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" 55f46b5a66SChristoph Hellwig 566cbff00fSChristoph Hellwig /* Mask out flags that are inappropriate for the given type of inode. */ 576cbff00fSChristoph Hellwig static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags) 586cbff00fSChristoph Hellwig { 596cbff00fSChristoph Hellwig if (S_ISDIR(mode)) 606cbff00fSChristoph Hellwig return flags; 616cbff00fSChristoph Hellwig else if (S_ISREG(mode)) 626cbff00fSChristoph Hellwig return flags & ~FS_DIRSYNC_FL; 636cbff00fSChristoph Hellwig else 646cbff00fSChristoph Hellwig return flags & (FS_NODUMP_FL | FS_NOATIME_FL); 656cbff00fSChristoph Hellwig } 66f46b5a66SChristoph Hellwig 676cbff00fSChristoph Hellwig /* 686cbff00fSChristoph Hellwig * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl. 696cbff00fSChristoph Hellwig */ 706cbff00fSChristoph Hellwig static unsigned int btrfs_flags_to_ioctl(unsigned int flags) 716cbff00fSChristoph Hellwig { 726cbff00fSChristoph Hellwig unsigned int iflags = 0; 736cbff00fSChristoph Hellwig 746cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_SYNC) 756cbff00fSChristoph Hellwig iflags |= FS_SYNC_FL; 766cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_IMMUTABLE) 776cbff00fSChristoph Hellwig iflags |= FS_IMMUTABLE_FL; 786cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_APPEND) 796cbff00fSChristoph Hellwig iflags |= FS_APPEND_FL; 806cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NODUMP) 816cbff00fSChristoph Hellwig iflags |= FS_NODUMP_FL; 826cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NOATIME) 836cbff00fSChristoph Hellwig iflags |= FS_NOATIME_FL; 846cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_DIRSYNC) 856cbff00fSChristoph Hellwig iflags |= FS_DIRSYNC_FL; 86d0092bddSLi Zefan if (flags & BTRFS_INODE_NODATACOW) 87d0092bddSLi Zefan iflags |= FS_NOCOW_FL; 88d0092bddSLi Zefan 89d0092bddSLi Zefan if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS)) 90d0092bddSLi Zefan iflags |= FS_COMPR_FL; 91d0092bddSLi Zefan else if (flags & BTRFS_INODE_NOCOMPRESS) 92d0092bddSLi Zefan iflags |= FS_NOCOMP_FL; 936cbff00fSChristoph Hellwig 946cbff00fSChristoph Hellwig return iflags; 956cbff00fSChristoph Hellwig } 966cbff00fSChristoph Hellwig 976cbff00fSChristoph Hellwig /* 986cbff00fSChristoph Hellwig * Update inode->i_flags based on the btrfs internal flags. 996cbff00fSChristoph Hellwig */ 1006cbff00fSChristoph Hellwig void btrfs_update_iflags(struct inode *inode) 1016cbff00fSChristoph Hellwig { 1026cbff00fSChristoph Hellwig struct btrfs_inode *ip = BTRFS_I(inode); 1036cbff00fSChristoph Hellwig 1046cbff00fSChristoph Hellwig inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC); 1056cbff00fSChristoph Hellwig 1066cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_SYNC) 1076cbff00fSChristoph Hellwig inode->i_flags |= S_SYNC; 1086cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_IMMUTABLE) 1096cbff00fSChristoph Hellwig inode->i_flags |= S_IMMUTABLE; 1106cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_APPEND) 1116cbff00fSChristoph Hellwig inode->i_flags |= S_APPEND; 1126cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_NOATIME) 1136cbff00fSChristoph Hellwig inode->i_flags |= S_NOATIME; 1146cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_DIRSYNC) 1156cbff00fSChristoph Hellwig inode->i_flags |= S_DIRSYNC; 1166cbff00fSChristoph Hellwig } 1176cbff00fSChristoph Hellwig 1186cbff00fSChristoph Hellwig /* 1196cbff00fSChristoph Hellwig * Inherit flags from the parent inode. 1206cbff00fSChristoph Hellwig * 121e27425d6SJosef Bacik * Currently only the compression flags and the cow flags are inherited. 1226cbff00fSChristoph Hellwig */ 1236cbff00fSChristoph Hellwig void btrfs_inherit_iflags(struct inode *inode, struct inode *dir) 1246cbff00fSChristoph Hellwig { 1250b4dcea5SChris Mason unsigned int flags; 1260b4dcea5SChris Mason 1270b4dcea5SChris Mason if (!dir) 1280b4dcea5SChris Mason return; 1290b4dcea5SChris Mason 1300b4dcea5SChris Mason flags = BTRFS_I(dir)->flags; 1316cbff00fSChristoph Hellwig 132e27425d6SJosef Bacik if (flags & BTRFS_INODE_NOCOMPRESS) { 133e27425d6SJosef Bacik BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS; 134e27425d6SJosef Bacik BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS; 135e27425d6SJosef Bacik } else if (flags & BTRFS_INODE_COMPRESS) { 136e27425d6SJosef Bacik BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS; 137e27425d6SJosef Bacik BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS; 138e27425d6SJosef Bacik } 1396cbff00fSChristoph Hellwig 140e27425d6SJosef Bacik if (flags & BTRFS_INODE_NODATACOW) 141e27425d6SJosef Bacik BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW; 142e27425d6SJosef Bacik 1436cbff00fSChristoph Hellwig btrfs_update_iflags(inode); 1446cbff00fSChristoph Hellwig } 1456cbff00fSChristoph Hellwig 1466cbff00fSChristoph Hellwig static int btrfs_ioctl_getflags(struct file *file, void __user *arg) 1476cbff00fSChristoph Hellwig { 1486cbff00fSChristoph Hellwig struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode); 1496cbff00fSChristoph Hellwig unsigned int flags = btrfs_flags_to_ioctl(ip->flags); 1506cbff00fSChristoph Hellwig 1516cbff00fSChristoph Hellwig if (copy_to_user(arg, &flags, sizeof(flags))) 1526cbff00fSChristoph Hellwig return -EFAULT; 1536cbff00fSChristoph Hellwig return 0; 1546cbff00fSChristoph Hellwig } 1556cbff00fSChristoph Hellwig 15675e7cb7fSLiu Bo static int check_flags(unsigned int flags) 15775e7cb7fSLiu Bo { 15875e7cb7fSLiu Bo if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \ 15975e7cb7fSLiu Bo FS_NOATIME_FL | FS_NODUMP_FL | \ 16075e7cb7fSLiu Bo FS_SYNC_FL | FS_DIRSYNC_FL | \ 161e1e8fb6aSLi Zefan FS_NOCOMP_FL | FS_COMPR_FL | 162e1e8fb6aSLi Zefan FS_NOCOW_FL)) 16375e7cb7fSLiu Bo return -EOPNOTSUPP; 16475e7cb7fSLiu Bo 16575e7cb7fSLiu Bo if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL)) 16675e7cb7fSLiu Bo return -EINVAL; 16775e7cb7fSLiu Bo 16875e7cb7fSLiu Bo return 0; 16975e7cb7fSLiu Bo } 17075e7cb7fSLiu Bo 1716cbff00fSChristoph Hellwig static int btrfs_ioctl_setflags(struct file *file, void __user *arg) 1726cbff00fSChristoph Hellwig { 1736cbff00fSChristoph Hellwig struct inode *inode = file->f_path.dentry->d_inode; 1746cbff00fSChristoph Hellwig struct btrfs_inode *ip = BTRFS_I(inode); 1756cbff00fSChristoph Hellwig struct btrfs_root *root = ip->root; 1766cbff00fSChristoph Hellwig struct btrfs_trans_handle *trans; 1776cbff00fSChristoph Hellwig unsigned int flags, oldflags; 1786cbff00fSChristoph Hellwig int ret; 179f062abf0SLi Zefan u64 ip_oldflags; 180f062abf0SLi Zefan unsigned int i_oldflags; 1816cbff00fSChristoph Hellwig 182b83cc969SLi Zefan if (btrfs_root_readonly(root)) 183b83cc969SLi Zefan return -EROFS; 184b83cc969SLi Zefan 1856cbff00fSChristoph Hellwig if (copy_from_user(&flags, arg, sizeof(flags))) 1866cbff00fSChristoph Hellwig return -EFAULT; 1876cbff00fSChristoph Hellwig 18875e7cb7fSLiu Bo ret = check_flags(flags); 18975e7cb7fSLiu Bo if (ret) 19075e7cb7fSLiu Bo return ret; 1916cbff00fSChristoph Hellwig 1922e149670SSerge E. Hallyn if (!inode_owner_or_capable(inode)) 1936cbff00fSChristoph Hellwig return -EACCES; 1946cbff00fSChristoph Hellwig 1956cbff00fSChristoph Hellwig mutex_lock(&inode->i_mutex); 1966cbff00fSChristoph Hellwig 197f062abf0SLi Zefan ip_oldflags = ip->flags; 198f062abf0SLi Zefan i_oldflags = inode->i_flags; 199f062abf0SLi Zefan 2006cbff00fSChristoph Hellwig flags = btrfs_mask_flags(inode->i_mode, flags); 2016cbff00fSChristoph Hellwig oldflags = btrfs_flags_to_ioctl(ip->flags); 2026cbff00fSChristoph Hellwig if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) { 2036cbff00fSChristoph Hellwig if (!capable(CAP_LINUX_IMMUTABLE)) { 2046cbff00fSChristoph Hellwig ret = -EPERM; 2056cbff00fSChristoph Hellwig goto out_unlock; 2066cbff00fSChristoph Hellwig } 2076cbff00fSChristoph Hellwig } 2086cbff00fSChristoph Hellwig 2096cbff00fSChristoph Hellwig ret = mnt_want_write(file->f_path.mnt); 2106cbff00fSChristoph Hellwig if (ret) 2116cbff00fSChristoph Hellwig goto out_unlock; 2126cbff00fSChristoph Hellwig 2136cbff00fSChristoph Hellwig if (flags & FS_SYNC_FL) 2146cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_SYNC; 2156cbff00fSChristoph Hellwig else 2166cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_SYNC; 2176cbff00fSChristoph Hellwig if (flags & FS_IMMUTABLE_FL) 2186cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_IMMUTABLE; 2196cbff00fSChristoph Hellwig else 2206cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_IMMUTABLE; 2216cbff00fSChristoph Hellwig if (flags & FS_APPEND_FL) 2226cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_APPEND; 2236cbff00fSChristoph Hellwig else 2246cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_APPEND; 2256cbff00fSChristoph Hellwig if (flags & FS_NODUMP_FL) 2266cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_NODUMP; 2276cbff00fSChristoph Hellwig else 2286cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_NODUMP; 2296cbff00fSChristoph Hellwig if (flags & FS_NOATIME_FL) 2306cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_NOATIME; 2316cbff00fSChristoph Hellwig else 2326cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_NOATIME; 2336cbff00fSChristoph Hellwig if (flags & FS_DIRSYNC_FL) 2346cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_DIRSYNC; 2356cbff00fSChristoph Hellwig else 2366cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_DIRSYNC; 237e1e8fb6aSLi Zefan if (flags & FS_NOCOW_FL) 238e1e8fb6aSLi Zefan ip->flags |= BTRFS_INODE_NODATACOW; 239e1e8fb6aSLi Zefan else 240e1e8fb6aSLi Zefan ip->flags &= ~BTRFS_INODE_NODATACOW; 2416cbff00fSChristoph Hellwig 24275e7cb7fSLiu Bo /* 24375e7cb7fSLiu Bo * The COMPRESS flag can only be changed by users, while the NOCOMPRESS 24475e7cb7fSLiu Bo * flag may be changed automatically if compression code won't make 24575e7cb7fSLiu Bo * things smaller. 24675e7cb7fSLiu Bo */ 24775e7cb7fSLiu Bo if (flags & FS_NOCOMP_FL) { 24875e7cb7fSLiu Bo ip->flags &= ~BTRFS_INODE_COMPRESS; 24975e7cb7fSLiu Bo ip->flags |= BTRFS_INODE_NOCOMPRESS; 25075e7cb7fSLiu Bo } else if (flags & FS_COMPR_FL) { 25175e7cb7fSLiu Bo ip->flags |= BTRFS_INODE_COMPRESS; 25275e7cb7fSLiu Bo ip->flags &= ~BTRFS_INODE_NOCOMPRESS; 253ebcb904dSLi Zefan } else { 254ebcb904dSLi Zefan ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS); 25575e7cb7fSLiu Bo } 2566cbff00fSChristoph Hellwig 2574da6f1a3SLi Zefan trans = btrfs_start_transaction(root, 1); 258f062abf0SLi Zefan if (IS_ERR(trans)) { 259f062abf0SLi Zefan ret = PTR_ERR(trans); 260f062abf0SLi Zefan goto out_drop; 261f062abf0SLi Zefan } 2626cbff00fSChristoph Hellwig 263306424ccSLi Zefan btrfs_update_iflags(inode); 264306424ccSLi Zefan inode->i_ctime = CURRENT_TIME; 2656cbff00fSChristoph Hellwig ret = btrfs_update_inode(trans, root, inode); 2666cbff00fSChristoph Hellwig 2676cbff00fSChristoph Hellwig btrfs_end_transaction(trans, root); 268f062abf0SLi Zefan out_drop: 269f062abf0SLi Zefan if (ret) { 270f062abf0SLi Zefan ip->flags = ip_oldflags; 271f062abf0SLi Zefan inode->i_flags = i_oldflags; 272f062abf0SLi Zefan } 2736cbff00fSChristoph Hellwig 2746cbff00fSChristoph Hellwig mnt_drop_write(file->f_path.mnt); 2756cbff00fSChristoph Hellwig out_unlock: 2766cbff00fSChristoph Hellwig mutex_unlock(&inode->i_mutex); 2772d4e6f6aSliubo return ret; 2786cbff00fSChristoph Hellwig } 2796cbff00fSChristoph Hellwig 2806cbff00fSChristoph Hellwig static int btrfs_ioctl_getversion(struct file *file, int __user *arg) 2816cbff00fSChristoph Hellwig { 2826cbff00fSChristoph Hellwig struct inode *inode = file->f_path.dentry->d_inode; 2836cbff00fSChristoph Hellwig 2846cbff00fSChristoph Hellwig return put_user(inode->i_generation, arg); 2856cbff00fSChristoph Hellwig } 286f46b5a66SChristoph Hellwig 287f7039b1dSLi Dongyang static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg) 288f7039b1dSLi Dongyang { 289f7039b1dSLi Dongyang struct btrfs_root *root = fdentry(file)->d_sb->s_fs_info; 290f7039b1dSLi Dongyang struct btrfs_fs_info *fs_info = root->fs_info; 291f7039b1dSLi Dongyang struct btrfs_device *device; 292f7039b1dSLi Dongyang struct request_queue *q; 293f7039b1dSLi Dongyang struct fstrim_range range; 294f7039b1dSLi Dongyang u64 minlen = ULLONG_MAX; 295f7039b1dSLi Dongyang u64 num_devices = 0; 2966c41761fSDavid Sterba u64 total_bytes = btrfs_super_total_bytes(root->fs_info->super_copy); 297f7039b1dSLi Dongyang int ret; 298f7039b1dSLi Dongyang 299f7039b1dSLi Dongyang if (!capable(CAP_SYS_ADMIN)) 300f7039b1dSLi Dongyang return -EPERM; 301f7039b1dSLi Dongyang 3021f78160cSXiao Guangrong rcu_read_lock(); 3031f78160cSXiao Guangrong list_for_each_entry_rcu(device, &fs_info->fs_devices->devices, 3041f78160cSXiao Guangrong dev_list) { 305f7039b1dSLi Dongyang if (!device->bdev) 306f7039b1dSLi Dongyang continue; 307f7039b1dSLi Dongyang q = bdev_get_queue(device->bdev); 308f7039b1dSLi Dongyang if (blk_queue_discard(q)) { 309f7039b1dSLi Dongyang num_devices++; 310f7039b1dSLi Dongyang minlen = min((u64)q->limits.discard_granularity, 311f7039b1dSLi Dongyang minlen); 312f7039b1dSLi Dongyang } 313f7039b1dSLi Dongyang } 3141f78160cSXiao Guangrong rcu_read_unlock(); 315f4c697e6SLukas Czerner 316f7039b1dSLi Dongyang if (!num_devices) 317f7039b1dSLi Dongyang return -EOPNOTSUPP; 318f7039b1dSLi Dongyang if (copy_from_user(&range, arg, sizeof(range))) 319f7039b1dSLi Dongyang return -EFAULT; 320f4c697e6SLukas Czerner if (range.start > total_bytes) 321f4c697e6SLukas Czerner return -EINVAL; 322f7039b1dSLi Dongyang 323f4c697e6SLukas Czerner range.len = min(range.len, total_bytes - range.start); 324f7039b1dSLi Dongyang range.minlen = max(range.minlen, minlen); 325f7039b1dSLi Dongyang ret = btrfs_trim_fs(root, &range); 326f7039b1dSLi Dongyang if (ret < 0) 327f7039b1dSLi Dongyang return ret; 328f7039b1dSLi Dongyang 329f7039b1dSLi Dongyang if (copy_to_user(arg, &range, sizeof(range))) 330f7039b1dSLi Dongyang return -EFAULT; 331f7039b1dSLi Dongyang 332f7039b1dSLi Dongyang return 0; 333f7039b1dSLi Dongyang } 334f7039b1dSLi Dongyang 335cb8e7090SChristoph Hellwig static noinline int create_subvol(struct btrfs_root *root, 336cb8e7090SChristoph Hellwig struct dentry *dentry, 33772fd032eSSage Weil char *name, int namelen, 33872fd032eSSage Weil u64 *async_transid) 339f46b5a66SChristoph Hellwig { 340f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 341f46b5a66SChristoph Hellwig struct btrfs_key key; 342f46b5a66SChristoph Hellwig struct btrfs_root_item root_item; 343f46b5a66SChristoph Hellwig struct btrfs_inode_item *inode_item; 344f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 34576dda93cSYan, Zheng struct btrfs_root *new_root; 3462fbe8c8aSAl Viro struct dentry *parent = dentry->d_parent; 3476a912213SJosef Bacik struct inode *dir; 348f46b5a66SChristoph Hellwig int ret; 349f46b5a66SChristoph Hellwig int err; 350f46b5a66SChristoph Hellwig u64 objectid; 351f46b5a66SChristoph Hellwig u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; 3523de4586cSChris Mason u64 index = 0; 353f46b5a66SChristoph Hellwig 354581bb050SLi Zefan ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid); 3552fbe8c8aSAl Viro if (ret) 356a22285a6SYan, Zheng return ret; 3576a912213SJosef Bacik 3586a912213SJosef Bacik dir = parent->d_inode; 3596a912213SJosef Bacik 3609ed74f2dSJosef Bacik /* 3619ed74f2dSJosef Bacik * 1 - inode item 3629ed74f2dSJosef Bacik * 2 - refs 3639ed74f2dSJosef Bacik * 1 - root item 3649ed74f2dSJosef Bacik * 2 - dir items 3659ed74f2dSJosef Bacik */ 366a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 6); 3672fbe8c8aSAl Viro if (IS_ERR(trans)) 368a22285a6SYan, Zheng return PTR_ERR(trans); 369f46b5a66SChristoph Hellwig 3705d4f98a2SYan Zheng leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 37166d7e7f0SArne Jansen 0, objectid, NULL, 0, 0, 0, 0); 3728e8a1e31SJosef Bacik if (IS_ERR(leaf)) { 3738e8a1e31SJosef Bacik ret = PTR_ERR(leaf); 3748e8a1e31SJosef Bacik goto fail; 3758e8a1e31SJosef Bacik } 376f46b5a66SChristoph Hellwig 3775d4f98a2SYan Zheng memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); 378f46b5a66SChristoph Hellwig btrfs_set_header_bytenr(leaf, leaf->start); 379f46b5a66SChristoph Hellwig btrfs_set_header_generation(leaf, trans->transid); 3805d4f98a2SYan Zheng btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 381f46b5a66SChristoph Hellwig btrfs_set_header_owner(leaf, objectid); 382f46b5a66SChristoph Hellwig 383f46b5a66SChristoph Hellwig write_extent_buffer(leaf, root->fs_info->fsid, 384f46b5a66SChristoph Hellwig (unsigned long)btrfs_header_fsid(leaf), 385f46b5a66SChristoph Hellwig BTRFS_FSID_SIZE); 3865d4f98a2SYan Zheng write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid, 3875d4f98a2SYan Zheng (unsigned long)btrfs_header_chunk_tree_uuid(leaf), 3885d4f98a2SYan Zheng BTRFS_UUID_SIZE); 389f46b5a66SChristoph Hellwig btrfs_mark_buffer_dirty(leaf); 390f46b5a66SChristoph Hellwig 391f46b5a66SChristoph Hellwig inode_item = &root_item.inode; 392f46b5a66SChristoph Hellwig memset(inode_item, 0, sizeof(*inode_item)); 393f46b5a66SChristoph Hellwig inode_item->generation = cpu_to_le64(1); 394f46b5a66SChristoph Hellwig inode_item->size = cpu_to_le64(3); 395f46b5a66SChristoph Hellwig inode_item->nlink = cpu_to_le32(1); 396a76a3cd4SYan Zheng inode_item->nbytes = cpu_to_le64(root->leafsize); 397f46b5a66SChristoph Hellwig inode_item->mode = cpu_to_le32(S_IFDIR | 0755); 398f46b5a66SChristoph Hellwig 39908fe4db1SLi Zefan root_item.flags = 0; 40008fe4db1SLi Zefan root_item.byte_limit = 0; 40108fe4db1SLi Zefan inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT); 40208fe4db1SLi Zefan 403f46b5a66SChristoph Hellwig btrfs_set_root_bytenr(&root_item, leaf->start); 40484234f3aSYan Zheng btrfs_set_root_generation(&root_item, trans->transid); 405f46b5a66SChristoph Hellwig btrfs_set_root_level(&root_item, 0); 406f46b5a66SChristoph Hellwig btrfs_set_root_refs(&root_item, 1); 40786b9f2ecSYan, Zheng btrfs_set_root_used(&root_item, leaf->len); 40880ff3856SYan Zheng btrfs_set_root_last_snapshot(&root_item, 0); 409f46b5a66SChristoph Hellwig 410f46b5a66SChristoph Hellwig memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress)); 411f46b5a66SChristoph Hellwig root_item.drop_level = 0; 412f46b5a66SChristoph Hellwig 413925baeddSChris Mason btrfs_tree_unlock(leaf); 414f46b5a66SChristoph Hellwig free_extent_buffer(leaf); 415f46b5a66SChristoph Hellwig leaf = NULL; 416f46b5a66SChristoph Hellwig 417f46b5a66SChristoph Hellwig btrfs_set_root_dirid(&root_item, new_dirid); 418f46b5a66SChristoph Hellwig 419f46b5a66SChristoph Hellwig key.objectid = objectid; 4205d4f98a2SYan Zheng key.offset = 0; 421f46b5a66SChristoph Hellwig btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); 422f46b5a66SChristoph Hellwig ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, 423f46b5a66SChristoph Hellwig &root_item); 424f46b5a66SChristoph Hellwig if (ret) 425f46b5a66SChristoph Hellwig goto fail; 426f46b5a66SChristoph Hellwig 42776dda93cSYan, Zheng key.offset = (u64)-1; 42876dda93cSYan, Zheng new_root = btrfs_read_fs_root_no_name(root->fs_info, &key); 42976dda93cSYan, Zheng BUG_ON(IS_ERR(new_root)); 43076dda93cSYan, Zheng 43176dda93cSYan, Zheng btrfs_record_root_in_trans(trans, new_root); 43276dda93cSYan, Zheng 433d82a6f1dSJosef Bacik ret = btrfs_create_subvol_root(trans, new_root, new_dirid); 434f46b5a66SChristoph Hellwig /* 435f46b5a66SChristoph Hellwig * insert the directory item 436f46b5a66SChristoph Hellwig */ 4373de4586cSChris Mason ret = btrfs_set_inode_index(dir, &index); 4383de4586cSChris Mason BUG_ON(ret); 4393de4586cSChris Mason 4403de4586cSChris Mason ret = btrfs_insert_dir_item(trans, root, 44116cdcec7SMiao Xie name, namelen, dir, &key, 4423de4586cSChris Mason BTRFS_FT_DIR, index); 443f46b5a66SChristoph Hellwig if (ret) 444f46b5a66SChristoph Hellwig goto fail; 4450660b5afSChris Mason 44652c26179SYan Zheng btrfs_i_size_write(dir, dir->i_size + namelen * 2); 44752c26179SYan Zheng ret = btrfs_update_inode(trans, root, dir); 44852c26179SYan Zheng BUG_ON(ret); 44952c26179SYan Zheng 4500660b5afSChris Mason ret = btrfs_add_root_ref(trans, root->fs_info->tree_root, 4514df27c4dSYan, Zheng objectid, root->root_key.objectid, 45233345d01SLi Zefan btrfs_ino(dir), index, name, namelen); 45376dda93cSYan, Zheng 4540660b5afSChris Mason BUG_ON(ret); 4550660b5afSChris Mason 45676dda93cSYan, Zheng d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry)); 457f46b5a66SChristoph Hellwig fail: 45872fd032eSSage Weil if (async_transid) { 45972fd032eSSage Weil *async_transid = trans->transid; 46072fd032eSSage Weil err = btrfs_commit_transaction_async(trans, root, 1); 46172fd032eSSage Weil } else { 46276dda93cSYan, Zheng err = btrfs_commit_transaction(trans, root); 46372fd032eSSage Weil } 464f46b5a66SChristoph Hellwig if (err && !ret) 465f46b5a66SChristoph Hellwig ret = err; 466f46b5a66SChristoph Hellwig return ret; 467f46b5a66SChristoph Hellwig } 468f46b5a66SChristoph Hellwig 46972fd032eSSage Weil static int create_snapshot(struct btrfs_root *root, struct dentry *dentry, 470b83cc969SLi Zefan char *name, int namelen, u64 *async_transid, 471b83cc969SLi Zefan bool readonly) 472f46b5a66SChristoph Hellwig { 4732e4bfab9SYan, Zheng struct inode *inode; 474f46b5a66SChristoph Hellwig struct btrfs_pending_snapshot *pending_snapshot; 475f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 4762e4bfab9SYan, Zheng int ret; 477f46b5a66SChristoph Hellwig 478f46b5a66SChristoph Hellwig if (!root->ref_cows) 479f46b5a66SChristoph Hellwig return -EINVAL; 480f46b5a66SChristoph Hellwig 481a22285a6SYan, Zheng pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS); 482a22285a6SYan, Zheng if (!pending_snapshot) 483a22285a6SYan, Zheng return -ENOMEM; 484a22285a6SYan, Zheng 485a22285a6SYan, Zheng btrfs_init_block_rsv(&pending_snapshot->block_rsv); 486a22285a6SYan, Zheng pending_snapshot->dentry = dentry; 487a22285a6SYan, Zheng pending_snapshot->root = root; 488b83cc969SLi Zefan pending_snapshot->readonly = readonly; 489a22285a6SYan, Zheng 490a22285a6SYan, Zheng trans = btrfs_start_transaction(root->fs_info->extent_root, 5); 491a22285a6SYan, Zheng if (IS_ERR(trans)) { 492a22285a6SYan, Zheng ret = PTR_ERR(trans); 493a22285a6SYan, Zheng goto fail; 494a22285a6SYan, Zheng } 495a22285a6SYan, Zheng 496a22285a6SYan, Zheng ret = btrfs_snap_reserve_metadata(trans, pending_snapshot); 497a22285a6SYan, Zheng BUG_ON(ret); 498a22285a6SYan, Zheng 4998351583eSJosef Bacik spin_lock(&root->fs_info->trans_lock); 500a22285a6SYan, Zheng list_add(&pending_snapshot->list, 501a22285a6SYan, Zheng &trans->transaction->pending_snapshots); 5028351583eSJosef Bacik spin_unlock(&root->fs_info->trans_lock); 50372fd032eSSage Weil if (async_transid) { 50472fd032eSSage Weil *async_transid = trans->transid; 50572fd032eSSage Weil ret = btrfs_commit_transaction_async(trans, 50672fd032eSSage Weil root->fs_info->extent_root, 1); 50772fd032eSSage Weil } else { 50872fd032eSSage Weil ret = btrfs_commit_transaction(trans, 50972fd032eSSage Weil root->fs_info->extent_root); 51072fd032eSSage Weil } 511a22285a6SYan, Zheng BUG_ON(ret); 512a22285a6SYan, Zheng 513a22285a6SYan, Zheng ret = pending_snapshot->error; 514f46b5a66SChristoph Hellwig if (ret) 5152e4bfab9SYan, Zheng goto fail; 516f46b5a66SChristoph Hellwig 51766b4ffd1SJosef Bacik ret = btrfs_orphan_cleanup(pending_snapshot->snap); 51866b4ffd1SJosef Bacik if (ret) 51966b4ffd1SJosef Bacik goto fail; 520f46b5a66SChristoph Hellwig 5212fbe8c8aSAl Viro inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry); 5222e4bfab9SYan, Zheng if (IS_ERR(inode)) { 5232e4bfab9SYan, Zheng ret = PTR_ERR(inode); 5242e4bfab9SYan, Zheng goto fail; 5252e4bfab9SYan, Zheng } 5262e4bfab9SYan, Zheng BUG_ON(!inode); 5272e4bfab9SYan, Zheng d_instantiate(dentry, inode); 5282e4bfab9SYan, Zheng ret = 0; 5292e4bfab9SYan, Zheng fail: 530a22285a6SYan, Zheng kfree(pending_snapshot); 531f46b5a66SChristoph Hellwig return ret; 532f46b5a66SChristoph Hellwig } 533f46b5a66SChristoph Hellwig 5344260f7c7SSage Weil /* copy of check_sticky in fs/namei.c() 5354260f7c7SSage Weil * It's inline, so penalty for filesystems that don't use sticky bit is 5364260f7c7SSage Weil * minimal. 5374260f7c7SSage Weil */ 5384260f7c7SSage Weil static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode) 5394260f7c7SSage Weil { 5404260f7c7SSage Weil uid_t fsuid = current_fsuid(); 5414260f7c7SSage Weil 5424260f7c7SSage Weil if (!(dir->i_mode & S_ISVTX)) 5434260f7c7SSage Weil return 0; 5444260f7c7SSage Weil if (inode->i_uid == fsuid) 5454260f7c7SSage Weil return 0; 5464260f7c7SSage Weil if (dir->i_uid == fsuid) 5474260f7c7SSage Weil return 0; 5484260f7c7SSage Weil return !capable(CAP_FOWNER); 5494260f7c7SSage Weil } 5504260f7c7SSage Weil 5514260f7c7SSage Weil /* copy of may_delete in fs/namei.c() 5524260f7c7SSage Weil * Check whether we can remove a link victim from directory dir, check 5534260f7c7SSage Weil * whether the type of victim is right. 5544260f7c7SSage Weil * 1. We can't do it if dir is read-only (done in permission()) 5554260f7c7SSage Weil * 2. We should have write and exec permissions on dir 5564260f7c7SSage Weil * 3. We can't remove anything from append-only dir 5574260f7c7SSage Weil * 4. We can't do anything with immutable dir (done in permission()) 5584260f7c7SSage Weil * 5. If the sticky bit on dir is set we should either 5594260f7c7SSage Weil * a. be owner of dir, or 5604260f7c7SSage Weil * b. be owner of victim, or 5614260f7c7SSage Weil * c. have CAP_FOWNER capability 5624260f7c7SSage Weil * 6. If the victim is append-only or immutable we can't do antyhing with 5634260f7c7SSage Weil * links pointing to it. 5644260f7c7SSage Weil * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. 5654260f7c7SSage Weil * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. 5664260f7c7SSage Weil * 9. We can't remove a root or mountpoint. 5674260f7c7SSage Weil * 10. We don't allow removal of NFS sillyrenamed files; it's handled by 5684260f7c7SSage Weil * nfs_async_unlink(). 5694260f7c7SSage Weil */ 5704260f7c7SSage Weil 5714260f7c7SSage Weil static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir) 5724260f7c7SSage Weil { 5734260f7c7SSage Weil int error; 5744260f7c7SSage Weil 5754260f7c7SSage Weil if (!victim->d_inode) 5764260f7c7SSage Weil return -ENOENT; 5774260f7c7SSage Weil 5784260f7c7SSage Weil BUG_ON(victim->d_parent->d_inode != dir); 5794260f7c7SSage Weil audit_inode_child(victim, dir); 5804260f7c7SSage Weil 5814260f7c7SSage Weil error = inode_permission(dir, MAY_WRITE | MAY_EXEC); 5824260f7c7SSage Weil if (error) 5834260f7c7SSage Weil return error; 5844260f7c7SSage Weil if (IS_APPEND(dir)) 5854260f7c7SSage Weil return -EPERM; 5864260f7c7SSage Weil if (btrfs_check_sticky(dir, victim->d_inode)|| 5874260f7c7SSage Weil IS_APPEND(victim->d_inode)|| 5884260f7c7SSage Weil IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode)) 5894260f7c7SSage Weil return -EPERM; 5904260f7c7SSage Weil if (isdir) { 5914260f7c7SSage Weil if (!S_ISDIR(victim->d_inode->i_mode)) 5924260f7c7SSage Weil return -ENOTDIR; 5934260f7c7SSage Weil if (IS_ROOT(victim)) 5944260f7c7SSage Weil return -EBUSY; 5954260f7c7SSage Weil } else if (S_ISDIR(victim->d_inode->i_mode)) 5964260f7c7SSage Weil return -EISDIR; 5974260f7c7SSage Weil if (IS_DEADDIR(dir)) 5984260f7c7SSage Weil return -ENOENT; 5994260f7c7SSage Weil if (victim->d_flags & DCACHE_NFSFS_RENAMED) 6004260f7c7SSage Weil return -EBUSY; 6014260f7c7SSage Weil return 0; 6024260f7c7SSage Weil } 6034260f7c7SSage Weil 604cb8e7090SChristoph Hellwig /* copy of may_create in fs/namei.c() */ 605cb8e7090SChristoph Hellwig static inline int btrfs_may_create(struct inode *dir, struct dentry *child) 606cb8e7090SChristoph Hellwig { 607cb8e7090SChristoph Hellwig if (child->d_inode) 608cb8e7090SChristoph Hellwig return -EEXIST; 609cb8e7090SChristoph Hellwig if (IS_DEADDIR(dir)) 610cb8e7090SChristoph Hellwig return -ENOENT; 611cb8e7090SChristoph Hellwig return inode_permission(dir, MAY_WRITE | MAY_EXEC); 612cb8e7090SChristoph Hellwig } 613cb8e7090SChristoph Hellwig 614cb8e7090SChristoph Hellwig /* 615cb8e7090SChristoph Hellwig * Create a new subvolume below @parent. This is largely modeled after 616cb8e7090SChristoph Hellwig * sys_mkdirat and vfs_mkdir, but we only do a single component lookup 617cb8e7090SChristoph Hellwig * inside this filesystem so it's quite a bit simpler. 618cb8e7090SChristoph Hellwig */ 61976dda93cSYan, Zheng static noinline int btrfs_mksubvol(struct path *parent, 62076dda93cSYan, Zheng char *name, int namelen, 62172fd032eSSage Weil struct btrfs_root *snap_src, 622b83cc969SLi Zefan u64 *async_transid, bool readonly) 623cb8e7090SChristoph Hellwig { 62476dda93cSYan, Zheng struct inode *dir = parent->dentry->d_inode; 625cb8e7090SChristoph Hellwig struct dentry *dentry; 626cb8e7090SChristoph Hellwig int error; 627cb8e7090SChristoph Hellwig 62876dda93cSYan, Zheng mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT); 629cb8e7090SChristoph Hellwig 630cb8e7090SChristoph Hellwig dentry = lookup_one_len(name, parent->dentry, namelen); 631cb8e7090SChristoph Hellwig error = PTR_ERR(dentry); 632cb8e7090SChristoph Hellwig if (IS_ERR(dentry)) 633cb8e7090SChristoph Hellwig goto out_unlock; 634cb8e7090SChristoph Hellwig 635cb8e7090SChristoph Hellwig error = -EEXIST; 636cb8e7090SChristoph Hellwig if (dentry->d_inode) 637cb8e7090SChristoph Hellwig goto out_dput; 638cb8e7090SChristoph Hellwig 639cb8e7090SChristoph Hellwig error = mnt_want_write(parent->mnt); 640cb8e7090SChristoph Hellwig if (error) 641cb8e7090SChristoph Hellwig goto out_dput; 642cb8e7090SChristoph Hellwig 64376dda93cSYan, Zheng error = btrfs_may_create(dir, dentry); 644cb8e7090SChristoph Hellwig if (error) 645cb8e7090SChristoph Hellwig goto out_drop_write; 646cb8e7090SChristoph Hellwig 64776dda93cSYan, Zheng down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem); 64876dda93cSYan, Zheng 64976dda93cSYan, Zheng if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) 65076dda93cSYan, Zheng goto out_up_read; 65176dda93cSYan, Zheng 6523de4586cSChris Mason if (snap_src) { 65372fd032eSSage Weil error = create_snapshot(snap_src, dentry, 654b83cc969SLi Zefan name, namelen, async_transid, readonly); 6553de4586cSChris Mason } else { 65676dda93cSYan, Zheng error = create_subvol(BTRFS_I(dir)->root, dentry, 65772fd032eSSage Weil name, namelen, async_transid); 6583de4586cSChris Mason } 65976dda93cSYan, Zheng if (!error) 66076dda93cSYan, Zheng fsnotify_mkdir(dir, dentry); 66176dda93cSYan, Zheng out_up_read: 66276dda93cSYan, Zheng up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem); 663cb8e7090SChristoph Hellwig out_drop_write: 664cb8e7090SChristoph Hellwig mnt_drop_write(parent->mnt); 665cb8e7090SChristoph Hellwig out_dput: 666cb8e7090SChristoph Hellwig dput(dentry); 667cb8e7090SChristoph Hellwig out_unlock: 66876dda93cSYan, Zheng mutex_unlock(&dir->i_mutex); 669cb8e7090SChristoph Hellwig return error; 670cb8e7090SChristoph Hellwig } 671cb8e7090SChristoph Hellwig 6724cb5300bSChris Mason /* 6734cb5300bSChris Mason * When we're defragging a range, we don't want to kick it off again 6744cb5300bSChris Mason * if it is really just waiting for delalloc to send it down. 6754cb5300bSChris Mason * If we find a nice big extent or delalloc range for the bytes in the 6764cb5300bSChris Mason * file you want to defrag, we return 0 to let you know to skip this 6774cb5300bSChris Mason * part of the file 6784cb5300bSChris Mason */ 6794cb5300bSChris Mason static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh) 6804cb5300bSChris Mason { 6814cb5300bSChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 6824cb5300bSChris Mason struct extent_map *em = NULL; 6834cb5300bSChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 6844cb5300bSChris Mason u64 end; 6854cb5300bSChris Mason 6864cb5300bSChris Mason read_lock(&em_tree->lock); 6874cb5300bSChris Mason em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE); 6884cb5300bSChris Mason read_unlock(&em_tree->lock); 6894cb5300bSChris Mason 6904cb5300bSChris Mason if (em) { 6914cb5300bSChris Mason end = extent_map_end(em); 6924cb5300bSChris Mason free_extent_map(em); 6934cb5300bSChris Mason if (end - offset > thresh) 6944cb5300bSChris Mason return 0; 6954cb5300bSChris Mason } 6964cb5300bSChris Mason /* if we already have a nice delalloc here, just stop */ 6974cb5300bSChris Mason thresh /= 2; 6984cb5300bSChris Mason end = count_range_bits(io_tree, &offset, offset + thresh, 6994cb5300bSChris Mason thresh, EXTENT_DELALLOC, 1); 7004cb5300bSChris Mason if (end >= thresh) 7014cb5300bSChris Mason return 0; 7024cb5300bSChris Mason return 1; 7034cb5300bSChris Mason } 7044cb5300bSChris Mason 7054cb5300bSChris Mason /* 7064cb5300bSChris Mason * helper function to walk through a file and find extents 7074cb5300bSChris Mason * newer than a specific transid, and smaller than thresh. 7084cb5300bSChris Mason * 7094cb5300bSChris Mason * This is used by the defragging code to find new and small 7104cb5300bSChris Mason * extents 7114cb5300bSChris Mason */ 7124cb5300bSChris Mason static int find_new_extents(struct btrfs_root *root, 7134cb5300bSChris Mason struct inode *inode, u64 newer_than, 7144cb5300bSChris Mason u64 *off, int thresh) 7154cb5300bSChris Mason { 7164cb5300bSChris Mason struct btrfs_path *path; 7174cb5300bSChris Mason struct btrfs_key min_key; 7184cb5300bSChris Mason struct btrfs_key max_key; 7194cb5300bSChris Mason struct extent_buffer *leaf; 7204cb5300bSChris Mason struct btrfs_file_extent_item *extent; 7214cb5300bSChris Mason int type; 7224cb5300bSChris Mason int ret; 723a4689d2bSDavid Sterba u64 ino = btrfs_ino(inode); 7244cb5300bSChris Mason 7254cb5300bSChris Mason path = btrfs_alloc_path(); 7264cb5300bSChris Mason if (!path) 7274cb5300bSChris Mason return -ENOMEM; 7284cb5300bSChris Mason 729a4689d2bSDavid Sterba min_key.objectid = ino; 7304cb5300bSChris Mason min_key.type = BTRFS_EXTENT_DATA_KEY; 7314cb5300bSChris Mason min_key.offset = *off; 7324cb5300bSChris Mason 733a4689d2bSDavid Sterba max_key.objectid = ino; 7344cb5300bSChris Mason max_key.type = (u8)-1; 7354cb5300bSChris Mason max_key.offset = (u64)-1; 7364cb5300bSChris Mason 7374cb5300bSChris Mason path->keep_locks = 1; 7384cb5300bSChris Mason 7394cb5300bSChris Mason while(1) { 7404cb5300bSChris Mason ret = btrfs_search_forward(root, &min_key, &max_key, 7414cb5300bSChris Mason path, 0, newer_than); 7424cb5300bSChris Mason if (ret != 0) 7434cb5300bSChris Mason goto none; 744a4689d2bSDavid Sterba if (min_key.objectid != ino) 7454cb5300bSChris Mason goto none; 7464cb5300bSChris Mason if (min_key.type != BTRFS_EXTENT_DATA_KEY) 7474cb5300bSChris Mason goto none; 7484cb5300bSChris Mason 7494cb5300bSChris Mason leaf = path->nodes[0]; 7504cb5300bSChris Mason extent = btrfs_item_ptr(leaf, path->slots[0], 7514cb5300bSChris Mason struct btrfs_file_extent_item); 7524cb5300bSChris Mason 7534cb5300bSChris Mason type = btrfs_file_extent_type(leaf, extent); 7544cb5300bSChris Mason if (type == BTRFS_FILE_EXTENT_REG && 7554cb5300bSChris Mason btrfs_file_extent_num_bytes(leaf, extent) < thresh && 7564cb5300bSChris Mason check_defrag_in_cache(inode, min_key.offset, thresh)) { 7574cb5300bSChris Mason *off = min_key.offset; 7584cb5300bSChris Mason btrfs_free_path(path); 7594cb5300bSChris Mason return 0; 7604cb5300bSChris Mason } 7614cb5300bSChris Mason 7624cb5300bSChris Mason if (min_key.offset == (u64)-1) 7634cb5300bSChris Mason goto none; 7644cb5300bSChris Mason 7654cb5300bSChris Mason min_key.offset++; 7664cb5300bSChris Mason btrfs_release_path(path); 7674cb5300bSChris Mason } 7684cb5300bSChris Mason none: 7694cb5300bSChris Mason btrfs_free_path(path); 7704cb5300bSChris Mason return -ENOENT; 7714cb5300bSChris Mason } 7724cb5300bSChris Mason 773940100a4SChris Mason static int should_defrag_range(struct inode *inode, u64 start, u64 len, 7741e701a32SChris Mason int thresh, u64 *last_len, u64 *skip, 7751e701a32SChris Mason u64 *defrag_end) 776940100a4SChris Mason { 777940100a4SChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 778940100a4SChris Mason struct extent_map *em = NULL; 779940100a4SChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 780940100a4SChris Mason int ret = 1; 781940100a4SChris Mason 782940100a4SChris Mason /* 783008873eaSLi Zefan * make sure that once we start defragging an extent, we keep on 784940100a4SChris Mason * defragging it 785940100a4SChris Mason */ 786940100a4SChris Mason if (start < *defrag_end) 787940100a4SChris Mason return 1; 788940100a4SChris Mason 789940100a4SChris Mason *skip = 0; 790940100a4SChris Mason 791940100a4SChris Mason /* 792940100a4SChris Mason * hopefully we have this extent in the tree already, try without 793940100a4SChris Mason * the full extent lock 794940100a4SChris Mason */ 795940100a4SChris Mason read_lock(&em_tree->lock); 796940100a4SChris Mason em = lookup_extent_mapping(em_tree, start, len); 797940100a4SChris Mason read_unlock(&em_tree->lock); 798940100a4SChris Mason 799940100a4SChris Mason if (!em) { 800940100a4SChris Mason /* get the big lock and read metadata off disk */ 801940100a4SChris Mason lock_extent(io_tree, start, start + len - 1, GFP_NOFS); 802940100a4SChris Mason em = btrfs_get_extent(inode, NULL, 0, start, len, 0); 803940100a4SChris Mason unlock_extent(io_tree, start, start + len - 1, GFP_NOFS); 804940100a4SChris Mason 8056cf8bfbfSDan Carpenter if (IS_ERR(em)) 806940100a4SChris Mason return 0; 807940100a4SChris Mason } 808940100a4SChris Mason 809940100a4SChris Mason /* this will cover holes, and inline extents */ 810940100a4SChris Mason if (em->block_start >= EXTENT_MAP_LAST_BYTE) 811940100a4SChris Mason ret = 0; 812940100a4SChris Mason 813940100a4SChris Mason /* 814940100a4SChris Mason * we hit a real extent, if it is big don't bother defragging it again 815940100a4SChris Mason */ 8161e701a32SChris Mason if ((*last_len == 0 || *last_len >= thresh) && em->len >= thresh) 817940100a4SChris Mason ret = 0; 818940100a4SChris Mason 819940100a4SChris Mason /* 820940100a4SChris Mason * last_len ends up being a counter of how many bytes we've defragged. 821940100a4SChris Mason * every time we choose not to defrag an extent, we reset *last_len 822940100a4SChris Mason * so that the next tiny extent will force a defrag. 823940100a4SChris Mason * 824940100a4SChris Mason * The end result of this is that tiny extents before a single big 825940100a4SChris Mason * extent will force at least part of that big extent to be defragged. 826940100a4SChris Mason */ 827940100a4SChris Mason if (ret) { 828940100a4SChris Mason *defrag_end = extent_map_end(em); 829940100a4SChris Mason } else { 830940100a4SChris Mason *last_len = 0; 831940100a4SChris Mason *skip = extent_map_end(em); 832940100a4SChris Mason *defrag_end = 0; 833940100a4SChris Mason } 834940100a4SChris Mason 835940100a4SChris Mason free_extent_map(em); 836940100a4SChris Mason return ret; 837940100a4SChris Mason } 838940100a4SChris Mason 8394cb5300bSChris Mason /* 8404cb5300bSChris Mason * it doesn't do much good to defrag one or two pages 8414cb5300bSChris Mason * at a time. This pulls in a nice chunk of pages 8424cb5300bSChris Mason * to COW and defrag. 8434cb5300bSChris Mason * 8444cb5300bSChris Mason * It also makes sure the delalloc code has enough 8454cb5300bSChris Mason * dirty data to avoid making new small extents as part 8464cb5300bSChris Mason * of the defrag 8474cb5300bSChris Mason * 8484cb5300bSChris Mason * It's a good idea to start RA on this range 8494cb5300bSChris Mason * before calling this. 8504cb5300bSChris Mason */ 8514cb5300bSChris Mason static int cluster_pages_for_defrag(struct inode *inode, 8524cb5300bSChris Mason struct page **pages, 8534cb5300bSChris Mason unsigned long start_index, 8544cb5300bSChris Mason int num_pages) 855f46b5a66SChristoph Hellwig { 8564cb5300bSChris Mason unsigned long file_end; 8574cb5300bSChris Mason u64 isize = i_size_read(inode); 858f46b5a66SChristoph Hellwig u64 page_start; 859f46b5a66SChristoph Hellwig u64 page_end; 8604cb5300bSChris Mason int ret; 8614cb5300bSChris Mason int i; 8624cb5300bSChris Mason int i_done; 8634cb5300bSChris Mason struct btrfs_ordered_extent *ordered; 8644cb5300bSChris Mason struct extent_state *cached_state = NULL; 865600a45e1SMiao Xie struct extent_io_tree *tree; 8663b16a4e3SJosef Bacik gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping); 8674cb5300bSChris Mason 8684cb5300bSChris Mason if (isize == 0) 8694cb5300bSChris Mason return 0; 8704cb5300bSChris Mason file_end = (isize - 1) >> PAGE_CACHE_SHIFT; 8714cb5300bSChris Mason 8724cb5300bSChris Mason ret = btrfs_delalloc_reserve_space(inode, 8734cb5300bSChris Mason num_pages << PAGE_CACHE_SHIFT); 8744cb5300bSChris Mason if (ret) 8754cb5300bSChris Mason return ret; 8764cb5300bSChris Mason i_done = 0; 877600a45e1SMiao Xie tree = &BTRFS_I(inode)->io_tree; 8784cb5300bSChris Mason 8794cb5300bSChris Mason /* step one, lock all the pages */ 8804cb5300bSChris Mason for (i = 0; i < num_pages; i++) { 8814cb5300bSChris Mason struct page *page; 882600a45e1SMiao Xie again: 883a94733d0SJosef Bacik page = find_or_create_page(inode->i_mapping, 8843b16a4e3SJosef Bacik start_index + i, mask); 8854cb5300bSChris Mason if (!page) 8864cb5300bSChris Mason break; 8874cb5300bSChris Mason 888600a45e1SMiao Xie page_start = page_offset(page); 889600a45e1SMiao Xie page_end = page_start + PAGE_CACHE_SIZE - 1; 890600a45e1SMiao Xie while (1) { 891600a45e1SMiao Xie lock_extent(tree, page_start, page_end, GFP_NOFS); 892600a45e1SMiao Xie ordered = btrfs_lookup_ordered_extent(inode, 893600a45e1SMiao Xie page_start); 894600a45e1SMiao Xie unlock_extent(tree, page_start, page_end, GFP_NOFS); 895600a45e1SMiao Xie if (!ordered) 896600a45e1SMiao Xie break; 897600a45e1SMiao Xie 898600a45e1SMiao Xie unlock_page(page); 899600a45e1SMiao Xie btrfs_start_ordered_extent(inode, ordered, 1); 900600a45e1SMiao Xie btrfs_put_ordered_extent(ordered); 901600a45e1SMiao Xie lock_page(page); 902600a45e1SMiao Xie } 903600a45e1SMiao Xie 9044cb5300bSChris Mason if (!PageUptodate(page)) { 9054cb5300bSChris Mason btrfs_readpage(NULL, page); 9064cb5300bSChris Mason lock_page(page); 9074cb5300bSChris Mason if (!PageUptodate(page)) { 9084cb5300bSChris Mason unlock_page(page); 9094cb5300bSChris Mason page_cache_release(page); 9104cb5300bSChris Mason ret = -EIO; 9114cb5300bSChris Mason break; 9124cb5300bSChris Mason } 9134cb5300bSChris Mason } 914600a45e1SMiao Xie 9154cb5300bSChris Mason isize = i_size_read(inode); 9164cb5300bSChris Mason file_end = (isize - 1) >> PAGE_CACHE_SHIFT; 917600a45e1SMiao Xie if (!isize || page->index > file_end) { 9184cb5300bSChris Mason /* whoops, we blew past eof, skip this page */ 9194cb5300bSChris Mason unlock_page(page); 9204cb5300bSChris Mason page_cache_release(page); 9214cb5300bSChris Mason break; 9224cb5300bSChris Mason } 923600a45e1SMiao Xie 924600a45e1SMiao Xie if (page->mapping != inode->i_mapping) { 925600a45e1SMiao Xie unlock_page(page); 926600a45e1SMiao Xie page_cache_release(page); 927600a45e1SMiao Xie goto again; 928600a45e1SMiao Xie } 929600a45e1SMiao Xie 9304cb5300bSChris Mason pages[i] = page; 9314cb5300bSChris Mason i_done++; 9324cb5300bSChris Mason } 9334cb5300bSChris Mason if (!i_done || ret) 9344cb5300bSChris Mason goto out; 9354cb5300bSChris Mason 9364cb5300bSChris Mason if (!(inode->i_sb->s_flags & MS_ACTIVE)) 9374cb5300bSChris Mason goto out; 9384cb5300bSChris Mason 9394cb5300bSChris Mason /* 9404cb5300bSChris Mason * so now we have a nice long stream of locked 9414cb5300bSChris Mason * and up to date pages, lets wait on them 9424cb5300bSChris Mason */ 9434cb5300bSChris Mason for (i = 0; i < i_done; i++) 9444cb5300bSChris Mason wait_on_page_writeback(pages[i]); 9454cb5300bSChris Mason 9464cb5300bSChris Mason page_start = page_offset(pages[0]); 9474cb5300bSChris Mason page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE; 9484cb5300bSChris Mason 9494cb5300bSChris Mason lock_extent_bits(&BTRFS_I(inode)->io_tree, 9504cb5300bSChris Mason page_start, page_end - 1, 0, &cached_state, 9514cb5300bSChris Mason GFP_NOFS); 9524cb5300bSChris Mason clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, 9534cb5300bSChris Mason page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC | 9544cb5300bSChris Mason EXTENT_DO_ACCOUNTING, 0, 0, &cached_state, 9554cb5300bSChris Mason GFP_NOFS); 9564cb5300bSChris Mason 9574cb5300bSChris Mason if (i_done != num_pages) { 9589e0baf60SJosef Bacik spin_lock(&BTRFS_I(inode)->lock); 9599e0baf60SJosef Bacik BTRFS_I(inode)->outstanding_extents++; 9609e0baf60SJosef Bacik spin_unlock(&BTRFS_I(inode)->lock); 9614cb5300bSChris Mason btrfs_delalloc_release_space(inode, 9624cb5300bSChris Mason (num_pages - i_done) << PAGE_CACHE_SHIFT); 9634cb5300bSChris Mason } 9644cb5300bSChris Mason 9654cb5300bSChris Mason 9664cb5300bSChris Mason btrfs_set_extent_delalloc(inode, page_start, page_end - 1, 9674cb5300bSChris Mason &cached_state); 9684cb5300bSChris Mason 9694cb5300bSChris Mason unlock_extent_cached(&BTRFS_I(inode)->io_tree, 9704cb5300bSChris Mason page_start, page_end - 1, &cached_state, 9714cb5300bSChris Mason GFP_NOFS); 9724cb5300bSChris Mason 9734cb5300bSChris Mason for (i = 0; i < i_done; i++) { 9744cb5300bSChris Mason clear_page_dirty_for_io(pages[i]); 9754cb5300bSChris Mason ClearPageChecked(pages[i]); 9764cb5300bSChris Mason set_page_extent_mapped(pages[i]); 9774cb5300bSChris Mason set_page_dirty(pages[i]); 9784cb5300bSChris Mason unlock_page(pages[i]); 9794cb5300bSChris Mason page_cache_release(pages[i]); 9804cb5300bSChris Mason } 9814cb5300bSChris Mason return i_done; 9824cb5300bSChris Mason out: 9834cb5300bSChris Mason for (i = 0; i < i_done; i++) { 9844cb5300bSChris Mason unlock_page(pages[i]); 9854cb5300bSChris Mason page_cache_release(pages[i]); 9864cb5300bSChris Mason } 9874cb5300bSChris Mason btrfs_delalloc_release_space(inode, num_pages << PAGE_CACHE_SHIFT); 9884cb5300bSChris Mason return ret; 9894cb5300bSChris Mason 9904cb5300bSChris Mason } 9914cb5300bSChris Mason 9924cb5300bSChris Mason int btrfs_defrag_file(struct inode *inode, struct file *file, 9934cb5300bSChris Mason struct btrfs_ioctl_defrag_range_args *range, 9944cb5300bSChris Mason u64 newer_than, unsigned long max_to_defrag) 9954cb5300bSChris Mason { 9964cb5300bSChris Mason struct btrfs_root *root = BTRFS_I(inode)->root; 9974cb5300bSChris Mason struct btrfs_super_block *disk_super; 9984cb5300bSChris Mason struct file_ra_state *ra = NULL; 9994cb5300bSChris Mason unsigned long last_index; 1000151a31b2SLi Zefan u64 isize = i_size_read(inode); 10014cb5300bSChris Mason u64 features; 1002940100a4SChris Mason u64 last_len = 0; 1003940100a4SChris Mason u64 skip = 0; 1004940100a4SChris Mason u64 defrag_end = 0; 10054cb5300bSChris Mason u64 newer_off = range->start; 1006f46b5a66SChristoph Hellwig unsigned long i; 1007008873eaSLi Zefan unsigned long ra_index = 0; 1008f46b5a66SChristoph Hellwig int ret; 10094cb5300bSChris Mason int defrag_count = 0; 10101a419d85SLi Zefan int compress_type = BTRFS_COMPRESS_ZLIB; 10114cb5300bSChris Mason int extent_thresh = range->extent_thresh; 1012008873eaSLi Zefan int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT; 1013008873eaSLi Zefan int cluster = max_cluster; 10144cb5300bSChris Mason u64 new_align = ~((u64)128 * 1024 - 1); 10154cb5300bSChris Mason struct page **pages = NULL; 10164cb5300bSChris Mason 10174cb5300bSChris Mason if (extent_thresh == 0) 10184cb5300bSChris Mason extent_thresh = 256 * 1024; 10191a419d85SLi Zefan 10201a419d85SLi Zefan if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) { 10211a419d85SLi Zefan if (range->compress_type > BTRFS_COMPRESS_TYPES) 10221a419d85SLi Zefan return -EINVAL; 10231a419d85SLi Zefan if (range->compress_type) 10241a419d85SLi Zefan compress_type = range->compress_type; 10251a419d85SLi Zefan } 1026f46b5a66SChristoph Hellwig 1027151a31b2SLi Zefan if (isize == 0) 1028940100a4SChris Mason return 0; 1029f46b5a66SChristoph Hellwig 10304cb5300bSChris Mason /* 10314cb5300bSChris Mason * if we were not given a file, allocate a readahead 10324cb5300bSChris Mason * context 10334cb5300bSChris Mason */ 10344cb5300bSChris Mason if (!file) { 10354cb5300bSChris Mason ra = kzalloc(sizeof(*ra), GFP_NOFS); 10364cb5300bSChris Mason if (!ra) 10374cb5300bSChris Mason return -ENOMEM; 10384cb5300bSChris Mason file_ra_state_init(ra, inode->i_mapping); 10394cb5300bSChris Mason } else { 10404cb5300bSChris Mason ra = &file->f_ra; 10414cb5300bSChris Mason } 10424cb5300bSChris Mason 1043008873eaSLi Zefan pages = kmalloc(sizeof(struct page *) * max_cluster, 10444cb5300bSChris Mason GFP_NOFS); 10454cb5300bSChris Mason if (!pages) { 10464cb5300bSChris Mason ret = -ENOMEM; 10474cb5300bSChris Mason goto out_ra; 10484cb5300bSChris Mason } 10494cb5300bSChris Mason 10504cb5300bSChris Mason /* find the last page to defrag */ 10511e701a32SChris Mason if (range->start + range->len > range->start) { 1052151a31b2SLi Zefan last_index = min_t(u64, isize - 1, 10531e701a32SChris Mason range->start + range->len - 1) >> PAGE_CACHE_SHIFT; 10541e701a32SChris Mason } else { 1055151a31b2SLi Zefan last_index = (isize - 1) >> PAGE_CACHE_SHIFT; 10561e701a32SChris Mason } 10571e701a32SChris Mason 10584cb5300bSChris Mason if (newer_than) { 10594cb5300bSChris Mason ret = find_new_extents(root, inode, newer_than, 10604cb5300bSChris Mason &newer_off, 64 * 1024); 10614cb5300bSChris Mason if (!ret) { 10624cb5300bSChris Mason range->start = newer_off; 10634cb5300bSChris Mason /* 10644cb5300bSChris Mason * we always align our defrag to help keep 10654cb5300bSChris Mason * the extents in the file evenly spaced 10664cb5300bSChris Mason */ 10674cb5300bSChris Mason i = (newer_off & new_align) >> PAGE_CACHE_SHIFT; 10684cb5300bSChris Mason } else 10694cb5300bSChris Mason goto out_ra; 10704cb5300bSChris Mason } else { 10711e701a32SChris Mason i = range->start >> PAGE_CACHE_SHIFT; 10724cb5300bSChris Mason } 10734cb5300bSChris Mason if (!max_to_defrag) 10747ec31b54SLiu Bo max_to_defrag = last_index + 1; 10754cb5300bSChris Mason 10762a0f7f57SLi Zefan /* 10772a0f7f57SLi Zefan * make writeback starts from i, so the defrag range can be 10782a0f7f57SLi Zefan * written sequentially. 10792a0f7f57SLi Zefan */ 10802a0f7f57SLi Zefan if (i < inode->i_mapping->writeback_index) 10812a0f7f57SLi Zefan inode->i_mapping->writeback_index = i; 10822a0f7f57SLi Zefan 1083f7f43cc8SChris Mason while (i <= last_index && defrag_count < max_to_defrag && 1084f7f43cc8SChris Mason (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >> 1085f7f43cc8SChris Mason PAGE_CACHE_SHIFT)) { 10864cb5300bSChris Mason /* 10874cb5300bSChris Mason * make sure we stop running if someone unmounts 10884cb5300bSChris Mason * the FS 10894cb5300bSChris Mason */ 10904cb5300bSChris Mason if (!(inode->i_sb->s_flags & MS_ACTIVE)) 10914cb5300bSChris Mason break; 10924cb5300bSChris Mason 10934cb5300bSChris Mason if (!newer_than && 10944cb5300bSChris Mason !should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT, 10951e701a32SChris Mason PAGE_CACHE_SIZE, 10964cb5300bSChris Mason extent_thresh, 10971e701a32SChris Mason &last_len, &skip, 1098940100a4SChris Mason &defrag_end)) { 1099940100a4SChris Mason unsigned long next; 1100940100a4SChris Mason /* 1101940100a4SChris Mason * the should_defrag function tells us how much to skip 1102940100a4SChris Mason * bump our counter by the suggested amount 1103940100a4SChris Mason */ 1104940100a4SChris Mason next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT; 1105940100a4SChris Mason i = max(i + 1, next); 1106940100a4SChris Mason continue; 1107940100a4SChris Mason } 1108008873eaSLi Zefan 1109008873eaSLi Zefan if (!newer_than) { 1110008873eaSLi Zefan cluster = (PAGE_CACHE_ALIGN(defrag_end) >> 1111008873eaSLi Zefan PAGE_CACHE_SHIFT) - i; 1112008873eaSLi Zefan cluster = min(cluster, max_cluster); 1113008873eaSLi Zefan } else { 1114008873eaSLi Zefan cluster = max_cluster; 1115008873eaSLi Zefan } 1116008873eaSLi Zefan 11171e701a32SChris Mason if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) 11181a419d85SLi Zefan BTRFS_I(inode)->force_compress = compress_type; 1119940100a4SChris Mason 1120008873eaSLi Zefan if (i + cluster > ra_index) { 1121008873eaSLi Zefan ra_index = max(i, ra_index); 1122008873eaSLi Zefan btrfs_force_ra(inode->i_mapping, ra, file, ra_index, 1123008873eaSLi Zefan cluster); 1124008873eaSLi Zefan ra_index += max_cluster; 1125008873eaSLi Zefan } 11264cb5300bSChris Mason 1127008873eaSLi Zefan ret = cluster_pages_for_defrag(inode, pages, i, cluster); 11284cb5300bSChris Mason if (ret < 0) 11294cb5300bSChris Mason goto out_ra; 11304cb5300bSChris Mason 11314cb5300bSChris Mason defrag_count += ret; 11324cb5300bSChris Mason balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret); 11334cb5300bSChris Mason 11344cb5300bSChris Mason if (newer_than) { 11354cb5300bSChris Mason if (newer_off == (u64)-1) 11364cb5300bSChris Mason break; 11374cb5300bSChris Mason 11384cb5300bSChris Mason newer_off = max(newer_off + 1, 11394cb5300bSChris Mason (u64)i << PAGE_CACHE_SHIFT); 11404cb5300bSChris Mason 11414cb5300bSChris Mason ret = find_new_extents(root, inode, 11424cb5300bSChris Mason newer_than, &newer_off, 11434cb5300bSChris Mason 64 * 1024); 11444cb5300bSChris Mason if (!ret) { 11454cb5300bSChris Mason range->start = newer_off; 11464cb5300bSChris Mason i = (newer_off & new_align) >> PAGE_CACHE_SHIFT; 11474cb5300bSChris Mason } else { 11484cb5300bSChris Mason break; 1149940100a4SChris Mason } 11504cb5300bSChris Mason } else { 1151008873eaSLi Zefan if (ret > 0) { 1152cbcc8326SLi Zefan i += ret; 1153008873eaSLi Zefan last_len += ret << PAGE_CACHE_SHIFT; 1154008873eaSLi Zefan } else { 1155940100a4SChris Mason i++; 1156008873eaSLi Zefan last_len = 0; 1157008873eaSLi Zefan } 1158f46b5a66SChristoph Hellwig } 11594cb5300bSChris Mason } 1160f46b5a66SChristoph Hellwig 11611e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) 11621e701a32SChris Mason filemap_flush(inode->i_mapping); 11631e701a32SChris Mason 11641e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 11651e701a32SChris Mason /* the filemap_flush will queue IO into the worker threads, but 11661e701a32SChris Mason * we have to make sure the IO is actually started and that 11671e701a32SChris Mason * ordered extents get created before we return 11681e701a32SChris Mason */ 11691e701a32SChris Mason atomic_inc(&root->fs_info->async_submit_draining); 11701e701a32SChris Mason while (atomic_read(&root->fs_info->nr_async_submits) || 11711e701a32SChris Mason atomic_read(&root->fs_info->async_delalloc_pages)) { 11721e701a32SChris Mason wait_event(root->fs_info->async_submit_wait, 11731e701a32SChris Mason (atomic_read(&root->fs_info->nr_async_submits) == 0 && 11741e701a32SChris Mason atomic_read(&root->fs_info->async_delalloc_pages) == 0)); 11751e701a32SChris Mason } 11761e701a32SChris Mason atomic_dec(&root->fs_info->async_submit_draining); 11771e701a32SChris Mason 11781e701a32SChris Mason mutex_lock(&inode->i_mutex); 1179261507a0SLi Zefan BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE; 11801e701a32SChris Mason mutex_unlock(&inode->i_mutex); 11811e701a32SChris Mason } 11821e701a32SChris Mason 11836c41761fSDavid Sterba disk_super = root->fs_info->super_copy; 11841a419d85SLi Zefan features = btrfs_super_incompat_flags(disk_super); 11851a419d85SLi Zefan if (range->compress_type == BTRFS_COMPRESS_LZO) { 11861a419d85SLi Zefan features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; 11871a419d85SLi Zefan btrfs_set_super_incompat_flags(disk_super, features); 11881a419d85SLi Zefan } 11891a419d85SLi Zefan 119060ccf82fSDiego Calleja ret = defrag_count; 1191940100a4SChris Mason 11924cb5300bSChris Mason out_ra: 11934cb5300bSChris Mason if (!file) 11944cb5300bSChris Mason kfree(ra); 11954cb5300bSChris Mason kfree(pages); 1196940100a4SChris Mason return ret; 1197f46b5a66SChristoph Hellwig } 1198f46b5a66SChristoph Hellwig 119976dda93cSYan, Zheng static noinline int btrfs_ioctl_resize(struct btrfs_root *root, 120076dda93cSYan, Zheng void __user *arg) 1201f46b5a66SChristoph Hellwig { 1202f46b5a66SChristoph Hellwig u64 new_size; 1203f46b5a66SChristoph Hellwig u64 old_size; 1204f46b5a66SChristoph Hellwig u64 devid = 1; 1205f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 1206f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 1207f46b5a66SChristoph Hellwig struct btrfs_device *device = NULL; 1208f46b5a66SChristoph Hellwig char *sizestr; 1209f46b5a66SChristoph Hellwig char *devstr = NULL; 1210f46b5a66SChristoph Hellwig int ret = 0; 1211f46b5a66SChristoph Hellwig int mod = 0; 1212f46b5a66SChristoph Hellwig 1213c146afadSYan Zheng if (root->fs_info->sb->s_flags & MS_RDONLY) 1214c146afadSYan Zheng return -EROFS; 1215c146afadSYan Zheng 1216e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 1217e441d54dSChris Mason return -EPERM; 1218e441d54dSChris Mason 1219c9e9f97bSIlya Dryomov mutex_lock(&root->fs_info->volume_mutex); 1220c9e9f97bSIlya Dryomov if (root->fs_info->balance_ctl) { 1221c9e9f97bSIlya Dryomov printk(KERN_INFO "btrfs: balance in progress\n"); 1222c9e9f97bSIlya Dryomov ret = -EINVAL; 1223c9e9f97bSIlya Dryomov goto out; 1224c9e9f97bSIlya Dryomov } 1225c9e9f97bSIlya Dryomov 1226dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1227c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 1228c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 1229c9e9f97bSIlya Dryomov goto out; 1230c9e9f97bSIlya Dryomov } 12315516e595SMark Fasheh 12325516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1233f46b5a66SChristoph Hellwig 1234f46b5a66SChristoph Hellwig sizestr = vol_args->name; 1235f46b5a66SChristoph Hellwig devstr = strchr(sizestr, ':'); 1236f46b5a66SChristoph Hellwig if (devstr) { 1237f46b5a66SChristoph Hellwig char *end; 1238f46b5a66SChristoph Hellwig sizestr = devstr + 1; 1239f46b5a66SChristoph Hellwig *devstr = '\0'; 1240f46b5a66SChristoph Hellwig devstr = vol_args->name; 1241f46b5a66SChristoph Hellwig devid = simple_strtoull(devstr, &end, 10); 12425bb14682SArnd Hannemann printk(KERN_INFO "btrfs: resizing devid %llu\n", 124321380931SJoel Becker (unsigned long long)devid); 1244f46b5a66SChristoph Hellwig } 12452b82032cSYan Zheng device = btrfs_find_device(root, devid, NULL, NULL); 1246f46b5a66SChristoph Hellwig if (!device) { 12475bb14682SArnd Hannemann printk(KERN_INFO "btrfs: resizer unable to find device %llu\n", 124821380931SJoel Becker (unsigned long long)devid); 1249f46b5a66SChristoph Hellwig ret = -EINVAL; 1250c9e9f97bSIlya Dryomov goto out_free; 1251f46b5a66SChristoph Hellwig } 1252f46b5a66SChristoph Hellwig if (!strcmp(sizestr, "max")) 1253f46b5a66SChristoph Hellwig new_size = device->bdev->bd_inode->i_size; 1254f46b5a66SChristoph Hellwig else { 1255f46b5a66SChristoph Hellwig if (sizestr[0] == '-') { 1256f46b5a66SChristoph Hellwig mod = -1; 1257f46b5a66SChristoph Hellwig sizestr++; 1258f46b5a66SChristoph Hellwig } else if (sizestr[0] == '+') { 1259f46b5a66SChristoph Hellwig mod = 1; 1260f46b5a66SChristoph Hellwig sizestr++; 1261f46b5a66SChristoph Hellwig } 126291748467SAkinobu Mita new_size = memparse(sizestr, NULL); 1263f46b5a66SChristoph Hellwig if (new_size == 0) { 1264f46b5a66SChristoph Hellwig ret = -EINVAL; 1265c9e9f97bSIlya Dryomov goto out_free; 1266f46b5a66SChristoph Hellwig } 1267f46b5a66SChristoph Hellwig } 1268f46b5a66SChristoph Hellwig 1269f46b5a66SChristoph Hellwig old_size = device->total_bytes; 1270f46b5a66SChristoph Hellwig 1271f46b5a66SChristoph Hellwig if (mod < 0) { 1272f46b5a66SChristoph Hellwig if (new_size > old_size) { 1273f46b5a66SChristoph Hellwig ret = -EINVAL; 1274c9e9f97bSIlya Dryomov goto out_free; 1275f46b5a66SChristoph Hellwig } 1276f46b5a66SChristoph Hellwig new_size = old_size - new_size; 1277f46b5a66SChristoph Hellwig } else if (mod > 0) { 1278f46b5a66SChristoph Hellwig new_size = old_size + new_size; 1279f46b5a66SChristoph Hellwig } 1280f46b5a66SChristoph Hellwig 1281f46b5a66SChristoph Hellwig if (new_size < 256 * 1024 * 1024) { 1282f46b5a66SChristoph Hellwig ret = -EINVAL; 1283c9e9f97bSIlya Dryomov goto out_free; 1284f46b5a66SChristoph Hellwig } 1285f46b5a66SChristoph Hellwig if (new_size > device->bdev->bd_inode->i_size) { 1286f46b5a66SChristoph Hellwig ret = -EFBIG; 1287c9e9f97bSIlya Dryomov goto out_free; 1288f46b5a66SChristoph Hellwig } 1289f46b5a66SChristoph Hellwig 1290f46b5a66SChristoph Hellwig do_div(new_size, root->sectorsize); 1291f46b5a66SChristoph Hellwig new_size *= root->sectorsize; 1292f46b5a66SChristoph Hellwig 12935bb14682SArnd Hannemann printk(KERN_INFO "btrfs: new size for %s is %llu\n", 1294f46b5a66SChristoph Hellwig device->name, (unsigned long long)new_size); 1295f46b5a66SChristoph Hellwig 1296f46b5a66SChristoph Hellwig if (new_size > old_size) { 1297a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 129898d5dc13STsutomu Itoh if (IS_ERR(trans)) { 129998d5dc13STsutomu Itoh ret = PTR_ERR(trans); 1300c9e9f97bSIlya Dryomov goto out_free; 130198d5dc13STsutomu Itoh } 1302f46b5a66SChristoph Hellwig ret = btrfs_grow_device(trans, device, new_size); 1303f46b5a66SChristoph Hellwig btrfs_commit_transaction(trans, root); 1304ece7d20eSMike Fleetwood } else if (new_size < old_size) { 1305f46b5a66SChristoph Hellwig ret = btrfs_shrink_device(device, new_size); 1306f46b5a66SChristoph Hellwig } 1307f46b5a66SChristoph Hellwig 1308c9e9f97bSIlya Dryomov out_free: 1309f46b5a66SChristoph Hellwig kfree(vol_args); 1310c9e9f97bSIlya Dryomov out: 1311c9e9f97bSIlya Dryomov mutex_unlock(&root->fs_info->volume_mutex); 1312f46b5a66SChristoph Hellwig return ret; 1313f46b5a66SChristoph Hellwig } 1314f46b5a66SChristoph Hellwig 131572fd032eSSage Weil static noinline int btrfs_ioctl_snap_create_transid(struct file *file, 131672fd032eSSage Weil char *name, 131772fd032eSSage Weil unsigned long fd, 131872fd032eSSage Weil int subvol, 1319b83cc969SLi Zefan u64 *transid, 1320b83cc969SLi Zefan bool readonly) 1321f46b5a66SChristoph Hellwig { 1322cb8e7090SChristoph Hellwig struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; 13233de4586cSChris Mason struct file *src_file; 1324f46b5a66SChristoph Hellwig int namelen; 13253de4586cSChris Mason int ret = 0; 1326f46b5a66SChristoph Hellwig 1327c146afadSYan Zheng if (root->fs_info->sb->s_flags & MS_RDONLY) 1328c146afadSYan Zheng return -EROFS; 1329c146afadSYan Zheng 133072fd032eSSage Weil namelen = strlen(name); 133172fd032eSSage Weil if (strchr(name, '/')) { 1332f46b5a66SChristoph Hellwig ret = -EINVAL; 1333f46b5a66SChristoph Hellwig goto out; 1334f46b5a66SChristoph Hellwig } 1335f46b5a66SChristoph Hellwig 1336*16780cabSChris Mason if (name[0] == '.' && 1337*16780cabSChris Mason (namelen == 1 || (name[1] == '.' && namelen == 2))) { 1338*16780cabSChris Mason ret = -EEXIST; 1339*16780cabSChris Mason goto out; 1340*16780cabSChris Mason } 1341*16780cabSChris Mason 13423de4586cSChris Mason if (subvol) { 134372fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 1344b83cc969SLi Zefan NULL, transid, readonly); 1345cb8e7090SChristoph Hellwig } else { 13463de4586cSChris Mason struct inode *src_inode; 134772fd032eSSage Weil src_file = fget(fd); 13483de4586cSChris Mason if (!src_file) { 13493de4586cSChris Mason ret = -EINVAL; 13503de4586cSChris Mason goto out; 13513de4586cSChris Mason } 13523de4586cSChris Mason 13533de4586cSChris Mason src_inode = src_file->f_path.dentry->d_inode; 13543de4586cSChris Mason if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) { 1355d397712bSChris Mason printk(KERN_INFO "btrfs: Snapshot src from " 1356d397712bSChris Mason "another FS\n"); 13573de4586cSChris Mason ret = -EINVAL; 13583de4586cSChris Mason fput(src_file); 13593de4586cSChris Mason goto out; 13603de4586cSChris Mason } 136172fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 136272fd032eSSage Weil BTRFS_I(src_inode)->root, 1363b83cc969SLi Zefan transid, readonly); 13643de4586cSChris Mason fput(src_file); 1365cb8e7090SChristoph Hellwig } 1366f46b5a66SChristoph Hellwig out: 136772fd032eSSage Weil return ret; 136872fd032eSSage Weil } 136972fd032eSSage Weil 137072fd032eSSage Weil static noinline int btrfs_ioctl_snap_create(struct file *file, 1371fa0d2b9bSLi Zefan void __user *arg, int subvol) 137272fd032eSSage Weil { 1373fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args *vol_args; 137472fd032eSSage Weil int ret; 137572fd032eSSage Weil 1376fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1377fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1378fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1379fa0d2b9bSLi Zefan vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1380fa0d2b9bSLi Zefan 1381fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 1382b83cc969SLi Zefan vol_args->fd, subvol, 1383b83cc969SLi Zefan NULL, false); 1384fa0d2b9bSLi Zefan 1385fa0d2b9bSLi Zefan kfree(vol_args); 1386fa0d2b9bSLi Zefan return ret; 1387fa0d2b9bSLi Zefan } 1388fa0d2b9bSLi Zefan 1389fa0d2b9bSLi Zefan static noinline int btrfs_ioctl_snap_create_v2(struct file *file, 1390fa0d2b9bSLi Zefan void __user *arg, int subvol) 1391fa0d2b9bSLi Zefan { 1392fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args_v2 *vol_args; 1393fa0d2b9bSLi Zefan int ret; 1394fdfb1e4fSLi Zefan u64 transid = 0; 1395fdfb1e4fSLi Zefan u64 *ptr = NULL; 1396b83cc969SLi Zefan bool readonly = false; 139772fd032eSSage Weil 1398fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1399fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1400fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1401fa0d2b9bSLi Zefan vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 1402fdfb1e4fSLi Zefan 1403b83cc969SLi Zefan if (vol_args->flags & 1404b83cc969SLi Zefan ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) { 1405b83cc969SLi Zefan ret = -EOPNOTSUPP; 1406fdfb1e4fSLi Zefan goto out; 1407fdfb1e4fSLi Zefan } 1408fdfb1e4fSLi Zefan 1409fa0d2b9bSLi Zefan if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC) 1410fdfb1e4fSLi Zefan ptr = &transid; 1411b83cc969SLi Zefan if (vol_args->flags & BTRFS_SUBVOL_RDONLY) 1412b83cc969SLi Zefan readonly = true; 141375eaa0e2SSage Weil 1414fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 1415b83cc969SLi Zefan vol_args->fd, subvol, 1416b83cc969SLi Zefan ptr, readonly); 141775eaa0e2SSage Weil 1418fdfb1e4fSLi Zefan if (ret == 0 && ptr && 141975eaa0e2SSage Weil copy_to_user(arg + 1420fdfb1e4fSLi Zefan offsetof(struct btrfs_ioctl_vol_args_v2, 1421fdfb1e4fSLi Zefan transid), ptr, sizeof(*ptr))) 142275eaa0e2SSage Weil ret = -EFAULT; 1423fdfb1e4fSLi Zefan out: 1424f46b5a66SChristoph Hellwig kfree(vol_args); 1425f46b5a66SChristoph Hellwig return ret; 1426f46b5a66SChristoph Hellwig } 1427f46b5a66SChristoph Hellwig 14280caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_getflags(struct file *file, 14290caa102dSLi Zefan void __user *arg) 14300caa102dSLi Zefan { 14310caa102dSLi Zefan struct inode *inode = fdentry(file)->d_inode; 14320caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 14330caa102dSLi Zefan int ret = 0; 14340caa102dSLi Zefan u64 flags = 0; 14350caa102dSLi Zefan 143633345d01SLi Zefan if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) 14370caa102dSLi Zefan return -EINVAL; 14380caa102dSLi Zefan 14390caa102dSLi Zefan down_read(&root->fs_info->subvol_sem); 14400caa102dSLi Zefan if (btrfs_root_readonly(root)) 14410caa102dSLi Zefan flags |= BTRFS_SUBVOL_RDONLY; 14420caa102dSLi Zefan up_read(&root->fs_info->subvol_sem); 14430caa102dSLi Zefan 14440caa102dSLi Zefan if (copy_to_user(arg, &flags, sizeof(flags))) 14450caa102dSLi Zefan ret = -EFAULT; 14460caa102dSLi Zefan 14470caa102dSLi Zefan return ret; 14480caa102dSLi Zefan } 14490caa102dSLi Zefan 14500caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_setflags(struct file *file, 14510caa102dSLi Zefan void __user *arg) 14520caa102dSLi Zefan { 14530caa102dSLi Zefan struct inode *inode = fdentry(file)->d_inode; 14540caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 14550caa102dSLi Zefan struct btrfs_trans_handle *trans; 14560caa102dSLi Zefan u64 root_flags; 14570caa102dSLi Zefan u64 flags; 14580caa102dSLi Zefan int ret = 0; 14590caa102dSLi Zefan 14600caa102dSLi Zefan if (root->fs_info->sb->s_flags & MS_RDONLY) 14610caa102dSLi Zefan return -EROFS; 14620caa102dSLi Zefan 146333345d01SLi Zefan if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) 14640caa102dSLi Zefan return -EINVAL; 14650caa102dSLi Zefan 14660caa102dSLi Zefan if (copy_from_user(&flags, arg, sizeof(flags))) 14670caa102dSLi Zefan return -EFAULT; 14680caa102dSLi Zefan 1469b4dc2b8cSLi Zefan if (flags & BTRFS_SUBVOL_CREATE_ASYNC) 14700caa102dSLi Zefan return -EINVAL; 14710caa102dSLi Zefan 14720caa102dSLi Zefan if (flags & ~BTRFS_SUBVOL_RDONLY) 14730caa102dSLi Zefan return -EOPNOTSUPP; 14740caa102dSLi Zefan 14752e149670SSerge E. Hallyn if (!inode_owner_or_capable(inode)) 1476b4dc2b8cSLi Zefan return -EACCES; 1477b4dc2b8cSLi Zefan 14780caa102dSLi Zefan down_write(&root->fs_info->subvol_sem); 14790caa102dSLi Zefan 14800caa102dSLi Zefan /* nothing to do */ 14810caa102dSLi Zefan if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) 14820caa102dSLi Zefan goto out; 14830caa102dSLi Zefan 14840caa102dSLi Zefan root_flags = btrfs_root_flags(&root->root_item); 14850caa102dSLi Zefan if (flags & BTRFS_SUBVOL_RDONLY) 14860caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 14870caa102dSLi Zefan root_flags | BTRFS_ROOT_SUBVOL_RDONLY); 14880caa102dSLi Zefan else 14890caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 14900caa102dSLi Zefan root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); 14910caa102dSLi Zefan 14920caa102dSLi Zefan trans = btrfs_start_transaction(root, 1); 14930caa102dSLi Zefan if (IS_ERR(trans)) { 14940caa102dSLi Zefan ret = PTR_ERR(trans); 14950caa102dSLi Zefan goto out_reset; 14960caa102dSLi Zefan } 14970caa102dSLi Zefan 1498b4dc2b8cSLi Zefan ret = btrfs_update_root(trans, root->fs_info->tree_root, 14990caa102dSLi Zefan &root->root_key, &root->root_item); 15000caa102dSLi Zefan 15010caa102dSLi Zefan btrfs_commit_transaction(trans, root); 15020caa102dSLi Zefan out_reset: 15030caa102dSLi Zefan if (ret) 15040caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, root_flags); 15050caa102dSLi Zefan out: 15060caa102dSLi Zefan up_write(&root->fs_info->subvol_sem); 15070caa102dSLi Zefan return ret; 15080caa102dSLi Zefan } 15090caa102dSLi Zefan 151076dda93cSYan, Zheng /* 151176dda93cSYan, Zheng * helper to check if the subvolume references other subvolumes 151276dda93cSYan, Zheng */ 151376dda93cSYan, Zheng static noinline int may_destroy_subvol(struct btrfs_root *root) 151476dda93cSYan, Zheng { 151576dda93cSYan, Zheng struct btrfs_path *path; 151676dda93cSYan, Zheng struct btrfs_key key; 151776dda93cSYan, Zheng int ret; 151876dda93cSYan, Zheng 151976dda93cSYan, Zheng path = btrfs_alloc_path(); 152076dda93cSYan, Zheng if (!path) 152176dda93cSYan, Zheng return -ENOMEM; 152276dda93cSYan, Zheng 152376dda93cSYan, Zheng key.objectid = root->root_key.objectid; 152476dda93cSYan, Zheng key.type = BTRFS_ROOT_REF_KEY; 152576dda93cSYan, Zheng key.offset = (u64)-1; 152676dda93cSYan, Zheng 152776dda93cSYan, Zheng ret = btrfs_search_slot(NULL, root->fs_info->tree_root, 152876dda93cSYan, Zheng &key, path, 0, 0); 152976dda93cSYan, Zheng if (ret < 0) 153076dda93cSYan, Zheng goto out; 153176dda93cSYan, Zheng BUG_ON(ret == 0); 153276dda93cSYan, Zheng 153376dda93cSYan, Zheng ret = 0; 153476dda93cSYan, Zheng if (path->slots[0] > 0) { 153576dda93cSYan, Zheng path->slots[0]--; 153676dda93cSYan, Zheng btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 153776dda93cSYan, Zheng if (key.objectid == root->root_key.objectid && 153876dda93cSYan, Zheng key.type == BTRFS_ROOT_REF_KEY) 153976dda93cSYan, Zheng ret = -ENOTEMPTY; 154076dda93cSYan, Zheng } 154176dda93cSYan, Zheng out: 154276dda93cSYan, Zheng btrfs_free_path(path); 154376dda93cSYan, Zheng return ret; 154476dda93cSYan, Zheng } 154576dda93cSYan, Zheng 1546ac8e9819SChris Mason static noinline int key_in_sk(struct btrfs_key *key, 1547ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk) 1548ac8e9819SChris Mason { 1549abc6e134SChris Mason struct btrfs_key test; 1550abc6e134SChris Mason int ret; 1551abc6e134SChris Mason 1552abc6e134SChris Mason test.objectid = sk->min_objectid; 1553abc6e134SChris Mason test.type = sk->min_type; 1554abc6e134SChris Mason test.offset = sk->min_offset; 1555abc6e134SChris Mason 1556abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1557abc6e134SChris Mason if (ret < 0) 1558ac8e9819SChris Mason return 0; 1559abc6e134SChris Mason 1560abc6e134SChris Mason test.objectid = sk->max_objectid; 1561abc6e134SChris Mason test.type = sk->max_type; 1562abc6e134SChris Mason test.offset = sk->max_offset; 1563abc6e134SChris Mason 1564abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1565abc6e134SChris Mason if (ret > 0) 1566ac8e9819SChris Mason return 0; 1567ac8e9819SChris Mason return 1; 1568ac8e9819SChris Mason } 1569ac8e9819SChris Mason 1570ac8e9819SChris Mason static noinline int copy_to_sk(struct btrfs_root *root, 1571ac8e9819SChris Mason struct btrfs_path *path, 1572ac8e9819SChris Mason struct btrfs_key *key, 1573ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk, 1574ac8e9819SChris Mason char *buf, 1575ac8e9819SChris Mason unsigned long *sk_offset, 1576ac8e9819SChris Mason int *num_found) 1577ac8e9819SChris Mason { 1578ac8e9819SChris Mason u64 found_transid; 1579ac8e9819SChris Mason struct extent_buffer *leaf; 1580ac8e9819SChris Mason struct btrfs_ioctl_search_header sh; 1581ac8e9819SChris Mason unsigned long item_off; 1582ac8e9819SChris Mason unsigned long item_len; 1583ac8e9819SChris Mason int nritems; 1584ac8e9819SChris Mason int i; 1585ac8e9819SChris Mason int slot; 1586ac8e9819SChris Mason int ret = 0; 1587ac8e9819SChris Mason 1588ac8e9819SChris Mason leaf = path->nodes[0]; 1589ac8e9819SChris Mason slot = path->slots[0]; 1590ac8e9819SChris Mason nritems = btrfs_header_nritems(leaf); 1591ac8e9819SChris Mason 1592ac8e9819SChris Mason if (btrfs_header_generation(leaf) > sk->max_transid) { 1593ac8e9819SChris Mason i = nritems; 1594ac8e9819SChris Mason goto advance_key; 1595ac8e9819SChris Mason } 1596ac8e9819SChris Mason found_transid = btrfs_header_generation(leaf); 1597ac8e9819SChris Mason 1598ac8e9819SChris Mason for (i = slot; i < nritems; i++) { 1599ac8e9819SChris Mason item_off = btrfs_item_ptr_offset(leaf, i); 1600ac8e9819SChris Mason item_len = btrfs_item_size_nr(leaf, i); 1601ac8e9819SChris Mason 1602ac8e9819SChris Mason if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE) 1603ac8e9819SChris Mason item_len = 0; 1604ac8e9819SChris Mason 1605ac8e9819SChris Mason if (sizeof(sh) + item_len + *sk_offset > 1606ac8e9819SChris Mason BTRFS_SEARCH_ARGS_BUFSIZE) { 1607ac8e9819SChris Mason ret = 1; 1608ac8e9819SChris Mason goto overflow; 1609ac8e9819SChris Mason } 1610ac8e9819SChris Mason 1611ac8e9819SChris Mason btrfs_item_key_to_cpu(leaf, key, i); 1612ac8e9819SChris Mason if (!key_in_sk(key, sk)) 1613ac8e9819SChris Mason continue; 1614ac8e9819SChris Mason 1615ac8e9819SChris Mason sh.objectid = key->objectid; 1616ac8e9819SChris Mason sh.offset = key->offset; 1617ac8e9819SChris Mason sh.type = key->type; 1618ac8e9819SChris Mason sh.len = item_len; 1619ac8e9819SChris Mason sh.transid = found_transid; 1620ac8e9819SChris Mason 1621ac8e9819SChris Mason /* copy search result header */ 1622ac8e9819SChris Mason memcpy(buf + *sk_offset, &sh, sizeof(sh)); 1623ac8e9819SChris Mason *sk_offset += sizeof(sh); 1624ac8e9819SChris Mason 1625ac8e9819SChris Mason if (item_len) { 1626ac8e9819SChris Mason char *p = buf + *sk_offset; 1627ac8e9819SChris Mason /* copy the item */ 1628ac8e9819SChris Mason read_extent_buffer(leaf, p, 1629ac8e9819SChris Mason item_off, item_len); 1630ac8e9819SChris Mason *sk_offset += item_len; 1631ac8e9819SChris Mason } 1632e2156867SHugo Mills (*num_found)++; 1633ac8e9819SChris Mason 1634ac8e9819SChris Mason if (*num_found >= sk->nr_items) 1635ac8e9819SChris Mason break; 1636ac8e9819SChris Mason } 1637ac8e9819SChris Mason advance_key: 1638ac8e9819SChris Mason ret = 0; 1639abc6e134SChris Mason if (key->offset < (u64)-1 && key->offset < sk->max_offset) 1640abc6e134SChris Mason key->offset++; 1641abc6e134SChris Mason else if (key->type < (u8)-1 && key->type < sk->max_type) { 1642abc6e134SChris Mason key->offset = 0; 1643abc6e134SChris Mason key->type++; 1644abc6e134SChris Mason } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) { 1645abc6e134SChris Mason key->offset = 0; 1646abc6e134SChris Mason key->type = 0; 1647abc6e134SChris Mason key->objectid++; 1648abc6e134SChris Mason } else 1649abc6e134SChris Mason ret = 1; 1650ac8e9819SChris Mason overflow: 1651ac8e9819SChris Mason return ret; 1652ac8e9819SChris Mason } 1653ac8e9819SChris Mason 1654ac8e9819SChris Mason static noinline int search_ioctl(struct inode *inode, 1655ac8e9819SChris Mason struct btrfs_ioctl_search_args *args) 1656ac8e9819SChris Mason { 1657ac8e9819SChris Mason struct btrfs_root *root; 1658ac8e9819SChris Mason struct btrfs_key key; 1659ac8e9819SChris Mason struct btrfs_key max_key; 1660ac8e9819SChris Mason struct btrfs_path *path; 1661ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk = &args->key; 1662ac8e9819SChris Mason struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info; 1663ac8e9819SChris Mason int ret; 1664ac8e9819SChris Mason int num_found = 0; 1665ac8e9819SChris Mason unsigned long sk_offset = 0; 1666ac8e9819SChris Mason 1667ac8e9819SChris Mason path = btrfs_alloc_path(); 1668ac8e9819SChris Mason if (!path) 1669ac8e9819SChris Mason return -ENOMEM; 1670ac8e9819SChris Mason 1671ac8e9819SChris Mason if (sk->tree_id == 0) { 1672ac8e9819SChris Mason /* search the root of the inode that was passed */ 1673ac8e9819SChris Mason root = BTRFS_I(inode)->root; 1674ac8e9819SChris Mason } else { 1675ac8e9819SChris Mason key.objectid = sk->tree_id; 1676ac8e9819SChris Mason key.type = BTRFS_ROOT_ITEM_KEY; 1677ac8e9819SChris Mason key.offset = (u64)-1; 1678ac8e9819SChris Mason root = btrfs_read_fs_root_no_name(info, &key); 1679ac8e9819SChris Mason if (IS_ERR(root)) { 1680ac8e9819SChris Mason printk(KERN_ERR "could not find root %llu\n", 1681ac8e9819SChris Mason sk->tree_id); 1682ac8e9819SChris Mason btrfs_free_path(path); 1683ac8e9819SChris Mason return -ENOENT; 1684ac8e9819SChris Mason } 1685ac8e9819SChris Mason } 1686ac8e9819SChris Mason 1687ac8e9819SChris Mason key.objectid = sk->min_objectid; 1688ac8e9819SChris Mason key.type = sk->min_type; 1689ac8e9819SChris Mason key.offset = sk->min_offset; 1690ac8e9819SChris Mason 1691ac8e9819SChris Mason max_key.objectid = sk->max_objectid; 1692ac8e9819SChris Mason max_key.type = sk->max_type; 1693ac8e9819SChris Mason max_key.offset = sk->max_offset; 1694ac8e9819SChris Mason 1695ac8e9819SChris Mason path->keep_locks = 1; 1696ac8e9819SChris Mason 1697ac8e9819SChris Mason while(1) { 1698ac8e9819SChris Mason ret = btrfs_search_forward(root, &key, &max_key, path, 0, 1699ac8e9819SChris Mason sk->min_transid); 1700ac8e9819SChris Mason if (ret != 0) { 1701ac8e9819SChris Mason if (ret > 0) 1702ac8e9819SChris Mason ret = 0; 1703ac8e9819SChris Mason goto err; 1704ac8e9819SChris Mason } 1705ac8e9819SChris Mason ret = copy_to_sk(root, path, &key, sk, args->buf, 1706ac8e9819SChris Mason &sk_offset, &num_found); 1707b3b4aa74SDavid Sterba btrfs_release_path(path); 1708ac8e9819SChris Mason if (ret || num_found >= sk->nr_items) 1709ac8e9819SChris Mason break; 1710ac8e9819SChris Mason 1711ac8e9819SChris Mason } 1712ac8e9819SChris Mason ret = 0; 1713ac8e9819SChris Mason err: 1714ac8e9819SChris Mason sk->nr_items = num_found; 1715ac8e9819SChris Mason btrfs_free_path(path); 1716ac8e9819SChris Mason return ret; 1717ac8e9819SChris Mason } 1718ac8e9819SChris Mason 1719ac8e9819SChris Mason static noinline int btrfs_ioctl_tree_search(struct file *file, 1720ac8e9819SChris Mason void __user *argp) 1721ac8e9819SChris Mason { 1722ac8e9819SChris Mason struct btrfs_ioctl_search_args *args; 1723ac8e9819SChris Mason struct inode *inode; 1724ac8e9819SChris Mason int ret; 1725ac8e9819SChris Mason 1726ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 1727ac8e9819SChris Mason return -EPERM; 1728ac8e9819SChris Mason 17292354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 17302354d08fSJulia Lawall if (IS_ERR(args)) 17312354d08fSJulia Lawall return PTR_ERR(args); 1732ac8e9819SChris Mason 1733ac8e9819SChris Mason inode = fdentry(file)->d_inode; 1734ac8e9819SChris Mason ret = search_ioctl(inode, args); 1735ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 1736ac8e9819SChris Mason ret = -EFAULT; 1737ac8e9819SChris Mason kfree(args); 1738ac8e9819SChris Mason return ret; 1739ac8e9819SChris Mason } 1740ac8e9819SChris Mason 174198d377a0STARUISI Hiroaki /* 1742ac8e9819SChris Mason * Search INODE_REFs to identify path name of 'dirid' directory 1743ac8e9819SChris Mason * in a 'tree_id' tree. and sets path name to 'name'. 174498d377a0STARUISI Hiroaki */ 174598d377a0STARUISI Hiroaki static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, 174698d377a0STARUISI Hiroaki u64 tree_id, u64 dirid, char *name) 174798d377a0STARUISI Hiroaki { 174898d377a0STARUISI Hiroaki struct btrfs_root *root; 174998d377a0STARUISI Hiroaki struct btrfs_key key; 1750ac8e9819SChris Mason char *ptr; 175198d377a0STARUISI Hiroaki int ret = -1; 175298d377a0STARUISI Hiroaki int slot; 175398d377a0STARUISI Hiroaki int len; 175498d377a0STARUISI Hiroaki int total_len = 0; 175598d377a0STARUISI Hiroaki struct btrfs_inode_ref *iref; 175698d377a0STARUISI Hiroaki struct extent_buffer *l; 175798d377a0STARUISI Hiroaki struct btrfs_path *path; 175898d377a0STARUISI Hiroaki 175998d377a0STARUISI Hiroaki if (dirid == BTRFS_FIRST_FREE_OBJECTID) { 176098d377a0STARUISI Hiroaki name[0]='\0'; 176198d377a0STARUISI Hiroaki return 0; 176298d377a0STARUISI Hiroaki } 176398d377a0STARUISI Hiroaki 176498d377a0STARUISI Hiroaki path = btrfs_alloc_path(); 176598d377a0STARUISI Hiroaki if (!path) 176698d377a0STARUISI Hiroaki return -ENOMEM; 176798d377a0STARUISI Hiroaki 1768ac8e9819SChris Mason ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX]; 176998d377a0STARUISI Hiroaki 177098d377a0STARUISI Hiroaki key.objectid = tree_id; 177198d377a0STARUISI Hiroaki key.type = BTRFS_ROOT_ITEM_KEY; 177298d377a0STARUISI Hiroaki key.offset = (u64)-1; 177398d377a0STARUISI Hiroaki root = btrfs_read_fs_root_no_name(info, &key); 177498d377a0STARUISI Hiroaki if (IS_ERR(root)) { 177598d377a0STARUISI Hiroaki printk(KERN_ERR "could not find root %llu\n", tree_id); 17768ad6fcabSChris Mason ret = -ENOENT; 17778ad6fcabSChris Mason goto out; 177898d377a0STARUISI Hiroaki } 177998d377a0STARUISI Hiroaki 178098d377a0STARUISI Hiroaki key.objectid = dirid; 178198d377a0STARUISI Hiroaki key.type = BTRFS_INODE_REF_KEY; 17828ad6fcabSChris Mason key.offset = (u64)-1; 178398d377a0STARUISI Hiroaki 178498d377a0STARUISI Hiroaki while(1) { 178598d377a0STARUISI Hiroaki ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 178698d377a0STARUISI Hiroaki if (ret < 0) 178798d377a0STARUISI Hiroaki goto out; 178898d377a0STARUISI Hiroaki 178998d377a0STARUISI Hiroaki l = path->nodes[0]; 179098d377a0STARUISI Hiroaki slot = path->slots[0]; 17918ad6fcabSChris Mason if (ret > 0 && slot > 0) 17928ad6fcabSChris Mason slot--; 179398d377a0STARUISI Hiroaki btrfs_item_key_to_cpu(l, &key, slot); 179498d377a0STARUISI Hiroaki 179598d377a0STARUISI Hiroaki if (ret > 0 && (key.objectid != dirid || 1796ac8e9819SChris Mason key.type != BTRFS_INODE_REF_KEY)) { 1797ac8e9819SChris Mason ret = -ENOENT; 179898d377a0STARUISI Hiroaki goto out; 1799ac8e9819SChris Mason } 180098d377a0STARUISI Hiroaki 180198d377a0STARUISI Hiroaki iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); 180298d377a0STARUISI Hiroaki len = btrfs_inode_ref_name_len(l, iref); 180398d377a0STARUISI Hiroaki ptr -= len + 1; 180498d377a0STARUISI Hiroaki total_len += len + 1; 1805ac8e9819SChris Mason if (ptr < name) 180698d377a0STARUISI Hiroaki goto out; 180798d377a0STARUISI Hiroaki 180898d377a0STARUISI Hiroaki *(ptr + len) = '/'; 180998d377a0STARUISI Hiroaki read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len); 181098d377a0STARUISI Hiroaki 181198d377a0STARUISI Hiroaki if (key.offset == BTRFS_FIRST_FREE_OBJECTID) 181298d377a0STARUISI Hiroaki break; 181398d377a0STARUISI Hiroaki 1814b3b4aa74SDavid Sterba btrfs_release_path(path); 181598d377a0STARUISI Hiroaki key.objectid = key.offset; 18168ad6fcabSChris Mason key.offset = (u64)-1; 181798d377a0STARUISI Hiroaki dirid = key.objectid; 181898d377a0STARUISI Hiroaki } 1819ac8e9819SChris Mason if (ptr < name) 182098d377a0STARUISI Hiroaki goto out; 182177906a50SLi Zefan memmove(name, ptr, total_len); 182298d377a0STARUISI Hiroaki name[total_len]='\0'; 182398d377a0STARUISI Hiroaki ret = 0; 182498d377a0STARUISI Hiroaki out: 182598d377a0STARUISI Hiroaki btrfs_free_path(path); 1826ac8e9819SChris Mason return ret; 1827ac8e9819SChris Mason } 1828ac8e9819SChris Mason 1829ac8e9819SChris Mason static noinline int btrfs_ioctl_ino_lookup(struct file *file, 1830ac8e9819SChris Mason void __user *argp) 1831ac8e9819SChris Mason { 1832ac8e9819SChris Mason struct btrfs_ioctl_ino_lookup_args *args; 1833ac8e9819SChris Mason struct inode *inode; 1834ac8e9819SChris Mason int ret; 1835ac8e9819SChris Mason 1836ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 1837ac8e9819SChris Mason return -EPERM; 1838ac8e9819SChris Mason 18392354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 18402354d08fSJulia Lawall if (IS_ERR(args)) 18412354d08fSJulia Lawall return PTR_ERR(args); 1842c2b96929SDan Carpenter 1843ac8e9819SChris Mason inode = fdentry(file)->d_inode; 1844ac8e9819SChris Mason 18451b53ac4dSChris Mason if (args->treeid == 0) 18461b53ac4dSChris Mason args->treeid = BTRFS_I(inode)->root->root_key.objectid; 18471b53ac4dSChris Mason 1848ac8e9819SChris Mason ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, 1849ac8e9819SChris Mason args->treeid, args->objectid, 1850ac8e9819SChris Mason args->name); 1851ac8e9819SChris Mason 1852ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 1853ac8e9819SChris Mason ret = -EFAULT; 1854ac8e9819SChris Mason 1855ac8e9819SChris Mason kfree(args); 185698d377a0STARUISI Hiroaki return ret; 185798d377a0STARUISI Hiroaki } 185898d377a0STARUISI Hiroaki 185976dda93cSYan, Zheng static noinline int btrfs_ioctl_snap_destroy(struct file *file, 186076dda93cSYan, Zheng void __user *arg) 186176dda93cSYan, Zheng { 186276dda93cSYan, Zheng struct dentry *parent = fdentry(file); 186376dda93cSYan, Zheng struct dentry *dentry; 186476dda93cSYan, Zheng struct inode *dir = parent->d_inode; 186576dda93cSYan, Zheng struct inode *inode; 186676dda93cSYan, Zheng struct btrfs_root *root = BTRFS_I(dir)->root; 186776dda93cSYan, Zheng struct btrfs_root *dest = NULL; 186876dda93cSYan, Zheng struct btrfs_ioctl_vol_args *vol_args; 186976dda93cSYan, Zheng struct btrfs_trans_handle *trans; 187076dda93cSYan, Zheng int namelen; 187176dda93cSYan, Zheng int ret; 187276dda93cSYan, Zheng int err = 0; 187376dda93cSYan, Zheng 187476dda93cSYan, Zheng vol_args = memdup_user(arg, sizeof(*vol_args)); 187576dda93cSYan, Zheng if (IS_ERR(vol_args)) 187676dda93cSYan, Zheng return PTR_ERR(vol_args); 187776dda93cSYan, Zheng 187876dda93cSYan, Zheng vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 187976dda93cSYan, Zheng namelen = strlen(vol_args->name); 188076dda93cSYan, Zheng if (strchr(vol_args->name, '/') || 188176dda93cSYan, Zheng strncmp(vol_args->name, "..", namelen) == 0) { 188276dda93cSYan, Zheng err = -EINVAL; 188376dda93cSYan, Zheng goto out; 188476dda93cSYan, Zheng } 188576dda93cSYan, Zheng 188676dda93cSYan, Zheng err = mnt_want_write(file->f_path.mnt); 188776dda93cSYan, Zheng if (err) 188876dda93cSYan, Zheng goto out; 188976dda93cSYan, Zheng 189076dda93cSYan, Zheng mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT); 189176dda93cSYan, Zheng dentry = lookup_one_len(vol_args->name, parent, namelen); 189276dda93cSYan, Zheng if (IS_ERR(dentry)) { 189376dda93cSYan, Zheng err = PTR_ERR(dentry); 189476dda93cSYan, Zheng goto out_unlock_dir; 189576dda93cSYan, Zheng } 189676dda93cSYan, Zheng 189776dda93cSYan, Zheng if (!dentry->d_inode) { 189876dda93cSYan, Zheng err = -ENOENT; 189976dda93cSYan, Zheng goto out_dput; 190076dda93cSYan, Zheng } 190176dda93cSYan, Zheng 190276dda93cSYan, Zheng inode = dentry->d_inode; 19034260f7c7SSage Weil dest = BTRFS_I(inode)->root; 19044260f7c7SSage Weil if (!capable(CAP_SYS_ADMIN)){ 19054260f7c7SSage Weil /* 19064260f7c7SSage Weil * Regular user. Only allow this with a special mount 19074260f7c7SSage Weil * option, when the user has write+exec access to the 19084260f7c7SSage Weil * subvol root, and when rmdir(2) would have been 19094260f7c7SSage Weil * allowed. 19104260f7c7SSage Weil * 19114260f7c7SSage Weil * Note that this is _not_ check that the subvol is 19124260f7c7SSage Weil * empty or doesn't contain data that we wouldn't 19134260f7c7SSage Weil * otherwise be able to delete. 19144260f7c7SSage Weil * 19154260f7c7SSage Weil * Users who want to delete empty subvols should try 19164260f7c7SSage Weil * rmdir(2). 19174260f7c7SSage Weil */ 19184260f7c7SSage Weil err = -EPERM; 19194260f7c7SSage Weil if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED)) 19204260f7c7SSage Weil goto out_dput; 19214260f7c7SSage Weil 19224260f7c7SSage Weil /* 19234260f7c7SSage Weil * Do not allow deletion if the parent dir is the same 19244260f7c7SSage Weil * as the dir to be deleted. That means the ioctl 19254260f7c7SSage Weil * must be called on the dentry referencing the root 19264260f7c7SSage Weil * of the subvol, not a random directory contained 19274260f7c7SSage Weil * within it. 19284260f7c7SSage Weil */ 19294260f7c7SSage Weil err = -EINVAL; 19304260f7c7SSage Weil if (root == dest) 19314260f7c7SSage Weil goto out_dput; 19324260f7c7SSage Weil 19334260f7c7SSage Weil err = inode_permission(inode, MAY_WRITE | MAY_EXEC); 19344260f7c7SSage Weil if (err) 19354260f7c7SSage Weil goto out_dput; 19364260f7c7SSage Weil 19374260f7c7SSage Weil /* check if subvolume may be deleted by a non-root user */ 19384260f7c7SSage Weil err = btrfs_may_delete(dir, dentry, 1); 19394260f7c7SSage Weil if (err) 19404260f7c7SSage Weil goto out_dput; 19414260f7c7SSage Weil } 19424260f7c7SSage Weil 194333345d01SLi Zefan if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { 194476dda93cSYan, Zheng err = -EINVAL; 194576dda93cSYan, Zheng goto out_dput; 194676dda93cSYan, Zheng } 194776dda93cSYan, Zheng 194876dda93cSYan, Zheng mutex_lock(&inode->i_mutex); 194976dda93cSYan, Zheng err = d_invalidate(dentry); 195076dda93cSYan, Zheng if (err) 195176dda93cSYan, Zheng goto out_unlock; 195276dda93cSYan, Zheng 195376dda93cSYan, Zheng down_write(&root->fs_info->subvol_sem); 195476dda93cSYan, Zheng 195576dda93cSYan, Zheng err = may_destroy_subvol(dest); 195676dda93cSYan, Zheng if (err) 195776dda93cSYan, Zheng goto out_up_write; 195876dda93cSYan, Zheng 1959a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 1960a22285a6SYan, Zheng if (IS_ERR(trans)) { 1961a22285a6SYan, Zheng err = PTR_ERR(trans); 1962d327099aSDan Carpenter goto out_up_write; 1963a22285a6SYan, Zheng } 1964a22285a6SYan, Zheng trans->block_rsv = &root->fs_info->global_block_rsv; 1965a22285a6SYan, Zheng 196676dda93cSYan, Zheng ret = btrfs_unlink_subvol(trans, root, dir, 196776dda93cSYan, Zheng dest->root_key.objectid, 196876dda93cSYan, Zheng dentry->d_name.name, 196976dda93cSYan, Zheng dentry->d_name.len); 197076dda93cSYan, Zheng BUG_ON(ret); 197176dda93cSYan, Zheng 197276dda93cSYan, Zheng btrfs_record_root_in_trans(trans, dest); 197376dda93cSYan, Zheng 197476dda93cSYan, Zheng memset(&dest->root_item.drop_progress, 0, 197576dda93cSYan, Zheng sizeof(dest->root_item.drop_progress)); 197676dda93cSYan, Zheng dest->root_item.drop_level = 0; 197776dda93cSYan, Zheng btrfs_set_root_refs(&dest->root_item, 0); 197876dda93cSYan, Zheng 1979d68fc57bSYan, Zheng if (!xchg(&dest->orphan_item_inserted, 1)) { 198076dda93cSYan, Zheng ret = btrfs_insert_orphan_item(trans, 198176dda93cSYan, Zheng root->fs_info->tree_root, 198276dda93cSYan, Zheng dest->root_key.objectid); 198376dda93cSYan, Zheng BUG_ON(ret); 1984d68fc57bSYan, Zheng } 198576dda93cSYan, Zheng 1986531cb13fSSage Weil ret = btrfs_end_transaction(trans, root); 198776dda93cSYan, Zheng BUG_ON(ret); 198876dda93cSYan, Zheng inode->i_flags |= S_DEAD; 198976dda93cSYan, Zheng out_up_write: 199076dda93cSYan, Zheng up_write(&root->fs_info->subvol_sem); 199176dda93cSYan, Zheng out_unlock: 199276dda93cSYan, Zheng mutex_unlock(&inode->i_mutex); 199376dda93cSYan, Zheng if (!err) { 1994efefb143SYan, Zheng shrink_dcache_sb(root->fs_info->sb); 199576dda93cSYan, Zheng btrfs_invalidate_inodes(dest); 199676dda93cSYan, Zheng d_delete(dentry); 199776dda93cSYan, Zheng } 199876dda93cSYan, Zheng out_dput: 199976dda93cSYan, Zheng dput(dentry); 200076dda93cSYan, Zheng out_unlock_dir: 200176dda93cSYan, Zheng mutex_unlock(&dir->i_mutex); 200276dda93cSYan, Zheng mnt_drop_write(file->f_path.mnt); 200376dda93cSYan, Zheng out: 200476dda93cSYan, Zheng kfree(vol_args); 200576dda93cSYan, Zheng return err; 200676dda93cSYan, Zheng } 200776dda93cSYan, Zheng 20081e701a32SChris Mason static int btrfs_ioctl_defrag(struct file *file, void __user *argp) 2009f46b5a66SChristoph Hellwig { 2010f46b5a66SChristoph Hellwig struct inode *inode = fdentry(file)->d_inode; 2011f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 20121e701a32SChris Mason struct btrfs_ioctl_defrag_range_args *range; 2013c146afadSYan Zheng int ret; 2014c146afadSYan Zheng 2015b83cc969SLi Zefan if (btrfs_root_readonly(root)) 2016b83cc969SLi Zefan return -EROFS; 2017b83cc969SLi Zefan 2018c146afadSYan Zheng ret = mnt_want_write(file->f_path.mnt); 2019c146afadSYan Zheng if (ret) 2020c146afadSYan Zheng return ret; 2021f46b5a66SChristoph Hellwig 2022f46b5a66SChristoph Hellwig switch (inode->i_mode & S_IFMT) { 2023f46b5a66SChristoph Hellwig case S_IFDIR: 2024e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) { 2025e441d54dSChris Mason ret = -EPERM; 2026e441d54dSChris Mason goto out; 2027e441d54dSChris Mason } 20288929ecfaSYan, Zheng ret = btrfs_defrag_root(root, 0); 20298929ecfaSYan, Zheng if (ret) 20308929ecfaSYan, Zheng goto out; 20318929ecfaSYan, Zheng ret = btrfs_defrag_root(root->fs_info->extent_root, 0); 2032f46b5a66SChristoph Hellwig break; 2033f46b5a66SChristoph Hellwig case S_IFREG: 2034e441d54dSChris Mason if (!(file->f_mode & FMODE_WRITE)) { 2035e441d54dSChris Mason ret = -EINVAL; 2036e441d54dSChris Mason goto out; 2037e441d54dSChris Mason } 20381e701a32SChris Mason 20391e701a32SChris Mason range = kzalloc(sizeof(*range), GFP_KERNEL); 20401e701a32SChris Mason if (!range) { 20411e701a32SChris Mason ret = -ENOMEM; 20421e701a32SChris Mason goto out; 20431e701a32SChris Mason } 20441e701a32SChris Mason 20451e701a32SChris Mason if (argp) { 20461e701a32SChris Mason if (copy_from_user(range, argp, 20471e701a32SChris Mason sizeof(*range))) { 20481e701a32SChris Mason ret = -EFAULT; 20491e701a32SChris Mason kfree(range); 2050683be16eSDan Carpenter goto out; 20511e701a32SChris Mason } 20521e701a32SChris Mason /* compression requires us to start the IO */ 20531e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 20541e701a32SChris Mason range->flags |= BTRFS_DEFRAG_RANGE_START_IO; 20551e701a32SChris Mason range->extent_thresh = (u32)-1; 20561e701a32SChris Mason } 20571e701a32SChris Mason } else { 20581e701a32SChris Mason /* the rest are all set to zero by kzalloc */ 20591e701a32SChris Mason range->len = (u64)-1; 20601e701a32SChris Mason } 20614cb5300bSChris Mason ret = btrfs_defrag_file(fdentry(file)->d_inode, file, 20624cb5300bSChris Mason range, 0, 0); 20634cb5300bSChris Mason if (ret > 0) 20644cb5300bSChris Mason ret = 0; 20651e701a32SChris Mason kfree(range); 2066f46b5a66SChristoph Hellwig break; 20678929ecfaSYan, Zheng default: 20688929ecfaSYan, Zheng ret = -EINVAL; 2069f46b5a66SChristoph Hellwig } 2070e441d54dSChris Mason out: 2071ab67b7c1SYan Zheng mnt_drop_write(file->f_path.mnt); 2072e441d54dSChris Mason return ret; 2073f46b5a66SChristoph Hellwig } 2074f46b5a66SChristoph Hellwig 2075b2950863SChristoph Hellwig static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg) 2076f46b5a66SChristoph Hellwig { 2077f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2078f46b5a66SChristoph Hellwig int ret; 2079f46b5a66SChristoph Hellwig 2080e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2081e441d54dSChris Mason return -EPERM; 2082e441d54dSChris Mason 2083c9e9f97bSIlya Dryomov mutex_lock(&root->fs_info->volume_mutex); 2084c9e9f97bSIlya Dryomov if (root->fs_info->balance_ctl) { 2085c9e9f97bSIlya Dryomov printk(KERN_INFO "btrfs: balance in progress\n"); 2086c9e9f97bSIlya Dryomov ret = -EINVAL; 2087c9e9f97bSIlya Dryomov goto out; 2088c9e9f97bSIlya Dryomov } 2089c9e9f97bSIlya Dryomov 2090dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2091c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 2092c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 2093c9e9f97bSIlya Dryomov goto out; 2094c9e9f97bSIlya Dryomov } 2095f46b5a66SChristoph Hellwig 20965516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 2097f46b5a66SChristoph Hellwig ret = btrfs_init_new_device(root, vol_args->name); 2098f46b5a66SChristoph Hellwig 2099f46b5a66SChristoph Hellwig kfree(vol_args); 2100c9e9f97bSIlya Dryomov out: 2101c9e9f97bSIlya Dryomov mutex_unlock(&root->fs_info->volume_mutex); 2102f46b5a66SChristoph Hellwig return ret; 2103f46b5a66SChristoph Hellwig } 2104f46b5a66SChristoph Hellwig 2105b2950863SChristoph Hellwig static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg) 2106f46b5a66SChristoph Hellwig { 2107f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2108f46b5a66SChristoph Hellwig int ret; 2109f46b5a66SChristoph Hellwig 2110e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2111e441d54dSChris Mason return -EPERM; 2112e441d54dSChris Mason 2113c146afadSYan Zheng if (root->fs_info->sb->s_flags & MS_RDONLY) 2114c146afadSYan Zheng return -EROFS; 2115c146afadSYan Zheng 2116c9e9f97bSIlya Dryomov mutex_lock(&root->fs_info->volume_mutex); 2117c9e9f97bSIlya Dryomov if (root->fs_info->balance_ctl) { 2118c9e9f97bSIlya Dryomov printk(KERN_INFO "btrfs: balance in progress\n"); 2119c9e9f97bSIlya Dryomov ret = -EINVAL; 2120c9e9f97bSIlya Dryomov goto out; 2121c9e9f97bSIlya Dryomov } 2122c9e9f97bSIlya Dryomov 2123dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2124c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 2125c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 2126c9e9f97bSIlya Dryomov goto out; 2127c9e9f97bSIlya Dryomov } 2128f46b5a66SChristoph Hellwig 21295516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 2130f46b5a66SChristoph Hellwig ret = btrfs_rm_device(root, vol_args->name); 2131f46b5a66SChristoph Hellwig 2132f46b5a66SChristoph Hellwig kfree(vol_args); 2133c9e9f97bSIlya Dryomov out: 2134c9e9f97bSIlya Dryomov mutex_unlock(&root->fs_info->volume_mutex); 2135f46b5a66SChristoph Hellwig return ret; 2136f46b5a66SChristoph Hellwig } 2137f46b5a66SChristoph Hellwig 2138475f6387SJan Schmidt static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg) 2139475f6387SJan Schmidt { 2140027ed2f0SLi Zefan struct btrfs_ioctl_fs_info_args *fi_args; 2141475f6387SJan Schmidt struct btrfs_device *device; 2142475f6387SJan Schmidt struct btrfs_device *next; 2143475f6387SJan Schmidt struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; 2144027ed2f0SLi Zefan int ret = 0; 2145475f6387SJan Schmidt 2146475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 2147475f6387SJan Schmidt return -EPERM; 2148475f6387SJan Schmidt 2149027ed2f0SLi Zefan fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL); 2150027ed2f0SLi Zefan if (!fi_args) 2151027ed2f0SLi Zefan return -ENOMEM; 2152027ed2f0SLi Zefan 2153027ed2f0SLi Zefan fi_args->num_devices = fs_devices->num_devices; 2154027ed2f0SLi Zefan memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid)); 2155475f6387SJan Schmidt 2156475f6387SJan Schmidt mutex_lock(&fs_devices->device_list_mutex); 2157475f6387SJan Schmidt list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) { 2158027ed2f0SLi Zefan if (device->devid > fi_args->max_id) 2159027ed2f0SLi Zefan fi_args->max_id = device->devid; 2160475f6387SJan Schmidt } 2161475f6387SJan Schmidt mutex_unlock(&fs_devices->device_list_mutex); 2162475f6387SJan Schmidt 2163027ed2f0SLi Zefan if (copy_to_user(arg, fi_args, sizeof(*fi_args))) 2164027ed2f0SLi Zefan ret = -EFAULT; 2165475f6387SJan Schmidt 2166027ed2f0SLi Zefan kfree(fi_args); 2167027ed2f0SLi Zefan return ret; 2168475f6387SJan Schmidt } 2169475f6387SJan Schmidt 2170475f6387SJan Schmidt static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg) 2171475f6387SJan Schmidt { 2172475f6387SJan Schmidt struct btrfs_ioctl_dev_info_args *di_args; 2173475f6387SJan Schmidt struct btrfs_device *dev; 2174475f6387SJan Schmidt struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; 2175475f6387SJan Schmidt int ret = 0; 2176475f6387SJan Schmidt char *s_uuid = NULL; 2177475f6387SJan Schmidt char empty_uuid[BTRFS_UUID_SIZE] = {0}; 2178475f6387SJan Schmidt 2179475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 2180475f6387SJan Schmidt return -EPERM; 2181475f6387SJan Schmidt 2182475f6387SJan Schmidt di_args = memdup_user(arg, sizeof(*di_args)); 2183475f6387SJan Schmidt if (IS_ERR(di_args)) 2184475f6387SJan Schmidt return PTR_ERR(di_args); 2185475f6387SJan Schmidt 2186475f6387SJan Schmidt if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0) 2187475f6387SJan Schmidt s_uuid = di_args->uuid; 2188475f6387SJan Schmidt 2189475f6387SJan Schmidt mutex_lock(&fs_devices->device_list_mutex); 2190475f6387SJan Schmidt dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL); 2191475f6387SJan Schmidt mutex_unlock(&fs_devices->device_list_mutex); 2192475f6387SJan Schmidt 2193475f6387SJan Schmidt if (!dev) { 2194475f6387SJan Schmidt ret = -ENODEV; 2195475f6387SJan Schmidt goto out; 2196475f6387SJan Schmidt } 2197475f6387SJan Schmidt 2198475f6387SJan Schmidt di_args->devid = dev->devid; 2199475f6387SJan Schmidt di_args->bytes_used = dev->bytes_used; 2200475f6387SJan Schmidt di_args->total_bytes = dev->total_bytes; 2201475f6387SJan Schmidt memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); 2202475f6387SJan Schmidt strncpy(di_args->path, dev->name, sizeof(di_args->path)); 2203475f6387SJan Schmidt 2204475f6387SJan Schmidt out: 2205475f6387SJan Schmidt if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) 2206475f6387SJan Schmidt ret = -EFAULT; 2207475f6387SJan Schmidt 2208475f6387SJan Schmidt kfree(di_args); 2209475f6387SJan Schmidt return ret; 2210475f6387SJan Schmidt } 2211475f6387SJan Schmidt 221276dda93cSYan, Zheng static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, 2213b2950863SChristoph Hellwig u64 off, u64 olen, u64 destoff) 2214f46b5a66SChristoph Hellwig { 2215f46b5a66SChristoph Hellwig struct inode *inode = fdentry(file)->d_inode; 2216f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 2217f46b5a66SChristoph Hellwig struct file *src_file; 2218f46b5a66SChristoph Hellwig struct inode *src; 2219f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 2220f46b5a66SChristoph Hellwig struct btrfs_path *path; 2221f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 2222ae01a0abSYan Zheng char *buf; 2223ae01a0abSYan Zheng struct btrfs_key key; 2224f46b5a66SChristoph Hellwig u32 nritems; 2225f46b5a66SChristoph Hellwig int slot; 2226ae01a0abSYan Zheng int ret; 2227c5c9cd4dSSage Weil u64 len = olen; 2228c5c9cd4dSSage Weil u64 bs = root->fs_info->sb->s_blocksize; 2229c5c9cd4dSSage Weil u64 hint_byte; 2230d20f7043SChris Mason 2231c5c9cd4dSSage Weil /* 2232c5c9cd4dSSage Weil * TODO: 2233c5c9cd4dSSage Weil * - split compressed inline extents. annoying: we need to 2234c5c9cd4dSSage Weil * decompress into destination's address_space (the file offset 2235c5c9cd4dSSage Weil * may change, so source mapping won't do), then recompress (or 2236c5c9cd4dSSage Weil * otherwise reinsert) a subrange. 2237c5c9cd4dSSage Weil * - allow ranges within the same file to be cloned (provided 2238c5c9cd4dSSage Weil * they don't overlap)? 2239c5c9cd4dSSage Weil */ 2240c5c9cd4dSSage Weil 2241e441d54dSChris Mason /* the destination must be opened for writing */ 22422ebc3464SDan Rosenberg if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND)) 2243e441d54dSChris Mason return -EINVAL; 2244e441d54dSChris Mason 2245b83cc969SLi Zefan if (btrfs_root_readonly(root)) 2246b83cc969SLi Zefan return -EROFS; 2247b83cc969SLi Zefan 2248c146afadSYan Zheng ret = mnt_want_write(file->f_path.mnt); 2249c146afadSYan Zheng if (ret) 2250c146afadSYan Zheng return ret; 2251c146afadSYan Zheng 2252c5c9cd4dSSage Weil src_file = fget(srcfd); 2253ab67b7c1SYan Zheng if (!src_file) { 2254ab67b7c1SYan Zheng ret = -EBADF; 2255ab67b7c1SYan Zheng goto out_drop_write; 2256ab67b7c1SYan Zheng } 22575dc64164SDan Rosenberg 2258f46b5a66SChristoph Hellwig src = src_file->f_dentry->d_inode; 2259f46b5a66SChristoph Hellwig 2260c5c9cd4dSSage Weil ret = -EINVAL; 2261c5c9cd4dSSage Weil if (src == inode) 2262c5c9cd4dSSage Weil goto out_fput; 2263c5c9cd4dSSage Weil 22645dc64164SDan Rosenberg /* the src must be open for reading */ 22655dc64164SDan Rosenberg if (!(src_file->f_mode & FMODE_READ)) 22665dc64164SDan Rosenberg goto out_fput; 22675dc64164SDan Rosenberg 22680e7b824cSLi Zefan /* don't make the dst file partly checksummed */ 22690e7b824cSLi Zefan if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != 22700e7b824cSLi Zefan (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) 22710e7b824cSLi Zefan goto out_fput; 22720e7b824cSLi Zefan 2273ae01a0abSYan Zheng ret = -EISDIR; 2274ae01a0abSYan Zheng if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode)) 2275f46b5a66SChristoph Hellwig goto out_fput; 2276f46b5a66SChristoph Hellwig 2277ae01a0abSYan Zheng ret = -EXDEV; 2278ae01a0abSYan Zheng if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root) 2279ae01a0abSYan Zheng goto out_fput; 2280ae01a0abSYan Zheng 2281ae01a0abSYan Zheng ret = -ENOMEM; 2282ae01a0abSYan Zheng buf = vmalloc(btrfs_level_size(root, 0)); 2283ae01a0abSYan Zheng if (!buf) 2284ae01a0abSYan Zheng goto out_fput; 2285ae01a0abSYan Zheng 2286ae01a0abSYan Zheng path = btrfs_alloc_path(); 2287ae01a0abSYan Zheng if (!path) { 2288ae01a0abSYan Zheng vfree(buf); 2289ae01a0abSYan Zheng goto out_fput; 2290ae01a0abSYan Zheng } 2291ae01a0abSYan Zheng path->reada = 2; 2292ae01a0abSYan Zheng 2293f46b5a66SChristoph Hellwig if (inode < src) { 2294fccdae43SSage Weil mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT); 2295fccdae43SSage Weil mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD); 2296f46b5a66SChristoph Hellwig } else { 2297fccdae43SSage Weil mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT); 2298fccdae43SSage Weil mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD); 2299f46b5a66SChristoph Hellwig } 2300f46b5a66SChristoph Hellwig 2301c5c9cd4dSSage Weil /* determine range to clone */ 2302c5c9cd4dSSage Weil ret = -EINVAL; 23032ebc3464SDan Rosenberg if (off + len > src->i_size || off + len < off) 2304f46b5a66SChristoph Hellwig goto out_unlock; 2305c5c9cd4dSSage Weil if (len == 0) 2306c5c9cd4dSSage Weil olen = len = src->i_size - off; 2307c5c9cd4dSSage Weil /* if we extend to eof, continue to block boundary */ 2308c5c9cd4dSSage Weil if (off + len == src->i_size) 23092a6b8daeSLi Zefan len = ALIGN(src->i_size, bs) - off; 2310c5c9cd4dSSage Weil 2311c5c9cd4dSSage Weil /* verify the end result is block aligned */ 23122a6b8daeSLi Zefan if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) || 23132a6b8daeSLi Zefan !IS_ALIGNED(destoff, bs)) 2314c5c9cd4dSSage Weil goto out_unlock; 2315c5c9cd4dSSage Weil 2316d525e8abSLi Zefan if (destoff > inode->i_size) { 2317d525e8abSLi Zefan ret = btrfs_cont_expand(inode, inode->i_size, destoff); 2318d525e8abSLi Zefan if (ret) 2319d525e8abSLi Zefan goto out_unlock; 2320d525e8abSLi Zefan } 2321d525e8abSLi Zefan 232271ef0786SLi Zefan /* truncate page cache pages from target inode range */ 232371ef0786SLi Zefan truncate_inode_pages_range(&inode->i_data, destoff, 232471ef0786SLi Zefan PAGE_CACHE_ALIGN(destoff + len) - 1); 232571ef0786SLi Zefan 2326f46b5a66SChristoph Hellwig /* do any pending delalloc/csum calc on src, one way or 2327f46b5a66SChristoph Hellwig another, and lock file content */ 2328f46b5a66SChristoph Hellwig while (1) { 232931840ae1SZheng Yan struct btrfs_ordered_extent *ordered; 2330c5c9cd4dSSage Weil lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS); 23319a019196SSage Weil ordered = btrfs_lookup_first_ordered_extent(src, off+len); 23329a019196SSage Weil if (!ordered && 23339a019196SSage Weil !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len, 23349a019196SSage Weil EXTENT_DELALLOC, 0, NULL)) 2335f46b5a66SChristoph Hellwig break; 2336c5c9cd4dSSage Weil unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS); 2337ae01a0abSYan Zheng if (ordered) 2338ae01a0abSYan Zheng btrfs_put_ordered_extent(ordered); 23399a019196SSage Weil btrfs_wait_ordered_range(src, off, len); 2340f46b5a66SChristoph Hellwig } 2341f46b5a66SChristoph Hellwig 2342c5c9cd4dSSage Weil /* clone data */ 234333345d01SLi Zefan key.objectid = btrfs_ino(src); 2344ae01a0abSYan Zheng key.type = BTRFS_EXTENT_DATA_KEY; 2345ae01a0abSYan Zheng key.offset = 0; 2346f46b5a66SChristoph Hellwig 2347f46b5a66SChristoph Hellwig while (1) { 2348f46b5a66SChristoph Hellwig /* 2349f46b5a66SChristoph Hellwig * note the key will change type as we walk through the 2350f46b5a66SChristoph Hellwig * tree. 2351f46b5a66SChristoph Hellwig */ 2352a22285a6SYan, Zheng ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 2353f46b5a66SChristoph Hellwig if (ret < 0) 2354f46b5a66SChristoph Hellwig goto out; 2355f46b5a66SChristoph Hellwig 2356ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 2357ae01a0abSYan Zheng if (path->slots[0] >= nritems) { 2358f46b5a66SChristoph Hellwig ret = btrfs_next_leaf(root, path); 2359f46b5a66SChristoph Hellwig if (ret < 0) 2360f46b5a66SChristoph Hellwig goto out; 2361f46b5a66SChristoph Hellwig if (ret > 0) 2362f46b5a66SChristoph Hellwig break; 2363ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 2364f46b5a66SChristoph Hellwig } 2365f46b5a66SChristoph Hellwig leaf = path->nodes[0]; 2366f46b5a66SChristoph Hellwig slot = path->slots[0]; 2367f46b5a66SChristoph Hellwig 2368ae01a0abSYan Zheng btrfs_item_key_to_cpu(leaf, &key, slot); 2369d20f7043SChris Mason if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY || 237033345d01SLi Zefan key.objectid != btrfs_ino(src)) 2371f46b5a66SChristoph Hellwig break; 2372f46b5a66SChristoph Hellwig 2373c5c9cd4dSSage Weil if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) { 2374c5c9cd4dSSage Weil struct btrfs_file_extent_item *extent; 2375c5c9cd4dSSage Weil int type; 237631840ae1SZheng Yan u32 size; 237731840ae1SZheng Yan struct btrfs_key new_key; 2378c5c9cd4dSSage Weil u64 disko = 0, diskl = 0; 2379c5c9cd4dSSage Weil u64 datao = 0, datal = 0; 2380c5c9cd4dSSage Weil u8 comp; 2381b5384d48SSage Weil u64 endoff; 238231840ae1SZheng Yan 238331840ae1SZheng Yan size = btrfs_item_size_nr(leaf, slot); 238431840ae1SZheng Yan read_extent_buffer(leaf, buf, 238531840ae1SZheng Yan btrfs_item_ptr_offset(leaf, slot), 238631840ae1SZheng Yan size); 2387c5c9cd4dSSage Weil 2388c5c9cd4dSSage Weil extent = btrfs_item_ptr(leaf, slot, 2389c5c9cd4dSSage Weil struct btrfs_file_extent_item); 2390c5c9cd4dSSage Weil comp = btrfs_file_extent_compression(leaf, extent); 2391c5c9cd4dSSage Weil type = btrfs_file_extent_type(leaf, extent); 2392c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 2393c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 2394d397712bSChris Mason disko = btrfs_file_extent_disk_bytenr(leaf, 2395d397712bSChris Mason extent); 2396d397712bSChris Mason diskl = btrfs_file_extent_disk_num_bytes(leaf, 2397d397712bSChris Mason extent); 2398c5c9cd4dSSage Weil datao = btrfs_file_extent_offset(leaf, extent); 2399d397712bSChris Mason datal = btrfs_file_extent_num_bytes(leaf, 2400d397712bSChris Mason extent); 2401c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 2402c5c9cd4dSSage Weil /* take upper bound, may be compressed */ 2403c5c9cd4dSSage Weil datal = btrfs_file_extent_ram_bytes(leaf, 2404c5c9cd4dSSage Weil extent); 2405c5c9cd4dSSage Weil } 2406b3b4aa74SDavid Sterba btrfs_release_path(path); 240731840ae1SZheng Yan 2408050006a7SSage Weil if (key.offset + datal <= off || 2409c5c9cd4dSSage Weil key.offset >= off+len) 2410c5c9cd4dSSage Weil goto next; 2411c5c9cd4dSSage Weil 241231840ae1SZheng Yan memcpy(&new_key, &key, sizeof(new_key)); 241333345d01SLi Zefan new_key.objectid = btrfs_ino(inode); 24144d728ec7SLi Zefan if (off <= key.offset) 2415c5c9cd4dSSage Weil new_key.offset = key.offset + destoff - off; 24164d728ec7SLi Zefan else 24174d728ec7SLi Zefan new_key.offset = destoff; 2418c5c9cd4dSSage Weil 2419b6f3409bSSage Weil /* 2420b6f3409bSSage Weil * 1 - adjusting old extent (we may have to split it) 2421b6f3409bSSage Weil * 1 - add new extent 2422b6f3409bSSage Weil * 1 - inode update 2423b6f3409bSSage Weil */ 2424b6f3409bSSage Weil trans = btrfs_start_transaction(root, 3); 2425a22285a6SYan, Zheng if (IS_ERR(trans)) { 2426a22285a6SYan, Zheng ret = PTR_ERR(trans); 2427a22285a6SYan, Zheng goto out; 2428a22285a6SYan, Zheng } 2429a22285a6SYan, Zheng 2430c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 2431c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 2432d72c0842SLi Zefan /* 2433d72c0842SLi Zefan * a | --- range to clone ---| b 2434d72c0842SLi Zefan * | ------------- extent ------------- | 2435d72c0842SLi Zefan */ 2436d72c0842SLi Zefan 2437d72c0842SLi Zefan /* substract range b */ 2438d72c0842SLi Zefan if (key.offset + datal > off + len) 2439d72c0842SLi Zefan datal = off + len - key.offset; 2440d72c0842SLi Zefan 2441d72c0842SLi Zefan /* substract range a */ 2442a22285a6SYan, Zheng if (off > key.offset) { 2443a22285a6SYan, Zheng datao += off - key.offset; 2444a22285a6SYan, Zheng datal -= off - key.offset; 2445a22285a6SYan, Zheng } 2446a22285a6SYan, Zheng 2447a22285a6SYan, Zheng ret = btrfs_drop_extents(trans, inode, 2448a22285a6SYan, Zheng new_key.offset, 2449a22285a6SYan, Zheng new_key.offset + datal, 2450a22285a6SYan, Zheng &hint_byte, 1); 2451a22285a6SYan, Zheng BUG_ON(ret); 2452a22285a6SYan, Zheng 245331840ae1SZheng Yan ret = btrfs_insert_empty_item(trans, root, path, 245431840ae1SZheng Yan &new_key, size); 2455a22285a6SYan, Zheng BUG_ON(ret); 245631840ae1SZheng Yan 245731840ae1SZheng Yan leaf = path->nodes[0]; 245831840ae1SZheng Yan slot = path->slots[0]; 245931840ae1SZheng Yan write_extent_buffer(leaf, buf, 246031840ae1SZheng Yan btrfs_item_ptr_offset(leaf, slot), 246131840ae1SZheng Yan size); 2462ae01a0abSYan Zheng 2463f46b5a66SChristoph Hellwig extent = btrfs_item_ptr(leaf, slot, 2464f46b5a66SChristoph Hellwig struct btrfs_file_extent_item); 2465c5c9cd4dSSage Weil 2466c5c9cd4dSSage Weil /* disko == 0 means it's a hole */ 2467c5c9cd4dSSage Weil if (!disko) 2468c5c9cd4dSSage Weil datao = 0; 2469c5c9cd4dSSage Weil 2470c5c9cd4dSSage Weil btrfs_set_file_extent_offset(leaf, extent, 2471c5c9cd4dSSage Weil datao); 2472c5c9cd4dSSage Weil btrfs_set_file_extent_num_bytes(leaf, extent, 2473c5c9cd4dSSage Weil datal); 2474c5c9cd4dSSage Weil if (disko) { 2475c5c9cd4dSSage Weil inode_add_bytes(inode, datal); 2476ae01a0abSYan Zheng ret = btrfs_inc_extent_ref(trans, root, 24775d4f98a2SYan Zheng disko, diskl, 0, 2478f46b5a66SChristoph Hellwig root->root_key.objectid, 247933345d01SLi Zefan btrfs_ino(inode), 248066d7e7f0SArne Jansen new_key.offset - datao, 248166d7e7f0SArne Jansen 0); 2482ae01a0abSYan Zheng BUG_ON(ret); 2483f46b5a66SChristoph Hellwig } 2484c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 2485c5c9cd4dSSage Weil u64 skip = 0; 2486c5c9cd4dSSage Weil u64 trim = 0; 2487c5c9cd4dSSage Weil if (off > key.offset) { 2488c5c9cd4dSSage Weil skip = off - key.offset; 2489c5c9cd4dSSage Weil new_key.offset += skip; 249031840ae1SZheng Yan } 2491d397712bSChris Mason 2492c5c9cd4dSSage Weil if (key.offset + datal > off+len) 2493c5c9cd4dSSage Weil trim = key.offset + datal - (off+len); 2494d397712bSChris Mason 2495c5c9cd4dSSage Weil if (comp && (skip || trim)) { 2496c5c9cd4dSSage Weil ret = -EINVAL; 2497a22285a6SYan, Zheng btrfs_end_transaction(trans, root); 2498c5c9cd4dSSage Weil goto out; 249931840ae1SZheng Yan } 2500c5c9cd4dSSage Weil size -= skip + trim; 2501c5c9cd4dSSage Weil datal -= skip + trim; 2502a22285a6SYan, Zheng 2503a22285a6SYan, Zheng ret = btrfs_drop_extents(trans, inode, 2504a22285a6SYan, Zheng new_key.offset, 2505a22285a6SYan, Zheng new_key.offset + datal, 2506a22285a6SYan, Zheng &hint_byte, 1); 2507a22285a6SYan, Zheng BUG_ON(ret); 2508a22285a6SYan, Zheng 2509c5c9cd4dSSage Weil ret = btrfs_insert_empty_item(trans, root, path, 2510c5c9cd4dSSage Weil &new_key, size); 2511a22285a6SYan, Zheng BUG_ON(ret); 2512c5c9cd4dSSage Weil 2513c5c9cd4dSSage Weil if (skip) { 2514d397712bSChris Mason u32 start = 2515d397712bSChris Mason btrfs_file_extent_calc_inline_size(0); 2516c5c9cd4dSSage Weil memmove(buf+start, buf+start+skip, 2517c5c9cd4dSSage Weil datal); 2518c5c9cd4dSSage Weil } 2519c5c9cd4dSSage Weil 2520c5c9cd4dSSage Weil leaf = path->nodes[0]; 2521c5c9cd4dSSage Weil slot = path->slots[0]; 2522c5c9cd4dSSage Weil write_extent_buffer(leaf, buf, 2523c5c9cd4dSSage Weil btrfs_item_ptr_offset(leaf, slot), 2524c5c9cd4dSSage Weil size); 2525c5c9cd4dSSage Weil inode_add_bytes(inode, datal); 2526c5c9cd4dSSage Weil } 2527c5c9cd4dSSage Weil 2528c5c9cd4dSSage Weil btrfs_mark_buffer_dirty(leaf); 2529b3b4aa74SDavid Sterba btrfs_release_path(path); 2530c5c9cd4dSSage Weil 2531a22285a6SYan, Zheng inode->i_mtime = inode->i_ctime = CURRENT_TIME; 2532b5384d48SSage Weil 2533b5384d48SSage Weil /* 2534b5384d48SSage Weil * we round up to the block size at eof when 2535b5384d48SSage Weil * determining which extents to clone above, 2536b5384d48SSage Weil * but shouldn't round up the file size 2537b5384d48SSage Weil */ 2538b5384d48SSage Weil endoff = new_key.offset + datal; 25395f3888ffSLi Zefan if (endoff > destoff+olen) 25405f3888ffSLi Zefan endoff = destoff+olen; 2541b5384d48SSage Weil if (endoff > inode->i_size) 2542b5384d48SSage Weil btrfs_i_size_write(inode, endoff); 2543b5384d48SSage Weil 2544a22285a6SYan, Zheng ret = btrfs_update_inode(trans, root, inode); 2545a22285a6SYan, Zheng BUG_ON(ret); 2546a22285a6SYan, Zheng btrfs_end_transaction(trans, root); 2547a22285a6SYan, Zheng } 2548c5c9cd4dSSage Weil next: 2549b3b4aa74SDavid Sterba btrfs_release_path(path); 2550ae01a0abSYan Zheng key.offset++; 2551ae01a0abSYan Zheng } 2552f46b5a66SChristoph Hellwig ret = 0; 2553f46b5a66SChristoph Hellwig out: 2554b3b4aa74SDavid Sterba btrfs_release_path(path); 2555c5c9cd4dSSage Weil unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS); 2556f46b5a66SChristoph Hellwig out_unlock: 2557f46b5a66SChristoph Hellwig mutex_unlock(&src->i_mutex); 2558f46b5a66SChristoph Hellwig mutex_unlock(&inode->i_mutex); 2559ae01a0abSYan Zheng vfree(buf); 2560ae01a0abSYan Zheng btrfs_free_path(path); 2561f46b5a66SChristoph Hellwig out_fput: 2562f46b5a66SChristoph Hellwig fput(src_file); 2563ab67b7c1SYan Zheng out_drop_write: 2564ab67b7c1SYan Zheng mnt_drop_write(file->f_path.mnt); 2565f46b5a66SChristoph Hellwig return ret; 2566f46b5a66SChristoph Hellwig } 2567f46b5a66SChristoph Hellwig 25687a865e8aSChristoph Hellwig static long btrfs_ioctl_clone_range(struct file *file, void __user *argp) 2569c5c9cd4dSSage Weil { 2570c5c9cd4dSSage Weil struct btrfs_ioctl_clone_range_args args; 2571c5c9cd4dSSage Weil 25727a865e8aSChristoph Hellwig if (copy_from_user(&args, argp, sizeof(args))) 2573c5c9cd4dSSage Weil return -EFAULT; 2574c5c9cd4dSSage Weil return btrfs_ioctl_clone(file, args.src_fd, args.src_offset, 2575c5c9cd4dSSage Weil args.src_length, args.dest_offset); 2576c5c9cd4dSSage Weil } 2577c5c9cd4dSSage Weil 2578f46b5a66SChristoph Hellwig /* 2579f46b5a66SChristoph Hellwig * there are many ways the trans_start and trans_end ioctls can lead 2580f46b5a66SChristoph Hellwig * to deadlocks. They should only be used by applications that 2581f46b5a66SChristoph Hellwig * basically own the machine, and have a very in depth understanding 2582f46b5a66SChristoph Hellwig * of all the possible deadlocks and enospc problems. 2583f46b5a66SChristoph Hellwig */ 2584b2950863SChristoph Hellwig static long btrfs_ioctl_trans_start(struct file *file) 2585f46b5a66SChristoph Hellwig { 2586f46b5a66SChristoph Hellwig struct inode *inode = fdentry(file)->d_inode; 2587f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 2588f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 25891ab86aedSSage Weil int ret; 2590f46b5a66SChristoph Hellwig 25911ab86aedSSage Weil ret = -EPERM; 2592df5b5520SChristoph Hellwig if (!capable(CAP_SYS_ADMIN)) 2593f46b5a66SChristoph Hellwig goto out; 25941ab86aedSSage Weil 25951ab86aedSSage Weil ret = -EINPROGRESS; 25961ab86aedSSage Weil if (file->private_data) 25971ab86aedSSage Weil goto out; 25989ca9ee09SSage Weil 2599b83cc969SLi Zefan ret = -EROFS; 2600b83cc969SLi Zefan if (btrfs_root_readonly(root)) 2601b83cc969SLi Zefan goto out; 2602b83cc969SLi Zefan 2603c146afadSYan Zheng ret = mnt_want_write(file->f_path.mnt); 2604c146afadSYan Zheng if (ret) 2605c146afadSYan Zheng goto out; 2606c146afadSYan Zheng 2607a4abeea4SJosef Bacik atomic_inc(&root->fs_info->open_ioctl_trans); 26089ca9ee09SSage Weil 2609f46b5a66SChristoph Hellwig ret = -ENOMEM; 26107a7eaa40SJosef Bacik trans = btrfs_start_ioctl_transaction(root); 2611abd30bb0STsutomu Itoh if (IS_ERR(trans)) 26121ab86aedSSage Weil goto out_drop; 26131ab86aedSSage Weil 26141ab86aedSSage Weil file->private_data = trans; 26151ab86aedSSage Weil return 0; 26161ab86aedSSage Weil 26171ab86aedSSage Weil out_drop: 2618a4abeea4SJosef Bacik atomic_dec(&root->fs_info->open_ioctl_trans); 26191ab86aedSSage Weil mnt_drop_write(file->f_path.mnt); 2620f46b5a66SChristoph Hellwig out: 2621f46b5a66SChristoph Hellwig return ret; 2622f46b5a66SChristoph Hellwig } 2623f46b5a66SChristoph Hellwig 26246ef5ed0dSJosef Bacik static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) 26256ef5ed0dSJosef Bacik { 26266ef5ed0dSJosef Bacik struct inode *inode = fdentry(file)->d_inode; 26276ef5ed0dSJosef Bacik struct btrfs_root *root = BTRFS_I(inode)->root; 26286ef5ed0dSJosef Bacik struct btrfs_root *new_root; 26296ef5ed0dSJosef Bacik struct btrfs_dir_item *di; 26306ef5ed0dSJosef Bacik struct btrfs_trans_handle *trans; 26316ef5ed0dSJosef Bacik struct btrfs_path *path; 26326ef5ed0dSJosef Bacik struct btrfs_key location; 26336ef5ed0dSJosef Bacik struct btrfs_disk_key disk_key; 26346ef5ed0dSJosef Bacik struct btrfs_super_block *disk_super; 26356ef5ed0dSJosef Bacik u64 features; 26366ef5ed0dSJosef Bacik u64 objectid = 0; 26376ef5ed0dSJosef Bacik u64 dir_id; 26386ef5ed0dSJosef Bacik 26396ef5ed0dSJosef Bacik if (!capable(CAP_SYS_ADMIN)) 26406ef5ed0dSJosef Bacik return -EPERM; 26416ef5ed0dSJosef Bacik 26426ef5ed0dSJosef Bacik if (copy_from_user(&objectid, argp, sizeof(objectid))) 26436ef5ed0dSJosef Bacik return -EFAULT; 26446ef5ed0dSJosef Bacik 26456ef5ed0dSJosef Bacik if (!objectid) 26466ef5ed0dSJosef Bacik objectid = root->root_key.objectid; 26476ef5ed0dSJosef Bacik 26486ef5ed0dSJosef Bacik location.objectid = objectid; 26496ef5ed0dSJosef Bacik location.type = BTRFS_ROOT_ITEM_KEY; 26506ef5ed0dSJosef Bacik location.offset = (u64)-1; 26516ef5ed0dSJosef Bacik 26526ef5ed0dSJosef Bacik new_root = btrfs_read_fs_root_no_name(root->fs_info, &location); 26536ef5ed0dSJosef Bacik if (IS_ERR(new_root)) 26546ef5ed0dSJosef Bacik return PTR_ERR(new_root); 26556ef5ed0dSJosef Bacik 26566ef5ed0dSJosef Bacik if (btrfs_root_refs(&new_root->root_item) == 0) 26576ef5ed0dSJosef Bacik return -ENOENT; 26586ef5ed0dSJosef Bacik 26596ef5ed0dSJosef Bacik path = btrfs_alloc_path(); 26606ef5ed0dSJosef Bacik if (!path) 26616ef5ed0dSJosef Bacik return -ENOMEM; 26626ef5ed0dSJosef Bacik path->leave_spinning = 1; 26636ef5ed0dSJosef Bacik 26646ef5ed0dSJosef Bacik trans = btrfs_start_transaction(root, 1); 266598d5dc13STsutomu Itoh if (IS_ERR(trans)) { 26666ef5ed0dSJosef Bacik btrfs_free_path(path); 266798d5dc13STsutomu Itoh return PTR_ERR(trans); 26686ef5ed0dSJosef Bacik } 26696ef5ed0dSJosef Bacik 26706c41761fSDavid Sterba dir_id = btrfs_super_root_dir(root->fs_info->super_copy); 26716ef5ed0dSJosef Bacik di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path, 26726ef5ed0dSJosef Bacik dir_id, "default", 7, 1); 2673cf1e99a4SDan Carpenter if (IS_ERR_OR_NULL(di)) { 26746ef5ed0dSJosef Bacik btrfs_free_path(path); 26756ef5ed0dSJosef Bacik btrfs_end_transaction(trans, root); 26766ef5ed0dSJosef Bacik printk(KERN_ERR "Umm, you don't have the default dir item, " 26776ef5ed0dSJosef Bacik "this isn't going to work\n"); 26786ef5ed0dSJosef Bacik return -ENOENT; 26796ef5ed0dSJosef Bacik } 26806ef5ed0dSJosef Bacik 26816ef5ed0dSJosef Bacik btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); 26826ef5ed0dSJosef Bacik btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); 26836ef5ed0dSJosef Bacik btrfs_mark_buffer_dirty(path->nodes[0]); 26846ef5ed0dSJosef Bacik btrfs_free_path(path); 26856ef5ed0dSJosef Bacik 26866c41761fSDavid Sterba disk_super = root->fs_info->super_copy; 26876ef5ed0dSJosef Bacik features = btrfs_super_incompat_flags(disk_super); 26886ef5ed0dSJosef Bacik if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) { 26896ef5ed0dSJosef Bacik features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL; 26906ef5ed0dSJosef Bacik btrfs_set_super_incompat_flags(disk_super, features); 26916ef5ed0dSJosef Bacik } 26926ef5ed0dSJosef Bacik btrfs_end_transaction(trans, root); 26936ef5ed0dSJosef Bacik 26946ef5ed0dSJosef Bacik return 0; 26956ef5ed0dSJosef Bacik } 26966ef5ed0dSJosef Bacik 2697bf5fc093SJosef Bacik static void get_block_group_info(struct list_head *groups_list, 2698bf5fc093SJosef Bacik struct btrfs_ioctl_space_info *space) 2699bf5fc093SJosef Bacik { 2700bf5fc093SJosef Bacik struct btrfs_block_group_cache *block_group; 2701bf5fc093SJosef Bacik 2702bf5fc093SJosef Bacik space->total_bytes = 0; 2703bf5fc093SJosef Bacik space->used_bytes = 0; 2704bf5fc093SJosef Bacik space->flags = 0; 2705bf5fc093SJosef Bacik list_for_each_entry(block_group, groups_list, list) { 2706bf5fc093SJosef Bacik space->flags = block_group->flags; 2707bf5fc093SJosef Bacik space->total_bytes += block_group->key.offset; 2708bf5fc093SJosef Bacik space->used_bytes += 2709bf5fc093SJosef Bacik btrfs_block_group_used(&block_group->item); 2710bf5fc093SJosef Bacik } 2711bf5fc093SJosef Bacik } 2712bf5fc093SJosef Bacik 27131406e432SJosef Bacik long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg) 27141406e432SJosef Bacik { 27151406e432SJosef Bacik struct btrfs_ioctl_space_args space_args; 27161406e432SJosef Bacik struct btrfs_ioctl_space_info space; 27171406e432SJosef Bacik struct btrfs_ioctl_space_info *dest; 27187fde62bfSChris Mason struct btrfs_ioctl_space_info *dest_orig; 271913f2696fSDaniel J Blueman struct btrfs_ioctl_space_info __user *user_dest; 27201406e432SJosef Bacik struct btrfs_space_info *info; 2721bf5fc093SJosef Bacik u64 types[] = {BTRFS_BLOCK_GROUP_DATA, 2722bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_SYSTEM, 2723bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_METADATA, 2724bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA}; 2725bf5fc093SJosef Bacik int num_types = 4; 27267fde62bfSChris Mason int alloc_size; 27271406e432SJosef Bacik int ret = 0; 272851788b1bSDan Rosenberg u64 slot_count = 0; 2729bf5fc093SJosef Bacik int i, c; 27301406e432SJosef Bacik 27311406e432SJosef Bacik if (copy_from_user(&space_args, 27321406e432SJosef Bacik (struct btrfs_ioctl_space_args __user *)arg, 27331406e432SJosef Bacik sizeof(space_args))) 27341406e432SJosef Bacik return -EFAULT; 27351406e432SJosef Bacik 2736bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 2737bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 2738bf5fc093SJosef Bacik 2739bf5fc093SJosef Bacik info = NULL; 27407fde62bfSChris Mason rcu_read_lock(); 2741bf5fc093SJosef Bacik list_for_each_entry_rcu(tmp, &root->fs_info->space_info, 2742bf5fc093SJosef Bacik list) { 2743bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 2744bf5fc093SJosef Bacik info = tmp; 2745bf5fc093SJosef Bacik break; 2746bf5fc093SJosef Bacik } 2747bf5fc093SJosef Bacik } 27487fde62bfSChris Mason rcu_read_unlock(); 27491406e432SJosef Bacik 2750bf5fc093SJosef Bacik if (!info) 2751bf5fc093SJosef Bacik continue; 2752bf5fc093SJosef Bacik 2753bf5fc093SJosef Bacik down_read(&info->groups_sem); 2754bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 2755bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) 2756bf5fc093SJosef Bacik slot_count++; 2757bf5fc093SJosef Bacik } 2758bf5fc093SJosef Bacik up_read(&info->groups_sem); 2759bf5fc093SJosef Bacik } 2760bf5fc093SJosef Bacik 27617fde62bfSChris Mason /* space_slots == 0 means they are asking for a count */ 27627fde62bfSChris Mason if (space_args.space_slots == 0) { 27637fde62bfSChris Mason space_args.total_spaces = slot_count; 27647fde62bfSChris Mason goto out; 27657fde62bfSChris Mason } 2766bf5fc093SJosef Bacik 276751788b1bSDan Rosenberg slot_count = min_t(u64, space_args.space_slots, slot_count); 2768bf5fc093SJosef Bacik 27697fde62bfSChris Mason alloc_size = sizeof(*dest) * slot_count; 2770bf5fc093SJosef Bacik 27717fde62bfSChris Mason /* we generally have at most 6 or so space infos, one for each raid 27727fde62bfSChris Mason * level. So, a whole page should be more than enough for everyone 27737fde62bfSChris Mason */ 27747fde62bfSChris Mason if (alloc_size > PAGE_CACHE_SIZE) 27757fde62bfSChris Mason return -ENOMEM; 27767fde62bfSChris Mason 27777fde62bfSChris Mason space_args.total_spaces = 0; 27787fde62bfSChris Mason dest = kmalloc(alloc_size, GFP_NOFS); 27797fde62bfSChris Mason if (!dest) 27807fde62bfSChris Mason return -ENOMEM; 27817fde62bfSChris Mason dest_orig = dest; 27827fde62bfSChris Mason 27837fde62bfSChris Mason /* now we have a buffer to copy into */ 2784bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 2785bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 2786bf5fc093SJosef Bacik 278751788b1bSDan Rosenberg if (!slot_count) 278851788b1bSDan Rosenberg break; 278951788b1bSDan Rosenberg 2790bf5fc093SJosef Bacik info = NULL; 27911406e432SJosef Bacik rcu_read_lock(); 2792bf5fc093SJosef Bacik list_for_each_entry_rcu(tmp, &root->fs_info->space_info, 2793bf5fc093SJosef Bacik list) { 2794bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 2795bf5fc093SJosef Bacik info = tmp; 27967fde62bfSChris Mason break; 2797bf5fc093SJosef Bacik } 2798bf5fc093SJosef Bacik } 2799bf5fc093SJosef Bacik rcu_read_unlock(); 28007fde62bfSChris Mason 2801bf5fc093SJosef Bacik if (!info) 2802bf5fc093SJosef Bacik continue; 2803bf5fc093SJosef Bacik down_read(&info->groups_sem); 2804bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 2805bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) { 2806bf5fc093SJosef Bacik get_block_group_info(&info->block_groups[c], 2807bf5fc093SJosef Bacik &space); 28087fde62bfSChris Mason memcpy(dest, &space, sizeof(space)); 28091406e432SJosef Bacik dest++; 28101406e432SJosef Bacik space_args.total_spaces++; 281151788b1bSDan Rosenberg slot_count--; 28121406e432SJosef Bacik } 281351788b1bSDan Rosenberg if (!slot_count) 281451788b1bSDan Rosenberg break; 2815bf5fc093SJosef Bacik } 2816bf5fc093SJosef Bacik up_read(&info->groups_sem); 2817bf5fc093SJosef Bacik } 28181406e432SJosef Bacik 28197fde62bfSChris Mason user_dest = (struct btrfs_ioctl_space_info *) 28207fde62bfSChris Mason (arg + sizeof(struct btrfs_ioctl_space_args)); 28217fde62bfSChris Mason 28227fde62bfSChris Mason if (copy_to_user(user_dest, dest_orig, alloc_size)) 28237fde62bfSChris Mason ret = -EFAULT; 28247fde62bfSChris Mason 28257fde62bfSChris Mason kfree(dest_orig); 28267fde62bfSChris Mason out: 28277fde62bfSChris Mason if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) 28281406e432SJosef Bacik ret = -EFAULT; 28291406e432SJosef Bacik 28301406e432SJosef Bacik return ret; 28311406e432SJosef Bacik } 28321406e432SJosef Bacik 2833f46b5a66SChristoph Hellwig /* 2834f46b5a66SChristoph Hellwig * there are many ways the trans_start and trans_end ioctls can lead 2835f46b5a66SChristoph Hellwig * to deadlocks. They should only be used by applications that 2836f46b5a66SChristoph Hellwig * basically own the machine, and have a very in depth understanding 2837f46b5a66SChristoph Hellwig * of all the possible deadlocks and enospc problems. 2838f46b5a66SChristoph Hellwig */ 2839f46b5a66SChristoph Hellwig long btrfs_ioctl_trans_end(struct file *file) 2840f46b5a66SChristoph Hellwig { 2841f46b5a66SChristoph Hellwig struct inode *inode = fdentry(file)->d_inode; 2842f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 2843f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 2844f46b5a66SChristoph Hellwig 2845f46b5a66SChristoph Hellwig trans = file->private_data; 28461ab86aedSSage Weil if (!trans) 28471ab86aedSSage Weil return -EINVAL; 2848b214107eSChristoph Hellwig file->private_data = NULL; 28499ca9ee09SSage Weil 28501ab86aedSSage Weil btrfs_end_transaction(trans, root); 28511ab86aedSSage Weil 2852a4abeea4SJosef Bacik atomic_dec(&root->fs_info->open_ioctl_trans); 28539ca9ee09SSage Weil 2854cfc8ea87SSage Weil mnt_drop_write(file->f_path.mnt); 28551ab86aedSSage Weil return 0; 2856f46b5a66SChristoph Hellwig } 2857f46b5a66SChristoph Hellwig 285846204592SSage Weil static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp) 285946204592SSage Weil { 286046204592SSage Weil struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root; 286146204592SSage Weil struct btrfs_trans_handle *trans; 286246204592SSage Weil u64 transid; 2863db5b493aSTsutomu Itoh int ret; 286446204592SSage Weil 286546204592SSage Weil trans = btrfs_start_transaction(root, 0); 286698d5dc13STsutomu Itoh if (IS_ERR(trans)) 286798d5dc13STsutomu Itoh return PTR_ERR(trans); 286846204592SSage Weil transid = trans->transid; 2869db5b493aSTsutomu Itoh ret = btrfs_commit_transaction_async(trans, root, 0); 28708b2b2d3cSTsutomu Itoh if (ret) { 28718b2b2d3cSTsutomu Itoh btrfs_end_transaction(trans, root); 2872db5b493aSTsutomu Itoh return ret; 28738b2b2d3cSTsutomu Itoh } 287446204592SSage Weil 287546204592SSage Weil if (argp) 287646204592SSage Weil if (copy_to_user(argp, &transid, sizeof(transid))) 287746204592SSage Weil return -EFAULT; 287846204592SSage Weil return 0; 287946204592SSage Weil } 288046204592SSage Weil 288146204592SSage Weil static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp) 288246204592SSage Weil { 288346204592SSage Weil struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root; 288446204592SSage Weil u64 transid; 288546204592SSage Weil 288646204592SSage Weil if (argp) { 288746204592SSage Weil if (copy_from_user(&transid, argp, sizeof(transid))) 288846204592SSage Weil return -EFAULT; 288946204592SSage Weil } else { 289046204592SSage Weil transid = 0; /* current trans */ 289146204592SSage Weil } 289246204592SSage Weil return btrfs_wait_for_commit(root, transid); 289346204592SSage Weil } 289446204592SSage Weil 2895475f6387SJan Schmidt static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg) 2896475f6387SJan Schmidt { 2897475f6387SJan Schmidt int ret; 2898475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 2899475f6387SJan Schmidt 2900475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 2901475f6387SJan Schmidt return -EPERM; 2902475f6387SJan Schmidt 2903475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 2904475f6387SJan Schmidt if (IS_ERR(sa)) 2905475f6387SJan Schmidt return PTR_ERR(sa); 2906475f6387SJan Schmidt 2907475f6387SJan Schmidt ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end, 29088628764eSArne Jansen &sa->progress, sa->flags & BTRFS_SCRUB_READONLY); 2909475f6387SJan Schmidt 2910475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 2911475f6387SJan Schmidt ret = -EFAULT; 2912475f6387SJan Schmidt 2913475f6387SJan Schmidt kfree(sa); 2914475f6387SJan Schmidt return ret; 2915475f6387SJan Schmidt } 2916475f6387SJan Schmidt 2917475f6387SJan Schmidt static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg) 2918475f6387SJan Schmidt { 2919475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 2920475f6387SJan Schmidt return -EPERM; 2921475f6387SJan Schmidt 2922475f6387SJan Schmidt return btrfs_scrub_cancel(root); 2923475f6387SJan Schmidt } 2924475f6387SJan Schmidt 2925475f6387SJan Schmidt static long btrfs_ioctl_scrub_progress(struct btrfs_root *root, 2926475f6387SJan Schmidt void __user *arg) 2927475f6387SJan Schmidt { 2928475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 2929475f6387SJan Schmidt int ret; 2930475f6387SJan Schmidt 2931475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 2932475f6387SJan Schmidt return -EPERM; 2933475f6387SJan Schmidt 2934475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 2935475f6387SJan Schmidt if (IS_ERR(sa)) 2936475f6387SJan Schmidt return PTR_ERR(sa); 2937475f6387SJan Schmidt 2938475f6387SJan Schmidt ret = btrfs_scrub_progress(root, sa->devid, &sa->progress); 2939475f6387SJan Schmidt 2940475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 2941475f6387SJan Schmidt ret = -EFAULT; 2942475f6387SJan Schmidt 2943475f6387SJan Schmidt kfree(sa); 2944475f6387SJan Schmidt return ret; 2945475f6387SJan Schmidt } 2946475f6387SJan Schmidt 2947d7728c96SJan Schmidt static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) 2948d7728c96SJan Schmidt { 2949d7728c96SJan Schmidt int ret = 0; 2950d7728c96SJan Schmidt int i; 2951740c3d22SChris Mason u64 rel_ptr; 2952d7728c96SJan Schmidt int size; 2953806468f8SChris Mason struct btrfs_ioctl_ino_path_args *ipa = NULL; 2954d7728c96SJan Schmidt struct inode_fs_paths *ipath = NULL; 2955d7728c96SJan Schmidt struct btrfs_path *path; 2956d7728c96SJan Schmidt 2957d7728c96SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 2958d7728c96SJan Schmidt return -EPERM; 2959d7728c96SJan Schmidt 2960d7728c96SJan Schmidt path = btrfs_alloc_path(); 2961d7728c96SJan Schmidt if (!path) { 2962d7728c96SJan Schmidt ret = -ENOMEM; 2963d7728c96SJan Schmidt goto out; 2964d7728c96SJan Schmidt } 2965d7728c96SJan Schmidt 2966d7728c96SJan Schmidt ipa = memdup_user(arg, sizeof(*ipa)); 2967d7728c96SJan Schmidt if (IS_ERR(ipa)) { 2968d7728c96SJan Schmidt ret = PTR_ERR(ipa); 2969d7728c96SJan Schmidt ipa = NULL; 2970d7728c96SJan Schmidt goto out; 2971d7728c96SJan Schmidt } 2972d7728c96SJan Schmidt 2973d7728c96SJan Schmidt size = min_t(u32, ipa->size, 4096); 2974d7728c96SJan Schmidt ipath = init_ipath(size, root, path); 2975d7728c96SJan Schmidt if (IS_ERR(ipath)) { 2976d7728c96SJan Schmidt ret = PTR_ERR(ipath); 2977d7728c96SJan Schmidt ipath = NULL; 2978d7728c96SJan Schmidt goto out; 2979d7728c96SJan Schmidt } 2980d7728c96SJan Schmidt 2981d7728c96SJan Schmidt ret = paths_from_inode(ipa->inum, ipath); 2982d7728c96SJan Schmidt if (ret < 0) 2983d7728c96SJan Schmidt goto out; 2984d7728c96SJan Schmidt 2985d7728c96SJan Schmidt for (i = 0; i < ipath->fspath->elem_cnt; ++i) { 2986745c4d8eSJeff Mahoney rel_ptr = ipath->fspath->val[i] - 2987745c4d8eSJeff Mahoney (u64)(unsigned long)ipath->fspath->val; 2988740c3d22SChris Mason ipath->fspath->val[i] = rel_ptr; 2989d7728c96SJan Schmidt } 2990d7728c96SJan Schmidt 2991745c4d8eSJeff Mahoney ret = copy_to_user((void *)(unsigned long)ipa->fspath, 2992745c4d8eSJeff Mahoney (void *)(unsigned long)ipath->fspath, size); 2993d7728c96SJan Schmidt if (ret) { 2994d7728c96SJan Schmidt ret = -EFAULT; 2995d7728c96SJan Schmidt goto out; 2996d7728c96SJan Schmidt } 2997d7728c96SJan Schmidt 2998d7728c96SJan Schmidt out: 2999d7728c96SJan Schmidt btrfs_free_path(path); 3000d7728c96SJan Schmidt free_ipath(ipath); 3001d7728c96SJan Schmidt kfree(ipa); 3002d7728c96SJan Schmidt 3003d7728c96SJan Schmidt return ret; 3004d7728c96SJan Schmidt } 3005d7728c96SJan Schmidt 3006d7728c96SJan Schmidt static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx) 3007d7728c96SJan Schmidt { 3008d7728c96SJan Schmidt struct btrfs_data_container *inodes = ctx; 3009d7728c96SJan Schmidt const size_t c = 3 * sizeof(u64); 3010d7728c96SJan Schmidt 3011d7728c96SJan Schmidt if (inodes->bytes_left >= c) { 3012d7728c96SJan Schmidt inodes->bytes_left -= c; 3013d7728c96SJan Schmidt inodes->val[inodes->elem_cnt] = inum; 3014d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 1] = offset; 3015d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 2] = root; 3016d7728c96SJan Schmidt inodes->elem_cnt += 3; 3017d7728c96SJan Schmidt } else { 3018d7728c96SJan Schmidt inodes->bytes_missing += c - inodes->bytes_left; 3019d7728c96SJan Schmidt inodes->bytes_left = 0; 3020d7728c96SJan Schmidt inodes->elem_missed += 3; 3021d7728c96SJan Schmidt } 3022d7728c96SJan Schmidt 3023d7728c96SJan Schmidt return 0; 3024d7728c96SJan Schmidt } 3025d7728c96SJan Schmidt 3026d7728c96SJan Schmidt static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root, 3027d7728c96SJan Schmidt void __user *arg) 3028d7728c96SJan Schmidt { 3029d7728c96SJan Schmidt int ret = 0; 3030d7728c96SJan Schmidt int size; 30314692cf58SJan Schmidt u64 extent_item_pos; 3032d7728c96SJan Schmidt struct btrfs_ioctl_logical_ino_args *loi; 3033d7728c96SJan Schmidt struct btrfs_data_container *inodes = NULL; 3034d7728c96SJan Schmidt struct btrfs_path *path = NULL; 3035d7728c96SJan Schmidt struct btrfs_key key; 3036d7728c96SJan Schmidt 3037d7728c96SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 3038d7728c96SJan Schmidt return -EPERM; 3039d7728c96SJan Schmidt 3040d7728c96SJan Schmidt loi = memdup_user(arg, sizeof(*loi)); 3041d7728c96SJan Schmidt if (IS_ERR(loi)) { 3042d7728c96SJan Schmidt ret = PTR_ERR(loi); 3043d7728c96SJan Schmidt loi = NULL; 3044d7728c96SJan Schmidt goto out; 3045d7728c96SJan Schmidt } 3046d7728c96SJan Schmidt 3047d7728c96SJan Schmidt path = btrfs_alloc_path(); 3048d7728c96SJan Schmidt if (!path) { 3049d7728c96SJan Schmidt ret = -ENOMEM; 3050d7728c96SJan Schmidt goto out; 3051d7728c96SJan Schmidt } 3052d7728c96SJan Schmidt 3053d7728c96SJan Schmidt size = min_t(u32, loi->size, 4096); 3054d7728c96SJan Schmidt inodes = init_data_container(size); 3055d7728c96SJan Schmidt if (IS_ERR(inodes)) { 3056d7728c96SJan Schmidt ret = PTR_ERR(inodes); 3057d7728c96SJan Schmidt inodes = NULL; 3058d7728c96SJan Schmidt goto out; 3059d7728c96SJan Schmidt } 3060d7728c96SJan Schmidt 3061d7728c96SJan Schmidt ret = extent_from_logical(root->fs_info, loi->logical, path, &key); 30624692cf58SJan Schmidt btrfs_release_path(path); 3063d7728c96SJan Schmidt 3064d7728c96SJan Schmidt if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK) 3065d7728c96SJan Schmidt ret = -ENOENT; 3066d7728c96SJan Schmidt if (ret < 0) 3067d7728c96SJan Schmidt goto out; 3068d7728c96SJan Schmidt 30694692cf58SJan Schmidt extent_item_pos = loi->logical - key.objectid; 3070d7728c96SJan Schmidt ret = iterate_extent_inodes(root->fs_info, path, key.objectid, 30714692cf58SJan Schmidt extent_item_pos, build_ino_list, 30724692cf58SJan Schmidt inodes); 3073d7728c96SJan Schmidt 3074d7728c96SJan Schmidt if (ret < 0) 3075d7728c96SJan Schmidt goto out; 3076d7728c96SJan Schmidt 3077745c4d8eSJeff Mahoney ret = copy_to_user((void *)(unsigned long)loi->inodes, 3078745c4d8eSJeff Mahoney (void *)(unsigned long)inodes, size); 3079d7728c96SJan Schmidt if (ret) 3080d7728c96SJan Schmidt ret = -EFAULT; 3081d7728c96SJan Schmidt 3082d7728c96SJan Schmidt out: 3083d7728c96SJan Schmidt btrfs_free_path(path); 3084d7728c96SJan Schmidt kfree(inodes); 3085d7728c96SJan Schmidt kfree(loi); 3086d7728c96SJan Schmidt 3087d7728c96SJan Schmidt return ret; 3088d7728c96SJan Schmidt } 3089d7728c96SJan Schmidt 309019a39dceSIlya Dryomov void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock, 3091c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs) 3092c9e9f97bSIlya Dryomov { 3093c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl = fs_info->balance_ctl; 3094c9e9f97bSIlya Dryomov 3095c9e9f97bSIlya Dryomov bargs->flags = bctl->flags; 3096c9e9f97bSIlya Dryomov 3097837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_running)) 3098837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_RUNNING; 3099837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_pause_req)) 3100837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; 3101a7e99c69SIlya Dryomov if (atomic_read(&fs_info->balance_cancel_req)) 3102a7e99c69SIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; 3103837d5b6eSIlya Dryomov 3104c9e9f97bSIlya Dryomov memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); 3105c9e9f97bSIlya Dryomov memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); 3106c9e9f97bSIlya Dryomov memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); 310719a39dceSIlya Dryomov 310819a39dceSIlya Dryomov if (lock) { 310919a39dceSIlya Dryomov spin_lock(&fs_info->balance_lock); 311019a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 311119a39dceSIlya Dryomov spin_unlock(&fs_info->balance_lock); 311219a39dceSIlya Dryomov } else { 311319a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 311419a39dceSIlya Dryomov } 3115c9e9f97bSIlya Dryomov } 3116c9e9f97bSIlya Dryomov 3117c9e9f97bSIlya Dryomov static long btrfs_ioctl_balance(struct btrfs_root *root, void __user *arg) 3118c9e9f97bSIlya Dryomov { 3119c9e9f97bSIlya Dryomov struct btrfs_fs_info *fs_info = root->fs_info; 3120c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 3121c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl; 3122c9e9f97bSIlya Dryomov int ret; 3123c9e9f97bSIlya Dryomov 3124c9e9f97bSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 3125c9e9f97bSIlya Dryomov return -EPERM; 3126c9e9f97bSIlya Dryomov 3127c9e9f97bSIlya Dryomov if (fs_info->sb->s_flags & MS_RDONLY) 3128c9e9f97bSIlya Dryomov return -EROFS; 3129c9e9f97bSIlya Dryomov 3130c9e9f97bSIlya Dryomov mutex_lock(&fs_info->volume_mutex); 3131c9e9f97bSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 3132c9e9f97bSIlya Dryomov 3133c9e9f97bSIlya Dryomov if (arg) { 3134c9e9f97bSIlya Dryomov bargs = memdup_user(arg, sizeof(*bargs)); 3135c9e9f97bSIlya Dryomov if (IS_ERR(bargs)) { 3136c9e9f97bSIlya Dryomov ret = PTR_ERR(bargs); 3137c9e9f97bSIlya Dryomov goto out; 3138c9e9f97bSIlya Dryomov } 3139de322263SIlya Dryomov 3140de322263SIlya Dryomov if (bargs->flags & BTRFS_BALANCE_RESUME) { 3141de322263SIlya Dryomov if (!fs_info->balance_ctl) { 3142de322263SIlya Dryomov ret = -ENOTCONN; 3143de322263SIlya Dryomov goto out_bargs; 3144de322263SIlya Dryomov } 3145de322263SIlya Dryomov 3146de322263SIlya Dryomov bctl = fs_info->balance_ctl; 3147de322263SIlya Dryomov spin_lock(&fs_info->balance_lock); 3148de322263SIlya Dryomov bctl->flags |= BTRFS_BALANCE_RESUME; 3149de322263SIlya Dryomov spin_unlock(&fs_info->balance_lock); 3150de322263SIlya Dryomov 3151de322263SIlya Dryomov goto do_balance; 3152de322263SIlya Dryomov } 3153c9e9f97bSIlya Dryomov } else { 3154c9e9f97bSIlya Dryomov bargs = NULL; 3155c9e9f97bSIlya Dryomov } 3156c9e9f97bSIlya Dryomov 3157837d5b6eSIlya Dryomov if (fs_info->balance_ctl) { 3158837d5b6eSIlya Dryomov ret = -EINPROGRESS; 3159837d5b6eSIlya Dryomov goto out_bargs; 3160837d5b6eSIlya Dryomov } 3161837d5b6eSIlya Dryomov 3162c9e9f97bSIlya Dryomov bctl = kzalloc(sizeof(*bctl), GFP_NOFS); 3163c9e9f97bSIlya Dryomov if (!bctl) { 3164c9e9f97bSIlya Dryomov ret = -ENOMEM; 3165c9e9f97bSIlya Dryomov goto out_bargs; 3166c9e9f97bSIlya Dryomov } 3167c9e9f97bSIlya Dryomov 3168c9e9f97bSIlya Dryomov bctl->fs_info = fs_info; 3169c9e9f97bSIlya Dryomov if (arg) { 3170c9e9f97bSIlya Dryomov memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); 3171c9e9f97bSIlya Dryomov memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); 3172c9e9f97bSIlya Dryomov memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); 3173c9e9f97bSIlya Dryomov 3174c9e9f97bSIlya Dryomov bctl->flags = bargs->flags; 3175f43ffb60SIlya Dryomov } else { 3176f43ffb60SIlya Dryomov /* balance everything - no filters */ 3177f43ffb60SIlya Dryomov bctl->flags |= BTRFS_BALANCE_TYPE_MASK; 3178c9e9f97bSIlya Dryomov } 3179c9e9f97bSIlya Dryomov 3180de322263SIlya Dryomov do_balance: 3181c9e9f97bSIlya Dryomov ret = btrfs_balance(bctl, bargs); 3182c9e9f97bSIlya Dryomov /* 3183837d5b6eSIlya Dryomov * bctl is freed in __cancel_balance or in free_fs_info if 3184837d5b6eSIlya Dryomov * restriper was paused all the way until unmount 3185c9e9f97bSIlya Dryomov */ 3186c9e9f97bSIlya Dryomov if (arg) { 3187c9e9f97bSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 3188c9e9f97bSIlya Dryomov ret = -EFAULT; 3189c9e9f97bSIlya Dryomov } 3190c9e9f97bSIlya Dryomov 3191c9e9f97bSIlya Dryomov out_bargs: 3192c9e9f97bSIlya Dryomov kfree(bargs); 3193c9e9f97bSIlya Dryomov out: 3194c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 3195c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->volume_mutex); 3196c9e9f97bSIlya Dryomov return ret; 3197c9e9f97bSIlya Dryomov } 3198c9e9f97bSIlya Dryomov 3199837d5b6eSIlya Dryomov static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd) 3200837d5b6eSIlya Dryomov { 3201837d5b6eSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 3202837d5b6eSIlya Dryomov return -EPERM; 3203837d5b6eSIlya Dryomov 3204837d5b6eSIlya Dryomov switch (cmd) { 3205837d5b6eSIlya Dryomov case BTRFS_BALANCE_CTL_PAUSE: 3206837d5b6eSIlya Dryomov return btrfs_pause_balance(root->fs_info); 3207a7e99c69SIlya Dryomov case BTRFS_BALANCE_CTL_CANCEL: 3208a7e99c69SIlya Dryomov return btrfs_cancel_balance(root->fs_info); 3209837d5b6eSIlya Dryomov } 3210837d5b6eSIlya Dryomov 3211837d5b6eSIlya Dryomov return -EINVAL; 3212837d5b6eSIlya Dryomov } 3213837d5b6eSIlya Dryomov 321419a39dceSIlya Dryomov static long btrfs_ioctl_balance_progress(struct btrfs_root *root, 321519a39dceSIlya Dryomov void __user *arg) 321619a39dceSIlya Dryomov { 321719a39dceSIlya Dryomov struct btrfs_fs_info *fs_info = root->fs_info; 321819a39dceSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 321919a39dceSIlya Dryomov int ret = 0; 322019a39dceSIlya Dryomov 322119a39dceSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 322219a39dceSIlya Dryomov return -EPERM; 322319a39dceSIlya Dryomov 322419a39dceSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 322519a39dceSIlya Dryomov if (!fs_info->balance_ctl) { 322619a39dceSIlya Dryomov ret = -ENOTCONN; 322719a39dceSIlya Dryomov goto out; 322819a39dceSIlya Dryomov } 322919a39dceSIlya Dryomov 323019a39dceSIlya Dryomov bargs = kzalloc(sizeof(*bargs), GFP_NOFS); 323119a39dceSIlya Dryomov if (!bargs) { 323219a39dceSIlya Dryomov ret = -ENOMEM; 323319a39dceSIlya Dryomov goto out; 323419a39dceSIlya Dryomov } 323519a39dceSIlya Dryomov 323619a39dceSIlya Dryomov update_ioctl_balance_args(fs_info, 1, bargs); 323719a39dceSIlya Dryomov 323819a39dceSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 323919a39dceSIlya Dryomov ret = -EFAULT; 324019a39dceSIlya Dryomov 324119a39dceSIlya Dryomov kfree(bargs); 324219a39dceSIlya Dryomov out: 324319a39dceSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 324419a39dceSIlya Dryomov return ret; 324519a39dceSIlya Dryomov } 324619a39dceSIlya Dryomov 3247f46b5a66SChristoph Hellwig long btrfs_ioctl(struct file *file, unsigned int 3248f46b5a66SChristoph Hellwig cmd, unsigned long arg) 3249f46b5a66SChristoph Hellwig { 3250f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; 32514bcabaa3SChristoph Hellwig void __user *argp = (void __user *)arg; 3252f46b5a66SChristoph Hellwig 3253f46b5a66SChristoph Hellwig switch (cmd) { 32546cbff00fSChristoph Hellwig case FS_IOC_GETFLAGS: 32556cbff00fSChristoph Hellwig return btrfs_ioctl_getflags(file, argp); 32566cbff00fSChristoph Hellwig case FS_IOC_SETFLAGS: 32576cbff00fSChristoph Hellwig return btrfs_ioctl_setflags(file, argp); 32586cbff00fSChristoph Hellwig case FS_IOC_GETVERSION: 32596cbff00fSChristoph Hellwig return btrfs_ioctl_getversion(file, argp); 3260f7039b1dSLi Dongyang case FITRIM: 3261f7039b1dSLi Dongyang return btrfs_ioctl_fitrim(file, argp); 3262f46b5a66SChristoph Hellwig case BTRFS_IOC_SNAP_CREATE: 3263fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 0); 3264fdfb1e4fSLi Zefan case BTRFS_IOC_SNAP_CREATE_V2: 3265fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create_v2(file, argp, 0); 32663de4586cSChris Mason case BTRFS_IOC_SUBVOL_CREATE: 3267fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 1); 326876dda93cSYan, Zheng case BTRFS_IOC_SNAP_DESTROY: 326976dda93cSYan, Zheng return btrfs_ioctl_snap_destroy(file, argp); 32700caa102dSLi Zefan case BTRFS_IOC_SUBVOL_GETFLAGS: 32710caa102dSLi Zefan return btrfs_ioctl_subvol_getflags(file, argp); 32720caa102dSLi Zefan case BTRFS_IOC_SUBVOL_SETFLAGS: 32730caa102dSLi Zefan return btrfs_ioctl_subvol_setflags(file, argp); 32746ef5ed0dSJosef Bacik case BTRFS_IOC_DEFAULT_SUBVOL: 32756ef5ed0dSJosef Bacik return btrfs_ioctl_default_subvol(file, argp); 3276f46b5a66SChristoph Hellwig case BTRFS_IOC_DEFRAG: 32771e701a32SChris Mason return btrfs_ioctl_defrag(file, NULL); 32781e701a32SChris Mason case BTRFS_IOC_DEFRAG_RANGE: 32791e701a32SChris Mason return btrfs_ioctl_defrag(file, argp); 3280f46b5a66SChristoph Hellwig case BTRFS_IOC_RESIZE: 32814bcabaa3SChristoph Hellwig return btrfs_ioctl_resize(root, argp); 3282f46b5a66SChristoph Hellwig case BTRFS_IOC_ADD_DEV: 32834bcabaa3SChristoph Hellwig return btrfs_ioctl_add_dev(root, argp); 3284f46b5a66SChristoph Hellwig case BTRFS_IOC_RM_DEV: 32854bcabaa3SChristoph Hellwig return btrfs_ioctl_rm_dev(root, argp); 3286475f6387SJan Schmidt case BTRFS_IOC_FS_INFO: 3287475f6387SJan Schmidt return btrfs_ioctl_fs_info(root, argp); 3288475f6387SJan Schmidt case BTRFS_IOC_DEV_INFO: 3289475f6387SJan Schmidt return btrfs_ioctl_dev_info(root, argp); 3290f46b5a66SChristoph Hellwig case BTRFS_IOC_BALANCE: 3291c9e9f97bSIlya Dryomov return btrfs_ioctl_balance(root, NULL); 3292f46b5a66SChristoph Hellwig case BTRFS_IOC_CLONE: 3293c5c9cd4dSSage Weil return btrfs_ioctl_clone(file, arg, 0, 0, 0); 3294c5c9cd4dSSage Weil case BTRFS_IOC_CLONE_RANGE: 32957a865e8aSChristoph Hellwig return btrfs_ioctl_clone_range(file, argp); 3296f46b5a66SChristoph Hellwig case BTRFS_IOC_TRANS_START: 3297f46b5a66SChristoph Hellwig return btrfs_ioctl_trans_start(file); 3298f46b5a66SChristoph Hellwig case BTRFS_IOC_TRANS_END: 3299f46b5a66SChristoph Hellwig return btrfs_ioctl_trans_end(file); 3300ac8e9819SChris Mason case BTRFS_IOC_TREE_SEARCH: 3301ac8e9819SChris Mason return btrfs_ioctl_tree_search(file, argp); 3302ac8e9819SChris Mason case BTRFS_IOC_INO_LOOKUP: 3303ac8e9819SChris Mason return btrfs_ioctl_ino_lookup(file, argp); 3304d7728c96SJan Schmidt case BTRFS_IOC_INO_PATHS: 3305d7728c96SJan Schmidt return btrfs_ioctl_ino_to_path(root, argp); 3306d7728c96SJan Schmidt case BTRFS_IOC_LOGICAL_INO: 3307d7728c96SJan Schmidt return btrfs_ioctl_logical_to_ino(root, argp); 33081406e432SJosef Bacik case BTRFS_IOC_SPACE_INFO: 33091406e432SJosef Bacik return btrfs_ioctl_space_info(root, argp); 3310f46b5a66SChristoph Hellwig case BTRFS_IOC_SYNC: 3311f46b5a66SChristoph Hellwig btrfs_sync_fs(file->f_dentry->d_sb, 1); 3312f46b5a66SChristoph Hellwig return 0; 331346204592SSage Weil case BTRFS_IOC_START_SYNC: 331446204592SSage Weil return btrfs_ioctl_start_sync(file, argp); 331546204592SSage Weil case BTRFS_IOC_WAIT_SYNC: 331646204592SSage Weil return btrfs_ioctl_wait_sync(file, argp); 3317475f6387SJan Schmidt case BTRFS_IOC_SCRUB: 3318475f6387SJan Schmidt return btrfs_ioctl_scrub(root, argp); 3319475f6387SJan Schmidt case BTRFS_IOC_SCRUB_CANCEL: 3320475f6387SJan Schmidt return btrfs_ioctl_scrub_cancel(root, argp); 3321475f6387SJan Schmidt case BTRFS_IOC_SCRUB_PROGRESS: 3322475f6387SJan Schmidt return btrfs_ioctl_scrub_progress(root, argp); 3323c9e9f97bSIlya Dryomov case BTRFS_IOC_BALANCE_V2: 3324c9e9f97bSIlya Dryomov return btrfs_ioctl_balance(root, argp); 3325837d5b6eSIlya Dryomov case BTRFS_IOC_BALANCE_CTL: 3326837d5b6eSIlya Dryomov return btrfs_ioctl_balance_ctl(root, arg); 332719a39dceSIlya Dryomov case BTRFS_IOC_BALANCE_PROGRESS: 332819a39dceSIlya Dryomov return btrfs_ioctl_balance_progress(root, argp); 3329f46b5a66SChristoph Hellwig } 3330f46b5a66SChristoph Hellwig 3331f46b5a66SChristoph Hellwig return -ENOTTY; 3332f46b5a66SChristoph Hellwig } 3333