = 0;
2097
Gerhard Heift9b6e8172014-01-30 16:24:00 +01002098 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2099 *buf_size = sizeof(struct btrfs_ioctl_search_header);
Gerhard Heift12544442014-01-30 16:23:58 +01002100 return -EOVERFLOW;
Gerhard Heift9b6e8172014-01-30 16:24:00 +01002101 }
Gerhard Heift12544442014-01-30 16:23:58 +01002102
Chris Masonac8e9812010-02-28 15:39:26 -05002103 path = btrfs_alloc_path();
2104 if (!path)
2105 return -ENOMEM;
2106
2107 if (sk->tree_id == 0) {
2108 /* search the root of the inode that was passed */
2109 root = BTRFS_I(inode)->root;
2110 } else {
2111 key.objectid = sk->tree_id;
2112 key.type = BTRFS_ROOT_ITEM_KEY;
2113 key.offset = (u64)-1;
2114 root = btrfs_read_fs_root_no_name(info, &key);
2115 if (IS_ERR(root)) {
Chris Masonac8e9812010-02-28 15:39:26 -05002116 btrfs_free_path(path);
2117 return -ENOENT;
2118 }
2119 }
2120
2121 key.objectid = sk->min_objectid;
2122 key.type = sk->min_type;
2123 key.offset = sk->min_offset;
2124
Dulshani Gunawardhana67871252013-10-31 10:33:04 +05302125 while (1) {
Filipe David Borba Manana6174d3c2013-10-01 16:13:42 +01002126 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
Chris Masonac8e9812010-02-28 15:39:26 -05002127 if (ret != 0) {
2128 if (ret > 0)
2129 ret = 0;
2130 goto err;
2131 }
Jeff Mahoneydf397562016-06-21 20:18:21 -04002132 ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
Chris Masonac8e9812010-02-28 15:39:26 -05002133 &sk_offset, &num_found);
David Sterbab3b4aa72011-04-21 01:20:15 +02002134 btrfs_release_path(path);
Gerhard Heift25c9bc22014-01-30 16:23:57 +01002135 if (ret)
Chris Masonac8e9812010-02-28 15:39:26 -05002136 break;
2137
2138 }
Gerhard Heift8f5f6172014-01-30 16:23:59 +01002139 if (ret > 0)
2140 ret = 0;
Chris Masonac8e9812010-02-28 15:39:26 -05002141err:
2142 sk->nr_items = num_found;
2143 btrfs_free_path(path);
2144 return ret;
2145}
2146
2147static noinline int btrfs_ioctl_tree_search(struct file *file,
2148 void __user *argp)
2149{
Gerhard Heiftba346b32014-01-30 16:24:02 +01002150 struct btrfs_ioctl_search_args __user *uargs;
2151 struct btrfs_ioctl_search_key sk;
Gerhard Heift9b6e8172014-01-30 16:24:00 +01002152 struct inode *inode;
2153 int ret;
2154 size_t buf_size;
Chris Masonac8e9812010-02-28 15:39:26 -05002155
2156 if (!capable(CAP_SYS_ADMIN))
2157 return -EPERM;
2158
Gerhard Heiftba346b32014-01-30 16:24:02 +01002159 uargs = (struct btrfs_ioctl_search_args __user *)argp;
Chris Masonac8e9812010-02-28 15:39:26 -05002160
Gerhard Heiftba346b32014-01-30 16:24:02 +01002161 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2162 return -EFAULT;
2163
2164 buf_size = sizeof(uargs->buf);
Chris Masonac8e9812010-02-28 15:39:26 -05002165
Al Viro496ad9a2013-01-23 17:07:38 -05002166 inode = file_inode(file);
Gerhard Heiftba346b32014-01-30 16:24:02 +01002167 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
Gerhard Heift8f5f6172014-01-30 16:23:59 +01002168
2169 /*
2170 * In the origin implementation an overflow is handled by returning a
2171 * search header with a len of zero, so reset ret.
2172 */
2173 if (ret == -EOVERFLOW)
2174 ret = 0;
2175
Gerhard Heiftba346b32014-01-30 16:24:02 +01002176 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
Chris Masonac8e9812010-02-28 15:39:26 -05002177 ret = -EFAULT;
Chris Masonac8e9812010-02-28 15:39:26 -05002178 return ret;
2179}
2180
Gerhard Heiftcc68a8a2014-01-30 16:24:03 +01002181static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2182 void __user *argp)
2183{
2184 struct btrfs_ioctl_search_args_v2 __user *uarg;
2185 struct btrfs_ioctl_search_args_v2 args;
2186 struct inode *inode;
2187 int ret;
2188 size_t buf_size;
Byongho Leeee221842015-12-15 01:42:10 +09002189 const size_t buf_limit = SZ_16M;
Gerhard Heiftcc68a8a2014-01-30 16:24:03 +01002190
2191 if (!capable(CAP_SYS_ADMIN))
2192 return -EPERM;
2193
2194 /* copy search header and buffer size */
2195 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2196 if (copy_from_user(&args, uarg, sizeof(args)))
2197 return -EFAULT;
2198
2199 buf_size = args.buf_size;
2200
2201 if (buf_size < sizeof(struct btrfs_ioctl_search_header))
2202 return -EOVERFLOW;
2203
2204 /* limit result size to 16MB */
2205 if (buf_size > buf_limit)
2206 buf_size = buf_limit;
2207
2208 inode = file_inode(file);
2209 ret = search_ioctl(inode, &args.key, &buf_size,
2210 (char *)(&uarg->buf[0]));
2211 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2212 ret = -EFAULT;
2213 else if (ret == -EOVERFLOW &&
2214 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2215 ret = -EFAULT;
2216
Yan, Zheng76dda932009-09-21 16:00:26 -04002217 return ret;
2218}
2219
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002220/*
Chris Masonac8e9812010-02-28 15:39:26 -05002221 * Search INODE_REFs to identify path name of 'dirid' directory
2222 * in a 'tree_id' tree. and sets path name to 'name'.
2223 */
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002224static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2225 u64 tree_id, u64 dirid, char *name)
2226{
2227 struct btrfs_root *root;
2228 struct btrfs_key key;
Chris Masonac8e9812010-02-28 15:39:26 -05002229 char *ptr;
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002230 int ret = -1;
2231 int slot;
2232 int len;
2233 int total_len = 0;
2234 struct btrfs_inode_ref *iref;
2235 struct extent_buffer *l;
2236 struct btrfs_path *path;
2237
2238 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2239 name[0]='\0';
2240 return 0;
2241 }
2242
2243 path = btrfs_alloc_path();
2244 if (!path)
2245 return -ENOMEM;
2246
Chris Masonac8e9812010-02-28 15:39:26 -05002247 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002248
2249 key.objectid = tree_id;
2250 key.type = BTRFS_ROOT_ITEM_KEY;
2251 key.offset = (u64)-1;
2252 root = btrfs_read_fs_root_no_name(info, &key);
2253 if (IS_ERR(root)) {
David Sterbaf14d1042015-10-08 11:37:06 +02002254 btrfs_err(info, "could not find root %llu", tree_id);
Chris Mason8ad6fca2010-03-18 12:23:10 -04002255 ret = -ENOENT;
2256 goto out;
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002257 }
2258
2259 key.objectid = dirid;
2260 key.type = BTRFS_INODE_REF_KEY;
Chris Mason8ad6fca2010-03-18 12:23:10 -04002261 key.offset = (u64)-1;
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002262
Dulshani Gunawardhana67871252013-10-31 10:33:04 +05302263 while (1) {
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002264 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2265 if (ret < 0)
2266 goto out;
Filipe David Borba Manana18674c62013-08-14 03:00:21 +01002267 else if (ret > 0) {
2268 ret = btrfs_previous_item(root, path, dirid,
2269 BTRFS_INODE_REF_KEY);
2270 if (ret < 0)
2271 goto out;
2272 else if (ret > 0) {
2273 ret = -ENOENT;
2274 goto out;
2275 }
2276 }
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002277
2278 l = path->nodes[0];
2279 slot = path->slots[0];
2280 btrfs_item_key_to_cpu(l, &key, slot);
2281
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002282 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2283 len = btrfs_inode_ref_name_len(l, iref);
2284 ptr -= len + 1;
2285 total_len += len + 1;
Filipe David Borba Mananaa696cf32013-08-14 03:00:20 +01002286 if (ptr < name) {
2287 ret = -ENAMETOOLONG;
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002288 goto out;
Filipe David Borba Mananaa696cf32013-08-14 03:00:20 +01002289 }
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002290
2291 *(ptr + len) = '/';
Dulshani Gunawardhana67871252013-10-31 10:33:04 +05302292 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002293
2294 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2295 break;
2296
David Sterbab3b4aa72011-04-21 01:20:15 +02002297 btrfs_release_path(path);
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002298 key.objectid = key.offset;
Chris Mason8ad6fca2010-03-18 12:23:10 -04002299 key.offset = (u64)-1;
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002300 dirid = key.objectid;
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002301 }
Li Zefan77906a502011-07-14 03:16:00 +00002302 memmove(name, ptr, total_len);
Dulshani Gunawardhana67871252013-10-31 10:33:04 +05302303 name[total_len] = '\0';
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002304 ret = 0;
2305out:
2306 btrfs_free_path(path);
Chris Masonac8e9812010-02-28 15:39:26 -05002307 return ret;
2308}
2309
2310static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2311 void __user *argp)
2312{
2313 struct btrfs_ioctl_ino_lookup_args *args;
2314 struct inode *inode;
David Sterba01b810b2015-05-12 19:14:49 +02002315 int ret = 0;
Chris Masonac8e9812010-02-28 15:39:26 -05002316
Julia Lawall2354d08f2010-10-29 15:14:18 -04002317 args = memdup_user(argp, sizeof(*args));
2318 if (IS_ERR(args))
2319 return PTR_ERR(args);
Dan Carpenterc2b96922010-03-20 11:24:15 +00002320
Al Viro496ad9a2013-01-23 17:07:38 -05002321 inode = file_inode(file);
Chris Masonac8e9812010-02-28 15:39:26 -05002322
David Sterba01b810b2015-05-12 19:14:49 +02002323 /*
2324 * Unprivileged query to obtain the containing subvolume root id. The
2325 * path is reset so it's consistent with btrfs_search_path_in_tree.
2326 */
Chris Mason1b53ac42010-03-18 12:17:05 -04002327 if (args->treeid == 0)
2328 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2329
David Sterba01b810b2015-05-12 19:14:49 +02002330 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2331 args->name[0] = 0;
2332 goto out;
2333 }
2334
2335 if (!capable(CAP_SYS_ADMIN)) {
2336 ret = -EPERM;
2337 goto out;
2338 }
2339
Chris Masonac8e9812010-02-28 15:39:26 -05002340 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2341 args->treeid, args->objectid,
2342 args->name);
2343
David Sterba01b810b2015-05-12 19:14:49 +02002344out:
Chris Masonac8e9812010-02-28 15:39:26 -05002345 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2346 ret = -EFAULT;
2347
2348 kfree(args);
TARUISI Hiroaki98d377a2009-11-18 05:42:14 +00002349 return ret;
2350}
2351
Yan, Zheng76dda932009-09-21 16:00:26 -04002352static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2353 void __user *arg)
2354{
Al Viro54563d42013-09-01 15:57:51 -04002355 struct dentry *parent = file->f_path.dentry;
Yan, Zheng76dda932009-09-21 16:00:26 -04002356 struct dentry *dentry;
David Howells2b0143b2015-03-17 22:25:59 +00002357 struct inode *dir = d_inode(parent);
Yan, Zheng76dda932009-09-21 16:00:26 -04002358 struct inode *inode;
2359 struct btrfs_root *root = BTRFS_I(dir)->root;
2360 struct btrfs_root *dest = NULL;
2361 struct btrfs_ioctl_vol_args *vol_args;
2362 struct btrfs_trans_handle *trans;
Miao Xiec58aaad2013-02-28 10:05:36 +00002363 struct btrfs_block_rsv block_rsv;
David Sterba521e0542014-04-15 16:41:44 +02002364 u64 root_flags;
Miao Xiec58aaad2013-02-28 10:05:36 +00002365 u64 qgroup_reserved;
Yan, Zheng76dda932009-09-21 16:00:26 -04002366 int namelen;
2367 int ret;
2368 int err = 0;
2369
Jeff Mahoney325c50e2016-09-21 08:31:29 -04002370 if (!S_ISDIR(dir->i_mode))
2371 return -ENOTDIR;
2372
Yan, Zheng76dda932009-09-21 16:00:26 -04002373 vol_args = memdup_user(arg, sizeof(*vol_args));
2374 if (IS_ERR(vol_args))
2375 return PTR_ERR(vol_args);
2376
2377 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2378 namelen = strlen(vol_args->name);
2379 if (strchr(vol_args->name, '/') ||
2380 strncmp(vol_args->name, "..", namelen) == 0) {
2381 err = -EINVAL;
2382 goto out;
2383 }
2384
Al Viroa561be72011-11-23 11:57:51 -05002385 err = mnt_want_write_file(file);
Yan, Zheng76dda932009-09-21 16:00:26 -04002386 if (err)
2387 goto out;
2388
David Sterba521e0542014-04-15 16:41:44 +02002389
Al Viro00235412016-05-26 00:05:12 -04002390 err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
2391 if (err == -EINTR)
2392 goto out_drop_write;
Yan, Zheng76dda932009-09-21 16:00:26 -04002393 dentry = lookup_one_len(vol_args->name, parent, namelen);
2394 if (IS_ERR(dentry)) {
2395 err = PTR_ERR(dentry);
2396 goto out_unlock_dir;
2397 }
2398
David Howells2b0143b2015-03-17 22:25:59 +00002399 if (d_really_is_negative(dentry)) {
Yan, Zheng76dda932009-09-21 16:00:26 -04002400 err = -ENOENT;
2401 goto out_dput;
2402 }
2403
David Howells2b0143b2015-03-17 22:25:59 +00002404 inode = d_inode(dentry);
Sage Weil4260f7c2010-10-29 15:46:43 -04002405 dest = BTRFS_I(inode)->root;
Dulshani Gunawardhana67871252013-10-31 10:33:04 +05302406 if (!capable(CAP_SYS_ADMIN)) {
Sage Weil4260f7c2010-10-29 15:46:43 -04002407 /*
2408 * Regular user. Only allow this with a special mount
2409 * option, when the user has write+exec access to the
2410 * subvol root, and when rmdir(2) would have been
2411 * allowed.
2412 *
2413 * Note that this is _not_ check that the subvol is
2414 * empty or doesn't contain data that we wouldn't
2415 * otherwise be able to delete.
2416 *
2417 * Users who want to delete empty subvols should try
2418 * rmdir(2).
2419 */
2420 err = -EPERM;
Jeff Mahoney3cdde222016-06-09 21:38:35 -04002421 if (!btrfs_test_opt(root->fs_info, USER_SUBVOL_RM_ALLOWED))
Sage Weil4260f7c2010-10-29 15:46:43 -04002422 goto out_dput;
2423
2424 /*
2425 * Do not allow deletion if the parent dir is the same
2426 * as the dir to be deleted. That means the ioctl
2427 * must be called on the dentry referencing the root
2428 * of the subvol, not a random directory contained
2429 * within it.
2430 */
2431 err = -EINVAL;
2432 if (root == dest)
2433 goto out_dput;
2434
2435 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2436 if (err)
2437 goto out_dput;
Sage Weil4260f7c2010-10-29 15:46:43 -04002438 }
2439
Miao Xie5c39da52012-10-22 11:39:53 +00002440 /* check if subvolume may be deleted by a user */
2441 err = btrfs_may_delete(dir, dentry, 1);
2442 if (err)
2443 goto out_dput;
2444
Li Zefan33345d012011-04-20 10:31:50 +08002445 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
Yan, Zheng76dda932009-09-21 16:00:26 -04002446 err = -EINVAL;
2447 goto out_dput;
2448 }
2449
Al Viro59551022016-01-22 15:40:57 -05002450 inode_lock(inode);
David Sterba521e0542014-04-15 16:41:44 +02002451
2452 /*
2453 * Don't allow to delete a subvolume with send in progress. This is
2454 * inside the i_mutex so the error handling that has to drop the bit
2455 * again is not run concurrently.
2456 */
2457 spin_lock(&dest->root_item_lock);
Filipe Mananac55bfa62014-05-25 03:55:44 +01002458 root_flags = btrfs_root_flags(&dest->root_item);
2459 if (dest->send_in_progress == 0) {
2460 btrfs_set_root_flags(&dest->root_item,
David Sterba521e0542014-04-15 16:41:44 +02002461 root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2462 spin_unlock(&dest->root_item_lock);
2463 } else {
2464 spin_unlock(&dest->root_item_lock);
2465 btrfs_warn(root->fs_info,
2466 "Attempt to delete subvolume %llu during send",
Filipe Mananac55bfa62014-05-25 03:55:44 +01002467 dest->root_key.objectid);
David Sterba521e0542014-04-15 16:41:44 +02002468 err = -EPERM;
Omar Sandoval909e26d2015-04-10 14:20:40 -07002469 goto out_unlock_inode;
David Sterba521e0542014-04-15 16:41:44 +02002470 }
2471
Yan, Zheng76dda932009-09-21 16:00:26 -04002472 down_write(&root->fs_info->subvol_sem);
2473
2474 err = may_destroy_subvol(dest);
2475 if (err)
2476 goto out_up_write;
2477
Miao Xiec58aaad2013-02-28 10:05:36 +00002478 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2479 /*
2480 * One for dir inode, two for dir entries, two for root
2481 * ref/backref.
2482 */
2483 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
Jeff Mahoneyee3441b2013-07-09 16:37:21 -04002484 5, &qgroup_reserved, true);
Miao Xiec58aaad2013-02-28 10:05:36 +00002485 if (err)
2486 goto out_up_write;
2487
Yan, Zhenga22285a2010-05-16 10:48:46 -04002488 trans = btrfs_start_transaction(root, 0);
2489 if (IS_ERR(trans)) {
2490 err = PTR_ERR(trans);
Miao Xiec58aaad2013-02-28 10:05:36 +00002491 goto out_release;
Yan, Zhenga22285a2010-05-16 10:48:46 -04002492 }
Miao Xiec58aaad2013-02-28 10:05:36 +00002493 trans->block_rsv = &block_rsv;
2494 trans->bytes_reserved = block_rsv.size;
Yan, Zhenga22285a2010-05-16 10:48:46 -04002495
Filipe Manana2be63d52016-02-12 11:34:23 +00002496 btrfs_record_snapshot_destroy(trans, dir);
2497
Yan, Zheng76dda932009-09-21 16:00:26 -04002498 ret = btrfs_unlink_subvol(trans, root, dir,
2499 dest->root_key.objectid,
2500 dentry->d_name.name,
2501 dentry->d_name.len);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002502 if (ret) {
2503 err = ret;
Jeff Mahoney66642832016-06-10 18:19:25 -04002504 btrfs_abort_transaction(trans, ret);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002505 goto out_end_trans;
2506 }
Yan, Zheng76dda932009-09-21 16:00:26 -04002507
2508 btrfs_record_root_in_trans(trans, dest);
2509
2510 memset(&dest->root_item.drop_progress, 0,
2511 sizeof(dest->root_item.drop_progress));
2512 dest->root_item.drop_level = 0;
2513 btrfs_set_root_refs(&dest->root_item, 0);
2514
Miao Xie27cdeb72014-04-02 19:51:05 +08002515 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
Yan, Zhengd68fc572010-05-16 10:49:58 -04002516 ret = btrfs_insert_orphan_item(trans,
2517 root->fs_info->tree_root,
2518 dest->root_key.objectid);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002519 if (ret) {
Jeff Mahoney66642832016-06-10 18:19:25 -04002520 btrfs_abort_transaction(trans, ret);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002521 err = ret;
2522 goto out_end_trans;
2523 }
Yan, Zhengd68fc572010-05-16 10:49:58 -04002524 }
Stefan Behrensdd5f9612013-08-15 17:11:20 +02002525
2526 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2527 dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
2528 dest->root_key.objectid);
2529 if (ret && ret != -ENOENT) {
Jeff Mahoney66642832016-06-10 18:19:25 -04002530 btrfs_abort_transaction(trans, ret);
Stefan Behrensdd5f9612013-08-15 17:11:20 +02002531 err = ret;
2532 goto out_end_trans;
2533 }
2534 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2535 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2536 dest->root_item.received_uuid,
2537 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2538 dest->root_key.objectid);
2539 if (ret && ret != -ENOENT) {
Jeff Mahoney66642832016-06-10 18:19:25 -04002540 btrfs_abort_transaction(trans, ret);
Stefan Behrensdd5f9612013-08-15 17:11:20 +02002541 err = ret;
2542 goto out_end_trans;
2543 }
2544 }
2545
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002546out_end_trans:
Miao Xiec58aaad2013-02-28 10:05:36 +00002547 trans->block_rsv = NULL;
2548 trans->bytes_reserved = 0;
Sage Weil531cb132010-10-29 15:41:32 -04002549 ret = btrfs_end_transaction(trans, root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002550 if (ret && !err)
2551 err = ret;
Yan, Zheng76dda932009-09-21 16:00:26 -04002552 inode->i_flags |= S_DEAD;
Miao Xiec58aaad2013-02-28 10:05:36 +00002553out_release:
2554 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
Yan, Zheng76dda932009-09-21 16:00:26 -04002555out_up_write:
2556 up_write(&root->fs_info->subvol_sem);
David Sterba521e0542014-04-15 16:41:44 +02002557 if (err) {
2558 spin_lock(&dest->root_item_lock);
Filipe Mananac55bfa62014-05-25 03:55:44 +01002559 root_flags = btrfs_root_flags(&dest->root_item);
2560 btrfs_set_root_flags(&dest->root_item,
David Sterba521e0542014-04-15 16:41:44 +02002561 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2562 spin_unlock(&dest->root_item_lock);
2563 }
Omar Sandoval909e26d2015-04-10 14:20:40 -07002564out_unlock_inode:
Al Viro59551022016-01-22 15:40:57 -05002565 inode_unlock(inode);
Yan, Zheng76dda932009-09-21 16:00:26 -04002566 if (!err) {
Omar Sandoval64ad6c42015-06-02 17:31:00 -07002567 d_invalidate(dentry);
Yan, Zheng76dda932009-09-21 16:00:26 -04002568 btrfs_invalidate_inodes(dest);
2569 d_delete(dentry);
David Sterba61155aa2014-04-15 16:42:03 +02002570 ASSERT(dest->send_in_progress == 0);
Liu Bofa6ac872013-02-20 14:10:23 +00002571
2572 /* the last ref */
David Sterba57cdc8d2014-02-05 02:37:48 +01002573 if (dest->ino_cache_inode) {
2574 iput(dest->ino_cache_inode);
2575 dest->ino_cache_inode = NULL;
Liu Bofa6ac872013-02-20 14:10:23 +00002576 }
Yan, Zheng76dda932009-09-21 16:00:26 -04002577 }
2578out_dput:
2579 dput(dentry);
2580out_unlock_dir:
Al Viro59551022016-01-22 15:40:57 -05002581 inode_unlock(dir);
Al Viro00235412016-05-26 00:05:12 -04002582out_drop_write:
Al Viro2a79f172011-12-09 08:06:57 -05002583 mnt_drop_write_file(file);
Yan, Zheng76dda932009-09-21 16:00:26 -04002584out:
2585 kfree(vol_args);
2586 return err;
2587}
2588
Chris Mason1e701a32010-03-11 09:42:04 -05002589static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002590{
Al Viro496ad9a2013-01-23 17:07:38 -05002591 struct inode *inode = file_inode(file);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002592 struct btrfs_root *root = BTRFS_I(inode)->root;
Chris Mason1e701a32010-03-11 09:42:04 -05002593 struct btrfs_ioctl_defrag_range_args *range;
Yan Zhengc146afa2008-11-12 14:34:12 -05002594 int ret;
2595
Ilya Dryomov25122d12013-01-20 15:57:57 +02002596 ret = mnt_want_write_file(file);
2597 if (ret)
2598 return ret;
Li Zefanb83cc962010-12-20 16:04:08 +08002599
Ilya Dryomov25122d12013-01-20 15:57:57 +02002600 if (btrfs_root_readonly(root)) {
2601 ret = -EROFS;
2602 goto out;
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01002603 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002604
2605 switch (inode->i_mode & S_IFMT) {
2606 case S_IFDIR:
Chris Masone441d542009-01-05 16:57:23 -05002607 if (!capable(CAP_SYS_ADMIN)) {
2608 ret = -EPERM;
2609 goto out;
2610 }
Eric Sandeende78b512013-01-31 18:21:12 +00002611 ret = btrfs_defrag_root(root);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002612 if (ret)
2613 goto out;
Eric Sandeende78b512013-01-31 18:21:12 +00002614 ret = btrfs_defrag_root(root->fs_info->extent_root);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002615 break;
2616 case S_IFREG:
Chris Masone441d542009-01-05 16:57:23 -05002617 if (!(file->f_mode & FMODE_WRITE)) {
2618 ret = -EINVAL;
2619 goto out;
2620 }
Chris Mason1e701a32010-03-11 09:42:04 -05002621
2622 range = kzalloc(sizeof(*range), GFP_KERNEL);
2623 if (!range) {
2624 ret = -ENOMEM;
2625 goto out;
2626 }
2627
2628 if (argp) {
2629 if (copy_from_user(range, argp,
2630 sizeof(*range))) {
2631 ret = -EFAULT;
2632 kfree(range);
Dan Carpenter683be162010-03-20 11:24:48 +00002633 goto out;
Chris Mason1e701a32010-03-11 09:42:04 -05002634 }
2635 /* compression requires us to start the IO */
2636 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2637 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2638 range->extent_thresh = (u32)-1;
2639 }
2640 } else {
2641 /* the rest are all set to zero by kzalloc */
2642 range->len = (u64)-1;
2643 }
Al Viro496ad9a2013-01-23 17:07:38 -05002644 ret = btrfs_defrag_file(file_inode(file), file,
Chris Mason4cb53002011-05-24 15:35:30 -04002645 range, 0, 0);
2646 if (ret > 0)
2647 ret = 0;
Chris Mason1e701a32010-03-11 09:42:04 -05002648 kfree(range);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002649 break;
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002650 default:
2651 ret = -EINVAL;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002652 }
Chris Masone441d542009-01-05 16:57:23 -05002653out:
Ilya Dryomov25122d12013-01-20 15:57:57 +02002654 mnt_drop_write_file(file);
Chris Masone441d542009-01-05 16:57:23 -05002655 return ret;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002656}
2657
Christoph Hellwigb2950862008-12-02 09:54:17 -05002658static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002659{
2660 struct btrfs_ioctl_vol_args *vol_args;
2661 int ret;
2662
Chris Masone441d542009-01-05 16:57:23 -05002663 if (!capable(CAP_SYS_ADMIN))
2664 return -EPERM;
2665
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01002666 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2667 1)) {
Anand Jaine57138b2013-08-21 11:44:48 +08002668 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002669 }
2670
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01002671 mutex_lock(&root->fs_info->volume_mutex);
Li Zefandae7b662009-04-08 15:06:54 +08002672 vol_args = memdup_user(arg, sizeof(*vol_args));
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002673 if (IS_ERR(vol_args)) {
2674 ret = PTR_ERR(vol_args);
2675 goto out;
2676 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002677
Mark Fasheh5516e592008-07-24 12:20:14 -04002678 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002679 ret = btrfs_init_new_device(root, vol_args->name);
2680
Anand Jain43d20762014-07-01 00:58:56 +08002681 if (!ret)
2682 btrfs_info(root->fs_info, "disk added %s",vol_args->name);
2683
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002684 kfree(vol_args);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002685out:
2686 mutex_unlock(&root->fs_info->volume_mutex);
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01002687 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002688 return ret;
2689}
2690
Anand Jain6b526ed2016-02-13 10:01:39 +08002691static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002692{
Al Viro496ad9a2013-01-23 17:07:38 -05002693 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Anand Jain6b526ed2016-02-13 10:01:39 +08002694 struct btrfs_ioctl_vol_args_v2 *vol_args;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002695 int ret;
2696
Chris Masone441d542009-01-05 16:57:23 -05002697 if (!capable(CAP_SYS_ADMIN))
2698 return -EPERM;
2699
Miao Xieda249272012-11-26 08:44:50 +00002700 ret = mnt_want_write_file(file);
2701 if (ret)
2702 return ret;
Yan Zhengc146afa2008-11-12 14:34:12 -05002703
Li Zefandae7b662009-04-08 15:06:54 +08002704 vol_args = memdup_user(arg, sizeof(*vol_args));
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002705 if (IS_ERR(vol_args)) {
2706 ret = PTR_ERR(vol_args);
Dan Carpenterc47ca322014-09-04 14:09:15 +03002707 goto err_drop;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002708 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002709
Anand Jain6b526ed2016-02-13 10:01:39 +08002710 /* Check for compatibility reject unknown flags */
David Sterba735654e2016-02-15 18:15:21 +01002711 if (vol_args->flags & ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED)
2712 return -EOPNOTSUPP;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002713
Anand Jain183860f2013-05-17 10:52:45 +00002714 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2715 1)) {
2716 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2717 goto out;
2718 }
2719
2720 mutex_lock(&root->fs_info->volume_mutex);
David Sterba735654e2016-02-15 18:15:21 +01002721 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
Anand Jain6b526ed2016-02-13 10:01:39 +08002722 ret = btrfs_rm_device(root, NULL, vol_args->devid);
2723 } else {
2724 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
2725 ret = btrfs_rm_device(root, vol_args->name, 0);
2726 }
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002727 mutex_unlock(&root->fs_info->volume_mutex);
Stefan Behrens5ac00ad2012-11-05 17:54:08 +01002728 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
Anand Jain183860f2013-05-17 10:52:45 +00002729
Anand Jain6b526ed2016-02-13 10:01:39 +08002730 if (!ret) {
David Sterba735654e2016-02-15 18:15:21 +01002731 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
Anand Jain6b526ed2016-02-13 10:01:39 +08002732 btrfs_info(root->fs_info, "device deleted: id %llu",
2733 vol_args->devid);
2734 else
2735 btrfs_info(root->fs_info, "device deleted: %s",
2736 vol_args->name);
2737 }
Anand Jain183860f2013-05-17 10:52:45 +00002738out:
2739 kfree(vol_args);
Dan Carpenterc47ca322014-09-04 14:09:15 +03002740err_drop:
Ilya Dryomov4ac20c72013-01-20 15:57:57 +02002741 mnt_drop_write_file(file);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002742 return ret;
2743}
2744
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002745static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2746{
2747 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2748 struct btrfs_ioctl_vol_args *vol_args;
2749 int ret;
2750
2751 if (!capable(CAP_SYS_ADMIN))
2752 return -EPERM;
2753
2754 ret = mnt_want_write_file(file);
2755 if (ret)
2756 return ret;
2757
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002758 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2759 1)) {
2760 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
David Sterba58d7bbf2016-05-04 14:10:47 +02002761 goto out_drop_write;
2762 }
2763
2764 vol_args = memdup_user(arg, sizeof(*vol_args));
2765 if (IS_ERR(vol_args)) {
2766 ret = PTR_ERR(vol_args);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002767 goto out;
2768 }
2769
David Sterba58d7bbf2016-05-04 14:10:47 +02002770 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002771 mutex_lock(&root->fs_info->volume_mutex);
Anand Jain6b526ed2016-02-13 10:01:39 +08002772 ret = btrfs_rm_device(root, vol_args->name, 0);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002773 mutex_unlock(&root->fs_info->volume_mutex);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002774
2775 if (!ret)
2776 btrfs_info(root->fs_info, "disk deleted %s",vol_args->name);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002777 kfree(vol_args);
David Sterba58d7bbf2016-05-04 14:10:47 +02002778out:
2779 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2780out_drop_write:
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002781 mnt_drop_write_file(file);
David Sterba58d7bbf2016-05-04 14:10:47 +02002782
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04002783 return ret;
2784}
2785
Jan Schmidt475f6382011-03-11 15:41:01 +01002786static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2787{
Li Zefan027ed2f2011-06-08 08:27:56 +00002788 struct btrfs_ioctl_fs_info_args *fi_args;
Jan Schmidt475f6382011-03-11 15:41:01 +01002789 struct btrfs_device *device;
Jan Schmidt475f6382011-03-11 15:41:01 +01002790 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
Li Zefan027ed2f2011-06-08 08:27:56 +00002791 int ret = 0;
Jan Schmidt475f6382011-03-11 15:41:01 +01002792
Li Zefan027ed2f2011-06-08 08:27:56 +00002793 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2794 if (!fi_args)
2795 return -ENOMEM;
2796
Filipe David Borba Mananaf7171752013-08-12 20:56:58 +01002797 mutex_lock(&fs_devices->device_list_mutex);
Li Zefan027ed2f2011-06-08 08:27:56 +00002798 fi_args->num_devices = fs_devices->num_devices;
2799 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
Jan Schmidt475f6382011-03-11 15:41:01 +01002800
Byongho Leed7641a42015-09-01 23:10:57 +09002801 list_for_each_entry(device, &fs_devices->devices, dev_list) {
Li Zefan027ed2f2011-06-08 08:27:56 +00002802 if (device->devid > fi_args->max_id)
2803 fi_args->max_id = device->devid;
Jan Schmidt475f6382011-03-11 15:41:01 +01002804 }
2805 mutex_unlock(&fs_devices->device_list_mutex);
2806
David Sterba80a773f2014-05-07 18:17:06 +02002807 fi_args->nodesize = root->fs_info->super_copy->nodesize;
2808 fi_args->sectorsize = root->fs_info->super_copy->sectorsize;
2809 fi_args->clone_alignment = root->fs_info->super_copy->sectorsize;
2810
Li Zefan027ed2f2011-06-08 08:27:56 +00002811 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2812 ret = -EFAULT;
Jan Schmidt475f6382011-03-11 15:41:01 +01002813
Li Zefan027ed2f2011-06-08 08:27:56 +00002814 kfree(fi_args);
2815 return ret;
Jan Schmidt475f6382011-03-11 15:41:01 +01002816}
2817
2818static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2819{
2820 struct btrfs_ioctl_dev_info_args *di_args;
2821 struct btrfs_device *dev;
2822 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2823 int ret = 0;
2824 char *s_uuid = NULL;
Jan Schmidt475f6382011-03-11 15:41:01 +01002825
Jan Schmidt475f6382011-03-11 15:41:01 +01002826 di_args = memdup_user(arg, sizeof(*di_args));
2827 if (IS_ERR(di_args))
2828 return PTR_ERR(di_args);
2829
Stefan Behrensdd5f9612013-08-15 17:11:20 +02002830 if (!btrfs_is_empty_uuid(di_args->uuid))
Jan Schmidt475f6382011-03-11 15:41:01 +01002831 s_uuid = di_args->uuid;
2832
2833 mutex_lock(&fs_devices->device_list_mutex);
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01002834 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
Jan Schmidt475f6382011-03-11 15:41:01 +01002835
2836 if (!dev) {
2837 ret = -ENODEV;
2838 goto out;
2839 }
2840
2841 di_args->devid = dev->devid;
Miao Xie7cc8e582014-09-03 21:35:38 +08002842 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2843 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
Jan Schmidt475f6382011-03-11 15:41:01 +01002844 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
Jim Meyeringa27202f2012-04-26 18:36:56 +02002845 if (dev->name) {
Josef Bacik606686e2012-06-04 14:03:51 -04002846 struct rcu_string *name;
2847
2848 rcu_read_lock();
2849 name = rcu_dereference(dev->name);
2850 strncpy(di_args->path, name->str, sizeof(di_args->path));
2851 rcu_read_unlock();
Jim Meyeringa27202f2012-04-26 18:36:56 +02002852 di_args->path[sizeof(di_args->path) - 1] = 0;
2853 } else {
Stefan Behrens99ba55a2012-03-19 16:17:22 +01002854 di_args->path[0] = '\0';
Jim Meyeringa27202f2012-04-26 18:36:56 +02002855 }
Jan Schmidt475f6382011-03-11 15:41:01 +01002856
2857out:
David Sterba55793c02013-04-26 15:20:23 +00002858 mutex_unlock(&fs_devices->device_list_mutex);
Jan Schmidt475f6382011-03-11 15:41:01 +01002859 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2860 ret = -EFAULT;
2861
2862 kfree(di_args);
2863 return ret;
2864}
2865
Mark Fashehf4414602015-06-30 14:42:05 -07002866static struct page *extent_same_get_page(struct inode *inode, pgoff_t index)
Mark Fasheh416161d2013-08-06 11:42:51 -07002867{
2868 struct page *page;
Mark Fasheh416161d2013-08-06 11:42:51 -07002869
Mark Fasheh416161d2013-08-06 11:42:51 -07002870 page = grab_cache_page(inode->i_mapping, index);
2871 if (!page)
Filipe Manana31314002016-01-27 18:37:47 +00002872 return ERR_PTR(-ENOMEM);
Mark Fasheh416161d2013-08-06 11:42:51 -07002873
2874 if (!PageUptodate(page)) {
Filipe Manana31314002016-01-27 18:37:47 +00002875 int ret;
2876
2877 ret = btrfs_readpage(NULL, page);
2878 if (ret)
2879 return ERR_PTR(ret);
Mark Fasheh416161d2013-08-06 11:42:51 -07002880 lock_page(page);
2881 if (!PageUptodate(page)) {
2882 unlock_page(page);
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03002883 put_page(page);
Filipe Manana31314002016-01-27 18:37:47 +00002884 return ERR_PTR(-EIO);
2885 }
2886 if (page->mapping != inode->i_mapping) {
2887 unlock_page(page);
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03002888 put_page(page);
Filipe Manana31314002016-01-27 18:37:47 +00002889 return ERR_PTR(-EAGAIN);
Mark Fasheh416161d2013-08-06 11:42:51 -07002890 }
2891 }
Mark Fasheh416161d2013-08-06 11:42:51 -07002892
2893 return page;
2894}
2895
Mark Fashehf4414602015-06-30 14:42:05 -07002896static int gather_extent_pages(struct inode *inode, struct page **pages,
2897 int num_pages, u64 off)
2898{
2899 int i;
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03002900 pgoff_t index = off >> PAGE_SHIFT;
Mark Fashehf4414602015-06-30 14:42:05 -07002901
2902 for (i = 0; i < num_pages; i++) {
Filipe Manana31314002016-01-27 18:37:47 +00002903again:
Mark Fashehf4414602015-06-30 14:42:05 -07002904 pages[i] = extent_same_get_page(inode, index + i);
Filipe Manana31314002016-01-27 18:37:47 +00002905 if (IS_ERR(pages[i])) {
2906 int err = PTR_ERR(pages[i]);
2907
2908 if (err == -EAGAIN)
2909 goto again;
2910 pages[i] = NULL;
2911 return err;
2912 }
Mark Fashehf4414602015-06-30 14:42:05 -07002913 }
2914 return 0;
2915}
2916
Filipe Mananae0bd70c2016-01-27 10:20:58 +00002917static int lock_extent_range(struct inode *inode, u64 off, u64 len,
2918 bool retry_range_locking)
Mark Fasheh77fe20dc2013-08-06 11:42:48 -07002919{
Filipe Mananae0bd70c2016-01-27 10:20:58 +00002920 /*
2921 * Do any pending delalloc/csum calculations on inode, one way or
2922 * another, and lock file content.
2923 * The locking order is:
2924 *
2925 * 1) pages
2926 * 2) range in the inode's io tree
2927 */
Mark Fasheh77fe20dc2013-08-06 11:42:48 -07002928 while (1) {
2929 struct btrfs_ordered_extent *ordered;
2930 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2931 ordered = btrfs_lookup_first_ordered_extent(inode,
2932 off + len - 1);
Filipe Mananaff5df9b2014-05-30 17:56:24 +01002933 if ((!ordered ||
2934 ordered->file_offset + ordered->len <= off ||
2935 ordered->file_offset >= off + len) &&
Mark Fasheh77fe20dc2013-08-06 11:42:48 -07002936 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
Filipe Mananaff5df9b2014-05-30 17:56:24 +01002937 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2938 if (ordered)
2939 btrfs_put_ordered_extent(ordered);
Mark Fasheh77fe20dc2013-08-06 11:42:48 -07002940 break;
Filipe Mananaff5df9b2014-05-30 17:56:24 +01002941 }
Mark Fasheh77fe20dc2013-08-06 11:42:48 -07002942 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2943 if (ordered)
2944 btrfs_put_ordered_extent(ordered);
Filipe Mananae0bd70c2016-01-27 10:20:58 +00002945 if (!retry_range_locking)
2946 return -EAGAIN;
Mark Fasheh77fe20dc2013-08-06 11:42:48 -07002947 btrfs_wait_ordered_range(inode, off, len);
2948 }
Filipe Mananae0bd70c2016-01-27 10:20:58 +00002949 return 0;
Mark Fasheh77fe20dc2013-08-06 11:42:48 -07002950}
2951
Mark Fashehf4414602015-06-30 14:42:05 -07002952static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2)
Mark Fasheh416161d2013-08-06 11:42:51 -07002953{
Al Viro59551022016-01-22 15:40:57 -05002954 inode_unlock(inode1);
2955 inode_unlock(inode2);
Mark Fasheh416161d2013-08-06 11:42:51 -07002956}
2957
Mark Fashehf4414602015-06-30 14:42:05 -07002958static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2)
2959{
2960 if (inode1 < inode2)
2961 swap(inode1, inode2);
2962
Al Viro59551022016-01-22 15:40:57 -05002963 inode_lock_nested(inode1, I_MUTEX_PARENT);
2964 inode_lock_nested(inode2, I_MUTEX_CHILD);
Mark Fashehf4414602015-06-30 14:42:05 -07002965}
2966
2967static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1,
2968 struct inode *inode2, u64 loff2, u64 len)
2969{
2970 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2971 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2972}
2973
Filipe Mananae0bd70c2016-01-27 10:20:58 +00002974static int btrfs_double_extent_lock(struct inode *inode1, u64 loff1,
2975 struct inode *inode2, u64 loff2, u64 len,
2976 bool retry_range_locking)
Mark Fasheh416161d2013-08-06 11:42:51 -07002977{
Filipe Mananae0bd70c2016-01-27 10:20:58 +00002978 int ret;
2979
Mark Fasheh416161d2013-08-06 11:42:51 -07002980 if (inode1 < inode2) {
2981 swap(inode1, inode2);
2982 swap(loff1, loff2);
2983 }
Filipe Mananae0bd70c2016-01-27 10:20:58 +00002984 ret = lock_extent_range(inode1, loff1, len, retry_range_locking);
2985 if (ret)
2986 return ret;
2987 ret = lock_extent_range(inode2, loff2, len, retry_range_locking);
2988 if (ret)
2989 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1,
2990 loff1 + len - 1);
2991 return ret;
Mark Fashehf4414602015-06-30 14:42:05 -07002992}
2993
2994struct cmp_pages {
2995 int num_pages;
2996 struct page **src_pages;
2997 struct page **dst_pages;
2998};
2999
3000static void btrfs_cmp_data_free(struct cmp_pages *cmp)
3001{
3002 int i;
3003 struct page *pg;
3004
3005 for (i = 0; i < cmp->num_pages; i++) {
3006 pg = cmp->src_pages[i];
Filipe Mananae0bd70c2016-01-27 10:20:58 +00003007 if (pg) {
3008 unlock_page(pg);
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03003009 put_page(pg);
Filipe Mananae0bd70c2016-01-27 10:20:58 +00003010 }
Mark Fashehf4414602015-06-30 14:42:05 -07003011 pg = cmp->dst_pages[i];
Filipe Mananae0bd70c2016-01-27 10:20:58 +00003012 if (pg) {
3013 unlock_page(pg);
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03003014 put_page(pg);
Filipe Mananae0bd70c2016-01-27 10:20:58 +00003015 }
Mark Fasheh416161d2013-08-06 11:42:51 -07003016 }
Mark Fashehf4414602015-06-30 14:42:05 -07003017 kfree(cmp->src_pages);
3018 kfree(cmp->dst_pages);
3019}
3020
3021static int btrfs_cmp_data_prepare(struct inode *src, u64 loff,
3022 struct inode *dst, u64 dst_loff,
3023 u64 len, struct cmp_pages *cmp)
3024{
3025 int ret;
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03003026 int num_pages = PAGE_ALIGN(len) >> PAGE_SHIFT;
Mark Fashehf4414602015-06-30 14:42:05 -07003027 struct page **src_pgarr, **dst_pgarr;
3028
3029 /*
3030 * We must gather up all the pages before we initiate our
3031 * extent locking. We use an array for the page pointers. Size
3032 * of the array is bounded by len, which is in turn bounded by
3033 * BTRFS_MAX_DEDUPE_LEN.
3034 */
David Sterba66722f72016-02-11 15:01:38 +01003035 src_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL);
3036 dst_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL);
Mark Fashehf4414602015-06-30 14:42:05 -07003037 if (!src_pgarr || !dst_pgarr) {
3038 kfree(src_pgarr);
3039 kfree(dst_pgarr);
3040 return -ENOMEM;
3041 }
3042 cmp->num_pages = num_pages;
3043 cmp->src_pages = src_pgarr;
3044 cmp->dst_pages = dst_pgarr;
3045
3046 ret = gather_extent_pages(src, cmp->src_pages, cmp->num_pages, loff);
3047 if (ret)
3048 goto out;
3049
3050 ret = gather_extent_pages(dst, cmp->dst_pages, cmp->num_pages, dst_loff);
3051
3052out:
3053 if (ret)
3054 btrfs_cmp_data_free(cmp);
3055 return 0;
Mark Fasheh416161d2013-08-06 11:42:51 -07003056}
3057
3058static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
Mark Fashehf4414602015-06-30 14:42:05 -07003059 u64 dst_loff, u64 len, struct cmp_pages *cmp)
Mark Fasheh416161d2013-08-06 11:42:51 -07003060{
3061 int ret = 0;
Mark Fashehf4414602015-06-30 14:42:05 -07003062 int i;
Mark Fasheh416161d2013-08-06 11:42:51 -07003063 struct page *src_page, *dst_page;
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03003064 unsigned int cmp_len = PAGE_SIZE;
Mark Fasheh416161d2013-08-06 11:42:51 -07003065 void *addr, *dst_addr;
3066
Mark Fashehf4414602015-06-30 14:42:05 -07003067 i = 0;
Mark Fasheh416161d2013-08-06 11:42:51 -07003068 while (len) {
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03003069 if (len < PAGE_SIZE)
Mark Fasheh416161d2013-08-06 11:42:51 -07003070 cmp_len = len;
3071
Mark Fashehf4414602015-06-30 14:42:05 -07003072 BUG_ON(i >= cmp->num_pages);
3073
3074 src_page = cmp->src_pages[i];
3075 dst_page = cmp->dst_pages[i];
Filipe Mananae0bd70c2016-01-27 10:20:58 +00003076 ASSERT(PageLocked(src_page));
3077 ASSERT(PageLocked(dst_page));
Mark Fashehf4414602015-06-30 14:42:05 -07003078
Mark Fasheh416161d2013-08-06 11:42:51 -07003079 addr = kmap_atomic(src_page);
3080 dst_addr = kmap_atomic(dst_page);
3081
3082 flush_dcache_page(src_page);
3083 flush_dcache_page(dst_page);
3084
3085 if (memcmp(addr, dst_addr, cmp_len))
Darrick J. Wong2b3909f2015-12-19 00:56:05 -08003086 ret = -EBADE;
Mark Fasheh416161d2013-08-06 11:42:51 -07003087
3088 kunmap_atomic(addr);
3089 kunmap_atomic(dst_addr);
Mark Fasheh416161d2013-08-06 11:42:51 -07003090
3091 if (ret)
3092 break;
3093
Mark Fasheh416161d2013-08-06 11:42:51 -07003094 len -= cmp_len;
Mark Fashehf4414602015-06-30 14:42:05 -07003095 i++;
Mark Fasheh416161d2013-08-06 11:42:51 -07003096 }
3097
3098 return ret;
3099}
3100
Mark Fashehe1d227a2015-06-08 15:05:25 -07003101static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen,
3102 u64 olen)
Mark Fasheh416161d2013-08-06 11:42:51 -07003103{
Mark Fashehe1d227a2015-06-08 15:05:25 -07003104 u64 len = *plen;
Mark Fasheh416161d2013-08-06 11:42:51 -07003105 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
3106
Mark Fashehe1d227a2015-06-08 15:05:25 -07003107 if (off + olen > inode->i_size || off + olen < off)
Mark Fasheh416161d2013-08-06 11:42:51 -07003108 return -EINVAL;
Mark Fashehe1d227a2015-06-08 15:05:25 -07003109
3110 /* if we extend to eof, continue to block boundary */
3111 if (off + len == inode->i_size)
3112 *plen = len = ALIGN(inode->i_size, bs) - off;
3113
Mark Fasheh416161d2013-08-06 11:42:51 -07003114 /* Check that we are block aligned - btrfs_clone() requires this */
3115 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
3116 return -EINVAL;
3117
3118 return 0;
3119}
3120
Mark Fashehe1d227a2015-06-08 15:05:25 -07003121static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen,
Mark Fasheh416161d2013-08-06 11:42:51 -07003122 struct inode *dst, u64 dst_loff)
3123{
3124 int ret;
Mark Fashehe1d227a2015-06-08 15:05:25 -07003125 u64 len = olen;
Mark Fashehf4414602015-06-30 14:42:05 -07003126 struct cmp_pages cmp;
Mark Fasheh0efa9f42015-06-30 14:42:07 -07003127 int same_inode = 0;
3128 u64 same_lock_start = 0;
3129 u64 same_lock_len = 0;
Mark Fasheh416161d2013-08-06 11:42:51 -07003130
Mark Fasheh416161d2013-08-06 11:42:51 -07003131 if (src == dst)
Mark Fasheh0efa9f42015-06-30 14:42:07 -07003132 same_inode = 1;
Mark Fasheh416161d2013-08-06 11:42:51 -07003133
Filipe Manana113e8282015-03-30 18:26:47 +01003134 if (len == 0)
3135 return 0;
3136
Mark Fasheh0efa9f42015-06-30 14:42:07 -07003137 if (same_inode) {
Al Viro59551022016-01-22 15:40:57 -05003138 inode_lock(src);
Mark Fasheh416161d2013-08-06 11:42:51 -07003139
Mark Fasheh0efa9f42015-06-30 14:42:07 -07003140 ret = extent_same_check_offsets(src, loff, &len, olen);
3141 if (ret)
3142 goto out_unlock;
Filipe Mananaf4dfe682016-02-12 14:44:00 +00003143 ret = extent_same_check_offsets(src, dst_loff, &len, olen);
3144 if (ret)
3145 goto out_unlock;
Mark Fasheh416161d2013-08-06 11:42:51 -07003146
Mark Fasheh0efa9f42015-06-30 14:42:07 -07003147 /*
3148 * Single inode case wants the same checks, except we
3149 * don't want our length pushed out past i_size as
3150 * comparing that data range makes no sense.
3151 *
3152 * extent_same_check_offsets() will do this for an
3153 * unaligned length at i_size, so catch it here and
3154 * reject the request.
3155 *
3156 * This effectively means we require aligned extents
3157 * for the single-inode case, whereas the other cases
3158 * allow an unaligned length so long as it ends at
3159 * i_size.
3160 */
3161 if (len != olen) {
3162 ret = -EINVAL;
3163 goto out_unlock;
3164 }
3165
3166 /* Check for overlapping ranges */
3167 if (dst_loff + len > loff && dst_loff < loff + len) {
3168 ret = -EINVAL;
3169 goto out_unlock;
3170 }
3171
3172 same_lock_start = min_t(u64, loff, dst_loff);
3173 same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start;
3174 } else {
3175 btrfs_double_inode_lock(src, dst);
3176
3177 ret = extent_same_check_offsets(src, loff, &len, olen);
3178 if (ret)
3179 goto out_unlock;
3180
3181 ret = extent_same_check_offsets(dst, dst_loff, &len, olen);
3182 if (ret)
3183 goto out_unlock;
3184 }
Mark Fasheh416161d2013-08-06 11:42:51 -07003185
3186 /* don't make the dst file partly checksummed */
3187 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3188 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
3189 ret = -EINVAL;
3190 goto out_unlock;
3191 }
3192
Filipe Mananae0bd70c2016-01-27 10:20:58 +00003193again:
Mark Fashehf4414602015-06-30 14:42:05 -07003194 ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, &cmp);
3195 if (ret)
3196 goto out_unlock;
3197
Mark Fasheh0efa9f42015-06-30 14:42:07 -07003198 if (same_inode)
Filipe Mananae0bd70c2016-01-27 10:20:58 +00003199 ret = lock_extent_range(src, same_lock_start, same_lock_len,
3200 false);
Mark Fasheh0efa9f42015-06-30 14:42:07 -07003201 else
Filipe Mananae0bd70c2016-01-27 10:20:58 +00003202 ret = btrfs_double_extent_lock(src, loff, dst, dst_loff, len,
3203 false);
3204 /*
3205 * If one of the inodes has dirty pages in the respective range or
3206 * ordered extents, we need to flush dellaloc and wait for all ordered
3207 * extents in the range. We must unlock the pages and the ranges in the
3208 * io trees to avoid deadlocks when flushing delalloc (requires locking
3209 * pages) and when waiting for ordered extents to complete (they require
3210 * range locking).
3211 */
3212 if (ret == -EAGAIN) {
3213 /*
3214 * Ranges in the io trees already unlocked. Now unlock all
3215 * pages before waiting for all IO to complete.
3216 */
3217 btrfs_cmp_data_free(&cmp);
3218 if (same_inode) {
3219 btrfs_wait_ordered_range(src, same_lock_start,
3220 same_lock_len);
3221 } else {
3222 btrfs_wait_ordered_range(src, loff, len);
3223 btrfs_wait_ordered_range(dst, dst_loff, len);
3224 }
3225 goto again;
3226 }
3227 ASSERT(ret == 0);
3228 if (WARN_ON(ret)) {
3229 /* ranges in the io trees already unlocked */
3230 btrfs_cmp_data_free(&cmp);
3231 return ret;
3232 }
Mark Fashehf4414602015-06-30 14:42:05 -07003233
Mark Fasheh207910d2015-06-30 14:42:04 -07003234 /* pass original length for comparison so we stay within i_size */
Mark Fashehf4414602015-06-30 14:42:05 -07003235 ret = btrfs_cmp_data(src, loff, dst, dst_loff, olen, &cmp);
Mark Fasheh416161d2013-08-06 11:42:51 -07003236 if (ret == 0)
Mark Fasheh1c919a52015-06-30 14:42:08 -07003237 ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1);
Mark Fasheh416161d2013-08-06 11:42:51 -07003238
Mark Fasheh0efa9f42015-06-30 14:42:07 -07003239 if (same_inode)
3240 unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start,
3241 same_lock_start + same_lock_len - 1);
3242 else
3243 btrfs_double_extent_unlock(src, loff, dst, dst_loff, len);
Mark Fashehf4414602015-06-30 14:42:05 -07003244
3245 btrfs_cmp_data_free(&cmp);
Mark Fasheh416161d2013-08-06 11:42:51 -07003246out_unlock:
Mark Fasheh0efa9f42015-06-30 14:42:07 -07003247 if (same_inode)
Al Viro59551022016-01-22 15:40:57 -05003248 inode_unlock(src);
Mark Fasheh0efa9f42015-06-30 14:42:07 -07003249 else
3250 btrfs_double_inode_unlock(src, dst);
Mark Fasheh416161d2013-08-06 11:42:51 -07003251
3252 return ret;
3253}
3254
Byongho Leeee221842015-12-15 01:42:10 +09003255#define BTRFS_MAX_DEDUPE_LEN SZ_16M
Mark Fasheh416161d2013-08-06 11:42:51 -07003256
Darrick J. Wong2b3909f2015-12-19 00:56:05 -08003257ssize_t btrfs_dedupe_file_range(struct file *src_file, u64 loff, u64 olen,
3258 struct file *dst_file, u64 dst_loff)
Mark Fasheh416161d2013-08-06 11:42:51 -07003259{
Darrick J. Wong2b3909f2015-12-19 00:56:05 -08003260 struct inode *src = file_inode(src_file);
3261 struct inode *dst = file_inode(dst_file);
Mark Fasheh416161d2013-08-06 11:42:51 -07003262 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
Darrick J. Wong2b3909f2015-12-19 00:56:05 -08003263 ssize_t res;
Mark Fasheh416161d2013-08-06 11:42:51 -07003264
Darrick J. Wong2b3909f2015-12-19 00:56:05 -08003265 if (olen > BTRFS_MAX_DEDUPE_LEN)
3266 olen = BTRFS_MAX_DEDUPE_LEN;
Mark Fasheh416161d2013-08-06 11:42:51 -07003267
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03003268 if (WARN_ON_ONCE(bs < PAGE_SIZE)) {
Mark Fasheh416161d2013-08-06 11:42:51 -07003269 /*
3270 * Btrfs does not support blocksize < page_size. As a
3271 * result, btrfs_cmp_data() won't correctly handle
3272 * this situation without an update.
3273 */
Darrick J. Wong2b3909f2015-12-19 00:56:05 -08003274 return -EINVAL;
Mark Fasheh416161d2013-08-06 11:42:51 -07003275 }
3276
Darrick J. Wong2b3909f2015-12-19 00:56:05 -08003277 res = btrfs_extent_same(src, loff, olen, dst, dst_loff);
3278 if (res)
3279 return res;
3280 return olen;
Mark Fasheh416161d2013-08-06 11:42:51 -07003281}
3282
Filipe Mananaf82a9902014-06-01 01:50:28 +01003283static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3284 struct inode *inode,
3285 u64 endoff,
3286 const u64 destoff,
Mark Fasheh1c919a52015-06-30 14:42:08 -07003287 const u64 olen,
3288 int no_time_update)
Filipe Mananaf82a9902014-06-01 01:50:28 +01003289{
3290 struct btrfs_root *root = BTRFS_I(inode)->root;
3291 int ret;
3292
3293 inode_inc_iversion(inode);
Mark Fasheh1c919a52015-06-30 14:42:08 -07003294 if (!no_time_update)
Deepa Dinamani04b285f2016-02-06 23:57:21 -08003295 inode->i_mtime = inode->i_ctime = current_fs_time(inode->i_sb);
Filipe Mananaf82a9902014-06-01 01:50:28 +01003296 /*
3297 * We round up to the block size at eof when determining which
3298 * extents to clone above, but shouldn't round up the file size.
3299 */
3300 if (endoff > destoff + olen)
3301 endoff = destoff + olen;
3302 if (endoff > inode->i_size)
3303 btrfs_i_size_write(inode, endoff);
3304
3305 ret = btrfs_update_inode(trans, root, inode);
3306 if (ret) {
Jeff Mahoney66642832016-06-10 18:19:25 -04003307 btrfs_abort_transaction(trans, ret);
Filipe Mananaf82a9902014-06-01 01:50:28 +01003308 btrfs_end_transaction(trans, root);
3309 goto out;
3310 }
3311 ret = btrfs_end_transaction(trans, root);
3312out:
3313 return ret;
3314}
3315
Filipe Manana7ffbb592014-06-09 03:48:05 +01003316static void clone_update_extent_map(struct inode *inode,
3317 const struct btrfs_trans_handle *trans,
3318 const struct btrfs_path *path,
Filipe Manana7ffbb592014-06-09 03:48:05 +01003319 const u64 hole_offset,
3320 const u64 hole_len)
3321{
3322 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
3323 struct extent_map *em;
3324 int ret;
3325
3326 em = alloc_extent_map();
3327 if (!em) {
3328 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3329 &BTRFS_I(inode)->runtime_flags);
3330 return;
3331 }
3332
Filipe Manana14f59792014-06-29 21:45:40 +01003333 if (path) {
3334 struct btrfs_file_extent_item *fi;
3335
3336 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3337 struct btrfs_file_extent_item);
Filipe Manana7ffbb592014-06-09 03:48:05 +01003338 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3339 em->generation = -1;
3340 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3341 BTRFS_FILE_EXTENT_INLINE)
3342 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3343 &BTRFS_I(inode)->runtime_flags);
3344 } else {
3345 em->start = hole_offset;
3346 em->len = hole_len;
3347 em->ram_bytes = em->len;
3348 em->orig_start = hole_offset;
3349 em->block_start = EXTENT_MAP_HOLE;
3350 em->block_len = 0;
3351 em->orig_block_len = 0;
3352 em->compress_type = BTRFS_COMPRESS_NONE;
3353 em->generation = trans->transid;
3354 }
3355
3356 while (1) {
3357 write_lock(&em_tree->lock);
3358 ret = add_extent_mapping(em_tree, em, 1);
3359 write_unlock(&em_tree->lock);
3360 if (ret != -EEXIST) {
3361 free_extent_map(em);
3362 break;
3363 }
3364 btrfs_drop_extent_cache(inode, em->start,
3365 em->start + em->len - 1, 0);
3366 }
3367
David Sterbaee39b432014-09-30 01:33:33 +02003368 if (ret)
Filipe Manana7ffbb592014-06-09 03:48:05 +01003369 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3370 &BTRFS_I(inode)->runtime_flags);
3371}
3372
Filipe Manana8039d872015-10-13 15:15:00 +01003373/*
3374 * Make sure we do not end up inserting an inline extent into a file that has
3375 * already other (non-inline) extents. If a file has an inline extent it can
3376 * not have any other extents and the (single) inline extent must start at the
3377 * file offset 0. Failing to respect these rules will lead to file corruption,
3378 * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc
3379 *
3380 * We can have extents that have been already written to disk or we can have
3381 * dirty ranges still in delalloc, in which case the extent maps and items are
3382 * created only when we run delalloc, and the delalloc ranges might fall outside
3383 * the range we are currently locking in the inode's io tree. So we check the
3384 * inode's i_size because of that (i_size updates are done while holding the
3385 * i_mutex, which we are holding here).
3386 * We also check to see if the inode has a size not greater than "datal" but has
3387 * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are
3388 * protected against such concurrent fallocate calls by the i_mutex).
3389 *
3390 * If the file has no extents but a size greater than datal, do not allow the
3391 * copy because we would need turn the inline extent into a non-inline one (even
3392 * with NO_HOLES enabled). If we find our destination inode only has one inline
3393 * extent, just overwrite it with the source inline extent if its size is less
3394 * than the source extent's size, or we could copy the source inline extent's
3395 * data into the destination inode's inline extent if the later is greater then
3396 * the former.
3397 */
3398static int clone_copy_inline_extent(struct inode *src,
3399 struct inode *dst,
3400 struct btrfs_trans_handle *trans,
3401 struct btrfs_path *path,
3402 struct btrfs_key *new_key,
3403 const u64 drop_start,
3404 const u64 datal,
3405 const u64 skip,
3406 const u64 size,
3407 char *inline_data)
3408{
3409 struct btrfs_root *root = BTRFS_I(dst)->root;
3410 const u64 aligned_end = ALIGN(new_key->offset + datal,
3411 root->sectorsize);
3412 int ret;
3413 struct btrfs_key key;
3414
3415 if (new_key->offset > 0)
3416 return -EOPNOTSUPP;
3417
3418 key.objectid = btrfs_ino(dst);
3419 key.type = BTRFS_EXTENT_DATA_KEY;
3420 key.offset = 0;
3421 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3422 if (ret < 0) {
3423 return ret;
3424 } else if (ret > 0) {
3425 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
3426 ret = btrfs_next_leaf(root, path);
3427 if (ret < 0)
3428 return ret;
3429 else if (ret > 0)
3430 goto copy_inline_extent;
3431 }
3432 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
3433 if (key.objectid == btrfs_ino(dst) &&
3434 key.type == BTRFS_EXTENT_DATA_KEY) {
3435 ASSERT(key.offset > 0);
3436 return -EOPNOTSUPP;
3437 }
3438 } else if (i_size_read(dst) <= datal) {
3439 struct btrfs_file_extent_item *ei;
3440 u64 ext_len;
3441
3442 /*
3443 * If the file size is <= datal, make sure there are no other
3444 * extents following (can happen do to an fallocate call with
3445 * the flag FALLOC_FL_KEEP_SIZE).
3446 */
3447 ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
3448 struct btrfs_file_extent_item);
3449 /*
3450 * If it's an inline extent, it can not have other extents
3451 * following it.
3452 */
3453 if (btrfs_file_extent_type(path->nodes[0], ei) ==
3454 BTRFS_FILE_EXTENT_INLINE)
3455 goto copy_inline_extent;
3456
3457 ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei);
3458 if (ext_len > aligned_end)
3459 return -EOPNOTSUPP;
3460
3461 ret = btrfs_next_item(root, path);
3462 if (ret < 0) {
3463 return ret;
3464 } else if (ret == 0) {
3465 btrfs_item_key_to_cpu(path->nodes[0], &key,
3466 path->slots[0]);
3467 if (key.objectid == btrfs_ino(dst) &&
3468 key.type == BTRFS_EXTENT_DATA_KEY)
3469 return -EOPNOTSUPP;
3470 }
3471 }
3472
3473copy_inline_extent:
3474 /*
3475 * We have no extent items, or we have an extent at offset 0 which may
3476 * or may not be inlined. All these cases are dealt the same way.
3477 */
3478 if (i_size_read(dst) > datal) {
3479 /*
3480 * If the destination inode has an inline extent...
3481 * This would require copying the data from the source inline
3482 * extent into the beginning of the destination's inline extent.
3483 * But this is really complex, both extents can be compressed
3484 * or just one of them, which would require decompressing and
3485 * re-compressing data (which could increase the new compressed
3486 * size, not allowing the compressed data to fit anymore in an
3487 * inline extent).
3488 * So just don't support this case for now (it should be rare,
3489 * we are not really saving space when cloning inline extents).
3490 */
3491 return -EOPNOTSUPP;
3492 }
3493
3494 btrfs_release_path(path);
3495 ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1);
3496 if (ret)
3497 return ret;
3498 ret = btrfs_insert_empty_item(trans, root, path, new_key, size);
3499 if (ret)
3500 return ret;
3501
3502 if (skip) {
3503 const u32 start = btrfs_file_extent_calc_inline_size(0);
3504
3505 memmove(inline_data + start, inline_data + start + skip, datal);
3506 }
3507
3508 write_extent_buffer(path->nodes[0], inline_data,
3509 btrfs_item_ptr_offset(path->nodes[0],
3510 path->slots[0]),
3511 size);
3512 inode_add_bytes(dst, datal);
3513
3514 return 0;
3515}
3516
Mark Fasheh32b7c682013-08-06 11:42:49 -07003517/**
3518 * btrfs_clone() - clone a range from inode file to another
3519 *
3520 * @src: Inode to clone from
3521 * @inode: Inode to clone to
3522 * @off: Offset within source to start clone from
3523 * @olen: Original length, passed by user, of range to clone
Mark Fasheh1c919a52015-06-30 14:42:08 -07003524 * @olen_aligned: Block-aligned value of olen
Mark Fasheh32b7c682013-08-06 11:42:49 -07003525 * @destoff: Offset within @inode to start clone
Mark Fasheh1c919a52015-06-30 14:42:08 -07003526 * @no_time_update: Whether to update mtime/ctime on the target inode
Mark Fasheh32b7c682013-08-06 11:42:49 -07003527 */
3528static int btrfs_clone(struct inode *src, struct inode *inode,
Filipe Mananaf82a9902014-06-01 01:50:28 +01003529 const u64 off, const u64 olen, const u64 olen_aligned,
Mark Fasheh1c919a52015-06-30 14:42:08 -07003530 const u64 destoff, int no_time_update)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003531{
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003532 struct btrfs_root *root = BTRFS_I(inode)->root;
Mark Fasheh32b7c682013-08-06 11:42:49 -07003533 struct btrfs_path *path = NULL;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003534 struct extent_buffer *leaf;
Mark Fasheh32b7c682013-08-06 11:42:49 -07003535 struct btrfs_trans_handle *trans;
3536 char *buf = NULL;
Yan Zhengae01a0a2008-08-04 23:23:47 -04003537 struct btrfs_key key;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003538 u32 nritems;
3539 int slot;
Yan Zhengae01a0a2008-08-04 23:23:47 -04003540 int ret;
Filipe Mananaf82a9902014-06-01 01:50:28 +01003541 const u64 len = olen_aligned;
Filipe Mananaf82a9902014-06-01 01:50:28 +01003542 u64 last_dest_end = destoff;
Yan Zhengae01a0a2008-08-04 23:23:47 -04003543
3544 ret = -ENOMEM;
David Sterba15351952016-04-11 18:40:08 +02003545 buf = kmalloc(root->nodesize, GFP_KERNEL | __GFP_NOWARN);
3546 if (!buf) {
3547 buf = vmalloc(root->nodesize);
3548 if (!buf)
3549 return ret;
3550 }
Yan Zhengae01a0a2008-08-04 23:23:47 -04003551
3552 path = btrfs_alloc_path();
3553 if (!path) {
David Sterba15351952016-04-11 18:40:08 +02003554 kvfree(buf);
Mark Fasheh32b7c682013-08-06 11:42:49 -07003555 return ret;
Yan Zhengae01a0a2008-08-04 23:23:47 -04003556 }
Mark Fasheh32b7c682013-08-06 11:42:49 -07003557
David Sterbae4058b52015-11-27 16:31:35 +01003558 path->reada = READA_FORWARD;
Sage Weilc5c9cd42008-11-12 14:32:25 -05003559 /* clone data */
Li Zefan33345d012011-04-20 10:31:50 +08003560 key.objectid = btrfs_ino(src);
Yan Zhengae01a0a2008-08-04 23:23:47 -04003561 key.type = BTRFS_EXTENT_DATA_KEY;
Filipe Manana2c463822014-05-31 02:31:05 +01003562 key.offset = off;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003563
3564 while (1) {
Chris Masonde249e62015-04-11 05:09:06 -07003565 u64 next_key_min_offset = key.offset + 1;
Filipe Mananadf858e72015-03-31 14:56:46 +01003566
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003567 /*
3568 * note the key will change type as we walk through the
3569 * tree.
3570 */
Filipe David Borba Mananae4355f32014-01-13 19:35:01 +00003571 path->leave_spinning = 1;
David Sterba362a20c2011-08-01 18:11:57 +02003572 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3573 0, 0);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003574 if (ret < 0)
3575 goto out;
Filipe Manana2c463822014-05-31 02:31:05 +01003576 /*
3577 * First search, if no extent item that starts at offset off was
3578 * found but the previous item is an extent item, it's possible
3579 * it might overlap our target range, therefore process it.
3580 */
3581 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3582 btrfs_item_key_to_cpu(path->nodes[0], &key,
3583 path->slots[0] - 1);
3584 if (key.type == BTRFS_EXTENT_DATA_KEY)
3585 path->slots[0]--;
3586 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003587
Yan Zhengae01a0a2008-08-04 23:23:47 -04003588 nritems = btrfs_header_nritems(path->nodes[0]);
Filipe David Borba Mananae4355f32014-01-13 19:35:01 +00003589process_slot:
Yan Zhengae01a0a2008-08-04 23:23:47 -04003590 if (path->slots[0] >= nritems) {
David Sterba362a20c2011-08-01 18:11:57 +02003591 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003592 if (ret < 0)
3593 goto out;
3594 if (ret > 0)
3595 break;
Yan Zhengae01a0a2008-08-04 23:23:47 -04003596 nritems = btrfs_header_nritems(path->nodes[0]);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003597 }
3598 leaf = path->nodes[0];
3599 slot = path->slots[0];
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003600
Yan Zhengae01a0a2008-08-04 23:23:47 -04003601 btrfs_item_key_to_cpu(leaf, &key, slot);
David Sterba962a2982014-06-04 18:41:45 +02003602 if (key.type > BTRFS_EXTENT_DATA_KEY ||
Li Zefan33345d012011-04-20 10:31:50 +08003603 key.objectid != btrfs_ino(src))
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003604 break;
3605
David Sterba962a2982014-06-04 18:41:45 +02003606 if (key.type == BTRFS_EXTENT_DATA_KEY) {
Sage Weilc5c9cd42008-11-12 14:32:25 -05003607 struct btrfs_file_extent_item *extent;
3608 int type;
Zheng Yan31840ae2008-09-23 13:14:14 -04003609 u32 size;
3610 struct btrfs_key new_key;
Sage Weilc5c9cd42008-11-12 14:32:25 -05003611 u64 disko = 0, diskl = 0;
3612 u64 datao = 0, datal = 0;
3613 u8 comp;
Filipe Mananaf82a9902014-06-01 01:50:28 +01003614 u64 drop_start;
Zheng Yan31840ae2008-09-23 13:14:14 -04003615
Sage Weilc5c9cd42008-11-12 14:32:25 -05003616 extent = btrfs_item_ptr(leaf, slot,
3617 struct btrfs_file_extent_item);
3618 comp = btrfs_file_extent_compression(leaf, extent);
3619 type = btrfs_file_extent_type(leaf, extent);
Chris Masonc8a894d2009-06-27 21:07:03 -04003620 if (type == BTRFS_FILE_EXTENT_REG ||
3621 type == BTRFS_FILE_EXTENT_PREALLOC) {
Chris Masond3977122009-01-05 21:25:51 -05003622 disko = btrfs_file_extent_disk_bytenr(leaf,
3623 extent);
3624 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3625 extent);
Sage Weilc5c9cd42008-11-12 14:32:25 -05003626 datao = btrfs_file_extent_offset(leaf, extent);
Chris Masond3977122009-01-05 21:25:51 -05003627 datal = btrfs_file_extent_num_bytes(leaf,
3628 extent);
Sage Weilc5c9cd42008-11-12 14:32:25 -05003629 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3630 /* take upper bound, may be compressed */
3631 datal = btrfs_file_extent_ram_bytes(leaf,
3632 extent);
3633 }
Zheng Yan31840ae2008-09-23 13:14:14 -04003634
Filipe Manana2c463822014-05-31 02:31:05 +01003635 /*
3636 * The first search might have left us at an extent
3637 * item that ends before our target range's start, can
3638 * happen if we have holes and NO_HOLES feature enabled.
3639 */
3640 if (key.offset + datal <= off) {
Filipe David Borba Mananae4355f32014-01-13 19:35:01 +00003641 path->slots[0]++;
3642 goto process_slot;
Filipe Manana2c463822014-05-31 02:31:05 +01003643 } else if (key.offset >= off + len) {
3644 break;
Filipe David Borba Mananae4355f32014-01-13 19:35:01 +00003645 }
Filipe Mananadf858e72015-03-31 14:56:46 +01003646 next_key_min_offset = key.offset + datal;
Filipe David Borba Mananae4355f32014-01-13 19:35:01 +00003647 size = btrfs_item_size_nr(leaf, slot);
3648 read_extent_buffer(leaf, buf,
3649 btrfs_item_ptr_offset(leaf, slot),
3650 size);
3651
3652 btrfs_release_path(path);
3653 path->leave_spinning = 0;
Sage Weilc5c9cd42008-11-12 14:32:25 -05003654
Zheng Yan31840ae2008-09-23 13:14:14 -04003655 memcpy(&new_key, &key, sizeof(new_key));
Li Zefan33345d012011-04-20 10:31:50 +08003656 new_key.objectid = btrfs_ino(inode);
Li Zefan4d728ec2011-01-26 14:10:43 +08003657 if (off <= key.offset)
3658 new_key.offset = key.offset + destoff - off;
3659 else
3660 new_key.offset = destoff;
Sage Weilc5c9cd42008-11-12 14:32:25 -05003661
Sage Weilb6f34092011-09-20 14:48:51 -04003662 /*
Filipe Mananaf82a9902014-06-01 01:50:28 +01003663 * Deal with a hole that doesn't have an extent item
3664 * that represents it (NO_HOLES feature enabled).
3665 * This hole is either in the middle of the cloning
3666 * range or at the beginning (fully overlaps it or
3667 * partially overlaps it).
3668 */
3669 if (new_key.offset != last_dest_end)
3670 drop_start = last_dest_end;
3671 else
3672 drop_start = new_key.offset;
3673
3674 /*
Sage Weilb6f34092011-09-20 14:48:51 -04003675 * 1 - adjusting old extent (we may have to split it)
3676 * 1 - add new extent
3677 * 1 - inode update
3678 */
3679 trans = btrfs_start_transaction(root, 3);
Yan, Zhenga22285a2010-05-16 10:48:46 -04003680 if (IS_ERR(trans)) {
3681 ret = PTR_ERR(trans);
3682 goto out;
3683 }
3684
Chris Masonc8a894d2009-06-27 21:07:03 -04003685 if (type == BTRFS_FILE_EXTENT_REG ||
3686 type == BTRFS_FILE_EXTENT_PREALLOC) {
Li Zefand72c0842011-09-11 10:52:25 -04003687 /*
3688 * a | --- range to clone ---| b
3689 * | ------------- extent ------------- |
3690 */
3691
Antonio Ospite93915582014-06-04 14:03:48 +02003692 /* subtract range b */
Li Zefand72c0842011-09-11 10:52:25 -04003693 if (key.offset + datal > off + len)
3694 datal = off + len - key.offset;
3695
Antonio Ospite93915582014-06-04 14:03:48 +02003696 /* subtract range a */
Yan, Zhenga22285a2010-05-16 10:48:46 -04003697 if (off > key.offset) {
3698 datao += off - key.offset;
3699 datal -= off - key.offset;
3700 }
3701
Josef Bacik5dc562c2012-08-17 13:14:17 -04003702 ret = btrfs_drop_extents(trans, root, inode,
Filipe Mananaf82a9902014-06-01 01:50:28 +01003703 drop_start,
Yan, Zhenga22285a2010-05-16 10:48:46 -04003704 new_key.offset + datal,
Josef Bacik26714852012-08-29 12:24:27 -04003705 1);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003706 if (ret) {
David Sterba3f9e3df2014-04-15 18:50:17 +02003707 if (ret != -EOPNOTSUPP)
Liu Bo00fdf132014-03-10 18:56:07 +08003708 btrfs_abort_transaction(trans,
Jeff Mahoney66642832016-06-10 18:19:25 -04003709 ret);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003710 btrfs_end_transaction(trans, root);
3711 goto out;
3712 }
Yan, Zhenga22285a2010-05-16 10:48:46 -04003713
Sage Weilc5c9cd42008-11-12 14:32:25 -05003714 ret = btrfs_insert_empty_item(trans, root, path,
3715 &new_key, size);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003716 if (ret) {
Jeff Mahoney66642832016-06-10 18:19:25 -04003717 btrfs_abort_transaction(trans, ret);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003718 btrfs_end_transaction(trans, root);
3719 goto out;
3720 }
Sage Weilc5c9cd42008-11-12 14:32:25 -05003721
3722 leaf = path->nodes[0];
3723 slot = path->slots[0];
3724 write_extent_buffer(leaf, buf,
3725 btrfs_item_ptr_offset(leaf, slot),
3726 size);
3727
3728 extent = btrfs_item_ptr(leaf, slot,
3729 struct btrfs_file_extent_item);
Sage Weilc5c9cd42008-11-12 14:32:25 -05003730
Sage Weilc5c9cd42008-11-12 14:32:25 -05003731 /* disko == 0 means it's a hole */
3732 if (!disko)
3733 datao = 0;
Sage Weilc5c9cd42008-11-12 14:32:25 -05003734
3735 btrfs_set_file_extent_offset(leaf, extent,
3736 datao);
3737 btrfs_set_file_extent_num_bytes(leaf, extent,
3738 datal);
Josef Bacikfcebe452014-05-13 17:30:47 -07003739
Sage Weilc5c9cd42008-11-12 14:32:25 -05003740 if (disko) {
3741 inode_add_bytes(inode, datal);
3742 ret = btrfs_inc_extent_ref(trans, root,
Yan Zheng5d4f98a2009-06-10 10:45:14 -04003743 disko, diskl, 0,
3744 root->root_key.objectid,
Li Zefan33345d012011-04-20 10:31:50 +08003745 btrfs_ino(inode),
Filipe Mananab06c4bf2015-10-23 07:52:54 +01003746 new_key.offset - datao);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003747 if (ret) {
3748 btrfs_abort_transaction(trans,
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003749 ret);
3750 btrfs_end_transaction(trans,
3751 root);
3752 goto out;
3753
3754 }
Sage Weilc5c9cd42008-11-12 14:32:25 -05003755 }
3756 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3757 u64 skip = 0;
3758 u64 trim = 0;
Filipe Mananaed958762015-07-14 16:09:39 +01003759
Sage Weilc5c9cd42008-11-12 14:32:25 -05003760 if (off > key.offset) {
3761 skip = off - key.offset;
3762 new_key.offset += skip;
3763 }
Chris Masond3977122009-01-05 21:25:51 -05003764
Liu Boaa42ffd2012-09-18 03:52:23 -06003765 if (key.offset + datal > off + len)
3766 trim = key.offset + datal - (off + len);
Chris Masond3977122009-01-05 21:25:51 -05003767
Sage Weilc5c9cd42008-11-12 14:32:25 -05003768 if (comp && (skip || trim)) {
Sage Weilc5c9cd42008-11-12 14:32:25 -05003769 ret = -EINVAL;
Yan, Zhenga22285a2010-05-16 10:48:46 -04003770 btrfs_end_transaction(trans, root);
Sage Weilc5c9cd42008-11-12 14:32:25 -05003771 goto out;
3772 }
3773 size -= skip + trim;
3774 datal -= skip + trim;
Yan, Zhenga22285a2010-05-16 10:48:46 -04003775
Filipe Manana8039d872015-10-13 15:15:00 +01003776 ret = clone_copy_inline_extent(src, inode,
3777 trans, path,
3778 &new_key,
3779 drop_start,
3780 datal,
3781 skip, size, buf);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003782 if (ret) {
David Sterba3f9e3df2014-04-15 18:50:17 +02003783 if (ret != -EOPNOTSUPP)
Chris Mason3a29bc02014-04-07 07:10:40 -07003784 btrfs_abort_transaction(trans,
Filipe Manana8039d872015-10-13 15:15:00 +01003785 ret);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003786 btrfs_end_transaction(trans, root);
3787 goto out;
3788 }
Sage Weilc5c9cd42008-11-12 14:32:25 -05003789 leaf = path->nodes[0];
3790 slot = path->slots[0];
Sage Weilc5c9cd42008-11-12 14:32:25 -05003791 }
3792
Filipe Manana7ffbb592014-06-09 03:48:05 +01003793 /* If we have an implicit hole (NO_HOLES feature). */
3794 if (drop_start < new_key.offset)
3795 clone_update_extent_map(inode, trans,
Filipe Manana14f59792014-06-29 21:45:40 +01003796 NULL, drop_start,
Filipe Manana7ffbb592014-06-09 03:48:05 +01003797 new_key.offset - drop_start);
3798
Filipe Manana14f59792014-06-29 21:45:40 +01003799 clone_update_extent_map(inode, trans, path, 0, 0);
Filipe Manana7ffbb592014-06-09 03:48:05 +01003800
Sage Weilc5c9cd42008-11-12 14:32:25 -05003801 btrfs_mark_buffer_dirty(leaf);
David Sterbab3b4aa72011-04-21 01:20:15 +02003802 btrfs_release_path(path);
Sage Weilc5c9cd42008-11-12 14:32:25 -05003803
Filipe Manana62e23902014-08-08 02:47:06 +01003804 last_dest_end = ALIGN(new_key.offset + datal,
3805 root->sectorsize);
Filipe Mananaf82a9902014-06-01 01:50:28 +01003806 ret = clone_finish_inode_update(trans, inode,
3807 last_dest_end,
Mark Fasheh1c919a52015-06-30 14:42:08 -07003808 destoff, olen,
3809 no_time_update);
Filipe Mananaf82a9902014-06-01 01:50:28 +01003810 if (ret)
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003811 goto out;
Filipe Manana2c463822014-05-31 02:31:05 +01003812 if (new_key.offset + datal >= destoff + len)
3813 break;
Yan, Zhenga22285a2010-05-16 10:48:46 -04003814 }
David Sterbab3b4aa72011-04-21 01:20:15 +02003815 btrfs_release_path(path);
Filipe Mananadf858e72015-03-31 14:56:46 +01003816 key.offset = next_key_min_offset;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003817 }
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003818 ret = 0;
Mark Fasheh32b7c682013-08-06 11:42:49 -07003819
Filipe Mananaf82a9902014-06-01 01:50:28 +01003820 if (last_dest_end < destoff + len) {
3821 /*
3822 * We have an implicit hole (NO_HOLES feature is enabled) that
3823 * fully or partially overlaps our cloning range at its end.
3824 */
3825 btrfs_release_path(path);
3826
3827 /*
3828 * 1 - remove extent(s)
3829 * 1 - inode update
3830 */
3831 trans = btrfs_start_transaction(root, 2);
3832 if (IS_ERR(trans)) {
3833 ret = PTR_ERR(trans);
3834 goto out;
3835 }
3836 ret = btrfs_drop_extents(trans, root, inode,
3837 last_dest_end, destoff + len, 1);
3838 if (ret) {
3839 if (ret != -EOPNOTSUPP)
Jeff Mahoney66642832016-06-10 18:19:25 -04003840 btrfs_abort_transaction(trans, ret);
Filipe Mananaf82a9902014-06-01 01:50:28 +01003841 btrfs_end_transaction(trans, root);
3842 goto out;
3843 }
Filipe Manana14f59792014-06-29 21:45:40 +01003844 clone_update_extent_map(inode, trans, NULL, last_dest_end,
3845 destoff + len - last_dest_end);
Filipe Mananaf82a9902014-06-01 01:50:28 +01003846 ret = clone_finish_inode_update(trans, inode, destoff + len,
Mark Fasheh1c919a52015-06-30 14:42:08 -07003847 destoff, olen, no_time_update);
Filipe Mananaf82a9902014-06-01 01:50:28 +01003848 }
3849
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003850out:
Mark Fasheh32b7c682013-08-06 11:42:49 -07003851 btrfs_free_path(path);
David Sterba15351952016-04-11 18:40:08 +02003852 kvfree(buf);
Mark Fasheh32b7c682013-08-06 11:42:49 -07003853 return ret;
3854}
3855
Zach Brown3db11b22015-11-10 16:53:32 -05003856static noinline int btrfs_clone_files(struct file *file, struct file *file_src,
3857 u64 off, u64 olen, u64 destoff)
Mark Fasheh32b7c682013-08-06 11:42:49 -07003858{
Al Viro54563d42013-09-01 15:57:51 -04003859 struct inode *inode = file_inode(file);
Zach Brown3db11b22015-11-10 16:53:32 -05003860 struct inode *src = file_inode(file_src);
Mark Fasheh32b7c682013-08-06 11:42:49 -07003861 struct btrfs_root *root = BTRFS_I(inode)->root;
Mark Fasheh32b7c682013-08-06 11:42:49 -07003862 int ret;
3863 u64 len = olen;
3864 u64 bs = root->fs_info->sb->s_blocksize;
Zach Brown3db11b22015-11-10 16:53:32 -05003865 int same_inode = src == inode;
Mark Fasheh32b7c682013-08-06 11:42:49 -07003866
3867 /*
3868 * TODO:
3869 * - split compressed inline extents. annoying: we need to
3870 * decompress into destination's address_space (the file offset
3871 * may change, so source mapping won't do), then recompress (or
3872 * otherwise reinsert) a subrange.
Liu Bo00fdf132014-03-10 18:56:07 +08003873 *
3874 * - split destination inode's inline extents. The inline extents can
3875 * be either compressed or non-compressed.
Mark Fasheh32b7c682013-08-06 11:42:49 -07003876 */
3877
Mark Fasheh32b7c682013-08-06 11:42:49 -07003878 if (btrfs_root_readonly(root))
3879 return -EROFS;
3880
Zach Brown3db11b22015-11-10 16:53:32 -05003881 if (file_src->f_path.mnt != file->f_path.mnt ||
3882 src->i_sb != inode->i_sb)
3883 return -EXDEV;
Mark Fasheh32b7c682013-08-06 11:42:49 -07003884
3885 /* don't make the dst file partly checksummed */
3886 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3887 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
Zach Brown3db11b22015-11-10 16:53:32 -05003888 return -EINVAL;
Mark Fasheh32b7c682013-08-06 11:42:49 -07003889
Mark Fasheh32b7c682013-08-06 11:42:49 -07003890 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
Zach Brown3db11b22015-11-10 16:53:32 -05003891 return -EISDIR;
Mark Fasheh32b7c682013-08-06 11:42:49 -07003892
3893 if (!same_inode) {
Mark Fasheh293a8482015-06-30 14:42:06 -07003894 btrfs_double_inode_lock(src, inode);
Mark Fasheh32b7c682013-08-06 11:42:49 -07003895 } else {
Al Viro59551022016-01-22 15:40:57 -05003896 inode_lock(src);
Mark Fasheh32b7c682013-08-06 11:42:49 -07003897 }
3898
3899 /* determine range to clone */
3900 ret = -EINVAL;
3901 if (off + len > src->i_size || off + len < off)
3902 goto out_unlock;
3903 if (len == 0)
3904 olen = len = src->i_size - off;
3905 /* if we extend to eof, continue to block boundary */
3906 if (off + len == src->i_size)
3907 len = ALIGN(src->i_size, bs) - off;
3908
Filipe Mananaccccf3d62015-03-30 18:23:59 +01003909 if (len == 0) {
3910 ret = 0;
3911 goto out_unlock;
3912 }
3913
Mark Fasheh32b7c682013-08-06 11:42:49 -07003914 /* verify the end result is block aligned */
3915 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3916 !IS_ALIGNED(destoff, bs))
3917 goto out_unlock;
3918
3919 /* verify if ranges are overlapped within the same file */
3920 if (same_inode) {
3921 if (destoff + len > off && destoff < off + len)
3922 goto out_unlock;
3923 }
3924
3925 if (destoff > inode->i_size) {
3926 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3927 if (ret)
3928 goto out_unlock;
3929 }
3930
Filipe Mananac125b8b2014-05-23 05:03:34 +01003931 /*
3932 * Lock the target range too. Right after we replace the file extent
3933 * items in the fs tree (which now point to the cloned data), we might
3934 * have a worker replace them with extent items relative to a write
3935 * operation that was issued before this clone operation (i.e. confront
3936 * with inode.c:btrfs_finish_ordered_io).
3937 */
3938 if (same_inode) {
3939 u64 lock_start = min_t(u64, off, destoff);
3940 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
Mark Fasheh32b7c682013-08-06 11:42:49 -07003941
Filipe Mananae0bd70c2016-01-27 10:20:58 +00003942 ret = lock_extent_range(src, lock_start, lock_len, true);
Filipe Mananac125b8b2014-05-23 05:03:34 +01003943 } else {
Filipe Mananae0bd70c2016-01-27 10:20:58 +00003944 ret = btrfs_double_extent_lock(src, off, inode, destoff, len,
3945 true);
3946 }
3947 ASSERT(ret == 0);
3948 if (WARN_ON(ret)) {
3949 /* ranges in the io trees already unlocked */
3950 goto out_unlock;
Filipe Mananac125b8b2014-05-23 05:03:34 +01003951 }
Mark Fasheh32b7c682013-08-06 11:42:49 -07003952
Mark Fasheh1c919a52015-06-30 14:42:08 -07003953 ret = btrfs_clone(src, inode, off, olen, len, destoff, 0);
Mark Fasheh32b7c682013-08-06 11:42:49 -07003954
Filipe Mananac125b8b2014-05-23 05:03:34 +01003955 if (same_inode) {
3956 u64 lock_start = min_t(u64, off, destoff);
3957 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3958
3959 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3960 } else {
Mark Fasheh293a8482015-06-30 14:42:06 -07003961 btrfs_double_extent_unlock(src, off, inode, destoff, len);
Filipe Mananac125b8b2014-05-23 05:03:34 +01003962 }
3963 /*
3964 * Truncate page cache pages so that future reads will see the cloned
3965 * data immediately and not the previous data.
3966 */
Chandan Rajendra65bfa652016-01-21 15:56:04 +05303967 truncate_inode_pages_range(&inode->i_data,
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03003968 round_down(destoff, PAGE_SIZE),
3969 round_up(destoff + len, PAGE_SIZE) - 1);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003970out_unlock:
Mark Fasheh293a8482015-06-30 14:42:06 -07003971 if (!same_inode)
3972 btrfs_double_inode_unlock(src, inode);
3973 else
Al Viro59551022016-01-22 15:40:57 -05003974 inode_unlock(src);
Zach Brown3db11b22015-11-10 16:53:32 -05003975 return ret;
3976}
3977
3978ssize_t btrfs_copy_file_range(struct file *file_in, loff_t pos_in,
3979 struct file *file_out, loff_t pos_out,
3980 size_t len, unsigned int flags)
3981{
3982 ssize_t ret;
3983
3984 ret = btrfs_clone_files(file_out, file_in, pos_in, len, pos_out);
3985 if (ret == 0)
3986 ret = len;
3987 return ret;
3988}
3989
Christoph Hellwig04b38d62015-12-03 12:59:50 +01003990int btrfs_clone_file_range(struct file *src_file, loff_t off,
3991 struct file *dst_file, loff_t destoff, u64 len)
Zach Brown3db11b22015-11-10 16:53:32 -05003992{
Christoph Hellwig04b38d62015-12-03 12:59:50 +01003993 return btrfs_clone_files(dst_file, src_file, off, len, destoff);
Sage Weilc5c9cd42008-11-12 14:32:25 -05003994}
3995
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04003996/*
3997 * there are many ways the trans_start and trans_end ioctls can lead
3998 * to deadlocks. They should only be used by applications that
3999 * basically own the machine, and have a very in depth understanding
4000 * of all the possible deadlocks and enospc problems.
4001 */
Christoph Hellwigb2950862008-12-02 09:54:17 -05004002static long btrfs_ioctl_trans_start(struct file *file)
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04004003{
Al Viro496ad9a2013-01-23 17:07:38 -05004004 struct inode *inode = file_inode(file);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04004005 struct btrfs_root *root = BTRFS_I(inode)->root;
4006 struct btrfs_trans_handle *trans;
Sage Weil1ab86ae2009-09-29 18:38:44 -04004007 int ret;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04004008
Sage Weil1ab86ae2009-09-29 18:38:44 -04004009 ret = -EPERM;
Christoph Hellwigdf5b5522008-06-11 21:53:58 -04004010 if (!capable(CAP_SYS_ADMIN))
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04004011 goto out;
Sage Weil1ab86ae2009-09-29 18:38:44 -04004012
4013 ret = -EINPROGRESS;
4014 if (file->private_data)
4015 goto out;
Sage Weil9ca9ee02008-08-04 10:41:27 -04004016
Li Zefanb83cc962010-12-20 16:04:08 +08004017 ret = -EROFS;
4018 if (btrfs_root_readonly(root))
4019 goto out;
4020
Al Viroa561be72011-11-23 11:57:51 -05004021 ret = mnt_want_write_file(file);
Yan Zhengc146afa2008-11-12 14:34:12 -05004022 if (ret)
4023 goto out;
4024
Josef Bacika4abeea2011-04-11 17:25:13 -04004025 atomic_inc(&root->fs_info->open_ioctl_trans);
Sage Weil9ca9ee02008-08-04 10:41:27 -04004026
Sage Weil1ab86ae2009-09-29 18:38:44 -04004027 ret = -ENOMEM;
Josef Bacik7a7eaa42011-04-13 12:54:33 -04004028 trans = btrfs_start_ioctl_transaction(root);
Tsutomu Itohabd30bb2011-01-24 00:57:10 +00004029 if (IS_ERR(trans))
Sage Weil1ab86ae2009-09-29 18:38:44 -04004030 goto out_drop;
4031
4032 file->private_data = trans;
4033 return 0;
4034
4035out_drop:
Josef Bacika4abeea2011-04-11 17:25:13 -04004036 atomic_dec(&root->fs_info->open_ioctl_trans);
Al Viro2a79f172011-12-09 08:06:57 -05004037 mnt_drop_write_file(file);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04004038out:
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04004039 return ret;
4040}
4041
Josef Bacik6ef5ed02009-12-11 21:11:29 +00004042static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
4043{
Al Viro496ad9a2013-01-23 17:07:38 -05004044 struct inode *inode = file_inode(file);
Josef Bacik6ef5ed02009-12-11 21:11:29 +00004045 struct btrfs_root *root = BTRFS_I(inode)->root;
4046 struct btrfs_root *new_root;
4047 struct btrfs_dir_item *di;
4048 struct btrfs_trans_handle *trans;
4049 struct btrfs_path *path;
4050 struct btrfs_key location;
4051 struct btrfs_disk_key disk_key;
Josef Bacik6ef5ed02009-12-11 21:11:29 +00004052 u64 objectid = 0;
4053 u64 dir_id;
Miao Xie3c04ce02012-11-26 08:43:07 +00004054 int ret;
Josef Bacik6ef5ed02009-12-11 21:11:29 +00004055
4056 if (!capable(CAP_SYS_ADMIN))
4057 return -EPERM;
4058
Miao Xie3c04ce02012-11-26 08:43:07 +00004059 ret = mnt_want_write_file(file);
4060 if (ret)
4061 return ret;
4062
4063 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
4064 ret = -EFAULT;
4065 goto out;
4066 }
Josef Bacik6ef5ed02009-12-11 21:11:29 +00004067
4068 if (!objectid)
chandan1cecf572013-09-13 19:34:10 +05304069 objectid = BTRFS_FS_TREE_OBJECTID;
Josef Bacik6ef5ed02009-12-11 21:11:29 +00004070
4071 location.objectid = objectid;
4072 location.type = BTRFS_ROOT_ITEM_KEY;
4073 location.offset = (u64)-1;
4074
4075 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
Miao Xie3c04ce02012-11-26 08:43:07 +00004076 if (IS_ERR(new_root)) {
4077 ret = PTR_ERR(new_root);
4078 goto out;
4079 }
Josef Bacik6ef5ed02009-12-11 21:11:29 +00004080
Josef Bacik6ef5ed02009-12-11 21:11:29 +00004081 path = btrfs_alloc_path();
Miao Xie3c04ce02012-11-26 08:43:07 +00004082 if (!path) {
4083 ret = -ENOMEM;
4084 goto out;
4085 }
Josef Bacik6ef5ed02009-12-11 21:11:29 +00004086 path->leave_spinning = 1;
4087
4088 trans = btrfs_start_transaction(root, 1);
Tsutomu Itoh98d5dc12011-01-20 06:19:37 +00004089 if (IS_ERR(trans)) {
Josef Bacik6ef5ed02009-12-11 21:11:29 +00004090 btrfs_free_path(path);
Miao Xie3c04ce02012-11-26 08:43:07 +00004091 ret = PTR_ERR(trans);
4092 goto out;
Josef Bacik6ef5ed02009-12-11 21:11:29 +00004093 }
4094
David Sterba6c417612011-04-13 15:41:04 +02004095 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
Josef Bacik6ef5ed02009-12-11 21:11:29 +00004096 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
4097 dir_id, "default", 7, 1);
Dan Carpentercf1e99a2010-05-29 09:47:24 +00004098 if (IS_ERR_OR_NULL(di)) {
Josef Bacik6ef5ed02009-12-11 21:11:29 +00004099 btrfs_free_path(path);
4100 btrfs_end_transaction(trans, root);
Jeff Mahoney5d163e02016-09-20 10:05:00 -04004101 btrfs_err(new_root->fs_info,
4102 "Umm, you don't have the default diritem, this isn't going to work");
Miao Xie3c04ce02012-11-26 08:43:07 +00004103 ret = -ENOENT;
4104 goto out;
Josef Bacik6ef5ed02009-12-11 21:11:29 +00004105 }
4106
4107 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
4108 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
4109 btrfs_mark_buffer_dirty(path->nodes[0]);
4110 btrfs_free_path(path);
4111
Mitch Harder2b0ce2c2012-07-24 11:58:43 -06004112 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
Josef Bacik6ef5ed02009-12-11 21:11:29 +00004113 btrfs_end_transaction(trans, root);
Miao Xie3c04ce02012-11-26 08:43:07 +00004114out:
4115 mnt_drop_write_file(file);
4116 return ret;
Josef Bacik6ef5ed02009-12-11 21:11:29 +00004117}
4118
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02004119void btrfs_get_block_group_info(struct list_head *groups_list,
4120 struct btrfs_ioctl_space_info *space)
Josef Bacikbf5fc0932010-09-29 11:22:36 -04004121{
4122 struct btrfs_block_group_cache *block_group;
4123
4124 space->total_bytes = 0;
4125 space->used_bytes = 0;
4126 space->flags = 0;
4127 list_for_each_entry(block_group, groups_list, list) {
4128 space->flags = block_group->flags;
4129 space->total_bytes += block_group->key.offset;
4130 space->used_bytes +=
4131 btrfs_block_group_used(&block_group->item);
4132 }
4133}
4134
Eric Sandeen48a3b632013-04-25 20:41:01 +00004135static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
Josef Bacik1406e432010-01-13 18:19:06 +00004136{
4137 struct btrfs_ioctl_space_args space_args;
4138 struct btrfs_ioctl_space_info space;
4139 struct btrfs_ioctl_space_info *dest;
Chris Mason7fde62b2010-03-16 15:40:10 -04004140 struct btrfs_ioctl_space_info *dest_orig;
Daniel J Blueman13f26962011-04-11 15:56:31 +00004141 struct btrfs_ioctl_space_info __user *user_dest;
Josef Bacik1406e432010-01-13 18:19:06 +00004142 struct btrfs_space_info *info;
Josef Bacikbf5fc0932010-09-29 11:22:36 -04004143 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
4144 BTRFS_BLOCK_GROUP_SYSTEM,
4145 BTRFS_BLOCK_GROUP_METADATA,
4146 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
4147 int num_types = 4;
Chris Mason7fde62b2010-03-16 15:40:10 -04004148 int alloc_size;
Josef Bacik1406e432010-01-13 18:19:06 +00004149 int ret = 0;
Dan Rosenberg51788b12011-02-14 16:04:23 -05004150 u64 slot_count = 0;
Josef Bacikbf5fc0932010-09-29 11:22:36 -04004151 int i, c;
Josef Bacik1406e432010-01-13 18:19:06 +00004152
4153 if (copy_from_user(&space_args,
4154 (struct btrfs_ioctl_space_args __user *)arg,
4155 sizeof(space_args)))
4156 return -EFAULT;
4157
Josef Bacikbf5fc0932010-09-29 11:22:36 -04004158 for (i = 0; i < num_types; i++) {
4159 struct btrfs_space_info *tmp;
4160
4161 info = NULL;
4162 rcu_read_lock();
4163 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
4164 list) {
4165 if (tmp->flags == types[i]) {
4166 info = tmp;
4167 break;
4168 }
4169 }
4170 rcu_read_unlock();
4171
4172 if (!info)
4173 continue;
4174
4175 down_read(&info->groups_sem);
4176 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4177 if (!list_empty(&info->block_groups[c]))
4178 slot_count++;
4179 }
4180 up_read(&info->groups_sem);
4181 }
Josef Bacik1406e432010-01-13 18:19:06 +00004182
David Sterba36523e952014-02-07 14:34:12 +01004183 /*
4184 * Global block reserve, exported as a space_info
4185 */
4186 slot_count++;
4187
Chris Mason7fde62b2010-03-16 15:40:10 -04004188 /* space_slots == 0 means they are asking for a count */
4189 if (space_args.space_slots == 0) {
4190 space_args.total_spaces = slot_count;
4191 goto out;
4192 }
Josef Bacikbf5fc0932010-09-29 11:22:36 -04004193
Dan Rosenberg51788b12011-02-14 16:04:23 -05004194 slot_count = min_t(u64, space_args.space_slots, slot_count);
Josef Bacikbf5fc0932010-09-29 11:22:36 -04004195
Chris Mason7fde62b2010-03-16 15:40:10 -04004196 alloc_size = sizeof(*dest) * slot_count;
Josef Bacikbf5fc0932010-09-29 11:22:36 -04004197
Chris Mason7fde62b2010-03-16 15:40:10 -04004198 /* we generally have at most 6 or so space infos, one for each raid
4199 * level. So, a whole page should be more than enough for everyone
4200 */
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03004201 if (alloc_size > PAGE_SIZE)
Chris Mason7fde62b2010-03-16 15:40:10 -04004202 return -ENOMEM;
4203
4204 space_args.total_spaces = 0;
David Sterba8d2db782015-11-04 15:38:29 +01004205 dest = kmalloc(alloc_size, GFP_KERNEL);
Chris Mason7fde62b2010-03-16 15:40:10 -04004206 if (!dest)
4207 return -ENOMEM;
4208 dest_orig = dest;
4209
4210 /* now we have a buffer to copy into */
Josef Bacikbf5fc0932010-09-29 11:22:36 -04004211 for (i = 0; i < num_types; i++) {
4212 struct btrfs_space_info *tmp;
Chris Mason7fde62b2010-03-16 15:40:10 -04004213
Dan Rosenberg51788b12011-02-14 16:04:23 -05004214 if (!slot_count)
4215 break;
4216
Josef Bacikbf5fc0932010-09-29 11:22:36 -04004217 info = NULL;
4218 rcu_read_lock();
4219 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
4220 list) {
4221 if (tmp->flags == types[i]) {
4222 info = tmp;
4223 break;
4224 }
4225 }
4226 rcu_read_unlock();
Chris Mason7fde62b2010-03-16 15:40:10 -04004227
Josef Bacikbf5fc0932010-09-29 11:22:36 -04004228 if (!info)
4229 continue;
4230 down_read(&info->groups_sem);
4231 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4232 if (!list_empty(&info->block_groups[c])) {
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02004233 btrfs_get_block_group_info(
4234 &info->block_groups[c], &space);
Josef Bacikbf5fc0932010-09-29 11:22:36 -04004235 memcpy(dest, &space, sizeof(space));
4236 dest++;
4237 space_args.total_spaces++;
Dan Rosenberg51788b12011-02-14 16:04:23 -05004238 slot_count--;
Josef Bacikbf5fc0932010-09-29 11:22:36 -04004239 }
Dan Rosenberg51788b12011-02-14 16:04:23 -05004240 if (!slot_count)
4241 break;
Josef Bacikbf5fc0932010-09-29 11:22:36 -04004242 }
4243 up_read(&info->groups_sem);
Josef Bacik1406e432010-01-13 18:19:06 +00004244 }
Josef Bacik1406e432010-01-13 18:19:06 +00004245
David Sterba36523e952014-02-07 14:34:12 +01004246 /*
4247 * Add global block reserve
4248 */
4249 if (slot_count) {
4250 struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv;
4251
4252 spin_lock(&block_rsv->lock);
4253 space.total_bytes = block_rsv->size;
4254 space.used_bytes = block_rsv->size - block_rsv->reserved;
4255 spin_unlock(&block_rsv->lock);
4256 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
4257 memcpy(dest, &space, sizeof(space));
4258 space_args.total_spaces++;
4259 }
4260
Daniel J Blueman2eec6c82012-04-26 00:37:14 +08004261 user_dest = (struct btrfs_ioctl_space_info __user *)
Chris Mason7fde62b2010-03-16 15:40:10 -04004262 (arg + sizeof(struct btrfs_ioctl_space_args));
4263
4264 if (copy_to_user(user_dest, dest_orig, alloc_size))
4265 ret = -EFAULT;
4266
4267 kfree(dest_orig);
4268out:
4269 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
Josef Bacik1406e432010-01-13 18:19:06 +00004270 ret = -EFAULT;
4271
4272 return ret;
4273}
4274
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04004275/*
4276 * there are many ways the trans_start and trans_end ioctls can lead
4277 * to deadlocks. They should only be used by applications that
4278 * basically own the machine, and have a very in depth understanding
4279 * of all the possible deadlocks and enospc problems.
4280 */
4281long btrfs_ioctl_trans_end(struct file *file)
4282{
Al Viro496ad9a2013-01-23 17:07:38 -05004283 struct inode *inode = file_inode(file);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04004284 struct btrfs_root *root = BTRFS_I(inode)->root;
4285 struct btrfs_trans_handle *trans;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04004286
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04004287 trans = file->private_data;
Sage Weil1ab86ae2009-09-29 18:38:44 -04004288 if (!trans)
4289 return -EINVAL;
Christoph Hellwigb2141072008-09-05 16:43:31 -04004290 file->private_data = NULL;
Sage Weil9ca9ee02008-08-04 10:41:27 -04004291
Sage Weil1ab86ae2009-09-29 18:38:44 -04004292 btrfs_end_transaction(trans, root);
4293
Josef Bacika4abeea2011-04-11 17:25:13 -04004294 atomic_dec(&root->fs_info->open_ioctl_trans);
Sage Weil9ca9ee02008-08-04 10:41:27 -04004295
Al Viro2a79f172011-12-09 08:06:57 -05004296 mnt_drop_write_file(file);
Sage Weil1ab86ae2009-09-29 18:38:44 -04004297 return 0;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04004298}
4299
Miao Xie9a8c28b2012-11-26 08:40:43 +00004300static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4301 void __user *argp)
Sage Weil46204592010-10-29 15:41:32 -04004302{
Sage Weil46204592010-10-29 15:41:32 -04004303 struct btrfs_trans_handle *trans;
4304 u64 transid;
Tsutomu Itohdb5b4932011-03-23 08:14:16 +00004305 int ret;
Sage Weil46204592010-10-29 15:41:32 -04004306
Miao Xied4edf392013-02-20 09:17:06 +00004307 trans = btrfs_attach_transaction_barrier(root);
Miao Xieff7c1d32012-11-26 08:41:29 +00004308 if (IS_ERR(trans)) {
4309 if (PTR_ERR(trans) != -ENOENT)
4310 return PTR_ERR(trans);
4311
4312 /* No running transaction, don't bother */
4313 transid = root->fs_info->last_trans_committed;
4314 goto out;
4315 }
Sage Weil46204592010-10-29 15:41:32 -04004316 transid = trans->transid;
Tsutomu Itohdb5b4932011-03-23 08:14:16 +00004317 ret = btrfs_commit_transaction_async(trans, root, 0);
Tsutomu Itoh8b2b2d32011-04-04 01:52:13 +00004318 if (ret) {
4319 btrfs_end_transaction(trans, root);
Tsutomu Itohdb5b4932011-03-23 08:14:16 +00004320 return ret;
Tsutomu Itoh8b2b2d32011-04-04 01:52:13 +00004321 }
Miao Xieff7c1d32012-11-26 08:41:29 +00004322out:
Sage Weil46204592010-10-29 15:41:32 -04004323 if (argp)
4324 if (copy_to_user(argp, &transid, sizeof(transid)))
4325 return -EFAULT;
4326 return 0;
4327}
4328
Miao Xie9a8c28b2012-11-26 08:40:43 +00004329static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
4330 void __user *argp)
Sage Weil46204592010-10-29 15:41:32 -04004331{
Sage Weil46204592010-10-29 15:41:32 -04004332 u64 transid;
4333
4334 if (argp) {
4335 if (copy_from_user(&transid, argp, sizeof(transid)))
4336 return -EFAULT;
4337 } else {
4338 transid = 0; /* current trans */
4339 }
4340 return btrfs_wait_for_commit(root, transid);
4341}
4342
Miao Xieb8e95482012-11-26 08:48:01 +00004343static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
Jan Schmidt475f6382011-03-11 15:41:01 +01004344{
Al Viro496ad9a2013-01-23 17:07:38 -05004345 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Jan Schmidt475f6382011-03-11 15:41:01 +01004346 struct btrfs_ioctl_scrub_args *sa;
Miao Xieb8e95482012-11-26 08:48:01 +00004347 int ret;
Jan Schmidt475f6382011-03-11 15:41:01 +01004348
4349 if (!capable(CAP_SYS_ADMIN))
4350 return -EPERM;
4351
4352 sa = memdup_user(arg, sizeof(*sa));
4353 if (IS_ERR(sa))
4354 return PTR_ERR(sa);
4355
Miao Xieb8e95482012-11-26 08:48:01 +00004356 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4357 ret = mnt_want_write_file(file);
4358 if (ret)
4359 goto out;
4360 }
4361
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01004362 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
Stefan Behrens63a212a2012-11-05 18:29:28 +01004363 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4364 0);
Jan Schmidt475f6382011-03-11 15:41:01 +01004365
4366 if (copy_to_user(arg, sa, sizeof(*sa)))
4367 ret = -EFAULT;
4368
Miao Xieb8e95482012-11-26 08:48:01 +00004369 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4370 mnt_drop_write_file(file);
4371out:
Jan Schmidt475f6382011-03-11 15:41:01 +01004372 kfree(sa);
4373 return ret;
4374}
4375
4376static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
4377{
4378 if (!capable(CAP_SYS_ADMIN))
4379 return -EPERM;
4380
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01004381 return btrfs_scrub_cancel(root->fs_info);
Jan Schmidt475f6382011-03-11 15:41:01 +01004382}
4383
4384static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
4385 void __user *arg)
4386{
4387 struct btrfs_ioctl_scrub_args *sa;
4388 int ret;
4389
4390 if (!capable(CAP_SYS_ADMIN))
4391 return -EPERM;
4392
4393 sa = memdup_user(arg, sizeof(*sa));
4394 if (IS_ERR(sa))
4395 return PTR_ERR(sa);
4396
4397 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
4398
4399 if (copy_to_user(arg, sa, sizeof(*sa)))
4400 ret = -EFAULT;
4401
4402 kfree(sa);
4403 return ret;
4404}
4405
Stefan Behrensc11d2c22012-05-25 16:06:09 +02004406static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
David Sterbab27f7c02012-06-22 06:30:39 -06004407 void __user *arg)
Stefan Behrensc11d2c22012-05-25 16:06:09 +02004408{
4409 struct btrfs_ioctl_get_dev_stats *sa;
4410 int ret;
4411
Stefan Behrensc11d2c22012-05-25 16:06:09 +02004412 sa = memdup_user(arg, sizeof(*sa));
4413 if (IS_ERR(sa))
4414 return PTR_ERR(sa);
4415
David Sterbab27f7c02012-06-22 06:30:39 -06004416 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4417 kfree(sa);
4418 return -EPERM;
4419 }
4420
4421 ret = btrfs_get_dev_stats(root, sa);
Stefan Behrensc11d2c22012-05-25 16:06:09 +02004422
4423 if (copy_to_user(arg, sa, sizeof(*sa)))
4424 ret = -EFAULT;
4425
4426 kfree(sa);
4427 return ret;
4428}
4429
Stefan Behrens3f6bcfb2012-11-06 15:08:53 +01004430static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
4431{
4432 struct btrfs_ioctl_dev_replace_args *p;
4433 int ret;
4434
4435 if (!capable(CAP_SYS_ADMIN))
4436 return -EPERM;
4437
4438 p = memdup_user(arg, sizeof(*p));
4439 if (IS_ERR(p))
4440 return PTR_ERR(p);
4441
4442 switch (p->cmd) {
4443 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
Ilya Dryomovadfa97c2013-10-10 20:39:28 +03004444 if (root->fs_info->sb->s_flags & MS_RDONLY) {
4445 ret = -EROFS;
4446 goto out;
4447 }
Stefan Behrens3f6bcfb2012-11-06 15:08:53 +01004448 if (atomic_xchg(
4449 &root->fs_info->mutually_exclusive_operation_running,
4450 1)) {
Anand Jaine57138b2013-08-21 11:44:48 +08004451 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
Stefan Behrens3f6bcfb2012-11-06 15:08:53 +01004452 } else {
Anand Jainb5255452016-03-24 18:48:14 +08004453 ret = btrfs_dev_replace_by_ioctl(root, p);
Stefan Behrens3f6bcfb2012-11-06 15:08:53 +01004454 atomic_set(
4455 &root->fs_info->mutually_exclusive_operation_running,
4456 0);
4457 }
4458 break;
4459 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4460 btrfs_dev_replace_status(root->fs_info, p);
4461 ret = 0;
4462 break;
4463 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4464 ret = btrfs_dev_replace_cancel(root->fs_info, p);
4465 break;
4466 default:
4467 ret = -EINVAL;
4468 break;
4469 }
4470
4471 if (copy_to_user(arg, p, sizeof(*p)))
4472 ret = -EFAULT;
Ilya Dryomovadfa97c2013-10-10 20:39:28 +03004473out:
Stefan Behrens3f6bcfb2012-11-06 15:08:53 +01004474 kfree(p);
4475 return ret;
4476}
4477
Jan Schmidtd7728c92011-07-07 16:48:38 +02004478static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4479{
4480 int ret = 0;
4481 int i;
Chris Mason740c3d22011-11-02 15:48:34 -04004482 u64 rel_ptr;
Jan Schmidtd7728c92011-07-07 16:48:38 +02004483 int size;
Chris Mason806468f2011-11-06 03:07:10 -05004484 struct btrfs_ioctl_ino_path_args *ipa = NULL;
Jan Schmidtd7728c92011-07-07 16:48:38 +02004485 struct inode_fs_paths *ipath = NULL;
4486 struct btrfs_path *path;
4487
Kusanagi Kouichi82b22ac2013-01-28 11:33:31 +00004488 if (!capable(CAP_DAC_READ_SEARCH))
Jan Schmidtd7728c92011-07-07 16:48:38 +02004489 return -EPERM;
4490
4491 path = btrfs_alloc_path();
4492 if (!path) {
4493 ret = -ENOMEM;
4494 goto out;
4495 }
4496
4497 ipa = memdup_user(arg, sizeof(*ipa));
4498 if (IS_ERR(ipa)) {
4499 ret = PTR_ERR(ipa);
4500 ipa = NULL;
4501 goto out;
4502 }
4503
4504 size = min_t(u32, ipa->size, 4096);
4505 ipath = init_ipath(size, root, path);
4506 if (IS_ERR(ipath)) {
4507 ret = PTR_ERR(ipath);
4508 ipath = NULL;
4509 goto out;
4510 }
4511
4512 ret = paths_from_inode(ipa->inum, ipath);
4513 if (ret < 0)
4514 goto out;
4515
4516 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
Jeff Mahoney745c4d82011-11-20 07:31:57 -05004517 rel_ptr = ipath->fspath->val[i] -
4518 (u64)(unsigned long)ipath->fspath->val;
Chris Mason740c3d22011-11-02 15:48:34 -04004519 ipath->fspath->val[i] = rel_ptr;
Jan Schmidtd7728c92011-07-07 16:48:38 +02004520 }
4521
Jeff Mahoney745c4d82011-11-20 07:31:57 -05004522 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
4523 (void *)(unsigned long)ipath->fspath, size);
Jan Schmidtd7728c92011-07-07 16:48:38 +02004524 if (ret) {
4525 ret = -EFAULT;
4526 goto out;
4527 }
4528
4529out:
4530 btrfs_free_path(path);
4531 free_ipath(ipath);
4532 kfree(ipa);
4533
4534 return ret;
4535}
4536
4537static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4538{
4539 struct btrfs_data_container *inodes = ctx;
4540 const size_t c = 3 * sizeof(u64);
4541
4542 if (inodes->bytes_left >= c) {
4543 inodes->bytes_left -= c;
4544 inodes->val[inodes->elem_cnt] = inum;
4545 inodes->val[inodes->elem_cnt + 1] = offset;
4546 inodes->val[inodes->elem_cnt + 2] = root;
4547 inodes->elem_cnt += 3;
4548 } else {
4549 inodes->bytes_missing += c - inodes->bytes_left;
4550 inodes->bytes_left = 0;
4551 inodes->elem_missed += 3;
4552 }
4553
4554 return 0;
4555}
4556
4557static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
4558 void __user *arg)
4559{
4560 int ret = 0;
4561 int size;
Jan Schmidtd7728c92011-07-07 16:48:38 +02004562 struct btrfs_ioctl_logical_ino_args *loi;
4563 struct btrfs_data_container *inodes = NULL;
4564 struct btrfs_path *path = NULL;
Jan Schmidtd7728c92011-07-07 16:48:38 +02004565
4566 if (!capable(CAP_SYS_ADMIN))
4567 return -EPERM;
4568
4569 loi = memdup_user(arg, sizeof(*loi));
4570 if (IS_ERR(loi)) {
4571 ret = PTR_ERR(loi);
4572 loi = NULL;
4573 goto out;
4574 }
4575
4576 path = btrfs_alloc_path();
4577 if (!path) {
4578 ret = -ENOMEM;
4579 goto out;
4580 }
4581
Byongho Leeee221842015-12-15 01:42:10 +09004582 size = min_t(u32, loi->size, SZ_64K);
Jan Schmidtd7728c92011-07-07 16:48:38 +02004583 inodes = init_data_container(size);
4584 if (IS_ERR(inodes)) {
4585 ret = PTR_ERR(inodes);
4586 inodes = NULL;
4587 goto out;
4588 }
4589
Liu Bodf031f02012-09-07 20:01:29 -06004590 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
4591 build_ino_list, inodes);
4592 if (ret == -EINVAL)
Jan Schmidtd7728c92011-07-07 16:48:38 +02004593 ret = -ENOENT;
4594 if (ret < 0)
4595 goto out;
4596
Jeff Mahoney745c4d82011-11-20 07:31:57 -05004597 ret = copy_to_user((void *)(unsigned long)loi->inodes,
4598 (void *)(unsigned long)inodes, size);
Jan Schmidtd7728c92011-07-07 16:48:38 +02004599 if (ret)
4600 ret = -EFAULT;
4601
4602out:
4603 btrfs_free_path(path);
Liu Bo425d17a2012-09-07 20:01:30 -06004604 vfree(inodes);
Jan Schmidtd7728c92011-07-07 16:48:38 +02004605 kfree(loi);
4606
4607 return ret;
4608}
4609
Ilya Dryomov19a39dc2012-01-16 22:04:49 +02004610void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004611 struct btrfs_ioctl_balance_args *bargs)
4612{
4613 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4614
4615 bargs->flags = bctl->flags;
4616
Ilya Dryomov837d5b62012-01-16 22:04:49 +02004617 if (atomic_read(&fs_info->balance_running))
4618 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4619 if (atomic_read(&fs_info->balance_pause_req))
4620 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
Ilya Dryomova7e99c62012-01-16 22:04:49 +02004621 if (atomic_read(&fs_info->balance_cancel_req))
4622 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
Ilya Dryomov837d5b62012-01-16 22:04:49 +02004623
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004624 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4625 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4626 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
Ilya Dryomov19a39dc2012-01-16 22:04:49 +02004627
4628 if (lock) {
4629 spin_lock(&fs_info->balance_lock);
4630 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4631 spin_unlock(&fs_info->balance_lock);
4632 } else {
4633 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4634 }
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004635}
4636
Liu Bo9ba1f6e2012-05-11 18:11:26 +08004637static long btrfs_ioctl_balance(struct file *file, void __user *arg)
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004638{
Al Viro496ad9a2013-01-23 17:07:38 -05004639 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004640 struct btrfs_fs_info *fs_info = root->fs_info;
4641 struct btrfs_ioctl_balance_args *bargs;
4642 struct btrfs_balance_control *bctl;
Ilya Dryomoved0fb782013-01-20 15:57:57 +02004643 bool need_unlock; /* for mut. excl. ops lock */
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004644 int ret;
4645
4646 if (!capable(CAP_SYS_ADMIN))
4647 return -EPERM;
4648
Liu Boe54bfa312012-06-29 03:58:48 -06004649 ret = mnt_want_write_file(file);
Liu Bo9ba1f6e2012-05-11 18:11:26 +08004650 if (ret)
4651 return ret;
4652
Ilya Dryomoved0fb782013-01-20 15:57:57 +02004653again:
4654 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
4655 mutex_lock(&fs_info->volume_mutex);
4656 mutex_lock(&fs_info->balance_mutex);
4657 need_unlock = true;
4658 goto locked;
4659 }
4660
4661 /*
Nicholas D Steeves01327612016-05-19 21:18:45 -04004662 * mut. excl. ops lock is locked. Three possibilities:
Ilya Dryomoved0fb782013-01-20 15:57:57 +02004663 * (1) some other op is running
4664 * (2) balance is running
4665 * (3) balance is paused -- special case (think resume)
4666 */
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004667 mutex_lock(&fs_info->balance_mutex);
Ilya Dryomoved0fb782013-01-20 15:57:57 +02004668 if (fs_info->balance_ctl) {
4669 /* this is either (2) or (3) */
4670 if (!atomic_read(&fs_info->balance_running)) {
4671 mutex_unlock(&fs_info->balance_mutex);
4672 if (!mutex_trylock(&fs_info->volume_mutex))
4673 goto again;
4674 mutex_lock(&fs_info->balance_mutex);
4675
4676 if (fs_info->balance_ctl &&
4677 !atomic_read(&fs_info->balance_running)) {
4678 /* this is (3) */
4679 need_unlock = false;
4680 goto locked;
4681 }
4682
4683 mutex_unlock(&fs_info->balance_mutex);
4684 mutex_unlock(&fs_info->volume_mutex);
4685 goto again;
4686 } else {
4687 /* this is (2) */
4688 mutex_unlock(&fs_info->balance_mutex);
4689 ret = -EINPROGRESS;
4690 goto out;
4691 }
4692 } else {
4693 /* this is (1) */
4694 mutex_unlock(&fs_info->balance_mutex);
Anand Jaine57138b2013-08-21 11:44:48 +08004695 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
Ilya Dryomoved0fb782013-01-20 15:57:57 +02004696 goto out;
4697 }
4698
4699locked:
4700 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004701
4702 if (arg) {
4703 bargs = memdup_user(arg, sizeof(*bargs));
4704 if (IS_ERR(bargs)) {
4705 ret = PTR_ERR(bargs);
Ilya Dryomoved0fb782013-01-20 15:57:57 +02004706 goto out_unlock;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004707 }
Ilya Dryomovde322262012-01-16 22:04:49 +02004708
4709 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4710 if (!fs_info->balance_ctl) {
4711 ret = -ENOTCONN;
4712 goto out_bargs;
4713 }
4714
4715 bctl = fs_info->balance_ctl;
4716 spin_lock(&fs_info->balance_lock);
4717 bctl->flags |= BTRFS_BALANCE_RESUME;
4718 spin_unlock(&fs_info->balance_lock);
4719
4720 goto do_balance;
4721 }
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004722 } else {
4723 bargs = NULL;
4724 }
4725
Ilya Dryomoved0fb782013-01-20 15:57:57 +02004726 if (fs_info->balance_ctl) {
Ilya Dryomov837d5b62012-01-16 22:04:49 +02004727 ret = -EINPROGRESS;
4728 goto out_bargs;
4729 }
4730
David Sterba8d2db782015-11-04 15:38:29 +01004731 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004732 if (!bctl) {
4733 ret = -ENOMEM;
4734 goto out_bargs;
4735 }
4736
4737 bctl->fs_info = fs_info;
4738 if (arg) {
4739 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4740 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4741 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4742
4743 bctl->flags = bargs->flags;
Ilya Dryomovf43ffb62012-01-16 22:04:47 +02004744 } else {
4745 /* balance everything - no filters */
4746 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004747 }
4748
David Sterba8eb93452015-10-12 16:55:54 +02004749 if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
4750 ret = -EINVAL;
Christian Engelmayer0f89abf2015-10-21 00:50:06 +02004751 goto out_bctl;
David Sterba8eb93452015-10-12 16:55:54 +02004752 }
4753
Ilya Dryomovde322262012-01-16 22:04:49 +02004754do_balance:
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004755 /*
Ilya Dryomoved0fb782013-01-20 15:57:57 +02004756 * Ownership of bctl and mutually_exclusive_operation_running
4757 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4758 * or, if restriper was paused all the way until unmount, in
4759 * free_fs_info. mutually_exclusive_operation_running is
4760 * cleared in __cancel_balance.
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004761 */
Ilya Dryomoved0fb782013-01-20 15:57:57 +02004762 need_unlock = false;
4763
4764 ret = btrfs_balance(bctl, bargs);
Christian Engelmayer0f89abf2015-10-21 00:50:06 +02004765 bctl = NULL;
Ilya Dryomoved0fb782013-01-20 15:57:57 +02004766
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004767 if (arg) {
4768 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4769 ret = -EFAULT;
4770 }
4771
Christian Engelmayer0f89abf2015-10-21 00:50:06 +02004772out_bctl:
4773 kfree(bctl);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004774out_bargs:
4775 kfree(bargs);
Ilya Dryomoved0fb782013-01-20 15:57:57 +02004776out_unlock:
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004777 mutex_unlock(&fs_info->balance_mutex);
4778 mutex_unlock(&fs_info->volume_mutex);
Ilya Dryomoved0fb782013-01-20 15:57:57 +02004779 if (need_unlock)
4780 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4781out:
Liu Boe54bfa312012-06-29 03:58:48 -06004782 mnt_drop_write_file(file);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02004783 return ret;
4784}
4785
Ilya Dryomov837d5b62012-01-16 22:04:49 +02004786static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4787{
4788 if (!capable(CAP_SYS_ADMIN))
4789 return -EPERM;
4790
4791 switch (cmd) {
4792 case BTRFS_BALANCE_CTL_PAUSE:
4793 return btrfs_pause_balance(root->fs_info);
Ilya Dryomova7e99c62012-01-16 22:04:49 +02004794 case BTRFS_BALANCE_CTL_CANCEL:
4795 return btrfs_cancel_balance(root->fs_info);
Ilya Dryomov837d5b62012-01-16 22:04:49 +02004796 }
4797
4798 return -EINVAL;
4799}
4800
Ilya Dryomov19a39dc2012-01-16 22:04:49 +02004801static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4802 void __user *arg)
4803{
4804 struct btrfs_fs_info *fs_info = root->fs_info;
4805 struct btrfs_ioctl_balance_args *bargs;
4806 int ret = 0;
4807
4808 if (!capable(CAP_SYS_ADMIN))
4809 return -EPERM;
4810
4811 mutex_lock(&fs_info->balance_mutex);
4812 if (!fs_info->balance_ctl) {
4813 ret = -ENOTCONN;
4814 goto out;
4815 }
4816
David Sterba8d2db782015-11-04 15:38:29 +01004817 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
Ilya Dryomov19a39dc2012-01-16 22:04:49 +02004818 if (!bargs) {
4819 ret = -ENOMEM;
4820 goto out;
4821 }
4822
4823 update_ioctl_balance_args(fs_info, 1, bargs);
4824
4825 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4826 ret = -EFAULT;
4827
4828 kfree(bargs);
4829out:
4830 mutex_unlock(&fs_info->balance_mutex);
4831 return ret;
4832}
4833
Miao Xie905b0dd2012-11-26 08:50:11 +00004834static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
Arne Jansen5d13a372011-09-14 15:53:51 +02004835{
Al Viro496ad9a2013-01-23 17:07:38 -05004836 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Arne Jansen5d13a372011-09-14 15:53:51 +02004837 struct btrfs_ioctl_quota_ctl_args *sa;
4838 struct btrfs_trans_handle *trans = NULL;
4839 int ret;
4840 int err;
4841
4842 if (!capable(CAP_SYS_ADMIN))
4843 return -EPERM;
4844
Miao Xie905b0dd2012-11-26 08:50:11 +00004845 ret = mnt_want_write_file(file);
4846 if (ret)
4847 return ret;
Arne Jansen5d13a372011-09-14 15:53:51 +02004848
4849 sa = memdup_user(arg, sizeof(*sa));
Miao Xie905b0dd2012-11-26 08:50:11 +00004850 if (IS_ERR(sa)) {
4851 ret = PTR_ERR(sa);
4852 goto drop_write;
4853 }
Arne Jansen5d13a372011-09-14 15:53:51 +02004854
Wang Shilong7708f022013-04-07 10:24:57 +00004855 down_write(&root->fs_info->subvol_sem);
Jan Schmidt2f232032013-04-25 16:04:51 +00004856 trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4857 if (IS_ERR(trans)) {
4858 ret = PTR_ERR(trans);
4859 goto out;
Arne Jansen5d13a372011-09-14 15:53:51 +02004860 }
4861
4862 switch (sa->cmd) {
4863 case BTRFS_QUOTA_CTL_ENABLE:
4864 ret = btrfs_quota_enable(trans, root->fs_info);
4865 break;
4866 case BTRFS_QUOTA_CTL_DISABLE:
4867 ret = btrfs_quota_disable(trans, root->fs_info);
4868 break;
Arne Jansen5d13a372011-09-14 15:53:51 +02004869 default:
4870 ret = -EINVAL;
4871 break;
4872 }
4873
Jan Schmidt2f232032013-04-25 16:04:51 +00004874 err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4875 if (err && !ret)
4876 ret = err;
Arne Jansen5d13a372011-09-14 15:53:51 +02004877out:
4878 kfree(sa);
Wang Shilong7708f022013-04-07 10:24:57 +00004879 up_write(&root->fs_info->subvol_sem);
Miao Xie905b0dd2012-11-26 08:50:11 +00004880drop_write:
4881 mnt_drop_write_file(file);
Arne Jansen5d13a372011-09-14 15:53:51 +02004882 return ret;
4883}
4884
Miao Xie905b0dd2012-11-26 08:50:11 +00004885static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
Arne Jansen5d13a372011-09-14 15:53:51 +02004886{
Al Viro496ad9a2013-01-23 17:07:38 -05004887 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Arne Jansen5d13a372011-09-14 15:53:51 +02004888 struct btrfs_ioctl_qgroup_assign_args *sa;
4889 struct btrfs_trans_handle *trans;
4890 int ret;
4891 int err;
4892
4893 if (!capable(CAP_SYS_ADMIN))
4894 return -EPERM;
4895
Miao Xie905b0dd2012-11-26 08:50:11 +00004896 ret = mnt_want_write_file(file);
4897 if (ret)
4898 return ret;
Arne Jansen5d13a372011-09-14 15:53:51 +02004899
4900 sa = memdup_user(arg, sizeof(*sa));
Miao Xie905b0dd2012-11-26 08:50:11 +00004901 if (IS_ERR(sa)) {
4902 ret = PTR_ERR(sa);
4903 goto drop_write;
4904 }
Arne Jansen5d13a372011-09-14 15:53:51 +02004905
4906 trans = btrfs_join_transaction(root);
4907 if (IS_ERR(trans)) {
4908 ret = PTR_ERR(trans);
4909 goto out;
4910 }
4911
4912 /* FIXME: check if the IDs really exist */
4913 if (sa->assign) {
4914 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4915 sa->src, sa->dst);
4916 } else {
4917 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4918 sa->src, sa->dst);
4919 }
4920
Qu Wenruoe082f562015-02-27 16:24:28 +08004921 /* update qgroup status and info */
4922 err = btrfs_run_qgroups(trans, root->fs_info);
4923 if (err < 0)
Anand Jainad8403d2016-03-10 12:22:15 +08004924 btrfs_handle_fs_error(root->fs_info, err,
4925 "failed to update qgroup status and info");
Arne Jansen5d13a372011-09-14 15:53:51 +02004926 err = btrfs_end_transaction(trans, root);
4927 if (err && !ret)
4928 ret = err;
4929
4930out:
4931 kfree(sa);
Miao Xie905b0dd2012-11-26 08:50:11 +00004932drop_write:
4933 mnt_drop_write_file(file);
Arne Jansen5d13a372011-09-14 15:53:51 +02004934 return ret;
4935}
4936
Miao Xie905b0dd2012-11-26 08:50:11 +00004937static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
Arne Jansen5d13a372011-09-14 15:53:51 +02004938{
Al Viro496ad9a2013-01-23 17:07:38 -05004939 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Arne Jansen5d13a372011-09-14 15:53:51 +02004940 struct btrfs_ioctl_qgroup_create_args *sa;
4941 struct btrfs_trans_handle *trans;
4942 int ret;
4943 int err;
4944
4945 if (!capable(CAP_SYS_ADMIN))
4946 return -EPERM;
4947
Miao Xie905b0dd2012-11-26 08:50:11 +00004948 ret = mnt_want_write_file(file);
4949 if (ret)
4950 return ret;
Arne Jansen5d13a372011-09-14 15:53:51 +02004951
4952 sa = memdup_user(arg, sizeof(*sa));
Miao Xie905b0dd2012-11-26 08:50:11 +00004953 if (IS_ERR(sa)) {
4954 ret = PTR_ERR(sa);
4955 goto drop_write;
4956 }
Arne Jansen5d13a372011-09-14 15:53:51 +02004957
Miao Xied86e56c2012-11-15 11:35:41 +00004958 if (!sa->qgroupid) {
4959 ret = -EINVAL;
4960 goto out;
4961 }
4962
Arne Jansen5d13a372011-09-14 15:53:51 +02004963 trans = btrfs_join_transaction(root);
4964 if (IS_ERR(trans)) {
4965 ret = PTR_ERR(trans);
4966 goto out;
4967 }
4968
4969 /* FIXME: check if the IDs really exist */
4970 if (sa->create) {
Dongsheng Yang4087cf22015-01-18 10:59:23 -05004971 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid);
Arne Jansen5d13a372011-09-14 15:53:51 +02004972 } else {
4973 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4974 }
4975
4976 err = btrfs_end_transaction(trans, root);
4977 if (err && !ret)
4978 ret = err;
4979
4980out:
4981 kfree(sa);
Miao Xie905b0dd2012-11-26 08:50:11 +00004982drop_write:
4983 mnt_drop_write_file(file);
Arne Jansen5d13a372011-09-14 15:53:51 +02004984 return ret;
4985}
4986
Miao Xie905b0dd2012-11-26 08:50:11 +00004987static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
Arne Jansen5d13a372011-09-14 15:53:51 +02004988{
Al Viro496ad9a2013-01-23 17:07:38 -05004989 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Arne Jansen5d13a372011-09-14 15:53:51 +02004990 struct btrfs_ioctl_qgroup_limit_args *sa;
4991 struct btrfs_trans_handle *trans;
4992 int ret;
4993 int err;
4994 u64 qgroupid;
4995
4996 if (!capable(CAP_SYS_ADMIN))
4997 return -EPERM;
4998
Miao Xie905b0dd2012-11-26 08:50:11 +00004999 ret = mnt_want_write_file(file);
5000 if (ret)
5001 return ret;
Arne Jansen5d13a372011-09-14 15:53:51 +02005002
5003 sa = memdup_user(arg, sizeof(*sa));
Miao Xie905b0dd2012-11-26 08:50:11 +00005004 if (IS_ERR(sa)) {
5005 ret = PTR_ERR(sa);
5006 goto drop_write;
5007 }
Arne Jansen5d13a372011-09-14 15:53:51 +02005008
5009 trans = btrfs_join_transaction(root);
5010 if (IS_ERR(trans)) {
5011 ret = PTR_ERR(trans);
5012 goto out;
5013 }
5014
5015 qgroupid = sa->qgroupid;
5016 if (!qgroupid) {
5017 /* take the current subvol as qgroup */
5018 qgroupid = root->root_key.objectid;
5019 }
5020
5021 /* FIXME: check if the IDs really exist */
5022 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
5023
5024 err = btrfs_end_transaction(trans, root);
5025 if (err && !ret)
5026 ret = err;
5027
5028out:
5029 kfree(sa);
Miao Xie905b0dd2012-11-26 08:50:11 +00005030drop_write:
5031 mnt_drop_write_file(file);
Arne Jansen5d13a372011-09-14 15:53:51 +02005032 return ret;
5033}
5034
Jan Schmidt2f232032013-04-25 16:04:51 +00005035static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
5036{
Al Viro6d0379e2013-06-16 19:32:35 +04005037 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Jan Schmidt2f232032013-04-25 16:04:51 +00005038 struct btrfs_ioctl_quota_rescan_args *qsa;
5039 int ret;
5040
5041 if (!capable(CAP_SYS_ADMIN))
5042 return -EPERM;
5043
5044 ret = mnt_want_write_file(file);
5045 if (ret)
5046 return ret;
5047
5048 qsa = memdup_user(arg, sizeof(*qsa));
5049 if (IS_ERR(qsa)) {
5050 ret = PTR_ERR(qsa);
5051 goto drop_write;
5052 }
5053
5054 if (qsa->flags) {
5055 ret = -EINVAL;
5056 goto out;
5057 }
5058
5059 ret = btrfs_qgroup_rescan(root->fs_info);
5060
5061out:
5062 kfree(qsa);
5063drop_write:
5064 mnt_drop_write_file(file);
5065 return ret;
5066}
5067
5068static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
5069{
Al Viro6d0379e2013-06-16 19:32:35 +04005070 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Jan Schmidt2f232032013-04-25 16:04:51 +00005071 struct btrfs_ioctl_quota_rescan_args *qsa;
5072 int ret = 0;
5073
5074 if (!capable(CAP_SYS_ADMIN))
5075 return -EPERM;
5076
David Sterba8d2db782015-11-04 15:38:29 +01005077 qsa = kzalloc(sizeof(*qsa), GFP_KERNEL);
Jan Schmidt2f232032013-04-25 16:04:51 +00005078 if (!qsa)
5079 return -ENOMEM;
5080
5081 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
5082 qsa->flags = 1;
5083 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
5084 }
5085
5086 if (copy_to_user(arg, qsa, sizeof(*qsa)))
5087 ret = -EFAULT;
5088
5089 kfree(qsa);
5090 return ret;
5091}
5092
Jan Schmidt57254b6e2013-05-06 19:14:17 +00005093static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
5094{
Al Viro54563d42013-09-01 15:57:51 -04005095 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Jan Schmidt57254b6e2013-05-06 19:14:17 +00005096
5097 if (!capable(CAP_SYS_ADMIN))
5098 return -EPERM;
5099
Jeff Mahoneyd06f23d2016-08-08 22:08:06 -04005100 return btrfs_qgroup_wait_for_completion(root->fs_info, true);
Jan Schmidt57254b6e2013-05-06 19:14:17 +00005101}
5102
Hugo Millsabccd002014-01-30 20:17:00 +00005103static long _btrfs_ioctl_set_received_subvol(struct file *file,
5104 struct btrfs_ioctl_received_subvol_args *sa)
Alexander Block8ea05e32012-07-25 17:35:53 +02005105{
Al Viro496ad9a2013-01-23 17:07:38 -05005106 struct inode *inode = file_inode(file);
Alexander Block8ea05e32012-07-25 17:35:53 +02005107 struct btrfs_root *root = BTRFS_I(inode)->root;
5108 struct btrfs_root_item *root_item = &root->root_item;
5109 struct btrfs_trans_handle *trans;
Deepa Dinamani04b285f2016-02-06 23:57:21 -08005110 struct timespec ct = current_fs_time(inode->i_sb);
Alexander Block8ea05e32012-07-25 17:35:53 +02005111 int ret = 0;
Stefan Behrensdd5f9612013-08-15 17:11:20 +02005112 int received_uuid_changed;
Alexander Block8ea05e32012-07-25 17:35:53 +02005113
David Sterbabd60ea02014-01-16 15:50:22 +01005114 if (!inode_owner_or_capable(inode))
5115 return -EPERM;
5116
Alexander Block8ea05e32012-07-25 17:35:53 +02005117 ret = mnt_want_write_file(file);
5118 if (ret < 0)
5119 return ret;
5120
5121 down_write(&root->fs_info->subvol_sem);
5122
5123 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
5124 ret = -EINVAL;
5125 goto out;
5126 }
5127
5128 if (btrfs_root_readonly(root)) {
5129 ret = -EROFS;
5130 goto out;
5131 }
5132
Stefan Behrensdd5f9612013-08-15 17:11:20 +02005133 /*
5134 * 1 - root item
5135 * 2 - uuid items (received uuid + subvol uuid)
5136 */
5137 trans = btrfs_start_transaction(root, 3);
Alexander Block8ea05e32012-07-25 17:35:53 +02005138 if (IS_ERR(trans)) {
5139 ret = PTR_ERR(trans);
5140 trans = NULL;
5141 goto out;
5142 }
5143
5144 sa->rtransid = trans->transid;
5145 sa->rtime.sec = ct.tv_sec;
5146 sa->rtime.nsec = ct.tv_nsec;
5147
Stefan Behrensdd5f9612013-08-15 17:11:20 +02005148 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
5149 BTRFS_UUID_SIZE);
5150 if (received_uuid_changed &&
5151 !btrfs_is_empty_uuid(root_item->received_uuid))
5152 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
5153 root_item->received_uuid,
5154 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5155 root->root_key.objectid);
Alexander Block8ea05e32012-07-25 17:35:53 +02005156 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
5157 btrfs_set_root_stransid(root_item, sa->stransid);
5158 btrfs_set_root_rtransid(root_item, sa->rtransid);
Qu Wenruo3cae2102013-07-16 11:19:18 +08005159 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
5160 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
5161 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
5162 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
Alexander Block8ea05e32012-07-25 17:35:53 +02005163
5164 ret = btrfs_update_root(trans, root->fs_info->tree_root,
5165 &root->root_key, &root->root_item);
5166 if (ret < 0) {
5167 btrfs_end_transaction(trans, root);
Alexander Block8ea05e32012-07-25 17:35:53 +02005168 goto out;
Stefan Behrensdd5f9612013-08-15 17:11:20 +02005169 }
5170 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
5171 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
5172 sa->uuid,
5173 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5174 root->root_key.objectid);
5175 if (ret < 0 && ret != -EEXIST) {
Jeff Mahoney66642832016-06-10 18:19:25 -04005176 btrfs_abort_transaction(trans, ret);
Alexander Block8ea05e32012-07-25 17:35:53 +02005177 goto out;
Stefan Behrensdd5f9612013-08-15 17:11:20 +02005178 }
5179 }
5180 ret = btrfs_commit_transaction(trans, root);
5181 if (ret < 0) {
Jeff Mahoney66642832016-06-10 18:19:25 -04005182 btrfs_abort_transaction(trans, ret);
Stefan Behrensdd5f9612013-08-15 17:11:20 +02005183 goto out;
Alexander Block8ea05e32012-07-25 17:35:53 +02005184 }
5185
Hugo Millsabccd002014-01-30 20:17:00 +00005186out:
5187 up_write(&root->fs_info->subvol_sem);
5188 mnt_drop_write_file(file);
5189 return ret;
5190}
5191
5192#ifdef CONFIG_64BIT
5193static long btrfs_ioctl_set_received_subvol_32(struct file *file,
5194 void __user *arg)
5195{
5196 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
5197 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
5198 int ret = 0;
5199
5200 args32 = memdup_user(arg, sizeof(*args32));
5201 if (IS_ERR(args32)) {
5202 ret = PTR_ERR(args32);
5203 args32 = NULL;
5204 goto out;
5205 }
5206
David Sterba8d2db782015-11-04 15:38:29 +01005207 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
Dan Carpenter84dbeb82014-03-28 11:06:00 +03005208 if (!args64) {
5209 ret = -ENOMEM;
Hugo Millsabccd002014-01-30 20:17:00 +00005210 goto out;
5211 }
5212
5213 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
5214 args64->stransid = args32->stransid;
5215 args64->rtransid = args32->rtransid;
5216 args64->stime.sec = args32->stime.sec;
5217 args64->stime.nsec = args32->stime.nsec;
5218 args64->rtime.sec = args32->rtime.sec;
5219 args64->rtime.nsec = args32->rtime.nsec;
5220 args64->flags = args32->flags;
5221
5222 ret = _btrfs_ioctl_set_received_subvol(file, args64);
5223 if (ret)
5224 goto out;
5225
5226 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
5227 args32->stransid = args64->stransid;
5228 args32->rtransid = args64->rtransid;
5229 args32->stime.sec = args64->stime.sec;
5230 args32->stime.nsec = args64->stime.nsec;
5231 args32->rtime.sec = args64->rtime.sec;
5232 args32->rtime.nsec = args64->rtime.nsec;
5233 args32->flags = args64->flags;
5234
5235 ret = copy_to_user(arg, args32, sizeof(*args32));
5236 if (ret)
5237 ret = -EFAULT;
5238
5239out:
5240 kfree(args32);
5241 kfree(args64);
5242 return ret;
5243}
5244#endif
5245
5246static long btrfs_ioctl_set_received_subvol(struct file *file,
5247 void __user *arg)
5248{
5249 struct btrfs_ioctl_received_subvol_args *sa = NULL;
5250 int ret = 0;
5251
5252 sa = memdup_user(arg, sizeof(*sa));
5253 if (IS_ERR(sa)) {
5254 ret = PTR_ERR(sa);
5255 sa = NULL;
5256 goto out;
5257 }
5258
5259 ret = _btrfs_ioctl_set_received_subvol(file, sa);
5260
5261 if (ret)
5262 goto out;
5263
Alexander Block8ea05e32012-07-25 17:35:53 +02005264 ret = copy_to_user(arg, sa, sizeof(*sa));
5265 if (ret)
5266 ret = -EFAULT;
5267
5268out:
5269 kfree(sa);
Alexander Block8ea05e32012-07-25 17:35:53 +02005270 return ret;
5271}
5272
jeff.liu867ab662013-01-05 02:48:01 +00005273static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
5274{
Al Viro6d0379e2013-06-16 19:32:35 +04005275 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Anand Jaina1b83ac2013-07-19 17:39:32 +08005276 size_t len;
jeff.liu867ab662013-01-05 02:48:01 +00005277 int ret;
Anand Jaina1b83ac2013-07-19 17:39:32 +08005278 char label[BTRFS_LABEL_SIZE];
5279
5280 spin_lock(&root->fs_info->super_lock);
5281 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5282 spin_unlock(&root->fs_info->super_lock);
5283
5284 len = strnlen(label, BTRFS_LABEL_SIZE);
jeff.liu867ab662013-01-05 02:48:01 +00005285
5286 if (len == BTRFS_LABEL_SIZE) {
Frank Holtonefe120a2013-12-20 11:37:06 -05005287 btrfs_warn(root->fs_info,
5288 "label is too long, return the first %zu bytes", --len);
jeff.liu867ab662013-01-05 02:48:01 +00005289 }
5290
jeff.liu867ab662013-01-05 02:48:01 +00005291 ret = copy_to_user(arg, label, len);
jeff.liu867ab662013-01-05 02:48:01 +00005292
5293 return ret ? -EFAULT : 0;
5294}
5295
jeff.liua8bfd4a2013-01-05 02:48:08 +00005296static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5297{
Al Viro6d0379e2013-06-16 19:32:35 +04005298 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
jeff.liua8bfd4a2013-01-05 02:48:08 +00005299 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5300 struct btrfs_trans_handle *trans;
5301 char label[BTRFS_LABEL_SIZE];
5302 int ret;
5303
5304 if (!capable(CAP_SYS_ADMIN))
5305 return -EPERM;
5306
5307 if (copy_from_user(label, arg, sizeof(label)))
5308 return -EFAULT;
5309
5310 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
Jeff Mahoney5d163e02016-09-20 10:05:00 -04005311 btrfs_err(root->fs_info,
5312 "unable to set label with more than %d bytes",
5313 BTRFS_LABEL_SIZE - 1);
jeff.liua8bfd4a2013-01-05 02:48:08 +00005314 return -EINVAL;
5315 }
5316
5317 ret = mnt_want_write_file(file);
5318 if (ret)
5319 return ret;
5320
jeff.liua8bfd4a2013-01-05 02:48:08 +00005321 trans = btrfs_start_transaction(root, 0);
5322 if (IS_ERR(trans)) {
5323 ret = PTR_ERR(trans);
5324 goto out_unlock;
5325 }
5326
Anand Jaina1b83ac2013-07-19 17:39:32 +08005327 spin_lock(&root->fs_info->super_lock);
jeff.liua8bfd4a2013-01-05 02:48:08 +00005328 strcpy(super_block->label, label);
Anand Jaina1b83ac2013-07-19 17:39:32 +08005329 spin_unlock(&root->fs_info->super_lock);
Jeff Mahoneyd0270ac2014-02-07 14:33:57 +01005330 ret = btrfs_commit_transaction(trans, root);
jeff.liua8bfd4a2013-01-05 02:48:08 +00005331
5332out_unlock:
jeff.liua8bfd4a2013-01-05 02:48:08 +00005333 mnt_drop_write_file(file);
5334 return ret;
5335}
5336
Jeff Mahoney2eaa0552013-11-15 15:33:55 -05005337#define INIT_FEATURE_FLAGS(suffix) \
5338 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5339 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5340 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5341
David Sterbad5131b62016-02-17 15:26:27 +01005342int btrfs_ioctl_get_supported_features(void __user *arg)
Jeff Mahoney2eaa0552013-11-15 15:33:55 -05005343{
David Sterba4d4ab6d2015-11-19 11:42:31 +01005344 static const struct btrfs_ioctl_feature_flags features[3] = {
Jeff Mahoney2eaa0552013-11-15 15:33:55 -05005345 INIT_FEATURE_FLAGS(SUPP),
5346 INIT_FEATURE_FLAGS(SAFE_SET),
5347 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5348 };
5349
5350 if (copy_to_user(arg, &features, sizeof(features)))
5351 return -EFAULT;
5352
5353 return 0;
5354}
5355
5356static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5357{
5358 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5359 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5360 struct btrfs_ioctl_feature_flags features;
5361
5362 features.compat_flags = btrfs_super_compat_flags(super_block);
5363 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5364 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5365
5366 if (copy_to_user(arg, &features, sizeof(features)))
5367 return -EFAULT;
5368
5369 return 0;
5370}
5371
Jeff Mahoney3b02a682013-11-01 13:07:02 -04005372static int check_feature_bits(struct btrfs_root *root,
5373 enum btrfs_feature_set set,
Jeff Mahoney2eaa0552013-11-15 15:33:55 -05005374 u64 change_mask, u64 flags, u64 supported_flags,
5375 u64 safe_set, u64 safe_clear)
5376{
Jeff Mahoney3b02a682013-11-01 13:07:02 -04005377 const char *type = btrfs_feature_set_names[set];
5378 char *names;
Jeff Mahoney2eaa0552013-11-15 15:33:55 -05005379 u64 disallowed, unsupported;
5380 u64 set_mask = flags & change_mask;
5381 u64 clear_mask = ~flags & change_mask;
5382
5383 unsupported = set_mask & ~supported_flags;
5384 if (unsupported) {
Jeff Mahoney3b02a682013-11-01 13:07:02 -04005385 names = btrfs_printable_features(set, unsupported);
5386 if (names) {
5387 btrfs_warn(root->fs_info,
5388 "this kernel does not support the %s feature bit%s",
5389 names, strchr(names, ',') ? "s" : "");
5390 kfree(names);
5391 } else
5392 btrfs_warn(root->fs_info,
Jeff Mahoney2eaa0552013-11-15 15:33:55 -05005393 "this kernel does not support %s bits 0x%llx",
5394 type, unsupported);
5395 return -EOPNOTSUPP;
5396 }
5397
5398 disallowed = set_mask & ~safe_set;
5399 if (disallowed) {
Jeff Mahoney3b02a682013-11-01 13:07:02 -04005400 names = btrfs_printable_features(set, disallowed);
5401 if (names) {
5402 btrfs_warn(root->fs_info,
5403 "can't set the %s feature bit%s while mounted",
5404 names, strchr(names, ',') ? "s" : "");
5405 kfree(names);
5406 } else
5407 btrfs_warn(root->fs_info,
Jeff Mahoney2eaa0552013-11-15 15:33:55 -05005408 "can't set %s bits 0x%llx while mounted",
5409 type, disallowed);
5410 return -EPERM;
5411 }
5412
5413 disallowed = clear_mask & ~safe_clear;
5414 if (disallowed) {
Jeff Mahoney3b02a682013-11-01 13:07:02 -04005415 names = btrfs_printable_features(set, disallowed);
5416 if (names) {
5417 btrfs_warn(root->fs_info,
5418 "can't clear the %s feature bit%s while mounted",
5419 names, strchr(names, ',') ? "s" : "");
5420 kfree(names);
5421 } else
5422 btrfs_warn(root->fs_info,
Jeff Mahoney2eaa0552013-11-15 15:33:55 -05005423 "can't clear %s bits 0x%llx while mounted",
5424 type, disallowed);
5425 return -EPERM;
5426 }
5427
5428 return 0;
5429}
5430
5431#define check_feature(root, change_mask, flags, mask_base) \
Jeff Mahoney3b02a682013-11-01 13:07:02 -04005432check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
Jeff Mahoney2eaa0552013-11-15 15:33:55 -05005433 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5434 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5435 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5436
5437static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5438{
5439 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5440 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5441 struct btrfs_ioctl_feature_flags flags[2];
5442 struct btrfs_trans_handle *trans;
5443 u64 newflags;
5444 int ret;
5445
5446 if (!capable(CAP_SYS_ADMIN))
5447 return -EPERM;
5448
5449 if (copy_from_user(flags, arg, sizeof(flags)))
5450 return -EFAULT;
5451
5452 /* Nothing to do */
5453 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5454 !flags[0].incompat_flags)
5455 return 0;
5456
5457 ret = check_feature(root, flags[0].compat_flags,
5458 flags[1].compat_flags, COMPAT);
5459 if (ret)
5460 return ret;
5461
5462 ret = check_feature(root, flags[0].compat_ro_flags,
5463 flags[1].compat_ro_flags, COMPAT_RO);
5464 if (ret)
5465 return ret;
5466
5467 ret = check_feature(root, flags[0].incompat_flags,
5468 flags[1].incompat_flags, INCOMPAT);
5469 if (ret)
5470 return ret;
5471
David Sterba7ab19622016-05-04 11:32:00 +02005472 ret = mnt_want_write_file(file);
5473 if (ret)
5474 return ret;
5475
David Sterba8051aa12014-02-07 14:34:04 +01005476 trans = btrfs_start_transaction(root, 0);
David Sterba7ab19622016-05-04 11:32:00 +02005477 if (IS_ERR(trans)) {
5478 ret = PTR_ERR(trans);
5479 goto out_drop_write;
5480 }
Jeff Mahoney2eaa0552013-11-15 15:33:55 -05005481
5482 spin_lock(&root->fs_info->super_lock);
5483 newflags = btrfs_super_compat_flags(super_block);
5484 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5485 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5486 btrfs_set_super_compat_flags(super_block, newflags);
5487
5488 newflags = btrfs_super_compat_ro_flags(super_block);
5489 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5490 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5491 btrfs_set_super_compat_ro_flags(super_block, newflags);
5492
5493 newflags = btrfs_super_incompat_flags(super_block);
5494 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5495 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5496 btrfs_set_super_incompat_flags(super_block, newflags);
5497 spin_unlock(&root->fs_info->super_lock);
5498
David Sterba7ab19622016-05-04 11:32:00 +02005499 ret = btrfs_commit_transaction(trans, root);
5500out_drop_write:
5501 mnt_drop_write_file(file);
5502
5503 return ret;
Jeff Mahoney2eaa0552013-11-15 15:33:55 -05005504}
5505
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04005506long btrfs_ioctl(struct file *file, unsigned int
5507 cmd, unsigned long arg)
5508{
Al Viro496ad9a2013-01-23 17:07:38 -05005509 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
Christoph Hellwig4bcabaa2008-12-02 06:36:08 -05005510 void __user *argp = (void __user *)arg;
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04005511
5512 switch (cmd) {
Christoph Hellwig6cbff002009-04-17 10:37:41 +02005513 case FS_IOC_GETFLAGS:
5514 return btrfs_ioctl_getflags(file, argp);
5515 case FS_IOC_SETFLAGS:
5516 return btrfs_ioctl_setflags(file, argp);
5517 case FS_IOC_GETVERSION:
5518 return btrfs_ioctl_getversion(file, argp);
Li Dongyangf7039b12011-03-24 10:24:28 +00005519 case FITRIM:
5520 return btrfs_ioctl_fitrim(file, argp);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04005521 case BTRFS_IOC_SNAP_CREATE:
Li Zefanfa0d2b92010-12-20 15:53:28 +08005522 return btrfs_ioctl_snap_create(file, argp, 0);
Li Zefanfdfb1e42010-12-10 06:41:56 +00005523 case BTRFS_IOC_SNAP_CREATE_V2:
Li Zefanfa0d2b92010-12-20 15:53:28 +08005524 return btrfs_ioctl_snap_create_v2(file, argp, 0);
Chris Mason3de45862008-11-17 21:02:50 -05005525 case BTRFS_IOC_SUBVOL_CREATE:
Li Zefanfa0d2b92010-12-20 15:53:28 +08005526 return btrfs_ioctl_snap_create(file, argp, 1);
Arne Jansen6f72c7e2011-09-14 15:58:21 +02005527 case BTRFS_IOC_SUBVOL_CREATE_V2:
5528 return btrfs_ioctl_snap_create_v2(file, argp, 1);
Yan, Zheng76dda932009-09-21 16:00:26 -04005529 case BTRFS_IOC_SNAP_DESTROY:
5530 return btrfs_ioctl_snap_destroy(file, argp);
Li Zefan0caa1022010-12-20 16:30:25 +08005531 case BTRFS_IOC_SUBVOL_GETFLAGS:
5532 return btrfs_ioctl_subvol_getflags(file, argp);
5533 case BTRFS_IOC_SUBVOL_SETFLAGS:
5534 return btrfs_ioctl_subvol_setflags(file, argp);
Josef Bacik6ef5ed02009-12-11 21:11:29 +00005535 case BTRFS_IOC_DEFAULT_SUBVOL:
5536 return btrfs_ioctl_default_subvol(file, argp);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04005537 case BTRFS_IOC_DEFRAG:
Chris Mason1e701a32010-03-11 09:42:04 -05005538 return btrfs_ioctl_defrag(file, NULL);
5539 case BTRFS_IOC_DEFRAG_RANGE:
5540 return btrfs_ioctl_defrag(file, argp);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04005541 case BTRFS_IOC_RESIZE:
Miao Xie198605a2012-11-26 08:43:45 +00005542 return btrfs_ioctl_resize(file, argp);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04005543 case BTRFS_IOC_ADD_DEV:
Christoph Hellwig4bcabaa2008-12-02 06:36:08 -05005544 return btrfs_ioctl_add_dev(root, argp);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04005545 case BTRFS_IOC_RM_DEV:
Miao Xieda249272012-11-26 08:44:50 +00005546 return btrfs_ioctl_rm_dev(file, argp);
Anand Jain6b526ed2016-02-13 10:01:39 +08005547 case BTRFS_IOC_RM_DEV_V2:
5548 return btrfs_ioctl_rm_dev_v2(file, argp);
Jan Schmidt475f6382011-03-11 15:41:01 +01005549 case BTRFS_IOC_FS_INFO:
5550 return btrfs_ioctl_fs_info(root, argp);
5551 case BTRFS_IOC_DEV_INFO:
5552 return btrfs_ioctl_dev_info(root, argp);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04005553 case BTRFS_IOC_BALANCE:
Liu Bo9ba1f6e2012-05-11 18:11:26 +08005554 return btrfs_ioctl_balance(file, NULL);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04005555 case BTRFS_IOC_TRANS_START:
5556 return btrfs_ioctl_trans_start(file);
5557 case BTRFS_IOC_TRANS_END:
5558 return btrfs_ioctl_trans_end(file);
Chris Masonac8e9812010-02-28 15:39:26 -05005559 case BTRFS_IOC_TREE_SEARCH:
5560 return btrfs_ioctl_tree_search(file, argp);
Gerhard Heiftcc68a8a2014-01-30 16:24:03 +01005561 case BTRFS_IOC_TREE_SEARCH_V2:
5562 return btrfs_ioctl_tree_search_v2(file, argp);
Chris Masonac8e9812010-02-28 15:39:26 -05005563 case BTRFS_IOC_INO_LOOKUP:
5564 return btrfs_ioctl_ino_lookup(file, argp);
Jan Schmidtd7728c92011-07-07 16:48:38 +02005565 case BTRFS_IOC_INO_PATHS:
5566 return btrfs_ioctl_ino_to_path(root, argp);
5567 case BTRFS_IOC_LOGICAL_INO:
5568 return btrfs_ioctl_logical_to_ino(root, argp);
Josef Bacik1406e432010-01-13 18:19:06 +00005569 case BTRFS_IOC_SPACE_INFO:
5570 return btrfs_ioctl_space_info(root, argp);
Filipe David Borba Manana9b199852013-09-23 11:35:11 +01005571 case BTRFS_IOC_SYNC: {
5572 int ret;
5573
Miao Xie6c255e62014-03-06 13:55:01 +08005574 ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1);
Filipe David Borba Manana9b199852013-09-23 11:35:11 +01005575 if (ret)
5576 return ret;
Al Viroddb52f42014-10-21 20:21:18 -04005577 ret = btrfs_sync_fs(file_inode(file)->i_sb, 1);
David Sterba2fad4e82014-07-23 14:39:35 +02005578 /*
5579 * The transaction thread may want to do more work,
Nicholas D Steeves01327612016-05-19 21:18:45 -04005580 * namely it pokes the cleaner kthread that will start
David Sterba2fad4e82014-07-23 14:39:35 +02005581 * processing uncleaned subvols.
5582 */
5583 wake_up_process(root->fs_info->transaction_kthread);
Filipe David Borba Manana9b199852013-09-23 11:35:11 +01005584 return ret;
5585 }
Sage Weil46204592010-10-29 15:41:32 -04005586 case BTRFS_IOC_START_SYNC:
Miao Xie9a8c28b2012-11-26 08:40:43 +00005587 return btrfs_ioctl_start_sync(root, argp);
Sage Weil46204592010-10-29 15:41:32 -04005588 case BTRFS_IOC_WAIT_SYNC:
Miao Xie9a8c28b2012-11-26 08:40:43 +00005589 return btrfs_ioctl_wait_sync(root, argp);
Jan Schmidt475f6382011-03-11 15:41:01 +01005590 case BTRFS_IOC_SCRUB:
Miao Xieb8e95482012-11-26 08:48:01 +00005591 return btrfs_ioctl_scrub(file, argp);
Jan Schmidt475f6382011-03-11 15:41:01 +01005592 case BTRFS_IOC_SCRUB_CANCEL:
5593 return btrfs_ioctl_scrub_cancel(root, argp);
5594 case BTRFS_IOC_SCRUB_PROGRESS:
5595 return btrfs_ioctl_scrub_progress(root, argp);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02005596 case BTRFS_IOC_BALANCE_V2:
Liu Bo9ba1f6e2012-05-11 18:11:26 +08005597 return btrfs_ioctl_balance(file, argp);
Ilya Dryomov837d5b62012-01-16 22:04:49 +02005598 case BTRFS_IOC_BALANCE_CTL:
5599 return btrfs_ioctl_balance_ctl(root, arg);
Ilya Dryomov19a39dc2012-01-16 22:04:49 +02005600 case BTRFS_IOC_BALANCE_PROGRESS:
5601 return btrfs_ioctl_balance_progress(root, argp);
Alexander Block8ea05e32012-07-25 17:35:53 +02005602 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5603 return btrfs_ioctl_set_received_subvol(file, argp);
Hugo Millsabccd002014-01-30 20:17:00 +00005604#ifdef CONFIG_64BIT
5605 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5606 return btrfs_ioctl_set_received_subvol_32(file, argp);
5607#endif
Alexander Block31db9f72012-07-25 23:19:24 +02005608 case BTRFS_IOC_SEND:
5609 return btrfs_ioctl_send(file, argp);
Stefan Behrensc11d2c22012-05-25 16:06:09 +02005610 case BTRFS_IOC_GET_DEV_STATS:
David Sterbab27f7c02012-06-22 06:30:39 -06005611 return btrfs_ioctl_get_dev_stats(root, argp);
Arne Jansen5d13a372011-09-14 15:53:51 +02005612 case BTRFS_IOC_QUOTA_CTL:
Miao Xie905b0dd2012-11-26 08:50:11 +00005613 return btrfs_ioctl_quota_ctl(file, argp);
Arne Jansen5d13a372011-09-14 15:53:51 +02005614 case BTRFS_IOC_QGROUP_ASSIGN:
Miao Xie905b0dd2012-11-26 08:50:11 +00005615 return btrfs_ioctl_qgroup_assign(file, argp);
Arne Jansen5d13a372011-09-14 15:53:51 +02005616 case BTRFS_IOC_QGROUP_CREATE:
Miao Xie905b0dd2012-11-26 08:50:11 +00005617 return btrfs_ioctl_qgroup_create(file, argp);
Arne Jansen5d13a372011-09-14 15:53:51 +02005618 case BTRFS_IOC_QGROUP_LIMIT:
Miao Xie905b0dd2012-11-26 08:50:11 +00005619 return btrfs_ioctl_qgroup_limit(file, argp);
Jan Schmidt2f232032013-04-25 16:04:51 +00005620 case BTRFS_IOC_QUOTA_RESCAN:
5621 return btrfs_ioctl_quota_rescan(file, argp);
5622 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5623 return btrfs_ioctl_quota_rescan_status(file, argp);
Jan Schmidt57254b6e2013-05-06 19:14:17 +00005624 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5625 return btrfs_ioctl_quota_rescan_wait(file, argp);
Stefan Behrens3f6bcfb2012-11-06 15:08:53 +01005626 case BTRFS_IOC_DEV_REPLACE:
5627 return btrfs_ioctl_dev_replace(root, argp);
jeff.liu867ab662013-01-05 02:48:01 +00005628 case BTRFS_IOC_GET_FSLABEL:
5629 return btrfs_ioctl_get_fslabel(file, argp);
jeff.liua8bfd4a2013-01-05 02:48:08 +00005630 case BTRFS_IOC_SET_FSLABEL:
5631 return btrfs_ioctl_set_fslabel(file, argp);
Jeff Mahoney2eaa0552013-11-15 15:33:55 -05005632 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
David Sterbad5131b62016-02-17 15:26:27 +01005633 return btrfs_ioctl_get_supported_features(argp);
Jeff Mahoney2eaa0552013-11-15 15:33:55 -05005634 case BTRFS_IOC_GET_FEATURES:
5635 return btrfs_ioctl_get_features(file, argp);
5636 case BTRFS_IOC_SET_FEATURES:
5637 return btrfs_ioctl_set_features(file, argp);
Christoph Hellwigf46b5a62008-06-11 21:53:53 -04005638 }
5639
5640 return -ENOTTY;
5641}
Luke Dashjr4c63c242015-10-29 08:22:21 +00005642
5643#ifdef CONFIG_COMPAT
5644long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5645{
5646 switch (cmd) {
5647 case FS_IOC32_GETFLAGS:
5648 cmd = FS_IOC_GETFLAGS;
5649 break;
5650 case FS_IOC32_SETFLAGS:
5651 cmd = FS_IOC_SETFLAGS;
5652 break;
5653 case FS_IOC32_GETVERSION:
5654 cmd = FS_IOC_GETVERSION;
5655 break;
5656 default:
5657 return -ENOIOCTLCMD;
5658 }
5659
5660 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
5661}
5662#endif