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; 3266a912213SJosef Bacik struct dentry *parent = dget_parent(dentry); 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); 3356a912213SJosef Bacik if (ret) { 3366a912213SJosef Bacik dput(parent); 337a22285a6SYan, Zheng return ret; 3386a912213SJosef Bacik } 3396a912213SJosef Bacik 3406a912213SJosef Bacik dir = parent->d_inode; 3416a912213SJosef Bacik 3429ed74f2dSJosef Bacik /* 3439ed74f2dSJosef Bacik * 1 - inode item 3449ed74f2dSJosef Bacik * 2 - refs 3459ed74f2dSJosef Bacik * 1 - root item 3469ed74f2dSJosef Bacik * 2 - dir items 3479ed74f2dSJosef Bacik */ 348a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 6); 3496a912213SJosef Bacik if (IS_ERR(trans)) { 3506a912213SJosef Bacik dput(parent); 351a22285a6SYan, Zheng return PTR_ERR(trans); 3526a912213SJosef Bacik } 353f46b5a66SChristoph Hellwig 3545d4f98a2SYan Zheng leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 3555d4f98a2SYan Zheng 0, objectid, NULL, 0, 0, 0); 3568e8a1e31SJosef Bacik if (IS_ERR(leaf)) { 3578e8a1e31SJosef Bacik ret = PTR_ERR(leaf); 3588e8a1e31SJosef Bacik goto fail; 3598e8a1e31SJosef Bacik } 360f46b5a66SChristoph Hellwig 3615d4f98a2SYan Zheng memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); 362f46b5a66SChristoph Hellwig btrfs_set_header_bytenr(leaf, leaf->start); 363f46b5a66SChristoph Hellwig btrfs_set_header_generation(leaf, trans->transid); 3645d4f98a2SYan Zheng btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 365f46b5a66SChristoph Hellwig btrfs_set_header_owner(leaf, objectid); 366f46b5a66SChristoph Hellwig 367f46b5a66SChristoph Hellwig write_extent_buffer(leaf, root->fs_info->fsid, 368f46b5a66SChristoph Hellwig (unsigned long)btrfs_header_fsid(leaf), 369f46b5a66SChristoph Hellwig BTRFS_FSID_SIZE); 3705d4f98a2SYan Zheng write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid, 3715d4f98a2SYan Zheng (unsigned long)btrfs_header_chunk_tree_uuid(leaf), 3725d4f98a2SYan Zheng BTRFS_UUID_SIZE); 373f46b5a66SChristoph Hellwig btrfs_mark_buffer_dirty(leaf); 374f46b5a66SChristoph Hellwig 375f46b5a66SChristoph Hellwig inode_item = &root_item.inode; 376f46b5a66SChristoph Hellwig memset(inode_item, 0, sizeof(*inode_item)); 377f46b5a66SChristoph Hellwig inode_item->generation = cpu_to_le64(1); 378f46b5a66SChristoph Hellwig inode_item->size = cpu_to_le64(3); 379f46b5a66SChristoph Hellwig inode_item->nlink = cpu_to_le32(1); 380a76a3cd4SYan Zheng inode_item->nbytes = cpu_to_le64(root->leafsize); 381f46b5a66SChristoph Hellwig inode_item->mode = cpu_to_le32(S_IFDIR | 0755); 382f46b5a66SChristoph Hellwig 38308fe4db1SLi Zefan root_item.flags = 0; 38408fe4db1SLi Zefan root_item.byte_limit = 0; 38508fe4db1SLi Zefan inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT); 38608fe4db1SLi Zefan 387f46b5a66SChristoph Hellwig btrfs_set_root_bytenr(&root_item, leaf->start); 38884234f3aSYan Zheng btrfs_set_root_generation(&root_item, trans->transid); 389f46b5a66SChristoph Hellwig btrfs_set_root_level(&root_item, 0); 390f46b5a66SChristoph Hellwig btrfs_set_root_refs(&root_item, 1); 39186b9f2ecSYan, Zheng btrfs_set_root_used(&root_item, leaf->len); 39280ff3856SYan Zheng btrfs_set_root_last_snapshot(&root_item, 0); 393f46b5a66SChristoph Hellwig 394f46b5a66SChristoph Hellwig memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress)); 395f46b5a66SChristoph Hellwig root_item.drop_level = 0; 396f46b5a66SChristoph Hellwig 397925baeddSChris Mason btrfs_tree_unlock(leaf); 398f46b5a66SChristoph Hellwig free_extent_buffer(leaf); 399f46b5a66SChristoph Hellwig leaf = NULL; 400f46b5a66SChristoph Hellwig 401f46b5a66SChristoph Hellwig btrfs_set_root_dirid(&root_item, new_dirid); 402f46b5a66SChristoph Hellwig 403f46b5a66SChristoph Hellwig key.objectid = objectid; 4045d4f98a2SYan Zheng key.offset = 0; 405f46b5a66SChristoph Hellwig btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); 406f46b5a66SChristoph Hellwig ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, 407f46b5a66SChristoph Hellwig &root_item); 408f46b5a66SChristoph Hellwig if (ret) 409f46b5a66SChristoph Hellwig goto fail; 410f46b5a66SChristoph Hellwig 41176dda93cSYan, Zheng key.offset = (u64)-1; 41276dda93cSYan, Zheng new_root = btrfs_read_fs_root_no_name(root->fs_info, &key); 41376dda93cSYan, Zheng BUG_ON(IS_ERR(new_root)); 41476dda93cSYan, Zheng 41576dda93cSYan, Zheng btrfs_record_root_in_trans(trans, new_root); 41676dda93cSYan, Zheng 417d82a6f1dSJosef Bacik ret = btrfs_create_subvol_root(trans, new_root, new_dirid); 418f46b5a66SChristoph Hellwig /* 419f46b5a66SChristoph Hellwig * insert the directory item 420f46b5a66SChristoph Hellwig */ 4213de4586cSChris Mason ret = btrfs_set_inode_index(dir, &index); 4223de4586cSChris Mason BUG_ON(ret); 4233de4586cSChris Mason 4243de4586cSChris Mason ret = btrfs_insert_dir_item(trans, root, 42516cdcec7SMiao Xie name, namelen, dir, &key, 4263de4586cSChris Mason BTRFS_FT_DIR, index); 427f46b5a66SChristoph Hellwig if (ret) 428f46b5a66SChristoph Hellwig goto fail; 4290660b5afSChris Mason 43052c26179SYan Zheng btrfs_i_size_write(dir, dir->i_size + namelen * 2); 43152c26179SYan Zheng ret = btrfs_update_inode(trans, root, dir); 43252c26179SYan Zheng BUG_ON(ret); 43352c26179SYan Zheng 4340660b5afSChris Mason ret = btrfs_add_root_ref(trans, root->fs_info->tree_root, 4354df27c4dSYan, Zheng objectid, root->root_key.objectid, 43633345d01SLi Zefan btrfs_ino(dir), index, name, namelen); 43776dda93cSYan, Zheng 4380660b5afSChris Mason BUG_ON(ret); 4390660b5afSChris Mason 44076dda93cSYan, Zheng d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry)); 441f46b5a66SChristoph Hellwig fail: 4426a912213SJosef Bacik dput(parent); 44372fd032eSSage Weil if (async_transid) { 44472fd032eSSage Weil *async_transid = trans->transid; 44572fd032eSSage Weil err = btrfs_commit_transaction_async(trans, root, 1); 44672fd032eSSage Weil } else { 44776dda93cSYan, Zheng err = btrfs_commit_transaction(trans, root); 44872fd032eSSage Weil } 449f46b5a66SChristoph Hellwig if (err && !ret) 450f46b5a66SChristoph Hellwig ret = err; 451f46b5a66SChristoph Hellwig return ret; 452f46b5a66SChristoph Hellwig } 453f46b5a66SChristoph Hellwig 45472fd032eSSage Weil static int create_snapshot(struct btrfs_root *root, struct dentry *dentry, 455b83cc969SLi Zefan char *name, int namelen, u64 *async_transid, 456b83cc969SLi Zefan bool readonly) 457f46b5a66SChristoph Hellwig { 4582e4bfab9SYan, Zheng struct inode *inode; 4596a912213SJosef Bacik struct dentry *parent; 460f46b5a66SChristoph Hellwig struct btrfs_pending_snapshot *pending_snapshot; 461f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 4622e4bfab9SYan, Zheng int ret; 463f46b5a66SChristoph Hellwig 464f46b5a66SChristoph Hellwig if (!root->ref_cows) 465f46b5a66SChristoph Hellwig return -EINVAL; 466f46b5a66SChristoph Hellwig 467a22285a6SYan, Zheng pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS); 468a22285a6SYan, Zheng if (!pending_snapshot) 469a22285a6SYan, Zheng return -ENOMEM; 470a22285a6SYan, Zheng 471a22285a6SYan, Zheng btrfs_init_block_rsv(&pending_snapshot->block_rsv); 472a22285a6SYan, Zheng pending_snapshot->dentry = dentry; 473a22285a6SYan, Zheng pending_snapshot->root = root; 474b83cc969SLi Zefan pending_snapshot->readonly = readonly; 475a22285a6SYan, Zheng 476a22285a6SYan, Zheng trans = btrfs_start_transaction(root->fs_info->extent_root, 5); 477a22285a6SYan, Zheng if (IS_ERR(trans)) { 478a22285a6SYan, Zheng ret = PTR_ERR(trans); 479a22285a6SYan, Zheng goto fail; 480a22285a6SYan, Zheng } 481a22285a6SYan, Zheng 482a22285a6SYan, Zheng ret = btrfs_snap_reserve_metadata(trans, pending_snapshot); 483a22285a6SYan, Zheng BUG_ON(ret); 484a22285a6SYan, Zheng 4858351583eSJosef Bacik spin_lock(&root->fs_info->trans_lock); 486a22285a6SYan, Zheng list_add(&pending_snapshot->list, 487a22285a6SYan, Zheng &trans->transaction->pending_snapshots); 4888351583eSJosef Bacik spin_unlock(&root->fs_info->trans_lock); 48972fd032eSSage Weil if (async_transid) { 49072fd032eSSage Weil *async_transid = trans->transid; 49172fd032eSSage Weil ret = btrfs_commit_transaction_async(trans, 49272fd032eSSage Weil root->fs_info->extent_root, 1); 49372fd032eSSage Weil } else { 49472fd032eSSage Weil ret = btrfs_commit_transaction(trans, 49572fd032eSSage Weil root->fs_info->extent_root); 49672fd032eSSage Weil } 497a22285a6SYan, Zheng BUG_ON(ret); 498a22285a6SYan, Zheng 499a22285a6SYan, Zheng ret = pending_snapshot->error; 500f46b5a66SChristoph Hellwig if (ret) 5012e4bfab9SYan, Zheng goto fail; 502f46b5a66SChristoph Hellwig 50366b4ffd1SJosef Bacik ret = btrfs_orphan_cleanup(pending_snapshot->snap); 50466b4ffd1SJosef Bacik if (ret) 50566b4ffd1SJosef Bacik goto fail; 506f46b5a66SChristoph Hellwig 5076a912213SJosef Bacik parent = dget_parent(dentry); 5086a912213SJosef Bacik inode = btrfs_lookup_dentry(parent->d_inode, dentry); 5096a912213SJosef Bacik dput(parent); 5102e4bfab9SYan, Zheng if (IS_ERR(inode)) { 5112e4bfab9SYan, Zheng ret = PTR_ERR(inode); 5122e4bfab9SYan, Zheng goto fail; 5132e4bfab9SYan, Zheng } 5142e4bfab9SYan, Zheng BUG_ON(!inode); 5152e4bfab9SYan, Zheng d_instantiate(dentry, inode); 5162e4bfab9SYan, Zheng ret = 0; 5172e4bfab9SYan, Zheng fail: 518a22285a6SYan, Zheng kfree(pending_snapshot); 519f46b5a66SChristoph Hellwig return ret; 520f46b5a66SChristoph Hellwig } 521f46b5a66SChristoph Hellwig 5224260f7c7SSage Weil /* copy of check_sticky in fs/namei.c() 5234260f7c7SSage Weil * It's inline, so penalty for filesystems that don't use sticky bit is 5244260f7c7SSage Weil * minimal. 5254260f7c7SSage Weil */ 5264260f7c7SSage Weil static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode) 5274260f7c7SSage Weil { 5284260f7c7SSage Weil uid_t fsuid = current_fsuid(); 5294260f7c7SSage Weil 5304260f7c7SSage Weil if (!(dir->i_mode & S_ISVTX)) 5314260f7c7SSage Weil return 0; 5324260f7c7SSage Weil if (inode->i_uid == fsuid) 5334260f7c7SSage Weil return 0; 5344260f7c7SSage Weil if (dir->i_uid == fsuid) 5354260f7c7SSage Weil return 0; 5364260f7c7SSage Weil return !capable(CAP_FOWNER); 5374260f7c7SSage Weil } 5384260f7c7SSage Weil 5394260f7c7SSage Weil /* copy of may_delete in fs/namei.c() 5404260f7c7SSage Weil * Check whether we can remove a link victim from directory dir, check 5414260f7c7SSage Weil * whether the type of victim is right. 5424260f7c7SSage Weil * 1. We can't do it if dir is read-only (done in permission()) 5434260f7c7SSage Weil * 2. We should have write and exec permissions on dir 5444260f7c7SSage Weil * 3. We can't remove anything from append-only dir 5454260f7c7SSage Weil * 4. We can't do anything with immutable dir (done in permission()) 5464260f7c7SSage Weil * 5. If the sticky bit on dir is set we should either 5474260f7c7SSage Weil * a. be owner of dir, or 5484260f7c7SSage Weil * b. be owner of victim, or 5494260f7c7SSage Weil * c. have CAP_FOWNER capability 5504260f7c7SSage Weil * 6. If the victim is append-only or immutable we can't do antyhing with 5514260f7c7SSage Weil * links pointing to it. 5524260f7c7SSage Weil * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. 5534260f7c7SSage Weil * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. 5544260f7c7SSage Weil * 9. We can't remove a root or mountpoint. 5554260f7c7SSage Weil * 10. We don't allow removal of NFS sillyrenamed files; it's handled by 5564260f7c7SSage Weil * nfs_async_unlink(). 5574260f7c7SSage Weil */ 5584260f7c7SSage Weil 5594260f7c7SSage Weil static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir) 5604260f7c7SSage Weil { 5614260f7c7SSage Weil int error; 5624260f7c7SSage Weil 5634260f7c7SSage Weil if (!victim->d_inode) 5644260f7c7SSage Weil return -ENOENT; 5654260f7c7SSage Weil 5664260f7c7SSage Weil BUG_ON(victim->d_parent->d_inode != dir); 5674260f7c7SSage Weil audit_inode_child(victim, dir); 5684260f7c7SSage Weil 5694260f7c7SSage Weil error = inode_permission(dir, MAY_WRITE | MAY_EXEC); 5704260f7c7SSage Weil if (error) 5714260f7c7SSage Weil return error; 5724260f7c7SSage Weil if (IS_APPEND(dir)) 5734260f7c7SSage Weil return -EPERM; 5744260f7c7SSage Weil if (btrfs_check_sticky(dir, victim->d_inode)|| 5754260f7c7SSage Weil IS_APPEND(victim->d_inode)|| 5764260f7c7SSage Weil IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode)) 5774260f7c7SSage Weil return -EPERM; 5784260f7c7SSage Weil if (isdir) { 5794260f7c7SSage Weil if (!S_ISDIR(victim->d_inode->i_mode)) 5804260f7c7SSage Weil return -ENOTDIR; 5814260f7c7SSage Weil if (IS_ROOT(victim)) 5824260f7c7SSage Weil return -EBUSY; 5834260f7c7SSage Weil } else if (S_ISDIR(victim->d_inode->i_mode)) 5844260f7c7SSage Weil return -EISDIR; 5854260f7c7SSage Weil if (IS_DEADDIR(dir)) 5864260f7c7SSage Weil return -ENOENT; 5874260f7c7SSage Weil if (victim->d_flags & DCACHE_NFSFS_RENAMED) 5884260f7c7SSage Weil return -EBUSY; 5894260f7c7SSage Weil return 0; 5904260f7c7SSage Weil } 5914260f7c7SSage Weil 592cb8e7090SChristoph Hellwig /* copy of may_create in fs/namei.c() */ 593cb8e7090SChristoph Hellwig static inline int btrfs_may_create(struct inode *dir, struct dentry *child) 594cb8e7090SChristoph Hellwig { 595cb8e7090SChristoph Hellwig if (child->d_inode) 596cb8e7090SChristoph Hellwig return -EEXIST; 597cb8e7090SChristoph Hellwig if (IS_DEADDIR(dir)) 598cb8e7090SChristoph Hellwig return -ENOENT; 599cb8e7090SChristoph Hellwig return inode_permission(dir, MAY_WRITE | MAY_EXEC); 600cb8e7090SChristoph Hellwig } 601cb8e7090SChristoph Hellwig 602cb8e7090SChristoph Hellwig /* 603cb8e7090SChristoph Hellwig * Create a new subvolume below @parent. This is largely modeled after 604cb8e7090SChristoph Hellwig * sys_mkdirat and vfs_mkdir, but we only do a single component lookup 605cb8e7090SChristoph Hellwig * inside this filesystem so it's quite a bit simpler. 606cb8e7090SChristoph Hellwig */ 60776dda93cSYan, Zheng static noinline int btrfs_mksubvol(struct path *parent, 60876dda93cSYan, Zheng char *name, int namelen, 60972fd032eSSage Weil struct btrfs_root *snap_src, 610b83cc969SLi Zefan u64 *async_transid, bool readonly) 611cb8e7090SChristoph Hellwig { 61276dda93cSYan, Zheng struct inode *dir = parent->dentry->d_inode; 613cb8e7090SChristoph Hellwig struct dentry *dentry; 614cb8e7090SChristoph Hellwig int error; 615cb8e7090SChristoph Hellwig 61676dda93cSYan, Zheng mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT); 617cb8e7090SChristoph Hellwig 618cb8e7090SChristoph Hellwig dentry = lookup_one_len(name, parent->dentry, namelen); 619cb8e7090SChristoph Hellwig error = PTR_ERR(dentry); 620cb8e7090SChristoph Hellwig if (IS_ERR(dentry)) 621cb8e7090SChristoph Hellwig goto out_unlock; 622cb8e7090SChristoph Hellwig 623cb8e7090SChristoph Hellwig error = -EEXIST; 624cb8e7090SChristoph Hellwig if (dentry->d_inode) 625cb8e7090SChristoph Hellwig goto out_dput; 626cb8e7090SChristoph Hellwig 627cb8e7090SChristoph Hellwig error = mnt_want_write(parent->mnt); 628cb8e7090SChristoph Hellwig if (error) 629cb8e7090SChristoph Hellwig goto out_dput; 630cb8e7090SChristoph Hellwig 63176dda93cSYan, Zheng error = btrfs_may_create(dir, dentry); 632cb8e7090SChristoph Hellwig if (error) 633cb8e7090SChristoph Hellwig goto out_drop_write; 634cb8e7090SChristoph Hellwig 63576dda93cSYan, Zheng down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem); 63676dda93cSYan, Zheng 63776dda93cSYan, Zheng if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) 63876dda93cSYan, Zheng goto out_up_read; 63976dda93cSYan, Zheng 6403de4586cSChris Mason if (snap_src) { 64172fd032eSSage Weil error = create_snapshot(snap_src, dentry, 642b83cc969SLi Zefan name, namelen, async_transid, readonly); 6433de4586cSChris Mason } else { 64476dda93cSYan, Zheng error = create_subvol(BTRFS_I(dir)->root, dentry, 64572fd032eSSage Weil name, namelen, async_transid); 6463de4586cSChris Mason } 64776dda93cSYan, Zheng if (!error) 64876dda93cSYan, Zheng fsnotify_mkdir(dir, dentry); 64976dda93cSYan, Zheng out_up_read: 65076dda93cSYan, Zheng up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem); 651cb8e7090SChristoph Hellwig out_drop_write: 652cb8e7090SChristoph Hellwig mnt_drop_write(parent->mnt); 653cb8e7090SChristoph Hellwig out_dput: 654cb8e7090SChristoph Hellwig dput(dentry); 655cb8e7090SChristoph Hellwig out_unlock: 65676dda93cSYan, Zheng mutex_unlock(&dir->i_mutex); 657cb8e7090SChristoph Hellwig return error; 658cb8e7090SChristoph Hellwig } 659cb8e7090SChristoph Hellwig 6604cb5300bSChris Mason /* 6614cb5300bSChris Mason * When we're defragging a range, we don't want to kick it off again 6624cb5300bSChris Mason * if it is really just waiting for delalloc to send it down. 6634cb5300bSChris Mason * If we find a nice big extent or delalloc range for the bytes in the 6644cb5300bSChris Mason * file you want to defrag, we return 0 to let you know to skip this 6654cb5300bSChris Mason * part of the file 6664cb5300bSChris Mason */ 6674cb5300bSChris Mason static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh) 6684cb5300bSChris Mason { 6694cb5300bSChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 6704cb5300bSChris Mason struct extent_map *em = NULL; 6714cb5300bSChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 6724cb5300bSChris Mason u64 end; 6734cb5300bSChris Mason 6744cb5300bSChris Mason read_lock(&em_tree->lock); 6754cb5300bSChris Mason em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE); 6764cb5300bSChris Mason read_unlock(&em_tree->lock); 6774cb5300bSChris Mason 6784cb5300bSChris Mason if (em) { 6794cb5300bSChris Mason end = extent_map_end(em); 6804cb5300bSChris Mason free_extent_map(em); 6814cb5300bSChris Mason if (end - offset > thresh) 6824cb5300bSChris Mason return 0; 6834cb5300bSChris Mason } 6844cb5300bSChris Mason /* if we already have a nice delalloc here, just stop */ 6854cb5300bSChris Mason thresh /= 2; 6864cb5300bSChris Mason end = count_range_bits(io_tree, &offset, offset + thresh, 6874cb5300bSChris Mason thresh, EXTENT_DELALLOC, 1); 6884cb5300bSChris Mason if (end >= thresh) 6894cb5300bSChris Mason return 0; 6904cb5300bSChris Mason return 1; 6914cb5300bSChris Mason } 6924cb5300bSChris Mason 6934cb5300bSChris Mason /* 6944cb5300bSChris Mason * helper function to walk through a file and find extents 6954cb5300bSChris Mason * newer than a specific transid, and smaller than thresh. 6964cb5300bSChris Mason * 6974cb5300bSChris Mason * This is used by the defragging code to find new and small 6984cb5300bSChris Mason * extents 6994cb5300bSChris Mason */ 7004cb5300bSChris Mason static int find_new_extents(struct btrfs_root *root, 7014cb5300bSChris Mason struct inode *inode, u64 newer_than, 7024cb5300bSChris Mason u64 *off, int thresh) 7034cb5300bSChris Mason { 7044cb5300bSChris Mason struct btrfs_path *path; 7054cb5300bSChris Mason struct btrfs_key min_key; 7064cb5300bSChris Mason struct btrfs_key max_key; 7074cb5300bSChris Mason struct extent_buffer *leaf; 7084cb5300bSChris Mason struct btrfs_file_extent_item *extent; 7094cb5300bSChris Mason int type; 7104cb5300bSChris Mason int ret; 711a4689d2bSDavid Sterba u64 ino = btrfs_ino(inode); 7124cb5300bSChris Mason 7134cb5300bSChris Mason path = btrfs_alloc_path(); 7144cb5300bSChris Mason if (!path) 7154cb5300bSChris Mason return -ENOMEM; 7164cb5300bSChris Mason 717a4689d2bSDavid Sterba min_key.objectid = ino; 7184cb5300bSChris Mason min_key.type = BTRFS_EXTENT_DATA_KEY; 7194cb5300bSChris Mason min_key.offset = *off; 7204cb5300bSChris Mason 721a4689d2bSDavid Sterba max_key.objectid = ino; 7224cb5300bSChris Mason max_key.type = (u8)-1; 7234cb5300bSChris Mason max_key.offset = (u64)-1; 7244cb5300bSChris Mason 7254cb5300bSChris Mason path->keep_locks = 1; 7264cb5300bSChris Mason 7274cb5300bSChris Mason while(1) { 7284cb5300bSChris Mason ret = btrfs_search_forward(root, &min_key, &max_key, 7294cb5300bSChris Mason path, 0, newer_than); 7304cb5300bSChris Mason if (ret != 0) 7314cb5300bSChris Mason goto none; 732a4689d2bSDavid Sterba if (min_key.objectid != ino) 7334cb5300bSChris Mason goto none; 7344cb5300bSChris Mason if (min_key.type != BTRFS_EXTENT_DATA_KEY) 7354cb5300bSChris Mason goto none; 7364cb5300bSChris Mason 7374cb5300bSChris Mason leaf = path->nodes[0]; 7384cb5300bSChris Mason extent = btrfs_item_ptr(leaf, path->slots[0], 7394cb5300bSChris Mason struct btrfs_file_extent_item); 7404cb5300bSChris Mason 7414cb5300bSChris Mason type = btrfs_file_extent_type(leaf, extent); 7424cb5300bSChris Mason if (type == BTRFS_FILE_EXTENT_REG && 7434cb5300bSChris Mason btrfs_file_extent_num_bytes(leaf, extent) < thresh && 7444cb5300bSChris Mason check_defrag_in_cache(inode, min_key.offset, thresh)) { 7454cb5300bSChris Mason *off = min_key.offset; 7464cb5300bSChris Mason btrfs_free_path(path); 7474cb5300bSChris Mason return 0; 7484cb5300bSChris Mason } 7494cb5300bSChris Mason 7504cb5300bSChris Mason if (min_key.offset == (u64)-1) 7514cb5300bSChris Mason goto none; 7524cb5300bSChris Mason 7534cb5300bSChris Mason min_key.offset++; 7544cb5300bSChris Mason btrfs_release_path(path); 7554cb5300bSChris Mason } 7564cb5300bSChris Mason none: 7574cb5300bSChris Mason btrfs_free_path(path); 7584cb5300bSChris Mason return -ENOENT; 7594cb5300bSChris Mason } 7604cb5300bSChris Mason 761940100a4SChris Mason static int should_defrag_range(struct inode *inode, u64 start, u64 len, 7621e701a32SChris Mason int thresh, u64 *last_len, u64 *skip, 7631e701a32SChris Mason u64 *defrag_end) 764940100a4SChris Mason { 765940100a4SChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 766940100a4SChris Mason struct extent_map *em = NULL; 767940100a4SChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 768940100a4SChris Mason int ret = 1; 769940100a4SChris Mason 770940100a4SChris Mason /* 771940100a4SChris Mason * make sure that once we start defragging and extent, we keep on 772940100a4SChris Mason * defragging it 773940100a4SChris Mason */ 774940100a4SChris Mason if (start < *defrag_end) 775940100a4SChris Mason return 1; 776940100a4SChris Mason 777940100a4SChris Mason *skip = 0; 778940100a4SChris Mason 779940100a4SChris Mason /* 780940100a4SChris Mason * hopefully we have this extent in the tree already, try without 781940100a4SChris Mason * the full extent lock 782940100a4SChris Mason */ 783940100a4SChris Mason read_lock(&em_tree->lock); 784940100a4SChris Mason em = lookup_extent_mapping(em_tree, start, len); 785940100a4SChris Mason read_unlock(&em_tree->lock); 786940100a4SChris Mason 787940100a4SChris Mason if (!em) { 788940100a4SChris Mason /* get the big lock and read metadata off disk */ 789940100a4SChris Mason lock_extent(io_tree, start, start + len - 1, GFP_NOFS); 790940100a4SChris Mason em = btrfs_get_extent(inode, NULL, 0, start, len, 0); 791940100a4SChris Mason unlock_extent(io_tree, start, start + len - 1, GFP_NOFS); 792940100a4SChris Mason 7936cf8bfbfSDan Carpenter if (IS_ERR(em)) 794940100a4SChris Mason return 0; 795940100a4SChris Mason } 796940100a4SChris Mason 797940100a4SChris Mason /* this will cover holes, and inline extents */ 798940100a4SChris Mason if (em->block_start >= EXTENT_MAP_LAST_BYTE) 799940100a4SChris Mason ret = 0; 800940100a4SChris Mason 801940100a4SChris Mason /* 802940100a4SChris Mason * we hit a real extent, if it is big don't bother defragging it again 803940100a4SChris Mason */ 8041e701a32SChris Mason if ((*last_len == 0 || *last_len >= thresh) && em->len >= thresh) 805940100a4SChris Mason ret = 0; 806940100a4SChris Mason 807940100a4SChris Mason /* 808940100a4SChris Mason * last_len ends up being a counter of how many bytes we've defragged. 809940100a4SChris Mason * every time we choose not to defrag an extent, we reset *last_len 810940100a4SChris Mason * so that the next tiny extent will force a defrag. 811940100a4SChris Mason * 812940100a4SChris Mason * The end result of this is that tiny extents before a single big 813940100a4SChris Mason * extent will force at least part of that big extent to be defragged. 814940100a4SChris Mason */ 815940100a4SChris Mason if (ret) { 816940100a4SChris Mason *last_len += len; 817940100a4SChris Mason *defrag_end = extent_map_end(em); 818940100a4SChris Mason } else { 819940100a4SChris Mason *last_len = 0; 820940100a4SChris Mason *skip = extent_map_end(em); 821940100a4SChris Mason *defrag_end = 0; 822940100a4SChris Mason } 823940100a4SChris Mason 824940100a4SChris Mason free_extent_map(em); 825940100a4SChris Mason return ret; 826940100a4SChris Mason } 827940100a4SChris Mason 8284cb5300bSChris Mason /* 8294cb5300bSChris Mason * it doesn't do much good to defrag one or two pages 8304cb5300bSChris Mason * at a time. This pulls in a nice chunk of pages 8314cb5300bSChris Mason * to COW and defrag. 8324cb5300bSChris Mason * 8334cb5300bSChris Mason * It also makes sure the delalloc code has enough 8344cb5300bSChris Mason * dirty data to avoid making new small extents as part 8354cb5300bSChris Mason * of the defrag 8364cb5300bSChris Mason * 8374cb5300bSChris Mason * It's a good idea to start RA on this range 8384cb5300bSChris Mason * before calling this. 8394cb5300bSChris Mason */ 8404cb5300bSChris Mason static int cluster_pages_for_defrag(struct inode *inode, 8414cb5300bSChris Mason struct page **pages, 8424cb5300bSChris Mason unsigned long start_index, 8434cb5300bSChris Mason int num_pages) 844f46b5a66SChristoph Hellwig { 8454cb5300bSChris Mason unsigned long file_end; 8464cb5300bSChris Mason u64 isize = i_size_read(inode); 847f46b5a66SChristoph Hellwig u64 page_start; 848f46b5a66SChristoph Hellwig u64 page_end; 8494cb5300bSChris Mason int ret; 8504cb5300bSChris Mason int i; 8514cb5300bSChris Mason int i_done; 8524cb5300bSChris Mason struct btrfs_ordered_extent *ordered; 8534cb5300bSChris Mason struct extent_state *cached_state = NULL; 8544cb5300bSChris Mason 8554cb5300bSChris Mason if (isize == 0) 8564cb5300bSChris Mason return 0; 8574cb5300bSChris Mason file_end = (isize - 1) >> PAGE_CACHE_SHIFT; 8584cb5300bSChris Mason 8594cb5300bSChris Mason ret = btrfs_delalloc_reserve_space(inode, 8604cb5300bSChris Mason num_pages << PAGE_CACHE_SHIFT); 8614cb5300bSChris Mason if (ret) 8624cb5300bSChris Mason return ret; 8634cb5300bSChris Mason again: 8644cb5300bSChris Mason ret = 0; 8654cb5300bSChris Mason i_done = 0; 8664cb5300bSChris Mason 8674cb5300bSChris Mason /* step one, lock all the pages */ 8684cb5300bSChris Mason for (i = 0; i < num_pages; i++) { 8694cb5300bSChris Mason struct page *page; 870a94733d0SJosef Bacik page = find_or_create_page(inode->i_mapping, 871a94733d0SJosef Bacik start_index + i, GFP_NOFS); 8724cb5300bSChris Mason if (!page) 8734cb5300bSChris Mason break; 8744cb5300bSChris Mason 8754cb5300bSChris Mason if (!PageUptodate(page)) { 8764cb5300bSChris Mason btrfs_readpage(NULL, page); 8774cb5300bSChris Mason lock_page(page); 8784cb5300bSChris Mason if (!PageUptodate(page)) { 8794cb5300bSChris Mason unlock_page(page); 8804cb5300bSChris Mason page_cache_release(page); 8814cb5300bSChris Mason ret = -EIO; 8824cb5300bSChris Mason break; 8834cb5300bSChris Mason } 8844cb5300bSChris Mason } 8854cb5300bSChris Mason isize = i_size_read(inode); 8864cb5300bSChris Mason file_end = (isize - 1) >> PAGE_CACHE_SHIFT; 8874cb5300bSChris Mason if (!isize || page->index > file_end || 8884cb5300bSChris Mason page->mapping != inode->i_mapping) { 8894cb5300bSChris Mason /* whoops, we blew past eof, skip this page */ 8904cb5300bSChris Mason unlock_page(page); 8914cb5300bSChris Mason page_cache_release(page); 8924cb5300bSChris Mason break; 8934cb5300bSChris Mason } 8944cb5300bSChris Mason pages[i] = page; 8954cb5300bSChris Mason i_done++; 8964cb5300bSChris Mason } 8974cb5300bSChris Mason if (!i_done || ret) 8984cb5300bSChris Mason goto out; 8994cb5300bSChris Mason 9004cb5300bSChris Mason if (!(inode->i_sb->s_flags & MS_ACTIVE)) 9014cb5300bSChris Mason goto out; 9024cb5300bSChris Mason 9034cb5300bSChris Mason /* 9044cb5300bSChris Mason * so now we have a nice long stream of locked 9054cb5300bSChris Mason * and up to date pages, lets wait on them 9064cb5300bSChris Mason */ 9074cb5300bSChris Mason for (i = 0; i < i_done; i++) 9084cb5300bSChris Mason wait_on_page_writeback(pages[i]); 9094cb5300bSChris Mason 9104cb5300bSChris Mason page_start = page_offset(pages[0]); 9114cb5300bSChris Mason page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE; 9124cb5300bSChris Mason 9134cb5300bSChris Mason lock_extent_bits(&BTRFS_I(inode)->io_tree, 9144cb5300bSChris Mason page_start, page_end - 1, 0, &cached_state, 9154cb5300bSChris Mason GFP_NOFS); 9164cb5300bSChris Mason ordered = btrfs_lookup_first_ordered_extent(inode, page_end - 1); 9174cb5300bSChris Mason if (ordered && 9184cb5300bSChris Mason ordered->file_offset + ordered->len > page_start && 9194cb5300bSChris Mason ordered->file_offset < page_end) { 9204cb5300bSChris Mason btrfs_put_ordered_extent(ordered); 9214cb5300bSChris Mason unlock_extent_cached(&BTRFS_I(inode)->io_tree, 9224cb5300bSChris Mason page_start, page_end - 1, 9234cb5300bSChris Mason &cached_state, GFP_NOFS); 9244cb5300bSChris Mason for (i = 0; i < i_done; i++) { 9254cb5300bSChris Mason unlock_page(pages[i]); 9264cb5300bSChris Mason page_cache_release(pages[i]); 9274cb5300bSChris Mason } 9284cb5300bSChris Mason btrfs_wait_ordered_range(inode, page_start, 9294cb5300bSChris Mason page_end - page_start); 9304cb5300bSChris Mason goto again; 9314cb5300bSChris Mason } 9324cb5300bSChris Mason if (ordered) 9334cb5300bSChris Mason btrfs_put_ordered_extent(ordered); 9344cb5300bSChris Mason 9354cb5300bSChris Mason clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, 9364cb5300bSChris Mason page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC | 9374cb5300bSChris Mason EXTENT_DO_ACCOUNTING, 0, 0, &cached_state, 9384cb5300bSChris Mason GFP_NOFS); 9394cb5300bSChris Mason 9404cb5300bSChris Mason if (i_done != num_pages) { 9419e0baf60SJosef Bacik spin_lock(&BTRFS_I(inode)->lock); 9429e0baf60SJosef Bacik BTRFS_I(inode)->outstanding_extents++; 9439e0baf60SJosef Bacik spin_unlock(&BTRFS_I(inode)->lock); 9444cb5300bSChris Mason btrfs_delalloc_release_space(inode, 9454cb5300bSChris Mason (num_pages - i_done) << PAGE_CACHE_SHIFT); 9464cb5300bSChris Mason } 9474cb5300bSChris Mason 9484cb5300bSChris Mason 9494cb5300bSChris Mason btrfs_set_extent_delalloc(inode, page_start, page_end - 1, 9504cb5300bSChris Mason &cached_state); 9514cb5300bSChris Mason 9524cb5300bSChris Mason unlock_extent_cached(&BTRFS_I(inode)->io_tree, 9534cb5300bSChris Mason page_start, page_end - 1, &cached_state, 9544cb5300bSChris Mason GFP_NOFS); 9554cb5300bSChris Mason 9564cb5300bSChris Mason for (i = 0; i < i_done; i++) { 9574cb5300bSChris Mason clear_page_dirty_for_io(pages[i]); 9584cb5300bSChris Mason ClearPageChecked(pages[i]); 9594cb5300bSChris Mason set_page_extent_mapped(pages[i]); 9604cb5300bSChris Mason set_page_dirty(pages[i]); 9614cb5300bSChris Mason unlock_page(pages[i]); 9624cb5300bSChris Mason page_cache_release(pages[i]); 9634cb5300bSChris Mason } 9644cb5300bSChris Mason return i_done; 9654cb5300bSChris Mason out: 9664cb5300bSChris Mason for (i = 0; i < i_done; i++) { 9674cb5300bSChris Mason unlock_page(pages[i]); 9684cb5300bSChris Mason page_cache_release(pages[i]); 9694cb5300bSChris Mason } 9704cb5300bSChris Mason btrfs_delalloc_release_space(inode, num_pages << PAGE_CACHE_SHIFT); 9714cb5300bSChris Mason return ret; 9724cb5300bSChris Mason 9734cb5300bSChris Mason } 9744cb5300bSChris Mason 9754cb5300bSChris Mason int btrfs_defrag_file(struct inode *inode, struct file *file, 9764cb5300bSChris Mason struct btrfs_ioctl_defrag_range_args *range, 9774cb5300bSChris Mason u64 newer_than, unsigned long max_to_defrag) 9784cb5300bSChris Mason { 9794cb5300bSChris Mason struct btrfs_root *root = BTRFS_I(inode)->root; 9804cb5300bSChris Mason struct btrfs_super_block *disk_super; 9814cb5300bSChris Mason struct file_ra_state *ra = NULL; 9824cb5300bSChris Mason unsigned long last_index; 9834cb5300bSChris Mason u64 features; 984940100a4SChris Mason u64 last_len = 0; 985940100a4SChris Mason u64 skip = 0; 986940100a4SChris Mason u64 defrag_end = 0; 9874cb5300bSChris Mason u64 newer_off = range->start; 9884cb5300bSChris Mason int newer_left = 0; 989f46b5a66SChristoph Hellwig unsigned long i; 990f46b5a66SChristoph Hellwig int ret; 9914cb5300bSChris Mason int defrag_count = 0; 9921a419d85SLi Zefan int compress_type = BTRFS_COMPRESS_ZLIB; 9934cb5300bSChris Mason int extent_thresh = range->extent_thresh; 9944cb5300bSChris Mason int newer_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT; 9954cb5300bSChris Mason u64 new_align = ~((u64)128 * 1024 - 1); 9964cb5300bSChris Mason struct page **pages = NULL; 9974cb5300bSChris Mason 9984cb5300bSChris Mason if (extent_thresh == 0) 9994cb5300bSChris Mason extent_thresh = 256 * 1024; 10001a419d85SLi Zefan 10011a419d85SLi Zefan if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) { 10021a419d85SLi Zefan if (range->compress_type > BTRFS_COMPRESS_TYPES) 10031a419d85SLi Zefan return -EINVAL; 10041a419d85SLi Zefan if (range->compress_type) 10051a419d85SLi Zefan compress_type = range->compress_type; 10061a419d85SLi Zefan } 1007f46b5a66SChristoph Hellwig 1008940100a4SChris Mason if (inode->i_size == 0) 1009940100a4SChris Mason return 0; 1010f46b5a66SChristoph Hellwig 10114cb5300bSChris Mason /* 10124cb5300bSChris Mason * if we were not given a file, allocate a readahead 10134cb5300bSChris Mason * context 10144cb5300bSChris Mason */ 10154cb5300bSChris Mason if (!file) { 10164cb5300bSChris Mason ra = kzalloc(sizeof(*ra), GFP_NOFS); 10174cb5300bSChris Mason if (!ra) 10184cb5300bSChris Mason return -ENOMEM; 10194cb5300bSChris Mason file_ra_state_init(ra, inode->i_mapping); 10204cb5300bSChris Mason } else { 10214cb5300bSChris Mason ra = &file->f_ra; 10224cb5300bSChris Mason } 10234cb5300bSChris Mason 10244cb5300bSChris Mason pages = kmalloc(sizeof(struct page *) * newer_cluster, 10254cb5300bSChris Mason GFP_NOFS); 10264cb5300bSChris Mason if (!pages) { 10274cb5300bSChris Mason ret = -ENOMEM; 10284cb5300bSChris Mason goto out_ra; 10294cb5300bSChris Mason } 10304cb5300bSChris Mason 10314cb5300bSChris Mason /* find the last page to defrag */ 10321e701a32SChris Mason if (range->start + range->len > range->start) { 10331e701a32SChris Mason last_index = min_t(u64, inode->i_size - 1, 10341e701a32SChris Mason range->start + range->len - 1) >> PAGE_CACHE_SHIFT; 10351e701a32SChris Mason } else { 1036940100a4SChris Mason last_index = (inode->i_size - 1) >> PAGE_CACHE_SHIFT; 10371e701a32SChris Mason } 10381e701a32SChris Mason 10394cb5300bSChris Mason if (newer_than) { 10404cb5300bSChris Mason ret = find_new_extents(root, inode, newer_than, 10414cb5300bSChris Mason &newer_off, 64 * 1024); 10424cb5300bSChris Mason if (!ret) { 10434cb5300bSChris Mason range->start = newer_off; 10444cb5300bSChris Mason /* 10454cb5300bSChris Mason * we always align our defrag to help keep 10464cb5300bSChris Mason * the extents in the file evenly spaced 10474cb5300bSChris Mason */ 10484cb5300bSChris Mason i = (newer_off & new_align) >> PAGE_CACHE_SHIFT; 10494cb5300bSChris Mason newer_left = newer_cluster; 10504cb5300bSChris Mason } else 10514cb5300bSChris Mason goto out_ra; 10524cb5300bSChris Mason } else { 10531e701a32SChris Mason i = range->start >> PAGE_CACHE_SHIFT; 10544cb5300bSChris Mason } 10554cb5300bSChris Mason if (!max_to_defrag) 10564cb5300bSChris Mason max_to_defrag = last_index - 1; 10574cb5300bSChris Mason 10584cb5300bSChris Mason while (i <= last_index && defrag_count < max_to_defrag) { 10594cb5300bSChris Mason /* 10604cb5300bSChris Mason * make sure we stop running if someone unmounts 10614cb5300bSChris Mason * the FS 10624cb5300bSChris Mason */ 10634cb5300bSChris Mason if (!(inode->i_sb->s_flags & MS_ACTIVE)) 10644cb5300bSChris Mason break; 10654cb5300bSChris Mason 10664cb5300bSChris Mason if (!newer_than && 10674cb5300bSChris Mason !should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT, 10681e701a32SChris Mason PAGE_CACHE_SIZE, 10694cb5300bSChris Mason extent_thresh, 10701e701a32SChris Mason &last_len, &skip, 1071940100a4SChris Mason &defrag_end)) { 1072940100a4SChris Mason unsigned long next; 1073940100a4SChris Mason /* 1074940100a4SChris Mason * the should_defrag function tells us how much to skip 1075940100a4SChris Mason * bump our counter by the suggested amount 1076940100a4SChris Mason */ 1077940100a4SChris Mason next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT; 1078940100a4SChris Mason i = max(i + 1, next); 1079940100a4SChris Mason continue; 1080940100a4SChris Mason } 10811e701a32SChris Mason if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) 10821a419d85SLi Zefan BTRFS_I(inode)->force_compress = compress_type; 1083940100a4SChris Mason 10844cb5300bSChris Mason btrfs_force_ra(inode->i_mapping, ra, file, i, newer_cluster); 10854cb5300bSChris Mason 10864cb5300bSChris Mason ret = cluster_pages_for_defrag(inode, pages, i, newer_cluster); 10874cb5300bSChris Mason if (ret < 0) 10884cb5300bSChris Mason goto out_ra; 10894cb5300bSChris Mason 10904cb5300bSChris Mason defrag_count += ret; 10914cb5300bSChris Mason balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret); 10924cb5300bSChris Mason i += ret; 10934cb5300bSChris Mason 10944cb5300bSChris Mason if (newer_than) { 10954cb5300bSChris Mason if (newer_off == (u64)-1) 10964cb5300bSChris Mason break; 10974cb5300bSChris Mason 10984cb5300bSChris Mason newer_off = max(newer_off + 1, 10994cb5300bSChris Mason (u64)i << PAGE_CACHE_SHIFT); 11004cb5300bSChris Mason 11014cb5300bSChris Mason ret = find_new_extents(root, inode, 11024cb5300bSChris Mason newer_than, &newer_off, 11034cb5300bSChris Mason 64 * 1024); 11044cb5300bSChris Mason if (!ret) { 11054cb5300bSChris Mason range->start = newer_off; 11064cb5300bSChris Mason i = (newer_off & new_align) >> PAGE_CACHE_SHIFT; 11074cb5300bSChris Mason newer_left = newer_cluster; 11084cb5300bSChris Mason } else { 11094cb5300bSChris Mason break; 1110940100a4SChris Mason } 11114cb5300bSChris Mason } else { 1112940100a4SChris Mason i++; 1113f46b5a66SChristoph Hellwig } 11144cb5300bSChris Mason } 1115f46b5a66SChristoph Hellwig 11161e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) 11171e701a32SChris Mason filemap_flush(inode->i_mapping); 11181e701a32SChris Mason 11191e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 11201e701a32SChris Mason /* the filemap_flush will queue IO into the worker threads, but 11211e701a32SChris Mason * we have to make sure the IO is actually started and that 11221e701a32SChris Mason * ordered extents get created before we return 11231e701a32SChris Mason */ 11241e701a32SChris Mason atomic_inc(&root->fs_info->async_submit_draining); 11251e701a32SChris Mason while (atomic_read(&root->fs_info->nr_async_submits) || 11261e701a32SChris Mason atomic_read(&root->fs_info->async_delalloc_pages)) { 11271e701a32SChris Mason wait_event(root->fs_info->async_submit_wait, 11281e701a32SChris Mason (atomic_read(&root->fs_info->nr_async_submits) == 0 && 11291e701a32SChris Mason atomic_read(&root->fs_info->async_delalloc_pages) == 0)); 11301e701a32SChris Mason } 11311e701a32SChris Mason atomic_dec(&root->fs_info->async_submit_draining); 11321e701a32SChris Mason 11331e701a32SChris Mason mutex_lock(&inode->i_mutex); 1134261507a0SLi Zefan BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE; 11351e701a32SChris Mason mutex_unlock(&inode->i_mutex); 11361e701a32SChris Mason } 11371e701a32SChris Mason 11381a419d85SLi Zefan disk_super = &root->fs_info->super_copy; 11391a419d85SLi Zefan features = btrfs_super_incompat_flags(disk_super); 11401a419d85SLi Zefan if (range->compress_type == BTRFS_COMPRESS_LZO) { 11411a419d85SLi Zefan features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; 11421a419d85SLi Zefan btrfs_set_super_incompat_flags(disk_super, features); 11431a419d85SLi Zefan } 11441a419d85SLi Zefan 11454cb5300bSChris Mason if (!file) 11464cb5300bSChris Mason kfree(ra); 11474cb5300bSChris Mason return defrag_count; 1148940100a4SChris Mason 11494cb5300bSChris Mason out_ra: 11504cb5300bSChris Mason if (!file) 11514cb5300bSChris Mason kfree(ra); 11524cb5300bSChris Mason kfree(pages); 1153940100a4SChris Mason return ret; 1154f46b5a66SChristoph Hellwig } 1155f46b5a66SChristoph Hellwig 115676dda93cSYan, Zheng static noinline int btrfs_ioctl_resize(struct btrfs_root *root, 115776dda93cSYan, Zheng void __user *arg) 1158f46b5a66SChristoph Hellwig { 1159f46b5a66SChristoph Hellwig u64 new_size; 1160f46b5a66SChristoph Hellwig u64 old_size; 1161f46b5a66SChristoph Hellwig u64 devid = 1; 1162f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 1163f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 1164f46b5a66SChristoph Hellwig struct btrfs_device *device = NULL; 1165f46b5a66SChristoph Hellwig char *sizestr; 1166f46b5a66SChristoph Hellwig char *devstr = NULL; 1167f46b5a66SChristoph Hellwig int ret = 0; 1168f46b5a66SChristoph Hellwig int mod = 0; 1169f46b5a66SChristoph Hellwig 1170c146afadSYan Zheng if (root->fs_info->sb->s_flags & MS_RDONLY) 1171c146afadSYan Zheng return -EROFS; 1172c146afadSYan Zheng 1173e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 1174e441d54dSChris Mason return -EPERM; 1175e441d54dSChris Mason 1176dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1177dae7b665SLi Zefan if (IS_ERR(vol_args)) 1178dae7b665SLi Zefan return PTR_ERR(vol_args); 11795516e595SMark Fasheh 11805516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1181f46b5a66SChristoph Hellwig 11827d9eb12cSChris Mason mutex_lock(&root->fs_info->volume_mutex); 1183f46b5a66SChristoph Hellwig sizestr = vol_args->name; 1184f46b5a66SChristoph Hellwig devstr = strchr(sizestr, ':'); 1185f46b5a66SChristoph Hellwig if (devstr) { 1186f46b5a66SChristoph Hellwig char *end; 1187f46b5a66SChristoph Hellwig sizestr = devstr + 1; 1188f46b5a66SChristoph Hellwig *devstr = '\0'; 1189f46b5a66SChristoph Hellwig devstr = vol_args->name; 1190f46b5a66SChristoph Hellwig devid = simple_strtoull(devstr, &end, 10); 119121380931SJoel Becker printk(KERN_INFO "resizing devid %llu\n", 119221380931SJoel Becker (unsigned long long)devid); 1193f46b5a66SChristoph Hellwig } 11942b82032cSYan Zheng device = btrfs_find_device(root, devid, NULL, NULL); 1195f46b5a66SChristoph Hellwig if (!device) { 119621380931SJoel Becker printk(KERN_INFO "resizer unable to find device %llu\n", 119721380931SJoel Becker (unsigned long long)devid); 1198f46b5a66SChristoph Hellwig ret = -EINVAL; 1199f46b5a66SChristoph Hellwig goto out_unlock; 1200f46b5a66SChristoph Hellwig } 1201f46b5a66SChristoph Hellwig if (!strcmp(sizestr, "max")) 1202f46b5a66SChristoph Hellwig new_size = device->bdev->bd_inode->i_size; 1203f46b5a66SChristoph Hellwig else { 1204f46b5a66SChristoph Hellwig if (sizestr[0] == '-') { 1205f46b5a66SChristoph Hellwig mod = -1; 1206f46b5a66SChristoph Hellwig sizestr++; 1207f46b5a66SChristoph Hellwig } else if (sizestr[0] == '+') { 1208f46b5a66SChristoph Hellwig mod = 1; 1209f46b5a66SChristoph Hellwig sizestr++; 1210f46b5a66SChristoph Hellwig } 121191748467SAkinobu Mita new_size = memparse(sizestr, NULL); 1212f46b5a66SChristoph Hellwig if (new_size == 0) { 1213f46b5a66SChristoph Hellwig ret = -EINVAL; 1214f46b5a66SChristoph Hellwig goto out_unlock; 1215f46b5a66SChristoph Hellwig } 1216f46b5a66SChristoph Hellwig } 1217f46b5a66SChristoph Hellwig 1218f46b5a66SChristoph Hellwig old_size = device->total_bytes; 1219f46b5a66SChristoph Hellwig 1220f46b5a66SChristoph Hellwig if (mod < 0) { 1221f46b5a66SChristoph Hellwig if (new_size > old_size) { 1222f46b5a66SChristoph Hellwig ret = -EINVAL; 1223f46b5a66SChristoph Hellwig goto out_unlock; 1224f46b5a66SChristoph Hellwig } 1225f46b5a66SChristoph Hellwig new_size = old_size - new_size; 1226f46b5a66SChristoph Hellwig } else if (mod > 0) { 1227f46b5a66SChristoph Hellwig new_size = old_size + new_size; 1228f46b5a66SChristoph Hellwig } 1229f46b5a66SChristoph Hellwig 1230f46b5a66SChristoph Hellwig if (new_size < 256 * 1024 * 1024) { 1231f46b5a66SChristoph Hellwig ret = -EINVAL; 1232f46b5a66SChristoph Hellwig goto out_unlock; 1233f46b5a66SChristoph Hellwig } 1234f46b5a66SChristoph Hellwig if (new_size > device->bdev->bd_inode->i_size) { 1235f46b5a66SChristoph Hellwig ret = -EFBIG; 1236f46b5a66SChristoph Hellwig goto out_unlock; 1237f46b5a66SChristoph Hellwig } 1238f46b5a66SChristoph Hellwig 1239f46b5a66SChristoph Hellwig do_div(new_size, root->sectorsize); 1240f46b5a66SChristoph Hellwig new_size *= root->sectorsize; 1241f46b5a66SChristoph Hellwig 1242f46b5a66SChristoph Hellwig printk(KERN_INFO "new size for %s is %llu\n", 1243f46b5a66SChristoph Hellwig device->name, (unsigned long long)new_size); 1244f46b5a66SChristoph Hellwig 1245f46b5a66SChristoph Hellwig if (new_size > old_size) { 1246a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 124798d5dc13STsutomu Itoh if (IS_ERR(trans)) { 124898d5dc13STsutomu Itoh ret = PTR_ERR(trans); 124998d5dc13STsutomu Itoh goto out_unlock; 125098d5dc13STsutomu Itoh } 1251f46b5a66SChristoph Hellwig ret = btrfs_grow_device(trans, device, new_size); 1252f46b5a66SChristoph Hellwig btrfs_commit_transaction(trans, root); 1253f46b5a66SChristoph Hellwig } else { 1254f46b5a66SChristoph Hellwig ret = btrfs_shrink_device(device, new_size); 1255f46b5a66SChristoph Hellwig } 1256f46b5a66SChristoph Hellwig 1257f46b5a66SChristoph Hellwig out_unlock: 12587d9eb12cSChris Mason mutex_unlock(&root->fs_info->volume_mutex); 1259f46b5a66SChristoph Hellwig kfree(vol_args); 1260f46b5a66SChristoph Hellwig return ret; 1261f46b5a66SChristoph Hellwig } 1262f46b5a66SChristoph Hellwig 126372fd032eSSage Weil static noinline int btrfs_ioctl_snap_create_transid(struct file *file, 126472fd032eSSage Weil char *name, 126572fd032eSSage Weil unsigned long fd, 126672fd032eSSage Weil int subvol, 1267b83cc969SLi Zefan u64 *transid, 1268b83cc969SLi Zefan bool readonly) 1269f46b5a66SChristoph Hellwig { 1270cb8e7090SChristoph Hellwig struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; 12713de4586cSChris Mason struct file *src_file; 1272f46b5a66SChristoph Hellwig int namelen; 12733de4586cSChris Mason int ret = 0; 1274f46b5a66SChristoph Hellwig 1275c146afadSYan Zheng if (root->fs_info->sb->s_flags & MS_RDONLY) 1276c146afadSYan Zheng return -EROFS; 1277c146afadSYan Zheng 127872fd032eSSage Weil namelen = strlen(name); 127972fd032eSSage Weil if (strchr(name, '/')) { 1280f46b5a66SChristoph Hellwig ret = -EINVAL; 1281f46b5a66SChristoph Hellwig goto out; 1282f46b5a66SChristoph Hellwig } 1283f46b5a66SChristoph Hellwig 12843de4586cSChris Mason if (subvol) { 128572fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 1286b83cc969SLi Zefan NULL, transid, readonly); 1287cb8e7090SChristoph Hellwig } else { 12883de4586cSChris Mason struct inode *src_inode; 128972fd032eSSage Weil src_file = fget(fd); 12903de4586cSChris Mason if (!src_file) { 12913de4586cSChris Mason ret = -EINVAL; 12923de4586cSChris Mason goto out; 12933de4586cSChris Mason } 12943de4586cSChris Mason 12953de4586cSChris Mason src_inode = src_file->f_path.dentry->d_inode; 12963de4586cSChris Mason if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) { 1297d397712bSChris Mason printk(KERN_INFO "btrfs: Snapshot src from " 1298d397712bSChris Mason "another FS\n"); 12993de4586cSChris Mason ret = -EINVAL; 13003de4586cSChris Mason fput(src_file); 13013de4586cSChris Mason goto out; 13023de4586cSChris Mason } 130372fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 130472fd032eSSage Weil BTRFS_I(src_inode)->root, 1305b83cc969SLi Zefan transid, readonly); 13063de4586cSChris Mason fput(src_file); 1307cb8e7090SChristoph Hellwig } 1308f46b5a66SChristoph Hellwig out: 130972fd032eSSage Weil return ret; 131072fd032eSSage Weil } 131172fd032eSSage Weil 131272fd032eSSage Weil static noinline int btrfs_ioctl_snap_create(struct file *file, 1313fa0d2b9bSLi Zefan void __user *arg, int subvol) 131472fd032eSSage Weil { 1315fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args *vol_args; 131672fd032eSSage Weil int ret; 131772fd032eSSage Weil 1318fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1319fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1320fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1321fa0d2b9bSLi Zefan vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1322fa0d2b9bSLi Zefan 1323fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 1324b83cc969SLi Zefan vol_args->fd, subvol, 1325b83cc969SLi Zefan NULL, false); 1326fa0d2b9bSLi Zefan 1327fa0d2b9bSLi Zefan kfree(vol_args); 1328fa0d2b9bSLi Zefan return ret; 1329fa0d2b9bSLi Zefan } 1330fa0d2b9bSLi Zefan 1331fa0d2b9bSLi Zefan static noinline int btrfs_ioctl_snap_create_v2(struct file *file, 1332fa0d2b9bSLi Zefan void __user *arg, int subvol) 1333fa0d2b9bSLi Zefan { 1334fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args_v2 *vol_args; 1335fa0d2b9bSLi Zefan int ret; 1336fdfb1e4fSLi Zefan u64 transid = 0; 1337fdfb1e4fSLi Zefan u64 *ptr = NULL; 1338b83cc969SLi Zefan bool readonly = false; 133972fd032eSSage Weil 1340fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1341fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1342fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1343fa0d2b9bSLi Zefan vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 1344fdfb1e4fSLi Zefan 1345b83cc969SLi Zefan if (vol_args->flags & 1346b83cc969SLi Zefan ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) { 1347b83cc969SLi Zefan ret = -EOPNOTSUPP; 1348fdfb1e4fSLi Zefan goto out; 1349fdfb1e4fSLi Zefan } 1350fdfb1e4fSLi Zefan 1351fa0d2b9bSLi Zefan if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC) 1352fdfb1e4fSLi Zefan ptr = &transid; 1353b83cc969SLi Zefan if (vol_args->flags & BTRFS_SUBVOL_RDONLY) 1354b83cc969SLi Zefan readonly = true; 135575eaa0e2SSage Weil 1356fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 1357b83cc969SLi Zefan vol_args->fd, subvol, 1358b83cc969SLi Zefan ptr, readonly); 135975eaa0e2SSage Weil 1360fdfb1e4fSLi Zefan if (ret == 0 && ptr && 136175eaa0e2SSage Weil copy_to_user(arg + 1362fdfb1e4fSLi Zefan offsetof(struct btrfs_ioctl_vol_args_v2, 1363fdfb1e4fSLi Zefan transid), ptr, sizeof(*ptr))) 136475eaa0e2SSage Weil ret = -EFAULT; 1365fdfb1e4fSLi Zefan out: 1366f46b5a66SChristoph Hellwig kfree(vol_args); 1367f46b5a66SChristoph Hellwig return ret; 1368f46b5a66SChristoph Hellwig } 1369f46b5a66SChristoph Hellwig 13700caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_getflags(struct file *file, 13710caa102dSLi Zefan void __user *arg) 13720caa102dSLi Zefan { 13730caa102dSLi Zefan struct inode *inode = fdentry(file)->d_inode; 13740caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 13750caa102dSLi Zefan int ret = 0; 13760caa102dSLi Zefan u64 flags = 0; 13770caa102dSLi Zefan 137833345d01SLi Zefan if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) 13790caa102dSLi Zefan return -EINVAL; 13800caa102dSLi Zefan 13810caa102dSLi Zefan down_read(&root->fs_info->subvol_sem); 13820caa102dSLi Zefan if (btrfs_root_readonly(root)) 13830caa102dSLi Zefan flags |= BTRFS_SUBVOL_RDONLY; 13840caa102dSLi Zefan up_read(&root->fs_info->subvol_sem); 13850caa102dSLi Zefan 13860caa102dSLi Zefan if (copy_to_user(arg, &flags, sizeof(flags))) 13870caa102dSLi Zefan ret = -EFAULT; 13880caa102dSLi Zefan 13890caa102dSLi Zefan return ret; 13900caa102dSLi Zefan } 13910caa102dSLi Zefan 13920caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_setflags(struct file *file, 13930caa102dSLi Zefan void __user *arg) 13940caa102dSLi Zefan { 13950caa102dSLi Zefan struct inode *inode = fdentry(file)->d_inode; 13960caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 13970caa102dSLi Zefan struct btrfs_trans_handle *trans; 13980caa102dSLi Zefan u64 root_flags; 13990caa102dSLi Zefan u64 flags; 14000caa102dSLi Zefan int ret = 0; 14010caa102dSLi Zefan 14020caa102dSLi Zefan if (root->fs_info->sb->s_flags & MS_RDONLY) 14030caa102dSLi Zefan return -EROFS; 14040caa102dSLi Zefan 140533345d01SLi Zefan if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) 14060caa102dSLi Zefan return -EINVAL; 14070caa102dSLi Zefan 14080caa102dSLi Zefan if (copy_from_user(&flags, arg, sizeof(flags))) 14090caa102dSLi Zefan return -EFAULT; 14100caa102dSLi Zefan 1411b4dc2b8cSLi Zefan if (flags & BTRFS_SUBVOL_CREATE_ASYNC) 14120caa102dSLi Zefan return -EINVAL; 14130caa102dSLi Zefan 14140caa102dSLi Zefan if (flags & ~BTRFS_SUBVOL_RDONLY) 14150caa102dSLi Zefan return -EOPNOTSUPP; 14160caa102dSLi Zefan 14172e149670SSerge E. Hallyn if (!inode_owner_or_capable(inode)) 1418b4dc2b8cSLi Zefan return -EACCES; 1419b4dc2b8cSLi Zefan 14200caa102dSLi Zefan down_write(&root->fs_info->subvol_sem); 14210caa102dSLi Zefan 14220caa102dSLi Zefan /* nothing to do */ 14230caa102dSLi Zefan if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) 14240caa102dSLi Zefan goto out; 14250caa102dSLi Zefan 14260caa102dSLi Zefan root_flags = btrfs_root_flags(&root->root_item); 14270caa102dSLi Zefan if (flags & BTRFS_SUBVOL_RDONLY) 14280caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 14290caa102dSLi Zefan root_flags | BTRFS_ROOT_SUBVOL_RDONLY); 14300caa102dSLi Zefan else 14310caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 14320caa102dSLi Zefan root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); 14330caa102dSLi Zefan 14340caa102dSLi Zefan trans = btrfs_start_transaction(root, 1); 14350caa102dSLi Zefan if (IS_ERR(trans)) { 14360caa102dSLi Zefan ret = PTR_ERR(trans); 14370caa102dSLi Zefan goto out_reset; 14380caa102dSLi Zefan } 14390caa102dSLi Zefan 1440b4dc2b8cSLi Zefan ret = btrfs_update_root(trans, root->fs_info->tree_root, 14410caa102dSLi Zefan &root->root_key, &root->root_item); 14420caa102dSLi Zefan 14430caa102dSLi Zefan btrfs_commit_transaction(trans, root); 14440caa102dSLi Zefan out_reset: 14450caa102dSLi Zefan if (ret) 14460caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, root_flags); 14470caa102dSLi Zefan out: 14480caa102dSLi Zefan up_write(&root->fs_info->subvol_sem); 14490caa102dSLi Zefan return ret; 14500caa102dSLi Zefan } 14510caa102dSLi Zefan 145276dda93cSYan, Zheng /* 145376dda93cSYan, Zheng * helper to check if the subvolume references other subvolumes 145476dda93cSYan, Zheng */ 145576dda93cSYan, Zheng static noinline int may_destroy_subvol(struct btrfs_root *root) 145676dda93cSYan, Zheng { 145776dda93cSYan, Zheng struct btrfs_path *path; 145876dda93cSYan, Zheng struct btrfs_key key; 145976dda93cSYan, Zheng int ret; 146076dda93cSYan, Zheng 146176dda93cSYan, Zheng path = btrfs_alloc_path(); 146276dda93cSYan, Zheng if (!path) 146376dda93cSYan, Zheng return -ENOMEM; 146476dda93cSYan, Zheng 146576dda93cSYan, Zheng key.objectid = root->root_key.objectid; 146676dda93cSYan, Zheng key.type = BTRFS_ROOT_REF_KEY; 146776dda93cSYan, Zheng key.offset = (u64)-1; 146876dda93cSYan, Zheng 146976dda93cSYan, Zheng ret = btrfs_search_slot(NULL, root->fs_info->tree_root, 147076dda93cSYan, Zheng &key, path, 0, 0); 147176dda93cSYan, Zheng if (ret < 0) 147276dda93cSYan, Zheng goto out; 147376dda93cSYan, Zheng BUG_ON(ret == 0); 147476dda93cSYan, Zheng 147576dda93cSYan, Zheng ret = 0; 147676dda93cSYan, Zheng if (path->slots[0] > 0) { 147776dda93cSYan, Zheng path->slots[0]--; 147876dda93cSYan, Zheng btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 147976dda93cSYan, Zheng if (key.objectid == root->root_key.objectid && 148076dda93cSYan, Zheng key.type == BTRFS_ROOT_REF_KEY) 148176dda93cSYan, Zheng ret = -ENOTEMPTY; 148276dda93cSYan, Zheng } 148376dda93cSYan, Zheng out: 148476dda93cSYan, Zheng btrfs_free_path(path); 148576dda93cSYan, Zheng return ret; 148676dda93cSYan, Zheng } 148776dda93cSYan, Zheng 1488ac8e9819SChris Mason static noinline int key_in_sk(struct btrfs_key *key, 1489ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk) 1490ac8e9819SChris Mason { 1491abc6e134SChris Mason struct btrfs_key test; 1492abc6e134SChris Mason int ret; 1493abc6e134SChris Mason 1494abc6e134SChris Mason test.objectid = sk->min_objectid; 1495abc6e134SChris Mason test.type = sk->min_type; 1496abc6e134SChris Mason test.offset = sk->min_offset; 1497abc6e134SChris Mason 1498abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1499abc6e134SChris Mason if (ret < 0) 1500ac8e9819SChris Mason return 0; 1501abc6e134SChris Mason 1502abc6e134SChris Mason test.objectid = sk->max_objectid; 1503abc6e134SChris Mason test.type = sk->max_type; 1504abc6e134SChris Mason test.offset = sk->max_offset; 1505abc6e134SChris Mason 1506abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1507abc6e134SChris Mason if (ret > 0) 1508ac8e9819SChris Mason return 0; 1509ac8e9819SChris Mason return 1; 1510ac8e9819SChris Mason } 1511ac8e9819SChris Mason 1512ac8e9819SChris Mason static noinline int copy_to_sk(struct btrfs_root *root, 1513ac8e9819SChris Mason struct btrfs_path *path, 1514ac8e9819SChris Mason struct btrfs_key *key, 1515ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk, 1516ac8e9819SChris Mason char *buf, 1517ac8e9819SChris Mason unsigned long *sk_offset, 1518ac8e9819SChris Mason int *num_found) 1519ac8e9819SChris Mason { 1520ac8e9819SChris Mason u64 found_transid; 1521ac8e9819SChris Mason struct extent_buffer *leaf; 1522ac8e9819SChris Mason struct btrfs_ioctl_search_header sh; 1523ac8e9819SChris Mason unsigned long item_off; 1524ac8e9819SChris Mason unsigned long item_len; 1525ac8e9819SChris Mason int nritems; 1526ac8e9819SChris Mason int i; 1527ac8e9819SChris Mason int slot; 1528ac8e9819SChris Mason int ret = 0; 1529ac8e9819SChris Mason 1530ac8e9819SChris Mason leaf = path->nodes[0]; 1531ac8e9819SChris Mason slot = path->slots[0]; 1532ac8e9819SChris Mason nritems = btrfs_header_nritems(leaf); 1533ac8e9819SChris Mason 1534ac8e9819SChris Mason if (btrfs_header_generation(leaf) > sk->max_transid) { 1535ac8e9819SChris Mason i = nritems; 1536ac8e9819SChris Mason goto advance_key; 1537ac8e9819SChris Mason } 1538ac8e9819SChris Mason found_transid = btrfs_header_generation(leaf); 1539ac8e9819SChris Mason 1540ac8e9819SChris Mason for (i = slot; i < nritems; i++) { 1541ac8e9819SChris Mason item_off = btrfs_item_ptr_offset(leaf, i); 1542ac8e9819SChris Mason item_len = btrfs_item_size_nr(leaf, i); 1543ac8e9819SChris Mason 1544ac8e9819SChris Mason if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE) 1545ac8e9819SChris Mason item_len = 0; 1546ac8e9819SChris Mason 1547ac8e9819SChris Mason if (sizeof(sh) + item_len + *sk_offset > 1548ac8e9819SChris Mason BTRFS_SEARCH_ARGS_BUFSIZE) { 1549ac8e9819SChris Mason ret = 1; 1550ac8e9819SChris Mason goto overflow; 1551ac8e9819SChris Mason } 1552ac8e9819SChris Mason 1553ac8e9819SChris Mason btrfs_item_key_to_cpu(leaf, key, i); 1554ac8e9819SChris Mason if (!key_in_sk(key, sk)) 1555ac8e9819SChris Mason continue; 1556ac8e9819SChris Mason 1557ac8e9819SChris Mason sh.objectid = key->objectid; 1558ac8e9819SChris Mason sh.offset = key->offset; 1559ac8e9819SChris Mason sh.type = key->type; 1560ac8e9819SChris Mason sh.len = item_len; 1561ac8e9819SChris Mason sh.transid = found_transid; 1562ac8e9819SChris Mason 1563ac8e9819SChris Mason /* copy search result header */ 1564ac8e9819SChris Mason memcpy(buf + *sk_offset, &sh, sizeof(sh)); 1565ac8e9819SChris Mason *sk_offset += sizeof(sh); 1566ac8e9819SChris Mason 1567ac8e9819SChris Mason if (item_len) { 1568ac8e9819SChris Mason char *p = buf + *sk_offset; 1569ac8e9819SChris Mason /* copy the item */ 1570ac8e9819SChris Mason read_extent_buffer(leaf, p, 1571ac8e9819SChris Mason item_off, item_len); 1572ac8e9819SChris Mason *sk_offset += item_len; 1573ac8e9819SChris Mason } 1574e2156867SHugo Mills (*num_found)++; 1575ac8e9819SChris Mason 1576ac8e9819SChris Mason if (*num_found >= sk->nr_items) 1577ac8e9819SChris Mason break; 1578ac8e9819SChris Mason } 1579ac8e9819SChris Mason advance_key: 1580ac8e9819SChris Mason ret = 0; 1581abc6e134SChris Mason if (key->offset < (u64)-1 && key->offset < sk->max_offset) 1582abc6e134SChris Mason key->offset++; 1583abc6e134SChris Mason else if (key->type < (u8)-1 && key->type < sk->max_type) { 1584abc6e134SChris Mason key->offset = 0; 1585abc6e134SChris Mason key->type++; 1586abc6e134SChris Mason } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) { 1587abc6e134SChris Mason key->offset = 0; 1588abc6e134SChris Mason key->type = 0; 1589abc6e134SChris Mason key->objectid++; 1590abc6e134SChris Mason } else 1591abc6e134SChris Mason ret = 1; 1592ac8e9819SChris Mason overflow: 1593ac8e9819SChris Mason return ret; 1594ac8e9819SChris Mason } 1595ac8e9819SChris Mason 1596ac8e9819SChris Mason static noinline int search_ioctl(struct inode *inode, 1597ac8e9819SChris Mason struct btrfs_ioctl_search_args *args) 1598ac8e9819SChris Mason { 1599ac8e9819SChris Mason struct btrfs_root *root; 1600ac8e9819SChris Mason struct btrfs_key key; 1601ac8e9819SChris Mason struct btrfs_key max_key; 1602ac8e9819SChris Mason struct btrfs_path *path; 1603ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk = &args->key; 1604ac8e9819SChris Mason struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info; 1605ac8e9819SChris Mason int ret; 1606ac8e9819SChris Mason int num_found = 0; 1607ac8e9819SChris Mason unsigned long sk_offset = 0; 1608ac8e9819SChris Mason 1609ac8e9819SChris Mason path = btrfs_alloc_path(); 1610ac8e9819SChris Mason if (!path) 1611ac8e9819SChris Mason return -ENOMEM; 1612ac8e9819SChris Mason 1613ac8e9819SChris Mason if (sk->tree_id == 0) { 1614ac8e9819SChris Mason /* search the root of the inode that was passed */ 1615ac8e9819SChris Mason root = BTRFS_I(inode)->root; 1616ac8e9819SChris Mason } else { 1617ac8e9819SChris Mason key.objectid = sk->tree_id; 1618ac8e9819SChris Mason key.type = BTRFS_ROOT_ITEM_KEY; 1619ac8e9819SChris Mason key.offset = (u64)-1; 1620ac8e9819SChris Mason root = btrfs_read_fs_root_no_name(info, &key); 1621ac8e9819SChris Mason if (IS_ERR(root)) { 1622ac8e9819SChris Mason printk(KERN_ERR "could not find root %llu\n", 1623ac8e9819SChris Mason sk->tree_id); 1624ac8e9819SChris Mason btrfs_free_path(path); 1625ac8e9819SChris Mason return -ENOENT; 1626ac8e9819SChris Mason } 1627ac8e9819SChris Mason } 1628ac8e9819SChris Mason 1629ac8e9819SChris Mason key.objectid = sk->min_objectid; 1630ac8e9819SChris Mason key.type = sk->min_type; 1631ac8e9819SChris Mason key.offset = sk->min_offset; 1632ac8e9819SChris Mason 1633ac8e9819SChris Mason max_key.objectid = sk->max_objectid; 1634ac8e9819SChris Mason max_key.type = sk->max_type; 1635ac8e9819SChris Mason max_key.offset = sk->max_offset; 1636ac8e9819SChris Mason 1637ac8e9819SChris Mason path->keep_locks = 1; 1638ac8e9819SChris Mason 1639ac8e9819SChris Mason while(1) { 1640ac8e9819SChris Mason ret = btrfs_search_forward(root, &key, &max_key, path, 0, 1641ac8e9819SChris Mason sk->min_transid); 1642ac8e9819SChris Mason if (ret != 0) { 1643ac8e9819SChris Mason if (ret > 0) 1644ac8e9819SChris Mason ret = 0; 1645ac8e9819SChris Mason goto err; 1646ac8e9819SChris Mason } 1647ac8e9819SChris Mason ret = copy_to_sk(root, path, &key, sk, args->buf, 1648ac8e9819SChris Mason &sk_offset, &num_found); 1649b3b4aa74SDavid Sterba btrfs_release_path(path); 1650ac8e9819SChris Mason if (ret || num_found >= sk->nr_items) 1651ac8e9819SChris Mason break; 1652ac8e9819SChris Mason 1653ac8e9819SChris Mason } 1654ac8e9819SChris Mason ret = 0; 1655ac8e9819SChris Mason err: 1656ac8e9819SChris Mason sk->nr_items = num_found; 1657ac8e9819SChris Mason btrfs_free_path(path); 1658ac8e9819SChris Mason return ret; 1659ac8e9819SChris Mason } 1660ac8e9819SChris Mason 1661ac8e9819SChris Mason static noinline int btrfs_ioctl_tree_search(struct file *file, 1662ac8e9819SChris Mason void __user *argp) 1663ac8e9819SChris Mason { 1664ac8e9819SChris Mason struct btrfs_ioctl_search_args *args; 1665ac8e9819SChris Mason struct inode *inode; 1666ac8e9819SChris Mason int ret; 1667ac8e9819SChris Mason 1668ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 1669ac8e9819SChris Mason return -EPERM; 1670ac8e9819SChris Mason 16712354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 16722354d08fSJulia Lawall if (IS_ERR(args)) 16732354d08fSJulia Lawall return PTR_ERR(args); 1674ac8e9819SChris Mason 1675ac8e9819SChris Mason inode = fdentry(file)->d_inode; 1676ac8e9819SChris Mason ret = search_ioctl(inode, args); 1677ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 1678ac8e9819SChris Mason ret = -EFAULT; 1679ac8e9819SChris Mason kfree(args); 1680ac8e9819SChris Mason return ret; 1681ac8e9819SChris Mason } 1682ac8e9819SChris Mason 168398d377a0STARUISI Hiroaki /* 1684ac8e9819SChris Mason * Search INODE_REFs to identify path name of 'dirid' directory 1685ac8e9819SChris Mason * in a 'tree_id' tree. and sets path name to 'name'. 168698d377a0STARUISI Hiroaki */ 168798d377a0STARUISI Hiroaki static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, 168898d377a0STARUISI Hiroaki u64 tree_id, u64 dirid, char *name) 168998d377a0STARUISI Hiroaki { 169098d377a0STARUISI Hiroaki struct btrfs_root *root; 169198d377a0STARUISI Hiroaki struct btrfs_key key; 1692ac8e9819SChris Mason char *ptr; 169398d377a0STARUISI Hiroaki int ret = -1; 169498d377a0STARUISI Hiroaki int slot; 169598d377a0STARUISI Hiroaki int len; 169698d377a0STARUISI Hiroaki int total_len = 0; 169798d377a0STARUISI Hiroaki struct btrfs_inode_ref *iref; 169898d377a0STARUISI Hiroaki struct extent_buffer *l; 169998d377a0STARUISI Hiroaki struct btrfs_path *path; 170098d377a0STARUISI Hiroaki 170198d377a0STARUISI Hiroaki if (dirid == BTRFS_FIRST_FREE_OBJECTID) { 170298d377a0STARUISI Hiroaki name[0]='\0'; 170398d377a0STARUISI Hiroaki return 0; 170498d377a0STARUISI Hiroaki } 170598d377a0STARUISI Hiroaki 170698d377a0STARUISI Hiroaki path = btrfs_alloc_path(); 170798d377a0STARUISI Hiroaki if (!path) 170898d377a0STARUISI Hiroaki return -ENOMEM; 170998d377a0STARUISI Hiroaki 1710ac8e9819SChris Mason ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX]; 171198d377a0STARUISI Hiroaki 171298d377a0STARUISI Hiroaki key.objectid = tree_id; 171398d377a0STARUISI Hiroaki key.type = BTRFS_ROOT_ITEM_KEY; 171498d377a0STARUISI Hiroaki key.offset = (u64)-1; 171598d377a0STARUISI Hiroaki root = btrfs_read_fs_root_no_name(info, &key); 171698d377a0STARUISI Hiroaki if (IS_ERR(root)) { 171798d377a0STARUISI Hiroaki printk(KERN_ERR "could not find root %llu\n", tree_id); 17188ad6fcabSChris Mason ret = -ENOENT; 17198ad6fcabSChris Mason goto out; 172098d377a0STARUISI Hiroaki } 172198d377a0STARUISI Hiroaki 172298d377a0STARUISI Hiroaki key.objectid = dirid; 172398d377a0STARUISI Hiroaki key.type = BTRFS_INODE_REF_KEY; 17248ad6fcabSChris Mason key.offset = (u64)-1; 172598d377a0STARUISI Hiroaki 172698d377a0STARUISI Hiroaki while(1) { 172798d377a0STARUISI Hiroaki ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 172898d377a0STARUISI Hiroaki if (ret < 0) 172998d377a0STARUISI Hiroaki goto out; 173098d377a0STARUISI Hiroaki 173198d377a0STARUISI Hiroaki l = path->nodes[0]; 173298d377a0STARUISI Hiroaki slot = path->slots[0]; 17338ad6fcabSChris Mason if (ret > 0 && slot > 0) 17348ad6fcabSChris Mason slot--; 173598d377a0STARUISI Hiroaki btrfs_item_key_to_cpu(l, &key, slot); 173698d377a0STARUISI Hiroaki 173798d377a0STARUISI Hiroaki if (ret > 0 && (key.objectid != dirid || 1738ac8e9819SChris Mason key.type != BTRFS_INODE_REF_KEY)) { 1739ac8e9819SChris Mason ret = -ENOENT; 174098d377a0STARUISI Hiroaki goto out; 1741ac8e9819SChris Mason } 174298d377a0STARUISI Hiroaki 174398d377a0STARUISI Hiroaki iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); 174498d377a0STARUISI Hiroaki len = btrfs_inode_ref_name_len(l, iref); 174598d377a0STARUISI Hiroaki ptr -= len + 1; 174698d377a0STARUISI Hiroaki total_len += len + 1; 1747ac8e9819SChris Mason if (ptr < name) 174898d377a0STARUISI Hiroaki goto out; 174998d377a0STARUISI Hiroaki 175098d377a0STARUISI Hiroaki *(ptr + len) = '/'; 175198d377a0STARUISI Hiroaki read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len); 175298d377a0STARUISI Hiroaki 175398d377a0STARUISI Hiroaki if (key.offset == BTRFS_FIRST_FREE_OBJECTID) 175498d377a0STARUISI Hiroaki break; 175598d377a0STARUISI Hiroaki 1756b3b4aa74SDavid Sterba btrfs_release_path(path); 175798d377a0STARUISI Hiroaki key.objectid = key.offset; 17588ad6fcabSChris Mason key.offset = (u64)-1; 175998d377a0STARUISI Hiroaki dirid = key.objectid; 176098d377a0STARUISI Hiroaki } 1761ac8e9819SChris Mason if (ptr < name) 176298d377a0STARUISI Hiroaki goto out; 176377906a50SLi Zefan memmove(name, ptr, total_len); 176498d377a0STARUISI Hiroaki name[total_len]='\0'; 176598d377a0STARUISI Hiroaki ret = 0; 176698d377a0STARUISI Hiroaki out: 176798d377a0STARUISI Hiroaki btrfs_free_path(path); 1768ac8e9819SChris Mason return ret; 1769ac8e9819SChris Mason } 1770ac8e9819SChris Mason 1771ac8e9819SChris Mason static noinline int btrfs_ioctl_ino_lookup(struct file *file, 1772ac8e9819SChris Mason void __user *argp) 1773ac8e9819SChris Mason { 1774ac8e9819SChris Mason struct btrfs_ioctl_ino_lookup_args *args; 1775ac8e9819SChris Mason struct inode *inode; 1776ac8e9819SChris Mason int ret; 1777ac8e9819SChris Mason 1778ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 1779ac8e9819SChris Mason return -EPERM; 1780ac8e9819SChris Mason 17812354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 17822354d08fSJulia Lawall if (IS_ERR(args)) 17832354d08fSJulia Lawall return PTR_ERR(args); 1784c2b96929SDan Carpenter 1785ac8e9819SChris Mason inode = fdentry(file)->d_inode; 1786ac8e9819SChris Mason 17871b53ac4dSChris Mason if (args->treeid == 0) 17881b53ac4dSChris Mason args->treeid = BTRFS_I(inode)->root->root_key.objectid; 17891b53ac4dSChris Mason 1790ac8e9819SChris Mason ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, 1791ac8e9819SChris Mason args->treeid, args->objectid, 1792ac8e9819SChris Mason args->name); 1793ac8e9819SChris Mason 1794ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 1795ac8e9819SChris Mason ret = -EFAULT; 1796ac8e9819SChris Mason 1797ac8e9819SChris Mason kfree(args); 179898d377a0STARUISI Hiroaki return ret; 179998d377a0STARUISI Hiroaki } 180098d377a0STARUISI Hiroaki 180176dda93cSYan, Zheng static noinline int btrfs_ioctl_snap_destroy(struct file *file, 180276dda93cSYan, Zheng void __user *arg) 180376dda93cSYan, Zheng { 180476dda93cSYan, Zheng struct dentry *parent = fdentry(file); 180576dda93cSYan, Zheng struct dentry *dentry; 180676dda93cSYan, Zheng struct inode *dir = parent->d_inode; 180776dda93cSYan, Zheng struct inode *inode; 180876dda93cSYan, Zheng struct btrfs_root *root = BTRFS_I(dir)->root; 180976dda93cSYan, Zheng struct btrfs_root *dest = NULL; 181076dda93cSYan, Zheng struct btrfs_ioctl_vol_args *vol_args; 181176dda93cSYan, Zheng struct btrfs_trans_handle *trans; 181276dda93cSYan, Zheng int namelen; 181376dda93cSYan, Zheng int ret; 181476dda93cSYan, Zheng int err = 0; 181576dda93cSYan, Zheng 181676dda93cSYan, Zheng vol_args = memdup_user(arg, sizeof(*vol_args)); 181776dda93cSYan, Zheng if (IS_ERR(vol_args)) 181876dda93cSYan, Zheng return PTR_ERR(vol_args); 181976dda93cSYan, Zheng 182076dda93cSYan, Zheng vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 182176dda93cSYan, Zheng namelen = strlen(vol_args->name); 182276dda93cSYan, Zheng if (strchr(vol_args->name, '/') || 182376dda93cSYan, Zheng strncmp(vol_args->name, "..", namelen) == 0) { 182476dda93cSYan, Zheng err = -EINVAL; 182576dda93cSYan, Zheng goto out; 182676dda93cSYan, Zheng } 182776dda93cSYan, Zheng 182876dda93cSYan, Zheng err = mnt_want_write(file->f_path.mnt); 182976dda93cSYan, Zheng if (err) 183076dda93cSYan, Zheng goto out; 183176dda93cSYan, Zheng 183276dda93cSYan, Zheng mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT); 183376dda93cSYan, Zheng dentry = lookup_one_len(vol_args->name, parent, namelen); 183476dda93cSYan, Zheng if (IS_ERR(dentry)) { 183576dda93cSYan, Zheng err = PTR_ERR(dentry); 183676dda93cSYan, Zheng goto out_unlock_dir; 183776dda93cSYan, Zheng } 183876dda93cSYan, Zheng 183976dda93cSYan, Zheng if (!dentry->d_inode) { 184076dda93cSYan, Zheng err = -ENOENT; 184176dda93cSYan, Zheng goto out_dput; 184276dda93cSYan, Zheng } 184376dda93cSYan, Zheng 184476dda93cSYan, Zheng inode = dentry->d_inode; 18454260f7c7SSage Weil dest = BTRFS_I(inode)->root; 18464260f7c7SSage Weil if (!capable(CAP_SYS_ADMIN)){ 18474260f7c7SSage Weil /* 18484260f7c7SSage Weil * Regular user. Only allow this with a special mount 18494260f7c7SSage Weil * option, when the user has write+exec access to the 18504260f7c7SSage Weil * subvol root, and when rmdir(2) would have been 18514260f7c7SSage Weil * allowed. 18524260f7c7SSage Weil * 18534260f7c7SSage Weil * Note that this is _not_ check that the subvol is 18544260f7c7SSage Weil * empty or doesn't contain data that we wouldn't 18554260f7c7SSage Weil * otherwise be able to delete. 18564260f7c7SSage Weil * 18574260f7c7SSage Weil * Users who want to delete empty subvols should try 18584260f7c7SSage Weil * rmdir(2). 18594260f7c7SSage Weil */ 18604260f7c7SSage Weil err = -EPERM; 18614260f7c7SSage Weil if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED)) 18624260f7c7SSage Weil goto out_dput; 18634260f7c7SSage Weil 18644260f7c7SSage Weil /* 18654260f7c7SSage Weil * Do not allow deletion if the parent dir is the same 18664260f7c7SSage Weil * as the dir to be deleted. That means the ioctl 18674260f7c7SSage Weil * must be called on the dentry referencing the root 18684260f7c7SSage Weil * of the subvol, not a random directory contained 18694260f7c7SSage Weil * within it. 18704260f7c7SSage Weil */ 18714260f7c7SSage Weil err = -EINVAL; 18724260f7c7SSage Weil if (root == dest) 18734260f7c7SSage Weil goto out_dput; 18744260f7c7SSage Weil 18754260f7c7SSage Weil err = inode_permission(inode, MAY_WRITE | MAY_EXEC); 18764260f7c7SSage Weil if (err) 18774260f7c7SSage Weil goto out_dput; 18784260f7c7SSage Weil 18794260f7c7SSage Weil /* check if subvolume may be deleted by a non-root user */ 18804260f7c7SSage Weil err = btrfs_may_delete(dir, dentry, 1); 18814260f7c7SSage Weil if (err) 18824260f7c7SSage Weil goto out_dput; 18834260f7c7SSage Weil } 18844260f7c7SSage Weil 188533345d01SLi Zefan if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { 188676dda93cSYan, Zheng err = -EINVAL; 188776dda93cSYan, Zheng goto out_dput; 188876dda93cSYan, Zheng } 188976dda93cSYan, Zheng 189076dda93cSYan, Zheng mutex_lock(&inode->i_mutex); 189176dda93cSYan, Zheng err = d_invalidate(dentry); 189276dda93cSYan, Zheng if (err) 189376dda93cSYan, Zheng goto out_unlock; 189476dda93cSYan, Zheng 189576dda93cSYan, Zheng down_write(&root->fs_info->subvol_sem); 189676dda93cSYan, Zheng 189776dda93cSYan, Zheng err = may_destroy_subvol(dest); 189876dda93cSYan, Zheng if (err) 189976dda93cSYan, Zheng goto out_up_write; 190076dda93cSYan, Zheng 1901a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 1902a22285a6SYan, Zheng if (IS_ERR(trans)) { 1903a22285a6SYan, Zheng err = PTR_ERR(trans); 1904d327099aSDan Carpenter goto out_up_write; 1905a22285a6SYan, Zheng } 1906a22285a6SYan, Zheng trans->block_rsv = &root->fs_info->global_block_rsv; 1907a22285a6SYan, Zheng 190876dda93cSYan, Zheng ret = btrfs_unlink_subvol(trans, root, dir, 190976dda93cSYan, Zheng dest->root_key.objectid, 191076dda93cSYan, Zheng dentry->d_name.name, 191176dda93cSYan, Zheng dentry->d_name.len); 191276dda93cSYan, Zheng BUG_ON(ret); 191376dda93cSYan, Zheng 191476dda93cSYan, Zheng btrfs_record_root_in_trans(trans, dest); 191576dda93cSYan, Zheng 191676dda93cSYan, Zheng memset(&dest->root_item.drop_progress, 0, 191776dda93cSYan, Zheng sizeof(dest->root_item.drop_progress)); 191876dda93cSYan, Zheng dest->root_item.drop_level = 0; 191976dda93cSYan, Zheng btrfs_set_root_refs(&dest->root_item, 0); 192076dda93cSYan, Zheng 1921d68fc57bSYan, Zheng if (!xchg(&dest->orphan_item_inserted, 1)) { 192276dda93cSYan, Zheng ret = btrfs_insert_orphan_item(trans, 192376dda93cSYan, Zheng root->fs_info->tree_root, 192476dda93cSYan, Zheng dest->root_key.objectid); 192576dda93cSYan, Zheng BUG_ON(ret); 1926d68fc57bSYan, Zheng } 192776dda93cSYan, Zheng 1928531cb13fSSage Weil ret = btrfs_end_transaction(trans, root); 192976dda93cSYan, Zheng BUG_ON(ret); 193076dda93cSYan, Zheng inode->i_flags |= S_DEAD; 193176dda93cSYan, Zheng out_up_write: 193276dda93cSYan, Zheng up_write(&root->fs_info->subvol_sem); 193376dda93cSYan, Zheng out_unlock: 193476dda93cSYan, Zheng mutex_unlock(&inode->i_mutex); 193576dda93cSYan, Zheng if (!err) { 1936efefb143SYan, Zheng shrink_dcache_sb(root->fs_info->sb); 193776dda93cSYan, Zheng btrfs_invalidate_inodes(dest); 193876dda93cSYan, Zheng d_delete(dentry); 193976dda93cSYan, Zheng } 194076dda93cSYan, Zheng out_dput: 194176dda93cSYan, Zheng dput(dentry); 194276dda93cSYan, Zheng out_unlock_dir: 194376dda93cSYan, Zheng mutex_unlock(&dir->i_mutex); 194476dda93cSYan, Zheng mnt_drop_write(file->f_path.mnt); 194576dda93cSYan, Zheng out: 194676dda93cSYan, Zheng kfree(vol_args); 194776dda93cSYan, Zheng return err; 194876dda93cSYan, Zheng } 194976dda93cSYan, Zheng 19501e701a32SChris Mason static int btrfs_ioctl_defrag(struct file *file, void __user *argp) 1951f46b5a66SChristoph Hellwig { 1952f46b5a66SChristoph Hellwig struct inode *inode = fdentry(file)->d_inode; 1953f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 19541e701a32SChris Mason struct btrfs_ioctl_defrag_range_args *range; 1955c146afadSYan Zheng int ret; 1956c146afadSYan Zheng 1957b83cc969SLi Zefan if (btrfs_root_readonly(root)) 1958b83cc969SLi Zefan return -EROFS; 1959b83cc969SLi Zefan 1960c146afadSYan Zheng ret = mnt_want_write(file->f_path.mnt); 1961c146afadSYan Zheng if (ret) 1962c146afadSYan Zheng return ret; 1963f46b5a66SChristoph Hellwig 1964f46b5a66SChristoph Hellwig switch (inode->i_mode & S_IFMT) { 1965f46b5a66SChristoph Hellwig case S_IFDIR: 1966e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) { 1967e441d54dSChris Mason ret = -EPERM; 1968e441d54dSChris Mason goto out; 1969e441d54dSChris Mason } 19708929ecfaSYan, Zheng ret = btrfs_defrag_root(root, 0); 19718929ecfaSYan, Zheng if (ret) 19728929ecfaSYan, Zheng goto out; 19738929ecfaSYan, Zheng ret = btrfs_defrag_root(root->fs_info->extent_root, 0); 1974f46b5a66SChristoph Hellwig break; 1975f46b5a66SChristoph Hellwig case S_IFREG: 1976e441d54dSChris Mason if (!(file->f_mode & FMODE_WRITE)) { 1977e441d54dSChris Mason ret = -EINVAL; 1978e441d54dSChris Mason goto out; 1979e441d54dSChris Mason } 19801e701a32SChris Mason 19811e701a32SChris Mason range = kzalloc(sizeof(*range), GFP_KERNEL); 19821e701a32SChris Mason if (!range) { 19831e701a32SChris Mason ret = -ENOMEM; 19841e701a32SChris Mason goto out; 19851e701a32SChris Mason } 19861e701a32SChris Mason 19871e701a32SChris Mason if (argp) { 19881e701a32SChris Mason if (copy_from_user(range, argp, 19891e701a32SChris Mason sizeof(*range))) { 19901e701a32SChris Mason ret = -EFAULT; 19911e701a32SChris Mason kfree(range); 1992683be16eSDan Carpenter goto out; 19931e701a32SChris Mason } 19941e701a32SChris Mason /* compression requires us to start the IO */ 19951e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 19961e701a32SChris Mason range->flags |= BTRFS_DEFRAG_RANGE_START_IO; 19971e701a32SChris Mason range->extent_thresh = (u32)-1; 19981e701a32SChris Mason } 19991e701a32SChris Mason } else { 20001e701a32SChris Mason /* the rest are all set to zero by kzalloc */ 20011e701a32SChris Mason range->len = (u64)-1; 20021e701a32SChris Mason } 20034cb5300bSChris Mason ret = btrfs_defrag_file(fdentry(file)->d_inode, file, 20044cb5300bSChris Mason range, 0, 0); 20054cb5300bSChris Mason if (ret > 0) 20064cb5300bSChris Mason ret = 0; 20071e701a32SChris Mason kfree(range); 2008f46b5a66SChristoph Hellwig break; 20098929ecfaSYan, Zheng default: 20108929ecfaSYan, Zheng ret = -EINVAL; 2011f46b5a66SChristoph Hellwig } 2012e441d54dSChris Mason out: 2013ab67b7c1SYan Zheng mnt_drop_write(file->f_path.mnt); 2014e441d54dSChris Mason return ret; 2015f46b5a66SChristoph Hellwig } 2016f46b5a66SChristoph Hellwig 2017b2950863SChristoph Hellwig static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg) 2018f46b5a66SChristoph Hellwig { 2019f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2020f46b5a66SChristoph Hellwig int ret; 2021f46b5a66SChristoph Hellwig 2022e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2023e441d54dSChris Mason return -EPERM; 2024e441d54dSChris Mason 2025dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2026dae7b665SLi Zefan if (IS_ERR(vol_args)) 2027dae7b665SLi Zefan return PTR_ERR(vol_args); 2028f46b5a66SChristoph Hellwig 20295516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 2030f46b5a66SChristoph Hellwig ret = btrfs_init_new_device(root, vol_args->name); 2031f46b5a66SChristoph Hellwig 2032f46b5a66SChristoph Hellwig kfree(vol_args); 2033f46b5a66SChristoph Hellwig return ret; 2034f46b5a66SChristoph Hellwig } 2035f46b5a66SChristoph Hellwig 2036b2950863SChristoph Hellwig static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg) 2037f46b5a66SChristoph Hellwig { 2038f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2039f46b5a66SChristoph Hellwig int ret; 2040f46b5a66SChristoph Hellwig 2041e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2042e441d54dSChris Mason return -EPERM; 2043e441d54dSChris Mason 2044c146afadSYan Zheng if (root->fs_info->sb->s_flags & MS_RDONLY) 2045c146afadSYan Zheng return -EROFS; 2046c146afadSYan Zheng 2047dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2048dae7b665SLi Zefan if (IS_ERR(vol_args)) 2049dae7b665SLi Zefan return PTR_ERR(vol_args); 2050f46b5a66SChristoph Hellwig 20515516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 2052f46b5a66SChristoph Hellwig ret = btrfs_rm_device(root, vol_args->name); 2053f46b5a66SChristoph Hellwig 2054f46b5a66SChristoph Hellwig kfree(vol_args); 2055f46b5a66SChristoph Hellwig return ret; 2056f46b5a66SChristoph Hellwig } 2057f46b5a66SChristoph Hellwig 2058475f6387SJan Schmidt static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg) 2059475f6387SJan Schmidt { 2060027ed2f0SLi Zefan struct btrfs_ioctl_fs_info_args *fi_args; 2061475f6387SJan Schmidt struct btrfs_device *device; 2062475f6387SJan Schmidt struct btrfs_device *next; 2063475f6387SJan Schmidt struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; 2064027ed2f0SLi Zefan int ret = 0; 2065475f6387SJan Schmidt 2066475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 2067475f6387SJan Schmidt return -EPERM; 2068475f6387SJan Schmidt 2069027ed2f0SLi Zefan fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL); 2070027ed2f0SLi Zefan if (!fi_args) 2071027ed2f0SLi Zefan return -ENOMEM; 2072027ed2f0SLi Zefan 2073027ed2f0SLi Zefan fi_args->num_devices = fs_devices->num_devices; 2074027ed2f0SLi Zefan memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid)); 2075475f6387SJan Schmidt 2076475f6387SJan Schmidt mutex_lock(&fs_devices->device_list_mutex); 2077475f6387SJan Schmidt list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) { 2078027ed2f0SLi Zefan if (device->devid > fi_args->max_id) 2079027ed2f0SLi Zefan fi_args->max_id = device->devid; 2080475f6387SJan Schmidt } 2081475f6387SJan Schmidt mutex_unlock(&fs_devices->device_list_mutex); 2082475f6387SJan Schmidt 2083027ed2f0SLi Zefan if (copy_to_user(arg, fi_args, sizeof(*fi_args))) 2084027ed2f0SLi Zefan ret = -EFAULT; 2085475f6387SJan Schmidt 2086027ed2f0SLi Zefan kfree(fi_args); 2087027ed2f0SLi Zefan return ret; 2088475f6387SJan Schmidt } 2089475f6387SJan Schmidt 2090475f6387SJan Schmidt static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg) 2091475f6387SJan Schmidt { 2092475f6387SJan Schmidt struct btrfs_ioctl_dev_info_args *di_args; 2093475f6387SJan Schmidt struct btrfs_device *dev; 2094475f6387SJan Schmidt struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; 2095475f6387SJan Schmidt int ret = 0; 2096475f6387SJan Schmidt char *s_uuid = NULL; 2097475f6387SJan Schmidt char empty_uuid[BTRFS_UUID_SIZE] = {0}; 2098475f6387SJan Schmidt 2099475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 2100475f6387SJan Schmidt return -EPERM; 2101475f6387SJan Schmidt 2102475f6387SJan Schmidt di_args = memdup_user(arg, sizeof(*di_args)); 2103475f6387SJan Schmidt if (IS_ERR(di_args)) 2104475f6387SJan Schmidt return PTR_ERR(di_args); 2105475f6387SJan Schmidt 2106475f6387SJan Schmidt if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0) 2107475f6387SJan Schmidt s_uuid = di_args->uuid; 2108475f6387SJan Schmidt 2109475f6387SJan Schmidt mutex_lock(&fs_devices->device_list_mutex); 2110475f6387SJan Schmidt dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL); 2111475f6387SJan Schmidt mutex_unlock(&fs_devices->device_list_mutex); 2112475f6387SJan Schmidt 2113475f6387SJan Schmidt if (!dev) { 2114475f6387SJan Schmidt ret = -ENODEV; 2115475f6387SJan Schmidt goto out; 2116475f6387SJan Schmidt } 2117475f6387SJan Schmidt 2118475f6387SJan Schmidt di_args->devid = dev->devid; 2119475f6387SJan Schmidt di_args->bytes_used = dev->bytes_used; 2120475f6387SJan Schmidt di_args->total_bytes = dev->total_bytes; 2121475f6387SJan Schmidt memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); 2122475f6387SJan Schmidt strncpy(di_args->path, dev->name, sizeof(di_args->path)); 2123475f6387SJan Schmidt 2124475f6387SJan Schmidt out: 2125475f6387SJan Schmidt if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) 2126475f6387SJan Schmidt ret = -EFAULT; 2127475f6387SJan Schmidt 2128475f6387SJan Schmidt kfree(di_args); 2129475f6387SJan Schmidt return ret; 2130475f6387SJan Schmidt } 2131475f6387SJan Schmidt 213276dda93cSYan, Zheng static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, 2133b2950863SChristoph Hellwig u64 off, u64 olen, u64 destoff) 2134f46b5a66SChristoph Hellwig { 2135f46b5a66SChristoph Hellwig struct inode *inode = fdentry(file)->d_inode; 2136f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 2137f46b5a66SChristoph Hellwig struct file *src_file; 2138f46b5a66SChristoph Hellwig struct inode *src; 2139f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 2140f46b5a66SChristoph Hellwig struct btrfs_path *path; 2141f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 2142ae01a0abSYan Zheng char *buf; 2143ae01a0abSYan Zheng struct btrfs_key key; 2144f46b5a66SChristoph Hellwig u32 nritems; 2145f46b5a66SChristoph Hellwig int slot; 2146ae01a0abSYan Zheng int ret; 2147c5c9cd4dSSage Weil u64 len = olen; 2148c5c9cd4dSSage Weil u64 bs = root->fs_info->sb->s_blocksize; 2149c5c9cd4dSSage Weil u64 hint_byte; 2150d20f7043SChris Mason 2151c5c9cd4dSSage Weil /* 2152c5c9cd4dSSage Weil * TODO: 2153c5c9cd4dSSage Weil * - split compressed inline extents. annoying: we need to 2154c5c9cd4dSSage Weil * decompress into destination's address_space (the file offset 2155c5c9cd4dSSage Weil * may change, so source mapping won't do), then recompress (or 2156c5c9cd4dSSage Weil * otherwise reinsert) a subrange. 2157c5c9cd4dSSage Weil * - allow ranges within the same file to be cloned (provided 2158c5c9cd4dSSage Weil * they don't overlap)? 2159c5c9cd4dSSage Weil */ 2160c5c9cd4dSSage Weil 2161e441d54dSChris Mason /* the destination must be opened for writing */ 21622ebc3464SDan Rosenberg if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND)) 2163e441d54dSChris Mason return -EINVAL; 2164e441d54dSChris Mason 2165b83cc969SLi Zefan if (btrfs_root_readonly(root)) 2166b83cc969SLi Zefan return -EROFS; 2167b83cc969SLi Zefan 2168c146afadSYan Zheng ret = mnt_want_write(file->f_path.mnt); 2169c146afadSYan Zheng if (ret) 2170c146afadSYan Zheng return ret; 2171c146afadSYan Zheng 2172c5c9cd4dSSage Weil src_file = fget(srcfd); 2173ab67b7c1SYan Zheng if (!src_file) { 2174ab67b7c1SYan Zheng ret = -EBADF; 2175ab67b7c1SYan Zheng goto out_drop_write; 2176ab67b7c1SYan Zheng } 21775dc64164SDan Rosenberg 2178f46b5a66SChristoph Hellwig src = src_file->f_dentry->d_inode; 2179f46b5a66SChristoph Hellwig 2180c5c9cd4dSSage Weil ret = -EINVAL; 2181c5c9cd4dSSage Weil if (src == inode) 2182c5c9cd4dSSage Weil goto out_fput; 2183c5c9cd4dSSage Weil 21845dc64164SDan Rosenberg /* the src must be open for reading */ 21855dc64164SDan Rosenberg if (!(src_file->f_mode & FMODE_READ)) 21865dc64164SDan Rosenberg goto out_fput; 21875dc64164SDan Rosenberg 2188ae01a0abSYan Zheng ret = -EISDIR; 2189ae01a0abSYan Zheng if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode)) 2190f46b5a66SChristoph Hellwig goto out_fput; 2191f46b5a66SChristoph Hellwig 2192ae01a0abSYan Zheng ret = -EXDEV; 2193ae01a0abSYan Zheng if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root) 2194ae01a0abSYan Zheng goto out_fput; 2195ae01a0abSYan Zheng 2196ae01a0abSYan Zheng ret = -ENOMEM; 2197ae01a0abSYan Zheng buf = vmalloc(btrfs_level_size(root, 0)); 2198ae01a0abSYan Zheng if (!buf) 2199ae01a0abSYan Zheng goto out_fput; 2200ae01a0abSYan Zheng 2201ae01a0abSYan Zheng path = btrfs_alloc_path(); 2202ae01a0abSYan Zheng if (!path) { 2203ae01a0abSYan Zheng vfree(buf); 2204ae01a0abSYan Zheng goto out_fput; 2205ae01a0abSYan Zheng } 2206ae01a0abSYan Zheng path->reada = 2; 2207ae01a0abSYan Zheng 2208f46b5a66SChristoph Hellwig if (inode < src) { 2209fccdae43SSage Weil mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT); 2210fccdae43SSage Weil mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD); 2211f46b5a66SChristoph Hellwig } else { 2212fccdae43SSage Weil mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT); 2213fccdae43SSage Weil mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD); 2214f46b5a66SChristoph Hellwig } 2215f46b5a66SChristoph Hellwig 2216c5c9cd4dSSage Weil /* determine range to clone */ 2217c5c9cd4dSSage Weil ret = -EINVAL; 22182ebc3464SDan Rosenberg if (off + len > src->i_size || off + len < off) 2219f46b5a66SChristoph Hellwig goto out_unlock; 2220c5c9cd4dSSage Weil if (len == 0) 2221c5c9cd4dSSage Weil olen = len = src->i_size - off; 2222c5c9cd4dSSage Weil /* if we extend to eof, continue to block boundary */ 2223c5c9cd4dSSage Weil if (off + len == src->i_size) 22242a6b8daeSLi Zefan len = ALIGN(src->i_size, bs) - off; 2225c5c9cd4dSSage Weil 2226c5c9cd4dSSage Weil /* verify the end result is block aligned */ 22272a6b8daeSLi Zefan if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) || 22282a6b8daeSLi Zefan !IS_ALIGNED(destoff, bs)) 2229c5c9cd4dSSage Weil goto out_unlock; 2230c5c9cd4dSSage Weil 2231f46b5a66SChristoph Hellwig /* do any pending delalloc/csum calc on src, one way or 2232f46b5a66SChristoph Hellwig another, and lock file content */ 2233f46b5a66SChristoph Hellwig while (1) { 223431840ae1SZheng Yan struct btrfs_ordered_extent *ordered; 2235c5c9cd4dSSage Weil lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS); 22369a019196SSage Weil ordered = btrfs_lookup_first_ordered_extent(src, off+len); 22379a019196SSage Weil if (!ordered && 22389a019196SSage Weil !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len, 22399a019196SSage Weil EXTENT_DELALLOC, 0, NULL)) 2240f46b5a66SChristoph Hellwig break; 2241c5c9cd4dSSage Weil unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS); 2242ae01a0abSYan Zheng if (ordered) 2243ae01a0abSYan Zheng btrfs_put_ordered_extent(ordered); 22449a019196SSage Weil btrfs_wait_ordered_range(src, off, len); 2245f46b5a66SChristoph Hellwig } 2246f46b5a66SChristoph Hellwig 2247c5c9cd4dSSage Weil /* clone data */ 224833345d01SLi Zefan key.objectid = btrfs_ino(src); 2249ae01a0abSYan Zheng key.type = BTRFS_EXTENT_DATA_KEY; 2250ae01a0abSYan Zheng key.offset = 0; 2251f46b5a66SChristoph Hellwig 2252f46b5a66SChristoph Hellwig while (1) { 2253f46b5a66SChristoph Hellwig /* 2254f46b5a66SChristoph Hellwig * note the key will change type as we walk through the 2255f46b5a66SChristoph Hellwig * tree. 2256f46b5a66SChristoph Hellwig */ 2257a22285a6SYan, Zheng ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 2258f46b5a66SChristoph Hellwig if (ret < 0) 2259f46b5a66SChristoph Hellwig goto out; 2260f46b5a66SChristoph Hellwig 2261ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 2262ae01a0abSYan Zheng if (path->slots[0] >= nritems) { 2263f46b5a66SChristoph Hellwig ret = btrfs_next_leaf(root, path); 2264f46b5a66SChristoph Hellwig if (ret < 0) 2265f46b5a66SChristoph Hellwig goto out; 2266f46b5a66SChristoph Hellwig if (ret > 0) 2267f46b5a66SChristoph Hellwig break; 2268ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 2269f46b5a66SChristoph Hellwig } 2270f46b5a66SChristoph Hellwig leaf = path->nodes[0]; 2271f46b5a66SChristoph Hellwig slot = path->slots[0]; 2272f46b5a66SChristoph Hellwig 2273ae01a0abSYan Zheng btrfs_item_key_to_cpu(leaf, &key, slot); 2274d20f7043SChris Mason if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY || 227533345d01SLi Zefan key.objectid != btrfs_ino(src)) 2276f46b5a66SChristoph Hellwig break; 2277f46b5a66SChristoph Hellwig 2278c5c9cd4dSSage Weil if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) { 2279c5c9cd4dSSage Weil struct btrfs_file_extent_item *extent; 2280c5c9cd4dSSage Weil int type; 228131840ae1SZheng Yan u32 size; 228231840ae1SZheng Yan struct btrfs_key new_key; 2283c5c9cd4dSSage Weil u64 disko = 0, diskl = 0; 2284c5c9cd4dSSage Weil u64 datao = 0, datal = 0; 2285c5c9cd4dSSage Weil u8 comp; 2286b5384d48SSage Weil u64 endoff; 228731840ae1SZheng Yan 228831840ae1SZheng Yan size = btrfs_item_size_nr(leaf, slot); 228931840ae1SZheng Yan read_extent_buffer(leaf, buf, 229031840ae1SZheng Yan btrfs_item_ptr_offset(leaf, slot), 229131840ae1SZheng Yan size); 2292c5c9cd4dSSage Weil 2293c5c9cd4dSSage Weil extent = btrfs_item_ptr(leaf, slot, 2294c5c9cd4dSSage Weil struct btrfs_file_extent_item); 2295c5c9cd4dSSage Weil comp = btrfs_file_extent_compression(leaf, extent); 2296c5c9cd4dSSage Weil type = btrfs_file_extent_type(leaf, extent); 2297c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 2298c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 2299d397712bSChris Mason disko = btrfs_file_extent_disk_bytenr(leaf, 2300d397712bSChris Mason extent); 2301d397712bSChris Mason diskl = btrfs_file_extent_disk_num_bytes(leaf, 2302d397712bSChris Mason extent); 2303c5c9cd4dSSage Weil datao = btrfs_file_extent_offset(leaf, extent); 2304d397712bSChris Mason datal = btrfs_file_extent_num_bytes(leaf, 2305d397712bSChris Mason extent); 2306c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 2307c5c9cd4dSSage Weil /* take upper bound, may be compressed */ 2308c5c9cd4dSSage Weil datal = btrfs_file_extent_ram_bytes(leaf, 2309c5c9cd4dSSage Weil extent); 2310c5c9cd4dSSage Weil } 2311b3b4aa74SDavid Sterba btrfs_release_path(path); 231231840ae1SZheng Yan 2313050006a7SSage Weil if (key.offset + datal <= off || 2314c5c9cd4dSSage Weil key.offset >= off+len) 2315c5c9cd4dSSage Weil goto next; 2316c5c9cd4dSSage Weil 231731840ae1SZheng Yan memcpy(&new_key, &key, sizeof(new_key)); 231833345d01SLi Zefan new_key.objectid = btrfs_ino(inode); 23194d728ec7SLi Zefan if (off <= key.offset) 2320c5c9cd4dSSage Weil new_key.offset = key.offset + destoff - off; 23214d728ec7SLi Zefan else 23224d728ec7SLi Zefan new_key.offset = destoff; 2323c5c9cd4dSSage Weil 2324a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 1); 2325a22285a6SYan, Zheng if (IS_ERR(trans)) { 2326a22285a6SYan, Zheng ret = PTR_ERR(trans); 2327a22285a6SYan, Zheng goto out; 2328a22285a6SYan, Zheng } 2329a22285a6SYan, Zheng 2330c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 2331c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 2332a22285a6SYan, Zheng if (off > key.offset) { 2333a22285a6SYan, Zheng datao += off - key.offset; 2334a22285a6SYan, Zheng datal -= off - key.offset; 2335a22285a6SYan, Zheng } 2336a22285a6SYan, Zheng 2337a22285a6SYan, Zheng if (key.offset + datal > off + len) 2338a22285a6SYan, Zheng datal = off + len - key.offset; 2339a22285a6SYan, Zheng 2340a22285a6SYan, Zheng ret = btrfs_drop_extents(trans, inode, 2341a22285a6SYan, Zheng new_key.offset, 2342a22285a6SYan, Zheng new_key.offset + datal, 2343a22285a6SYan, Zheng &hint_byte, 1); 2344a22285a6SYan, Zheng BUG_ON(ret); 2345a22285a6SYan, Zheng 234631840ae1SZheng Yan ret = btrfs_insert_empty_item(trans, root, path, 234731840ae1SZheng Yan &new_key, size); 2348a22285a6SYan, Zheng BUG_ON(ret); 234931840ae1SZheng Yan 235031840ae1SZheng Yan leaf = path->nodes[0]; 235131840ae1SZheng Yan slot = path->slots[0]; 235231840ae1SZheng Yan write_extent_buffer(leaf, buf, 235331840ae1SZheng Yan btrfs_item_ptr_offset(leaf, slot), 235431840ae1SZheng Yan size); 2355ae01a0abSYan Zheng 2356f46b5a66SChristoph Hellwig extent = btrfs_item_ptr(leaf, slot, 2357f46b5a66SChristoph Hellwig struct btrfs_file_extent_item); 2358c5c9cd4dSSage Weil 2359c5c9cd4dSSage Weil /* disko == 0 means it's a hole */ 2360c5c9cd4dSSage Weil if (!disko) 2361c5c9cd4dSSage Weil datao = 0; 2362c5c9cd4dSSage Weil 2363c5c9cd4dSSage Weil btrfs_set_file_extent_offset(leaf, extent, 2364c5c9cd4dSSage Weil datao); 2365c5c9cd4dSSage Weil btrfs_set_file_extent_num_bytes(leaf, extent, 2366c5c9cd4dSSage Weil datal); 2367c5c9cd4dSSage Weil if (disko) { 2368c5c9cd4dSSage Weil inode_add_bytes(inode, datal); 2369ae01a0abSYan Zheng ret = btrfs_inc_extent_ref(trans, root, 23705d4f98a2SYan Zheng disko, diskl, 0, 2371f46b5a66SChristoph Hellwig root->root_key.objectid, 237233345d01SLi Zefan btrfs_ino(inode), 23735d4f98a2SYan Zheng new_key.offset - datao); 2374ae01a0abSYan Zheng BUG_ON(ret); 2375f46b5a66SChristoph Hellwig } 2376c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 2377c5c9cd4dSSage Weil u64 skip = 0; 2378c5c9cd4dSSage Weil u64 trim = 0; 2379c5c9cd4dSSage Weil if (off > key.offset) { 2380c5c9cd4dSSage Weil skip = off - key.offset; 2381c5c9cd4dSSage Weil new_key.offset += skip; 238231840ae1SZheng Yan } 2383d397712bSChris Mason 2384c5c9cd4dSSage Weil if (key.offset + datal > off+len) 2385c5c9cd4dSSage Weil trim = key.offset + datal - (off+len); 2386d397712bSChris Mason 2387c5c9cd4dSSage Weil if (comp && (skip || trim)) { 2388c5c9cd4dSSage Weil ret = -EINVAL; 2389a22285a6SYan, Zheng btrfs_end_transaction(trans, root); 2390c5c9cd4dSSage Weil goto out; 239131840ae1SZheng Yan } 2392c5c9cd4dSSage Weil size -= skip + trim; 2393c5c9cd4dSSage Weil datal -= skip + trim; 2394a22285a6SYan, Zheng 2395a22285a6SYan, Zheng ret = btrfs_drop_extents(trans, inode, 2396a22285a6SYan, Zheng new_key.offset, 2397a22285a6SYan, Zheng new_key.offset + datal, 2398a22285a6SYan, Zheng &hint_byte, 1); 2399a22285a6SYan, Zheng BUG_ON(ret); 2400a22285a6SYan, Zheng 2401c5c9cd4dSSage Weil ret = btrfs_insert_empty_item(trans, root, path, 2402c5c9cd4dSSage Weil &new_key, size); 2403a22285a6SYan, Zheng BUG_ON(ret); 2404c5c9cd4dSSage Weil 2405c5c9cd4dSSage Weil if (skip) { 2406d397712bSChris Mason u32 start = 2407d397712bSChris Mason btrfs_file_extent_calc_inline_size(0); 2408c5c9cd4dSSage Weil memmove(buf+start, buf+start+skip, 2409c5c9cd4dSSage Weil datal); 2410c5c9cd4dSSage Weil } 2411c5c9cd4dSSage Weil 2412c5c9cd4dSSage Weil leaf = path->nodes[0]; 2413c5c9cd4dSSage Weil slot = path->slots[0]; 2414c5c9cd4dSSage Weil write_extent_buffer(leaf, buf, 2415c5c9cd4dSSage Weil btrfs_item_ptr_offset(leaf, slot), 2416c5c9cd4dSSage Weil size); 2417c5c9cd4dSSage Weil inode_add_bytes(inode, datal); 2418c5c9cd4dSSage Weil } 2419c5c9cd4dSSage Weil 2420c5c9cd4dSSage Weil btrfs_mark_buffer_dirty(leaf); 2421b3b4aa74SDavid Sterba btrfs_release_path(path); 2422c5c9cd4dSSage Weil 2423a22285a6SYan, Zheng inode->i_mtime = inode->i_ctime = CURRENT_TIME; 2424b5384d48SSage Weil 2425b5384d48SSage Weil /* 2426b5384d48SSage Weil * we round up to the block size at eof when 2427b5384d48SSage Weil * determining which extents to clone above, 2428b5384d48SSage Weil * but shouldn't round up the file size 2429b5384d48SSage Weil */ 2430b5384d48SSage Weil endoff = new_key.offset + datal; 24315f3888ffSLi Zefan if (endoff > destoff+olen) 24325f3888ffSLi Zefan endoff = destoff+olen; 2433b5384d48SSage Weil if (endoff > inode->i_size) 2434b5384d48SSage Weil btrfs_i_size_write(inode, endoff); 2435b5384d48SSage Weil 2436a22285a6SYan, Zheng BTRFS_I(inode)->flags = BTRFS_I(src)->flags; 2437a22285a6SYan, Zheng ret = btrfs_update_inode(trans, root, inode); 2438a22285a6SYan, Zheng BUG_ON(ret); 2439a22285a6SYan, Zheng btrfs_end_transaction(trans, root); 2440a22285a6SYan, Zheng } 2441c5c9cd4dSSage Weil next: 2442b3b4aa74SDavid Sterba btrfs_release_path(path); 2443ae01a0abSYan Zheng key.offset++; 2444ae01a0abSYan Zheng } 2445f46b5a66SChristoph Hellwig ret = 0; 2446f46b5a66SChristoph Hellwig out: 2447b3b4aa74SDavid Sterba btrfs_release_path(path); 2448c5c9cd4dSSage Weil unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS); 2449f46b5a66SChristoph Hellwig out_unlock: 2450f46b5a66SChristoph Hellwig mutex_unlock(&src->i_mutex); 2451f46b5a66SChristoph Hellwig mutex_unlock(&inode->i_mutex); 2452ae01a0abSYan Zheng vfree(buf); 2453ae01a0abSYan Zheng btrfs_free_path(path); 2454f46b5a66SChristoph Hellwig out_fput: 2455f46b5a66SChristoph Hellwig fput(src_file); 2456ab67b7c1SYan Zheng out_drop_write: 2457ab67b7c1SYan Zheng mnt_drop_write(file->f_path.mnt); 2458f46b5a66SChristoph Hellwig return ret; 2459f46b5a66SChristoph Hellwig } 2460f46b5a66SChristoph Hellwig 24617a865e8aSChristoph Hellwig static long btrfs_ioctl_clone_range(struct file *file, void __user *argp) 2462c5c9cd4dSSage Weil { 2463c5c9cd4dSSage Weil struct btrfs_ioctl_clone_range_args args; 2464c5c9cd4dSSage Weil 24657a865e8aSChristoph Hellwig if (copy_from_user(&args, argp, sizeof(args))) 2466c5c9cd4dSSage Weil return -EFAULT; 2467c5c9cd4dSSage Weil return btrfs_ioctl_clone(file, args.src_fd, args.src_offset, 2468c5c9cd4dSSage Weil args.src_length, args.dest_offset); 2469c5c9cd4dSSage Weil } 2470c5c9cd4dSSage Weil 2471f46b5a66SChristoph Hellwig /* 2472f46b5a66SChristoph Hellwig * there are many ways the trans_start and trans_end ioctls can lead 2473f46b5a66SChristoph Hellwig * to deadlocks. They should only be used by applications that 2474f46b5a66SChristoph Hellwig * basically own the machine, and have a very in depth understanding 2475f46b5a66SChristoph Hellwig * of all the possible deadlocks and enospc problems. 2476f46b5a66SChristoph Hellwig */ 2477b2950863SChristoph Hellwig static long btrfs_ioctl_trans_start(struct file *file) 2478f46b5a66SChristoph Hellwig { 2479f46b5a66SChristoph Hellwig struct inode *inode = fdentry(file)->d_inode; 2480f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 2481f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 24821ab86aedSSage Weil int ret; 2483f46b5a66SChristoph Hellwig 24841ab86aedSSage Weil ret = -EPERM; 2485df5b5520SChristoph Hellwig if (!capable(CAP_SYS_ADMIN)) 2486f46b5a66SChristoph Hellwig goto out; 24871ab86aedSSage Weil 24881ab86aedSSage Weil ret = -EINPROGRESS; 24891ab86aedSSage Weil if (file->private_data) 24901ab86aedSSage Weil goto out; 24919ca9ee09SSage Weil 2492b83cc969SLi Zefan ret = -EROFS; 2493b83cc969SLi Zefan if (btrfs_root_readonly(root)) 2494b83cc969SLi Zefan goto out; 2495b83cc969SLi Zefan 2496c146afadSYan Zheng ret = mnt_want_write(file->f_path.mnt); 2497c146afadSYan Zheng if (ret) 2498c146afadSYan Zheng goto out; 2499c146afadSYan Zheng 2500a4abeea4SJosef Bacik atomic_inc(&root->fs_info->open_ioctl_trans); 25019ca9ee09SSage Weil 2502f46b5a66SChristoph Hellwig ret = -ENOMEM; 25037a7eaa40SJosef Bacik trans = btrfs_start_ioctl_transaction(root); 2504abd30bb0STsutomu Itoh if (IS_ERR(trans)) 25051ab86aedSSage Weil goto out_drop; 25061ab86aedSSage Weil 25071ab86aedSSage Weil file->private_data = trans; 25081ab86aedSSage Weil return 0; 25091ab86aedSSage Weil 25101ab86aedSSage Weil out_drop: 2511a4abeea4SJosef Bacik atomic_dec(&root->fs_info->open_ioctl_trans); 25121ab86aedSSage Weil mnt_drop_write(file->f_path.mnt); 2513f46b5a66SChristoph Hellwig out: 2514f46b5a66SChristoph Hellwig return ret; 2515f46b5a66SChristoph Hellwig } 2516f46b5a66SChristoph Hellwig 25176ef5ed0dSJosef Bacik static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) 25186ef5ed0dSJosef Bacik { 25196ef5ed0dSJosef Bacik struct inode *inode = fdentry(file)->d_inode; 25206ef5ed0dSJosef Bacik struct btrfs_root *root = BTRFS_I(inode)->root; 25216ef5ed0dSJosef Bacik struct btrfs_root *new_root; 25226ef5ed0dSJosef Bacik struct btrfs_dir_item *di; 25236ef5ed0dSJosef Bacik struct btrfs_trans_handle *trans; 25246ef5ed0dSJosef Bacik struct btrfs_path *path; 25256ef5ed0dSJosef Bacik struct btrfs_key location; 25266ef5ed0dSJosef Bacik struct btrfs_disk_key disk_key; 25276ef5ed0dSJosef Bacik struct btrfs_super_block *disk_super; 25286ef5ed0dSJosef Bacik u64 features; 25296ef5ed0dSJosef Bacik u64 objectid = 0; 25306ef5ed0dSJosef Bacik u64 dir_id; 25316ef5ed0dSJosef Bacik 25326ef5ed0dSJosef Bacik if (!capable(CAP_SYS_ADMIN)) 25336ef5ed0dSJosef Bacik return -EPERM; 25346ef5ed0dSJosef Bacik 25356ef5ed0dSJosef Bacik if (copy_from_user(&objectid, argp, sizeof(objectid))) 25366ef5ed0dSJosef Bacik return -EFAULT; 25376ef5ed0dSJosef Bacik 25386ef5ed0dSJosef Bacik if (!objectid) 25396ef5ed0dSJosef Bacik objectid = root->root_key.objectid; 25406ef5ed0dSJosef Bacik 25416ef5ed0dSJosef Bacik location.objectid = objectid; 25426ef5ed0dSJosef Bacik location.type = BTRFS_ROOT_ITEM_KEY; 25436ef5ed0dSJosef Bacik location.offset = (u64)-1; 25446ef5ed0dSJosef Bacik 25456ef5ed0dSJosef Bacik new_root = btrfs_read_fs_root_no_name(root->fs_info, &location); 25466ef5ed0dSJosef Bacik if (IS_ERR(new_root)) 25476ef5ed0dSJosef Bacik return PTR_ERR(new_root); 25486ef5ed0dSJosef Bacik 25496ef5ed0dSJosef Bacik if (btrfs_root_refs(&new_root->root_item) == 0) 25506ef5ed0dSJosef Bacik return -ENOENT; 25516ef5ed0dSJosef Bacik 25526ef5ed0dSJosef Bacik path = btrfs_alloc_path(); 25536ef5ed0dSJosef Bacik if (!path) 25546ef5ed0dSJosef Bacik return -ENOMEM; 25556ef5ed0dSJosef Bacik path->leave_spinning = 1; 25566ef5ed0dSJosef Bacik 25576ef5ed0dSJosef Bacik trans = btrfs_start_transaction(root, 1); 255898d5dc13STsutomu Itoh if (IS_ERR(trans)) { 25596ef5ed0dSJosef Bacik btrfs_free_path(path); 256098d5dc13STsutomu Itoh return PTR_ERR(trans); 25616ef5ed0dSJosef Bacik } 25626ef5ed0dSJosef Bacik 25636ef5ed0dSJosef Bacik dir_id = btrfs_super_root_dir(&root->fs_info->super_copy); 25646ef5ed0dSJosef Bacik di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path, 25656ef5ed0dSJosef Bacik dir_id, "default", 7, 1); 2566cf1e99a4SDan Carpenter if (IS_ERR_OR_NULL(di)) { 25676ef5ed0dSJosef Bacik btrfs_free_path(path); 25686ef5ed0dSJosef Bacik btrfs_end_transaction(trans, root); 25696ef5ed0dSJosef Bacik printk(KERN_ERR "Umm, you don't have the default dir item, " 25706ef5ed0dSJosef Bacik "this isn't going to work\n"); 25716ef5ed0dSJosef Bacik return -ENOENT; 25726ef5ed0dSJosef Bacik } 25736ef5ed0dSJosef Bacik 25746ef5ed0dSJosef Bacik btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); 25756ef5ed0dSJosef Bacik btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); 25766ef5ed0dSJosef Bacik btrfs_mark_buffer_dirty(path->nodes[0]); 25776ef5ed0dSJosef Bacik btrfs_free_path(path); 25786ef5ed0dSJosef Bacik 25796ef5ed0dSJosef Bacik disk_super = &root->fs_info->super_copy; 25806ef5ed0dSJosef Bacik features = btrfs_super_incompat_flags(disk_super); 25816ef5ed0dSJosef Bacik if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) { 25826ef5ed0dSJosef Bacik features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL; 25836ef5ed0dSJosef Bacik btrfs_set_super_incompat_flags(disk_super, features); 25846ef5ed0dSJosef Bacik } 25856ef5ed0dSJosef Bacik btrfs_end_transaction(trans, root); 25866ef5ed0dSJosef Bacik 25876ef5ed0dSJosef Bacik return 0; 25886ef5ed0dSJosef Bacik } 25896ef5ed0dSJosef Bacik 2590bf5fc093SJosef Bacik static void get_block_group_info(struct list_head *groups_list, 2591bf5fc093SJosef Bacik struct btrfs_ioctl_space_info *space) 2592bf5fc093SJosef Bacik { 2593bf5fc093SJosef Bacik struct btrfs_block_group_cache *block_group; 2594bf5fc093SJosef Bacik 2595bf5fc093SJosef Bacik space->total_bytes = 0; 2596bf5fc093SJosef Bacik space->used_bytes = 0; 2597bf5fc093SJosef Bacik space->flags = 0; 2598bf5fc093SJosef Bacik list_for_each_entry(block_group, groups_list, list) { 2599bf5fc093SJosef Bacik space->flags = block_group->flags; 2600bf5fc093SJosef Bacik space->total_bytes += block_group->key.offset; 2601bf5fc093SJosef Bacik space->used_bytes += 2602bf5fc093SJosef Bacik btrfs_block_group_used(&block_group->item); 2603bf5fc093SJosef Bacik } 2604bf5fc093SJosef Bacik } 2605bf5fc093SJosef Bacik 26061406e432SJosef Bacik long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg) 26071406e432SJosef Bacik { 26081406e432SJosef Bacik struct btrfs_ioctl_space_args space_args; 26091406e432SJosef Bacik struct btrfs_ioctl_space_info space; 26101406e432SJosef Bacik struct btrfs_ioctl_space_info *dest; 26117fde62bfSChris Mason struct btrfs_ioctl_space_info *dest_orig; 261213f2696fSDaniel J Blueman struct btrfs_ioctl_space_info __user *user_dest; 26131406e432SJosef Bacik struct btrfs_space_info *info; 2614bf5fc093SJosef Bacik u64 types[] = {BTRFS_BLOCK_GROUP_DATA, 2615bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_SYSTEM, 2616bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_METADATA, 2617bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA}; 2618bf5fc093SJosef Bacik int num_types = 4; 26197fde62bfSChris Mason int alloc_size; 26201406e432SJosef Bacik int ret = 0; 262151788b1bSDan Rosenberg u64 slot_count = 0; 2622bf5fc093SJosef Bacik int i, c; 26231406e432SJosef Bacik 26241406e432SJosef Bacik if (copy_from_user(&space_args, 26251406e432SJosef Bacik (struct btrfs_ioctl_space_args __user *)arg, 26261406e432SJosef Bacik sizeof(space_args))) 26271406e432SJosef Bacik return -EFAULT; 26281406e432SJosef Bacik 2629bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 2630bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 2631bf5fc093SJosef Bacik 2632bf5fc093SJosef Bacik info = NULL; 26337fde62bfSChris Mason rcu_read_lock(); 2634bf5fc093SJosef Bacik list_for_each_entry_rcu(tmp, &root->fs_info->space_info, 2635bf5fc093SJosef Bacik list) { 2636bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 2637bf5fc093SJosef Bacik info = tmp; 2638bf5fc093SJosef Bacik break; 2639bf5fc093SJosef Bacik } 2640bf5fc093SJosef Bacik } 26417fde62bfSChris Mason rcu_read_unlock(); 26421406e432SJosef Bacik 2643bf5fc093SJosef Bacik if (!info) 2644bf5fc093SJosef Bacik continue; 2645bf5fc093SJosef Bacik 2646bf5fc093SJosef Bacik down_read(&info->groups_sem); 2647bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 2648bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) 2649bf5fc093SJosef Bacik slot_count++; 2650bf5fc093SJosef Bacik } 2651bf5fc093SJosef Bacik up_read(&info->groups_sem); 2652bf5fc093SJosef Bacik } 2653bf5fc093SJosef Bacik 26547fde62bfSChris Mason /* space_slots == 0 means they are asking for a count */ 26557fde62bfSChris Mason if (space_args.space_slots == 0) { 26567fde62bfSChris Mason space_args.total_spaces = slot_count; 26577fde62bfSChris Mason goto out; 26587fde62bfSChris Mason } 2659bf5fc093SJosef Bacik 266051788b1bSDan Rosenberg slot_count = min_t(u64, space_args.space_slots, slot_count); 2661bf5fc093SJosef Bacik 26627fde62bfSChris Mason alloc_size = sizeof(*dest) * slot_count; 2663bf5fc093SJosef Bacik 26647fde62bfSChris Mason /* we generally have at most 6 or so space infos, one for each raid 26657fde62bfSChris Mason * level. So, a whole page should be more than enough for everyone 26667fde62bfSChris Mason */ 26677fde62bfSChris Mason if (alloc_size > PAGE_CACHE_SIZE) 26687fde62bfSChris Mason return -ENOMEM; 26697fde62bfSChris Mason 26707fde62bfSChris Mason space_args.total_spaces = 0; 26717fde62bfSChris Mason dest = kmalloc(alloc_size, GFP_NOFS); 26727fde62bfSChris Mason if (!dest) 26737fde62bfSChris Mason return -ENOMEM; 26747fde62bfSChris Mason dest_orig = dest; 26757fde62bfSChris Mason 26767fde62bfSChris Mason /* now we have a buffer to copy into */ 2677bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 2678bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 2679bf5fc093SJosef Bacik 268051788b1bSDan Rosenberg if (!slot_count) 268151788b1bSDan Rosenberg break; 268251788b1bSDan Rosenberg 2683bf5fc093SJosef Bacik info = NULL; 26841406e432SJosef Bacik rcu_read_lock(); 2685bf5fc093SJosef Bacik list_for_each_entry_rcu(tmp, &root->fs_info->space_info, 2686bf5fc093SJosef Bacik list) { 2687bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 2688bf5fc093SJosef Bacik info = tmp; 26897fde62bfSChris Mason break; 2690bf5fc093SJosef Bacik } 2691bf5fc093SJosef Bacik } 2692bf5fc093SJosef Bacik rcu_read_unlock(); 26937fde62bfSChris Mason 2694bf5fc093SJosef Bacik if (!info) 2695bf5fc093SJosef Bacik continue; 2696bf5fc093SJosef Bacik down_read(&info->groups_sem); 2697bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 2698bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) { 2699bf5fc093SJosef Bacik get_block_group_info(&info->block_groups[c], 2700bf5fc093SJosef Bacik &space); 27017fde62bfSChris Mason memcpy(dest, &space, sizeof(space)); 27021406e432SJosef Bacik dest++; 27031406e432SJosef Bacik space_args.total_spaces++; 270451788b1bSDan Rosenberg slot_count--; 27051406e432SJosef Bacik } 270651788b1bSDan Rosenberg if (!slot_count) 270751788b1bSDan Rosenberg break; 2708bf5fc093SJosef Bacik } 2709bf5fc093SJosef Bacik up_read(&info->groups_sem); 2710bf5fc093SJosef Bacik } 27111406e432SJosef Bacik 27127fde62bfSChris Mason user_dest = (struct btrfs_ioctl_space_info *) 27137fde62bfSChris Mason (arg + sizeof(struct btrfs_ioctl_space_args)); 27147fde62bfSChris Mason 27157fde62bfSChris Mason if (copy_to_user(user_dest, dest_orig, alloc_size)) 27167fde62bfSChris Mason ret = -EFAULT; 27177fde62bfSChris Mason 27187fde62bfSChris Mason kfree(dest_orig); 27197fde62bfSChris Mason out: 27207fde62bfSChris Mason if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) 27211406e432SJosef Bacik ret = -EFAULT; 27221406e432SJosef Bacik 27231406e432SJosef Bacik return ret; 27241406e432SJosef Bacik } 27251406e432SJosef Bacik 2726f46b5a66SChristoph Hellwig /* 2727f46b5a66SChristoph Hellwig * there are many ways the trans_start and trans_end ioctls can lead 2728f46b5a66SChristoph Hellwig * to deadlocks. They should only be used by applications that 2729f46b5a66SChristoph Hellwig * basically own the machine, and have a very in depth understanding 2730f46b5a66SChristoph Hellwig * of all the possible deadlocks and enospc problems. 2731f46b5a66SChristoph Hellwig */ 2732f46b5a66SChristoph Hellwig long btrfs_ioctl_trans_end(struct file *file) 2733f46b5a66SChristoph Hellwig { 2734f46b5a66SChristoph Hellwig struct inode *inode = fdentry(file)->d_inode; 2735f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 2736f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 2737f46b5a66SChristoph Hellwig 2738f46b5a66SChristoph Hellwig trans = file->private_data; 27391ab86aedSSage Weil if (!trans) 27401ab86aedSSage Weil return -EINVAL; 2741b214107eSChristoph Hellwig file->private_data = NULL; 27429ca9ee09SSage Weil 27431ab86aedSSage Weil btrfs_end_transaction(trans, root); 27441ab86aedSSage Weil 2745a4abeea4SJosef Bacik atomic_dec(&root->fs_info->open_ioctl_trans); 27469ca9ee09SSage Weil 2747cfc8ea87SSage Weil mnt_drop_write(file->f_path.mnt); 27481ab86aedSSage Weil return 0; 2749f46b5a66SChristoph Hellwig } 2750f46b5a66SChristoph Hellwig 275146204592SSage Weil static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp) 275246204592SSage Weil { 275346204592SSage Weil struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root; 275446204592SSage Weil struct btrfs_trans_handle *trans; 275546204592SSage Weil u64 transid; 2756db5b493aSTsutomu Itoh int ret; 275746204592SSage Weil 275846204592SSage Weil trans = btrfs_start_transaction(root, 0); 275998d5dc13STsutomu Itoh if (IS_ERR(trans)) 276098d5dc13STsutomu Itoh return PTR_ERR(trans); 276146204592SSage Weil transid = trans->transid; 2762db5b493aSTsutomu Itoh ret = btrfs_commit_transaction_async(trans, root, 0); 27638b2b2d3cSTsutomu Itoh if (ret) { 27648b2b2d3cSTsutomu Itoh btrfs_end_transaction(trans, root); 2765db5b493aSTsutomu Itoh return ret; 27668b2b2d3cSTsutomu Itoh } 276746204592SSage Weil 276846204592SSage Weil if (argp) 276946204592SSage Weil if (copy_to_user(argp, &transid, sizeof(transid))) 277046204592SSage Weil return -EFAULT; 277146204592SSage Weil return 0; 277246204592SSage Weil } 277346204592SSage Weil 277446204592SSage Weil static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp) 277546204592SSage Weil { 277646204592SSage Weil struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root; 277746204592SSage Weil u64 transid; 277846204592SSage Weil 277946204592SSage Weil if (argp) { 278046204592SSage Weil if (copy_from_user(&transid, argp, sizeof(transid))) 278146204592SSage Weil return -EFAULT; 278246204592SSage Weil } else { 278346204592SSage Weil transid = 0; /* current trans */ 278446204592SSage Weil } 278546204592SSage Weil return btrfs_wait_for_commit(root, transid); 278646204592SSage Weil } 278746204592SSage Weil 2788475f6387SJan Schmidt static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg) 2789475f6387SJan Schmidt { 2790475f6387SJan Schmidt int ret; 2791475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 2792475f6387SJan Schmidt 2793475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 2794475f6387SJan Schmidt return -EPERM; 2795475f6387SJan Schmidt 2796475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 2797475f6387SJan Schmidt if (IS_ERR(sa)) 2798475f6387SJan Schmidt return PTR_ERR(sa); 2799475f6387SJan Schmidt 2800475f6387SJan Schmidt ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end, 28018628764eSArne Jansen &sa->progress, sa->flags & BTRFS_SCRUB_READONLY); 2802475f6387SJan Schmidt 2803475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 2804475f6387SJan Schmidt ret = -EFAULT; 2805475f6387SJan Schmidt 2806475f6387SJan Schmidt kfree(sa); 2807475f6387SJan Schmidt return ret; 2808475f6387SJan Schmidt } 2809475f6387SJan Schmidt 2810475f6387SJan Schmidt static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg) 2811475f6387SJan Schmidt { 2812475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 2813475f6387SJan Schmidt return -EPERM; 2814475f6387SJan Schmidt 2815475f6387SJan Schmidt return btrfs_scrub_cancel(root); 2816475f6387SJan Schmidt } 2817475f6387SJan Schmidt 2818475f6387SJan Schmidt static long btrfs_ioctl_scrub_progress(struct btrfs_root *root, 2819475f6387SJan Schmidt void __user *arg) 2820475f6387SJan Schmidt { 2821475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 2822475f6387SJan Schmidt int ret; 2823475f6387SJan Schmidt 2824475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 2825475f6387SJan Schmidt return -EPERM; 2826475f6387SJan Schmidt 2827475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 2828475f6387SJan Schmidt if (IS_ERR(sa)) 2829475f6387SJan Schmidt return PTR_ERR(sa); 2830475f6387SJan Schmidt 2831475f6387SJan Schmidt ret = btrfs_scrub_progress(root, sa->devid, &sa->progress); 2832475f6387SJan Schmidt 2833475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 2834475f6387SJan Schmidt ret = -EFAULT; 2835475f6387SJan Schmidt 2836475f6387SJan Schmidt kfree(sa); 2837475f6387SJan Schmidt return ret; 2838475f6387SJan Schmidt } 2839475f6387SJan Schmidt 2840f46b5a66SChristoph Hellwig long btrfs_ioctl(struct file *file, unsigned int 2841f46b5a66SChristoph Hellwig cmd, unsigned long arg) 2842f46b5a66SChristoph Hellwig { 2843f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; 28444bcabaa3SChristoph Hellwig void __user *argp = (void __user *)arg; 2845f46b5a66SChristoph Hellwig 2846f46b5a66SChristoph Hellwig switch (cmd) { 28476cbff00fSChristoph Hellwig case FS_IOC_GETFLAGS: 28486cbff00fSChristoph Hellwig return btrfs_ioctl_getflags(file, argp); 28496cbff00fSChristoph Hellwig case FS_IOC_SETFLAGS: 28506cbff00fSChristoph Hellwig return btrfs_ioctl_setflags(file, argp); 28516cbff00fSChristoph Hellwig case FS_IOC_GETVERSION: 28526cbff00fSChristoph Hellwig return btrfs_ioctl_getversion(file, argp); 2853f7039b1dSLi Dongyang case FITRIM: 2854f7039b1dSLi Dongyang return btrfs_ioctl_fitrim(file, argp); 2855f46b5a66SChristoph Hellwig case BTRFS_IOC_SNAP_CREATE: 2856fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 0); 2857fdfb1e4fSLi Zefan case BTRFS_IOC_SNAP_CREATE_V2: 2858fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create_v2(file, argp, 0); 28593de4586cSChris Mason case BTRFS_IOC_SUBVOL_CREATE: 2860fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 1); 286176dda93cSYan, Zheng case BTRFS_IOC_SNAP_DESTROY: 286276dda93cSYan, Zheng return btrfs_ioctl_snap_destroy(file, argp); 28630caa102dSLi Zefan case BTRFS_IOC_SUBVOL_GETFLAGS: 28640caa102dSLi Zefan return btrfs_ioctl_subvol_getflags(file, argp); 28650caa102dSLi Zefan case BTRFS_IOC_SUBVOL_SETFLAGS: 28660caa102dSLi Zefan return btrfs_ioctl_subvol_setflags(file, argp); 28676ef5ed0dSJosef Bacik case BTRFS_IOC_DEFAULT_SUBVOL: 28686ef5ed0dSJosef Bacik return btrfs_ioctl_default_subvol(file, argp); 2869f46b5a66SChristoph Hellwig case BTRFS_IOC_DEFRAG: 28701e701a32SChris Mason return btrfs_ioctl_defrag(file, NULL); 28711e701a32SChris Mason case BTRFS_IOC_DEFRAG_RANGE: 28721e701a32SChris Mason return btrfs_ioctl_defrag(file, argp); 2873f46b5a66SChristoph Hellwig case BTRFS_IOC_RESIZE: 28744bcabaa3SChristoph Hellwig return btrfs_ioctl_resize(root, argp); 2875f46b5a66SChristoph Hellwig case BTRFS_IOC_ADD_DEV: 28764bcabaa3SChristoph Hellwig return btrfs_ioctl_add_dev(root, argp); 2877f46b5a66SChristoph Hellwig case BTRFS_IOC_RM_DEV: 28784bcabaa3SChristoph Hellwig return btrfs_ioctl_rm_dev(root, argp); 2879475f6387SJan Schmidt case BTRFS_IOC_FS_INFO: 2880475f6387SJan Schmidt return btrfs_ioctl_fs_info(root, argp); 2881475f6387SJan Schmidt case BTRFS_IOC_DEV_INFO: 2882475f6387SJan Schmidt return btrfs_ioctl_dev_info(root, argp); 2883f46b5a66SChristoph Hellwig case BTRFS_IOC_BALANCE: 2884f46b5a66SChristoph Hellwig return btrfs_balance(root->fs_info->dev_root); 2885f46b5a66SChristoph Hellwig case BTRFS_IOC_CLONE: 2886c5c9cd4dSSage Weil return btrfs_ioctl_clone(file, arg, 0, 0, 0); 2887c5c9cd4dSSage Weil case BTRFS_IOC_CLONE_RANGE: 28887a865e8aSChristoph Hellwig return btrfs_ioctl_clone_range(file, argp); 2889f46b5a66SChristoph Hellwig case BTRFS_IOC_TRANS_START: 2890f46b5a66SChristoph Hellwig return btrfs_ioctl_trans_start(file); 2891f46b5a66SChristoph Hellwig case BTRFS_IOC_TRANS_END: 2892f46b5a66SChristoph Hellwig return btrfs_ioctl_trans_end(file); 2893ac8e9819SChris Mason case BTRFS_IOC_TREE_SEARCH: 2894ac8e9819SChris Mason return btrfs_ioctl_tree_search(file, argp); 2895ac8e9819SChris Mason case BTRFS_IOC_INO_LOOKUP: 2896ac8e9819SChris Mason return btrfs_ioctl_ino_lookup(file, argp); 28971406e432SJosef Bacik case BTRFS_IOC_SPACE_INFO: 28981406e432SJosef Bacik return btrfs_ioctl_space_info(root, argp); 2899f46b5a66SChristoph Hellwig case BTRFS_IOC_SYNC: 2900f46b5a66SChristoph Hellwig btrfs_sync_fs(file->f_dentry->d_sb, 1); 2901f46b5a66SChristoph Hellwig return 0; 290246204592SSage Weil case BTRFS_IOC_START_SYNC: 290346204592SSage Weil return btrfs_ioctl_start_sync(file, argp); 290446204592SSage Weil case BTRFS_IOC_WAIT_SYNC: 290546204592SSage Weil return btrfs_ioctl_wait_sync(file, argp); 2906475f6387SJan Schmidt case BTRFS_IOC_SCRUB: 2907475f6387SJan Schmidt return btrfs_ioctl_scrub(root, argp); 2908475f6387SJan Schmidt case BTRFS_IOC_SCRUB_CANCEL: 2909475f6387SJan Schmidt return btrfs_ioctl_scrub_cancel(root, argp); 2910475f6387SJan Schmidt case BTRFS_IOC_SCRUB_PROGRESS: 2911475f6387SJan Schmidt return btrfs_ioctl_scrub_progress(root, argp); 2912f46b5a66SChristoph Hellwig } 2913f46b5a66SChristoph Hellwig 2914f46b5a66SChristoph Hellwig return -ENOTTY; 2915f46b5a66SChristoph Hellwig } 2916