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> 47f46b5a66SChristoph Hellwig #include "ctree.h" 48f46b5a66SChristoph Hellwig #include "disk-io.h" 49f46b5a66SChristoph Hellwig #include "transaction.h" 50f46b5a66SChristoph Hellwig #include "btrfs_inode.h" 51f46b5a66SChristoph Hellwig #include "print-tree.h" 52f46b5a66SChristoph Hellwig #include "volumes.h" 53925baeddSChris Mason #include "locking.h" 54581bb050SLi Zefan #include "inode-map.h" 55d7728c96SJan Schmidt #include "backref.h" 56606686eeSJosef Bacik #include "rcu-string.h" 5731db9f7cSAlexander Block #include "send.h" 583f6bcfbdSStefan Behrens #include "dev-replace.h" 5963541927SFilipe David Borba Manana #include "props.h" 603b02a68aSJeff Mahoney #include "sysfs.h" 61fcebe456SJosef Bacik #include "qgroup.h" 62f46b5a66SChristoph Hellwig 63abccd00fSHugo Mills #ifdef CONFIG_64BIT 64abccd00fSHugo Mills /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI 65abccd00fSHugo Mills * structures are incorrect, as the timespec structure from userspace 66abccd00fSHugo Mills * is 4 bytes too small. We define these alternatives here to teach 67abccd00fSHugo Mills * the kernel about the 32-bit struct packing. 68abccd00fSHugo Mills */ 69abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 { 70abccd00fSHugo Mills __u64 sec; 71abccd00fSHugo Mills __u32 nsec; 72abccd00fSHugo Mills } __attribute__ ((__packed__)); 73abccd00fSHugo Mills 74abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 { 75abccd00fSHugo Mills char uuid[BTRFS_UUID_SIZE]; /* in */ 76abccd00fSHugo Mills __u64 stransid; /* in */ 77abccd00fSHugo Mills __u64 rtransid; /* out */ 78abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 stime; /* in */ 79abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 rtime; /* out */ 80abccd00fSHugo Mills __u64 flags; /* in */ 81abccd00fSHugo Mills __u64 reserved[16]; /* in */ 82abccd00fSHugo Mills } __attribute__ ((__packed__)); 83abccd00fSHugo Mills 84abccd00fSHugo Mills #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \ 85abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32) 86abccd00fSHugo Mills #endif 87abccd00fSHugo Mills 88abccd00fSHugo Mills 89416161dbSMark Fasheh static int btrfs_clone(struct inode *src, struct inode *inode, 90416161dbSMark Fasheh u64 off, u64 olen, u64 olen_aligned, u64 destoff); 91416161dbSMark Fasheh 926cbff00fSChristoph Hellwig /* Mask out flags that are inappropriate for the given type of inode. */ 936cbff00fSChristoph Hellwig static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags) 946cbff00fSChristoph Hellwig { 956cbff00fSChristoph Hellwig if (S_ISDIR(mode)) 966cbff00fSChristoph Hellwig return flags; 976cbff00fSChristoph Hellwig else if (S_ISREG(mode)) 986cbff00fSChristoph Hellwig return flags & ~FS_DIRSYNC_FL; 996cbff00fSChristoph Hellwig else 1006cbff00fSChristoph Hellwig return flags & (FS_NODUMP_FL | FS_NOATIME_FL); 1016cbff00fSChristoph Hellwig } 102f46b5a66SChristoph Hellwig 1036cbff00fSChristoph Hellwig /* 1046cbff00fSChristoph Hellwig * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl. 1056cbff00fSChristoph Hellwig */ 1066cbff00fSChristoph Hellwig static unsigned int btrfs_flags_to_ioctl(unsigned int flags) 1076cbff00fSChristoph Hellwig { 1086cbff00fSChristoph Hellwig unsigned int iflags = 0; 1096cbff00fSChristoph Hellwig 1106cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_SYNC) 1116cbff00fSChristoph Hellwig iflags |= FS_SYNC_FL; 1126cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_IMMUTABLE) 1136cbff00fSChristoph Hellwig iflags |= FS_IMMUTABLE_FL; 1146cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_APPEND) 1156cbff00fSChristoph Hellwig iflags |= FS_APPEND_FL; 1166cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NODUMP) 1176cbff00fSChristoph Hellwig iflags |= FS_NODUMP_FL; 1186cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NOATIME) 1196cbff00fSChristoph Hellwig iflags |= FS_NOATIME_FL; 1206cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_DIRSYNC) 1216cbff00fSChristoph Hellwig iflags |= FS_DIRSYNC_FL; 122d0092bddSLi Zefan if (flags & BTRFS_INODE_NODATACOW) 123d0092bddSLi Zefan iflags |= FS_NOCOW_FL; 124d0092bddSLi Zefan 125d0092bddSLi Zefan if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS)) 126d0092bddSLi Zefan iflags |= FS_COMPR_FL; 127d0092bddSLi Zefan else if (flags & BTRFS_INODE_NOCOMPRESS) 128d0092bddSLi Zefan iflags |= FS_NOCOMP_FL; 1296cbff00fSChristoph Hellwig 1306cbff00fSChristoph Hellwig return iflags; 1316cbff00fSChristoph Hellwig } 1326cbff00fSChristoph Hellwig 1336cbff00fSChristoph Hellwig /* 1346cbff00fSChristoph Hellwig * Update inode->i_flags based on the btrfs internal flags. 1356cbff00fSChristoph Hellwig */ 1366cbff00fSChristoph Hellwig void btrfs_update_iflags(struct inode *inode) 1376cbff00fSChristoph Hellwig { 1386cbff00fSChristoph Hellwig struct btrfs_inode *ip = BTRFS_I(inode); 1393cc79392SFilipe Manana unsigned int new_fl = 0; 1406cbff00fSChristoph Hellwig 1416cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_SYNC) 1423cc79392SFilipe Manana new_fl |= S_SYNC; 1436cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_IMMUTABLE) 1443cc79392SFilipe Manana new_fl |= S_IMMUTABLE; 1456cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_APPEND) 1463cc79392SFilipe Manana new_fl |= S_APPEND; 1476cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_NOATIME) 1483cc79392SFilipe Manana new_fl |= S_NOATIME; 1496cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_DIRSYNC) 1503cc79392SFilipe Manana new_fl |= S_DIRSYNC; 1513cc79392SFilipe Manana 1523cc79392SFilipe Manana set_mask_bits(&inode->i_flags, 1533cc79392SFilipe Manana S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC, 1543cc79392SFilipe Manana new_fl); 1556cbff00fSChristoph Hellwig } 1566cbff00fSChristoph Hellwig 1576cbff00fSChristoph Hellwig /* 1586cbff00fSChristoph Hellwig * Inherit flags from the parent inode. 1596cbff00fSChristoph Hellwig * 160e27425d6SJosef Bacik * Currently only the compression flags and the cow flags are inherited. 1616cbff00fSChristoph Hellwig */ 1626cbff00fSChristoph Hellwig void btrfs_inherit_iflags(struct inode *inode, struct inode *dir) 1636cbff00fSChristoph Hellwig { 1640b4dcea5SChris Mason unsigned int flags; 1650b4dcea5SChris Mason 1660b4dcea5SChris Mason if (!dir) 1670b4dcea5SChris Mason return; 1680b4dcea5SChris Mason 1690b4dcea5SChris Mason flags = BTRFS_I(dir)->flags; 1706cbff00fSChristoph Hellwig 171e27425d6SJosef Bacik if (flags & BTRFS_INODE_NOCOMPRESS) { 172e27425d6SJosef Bacik BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS; 173e27425d6SJosef Bacik BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS; 174e27425d6SJosef Bacik } else if (flags & BTRFS_INODE_COMPRESS) { 175e27425d6SJosef Bacik BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS; 176e27425d6SJosef Bacik BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS; 177e27425d6SJosef Bacik } 1786cbff00fSChristoph Hellwig 179213490b3SLiu Bo if (flags & BTRFS_INODE_NODATACOW) { 180e27425d6SJosef Bacik BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW; 181213490b3SLiu Bo if (S_ISREG(inode->i_mode)) 182213490b3SLiu Bo BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM; 183213490b3SLiu Bo } 184e27425d6SJosef Bacik 1856cbff00fSChristoph Hellwig btrfs_update_iflags(inode); 1866cbff00fSChristoph Hellwig } 1876cbff00fSChristoph Hellwig 1886cbff00fSChristoph Hellwig static int btrfs_ioctl_getflags(struct file *file, void __user *arg) 1896cbff00fSChristoph Hellwig { 190496ad9aaSAl Viro struct btrfs_inode *ip = BTRFS_I(file_inode(file)); 1916cbff00fSChristoph Hellwig unsigned int flags = btrfs_flags_to_ioctl(ip->flags); 1926cbff00fSChristoph Hellwig 1936cbff00fSChristoph Hellwig if (copy_to_user(arg, &flags, sizeof(flags))) 1946cbff00fSChristoph Hellwig return -EFAULT; 1956cbff00fSChristoph Hellwig return 0; 1966cbff00fSChristoph Hellwig } 1976cbff00fSChristoph Hellwig 19875e7cb7fSLiu Bo static int check_flags(unsigned int flags) 19975e7cb7fSLiu Bo { 20075e7cb7fSLiu Bo if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \ 20175e7cb7fSLiu Bo FS_NOATIME_FL | FS_NODUMP_FL | \ 20275e7cb7fSLiu Bo FS_SYNC_FL | FS_DIRSYNC_FL | \ 203e1e8fb6aSLi Zefan FS_NOCOMP_FL | FS_COMPR_FL | 204e1e8fb6aSLi Zefan FS_NOCOW_FL)) 20575e7cb7fSLiu Bo return -EOPNOTSUPP; 20675e7cb7fSLiu Bo 20775e7cb7fSLiu Bo if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL)) 20875e7cb7fSLiu Bo return -EINVAL; 20975e7cb7fSLiu Bo 21075e7cb7fSLiu Bo return 0; 21175e7cb7fSLiu Bo } 21275e7cb7fSLiu Bo 2136cbff00fSChristoph Hellwig static int btrfs_ioctl_setflags(struct file *file, void __user *arg) 2146cbff00fSChristoph Hellwig { 215496ad9aaSAl Viro struct inode *inode = file_inode(file); 2166cbff00fSChristoph Hellwig struct btrfs_inode *ip = BTRFS_I(inode); 2176cbff00fSChristoph Hellwig struct btrfs_root *root = ip->root; 2186cbff00fSChristoph Hellwig struct btrfs_trans_handle *trans; 2196cbff00fSChristoph Hellwig unsigned int flags, oldflags; 2206cbff00fSChristoph Hellwig int ret; 221f062abf0SLi Zefan u64 ip_oldflags; 222f062abf0SLi Zefan unsigned int i_oldflags; 2237e97b8daSDavid Sterba umode_t mode; 2246cbff00fSChristoph Hellwig 225bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 226bd60ea0fSDavid Sterba return -EPERM; 227bd60ea0fSDavid Sterba 228b83cc969SLi Zefan if (btrfs_root_readonly(root)) 229b83cc969SLi Zefan return -EROFS; 230b83cc969SLi Zefan 2316cbff00fSChristoph Hellwig if (copy_from_user(&flags, arg, sizeof(flags))) 2326cbff00fSChristoph Hellwig return -EFAULT; 2336cbff00fSChristoph Hellwig 23475e7cb7fSLiu Bo ret = check_flags(flags); 23575e7cb7fSLiu Bo if (ret) 23675e7cb7fSLiu Bo return ret; 2376cbff00fSChristoph Hellwig 238e7848683SJan Kara ret = mnt_want_write_file(file); 239e7848683SJan Kara if (ret) 240e7848683SJan Kara return ret; 241e7848683SJan Kara 2426cbff00fSChristoph Hellwig mutex_lock(&inode->i_mutex); 2436cbff00fSChristoph Hellwig 244f062abf0SLi Zefan ip_oldflags = ip->flags; 245f062abf0SLi Zefan i_oldflags = inode->i_flags; 2467e97b8daSDavid Sterba mode = inode->i_mode; 247f062abf0SLi Zefan 2486cbff00fSChristoph Hellwig flags = btrfs_mask_flags(inode->i_mode, flags); 2496cbff00fSChristoph Hellwig oldflags = btrfs_flags_to_ioctl(ip->flags); 2506cbff00fSChristoph Hellwig if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) { 2516cbff00fSChristoph Hellwig if (!capable(CAP_LINUX_IMMUTABLE)) { 2526cbff00fSChristoph Hellwig ret = -EPERM; 2536cbff00fSChristoph Hellwig goto out_unlock; 2546cbff00fSChristoph Hellwig } 2556cbff00fSChristoph Hellwig } 2566cbff00fSChristoph Hellwig 2576cbff00fSChristoph Hellwig if (flags & FS_SYNC_FL) 2586cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_SYNC; 2596cbff00fSChristoph Hellwig else 2606cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_SYNC; 2616cbff00fSChristoph Hellwig if (flags & FS_IMMUTABLE_FL) 2626cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_IMMUTABLE; 2636cbff00fSChristoph Hellwig else 2646cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_IMMUTABLE; 2656cbff00fSChristoph Hellwig if (flags & FS_APPEND_FL) 2666cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_APPEND; 2676cbff00fSChristoph Hellwig else 2686cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_APPEND; 2696cbff00fSChristoph Hellwig if (flags & FS_NODUMP_FL) 2706cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_NODUMP; 2716cbff00fSChristoph Hellwig else 2726cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_NODUMP; 2736cbff00fSChristoph Hellwig if (flags & FS_NOATIME_FL) 2746cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_NOATIME; 2756cbff00fSChristoph Hellwig else 2766cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_NOATIME; 2776cbff00fSChristoph Hellwig if (flags & FS_DIRSYNC_FL) 2786cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_DIRSYNC; 2796cbff00fSChristoph Hellwig else 2806cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_DIRSYNC; 2817e97b8daSDavid Sterba if (flags & FS_NOCOW_FL) { 2827e97b8daSDavid Sterba if (S_ISREG(mode)) { 2837e97b8daSDavid Sterba /* 2847e97b8daSDavid Sterba * It's safe to turn csums off here, no extents exist. 2857e97b8daSDavid Sterba * Otherwise we want the flag to reflect the real COW 2867e97b8daSDavid Sterba * status of the file and will not set it. 2877e97b8daSDavid Sterba */ 2887e97b8daSDavid Sterba if (inode->i_size == 0) 2897e97b8daSDavid Sterba ip->flags |= BTRFS_INODE_NODATACOW 2907e97b8daSDavid Sterba | BTRFS_INODE_NODATASUM; 2917e97b8daSDavid Sterba } else { 292e1e8fb6aSLi Zefan ip->flags |= BTRFS_INODE_NODATACOW; 2937e97b8daSDavid Sterba } 2947e97b8daSDavid Sterba } else { 2957e97b8daSDavid Sterba /* 2967e97b8daSDavid Sterba * Revert back under same assuptions as above 2977e97b8daSDavid Sterba */ 2987e97b8daSDavid Sterba if (S_ISREG(mode)) { 2997e97b8daSDavid Sterba if (inode->i_size == 0) 3007e97b8daSDavid Sterba ip->flags &= ~(BTRFS_INODE_NODATACOW 3017e97b8daSDavid Sterba | BTRFS_INODE_NODATASUM); 3027e97b8daSDavid Sterba } else { 303e1e8fb6aSLi Zefan ip->flags &= ~BTRFS_INODE_NODATACOW; 3047e97b8daSDavid Sterba } 3057e97b8daSDavid Sterba } 3066cbff00fSChristoph Hellwig 30775e7cb7fSLiu Bo /* 30875e7cb7fSLiu Bo * The COMPRESS flag can only be changed by users, while the NOCOMPRESS 30975e7cb7fSLiu Bo * flag may be changed automatically if compression code won't make 31075e7cb7fSLiu Bo * things smaller. 31175e7cb7fSLiu Bo */ 31275e7cb7fSLiu Bo if (flags & FS_NOCOMP_FL) { 31375e7cb7fSLiu Bo ip->flags &= ~BTRFS_INODE_COMPRESS; 31475e7cb7fSLiu Bo ip->flags |= BTRFS_INODE_NOCOMPRESS; 31563541927SFilipe David Borba Manana 31663541927SFilipe David Borba Manana ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0); 31763541927SFilipe David Borba Manana if (ret && ret != -ENODATA) 31863541927SFilipe David Borba Manana goto out_drop; 31975e7cb7fSLiu Bo } else if (flags & FS_COMPR_FL) { 32063541927SFilipe David Borba Manana const char *comp; 32163541927SFilipe David Borba Manana 32275e7cb7fSLiu Bo ip->flags |= BTRFS_INODE_COMPRESS; 32375e7cb7fSLiu Bo ip->flags &= ~BTRFS_INODE_NOCOMPRESS; 32463541927SFilipe David Borba Manana 32563541927SFilipe David Borba Manana if (root->fs_info->compress_type == BTRFS_COMPRESS_LZO) 32663541927SFilipe David Borba Manana comp = "lzo"; 32763541927SFilipe David Borba Manana else 32863541927SFilipe David Borba Manana comp = "zlib"; 32963541927SFilipe David Borba Manana ret = btrfs_set_prop(inode, "btrfs.compression", 33063541927SFilipe David Borba Manana comp, strlen(comp), 0); 33163541927SFilipe David Borba Manana if (ret) 33263541927SFilipe David Borba Manana goto out_drop; 33363541927SFilipe David Borba Manana 334ebcb904dSLi Zefan } else { 335ebcb904dSLi Zefan ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS); 33675e7cb7fSLiu Bo } 3376cbff00fSChristoph Hellwig 3384da6f1a3SLi Zefan trans = btrfs_start_transaction(root, 1); 339f062abf0SLi Zefan if (IS_ERR(trans)) { 340f062abf0SLi Zefan ret = PTR_ERR(trans); 341f062abf0SLi Zefan goto out_drop; 342f062abf0SLi Zefan } 3436cbff00fSChristoph Hellwig 344306424ccSLi Zefan btrfs_update_iflags(inode); 3450c4d2d95SJosef Bacik inode_inc_iversion(inode); 346306424ccSLi Zefan inode->i_ctime = CURRENT_TIME; 3476cbff00fSChristoph Hellwig ret = btrfs_update_inode(trans, root, inode); 3486cbff00fSChristoph Hellwig 3496cbff00fSChristoph Hellwig btrfs_end_transaction(trans, root); 350f062abf0SLi Zefan out_drop: 351f062abf0SLi Zefan if (ret) { 352f062abf0SLi Zefan ip->flags = ip_oldflags; 353f062abf0SLi Zefan inode->i_flags = i_oldflags; 354f062abf0SLi Zefan } 3556cbff00fSChristoph Hellwig 3566cbff00fSChristoph Hellwig out_unlock: 3576cbff00fSChristoph Hellwig mutex_unlock(&inode->i_mutex); 358e7848683SJan Kara mnt_drop_write_file(file); 3592d4e6f6aSliubo return ret; 3606cbff00fSChristoph Hellwig } 3616cbff00fSChristoph Hellwig 3626cbff00fSChristoph Hellwig static int btrfs_ioctl_getversion(struct file *file, int __user *arg) 3636cbff00fSChristoph Hellwig { 364496ad9aaSAl Viro struct inode *inode = file_inode(file); 3656cbff00fSChristoph Hellwig 3666cbff00fSChristoph Hellwig return put_user(inode->i_generation, arg); 3676cbff00fSChristoph Hellwig } 368f46b5a66SChristoph Hellwig 369f7039b1dSLi Dongyang static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg) 370f7039b1dSLi Dongyang { 37154563d41SAl Viro struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb); 372f7039b1dSLi Dongyang struct btrfs_device *device; 373f7039b1dSLi Dongyang struct request_queue *q; 374f7039b1dSLi Dongyang struct fstrim_range range; 375f7039b1dSLi Dongyang u64 minlen = ULLONG_MAX; 376f7039b1dSLi Dongyang u64 num_devices = 0; 377815745cfSAl Viro u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy); 378f7039b1dSLi Dongyang int ret; 379f7039b1dSLi Dongyang 380f7039b1dSLi Dongyang if (!capable(CAP_SYS_ADMIN)) 381f7039b1dSLi Dongyang return -EPERM; 382f7039b1dSLi Dongyang 3831f78160cSXiao Guangrong rcu_read_lock(); 3841f78160cSXiao Guangrong list_for_each_entry_rcu(device, &fs_info->fs_devices->devices, 3851f78160cSXiao Guangrong dev_list) { 386f7039b1dSLi Dongyang if (!device->bdev) 387f7039b1dSLi Dongyang continue; 388f7039b1dSLi Dongyang q = bdev_get_queue(device->bdev); 389f7039b1dSLi Dongyang if (blk_queue_discard(q)) { 390f7039b1dSLi Dongyang num_devices++; 391f7039b1dSLi Dongyang minlen = min((u64)q->limits.discard_granularity, 392f7039b1dSLi Dongyang minlen); 393f7039b1dSLi Dongyang } 394f7039b1dSLi Dongyang } 3951f78160cSXiao Guangrong rcu_read_unlock(); 396f4c697e6SLukas Czerner 397f7039b1dSLi Dongyang if (!num_devices) 398f7039b1dSLi Dongyang return -EOPNOTSUPP; 399f7039b1dSLi Dongyang if (copy_from_user(&range, arg, sizeof(range))) 400f7039b1dSLi Dongyang return -EFAULT; 401e515c18bSLukas Czerner if (range.start > total_bytes || 402e515c18bSLukas Czerner range.len < fs_info->sb->s_blocksize) 403f4c697e6SLukas Czerner return -EINVAL; 404f7039b1dSLi Dongyang 405f4c697e6SLukas Czerner range.len = min(range.len, total_bytes - range.start); 406f7039b1dSLi Dongyang range.minlen = max(range.minlen, minlen); 407815745cfSAl Viro ret = btrfs_trim_fs(fs_info->tree_root, &range); 408f7039b1dSLi Dongyang if (ret < 0) 409f7039b1dSLi Dongyang return ret; 410f7039b1dSLi Dongyang 411f7039b1dSLi Dongyang if (copy_to_user(arg, &range, sizeof(range))) 412f7039b1dSLi Dongyang return -EFAULT; 413f7039b1dSLi Dongyang 414f7039b1dSLi Dongyang return 0; 415f7039b1dSLi Dongyang } 416f7039b1dSLi Dongyang 417dd5f9615SStefan Behrens int btrfs_is_empty_uuid(u8 *uuid) 418dd5f9615SStefan Behrens { 41946e0f66aSChris Mason int i; 42046e0f66aSChris Mason 42146e0f66aSChris Mason for (i = 0; i < BTRFS_UUID_SIZE; i++) { 42246e0f66aSChris Mason if (uuid[i]) 42346e0f66aSChris Mason return 0; 42446e0f66aSChris Mason } 42546e0f66aSChris Mason return 1; 426dd5f9615SStefan Behrens } 427dd5f9615SStefan Behrens 428d5c12070SMiao Xie static noinline int create_subvol(struct inode *dir, 429cb8e7090SChristoph Hellwig struct dentry *dentry, 43072fd032eSSage Weil char *name, int namelen, 4316f72c7e2SArne Jansen u64 *async_transid, 4328696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 433f46b5a66SChristoph Hellwig { 434f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 435f46b5a66SChristoph Hellwig struct btrfs_key key; 436f46b5a66SChristoph Hellwig struct btrfs_root_item root_item; 437f46b5a66SChristoph Hellwig struct btrfs_inode_item *inode_item; 438f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 439d5c12070SMiao Xie struct btrfs_root *root = BTRFS_I(dir)->root; 44076dda93cSYan, Zheng struct btrfs_root *new_root; 441d5c12070SMiao Xie struct btrfs_block_rsv block_rsv; 4428ea05e3aSAlexander Block struct timespec cur_time = CURRENT_TIME; 4435662344bSTsutomu Itoh struct inode *inode; 444f46b5a66SChristoph Hellwig int ret; 445f46b5a66SChristoph Hellwig int err; 446f46b5a66SChristoph Hellwig u64 objectid; 447f46b5a66SChristoph Hellwig u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; 4483de4586cSChris Mason u64 index = 0; 449d5c12070SMiao Xie u64 qgroup_reserved; 4508ea05e3aSAlexander Block uuid_le new_uuid; 451f46b5a66SChristoph Hellwig 452581bb050SLi Zefan ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid); 4532fbe8c8aSAl Viro if (ret) 454a22285a6SYan, Zheng return ret; 4556a912213SJosef Bacik 456d5c12070SMiao Xie btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); 4579ed74f2dSJosef Bacik /* 458d5c12070SMiao Xie * The same as the snapshot creation, please see the comment 459d5c12070SMiao Xie * of create_snapshot(). 4609ed74f2dSJosef Bacik */ 461d5c12070SMiao Xie ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 462dd5f9615SStefan Behrens 8, &qgroup_reserved, false); 463d5c12070SMiao Xie if (ret) 464d5c12070SMiao Xie return ret; 465f46b5a66SChristoph Hellwig 466d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 467d5c12070SMiao Xie if (IS_ERR(trans)) { 468d5c12070SMiao Xie ret = PTR_ERR(trans); 469de6e8200SLiu Bo btrfs_subvolume_release_metadata(root, &block_rsv, 470de6e8200SLiu Bo qgroup_reserved); 471de6e8200SLiu Bo return ret; 472d5c12070SMiao Xie } 473d5c12070SMiao Xie trans->block_rsv = &block_rsv; 474d5c12070SMiao Xie trans->bytes_reserved = block_rsv.size; 475f46b5a66SChristoph Hellwig 4768696c533SMiao Xie ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit); 4776f72c7e2SArne Jansen if (ret) 4786f72c7e2SArne Jansen goto fail; 4796f72c7e2SArne Jansen 4805d4f98a2SYan Zheng leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 4815581a51aSJan Schmidt 0, objectid, NULL, 0, 0, 0); 4828e8a1e31SJosef Bacik if (IS_ERR(leaf)) { 4838e8a1e31SJosef Bacik ret = PTR_ERR(leaf); 4848e8a1e31SJosef Bacik goto fail; 4858e8a1e31SJosef Bacik } 486f46b5a66SChristoph Hellwig 4875d4f98a2SYan Zheng memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); 488f46b5a66SChristoph Hellwig btrfs_set_header_bytenr(leaf, leaf->start); 489f46b5a66SChristoph Hellwig btrfs_set_header_generation(leaf, trans->transid); 4905d4f98a2SYan Zheng btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 491f46b5a66SChristoph Hellwig btrfs_set_header_owner(leaf, objectid); 492f46b5a66SChristoph Hellwig 4930a4e5586SRoss Kirk write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(), 494f46b5a66SChristoph Hellwig BTRFS_FSID_SIZE); 4955d4f98a2SYan Zheng write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid, 496b308bc2fSGeert Uytterhoeven btrfs_header_chunk_tree_uuid(leaf), 4975d4f98a2SYan Zheng BTRFS_UUID_SIZE); 498f46b5a66SChristoph Hellwig btrfs_mark_buffer_dirty(leaf); 499f46b5a66SChristoph Hellwig 5008ea05e3aSAlexander Block memset(&root_item, 0, sizeof(root_item)); 5018ea05e3aSAlexander Block 502f46b5a66SChristoph Hellwig inode_item = &root_item.inode; 5033cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 5043cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 5053cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 5063cae210fSQu Wenruo btrfs_set_stack_inode_nbytes(inode_item, root->leafsize); 5073cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 508f46b5a66SChristoph Hellwig 5093cae210fSQu Wenruo btrfs_set_root_flags(&root_item, 0); 5103cae210fSQu Wenruo btrfs_set_root_limit(&root_item, 0); 5113cae210fSQu Wenruo btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT); 51208fe4db1SLi Zefan 513f46b5a66SChristoph Hellwig btrfs_set_root_bytenr(&root_item, leaf->start); 51484234f3aSYan Zheng btrfs_set_root_generation(&root_item, trans->transid); 515f46b5a66SChristoph Hellwig btrfs_set_root_level(&root_item, 0); 516f46b5a66SChristoph Hellwig btrfs_set_root_refs(&root_item, 1); 51786b9f2ecSYan, Zheng btrfs_set_root_used(&root_item, leaf->len); 51880ff3856SYan Zheng btrfs_set_root_last_snapshot(&root_item, 0); 519f46b5a66SChristoph Hellwig 5208ea05e3aSAlexander Block btrfs_set_root_generation_v2(&root_item, 5218ea05e3aSAlexander Block btrfs_root_generation(&root_item)); 5228ea05e3aSAlexander Block uuid_le_gen(&new_uuid); 5238ea05e3aSAlexander Block memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE); 5243cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec); 5253cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec); 5268ea05e3aSAlexander Block root_item.ctime = root_item.otime; 5278ea05e3aSAlexander Block btrfs_set_root_ctransid(&root_item, trans->transid); 5288ea05e3aSAlexander Block btrfs_set_root_otransid(&root_item, trans->transid); 529f46b5a66SChristoph Hellwig 530925baeddSChris Mason btrfs_tree_unlock(leaf); 531f46b5a66SChristoph Hellwig free_extent_buffer(leaf); 532f46b5a66SChristoph Hellwig leaf = NULL; 533f46b5a66SChristoph Hellwig 534f46b5a66SChristoph Hellwig btrfs_set_root_dirid(&root_item, new_dirid); 535f46b5a66SChristoph Hellwig 536f46b5a66SChristoph Hellwig key.objectid = objectid; 5375d4f98a2SYan Zheng key.offset = 0; 538f46b5a66SChristoph Hellwig btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); 539f46b5a66SChristoph Hellwig ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, 540f46b5a66SChristoph Hellwig &root_item); 541f46b5a66SChristoph Hellwig if (ret) 542f46b5a66SChristoph Hellwig goto fail; 543f46b5a66SChristoph Hellwig 54476dda93cSYan, Zheng key.offset = (u64)-1; 54576dda93cSYan, Zheng new_root = btrfs_read_fs_root_no_name(root->fs_info, &key); 54679787eaaSJeff Mahoney if (IS_ERR(new_root)) { 54779787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, PTR_ERR(new_root)); 54879787eaaSJeff Mahoney ret = PTR_ERR(new_root); 54979787eaaSJeff Mahoney goto fail; 55079787eaaSJeff Mahoney } 55176dda93cSYan, Zheng 55276dda93cSYan, Zheng btrfs_record_root_in_trans(trans, new_root); 55376dda93cSYan, Zheng 55463541927SFilipe David Borba Manana ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid); 555ce598979SMark Fasheh if (ret) { 556ce598979SMark Fasheh /* We potentially lose an unused inode item here */ 55779787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, ret); 558ce598979SMark Fasheh goto fail; 559ce598979SMark Fasheh } 560ce598979SMark Fasheh 561f46b5a66SChristoph Hellwig /* 562f46b5a66SChristoph Hellwig * insert the directory item 563f46b5a66SChristoph Hellwig */ 5643de4586cSChris Mason ret = btrfs_set_inode_index(dir, &index); 56579787eaaSJeff Mahoney if (ret) { 56679787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, ret); 56779787eaaSJeff Mahoney goto fail; 56879787eaaSJeff Mahoney } 5693de4586cSChris Mason 5703de4586cSChris Mason ret = btrfs_insert_dir_item(trans, root, 57116cdcec7SMiao Xie name, namelen, dir, &key, 5723de4586cSChris Mason BTRFS_FT_DIR, index); 57379787eaaSJeff Mahoney if (ret) { 57479787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, ret); 575f46b5a66SChristoph Hellwig goto fail; 57679787eaaSJeff Mahoney } 5770660b5afSChris Mason 57852c26179SYan Zheng btrfs_i_size_write(dir, dir->i_size + namelen * 2); 57952c26179SYan Zheng ret = btrfs_update_inode(trans, root, dir); 58052c26179SYan Zheng BUG_ON(ret); 58152c26179SYan Zheng 5820660b5afSChris Mason ret = btrfs_add_root_ref(trans, root->fs_info->tree_root, 5834df27c4dSYan, Zheng objectid, root->root_key.objectid, 58433345d01SLi Zefan btrfs_ino(dir), index, name, namelen); 5850660b5afSChris Mason BUG_ON(ret); 5860660b5afSChris Mason 587dd5f9615SStefan Behrens ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root, 588dd5f9615SStefan Behrens root_item.uuid, BTRFS_UUID_KEY_SUBVOL, 589dd5f9615SStefan Behrens objectid); 590dd5f9615SStefan Behrens if (ret) 591dd5f9615SStefan Behrens btrfs_abort_transaction(trans, root, ret); 592dd5f9615SStefan Behrens 593f46b5a66SChristoph Hellwig fail: 594d5c12070SMiao Xie trans->block_rsv = NULL; 595d5c12070SMiao Xie trans->bytes_reserved = 0; 596de6e8200SLiu Bo btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved); 597de6e8200SLiu Bo 59872fd032eSSage Weil if (async_transid) { 59972fd032eSSage Weil *async_transid = trans->transid; 60072fd032eSSage Weil err = btrfs_commit_transaction_async(trans, root, 1); 60100d71c9cSMiao Xie if (err) 60200d71c9cSMiao Xie err = btrfs_commit_transaction(trans, root); 60372fd032eSSage Weil } else { 60476dda93cSYan, Zheng err = btrfs_commit_transaction(trans, root); 60572fd032eSSage Weil } 606f46b5a66SChristoph Hellwig if (err && !ret) 607f46b5a66SChristoph Hellwig ret = err; 6081a65e24bSChris Mason 6095662344bSTsutomu Itoh if (!ret) { 6105662344bSTsutomu Itoh inode = btrfs_lookup_dentry(dir, dentry); 611de6e8200SLiu Bo if (IS_ERR(inode)) 612de6e8200SLiu Bo return PTR_ERR(inode); 6135662344bSTsutomu Itoh d_instantiate(dentry, inode); 6145662344bSTsutomu Itoh } 615f46b5a66SChristoph Hellwig return ret; 616f46b5a66SChristoph Hellwig } 617f46b5a66SChristoph Hellwig 6188257b2dcSMiao Xie static void btrfs_wait_nocow_write(struct btrfs_root *root) 6198257b2dcSMiao Xie { 6208257b2dcSMiao Xie s64 writers; 6218257b2dcSMiao Xie DEFINE_WAIT(wait); 6228257b2dcSMiao Xie 6238257b2dcSMiao Xie do { 6248257b2dcSMiao Xie prepare_to_wait(&root->subv_writers->wait, &wait, 6258257b2dcSMiao Xie TASK_UNINTERRUPTIBLE); 6268257b2dcSMiao Xie 6278257b2dcSMiao Xie writers = percpu_counter_sum(&root->subv_writers->counter); 6288257b2dcSMiao Xie if (writers) 6298257b2dcSMiao Xie schedule(); 6308257b2dcSMiao Xie 6318257b2dcSMiao Xie finish_wait(&root->subv_writers->wait, &wait); 6328257b2dcSMiao Xie } while (writers); 6338257b2dcSMiao Xie } 6348257b2dcSMiao Xie 635e9662f70SMiao Xie static int create_snapshot(struct btrfs_root *root, struct inode *dir, 636e9662f70SMiao Xie struct dentry *dentry, char *name, int namelen, 637e9662f70SMiao Xie u64 *async_transid, bool readonly, 638e9662f70SMiao Xie struct btrfs_qgroup_inherit *inherit) 639f46b5a66SChristoph Hellwig { 6402e4bfab9SYan, Zheng struct inode *inode; 641f46b5a66SChristoph Hellwig struct btrfs_pending_snapshot *pending_snapshot; 642f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 6432e4bfab9SYan, Zheng int ret; 644f46b5a66SChristoph Hellwig 64527cdeb70SMiao Xie if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state)) 646f46b5a66SChristoph Hellwig return -EINVAL; 647f46b5a66SChristoph Hellwig 6488257b2dcSMiao Xie atomic_inc(&root->will_be_snapshoted); 6498257b2dcSMiao Xie smp_mb__after_atomic_inc(); 6508257b2dcSMiao Xie btrfs_wait_nocow_write(root); 6518257b2dcSMiao Xie 6526a03843dSMiao Xie ret = btrfs_start_delalloc_inodes(root, 0); 6536a03843dSMiao Xie if (ret) 6548257b2dcSMiao Xie goto out; 6556a03843dSMiao Xie 656b0244199SMiao Xie btrfs_wait_ordered_extents(root, -1); 6576a03843dSMiao Xie 658a22285a6SYan, Zheng pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS); 6598257b2dcSMiao Xie if (!pending_snapshot) { 6608257b2dcSMiao Xie ret = -ENOMEM; 6618257b2dcSMiao Xie goto out; 6628257b2dcSMiao Xie } 663a22285a6SYan, Zheng 66466d8f3ddSMiao Xie btrfs_init_block_rsv(&pending_snapshot->block_rsv, 66566d8f3ddSMiao Xie BTRFS_BLOCK_RSV_TEMP); 666d5c12070SMiao Xie /* 667d5c12070SMiao Xie * 1 - parent dir inode 668d5c12070SMiao Xie * 2 - dir entries 669d5c12070SMiao Xie * 1 - root item 670d5c12070SMiao Xie * 2 - root ref/backref 671d5c12070SMiao Xie * 1 - root of snapshot 672dd5f9615SStefan Behrens * 1 - UUID item 673d5c12070SMiao Xie */ 674d5c12070SMiao Xie ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, 675dd5f9615SStefan Behrens &pending_snapshot->block_rsv, 8, 676ee3441b4SJeff Mahoney &pending_snapshot->qgroup_reserved, 677ee3441b4SJeff Mahoney false); 678d5c12070SMiao Xie if (ret) 6798257b2dcSMiao Xie goto free; 680d5c12070SMiao Xie 681a22285a6SYan, Zheng pending_snapshot->dentry = dentry; 682a22285a6SYan, Zheng pending_snapshot->root = root; 683b83cc969SLi Zefan pending_snapshot->readonly = readonly; 684e9662f70SMiao Xie pending_snapshot->dir = dir; 6858696c533SMiao Xie pending_snapshot->inherit = inherit; 686a22285a6SYan, Zheng 687d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 688a22285a6SYan, Zheng if (IS_ERR(trans)) { 689a22285a6SYan, Zheng ret = PTR_ERR(trans); 690a22285a6SYan, Zheng goto fail; 691a22285a6SYan, Zheng } 692a22285a6SYan, Zheng 6938351583eSJosef Bacik spin_lock(&root->fs_info->trans_lock); 694a22285a6SYan, Zheng list_add(&pending_snapshot->list, 695a22285a6SYan, Zheng &trans->transaction->pending_snapshots); 6968351583eSJosef Bacik spin_unlock(&root->fs_info->trans_lock); 69772fd032eSSage Weil if (async_transid) { 69872fd032eSSage Weil *async_transid = trans->transid; 69972fd032eSSage Weil ret = btrfs_commit_transaction_async(trans, 70072fd032eSSage Weil root->fs_info->extent_root, 1); 70100d71c9cSMiao Xie if (ret) 70200d71c9cSMiao Xie ret = btrfs_commit_transaction(trans, root); 70372fd032eSSage Weil } else { 70472fd032eSSage Weil ret = btrfs_commit_transaction(trans, 70572fd032eSSage Weil root->fs_info->extent_root); 70672fd032eSSage Weil } 707aec8030aSMiao Xie if (ret) 708c37b2b62SJosef Bacik goto fail; 709a22285a6SYan, Zheng 710a22285a6SYan, Zheng ret = pending_snapshot->error; 711f46b5a66SChristoph Hellwig if (ret) 7122e4bfab9SYan, Zheng goto fail; 713f46b5a66SChristoph Hellwig 71466b4ffd1SJosef Bacik ret = btrfs_orphan_cleanup(pending_snapshot->snap); 71566b4ffd1SJosef Bacik if (ret) 71666b4ffd1SJosef Bacik goto fail; 717f46b5a66SChristoph Hellwig 7183821f348SFilipe Manana /* 7193821f348SFilipe Manana * If orphan cleanup did remove any orphans, it means the tree was 7203821f348SFilipe Manana * modified and therefore the commit root is not the same as the 7213821f348SFilipe Manana * current root anymore. This is a problem, because send uses the 7223821f348SFilipe Manana * commit root and therefore can see inode items that don't exist 7233821f348SFilipe Manana * in the current root anymore, and for example make calls to 7243821f348SFilipe Manana * btrfs_iget, which will do tree lookups based on the current root 7253821f348SFilipe Manana * and not on the commit root. Those lookups will fail, returning a 7263821f348SFilipe Manana * -ESTALE error, and making send fail with that error. So make sure 7273821f348SFilipe Manana * a send does not see any orphans we have just removed, and that it 7283821f348SFilipe Manana * will see the same inodes regardless of whether a transaction 7293821f348SFilipe Manana * commit happened before it started (meaning that the commit root 7303821f348SFilipe Manana * will be the same as the current root) or not. 7313821f348SFilipe Manana */ 7323821f348SFilipe Manana if (readonly && pending_snapshot->snap->node != 7333821f348SFilipe Manana pending_snapshot->snap->commit_root) { 7343821f348SFilipe Manana trans = btrfs_join_transaction(pending_snapshot->snap); 7353821f348SFilipe Manana if (IS_ERR(trans) && PTR_ERR(trans) != -ENOENT) { 7363821f348SFilipe Manana ret = PTR_ERR(trans); 7373821f348SFilipe Manana goto fail; 7383821f348SFilipe Manana } 7393821f348SFilipe Manana if (!IS_ERR(trans)) { 7403821f348SFilipe Manana ret = btrfs_commit_transaction(trans, 7413821f348SFilipe Manana pending_snapshot->snap); 7423821f348SFilipe Manana if (ret) 7433821f348SFilipe Manana goto fail; 7443821f348SFilipe Manana } 7453821f348SFilipe Manana } 7463821f348SFilipe Manana 7472fbe8c8aSAl Viro inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry); 7482e4bfab9SYan, Zheng if (IS_ERR(inode)) { 7492e4bfab9SYan, Zheng ret = PTR_ERR(inode); 7502e4bfab9SYan, Zheng goto fail; 7512e4bfab9SYan, Zheng } 7525662344bSTsutomu Itoh 7532e4bfab9SYan, Zheng d_instantiate(dentry, inode); 7542e4bfab9SYan, Zheng ret = 0; 7552e4bfab9SYan, Zheng fail: 756d5c12070SMiao Xie btrfs_subvolume_release_metadata(BTRFS_I(dir)->root, 757d5c12070SMiao Xie &pending_snapshot->block_rsv, 758d5c12070SMiao Xie pending_snapshot->qgroup_reserved); 7598257b2dcSMiao Xie free: 760a22285a6SYan, Zheng kfree(pending_snapshot); 7618257b2dcSMiao Xie out: 7628257b2dcSMiao Xie atomic_dec(&root->will_be_snapshoted); 763f46b5a66SChristoph Hellwig return ret; 764f46b5a66SChristoph Hellwig } 765f46b5a66SChristoph Hellwig 7664260f7c7SSage Weil /* copy of check_sticky in fs/namei.c() 7674260f7c7SSage Weil * It's inline, so penalty for filesystems that don't use sticky bit is 7684260f7c7SSage Weil * minimal. 7694260f7c7SSage Weil */ 7704260f7c7SSage Weil static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode) 7714260f7c7SSage Weil { 7722f2f43d3SEric W. Biederman kuid_t fsuid = current_fsuid(); 7734260f7c7SSage Weil 7744260f7c7SSage Weil if (!(dir->i_mode & S_ISVTX)) 7754260f7c7SSage Weil return 0; 7762f2f43d3SEric W. Biederman if (uid_eq(inode->i_uid, fsuid)) 7774260f7c7SSage Weil return 0; 7782f2f43d3SEric W. Biederman if (uid_eq(dir->i_uid, fsuid)) 7794260f7c7SSage Weil return 0; 7804260f7c7SSage Weil return !capable(CAP_FOWNER); 7814260f7c7SSage Weil } 7824260f7c7SSage Weil 7834260f7c7SSage Weil /* copy of may_delete in fs/namei.c() 7844260f7c7SSage Weil * Check whether we can remove a link victim from directory dir, check 7854260f7c7SSage Weil * whether the type of victim is right. 7864260f7c7SSage Weil * 1. We can't do it if dir is read-only (done in permission()) 7874260f7c7SSage Weil * 2. We should have write and exec permissions on dir 7884260f7c7SSage Weil * 3. We can't remove anything from append-only dir 7894260f7c7SSage Weil * 4. We can't do anything with immutable dir (done in permission()) 7904260f7c7SSage Weil * 5. If the sticky bit on dir is set we should either 7914260f7c7SSage Weil * a. be owner of dir, or 7924260f7c7SSage Weil * b. be owner of victim, or 7934260f7c7SSage Weil * c. have CAP_FOWNER capability 7944260f7c7SSage Weil * 6. If the victim is append-only or immutable we can't do antyhing with 7954260f7c7SSage Weil * links pointing to it. 7964260f7c7SSage Weil * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. 7974260f7c7SSage Weil * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. 7984260f7c7SSage Weil * 9. We can't remove a root or mountpoint. 7994260f7c7SSage Weil * 10. We don't allow removal of NFS sillyrenamed files; it's handled by 8004260f7c7SSage Weil * nfs_async_unlink(). 8014260f7c7SSage Weil */ 8024260f7c7SSage Weil 8034260f7c7SSage Weil static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir) 8044260f7c7SSage Weil { 8054260f7c7SSage Weil int error; 8064260f7c7SSage Weil 8074260f7c7SSage Weil if (!victim->d_inode) 8084260f7c7SSage Weil return -ENOENT; 8094260f7c7SSage Weil 8104260f7c7SSage Weil BUG_ON(victim->d_parent->d_inode != dir); 8114fa6b5ecSJeff Layton audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); 8124260f7c7SSage Weil 8134260f7c7SSage Weil error = inode_permission(dir, MAY_WRITE | MAY_EXEC); 8144260f7c7SSage Weil if (error) 8154260f7c7SSage Weil return error; 8164260f7c7SSage Weil if (IS_APPEND(dir)) 8174260f7c7SSage Weil return -EPERM; 8184260f7c7SSage Weil if (btrfs_check_sticky(dir, victim->d_inode)|| 8194260f7c7SSage Weil IS_APPEND(victim->d_inode)|| 8204260f7c7SSage Weil IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode)) 8214260f7c7SSage Weil return -EPERM; 8224260f7c7SSage Weil if (isdir) { 8234260f7c7SSage Weil if (!S_ISDIR(victim->d_inode->i_mode)) 8244260f7c7SSage Weil return -ENOTDIR; 8254260f7c7SSage Weil if (IS_ROOT(victim)) 8264260f7c7SSage Weil return -EBUSY; 8274260f7c7SSage Weil } else if (S_ISDIR(victim->d_inode->i_mode)) 8284260f7c7SSage Weil return -EISDIR; 8294260f7c7SSage Weil if (IS_DEADDIR(dir)) 8304260f7c7SSage Weil return -ENOENT; 8314260f7c7SSage Weil if (victim->d_flags & DCACHE_NFSFS_RENAMED) 8324260f7c7SSage Weil return -EBUSY; 8334260f7c7SSage Weil return 0; 8344260f7c7SSage Weil } 8354260f7c7SSage Weil 836cb8e7090SChristoph Hellwig /* copy of may_create in fs/namei.c() */ 837cb8e7090SChristoph Hellwig static inline int btrfs_may_create(struct inode *dir, struct dentry *child) 838cb8e7090SChristoph Hellwig { 839cb8e7090SChristoph Hellwig if (child->d_inode) 840cb8e7090SChristoph Hellwig return -EEXIST; 841cb8e7090SChristoph Hellwig if (IS_DEADDIR(dir)) 842cb8e7090SChristoph Hellwig return -ENOENT; 843cb8e7090SChristoph Hellwig return inode_permission(dir, MAY_WRITE | MAY_EXEC); 844cb8e7090SChristoph Hellwig } 845cb8e7090SChristoph Hellwig 846cb8e7090SChristoph Hellwig /* 847cb8e7090SChristoph Hellwig * Create a new subvolume below @parent. This is largely modeled after 848cb8e7090SChristoph Hellwig * sys_mkdirat and vfs_mkdir, but we only do a single component lookup 849cb8e7090SChristoph Hellwig * inside this filesystem so it's quite a bit simpler. 850cb8e7090SChristoph Hellwig */ 85176dda93cSYan, Zheng static noinline int btrfs_mksubvol(struct path *parent, 85276dda93cSYan, Zheng char *name, int namelen, 85372fd032eSSage Weil struct btrfs_root *snap_src, 8546f72c7e2SArne Jansen u64 *async_transid, bool readonly, 8558696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 856cb8e7090SChristoph Hellwig { 85776dda93cSYan, Zheng struct inode *dir = parent->dentry->d_inode; 858cb8e7090SChristoph Hellwig struct dentry *dentry; 859cb8e7090SChristoph Hellwig int error; 860cb8e7090SChristoph Hellwig 8615c50c9b8SDavid Sterba error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT); 8625c50c9b8SDavid Sterba if (error == -EINTR) 8635c50c9b8SDavid Sterba return error; 864cb8e7090SChristoph Hellwig 865cb8e7090SChristoph Hellwig dentry = lookup_one_len(name, parent->dentry, namelen); 866cb8e7090SChristoph Hellwig error = PTR_ERR(dentry); 867cb8e7090SChristoph Hellwig if (IS_ERR(dentry)) 868cb8e7090SChristoph Hellwig goto out_unlock; 869cb8e7090SChristoph Hellwig 870cb8e7090SChristoph Hellwig error = -EEXIST; 871cb8e7090SChristoph Hellwig if (dentry->d_inode) 872cb8e7090SChristoph Hellwig goto out_dput; 873cb8e7090SChristoph Hellwig 87476dda93cSYan, Zheng error = btrfs_may_create(dir, dentry); 875cb8e7090SChristoph Hellwig if (error) 876a874a63eSLiu Bo goto out_dput; 877cb8e7090SChristoph Hellwig 8789c52057cSChris Mason /* 8799c52057cSChris Mason * even if this name doesn't exist, we may get hash collisions. 8809c52057cSChris Mason * check for them now when we can safely fail 8819c52057cSChris Mason */ 8829c52057cSChris Mason error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, 8839c52057cSChris Mason dir->i_ino, name, 8849c52057cSChris Mason namelen); 8859c52057cSChris Mason if (error) 8869c52057cSChris Mason goto out_dput; 8879c52057cSChris Mason 88876dda93cSYan, Zheng down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem); 88976dda93cSYan, Zheng 89076dda93cSYan, Zheng if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) 89176dda93cSYan, Zheng goto out_up_read; 89276dda93cSYan, Zheng 8933de4586cSChris Mason if (snap_src) { 894e9662f70SMiao Xie error = create_snapshot(snap_src, dir, dentry, name, namelen, 8956f72c7e2SArne Jansen async_transid, readonly, inherit); 8963de4586cSChris Mason } else { 897d5c12070SMiao Xie error = create_subvol(dir, dentry, name, namelen, 898d5c12070SMiao Xie async_transid, inherit); 8993de4586cSChris Mason } 90076dda93cSYan, Zheng if (!error) 90176dda93cSYan, Zheng fsnotify_mkdir(dir, dentry); 90276dda93cSYan, Zheng out_up_read: 90376dda93cSYan, Zheng up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem); 904cb8e7090SChristoph Hellwig out_dput: 905cb8e7090SChristoph Hellwig dput(dentry); 906cb8e7090SChristoph Hellwig out_unlock: 90776dda93cSYan, Zheng mutex_unlock(&dir->i_mutex); 908cb8e7090SChristoph Hellwig return error; 909cb8e7090SChristoph Hellwig } 910cb8e7090SChristoph Hellwig 9114cb5300bSChris Mason /* 9124cb5300bSChris Mason * When we're defragging a range, we don't want to kick it off again 9134cb5300bSChris Mason * if it is really just waiting for delalloc to send it down. 9144cb5300bSChris Mason * If we find a nice big extent or delalloc range for the bytes in the 9154cb5300bSChris Mason * file you want to defrag, we return 0 to let you know to skip this 9164cb5300bSChris Mason * part of the file 9174cb5300bSChris Mason */ 9184cb5300bSChris Mason static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh) 9194cb5300bSChris Mason { 9204cb5300bSChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 9214cb5300bSChris Mason struct extent_map *em = NULL; 9224cb5300bSChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 9234cb5300bSChris Mason u64 end; 9244cb5300bSChris Mason 9254cb5300bSChris Mason read_lock(&em_tree->lock); 9264cb5300bSChris Mason em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE); 9274cb5300bSChris Mason read_unlock(&em_tree->lock); 9284cb5300bSChris Mason 9294cb5300bSChris Mason if (em) { 9304cb5300bSChris Mason end = extent_map_end(em); 9314cb5300bSChris Mason free_extent_map(em); 9324cb5300bSChris Mason if (end - offset > thresh) 9334cb5300bSChris Mason return 0; 9344cb5300bSChris Mason } 9354cb5300bSChris Mason /* if we already have a nice delalloc here, just stop */ 9364cb5300bSChris Mason thresh /= 2; 9374cb5300bSChris Mason end = count_range_bits(io_tree, &offset, offset + thresh, 9384cb5300bSChris Mason thresh, EXTENT_DELALLOC, 1); 9394cb5300bSChris Mason if (end >= thresh) 9404cb5300bSChris Mason return 0; 9414cb5300bSChris Mason return 1; 9424cb5300bSChris Mason } 9434cb5300bSChris Mason 9444cb5300bSChris Mason /* 9454cb5300bSChris Mason * helper function to walk through a file and find extents 9464cb5300bSChris Mason * newer than a specific transid, and smaller than thresh. 9474cb5300bSChris Mason * 9484cb5300bSChris Mason * This is used by the defragging code to find new and small 9494cb5300bSChris Mason * extents 9504cb5300bSChris Mason */ 9514cb5300bSChris Mason static int find_new_extents(struct btrfs_root *root, 9524cb5300bSChris Mason struct inode *inode, u64 newer_than, 9534cb5300bSChris Mason u64 *off, int thresh) 9544cb5300bSChris Mason { 9554cb5300bSChris Mason struct btrfs_path *path; 9564cb5300bSChris Mason struct btrfs_key min_key; 9574cb5300bSChris Mason struct extent_buffer *leaf; 9584cb5300bSChris Mason struct btrfs_file_extent_item *extent; 9594cb5300bSChris Mason int type; 9604cb5300bSChris Mason int ret; 961a4689d2bSDavid Sterba u64 ino = btrfs_ino(inode); 9624cb5300bSChris Mason 9634cb5300bSChris Mason path = btrfs_alloc_path(); 9644cb5300bSChris Mason if (!path) 9654cb5300bSChris Mason return -ENOMEM; 9664cb5300bSChris Mason 967a4689d2bSDavid Sterba min_key.objectid = ino; 9684cb5300bSChris Mason min_key.type = BTRFS_EXTENT_DATA_KEY; 9694cb5300bSChris Mason min_key.offset = *off; 9704cb5300bSChris Mason 9714cb5300bSChris Mason while (1) { 972f094c9bdSFilipe Manana path->keep_locks = 1; 9736174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &min_key, path, newer_than); 9744cb5300bSChris Mason if (ret != 0) 9754cb5300bSChris Mason goto none; 976f094c9bdSFilipe Manana path->keep_locks = 0; 977f094c9bdSFilipe Manana btrfs_unlock_up_safe(path, 1); 978f094c9bdSFilipe Manana process_slot: 979a4689d2bSDavid Sterba if (min_key.objectid != ino) 9804cb5300bSChris Mason goto none; 9814cb5300bSChris Mason if (min_key.type != BTRFS_EXTENT_DATA_KEY) 9824cb5300bSChris Mason goto none; 9834cb5300bSChris Mason 9844cb5300bSChris Mason leaf = path->nodes[0]; 9854cb5300bSChris Mason extent = btrfs_item_ptr(leaf, path->slots[0], 9864cb5300bSChris Mason struct btrfs_file_extent_item); 9874cb5300bSChris Mason 9884cb5300bSChris Mason type = btrfs_file_extent_type(leaf, extent); 9894cb5300bSChris Mason if (type == BTRFS_FILE_EXTENT_REG && 9904cb5300bSChris Mason btrfs_file_extent_num_bytes(leaf, extent) < thresh && 9914cb5300bSChris Mason check_defrag_in_cache(inode, min_key.offset, thresh)) { 9924cb5300bSChris Mason *off = min_key.offset; 9934cb5300bSChris Mason btrfs_free_path(path); 9944cb5300bSChris Mason return 0; 9954cb5300bSChris Mason } 9964cb5300bSChris Mason 997f094c9bdSFilipe Manana path->slots[0]++; 998f094c9bdSFilipe Manana if (path->slots[0] < btrfs_header_nritems(leaf)) { 999f094c9bdSFilipe Manana btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]); 1000f094c9bdSFilipe Manana goto process_slot; 1001f094c9bdSFilipe Manana } 1002f094c9bdSFilipe Manana 10034cb5300bSChris Mason if (min_key.offset == (u64)-1) 10044cb5300bSChris Mason goto none; 10054cb5300bSChris Mason 10064cb5300bSChris Mason min_key.offset++; 10074cb5300bSChris Mason btrfs_release_path(path); 10084cb5300bSChris Mason } 10094cb5300bSChris Mason none: 10104cb5300bSChris Mason btrfs_free_path(path); 10114cb5300bSChris Mason return -ENOENT; 10124cb5300bSChris Mason } 10134cb5300bSChris Mason 10146c282eb4SLi Zefan static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start) 101517ce6ef8SLiu Bo { 101617ce6ef8SLiu Bo struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 1017940100a4SChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 10186c282eb4SLi Zefan struct extent_map *em; 10196c282eb4SLi Zefan u64 len = PAGE_CACHE_SIZE; 1020940100a4SChris Mason 1021940100a4SChris Mason /* 1022940100a4SChris Mason * hopefully we have this extent in the tree already, try without 1023940100a4SChris Mason * the full extent lock 1024940100a4SChris Mason */ 1025940100a4SChris Mason read_lock(&em_tree->lock); 1026940100a4SChris Mason em = lookup_extent_mapping(em_tree, start, len); 1027940100a4SChris Mason read_unlock(&em_tree->lock); 1028940100a4SChris Mason 1029940100a4SChris Mason if (!em) { 1030308d9800SFilipe Manana struct extent_state *cached = NULL; 1031308d9800SFilipe Manana u64 end = start + len - 1; 1032308d9800SFilipe Manana 1033940100a4SChris Mason /* get the big lock and read metadata off disk */ 1034308d9800SFilipe Manana lock_extent_bits(io_tree, start, end, 0, &cached); 1035940100a4SChris Mason em = btrfs_get_extent(inode, NULL, 0, start, len, 0); 1036308d9800SFilipe Manana unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS); 1037940100a4SChris Mason 10386cf8bfbfSDan Carpenter if (IS_ERR(em)) 10396c282eb4SLi Zefan return NULL; 1040940100a4SChris Mason } 1041940100a4SChris Mason 10426c282eb4SLi Zefan return em; 10436c282eb4SLi Zefan } 10446c282eb4SLi Zefan 10456c282eb4SLi Zefan static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em) 10466c282eb4SLi Zefan { 10476c282eb4SLi Zefan struct extent_map *next; 10486c282eb4SLi Zefan bool ret = true; 10496c282eb4SLi Zefan 10506c282eb4SLi Zefan /* this is the last extent */ 10516c282eb4SLi Zefan if (em->start + em->len >= i_size_read(inode)) 10526c282eb4SLi Zefan return false; 10536c282eb4SLi Zefan 10546c282eb4SLi Zefan next = defrag_lookup_extent(inode, em->start + em->len); 1055e2127cf0SFilipe Manana if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE || 1056e2127cf0SFilipe Manana (em->block_start + em->block_len == next->block_start)) 10576c282eb4SLi Zefan ret = false; 10586c282eb4SLi Zefan 10596c282eb4SLi Zefan free_extent_map(next); 10606c282eb4SLi Zefan return ret; 10616c282eb4SLi Zefan } 10626c282eb4SLi Zefan 10636c282eb4SLi Zefan static int should_defrag_range(struct inode *inode, u64 start, int thresh, 1064a43a2111SAndrew Mahone u64 *last_len, u64 *skip, u64 *defrag_end, 1065a43a2111SAndrew Mahone int compress) 10666c282eb4SLi Zefan { 10676c282eb4SLi Zefan struct extent_map *em; 10686c282eb4SLi Zefan int ret = 1; 10696c282eb4SLi Zefan bool next_mergeable = true; 10706c282eb4SLi Zefan 10716c282eb4SLi Zefan /* 10726c282eb4SLi Zefan * make sure that once we start defragging an extent, we keep on 10736c282eb4SLi Zefan * defragging it 10746c282eb4SLi Zefan */ 10756c282eb4SLi Zefan if (start < *defrag_end) 10766c282eb4SLi Zefan return 1; 10776c282eb4SLi Zefan 10786c282eb4SLi Zefan *skip = 0; 10796c282eb4SLi Zefan 10806c282eb4SLi Zefan em = defrag_lookup_extent(inode, start); 10816c282eb4SLi Zefan if (!em) 10826c282eb4SLi Zefan return 0; 10836c282eb4SLi Zefan 1084940100a4SChris Mason /* this will cover holes, and inline extents */ 108517ce6ef8SLiu Bo if (em->block_start >= EXTENT_MAP_LAST_BYTE) { 1086940100a4SChris Mason ret = 0; 108717ce6ef8SLiu Bo goto out; 108817ce6ef8SLiu Bo } 108917ce6ef8SLiu Bo 10906c282eb4SLi Zefan next_mergeable = defrag_check_next_extent(inode, em); 1091940100a4SChris Mason 1092940100a4SChris Mason /* 10936c282eb4SLi Zefan * we hit a real extent, if it is big or the next extent is not a 10946c282eb4SLi Zefan * real extent, don't bother defragging it 1095940100a4SChris Mason */ 1096a43a2111SAndrew Mahone if (!compress && (*last_len == 0 || *last_len >= thresh) && 10976c282eb4SLi Zefan (em->len >= thresh || !next_mergeable)) 1098940100a4SChris Mason ret = 0; 109917ce6ef8SLiu Bo out: 1100940100a4SChris Mason /* 1101940100a4SChris Mason * last_len ends up being a counter of how many bytes we've defragged. 1102940100a4SChris Mason * every time we choose not to defrag an extent, we reset *last_len 1103940100a4SChris Mason * so that the next tiny extent will force a defrag. 1104940100a4SChris Mason * 1105940100a4SChris Mason * The end result of this is that tiny extents before a single big 1106940100a4SChris Mason * extent will force at least part of that big extent to be defragged. 1107940100a4SChris Mason */ 1108940100a4SChris Mason if (ret) { 1109940100a4SChris Mason *defrag_end = extent_map_end(em); 1110940100a4SChris Mason } else { 1111940100a4SChris Mason *last_len = 0; 1112940100a4SChris Mason *skip = extent_map_end(em); 1113940100a4SChris Mason *defrag_end = 0; 1114940100a4SChris Mason } 1115940100a4SChris Mason 1116940100a4SChris Mason free_extent_map(em); 1117940100a4SChris Mason return ret; 1118940100a4SChris Mason } 1119940100a4SChris Mason 11204cb5300bSChris Mason /* 11214cb5300bSChris Mason * it doesn't do much good to defrag one or two pages 11224cb5300bSChris Mason * at a time. This pulls in a nice chunk of pages 11234cb5300bSChris Mason * to COW and defrag. 11244cb5300bSChris Mason * 11254cb5300bSChris Mason * It also makes sure the delalloc code has enough 11264cb5300bSChris Mason * dirty data to avoid making new small extents as part 11274cb5300bSChris Mason * of the defrag 11284cb5300bSChris Mason * 11294cb5300bSChris Mason * It's a good idea to start RA on this range 11304cb5300bSChris Mason * before calling this. 11314cb5300bSChris Mason */ 11324cb5300bSChris Mason static int cluster_pages_for_defrag(struct inode *inode, 11334cb5300bSChris Mason struct page **pages, 11344cb5300bSChris Mason unsigned long start_index, 1135c41570c9SJustin Maggard unsigned long num_pages) 1136f46b5a66SChristoph Hellwig { 11374cb5300bSChris Mason unsigned long file_end; 11384cb5300bSChris Mason u64 isize = i_size_read(inode); 1139f46b5a66SChristoph Hellwig u64 page_start; 1140f46b5a66SChristoph Hellwig u64 page_end; 11411f12bd06SLiu Bo u64 page_cnt; 11424cb5300bSChris Mason int ret; 11434cb5300bSChris Mason int i; 11444cb5300bSChris Mason int i_done; 11454cb5300bSChris Mason struct btrfs_ordered_extent *ordered; 11464cb5300bSChris Mason struct extent_state *cached_state = NULL; 1147600a45e1SMiao Xie struct extent_io_tree *tree; 11483b16a4e3SJosef Bacik gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping); 11494cb5300bSChris Mason 11504cb5300bSChris Mason file_end = (isize - 1) >> PAGE_CACHE_SHIFT; 11511f12bd06SLiu Bo if (!isize || start_index > file_end) 11521f12bd06SLiu Bo return 0; 11531f12bd06SLiu Bo 11541f12bd06SLiu Bo page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1); 11554cb5300bSChris Mason 11564cb5300bSChris Mason ret = btrfs_delalloc_reserve_space(inode, 11571f12bd06SLiu Bo page_cnt << PAGE_CACHE_SHIFT); 11584cb5300bSChris Mason if (ret) 11594cb5300bSChris Mason return ret; 11604cb5300bSChris Mason i_done = 0; 1161600a45e1SMiao Xie tree = &BTRFS_I(inode)->io_tree; 11624cb5300bSChris Mason 11634cb5300bSChris Mason /* step one, lock all the pages */ 11641f12bd06SLiu Bo for (i = 0; i < page_cnt; i++) { 11654cb5300bSChris Mason struct page *page; 1166600a45e1SMiao Xie again: 1167a94733d0SJosef Bacik page = find_or_create_page(inode->i_mapping, 11683b16a4e3SJosef Bacik start_index + i, mask); 11694cb5300bSChris Mason if (!page) 11704cb5300bSChris Mason break; 11714cb5300bSChris Mason 1172600a45e1SMiao Xie page_start = page_offset(page); 1173600a45e1SMiao Xie page_end = page_start + PAGE_CACHE_SIZE - 1; 1174600a45e1SMiao Xie while (1) { 1175308d9800SFilipe Manana lock_extent_bits(tree, page_start, page_end, 1176308d9800SFilipe Manana 0, &cached_state); 1177600a45e1SMiao Xie ordered = btrfs_lookup_ordered_extent(inode, 1178600a45e1SMiao Xie page_start); 1179308d9800SFilipe Manana unlock_extent_cached(tree, page_start, page_end, 1180308d9800SFilipe Manana &cached_state, GFP_NOFS); 1181600a45e1SMiao Xie if (!ordered) 1182600a45e1SMiao Xie break; 1183600a45e1SMiao Xie 1184600a45e1SMiao Xie unlock_page(page); 1185600a45e1SMiao Xie btrfs_start_ordered_extent(inode, ordered, 1); 1186600a45e1SMiao Xie btrfs_put_ordered_extent(ordered); 1187600a45e1SMiao Xie lock_page(page); 11881f12bd06SLiu Bo /* 11891f12bd06SLiu Bo * we unlocked the page above, so we need check if 11901f12bd06SLiu Bo * it was released or not. 11911f12bd06SLiu Bo */ 11921f12bd06SLiu Bo if (page->mapping != inode->i_mapping) { 11931f12bd06SLiu Bo unlock_page(page); 11941f12bd06SLiu Bo page_cache_release(page); 11951f12bd06SLiu Bo goto again; 11961f12bd06SLiu Bo } 1197600a45e1SMiao Xie } 1198600a45e1SMiao Xie 11994cb5300bSChris Mason if (!PageUptodate(page)) { 12004cb5300bSChris Mason btrfs_readpage(NULL, page); 12014cb5300bSChris Mason lock_page(page); 12024cb5300bSChris Mason if (!PageUptodate(page)) { 12034cb5300bSChris Mason unlock_page(page); 12044cb5300bSChris Mason page_cache_release(page); 12054cb5300bSChris Mason ret = -EIO; 12064cb5300bSChris Mason break; 12074cb5300bSChris Mason } 12084cb5300bSChris Mason } 1209600a45e1SMiao Xie 1210600a45e1SMiao Xie if (page->mapping != inode->i_mapping) { 1211600a45e1SMiao Xie unlock_page(page); 1212600a45e1SMiao Xie page_cache_release(page); 1213600a45e1SMiao Xie goto again; 1214600a45e1SMiao Xie } 1215600a45e1SMiao Xie 12164cb5300bSChris Mason pages[i] = page; 12174cb5300bSChris Mason i_done++; 12184cb5300bSChris Mason } 12194cb5300bSChris Mason if (!i_done || ret) 12204cb5300bSChris Mason goto out; 12214cb5300bSChris Mason 12224cb5300bSChris Mason if (!(inode->i_sb->s_flags & MS_ACTIVE)) 12234cb5300bSChris Mason goto out; 12244cb5300bSChris Mason 12254cb5300bSChris Mason /* 12264cb5300bSChris Mason * so now we have a nice long stream of locked 12274cb5300bSChris Mason * and up to date pages, lets wait on them 12284cb5300bSChris Mason */ 12294cb5300bSChris Mason for (i = 0; i < i_done; i++) 12304cb5300bSChris Mason wait_on_page_writeback(pages[i]); 12314cb5300bSChris Mason 12324cb5300bSChris Mason page_start = page_offset(pages[0]); 12334cb5300bSChris Mason page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE; 12344cb5300bSChris Mason 12354cb5300bSChris Mason lock_extent_bits(&BTRFS_I(inode)->io_tree, 1236d0082371SJeff Mahoney page_start, page_end - 1, 0, &cached_state); 12374cb5300bSChris Mason clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, 12384cb5300bSChris Mason page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC | 12399e8a4a8bSLiu Bo EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0, 12409e8a4a8bSLiu Bo &cached_state, GFP_NOFS); 12414cb5300bSChris Mason 12421f12bd06SLiu Bo if (i_done != page_cnt) { 12439e0baf60SJosef Bacik spin_lock(&BTRFS_I(inode)->lock); 12449e0baf60SJosef Bacik BTRFS_I(inode)->outstanding_extents++; 12459e0baf60SJosef Bacik spin_unlock(&BTRFS_I(inode)->lock); 12464cb5300bSChris Mason btrfs_delalloc_release_space(inode, 12471f12bd06SLiu Bo (page_cnt - i_done) << PAGE_CACHE_SHIFT); 12484cb5300bSChris Mason } 12494cb5300bSChris Mason 12504cb5300bSChris Mason 12519e8a4a8bSLiu Bo set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1, 12529e8a4a8bSLiu Bo &cached_state, GFP_NOFS); 12534cb5300bSChris Mason 12544cb5300bSChris Mason unlock_extent_cached(&BTRFS_I(inode)->io_tree, 12554cb5300bSChris Mason page_start, page_end - 1, &cached_state, 12564cb5300bSChris Mason GFP_NOFS); 12574cb5300bSChris Mason 12584cb5300bSChris Mason for (i = 0; i < i_done; i++) { 12594cb5300bSChris Mason clear_page_dirty_for_io(pages[i]); 12604cb5300bSChris Mason ClearPageChecked(pages[i]); 12614cb5300bSChris Mason set_page_extent_mapped(pages[i]); 12624cb5300bSChris Mason set_page_dirty(pages[i]); 12634cb5300bSChris Mason unlock_page(pages[i]); 12644cb5300bSChris Mason page_cache_release(pages[i]); 12654cb5300bSChris Mason } 12664cb5300bSChris Mason return i_done; 12674cb5300bSChris Mason out: 12684cb5300bSChris Mason for (i = 0; i < i_done; i++) { 12694cb5300bSChris Mason unlock_page(pages[i]); 12704cb5300bSChris Mason page_cache_release(pages[i]); 12714cb5300bSChris Mason } 12721f12bd06SLiu Bo btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT); 12734cb5300bSChris Mason return ret; 12744cb5300bSChris Mason 12754cb5300bSChris Mason } 12764cb5300bSChris Mason 12774cb5300bSChris Mason int btrfs_defrag_file(struct inode *inode, struct file *file, 12784cb5300bSChris Mason struct btrfs_ioctl_defrag_range_args *range, 12794cb5300bSChris Mason u64 newer_than, unsigned long max_to_defrag) 12804cb5300bSChris Mason { 12814cb5300bSChris Mason struct btrfs_root *root = BTRFS_I(inode)->root; 12824cb5300bSChris Mason struct file_ra_state *ra = NULL; 12834cb5300bSChris Mason unsigned long last_index; 1284151a31b2SLi Zefan u64 isize = i_size_read(inode); 1285940100a4SChris Mason u64 last_len = 0; 1286940100a4SChris Mason u64 skip = 0; 1287940100a4SChris Mason u64 defrag_end = 0; 12884cb5300bSChris Mason u64 newer_off = range->start; 1289f46b5a66SChristoph Hellwig unsigned long i; 1290008873eaSLi Zefan unsigned long ra_index = 0; 1291f46b5a66SChristoph Hellwig int ret; 12924cb5300bSChris Mason int defrag_count = 0; 12931a419d85SLi Zefan int compress_type = BTRFS_COMPRESS_ZLIB; 12944cb5300bSChris Mason int extent_thresh = range->extent_thresh; 1295c41570c9SJustin Maggard unsigned long max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT; 1296c41570c9SJustin Maggard unsigned long cluster = max_cluster; 12974cb5300bSChris Mason u64 new_align = ~((u64)128 * 1024 - 1); 12984cb5300bSChris Mason struct page **pages = NULL; 12994cb5300bSChris Mason 13000abd5b17SLiu Bo if (isize == 0) 13010abd5b17SLiu Bo return 0; 13020abd5b17SLiu Bo 13030abd5b17SLiu Bo if (range->start >= isize) 13040abd5b17SLiu Bo return -EINVAL; 13051a419d85SLi Zefan 13061a419d85SLi Zefan if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) { 13071a419d85SLi Zefan if (range->compress_type > BTRFS_COMPRESS_TYPES) 13081a419d85SLi Zefan return -EINVAL; 13091a419d85SLi Zefan if (range->compress_type) 13101a419d85SLi Zefan compress_type = range->compress_type; 13111a419d85SLi Zefan } 1312f46b5a66SChristoph Hellwig 13130abd5b17SLiu Bo if (extent_thresh == 0) 13140abd5b17SLiu Bo extent_thresh = 256 * 1024; 1315f46b5a66SChristoph Hellwig 13164cb5300bSChris Mason /* 13174cb5300bSChris Mason * if we were not given a file, allocate a readahead 13184cb5300bSChris Mason * context 13194cb5300bSChris Mason */ 13204cb5300bSChris Mason if (!file) { 13214cb5300bSChris Mason ra = kzalloc(sizeof(*ra), GFP_NOFS); 13224cb5300bSChris Mason if (!ra) 13234cb5300bSChris Mason return -ENOMEM; 13244cb5300bSChris Mason file_ra_state_init(ra, inode->i_mapping); 13254cb5300bSChris Mason } else { 13264cb5300bSChris Mason ra = &file->f_ra; 13274cb5300bSChris Mason } 13284cb5300bSChris Mason 1329d9b0d9baSDulshani Gunawardhana pages = kmalloc_array(max_cluster, sizeof(struct page *), 13304cb5300bSChris Mason GFP_NOFS); 13314cb5300bSChris Mason if (!pages) { 13324cb5300bSChris Mason ret = -ENOMEM; 13334cb5300bSChris Mason goto out_ra; 13344cb5300bSChris Mason } 13354cb5300bSChris Mason 13364cb5300bSChris Mason /* find the last page to defrag */ 13371e701a32SChris Mason if (range->start + range->len > range->start) { 1338151a31b2SLi Zefan last_index = min_t(u64, isize - 1, 13391e701a32SChris Mason range->start + range->len - 1) >> PAGE_CACHE_SHIFT; 13401e701a32SChris Mason } else { 1341151a31b2SLi Zefan last_index = (isize - 1) >> PAGE_CACHE_SHIFT; 13421e701a32SChris Mason } 13431e701a32SChris Mason 13444cb5300bSChris Mason if (newer_than) { 13454cb5300bSChris Mason ret = find_new_extents(root, inode, newer_than, 13464cb5300bSChris Mason &newer_off, 64 * 1024); 13474cb5300bSChris Mason if (!ret) { 13484cb5300bSChris Mason range->start = newer_off; 13494cb5300bSChris Mason /* 13504cb5300bSChris Mason * we always align our defrag to help keep 13514cb5300bSChris Mason * the extents in the file evenly spaced 13524cb5300bSChris Mason */ 13534cb5300bSChris Mason i = (newer_off & new_align) >> PAGE_CACHE_SHIFT; 13544cb5300bSChris Mason } else 13554cb5300bSChris Mason goto out_ra; 13564cb5300bSChris Mason } else { 13571e701a32SChris Mason i = range->start >> PAGE_CACHE_SHIFT; 13584cb5300bSChris Mason } 13594cb5300bSChris Mason if (!max_to_defrag) 13607ec31b54SLiu Bo max_to_defrag = last_index + 1; 13614cb5300bSChris Mason 13622a0f7f57SLi Zefan /* 13632a0f7f57SLi Zefan * make writeback starts from i, so the defrag range can be 13642a0f7f57SLi Zefan * written sequentially. 13652a0f7f57SLi Zefan */ 13662a0f7f57SLi Zefan if (i < inode->i_mapping->writeback_index) 13672a0f7f57SLi Zefan inode->i_mapping->writeback_index = i; 13682a0f7f57SLi Zefan 1369f7f43cc8SChris Mason while (i <= last_index && defrag_count < max_to_defrag && 1370f7f43cc8SChris Mason (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >> 1371f7f43cc8SChris Mason PAGE_CACHE_SHIFT)) { 13724cb5300bSChris Mason /* 13734cb5300bSChris Mason * make sure we stop running if someone unmounts 13744cb5300bSChris Mason * the FS 13754cb5300bSChris Mason */ 13764cb5300bSChris Mason if (!(inode->i_sb->s_flags & MS_ACTIVE)) 13774cb5300bSChris Mason break; 13784cb5300bSChris Mason 1379210549ebSDavid Sterba if (btrfs_defrag_cancelled(root->fs_info)) { 1380efe120a0SFrank Holton printk(KERN_DEBUG "BTRFS: defrag_file cancelled\n"); 1381210549ebSDavid Sterba ret = -EAGAIN; 1382210549ebSDavid Sterba break; 1383210549ebSDavid Sterba } 1384210549ebSDavid Sterba 138566c26892SLiu Bo if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT, 13866c282eb4SLi Zefan extent_thresh, &last_len, &skip, 1387a43a2111SAndrew Mahone &defrag_end, range->flags & 1388a43a2111SAndrew Mahone BTRFS_DEFRAG_RANGE_COMPRESS)) { 1389940100a4SChris Mason unsigned long next; 1390940100a4SChris Mason /* 1391940100a4SChris Mason * the should_defrag function tells us how much to skip 1392940100a4SChris Mason * bump our counter by the suggested amount 1393940100a4SChris Mason */ 1394940100a4SChris Mason next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT; 1395940100a4SChris Mason i = max(i + 1, next); 1396940100a4SChris Mason continue; 1397940100a4SChris Mason } 1398008873eaSLi Zefan 1399008873eaSLi Zefan if (!newer_than) { 1400008873eaSLi Zefan cluster = (PAGE_CACHE_ALIGN(defrag_end) >> 1401008873eaSLi Zefan PAGE_CACHE_SHIFT) - i; 1402008873eaSLi Zefan cluster = min(cluster, max_cluster); 1403008873eaSLi Zefan } else { 1404008873eaSLi Zefan cluster = max_cluster; 1405008873eaSLi Zefan } 1406008873eaSLi Zefan 1407008873eaSLi Zefan if (i + cluster > ra_index) { 1408008873eaSLi Zefan ra_index = max(i, ra_index); 1409008873eaSLi Zefan btrfs_force_ra(inode->i_mapping, ra, file, ra_index, 1410008873eaSLi Zefan cluster); 1411008873eaSLi Zefan ra_index += max_cluster; 1412008873eaSLi Zefan } 14134cb5300bSChris Mason 1414ecb8bea8SLiu Bo mutex_lock(&inode->i_mutex); 1415633085c7SFilipe David Borba Manana if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) 1416633085c7SFilipe David Borba Manana BTRFS_I(inode)->force_compress = compress_type; 1417008873eaSLi Zefan ret = cluster_pages_for_defrag(inode, pages, i, cluster); 1418ecb8bea8SLiu Bo if (ret < 0) { 1419ecb8bea8SLiu Bo mutex_unlock(&inode->i_mutex); 14204cb5300bSChris Mason goto out_ra; 1421ecb8bea8SLiu Bo } 14224cb5300bSChris Mason 14234cb5300bSChris Mason defrag_count += ret; 1424d0e1d66bSNamjae Jeon balance_dirty_pages_ratelimited(inode->i_mapping); 1425ecb8bea8SLiu Bo mutex_unlock(&inode->i_mutex); 14264cb5300bSChris Mason 14274cb5300bSChris Mason if (newer_than) { 14284cb5300bSChris Mason if (newer_off == (u64)-1) 14294cb5300bSChris Mason break; 14304cb5300bSChris Mason 1431e1f041e1SLiu Bo if (ret > 0) 1432e1f041e1SLiu Bo i += ret; 1433e1f041e1SLiu Bo 14344cb5300bSChris Mason newer_off = max(newer_off + 1, 14354cb5300bSChris Mason (u64)i << PAGE_CACHE_SHIFT); 14364cb5300bSChris Mason 14374cb5300bSChris Mason ret = find_new_extents(root, inode, 14384cb5300bSChris Mason newer_than, &newer_off, 14394cb5300bSChris Mason 64 * 1024); 14404cb5300bSChris Mason if (!ret) { 14414cb5300bSChris Mason range->start = newer_off; 14424cb5300bSChris Mason i = (newer_off & new_align) >> PAGE_CACHE_SHIFT; 14434cb5300bSChris Mason } else { 14444cb5300bSChris Mason break; 1445940100a4SChris Mason } 14464cb5300bSChris Mason } else { 1447008873eaSLi Zefan if (ret > 0) { 1448cbcc8326SLi Zefan i += ret; 1449008873eaSLi Zefan last_len += ret << PAGE_CACHE_SHIFT; 1450008873eaSLi Zefan } else { 1451940100a4SChris Mason i++; 1452008873eaSLi Zefan last_len = 0; 1453008873eaSLi Zefan } 1454f46b5a66SChristoph Hellwig } 14554cb5300bSChris Mason } 1456f46b5a66SChristoph Hellwig 1457dec8ef90SFilipe Manana if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) { 14581e701a32SChris Mason filemap_flush(inode->i_mapping); 1459dec8ef90SFilipe Manana if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT, 1460dec8ef90SFilipe Manana &BTRFS_I(inode)->runtime_flags)) 1461dec8ef90SFilipe Manana filemap_flush(inode->i_mapping); 1462dec8ef90SFilipe Manana } 14631e701a32SChris Mason 14641e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 14651e701a32SChris Mason /* the filemap_flush will queue IO into the worker threads, but 14661e701a32SChris Mason * we have to make sure the IO is actually started and that 14671e701a32SChris Mason * ordered extents get created before we return 14681e701a32SChris Mason */ 14691e701a32SChris Mason atomic_inc(&root->fs_info->async_submit_draining); 14701e701a32SChris Mason while (atomic_read(&root->fs_info->nr_async_submits) || 14711e701a32SChris Mason atomic_read(&root->fs_info->async_delalloc_pages)) { 14721e701a32SChris Mason wait_event(root->fs_info->async_submit_wait, 14731e701a32SChris Mason (atomic_read(&root->fs_info->nr_async_submits) == 0 && 14741e701a32SChris Mason atomic_read(&root->fs_info->async_delalloc_pages) == 0)); 14751e701a32SChris Mason } 14761e701a32SChris Mason atomic_dec(&root->fs_info->async_submit_draining); 14771e701a32SChris Mason } 14781e701a32SChris Mason 14791a419d85SLi Zefan if (range->compress_type == BTRFS_COMPRESS_LZO) { 14802b0ce2c2SMitch Harder btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO); 14811a419d85SLi Zefan } 14821a419d85SLi Zefan 148360ccf82fSDiego Calleja ret = defrag_count; 1484940100a4SChris Mason 14854cb5300bSChris Mason out_ra: 1486633085c7SFilipe David Borba Manana if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) { 1487633085c7SFilipe David Borba Manana mutex_lock(&inode->i_mutex); 1488633085c7SFilipe David Borba Manana BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE; 1489633085c7SFilipe David Borba Manana mutex_unlock(&inode->i_mutex); 1490633085c7SFilipe David Borba Manana } 14914cb5300bSChris Mason if (!file) 14924cb5300bSChris Mason kfree(ra); 14934cb5300bSChris Mason kfree(pages); 1494940100a4SChris Mason return ret; 1495f46b5a66SChristoph Hellwig } 1496f46b5a66SChristoph Hellwig 1497198605a8SMiao Xie static noinline int btrfs_ioctl_resize(struct file *file, 149876dda93cSYan, Zheng void __user *arg) 1499f46b5a66SChristoph Hellwig { 1500f46b5a66SChristoph Hellwig u64 new_size; 1501f46b5a66SChristoph Hellwig u64 old_size; 1502f46b5a66SChristoph Hellwig u64 devid = 1; 1503496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 1504f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 1505f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 1506f46b5a66SChristoph Hellwig struct btrfs_device *device = NULL; 1507f46b5a66SChristoph Hellwig char *sizestr; 15089a40f122SGui Hecheng char *retptr; 1509f46b5a66SChristoph Hellwig char *devstr = NULL; 1510f46b5a66SChristoph Hellwig int ret = 0; 1511f46b5a66SChristoph Hellwig int mod = 0; 1512f46b5a66SChristoph Hellwig 1513e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 1514e441d54dSChris Mason return -EPERM; 1515e441d54dSChris Mason 1516198605a8SMiao Xie ret = mnt_want_write_file(file); 1517198605a8SMiao Xie if (ret) 1518198605a8SMiao Xie return ret; 1519198605a8SMiao Xie 15205ac00addSStefan Behrens if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running, 15215ac00addSStefan Behrens 1)) { 152297547676SMiao Xie mnt_drop_write_file(file); 1523e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 1524c9e9f97bSIlya Dryomov } 1525c9e9f97bSIlya Dryomov 15265ac00addSStefan Behrens mutex_lock(&root->fs_info->volume_mutex); 1527dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1528c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 1529c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 1530c9e9f97bSIlya Dryomov goto out; 1531c9e9f97bSIlya Dryomov } 15325516e595SMark Fasheh 15335516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1534f46b5a66SChristoph Hellwig 1535f46b5a66SChristoph Hellwig sizestr = vol_args->name; 1536f46b5a66SChristoph Hellwig devstr = strchr(sizestr, ':'); 1537f46b5a66SChristoph Hellwig if (devstr) { 1538f46b5a66SChristoph Hellwig sizestr = devstr + 1; 1539f46b5a66SChristoph Hellwig *devstr = '\0'; 1540f46b5a66SChristoph Hellwig devstr = vol_args->name; 154158dfae63SZhangZhen ret = kstrtoull(devstr, 10, &devid); 154258dfae63SZhangZhen if (ret) 154358dfae63SZhangZhen goto out_free; 1544dfd79829SMiao Xie if (!devid) { 1545dfd79829SMiao Xie ret = -EINVAL; 1546dfd79829SMiao Xie goto out_free; 1547dfd79829SMiao Xie } 1548efe120a0SFrank Holton btrfs_info(root->fs_info, "resizing devid %llu", devid); 1549f46b5a66SChristoph Hellwig } 1550dba60f3fSMiao Xie 1551aa1b8cd4SStefan Behrens device = btrfs_find_device(root->fs_info, devid, NULL, NULL); 1552f46b5a66SChristoph Hellwig if (!device) { 1553efe120a0SFrank Holton btrfs_info(root->fs_info, "resizer unable to find device %llu", 1554c1c9ff7cSGeert Uytterhoeven devid); 1555dfd79829SMiao Xie ret = -ENODEV; 1556c9e9f97bSIlya Dryomov goto out_free; 1557f46b5a66SChristoph Hellwig } 1558dba60f3fSMiao Xie 1559dba60f3fSMiao Xie if (!device->writeable) { 1560efe120a0SFrank Holton btrfs_info(root->fs_info, 1561efe120a0SFrank Holton "resizer unable to apply on readonly device %llu", 1562c1c9ff7cSGeert Uytterhoeven devid); 1563dfd79829SMiao Xie ret = -EPERM; 15644e42ae1bSLiu Bo goto out_free; 15654e42ae1bSLiu Bo } 15664e42ae1bSLiu Bo 1567f46b5a66SChristoph Hellwig if (!strcmp(sizestr, "max")) 1568f46b5a66SChristoph Hellwig new_size = device->bdev->bd_inode->i_size; 1569f46b5a66SChristoph Hellwig else { 1570f46b5a66SChristoph Hellwig if (sizestr[0] == '-') { 1571f46b5a66SChristoph Hellwig mod = -1; 1572f46b5a66SChristoph Hellwig sizestr++; 1573f46b5a66SChristoph Hellwig } else if (sizestr[0] == '+') { 1574f46b5a66SChristoph Hellwig mod = 1; 1575f46b5a66SChristoph Hellwig sizestr++; 1576f46b5a66SChristoph Hellwig } 15779a40f122SGui Hecheng new_size = memparse(sizestr, &retptr); 15789a40f122SGui Hecheng if (*retptr != '\0' || new_size == 0) { 1579f46b5a66SChristoph Hellwig ret = -EINVAL; 1580c9e9f97bSIlya Dryomov goto out_free; 1581f46b5a66SChristoph Hellwig } 1582f46b5a66SChristoph Hellwig } 1583f46b5a66SChristoph Hellwig 158463a212abSStefan Behrens if (device->is_tgtdev_for_dev_replace) { 1585dfd79829SMiao Xie ret = -EPERM; 158663a212abSStefan Behrens goto out_free; 158763a212abSStefan Behrens } 158863a212abSStefan Behrens 1589f46b5a66SChristoph Hellwig old_size = device->total_bytes; 1590f46b5a66SChristoph Hellwig 1591f46b5a66SChristoph Hellwig if (mod < 0) { 1592f46b5a66SChristoph Hellwig if (new_size > old_size) { 1593f46b5a66SChristoph Hellwig ret = -EINVAL; 1594c9e9f97bSIlya Dryomov goto out_free; 1595f46b5a66SChristoph Hellwig } 1596f46b5a66SChristoph Hellwig new_size = old_size - new_size; 1597f46b5a66SChristoph Hellwig } else if (mod > 0) { 1598eb8052e0SWenliang Fan if (new_size > ULLONG_MAX - old_size) { 1599902c68a4SGui Hecheng ret = -ERANGE; 1600eb8052e0SWenliang Fan goto out_free; 1601eb8052e0SWenliang Fan } 1602f46b5a66SChristoph Hellwig new_size = old_size + new_size; 1603f46b5a66SChristoph Hellwig } 1604f46b5a66SChristoph Hellwig 1605f46b5a66SChristoph Hellwig if (new_size < 256 * 1024 * 1024) { 1606f46b5a66SChristoph Hellwig ret = -EINVAL; 1607c9e9f97bSIlya Dryomov goto out_free; 1608f46b5a66SChristoph Hellwig } 1609f46b5a66SChristoph Hellwig if (new_size > device->bdev->bd_inode->i_size) { 1610f46b5a66SChristoph Hellwig ret = -EFBIG; 1611c9e9f97bSIlya Dryomov goto out_free; 1612f46b5a66SChristoph Hellwig } 1613f46b5a66SChristoph Hellwig 1614f46b5a66SChristoph Hellwig do_div(new_size, root->sectorsize); 1615f46b5a66SChristoph Hellwig new_size *= root->sectorsize; 1616f46b5a66SChristoph Hellwig 1617efe120a0SFrank Holton printk_in_rcu(KERN_INFO "BTRFS: new size for %s is %llu\n", 1618c1c9ff7cSGeert Uytterhoeven rcu_str_deref(device->name), new_size); 1619f46b5a66SChristoph Hellwig 1620f46b5a66SChristoph Hellwig if (new_size > old_size) { 1621a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 162298d5dc13STsutomu Itoh if (IS_ERR(trans)) { 162398d5dc13STsutomu Itoh ret = PTR_ERR(trans); 1624c9e9f97bSIlya Dryomov goto out_free; 162598d5dc13STsutomu Itoh } 1626f46b5a66SChristoph Hellwig ret = btrfs_grow_device(trans, device, new_size); 1627f46b5a66SChristoph Hellwig btrfs_commit_transaction(trans, root); 1628ece7d20eSMike Fleetwood } else if (new_size < old_size) { 1629f46b5a66SChristoph Hellwig ret = btrfs_shrink_device(device, new_size); 16300253f40eSjeff.liu } /* equal, nothing need to do */ 1631f46b5a66SChristoph Hellwig 1632c9e9f97bSIlya Dryomov out_free: 1633f46b5a66SChristoph Hellwig kfree(vol_args); 1634c9e9f97bSIlya Dryomov out: 1635c9e9f97bSIlya Dryomov mutex_unlock(&root->fs_info->volume_mutex); 16365ac00addSStefan Behrens atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0); 163718f39c41SIlya Dryomov mnt_drop_write_file(file); 1638f46b5a66SChristoph Hellwig return ret; 1639f46b5a66SChristoph Hellwig } 1640f46b5a66SChristoph Hellwig 164172fd032eSSage Weil static noinline int btrfs_ioctl_snap_create_transid(struct file *file, 16426f72c7e2SArne Jansen char *name, unsigned long fd, int subvol, 16436f72c7e2SArne Jansen u64 *transid, bool readonly, 16448696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 1645f46b5a66SChristoph Hellwig { 1646f46b5a66SChristoph Hellwig int namelen; 16473de4586cSChris Mason int ret = 0; 1648f46b5a66SChristoph Hellwig 1649a874a63eSLiu Bo ret = mnt_want_write_file(file); 1650a874a63eSLiu Bo if (ret) 1651a874a63eSLiu Bo goto out; 1652a874a63eSLiu Bo 165372fd032eSSage Weil namelen = strlen(name); 165472fd032eSSage Weil if (strchr(name, '/')) { 1655f46b5a66SChristoph Hellwig ret = -EINVAL; 1656a874a63eSLiu Bo goto out_drop_write; 1657f46b5a66SChristoph Hellwig } 1658f46b5a66SChristoph Hellwig 165916780cabSChris Mason if (name[0] == '.' && 166016780cabSChris Mason (namelen == 1 || (name[1] == '.' && namelen == 2))) { 166116780cabSChris Mason ret = -EEXIST; 1662a874a63eSLiu Bo goto out_drop_write; 166316780cabSChris Mason } 166416780cabSChris Mason 16653de4586cSChris Mason if (subvol) { 166672fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 16676f72c7e2SArne Jansen NULL, transid, readonly, inherit); 1668cb8e7090SChristoph Hellwig } else { 16692903ff01SAl Viro struct fd src = fdget(fd); 16703de4586cSChris Mason struct inode *src_inode; 16712903ff01SAl Viro if (!src.file) { 16723de4586cSChris Mason ret = -EINVAL; 1673a874a63eSLiu Bo goto out_drop_write; 16743de4586cSChris Mason } 16753de4586cSChris Mason 1676496ad9aaSAl Viro src_inode = file_inode(src.file); 1677496ad9aaSAl Viro if (src_inode->i_sb != file_inode(file)->i_sb) { 1678efe120a0SFrank Holton btrfs_info(BTRFS_I(src_inode)->root->fs_info, 1679efe120a0SFrank Holton "Snapshot src from another FS"); 168023ad5b17SKusanagi Kouichi ret = -EXDEV; 1681d0242061SDavid Sterba } else if (!inode_owner_or_capable(src_inode)) { 1682d0242061SDavid Sterba /* 1683d0242061SDavid Sterba * Subvolume creation is not restricted, but snapshots 1684d0242061SDavid Sterba * are limited to own subvolumes only 1685d0242061SDavid Sterba */ 1686d0242061SDavid Sterba ret = -EPERM; 1687ecd18815SAl Viro } else { 168872fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 168972fd032eSSage Weil BTRFS_I(src_inode)->root, 16906f72c7e2SArne Jansen transid, readonly, inherit); 1691ecd18815SAl Viro } 16922903ff01SAl Viro fdput(src); 1693cb8e7090SChristoph Hellwig } 1694a874a63eSLiu Bo out_drop_write: 1695a874a63eSLiu Bo mnt_drop_write_file(file); 1696f46b5a66SChristoph Hellwig out: 169772fd032eSSage Weil return ret; 169872fd032eSSage Weil } 169972fd032eSSage Weil 170072fd032eSSage Weil static noinline int btrfs_ioctl_snap_create(struct file *file, 1701fa0d2b9bSLi Zefan void __user *arg, int subvol) 170272fd032eSSage Weil { 1703fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args *vol_args; 170472fd032eSSage Weil int ret; 170572fd032eSSage Weil 1706fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1707fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1708fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1709fa0d2b9bSLi Zefan vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1710fa0d2b9bSLi Zefan 1711fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 1712b83cc969SLi Zefan vol_args->fd, subvol, 17136f72c7e2SArne Jansen NULL, false, NULL); 1714fa0d2b9bSLi Zefan 1715fa0d2b9bSLi Zefan kfree(vol_args); 1716fa0d2b9bSLi Zefan return ret; 1717fa0d2b9bSLi Zefan } 1718fa0d2b9bSLi Zefan 1719fa0d2b9bSLi Zefan static noinline int btrfs_ioctl_snap_create_v2(struct file *file, 1720fa0d2b9bSLi Zefan void __user *arg, int subvol) 1721fa0d2b9bSLi Zefan { 1722fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args_v2 *vol_args; 1723fa0d2b9bSLi Zefan int ret; 1724fdfb1e4fSLi Zefan u64 transid = 0; 1725fdfb1e4fSLi Zefan u64 *ptr = NULL; 1726b83cc969SLi Zefan bool readonly = false; 17276f72c7e2SArne Jansen struct btrfs_qgroup_inherit *inherit = NULL; 172872fd032eSSage Weil 1729fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1730fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1731fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1732fa0d2b9bSLi Zefan vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 1733fdfb1e4fSLi Zefan 1734b83cc969SLi Zefan if (vol_args->flags & 17356f72c7e2SArne Jansen ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY | 17366f72c7e2SArne Jansen BTRFS_SUBVOL_QGROUP_INHERIT)) { 1737b83cc969SLi Zefan ret = -EOPNOTSUPP; 1738fdfb1e4fSLi Zefan goto out; 1739fdfb1e4fSLi Zefan } 1740fdfb1e4fSLi Zefan 1741fa0d2b9bSLi Zefan if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC) 1742fdfb1e4fSLi Zefan ptr = &transid; 1743b83cc969SLi Zefan if (vol_args->flags & BTRFS_SUBVOL_RDONLY) 1744b83cc969SLi Zefan readonly = true; 17456f72c7e2SArne Jansen if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) { 17466f72c7e2SArne Jansen if (vol_args->size > PAGE_CACHE_SIZE) { 17476f72c7e2SArne Jansen ret = -EINVAL; 17486f72c7e2SArne Jansen goto out; 17496f72c7e2SArne Jansen } 17506f72c7e2SArne Jansen inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size); 17516f72c7e2SArne Jansen if (IS_ERR(inherit)) { 17526f72c7e2SArne Jansen ret = PTR_ERR(inherit); 17536f72c7e2SArne Jansen goto out; 17546f72c7e2SArne Jansen } 17556f72c7e2SArne Jansen } 175675eaa0e2SSage Weil 1757fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 17586f72c7e2SArne Jansen vol_args->fd, subvol, ptr, 17598696c533SMiao Xie readonly, inherit); 176075eaa0e2SSage Weil 1761fdfb1e4fSLi Zefan if (ret == 0 && ptr && 176275eaa0e2SSage Weil copy_to_user(arg + 1763fdfb1e4fSLi Zefan offsetof(struct btrfs_ioctl_vol_args_v2, 1764fdfb1e4fSLi Zefan transid), ptr, sizeof(*ptr))) 176575eaa0e2SSage Weil ret = -EFAULT; 1766fdfb1e4fSLi Zefan out: 1767f46b5a66SChristoph Hellwig kfree(vol_args); 17686f72c7e2SArne Jansen kfree(inherit); 1769f46b5a66SChristoph Hellwig return ret; 1770f46b5a66SChristoph Hellwig } 1771f46b5a66SChristoph Hellwig 17720caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_getflags(struct file *file, 17730caa102dSLi Zefan void __user *arg) 17740caa102dSLi Zefan { 1775496ad9aaSAl Viro struct inode *inode = file_inode(file); 17760caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 17770caa102dSLi Zefan int ret = 0; 17780caa102dSLi Zefan u64 flags = 0; 17790caa102dSLi Zefan 178033345d01SLi Zefan if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) 17810caa102dSLi Zefan return -EINVAL; 17820caa102dSLi Zefan 17830caa102dSLi Zefan down_read(&root->fs_info->subvol_sem); 17840caa102dSLi Zefan if (btrfs_root_readonly(root)) 17850caa102dSLi Zefan flags |= BTRFS_SUBVOL_RDONLY; 17860caa102dSLi Zefan up_read(&root->fs_info->subvol_sem); 17870caa102dSLi Zefan 17880caa102dSLi Zefan if (copy_to_user(arg, &flags, sizeof(flags))) 17890caa102dSLi Zefan ret = -EFAULT; 17900caa102dSLi Zefan 17910caa102dSLi Zefan return ret; 17920caa102dSLi Zefan } 17930caa102dSLi Zefan 17940caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_setflags(struct file *file, 17950caa102dSLi Zefan void __user *arg) 17960caa102dSLi Zefan { 1797496ad9aaSAl Viro struct inode *inode = file_inode(file); 17980caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 17990caa102dSLi Zefan struct btrfs_trans_handle *trans; 18000caa102dSLi Zefan u64 root_flags; 18010caa102dSLi Zefan u64 flags; 18020caa102dSLi Zefan int ret = 0; 18030caa102dSLi Zefan 1804bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 1805bd60ea0fSDavid Sterba return -EPERM; 1806bd60ea0fSDavid Sterba 1807b9ca0664SLiu Bo ret = mnt_want_write_file(file); 1808b9ca0664SLiu Bo if (ret) 1809b9ca0664SLiu Bo goto out; 18100caa102dSLi Zefan 1811b9ca0664SLiu Bo if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { 1812b9ca0664SLiu Bo ret = -EINVAL; 1813b9ca0664SLiu Bo goto out_drop_write; 1814b9ca0664SLiu Bo } 18150caa102dSLi Zefan 1816b9ca0664SLiu Bo if (copy_from_user(&flags, arg, sizeof(flags))) { 1817b9ca0664SLiu Bo ret = -EFAULT; 1818b9ca0664SLiu Bo goto out_drop_write; 1819b9ca0664SLiu Bo } 18200caa102dSLi Zefan 1821b9ca0664SLiu Bo if (flags & BTRFS_SUBVOL_CREATE_ASYNC) { 1822b9ca0664SLiu Bo ret = -EINVAL; 1823b9ca0664SLiu Bo goto out_drop_write; 1824b9ca0664SLiu Bo } 18250caa102dSLi Zefan 1826b9ca0664SLiu Bo if (flags & ~BTRFS_SUBVOL_RDONLY) { 1827b9ca0664SLiu Bo ret = -EOPNOTSUPP; 1828b9ca0664SLiu Bo goto out_drop_write; 1829b9ca0664SLiu Bo } 18300caa102dSLi Zefan 18310caa102dSLi Zefan down_write(&root->fs_info->subvol_sem); 18320caa102dSLi Zefan 18330caa102dSLi Zefan /* nothing to do */ 18340caa102dSLi Zefan if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) 1835b9ca0664SLiu Bo goto out_drop_sem; 18360caa102dSLi Zefan 18370caa102dSLi Zefan root_flags = btrfs_root_flags(&root->root_item); 18382c686537SDavid Sterba if (flags & BTRFS_SUBVOL_RDONLY) { 18390caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 18400caa102dSLi Zefan root_flags | BTRFS_ROOT_SUBVOL_RDONLY); 18412c686537SDavid Sterba } else { 18422c686537SDavid Sterba /* 18432c686537SDavid Sterba * Block RO -> RW transition if this subvolume is involved in 18442c686537SDavid Sterba * send 18452c686537SDavid Sterba */ 18462c686537SDavid Sterba spin_lock(&root->root_item_lock); 18472c686537SDavid Sterba if (root->send_in_progress == 0) { 18480caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 18490caa102dSLi Zefan root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); 18502c686537SDavid Sterba spin_unlock(&root->root_item_lock); 18512c686537SDavid Sterba } else { 18522c686537SDavid Sterba spin_unlock(&root->root_item_lock); 18532c686537SDavid Sterba btrfs_warn(root->fs_info, 18542c686537SDavid Sterba "Attempt to set subvolume %llu read-write during send", 18552c686537SDavid Sterba root->root_key.objectid); 18562c686537SDavid Sterba ret = -EPERM; 18572c686537SDavid Sterba goto out_drop_sem; 18582c686537SDavid Sterba } 18592c686537SDavid Sterba } 18600caa102dSLi Zefan 18610caa102dSLi Zefan trans = btrfs_start_transaction(root, 1); 18620caa102dSLi Zefan if (IS_ERR(trans)) { 18630caa102dSLi Zefan ret = PTR_ERR(trans); 18640caa102dSLi Zefan goto out_reset; 18650caa102dSLi Zefan } 18660caa102dSLi Zefan 1867b4dc2b8cSLi Zefan ret = btrfs_update_root(trans, root->fs_info->tree_root, 18680caa102dSLi Zefan &root->root_key, &root->root_item); 18690caa102dSLi Zefan 18700caa102dSLi Zefan btrfs_commit_transaction(trans, root); 18710caa102dSLi Zefan out_reset: 18720caa102dSLi Zefan if (ret) 18730caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, root_flags); 1874b9ca0664SLiu Bo out_drop_sem: 18750caa102dSLi Zefan up_write(&root->fs_info->subvol_sem); 1876b9ca0664SLiu Bo out_drop_write: 1877b9ca0664SLiu Bo mnt_drop_write_file(file); 1878b9ca0664SLiu Bo out: 18790caa102dSLi Zefan return ret; 18800caa102dSLi Zefan } 18810caa102dSLi Zefan 188276dda93cSYan, Zheng /* 188376dda93cSYan, Zheng * helper to check if the subvolume references other subvolumes 188476dda93cSYan, Zheng */ 188576dda93cSYan, Zheng static noinline int may_destroy_subvol(struct btrfs_root *root) 188676dda93cSYan, Zheng { 188776dda93cSYan, Zheng struct btrfs_path *path; 1888175a2b87SJosef Bacik struct btrfs_dir_item *di; 188976dda93cSYan, Zheng struct btrfs_key key; 1890175a2b87SJosef Bacik u64 dir_id; 189176dda93cSYan, Zheng int ret; 189276dda93cSYan, Zheng 189376dda93cSYan, Zheng path = btrfs_alloc_path(); 189476dda93cSYan, Zheng if (!path) 189576dda93cSYan, Zheng return -ENOMEM; 189676dda93cSYan, Zheng 1897175a2b87SJosef Bacik /* Make sure this root isn't set as the default subvol */ 1898175a2b87SJosef Bacik dir_id = btrfs_super_root_dir(root->fs_info->super_copy); 1899175a2b87SJosef Bacik di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path, 1900175a2b87SJosef Bacik dir_id, "default", 7, 0); 1901175a2b87SJosef Bacik if (di && !IS_ERR(di)) { 1902175a2b87SJosef Bacik btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key); 1903175a2b87SJosef Bacik if (key.objectid == root->root_key.objectid) { 190472de6b53SGuangyu Sun ret = -EPERM; 190572de6b53SGuangyu Sun btrfs_err(root->fs_info, "deleting default subvolume " 190672de6b53SGuangyu Sun "%llu is not allowed", key.objectid); 1907175a2b87SJosef Bacik goto out; 1908175a2b87SJosef Bacik } 1909175a2b87SJosef Bacik btrfs_release_path(path); 1910175a2b87SJosef Bacik } 1911175a2b87SJosef Bacik 191276dda93cSYan, Zheng key.objectid = root->root_key.objectid; 191376dda93cSYan, Zheng key.type = BTRFS_ROOT_REF_KEY; 191476dda93cSYan, Zheng key.offset = (u64)-1; 191576dda93cSYan, Zheng 191676dda93cSYan, Zheng ret = btrfs_search_slot(NULL, root->fs_info->tree_root, 191776dda93cSYan, Zheng &key, path, 0, 0); 191876dda93cSYan, Zheng if (ret < 0) 191976dda93cSYan, Zheng goto out; 192076dda93cSYan, Zheng BUG_ON(ret == 0); 192176dda93cSYan, Zheng 192276dda93cSYan, Zheng ret = 0; 192376dda93cSYan, Zheng if (path->slots[0] > 0) { 192476dda93cSYan, Zheng path->slots[0]--; 192576dda93cSYan, Zheng btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 192676dda93cSYan, Zheng if (key.objectid == root->root_key.objectid && 192776dda93cSYan, Zheng key.type == BTRFS_ROOT_REF_KEY) 192876dda93cSYan, Zheng ret = -ENOTEMPTY; 192976dda93cSYan, Zheng } 193076dda93cSYan, Zheng out: 193176dda93cSYan, Zheng btrfs_free_path(path); 193276dda93cSYan, Zheng return ret; 193376dda93cSYan, Zheng } 193476dda93cSYan, Zheng 1935ac8e9819SChris Mason static noinline int key_in_sk(struct btrfs_key *key, 1936ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk) 1937ac8e9819SChris Mason { 1938abc6e134SChris Mason struct btrfs_key test; 1939abc6e134SChris Mason int ret; 1940abc6e134SChris Mason 1941abc6e134SChris Mason test.objectid = sk->min_objectid; 1942abc6e134SChris Mason test.type = sk->min_type; 1943abc6e134SChris Mason test.offset = sk->min_offset; 1944abc6e134SChris Mason 1945abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1946abc6e134SChris Mason if (ret < 0) 1947ac8e9819SChris Mason return 0; 1948abc6e134SChris Mason 1949abc6e134SChris Mason test.objectid = sk->max_objectid; 1950abc6e134SChris Mason test.type = sk->max_type; 1951abc6e134SChris Mason test.offset = sk->max_offset; 1952abc6e134SChris Mason 1953abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1954abc6e134SChris Mason if (ret > 0) 1955ac8e9819SChris Mason return 0; 1956ac8e9819SChris Mason return 1; 1957ac8e9819SChris Mason } 1958ac8e9819SChris Mason 1959ac8e9819SChris Mason static noinline int copy_to_sk(struct btrfs_root *root, 1960ac8e9819SChris Mason struct btrfs_path *path, 1961ac8e9819SChris Mason struct btrfs_key *key, 1962ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk, 19639b6e817dSGerhard Heift size_t *buf_size, 1964ba346b35SGerhard Heift char __user *ubuf, 1965ac8e9819SChris Mason unsigned long *sk_offset, 1966ac8e9819SChris Mason int *num_found) 1967ac8e9819SChris Mason { 1968ac8e9819SChris Mason u64 found_transid; 1969ac8e9819SChris Mason struct extent_buffer *leaf; 1970ac8e9819SChris Mason struct btrfs_ioctl_search_header sh; 1971ac8e9819SChris Mason unsigned long item_off; 1972ac8e9819SChris Mason unsigned long item_len; 1973ac8e9819SChris Mason int nritems; 1974ac8e9819SChris Mason int i; 1975ac8e9819SChris Mason int slot; 1976ac8e9819SChris Mason int ret = 0; 1977ac8e9819SChris Mason 1978ac8e9819SChris Mason leaf = path->nodes[0]; 1979ac8e9819SChris Mason slot = path->slots[0]; 1980ac8e9819SChris Mason nritems = btrfs_header_nritems(leaf); 1981ac8e9819SChris Mason 1982ac8e9819SChris Mason if (btrfs_header_generation(leaf) > sk->max_transid) { 1983ac8e9819SChris Mason i = nritems; 1984ac8e9819SChris Mason goto advance_key; 1985ac8e9819SChris Mason } 1986ac8e9819SChris Mason found_transid = btrfs_header_generation(leaf); 1987ac8e9819SChris Mason 1988ac8e9819SChris Mason for (i = slot; i < nritems; i++) { 1989ac8e9819SChris Mason item_off = btrfs_item_ptr_offset(leaf, i); 1990ac8e9819SChris Mason item_len = btrfs_item_size_nr(leaf, i); 1991ac8e9819SChris Mason 199203b71c6cSGabriel de Perthuis btrfs_item_key_to_cpu(leaf, key, i); 199303b71c6cSGabriel de Perthuis if (!key_in_sk(key, sk)) 199403b71c6cSGabriel de Perthuis continue; 199503b71c6cSGabriel de Perthuis 19969b6e817dSGerhard Heift if (sizeof(sh) + item_len > *buf_size) { 19978f5f6178SGerhard Heift if (*num_found) { 19988f5f6178SGerhard Heift ret = 1; 19998f5f6178SGerhard Heift goto out; 20008f5f6178SGerhard Heift } 20018f5f6178SGerhard Heift 20028f5f6178SGerhard Heift /* 20038f5f6178SGerhard Heift * return one empty item back for v1, which does not 20048f5f6178SGerhard Heift * handle -EOVERFLOW 20058f5f6178SGerhard Heift */ 20068f5f6178SGerhard Heift 20079b6e817dSGerhard Heift *buf_size = sizeof(sh) + item_len; 2008ac8e9819SChris Mason item_len = 0; 20098f5f6178SGerhard Heift ret = -EOVERFLOW; 20108f5f6178SGerhard Heift } 2011ac8e9819SChris Mason 20129b6e817dSGerhard Heift if (sizeof(sh) + item_len + *sk_offset > *buf_size) { 2013ac8e9819SChris Mason ret = 1; 201425c9bc2eSGerhard Heift goto out; 2015ac8e9819SChris Mason } 2016ac8e9819SChris Mason 2017ac8e9819SChris Mason sh.objectid = key->objectid; 2018ac8e9819SChris Mason sh.offset = key->offset; 2019ac8e9819SChris Mason sh.type = key->type; 2020ac8e9819SChris Mason sh.len = item_len; 2021ac8e9819SChris Mason sh.transid = found_transid; 2022ac8e9819SChris Mason 2023ac8e9819SChris Mason /* copy search result header */ 2024ba346b35SGerhard Heift if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) { 2025ba346b35SGerhard Heift ret = -EFAULT; 2026ba346b35SGerhard Heift goto out; 2027ba346b35SGerhard Heift } 2028ba346b35SGerhard Heift 2029ac8e9819SChris Mason *sk_offset += sizeof(sh); 2030ac8e9819SChris Mason 2031ac8e9819SChris Mason if (item_len) { 2032ba346b35SGerhard Heift char __user *up = ubuf + *sk_offset; 2033ac8e9819SChris Mason /* copy the item */ 2034ba346b35SGerhard Heift if (read_extent_buffer_to_user(leaf, up, 2035ba346b35SGerhard Heift item_off, item_len)) { 2036ba346b35SGerhard Heift ret = -EFAULT; 2037ba346b35SGerhard Heift goto out; 2038ba346b35SGerhard Heift } 2039ba346b35SGerhard Heift 2040ac8e9819SChris Mason *sk_offset += item_len; 2041ac8e9819SChris Mason } 2042e2156867SHugo Mills (*num_found)++; 2043ac8e9819SChris Mason 20448f5f6178SGerhard Heift if (ret) /* -EOVERFLOW from above */ 20458f5f6178SGerhard Heift goto out; 20468f5f6178SGerhard Heift 204725c9bc2eSGerhard Heift if (*num_found >= sk->nr_items) { 204825c9bc2eSGerhard Heift ret = 1; 204925c9bc2eSGerhard Heift goto out; 205025c9bc2eSGerhard Heift } 2051ac8e9819SChris Mason } 2052ac8e9819SChris Mason advance_key: 2053ac8e9819SChris Mason ret = 0; 2054abc6e134SChris Mason if (key->offset < (u64)-1 && key->offset < sk->max_offset) 2055abc6e134SChris Mason key->offset++; 2056abc6e134SChris Mason else if (key->type < (u8)-1 && key->type < sk->max_type) { 2057abc6e134SChris Mason key->offset = 0; 2058abc6e134SChris Mason key->type++; 2059abc6e134SChris Mason } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) { 2060abc6e134SChris Mason key->offset = 0; 2061abc6e134SChris Mason key->type = 0; 2062abc6e134SChris Mason key->objectid++; 2063abc6e134SChris Mason } else 2064abc6e134SChris Mason ret = 1; 206525c9bc2eSGerhard Heift out: 2066ba346b35SGerhard Heift /* 2067ba346b35SGerhard Heift * 0: all items from this leaf copied, continue with next 2068ba346b35SGerhard Heift * 1: * more items can be copied, but unused buffer is too small 2069ba346b35SGerhard Heift * * all items were found 2070ba346b35SGerhard Heift * Either way, it will stops the loop which iterates to the next 2071ba346b35SGerhard Heift * leaf 2072ba346b35SGerhard Heift * -EOVERFLOW: item was to large for buffer 2073ba346b35SGerhard Heift * -EFAULT: could not copy extent buffer back to userspace 2074ba346b35SGerhard Heift */ 2075ac8e9819SChris Mason return ret; 2076ac8e9819SChris Mason } 2077ac8e9819SChris Mason 2078ac8e9819SChris Mason static noinline int search_ioctl(struct inode *inode, 207912544442SGerhard Heift struct btrfs_ioctl_search_key *sk, 20809b6e817dSGerhard Heift size_t *buf_size, 2081ba346b35SGerhard Heift char __user *ubuf) 2082ac8e9819SChris Mason { 2083ac8e9819SChris Mason struct btrfs_root *root; 2084ac8e9819SChris Mason struct btrfs_key key; 2085ac8e9819SChris Mason struct btrfs_path *path; 2086ac8e9819SChris Mason struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info; 2087ac8e9819SChris Mason int ret; 2088ac8e9819SChris Mason int num_found = 0; 2089ac8e9819SChris Mason unsigned long sk_offset = 0; 2090ac8e9819SChris Mason 20919b6e817dSGerhard Heift if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) { 20929b6e817dSGerhard Heift *buf_size = sizeof(struct btrfs_ioctl_search_header); 209312544442SGerhard Heift return -EOVERFLOW; 20949b6e817dSGerhard Heift } 209512544442SGerhard Heift 2096ac8e9819SChris Mason path = btrfs_alloc_path(); 2097ac8e9819SChris Mason if (!path) 2098ac8e9819SChris Mason return -ENOMEM; 2099ac8e9819SChris Mason 2100ac8e9819SChris Mason if (sk->tree_id == 0) { 2101ac8e9819SChris Mason /* search the root of the inode that was passed */ 2102ac8e9819SChris Mason root = BTRFS_I(inode)->root; 2103ac8e9819SChris Mason } else { 2104ac8e9819SChris Mason key.objectid = sk->tree_id; 2105ac8e9819SChris Mason key.type = BTRFS_ROOT_ITEM_KEY; 2106ac8e9819SChris Mason key.offset = (u64)-1; 2107ac8e9819SChris Mason root = btrfs_read_fs_root_no_name(info, &key); 2108ac8e9819SChris Mason if (IS_ERR(root)) { 2109efe120a0SFrank Holton printk(KERN_ERR "BTRFS: could not find root %llu\n", 2110ac8e9819SChris Mason sk->tree_id); 2111ac8e9819SChris Mason btrfs_free_path(path); 2112ac8e9819SChris Mason return -ENOENT; 2113ac8e9819SChris Mason } 2114ac8e9819SChris Mason } 2115ac8e9819SChris Mason 2116ac8e9819SChris Mason key.objectid = sk->min_objectid; 2117ac8e9819SChris Mason key.type = sk->min_type; 2118ac8e9819SChris Mason key.offset = sk->min_offset; 2119ac8e9819SChris Mason 2120ac8e9819SChris Mason path->keep_locks = 1; 2121ac8e9819SChris Mason 2122ac8e9819SChris Mason while (1) { 21236174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &key, path, sk->min_transid); 2124ac8e9819SChris Mason if (ret != 0) { 2125ac8e9819SChris Mason if (ret > 0) 2126ac8e9819SChris Mason ret = 0; 2127ac8e9819SChris Mason goto err; 2128ac8e9819SChris Mason } 2129ba346b35SGerhard Heift ret = copy_to_sk(root, path, &key, sk, buf_size, ubuf, 2130ac8e9819SChris Mason &sk_offset, &num_found); 2131b3b4aa74SDavid Sterba btrfs_release_path(path); 213225c9bc2eSGerhard Heift if (ret) 2133ac8e9819SChris Mason break; 2134ac8e9819SChris Mason 2135ac8e9819SChris Mason } 21368f5f6178SGerhard Heift if (ret > 0) 2137ac8e9819SChris Mason ret = 0; 2138ac8e9819SChris Mason err: 2139ac8e9819SChris Mason sk->nr_items = num_found; 2140ac8e9819SChris Mason btrfs_free_path(path); 2141ac8e9819SChris Mason return ret; 2142ac8e9819SChris Mason } 2143ac8e9819SChris Mason 2144ac8e9819SChris Mason static noinline int btrfs_ioctl_tree_search(struct file *file, 2145ac8e9819SChris Mason void __user *argp) 2146ac8e9819SChris Mason { 2147ba346b35SGerhard Heift struct btrfs_ioctl_search_args __user *uargs; 2148ba346b35SGerhard Heift struct btrfs_ioctl_search_key sk; 2149ac8e9819SChris Mason struct inode *inode; 2150ac8e9819SChris Mason int ret; 21519b6e817dSGerhard Heift size_t buf_size; 2152ac8e9819SChris Mason 2153ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 2154ac8e9819SChris Mason return -EPERM; 2155ac8e9819SChris Mason 2156ba346b35SGerhard Heift uargs = (struct btrfs_ioctl_search_args __user *)argp; 2157ac8e9819SChris Mason 2158ba346b35SGerhard Heift if (copy_from_user(&sk, &uargs->key, sizeof(sk))) 2159ba346b35SGerhard Heift return -EFAULT; 2160ba346b35SGerhard Heift 2161ba346b35SGerhard Heift buf_size = sizeof(uargs->buf); 21629b6e817dSGerhard Heift 2163496ad9aaSAl Viro inode = file_inode(file); 2164ba346b35SGerhard Heift ret = search_ioctl(inode, &sk, &buf_size, uargs->buf); 21658f5f6178SGerhard Heift 21668f5f6178SGerhard Heift /* 21678f5f6178SGerhard Heift * In the origin implementation an overflow is handled by returning a 21688f5f6178SGerhard Heift * search header with a len of zero, so reset ret. 21698f5f6178SGerhard Heift */ 21708f5f6178SGerhard Heift if (ret == -EOVERFLOW) 21718f5f6178SGerhard Heift ret = 0; 21728f5f6178SGerhard Heift 2173ba346b35SGerhard Heift if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk))) 2174ac8e9819SChris Mason ret = -EFAULT; 2175ac8e9819SChris Mason return ret; 2176ac8e9819SChris Mason } 2177ac8e9819SChris Mason 2178cc68a8a5SGerhard Heift static noinline int btrfs_ioctl_tree_search_v2(struct file *file, 2179cc68a8a5SGerhard Heift void __user *argp) 2180cc68a8a5SGerhard Heift { 2181cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 __user *uarg; 2182cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 args; 2183cc68a8a5SGerhard Heift struct inode *inode; 2184cc68a8a5SGerhard Heift int ret; 2185cc68a8a5SGerhard Heift size_t buf_size; 2186cc68a8a5SGerhard Heift const size_t buf_limit = 16 * 1024 * 1024; 2187cc68a8a5SGerhard Heift 2188cc68a8a5SGerhard Heift if (!capable(CAP_SYS_ADMIN)) 2189cc68a8a5SGerhard Heift return -EPERM; 2190cc68a8a5SGerhard Heift 2191cc68a8a5SGerhard Heift /* copy search header and buffer size */ 2192cc68a8a5SGerhard Heift uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp; 2193cc68a8a5SGerhard Heift if (copy_from_user(&args, uarg, sizeof(args))) 2194cc68a8a5SGerhard Heift return -EFAULT; 2195cc68a8a5SGerhard Heift 2196cc68a8a5SGerhard Heift buf_size = args.buf_size; 2197cc68a8a5SGerhard Heift 2198cc68a8a5SGerhard Heift if (buf_size < sizeof(struct btrfs_ioctl_search_header)) 2199cc68a8a5SGerhard Heift return -EOVERFLOW; 2200cc68a8a5SGerhard Heift 2201cc68a8a5SGerhard Heift /* limit result size to 16MB */ 2202cc68a8a5SGerhard Heift if (buf_size > buf_limit) 2203cc68a8a5SGerhard Heift buf_size = buf_limit; 2204cc68a8a5SGerhard Heift 2205cc68a8a5SGerhard Heift inode = file_inode(file); 2206cc68a8a5SGerhard Heift ret = search_ioctl(inode, &args.key, &buf_size, 2207cc68a8a5SGerhard Heift (char *)(&uarg->buf[0])); 2208cc68a8a5SGerhard Heift if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key))) 2209cc68a8a5SGerhard Heift ret = -EFAULT; 2210cc68a8a5SGerhard Heift else if (ret == -EOVERFLOW && 2211cc68a8a5SGerhard Heift copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size))) 2212cc68a8a5SGerhard Heift ret = -EFAULT; 2213cc68a8a5SGerhard Heift 2214cc68a8a5SGerhard Heift return ret; 2215cc68a8a5SGerhard Heift } 2216cc68a8a5SGerhard Heift 221798d377a0STARUISI Hiroaki /* 2218ac8e9819SChris Mason * Search INODE_REFs to identify path name of 'dirid' directory 2219ac8e9819SChris Mason * in a 'tree_id' tree. and sets path name to 'name'. 222098d377a0STARUISI Hiroaki */ 222198d377a0STARUISI Hiroaki static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, 222298d377a0STARUISI Hiroaki u64 tree_id, u64 dirid, char *name) 222398d377a0STARUISI Hiroaki { 222498d377a0STARUISI Hiroaki struct btrfs_root *root; 222598d377a0STARUISI Hiroaki struct btrfs_key key; 2226ac8e9819SChris Mason char *ptr; 222798d377a0STARUISI Hiroaki int ret = -1; 222898d377a0STARUISI Hiroaki int slot; 222998d377a0STARUISI Hiroaki int len; 223098d377a0STARUISI Hiroaki int total_len = 0; 223198d377a0STARUISI Hiroaki struct btrfs_inode_ref *iref; 223298d377a0STARUISI Hiroaki struct extent_buffer *l; 223398d377a0STARUISI Hiroaki struct btrfs_path *path; 223498d377a0STARUISI Hiroaki 223598d377a0STARUISI Hiroaki if (dirid == BTRFS_FIRST_FREE_OBJECTID) { 223698d377a0STARUISI Hiroaki name[0]='\0'; 223798d377a0STARUISI Hiroaki return 0; 223898d377a0STARUISI Hiroaki } 223998d377a0STARUISI Hiroaki 224098d377a0STARUISI Hiroaki path = btrfs_alloc_path(); 224198d377a0STARUISI Hiroaki if (!path) 224298d377a0STARUISI Hiroaki return -ENOMEM; 224398d377a0STARUISI Hiroaki 2244ac8e9819SChris Mason ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX]; 224598d377a0STARUISI Hiroaki 224698d377a0STARUISI Hiroaki key.objectid = tree_id; 224798d377a0STARUISI Hiroaki key.type = BTRFS_ROOT_ITEM_KEY; 224898d377a0STARUISI Hiroaki key.offset = (u64)-1; 224998d377a0STARUISI Hiroaki root = btrfs_read_fs_root_no_name(info, &key); 225098d377a0STARUISI Hiroaki if (IS_ERR(root)) { 2251efe120a0SFrank Holton printk(KERN_ERR "BTRFS: could not find root %llu\n", tree_id); 22528ad6fcabSChris Mason ret = -ENOENT; 22538ad6fcabSChris Mason goto out; 225498d377a0STARUISI Hiroaki } 225598d377a0STARUISI Hiroaki 225698d377a0STARUISI Hiroaki key.objectid = dirid; 225798d377a0STARUISI Hiroaki key.type = BTRFS_INODE_REF_KEY; 22588ad6fcabSChris Mason key.offset = (u64)-1; 225998d377a0STARUISI Hiroaki 226098d377a0STARUISI Hiroaki while (1) { 226198d377a0STARUISI Hiroaki ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 226298d377a0STARUISI Hiroaki if (ret < 0) 226398d377a0STARUISI Hiroaki goto out; 226418674c6cSFilipe David Borba Manana else if (ret > 0) { 226518674c6cSFilipe David Borba Manana ret = btrfs_previous_item(root, path, dirid, 226618674c6cSFilipe David Borba Manana BTRFS_INODE_REF_KEY); 226718674c6cSFilipe David Borba Manana if (ret < 0) 226818674c6cSFilipe David Borba Manana goto out; 226918674c6cSFilipe David Borba Manana else if (ret > 0) { 2270ac8e9819SChris Mason ret = -ENOENT; 227198d377a0STARUISI Hiroaki goto out; 2272ac8e9819SChris Mason } 227318674c6cSFilipe David Borba Manana } 227418674c6cSFilipe David Borba Manana 227518674c6cSFilipe David Borba Manana l = path->nodes[0]; 227618674c6cSFilipe David Borba Manana slot = path->slots[0]; 227718674c6cSFilipe David Borba Manana btrfs_item_key_to_cpu(l, &key, slot); 227898d377a0STARUISI Hiroaki 227998d377a0STARUISI Hiroaki iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); 228098d377a0STARUISI Hiroaki len = btrfs_inode_ref_name_len(l, iref); 228198d377a0STARUISI Hiroaki ptr -= len + 1; 228298d377a0STARUISI Hiroaki total_len += len + 1; 2283a696cf35SFilipe David Borba Manana if (ptr < name) { 2284a696cf35SFilipe David Borba Manana ret = -ENAMETOOLONG; 228598d377a0STARUISI Hiroaki goto out; 2286a696cf35SFilipe David Borba Manana } 228798d377a0STARUISI Hiroaki 228898d377a0STARUISI Hiroaki *(ptr + len) = '/'; 228998d377a0STARUISI Hiroaki read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len); 229098d377a0STARUISI Hiroaki 229198d377a0STARUISI Hiroaki if (key.offset == BTRFS_FIRST_FREE_OBJECTID) 229298d377a0STARUISI Hiroaki break; 229398d377a0STARUISI Hiroaki 2294b3b4aa74SDavid Sterba btrfs_release_path(path); 229598d377a0STARUISI Hiroaki key.objectid = key.offset; 22968ad6fcabSChris Mason key.offset = (u64)-1; 229798d377a0STARUISI Hiroaki dirid = key.objectid; 229898d377a0STARUISI Hiroaki } 229977906a50SLi Zefan memmove(name, ptr, total_len); 230098d377a0STARUISI Hiroaki name[total_len] = '\0'; 230198d377a0STARUISI Hiroaki ret = 0; 230298d377a0STARUISI Hiroaki out: 230398d377a0STARUISI Hiroaki btrfs_free_path(path); 2304ac8e9819SChris Mason return ret; 2305ac8e9819SChris Mason } 2306ac8e9819SChris Mason 2307ac8e9819SChris Mason static noinline int btrfs_ioctl_ino_lookup(struct file *file, 2308ac8e9819SChris Mason void __user *argp) 2309ac8e9819SChris Mason { 2310ac8e9819SChris Mason struct btrfs_ioctl_ino_lookup_args *args; 2311ac8e9819SChris Mason struct inode *inode; 2312ac8e9819SChris Mason int ret; 2313ac8e9819SChris Mason 2314ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 2315ac8e9819SChris Mason return -EPERM; 2316ac8e9819SChris Mason 23172354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 23182354d08fSJulia Lawall if (IS_ERR(args)) 23192354d08fSJulia Lawall return PTR_ERR(args); 2320c2b96929SDan Carpenter 2321496ad9aaSAl Viro inode = file_inode(file); 2322ac8e9819SChris Mason 23231b53ac4dSChris Mason if (args->treeid == 0) 23241b53ac4dSChris Mason args->treeid = BTRFS_I(inode)->root->root_key.objectid; 23251b53ac4dSChris Mason 2326ac8e9819SChris Mason ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, 2327ac8e9819SChris Mason args->treeid, args->objectid, 2328ac8e9819SChris Mason args->name); 2329ac8e9819SChris Mason 2330ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 2331ac8e9819SChris Mason ret = -EFAULT; 2332ac8e9819SChris Mason 2333ac8e9819SChris Mason kfree(args); 233498d377a0STARUISI Hiroaki return ret; 233598d377a0STARUISI Hiroaki } 233698d377a0STARUISI Hiroaki 233776dda93cSYan, Zheng static noinline int btrfs_ioctl_snap_destroy(struct file *file, 233876dda93cSYan, Zheng void __user *arg) 233976dda93cSYan, Zheng { 234054563d41SAl Viro struct dentry *parent = file->f_path.dentry; 234176dda93cSYan, Zheng struct dentry *dentry; 234276dda93cSYan, Zheng struct inode *dir = parent->d_inode; 234376dda93cSYan, Zheng struct inode *inode; 234476dda93cSYan, Zheng struct btrfs_root *root = BTRFS_I(dir)->root; 234576dda93cSYan, Zheng struct btrfs_root *dest = NULL; 234676dda93cSYan, Zheng struct btrfs_ioctl_vol_args *vol_args; 234776dda93cSYan, Zheng struct btrfs_trans_handle *trans; 2348c58aaad2SMiao Xie struct btrfs_block_rsv block_rsv; 2349521e0546SDavid Sterba u64 root_flags; 2350c58aaad2SMiao Xie u64 qgroup_reserved; 235176dda93cSYan, Zheng int namelen; 235276dda93cSYan, Zheng int ret; 235376dda93cSYan, Zheng int err = 0; 235476dda93cSYan, Zheng 235576dda93cSYan, Zheng vol_args = memdup_user(arg, sizeof(*vol_args)); 235676dda93cSYan, Zheng if (IS_ERR(vol_args)) 235776dda93cSYan, Zheng return PTR_ERR(vol_args); 235876dda93cSYan, Zheng 235976dda93cSYan, Zheng vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 236076dda93cSYan, Zheng namelen = strlen(vol_args->name); 236176dda93cSYan, Zheng if (strchr(vol_args->name, '/') || 236276dda93cSYan, Zheng strncmp(vol_args->name, "..", namelen) == 0) { 236376dda93cSYan, Zheng err = -EINVAL; 236476dda93cSYan, Zheng goto out; 236576dda93cSYan, Zheng } 236676dda93cSYan, Zheng 2367a561be71SAl Viro err = mnt_want_write_file(file); 236876dda93cSYan, Zheng if (err) 236976dda93cSYan, Zheng goto out; 237076dda93cSYan, Zheng 2371521e0546SDavid Sterba 23725c50c9b8SDavid Sterba err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT); 23735c50c9b8SDavid Sterba if (err == -EINTR) 2374e43f998eSDavid Sterba goto out_drop_write; 237576dda93cSYan, Zheng dentry = lookup_one_len(vol_args->name, parent, namelen); 237676dda93cSYan, Zheng if (IS_ERR(dentry)) { 237776dda93cSYan, Zheng err = PTR_ERR(dentry); 237876dda93cSYan, Zheng goto out_unlock_dir; 237976dda93cSYan, Zheng } 238076dda93cSYan, Zheng 238176dda93cSYan, Zheng if (!dentry->d_inode) { 238276dda93cSYan, Zheng err = -ENOENT; 238376dda93cSYan, Zheng goto out_dput; 238476dda93cSYan, Zheng } 238576dda93cSYan, Zheng 238676dda93cSYan, Zheng inode = dentry->d_inode; 23874260f7c7SSage Weil dest = BTRFS_I(inode)->root; 23884260f7c7SSage Weil if (!capable(CAP_SYS_ADMIN)) { 23894260f7c7SSage Weil /* 23904260f7c7SSage Weil * Regular user. Only allow this with a special mount 23914260f7c7SSage Weil * option, when the user has write+exec access to the 23924260f7c7SSage Weil * subvol root, and when rmdir(2) would have been 23934260f7c7SSage Weil * allowed. 23944260f7c7SSage Weil * 23954260f7c7SSage Weil * Note that this is _not_ check that the subvol is 23964260f7c7SSage Weil * empty or doesn't contain data that we wouldn't 23974260f7c7SSage Weil * otherwise be able to delete. 23984260f7c7SSage Weil * 23994260f7c7SSage Weil * Users who want to delete empty subvols should try 24004260f7c7SSage Weil * rmdir(2). 24014260f7c7SSage Weil */ 24024260f7c7SSage Weil err = -EPERM; 24034260f7c7SSage Weil if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED)) 24044260f7c7SSage Weil goto out_dput; 24054260f7c7SSage Weil 24064260f7c7SSage Weil /* 24074260f7c7SSage Weil * Do not allow deletion if the parent dir is the same 24084260f7c7SSage Weil * as the dir to be deleted. That means the ioctl 24094260f7c7SSage Weil * must be called on the dentry referencing the root 24104260f7c7SSage Weil * of the subvol, not a random directory contained 24114260f7c7SSage Weil * within it. 24124260f7c7SSage Weil */ 24134260f7c7SSage Weil err = -EINVAL; 24144260f7c7SSage Weil if (root == dest) 24154260f7c7SSage Weil goto out_dput; 24164260f7c7SSage Weil 24174260f7c7SSage Weil err = inode_permission(inode, MAY_WRITE | MAY_EXEC); 24184260f7c7SSage Weil if (err) 24194260f7c7SSage Weil goto out_dput; 24205c39da5bSMiao Xie } 24214260f7c7SSage Weil 24225c39da5bSMiao Xie /* check if subvolume may be deleted by a user */ 24234260f7c7SSage Weil err = btrfs_may_delete(dir, dentry, 1); 24244260f7c7SSage Weil if (err) 24254260f7c7SSage Weil goto out_dput; 24264260f7c7SSage Weil 242733345d01SLi Zefan if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { 242876dda93cSYan, Zheng err = -EINVAL; 242976dda93cSYan, Zheng goto out_dput; 243076dda93cSYan, Zheng } 243176dda93cSYan, Zheng 243276dda93cSYan, Zheng mutex_lock(&inode->i_mutex); 2433521e0546SDavid Sterba 2434521e0546SDavid Sterba /* 2435521e0546SDavid Sterba * Don't allow to delete a subvolume with send in progress. This is 2436521e0546SDavid Sterba * inside the i_mutex so the error handling that has to drop the bit 2437521e0546SDavid Sterba * again is not run concurrently. 2438521e0546SDavid Sterba */ 2439521e0546SDavid Sterba spin_lock(&dest->root_item_lock); 2440c55bfa67SFilipe Manana root_flags = btrfs_root_flags(&dest->root_item); 2441c55bfa67SFilipe Manana if (dest->send_in_progress == 0) { 2442c55bfa67SFilipe Manana btrfs_set_root_flags(&dest->root_item, 2443521e0546SDavid Sterba root_flags | BTRFS_ROOT_SUBVOL_DEAD); 2444521e0546SDavid Sterba spin_unlock(&dest->root_item_lock); 2445521e0546SDavid Sterba } else { 2446521e0546SDavid Sterba spin_unlock(&dest->root_item_lock); 2447521e0546SDavid Sterba btrfs_warn(root->fs_info, 2448521e0546SDavid Sterba "Attempt to delete subvolume %llu during send", 2449c55bfa67SFilipe Manana dest->root_key.objectid); 2450521e0546SDavid Sterba err = -EPERM; 2451521e0546SDavid Sterba goto out_dput; 2452521e0546SDavid Sterba } 2453521e0546SDavid Sterba 245476dda93cSYan, Zheng err = d_invalidate(dentry); 245576dda93cSYan, Zheng if (err) 245676dda93cSYan, Zheng goto out_unlock; 245776dda93cSYan, Zheng 245876dda93cSYan, Zheng down_write(&root->fs_info->subvol_sem); 245976dda93cSYan, Zheng 246076dda93cSYan, Zheng err = may_destroy_subvol(dest); 246176dda93cSYan, Zheng if (err) 246276dda93cSYan, Zheng goto out_up_write; 246376dda93cSYan, Zheng 2464c58aaad2SMiao Xie btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); 2465c58aaad2SMiao Xie /* 2466c58aaad2SMiao Xie * One for dir inode, two for dir entries, two for root 2467c58aaad2SMiao Xie * ref/backref. 2468c58aaad2SMiao Xie */ 2469c58aaad2SMiao Xie err = btrfs_subvolume_reserve_metadata(root, &block_rsv, 2470ee3441b4SJeff Mahoney 5, &qgroup_reserved, true); 2471c58aaad2SMiao Xie if (err) 2472c58aaad2SMiao Xie goto out_up_write; 2473c58aaad2SMiao Xie 2474a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 2475a22285a6SYan, Zheng if (IS_ERR(trans)) { 2476a22285a6SYan, Zheng err = PTR_ERR(trans); 2477c58aaad2SMiao Xie goto out_release; 2478a22285a6SYan, Zheng } 2479c58aaad2SMiao Xie trans->block_rsv = &block_rsv; 2480c58aaad2SMiao Xie trans->bytes_reserved = block_rsv.size; 2481a22285a6SYan, Zheng 248276dda93cSYan, Zheng ret = btrfs_unlink_subvol(trans, root, dir, 248376dda93cSYan, Zheng dest->root_key.objectid, 248476dda93cSYan, Zheng dentry->d_name.name, 248576dda93cSYan, Zheng dentry->d_name.len); 248679787eaaSJeff Mahoney if (ret) { 248779787eaaSJeff Mahoney err = ret; 248879787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, ret); 248979787eaaSJeff Mahoney goto out_end_trans; 249079787eaaSJeff Mahoney } 249176dda93cSYan, Zheng 249276dda93cSYan, Zheng btrfs_record_root_in_trans(trans, dest); 249376dda93cSYan, Zheng 249476dda93cSYan, Zheng memset(&dest->root_item.drop_progress, 0, 249576dda93cSYan, Zheng sizeof(dest->root_item.drop_progress)); 249676dda93cSYan, Zheng dest->root_item.drop_level = 0; 249776dda93cSYan, Zheng btrfs_set_root_refs(&dest->root_item, 0); 249876dda93cSYan, Zheng 249927cdeb70SMiao Xie if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) { 250076dda93cSYan, Zheng ret = btrfs_insert_orphan_item(trans, 250176dda93cSYan, Zheng root->fs_info->tree_root, 250276dda93cSYan, Zheng dest->root_key.objectid); 250379787eaaSJeff Mahoney if (ret) { 250479787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, ret); 250579787eaaSJeff Mahoney err = ret; 250679787eaaSJeff Mahoney goto out_end_trans; 2507d68fc57bSYan, Zheng } 250879787eaaSJeff Mahoney } 2509dd5f9615SStefan Behrens 2510dd5f9615SStefan Behrens ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root, 2511dd5f9615SStefan Behrens dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL, 2512dd5f9615SStefan Behrens dest->root_key.objectid); 2513dd5f9615SStefan Behrens if (ret && ret != -ENOENT) { 2514dd5f9615SStefan Behrens btrfs_abort_transaction(trans, root, ret); 2515dd5f9615SStefan Behrens err = ret; 2516dd5f9615SStefan Behrens goto out_end_trans; 2517dd5f9615SStefan Behrens } 2518dd5f9615SStefan Behrens if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) { 2519dd5f9615SStefan Behrens ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root, 2520dd5f9615SStefan Behrens dest->root_item.received_uuid, 2521dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 2522dd5f9615SStefan Behrens dest->root_key.objectid); 2523dd5f9615SStefan Behrens if (ret && ret != -ENOENT) { 2524dd5f9615SStefan Behrens btrfs_abort_transaction(trans, root, ret); 2525dd5f9615SStefan Behrens err = ret; 2526dd5f9615SStefan Behrens goto out_end_trans; 2527dd5f9615SStefan Behrens } 2528dd5f9615SStefan Behrens } 2529dd5f9615SStefan Behrens 253079787eaaSJeff Mahoney out_end_trans: 2531c58aaad2SMiao Xie trans->block_rsv = NULL; 2532c58aaad2SMiao Xie trans->bytes_reserved = 0; 2533531cb13fSSage Weil ret = btrfs_end_transaction(trans, root); 253479787eaaSJeff Mahoney if (ret && !err) 253579787eaaSJeff Mahoney err = ret; 253676dda93cSYan, Zheng inode->i_flags |= S_DEAD; 2537c58aaad2SMiao Xie out_release: 2538c58aaad2SMiao Xie btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved); 253976dda93cSYan, Zheng out_up_write: 254076dda93cSYan, Zheng up_write(&root->fs_info->subvol_sem); 254176dda93cSYan, Zheng out_unlock: 2542521e0546SDavid Sterba if (err) { 2543521e0546SDavid Sterba spin_lock(&dest->root_item_lock); 2544c55bfa67SFilipe Manana root_flags = btrfs_root_flags(&dest->root_item); 2545c55bfa67SFilipe Manana btrfs_set_root_flags(&dest->root_item, 2546521e0546SDavid Sterba root_flags & ~BTRFS_ROOT_SUBVOL_DEAD); 2547521e0546SDavid Sterba spin_unlock(&dest->root_item_lock); 2548521e0546SDavid Sterba } 254976dda93cSYan, Zheng mutex_unlock(&inode->i_mutex); 255076dda93cSYan, Zheng if (!err) { 2551efefb143SYan, Zheng shrink_dcache_sb(root->fs_info->sb); 255276dda93cSYan, Zheng btrfs_invalidate_inodes(dest); 255376dda93cSYan, Zheng d_delete(dentry); 255461155aa0SDavid Sterba ASSERT(dest->send_in_progress == 0); 2555fa6ac876SLiu Bo 2556fa6ac876SLiu Bo /* the last ref */ 2557fa6ac876SLiu Bo if (dest->cache_inode) { 2558fa6ac876SLiu Bo iput(dest->cache_inode); 2559fa6ac876SLiu Bo dest->cache_inode = NULL; 2560fa6ac876SLiu Bo } 256176dda93cSYan, Zheng } 256276dda93cSYan, Zheng out_dput: 256376dda93cSYan, Zheng dput(dentry); 256476dda93cSYan, Zheng out_unlock_dir: 256576dda93cSYan, Zheng mutex_unlock(&dir->i_mutex); 2566e43f998eSDavid Sterba out_drop_write: 25672a79f17eSAl Viro mnt_drop_write_file(file); 256876dda93cSYan, Zheng out: 256976dda93cSYan, Zheng kfree(vol_args); 257076dda93cSYan, Zheng return err; 257176dda93cSYan, Zheng } 257276dda93cSYan, Zheng 25731e701a32SChris Mason static int btrfs_ioctl_defrag(struct file *file, void __user *argp) 2574f46b5a66SChristoph Hellwig { 2575496ad9aaSAl Viro struct inode *inode = file_inode(file); 2576f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 25771e701a32SChris Mason struct btrfs_ioctl_defrag_range_args *range; 2578c146afadSYan Zheng int ret; 2579c146afadSYan Zheng 258025122d15SIlya Dryomov ret = mnt_want_write_file(file); 258125122d15SIlya Dryomov if (ret) 258225122d15SIlya Dryomov return ret; 2583b83cc969SLi Zefan 258425122d15SIlya Dryomov if (btrfs_root_readonly(root)) { 258525122d15SIlya Dryomov ret = -EROFS; 258625122d15SIlya Dryomov goto out; 25875ac00addSStefan Behrens } 2588f46b5a66SChristoph Hellwig 2589f46b5a66SChristoph Hellwig switch (inode->i_mode & S_IFMT) { 2590f46b5a66SChristoph Hellwig case S_IFDIR: 2591e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) { 2592e441d54dSChris Mason ret = -EPERM; 2593e441d54dSChris Mason goto out; 2594e441d54dSChris Mason } 2595de78b51aSEric Sandeen ret = btrfs_defrag_root(root); 25968929ecfaSYan, Zheng if (ret) 25978929ecfaSYan, Zheng goto out; 2598de78b51aSEric Sandeen ret = btrfs_defrag_root(root->fs_info->extent_root); 2599f46b5a66SChristoph Hellwig break; 2600f46b5a66SChristoph Hellwig case S_IFREG: 2601e441d54dSChris Mason if (!(file->f_mode & FMODE_WRITE)) { 2602e441d54dSChris Mason ret = -EINVAL; 2603e441d54dSChris Mason goto out; 2604e441d54dSChris Mason } 26051e701a32SChris Mason 26061e701a32SChris Mason range = kzalloc(sizeof(*range), GFP_KERNEL); 26071e701a32SChris Mason if (!range) { 26081e701a32SChris Mason ret = -ENOMEM; 26091e701a32SChris Mason goto out; 26101e701a32SChris Mason } 26111e701a32SChris Mason 26121e701a32SChris Mason if (argp) { 26131e701a32SChris Mason if (copy_from_user(range, argp, 26141e701a32SChris Mason sizeof(*range))) { 26151e701a32SChris Mason ret = -EFAULT; 26161e701a32SChris Mason kfree(range); 2617683be16eSDan Carpenter goto out; 26181e701a32SChris Mason } 26191e701a32SChris Mason /* compression requires us to start the IO */ 26201e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 26211e701a32SChris Mason range->flags |= BTRFS_DEFRAG_RANGE_START_IO; 26221e701a32SChris Mason range->extent_thresh = (u32)-1; 26231e701a32SChris Mason } 26241e701a32SChris Mason } else { 26251e701a32SChris Mason /* the rest are all set to zero by kzalloc */ 26261e701a32SChris Mason range->len = (u64)-1; 26271e701a32SChris Mason } 2628496ad9aaSAl Viro ret = btrfs_defrag_file(file_inode(file), file, 26294cb5300bSChris Mason range, 0, 0); 26304cb5300bSChris Mason if (ret > 0) 26314cb5300bSChris Mason ret = 0; 26321e701a32SChris Mason kfree(range); 2633f46b5a66SChristoph Hellwig break; 26348929ecfaSYan, Zheng default: 26358929ecfaSYan, Zheng ret = -EINVAL; 2636f46b5a66SChristoph Hellwig } 2637e441d54dSChris Mason out: 263825122d15SIlya Dryomov mnt_drop_write_file(file); 2639e441d54dSChris Mason return ret; 2640f46b5a66SChristoph Hellwig } 2641f46b5a66SChristoph Hellwig 2642b2950863SChristoph Hellwig static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg) 2643f46b5a66SChristoph Hellwig { 2644f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2645f46b5a66SChristoph Hellwig int ret; 2646f46b5a66SChristoph Hellwig 2647e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2648e441d54dSChris Mason return -EPERM; 2649e441d54dSChris Mason 26505ac00addSStefan Behrens if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running, 26515ac00addSStefan Behrens 1)) { 2652e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 2653c9e9f97bSIlya Dryomov } 2654c9e9f97bSIlya Dryomov 26555ac00addSStefan Behrens mutex_lock(&root->fs_info->volume_mutex); 2656dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2657c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 2658c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 2659c9e9f97bSIlya Dryomov goto out; 2660c9e9f97bSIlya Dryomov } 2661f46b5a66SChristoph Hellwig 26625516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 2663f46b5a66SChristoph Hellwig ret = btrfs_init_new_device(root, vol_args->name); 2664f46b5a66SChristoph Hellwig 2665f46b5a66SChristoph Hellwig kfree(vol_args); 2666c9e9f97bSIlya Dryomov out: 2667c9e9f97bSIlya Dryomov mutex_unlock(&root->fs_info->volume_mutex); 26685ac00addSStefan Behrens atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0); 2669f46b5a66SChristoph Hellwig return ret; 2670f46b5a66SChristoph Hellwig } 2671f46b5a66SChristoph Hellwig 2672da24927bSMiao Xie static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg) 2673f46b5a66SChristoph Hellwig { 2674496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 2675f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2676f46b5a66SChristoph Hellwig int ret; 2677f46b5a66SChristoph Hellwig 2678e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2679e441d54dSChris Mason return -EPERM; 2680e441d54dSChris Mason 2681da24927bSMiao Xie ret = mnt_want_write_file(file); 2682da24927bSMiao Xie if (ret) 2683da24927bSMiao Xie return ret; 2684c146afadSYan Zheng 2685dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2686c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 2687c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 2688c9e9f97bSIlya Dryomov goto out; 2689c9e9f97bSIlya Dryomov } 2690f46b5a66SChristoph Hellwig 26915516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 2692f46b5a66SChristoph Hellwig 2693183860f6SAnand Jain if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running, 2694183860f6SAnand Jain 1)) { 2695183860f6SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 2696183860f6SAnand Jain goto out; 2697183860f6SAnand Jain } 2698183860f6SAnand Jain 2699183860f6SAnand Jain mutex_lock(&root->fs_info->volume_mutex); 2700183860f6SAnand Jain ret = btrfs_rm_device(root, vol_args->name); 2701c9e9f97bSIlya Dryomov mutex_unlock(&root->fs_info->volume_mutex); 27025ac00addSStefan Behrens atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0); 2703183860f6SAnand Jain 2704183860f6SAnand Jain out: 2705183860f6SAnand Jain kfree(vol_args); 27064ac20c70SIlya Dryomov mnt_drop_write_file(file); 2707f46b5a66SChristoph Hellwig return ret; 2708f46b5a66SChristoph Hellwig } 2709f46b5a66SChristoph Hellwig 2710475f6387SJan Schmidt static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg) 2711475f6387SJan Schmidt { 2712027ed2f0SLi Zefan struct btrfs_ioctl_fs_info_args *fi_args; 2713475f6387SJan Schmidt struct btrfs_device *device; 2714475f6387SJan Schmidt struct btrfs_device *next; 2715475f6387SJan Schmidt struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; 2716027ed2f0SLi Zefan int ret = 0; 2717475f6387SJan Schmidt 2718027ed2f0SLi Zefan fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL); 2719027ed2f0SLi Zefan if (!fi_args) 2720027ed2f0SLi Zefan return -ENOMEM; 2721027ed2f0SLi Zefan 2722f7171750SFilipe David Borba Manana mutex_lock(&fs_devices->device_list_mutex); 2723027ed2f0SLi Zefan fi_args->num_devices = fs_devices->num_devices; 2724027ed2f0SLi Zefan memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid)); 2725475f6387SJan Schmidt 2726475f6387SJan Schmidt list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) { 2727027ed2f0SLi Zefan if (device->devid > fi_args->max_id) 2728027ed2f0SLi Zefan fi_args->max_id = device->devid; 2729475f6387SJan Schmidt } 2730475f6387SJan Schmidt mutex_unlock(&fs_devices->device_list_mutex); 2731475f6387SJan Schmidt 273280a773fbSDavid Sterba fi_args->nodesize = root->fs_info->super_copy->nodesize; 273380a773fbSDavid Sterba fi_args->sectorsize = root->fs_info->super_copy->sectorsize; 273480a773fbSDavid Sterba fi_args->clone_alignment = root->fs_info->super_copy->sectorsize; 273580a773fbSDavid Sterba 2736027ed2f0SLi Zefan if (copy_to_user(arg, fi_args, sizeof(*fi_args))) 2737027ed2f0SLi Zefan ret = -EFAULT; 2738475f6387SJan Schmidt 2739027ed2f0SLi Zefan kfree(fi_args); 2740027ed2f0SLi Zefan return ret; 2741475f6387SJan Schmidt } 2742475f6387SJan Schmidt 2743475f6387SJan Schmidt static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg) 2744475f6387SJan Schmidt { 2745475f6387SJan Schmidt struct btrfs_ioctl_dev_info_args *di_args; 2746475f6387SJan Schmidt struct btrfs_device *dev; 2747475f6387SJan Schmidt struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; 2748475f6387SJan Schmidt int ret = 0; 2749475f6387SJan Schmidt char *s_uuid = NULL; 2750475f6387SJan Schmidt 2751475f6387SJan Schmidt di_args = memdup_user(arg, sizeof(*di_args)); 2752475f6387SJan Schmidt if (IS_ERR(di_args)) 2753475f6387SJan Schmidt return PTR_ERR(di_args); 2754475f6387SJan Schmidt 2755dd5f9615SStefan Behrens if (!btrfs_is_empty_uuid(di_args->uuid)) 2756475f6387SJan Schmidt s_uuid = di_args->uuid; 2757475f6387SJan Schmidt 2758475f6387SJan Schmidt mutex_lock(&fs_devices->device_list_mutex); 2759aa1b8cd4SStefan Behrens dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL); 2760475f6387SJan Schmidt 2761475f6387SJan Schmidt if (!dev) { 2762475f6387SJan Schmidt ret = -ENODEV; 2763475f6387SJan Schmidt goto out; 2764475f6387SJan Schmidt } 2765475f6387SJan Schmidt 2766475f6387SJan Schmidt di_args->devid = dev->devid; 2767475f6387SJan Schmidt di_args->bytes_used = dev->bytes_used; 2768475f6387SJan Schmidt di_args->total_bytes = dev->total_bytes; 2769475f6387SJan Schmidt memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); 2770a27202fbSJim Meyering if (dev->name) { 2771606686eeSJosef Bacik struct rcu_string *name; 2772606686eeSJosef Bacik 2773606686eeSJosef Bacik rcu_read_lock(); 2774606686eeSJosef Bacik name = rcu_dereference(dev->name); 2775606686eeSJosef Bacik strncpy(di_args->path, name->str, sizeof(di_args->path)); 2776606686eeSJosef Bacik rcu_read_unlock(); 2777a27202fbSJim Meyering di_args->path[sizeof(di_args->path) - 1] = 0; 2778a27202fbSJim Meyering } else { 277999ba55adSStefan Behrens di_args->path[0] = '\0'; 2780a27202fbSJim Meyering } 2781475f6387SJan Schmidt 2782475f6387SJan Schmidt out: 278355793c0dSDavid Sterba mutex_unlock(&fs_devices->device_list_mutex); 2784475f6387SJan Schmidt if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) 2785475f6387SJan Schmidt ret = -EFAULT; 2786475f6387SJan Schmidt 2787475f6387SJan Schmidt kfree(di_args); 2788475f6387SJan Schmidt return ret; 2789475f6387SJan Schmidt } 2790475f6387SJan Schmidt 2791416161dbSMark Fasheh static struct page *extent_same_get_page(struct inode *inode, u64 off) 2792416161dbSMark Fasheh { 2793416161dbSMark Fasheh struct page *page; 2794416161dbSMark Fasheh pgoff_t index; 2795416161dbSMark Fasheh struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree; 2796416161dbSMark Fasheh 2797416161dbSMark Fasheh index = off >> PAGE_CACHE_SHIFT; 2798416161dbSMark Fasheh 2799416161dbSMark Fasheh page = grab_cache_page(inode->i_mapping, index); 2800416161dbSMark Fasheh if (!page) 2801416161dbSMark Fasheh return NULL; 2802416161dbSMark Fasheh 2803416161dbSMark Fasheh if (!PageUptodate(page)) { 2804416161dbSMark Fasheh if (extent_read_full_page_nolock(tree, page, btrfs_get_extent, 2805416161dbSMark Fasheh 0)) 2806416161dbSMark Fasheh return NULL; 2807416161dbSMark Fasheh lock_page(page); 2808416161dbSMark Fasheh if (!PageUptodate(page)) { 2809416161dbSMark Fasheh unlock_page(page); 2810416161dbSMark Fasheh page_cache_release(page); 2811416161dbSMark Fasheh return NULL; 2812416161dbSMark Fasheh } 2813416161dbSMark Fasheh } 2814416161dbSMark Fasheh unlock_page(page); 2815416161dbSMark Fasheh 2816416161dbSMark Fasheh return page; 2817416161dbSMark Fasheh } 2818416161dbSMark Fasheh 281977fe20dcSMark Fasheh static inline void lock_extent_range(struct inode *inode, u64 off, u64 len) 282077fe20dcSMark Fasheh { 282177fe20dcSMark Fasheh /* do any pending delalloc/csum calc on src, one way or 282277fe20dcSMark Fasheh another, and lock file content */ 282377fe20dcSMark Fasheh while (1) { 282477fe20dcSMark Fasheh struct btrfs_ordered_extent *ordered; 282577fe20dcSMark Fasheh lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); 282677fe20dcSMark Fasheh ordered = btrfs_lookup_first_ordered_extent(inode, 282777fe20dcSMark Fasheh off + len - 1); 2828ff5df9b8SFilipe Manana if ((!ordered || 2829ff5df9b8SFilipe Manana ordered->file_offset + ordered->len <= off || 2830ff5df9b8SFilipe Manana ordered->file_offset >= off + len) && 283177fe20dcSMark Fasheh !test_range_bit(&BTRFS_I(inode)->io_tree, off, 2832ff5df9b8SFilipe Manana off + len - 1, EXTENT_DELALLOC, 0, NULL)) { 2833ff5df9b8SFilipe Manana if (ordered) 2834ff5df9b8SFilipe Manana btrfs_put_ordered_extent(ordered); 283577fe20dcSMark Fasheh break; 2836ff5df9b8SFilipe Manana } 283777fe20dcSMark Fasheh unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); 283877fe20dcSMark Fasheh if (ordered) 283977fe20dcSMark Fasheh btrfs_put_ordered_extent(ordered); 284077fe20dcSMark Fasheh btrfs_wait_ordered_range(inode, off, len); 284177fe20dcSMark Fasheh } 284277fe20dcSMark Fasheh } 284377fe20dcSMark Fasheh 2844416161dbSMark Fasheh static void btrfs_double_unlock(struct inode *inode1, u64 loff1, 2845416161dbSMark Fasheh struct inode *inode2, u64 loff2, u64 len) 2846416161dbSMark Fasheh { 2847416161dbSMark Fasheh unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); 2848416161dbSMark Fasheh unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); 2849416161dbSMark Fasheh 2850416161dbSMark Fasheh mutex_unlock(&inode1->i_mutex); 2851416161dbSMark Fasheh mutex_unlock(&inode2->i_mutex); 2852416161dbSMark Fasheh } 2853416161dbSMark Fasheh 2854416161dbSMark Fasheh static void btrfs_double_lock(struct inode *inode1, u64 loff1, 2855416161dbSMark Fasheh struct inode *inode2, u64 loff2, u64 len) 2856416161dbSMark Fasheh { 2857416161dbSMark Fasheh if (inode1 < inode2) { 2858416161dbSMark Fasheh swap(inode1, inode2); 2859416161dbSMark Fasheh swap(loff1, loff2); 2860416161dbSMark Fasheh } 2861416161dbSMark Fasheh 2862416161dbSMark Fasheh mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT); 2863416161dbSMark Fasheh lock_extent_range(inode1, loff1, len); 2864416161dbSMark Fasheh if (inode1 != inode2) { 2865416161dbSMark Fasheh mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD); 2866416161dbSMark Fasheh lock_extent_range(inode2, loff2, len); 2867416161dbSMark Fasheh } 2868416161dbSMark Fasheh } 2869416161dbSMark Fasheh 2870416161dbSMark Fasheh static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst, 2871416161dbSMark Fasheh u64 dst_loff, u64 len) 2872416161dbSMark Fasheh { 2873416161dbSMark Fasheh int ret = 0; 2874416161dbSMark Fasheh struct page *src_page, *dst_page; 2875416161dbSMark Fasheh unsigned int cmp_len = PAGE_CACHE_SIZE; 2876416161dbSMark Fasheh void *addr, *dst_addr; 2877416161dbSMark Fasheh 2878416161dbSMark Fasheh while (len) { 2879416161dbSMark Fasheh if (len < PAGE_CACHE_SIZE) 2880416161dbSMark Fasheh cmp_len = len; 2881416161dbSMark Fasheh 2882416161dbSMark Fasheh src_page = extent_same_get_page(src, loff); 2883416161dbSMark Fasheh if (!src_page) 2884416161dbSMark Fasheh return -EINVAL; 2885416161dbSMark Fasheh dst_page = extent_same_get_page(dst, dst_loff); 2886416161dbSMark Fasheh if (!dst_page) { 2887416161dbSMark Fasheh page_cache_release(src_page); 2888416161dbSMark Fasheh return -EINVAL; 2889416161dbSMark Fasheh } 2890416161dbSMark Fasheh addr = kmap_atomic(src_page); 2891416161dbSMark Fasheh dst_addr = kmap_atomic(dst_page); 2892416161dbSMark Fasheh 2893416161dbSMark Fasheh flush_dcache_page(src_page); 2894416161dbSMark Fasheh flush_dcache_page(dst_page); 2895416161dbSMark Fasheh 2896416161dbSMark Fasheh if (memcmp(addr, dst_addr, cmp_len)) 2897416161dbSMark Fasheh ret = BTRFS_SAME_DATA_DIFFERS; 2898416161dbSMark Fasheh 2899416161dbSMark Fasheh kunmap_atomic(addr); 2900416161dbSMark Fasheh kunmap_atomic(dst_addr); 2901416161dbSMark Fasheh page_cache_release(src_page); 2902416161dbSMark Fasheh page_cache_release(dst_page); 2903416161dbSMark Fasheh 2904416161dbSMark Fasheh if (ret) 2905416161dbSMark Fasheh break; 2906416161dbSMark Fasheh 2907416161dbSMark Fasheh loff += cmp_len; 2908416161dbSMark Fasheh dst_loff += cmp_len; 2909416161dbSMark Fasheh len -= cmp_len; 2910416161dbSMark Fasheh } 2911416161dbSMark Fasheh 2912416161dbSMark Fasheh return ret; 2913416161dbSMark Fasheh } 2914416161dbSMark Fasheh 2915416161dbSMark Fasheh static int extent_same_check_offsets(struct inode *inode, u64 off, u64 len) 2916416161dbSMark Fasheh { 2917416161dbSMark Fasheh u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize; 2918416161dbSMark Fasheh 2919416161dbSMark Fasheh if (off + len > inode->i_size || off + len < off) 2920416161dbSMark Fasheh return -EINVAL; 2921416161dbSMark Fasheh /* Check that we are block aligned - btrfs_clone() requires this */ 2922416161dbSMark Fasheh if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs)) 2923416161dbSMark Fasheh return -EINVAL; 2924416161dbSMark Fasheh 2925416161dbSMark Fasheh return 0; 2926416161dbSMark Fasheh } 2927416161dbSMark Fasheh 2928416161dbSMark Fasheh static int btrfs_extent_same(struct inode *src, u64 loff, u64 len, 2929416161dbSMark Fasheh struct inode *dst, u64 dst_loff) 2930416161dbSMark Fasheh { 2931416161dbSMark Fasheh int ret; 2932416161dbSMark Fasheh 2933416161dbSMark Fasheh /* 2934416161dbSMark Fasheh * btrfs_clone() can't handle extents in the same file 2935416161dbSMark Fasheh * yet. Once that works, we can drop this check and replace it 2936416161dbSMark Fasheh * with a check for the same inode, but overlapping extents. 2937416161dbSMark Fasheh */ 2938416161dbSMark Fasheh if (src == dst) 2939416161dbSMark Fasheh return -EINVAL; 2940416161dbSMark Fasheh 2941416161dbSMark Fasheh btrfs_double_lock(src, loff, dst, dst_loff, len); 2942416161dbSMark Fasheh 2943416161dbSMark Fasheh ret = extent_same_check_offsets(src, loff, len); 2944416161dbSMark Fasheh if (ret) 2945416161dbSMark Fasheh goto out_unlock; 2946416161dbSMark Fasheh 2947416161dbSMark Fasheh ret = extent_same_check_offsets(dst, dst_loff, len); 2948416161dbSMark Fasheh if (ret) 2949416161dbSMark Fasheh goto out_unlock; 2950416161dbSMark Fasheh 2951416161dbSMark Fasheh /* don't make the dst file partly checksummed */ 2952416161dbSMark Fasheh if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != 2953416161dbSMark Fasheh (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) { 2954416161dbSMark Fasheh ret = -EINVAL; 2955416161dbSMark Fasheh goto out_unlock; 2956416161dbSMark Fasheh } 2957416161dbSMark Fasheh 2958416161dbSMark Fasheh ret = btrfs_cmp_data(src, loff, dst, dst_loff, len); 2959416161dbSMark Fasheh if (ret == 0) 2960416161dbSMark Fasheh ret = btrfs_clone(src, dst, loff, len, len, dst_loff); 2961416161dbSMark Fasheh 2962416161dbSMark Fasheh out_unlock: 2963416161dbSMark Fasheh btrfs_double_unlock(src, loff, dst, dst_loff, len); 2964416161dbSMark Fasheh 2965416161dbSMark Fasheh return ret; 2966416161dbSMark Fasheh } 2967416161dbSMark Fasheh 2968416161dbSMark Fasheh #define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024) 2969416161dbSMark Fasheh 2970416161dbSMark Fasheh static long btrfs_ioctl_file_extent_same(struct file *file, 29711c1c8747SAl Viro struct btrfs_ioctl_same_args __user *argp) 2972416161dbSMark Fasheh { 2973cbf8b8caSMark Fasheh struct btrfs_ioctl_same_args *same; 2974cbf8b8caSMark Fasheh struct btrfs_ioctl_same_extent_info *info; 29751c1c8747SAl Viro struct inode *src = file_inode(file); 2976416161dbSMark Fasheh u64 off; 2977416161dbSMark Fasheh u64 len; 2978416161dbSMark Fasheh int i; 2979416161dbSMark Fasheh int ret; 2980cbf8b8caSMark Fasheh unsigned long size; 2981416161dbSMark Fasheh u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize; 2982416161dbSMark Fasheh bool is_admin = capable(CAP_SYS_ADMIN); 29831c1c8747SAl Viro u16 count; 2984416161dbSMark Fasheh 2985416161dbSMark Fasheh if (!(file->f_mode & FMODE_READ)) 2986416161dbSMark Fasheh return -EINVAL; 2987416161dbSMark Fasheh 2988416161dbSMark Fasheh ret = mnt_want_write_file(file); 2989416161dbSMark Fasheh if (ret) 2990416161dbSMark Fasheh return ret; 2991416161dbSMark Fasheh 29921c1c8747SAl Viro if (get_user(count, &argp->dest_count)) { 2993416161dbSMark Fasheh ret = -EFAULT; 2994416161dbSMark Fasheh goto out; 2995416161dbSMark Fasheh } 2996416161dbSMark Fasheh 29971c1c8747SAl Viro size = offsetof(struct btrfs_ioctl_same_args __user, info[count]); 2998cbf8b8caSMark Fasheh 29991c1c8747SAl Viro same = memdup_user(argp, size); 3000cbf8b8caSMark Fasheh 3001229eed43SGeyslan G. Bem if (IS_ERR(same)) { 3002229eed43SGeyslan G. Bem ret = PTR_ERR(same); 3003cbf8b8caSMark Fasheh goto out; 3004cbf8b8caSMark Fasheh } 3005cbf8b8caSMark Fasheh 3006cbf8b8caSMark Fasheh off = same->logical_offset; 3007cbf8b8caSMark Fasheh len = same->length; 3008416161dbSMark Fasheh 3009416161dbSMark Fasheh /* 3010416161dbSMark Fasheh * Limit the total length we will dedupe for each operation. 3011416161dbSMark Fasheh * This is intended to bound the total time spent in this 3012416161dbSMark Fasheh * ioctl to something sane. 3013416161dbSMark Fasheh */ 3014416161dbSMark Fasheh if (len > BTRFS_MAX_DEDUPE_LEN) 3015416161dbSMark Fasheh len = BTRFS_MAX_DEDUPE_LEN; 3016416161dbSMark Fasheh 3017416161dbSMark Fasheh if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) { 3018416161dbSMark Fasheh /* 3019416161dbSMark Fasheh * Btrfs does not support blocksize < page_size. As a 3020416161dbSMark Fasheh * result, btrfs_cmp_data() won't correctly handle 3021416161dbSMark Fasheh * this situation without an update. 3022416161dbSMark Fasheh */ 3023416161dbSMark Fasheh ret = -EINVAL; 3024416161dbSMark Fasheh goto out; 3025416161dbSMark Fasheh } 3026416161dbSMark Fasheh 3027416161dbSMark Fasheh ret = -EISDIR; 3028416161dbSMark Fasheh if (S_ISDIR(src->i_mode)) 3029416161dbSMark Fasheh goto out; 3030416161dbSMark Fasheh 3031416161dbSMark Fasheh ret = -EACCES; 3032416161dbSMark Fasheh if (!S_ISREG(src->i_mode)) 3033416161dbSMark Fasheh goto out; 3034416161dbSMark Fasheh 3035cbf8b8caSMark Fasheh /* pre-format output fields to sane values */ 30361c1c8747SAl Viro for (i = 0; i < count; i++) { 3037cbf8b8caSMark Fasheh same->info[i].bytes_deduped = 0ULL; 3038cbf8b8caSMark Fasheh same->info[i].status = 0; 3039416161dbSMark Fasheh } 3040416161dbSMark Fasheh 30411c1c8747SAl Viro for (i = 0, info = same->info; i < count; i++, info++) { 30421c1c8747SAl Viro struct inode *dst; 30431c1c8747SAl Viro struct fd dst_file = fdget(info->fd); 30441c1c8747SAl Viro if (!dst_file.file) { 3045cbf8b8caSMark Fasheh info->status = -EBADF; 30461c1c8747SAl Viro continue; 3047416161dbSMark Fasheh } 30481c1c8747SAl Viro dst = file_inode(dst_file.file); 3049416161dbSMark Fasheh 30501c1c8747SAl Viro if (!(is_admin || (dst_file.file->f_mode & FMODE_WRITE))) { 3051cbf8b8caSMark Fasheh info->status = -EINVAL; 30521c1c8747SAl Viro } else if (file->f_path.mnt != dst_file.file->f_path.mnt) { 3053cbf8b8caSMark Fasheh info->status = -EXDEV; 30541c1c8747SAl Viro } else if (S_ISDIR(dst->i_mode)) { 3055cbf8b8caSMark Fasheh info->status = -EISDIR; 30561c1c8747SAl Viro } else if (!S_ISREG(dst->i_mode)) { 3057cbf8b8caSMark Fasheh info->status = -EACCES; 30581c1c8747SAl Viro } else { 3059cbf8b8caSMark Fasheh info->status = btrfs_extent_same(src, off, len, dst, 3060cbf8b8caSMark Fasheh info->logical_offset); 3061cbf8b8caSMark Fasheh if (info->status == 0) 3062cbf8b8caSMark Fasheh info->bytes_deduped += len; 30631c1c8747SAl Viro } 30641c1c8747SAl Viro fdput(dst_file); 3065cbf8b8caSMark Fasheh } 3066416161dbSMark Fasheh 3067cbf8b8caSMark Fasheh ret = copy_to_user(argp, same, size); 3068cbf8b8caSMark Fasheh if (ret) 3069416161dbSMark Fasheh ret = -EFAULT; 3070416161dbSMark Fasheh 3071416161dbSMark Fasheh out: 3072416161dbSMark Fasheh mnt_drop_write_file(file); 3073416161dbSMark Fasheh return ret; 3074416161dbSMark Fasheh } 3075416161dbSMark Fasheh 3076fcebe456SJosef Bacik /* Helper to check and see if this root currently has a ref on the given disk 3077fcebe456SJosef Bacik * bytenr. If it does then we need to update the quota for this root. This 3078fcebe456SJosef Bacik * doesn't do anything if quotas aren't enabled. 3079fcebe456SJosef Bacik */ 3080fcebe456SJosef Bacik static int check_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, 3081fcebe456SJosef Bacik u64 disko) 3082fcebe456SJosef Bacik { 3083fcebe456SJosef Bacik struct seq_list tree_mod_seq_elem = {}; 3084fcebe456SJosef Bacik struct ulist *roots; 3085fcebe456SJosef Bacik struct ulist_iterator uiter; 3086fcebe456SJosef Bacik struct ulist_node *root_node = NULL; 3087fcebe456SJosef Bacik int ret; 3088fcebe456SJosef Bacik 3089fcebe456SJosef Bacik if (!root->fs_info->quota_enabled) 3090fcebe456SJosef Bacik return 1; 3091fcebe456SJosef Bacik 3092fcebe456SJosef Bacik btrfs_get_tree_mod_seq(root->fs_info, &tree_mod_seq_elem); 3093fcebe456SJosef Bacik ret = btrfs_find_all_roots(trans, root->fs_info, disko, 3094fcebe456SJosef Bacik tree_mod_seq_elem.seq, &roots); 3095fcebe456SJosef Bacik if (ret < 0) 3096fcebe456SJosef Bacik goto out; 3097fcebe456SJosef Bacik ret = 0; 3098fcebe456SJosef Bacik ULIST_ITER_INIT(&uiter); 3099fcebe456SJosef Bacik while ((root_node = ulist_next(roots, &uiter))) { 3100fcebe456SJosef Bacik if (root_node->val == root->objectid) { 3101fcebe456SJosef Bacik ret = 1; 3102fcebe456SJosef Bacik break; 3103fcebe456SJosef Bacik } 3104fcebe456SJosef Bacik } 3105fcebe456SJosef Bacik ulist_free(roots); 3106fcebe456SJosef Bacik out: 3107fcebe456SJosef Bacik btrfs_put_tree_mod_seq(root->fs_info, &tree_mod_seq_elem); 3108fcebe456SJosef Bacik return ret; 3109fcebe456SJosef Bacik } 3110fcebe456SJosef Bacik 3111f82a9901SFilipe Manana static int clone_finish_inode_update(struct btrfs_trans_handle *trans, 3112f82a9901SFilipe Manana struct inode *inode, 3113f82a9901SFilipe Manana u64 endoff, 3114f82a9901SFilipe Manana const u64 destoff, 3115f82a9901SFilipe Manana const u64 olen) 3116f82a9901SFilipe Manana { 3117f82a9901SFilipe Manana struct btrfs_root *root = BTRFS_I(inode)->root; 3118f82a9901SFilipe Manana int ret; 3119f82a9901SFilipe Manana 3120f82a9901SFilipe Manana inode_inc_iversion(inode); 3121f82a9901SFilipe Manana inode->i_mtime = inode->i_ctime = CURRENT_TIME; 3122f82a9901SFilipe Manana /* 3123f82a9901SFilipe Manana * We round up to the block size at eof when determining which 3124f82a9901SFilipe Manana * extents to clone above, but shouldn't round up the file size. 3125f82a9901SFilipe Manana */ 3126f82a9901SFilipe Manana if (endoff > destoff + olen) 3127f82a9901SFilipe Manana endoff = destoff + olen; 3128f82a9901SFilipe Manana if (endoff > inode->i_size) 3129f82a9901SFilipe Manana btrfs_i_size_write(inode, endoff); 3130f82a9901SFilipe Manana 3131f82a9901SFilipe Manana ret = btrfs_update_inode(trans, root, inode); 3132f82a9901SFilipe Manana if (ret) { 3133f82a9901SFilipe Manana btrfs_abort_transaction(trans, root, ret); 3134f82a9901SFilipe Manana btrfs_end_transaction(trans, root); 3135f82a9901SFilipe Manana goto out; 3136f82a9901SFilipe Manana } 3137f82a9901SFilipe Manana ret = btrfs_end_transaction(trans, root); 3138f82a9901SFilipe Manana out: 3139f82a9901SFilipe Manana return ret; 3140f82a9901SFilipe Manana } 3141f82a9901SFilipe Manana 31427ffbb598SFilipe Manana static void clone_update_extent_map(struct inode *inode, 31437ffbb598SFilipe Manana const struct btrfs_trans_handle *trans, 31447ffbb598SFilipe Manana const struct btrfs_path *path, 31457ffbb598SFilipe Manana struct btrfs_file_extent_item *fi, 31467ffbb598SFilipe Manana const u64 hole_offset, 31477ffbb598SFilipe Manana const u64 hole_len) 31487ffbb598SFilipe Manana { 31497ffbb598SFilipe Manana struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 31507ffbb598SFilipe Manana struct extent_map *em; 31517ffbb598SFilipe Manana int ret; 31527ffbb598SFilipe Manana 31537ffbb598SFilipe Manana em = alloc_extent_map(); 31547ffbb598SFilipe Manana if (!em) { 31557ffbb598SFilipe Manana set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, 31567ffbb598SFilipe Manana &BTRFS_I(inode)->runtime_flags); 31577ffbb598SFilipe Manana return; 31587ffbb598SFilipe Manana } 31597ffbb598SFilipe Manana 31607ffbb598SFilipe Manana if (fi) { 31617ffbb598SFilipe Manana btrfs_extent_item_to_extent_map(inode, path, fi, false, em); 31627ffbb598SFilipe Manana em->generation = -1; 31637ffbb598SFilipe Manana if (btrfs_file_extent_type(path->nodes[0], fi) == 31647ffbb598SFilipe Manana BTRFS_FILE_EXTENT_INLINE) 31657ffbb598SFilipe Manana set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, 31667ffbb598SFilipe Manana &BTRFS_I(inode)->runtime_flags); 31677ffbb598SFilipe Manana } else { 31687ffbb598SFilipe Manana em->start = hole_offset; 31697ffbb598SFilipe Manana em->len = hole_len; 31707ffbb598SFilipe Manana em->ram_bytes = em->len; 31717ffbb598SFilipe Manana em->orig_start = hole_offset; 31727ffbb598SFilipe Manana em->block_start = EXTENT_MAP_HOLE; 31737ffbb598SFilipe Manana em->block_len = 0; 31747ffbb598SFilipe Manana em->orig_block_len = 0; 31757ffbb598SFilipe Manana em->compress_type = BTRFS_COMPRESS_NONE; 31767ffbb598SFilipe Manana em->generation = trans->transid; 31777ffbb598SFilipe Manana } 31787ffbb598SFilipe Manana 31797ffbb598SFilipe Manana while (1) { 31807ffbb598SFilipe Manana write_lock(&em_tree->lock); 31817ffbb598SFilipe Manana ret = add_extent_mapping(em_tree, em, 1); 31827ffbb598SFilipe Manana write_unlock(&em_tree->lock); 31837ffbb598SFilipe Manana if (ret != -EEXIST) { 31847ffbb598SFilipe Manana free_extent_map(em); 31857ffbb598SFilipe Manana break; 31867ffbb598SFilipe Manana } 31877ffbb598SFilipe Manana btrfs_drop_extent_cache(inode, em->start, 31887ffbb598SFilipe Manana em->start + em->len - 1, 0); 31897ffbb598SFilipe Manana } 31907ffbb598SFilipe Manana 31917ffbb598SFilipe Manana if (unlikely(ret)) 31927ffbb598SFilipe Manana set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, 31937ffbb598SFilipe Manana &BTRFS_I(inode)->runtime_flags); 31947ffbb598SFilipe Manana } 31957ffbb598SFilipe Manana 319632b7c687SMark Fasheh /** 319732b7c687SMark Fasheh * btrfs_clone() - clone a range from inode file to another 319832b7c687SMark Fasheh * 319932b7c687SMark Fasheh * @src: Inode to clone from 320032b7c687SMark Fasheh * @inode: Inode to clone to 320132b7c687SMark Fasheh * @off: Offset within source to start clone from 320232b7c687SMark Fasheh * @olen: Original length, passed by user, of range to clone 320332b7c687SMark Fasheh * @olen_aligned: Block-aligned value of olen, extent_same uses 320432b7c687SMark Fasheh * identical values here 320532b7c687SMark Fasheh * @destoff: Offset within @inode to start clone 320632b7c687SMark Fasheh */ 320732b7c687SMark Fasheh static int btrfs_clone(struct inode *src, struct inode *inode, 3208f82a9901SFilipe Manana const u64 off, const u64 olen, const u64 olen_aligned, 3209f82a9901SFilipe Manana const u64 destoff) 3210f46b5a66SChristoph Hellwig { 3211f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 321232b7c687SMark Fasheh struct btrfs_path *path = NULL; 3213f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 321432b7c687SMark Fasheh struct btrfs_trans_handle *trans; 321532b7c687SMark Fasheh char *buf = NULL; 3216ae01a0abSYan Zheng struct btrfs_key key; 3217f46b5a66SChristoph Hellwig u32 nritems; 3218f46b5a66SChristoph Hellwig int slot; 3219ae01a0abSYan Zheng int ret; 3220fcebe456SJosef Bacik int no_quota; 3221f82a9901SFilipe Manana const u64 len = olen_aligned; 3222fcebe456SJosef Bacik u64 last_disko = 0; 3223f82a9901SFilipe Manana u64 last_dest_end = destoff; 3224ae01a0abSYan Zheng 3225ae01a0abSYan Zheng ret = -ENOMEM; 3226ae01a0abSYan Zheng buf = vmalloc(btrfs_level_size(root, 0)); 3227ae01a0abSYan Zheng if (!buf) 322832b7c687SMark Fasheh return ret; 3229ae01a0abSYan Zheng 3230ae01a0abSYan Zheng path = btrfs_alloc_path(); 3231ae01a0abSYan Zheng if (!path) { 3232ae01a0abSYan Zheng vfree(buf); 323332b7c687SMark Fasheh return ret; 3234ae01a0abSYan Zheng } 323532b7c687SMark Fasheh 3236ae01a0abSYan Zheng path->reada = 2; 3237c5c9cd4dSSage Weil /* clone data */ 323833345d01SLi Zefan key.objectid = btrfs_ino(src); 3239ae01a0abSYan Zheng key.type = BTRFS_EXTENT_DATA_KEY; 32402c463823SFilipe Manana key.offset = off; 3241f46b5a66SChristoph Hellwig 3242f46b5a66SChristoph Hellwig while (1) { 3243f46b5a66SChristoph Hellwig /* 3244f46b5a66SChristoph Hellwig * note the key will change type as we walk through the 3245f46b5a66SChristoph Hellwig * tree. 3246f46b5a66SChristoph Hellwig */ 3247e4355f34SFilipe David Borba Manana path->leave_spinning = 1; 3248362a20c5SDavid Sterba ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path, 3249362a20c5SDavid Sterba 0, 0); 3250f46b5a66SChristoph Hellwig if (ret < 0) 3251f46b5a66SChristoph Hellwig goto out; 32522c463823SFilipe Manana /* 32532c463823SFilipe Manana * First search, if no extent item that starts at offset off was 32542c463823SFilipe Manana * found but the previous item is an extent item, it's possible 32552c463823SFilipe Manana * it might overlap our target range, therefore process it. 32562c463823SFilipe Manana */ 32572c463823SFilipe Manana if (key.offset == off && ret > 0 && path->slots[0] > 0) { 32582c463823SFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 32592c463823SFilipe Manana path->slots[0] - 1); 32602c463823SFilipe Manana if (key.type == BTRFS_EXTENT_DATA_KEY) 32612c463823SFilipe Manana path->slots[0]--; 32622c463823SFilipe Manana } 3263f46b5a66SChristoph Hellwig 3264ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3265e4355f34SFilipe David Borba Manana process_slot: 3266fcebe456SJosef Bacik no_quota = 1; 3267ae01a0abSYan Zheng if (path->slots[0] >= nritems) { 3268362a20c5SDavid Sterba ret = btrfs_next_leaf(BTRFS_I(src)->root, path); 3269f46b5a66SChristoph Hellwig if (ret < 0) 3270f46b5a66SChristoph Hellwig goto out; 3271f46b5a66SChristoph Hellwig if (ret > 0) 3272f46b5a66SChristoph Hellwig break; 3273ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3274f46b5a66SChristoph Hellwig } 3275f46b5a66SChristoph Hellwig leaf = path->nodes[0]; 3276f46b5a66SChristoph Hellwig slot = path->slots[0]; 3277f46b5a66SChristoph Hellwig 3278ae01a0abSYan Zheng btrfs_item_key_to_cpu(leaf, &key, slot); 3279d20f7043SChris Mason if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY || 328033345d01SLi Zefan key.objectid != btrfs_ino(src)) 3281f46b5a66SChristoph Hellwig break; 3282f46b5a66SChristoph Hellwig 3283c5c9cd4dSSage Weil if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) { 3284c5c9cd4dSSage Weil struct btrfs_file_extent_item *extent; 3285c5c9cd4dSSage Weil int type; 328631840ae1SZheng Yan u32 size; 328731840ae1SZheng Yan struct btrfs_key new_key; 3288c5c9cd4dSSage Weil u64 disko = 0, diskl = 0; 3289c5c9cd4dSSage Weil u64 datao = 0, datal = 0; 3290c5c9cd4dSSage Weil u8 comp; 3291f82a9901SFilipe Manana u64 drop_start; 329231840ae1SZheng Yan 3293c5c9cd4dSSage Weil extent = btrfs_item_ptr(leaf, slot, 3294c5c9cd4dSSage Weil struct btrfs_file_extent_item); 3295c5c9cd4dSSage Weil comp = btrfs_file_extent_compression(leaf, extent); 3296c5c9cd4dSSage Weil type = btrfs_file_extent_type(leaf, extent); 3297c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3298c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3299d397712bSChris Mason disko = btrfs_file_extent_disk_bytenr(leaf, 3300d397712bSChris Mason extent); 3301d397712bSChris Mason diskl = btrfs_file_extent_disk_num_bytes(leaf, 3302d397712bSChris Mason extent); 3303c5c9cd4dSSage Weil datao = btrfs_file_extent_offset(leaf, extent); 3304d397712bSChris Mason datal = btrfs_file_extent_num_bytes(leaf, 3305d397712bSChris Mason extent); 3306c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3307c5c9cd4dSSage Weil /* take upper bound, may be compressed */ 3308c5c9cd4dSSage Weil datal = btrfs_file_extent_ram_bytes(leaf, 3309c5c9cd4dSSage Weil extent); 3310c5c9cd4dSSage Weil } 331131840ae1SZheng Yan 33122c463823SFilipe Manana /* 33132c463823SFilipe Manana * The first search might have left us at an extent 33142c463823SFilipe Manana * item that ends before our target range's start, can 33152c463823SFilipe Manana * happen if we have holes and NO_HOLES feature enabled. 33162c463823SFilipe Manana */ 33172c463823SFilipe Manana if (key.offset + datal <= off) { 3318e4355f34SFilipe David Borba Manana path->slots[0]++; 3319e4355f34SFilipe David Borba Manana goto process_slot; 33202c463823SFilipe Manana } else if (key.offset >= off + len) { 33212c463823SFilipe Manana break; 3322e4355f34SFilipe David Borba Manana } 3323e4355f34SFilipe David Borba Manana 3324e4355f34SFilipe David Borba Manana size = btrfs_item_size_nr(leaf, slot); 3325e4355f34SFilipe David Borba Manana read_extent_buffer(leaf, buf, 3326e4355f34SFilipe David Borba Manana btrfs_item_ptr_offset(leaf, slot), 3327e4355f34SFilipe David Borba Manana size); 3328e4355f34SFilipe David Borba Manana 3329e4355f34SFilipe David Borba Manana btrfs_release_path(path); 3330e4355f34SFilipe David Borba Manana path->leave_spinning = 0; 3331c5c9cd4dSSage Weil 333231840ae1SZheng Yan memcpy(&new_key, &key, sizeof(new_key)); 333333345d01SLi Zefan new_key.objectid = btrfs_ino(inode); 33344d728ec7SLi Zefan if (off <= key.offset) 3335c5c9cd4dSSage Weil new_key.offset = key.offset + destoff - off; 33364d728ec7SLi Zefan else 33374d728ec7SLi Zefan new_key.offset = destoff; 3338c5c9cd4dSSage Weil 3339b6f3409bSSage Weil /* 3340f82a9901SFilipe Manana * Deal with a hole that doesn't have an extent item 3341f82a9901SFilipe Manana * that represents it (NO_HOLES feature enabled). 3342f82a9901SFilipe Manana * This hole is either in the middle of the cloning 3343f82a9901SFilipe Manana * range or at the beginning (fully overlaps it or 3344f82a9901SFilipe Manana * partially overlaps it). 3345f82a9901SFilipe Manana */ 3346f82a9901SFilipe Manana if (new_key.offset != last_dest_end) 3347f82a9901SFilipe Manana drop_start = last_dest_end; 3348f82a9901SFilipe Manana else 3349f82a9901SFilipe Manana drop_start = new_key.offset; 3350f82a9901SFilipe Manana 3351f82a9901SFilipe Manana /* 3352b6f3409bSSage Weil * 1 - adjusting old extent (we may have to split it) 3353b6f3409bSSage Weil * 1 - add new extent 3354b6f3409bSSage Weil * 1 - inode update 3355b6f3409bSSage Weil */ 3356b6f3409bSSage Weil trans = btrfs_start_transaction(root, 3); 3357a22285a6SYan, Zheng if (IS_ERR(trans)) { 3358a22285a6SYan, Zheng ret = PTR_ERR(trans); 3359a22285a6SYan, Zheng goto out; 3360a22285a6SYan, Zheng } 3361a22285a6SYan, Zheng 3362c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3363c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3364d72c0842SLi Zefan /* 3365d72c0842SLi Zefan * a | --- range to clone ---| b 3366d72c0842SLi Zefan * | ------------- extent ------------- | 3367d72c0842SLi Zefan */ 3368d72c0842SLi Zefan 336993915584SAntonio Ospite /* subtract range b */ 3370d72c0842SLi Zefan if (key.offset + datal > off + len) 3371d72c0842SLi Zefan datal = off + len - key.offset; 3372d72c0842SLi Zefan 337393915584SAntonio Ospite /* subtract range a */ 3374a22285a6SYan, Zheng if (off > key.offset) { 3375a22285a6SYan, Zheng datao += off - key.offset; 3376a22285a6SYan, Zheng datal -= off - key.offset; 3377a22285a6SYan, Zheng } 3378a22285a6SYan, Zheng 33795dc562c5SJosef Bacik ret = btrfs_drop_extents(trans, root, inode, 3380f82a9901SFilipe Manana drop_start, 3381a22285a6SYan, Zheng new_key.offset + datal, 33822671485dSJosef Bacik 1); 338379787eaaSJeff Mahoney if (ret) { 33843f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 338500fdf13aSLiu Bo btrfs_abort_transaction(trans, 338600fdf13aSLiu Bo root, ret); 338779787eaaSJeff Mahoney btrfs_end_transaction(trans, root); 338879787eaaSJeff Mahoney goto out; 338979787eaaSJeff Mahoney } 3390a22285a6SYan, Zheng 339131840ae1SZheng Yan ret = btrfs_insert_empty_item(trans, root, path, 339231840ae1SZheng Yan &new_key, size); 339379787eaaSJeff Mahoney if (ret) { 339479787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, 339579787eaaSJeff Mahoney ret); 339679787eaaSJeff Mahoney btrfs_end_transaction(trans, root); 339779787eaaSJeff Mahoney goto out; 339879787eaaSJeff Mahoney } 339931840ae1SZheng Yan 340031840ae1SZheng Yan leaf = path->nodes[0]; 340131840ae1SZheng Yan slot = path->slots[0]; 340231840ae1SZheng Yan write_extent_buffer(leaf, buf, 340331840ae1SZheng Yan btrfs_item_ptr_offset(leaf, slot), 340431840ae1SZheng Yan size); 3405ae01a0abSYan Zheng 3406f46b5a66SChristoph Hellwig extent = btrfs_item_ptr(leaf, slot, 3407f46b5a66SChristoph Hellwig struct btrfs_file_extent_item); 3408c5c9cd4dSSage Weil 3409c5c9cd4dSSage Weil /* disko == 0 means it's a hole */ 3410c5c9cd4dSSage Weil if (!disko) 3411c5c9cd4dSSage Weil datao = 0; 3412c5c9cd4dSSage Weil 3413c5c9cd4dSSage Weil btrfs_set_file_extent_offset(leaf, extent, 3414c5c9cd4dSSage Weil datao); 3415c5c9cd4dSSage Weil btrfs_set_file_extent_num_bytes(leaf, extent, 3416c5c9cd4dSSage Weil datal); 3417fcebe456SJosef Bacik 3418fcebe456SJosef Bacik /* 3419fcebe456SJosef Bacik * We need to look up the roots that point at 3420fcebe456SJosef Bacik * this bytenr and see if the new root does. If 3421fcebe456SJosef Bacik * it does not we need to make sure we update 3422fcebe456SJosef Bacik * quotas appropriately. 3423fcebe456SJosef Bacik */ 3424fcebe456SJosef Bacik if (disko && root != BTRFS_I(src)->root && 3425fcebe456SJosef Bacik disko != last_disko) { 3426fcebe456SJosef Bacik no_quota = check_ref(trans, root, 3427fcebe456SJosef Bacik disko); 3428fcebe456SJosef Bacik if (no_quota < 0) { 3429fcebe456SJosef Bacik btrfs_abort_transaction(trans, 3430fcebe456SJosef Bacik root, 3431fcebe456SJosef Bacik ret); 3432fcebe456SJosef Bacik btrfs_end_transaction(trans, 3433fcebe456SJosef Bacik root); 3434fcebe456SJosef Bacik ret = no_quota; 3435fcebe456SJosef Bacik goto out; 3436fcebe456SJosef Bacik } 3437fcebe456SJosef Bacik } 3438fcebe456SJosef Bacik 3439c5c9cd4dSSage Weil if (disko) { 3440c5c9cd4dSSage Weil inode_add_bytes(inode, datal); 3441ae01a0abSYan Zheng ret = btrfs_inc_extent_ref(trans, root, 34425d4f98a2SYan Zheng disko, diskl, 0, 3443f46b5a66SChristoph Hellwig root->root_key.objectid, 344433345d01SLi Zefan btrfs_ino(inode), 344566d7e7f0SArne Jansen new_key.offset - datao, 3446fcebe456SJosef Bacik no_quota); 344779787eaaSJeff Mahoney if (ret) { 344879787eaaSJeff Mahoney btrfs_abort_transaction(trans, 344979787eaaSJeff Mahoney root, 345079787eaaSJeff Mahoney ret); 345179787eaaSJeff Mahoney btrfs_end_transaction(trans, 345279787eaaSJeff Mahoney root); 345379787eaaSJeff Mahoney goto out; 345479787eaaSJeff Mahoney 345579787eaaSJeff Mahoney } 3456f46b5a66SChristoph Hellwig } 3457c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3458c5c9cd4dSSage Weil u64 skip = 0; 3459c5c9cd4dSSage Weil u64 trim = 0; 3460d3ecfcdfSLiu Bo u64 aligned_end = 0; 3461d3ecfcdfSLiu Bo 3462c5c9cd4dSSage Weil if (off > key.offset) { 3463c5c9cd4dSSage Weil skip = off - key.offset; 3464c5c9cd4dSSage Weil new_key.offset += skip; 346531840ae1SZheng Yan } 3466d397712bSChris Mason 3467c5c9cd4dSSage Weil if (key.offset + datal > off + len) 3468c5c9cd4dSSage Weil trim = key.offset + datal - (off + len); 3469d397712bSChris Mason 3470c5c9cd4dSSage Weil if (comp && (skip || trim)) { 3471c5c9cd4dSSage Weil ret = -EINVAL; 3472a22285a6SYan, Zheng btrfs_end_transaction(trans, root); 3473c5c9cd4dSSage Weil goto out; 347431840ae1SZheng Yan } 3475c5c9cd4dSSage Weil size -= skip + trim; 3476c5c9cd4dSSage Weil datal -= skip + trim; 3477a22285a6SYan, Zheng 3478d3ecfcdfSLiu Bo aligned_end = ALIGN(new_key.offset + datal, 3479d3ecfcdfSLiu Bo root->sectorsize); 34805dc562c5SJosef Bacik ret = btrfs_drop_extents(trans, root, inode, 3481f82a9901SFilipe Manana drop_start, 3482d3ecfcdfSLiu Bo aligned_end, 34832671485dSJosef Bacik 1); 348479787eaaSJeff Mahoney if (ret) { 34853f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 34863a29bc09SChris Mason btrfs_abort_transaction(trans, 34873a29bc09SChris Mason root, ret); 348879787eaaSJeff Mahoney btrfs_end_transaction(trans, root); 348979787eaaSJeff Mahoney goto out; 349079787eaaSJeff Mahoney } 3491a22285a6SYan, Zheng 3492c5c9cd4dSSage Weil ret = btrfs_insert_empty_item(trans, root, path, 3493c5c9cd4dSSage Weil &new_key, size); 349479787eaaSJeff Mahoney if (ret) { 349579787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, 349679787eaaSJeff Mahoney ret); 349779787eaaSJeff Mahoney btrfs_end_transaction(trans, root); 349879787eaaSJeff Mahoney goto out; 349979787eaaSJeff Mahoney } 3500c5c9cd4dSSage Weil 3501c5c9cd4dSSage Weil if (skip) { 3502d397712bSChris Mason u32 start = 3503d397712bSChris Mason btrfs_file_extent_calc_inline_size(0); 3504c5c9cd4dSSage Weil memmove(buf+start, buf+start+skip, 3505c5c9cd4dSSage Weil datal); 3506c5c9cd4dSSage Weil } 3507c5c9cd4dSSage Weil 3508c5c9cd4dSSage Weil leaf = path->nodes[0]; 3509c5c9cd4dSSage Weil slot = path->slots[0]; 3510c5c9cd4dSSage Weil write_extent_buffer(leaf, buf, 3511c5c9cd4dSSage Weil btrfs_item_ptr_offset(leaf, slot), 3512c5c9cd4dSSage Weil size); 3513c5c9cd4dSSage Weil inode_add_bytes(inode, datal); 35147ffbb598SFilipe Manana extent = btrfs_item_ptr(leaf, slot, 35157ffbb598SFilipe Manana struct btrfs_file_extent_item); 3516c5c9cd4dSSage Weil } 3517c5c9cd4dSSage Weil 35187ffbb598SFilipe Manana /* If we have an implicit hole (NO_HOLES feature). */ 35197ffbb598SFilipe Manana if (drop_start < new_key.offset) 35207ffbb598SFilipe Manana clone_update_extent_map(inode, trans, 35217ffbb598SFilipe Manana path, NULL, drop_start, 35227ffbb598SFilipe Manana new_key.offset - drop_start); 35237ffbb598SFilipe Manana 35247ffbb598SFilipe Manana clone_update_extent_map(inode, trans, path, 35257ffbb598SFilipe Manana extent, 0, 0); 35267ffbb598SFilipe Manana 3527c5c9cd4dSSage Weil btrfs_mark_buffer_dirty(leaf); 3528b3b4aa74SDavid Sterba btrfs_release_path(path); 3529c5c9cd4dSSage Weil 3530f82a9901SFilipe Manana last_dest_end = new_key.offset + datal; 3531f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, 3532f82a9901SFilipe Manana last_dest_end, 3533f82a9901SFilipe Manana destoff, olen); 3534f82a9901SFilipe Manana if (ret) 353579787eaaSJeff Mahoney goto out; 35362c463823SFilipe Manana if (new_key.offset + datal >= destoff + len) 35372c463823SFilipe Manana break; 3538a22285a6SYan, Zheng } 3539b3b4aa74SDavid Sterba btrfs_release_path(path); 3540ae01a0abSYan Zheng key.offset++; 3541ae01a0abSYan Zheng } 3542f46b5a66SChristoph Hellwig ret = 0; 354332b7c687SMark Fasheh 3544f82a9901SFilipe Manana if (last_dest_end < destoff + len) { 3545f82a9901SFilipe Manana /* 3546f82a9901SFilipe Manana * We have an implicit hole (NO_HOLES feature is enabled) that 3547f82a9901SFilipe Manana * fully or partially overlaps our cloning range at its end. 3548f82a9901SFilipe Manana */ 3549f82a9901SFilipe Manana btrfs_release_path(path); 3550f82a9901SFilipe Manana 3551f82a9901SFilipe Manana /* 3552f82a9901SFilipe Manana * 1 - remove extent(s) 3553f82a9901SFilipe Manana * 1 - inode update 3554f82a9901SFilipe Manana */ 3555f82a9901SFilipe Manana trans = btrfs_start_transaction(root, 2); 3556f82a9901SFilipe Manana if (IS_ERR(trans)) { 3557f82a9901SFilipe Manana ret = PTR_ERR(trans); 3558f82a9901SFilipe Manana goto out; 3559f82a9901SFilipe Manana } 3560f82a9901SFilipe Manana ret = btrfs_drop_extents(trans, root, inode, 3561f82a9901SFilipe Manana last_dest_end, destoff + len, 1); 3562f82a9901SFilipe Manana if (ret) { 3563f82a9901SFilipe Manana if (ret != -EOPNOTSUPP) 3564f82a9901SFilipe Manana btrfs_abort_transaction(trans, root, ret); 3565f82a9901SFilipe Manana btrfs_end_transaction(trans, root); 3566f82a9901SFilipe Manana goto out; 3567f82a9901SFilipe Manana } 3568f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, destoff + len, 3569f82a9901SFilipe Manana destoff, olen); 35707ffbb598SFilipe Manana if (ret) 35717ffbb598SFilipe Manana goto out; 35727ffbb598SFilipe Manana clone_update_extent_map(inode, trans, path, NULL, last_dest_end, 35737ffbb598SFilipe Manana destoff + len - last_dest_end); 3574f82a9901SFilipe Manana } 3575f82a9901SFilipe Manana 3576f46b5a66SChristoph Hellwig out: 357732b7c687SMark Fasheh btrfs_free_path(path); 357832b7c687SMark Fasheh vfree(buf); 357932b7c687SMark Fasheh return ret; 358032b7c687SMark Fasheh } 358132b7c687SMark Fasheh 358232b7c687SMark Fasheh static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, 358332b7c687SMark Fasheh u64 off, u64 olen, u64 destoff) 358432b7c687SMark Fasheh { 358554563d41SAl Viro struct inode *inode = file_inode(file); 358632b7c687SMark Fasheh struct btrfs_root *root = BTRFS_I(inode)->root; 358732b7c687SMark Fasheh struct fd src_file; 358832b7c687SMark Fasheh struct inode *src; 358932b7c687SMark Fasheh int ret; 359032b7c687SMark Fasheh u64 len = olen; 359132b7c687SMark Fasheh u64 bs = root->fs_info->sb->s_blocksize; 359232b7c687SMark Fasheh int same_inode = 0; 359332b7c687SMark Fasheh 359432b7c687SMark Fasheh /* 359532b7c687SMark Fasheh * TODO: 359632b7c687SMark Fasheh * - split compressed inline extents. annoying: we need to 359732b7c687SMark Fasheh * decompress into destination's address_space (the file offset 359832b7c687SMark Fasheh * may change, so source mapping won't do), then recompress (or 359932b7c687SMark Fasheh * otherwise reinsert) a subrange. 360000fdf13aSLiu Bo * 360100fdf13aSLiu Bo * - split destination inode's inline extents. The inline extents can 360200fdf13aSLiu Bo * be either compressed or non-compressed. 360332b7c687SMark Fasheh */ 360432b7c687SMark Fasheh 360532b7c687SMark Fasheh /* the destination must be opened for writing */ 360632b7c687SMark Fasheh if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND)) 360732b7c687SMark Fasheh return -EINVAL; 360832b7c687SMark Fasheh 360932b7c687SMark Fasheh if (btrfs_root_readonly(root)) 361032b7c687SMark Fasheh return -EROFS; 361132b7c687SMark Fasheh 361232b7c687SMark Fasheh ret = mnt_want_write_file(file); 361332b7c687SMark Fasheh if (ret) 361432b7c687SMark Fasheh return ret; 361532b7c687SMark Fasheh 361632b7c687SMark Fasheh src_file = fdget(srcfd); 361732b7c687SMark Fasheh if (!src_file.file) { 361832b7c687SMark Fasheh ret = -EBADF; 361932b7c687SMark Fasheh goto out_drop_write; 362032b7c687SMark Fasheh } 362132b7c687SMark Fasheh 362232b7c687SMark Fasheh ret = -EXDEV; 362332b7c687SMark Fasheh if (src_file.file->f_path.mnt != file->f_path.mnt) 362432b7c687SMark Fasheh goto out_fput; 362532b7c687SMark Fasheh 362632b7c687SMark Fasheh src = file_inode(src_file.file); 362732b7c687SMark Fasheh 362832b7c687SMark Fasheh ret = -EINVAL; 362932b7c687SMark Fasheh if (src == inode) 363032b7c687SMark Fasheh same_inode = 1; 363132b7c687SMark Fasheh 363232b7c687SMark Fasheh /* the src must be open for reading */ 363332b7c687SMark Fasheh if (!(src_file.file->f_mode & FMODE_READ)) 363432b7c687SMark Fasheh goto out_fput; 363532b7c687SMark Fasheh 363632b7c687SMark Fasheh /* don't make the dst file partly checksummed */ 363732b7c687SMark Fasheh if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != 363832b7c687SMark Fasheh (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) 363932b7c687SMark Fasheh goto out_fput; 364032b7c687SMark Fasheh 364132b7c687SMark Fasheh ret = -EISDIR; 364232b7c687SMark Fasheh if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode)) 364332b7c687SMark Fasheh goto out_fput; 364432b7c687SMark Fasheh 364532b7c687SMark Fasheh ret = -EXDEV; 364632b7c687SMark Fasheh if (src->i_sb != inode->i_sb) 364732b7c687SMark Fasheh goto out_fput; 364832b7c687SMark Fasheh 364932b7c687SMark Fasheh if (!same_inode) { 365032b7c687SMark Fasheh if (inode < src) { 365132b7c687SMark Fasheh mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT); 365232b7c687SMark Fasheh mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD); 365332b7c687SMark Fasheh } else { 365432b7c687SMark Fasheh mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT); 365532b7c687SMark Fasheh mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD); 365632b7c687SMark Fasheh } 365732b7c687SMark Fasheh } else { 365832b7c687SMark Fasheh mutex_lock(&src->i_mutex); 365932b7c687SMark Fasheh } 366032b7c687SMark Fasheh 366132b7c687SMark Fasheh /* determine range to clone */ 366232b7c687SMark Fasheh ret = -EINVAL; 366332b7c687SMark Fasheh if (off + len > src->i_size || off + len < off) 366432b7c687SMark Fasheh goto out_unlock; 366532b7c687SMark Fasheh if (len == 0) 366632b7c687SMark Fasheh olen = len = src->i_size - off; 366732b7c687SMark Fasheh /* if we extend to eof, continue to block boundary */ 366832b7c687SMark Fasheh if (off + len == src->i_size) 366932b7c687SMark Fasheh len = ALIGN(src->i_size, bs) - off; 367032b7c687SMark Fasheh 367132b7c687SMark Fasheh /* verify the end result is block aligned */ 367232b7c687SMark Fasheh if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) || 367332b7c687SMark Fasheh !IS_ALIGNED(destoff, bs)) 367432b7c687SMark Fasheh goto out_unlock; 367532b7c687SMark Fasheh 367632b7c687SMark Fasheh /* verify if ranges are overlapped within the same file */ 367732b7c687SMark Fasheh if (same_inode) { 367832b7c687SMark Fasheh if (destoff + len > off && destoff < off + len) 367932b7c687SMark Fasheh goto out_unlock; 368032b7c687SMark Fasheh } 368132b7c687SMark Fasheh 368232b7c687SMark Fasheh if (destoff > inode->i_size) { 368332b7c687SMark Fasheh ret = btrfs_cont_expand(inode, inode->i_size, destoff); 368432b7c687SMark Fasheh if (ret) 368532b7c687SMark Fasheh goto out_unlock; 368632b7c687SMark Fasheh } 368732b7c687SMark Fasheh 3688c125b8bfSFilipe Manana /* 3689c125b8bfSFilipe Manana * Lock the target range too. Right after we replace the file extent 3690c125b8bfSFilipe Manana * items in the fs tree (which now point to the cloned data), we might 3691c125b8bfSFilipe Manana * have a worker replace them with extent items relative to a write 3692c125b8bfSFilipe Manana * operation that was issued before this clone operation (i.e. confront 3693c125b8bfSFilipe Manana * with inode.c:btrfs_finish_ordered_io). 3694c125b8bfSFilipe Manana */ 3695c125b8bfSFilipe Manana if (same_inode) { 3696c125b8bfSFilipe Manana u64 lock_start = min_t(u64, off, destoff); 3697c125b8bfSFilipe Manana u64 lock_len = max_t(u64, off, destoff) + len - lock_start; 369832b7c687SMark Fasheh 3699c125b8bfSFilipe Manana lock_extent_range(src, lock_start, lock_len); 3700c125b8bfSFilipe Manana } else { 370132b7c687SMark Fasheh lock_extent_range(src, off, len); 3702c125b8bfSFilipe Manana lock_extent_range(inode, destoff, len); 3703c125b8bfSFilipe Manana } 370432b7c687SMark Fasheh 370532b7c687SMark Fasheh ret = btrfs_clone(src, inode, off, olen, len, destoff); 370632b7c687SMark Fasheh 3707c125b8bfSFilipe Manana if (same_inode) { 3708c125b8bfSFilipe Manana u64 lock_start = min_t(u64, off, destoff); 3709c125b8bfSFilipe Manana u64 lock_end = max_t(u64, off, destoff) + len - 1; 3710c125b8bfSFilipe Manana 3711c125b8bfSFilipe Manana unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end); 3712c125b8bfSFilipe Manana } else { 3713aa42ffd9SLiu Bo unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1); 3714c125b8bfSFilipe Manana unlock_extent(&BTRFS_I(inode)->io_tree, destoff, 3715c125b8bfSFilipe Manana destoff + len - 1); 3716c125b8bfSFilipe Manana } 3717c125b8bfSFilipe Manana /* 3718c125b8bfSFilipe Manana * Truncate page cache pages so that future reads will see the cloned 3719c125b8bfSFilipe Manana * data immediately and not the previous data. 3720c125b8bfSFilipe Manana */ 3721c125b8bfSFilipe Manana truncate_inode_pages_range(&inode->i_data, destoff, 3722c125b8bfSFilipe Manana PAGE_CACHE_ALIGN(destoff + len) - 1); 3723f46b5a66SChristoph Hellwig out_unlock: 3724c57c2b3eSFilipe David Borba Manana if (!same_inode) { 3725c57c2b3eSFilipe David Borba Manana if (inode < src) { 3726f46b5a66SChristoph Hellwig mutex_unlock(&src->i_mutex); 3727f46b5a66SChristoph Hellwig mutex_unlock(&inode->i_mutex); 3728c57c2b3eSFilipe David Borba Manana } else { 3729c57c2b3eSFilipe David Borba Manana mutex_unlock(&inode->i_mutex); 3730c57c2b3eSFilipe David Borba Manana mutex_unlock(&src->i_mutex); 3731c57c2b3eSFilipe David Borba Manana } 3732c57c2b3eSFilipe David Borba Manana } else { 3733c57c2b3eSFilipe David Borba Manana mutex_unlock(&src->i_mutex); 3734c57c2b3eSFilipe David Borba Manana } 3735f46b5a66SChristoph Hellwig out_fput: 37362903ff01SAl Viro fdput(src_file); 3737ab67b7c1SYan Zheng out_drop_write: 37382a79f17eSAl Viro mnt_drop_write_file(file); 3739f46b5a66SChristoph Hellwig return ret; 3740f46b5a66SChristoph Hellwig } 3741f46b5a66SChristoph Hellwig 37427a865e8aSChristoph Hellwig static long btrfs_ioctl_clone_range(struct file *file, void __user *argp) 3743c5c9cd4dSSage Weil { 3744c5c9cd4dSSage Weil struct btrfs_ioctl_clone_range_args args; 3745c5c9cd4dSSage Weil 37467a865e8aSChristoph Hellwig if (copy_from_user(&args, argp, sizeof(args))) 3747c5c9cd4dSSage Weil return -EFAULT; 3748c5c9cd4dSSage Weil return btrfs_ioctl_clone(file, args.src_fd, args.src_offset, 3749c5c9cd4dSSage Weil args.src_length, args.dest_offset); 3750c5c9cd4dSSage Weil } 3751c5c9cd4dSSage Weil 3752f46b5a66SChristoph Hellwig /* 3753f46b5a66SChristoph Hellwig * there are many ways the trans_start and trans_end ioctls can lead 3754f46b5a66SChristoph Hellwig * to deadlocks. They should only be used by applications that 3755f46b5a66SChristoph Hellwig * basically own the machine, and have a very in depth understanding 3756f46b5a66SChristoph Hellwig * of all the possible deadlocks and enospc problems. 3757f46b5a66SChristoph Hellwig */ 3758b2950863SChristoph Hellwig static long btrfs_ioctl_trans_start(struct file *file) 3759f46b5a66SChristoph Hellwig { 3760496ad9aaSAl Viro struct inode *inode = file_inode(file); 3761f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 3762f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 37631ab86aedSSage Weil int ret; 3764f46b5a66SChristoph Hellwig 37651ab86aedSSage Weil ret = -EPERM; 3766df5b5520SChristoph Hellwig if (!capable(CAP_SYS_ADMIN)) 3767f46b5a66SChristoph Hellwig goto out; 37681ab86aedSSage Weil 37691ab86aedSSage Weil ret = -EINPROGRESS; 37701ab86aedSSage Weil if (file->private_data) 37711ab86aedSSage Weil goto out; 37729ca9ee09SSage Weil 3773b83cc969SLi Zefan ret = -EROFS; 3774b83cc969SLi Zefan if (btrfs_root_readonly(root)) 3775b83cc969SLi Zefan goto out; 3776b83cc969SLi Zefan 3777a561be71SAl Viro ret = mnt_want_write_file(file); 3778c146afadSYan Zheng if (ret) 3779c146afadSYan Zheng goto out; 3780c146afadSYan Zheng 3781a4abeea4SJosef Bacik atomic_inc(&root->fs_info->open_ioctl_trans); 37829ca9ee09SSage Weil 3783f46b5a66SChristoph Hellwig ret = -ENOMEM; 37847a7eaa40SJosef Bacik trans = btrfs_start_ioctl_transaction(root); 3785abd30bb0STsutomu Itoh if (IS_ERR(trans)) 37861ab86aedSSage Weil goto out_drop; 37871ab86aedSSage Weil 37881ab86aedSSage Weil file->private_data = trans; 37891ab86aedSSage Weil return 0; 37901ab86aedSSage Weil 37911ab86aedSSage Weil out_drop: 3792a4abeea4SJosef Bacik atomic_dec(&root->fs_info->open_ioctl_trans); 37932a79f17eSAl Viro mnt_drop_write_file(file); 3794f46b5a66SChristoph Hellwig out: 3795f46b5a66SChristoph Hellwig return ret; 3796f46b5a66SChristoph Hellwig } 3797f46b5a66SChristoph Hellwig 37986ef5ed0dSJosef Bacik static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) 37996ef5ed0dSJosef Bacik { 3800496ad9aaSAl Viro struct inode *inode = file_inode(file); 38016ef5ed0dSJosef Bacik struct btrfs_root *root = BTRFS_I(inode)->root; 38026ef5ed0dSJosef Bacik struct btrfs_root *new_root; 38036ef5ed0dSJosef Bacik struct btrfs_dir_item *di; 38046ef5ed0dSJosef Bacik struct btrfs_trans_handle *trans; 38056ef5ed0dSJosef Bacik struct btrfs_path *path; 38066ef5ed0dSJosef Bacik struct btrfs_key location; 38076ef5ed0dSJosef Bacik struct btrfs_disk_key disk_key; 38086ef5ed0dSJosef Bacik u64 objectid = 0; 38096ef5ed0dSJosef Bacik u64 dir_id; 38103c04ce01SMiao Xie int ret; 38116ef5ed0dSJosef Bacik 38126ef5ed0dSJosef Bacik if (!capable(CAP_SYS_ADMIN)) 38136ef5ed0dSJosef Bacik return -EPERM; 38146ef5ed0dSJosef Bacik 38153c04ce01SMiao Xie ret = mnt_want_write_file(file); 38163c04ce01SMiao Xie if (ret) 38173c04ce01SMiao Xie return ret; 38183c04ce01SMiao Xie 38193c04ce01SMiao Xie if (copy_from_user(&objectid, argp, sizeof(objectid))) { 38203c04ce01SMiao Xie ret = -EFAULT; 38213c04ce01SMiao Xie goto out; 38223c04ce01SMiao Xie } 38236ef5ed0dSJosef Bacik 38246ef5ed0dSJosef Bacik if (!objectid) 38251cecf579Schandan objectid = BTRFS_FS_TREE_OBJECTID; 38266ef5ed0dSJosef Bacik 38276ef5ed0dSJosef Bacik location.objectid = objectid; 38286ef5ed0dSJosef Bacik location.type = BTRFS_ROOT_ITEM_KEY; 38296ef5ed0dSJosef Bacik location.offset = (u64)-1; 38306ef5ed0dSJosef Bacik 38316ef5ed0dSJosef Bacik new_root = btrfs_read_fs_root_no_name(root->fs_info, &location); 38323c04ce01SMiao Xie if (IS_ERR(new_root)) { 38333c04ce01SMiao Xie ret = PTR_ERR(new_root); 38343c04ce01SMiao Xie goto out; 38353c04ce01SMiao Xie } 38366ef5ed0dSJosef Bacik 38376ef5ed0dSJosef Bacik path = btrfs_alloc_path(); 38383c04ce01SMiao Xie if (!path) { 38393c04ce01SMiao Xie ret = -ENOMEM; 38403c04ce01SMiao Xie goto out; 38413c04ce01SMiao Xie } 38426ef5ed0dSJosef Bacik path->leave_spinning = 1; 38436ef5ed0dSJosef Bacik 38446ef5ed0dSJosef Bacik trans = btrfs_start_transaction(root, 1); 384598d5dc13STsutomu Itoh if (IS_ERR(trans)) { 38466ef5ed0dSJosef Bacik btrfs_free_path(path); 38473c04ce01SMiao Xie ret = PTR_ERR(trans); 38483c04ce01SMiao Xie goto out; 38496ef5ed0dSJosef Bacik } 38506ef5ed0dSJosef Bacik 38516c41761fSDavid Sterba dir_id = btrfs_super_root_dir(root->fs_info->super_copy); 38526ef5ed0dSJosef Bacik di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path, 38536ef5ed0dSJosef Bacik dir_id, "default", 7, 1); 3854cf1e99a4SDan Carpenter if (IS_ERR_OR_NULL(di)) { 38556ef5ed0dSJosef Bacik btrfs_free_path(path); 38566ef5ed0dSJosef Bacik btrfs_end_transaction(trans, root); 3857efe120a0SFrank Holton btrfs_err(new_root->fs_info, "Umm, you don't have the default dir" 3858efe120a0SFrank Holton "item, this isn't going to work"); 38593c04ce01SMiao Xie ret = -ENOENT; 38603c04ce01SMiao Xie goto out; 38616ef5ed0dSJosef Bacik } 38626ef5ed0dSJosef Bacik 38636ef5ed0dSJosef Bacik btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); 38646ef5ed0dSJosef Bacik btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); 38656ef5ed0dSJosef Bacik btrfs_mark_buffer_dirty(path->nodes[0]); 38666ef5ed0dSJosef Bacik btrfs_free_path(path); 38676ef5ed0dSJosef Bacik 38682b0ce2c2SMitch Harder btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL); 38696ef5ed0dSJosef Bacik btrfs_end_transaction(trans, root); 38703c04ce01SMiao Xie out: 38713c04ce01SMiao Xie mnt_drop_write_file(file); 38723c04ce01SMiao Xie return ret; 38736ef5ed0dSJosef Bacik } 38746ef5ed0dSJosef Bacik 38755af3e8ccSStefan Behrens void btrfs_get_block_group_info(struct list_head *groups_list, 3876bf5fc093SJosef Bacik struct btrfs_ioctl_space_info *space) 3877bf5fc093SJosef Bacik { 3878bf5fc093SJosef Bacik struct btrfs_block_group_cache *block_group; 3879bf5fc093SJosef Bacik 3880bf5fc093SJosef Bacik space->total_bytes = 0; 3881bf5fc093SJosef Bacik space->used_bytes = 0; 3882bf5fc093SJosef Bacik space->flags = 0; 3883bf5fc093SJosef Bacik list_for_each_entry(block_group, groups_list, list) { 3884bf5fc093SJosef Bacik space->flags = block_group->flags; 3885bf5fc093SJosef Bacik space->total_bytes += block_group->key.offset; 3886bf5fc093SJosef Bacik space->used_bytes += 3887bf5fc093SJosef Bacik btrfs_block_group_used(&block_group->item); 3888bf5fc093SJosef Bacik } 3889bf5fc093SJosef Bacik } 3890bf5fc093SJosef Bacik 389148a3b636SEric Sandeen static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg) 38921406e432SJosef Bacik { 38931406e432SJosef Bacik struct btrfs_ioctl_space_args space_args; 38941406e432SJosef Bacik struct btrfs_ioctl_space_info space; 38951406e432SJosef Bacik struct btrfs_ioctl_space_info *dest; 38967fde62bfSChris Mason struct btrfs_ioctl_space_info *dest_orig; 389713f2696fSDaniel J Blueman struct btrfs_ioctl_space_info __user *user_dest; 38981406e432SJosef Bacik struct btrfs_space_info *info; 3899bf5fc093SJosef Bacik u64 types[] = {BTRFS_BLOCK_GROUP_DATA, 3900bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_SYSTEM, 3901bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_METADATA, 3902bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA}; 3903bf5fc093SJosef Bacik int num_types = 4; 39047fde62bfSChris Mason int alloc_size; 39051406e432SJosef Bacik int ret = 0; 390651788b1bSDan Rosenberg u64 slot_count = 0; 3907bf5fc093SJosef Bacik int i, c; 39081406e432SJosef Bacik 39091406e432SJosef Bacik if (copy_from_user(&space_args, 39101406e432SJosef Bacik (struct btrfs_ioctl_space_args __user *)arg, 39111406e432SJosef Bacik sizeof(space_args))) 39121406e432SJosef Bacik return -EFAULT; 39131406e432SJosef Bacik 3914bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 3915bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 3916bf5fc093SJosef Bacik 3917bf5fc093SJosef Bacik info = NULL; 39187fde62bfSChris Mason rcu_read_lock(); 3919bf5fc093SJosef Bacik list_for_each_entry_rcu(tmp, &root->fs_info->space_info, 3920bf5fc093SJosef Bacik list) { 3921bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 3922bf5fc093SJosef Bacik info = tmp; 3923bf5fc093SJosef Bacik break; 3924bf5fc093SJosef Bacik } 3925bf5fc093SJosef Bacik } 39267fde62bfSChris Mason rcu_read_unlock(); 39271406e432SJosef Bacik 3928bf5fc093SJosef Bacik if (!info) 3929bf5fc093SJosef Bacik continue; 3930bf5fc093SJosef Bacik 3931bf5fc093SJosef Bacik down_read(&info->groups_sem); 3932bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 3933bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) 3934bf5fc093SJosef Bacik slot_count++; 3935bf5fc093SJosef Bacik } 3936bf5fc093SJosef Bacik up_read(&info->groups_sem); 3937bf5fc093SJosef Bacik } 3938bf5fc093SJosef Bacik 393936523e95SDavid Sterba /* 394036523e95SDavid Sterba * Global block reserve, exported as a space_info 394136523e95SDavid Sterba */ 394236523e95SDavid Sterba slot_count++; 394336523e95SDavid Sterba 39447fde62bfSChris Mason /* space_slots == 0 means they are asking for a count */ 39457fde62bfSChris Mason if (space_args.space_slots == 0) { 39467fde62bfSChris Mason space_args.total_spaces = slot_count; 39477fde62bfSChris Mason goto out; 39487fde62bfSChris Mason } 3949bf5fc093SJosef Bacik 395051788b1bSDan Rosenberg slot_count = min_t(u64, space_args.space_slots, slot_count); 3951bf5fc093SJosef Bacik 39527fde62bfSChris Mason alloc_size = sizeof(*dest) * slot_count; 3953bf5fc093SJosef Bacik 39547fde62bfSChris Mason /* we generally have at most 6 or so space infos, one for each raid 39557fde62bfSChris Mason * level. So, a whole page should be more than enough for everyone 39567fde62bfSChris Mason */ 39577fde62bfSChris Mason if (alloc_size > PAGE_CACHE_SIZE) 39587fde62bfSChris Mason return -ENOMEM; 39597fde62bfSChris Mason 39607fde62bfSChris Mason space_args.total_spaces = 0; 39617fde62bfSChris Mason dest = kmalloc(alloc_size, GFP_NOFS); 39627fde62bfSChris Mason if (!dest) 39637fde62bfSChris Mason return -ENOMEM; 39647fde62bfSChris Mason dest_orig = dest; 39657fde62bfSChris Mason 39667fde62bfSChris Mason /* now we have a buffer to copy into */ 3967bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 3968bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 3969bf5fc093SJosef Bacik 397051788b1bSDan Rosenberg if (!slot_count) 397151788b1bSDan Rosenberg break; 397251788b1bSDan Rosenberg 3973bf5fc093SJosef Bacik info = NULL; 39741406e432SJosef Bacik rcu_read_lock(); 3975bf5fc093SJosef Bacik list_for_each_entry_rcu(tmp, &root->fs_info->space_info, 3976bf5fc093SJosef Bacik list) { 3977bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 3978bf5fc093SJosef Bacik info = tmp; 39797fde62bfSChris Mason break; 3980bf5fc093SJosef Bacik } 3981bf5fc093SJosef Bacik } 3982bf5fc093SJosef Bacik rcu_read_unlock(); 39837fde62bfSChris Mason 3984bf5fc093SJosef Bacik if (!info) 3985bf5fc093SJosef Bacik continue; 3986bf5fc093SJosef Bacik down_read(&info->groups_sem); 3987bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 3988bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) { 39895af3e8ccSStefan Behrens btrfs_get_block_group_info( 39905af3e8ccSStefan Behrens &info->block_groups[c], &space); 39917fde62bfSChris Mason memcpy(dest, &space, sizeof(space)); 39921406e432SJosef Bacik dest++; 39931406e432SJosef Bacik space_args.total_spaces++; 399451788b1bSDan Rosenberg slot_count--; 39951406e432SJosef Bacik } 399651788b1bSDan Rosenberg if (!slot_count) 399751788b1bSDan Rosenberg break; 3998bf5fc093SJosef Bacik } 3999bf5fc093SJosef Bacik up_read(&info->groups_sem); 4000bf5fc093SJosef Bacik } 40011406e432SJosef Bacik 400236523e95SDavid Sterba /* 400336523e95SDavid Sterba * Add global block reserve 400436523e95SDavid Sterba */ 400536523e95SDavid Sterba if (slot_count) { 400636523e95SDavid Sterba struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv; 400736523e95SDavid Sterba 400836523e95SDavid Sterba spin_lock(&block_rsv->lock); 400936523e95SDavid Sterba space.total_bytes = block_rsv->size; 401036523e95SDavid Sterba space.used_bytes = block_rsv->size - block_rsv->reserved; 401136523e95SDavid Sterba spin_unlock(&block_rsv->lock); 401236523e95SDavid Sterba space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV; 401336523e95SDavid Sterba memcpy(dest, &space, sizeof(space)); 401436523e95SDavid Sterba space_args.total_spaces++; 401536523e95SDavid Sterba } 401636523e95SDavid Sterba 40172eec6c81SDaniel J Blueman user_dest = (struct btrfs_ioctl_space_info __user *) 40187fde62bfSChris Mason (arg + sizeof(struct btrfs_ioctl_space_args)); 40197fde62bfSChris Mason 40207fde62bfSChris Mason if (copy_to_user(user_dest, dest_orig, alloc_size)) 40217fde62bfSChris Mason ret = -EFAULT; 40227fde62bfSChris Mason 40237fde62bfSChris Mason kfree(dest_orig); 40247fde62bfSChris Mason out: 40257fde62bfSChris Mason if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) 40261406e432SJosef Bacik ret = -EFAULT; 40271406e432SJosef Bacik 40281406e432SJosef Bacik return ret; 40291406e432SJosef Bacik } 40301406e432SJosef Bacik 4031f46b5a66SChristoph Hellwig /* 4032f46b5a66SChristoph Hellwig * there are many ways the trans_start and trans_end ioctls can lead 4033f46b5a66SChristoph Hellwig * to deadlocks. They should only be used by applications that 4034f46b5a66SChristoph Hellwig * basically own the machine, and have a very in depth understanding 4035f46b5a66SChristoph Hellwig * of all the possible deadlocks and enospc problems. 4036f46b5a66SChristoph Hellwig */ 4037f46b5a66SChristoph Hellwig long btrfs_ioctl_trans_end(struct file *file) 4038f46b5a66SChristoph Hellwig { 4039496ad9aaSAl Viro struct inode *inode = file_inode(file); 4040f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 4041f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 4042f46b5a66SChristoph Hellwig 4043f46b5a66SChristoph Hellwig trans = file->private_data; 40441ab86aedSSage Weil if (!trans) 40451ab86aedSSage Weil return -EINVAL; 4046b214107eSChristoph Hellwig file->private_data = NULL; 40479ca9ee09SSage Weil 40481ab86aedSSage Weil btrfs_end_transaction(trans, root); 40491ab86aedSSage Weil 4050a4abeea4SJosef Bacik atomic_dec(&root->fs_info->open_ioctl_trans); 40519ca9ee09SSage Weil 40522a79f17eSAl Viro mnt_drop_write_file(file); 40531ab86aedSSage Weil return 0; 4054f46b5a66SChristoph Hellwig } 4055f46b5a66SChristoph Hellwig 40569a8c28beSMiao Xie static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, 40579a8c28beSMiao Xie void __user *argp) 405846204592SSage Weil { 405946204592SSage Weil struct btrfs_trans_handle *trans; 406046204592SSage Weil u64 transid; 4061db5b493aSTsutomu Itoh int ret; 406246204592SSage Weil 4063d4edf39bSMiao Xie trans = btrfs_attach_transaction_barrier(root); 4064ff7c1d33SMiao Xie if (IS_ERR(trans)) { 4065ff7c1d33SMiao Xie if (PTR_ERR(trans) != -ENOENT) 406698d5dc13STsutomu Itoh return PTR_ERR(trans); 4067ff7c1d33SMiao Xie 4068ff7c1d33SMiao Xie /* No running transaction, don't bother */ 4069ff7c1d33SMiao Xie transid = root->fs_info->last_trans_committed; 4070ff7c1d33SMiao Xie goto out; 4071ff7c1d33SMiao Xie } 407246204592SSage Weil transid = trans->transid; 4073db5b493aSTsutomu Itoh ret = btrfs_commit_transaction_async(trans, root, 0); 40748b2b2d3cSTsutomu Itoh if (ret) { 40758b2b2d3cSTsutomu Itoh btrfs_end_transaction(trans, root); 4076db5b493aSTsutomu Itoh return ret; 40778b2b2d3cSTsutomu Itoh } 4078ff7c1d33SMiao Xie out: 407946204592SSage Weil if (argp) 408046204592SSage Weil if (copy_to_user(argp, &transid, sizeof(transid))) 408146204592SSage Weil return -EFAULT; 408246204592SSage Weil return 0; 408346204592SSage Weil } 408446204592SSage Weil 40859a8c28beSMiao Xie static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root, 40869a8c28beSMiao Xie void __user *argp) 408746204592SSage Weil { 408846204592SSage Weil u64 transid; 408946204592SSage Weil 409046204592SSage Weil if (argp) { 409146204592SSage Weil if (copy_from_user(&transid, argp, sizeof(transid))) 409246204592SSage Weil return -EFAULT; 409346204592SSage Weil } else { 409446204592SSage Weil transid = 0; /* current trans */ 409546204592SSage Weil } 409646204592SSage Weil return btrfs_wait_for_commit(root, transid); 409746204592SSage Weil } 409846204592SSage Weil 4099b8e95489SMiao Xie static long btrfs_ioctl_scrub(struct file *file, void __user *arg) 4100475f6387SJan Schmidt { 4101496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 4102475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4103b8e95489SMiao Xie int ret; 4104475f6387SJan Schmidt 4105475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4106475f6387SJan Schmidt return -EPERM; 4107475f6387SJan Schmidt 4108475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4109475f6387SJan Schmidt if (IS_ERR(sa)) 4110475f6387SJan Schmidt return PTR_ERR(sa); 4111475f6387SJan Schmidt 4112b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) { 4113b8e95489SMiao Xie ret = mnt_want_write_file(file); 4114b8e95489SMiao Xie if (ret) 4115b8e95489SMiao Xie goto out; 4116b8e95489SMiao Xie } 4117b8e95489SMiao Xie 4118aa1b8cd4SStefan Behrens ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end, 411963a212abSStefan Behrens &sa->progress, sa->flags & BTRFS_SCRUB_READONLY, 412063a212abSStefan Behrens 0); 4121475f6387SJan Schmidt 4122475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 4123475f6387SJan Schmidt ret = -EFAULT; 4124475f6387SJan Schmidt 4125b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) 4126b8e95489SMiao Xie mnt_drop_write_file(file); 4127b8e95489SMiao Xie out: 4128475f6387SJan Schmidt kfree(sa); 4129475f6387SJan Schmidt return ret; 4130475f6387SJan Schmidt } 4131475f6387SJan Schmidt 4132475f6387SJan Schmidt static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg) 4133475f6387SJan Schmidt { 4134475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4135475f6387SJan Schmidt return -EPERM; 4136475f6387SJan Schmidt 4137aa1b8cd4SStefan Behrens return btrfs_scrub_cancel(root->fs_info); 4138475f6387SJan Schmidt } 4139475f6387SJan Schmidt 4140475f6387SJan Schmidt static long btrfs_ioctl_scrub_progress(struct btrfs_root *root, 4141475f6387SJan Schmidt void __user *arg) 4142475f6387SJan Schmidt { 4143475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4144475f6387SJan Schmidt int ret; 4145475f6387SJan Schmidt 4146475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4147475f6387SJan Schmidt return -EPERM; 4148475f6387SJan Schmidt 4149475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4150475f6387SJan Schmidt if (IS_ERR(sa)) 4151475f6387SJan Schmidt return PTR_ERR(sa); 4152475f6387SJan Schmidt 4153475f6387SJan Schmidt ret = btrfs_scrub_progress(root, sa->devid, &sa->progress); 4154475f6387SJan Schmidt 4155475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 4156475f6387SJan Schmidt ret = -EFAULT; 4157475f6387SJan Schmidt 4158475f6387SJan Schmidt kfree(sa); 4159475f6387SJan Schmidt return ret; 4160475f6387SJan Schmidt } 4161475f6387SJan Schmidt 4162c11d2c23SStefan Behrens static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root, 4163b27f7c0cSDavid Sterba void __user *arg) 4164c11d2c23SStefan Behrens { 4165c11d2c23SStefan Behrens struct btrfs_ioctl_get_dev_stats *sa; 4166c11d2c23SStefan Behrens int ret; 4167c11d2c23SStefan Behrens 4168c11d2c23SStefan Behrens sa = memdup_user(arg, sizeof(*sa)); 4169c11d2c23SStefan Behrens if (IS_ERR(sa)) 4170c11d2c23SStefan Behrens return PTR_ERR(sa); 4171c11d2c23SStefan Behrens 4172b27f7c0cSDavid Sterba if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) { 4173b27f7c0cSDavid Sterba kfree(sa); 4174b27f7c0cSDavid Sterba return -EPERM; 4175b27f7c0cSDavid Sterba } 4176b27f7c0cSDavid Sterba 4177b27f7c0cSDavid Sterba ret = btrfs_get_dev_stats(root, sa); 4178c11d2c23SStefan Behrens 4179c11d2c23SStefan Behrens if (copy_to_user(arg, sa, sizeof(*sa))) 4180c11d2c23SStefan Behrens ret = -EFAULT; 4181c11d2c23SStefan Behrens 4182c11d2c23SStefan Behrens kfree(sa); 4183c11d2c23SStefan Behrens return ret; 4184c11d2c23SStefan Behrens } 4185c11d2c23SStefan Behrens 41863f6bcfbdSStefan Behrens static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg) 41873f6bcfbdSStefan Behrens { 41883f6bcfbdSStefan Behrens struct btrfs_ioctl_dev_replace_args *p; 41893f6bcfbdSStefan Behrens int ret; 41903f6bcfbdSStefan Behrens 41913f6bcfbdSStefan Behrens if (!capable(CAP_SYS_ADMIN)) 41923f6bcfbdSStefan Behrens return -EPERM; 41933f6bcfbdSStefan Behrens 41943f6bcfbdSStefan Behrens p = memdup_user(arg, sizeof(*p)); 41953f6bcfbdSStefan Behrens if (IS_ERR(p)) 41963f6bcfbdSStefan Behrens return PTR_ERR(p); 41973f6bcfbdSStefan Behrens 41983f6bcfbdSStefan Behrens switch (p->cmd) { 41993f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_START: 4200adfa97cbSIlya Dryomov if (root->fs_info->sb->s_flags & MS_RDONLY) { 4201adfa97cbSIlya Dryomov ret = -EROFS; 4202adfa97cbSIlya Dryomov goto out; 4203adfa97cbSIlya Dryomov } 42043f6bcfbdSStefan Behrens if (atomic_xchg( 42053f6bcfbdSStefan Behrens &root->fs_info->mutually_exclusive_operation_running, 42063f6bcfbdSStefan Behrens 1)) { 4207e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 42083f6bcfbdSStefan Behrens } else { 42093f6bcfbdSStefan Behrens ret = btrfs_dev_replace_start(root, p); 42103f6bcfbdSStefan Behrens atomic_set( 42113f6bcfbdSStefan Behrens &root->fs_info->mutually_exclusive_operation_running, 42123f6bcfbdSStefan Behrens 0); 42133f6bcfbdSStefan Behrens } 42143f6bcfbdSStefan Behrens break; 42153f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS: 42163f6bcfbdSStefan Behrens btrfs_dev_replace_status(root->fs_info, p); 42173f6bcfbdSStefan Behrens ret = 0; 42183f6bcfbdSStefan Behrens break; 42193f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL: 42203f6bcfbdSStefan Behrens ret = btrfs_dev_replace_cancel(root->fs_info, p); 42213f6bcfbdSStefan Behrens break; 42223f6bcfbdSStefan Behrens default: 42233f6bcfbdSStefan Behrens ret = -EINVAL; 42243f6bcfbdSStefan Behrens break; 42253f6bcfbdSStefan Behrens } 42263f6bcfbdSStefan Behrens 42273f6bcfbdSStefan Behrens if (copy_to_user(arg, p, sizeof(*p))) 42283f6bcfbdSStefan Behrens ret = -EFAULT; 4229adfa97cbSIlya Dryomov out: 42303f6bcfbdSStefan Behrens kfree(p); 42313f6bcfbdSStefan Behrens return ret; 42323f6bcfbdSStefan Behrens } 42333f6bcfbdSStefan Behrens 4234d7728c96SJan Schmidt static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) 4235d7728c96SJan Schmidt { 4236d7728c96SJan Schmidt int ret = 0; 4237d7728c96SJan Schmidt int i; 4238740c3d22SChris Mason u64 rel_ptr; 4239d7728c96SJan Schmidt int size; 4240806468f8SChris Mason struct btrfs_ioctl_ino_path_args *ipa = NULL; 4241d7728c96SJan Schmidt struct inode_fs_paths *ipath = NULL; 4242d7728c96SJan Schmidt struct btrfs_path *path; 4243d7728c96SJan Schmidt 424482b22ac8SKusanagi Kouichi if (!capable(CAP_DAC_READ_SEARCH)) 4245d7728c96SJan Schmidt return -EPERM; 4246d7728c96SJan Schmidt 4247d7728c96SJan Schmidt path = btrfs_alloc_path(); 4248d7728c96SJan Schmidt if (!path) { 4249d7728c96SJan Schmidt ret = -ENOMEM; 4250d7728c96SJan Schmidt goto out; 4251d7728c96SJan Schmidt } 4252d7728c96SJan Schmidt 4253d7728c96SJan Schmidt ipa = memdup_user(arg, sizeof(*ipa)); 4254d7728c96SJan Schmidt if (IS_ERR(ipa)) { 4255d7728c96SJan Schmidt ret = PTR_ERR(ipa); 4256d7728c96SJan Schmidt ipa = NULL; 4257d7728c96SJan Schmidt goto out; 4258d7728c96SJan Schmidt } 4259d7728c96SJan Schmidt 4260d7728c96SJan Schmidt size = min_t(u32, ipa->size, 4096); 4261d7728c96SJan Schmidt ipath = init_ipath(size, root, path); 4262d7728c96SJan Schmidt if (IS_ERR(ipath)) { 4263d7728c96SJan Schmidt ret = PTR_ERR(ipath); 4264d7728c96SJan Schmidt ipath = NULL; 4265d7728c96SJan Schmidt goto out; 4266d7728c96SJan Schmidt } 4267d7728c96SJan Schmidt 4268d7728c96SJan Schmidt ret = paths_from_inode(ipa->inum, ipath); 4269d7728c96SJan Schmidt if (ret < 0) 4270d7728c96SJan Schmidt goto out; 4271d7728c96SJan Schmidt 4272d7728c96SJan Schmidt for (i = 0; i < ipath->fspath->elem_cnt; ++i) { 4273745c4d8eSJeff Mahoney rel_ptr = ipath->fspath->val[i] - 4274745c4d8eSJeff Mahoney (u64)(unsigned long)ipath->fspath->val; 4275740c3d22SChris Mason ipath->fspath->val[i] = rel_ptr; 4276d7728c96SJan Schmidt } 4277d7728c96SJan Schmidt 4278745c4d8eSJeff Mahoney ret = copy_to_user((void *)(unsigned long)ipa->fspath, 4279745c4d8eSJeff Mahoney (void *)(unsigned long)ipath->fspath, size); 4280d7728c96SJan Schmidt if (ret) { 4281d7728c96SJan Schmidt ret = -EFAULT; 4282d7728c96SJan Schmidt goto out; 4283d7728c96SJan Schmidt } 4284d7728c96SJan Schmidt 4285d7728c96SJan Schmidt out: 4286d7728c96SJan Schmidt btrfs_free_path(path); 4287d7728c96SJan Schmidt free_ipath(ipath); 4288d7728c96SJan Schmidt kfree(ipa); 4289d7728c96SJan Schmidt 4290d7728c96SJan Schmidt return ret; 4291d7728c96SJan Schmidt } 4292d7728c96SJan Schmidt 4293d7728c96SJan Schmidt static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx) 4294d7728c96SJan Schmidt { 4295d7728c96SJan Schmidt struct btrfs_data_container *inodes = ctx; 4296d7728c96SJan Schmidt const size_t c = 3 * sizeof(u64); 4297d7728c96SJan Schmidt 4298d7728c96SJan Schmidt if (inodes->bytes_left >= c) { 4299d7728c96SJan Schmidt inodes->bytes_left -= c; 4300d7728c96SJan Schmidt inodes->val[inodes->elem_cnt] = inum; 4301d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 1] = offset; 4302d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 2] = root; 4303d7728c96SJan Schmidt inodes->elem_cnt += 3; 4304d7728c96SJan Schmidt } else { 4305d7728c96SJan Schmidt inodes->bytes_missing += c - inodes->bytes_left; 4306d7728c96SJan Schmidt inodes->bytes_left = 0; 4307d7728c96SJan Schmidt inodes->elem_missed += 3; 4308d7728c96SJan Schmidt } 4309d7728c96SJan Schmidt 4310d7728c96SJan Schmidt return 0; 4311d7728c96SJan Schmidt } 4312d7728c96SJan Schmidt 4313d7728c96SJan Schmidt static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root, 4314d7728c96SJan Schmidt void __user *arg) 4315d7728c96SJan Schmidt { 4316d7728c96SJan Schmidt int ret = 0; 4317d7728c96SJan Schmidt int size; 4318d7728c96SJan Schmidt struct btrfs_ioctl_logical_ino_args *loi; 4319d7728c96SJan Schmidt struct btrfs_data_container *inodes = NULL; 4320d7728c96SJan Schmidt struct btrfs_path *path = NULL; 4321d7728c96SJan Schmidt 4322d7728c96SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4323d7728c96SJan Schmidt return -EPERM; 4324d7728c96SJan Schmidt 4325d7728c96SJan Schmidt loi = memdup_user(arg, sizeof(*loi)); 4326d7728c96SJan Schmidt if (IS_ERR(loi)) { 4327d7728c96SJan Schmidt ret = PTR_ERR(loi); 4328d7728c96SJan Schmidt loi = NULL; 4329d7728c96SJan Schmidt goto out; 4330d7728c96SJan Schmidt } 4331d7728c96SJan Schmidt 4332d7728c96SJan Schmidt path = btrfs_alloc_path(); 4333d7728c96SJan Schmidt if (!path) { 4334d7728c96SJan Schmidt ret = -ENOMEM; 4335d7728c96SJan Schmidt goto out; 4336d7728c96SJan Schmidt } 4337d7728c96SJan Schmidt 4338425d17a2SLiu Bo size = min_t(u32, loi->size, 64 * 1024); 4339d7728c96SJan Schmidt inodes = init_data_container(size); 4340d7728c96SJan Schmidt if (IS_ERR(inodes)) { 4341d7728c96SJan Schmidt ret = PTR_ERR(inodes); 4342d7728c96SJan Schmidt inodes = NULL; 4343d7728c96SJan Schmidt goto out; 4344d7728c96SJan Schmidt } 4345d7728c96SJan Schmidt 4346df031f07SLiu Bo ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path, 4347df031f07SLiu Bo build_ino_list, inodes); 4348df031f07SLiu Bo if (ret == -EINVAL) 4349d7728c96SJan Schmidt ret = -ENOENT; 4350d7728c96SJan Schmidt if (ret < 0) 4351d7728c96SJan Schmidt goto out; 4352d7728c96SJan Schmidt 4353745c4d8eSJeff Mahoney ret = copy_to_user((void *)(unsigned long)loi->inodes, 4354745c4d8eSJeff Mahoney (void *)(unsigned long)inodes, size); 4355d7728c96SJan Schmidt if (ret) 4356d7728c96SJan Schmidt ret = -EFAULT; 4357d7728c96SJan Schmidt 4358d7728c96SJan Schmidt out: 4359d7728c96SJan Schmidt btrfs_free_path(path); 4360425d17a2SLiu Bo vfree(inodes); 4361d7728c96SJan Schmidt kfree(loi); 4362d7728c96SJan Schmidt 4363d7728c96SJan Schmidt return ret; 4364d7728c96SJan Schmidt } 4365d7728c96SJan Schmidt 436619a39dceSIlya Dryomov void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock, 4367c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs) 4368c9e9f97bSIlya Dryomov { 4369c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl = fs_info->balance_ctl; 4370c9e9f97bSIlya Dryomov 4371c9e9f97bSIlya Dryomov bargs->flags = bctl->flags; 4372c9e9f97bSIlya Dryomov 4373837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_running)) 4374837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_RUNNING; 4375837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_pause_req)) 4376837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; 4377a7e99c69SIlya Dryomov if (atomic_read(&fs_info->balance_cancel_req)) 4378a7e99c69SIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; 4379837d5b6eSIlya Dryomov 4380c9e9f97bSIlya Dryomov memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); 4381c9e9f97bSIlya Dryomov memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); 4382c9e9f97bSIlya Dryomov memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); 438319a39dceSIlya Dryomov 438419a39dceSIlya Dryomov if (lock) { 438519a39dceSIlya Dryomov spin_lock(&fs_info->balance_lock); 438619a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 438719a39dceSIlya Dryomov spin_unlock(&fs_info->balance_lock); 438819a39dceSIlya Dryomov } else { 438919a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 439019a39dceSIlya Dryomov } 4391c9e9f97bSIlya Dryomov } 4392c9e9f97bSIlya Dryomov 43939ba1f6e4SLiu Bo static long btrfs_ioctl_balance(struct file *file, void __user *arg) 4394c9e9f97bSIlya Dryomov { 4395496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 4396c9e9f97bSIlya Dryomov struct btrfs_fs_info *fs_info = root->fs_info; 4397c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 4398c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl; 4399ed0fb78fSIlya Dryomov bool need_unlock; /* for mut. excl. ops lock */ 4400c9e9f97bSIlya Dryomov int ret; 4401c9e9f97bSIlya Dryomov 4402c9e9f97bSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4403c9e9f97bSIlya Dryomov return -EPERM; 4404c9e9f97bSIlya Dryomov 4405e54bfa31SLiu Bo ret = mnt_want_write_file(file); 44069ba1f6e4SLiu Bo if (ret) 44079ba1f6e4SLiu Bo return ret; 44089ba1f6e4SLiu Bo 4409ed0fb78fSIlya Dryomov again: 4410ed0fb78fSIlya Dryomov if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) { 4411c9e9f97bSIlya Dryomov mutex_lock(&fs_info->volume_mutex); 4412c9e9f97bSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4413ed0fb78fSIlya Dryomov need_unlock = true; 4414ed0fb78fSIlya Dryomov goto locked; 4415ed0fb78fSIlya Dryomov } 4416ed0fb78fSIlya Dryomov 4417ed0fb78fSIlya Dryomov /* 4418ed0fb78fSIlya Dryomov * mut. excl. ops lock is locked. Three possibilites: 4419ed0fb78fSIlya Dryomov * (1) some other op is running 4420ed0fb78fSIlya Dryomov * (2) balance is running 4421ed0fb78fSIlya Dryomov * (3) balance is paused -- special case (think resume) 4422ed0fb78fSIlya Dryomov */ 4423ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4424ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4425ed0fb78fSIlya Dryomov /* this is either (2) or (3) */ 4426ed0fb78fSIlya Dryomov if (!atomic_read(&fs_info->balance_running)) { 4427ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4428ed0fb78fSIlya Dryomov if (!mutex_trylock(&fs_info->volume_mutex)) 4429ed0fb78fSIlya Dryomov goto again; 4430ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4431ed0fb78fSIlya Dryomov 4432ed0fb78fSIlya Dryomov if (fs_info->balance_ctl && 4433ed0fb78fSIlya Dryomov !atomic_read(&fs_info->balance_running)) { 4434ed0fb78fSIlya Dryomov /* this is (3) */ 4435ed0fb78fSIlya Dryomov need_unlock = false; 4436ed0fb78fSIlya Dryomov goto locked; 4437ed0fb78fSIlya Dryomov } 4438ed0fb78fSIlya Dryomov 4439ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4440ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->volume_mutex); 4441ed0fb78fSIlya Dryomov goto again; 4442ed0fb78fSIlya Dryomov } else { 4443ed0fb78fSIlya Dryomov /* this is (2) */ 4444ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4445ed0fb78fSIlya Dryomov ret = -EINPROGRESS; 4446ed0fb78fSIlya Dryomov goto out; 4447ed0fb78fSIlya Dryomov } 4448ed0fb78fSIlya Dryomov } else { 4449ed0fb78fSIlya Dryomov /* this is (1) */ 4450ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4451e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 4452ed0fb78fSIlya Dryomov goto out; 4453ed0fb78fSIlya Dryomov } 4454ed0fb78fSIlya Dryomov 4455ed0fb78fSIlya Dryomov locked: 4456ed0fb78fSIlya Dryomov BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running)); 4457c9e9f97bSIlya Dryomov 4458c9e9f97bSIlya Dryomov if (arg) { 4459c9e9f97bSIlya Dryomov bargs = memdup_user(arg, sizeof(*bargs)); 4460c9e9f97bSIlya Dryomov if (IS_ERR(bargs)) { 4461c9e9f97bSIlya Dryomov ret = PTR_ERR(bargs); 4462ed0fb78fSIlya Dryomov goto out_unlock; 4463c9e9f97bSIlya Dryomov } 4464de322263SIlya Dryomov 4465de322263SIlya Dryomov if (bargs->flags & BTRFS_BALANCE_RESUME) { 4466de322263SIlya Dryomov if (!fs_info->balance_ctl) { 4467de322263SIlya Dryomov ret = -ENOTCONN; 4468de322263SIlya Dryomov goto out_bargs; 4469de322263SIlya Dryomov } 4470de322263SIlya Dryomov 4471de322263SIlya Dryomov bctl = fs_info->balance_ctl; 4472de322263SIlya Dryomov spin_lock(&fs_info->balance_lock); 4473de322263SIlya Dryomov bctl->flags |= BTRFS_BALANCE_RESUME; 4474de322263SIlya Dryomov spin_unlock(&fs_info->balance_lock); 4475de322263SIlya Dryomov 4476de322263SIlya Dryomov goto do_balance; 4477de322263SIlya Dryomov } 4478c9e9f97bSIlya Dryomov } else { 4479c9e9f97bSIlya Dryomov bargs = NULL; 4480c9e9f97bSIlya Dryomov } 4481c9e9f97bSIlya Dryomov 4482ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4483837d5b6eSIlya Dryomov ret = -EINPROGRESS; 4484837d5b6eSIlya Dryomov goto out_bargs; 4485837d5b6eSIlya Dryomov } 4486837d5b6eSIlya Dryomov 4487c9e9f97bSIlya Dryomov bctl = kzalloc(sizeof(*bctl), GFP_NOFS); 4488c9e9f97bSIlya Dryomov if (!bctl) { 4489c9e9f97bSIlya Dryomov ret = -ENOMEM; 4490c9e9f97bSIlya Dryomov goto out_bargs; 4491c9e9f97bSIlya Dryomov } 4492c9e9f97bSIlya Dryomov 4493c9e9f97bSIlya Dryomov bctl->fs_info = fs_info; 4494c9e9f97bSIlya Dryomov if (arg) { 4495c9e9f97bSIlya Dryomov memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); 4496c9e9f97bSIlya Dryomov memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); 4497c9e9f97bSIlya Dryomov memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); 4498c9e9f97bSIlya Dryomov 4499c9e9f97bSIlya Dryomov bctl->flags = bargs->flags; 4500f43ffb60SIlya Dryomov } else { 4501f43ffb60SIlya Dryomov /* balance everything - no filters */ 4502f43ffb60SIlya Dryomov bctl->flags |= BTRFS_BALANCE_TYPE_MASK; 4503c9e9f97bSIlya Dryomov } 4504c9e9f97bSIlya Dryomov 4505de322263SIlya Dryomov do_balance: 4506c9e9f97bSIlya Dryomov /* 4507ed0fb78fSIlya Dryomov * Ownership of bctl and mutually_exclusive_operation_running 4508ed0fb78fSIlya Dryomov * goes to to btrfs_balance. bctl is freed in __cancel_balance, 4509ed0fb78fSIlya Dryomov * or, if restriper was paused all the way until unmount, in 4510ed0fb78fSIlya Dryomov * free_fs_info. mutually_exclusive_operation_running is 4511ed0fb78fSIlya Dryomov * cleared in __cancel_balance. 4512c9e9f97bSIlya Dryomov */ 4513ed0fb78fSIlya Dryomov need_unlock = false; 4514ed0fb78fSIlya Dryomov 4515ed0fb78fSIlya Dryomov ret = btrfs_balance(bctl, bargs); 4516ed0fb78fSIlya Dryomov 4517c9e9f97bSIlya Dryomov if (arg) { 4518c9e9f97bSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 4519c9e9f97bSIlya Dryomov ret = -EFAULT; 4520c9e9f97bSIlya Dryomov } 4521c9e9f97bSIlya Dryomov 4522c9e9f97bSIlya Dryomov out_bargs: 4523c9e9f97bSIlya Dryomov kfree(bargs); 4524ed0fb78fSIlya Dryomov out_unlock: 4525c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4526c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->volume_mutex); 4527ed0fb78fSIlya Dryomov if (need_unlock) 4528ed0fb78fSIlya Dryomov atomic_set(&fs_info->mutually_exclusive_operation_running, 0); 4529ed0fb78fSIlya Dryomov out: 4530e54bfa31SLiu Bo mnt_drop_write_file(file); 4531c9e9f97bSIlya Dryomov return ret; 4532c9e9f97bSIlya Dryomov } 4533c9e9f97bSIlya Dryomov 4534837d5b6eSIlya Dryomov static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd) 4535837d5b6eSIlya Dryomov { 4536837d5b6eSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4537837d5b6eSIlya Dryomov return -EPERM; 4538837d5b6eSIlya Dryomov 4539837d5b6eSIlya Dryomov switch (cmd) { 4540837d5b6eSIlya Dryomov case BTRFS_BALANCE_CTL_PAUSE: 4541837d5b6eSIlya Dryomov return btrfs_pause_balance(root->fs_info); 4542a7e99c69SIlya Dryomov case BTRFS_BALANCE_CTL_CANCEL: 4543a7e99c69SIlya Dryomov return btrfs_cancel_balance(root->fs_info); 4544837d5b6eSIlya Dryomov } 4545837d5b6eSIlya Dryomov 4546837d5b6eSIlya Dryomov return -EINVAL; 4547837d5b6eSIlya Dryomov } 4548837d5b6eSIlya Dryomov 454919a39dceSIlya Dryomov static long btrfs_ioctl_balance_progress(struct btrfs_root *root, 455019a39dceSIlya Dryomov void __user *arg) 455119a39dceSIlya Dryomov { 455219a39dceSIlya Dryomov struct btrfs_fs_info *fs_info = root->fs_info; 455319a39dceSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 455419a39dceSIlya Dryomov int ret = 0; 455519a39dceSIlya Dryomov 455619a39dceSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 455719a39dceSIlya Dryomov return -EPERM; 455819a39dceSIlya Dryomov 455919a39dceSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 456019a39dceSIlya Dryomov if (!fs_info->balance_ctl) { 456119a39dceSIlya Dryomov ret = -ENOTCONN; 456219a39dceSIlya Dryomov goto out; 456319a39dceSIlya Dryomov } 456419a39dceSIlya Dryomov 456519a39dceSIlya Dryomov bargs = kzalloc(sizeof(*bargs), GFP_NOFS); 456619a39dceSIlya Dryomov if (!bargs) { 456719a39dceSIlya Dryomov ret = -ENOMEM; 456819a39dceSIlya Dryomov goto out; 456919a39dceSIlya Dryomov } 457019a39dceSIlya Dryomov 457119a39dceSIlya Dryomov update_ioctl_balance_args(fs_info, 1, bargs); 457219a39dceSIlya Dryomov 457319a39dceSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 457419a39dceSIlya Dryomov ret = -EFAULT; 457519a39dceSIlya Dryomov 457619a39dceSIlya Dryomov kfree(bargs); 457719a39dceSIlya Dryomov out: 457819a39dceSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 457919a39dceSIlya Dryomov return ret; 458019a39dceSIlya Dryomov } 458119a39dceSIlya Dryomov 4582905b0ddaSMiao Xie static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) 45835d13a37bSArne Jansen { 4584496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 45855d13a37bSArne Jansen struct btrfs_ioctl_quota_ctl_args *sa; 45865d13a37bSArne Jansen struct btrfs_trans_handle *trans = NULL; 45875d13a37bSArne Jansen int ret; 45885d13a37bSArne Jansen int err; 45895d13a37bSArne Jansen 45905d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 45915d13a37bSArne Jansen return -EPERM; 45925d13a37bSArne Jansen 4593905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4594905b0ddaSMiao Xie if (ret) 4595905b0ddaSMiao Xie return ret; 45965d13a37bSArne Jansen 45975d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4598905b0ddaSMiao Xie if (IS_ERR(sa)) { 4599905b0ddaSMiao Xie ret = PTR_ERR(sa); 4600905b0ddaSMiao Xie goto drop_write; 4601905b0ddaSMiao Xie } 46025d13a37bSArne Jansen 46037708f029SWang Shilong down_write(&root->fs_info->subvol_sem); 460492f183aaSWang Shilong trans = btrfs_start_transaction(root->fs_info->tree_root, 2); 46055d13a37bSArne Jansen if (IS_ERR(trans)) { 46065d13a37bSArne Jansen ret = PTR_ERR(trans); 46075d13a37bSArne Jansen goto out; 46085d13a37bSArne Jansen } 46095d13a37bSArne Jansen 46105d13a37bSArne Jansen switch (sa->cmd) { 46115d13a37bSArne Jansen case BTRFS_QUOTA_CTL_ENABLE: 46125d13a37bSArne Jansen ret = btrfs_quota_enable(trans, root->fs_info); 46135d13a37bSArne Jansen break; 46145d13a37bSArne Jansen case BTRFS_QUOTA_CTL_DISABLE: 46155d13a37bSArne Jansen ret = btrfs_quota_disable(trans, root->fs_info); 46165d13a37bSArne Jansen break; 46175d13a37bSArne Jansen default: 46185d13a37bSArne Jansen ret = -EINVAL; 46195d13a37bSArne Jansen break; 46205d13a37bSArne Jansen } 46215d13a37bSArne Jansen 462292f183aaSWang Shilong err = btrfs_commit_transaction(trans, root->fs_info->tree_root); 46235d13a37bSArne Jansen if (err && !ret) 46245d13a37bSArne Jansen ret = err; 46255d13a37bSArne Jansen out: 46265d13a37bSArne Jansen kfree(sa); 46277708f029SWang Shilong up_write(&root->fs_info->subvol_sem); 4628905b0ddaSMiao Xie drop_write: 4629905b0ddaSMiao Xie mnt_drop_write_file(file); 46305d13a37bSArne Jansen return ret; 46315d13a37bSArne Jansen } 46325d13a37bSArne Jansen 4633905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg) 46345d13a37bSArne Jansen { 4635496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 46365d13a37bSArne Jansen struct btrfs_ioctl_qgroup_assign_args *sa; 46375d13a37bSArne Jansen struct btrfs_trans_handle *trans; 46385d13a37bSArne Jansen int ret; 46395d13a37bSArne Jansen int err; 46405d13a37bSArne Jansen 46415d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 46425d13a37bSArne Jansen return -EPERM; 46435d13a37bSArne Jansen 4644905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4645905b0ddaSMiao Xie if (ret) 4646905b0ddaSMiao Xie return ret; 46475d13a37bSArne Jansen 46485d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4649905b0ddaSMiao Xie if (IS_ERR(sa)) { 4650905b0ddaSMiao Xie ret = PTR_ERR(sa); 4651905b0ddaSMiao Xie goto drop_write; 4652905b0ddaSMiao Xie } 46535d13a37bSArne Jansen 46545d13a37bSArne Jansen trans = btrfs_join_transaction(root); 46555d13a37bSArne Jansen if (IS_ERR(trans)) { 46565d13a37bSArne Jansen ret = PTR_ERR(trans); 46575d13a37bSArne Jansen goto out; 46585d13a37bSArne Jansen } 46595d13a37bSArne Jansen 46605d13a37bSArne Jansen /* FIXME: check if the IDs really exist */ 46615d13a37bSArne Jansen if (sa->assign) { 46625d13a37bSArne Jansen ret = btrfs_add_qgroup_relation(trans, root->fs_info, 46635d13a37bSArne Jansen sa->src, sa->dst); 46645d13a37bSArne Jansen } else { 46655d13a37bSArne Jansen ret = btrfs_del_qgroup_relation(trans, root->fs_info, 46665d13a37bSArne Jansen sa->src, sa->dst); 46675d13a37bSArne Jansen } 46685d13a37bSArne Jansen 46695d13a37bSArne Jansen err = btrfs_end_transaction(trans, root); 46705d13a37bSArne Jansen if (err && !ret) 46715d13a37bSArne Jansen ret = err; 46725d13a37bSArne Jansen 46735d13a37bSArne Jansen out: 46745d13a37bSArne Jansen kfree(sa); 4675905b0ddaSMiao Xie drop_write: 4676905b0ddaSMiao Xie mnt_drop_write_file(file); 46775d13a37bSArne Jansen return ret; 46785d13a37bSArne Jansen } 46795d13a37bSArne Jansen 4680905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg) 46815d13a37bSArne Jansen { 4682496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 46835d13a37bSArne Jansen struct btrfs_ioctl_qgroup_create_args *sa; 46845d13a37bSArne Jansen struct btrfs_trans_handle *trans; 46855d13a37bSArne Jansen int ret; 46865d13a37bSArne Jansen int err; 46875d13a37bSArne Jansen 46885d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 46895d13a37bSArne Jansen return -EPERM; 46905d13a37bSArne Jansen 4691905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4692905b0ddaSMiao Xie if (ret) 4693905b0ddaSMiao Xie return ret; 46945d13a37bSArne Jansen 46955d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4696905b0ddaSMiao Xie if (IS_ERR(sa)) { 4697905b0ddaSMiao Xie ret = PTR_ERR(sa); 4698905b0ddaSMiao Xie goto drop_write; 4699905b0ddaSMiao Xie } 47005d13a37bSArne Jansen 4701d86e56cfSMiao Xie if (!sa->qgroupid) { 4702d86e56cfSMiao Xie ret = -EINVAL; 4703d86e56cfSMiao Xie goto out; 4704d86e56cfSMiao Xie } 4705d86e56cfSMiao Xie 47065d13a37bSArne Jansen trans = btrfs_join_transaction(root); 47075d13a37bSArne Jansen if (IS_ERR(trans)) { 47085d13a37bSArne Jansen ret = PTR_ERR(trans); 47095d13a37bSArne Jansen goto out; 47105d13a37bSArne Jansen } 47115d13a37bSArne Jansen 47125d13a37bSArne Jansen /* FIXME: check if the IDs really exist */ 47135d13a37bSArne Jansen if (sa->create) { 47145d13a37bSArne Jansen ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid, 47155d13a37bSArne Jansen NULL); 47165d13a37bSArne Jansen } else { 47175d13a37bSArne Jansen ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid); 47185d13a37bSArne Jansen } 47195d13a37bSArne Jansen 47205d13a37bSArne Jansen err = btrfs_end_transaction(trans, root); 47215d13a37bSArne Jansen if (err && !ret) 47225d13a37bSArne Jansen ret = err; 47235d13a37bSArne Jansen 47245d13a37bSArne Jansen out: 47255d13a37bSArne Jansen kfree(sa); 4726905b0ddaSMiao Xie drop_write: 4727905b0ddaSMiao Xie mnt_drop_write_file(file); 47285d13a37bSArne Jansen return ret; 47295d13a37bSArne Jansen } 47305d13a37bSArne Jansen 4731905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg) 47325d13a37bSArne Jansen { 4733496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 47345d13a37bSArne Jansen struct btrfs_ioctl_qgroup_limit_args *sa; 47355d13a37bSArne Jansen struct btrfs_trans_handle *trans; 47365d13a37bSArne Jansen int ret; 47375d13a37bSArne Jansen int err; 47385d13a37bSArne Jansen u64 qgroupid; 47395d13a37bSArne Jansen 47405d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 47415d13a37bSArne Jansen return -EPERM; 47425d13a37bSArne Jansen 4743905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4744905b0ddaSMiao Xie if (ret) 4745905b0ddaSMiao Xie return ret; 47465d13a37bSArne Jansen 47475d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4748905b0ddaSMiao Xie if (IS_ERR(sa)) { 4749905b0ddaSMiao Xie ret = PTR_ERR(sa); 4750905b0ddaSMiao Xie goto drop_write; 4751905b0ddaSMiao Xie } 47525d13a37bSArne Jansen 47535d13a37bSArne Jansen trans = btrfs_join_transaction(root); 47545d13a37bSArne Jansen if (IS_ERR(trans)) { 47555d13a37bSArne Jansen ret = PTR_ERR(trans); 47565d13a37bSArne Jansen goto out; 47575d13a37bSArne Jansen } 47585d13a37bSArne Jansen 47595d13a37bSArne Jansen qgroupid = sa->qgroupid; 47605d13a37bSArne Jansen if (!qgroupid) { 47615d13a37bSArne Jansen /* take the current subvol as qgroup */ 47625d13a37bSArne Jansen qgroupid = root->root_key.objectid; 47635d13a37bSArne Jansen } 47645d13a37bSArne Jansen 47655d13a37bSArne Jansen /* FIXME: check if the IDs really exist */ 47665d13a37bSArne Jansen ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim); 47675d13a37bSArne Jansen 47685d13a37bSArne Jansen err = btrfs_end_transaction(trans, root); 47695d13a37bSArne Jansen if (err && !ret) 47705d13a37bSArne Jansen ret = err; 47715d13a37bSArne Jansen 47725d13a37bSArne Jansen out: 47735d13a37bSArne Jansen kfree(sa); 4774905b0ddaSMiao Xie drop_write: 4775905b0ddaSMiao Xie mnt_drop_write_file(file); 47765d13a37bSArne Jansen return ret; 47775d13a37bSArne Jansen } 47785d13a37bSArne Jansen 47792f232036SJan Schmidt static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) 47802f232036SJan Schmidt { 47816d0379ecSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 47822f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 47832f232036SJan Schmidt int ret; 47842f232036SJan Schmidt 47852f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 47862f232036SJan Schmidt return -EPERM; 47872f232036SJan Schmidt 47882f232036SJan Schmidt ret = mnt_want_write_file(file); 47892f232036SJan Schmidt if (ret) 47902f232036SJan Schmidt return ret; 47912f232036SJan Schmidt 47922f232036SJan Schmidt qsa = memdup_user(arg, sizeof(*qsa)); 47932f232036SJan Schmidt if (IS_ERR(qsa)) { 47942f232036SJan Schmidt ret = PTR_ERR(qsa); 47952f232036SJan Schmidt goto drop_write; 47962f232036SJan Schmidt } 47972f232036SJan Schmidt 47982f232036SJan Schmidt if (qsa->flags) { 47992f232036SJan Schmidt ret = -EINVAL; 48002f232036SJan Schmidt goto out; 48012f232036SJan Schmidt } 48022f232036SJan Schmidt 48032f232036SJan Schmidt ret = btrfs_qgroup_rescan(root->fs_info); 48042f232036SJan Schmidt 48052f232036SJan Schmidt out: 48062f232036SJan Schmidt kfree(qsa); 48072f232036SJan Schmidt drop_write: 48082f232036SJan Schmidt mnt_drop_write_file(file); 48092f232036SJan Schmidt return ret; 48102f232036SJan Schmidt } 48112f232036SJan Schmidt 48122f232036SJan Schmidt static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg) 48132f232036SJan Schmidt { 48146d0379ecSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 48152f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 48162f232036SJan Schmidt int ret = 0; 48172f232036SJan Schmidt 48182f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 48192f232036SJan Schmidt return -EPERM; 48202f232036SJan Schmidt 48212f232036SJan Schmidt qsa = kzalloc(sizeof(*qsa), GFP_NOFS); 48222f232036SJan Schmidt if (!qsa) 48232f232036SJan Schmidt return -ENOMEM; 48242f232036SJan Schmidt 48252f232036SJan Schmidt if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { 48262f232036SJan Schmidt qsa->flags = 1; 48272f232036SJan Schmidt qsa->progress = root->fs_info->qgroup_rescan_progress.objectid; 48282f232036SJan Schmidt } 48292f232036SJan Schmidt 48302f232036SJan Schmidt if (copy_to_user(arg, qsa, sizeof(*qsa))) 48312f232036SJan Schmidt ret = -EFAULT; 48322f232036SJan Schmidt 48332f232036SJan Schmidt kfree(qsa); 48342f232036SJan Schmidt return ret; 48352f232036SJan Schmidt } 48362f232036SJan Schmidt 483757254b6eSJan Schmidt static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg) 483857254b6eSJan Schmidt { 483954563d41SAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 484057254b6eSJan Schmidt 484157254b6eSJan Schmidt if (!capable(CAP_SYS_ADMIN)) 484257254b6eSJan Schmidt return -EPERM; 484357254b6eSJan Schmidt 484457254b6eSJan Schmidt return btrfs_qgroup_wait_for_completion(root->fs_info); 484557254b6eSJan Schmidt } 484657254b6eSJan Schmidt 4847abccd00fSHugo Mills static long _btrfs_ioctl_set_received_subvol(struct file *file, 4848abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa) 48498ea05e3aSAlexander Block { 4850496ad9aaSAl Viro struct inode *inode = file_inode(file); 48518ea05e3aSAlexander Block struct btrfs_root *root = BTRFS_I(inode)->root; 48528ea05e3aSAlexander Block struct btrfs_root_item *root_item = &root->root_item; 48538ea05e3aSAlexander Block struct btrfs_trans_handle *trans; 48548ea05e3aSAlexander Block struct timespec ct = CURRENT_TIME; 48558ea05e3aSAlexander Block int ret = 0; 4856dd5f9615SStefan Behrens int received_uuid_changed; 48578ea05e3aSAlexander Block 4858bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 4859bd60ea0fSDavid Sterba return -EPERM; 4860bd60ea0fSDavid Sterba 48618ea05e3aSAlexander Block ret = mnt_want_write_file(file); 48628ea05e3aSAlexander Block if (ret < 0) 48638ea05e3aSAlexander Block return ret; 48648ea05e3aSAlexander Block 48658ea05e3aSAlexander Block down_write(&root->fs_info->subvol_sem); 48668ea05e3aSAlexander Block 48678ea05e3aSAlexander Block if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { 48688ea05e3aSAlexander Block ret = -EINVAL; 48698ea05e3aSAlexander Block goto out; 48708ea05e3aSAlexander Block } 48718ea05e3aSAlexander Block 48728ea05e3aSAlexander Block if (btrfs_root_readonly(root)) { 48738ea05e3aSAlexander Block ret = -EROFS; 48748ea05e3aSAlexander Block goto out; 48758ea05e3aSAlexander Block } 48768ea05e3aSAlexander Block 4877dd5f9615SStefan Behrens /* 4878dd5f9615SStefan Behrens * 1 - root item 4879dd5f9615SStefan Behrens * 2 - uuid items (received uuid + subvol uuid) 4880dd5f9615SStefan Behrens */ 4881dd5f9615SStefan Behrens trans = btrfs_start_transaction(root, 3); 48828ea05e3aSAlexander Block if (IS_ERR(trans)) { 48838ea05e3aSAlexander Block ret = PTR_ERR(trans); 48848ea05e3aSAlexander Block trans = NULL; 48858ea05e3aSAlexander Block goto out; 48868ea05e3aSAlexander Block } 48878ea05e3aSAlexander Block 48888ea05e3aSAlexander Block sa->rtransid = trans->transid; 48898ea05e3aSAlexander Block sa->rtime.sec = ct.tv_sec; 48908ea05e3aSAlexander Block sa->rtime.nsec = ct.tv_nsec; 48918ea05e3aSAlexander Block 4892dd5f9615SStefan Behrens received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid, 4893dd5f9615SStefan Behrens BTRFS_UUID_SIZE); 4894dd5f9615SStefan Behrens if (received_uuid_changed && 4895dd5f9615SStefan Behrens !btrfs_is_empty_uuid(root_item->received_uuid)) 4896dd5f9615SStefan Behrens btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root, 4897dd5f9615SStefan Behrens root_item->received_uuid, 4898dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 4899dd5f9615SStefan Behrens root->root_key.objectid); 49008ea05e3aSAlexander Block memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE); 49018ea05e3aSAlexander Block btrfs_set_root_stransid(root_item, sa->stransid); 49028ea05e3aSAlexander Block btrfs_set_root_rtransid(root_item, sa->rtransid); 49033cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec); 49043cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec); 49053cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec); 49063cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec); 49078ea05e3aSAlexander Block 49088ea05e3aSAlexander Block ret = btrfs_update_root(trans, root->fs_info->tree_root, 49098ea05e3aSAlexander Block &root->root_key, &root->root_item); 49108ea05e3aSAlexander Block if (ret < 0) { 49118ea05e3aSAlexander Block btrfs_end_transaction(trans, root); 49128ea05e3aSAlexander Block goto out; 4913dd5f9615SStefan Behrens } 4914dd5f9615SStefan Behrens if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { 4915dd5f9615SStefan Behrens ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root, 4916dd5f9615SStefan Behrens sa->uuid, 4917dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 4918dd5f9615SStefan Behrens root->root_key.objectid); 4919dd5f9615SStefan Behrens if (ret < 0 && ret != -EEXIST) { 4920dd5f9615SStefan Behrens btrfs_abort_transaction(trans, root, ret); 4921dd5f9615SStefan Behrens goto out; 4922dd5f9615SStefan Behrens } 4923dd5f9615SStefan Behrens } 49248ea05e3aSAlexander Block ret = btrfs_commit_transaction(trans, root); 4925dd5f9615SStefan Behrens if (ret < 0) { 4926dd5f9615SStefan Behrens btrfs_abort_transaction(trans, root, ret); 49278ea05e3aSAlexander Block goto out; 49288ea05e3aSAlexander Block } 49298ea05e3aSAlexander Block 4930abccd00fSHugo Mills out: 4931abccd00fSHugo Mills up_write(&root->fs_info->subvol_sem); 4932abccd00fSHugo Mills mnt_drop_write_file(file); 4933abccd00fSHugo Mills return ret; 4934abccd00fSHugo Mills } 4935abccd00fSHugo Mills 4936abccd00fSHugo Mills #ifdef CONFIG_64BIT 4937abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol_32(struct file *file, 4938abccd00fSHugo Mills void __user *arg) 4939abccd00fSHugo Mills { 4940abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL; 4941abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *args64 = NULL; 4942abccd00fSHugo Mills int ret = 0; 4943abccd00fSHugo Mills 4944abccd00fSHugo Mills args32 = memdup_user(arg, sizeof(*args32)); 4945abccd00fSHugo Mills if (IS_ERR(args32)) { 4946abccd00fSHugo Mills ret = PTR_ERR(args32); 4947abccd00fSHugo Mills args32 = NULL; 4948abccd00fSHugo Mills goto out; 4949abccd00fSHugo Mills } 4950abccd00fSHugo Mills 4951abccd00fSHugo Mills args64 = kmalloc(sizeof(*args64), GFP_NOFS); 495284dbeb87SDan Carpenter if (!args64) { 495384dbeb87SDan Carpenter ret = -ENOMEM; 4954abccd00fSHugo Mills goto out; 4955abccd00fSHugo Mills } 4956abccd00fSHugo Mills 4957abccd00fSHugo Mills memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE); 4958abccd00fSHugo Mills args64->stransid = args32->stransid; 4959abccd00fSHugo Mills args64->rtransid = args32->rtransid; 4960abccd00fSHugo Mills args64->stime.sec = args32->stime.sec; 4961abccd00fSHugo Mills args64->stime.nsec = args32->stime.nsec; 4962abccd00fSHugo Mills args64->rtime.sec = args32->rtime.sec; 4963abccd00fSHugo Mills args64->rtime.nsec = args32->rtime.nsec; 4964abccd00fSHugo Mills args64->flags = args32->flags; 4965abccd00fSHugo Mills 4966abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, args64); 4967abccd00fSHugo Mills if (ret) 4968abccd00fSHugo Mills goto out; 4969abccd00fSHugo Mills 4970abccd00fSHugo Mills memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE); 4971abccd00fSHugo Mills args32->stransid = args64->stransid; 4972abccd00fSHugo Mills args32->rtransid = args64->rtransid; 4973abccd00fSHugo Mills args32->stime.sec = args64->stime.sec; 4974abccd00fSHugo Mills args32->stime.nsec = args64->stime.nsec; 4975abccd00fSHugo Mills args32->rtime.sec = args64->rtime.sec; 4976abccd00fSHugo Mills args32->rtime.nsec = args64->rtime.nsec; 4977abccd00fSHugo Mills args32->flags = args64->flags; 4978abccd00fSHugo Mills 4979abccd00fSHugo Mills ret = copy_to_user(arg, args32, sizeof(*args32)); 4980abccd00fSHugo Mills if (ret) 4981abccd00fSHugo Mills ret = -EFAULT; 4982abccd00fSHugo Mills 4983abccd00fSHugo Mills out: 4984abccd00fSHugo Mills kfree(args32); 4985abccd00fSHugo Mills kfree(args64); 4986abccd00fSHugo Mills return ret; 4987abccd00fSHugo Mills } 4988abccd00fSHugo Mills #endif 4989abccd00fSHugo Mills 4990abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol(struct file *file, 4991abccd00fSHugo Mills void __user *arg) 4992abccd00fSHugo Mills { 4993abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa = NULL; 4994abccd00fSHugo Mills int ret = 0; 4995abccd00fSHugo Mills 4996abccd00fSHugo Mills sa = memdup_user(arg, sizeof(*sa)); 4997abccd00fSHugo Mills if (IS_ERR(sa)) { 4998abccd00fSHugo Mills ret = PTR_ERR(sa); 4999abccd00fSHugo Mills sa = NULL; 5000abccd00fSHugo Mills goto out; 5001abccd00fSHugo Mills } 5002abccd00fSHugo Mills 5003abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, sa); 5004abccd00fSHugo Mills 5005abccd00fSHugo Mills if (ret) 5006abccd00fSHugo Mills goto out; 5007abccd00fSHugo Mills 50088ea05e3aSAlexander Block ret = copy_to_user(arg, sa, sizeof(*sa)); 50098ea05e3aSAlexander Block if (ret) 50108ea05e3aSAlexander Block ret = -EFAULT; 50118ea05e3aSAlexander Block 50128ea05e3aSAlexander Block out: 50138ea05e3aSAlexander Block kfree(sa); 50148ea05e3aSAlexander Block return ret; 50158ea05e3aSAlexander Block } 50168ea05e3aSAlexander Block 5017867ab667Sjeff.liu static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg) 5018867ab667Sjeff.liu { 50196d0379ecSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 5020a1b83ac5SAnand Jain size_t len; 5021867ab667Sjeff.liu int ret; 5022a1b83ac5SAnand Jain char label[BTRFS_LABEL_SIZE]; 5023a1b83ac5SAnand Jain 5024a1b83ac5SAnand Jain spin_lock(&root->fs_info->super_lock); 5025a1b83ac5SAnand Jain memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE); 5026a1b83ac5SAnand Jain spin_unlock(&root->fs_info->super_lock); 5027a1b83ac5SAnand Jain 5028a1b83ac5SAnand Jain len = strnlen(label, BTRFS_LABEL_SIZE); 5029867ab667Sjeff.liu 5030867ab667Sjeff.liu if (len == BTRFS_LABEL_SIZE) { 5031efe120a0SFrank Holton btrfs_warn(root->fs_info, 5032efe120a0SFrank Holton "label is too long, return the first %zu bytes", --len); 5033867ab667Sjeff.liu } 5034867ab667Sjeff.liu 5035867ab667Sjeff.liu ret = copy_to_user(arg, label, len); 5036867ab667Sjeff.liu 5037867ab667Sjeff.liu return ret ? -EFAULT : 0; 5038867ab667Sjeff.liu } 5039867ab667Sjeff.liu 5040a8bfd4abSjeff.liu static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg) 5041a8bfd4abSjeff.liu { 50426d0379ecSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 5043a8bfd4abSjeff.liu struct btrfs_super_block *super_block = root->fs_info->super_copy; 5044a8bfd4abSjeff.liu struct btrfs_trans_handle *trans; 5045a8bfd4abSjeff.liu char label[BTRFS_LABEL_SIZE]; 5046a8bfd4abSjeff.liu int ret; 5047a8bfd4abSjeff.liu 5048a8bfd4abSjeff.liu if (!capable(CAP_SYS_ADMIN)) 5049a8bfd4abSjeff.liu return -EPERM; 5050a8bfd4abSjeff.liu 5051a8bfd4abSjeff.liu if (copy_from_user(label, arg, sizeof(label))) 5052a8bfd4abSjeff.liu return -EFAULT; 5053a8bfd4abSjeff.liu 5054a8bfd4abSjeff.liu if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) { 5055efe120a0SFrank Holton btrfs_err(root->fs_info, "unable to set label with more than %d bytes", 5056a8bfd4abSjeff.liu BTRFS_LABEL_SIZE - 1); 5057a8bfd4abSjeff.liu return -EINVAL; 5058a8bfd4abSjeff.liu } 5059a8bfd4abSjeff.liu 5060a8bfd4abSjeff.liu ret = mnt_want_write_file(file); 5061a8bfd4abSjeff.liu if (ret) 5062a8bfd4abSjeff.liu return ret; 5063a8bfd4abSjeff.liu 5064a8bfd4abSjeff.liu trans = btrfs_start_transaction(root, 0); 5065a8bfd4abSjeff.liu if (IS_ERR(trans)) { 5066a8bfd4abSjeff.liu ret = PTR_ERR(trans); 5067a8bfd4abSjeff.liu goto out_unlock; 5068a8bfd4abSjeff.liu } 5069a8bfd4abSjeff.liu 5070a1b83ac5SAnand Jain spin_lock(&root->fs_info->super_lock); 5071a8bfd4abSjeff.liu strcpy(super_block->label, label); 5072a1b83ac5SAnand Jain spin_unlock(&root->fs_info->super_lock); 5073d0270acaSJeff Mahoney ret = btrfs_commit_transaction(trans, root); 5074a8bfd4abSjeff.liu 5075a8bfd4abSjeff.liu out_unlock: 5076a8bfd4abSjeff.liu mnt_drop_write_file(file); 5077a8bfd4abSjeff.liu return ret; 5078a8bfd4abSjeff.liu } 5079a8bfd4abSjeff.liu 50802eaa055fSJeff Mahoney #define INIT_FEATURE_FLAGS(suffix) \ 50812eaa055fSJeff Mahoney { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \ 50822eaa055fSJeff Mahoney .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \ 50832eaa055fSJeff Mahoney .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix } 50842eaa055fSJeff Mahoney 50852eaa055fSJeff Mahoney static int btrfs_ioctl_get_supported_features(struct file *file, 50862eaa055fSJeff Mahoney void __user *arg) 50872eaa055fSJeff Mahoney { 50882eaa055fSJeff Mahoney static struct btrfs_ioctl_feature_flags features[3] = { 50892eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SUPP), 50902eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_SET), 50912eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_CLEAR) 50922eaa055fSJeff Mahoney }; 50932eaa055fSJeff Mahoney 50942eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 50952eaa055fSJeff Mahoney return -EFAULT; 50962eaa055fSJeff Mahoney 50972eaa055fSJeff Mahoney return 0; 50982eaa055fSJeff Mahoney } 50992eaa055fSJeff Mahoney 51002eaa055fSJeff Mahoney static int btrfs_ioctl_get_features(struct file *file, void __user *arg) 51012eaa055fSJeff Mahoney { 51022eaa055fSJeff Mahoney struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 51032eaa055fSJeff Mahoney struct btrfs_super_block *super_block = root->fs_info->super_copy; 51042eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags features; 51052eaa055fSJeff Mahoney 51062eaa055fSJeff Mahoney features.compat_flags = btrfs_super_compat_flags(super_block); 51072eaa055fSJeff Mahoney features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block); 51082eaa055fSJeff Mahoney features.incompat_flags = btrfs_super_incompat_flags(super_block); 51092eaa055fSJeff Mahoney 51102eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 51112eaa055fSJeff Mahoney return -EFAULT; 51122eaa055fSJeff Mahoney 51132eaa055fSJeff Mahoney return 0; 51142eaa055fSJeff Mahoney } 51152eaa055fSJeff Mahoney 51163b02a68aSJeff Mahoney static int check_feature_bits(struct btrfs_root *root, 51173b02a68aSJeff Mahoney enum btrfs_feature_set set, 51182eaa055fSJeff Mahoney u64 change_mask, u64 flags, u64 supported_flags, 51192eaa055fSJeff Mahoney u64 safe_set, u64 safe_clear) 51202eaa055fSJeff Mahoney { 51213b02a68aSJeff Mahoney const char *type = btrfs_feature_set_names[set]; 51223b02a68aSJeff Mahoney char *names; 51232eaa055fSJeff Mahoney u64 disallowed, unsupported; 51242eaa055fSJeff Mahoney u64 set_mask = flags & change_mask; 51252eaa055fSJeff Mahoney u64 clear_mask = ~flags & change_mask; 51262eaa055fSJeff Mahoney 51272eaa055fSJeff Mahoney unsupported = set_mask & ~supported_flags; 51282eaa055fSJeff Mahoney if (unsupported) { 51293b02a68aSJeff Mahoney names = btrfs_printable_features(set, unsupported); 51303b02a68aSJeff Mahoney if (names) { 51313b02a68aSJeff Mahoney btrfs_warn(root->fs_info, 51323b02a68aSJeff Mahoney "this kernel does not support the %s feature bit%s", 51333b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 51343b02a68aSJeff Mahoney kfree(names); 51353b02a68aSJeff Mahoney } else 51362eaa055fSJeff Mahoney btrfs_warn(root->fs_info, 51372eaa055fSJeff Mahoney "this kernel does not support %s bits 0x%llx", 51382eaa055fSJeff Mahoney type, unsupported); 51392eaa055fSJeff Mahoney return -EOPNOTSUPP; 51402eaa055fSJeff Mahoney } 51412eaa055fSJeff Mahoney 51422eaa055fSJeff Mahoney disallowed = set_mask & ~safe_set; 51432eaa055fSJeff Mahoney if (disallowed) { 51443b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 51453b02a68aSJeff Mahoney if (names) { 51463b02a68aSJeff Mahoney btrfs_warn(root->fs_info, 51473b02a68aSJeff Mahoney "can't set the %s feature bit%s while mounted", 51483b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 51493b02a68aSJeff Mahoney kfree(names); 51503b02a68aSJeff Mahoney } else 51512eaa055fSJeff Mahoney btrfs_warn(root->fs_info, 51522eaa055fSJeff Mahoney "can't set %s bits 0x%llx while mounted", 51532eaa055fSJeff Mahoney type, disallowed); 51542eaa055fSJeff Mahoney return -EPERM; 51552eaa055fSJeff Mahoney } 51562eaa055fSJeff Mahoney 51572eaa055fSJeff Mahoney disallowed = clear_mask & ~safe_clear; 51582eaa055fSJeff Mahoney if (disallowed) { 51593b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 51603b02a68aSJeff Mahoney if (names) { 51613b02a68aSJeff Mahoney btrfs_warn(root->fs_info, 51623b02a68aSJeff Mahoney "can't clear the %s feature bit%s while mounted", 51633b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 51643b02a68aSJeff Mahoney kfree(names); 51653b02a68aSJeff Mahoney } else 51662eaa055fSJeff Mahoney btrfs_warn(root->fs_info, 51672eaa055fSJeff Mahoney "can't clear %s bits 0x%llx while mounted", 51682eaa055fSJeff Mahoney type, disallowed); 51692eaa055fSJeff Mahoney return -EPERM; 51702eaa055fSJeff Mahoney } 51712eaa055fSJeff Mahoney 51722eaa055fSJeff Mahoney return 0; 51732eaa055fSJeff Mahoney } 51742eaa055fSJeff Mahoney 51752eaa055fSJeff Mahoney #define check_feature(root, change_mask, flags, mask_base) \ 51763b02a68aSJeff Mahoney check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \ 51772eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SUPP, \ 51782eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \ 51792eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR) 51802eaa055fSJeff Mahoney 51812eaa055fSJeff Mahoney static int btrfs_ioctl_set_features(struct file *file, void __user *arg) 51822eaa055fSJeff Mahoney { 51832eaa055fSJeff Mahoney struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 51842eaa055fSJeff Mahoney struct btrfs_super_block *super_block = root->fs_info->super_copy; 51852eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags flags[2]; 51862eaa055fSJeff Mahoney struct btrfs_trans_handle *trans; 51872eaa055fSJeff Mahoney u64 newflags; 51882eaa055fSJeff Mahoney int ret; 51892eaa055fSJeff Mahoney 51902eaa055fSJeff Mahoney if (!capable(CAP_SYS_ADMIN)) 51912eaa055fSJeff Mahoney return -EPERM; 51922eaa055fSJeff Mahoney 51932eaa055fSJeff Mahoney if (copy_from_user(flags, arg, sizeof(flags))) 51942eaa055fSJeff Mahoney return -EFAULT; 51952eaa055fSJeff Mahoney 51962eaa055fSJeff Mahoney /* Nothing to do */ 51972eaa055fSJeff Mahoney if (!flags[0].compat_flags && !flags[0].compat_ro_flags && 51982eaa055fSJeff Mahoney !flags[0].incompat_flags) 51992eaa055fSJeff Mahoney return 0; 52002eaa055fSJeff Mahoney 52012eaa055fSJeff Mahoney ret = check_feature(root, flags[0].compat_flags, 52022eaa055fSJeff Mahoney flags[1].compat_flags, COMPAT); 52032eaa055fSJeff Mahoney if (ret) 52042eaa055fSJeff Mahoney return ret; 52052eaa055fSJeff Mahoney 52062eaa055fSJeff Mahoney ret = check_feature(root, flags[0].compat_ro_flags, 52072eaa055fSJeff Mahoney flags[1].compat_ro_flags, COMPAT_RO); 52082eaa055fSJeff Mahoney if (ret) 52092eaa055fSJeff Mahoney return ret; 52102eaa055fSJeff Mahoney 52112eaa055fSJeff Mahoney ret = check_feature(root, flags[0].incompat_flags, 52122eaa055fSJeff Mahoney flags[1].incompat_flags, INCOMPAT); 52132eaa055fSJeff Mahoney if (ret) 52142eaa055fSJeff Mahoney return ret; 52152eaa055fSJeff Mahoney 52168051aa1aSDavid Sterba trans = btrfs_start_transaction(root, 0); 52172eaa055fSJeff Mahoney if (IS_ERR(trans)) 52182eaa055fSJeff Mahoney return PTR_ERR(trans); 52192eaa055fSJeff Mahoney 52202eaa055fSJeff Mahoney spin_lock(&root->fs_info->super_lock); 52212eaa055fSJeff Mahoney newflags = btrfs_super_compat_flags(super_block); 52222eaa055fSJeff Mahoney newflags |= flags[0].compat_flags & flags[1].compat_flags; 52232eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags); 52242eaa055fSJeff Mahoney btrfs_set_super_compat_flags(super_block, newflags); 52252eaa055fSJeff Mahoney 52262eaa055fSJeff Mahoney newflags = btrfs_super_compat_ro_flags(super_block); 52272eaa055fSJeff Mahoney newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags; 52282eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags); 52292eaa055fSJeff Mahoney btrfs_set_super_compat_ro_flags(super_block, newflags); 52302eaa055fSJeff Mahoney 52312eaa055fSJeff Mahoney newflags = btrfs_super_incompat_flags(super_block); 52322eaa055fSJeff Mahoney newflags |= flags[0].incompat_flags & flags[1].incompat_flags; 52332eaa055fSJeff Mahoney newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags); 52342eaa055fSJeff Mahoney btrfs_set_super_incompat_flags(super_block, newflags); 52352eaa055fSJeff Mahoney spin_unlock(&root->fs_info->super_lock); 52362eaa055fSJeff Mahoney 5237d0270acaSJeff Mahoney return btrfs_commit_transaction(trans, root); 52382eaa055fSJeff Mahoney } 52392eaa055fSJeff Mahoney 5240f46b5a66SChristoph Hellwig long btrfs_ioctl(struct file *file, unsigned int 5241f46b5a66SChristoph Hellwig cmd, unsigned long arg) 5242f46b5a66SChristoph Hellwig { 5243496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 52444bcabaa3SChristoph Hellwig void __user *argp = (void __user *)arg; 5245f46b5a66SChristoph Hellwig 5246f46b5a66SChristoph Hellwig switch (cmd) { 52476cbff00fSChristoph Hellwig case FS_IOC_GETFLAGS: 52486cbff00fSChristoph Hellwig return btrfs_ioctl_getflags(file, argp); 52496cbff00fSChristoph Hellwig case FS_IOC_SETFLAGS: 52506cbff00fSChristoph Hellwig return btrfs_ioctl_setflags(file, argp); 52516cbff00fSChristoph Hellwig case FS_IOC_GETVERSION: 52526cbff00fSChristoph Hellwig return btrfs_ioctl_getversion(file, argp); 5253f7039b1dSLi Dongyang case FITRIM: 5254f7039b1dSLi Dongyang return btrfs_ioctl_fitrim(file, argp); 5255f46b5a66SChristoph Hellwig case BTRFS_IOC_SNAP_CREATE: 5256fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 0); 5257fdfb1e4fSLi Zefan case BTRFS_IOC_SNAP_CREATE_V2: 5258fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create_v2(file, argp, 0); 52593de4586cSChris Mason case BTRFS_IOC_SUBVOL_CREATE: 5260fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 1); 52616f72c7e2SArne Jansen case BTRFS_IOC_SUBVOL_CREATE_V2: 52626f72c7e2SArne Jansen return btrfs_ioctl_snap_create_v2(file, argp, 1); 526376dda93cSYan, Zheng case BTRFS_IOC_SNAP_DESTROY: 526476dda93cSYan, Zheng return btrfs_ioctl_snap_destroy(file, argp); 52650caa102dSLi Zefan case BTRFS_IOC_SUBVOL_GETFLAGS: 52660caa102dSLi Zefan return btrfs_ioctl_subvol_getflags(file, argp); 52670caa102dSLi Zefan case BTRFS_IOC_SUBVOL_SETFLAGS: 52680caa102dSLi Zefan return btrfs_ioctl_subvol_setflags(file, argp); 52696ef5ed0dSJosef Bacik case BTRFS_IOC_DEFAULT_SUBVOL: 52706ef5ed0dSJosef Bacik return btrfs_ioctl_default_subvol(file, argp); 5271f46b5a66SChristoph Hellwig case BTRFS_IOC_DEFRAG: 52721e701a32SChris Mason return btrfs_ioctl_defrag(file, NULL); 52731e701a32SChris Mason case BTRFS_IOC_DEFRAG_RANGE: 52741e701a32SChris Mason return btrfs_ioctl_defrag(file, argp); 5275f46b5a66SChristoph Hellwig case BTRFS_IOC_RESIZE: 5276198605a8SMiao Xie return btrfs_ioctl_resize(file, argp); 5277f46b5a66SChristoph Hellwig case BTRFS_IOC_ADD_DEV: 52784bcabaa3SChristoph Hellwig return btrfs_ioctl_add_dev(root, argp); 5279f46b5a66SChristoph Hellwig case BTRFS_IOC_RM_DEV: 5280da24927bSMiao Xie return btrfs_ioctl_rm_dev(file, argp); 5281475f6387SJan Schmidt case BTRFS_IOC_FS_INFO: 5282475f6387SJan Schmidt return btrfs_ioctl_fs_info(root, argp); 5283475f6387SJan Schmidt case BTRFS_IOC_DEV_INFO: 5284475f6387SJan Schmidt return btrfs_ioctl_dev_info(root, argp); 5285f46b5a66SChristoph Hellwig case BTRFS_IOC_BALANCE: 52869ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, NULL); 5287f46b5a66SChristoph Hellwig case BTRFS_IOC_CLONE: 5288c5c9cd4dSSage Weil return btrfs_ioctl_clone(file, arg, 0, 0, 0); 5289c5c9cd4dSSage Weil case BTRFS_IOC_CLONE_RANGE: 52907a865e8aSChristoph Hellwig return btrfs_ioctl_clone_range(file, argp); 5291f46b5a66SChristoph Hellwig case BTRFS_IOC_TRANS_START: 5292f46b5a66SChristoph Hellwig return btrfs_ioctl_trans_start(file); 5293f46b5a66SChristoph Hellwig case BTRFS_IOC_TRANS_END: 5294f46b5a66SChristoph Hellwig return btrfs_ioctl_trans_end(file); 5295ac8e9819SChris Mason case BTRFS_IOC_TREE_SEARCH: 5296ac8e9819SChris Mason return btrfs_ioctl_tree_search(file, argp); 5297cc68a8a5SGerhard Heift case BTRFS_IOC_TREE_SEARCH_V2: 5298cc68a8a5SGerhard Heift return btrfs_ioctl_tree_search_v2(file, argp); 5299ac8e9819SChris Mason case BTRFS_IOC_INO_LOOKUP: 5300ac8e9819SChris Mason return btrfs_ioctl_ino_lookup(file, argp); 5301d7728c96SJan Schmidt case BTRFS_IOC_INO_PATHS: 5302d7728c96SJan Schmidt return btrfs_ioctl_ino_to_path(root, argp); 5303d7728c96SJan Schmidt case BTRFS_IOC_LOGICAL_INO: 5304d7728c96SJan Schmidt return btrfs_ioctl_logical_to_ino(root, argp); 53051406e432SJosef Bacik case BTRFS_IOC_SPACE_INFO: 53061406e432SJosef Bacik return btrfs_ioctl_space_info(root, argp); 53079b199859SFilipe David Borba Manana case BTRFS_IOC_SYNC: { 53089b199859SFilipe David Borba Manana int ret; 53099b199859SFilipe David Borba Manana 53106c255e67SMiao Xie ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1); 53119b199859SFilipe David Borba Manana if (ret) 53129b199859SFilipe David Borba Manana return ret; 53139b199859SFilipe David Borba Manana ret = btrfs_sync_fs(file->f_dentry->d_sb, 1); 53149b199859SFilipe David Borba Manana return ret; 53159b199859SFilipe David Borba Manana } 531646204592SSage Weil case BTRFS_IOC_START_SYNC: 53179a8c28beSMiao Xie return btrfs_ioctl_start_sync(root, argp); 531846204592SSage Weil case BTRFS_IOC_WAIT_SYNC: 53199a8c28beSMiao Xie return btrfs_ioctl_wait_sync(root, argp); 5320475f6387SJan Schmidt case BTRFS_IOC_SCRUB: 5321b8e95489SMiao Xie return btrfs_ioctl_scrub(file, argp); 5322475f6387SJan Schmidt case BTRFS_IOC_SCRUB_CANCEL: 5323475f6387SJan Schmidt return btrfs_ioctl_scrub_cancel(root, argp); 5324475f6387SJan Schmidt case BTRFS_IOC_SCRUB_PROGRESS: 5325475f6387SJan Schmidt return btrfs_ioctl_scrub_progress(root, argp); 5326c9e9f97bSIlya Dryomov case BTRFS_IOC_BALANCE_V2: 53279ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, argp); 5328837d5b6eSIlya Dryomov case BTRFS_IOC_BALANCE_CTL: 5329837d5b6eSIlya Dryomov return btrfs_ioctl_balance_ctl(root, arg); 533019a39dceSIlya Dryomov case BTRFS_IOC_BALANCE_PROGRESS: 533119a39dceSIlya Dryomov return btrfs_ioctl_balance_progress(root, argp); 53328ea05e3aSAlexander Block case BTRFS_IOC_SET_RECEIVED_SUBVOL: 53338ea05e3aSAlexander Block return btrfs_ioctl_set_received_subvol(file, argp); 5334abccd00fSHugo Mills #ifdef CONFIG_64BIT 5335abccd00fSHugo Mills case BTRFS_IOC_SET_RECEIVED_SUBVOL_32: 5336abccd00fSHugo Mills return btrfs_ioctl_set_received_subvol_32(file, argp); 5337abccd00fSHugo Mills #endif 533831db9f7cSAlexander Block case BTRFS_IOC_SEND: 533931db9f7cSAlexander Block return btrfs_ioctl_send(file, argp); 5340c11d2c23SStefan Behrens case BTRFS_IOC_GET_DEV_STATS: 5341b27f7c0cSDavid Sterba return btrfs_ioctl_get_dev_stats(root, argp); 53425d13a37bSArne Jansen case BTRFS_IOC_QUOTA_CTL: 5343905b0ddaSMiao Xie return btrfs_ioctl_quota_ctl(file, argp); 53445d13a37bSArne Jansen case BTRFS_IOC_QGROUP_ASSIGN: 5345905b0ddaSMiao Xie return btrfs_ioctl_qgroup_assign(file, argp); 53465d13a37bSArne Jansen case BTRFS_IOC_QGROUP_CREATE: 5347905b0ddaSMiao Xie return btrfs_ioctl_qgroup_create(file, argp); 53485d13a37bSArne Jansen case BTRFS_IOC_QGROUP_LIMIT: 5349905b0ddaSMiao Xie return btrfs_ioctl_qgroup_limit(file, argp); 53502f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN: 53512f232036SJan Schmidt return btrfs_ioctl_quota_rescan(file, argp); 53522f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_STATUS: 53532f232036SJan Schmidt return btrfs_ioctl_quota_rescan_status(file, argp); 535457254b6eSJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_WAIT: 535557254b6eSJan Schmidt return btrfs_ioctl_quota_rescan_wait(file, argp); 53563f6bcfbdSStefan Behrens case BTRFS_IOC_DEV_REPLACE: 53573f6bcfbdSStefan Behrens return btrfs_ioctl_dev_replace(root, argp); 5358867ab667Sjeff.liu case BTRFS_IOC_GET_FSLABEL: 5359867ab667Sjeff.liu return btrfs_ioctl_get_fslabel(file, argp); 5360a8bfd4abSjeff.liu case BTRFS_IOC_SET_FSLABEL: 5361a8bfd4abSjeff.liu return btrfs_ioctl_set_fslabel(file, argp); 5362416161dbSMark Fasheh case BTRFS_IOC_FILE_EXTENT_SAME: 5363416161dbSMark Fasheh return btrfs_ioctl_file_extent_same(file, argp); 53642eaa055fSJeff Mahoney case BTRFS_IOC_GET_SUPPORTED_FEATURES: 53652eaa055fSJeff Mahoney return btrfs_ioctl_get_supported_features(file, argp); 53662eaa055fSJeff Mahoney case BTRFS_IOC_GET_FEATURES: 53672eaa055fSJeff Mahoney return btrfs_ioctl_get_features(file, argp); 53682eaa055fSJeff Mahoney case BTRFS_IOC_SET_FEATURES: 53692eaa055fSJeff Mahoney return btrfs_ioctl_set_features(file, argp); 5370f46b5a66SChristoph Hellwig } 5371f46b5a66SChristoph Hellwig 5372f46b5a66SChristoph Hellwig return -ENOTTY; 5373f46b5a66SChristoph Hellwig } 5374