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" 54f46b5a66SChristoph Hellwig 556cbff00fSChristoph Hellwig /* Mask out flags that are inappropriate for the given type of inode. */ 566cbff00fSChristoph Hellwig static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags) 576cbff00fSChristoph Hellwig { 586cbff00fSChristoph Hellwig if (S_ISDIR(mode)) 596cbff00fSChristoph Hellwig return flags; 606cbff00fSChristoph Hellwig else if (S_ISREG(mode)) 616cbff00fSChristoph Hellwig return flags & ~FS_DIRSYNC_FL; 626cbff00fSChristoph Hellwig else 636cbff00fSChristoph Hellwig return flags & (FS_NODUMP_FL | FS_NOATIME_FL); 646cbff00fSChristoph Hellwig } 65f46b5a66SChristoph Hellwig 666cbff00fSChristoph Hellwig /* 676cbff00fSChristoph Hellwig * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl. 686cbff00fSChristoph Hellwig */ 696cbff00fSChristoph Hellwig static unsigned int btrfs_flags_to_ioctl(unsigned int flags) 706cbff00fSChristoph Hellwig { 716cbff00fSChristoph Hellwig unsigned int iflags = 0; 726cbff00fSChristoph Hellwig 736cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_SYNC) 746cbff00fSChristoph Hellwig iflags |= FS_SYNC_FL; 756cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_IMMUTABLE) 766cbff00fSChristoph Hellwig iflags |= FS_IMMUTABLE_FL; 776cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_APPEND) 786cbff00fSChristoph Hellwig iflags |= FS_APPEND_FL; 796cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NODUMP) 806cbff00fSChristoph Hellwig iflags |= FS_NODUMP_FL; 816cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NOATIME) 826cbff00fSChristoph Hellwig iflags |= FS_NOATIME_FL; 836cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_DIRSYNC) 846cbff00fSChristoph Hellwig iflags |= FS_DIRSYNC_FL; 85d0092bddSLi Zefan if (flags & BTRFS_INODE_NODATACOW) 86d0092bddSLi Zefan iflags |= FS_NOCOW_FL; 87d0092bddSLi Zefan 88d0092bddSLi Zefan if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS)) 89d0092bddSLi Zefan iflags |= FS_COMPR_FL; 90d0092bddSLi Zefan else if (flags & BTRFS_INODE_NOCOMPRESS) 91d0092bddSLi Zefan iflags |= FS_NOCOMP_FL; 926cbff00fSChristoph Hellwig 936cbff00fSChristoph Hellwig return iflags; 946cbff00fSChristoph Hellwig } 956cbff00fSChristoph Hellwig 966cbff00fSChristoph Hellwig /* 976cbff00fSChristoph Hellwig * Update inode->i_flags based on the btrfs internal flags. 986cbff00fSChristoph Hellwig */ 996cbff00fSChristoph Hellwig void btrfs_update_iflags(struct inode *inode) 1006cbff00fSChristoph Hellwig { 1016cbff00fSChristoph Hellwig struct btrfs_inode *ip = BTRFS_I(inode); 1026cbff00fSChristoph Hellwig 1036cbff00fSChristoph Hellwig inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC); 1046cbff00fSChristoph Hellwig 1056cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_SYNC) 1066cbff00fSChristoph Hellwig inode->i_flags |= S_SYNC; 1076cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_IMMUTABLE) 1086cbff00fSChristoph Hellwig inode->i_flags |= S_IMMUTABLE; 1096cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_APPEND) 1106cbff00fSChristoph Hellwig inode->i_flags |= S_APPEND; 1116cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_NOATIME) 1126cbff00fSChristoph Hellwig inode->i_flags |= S_NOATIME; 1136cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_DIRSYNC) 1146cbff00fSChristoph Hellwig inode->i_flags |= S_DIRSYNC; 1156cbff00fSChristoph Hellwig } 1166cbff00fSChristoph Hellwig 1176cbff00fSChristoph Hellwig /* 1186cbff00fSChristoph Hellwig * Inherit flags from the parent inode. 1196cbff00fSChristoph Hellwig * 1206cbff00fSChristoph Hellwig * Unlike extN we don't have any flags we don't want to inherit currently. 1216cbff00fSChristoph Hellwig */ 1226cbff00fSChristoph Hellwig void btrfs_inherit_iflags(struct inode *inode, struct inode *dir) 1236cbff00fSChristoph Hellwig { 1240b4dcea5SChris Mason unsigned int flags; 1250b4dcea5SChris Mason 1260b4dcea5SChris Mason if (!dir) 1270b4dcea5SChris Mason return; 1280b4dcea5SChris Mason 1290b4dcea5SChris Mason flags = BTRFS_I(dir)->flags; 1306cbff00fSChristoph Hellwig 1316cbff00fSChristoph Hellwig if (S_ISREG(inode->i_mode)) 1326cbff00fSChristoph Hellwig flags &= ~BTRFS_INODE_DIRSYNC; 1336cbff00fSChristoph Hellwig else if (!S_ISDIR(inode->i_mode)) 1346cbff00fSChristoph Hellwig flags &= (BTRFS_INODE_NODUMP | BTRFS_INODE_NOATIME); 1356cbff00fSChristoph Hellwig 1366cbff00fSChristoph Hellwig BTRFS_I(inode)->flags = flags; 1376cbff00fSChristoph Hellwig btrfs_update_iflags(inode); 1386cbff00fSChristoph Hellwig } 1396cbff00fSChristoph Hellwig 1406cbff00fSChristoph Hellwig static int btrfs_ioctl_getflags(struct file *file, void __user *arg) 1416cbff00fSChristoph Hellwig { 1426cbff00fSChristoph Hellwig struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode); 1436cbff00fSChristoph Hellwig unsigned int flags = btrfs_flags_to_ioctl(ip->flags); 1446cbff00fSChristoph Hellwig 1456cbff00fSChristoph Hellwig if (copy_to_user(arg, &flags, sizeof(flags))) 1466cbff00fSChristoph Hellwig return -EFAULT; 1476cbff00fSChristoph Hellwig return 0; 1486cbff00fSChristoph Hellwig } 1496cbff00fSChristoph Hellwig 15075e7cb7fSLiu Bo static int check_flags(unsigned int flags) 15175e7cb7fSLiu Bo { 15275e7cb7fSLiu Bo if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \ 15375e7cb7fSLiu Bo FS_NOATIME_FL | FS_NODUMP_FL | \ 15475e7cb7fSLiu Bo FS_SYNC_FL | FS_DIRSYNC_FL | \ 155e1e8fb6aSLi Zefan FS_NOCOMP_FL | FS_COMPR_FL | 156e1e8fb6aSLi Zefan FS_NOCOW_FL)) 15775e7cb7fSLiu Bo return -EOPNOTSUPP; 15875e7cb7fSLiu Bo 15975e7cb7fSLiu Bo if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL)) 16075e7cb7fSLiu Bo return -EINVAL; 16175e7cb7fSLiu Bo 16275e7cb7fSLiu Bo return 0; 16375e7cb7fSLiu Bo } 16475e7cb7fSLiu Bo 1656cbff00fSChristoph Hellwig static int btrfs_ioctl_setflags(struct file *file, void __user *arg) 1666cbff00fSChristoph Hellwig { 1676cbff00fSChristoph Hellwig struct inode *inode = file->f_path.dentry->d_inode; 1686cbff00fSChristoph Hellwig struct btrfs_inode *ip = BTRFS_I(inode); 1696cbff00fSChristoph Hellwig struct btrfs_root *root = ip->root; 1706cbff00fSChristoph Hellwig struct btrfs_trans_handle *trans; 1716cbff00fSChristoph Hellwig unsigned int flags, oldflags; 1726cbff00fSChristoph Hellwig int ret; 1736cbff00fSChristoph Hellwig 174b83cc969SLi Zefan if (btrfs_root_readonly(root)) 175b83cc969SLi Zefan return -EROFS; 176b83cc969SLi Zefan 1776cbff00fSChristoph Hellwig if (copy_from_user(&flags, arg, sizeof(flags))) 1786cbff00fSChristoph Hellwig return -EFAULT; 1796cbff00fSChristoph Hellwig 18075e7cb7fSLiu Bo ret = check_flags(flags); 18175e7cb7fSLiu Bo if (ret) 18275e7cb7fSLiu Bo return ret; 1836cbff00fSChristoph Hellwig 1842e149670SSerge E. Hallyn if (!inode_owner_or_capable(inode)) 1856cbff00fSChristoph Hellwig return -EACCES; 1866cbff00fSChristoph Hellwig 1876cbff00fSChristoph Hellwig mutex_lock(&inode->i_mutex); 1886cbff00fSChristoph Hellwig 1896cbff00fSChristoph Hellwig flags = btrfs_mask_flags(inode->i_mode, flags); 1906cbff00fSChristoph Hellwig oldflags = btrfs_flags_to_ioctl(ip->flags); 1916cbff00fSChristoph Hellwig if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) { 1926cbff00fSChristoph Hellwig if (!capable(CAP_LINUX_IMMUTABLE)) { 1936cbff00fSChristoph Hellwig ret = -EPERM; 1946cbff00fSChristoph Hellwig goto out_unlock; 1956cbff00fSChristoph Hellwig } 1966cbff00fSChristoph Hellwig } 1976cbff00fSChristoph Hellwig 1986cbff00fSChristoph Hellwig ret = mnt_want_write(file->f_path.mnt); 1996cbff00fSChristoph Hellwig if (ret) 2006cbff00fSChristoph Hellwig goto out_unlock; 2016cbff00fSChristoph Hellwig 2026cbff00fSChristoph Hellwig if (flags & FS_SYNC_FL) 2036cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_SYNC; 2046cbff00fSChristoph Hellwig else 2056cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_SYNC; 2066cbff00fSChristoph Hellwig if (flags & FS_IMMUTABLE_FL) 2076cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_IMMUTABLE; 2086cbff00fSChristoph Hellwig else 2096cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_IMMUTABLE; 2106cbff00fSChristoph Hellwig if (flags & FS_APPEND_FL) 2116cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_APPEND; 2126cbff00fSChristoph Hellwig else 2136cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_APPEND; 2146cbff00fSChristoph Hellwig if (flags & FS_NODUMP_FL) 2156cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_NODUMP; 2166cbff00fSChristoph Hellwig else 2176cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_NODUMP; 2186cbff00fSChristoph Hellwig if (flags & FS_NOATIME_FL) 2196cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_NOATIME; 2206cbff00fSChristoph Hellwig else 2216cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_NOATIME; 2226cbff00fSChristoph Hellwig if (flags & FS_DIRSYNC_FL) 2236cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_DIRSYNC; 2246cbff00fSChristoph Hellwig else 2256cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_DIRSYNC; 226e1e8fb6aSLi Zefan if (flags & FS_NOCOW_FL) 227e1e8fb6aSLi Zefan ip->flags |= BTRFS_INODE_NODATACOW; 228e1e8fb6aSLi Zefan else 229e1e8fb6aSLi Zefan ip->flags &= ~BTRFS_INODE_NODATACOW; 2306cbff00fSChristoph Hellwig 23175e7cb7fSLiu Bo /* 23275e7cb7fSLiu Bo * The COMPRESS flag can only be changed by users, while the NOCOMPRESS 23375e7cb7fSLiu Bo * flag may be changed automatically if compression code won't make 23475e7cb7fSLiu Bo * things smaller. 23575e7cb7fSLiu Bo */ 23675e7cb7fSLiu Bo if (flags & FS_NOCOMP_FL) { 23775e7cb7fSLiu Bo ip->flags &= ~BTRFS_INODE_COMPRESS; 23875e7cb7fSLiu Bo ip->flags |= BTRFS_INODE_NOCOMPRESS; 23975e7cb7fSLiu Bo } else if (flags & FS_COMPR_FL) { 24075e7cb7fSLiu Bo ip->flags |= BTRFS_INODE_COMPRESS; 24175e7cb7fSLiu Bo ip->flags &= ~BTRFS_INODE_NOCOMPRESS; 242ebcb904dSLi Zefan } else { 243ebcb904dSLi Zefan ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS); 24475e7cb7fSLiu Bo } 2456cbff00fSChristoph Hellwig 2467a7eaa40SJosef Bacik trans = btrfs_join_transaction(root); 2473612b495STsutomu Itoh BUG_ON(IS_ERR(trans)); 2486cbff00fSChristoph Hellwig 2496cbff00fSChristoph Hellwig ret = btrfs_update_inode(trans, root, inode); 2506cbff00fSChristoph Hellwig BUG_ON(ret); 2516cbff00fSChristoph Hellwig 2526cbff00fSChristoph Hellwig btrfs_update_iflags(inode); 2536cbff00fSChristoph Hellwig inode->i_ctime = CURRENT_TIME; 2546cbff00fSChristoph Hellwig btrfs_end_transaction(trans, root); 2556cbff00fSChristoph Hellwig 2566cbff00fSChristoph Hellwig mnt_drop_write(file->f_path.mnt); 2572d4e6f6aSliubo 2582d4e6f6aSliubo ret = 0; 2596cbff00fSChristoph Hellwig out_unlock: 2606cbff00fSChristoph Hellwig mutex_unlock(&inode->i_mutex); 2612d4e6f6aSliubo return ret; 2626cbff00fSChristoph Hellwig } 2636cbff00fSChristoph Hellwig 2646cbff00fSChristoph Hellwig static int btrfs_ioctl_getversion(struct file *file, int __user *arg) 2656cbff00fSChristoph Hellwig { 2666cbff00fSChristoph Hellwig struct inode *inode = file->f_path.dentry->d_inode; 2676cbff00fSChristoph Hellwig 2686cbff00fSChristoph Hellwig return put_user(inode->i_generation, arg); 2696cbff00fSChristoph Hellwig } 270f46b5a66SChristoph Hellwig 271f7039b1dSLi Dongyang static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg) 272f7039b1dSLi Dongyang { 273f7039b1dSLi Dongyang struct btrfs_root *root = fdentry(file)->d_sb->s_fs_info; 274f7039b1dSLi Dongyang struct btrfs_fs_info *fs_info = root->fs_info; 275f7039b1dSLi Dongyang struct btrfs_device *device; 276f7039b1dSLi Dongyang struct request_queue *q; 277f7039b1dSLi Dongyang struct fstrim_range range; 278f7039b1dSLi Dongyang u64 minlen = ULLONG_MAX; 279f7039b1dSLi Dongyang u64 num_devices = 0; 280f7039b1dSLi Dongyang int ret; 281f7039b1dSLi Dongyang 282f7039b1dSLi Dongyang if (!capable(CAP_SYS_ADMIN)) 283f7039b1dSLi Dongyang return -EPERM; 284f7039b1dSLi Dongyang 2851f78160cSXiao Guangrong rcu_read_lock(); 2861f78160cSXiao Guangrong list_for_each_entry_rcu(device, &fs_info->fs_devices->devices, 2871f78160cSXiao Guangrong dev_list) { 288f7039b1dSLi Dongyang if (!device->bdev) 289f7039b1dSLi Dongyang continue; 290f7039b1dSLi Dongyang q = bdev_get_queue(device->bdev); 291f7039b1dSLi Dongyang if (blk_queue_discard(q)) { 292f7039b1dSLi Dongyang num_devices++; 293f7039b1dSLi Dongyang minlen = min((u64)q->limits.discard_granularity, 294f7039b1dSLi Dongyang minlen); 295f7039b1dSLi Dongyang } 296f7039b1dSLi Dongyang } 2971f78160cSXiao Guangrong rcu_read_unlock(); 298f7039b1dSLi Dongyang if (!num_devices) 299f7039b1dSLi Dongyang return -EOPNOTSUPP; 300f7039b1dSLi Dongyang 301f7039b1dSLi Dongyang if (copy_from_user(&range, arg, sizeof(range))) 302f7039b1dSLi Dongyang return -EFAULT; 303f7039b1dSLi Dongyang 304f7039b1dSLi Dongyang range.minlen = max(range.minlen, minlen); 305f7039b1dSLi Dongyang ret = btrfs_trim_fs(root, &range); 306f7039b1dSLi Dongyang if (ret < 0) 307f7039b1dSLi Dongyang return ret; 308f7039b1dSLi Dongyang 309f7039b1dSLi Dongyang if (copy_to_user(arg, &range, sizeof(range))) 310f7039b1dSLi Dongyang return -EFAULT; 311f7039b1dSLi Dongyang 312f7039b1dSLi Dongyang return 0; 313f7039b1dSLi Dongyang } 314f7039b1dSLi Dongyang 315cb8e7090SChristoph Hellwig static noinline int create_subvol(struct btrfs_root *root, 316cb8e7090SChristoph Hellwig struct dentry *dentry, 31772fd032eSSage Weil char *name, int namelen, 31872fd032eSSage Weil u64 *async_transid) 319f46b5a66SChristoph Hellwig { 320f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 321f46b5a66SChristoph Hellwig struct btrfs_key key; 322f46b5a66SChristoph Hellwig struct btrfs_root_item root_item; 323f46b5a66SChristoph Hellwig struct btrfs_inode_item *inode_item; 324f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 32576dda93cSYan, Zheng struct btrfs_root *new_root; 3262fbe8c8aSAl Viro struct dentry *parent = dentry->d_parent; 3276a912213SJosef Bacik struct inode *dir; 328f46b5a66SChristoph Hellwig int ret; 329f46b5a66SChristoph Hellwig int err; 330f46b5a66SChristoph Hellwig u64 objectid; 331f46b5a66SChristoph Hellwig u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; 3323de4586cSChris Mason u64 index = 0; 333f46b5a66SChristoph Hellwig 334581bb050SLi Zefan ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid); 3352fbe8c8aSAl Viro if (ret) 336a22285a6SYan, Zheng return ret; 3376a912213SJosef Bacik 3386a912213SJosef Bacik dir = parent->d_inode; 3396a912213SJosef Bacik 3409ed74f2dSJosef Bacik /* 3419ed74f2dSJosef Bacik * 1 - inode item 3429ed74f2dSJosef Bacik * 2 - refs 3439ed74f2dSJosef Bacik * 1 - root item 3449ed74f2dSJosef Bacik * 2 - dir items 3459ed74f2dSJosef Bacik */ 346a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 6); 3472fbe8c8aSAl Viro if (IS_ERR(trans)) 348a22285a6SYan, Zheng return PTR_ERR(trans); 349f46b5a66SChristoph Hellwig 3505d4f98a2SYan Zheng leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 3515d4f98a2SYan Zheng 0, objectid, NULL, 0, 0, 0); 3528e8a1e31SJosef Bacik if (IS_ERR(leaf)) { 3538e8a1e31SJosef Bacik ret = PTR_ERR(leaf); 3548e8a1e31SJosef Bacik goto fail; 3558e8a1e31SJosef Bacik } 356f46b5a66SChristoph Hellwig 3575d4f98a2SYan Zheng memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); 358f46b5a66SChristoph Hellwig btrfs_set_header_bytenr(leaf, leaf->start); 359f46b5a66SChristoph Hellwig btrfs_set_header_generation(leaf, trans->transid); 3605d4f98a2SYan Zheng btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 361f46b5a66SChristoph Hellwig btrfs_set_header_owner(leaf, objectid); 362f46b5a66SChristoph Hellwig 363f46b5a66SChristoph Hellwig write_extent_buffer(leaf, root->fs_info->fsid, 364f46b5a66SChristoph Hellwig (unsigned long)btrfs_header_fsid(leaf), 365f46b5a66SChristoph Hellwig BTRFS_FSID_SIZE); 3665d4f98a2SYan Zheng write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid, 3675d4f98a2SYan Zheng (unsigned long)btrfs_header_chunk_tree_uuid(leaf), 3685d4f98a2SYan Zheng BTRFS_UUID_SIZE); 369f46b5a66SChristoph Hellwig btrfs_mark_buffer_dirty(leaf); 370f46b5a66SChristoph Hellwig 371f46b5a66SChristoph Hellwig inode_item = &root_item.inode; 372f46b5a66SChristoph Hellwig memset(inode_item, 0, sizeof(*inode_item)); 373f46b5a66SChristoph Hellwig inode_item->generation = cpu_to_le64(1); 374f46b5a66SChristoph Hellwig inode_item->size = cpu_to_le64(3); 375f46b5a66SChristoph Hellwig inode_item->nlink = cpu_to_le32(1); 376a76a3cd4SYan Zheng inode_item->nbytes = cpu_to_le64(root->leafsize); 377f46b5a66SChristoph Hellwig inode_item->mode = cpu_to_le32(S_IFDIR | 0755); 378f46b5a66SChristoph Hellwig 37908fe4db1SLi Zefan root_item.flags = 0; 38008fe4db1SLi Zefan root_item.byte_limit = 0; 38108fe4db1SLi Zefan inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT); 38208fe4db1SLi Zefan 383f46b5a66SChristoph Hellwig btrfs_set_root_bytenr(&root_item, leaf->start); 38484234f3aSYan Zheng btrfs_set_root_generation(&root_item, trans->transid); 385f46b5a66SChristoph Hellwig btrfs_set_root_level(&root_item, 0); 386f46b5a66SChristoph Hellwig btrfs_set_root_refs(&root_item, 1); 38786b9f2ecSYan, Zheng btrfs_set_root_used(&root_item, leaf->len); 38880ff3856SYan Zheng btrfs_set_root_last_snapshot(&root_item, 0); 389f46b5a66SChristoph Hellwig 390f46b5a66SChristoph Hellwig memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress)); 391f46b5a66SChristoph Hellwig root_item.drop_level = 0; 392f46b5a66SChristoph Hellwig 393925baeddSChris Mason btrfs_tree_unlock(leaf); 394f46b5a66SChristoph Hellwig free_extent_buffer(leaf); 395f46b5a66SChristoph Hellwig leaf = NULL; 396f46b5a66SChristoph Hellwig 397f46b5a66SChristoph Hellwig btrfs_set_root_dirid(&root_item, new_dirid); 398f46b5a66SChristoph Hellwig 399f46b5a66SChristoph Hellwig key.objectid = objectid; 4005d4f98a2SYan Zheng key.offset = 0; 401f46b5a66SChristoph Hellwig btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); 402f46b5a66SChristoph Hellwig ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, 403f46b5a66SChristoph Hellwig &root_item); 404f46b5a66SChristoph Hellwig if (ret) 405f46b5a66SChristoph Hellwig goto fail; 406f46b5a66SChristoph Hellwig 40776dda93cSYan, Zheng key.offset = (u64)-1; 40876dda93cSYan, Zheng new_root = btrfs_read_fs_root_no_name(root->fs_info, &key); 40976dda93cSYan, Zheng BUG_ON(IS_ERR(new_root)); 41076dda93cSYan, Zheng 41176dda93cSYan, Zheng btrfs_record_root_in_trans(trans, new_root); 41276dda93cSYan, Zheng 413d82a6f1dSJosef Bacik ret = btrfs_create_subvol_root(trans, new_root, new_dirid); 414f46b5a66SChristoph Hellwig /* 415f46b5a66SChristoph Hellwig * insert the directory item 416f46b5a66SChristoph Hellwig */ 4173de4586cSChris Mason ret = btrfs_set_inode_index(dir, &index); 4183de4586cSChris Mason BUG_ON(ret); 4193de4586cSChris Mason 4203de4586cSChris Mason ret = btrfs_insert_dir_item(trans, root, 42116cdcec7SMiao Xie name, namelen, dir, &key, 4223de4586cSChris Mason BTRFS_FT_DIR, index); 423f46b5a66SChristoph Hellwig if (ret) 424f46b5a66SChristoph Hellwig goto fail; 4250660b5afSChris Mason 42652c26179SYan Zheng btrfs_i_size_write(dir, dir->i_size + namelen * 2); 42752c26179SYan Zheng ret = btrfs_update_inode(trans, root, dir); 42852c26179SYan Zheng BUG_ON(ret); 42952c26179SYan Zheng 4300660b5afSChris Mason ret = btrfs_add_root_ref(trans, root->fs_info->tree_root, 4314df27c4dSYan, Zheng objectid, root->root_key.objectid, 43233345d01SLi Zefan btrfs_ino(dir), index, name, namelen); 43376dda93cSYan, Zheng 4340660b5afSChris Mason BUG_ON(ret); 4350660b5afSChris Mason 43676dda93cSYan, Zheng d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry)); 437f46b5a66SChristoph Hellwig fail: 43872fd032eSSage Weil if (async_transid) { 43972fd032eSSage Weil *async_transid = trans->transid; 44072fd032eSSage Weil err = btrfs_commit_transaction_async(trans, root, 1); 44172fd032eSSage Weil } else { 44276dda93cSYan, Zheng err = btrfs_commit_transaction(trans, root); 44372fd032eSSage Weil } 444f46b5a66SChristoph Hellwig if (err && !ret) 445f46b5a66SChristoph Hellwig ret = err; 446f46b5a66SChristoph Hellwig return ret; 447f46b5a66SChristoph Hellwig } 448f46b5a66SChristoph Hellwig 44972fd032eSSage Weil static int create_snapshot(struct btrfs_root *root, struct dentry *dentry, 450b83cc969SLi Zefan char *name, int namelen, u64 *async_transid, 451b83cc969SLi Zefan bool readonly) 452f46b5a66SChristoph Hellwig { 4532e4bfab9SYan, Zheng struct inode *inode; 454f46b5a66SChristoph Hellwig struct btrfs_pending_snapshot *pending_snapshot; 455f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 4562e4bfab9SYan, Zheng int ret; 457f46b5a66SChristoph Hellwig 458f46b5a66SChristoph Hellwig if (!root->ref_cows) 459f46b5a66SChristoph Hellwig return -EINVAL; 460f46b5a66SChristoph Hellwig 461a22285a6SYan, Zheng pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS); 462a22285a6SYan, Zheng if (!pending_snapshot) 463a22285a6SYan, Zheng return -ENOMEM; 464a22285a6SYan, Zheng 465a22285a6SYan, Zheng btrfs_init_block_rsv(&pending_snapshot->block_rsv); 466a22285a6SYan, Zheng pending_snapshot->dentry = dentry; 467a22285a6SYan, Zheng pending_snapshot->root = root; 468b83cc969SLi Zefan pending_snapshot->readonly = readonly; 469a22285a6SYan, Zheng 470a22285a6SYan, Zheng trans = btrfs_start_transaction(root->fs_info->extent_root, 5); 471a22285a6SYan, Zheng if (IS_ERR(trans)) { 472a22285a6SYan, Zheng ret = PTR_ERR(trans); 473a22285a6SYan, Zheng goto fail; 474a22285a6SYan, Zheng } 475a22285a6SYan, Zheng 476a22285a6SYan, Zheng ret = btrfs_snap_reserve_metadata(trans, pending_snapshot); 477a22285a6SYan, Zheng BUG_ON(ret); 478a22285a6SYan, Zheng 4798351583eSJosef Bacik spin_lock(&root->fs_info->trans_lock); 480a22285a6SYan, Zheng list_add(&pending_snapshot->list, 481a22285a6SYan, Zheng &trans->transaction->pending_snapshots); 4828351583eSJosef Bacik spin_unlock(&root->fs_info->trans_lock); 48372fd032eSSage Weil if (async_transid) { 48472fd032eSSage Weil *async_transid = trans->transid; 48572fd032eSSage Weil ret = btrfs_commit_transaction_async(trans, 48672fd032eSSage Weil root->fs_info->extent_root, 1); 48772fd032eSSage Weil } else { 48872fd032eSSage Weil ret = btrfs_commit_transaction(trans, 48972fd032eSSage Weil root->fs_info->extent_root); 49072fd032eSSage Weil } 491a22285a6SYan, Zheng BUG_ON(ret); 492a22285a6SYan, Zheng 493a22285a6SYan, Zheng ret = pending_snapshot->error; 494f46b5a66SChristoph Hellwig if (ret) 4952e4bfab9SYan, Zheng goto fail; 496f46b5a66SChristoph Hellwig 49766b4ffd1SJosef Bacik ret = btrfs_orphan_cleanup(pending_snapshot->snap); 49866b4ffd1SJosef Bacik if (ret) 49966b4ffd1SJosef Bacik goto fail; 500f46b5a66SChristoph Hellwig 5012fbe8c8aSAl Viro inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry); 5022e4bfab9SYan, Zheng if (IS_ERR(inode)) { 5032e4bfab9SYan, Zheng ret = PTR_ERR(inode); 5042e4bfab9SYan, Zheng goto fail; 5052e4bfab9SYan, Zheng } 5062e4bfab9SYan, Zheng BUG_ON(!inode); 5072e4bfab9SYan, Zheng d_instantiate(dentry, inode); 5082e4bfab9SYan, Zheng ret = 0; 5092e4bfab9SYan, Zheng fail: 510a22285a6SYan, Zheng kfree(pending_snapshot); 511f46b5a66SChristoph Hellwig return ret; 512f46b5a66SChristoph Hellwig } 513f46b5a66SChristoph Hellwig 5144260f7c7SSage Weil /* copy of check_sticky in fs/namei.c() 5154260f7c7SSage Weil * It's inline, so penalty for filesystems that don't use sticky bit is 5164260f7c7SSage Weil * minimal. 5174260f7c7SSage Weil */ 5184260f7c7SSage Weil static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode) 5194260f7c7SSage Weil { 5204260f7c7SSage Weil uid_t fsuid = current_fsuid(); 5214260f7c7SSage Weil 5224260f7c7SSage Weil if (!(dir->i_mode & S_ISVTX)) 5234260f7c7SSage Weil return 0; 5244260f7c7SSage Weil if (inode->i_uid == fsuid) 5254260f7c7SSage Weil return 0; 5264260f7c7SSage Weil if (dir->i_uid == fsuid) 5274260f7c7SSage Weil return 0; 5284260f7c7SSage Weil return !capable(CAP_FOWNER); 5294260f7c7SSage Weil } 5304260f7c7SSage Weil 5314260f7c7SSage Weil /* copy of may_delete in fs/namei.c() 5324260f7c7SSage Weil * Check whether we can remove a link victim from directory dir, check 5334260f7c7SSage Weil * whether the type of victim is right. 5344260f7c7SSage Weil * 1. We can't do it if dir is read-only (done in permission()) 5354260f7c7SSage Weil * 2. We should have write and exec permissions on dir 5364260f7c7SSage Weil * 3. We can't remove anything from append-only dir 5374260f7c7SSage Weil * 4. We can't do anything with immutable dir (done in permission()) 5384260f7c7SSage Weil * 5. If the sticky bit on dir is set we should either 5394260f7c7SSage Weil * a. be owner of dir, or 5404260f7c7SSage Weil * b. be owner of victim, or 5414260f7c7SSage Weil * c. have CAP_FOWNER capability 5424260f7c7SSage Weil * 6. If the victim is append-only or immutable we can't do antyhing with 5434260f7c7SSage Weil * links pointing to it. 5444260f7c7SSage Weil * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. 5454260f7c7SSage Weil * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. 5464260f7c7SSage Weil * 9. We can't remove a root or mountpoint. 5474260f7c7SSage Weil * 10. We don't allow removal of NFS sillyrenamed files; it's handled by 5484260f7c7SSage Weil * nfs_async_unlink(). 5494260f7c7SSage Weil */ 5504260f7c7SSage Weil 5514260f7c7SSage Weil static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir) 5524260f7c7SSage Weil { 5534260f7c7SSage Weil int error; 5544260f7c7SSage Weil 5554260f7c7SSage Weil if (!victim->d_inode) 5564260f7c7SSage Weil return -ENOENT; 5574260f7c7SSage Weil 5584260f7c7SSage Weil BUG_ON(victim->d_parent->d_inode != dir); 5594260f7c7SSage Weil audit_inode_child(victim, dir); 5604260f7c7SSage Weil 5614260f7c7SSage Weil error = inode_permission(dir, MAY_WRITE | MAY_EXEC); 5624260f7c7SSage Weil if (error) 5634260f7c7SSage Weil return error; 5644260f7c7SSage Weil if (IS_APPEND(dir)) 5654260f7c7SSage Weil return -EPERM; 5664260f7c7SSage Weil if (btrfs_check_sticky(dir, victim->d_inode)|| 5674260f7c7SSage Weil IS_APPEND(victim->d_inode)|| 5684260f7c7SSage Weil IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode)) 5694260f7c7SSage Weil return -EPERM; 5704260f7c7SSage Weil if (isdir) { 5714260f7c7SSage Weil if (!S_ISDIR(victim->d_inode->i_mode)) 5724260f7c7SSage Weil return -ENOTDIR; 5734260f7c7SSage Weil if (IS_ROOT(victim)) 5744260f7c7SSage Weil return -EBUSY; 5754260f7c7SSage Weil } else if (S_ISDIR(victim->d_inode->i_mode)) 5764260f7c7SSage Weil return -EISDIR; 5774260f7c7SSage Weil if (IS_DEADDIR(dir)) 5784260f7c7SSage Weil return -ENOENT; 5794260f7c7SSage Weil if (victim->d_flags & DCACHE_NFSFS_RENAMED) 5804260f7c7SSage Weil return -EBUSY; 5814260f7c7SSage Weil return 0; 5824260f7c7SSage Weil } 5834260f7c7SSage Weil 584cb8e7090SChristoph Hellwig /* copy of may_create in fs/namei.c() */ 585cb8e7090SChristoph Hellwig static inline int btrfs_may_create(struct inode *dir, struct dentry *child) 586cb8e7090SChristoph Hellwig { 587cb8e7090SChristoph Hellwig if (child->d_inode) 588cb8e7090SChristoph Hellwig return -EEXIST; 589cb8e7090SChristoph Hellwig if (IS_DEADDIR(dir)) 590cb8e7090SChristoph Hellwig return -ENOENT; 591cb8e7090SChristoph Hellwig return inode_permission(dir, MAY_WRITE | MAY_EXEC); 592cb8e7090SChristoph Hellwig } 593cb8e7090SChristoph Hellwig 594cb8e7090SChristoph Hellwig /* 595cb8e7090SChristoph Hellwig * Create a new subvolume below @parent. This is largely modeled after 596cb8e7090SChristoph Hellwig * sys_mkdirat and vfs_mkdir, but we only do a single component lookup 597cb8e7090SChristoph Hellwig * inside this filesystem so it's quite a bit simpler. 598cb8e7090SChristoph Hellwig */ 59976dda93cSYan, Zheng static noinline int btrfs_mksubvol(struct path *parent, 60076dda93cSYan, Zheng char *name, int namelen, 60172fd032eSSage Weil struct btrfs_root *snap_src, 602b83cc969SLi Zefan u64 *async_transid, bool readonly) 603cb8e7090SChristoph Hellwig { 60476dda93cSYan, Zheng struct inode *dir = parent->dentry->d_inode; 605cb8e7090SChristoph Hellwig struct dentry *dentry; 606cb8e7090SChristoph Hellwig int error; 607cb8e7090SChristoph Hellwig 60876dda93cSYan, Zheng mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT); 609cb8e7090SChristoph Hellwig 610cb8e7090SChristoph Hellwig dentry = lookup_one_len(name, parent->dentry, namelen); 611cb8e7090SChristoph Hellwig error = PTR_ERR(dentry); 612cb8e7090SChristoph Hellwig if (IS_ERR(dentry)) 613cb8e7090SChristoph Hellwig goto out_unlock; 614cb8e7090SChristoph Hellwig 615cb8e7090SChristoph Hellwig error = -EEXIST; 616cb8e7090SChristoph Hellwig if (dentry->d_inode) 617cb8e7090SChristoph Hellwig goto out_dput; 618cb8e7090SChristoph Hellwig 619cb8e7090SChristoph Hellwig error = mnt_want_write(parent->mnt); 620cb8e7090SChristoph Hellwig if (error) 621cb8e7090SChristoph Hellwig goto out_dput; 622cb8e7090SChristoph Hellwig 62376dda93cSYan, Zheng error = btrfs_may_create(dir, dentry); 624cb8e7090SChristoph Hellwig if (error) 625cb8e7090SChristoph Hellwig goto out_drop_write; 626cb8e7090SChristoph Hellwig 62776dda93cSYan, Zheng down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem); 62876dda93cSYan, Zheng 62976dda93cSYan, Zheng if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) 63076dda93cSYan, Zheng goto out_up_read; 63176dda93cSYan, Zheng 6323de4586cSChris Mason if (snap_src) { 63372fd032eSSage Weil error = create_snapshot(snap_src, dentry, 634b83cc969SLi Zefan name, namelen, async_transid, readonly); 6353de4586cSChris Mason } else { 63676dda93cSYan, Zheng error = create_subvol(BTRFS_I(dir)->root, dentry, 63772fd032eSSage Weil name, namelen, async_transid); 6383de4586cSChris Mason } 63976dda93cSYan, Zheng if (!error) 64076dda93cSYan, Zheng fsnotify_mkdir(dir, dentry); 64176dda93cSYan, Zheng out_up_read: 64276dda93cSYan, Zheng up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem); 643cb8e7090SChristoph Hellwig out_drop_write: 644cb8e7090SChristoph Hellwig mnt_drop_write(parent->mnt); 645cb8e7090SChristoph Hellwig out_dput: 646cb8e7090SChristoph Hellwig dput(dentry); 647cb8e7090SChristoph Hellwig out_unlock: 64876dda93cSYan, Zheng mutex_unlock(&dir->i_mutex); 649cb8e7090SChristoph Hellwig return error; 650cb8e7090SChristoph Hellwig } 651cb8e7090SChristoph Hellwig 6524cb5300bSChris Mason /* 6534cb5300bSChris Mason * When we're defragging a range, we don't want to kick it off again 6544cb5300bSChris Mason * if it is really just waiting for delalloc to send it down. 6554cb5300bSChris Mason * If we find a nice big extent or delalloc range for the bytes in the 6564cb5300bSChris Mason * file you want to defrag, we return 0 to let you know to skip this 6574cb5300bSChris Mason * part of the file 6584cb5300bSChris Mason */ 6594cb5300bSChris Mason static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh) 6604cb5300bSChris Mason { 6614cb5300bSChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 6624cb5300bSChris Mason struct extent_map *em = NULL; 6634cb5300bSChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 6644cb5300bSChris Mason u64 end; 6654cb5300bSChris Mason 6664cb5300bSChris Mason read_lock(&em_tree->lock); 6674cb5300bSChris Mason em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE); 6684cb5300bSChris Mason read_unlock(&em_tree->lock); 6694cb5300bSChris Mason 6704cb5300bSChris Mason if (em) { 6714cb5300bSChris Mason end = extent_map_end(em); 6724cb5300bSChris Mason free_extent_map(em); 6734cb5300bSChris Mason if (end - offset > thresh) 6744cb5300bSChris Mason return 0; 6754cb5300bSChris Mason } 6764cb5300bSChris Mason /* if we already have a nice delalloc here, just stop */ 6774cb5300bSChris Mason thresh /= 2; 6784cb5300bSChris Mason end = count_range_bits(io_tree, &offset, offset + thresh, 6794cb5300bSChris Mason thresh, EXTENT_DELALLOC, 1); 6804cb5300bSChris Mason if (end >= thresh) 6814cb5300bSChris Mason return 0; 6824cb5300bSChris Mason return 1; 6834cb5300bSChris Mason } 6844cb5300bSChris Mason 6854cb5300bSChris Mason /* 6864cb5300bSChris Mason * helper function to walk through a file and find extents 6874cb5300bSChris Mason * newer than a specific transid, and smaller than thresh. 6884cb5300bSChris Mason * 6894cb5300bSChris Mason * This is used by the defragging code to find new and small 6904cb5300bSChris Mason * extents 6914cb5300bSChris Mason */ 6924cb5300bSChris Mason static int find_new_extents(struct btrfs_root *root, 6934cb5300bSChris Mason struct inode *inode, u64 newer_than, 6944cb5300bSChris Mason u64 *off, int thresh) 6954cb5300bSChris Mason { 6964cb5300bSChris Mason struct btrfs_path *path; 6974cb5300bSChris Mason struct btrfs_key min_key; 6984cb5300bSChris Mason struct btrfs_key max_key; 6994cb5300bSChris Mason struct extent_buffer *leaf; 7004cb5300bSChris Mason struct btrfs_file_extent_item *extent; 7014cb5300bSChris Mason int type; 7024cb5300bSChris Mason int ret; 703a4689d2bSDavid Sterba u64 ino = btrfs_ino(inode); 7044cb5300bSChris Mason 7054cb5300bSChris Mason path = btrfs_alloc_path(); 7064cb5300bSChris Mason if (!path) 7074cb5300bSChris Mason return -ENOMEM; 7084cb5300bSChris Mason 709a4689d2bSDavid Sterba min_key.objectid = ino; 7104cb5300bSChris Mason min_key.type = BTRFS_EXTENT_DATA_KEY; 7114cb5300bSChris Mason min_key.offset = *off; 7124cb5300bSChris Mason 713a4689d2bSDavid Sterba max_key.objectid = ino; 7144cb5300bSChris Mason max_key.type = (u8)-1; 7154cb5300bSChris Mason max_key.offset = (u64)-1; 7164cb5300bSChris Mason 7174cb5300bSChris Mason path->keep_locks = 1; 7184cb5300bSChris Mason 7194cb5300bSChris Mason while(1) { 7204cb5300bSChris Mason ret = btrfs_search_forward(root, &min_key, &max_key, 7214cb5300bSChris Mason path, 0, newer_than); 7224cb5300bSChris Mason if (ret != 0) 7234cb5300bSChris Mason goto none; 724a4689d2bSDavid Sterba if (min_key.objectid != ino) 7254cb5300bSChris Mason goto none; 7264cb5300bSChris Mason if (min_key.type != BTRFS_EXTENT_DATA_KEY) 7274cb5300bSChris Mason goto none; 7284cb5300bSChris Mason 7294cb5300bSChris Mason leaf = path->nodes[0]; 7304cb5300bSChris Mason extent = btrfs_item_ptr(leaf, path->slots[0], 7314cb5300bSChris Mason struct btrfs_file_extent_item); 7324cb5300bSChris Mason 7334cb5300bSChris Mason type = btrfs_file_extent_type(leaf, extent); 7344cb5300bSChris Mason if (type == BTRFS_FILE_EXTENT_REG && 7354cb5300bSChris Mason btrfs_file_extent_num_bytes(leaf, extent) < thresh && 7364cb5300bSChris Mason check_defrag_in_cache(inode, min_key.offset, thresh)) { 7374cb5300bSChris Mason *off = min_key.offset; 7384cb5300bSChris Mason btrfs_free_path(path); 7394cb5300bSChris Mason return 0; 7404cb5300bSChris Mason } 7414cb5300bSChris Mason 7424cb5300bSChris Mason if (min_key.offset == (u64)-1) 7434cb5300bSChris Mason goto none; 7444cb5300bSChris Mason 7454cb5300bSChris Mason min_key.offset++; 7464cb5300bSChris Mason btrfs_release_path(path); 7474cb5300bSChris Mason } 7484cb5300bSChris Mason none: 7494cb5300bSChris Mason btrfs_free_path(path); 7504cb5300bSChris Mason return -ENOENT; 7514cb5300bSChris Mason } 7524cb5300bSChris Mason 753940100a4SChris Mason static int should_defrag_range(struct inode *inode, u64 start, u64 len, 7541e701a32SChris Mason int thresh, u64 *last_len, u64 *skip, 7551e701a32SChris Mason u64 *defrag_end) 756940100a4SChris Mason { 757940100a4SChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 758940100a4SChris Mason struct extent_map *em = NULL; 759940100a4SChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 760940100a4SChris Mason int ret = 1; 761940100a4SChris Mason 762940100a4SChris Mason /* 763940100a4SChris Mason * make sure that once we start defragging and extent, we keep on 764940100a4SChris Mason * defragging it 765940100a4SChris Mason */ 766940100a4SChris Mason if (start < *defrag_end) 767940100a4SChris Mason return 1; 768940100a4SChris Mason 769940100a4SChris Mason *skip = 0; 770940100a4SChris Mason 771940100a4SChris Mason /* 772940100a4SChris Mason * hopefully we have this extent in the tree already, try without 773940100a4SChris Mason * the full extent lock 774940100a4SChris Mason */ 775940100a4SChris Mason read_lock(&em_tree->lock); 776940100a4SChris Mason em = lookup_extent_mapping(em_tree, start, len); 777940100a4SChris Mason read_unlock(&em_tree->lock); 778940100a4SChris Mason 779940100a4SChris Mason if (!em) { 780940100a4SChris Mason /* get the big lock and read metadata off disk */ 781940100a4SChris Mason lock_extent(io_tree, start, start + len - 1, GFP_NOFS); 782940100a4SChris Mason em = btrfs_get_extent(inode, NULL, 0, start, len, 0); 783940100a4SChris Mason unlock_extent(io_tree, start, start + len - 1, GFP_NOFS); 784940100a4SChris Mason 7856cf8bfbfSDan Carpenter if (IS_ERR(em)) 786940100a4SChris Mason return 0; 787940100a4SChris Mason } 788940100a4SChris Mason 789940100a4SChris Mason /* this will cover holes, and inline extents */ 790940100a4SChris Mason if (em->block_start >= EXTENT_MAP_LAST_BYTE) 791940100a4SChris Mason ret = 0; 792940100a4SChris Mason 793940100a4SChris Mason /* 794940100a4SChris Mason * we hit a real extent, if it is big don't bother defragging it again 795940100a4SChris Mason */ 7961e701a32SChris Mason if ((*last_len == 0 || *last_len >= thresh) && em->len >= thresh) 797940100a4SChris Mason ret = 0; 798940100a4SChris Mason 799940100a4SChris Mason /* 800940100a4SChris Mason * last_len ends up being a counter of how many bytes we've defragged. 801940100a4SChris Mason * every time we choose not to defrag an extent, we reset *last_len 802940100a4SChris Mason * so that the next tiny extent will force a defrag. 803940100a4SChris Mason * 804940100a4SChris Mason * The end result of this is that tiny extents before a single big 805940100a4SChris Mason * extent will force at least part of that big extent to be defragged. 806940100a4SChris Mason */ 807940100a4SChris Mason if (ret) { 808940100a4SChris Mason *last_len += len; 809940100a4SChris Mason *defrag_end = extent_map_end(em); 810940100a4SChris Mason } else { 811940100a4SChris Mason *last_len = 0; 812940100a4SChris Mason *skip = extent_map_end(em); 813940100a4SChris Mason *defrag_end = 0; 814940100a4SChris Mason } 815940100a4SChris Mason 816940100a4SChris Mason free_extent_map(em); 817940100a4SChris Mason return ret; 818940100a4SChris Mason } 819940100a4SChris Mason 8204cb5300bSChris Mason /* 8214cb5300bSChris Mason * it doesn't do much good to defrag one or two pages 8224cb5300bSChris Mason * at a time. This pulls in a nice chunk of pages 8234cb5300bSChris Mason * to COW and defrag. 8244cb5300bSChris Mason * 8254cb5300bSChris Mason * It also makes sure the delalloc code has enough 8264cb5300bSChris Mason * dirty data to avoid making new small extents as part 8274cb5300bSChris Mason * of the defrag 8284cb5300bSChris Mason * 8294cb5300bSChris Mason * It's a good idea to start RA on this range 8304cb5300bSChris Mason * before calling this. 8314cb5300bSChris Mason */ 8324cb5300bSChris Mason static int cluster_pages_for_defrag(struct inode *inode, 8334cb5300bSChris Mason struct page **pages, 8344cb5300bSChris Mason unsigned long start_index, 8354cb5300bSChris Mason int num_pages) 836f46b5a66SChristoph Hellwig { 8374cb5300bSChris Mason unsigned long file_end; 8384cb5300bSChris Mason u64 isize = i_size_read(inode); 839f46b5a66SChristoph Hellwig u64 page_start; 840f46b5a66SChristoph Hellwig u64 page_end; 8414cb5300bSChris Mason int ret; 8424cb5300bSChris Mason int i; 8434cb5300bSChris Mason int i_done; 8444cb5300bSChris Mason struct btrfs_ordered_extent *ordered; 8454cb5300bSChris Mason struct extent_state *cached_state = NULL; 8464cb5300bSChris Mason 8474cb5300bSChris Mason if (isize == 0) 8484cb5300bSChris Mason return 0; 8494cb5300bSChris Mason file_end = (isize - 1) >> PAGE_CACHE_SHIFT; 8504cb5300bSChris Mason 8514cb5300bSChris Mason ret = btrfs_delalloc_reserve_space(inode, 8524cb5300bSChris Mason num_pages << PAGE_CACHE_SHIFT); 8534cb5300bSChris Mason if (ret) 8544cb5300bSChris Mason return ret; 8554cb5300bSChris Mason again: 8564cb5300bSChris Mason ret = 0; 8574cb5300bSChris Mason i_done = 0; 8584cb5300bSChris Mason 8594cb5300bSChris Mason /* step one, lock all the pages */ 8604cb5300bSChris Mason for (i = 0; i < num_pages; i++) { 8614cb5300bSChris Mason struct page *page; 8624cb5300bSChris Mason page = grab_cache_page(inode->i_mapping, 8634cb5300bSChris Mason start_index + i); 8644cb5300bSChris Mason if (!page) 8654cb5300bSChris Mason break; 8664cb5300bSChris Mason 8674cb5300bSChris Mason if (!PageUptodate(page)) { 8684cb5300bSChris Mason btrfs_readpage(NULL, page); 8694cb5300bSChris Mason lock_page(page); 8704cb5300bSChris Mason if (!PageUptodate(page)) { 8714cb5300bSChris Mason unlock_page(page); 8724cb5300bSChris Mason page_cache_release(page); 8734cb5300bSChris Mason ret = -EIO; 8744cb5300bSChris Mason break; 8754cb5300bSChris Mason } 8764cb5300bSChris Mason } 8774cb5300bSChris Mason isize = i_size_read(inode); 8784cb5300bSChris Mason file_end = (isize - 1) >> PAGE_CACHE_SHIFT; 8794cb5300bSChris Mason if (!isize || page->index > file_end || 8804cb5300bSChris Mason page->mapping != inode->i_mapping) { 8814cb5300bSChris Mason /* whoops, we blew past eof, skip this page */ 8824cb5300bSChris Mason unlock_page(page); 8834cb5300bSChris Mason page_cache_release(page); 8844cb5300bSChris Mason break; 8854cb5300bSChris Mason } 8864cb5300bSChris Mason pages[i] = page; 8874cb5300bSChris Mason i_done++; 8884cb5300bSChris Mason } 8894cb5300bSChris Mason if (!i_done || ret) 8904cb5300bSChris Mason goto out; 8914cb5300bSChris Mason 8924cb5300bSChris Mason if (!(inode->i_sb->s_flags & MS_ACTIVE)) 8934cb5300bSChris Mason goto out; 8944cb5300bSChris Mason 8954cb5300bSChris Mason /* 8964cb5300bSChris Mason * so now we have a nice long stream of locked 8974cb5300bSChris Mason * and up to date pages, lets wait on them 8984cb5300bSChris Mason */ 8994cb5300bSChris Mason for (i = 0; i < i_done; i++) 9004cb5300bSChris Mason wait_on_page_writeback(pages[i]); 9014cb5300bSChris Mason 9024cb5300bSChris Mason page_start = page_offset(pages[0]); 9034cb5300bSChris Mason page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE; 9044cb5300bSChris Mason 9054cb5300bSChris Mason lock_extent_bits(&BTRFS_I(inode)->io_tree, 9064cb5300bSChris Mason page_start, page_end - 1, 0, &cached_state, 9074cb5300bSChris Mason GFP_NOFS); 9084cb5300bSChris Mason ordered = btrfs_lookup_first_ordered_extent(inode, page_end - 1); 9094cb5300bSChris Mason if (ordered && 9104cb5300bSChris Mason ordered->file_offset + ordered->len > page_start && 9114cb5300bSChris Mason ordered->file_offset < page_end) { 9124cb5300bSChris Mason btrfs_put_ordered_extent(ordered); 9134cb5300bSChris Mason unlock_extent_cached(&BTRFS_I(inode)->io_tree, 9144cb5300bSChris Mason page_start, page_end - 1, 9154cb5300bSChris Mason &cached_state, GFP_NOFS); 9164cb5300bSChris Mason for (i = 0; i < i_done; i++) { 9174cb5300bSChris Mason unlock_page(pages[i]); 9184cb5300bSChris Mason page_cache_release(pages[i]); 9194cb5300bSChris Mason } 9204cb5300bSChris Mason btrfs_wait_ordered_range(inode, page_start, 9214cb5300bSChris Mason page_end - page_start); 9224cb5300bSChris Mason goto again; 9234cb5300bSChris Mason } 9244cb5300bSChris Mason if (ordered) 9254cb5300bSChris Mason btrfs_put_ordered_extent(ordered); 9264cb5300bSChris Mason 9274cb5300bSChris Mason clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, 9284cb5300bSChris Mason page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC | 9294cb5300bSChris Mason EXTENT_DO_ACCOUNTING, 0, 0, &cached_state, 9304cb5300bSChris Mason GFP_NOFS); 9314cb5300bSChris Mason 9324cb5300bSChris Mason if (i_done != num_pages) { 9334cb5300bSChris Mason atomic_inc(&BTRFS_I(inode)->outstanding_extents); 9344cb5300bSChris Mason btrfs_delalloc_release_space(inode, 9354cb5300bSChris Mason (num_pages - i_done) << PAGE_CACHE_SHIFT); 9364cb5300bSChris Mason } 9374cb5300bSChris Mason 9384cb5300bSChris Mason 9394cb5300bSChris Mason btrfs_set_extent_delalloc(inode, page_start, page_end - 1, 9404cb5300bSChris Mason &cached_state); 9414cb5300bSChris Mason 9424cb5300bSChris Mason unlock_extent_cached(&BTRFS_I(inode)->io_tree, 9434cb5300bSChris Mason page_start, page_end - 1, &cached_state, 9444cb5300bSChris Mason GFP_NOFS); 9454cb5300bSChris Mason 9464cb5300bSChris Mason for (i = 0; i < i_done; i++) { 9474cb5300bSChris Mason clear_page_dirty_for_io(pages[i]); 9484cb5300bSChris Mason ClearPageChecked(pages[i]); 9494cb5300bSChris Mason set_page_extent_mapped(pages[i]); 9504cb5300bSChris Mason set_page_dirty(pages[i]); 9514cb5300bSChris Mason unlock_page(pages[i]); 9524cb5300bSChris Mason page_cache_release(pages[i]); 9534cb5300bSChris Mason } 9544cb5300bSChris Mason return i_done; 9554cb5300bSChris Mason out: 9564cb5300bSChris Mason for (i = 0; i < i_done; i++) { 9574cb5300bSChris Mason unlock_page(pages[i]); 9584cb5300bSChris Mason page_cache_release(pages[i]); 9594cb5300bSChris Mason } 9604cb5300bSChris Mason btrfs_delalloc_release_space(inode, num_pages << PAGE_CACHE_SHIFT); 9614cb5300bSChris Mason return ret; 9624cb5300bSChris Mason 9634cb5300bSChris Mason } 9644cb5300bSChris Mason 9654cb5300bSChris Mason int btrfs_defrag_file(struct inode *inode, struct file *file, 9664cb5300bSChris Mason struct btrfs_ioctl_defrag_range_args *range, 9674cb5300bSChris Mason u64 newer_than, unsigned long max_to_defrag) 9684cb5300bSChris Mason { 9694cb5300bSChris Mason struct btrfs_root *root = BTRFS_I(inode)->root; 9704cb5300bSChris Mason struct btrfs_super_block *disk_super; 9714cb5300bSChris Mason struct file_ra_state *ra = NULL; 9724cb5300bSChris Mason unsigned long last_index; 9734cb5300bSChris Mason u64 features; 974940100a4SChris Mason u64 last_len = 0; 975940100a4SChris Mason u64 skip = 0; 976940100a4SChris Mason u64 defrag_end = 0; 9774cb5300bSChris Mason u64 newer_off = range->start; 9784cb5300bSChris Mason int newer_left = 0; 979f46b5a66SChristoph Hellwig unsigned long i; 980f46b5a66SChristoph Hellwig int ret; 9814cb5300bSChris Mason int defrag_count = 0; 9821a419d85SLi Zefan int compress_type = BTRFS_COMPRESS_ZLIB; 9834cb5300bSChris Mason int extent_thresh = range->extent_thresh; 9844cb5300bSChris Mason int newer_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT; 9854cb5300bSChris Mason u64 new_align = ~((u64)128 * 1024 - 1); 9864cb5300bSChris Mason struct page **pages = NULL; 9874cb5300bSChris Mason 9884cb5300bSChris Mason if (extent_thresh == 0) 9894cb5300bSChris Mason extent_thresh = 256 * 1024; 9901a419d85SLi Zefan 9911a419d85SLi Zefan if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) { 9921a419d85SLi Zefan if (range->compress_type > BTRFS_COMPRESS_TYPES) 9931a419d85SLi Zefan return -EINVAL; 9941a419d85SLi Zefan if (range->compress_type) 9951a419d85SLi Zefan compress_type = range->compress_type; 9961a419d85SLi Zefan } 997f46b5a66SChristoph Hellwig 998940100a4SChris Mason if (inode->i_size == 0) 999940100a4SChris Mason return 0; 1000f46b5a66SChristoph Hellwig 10014cb5300bSChris Mason /* 10024cb5300bSChris Mason * if we were not given a file, allocate a readahead 10034cb5300bSChris Mason * context 10044cb5300bSChris Mason */ 10054cb5300bSChris Mason if (!file) { 10064cb5300bSChris Mason ra = kzalloc(sizeof(*ra), GFP_NOFS); 10074cb5300bSChris Mason if (!ra) 10084cb5300bSChris Mason return -ENOMEM; 10094cb5300bSChris Mason file_ra_state_init(ra, inode->i_mapping); 10104cb5300bSChris Mason } else { 10114cb5300bSChris Mason ra = &file->f_ra; 10124cb5300bSChris Mason } 10134cb5300bSChris Mason 10144cb5300bSChris Mason pages = kmalloc(sizeof(struct page *) * newer_cluster, 10154cb5300bSChris Mason GFP_NOFS); 10164cb5300bSChris Mason if (!pages) { 10174cb5300bSChris Mason ret = -ENOMEM; 10184cb5300bSChris Mason goto out_ra; 10194cb5300bSChris Mason } 10204cb5300bSChris Mason 10214cb5300bSChris Mason /* find the last page to defrag */ 10221e701a32SChris Mason if (range->start + range->len > range->start) { 10231e701a32SChris Mason last_index = min_t(u64, inode->i_size - 1, 10241e701a32SChris Mason range->start + range->len - 1) >> PAGE_CACHE_SHIFT; 10251e701a32SChris Mason } else { 1026940100a4SChris Mason last_index = (inode->i_size - 1) >> PAGE_CACHE_SHIFT; 10271e701a32SChris Mason } 10281e701a32SChris Mason 10294cb5300bSChris Mason if (newer_than) { 10304cb5300bSChris Mason ret = find_new_extents(root, inode, newer_than, 10314cb5300bSChris Mason &newer_off, 64 * 1024); 10324cb5300bSChris Mason if (!ret) { 10334cb5300bSChris Mason range->start = newer_off; 10344cb5300bSChris Mason /* 10354cb5300bSChris Mason * we always align our defrag to help keep 10364cb5300bSChris Mason * the extents in the file evenly spaced 10374cb5300bSChris Mason */ 10384cb5300bSChris Mason i = (newer_off & new_align) >> PAGE_CACHE_SHIFT; 10394cb5300bSChris Mason newer_left = newer_cluster; 10404cb5300bSChris Mason } else 10414cb5300bSChris Mason goto out_ra; 10424cb5300bSChris Mason } else { 10431e701a32SChris Mason i = range->start >> PAGE_CACHE_SHIFT; 10444cb5300bSChris Mason } 10454cb5300bSChris Mason if (!max_to_defrag) 10464cb5300bSChris Mason max_to_defrag = last_index - 1; 10474cb5300bSChris Mason 10484cb5300bSChris Mason while (i <= last_index && defrag_count < max_to_defrag) { 10494cb5300bSChris Mason /* 10504cb5300bSChris Mason * make sure we stop running if someone unmounts 10514cb5300bSChris Mason * the FS 10524cb5300bSChris Mason */ 10534cb5300bSChris Mason if (!(inode->i_sb->s_flags & MS_ACTIVE)) 10544cb5300bSChris Mason break; 10554cb5300bSChris Mason 10564cb5300bSChris Mason if (!newer_than && 10574cb5300bSChris Mason !should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT, 10581e701a32SChris Mason PAGE_CACHE_SIZE, 10594cb5300bSChris Mason extent_thresh, 10601e701a32SChris Mason &last_len, &skip, 1061940100a4SChris Mason &defrag_end)) { 1062940100a4SChris Mason unsigned long next; 1063940100a4SChris Mason /* 1064940100a4SChris Mason * the should_defrag function tells us how much to skip 1065940100a4SChris Mason * bump our counter by the suggested amount 1066940100a4SChris Mason */ 1067940100a4SChris Mason next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT; 1068940100a4SChris Mason i = max(i + 1, next); 1069940100a4SChris Mason continue; 1070940100a4SChris Mason } 10711e701a32SChris Mason if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) 10721a419d85SLi Zefan BTRFS_I(inode)->force_compress = compress_type; 1073940100a4SChris Mason 10744cb5300bSChris Mason btrfs_force_ra(inode->i_mapping, ra, file, i, newer_cluster); 10754cb5300bSChris Mason 10764cb5300bSChris Mason ret = cluster_pages_for_defrag(inode, pages, i, newer_cluster); 10774cb5300bSChris Mason if (ret < 0) 10784cb5300bSChris Mason goto out_ra; 10794cb5300bSChris Mason 10804cb5300bSChris Mason defrag_count += ret; 10814cb5300bSChris Mason balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret); 10824cb5300bSChris Mason i += ret; 10834cb5300bSChris Mason 10844cb5300bSChris Mason if (newer_than) { 10854cb5300bSChris Mason if (newer_off == (u64)-1) 10864cb5300bSChris Mason break; 10874cb5300bSChris Mason 10884cb5300bSChris Mason newer_off = max(newer_off + 1, 10894cb5300bSChris Mason (u64)i << PAGE_CACHE_SHIFT); 10904cb5300bSChris Mason 10914cb5300bSChris Mason ret = find_new_extents(root, inode, 10924cb5300bSChris Mason newer_than, &newer_off, 10934cb5300bSChris Mason 64 * 1024); 10944cb5300bSChris Mason if (!ret) { 10954cb5300bSChris Mason range->start = newer_off; 10964cb5300bSChris Mason i = (newer_off & new_align) >> PAGE_CACHE_SHIFT; 10974cb5300bSChris Mason newer_left = newer_cluster; 10984cb5300bSChris Mason } else { 10994cb5300bSChris Mason break; 1100940100a4SChris Mason } 11014cb5300bSChris Mason } else { 1102940100a4SChris Mason i++; 1103f46b5a66SChristoph Hellwig } 11044cb5300bSChris Mason } 1105f46b5a66SChristoph Hellwig 11061e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) 11071e701a32SChris Mason filemap_flush(inode->i_mapping); 11081e701a32SChris Mason 11091e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 11101e701a32SChris Mason /* the filemap_flush will queue IO into the worker threads, but 11111e701a32SChris Mason * we have to make sure the IO is actually started and that 11121e701a32SChris Mason * ordered extents get created before we return 11131e701a32SChris Mason */ 11141e701a32SChris Mason atomic_inc(&root->fs_info->async_submit_draining); 11151e701a32SChris Mason while (atomic_read(&root->fs_info->nr_async_submits) || 11161e701a32SChris Mason atomic_read(&root->fs_info->async_delalloc_pages)) { 11171e701a32SChris Mason wait_event(root->fs_info->async_submit_wait, 11181e701a32SChris Mason (atomic_read(&root->fs_info->nr_async_submits) == 0 && 11191e701a32SChris Mason atomic_read(&root->fs_info->async_delalloc_pages) == 0)); 11201e701a32SChris Mason } 11211e701a32SChris Mason atomic_dec(&root->fs_info->async_submit_draining); 11221e701a32SChris Mason 11231e701a32SChris Mason mutex_lock(&inode->i_mutex); 1124261507a0SLi Zefan BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE; 11251e701a32SChris Mason mutex_unlock(&inode->i_mutex); 11261e701a32SChris Mason } 11271e701a32SChris Mason 11281a419d85SLi Zefan disk_super = &root->fs_info->super_copy; 11291a419d85SLi Zefan features = btrfs_super_incompat_flags(disk_super); 11301a419d85SLi Zefan if (range->compress_type == BTRFS_COMPRESS_LZO) { 11311a419d85SLi Zefan features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; 11321a419d85SLi Zefan btrfs_set_super_incompat_flags(disk_super, features); 11331a419d85SLi Zefan } 11341a419d85SLi Zefan 11354cb5300bSChris Mason if (!file) 11364cb5300bSChris Mason kfree(ra); 11374cb5300bSChris Mason return defrag_count; 1138940100a4SChris Mason 11394cb5300bSChris Mason out_ra: 11404cb5300bSChris Mason if (!file) 11414cb5300bSChris Mason kfree(ra); 11424cb5300bSChris Mason kfree(pages); 1143940100a4SChris Mason return ret; 1144f46b5a66SChristoph Hellwig } 1145f46b5a66SChristoph Hellwig 114676dda93cSYan, Zheng static noinline int btrfs_ioctl_resize(struct btrfs_root *root, 114776dda93cSYan, Zheng void __user *arg) 1148f46b5a66SChristoph Hellwig { 1149f46b5a66SChristoph Hellwig u64 new_size; 1150f46b5a66SChristoph Hellwig u64 old_size; 1151f46b5a66SChristoph Hellwig u64 devid = 1; 1152f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 1153f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 1154f46b5a66SChristoph Hellwig struct btrfs_device *device = NULL; 1155f46b5a66SChristoph Hellwig char *sizestr; 1156f46b5a66SChristoph Hellwig char *devstr = NULL; 1157f46b5a66SChristoph Hellwig int ret = 0; 1158f46b5a66SChristoph Hellwig int mod = 0; 1159f46b5a66SChristoph Hellwig 1160c146afadSYan Zheng if (root->fs_info->sb->s_flags & MS_RDONLY) 1161c146afadSYan Zheng return -EROFS; 1162c146afadSYan Zheng 1163e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 1164e441d54dSChris Mason return -EPERM; 1165e441d54dSChris Mason 1166dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1167dae7b665SLi Zefan if (IS_ERR(vol_args)) 1168dae7b665SLi Zefan return PTR_ERR(vol_args); 11695516e595SMark Fasheh 11705516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1171f46b5a66SChristoph Hellwig 11727d9eb12cSChris Mason mutex_lock(&root->fs_info->volume_mutex); 1173f46b5a66SChristoph Hellwig sizestr = vol_args->name; 1174f46b5a66SChristoph Hellwig devstr = strchr(sizestr, ':'); 1175f46b5a66SChristoph Hellwig if (devstr) { 1176f46b5a66SChristoph Hellwig char *end; 1177f46b5a66SChristoph Hellwig sizestr = devstr + 1; 1178f46b5a66SChristoph Hellwig *devstr = '\0'; 1179f46b5a66SChristoph Hellwig devstr = vol_args->name; 1180f46b5a66SChristoph Hellwig devid = simple_strtoull(devstr, &end, 10); 118121380931SJoel Becker printk(KERN_INFO "resizing devid %llu\n", 118221380931SJoel Becker (unsigned long long)devid); 1183f46b5a66SChristoph Hellwig } 11842b82032cSYan Zheng device = btrfs_find_device(root, devid, NULL, NULL); 1185f46b5a66SChristoph Hellwig if (!device) { 118621380931SJoel Becker printk(KERN_INFO "resizer unable to find device %llu\n", 118721380931SJoel Becker (unsigned long long)devid); 1188f46b5a66SChristoph Hellwig ret = -EINVAL; 1189f46b5a66SChristoph Hellwig goto out_unlock; 1190f46b5a66SChristoph Hellwig } 1191f46b5a66SChristoph Hellwig if (!strcmp(sizestr, "max")) 1192f46b5a66SChristoph Hellwig new_size = device->bdev->bd_inode->i_size; 1193f46b5a66SChristoph Hellwig else { 1194f46b5a66SChristoph Hellwig if (sizestr[0] == '-') { 1195f46b5a66SChristoph Hellwig mod = -1; 1196f46b5a66SChristoph Hellwig sizestr++; 1197f46b5a66SChristoph Hellwig } else if (sizestr[0] == '+') { 1198f46b5a66SChristoph Hellwig mod = 1; 1199f46b5a66SChristoph Hellwig sizestr++; 1200f46b5a66SChristoph Hellwig } 120191748467SAkinobu Mita new_size = memparse(sizestr, NULL); 1202f46b5a66SChristoph Hellwig if (new_size == 0) { 1203f46b5a66SChristoph Hellwig ret = -EINVAL; 1204f46b5a66SChristoph Hellwig goto out_unlock; 1205f46b5a66SChristoph Hellwig } 1206f46b5a66SChristoph Hellwig } 1207f46b5a66SChristoph Hellwig 1208f46b5a66SChristoph Hellwig old_size = device->total_bytes; 1209f46b5a66SChristoph Hellwig 1210f46b5a66SChristoph Hellwig if (mod < 0) { 1211f46b5a66SChristoph Hellwig if (new_size > old_size) { 1212f46b5a66SChristoph Hellwig ret = -EINVAL; 1213f46b5a66SChristoph Hellwig goto out_unlock; 1214f46b5a66SChristoph Hellwig } 1215f46b5a66SChristoph Hellwig new_size = old_size - new_size; 1216f46b5a66SChristoph Hellwig } else if (mod > 0) { 1217f46b5a66SChristoph Hellwig new_size = old_size + new_size; 1218f46b5a66SChristoph Hellwig } 1219f46b5a66SChristoph Hellwig 1220f46b5a66SChristoph Hellwig if (new_size < 256 * 1024 * 1024) { 1221f46b5a66SChristoph Hellwig ret = -EINVAL; 1222f46b5a66SChristoph Hellwig goto out_unlock; 1223f46b5a66SChristoph Hellwig } 1224f46b5a66SChristoph Hellwig if (new_size > device->bdev->bd_inode->i_size) { 1225f46b5a66SChristoph Hellwig ret = -EFBIG; 1226f46b5a66SChristoph Hellwig goto out_unlock; 1227f46b5a66SChristoph Hellwig } 1228f46b5a66SChristoph Hellwig 1229f46b5a66SChristoph Hellwig do_div(new_size, root->sectorsize); 1230f46b5a66SChristoph Hellwig new_size *= root->sectorsize; 1231f46b5a66SChristoph Hellwig 1232f46b5a66SChristoph Hellwig printk(KERN_INFO "new size for %s is %llu\n", 1233f46b5a66SChristoph Hellwig device->name, (unsigned long long)new_size); 1234f46b5a66SChristoph Hellwig 1235f46b5a66SChristoph Hellwig if (new_size > old_size) { 1236a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 123798d5dc13STsutomu Itoh if (IS_ERR(trans)) { 123898d5dc13STsutomu Itoh ret = PTR_ERR(trans); 123998d5dc13STsutomu Itoh goto out_unlock; 124098d5dc13STsutomu Itoh } 1241f46b5a66SChristoph Hellwig ret = btrfs_grow_device(trans, device, new_size); 1242f46b5a66SChristoph Hellwig btrfs_commit_transaction(trans, root); 1243f46b5a66SChristoph Hellwig } else { 1244f46b5a66SChristoph Hellwig ret = btrfs_shrink_device(device, new_size); 1245f46b5a66SChristoph Hellwig } 1246f46b5a66SChristoph Hellwig 1247f46b5a66SChristoph Hellwig out_unlock: 12487d9eb12cSChris Mason mutex_unlock(&root->fs_info->volume_mutex); 1249f46b5a66SChristoph Hellwig kfree(vol_args); 1250f46b5a66SChristoph Hellwig return ret; 1251f46b5a66SChristoph Hellwig } 1252f46b5a66SChristoph Hellwig 125372fd032eSSage Weil static noinline int btrfs_ioctl_snap_create_transid(struct file *file, 125472fd032eSSage Weil char *name, 125572fd032eSSage Weil unsigned long fd, 125672fd032eSSage Weil int subvol, 1257b83cc969SLi Zefan u64 *transid, 1258b83cc969SLi Zefan bool readonly) 1259f46b5a66SChristoph Hellwig { 1260cb8e7090SChristoph Hellwig struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; 12613de4586cSChris Mason struct file *src_file; 1262f46b5a66SChristoph Hellwig int namelen; 12633de4586cSChris Mason int ret = 0; 1264f46b5a66SChristoph Hellwig 1265c146afadSYan Zheng if (root->fs_info->sb->s_flags & MS_RDONLY) 1266c146afadSYan Zheng return -EROFS; 1267c146afadSYan Zheng 126872fd032eSSage Weil namelen = strlen(name); 126972fd032eSSage Weil if (strchr(name, '/')) { 1270f46b5a66SChristoph Hellwig ret = -EINVAL; 1271f46b5a66SChristoph Hellwig goto out; 1272f46b5a66SChristoph Hellwig } 1273f46b5a66SChristoph Hellwig 12743de4586cSChris Mason if (subvol) { 127572fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 1276b83cc969SLi Zefan NULL, transid, readonly); 1277cb8e7090SChristoph Hellwig } else { 12783de4586cSChris Mason struct inode *src_inode; 127972fd032eSSage Weil src_file = fget(fd); 12803de4586cSChris Mason if (!src_file) { 12813de4586cSChris Mason ret = -EINVAL; 12823de4586cSChris Mason goto out; 12833de4586cSChris Mason } 12843de4586cSChris Mason 12853de4586cSChris Mason src_inode = src_file->f_path.dentry->d_inode; 12863de4586cSChris Mason if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) { 1287d397712bSChris Mason printk(KERN_INFO "btrfs: Snapshot src from " 1288d397712bSChris Mason "another FS\n"); 12893de4586cSChris Mason ret = -EINVAL; 12903de4586cSChris Mason fput(src_file); 12913de4586cSChris Mason goto out; 12923de4586cSChris Mason } 129372fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 129472fd032eSSage Weil BTRFS_I(src_inode)->root, 1295b83cc969SLi Zefan transid, readonly); 12963de4586cSChris Mason fput(src_file); 1297cb8e7090SChristoph Hellwig } 1298f46b5a66SChristoph Hellwig out: 129972fd032eSSage Weil return ret; 130072fd032eSSage Weil } 130172fd032eSSage Weil 130272fd032eSSage Weil static noinline int btrfs_ioctl_snap_create(struct file *file, 1303fa0d2b9bSLi Zefan void __user *arg, int subvol) 130472fd032eSSage Weil { 1305fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args *vol_args; 130672fd032eSSage Weil int ret; 130772fd032eSSage Weil 1308fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1309fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1310fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1311fa0d2b9bSLi Zefan vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1312fa0d2b9bSLi Zefan 1313fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 1314b83cc969SLi Zefan vol_args->fd, subvol, 1315b83cc969SLi Zefan NULL, false); 1316fa0d2b9bSLi Zefan 1317fa0d2b9bSLi Zefan kfree(vol_args); 1318fa0d2b9bSLi Zefan return ret; 1319fa0d2b9bSLi Zefan } 1320fa0d2b9bSLi Zefan 1321fa0d2b9bSLi Zefan static noinline int btrfs_ioctl_snap_create_v2(struct file *file, 1322fa0d2b9bSLi Zefan void __user *arg, int subvol) 1323fa0d2b9bSLi Zefan { 1324fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args_v2 *vol_args; 1325fa0d2b9bSLi Zefan int ret; 1326fdfb1e4fSLi Zefan u64 transid = 0; 1327fdfb1e4fSLi Zefan u64 *ptr = NULL; 1328b83cc969SLi Zefan bool readonly = false; 132972fd032eSSage Weil 1330fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1331fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1332fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1333fa0d2b9bSLi Zefan vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 1334fdfb1e4fSLi Zefan 1335b83cc969SLi Zefan if (vol_args->flags & 1336b83cc969SLi Zefan ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) { 1337b83cc969SLi Zefan ret = -EOPNOTSUPP; 1338fdfb1e4fSLi Zefan goto out; 1339fdfb1e4fSLi Zefan } 1340fdfb1e4fSLi Zefan 1341fa0d2b9bSLi Zefan if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC) 1342fdfb1e4fSLi Zefan ptr = &transid; 1343b83cc969SLi Zefan if (vol_args->flags & BTRFS_SUBVOL_RDONLY) 1344b83cc969SLi Zefan readonly = true; 134575eaa0e2SSage Weil 1346fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 1347b83cc969SLi Zefan vol_args->fd, subvol, 1348b83cc969SLi Zefan ptr, readonly); 134975eaa0e2SSage Weil 1350fdfb1e4fSLi Zefan if (ret == 0 && ptr && 135175eaa0e2SSage Weil copy_to_user(arg + 1352fdfb1e4fSLi Zefan offsetof(struct btrfs_ioctl_vol_args_v2, 1353fdfb1e4fSLi Zefan transid), ptr, sizeof(*ptr))) 135475eaa0e2SSage Weil ret = -EFAULT; 1355fdfb1e4fSLi Zefan out: 1356f46b5a66SChristoph Hellwig kfree(vol_args); 1357f46b5a66SChristoph Hellwig return ret; 1358f46b5a66SChristoph Hellwig } 1359f46b5a66SChristoph Hellwig 13600caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_getflags(struct file *file, 13610caa102dSLi Zefan void __user *arg) 13620caa102dSLi Zefan { 13630caa102dSLi Zefan struct inode *inode = fdentry(file)->d_inode; 13640caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 13650caa102dSLi Zefan int ret = 0; 13660caa102dSLi Zefan u64 flags = 0; 13670caa102dSLi Zefan 136833345d01SLi Zefan if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) 13690caa102dSLi Zefan return -EINVAL; 13700caa102dSLi Zefan 13710caa102dSLi Zefan down_read(&root->fs_info->subvol_sem); 13720caa102dSLi Zefan if (btrfs_root_readonly(root)) 13730caa102dSLi Zefan flags |= BTRFS_SUBVOL_RDONLY; 13740caa102dSLi Zefan up_read(&root->fs_info->subvol_sem); 13750caa102dSLi Zefan 13760caa102dSLi Zefan if (copy_to_user(arg, &flags, sizeof(flags))) 13770caa102dSLi Zefan ret = -EFAULT; 13780caa102dSLi Zefan 13790caa102dSLi Zefan return ret; 13800caa102dSLi Zefan } 13810caa102dSLi Zefan 13820caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_setflags(struct file *file, 13830caa102dSLi Zefan void __user *arg) 13840caa102dSLi Zefan { 13850caa102dSLi Zefan struct inode *inode = fdentry(file)->d_inode; 13860caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 13870caa102dSLi Zefan struct btrfs_trans_handle *trans; 13880caa102dSLi Zefan u64 root_flags; 13890caa102dSLi Zefan u64 flags; 13900caa102dSLi Zefan int ret = 0; 13910caa102dSLi Zefan 13920caa102dSLi Zefan if (root->fs_info->sb->s_flags & MS_RDONLY) 13930caa102dSLi Zefan return -EROFS; 13940caa102dSLi Zefan 139533345d01SLi Zefan if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) 13960caa102dSLi Zefan return -EINVAL; 13970caa102dSLi Zefan 13980caa102dSLi Zefan if (copy_from_user(&flags, arg, sizeof(flags))) 13990caa102dSLi Zefan return -EFAULT; 14000caa102dSLi Zefan 1401b4dc2b8cSLi Zefan if (flags & BTRFS_SUBVOL_CREATE_ASYNC) 14020caa102dSLi Zefan return -EINVAL; 14030caa102dSLi Zefan 14040caa102dSLi Zefan if (flags & ~BTRFS_SUBVOL_RDONLY) 14050caa102dSLi Zefan return -EOPNOTSUPP; 14060caa102dSLi Zefan 14072e149670SSerge E. Hallyn if (!inode_owner_or_capable(inode)) 1408b4dc2b8cSLi Zefan return -EACCES; 1409b4dc2b8cSLi Zefan 14100caa102dSLi Zefan down_write(&root->fs_info->subvol_sem); 14110caa102dSLi Zefan 14120caa102dSLi Zefan /* nothing to do */ 14130caa102dSLi Zefan if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) 14140caa102dSLi Zefan goto out; 14150caa102dSLi Zefan 14160caa102dSLi Zefan root_flags = btrfs_root_flags(&root->root_item); 14170caa102dSLi Zefan if (flags & BTRFS_SUBVOL_RDONLY) 14180caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 14190caa102dSLi Zefan root_flags | BTRFS_ROOT_SUBVOL_RDONLY); 14200caa102dSLi Zefan else 14210caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 14220caa102dSLi Zefan root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); 14230caa102dSLi Zefan 14240caa102dSLi Zefan trans = btrfs_start_transaction(root, 1); 14250caa102dSLi Zefan if (IS_ERR(trans)) { 14260caa102dSLi Zefan ret = PTR_ERR(trans); 14270caa102dSLi Zefan goto out_reset; 14280caa102dSLi Zefan } 14290caa102dSLi Zefan 1430b4dc2b8cSLi Zefan ret = btrfs_update_root(trans, root->fs_info->tree_root, 14310caa102dSLi Zefan &root->root_key, &root->root_item); 14320caa102dSLi Zefan 14330caa102dSLi Zefan btrfs_commit_transaction(trans, root); 14340caa102dSLi Zefan out_reset: 14350caa102dSLi Zefan if (ret) 14360caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, root_flags); 14370caa102dSLi Zefan out: 14380caa102dSLi Zefan up_write(&root->fs_info->subvol_sem); 14390caa102dSLi Zefan return ret; 14400caa102dSLi Zefan } 14410caa102dSLi Zefan 144276dda93cSYan, Zheng /* 144376dda93cSYan, Zheng * helper to check if the subvolume references other subvolumes 144476dda93cSYan, Zheng */ 144576dda93cSYan, Zheng static noinline int may_destroy_subvol(struct btrfs_root *root) 144676dda93cSYan, Zheng { 144776dda93cSYan, Zheng struct btrfs_path *path; 144876dda93cSYan, Zheng struct btrfs_key key; 144976dda93cSYan, Zheng int ret; 145076dda93cSYan, Zheng 145176dda93cSYan, Zheng path = btrfs_alloc_path(); 145276dda93cSYan, Zheng if (!path) 145376dda93cSYan, Zheng return -ENOMEM; 145476dda93cSYan, Zheng 145576dda93cSYan, Zheng key.objectid = root->root_key.objectid; 145676dda93cSYan, Zheng key.type = BTRFS_ROOT_REF_KEY; 145776dda93cSYan, Zheng key.offset = (u64)-1; 145876dda93cSYan, Zheng 145976dda93cSYan, Zheng ret = btrfs_search_slot(NULL, root->fs_info->tree_root, 146076dda93cSYan, Zheng &key, path, 0, 0); 146176dda93cSYan, Zheng if (ret < 0) 146276dda93cSYan, Zheng goto out; 146376dda93cSYan, Zheng BUG_ON(ret == 0); 146476dda93cSYan, Zheng 146576dda93cSYan, Zheng ret = 0; 146676dda93cSYan, Zheng if (path->slots[0] > 0) { 146776dda93cSYan, Zheng path->slots[0]--; 146876dda93cSYan, Zheng btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 146976dda93cSYan, Zheng if (key.objectid == root->root_key.objectid && 147076dda93cSYan, Zheng key.type == BTRFS_ROOT_REF_KEY) 147176dda93cSYan, Zheng ret = -ENOTEMPTY; 147276dda93cSYan, Zheng } 147376dda93cSYan, Zheng out: 147476dda93cSYan, Zheng btrfs_free_path(path); 147576dda93cSYan, Zheng return ret; 147676dda93cSYan, Zheng } 147776dda93cSYan, Zheng 1478ac8e9819SChris Mason static noinline int key_in_sk(struct btrfs_key *key, 1479ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk) 1480ac8e9819SChris Mason { 1481abc6e134SChris Mason struct btrfs_key test; 1482abc6e134SChris Mason int ret; 1483abc6e134SChris Mason 1484abc6e134SChris Mason test.objectid = sk->min_objectid; 1485abc6e134SChris Mason test.type = sk->min_type; 1486abc6e134SChris Mason test.offset = sk->min_offset; 1487abc6e134SChris Mason 1488abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1489abc6e134SChris Mason if (ret < 0) 1490ac8e9819SChris Mason return 0; 1491abc6e134SChris Mason 1492abc6e134SChris Mason test.objectid = sk->max_objectid; 1493abc6e134SChris Mason test.type = sk->max_type; 1494abc6e134SChris Mason test.offset = sk->max_offset; 1495abc6e134SChris Mason 1496abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1497abc6e134SChris Mason if (ret > 0) 1498ac8e9819SChris Mason return 0; 1499ac8e9819SChris Mason return 1; 1500ac8e9819SChris Mason } 1501ac8e9819SChris Mason 1502ac8e9819SChris Mason static noinline int copy_to_sk(struct btrfs_root *root, 1503ac8e9819SChris Mason struct btrfs_path *path, 1504ac8e9819SChris Mason struct btrfs_key *key, 1505ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk, 1506ac8e9819SChris Mason char *buf, 1507ac8e9819SChris Mason unsigned long *sk_offset, 1508ac8e9819SChris Mason int *num_found) 1509ac8e9819SChris Mason { 1510ac8e9819SChris Mason u64 found_transid; 1511ac8e9819SChris Mason struct extent_buffer *leaf; 1512ac8e9819SChris Mason struct btrfs_ioctl_search_header sh; 1513ac8e9819SChris Mason unsigned long item_off; 1514ac8e9819SChris Mason unsigned long item_len; 1515ac8e9819SChris Mason int nritems; 1516ac8e9819SChris Mason int i; 1517ac8e9819SChris Mason int slot; 1518ac8e9819SChris Mason int ret = 0; 1519ac8e9819SChris Mason 1520ac8e9819SChris Mason leaf = path->nodes[0]; 1521ac8e9819SChris Mason slot = path->slots[0]; 1522ac8e9819SChris Mason nritems = btrfs_header_nritems(leaf); 1523ac8e9819SChris Mason 1524ac8e9819SChris Mason if (btrfs_header_generation(leaf) > sk->max_transid) { 1525ac8e9819SChris Mason i = nritems; 1526ac8e9819SChris Mason goto advance_key; 1527ac8e9819SChris Mason } 1528ac8e9819SChris Mason found_transid = btrfs_header_generation(leaf); 1529ac8e9819SChris Mason 1530ac8e9819SChris Mason for (i = slot; i < nritems; i++) { 1531ac8e9819SChris Mason item_off = btrfs_item_ptr_offset(leaf, i); 1532ac8e9819SChris Mason item_len = btrfs_item_size_nr(leaf, i); 1533ac8e9819SChris Mason 1534ac8e9819SChris Mason if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE) 1535ac8e9819SChris Mason item_len = 0; 1536ac8e9819SChris Mason 1537ac8e9819SChris Mason if (sizeof(sh) + item_len + *sk_offset > 1538ac8e9819SChris Mason BTRFS_SEARCH_ARGS_BUFSIZE) { 1539ac8e9819SChris Mason ret = 1; 1540ac8e9819SChris Mason goto overflow; 1541ac8e9819SChris Mason } 1542ac8e9819SChris Mason 1543ac8e9819SChris Mason btrfs_item_key_to_cpu(leaf, key, i); 1544ac8e9819SChris Mason if (!key_in_sk(key, sk)) 1545ac8e9819SChris Mason continue; 1546ac8e9819SChris Mason 1547ac8e9819SChris Mason sh.objectid = key->objectid; 1548ac8e9819SChris Mason sh.offset = key->offset; 1549ac8e9819SChris Mason sh.type = key->type; 1550ac8e9819SChris Mason sh.len = item_len; 1551ac8e9819SChris Mason sh.transid = found_transid; 1552ac8e9819SChris Mason 1553ac8e9819SChris Mason /* copy search result header */ 1554ac8e9819SChris Mason memcpy(buf + *sk_offset, &sh, sizeof(sh)); 1555ac8e9819SChris Mason *sk_offset += sizeof(sh); 1556ac8e9819SChris Mason 1557ac8e9819SChris Mason if (item_len) { 1558ac8e9819SChris Mason char *p = buf + *sk_offset; 1559ac8e9819SChris Mason /* copy the item */ 1560ac8e9819SChris Mason read_extent_buffer(leaf, p, 1561ac8e9819SChris Mason item_off, item_len); 1562ac8e9819SChris Mason *sk_offset += item_len; 1563ac8e9819SChris Mason } 1564e2156867SHugo Mills (*num_found)++; 1565ac8e9819SChris Mason 1566ac8e9819SChris Mason if (*num_found >= sk->nr_items) 1567ac8e9819SChris Mason break; 1568ac8e9819SChris Mason } 1569ac8e9819SChris Mason advance_key: 1570ac8e9819SChris Mason ret = 0; 1571abc6e134SChris Mason if (key->offset < (u64)-1 && key->offset < sk->max_offset) 1572abc6e134SChris Mason key->offset++; 1573abc6e134SChris Mason else if (key->type < (u8)-1 && key->type < sk->max_type) { 1574abc6e134SChris Mason key->offset = 0; 1575abc6e134SChris Mason key->type++; 1576abc6e134SChris Mason } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) { 1577abc6e134SChris Mason key->offset = 0; 1578abc6e134SChris Mason key->type = 0; 1579abc6e134SChris Mason key->objectid++; 1580abc6e134SChris Mason } else 1581abc6e134SChris Mason ret = 1; 1582ac8e9819SChris Mason overflow: 1583ac8e9819SChris Mason return ret; 1584ac8e9819SChris Mason } 1585ac8e9819SChris Mason 1586ac8e9819SChris Mason static noinline int search_ioctl(struct inode *inode, 1587ac8e9819SChris Mason struct btrfs_ioctl_search_args *args) 1588ac8e9819SChris Mason { 1589ac8e9819SChris Mason struct btrfs_root *root; 1590ac8e9819SChris Mason struct btrfs_key key; 1591ac8e9819SChris Mason struct btrfs_key max_key; 1592ac8e9819SChris Mason struct btrfs_path *path; 1593ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk = &args->key; 1594ac8e9819SChris Mason struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info; 1595ac8e9819SChris Mason int ret; 1596ac8e9819SChris Mason int num_found = 0; 1597ac8e9819SChris Mason unsigned long sk_offset = 0; 1598ac8e9819SChris Mason 1599ac8e9819SChris Mason path = btrfs_alloc_path(); 1600ac8e9819SChris Mason if (!path) 1601ac8e9819SChris Mason return -ENOMEM; 1602ac8e9819SChris Mason 1603ac8e9819SChris Mason if (sk->tree_id == 0) { 1604ac8e9819SChris Mason /* search the root of the inode that was passed */ 1605ac8e9819SChris Mason root = BTRFS_I(inode)->root; 1606ac8e9819SChris Mason } else { 1607ac8e9819SChris Mason key.objectid = sk->tree_id; 1608ac8e9819SChris Mason key.type = BTRFS_ROOT_ITEM_KEY; 1609ac8e9819SChris Mason key.offset = (u64)-1; 1610ac8e9819SChris Mason root = btrfs_read_fs_root_no_name(info, &key); 1611ac8e9819SChris Mason if (IS_ERR(root)) { 1612ac8e9819SChris Mason printk(KERN_ERR "could not find root %llu\n", 1613ac8e9819SChris Mason sk->tree_id); 1614ac8e9819SChris Mason btrfs_free_path(path); 1615ac8e9819SChris Mason return -ENOENT; 1616ac8e9819SChris Mason } 1617ac8e9819SChris Mason } 1618ac8e9819SChris Mason 1619ac8e9819SChris Mason key.objectid = sk->min_objectid; 1620ac8e9819SChris Mason key.type = sk->min_type; 1621ac8e9819SChris Mason key.offset = sk->min_offset; 1622ac8e9819SChris Mason 1623ac8e9819SChris Mason max_key.objectid = sk->max_objectid; 1624ac8e9819SChris Mason max_key.type = sk->max_type; 1625ac8e9819SChris Mason max_key.offset = sk->max_offset; 1626ac8e9819SChris Mason 1627ac8e9819SChris Mason path->keep_locks = 1; 1628ac8e9819SChris Mason 1629ac8e9819SChris Mason while(1) { 1630ac8e9819SChris Mason ret = btrfs_search_forward(root, &key, &max_key, path, 0, 1631ac8e9819SChris Mason sk->min_transid); 1632ac8e9819SChris Mason if (ret != 0) { 1633ac8e9819SChris Mason if (ret > 0) 1634ac8e9819SChris Mason ret = 0; 1635ac8e9819SChris Mason goto err; 1636ac8e9819SChris Mason } 1637ac8e9819SChris Mason ret = copy_to_sk(root, path, &key, sk, args->buf, 1638ac8e9819SChris Mason &sk_offset, &num_found); 1639b3b4aa74SDavid Sterba btrfs_release_path(path); 1640ac8e9819SChris Mason if (ret || num_found >= sk->nr_items) 1641ac8e9819SChris Mason break; 1642ac8e9819SChris Mason 1643ac8e9819SChris Mason } 1644ac8e9819SChris Mason ret = 0; 1645ac8e9819SChris Mason err: 1646ac8e9819SChris Mason sk->nr_items = num_found; 1647ac8e9819SChris Mason btrfs_free_path(path); 1648ac8e9819SChris Mason return ret; 1649ac8e9819SChris Mason } 1650ac8e9819SChris Mason 1651ac8e9819SChris Mason static noinline int btrfs_ioctl_tree_search(struct file *file, 1652ac8e9819SChris Mason void __user *argp) 1653ac8e9819SChris Mason { 1654ac8e9819SChris Mason struct btrfs_ioctl_search_args *args; 1655ac8e9819SChris Mason struct inode *inode; 1656ac8e9819SChris Mason int ret; 1657ac8e9819SChris Mason 1658ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 1659ac8e9819SChris Mason return -EPERM; 1660ac8e9819SChris Mason 16612354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 16622354d08fSJulia Lawall if (IS_ERR(args)) 16632354d08fSJulia Lawall return PTR_ERR(args); 1664ac8e9819SChris Mason 1665ac8e9819SChris Mason inode = fdentry(file)->d_inode; 1666ac8e9819SChris Mason ret = search_ioctl(inode, args); 1667ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 1668ac8e9819SChris Mason ret = -EFAULT; 1669ac8e9819SChris Mason kfree(args); 1670ac8e9819SChris Mason return ret; 1671ac8e9819SChris Mason } 1672ac8e9819SChris Mason 167398d377a0STARUISI Hiroaki /* 1674ac8e9819SChris Mason * Search INODE_REFs to identify path name of 'dirid' directory 1675ac8e9819SChris Mason * in a 'tree_id' tree. and sets path name to 'name'. 167698d377a0STARUISI Hiroaki */ 167798d377a0STARUISI Hiroaki static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, 167898d377a0STARUISI Hiroaki u64 tree_id, u64 dirid, char *name) 167998d377a0STARUISI Hiroaki { 168098d377a0STARUISI Hiroaki struct btrfs_root *root; 168198d377a0STARUISI Hiroaki struct btrfs_key key; 1682ac8e9819SChris Mason char *ptr; 168398d377a0STARUISI Hiroaki int ret = -1; 168498d377a0STARUISI Hiroaki int slot; 168598d377a0STARUISI Hiroaki int len; 168698d377a0STARUISI Hiroaki int total_len = 0; 168798d377a0STARUISI Hiroaki struct btrfs_inode_ref *iref; 168898d377a0STARUISI Hiroaki struct extent_buffer *l; 168998d377a0STARUISI Hiroaki struct btrfs_path *path; 169098d377a0STARUISI Hiroaki 169198d377a0STARUISI Hiroaki if (dirid == BTRFS_FIRST_FREE_OBJECTID) { 169298d377a0STARUISI Hiroaki name[0]='\0'; 169398d377a0STARUISI Hiroaki return 0; 169498d377a0STARUISI Hiroaki } 169598d377a0STARUISI Hiroaki 169698d377a0STARUISI Hiroaki path = btrfs_alloc_path(); 169798d377a0STARUISI Hiroaki if (!path) 169898d377a0STARUISI Hiroaki return -ENOMEM; 169998d377a0STARUISI Hiroaki 1700ac8e9819SChris Mason ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX]; 170198d377a0STARUISI Hiroaki 170298d377a0STARUISI Hiroaki key.objectid = tree_id; 170398d377a0STARUISI Hiroaki key.type = BTRFS_ROOT_ITEM_KEY; 170498d377a0STARUISI Hiroaki key.offset = (u64)-1; 170598d377a0STARUISI Hiroaki root = btrfs_read_fs_root_no_name(info, &key); 170698d377a0STARUISI Hiroaki if (IS_ERR(root)) { 170798d377a0STARUISI Hiroaki printk(KERN_ERR "could not find root %llu\n", tree_id); 17088ad6fcabSChris Mason ret = -ENOENT; 17098ad6fcabSChris Mason goto out; 171098d377a0STARUISI Hiroaki } 171198d377a0STARUISI Hiroaki 171298d377a0STARUISI Hiroaki key.objectid = dirid; 171398d377a0STARUISI Hiroaki key.type = BTRFS_INODE_REF_KEY; 17148ad6fcabSChris Mason key.offset = (u64)-1; 171598d377a0STARUISI Hiroaki 171698d377a0STARUISI Hiroaki while(1) { 171798d377a0STARUISI Hiroaki ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 171898d377a0STARUISI Hiroaki if (ret < 0) 171998d377a0STARUISI Hiroaki goto out; 172098d377a0STARUISI Hiroaki 172198d377a0STARUISI Hiroaki l = path->nodes[0]; 172298d377a0STARUISI Hiroaki slot = path->slots[0]; 17238ad6fcabSChris Mason if (ret > 0 && slot > 0) 17248ad6fcabSChris Mason slot--; 172598d377a0STARUISI Hiroaki btrfs_item_key_to_cpu(l, &key, slot); 172698d377a0STARUISI Hiroaki 172798d377a0STARUISI Hiroaki if (ret > 0 && (key.objectid != dirid || 1728ac8e9819SChris Mason key.type != BTRFS_INODE_REF_KEY)) { 1729ac8e9819SChris Mason ret = -ENOENT; 173098d377a0STARUISI Hiroaki goto out; 1731ac8e9819SChris Mason } 173298d377a0STARUISI Hiroaki 173398d377a0STARUISI Hiroaki iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); 173498d377a0STARUISI Hiroaki len = btrfs_inode_ref_name_len(l, iref); 173598d377a0STARUISI Hiroaki ptr -= len + 1; 173698d377a0STARUISI Hiroaki total_len += len + 1; 1737ac8e9819SChris Mason if (ptr < name) 173898d377a0STARUISI Hiroaki goto out; 173998d377a0STARUISI Hiroaki 174098d377a0STARUISI Hiroaki *(ptr + len) = '/'; 174198d377a0STARUISI Hiroaki read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len); 174298d377a0STARUISI Hiroaki 174398d377a0STARUISI Hiroaki if (key.offset == BTRFS_FIRST_FREE_OBJECTID) 174498d377a0STARUISI Hiroaki break; 174598d377a0STARUISI Hiroaki 1746b3b4aa74SDavid Sterba btrfs_release_path(path); 174798d377a0STARUISI Hiroaki key.objectid = key.offset; 17488ad6fcabSChris Mason key.offset = (u64)-1; 174998d377a0STARUISI Hiroaki dirid = key.objectid; 175098d377a0STARUISI Hiroaki 175198d377a0STARUISI Hiroaki } 1752ac8e9819SChris Mason if (ptr < name) 175398d377a0STARUISI Hiroaki goto out; 1754ac8e9819SChris Mason memcpy(name, ptr, total_len); 175598d377a0STARUISI Hiroaki name[total_len]='\0'; 175698d377a0STARUISI Hiroaki ret = 0; 175798d377a0STARUISI Hiroaki out: 175898d377a0STARUISI Hiroaki btrfs_free_path(path); 1759ac8e9819SChris Mason return ret; 1760ac8e9819SChris Mason } 1761ac8e9819SChris Mason 1762ac8e9819SChris Mason static noinline int btrfs_ioctl_ino_lookup(struct file *file, 1763ac8e9819SChris Mason void __user *argp) 1764ac8e9819SChris Mason { 1765ac8e9819SChris Mason struct btrfs_ioctl_ino_lookup_args *args; 1766ac8e9819SChris Mason struct inode *inode; 1767ac8e9819SChris Mason int ret; 1768ac8e9819SChris Mason 1769ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 1770ac8e9819SChris Mason return -EPERM; 1771ac8e9819SChris Mason 17722354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 17732354d08fSJulia Lawall if (IS_ERR(args)) 17742354d08fSJulia Lawall return PTR_ERR(args); 1775c2b96929SDan Carpenter 1776ac8e9819SChris Mason inode = fdentry(file)->d_inode; 1777ac8e9819SChris Mason 17781b53ac4dSChris Mason if (args->treeid == 0) 17791b53ac4dSChris Mason args->treeid = BTRFS_I(inode)->root->root_key.objectid; 17801b53ac4dSChris Mason 1781ac8e9819SChris Mason ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, 1782ac8e9819SChris Mason args->treeid, args->objectid, 1783ac8e9819SChris Mason args->name); 1784ac8e9819SChris Mason 1785ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 1786ac8e9819SChris Mason ret = -EFAULT; 1787ac8e9819SChris Mason 1788ac8e9819SChris Mason kfree(args); 178998d377a0STARUISI Hiroaki return ret; 179098d377a0STARUISI Hiroaki } 179198d377a0STARUISI Hiroaki 179276dda93cSYan, Zheng static noinline int btrfs_ioctl_snap_destroy(struct file *file, 179376dda93cSYan, Zheng void __user *arg) 179476dda93cSYan, Zheng { 179576dda93cSYan, Zheng struct dentry *parent = fdentry(file); 179676dda93cSYan, Zheng struct dentry *dentry; 179776dda93cSYan, Zheng struct inode *dir = parent->d_inode; 179876dda93cSYan, Zheng struct inode *inode; 179976dda93cSYan, Zheng struct btrfs_root *root = BTRFS_I(dir)->root; 180076dda93cSYan, Zheng struct btrfs_root *dest = NULL; 180176dda93cSYan, Zheng struct btrfs_ioctl_vol_args *vol_args; 180276dda93cSYan, Zheng struct btrfs_trans_handle *trans; 180376dda93cSYan, Zheng int namelen; 180476dda93cSYan, Zheng int ret; 180576dda93cSYan, Zheng int err = 0; 180676dda93cSYan, Zheng 180776dda93cSYan, Zheng vol_args = memdup_user(arg, sizeof(*vol_args)); 180876dda93cSYan, Zheng if (IS_ERR(vol_args)) 180976dda93cSYan, Zheng return PTR_ERR(vol_args); 181076dda93cSYan, Zheng 181176dda93cSYan, Zheng vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 181276dda93cSYan, Zheng namelen = strlen(vol_args->name); 181376dda93cSYan, Zheng if (strchr(vol_args->name, '/') || 181476dda93cSYan, Zheng strncmp(vol_args->name, "..", namelen) == 0) { 181576dda93cSYan, Zheng err = -EINVAL; 181676dda93cSYan, Zheng goto out; 181776dda93cSYan, Zheng } 181876dda93cSYan, Zheng 181976dda93cSYan, Zheng err = mnt_want_write(file->f_path.mnt); 182076dda93cSYan, Zheng if (err) 182176dda93cSYan, Zheng goto out; 182276dda93cSYan, Zheng 182376dda93cSYan, Zheng mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT); 182476dda93cSYan, Zheng dentry = lookup_one_len(vol_args->name, parent, namelen); 182576dda93cSYan, Zheng if (IS_ERR(dentry)) { 182676dda93cSYan, Zheng err = PTR_ERR(dentry); 182776dda93cSYan, Zheng goto out_unlock_dir; 182876dda93cSYan, Zheng } 182976dda93cSYan, Zheng 183076dda93cSYan, Zheng if (!dentry->d_inode) { 183176dda93cSYan, Zheng err = -ENOENT; 183276dda93cSYan, Zheng goto out_dput; 183376dda93cSYan, Zheng } 183476dda93cSYan, Zheng 183576dda93cSYan, Zheng inode = dentry->d_inode; 18364260f7c7SSage Weil dest = BTRFS_I(inode)->root; 18374260f7c7SSage Weil if (!capable(CAP_SYS_ADMIN)){ 18384260f7c7SSage Weil /* 18394260f7c7SSage Weil * Regular user. Only allow this with a special mount 18404260f7c7SSage Weil * option, when the user has write+exec access to the 18414260f7c7SSage Weil * subvol root, and when rmdir(2) would have been 18424260f7c7SSage Weil * allowed. 18434260f7c7SSage Weil * 18444260f7c7SSage Weil * Note that this is _not_ check that the subvol is 18454260f7c7SSage Weil * empty or doesn't contain data that we wouldn't 18464260f7c7SSage Weil * otherwise be able to delete. 18474260f7c7SSage Weil * 18484260f7c7SSage Weil * Users who want to delete empty subvols should try 18494260f7c7SSage Weil * rmdir(2). 18504260f7c7SSage Weil */ 18514260f7c7SSage Weil err = -EPERM; 18524260f7c7SSage Weil if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED)) 18534260f7c7SSage Weil goto out_dput; 18544260f7c7SSage Weil 18554260f7c7SSage Weil /* 18564260f7c7SSage Weil * Do not allow deletion if the parent dir is the same 18574260f7c7SSage Weil * as the dir to be deleted. That means the ioctl 18584260f7c7SSage Weil * must be called on the dentry referencing the root 18594260f7c7SSage Weil * of the subvol, not a random directory contained 18604260f7c7SSage Weil * within it. 18614260f7c7SSage Weil */ 18624260f7c7SSage Weil err = -EINVAL; 18634260f7c7SSage Weil if (root == dest) 18644260f7c7SSage Weil goto out_dput; 18654260f7c7SSage Weil 18664260f7c7SSage Weil err = inode_permission(inode, MAY_WRITE | MAY_EXEC); 18674260f7c7SSage Weil if (err) 18684260f7c7SSage Weil goto out_dput; 18694260f7c7SSage Weil 18704260f7c7SSage Weil /* check if subvolume may be deleted by a non-root user */ 18714260f7c7SSage Weil err = btrfs_may_delete(dir, dentry, 1); 18724260f7c7SSage Weil if (err) 18734260f7c7SSage Weil goto out_dput; 18744260f7c7SSage Weil } 18754260f7c7SSage Weil 187633345d01SLi Zefan if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { 187776dda93cSYan, Zheng err = -EINVAL; 187876dda93cSYan, Zheng goto out_dput; 187976dda93cSYan, Zheng } 188076dda93cSYan, Zheng 188176dda93cSYan, Zheng mutex_lock(&inode->i_mutex); 188276dda93cSYan, Zheng err = d_invalidate(dentry); 188376dda93cSYan, Zheng if (err) 188476dda93cSYan, Zheng goto out_unlock; 188576dda93cSYan, Zheng 188676dda93cSYan, Zheng down_write(&root->fs_info->subvol_sem); 188776dda93cSYan, Zheng 188876dda93cSYan, Zheng err = may_destroy_subvol(dest); 188976dda93cSYan, Zheng if (err) 189076dda93cSYan, Zheng goto out_up_write; 189176dda93cSYan, Zheng 1892a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 1893a22285a6SYan, Zheng if (IS_ERR(trans)) { 1894a22285a6SYan, Zheng err = PTR_ERR(trans); 1895d327099aSDan Carpenter goto out_up_write; 1896a22285a6SYan, Zheng } 1897a22285a6SYan, Zheng trans->block_rsv = &root->fs_info->global_block_rsv; 1898a22285a6SYan, Zheng 189976dda93cSYan, Zheng ret = btrfs_unlink_subvol(trans, root, dir, 190076dda93cSYan, Zheng dest->root_key.objectid, 190176dda93cSYan, Zheng dentry->d_name.name, 190276dda93cSYan, Zheng dentry->d_name.len); 190376dda93cSYan, Zheng BUG_ON(ret); 190476dda93cSYan, Zheng 190576dda93cSYan, Zheng btrfs_record_root_in_trans(trans, dest); 190676dda93cSYan, Zheng 190776dda93cSYan, Zheng memset(&dest->root_item.drop_progress, 0, 190876dda93cSYan, Zheng sizeof(dest->root_item.drop_progress)); 190976dda93cSYan, Zheng dest->root_item.drop_level = 0; 191076dda93cSYan, Zheng btrfs_set_root_refs(&dest->root_item, 0); 191176dda93cSYan, Zheng 1912d68fc57bSYan, Zheng if (!xchg(&dest->orphan_item_inserted, 1)) { 191376dda93cSYan, Zheng ret = btrfs_insert_orphan_item(trans, 191476dda93cSYan, Zheng root->fs_info->tree_root, 191576dda93cSYan, Zheng dest->root_key.objectid); 191676dda93cSYan, Zheng BUG_ON(ret); 1917d68fc57bSYan, Zheng } 191876dda93cSYan, Zheng 1919531cb13fSSage Weil ret = btrfs_end_transaction(trans, root); 192076dda93cSYan, Zheng BUG_ON(ret); 192176dda93cSYan, Zheng inode->i_flags |= S_DEAD; 192276dda93cSYan, Zheng out_up_write: 192376dda93cSYan, Zheng up_write(&root->fs_info->subvol_sem); 192476dda93cSYan, Zheng out_unlock: 192576dda93cSYan, Zheng mutex_unlock(&inode->i_mutex); 192676dda93cSYan, Zheng if (!err) { 1927efefb143SYan, Zheng shrink_dcache_sb(root->fs_info->sb); 192876dda93cSYan, Zheng btrfs_invalidate_inodes(dest); 192976dda93cSYan, Zheng d_delete(dentry); 193076dda93cSYan, Zheng } 193176dda93cSYan, Zheng out_dput: 193276dda93cSYan, Zheng dput(dentry); 193376dda93cSYan, Zheng out_unlock_dir: 193476dda93cSYan, Zheng mutex_unlock(&dir->i_mutex); 193576dda93cSYan, Zheng mnt_drop_write(file->f_path.mnt); 193676dda93cSYan, Zheng out: 193776dda93cSYan, Zheng kfree(vol_args); 193876dda93cSYan, Zheng return err; 193976dda93cSYan, Zheng } 194076dda93cSYan, Zheng 19411e701a32SChris Mason static int btrfs_ioctl_defrag(struct file *file, void __user *argp) 1942f46b5a66SChristoph Hellwig { 1943f46b5a66SChristoph Hellwig struct inode *inode = fdentry(file)->d_inode; 1944f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 19451e701a32SChris Mason struct btrfs_ioctl_defrag_range_args *range; 1946c146afadSYan Zheng int ret; 1947c146afadSYan Zheng 1948b83cc969SLi Zefan if (btrfs_root_readonly(root)) 1949b83cc969SLi Zefan return -EROFS; 1950b83cc969SLi Zefan 1951c146afadSYan Zheng ret = mnt_want_write(file->f_path.mnt); 1952c146afadSYan Zheng if (ret) 1953c146afadSYan Zheng return ret; 1954f46b5a66SChristoph Hellwig 1955f46b5a66SChristoph Hellwig switch (inode->i_mode & S_IFMT) { 1956f46b5a66SChristoph Hellwig case S_IFDIR: 1957e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) { 1958e441d54dSChris Mason ret = -EPERM; 1959e441d54dSChris Mason goto out; 1960e441d54dSChris Mason } 19618929ecfaSYan, Zheng ret = btrfs_defrag_root(root, 0); 19628929ecfaSYan, Zheng if (ret) 19638929ecfaSYan, Zheng goto out; 19648929ecfaSYan, Zheng ret = btrfs_defrag_root(root->fs_info->extent_root, 0); 1965f46b5a66SChristoph Hellwig break; 1966f46b5a66SChristoph Hellwig case S_IFREG: 1967e441d54dSChris Mason if (!(file->f_mode & FMODE_WRITE)) { 1968e441d54dSChris Mason ret = -EINVAL; 1969e441d54dSChris Mason goto out; 1970e441d54dSChris Mason } 19711e701a32SChris Mason 19721e701a32SChris Mason range = kzalloc(sizeof(*range), GFP_KERNEL); 19731e701a32SChris Mason if (!range) { 19741e701a32SChris Mason ret = -ENOMEM; 19751e701a32SChris Mason goto out; 19761e701a32SChris Mason } 19771e701a32SChris Mason 19781e701a32SChris Mason if (argp) { 19791e701a32SChris Mason if (copy_from_user(range, argp, 19801e701a32SChris Mason sizeof(*range))) { 19811e701a32SChris Mason ret = -EFAULT; 19821e701a32SChris Mason kfree(range); 1983683be16eSDan Carpenter goto out; 19841e701a32SChris Mason } 19851e701a32SChris Mason /* compression requires us to start the IO */ 19861e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 19871e701a32SChris Mason range->flags |= BTRFS_DEFRAG_RANGE_START_IO; 19881e701a32SChris Mason range->extent_thresh = (u32)-1; 19891e701a32SChris Mason } 19901e701a32SChris Mason } else { 19911e701a32SChris Mason /* the rest are all set to zero by kzalloc */ 19921e701a32SChris Mason range->len = (u64)-1; 19931e701a32SChris Mason } 19944cb5300bSChris Mason ret = btrfs_defrag_file(fdentry(file)->d_inode, file, 19954cb5300bSChris Mason range, 0, 0); 19964cb5300bSChris Mason if (ret > 0) 19974cb5300bSChris Mason ret = 0; 19981e701a32SChris Mason kfree(range); 1999f46b5a66SChristoph Hellwig break; 20008929ecfaSYan, Zheng default: 20018929ecfaSYan, Zheng ret = -EINVAL; 2002f46b5a66SChristoph Hellwig } 2003e441d54dSChris Mason out: 2004ab67b7c1SYan Zheng mnt_drop_write(file->f_path.mnt); 2005e441d54dSChris Mason return ret; 2006f46b5a66SChristoph Hellwig } 2007f46b5a66SChristoph Hellwig 2008b2950863SChristoph Hellwig static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg) 2009f46b5a66SChristoph Hellwig { 2010f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2011f46b5a66SChristoph Hellwig int ret; 2012f46b5a66SChristoph Hellwig 2013e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2014e441d54dSChris Mason return -EPERM; 2015e441d54dSChris Mason 2016dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2017dae7b665SLi Zefan if (IS_ERR(vol_args)) 2018dae7b665SLi Zefan return PTR_ERR(vol_args); 2019f46b5a66SChristoph Hellwig 20205516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 2021f46b5a66SChristoph Hellwig ret = btrfs_init_new_device(root, vol_args->name); 2022f46b5a66SChristoph Hellwig 2023f46b5a66SChristoph Hellwig kfree(vol_args); 2024f46b5a66SChristoph Hellwig return ret; 2025f46b5a66SChristoph Hellwig } 2026f46b5a66SChristoph Hellwig 2027b2950863SChristoph Hellwig static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg) 2028f46b5a66SChristoph Hellwig { 2029f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2030f46b5a66SChristoph Hellwig int ret; 2031f46b5a66SChristoph Hellwig 2032e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2033e441d54dSChris Mason return -EPERM; 2034e441d54dSChris Mason 2035c146afadSYan Zheng if (root->fs_info->sb->s_flags & MS_RDONLY) 2036c146afadSYan Zheng return -EROFS; 2037c146afadSYan Zheng 2038dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2039dae7b665SLi Zefan if (IS_ERR(vol_args)) 2040dae7b665SLi Zefan return PTR_ERR(vol_args); 2041f46b5a66SChristoph Hellwig 20425516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 2043f46b5a66SChristoph Hellwig ret = btrfs_rm_device(root, vol_args->name); 2044f46b5a66SChristoph Hellwig 2045f46b5a66SChristoph Hellwig kfree(vol_args); 2046f46b5a66SChristoph Hellwig return ret; 2047f46b5a66SChristoph Hellwig } 2048f46b5a66SChristoph Hellwig 2049475f6387SJan Schmidt static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg) 2050475f6387SJan Schmidt { 2051027ed2f0SLi Zefan struct btrfs_ioctl_fs_info_args *fi_args; 2052475f6387SJan Schmidt struct btrfs_device *device; 2053475f6387SJan Schmidt struct btrfs_device *next; 2054475f6387SJan Schmidt struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; 2055027ed2f0SLi Zefan int ret = 0; 2056475f6387SJan Schmidt 2057475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 2058475f6387SJan Schmidt return -EPERM; 2059475f6387SJan Schmidt 2060027ed2f0SLi Zefan fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL); 2061027ed2f0SLi Zefan if (!fi_args) 2062027ed2f0SLi Zefan return -ENOMEM; 2063027ed2f0SLi Zefan 2064027ed2f0SLi Zefan fi_args->num_devices = fs_devices->num_devices; 2065027ed2f0SLi Zefan memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid)); 2066475f6387SJan Schmidt 2067475f6387SJan Schmidt mutex_lock(&fs_devices->device_list_mutex); 2068475f6387SJan Schmidt list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) { 2069027ed2f0SLi Zefan if (device->devid > fi_args->max_id) 2070027ed2f0SLi Zefan fi_args->max_id = device->devid; 2071475f6387SJan Schmidt } 2072475f6387SJan Schmidt mutex_unlock(&fs_devices->device_list_mutex); 2073475f6387SJan Schmidt 2074027ed2f0SLi Zefan if (copy_to_user(arg, fi_args, sizeof(*fi_args))) 2075027ed2f0SLi Zefan ret = -EFAULT; 2076475f6387SJan Schmidt 2077027ed2f0SLi Zefan kfree(fi_args); 2078027ed2f0SLi Zefan return ret; 2079475f6387SJan Schmidt } 2080475f6387SJan Schmidt 2081475f6387SJan Schmidt static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg) 2082475f6387SJan Schmidt { 2083475f6387SJan Schmidt struct btrfs_ioctl_dev_info_args *di_args; 2084475f6387SJan Schmidt struct btrfs_device *dev; 2085475f6387SJan Schmidt struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; 2086475f6387SJan Schmidt int ret = 0; 2087475f6387SJan Schmidt char *s_uuid = NULL; 2088475f6387SJan Schmidt char empty_uuid[BTRFS_UUID_SIZE] = {0}; 2089475f6387SJan Schmidt 2090475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 2091475f6387SJan Schmidt return -EPERM; 2092475f6387SJan Schmidt 2093475f6387SJan Schmidt di_args = memdup_user(arg, sizeof(*di_args)); 2094475f6387SJan Schmidt if (IS_ERR(di_args)) 2095475f6387SJan Schmidt return PTR_ERR(di_args); 2096475f6387SJan Schmidt 2097475f6387SJan Schmidt if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0) 2098475f6387SJan Schmidt s_uuid = di_args->uuid; 2099475f6387SJan Schmidt 2100475f6387SJan Schmidt mutex_lock(&fs_devices->device_list_mutex); 2101475f6387SJan Schmidt dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL); 2102475f6387SJan Schmidt mutex_unlock(&fs_devices->device_list_mutex); 2103475f6387SJan Schmidt 2104475f6387SJan Schmidt if (!dev) { 2105475f6387SJan Schmidt ret = -ENODEV; 2106475f6387SJan Schmidt goto out; 2107475f6387SJan Schmidt } 2108475f6387SJan Schmidt 2109475f6387SJan Schmidt di_args->devid = dev->devid; 2110475f6387SJan Schmidt di_args->bytes_used = dev->bytes_used; 2111475f6387SJan Schmidt di_args->total_bytes = dev->total_bytes; 2112475f6387SJan Schmidt memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); 2113475f6387SJan Schmidt strncpy(di_args->path, dev->name, sizeof(di_args->path)); 2114475f6387SJan Schmidt 2115475f6387SJan Schmidt out: 2116475f6387SJan Schmidt if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) 2117475f6387SJan Schmidt ret = -EFAULT; 2118475f6387SJan Schmidt 2119475f6387SJan Schmidt kfree(di_args); 2120475f6387SJan Schmidt return ret; 2121475f6387SJan Schmidt } 2122475f6387SJan Schmidt 212376dda93cSYan, Zheng static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, 2124b2950863SChristoph Hellwig u64 off, u64 olen, u64 destoff) 2125f46b5a66SChristoph Hellwig { 2126f46b5a66SChristoph Hellwig struct inode *inode = fdentry(file)->d_inode; 2127f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 2128f46b5a66SChristoph Hellwig struct file *src_file; 2129f46b5a66SChristoph Hellwig struct inode *src; 2130f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 2131f46b5a66SChristoph Hellwig struct btrfs_path *path; 2132f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 2133ae01a0abSYan Zheng char *buf; 2134ae01a0abSYan Zheng struct btrfs_key key; 2135f46b5a66SChristoph Hellwig u32 nritems; 2136f46b5a66SChristoph Hellwig int slot; 2137ae01a0abSYan Zheng int ret; 2138c5c9cd4dSSage Weil u64 len = olen; 2139c5c9cd4dSSage Weil u64 bs = root->fs_info->sb->s_blocksize; 2140c5c9cd4dSSage Weil u64 hint_byte; 2141d20f7043SChris Mason 2142c5c9cd4dSSage Weil /* 2143c5c9cd4dSSage Weil * TODO: 2144c5c9cd4dSSage Weil * - split compressed inline extents. annoying: we need to 2145c5c9cd4dSSage Weil * decompress into destination's address_space (the file offset 2146c5c9cd4dSSage Weil * may change, so source mapping won't do), then recompress (or 2147c5c9cd4dSSage Weil * otherwise reinsert) a subrange. 2148c5c9cd4dSSage Weil * - allow ranges within the same file to be cloned (provided 2149c5c9cd4dSSage Weil * they don't overlap)? 2150c5c9cd4dSSage Weil */ 2151c5c9cd4dSSage Weil 2152e441d54dSChris Mason /* the destination must be opened for writing */ 21532ebc3464SDan Rosenberg if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND)) 2154e441d54dSChris Mason return -EINVAL; 2155e441d54dSChris Mason 2156b83cc969SLi Zefan if (btrfs_root_readonly(root)) 2157b83cc969SLi Zefan return -EROFS; 2158b83cc969SLi Zefan 2159c146afadSYan Zheng ret = mnt_want_write(file->f_path.mnt); 2160c146afadSYan Zheng if (ret) 2161c146afadSYan Zheng return ret; 2162c146afadSYan Zheng 2163c5c9cd4dSSage Weil src_file = fget(srcfd); 2164ab67b7c1SYan Zheng if (!src_file) { 2165ab67b7c1SYan Zheng ret = -EBADF; 2166ab67b7c1SYan Zheng goto out_drop_write; 2167ab67b7c1SYan Zheng } 21685dc64164SDan Rosenberg 2169f46b5a66SChristoph Hellwig src = src_file->f_dentry->d_inode; 2170f46b5a66SChristoph Hellwig 2171c5c9cd4dSSage Weil ret = -EINVAL; 2172c5c9cd4dSSage Weil if (src == inode) 2173c5c9cd4dSSage Weil goto out_fput; 2174c5c9cd4dSSage Weil 21755dc64164SDan Rosenberg /* the src must be open for reading */ 21765dc64164SDan Rosenberg if (!(src_file->f_mode & FMODE_READ)) 21775dc64164SDan Rosenberg goto out_fput; 21785dc64164SDan Rosenberg 2179ae01a0abSYan Zheng ret = -EISDIR; 2180ae01a0abSYan Zheng if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode)) 2181f46b5a66SChristoph Hellwig goto out_fput; 2182f46b5a66SChristoph Hellwig 2183ae01a0abSYan Zheng ret = -EXDEV; 2184ae01a0abSYan Zheng if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root) 2185ae01a0abSYan Zheng goto out_fput; 2186ae01a0abSYan Zheng 2187ae01a0abSYan Zheng ret = -ENOMEM; 2188ae01a0abSYan Zheng buf = vmalloc(btrfs_level_size(root, 0)); 2189ae01a0abSYan Zheng if (!buf) 2190ae01a0abSYan Zheng goto out_fput; 2191ae01a0abSYan Zheng 2192ae01a0abSYan Zheng path = btrfs_alloc_path(); 2193ae01a0abSYan Zheng if (!path) { 2194ae01a0abSYan Zheng vfree(buf); 2195ae01a0abSYan Zheng goto out_fput; 2196ae01a0abSYan Zheng } 2197ae01a0abSYan Zheng path->reada = 2; 2198ae01a0abSYan Zheng 2199f46b5a66SChristoph Hellwig if (inode < src) { 2200fccdae43SSage Weil mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT); 2201fccdae43SSage Weil mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD); 2202f46b5a66SChristoph Hellwig } else { 2203fccdae43SSage Weil mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT); 2204fccdae43SSage Weil mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD); 2205f46b5a66SChristoph Hellwig } 2206f46b5a66SChristoph Hellwig 2207c5c9cd4dSSage Weil /* determine range to clone */ 2208c5c9cd4dSSage Weil ret = -EINVAL; 22092ebc3464SDan Rosenberg if (off + len > src->i_size || off + len < off) 2210f46b5a66SChristoph Hellwig goto out_unlock; 2211c5c9cd4dSSage Weil if (len == 0) 2212c5c9cd4dSSage Weil olen = len = src->i_size - off; 2213c5c9cd4dSSage Weil /* if we extend to eof, continue to block boundary */ 2214c5c9cd4dSSage Weil if (off + len == src->i_size) 22152a6b8daeSLi Zefan len = ALIGN(src->i_size, bs) - off; 2216c5c9cd4dSSage Weil 2217c5c9cd4dSSage Weil /* verify the end result is block aligned */ 22182a6b8daeSLi Zefan if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) || 22192a6b8daeSLi Zefan !IS_ALIGNED(destoff, bs)) 2220c5c9cd4dSSage Weil goto out_unlock; 2221c5c9cd4dSSage Weil 2222f46b5a66SChristoph Hellwig /* do any pending delalloc/csum calc on src, one way or 2223f46b5a66SChristoph Hellwig another, and lock file content */ 2224f46b5a66SChristoph Hellwig while (1) { 222531840ae1SZheng Yan struct btrfs_ordered_extent *ordered; 2226c5c9cd4dSSage Weil lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS); 22279a019196SSage Weil ordered = btrfs_lookup_first_ordered_extent(src, off+len); 22289a019196SSage Weil if (!ordered && 22299a019196SSage Weil !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len, 22309a019196SSage Weil EXTENT_DELALLOC, 0, NULL)) 2231f46b5a66SChristoph Hellwig break; 2232c5c9cd4dSSage Weil unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS); 2233ae01a0abSYan Zheng if (ordered) 2234ae01a0abSYan Zheng btrfs_put_ordered_extent(ordered); 22359a019196SSage Weil btrfs_wait_ordered_range(src, off, len); 2236f46b5a66SChristoph Hellwig } 2237f46b5a66SChristoph Hellwig 2238c5c9cd4dSSage Weil /* clone data */ 223933345d01SLi Zefan key.objectid = btrfs_ino(src); 2240ae01a0abSYan Zheng key.type = BTRFS_EXTENT_DATA_KEY; 2241ae01a0abSYan Zheng key.offset = 0; 2242f46b5a66SChristoph Hellwig 2243f46b5a66SChristoph Hellwig while (1) { 2244f46b5a66SChristoph Hellwig /* 2245f46b5a66SChristoph Hellwig * note the key will change type as we walk through the 2246f46b5a66SChristoph Hellwig * tree. 2247f46b5a66SChristoph Hellwig */ 2248a22285a6SYan, Zheng ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 2249f46b5a66SChristoph Hellwig if (ret < 0) 2250f46b5a66SChristoph Hellwig goto out; 2251f46b5a66SChristoph Hellwig 2252ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 2253ae01a0abSYan Zheng if (path->slots[0] >= nritems) { 2254f46b5a66SChristoph Hellwig ret = btrfs_next_leaf(root, path); 2255f46b5a66SChristoph Hellwig if (ret < 0) 2256f46b5a66SChristoph Hellwig goto out; 2257f46b5a66SChristoph Hellwig if (ret > 0) 2258f46b5a66SChristoph Hellwig break; 2259ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 2260f46b5a66SChristoph Hellwig } 2261f46b5a66SChristoph Hellwig leaf = path->nodes[0]; 2262f46b5a66SChristoph Hellwig slot = path->slots[0]; 2263f46b5a66SChristoph Hellwig 2264ae01a0abSYan Zheng btrfs_item_key_to_cpu(leaf, &key, slot); 2265d20f7043SChris Mason if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY || 226633345d01SLi Zefan key.objectid != btrfs_ino(src)) 2267f46b5a66SChristoph Hellwig break; 2268f46b5a66SChristoph Hellwig 2269c5c9cd4dSSage Weil if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) { 2270c5c9cd4dSSage Weil struct btrfs_file_extent_item *extent; 2271c5c9cd4dSSage Weil int type; 227231840ae1SZheng Yan u32 size; 227331840ae1SZheng Yan struct btrfs_key new_key; 2274c5c9cd4dSSage Weil u64 disko = 0, diskl = 0; 2275c5c9cd4dSSage Weil u64 datao = 0, datal = 0; 2276c5c9cd4dSSage Weil u8 comp; 2277b5384d48SSage Weil u64 endoff; 227831840ae1SZheng Yan 227931840ae1SZheng Yan size = btrfs_item_size_nr(leaf, slot); 228031840ae1SZheng Yan read_extent_buffer(leaf, buf, 228131840ae1SZheng Yan btrfs_item_ptr_offset(leaf, slot), 228231840ae1SZheng Yan size); 2283c5c9cd4dSSage Weil 2284c5c9cd4dSSage Weil extent = btrfs_item_ptr(leaf, slot, 2285c5c9cd4dSSage Weil struct btrfs_file_extent_item); 2286c5c9cd4dSSage Weil comp = btrfs_file_extent_compression(leaf, extent); 2287c5c9cd4dSSage Weil type = btrfs_file_extent_type(leaf, extent); 2288c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 2289c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 2290d397712bSChris Mason disko = btrfs_file_extent_disk_bytenr(leaf, 2291d397712bSChris Mason extent); 2292d397712bSChris Mason diskl = btrfs_file_extent_disk_num_bytes(leaf, 2293d397712bSChris Mason extent); 2294c5c9cd4dSSage Weil datao = btrfs_file_extent_offset(leaf, extent); 2295d397712bSChris Mason datal = btrfs_file_extent_num_bytes(leaf, 2296d397712bSChris Mason extent); 2297c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 2298c5c9cd4dSSage Weil /* take upper bound, may be compressed */ 2299c5c9cd4dSSage Weil datal = btrfs_file_extent_ram_bytes(leaf, 2300c5c9cd4dSSage Weil extent); 2301c5c9cd4dSSage Weil } 2302b3b4aa74SDavid Sterba btrfs_release_path(path); 230331840ae1SZheng Yan 2304050006a7SSage Weil if (key.offset + datal <= off || 2305c5c9cd4dSSage Weil key.offset >= off+len) 2306c5c9cd4dSSage Weil goto next; 2307c5c9cd4dSSage Weil 230831840ae1SZheng Yan memcpy(&new_key, &key, sizeof(new_key)); 230933345d01SLi Zefan new_key.objectid = btrfs_ino(inode); 23104d728ec7SLi Zefan if (off <= key.offset) 2311c5c9cd4dSSage Weil new_key.offset = key.offset + destoff - off; 23124d728ec7SLi Zefan else 23134d728ec7SLi Zefan new_key.offset = destoff; 2314c5c9cd4dSSage Weil 2315a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 1); 2316a22285a6SYan, Zheng if (IS_ERR(trans)) { 2317a22285a6SYan, Zheng ret = PTR_ERR(trans); 2318a22285a6SYan, Zheng goto out; 2319a22285a6SYan, Zheng } 2320a22285a6SYan, Zheng 2321c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 2322c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 2323a22285a6SYan, Zheng if (off > key.offset) { 2324a22285a6SYan, Zheng datao += off - key.offset; 2325a22285a6SYan, Zheng datal -= off - key.offset; 2326a22285a6SYan, Zheng } 2327a22285a6SYan, Zheng 2328a22285a6SYan, Zheng if (key.offset + datal > off + len) 2329a22285a6SYan, Zheng datal = off + len - key.offset; 2330a22285a6SYan, Zheng 2331a22285a6SYan, Zheng ret = btrfs_drop_extents(trans, inode, 2332a22285a6SYan, Zheng new_key.offset, 2333a22285a6SYan, Zheng new_key.offset + datal, 2334a22285a6SYan, Zheng &hint_byte, 1); 2335a22285a6SYan, Zheng BUG_ON(ret); 2336a22285a6SYan, Zheng 233731840ae1SZheng Yan ret = btrfs_insert_empty_item(trans, root, path, 233831840ae1SZheng Yan &new_key, size); 2339a22285a6SYan, Zheng BUG_ON(ret); 234031840ae1SZheng Yan 234131840ae1SZheng Yan leaf = path->nodes[0]; 234231840ae1SZheng Yan slot = path->slots[0]; 234331840ae1SZheng Yan write_extent_buffer(leaf, buf, 234431840ae1SZheng Yan btrfs_item_ptr_offset(leaf, slot), 234531840ae1SZheng Yan size); 2346ae01a0abSYan Zheng 2347f46b5a66SChristoph Hellwig extent = btrfs_item_ptr(leaf, slot, 2348f46b5a66SChristoph Hellwig struct btrfs_file_extent_item); 2349c5c9cd4dSSage Weil 2350c5c9cd4dSSage Weil /* disko == 0 means it's a hole */ 2351c5c9cd4dSSage Weil if (!disko) 2352c5c9cd4dSSage Weil datao = 0; 2353c5c9cd4dSSage Weil 2354c5c9cd4dSSage Weil btrfs_set_file_extent_offset(leaf, extent, 2355c5c9cd4dSSage Weil datao); 2356c5c9cd4dSSage Weil btrfs_set_file_extent_num_bytes(leaf, extent, 2357c5c9cd4dSSage Weil datal); 2358c5c9cd4dSSage Weil if (disko) { 2359c5c9cd4dSSage Weil inode_add_bytes(inode, datal); 2360ae01a0abSYan Zheng ret = btrfs_inc_extent_ref(trans, root, 23615d4f98a2SYan Zheng disko, diskl, 0, 2362f46b5a66SChristoph Hellwig root->root_key.objectid, 236333345d01SLi Zefan btrfs_ino(inode), 23645d4f98a2SYan Zheng new_key.offset - datao); 2365ae01a0abSYan Zheng BUG_ON(ret); 2366f46b5a66SChristoph Hellwig } 2367c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 2368c5c9cd4dSSage Weil u64 skip = 0; 2369c5c9cd4dSSage Weil u64 trim = 0; 2370c5c9cd4dSSage Weil if (off > key.offset) { 2371c5c9cd4dSSage Weil skip = off - key.offset; 2372c5c9cd4dSSage Weil new_key.offset += skip; 237331840ae1SZheng Yan } 2374d397712bSChris Mason 2375c5c9cd4dSSage Weil if (key.offset + datal > off+len) 2376c5c9cd4dSSage Weil trim = key.offset + datal - (off+len); 2377d397712bSChris Mason 2378c5c9cd4dSSage Weil if (comp && (skip || trim)) { 2379c5c9cd4dSSage Weil ret = -EINVAL; 2380a22285a6SYan, Zheng btrfs_end_transaction(trans, root); 2381c5c9cd4dSSage Weil goto out; 238231840ae1SZheng Yan } 2383c5c9cd4dSSage Weil size -= skip + trim; 2384c5c9cd4dSSage Weil datal -= skip + trim; 2385a22285a6SYan, Zheng 2386a22285a6SYan, Zheng ret = btrfs_drop_extents(trans, inode, 2387a22285a6SYan, Zheng new_key.offset, 2388a22285a6SYan, Zheng new_key.offset + datal, 2389a22285a6SYan, Zheng &hint_byte, 1); 2390a22285a6SYan, Zheng BUG_ON(ret); 2391a22285a6SYan, Zheng 2392c5c9cd4dSSage Weil ret = btrfs_insert_empty_item(trans, root, path, 2393c5c9cd4dSSage Weil &new_key, size); 2394a22285a6SYan, Zheng BUG_ON(ret); 2395c5c9cd4dSSage Weil 2396c5c9cd4dSSage Weil if (skip) { 2397d397712bSChris Mason u32 start = 2398d397712bSChris Mason btrfs_file_extent_calc_inline_size(0); 2399c5c9cd4dSSage Weil memmove(buf+start, buf+start+skip, 2400c5c9cd4dSSage Weil datal); 2401c5c9cd4dSSage Weil } 2402c5c9cd4dSSage Weil 2403c5c9cd4dSSage Weil leaf = path->nodes[0]; 2404c5c9cd4dSSage Weil slot = path->slots[0]; 2405c5c9cd4dSSage Weil write_extent_buffer(leaf, buf, 2406c5c9cd4dSSage Weil btrfs_item_ptr_offset(leaf, slot), 2407c5c9cd4dSSage Weil size); 2408c5c9cd4dSSage Weil inode_add_bytes(inode, datal); 2409c5c9cd4dSSage Weil } 2410c5c9cd4dSSage Weil 2411c5c9cd4dSSage Weil btrfs_mark_buffer_dirty(leaf); 2412b3b4aa74SDavid Sterba btrfs_release_path(path); 2413c5c9cd4dSSage Weil 2414a22285a6SYan, Zheng inode->i_mtime = inode->i_ctime = CURRENT_TIME; 2415b5384d48SSage Weil 2416b5384d48SSage Weil /* 2417b5384d48SSage Weil * we round up to the block size at eof when 2418b5384d48SSage Weil * determining which extents to clone above, 2419b5384d48SSage Weil * but shouldn't round up the file size 2420b5384d48SSage Weil */ 2421b5384d48SSage Weil endoff = new_key.offset + datal; 24225f3888ffSLi Zefan if (endoff > destoff+olen) 24235f3888ffSLi Zefan endoff = destoff+olen; 2424b5384d48SSage Weil if (endoff > inode->i_size) 2425b5384d48SSage Weil btrfs_i_size_write(inode, endoff); 2426b5384d48SSage Weil 2427a22285a6SYan, Zheng BTRFS_I(inode)->flags = BTRFS_I(src)->flags; 2428a22285a6SYan, Zheng ret = btrfs_update_inode(trans, root, inode); 2429a22285a6SYan, Zheng BUG_ON(ret); 2430a22285a6SYan, Zheng btrfs_end_transaction(trans, root); 2431a22285a6SYan, Zheng } 2432c5c9cd4dSSage Weil next: 2433b3b4aa74SDavid Sterba btrfs_release_path(path); 2434ae01a0abSYan Zheng key.offset++; 2435ae01a0abSYan Zheng } 2436f46b5a66SChristoph Hellwig ret = 0; 2437f46b5a66SChristoph Hellwig out: 2438b3b4aa74SDavid Sterba btrfs_release_path(path); 2439c5c9cd4dSSage Weil unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS); 2440f46b5a66SChristoph Hellwig out_unlock: 2441f46b5a66SChristoph Hellwig mutex_unlock(&src->i_mutex); 2442f46b5a66SChristoph Hellwig mutex_unlock(&inode->i_mutex); 2443ae01a0abSYan Zheng vfree(buf); 2444ae01a0abSYan Zheng btrfs_free_path(path); 2445f46b5a66SChristoph Hellwig out_fput: 2446f46b5a66SChristoph Hellwig fput(src_file); 2447ab67b7c1SYan Zheng out_drop_write: 2448ab67b7c1SYan Zheng mnt_drop_write(file->f_path.mnt); 2449f46b5a66SChristoph Hellwig return ret; 2450f46b5a66SChristoph Hellwig } 2451f46b5a66SChristoph Hellwig 24527a865e8aSChristoph Hellwig static long btrfs_ioctl_clone_range(struct file *file, void __user *argp) 2453c5c9cd4dSSage Weil { 2454c5c9cd4dSSage Weil struct btrfs_ioctl_clone_range_args args; 2455c5c9cd4dSSage Weil 24567a865e8aSChristoph Hellwig if (copy_from_user(&args, argp, sizeof(args))) 2457c5c9cd4dSSage Weil return -EFAULT; 2458c5c9cd4dSSage Weil return btrfs_ioctl_clone(file, args.src_fd, args.src_offset, 2459c5c9cd4dSSage Weil args.src_length, args.dest_offset); 2460c5c9cd4dSSage Weil } 2461c5c9cd4dSSage Weil 2462f46b5a66SChristoph Hellwig /* 2463f46b5a66SChristoph Hellwig * there are many ways the trans_start and trans_end ioctls can lead 2464f46b5a66SChristoph Hellwig * to deadlocks. They should only be used by applications that 2465f46b5a66SChristoph Hellwig * basically own the machine, and have a very in depth understanding 2466f46b5a66SChristoph Hellwig * of all the possible deadlocks and enospc problems. 2467f46b5a66SChristoph Hellwig */ 2468b2950863SChristoph Hellwig static long btrfs_ioctl_trans_start(struct file *file) 2469f46b5a66SChristoph Hellwig { 2470f46b5a66SChristoph Hellwig struct inode *inode = fdentry(file)->d_inode; 2471f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 2472f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 24731ab86aedSSage Weil int ret; 2474f46b5a66SChristoph Hellwig 24751ab86aedSSage Weil ret = -EPERM; 2476df5b5520SChristoph Hellwig if (!capable(CAP_SYS_ADMIN)) 2477f46b5a66SChristoph Hellwig goto out; 24781ab86aedSSage Weil 24791ab86aedSSage Weil ret = -EINPROGRESS; 24801ab86aedSSage Weil if (file->private_data) 24811ab86aedSSage Weil goto out; 24829ca9ee09SSage Weil 2483b83cc969SLi Zefan ret = -EROFS; 2484b83cc969SLi Zefan if (btrfs_root_readonly(root)) 2485b83cc969SLi Zefan goto out; 2486b83cc969SLi Zefan 2487c146afadSYan Zheng ret = mnt_want_write(file->f_path.mnt); 2488c146afadSYan Zheng if (ret) 2489c146afadSYan Zheng goto out; 2490c146afadSYan Zheng 2491a4abeea4SJosef Bacik atomic_inc(&root->fs_info->open_ioctl_trans); 24929ca9ee09SSage Weil 2493f46b5a66SChristoph Hellwig ret = -ENOMEM; 24947a7eaa40SJosef Bacik trans = btrfs_start_ioctl_transaction(root); 2495abd30bb0STsutomu Itoh if (IS_ERR(trans)) 24961ab86aedSSage Weil goto out_drop; 24971ab86aedSSage Weil 24981ab86aedSSage Weil file->private_data = trans; 24991ab86aedSSage Weil return 0; 25001ab86aedSSage Weil 25011ab86aedSSage Weil out_drop: 2502a4abeea4SJosef Bacik atomic_dec(&root->fs_info->open_ioctl_trans); 25031ab86aedSSage Weil mnt_drop_write(file->f_path.mnt); 2504f46b5a66SChristoph Hellwig out: 2505f46b5a66SChristoph Hellwig return ret; 2506f46b5a66SChristoph Hellwig } 2507f46b5a66SChristoph Hellwig 25086ef5ed0dSJosef Bacik static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) 25096ef5ed0dSJosef Bacik { 25106ef5ed0dSJosef Bacik struct inode *inode = fdentry(file)->d_inode; 25116ef5ed0dSJosef Bacik struct btrfs_root *root = BTRFS_I(inode)->root; 25126ef5ed0dSJosef Bacik struct btrfs_root *new_root; 25136ef5ed0dSJosef Bacik struct btrfs_dir_item *di; 25146ef5ed0dSJosef Bacik struct btrfs_trans_handle *trans; 25156ef5ed0dSJosef Bacik struct btrfs_path *path; 25166ef5ed0dSJosef Bacik struct btrfs_key location; 25176ef5ed0dSJosef Bacik struct btrfs_disk_key disk_key; 25186ef5ed0dSJosef Bacik struct btrfs_super_block *disk_super; 25196ef5ed0dSJosef Bacik u64 features; 25206ef5ed0dSJosef Bacik u64 objectid = 0; 25216ef5ed0dSJosef Bacik u64 dir_id; 25226ef5ed0dSJosef Bacik 25236ef5ed0dSJosef Bacik if (!capable(CAP_SYS_ADMIN)) 25246ef5ed0dSJosef Bacik return -EPERM; 25256ef5ed0dSJosef Bacik 25266ef5ed0dSJosef Bacik if (copy_from_user(&objectid, argp, sizeof(objectid))) 25276ef5ed0dSJosef Bacik return -EFAULT; 25286ef5ed0dSJosef Bacik 25296ef5ed0dSJosef Bacik if (!objectid) 25306ef5ed0dSJosef Bacik objectid = root->root_key.objectid; 25316ef5ed0dSJosef Bacik 25326ef5ed0dSJosef Bacik location.objectid = objectid; 25336ef5ed0dSJosef Bacik location.type = BTRFS_ROOT_ITEM_KEY; 25346ef5ed0dSJosef Bacik location.offset = (u64)-1; 25356ef5ed0dSJosef Bacik 25366ef5ed0dSJosef Bacik new_root = btrfs_read_fs_root_no_name(root->fs_info, &location); 25376ef5ed0dSJosef Bacik if (IS_ERR(new_root)) 25386ef5ed0dSJosef Bacik return PTR_ERR(new_root); 25396ef5ed0dSJosef Bacik 25406ef5ed0dSJosef Bacik if (btrfs_root_refs(&new_root->root_item) == 0) 25416ef5ed0dSJosef Bacik return -ENOENT; 25426ef5ed0dSJosef Bacik 25436ef5ed0dSJosef Bacik path = btrfs_alloc_path(); 25446ef5ed0dSJosef Bacik if (!path) 25456ef5ed0dSJosef Bacik return -ENOMEM; 25466ef5ed0dSJosef Bacik path->leave_spinning = 1; 25476ef5ed0dSJosef Bacik 25486ef5ed0dSJosef Bacik trans = btrfs_start_transaction(root, 1); 254998d5dc13STsutomu Itoh if (IS_ERR(trans)) { 25506ef5ed0dSJosef Bacik btrfs_free_path(path); 255198d5dc13STsutomu Itoh return PTR_ERR(trans); 25526ef5ed0dSJosef Bacik } 25536ef5ed0dSJosef Bacik 25546ef5ed0dSJosef Bacik dir_id = btrfs_super_root_dir(&root->fs_info->super_copy); 25556ef5ed0dSJosef Bacik di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path, 25566ef5ed0dSJosef Bacik dir_id, "default", 7, 1); 2557cf1e99a4SDan Carpenter if (IS_ERR_OR_NULL(di)) { 25586ef5ed0dSJosef Bacik btrfs_free_path(path); 25596ef5ed0dSJosef Bacik btrfs_end_transaction(trans, root); 25606ef5ed0dSJosef Bacik printk(KERN_ERR "Umm, you don't have the default dir item, " 25616ef5ed0dSJosef Bacik "this isn't going to work\n"); 25626ef5ed0dSJosef Bacik return -ENOENT; 25636ef5ed0dSJosef Bacik } 25646ef5ed0dSJosef Bacik 25656ef5ed0dSJosef Bacik btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); 25666ef5ed0dSJosef Bacik btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); 25676ef5ed0dSJosef Bacik btrfs_mark_buffer_dirty(path->nodes[0]); 25686ef5ed0dSJosef Bacik btrfs_free_path(path); 25696ef5ed0dSJosef Bacik 25706ef5ed0dSJosef Bacik disk_super = &root->fs_info->super_copy; 25716ef5ed0dSJosef Bacik features = btrfs_super_incompat_flags(disk_super); 25726ef5ed0dSJosef Bacik if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) { 25736ef5ed0dSJosef Bacik features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL; 25746ef5ed0dSJosef Bacik btrfs_set_super_incompat_flags(disk_super, features); 25756ef5ed0dSJosef Bacik } 25766ef5ed0dSJosef Bacik btrfs_end_transaction(trans, root); 25776ef5ed0dSJosef Bacik 25786ef5ed0dSJosef Bacik return 0; 25796ef5ed0dSJosef Bacik } 25806ef5ed0dSJosef Bacik 2581bf5fc093SJosef Bacik static void get_block_group_info(struct list_head *groups_list, 2582bf5fc093SJosef Bacik struct btrfs_ioctl_space_info *space) 2583bf5fc093SJosef Bacik { 2584bf5fc093SJosef Bacik struct btrfs_block_group_cache *block_group; 2585bf5fc093SJosef Bacik 2586bf5fc093SJosef Bacik space->total_bytes = 0; 2587bf5fc093SJosef Bacik space->used_bytes = 0; 2588bf5fc093SJosef Bacik space->flags = 0; 2589bf5fc093SJosef Bacik list_for_each_entry(block_group, groups_list, list) { 2590bf5fc093SJosef Bacik space->flags = block_group->flags; 2591bf5fc093SJosef Bacik space->total_bytes += block_group->key.offset; 2592bf5fc093SJosef Bacik space->used_bytes += 2593bf5fc093SJosef Bacik btrfs_block_group_used(&block_group->item); 2594bf5fc093SJosef Bacik } 2595bf5fc093SJosef Bacik } 2596bf5fc093SJosef Bacik 25971406e432SJosef Bacik long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg) 25981406e432SJosef Bacik { 25991406e432SJosef Bacik struct btrfs_ioctl_space_args space_args; 26001406e432SJosef Bacik struct btrfs_ioctl_space_info space; 26011406e432SJosef Bacik struct btrfs_ioctl_space_info *dest; 26027fde62bfSChris Mason struct btrfs_ioctl_space_info *dest_orig; 260313f2696fSDaniel J Blueman struct btrfs_ioctl_space_info __user *user_dest; 26041406e432SJosef Bacik struct btrfs_space_info *info; 2605bf5fc093SJosef Bacik u64 types[] = {BTRFS_BLOCK_GROUP_DATA, 2606bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_SYSTEM, 2607bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_METADATA, 2608bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA}; 2609bf5fc093SJosef Bacik int num_types = 4; 26107fde62bfSChris Mason int alloc_size; 26111406e432SJosef Bacik int ret = 0; 261251788b1bSDan Rosenberg u64 slot_count = 0; 2613bf5fc093SJosef Bacik int i, c; 26141406e432SJosef Bacik 26151406e432SJosef Bacik if (copy_from_user(&space_args, 26161406e432SJosef Bacik (struct btrfs_ioctl_space_args __user *)arg, 26171406e432SJosef Bacik sizeof(space_args))) 26181406e432SJosef Bacik return -EFAULT; 26191406e432SJosef Bacik 2620bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 2621bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 2622bf5fc093SJosef Bacik 2623bf5fc093SJosef Bacik info = NULL; 26247fde62bfSChris Mason rcu_read_lock(); 2625bf5fc093SJosef Bacik list_for_each_entry_rcu(tmp, &root->fs_info->space_info, 2626bf5fc093SJosef Bacik list) { 2627bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 2628bf5fc093SJosef Bacik info = tmp; 2629bf5fc093SJosef Bacik break; 2630bf5fc093SJosef Bacik } 2631bf5fc093SJosef Bacik } 26327fde62bfSChris Mason rcu_read_unlock(); 26331406e432SJosef Bacik 2634bf5fc093SJosef Bacik if (!info) 2635bf5fc093SJosef Bacik continue; 2636bf5fc093SJosef Bacik 2637bf5fc093SJosef Bacik down_read(&info->groups_sem); 2638bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 2639bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) 2640bf5fc093SJosef Bacik slot_count++; 2641bf5fc093SJosef Bacik } 2642bf5fc093SJosef Bacik up_read(&info->groups_sem); 2643bf5fc093SJosef Bacik } 2644bf5fc093SJosef Bacik 26457fde62bfSChris Mason /* space_slots == 0 means they are asking for a count */ 26467fde62bfSChris Mason if (space_args.space_slots == 0) { 26477fde62bfSChris Mason space_args.total_spaces = slot_count; 26487fde62bfSChris Mason goto out; 26497fde62bfSChris Mason } 2650bf5fc093SJosef Bacik 265151788b1bSDan Rosenberg slot_count = min_t(u64, space_args.space_slots, slot_count); 2652bf5fc093SJosef Bacik 26537fde62bfSChris Mason alloc_size = sizeof(*dest) * slot_count; 2654bf5fc093SJosef Bacik 26557fde62bfSChris Mason /* we generally have at most 6 or so space infos, one for each raid 26567fde62bfSChris Mason * level. So, a whole page should be more than enough for everyone 26577fde62bfSChris Mason */ 26587fde62bfSChris Mason if (alloc_size > PAGE_CACHE_SIZE) 26597fde62bfSChris Mason return -ENOMEM; 26607fde62bfSChris Mason 26617fde62bfSChris Mason space_args.total_spaces = 0; 26627fde62bfSChris Mason dest = kmalloc(alloc_size, GFP_NOFS); 26637fde62bfSChris Mason if (!dest) 26647fde62bfSChris Mason return -ENOMEM; 26657fde62bfSChris Mason dest_orig = dest; 26667fde62bfSChris Mason 26677fde62bfSChris Mason /* now we have a buffer to copy into */ 2668bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 2669bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 2670bf5fc093SJosef Bacik 267151788b1bSDan Rosenberg if (!slot_count) 267251788b1bSDan Rosenberg break; 267351788b1bSDan Rosenberg 2674bf5fc093SJosef Bacik info = NULL; 26751406e432SJosef Bacik rcu_read_lock(); 2676bf5fc093SJosef Bacik list_for_each_entry_rcu(tmp, &root->fs_info->space_info, 2677bf5fc093SJosef Bacik list) { 2678bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 2679bf5fc093SJosef Bacik info = tmp; 26807fde62bfSChris Mason break; 2681bf5fc093SJosef Bacik } 2682bf5fc093SJosef Bacik } 2683bf5fc093SJosef Bacik rcu_read_unlock(); 26847fde62bfSChris Mason 2685bf5fc093SJosef Bacik if (!info) 2686bf5fc093SJosef Bacik continue; 2687bf5fc093SJosef Bacik down_read(&info->groups_sem); 2688bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 2689bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) { 2690bf5fc093SJosef Bacik get_block_group_info(&info->block_groups[c], 2691bf5fc093SJosef Bacik &space); 26927fde62bfSChris Mason memcpy(dest, &space, sizeof(space)); 26931406e432SJosef Bacik dest++; 26941406e432SJosef Bacik space_args.total_spaces++; 269551788b1bSDan Rosenberg slot_count--; 26961406e432SJosef Bacik } 269751788b1bSDan Rosenberg if (!slot_count) 269851788b1bSDan Rosenberg break; 2699bf5fc093SJosef Bacik } 2700bf5fc093SJosef Bacik up_read(&info->groups_sem); 2701bf5fc093SJosef Bacik } 27021406e432SJosef Bacik 27037fde62bfSChris Mason user_dest = (struct btrfs_ioctl_space_info *) 27047fde62bfSChris Mason (arg + sizeof(struct btrfs_ioctl_space_args)); 27057fde62bfSChris Mason 27067fde62bfSChris Mason if (copy_to_user(user_dest, dest_orig, alloc_size)) 27077fde62bfSChris Mason ret = -EFAULT; 27087fde62bfSChris Mason 27097fde62bfSChris Mason kfree(dest_orig); 27107fde62bfSChris Mason out: 27117fde62bfSChris Mason if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) 27121406e432SJosef Bacik ret = -EFAULT; 27131406e432SJosef Bacik 27141406e432SJosef Bacik return ret; 27151406e432SJosef Bacik } 27161406e432SJosef Bacik 2717f46b5a66SChristoph Hellwig /* 2718f46b5a66SChristoph Hellwig * there are many ways the trans_start and trans_end ioctls can lead 2719f46b5a66SChristoph Hellwig * to deadlocks. They should only be used by applications that 2720f46b5a66SChristoph Hellwig * basically own the machine, and have a very in depth understanding 2721f46b5a66SChristoph Hellwig * of all the possible deadlocks and enospc problems. 2722f46b5a66SChristoph Hellwig */ 2723f46b5a66SChristoph Hellwig long btrfs_ioctl_trans_end(struct file *file) 2724f46b5a66SChristoph Hellwig { 2725f46b5a66SChristoph Hellwig struct inode *inode = fdentry(file)->d_inode; 2726f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 2727f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 2728f46b5a66SChristoph Hellwig 2729f46b5a66SChristoph Hellwig trans = file->private_data; 27301ab86aedSSage Weil if (!trans) 27311ab86aedSSage Weil return -EINVAL; 2732b214107eSChristoph Hellwig file->private_data = NULL; 27339ca9ee09SSage Weil 27341ab86aedSSage Weil btrfs_end_transaction(trans, root); 27351ab86aedSSage Weil 2736a4abeea4SJosef Bacik atomic_dec(&root->fs_info->open_ioctl_trans); 27379ca9ee09SSage Weil 2738cfc8ea87SSage Weil mnt_drop_write(file->f_path.mnt); 27391ab86aedSSage Weil return 0; 2740f46b5a66SChristoph Hellwig } 2741f46b5a66SChristoph Hellwig 274246204592SSage Weil static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp) 274346204592SSage Weil { 274446204592SSage Weil struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root; 274546204592SSage Weil struct btrfs_trans_handle *trans; 274646204592SSage Weil u64 transid; 2747db5b493aSTsutomu Itoh int ret; 274846204592SSage Weil 274946204592SSage Weil trans = btrfs_start_transaction(root, 0); 275098d5dc13STsutomu Itoh if (IS_ERR(trans)) 275198d5dc13STsutomu Itoh return PTR_ERR(trans); 275246204592SSage Weil transid = trans->transid; 2753db5b493aSTsutomu Itoh ret = btrfs_commit_transaction_async(trans, root, 0); 27548b2b2d3cSTsutomu Itoh if (ret) { 27558b2b2d3cSTsutomu Itoh btrfs_end_transaction(trans, root); 2756db5b493aSTsutomu Itoh return ret; 27578b2b2d3cSTsutomu Itoh } 275846204592SSage Weil 275946204592SSage Weil if (argp) 276046204592SSage Weil if (copy_to_user(argp, &transid, sizeof(transid))) 276146204592SSage Weil return -EFAULT; 276246204592SSage Weil return 0; 276346204592SSage Weil } 276446204592SSage Weil 276546204592SSage Weil static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp) 276646204592SSage Weil { 276746204592SSage Weil struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root; 276846204592SSage Weil u64 transid; 276946204592SSage Weil 277046204592SSage Weil if (argp) { 277146204592SSage Weil if (copy_from_user(&transid, argp, sizeof(transid))) 277246204592SSage Weil return -EFAULT; 277346204592SSage Weil } else { 277446204592SSage Weil transid = 0; /* current trans */ 277546204592SSage Weil } 277646204592SSage Weil return btrfs_wait_for_commit(root, transid); 277746204592SSage Weil } 277846204592SSage Weil 2779475f6387SJan Schmidt static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg) 2780475f6387SJan Schmidt { 2781475f6387SJan Schmidt int ret; 2782475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 2783475f6387SJan Schmidt 2784475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 2785475f6387SJan Schmidt return -EPERM; 2786475f6387SJan Schmidt 2787475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 2788475f6387SJan Schmidt if (IS_ERR(sa)) 2789475f6387SJan Schmidt return PTR_ERR(sa); 2790475f6387SJan Schmidt 2791475f6387SJan Schmidt ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end, 27928628764eSArne Jansen &sa->progress, sa->flags & BTRFS_SCRUB_READONLY); 2793475f6387SJan Schmidt 2794475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 2795475f6387SJan Schmidt ret = -EFAULT; 2796475f6387SJan Schmidt 2797475f6387SJan Schmidt kfree(sa); 2798475f6387SJan Schmidt return ret; 2799475f6387SJan Schmidt } 2800475f6387SJan Schmidt 2801475f6387SJan Schmidt static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg) 2802475f6387SJan Schmidt { 2803475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 2804475f6387SJan Schmidt return -EPERM; 2805475f6387SJan Schmidt 2806475f6387SJan Schmidt return btrfs_scrub_cancel(root); 2807475f6387SJan Schmidt } 2808475f6387SJan Schmidt 2809475f6387SJan Schmidt static long btrfs_ioctl_scrub_progress(struct btrfs_root *root, 2810475f6387SJan Schmidt void __user *arg) 2811475f6387SJan Schmidt { 2812475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 2813475f6387SJan Schmidt int ret; 2814475f6387SJan Schmidt 2815475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 2816475f6387SJan Schmidt return -EPERM; 2817475f6387SJan Schmidt 2818475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 2819475f6387SJan Schmidt if (IS_ERR(sa)) 2820475f6387SJan Schmidt return PTR_ERR(sa); 2821475f6387SJan Schmidt 2822475f6387SJan Schmidt ret = btrfs_scrub_progress(root, sa->devid, &sa->progress); 2823475f6387SJan Schmidt 2824475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 2825475f6387SJan Schmidt ret = -EFAULT; 2826475f6387SJan Schmidt 2827475f6387SJan Schmidt kfree(sa); 2828475f6387SJan Schmidt return ret; 2829475f6387SJan Schmidt } 2830475f6387SJan Schmidt 2831f46b5a66SChristoph Hellwig long btrfs_ioctl(struct file *file, unsigned int 2832f46b5a66SChristoph Hellwig cmd, unsigned long arg) 2833f46b5a66SChristoph Hellwig { 2834f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; 28354bcabaa3SChristoph Hellwig void __user *argp = (void __user *)arg; 2836f46b5a66SChristoph Hellwig 2837f46b5a66SChristoph Hellwig switch (cmd) { 28386cbff00fSChristoph Hellwig case FS_IOC_GETFLAGS: 28396cbff00fSChristoph Hellwig return btrfs_ioctl_getflags(file, argp); 28406cbff00fSChristoph Hellwig case FS_IOC_SETFLAGS: 28416cbff00fSChristoph Hellwig return btrfs_ioctl_setflags(file, argp); 28426cbff00fSChristoph Hellwig case FS_IOC_GETVERSION: 28436cbff00fSChristoph Hellwig return btrfs_ioctl_getversion(file, argp); 2844f7039b1dSLi Dongyang case FITRIM: 2845f7039b1dSLi Dongyang return btrfs_ioctl_fitrim(file, argp); 2846f46b5a66SChristoph Hellwig case BTRFS_IOC_SNAP_CREATE: 2847fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 0); 2848fdfb1e4fSLi Zefan case BTRFS_IOC_SNAP_CREATE_V2: 2849fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create_v2(file, argp, 0); 28503de4586cSChris Mason case BTRFS_IOC_SUBVOL_CREATE: 2851fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 1); 285276dda93cSYan, Zheng case BTRFS_IOC_SNAP_DESTROY: 285376dda93cSYan, Zheng return btrfs_ioctl_snap_destroy(file, argp); 28540caa102dSLi Zefan case BTRFS_IOC_SUBVOL_GETFLAGS: 28550caa102dSLi Zefan return btrfs_ioctl_subvol_getflags(file, argp); 28560caa102dSLi Zefan case BTRFS_IOC_SUBVOL_SETFLAGS: 28570caa102dSLi Zefan return btrfs_ioctl_subvol_setflags(file, argp); 28586ef5ed0dSJosef Bacik case BTRFS_IOC_DEFAULT_SUBVOL: 28596ef5ed0dSJosef Bacik return btrfs_ioctl_default_subvol(file, argp); 2860f46b5a66SChristoph Hellwig case BTRFS_IOC_DEFRAG: 28611e701a32SChris Mason return btrfs_ioctl_defrag(file, NULL); 28621e701a32SChris Mason case BTRFS_IOC_DEFRAG_RANGE: 28631e701a32SChris Mason return btrfs_ioctl_defrag(file, argp); 2864f46b5a66SChristoph Hellwig case BTRFS_IOC_RESIZE: 28654bcabaa3SChristoph Hellwig return btrfs_ioctl_resize(root, argp); 2866f46b5a66SChristoph Hellwig case BTRFS_IOC_ADD_DEV: 28674bcabaa3SChristoph Hellwig return btrfs_ioctl_add_dev(root, argp); 2868f46b5a66SChristoph Hellwig case BTRFS_IOC_RM_DEV: 28694bcabaa3SChristoph Hellwig return btrfs_ioctl_rm_dev(root, argp); 2870475f6387SJan Schmidt case BTRFS_IOC_FS_INFO: 2871475f6387SJan Schmidt return btrfs_ioctl_fs_info(root, argp); 2872475f6387SJan Schmidt case BTRFS_IOC_DEV_INFO: 2873475f6387SJan Schmidt return btrfs_ioctl_dev_info(root, argp); 2874f46b5a66SChristoph Hellwig case BTRFS_IOC_BALANCE: 2875f46b5a66SChristoph Hellwig return btrfs_balance(root->fs_info->dev_root); 2876f46b5a66SChristoph Hellwig case BTRFS_IOC_CLONE: 2877c5c9cd4dSSage Weil return btrfs_ioctl_clone(file, arg, 0, 0, 0); 2878c5c9cd4dSSage Weil case BTRFS_IOC_CLONE_RANGE: 28797a865e8aSChristoph Hellwig return btrfs_ioctl_clone_range(file, argp); 2880f46b5a66SChristoph Hellwig case BTRFS_IOC_TRANS_START: 2881f46b5a66SChristoph Hellwig return btrfs_ioctl_trans_start(file); 2882f46b5a66SChristoph Hellwig case BTRFS_IOC_TRANS_END: 2883f46b5a66SChristoph Hellwig return btrfs_ioctl_trans_end(file); 2884ac8e9819SChris Mason case BTRFS_IOC_TREE_SEARCH: 2885ac8e9819SChris Mason return btrfs_ioctl_tree_search(file, argp); 2886ac8e9819SChris Mason case BTRFS_IOC_INO_LOOKUP: 2887ac8e9819SChris Mason return btrfs_ioctl_ino_lookup(file, argp); 28881406e432SJosef Bacik case BTRFS_IOC_SPACE_INFO: 28891406e432SJosef Bacik return btrfs_ioctl_space_info(root, argp); 2890f46b5a66SChristoph Hellwig case BTRFS_IOC_SYNC: 2891f46b5a66SChristoph Hellwig btrfs_sync_fs(file->f_dentry->d_sb, 1); 2892f46b5a66SChristoph Hellwig return 0; 289346204592SSage Weil case BTRFS_IOC_START_SYNC: 289446204592SSage Weil return btrfs_ioctl_start_sync(file, argp); 289546204592SSage Weil case BTRFS_IOC_WAIT_SYNC: 289646204592SSage Weil return btrfs_ioctl_wait_sync(file, argp); 2897475f6387SJan Schmidt case BTRFS_IOC_SCRUB: 2898475f6387SJan Schmidt return btrfs_ioctl_scrub(root, argp); 2899475f6387SJan Schmidt case BTRFS_IOC_SCRUB_CANCEL: 2900475f6387SJan Schmidt return btrfs_ioctl_scrub_cancel(root, argp); 2901475f6387SJan Schmidt case BTRFS_IOC_SCRUB_PROGRESS: 2902475f6387SJan Schmidt return btrfs_ioctl_scrub_progress(root, argp); 2903f46b5a66SChristoph Hellwig } 2904f46b5a66SChristoph Hellwig 2905f46b5a66SChristoph Hellwig return -ENOTTY; 2906f46b5a66SChristoph Hellwig } 2907