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> 448ea05e3aSAlexander Block #include <linux/uuid.h> 4555e301fdSFilipe Brandenburger #include <linux/btrfs.h> 46416161dbSMark Fasheh #include <linux/uaccess.h> 474b4e25f2SChris Mason #include "compat.h" 48f46b5a66SChristoph Hellwig #include "ctree.h" 49f46b5a66SChristoph Hellwig #include "disk-io.h" 50f46b5a66SChristoph Hellwig #include "transaction.h" 51f46b5a66SChristoph Hellwig #include "btrfs_inode.h" 52f46b5a66SChristoph Hellwig #include "print-tree.h" 53f46b5a66SChristoph Hellwig #include "volumes.h" 54925baeddSChris Mason #include "locking.h" 55581bb050SLi Zefan #include "inode-map.h" 56d7728c96SJan Schmidt #include "backref.h" 57606686eeSJosef Bacik #include "rcu-string.h" 5831db9f7cSAlexander Block #include "send.h" 593f6bcfbdSStefan Behrens #include "dev-replace.h" 60f46b5a66SChristoph Hellwig 61416161dbSMark Fasheh static int btrfs_clone(struct inode *src, struct inode *inode, 62416161dbSMark Fasheh u64 off, u64 olen, u64 olen_aligned, u64 destoff); 63416161dbSMark Fasheh 646cbff00fSChristoph Hellwig /* Mask out flags that are inappropriate for the given type of inode. */ 656cbff00fSChristoph Hellwig static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags) 666cbff00fSChristoph Hellwig { 676cbff00fSChristoph Hellwig if (S_ISDIR(mode)) 686cbff00fSChristoph Hellwig return flags; 696cbff00fSChristoph Hellwig else if (S_ISREG(mode)) 706cbff00fSChristoph Hellwig return flags & ~FS_DIRSYNC_FL; 716cbff00fSChristoph Hellwig else 726cbff00fSChristoph Hellwig return flags & (FS_NODUMP_FL | FS_NOATIME_FL); 736cbff00fSChristoph Hellwig } 74f46b5a66SChristoph Hellwig 756cbff00fSChristoph Hellwig /* 766cbff00fSChristoph Hellwig * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl. 776cbff00fSChristoph Hellwig */ 786cbff00fSChristoph Hellwig static unsigned int btrfs_flags_to_ioctl(unsigned int flags) 796cbff00fSChristoph Hellwig { 806cbff00fSChristoph Hellwig unsigned int iflags = 0; 816cbff00fSChristoph Hellwig 826cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_SYNC) 836cbff00fSChristoph Hellwig iflags |= FS_SYNC_FL; 846cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_IMMUTABLE) 856cbff00fSChristoph Hellwig iflags |= FS_IMMUTABLE_FL; 866cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_APPEND) 876cbff00fSChristoph Hellwig iflags |= FS_APPEND_FL; 886cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NODUMP) 896cbff00fSChristoph Hellwig iflags |= FS_NODUMP_FL; 906cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NOATIME) 916cbff00fSChristoph Hellwig iflags |= FS_NOATIME_FL; 926cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_DIRSYNC) 936cbff00fSChristoph Hellwig iflags |= FS_DIRSYNC_FL; 94d0092bddSLi Zefan if (flags & BTRFS_INODE_NODATACOW) 95d0092bddSLi Zefan iflags |= FS_NOCOW_FL; 96d0092bddSLi Zefan 97d0092bddSLi Zefan if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS)) 98d0092bddSLi Zefan iflags |= FS_COMPR_FL; 99d0092bddSLi Zefan else if (flags & BTRFS_INODE_NOCOMPRESS) 100d0092bddSLi Zefan iflags |= FS_NOCOMP_FL; 1016cbff00fSChristoph Hellwig 1026cbff00fSChristoph Hellwig return iflags; 1036cbff00fSChristoph Hellwig } 1046cbff00fSChristoph Hellwig 1056cbff00fSChristoph Hellwig /* 1066cbff00fSChristoph Hellwig * Update inode->i_flags based on the btrfs internal flags. 1076cbff00fSChristoph Hellwig */ 1086cbff00fSChristoph Hellwig void btrfs_update_iflags(struct inode *inode) 1096cbff00fSChristoph Hellwig { 1106cbff00fSChristoph Hellwig struct btrfs_inode *ip = BTRFS_I(inode); 1116cbff00fSChristoph Hellwig 1126cbff00fSChristoph Hellwig inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC); 1136cbff00fSChristoph Hellwig 1146cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_SYNC) 1156cbff00fSChristoph Hellwig inode->i_flags |= S_SYNC; 1166cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_IMMUTABLE) 1176cbff00fSChristoph Hellwig inode->i_flags |= S_IMMUTABLE; 1186cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_APPEND) 1196cbff00fSChristoph Hellwig inode->i_flags |= S_APPEND; 1206cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_NOATIME) 1216cbff00fSChristoph Hellwig inode->i_flags |= S_NOATIME; 1226cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_DIRSYNC) 1236cbff00fSChristoph Hellwig inode->i_flags |= S_DIRSYNC; 1246cbff00fSChristoph Hellwig } 1256cbff00fSChristoph Hellwig 1266cbff00fSChristoph Hellwig /* 1276cbff00fSChristoph Hellwig * Inherit flags from the parent inode. 1286cbff00fSChristoph Hellwig * 129e27425d6SJosef Bacik * Currently only the compression flags and the cow flags are inherited. 1306cbff00fSChristoph Hellwig */ 1316cbff00fSChristoph Hellwig void btrfs_inherit_iflags(struct inode *inode, struct inode *dir) 1326cbff00fSChristoph Hellwig { 1330b4dcea5SChris Mason unsigned int flags; 1340b4dcea5SChris Mason 1350b4dcea5SChris Mason if (!dir) 1360b4dcea5SChris Mason return; 1370b4dcea5SChris Mason 1380b4dcea5SChris Mason flags = BTRFS_I(dir)->flags; 1396cbff00fSChristoph Hellwig 140e27425d6SJosef Bacik if (flags & BTRFS_INODE_NOCOMPRESS) { 141e27425d6SJosef Bacik BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS; 142e27425d6SJosef Bacik BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS; 143e27425d6SJosef Bacik } else if (flags & BTRFS_INODE_COMPRESS) { 144e27425d6SJosef Bacik BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS; 145e27425d6SJosef Bacik BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS; 146e27425d6SJosef Bacik } 1476cbff00fSChristoph Hellwig 148213490b3SLiu Bo if (flags & BTRFS_INODE_NODATACOW) { 149e27425d6SJosef Bacik BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW; 150213490b3SLiu Bo if (S_ISREG(inode->i_mode)) 151213490b3SLiu Bo BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM; 152213490b3SLiu Bo } 153e27425d6SJosef Bacik 1546cbff00fSChristoph Hellwig btrfs_update_iflags(inode); 1556cbff00fSChristoph Hellwig } 1566cbff00fSChristoph Hellwig 1576cbff00fSChristoph Hellwig static int btrfs_ioctl_getflags(struct file *file, void __user *arg) 1586cbff00fSChristoph Hellwig { 159496ad9aaSAl Viro struct btrfs_inode *ip = BTRFS_I(file_inode(file)); 1606cbff00fSChristoph Hellwig unsigned int flags = btrfs_flags_to_ioctl(ip->flags); 1616cbff00fSChristoph Hellwig 1626cbff00fSChristoph Hellwig if (copy_to_user(arg, &flags, sizeof(flags))) 1636cbff00fSChristoph Hellwig return -EFAULT; 1646cbff00fSChristoph Hellwig return 0; 1656cbff00fSChristoph Hellwig } 1666cbff00fSChristoph Hellwig 16775e7cb7fSLiu Bo static int check_flags(unsigned int flags) 16875e7cb7fSLiu Bo { 16975e7cb7fSLiu Bo if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \ 17075e7cb7fSLiu Bo FS_NOATIME_FL | FS_NODUMP_FL | \ 17175e7cb7fSLiu Bo FS_SYNC_FL | FS_DIRSYNC_FL | \ 172e1e8fb6aSLi Zefan FS_NOCOMP_FL | FS_COMPR_FL | 173e1e8fb6aSLi Zefan FS_NOCOW_FL)) 17475e7cb7fSLiu Bo return -EOPNOTSUPP; 17575e7cb7fSLiu Bo 17675e7cb7fSLiu Bo if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL)) 17775e7cb7fSLiu Bo return -EINVAL; 17875e7cb7fSLiu Bo 17975e7cb7fSLiu Bo return 0; 18075e7cb7fSLiu Bo } 18175e7cb7fSLiu Bo 1826cbff00fSChristoph Hellwig static int btrfs_ioctl_setflags(struct file *file, void __user *arg) 1836cbff00fSChristoph Hellwig { 184496ad9aaSAl Viro struct inode *inode = file_inode(file); 1856cbff00fSChristoph Hellwig struct btrfs_inode *ip = BTRFS_I(inode); 1866cbff00fSChristoph Hellwig struct btrfs_root *root = ip->root; 1876cbff00fSChristoph Hellwig struct btrfs_trans_handle *trans; 1886cbff00fSChristoph Hellwig unsigned int flags, oldflags; 1896cbff00fSChristoph Hellwig int ret; 190f062abf0SLi Zefan u64 ip_oldflags; 191f062abf0SLi Zefan unsigned int i_oldflags; 1927e97b8daSDavid Sterba umode_t mode; 1936cbff00fSChristoph Hellwig 194b83cc969SLi Zefan if (btrfs_root_readonly(root)) 195b83cc969SLi Zefan return -EROFS; 196b83cc969SLi Zefan 1976cbff00fSChristoph Hellwig if (copy_from_user(&flags, arg, sizeof(flags))) 1986cbff00fSChristoph Hellwig return -EFAULT; 1996cbff00fSChristoph Hellwig 20075e7cb7fSLiu Bo ret = check_flags(flags); 20175e7cb7fSLiu Bo if (ret) 20275e7cb7fSLiu Bo return ret; 2036cbff00fSChristoph Hellwig 2042e149670SSerge E. Hallyn if (!inode_owner_or_capable(inode)) 2056cbff00fSChristoph Hellwig return -EACCES; 2066cbff00fSChristoph Hellwig 207e7848683SJan Kara ret = mnt_want_write_file(file); 208e7848683SJan Kara if (ret) 209e7848683SJan Kara return ret; 210e7848683SJan Kara 2116cbff00fSChristoph Hellwig mutex_lock(&inode->i_mutex); 2126cbff00fSChristoph Hellwig 213f062abf0SLi Zefan ip_oldflags = ip->flags; 214f062abf0SLi Zefan i_oldflags = inode->i_flags; 2157e97b8daSDavid Sterba mode = inode->i_mode; 216f062abf0SLi Zefan 2176cbff00fSChristoph Hellwig flags = btrfs_mask_flags(inode->i_mode, flags); 2186cbff00fSChristoph Hellwig oldflags = btrfs_flags_to_ioctl(ip->flags); 2196cbff00fSChristoph Hellwig if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) { 2206cbff00fSChristoph Hellwig if (!capable(CAP_LINUX_IMMUTABLE)) { 2216cbff00fSChristoph Hellwig ret = -EPERM; 2226cbff00fSChristoph Hellwig goto out_unlock; 2236cbff00fSChristoph Hellwig } 2246cbff00fSChristoph Hellwig } 2256cbff00fSChristoph Hellwig 2266cbff00fSChristoph Hellwig if (flags & FS_SYNC_FL) 2276cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_SYNC; 2286cbff00fSChristoph Hellwig else 2296cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_SYNC; 2306cbff00fSChristoph Hellwig if (flags & FS_IMMUTABLE_FL) 2316cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_IMMUTABLE; 2326cbff00fSChristoph Hellwig else 2336cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_IMMUTABLE; 2346cbff00fSChristoph Hellwig if (flags & FS_APPEND_FL) 2356cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_APPEND; 2366cbff00fSChristoph Hellwig else 2376cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_APPEND; 2386cbff00fSChristoph Hellwig if (flags & FS_NODUMP_FL) 2396cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_NODUMP; 2406cbff00fSChristoph Hellwig else 2416cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_NODUMP; 2426cbff00fSChristoph Hellwig if (flags & FS_NOATIME_FL) 2436cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_NOATIME; 2446cbff00fSChristoph Hellwig else 2456cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_NOATIME; 2466cbff00fSChristoph Hellwig if (flags & FS_DIRSYNC_FL) 2476cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_DIRSYNC; 2486cbff00fSChristoph Hellwig else 2496cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_DIRSYNC; 2507e97b8daSDavid Sterba if (flags & FS_NOCOW_FL) { 2517e97b8daSDavid Sterba if (S_ISREG(mode)) { 2527e97b8daSDavid Sterba /* 2537e97b8daSDavid Sterba * It's safe to turn csums off here, no extents exist. 2547e97b8daSDavid Sterba * Otherwise we want the flag to reflect the real COW 2557e97b8daSDavid Sterba * status of the file and will not set it. 2567e97b8daSDavid Sterba */ 2577e97b8daSDavid Sterba if (inode->i_size == 0) 2587e97b8daSDavid Sterba ip->flags |= BTRFS_INODE_NODATACOW 2597e97b8daSDavid Sterba | BTRFS_INODE_NODATASUM; 2607e97b8daSDavid Sterba } else { 261e1e8fb6aSLi Zefan ip->flags |= BTRFS_INODE_NODATACOW; 2627e97b8daSDavid Sterba } 2637e97b8daSDavid Sterba } else { 2647e97b8daSDavid Sterba /* 2657e97b8daSDavid Sterba * Revert back under same assuptions as above 2667e97b8daSDavid Sterba */ 2677e97b8daSDavid Sterba if (S_ISREG(mode)) { 2687e97b8daSDavid Sterba if (inode->i_size == 0) 2697e97b8daSDavid Sterba ip->flags &= ~(BTRFS_INODE_NODATACOW 2707e97b8daSDavid Sterba | BTRFS_INODE_NODATASUM); 2717e97b8daSDavid Sterba } else { 272e1e8fb6aSLi Zefan ip->flags &= ~BTRFS_INODE_NODATACOW; 2737e97b8daSDavid Sterba } 2747e97b8daSDavid Sterba } 2756cbff00fSChristoph Hellwig 27675e7cb7fSLiu Bo /* 27775e7cb7fSLiu Bo * The COMPRESS flag can only be changed by users, while the NOCOMPRESS 27875e7cb7fSLiu Bo * flag may be changed automatically if compression code won't make 27975e7cb7fSLiu Bo * things smaller. 28075e7cb7fSLiu Bo */ 28175e7cb7fSLiu Bo if (flags & FS_NOCOMP_FL) { 28275e7cb7fSLiu Bo ip->flags &= ~BTRFS_INODE_COMPRESS; 28375e7cb7fSLiu Bo ip->flags |= BTRFS_INODE_NOCOMPRESS; 28475e7cb7fSLiu Bo } else if (flags & FS_COMPR_FL) { 28575e7cb7fSLiu Bo ip->flags |= BTRFS_INODE_COMPRESS; 28675e7cb7fSLiu Bo ip->flags &= ~BTRFS_INODE_NOCOMPRESS; 287ebcb904dSLi Zefan } else { 288ebcb904dSLi Zefan ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS); 28975e7cb7fSLiu Bo } 2906cbff00fSChristoph Hellwig 2914da6f1a3SLi Zefan trans = btrfs_start_transaction(root, 1); 292f062abf0SLi Zefan if (IS_ERR(trans)) { 293f062abf0SLi Zefan ret = PTR_ERR(trans); 294f062abf0SLi Zefan goto out_drop; 295f062abf0SLi Zefan } 2966cbff00fSChristoph Hellwig 297306424ccSLi Zefan btrfs_update_iflags(inode); 2980c4d2d95SJosef Bacik inode_inc_iversion(inode); 299306424ccSLi Zefan inode->i_ctime = CURRENT_TIME; 3006cbff00fSChristoph Hellwig ret = btrfs_update_inode(trans, root, inode); 3016cbff00fSChristoph Hellwig 3026cbff00fSChristoph Hellwig btrfs_end_transaction(trans, root); 303f062abf0SLi Zefan out_drop: 304f062abf0SLi Zefan if (ret) { 305f062abf0SLi Zefan ip->flags = ip_oldflags; 306f062abf0SLi Zefan inode->i_flags = i_oldflags; 307f062abf0SLi Zefan } 3086cbff00fSChristoph Hellwig 3096cbff00fSChristoph Hellwig out_unlock: 3106cbff00fSChristoph Hellwig mutex_unlock(&inode->i_mutex); 311e7848683SJan Kara mnt_drop_write_file(file); 3122d4e6f6aSliubo return ret; 3136cbff00fSChristoph Hellwig } 3146cbff00fSChristoph Hellwig 3156cbff00fSChristoph Hellwig static int btrfs_ioctl_getversion(struct file *file, int __user *arg) 3166cbff00fSChristoph Hellwig { 317496ad9aaSAl Viro struct inode *inode = file_inode(file); 3186cbff00fSChristoph Hellwig 3196cbff00fSChristoph Hellwig return put_user(inode->i_generation, arg); 3206cbff00fSChristoph Hellwig } 321f46b5a66SChristoph Hellwig 322f7039b1dSLi Dongyang static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg) 323f7039b1dSLi Dongyang { 324815745cfSAl Viro struct btrfs_fs_info *fs_info = btrfs_sb(fdentry(file)->d_sb); 325f7039b1dSLi Dongyang struct btrfs_device *device; 326f7039b1dSLi Dongyang struct request_queue *q; 327f7039b1dSLi Dongyang struct fstrim_range range; 328f7039b1dSLi Dongyang u64 minlen = ULLONG_MAX; 329f7039b1dSLi Dongyang u64 num_devices = 0; 330815745cfSAl Viro u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy); 331f7039b1dSLi Dongyang int ret; 332f7039b1dSLi Dongyang 333f7039b1dSLi Dongyang if (!capable(CAP_SYS_ADMIN)) 334f7039b1dSLi Dongyang return -EPERM; 335f7039b1dSLi Dongyang 3361f78160cSXiao Guangrong rcu_read_lock(); 3371f78160cSXiao Guangrong list_for_each_entry_rcu(device, &fs_info->fs_devices->devices, 3381f78160cSXiao Guangrong dev_list) { 339f7039b1dSLi Dongyang if (!device->bdev) 340f7039b1dSLi Dongyang continue; 341f7039b1dSLi Dongyang q = bdev_get_queue(device->bdev); 342f7039b1dSLi Dongyang if (blk_queue_discard(q)) { 343f7039b1dSLi Dongyang num_devices++; 344f7039b1dSLi Dongyang minlen = min((u64)q->limits.discard_granularity, 345f7039b1dSLi Dongyang minlen); 346f7039b1dSLi Dongyang } 347f7039b1dSLi Dongyang } 3481f78160cSXiao Guangrong rcu_read_unlock(); 349f4c697e6SLukas Czerner 350f7039b1dSLi Dongyang if (!num_devices) 351f7039b1dSLi Dongyang return -EOPNOTSUPP; 352f7039b1dSLi Dongyang if (copy_from_user(&range, arg, sizeof(range))) 353f7039b1dSLi Dongyang return -EFAULT; 354e515c18bSLukas Czerner if (range.start > total_bytes || 355e515c18bSLukas Czerner range.len < fs_info->sb->s_blocksize) 356f4c697e6SLukas Czerner return -EINVAL; 357f7039b1dSLi Dongyang 358f4c697e6SLukas Czerner range.len = min(range.len, total_bytes - range.start); 359f7039b1dSLi Dongyang range.minlen = max(range.minlen, minlen); 360815745cfSAl Viro ret = btrfs_trim_fs(fs_info->tree_root, &range); 361f7039b1dSLi Dongyang if (ret < 0) 362f7039b1dSLi Dongyang return ret; 363f7039b1dSLi Dongyang 364f7039b1dSLi Dongyang if (copy_to_user(arg, &range, sizeof(range))) 365f7039b1dSLi Dongyang return -EFAULT; 366f7039b1dSLi Dongyang 367f7039b1dSLi Dongyang return 0; 368f7039b1dSLi Dongyang } 369f7039b1dSLi Dongyang 370d5c12070SMiao Xie static noinline int create_subvol(struct inode *dir, 371cb8e7090SChristoph Hellwig struct dentry *dentry, 37272fd032eSSage Weil char *name, int namelen, 3736f72c7e2SArne Jansen u64 *async_transid, 3748696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 375f46b5a66SChristoph Hellwig { 376f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 377f46b5a66SChristoph Hellwig struct btrfs_key key; 378f46b5a66SChristoph Hellwig struct btrfs_root_item root_item; 379f46b5a66SChristoph Hellwig struct btrfs_inode_item *inode_item; 380f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 381d5c12070SMiao Xie struct btrfs_root *root = BTRFS_I(dir)->root; 38276dda93cSYan, Zheng struct btrfs_root *new_root; 383d5c12070SMiao Xie struct btrfs_block_rsv block_rsv; 3848ea05e3aSAlexander Block struct timespec cur_time = CURRENT_TIME; 385f46b5a66SChristoph Hellwig int ret; 386f46b5a66SChristoph Hellwig int err; 387f46b5a66SChristoph Hellwig u64 objectid; 388f46b5a66SChristoph Hellwig u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; 3893de4586cSChris Mason u64 index = 0; 390d5c12070SMiao Xie u64 qgroup_reserved; 3918ea05e3aSAlexander Block uuid_le new_uuid; 392f46b5a66SChristoph Hellwig 393581bb050SLi Zefan ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid); 3942fbe8c8aSAl Viro if (ret) 395a22285a6SYan, Zheng return ret; 3966a912213SJosef Bacik 397d5c12070SMiao Xie btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); 3989ed74f2dSJosef Bacik /* 399d5c12070SMiao Xie * The same as the snapshot creation, please see the comment 400d5c12070SMiao Xie * of create_snapshot(). 4019ed74f2dSJosef Bacik */ 402d5c12070SMiao Xie ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 403ee3441b4SJeff Mahoney 7, &qgroup_reserved, false); 404d5c12070SMiao Xie if (ret) 405d5c12070SMiao Xie return ret; 406f46b5a66SChristoph Hellwig 407d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 408d5c12070SMiao Xie if (IS_ERR(trans)) { 409d5c12070SMiao Xie ret = PTR_ERR(trans); 410d5c12070SMiao Xie goto out; 411d5c12070SMiao Xie } 412d5c12070SMiao Xie trans->block_rsv = &block_rsv; 413d5c12070SMiao Xie trans->bytes_reserved = block_rsv.size; 414f46b5a66SChristoph Hellwig 4158696c533SMiao Xie ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit); 4166f72c7e2SArne Jansen if (ret) 4176f72c7e2SArne Jansen goto fail; 4186f72c7e2SArne Jansen 4195d4f98a2SYan Zheng leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 4205581a51aSJan Schmidt 0, objectid, NULL, 0, 0, 0); 4218e8a1e31SJosef Bacik if (IS_ERR(leaf)) { 4228e8a1e31SJosef Bacik ret = PTR_ERR(leaf); 4238e8a1e31SJosef Bacik goto fail; 4248e8a1e31SJosef Bacik } 425f46b5a66SChristoph Hellwig 4265d4f98a2SYan Zheng memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); 427f46b5a66SChristoph Hellwig btrfs_set_header_bytenr(leaf, leaf->start); 428f46b5a66SChristoph Hellwig btrfs_set_header_generation(leaf, trans->transid); 4295d4f98a2SYan Zheng btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 430f46b5a66SChristoph Hellwig btrfs_set_header_owner(leaf, objectid); 431f46b5a66SChristoph Hellwig 432f46b5a66SChristoph Hellwig write_extent_buffer(leaf, root->fs_info->fsid, 433f46b5a66SChristoph Hellwig (unsigned long)btrfs_header_fsid(leaf), 434f46b5a66SChristoph Hellwig BTRFS_FSID_SIZE); 4355d4f98a2SYan Zheng write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid, 4365d4f98a2SYan Zheng (unsigned long)btrfs_header_chunk_tree_uuid(leaf), 4375d4f98a2SYan Zheng BTRFS_UUID_SIZE); 438f46b5a66SChristoph Hellwig btrfs_mark_buffer_dirty(leaf); 439f46b5a66SChristoph Hellwig 4408ea05e3aSAlexander Block memset(&root_item, 0, sizeof(root_item)); 4418ea05e3aSAlexander Block 442f46b5a66SChristoph Hellwig inode_item = &root_item.inode; 4433cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 4443cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 4453cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 4463cae210fSQu Wenruo btrfs_set_stack_inode_nbytes(inode_item, root->leafsize); 4473cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 448f46b5a66SChristoph Hellwig 4493cae210fSQu Wenruo btrfs_set_root_flags(&root_item, 0); 4503cae210fSQu Wenruo btrfs_set_root_limit(&root_item, 0); 4513cae210fSQu Wenruo btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT); 45208fe4db1SLi Zefan 453f46b5a66SChristoph Hellwig btrfs_set_root_bytenr(&root_item, leaf->start); 45484234f3aSYan Zheng btrfs_set_root_generation(&root_item, trans->transid); 455f46b5a66SChristoph Hellwig btrfs_set_root_level(&root_item, 0); 456f46b5a66SChristoph Hellwig btrfs_set_root_refs(&root_item, 1); 45786b9f2ecSYan, Zheng btrfs_set_root_used(&root_item, leaf->len); 45880ff3856SYan Zheng btrfs_set_root_last_snapshot(&root_item, 0); 459f46b5a66SChristoph Hellwig 4608ea05e3aSAlexander Block btrfs_set_root_generation_v2(&root_item, 4618ea05e3aSAlexander Block btrfs_root_generation(&root_item)); 4628ea05e3aSAlexander Block uuid_le_gen(&new_uuid); 4638ea05e3aSAlexander Block memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE); 4643cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec); 4653cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec); 4668ea05e3aSAlexander Block root_item.ctime = root_item.otime; 4678ea05e3aSAlexander Block btrfs_set_root_ctransid(&root_item, trans->transid); 4688ea05e3aSAlexander Block btrfs_set_root_otransid(&root_item, trans->transid); 469f46b5a66SChristoph Hellwig 470925baeddSChris Mason btrfs_tree_unlock(leaf); 471f46b5a66SChristoph Hellwig free_extent_buffer(leaf); 472f46b5a66SChristoph Hellwig leaf = NULL; 473f46b5a66SChristoph Hellwig 474f46b5a66SChristoph Hellwig btrfs_set_root_dirid(&root_item, new_dirid); 475f46b5a66SChristoph Hellwig 476f46b5a66SChristoph Hellwig key.objectid = objectid; 4775d4f98a2SYan Zheng key.offset = 0; 478f46b5a66SChristoph Hellwig btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); 479f46b5a66SChristoph Hellwig ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, 480f46b5a66SChristoph Hellwig &root_item); 481f46b5a66SChristoph Hellwig if (ret) 482f46b5a66SChristoph Hellwig goto fail; 483f46b5a66SChristoph Hellwig 48476dda93cSYan, Zheng key.offset = (u64)-1; 48576dda93cSYan, Zheng new_root = btrfs_read_fs_root_no_name(root->fs_info, &key); 48679787eaaSJeff Mahoney if (IS_ERR(new_root)) { 48779787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, PTR_ERR(new_root)); 48879787eaaSJeff Mahoney ret = PTR_ERR(new_root); 48979787eaaSJeff Mahoney goto fail; 49079787eaaSJeff Mahoney } 49176dda93cSYan, Zheng 49276dda93cSYan, Zheng btrfs_record_root_in_trans(trans, new_root); 49376dda93cSYan, Zheng 494d82a6f1dSJosef Bacik ret = btrfs_create_subvol_root(trans, new_root, new_dirid); 495ce598979SMark Fasheh if (ret) { 496ce598979SMark Fasheh /* We potentially lose an unused inode item here */ 49779787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, ret); 498ce598979SMark Fasheh goto fail; 499ce598979SMark Fasheh } 500ce598979SMark Fasheh 501f46b5a66SChristoph Hellwig /* 502f46b5a66SChristoph Hellwig * insert the directory item 503f46b5a66SChristoph Hellwig */ 5043de4586cSChris Mason ret = btrfs_set_inode_index(dir, &index); 50579787eaaSJeff Mahoney if (ret) { 50679787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, ret); 50779787eaaSJeff Mahoney goto fail; 50879787eaaSJeff Mahoney } 5093de4586cSChris Mason 5103de4586cSChris Mason ret = btrfs_insert_dir_item(trans, root, 51116cdcec7SMiao Xie name, namelen, dir, &key, 5123de4586cSChris Mason BTRFS_FT_DIR, index); 51379787eaaSJeff Mahoney if (ret) { 51479787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, ret); 515f46b5a66SChristoph Hellwig goto fail; 51679787eaaSJeff Mahoney } 5170660b5afSChris Mason 51852c26179SYan Zheng btrfs_i_size_write(dir, dir->i_size + namelen * 2); 51952c26179SYan Zheng ret = btrfs_update_inode(trans, root, dir); 52052c26179SYan Zheng BUG_ON(ret); 52152c26179SYan Zheng 5220660b5afSChris Mason ret = btrfs_add_root_ref(trans, root->fs_info->tree_root, 5234df27c4dSYan, Zheng objectid, root->root_key.objectid, 52433345d01SLi Zefan btrfs_ino(dir), index, name, namelen); 52576dda93cSYan, Zheng 5260660b5afSChris Mason BUG_ON(ret); 5270660b5afSChris Mason 528f46b5a66SChristoph Hellwig fail: 529d5c12070SMiao Xie trans->block_rsv = NULL; 530d5c12070SMiao Xie trans->bytes_reserved = 0; 53172fd032eSSage Weil if (async_transid) { 53272fd032eSSage Weil *async_transid = trans->transid; 53372fd032eSSage Weil err = btrfs_commit_transaction_async(trans, root, 1); 53400d71c9cSMiao Xie if (err) 53500d71c9cSMiao Xie err = btrfs_commit_transaction(trans, root); 53672fd032eSSage Weil } else { 53776dda93cSYan, Zheng err = btrfs_commit_transaction(trans, root); 53872fd032eSSage Weil } 539f46b5a66SChristoph Hellwig if (err && !ret) 540f46b5a66SChristoph Hellwig ret = err; 5411a65e24bSChris Mason 5421a65e24bSChris Mason if (!ret) 5431a65e24bSChris Mason d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry)); 544d5c12070SMiao Xie out: 545d5c12070SMiao Xie btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved); 546f46b5a66SChristoph Hellwig return ret; 547f46b5a66SChristoph Hellwig } 548f46b5a66SChristoph Hellwig 549e9662f70SMiao Xie static int create_snapshot(struct btrfs_root *root, struct inode *dir, 550e9662f70SMiao Xie struct dentry *dentry, char *name, int namelen, 551e9662f70SMiao Xie u64 *async_transid, bool readonly, 552e9662f70SMiao Xie struct btrfs_qgroup_inherit *inherit) 553f46b5a66SChristoph Hellwig { 5542e4bfab9SYan, Zheng struct inode *inode; 555f46b5a66SChristoph Hellwig struct btrfs_pending_snapshot *pending_snapshot; 556f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 5572e4bfab9SYan, Zheng int ret; 558f46b5a66SChristoph Hellwig 559f46b5a66SChristoph Hellwig if (!root->ref_cows) 560f46b5a66SChristoph Hellwig return -EINVAL; 561f46b5a66SChristoph Hellwig 5626a03843dSMiao Xie ret = btrfs_start_delalloc_inodes(root, 0); 5636a03843dSMiao Xie if (ret) 5646a03843dSMiao Xie return ret; 5656a03843dSMiao Xie 5666a03843dSMiao Xie btrfs_wait_ordered_extents(root, 0); 5676a03843dSMiao Xie 568a22285a6SYan, Zheng pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS); 569a22285a6SYan, Zheng if (!pending_snapshot) 570a22285a6SYan, Zheng return -ENOMEM; 571a22285a6SYan, Zheng 57266d8f3ddSMiao Xie btrfs_init_block_rsv(&pending_snapshot->block_rsv, 57366d8f3ddSMiao Xie BTRFS_BLOCK_RSV_TEMP); 574d5c12070SMiao Xie /* 575d5c12070SMiao Xie * 1 - parent dir inode 576d5c12070SMiao Xie * 2 - dir entries 577d5c12070SMiao Xie * 1 - root item 578d5c12070SMiao Xie * 2 - root ref/backref 579d5c12070SMiao Xie * 1 - root of snapshot 580d5c12070SMiao Xie */ 581d5c12070SMiao Xie ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, 582d5c12070SMiao Xie &pending_snapshot->block_rsv, 7, 583ee3441b4SJeff Mahoney &pending_snapshot->qgroup_reserved, 584ee3441b4SJeff Mahoney false); 585d5c12070SMiao Xie if (ret) 586d5c12070SMiao Xie goto out; 587d5c12070SMiao Xie 588a22285a6SYan, Zheng pending_snapshot->dentry = dentry; 589a22285a6SYan, Zheng pending_snapshot->root = root; 590b83cc969SLi Zefan pending_snapshot->readonly = readonly; 591e9662f70SMiao Xie pending_snapshot->dir = dir; 5928696c533SMiao Xie pending_snapshot->inherit = inherit; 593a22285a6SYan, Zheng 594d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 595a22285a6SYan, Zheng if (IS_ERR(trans)) { 596a22285a6SYan, Zheng ret = PTR_ERR(trans); 597a22285a6SYan, Zheng goto fail; 598a22285a6SYan, Zheng } 599a22285a6SYan, Zheng 6008351583eSJosef Bacik spin_lock(&root->fs_info->trans_lock); 601a22285a6SYan, Zheng list_add(&pending_snapshot->list, 602a22285a6SYan, Zheng &trans->transaction->pending_snapshots); 6038351583eSJosef Bacik spin_unlock(&root->fs_info->trans_lock); 60472fd032eSSage Weil if (async_transid) { 60572fd032eSSage Weil *async_transid = trans->transid; 60672fd032eSSage Weil ret = btrfs_commit_transaction_async(trans, 60772fd032eSSage Weil root->fs_info->extent_root, 1); 60800d71c9cSMiao Xie if (ret) 60900d71c9cSMiao Xie ret = btrfs_commit_transaction(trans, root); 61072fd032eSSage Weil } else { 61172fd032eSSage Weil ret = btrfs_commit_transaction(trans, 61272fd032eSSage Weil root->fs_info->extent_root); 61372fd032eSSage Weil } 614aec8030aSMiao Xie if (ret) 615c37b2b62SJosef Bacik goto fail; 616a22285a6SYan, Zheng 617a22285a6SYan, Zheng ret = pending_snapshot->error; 618f46b5a66SChristoph Hellwig if (ret) 6192e4bfab9SYan, Zheng goto fail; 620f46b5a66SChristoph Hellwig 62166b4ffd1SJosef Bacik ret = btrfs_orphan_cleanup(pending_snapshot->snap); 62266b4ffd1SJosef Bacik if (ret) 62366b4ffd1SJosef Bacik goto fail; 624f46b5a66SChristoph Hellwig 6252fbe8c8aSAl Viro inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry); 6262e4bfab9SYan, Zheng if (IS_ERR(inode)) { 6272e4bfab9SYan, Zheng ret = PTR_ERR(inode); 6282e4bfab9SYan, Zheng goto fail; 6292e4bfab9SYan, Zheng } 6302e4bfab9SYan, Zheng BUG_ON(!inode); 6312e4bfab9SYan, Zheng d_instantiate(dentry, inode); 6322e4bfab9SYan, Zheng ret = 0; 6332e4bfab9SYan, Zheng fail: 634d5c12070SMiao Xie btrfs_subvolume_release_metadata(BTRFS_I(dir)->root, 635d5c12070SMiao Xie &pending_snapshot->block_rsv, 636d5c12070SMiao Xie pending_snapshot->qgroup_reserved); 637d5c12070SMiao Xie out: 638a22285a6SYan, Zheng kfree(pending_snapshot); 639f46b5a66SChristoph Hellwig return ret; 640f46b5a66SChristoph Hellwig } 641f46b5a66SChristoph Hellwig 6424260f7c7SSage Weil /* copy of check_sticky in fs/namei.c() 6434260f7c7SSage Weil * It's inline, so penalty for filesystems that don't use sticky bit is 6444260f7c7SSage Weil * minimal. 6454260f7c7SSage Weil */ 6464260f7c7SSage Weil static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode) 6474260f7c7SSage Weil { 6482f2f43d3SEric W. Biederman kuid_t fsuid = current_fsuid(); 6494260f7c7SSage Weil 6504260f7c7SSage Weil if (!(dir->i_mode & S_ISVTX)) 6514260f7c7SSage Weil return 0; 6522f2f43d3SEric W. Biederman if (uid_eq(inode->i_uid, fsuid)) 6534260f7c7SSage Weil return 0; 6542f2f43d3SEric W. Biederman if (uid_eq(dir->i_uid, fsuid)) 6554260f7c7SSage Weil return 0; 6564260f7c7SSage Weil return !capable(CAP_FOWNER); 6574260f7c7SSage Weil } 6584260f7c7SSage Weil 6594260f7c7SSage Weil /* copy of may_delete in fs/namei.c() 6604260f7c7SSage Weil * Check whether we can remove a link victim from directory dir, check 6614260f7c7SSage Weil * whether the type of victim is right. 6624260f7c7SSage Weil * 1. We can't do it if dir is read-only (done in permission()) 6634260f7c7SSage Weil * 2. We should have write and exec permissions on dir 6644260f7c7SSage Weil * 3. We can't remove anything from append-only dir 6654260f7c7SSage Weil * 4. We can't do anything with immutable dir (done in permission()) 6664260f7c7SSage Weil * 5. If the sticky bit on dir is set we should either 6674260f7c7SSage Weil * a. be owner of dir, or 6684260f7c7SSage Weil * b. be owner of victim, or 6694260f7c7SSage Weil * c. have CAP_FOWNER capability 6704260f7c7SSage Weil * 6. If the victim is append-only or immutable we can't do antyhing with 6714260f7c7SSage Weil * links pointing to it. 6724260f7c7SSage Weil * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. 6734260f7c7SSage Weil * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. 6744260f7c7SSage Weil * 9. We can't remove a root or mountpoint. 6754260f7c7SSage Weil * 10. We don't allow removal of NFS sillyrenamed files; it's handled by 6764260f7c7SSage Weil * nfs_async_unlink(). 6774260f7c7SSage Weil */ 6784260f7c7SSage Weil 6794260f7c7SSage Weil static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir) 6804260f7c7SSage Weil { 6814260f7c7SSage Weil int error; 6824260f7c7SSage Weil 6834260f7c7SSage Weil if (!victim->d_inode) 6844260f7c7SSage Weil return -ENOENT; 6854260f7c7SSage Weil 6864260f7c7SSage Weil BUG_ON(victim->d_parent->d_inode != dir); 6874fa6b5ecSJeff Layton audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); 6884260f7c7SSage Weil 6894260f7c7SSage Weil error = inode_permission(dir, MAY_WRITE | MAY_EXEC); 6904260f7c7SSage Weil if (error) 6914260f7c7SSage Weil return error; 6924260f7c7SSage Weil if (IS_APPEND(dir)) 6934260f7c7SSage Weil return -EPERM; 6944260f7c7SSage Weil if (btrfs_check_sticky(dir, victim->d_inode)|| 6954260f7c7SSage Weil IS_APPEND(victim->d_inode)|| 6964260f7c7SSage Weil IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode)) 6974260f7c7SSage Weil return -EPERM; 6984260f7c7SSage Weil if (isdir) { 6994260f7c7SSage Weil if (!S_ISDIR(victim->d_inode->i_mode)) 7004260f7c7SSage Weil return -ENOTDIR; 7014260f7c7SSage Weil if (IS_ROOT(victim)) 7024260f7c7SSage Weil return -EBUSY; 7034260f7c7SSage Weil } else if (S_ISDIR(victim->d_inode->i_mode)) 7044260f7c7SSage Weil return -EISDIR; 7054260f7c7SSage Weil if (IS_DEADDIR(dir)) 7064260f7c7SSage Weil return -ENOENT; 7074260f7c7SSage Weil if (victim->d_flags & DCACHE_NFSFS_RENAMED) 7084260f7c7SSage Weil return -EBUSY; 7094260f7c7SSage Weil return 0; 7104260f7c7SSage Weil } 7114260f7c7SSage Weil 712cb8e7090SChristoph Hellwig /* copy of may_create in fs/namei.c() */ 713cb8e7090SChristoph Hellwig static inline int btrfs_may_create(struct inode *dir, struct dentry *child) 714cb8e7090SChristoph Hellwig { 715cb8e7090SChristoph Hellwig if (child->d_inode) 716cb8e7090SChristoph Hellwig return -EEXIST; 717cb8e7090SChristoph Hellwig if (IS_DEADDIR(dir)) 718cb8e7090SChristoph Hellwig return -ENOENT; 719cb8e7090SChristoph Hellwig return inode_permission(dir, MAY_WRITE | MAY_EXEC); 720cb8e7090SChristoph Hellwig } 721cb8e7090SChristoph Hellwig 722cb8e7090SChristoph Hellwig /* 723cb8e7090SChristoph Hellwig * Create a new subvolume below @parent. This is largely modeled after 724cb8e7090SChristoph Hellwig * sys_mkdirat and vfs_mkdir, but we only do a single component lookup 725cb8e7090SChristoph Hellwig * inside this filesystem so it's quite a bit simpler. 726cb8e7090SChristoph Hellwig */ 72776dda93cSYan, Zheng static noinline int btrfs_mksubvol(struct path *parent, 72876dda93cSYan, Zheng char *name, int namelen, 72972fd032eSSage Weil struct btrfs_root *snap_src, 7306f72c7e2SArne Jansen u64 *async_transid, bool readonly, 7318696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 732cb8e7090SChristoph Hellwig { 73376dda93cSYan, Zheng struct inode *dir = parent->dentry->d_inode; 734cb8e7090SChristoph Hellwig struct dentry *dentry; 735cb8e7090SChristoph Hellwig int error; 736cb8e7090SChristoph Hellwig 7375c50c9b8SDavid Sterba error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT); 7385c50c9b8SDavid Sterba if (error == -EINTR) 7395c50c9b8SDavid Sterba return error; 740cb8e7090SChristoph Hellwig 741cb8e7090SChristoph Hellwig dentry = lookup_one_len(name, parent->dentry, namelen); 742cb8e7090SChristoph Hellwig error = PTR_ERR(dentry); 743cb8e7090SChristoph Hellwig if (IS_ERR(dentry)) 744cb8e7090SChristoph Hellwig goto out_unlock; 745cb8e7090SChristoph Hellwig 746cb8e7090SChristoph Hellwig error = -EEXIST; 747cb8e7090SChristoph Hellwig if (dentry->d_inode) 748cb8e7090SChristoph Hellwig goto out_dput; 749cb8e7090SChristoph Hellwig 75076dda93cSYan, Zheng error = btrfs_may_create(dir, dentry); 751cb8e7090SChristoph Hellwig if (error) 752a874a63eSLiu Bo goto out_dput; 753cb8e7090SChristoph Hellwig 7549c52057cSChris Mason /* 7559c52057cSChris Mason * even if this name doesn't exist, we may get hash collisions. 7569c52057cSChris Mason * check for them now when we can safely fail 7579c52057cSChris Mason */ 7589c52057cSChris Mason error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, 7599c52057cSChris Mason dir->i_ino, name, 7609c52057cSChris Mason namelen); 7619c52057cSChris Mason if (error) 7629c52057cSChris Mason goto out_dput; 7639c52057cSChris Mason 76476dda93cSYan, Zheng down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem); 76576dda93cSYan, Zheng 76676dda93cSYan, Zheng if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) 76776dda93cSYan, Zheng goto out_up_read; 76876dda93cSYan, Zheng 7693de4586cSChris Mason if (snap_src) { 770e9662f70SMiao Xie error = create_snapshot(snap_src, dir, dentry, name, namelen, 7716f72c7e2SArne Jansen async_transid, readonly, inherit); 7723de4586cSChris Mason } else { 773d5c12070SMiao Xie error = create_subvol(dir, dentry, name, namelen, 774d5c12070SMiao Xie async_transid, inherit); 7753de4586cSChris Mason } 77676dda93cSYan, Zheng if (!error) 77776dda93cSYan, Zheng fsnotify_mkdir(dir, dentry); 77876dda93cSYan, Zheng out_up_read: 77976dda93cSYan, Zheng up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem); 780cb8e7090SChristoph Hellwig out_dput: 781cb8e7090SChristoph Hellwig dput(dentry); 782cb8e7090SChristoph Hellwig out_unlock: 78376dda93cSYan, Zheng mutex_unlock(&dir->i_mutex); 784cb8e7090SChristoph Hellwig return error; 785cb8e7090SChristoph Hellwig } 786cb8e7090SChristoph Hellwig 7874cb5300bSChris Mason /* 7884cb5300bSChris Mason * When we're defragging a range, we don't want to kick it off again 7894cb5300bSChris Mason * if it is really just waiting for delalloc to send it down. 7904cb5300bSChris Mason * If we find a nice big extent or delalloc range for the bytes in the 7914cb5300bSChris Mason * file you want to defrag, we return 0 to let you know to skip this 7924cb5300bSChris Mason * part of the file 7934cb5300bSChris Mason */ 7944cb5300bSChris Mason static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh) 7954cb5300bSChris Mason { 7964cb5300bSChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 7974cb5300bSChris Mason struct extent_map *em = NULL; 7984cb5300bSChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 7994cb5300bSChris Mason u64 end; 8004cb5300bSChris Mason 8014cb5300bSChris Mason read_lock(&em_tree->lock); 8024cb5300bSChris Mason em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE); 8034cb5300bSChris Mason read_unlock(&em_tree->lock); 8044cb5300bSChris Mason 8054cb5300bSChris Mason if (em) { 8064cb5300bSChris Mason end = extent_map_end(em); 8074cb5300bSChris Mason free_extent_map(em); 8084cb5300bSChris Mason if (end - offset > thresh) 8094cb5300bSChris Mason return 0; 8104cb5300bSChris Mason } 8114cb5300bSChris Mason /* if we already have a nice delalloc here, just stop */ 8124cb5300bSChris Mason thresh /= 2; 8134cb5300bSChris Mason end = count_range_bits(io_tree, &offset, offset + thresh, 8144cb5300bSChris Mason thresh, EXTENT_DELALLOC, 1); 8154cb5300bSChris Mason if (end >= thresh) 8164cb5300bSChris Mason return 0; 8174cb5300bSChris Mason return 1; 8184cb5300bSChris Mason } 8194cb5300bSChris Mason 8204cb5300bSChris Mason /* 8214cb5300bSChris Mason * helper function to walk through a file and find extents 8224cb5300bSChris Mason * newer than a specific transid, and smaller than thresh. 8234cb5300bSChris Mason * 8244cb5300bSChris Mason * This is used by the defragging code to find new and small 8254cb5300bSChris Mason * extents 8264cb5300bSChris Mason */ 8274cb5300bSChris Mason static int find_new_extents(struct btrfs_root *root, 8284cb5300bSChris Mason struct inode *inode, u64 newer_than, 8294cb5300bSChris Mason u64 *off, int thresh) 8304cb5300bSChris Mason { 8314cb5300bSChris Mason struct btrfs_path *path; 8324cb5300bSChris Mason struct btrfs_key min_key; 8334cb5300bSChris Mason struct btrfs_key max_key; 8344cb5300bSChris Mason struct extent_buffer *leaf; 8354cb5300bSChris Mason struct btrfs_file_extent_item *extent; 8364cb5300bSChris Mason int type; 8374cb5300bSChris Mason int ret; 838a4689d2bSDavid Sterba u64 ino = btrfs_ino(inode); 8394cb5300bSChris Mason 8404cb5300bSChris Mason path = btrfs_alloc_path(); 8414cb5300bSChris Mason if (!path) 8424cb5300bSChris Mason return -ENOMEM; 8434cb5300bSChris Mason 844a4689d2bSDavid Sterba min_key.objectid = ino; 8454cb5300bSChris Mason min_key.type = BTRFS_EXTENT_DATA_KEY; 8464cb5300bSChris Mason min_key.offset = *off; 8474cb5300bSChris Mason 848a4689d2bSDavid Sterba max_key.objectid = ino; 8494cb5300bSChris Mason max_key.type = (u8)-1; 8504cb5300bSChris Mason max_key.offset = (u64)-1; 8514cb5300bSChris Mason 8524cb5300bSChris Mason path->keep_locks = 1; 8534cb5300bSChris Mason 8544cb5300bSChris Mason while(1) { 8554cb5300bSChris Mason ret = btrfs_search_forward(root, &min_key, &max_key, 856de78b51aSEric Sandeen path, newer_than); 8574cb5300bSChris Mason if (ret != 0) 8584cb5300bSChris Mason goto none; 859a4689d2bSDavid Sterba if (min_key.objectid != ino) 8604cb5300bSChris Mason goto none; 8614cb5300bSChris Mason if (min_key.type != BTRFS_EXTENT_DATA_KEY) 8624cb5300bSChris Mason goto none; 8634cb5300bSChris Mason 8644cb5300bSChris Mason leaf = path->nodes[0]; 8654cb5300bSChris Mason extent = btrfs_item_ptr(leaf, path->slots[0], 8664cb5300bSChris Mason struct btrfs_file_extent_item); 8674cb5300bSChris Mason 8684cb5300bSChris Mason type = btrfs_file_extent_type(leaf, extent); 8694cb5300bSChris Mason if (type == BTRFS_FILE_EXTENT_REG && 8704cb5300bSChris Mason btrfs_file_extent_num_bytes(leaf, extent) < thresh && 8714cb5300bSChris Mason check_defrag_in_cache(inode, min_key.offset, thresh)) { 8724cb5300bSChris Mason *off = min_key.offset; 8734cb5300bSChris Mason btrfs_free_path(path); 8744cb5300bSChris Mason return 0; 8754cb5300bSChris Mason } 8764cb5300bSChris Mason 8774cb5300bSChris Mason if (min_key.offset == (u64)-1) 8784cb5300bSChris Mason goto none; 8794cb5300bSChris Mason 8804cb5300bSChris Mason min_key.offset++; 8814cb5300bSChris Mason btrfs_release_path(path); 8824cb5300bSChris Mason } 8834cb5300bSChris Mason none: 8844cb5300bSChris Mason btrfs_free_path(path); 8854cb5300bSChris Mason return -ENOENT; 8864cb5300bSChris Mason } 8874cb5300bSChris Mason 8886c282eb4SLi Zefan static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start) 88917ce6ef8SLiu Bo { 89017ce6ef8SLiu Bo struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 891940100a4SChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 8926c282eb4SLi Zefan struct extent_map *em; 8936c282eb4SLi Zefan u64 len = PAGE_CACHE_SIZE; 894940100a4SChris Mason 895940100a4SChris Mason /* 896940100a4SChris Mason * hopefully we have this extent in the tree already, try without 897940100a4SChris Mason * the full extent lock 898940100a4SChris Mason */ 899940100a4SChris Mason read_lock(&em_tree->lock); 900940100a4SChris Mason em = lookup_extent_mapping(em_tree, start, len); 901940100a4SChris Mason read_unlock(&em_tree->lock); 902940100a4SChris Mason 903940100a4SChris Mason if (!em) { 904940100a4SChris Mason /* get the big lock and read metadata off disk */ 905d0082371SJeff Mahoney lock_extent(io_tree, start, start + len - 1); 906940100a4SChris Mason em = btrfs_get_extent(inode, NULL, 0, start, len, 0); 907d0082371SJeff Mahoney unlock_extent(io_tree, start, start + len - 1); 908940100a4SChris Mason 9096cf8bfbfSDan Carpenter if (IS_ERR(em)) 9106c282eb4SLi Zefan return NULL; 911940100a4SChris Mason } 912940100a4SChris Mason 9136c282eb4SLi Zefan return em; 9146c282eb4SLi Zefan } 9156c282eb4SLi Zefan 9166c282eb4SLi Zefan static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em) 9176c282eb4SLi Zefan { 9186c282eb4SLi Zefan struct extent_map *next; 9196c282eb4SLi Zefan bool ret = true; 9206c282eb4SLi Zefan 9216c282eb4SLi Zefan /* this is the last extent */ 9226c282eb4SLi Zefan if (em->start + em->len >= i_size_read(inode)) 9236c282eb4SLi Zefan return false; 9246c282eb4SLi Zefan 9256c282eb4SLi Zefan next = defrag_lookup_extent(inode, em->start + em->len); 9266c282eb4SLi Zefan if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE) 9276c282eb4SLi Zefan ret = false; 9286c282eb4SLi Zefan 9296c282eb4SLi Zefan free_extent_map(next); 9306c282eb4SLi Zefan return ret; 9316c282eb4SLi Zefan } 9326c282eb4SLi Zefan 9336c282eb4SLi Zefan static int should_defrag_range(struct inode *inode, u64 start, int thresh, 934a43a2111SAndrew Mahone u64 *last_len, u64 *skip, u64 *defrag_end, 935a43a2111SAndrew Mahone int compress) 9366c282eb4SLi Zefan { 9376c282eb4SLi Zefan struct extent_map *em; 9386c282eb4SLi Zefan int ret = 1; 9396c282eb4SLi Zefan bool next_mergeable = true; 9406c282eb4SLi Zefan 9416c282eb4SLi Zefan /* 9426c282eb4SLi Zefan * make sure that once we start defragging an extent, we keep on 9436c282eb4SLi Zefan * defragging it 9446c282eb4SLi Zefan */ 9456c282eb4SLi Zefan if (start < *defrag_end) 9466c282eb4SLi Zefan return 1; 9476c282eb4SLi Zefan 9486c282eb4SLi Zefan *skip = 0; 9496c282eb4SLi Zefan 9506c282eb4SLi Zefan em = defrag_lookup_extent(inode, start); 9516c282eb4SLi Zefan if (!em) 9526c282eb4SLi Zefan return 0; 9536c282eb4SLi Zefan 954940100a4SChris Mason /* this will cover holes, and inline extents */ 95517ce6ef8SLiu Bo if (em->block_start >= EXTENT_MAP_LAST_BYTE) { 956940100a4SChris Mason ret = 0; 95717ce6ef8SLiu Bo goto out; 95817ce6ef8SLiu Bo } 95917ce6ef8SLiu Bo 9606c282eb4SLi Zefan next_mergeable = defrag_check_next_extent(inode, em); 961940100a4SChris Mason 962940100a4SChris Mason /* 9636c282eb4SLi Zefan * we hit a real extent, if it is big or the next extent is not a 9646c282eb4SLi Zefan * real extent, don't bother defragging it 965940100a4SChris Mason */ 966a43a2111SAndrew Mahone if (!compress && (*last_len == 0 || *last_len >= thresh) && 9676c282eb4SLi Zefan (em->len >= thresh || !next_mergeable)) 968940100a4SChris Mason ret = 0; 96917ce6ef8SLiu Bo out: 970940100a4SChris Mason /* 971940100a4SChris Mason * last_len ends up being a counter of how many bytes we've defragged. 972940100a4SChris Mason * every time we choose not to defrag an extent, we reset *last_len 973940100a4SChris Mason * so that the next tiny extent will force a defrag. 974940100a4SChris Mason * 975940100a4SChris Mason * The end result of this is that tiny extents before a single big 976940100a4SChris Mason * extent will force at least part of that big extent to be defragged. 977940100a4SChris Mason */ 978940100a4SChris Mason if (ret) { 979940100a4SChris Mason *defrag_end = extent_map_end(em); 980940100a4SChris Mason } else { 981940100a4SChris Mason *last_len = 0; 982940100a4SChris Mason *skip = extent_map_end(em); 983940100a4SChris Mason *defrag_end = 0; 984940100a4SChris Mason } 985940100a4SChris Mason 986940100a4SChris Mason free_extent_map(em); 987940100a4SChris Mason return ret; 988940100a4SChris Mason } 989940100a4SChris Mason 9904cb5300bSChris Mason /* 9914cb5300bSChris Mason * it doesn't do much good to defrag one or two pages 9924cb5300bSChris Mason * at a time. This pulls in a nice chunk of pages 9934cb5300bSChris Mason * to COW and defrag. 9944cb5300bSChris Mason * 9954cb5300bSChris Mason * It also makes sure the delalloc code has enough 9964cb5300bSChris Mason * dirty data to avoid making new small extents as part 9974cb5300bSChris Mason * of the defrag 9984cb5300bSChris Mason * 9994cb5300bSChris Mason * It's a good idea to start RA on this range 10004cb5300bSChris Mason * before calling this. 10014cb5300bSChris Mason */ 10024cb5300bSChris Mason static int cluster_pages_for_defrag(struct inode *inode, 10034cb5300bSChris Mason struct page **pages, 10044cb5300bSChris Mason unsigned long start_index, 10054cb5300bSChris Mason int num_pages) 1006f46b5a66SChristoph Hellwig { 10074cb5300bSChris Mason unsigned long file_end; 10084cb5300bSChris Mason u64 isize = i_size_read(inode); 1009f46b5a66SChristoph Hellwig u64 page_start; 1010f46b5a66SChristoph Hellwig u64 page_end; 10111f12bd06SLiu Bo u64 page_cnt; 10124cb5300bSChris Mason int ret; 10134cb5300bSChris Mason int i; 10144cb5300bSChris Mason int i_done; 10154cb5300bSChris Mason struct btrfs_ordered_extent *ordered; 10164cb5300bSChris Mason struct extent_state *cached_state = NULL; 1017600a45e1SMiao Xie struct extent_io_tree *tree; 10183b16a4e3SJosef Bacik gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping); 10194cb5300bSChris Mason 10204cb5300bSChris Mason file_end = (isize - 1) >> PAGE_CACHE_SHIFT; 10211f12bd06SLiu Bo if (!isize || start_index > file_end) 10221f12bd06SLiu Bo return 0; 10231f12bd06SLiu Bo 10241f12bd06SLiu Bo page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1); 10254cb5300bSChris Mason 10264cb5300bSChris Mason ret = btrfs_delalloc_reserve_space(inode, 10271f12bd06SLiu Bo page_cnt << PAGE_CACHE_SHIFT); 10284cb5300bSChris Mason if (ret) 10294cb5300bSChris Mason return ret; 10304cb5300bSChris Mason i_done = 0; 1031600a45e1SMiao Xie tree = &BTRFS_I(inode)->io_tree; 10324cb5300bSChris Mason 10334cb5300bSChris Mason /* step one, lock all the pages */ 10341f12bd06SLiu Bo for (i = 0; i < page_cnt; i++) { 10354cb5300bSChris Mason struct page *page; 1036600a45e1SMiao Xie again: 1037a94733d0SJosef Bacik page = find_or_create_page(inode->i_mapping, 10383b16a4e3SJosef Bacik start_index + i, mask); 10394cb5300bSChris Mason if (!page) 10404cb5300bSChris Mason break; 10414cb5300bSChris Mason 1042600a45e1SMiao Xie page_start = page_offset(page); 1043600a45e1SMiao Xie page_end = page_start + PAGE_CACHE_SIZE - 1; 1044600a45e1SMiao Xie while (1) { 1045d0082371SJeff Mahoney lock_extent(tree, page_start, page_end); 1046600a45e1SMiao Xie ordered = btrfs_lookup_ordered_extent(inode, 1047600a45e1SMiao Xie page_start); 1048d0082371SJeff Mahoney unlock_extent(tree, page_start, page_end); 1049600a45e1SMiao Xie if (!ordered) 1050600a45e1SMiao Xie break; 1051600a45e1SMiao Xie 1052600a45e1SMiao Xie unlock_page(page); 1053600a45e1SMiao Xie btrfs_start_ordered_extent(inode, ordered, 1); 1054600a45e1SMiao Xie btrfs_put_ordered_extent(ordered); 1055600a45e1SMiao Xie lock_page(page); 10561f12bd06SLiu Bo /* 10571f12bd06SLiu Bo * we unlocked the page above, so we need check if 10581f12bd06SLiu Bo * it was released or not. 10591f12bd06SLiu Bo */ 10601f12bd06SLiu Bo if (page->mapping != inode->i_mapping) { 10611f12bd06SLiu Bo unlock_page(page); 10621f12bd06SLiu Bo page_cache_release(page); 10631f12bd06SLiu Bo goto again; 10641f12bd06SLiu Bo } 1065600a45e1SMiao Xie } 1066600a45e1SMiao Xie 10674cb5300bSChris Mason if (!PageUptodate(page)) { 10684cb5300bSChris Mason btrfs_readpage(NULL, page); 10694cb5300bSChris Mason lock_page(page); 10704cb5300bSChris Mason if (!PageUptodate(page)) { 10714cb5300bSChris Mason unlock_page(page); 10724cb5300bSChris Mason page_cache_release(page); 10734cb5300bSChris Mason ret = -EIO; 10744cb5300bSChris Mason break; 10754cb5300bSChris Mason } 10764cb5300bSChris Mason } 1077600a45e1SMiao Xie 1078600a45e1SMiao Xie if (page->mapping != inode->i_mapping) { 1079600a45e1SMiao Xie unlock_page(page); 1080600a45e1SMiao Xie page_cache_release(page); 1081600a45e1SMiao Xie goto again; 1082600a45e1SMiao Xie } 1083600a45e1SMiao Xie 10844cb5300bSChris Mason pages[i] = page; 10854cb5300bSChris Mason i_done++; 10864cb5300bSChris Mason } 10874cb5300bSChris Mason if (!i_done || ret) 10884cb5300bSChris Mason goto out; 10894cb5300bSChris Mason 10904cb5300bSChris Mason if (!(inode->i_sb->s_flags & MS_ACTIVE)) 10914cb5300bSChris Mason goto out; 10924cb5300bSChris Mason 10934cb5300bSChris Mason /* 10944cb5300bSChris Mason * so now we have a nice long stream of locked 10954cb5300bSChris Mason * and up to date pages, lets wait on them 10964cb5300bSChris Mason */ 10974cb5300bSChris Mason for (i = 0; i < i_done; i++) 10984cb5300bSChris Mason wait_on_page_writeback(pages[i]); 10994cb5300bSChris Mason 11004cb5300bSChris Mason page_start = page_offset(pages[0]); 11014cb5300bSChris Mason page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE; 11024cb5300bSChris Mason 11034cb5300bSChris Mason lock_extent_bits(&BTRFS_I(inode)->io_tree, 1104d0082371SJeff Mahoney page_start, page_end - 1, 0, &cached_state); 11054cb5300bSChris Mason clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, 11064cb5300bSChris Mason page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC | 11079e8a4a8bSLiu Bo EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0, 11089e8a4a8bSLiu Bo &cached_state, GFP_NOFS); 11094cb5300bSChris Mason 11101f12bd06SLiu Bo if (i_done != page_cnt) { 11119e0baf60SJosef Bacik spin_lock(&BTRFS_I(inode)->lock); 11129e0baf60SJosef Bacik BTRFS_I(inode)->outstanding_extents++; 11139e0baf60SJosef Bacik spin_unlock(&BTRFS_I(inode)->lock); 11144cb5300bSChris Mason btrfs_delalloc_release_space(inode, 11151f12bd06SLiu Bo (page_cnt - i_done) << PAGE_CACHE_SHIFT); 11164cb5300bSChris Mason } 11174cb5300bSChris Mason 11184cb5300bSChris Mason 11199e8a4a8bSLiu Bo set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1, 11209e8a4a8bSLiu Bo &cached_state, GFP_NOFS); 11214cb5300bSChris Mason 11224cb5300bSChris Mason unlock_extent_cached(&BTRFS_I(inode)->io_tree, 11234cb5300bSChris Mason page_start, page_end - 1, &cached_state, 11244cb5300bSChris Mason GFP_NOFS); 11254cb5300bSChris Mason 11264cb5300bSChris Mason for (i = 0; i < i_done; i++) { 11274cb5300bSChris Mason clear_page_dirty_for_io(pages[i]); 11284cb5300bSChris Mason ClearPageChecked(pages[i]); 11294cb5300bSChris Mason set_page_extent_mapped(pages[i]); 11304cb5300bSChris Mason set_page_dirty(pages[i]); 11314cb5300bSChris Mason unlock_page(pages[i]); 11324cb5300bSChris Mason page_cache_release(pages[i]); 11334cb5300bSChris Mason } 11344cb5300bSChris Mason return i_done; 11354cb5300bSChris Mason out: 11364cb5300bSChris Mason for (i = 0; i < i_done; i++) { 11374cb5300bSChris Mason unlock_page(pages[i]); 11384cb5300bSChris Mason page_cache_release(pages[i]); 11394cb5300bSChris Mason } 11401f12bd06SLiu Bo btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT); 11414cb5300bSChris Mason return ret; 11424cb5300bSChris Mason 11434cb5300bSChris Mason } 11444cb5300bSChris Mason 11454cb5300bSChris Mason int btrfs_defrag_file(struct inode *inode, struct file *file, 11464cb5300bSChris Mason struct btrfs_ioctl_defrag_range_args *range, 11474cb5300bSChris Mason u64 newer_than, unsigned long max_to_defrag) 11484cb5300bSChris Mason { 11494cb5300bSChris Mason struct btrfs_root *root = BTRFS_I(inode)->root; 11504cb5300bSChris Mason struct file_ra_state *ra = NULL; 11514cb5300bSChris Mason unsigned long last_index; 1152151a31b2SLi Zefan u64 isize = i_size_read(inode); 1153940100a4SChris Mason u64 last_len = 0; 1154940100a4SChris Mason u64 skip = 0; 1155940100a4SChris Mason u64 defrag_end = 0; 11564cb5300bSChris Mason u64 newer_off = range->start; 1157f46b5a66SChristoph Hellwig unsigned long i; 1158008873eaSLi Zefan unsigned long ra_index = 0; 1159f46b5a66SChristoph Hellwig int ret; 11604cb5300bSChris Mason int defrag_count = 0; 11611a419d85SLi Zefan int compress_type = BTRFS_COMPRESS_ZLIB; 11624cb5300bSChris Mason int extent_thresh = range->extent_thresh; 1163008873eaSLi Zefan int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT; 1164008873eaSLi Zefan int cluster = max_cluster; 11654cb5300bSChris Mason u64 new_align = ~((u64)128 * 1024 - 1); 11664cb5300bSChris Mason struct page **pages = NULL; 11674cb5300bSChris Mason 11680abd5b17SLiu Bo if (isize == 0) 11690abd5b17SLiu Bo return 0; 11700abd5b17SLiu Bo 11710abd5b17SLiu Bo if (range->start >= isize) 11720abd5b17SLiu Bo return -EINVAL; 11731a419d85SLi Zefan 11741a419d85SLi Zefan if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) { 11751a419d85SLi Zefan if (range->compress_type > BTRFS_COMPRESS_TYPES) 11761a419d85SLi Zefan return -EINVAL; 11771a419d85SLi Zefan if (range->compress_type) 11781a419d85SLi Zefan compress_type = range->compress_type; 11791a419d85SLi Zefan } 1180f46b5a66SChristoph Hellwig 11810abd5b17SLiu Bo if (extent_thresh == 0) 11820abd5b17SLiu Bo extent_thresh = 256 * 1024; 1183f46b5a66SChristoph Hellwig 11844cb5300bSChris Mason /* 11854cb5300bSChris Mason * if we were not given a file, allocate a readahead 11864cb5300bSChris Mason * context 11874cb5300bSChris Mason */ 11884cb5300bSChris Mason if (!file) { 11894cb5300bSChris Mason ra = kzalloc(sizeof(*ra), GFP_NOFS); 11904cb5300bSChris Mason if (!ra) 11914cb5300bSChris Mason return -ENOMEM; 11924cb5300bSChris Mason file_ra_state_init(ra, inode->i_mapping); 11934cb5300bSChris Mason } else { 11944cb5300bSChris Mason ra = &file->f_ra; 11954cb5300bSChris Mason } 11964cb5300bSChris Mason 1197008873eaSLi Zefan pages = kmalloc(sizeof(struct page *) * max_cluster, 11984cb5300bSChris Mason GFP_NOFS); 11994cb5300bSChris Mason if (!pages) { 12004cb5300bSChris Mason ret = -ENOMEM; 12014cb5300bSChris Mason goto out_ra; 12024cb5300bSChris Mason } 12034cb5300bSChris Mason 12044cb5300bSChris Mason /* find the last page to defrag */ 12051e701a32SChris Mason if (range->start + range->len > range->start) { 1206151a31b2SLi Zefan last_index = min_t(u64, isize - 1, 12071e701a32SChris Mason range->start + range->len - 1) >> PAGE_CACHE_SHIFT; 12081e701a32SChris Mason } else { 1209151a31b2SLi Zefan last_index = (isize - 1) >> PAGE_CACHE_SHIFT; 12101e701a32SChris Mason } 12111e701a32SChris Mason 12124cb5300bSChris Mason if (newer_than) { 12134cb5300bSChris Mason ret = find_new_extents(root, inode, newer_than, 12144cb5300bSChris Mason &newer_off, 64 * 1024); 12154cb5300bSChris Mason if (!ret) { 12164cb5300bSChris Mason range->start = newer_off; 12174cb5300bSChris Mason /* 12184cb5300bSChris Mason * we always align our defrag to help keep 12194cb5300bSChris Mason * the extents in the file evenly spaced 12204cb5300bSChris Mason */ 12214cb5300bSChris Mason i = (newer_off & new_align) >> PAGE_CACHE_SHIFT; 12224cb5300bSChris Mason } else 12234cb5300bSChris Mason goto out_ra; 12244cb5300bSChris Mason } else { 12251e701a32SChris Mason i = range->start >> PAGE_CACHE_SHIFT; 12264cb5300bSChris Mason } 12274cb5300bSChris Mason if (!max_to_defrag) 12287ec31b54SLiu Bo max_to_defrag = last_index + 1; 12294cb5300bSChris Mason 12302a0f7f57SLi Zefan /* 12312a0f7f57SLi Zefan * make writeback starts from i, so the defrag range can be 12322a0f7f57SLi Zefan * written sequentially. 12332a0f7f57SLi Zefan */ 12342a0f7f57SLi Zefan if (i < inode->i_mapping->writeback_index) 12352a0f7f57SLi Zefan inode->i_mapping->writeback_index = i; 12362a0f7f57SLi Zefan 1237f7f43cc8SChris Mason while (i <= last_index && defrag_count < max_to_defrag && 1238f7f43cc8SChris Mason (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >> 1239f7f43cc8SChris Mason PAGE_CACHE_SHIFT)) { 12404cb5300bSChris Mason /* 12414cb5300bSChris Mason * make sure we stop running if someone unmounts 12424cb5300bSChris Mason * the FS 12434cb5300bSChris Mason */ 12444cb5300bSChris Mason if (!(inode->i_sb->s_flags & MS_ACTIVE)) 12454cb5300bSChris Mason break; 12464cb5300bSChris Mason 1247210549ebSDavid Sterba if (btrfs_defrag_cancelled(root->fs_info)) { 1248210549ebSDavid Sterba printk(KERN_DEBUG "btrfs: defrag_file cancelled\n"); 1249210549ebSDavid Sterba ret = -EAGAIN; 1250210549ebSDavid Sterba break; 1251210549ebSDavid Sterba } 1252210549ebSDavid Sterba 125366c26892SLiu Bo if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT, 12546c282eb4SLi Zefan extent_thresh, &last_len, &skip, 1255a43a2111SAndrew Mahone &defrag_end, range->flags & 1256a43a2111SAndrew Mahone BTRFS_DEFRAG_RANGE_COMPRESS)) { 1257940100a4SChris Mason unsigned long next; 1258940100a4SChris Mason /* 1259940100a4SChris Mason * the should_defrag function tells us how much to skip 1260940100a4SChris Mason * bump our counter by the suggested amount 1261940100a4SChris Mason */ 1262940100a4SChris Mason next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT; 1263940100a4SChris Mason i = max(i + 1, next); 1264940100a4SChris Mason continue; 1265940100a4SChris Mason } 1266008873eaSLi Zefan 1267008873eaSLi Zefan if (!newer_than) { 1268008873eaSLi Zefan cluster = (PAGE_CACHE_ALIGN(defrag_end) >> 1269008873eaSLi Zefan PAGE_CACHE_SHIFT) - i; 1270008873eaSLi Zefan cluster = min(cluster, max_cluster); 1271008873eaSLi Zefan } else { 1272008873eaSLi Zefan cluster = max_cluster; 1273008873eaSLi Zefan } 1274008873eaSLi Zefan 12751e701a32SChris Mason if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) 12761a419d85SLi Zefan BTRFS_I(inode)->force_compress = compress_type; 1277940100a4SChris Mason 1278008873eaSLi Zefan if (i + cluster > ra_index) { 1279008873eaSLi Zefan ra_index = max(i, ra_index); 1280008873eaSLi Zefan btrfs_force_ra(inode->i_mapping, ra, file, ra_index, 1281008873eaSLi Zefan cluster); 1282008873eaSLi Zefan ra_index += max_cluster; 1283008873eaSLi Zefan } 12844cb5300bSChris Mason 1285ecb8bea8SLiu Bo mutex_lock(&inode->i_mutex); 1286008873eaSLi Zefan ret = cluster_pages_for_defrag(inode, pages, i, cluster); 1287ecb8bea8SLiu Bo if (ret < 0) { 1288ecb8bea8SLiu Bo mutex_unlock(&inode->i_mutex); 12894cb5300bSChris Mason goto out_ra; 1290ecb8bea8SLiu Bo } 12914cb5300bSChris Mason 12924cb5300bSChris Mason defrag_count += ret; 1293d0e1d66bSNamjae Jeon balance_dirty_pages_ratelimited(inode->i_mapping); 1294ecb8bea8SLiu Bo mutex_unlock(&inode->i_mutex); 12954cb5300bSChris Mason 12964cb5300bSChris Mason if (newer_than) { 12974cb5300bSChris Mason if (newer_off == (u64)-1) 12984cb5300bSChris Mason break; 12994cb5300bSChris Mason 1300e1f041e1SLiu Bo if (ret > 0) 1301e1f041e1SLiu Bo i += ret; 1302e1f041e1SLiu Bo 13034cb5300bSChris Mason newer_off = max(newer_off + 1, 13044cb5300bSChris Mason (u64)i << PAGE_CACHE_SHIFT); 13054cb5300bSChris Mason 13064cb5300bSChris Mason ret = find_new_extents(root, inode, 13074cb5300bSChris Mason newer_than, &newer_off, 13084cb5300bSChris Mason 64 * 1024); 13094cb5300bSChris Mason if (!ret) { 13104cb5300bSChris Mason range->start = newer_off; 13114cb5300bSChris Mason i = (newer_off & new_align) >> PAGE_CACHE_SHIFT; 13124cb5300bSChris Mason } else { 13134cb5300bSChris Mason break; 1314940100a4SChris Mason } 13154cb5300bSChris Mason } else { 1316008873eaSLi Zefan if (ret > 0) { 1317cbcc8326SLi Zefan i += ret; 1318008873eaSLi Zefan last_len += ret << PAGE_CACHE_SHIFT; 1319008873eaSLi Zefan } else { 1320940100a4SChris Mason i++; 1321008873eaSLi Zefan last_len = 0; 1322008873eaSLi Zefan } 1323f46b5a66SChristoph Hellwig } 13244cb5300bSChris Mason } 1325f46b5a66SChristoph Hellwig 13261e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) 13271e701a32SChris Mason filemap_flush(inode->i_mapping); 13281e701a32SChris Mason 13291e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 13301e701a32SChris Mason /* the filemap_flush will queue IO into the worker threads, but 13311e701a32SChris Mason * we have to make sure the IO is actually started and that 13321e701a32SChris Mason * ordered extents get created before we return 13331e701a32SChris Mason */ 13341e701a32SChris Mason atomic_inc(&root->fs_info->async_submit_draining); 13351e701a32SChris Mason while (atomic_read(&root->fs_info->nr_async_submits) || 13361e701a32SChris Mason atomic_read(&root->fs_info->async_delalloc_pages)) { 13371e701a32SChris Mason wait_event(root->fs_info->async_submit_wait, 13381e701a32SChris Mason (atomic_read(&root->fs_info->nr_async_submits) == 0 && 13391e701a32SChris Mason atomic_read(&root->fs_info->async_delalloc_pages) == 0)); 13401e701a32SChris Mason } 13411e701a32SChris Mason atomic_dec(&root->fs_info->async_submit_draining); 13421e701a32SChris Mason 13431e701a32SChris Mason mutex_lock(&inode->i_mutex); 1344261507a0SLi Zefan BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE; 13451e701a32SChris Mason mutex_unlock(&inode->i_mutex); 13461e701a32SChris Mason } 13471e701a32SChris Mason 13481a419d85SLi Zefan if (range->compress_type == BTRFS_COMPRESS_LZO) { 13492b0ce2c2SMitch Harder btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO); 13501a419d85SLi Zefan } 13511a419d85SLi Zefan 135260ccf82fSDiego Calleja ret = defrag_count; 1353940100a4SChris Mason 13544cb5300bSChris Mason out_ra: 13554cb5300bSChris Mason if (!file) 13564cb5300bSChris Mason kfree(ra); 13574cb5300bSChris Mason kfree(pages); 1358940100a4SChris Mason return ret; 1359f46b5a66SChristoph Hellwig } 1360f46b5a66SChristoph Hellwig 1361198605a8SMiao Xie static noinline int btrfs_ioctl_resize(struct file *file, 136276dda93cSYan, Zheng void __user *arg) 1363f46b5a66SChristoph Hellwig { 1364f46b5a66SChristoph Hellwig u64 new_size; 1365f46b5a66SChristoph Hellwig u64 old_size; 1366f46b5a66SChristoph Hellwig u64 devid = 1; 1367496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 1368f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 1369f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 1370f46b5a66SChristoph Hellwig struct btrfs_device *device = NULL; 1371f46b5a66SChristoph Hellwig char *sizestr; 1372f46b5a66SChristoph Hellwig char *devstr = NULL; 1373f46b5a66SChristoph Hellwig int ret = 0; 1374f46b5a66SChristoph Hellwig int mod = 0; 1375f46b5a66SChristoph Hellwig 1376e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 1377e441d54dSChris Mason return -EPERM; 1378e441d54dSChris Mason 1379198605a8SMiao Xie ret = mnt_want_write_file(file); 1380198605a8SMiao Xie if (ret) 1381198605a8SMiao Xie return ret; 1382198605a8SMiao Xie 13835ac00addSStefan Behrens if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running, 13845ac00addSStefan Behrens 1)) { 13855ac00addSStefan Behrens pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n"); 138697547676SMiao Xie mnt_drop_write_file(file); 13872c0c9da0SIlya Dryomov return -EINVAL; 1388c9e9f97bSIlya Dryomov } 1389c9e9f97bSIlya Dryomov 13905ac00addSStefan Behrens mutex_lock(&root->fs_info->volume_mutex); 1391dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1392c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 1393c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 1394c9e9f97bSIlya Dryomov goto out; 1395c9e9f97bSIlya Dryomov } 13965516e595SMark Fasheh 13975516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1398f46b5a66SChristoph Hellwig 1399f46b5a66SChristoph Hellwig sizestr = vol_args->name; 1400f46b5a66SChristoph Hellwig devstr = strchr(sizestr, ':'); 1401f46b5a66SChristoph Hellwig if (devstr) { 1402f46b5a66SChristoph Hellwig char *end; 1403f46b5a66SChristoph Hellwig sizestr = devstr + 1; 1404f46b5a66SChristoph Hellwig *devstr = '\0'; 1405f46b5a66SChristoph Hellwig devstr = vol_args->name; 1406f46b5a66SChristoph Hellwig devid = simple_strtoull(devstr, &end, 10); 1407dfd79829SMiao Xie if (!devid) { 1408dfd79829SMiao Xie ret = -EINVAL; 1409dfd79829SMiao Xie goto out_free; 1410dfd79829SMiao Xie } 14115bb14682SArnd Hannemann printk(KERN_INFO "btrfs: resizing devid %llu\n", 141221380931SJoel Becker (unsigned long long)devid); 1413f46b5a66SChristoph Hellwig } 1414dba60f3fSMiao Xie 1415aa1b8cd4SStefan Behrens device = btrfs_find_device(root->fs_info, devid, NULL, NULL); 1416f46b5a66SChristoph Hellwig if (!device) { 14175bb14682SArnd Hannemann printk(KERN_INFO "btrfs: resizer unable to find device %llu\n", 141821380931SJoel Becker (unsigned long long)devid); 1419dfd79829SMiao Xie ret = -ENODEV; 1420c9e9f97bSIlya Dryomov goto out_free; 1421f46b5a66SChristoph Hellwig } 1422dba60f3fSMiao Xie 1423dba60f3fSMiao Xie if (!device->writeable) { 14244e42ae1bSLiu Bo printk(KERN_INFO "btrfs: resizer unable to apply on " 1425dba60f3fSMiao Xie "readonly device %llu\n", 1426a8c4a33bSChris Mason (unsigned long long)devid); 1427dfd79829SMiao Xie ret = -EPERM; 14284e42ae1bSLiu Bo goto out_free; 14294e42ae1bSLiu Bo } 14304e42ae1bSLiu Bo 1431f46b5a66SChristoph Hellwig if (!strcmp(sizestr, "max")) 1432f46b5a66SChristoph Hellwig new_size = device->bdev->bd_inode->i_size; 1433f46b5a66SChristoph Hellwig else { 1434f46b5a66SChristoph Hellwig if (sizestr[0] == '-') { 1435f46b5a66SChristoph Hellwig mod = -1; 1436f46b5a66SChristoph Hellwig sizestr++; 1437f46b5a66SChristoph Hellwig } else if (sizestr[0] == '+') { 1438f46b5a66SChristoph Hellwig mod = 1; 1439f46b5a66SChristoph Hellwig sizestr++; 1440f46b5a66SChristoph Hellwig } 144191748467SAkinobu Mita new_size = memparse(sizestr, NULL); 1442f46b5a66SChristoph Hellwig if (new_size == 0) { 1443f46b5a66SChristoph Hellwig ret = -EINVAL; 1444c9e9f97bSIlya Dryomov goto out_free; 1445f46b5a66SChristoph Hellwig } 1446f46b5a66SChristoph Hellwig } 1447f46b5a66SChristoph Hellwig 144863a212abSStefan Behrens if (device->is_tgtdev_for_dev_replace) { 1449dfd79829SMiao Xie ret = -EPERM; 145063a212abSStefan Behrens goto out_free; 145163a212abSStefan Behrens } 145263a212abSStefan Behrens 1453f46b5a66SChristoph Hellwig old_size = device->total_bytes; 1454f46b5a66SChristoph Hellwig 1455f46b5a66SChristoph Hellwig if (mod < 0) { 1456f46b5a66SChristoph Hellwig if (new_size > old_size) { 1457f46b5a66SChristoph Hellwig ret = -EINVAL; 1458c9e9f97bSIlya Dryomov goto out_free; 1459f46b5a66SChristoph Hellwig } 1460f46b5a66SChristoph Hellwig new_size = old_size - new_size; 1461f46b5a66SChristoph Hellwig } else if (mod > 0) { 1462f46b5a66SChristoph Hellwig new_size = old_size + new_size; 1463f46b5a66SChristoph Hellwig } 1464f46b5a66SChristoph Hellwig 1465f46b5a66SChristoph Hellwig if (new_size < 256 * 1024 * 1024) { 1466f46b5a66SChristoph Hellwig ret = -EINVAL; 1467c9e9f97bSIlya Dryomov goto out_free; 1468f46b5a66SChristoph Hellwig } 1469f46b5a66SChristoph Hellwig if (new_size > device->bdev->bd_inode->i_size) { 1470f46b5a66SChristoph Hellwig ret = -EFBIG; 1471c9e9f97bSIlya Dryomov goto out_free; 1472f46b5a66SChristoph Hellwig } 1473f46b5a66SChristoph Hellwig 1474f46b5a66SChristoph Hellwig do_div(new_size, root->sectorsize); 1475f46b5a66SChristoph Hellwig new_size *= root->sectorsize; 1476f46b5a66SChristoph Hellwig 1477606686eeSJosef Bacik printk_in_rcu(KERN_INFO "btrfs: new size for %s is %llu\n", 1478606686eeSJosef Bacik rcu_str_deref(device->name), 1479606686eeSJosef Bacik (unsigned long long)new_size); 1480f46b5a66SChristoph Hellwig 1481f46b5a66SChristoph Hellwig if (new_size > old_size) { 1482a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 148398d5dc13STsutomu Itoh if (IS_ERR(trans)) { 148498d5dc13STsutomu Itoh ret = PTR_ERR(trans); 1485c9e9f97bSIlya Dryomov goto out_free; 148698d5dc13STsutomu Itoh } 1487f46b5a66SChristoph Hellwig ret = btrfs_grow_device(trans, device, new_size); 1488f46b5a66SChristoph Hellwig btrfs_commit_transaction(trans, root); 1489ece7d20eSMike Fleetwood } else if (new_size < old_size) { 1490f46b5a66SChristoph Hellwig ret = btrfs_shrink_device(device, new_size); 14910253f40eSjeff.liu } /* equal, nothing need to do */ 1492f46b5a66SChristoph Hellwig 1493c9e9f97bSIlya Dryomov out_free: 1494f46b5a66SChristoph Hellwig kfree(vol_args); 1495c9e9f97bSIlya Dryomov out: 1496c9e9f97bSIlya Dryomov mutex_unlock(&root->fs_info->volume_mutex); 14975ac00addSStefan Behrens atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0); 149818f39c41SIlya Dryomov mnt_drop_write_file(file); 1499f46b5a66SChristoph Hellwig return ret; 1500f46b5a66SChristoph Hellwig } 1501f46b5a66SChristoph Hellwig 150272fd032eSSage Weil static noinline int btrfs_ioctl_snap_create_transid(struct file *file, 15036f72c7e2SArne Jansen char *name, unsigned long fd, int subvol, 15046f72c7e2SArne Jansen u64 *transid, bool readonly, 15058696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 1506f46b5a66SChristoph Hellwig { 1507f46b5a66SChristoph Hellwig int namelen; 15083de4586cSChris Mason int ret = 0; 1509f46b5a66SChristoph Hellwig 1510a874a63eSLiu Bo ret = mnt_want_write_file(file); 1511a874a63eSLiu Bo if (ret) 1512a874a63eSLiu Bo goto out; 1513a874a63eSLiu Bo 151472fd032eSSage Weil namelen = strlen(name); 151572fd032eSSage Weil if (strchr(name, '/')) { 1516f46b5a66SChristoph Hellwig ret = -EINVAL; 1517a874a63eSLiu Bo goto out_drop_write; 1518f46b5a66SChristoph Hellwig } 1519f46b5a66SChristoph Hellwig 152016780cabSChris Mason if (name[0] == '.' && 152116780cabSChris Mason (namelen == 1 || (name[1] == '.' && namelen == 2))) { 152216780cabSChris Mason ret = -EEXIST; 1523a874a63eSLiu Bo goto out_drop_write; 152416780cabSChris Mason } 152516780cabSChris Mason 15263de4586cSChris Mason if (subvol) { 152772fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 15286f72c7e2SArne Jansen NULL, transid, readonly, inherit); 1529cb8e7090SChristoph Hellwig } else { 15302903ff01SAl Viro struct fd src = fdget(fd); 15313de4586cSChris Mason struct inode *src_inode; 15322903ff01SAl Viro if (!src.file) { 15333de4586cSChris Mason ret = -EINVAL; 1534a874a63eSLiu Bo goto out_drop_write; 15353de4586cSChris Mason } 15363de4586cSChris Mason 1537496ad9aaSAl Viro src_inode = file_inode(src.file); 1538496ad9aaSAl Viro if (src_inode->i_sb != file_inode(file)->i_sb) { 1539d397712bSChris Mason printk(KERN_INFO "btrfs: Snapshot src from " 1540d397712bSChris Mason "another FS\n"); 15413de4586cSChris Mason ret = -EINVAL; 1542ecd18815SAl Viro } else { 154372fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 154472fd032eSSage Weil BTRFS_I(src_inode)->root, 15456f72c7e2SArne Jansen transid, readonly, inherit); 1546ecd18815SAl Viro } 15472903ff01SAl Viro fdput(src); 1548cb8e7090SChristoph Hellwig } 1549a874a63eSLiu Bo out_drop_write: 1550a874a63eSLiu Bo mnt_drop_write_file(file); 1551f46b5a66SChristoph Hellwig out: 155272fd032eSSage Weil return ret; 155372fd032eSSage Weil } 155472fd032eSSage Weil 155572fd032eSSage Weil static noinline int btrfs_ioctl_snap_create(struct file *file, 1556fa0d2b9bSLi Zefan void __user *arg, int subvol) 155772fd032eSSage Weil { 1558fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args *vol_args; 155972fd032eSSage Weil int ret; 156072fd032eSSage Weil 1561fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1562fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1563fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1564fa0d2b9bSLi Zefan vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1565fa0d2b9bSLi Zefan 1566fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 1567b83cc969SLi Zefan vol_args->fd, subvol, 15686f72c7e2SArne Jansen NULL, false, NULL); 1569fa0d2b9bSLi Zefan 1570fa0d2b9bSLi Zefan kfree(vol_args); 1571fa0d2b9bSLi Zefan return ret; 1572fa0d2b9bSLi Zefan } 1573fa0d2b9bSLi Zefan 1574fa0d2b9bSLi Zefan static noinline int btrfs_ioctl_snap_create_v2(struct file *file, 1575fa0d2b9bSLi Zefan void __user *arg, int subvol) 1576fa0d2b9bSLi Zefan { 1577fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args_v2 *vol_args; 1578fa0d2b9bSLi Zefan int ret; 1579fdfb1e4fSLi Zefan u64 transid = 0; 1580fdfb1e4fSLi Zefan u64 *ptr = NULL; 1581b83cc969SLi Zefan bool readonly = false; 15826f72c7e2SArne Jansen struct btrfs_qgroup_inherit *inherit = NULL; 158372fd032eSSage Weil 1584fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1585fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1586fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1587fa0d2b9bSLi Zefan vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 1588fdfb1e4fSLi Zefan 1589b83cc969SLi Zefan if (vol_args->flags & 15906f72c7e2SArne Jansen ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY | 15916f72c7e2SArne Jansen BTRFS_SUBVOL_QGROUP_INHERIT)) { 1592b83cc969SLi Zefan ret = -EOPNOTSUPP; 1593fdfb1e4fSLi Zefan goto out; 1594fdfb1e4fSLi Zefan } 1595fdfb1e4fSLi Zefan 1596fa0d2b9bSLi Zefan if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC) 1597fdfb1e4fSLi Zefan ptr = &transid; 1598b83cc969SLi Zefan if (vol_args->flags & BTRFS_SUBVOL_RDONLY) 1599b83cc969SLi Zefan readonly = true; 16006f72c7e2SArne Jansen if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) { 16016f72c7e2SArne Jansen if (vol_args->size > PAGE_CACHE_SIZE) { 16026f72c7e2SArne Jansen ret = -EINVAL; 16036f72c7e2SArne Jansen goto out; 16046f72c7e2SArne Jansen } 16056f72c7e2SArne Jansen inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size); 16066f72c7e2SArne Jansen if (IS_ERR(inherit)) { 16076f72c7e2SArne Jansen ret = PTR_ERR(inherit); 16086f72c7e2SArne Jansen goto out; 16096f72c7e2SArne Jansen } 16106f72c7e2SArne Jansen } 161175eaa0e2SSage Weil 1612fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 16136f72c7e2SArne Jansen vol_args->fd, subvol, ptr, 16148696c533SMiao Xie readonly, inherit); 161575eaa0e2SSage Weil 1616fdfb1e4fSLi Zefan if (ret == 0 && ptr && 161775eaa0e2SSage Weil copy_to_user(arg + 1618fdfb1e4fSLi Zefan offsetof(struct btrfs_ioctl_vol_args_v2, 1619fdfb1e4fSLi Zefan transid), ptr, sizeof(*ptr))) 162075eaa0e2SSage Weil ret = -EFAULT; 1621fdfb1e4fSLi Zefan out: 1622f46b5a66SChristoph Hellwig kfree(vol_args); 16236f72c7e2SArne Jansen kfree(inherit); 1624f46b5a66SChristoph Hellwig return ret; 1625f46b5a66SChristoph Hellwig } 1626f46b5a66SChristoph Hellwig 16270caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_getflags(struct file *file, 16280caa102dSLi Zefan void __user *arg) 16290caa102dSLi Zefan { 1630496ad9aaSAl Viro struct inode *inode = file_inode(file); 16310caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 16320caa102dSLi Zefan int ret = 0; 16330caa102dSLi Zefan u64 flags = 0; 16340caa102dSLi Zefan 163533345d01SLi Zefan if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) 16360caa102dSLi Zefan return -EINVAL; 16370caa102dSLi Zefan 16380caa102dSLi Zefan down_read(&root->fs_info->subvol_sem); 16390caa102dSLi Zefan if (btrfs_root_readonly(root)) 16400caa102dSLi Zefan flags |= BTRFS_SUBVOL_RDONLY; 16410caa102dSLi Zefan up_read(&root->fs_info->subvol_sem); 16420caa102dSLi Zefan 16430caa102dSLi Zefan if (copy_to_user(arg, &flags, sizeof(flags))) 16440caa102dSLi Zefan ret = -EFAULT; 16450caa102dSLi Zefan 16460caa102dSLi Zefan return ret; 16470caa102dSLi Zefan } 16480caa102dSLi Zefan 16490caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_setflags(struct file *file, 16500caa102dSLi Zefan void __user *arg) 16510caa102dSLi Zefan { 1652496ad9aaSAl Viro struct inode *inode = file_inode(file); 16530caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 16540caa102dSLi Zefan struct btrfs_trans_handle *trans; 16550caa102dSLi Zefan u64 root_flags; 16560caa102dSLi Zefan u64 flags; 16570caa102dSLi Zefan int ret = 0; 16580caa102dSLi Zefan 1659b9ca0664SLiu Bo ret = mnt_want_write_file(file); 1660b9ca0664SLiu Bo if (ret) 1661b9ca0664SLiu Bo goto out; 16620caa102dSLi Zefan 1663b9ca0664SLiu Bo if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { 1664b9ca0664SLiu Bo ret = -EINVAL; 1665b9ca0664SLiu Bo goto out_drop_write; 1666b9ca0664SLiu Bo } 16670caa102dSLi Zefan 1668b9ca0664SLiu Bo if (copy_from_user(&flags, arg, sizeof(flags))) { 1669b9ca0664SLiu Bo ret = -EFAULT; 1670b9ca0664SLiu Bo goto out_drop_write; 1671b9ca0664SLiu Bo } 16720caa102dSLi Zefan 1673b9ca0664SLiu Bo if (flags & BTRFS_SUBVOL_CREATE_ASYNC) { 1674b9ca0664SLiu Bo ret = -EINVAL; 1675b9ca0664SLiu Bo goto out_drop_write; 1676b9ca0664SLiu Bo } 16770caa102dSLi Zefan 1678b9ca0664SLiu Bo if (flags & ~BTRFS_SUBVOL_RDONLY) { 1679b9ca0664SLiu Bo ret = -EOPNOTSUPP; 1680b9ca0664SLiu Bo goto out_drop_write; 1681b9ca0664SLiu Bo } 16820caa102dSLi Zefan 1683b9ca0664SLiu Bo if (!inode_owner_or_capable(inode)) { 1684b9ca0664SLiu Bo ret = -EACCES; 1685b9ca0664SLiu Bo goto out_drop_write; 1686b9ca0664SLiu Bo } 1687b4dc2b8cSLi Zefan 16880caa102dSLi Zefan down_write(&root->fs_info->subvol_sem); 16890caa102dSLi Zefan 16900caa102dSLi Zefan /* nothing to do */ 16910caa102dSLi Zefan if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) 1692b9ca0664SLiu Bo goto out_drop_sem; 16930caa102dSLi Zefan 16940caa102dSLi Zefan root_flags = btrfs_root_flags(&root->root_item); 16950caa102dSLi Zefan if (flags & BTRFS_SUBVOL_RDONLY) 16960caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 16970caa102dSLi Zefan root_flags | BTRFS_ROOT_SUBVOL_RDONLY); 16980caa102dSLi Zefan else 16990caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 17000caa102dSLi Zefan root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); 17010caa102dSLi Zefan 17020caa102dSLi Zefan trans = btrfs_start_transaction(root, 1); 17030caa102dSLi Zefan if (IS_ERR(trans)) { 17040caa102dSLi Zefan ret = PTR_ERR(trans); 17050caa102dSLi Zefan goto out_reset; 17060caa102dSLi Zefan } 17070caa102dSLi Zefan 1708b4dc2b8cSLi Zefan ret = btrfs_update_root(trans, root->fs_info->tree_root, 17090caa102dSLi Zefan &root->root_key, &root->root_item); 17100caa102dSLi Zefan 17110caa102dSLi Zefan btrfs_commit_transaction(trans, root); 17120caa102dSLi Zefan out_reset: 17130caa102dSLi Zefan if (ret) 17140caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, root_flags); 1715b9ca0664SLiu Bo out_drop_sem: 17160caa102dSLi Zefan up_write(&root->fs_info->subvol_sem); 1717b9ca0664SLiu Bo out_drop_write: 1718b9ca0664SLiu Bo mnt_drop_write_file(file); 1719b9ca0664SLiu Bo out: 17200caa102dSLi Zefan return ret; 17210caa102dSLi Zefan } 17220caa102dSLi Zefan 172376dda93cSYan, Zheng /* 172476dda93cSYan, Zheng * helper to check if the subvolume references other subvolumes 172576dda93cSYan, Zheng */ 172676dda93cSYan, Zheng static noinline int may_destroy_subvol(struct btrfs_root *root) 172776dda93cSYan, Zheng { 172876dda93cSYan, Zheng struct btrfs_path *path; 1729175a2b87SJosef Bacik struct btrfs_dir_item *di; 173076dda93cSYan, Zheng struct btrfs_key key; 1731175a2b87SJosef Bacik u64 dir_id; 173276dda93cSYan, Zheng int ret; 173376dda93cSYan, Zheng 173476dda93cSYan, Zheng path = btrfs_alloc_path(); 173576dda93cSYan, Zheng if (!path) 173676dda93cSYan, Zheng return -ENOMEM; 173776dda93cSYan, Zheng 1738175a2b87SJosef Bacik /* Make sure this root isn't set as the default subvol */ 1739175a2b87SJosef Bacik dir_id = btrfs_super_root_dir(root->fs_info->super_copy); 1740175a2b87SJosef Bacik di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path, 1741175a2b87SJosef Bacik dir_id, "default", 7, 0); 1742175a2b87SJosef Bacik if (di && !IS_ERR(di)) { 1743175a2b87SJosef Bacik btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key); 1744175a2b87SJosef Bacik if (key.objectid == root->root_key.objectid) { 1745175a2b87SJosef Bacik ret = -ENOTEMPTY; 1746175a2b87SJosef Bacik goto out; 1747175a2b87SJosef Bacik } 1748175a2b87SJosef Bacik btrfs_release_path(path); 1749175a2b87SJosef Bacik } 1750175a2b87SJosef Bacik 175176dda93cSYan, Zheng key.objectid = root->root_key.objectid; 175276dda93cSYan, Zheng key.type = BTRFS_ROOT_REF_KEY; 175376dda93cSYan, Zheng key.offset = (u64)-1; 175476dda93cSYan, Zheng 175576dda93cSYan, Zheng ret = btrfs_search_slot(NULL, root->fs_info->tree_root, 175676dda93cSYan, Zheng &key, path, 0, 0); 175776dda93cSYan, Zheng if (ret < 0) 175876dda93cSYan, Zheng goto out; 175976dda93cSYan, Zheng BUG_ON(ret == 0); 176076dda93cSYan, Zheng 176176dda93cSYan, Zheng ret = 0; 176276dda93cSYan, Zheng if (path->slots[0] > 0) { 176376dda93cSYan, Zheng path->slots[0]--; 176476dda93cSYan, Zheng btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 176576dda93cSYan, Zheng if (key.objectid == root->root_key.objectid && 176676dda93cSYan, Zheng key.type == BTRFS_ROOT_REF_KEY) 176776dda93cSYan, Zheng ret = -ENOTEMPTY; 176876dda93cSYan, Zheng } 176976dda93cSYan, Zheng out: 177076dda93cSYan, Zheng btrfs_free_path(path); 177176dda93cSYan, Zheng return ret; 177276dda93cSYan, Zheng } 177376dda93cSYan, Zheng 1774ac8e9819SChris Mason static noinline int key_in_sk(struct btrfs_key *key, 1775ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk) 1776ac8e9819SChris Mason { 1777abc6e134SChris Mason struct btrfs_key test; 1778abc6e134SChris Mason int ret; 1779abc6e134SChris Mason 1780abc6e134SChris Mason test.objectid = sk->min_objectid; 1781abc6e134SChris Mason test.type = sk->min_type; 1782abc6e134SChris Mason test.offset = sk->min_offset; 1783abc6e134SChris Mason 1784abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1785abc6e134SChris Mason if (ret < 0) 1786ac8e9819SChris Mason return 0; 1787abc6e134SChris Mason 1788abc6e134SChris Mason test.objectid = sk->max_objectid; 1789abc6e134SChris Mason test.type = sk->max_type; 1790abc6e134SChris Mason test.offset = sk->max_offset; 1791abc6e134SChris Mason 1792abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1793abc6e134SChris Mason if (ret > 0) 1794ac8e9819SChris Mason return 0; 1795ac8e9819SChris Mason return 1; 1796ac8e9819SChris Mason } 1797ac8e9819SChris Mason 1798ac8e9819SChris Mason static noinline int copy_to_sk(struct btrfs_root *root, 1799ac8e9819SChris Mason struct btrfs_path *path, 1800ac8e9819SChris Mason struct btrfs_key *key, 1801ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk, 1802ac8e9819SChris Mason char *buf, 1803ac8e9819SChris Mason unsigned long *sk_offset, 1804ac8e9819SChris Mason int *num_found) 1805ac8e9819SChris Mason { 1806ac8e9819SChris Mason u64 found_transid; 1807ac8e9819SChris Mason struct extent_buffer *leaf; 1808ac8e9819SChris Mason struct btrfs_ioctl_search_header sh; 1809ac8e9819SChris Mason unsigned long item_off; 1810ac8e9819SChris Mason unsigned long item_len; 1811ac8e9819SChris Mason int nritems; 1812ac8e9819SChris Mason int i; 1813ac8e9819SChris Mason int slot; 1814ac8e9819SChris Mason int ret = 0; 1815ac8e9819SChris Mason 1816ac8e9819SChris Mason leaf = path->nodes[0]; 1817ac8e9819SChris Mason slot = path->slots[0]; 1818ac8e9819SChris Mason nritems = btrfs_header_nritems(leaf); 1819ac8e9819SChris Mason 1820ac8e9819SChris Mason if (btrfs_header_generation(leaf) > sk->max_transid) { 1821ac8e9819SChris Mason i = nritems; 1822ac8e9819SChris Mason goto advance_key; 1823ac8e9819SChris Mason } 1824ac8e9819SChris Mason found_transid = btrfs_header_generation(leaf); 1825ac8e9819SChris Mason 1826ac8e9819SChris Mason for (i = slot; i < nritems; i++) { 1827ac8e9819SChris Mason item_off = btrfs_item_ptr_offset(leaf, i); 1828ac8e9819SChris Mason item_len = btrfs_item_size_nr(leaf, i); 1829ac8e9819SChris Mason 183003b71c6cSGabriel de Perthuis btrfs_item_key_to_cpu(leaf, key, i); 183103b71c6cSGabriel de Perthuis if (!key_in_sk(key, sk)) 183203b71c6cSGabriel de Perthuis continue; 183303b71c6cSGabriel de Perthuis 183403b71c6cSGabriel de Perthuis if (sizeof(sh) + item_len > BTRFS_SEARCH_ARGS_BUFSIZE) 1835ac8e9819SChris Mason item_len = 0; 1836ac8e9819SChris Mason 1837ac8e9819SChris Mason if (sizeof(sh) + item_len + *sk_offset > 1838ac8e9819SChris Mason BTRFS_SEARCH_ARGS_BUFSIZE) { 1839ac8e9819SChris Mason ret = 1; 1840ac8e9819SChris Mason goto overflow; 1841ac8e9819SChris Mason } 1842ac8e9819SChris Mason 1843ac8e9819SChris Mason sh.objectid = key->objectid; 1844ac8e9819SChris Mason sh.offset = key->offset; 1845ac8e9819SChris Mason sh.type = key->type; 1846ac8e9819SChris Mason sh.len = item_len; 1847ac8e9819SChris Mason sh.transid = found_transid; 1848ac8e9819SChris Mason 1849ac8e9819SChris Mason /* copy search result header */ 1850ac8e9819SChris Mason memcpy(buf + *sk_offset, &sh, sizeof(sh)); 1851ac8e9819SChris Mason *sk_offset += sizeof(sh); 1852ac8e9819SChris Mason 1853ac8e9819SChris Mason if (item_len) { 1854ac8e9819SChris Mason char *p = buf + *sk_offset; 1855ac8e9819SChris Mason /* copy the item */ 1856ac8e9819SChris Mason read_extent_buffer(leaf, p, 1857ac8e9819SChris Mason item_off, item_len); 1858ac8e9819SChris Mason *sk_offset += item_len; 1859ac8e9819SChris Mason } 1860e2156867SHugo Mills (*num_found)++; 1861ac8e9819SChris Mason 1862ac8e9819SChris Mason if (*num_found >= sk->nr_items) 1863ac8e9819SChris Mason break; 1864ac8e9819SChris Mason } 1865ac8e9819SChris Mason advance_key: 1866ac8e9819SChris Mason ret = 0; 1867abc6e134SChris Mason if (key->offset < (u64)-1 && key->offset < sk->max_offset) 1868abc6e134SChris Mason key->offset++; 1869abc6e134SChris Mason else if (key->type < (u8)-1 && key->type < sk->max_type) { 1870abc6e134SChris Mason key->offset = 0; 1871abc6e134SChris Mason key->type++; 1872abc6e134SChris Mason } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) { 1873abc6e134SChris Mason key->offset = 0; 1874abc6e134SChris Mason key->type = 0; 1875abc6e134SChris Mason key->objectid++; 1876abc6e134SChris Mason } else 1877abc6e134SChris Mason ret = 1; 1878ac8e9819SChris Mason overflow: 1879ac8e9819SChris Mason return ret; 1880ac8e9819SChris Mason } 1881ac8e9819SChris Mason 1882ac8e9819SChris Mason static noinline int search_ioctl(struct inode *inode, 1883ac8e9819SChris Mason struct btrfs_ioctl_search_args *args) 1884ac8e9819SChris Mason { 1885ac8e9819SChris Mason struct btrfs_root *root; 1886ac8e9819SChris Mason struct btrfs_key key; 1887ac8e9819SChris Mason struct btrfs_key max_key; 1888ac8e9819SChris Mason struct btrfs_path *path; 1889ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk = &args->key; 1890ac8e9819SChris Mason struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info; 1891ac8e9819SChris Mason int ret; 1892ac8e9819SChris Mason int num_found = 0; 1893ac8e9819SChris Mason unsigned long sk_offset = 0; 1894ac8e9819SChris Mason 1895ac8e9819SChris Mason path = btrfs_alloc_path(); 1896ac8e9819SChris Mason if (!path) 1897ac8e9819SChris Mason return -ENOMEM; 1898ac8e9819SChris Mason 1899ac8e9819SChris Mason if (sk->tree_id == 0) { 1900ac8e9819SChris Mason /* search the root of the inode that was passed */ 1901ac8e9819SChris Mason root = BTRFS_I(inode)->root; 1902ac8e9819SChris Mason } else { 1903ac8e9819SChris Mason key.objectid = sk->tree_id; 1904ac8e9819SChris Mason key.type = BTRFS_ROOT_ITEM_KEY; 1905ac8e9819SChris Mason key.offset = (u64)-1; 1906ac8e9819SChris Mason root = btrfs_read_fs_root_no_name(info, &key); 1907ac8e9819SChris Mason if (IS_ERR(root)) { 1908ac8e9819SChris Mason printk(KERN_ERR "could not find root %llu\n", 1909ac8e9819SChris Mason sk->tree_id); 1910ac8e9819SChris Mason btrfs_free_path(path); 1911ac8e9819SChris Mason return -ENOENT; 1912ac8e9819SChris Mason } 1913ac8e9819SChris Mason } 1914ac8e9819SChris Mason 1915ac8e9819SChris Mason key.objectid = sk->min_objectid; 1916ac8e9819SChris Mason key.type = sk->min_type; 1917ac8e9819SChris Mason key.offset = sk->min_offset; 1918ac8e9819SChris Mason 1919ac8e9819SChris Mason max_key.objectid = sk->max_objectid; 1920ac8e9819SChris Mason max_key.type = sk->max_type; 1921ac8e9819SChris Mason max_key.offset = sk->max_offset; 1922ac8e9819SChris Mason 1923ac8e9819SChris Mason path->keep_locks = 1; 1924ac8e9819SChris Mason 1925ac8e9819SChris Mason while(1) { 1926de78b51aSEric Sandeen ret = btrfs_search_forward(root, &key, &max_key, path, 1927ac8e9819SChris Mason sk->min_transid); 1928ac8e9819SChris Mason if (ret != 0) { 1929ac8e9819SChris Mason if (ret > 0) 1930ac8e9819SChris Mason ret = 0; 1931ac8e9819SChris Mason goto err; 1932ac8e9819SChris Mason } 1933ac8e9819SChris Mason ret = copy_to_sk(root, path, &key, sk, args->buf, 1934ac8e9819SChris Mason &sk_offset, &num_found); 1935b3b4aa74SDavid Sterba btrfs_release_path(path); 1936ac8e9819SChris Mason if (ret || num_found >= sk->nr_items) 1937ac8e9819SChris Mason break; 1938ac8e9819SChris Mason 1939ac8e9819SChris Mason } 1940ac8e9819SChris Mason ret = 0; 1941ac8e9819SChris Mason err: 1942ac8e9819SChris Mason sk->nr_items = num_found; 1943ac8e9819SChris Mason btrfs_free_path(path); 1944ac8e9819SChris Mason return ret; 1945ac8e9819SChris Mason } 1946ac8e9819SChris Mason 1947ac8e9819SChris Mason static noinline int btrfs_ioctl_tree_search(struct file *file, 1948ac8e9819SChris Mason void __user *argp) 1949ac8e9819SChris Mason { 1950ac8e9819SChris Mason struct btrfs_ioctl_search_args *args; 1951ac8e9819SChris Mason struct inode *inode; 1952ac8e9819SChris Mason int ret; 1953ac8e9819SChris Mason 1954ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 1955ac8e9819SChris Mason return -EPERM; 1956ac8e9819SChris Mason 19572354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 19582354d08fSJulia Lawall if (IS_ERR(args)) 19592354d08fSJulia Lawall return PTR_ERR(args); 1960ac8e9819SChris Mason 1961496ad9aaSAl Viro inode = file_inode(file); 1962ac8e9819SChris Mason ret = search_ioctl(inode, args); 1963ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 1964ac8e9819SChris Mason ret = -EFAULT; 1965ac8e9819SChris Mason kfree(args); 1966ac8e9819SChris Mason return ret; 1967ac8e9819SChris Mason } 1968ac8e9819SChris Mason 196998d377a0STARUISI Hiroaki /* 1970ac8e9819SChris Mason * Search INODE_REFs to identify path name of 'dirid' directory 1971ac8e9819SChris Mason * in a 'tree_id' tree. and sets path name to 'name'. 197298d377a0STARUISI Hiroaki */ 197398d377a0STARUISI Hiroaki static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, 197498d377a0STARUISI Hiroaki u64 tree_id, u64 dirid, char *name) 197598d377a0STARUISI Hiroaki { 197698d377a0STARUISI Hiroaki struct btrfs_root *root; 197798d377a0STARUISI Hiroaki struct btrfs_key key; 1978ac8e9819SChris Mason char *ptr; 197998d377a0STARUISI Hiroaki int ret = -1; 198098d377a0STARUISI Hiroaki int slot; 198198d377a0STARUISI Hiroaki int len; 198298d377a0STARUISI Hiroaki int total_len = 0; 198398d377a0STARUISI Hiroaki struct btrfs_inode_ref *iref; 198498d377a0STARUISI Hiroaki struct extent_buffer *l; 198598d377a0STARUISI Hiroaki struct btrfs_path *path; 198698d377a0STARUISI Hiroaki 198798d377a0STARUISI Hiroaki if (dirid == BTRFS_FIRST_FREE_OBJECTID) { 198898d377a0STARUISI Hiroaki name[0]='\0'; 198998d377a0STARUISI Hiroaki return 0; 199098d377a0STARUISI Hiroaki } 199198d377a0STARUISI Hiroaki 199298d377a0STARUISI Hiroaki path = btrfs_alloc_path(); 199398d377a0STARUISI Hiroaki if (!path) 199498d377a0STARUISI Hiroaki return -ENOMEM; 199598d377a0STARUISI Hiroaki 1996ac8e9819SChris Mason ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX]; 199798d377a0STARUISI Hiroaki 199898d377a0STARUISI Hiroaki key.objectid = tree_id; 199998d377a0STARUISI Hiroaki key.type = BTRFS_ROOT_ITEM_KEY; 200098d377a0STARUISI Hiroaki key.offset = (u64)-1; 200198d377a0STARUISI Hiroaki root = btrfs_read_fs_root_no_name(info, &key); 200298d377a0STARUISI Hiroaki if (IS_ERR(root)) { 200398d377a0STARUISI Hiroaki printk(KERN_ERR "could not find root %llu\n", tree_id); 20048ad6fcabSChris Mason ret = -ENOENT; 20058ad6fcabSChris Mason goto out; 200698d377a0STARUISI Hiroaki } 200798d377a0STARUISI Hiroaki 200898d377a0STARUISI Hiroaki key.objectid = dirid; 200998d377a0STARUISI Hiroaki key.type = BTRFS_INODE_REF_KEY; 20108ad6fcabSChris Mason key.offset = (u64)-1; 201198d377a0STARUISI Hiroaki 201298d377a0STARUISI Hiroaki while(1) { 201398d377a0STARUISI Hiroaki ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 201498d377a0STARUISI Hiroaki if (ret < 0) 201598d377a0STARUISI Hiroaki goto out; 2016*18674c6cSFilipe David Borba Manana else if (ret > 0) { 2017*18674c6cSFilipe David Borba Manana ret = btrfs_previous_item(root, path, dirid, 2018*18674c6cSFilipe David Borba Manana BTRFS_INODE_REF_KEY); 2019*18674c6cSFilipe David Borba Manana if (ret < 0) 2020*18674c6cSFilipe David Borba Manana goto out; 2021*18674c6cSFilipe David Borba Manana else if (ret > 0) { 2022ac8e9819SChris Mason ret = -ENOENT; 202398d377a0STARUISI Hiroaki goto out; 2024ac8e9819SChris Mason } 2025*18674c6cSFilipe David Borba Manana } 2026*18674c6cSFilipe David Borba Manana 2027*18674c6cSFilipe David Borba Manana l = path->nodes[0]; 2028*18674c6cSFilipe David Borba Manana slot = path->slots[0]; 2029*18674c6cSFilipe David Borba Manana btrfs_item_key_to_cpu(l, &key, slot); 203098d377a0STARUISI Hiroaki 203198d377a0STARUISI Hiroaki iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); 203298d377a0STARUISI Hiroaki len = btrfs_inode_ref_name_len(l, iref); 203398d377a0STARUISI Hiroaki ptr -= len + 1; 203498d377a0STARUISI Hiroaki total_len += len + 1; 2035a696cf35SFilipe David Borba Manana if (ptr < name) { 2036a696cf35SFilipe David Borba Manana ret = -ENAMETOOLONG; 203798d377a0STARUISI Hiroaki goto out; 2038a696cf35SFilipe David Borba Manana } 203998d377a0STARUISI Hiroaki 204098d377a0STARUISI Hiroaki *(ptr + len) = '/'; 204198d377a0STARUISI Hiroaki read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len); 204298d377a0STARUISI Hiroaki 204398d377a0STARUISI Hiroaki if (key.offset == BTRFS_FIRST_FREE_OBJECTID) 204498d377a0STARUISI Hiroaki break; 204598d377a0STARUISI Hiroaki 2046b3b4aa74SDavid Sterba btrfs_release_path(path); 204798d377a0STARUISI Hiroaki key.objectid = key.offset; 20488ad6fcabSChris Mason key.offset = (u64)-1; 204998d377a0STARUISI Hiroaki dirid = key.objectid; 205098d377a0STARUISI Hiroaki } 205177906a50SLi Zefan memmove(name, ptr, total_len); 205298d377a0STARUISI Hiroaki name[total_len]='\0'; 205398d377a0STARUISI Hiroaki ret = 0; 205498d377a0STARUISI Hiroaki out: 205598d377a0STARUISI Hiroaki btrfs_free_path(path); 2056ac8e9819SChris Mason return ret; 2057ac8e9819SChris Mason } 2058ac8e9819SChris Mason 2059ac8e9819SChris Mason static noinline int btrfs_ioctl_ino_lookup(struct file *file, 2060ac8e9819SChris Mason void __user *argp) 2061ac8e9819SChris Mason { 2062ac8e9819SChris Mason struct btrfs_ioctl_ino_lookup_args *args; 2063ac8e9819SChris Mason struct inode *inode; 2064ac8e9819SChris Mason int ret; 2065ac8e9819SChris Mason 2066ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 2067ac8e9819SChris Mason return -EPERM; 2068ac8e9819SChris Mason 20692354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 20702354d08fSJulia Lawall if (IS_ERR(args)) 20712354d08fSJulia Lawall return PTR_ERR(args); 2072c2b96929SDan Carpenter 2073496ad9aaSAl Viro inode = file_inode(file); 2074ac8e9819SChris Mason 20751b53ac4dSChris Mason if (args->treeid == 0) 20761b53ac4dSChris Mason args->treeid = BTRFS_I(inode)->root->root_key.objectid; 20771b53ac4dSChris Mason 2078ac8e9819SChris Mason ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, 2079ac8e9819SChris Mason args->treeid, args->objectid, 2080ac8e9819SChris Mason args->name); 2081ac8e9819SChris Mason 2082ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 2083ac8e9819SChris Mason ret = -EFAULT; 2084ac8e9819SChris Mason 2085ac8e9819SChris Mason kfree(args); 208698d377a0STARUISI Hiroaki return ret; 208798d377a0STARUISI Hiroaki } 208898d377a0STARUISI Hiroaki 208976dda93cSYan, Zheng static noinline int btrfs_ioctl_snap_destroy(struct file *file, 209076dda93cSYan, Zheng void __user *arg) 209176dda93cSYan, Zheng { 209276dda93cSYan, Zheng struct dentry *parent = fdentry(file); 209376dda93cSYan, Zheng struct dentry *dentry; 209476dda93cSYan, Zheng struct inode *dir = parent->d_inode; 209576dda93cSYan, Zheng struct inode *inode; 209676dda93cSYan, Zheng struct btrfs_root *root = BTRFS_I(dir)->root; 209776dda93cSYan, Zheng struct btrfs_root *dest = NULL; 209876dda93cSYan, Zheng struct btrfs_ioctl_vol_args *vol_args; 209976dda93cSYan, Zheng struct btrfs_trans_handle *trans; 2100c58aaad2SMiao Xie struct btrfs_block_rsv block_rsv; 2101c58aaad2SMiao Xie u64 qgroup_reserved; 210276dda93cSYan, Zheng int namelen; 210376dda93cSYan, Zheng int ret; 210476dda93cSYan, Zheng int err = 0; 210576dda93cSYan, Zheng 210676dda93cSYan, Zheng vol_args = memdup_user(arg, sizeof(*vol_args)); 210776dda93cSYan, Zheng if (IS_ERR(vol_args)) 210876dda93cSYan, Zheng return PTR_ERR(vol_args); 210976dda93cSYan, Zheng 211076dda93cSYan, Zheng vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 211176dda93cSYan, Zheng namelen = strlen(vol_args->name); 211276dda93cSYan, Zheng if (strchr(vol_args->name, '/') || 211376dda93cSYan, Zheng strncmp(vol_args->name, "..", namelen) == 0) { 211476dda93cSYan, Zheng err = -EINVAL; 211576dda93cSYan, Zheng goto out; 211676dda93cSYan, Zheng } 211776dda93cSYan, Zheng 2118a561be71SAl Viro err = mnt_want_write_file(file); 211976dda93cSYan, Zheng if (err) 212076dda93cSYan, Zheng goto out; 212176dda93cSYan, Zheng 21225c50c9b8SDavid Sterba err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT); 21235c50c9b8SDavid Sterba if (err == -EINTR) 21245c50c9b8SDavid Sterba goto out; 212576dda93cSYan, Zheng dentry = lookup_one_len(vol_args->name, parent, namelen); 212676dda93cSYan, Zheng if (IS_ERR(dentry)) { 212776dda93cSYan, Zheng err = PTR_ERR(dentry); 212876dda93cSYan, Zheng goto out_unlock_dir; 212976dda93cSYan, Zheng } 213076dda93cSYan, Zheng 213176dda93cSYan, Zheng if (!dentry->d_inode) { 213276dda93cSYan, Zheng err = -ENOENT; 213376dda93cSYan, Zheng goto out_dput; 213476dda93cSYan, Zheng } 213576dda93cSYan, Zheng 213676dda93cSYan, Zheng inode = dentry->d_inode; 21374260f7c7SSage Weil dest = BTRFS_I(inode)->root; 21384260f7c7SSage Weil if (!capable(CAP_SYS_ADMIN)){ 21394260f7c7SSage Weil /* 21404260f7c7SSage Weil * Regular user. Only allow this with a special mount 21414260f7c7SSage Weil * option, when the user has write+exec access to the 21424260f7c7SSage Weil * subvol root, and when rmdir(2) would have been 21434260f7c7SSage Weil * allowed. 21444260f7c7SSage Weil * 21454260f7c7SSage Weil * Note that this is _not_ check that the subvol is 21464260f7c7SSage Weil * empty or doesn't contain data that we wouldn't 21474260f7c7SSage Weil * otherwise be able to delete. 21484260f7c7SSage Weil * 21494260f7c7SSage Weil * Users who want to delete empty subvols should try 21504260f7c7SSage Weil * rmdir(2). 21514260f7c7SSage Weil */ 21524260f7c7SSage Weil err = -EPERM; 21534260f7c7SSage Weil if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED)) 21544260f7c7SSage Weil goto out_dput; 21554260f7c7SSage Weil 21564260f7c7SSage Weil /* 21574260f7c7SSage Weil * Do not allow deletion if the parent dir is the same 21584260f7c7SSage Weil * as the dir to be deleted. That means the ioctl 21594260f7c7SSage Weil * must be called on the dentry referencing the root 21604260f7c7SSage Weil * of the subvol, not a random directory contained 21614260f7c7SSage Weil * within it. 21624260f7c7SSage Weil */ 21634260f7c7SSage Weil err = -EINVAL; 21644260f7c7SSage Weil if (root == dest) 21654260f7c7SSage Weil goto out_dput; 21664260f7c7SSage Weil 21674260f7c7SSage Weil err = inode_permission(inode, MAY_WRITE | MAY_EXEC); 21684260f7c7SSage Weil if (err) 21694260f7c7SSage Weil goto out_dput; 21705c39da5bSMiao Xie } 21714260f7c7SSage Weil 21725c39da5bSMiao Xie /* check if subvolume may be deleted by a user */ 21734260f7c7SSage Weil err = btrfs_may_delete(dir, dentry, 1); 21744260f7c7SSage Weil if (err) 21754260f7c7SSage Weil goto out_dput; 21764260f7c7SSage Weil 217733345d01SLi Zefan if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { 217876dda93cSYan, Zheng err = -EINVAL; 217976dda93cSYan, Zheng goto out_dput; 218076dda93cSYan, Zheng } 218176dda93cSYan, Zheng 218276dda93cSYan, Zheng mutex_lock(&inode->i_mutex); 218376dda93cSYan, Zheng err = d_invalidate(dentry); 218476dda93cSYan, Zheng if (err) 218576dda93cSYan, Zheng goto out_unlock; 218676dda93cSYan, Zheng 218776dda93cSYan, Zheng down_write(&root->fs_info->subvol_sem); 218876dda93cSYan, Zheng 218976dda93cSYan, Zheng err = may_destroy_subvol(dest); 219076dda93cSYan, Zheng if (err) 219176dda93cSYan, Zheng goto out_up_write; 219276dda93cSYan, Zheng 2193c58aaad2SMiao Xie btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); 2194c58aaad2SMiao Xie /* 2195c58aaad2SMiao Xie * One for dir inode, two for dir entries, two for root 2196c58aaad2SMiao Xie * ref/backref. 2197c58aaad2SMiao Xie */ 2198c58aaad2SMiao Xie err = btrfs_subvolume_reserve_metadata(root, &block_rsv, 2199ee3441b4SJeff Mahoney 5, &qgroup_reserved, true); 2200c58aaad2SMiao Xie if (err) 2201c58aaad2SMiao Xie goto out_up_write; 2202c58aaad2SMiao Xie 2203a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 2204a22285a6SYan, Zheng if (IS_ERR(trans)) { 2205a22285a6SYan, Zheng err = PTR_ERR(trans); 2206c58aaad2SMiao Xie goto out_release; 2207a22285a6SYan, Zheng } 2208c58aaad2SMiao Xie trans->block_rsv = &block_rsv; 2209c58aaad2SMiao Xie trans->bytes_reserved = block_rsv.size; 2210a22285a6SYan, Zheng 221176dda93cSYan, Zheng ret = btrfs_unlink_subvol(trans, root, dir, 221276dda93cSYan, Zheng dest->root_key.objectid, 221376dda93cSYan, Zheng dentry->d_name.name, 221476dda93cSYan, Zheng dentry->d_name.len); 221579787eaaSJeff Mahoney if (ret) { 221679787eaaSJeff Mahoney err = ret; 221779787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, ret); 221879787eaaSJeff Mahoney goto out_end_trans; 221979787eaaSJeff Mahoney } 222076dda93cSYan, Zheng 222176dda93cSYan, Zheng btrfs_record_root_in_trans(trans, dest); 222276dda93cSYan, Zheng 222376dda93cSYan, Zheng memset(&dest->root_item.drop_progress, 0, 222476dda93cSYan, Zheng sizeof(dest->root_item.drop_progress)); 222576dda93cSYan, Zheng dest->root_item.drop_level = 0; 222676dda93cSYan, Zheng btrfs_set_root_refs(&dest->root_item, 0); 222776dda93cSYan, Zheng 2228d68fc57bSYan, Zheng if (!xchg(&dest->orphan_item_inserted, 1)) { 222976dda93cSYan, Zheng ret = btrfs_insert_orphan_item(trans, 223076dda93cSYan, Zheng root->fs_info->tree_root, 223176dda93cSYan, Zheng dest->root_key.objectid); 223279787eaaSJeff Mahoney if (ret) { 223379787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, ret); 223479787eaaSJeff Mahoney err = ret; 223579787eaaSJeff Mahoney goto out_end_trans; 2236d68fc57bSYan, Zheng } 223779787eaaSJeff Mahoney } 223879787eaaSJeff Mahoney out_end_trans: 2239c58aaad2SMiao Xie trans->block_rsv = NULL; 2240c58aaad2SMiao Xie trans->bytes_reserved = 0; 2241531cb13fSSage Weil ret = btrfs_end_transaction(trans, root); 224279787eaaSJeff Mahoney if (ret && !err) 224379787eaaSJeff Mahoney err = ret; 224476dda93cSYan, Zheng inode->i_flags |= S_DEAD; 2245c58aaad2SMiao Xie out_release: 2246c58aaad2SMiao Xie btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved); 224776dda93cSYan, Zheng out_up_write: 224876dda93cSYan, Zheng up_write(&root->fs_info->subvol_sem); 224976dda93cSYan, Zheng out_unlock: 225076dda93cSYan, Zheng mutex_unlock(&inode->i_mutex); 225176dda93cSYan, Zheng if (!err) { 2252efefb143SYan, Zheng shrink_dcache_sb(root->fs_info->sb); 225376dda93cSYan, Zheng btrfs_invalidate_inodes(dest); 225476dda93cSYan, Zheng d_delete(dentry); 2255fa6ac876SLiu Bo 2256fa6ac876SLiu Bo /* the last ref */ 2257fa6ac876SLiu Bo if (dest->cache_inode) { 2258fa6ac876SLiu Bo iput(dest->cache_inode); 2259fa6ac876SLiu Bo dest->cache_inode = NULL; 2260fa6ac876SLiu Bo } 226176dda93cSYan, Zheng } 226276dda93cSYan, Zheng out_dput: 226376dda93cSYan, Zheng dput(dentry); 226476dda93cSYan, Zheng out_unlock_dir: 226576dda93cSYan, Zheng mutex_unlock(&dir->i_mutex); 22662a79f17eSAl Viro mnt_drop_write_file(file); 226776dda93cSYan, Zheng out: 226876dda93cSYan, Zheng kfree(vol_args); 226976dda93cSYan, Zheng return err; 227076dda93cSYan, Zheng } 227176dda93cSYan, Zheng 22721e701a32SChris Mason static int btrfs_ioctl_defrag(struct file *file, void __user *argp) 2273f46b5a66SChristoph Hellwig { 2274496ad9aaSAl Viro struct inode *inode = file_inode(file); 2275f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 22761e701a32SChris Mason struct btrfs_ioctl_defrag_range_args *range; 2277c146afadSYan Zheng int ret; 2278c146afadSYan Zheng 227925122d15SIlya Dryomov ret = mnt_want_write_file(file); 228025122d15SIlya Dryomov if (ret) 228125122d15SIlya Dryomov return ret; 2282b83cc969SLi Zefan 228325122d15SIlya Dryomov if (btrfs_root_readonly(root)) { 228425122d15SIlya Dryomov ret = -EROFS; 228525122d15SIlya Dryomov goto out; 22865ac00addSStefan Behrens } 2287f46b5a66SChristoph Hellwig 2288f46b5a66SChristoph Hellwig switch (inode->i_mode & S_IFMT) { 2289f46b5a66SChristoph Hellwig case S_IFDIR: 2290e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) { 2291e441d54dSChris Mason ret = -EPERM; 2292e441d54dSChris Mason goto out; 2293e441d54dSChris Mason } 2294de78b51aSEric Sandeen ret = btrfs_defrag_root(root); 22958929ecfaSYan, Zheng if (ret) 22968929ecfaSYan, Zheng goto out; 2297de78b51aSEric Sandeen ret = btrfs_defrag_root(root->fs_info->extent_root); 2298f46b5a66SChristoph Hellwig break; 2299f46b5a66SChristoph Hellwig case S_IFREG: 2300e441d54dSChris Mason if (!(file->f_mode & FMODE_WRITE)) { 2301e441d54dSChris Mason ret = -EINVAL; 2302e441d54dSChris Mason goto out; 2303e441d54dSChris Mason } 23041e701a32SChris Mason 23051e701a32SChris Mason range = kzalloc(sizeof(*range), GFP_KERNEL); 23061e701a32SChris Mason if (!range) { 23071e701a32SChris Mason ret = -ENOMEM; 23081e701a32SChris Mason goto out; 23091e701a32SChris Mason } 23101e701a32SChris Mason 23111e701a32SChris Mason if (argp) { 23121e701a32SChris Mason if (copy_from_user(range, argp, 23131e701a32SChris Mason sizeof(*range))) { 23141e701a32SChris Mason ret = -EFAULT; 23151e701a32SChris Mason kfree(range); 2316683be16eSDan Carpenter goto out; 23171e701a32SChris Mason } 23181e701a32SChris Mason /* compression requires us to start the IO */ 23191e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 23201e701a32SChris Mason range->flags |= BTRFS_DEFRAG_RANGE_START_IO; 23211e701a32SChris Mason range->extent_thresh = (u32)-1; 23221e701a32SChris Mason } 23231e701a32SChris Mason } else { 23241e701a32SChris Mason /* the rest are all set to zero by kzalloc */ 23251e701a32SChris Mason range->len = (u64)-1; 23261e701a32SChris Mason } 2327496ad9aaSAl Viro ret = btrfs_defrag_file(file_inode(file), file, 23284cb5300bSChris Mason range, 0, 0); 23294cb5300bSChris Mason if (ret > 0) 23304cb5300bSChris Mason ret = 0; 23311e701a32SChris Mason kfree(range); 2332f46b5a66SChristoph Hellwig break; 23338929ecfaSYan, Zheng default: 23348929ecfaSYan, Zheng ret = -EINVAL; 2335f46b5a66SChristoph Hellwig } 2336e441d54dSChris Mason out: 233725122d15SIlya Dryomov mnt_drop_write_file(file); 2338e441d54dSChris Mason return ret; 2339f46b5a66SChristoph Hellwig } 2340f46b5a66SChristoph Hellwig 2341b2950863SChristoph Hellwig static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg) 2342f46b5a66SChristoph Hellwig { 2343f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2344f46b5a66SChristoph Hellwig int ret; 2345f46b5a66SChristoph Hellwig 2346e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2347e441d54dSChris Mason return -EPERM; 2348e441d54dSChris Mason 23495ac00addSStefan Behrens if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running, 23505ac00addSStefan Behrens 1)) { 23515ac00addSStefan Behrens pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n"); 23522c0c9da0SIlya Dryomov return -EINVAL; 2353c9e9f97bSIlya Dryomov } 2354c9e9f97bSIlya Dryomov 23555ac00addSStefan Behrens mutex_lock(&root->fs_info->volume_mutex); 2356dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2357c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 2358c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 2359c9e9f97bSIlya Dryomov goto out; 2360c9e9f97bSIlya Dryomov } 2361f46b5a66SChristoph Hellwig 23625516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 2363f46b5a66SChristoph Hellwig ret = btrfs_init_new_device(root, vol_args->name); 2364f46b5a66SChristoph Hellwig 2365f46b5a66SChristoph Hellwig kfree(vol_args); 2366c9e9f97bSIlya Dryomov out: 2367c9e9f97bSIlya Dryomov mutex_unlock(&root->fs_info->volume_mutex); 23685ac00addSStefan Behrens atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0); 2369f46b5a66SChristoph Hellwig return ret; 2370f46b5a66SChristoph Hellwig } 2371f46b5a66SChristoph Hellwig 2372da24927bSMiao Xie static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg) 2373f46b5a66SChristoph Hellwig { 2374496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 2375f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2376f46b5a66SChristoph Hellwig int ret; 2377f46b5a66SChristoph Hellwig 2378e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2379e441d54dSChris Mason return -EPERM; 2380e441d54dSChris Mason 2381da24927bSMiao Xie ret = mnt_want_write_file(file); 2382da24927bSMiao Xie if (ret) 2383da24927bSMiao Xie return ret; 2384c146afadSYan Zheng 2385dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2386c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 2387c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 2388c9e9f97bSIlya Dryomov goto out; 2389c9e9f97bSIlya Dryomov } 2390f46b5a66SChristoph Hellwig 23915516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 2392f46b5a66SChristoph Hellwig 2393183860f6SAnand Jain if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running, 2394183860f6SAnand Jain 1)) { 2395183860f6SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 2396183860f6SAnand Jain goto out; 2397183860f6SAnand Jain } 2398183860f6SAnand Jain 2399183860f6SAnand Jain mutex_lock(&root->fs_info->volume_mutex); 2400183860f6SAnand Jain ret = btrfs_rm_device(root, vol_args->name); 2401c9e9f97bSIlya Dryomov mutex_unlock(&root->fs_info->volume_mutex); 24025ac00addSStefan Behrens atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0); 2403183860f6SAnand Jain 2404183860f6SAnand Jain out: 2405183860f6SAnand Jain kfree(vol_args); 24064ac20c70SIlya Dryomov mnt_drop_write_file(file); 2407f46b5a66SChristoph Hellwig return ret; 2408f46b5a66SChristoph Hellwig } 2409f46b5a66SChristoph Hellwig 2410475f6387SJan Schmidt static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg) 2411475f6387SJan Schmidt { 2412027ed2f0SLi Zefan struct btrfs_ioctl_fs_info_args *fi_args; 2413475f6387SJan Schmidt struct btrfs_device *device; 2414475f6387SJan Schmidt struct btrfs_device *next; 2415475f6387SJan Schmidt struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; 2416027ed2f0SLi Zefan int ret = 0; 2417475f6387SJan Schmidt 2418475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 2419475f6387SJan Schmidt return -EPERM; 2420475f6387SJan Schmidt 2421027ed2f0SLi Zefan fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL); 2422027ed2f0SLi Zefan if (!fi_args) 2423027ed2f0SLi Zefan return -ENOMEM; 2424027ed2f0SLi Zefan 2425027ed2f0SLi Zefan fi_args->num_devices = fs_devices->num_devices; 2426027ed2f0SLi Zefan memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid)); 2427475f6387SJan Schmidt 2428475f6387SJan Schmidt mutex_lock(&fs_devices->device_list_mutex); 2429475f6387SJan Schmidt list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) { 2430027ed2f0SLi Zefan if (device->devid > fi_args->max_id) 2431027ed2f0SLi Zefan fi_args->max_id = device->devid; 2432475f6387SJan Schmidt } 2433475f6387SJan Schmidt mutex_unlock(&fs_devices->device_list_mutex); 2434475f6387SJan Schmidt 2435027ed2f0SLi Zefan if (copy_to_user(arg, fi_args, sizeof(*fi_args))) 2436027ed2f0SLi Zefan ret = -EFAULT; 2437475f6387SJan Schmidt 2438027ed2f0SLi Zefan kfree(fi_args); 2439027ed2f0SLi Zefan return ret; 2440475f6387SJan Schmidt } 2441475f6387SJan Schmidt 2442475f6387SJan Schmidt static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg) 2443475f6387SJan Schmidt { 2444475f6387SJan Schmidt struct btrfs_ioctl_dev_info_args *di_args; 2445475f6387SJan Schmidt struct btrfs_device *dev; 2446475f6387SJan Schmidt struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; 2447475f6387SJan Schmidt int ret = 0; 2448475f6387SJan Schmidt char *s_uuid = NULL; 2449475f6387SJan Schmidt char empty_uuid[BTRFS_UUID_SIZE] = {0}; 2450475f6387SJan Schmidt 2451475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 2452475f6387SJan Schmidt return -EPERM; 2453475f6387SJan Schmidt 2454475f6387SJan Schmidt di_args = memdup_user(arg, sizeof(*di_args)); 2455475f6387SJan Schmidt if (IS_ERR(di_args)) 2456475f6387SJan Schmidt return PTR_ERR(di_args); 2457475f6387SJan Schmidt 2458475f6387SJan Schmidt if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0) 2459475f6387SJan Schmidt s_uuid = di_args->uuid; 2460475f6387SJan Schmidt 2461475f6387SJan Schmidt mutex_lock(&fs_devices->device_list_mutex); 2462aa1b8cd4SStefan Behrens dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL); 2463475f6387SJan Schmidt 2464475f6387SJan Schmidt if (!dev) { 2465475f6387SJan Schmidt ret = -ENODEV; 2466475f6387SJan Schmidt goto out; 2467475f6387SJan Schmidt } 2468475f6387SJan Schmidt 2469475f6387SJan Schmidt di_args->devid = dev->devid; 2470475f6387SJan Schmidt di_args->bytes_used = dev->bytes_used; 2471475f6387SJan Schmidt di_args->total_bytes = dev->total_bytes; 2472475f6387SJan Schmidt memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); 2473a27202fbSJim Meyering if (dev->name) { 2474606686eeSJosef Bacik struct rcu_string *name; 2475606686eeSJosef Bacik 2476606686eeSJosef Bacik rcu_read_lock(); 2477606686eeSJosef Bacik name = rcu_dereference(dev->name); 2478606686eeSJosef Bacik strncpy(di_args->path, name->str, sizeof(di_args->path)); 2479606686eeSJosef Bacik rcu_read_unlock(); 2480a27202fbSJim Meyering di_args->path[sizeof(di_args->path) - 1] = 0; 2481a27202fbSJim Meyering } else { 248299ba55adSStefan Behrens di_args->path[0] = '\0'; 2483a27202fbSJim Meyering } 2484475f6387SJan Schmidt 2485475f6387SJan Schmidt out: 248655793c0dSDavid Sterba mutex_unlock(&fs_devices->device_list_mutex); 2487475f6387SJan Schmidt if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) 2488475f6387SJan Schmidt ret = -EFAULT; 2489475f6387SJan Schmidt 2490475f6387SJan Schmidt kfree(di_args); 2491475f6387SJan Schmidt return ret; 2492475f6387SJan Schmidt } 2493475f6387SJan Schmidt 2494416161dbSMark Fasheh static struct page *extent_same_get_page(struct inode *inode, u64 off) 2495416161dbSMark Fasheh { 2496416161dbSMark Fasheh struct page *page; 2497416161dbSMark Fasheh pgoff_t index; 2498416161dbSMark Fasheh struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree; 2499416161dbSMark Fasheh 2500416161dbSMark Fasheh index = off >> PAGE_CACHE_SHIFT; 2501416161dbSMark Fasheh 2502416161dbSMark Fasheh page = grab_cache_page(inode->i_mapping, index); 2503416161dbSMark Fasheh if (!page) 2504416161dbSMark Fasheh return NULL; 2505416161dbSMark Fasheh 2506416161dbSMark Fasheh if (!PageUptodate(page)) { 2507416161dbSMark Fasheh if (extent_read_full_page_nolock(tree, page, btrfs_get_extent, 2508416161dbSMark Fasheh 0)) 2509416161dbSMark Fasheh return NULL; 2510416161dbSMark Fasheh lock_page(page); 2511416161dbSMark Fasheh if (!PageUptodate(page)) { 2512416161dbSMark Fasheh unlock_page(page); 2513416161dbSMark Fasheh page_cache_release(page); 2514416161dbSMark Fasheh return NULL; 2515416161dbSMark Fasheh } 2516416161dbSMark Fasheh } 2517416161dbSMark Fasheh unlock_page(page); 2518416161dbSMark Fasheh 2519416161dbSMark Fasheh return page; 2520416161dbSMark Fasheh } 2521416161dbSMark Fasheh 252277fe20dcSMark Fasheh static inline void lock_extent_range(struct inode *inode, u64 off, u64 len) 252377fe20dcSMark Fasheh { 252477fe20dcSMark Fasheh /* do any pending delalloc/csum calc on src, one way or 252577fe20dcSMark Fasheh another, and lock file content */ 252677fe20dcSMark Fasheh while (1) { 252777fe20dcSMark Fasheh struct btrfs_ordered_extent *ordered; 252877fe20dcSMark Fasheh lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); 252977fe20dcSMark Fasheh ordered = btrfs_lookup_first_ordered_extent(inode, 253077fe20dcSMark Fasheh off + len - 1); 253177fe20dcSMark Fasheh if (!ordered && 253277fe20dcSMark Fasheh !test_range_bit(&BTRFS_I(inode)->io_tree, off, 253377fe20dcSMark Fasheh off + len - 1, EXTENT_DELALLOC, 0, NULL)) 253477fe20dcSMark Fasheh break; 253577fe20dcSMark Fasheh unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); 253677fe20dcSMark Fasheh if (ordered) 253777fe20dcSMark Fasheh btrfs_put_ordered_extent(ordered); 253877fe20dcSMark Fasheh btrfs_wait_ordered_range(inode, off, len); 253977fe20dcSMark Fasheh } 254077fe20dcSMark Fasheh } 254177fe20dcSMark Fasheh 2542416161dbSMark Fasheh static void btrfs_double_unlock(struct inode *inode1, u64 loff1, 2543416161dbSMark Fasheh struct inode *inode2, u64 loff2, u64 len) 2544416161dbSMark Fasheh { 2545416161dbSMark Fasheh unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); 2546416161dbSMark Fasheh unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); 2547416161dbSMark Fasheh 2548416161dbSMark Fasheh mutex_unlock(&inode1->i_mutex); 2549416161dbSMark Fasheh mutex_unlock(&inode2->i_mutex); 2550416161dbSMark Fasheh } 2551416161dbSMark Fasheh 2552416161dbSMark Fasheh static void btrfs_double_lock(struct inode *inode1, u64 loff1, 2553416161dbSMark Fasheh struct inode *inode2, u64 loff2, u64 len) 2554416161dbSMark Fasheh { 2555416161dbSMark Fasheh if (inode1 < inode2) { 2556416161dbSMark Fasheh swap(inode1, inode2); 2557416161dbSMark Fasheh swap(loff1, loff2); 2558416161dbSMark Fasheh } 2559416161dbSMark Fasheh 2560416161dbSMark Fasheh mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT); 2561416161dbSMark Fasheh lock_extent_range(inode1, loff1, len); 2562416161dbSMark Fasheh if (inode1 != inode2) { 2563416161dbSMark Fasheh mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD); 2564416161dbSMark Fasheh lock_extent_range(inode2, loff2, len); 2565416161dbSMark Fasheh } 2566416161dbSMark Fasheh } 2567416161dbSMark Fasheh 2568416161dbSMark Fasheh static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst, 2569416161dbSMark Fasheh u64 dst_loff, u64 len) 2570416161dbSMark Fasheh { 2571416161dbSMark Fasheh int ret = 0; 2572416161dbSMark Fasheh struct page *src_page, *dst_page; 2573416161dbSMark Fasheh unsigned int cmp_len = PAGE_CACHE_SIZE; 2574416161dbSMark Fasheh void *addr, *dst_addr; 2575416161dbSMark Fasheh 2576416161dbSMark Fasheh while (len) { 2577416161dbSMark Fasheh if (len < PAGE_CACHE_SIZE) 2578416161dbSMark Fasheh cmp_len = len; 2579416161dbSMark Fasheh 2580416161dbSMark Fasheh src_page = extent_same_get_page(src, loff); 2581416161dbSMark Fasheh if (!src_page) 2582416161dbSMark Fasheh return -EINVAL; 2583416161dbSMark Fasheh dst_page = extent_same_get_page(dst, dst_loff); 2584416161dbSMark Fasheh if (!dst_page) { 2585416161dbSMark Fasheh page_cache_release(src_page); 2586416161dbSMark Fasheh return -EINVAL; 2587416161dbSMark Fasheh } 2588416161dbSMark Fasheh addr = kmap_atomic(src_page); 2589416161dbSMark Fasheh dst_addr = kmap_atomic(dst_page); 2590416161dbSMark Fasheh 2591416161dbSMark Fasheh flush_dcache_page(src_page); 2592416161dbSMark Fasheh flush_dcache_page(dst_page); 2593416161dbSMark Fasheh 2594416161dbSMark Fasheh if (memcmp(addr, dst_addr, cmp_len)) 2595416161dbSMark Fasheh ret = BTRFS_SAME_DATA_DIFFERS; 2596416161dbSMark Fasheh 2597416161dbSMark Fasheh kunmap_atomic(addr); 2598416161dbSMark Fasheh kunmap_atomic(dst_addr); 2599416161dbSMark Fasheh page_cache_release(src_page); 2600416161dbSMark Fasheh page_cache_release(dst_page); 2601416161dbSMark Fasheh 2602416161dbSMark Fasheh if (ret) 2603416161dbSMark Fasheh break; 2604416161dbSMark Fasheh 2605416161dbSMark Fasheh loff += cmp_len; 2606416161dbSMark Fasheh dst_loff += cmp_len; 2607416161dbSMark Fasheh len -= cmp_len; 2608416161dbSMark Fasheh } 2609416161dbSMark Fasheh 2610416161dbSMark Fasheh return ret; 2611416161dbSMark Fasheh } 2612416161dbSMark Fasheh 2613416161dbSMark Fasheh static int extent_same_check_offsets(struct inode *inode, u64 off, u64 len) 2614416161dbSMark Fasheh { 2615416161dbSMark Fasheh u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize; 2616416161dbSMark Fasheh 2617416161dbSMark Fasheh if (off + len > inode->i_size || off + len < off) 2618416161dbSMark Fasheh return -EINVAL; 2619416161dbSMark Fasheh /* Check that we are block aligned - btrfs_clone() requires this */ 2620416161dbSMark Fasheh if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs)) 2621416161dbSMark Fasheh return -EINVAL; 2622416161dbSMark Fasheh 2623416161dbSMark Fasheh return 0; 2624416161dbSMark Fasheh } 2625416161dbSMark Fasheh 2626416161dbSMark Fasheh static int btrfs_extent_same(struct inode *src, u64 loff, u64 len, 2627416161dbSMark Fasheh struct inode *dst, u64 dst_loff) 2628416161dbSMark Fasheh { 2629416161dbSMark Fasheh int ret; 2630416161dbSMark Fasheh 2631416161dbSMark Fasheh /* 2632416161dbSMark Fasheh * btrfs_clone() can't handle extents in the same file 2633416161dbSMark Fasheh * yet. Once that works, we can drop this check and replace it 2634416161dbSMark Fasheh * with a check for the same inode, but overlapping extents. 2635416161dbSMark Fasheh */ 2636416161dbSMark Fasheh if (src == dst) 2637416161dbSMark Fasheh return -EINVAL; 2638416161dbSMark Fasheh 2639416161dbSMark Fasheh btrfs_double_lock(src, loff, dst, dst_loff, len); 2640416161dbSMark Fasheh 2641416161dbSMark Fasheh ret = extent_same_check_offsets(src, loff, len); 2642416161dbSMark Fasheh if (ret) 2643416161dbSMark Fasheh goto out_unlock; 2644416161dbSMark Fasheh 2645416161dbSMark Fasheh ret = extent_same_check_offsets(dst, dst_loff, len); 2646416161dbSMark Fasheh if (ret) 2647416161dbSMark Fasheh goto out_unlock; 2648416161dbSMark Fasheh 2649416161dbSMark Fasheh /* don't make the dst file partly checksummed */ 2650416161dbSMark Fasheh if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != 2651416161dbSMark Fasheh (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) { 2652416161dbSMark Fasheh ret = -EINVAL; 2653416161dbSMark Fasheh goto out_unlock; 2654416161dbSMark Fasheh } 2655416161dbSMark Fasheh 2656416161dbSMark Fasheh ret = btrfs_cmp_data(src, loff, dst, dst_loff, len); 2657416161dbSMark Fasheh if (ret == 0) 2658416161dbSMark Fasheh ret = btrfs_clone(src, dst, loff, len, len, dst_loff); 2659416161dbSMark Fasheh 2660416161dbSMark Fasheh out_unlock: 2661416161dbSMark Fasheh btrfs_double_unlock(src, loff, dst, dst_loff, len); 2662416161dbSMark Fasheh 2663416161dbSMark Fasheh return ret; 2664416161dbSMark Fasheh } 2665416161dbSMark Fasheh 2666416161dbSMark Fasheh #define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024) 2667416161dbSMark Fasheh 2668416161dbSMark Fasheh static long btrfs_ioctl_file_extent_same(struct file *file, 2669416161dbSMark Fasheh void __user *argp) 2670416161dbSMark Fasheh { 2671416161dbSMark Fasheh struct btrfs_ioctl_same_args *args = argp; 2672416161dbSMark Fasheh struct btrfs_ioctl_same_args same; 2673416161dbSMark Fasheh struct btrfs_ioctl_same_extent_info info; 2674416161dbSMark Fasheh struct inode *src = file->f_dentry->d_inode; 2675416161dbSMark Fasheh struct file *dst_file = NULL; 2676416161dbSMark Fasheh struct inode *dst; 2677416161dbSMark Fasheh u64 off; 2678416161dbSMark Fasheh u64 len; 2679416161dbSMark Fasheh int i; 2680416161dbSMark Fasheh int ret; 2681416161dbSMark Fasheh u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize; 2682416161dbSMark Fasheh bool is_admin = capable(CAP_SYS_ADMIN); 2683416161dbSMark Fasheh 2684416161dbSMark Fasheh if (!(file->f_mode & FMODE_READ)) 2685416161dbSMark Fasheh return -EINVAL; 2686416161dbSMark Fasheh 2687416161dbSMark Fasheh ret = mnt_want_write_file(file); 2688416161dbSMark Fasheh if (ret) 2689416161dbSMark Fasheh return ret; 2690416161dbSMark Fasheh 2691416161dbSMark Fasheh if (copy_from_user(&same, 2692416161dbSMark Fasheh (struct btrfs_ioctl_same_args __user *)argp, 2693416161dbSMark Fasheh sizeof(same))) { 2694416161dbSMark Fasheh ret = -EFAULT; 2695416161dbSMark Fasheh goto out; 2696416161dbSMark Fasheh } 2697416161dbSMark Fasheh 2698416161dbSMark Fasheh off = same.logical_offset; 2699416161dbSMark Fasheh len = same.length; 2700416161dbSMark Fasheh 2701416161dbSMark Fasheh /* 2702416161dbSMark Fasheh * Limit the total length we will dedupe for each operation. 2703416161dbSMark Fasheh * This is intended to bound the total time spent in this 2704416161dbSMark Fasheh * ioctl to something sane. 2705416161dbSMark Fasheh */ 2706416161dbSMark Fasheh if (len > BTRFS_MAX_DEDUPE_LEN) 2707416161dbSMark Fasheh len = BTRFS_MAX_DEDUPE_LEN; 2708416161dbSMark Fasheh 2709416161dbSMark Fasheh if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) { 2710416161dbSMark Fasheh /* 2711416161dbSMark Fasheh * Btrfs does not support blocksize < page_size. As a 2712416161dbSMark Fasheh * result, btrfs_cmp_data() won't correctly handle 2713416161dbSMark Fasheh * this situation without an update. 2714416161dbSMark Fasheh */ 2715416161dbSMark Fasheh ret = -EINVAL; 2716416161dbSMark Fasheh goto out; 2717416161dbSMark Fasheh } 2718416161dbSMark Fasheh 2719416161dbSMark Fasheh ret = -EISDIR; 2720416161dbSMark Fasheh if (S_ISDIR(src->i_mode)) 2721416161dbSMark Fasheh goto out; 2722416161dbSMark Fasheh 2723416161dbSMark Fasheh ret = -EACCES; 2724416161dbSMark Fasheh if (!S_ISREG(src->i_mode)) 2725416161dbSMark Fasheh goto out; 2726416161dbSMark Fasheh 2727416161dbSMark Fasheh ret = 0; 2728416161dbSMark Fasheh for (i = 0; i < same.dest_count; i++) { 2729416161dbSMark Fasheh if (copy_from_user(&info, &args->info[i], sizeof(info))) { 2730416161dbSMark Fasheh ret = -EFAULT; 2731416161dbSMark Fasheh goto out; 2732416161dbSMark Fasheh } 2733416161dbSMark Fasheh 2734416161dbSMark Fasheh info.bytes_deduped = 0; 2735416161dbSMark Fasheh 2736416161dbSMark Fasheh dst_file = fget(info.fd); 2737416161dbSMark Fasheh if (!dst_file) { 2738416161dbSMark Fasheh info.status = -EBADF; 2739416161dbSMark Fasheh goto next; 2740416161dbSMark Fasheh } 2741416161dbSMark Fasheh 2742416161dbSMark Fasheh if (!(is_admin || (dst_file->f_mode & FMODE_WRITE))) { 2743416161dbSMark Fasheh info.status = -EINVAL; 2744416161dbSMark Fasheh goto next; 2745416161dbSMark Fasheh } 2746416161dbSMark Fasheh 2747416161dbSMark Fasheh info.status = -EXDEV; 2748416161dbSMark Fasheh if (file->f_path.mnt != dst_file->f_path.mnt) 2749416161dbSMark Fasheh goto next; 2750416161dbSMark Fasheh 2751416161dbSMark Fasheh dst = dst_file->f_dentry->d_inode; 2752416161dbSMark Fasheh if (src->i_sb != dst->i_sb) 2753416161dbSMark Fasheh goto next; 2754416161dbSMark Fasheh 2755416161dbSMark Fasheh if (S_ISDIR(dst->i_mode)) { 2756416161dbSMark Fasheh info.status = -EISDIR; 2757416161dbSMark Fasheh goto next; 2758416161dbSMark Fasheh } 2759416161dbSMark Fasheh 2760416161dbSMark Fasheh if (!S_ISREG(dst->i_mode)) { 2761416161dbSMark Fasheh info.status = -EACCES; 2762416161dbSMark Fasheh goto next; 2763416161dbSMark Fasheh } 2764416161dbSMark Fasheh 2765416161dbSMark Fasheh info.status = btrfs_extent_same(src, off, len, dst, 2766416161dbSMark Fasheh info.logical_offset); 2767416161dbSMark Fasheh if (info.status == 0) 2768416161dbSMark Fasheh info.bytes_deduped += len; 2769416161dbSMark Fasheh 2770416161dbSMark Fasheh next: 2771416161dbSMark Fasheh if (dst_file) 2772416161dbSMark Fasheh fput(dst_file); 2773416161dbSMark Fasheh 2774416161dbSMark Fasheh if (__put_user_unaligned(info.status, &args->info[i].status) || 2775416161dbSMark Fasheh __put_user_unaligned(info.bytes_deduped, 2776416161dbSMark Fasheh &args->info[i].bytes_deduped)) { 2777416161dbSMark Fasheh ret = -EFAULT; 2778416161dbSMark Fasheh goto out; 2779416161dbSMark Fasheh } 2780416161dbSMark Fasheh } 2781416161dbSMark Fasheh 2782416161dbSMark Fasheh out: 2783416161dbSMark Fasheh mnt_drop_write_file(file); 2784416161dbSMark Fasheh return ret; 2785416161dbSMark Fasheh } 2786416161dbSMark Fasheh 278732b7c687SMark Fasheh /** 278832b7c687SMark Fasheh * btrfs_clone() - clone a range from inode file to another 278932b7c687SMark Fasheh * 279032b7c687SMark Fasheh * @src: Inode to clone from 279132b7c687SMark Fasheh * @inode: Inode to clone to 279232b7c687SMark Fasheh * @off: Offset within source to start clone from 279332b7c687SMark Fasheh * @olen: Original length, passed by user, of range to clone 279432b7c687SMark Fasheh * @olen_aligned: Block-aligned value of olen, extent_same uses 279532b7c687SMark Fasheh * identical values here 279632b7c687SMark Fasheh * @destoff: Offset within @inode to start clone 279732b7c687SMark Fasheh */ 279832b7c687SMark Fasheh static int btrfs_clone(struct inode *src, struct inode *inode, 279932b7c687SMark Fasheh u64 off, u64 olen, u64 olen_aligned, u64 destoff) 2800f46b5a66SChristoph Hellwig { 2801f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 280232b7c687SMark Fasheh struct btrfs_path *path = NULL; 2803f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 280432b7c687SMark Fasheh struct btrfs_trans_handle *trans; 280532b7c687SMark Fasheh char *buf = NULL; 2806ae01a0abSYan Zheng struct btrfs_key key; 2807f46b5a66SChristoph Hellwig u32 nritems; 2808f46b5a66SChristoph Hellwig int slot; 2809ae01a0abSYan Zheng int ret; 281032b7c687SMark Fasheh u64 len = olen_aligned; 2811ae01a0abSYan Zheng 2812ae01a0abSYan Zheng ret = -ENOMEM; 2813ae01a0abSYan Zheng buf = vmalloc(btrfs_level_size(root, 0)); 2814ae01a0abSYan Zheng if (!buf) 281532b7c687SMark Fasheh return ret; 2816ae01a0abSYan Zheng 2817ae01a0abSYan Zheng path = btrfs_alloc_path(); 2818ae01a0abSYan Zheng if (!path) { 2819ae01a0abSYan Zheng vfree(buf); 282032b7c687SMark Fasheh return ret; 2821ae01a0abSYan Zheng } 282232b7c687SMark Fasheh 2823ae01a0abSYan Zheng path->reada = 2; 2824c5c9cd4dSSage Weil /* clone data */ 282533345d01SLi Zefan key.objectid = btrfs_ino(src); 2826ae01a0abSYan Zheng key.type = BTRFS_EXTENT_DATA_KEY; 2827ae01a0abSYan Zheng key.offset = 0; 2828f46b5a66SChristoph Hellwig 2829f46b5a66SChristoph Hellwig while (1) { 2830f46b5a66SChristoph Hellwig /* 2831f46b5a66SChristoph Hellwig * note the key will change type as we walk through the 2832f46b5a66SChristoph Hellwig * tree. 2833f46b5a66SChristoph Hellwig */ 2834362a20c5SDavid Sterba ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path, 2835362a20c5SDavid Sterba 0, 0); 2836f46b5a66SChristoph Hellwig if (ret < 0) 2837f46b5a66SChristoph Hellwig goto out; 2838f46b5a66SChristoph Hellwig 2839ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 2840ae01a0abSYan Zheng if (path->slots[0] >= nritems) { 2841362a20c5SDavid Sterba ret = btrfs_next_leaf(BTRFS_I(src)->root, path); 2842f46b5a66SChristoph Hellwig if (ret < 0) 2843f46b5a66SChristoph Hellwig goto out; 2844f46b5a66SChristoph Hellwig if (ret > 0) 2845f46b5a66SChristoph Hellwig break; 2846ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 2847f46b5a66SChristoph Hellwig } 2848f46b5a66SChristoph Hellwig leaf = path->nodes[0]; 2849f46b5a66SChristoph Hellwig slot = path->slots[0]; 2850f46b5a66SChristoph Hellwig 2851ae01a0abSYan Zheng btrfs_item_key_to_cpu(leaf, &key, slot); 2852d20f7043SChris Mason if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY || 285333345d01SLi Zefan key.objectid != btrfs_ino(src)) 2854f46b5a66SChristoph Hellwig break; 2855f46b5a66SChristoph Hellwig 2856c5c9cd4dSSage Weil if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) { 2857c5c9cd4dSSage Weil struct btrfs_file_extent_item *extent; 2858c5c9cd4dSSage Weil int type; 285931840ae1SZheng Yan u32 size; 286031840ae1SZheng Yan struct btrfs_key new_key; 2861c5c9cd4dSSage Weil u64 disko = 0, diskl = 0; 2862c5c9cd4dSSage Weil u64 datao = 0, datal = 0; 2863c5c9cd4dSSage Weil u8 comp; 2864b5384d48SSage Weil u64 endoff; 286531840ae1SZheng Yan 286631840ae1SZheng Yan size = btrfs_item_size_nr(leaf, slot); 286731840ae1SZheng Yan read_extent_buffer(leaf, buf, 286831840ae1SZheng Yan btrfs_item_ptr_offset(leaf, slot), 286931840ae1SZheng Yan size); 2870c5c9cd4dSSage Weil 2871c5c9cd4dSSage Weil extent = btrfs_item_ptr(leaf, slot, 2872c5c9cd4dSSage Weil struct btrfs_file_extent_item); 2873c5c9cd4dSSage Weil comp = btrfs_file_extent_compression(leaf, extent); 2874c5c9cd4dSSage Weil type = btrfs_file_extent_type(leaf, extent); 2875c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 2876c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 2877d397712bSChris Mason disko = btrfs_file_extent_disk_bytenr(leaf, 2878d397712bSChris Mason extent); 2879d397712bSChris Mason diskl = btrfs_file_extent_disk_num_bytes(leaf, 2880d397712bSChris Mason extent); 2881c5c9cd4dSSage Weil datao = btrfs_file_extent_offset(leaf, extent); 2882d397712bSChris Mason datal = btrfs_file_extent_num_bytes(leaf, 2883d397712bSChris Mason extent); 2884c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 2885c5c9cd4dSSage Weil /* take upper bound, may be compressed */ 2886c5c9cd4dSSage Weil datal = btrfs_file_extent_ram_bytes(leaf, 2887c5c9cd4dSSage Weil extent); 2888c5c9cd4dSSage Weil } 2889b3b4aa74SDavid Sterba btrfs_release_path(path); 289031840ae1SZheng Yan 2891050006a7SSage Weil if (key.offset + datal <= off || 2892aa42ffd9SLiu Bo key.offset >= off + len - 1) 2893c5c9cd4dSSage Weil goto next; 2894c5c9cd4dSSage Weil 289531840ae1SZheng Yan memcpy(&new_key, &key, sizeof(new_key)); 289633345d01SLi Zefan new_key.objectid = btrfs_ino(inode); 28974d728ec7SLi Zefan if (off <= key.offset) 2898c5c9cd4dSSage Weil new_key.offset = key.offset + destoff - off; 28994d728ec7SLi Zefan else 29004d728ec7SLi Zefan new_key.offset = destoff; 2901c5c9cd4dSSage Weil 2902b6f3409bSSage Weil /* 2903b6f3409bSSage Weil * 1 - adjusting old extent (we may have to split it) 2904b6f3409bSSage Weil * 1 - add new extent 2905b6f3409bSSage Weil * 1 - inode update 2906b6f3409bSSage Weil */ 2907b6f3409bSSage Weil trans = btrfs_start_transaction(root, 3); 2908a22285a6SYan, Zheng if (IS_ERR(trans)) { 2909a22285a6SYan, Zheng ret = PTR_ERR(trans); 2910a22285a6SYan, Zheng goto out; 2911a22285a6SYan, Zheng } 2912a22285a6SYan, Zheng 2913c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 2914c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 2915d72c0842SLi Zefan /* 2916d72c0842SLi Zefan * a | --- range to clone ---| b 2917d72c0842SLi Zefan * | ------------- extent ------------- | 2918d72c0842SLi Zefan */ 2919d72c0842SLi Zefan 2920d72c0842SLi Zefan /* substract range b */ 2921d72c0842SLi Zefan if (key.offset + datal > off + len) 2922d72c0842SLi Zefan datal = off + len - key.offset; 2923d72c0842SLi Zefan 2924d72c0842SLi Zefan /* substract range a */ 2925a22285a6SYan, Zheng if (off > key.offset) { 2926a22285a6SYan, Zheng datao += off - key.offset; 2927a22285a6SYan, Zheng datal -= off - key.offset; 2928a22285a6SYan, Zheng } 2929a22285a6SYan, Zheng 29305dc562c5SJosef Bacik ret = btrfs_drop_extents(trans, root, inode, 2931a22285a6SYan, Zheng new_key.offset, 2932a22285a6SYan, Zheng new_key.offset + datal, 29332671485dSJosef Bacik 1); 293479787eaaSJeff Mahoney if (ret) { 293579787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, 293679787eaaSJeff Mahoney ret); 293779787eaaSJeff Mahoney btrfs_end_transaction(trans, root); 293879787eaaSJeff Mahoney goto out; 293979787eaaSJeff Mahoney } 2940a22285a6SYan, Zheng 294131840ae1SZheng Yan ret = btrfs_insert_empty_item(trans, root, path, 294231840ae1SZheng Yan &new_key, size); 294379787eaaSJeff Mahoney if (ret) { 294479787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, 294579787eaaSJeff Mahoney ret); 294679787eaaSJeff Mahoney btrfs_end_transaction(trans, root); 294779787eaaSJeff Mahoney goto out; 294879787eaaSJeff Mahoney } 294931840ae1SZheng Yan 295031840ae1SZheng Yan leaf = path->nodes[0]; 295131840ae1SZheng Yan slot = path->slots[0]; 295231840ae1SZheng Yan write_extent_buffer(leaf, buf, 295331840ae1SZheng Yan btrfs_item_ptr_offset(leaf, slot), 295431840ae1SZheng Yan size); 2955ae01a0abSYan Zheng 2956f46b5a66SChristoph Hellwig extent = btrfs_item_ptr(leaf, slot, 2957f46b5a66SChristoph Hellwig struct btrfs_file_extent_item); 2958c5c9cd4dSSage Weil 2959c5c9cd4dSSage Weil /* disko == 0 means it's a hole */ 2960c5c9cd4dSSage Weil if (!disko) 2961c5c9cd4dSSage Weil datao = 0; 2962c5c9cd4dSSage Weil 2963c5c9cd4dSSage Weil btrfs_set_file_extent_offset(leaf, extent, 2964c5c9cd4dSSage Weil datao); 2965c5c9cd4dSSage Weil btrfs_set_file_extent_num_bytes(leaf, extent, 2966c5c9cd4dSSage Weil datal); 2967c5c9cd4dSSage Weil if (disko) { 2968c5c9cd4dSSage Weil inode_add_bytes(inode, datal); 2969ae01a0abSYan Zheng ret = btrfs_inc_extent_ref(trans, root, 29705d4f98a2SYan Zheng disko, diskl, 0, 2971f46b5a66SChristoph Hellwig root->root_key.objectid, 297233345d01SLi Zefan btrfs_ino(inode), 297366d7e7f0SArne Jansen new_key.offset - datao, 297466d7e7f0SArne Jansen 0); 297579787eaaSJeff Mahoney if (ret) { 297679787eaaSJeff Mahoney btrfs_abort_transaction(trans, 297779787eaaSJeff Mahoney root, 297879787eaaSJeff Mahoney ret); 297979787eaaSJeff Mahoney btrfs_end_transaction(trans, 298079787eaaSJeff Mahoney root); 298179787eaaSJeff Mahoney goto out; 298279787eaaSJeff Mahoney 298379787eaaSJeff Mahoney } 2984f46b5a66SChristoph Hellwig } 2985c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 2986c5c9cd4dSSage Weil u64 skip = 0; 2987c5c9cd4dSSage Weil u64 trim = 0; 2988c5c9cd4dSSage Weil if (off > key.offset) { 2989c5c9cd4dSSage Weil skip = off - key.offset; 2990c5c9cd4dSSage Weil new_key.offset += skip; 299131840ae1SZheng Yan } 2992d397712bSChris Mason 2993c5c9cd4dSSage Weil if (key.offset + datal > off + len) 2994c5c9cd4dSSage Weil trim = key.offset + datal - (off + len); 2995d397712bSChris Mason 2996c5c9cd4dSSage Weil if (comp && (skip || trim)) { 2997c5c9cd4dSSage Weil ret = -EINVAL; 2998a22285a6SYan, Zheng btrfs_end_transaction(trans, root); 2999c5c9cd4dSSage Weil goto out; 300031840ae1SZheng Yan } 3001c5c9cd4dSSage Weil size -= skip + trim; 3002c5c9cd4dSSage Weil datal -= skip + trim; 3003a22285a6SYan, Zheng 30045dc562c5SJosef Bacik ret = btrfs_drop_extents(trans, root, inode, 3005a22285a6SYan, Zheng new_key.offset, 3006a22285a6SYan, Zheng new_key.offset + datal, 30072671485dSJosef Bacik 1); 300879787eaaSJeff Mahoney if (ret) { 300979787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, 301079787eaaSJeff Mahoney ret); 301179787eaaSJeff Mahoney btrfs_end_transaction(trans, root); 301279787eaaSJeff Mahoney goto out; 301379787eaaSJeff Mahoney } 3014a22285a6SYan, Zheng 3015c5c9cd4dSSage Weil ret = btrfs_insert_empty_item(trans, root, path, 3016c5c9cd4dSSage Weil &new_key, size); 301779787eaaSJeff Mahoney if (ret) { 301879787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, 301979787eaaSJeff Mahoney ret); 302079787eaaSJeff Mahoney btrfs_end_transaction(trans, root); 302179787eaaSJeff Mahoney goto out; 302279787eaaSJeff Mahoney } 3023c5c9cd4dSSage Weil 3024c5c9cd4dSSage Weil if (skip) { 3025d397712bSChris Mason u32 start = 3026d397712bSChris Mason btrfs_file_extent_calc_inline_size(0); 3027c5c9cd4dSSage Weil memmove(buf+start, buf+start+skip, 3028c5c9cd4dSSage Weil datal); 3029c5c9cd4dSSage Weil } 3030c5c9cd4dSSage Weil 3031c5c9cd4dSSage Weil leaf = path->nodes[0]; 3032c5c9cd4dSSage Weil slot = path->slots[0]; 3033c5c9cd4dSSage Weil write_extent_buffer(leaf, buf, 3034c5c9cd4dSSage Weil btrfs_item_ptr_offset(leaf, slot), 3035c5c9cd4dSSage Weil size); 3036c5c9cd4dSSage Weil inode_add_bytes(inode, datal); 3037c5c9cd4dSSage Weil } 3038c5c9cd4dSSage Weil 3039c5c9cd4dSSage Weil btrfs_mark_buffer_dirty(leaf); 3040b3b4aa74SDavid Sterba btrfs_release_path(path); 3041c5c9cd4dSSage Weil 30420c4d2d95SJosef Bacik inode_inc_iversion(inode); 3043a22285a6SYan, Zheng inode->i_mtime = inode->i_ctime = CURRENT_TIME; 3044b5384d48SSage Weil 3045b5384d48SSage Weil /* 3046b5384d48SSage Weil * we round up to the block size at eof when 3047b5384d48SSage Weil * determining which extents to clone above, 3048b5384d48SSage Weil * but shouldn't round up the file size 3049b5384d48SSage Weil */ 3050b5384d48SSage Weil endoff = new_key.offset + datal; 30515f3888ffSLi Zefan if (endoff > destoff+olen) 30525f3888ffSLi Zefan endoff = destoff+olen; 3053b5384d48SSage Weil if (endoff > inode->i_size) 3054b5384d48SSage Weil btrfs_i_size_write(inode, endoff); 3055b5384d48SSage Weil 3056a22285a6SYan, Zheng ret = btrfs_update_inode(trans, root, inode); 305779787eaaSJeff Mahoney if (ret) { 305879787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, ret); 3059a22285a6SYan, Zheng btrfs_end_transaction(trans, root); 306079787eaaSJeff Mahoney goto out; 306179787eaaSJeff Mahoney } 306279787eaaSJeff Mahoney ret = btrfs_end_transaction(trans, root); 3063a22285a6SYan, Zheng } 3064c5c9cd4dSSage Weil next: 3065b3b4aa74SDavid Sterba btrfs_release_path(path); 3066ae01a0abSYan Zheng key.offset++; 3067ae01a0abSYan Zheng } 3068f46b5a66SChristoph Hellwig ret = 0; 306932b7c687SMark Fasheh 3070f46b5a66SChristoph Hellwig out: 3071b3b4aa74SDavid Sterba btrfs_release_path(path); 307232b7c687SMark Fasheh btrfs_free_path(path); 307332b7c687SMark Fasheh vfree(buf); 307432b7c687SMark Fasheh return ret; 307532b7c687SMark Fasheh } 307632b7c687SMark Fasheh 307732b7c687SMark Fasheh static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, 307832b7c687SMark Fasheh u64 off, u64 olen, u64 destoff) 307932b7c687SMark Fasheh { 308032b7c687SMark Fasheh struct inode *inode = fdentry(file)->d_inode; 308132b7c687SMark Fasheh struct btrfs_root *root = BTRFS_I(inode)->root; 308232b7c687SMark Fasheh struct fd src_file; 308332b7c687SMark Fasheh struct inode *src; 308432b7c687SMark Fasheh int ret; 308532b7c687SMark Fasheh u64 len = olen; 308632b7c687SMark Fasheh u64 bs = root->fs_info->sb->s_blocksize; 308732b7c687SMark Fasheh int same_inode = 0; 308832b7c687SMark Fasheh 308932b7c687SMark Fasheh /* 309032b7c687SMark Fasheh * TODO: 309132b7c687SMark Fasheh * - split compressed inline extents. annoying: we need to 309232b7c687SMark Fasheh * decompress into destination's address_space (the file offset 309332b7c687SMark Fasheh * may change, so source mapping won't do), then recompress (or 309432b7c687SMark Fasheh * otherwise reinsert) a subrange. 309532b7c687SMark Fasheh * - allow ranges within the same file to be cloned (provided 309632b7c687SMark Fasheh * they don't overlap)? 309732b7c687SMark Fasheh */ 309832b7c687SMark Fasheh 309932b7c687SMark Fasheh /* the destination must be opened for writing */ 310032b7c687SMark Fasheh if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND)) 310132b7c687SMark Fasheh return -EINVAL; 310232b7c687SMark Fasheh 310332b7c687SMark Fasheh if (btrfs_root_readonly(root)) 310432b7c687SMark Fasheh return -EROFS; 310532b7c687SMark Fasheh 310632b7c687SMark Fasheh ret = mnt_want_write_file(file); 310732b7c687SMark Fasheh if (ret) 310832b7c687SMark Fasheh return ret; 310932b7c687SMark Fasheh 311032b7c687SMark Fasheh src_file = fdget(srcfd); 311132b7c687SMark Fasheh if (!src_file.file) { 311232b7c687SMark Fasheh ret = -EBADF; 311332b7c687SMark Fasheh goto out_drop_write; 311432b7c687SMark Fasheh } 311532b7c687SMark Fasheh 311632b7c687SMark Fasheh ret = -EXDEV; 311732b7c687SMark Fasheh if (src_file.file->f_path.mnt != file->f_path.mnt) 311832b7c687SMark Fasheh goto out_fput; 311932b7c687SMark Fasheh 312032b7c687SMark Fasheh src = file_inode(src_file.file); 312132b7c687SMark Fasheh 312232b7c687SMark Fasheh ret = -EINVAL; 312332b7c687SMark Fasheh if (src == inode) 312432b7c687SMark Fasheh same_inode = 1; 312532b7c687SMark Fasheh 312632b7c687SMark Fasheh /* the src must be open for reading */ 312732b7c687SMark Fasheh if (!(src_file.file->f_mode & FMODE_READ)) 312832b7c687SMark Fasheh goto out_fput; 312932b7c687SMark Fasheh 313032b7c687SMark Fasheh /* don't make the dst file partly checksummed */ 313132b7c687SMark Fasheh if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != 313232b7c687SMark Fasheh (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) 313332b7c687SMark Fasheh goto out_fput; 313432b7c687SMark Fasheh 313532b7c687SMark Fasheh ret = -EISDIR; 313632b7c687SMark Fasheh if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode)) 313732b7c687SMark Fasheh goto out_fput; 313832b7c687SMark Fasheh 313932b7c687SMark Fasheh ret = -EXDEV; 314032b7c687SMark Fasheh if (src->i_sb != inode->i_sb) 314132b7c687SMark Fasheh goto out_fput; 314232b7c687SMark Fasheh 314332b7c687SMark Fasheh if (!same_inode) { 314432b7c687SMark Fasheh if (inode < src) { 314532b7c687SMark Fasheh mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT); 314632b7c687SMark Fasheh mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD); 314732b7c687SMark Fasheh } else { 314832b7c687SMark Fasheh mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT); 314932b7c687SMark Fasheh mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD); 315032b7c687SMark Fasheh } 315132b7c687SMark Fasheh } else { 315232b7c687SMark Fasheh mutex_lock(&src->i_mutex); 315332b7c687SMark Fasheh } 315432b7c687SMark Fasheh 315532b7c687SMark Fasheh /* determine range to clone */ 315632b7c687SMark Fasheh ret = -EINVAL; 315732b7c687SMark Fasheh if (off + len > src->i_size || off + len < off) 315832b7c687SMark Fasheh goto out_unlock; 315932b7c687SMark Fasheh if (len == 0) 316032b7c687SMark Fasheh olen = len = src->i_size - off; 316132b7c687SMark Fasheh /* if we extend to eof, continue to block boundary */ 316232b7c687SMark Fasheh if (off + len == src->i_size) 316332b7c687SMark Fasheh len = ALIGN(src->i_size, bs) - off; 316432b7c687SMark Fasheh 316532b7c687SMark Fasheh /* verify the end result is block aligned */ 316632b7c687SMark Fasheh if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) || 316732b7c687SMark Fasheh !IS_ALIGNED(destoff, bs)) 316832b7c687SMark Fasheh goto out_unlock; 316932b7c687SMark Fasheh 317032b7c687SMark Fasheh /* verify if ranges are overlapped within the same file */ 317132b7c687SMark Fasheh if (same_inode) { 317232b7c687SMark Fasheh if (destoff + len > off && destoff < off + len) 317332b7c687SMark Fasheh goto out_unlock; 317432b7c687SMark Fasheh } 317532b7c687SMark Fasheh 317632b7c687SMark Fasheh if (destoff > inode->i_size) { 317732b7c687SMark Fasheh ret = btrfs_cont_expand(inode, inode->i_size, destoff); 317832b7c687SMark Fasheh if (ret) 317932b7c687SMark Fasheh goto out_unlock; 318032b7c687SMark Fasheh } 318132b7c687SMark Fasheh 318232b7c687SMark Fasheh /* truncate page cache pages from target inode range */ 318332b7c687SMark Fasheh truncate_inode_pages_range(&inode->i_data, destoff, 318432b7c687SMark Fasheh PAGE_CACHE_ALIGN(destoff + len) - 1); 318532b7c687SMark Fasheh 318632b7c687SMark Fasheh lock_extent_range(src, off, len); 318732b7c687SMark Fasheh 318832b7c687SMark Fasheh ret = btrfs_clone(src, inode, off, olen, len, destoff); 318932b7c687SMark Fasheh 3190aa42ffd9SLiu Bo unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1); 3191f46b5a66SChristoph Hellwig out_unlock: 3192f46b5a66SChristoph Hellwig mutex_unlock(&src->i_mutex); 3193a96fbc72SLiu Bo if (!same_inode) 3194f46b5a66SChristoph Hellwig mutex_unlock(&inode->i_mutex); 3195f46b5a66SChristoph Hellwig out_fput: 31962903ff01SAl Viro fdput(src_file); 3197ab67b7c1SYan Zheng out_drop_write: 31982a79f17eSAl Viro mnt_drop_write_file(file); 3199f46b5a66SChristoph Hellwig return ret; 3200f46b5a66SChristoph Hellwig } 3201f46b5a66SChristoph Hellwig 32027a865e8aSChristoph Hellwig static long btrfs_ioctl_clone_range(struct file *file, void __user *argp) 3203c5c9cd4dSSage Weil { 3204c5c9cd4dSSage Weil struct btrfs_ioctl_clone_range_args args; 3205c5c9cd4dSSage Weil 32067a865e8aSChristoph Hellwig if (copy_from_user(&args, argp, sizeof(args))) 3207c5c9cd4dSSage Weil return -EFAULT; 3208c5c9cd4dSSage Weil return btrfs_ioctl_clone(file, args.src_fd, args.src_offset, 3209c5c9cd4dSSage Weil args.src_length, args.dest_offset); 3210c5c9cd4dSSage Weil } 3211c5c9cd4dSSage Weil 3212f46b5a66SChristoph Hellwig /* 3213f46b5a66SChristoph Hellwig * there are many ways the trans_start and trans_end ioctls can lead 3214f46b5a66SChristoph Hellwig * to deadlocks. They should only be used by applications that 3215f46b5a66SChristoph Hellwig * basically own the machine, and have a very in depth understanding 3216f46b5a66SChristoph Hellwig * of all the possible deadlocks and enospc problems. 3217f46b5a66SChristoph Hellwig */ 3218b2950863SChristoph Hellwig static long btrfs_ioctl_trans_start(struct file *file) 3219f46b5a66SChristoph Hellwig { 3220496ad9aaSAl Viro struct inode *inode = file_inode(file); 3221f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 3222f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 32231ab86aedSSage Weil int ret; 3224f46b5a66SChristoph Hellwig 32251ab86aedSSage Weil ret = -EPERM; 3226df5b5520SChristoph Hellwig if (!capable(CAP_SYS_ADMIN)) 3227f46b5a66SChristoph Hellwig goto out; 32281ab86aedSSage Weil 32291ab86aedSSage Weil ret = -EINPROGRESS; 32301ab86aedSSage Weil if (file->private_data) 32311ab86aedSSage Weil goto out; 32329ca9ee09SSage Weil 3233b83cc969SLi Zefan ret = -EROFS; 3234b83cc969SLi Zefan if (btrfs_root_readonly(root)) 3235b83cc969SLi Zefan goto out; 3236b83cc969SLi Zefan 3237a561be71SAl Viro ret = mnt_want_write_file(file); 3238c146afadSYan Zheng if (ret) 3239c146afadSYan Zheng goto out; 3240c146afadSYan Zheng 3241a4abeea4SJosef Bacik atomic_inc(&root->fs_info->open_ioctl_trans); 32429ca9ee09SSage Weil 3243f46b5a66SChristoph Hellwig ret = -ENOMEM; 32447a7eaa40SJosef Bacik trans = btrfs_start_ioctl_transaction(root); 3245abd30bb0STsutomu Itoh if (IS_ERR(trans)) 32461ab86aedSSage Weil goto out_drop; 32471ab86aedSSage Weil 32481ab86aedSSage Weil file->private_data = trans; 32491ab86aedSSage Weil return 0; 32501ab86aedSSage Weil 32511ab86aedSSage Weil out_drop: 3252a4abeea4SJosef Bacik atomic_dec(&root->fs_info->open_ioctl_trans); 32532a79f17eSAl Viro mnt_drop_write_file(file); 3254f46b5a66SChristoph Hellwig out: 3255f46b5a66SChristoph Hellwig return ret; 3256f46b5a66SChristoph Hellwig } 3257f46b5a66SChristoph Hellwig 32586ef5ed0dSJosef Bacik static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) 32596ef5ed0dSJosef Bacik { 3260496ad9aaSAl Viro struct inode *inode = file_inode(file); 32616ef5ed0dSJosef Bacik struct btrfs_root *root = BTRFS_I(inode)->root; 32626ef5ed0dSJosef Bacik struct btrfs_root *new_root; 32636ef5ed0dSJosef Bacik struct btrfs_dir_item *di; 32646ef5ed0dSJosef Bacik struct btrfs_trans_handle *trans; 32656ef5ed0dSJosef Bacik struct btrfs_path *path; 32666ef5ed0dSJosef Bacik struct btrfs_key location; 32676ef5ed0dSJosef Bacik struct btrfs_disk_key disk_key; 32686ef5ed0dSJosef Bacik u64 objectid = 0; 32696ef5ed0dSJosef Bacik u64 dir_id; 32703c04ce01SMiao Xie int ret; 32716ef5ed0dSJosef Bacik 32726ef5ed0dSJosef Bacik if (!capable(CAP_SYS_ADMIN)) 32736ef5ed0dSJosef Bacik return -EPERM; 32746ef5ed0dSJosef Bacik 32753c04ce01SMiao Xie ret = mnt_want_write_file(file); 32763c04ce01SMiao Xie if (ret) 32773c04ce01SMiao Xie return ret; 32783c04ce01SMiao Xie 32793c04ce01SMiao Xie if (copy_from_user(&objectid, argp, sizeof(objectid))) { 32803c04ce01SMiao Xie ret = -EFAULT; 32813c04ce01SMiao Xie goto out; 32823c04ce01SMiao Xie } 32836ef5ed0dSJosef Bacik 32846ef5ed0dSJosef Bacik if (!objectid) 32856ef5ed0dSJosef Bacik objectid = root->root_key.objectid; 32866ef5ed0dSJosef Bacik 32876ef5ed0dSJosef Bacik location.objectid = objectid; 32886ef5ed0dSJosef Bacik location.type = BTRFS_ROOT_ITEM_KEY; 32896ef5ed0dSJosef Bacik location.offset = (u64)-1; 32906ef5ed0dSJosef Bacik 32916ef5ed0dSJosef Bacik new_root = btrfs_read_fs_root_no_name(root->fs_info, &location); 32923c04ce01SMiao Xie if (IS_ERR(new_root)) { 32933c04ce01SMiao Xie ret = PTR_ERR(new_root); 32943c04ce01SMiao Xie goto out; 32953c04ce01SMiao Xie } 32966ef5ed0dSJosef Bacik 32976ef5ed0dSJosef Bacik path = btrfs_alloc_path(); 32983c04ce01SMiao Xie if (!path) { 32993c04ce01SMiao Xie ret = -ENOMEM; 33003c04ce01SMiao Xie goto out; 33013c04ce01SMiao Xie } 33026ef5ed0dSJosef Bacik path->leave_spinning = 1; 33036ef5ed0dSJosef Bacik 33046ef5ed0dSJosef Bacik trans = btrfs_start_transaction(root, 1); 330598d5dc13STsutomu Itoh if (IS_ERR(trans)) { 33066ef5ed0dSJosef Bacik btrfs_free_path(path); 33073c04ce01SMiao Xie ret = PTR_ERR(trans); 33083c04ce01SMiao Xie goto out; 33096ef5ed0dSJosef Bacik } 33106ef5ed0dSJosef Bacik 33116c41761fSDavid Sterba dir_id = btrfs_super_root_dir(root->fs_info->super_copy); 33126ef5ed0dSJosef Bacik di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path, 33136ef5ed0dSJosef Bacik dir_id, "default", 7, 1); 3314cf1e99a4SDan Carpenter if (IS_ERR_OR_NULL(di)) { 33156ef5ed0dSJosef Bacik btrfs_free_path(path); 33166ef5ed0dSJosef Bacik btrfs_end_transaction(trans, root); 33176ef5ed0dSJosef Bacik printk(KERN_ERR "Umm, you don't have the default dir item, " 33186ef5ed0dSJosef Bacik "this isn't going to work\n"); 33193c04ce01SMiao Xie ret = -ENOENT; 33203c04ce01SMiao Xie goto out; 33216ef5ed0dSJosef Bacik } 33226ef5ed0dSJosef Bacik 33236ef5ed0dSJosef Bacik btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); 33246ef5ed0dSJosef Bacik btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); 33256ef5ed0dSJosef Bacik btrfs_mark_buffer_dirty(path->nodes[0]); 33266ef5ed0dSJosef Bacik btrfs_free_path(path); 33276ef5ed0dSJosef Bacik 33282b0ce2c2SMitch Harder btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL); 33296ef5ed0dSJosef Bacik btrfs_end_transaction(trans, root); 33303c04ce01SMiao Xie out: 33313c04ce01SMiao Xie mnt_drop_write_file(file); 33323c04ce01SMiao Xie return ret; 33336ef5ed0dSJosef Bacik } 33346ef5ed0dSJosef Bacik 33355af3e8ccSStefan Behrens void btrfs_get_block_group_info(struct list_head *groups_list, 3336bf5fc093SJosef Bacik struct btrfs_ioctl_space_info *space) 3337bf5fc093SJosef Bacik { 3338bf5fc093SJosef Bacik struct btrfs_block_group_cache *block_group; 3339bf5fc093SJosef Bacik 3340bf5fc093SJosef Bacik space->total_bytes = 0; 3341bf5fc093SJosef Bacik space->used_bytes = 0; 3342bf5fc093SJosef Bacik space->flags = 0; 3343bf5fc093SJosef Bacik list_for_each_entry(block_group, groups_list, list) { 3344bf5fc093SJosef Bacik space->flags = block_group->flags; 3345bf5fc093SJosef Bacik space->total_bytes += block_group->key.offset; 3346bf5fc093SJosef Bacik space->used_bytes += 3347bf5fc093SJosef Bacik btrfs_block_group_used(&block_group->item); 3348bf5fc093SJosef Bacik } 3349bf5fc093SJosef Bacik } 3350bf5fc093SJosef Bacik 335148a3b636SEric Sandeen static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg) 33521406e432SJosef Bacik { 33531406e432SJosef Bacik struct btrfs_ioctl_space_args space_args; 33541406e432SJosef Bacik struct btrfs_ioctl_space_info space; 33551406e432SJosef Bacik struct btrfs_ioctl_space_info *dest; 33567fde62bfSChris Mason struct btrfs_ioctl_space_info *dest_orig; 335713f2696fSDaniel J Blueman struct btrfs_ioctl_space_info __user *user_dest; 33581406e432SJosef Bacik struct btrfs_space_info *info; 3359bf5fc093SJosef Bacik u64 types[] = {BTRFS_BLOCK_GROUP_DATA, 3360bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_SYSTEM, 3361bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_METADATA, 3362bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA}; 3363bf5fc093SJosef Bacik int num_types = 4; 33647fde62bfSChris Mason int alloc_size; 33651406e432SJosef Bacik int ret = 0; 336651788b1bSDan Rosenberg u64 slot_count = 0; 3367bf5fc093SJosef Bacik int i, c; 33681406e432SJosef Bacik 33691406e432SJosef Bacik if (copy_from_user(&space_args, 33701406e432SJosef Bacik (struct btrfs_ioctl_space_args __user *)arg, 33711406e432SJosef Bacik sizeof(space_args))) 33721406e432SJosef Bacik return -EFAULT; 33731406e432SJosef Bacik 3374bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 3375bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 3376bf5fc093SJosef Bacik 3377bf5fc093SJosef Bacik info = NULL; 33787fde62bfSChris Mason rcu_read_lock(); 3379bf5fc093SJosef Bacik list_for_each_entry_rcu(tmp, &root->fs_info->space_info, 3380bf5fc093SJosef Bacik list) { 3381bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 3382bf5fc093SJosef Bacik info = tmp; 3383bf5fc093SJosef Bacik break; 3384bf5fc093SJosef Bacik } 3385bf5fc093SJosef Bacik } 33867fde62bfSChris Mason rcu_read_unlock(); 33871406e432SJosef Bacik 3388bf5fc093SJosef Bacik if (!info) 3389bf5fc093SJosef Bacik continue; 3390bf5fc093SJosef Bacik 3391bf5fc093SJosef Bacik down_read(&info->groups_sem); 3392bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 3393bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) 3394bf5fc093SJosef Bacik slot_count++; 3395bf5fc093SJosef Bacik } 3396bf5fc093SJosef Bacik up_read(&info->groups_sem); 3397bf5fc093SJosef Bacik } 3398bf5fc093SJosef Bacik 33997fde62bfSChris Mason /* space_slots == 0 means they are asking for a count */ 34007fde62bfSChris Mason if (space_args.space_slots == 0) { 34017fde62bfSChris Mason space_args.total_spaces = slot_count; 34027fde62bfSChris Mason goto out; 34037fde62bfSChris Mason } 3404bf5fc093SJosef Bacik 340551788b1bSDan Rosenberg slot_count = min_t(u64, space_args.space_slots, slot_count); 3406bf5fc093SJosef Bacik 34077fde62bfSChris Mason alloc_size = sizeof(*dest) * slot_count; 3408bf5fc093SJosef Bacik 34097fde62bfSChris Mason /* we generally have at most 6 or so space infos, one for each raid 34107fde62bfSChris Mason * level. So, a whole page should be more than enough for everyone 34117fde62bfSChris Mason */ 34127fde62bfSChris Mason if (alloc_size > PAGE_CACHE_SIZE) 34137fde62bfSChris Mason return -ENOMEM; 34147fde62bfSChris Mason 34157fde62bfSChris Mason space_args.total_spaces = 0; 34167fde62bfSChris Mason dest = kmalloc(alloc_size, GFP_NOFS); 34177fde62bfSChris Mason if (!dest) 34187fde62bfSChris Mason return -ENOMEM; 34197fde62bfSChris Mason dest_orig = dest; 34207fde62bfSChris Mason 34217fde62bfSChris Mason /* now we have a buffer to copy into */ 3422bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 3423bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 3424bf5fc093SJosef Bacik 342551788b1bSDan Rosenberg if (!slot_count) 342651788b1bSDan Rosenberg break; 342751788b1bSDan Rosenberg 3428bf5fc093SJosef Bacik info = NULL; 34291406e432SJosef Bacik rcu_read_lock(); 3430bf5fc093SJosef Bacik list_for_each_entry_rcu(tmp, &root->fs_info->space_info, 3431bf5fc093SJosef Bacik list) { 3432bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 3433bf5fc093SJosef Bacik info = tmp; 34347fde62bfSChris Mason break; 3435bf5fc093SJosef Bacik } 3436bf5fc093SJosef Bacik } 3437bf5fc093SJosef Bacik rcu_read_unlock(); 34387fde62bfSChris Mason 3439bf5fc093SJosef Bacik if (!info) 3440bf5fc093SJosef Bacik continue; 3441bf5fc093SJosef Bacik down_read(&info->groups_sem); 3442bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 3443bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) { 34445af3e8ccSStefan Behrens btrfs_get_block_group_info( 34455af3e8ccSStefan Behrens &info->block_groups[c], &space); 34467fde62bfSChris Mason memcpy(dest, &space, sizeof(space)); 34471406e432SJosef Bacik dest++; 34481406e432SJosef Bacik space_args.total_spaces++; 344951788b1bSDan Rosenberg slot_count--; 34501406e432SJosef Bacik } 345151788b1bSDan Rosenberg if (!slot_count) 345251788b1bSDan Rosenberg break; 3453bf5fc093SJosef Bacik } 3454bf5fc093SJosef Bacik up_read(&info->groups_sem); 3455bf5fc093SJosef Bacik } 34561406e432SJosef Bacik 34572eec6c81SDaniel J Blueman user_dest = (struct btrfs_ioctl_space_info __user *) 34587fde62bfSChris Mason (arg + sizeof(struct btrfs_ioctl_space_args)); 34597fde62bfSChris Mason 34607fde62bfSChris Mason if (copy_to_user(user_dest, dest_orig, alloc_size)) 34617fde62bfSChris Mason ret = -EFAULT; 34627fde62bfSChris Mason 34637fde62bfSChris Mason kfree(dest_orig); 34647fde62bfSChris Mason out: 34657fde62bfSChris Mason if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) 34661406e432SJosef Bacik ret = -EFAULT; 34671406e432SJosef Bacik 34681406e432SJosef Bacik return ret; 34691406e432SJosef Bacik } 34701406e432SJosef Bacik 3471f46b5a66SChristoph Hellwig /* 3472f46b5a66SChristoph Hellwig * there are many ways the trans_start and trans_end ioctls can lead 3473f46b5a66SChristoph Hellwig * to deadlocks. They should only be used by applications that 3474f46b5a66SChristoph Hellwig * basically own the machine, and have a very in depth understanding 3475f46b5a66SChristoph Hellwig * of all the possible deadlocks and enospc problems. 3476f46b5a66SChristoph Hellwig */ 3477f46b5a66SChristoph Hellwig long btrfs_ioctl_trans_end(struct file *file) 3478f46b5a66SChristoph Hellwig { 3479496ad9aaSAl Viro struct inode *inode = file_inode(file); 3480f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 3481f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 3482f46b5a66SChristoph Hellwig 3483f46b5a66SChristoph Hellwig trans = file->private_data; 34841ab86aedSSage Weil if (!trans) 34851ab86aedSSage Weil return -EINVAL; 3486b214107eSChristoph Hellwig file->private_data = NULL; 34879ca9ee09SSage Weil 34881ab86aedSSage Weil btrfs_end_transaction(trans, root); 34891ab86aedSSage Weil 3490a4abeea4SJosef Bacik atomic_dec(&root->fs_info->open_ioctl_trans); 34919ca9ee09SSage Weil 34922a79f17eSAl Viro mnt_drop_write_file(file); 34931ab86aedSSage Weil return 0; 3494f46b5a66SChristoph Hellwig } 3495f46b5a66SChristoph Hellwig 34969a8c28beSMiao Xie static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, 34979a8c28beSMiao Xie void __user *argp) 349846204592SSage Weil { 349946204592SSage Weil struct btrfs_trans_handle *trans; 350046204592SSage Weil u64 transid; 3501db5b493aSTsutomu Itoh int ret; 350246204592SSage Weil 3503d4edf39bSMiao Xie trans = btrfs_attach_transaction_barrier(root); 3504ff7c1d33SMiao Xie if (IS_ERR(trans)) { 3505ff7c1d33SMiao Xie if (PTR_ERR(trans) != -ENOENT) 350698d5dc13STsutomu Itoh return PTR_ERR(trans); 3507ff7c1d33SMiao Xie 3508ff7c1d33SMiao Xie /* No running transaction, don't bother */ 3509ff7c1d33SMiao Xie transid = root->fs_info->last_trans_committed; 3510ff7c1d33SMiao Xie goto out; 3511ff7c1d33SMiao Xie } 351246204592SSage Weil transid = trans->transid; 3513db5b493aSTsutomu Itoh ret = btrfs_commit_transaction_async(trans, root, 0); 35148b2b2d3cSTsutomu Itoh if (ret) { 35158b2b2d3cSTsutomu Itoh btrfs_end_transaction(trans, root); 3516db5b493aSTsutomu Itoh return ret; 35178b2b2d3cSTsutomu Itoh } 3518ff7c1d33SMiao Xie out: 351946204592SSage Weil if (argp) 352046204592SSage Weil if (copy_to_user(argp, &transid, sizeof(transid))) 352146204592SSage Weil return -EFAULT; 352246204592SSage Weil return 0; 352346204592SSage Weil } 352446204592SSage Weil 35259a8c28beSMiao Xie static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root, 35269a8c28beSMiao Xie void __user *argp) 352746204592SSage Weil { 352846204592SSage Weil u64 transid; 352946204592SSage Weil 353046204592SSage Weil if (argp) { 353146204592SSage Weil if (copy_from_user(&transid, argp, sizeof(transid))) 353246204592SSage Weil return -EFAULT; 353346204592SSage Weil } else { 353446204592SSage Weil transid = 0; /* current trans */ 353546204592SSage Weil } 353646204592SSage Weil return btrfs_wait_for_commit(root, transid); 353746204592SSage Weil } 353846204592SSage Weil 3539b8e95489SMiao Xie static long btrfs_ioctl_scrub(struct file *file, void __user *arg) 3540475f6387SJan Schmidt { 3541496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 3542475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 3543b8e95489SMiao Xie int ret; 3544475f6387SJan Schmidt 3545475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 3546475f6387SJan Schmidt return -EPERM; 3547475f6387SJan Schmidt 3548475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 3549475f6387SJan Schmidt if (IS_ERR(sa)) 3550475f6387SJan Schmidt return PTR_ERR(sa); 3551475f6387SJan Schmidt 3552b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) { 3553b8e95489SMiao Xie ret = mnt_want_write_file(file); 3554b8e95489SMiao Xie if (ret) 3555b8e95489SMiao Xie goto out; 3556b8e95489SMiao Xie } 3557b8e95489SMiao Xie 3558aa1b8cd4SStefan Behrens ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end, 355963a212abSStefan Behrens &sa->progress, sa->flags & BTRFS_SCRUB_READONLY, 356063a212abSStefan Behrens 0); 3561475f6387SJan Schmidt 3562475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 3563475f6387SJan Schmidt ret = -EFAULT; 3564475f6387SJan Schmidt 3565b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) 3566b8e95489SMiao Xie mnt_drop_write_file(file); 3567b8e95489SMiao Xie out: 3568475f6387SJan Schmidt kfree(sa); 3569475f6387SJan Schmidt return ret; 3570475f6387SJan Schmidt } 3571475f6387SJan Schmidt 3572475f6387SJan Schmidt static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg) 3573475f6387SJan Schmidt { 3574475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 3575475f6387SJan Schmidt return -EPERM; 3576475f6387SJan Schmidt 3577aa1b8cd4SStefan Behrens return btrfs_scrub_cancel(root->fs_info); 3578475f6387SJan Schmidt } 3579475f6387SJan Schmidt 3580475f6387SJan Schmidt static long btrfs_ioctl_scrub_progress(struct btrfs_root *root, 3581475f6387SJan Schmidt void __user *arg) 3582475f6387SJan Schmidt { 3583475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 3584475f6387SJan Schmidt int ret; 3585475f6387SJan Schmidt 3586475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 3587475f6387SJan Schmidt return -EPERM; 3588475f6387SJan Schmidt 3589475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 3590475f6387SJan Schmidt if (IS_ERR(sa)) 3591475f6387SJan Schmidt return PTR_ERR(sa); 3592475f6387SJan Schmidt 3593475f6387SJan Schmidt ret = btrfs_scrub_progress(root, sa->devid, &sa->progress); 3594475f6387SJan Schmidt 3595475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 3596475f6387SJan Schmidt ret = -EFAULT; 3597475f6387SJan Schmidt 3598475f6387SJan Schmidt kfree(sa); 3599475f6387SJan Schmidt return ret; 3600475f6387SJan Schmidt } 3601475f6387SJan Schmidt 3602c11d2c23SStefan Behrens static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root, 3603b27f7c0cSDavid Sterba void __user *arg) 3604c11d2c23SStefan Behrens { 3605c11d2c23SStefan Behrens struct btrfs_ioctl_get_dev_stats *sa; 3606c11d2c23SStefan Behrens int ret; 3607c11d2c23SStefan Behrens 3608c11d2c23SStefan Behrens sa = memdup_user(arg, sizeof(*sa)); 3609c11d2c23SStefan Behrens if (IS_ERR(sa)) 3610c11d2c23SStefan Behrens return PTR_ERR(sa); 3611c11d2c23SStefan Behrens 3612b27f7c0cSDavid Sterba if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) { 3613b27f7c0cSDavid Sterba kfree(sa); 3614b27f7c0cSDavid Sterba return -EPERM; 3615b27f7c0cSDavid Sterba } 3616b27f7c0cSDavid Sterba 3617b27f7c0cSDavid Sterba ret = btrfs_get_dev_stats(root, sa); 3618c11d2c23SStefan Behrens 3619c11d2c23SStefan Behrens if (copy_to_user(arg, sa, sizeof(*sa))) 3620c11d2c23SStefan Behrens ret = -EFAULT; 3621c11d2c23SStefan Behrens 3622c11d2c23SStefan Behrens kfree(sa); 3623c11d2c23SStefan Behrens return ret; 3624c11d2c23SStefan Behrens } 3625c11d2c23SStefan Behrens 36263f6bcfbdSStefan Behrens static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg) 36273f6bcfbdSStefan Behrens { 36283f6bcfbdSStefan Behrens struct btrfs_ioctl_dev_replace_args *p; 36293f6bcfbdSStefan Behrens int ret; 36303f6bcfbdSStefan Behrens 36313f6bcfbdSStefan Behrens if (!capable(CAP_SYS_ADMIN)) 36323f6bcfbdSStefan Behrens return -EPERM; 36333f6bcfbdSStefan Behrens 36343f6bcfbdSStefan Behrens p = memdup_user(arg, sizeof(*p)); 36353f6bcfbdSStefan Behrens if (IS_ERR(p)) 36363f6bcfbdSStefan Behrens return PTR_ERR(p); 36373f6bcfbdSStefan Behrens 36383f6bcfbdSStefan Behrens switch (p->cmd) { 36393f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_START: 36403f6bcfbdSStefan Behrens if (atomic_xchg( 36413f6bcfbdSStefan Behrens &root->fs_info->mutually_exclusive_operation_running, 36423f6bcfbdSStefan Behrens 1)) { 36433f6bcfbdSStefan Behrens pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n"); 36443f6bcfbdSStefan Behrens ret = -EINPROGRESS; 36453f6bcfbdSStefan Behrens } else { 36463f6bcfbdSStefan Behrens ret = btrfs_dev_replace_start(root, p); 36473f6bcfbdSStefan Behrens atomic_set( 36483f6bcfbdSStefan Behrens &root->fs_info->mutually_exclusive_operation_running, 36493f6bcfbdSStefan Behrens 0); 36503f6bcfbdSStefan Behrens } 36513f6bcfbdSStefan Behrens break; 36523f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS: 36533f6bcfbdSStefan Behrens btrfs_dev_replace_status(root->fs_info, p); 36543f6bcfbdSStefan Behrens ret = 0; 36553f6bcfbdSStefan Behrens break; 36563f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL: 36573f6bcfbdSStefan Behrens ret = btrfs_dev_replace_cancel(root->fs_info, p); 36583f6bcfbdSStefan Behrens break; 36593f6bcfbdSStefan Behrens default: 36603f6bcfbdSStefan Behrens ret = -EINVAL; 36613f6bcfbdSStefan Behrens break; 36623f6bcfbdSStefan Behrens } 36633f6bcfbdSStefan Behrens 36643f6bcfbdSStefan Behrens if (copy_to_user(arg, p, sizeof(*p))) 36653f6bcfbdSStefan Behrens ret = -EFAULT; 36663f6bcfbdSStefan Behrens 36673f6bcfbdSStefan Behrens kfree(p); 36683f6bcfbdSStefan Behrens return ret; 36693f6bcfbdSStefan Behrens } 36703f6bcfbdSStefan Behrens 3671d7728c96SJan Schmidt static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) 3672d7728c96SJan Schmidt { 3673d7728c96SJan Schmidt int ret = 0; 3674d7728c96SJan Schmidt int i; 3675740c3d22SChris Mason u64 rel_ptr; 3676d7728c96SJan Schmidt int size; 3677806468f8SChris Mason struct btrfs_ioctl_ino_path_args *ipa = NULL; 3678d7728c96SJan Schmidt struct inode_fs_paths *ipath = NULL; 3679d7728c96SJan Schmidt struct btrfs_path *path; 3680d7728c96SJan Schmidt 368182b22ac8SKusanagi Kouichi if (!capable(CAP_DAC_READ_SEARCH)) 3682d7728c96SJan Schmidt return -EPERM; 3683d7728c96SJan Schmidt 3684d7728c96SJan Schmidt path = btrfs_alloc_path(); 3685d7728c96SJan Schmidt if (!path) { 3686d7728c96SJan Schmidt ret = -ENOMEM; 3687d7728c96SJan Schmidt goto out; 3688d7728c96SJan Schmidt } 3689d7728c96SJan Schmidt 3690d7728c96SJan Schmidt ipa = memdup_user(arg, sizeof(*ipa)); 3691d7728c96SJan Schmidt if (IS_ERR(ipa)) { 3692d7728c96SJan Schmidt ret = PTR_ERR(ipa); 3693d7728c96SJan Schmidt ipa = NULL; 3694d7728c96SJan Schmidt goto out; 3695d7728c96SJan Schmidt } 3696d7728c96SJan Schmidt 3697d7728c96SJan Schmidt size = min_t(u32, ipa->size, 4096); 3698d7728c96SJan Schmidt ipath = init_ipath(size, root, path); 3699d7728c96SJan Schmidt if (IS_ERR(ipath)) { 3700d7728c96SJan Schmidt ret = PTR_ERR(ipath); 3701d7728c96SJan Schmidt ipath = NULL; 3702d7728c96SJan Schmidt goto out; 3703d7728c96SJan Schmidt } 3704d7728c96SJan Schmidt 3705d7728c96SJan Schmidt ret = paths_from_inode(ipa->inum, ipath); 3706d7728c96SJan Schmidt if (ret < 0) 3707d7728c96SJan Schmidt goto out; 3708d7728c96SJan Schmidt 3709d7728c96SJan Schmidt for (i = 0; i < ipath->fspath->elem_cnt; ++i) { 3710745c4d8eSJeff Mahoney rel_ptr = ipath->fspath->val[i] - 3711745c4d8eSJeff Mahoney (u64)(unsigned long)ipath->fspath->val; 3712740c3d22SChris Mason ipath->fspath->val[i] = rel_ptr; 3713d7728c96SJan Schmidt } 3714d7728c96SJan Schmidt 3715745c4d8eSJeff Mahoney ret = copy_to_user((void *)(unsigned long)ipa->fspath, 3716745c4d8eSJeff Mahoney (void *)(unsigned long)ipath->fspath, size); 3717d7728c96SJan Schmidt if (ret) { 3718d7728c96SJan Schmidt ret = -EFAULT; 3719d7728c96SJan Schmidt goto out; 3720d7728c96SJan Schmidt } 3721d7728c96SJan Schmidt 3722d7728c96SJan Schmidt out: 3723d7728c96SJan Schmidt btrfs_free_path(path); 3724d7728c96SJan Schmidt free_ipath(ipath); 3725d7728c96SJan Schmidt kfree(ipa); 3726d7728c96SJan Schmidt 3727d7728c96SJan Schmidt return ret; 3728d7728c96SJan Schmidt } 3729d7728c96SJan Schmidt 3730d7728c96SJan Schmidt static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx) 3731d7728c96SJan Schmidt { 3732d7728c96SJan Schmidt struct btrfs_data_container *inodes = ctx; 3733d7728c96SJan Schmidt const size_t c = 3 * sizeof(u64); 3734d7728c96SJan Schmidt 3735d7728c96SJan Schmidt if (inodes->bytes_left >= c) { 3736d7728c96SJan Schmidt inodes->bytes_left -= c; 3737d7728c96SJan Schmidt inodes->val[inodes->elem_cnt] = inum; 3738d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 1] = offset; 3739d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 2] = root; 3740d7728c96SJan Schmidt inodes->elem_cnt += 3; 3741d7728c96SJan Schmidt } else { 3742d7728c96SJan Schmidt inodes->bytes_missing += c - inodes->bytes_left; 3743d7728c96SJan Schmidt inodes->bytes_left = 0; 3744d7728c96SJan Schmidt inodes->elem_missed += 3; 3745d7728c96SJan Schmidt } 3746d7728c96SJan Schmidt 3747d7728c96SJan Schmidt return 0; 3748d7728c96SJan Schmidt } 3749d7728c96SJan Schmidt 3750d7728c96SJan Schmidt static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root, 3751d7728c96SJan Schmidt void __user *arg) 3752d7728c96SJan Schmidt { 3753d7728c96SJan Schmidt int ret = 0; 3754d7728c96SJan Schmidt int size; 3755d7728c96SJan Schmidt struct btrfs_ioctl_logical_ino_args *loi; 3756d7728c96SJan Schmidt struct btrfs_data_container *inodes = NULL; 3757d7728c96SJan Schmidt struct btrfs_path *path = NULL; 3758d7728c96SJan Schmidt 3759d7728c96SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 3760d7728c96SJan Schmidt return -EPERM; 3761d7728c96SJan Schmidt 3762d7728c96SJan Schmidt loi = memdup_user(arg, sizeof(*loi)); 3763d7728c96SJan Schmidt if (IS_ERR(loi)) { 3764d7728c96SJan Schmidt ret = PTR_ERR(loi); 3765d7728c96SJan Schmidt loi = NULL; 3766d7728c96SJan Schmidt goto out; 3767d7728c96SJan Schmidt } 3768d7728c96SJan Schmidt 3769d7728c96SJan Schmidt path = btrfs_alloc_path(); 3770d7728c96SJan Schmidt if (!path) { 3771d7728c96SJan Schmidt ret = -ENOMEM; 3772d7728c96SJan Schmidt goto out; 3773d7728c96SJan Schmidt } 3774d7728c96SJan Schmidt 3775425d17a2SLiu Bo size = min_t(u32, loi->size, 64 * 1024); 3776d7728c96SJan Schmidt inodes = init_data_container(size); 3777d7728c96SJan Schmidt if (IS_ERR(inodes)) { 3778d7728c96SJan Schmidt ret = PTR_ERR(inodes); 3779d7728c96SJan Schmidt inodes = NULL; 3780d7728c96SJan Schmidt goto out; 3781d7728c96SJan Schmidt } 3782d7728c96SJan Schmidt 3783df031f07SLiu Bo ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path, 3784df031f07SLiu Bo build_ino_list, inodes); 3785df031f07SLiu Bo if (ret == -EINVAL) 3786d7728c96SJan Schmidt ret = -ENOENT; 3787d7728c96SJan Schmidt if (ret < 0) 3788d7728c96SJan Schmidt goto out; 3789d7728c96SJan Schmidt 3790745c4d8eSJeff Mahoney ret = copy_to_user((void *)(unsigned long)loi->inodes, 3791745c4d8eSJeff Mahoney (void *)(unsigned long)inodes, size); 3792d7728c96SJan Schmidt if (ret) 3793d7728c96SJan Schmidt ret = -EFAULT; 3794d7728c96SJan Schmidt 3795d7728c96SJan Schmidt out: 3796d7728c96SJan Schmidt btrfs_free_path(path); 3797425d17a2SLiu Bo vfree(inodes); 3798d7728c96SJan Schmidt kfree(loi); 3799d7728c96SJan Schmidt 3800d7728c96SJan Schmidt return ret; 3801d7728c96SJan Schmidt } 3802d7728c96SJan Schmidt 380319a39dceSIlya Dryomov void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock, 3804c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs) 3805c9e9f97bSIlya Dryomov { 3806c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl = fs_info->balance_ctl; 3807c9e9f97bSIlya Dryomov 3808c9e9f97bSIlya Dryomov bargs->flags = bctl->flags; 3809c9e9f97bSIlya Dryomov 3810837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_running)) 3811837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_RUNNING; 3812837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_pause_req)) 3813837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; 3814a7e99c69SIlya Dryomov if (atomic_read(&fs_info->balance_cancel_req)) 3815a7e99c69SIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; 3816837d5b6eSIlya Dryomov 3817c9e9f97bSIlya Dryomov memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); 3818c9e9f97bSIlya Dryomov memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); 3819c9e9f97bSIlya Dryomov memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); 382019a39dceSIlya Dryomov 382119a39dceSIlya Dryomov if (lock) { 382219a39dceSIlya Dryomov spin_lock(&fs_info->balance_lock); 382319a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 382419a39dceSIlya Dryomov spin_unlock(&fs_info->balance_lock); 382519a39dceSIlya Dryomov } else { 382619a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 382719a39dceSIlya Dryomov } 3828c9e9f97bSIlya Dryomov } 3829c9e9f97bSIlya Dryomov 38309ba1f6e4SLiu Bo static long btrfs_ioctl_balance(struct file *file, void __user *arg) 3831c9e9f97bSIlya Dryomov { 3832496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 3833c9e9f97bSIlya Dryomov struct btrfs_fs_info *fs_info = root->fs_info; 3834c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 3835c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl; 3836ed0fb78fSIlya Dryomov bool need_unlock; /* for mut. excl. ops lock */ 3837c9e9f97bSIlya Dryomov int ret; 3838c9e9f97bSIlya Dryomov 3839c9e9f97bSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 3840c9e9f97bSIlya Dryomov return -EPERM; 3841c9e9f97bSIlya Dryomov 3842e54bfa31SLiu Bo ret = mnt_want_write_file(file); 38439ba1f6e4SLiu Bo if (ret) 38449ba1f6e4SLiu Bo return ret; 38459ba1f6e4SLiu Bo 3846ed0fb78fSIlya Dryomov again: 3847ed0fb78fSIlya Dryomov if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) { 3848c9e9f97bSIlya Dryomov mutex_lock(&fs_info->volume_mutex); 3849c9e9f97bSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 3850ed0fb78fSIlya Dryomov need_unlock = true; 3851ed0fb78fSIlya Dryomov goto locked; 3852ed0fb78fSIlya Dryomov } 3853ed0fb78fSIlya Dryomov 3854ed0fb78fSIlya Dryomov /* 3855ed0fb78fSIlya Dryomov * mut. excl. ops lock is locked. Three possibilites: 3856ed0fb78fSIlya Dryomov * (1) some other op is running 3857ed0fb78fSIlya Dryomov * (2) balance is running 3858ed0fb78fSIlya Dryomov * (3) balance is paused -- special case (think resume) 3859ed0fb78fSIlya Dryomov */ 3860ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 3861ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 3862ed0fb78fSIlya Dryomov /* this is either (2) or (3) */ 3863ed0fb78fSIlya Dryomov if (!atomic_read(&fs_info->balance_running)) { 3864ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 3865ed0fb78fSIlya Dryomov if (!mutex_trylock(&fs_info->volume_mutex)) 3866ed0fb78fSIlya Dryomov goto again; 3867ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 3868ed0fb78fSIlya Dryomov 3869ed0fb78fSIlya Dryomov if (fs_info->balance_ctl && 3870ed0fb78fSIlya Dryomov !atomic_read(&fs_info->balance_running)) { 3871ed0fb78fSIlya Dryomov /* this is (3) */ 3872ed0fb78fSIlya Dryomov need_unlock = false; 3873ed0fb78fSIlya Dryomov goto locked; 3874ed0fb78fSIlya Dryomov } 3875ed0fb78fSIlya Dryomov 3876ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 3877ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->volume_mutex); 3878ed0fb78fSIlya Dryomov goto again; 3879ed0fb78fSIlya Dryomov } else { 3880ed0fb78fSIlya Dryomov /* this is (2) */ 3881ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 3882ed0fb78fSIlya Dryomov ret = -EINPROGRESS; 3883ed0fb78fSIlya Dryomov goto out; 3884ed0fb78fSIlya Dryomov } 3885ed0fb78fSIlya Dryomov } else { 3886ed0fb78fSIlya Dryomov /* this is (1) */ 3887ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 3888ed0fb78fSIlya Dryomov pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n"); 3889ed0fb78fSIlya Dryomov ret = -EINVAL; 3890ed0fb78fSIlya Dryomov goto out; 3891ed0fb78fSIlya Dryomov } 3892ed0fb78fSIlya Dryomov 3893ed0fb78fSIlya Dryomov locked: 3894ed0fb78fSIlya Dryomov BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running)); 3895c9e9f97bSIlya Dryomov 3896c9e9f97bSIlya Dryomov if (arg) { 3897c9e9f97bSIlya Dryomov bargs = memdup_user(arg, sizeof(*bargs)); 3898c9e9f97bSIlya Dryomov if (IS_ERR(bargs)) { 3899c9e9f97bSIlya Dryomov ret = PTR_ERR(bargs); 3900ed0fb78fSIlya Dryomov goto out_unlock; 3901c9e9f97bSIlya Dryomov } 3902de322263SIlya Dryomov 3903de322263SIlya Dryomov if (bargs->flags & BTRFS_BALANCE_RESUME) { 3904de322263SIlya Dryomov if (!fs_info->balance_ctl) { 3905de322263SIlya Dryomov ret = -ENOTCONN; 3906de322263SIlya Dryomov goto out_bargs; 3907de322263SIlya Dryomov } 3908de322263SIlya Dryomov 3909de322263SIlya Dryomov bctl = fs_info->balance_ctl; 3910de322263SIlya Dryomov spin_lock(&fs_info->balance_lock); 3911de322263SIlya Dryomov bctl->flags |= BTRFS_BALANCE_RESUME; 3912de322263SIlya Dryomov spin_unlock(&fs_info->balance_lock); 3913de322263SIlya Dryomov 3914de322263SIlya Dryomov goto do_balance; 3915de322263SIlya Dryomov } 3916c9e9f97bSIlya Dryomov } else { 3917c9e9f97bSIlya Dryomov bargs = NULL; 3918c9e9f97bSIlya Dryomov } 3919c9e9f97bSIlya Dryomov 3920ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 3921837d5b6eSIlya Dryomov ret = -EINPROGRESS; 3922837d5b6eSIlya Dryomov goto out_bargs; 3923837d5b6eSIlya Dryomov } 3924837d5b6eSIlya Dryomov 3925c9e9f97bSIlya Dryomov bctl = kzalloc(sizeof(*bctl), GFP_NOFS); 3926c9e9f97bSIlya Dryomov if (!bctl) { 3927c9e9f97bSIlya Dryomov ret = -ENOMEM; 3928c9e9f97bSIlya Dryomov goto out_bargs; 3929c9e9f97bSIlya Dryomov } 3930c9e9f97bSIlya Dryomov 3931c9e9f97bSIlya Dryomov bctl->fs_info = fs_info; 3932c9e9f97bSIlya Dryomov if (arg) { 3933c9e9f97bSIlya Dryomov memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); 3934c9e9f97bSIlya Dryomov memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); 3935c9e9f97bSIlya Dryomov memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); 3936c9e9f97bSIlya Dryomov 3937c9e9f97bSIlya Dryomov bctl->flags = bargs->flags; 3938f43ffb60SIlya Dryomov } else { 3939f43ffb60SIlya Dryomov /* balance everything - no filters */ 3940f43ffb60SIlya Dryomov bctl->flags |= BTRFS_BALANCE_TYPE_MASK; 3941c9e9f97bSIlya Dryomov } 3942c9e9f97bSIlya Dryomov 3943de322263SIlya Dryomov do_balance: 3944c9e9f97bSIlya Dryomov /* 3945ed0fb78fSIlya Dryomov * Ownership of bctl and mutually_exclusive_operation_running 3946ed0fb78fSIlya Dryomov * goes to to btrfs_balance. bctl is freed in __cancel_balance, 3947ed0fb78fSIlya Dryomov * or, if restriper was paused all the way until unmount, in 3948ed0fb78fSIlya Dryomov * free_fs_info. mutually_exclusive_operation_running is 3949ed0fb78fSIlya Dryomov * cleared in __cancel_balance. 3950c9e9f97bSIlya Dryomov */ 3951ed0fb78fSIlya Dryomov need_unlock = false; 3952ed0fb78fSIlya Dryomov 3953ed0fb78fSIlya Dryomov ret = btrfs_balance(bctl, bargs); 3954ed0fb78fSIlya Dryomov 3955c9e9f97bSIlya Dryomov if (arg) { 3956c9e9f97bSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 3957c9e9f97bSIlya Dryomov ret = -EFAULT; 3958c9e9f97bSIlya Dryomov } 3959c9e9f97bSIlya Dryomov 3960c9e9f97bSIlya Dryomov out_bargs: 3961c9e9f97bSIlya Dryomov kfree(bargs); 3962ed0fb78fSIlya Dryomov out_unlock: 3963c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 3964c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->volume_mutex); 3965ed0fb78fSIlya Dryomov if (need_unlock) 3966ed0fb78fSIlya Dryomov atomic_set(&fs_info->mutually_exclusive_operation_running, 0); 3967ed0fb78fSIlya Dryomov out: 3968e54bfa31SLiu Bo mnt_drop_write_file(file); 3969c9e9f97bSIlya Dryomov return ret; 3970c9e9f97bSIlya Dryomov } 3971c9e9f97bSIlya Dryomov 3972837d5b6eSIlya Dryomov static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd) 3973837d5b6eSIlya Dryomov { 3974837d5b6eSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 3975837d5b6eSIlya Dryomov return -EPERM; 3976837d5b6eSIlya Dryomov 3977837d5b6eSIlya Dryomov switch (cmd) { 3978837d5b6eSIlya Dryomov case BTRFS_BALANCE_CTL_PAUSE: 3979837d5b6eSIlya Dryomov return btrfs_pause_balance(root->fs_info); 3980a7e99c69SIlya Dryomov case BTRFS_BALANCE_CTL_CANCEL: 3981a7e99c69SIlya Dryomov return btrfs_cancel_balance(root->fs_info); 3982837d5b6eSIlya Dryomov } 3983837d5b6eSIlya Dryomov 3984837d5b6eSIlya Dryomov return -EINVAL; 3985837d5b6eSIlya Dryomov } 3986837d5b6eSIlya Dryomov 398719a39dceSIlya Dryomov static long btrfs_ioctl_balance_progress(struct btrfs_root *root, 398819a39dceSIlya Dryomov void __user *arg) 398919a39dceSIlya Dryomov { 399019a39dceSIlya Dryomov struct btrfs_fs_info *fs_info = root->fs_info; 399119a39dceSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 399219a39dceSIlya Dryomov int ret = 0; 399319a39dceSIlya Dryomov 399419a39dceSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 399519a39dceSIlya Dryomov return -EPERM; 399619a39dceSIlya Dryomov 399719a39dceSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 399819a39dceSIlya Dryomov if (!fs_info->balance_ctl) { 399919a39dceSIlya Dryomov ret = -ENOTCONN; 400019a39dceSIlya Dryomov goto out; 400119a39dceSIlya Dryomov } 400219a39dceSIlya Dryomov 400319a39dceSIlya Dryomov bargs = kzalloc(sizeof(*bargs), GFP_NOFS); 400419a39dceSIlya Dryomov if (!bargs) { 400519a39dceSIlya Dryomov ret = -ENOMEM; 400619a39dceSIlya Dryomov goto out; 400719a39dceSIlya Dryomov } 400819a39dceSIlya Dryomov 400919a39dceSIlya Dryomov update_ioctl_balance_args(fs_info, 1, bargs); 401019a39dceSIlya Dryomov 401119a39dceSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 401219a39dceSIlya Dryomov ret = -EFAULT; 401319a39dceSIlya Dryomov 401419a39dceSIlya Dryomov kfree(bargs); 401519a39dceSIlya Dryomov out: 401619a39dceSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 401719a39dceSIlya Dryomov return ret; 401819a39dceSIlya Dryomov } 401919a39dceSIlya Dryomov 4020905b0ddaSMiao Xie static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) 40215d13a37bSArne Jansen { 4022496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 40235d13a37bSArne Jansen struct btrfs_ioctl_quota_ctl_args *sa; 40245d13a37bSArne Jansen struct btrfs_trans_handle *trans = NULL; 40255d13a37bSArne Jansen int ret; 40265d13a37bSArne Jansen int err; 40275d13a37bSArne Jansen 40285d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 40295d13a37bSArne Jansen return -EPERM; 40305d13a37bSArne Jansen 4031905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4032905b0ddaSMiao Xie if (ret) 4033905b0ddaSMiao Xie return ret; 40345d13a37bSArne Jansen 40355d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4036905b0ddaSMiao Xie if (IS_ERR(sa)) { 4037905b0ddaSMiao Xie ret = PTR_ERR(sa); 4038905b0ddaSMiao Xie goto drop_write; 4039905b0ddaSMiao Xie } 40405d13a37bSArne Jansen 40417708f029SWang Shilong down_write(&root->fs_info->subvol_sem); 404292f183aaSWang Shilong trans = btrfs_start_transaction(root->fs_info->tree_root, 2); 40435d13a37bSArne Jansen if (IS_ERR(trans)) { 40445d13a37bSArne Jansen ret = PTR_ERR(trans); 40455d13a37bSArne Jansen goto out; 40465d13a37bSArne Jansen } 40475d13a37bSArne Jansen 40485d13a37bSArne Jansen switch (sa->cmd) { 40495d13a37bSArne Jansen case BTRFS_QUOTA_CTL_ENABLE: 40505d13a37bSArne Jansen ret = btrfs_quota_enable(trans, root->fs_info); 40515d13a37bSArne Jansen break; 40525d13a37bSArne Jansen case BTRFS_QUOTA_CTL_DISABLE: 40535d13a37bSArne Jansen ret = btrfs_quota_disable(trans, root->fs_info); 40545d13a37bSArne Jansen break; 40555d13a37bSArne Jansen default: 40565d13a37bSArne Jansen ret = -EINVAL; 40575d13a37bSArne Jansen break; 40585d13a37bSArne Jansen } 40595d13a37bSArne Jansen 406092f183aaSWang Shilong err = btrfs_commit_transaction(trans, root->fs_info->tree_root); 40615d13a37bSArne Jansen if (err && !ret) 40625d13a37bSArne Jansen ret = err; 40635d13a37bSArne Jansen out: 40645d13a37bSArne Jansen kfree(sa); 40657708f029SWang Shilong up_write(&root->fs_info->subvol_sem); 4066905b0ddaSMiao Xie drop_write: 4067905b0ddaSMiao Xie mnt_drop_write_file(file); 40685d13a37bSArne Jansen return ret; 40695d13a37bSArne Jansen } 40705d13a37bSArne Jansen 4071905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg) 40725d13a37bSArne Jansen { 4073496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 40745d13a37bSArne Jansen struct btrfs_ioctl_qgroup_assign_args *sa; 40755d13a37bSArne Jansen struct btrfs_trans_handle *trans; 40765d13a37bSArne Jansen int ret; 40775d13a37bSArne Jansen int err; 40785d13a37bSArne Jansen 40795d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 40805d13a37bSArne Jansen return -EPERM; 40815d13a37bSArne Jansen 4082905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4083905b0ddaSMiao Xie if (ret) 4084905b0ddaSMiao Xie return ret; 40855d13a37bSArne Jansen 40865d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4087905b0ddaSMiao Xie if (IS_ERR(sa)) { 4088905b0ddaSMiao Xie ret = PTR_ERR(sa); 4089905b0ddaSMiao Xie goto drop_write; 4090905b0ddaSMiao Xie } 40915d13a37bSArne Jansen 40925d13a37bSArne Jansen trans = btrfs_join_transaction(root); 40935d13a37bSArne Jansen if (IS_ERR(trans)) { 40945d13a37bSArne Jansen ret = PTR_ERR(trans); 40955d13a37bSArne Jansen goto out; 40965d13a37bSArne Jansen } 40975d13a37bSArne Jansen 40985d13a37bSArne Jansen /* FIXME: check if the IDs really exist */ 40995d13a37bSArne Jansen if (sa->assign) { 41005d13a37bSArne Jansen ret = btrfs_add_qgroup_relation(trans, root->fs_info, 41015d13a37bSArne Jansen sa->src, sa->dst); 41025d13a37bSArne Jansen } else { 41035d13a37bSArne Jansen ret = btrfs_del_qgroup_relation(trans, root->fs_info, 41045d13a37bSArne Jansen sa->src, sa->dst); 41055d13a37bSArne Jansen } 41065d13a37bSArne Jansen 41075d13a37bSArne Jansen err = btrfs_end_transaction(trans, root); 41085d13a37bSArne Jansen if (err && !ret) 41095d13a37bSArne Jansen ret = err; 41105d13a37bSArne Jansen 41115d13a37bSArne Jansen out: 41125d13a37bSArne Jansen kfree(sa); 4113905b0ddaSMiao Xie drop_write: 4114905b0ddaSMiao Xie mnt_drop_write_file(file); 41155d13a37bSArne Jansen return ret; 41165d13a37bSArne Jansen } 41175d13a37bSArne Jansen 4118905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg) 41195d13a37bSArne Jansen { 4120496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 41215d13a37bSArne Jansen struct btrfs_ioctl_qgroup_create_args *sa; 41225d13a37bSArne Jansen struct btrfs_trans_handle *trans; 41235d13a37bSArne Jansen int ret; 41245d13a37bSArne Jansen int err; 41255d13a37bSArne Jansen 41265d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 41275d13a37bSArne Jansen return -EPERM; 41285d13a37bSArne Jansen 4129905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4130905b0ddaSMiao Xie if (ret) 4131905b0ddaSMiao Xie return ret; 41325d13a37bSArne Jansen 41335d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4134905b0ddaSMiao Xie if (IS_ERR(sa)) { 4135905b0ddaSMiao Xie ret = PTR_ERR(sa); 4136905b0ddaSMiao Xie goto drop_write; 4137905b0ddaSMiao Xie } 41385d13a37bSArne Jansen 4139d86e56cfSMiao Xie if (!sa->qgroupid) { 4140d86e56cfSMiao Xie ret = -EINVAL; 4141d86e56cfSMiao Xie goto out; 4142d86e56cfSMiao Xie } 4143d86e56cfSMiao Xie 41445d13a37bSArne Jansen trans = btrfs_join_transaction(root); 41455d13a37bSArne Jansen if (IS_ERR(trans)) { 41465d13a37bSArne Jansen ret = PTR_ERR(trans); 41475d13a37bSArne Jansen goto out; 41485d13a37bSArne Jansen } 41495d13a37bSArne Jansen 41505d13a37bSArne Jansen /* FIXME: check if the IDs really exist */ 41515d13a37bSArne Jansen if (sa->create) { 41525d13a37bSArne Jansen ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid, 41535d13a37bSArne Jansen NULL); 41545d13a37bSArne Jansen } else { 41555d13a37bSArne Jansen ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid); 41565d13a37bSArne Jansen } 41575d13a37bSArne Jansen 41585d13a37bSArne Jansen err = btrfs_end_transaction(trans, root); 41595d13a37bSArne Jansen if (err && !ret) 41605d13a37bSArne Jansen ret = err; 41615d13a37bSArne Jansen 41625d13a37bSArne Jansen out: 41635d13a37bSArne Jansen kfree(sa); 4164905b0ddaSMiao Xie drop_write: 4165905b0ddaSMiao Xie mnt_drop_write_file(file); 41665d13a37bSArne Jansen return ret; 41675d13a37bSArne Jansen } 41685d13a37bSArne Jansen 4169905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg) 41705d13a37bSArne Jansen { 4171496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 41725d13a37bSArne Jansen struct btrfs_ioctl_qgroup_limit_args *sa; 41735d13a37bSArne Jansen struct btrfs_trans_handle *trans; 41745d13a37bSArne Jansen int ret; 41755d13a37bSArne Jansen int err; 41765d13a37bSArne Jansen u64 qgroupid; 41775d13a37bSArne Jansen 41785d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 41795d13a37bSArne Jansen return -EPERM; 41805d13a37bSArne Jansen 4181905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4182905b0ddaSMiao Xie if (ret) 4183905b0ddaSMiao Xie return ret; 41845d13a37bSArne Jansen 41855d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4186905b0ddaSMiao Xie if (IS_ERR(sa)) { 4187905b0ddaSMiao Xie ret = PTR_ERR(sa); 4188905b0ddaSMiao Xie goto drop_write; 4189905b0ddaSMiao Xie } 41905d13a37bSArne Jansen 41915d13a37bSArne Jansen trans = btrfs_join_transaction(root); 41925d13a37bSArne Jansen if (IS_ERR(trans)) { 41935d13a37bSArne Jansen ret = PTR_ERR(trans); 41945d13a37bSArne Jansen goto out; 41955d13a37bSArne Jansen } 41965d13a37bSArne Jansen 41975d13a37bSArne Jansen qgroupid = sa->qgroupid; 41985d13a37bSArne Jansen if (!qgroupid) { 41995d13a37bSArne Jansen /* take the current subvol as qgroup */ 42005d13a37bSArne Jansen qgroupid = root->root_key.objectid; 42015d13a37bSArne Jansen } 42025d13a37bSArne Jansen 42035d13a37bSArne Jansen /* FIXME: check if the IDs really exist */ 42045d13a37bSArne Jansen ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim); 42055d13a37bSArne Jansen 42065d13a37bSArne Jansen err = btrfs_end_transaction(trans, root); 42075d13a37bSArne Jansen if (err && !ret) 42085d13a37bSArne Jansen ret = err; 42095d13a37bSArne Jansen 42105d13a37bSArne Jansen out: 42115d13a37bSArne Jansen kfree(sa); 4212905b0ddaSMiao Xie drop_write: 4213905b0ddaSMiao Xie mnt_drop_write_file(file); 42145d13a37bSArne Jansen return ret; 42155d13a37bSArne Jansen } 42165d13a37bSArne Jansen 42172f232036SJan Schmidt static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) 42182f232036SJan Schmidt { 42196d0379ecSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 42202f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 42212f232036SJan Schmidt int ret; 42222f232036SJan Schmidt 42232f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 42242f232036SJan Schmidt return -EPERM; 42252f232036SJan Schmidt 42262f232036SJan Schmidt ret = mnt_want_write_file(file); 42272f232036SJan Schmidt if (ret) 42282f232036SJan Schmidt return ret; 42292f232036SJan Schmidt 42302f232036SJan Schmidt qsa = memdup_user(arg, sizeof(*qsa)); 42312f232036SJan Schmidt if (IS_ERR(qsa)) { 42322f232036SJan Schmidt ret = PTR_ERR(qsa); 42332f232036SJan Schmidt goto drop_write; 42342f232036SJan Schmidt } 42352f232036SJan Schmidt 42362f232036SJan Schmidt if (qsa->flags) { 42372f232036SJan Schmidt ret = -EINVAL; 42382f232036SJan Schmidt goto out; 42392f232036SJan Schmidt } 42402f232036SJan Schmidt 42412f232036SJan Schmidt ret = btrfs_qgroup_rescan(root->fs_info); 42422f232036SJan Schmidt 42432f232036SJan Schmidt out: 42442f232036SJan Schmidt kfree(qsa); 42452f232036SJan Schmidt drop_write: 42462f232036SJan Schmidt mnt_drop_write_file(file); 42472f232036SJan Schmidt return ret; 42482f232036SJan Schmidt } 42492f232036SJan Schmidt 42502f232036SJan Schmidt static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg) 42512f232036SJan Schmidt { 42526d0379ecSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 42532f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 42542f232036SJan Schmidt int ret = 0; 42552f232036SJan Schmidt 42562f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 42572f232036SJan Schmidt return -EPERM; 42582f232036SJan Schmidt 42592f232036SJan Schmidt qsa = kzalloc(sizeof(*qsa), GFP_NOFS); 42602f232036SJan Schmidt if (!qsa) 42612f232036SJan Schmidt return -ENOMEM; 42622f232036SJan Schmidt 42632f232036SJan Schmidt if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { 42642f232036SJan Schmidt qsa->flags = 1; 42652f232036SJan Schmidt qsa->progress = root->fs_info->qgroup_rescan_progress.objectid; 42662f232036SJan Schmidt } 42672f232036SJan Schmidt 42682f232036SJan Schmidt if (copy_to_user(arg, qsa, sizeof(*qsa))) 42692f232036SJan Schmidt ret = -EFAULT; 42702f232036SJan Schmidt 42712f232036SJan Schmidt kfree(qsa); 42722f232036SJan Schmidt return ret; 42732f232036SJan Schmidt } 42742f232036SJan Schmidt 427557254b6eSJan Schmidt static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg) 427657254b6eSJan Schmidt { 427757254b6eSJan Schmidt struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; 427857254b6eSJan Schmidt 427957254b6eSJan Schmidt if (!capable(CAP_SYS_ADMIN)) 428057254b6eSJan Schmidt return -EPERM; 428157254b6eSJan Schmidt 428257254b6eSJan Schmidt return btrfs_qgroup_wait_for_completion(root->fs_info); 428357254b6eSJan Schmidt } 428457254b6eSJan Schmidt 42858ea05e3aSAlexander Block static long btrfs_ioctl_set_received_subvol(struct file *file, 42868ea05e3aSAlexander Block void __user *arg) 42878ea05e3aSAlexander Block { 42888ea05e3aSAlexander Block struct btrfs_ioctl_received_subvol_args *sa = NULL; 4289496ad9aaSAl Viro struct inode *inode = file_inode(file); 42908ea05e3aSAlexander Block struct btrfs_root *root = BTRFS_I(inode)->root; 42918ea05e3aSAlexander Block struct btrfs_root_item *root_item = &root->root_item; 42928ea05e3aSAlexander Block struct btrfs_trans_handle *trans; 42938ea05e3aSAlexander Block struct timespec ct = CURRENT_TIME; 42948ea05e3aSAlexander Block int ret = 0; 42958ea05e3aSAlexander Block 42968ea05e3aSAlexander Block ret = mnt_want_write_file(file); 42978ea05e3aSAlexander Block if (ret < 0) 42988ea05e3aSAlexander Block return ret; 42998ea05e3aSAlexander Block 43008ea05e3aSAlexander Block down_write(&root->fs_info->subvol_sem); 43018ea05e3aSAlexander Block 43028ea05e3aSAlexander Block if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { 43038ea05e3aSAlexander Block ret = -EINVAL; 43048ea05e3aSAlexander Block goto out; 43058ea05e3aSAlexander Block } 43068ea05e3aSAlexander Block 43078ea05e3aSAlexander Block if (btrfs_root_readonly(root)) { 43088ea05e3aSAlexander Block ret = -EROFS; 43098ea05e3aSAlexander Block goto out; 43108ea05e3aSAlexander Block } 43118ea05e3aSAlexander Block 43128ea05e3aSAlexander Block if (!inode_owner_or_capable(inode)) { 43138ea05e3aSAlexander Block ret = -EACCES; 43148ea05e3aSAlexander Block goto out; 43158ea05e3aSAlexander Block } 43168ea05e3aSAlexander Block 43178ea05e3aSAlexander Block sa = memdup_user(arg, sizeof(*sa)); 43188ea05e3aSAlexander Block if (IS_ERR(sa)) { 43198ea05e3aSAlexander Block ret = PTR_ERR(sa); 43208ea05e3aSAlexander Block sa = NULL; 43218ea05e3aSAlexander Block goto out; 43228ea05e3aSAlexander Block } 43238ea05e3aSAlexander Block 43248ea05e3aSAlexander Block trans = btrfs_start_transaction(root, 1); 43258ea05e3aSAlexander Block if (IS_ERR(trans)) { 43268ea05e3aSAlexander Block ret = PTR_ERR(trans); 43278ea05e3aSAlexander Block trans = NULL; 43288ea05e3aSAlexander Block goto out; 43298ea05e3aSAlexander Block } 43308ea05e3aSAlexander Block 43318ea05e3aSAlexander Block sa->rtransid = trans->transid; 43328ea05e3aSAlexander Block sa->rtime.sec = ct.tv_sec; 43338ea05e3aSAlexander Block sa->rtime.nsec = ct.tv_nsec; 43348ea05e3aSAlexander Block 43358ea05e3aSAlexander Block memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE); 43368ea05e3aSAlexander Block btrfs_set_root_stransid(root_item, sa->stransid); 43378ea05e3aSAlexander Block btrfs_set_root_rtransid(root_item, sa->rtransid); 43383cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec); 43393cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec); 43403cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec); 43413cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec); 43428ea05e3aSAlexander Block 43438ea05e3aSAlexander Block ret = btrfs_update_root(trans, root->fs_info->tree_root, 43448ea05e3aSAlexander Block &root->root_key, &root->root_item); 43458ea05e3aSAlexander Block if (ret < 0) { 43468ea05e3aSAlexander Block btrfs_end_transaction(trans, root); 43478ea05e3aSAlexander Block trans = NULL; 43488ea05e3aSAlexander Block goto out; 43498ea05e3aSAlexander Block } else { 43508ea05e3aSAlexander Block ret = btrfs_commit_transaction(trans, root); 43518ea05e3aSAlexander Block if (ret < 0) 43528ea05e3aSAlexander Block goto out; 43538ea05e3aSAlexander Block } 43548ea05e3aSAlexander Block 43558ea05e3aSAlexander Block ret = copy_to_user(arg, sa, sizeof(*sa)); 43568ea05e3aSAlexander Block if (ret) 43578ea05e3aSAlexander Block ret = -EFAULT; 43588ea05e3aSAlexander Block 43598ea05e3aSAlexander Block out: 43608ea05e3aSAlexander Block kfree(sa); 43618ea05e3aSAlexander Block up_write(&root->fs_info->subvol_sem); 43628ea05e3aSAlexander Block mnt_drop_write_file(file); 43638ea05e3aSAlexander Block return ret; 43648ea05e3aSAlexander Block } 43658ea05e3aSAlexander Block 4366867ab667Sjeff.liu static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg) 4367867ab667Sjeff.liu { 43686d0379ecSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 4369a1b83ac5SAnand Jain size_t len; 4370867ab667Sjeff.liu int ret; 4371a1b83ac5SAnand Jain char label[BTRFS_LABEL_SIZE]; 4372a1b83ac5SAnand Jain 4373a1b83ac5SAnand Jain spin_lock(&root->fs_info->super_lock); 4374a1b83ac5SAnand Jain memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE); 4375a1b83ac5SAnand Jain spin_unlock(&root->fs_info->super_lock); 4376a1b83ac5SAnand Jain 4377a1b83ac5SAnand Jain len = strnlen(label, BTRFS_LABEL_SIZE); 4378867ab667Sjeff.liu 4379867ab667Sjeff.liu if (len == BTRFS_LABEL_SIZE) { 4380867ab667Sjeff.liu pr_warn("btrfs: label is too long, return the first %zu bytes\n", 4381867ab667Sjeff.liu --len); 4382867ab667Sjeff.liu } 4383867ab667Sjeff.liu 4384867ab667Sjeff.liu ret = copy_to_user(arg, label, len); 4385867ab667Sjeff.liu 4386867ab667Sjeff.liu return ret ? -EFAULT : 0; 4387867ab667Sjeff.liu } 4388867ab667Sjeff.liu 4389a8bfd4abSjeff.liu static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg) 4390a8bfd4abSjeff.liu { 43916d0379ecSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 4392a8bfd4abSjeff.liu struct btrfs_super_block *super_block = root->fs_info->super_copy; 4393a8bfd4abSjeff.liu struct btrfs_trans_handle *trans; 4394a8bfd4abSjeff.liu char label[BTRFS_LABEL_SIZE]; 4395a8bfd4abSjeff.liu int ret; 4396a8bfd4abSjeff.liu 4397a8bfd4abSjeff.liu if (!capable(CAP_SYS_ADMIN)) 4398a8bfd4abSjeff.liu return -EPERM; 4399a8bfd4abSjeff.liu 4400a8bfd4abSjeff.liu if (copy_from_user(label, arg, sizeof(label))) 4401a8bfd4abSjeff.liu return -EFAULT; 4402a8bfd4abSjeff.liu 4403a8bfd4abSjeff.liu if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) { 4404a8bfd4abSjeff.liu pr_err("btrfs: unable to set label with more than %d bytes\n", 4405a8bfd4abSjeff.liu BTRFS_LABEL_SIZE - 1); 4406a8bfd4abSjeff.liu return -EINVAL; 4407a8bfd4abSjeff.liu } 4408a8bfd4abSjeff.liu 4409a8bfd4abSjeff.liu ret = mnt_want_write_file(file); 4410a8bfd4abSjeff.liu if (ret) 4411a8bfd4abSjeff.liu return ret; 4412a8bfd4abSjeff.liu 4413a8bfd4abSjeff.liu trans = btrfs_start_transaction(root, 0); 4414a8bfd4abSjeff.liu if (IS_ERR(trans)) { 4415a8bfd4abSjeff.liu ret = PTR_ERR(trans); 4416a8bfd4abSjeff.liu goto out_unlock; 4417a8bfd4abSjeff.liu } 4418a8bfd4abSjeff.liu 4419a1b83ac5SAnand Jain spin_lock(&root->fs_info->super_lock); 4420a8bfd4abSjeff.liu strcpy(super_block->label, label); 4421a1b83ac5SAnand Jain spin_unlock(&root->fs_info->super_lock); 4422a8bfd4abSjeff.liu ret = btrfs_end_transaction(trans, root); 4423a8bfd4abSjeff.liu 4424a8bfd4abSjeff.liu out_unlock: 4425a8bfd4abSjeff.liu mnt_drop_write_file(file); 4426a8bfd4abSjeff.liu return ret; 4427a8bfd4abSjeff.liu } 4428a8bfd4abSjeff.liu 4429f46b5a66SChristoph Hellwig long btrfs_ioctl(struct file *file, unsigned int 4430f46b5a66SChristoph Hellwig cmd, unsigned long arg) 4431f46b5a66SChristoph Hellwig { 4432496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 44334bcabaa3SChristoph Hellwig void __user *argp = (void __user *)arg; 4434f46b5a66SChristoph Hellwig 4435f46b5a66SChristoph Hellwig switch (cmd) { 44366cbff00fSChristoph Hellwig case FS_IOC_GETFLAGS: 44376cbff00fSChristoph Hellwig return btrfs_ioctl_getflags(file, argp); 44386cbff00fSChristoph Hellwig case FS_IOC_SETFLAGS: 44396cbff00fSChristoph Hellwig return btrfs_ioctl_setflags(file, argp); 44406cbff00fSChristoph Hellwig case FS_IOC_GETVERSION: 44416cbff00fSChristoph Hellwig return btrfs_ioctl_getversion(file, argp); 4442f7039b1dSLi Dongyang case FITRIM: 4443f7039b1dSLi Dongyang return btrfs_ioctl_fitrim(file, argp); 4444f46b5a66SChristoph Hellwig case BTRFS_IOC_SNAP_CREATE: 4445fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 0); 4446fdfb1e4fSLi Zefan case BTRFS_IOC_SNAP_CREATE_V2: 4447fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create_v2(file, argp, 0); 44483de4586cSChris Mason case BTRFS_IOC_SUBVOL_CREATE: 4449fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 1); 44506f72c7e2SArne Jansen case BTRFS_IOC_SUBVOL_CREATE_V2: 44516f72c7e2SArne Jansen return btrfs_ioctl_snap_create_v2(file, argp, 1); 445276dda93cSYan, Zheng case BTRFS_IOC_SNAP_DESTROY: 445376dda93cSYan, Zheng return btrfs_ioctl_snap_destroy(file, argp); 44540caa102dSLi Zefan case BTRFS_IOC_SUBVOL_GETFLAGS: 44550caa102dSLi Zefan return btrfs_ioctl_subvol_getflags(file, argp); 44560caa102dSLi Zefan case BTRFS_IOC_SUBVOL_SETFLAGS: 44570caa102dSLi Zefan return btrfs_ioctl_subvol_setflags(file, argp); 44586ef5ed0dSJosef Bacik case BTRFS_IOC_DEFAULT_SUBVOL: 44596ef5ed0dSJosef Bacik return btrfs_ioctl_default_subvol(file, argp); 4460f46b5a66SChristoph Hellwig case BTRFS_IOC_DEFRAG: 44611e701a32SChris Mason return btrfs_ioctl_defrag(file, NULL); 44621e701a32SChris Mason case BTRFS_IOC_DEFRAG_RANGE: 44631e701a32SChris Mason return btrfs_ioctl_defrag(file, argp); 4464f46b5a66SChristoph Hellwig case BTRFS_IOC_RESIZE: 4465198605a8SMiao Xie return btrfs_ioctl_resize(file, argp); 4466f46b5a66SChristoph Hellwig case BTRFS_IOC_ADD_DEV: 44674bcabaa3SChristoph Hellwig return btrfs_ioctl_add_dev(root, argp); 4468f46b5a66SChristoph Hellwig case BTRFS_IOC_RM_DEV: 4469da24927bSMiao Xie return btrfs_ioctl_rm_dev(file, argp); 4470475f6387SJan Schmidt case BTRFS_IOC_FS_INFO: 4471475f6387SJan Schmidt return btrfs_ioctl_fs_info(root, argp); 4472475f6387SJan Schmidt case BTRFS_IOC_DEV_INFO: 4473475f6387SJan Schmidt return btrfs_ioctl_dev_info(root, argp); 4474f46b5a66SChristoph Hellwig case BTRFS_IOC_BALANCE: 44759ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, NULL); 4476f46b5a66SChristoph Hellwig case BTRFS_IOC_CLONE: 4477c5c9cd4dSSage Weil return btrfs_ioctl_clone(file, arg, 0, 0, 0); 4478c5c9cd4dSSage Weil case BTRFS_IOC_CLONE_RANGE: 44797a865e8aSChristoph Hellwig return btrfs_ioctl_clone_range(file, argp); 4480f46b5a66SChristoph Hellwig case BTRFS_IOC_TRANS_START: 4481f46b5a66SChristoph Hellwig return btrfs_ioctl_trans_start(file); 4482f46b5a66SChristoph Hellwig case BTRFS_IOC_TRANS_END: 4483f46b5a66SChristoph Hellwig return btrfs_ioctl_trans_end(file); 4484ac8e9819SChris Mason case BTRFS_IOC_TREE_SEARCH: 4485ac8e9819SChris Mason return btrfs_ioctl_tree_search(file, argp); 4486ac8e9819SChris Mason case BTRFS_IOC_INO_LOOKUP: 4487ac8e9819SChris Mason return btrfs_ioctl_ino_lookup(file, argp); 4488d7728c96SJan Schmidt case BTRFS_IOC_INO_PATHS: 4489d7728c96SJan Schmidt return btrfs_ioctl_ino_to_path(root, argp); 4490d7728c96SJan Schmidt case BTRFS_IOC_LOGICAL_INO: 4491d7728c96SJan Schmidt return btrfs_ioctl_logical_to_ino(root, argp); 44921406e432SJosef Bacik case BTRFS_IOC_SPACE_INFO: 44931406e432SJosef Bacik return btrfs_ioctl_space_info(root, argp); 4494f46b5a66SChristoph Hellwig case BTRFS_IOC_SYNC: 4495f46b5a66SChristoph Hellwig btrfs_sync_fs(file->f_dentry->d_sb, 1); 4496f46b5a66SChristoph Hellwig return 0; 449746204592SSage Weil case BTRFS_IOC_START_SYNC: 44989a8c28beSMiao Xie return btrfs_ioctl_start_sync(root, argp); 449946204592SSage Weil case BTRFS_IOC_WAIT_SYNC: 45009a8c28beSMiao Xie return btrfs_ioctl_wait_sync(root, argp); 4501475f6387SJan Schmidt case BTRFS_IOC_SCRUB: 4502b8e95489SMiao Xie return btrfs_ioctl_scrub(file, argp); 4503475f6387SJan Schmidt case BTRFS_IOC_SCRUB_CANCEL: 4504475f6387SJan Schmidt return btrfs_ioctl_scrub_cancel(root, argp); 4505475f6387SJan Schmidt case BTRFS_IOC_SCRUB_PROGRESS: 4506475f6387SJan Schmidt return btrfs_ioctl_scrub_progress(root, argp); 4507c9e9f97bSIlya Dryomov case BTRFS_IOC_BALANCE_V2: 45089ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, argp); 4509837d5b6eSIlya Dryomov case BTRFS_IOC_BALANCE_CTL: 4510837d5b6eSIlya Dryomov return btrfs_ioctl_balance_ctl(root, arg); 451119a39dceSIlya Dryomov case BTRFS_IOC_BALANCE_PROGRESS: 451219a39dceSIlya Dryomov return btrfs_ioctl_balance_progress(root, argp); 45138ea05e3aSAlexander Block case BTRFS_IOC_SET_RECEIVED_SUBVOL: 45148ea05e3aSAlexander Block return btrfs_ioctl_set_received_subvol(file, argp); 451531db9f7cSAlexander Block case BTRFS_IOC_SEND: 451631db9f7cSAlexander Block return btrfs_ioctl_send(file, argp); 4517c11d2c23SStefan Behrens case BTRFS_IOC_GET_DEV_STATS: 4518b27f7c0cSDavid Sterba return btrfs_ioctl_get_dev_stats(root, argp); 45195d13a37bSArne Jansen case BTRFS_IOC_QUOTA_CTL: 4520905b0ddaSMiao Xie return btrfs_ioctl_quota_ctl(file, argp); 45215d13a37bSArne Jansen case BTRFS_IOC_QGROUP_ASSIGN: 4522905b0ddaSMiao Xie return btrfs_ioctl_qgroup_assign(file, argp); 45235d13a37bSArne Jansen case BTRFS_IOC_QGROUP_CREATE: 4524905b0ddaSMiao Xie return btrfs_ioctl_qgroup_create(file, argp); 45255d13a37bSArne Jansen case BTRFS_IOC_QGROUP_LIMIT: 4526905b0ddaSMiao Xie return btrfs_ioctl_qgroup_limit(file, argp); 45272f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN: 45282f232036SJan Schmidt return btrfs_ioctl_quota_rescan(file, argp); 45292f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_STATUS: 45302f232036SJan Schmidt return btrfs_ioctl_quota_rescan_status(file, argp); 453157254b6eSJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_WAIT: 453257254b6eSJan Schmidt return btrfs_ioctl_quota_rescan_wait(file, argp); 45333f6bcfbdSStefan Behrens case BTRFS_IOC_DEV_REPLACE: 45343f6bcfbdSStefan Behrens return btrfs_ioctl_dev_replace(root, argp); 4535867ab667Sjeff.liu case BTRFS_IOC_GET_FSLABEL: 4536867ab667Sjeff.liu return btrfs_ioctl_get_fslabel(file, argp); 4537a8bfd4abSjeff.liu case BTRFS_IOC_SET_FSLABEL: 4538a8bfd4abSjeff.liu return btrfs_ioctl_set_fslabel(file, argp); 4539416161dbSMark Fasheh case BTRFS_IOC_FILE_EXTENT_SAME: 4540416161dbSMark Fasheh return btrfs_ioctl_file_extent_same(file, argp); 4541f46b5a66SChristoph Hellwig } 4542f46b5a66SChristoph Hellwig 4543f46b5a66SChristoph Hellwig return -ENOTTY; 4544f46b5a66SChristoph Hellwig } 4545