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 * 120e27425d6SJosef Bacik * Currently only the compression flags and the cow flags are inherited. 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 131e27425d6SJosef Bacik if (flags & BTRFS_INODE_NOCOMPRESS) { 132e27425d6SJosef Bacik BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS; 133e27425d6SJosef Bacik BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS; 134e27425d6SJosef Bacik } else if (flags & BTRFS_INODE_COMPRESS) { 135e27425d6SJosef Bacik BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS; 136e27425d6SJosef Bacik BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS; 137e27425d6SJosef Bacik } 1386cbff00fSChristoph Hellwig 139e27425d6SJosef Bacik if (flags & BTRFS_INODE_NODATACOW) 140e27425d6SJosef Bacik BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW; 141e27425d6SJosef Bacik 1426cbff00fSChristoph Hellwig btrfs_update_iflags(inode); 1436cbff00fSChristoph Hellwig } 1446cbff00fSChristoph Hellwig 1456cbff00fSChristoph Hellwig static int btrfs_ioctl_getflags(struct file *file, void __user *arg) 1466cbff00fSChristoph Hellwig { 1476cbff00fSChristoph Hellwig struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode); 1486cbff00fSChristoph Hellwig unsigned int flags = btrfs_flags_to_ioctl(ip->flags); 1496cbff00fSChristoph Hellwig 1506cbff00fSChristoph Hellwig if (copy_to_user(arg, &flags, sizeof(flags))) 1516cbff00fSChristoph Hellwig return -EFAULT; 1526cbff00fSChristoph Hellwig return 0; 1536cbff00fSChristoph Hellwig } 1546cbff00fSChristoph Hellwig 15575e7cb7fSLiu Bo static int check_flags(unsigned int flags) 15675e7cb7fSLiu Bo { 15775e7cb7fSLiu Bo if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \ 15875e7cb7fSLiu Bo FS_NOATIME_FL | FS_NODUMP_FL | \ 15975e7cb7fSLiu Bo FS_SYNC_FL | FS_DIRSYNC_FL | \ 160e1e8fb6aSLi Zefan FS_NOCOMP_FL | FS_COMPR_FL | 161e1e8fb6aSLi Zefan FS_NOCOW_FL)) 16275e7cb7fSLiu Bo return -EOPNOTSUPP; 16375e7cb7fSLiu Bo 16475e7cb7fSLiu Bo if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL)) 16575e7cb7fSLiu Bo return -EINVAL; 16675e7cb7fSLiu Bo 16775e7cb7fSLiu Bo return 0; 16875e7cb7fSLiu Bo } 16975e7cb7fSLiu Bo 1706cbff00fSChristoph Hellwig static int btrfs_ioctl_setflags(struct file *file, void __user *arg) 1716cbff00fSChristoph Hellwig { 1726cbff00fSChristoph Hellwig struct inode *inode = file->f_path.dentry->d_inode; 1736cbff00fSChristoph Hellwig struct btrfs_inode *ip = BTRFS_I(inode); 1746cbff00fSChristoph Hellwig struct btrfs_root *root = ip->root; 1756cbff00fSChristoph Hellwig struct btrfs_trans_handle *trans; 1766cbff00fSChristoph Hellwig unsigned int flags, oldflags; 1776cbff00fSChristoph Hellwig int ret; 1786cbff00fSChristoph Hellwig 179b83cc969SLi Zefan if (btrfs_root_readonly(root)) 180b83cc969SLi Zefan return -EROFS; 181b83cc969SLi Zefan 1826cbff00fSChristoph Hellwig if (copy_from_user(&flags, arg, sizeof(flags))) 1836cbff00fSChristoph Hellwig return -EFAULT; 1846cbff00fSChristoph Hellwig 18575e7cb7fSLiu Bo ret = check_flags(flags); 18675e7cb7fSLiu Bo if (ret) 18775e7cb7fSLiu Bo return ret; 1886cbff00fSChristoph Hellwig 1892e149670SSerge E. Hallyn if (!inode_owner_or_capable(inode)) 1906cbff00fSChristoph Hellwig return -EACCES; 1916cbff00fSChristoph Hellwig 1926cbff00fSChristoph Hellwig mutex_lock(&inode->i_mutex); 1936cbff00fSChristoph Hellwig 1946cbff00fSChristoph Hellwig flags = btrfs_mask_flags(inode->i_mode, flags); 1956cbff00fSChristoph Hellwig oldflags = btrfs_flags_to_ioctl(ip->flags); 1966cbff00fSChristoph Hellwig if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) { 1976cbff00fSChristoph Hellwig if (!capable(CAP_LINUX_IMMUTABLE)) { 1986cbff00fSChristoph Hellwig ret = -EPERM; 1996cbff00fSChristoph Hellwig goto out_unlock; 2006cbff00fSChristoph Hellwig } 2016cbff00fSChristoph Hellwig } 2026cbff00fSChristoph Hellwig 2036cbff00fSChristoph Hellwig ret = mnt_want_write(file->f_path.mnt); 2046cbff00fSChristoph Hellwig if (ret) 2056cbff00fSChristoph Hellwig goto out_unlock; 2066cbff00fSChristoph Hellwig 2076cbff00fSChristoph Hellwig if (flags & FS_SYNC_FL) 2086cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_SYNC; 2096cbff00fSChristoph Hellwig else 2106cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_SYNC; 2116cbff00fSChristoph Hellwig if (flags & FS_IMMUTABLE_FL) 2126cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_IMMUTABLE; 2136cbff00fSChristoph Hellwig else 2146cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_IMMUTABLE; 2156cbff00fSChristoph Hellwig if (flags & FS_APPEND_FL) 2166cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_APPEND; 2176cbff00fSChristoph Hellwig else 2186cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_APPEND; 2196cbff00fSChristoph Hellwig if (flags & FS_NODUMP_FL) 2206cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_NODUMP; 2216cbff00fSChristoph Hellwig else 2226cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_NODUMP; 2236cbff00fSChristoph Hellwig if (flags & FS_NOATIME_FL) 2246cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_NOATIME; 2256cbff00fSChristoph Hellwig else 2266cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_NOATIME; 2276cbff00fSChristoph Hellwig if (flags & FS_DIRSYNC_FL) 2286cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_DIRSYNC; 2296cbff00fSChristoph Hellwig else 2306cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_DIRSYNC; 231e1e8fb6aSLi Zefan if (flags & FS_NOCOW_FL) 232e1e8fb6aSLi Zefan ip->flags |= BTRFS_INODE_NODATACOW; 233e1e8fb6aSLi Zefan else 234e1e8fb6aSLi Zefan ip->flags &= ~BTRFS_INODE_NODATACOW; 2356cbff00fSChristoph Hellwig 23675e7cb7fSLiu Bo /* 23775e7cb7fSLiu Bo * The COMPRESS flag can only be changed by users, while the NOCOMPRESS 23875e7cb7fSLiu Bo * flag may be changed automatically if compression code won't make 23975e7cb7fSLiu Bo * things smaller. 24075e7cb7fSLiu Bo */ 24175e7cb7fSLiu Bo if (flags & FS_NOCOMP_FL) { 24275e7cb7fSLiu Bo ip->flags &= ~BTRFS_INODE_COMPRESS; 24375e7cb7fSLiu Bo ip->flags |= BTRFS_INODE_NOCOMPRESS; 24475e7cb7fSLiu Bo } else if (flags & FS_COMPR_FL) { 24575e7cb7fSLiu Bo ip->flags |= BTRFS_INODE_COMPRESS; 24675e7cb7fSLiu Bo ip->flags &= ~BTRFS_INODE_NOCOMPRESS; 247ebcb904dSLi Zefan } else { 248ebcb904dSLi Zefan ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS); 24975e7cb7fSLiu Bo } 2506cbff00fSChristoph Hellwig 2517a7eaa40SJosef Bacik trans = btrfs_join_transaction(root); 2523612b495STsutomu Itoh BUG_ON(IS_ERR(trans)); 2536cbff00fSChristoph Hellwig 2546cbff00fSChristoph Hellwig ret = btrfs_update_inode(trans, root, inode); 2556cbff00fSChristoph Hellwig BUG_ON(ret); 2566cbff00fSChristoph Hellwig 2576cbff00fSChristoph Hellwig btrfs_update_iflags(inode); 2586cbff00fSChristoph Hellwig inode->i_ctime = CURRENT_TIME; 2596cbff00fSChristoph Hellwig btrfs_end_transaction(trans, root); 2606cbff00fSChristoph Hellwig 2616cbff00fSChristoph Hellwig mnt_drop_write(file->f_path.mnt); 2622d4e6f6aSliubo 2632d4e6f6aSliubo ret = 0; 2646cbff00fSChristoph Hellwig out_unlock: 2656cbff00fSChristoph Hellwig mutex_unlock(&inode->i_mutex); 2662d4e6f6aSliubo return ret; 2676cbff00fSChristoph Hellwig } 2686cbff00fSChristoph Hellwig 2696cbff00fSChristoph Hellwig static int btrfs_ioctl_getversion(struct file *file, int __user *arg) 2706cbff00fSChristoph Hellwig { 2716cbff00fSChristoph Hellwig struct inode *inode = file->f_path.dentry->d_inode; 2726cbff00fSChristoph Hellwig 2736cbff00fSChristoph Hellwig return put_user(inode->i_generation, arg); 2746cbff00fSChristoph Hellwig } 275f46b5a66SChristoph Hellwig 276f7039b1dSLi Dongyang static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg) 277f7039b1dSLi Dongyang { 278f7039b1dSLi Dongyang struct btrfs_root *root = fdentry(file)->d_sb->s_fs_info; 279f7039b1dSLi Dongyang struct btrfs_fs_info *fs_info = root->fs_info; 280f7039b1dSLi Dongyang struct btrfs_device *device; 281f7039b1dSLi Dongyang struct request_queue *q; 282f7039b1dSLi Dongyang struct fstrim_range range; 283f7039b1dSLi Dongyang u64 minlen = ULLONG_MAX; 284f7039b1dSLi Dongyang u64 num_devices = 0; 285f7039b1dSLi Dongyang int ret; 286f7039b1dSLi Dongyang 287f7039b1dSLi Dongyang if (!capable(CAP_SYS_ADMIN)) 288f7039b1dSLi Dongyang return -EPERM; 289f7039b1dSLi Dongyang 2901f78160cSXiao Guangrong rcu_read_lock(); 2911f78160cSXiao Guangrong list_for_each_entry_rcu(device, &fs_info->fs_devices->devices, 2921f78160cSXiao Guangrong dev_list) { 293f7039b1dSLi Dongyang if (!device->bdev) 294f7039b1dSLi Dongyang continue; 295f7039b1dSLi Dongyang q = bdev_get_queue(device->bdev); 296f7039b1dSLi Dongyang if (blk_queue_discard(q)) { 297f7039b1dSLi Dongyang num_devices++; 298f7039b1dSLi Dongyang minlen = min((u64)q->limits.discard_granularity, 299f7039b1dSLi Dongyang minlen); 300f7039b1dSLi Dongyang } 301f7039b1dSLi Dongyang } 3021f78160cSXiao Guangrong rcu_read_unlock(); 303f7039b1dSLi Dongyang if (!num_devices) 304f7039b1dSLi Dongyang return -EOPNOTSUPP; 305f7039b1dSLi Dongyang 306f7039b1dSLi Dongyang if (copy_from_user(&range, arg, sizeof(range))) 307f7039b1dSLi Dongyang return -EFAULT; 308f7039b1dSLi Dongyang 309f7039b1dSLi Dongyang range.minlen = max(range.minlen, minlen); 310f7039b1dSLi Dongyang ret = btrfs_trim_fs(root, &range); 311f7039b1dSLi Dongyang if (ret < 0) 312f7039b1dSLi Dongyang return ret; 313f7039b1dSLi Dongyang 314f7039b1dSLi Dongyang if (copy_to_user(arg, &range, sizeof(range))) 315f7039b1dSLi Dongyang return -EFAULT; 316f7039b1dSLi Dongyang 317f7039b1dSLi Dongyang return 0; 318f7039b1dSLi Dongyang } 319f7039b1dSLi Dongyang 320cb8e7090SChristoph Hellwig static noinline int create_subvol(struct btrfs_root *root, 321cb8e7090SChristoph Hellwig struct dentry *dentry, 32272fd032eSSage Weil char *name, int namelen, 32372fd032eSSage Weil u64 *async_transid) 324f46b5a66SChristoph Hellwig { 325f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 326f46b5a66SChristoph Hellwig struct btrfs_key key; 327f46b5a66SChristoph Hellwig struct btrfs_root_item root_item; 328f46b5a66SChristoph Hellwig struct btrfs_inode_item *inode_item; 329f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 33076dda93cSYan, Zheng struct btrfs_root *new_root; 3312fbe8c8aSAl Viro struct dentry *parent = dentry->d_parent; 3326a912213SJosef Bacik struct inode *dir; 333f46b5a66SChristoph Hellwig int ret; 334f46b5a66SChristoph Hellwig int err; 335f46b5a66SChristoph Hellwig u64 objectid; 336f46b5a66SChristoph Hellwig u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; 3373de4586cSChris Mason u64 index = 0; 338f46b5a66SChristoph Hellwig 339581bb050SLi Zefan ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid); 3402fbe8c8aSAl Viro if (ret) 341a22285a6SYan, Zheng return ret; 3426a912213SJosef Bacik 3436a912213SJosef Bacik dir = parent->d_inode; 3446a912213SJosef Bacik 3459ed74f2dSJosef Bacik /* 3469ed74f2dSJosef Bacik * 1 - inode item 3479ed74f2dSJosef Bacik * 2 - refs 3489ed74f2dSJosef Bacik * 1 - root item 3499ed74f2dSJosef Bacik * 2 - dir items 3509ed74f2dSJosef Bacik */ 351a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 6); 3522fbe8c8aSAl Viro if (IS_ERR(trans)) 353a22285a6SYan, Zheng return PTR_ERR(trans); 354f46b5a66SChristoph Hellwig 3555d4f98a2SYan Zheng leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 3565d4f98a2SYan Zheng 0, objectid, NULL, 0, 0, 0); 3578e8a1e31SJosef Bacik if (IS_ERR(leaf)) { 3588e8a1e31SJosef Bacik ret = PTR_ERR(leaf); 3598e8a1e31SJosef Bacik goto fail; 3608e8a1e31SJosef Bacik } 361f46b5a66SChristoph Hellwig 3625d4f98a2SYan Zheng memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); 363f46b5a66SChristoph Hellwig btrfs_set_header_bytenr(leaf, leaf->start); 364f46b5a66SChristoph Hellwig btrfs_set_header_generation(leaf, trans->transid); 3655d4f98a2SYan Zheng btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 366f46b5a66SChristoph Hellwig btrfs_set_header_owner(leaf, objectid); 367f46b5a66SChristoph Hellwig 368f46b5a66SChristoph Hellwig write_extent_buffer(leaf, root->fs_info->fsid, 369f46b5a66SChristoph Hellwig (unsigned long)btrfs_header_fsid(leaf), 370f46b5a66SChristoph Hellwig BTRFS_FSID_SIZE); 3715d4f98a2SYan Zheng write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid, 3725d4f98a2SYan Zheng (unsigned long)btrfs_header_chunk_tree_uuid(leaf), 3735d4f98a2SYan Zheng BTRFS_UUID_SIZE); 374f46b5a66SChristoph Hellwig btrfs_mark_buffer_dirty(leaf); 375f46b5a66SChristoph Hellwig 376f46b5a66SChristoph Hellwig inode_item = &root_item.inode; 377f46b5a66SChristoph Hellwig memset(inode_item, 0, sizeof(*inode_item)); 378f46b5a66SChristoph Hellwig inode_item->generation = cpu_to_le64(1); 379f46b5a66SChristoph Hellwig inode_item->size = cpu_to_le64(3); 380f46b5a66SChristoph Hellwig inode_item->nlink = cpu_to_le32(1); 381a76a3cd4SYan Zheng inode_item->nbytes = cpu_to_le64(root->leafsize); 382f46b5a66SChristoph Hellwig inode_item->mode = cpu_to_le32(S_IFDIR | 0755); 383f46b5a66SChristoph Hellwig 38408fe4db1SLi Zefan root_item.flags = 0; 38508fe4db1SLi Zefan root_item.byte_limit = 0; 38608fe4db1SLi Zefan inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT); 38708fe4db1SLi Zefan 388f46b5a66SChristoph Hellwig btrfs_set_root_bytenr(&root_item, leaf->start); 38984234f3aSYan Zheng btrfs_set_root_generation(&root_item, trans->transid); 390f46b5a66SChristoph Hellwig btrfs_set_root_level(&root_item, 0); 391f46b5a66SChristoph Hellwig btrfs_set_root_refs(&root_item, 1); 39286b9f2ecSYan, Zheng btrfs_set_root_used(&root_item, leaf->len); 39380ff3856SYan Zheng btrfs_set_root_last_snapshot(&root_item, 0); 394f46b5a66SChristoph Hellwig 395f46b5a66SChristoph Hellwig memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress)); 396f46b5a66SChristoph Hellwig root_item.drop_level = 0; 397f46b5a66SChristoph Hellwig 398925baeddSChris Mason btrfs_tree_unlock(leaf); 399f46b5a66SChristoph Hellwig free_extent_buffer(leaf); 400f46b5a66SChristoph Hellwig leaf = NULL; 401f46b5a66SChristoph Hellwig 402f46b5a66SChristoph Hellwig btrfs_set_root_dirid(&root_item, new_dirid); 403f46b5a66SChristoph Hellwig 404f46b5a66SChristoph Hellwig key.objectid = objectid; 4055d4f98a2SYan Zheng key.offset = 0; 406f46b5a66SChristoph Hellwig btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); 407f46b5a66SChristoph Hellwig ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, 408f46b5a66SChristoph Hellwig &root_item); 409f46b5a66SChristoph Hellwig if (ret) 410f46b5a66SChristoph Hellwig goto fail; 411f46b5a66SChristoph Hellwig 41276dda93cSYan, Zheng key.offset = (u64)-1; 41376dda93cSYan, Zheng new_root = btrfs_read_fs_root_no_name(root->fs_info, &key); 41476dda93cSYan, Zheng BUG_ON(IS_ERR(new_root)); 41576dda93cSYan, Zheng 41676dda93cSYan, Zheng btrfs_record_root_in_trans(trans, new_root); 41776dda93cSYan, Zheng 418d82a6f1dSJosef Bacik ret = btrfs_create_subvol_root(trans, new_root, new_dirid); 419f46b5a66SChristoph Hellwig /* 420f46b5a66SChristoph Hellwig * insert the directory item 421f46b5a66SChristoph Hellwig */ 4223de4586cSChris Mason ret = btrfs_set_inode_index(dir, &index); 4233de4586cSChris Mason BUG_ON(ret); 4243de4586cSChris Mason 4253de4586cSChris Mason ret = btrfs_insert_dir_item(trans, root, 42616cdcec7SMiao Xie name, namelen, dir, &key, 4273de4586cSChris Mason BTRFS_FT_DIR, index); 428f46b5a66SChristoph Hellwig if (ret) 429f46b5a66SChristoph Hellwig goto fail; 4300660b5afSChris Mason 43152c26179SYan Zheng btrfs_i_size_write(dir, dir->i_size + namelen * 2); 43252c26179SYan Zheng ret = btrfs_update_inode(trans, root, dir); 43352c26179SYan Zheng BUG_ON(ret); 43452c26179SYan Zheng 4350660b5afSChris Mason ret = btrfs_add_root_ref(trans, root->fs_info->tree_root, 4364df27c4dSYan, Zheng objectid, root->root_key.objectid, 43733345d01SLi Zefan btrfs_ino(dir), index, name, namelen); 43876dda93cSYan, Zheng 4390660b5afSChris Mason BUG_ON(ret); 4400660b5afSChris Mason 44176dda93cSYan, Zheng d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry)); 442f46b5a66SChristoph Hellwig fail: 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; 459f46b5a66SChristoph Hellwig struct btrfs_pending_snapshot *pending_snapshot; 460f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 4612e4bfab9SYan, Zheng int ret; 462f46b5a66SChristoph Hellwig 463f46b5a66SChristoph Hellwig if (!root->ref_cows) 464f46b5a66SChristoph Hellwig return -EINVAL; 465f46b5a66SChristoph Hellwig 466a22285a6SYan, Zheng pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS); 467a22285a6SYan, Zheng if (!pending_snapshot) 468a22285a6SYan, Zheng return -ENOMEM; 469a22285a6SYan, Zheng 470a22285a6SYan, Zheng btrfs_init_block_rsv(&pending_snapshot->block_rsv); 471a22285a6SYan, Zheng pending_snapshot->dentry = dentry; 472a22285a6SYan, Zheng pending_snapshot->root = root; 473b83cc969SLi Zefan pending_snapshot->readonly = readonly; 474a22285a6SYan, Zheng 475a22285a6SYan, Zheng trans = btrfs_start_transaction(root->fs_info->extent_root, 5); 476a22285a6SYan, Zheng if (IS_ERR(trans)) { 477a22285a6SYan, Zheng ret = PTR_ERR(trans); 478a22285a6SYan, Zheng goto fail; 479a22285a6SYan, Zheng } 480a22285a6SYan, Zheng 481a22285a6SYan, Zheng ret = btrfs_snap_reserve_metadata(trans, pending_snapshot); 482a22285a6SYan, Zheng BUG_ON(ret); 483a22285a6SYan, Zheng 4848351583eSJosef Bacik spin_lock(&root->fs_info->trans_lock); 485a22285a6SYan, Zheng list_add(&pending_snapshot->list, 486a22285a6SYan, Zheng &trans->transaction->pending_snapshots); 4878351583eSJosef Bacik spin_unlock(&root->fs_info->trans_lock); 48872fd032eSSage Weil if (async_transid) { 48972fd032eSSage Weil *async_transid = trans->transid; 49072fd032eSSage Weil ret = btrfs_commit_transaction_async(trans, 49172fd032eSSage Weil root->fs_info->extent_root, 1); 49272fd032eSSage Weil } else { 49372fd032eSSage Weil ret = btrfs_commit_transaction(trans, 49472fd032eSSage Weil root->fs_info->extent_root); 49572fd032eSSage Weil } 496a22285a6SYan, Zheng BUG_ON(ret); 497a22285a6SYan, Zheng 498a22285a6SYan, Zheng ret = pending_snapshot->error; 499f46b5a66SChristoph Hellwig if (ret) 5002e4bfab9SYan, Zheng goto fail; 501f46b5a66SChristoph Hellwig 50266b4ffd1SJosef Bacik ret = btrfs_orphan_cleanup(pending_snapshot->snap); 50366b4ffd1SJosef Bacik if (ret) 50466b4ffd1SJosef Bacik goto fail; 505f46b5a66SChristoph Hellwig 5062fbe8c8aSAl Viro inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry); 5072e4bfab9SYan, Zheng if (IS_ERR(inode)) { 5082e4bfab9SYan, Zheng ret = PTR_ERR(inode); 5092e4bfab9SYan, Zheng goto fail; 5102e4bfab9SYan, Zheng } 5112e4bfab9SYan, Zheng BUG_ON(!inode); 5122e4bfab9SYan, Zheng d_instantiate(dentry, inode); 5132e4bfab9SYan, Zheng ret = 0; 5142e4bfab9SYan, Zheng fail: 515a22285a6SYan, Zheng kfree(pending_snapshot); 516f46b5a66SChristoph Hellwig return ret; 517f46b5a66SChristoph Hellwig } 518f46b5a66SChristoph Hellwig 5194260f7c7SSage Weil /* copy of check_sticky in fs/namei.c() 5204260f7c7SSage Weil * It's inline, so penalty for filesystems that don't use sticky bit is 5214260f7c7SSage Weil * minimal. 5224260f7c7SSage Weil */ 5234260f7c7SSage Weil static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode) 5244260f7c7SSage Weil { 5254260f7c7SSage Weil uid_t fsuid = current_fsuid(); 5264260f7c7SSage Weil 5274260f7c7SSage Weil if (!(dir->i_mode & S_ISVTX)) 5284260f7c7SSage Weil return 0; 5294260f7c7SSage Weil if (inode->i_uid == fsuid) 5304260f7c7SSage Weil return 0; 5314260f7c7SSage Weil if (dir->i_uid == fsuid) 5324260f7c7SSage Weil return 0; 5334260f7c7SSage Weil return !capable(CAP_FOWNER); 5344260f7c7SSage Weil } 5354260f7c7SSage Weil 5364260f7c7SSage Weil /* copy of may_delete in fs/namei.c() 5374260f7c7SSage Weil * Check whether we can remove a link victim from directory dir, check 5384260f7c7SSage Weil * whether the type of victim is right. 5394260f7c7SSage Weil * 1. We can't do it if dir is read-only (done in permission()) 5404260f7c7SSage Weil * 2. We should have write and exec permissions on dir 5414260f7c7SSage Weil * 3. We can't remove anything from append-only dir 5424260f7c7SSage Weil * 4. We can't do anything with immutable dir (done in permission()) 5434260f7c7SSage Weil * 5. If the sticky bit on dir is set we should either 5444260f7c7SSage Weil * a. be owner of dir, or 5454260f7c7SSage Weil * b. be owner of victim, or 5464260f7c7SSage Weil * c. have CAP_FOWNER capability 5474260f7c7SSage Weil * 6. If the victim is append-only or immutable we can't do antyhing with 5484260f7c7SSage Weil * links pointing to it. 5494260f7c7SSage Weil * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. 5504260f7c7SSage Weil * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. 5514260f7c7SSage Weil * 9. We can't remove a root or mountpoint. 5524260f7c7SSage Weil * 10. We don't allow removal of NFS sillyrenamed files; it's handled by 5534260f7c7SSage Weil * nfs_async_unlink(). 5544260f7c7SSage Weil */ 5554260f7c7SSage Weil 5564260f7c7SSage Weil static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir) 5574260f7c7SSage Weil { 5584260f7c7SSage Weil int error; 5594260f7c7SSage Weil 5604260f7c7SSage Weil if (!victim->d_inode) 5614260f7c7SSage Weil return -ENOENT; 5624260f7c7SSage Weil 5634260f7c7SSage Weil BUG_ON(victim->d_parent->d_inode != dir); 5644260f7c7SSage Weil audit_inode_child(victim, dir); 5654260f7c7SSage Weil 5664260f7c7SSage Weil error = inode_permission(dir, MAY_WRITE | MAY_EXEC); 5674260f7c7SSage Weil if (error) 5684260f7c7SSage Weil return error; 5694260f7c7SSage Weil if (IS_APPEND(dir)) 5704260f7c7SSage Weil return -EPERM; 5714260f7c7SSage Weil if (btrfs_check_sticky(dir, victim->d_inode)|| 5724260f7c7SSage Weil IS_APPEND(victim->d_inode)|| 5734260f7c7SSage Weil IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode)) 5744260f7c7SSage Weil return -EPERM; 5754260f7c7SSage Weil if (isdir) { 5764260f7c7SSage Weil if (!S_ISDIR(victim->d_inode->i_mode)) 5774260f7c7SSage Weil return -ENOTDIR; 5784260f7c7SSage Weil if (IS_ROOT(victim)) 5794260f7c7SSage Weil return -EBUSY; 5804260f7c7SSage Weil } else if (S_ISDIR(victim->d_inode->i_mode)) 5814260f7c7SSage Weil return -EISDIR; 5824260f7c7SSage Weil if (IS_DEADDIR(dir)) 5834260f7c7SSage Weil return -ENOENT; 5844260f7c7SSage Weil if (victim->d_flags & DCACHE_NFSFS_RENAMED) 5854260f7c7SSage Weil return -EBUSY; 5864260f7c7SSage Weil return 0; 5874260f7c7SSage Weil } 5884260f7c7SSage Weil 589cb8e7090SChristoph Hellwig /* copy of may_create in fs/namei.c() */ 590cb8e7090SChristoph Hellwig static inline int btrfs_may_create(struct inode *dir, struct dentry *child) 591cb8e7090SChristoph Hellwig { 592cb8e7090SChristoph Hellwig if (child->d_inode) 593cb8e7090SChristoph Hellwig return -EEXIST; 594cb8e7090SChristoph Hellwig if (IS_DEADDIR(dir)) 595cb8e7090SChristoph Hellwig return -ENOENT; 596cb8e7090SChristoph Hellwig return inode_permission(dir, MAY_WRITE | MAY_EXEC); 597cb8e7090SChristoph Hellwig } 598cb8e7090SChristoph Hellwig 599cb8e7090SChristoph Hellwig /* 600cb8e7090SChristoph Hellwig * Create a new subvolume below @parent. This is largely modeled after 601cb8e7090SChristoph Hellwig * sys_mkdirat and vfs_mkdir, but we only do a single component lookup 602cb8e7090SChristoph Hellwig * inside this filesystem so it's quite a bit simpler. 603cb8e7090SChristoph Hellwig */ 60476dda93cSYan, Zheng static noinline int btrfs_mksubvol(struct path *parent, 60576dda93cSYan, Zheng char *name, int namelen, 60672fd032eSSage Weil struct btrfs_root *snap_src, 607b83cc969SLi Zefan u64 *async_transid, bool readonly) 608cb8e7090SChristoph Hellwig { 60976dda93cSYan, Zheng struct inode *dir = parent->dentry->d_inode; 610cb8e7090SChristoph Hellwig struct dentry *dentry; 611cb8e7090SChristoph Hellwig int error; 612cb8e7090SChristoph Hellwig 61376dda93cSYan, Zheng mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT); 614cb8e7090SChristoph Hellwig 615cb8e7090SChristoph Hellwig dentry = lookup_one_len(name, parent->dentry, namelen); 616cb8e7090SChristoph Hellwig error = PTR_ERR(dentry); 617cb8e7090SChristoph Hellwig if (IS_ERR(dentry)) 618cb8e7090SChristoph Hellwig goto out_unlock; 619cb8e7090SChristoph Hellwig 620cb8e7090SChristoph Hellwig error = -EEXIST; 621cb8e7090SChristoph Hellwig if (dentry->d_inode) 622cb8e7090SChristoph Hellwig goto out_dput; 623cb8e7090SChristoph Hellwig 624cb8e7090SChristoph Hellwig error = mnt_want_write(parent->mnt); 625cb8e7090SChristoph Hellwig if (error) 626cb8e7090SChristoph Hellwig goto out_dput; 627cb8e7090SChristoph Hellwig 62876dda93cSYan, Zheng error = btrfs_may_create(dir, dentry); 629cb8e7090SChristoph Hellwig if (error) 630cb8e7090SChristoph Hellwig goto out_drop_write; 631cb8e7090SChristoph Hellwig 63276dda93cSYan, Zheng down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem); 63376dda93cSYan, Zheng 63476dda93cSYan, Zheng if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) 63576dda93cSYan, Zheng goto out_up_read; 63676dda93cSYan, Zheng 6373de4586cSChris Mason if (snap_src) { 63872fd032eSSage Weil error = create_snapshot(snap_src, dentry, 639b83cc969SLi Zefan name, namelen, async_transid, readonly); 6403de4586cSChris Mason } else { 64176dda93cSYan, Zheng error = create_subvol(BTRFS_I(dir)->root, dentry, 64272fd032eSSage Weil name, namelen, async_transid); 6433de4586cSChris Mason } 64476dda93cSYan, Zheng if (!error) 64576dda93cSYan, Zheng fsnotify_mkdir(dir, dentry); 64676dda93cSYan, Zheng out_up_read: 64776dda93cSYan, Zheng up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem); 648cb8e7090SChristoph Hellwig out_drop_write: 649cb8e7090SChristoph Hellwig mnt_drop_write(parent->mnt); 650cb8e7090SChristoph Hellwig out_dput: 651cb8e7090SChristoph Hellwig dput(dentry); 652cb8e7090SChristoph Hellwig out_unlock: 65376dda93cSYan, Zheng mutex_unlock(&dir->i_mutex); 654cb8e7090SChristoph Hellwig return error; 655cb8e7090SChristoph Hellwig } 656cb8e7090SChristoph Hellwig 6574cb5300bSChris Mason /* 6584cb5300bSChris Mason * When we're defragging a range, we don't want to kick it off again 6594cb5300bSChris Mason * if it is really just waiting for delalloc to send it down. 6604cb5300bSChris Mason * If we find a nice big extent or delalloc range for the bytes in the 6614cb5300bSChris Mason * file you want to defrag, we return 0 to let you know to skip this 6624cb5300bSChris Mason * part of the file 6634cb5300bSChris Mason */ 6644cb5300bSChris Mason static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh) 6654cb5300bSChris Mason { 6664cb5300bSChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 6674cb5300bSChris Mason struct extent_map *em = NULL; 6684cb5300bSChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 6694cb5300bSChris Mason u64 end; 6704cb5300bSChris Mason 6714cb5300bSChris Mason read_lock(&em_tree->lock); 6724cb5300bSChris Mason em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE); 6734cb5300bSChris Mason read_unlock(&em_tree->lock); 6744cb5300bSChris Mason 6754cb5300bSChris Mason if (em) { 6764cb5300bSChris Mason end = extent_map_end(em); 6774cb5300bSChris Mason free_extent_map(em); 6784cb5300bSChris Mason if (end - offset > thresh) 6794cb5300bSChris Mason return 0; 6804cb5300bSChris Mason } 6814cb5300bSChris Mason /* if we already have a nice delalloc here, just stop */ 6824cb5300bSChris Mason thresh /= 2; 6834cb5300bSChris Mason end = count_range_bits(io_tree, &offset, offset + thresh, 6844cb5300bSChris Mason thresh, EXTENT_DELALLOC, 1); 6854cb5300bSChris Mason if (end >= thresh) 6864cb5300bSChris Mason return 0; 6874cb5300bSChris Mason return 1; 6884cb5300bSChris Mason } 6894cb5300bSChris Mason 6904cb5300bSChris Mason /* 6914cb5300bSChris Mason * helper function to walk through a file and find extents 6924cb5300bSChris Mason * newer than a specific transid, and smaller than thresh. 6934cb5300bSChris Mason * 6944cb5300bSChris Mason * This is used by the defragging code to find new and small 6954cb5300bSChris Mason * extents 6964cb5300bSChris Mason */ 6974cb5300bSChris Mason static int find_new_extents(struct btrfs_root *root, 6984cb5300bSChris Mason struct inode *inode, u64 newer_than, 6994cb5300bSChris Mason u64 *off, int thresh) 7004cb5300bSChris Mason { 7014cb5300bSChris Mason struct btrfs_path *path; 7024cb5300bSChris Mason struct btrfs_key min_key; 7034cb5300bSChris Mason struct btrfs_key max_key; 7044cb5300bSChris Mason struct extent_buffer *leaf; 7054cb5300bSChris Mason struct btrfs_file_extent_item *extent; 7064cb5300bSChris Mason int type; 7074cb5300bSChris Mason int ret; 708a4689d2bSDavid Sterba u64 ino = btrfs_ino(inode); 7094cb5300bSChris Mason 7104cb5300bSChris Mason path = btrfs_alloc_path(); 7114cb5300bSChris Mason if (!path) 7124cb5300bSChris Mason return -ENOMEM; 7134cb5300bSChris Mason 714a4689d2bSDavid Sterba min_key.objectid = ino; 7154cb5300bSChris Mason min_key.type = BTRFS_EXTENT_DATA_KEY; 7164cb5300bSChris Mason min_key.offset = *off; 7174cb5300bSChris Mason 718a4689d2bSDavid Sterba max_key.objectid = ino; 7194cb5300bSChris Mason max_key.type = (u8)-1; 7204cb5300bSChris Mason max_key.offset = (u64)-1; 7214cb5300bSChris Mason 7224cb5300bSChris Mason path->keep_locks = 1; 7234cb5300bSChris Mason 7244cb5300bSChris Mason while(1) { 7254cb5300bSChris Mason ret = btrfs_search_forward(root, &min_key, &max_key, 7264cb5300bSChris Mason path, 0, newer_than); 7274cb5300bSChris Mason if (ret != 0) 7284cb5300bSChris Mason goto none; 729a4689d2bSDavid Sterba if (min_key.objectid != ino) 7304cb5300bSChris Mason goto none; 7314cb5300bSChris Mason if (min_key.type != BTRFS_EXTENT_DATA_KEY) 7324cb5300bSChris Mason goto none; 7334cb5300bSChris Mason 7344cb5300bSChris Mason leaf = path->nodes[0]; 7354cb5300bSChris Mason extent = btrfs_item_ptr(leaf, path->slots[0], 7364cb5300bSChris Mason struct btrfs_file_extent_item); 7374cb5300bSChris Mason 7384cb5300bSChris Mason type = btrfs_file_extent_type(leaf, extent); 7394cb5300bSChris Mason if (type == BTRFS_FILE_EXTENT_REG && 7404cb5300bSChris Mason btrfs_file_extent_num_bytes(leaf, extent) < thresh && 7414cb5300bSChris Mason check_defrag_in_cache(inode, min_key.offset, thresh)) { 7424cb5300bSChris Mason *off = min_key.offset; 7434cb5300bSChris Mason btrfs_free_path(path); 7444cb5300bSChris Mason return 0; 7454cb5300bSChris Mason } 7464cb5300bSChris Mason 7474cb5300bSChris Mason if (min_key.offset == (u64)-1) 7484cb5300bSChris Mason goto none; 7494cb5300bSChris Mason 7504cb5300bSChris Mason min_key.offset++; 7514cb5300bSChris Mason btrfs_release_path(path); 7524cb5300bSChris Mason } 7534cb5300bSChris Mason none: 7544cb5300bSChris Mason btrfs_free_path(path); 7554cb5300bSChris Mason return -ENOENT; 7564cb5300bSChris Mason } 7574cb5300bSChris Mason 758940100a4SChris Mason static int should_defrag_range(struct inode *inode, u64 start, u64 len, 7591e701a32SChris Mason int thresh, u64 *last_len, u64 *skip, 7601e701a32SChris Mason u64 *defrag_end) 761940100a4SChris Mason { 762940100a4SChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 763940100a4SChris Mason struct extent_map *em = NULL; 764940100a4SChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 765940100a4SChris Mason int ret = 1; 766940100a4SChris Mason 767940100a4SChris Mason /* 768940100a4SChris Mason * make sure that once we start defragging and extent, we keep on 769940100a4SChris Mason * defragging it 770940100a4SChris Mason */ 771940100a4SChris Mason if (start < *defrag_end) 772940100a4SChris Mason return 1; 773940100a4SChris Mason 774940100a4SChris Mason *skip = 0; 775940100a4SChris Mason 776940100a4SChris Mason /* 777940100a4SChris Mason * hopefully we have this extent in the tree already, try without 778940100a4SChris Mason * the full extent lock 779940100a4SChris Mason */ 780940100a4SChris Mason read_lock(&em_tree->lock); 781940100a4SChris Mason em = lookup_extent_mapping(em_tree, start, len); 782940100a4SChris Mason read_unlock(&em_tree->lock); 783940100a4SChris Mason 784940100a4SChris Mason if (!em) { 785940100a4SChris Mason /* get the big lock and read metadata off disk */ 786940100a4SChris Mason lock_extent(io_tree, start, start + len - 1, GFP_NOFS); 787940100a4SChris Mason em = btrfs_get_extent(inode, NULL, 0, start, len, 0); 788940100a4SChris Mason unlock_extent(io_tree, start, start + len - 1, GFP_NOFS); 789940100a4SChris Mason 7906cf8bfbfSDan Carpenter if (IS_ERR(em)) 791940100a4SChris Mason return 0; 792940100a4SChris Mason } 793940100a4SChris Mason 794940100a4SChris Mason /* this will cover holes, and inline extents */ 795940100a4SChris Mason if (em->block_start >= EXTENT_MAP_LAST_BYTE) 796940100a4SChris Mason ret = 0; 797940100a4SChris Mason 798940100a4SChris Mason /* 799940100a4SChris Mason * we hit a real extent, if it is big don't bother defragging it again 800940100a4SChris Mason */ 8011e701a32SChris Mason if ((*last_len == 0 || *last_len >= thresh) && em->len >= thresh) 802940100a4SChris Mason ret = 0; 803940100a4SChris Mason 804940100a4SChris Mason /* 805940100a4SChris Mason * last_len ends up being a counter of how many bytes we've defragged. 806940100a4SChris Mason * every time we choose not to defrag an extent, we reset *last_len 807940100a4SChris Mason * so that the next tiny extent will force a defrag. 808940100a4SChris Mason * 809940100a4SChris Mason * The end result of this is that tiny extents before a single big 810940100a4SChris Mason * extent will force at least part of that big extent to be defragged. 811940100a4SChris Mason */ 812940100a4SChris Mason if (ret) { 813940100a4SChris Mason *last_len += len; 814940100a4SChris Mason *defrag_end = extent_map_end(em); 815940100a4SChris Mason } else { 816940100a4SChris Mason *last_len = 0; 817940100a4SChris Mason *skip = extent_map_end(em); 818940100a4SChris Mason *defrag_end = 0; 819940100a4SChris Mason } 820940100a4SChris Mason 821940100a4SChris Mason free_extent_map(em); 822940100a4SChris Mason return ret; 823940100a4SChris Mason } 824940100a4SChris Mason 8254cb5300bSChris Mason /* 8264cb5300bSChris Mason * it doesn't do much good to defrag one or two pages 8274cb5300bSChris Mason * at a time. This pulls in a nice chunk of pages 8284cb5300bSChris Mason * to COW and defrag. 8294cb5300bSChris Mason * 8304cb5300bSChris Mason * It also makes sure the delalloc code has enough 8314cb5300bSChris Mason * dirty data to avoid making new small extents as part 8324cb5300bSChris Mason * of the defrag 8334cb5300bSChris Mason * 8344cb5300bSChris Mason * It's a good idea to start RA on this range 8354cb5300bSChris Mason * before calling this. 8364cb5300bSChris Mason */ 8374cb5300bSChris Mason static int cluster_pages_for_defrag(struct inode *inode, 8384cb5300bSChris Mason struct page **pages, 8394cb5300bSChris Mason unsigned long start_index, 8404cb5300bSChris Mason int num_pages) 841f46b5a66SChristoph Hellwig { 8424cb5300bSChris Mason unsigned long file_end; 8434cb5300bSChris Mason u64 isize = i_size_read(inode); 844f46b5a66SChristoph Hellwig u64 page_start; 845f46b5a66SChristoph Hellwig u64 page_end; 8464cb5300bSChris Mason int ret; 8474cb5300bSChris Mason int i; 8484cb5300bSChris Mason int i_done; 8494cb5300bSChris Mason struct btrfs_ordered_extent *ordered; 8504cb5300bSChris Mason struct extent_state *cached_state = NULL; 8513b16a4e3SJosef Bacik gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping); 8524cb5300bSChris Mason 8534cb5300bSChris Mason if (isize == 0) 8544cb5300bSChris Mason return 0; 8554cb5300bSChris Mason file_end = (isize - 1) >> PAGE_CACHE_SHIFT; 8564cb5300bSChris Mason 8574cb5300bSChris Mason ret = btrfs_delalloc_reserve_space(inode, 8584cb5300bSChris Mason num_pages << PAGE_CACHE_SHIFT); 8594cb5300bSChris Mason if (ret) 8604cb5300bSChris Mason return ret; 8614cb5300bSChris Mason again: 8624cb5300bSChris Mason ret = 0; 8634cb5300bSChris Mason i_done = 0; 8644cb5300bSChris Mason 8654cb5300bSChris Mason /* step one, lock all the pages */ 8664cb5300bSChris Mason for (i = 0; i < num_pages; i++) { 8674cb5300bSChris Mason struct page *page; 868a94733d0SJosef Bacik page = find_or_create_page(inode->i_mapping, 8693b16a4e3SJosef Bacik start_index + i, mask); 8704cb5300bSChris Mason if (!page) 8714cb5300bSChris Mason break; 8724cb5300bSChris Mason 8734cb5300bSChris Mason if (!PageUptodate(page)) { 8744cb5300bSChris Mason btrfs_readpage(NULL, page); 8754cb5300bSChris Mason lock_page(page); 8764cb5300bSChris Mason if (!PageUptodate(page)) { 8774cb5300bSChris Mason unlock_page(page); 8784cb5300bSChris Mason page_cache_release(page); 8794cb5300bSChris Mason ret = -EIO; 8804cb5300bSChris Mason break; 8814cb5300bSChris Mason } 8824cb5300bSChris Mason } 8834cb5300bSChris Mason isize = i_size_read(inode); 8844cb5300bSChris Mason file_end = (isize - 1) >> PAGE_CACHE_SHIFT; 8854cb5300bSChris Mason if (!isize || page->index > file_end || 8864cb5300bSChris Mason page->mapping != inode->i_mapping) { 8874cb5300bSChris Mason /* whoops, we blew past eof, skip this page */ 8884cb5300bSChris Mason unlock_page(page); 8894cb5300bSChris Mason page_cache_release(page); 8904cb5300bSChris Mason break; 8914cb5300bSChris Mason } 8924cb5300bSChris Mason pages[i] = page; 8934cb5300bSChris Mason i_done++; 8944cb5300bSChris Mason } 8954cb5300bSChris Mason if (!i_done || ret) 8964cb5300bSChris Mason goto out; 8974cb5300bSChris Mason 8984cb5300bSChris Mason if (!(inode->i_sb->s_flags & MS_ACTIVE)) 8994cb5300bSChris Mason goto out; 9004cb5300bSChris Mason 9014cb5300bSChris Mason /* 9024cb5300bSChris Mason * so now we have a nice long stream of locked 9034cb5300bSChris Mason * and up to date pages, lets wait on them 9044cb5300bSChris Mason */ 9054cb5300bSChris Mason for (i = 0; i < i_done; i++) 9064cb5300bSChris Mason wait_on_page_writeback(pages[i]); 9074cb5300bSChris Mason 9084cb5300bSChris Mason page_start = page_offset(pages[0]); 9094cb5300bSChris Mason page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE; 9104cb5300bSChris Mason 9114cb5300bSChris Mason lock_extent_bits(&BTRFS_I(inode)->io_tree, 9124cb5300bSChris Mason page_start, page_end - 1, 0, &cached_state, 9134cb5300bSChris Mason GFP_NOFS); 9144cb5300bSChris Mason ordered = btrfs_lookup_first_ordered_extent(inode, page_end - 1); 9154cb5300bSChris Mason if (ordered && 9164cb5300bSChris Mason ordered->file_offset + ordered->len > page_start && 9174cb5300bSChris Mason ordered->file_offset < page_end) { 9184cb5300bSChris Mason btrfs_put_ordered_extent(ordered); 9194cb5300bSChris Mason unlock_extent_cached(&BTRFS_I(inode)->io_tree, 9204cb5300bSChris Mason page_start, page_end - 1, 9214cb5300bSChris Mason &cached_state, GFP_NOFS); 9224cb5300bSChris Mason for (i = 0; i < i_done; i++) { 9234cb5300bSChris Mason unlock_page(pages[i]); 9244cb5300bSChris Mason page_cache_release(pages[i]); 9254cb5300bSChris Mason } 9264cb5300bSChris Mason btrfs_wait_ordered_range(inode, page_start, 9274cb5300bSChris Mason page_end - page_start); 9284cb5300bSChris Mason goto again; 9294cb5300bSChris Mason } 9304cb5300bSChris Mason if (ordered) 9314cb5300bSChris Mason btrfs_put_ordered_extent(ordered); 9324cb5300bSChris Mason 9334cb5300bSChris Mason clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, 9344cb5300bSChris Mason page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC | 9354cb5300bSChris Mason EXTENT_DO_ACCOUNTING, 0, 0, &cached_state, 9364cb5300bSChris Mason GFP_NOFS); 9374cb5300bSChris Mason 9384cb5300bSChris Mason if (i_done != num_pages) { 9399e0baf60SJosef Bacik spin_lock(&BTRFS_I(inode)->lock); 9409e0baf60SJosef Bacik BTRFS_I(inode)->outstanding_extents++; 9419e0baf60SJosef Bacik spin_unlock(&BTRFS_I(inode)->lock); 9424cb5300bSChris Mason btrfs_delalloc_release_space(inode, 9434cb5300bSChris Mason (num_pages - i_done) << PAGE_CACHE_SHIFT); 9444cb5300bSChris Mason } 9454cb5300bSChris Mason 9464cb5300bSChris Mason 9474cb5300bSChris Mason btrfs_set_extent_delalloc(inode, page_start, page_end - 1, 9484cb5300bSChris Mason &cached_state); 9494cb5300bSChris Mason 9504cb5300bSChris Mason unlock_extent_cached(&BTRFS_I(inode)->io_tree, 9514cb5300bSChris Mason page_start, page_end - 1, &cached_state, 9524cb5300bSChris Mason GFP_NOFS); 9534cb5300bSChris Mason 9544cb5300bSChris Mason for (i = 0; i < i_done; i++) { 9554cb5300bSChris Mason clear_page_dirty_for_io(pages[i]); 9564cb5300bSChris Mason ClearPageChecked(pages[i]); 9574cb5300bSChris Mason set_page_extent_mapped(pages[i]); 9584cb5300bSChris Mason set_page_dirty(pages[i]); 9594cb5300bSChris Mason unlock_page(pages[i]); 9604cb5300bSChris Mason page_cache_release(pages[i]); 9614cb5300bSChris Mason } 9624cb5300bSChris Mason return i_done; 9634cb5300bSChris Mason out: 9644cb5300bSChris Mason for (i = 0; i < i_done; i++) { 9654cb5300bSChris Mason unlock_page(pages[i]); 9664cb5300bSChris Mason page_cache_release(pages[i]); 9674cb5300bSChris Mason } 9684cb5300bSChris Mason btrfs_delalloc_release_space(inode, num_pages << PAGE_CACHE_SHIFT); 9694cb5300bSChris Mason return ret; 9704cb5300bSChris Mason 9714cb5300bSChris Mason } 9724cb5300bSChris Mason 9734cb5300bSChris Mason int btrfs_defrag_file(struct inode *inode, struct file *file, 9744cb5300bSChris Mason struct btrfs_ioctl_defrag_range_args *range, 9754cb5300bSChris Mason u64 newer_than, unsigned long max_to_defrag) 9764cb5300bSChris Mason { 9774cb5300bSChris Mason struct btrfs_root *root = BTRFS_I(inode)->root; 9784cb5300bSChris Mason struct btrfs_super_block *disk_super; 9794cb5300bSChris Mason struct file_ra_state *ra = NULL; 9804cb5300bSChris Mason unsigned long last_index; 981151a31b2SLi Zefan u64 isize = i_size_read(inode); 9824cb5300bSChris Mason u64 features; 983940100a4SChris Mason u64 last_len = 0; 984940100a4SChris Mason u64 skip = 0; 985940100a4SChris Mason u64 defrag_end = 0; 9864cb5300bSChris Mason u64 newer_off = range->start; 9874cb5300bSChris Mason int newer_left = 0; 988f46b5a66SChristoph Hellwig unsigned long i; 989f46b5a66SChristoph Hellwig int ret; 9904cb5300bSChris Mason int defrag_count = 0; 9911a419d85SLi Zefan int compress_type = BTRFS_COMPRESS_ZLIB; 9924cb5300bSChris Mason int extent_thresh = range->extent_thresh; 9934cb5300bSChris Mason int newer_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT; 9944cb5300bSChris Mason u64 new_align = ~((u64)128 * 1024 - 1); 9954cb5300bSChris Mason struct page **pages = NULL; 9964cb5300bSChris Mason 9974cb5300bSChris Mason if (extent_thresh == 0) 9984cb5300bSChris Mason extent_thresh = 256 * 1024; 9991a419d85SLi Zefan 10001a419d85SLi Zefan if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) { 10011a419d85SLi Zefan if (range->compress_type > BTRFS_COMPRESS_TYPES) 10021a419d85SLi Zefan return -EINVAL; 10031a419d85SLi Zefan if (range->compress_type) 10041a419d85SLi Zefan compress_type = range->compress_type; 10051a419d85SLi Zefan } 1006f46b5a66SChristoph Hellwig 1007151a31b2SLi Zefan if (isize == 0) 1008940100a4SChris Mason return 0; 1009f46b5a66SChristoph Hellwig 10104cb5300bSChris Mason /* 10114cb5300bSChris Mason * if we were not given a file, allocate a readahead 10124cb5300bSChris Mason * context 10134cb5300bSChris Mason */ 10144cb5300bSChris Mason if (!file) { 10154cb5300bSChris Mason ra = kzalloc(sizeof(*ra), GFP_NOFS); 10164cb5300bSChris Mason if (!ra) 10174cb5300bSChris Mason return -ENOMEM; 10184cb5300bSChris Mason file_ra_state_init(ra, inode->i_mapping); 10194cb5300bSChris Mason } else { 10204cb5300bSChris Mason ra = &file->f_ra; 10214cb5300bSChris Mason } 10224cb5300bSChris Mason 10234cb5300bSChris Mason pages = kmalloc(sizeof(struct page *) * newer_cluster, 10244cb5300bSChris Mason GFP_NOFS); 10254cb5300bSChris Mason if (!pages) { 10264cb5300bSChris Mason ret = -ENOMEM; 10274cb5300bSChris Mason goto out_ra; 10284cb5300bSChris Mason } 10294cb5300bSChris Mason 10304cb5300bSChris Mason /* find the last page to defrag */ 10311e701a32SChris Mason if (range->start + range->len > range->start) { 1032151a31b2SLi Zefan last_index = min_t(u64, isize - 1, 10331e701a32SChris Mason range->start + range->len - 1) >> PAGE_CACHE_SHIFT; 10341e701a32SChris Mason } else { 1035151a31b2SLi Zefan last_index = (isize - 1) >> PAGE_CACHE_SHIFT; 10361e701a32SChris Mason } 10371e701a32SChris Mason 10384cb5300bSChris Mason if (newer_than) { 10394cb5300bSChris Mason ret = find_new_extents(root, inode, newer_than, 10404cb5300bSChris Mason &newer_off, 64 * 1024); 10414cb5300bSChris Mason if (!ret) { 10424cb5300bSChris Mason range->start = newer_off; 10434cb5300bSChris Mason /* 10444cb5300bSChris Mason * we always align our defrag to help keep 10454cb5300bSChris Mason * the extents in the file evenly spaced 10464cb5300bSChris Mason */ 10474cb5300bSChris Mason i = (newer_off & new_align) >> PAGE_CACHE_SHIFT; 10484cb5300bSChris Mason newer_left = newer_cluster; 10494cb5300bSChris Mason } else 10504cb5300bSChris Mason goto out_ra; 10514cb5300bSChris Mason } else { 10521e701a32SChris Mason i = range->start >> PAGE_CACHE_SHIFT; 10534cb5300bSChris Mason } 10544cb5300bSChris Mason if (!max_to_defrag) 10555ca49660SLi Zefan max_to_defrag = last_index; 10564cb5300bSChris Mason 10574cb5300bSChris Mason while (i <= last_index && defrag_count < max_to_defrag) { 10584cb5300bSChris Mason /* 10594cb5300bSChris Mason * make sure we stop running if someone unmounts 10604cb5300bSChris Mason * the FS 10614cb5300bSChris Mason */ 10624cb5300bSChris Mason if (!(inode->i_sb->s_flags & MS_ACTIVE)) 10634cb5300bSChris Mason break; 10644cb5300bSChris Mason 10654cb5300bSChris Mason if (!newer_than && 10664cb5300bSChris Mason !should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT, 10671e701a32SChris Mason PAGE_CACHE_SIZE, 10684cb5300bSChris Mason extent_thresh, 10691e701a32SChris Mason &last_len, &skip, 1070940100a4SChris Mason &defrag_end)) { 1071940100a4SChris Mason unsigned long next; 1072940100a4SChris Mason /* 1073940100a4SChris Mason * the should_defrag function tells us how much to skip 1074940100a4SChris Mason * bump our counter by the suggested amount 1075940100a4SChris Mason */ 1076940100a4SChris Mason next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT; 1077940100a4SChris Mason i = max(i + 1, next); 1078940100a4SChris Mason continue; 1079940100a4SChris Mason } 10801e701a32SChris Mason if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) 10811a419d85SLi Zefan BTRFS_I(inode)->force_compress = compress_type; 1082940100a4SChris Mason 10834cb5300bSChris Mason btrfs_force_ra(inode->i_mapping, ra, file, i, newer_cluster); 10844cb5300bSChris Mason 10854cb5300bSChris Mason ret = cluster_pages_for_defrag(inode, pages, i, newer_cluster); 10864cb5300bSChris Mason if (ret < 0) 10874cb5300bSChris Mason goto out_ra; 10884cb5300bSChris Mason 10894cb5300bSChris Mason defrag_count += ret; 10904cb5300bSChris Mason balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret); 10914cb5300bSChris Mason 10924cb5300bSChris Mason if (newer_than) { 10934cb5300bSChris Mason if (newer_off == (u64)-1) 10944cb5300bSChris Mason break; 10954cb5300bSChris Mason 10964cb5300bSChris Mason newer_off = max(newer_off + 1, 10974cb5300bSChris Mason (u64)i << PAGE_CACHE_SHIFT); 10984cb5300bSChris Mason 10994cb5300bSChris Mason ret = find_new_extents(root, inode, 11004cb5300bSChris Mason newer_than, &newer_off, 11014cb5300bSChris Mason 64 * 1024); 11024cb5300bSChris Mason if (!ret) { 11034cb5300bSChris Mason range->start = newer_off; 11044cb5300bSChris Mason i = (newer_off & new_align) >> PAGE_CACHE_SHIFT; 11054cb5300bSChris Mason newer_left = newer_cluster; 11064cb5300bSChris Mason } else { 11074cb5300bSChris Mason break; 1108940100a4SChris Mason } 11094cb5300bSChris Mason } else { 1110cbcc8326SLi Zefan if (ret > 0) 1111cbcc8326SLi Zefan i += ret; 1112cbcc8326SLi Zefan else 1113940100a4SChris Mason i++; 1114f46b5a66SChristoph Hellwig } 11154cb5300bSChris Mason } 1116f46b5a66SChristoph Hellwig 11171e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) 11181e701a32SChris Mason filemap_flush(inode->i_mapping); 11191e701a32SChris Mason 11201e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 11211e701a32SChris Mason /* the filemap_flush will queue IO into the worker threads, but 11221e701a32SChris Mason * we have to make sure the IO is actually started and that 11231e701a32SChris Mason * ordered extents get created before we return 11241e701a32SChris Mason */ 11251e701a32SChris Mason atomic_inc(&root->fs_info->async_submit_draining); 11261e701a32SChris Mason while (atomic_read(&root->fs_info->nr_async_submits) || 11271e701a32SChris Mason atomic_read(&root->fs_info->async_delalloc_pages)) { 11281e701a32SChris Mason wait_event(root->fs_info->async_submit_wait, 11291e701a32SChris Mason (atomic_read(&root->fs_info->nr_async_submits) == 0 && 11301e701a32SChris Mason atomic_read(&root->fs_info->async_delalloc_pages) == 0)); 11311e701a32SChris Mason } 11321e701a32SChris Mason atomic_dec(&root->fs_info->async_submit_draining); 11331e701a32SChris Mason 11341e701a32SChris Mason mutex_lock(&inode->i_mutex); 1135261507a0SLi Zefan BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE; 11361e701a32SChris Mason mutex_unlock(&inode->i_mutex); 11371e701a32SChris Mason } 11381e701a32SChris Mason 11391a419d85SLi Zefan disk_super = &root->fs_info->super_copy; 11401a419d85SLi Zefan features = btrfs_super_incompat_flags(disk_super); 11411a419d85SLi Zefan if (range->compress_type == BTRFS_COMPRESS_LZO) { 11421a419d85SLi Zefan features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; 11431a419d85SLi Zefan btrfs_set_super_incompat_flags(disk_super, features); 11441a419d85SLi Zefan } 11451a419d85SLi Zefan 114660ccf82fSDiego Calleja ret = defrag_count; 1147940100a4SChris Mason 11484cb5300bSChris Mason out_ra: 11494cb5300bSChris Mason if (!file) 11504cb5300bSChris Mason kfree(ra); 11514cb5300bSChris Mason kfree(pages); 1152940100a4SChris Mason return ret; 1153f46b5a66SChristoph Hellwig } 1154f46b5a66SChristoph Hellwig 115576dda93cSYan, Zheng static noinline int btrfs_ioctl_resize(struct btrfs_root *root, 115676dda93cSYan, Zheng void __user *arg) 1157f46b5a66SChristoph Hellwig { 1158f46b5a66SChristoph Hellwig u64 new_size; 1159f46b5a66SChristoph Hellwig u64 old_size; 1160f46b5a66SChristoph Hellwig u64 devid = 1; 1161f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 1162f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 1163f46b5a66SChristoph Hellwig struct btrfs_device *device = NULL; 1164f46b5a66SChristoph Hellwig char *sizestr; 1165f46b5a66SChristoph Hellwig char *devstr = NULL; 1166f46b5a66SChristoph Hellwig int ret = 0; 1167f46b5a66SChristoph Hellwig int mod = 0; 1168f46b5a66SChristoph Hellwig 1169c146afadSYan Zheng if (root->fs_info->sb->s_flags & MS_RDONLY) 1170c146afadSYan Zheng return -EROFS; 1171c146afadSYan Zheng 1172e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 1173e441d54dSChris Mason return -EPERM; 1174e441d54dSChris Mason 1175dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1176dae7b665SLi Zefan if (IS_ERR(vol_args)) 1177dae7b665SLi Zefan return PTR_ERR(vol_args); 11785516e595SMark Fasheh 11795516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1180f46b5a66SChristoph Hellwig 11817d9eb12cSChris Mason mutex_lock(&root->fs_info->volume_mutex); 1182f46b5a66SChristoph Hellwig sizestr = vol_args->name; 1183f46b5a66SChristoph Hellwig devstr = strchr(sizestr, ':'); 1184f46b5a66SChristoph Hellwig if (devstr) { 1185f46b5a66SChristoph Hellwig char *end; 1186f46b5a66SChristoph Hellwig sizestr = devstr + 1; 1187f46b5a66SChristoph Hellwig *devstr = '\0'; 1188f46b5a66SChristoph Hellwig devstr = vol_args->name; 1189f46b5a66SChristoph Hellwig devid = simple_strtoull(devstr, &end, 10); 119021380931SJoel Becker printk(KERN_INFO "resizing devid %llu\n", 119121380931SJoel Becker (unsigned long long)devid); 1192f46b5a66SChristoph Hellwig } 11932b82032cSYan Zheng device = btrfs_find_device(root, devid, NULL, NULL); 1194f46b5a66SChristoph Hellwig if (!device) { 119521380931SJoel Becker printk(KERN_INFO "resizer unable to find device %llu\n", 119621380931SJoel Becker (unsigned long long)devid); 1197f46b5a66SChristoph Hellwig ret = -EINVAL; 1198f46b5a66SChristoph Hellwig goto out_unlock; 1199f46b5a66SChristoph Hellwig } 1200f46b5a66SChristoph Hellwig if (!strcmp(sizestr, "max")) 1201f46b5a66SChristoph Hellwig new_size = device->bdev->bd_inode->i_size; 1202f46b5a66SChristoph Hellwig else { 1203f46b5a66SChristoph Hellwig if (sizestr[0] == '-') { 1204f46b5a66SChristoph Hellwig mod = -1; 1205f46b5a66SChristoph Hellwig sizestr++; 1206f46b5a66SChristoph Hellwig } else if (sizestr[0] == '+') { 1207f46b5a66SChristoph Hellwig mod = 1; 1208f46b5a66SChristoph Hellwig sizestr++; 1209f46b5a66SChristoph Hellwig } 121091748467SAkinobu Mita new_size = memparse(sizestr, NULL); 1211f46b5a66SChristoph Hellwig if (new_size == 0) { 1212f46b5a66SChristoph Hellwig ret = -EINVAL; 1213f46b5a66SChristoph Hellwig goto out_unlock; 1214f46b5a66SChristoph Hellwig } 1215f46b5a66SChristoph Hellwig } 1216f46b5a66SChristoph Hellwig 1217f46b5a66SChristoph Hellwig old_size = device->total_bytes; 1218f46b5a66SChristoph Hellwig 1219f46b5a66SChristoph Hellwig if (mod < 0) { 1220f46b5a66SChristoph Hellwig if (new_size > old_size) { 1221f46b5a66SChristoph Hellwig ret = -EINVAL; 1222f46b5a66SChristoph Hellwig goto out_unlock; 1223f46b5a66SChristoph Hellwig } 1224f46b5a66SChristoph Hellwig new_size = old_size - new_size; 1225f46b5a66SChristoph Hellwig } else if (mod > 0) { 1226f46b5a66SChristoph Hellwig new_size = old_size + new_size; 1227f46b5a66SChristoph Hellwig } 1228f46b5a66SChristoph Hellwig 1229f46b5a66SChristoph Hellwig if (new_size < 256 * 1024 * 1024) { 1230f46b5a66SChristoph Hellwig ret = -EINVAL; 1231f46b5a66SChristoph Hellwig goto out_unlock; 1232f46b5a66SChristoph Hellwig } 1233f46b5a66SChristoph Hellwig if (new_size > device->bdev->bd_inode->i_size) { 1234f46b5a66SChristoph Hellwig ret = -EFBIG; 1235f46b5a66SChristoph Hellwig goto out_unlock; 1236f46b5a66SChristoph Hellwig } 1237f46b5a66SChristoph Hellwig 1238f46b5a66SChristoph Hellwig do_div(new_size, root->sectorsize); 1239f46b5a66SChristoph Hellwig new_size *= root->sectorsize; 1240f46b5a66SChristoph Hellwig 1241f46b5a66SChristoph Hellwig printk(KERN_INFO "new size for %s is %llu\n", 1242f46b5a66SChristoph Hellwig device->name, (unsigned long long)new_size); 1243f46b5a66SChristoph Hellwig 1244f46b5a66SChristoph Hellwig if (new_size > old_size) { 1245a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 124698d5dc13STsutomu Itoh if (IS_ERR(trans)) { 124798d5dc13STsutomu Itoh ret = PTR_ERR(trans); 124898d5dc13STsutomu Itoh goto out_unlock; 124998d5dc13STsutomu Itoh } 1250f46b5a66SChristoph Hellwig ret = btrfs_grow_device(trans, device, new_size); 1251f46b5a66SChristoph Hellwig btrfs_commit_transaction(trans, root); 1252f46b5a66SChristoph Hellwig } else { 1253f46b5a66SChristoph Hellwig ret = btrfs_shrink_device(device, new_size); 1254f46b5a66SChristoph Hellwig } 1255f46b5a66SChristoph Hellwig 1256f46b5a66SChristoph Hellwig out_unlock: 12577d9eb12cSChris Mason mutex_unlock(&root->fs_info->volume_mutex); 1258f46b5a66SChristoph Hellwig kfree(vol_args); 1259f46b5a66SChristoph Hellwig return ret; 1260f46b5a66SChristoph Hellwig } 1261f46b5a66SChristoph Hellwig 126272fd032eSSage Weil static noinline int btrfs_ioctl_snap_create_transid(struct file *file, 126372fd032eSSage Weil char *name, 126472fd032eSSage Weil unsigned long fd, 126572fd032eSSage Weil int subvol, 1266b83cc969SLi Zefan u64 *transid, 1267b83cc969SLi Zefan bool readonly) 1268f46b5a66SChristoph Hellwig { 1269cb8e7090SChristoph Hellwig struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; 12703de4586cSChris Mason struct file *src_file; 1271f46b5a66SChristoph Hellwig int namelen; 12723de4586cSChris Mason int ret = 0; 1273f46b5a66SChristoph Hellwig 1274c146afadSYan Zheng if (root->fs_info->sb->s_flags & MS_RDONLY) 1275c146afadSYan Zheng return -EROFS; 1276c146afadSYan Zheng 127772fd032eSSage Weil namelen = strlen(name); 127872fd032eSSage Weil if (strchr(name, '/')) { 1279f46b5a66SChristoph Hellwig ret = -EINVAL; 1280f46b5a66SChristoph Hellwig goto out; 1281f46b5a66SChristoph Hellwig } 1282f46b5a66SChristoph Hellwig 12833de4586cSChris Mason if (subvol) { 128472fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 1285b83cc969SLi Zefan NULL, transid, readonly); 1286cb8e7090SChristoph Hellwig } else { 12873de4586cSChris Mason struct inode *src_inode; 128872fd032eSSage Weil src_file = fget(fd); 12893de4586cSChris Mason if (!src_file) { 12903de4586cSChris Mason ret = -EINVAL; 12913de4586cSChris Mason goto out; 12923de4586cSChris Mason } 12933de4586cSChris Mason 12943de4586cSChris Mason src_inode = src_file->f_path.dentry->d_inode; 12953de4586cSChris Mason if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) { 1296d397712bSChris Mason printk(KERN_INFO "btrfs: Snapshot src from " 1297d397712bSChris Mason "another FS\n"); 12983de4586cSChris Mason ret = -EINVAL; 12993de4586cSChris Mason fput(src_file); 13003de4586cSChris Mason goto out; 13013de4586cSChris Mason } 130272fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 130372fd032eSSage Weil BTRFS_I(src_inode)->root, 1304b83cc969SLi Zefan transid, readonly); 13053de4586cSChris Mason fput(src_file); 1306cb8e7090SChristoph Hellwig } 1307f46b5a66SChristoph Hellwig out: 130872fd032eSSage Weil return ret; 130972fd032eSSage Weil } 131072fd032eSSage Weil 131172fd032eSSage Weil static noinline int btrfs_ioctl_snap_create(struct file *file, 1312fa0d2b9bSLi Zefan void __user *arg, int subvol) 131372fd032eSSage Weil { 1314fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args *vol_args; 131572fd032eSSage Weil int ret; 131672fd032eSSage Weil 1317fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1318fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1319fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1320fa0d2b9bSLi Zefan vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1321fa0d2b9bSLi Zefan 1322fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 1323b83cc969SLi Zefan vol_args->fd, subvol, 1324b83cc969SLi Zefan NULL, false); 1325fa0d2b9bSLi Zefan 1326fa0d2b9bSLi Zefan kfree(vol_args); 1327fa0d2b9bSLi Zefan return ret; 1328fa0d2b9bSLi Zefan } 1329fa0d2b9bSLi Zefan 1330fa0d2b9bSLi Zefan static noinline int btrfs_ioctl_snap_create_v2(struct file *file, 1331fa0d2b9bSLi Zefan void __user *arg, int subvol) 1332fa0d2b9bSLi Zefan { 1333fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args_v2 *vol_args; 1334fa0d2b9bSLi Zefan int ret; 1335fdfb1e4fSLi Zefan u64 transid = 0; 1336fdfb1e4fSLi Zefan u64 *ptr = NULL; 1337b83cc969SLi Zefan bool readonly = false; 133872fd032eSSage Weil 1339fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1340fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1341fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1342fa0d2b9bSLi Zefan vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 1343fdfb1e4fSLi Zefan 1344b83cc969SLi Zefan if (vol_args->flags & 1345b83cc969SLi Zefan ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) { 1346b83cc969SLi Zefan ret = -EOPNOTSUPP; 1347fdfb1e4fSLi Zefan goto out; 1348fdfb1e4fSLi Zefan } 1349fdfb1e4fSLi Zefan 1350fa0d2b9bSLi Zefan if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC) 1351fdfb1e4fSLi Zefan ptr = &transid; 1352b83cc969SLi Zefan if (vol_args->flags & BTRFS_SUBVOL_RDONLY) 1353b83cc969SLi Zefan readonly = true; 135475eaa0e2SSage Weil 1355fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 1356b83cc969SLi Zefan vol_args->fd, subvol, 1357b83cc969SLi Zefan ptr, readonly); 135875eaa0e2SSage Weil 1359fdfb1e4fSLi Zefan if (ret == 0 && ptr && 136075eaa0e2SSage Weil copy_to_user(arg + 1361fdfb1e4fSLi Zefan offsetof(struct btrfs_ioctl_vol_args_v2, 1362fdfb1e4fSLi Zefan transid), ptr, sizeof(*ptr))) 136375eaa0e2SSage Weil ret = -EFAULT; 1364fdfb1e4fSLi Zefan out: 1365f46b5a66SChristoph Hellwig kfree(vol_args); 1366f46b5a66SChristoph Hellwig return ret; 1367f46b5a66SChristoph Hellwig } 1368f46b5a66SChristoph Hellwig 13690caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_getflags(struct file *file, 13700caa102dSLi Zefan void __user *arg) 13710caa102dSLi Zefan { 13720caa102dSLi Zefan struct inode *inode = fdentry(file)->d_inode; 13730caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 13740caa102dSLi Zefan int ret = 0; 13750caa102dSLi Zefan u64 flags = 0; 13760caa102dSLi Zefan 137733345d01SLi Zefan if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) 13780caa102dSLi Zefan return -EINVAL; 13790caa102dSLi Zefan 13800caa102dSLi Zefan down_read(&root->fs_info->subvol_sem); 13810caa102dSLi Zefan if (btrfs_root_readonly(root)) 13820caa102dSLi Zefan flags |= BTRFS_SUBVOL_RDONLY; 13830caa102dSLi Zefan up_read(&root->fs_info->subvol_sem); 13840caa102dSLi Zefan 13850caa102dSLi Zefan if (copy_to_user(arg, &flags, sizeof(flags))) 13860caa102dSLi Zefan ret = -EFAULT; 13870caa102dSLi Zefan 13880caa102dSLi Zefan return ret; 13890caa102dSLi Zefan } 13900caa102dSLi Zefan 13910caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_setflags(struct file *file, 13920caa102dSLi Zefan void __user *arg) 13930caa102dSLi Zefan { 13940caa102dSLi Zefan struct inode *inode = fdentry(file)->d_inode; 13950caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 13960caa102dSLi Zefan struct btrfs_trans_handle *trans; 13970caa102dSLi Zefan u64 root_flags; 13980caa102dSLi Zefan u64 flags; 13990caa102dSLi Zefan int ret = 0; 14000caa102dSLi Zefan 14010caa102dSLi Zefan if (root->fs_info->sb->s_flags & MS_RDONLY) 14020caa102dSLi Zefan return -EROFS; 14030caa102dSLi Zefan 140433345d01SLi Zefan if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) 14050caa102dSLi Zefan return -EINVAL; 14060caa102dSLi Zefan 14070caa102dSLi Zefan if (copy_from_user(&flags, arg, sizeof(flags))) 14080caa102dSLi Zefan return -EFAULT; 14090caa102dSLi Zefan 1410b4dc2b8cSLi Zefan if (flags & BTRFS_SUBVOL_CREATE_ASYNC) 14110caa102dSLi Zefan return -EINVAL; 14120caa102dSLi Zefan 14130caa102dSLi Zefan if (flags & ~BTRFS_SUBVOL_RDONLY) 14140caa102dSLi Zefan return -EOPNOTSUPP; 14150caa102dSLi Zefan 14162e149670SSerge E. Hallyn if (!inode_owner_or_capable(inode)) 1417b4dc2b8cSLi Zefan return -EACCES; 1418b4dc2b8cSLi Zefan 14190caa102dSLi Zefan down_write(&root->fs_info->subvol_sem); 14200caa102dSLi Zefan 14210caa102dSLi Zefan /* nothing to do */ 14220caa102dSLi Zefan if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) 14230caa102dSLi Zefan goto out; 14240caa102dSLi Zefan 14250caa102dSLi Zefan root_flags = btrfs_root_flags(&root->root_item); 14260caa102dSLi Zefan if (flags & BTRFS_SUBVOL_RDONLY) 14270caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 14280caa102dSLi Zefan root_flags | BTRFS_ROOT_SUBVOL_RDONLY); 14290caa102dSLi Zefan else 14300caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 14310caa102dSLi Zefan root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); 14320caa102dSLi Zefan 14330caa102dSLi Zefan trans = btrfs_start_transaction(root, 1); 14340caa102dSLi Zefan if (IS_ERR(trans)) { 14350caa102dSLi Zefan ret = PTR_ERR(trans); 14360caa102dSLi Zefan goto out_reset; 14370caa102dSLi Zefan } 14380caa102dSLi Zefan 1439b4dc2b8cSLi Zefan ret = btrfs_update_root(trans, root->fs_info->tree_root, 14400caa102dSLi Zefan &root->root_key, &root->root_item); 14410caa102dSLi Zefan 14420caa102dSLi Zefan btrfs_commit_transaction(trans, root); 14430caa102dSLi Zefan out_reset: 14440caa102dSLi Zefan if (ret) 14450caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, root_flags); 14460caa102dSLi Zefan out: 14470caa102dSLi Zefan up_write(&root->fs_info->subvol_sem); 14480caa102dSLi Zefan return ret; 14490caa102dSLi Zefan } 14500caa102dSLi Zefan 145176dda93cSYan, Zheng /* 145276dda93cSYan, Zheng * helper to check if the subvolume references other subvolumes 145376dda93cSYan, Zheng */ 145476dda93cSYan, Zheng static noinline int may_destroy_subvol(struct btrfs_root *root) 145576dda93cSYan, Zheng { 145676dda93cSYan, Zheng struct btrfs_path *path; 145776dda93cSYan, Zheng struct btrfs_key key; 145876dda93cSYan, Zheng int ret; 145976dda93cSYan, Zheng 146076dda93cSYan, Zheng path = btrfs_alloc_path(); 146176dda93cSYan, Zheng if (!path) 146276dda93cSYan, Zheng return -ENOMEM; 146376dda93cSYan, Zheng 146476dda93cSYan, Zheng key.objectid = root->root_key.objectid; 146576dda93cSYan, Zheng key.type = BTRFS_ROOT_REF_KEY; 146676dda93cSYan, Zheng key.offset = (u64)-1; 146776dda93cSYan, Zheng 146876dda93cSYan, Zheng ret = btrfs_search_slot(NULL, root->fs_info->tree_root, 146976dda93cSYan, Zheng &key, path, 0, 0); 147076dda93cSYan, Zheng if (ret < 0) 147176dda93cSYan, Zheng goto out; 147276dda93cSYan, Zheng BUG_ON(ret == 0); 147376dda93cSYan, Zheng 147476dda93cSYan, Zheng ret = 0; 147576dda93cSYan, Zheng if (path->slots[0] > 0) { 147676dda93cSYan, Zheng path->slots[0]--; 147776dda93cSYan, Zheng btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 147876dda93cSYan, Zheng if (key.objectid == root->root_key.objectid && 147976dda93cSYan, Zheng key.type == BTRFS_ROOT_REF_KEY) 148076dda93cSYan, Zheng ret = -ENOTEMPTY; 148176dda93cSYan, Zheng } 148276dda93cSYan, Zheng out: 148376dda93cSYan, Zheng btrfs_free_path(path); 148476dda93cSYan, Zheng return ret; 148576dda93cSYan, Zheng } 148676dda93cSYan, Zheng 1487ac8e9819SChris Mason static noinline int key_in_sk(struct btrfs_key *key, 1488ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk) 1489ac8e9819SChris Mason { 1490abc6e134SChris Mason struct btrfs_key test; 1491abc6e134SChris Mason int ret; 1492abc6e134SChris Mason 1493abc6e134SChris Mason test.objectid = sk->min_objectid; 1494abc6e134SChris Mason test.type = sk->min_type; 1495abc6e134SChris Mason test.offset = sk->min_offset; 1496abc6e134SChris Mason 1497abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1498abc6e134SChris Mason if (ret < 0) 1499ac8e9819SChris Mason return 0; 1500abc6e134SChris Mason 1501abc6e134SChris Mason test.objectid = sk->max_objectid; 1502abc6e134SChris Mason test.type = sk->max_type; 1503abc6e134SChris Mason test.offset = sk->max_offset; 1504abc6e134SChris Mason 1505abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1506abc6e134SChris Mason if (ret > 0) 1507ac8e9819SChris Mason return 0; 1508ac8e9819SChris Mason return 1; 1509ac8e9819SChris Mason } 1510ac8e9819SChris Mason 1511ac8e9819SChris Mason static noinline int copy_to_sk(struct btrfs_root *root, 1512ac8e9819SChris Mason struct btrfs_path *path, 1513ac8e9819SChris Mason struct btrfs_key *key, 1514ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk, 1515ac8e9819SChris Mason char *buf, 1516ac8e9819SChris Mason unsigned long *sk_offset, 1517ac8e9819SChris Mason int *num_found) 1518ac8e9819SChris Mason { 1519ac8e9819SChris Mason u64 found_transid; 1520ac8e9819SChris Mason struct extent_buffer *leaf; 1521ac8e9819SChris Mason struct btrfs_ioctl_search_header sh; 1522ac8e9819SChris Mason unsigned long item_off; 1523ac8e9819SChris Mason unsigned long item_len; 1524ac8e9819SChris Mason int nritems; 1525ac8e9819SChris Mason int i; 1526ac8e9819SChris Mason int slot; 1527ac8e9819SChris Mason int ret = 0; 1528ac8e9819SChris Mason 1529ac8e9819SChris Mason leaf = path->nodes[0]; 1530ac8e9819SChris Mason slot = path->slots[0]; 1531ac8e9819SChris Mason nritems = btrfs_header_nritems(leaf); 1532ac8e9819SChris Mason 1533ac8e9819SChris Mason if (btrfs_header_generation(leaf) > sk->max_transid) { 1534ac8e9819SChris Mason i = nritems; 1535ac8e9819SChris Mason goto advance_key; 1536ac8e9819SChris Mason } 1537ac8e9819SChris Mason found_transid = btrfs_header_generation(leaf); 1538ac8e9819SChris Mason 1539ac8e9819SChris Mason for (i = slot; i < nritems; i++) { 1540ac8e9819SChris Mason item_off = btrfs_item_ptr_offset(leaf, i); 1541ac8e9819SChris Mason item_len = btrfs_item_size_nr(leaf, i); 1542ac8e9819SChris Mason 1543ac8e9819SChris Mason if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE) 1544ac8e9819SChris Mason item_len = 0; 1545ac8e9819SChris Mason 1546ac8e9819SChris Mason if (sizeof(sh) + item_len + *sk_offset > 1547ac8e9819SChris Mason BTRFS_SEARCH_ARGS_BUFSIZE) { 1548ac8e9819SChris Mason ret = 1; 1549ac8e9819SChris Mason goto overflow; 1550ac8e9819SChris Mason } 1551ac8e9819SChris Mason 1552ac8e9819SChris Mason btrfs_item_key_to_cpu(leaf, key, i); 1553ac8e9819SChris Mason if (!key_in_sk(key, sk)) 1554ac8e9819SChris Mason continue; 1555ac8e9819SChris Mason 1556ac8e9819SChris Mason sh.objectid = key->objectid; 1557ac8e9819SChris Mason sh.offset = key->offset; 1558ac8e9819SChris Mason sh.type = key->type; 1559ac8e9819SChris Mason sh.len = item_len; 1560ac8e9819SChris Mason sh.transid = found_transid; 1561ac8e9819SChris Mason 1562ac8e9819SChris Mason /* copy search result header */ 1563ac8e9819SChris Mason memcpy(buf + *sk_offset, &sh, sizeof(sh)); 1564ac8e9819SChris Mason *sk_offset += sizeof(sh); 1565ac8e9819SChris Mason 1566ac8e9819SChris Mason if (item_len) { 1567ac8e9819SChris Mason char *p = buf + *sk_offset; 1568ac8e9819SChris Mason /* copy the item */ 1569ac8e9819SChris Mason read_extent_buffer(leaf, p, 1570ac8e9819SChris Mason item_off, item_len); 1571ac8e9819SChris Mason *sk_offset += item_len; 1572ac8e9819SChris Mason } 1573e2156867SHugo Mills (*num_found)++; 1574ac8e9819SChris Mason 1575ac8e9819SChris Mason if (*num_found >= sk->nr_items) 1576ac8e9819SChris Mason break; 1577ac8e9819SChris Mason } 1578ac8e9819SChris Mason advance_key: 1579ac8e9819SChris Mason ret = 0; 1580abc6e134SChris Mason if (key->offset < (u64)-1 && key->offset < sk->max_offset) 1581abc6e134SChris Mason key->offset++; 1582abc6e134SChris Mason else if (key->type < (u8)-1 && key->type < sk->max_type) { 1583abc6e134SChris Mason key->offset = 0; 1584abc6e134SChris Mason key->type++; 1585abc6e134SChris Mason } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) { 1586abc6e134SChris Mason key->offset = 0; 1587abc6e134SChris Mason key->type = 0; 1588abc6e134SChris Mason key->objectid++; 1589abc6e134SChris Mason } else 1590abc6e134SChris Mason ret = 1; 1591ac8e9819SChris Mason overflow: 1592ac8e9819SChris Mason return ret; 1593ac8e9819SChris Mason } 1594ac8e9819SChris Mason 1595ac8e9819SChris Mason static noinline int search_ioctl(struct inode *inode, 1596ac8e9819SChris Mason struct btrfs_ioctl_search_args *args) 1597ac8e9819SChris Mason { 1598ac8e9819SChris Mason struct btrfs_root *root; 1599ac8e9819SChris Mason struct btrfs_key key; 1600ac8e9819SChris Mason struct btrfs_key max_key; 1601ac8e9819SChris Mason struct btrfs_path *path; 1602ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk = &args->key; 1603ac8e9819SChris Mason struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info; 1604ac8e9819SChris Mason int ret; 1605ac8e9819SChris Mason int num_found = 0; 1606ac8e9819SChris Mason unsigned long sk_offset = 0; 1607ac8e9819SChris Mason 1608ac8e9819SChris Mason path = btrfs_alloc_path(); 1609ac8e9819SChris Mason if (!path) 1610ac8e9819SChris Mason return -ENOMEM; 1611ac8e9819SChris Mason 1612ac8e9819SChris Mason if (sk->tree_id == 0) { 1613ac8e9819SChris Mason /* search the root of the inode that was passed */ 1614ac8e9819SChris Mason root = BTRFS_I(inode)->root; 1615ac8e9819SChris Mason } else { 1616ac8e9819SChris Mason key.objectid = sk->tree_id; 1617ac8e9819SChris Mason key.type = BTRFS_ROOT_ITEM_KEY; 1618ac8e9819SChris Mason key.offset = (u64)-1; 1619ac8e9819SChris Mason root = btrfs_read_fs_root_no_name(info, &key); 1620ac8e9819SChris Mason if (IS_ERR(root)) { 1621ac8e9819SChris Mason printk(KERN_ERR "could not find root %llu\n", 1622ac8e9819SChris Mason sk->tree_id); 1623ac8e9819SChris Mason btrfs_free_path(path); 1624ac8e9819SChris Mason return -ENOENT; 1625ac8e9819SChris Mason } 1626ac8e9819SChris Mason } 1627ac8e9819SChris Mason 1628ac8e9819SChris Mason key.objectid = sk->min_objectid; 1629ac8e9819SChris Mason key.type = sk->min_type; 1630ac8e9819SChris Mason key.offset = sk->min_offset; 1631ac8e9819SChris Mason 1632ac8e9819SChris Mason max_key.objectid = sk->max_objectid; 1633ac8e9819SChris Mason max_key.type = sk->max_type; 1634ac8e9819SChris Mason max_key.offset = sk->max_offset; 1635ac8e9819SChris Mason 1636ac8e9819SChris Mason path->keep_locks = 1; 1637ac8e9819SChris Mason 1638ac8e9819SChris Mason while(1) { 1639ac8e9819SChris Mason ret = btrfs_search_forward(root, &key, &max_key, path, 0, 1640ac8e9819SChris Mason sk->min_transid); 1641ac8e9819SChris Mason if (ret != 0) { 1642ac8e9819SChris Mason if (ret > 0) 1643ac8e9819SChris Mason ret = 0; 1644ac8e9819SChris Mason goto err; 1645ac8e9819SChris Mason } 1646ac8e9819SChris Mason ret = copy_to_sk(root, path, &key, sk, args->buf, 1647ac8e9819SChris Mason &sk_offset, &num_found); 1648b3b4aa74SDavid Sterba btrfs_release_path(path); 1649ac8e9819SChris Mason if (ret || num_found >= sk->nr_items) 1650ac8e9819SChris Mason break; 1651ac8e9819SChris Mason 1652ac8e9819SChris Mason } 1653ac8e9819SChris Mason ret = 0; 1654ac8e9819SChris Mason err: 1655ac8e9819SChris Mason sk->nr_items = num_found; 1656ac8e9819SChris Mason btrfs_free_path(path); 1657ac8e9819SChris Mason return ret; 1658ac8e9819SChris Mason } 1659ac8e9819SChris Mason 1660ac8e9819SChris Mason static noinline int btrfs_ioctl_tree_search(struct file *file, 1661ac8e9819SChris Mason void __user *argp) 1662ac8e9819SChris Mason { 1663ac8e9819SChris Mason struct btrfs_ioctl_search_args *args; 1664ac8e9819SChris Mason struct inode *inode; 1665ac8e9819SChris Mason int ret; 1666ac8e9819SChris Mason 1667ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 1668ac8e9819SChris Mason return -EPERM; 1669ac8e9819SChris Mason 16702354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 16712354d08fSJulia Lawall if (IS_ERR(args)) 16722354d08fSJulia Lawall return PTR_ERR(args); 1673ac8e9819SChris Mason 1674ac8e9819SChris Mason inode = fdentry(file)->d_inode; 1675ac8e9819SChris Mason ret = search_ioctl(inode, args); 1676ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 1677ac8e9819SChris Mason ret = -EFAULT; 1678ac8e9819SChris Mason kfree(args); 1679ac8e9819SChris Mason return ret; 1680ac8e9819SChris Mason } 1681ac8e9819SChris Mason 168298d377a0STARUISI Hiroaki /* 1683ac8e9819SChris Mason * Search INODE_REFs to identify path name of 'dirid' directory 1684ac8e9819SChris Mason * in a 'tree_id' tree. and sets path name to 'name'. 168598d377a0STARUISI Hiroaki */ 168698d377a0STARUISI Hiroaki static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, 168798d377a0STARUISI Hiroaki u64 tree_id, u64 dirid, char *name) 168898d377a0STARUISI Hiroaki { 168998d377a0STARUISI Hiroaki struct btrfs_root *root; 169098d377a0STARUISI Hiroaki struct btrfs_key key; 1691ac8e9819SChris Mason char *ptr; 169298d377a0STARUISI Hiroaki int ret = -1; 169398d377a0STARUISI Hiroaki int slot; 169498d377a0STARUISI Hiroaki int len; 169598d377a0STARUISI Hiroaki int total_len = 0; 169698d377a0STARUISI Hiroaki struct btrfs_inode_ref *iref; 169798d377a0STARUISI Hiroaki struct extent_buffer *l; 169898d377a0STARUISI Hiroaki struct btrfs_path *path; 169998d377a0STARUISI Hiroaki 170098d377a0STARUISI Hiroaki if (dirid == BTRFS_FIRST_FREE_OBJECTID) { 170198d377a0STARUISI Hiroaki name[0]='\0'; 170298d377a0STARUISI Hiroaki return 0; 170398d377a0STARUISI Hiroaki } 170498d377a0STARUISI Hiroaki 170598d377a0STARUISI Hiroaki path = btrfs_alloc_path(); 170698d377a0STARUISI Hiroaki if (!path) 170798d377a0STARUISI Hiroaki return -ENOMEM; 170898d377a0STARUISI Hiroaki 1709ac8e9819SChris Mason ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX]; 171098d377a0STARUISI Hiroaki 171198d377a0STARUISI Hiroaki key.objectid = tree_id; 171298d377a0STARUISI Hiroaki key.type = BTRFS_ROOT_ITEM_KEY; 171398d377a0STARUISI Hiroaki key.offset = (u64)-1; 171498d377a0STARUISI Hiroaki root = btrfs_read_fs_root_no_name(info, &key); 171598d377a0STARUISI Hiroaki if (IS_ERR(root)) { 171698d377a0STARUISI Hiroaki printk(KERN_ERR "could not find root %llu\n", tree_id); 17178ad6fcabSChris Mason ret = -ENOENT; 17188ad6fcabSChris Mason goto out; 171998d377a0STARUISI Hiroaki } 172098d377a0STARUISI Hiroaki 172198d377a0STARUISI Hiroaki key.objectid = dirid; 172298d377a0STARUISI Hiroaki key.type = BTRFS_INODE_REF_KEY; 17238ad6fcabSChris Mason key.offset = (u64)-1; 172498d377a0STARUISI Hiroaki 172598d377a0STARUISI Hiroaki while(1) { 172698d377a0STARUISI Hiroaki ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 172798d377a0STARUISI Hiroaki if (ret < 0) 172898d377a0STARUISI Hiroaki goto out; 172998d377a0STARUISI Hiroaki 173098d377a0STARUISI Hiroaki l = path->nodes[0]; 173198d377a0STARUISI Hiroaki slot = path->slots[0]; 17328ad6fcabSChris Mason if (ret > 0 && slot > 0) 17338ad6fcabSChris Mason slot--; 173498d377a0STARUISI Hiroaki btrfs_item_key_to_cpu(l, &key, slot); 173598d377a0STARUISI Hiroaki 173698d377a0STARUISI Hiroaki if (ret > 0 && (key.objectid != dirid || 1737ac8e9819SChris Mason key.type != BTRFS_INODE_REF_KEY)) { 1738ac8e9819SChris Mason ret = -ENOENT; 173998d377a0STARUISI Hiroaki goto out; 1740ac8e9819SChris Mason } 174198d377a0STARUISI Hiroaki 174298d377a0STARUISI Hiroaki iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); 174398d377a0STARUISI Hiroaki len = btrfs_inode_ref_name_len(l, iref); 174498d377a0STARUISI Hiroaki ptr -= len + 1; 174598d377a0STARUISI Hiroaki total_len += len + 1; 1746ac8e9819SChris Mason if (ptr < name) 174798d377a0STARUISI Hiroaki goto out; 174898d377a0STARUISI Hiroaki 174998d377a0STARUISI Hiroaki *(ptr + len) = '/'; 175098d377a0STARUISI Hiroaki read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len); 175198d377a0STARUISI Hiroaki 175298d377a0STARUISI Hiroaki if (key.offset == BTRFS_FIRST_FREE_OBJECTID) 175398d377a0STARUISI Hiroaki break; 175498d377a0STARUISI Hiroaki 1755b3b4aa74SDavid Sterba btrfs_release_path(path); 175698d377a0STARUISI Hiroaki key.objectid = key.offset; 17578ad6fcabSChris Mason key.offset = (u64)-1; 175898d377a0STARUISI Hiroaki dirid = key.objectid; 175998d377a0STARUISI Hiroaki } 1760ac8e9819SChris Mason if (ptr < name) 176198d377a0STARUISI Hiroaki goto out; 176277906a50SLi Zefan memmove(name, ptr, total_len); 176398d377a0STARUISI Hiroaki name[total_len]='\0'; 176498d377a0STARUISI Hiroaki ret = 0; 176598d377a0STARUISI Hiroaki out: 176698d377a0STARUISI Hiroaki btrfs_free_path(path); 1767ac8e9819SChris Mason return ret; 1768ac8e9819SChris Mason } 1769ac8e9819SChris Mason 1770ac8e9819SChris Mason static noinline int btrfs_ioctl_ino_lookup(struct file *file, 1771ac8e9819SChris Mason void __user *argp) 1772ac8e9819SChris Mason { 1773ac8e9819SChris Mason struct btrfs_ioctl_ino_lookup_args *args; 1774ac8e9819SChris Mason struct inode *inode; 1775ac8e9819SChris Mason int ret; 1776ac8e9819SChris Mason 1777ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 1778ac8e9819SChris Mason return -EPERM; 1779ac8e9819SChris Mason 17802354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 17812354d08fSJulia Lawall if (IS_ERR(args)) 17822354d08fSJulia Lawall return PTR_ERR(args); 1783c2b96929SDan Carpenter 1784ac8e9819SChris Mason inode = fdentry(file)->d_inode; 1785ac8e9819SChris Mason 17861b53ac4dSChris Mason if (args->treeid == 0) 17871b53ac4dSChris Mason args->treeid = BTRFS_I(inode)->root->root_key.objectid; 17881b53ac4dSChris Mason 1789ac8e9819SChris Mason ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, 1790ac8e9819SChris Mason args->treeid, args->objectid, 1791ac8e9819SChris Mason args->name); 1792ac8e9819SChris Mason 1793ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 1794ac8e9819SChris Mason ret = -EFAULT; 1795ac8e9819SChris Mason 1796ac8e9819SChris Mason kfree(args); 179798d377a0STARUISI Hiroaki return ret; 179898d377a0STARUISI Hiroaki } 179998d377a0STARUISI Hiroaki 180076dda93cSYan, Zheng static noinline int btrfs_ioctl_snap_destroy(struct file *file, 180176dda93cSYan, Zheng void __user *arg) 180276dda93cSYan, Zheng { 180376dda93cSYan, Zheng struct dentry *parent = fdentry(file); 180476dda93cSYan, Zheng struct dentry *dentry; 180576dda93cSYan, Zheng struct inode *dir = parent->d_inode; 180676dda93cSYan, Zheng struct inode *inode; 180776dda93cSYan, Zheng struct btrfs_root *root = BTRFS_I(dir)->root; 180876dda93cSYan, Zheng struct btrfs_root *dest = NULL; 180976dda93cSYan, Zheng struct btrfs_ioctl_vol_args *vol_args; 181076dda93cSYan, Zheng struct btrfs_trans_handle *trans; 181176dda93cSYan, Zheng int namelen; 181276dda93cSYan, Zheng int ret; 181376dda93cSYan, Zheng int err = 0; 181476dda93cSYan, Zheng 181576dda93cSYan, Zheng vol_args = memdup_user(arg, sizeof(*vol_args)); 181676dda93cSYan, Zheng if (IS_ERR(vol_args)) 181776dda93cSYan, Zheng return PTR_ERR(vol_args); 181876dda93cSYan, Zheng 181976dda93cSYan, Zheng vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 182076dda93cSYan, Zheng namelen = strlen(vol_args->name); 182176dda93cSYan, Zheng if (strchr(vol_args->name, '/') || 182276dda93cSYan, Zheng strncmp(vol_args->name, "..", namelen) == 0) { 182376dda93cSYan, Zheng err = -EINVAL; 182476dda93cSYan, Zheng goto out; 182576dda93cSYan, Zheng } 182676dda93cSYan, Zheng 182776dda93cSYan, Zheng err = mnt_want_write(file->f_path.mnt); 182876dda93cSYan, Zheng if (err) 182976dda93cSYan, Zheng goto out; 183076dda93cSYan, Zheng 183176dda93cSYan, Zheng mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT); 183276dda93cSYan, Zheng dentry = lookup_one_len(vol_args->name, parent, namelen); 183376dda93cSYan, Zheng if (IS_ERR(dentry)) { 183476dda93cSYan, Zheng err = PTR_ERR(dentry); 183576dda93cSYan, Zheng goto out_unlock_dir; 183676dda93cSYan, Zheng } 183776dda93cSYan, Zheng 183876dda93cSYan, Zheng if (!dentry->d_inode) { 183976dda93cSYan, Zheng err = -ENOENT; 184076dda93cSYan, Zheng goto out_dput; 184176dda93cSYan, Zheng } 184276dda93cSYan, Zheng 184376dda93cSYan, Zheng inode = dentry->d_inode; 18444260f7c7SSage Weil dest = BTRFS_I(inode)->root; 18454260f7c7SSage Weil if (!capable(CAP_SYS_ADMIN)){ 18464260f7c7SSage Weil /* 18474260f7c7SSage Weil * Regular user. Only allow this with a special mount 18484260f7c7SSage Weil * option, when the user has write+exec access to the 18494260f7c7SSage Weil * subvol root, and when rmdir(2) would have been 18504260f7c7SSage Weil * allowed. 18514260f7c7SSage Weil * 18524260f7c7SSage Weil * Note that this is _not_ check that the subvol is 18534260f7c7SSage Weil * empty or doesn't contain data that we wouldn't 18544260f7c7SSage Weil * otherwise be able to delete. 18554260f7c7SSage Weil * 18564260f7c7SSage Weil * Users who want to delete empty subvols should try 18574260f7c7SSage Weil * rmdir(2). 18584260f7c7SSage Weil */ 18594260f7c7SSage Weil err = -EPERM; 18604260f7c7SSage Weil if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED)) 18614260f7c7SSage Weil goto out_dput; 18624260f7c7SSage Weil 18634260f7c7SSage Weil /* 18644260f7c7SSage Weil * Do not allow deletion if the parent dir is the same 18654260f7c7SSage Weil * as the dir to be deleted. That means the ioctl 18664260f7c7SSage Weil * must be called on the dentry referencing the root 18674260f7c7SSage Weil * of the subvol, not a random directory contained 18684260f7c7SSage Weil * within it. 18694260f7c7SSage Weil */ 18704260f7c7SSage Weil err = -EINVAL; 18714260f7c7SSage Weil if (root == dest) 18724260f7c7SSage Weil goto out_dput; 18734260f7c7SSage Weil 18744260f7c7SSage Weil err = inode_permission(inode, MAY_WRITE | MAY_EXEC); 18754260f7c7SSage Weil if (err) 18764260f7c7SSage Weil goto out_dput; 18774260f7c7SSage Weil 18784260f7c7SSage Weil /* check if subvolume may be deleted by a non-root user */ 18794260f7c7SSage Weil err = btrfs_may_delete(dir, dentry, 1); 18804260f7c7SSage Weil if (err) 18814260f7c7SSage Weil goto out_dput; 18824260f7c7SSage Weil } 18834260f7c7SSage Weil 188433345d01SLi Zefan if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { 188576dda93cSYan, Zheng err = -EINVAL; 188676dda93cSYan, Zheng goto out_dput; 188776dda93cSYan, Zheng } 188876dda93cSYan, Zheng 188976dda93cSYan, Zheng mutex_lock(&inode->i_mutex); 189076dda93cSYan, Zheng err = d_invalidate(dentry); 189176dda93cSYan, Zheng if (err) 189276dda93cSYan, Zheng goto out_unlock; 189376dda93cSYan, Zheng 189476dda93cSYan, Zheng down_write(&root->fs_info->subvol_sem); 189576dda93cSYan, Zheng 189676dda93cSYan, Zheng err = may_destroy_subvol(dest); 189776dda93cSYan, Zheng if (err) 189876dda93cSYan, Zheng goto out_up_write; 189976dda93cSYan, Zheng 1900a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 1901a22285a6SYan, Zheng if (IS_ERR(trans)) { 1902a22285a6SYan, Zheng err = PTR_ERR(trans); 1903d327099aSDan Carpenter goto out_up_write; 1904a22285a6SYan, Zheng } 1905a22285a6SYan, Zheng trans->block_rsv = &root->fs_info->global_block_rsv; 1906a22285a6SYan, Zheng 190776dda93cSYan, Zheng ret = btrfs_unlink_subvol(trans, root, dir, 190876dda93cSYan, Zheng dest->root_key.objectid, 190976dda93cSYan, Zheng dentry->d_name.name, 191076dda93cSYan, Zheng dentry->d_name.len); 191176dda93cSYan, Zheng BUG_ON(ret); 191276dda93cSYan, Zheng 191376dda93cSYan, Zheng btrfs_record_root_in_trans(trans, dest); 191476dda93cSYan, Zheng 191576dda93cSYan, Zheng memset(&dest->root_item.drop_progress, 0, 191676dda93cSYan, Zheng sizeof(dest->root_item.drop_progress)); 191776dda93cSYan, Zheng dest->root_item.drop_level = 0; 191876dda93cSYan, Zheng btrfs_set_root_refs(&dest->root_item, 0); 191976dda93cSYan, Zheng 1920d68fc57bSYan, Zheng if (!xchg(&dest->orphan_item_inserted, 1)) { 192176dda93cSYan, Zheng ret = btrfs_insert_orphan_item(trans, 192276dda93cSYan, Zheng root->fs_info->tree_root, 192376dda93cSYan, Zheng dest->root_key.objectid); 192476dda93cSYan, Zheng BUG_ON(ret); 1925d68fc57bSYan, Zheng } 192676dda93cSYan, Zheng 1927531cb13fSSage Weil ret = btrfs_end_transaction(trans, root); 192876dda93cSYan, Zheng BUG_ON(ret); 192976dda93cSYan, Zheng inode->i_flags |= S_DEAD; 193076dda93cSYan, Zheng out_up_write: 193176dda93cSYan, Zheng up_write(&root->fs_info->subvol_sem); 193276dda93cSYan, Zheng out_unlock: 193376dda93cSYan, Zheng mutex_unlock(&inode->i_mutex); 193476dda93cSYan, Zheng if (!err) { 1935efefb143SYan, Zheng shrink_dcache_sb(root->fs_info->sb); 193676dda93cSYan, Zheng btrfs_invalidate_inodes(dest); 193776dda93cSYan, Zheng d_delete(dentry); 193876dda93cSYan, Zheng } 193976dda93cSYan, Zheng out_dput: 194076dda93cSYan, Zheng dput(dentry); 194176dda93cSYan, Zheng out_unlock_dir: 194276dda93cSYan, Zheng mutex_unlock(&dir->i_mutex); 194376dda93cSYan, Zheng mnt_drop_write(file->f_path.mnt); 194476dda93cSYan, Zheng out: 194576dda93cSYan, Zheng kfree(vol_args); 194676dda93cSYan, Zheng return err; 194776dda93cSYan, Zheng } 194876dda93cSYan, Zheng 19491e701a32SChris Mason static int btrfs_ioctl_defrag(struct file *file, void __user *argp) 1950f46b5a66SChristoph Hellwig { 1951f46b5a66SChristoph Hellwig struct inode *inode = fdentry(file)->d_inode; 1952f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 19531e701a32SChris Mason struct btrfs_ioctl_defrag_range_args *range; 1954c146afadSYan Zheng int ret; 1955c146afadSYan Zheng 1956b83cc969SLi Zefan if (btrfs_root_readonly(root)) 1957b83cc969SLi Zefan return -EROFS; 1958b83cc969SLi Zefan 1959c146afadSYan Zheng ret = mnt_want_write(file->f_path.mnt); 1960c146afadSYan Zheng if (ret) 1961c146afadSYan Zheng return ret; 1962f46b5a66SChristoph Hellwig 1963f46b5a66SChristoph Hellwig switch (inode->i_mode & S_IFMT) { 1964f46b5a66SChristoph Hellwig case S_IFDIR: 1965e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) { 1966e441d54dSChris Mason ret = -EPERM; 1967e441d54dSChris Mason goto out; 1968e441d54dSChris Mason } 19698929ecfaSYan, Zheng ret = btrfs_defrag_root(root, 0); 19708929ecfaSYan, Zheng if (ret) 19718929ecfaSYan, Zheng goto out; 19728929ecfaSYan, Zheng ret = btrfs_defrag_root(root->fs_info->extent_root, 0); 1973f46b5a66SChristoph Hellwig break; 1974f46b5a66SChristoph Hellwig case S_IFREG: 1975e441d54dSChris Mason if (!(file->f_mode & FMODE_WRITE)) { 1976e441d54dSChris Mason ret = -EINVAL; 1977e441d54dSChris Mason goto out; 1978e441d54dSChris Mason } 19791e701a32SChris Mason 19801e701a32SChris Mason range = kzalloc(sizeof(*range), GFP_KERNEL); 19811e701a32SChris Mason if (!range) { 19821e701a32SChris Mason ret = -ENOMEM; 19831e701a32SChris Mason goto out; 19841e701a32SChris Mason } 19851e701a32SChris Mason 19861e701a32SChris Mason if (argp) { 19871e701a32SChris Mason if (copy_from_user(range, argp, 19881e701a32SChris Mason sizeof(*range))) { 19891e701a32SChris Mason ret = -EFAULT; 19901e701a32SChris Mason kfree(range); 1991683be16eSDan Carpenter goto out; 19921e701a32SChris Mason } 19931e701a32SChris Mason /* compression requires us to start the IO */ 19941e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 19951e701a32SChris Mason range->flags |= BTRFS_DEFRAG_RANGE_START_IO; 19961e701a32SChris Mason range->extent_thresh = (u32)-1; 19971e701a32SChris Mason } 19981e701a32SChris Mason } else { 19991e701a32SChris Mason /* the rest are all set to zero by kzalloc */ 20001e701a32SChris Mason range->len = (u64)-1; 20011e701a32SChris Mason } 20024cb5300bSChris Mason ret = btrfs_defrag_file(fdentry(file)->d_inode, file, 20034cb5300bSChris Mason range, 0, 0); 20044cb5300bSChris Mason if (ret > 0) 20054cb5300bSChris Mason ret = 0; 20061e701a32SChris Mason kfree(range); 2007f46b5a66SChristoph Hellwig break; 20088929ecfaSYan, Zheng default: 20098929ecfaSYan, Zheng ret = -EINVAL; 2010f46b5a66SChristoph Hellwig } 2011e441d54dSChris Mason out: 2012ab67b7c1SYan Zheng mnt_drop_write(file->f_path.mnt); 2013e441d54dSChris Mason return ret; 2014f46b5a66SChristoph Hellwig } 2015f46b5a66SChristoph Hellwig 2016b2950863SChristoph Hellwig static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg) 2017f46b5a66SChristoph Hellwig { 2018f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2019f46b5a66SChristoph Hellwig int ret; 2020f46b5a66SChristoph Hellwig 2021e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2022e441d54dSChris Mason return -EPERM; 2023e441d54dSChris Mason 2024dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2025dae7b665SLi Zefan if (IS_ERR(vol_args)) 2026dae7b665SLi Zefan return PTR_ERR(vol_args); 2027f46b5a66SChristoph Hellwig 20285516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 2029f46b5a66SChristoph Hellwig ret = btrfs_init_new_device(root, vol_args->name); 2030f46b5a66SChristoph Hellwig 2031f46b5a66SChristoph Hellwig kfree(vol_args); 2032f46b5a66SChristoph Hellwig return ret; 2033f46b5a66SChristoph Hellwig } 2034f46b5a66SChristoph Hellwig 2035b2950863SChristoph Hellwig static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg) 2036f46b5a66SChristoph Hellwig { 2037f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2038f46b5a66SChristoph Hellwig int ret; 2039f46b5a66SChristoph Hellwig 2040e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2041e441d54dSChris Mason return -EPERM; 2042e441d54dSChris Mason 2043c146afadSYan Zheng if (root->fs_info->sb->s_flags & MS_RDONLY) 2044c146afadSYan Zheng return -EROFS; 2045c146afadSYan Zheng 2046dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2047dae7b665SLi Zefan if (IS_ERR(vol_args)) 2048dae7b665SLi Zefan return PTR_ERR(vol_args); 2049f46b5a66SChristoph Hellwig 20505516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 2051f46b5a66SChristoph Hellwig ret = btrfs_rm_device(root, vol_args->name); 2052f46b5a66SChristoph Hellwig 2053f46b5a66SChristoph Hellwig kfree(vol_args); 2054f46b5a66SChristoph Hellwig return ret; 2055f46b5a66SChristoph Hellwig } 2056f46b5a66SChristoph Hellwig 2057475f6387SJan Schmidt static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg) 2058475f6387SJan Schmidt { 2059027ed2f0SLi Zefan struct btrfs_ioctl_fs_info_args *fi_args; 2060475f6387SJan Schmidt struct btrfs_device *device; 2061475f6387SJan Schmidt struct btrfs_device *next; 2062475f6387SJan Schmidt struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; 2063027ed2f0SLi Zefan int ret = 0; 2064475f6387SJan Schmidt 2065475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 2066475f6387SJan Schmidt return -EPERM; 2067475f6387SJan Schmidt 2068027ed2f0SLi Zefan fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL); 2069027ed2f0SLi Zefan if (!fi_args) 2070027ed2f0SLi Zefan return -ENOMEM; 2071027ed2f0SLi Zefan 2072027ed2f0SLi Zefan fi_args->num_devices = fs_devices->num_devices; 2073027ed2f0SLi Zefan memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid)); 2074475f6387SJan Schmidt 2075475f6387SJan Schmidt mutex_lock(&fs_devices->device_list_mutex); 2076475f6387SJan Schmidt list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) { 2077027ed2f0SLi Zefan if (device->devid > fi_args->max_id) 2078027ed2f0SLi Zefan fi_args->max_id = device->devid; 2079475f6387SJan Schmidt } 2080475f6387SJan Schmidt mutex_unlock(&fs_devices->device_list_mutex); 2081475f6387SJan Schmidt 2082027ed2f0SLi Zefan if (copy_to_user(arg, fi_args, sizeof(*fi_args))) 2083027ed2f0SLi Zefan ret = -EFAULT; 2084475f6387SJan Schmidt 2085027ed2f0SLi Zefan kfree(fi_args); 2086027ed2f0SLi Zefan return ret; 2087475f6387SJan Schmidt } 2088475f6387SJan Schmidt 2089475f6387SJan Schmidt static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg) 2090475f6387SJan Schmidt { 2091475f6387SJan Schmidt struct btrfs_ioctl_dev_info_args *di_args; 2092475f6387SJan Schmidt struct btrfs_device *dev; 2093475f6387SJan Schmidt struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; 2094475f6387SJan Schmidt int ret = 0; 2095475f6387SJan Schmidt char *s_uuid = NULL; 2096475f6387SJan Schmidt char empty_uuid[BTRFS_UUID_SIZE] = {0}; 2097475f6387SJan Schmidt 2098475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 2099475f6387SJan Schmidt return -EPERM; 2100475f6387SJan Schmidt 2101475f6387SJan Schmidt di_args = memdup_user(arg, sizeof(*di_args)); 2102475f6387SJan Schmidt if (IS_ERR(di_args)) 2103475f6387SJan Schmidt return PTR_ERR(di_args); 2104475f6387SJan Schmidt 2105475f6387SJan Schmidt if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0) 2106475f6387SJan Schmidt s_uuid = di_args->uuid; 2107475f6387SJan Schmidt 2108475f6387SJan Schmidt mutex_lock(&fs_devices->device_list_mutex); 2109475f6387SJan Schmidt dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL); 2110475f6387SJan Schmidt mutex_unlock(&fs_devices->device_list_mutex); 2111475f6387SJan Schmidt 2112475f6387SJan Schmidt if (!dev) { 2113475f6387SJan Schmidt ret = -ENODEV; 2114475f6387SJan Schmidt goto out; 2115475f6387SJan Schmidt } 2116475f6387SJan Schmidt 2117475f6387SJan Schmidt di_args->devid = dev->devid; 2118475f6387SJan Schmidt di_args->bytes_used = dev->bytes_used; 2119475f6387SJan Schmidt di_args->total_bytes = dev->total_bytes; 2120475f6387SJan Schmidt memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); 2121475f6387SJan Schmidt strncpy(di_args->path, dev->name, sizeof(di_args->path)); 2122475f6387SJan Schmidt 2123475f6387SJan Schmidt out: 2124475f6387SJan Schmidt if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) 2125475f6387SJan Schmidt ret = -EFAULT; 2126475f6387SJan Schmidt 2127475f6387SJan Schmidt kfree(di_args); 2128475f6387SJan Schmidt return ret; 2129475f6387SJan Schmidt } 2130475f6387SJan Schmidt 213176dda93cSYan, Zheng static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, 2132b2950863SChristoph Hellwig u64 off, u64 olen, u64 destoff) 2133f46b5a66SChristoph Hellwig { 2134f46b5a66SChristoph Hellwig struct inode *inode = fdentry(file)->d_inode; 2135f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 2136f46b5a66SChristoph Hellwig struct file *src_file; 2137f46b5a66SChristoph Hellwig struct inode *src; 2138f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 2139f46b5a66SChristoph Hellwig struct btrfs_path *path; 2140f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 2141ae01a0abSYan Zheng char *buf; 2142ae01a0abSYan Zheng struct btrfs_key key; 2143f46b5a66SChristoph Hellwig u32 nritems; 2144f46b5a66SChristoph Hellwig int slot; 2145ae01a0abSYan Zheng int ret; 2146c5c9cd4dSSage Weil u64 len = olen; 2147c5c9cd4dSSage Weil u64 bs = root->fs_info->sb->s_blocksize; 2148c5c9cd4dSSage Weil u64 hint_byte; 2149d20f7043SChris Mason 2150c5c9cd4dSSage Weil /* 2151c5c9cd4dSSage Weil * TODO: 2152c5c9cd4dSSage Weil * - split compressed inline extents. annoying: we need to 2153c5c9cd4dSSage Weil * decompress into destination's address_space (the file offset 2154c5c9cd4dSSage Weil * may change, so source mapping won't do), then recompress (or 2155c5c9cd4dSSage Weil * otherwise reinsert) a subrange. 2156c5c9cd4dSSage Weil * - allow ranges within the same file to be cloned (provided 2157c5c9cd4dSSage Weil * they don't overlap)? 2158c5c9cd4dSSage Weil */ 2159c5c9cd4dSSage Weil 2160e441d54dSChris Mason /* the destination must be opened for writing */ 21612ebc3464SDan Rosenberg if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND)) 2162e441d54dSChris Mason return -EINVAL; 2163e441d54dSChris Mason 2164b83cc969SLi Zefan if (btrfs_root_readonly(root)) 2165b83cc969SLi Zefan return -EROFS; 2166b83cc969SLi Zefan 2167c146afadSYan Zheng ret = mnt_want_write(file->f_path.mnt); 2168c146afadSYan Zheng if (ret) 2169c146afadSYan Zheng return ret; 2170c146afadSYan Zheng 2171c5c9cd4dSSage Weil src_file = fget(srcfd); 2172ab67b7c1SYan Zheng if (!src_file) { 2173ab67b7c1SYan Zheng ret = -EBADF; 2174ab67b7c1SYan Zheng goto out_drop_write; 2175ab67b7c1SYan Zheng } 21765dc64164SDan Rosenberg 2177f46b5a66SChristoph Hellwig src = src_file->f_dentry->d_inode; 2178f46b5a66SChristoph Hellwig 2179c5c9cd4dSSage Weil ret = -EINVAL; 2180c5c9cd4dSSage Weil if (src == inode) 2181c5c9cd4dSSage Weil goto out_fput; 2182c5c9cd4dSSage Weil 21835dc64164SDan Rosenberg /* the src must be open for reading */ 21845dc64164SDan Rosenberg if (!(src_file->f_mode & FMODE_READ)) 21855dc64164SDan Rosenberg goto out_fput; 21865dc64164SDan Rosenberg 21870e7b824cSLi Zefan /* don't make the dst file partly checksummed */ 21880e7b824cSLi Zefan if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != 21890e7b824cSLi Zefan (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) 21900e7b824cSLi Zefan goto out_fput; 21910e7b824cSLi Zefan 2192ae01a0abSYan Zheng ret = -EISDIR; 2193ae01a0abSYan Zheng if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode)) 2194f46b5a66SChristoph Hellwig goto out_fput; 2195f46b5a66SChristoph Hellwig 2196ae01a0abSYan Zheng ret = -EXDEV; 2197ae01a0abSYan Zheng if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root) 2198ae01a0abSYan Zheng goto out_fput; 2199ae01a0abSYan Zheng 2200ae01a0abSYan Zheng ret = -ENOMEM; 2201ae01a0abSYan Zheng buf = vmalloc(btrfs_level_size(root, 0)); 2202ae01a0abSYan Zheng if (!buf) 2203ae01a0abSYan Zheng goto out_fput; 2204ae01a0abSYan Zheng 2205ae01a0abSYan Zheng path = btrfs_alloc_path(); 2206ae01a0abSYan Zheng if (!path) { 2207ae01a0abSYan Zheng vfree(buf); 2208ae01a0abSYan Zheng goto out_fput; 2209ae01a0abSYan Zheng } 2210ae01a0abSYan Zheng path->reada = 2; 2211ae01a0abSYan Zheng 2212f46b5a66SChristoph Hellwig if (inode < src) { 2213fccdae43SSage Weil mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT); 2214fccdae43SSage Weil mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD); 2215f46b5a66SChristoph Hellwig } else { 2216fccdae43SSage Weil mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT); 2217fccdae43SSage Weil mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD); 2218f46b5a66SChristoph Hellwig } 2219f46b5a66SChristoph Hellwig 2220c5c9cd4dSSage Weil /* determine range to clone */ 2221c5c9cd4dSSage Weil ret = -EINVAL; 22222ebc3464SDan Rosenberg if (off + len > src->i_size || off + len < off) 2223f46b5a66SChristoph Hellwig goto out_unlock; 2224c5c9cd4dSSage Weil if (len == 0) 2225c5c9cd4dSSage Weil olen = len = src->i_size - off; 2226c5c9cd4dSSage Weil /* if we extend to eof, continue to block boundary */ 2227c5c9cd4dSSage Weil if (off + len == src->i_size) 22282a6b8daeSLi Zefan len = ALIGN(src->i_size, bs) - off; 2229c5c9cd4dSSage Weil 2230c5c9cd4dSSage Weil /* verify the end result is block aligned */ 22312a6b8daeSLi Zefan if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) || 22322a6b8daeSLi Zefan !IS_ALIGNED(destoff, bs)) 2233c5c9cd4dSSage Weil goto out_unlock; 2234c5c9cd4dSSage Weil 2235d525e8abSLi Zefan if (destoff > inode->i_size) { 2236d525e8abSLi Zefan ret = btrfs_cont_expand(inode, inode->i_size, destoff); 2237d525e8abSLi Zefan if (ret) 2238d525e8abSLi Zefan goto out_unlock; 2239d525e8abSLi Zefan } 2240d525e8abSLi Zefan 224171ef0786SLi Zefan /* truncate page cache pages from target inode range */ 224271ef0786SLi Zefan truncate_inode_pages_range(&inode->i_data, destoff, 224371ef0786SLi Zefan PAGE_CACHE_ALIGN(destoff + len) - 1); 224471ef0786SLi Zefan 2245f46b5a66SChristoph Hellwig /* do any pending delalloc/csum calc on src, one way or 2246f46b5a66SChristoph Hellwig another, and lock file content */ 2247f46b5a66SChristoph Hellwig while (1) { 224831840ae1SZheng Yan struct btrfs_ordered_extent *ordered; 2249c5c9cd4dSSage Weil lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS); 22509a019196SSage Weil ordered = btrfs_lookup_first_ordered_extent(src, off+len); 22519a019196SSage Weil if (!ordered && 22529a019196SSage Weil !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len, 22539a019196SSage Weil EXTENT_DELALLOC, 0, NULL)) 2254f46b5a66SChristoph Hellwig break; 2255c5c9cd4dSSage Weil unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS); 2256ae01a0abSYan Zheng if (ordered) 2257ae01a0abSYan Zheng btrfs_put_ordered_extent(ordered); 22589a019196SSage Weil btrfs_wait_ordered_range(src, off, len); 2259f46b5a66SChristoph Hellwig } 2260f46b5a66SChristoph Hellwig 2261c5c9cd4dSSage Weil /* clone data */ 226233345d01SLi Zefan key.objectid = btrfs_ino(src); 2263ae01a0abSYan Zheng key.type = BTRFS_EXTENT_DATA_KEY; 2264ae01a0abSYan Zheng key.offset = 0; 2265f46b5a66SChristoph Hellwig 2266f46b5a66SChristoph Hellwig while (1) { 2267f46b5a66SChristoph Hellwig /* 2268f46b5a66SChristoph Hellwig * note the key will change type as we walk through the 2269f46b5a66SChristoph Hellwig * tree. 2270f46b5a66SChristoph Hellwig */ 2271a22285a6SYan, Zheng ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 2272f46b5a66SChristoph Hellwig if (ret < 0) 2273f46b5a66SChristoph Hellwig goto out; 2274f46b5a66SChristoph Hellwig 2275ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 2276ae01a0abSYan Zheng if (path->slots[0] >= nritems) { 2277f46b5a66SChristoph Hellwig ret = btrfs_next_leaf(root, path); 2278f46b5a66SChristoph Hellwig if (ret < 0) 2279f46b5a66SChristoph Hellwig goto out; 2280f46b5a66SChristoph Hellwig if (ret > 0) 2281f46b5a66SChristoph Hellwig break; 2282ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 2283f46b5a66SChristoph Hellwig } 2284f46b5a66SChristoph Hellwig leaf = path->nodes[0]; 2285f46b5a66SChristoph Hellwig slot = path->slots[0]; 2286f46b5a66SChristoph Hellwig 2287ae01a0abSYan Zheng btrfs_item_key_to_cpu(leaf, &key, slot); 2288d20f7043SChris Mason if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY || 228933345d01SLi Zefan key.objectid != btrfs_ino(src)) 2290f46b5a66SChristoph Hellwig break; 2291f46b5a66SChristoph Hellwig 2292c5c9cd4dSSage Weil if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) { 2293c5c9cd4dSSage Weil struct btrfs_file_extent_item *extent; 2294c5c9cd4dSSage Weil int type; 229531840ae1SZheng Yan u32 size; 229631840ae1SZheng Yan struct btrfs_key new_key; 2297c5c9cd4dSSage Weil u64 disko = 0, diskl = 0; 2298c5c9cd4dSSage Weil u64 datao = 0, datal = 0; 2299c5c9cd4dSSage Weil u8 comp; 2300b5384d48SSage Weil u64 endoff; 230131840ae1SZheng Yan 230231840ae1SZheng Yan size = btrfs_item_size_nr(leaf, slot); 230331840ae1SZheng Yan read_extent_buffer(leaf, buf, 230431840ae1SZheng Yan btrfs_item_ptr_offset(leaf, slot), 230531840ae1SZheng Yan size); 2306c5c9cd4dSSage Weil 2307c5c9cd4dSSage Weil extent = btrfs_item_ptr(leaf, slot, 2308c5c9cd4dSSage Weil struct btrfs_file_extent_item); 2309c5c9cd4dSSage Weil comp = btrfs_file_extent_compression(leaf, extent); 2310c5c9cd4dSSage Weil type = btrfs_file_extent_type(leaf, extent); 2311c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 2312c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 2313d397712bSChris Mason disko = btrfs_file_extent_disk_bytenr(leaf, 2314d397712bSChris Mason extent); 2315d397712bSChris Mason diskl = btrfs_file_extent_disk_num_bytes(leaf, 2316d397712bSChris Mason extent); 2317c5c9cd4dSSage Weil datao = btrfs_file_extent_offset(leaf, extent); 2318d397712bSChris Mason datal = btrfs_file_extent_num_bytes(leaf, 2319d397712bSChris Mason extent); 2320c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 2321c5c9cd4dSSage Weil /* take upper bound, may be compressed */ 2322c5c9cd4dSSage Weil datal = btrfs_file_extent_ram_bytes(leaf, 2323c5c9cd4dSSage Weil extent); 2324c5c9cd4dSSage Weil } 2325b3b4aa74SDavid Sterba btrfs_release_path(path); 232631840ae1SZheng Yan 2327050006a7SSage Weil if (key.offset + datal <= off || 2328c5c9cd4dSSage Weil key.offset >= off+len) 2329c5c9cd4dSSage Weil goto next; 2330c5c9cd4dSSage Weil 233131840ae1SZheng Yan memcpy(&new_key, &key, sizeof(new_key)); 233233345d01SLi Zefan new_key.objectid = btrfs_ino(inode); 23334d728ec7SLi Zefan if (off <= key.offset) 2334c5c9cd4dSSage Weil new_key.offset = key.offset + destoff - off; 23354d728ec7SLi Zefan else 23364d728ec7SLi Zefan new_key.offset = destoff; 2337c5c9cd4dSSage Weil 2338b6f3409bSSage Weil /* 2339b6f3409bSSage Weil * 1 - adjusting old extent (we may have to split it) 2340b6f3409bSSage Weil * 1 - add new extent 2341b6f3409bSSage Weil * 1 - inode update 2342b6f3409bSSage Weil */ 2343b6f3409bSSage Weil trans = btrfs_start_transaction(root, 3); 2344a22285a6SYan, Zheng if (IS_ERR(trans)) { 2345a22285a6SYan, Zheng ret = PTR_ERR(trans); 2346a22285a6SYan, Zheng goto out; 2347a22285a6SYan, Zheng } 2348a22285a6SYan, Zheng 2349c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 2350c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 2351d72c0842SLi Zefan /* 2352d72c0842SLi Zefan * a | --- range to clone ---| b 2353d72c0842SLi Zefan * | ------------- extent ------------- | 2354d72c0842SLi Zefan */ 2355d72c0842SLi Zefan 2356d72c0842SLi Zefan /* substract range b */ 2357d72c0842SLi Zefan if (key.offset + datal > off + len) 2358d72c0842SLi Zefan datal = off + len - key.offset; 2359d72c0842SLi Zefan 2360d72c0842SLi Zefan /* substract range a */ 2361a22285a6SYan, Zheng if (off > key.offset) { 2362a22285a6SYan, Zheng datao += off - key.offset; 2363a22285a6SYan, Zheng datal -= off - key.offset; 2364a22285a6SYan, Zheng } 2365a22285a6SYan, Zheng 2366a22285a6SYan, Zheng ret = btrfs_drop_extents(trans, inode, 2367a22285a6SYan, Zheng new_key.offset, 2368a22285a6SYan, Zheng new_key.offset + datal, 2369a22285a6SYan, Zheng &hint_byte, 1); 2370a22285a6SYan, Zheng BUG_ON(ret); 2371a22285a6SYan, Zheng 237231840ae1SZheng Yan ret = btrfs_insert_empty_item(trans, root, path, 237331840ae1SZheng Yan &new_key, size); 2374a22285a6SYan, Zheng BUG_ON(ret); 237531840ae1SZheng Yan 237631840ae1SZheng Yan leaf = path->nodes[0]; 237731840ae1SZheng Yan slot = path->slots[0]; 237831840ae1SZheng Yan write_extent_buffer(leaf, buf, 237931840ae1SZheng Yan btrfs_item_ptr_offset(leaf, slot), 238031840ae1SZheng Yan size); 2381ae01a0abSYan Zheng 2382f46b5a66SChristoph Hellwig extent = btrfs_item_ptr(leaf, slot, 2383f46b5a66SChristoph Hellwig struct btrfs_file_extent_item); 2384c5c9cd4dSSage Weil 2385c5c9cd4dSSage Weil /* disko == 0 means it's a hole */ 2386c5c9cd4dSSage Weil if (!disko) 2387c5c9cd4dSSage Weil datao = 0; 2388c5c9cd4dSSage Weil 2389c5c9cd4dSSage Weil btrfs_set_file_extent_offset(leaf, extent, 2390c5c9cd4dSSage Weil datao); 2391c5c9cd4dSSage Weil btrfs_set_file_extent_num_bytes(leaf, extent, 2392c5c9cd4dSSage Weil datal); 2393c5c9cd4dSSage Weil if (disko) { 2394c5c9cd4dSSage Weil inode_add_bytes(inode, datal); 2395ae01a0abSYan Zheng ret = btrfs_inc_extent_ref(trans, root, 23965d4f98a2SYan Zheng disko, diskl, 0, 2397f46b5a66SChristoph Hellwig root->root_key.objectid, 239833345d01SLi Zefan btrfs_ino(inode), 23995d4f98a2SYan Zheng new_key.offset - datao); 2400ae01a0abSYan Zheng BUG_ON(ret); 2401f46b5a66SChristoph Hellwig } 2402c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 2403c5c9cd4dSSage Weil u64 skip = 0; 2404c5c9cd4dSSage Weil u64 trim = 0; 2405c5c9cd4dSSage Weil if (off > key.offset) { 2406c5c9cd4dSSage Weil skip = off - key.offset; 2407c5c9cd4dSSage Weil new_key.offset += skip; 240831840ae1SZheng Yan } 2409d397712bSChris Mason 2410c5c9cd4dSSage Weil if (key.offset + datal > off+len) 2411c5c9cd4dSSage Weil trim = key.offset + datal - (off+len); 2412d397712bSChris Mason 2413c5c9cd4dSSage Weil if (comp && (skip || trim)) { 2414c5c9cd4dSSage Weil ret = -EINVAL; 2415a22285a6SYan, Zheng btrfs_end_transaction(trans, root); 2416c5c9cd4dSSage Weil goto out; 241731840ae1SZheng Yan } 2418c5c9cd4dSSage Weil size -= skip + trim; 2419c5c9cd4dSSage Weil datal -= skip + trim; 2420a22285a6SYan, Zheng 2421a22285a6SYan, Zheng ret = btrfs_drop_extents(trans, inode, 2422a22285a6SYan, Zheng new_key.offset, 2423a22285a6SYan, Zheng new_key.offset + datal, 2424a22285a6SYan, Zheng &hint_byte, 1); 2425a22285a6SYan, Zheng BUG_ON(ret); 2426a22285a6SYan, Zheng 2427c5c9cd4dSSage Weil ret = btrfs_insert_empty_item(trans, root, path, 2428c5c9cd4dSSage Weil &new_key, size); 2429a22285a6SYan, Zheng BUG_ON(ret); 2430c5c9cd4dSSage Weil 2431c5c9cd4dSSage Weil if (skip) { 2432d397712bSChris Mason u32 start = 2433d397712bSChris Mason btrfs_file_extent_calc_inline_size(0); 2434c5c9cd4dSSage Weil memmove(buf+start, buf+start+skip, 2435c5c9cd4dSSage Weil datal); 2436c5c9cd4dSSage Weil } 2437c5c9cd4dSSage Weil 2438c5c9cd4dSSage Weil leaf = path->nodes[0]; 2439c5c9cd4dSSage Weil slot = path->slots[0]; 2440c5c9cd4dSSage Weil write_extent_buffer(leaf, buf, 2441c5c9cd4dSSage Weil btrfs_item_ptr_offset(leaf, slot), 2442c5c9cd4dSSage Weil size); 2443c5c9cd4dSSage Weil inode_add_bytes(inode, datal); 2444c5c9cd4dSSage Weil } 2445c5c9cd4dSSage Weil 2446c5c9cd4dSSage Weil btrfs_mark_buffer_dirty(leaf); 2447b3b4aa74SDavid Sterba btrfs_release_path(path); 2448c5c9cd4dSSage Weil 2449a22285a6SYan, Zheng inode->i_mtime = inode->i_ctime = CURRENT_TIME; 2450b5384d48SSage Weil 2451b5384d48SSage Weil /* 2452b5384d48SSage Weil * we round up to the block size at eof when 2453b5384d48SSage Weil * determining which extents to clone above, 2454b5384d48SSage Weil * but shouldn't round up the file size 2455b5384d48SSage Weil */ 2456b5384d48SSage Weil endoff = new_key.offset + datal; 24575f3888ffSLi Zefan if (endoff > destoff+olen) 24585f3888ffSLi Zefan endoff = destoff+olen; 2459b5384d48SSage Weil if (endoff > inode->i_size) 2460b5384d48SSage Weil btrfs_i_size_write(inode, endoff); 2461b5384d48SSage Weil 2462a22285a6SYan, Zheng ret = btrfs_update_inode(trans, root, inode); 2463a22285a6SYan, Zheng BUG_ON(ret); 2464a22285a6SYan, Zheng btrfs_end_transaction(trans, root); 2465a22285a6SYan, Zheng } 2466c5c9cd4dSSage Weil next: 2467b3b4aa74SDavid Sterba btrfs_release_path(path); 2468ae01a0abSYan Zheng key.offset++; 2469ae01a0abSYan Zheng } 2470f46b5a66SChristoph Hellwig ret = 0; 2471f46b5a66SChristoph Hellwig out: 2472b3b4aa74SDavid Sterba btrfs_release_path(path); 2473c5c9cd4dSSage Weil unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS); 2474f46b5a66SChristoph Hellwig out_unlock: 2475f46b5a66SChristoph Hellwig mutex_unlock(&src->i_mutex); 2476f46b5a66SChristoph Hellwig mutex_unlock(&inode->i_mutex); 2477ae01a0abSYan Zheng vfree(buf); 2478ae01a0abSYan Zheng btrfs_free_path(path); 2479f46b5a66SChristoph Hellwig out_fput: 2480f46b5a66SChristoph Hellwig fput(src_file); 2481ab67b7c1SYan Zheng out_drop_write: 2482ab67b7c1SYan Zheng mnt_drop_write(file->f_path.mnt); 2483f46b5a66SChristoph Hellwig return ret; 2484f46b5a66SChristoph Hellwig } 2485f46b5a66SChristoph Hellwig 24867a865e8aSChristoph Hellwig static long btrfs_ioctl_clone_range(struct file *file, void __user *argp) 2487c5c9cd4dSSage Weil { 2488c5c9cd4dSSage Weil struct btrfs_ioctl_clone_range_args args; 2489c5c9cd4dSSage Weil 24907a865e8aSChristoph Hellwig if (copy_from_user(&args, argp, sizeof(args))) 2491c5c9cd4dSSage Weil return -EFAULT; 2492c5c9cd4dSSage Weil return btrfs_ioctl_clone(file, args.src_fd, args.src_offset, 2493c5c9cd4dSSage Weil args.src_length, args.dest_offset); 2494c5c9cd4dSSage Weil } 2495c5c9cd4dSSage Weil 2496f46b5a66SChristoph Hellwig /* 2497f46b5a66SChristoph Hellwig * there are many ways the trans_start and trans_end ioctls can lead 2498f46b5a66SChristoph Hellwig * to deadlocks. They should only be used by applications that 2499f46b5a66SChristoph Hellwig * basically own the machine, and have a very in depth understanding 2500f46b5a66SChristoph Hellwig * of all the possible deadlocks and enospc problems. 2501f46b5a66SChristoph Hellwig */ 2502b2950863SChristoph Hellwig static long btrfs_ioctl_trans_start(struct file *file) 2503f46b5a66SChristoph Hellwig { 2504f46b5a66SChristoph Hellwig struct inode *inode = fdentry(file)->d_inode; 2505f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 2506f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 25071ab86aedSSage Weil int ret; 2508f46b5a66SChristoph Hellwig 25091ab86aedSSage Weil ret = -EPERM; 2510df5b5520SChristoph Hellwig if (!capable(CAP_SYS_ADMIN)) 2511f46b5a66SChristoph Hellwig goto out; 25121ab86aedSSage Weil 25131ab86aedSSage Weil ret = -EINPROGRESS; 25141ab86aedSSage Weil if (file->private_data) 25151ab86aedSSage Weil goto out; 25169ca9ee09SSage Weil 2517b83cc969SLi Zefan ret = -EROFS; 2518b83cc969SLi Zefan if (btrfs_root_readonly(root)) 2519b83cc969SLi Zefan goto out; 2520b83cc969SLi Zefan 2521c146afadSYan Zheng ret = mnt_want_write(file->f_path.mnt); 2522c146afadSYan Zheng if (ret) 2523c146afadSYan Zheng goto out; 2524c146afadSYan Zheng 2525a4abeea4SJosef Bacik atomic_inc(&root->fs_info->open_ioctl_trans); 25269ca9ee09SSage Weil 2527f46b5a66SChristoph Hellwig ret = -ENOMEM; 25287a7eaa40SJosef Bacik trans = btrfs_start_ioctl_transaction(root); 2529abd30bb0STsutomu Itoh if (IS_ERR(trans)) 25301ab86aedSSage Weil goto out_drop; 25311ab86aedSSage Weil 25321ab86aedSSage Weil file->private_data = trans; 25331ab86aedSSage Weil return 0; 25341ab86aedSSage Weil 25351ab86aedSSage Weil out_drop: 2536a4abeea4SJosef Bacik atomic_dec(&root->fs_info->open_ioctl_trans); 25371ab86aedSSage Weil mnt_drop_write(file->f_path.mnt); 2538f46b5a66SChristoph Hellwig out: 2539f46b5a66SChristoph Hellwig return ret; 2540f46b5a66SChristoph Hellwig } 2541f46b5a66SChristoph Hellwig 25426ef5ed0dSJosef Bacik static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) 25436ef5ed0dSJosef Bacik { 25446ef5ed0dSJosef Bacik struct inode *inode = fdentry(file)->d_inode; 25456ef5ed0dSJosef Bacik struct btrfs_root *root = BTRFS_I(inode)->root; 25466ef5ed0dSJosef Bacik struct btrfs_root *new_root; 25476ef5ed0dSJosef Bacik struct btrfs_dir_item *di; 25486ef5ed0dSJosef Bacik struct btrfs_trans_handle *trans; 25496ef5ed0dSJosef Bacik struct btrfs_path *path; 25506ef5ed0dSJosef Bacik struct btrfs_key location; 25516ef5ed0dSJosef Bacik struct btrfs_disk_key disk_key; 25526ef5ed0dSJosef Bacik struct btrfs_super_block *disk_super; 25536ef5ed0dSJosef Bacik u64 features; 25546ef5ed0dSJosef Bacik u64 objectid = 0; 25556ef5ed0dSJosef Bacik u64 dir_id; 25566ef5ed0dSJosef Bacik 25576ef5ed0dSJosef Bacik if (!capable(CAP_SYS_ADMIN)) 25586ef5ed0dSJosef Bacik return -EPERM; 25596ef5ed0dSJosef Bacik 25606ef5ed0dSJosef Bacik if (copy_from_user(&objectid, argp, sizeof(objectid))) 25616ef5ed0dSJosef Bacik return -EFAULT; 25626ef5ed0dSJosef Bacik 25636ef5ed0dSJosef Bacik if (!objectid) 25646ef5ed0dSJosef Bacik objectid = root->root_key.objectid; 25656ef5ed0dSJosef Bacik 25666ef5ed0dSJosef Bacik location.objectid = objectid; 25676ef5ed0dSJosef Bacik location.type = BTRFS_ROOT_ITEM_KEY; 25686ef5ed0dSJosef Bacik location.offset = (u64)-1; 25696ef5ed0dSJosef Bacik 25706ef5ed0dSJosef Bacik new_root = btrfs_read_fs_root_no_name(root->fs_info, &location); 25716ef5ed0dSJosef Bacik if (IS_ERR(new_root)) 25726ef5ed0dSJosef Bacik return PTR_ERR(new_root); 25736ef5ed0dSJosef Bacik 25746ef5ed0dSJosef Bacik if (btrfs_root_refs(&new_root->root_item) == 0) 25756ef5ed0dSJosef Bacik return -ENOENT; 25766ef5ed0dSJosef Bacik 25776ef5ed0dSJosef Bacik path = btrfs_alloc_path(); 25786ef5ed0dSJosef Bacik if (!path) 25796ef5ed0dSJosef Bacik return -ENOMEM; 25806ef5ed0dSJosef Bacik path->leave_spinning = 1; 25816ef5ed0dSJosef Bacik 25826ef5ed0dSJosef Bacik trans = btrfs_start_transaction(root, 1); 258398d5dc13STsutomu Itoh if (IS_ERR(trans)) { 25846ef5ed0dSJosef Bacik btrfs_free_path(path); 258598d5dc13STsutomu Itoh return PTR_ERR(trans); 25866ef5ed0dSJosef Bacik } 25876ef5ed0dSJosef Bacik 25886ef5ed0dSJosef Bacik dir_id = btrfs_super_root_dir(&root->fs_info->super_copy); 25896ef5ed0dSJosef Bacik di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path, 25906ef5ed0dSJosef Bacik dir_id, "default", 7, 1); 2591cf1e99a4SDan Carpenter if (IS_ERR_OR_NULL(di)) { 25926ef5ed0dSJosef Bacik btrfs_free_path(path); 25936ef5ed0dSJosef Bacik btrfs_end_transaction(trans, root); 25946ef5ed0dSJosef Bacik printk(KERN_ERR "Umm, you don't have the default dir item, " 25956ef5ed0dSJosef Bacik "this isn't going to work\n"); 25966ef5ed0dSJosef Bacik return -ENOENT; 25976ef5ed0dSJosef Bacik } 25986ef5ed0dSJosef Bacik 25996ef5ed0dSJosef Bacik btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); 26006ef5ed0dSJosef Bacik btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); 26016ef5ed0dSJosef Bacik btrfs_mark_buffer_dirty(path->nodes[0]); 26026ef5ed0dSJosef Bacik btrfs_free_path(path); 26036ef5ed0dSJosef Bacik 26046ef5ed0dSJosef Bacik disk_super = &root->fs_info->super_copy; 26056ef5ed0dSJosef Bacik features = btrfs_super_incompat_flags(disk_super); 26066ef5ed0dSJosef Bacik if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) { 26076ef5ed0dSJosef Bacik features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL; 26086ef5ed0dSJosef Bacik btrfs_set_super_incompat_flags(disk_super, features); 26096ef5ed0dSJosef Bacik } 26106ef5ed0dSJosef Bacik btrfs_end_transaction(trans, root); 26116ef5ed0dSJosef Bacik 26126ef5ed0dSJosef Bacik return 0; 26136ef5ed0dSJosef Bacik } 26146ef5ed0dSJosef Bacik 2615bf5fc093SJosef Bacik static void get_block_group_info(struct list_head *groups_list, 2616bf5fc093SJosef Bacik struct btrfs_ioctl_space_info *space) 2617bf5fc093SJosef Bacik { 2618bf5fc093SJosef Bacik struct btrfs_block_group_cache *block_group; 2619bf5fc093SJosef Bacik 2620bf5fc093SJosef Bacik space->total_bytes = 0; 2621bf5fc093SJosef Bacik space->used_bytes = 0; 2622bf5fc093SJosef Bacik space->flags = 0; 2623bf5fc093SJosef Bacik list_for_each_entry(block_group, groups_list, list) { 2624bf5fc093SJosef Bacik space->flags = block_group->flags; 2625bf5fc093SJosef Bacik space->total_bytes += block_group->key.offset; 2626bf5fc093SJosef Bacik space->used_bytes += 2627bf5fc093SJosef Bacik btrfs_block_group_used(&block_group->item); 2628bf5fc093SJosef Bacik } 2629bf5fc093SJosef Bacik } 2630bf5fc093SJosef Bacik 26311406e432SJosef Bacik long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg) 26321406e432SJosef Bacik { 26331406e432SJosef Bacik struct btrfs_ioctl_space_args space_args; 26341406e432SJosef Bacik struct btrfs_ioctl_space_info space; 26351406e432SJosef Bacik struct btrfs_ioctl_space_info *dest; 26367fde62bfSChris Mason struct btrfs_ioctl_space_info *dest_orig; 263713f2696fSDaniel J Blueman struct btrfs_ioctl_space_info __user *user_dest; 26381406e432SJosef Bacik struct btrfs_space_info *info; 2639bf5fc093SJosef Bacik u64 types[] = {BTRFS_BLOCK_GROUP_DATA, 2640bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_SYSTEM, 2641bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_METADATA, 2642bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA}; 2643bf5fc093SJosef Bacik int num_types = 4; 26447fde62bfSChris Mason int alloc_size; 26451406e432SJosef Bacik int ret = 0; 264651788b1bSDan Rosenberg u64 slot_count = 0; 2647bf5fc093SJosef Bacik int i, c; 26481406e432SJosef Bacik 26491406e432SJosef Bacik if (copy_from_user(&space_args, 26501406e432SJosef Bacik (struct btrfs_ioctl_space_args __user *)arg, 26511406e432SJosef Bacik sizeof(space_args))) 26521406e432SJosef Bacik return -EFAULT; 26531406e432SJosef Bacik 2654bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 2655bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 2656bf5fc093SJosef Bacik 2657bf5fc093SJosef Bacik info = NULL; 26587fde62bfSChris Mason rcu_read_lock(); 2659bf5fc093SJosef Bacik list_for_each_entry_rcu(tmp, &root->fs_info->space_info, 2660bf5fc093SJosef Bacik list) { 2661bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 2662bf5fc093SJosef Bacik info = tmp; 2663bf5fc093SJosef Bacik break; 2664bf5fc093SJosef Bacik } 2665bf5fc093SJosef Bacik } 26667fde62bfSChris Mason rcu_read_unlock(); 26671406e432SJosef Bacik 2668bf5fc093SJosef Bacik if (!info) 2669bf5fc093SJosef Bacik continue; 2670bf5fc093SJosef Bacik 2671bf5fc093SJosef Bacik down_read(&info->groups_sem); 2672bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 2673bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) 2674bf5fc093SJosef Bacik slot_count++; 2675bf5fc093SJosef Bacik } 2676bf5fc093SJosef Bacik up_read(&info->groups_sem); 2677bf5fc093SJosef Bacik } 2678bf5fc093SJosef Bacik 26797fde62bfSChris Mason /* space_slots == 0 means they are asking for a count */ 26807fde62bfSChris Mason if (space_args.space_slots == 0) { 26817fde62bfSChris Mason space_args.total_spaces = slot_count; 26827fde62bfSChris Mason goto out; 26837fde62bfSChris Mason } 2684bf5fc093SJosef Bacik 268551788b1bSDan Rosenberg slot_count = min_t(u64, space_args.space_slots, slot_count); 2686bf5fc093SJosef Bacik 26877fde62bfSChris Mason alloc_size = sizeof(*dest) * slot_count; 2688bf5fc093SJosef Bacik 26897fde62bfSChris Mason /* we generally have at most 6 or so space infos, one for each raid 26907fde62bfSChris Mason * level. So, a whole page should be more than enough for everyone 26917fde62bfSChris Mason */ 26927fde62bfSChris Mason if (alloc_size > PAGE_CACHE_SIZE) 26937fde62bfSChris Mason return -ENOMEM; 26947fde62bfSChris Mason 26957fde62bfSChris Mason space_args.total_spaces = 0; 26967fde62bfSChris Mason dest = kmalloc(alloc_size, GFP_NOFS); 26977fde62bfSChris Mason if (!dest) 26987fde62bfSChris Mason return -ENOMEM; 26997fde62bfSChris Mason dest_orig = dest; 27007fde62bfSChris Mason 27017fde62bfSChris Mason /* now we have a buffer to copy into */ 2702bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 2703bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 2704bf5fc093SJosef Bacik 270551788b1bSDan Rosenberg if (!slot_count) 270651788b1bSDan Rosenberg break; 270751788b1bSDan Rosenberg 2708bf5fc093SJosef Bacik info = NULL; 27091406e432SJosef Bacik rcu_read_lock(); 2710bf5fc093SJosef Bacik list_for_each_entry_rcu(tmp, &root->fs_info->space_info, 2711bf5fc093SJosef Bacik list) { 2712bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 2713bf5fc093SJosef Bacik info = tmp; 27147fde62bfSChris Mason break; 2715bf5fc093SJosef Bacik } 2716bf5fc093SJosef Bacik } 2717bf5fc093SJosef Bacik rcu_read_unlock(); 27187fde62bfSChris Mason 2719bf5fc093SJosef Bacik if (!info) 2720bf5fc093SJosef Bacik continue; 2721bf5fc093SJosef Bacik down_read(&info->groups_sem); 2722bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 2723bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) { 2724bf5fc093SJosef Bacik get_block_group_info(&info->block_groups[c], 2725bf5fc093SJosef Bacik &space); 27267fde62bfSChris Mason memcpy(dest, &space, sizeof(space)); 27271406e432SJosef Bacik dest++; 27281406e432SJosef Bacik space_args.total_spaces++; 272951788b1bSDan Rosenberg slot_count--; 27301406e432SJosef Bacik } 273151788b1bSDan Rosenberg if (!slot_count) 273251788b1bSDan Rosenberg break; 2733bf5fc093SJosef Bacik } 2734bf5fc093SJosef Bacik up_read(&info->groups_sem); 2735bf5fc093SJosef Bacik } 27361406e432SJosef Bacik 27377fde62bfSChris Mason user_dest = (struct btrfs_ioctl_space_info *) 27387fde62bfSChris Mason (arg + sizeof(struct btrfs_ioctl_space_args)); 27397fde62bfSChris Mason 27407fde62bfSChris Mason if (copy_to_user(user_dest, dest_orig, alloc_size)) 27417fde62bfSChris Mason ret = -EFAULT; 27427fde62bfSChris Mason 27437fde62bfSChris Mason kfree(dest_orig); 27447fde62bfSChris Mason out: 27457fde62bfSChris Mason if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) 27461406e432SJosef Bacik ret = -EFAULT; 27471406e432SJosef Bacik 27481406e432SJosef Bacik return ret; 27491406e432SJosef Bacik } 27501406e432SJosef Bacik 2751f46b5a66SChristoph Hellwig /* 2752f46b5a66SChristoph Hellwig * there are many ways the trans_start and trans_end ioctls can lead 2753f46b5a66SChristoph Hellwig * to deadlocks. They should only be used by applications that 2754f46b5a66SChristoph Hellwig * basically own the machine, and have a very in depth understanding 2755f46b5a66SChristoph Hellwig * of all the possible deadlocks and enospc problems. 2756f46b5a66SChristoph Hellwig */ 2757f46b5a66SChristoph Hellwig long btrfs_ioctl_trans_end(struct file *file) 2758f46b5a66SChristoph Hellwig { 2759f46b5a66SChristoph Hellwig struct inode *inode = fdentry(file)->d_inode; 2760f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 2761f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 2762f46b5a66SChristoph Hellwig 2763f46b5a66SChristoph Hellwig trans = file->private_data; 27641ab86aedSSage Weil if (!trans) 27651ab86aedSSage Weil return -EINVAL; 2766b214107eSChristoph Hellwig file->private_data = NULL; 27679ca9ee09SSage Weil 27681ab86aedSSage Weil btrfs_end_transaction(trans, root); 27691ab86aedSSage Weil 2770a4abeea4SJosef Bacik atomic_dec(&root->fs_info->open_ioctl_trans); 27719ca9ee09SSage Weil 2772cfc8ea87SSage Weil mnt_drop_write(file->f_path.mnt); 27731ab86aedSSage Weil return 0; 2774f46b5a66SChristoph Hellwig } 2775f46b5a66SChristoph Hellwig 277646204592SSage Weil static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp) 277746204592SSage Weil { 277846204592SSage Weil struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root; 277946204592SSage Weil struct btrfs_trans_handle *trans; 278046204592SSage Weil u64 transid; 2781db5b493aSTsutomu Itoh int ret; 278246204592SSage Weil 278346204592SSage Weil trans = btrfs_start_transaction(root, 0); 278498d5dc13STsutomu Itoh if (IS_ERR(trans)) 278598d5dc13STsutomu Itoh return PTR_ERR(trans); 278646204592SSage Weil transid = trans->transid; 2787db5b493aSTsutomu Itoh ret = btrfs_commit_transaction_async(trans, root, 0); 27888b2b2d3cSTsutomu Itoh if (ret) { 27898b2b2d3cSTsutomu Itoh btrfs_end_transaction(trans, root); 2790db5b493aSTsutomu Itoh return ret; 27918b2b2d3cSTsutomu Itoh } 279246204592SSage Weil 279346204592SSage Weil if (argp) 279446204592SSage Weil if (copy_to_user(argp, &transid, sizeof(transid))) 279546204592SSage Weil return -EFAULT; 279646204592SSage Weil return 0; 279746204592SSage Weil } 279846204592SSage Weil 279946204592SSage Weil static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp) 280046204592SSage Weil { 280146204592SSage Weil struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root; 280246204592SSage Weil u64 transid; 280346204592SSage Weil 280446204592SSage Weil if (argp) { 280546204592SSage Weil if (copy_from_user(&transid, argp, sizeof(transid))) 280646204592SSage Weil return -EFAULT; 280746204592SSage Weil } else { 280846204592SSage Weil transid = 0; /* current trans */ 280946204592SSage Weil } 281046204592SSage Weil return btrfs_wait_for_commit(root, transid); 281146204592SSage Weil } 281246204592SSage Weil 2813475f6387SJan Schmidt static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg) 2814475f6387SJan Schmidt { 2815475f6387SJan Schmidt int ret; 2816475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 2817475f6387SJan Schmidt 2818475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 2819475f6387SJan Schmidt return -EPERM; 2820475f6387SJan Schmidt 2821475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 2822475f6387SJan Schmidt if (IS_ERR(sa)) 2823475f6387SJan Schmidt return PTR_ERR(sa); 2824475f6387SJan Schmidt 2825475f6387SJan Schmidt ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end, 28268628764eSArne Jansen &sa->progress, sa->flags & BTRFS_SCRUB_READONLY); 2827475f6387SJan Schmidt 2828475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 2829475f6387SJan Schmidt ret = -EFAULT; 2830475f6387SJan Schmidt 2831475f6387SJan Schmidt kfree(sa); 2832475f6387SJan Schmidt return ret; 2833475f6387SJan Schmidt } 2834475f6387SJan Schmidt 2835475f6387SJan Schmidt static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg) 2836475f6387SJan Schmidt { 2837475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 2838475f6387SJan Schmidt return -EPERM; 2839475f6387SJan Schmidt 2840475f6387SJan Schmidt return btrfs_scrub_cancel(root); 2841475f6387SJan Schmidt } 2842475f6387SJan Schmidt 2843475f6387SJan Schmidt static long btrfs_ioctl_scrub_progress(struct btrfs_root *root, 2844475f6387SJan Schmidt void __user *arg) 2845475f6387SJan Schmidt { 2846475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 2847475f6387SJan Schmidt int ret; 2848475f6387SJan Schmidt 2849475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 2850475f6387SJan Schmidt return -EPERM; 2851475f6387SJan Schmidt 2852475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 2853475f6387SJan Schmidt if (IS_ERR(sa)) 2854475f6387SJan Schmidt return PTR_ERR(sa); 2855475f6387SJan Schmidt 2856475f6387SJan Schmidt ret = btrfs_scrub_progress(root, sa->devid, &sa->progress); 2857475f6387SJan Schmidt 2858475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 2859475f6387SJan Schmidt ret = -EFAULT; 2860475f6387SJan Schmidt 2861475f6387SJan Schmidt kfree(sa); 2862475f6387SJan Schmidt return ret; 2863475f6387SJan Schmidt } 2864475f6387SJan Schmidt 2865f46b5a66SChristoph Hellwig long btrfs_ioctl(struct file *file, unsigned int 2866f46b5a66SChristoph Hellwig cmd, unsigned long arg) 2867f46b5a66SChristoph Hellwig { 2868f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; 28694bcabaa3SChristoph Hellwig void __user *argp = (void __user *)arg; 2870f46b5a66SChristoph Hellwig 2871f46b5a66SChristoph Hellwig switch (cmd) { 28726cbff00fSChristoph Hellwig case FS_IOC_GETFLAGS: 28736cbff00fSChristoph Hellwig return btrfs_ioctl_getflags(file, argp); 28746cbff00fSChristoph Hellwig case FS_IOC_SETFLAGS: 28756cbff00fSChristoph Hellwig return btrfs_ioctl_setflags(file, argp); 28766cbff00fSChristoph Hellwig case FS_IOC_GETVERSION: 28776cbff00fSChristoph Hellwig return btrfs_ioctl_getversion(file, argp); 2878f7039b1dSLi Dongyang case FITRIM: 2879f7039b1dSLi Dongyang return btrfs_ioctl_fitrim(file, argp); 2880f46b5a66SChristoph Hellwig case BTRFS_IOC_SNAP_CREATE: 2881fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 0); 2882fdfb1e4fSLi Zefan case BTRFS_IOC_SNAP_CREATE_V2: 2883fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create_v2(file, argp, 0); 28843de4586cSChris Mason case BTRFS_IOC_SUBVOL_CREATE: 2885fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 1); 288676dda93cSYan, Zheng case BTRFS_IOC_SNAP_DESTROY: 288776dda93cSYan, Zheng return btrfs_ioctl_snap_destroy(file, argp); 28880caa102dSLi Zefan case BTRFS_IOC_SUBVOL_GETFLAGS: 28890caa102dSLi Zefan return btrfs_ioctl_subvol_getflags(file, argp); 28900caa102dSLi Zefan case BTRFS_IOC_SUBVOL_SETFLAGS: 28910caa102dSLi Zefan return btrfs_ioctl_subvol_setflags(file, argp); 28926ef5ed0dSJosef Bacik case BTRFS_IOC_DEFAULT_SUBVOL: 28936ef5ed0dSJosef Bacik return btrfs_ioctl_default_subvol(file, argp); 2894f46b5a66SChristoph Hellwig case BTRFS_IOC_DEFRAG: 28951e701a32SChris Mason return btrfs_ioctl_defrag(file, NULL); 28961e701a32SChris Mason case BTRFS_IOC_DEFRAG_RANGE: 28971e701a32SChris Mason return btrfs_ioctl_defrag(file, argp); 2898f46b5a66SChristoph Hellwig case BTRFS_IOC_RESIZE: 28994bcabaa3SChristoph Hellwig return btrfs_ioctl_resize(root, argp); 2900f46b5a66SChristoph Hellwig case BTRFS_IOC_ADD_DEV: 29014bcabaa3SChristoph Hellwig return btrfs_ioctl_add_dev(root, argp); 2902f46b5a66SChristoph Hellwig case BTRFS_IOC_RM_DEV: 29034bcabaa3SChristoph Hellwig return btrfs_ioctl_rm_dev(root, argp); 2904475f6387SJan Schmidt case BTRFS_IOC_FS_INFO: 2905475f6387SJan Schmidt return btrfs_ioctl_fs_info(root, argp); 2906475f6387SJan Schmidt case BTRFS_IOC_DEV_INFO: 2907475f6387SJan Schmidt return btrfs_ioctl_dev_info(root, argp); 2908f46b5a66SChristoph Hellwig case BTRFS_IOC_BALANCE: 2909f46b5a66SChristoph Hellwig return btrfs_balance(root->fs_info->dev_root); 2910f46b5a66SChristoph Hellwig case BTRFS_IOC_CLONE: 2911c5c9cd4dSSage Weil return btrfs_ioctl_clone(file, arg, 0, 0, 0); 2912c5c9cd4dSSage Weil case BTRFS_IOC_CLONE_RANGE: 29137a865e8aSChristoph Hellwig return btrfs_ioctl_clone_range(file, argp); 2914f46b5a66SChristoph Hellwig case BTRFS_IOC_TRANS_START: 2915f46b5a66SChristoph Hellwig return btrfs_ioctl_trans_start(file); 2916f46b5a66SChristoph Hellwig case BTRFS_IOC_TRANS_END: 2917f46b5a66SChristoph Hellwig return btrfs_ioctl_trans_end(file); 2918ac8e9819SChris Mason case BTRFS_IOC_TREE_SEARCH: 2919ac8e9819SChris Mason return btrfs_ioctl_tree_search(file, argp); 2920ac8e9819SChris Mason case BTRFS_IOC_INO_LOOKUP: 2921ac8e9819SChris Mason return btrfs_ioctl_ino_lookup(file, argp); 29221406e432SJosef Bacik case BTRFS_IOC_SPACE_INFO: 29231406e432SJosef Bacik return btrfs_ioctl_space_info(root, argp); 2924f46b5a66SChristoph Hellwig case BTRFS_IOC_SYNC: 2925f46b5a66SChristoph Hellwig btrfs_sync_fs(file->f_dentry->d_sb, 1); 2926f46b5a66SChristoph Hellwig return 0; 292746204592SSage Weil case BTRFS_IOC_START_SYNC: 292846204592SSage Weil return btrfs_ioctl_start_sync(file, argp); 292946204592SSage Weil case BTRFS_IOC_WAIT_SYNC: 293046204592SSage Weil return btrfs_ioctl_wait_sync(file, argp); 2931475f6387SJan Schmidt case BTRFS_IOC_SCRUB: 2932475f6387SJan Schmidt return btrfs_ioctl_scrub(root, argp); 2933475f6387SJan Schmidt case BTRFS_IOC_SCRUB_CANCEL: 2934475f6387SJan Schmidt return btrfs_ioctl_scrub_cancel(root, argp); 2935475f6387SJan Schmidt case BTRFS_IOC_SCRUB_PROGRESS: 2936475f6387SJan Schmidt return btrfs_ioctl_scrub_progress(root, argp); 2937f46b5a66SChristoph Hellwig } 2938f46b5a66SChristoph Hellwig 2939f46b5a66SChristoph Hellwig return -ENOTTY; 2940f46b5a66SChristoph Hellwig } 2941