1f46b5a66SChristoph Hellwig /* 2f46b5a66SChristoph Hellwig * Copyright (C) 2007 Oracle. All rights reserved. 3f46b5a66SChristoph Hellwig * 4f46b5a66SChristoph Hellwig * This program is free software; you can redistribute it and/or 5f46b5a66SChristoph Hellwig * modify it under the terms of the GNU General Public 6f46b5a66SChristoph Hellwig * License v2 as published by the Free Software Foundation. 7f46b5a66SChristoph Hellwig * 8f46b5a66SChristoph Hellwig * This program is distributed in the hope that it will be useful, 9f46b5a66SChristoph Hellwig * but WITHOUT ANY WARRANTY; without even the implied warranty of 10f46b5a66SChristoph Hellwig * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 11f46b5a66SChristoph Hellwig * General Public License for more details. 12f46b5a66SChristoph Hellwig * 13f46b5a66SChristoph Hellwig * You should have received a copy of the GNU General Public 14f46b5a66SChristoph Hellwig * License along with this program; if not, write to the 15f46b5a66SChristoph Hellwig * Free Software Foundation, Inc., 59 Temple Place - Suite 330, 16f46b5a66SChristoph Hellwig * Boston, MA 021110-1307, USA. 17f46b5a66SChristoph Hellwig */ 18f46b5a66SChristoph Hellwig 19f46b5a66SChristoph Hellwig #include <linux/kernel.h> 20f46b5a66SChristoph Hellwig #include <linux/bio.h> 21f46b5a66SChristoph Hellwig #include <linux/buffer_head.h> 22f46b5a66SChristoph Hellwig #include <linux/file.h> 23f46b5a66SChristoph Hellwig #include <linux/fs.h> 24cb8e7090SChristoph Hellwig #include <linux/fsnotify.h> 25f46b5a66SChristoph Hellwig #include <linux/pagemap.h> 26f46b5a66SChristoph Hellwig #include <linux/highmem.h> 27f46b5a66SChristoph Hellwig #include <linux/time.h> 28f46b5a66SChristoph Hellwig #include <linux/init.h> 29f46b5a66SChristoph Hellwig #include <linux/string.h> 30f46b5a66SChristoph Hellwig #include <linux/backing-dev.h> 31cb8e7090SChristoph Hellwig #include <linux/mount.h> 32f46b5a66SChristoph Hellwig #include <linux/mpage.h> 33cb8e7090SChristoph Hellwig #include <linux/namei.h> 34f46b5a66SChristoph Hellwig #include <linux/swap.h> 35f46b5a66SChristoph Hellwig #include <linux/writeback.h> 36f46b5a66SChristoph Hellwig #include <linux/statfs.h> 37f46b5a66SChristoph Hellwig #include <linux/compat.h> 38f46b5a66SChristoph Hellwig #include <linux/bit_spinlock.h> 39cb8e7090SChristoph Hellwig #include <linux/security.h> 40f46b5a66SChristoph Hellwig #include <linux/xattr.h> 417ea394f1SYan Zheng #include <linux/vmalloc.h> 425a0e3ad6STejun Heo #include <linux/slab.h> 43f7039b1dSLi Dongyang #include <linux/blkdev.h> 444b4e25f2SChris Mason #include "compat.h" 45f46b5a66SChristoph Hellwig #include "ctree.h" 46f46b5a66SChristoph Hellwig #include "disk-io.h" 47f46b5a66SChristoph Hellwig #include "transaction.h" 48f46b5a66SChristoph Hellwig #include "btrfs_inode.h" 49f46b5a66SChristoph Hellwig #include "ioctl.h" 50f46b5a66SChristoph Hellwig #include "print-tree.h" 51f46b5a66SChristoph Hellwig #include "volumes.h" 52925baeddSChris Mason #include "locking.h" 53581bb050SLi Zefan #include "inode-map.h" 54d7728c96SJan Schmidt #include "backref.h" 55606686eeSJosef Bacik #include "rcu-string.h" 56f46b5a66SChristoph Hellwig 576cbff00fSChristoph Hellwig /* Mask out flags that are inappropriate for the given type of inode. */ 586cbff00fSChristoph Hellwig static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags) 596cbff00fSChristoph Hellwig { 606cbff00fSChristoph Hellwig if (S_ISDIR(mode)) 616cbff00fSChristoph Hellwig return flags; 626cbff00fSChristoph Hellwig else if (S_ISREG(mode)) 636cbff00fSChristoph Hellwig return flags & ~FS_DIRSYNC_FL; 646cbff00fSChristoph Hellwig else 656cbff00fSChristoph Hellwig return flags & (FS_NODUMP_FL | FS_NOATIME_FL); 666cbff00fSChristoph Hellwig } 67f46b5a66SChristoph Hellwig 686cbff00fSChristoph Hellwig /* 696cbff00fSChristoph Hellwig * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl. 706cbff00fSChristoph Hellwig */ 716cbff00fSChristoph Hellwig static unsigned int btrfs_flags_to_ioctl(unsigned int flags) 726cbff00fSChristoph Hellwig { 736cbff00fSChristoph Hellwig unsigned int iflags = 0; 746cbff00fSChristoph Hellwig 756cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_SYNC) 766cbff00fSChristoph Hellwig iflags |= FS_SYNC_FL; 776cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_IMMUTABLE) 786cbff00fSChristoph Hellwig iflags |= FS_IMMUTABLE_FL; 796cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_APPEND) 806cbff00fSChristoph Hellwig iflags |= FS_APPEND_FL; 816cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NODUMP) 826cbff00fSChristoph Hellwig iflags |= FS_NODUMP_FL; 836cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NOATIME) 846cbff00fSChristoph Hellwig iflags |= FS_NOATIME_FL; 856cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_DIRSYNC) 866cbff00fSChristoph Hellwig iflags |= FS_DIRSYNC_FL; 87d0092bddSLi Zefan if (flags & BTRFS_INODE_NODATACOW) 88d0092bddSLi Zefan iflags |= FS_NOCOW_FL; 89d0092bddSLi Zefan 90d0092bddSLi Zefan if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS)) 91d0092bddSLi Zefan iflags |= FS_COMPR_FL; 92d0092bddSLi Zefan else if (flags & BTRFS_INODE_NOCOMPRESS) 93d0092bddSLi Zefan iflags |= FS_NOCOMP_FL; 946cbff00fSChristoph Hellwig 956cbff00fSChristoph Hellwig return iflags; 966cbff00fSChristoph Hellwig } 976cbff00fSChristoph Hellwig 986cbff00fSChristoph Hellwig /* 996cbff00fSChristoph Hellwig * Update inode->i_flags based on the btrfs internal flags. 1006cbff00fSChristoph Hellwig */ 1016cbff00fSChristoph Hellwig void btrfs_update_iflags(struct inode *inode) 1026cbff00fSChristoph Hellwig { 1036cbff00fSChristoph Hellwig struct btrfs_inode *ip = BTRFS_I(inode); 1046cbff00fSChristoph Hellwig 1056cbff00fSChristoph Hellwig inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC); 1066cbff00fSChristoph Hellwig 1076cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_SYNC) 1086cbff00fSChristoph Hellwig inode->i_flags |= S_SYNC; 1096cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_IMMUTABLE) 1106cbff00fSChristoph Hellwig inode->i_flags |= S_IMMUTABLE; 1116cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_APPEND) 1126cbff00fSChristoph Hellwig inode->i_flags |= S_APPEND; 1136cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_NOATIME) 1146cbff00fSChristoph Hellwig inode->i_flags |= S_NOATIME; 1156cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_DIRSYNC) 1166cbff00fSChristoph Hellwig inode->i_flags |= S_DIRSYNC; 1176cbff00fSChristoph Hellwig } 1186cbff00fSChristoph Hellwig 1196cbff00fSChristoph Hellwig /* 1206cbff00fSChristoph Hellwig * Inherit flags from the parent inode. 1216cbff00fSChristoph Hellwig * 122e27425d6SJosef Bacik * Currently only the compression flags and the cow flags are inherited. 1236cbff00fSChristoph Hellwig */ 1246cbff00fSChristoph Hellwig void btrfs_inherit_iflags(struct inode *inode, struct inode *dir) 1256cbff00fSChristoph Hellwig { 1260b4dcea5SChris Mason unsigned int flags; 1270b4dcea5SChris Mason 1280b4dcea5SChris Mason if (!dir) 1290b4dcea5SChris Mason return; 1300b4dcea5SChris Mason 1310b4dcea5SChris Mason flags = BTRFS_I(dir)->flags; 1326cbff00fSChristoph Hellwig 133e27425d6SJosef Bacik if (flags & BTRFS_INODE_NOCOMPRESS) { 134e27425d6SJosef Bacik BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS; 135e27425d6SJosef Bacik BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS; 136e27425d6SJosef Bacik } else if (flags & BTRFS_INODE_COMPRESS) { 137e27425d6SJosef Bacik BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS; 138e27425d6SJosef Bacik BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS; 139e27425d6SJosef Bacik } 1406cbff00fSChristoph Hellwig 141e27425d6SJosef Bacik if (flags & BTRFS_INODE_NODATACOW) 142e27425d6SJosef Bacik BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW; 143e27425d6SJosef Bacik 1446cbff00fSChristoph Hellwig btrfs_update_iflags(inode); 1456cbff00fSChristoph Hellwig } 1466cbff00fSChristoph Hellwig 1476cbff00fSChristoph Hellwig static int btrfs_ioctl_getflags(struct file *file, void __user *arg) 1486cbff00fSChristoph Hellwig { 1496cbff00fSChristoph Hellwig struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode); 1506cbff00fSChristoph Hellwig unsigned int flags = btrfs_flags_to_ioctl(ip->flags); 1516cbff00fSChristoph Hellwig 1526cbff00fSChristoph Hellwig if (copy_to_user(arg, &flags, sizeof(flags))) 1536cbff00fSChristoph Hellwig return -EFAULT; 1546cbff00fSChristoph Hellwig return 0; 1556cbff00fSChristoph Hellwig } 1566cbff00fSChristoph Hellwig 15775e7cb7fSLiu Bo static int check_flags(unsigned int flags) 15875e7cb7fSLiu Bo { 15975e7cb7fSLiu Bo if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \ 16075e7cb7fSLiu Bo FS_NOATIME_FL | FS_NODUMP_FL | \ 16175e7cb7fSLiu Bo FS_SYNC_FL | FS_DIRSYNC_FL | \ 162e1e8fb6aSLi Zefan FS_NOCOMP_FL | FS_COMPR_FL | 163e1e8fb6aSLi Zefan FS_NOCOW_FL)) 16475e7cb7fSLiu Bo return -EOPNOTSUPP; 16575e7cb7fSLiu Bo 16675e7cb7fSLiu Bo if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL)) 16775e7cb7fSLiu Bo return -EINVAL; 16875e7cb7fSLiu Bo 16975e7cb7fSLiu Bo return 0; 17075e7cb7fSLiu Bo } 17175e7cb7fSLiu Bo 1726cbff00fSChristoph Hellwig static int btrfs_ioctl_setflags(struct file *file, void __user *arg) 1736cbff00fSChristoph Hellwig { 1746cbff00fSChristoph Hellwig struct inode *inode = file->f_path.dentry->d_inode; 1756cbff00fSChristoph Hellwig struct btrfs_inode *ip = BTRFS_I(inode); 1766cbff00fSChristoph Hellwig struct btrfs_root *root = ip->root; 1776cbff00fSChristoph Hellwig struct btrfs_trans_handle *trans; 1786cbff00fSChristoph Hellwig unsigned int flags, oldflags; 1796cbff00fSChristoph Hellwig int ret; 180f062abf0SLi Zefan u64 ip_oldflags; 181f062abf0SLi Zefan unsigned int i_oldflags; 1826cbff00fSChristoph Hellwig 183b83cc969SLi Zefan if (btrfs_root_readonly(root)) 184b83cc969SLi Zefan return -EROFS; 185b83cc969SLi Zefan 1866cbff00fSChristoph Hellwig if (copy_from_user(&flags, arg, sizeof(flags))) 1876cbff00fSChristoph Hellwig return -EFAULT; 1886cbff00fSChristoph Hellwig 18975e7cb7fSLiu Bo ret = check_flags(flags); 19075e7cb7fSLiu Bo if (ret) 19175e7cb7fSLiu Bo return ret; 1926cbff00fSChristoph Hellwig 1932e149670SSerge E. Hallyn if (!inode_owner_or_capable(inode)) 1946cbff00fSChristoph Hellwig return -EACCES; 1956cbff00fSChristoph Hellwig 196*e7848683SJan Kara ret = mnt_want_write_file(file); 197*e7848683SJan Kara if (ret) 198*e7848683SJan Kara return ret; 199*e7848683SJan Kara 2006cbff00fSChristoph Hellwig mutex_lock(&inode->i_mutex); 2016cbff00fSChristoph Hellwig 202f062abf0SLi Zefan ip_oldflags = ip->flags; 203f062abf0SLi Zefan i_oldflags = inode->i_flags; 204f062abf0SLi Zefan 2056cbff00fSChristoph Hellwig flags = btrfs_mask_flags(inode->i_mode, flags); 2066cbff00fSChristoph Hellwig oldflags = btrfs_flags_to_ioctl(ip->flags); 2076cbff00fSChristoph Hellwig if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) { 2086cbff00fSChristoph Hellwig if (!capable(CAP_LINUX_IMMUTABLE)) { 2096cbff00fSChristoph Hellwig ret = -EPERM; 2106cbff00fSChristoph Hellwig goto out_unlock; 2116cbff00fSChristoph Hellwig } 2126cbff00fSChristoph Hellwig } 2136cbff00fSChristoph Hellwig 2146cbff00fSChristoph Hellwig if (flags & FS_SYNC_FL) 2156cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_SYNC; 2166cbff00fSChristoph Hellwig else 2176cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_SYNC; 2186cbff00fSChristoph Hellwig if (flags & FS_IMMUTABLE_FL) 2196cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_IMMUTABLE; 2206cbff00fSChristoph Hellwig else 2216cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_IMMUTABLE; 2226cbff00fSChristoph Hellwig if (flags & FS_APPEND_FL) 2236cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_APPEND; 2246cbff00fSChristoph Hellwig else 2256cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_APPEND; 2266cbff00fSChristoph Hellwig if (flags & FS_NODUMP_FL) 2276cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_NODUMP; 2286cbff00fSChristoph Hellwig else 2296cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_NODUMP; 2306cbff00fSChristoph Hellwig if (flags & FS_NOATIME_FL) 2316cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_NOATIME; 2326cbff00fSChristoph Hellwig else 2336cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_NOATIME; 2346cbff00fSChristoph Hellwig if (flags & FS_DIRSYNC_FL) 2356cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_DIRSYNC; 2366cbff00fSChristoph Hellwig else 2376cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_DIRSYNC; 238e1e8fb6aSLi Zefan if (flags & FS_NOCOW_FL) 239e1e8fb6aSLi Zefan ip->flags |= BTRFS_INODE_NODATACOW; 240e1e8fb6aSLi Zefan else 241e1e8fb6aSLi Zefan ip->flags &= ~BTRFS_INODE_NODATACOW; 2426cbff00fSChristoph Hellwig 24375e7cb7fSLiu Bo /* 24475e7cb7fSLiu Bo * The COMPRESS flag can only be changed by users, while the NOCOMPRESS 24575e7cb7fSLiu Bo * flag may be changed automatically if compression code won't make 24675e7cb7fSLiu Bo * things smaller. 24775e7cb7fSLiu Bo */ 24875e7cb7fSLiu Bo if (flags & FS_NOCOMP_FL) { 24975e7cb7fSLiu Bo ip->flags &= ~BTRFS_INODE_COMPRESS; 25075e7cb7fSLiu Bo ip->flags |= BTRFS_INODE_NOCOMPRESS; 25175e7cb7fSLiu Bo } else if (flags & FS_COMPR_FL) { 25275e7cb7fSLiu Bo ip->flags |= BTRFS_INODE_COMPRESS; 25375e7cb7fSLiu Bo ip->flags &= ~BTRFS_INODE_NOCOMPRESS; 254ebcb904dSLi Zefan } else { 255ebcb904dSLi Zefan ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS); 25675e7cb7fSLiu Bo } 2576cbff00fSChristoph Hellwig 2584da6f1a3SLi Zefan trans = btrfs_start_transaction(root, 1); 259f062abf0SLi Zefan if (IS_ERR(trans)) { 260f062abf0SLi Zefan ret = PTR_ERR(trans); 261f062abf0SLi Zefan goto out_drop; 262f062abf0SLi Zefan } 2636cbff00fSChristoph Hellwig 264306424ccSLi Zefan btrfs_update_iflags(inode); 2650c4d2d95SJosef Bacik inode_inc_iversion(inode); 266306424ccSLi Zefan inode->i_ctime = CURRENT_TIME; 2676cbff00fSChristoph Hellwig ret = btrfs_update_inode(trans, root, inode); 2686cbff00fSChristoph Hellwig 2696cbff00fSChristoph Hellwig btrfs_end_transaction(trans, root); 270f062abf0SLi Zefan out_drop: 271f062abf0SLi Zefan if (ret) { 272f062abf0SLi Zefan ip->flags = ip_oldflags; 273f062abf0SLi Zefan inode->i_flags = i_oldflags; 274f062abf0SLi Zefan } 2756cbff00fSChristoph Hellwig 2766cbff00fSChristoph Hellwig out_unlock: 2776cbff00fSChristoph Hellwig mutex_unlock(&inode->i_mutex); 278*e7848683SJan Kara mnt_drop_write_file(file); 2792d4e6f6aSliubo return ret; 2806cbff00fSChristoph Hellwig } 2816cbff00fSChristoph Hellwig 2826cbff00fSChristoph Hellwig static int btrfs_ioctl_getversion(struct file *file, int __user *arg) 2836cbff00fSChristoph Hellwig { 2846cbff00fSChristoph Hellwig struct inode *inode = file->f_path.dentry->d_inode; 2856cbff00fSChristoph Hellwig 2866cbff00fSChristoph Hellwig return put_user(inode->i_generation, arg); 2876cbff00fSChristoph Hellwig } 288f46b5a66SChristoph Hellwig 289f7039b1dSLi Dongyang static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg) 290f7039b1dSLi Dongyang { 291815745cfSAl Viro struct btrfs_fs_info *fs_info = btrfs_sb(fdentry(file)->d_sb); 292f7039b1dSLi Dongyang struct btrfs_device *device; 293f7039b1dSLi Dongyang struct request_queue *q; 294f7039b1dSLi Dongyang struct fstrim_range range; 295f7039b1dSLi Dongyang u64 minlen = ULLONG_MAX; 296f7039b1dSLi Dongyang u64 num_devices = 0; 297815745cfSAl Viro u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy); 298f7039b1dSLi Dongyang int ret; 299f7039b1dSLi Dongyang 300f7039b1dSLi Dongyang if (!capable(CAP_SYS_ADMIN)) 301f7039b1dSLi Dongyang return -EPERM; 302f7039b1dSLi Dongyang 3031f78160cSXiao Guangrong rcu_read_lock(); 3041f78160cSXiao Guangrong list_for_each_entry_rcu(device, &fs_info->fs_devices->devices, 3051f78160cSXiao Guangrong dev_list) { 306f7039b1dSLi Dongyang if (!device->bdev) 307f7039b1dSLi Dongyang continue; 308f7039b1dSLi Dongyang q = bdev_get_queue(device->bdev); 309f7039b1dSLi Dongyang if (blk_queue_discard(q)) { 310f7039b1dSLi Dongyang num_devices++; 311f7039b1dSLi Dongyang minlen = min((u64)q->limits.discard_granularity, 312f7039b1dSLi Dongyang minlen); 313f7039b1dSLi Dongyang } 314f7039b1dSLi Dongyang } 3151f78160cSXiao Guangrong rcu_read_unlock(); 316f4c697e6SLukas Czerner 317f7039b1dSLi Dongyang if (!num_devices) 318f7039b1dSLi Dongyang return -EOPNOTSUPP; 319f7039b1dSLi Dongyang if (copy_from_user(&range, arg, sizeof(range))) 320f7039b1dSLi Dongyang return -EFAULT; 321f4c697e6SLukas Czerner if (range.start > total_bytes) 322f4c697e6SLukas Czerner return -EINVAL; 323f7039b1dSLi Dongyang 324f4c697e6SLukas Czerner range.len = min(range.len, total_bytes - range.start); 325f7039b1dSLi Dongyang range.minlen = max(range.minlen, minlen); 326815745cfSAl Viro ret = btrfs_trim_fs(fs_info->tree_root, &range); 327f7039b1dSLi Dongyang if (ret < 0) 328f7039b1dSLi Dongyang return ret; 329f7039b1dSLi Dongyang 330f7039b1dSLi Dongyang if (copy_to_user(arg, &range, sizeof(range))) 331f7039b1dSLi Dongyang return -EFAULT; 332f7039b1dSLi Dongyang 333f7039b1dSLi Dongyang return 0; 334f7039b1dSLi Dongyang } 335f7039b1dSLi Dongyang 336cb8e7090SChristoph Hellwig static noinline int create_subvol(struct btrfs_root *root, 337cb8e7090SChristoph Hellwig struct dentry *dentry, 33872fd032eSSage Weil char *name, int namelen, 33972fd032eSSage Weil u64 *async_transid) 340f46b5a66SChristoph Hellwig { 341f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 342f46b5a66SChristoph Hellwig struct btrfs_key key; 343f46b5a66SChristoph Hellwig struct btrfs_root_item root_item; 344f46b5a66SChristoph Hellwig struct btrfs_inode_item *inode_item; 345f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 34676dda93cSYan, Zheng struct btrfs_root *new_root; 3472fbe8c8aSAl Viro struct dentry *parent = dentry->d_parent; 3486a912213SJosef Bacik struct inode *dir; 349f46b5a66SChristoph Hellwig int ret; 350f46b5a66SChristoph Hellwig int err; 351f46b5a66SChristoph Hellwig u64 objectid; 352f46b5a66SChristoph Hellwig u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; 3533de4586cSChris Mason u64 index = 0; 354f46b5a66SChristoph Hellwig 355581bb050SLi Zefan ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid); 3562fbe8c8aSAl Viro if (ret) 357a22285a6SYan, Zheng return ret; 3586a912213SJosef Bacik 3596a912213SJosef Bacik dir = parent->d_inode; 3606a912213SJosef Bacik 3619ed74f2dSJosef Bacik /* 3629ed74f2dSJosef Bacik * 1 - inode item 3639ed74f2dSJosef Bacik * 2 - refs 3649ed74f2dSJosef Bacik * 1 - root item 3659ed74f2dSJosef Bacik * 2 - dir items 3669ed74f2dSJosef Bacik */ 367a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 6); 3682fbe8c8aSAl Viro if (IS_ERR(trans)) 369a22285a6SYan, Zheng return PTR_ERR(trans); 370f46b5a66SChristoph Hellwig 3715d4f98a2SYan Zheng leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 3725581a51aSJan Schmidt 0, objectid, NULL, 0, 0, 0); 3738e8a1e31SJosef Bacik if (IS_ERR(leaf)) { 3748e8a1e31SJosef Bacik ret = PTR_ERR(leaf); 3758e8a1e31SJosef Bacik goto fail; 3768e8a1e31SJosef Bacik } 377f46b5a66SChristoph Hellwig 3785d4f98a2SYan Zheng memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); 379f46b5a66SChristoph Hellwig btrfs_set_header_bytenr(leaf, leaf->start); 380f46b5a66SChristoph Hellwig btrfs_set_header_generation(leaf, trans->transid); 3815d4f98a2SYan Zheng btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 382f46b5a66SChristoph Hellwig btrfs_set_header_owner(leaf, objectid); 383f46b5a66SChristoph Hellwig 384f46b5a66SChristoph Hellwig write_extent_buffer(leaf, root->fs_info->fsid, 385f46b5a66SChristoph Hellwig (unsigned long)btrfs_header_fsid(leaf), 386f46b5a66SChristoph Hellwig BTRFS_FSID_SIZE); 3875d4f98a2SYan Zheng write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid, 3885d4f98a2SYan Zheng (unsigned long)btrfs_header_chunk_tree_uuid(leaf), 3895d4f98a2SYan Zheng BTRFS_UUID_SIZE); 390f46b5a66SChristoph Hellwig btrfs_mark_buffer_dirty(leaf); 391f46b5a66SChristoph Hellwig 392f46b5a66SChristoph Hellwig inode_item = &root_item.inode; 393f46b5a66SChristoph Hellwig memset(inode_item, 0, sizeof(*inode_item)); 394f46b5a66SChristoph Hellwig inode_item->generation = cpu_to_le64(1); 395f46b5a66SChristoph Hellwig inode_item->size = cpu_to_le64(3); 396f46b5a66SChristoph Hellwig inode_item->nlink = cpu_to_le32(1); 397a76a3cd4SYan Zheng inode_item->nbytes = cpu_to_le64(root->leafsize); 398f46b5a66SChristoph Hellwig inode_item->mode = cpu_to_le32(S_IFDIR | 0755); 399f46b5a66SChristoph Hellwig 40008fe4db1SLi Zefan root_item.flags = 0; 40108fe4db1SLi Zefan root_item.byte_limit = 0; 40208fe4db1SLi Zefan inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT); 40308fe4db1SLi Zefan 404f46b5a66SChristoph Hellwig btrfs_set_root_bytenr(&root_item, leaf->start); 40584234f3aSYan Zheng btrfs_set_root_generation(&root_item, trans->transid); 406f46b5a66SChristoph Hellwig btrfs_set_root_level(&root_item, 0); 407f46b5a66SChristoph Hellwig btrfs_set_root_refs(&root_item, 1); 40886b9f2ecSYan, Zheng btrfs_set_root_used(&root_item, leaf->len); 40980ff3856SYan Zheng btrfs_set_root_last_snapshot(&root_item, 0); 410f46b5a66SChristoph Hellwig 411f46b5a66SChristoph Hellwig memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress)); 412f46b5a66SChristoph Hellwig root_item.drop_level = 0; 413f46b5a66SChristoph Hellwig 414925baeddSChris Mason btrfs_tree_unlock(leaf); 415f46b5a66SChristoph Hellwig free_extent_buffer(leaf); 416f46b5a66SChristoph Hellwig leaf = NULL; 417f46b5a66SChristoph Hellwig 418f46b5a66SChristoph Hellwig btrfs_set_root_dirid(&root_item, new_dirid); 419f46b5a66SChristoph Hellwig 420f46b5a66SChristoph Hellwig key.objectid = objectid; 4215d4f98a2SYan Zheng key.offset = 0; 422f46b5a66SChristoph Hellwig btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); 423f46b5a66SChristoph Hellwig ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, 424f46b5a66SChristoph Hellwig &root_item); 425f46b5a66SChristoph Hellwig if (ret) 426f46b5a66SChristoph Hellwig goto fail; 427f46b5a66SChristoph Hellwig 42876dda93cSYan, Zheng key.offset = (u64)-1; 42976dda93cSYan, Zheng new_root = btrfs_read_fs_root_no_name(root->fs_info, &key); 43079787eaaSJeff Mahoney if (IS_ERR(new_root)) { 43179787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, PTR_ERR(new_root)); 43279787eaaSJeff Mahoney ret = PTR_ERR(new_root); 43379787eaaSJeff Mahoney goto fail; 43479787eaaSJeff Mahoney } 43576dda93cSYan, Zheng 43676dda93cSYan, Zheng btrfs_record_root_in_trans(trans, new_root); 43776dda93cSYan, Zheng 438d82a6f1dSJosef Bacik ret = btrfs_create_subvol_root(trans, new_root, new_dirid); 439ce598979SMark Fasheh if (ret) { 440ce598979SMark Fasheh /* We potentially lose an unused inode item here */ 44179787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, ret); 442ce598979SMark Fasheh goto fail; 443ce598979SMark Fasheh } 444ce598979SMark Fasheh 445f46b5a66SChristoph Hellwig /* 446f46b5a66SChristoph Hellwig * insert the directory item 447f46b5a66SChristoph Hellwig */ 4483de4586cSChris Mason ret = btrfs_set_inode_index(dir, &index); 44979787eaaSJeff Mahoney if (ret) { 45079787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, ret); 45179787eaaSJeff Mahoney goto fail; 45279787eaaSJeff Mahoney } 4533de4586cSChris Mason 4543de4586cSChris Mason ret = btrfs_insert_dir_item(trans, root, 45516cdcec7SMiao Xie name, namelen, dir, &key, 4563de4586cSChris Mason BTRFS_FT_DIR, index); 45779787eaaSJeff Mahoney if (ret) { 45879787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, ret); 459f46b5a66SChristoph Hellwig goto fail; 46079787eaaSJeff Mahoney } 4610660b5afSChris Mason 46252c26179SYan Zheng btrfs_i_size_write(dir, dir->i_size + namelen * 2); 46352c26179SYan Zheng ret = btrfs_update_inode(trans, root, dir); 46452c26179SYan Zheng BUG_ON(ret); 46552c26179SYan Zheng 4660660b5afSChris Mason ret = btrfs_add_root_ref(trans, root->fs_info->tree_root, 4674df27c4dSYan, Zheng objectid, root->root_key.objectid, 46833345d01SLi Zefan btrfs_ino(dir), index, name, namelen); 46976dda93cSYan, Zheng 4700660b5afSChris Mason BUG_ON(ret); 4710660b5afSChris Mason 47276dda93cSYan, Zheng d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry)); 473f46b5a66SChristoph Hellwig fail: 47472fd032eSSage Weil if (async_transid) { 47572fd032eSSage Weil *async_transid = trans->transid; 47672fd032eSSage Weil err = btrfs_commit_transaction_async(trans, root, 1); 47772fd032eSSage Weil } else { 47876dda93cSYan, Zheng err = btrfs_commit_transaction(trans, root); 47972fd032eSSage Weil } 480f46b5a66SChristoph Hellwig if (err && !ret) 481f46b5a66SChristoph Hellwig ret = err; 482f46b5a66SChristoph Hellwig return ret; 483f46b5a66SChristoph Hellwig } 484f46b5a66SChristoph Hellwig 48572fd032eSSage Weil static int create_snapshot(struct btrfs_root *root, struct dentry *dentry, 486b83cc969SLi Zefan char *name, int namelen, u64 *async_transid, 487b83cc969SLi Zefan bool readonly) 488f46b5a66SChristoph Hellwig { 4892e4bfab9SYan, Zheng struct inode *inode; 490f46b5a66SChristoph Hellwig struct btrfs_pending_snapshot *pending_snapshot; 491f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 4922e4bfab9SYan, Zheng int ret; 493f46b5a66SChristoph Hellwig 494f46b5a66SChristoph Hellwig if (!root->ref_cows) 495f46b5a66SChristoph Hellwig return -EINVAL; 496f46b5a66SChristoph Hellwig 497a22285a6SYan, Zheng pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS); 498a22285a6SYan, Zheng if (!pending_snapshot) 499a22285a6SYan, Zheng return -ENOMEM; 500a22285a6SYan, Zheng 501a22285a6SYan, Zheng btrfs_init_block_rsv(&pending_snapshot->block_rsv); 502a22285a6SYan, Zheng pending_snapshot->dentry = dentry; 503a22285a6SYan, Zheng pending_snapshot->root = root; 504b83cc969SLi Zefan pending_snapshot->readonly = readonly; 505a22285a6SYan, Zheng 506a22285a6SYan, Zheng trans = btrfs_start_transaction(root->fs_info->extent_root, 5); 507a22285a6SYan, Zheng if (IS_ERR(trans)) { 508a22285a6SYan, Zheng ret = PTR_ERR(trans); 509a22285a6SYan, Zheng goto fail; 510a22285a6SYan, Zheng } 511a22285a6SYan, Zheng 512a22285a6SYan, Zheng ret = btrfs_snap_reserve_metadata(trans, pending_snapshot); 513a22285a6SYan, Zheng BUG_ON(ret); 514a22285a6SYan, Zheng 5158351583eSJosef Bacik spin_lock(&root->fs_info->trans_lock); 516a22285a6SYan, Zheng list_add(&pending_snapshot->list, 517a22285a6SYan, Zheng &trans->transaction->pending_snapshots); 5188351583eSJosef Bacik spin_unlock(&root->fs_info->trans_lock); 51972fd032eSSage Weil if (async_transid) { 52072fd032eSSage Weil *async_transid = trans->transid; 52172fd032eSSage Weil ret = btrfs_commit_transaction_async(trans, 52272fd032eSSage Weil root->fs_info->extent_root, 1); 52372fd032eSSage Weil } else { 52472fd032eSSage Weil ret = btrfs_commit_transaction(trans, 52572fd032eSSage Weil root->fs_info->extent_root); 52672fd032eSSage Weil } 527a22285a6SYan, Zheng BUG_ON(ret); 528a22285a6SYan, Zheng 529a22285a6SYan, Zheng ret = pending_snapshot->error; 530f46b5a66SChristoph Hellwig if (ret) 5312e4bfab9SYan, Zheng goto fail; 532f46b5a66SChristoph Hellwig 53366b4ffd1SJosef Bacik ret = btrfs_orphan_cleanup(pending_snapshot->snap); 53466b4ffd1SJosef Bacik if (ret) 53566b4ffd1SJosef Bacik goto fail; 536f46b5a66SChristoph Hellwig 5372fbe8c8aSAl Viro inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry); 5382e4bfab9SYan, Zheng if (IS_ERR(inode)) { 5392e4bfab9SYan, Zheng ret = PTR_ERR(inode); 5402e4bfab9SYan, Zheng goto fail; 5412e4bfab9SYan, Zheng } 5422e4bfab9SYan, Zheng BUG_ON(!inode); 5432e4bfab9SYan, Zheng d_instantiate(dentry, inode); 5442e4bfab9SYan, Zheng ret = 0; 5452e4bfab9SYan, Zheng fail: 546a22285a6SYan, Zheng kfree(pending_snapshot); 547f46b5a66SChristoph Hellwig return ret; 548f46b5a66SChristoph Hellwig } 549f46b5a66SChristoph Hellwig 5504260f7c7SSage Weil /* copy of check_sticky in fs/namei.c() 5514260f7c7SSage Weil * It's inline, so penalty for filesystems that don't use sticky bit is 5524260f7c7SSage Weil * minimal. 5534260f7c7SSage Weil */ 5544260f7c7SSage Weil static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode) 5554260f7c7SSage Weil { 5564260f7c7SSage Weil uid_t fsuid = current_fsuid(); 5574260f7c7SSage Weil 5584260f7c7SSage Weil if (!(dir->i_mode & S_ISVTX)) 5594260f7c7SSage Weil return 0; 5604260f7c7SSage Weil if (inode->i_uid == fsuid) 5614260f7c7SSage Weil return 0; 5624260f7c7SSage Weil if (dir->i_uid == fsuid) 5634260f7c7SSage Weil return 0; 5644260f7c7SSage Weil return !capable(CAP_FOWNER); 5654260f7c7SSage Weil } 5664260f7c7SSage Weil 5674260f7c7SSage Weil /* copy of may_delete in fs/namei.c() 5684260f7c7SSage Weil * Check whether we can remove a link victim from directory dir, check 5694260f7c7SSage Weil * whether the type of victim is right. 5704260f7c7SSage Weil * 1. We can't do it if dir is read-only (done in permission()) 5714260f7c7SSage Weil * 2. We should have write and exec permissions on dir 5724260f7c7SSage Weil * 3. We can't remove anything from append-only dir 5734260f7c7SSage Weil * 4. We can't do anything with immutable dir (done in permission()) 5744260f7c7SSage Weil * 5. If the sticky bit on dir is set we should either 5754260f7c7SSage Weil * a. be owner of dir, or 5764260f7c7SSage Weil * b. be owner of victim, or 5774260f7c7SSage Weil * c. have CAP_FOWNER capability 5784260f7c7SSage Weil * 6. If the victim is append-only or immutable we can't do antyhing with 5794260f7c7SSage Weil * links pointing to it. 5804260f7c7SSage Weil * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. 5814260f7c7SSage Weil * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. 5824260f7c7SSage Weil * 9. We can't remove a root or mountpoint. 5834260f7c7SSage Weil * 10. We don't allow removal of NFS sillyrenamed files; it's handled by 5844260f7c7SSage Weil * nfs_async_unlink(). 5854260f7c7SSage Weil */ 5864260f7c7SSage Weil 5874260f7c7SSage Weil static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir) 5884260f7c7SSage Weil { 5894260f7c7SSage Weil int error; 5904260f7c7SSage Weil 5914260f7c7SSage Weil if (!victim->d_inode) 5924260f7c7SSage Weil return -ENOENT; 5934260f7c7SSage Weil 5944260f7c7SSage Weil BUG_ON(victim->d_parent->d_inode != dir); 5954260f7c7SSage Weil audit_inode_child(victim, dir); 5964260f7c7SSage Weil 5974260f7c7SSage Weil error = inode_permission(dir, MAY_WRITE | MAY_EXEC); 5984260f7c7SSage Weil if (error) 5994260f7c7SSage Weil return error; 6004260f7c7SSage Weil if (IS_APPEND(dir)) 6014260f7c7SSage Weil return -EPERM; 6024260f7c7SSage Weil if (btrfs_check_sticky(dir, victim->d_inode)|| 6034260f7c7SSage Weil IS_APPEND(victim->d_inode)|| 6044260f7c7SSage Weil IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode)) 6054260f7c7SSage Weil return -EPERM; 6064260f7c7SSage Weil if (isdir) { 6074260f7c7SSage Weil if (!S_ISDIR(victim->d_inode->i_mode)) 6084260f7c7SSage Weil return -ENOTDIR; 6094260f7c7SSage Weil if (IS_ROOT(victim)) 6104260f7c7SSage Weil return -EBUSY; 6114260f7c7SSage Weil } else if (S_ISDIR(victim->d_inode->i_mode)) 6124260f7c7SSage Weil return -EISDIR; 6134260f7c7SSage Weil if (IS_DEADDIR(dir)) 6144260f7c7SSage Weil return -ENOENT; 6154260f7c7SSage Weil if (victim->d_flags & DCACHE_NFSFS_RENAMED) 6164260f7c7SSage Weil return -EBUSY; 6174260f7c7SSage Weil return 0; 6184260f7c7SSage Weil } 6194260f7c7SSage Weil 620cb8e7090SChristoph Hellwig /* copy of may_create in fs/namei.c() */ 621cb8e7090SChristoph Hellwig static inline int btrfs_may_create(struct inode *dir, struct dentry *child) 622cb8e7090SChristoph Hellwig { 623cb8e7090SChristoph Hellwig if (child->d_inode) 624cb8e7090SChristoph Hellwig return -EEXIST; 625cb8e7090SChristoph Hellwig if (IS_DEADDIR(dir)) 626cb8e7090SChristoph Hellwig return -ENOENT; 627cb8e7090SChristoph Hellwig return inode_permission(dir, MAY_WRITE | MAY_EXEC); 628cb8e7090SChristoph Hellwig } 629cb8e7090SChristoph Hellwig 630cb8e7090SChristoph Hellwig /* 631cb8e7090SChristoph Hellwig * Create a new subvolume below @parent. This is largely modeled after 632cb8e7090SChristoph Hellwig * sys_mkdirat and vfs_mkdir, but we only do a single component lookup 633cb8e7090SChristoph Hellwig * inside this filesystem so it's quite a bit simpler. 634cb8e7090SChristoph Hellwig */ 63576dda93cSYan, Zheng static noinline int btrfs_mksubvol(struct path *parent, 63676dda93cSYan, Zheng char *name, int namelen, 63772fd032eSSage Weil struct btrfs_root *snap_src, 638b83cc969SLi Zefan u64 *async_transid, bool readonly) 639cb8e7090SChristoph Hellwig { 64076dda93cSYan, Zheng struct inode *dir = parent->dentry->d_inode; 641cb8e7090SChristoph Hellwig struct dentry *dentry; 642cb8e7090SChristoph Hellwig int error; 643cb8e7090SChristoph Hellwig 644*e7848683SJan Kara error = mnt_want_write(parent->mnt); 645*e7848683SJan Kara if (error) 646*e7848683SJan Kara return error; 647*e7848683SJan Kara 64876dda93cSYan, Zheng mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT); 649cb8e7090SChristoph Hellwig 650cb8e7090SChristoph Hellwig dentry = lookup_one_len(name, parent->dentry, namelen); 651cb8e7090SChristoph Hellwig error = PTR_ERR(dentry); 652cb8e7090SChristoph Hellwig if (IS_ERR(dentry)) 653cb8e7090SChristoph Hellwig goto out_unlock; 654cb8e7090SChristoph Hellwig 655cb8e7090SChristoph Hellwig error = -EEXIST; 656cb8e7090SChristoph Hellwig if (dentry->d_inode) 657cb8e7090SChristoph Hellwig goto out_dput; 658cb8e7090SChristoph Hellwig 65976dda93cSYan, Zheng error = btrfs_may_create(dir, dentry); 660cb8e7090SChristoph Hellwig if (error) 661*e7848683SJan Kara goto out_dput; 662cb8e7090SChristoph Hellwig 66376dda93cSYan, Zheng down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem); 66476dda93cSYan, Zheng 66576dda93cSYan, Zheng if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) 66676dda93cSYan, Zheng goto out_up_read; 66776dda93cSYan, Zheng 6683de4586cSChris Mason if (snap_src) { 66972fd032eSSage Weil error = create_snapshot(snap_src, dentry, 670b83cc969SLi Zefan name, namelen, async_transid, readonly); 6713de4586cSChris Mason } else { 67276dda93cSYan, Zheng error = create_subvol(BTRFS_I(dir)->root, dentry, 67372fd032eSSage Weil name, namelen, async_transid); 6743de4586cSChris Mason } 67576dda93cSYan, Zheng if (!error) 67676dda93cSYan, Zheng fsnotify_mkdir(dir, dentry); 67776dda93cSYan, Zheng out_up_read: 67876dda93cSYan, Zheng up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem); 679cb8e7090SChristoph Hellwig out_dput: 680cb8e7090SChristoph Hellwig dput(dentry); 681cb8e7090SChristoph Hellwig out_unlock: 68276dda93cSYan, Zheng mutex_unlock(&dir->i_mutex); 683*e7848683SJan Kara mnt_drop_write(parent->mnt); 684cb8e7090SChristoph Hellwig return error; 685cb8e7090SChristoph Hellwig } 686cb8e7090SChristoph Hellwig 6874cb5300bSChris Mason /* 6884cb5300bSChris Mason * When we're defragging a range, we don't want to kick it off again 6894cb5300bSChris Mason * if it is really just waiting for delalloc to send it down. 6904cb5300bSChris Mason * If we find a nice big extent or delalloc range for the bytes in the 6914cb5300bSChris Mason * file you want to defrag, we return 0 to let you know to skip this 6924cb5300bSChris Mason * part of the file 6934cb5300bSChris Mason */ 6944cb5300bSChris Mason static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh) 6954cb5300bSChris Mason { 6964cb5300bSChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 6974cb5300bSChris Mason struct extent_map *em = NULL; 6984cb5300bSChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 6994cb5300bSChris Mason u64 end; 7004cb5300bSChris Mason 7014cb5300bSChris Mason read_lock(&em_tree->lock); 7024cb5300bSChris Mason em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE); 7034cb5300bSChris Mason read_unlock(&em_tree->lock); 7044cb5300bSChris Mason 7054cb5300bSChris Mason if (em) { 7064cb5300bSChris Mason end = extent_map_end(em); 7074cb5300bSChris Mason free_extent_map(em); 7084cb5300bSChris Mason if (end - offset > thresh) 7094cb5300bSChris Mason return 0; 7104cb5300bSChris Mason } 7114cb5300bSChris Mason /* if we already have a nice delalloc here, just stop */ 7124cb5300bSChris Mason thresh /= 2; 7134cb5300bSChris Mason end = count_range_bits(io_tree, &offset, offset + thresh, 7144cb5300bSChris Mason thresh, EXTENT_DELALLOC, 1); 7154cb5300bSChris Mason if (end >= thresh) 7164cb5300bSChris Mason return 0; 7174cb5300bSChris Mason return 1; 7184cb5300bSChris Mason } 7194cb5300bSChris Mason 7204cb5300bSChris Mason /* 7214cb5300bSChris Mason * helper function to walk through a file and find extents 7224cb5300bSChris Mason * newer than a specific transid, and smaller than thresh. 7234cb5300bSChris Mason * 7244cb5300bSChris Mason * This is used by the defragging code to find new and small 7254cb5300bSChris Mason * extents 7264cb5300bSChris Mason */ 7274cb5300bSChris Mason static int find_new_extents(struct btrfs_root *root, 7284cb5300bSChris Mason struct inode *inode, u64 newer_than, 7294cb5300bSChris Mason u64 *off, int thresh) 7304cb5300bSChris Mason { 7314cb5300bSChris Mason struct btrfs_path *path; 7324cb5300bSChris Mason struct btrfs_key min_key; 7334cb5300bSChris Mason struct btrfs_key max_key; 7344cb5300bSChris Mason struct extent_buffer *leaf; 7354cb5300bSChris Mason struct btrfs_file_extent_item *extent; 7364cb5300bSChris Mason int type; 7374cb5300bSChris Mason int ret; 738a4689d2bSDavid Sterba u64 ino = btrfs_ino(inode); 7394cb5300bSChris Mason 7404cb5300bSChris Mason path = btrfs_alloc_path(); 7414cb5300bSChris Mason if (!path) 7424cb5300bSChris Mason return -ENOMEM; 7434cb5300bSChris Mason 744a4689d2bSDavid Sterba min_key.objectid = ino; 7454cb5300bSChris Mason min_key.type = BTRFS_EXTENT_DATA_KEY; 7464cb5300bSChris Mason min_key.offset = *off; 7474cb5300bSChris Mason 748a4689d2bSDavid Sterba max_key.objectid = ino; 7494cb5300bSChris Mason max_key.type = (u8)-1; 7504cb5300bSChris Mason max_key.offset = (u64)-1; 7514cb5300bSChris Mason 7524cb5300bSChris Mason path->keep_locks = 1; 7534cb5300bSChris Mason 7544cb5300bSChris Mason while(1) { 7554cb5300bSChris Mason ret = btrfs_search_forward(root, &min_key, &max_key, 7564cb5300bSChris Mason path, 0, newer_than); 7574cb5300bSChris Mason if (ret != 0) 7584cb5300bSChris Mason goto none; 759a4689d2bSDavid Sterba if (min_key.objectid != ino) 7604cb5300bSChris Mason goto none; 7614cb5300bSChris Mason if (min_key.type != BTRFS_EXTENT_DATA_KEY) 7624cb5300bSChris Mason goto none; 7634cb5300bSChris Mason 7644cb5300bSChris Mason leaf = path->nodes[0]; 7654cb5300bSChris Mason extent = btrfs_item_ptr(leaf, path->slots[0], 7664cb5300bSChris Mason struct btrfs_file_extent_item); 7674cb5300bSChris Mason 7684cb5300bSChris Mason type = btrfs_file_extent_type(leaf, extent); 7694cb5300bSChris Mason if (type == BTRFS_FILE_EXTENT_REG && 7704cb5300bSChris Mason btrfs_file_extent_num_bytes(leaf, extent) < thresh && 7714cb5300bSChris Mason check_defrag_in_cache(inode, min_key.offset, thresh)) { 7724cb5300bSChris Mason *off = min_key.offset; 7734cb5300bSChris Mason btrfs_free_path(path); 7744cb5300bSChris Mason return 0; 7754cb5300bSChris Mason } 7764cb5300bSChris Mason 7774cb5300bSChris Mason if (min_key.offset == (u64)-1) 7784cb5300bSChris Mason goto none; 7794cb5300bSChris Mason 7804cb5300bSChris Mason min_key.offset++; 7814cb5300bSChris Mason btrfs_release_path(path); 7824cb5300bSChris Mason } 7834cb5300bSChris Mason none: 7844cb5300bSChris Mason btrfs_free_path(path); 7854cb5300bSChris Mason return -ENOENT; 7864cb5300bSChris Mason } 7874cb5300bSChris Mason 7886c282eb4SLi Zefan static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start) 78917ce6ef8SLiu Bo { 79017ce6ef8SLiu Bo struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 791940100a4SChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 7926c282eb4SLi Zefan struct extent_map *em; 7936c282eb4SLi Zefan u64 len = PAGE_CACHE_SIZE; 794940100a4SChris Mason 795940100a4SChris Mason /* 796940100a4SChris Mason * hopefully we have this extent in the tree already, try without 797940100a4SChris Mason * the full extent lock 798940100a4SChris Mason */ 799940100a4SChris Mason read_lock(&em_tree->lock); 800940100a4SChris Mason em = lookup_extent_mapping(em_tree, start, len); 801940100a4SChris Mason read_unlock(&em_tree->lock); 802940100a4SChris Mason 803940100a4SChris Mason if (!em) { 804940100a4SChris Mason /* get the big lock and read metadata off disk */ 805d0082371SJeff Mahoney lock_extent(io_tree, start, start + len - 1); 806940100a4SChris Mason em = btrfs_get_extent(inode, NULL, 0, start, len, 0); 807d0082371SJeff Mahoney unlock_extent(io_tree, start, start + len - 1); 808940100a4SChris Mason 8096cf8bfbfSDan Carpenter if (IS_ERR(em)) 8106c282eb4SLi Zefan return NULL; 811940100a4SChris Mason } 812940100a4SChris Mason 8136c282eb4SLi Zefan return em; 8146c282eb4SLi Zefan } 8156c282eb4SLi Zefan 8166c282eb4SLi Zefan static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em) 8176c282eb4SLi Zefan { 8186c282eb4SLi Zefan struct extent_map *next; 8196c282eb4SLi Zefan bool ret = true; 8206c282eb4SLi Zefan 8216c282eb4SLi Zefan /* this is the last extent */ 8226c282eb4SLi Zefan if (em->start + em->len >= i_size_read(inode)) 8236c282eb4SLi Zefan return false; 8246c282eb4SLi Zefan 8256c282eb4SLi Zefan next = defrag_lookup_extent(inode, em->start + em->len); 8266c282eb4SLi Zefan if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE) 8276c282eb4SLi Zefan ret = false; 8286c282eb4SLi Zefan 8296c282eb4SLi Zefan free_extent_map(next); 8306c282eb4SLi Zefan return ret; 8316c282eb4SLi Zefan } 8326c282eb4SLi Zefan 8336c282eb4SLi Zefan static int should_defrag_range(struct inode *inode, u64 start, int thresh, 8346c282eb4SLi Zefan u64 *last_len, u64 *skip, u64 *defrag_end) 8356c282eb4SLi Zefan { 8366c282eb4SLi Zefan struct extent_map *em; 8376c282eb4SLi Zefan int ret = 1; 8386c282eb4SLi Zefan bool next_mergeable = true; 8396c282eb4SLi Zefan 8406c282eb4SLi Zefan /* 8416c282eb4SLi Zefan * make sure that once we start defragging an extent, we keep on 8426c282eb4SLi Zefan * defragging it 8436c282eb4SLi Zefan */ 8446c282eb4SLi Zefan if (start < *defrag_end) 8456c282eb4SLi Zefan return 1; 8466c282eb4SLi Zefan 8476c282eb4SLi Zefan *skip = 0; 8486c282eb4SLi Zefan 8496c282eb4SLi Zefan em = defrag_lookup_extent(inode, start); 8506c282eb4SLi Zefan if (!em) 8516c282eb4SLi Zefan return 0; 8526c282eb4SLi Zefan 853940100a4SChris Mason /* this will cover holes, and inline extents */ 85417ce6ef8SLiu Bo if (em->block_start >= EXTENT_MAP_LAST_BYTE) { 855940100a4SChris Mason ret = 0; 85617ce6ef8SLiu Bo goto out; 85717ce6ef8SLiu Bo } 85817ce6ef8SLiu Bo 8596c282eb4SLi Zefan next_mergeable = defrag_check_next_extent(inode, em); 860940100a4SChris Mason 861940100a4SChris Mason /* 8626c282eb4SLi Zefan * we hit a real extent, if it is big or the next extent is not a 8636c282eb4SLi Zefan * real extent, don't bother defragging it 864940100a4SChris Mason */ 8656c282eb4SLi Zefan if ((*last_len == 0 || *last_len >= thresh) && 8666c282eb4SLi Zefan (em->len >= thresh || !next_mergeable)) 867940100a4SChris Mason ret = 0; 86817ce6ef8SLiu Bo out: 869940100a4SChris Mason /* 870940100a4SChris Mason * last_len ends up being a counter of how many bytes we've defragged. 871940100a4SChris Mason * every time we choose not to defrag an extent, we reset *last_len 872940100a4SChris Mason * so that the next tiny extent will force a defrag. 873940100a4SChris Mason * 874940100a4SChris Mason * The end result of this is that tiny extents before a single big 875940100a4SChris Mason * extent will force at least part of that big extent to be defragged. 876940100a4SChris Mason */ 877940100a4SChris Mason if (ret) { 878940100a4SChris Mason *defrag_end = extent_map_end(em); 879940100a4SChris Mason } else { 880940100a4SChris Mason *last_len = 0; 881940100a4SChris Mason *skip = extent_map_end(em); 882940100a4SChris Mason *defrag_end = 0; 883940100a4SChris Mason } 884940100a4SChris Mason 885940100a4SChris Mason free_extent_map(em); 886940100a4SChris Mason return ret; 887940100a4SChris Mason } 888940100a4SChris Mason 8894cb5300bSChris Mason /* 8904cb5300bSChris Mason * it doesn't do much good to defrag one or two pages 8914cb5300bSChris Mason * at a time. This pulls in a nice chunk of pages 8924cb5300bSChris Mason * to COW and defrag. 8934cb5300bSChris Mason * 8944cb5300bSChris Mason * It also makes sure the delalloc code has enough 8954cb5300bSChris Mason * dirty data to avoid making new small extents as part 8964cb5300bSChris Mason * of the defrag 8974cb5300bSChris Mason * 8984cb5300bSChris Mason * It's a good idea to start RA on this range 8994cb5300bSChris Mason * before calling this. 9004cb5300bSChris Mason */ 9014cb5300bSChris Mason static int cluster_pages_for_defrag(struct inode *inode, 9024cb5300bSChris Mason struct page **pages, 9034cb5300bSChris Mason unsigned long start_index, 9044cb5300bSChris Mason int num_pages) 905f46b5a66SChristoph Hellwig { 9064cb5300bSChris Mason unsigned long file_end; 9074cb5300bSChris Mason u64 isize = i_size_read(inode); 908f46b5a66SChristoph Hellwig u64 page_start; 909f46b5a66SChristoph Hellwig u64 page_end; 9101f12bd06SLiu Bo u64 page_cnt; 9114cb5300bSChris Mason int ret; 9124cb5300bSChris Mason int i; 9134cb5300bSChris Mason int i_done; 9144cb5300bSChris Mason struct btrfs_ordered_extent *ordered; 9154cb5300bSChris Mason struct extent_state *cached_state = NULL; 916600a45e1SMiao Xie struct extent_io_tree *tree; 9173b16a4e3SJosef Bacik gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping); 9184cb5300bSChris Mason 9194cb5300bSChris Mason file_end = (isize - 1) >> PAGE_CACHE_SHIFT; 9201f12bd06SLiu Bo if (!isize || start_index > file_end) 9211f12bd06SLiu Bo return 0; 9221f12bd06SLiu Bo 9231f12bd06SLiu Bo page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1); 9244cb5300bSChris Mason 9254cb5300bSChris Mason ret = btrfs_delalloc_reserve_space(inode, 9261f12bd06SLiu Bo page_cnt << PAGE_CACHE_SHIFT); 9274cb5300bSChris Mason if (ret) 9284cb5300bSChris Mason return ret; 9294cb5300bSChris Mason i_done = 0; 930600a45e1SMiao Xie tree = &BTRFS_I(inode)->io_tree; 9314cb5300bSChris Mason 9324cb5300bSChris Mason /* step one, lock all the pages */ 9331f12bd06SLiu Bo for (i = 0; i < page_cnt; i++) { 9344cb5300bSChris Mason struct page *page; 935600a45e1SMiao Xie again: 936a94733d0SJosef Bacik page = find_or_create_page(inode->i_mapping, 9373b16a4e3SJosef Bacik start_index + i, mask); 9384cb5300bSChris Mason if (!page) 9394cb5300bSChris Mason break; 9404cb5300bSChris Mason 941600a45e1SMiao Xie page_start = page_offset(page); 942600a45e1SMiao Xie page_end = page_start + PAGE_CACHE_SIZE - 1; 943600a45e1SMiao Xie while (1) { 944d0082371SJeff Mahoney lock_extent(tree, page_start, page_end); 945600a45e1SMiao Xie ordered = btrfs_lookup_ordered_extent(inode, 946600a45e1SMiao Xie page_start); 947d0082371SJeff Mahoney unlock_extent(tree, page_start, page_end); 948600a45e1SMiao Xie if (!ordered) 949600a45e1SMiao Xie break; 950600a45e1SMiao Xie 951600a45e1SMiao Xie unlock_page(page); 952600a45e1SMiao Xie btrfs_start_ordered_extent(inode, ordered, 1); 953600a45e1SMiao Xie btrfs_put_ordered_extent(ordered); 954600a45e1SMiao Xie lock_page(page); 9551f12bd06SLiu Bo /* 9561f12bd06SLiu Bo * we unlocked the page above, so we need check if 9571f12bd06SLiu Bo * it was released or not. 9581f12bd06SLiu Bo */ 9591f12bd06SLiu Bo if (page->mapping != inode->i_mapping) { 9601f12bd06SLiu Bo unlock_page(page); 9611f12bd06SLiu Bo page_cache_release(page); 9621f12bd06SLiu Bo goto again; 9631f12bd06SLiu Bo } 964600a45e1SMiao Xie } 965600a45e1SMiao Xie 9664cb5300bSChris Mason if (!PageUptodate(page)) { 9674cb5300bSChris Mason btrfs_readpage(NULL, page); 9684cb5300bSChris Mason lock_page(page); 9694cb5300bSChris Mason if (!PageUptodate(page)) { 9704cb5300bSChris Mason unlock_page(page); 9714cb5300bSChris Mason page_cache_release(page); 9724cb5300bSChris Mason ret = -EIO; 9734cb5300bSChris Mason break; 9744cb5300bSChris Mason } 9754cb5300bSChris Mason } 976600a45e1SMiao Xie 977600a45e1SMiao Xie if (page->mapping != inode->i_mapping) { 978600a45e1SMiao Xie unlock_page(page); 979600a45e1SMiao Xie page_cache_release(page); 980600a45e1SMiao Xie goto again; 981600a45e1SMiao Xie } 982600a45e1SMiao Xie 9834cb5300bSChris Mason pages[i] = page; 9844cb5300bSChris Mason i_done++; 9854cb5300bSChris Mason } 9864cb5300bSChris Mason if (!i_done || ret) 9874cb5300bSChris Mason goto out; 9884cb5300bSChris Mason 9894cb5300bSChris Mason if (!(inode->i_sb->s_flags & MS_ACTIVE)) 9904cb5300bSChris Mason goto out; 9914cb5300bSChris Mason 9924cb5300bSChris Mason /* 9934cb5300bSChris Mason * so now we have a nice long stream of locked 9944cb5300bSChris Mason * and up to date pages, lets wait on them 9954cb5300bSChris Mason */ 9964cb5300bSChris Mason for (i = 0; i < i_done; i++) 9974cb5300bSChris Mason wait_on_page_writeback(pages[i]); 9984cb5300bSChris Mason 9994cb5300bSChris Mason page_start = page_offset(pages[0]); 10004cb5300bSChris Mason page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE; 10014cb5300bSChris Mason 10024cb5300bSChris Mason lock_extent_bits(&BTRFS_I(inode)->io_tree, 1003d0082371SJeff Mahoney page_start, page_end - 1, 0, &cached_state); 10044cb5300bSChris Mason clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, 10054cb5300bSChris Mason page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC | 10064cb5300bSChris Mason EXTENT_DO_ACCOUNTING, 0, 0, &cached_state, 10074cb5300bSChris Mason GFP_NOFS); 10084cb5300bSChris Mason 10091f12bd06SLiu Bo if (i_done != page_cnt) { 10109e0baf60SJosef Bacik spin_lock(&BTRFS_I(inode)->lock); 10119e0baf60SJosef Bacik BTRFS_I(inode)->outstanding_extents++; 10129e0baf60SJosef Bacik spin_unlock(&BTRFS_I(inode)->lock); 10134cb5300bSChris Mason btrfs_delalloc_release_space(inode, 10141f12bd06SLiu Bo (page_cnt - i_done) << PAGE_CACHE_SHIFT); 10154cb5300bSChris Mason } 10164cb5300bSChris Mason 10174cb5300bSChris Mason 10184cb5300bSChris Mason btrfs_set_extent_delalloc(inode, page_start, page_end - 1, 10194cb5300bSChris Mason &cached_state); 10204cb5300bSChris Mason 10214cb5300bSChris Mason unlock_extent_cached(&BTRFS_I(inode)->io_tree, 10224cb5300bSChris Mason page_start, page_end - 1, &cached_state, 10234cb5300bSChris Mason GFP_NOFS); 10244cb5300bSChris Mason 10254cb5300bSChris Mason for (i = 0; i < i_done; i++) { 10264cb5300bSChris Mason clear_page_dirty_for_io(pages[i]); 10274cb5300bSChris Mason ClearPageChecked(pages[i]); 10284cb5300bSChris Mason set_page_extent_mapped(pages[i]); 10294cb5300bSChris Mason set_page_dirty(pages[i]); 10304cb5300bSChris Mason unlock_page(pages[i]); 10314cb5300bSChris Mason page_cache_release(pages[i]); 10324cb5300bSChris Mason } 10334cb5300bSChris Mason return i_done; 10344cb5300bSChris Mason out: 10354cb5300bSChris Mason for (i = 0; i < i_done; i++) { 10364cb5300bSChris Mason unlock_page(pages[i]); 10374cb5300bSChris Mason page_cache_release(pages[i]); 10384cb5300bSChris Mason } 10391f12bd06SLiu Bo btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT); 10404cb5300bSChris Mason return ret; 10414cb5300bSChris Mason 10424cb5300bSChris Mason } 10434cb5300bSChris Mason 10444cb5300bSChris Mason int btrfs_defrag_file(struct inode *inode, struct file *file, 10454cb5300bSChris Mason struct btrfs_ioctl_defrag_range_args *range, 10464cb5300bSChris Mason u64 newer_than, unsigned long max_to_defrag) 10474cb5300bSChris Mason { 10484cb5300bSChris Mason struct btrfs_root *root = BTRFS_I(inode)->root; 10494cb5300bSChris Mason struct btrfs_super_block *disk_super; 10504cb5300bSChris Mason struct file_ra_state *ra = NULL; 10514cb5300bSChris Mason unsigned long last_index; 1052151a31b2SLi Zefan u64 isize = i_size_read(inode); 10534cb5300bSChris Mason u64 features; 1054940100a4SChris Mason u64 last_len = 0; 1055940100a4SChris Mason u64 skip = 0; 1056940100a4SChris Mason u64 defrag_end = 0; 10574cb5300bSChris Mason u64 newer_off = range->start; 1058f46b5a66SChristoph Hellwig unsigned long i; 1059008873eaSLi Zefan unsigned long ra_index = 0; 1060f46b5a66SChristoph Hellwig int ret; 10614cb5300bSChris Mason int defrag_count = 0; 10621a419d85SLi Zefan int compress_type = BTRFS_COMPRESS_ZLIB; 10634cb5300bSChris Mason int extent_thresh = range->extent_thresh; 1064008873eaSLi Zefan int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT; 1065008873eaSLi Zefan int cluster = max_cluster; 10664cb5300bSChris Mason u64 new_align = ~((u64)128 * 1024 - 1); 10674cb5300bSChris Mason struct page **pages = NULL; 10684cb5300bSChris Mason 10694cb5300bSChris Mason if (extent_thresh == 0) 10704cb5300bSChris Mason extent_thresh = 256 * 1024; 10711a419d85SLi Zefan 10721a419d85SLi Zefan if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) { 10731a419d85SLi Zefan if (range->compress_type > BTRFS_COMPRESS_TYPES) 10741a419d85SLi Zefan return -EINVAL; 10751a419d85SLi Zefan if (range->compress_type) 10761a419d85SLi Zefan compress_type = range->compress_type; 10771a419d85SLi Zefan } 1078f46b5a66SChristoph Hellwig 1079151a31b2SLi Zefan if (isize == 0) 1080940100a4SChris Mason return 0; 1081f46b5a66SChristoph Hellwig 10824cb5300bSChris Mason /* 10834cb5300bSChris Mason * if we were not given a file, allocate a readahead 10844cb5300bSChris Mason * context 10854cb5300bSChris Mason */ 10864cb5300bSChris Mason if (!file) { 10874cb5300bSChris Mason ra = kzalloc(sizeof(*ra), GFP_NOFS); 10884cb5300bSChris Mason if (!ra) 10894cb5300bSChris Mason return -ENOMEM; 10904cb5300bSChris Mason file_ra_state_init(ra, inode->i_mapping); 10914cb5300bSChris Mason } else { 10924cb5300bSChris Mason ra = &file->f_ra; 10934cb5300bSChris Mason } 10944cb5300bSChris Mason 1095008873eaSLi Zefan pages = kmalloc(sizeof(struct page *) * max_cluster, 10964cb5300bSChris Mason GFP_NOFS); 10974cb5300bSChris Mason if (!pages) { 10984cb5300bSChris Mason ret = -ENOMEM; 10994cb5300bSChris Mason goto out_ra; 11004cb5300bSChris Mason } 11014cb5300bSChris Mason 11024cb5300bSChris Mason /* find the last page to defrag */ 11031e701a32SChris Mason if (range->start + range->len > range->start) { 1104151a31b2SLi Zefan last_index = min_t(u64, isize - 1, 11051e701a32SChris Mason range->start + range->len - 1) >> PAGE_CACHE_SHIFT; 11061e701a32SChris Mason } else { 1107151a31b2SLi Zefan last_index = (isize - 1) >> PAGE_CACHE_SHIFT; 11081e701a32SChris Mason } 11091e701a32SChris Mason 11104cb5300bSChris Mason if (newer_than) { 11114cb5300bSChris Mason ret = find_new_extents(root, inode, newer_than, 11124cb5300bSChris Mason &newer_off, 64 * 1024); 11134cb5300bSChris Mason if (!ret) { 11144cb5300bSChris Mason range->start = newer_off; 11154cb5300bSChris Mason /* 11164cb5300bSChris Mason * we always align our defrag to help keep 11174cb5300bSChris Mason * the extents in the file evenly spaced 11184cb5300bSChris Mason */ 11194cb5300bSChris Mason i = (newer_off & new_align) >> PAGE_CACHE_SHIFT; 11204cb5300bSChris Mason } else 11214cb5300bSChris Mason goto out_ra; 11224cb5300bSChris Mason } else { 11231e701a32SChris Mason i = range->start >> PAGE_CACHE_SHIFT; 11244cb5300bSChris Mason } 11254cb5300bSChris Mason if (!max_to_defrag) 11267ec31b54SLiu Bo max_to_defrag = last_index + 1; 11274cb5300bSChris Mason 11282a0f7f57SLi Zefan /* 11292a0f7f57SLi Zefan * make writeback starts from i, so the defrag range can be 11302a0f7f57SLi Zefan * written sequentially. 11312a0f7f57SLi Zefan */ 11322a0f7f57SLi Zefan if (i < inode->i_mapping->writeback_index) 11332a0f7f57SLi Zefan inode->i_mapping->writeback_index = i; 11342a0f7f57SLi Zefan 1135f7f43cc8SChris Mason while (i <= last_index && defrag_count < max_to_defrag && 1136f7f43cc8SChris Mason (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >> 1137f7f43cc8SChris Mason PAGE_CACHE_SHIFT)) { 11384cb5300bSChris Mason /* 11394cb5300bSChris Mason * make sure we stop running if someone unmounts 11404cb5300bSChris Mason * the FS 11414cb5300bSChris Mason */ 11424cb5300bSChris Mason if (!(inode->i_sb->s_flags & MS_ACTIVE)) 11434cb5300bSChris Mason break; 11444cb5300bSChris Mason 114566c26892SLiu Bo if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT, 11466c282eb4SLi Zefan extent_thresh, &last_len, &skip, 11476c282eb4SLi Zefan &defrag_end)) { 1148940100a4SChris Mason unsigned long next; 1149940100a4SChris Mason /* 1150940100a4SChris Mason * the should_defrag function tells us how much to skip 1151940100a4SChris Mason * bump our counter by the suggested amount 1152940100a4SChris Mason */ 1153940100a4SChris Mason next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT; 1154940100a4SChris Mason i = max(i + 1, next); 1155940100a4SChris Mason continue; 1156940100a4SChris Mason } 1157008873eaSLi Zefan 1158008873eaSLi Zefan if (!newer_than) { 1159008873eaSLi Zefan cluster = (PAGE_CACHE_ALIGN(defrag_end) >> 1160008873eaSLi Zefan PAGE_CACHE_SHIFT) - i; 1161008873eaSLi Zefan cluster = min(cluster, max_cluster); 1162008873eaSLi Zefan } else { 1163008873eaSLi Zefan cluster = max_cluster; 1164008873eaSLi Zefan } 1165008873eaSLi Zefan 11661e701a32SChris Mason if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) 11671a419d85SLi Zefan BTRFS_I(inode)->force_compress = compress_type; 1168940100a4SChris Mason 1169008873eaSLi Zefan if (i + cluster > ra_index) { 1170008873eaSLi Zefan ra_index = max(i, ra_index); 1171008873eaSLi Zefan btrfs_force_ra(inode->i_mapping, ra, file, ra_index, 1172008873eaSLi Zefan cluster); 1173008873eaSLi Zefan ra_index += max_cluster; 1174008873eaSLi Zefan } 11754cb5300bSChris Mason 1176ecb8bea8SLiu Bo mutex_lock(&inode->i_mutex); 1177008873eaSLi Zefan ret = cluster_pages_for_defrag(inode, pages, i, cluster); 1178ecb8bea8SLiu Bo if (ret < 0) { 1179ecb8bea8SLiu Bo mutex_unlock(&inode->i_mutex); 11804cb5300bSChris Mason goto out_ra; 1181ecb8bea8SLiu Bo } 11824cb5300bSChris Mason 11834cb5300bSChris Mason defrag_count += ret; 11844cb5300bSChris Mason balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret); 1185ecb8bea8SLiu Bo mutex_unlock(&inode->i_mutex); 11864cb5300bSChris Mason 11874cb5300bSChris Mason if (newer_than) { 11884cb5300bSChris Mason if (newer_off == (u64)-1) 11894cb5300bSChris Mason break; 11904cb5300bSChris Mason 1191e1f041e1SLiu Bo if (ret > 0) 1192e1f041e1SLiu Bo i += ret; 1193e1f041e1SLiu Bo 11944cb5300bSChris Mason newer_off = max(newer_off + 1, 11954cb5300bSChris Mason (u64)i << PAGE_CACHE_SHIFT); 11964cb5300bSChris Mason 11974cb5300bSChris Mason ret = find_new_extents(root, inode, 11984cb5300bSChris Mason newer_than, &newer_off, 11994cb5300bSChris Mason 64 * 1024); 12004cb5300bSChris Mason if (!ret) { 12014cb5300bSChris Mason range->start = newer_off; 12024cb5300bSChris Mason i = (newer_off & new_align) >> PAGE_CACHE_SHIFT; 12034cb5300bSChris Mason } else { 12044cb5300bSChris Mason break; 1205940100a4SChris Mason } 12064cb5300bSChris Mason } else { 1207008873eaSLi Zefan if (ret > 0) { 1208cbcc8326SLi Zefan i += ret; 1209008873eaSLi Zefan last_len += ret << PAGE_CACHE_SHIFT; 1210008873eaSLi Zefan } else { 1211940100a4SChris Mason i++; 1212008873eaSLi Zefan last_len = 0; 1213008873eaSLi Zefan } 1214f46b5a66SChristoph Hellwig } 12154cb5300bSChris Mason } 1216f46b5a66SChristoph Hellwig 12171e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) 12181e701a32SChris Mason filemap_flush(inode->i_mapping); 12191e701a32SChris Mason 12201e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 12211e701a32SChris Mason /* the filemap_flush will queue IO into the worker threads, but 12221e701a32SChris Mason * we have to make sure the IO is actually started and that 12231e701a32SChris Mason * ordered extents get created before we return 12241e701a32SChris Mason */ 12251e701a32SChris Mason atomic_inc(&root->fs_info->async_submit_draining); 12261e701a32SChris Mason while (atomic_read(&root->fs_info->nr_async_submits) || 12271e701a32SChris Mason atomic_read(&root->fs_info->async_delalloc_pages)) { 12281e701a32SChris Mason wait_event(root->fs_info->async_submit_wait, 12291e701a32SChris Mason (atomic_read(&root->fs_info->nr_async_submits) == 0 && 12301e701a32SChris Mason atomic_read(&root->fs_info->async_delalloc_pages) == 0)); 12311e701a32SChris Mason } 12321e701a32SChris Mason atomic_dec(&root->fs_info->async_submit_draining); 12331e701a32SChris Mason 12341e701a32SChris Mason mutex_lock(&inode->i_mutex); 1235261507a0SLi Zefan BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE; 12361e701a32SChris Mason mutex_unlock(&inode->i_mutex); 12371e701a32SChris Mason } 12381e701a32SChris Mason 12396c41761fSDavid Sterba disk_super = root->fs_info->super_copy; 12401a419d85SLi Zefan features = btrfs_super_incompat_flags(disk_super); 12411a419d85SLi Zefan if (range->compress_type == BTRFS_COMPRESS_LZO) { 12421a419d85SLi Zefan features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; 12431a419d85SLi Zefan btrfs_set_super_incompat_flags(disk_super, features); 12441a419d85SLi Zefan } 12451a419d85SLi Zefan 124660ccf82fSDiego Calleja ret = defrag_count; 1247940100a4SChris Mason 12484cb5300bSChris Mason out_ra: 12494cb5300bSChris Mason if (!file) 12504cb5300bSChris Mason kfree(ra); 12514cb5300bSChris Mason kfree(pages); 1252940100a4SChris Mason return ret; 1253f46b5a66SChristoph Hellwig } 1254f46b5a66SChristoph Hellwig 125576dda93cSYan, Zheng static noinline int btrfs_ioctl_resize(struct btrfs_root *root, 125676dda93cSYan, Zheng void __user *arg) 1257f46b5a66SChristoph Hellwig { 1258f46b5a66SChristoph Hellwig u64 new_size; 1259f46b5a66SChristoph Hellwig u64 old_size; 1260f46b5a66SChristoph Hellwig u64 devid = 1; 1261f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 1262f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 1263f46b5a66SChristoph Hellwig struct btrfs_device *device = NULL; 1264f46b5a66SChristoph Hellwig char *sizestr; 1265f46b5a66SChristoph Hellwig char *devstr = NULL; 1266f46b5a66SChristoph Hellwig int ret = 0; 1267f46b5a66SChristoph Hellwig int mod = 0; 1268f46b5a66SChristoph Hellwig 1269c146afadSYan Zheng if (root->fs_info->sb->s_flags & MS_RDONLY) 1270c146afadSYan Zheng return -EROFS; 1271c146afadSYan Zheng 1272e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 1273e441d54dSChris Mason return -EPERM; 1274e441d54dSChris Mason 1275c9e9f97bSIlya Dryomov mutex_lock(&root->fs_info->volume_mutex); 1276c9e9f97bSIlya Dryomov if (root->fs_info->balance_ctl) { 1277c9e9f97bSIlya Dryomov printk(KERN_INFO "btrfs: balance in progress\n"); 1278c9e9f97bSIlya Dryomov ret = -EINVAL; 1279c9e9f97bSIlya Dryomov goto out; 1280c9e9f97bSIlya Dryomov } 1281c9e9f97bSIlya Dryomov 1282dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1283c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 1284c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 1285c9e9f97bSIlya Dryomov goto out; 1286c9e9f97bSIlya Dryomov } 12875516e595SMark Fasheh 12885516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1289f46b5a66SChristoph Hellwig 1290f46b5a66SChristoph Hellwig sizestr = vol_args->name; 1291f46b5a66SChristoph Hellwig devstr = strchr(sizestr, ':'); 1292f46b5a66SChristoph Hellwig if (devstr) { 1293f46b5a66SChristoph Hellwig char *end; 1294f46b5a66SChristoph Hellwig sizestr = devstr + 1; 1295f46b5a66SChristoph Hellwig *devstr = '\0'; 1296f46b5a66SChristoph Hellwig devstr = vol_args->name; 1297f46b5a66SChristoph Hellwig devid = simple_strtoull(devstr, &end, 10); 12985bb14682SArnd Hannemann printk(KERN_INFO "btrfs: resizing devid %llu\n", 129921380931SJoel Becker (unsigned long long)devid); 1300f46b5a66SChristoph Hellwig } 13012b82032cSYan Zheng device = btrfs_find_device(root, devid, NULL, NULL); 1302f46b5a66SChristoph Hellwig if (!device) { 13035bb14682SArnd Hannemann printk(KERN_INFO "btrfs: resizer unable to find device %llu\n", 130421380931SJoel Becker (unsigned long long)devid); 1305f46b5a66SChristoph Hellwig ret = -EINVAL; 1306c9e9f97bSIlya Dryomov goto out_free; 1307f46b5a66SChristoph Hellwig } 13084e42ae1bSLiu Bo if (device->fs_devices && device->fs_devices->seeding) { 13094e42ae1bSLiu Bo printk(KERN_INFO "btrfs: resizer unable to apply on " 1310a8c4a33bSChris Mason "seeding device %llu\n", 1311a8c4a33bSChris Mason (unsigned long long)devid); 13124e42ae1bSLiu Bo ret = -EINVAL; 13134e42ae1bSLiu Bo goto out_free; 13144e42ae1bSLiu Bo } 13154e42ae1bSLiu Bo 1316f46b5a66SChristoph Hellwig if (!strcmp(sizestr, "max")) 1317f46b5a66SChristoph Hellwig new_size = device->bdev->bd_inode->i_size; 1318f46b5a66SChristoph Hellwig else { 1319f46b5a66SChristoph Hellwig if (sizestr[0] == '-') { 1320f46b5a66SChristoph Hellwig mod = -1; 1321f46b5a66SChristoph Hellwig sizestr++; 1322f46b5a66SChristoph Hellwig } else if (sizestr[0] == '+') { 1323f46b5a66SChristoph Hellwig mod = 1; 1324f46b5a66SChristoph Hellwig sizestr++; 1325f46b5a66SChristoph Hellwig } 132691748467SAkinobu Mita new_size = memparse(sizestr, NULL); 1327f46b5a66SChristoph Hellwig if (new_size == 0) { 1328f46b5a66SChristoph Hellwig ret = -EINVAL; 1329c9e9f97bSIlya Dryomov goto out_free; 1330f46b5a66SChristoph Hellwig } 1331f46b5a66SChristoph Hellwig } 1332f46b5a66SChristoph Hellwig 1333f46b5a66SChristoph Hellwig old_size = device->total_bytes; 1334f46b5a66SChristoph Hellwig 1335f46b5a66SChristoph Hellwig if (mod < 0) { 1336f46b5a66SChristoph Hellwig if (new_size > old_size) { 1337f46b5a66SChristoph Hellwig ret = -EINVAL; 1338c9e9f97bSIlya Dryomov goto out_free; 1339f46b5a66SChristoph Hellwig } 1340f46b5a66SChristoph Hellwig new_size = old_size - new_size; 1341f46b5a66SChristoph Hellwig } else if (mod > 0) { 1342f46b5a66SChristoph Hellwig new_size = old_size + new_size; 1343f46b5a66SChristoph Hellwig } 1344f46b5a66SChristoph Hellwig 1345f46b5a66SChristoph Hellwig if (new_size < 256 * 1024 * 1024) { 1346f46b5a66SChristoph Hellwig ret = -EINVAL; 1347c9e9f97bSIlya Dryomov goto out_free; 1348f46b5a66SChristoph Hellwig } 1349f46b5a66SChristoph Hellwig if (new_size > device->bdev->bd_inode->i_size) { 1350f46b5a66SChristoph Hellwig ret = -EFBIG; 1351c9e9f97bSIlya Dryomov goto out_free; 1352f46b5a66SChristoph Hellwig } 1353f46b5a66SChristoph Hellwig 1354f46b5a66SChristoph Hellwig do_div(new_size, root->sectorsize); 1355f46b5a66SChristoph Hellwig new_size *= root->sectorsize; 1356f46b5a66SChristoph Hellwig 1357606686eeSJosef Bacik printk_in_rcu(KERN_INFO "btrfs: new size for %s is %llu\n", 1358606686eeSJosef Bacik rcu_str_deref(device->name), 1359606686eeSJosef Bacik (unsigned long long)new_size); 1360f46b5a66SChristoph Hellwig 1361f46b5a66SChristoph Hellwig if (new_size > old_size) { 1362a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 136398d5dc13STsutomu Itoh if (IS_ERR(trans)) { 136498d5dc13STsutomu Itoh ret = PTR_ERR(trans); 1365c9e9f97bSIlya Dryomov goto out_free; 136698d5dc13STsutomu Itoh } 1367f46b5a66SChristoph Hellwig ret = btrfs_grow_device(trans, device, new_size); 1368f46b5a66SChristoph Hellwig btrfs_commit_transaction(trans, root); 1369ece7d20eSMike Fleetwood } else if (new_size < old_size) { 1370f46b5a66SChristoph Hellwig ret = btrfs_shrink_device(device, new_size); 1371f46b5a66SChristoph Hellwig } 1372f46b5a66SChristoph Hellwig 1373c9e9f97bSIlya Dryomov out_free: 1374f46b5a66SChristoph Hellwig kfree(vol_args); 1375c9e9f97bSIlya Dryomov out: 1376c9e9f97bSIlya Dryomov mutex_unlock(&root->fs_info->volume_mutex); 1377f46b5a66SChristoph Hellwig return ret; 1378f46b5a66SChristoph Hellwig } 1379f46b5a66SChristoph Hellwig 138072fd032eSSage Weil static noinline int btrfs_ioctl_snap_create_transid(struct file *file, 138172fd032eSSage Weil char *name, 138272fd032eSSage Weil unsigned long fd, 138372fd032eSSage Weil int subvol, 1384b83cc969SLi Zefan u64 *transid, 1385b83cc969SLi Zefan bool readonly) 1386f46b5a66SChristoph Hellwig { 1387cb8e7090SChristoph Hellwig struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; 13883de4586cSChris Mason struct file *src_file; 1389f46b5a66SChristoph Hellwig int namelen; 13903de4586cSChris Mason int ret = 0; 1391f46b5a66SChristoph Hellwig 1392c146afadSYan Zheng if (root->fs_info->sb->s_flags & MS_RDONLY) 1393c146afadSYan Zheng return -EROFS; 1394c146afadSYan Zheng 139572fd032eSSage Weil namelen = strlen(name); 139672fd032eSSage Weil if (strchr(name, '/')) { 1397f46b5a66SChristoph Hellwig ret = -EINVAL; 1398f46b5a66SChristoph Hellwig goto out; 1399f46b5a66SChristoph Hellwig } 1400f46b5a66SChristoph Hellwig 140116780cabSChris Mason if (name[0] == '.' && 140216780cabSChris Mason (namelen == 1 || (name[1] == '.' && namelen == 2))) { 140316780cabSChris Mason ret = -EEXIST; 140416780cabSChris Mason goto out; 140516780cabSChris Mason } 140616780cabSChris Mason 14073de4586cSChris Mason if (subvol) { 140872fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 1409b83cc969SLi Zefan NULL, transid, readonly); 1410cb8e7090SChristoph Hellwig } else { 14113de4586cSChris Mason struct inode *src_inode; 141272fd032eSSage Weil src_file = fget(fd); 14133de4586cSChris Mason if (!src_file) { 14143de4586cSChris Mason ret = -EINVAL; 14153de4586cSChris Mason goto out; 14163de4586cSChris Mason } 14173de4586cSChris Mason 14183de4586cSChris Mason src_inode = src_file->f_path.dentry->d_inode; 14193de4586cSChris Mason if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) { 1420d397712bSChris Mason printk(KERN_INFO "btrfs: Snapshot src from " 1421d397712bSChris Mason "another FS\n"); 14223de4586cSChris Mason ret = -EINVAL; 14233de4586cSChris Mason fput(src_file); 14243de4586cSChris Mason goto out; 14253de4586cSChris Mason } 142672fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 142772fd032eSSage Weil BTRFS_I(src_inode)->root, 1428b83cc969SLi Zefan transid, readonly); 14293de4586cSChris Mason fput(src_file); 1430cb8e7090SChristoph Hellwig } 1431f46b5a66SChristoph Hellwig out: 143272fd032eSSage Weil return ret; 143372fd032eSSage Weil } 143472fd032eSSage Weil 143572fd032eSSage Weil static noinline int btrfs_ioctl_snap_create(struct file *file, 1436fa0d2b9bSLi Zefan void __user *arg, int subvol) 143772fd032eSSage Weil { 1438fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args *vol_args; 143972fd032eSSage Weil int ret; 144072fd032eSSage Weil 1441fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1442fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1443fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1444fa0d2b9bSLi Zefan vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1445fa0d2b9bSLi Zefan 1446fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 1447b83cc969SLi Zefan vol_args->fd, subvol, 1448b83cc969SLi Zefan NULL, false); 1449fa0d2b9bSLi Zefan 1450fa0d2b9bSLi Zefan kfree(vol_args); 1451fa0d2b9bSLi Zefan return ret; 1452fa0d2b9bSLi Zefan } 1453fa0d2b9bSLi Zefan 1454fa0d2b9bSLi Zefan static noinline int btrfs_ioctl_snap_create_v2(struct file *file, 1455fa0d2b9bSLi Zefan void __user *arg, int subvol) 1456fa0d2b9bSLi Zefan { 1457fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args_v2 *vol_args; 1458fa0d2b9bSLi Zefan int ret; 1459fdfb1e4fSLi Zefan u64 transid = 0; 1460fdfb1e4fSLi Zefan u64 *ptr = NULL; 1461b83cc969SLi Zefan bool readonly = false; 146272fd032eSSage Weil 1463fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1464fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1465fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1466fa0d2b9bSLi Zefan vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 1467fdfb1e4fSLi Zefan 1468b83cc969SLi Zefan if (vol_args->flags & 1469b83cc969SLi Zefan ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) { 1470b83cc969SLi Zefan ret = -EOPNOTSUPP; 1471fdfb1e4fSLi Zefan goto out; 1472fdfb1e4fSLi Zefan } 1473fdfb1e4fSLi Zefan 1474fa0d2b9bSLi Zefan if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC) 1475fdfb1e4fSLi Zefan ptr = &transid; 1476b83cc969SLi Zefan if (vol_args->flags & BTRFS_SUBVOL_RDONLY) 1477b83cc969SLi Zefan readonly = true; 147875eaa0e2SSage Weil 1479fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 1480b83cc969SLi Zefan vol_args->fd, subvol, 1481b83cc969SLi Zefan ptr, readonly); 148275eaa0e2SSage Weil 1483fdfb1e4fSLi Zefan if (ret == 0 && ptr && 148475eaa0e2SSage Weil copy_to_user(arg + 1485fdfb1e4fSLi Zefan offsetof(struct btrfs_ioctl_vol_args_v2, 1486fdfb1e4fSLi Zefan transid), ptr, sizeof(*ptr))) 148775eaa0e2SSage Weil ret = -EFAULT; 1488fdfb1e4fSLi Zefan out: 1489f46b5a66SChristoph Hellwig kfree(vol_args); 1490f46b5a66SChristoph Hellwig return ret; 1491f46b5a66SChristoph Hellwig } 1492f46b5a66SChristoph Hellwig 14930caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_getflags(struct file *file, 14940caa102dSLi Zefan void __user *arg) 14950caa102dSLi Zefan { 14960caa102dSLi Zefan struct inode *inode = fdentry(file)->d_inode; 14970caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 14980caa102dSLi Zefan int ret = 0; 14990caa102dSLi Zefan u64 flags = 0; 15000caa102dSLi Zefan 150133345d01SLi Zefan if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) 15020caa102dSLi Zefan return -EINVAL; 15030caa102dSLi Zefan 15040caa102dSLi Zefan down_read(&root->fs_info->subvol_sem); 15050caa102dSLi Zefan if (btrfs_root_readonly(root)) 15060caa102dSLi Zefan flags |= BTRFS_SUBVOL_RDONLY; 15070caa102dSLi Zefan up_read(&root->fs_info->subvol_sem); 15080caa102dSLi Zefan 15090caa102dSLi Zefan if (copy_to_user(arg, &flags, sizeof(flags))) 15100caa102dSLi Zefan ret = -EFAULT; 15110caa102dSLi Zefan 15120caa102dSLi Zefan return ret; 15130caa102dSLi Zefan } 15140caa102dSLi Zefan 15150caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_setflags(struct file *file, 15160caa102dSLi Zefan void __user *arg) 15170caa102dSLi Zefan { 15180caa102dSLi Zefan struct inode *inode = fdentry(file)->d_inode; 15190caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 15200caa102dSLi Zefan struct btrfs_trans_handle *trans; 15210caa102dSLi Zefan u64 root_flags; 15220caa102dSLi Zefan u64 flags; 15230caa102dSLi Zefan int ret = 0; 15240caa102dSLi Zefan 15250caa102dSLi Zefan if (root->fs_info->sb->s_flags & MS_RDONLY) 15260caa102dSLi Zefan return -EROFS; 15270caa102dSLi Zefan 152833345d01SLi Zefan if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) 15290caa102dSLi Zefan return -EINVAL; 15300caa102dSLi Zefan 15310caa102dSLi Zefan if (copy_from_user(&flags, arg, sizeof(flags))) 15320caa102dSLi Zefan return -EFAULT; 15330caa102dSLi Zefan 1534b4dc2b8cSLi Zefan if (flags & BTRFS_SUBVOL_CREATE_ASYNC) 15350caa102dSLi Zefan return -EINVAL; 15360caa102dSLi Zefan 15370caa102dSLi Zefan if (flags & ~BTRFS_SUBVOL_RDONLY) 15380caa102dSLi Zefan return -EOPNOTSUPP; 15390caa102dSLi Zefan 15402e149670SSerge E. Hallyn if (!inode_owner_or_capable(inode)) 1541b4dc2b8cSLi Zefan return -EACCES; 1542b4dc2b8cSLi Zefan 15430caa102dSLi Zefan down_write(&root->fs_info->subvol_sem); 15440caa102dSLi Zefan 15450caa102dSLi Zefan /* nothing to do */ 15460caa102dSLi Zefan if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) 15470caa102dSLi Zefan goto out; 15480caa102dSLi Zefan 15490caa102dSLi Zefan root_flags = btrfs_root_flags(&root->root_item); 15500caa102dSLi Zefan if (flags & BTRFS_SUBVOL_RDONLY) 15510caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 15520caa102dSLi Zefan root_flags | BTRFS_ROOT_SUBVOL_RDONLY); 15530caa102dSLi Zefan else 15540caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 15550caa102dSLi Zefan root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); 15560caa102dSLi Zefan 15570caa102dSLi Zefan trans = btrfs_start_transaction(root, 1); 15580caa102dSLi Zefan if (IS_ERR(trans)) { 15590caa102dSLi Zefan ret = PTR_ERR(trans); 15600caa102dSLi Zefan goto out_reset; 15610caa102dSLi Zefan } 15620caa102dSLi Zefan 1563b4dc2b8cSLi Zefan ret = btrfs_update_root(trans, root->fs_info->tree_root, 15640caa102dSLi Zefan &root->root_key, &root->root_item); 15650caa102dSLi Zefan 15660caa102dSLi Zefan btrfs_commit_transaction(trans, root); 15670caa102dSLi Zefan out_reset: 15680caa102dSLi Zefan if (ret) 15690caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, root_flags); 15700caa102dSLi Zefan out: 15710caa102dSLi Zefan up_write(&root->fs_info->subvol_sem); 15720caa102dSLi Zefan return ret; 15730caa102dSLi Zefan } 15740caa102dSLi Zefan 157576dda93cSYan, Zheng /* 157676dda93cSYan, Zheng * helper to check if the subvolume references other subvolumes 157776dda93cSYan, Zheng */ 157876dda93cSYan, Zheng static noinline int may_destroy_subvol(struct btrfs_root *root) 157976dda93cSYan, Zheng { 158076dda93cSYan, Zheng struct btrfs_path *path; 158176dda93cSYan, Zheng struct btrfs_key key; 158276dda93cSYan, Zheng int ret; 158376dda93cSYan, Zheng 158476dda93cSYan, Zheng path = btrfs_alloc_path(); 158576dda93cSYan, Zheng if (!path) 158676dda93cSYan, Zheng return -ENOMEM; 158776dda93cSYan, Zheng 158876dda93cSYan, Zheng key.objectid = root->root_key.objectid; 158976dda93cSYan, Zheng key.type = BTRFS_ROOT_REF_KEY; 159076dda93cSYan, Zheng key.offset = (u64)-1; 159176dda93cSYan, Zheng 159276dda93cSYan, Zheng ret = btrfs_search_slot(NULL, root->fs_info->tree_root, 159376dda93cSYan, Zheng &key, path, 0, 0); 159476dda93cSYan, Zheng if (ret < 0) 159576dda93cSYan, Zheng goto out; 159676dda93cSYan, Zheng BUG_ON(ret == 0); 159776dda93cSYan, Zheng 159876dda93cSYan, Zheng ret = 0; 159976dda93cSYan, Zheng if (path->slots[0] > 0) { 160076dda93cSYan, Zheng path->slots[0]--; 160176dda93cSYan, Zheng btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 160276dda93cSYan, Zheng if (key.objectid == root->root_key.objectid && 160376dda93cSYan, Zheng key.type == BTRFS_ROOT_REF_KEY) 160476dda93cSYan, Zheng ret = -ENOTEMPTY; 160576dda93cSYan, Zheng } 160676dda93cSYan, Zheng out: 160776dda93cSYan, Zheng btrfs_free_path(path); 160876dda93cSYan, Zheng return ret; 160976dda93cSYan, Zheng } 161076dda93cSYan, Zheng 1611ac8e9819SChris Mason static noinline int key_in_sk(struct btrfs_key *key, 1612ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk) 1613ac8e9819SChris Mason { 1614abc6e134SChris Mason struct btrfs_key test; 1615abc6e134SChris Mason int ret; 1616abc6e134SChris Mason 1617abc6e134SChris Mason test.objectid = sk->min_objectid; 1618abc6e134SChris Mason test.type = sk->min_type; 1619abc6e134SChris Mason test.offset = sk->min_offset; 1620abc6e134SChris Mason 1621abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1622abc6e134SChris Mason if (ret < 0) 1623ac8e9819SChris Mason return 0; 1624abc6e134SChris Mason 1625abc6e134SChris Mason test.objectid = sk->max_objectid; 1626abc6e134SChris Mason test.type = sk->max_type; 1627abc6e134SChris Mason test.offset = sk->max_offset; 1628abc6e134SChris Mason 1629abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1630abc6e134SChris Mason if (ret > 0) 1631ac8e9819SChris Mason return 0; 1632ac8e9819SChris Mason return 1; 1633ac8e9819SChris Mason } 1634ac8e9819SChris Mason 1635ac8e9819SChris Mason static noinline int copy_to_sk(struct btrfs_root *root, 1636ac8e9819SChris Mason struct btrfs_path *path, 1637ac8e9819SChris Mason struct btrfs_key *key, 1638ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk, 1639ac8e9819SChris Mason char *buf, 1640ac8e9819SChris Mason unsigned long *sk_offset, 1641ac8e9819SChris Mason int *num_found) 1642ac8e9819SChris Mason { 1643ac8e9819SChris Mason u64 found_transid; 1644ac8e9819SChris Mason struct extent_buffer *leaf; 1645ac8e9819SChris Mason struct btrfs_ioctl_search_header sh; 1646ac8e9819SChris Mason unsigned long item_off; 1647ac8e9819SChris Mason unsigned long item_len; 1648ac8e9819SChris Mason int nritems; 1649ac8e9819SChris Mason int i; 1650ac8e9819SChris Mason int slot; 1651ac8e9819SChris Mason int ret = 0; 1652ac8e9819SChris Mason 1653ac8e9819SChris Mason leaf = path->nodes[0]; 1654ac8e9819SChris Mason slot = path->slots[0]; 1655ac8e9819SChris Mason nritems = btrfs_header_nritems(leaf); 1656ac8e9819SChris Mason 1657ac8e9819SChris Mason if (btrfs_header_generation(leaf) > sk->max_transid) { 1658ac8e9819SChris Mason i = nritems; 1659ac8e9819SChris Mason goto advance_key; 1660ac8e9819SChris Mason } 1661ac8e9819SChris Mason found_transid = btrfs_header_generation(leaf); 1662ac8e9819SChris Mason 1663ac8e9819SChris Mason for (i = slot; i < nritems; i++) { 1664ac8e9819SChris Mason item_off = btrfs_item_ptr_offset(leaf, i); 1665ac8e9819SChris Mason item_len = btrfs_item_size_nr(leaf, i); 1666ac8e9819SChris Mason 1667ac8e9819SChris Mason if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE) 1668ac8e9819SChris Mason item_len = 0; 1669ac8e9819SChris Mason 1670ac8e9819SChris Mason if (sizeof(sh) + item_len + *sk_offset > 1671ac8e9819SChris Mason BTRFS_SEARCH_ARGS_BUFSIZE) { 1672ac8e9819SChris Mason ret = 1; 1673ac8e9819SChris Mason goto overflow; 1674ac8e9819SChris Mason } 1675ac8e9819SChris Mason 1676ac8e9819SChris Mason btrfs_item_key_to_cpu(leaf, key, i); 1677ac8e9819SChris Mason if (!key_in_sk(key, sk)) 1678ac8e9819SChris Mason continue; 1679ac8e9819SChris Mason 1680ac8e9819SChris Mason sh.objectid = key->objectid; 1681ac8e9819SChris Mason sh.offset = key->offset; 1682ac8e9819SChris Mason sh.type = key->type; 1683ac8e9819SChris Mason sh.len = item_len; 1684ac8e9819SChris Mason sh.transid = found_transid; 1685ac8e9819SChris Mason 1686ac8e9819SChris Mason /* copy search result header */ 1687ac8e9819SChris Mason memcpy(buf + *sk_offset, &sh, sizeof(sh)); 1688ac8e9819SChris Mason *sk_offset += sizeof(sh); 1689ac8e9819SChris Mason 1690ac8e9819SChris Mason if (item_len) { 1691ac8e9819SChris Mason char *p = buf + *sk_offset; 1692ac8e9819SChris Mason /* copy the item */ 1693ac8e9819SChris Mason read_extent_buffer(leaf, p, 1694ac8e9819SChris Mason item_off, item_len); 1695ac8e9819SChris Mason *sk_offset += item_len; 1696ac8e9819SChris Mason } 1697e2156867SHugo Mills (*num_found)++; 1698ac8e9819SChris Mason 1699ac8e9819SChris Mason if (*num_found >= sk->nr_items) 1700ac8e9819SChris Mason break; 1701ac8e9819SChris Mason } 1702ac8e9819SChris Mason advance_key: 1703ac8e9819SChris Mason ret = 0; 1704abc6e134SChris Mason if (key->offset < (u64)-1 && key->offset < sk->max_offset) 1705abc6e134SChris Mason key->offset++; 1706abc6e134SChris Mason else if (key->type < (u8)-1 && key->type < sk->max_type) { 1707abc6e134SChris Mason key->offset = 0; 1708abc6e134SChris Mason key->type++; 1709abc6e134SChris Mason } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) { 1710abc6e134SChris Mason key->offset = 0; 1711abc6e134SChris Mason key->type = 0; 1712abc6e134SChris Mason key->objectid++; 1713abc6e134SChris Mason } else 1714abc6e134SChris Mason ret = 1; 1715ac8e9819SChris Mason overflow: 1716ac8e9819SChris Mason return ret; 1717ac8e9819SChris Mason } 1718ac8e9819SChris Mason 1719ac8e9819SChris Mason static noinline int search_ioctl(struct inode *inode, 1720ac8e9819SChris Mason struct btrfs_ioctl_search_args *args) 1721ac8e9819SChris Mason { 1722ac8e9819SChris Mason struct btrfs_root *root; 1723ac8e9819SChris Mason struct btrfs_key key; 1724ac8e9819SChris Mason struct btrfs_key max_key; 1725ac8e9819SChris Mason struct btrfs_path *path; 1726ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk = &args->key; 1727ac8e9819SChris Mason struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info; 1728ac8e9819SChris Mason int ret; 1729ac8e9819SChris Mason int num_found = 0; 1730ac8e9819SChris Mason unsigned long sk_offset = 0; 1731ac8e9819SChris Mason 1732ac8e9819SChris Mason path = btrfs_alloc_path(); 1733ac8e9819SChris Mason if (!path) 1734ac8e9819SChris Mason return -ENOMEM; 1735ac8e9819SChris Mason 1736ac8e9819SChris Mason if (sk->tree_id == 0) { 1737ac8e9819SChris Mason /* search the root of the inode that was passed */ 1738ac8e9819SChris Mason root = BTRFS_I(inode)->root; 1739ac8e9819SChris Mason } else { 1740ac8e9819SChris Mason key.objectid = sk->tree_id; 1741ac8e9819SChris Mason key.type = BTRFS_ROOT_ITEM_KEY; 1742ac8e9819SChris Mason key.offset = (u64)-1; 1743ac8e9819SChris Mason root = btrfs_read_fs_root_no_name(info, &key); 1744ac8e9819SChris Mason if (IS_ERR(root)) { 1745ac8e9819SChris Mason printk(KERN_ERR "could not find root %llu\n", 1746ac8e9819SChris Mason sk->tree_id); 1747ac8e9819SChris Mason btrfs_free_path(path); 1748ac8e9819SChris Mason return -ENOENT; 1749ac8e9819SChris Mason } 1750ac8e9819SChris Mason } 1751ac8e9819SChris Mason 1752ac8e9819SChris Mason key.objectid = sk->min_objectid; 1753ac8e9819SChris Mason key.type = sk->min_type; 1754ac8e9819SChris Mason key.offset = sk->min_offset; 1755ac8e9819SChris Mason 1756ac8e9819SChris Mason max_key.objectid = sk->max_objectid; 1757ac8e9819SChris Mason max_key.type = sk->max_type; 1758ac8e9819SChris Mason max_key.offset = sk->max_offset; 1759ac8e9819SChris Mason 1760ac8e9819SChris Mason path->keep_locks = 1; 1761ac8e9819SChris Mason 1762ac8e9819SChris Mason while(1) { 1763ac8e9819SChris Mason ret = btrfs_search_forward(root, &key, &max_key, path, 0, 1764ac8e9819SChris Mason sk->min_transid); 1765ac8e9819SChris Mason if (ret != 0) { 1766ac8e9819SChris Mason if (ret > 0) 1767ac8e9819SChris Mason ret = 0; 1768ac8e9819SChris Mason goto err; 1769ac8e9819SChris Mason } 1770ac8e9819SChris Mason ret = copy_to_sk(root, path, &key, sk, args->buf, 1771ac8e9819SChris Mason &sk_offset, &num_found); 1772b3b4aa74SDavid Sterba btrfs_release_path(path); 1773ac8e9819SChris Mason if (ret || num_found >= sk->nr_items) 1774ac8e9819SChris Mason break; 1775ac8e9819SChris Mason 1776ac8e9819SChris Mason } 1777ac8e9819SChris Mason ret = 0; 1778ac8e9819SChris Mason err: 1779ac8e9819SChris Mason sk->nr_items = num_found; 1780ac8e9819SChris Mason btrfs_free_path(path); 1781ac8e9819SChris Mason return ret; 1782ac8e9819SChris Mason } 1783ac8e9819SChris Mason 1784ac8e9819SChris Mason static noinline int btrfs_ioctl_tree_search(struct file *file, 1785ac8e9819SChris Mason void __user *argp) 1786ac8e9819SChris Mason { 1787ac8e9819SChris Mason struct btrfs_ioctl_search_args *args; 1788ac8e9819SChris Mason struct inode *inode; 1789ac8e9819SChris Mason int ret; 1790ac8e9819SChris Mason 1791ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 1792ac8e9819SChris Mason return -EPERM; 1793ac8e9819SChris Mason 17942354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 17952354d08fSJulia Lawall if (IS_ERR(args)) 17962354d08fSJulia Lawall return PTR_ERR(args); 1797ac8e9819SChris Mason 1798ac8e9819SChris Mason inode = fdentry(file)->d_inode; 1799ac8e9819SChris Mason ret = search_ioctl(inode, args); 1800ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 1801ac8e9819SChris Mason ret = -EFAULT; 1802ac8e9819SChris Mason kfree(args); 1803ac8e9819SChris Mason return ret; 1804ac8e9819SChris Mason } 1805ac8e9819SChris Mason 180698d377a0STARUISI Hiroaki /* 1807ac8e9819SChris Mason * Search INODE_REFs to identify path name of 'dirid' directory 1808ac8e9819SChris Mason * in a 'tree_id' tree. and sets path name to 'name'. 180998d377a0STARUISI Hiroaki */ 181098d377a0STARUISI Hiroaki static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, 181198d377a0STARUISI Hiroaki u64 tree_id, u64 dirid, char *name) 181298d377a0STARUISI Hiroaki { 181398d377a0STARUISI Hiroaki struct btrfs_root *root; 181498d377a0STARUISI Hiroaki struct btrfs_key key; 1815ac8e9819SChris Mason char *ptr; 181698d377a0STARUISI Hiroaki int ret = -1; 181798d377a0STARUISI Hiroaki int slot; 181898d377a0STARUISI Hiroaki int len; 181998d377a0STARUISI Hiroaki int total_len = 0; 182098d377a0STARUISI Hiroaki struct btrfs_inode_ref *iref; 182198d377a0STARUISI Hiroaki struct extent_buffer *l; 182298d377a0STARUISI Hiroaki struct btrfs_path *path; 182398d377a0STARUISI Hiroaki 182498d377a0STARUISI Hiroaki if (dirid == BTRFS_FIRST_FREE_OBJECTID) { 182598d377a0STARUISI Hiroaki name[0]='\0'; 182698d377a0STARUISI Hiroaki return 0; 182798d377a0STARUISI Hiroaki } 182898d377a0STARUISI Hiroaki 182998d377a0STARUISI Hiroaki path = btrfs_alloc_path(); 183098d377a0STARUISI Hiroaki if (!path) 183198d377a0STARUISI Hiroaki return -ENOMEM; 183298d377a0STARUISI Hiroaki 1833ac8e9819SChris Mason ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX]; 183498d377a0STARUISI Hiroaki 183598d377a0STARUISI Hiroaki key.objectid = tree_id; 183698d377a0STARUISI Hiroaki key.type = BTRFS_ROOT_ITEM_KEY; 183798d377a0STARUISI Hiroaki key.offset = (u64)-1; 183898d377a0STARUISI Hiroaki root = btrfs_read_fs_root_no_name(info, &key); 183998d377a0STARUISI Hiroaki if (IS_ERR(root)) { 184098d377a0STARUISI Hiroaki printk(KERN_ERR "could not find root %llu\n", tree_id); 18418ad6fcabSChris Mason ret = -ENOENT; 18428ad6fcabSChris Mason goto out; 184398d377a0STARUISI Hiroaki } 184498d377a0STARUISI Hiroaki 184598d377a0STARUISI Hiroaki key.objectid = dirid; 184698d377a0STARUISI Hiroaki key.type = BTRFS_INODE_REF_KEY; 18478ad6fcabSChris Mason key.offset = (u64)-1; 184898d377a0STARUISI Hiroaki 184998d377a0STARUISI Hiroaki while(1) { 185098d377a0STARUISI Hiroaki ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 185198d377a0STARUISI Hiroaki if (ret < 0) 185298d377a0STARUISI Hiroaki goto out; 185398d377a0STARUISI Hiroaki 185498d377a0STARUISI Hiroaki l = path->nodes[0]; 185598d377a0STARUISI Hiroaki slot = path->slots[0]; 18568ad6fcabSChris Mason if (ret > 0 && slot > 0) 18578ad6fcabSChris Mason slot--; 185898d377a0STARUISI Hiroaki btrfs_item_key_to_cpu(l, &key, slot); 185998d377a0STARUISI Hiroaki 186098d377a0STARUISI Hiroaki if (ret > 0 && (key.objectid != dirid || 1861ac8e9819SChris Mason key.type != BTRFS_INODE_REF_KEY)) { 1862ac8e9819SChris Mason ret = -ENOENT; 186398d377a0STARUISI Hiroaki goto out; 1864ac8e9819SChris Mason } 186598d377a0STARUISI Hiroaki 186698d377a0STARUISI Hiroaki iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); 186798d377a0STARUISI Hiroaki len = btrfs_inode_ref_name_len(l, iref); 186898d377a0STARUISI Hiroaki ptr -= len + 1; 186998d377a0STARUISI Hiroaki total_len += len + 1; 1870ac8e9819SChris Mason if (ptr < name) 187198d377a0STARUISI Hiroaki goto out; 187298d377a0STARUISI Hiroaki 187398d377a0STARUISI Hiroaki *(ptr + len) = '/'; 187498d377a0STARUISI Hiroaki read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len); 187598d377a0STARUISI Hiroaki 187698d377a0STARUISI Hiroaki if (key.offset == BTRFS_FIRST_FREE_OBJECTID) 187798d377a0STARUISI Hiroaki break; 187898d377a0STARUISI Hiroaki 1879b3b4aa74SDavid Sterba btrfs_release_path(path); 188098d377a0STARUISI Hiroaki key.objectid = key.offset; 18818ad6fcabSChris Mason key.offset = (u64)-1; 188298d377a0STARUISI Hiroaki dirid = key.objectid; 188398d377a0STARUISI Hiroaki } 1884ac8e9819SChris Mason if (ptr < name) 188598d377a0STARUISI Hiroaki goto out; 188677906a50SLi Zefan memmove(name, ptr, total_len); 188798d377a0STARUISI Hiroaki name[total_len]='\0'; 188898d377a0STARUISI Hiroaki ret = 0; 188998d377a0STARUISI Hiroaki out: 189098d377a0STARUISI Hiroaki btrfs_free_path(path); 1891ac8e9819SChris Mason return ret; 1892ac8e9819SChris Mason } 1893ac8e9819SChris Mason 1894ac8e9819SChris Mason static noinline int btrfs_ioctl_ino_lookup(struct file *file, 1895ac8e9819SChris Mason void __user *argp) 1896ac8e9819SChris Mason { 1897ac8e9819SChris Mason struct btrfs_ioctl_ino_lookup_args *args; 1898ac8e9819SChris Mason struct inode *inode; 1899ac8e9819SChris Mason int ret; 1900ac8e9819SChris Mason 1901ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 1902ac8e9819SChris Mason return -EPERM; 1903ac8e9819SChris Mason 19042354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 19052354d08fSJulia Lawall if (IS_ERR(args)) 19062354d08fSJulia Lawall return PTR_ERR(args); 1907c2b96929SDan Carpenter 1908ac8e9819SChris Mason inode = fdentry(file)->d_inode; 1909ac8e9819SChris Mason 19101b53ac4dSChris Mason if (args->treeid == 0) 19111b53ac4dSChris Mason args->treeid = BTRFS_I(inode)->root->root_key.objectid; 19121b53ac4dSChris Mason 1913ac8e9819SChris Mason ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, 1914ac8e9819SChris Mason args->treeid, args->objectid, 1915ac8e9819SChris Mason args->name); 1916ac8e9819SChris Mason 1917ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 1918ac8e9819SChris Mason ret = -EFAULT; 1919ac8e9819SChris Mason 1920ac8e9819SChris Mason kfree(args); 192198d377a0STARUISI Hiroaki return ret; 192298d377a0STARUISI Hiroaki } 192398d377a0STARUISI Hiroaki 192476dda93cSYan, Zheng static noinline int btrfs_ioctl_snap_destroy(struct file *file, 192576dda93cSYan, Zheng void __user *arg) 192676dda93cSYan, Zheng { 192776dda93cSYan, Zheng struct dentry *parent = fdentry(file); 192876dda93cSYan, Zheng struct dentry *dentry; 192976dda93cSYan, Zheng struct inode *dir = parent->d_inode; 193076dda93cSYan, Zheng struct inode *inode; 193176dda93cSYan, Zheng struct btrfs_root *root = BTRFS_I(dir)->root; 193276dda93cSYan, Zheng struct btrfs_root *dest = NULL; 193376dda93cSYan, Zheng struct btrfs_ioctl_vol_args *vol_args; 193476dda93cSYan, Zheng struct btrfs_trans_handle *trans; 193576dda93cSYan, Zheng int namelen; 193676dda93cSYan, Zheng int ret; 193776dda93cSYan, Zheng int err = 0; 193876dda93cSYan, Zheng 193976dda93cSYan, Zheng vol_args = memdup_user(arg, sizeof(*vol_args)); 194076dda93cSYan, Zheng if (IS_ERR(vol_args)) 194176dda93cSYan, Zheng return PTR_ERR(vol_args); 194276dda93cSYan, Zheng 194376dda93cSYan, Zheng vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 194476dda93cSYan, Zheng namelen = strlen(vol_args->name); 194576dda93cSYan, Zheng if (strchr(vol_args->name, '/') || 194676dda93cSYan, Zheng strncmp(vol_args->name, "..", namelen) == 0) { 194776dda93cSYan, Zheng err = -EINVAL; 194876dda93cSYan, Zheng goto out; 194976dda93cSYan, Zheng } 195076dda93cSYan, Zheng 1951a561be71SAl Viro err = mnt_want_write_file(file); 195276dda93cSYan, Zheng if (err) 195376dda93cSYan, Zheng goto out; 195476dda93cSYan, Zheng 195576dda93cSYan, Zheng mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT); 195676dda93cSYan, Zheng dentry = lookup_one_len(vol_args->name, parent, namelen); 195776dda93cSYan, Zheng if (IS_ERR(dentry)) { 195876dda93cSYan, Zheng err = PTR_ERR(dentry); 195976dda93cSYan, Zheng goto out_unlock_dir; 196076dda93cSYan, Zheng } 196176dda93cSYan, Zheng 196276dda93cSYan, Zheng if (!dentry->d_inode) { 196376dda93cSYan, Zheng err = -ENOENT; 196476dda93cSYan, Zheng goto out_dput; 196576dda93cSYan, Zheng } 196676dda93cSYan, Zheng 196776dda93cSYan, Zheng inode = dentry->d_inode; 19684260f7c7SSage Weil dest = BTRFS_I(inode)->root; 19694260f7c7SSage Weil if (!capable(CAP_SYS_ADMIN)){ 19704260f7c7SSage Weil /* 19714260f7c7SSage Weil * Regular user. Only allow this with a special mount 19724260f7c7SSage Weil * option, when the user has write+exec access to the 19734260f7c7SSage Weil * subvol root, and when rmdir(2) would have been 19744260f7c7SSage Weil * allowed. 19754260f7c7SSage Weil * 19764260f7c7SSage Weil * Note that this is _not_ check that the subvol is 19774260f7c7SSage Weil * empty or doesn't contain data that we wouldn't 19784260f7c7SSage Weil * otherwise be able to delete. 19794260f7c7SSage Weil * 19804260f7c7SSage Weil * Users who want to delete empty subvols should try 19814260f7c7SSage Weil * rmdir(2). 19824260f7c7SSage Weil */ 19834260f7c7SSage Weil err = -EPERM; 19844260f7c7SSage Weil if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED)) 19854260f7c7SSage Weil goto out_dput; 19864260f7c7SSage Weil 19874260f7c7SSage Weil /* 19884260f7c7SSage Weil * Do not allow deletion if the parent dir is the same 19894260f7c7SSage Weil * as the dir to be deleted. That means the ioctl 19904260f7c7SSage Weil * must be called on the dentry referencing the root 19914260f7c7SSage Weil * of the subvol, not a random directory contained 19924260f7c7SSage Weil * within it. 19934260f7c7SSage Weil */ 19944260f7c7SSage Weil err = -EINVAL; 19954260f7c7SSage Weil if (root == dest) 19964260f7c7SSage Weil goto out_dput; 19974260f7c7SSage Weil 19984260f7c7SSage Weil err = inode_permission(inode, MAY_WRITE | MAY_EXEC); 19994260f7c7SSage Weil if (err) 20004260f7c7SSage Weil goto out_dput; 20014260f7c7SSage Weil 20024260f7c7SSage Weil /* check if subvolume may be deleted by a non-root user */ 20034260f7c7SSage Weil err = btrfs_may_delete(dir, dentry, 1); 20044260f7c7SSage Weil if (err) 20054260f7c7SSage Weil goto out_dput; 20064260f7c7SSage Weil } 20074260f7c7SSage Weil 200833345d01SLi Zefan if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { 200976dda93cSYan, Zheng err = -EINVAL; 201076dda93cSYan, Zheng goto out_dput; 201176dda93cSYan, Zheng } 201276dda93cSYan, Zheng 201376dda93cSYan, Zheng mutex_lock(&inode->i_mutex); 201476dda93cSYan, Zheng err = d_invalidate(dentry); 201576dda93cSYan, Zheng if (err) 201676dda93cSYan, Zheng goto out_unlock; 201776dda93cSYan, Zheng 201876dda93cSYan, Zheng down_write(&root->fs_info->subvol_sem); 201976dda93cSYan, Zheng 202076dda93cSYan, Zheng err = may_destroy_subvol(dest); 202176dda93cSYan, Zheng if (err) 202276dda93cSYan, Zheng goto out_up_write; 202376dda93cSYan, Zheng 2024a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 2025a22285a6SYan, Zheng if (IS_ERR(trans)) { 2026a22285a6SYan, Zheng err = PTR_ERR(trans); 2027d327099aSDan Carpenter goto out_up_write; 2028a22285a6SYan, Zheng } 2029a22285a6SYan, Zheng trans->block_rsv = &root->fs_info->global_block_rsv; 2030a22285a6SYan, Zheng 203176dda93cSYan, Zheng ret = btrfs_unlink_subvol(trans, root, dir, 203276dda93cSYan, Zheng dest->root_key.objectid, 203376dda93cSYan, Zheng dentry->d_name.name, 203476dda93cSYan, Zheng dentry->d_name.len); 203579787eaaSJeff Mahoney if (ret) { 203679787eaaSJeff Mahoney err = ret; 203779787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, ret); 203879787eaaSJeff Mahoney goto out_end_trans; 203979787eaaSJeff Mahoney } 204076dda93cSYan, Zheng 204176dda93cSYan, Zheng btrfs_record_root_in_trans(trans, dest); 204276dda93cSYan, Zheng 204376dda93cSYan, Zheng memset(&dest->root_item.drop_progress, 0, 204476dda93cSYan, Zheng sizeof(dest->root_item.drop_progress)); 204576dda93cSYan, Zheng dest->root_item.drop_level = 0; 204676dda93cSYan, Zheng btrfs_set_root_refs(&dest->root_item, 0); 204776dda93cSYan, Zheng 2048d68fc57bSYan, Zheng if (!xchg(&dest->orphan_item_inserted, 1)) { 204976dda93cSYan, Zheng ret = btrfs_insert_orphan_item(trans, 205076dda93cSYan, Zheng root->fs_info->tree_root, 205176dda93cSYan, Zheng dest->root_key.objectid); 205279787eaaSJeff Mahoney if (ret) { 205379787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, ret); 205479787eaaSJeff Mahoney err = ret; 205579787eaaSJeff Mahoney goto out_end_trans; 2056d68fc57bSYan, Zheng } 205779787eaaSJeff Mahoney } 205879787eaaSJeff Mahoney out_end_trans: 2059531cb13fSSage Weil ret = btrfs_end_transaction(trans, root); 206079787eaaSJeff Mahoney if (ret && !err) 206179787eaaSJeff Mahoney err = ret; 206276dda93cSYan, Zheng inode->i_flags |= S_DEAD; 206376dda93cSYan, Zheng out_up_write: 206476dda93cSYan, Zheng up_write(&root->fs_info->subvol_sem); 206576dda93cSYan, Zheng out_unlock: 206676dda93cSYan, Zheng mutex_unlock(&inode->i_mutex); 206776dda93cSYan, Zheng if (!err) { 2068efefb143SYan, Zheng shrink_dcache_sb(root->fs_info->sb); 206976dda93cSYan, Zheng btrfs_invalidate_inodes(dest); 207076dda93cSYan, Zheng d_delete(dentry); 207176dda93cSYan, Zheng } 207276dda93cSYan, Zheng out_dput: 207376dda93cSYan, Zheng dput(dentry); 207476dda93cSYan, Zheng out_unlock_dir: 207576dda93cSYan, Zheng mutex_unlock(&dir->i_mutex); 20762a79f17eSAl Viro mnt_drop_write_file(file); 207776dda93cSYan, Zheng out: 207876dda93cSYan, Zheng kfree(vol_args); 207976dda93cSYan, Zheng return err; 208076dda93cSYan, Zheng } 208176dda93cSYan, Zheng 20821e701a32SChris Mason static int btrfs_ioctl_defrag(struct file *file, void __user *argp) 2083f46b5a66SChristoph Hellwig { 2084f46b5a66SChristoph Hellwig struct inode *inode = fdentry(file)->d_inode; 2085f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 20861e701a32SChris Mason struct btrfs_ioctl_defrag_range_args *range; 2087c146afadSYan Zheng int ret; 2088c146afadSYan Zheng 2089b83cc969SLi Zefan if (btrfs_root_readonly(root)) 2090b83cc969SLi Zefan return -EROFS; 2091b83cc969SLi Zefan 2092a561be71SAl Viro ret = mnt_want_write_file(file); 2093c146afadSYan Zheng if (ret) 2094c146afadSYan Zheng return ret; 2095f46b5a66SChristoph Hellwig 2096f46b5a66SChristoph Hellwig switch (inode->i_mode & S_IFMT) { 2097f46b5a66SChristoph Hellwig case S_IFDIR: 2098e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) { 2099e441d54dSChris Mason ret = -EPERM; 2100e441d54dSChris Mason goto out; 2101e441d54dSChris Mason } 21028929ecfaSYan, Zheng ret = btrfs_defrag_root(root, 0); 21038929ecfaSYan, Zheng if (ret) 21048929ecfaSYan, Zheng goto out; 21058929ecfaSYan, Zheng ret = btrfs_defrag_root(root->fs_info->extent_root, 0); 2106f46b5a66SChristoph Hellwig break; 2107f46b5a66SChristoph Hellwig case S_IFREG: 2108e441d54dSChris Mason if (!(file->f_mode & FMODE_WRITE)) { 2109e441d54dSChris Mason ret = -EINVAL; 2110e441d54dSChris Mason goto out; 2111e441d54dSChris Mason } 21121e701a32SChris Mason 21131e701a32SChris Mason range = kzalloc(sizeof(*range), GFP_KERNEL); 21141e701a32SChris Mason if (!range) { 21151e701a32SChris Mason ret = -ENOMEM; 21161e701a32SChris Mason goto out; 21171e701a32SChris Mason } 21181e701a32SChris Mason 21191e701a32SChris Mason if (argp) { 21201e701a32SChris Mason if (copy_from_user(range, argp, 21211e701a32SChris Mason sizeof(*range))) { 21221e701a32SChris Mason ret = -EFAULT; 21231e701a32SChris Mason kfree(range); 2124683be16eSDan Carpenter goto out; 21251e701a32SChris Mason } 21261e701a32SChris Mason /* compression requires us to start the IO */ 21271e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 21281e701a32SChris Mason range->flags |= BTRFS_DEFRAG_RANGE_START_IO; 21291e701a32SChris Mason range->extent_thresh = (u32)-1; 21301e701a32SChris Mason } 21311e701a32SChris Mason } else { 21321e701a32SChris Mason /* the rest are all set to zero by kzalloc */ 21331e701a32SChris Mason range->len = (u64)-1; 21341e701a32SChris Mason } 21354cb5300bSChris Mason ret = btrfs_defrag_file(fdentry(file)->d_inode, file, 21364cb5300bSChris Mason range, 0, 0); 21374cb5300bSChris Mason if (ret > 0) 21384cb5300bSChris Mason ret = 0; 21391e701a32SChris Mason kfree(range); 2140f46b5a66SChristoph Hellwig break; 21418929ecfaSYan, Zheng default: 21428929ecfaSYan, Zheng ret = -EINVAL; 2143f46b5a66SChristoph Hellwig } 2144e441d54dSChris Mason out: 21452a79f17eSAl Viro mnt_drop_write_file(file); 2146e441d54dSChris Mason return ret; 2147f46b5a66SChristoph Hellwig } 2148f46b5a66SChristoph Hellwig 2149b2950863SChristoph Hellwig static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg) 2150f46b5a66SChristoph Hellwig { 2151f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2152f46b5a66SChristoph Hellwig int ret; 2153f46b5a66SChristoph Hellwig 2154e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2155e441d54dSChris Mason return -EPERM; 2156e441d54dSChris Mason 2157c9e9f97bSIlya Dryomov mutex_lock(&root->fs_info->volume_mutex); 2158c9e9f97bSIlya Dryomov if (root->fs_info->balance_ctl) { 2159c9e9f97bSIlya Dryomov printk(KERN_INFO "btrfs: balance in progress\n"); 2160c9e9f97bSIlya Dryomov ret = -EINVAL; 2161c9e9f97bSIlya Dryomov goto out; 2162c9e9f97bSIlya Dryomov } 2163c9e9f97bSIlya Dryomov 2164dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2165c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 2166c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 2167c9e9f97bSIlya Dryomov goto out; 2168c9e9f97bSIlya Dryomov } 2169f46b5a66SChristoph Hellwig 21705516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 2171f46b5a66SChristoph Hellwig ret = btrfs_init_new_device(root, vol_args->name); 2172f46b5a66SChristoph Hellwig 2173f46b5a66SChristoph Hellwig kfree(vol_args); 2174c9e9f97bSIlya Dryomov out: 2175c9e9f97bSIlya Dryomov mutex_unlock(&root->fs_info->volume_mutex); 2176f46b5a66SChristoph Hellwig return ret; 2177f46b5a66SChristoph Hellwig } 2178f46b5a66SChristoph Hellwig 2179b2950863SChristoph Hellwig static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg) 2180f46b5a66SChristoph Hellwig { 2181f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2182f46b5a66SChristoph Hellwig int ret; 2183f46b5a66SChristoph Hellwig 2184e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2185e441d54dSChris Mason return -EPERM; 2186e441d54dSChris Mason 2187c146afadSYan Zheng if (root->fs_info->sb->s_flags & MS_RDONLY) 2188c146afadSYan Zheng return -EROFS; 2189c146afadSYan Zheng 2190c9e9f97bSIlya Dryomov mutex_lock(&root->fs_info->volume_mutex); 2191c9e9f97bSIlya Dryomov if (root->fs_info->balance_ctl) { 2192c9e9f97bSIlya Dryomov printk(KERN_INFO "btrfs: balance in progress\n"); 2193c9e9f97bSIlya Dryomov ret = -EINVAL; 2194c9e9f97bSIlya Dryomov goto out; 2195c9e9f97bSIlya Dryomov } 2196c9e9f97bSIlya Dryomov 2197dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2198c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 2199c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 2200c9e9f97bSIlya Dryomov goto out; 2201c9e9f97bSIlya Dryomov } 2202f46b5a66SChristoph Hellwig 22035516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 2204f46b5a66SChristoph Hellwig ret = btrfs_rm_device(root, vol_args->name); 2205f46b5a66SChristoph Hellwig 2206f46b5a66SChristoph Hellwig kfree(vol_args); 2207c9e9f97bSIlya Dryomov out: 2208c9e9f97bSIlya Dryomov mutex_unlock(&root->fs_info->volume_mutex); 2209f46b5a66SChristoph Hellwig return ret; 2210f46b5a66SChristoph Hellwig } 2211f46b5a66SChristoph Hellwig 2212475f6387SJan Schmidt static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg) 2213475f6387SJan Schmidt { 2214027ed2f0SLi Zefan struct btrfs_ioctl_fs_info_args *fi_args; 2215475f6387SJan Schmidt struct btrfs_device *device; 2216475f6387SJan Schmidt struct btrfs_device *next; 2217475f6387SJan Schmidt struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; 2218027ed2f0SLi Zefan int ret = 0; 2219475f6387SJan Schmidt 2220475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 2221475f6387SJan Schmidt return -EPERM; 2222475f6387SJan Schmidt 2223027ed2f0SLi Zefan fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL); 2224027ed2f0SLi Zefan if (!fi_args) 2225027ed2f0SLi Zefan return -ENOMEM; 2226027ed2f0SLi Zefan 2227027ed2f0SLi Zefan fi_args->num_devices = fs_devices->num_devices; 2228027ed2f0SLi Zefan memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid)); 2229475f6387SJan Schmidt 2230475f6387SJan Schmidt mutex_lock(&fs_devices->device_list_mutex); 2231475f6387SJan Schmidt list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) { 2232027ed2f0SLi Zefan if (device->devid > fi_args->max_id) 2233027ed2f0SLi Zefan fi_args->max_id = device->devid; 2234475f6387SJan Schmidt } 2235475f6387SJan Schmidt mutex_unlock(&fs_devices->device_list_mutex); 2236475f6387SJan Schmidt 2237027ed2f0SLi Zefan if (copy_to_user(arg, fi_args, sizeof(*fi_args))) 2238027ed2f0SLi Zefan ret = -EFAULT; 2239475f6387SJan Schmidt 2240027ed2f0SLi Zefan kfree(fi_args); 2241027ed2f0SLi Zefan return ret; 2242475f6387SJan Schmidt } 2243475f6387SJan Schmidt 2244475f6387SJan Schmidt static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg) 2245475f6387SJan Schmidt { 2246475f6387SJan Schmidt struct btrfs_ioctl_dev_info_args *di_args; 2247475f6387SJan Schmidt struct btrfs_device *dev; 2248475f6387SJan Schmidt struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; 2249475f6387SJan Schmidt int ret = 0; 2250475f6387SJan Schmidt char *s_uuid = NULL; 2251475f6387SJan Schmidt char empty_uuid[BTRFS_UUID_SIZE] = {0}; 2252475f6387SJan Schmidt 2253475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 2254475f6387SJan Schmidt return -EPERM; 2255475f6387SJan Schmidt 2256475f6387SJan Schmidt di_args = memdup_user(arg, sizeof(*di_args)); 2257475f6387SJan Schmidt if (IS_ERR(di_args)) 2258475f6387SJan Schmidt return PTR_ERR(di_args); 2259475f6387SJan Schmidt 2260475f6387SJan Schmidt if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0) 2261475f6387SJan Schmidt s_uuid = di_args->uuid; 2262475f6387SJan Schmidt 2263475f6387SJan Schmidt mutex_lock(&fs_devices->device_list_mutex); 2264475f6387SJan Schmidt dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL); 2265475f6387SJan Schmidt mutex_unlock(&fs_devices->device_list_mutex); 2266475f6387SJan Schmidt 2267475f6387SJan Schmidt if (!dev) { 2268475f6387SJan Schmidt ret = -ENODEV; 2269475f6387SJan Schmidt goto out; 2270475f6387SJan Schmidt } 2271475f6387SJan Schmidt 2272475f6387SJan Schmidt di_args->devid = dev->devid; 2273475f6387SJan Schmidt di_args->bytes_used = dev->bytes_used; 2274475f6387SJan Schmidt di_args->total_bytes = dev->total_bytes; 2275475f6387SJan Schmidt memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); 2276a27202fbSJim Meyering if (dev->name) { 2277606686eeSJosef Bacik struct rcu_string *name; 2278606686eeSJosef Bacik 2279606686eeSJosef Bacik rcu_read_lock(); 2280606686eeSJosef Bacik name = rcu_dereference(dev->name); 2281606686eeSJosef Bacik strncpy(di_args->path, name->str, sizeof(di_args->path)); 2282606686eeSJosef Bacik rcu_read_unlock(); 2283a27202fbSJim Meyering di_args->path[sizeof(di_args->path) - 1] = 0; 2284a27202fbSJim Meyering } else { 228599ba55adSStefan Behrens di_args->path[0] = '\0'; 2286a27202fbSJim Meyering } 2287475f6387SJan Schmidt 2288475f6387SJan Schmidt out: 2289475f6387SJan Schmidt if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) 2290475f6387SJan Schmidt ret = -EFAULT; 2291475f6387SJan Schmidt 2292475f6387SJan Schmidt kfree(di_args); 2293475f6387SJan Schmidt return ret; 2294475f6387SJan Schmidt } 2295475f6387SJan Schmidt 229676dda93cSYan, Zheng static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, 2297b2950863SChristoph Hellwig u64 off, u64 olen, u64 destoff) 2298f46b5a66SChristoph Hellwig { 2299f46b5a66SChristoph Hellwig struct inode *inode = fdentry(file)->d_inode; 2300f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 2301f46b5a66SChristoph Hellwig struct file *src_file; 2302f46b5a66SChristoph Hellwig struct inode *src; 2303f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 2304f46b5a66SChristoph Hellwig struct btrfs_path *path; 2305f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 2306ae01a0abSYan Zheng char *buf; 2307ae01a0abSYan Zheng struct btrfs_key key; 2308f46b5a66SChristoph Hellwig u32 nritems; 2309f46b5a66SChristoph Hellwig int slot; 2310ae01a0abSYan Zheng int ret; 2311c5c9cd4dSSage Weil u64 len = olen; 2312c5c9cd4dSSage Weil u64 bs = root->fs_info->sb->s_blocksize; 2313c5c9cd4dSSage Weil u64 hint_byte; 2314d20f7043SChris Mason 2315c5c9cd4dSSage Weil /* 2316c5c9cd4dSSage Weil * TODO: 2317c5c9cd4dSSage Weil * - split compressed inline extents. annoying: we need to 2318c5c9cd4dSSage Weil * decompress into destination's address_space (the file offset 2319c5c9cd4dSSage Weil * may change, so source mapping won't do), then recompress (or 2320c5c9cd4dSSage Weil * otherwise reinsert) a subrange. 2321c5c9cd4dSSage Weil * - allow ranges within the same file to be cloned (provided 2322c5c9cd4dSSage Weil * they don't overlap)? 2323c5c9cd4dSSage Weil */ 2324c5c9cd4dSSage Weil 2325e441d54dSChris Mason /* the destination must be opened for writing */ 23262ebc3464SDan Rosenberg if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND)) 2327e441d54dSChris Mason return -EINVAL; 2328e441d54dSChris Mason 2329b83cc969SLi Zefan if (btrfs_root_readonly(root)) 2330b83cc969SLi Zefan return -EROFS; 2331b83cc969SLi Zefan 2332a561be71SAl Viro ret = mnt_want_write_file(file); 2333c146afadSYan Zheng if (ret) 2334c146afadSYan Zheng return ret; 2335c146afadSYan Zheng 2336c5c9cd4dSSage Weil src_file = fget(srcfd); 2337ab67b7c1SYan Zheng if (!src_file) { 2338ab67b7c1SYan Zheng ret = -EBADF; 2339ab67b7c1SYan Zheng goto out_drop_write; 2340ab67b7c1SYan Zheng } 23415dc64164SDan Rosenberg 2342f46b5a66SChristoph Hellwig src = src_file->f_dentry->d_inode; 2343f46b5a66SChristoph Hellwig 2344c5c9cd4dSSage Weil ret = -EINVAL; 2345c5c9cd4dSSage Weil if (src == inode) 2346c5c9cd4dSSage Weil goto out_fput; 2347c5c9cd4dSSage Weil 23485dc64164SDan Rosenberg /* the src must be open for reading */ 23495dc64164SDan Rosenberg if (!(src_file->f_mode & FMODE_READ)) 23505dc64164SDan Rosenberg goto out_fput; 23515dc64164SDan Rosenberg 23520e7b824cSLi Zefan /* don't make the dst file partly checksummed */ 23530e7b824cSLi Zefan if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != 23540e7b824cSLi Zefan (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) 23550e7b824cSLi Zefan goto out_fput; 23560e7b824cSLi Zefan 2357ae01a0abSYan Zheng ret = -EISDIR; 2358ae01a0abSYan Zheng if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode)) 2359f46b5a66SChristoph Hellwig goto out_fput; 2360f46b5a66SChristoph Hellwig 2361ae01a0abSYan Zheng ret = -EXDEV; 2362ae01a0abSYan Zheng if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root) 2363ae01a0abSYan Zheng goto out_fput; 2364ae01a0abSYan Zheng 2365ae01a0abSYan Zheng ret = -ENOMEM; 2366ae01a0abSYan Zheng buf = vmalloc(btrfs_level_size(root, 0)); 2367ae01a0abSYan Zheng if (!buf) 2368ae01a0abSYan Zheng goto out_fput; 2369ae01a0abSYan Zheng 2370ae01a0abSYan Zheng path = btrfs_alloc_path(); 2371ae01a0abSYan Zheng if (!path) { 2372ae01a0abSYan Zheng vfree(buf); 2373ae01a0abSYan Zheng goto out_fput; 2374ae01a0abSYan Zheng } 2375ae01a0abSYan Zheng path->reada = 2; 2376ae01a0abSYan Zheng 2377f46b5a66SChristoph Hellwig if (inode < src) { 2378fccdae43SSage Weil mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT); 2379fccdae43SSage Weil mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD); 2380f46b5a66SChristoph Hellwig } else { 2381fccdae43SSage Weil mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT); 2382fccdae43SSage Weil mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD); 2383f46b5a66SChristoph Hellwig } 2384f46b5a66SChristoph Hellwig 2385c5c9cd4dSSage Weil /* determine range to clone */ 2386c5c9cd4dSSage Weil ret = -EINVAL; 23872ebc3464SDan Rosenberg if (off + len > src->i_size || off + len < off) 2388f46b5a66SChristoph Hellwig goto out_unlock; 2389c5c9cd4dSSage Weil if (len == 0) 2390c5c9cd4dSSage Weil olen = len = src->i_size - off; 2391c5c9cd4dSSage Weil /* if we extend to eof, continue to block boundary */ 2392c5c9cd4dSSage Weil if (off + len == src->i_size) 23932a6b8daeSLi Zefan len = ALIGN(src->i_size, bs) - off; 2394c5c9cd4dSSage Weil 2395c5c9cd4dSSage Weil /* verify the end result is block aligned */ 23962a6b8daeSLi Zefan if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) || 23972a6b8daeSLi Zefan !IS_ALIGNED(destoff, bs)) 2398c5c9cd4dSSage Weil goto out_unlock; 2399c5c9cd4dSSage Weil 2400d525e8abSLi Zefan if (destoff > inode->i_size) { 2401d525e8abSLi Zefan ret = btrfs_cont_expand(inode, inode->i_size, destoff); 2402d525e8abSLi Zefan if (ret) 2403d525e8abSLi Zefan goto out_unlock; 2404d525e8abSLi Zefan } 2405d525e8abSLi Zefan 240671ef0786SLi Zefan /* truncate page cache pages from target inode range */ 240771ef0786SLi Zefan truncate_inode_pages_range(&inode->i_data, destoff, 240871ef0786SLi Zefan PAGE_CACHE_ALIGN(destoff + len) - 1); 240971ef0786SLi Zefan 2410f46b5a66SChristoph Hellwig /* do any pending delalloc/csum calc on src, one way or 2411f46b5a66SChristoph Hellwig another, and lock file content */ 2412f46b5a66SChristoph Hellwig while (1) { 241331840ae1SZheng Yan struct btrfs_ordered_extent *ordered; 2414d0082371SJeff Mahoney lock_extent(&BTRFS_I(src)->io_tree, off, off+len); 24159a019196SSage Weil ordered = btrfs_lookup_first_ordered_extent(src, off+len); 24169a019196SSage Weil if (!ordered && 24179a019196SSage Weil !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len, 24189a019196SSage Weil EXTENT_DELALLOC, 0, NULL)) 2419f46b5a66SChristoph Hellwig break; 2420d0082371SJeff Mahoney unlock_extent(&BTRFS_I(src)->io_tree, off, off+len); 2421ae01a0abSYan Zheng if (ordered) 2422ae01a0abSYan Zheng btrfs_put_ordered_extent(ordered); 24239a019196SSage Weil btrfs_wait_ordered_range(src, off, len); 2424f46b5a66SChristoph Hellwig } 2425f46b5a66SChristoph Hellwig 2426c5c9cd4dSSage Weil /* clone data */ 242733345d01SLi Zefan key.objectid = btrfs_ino(src); 2428ae01a0abSYan Zheng key.type = BTRFS_EXTENT_DATA_KEY; 2429ae01a0abSYan Zheng key.offset = 0; 2430f46b5a66SChristoph Hellwig 2431f46b5a66SChristoph Hellwig while (1) { 2432f46b5a66SChristoph Hellwig /* 2433f46b5a66SChristoph Hellwig * note the key will change type as we walk through the 2434f46b5a66SChristoph Hellwig * tree. 2435f46b5a66SChristoph Hellwig */ 2436a22285a6SYan, Zheng ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 2437f46b5a66SChristoph Hellwig if (ret < 0) 2438f46b5a66SChristoph Hellwig goto out; 2439f46b5a66SChristoph Hellwig 2440ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 2441ae01a0abSYan Zheng if (path->slots[0] >= nritems) { 2442f46b5a66SChristoph Hellwig ret = btrfs_next_leaf(root, path); 2443f46b5a66SChristoph Hellwig if (ret < 0) 2444f46b5a66SChristoph Hellwig goto out; 2445f46b5a66SChristoph Hellwig if (ret > 0) 2446f46b5a66SChristoph Hellwig break; 2447ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 2448f46b5a66SChristoph Hellwig } 2449f46b5a66SChristoph Hellwig leaf = path->nodes[0]; 2450f46b5a66SChristoph Hellwig slot = path->slots[0]; 2451f46b5a66SChristoph Hellwig 2452ae01a0abSYan Zheng btrfs_item_key_to_cpu(leaf, &key, slot); 2453d20f7043SChris Mason if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY || 245433345d01SLi Zefan key.objectid != btrfs_ino(src)) 2455f46b5a66SChristoph Hellwig break; 2456f46b5a66SChristoph Hellwig 2457c5c9cd4dSSage Weil if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) { 2458c5c9cd4dSSage Weil struct btrfs_file_extent_item *extent; 2459c5c9cd4dSSage Weil int type; 246031840ae1SZheng Yan u32 size; 246131840ae1SZheng Yan struct btrfs_key new_key; 2462c5c9cd4dSSage Weil u64 disko = 0, diskl = 0; 2463c5c9cd4dSSage Weil u64 datao = 0, datal = 0; 2464c5c9cd4dSSage Weil u8 comp; 2465b5384d48SSage Weil u64 endoff; 246631840ae1SZheng Yan 246731840ae1SZheng Yan size = btrfs_item_size_nr(leaf, slot); 246831840ae1SZheng Yan read_extent_buffer(leaf, buf, 246931840ae1SZheng Yan btrfs_item_ptr_offset(leaf, slot), 247031840ae1SZheng Yan size); 2471c5c9cd4dSSage Weil 2472c5c9cd4dSSage Weil extent = btrfs_item_ptr(leaf, slot, 2473c5c9cd4dSSage Weil struct btrfs_file_extent_item); 2474c5c9cd4dSSage Weil comp = btrfs_file_extent_compression(leaf, extent); 2475c5c9cd4dSSage Weil type = btrfs_file_extent_type(leaf, extent); 2476c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 2477c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 2478d397712bSChris Mason disko = btrfs_file_extent_disk_bytenr(leaf, 2479d397712bSChris Mason extent); 2480d397712bSChris Mason diskl = btrfs_file_extent_disk_num_bytes(leaf, 2481d397712bSChris Mason extent); 2482c5c9cd4dSSage Weil datao = btrfs_file_extent_offset(leaf, extent); 2483d397712bSChris Mason datal = btrfs_file_extent_num_bytes(leaf, 2484d397712bSChris Mason extent); 2485c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 2486c5c9cd4dSSage Weil /* take upper bound, may be compressed */ 2487c5c9cd4dSSage Weil datal = btrfs_file_extent_ram_bytes(leaf, 2488c5c9cd4dSSage Weil extent); 2489c5c9cd4dSSage Weil } 2490b3b4aa74SDavid Sterba btrfs_release_path(path); 249131840ae1SZheng Yan 2492050006a7SSage Weil if (key.offset + datal <= off || 2493c5c9cd4dSSage Weil key.offset >= off+len) 2494c5c9cd4dSSage Weil goto next; 2495c5c9cd4dSSage Weil 249631840ae1SZheng Yan memcpy(&new_key, &key, sizeof(new_key)); 249733345d01SLi Zefan new_key.objectid = btrfs_ino(inode); 24984d728ec7SLi Zefan if (off <= key.offset) 2499c5c9cd4dSSage Weil new_key.offset = key.offset + destoff - off; 25004d728ec7SLi Zefan else 25014d728ec7SLi Zefan new_key.offset = destoff; 2502c5c9cd4dSSage Weil 2503b6f3409bSSage Weil /* 2504b6f3409bSSage Weil * 1 - adjusting old extent (we may have to split it) 2505b6f3409bSSage Weil * 1 - add new extent 2506b6f3409bSSage Weil * 1 - inode update 2507b6f3409bSSage Weil */ 2508b6f3409bSSage Weil trans = btrfs_start_transaction(root, 3); 2509a22285a6SYan, Zheng if (IS_ERR(trans)) { 2510a22285a6SYan, Zheng ret = PTR_ERR(trans); 2511a22285a6SYan, Zheng goto out; 2512a22285a6SYan, Zheng } 2513a22285a6SYan, Zheng 2514c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 2515c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 2516d72c0842SLi Zefan /* 2517d72c0842SLi Zefan * a | --- range to clone ---| b 2518d72c0842SLi Zefan * | ------------- extent ------------- | 2519d72c0842SLi Zefan */ 2520d72c0842SLi Zefan 2521d72c0842SLi Zefan /* substract range b */ 2522d72c0842SLi Zefan if (key.offset + datal > off + len) 2523d72c0842SLi Zefan datal = off + len - key.offset; 2524d72c0842SLi Zefan 2525d72c0842SLi Zefan /* substract range a */ 2526a22285a6SYan, Zheng if (off > key.offset) { 2527a22285a6SYan, Zheng datao += off - key.offset; 2528a22285a6SYan, Zheng datal -= off - key.offset; 2529a22285a6SYan, Zheng } 2530a22285a6SYan, Zheng 2531a22285a6SYan, Zheng ret = btrfs_drop_extents(trans, inode, 2532a22285a6SYan, Zheng new_key.offset, 2533a22285a6SYan, Zheng new_key.offset + datal, 2534a22285a6SYan, Zheng &hint_byte, 1); 253579787eaaSJeff Mahoney if (ret) { 253679787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, 253779787eaaSJeff Mahoney ret); 253879787eaaSJeff Mahoney btrfs_end_transaction(trans, root); 253979787eaaSJeff Mahoney goto out; 254079787eaaSJeff Mahoney } 2541a22285a6SYan, Zheng 254231840ae1SZheng Yan ret = btrfs_insert_empty_item(trans, root, path, 254331840ae1SZheng Yan &new_key, size); 254479787eaaSJeff Mahoney if (ret) { 254579787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, 254679787eaaSJeff Mahoney ret); 254779787eaaSJeff Mahoney btrfs_end_transaction(trans, root); 254879787eaaSJeff Mahoney goto out; 254979787eaaSJeff Mahoney } 255031840ae1SZheng Yan 255131840ae1SZheng Yan leaf = path->nodes[0]; 255231840ae1SZheng Yan slot = path->slots[0]; 255331840ae1SZheng Yan write_extent_buffer(leaf, buf, 255431840ae1SZheng Yan btrfs_item_ptr_offset(leaf, slot), 255531840ae1SZheng Yan size); 2556ae01a0abSYan Zheng 2557f46b5a66SChristoph Hellwig extent = btrfs_item_ptr(leaf, slot, 2558f46b5a66SChristoph Hellwig struct btrfs_file_extent_item); 2559c5c9cd4dSSage Weil 2560c5c9cd4dSSage Weil /* disko == 0 means it's a hole */ 2561c5c9cd4dSSage Weil if (!disko) 2562c5c9cd4dSSage Weil datao = 0; 2563c5c9cd4dSSage Weil 2564c5c9cd4dSSage Weil btrfs_set_file_extent_offset(leaf, extent, 2565c5c9cd4dSSage Weil datao); 2566c5c9cd4dSSage Weil btrfs_set_file_extent_num_bytes(leaf, extent, 2567c5c9cd4dSSage Weil datal); 2568c5c9cd4dSSage Weil if (disko) { 2569c5c9cd4dSSage Weil inode_add_bytes(inode, datal); 2570ae01a0abSYan Zheng ret = btrfs_inc_extent_ref(trans, root, 25715d4f98a2SYan Zheng disko, diskl, 0, 2572f46b5a66SChristoph Hellwig root->root_key.objectid, 257333345d01SLi Zefan btrfs_ino(inode), 257466d7e7f0SArne Jansen new_key.offset - datao, 257566d7e7f0SArne Jansen 0); 257679787eaaSJeff Mahoney if (ret) { 257779787eaaSJeff Mahoney btrfs_abort_transaction(trans, 257879787eaaSJeff Mahoney root, 257979787eaaSJeff Mahoney ret); 258079787eaaSJeff Mahoney btrfs_end_transaction(trans, 258179787eaaSJeff Mahoney root); 258279787eaaSJeff Mahoney goto out; 258379787eaaSJeff Mahoney 258479787eaaSJeff Mahoney } 2585f46b5a66SChristoph Hellwig } 2586c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 2587c5c9cd4dSSage Weil u64 skip = 0; 2588c5c9cd4dSSage Weil u64 trim = 0; 2589c5c9cd4dSSage Weil if (off > key.offset) { 2590c5c9cd4dSSage Weil skip = off - key.offset; 2591c5c9cd4dSSage Weil new_key.offset += skip; 259231840ae1SZheng Yan } 2593d397712bSChris Mason 2594c5c9cd4dSSage Weil if (key.offset + datal > off+len) 2595c5c9cd4dSSage Weil trim = key.offset + datal - (off+len); 2596d397712bSChris Mason 2597c5c9cd4dSSage Weil if (comp && (skip || trim)) { 2598c5c9cd4dSSage Weil ret = -EINVAL; 2599a22285a6SYan, Zheng btrfs_end_transaction(trans, root); 2600c5c9cd4dSSage Weil goto out; 260131840ae1SZheng Yan } 2602c5c9cd4dSSage Weil size -= skip + trim; 2603c5c9cd4dSSage Weil datal -= skip + trim; 2604a22285a6SYan, Zheng 2605a22285a6SYan, Zheng ret = btrfs_drop_extents(trans, inode, 2606a22285a6SYan, Zheng new_key.offset, 2607a22285a6SYan, Zheng new_key.offset + datal, 2608a22285a6SYan, Zheng &hint_byte, 1); 260979787eaaSJeff Mahoney if (ret) { 261079787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, 261179787eaaSJeff Mahoney ret); 261279787eaaSJeff Mahoney btrfs_end_transaction(trans, root); 261379787eaaSJeff Mahoney goto out; 261479787eaaSJeff Mahoney } 2615a22285a6SYan, Zheng 2616c5c9cd4dSSage Weil ret = btrfs_insert_empty_item(trans, root, path, 2617c5c9cd4dSSage Weil &new_key, size); 261879787eaaSJeff Mahoney if (ret) { 261979787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, 262079787eaaSJeff Mahoney ret); 262179787eaaSJeff Mahoney btrfs_end_transaction(trans, root); 262279787eaaSJeff Mahoney goto out; 262379787eaaSJeff Mahoney } 2624c5c9cd4dSSage Weil 2625c5c9cd4dSSage Weil if (skip) { 2626d397712bSChris Mason u32 start = 2627d397712bSChris Mason btrfs_file_extent_calc_inline_size(0); 2628c5c9cd4dSSage Weil memmove(buf+start, buf+start+skip, 2629c5c9cd4dSSage Weil datal); 2630c5c9cd4dSSage Weil } 2631c5c9cd4dSSage Weil 2632c5c9cd4dSSage Weil leaf = path->nodes[0]; 2633c5c9cd4dSSage Weil slot = path->slots[0]; 2634c5c9cd4dSSage Weil write_extent_buffer(leaf, buf, 2635c5c9cd4dSSage Weil btrfs_item_ptr_offset(leaf, slot), 2636c5c9cd4dSSage Weil size); 2637c5c9cd4dSSage Weil inode_add_bytes(inode, datal); 2638c5c9cd4dSSage Weil } 2639c5c9cd4dSSage Weil 2640c5c9cd4dSSage Weil btrfs_mark_buffer_dirty(leaf); 2641b3b4aa74SDavid Sterba btrfs_release_path(path); 2642c5c9cd4dSSage Weil 26430c4d2d95SJosef Bacik inode_inc_iversion(inode); 2644a22285a6SYan, Zheng inode->i_mtime = inode->i_ctime = CURRENT_TIME; 2645b5384d48SSage Weil 2646b5384d48SSage Weil /* 2647b5384d48SSage Weil * we round up to the block size at eof when 2648b5384d48SSage Weil * determining which extents to clone above, 2649b5384d48SSage Weil * but shouldn't round up the file size 2650b5384d48SSage Weil */ 2651b5384d48SSage Weil endoff = new_key.offset + datal; 26525f3888ffSLi Zefan if (endoff > destoff+olen) 26535f3888ffSLi Zefan endoff = destoff+olen; 2654b5384d48SSage Weil if (endoff > inode->i_size) 2655b5384d48SSage Weil btrfs_i_size_write(inode, endoff); 2656b5384d48SSage Weil 2657a22285a6SYan, Zheng ret = btrfs_update_inode(trans, root, inode); 265879787eaaSJeff Mahoney if (ret) { 265979787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, ret); 2660a22285a6SYan, Zheng btrfs_end_transaction(trans, root); 266179787eaaSJeff Mahoney goto out; 266279787eaaSJeff Mahoney } 266379787eaaSJeff Mahoney ret = btrfs_end_transaction(trans, root); 2664a22285a6SYan, Zheng } 2665c5c9cd4dSSage Weil next: 2666b3b4aa74SDavid Sterba btrfs_release_path(path); 2667ae01a0abSYan Zheng key.offset++; 2668ae01a0abSYan Zheng } 2669f46b5a66SChristoph Hellwig ret = 0; 2670f46b5a66SChristoph Hellwig out: 2671b3b4aa74SDavid Sterba btrfs_release_path(path); 2672d0082371SJeff Mahoney unlock_extent(&BTRFS_I(src)->io_tree, off, off+len); 2673f46b5a66SChristoph Hellwig out_unlock: 2674f46b5a66SChristoph Hellwig mutex_unlock(&src->i_mutex); 2675f46b5a66SChristoph Hellwig mutex_unlock(&inode->i_mutex); 2676ae01a0abSYan Zheng vfree(buf); 2677ae01a0abSYan Zheng btrfs_free_path(path); 2678f46b5a66SChristoph Hellwig out_fput: 2679f46b5a66SChristoph Hellwig fput(src_file); 2680ab67b7c1SYan Zheng out_drop_write: 26812a79f17eSAl Viro mnt_drop_write_file(file); 2682f46b5a66SChristoph Hellwig return ret; 2683f46b5a66SChristoph Hellwig } 2684f46b5a66SChristoph Hellwig 26857a865e8aSChristoph Hellwig static long btrfs_ioctl_clone_range(struct file *file, void __user *argp) 2686c5c9cd4dSSage Weil { 2687c5c9cd4dSSage Weil struct btrfs_ioctl_clone_range_args args; 2688c5c9cd4dSSage Weil 26897a865e8aSChristoph Hellwig if (copy_from_user(&args, argp, sizeof(args))) 2690c5c9cd4dSSage Weil return -EFAULT; 2691c5c9cd4dSSage Weil return btrfs_ioctl_clone(file, args.src_fd, args.src_offset, 2692c5c9cd4dSSage Weil args.src_length, args.dest_offset); 2693c5c9cd4dSSage Weil } 2694c5c9cd4dSSage Weil 2695f46b5a66SChristoph Hellwig /* 2696f46b5a66SChristoph Hellwig * there are many ways the trans_start and trans_end ioctls can lead 2697f46b5a66SChristoph Hellwig * to deadlocks. They should only be used by applications that 2698f46b5a66SChristoph Hellwig * basically own the machine, and have a very in depth understanding 2699f46b5a66SChristoph Hellwig * of all the possible deadlocks and enospc problems. 2700f46b5a66SChristoph Hellwig */ 2701b2950863SChristoph Hellwig static long btrfs_ioctl_trans_start(struct file *file) 2702f46b5a66SChristoph Hellwig { 2703f46b5a66SChristoph Hellwig struct inode *inode = fdentry(file)->d_inode; 2704f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 2705f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 27061ab86aedSSage Weil int ret; 2707f46b5a66SChristoph Hellwig 27081ab86aedSSage Weil ret = -EPERM; 2709df5b5520SChristoph Hellwig if (!capable(CAP_SYS_ADMIN)) 2710f46b5a66SChristoph Hellwig goto out; 27111ab86aedSSage Weil 27121ab86aedSSage Weil ret = -EINPROGRESS; 27131ab86aedSSage Weil if (file->private_data) 27141ab86aedSSage Weil goto out; 27159ca9ee09SSage Weil 2716b83cc969SLi Zefan ret = -EROFS; 2717b83cc969SLi Zefan if (btrfs_root_readonly(root)) 2718b83cc969SLi Zefan goto out; 2719b83cc969SLi Zefan 2720a561be71SAl Viro ret = mnt_want_write_file(file); 2721c146afadSYan Zheng if (ret) 2722c146afadSYan Zheng goto out; 2723c146afadSYan Zheng 2724a4abeea4SJosef Bacik atomic_inc(&root->fs_info->open_ioctl_trans); 27259ca9ee09SSage Weil 2726f46b5a66SChristoph Hellwig ret = -ENOMEM; 27277a7eaa40SJosef Bacik trans = btrfs_start_ioctl_transaction(root); 2728abd30bb0STsutomu Itoh if (IS_ERR(trans)) 27291ab86aedSSage Weil goto out_drop; 27301ab86aedSSage Weil 27311ab86aedSSage Weil file->private_data = trans; 27321ab86aedSSage Weil return 0; 27331ab86aedSSage Weil 27341ab86aedSSage Weil out_drop: 2735a4abeea4SJosef Bacik atomic_dec(&root->fs_info->open_ioctl_trans); 27362a79f17eSAl Viro mnt_drop_write_file(file); 2737f46b5a66SChristoph Hellwig out: 2738f46b5a66SChristoph Hellwig return ret; 2739f46b5a66SChristoph Hellwig } 2740f46b5a66SChristoph Hellwig 27416ef5ed0dSJosef Bacik static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) 27426ef5ed0dSJosef Bacik { 27436ef5ed0dSJosef Bacik struct inode *inode = fdentry(file)->d_inode; 27446ef5ed0dSJosef Bacik struct btrfs_root *root = BTRFS_I(inode)->root; 27456ef5ed0dSJosef Bacik struct btrfs_root *new_root; 27466ef5ed0dSJosef Bacik struct btrfs_dir_item *di; 27476ef5ed0dSJosef Bacik struct btrfs_trans_handle *trans; 27486ef5ed0dSJosef Bacik struct btrfs_path *path; 27496ef5ed0dSJosef Bacik struct btrfs_key location; 27506ef5ed0dSJosef Bacik struct btrfs_disk_key disk_key; 27516ef5ed0dSJosef Bacik struct btrfs_super_block *disk_super; 27526ef5ed0dSJosef Bacik u64 features; 27536ef5ed0dSJosef Bacik u64 objectid = 0; 27546ef5ed0dSJosef Bacik u64 dir_id; 27556ef5ed0dSJosef Bacik 27566ef5ed0dSJosef Bacik if (!capable(CAP_SYS_ADMIN)) 27576ef5ed0dSJosef Bacik return -EPERM; 27586ef5ed0dSJosef Bacik 27596ef5ed0dSJosef Bacik if (copy_from_user(&objectid, argp, sizeof(objectid))) 27606ef5ed0dSJosef Bacik return -EFAULT; 27616ef5ed0dSJosef Bacik 27626ef5ed0dSJosef Bacik if (!objectid) 27636ef5ed0dSJosef Bacik objectid = root->root_key.objectid; 27646ef5ed0dSJosef Bacik 27656ef5ed0dSJosef Bacik location.objectid = objectid; 27666ef5ed0dSJosef Bacik location.type = BTRFS_ROOT_ITEM_KEY; 27676ef5ed0dSJosef Bacik location.offset = (u64)-1; 27686ef5ed0dSJosef Bacik 27696ef5ed0dSJosef Bacik new_root = btrfs_read_fs_root_no_name(root->fs_info, &location); 27706ef5ed0dSJosef Bacik if (IS_ERR(new_root)) 27716ef5ed0dSJosef Bacik return PTR_ERR(new_root); 27726ef5ed0dSJosef Bacik 27736ef5ed0dSJosef Bacik if (btrfs_root_refs(&new_root->root_item) == 0) 27746ef5ed0dSJosef Bacik return -ENOENT; 27756ef5ed0dSJosef Bacik 27766ef5ed0dSJosef Bacik path = btrfs_alloc_path(); 27776ef5ed0dSJosef Bacik if (!path) 27786ef5ed0dSJosef Bacik return -ENOMEM; 27796ef5ed0dSJosef Bacik path->leave_spinning = 1; 27806ef5ed0dSJosef Bacik 27816ef5ed0dSJosef Bacik trans = btrfs_start_transaction(root, 1); 278298d5dc13STsutomu Itoh if (IS_ERR(trans)) { 27836ef5ed0dSJosef Bacik btrfs_free_path(path); 278498d5dc13STsutomu Itoh return PTR_ERR(trans); 27856ef5ed0dSJosef Bacik } 27866ef5ed0dSJosef Bacik 27876c41761fSDavid Sterba dir_id = btrfs_super_root_dir(root->fs_info->super_copy); 27886ef5ed0dSJosef Bacik di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path, 27896ef5ed0dSJosef Bacik dir_id, "default", 7, 1); 2790cf1e99a4SDan Carpenter if (IS_ERR_OR_NULL(di)) { 27916ef5ed0dSJosef Bacik btrfs_free_path(path); 27926ef5ed0dSJosef Bacik btrfs_end_transaction(trans, root); 27936ef5ed0dSJosef Bacik printk(KERN_ERR "Umm, you don't have the default dir item, " 27946ef5ed0dSJosef Bacik "this isn't going to work\n"); 27956ef5ed0dSJosef Bacik return -ENOENT; 27966ef5ed0dSJosef Bacik } 27976ef5ed0dSJosef Bacik 27986ef5ed0dSJosef Bacik btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); 27996ef5ed0dSJosef Bacik btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); 28006ef5ed0dSJosef Bacik btrfs_mark_buffer_dirty(path->nodes[0]); 28016ef5ed0dSJosef Bacik btrfs_free_path(path); 28026ef5ed0dSJosef Bacik 28036c41761fSDavid Sterba disk_super = root->fs_info->super_copy; 28046ef5ed0dSJosef Bacik features = btrfs_super_incompat_flags(disk_super); 28056ef5ed0dSJosef Bacik if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) { 28066ef5ed0dSJosef Bacik features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL; 28076ef5ed0dSJosef Bacik btrfs_set_super_incompat_flags(disk_super, features); 28086ef5ed0dSJosef Bacik } 28096ef5ed0dSJosef Bacik btrfs_end_transaction(trans, root); 28106ef5ed0dSJosef Bacik 28116ef5ed0dSJosef Bacik return 0; 28126ef5ed0dSJosef Bacik } 28136ef5ed0dSJosef Bacik 2814bf5fc093SJosef Bacik static void get_block_group_info(struct list_head *groups_list, 2815bf5fc093SJosef Bacik struct btrfs_ioctl_space_info *space) 2816bf5fc093SJosef Bacik { 2817bf5fc093SJosef Bacik struct btrfs_block_group_cache *block_group; 2818bf5fc093SJosef Bacik 2819bf5fc093SJosef Bacik space->total_bytes = 0; 2820bf5fc093SJosef Bacik space->used_bytes = 0; 2821bf5fc093SJosef Bacik space->flags = 0; 2822bf5fc093SJosef Bacik list_for_each_entry(block_group, groups_list, list) { 2823bf5fc093SJosef Bacik space->flags = block_group->flags; 2824bf5fc093SJosef Bacik space->total_bytes += block_group->key.offset; 2825bf5fc093SJosef Bacik space->used_bytes += 2826bf5fc093SJosef Bacik btrfs_block_group_used(&block_group->item); 2827bf5fc093SJosef Bacik } 2828bf5fc093SJosef Bacik } 2829bf5fc093SJosef Bacik 28301406e432SJosef Bacik long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg) 28311406e432SJosef Bacik { 28321406e432SJosef Bacik struct btrfs_ioctl_space_args space_args; 28331406e432SJosef Bacik struct btrfs_ioctl_space_info space; 28341406e432SJosef Bacik struct btrfs_ioctl_space_info *dest; 28357fde62bfSChris Mason struct btrfs_ioctl_space_info *dest_orig; 283613f2696fSDaniel J Blueman struct btrfs_ioctl_space_info __user *user_dest; 28371406e432SJosef Bacik struct btrfs_space_info *info; 2838bf5fc093SJosef Bacik u64 types[] = {BTRFS_BLOCK_GROUP_DATA, 2839bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_SYSTEM, 2840bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_METADATA, 2841bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA}; 2842bf5fc093SJosef Bacik int num_types = 4; 28437fde62bfSChris Mason int alloc_size; 28441406e432SJosef Bacik int ret = 0; 284551788b1bSDan Rosenberg u64 slot_count = 0; 2846bf5fc093SJosef Bacik int i, c; 28471406e432SJosef Bacik 28481406e432SJosef Bacik if (copy_from_user(&space_args, 28491406e432SJosef Bacik (struct btrfs_ioctl_space_args __user *)arg, 28501406e432SJosef Bacik sizeof(space_args))) 28511406e432SJosef Bacik return -EFAULT; 28521406e432SJosef Bacik 2853bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 2854bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 2855bf5fc093SJosef Bacik 2856bf5fc093SJosef Bacik info = NULL; 28577fde62bfSChris Mason rcu_read_lock(); 2858bf5fc093SJosef Bacik list_for_each_entry_rcu(tmp, &root->fs_info->space_info, 2859bf5fc093SJosef Bacik list) { 2860bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 2861bf5fc093SJosef Bacik info = tmp; 2862bf5fc093SJosef Bacik break; 2863bf5fc093SJosef Bacik } 2864bf5fc093SJosef Bacik } 28657fde62bfSChris Mason rcu_read_unlock(); 28661406e432SJosef Bacik 2867bf5fc093SJosef Bacik if (!info) 2868bf5fc093SJosef Bacik continue; 2869bf5fc093SJosef Bacik 2870bf5fc093SJosef Bacik down_read(&info->groups_sem); 2871bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 2872bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) 2873bf5fc093SJosef Bacik slot_count++; 2874bf5fc093SJosef Bacik } 2875bf5fc093SJosef Bacik up_read(&info->groups_sem); 2876bf5fc093SJosef Bacik } 2877bf5fc093SJosef Bacik 28787fde62bfSChris Mason /* space_slots == 0 means they are asking for a count */ 28797fde62bfSChris Mason if (space_args.space_slots == 0) { 28807fde62bfSChris Mason space_args.total_spaces = slot_count; 28817fde62bfSChris Mason goto out; 28827fde62bfSChris Mason } 2883bf5fc093SJosef Bacik 288451788b1bSDan Rosenberg slot_count = min_t(u64, space_args.space_slots, slot_count); 2885bf5fc093SJosef Bacik 28867fde62bfSChris Mason alloc_size = sizeof(*dest) * slot_count; 2887bf5fc093SJosef Bacik 28887fde62bfSChris Mason /* we generally have at most 6 or so space infos, one for each raid 28897fde62bfSChris Mason * level. So, a whole page should be more than enough for everyone 28907fde62bfSChris Mason */ 28917fde62bfSChris Mason if (alloc_size > PAGE_CACHE_SIZE) 28927fde62bfSChris Mason return -ENOMEM; 28937fde62bfSChris Mason 28947fde62bfSChris Mason space_args.total_spaces = 0; 28957fde62bfSChris Mason dest = kmalloc(alloc_size, GFP_NOFS); 28967fde62bfSChris Mason if (!dest) 28977fde62bfSChris Mason return -ENOMEM; 28987fde62bfSChris Mason dest_orig = dest; 28997fde62bfSChris Mason 29007fde62bfSChris Mason /* now we have a buffer to copy into */ 2901bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 2902bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 2903bf5fc093SJosef Bacik 290451788b1bSDan Rosenberg if (!slot_count) 290551788b1bSDan Rosenberg break; 290651788b1bSDan Rosenberg 2907bf5fc093SJosef Bacik info = NULL; 29081406e432SJosef Bacik rcu_read_lock(); 2909bf5fc093SJosef Bacik list_for_each_entry_rcu(tmp, &root->fs_info->space_info, 2910bf5fc093SJosef Bacik list) { 2911bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 2912bf5fc093SJosef Bacik info = tmp; 29137fde62bfSChris Mason break; 2914bf5fc093SJosef Bacik } 2915bf5fc093SJosef Bacik } 2916bf5fc093SJosef Bacik rcu_read_unlock(); 29177fde62bfSChris Mason 2918bf5fc093SJosef Bacik if (!info) 2919bf5fc093SJosef Bacik continue; 2920bf5fc093SJosef Bacik down_read(&info->groups_sem); 2921bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 2922bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) { 2923bf5fc093SJosef Bacik get_block_group_info(&info->block_groups[c], 2924bf5fc093SJosef Bacik &space); 29257fde62bfSChris Mason memcpy(dest, &space, sizeof(space)); 29261406e432SJosef Bacik dest++; 29271406e432SJosef Bacik space_args.total_spaces++; 292851788b1bSDan Rosenberg slot_count--; 29291406e432SJosef Bacik } 293051788b1bSDan Rosenberg if (!slot_count) 293151788b1bSDan Rosenberg break; 2932bf5fc093SJosef Bacik } 2933bf5fc093SJosef Bacik up_read(&info->groups_sem); 2934bf5fc093SJosef Bacik } 29351406e432SJosef Bacik 29362eec6c81SDaniel J Blueman user_dest = (struct btrfs_ioctl_space_info __user *) 29377fde62bfSChris Mason (arg + sizeof(struct btrfs_ioctl_space_args)); 29387fde62bfSChris Mason 29397fde62bfSChris Mason if (copy_to_user(user_dest, dest_orig, alloc_size)) 29407fde62bfSChris Mason ret = -EFAULT; 29417fde62bfSChris Mason 29427fde62bfSChris Mason kfree(dest_orig); 29437fde62bfSChris Mason out: 29447fde62bfSChris Mason if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) 29451406e432SJosef Bacik ret = -EFAULT; 29461406e432SJosef Bacik 29471406e432SJosef Bacik return ret; 29481406e432SJosef Bacik } 29491406e432SJosef Bacik 2950f46b5a66SChristoph Hellwig /* 2951f46b5a66SChristoph Hellwig * there are many ways the trans_start and trans_end ioctls can lead 2952f46b5a66SChristoph Hellwig * to deadlocks. They should only be used by applications that 2953f46b5a66SChristoph Hellwig * basically own the machine, and have a very in depth understanding 2954f46b5a66SChristoph Hellwig * of all the possible deadlocks and enospc problems. 2955f46b5a66SChristoph Hellwig */ 2956f46b5a66SChristoph Hellwig long btrfs_ioctl_trans_end(struct file *file) 2957f46b5a66SChristoph Hellwig { 2958f46b5a66SChristoph Hellwig struct inode *inode = fdentry(file)->d_inode; 2959f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 2960f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 2961f46b5a66SChristoph Hellwig 2962f46b5a66SChristoph Hellwig trans = file->private_data; 29631ab86aedSSage Weil if (!trans) 29641ab86aedSSage Weil return -EINVAL; 2965b214107eSChristoph Hellwig file->private_data = NULL; 29669ca9ee09SSage Weil 29671ab86aedSSage Weil btrfs_end_transaction(trans, root); 29681ab86aedSSage Weil 2969a4abeea4SJosef Bacik atomic_dec(&root->fs_info->open_ioctl_trans); 29709ca9ee09SSage Weil 29712a79f17eSAl Viro mnt_drop_write_file(file); 29721ab86aedSSage Weil return 0; 2973f46b5a66SChristoph Hellwig } 2974f46b5a66SChristoph Hellwig 297546204592SSage Weil static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp) 297646204592SSage Weil { 297746204592SSage Weil struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root; 297846204592SSage Weil struct btrfs_trans_handle *trans; 297946204592SSage Weil u64 transid; 2980db5b493aSTsutomu Itoh int ret; 298146204592SSage Weil 298246204592SSage Weil trans = btrfs_start_transaction(root, 0); 298398d5dc13STsutomu Itoh if (IS_ERR(trans)) 298498d5dc13STsutomu Itoh return PTR_ERR(trans); 298546204592SSage Weil transid = trans->transid; 2986db5b493aSTsutomu Itoh ret = btrfs_commit_transaction_async(trans, root, 0); 29878b2b2d3cSTsutomu Itoh if (ret) { 29888b2b2d3cSTsutomu Itoh btrfs_end_transaction(trans, root); 2989db5b493aSTsutomu Itoh return ret; 29908b2b2d3cSTsutomu Itoh } 299146204592SSage Weil 299246204592SSage Weil if (argp) 299346204592SSage Weil if (copy_to_user(argp, &transid, sizeof(transid))) 299446204592SSage Weil return -EFAULT; 299546204592SSage Weil return 0; 299646204592SSage Weil } 299746204592SSage Weil 299846204592SSage Weil static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp) 299946204592SSage Weil { 300046204592SSage Weil struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root; 300146204592SSage Weil u64 transid; 300246204592SSage Weil 300346204592SSage Weil if (argp) { 300446204592SSage Weil if (copy_from_user(&transid, argp, sizeof(transid))) 300546204592SSage Weil return -EFAULT; 300646204592SSage Weil } else { 300746204592SSage Weil transid = 0; /* current trans */ 300846204592SSage Weil } 300946204592SSage Weil return btrfs_wait_for_commit(root, transid); 301046204592SSage Weil } 301146204592SSage Weil 3012475f6387SJan Schmidt static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg) 3013475f6387SJan Schmidt { 3014475f6387SJan Schmidt int ret; 3015475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 3016475f6387SJan Schmidt 3017475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 3018475f6387SJan Schmidt return -EPERM; 3019475f6387SJan Schmidt 3020475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 3021475f6387SJan Schmidt if (IS_ERR(sa)) 3022475f6387SJan Schmidt return PTR_ERR(sa); 3023475f6387SJan Schmidt 3024475f6387SJan Schmidt ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end, 30258628764eSArne Jansen &sa->progress, sa->flags & BTRFS_SCRUB_READONLY); 3026475f6387SJan Schmidt 3027475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 3028475f6387SJan Schmidt ret = -EFAULT; 3029475f6387SJan Schmidt 3030475f6387SJan Schmidt kfree(sa); 3031475f6387SJan Schmidt return ret; 3032475f6387SJan Schmidt } 3033475f6387SJan Schmidt 3034475f6387SJan Schmidt static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg) 3035475f6387SJan Schmidt { 3036475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 3037475f6387SJan Schmidt return -EPERM; 3038475f6387SJan Schmidt 3039475f6387SJan Schmidt return btrfs_scrub_cancel(root); 3040475f6387SJan Schmidt } 3041475f6387SJan Schmidt 3042475f6387SJan Schmidt static long btrfs_ioctl_scrub_progress(struct btrfs_root *root, 3043475f6387SJan Schmidt void __user *arg) 3044475f6387SJan Schmidt { 3045475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 3046475f6387SJan Schmidt int ret; 3047475f6387SJan Schmidt 3048475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 3049475f6387SJan Schmidt return -EPERM; 3050475f6387SJan Schmidt 3051475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 3052475f6387SJan Schmidt if (IS_ERR(sa)) 3053475f6387SJan Schmidt return PTR_ERR(sa); 3054475f6387SJan Schmidt 3055475f6387SJan Schmidt ret = btrfs_scrub_progress(root, sa->devid, &sa->progress); 3056475f6387SJan Schmidt 3057475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 3058475f6387SJan Schmidt ret = -EFAULT; 3059475f6387SJan Schmidt 3060475f6387SJan Schmidt kfree(sa); 3061475f6387SJan Schmidt return ret; 3062475f6387SJan Schmidt } 3063475f6387SJan Schmidt 3064c11d2c23SStefan Behrens static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root, 3065c11d2c23SStefan Behrens void __user *arg, int reset_after_read) 3066c11d2c23SStefan Behrens { 3067c11d2c23SStefan Behrens struct btrfs_ioctl_get_dev_stats *sa; 3068c11d2c23SStefan Behrens int ret; 3069c11d2c23SStefan Behrens 3070c11d2c23SStefan Behrens if (reset_after_read && !capable(CAP_SYS_ADMIN)) 3071c11d2c23SStefan Behrens return -EPERM; 3072c11d2c23SStefan Behrens 3073c11d2c23SStefan Behrens sa = memdup_user(arg, sizeof(*sa)); 3074c11d2c23SStefan Behrens if (IS_ERR(sa)) 3075c11d2c23SStefan Behrens return PTR_ERR(sa); 3076c11d2c23SStefan Behrens 3077c11d2c23SStefan Behrens ret = btrfs_get_dev_stats(root, sa, reset_after_read); 3078c11d2c23SStefan Behrens 3079c11d2c23SStefan Behrens if (copy_to_user(arg, sa, sizeof(*sa))) 3080c11d2c23SStefan Behrens ret = -EFAULT; 3081c11d2c23SStefan Behrens 3082c11d2c23SStefan Behrens kfree(sa); 3083c11d2c23SStefan Behrens return ret; 3084c11d2c23SStefan Behrens } 3085c11d2c23SStefan Behrens 3086d7728c96SJan Schmidt static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) 3087d7728c96SJan Schmidt { 3088d7728c96SJan Schmidt int ret = 0; 3089d7728c96SJan Schmidt int i; 3090740c3d22SChris Mason u64 rel_ptr; 3091d7728c96SJan Schmidt int size; 3092806468f8SChris Mason struct btrfs_ioctl_ino_path_args *ipa = NULL; 3093d7728c96SJan Schmidt struct inode_fs_paths *ipath = NULL; 3094d7728c96SJan Schmidt struct btrfs_path *path; 3095d7728c96SJan Schmidt 3096d7728c96SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 3097d7728c96SJan Schmidt return -EPERM; 3098d7728c96SJan Schmidt 3099d7728c96SJan Schmidt path = btrfs_alloc_path(); 3100d7728c96SJan Schmidt if (!path) { 3101d7728c96SJan Schmidt ret = -ENOMEM; 3102d7728c96SJan Schmidt goto out; 3103d7728c96SJan Schmidt } 3104d7728c96SJan Schmidt 3105d7728c96SJan Schmidt ipa = memdup_user(arg, sizeof(*ipa)); 3106d7728c96SJan Schmidt if (IS_ERR(ipa)) { 3107d7728c96SJan Schmidt ret = PTR_ERR(ipa); 3108d7728c96SJan Schmidt ipa = NULL; 3109d7728c96SJan Schmidt goto out; 3110d7728c96SJan Schmidt } 3111d7728c96SJan Schmidt 3112d7728c96SJan Schmidt size = min_t(u32, ipa->size, 4096); 3113d7728c96SJan Schmidt ipath = init_ipath(size, root, path); 3114d7728c96SJan Schmidt if (IS_ERR(ipath)) { 3115d7728c96SJan Schmidt ret = PTR_ERR(ipath); 3116d7728c96SJan Schmidt ipath = NULL; 3117d7728c96SJan Schmidt goto out; 3118d7728c96SJan Schmidt } 3119d7728c96SJan Schmidt 3120d7728c96SJan Schmidt ret = paths_from_inode(ipa->inum, ipath); 3121d7728c96SJan Schmidt if (ret < 0) 3122d7728c96SJan Schmidt goto out; 3123d7728c96SJan Schmidt 3124d7728c96SJan Schmidt for (i = 0; i < ipath->fspath->elem_cnt; ++i) { 3125745c4d8eSJeff Mahoney rel_ptr = ipath->fspath->val[i] - 3126745c4d8eSJeff Mahoney (u64)(unsigned long)ipath->fspath->val; 3127740c3d22SChris Mason ipath->fspath->val[i] = rel_ptr; 3128d7728c96SJan Schmidt } 3129d7728c96SJan Schmidt 3130745c4d8eSJeff Mahoney ret = copy_to_user((void *)(unsigned long)ipa->fspath, 3131745c4d8eSJeff Mahoney (void *)(unsigned long)ipath->fspath, size); 3132d7728c96SJan Schmidt if (ret) { 3133d7728c96SJan Schmidt ret = -EFAULT; 3134d7728c96SJan Schmidt goto out; 3135d7728c96SJan Schmidt } 3136d7728c96SJan Schmidt 3137d7728c96SJan Schmidt out: 3138d7728c96SJan Schmidt btrfs_free_path(path); 3139d7728c96SJan Schmidt free_ipath(ipath); 3140d7728c96SJan Schmidt kfree(ipa); 3141d7728c96SJan Schmidt 3142d7728c96SJan Schmidt return ret; 3143d7728c96SJan Schmidt } 3144d7728c96SJan Schmidt 3145d7728c96SJan Schmidt static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx) 3146d7728c96SJan Schmidt { 3147d7728c96SJan Schmidt struct btrfs_data_container *inodes = ctx; 3148d7728c96SJan Schmidt const size_t c = 3 * sizeof(u64); 3149d7728c96SJan Schmidt 3150d7728c96SJan Schmidt if (inodes->bytes_left >= c) { 3151d7728c96SJan Schmidt inodes->bytes_left -= c; 3152d7728c96SJan Schmidt inodes->val[inodes->elem_cnt] = inum; 3153d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 1] = offset; 3154d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 2] = root; 3155d7728c96SJan Schmidt inodes->elem_cnt += 3; 3156d7728c96SJan Schmidt } else { 3157d7728c96SJan Schmidt inodes->bytes_missing += c - inodes->bytes_left; 3158d7728c96SJan Schmidt inodes->bytes_left = 0; 3159d7728c96SJan Schmidt inodes->elem_missed += 3; 3160d7728c96SJan Schmidt } 3161d7728c96SJan Schmidt 3162d7728c96SJan Schmidt return 0; 3163d7728c96SJan Schmidt } 3164d7728c96SJan Schmidt 3165d7728c96SJan Schmidt static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root, 3166d7728c96SJan Schmidt void __user *arg) 3167d7728c96SJan Schmidt { 3168d7728c96SJan Schmidt int ret = 0; 3169d7728c96SJan Schmidt int size; 31704692cf58SJan Schmidt u64 extent_item_pos; 3171d7728c96SJan Schmidt struct btrfs_ioctl_logical_ino_args *loi; 3172d7728c96SJan Schmidt struct btrfs_data_container *inodes = NULL; 3173d7728c96SJan Schmidt struct btrfs_path *path = NULL; 3174d7728c96SJan Schmidt struct btrfs_key key; 3175d7728c96SJan Schmidt 3176d7728c96SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 3177d7728c96SJan Schmidt return -EPERM; 3178d7728c96SJan Schmidt 3179d7728c96SJan Schmidt loi = memdup_user(arg, sizeof(*loi)); 3180d7728c96SJan Schmidt if (IS_ERR(loi)) { 3181d7728c96SJan Schmidt ret = PTR_ERR(loi); 3182d7728c96SJan Schmidt loi = NULL; 3183d7728c96SJan Schmidt goto out; 3184d7728c96SJan Schmidt } 3185d7728c96SJan Schmidt 3186d7728c96SJan Schmidt path = btrfs_alloc_path(); 3187d7728c96SJan Schmidt if (!path) { 3188d7728c96SJan Schmidt ret = -ENOMEM; 3189d7728c96SJan Schmidt goto out; 3190d7728c96SJan Schmidt } 3191d7728c96SJan Schmidt 3192d7728c96SJan Schmidt size = min_t(u32, loi->size, 4096); 3193d7728c96SJan Schmidt inodes = init_data_container(size); 3194d7728c96SJan Schmidt if (IS_ERR(inodes)) { 3195d7728c96SJan Schmidt ret = PTR_ERR(inodes); 3196d7728c96SJan Schmidt inodes = NULL; 3197d7728c96SJan Schmidt goto out; 3198d7728c96SJan Schmidt } 3199d7728c96SJan Schmidt 3200d7728c96SJan Schmidt ret = extent_from_logical(root->fs_info, loi->logical, path, &key); 32014692cf58SJan Schmidt btrfs_release_path(path); 3202d7728c96SJan Schmidt 3203d7728c96SJan Schmidt if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK) 3204d7728c96SJan Schmidt ret = -ENOENT; 3205d7728c96SJan Schmidt if (ret < 0) 3206d7728c96SJan Schmidt goto out; 3207d7728c96SJan Schmidt 32084692cf58SJan Schmidt extent_item_pos = loi->logical - key.objectid; 32097a3ae2f8SJan Schmidt ret = iterate_extent_inodes(root->fs_info, key.objectid, 32107a3ae2f8SJan Schmidt extent_item_pos, 0, build_ino_list, 32114692cf58SJan Schmidt inodes); 3212d7728c96SJan Schmidt 3213d7728c96SJan Schmidt if (ret < 0) 3214d7728c96SJan Schmidt goto out; 3215d7728c96SJan Schmidt 3216745c4d8eSJeff Mahoney ret = copy_to_user((void *)(unsigned long)loi->inodes, 3217745c4d8eSJeff Mahoney (void *)(unsigned long)inodes, size); 3218d7728c96SJan Schmidt if (ret) 3219d7728c96SJan Schmidt ret = -EFAULT; 3220d7728c96SJan Schmidt 3221d7728c96SJan Schmidt out: 3222d7728c96SJan Schmidt btrfs_free_path(path); 3223d7728c96SJan Schmidt kfree(inodes); 3224d7728c96SJan Schmidt kfree(loi); 3225d7728c96SJan Schmidt 3226d7728c96SJan Schmidt return ret; 3227d7728c96SJan Schmidt } 3228d7728c96SJan Schmidt 322919a39dceSIlya Dryomov void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock, 3230c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs) 3231c9e9f97bSIlya Dryomov { 3232c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl = fs_info->balance_ctl; 3233c9e9f97bSIlya Dryomov 3234c9e9f97bSIlya Dryomov bargs->flags = bctl->flags; 3235c9e9f97bSIlya Dryomov 3236837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_running)) 3237837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_RUNNING; 3238837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_pause_req)) 3239837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; 3240a7e99c69SIlya Dryomov if (atomic_read(&fs_info->balance_cancel_req)) 3241a7e99c69SIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; 3242837d5b6eSIlya Dryomov 3243c9e9f97bSIlya Dryomov memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); 3244c9e9f97bSIlya Dryomov memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); 3245c9e9f97bSIlya Dryomov memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); 324619a39dceSIlya Dryomov 324719a39dceSIlya Dryomov if (lock) { 324819a39dceSIlya Dryomov spin_lock(&fs_info->balance_lock); 324919a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 325019a39dceSIlya Dryomov spin_unlock(&fs_info->balance_lock); 325119a39dceSIlya Dryomov } else { 325219a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 325319a39dceSIlya Dryomov } 3254c9e9f97bSIlya Dryomov } 3255c9e9f97bSIlya Dryomov 32569ba1f6e4SLiu Bo static long btrfs_ioctl_balance(struct file *file, void __user *arg) 3257c9e9f97bSIlya Dryomov { 32589ba1f6e4SLiu Bo struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; 3259c9e9f97bSIlya Dryomov struct btrfs_fs_info *fs_info = root->fs_info; 3260c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 3261c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl; 3262c9e9f97bSIlya Dryomov int ret; 3263c9e9f97bSIlya Dryomov 3264c9e9f97bSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 3265c9e9f97bSIlya Dryomov return -EPERM; 3266c9e9f97bSIlya Dryomov 3267c9e9f97bSIlya Dryomov if (fs_info->sb->s_flags & MS_RDONLY) 3268c9e9f97bSIlya Dryomov return -EROFS; 3269c9e9f97bSIlya Dryomov 327011e62a8fSAl Viro ret = mnt_want_write_file(file); 32719ba1f6e4SLiu Bo if (ret) 32729ba1f6e4SLiu Bo return ret; 32739ba1f6e4SLiu Bo 3274c9e9f97bSIlya Dryomov mutex_lock(&fs_info->volume_mutex); 3275c9e9f97bSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 3276c9e9f97bSIlya Dryomov 3277c9e9f97bSIlya Dryomov if (arg) { 3278c9e9f97bSIlya Dryomov bargs = memdup_user(arg, sizeof(*bargs)); 3279c9e9f97bSIlya Dryomov if (IS_ERR(bargs)) { 3280c9e9f97bSIlya Dryomov ret = PTR_ERR(bargs); 3281c9e9f97bSIlya Dryomov goto out; 3282c9e9f97bSIlya Dryomov } 3283de322263SIlya Dryomov 3284de322263SIlya Dryomov if (bargs->flags & BTRFS_BALANCE_RESUME) { 3285de322263SIlya Dryomov if (!fs_info->balance_ctl) { 3286de322263SIlya Dryomov ret = -ENOTCONN; 3287de322263SIlya Dryomov goto out_bargs; 3288de322263SIlya Dryomov } 3289de322263SIlya Dryomov 3290de322263SIlya Dryomov bctl = fs_info->balance_ctl; 3291de322263SIlya Dryomov spin_lock(&fs_info->balance_lock); 3292de322263SIlya Dryomov bctl->flags |= BTRFS_BALANCE_RESUME; 3293de322263SIlya Dryomov spin_unlock(&fs_info->balance_lock); 3294de322263SIlya Dryomov 3295de322263SIlya Dryomov goto do_balance; 3296de322263SIlya Dryomov } 3297c9e9f97bSIlya Dryomov } else { 3298c9e9f97bSIlya Dryomov bargs = NULL; 3299c9e9f97bSIlya Dryomov } 3300c9e9f97bSIlya Dryomov 3301837d5b6eSIlya Dryomov if (fs_info->balance_ctl) { 3302837d5b6eSIlya Dryomov ret = -EINPROGRESS; 3303837d5b6eSIlya Dryomov goto out_bargs; 3304837d5b6eSIlya Dryomov } 3305837d5b6eSIlya Dryomov 3306c9e9f97bSIlya Dryomov bctl = kzalloc(sizeof(*bctl), GFP_NOFS); 3307c9e9f97bSIlya Dryomov if (!bctl) { 3308c9e9f97bSIlya Dryomov ret = -ENOMEM; 3309c9e9f97bSIlya Dryomov goto out_bargs; 3310c9e9f97bSIlya Dryomov } 3311c9e9f97bSIlya Dryomov 3312c9e9f97bSIlya Dryomov bctl->fs_info = fs_info; 3313c9e9f97bSIlya Dryomov if (arg) { 3314c9e9f97bSIlya Dryomov memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); 3315c9e9f97bSIlya Dryomov memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); 3316c9e9f97bSIlya Dryomov memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); 3317c9e9f97bSIlya Dryomov 3318c9e9f97bSIlya Dryomov bctl->flags = bargs->flags; 3319f43ffb60SIlya Dryomov } else { 3320f43ffb60SIlya Dryomov /* balance everything - no filters */ 3321f43ffb60SIlya Dryomov bctl->flags |= BTRFS_BALANCE_TYPE_MASK; 3322c9e9f97bSIlya Dryomov } 3323c9e9f97bSIlya Dryomov 3324de322263SIlya Dryomov do_balance: 3325c9e9f97bSIlya Dryomov ret = btrfs_balance(bctl, bargs); 3326c9e9f97bSIlya Dryomov /* 3327837d5b6eSIlya Dryomov * bctl is freed in __cancel_balance or in free_fs_info if 3328837d5b6eSIlya Dryomov * restriper was paused all the way until unmount 3329c9e9f97bSIlya Dryomov */ 3330c9e9f97bSIlya Dryomov if (arg) { 3331c9e9f97bSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 3332c9e9f97bSIlya Dryomov ret = -EFAULT; 3333c9e9f97bSIlya Dryomov } 3334c9e9f97bSIlya Dryomov 3335c9e9f97bSIlya Dryomov out_bargs: 3336c9e9f97bSIlya Dryomov kfree(bargs); 3337c9e9f97bSIlya Dryomov out: 3338c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 3339c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->volume_mutex); 334011e62a8fSAl Viro mnt_drop_write_file(file); 3341c9e9f97bSIlya Dryomov return ret; 3342c9e9f97bSIlya Dryomov } 3343c9e9f97bSIlya Dryomov 3344837d5b6eSIlya Dryomov static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd) 3345837d5b6eSIlya Dryomov { 3346837d5b6eSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 3347837d5b6eSIlya Dryomov return -EPERM; 3348837d5b6eSIlya Dryomov 3349837d5b6eSIlya Dryomov switch (cmd) { 3350837d5b6eSIlya Dryomov case BTRFS_BALANCE_CTL_PAUSE: 3351837d5b6eSIlya Dryomov return btrfs_pause_balance(root->fs_info); 3352a7e99c69SIlya Dryomov case BTRFS_BALANCE_CTL_CANCEL: 3353a7e99c69SIlya Dryomov return btrfs_cancel_balance(root->fs_info); 3354837d5b6eSIlya Dryomov } 3355837d5b6eSIlya Dryomov 3356837d5b6eSIlya Dryomov return -EINVAL; 3357837d5b6eSIlya Dryomov } 3358837d5b6eSIlya Dryomov 335919a39dceSIlya Dryomov static long btrfs_ioctl_balance_progress(struct btrfs_root *root, 336019a39dceSIlya Dryomov void __user *arg) 336119a39dceSIlya Dryomov { 336219a39dceSIlya Dryomov struct btrfs_fs_info *fs_info = root->fs_info; 336319a39dceSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 336419a39dceSIlya Dryomov int ret = 0; 336519a39dceSIlya Dryomov 336619a39dceSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 336719a39dceSIlya Dryomov return -EPERM; 336819a39dceSIlya Dryomov 336919a39dceSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 337019a39dceSIlya Dryomov if (!fs_info->balance_ctl) { 337119a39dceSIlya Dryomov ret = -ENOTCONN; 337219a39dceSIlya Dryomov goto out; 337319a39dceSIlya Dryomov } 337419a39dceSIlya Dryomov 337519a39dceSIlya Dryomov bargs = kzalloc(sizeof(*bargs), GFP_NOFS); 337619a39dceSIlya Dryomov if (!bargs) { 337719a39dceSIlya Dryomov ret = -ENOMEM; 337819a39dceSIlya Dryomov goto out; 337919a39dceSIlya Dryomov } 338019a39dceSIlya Dryomov 338119a39dceSIlya Dryomov update_ioctl_balance_args(fs_info, 1, bargs); 338219a39dceSIlya Dryomov 338319a39dceSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 338419a39dceSIlya Dryomov ret = -EFAULT; 338519a39dceSIlya Dryomov 338619a39dceSIlya Dryomov kfree(bargs); 338719a39dceSIlya Dryomov out: 338819a39dceSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 338919a39dceSIlya Dryomov return ret; 339019a39dceSIlya Dryomov } 339119a39dceSIlya Dryomov 3392f46b5a66SChristoph Hellwig long btrfs_ioctl(struct file *file, unsigned int 3393f46b5a66SChristoph Hellwig cmd, unsigned long arg) 3394f46b5a66SChristoph Hellwig { 3395f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; 33964bcabaa3SChristoph Hellwig void __user *argp = (void __user *)arg; 3397f46b5a66SChristoph Hellwig 3398f46b5a66SChristoph Hellwig switch (cmd) { 33996cbff00fSChristoph Hellwig case FS_IOC_GETFLAGS: 34006cbff00fSChristoph Hellwig return btrfs_ioctl_getflags(file, argp); 34016cbff00fSChristoph Hellwig case FS_IOC_SETFLAGS: 34026cbff00fSChristoph Hellwig return btrfs_ioctl_setflags(file, argp); 34036cbff00fSChristoph Hellwig case FS_IOC_GETVERSION: 34046cbff00fSChristoph Hellwig return btrfs_ioctl_getversion(file, argp); 3405f7039b1dSLi Dongyang case FITRIM: 3406f7039b1dSLi Dongyang return btrfs_ioctl_fitrim(file, argp); 3407f46b5a66SChristoph Hellwig case BTRFS_IOC_SNAP_CREATE: 3408fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 0); 3409fdfb1e4fSLi Zefan case BTRFS_IOC_SNAP_CREATE_V2: 3410fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create_v2(file, argp, 0); 34113de4586cSChris Mason case BTRFS_IOC_SUBVOL_CREATE: 3412fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 1); 341376dda93cSYan, Zheng case BTRFS_IOC_SNAP_DESTROY: 341476dda93cSYan, Zheng return btrfs_ioctl_snap_destroy(file, argp); 34150caa102dSLi Zefan case BTRFS_IOC_SUBVOL_GETFLAGS: 34160caa102dSLi Zefan return btrfs_ioctl_subvol_getflags(file, argp); 34170caa102dSLi Zefan case BTRFS_IOC_SUBVOL_SETFLAGS: 34180caa102dSLi Zefan return btrfs_ioctl_subvol_setflags(file, argp); 34196ef5ed0dSJosef Bacik case BTRFS_IOC_DEFAULT_SUBVOL: 34206ef5ed0dSJosef Bacik return btrfs_ioctl_default_subvol(file, argp); 3421f46b5a66SChristoph Hellwig case BTRFS_IOC_DEFRAG: 34221e701a32SChris Mason return btrfs_ioctl_defrag(file, NULL); 34231e701a32SChris Mason case BTRFS_IOC_DEFRAG_RANGE: 34241e701a32SChris Mason return btrfs_ioctl_defrag(file, argp); 3425f46b5a66SChristoph Hellwig case BTRFS_IOC_RESIZE: 34264bcabaa3SChristoph Hellwig return btrfs_ioctl_resize(root, argp); 3427f46b5a66SChristoph Hellwig case BTRFS_IOC_ADD_DEV: 34284bcabaa3SChristoph Hellwig return btrfs_ioctl_add_dev(root, argp); 3429f46b5a66SChristoph Hellwig case BTRFS_IOC_RM_DEV: 34304bcabaa3SChristoph Hellwig return btrfs_ioctl_rm_dev(root, argp); 3431475f6387SJan Schmidt case BTRFS_IOC_FS_INFO: 3432475f6387SJan Schmidt return btrfs_ioctl_fs_info(root, argp); 3433475f6387SJan Schmidt case BTRFS_IOC_DEV_INFO: 3434475f6387SJan Schmidt return btrfs_ioctl_dev_info(root, argp); 3435f46b5a66SChristoph Hellwig case BTRFS_IOC_BALANCE: 34369ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, NULL); 3437f46b5a66SChristoph Hellwig case BTRFS_IOC_CLONE: 3438c5c9cd4dSSage Weil return btrfs_ioctl_clone(file, arg, 0, 0, 0); 3439c5c9cd4dSSage Weil case BTRFS_IOC_CLONE_RANGE: 34407a865e8aSChristoph Hellwig return btrfs_ioctl_clone_range(file, argp); 3441f46b5a66SChristoph Hellwig case BTRFS_IOC_TRANS_START: 3442f46b5a66SChristoph Hellwig return btrfs_ioctl_trans_start(file); 3443f46b5a66SChristoph Hellwig case BTRFS_IOC_TRANS_END: 3444f46b5a66SChristoph Hellwig return btrfs_ioctl_trans_end(file); 3445ac8e9819SChris Mason case BTRFS_IOC_TREE_SEARCH: 3446ac8e9819SChris Mason return btrfs_ioctl_tree_search(file, argp); 3447ac8e9819SChris Mason case BTRFS_IOC_INO_LOOKUP: 3448ac8e9819SChris Mason return btrfs_ioctl_ino_lookup(file, argp); 3449d7728c96SJan Schmidt case BTRFS_IOC_INO_PATHS: 3450d7728c96SJan Schmidt return btrfs_ioctl_ino_to_path(root, argp); 3451d7728c96SJan Schmidt case BTRFS_IOC_LOGICAL_INO: 3452d7728c96SJan Schmidt return btrfs_ioctl_logical_to_ino(root, argp); 34531406e432SJosef Bacik case BTRFS_IOC_SPACE_INFO: 34541406e432SJosef Bacik return btrfs_ioctl_space_info(root, argp); 3455f46b5a66SChristoph Hellwig case BTRFS_IOC_SYNC: 3456f46b5a66SChristoph Hellwig btrfs_sync_fs(file->f_dentry->d_sb, 1); 3457f46b5a66SChristoph Hellwig return 0; 345846204592SSage Weil case BTRFS_IOC_START_SYNC: 345946204592SSage Weil return btrfs_ioctl_start_sync(file, argp); 346046204592SSage Weil case BTRFS_IOC_WAIT_SYNC: 346146204592SSage Weil return btrfs_ioctl_wait_sync(file, argp); 3462475f6387SJan Schmidt case BTRFS_IOC_SCRUB: 3463475f6387SJan Schmidt return btrfs_ioctl_scrub(root, argp); 3464475f6387SJan Schmidt case BTRFS_IOC_SCRUB_CANCEL: 3465475f6387SJan Schmidt return btrfs_ioctl_scrub_cancel(root, argp); 3466475f6387SJan Schmidt case BTRFS_IOC_SCRUB_PROGRESS: 3467475f6387SJan Schmidt return btrfs_ioctl_scrub_progress(root, argp); 3468c9e9f97bSIlya Dryomov case BTRFS_IOC_BALANCE_V2: 34699ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, argp); 3470837d5b6eSIlya Dryomov case BTRFS_IOC_BALANCE_CTL: 3471837d5b6eSIlya Dryomov return btrfs_ioctl_balance_ctl(root, arg); 347219a39dceSIlya Dryomov case BTRFS_IOC_BALANCE_PROGRESS: 347319a39dceSIlya Dryomov return btrfs_ioctl_balance_progress(root, argp); 3474c11d2c23SStefan Behrens case BTRFS_IOC_GET_DEV_STATS: 3475c11d2c23SStefan Behrens return btrfs_ioctl_get_dev_stats(root, argp, 0); 3476c11d2c23SStefan Behrens case BTRFS_IOC_GET_AND_RESET_DEV_STATS: 3477c11d2c23SStefan Behrens return btrfs_ioctl_get_dev_stats(root, argp, 1); 3478f46b5a66SChristoph Hellwig } 3479f46b5a66SChristoph Hellwig 3480f46b5a66SChristoph Hellwig return -ENOTTY; 3481f46b5a66SChristoph Hellwig } 3482