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, 901c919a5eSMark Fasheh u64 off, u64 olen, u64 olen_aligned, u64 destoff, 911c919a5eSMark Fasheh int no_time_update); 92416161dbSMark Fasheh 936cbff00fSChristoph Hellwig /* Mask out flags that are inappropriate for the given type of inode. */ 946cbff00fSChristoph Hellwig static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags) 956cbff00fSChristoph Hellwig { 966cbff00fSChristoph Hellwig if (S_ISDIR(mode)) 976cbff00fSChristoph Hellwig return flags; 986cbff00fSChristoph Hellwig else if (S_ISREG(mode)) 996cbff00fSChristoph Hellwig return flags & ~FS_DIRSYNC_FL; 1006cbff00fSChristoph Hellwig else 1016cbff00fSChristoph Hellwig return flags & (FS_NODUMP_FL | FS_NOATIME_FL); 1026cbff00fSChristoph Hellwig } 103f46b5a66SChristoph Hellwig 1046cbff00fSChristoph Hellwig /* 1056cbff00fSChristoph Hellwig * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl. 1066cbff00fSChristoph Hellwig */ 1076cbff00fSChristoph Hellwig static unsigned int btrfs_flags_to_ioctl(unsigned int flags) 1086cbff00fSChristoph Hellwig { 1096cbff00fSChristoph Hellwig unsigned int iflags = 0; 1106cbff00fSChristoph Hellwig 1116cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_SYNC) 1126cbff00fSChristoph Hellwig iflags |= FS_SYNC_FL; 1136cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_IMMUTABLE) 1146cbff00fSChristoph Hellwig iflags |= FS_IMMUTABLE_FL; 1156cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_APPEND) 1166cbff00fSChristoph Hellwig iflags |= FS_APPEND_FL; 1176cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NODUMP) 1186cbff00fSChristoph Hellwig iflags |= FS_NODUMP_FL; 1196cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NOATIME) 1206cbff00fSChristoph Hellwig iflags |= FS_NOATIME_FL; 1216cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_DIRSYNC) 1226cbff00fSChristoph Hellwig iflags |= FS_DIRSYNC_FL; 123d0092bddSLi Zefan if (flags & BTRFS_INODE_NODATACOW) 124d0092bddSLi Zefan iflags |= FS_NOCOW_FL; 125d0092bddSLi Zefan 126d0092bddSLi Zefan if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS)) 127d0092bddSLi Zefan iflags |= FS_COMPR_FL; 128d0092bddSLi Zefan else if (flags & BTRFS_INODE_NOCOMPRESS) 129d0092bddSLi Zefan iflags |= FS_NOCOMP_FL; 1306cbff00fSChristoph Hellwig 1316cbff00fSChristoph Hellwig return iflags; 1326cbff00fSChristoph Hellwig } 1336cbff00fSChristoph Hellwig 1346cbff00fSChristoph Hellwig /* 1356cbff00fSChristoph Hellwig * Update inode->i_flags based on the btrfs internal flags. 1366cbff00fSChristoph Hellwig */ 1376cbff00fSChristoph Hellwig void btrfs_update_iflags(struct inode *inode) 1386cbff00fSChristoph Hellwig { 1396cbff00fSChristoph Hellwig struct btrfs_inode *ip = BTRFS_I(inode); 1403cc79392SFilipe Manana unsigned int new_fl = 0; 1416cbff00fSChristoph Hellwig 1426cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_SYNC) 1433cc79392SFilipe Manana new_fl |= S_SYNC; 1446cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_IMMUTABLE) 1453cc79392SFilipe Manana new_fl |= S_IMMUTABLE; 1466cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_APPEND) 1473cc79392SFilipe Manana new_fl |= S_APPEND; 1486cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_NOATIME) 1493cc79392SFilipe Manana new_fl |= S_NOATIME; 1506cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_DIRSYNC) 1513cc79392SFilipe Manana new_fl |= S_DIRSYNC; 1523cc79392SFilipe Manana 1533cc79392SFilipe Manana set_mask_bits(&inode->i_flags, 1543cc79392SFilipe Manana S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC, 1553cc79392SFilipe Manana new_fl); 1566cbff00fSChristoph Hellwig } 1576cbff00fSChristoph Hellwig 1586cbff00fSChristoph Hellwig /* 1596cbff00fSChristoph Hellwig * Inherit flags from the parent inode. 1606cbff00fSChristoph Hellwig * 161e27425d6SJosef Bacik * Currently only the compression flags and the cow flags are inherited. 1626cbff00fSChristoph Hellwig */ 1636cbff00fSChristoph Hellwig void btrfs_inherit_iflags(struct inode *inode, struct inode *dir) 1646cbff00fSChristoph Hellwig { 1650b4dcea5SChris Mason unsigned int flags; 1660b4dcea5SChris Mason 1670b4dcea5SChris Mason if (!dir) 1680b4dcea5SChris Mason return; 1690b4dcea5SChris Mason 1700b4dcea5SChris Mason flags = BTRFS_I(dir)->flags; 1716cbff00fSChristoph Hellwig 172e27425d6SJosef Bacik if (flags & BTRFS_INODE_NOCOMPRESS) { 173e27425d6SJosef Bacik BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS; 174e27425d6SJosef Bacik BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS; 175e27425d6SJosef Bacik } else if (flags & BTRFS_INODE_COMPRESS) { 176e27425d6SJosef Bacik BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS; 177e27425d6SJosef Bacik BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS; 178e27425d6SJosef Bacik } 1796cbff00fSChristoph Hellwig 180213490b3SLiu Bo if (flags & BTRFS_INODE_NODATACOW) { 181e27425d6SJosef Bacik BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW; 182213490b3SLiu Bo if (S_ISREG(inode->i_mode)) 183213490b3SLiu Bo BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM; 184213490b3SLiu Bo } 185e27425d6SJosef Bacik 1866cbff00fSChristoph Hellwig btrfs_update_iflags(inode); 1876cbff00fSChristoph Hellwig } 1886cbff00fSChristoph Hellwig 1896cbff00fSChristoph Hellwig static int btrfs_ioctl_getflags(struct file *file, void __user *arg) 1906cbff00fSChristoph Hellwig { 191496ad9aaSAl Viro struct btrfs_inode *ip = BTRFS_I(file_inode(file)); 1926cbff00fSChristoph Hellwig unsigned int flags = btrfs_flags_to_ioctl(ip->flags); 1936cbff00fSChristoph Hellwig 1946cbff00fSChristoph Hellwig if (copy_to_user(arg, &flags, sizeof(flags))) 1956cbff00fSChristoph Hellwig return -EFAULT; 1966cbff00fSChristoph Hellwig return 0; 1976cbff00fSChristoph Hellwig } 1986cbff00fSChristoph Hellwig 19975e7cb7fSLiu Bo static int check_flags(unsigned int flags) 20075e7cb7fSLiu Bo { 20175e7cb7fSLiu Bo if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \ 20275e7cb7fSLiu Bo FS_NOATIME_FL | FS_NODUMP_FL | \ 20375e7cb7fSLiu Bo FS_SYNC_FL | FS_DIRSYNC_FL | \ 204e1e8fb6aSLi Zefan FS_NOCOMP_FL | FS_COMPR_FL | 205e1e8fb6aSLi Zefan FS_NOCOW_FL)) 20675e7cb7fSLiu Bo return -EOPNOTSUPP; 20775e7cb7fSLiu Bo 20875e7cb7fSLiu Bo if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL)) 20975e7cb7fSLiu Bo return -EINVAL; 21075e7cb7fSLiu Bo 21175e7cb7fSLiu Bo return 0; 21275e7cb7fSLiu Bo } 21375e7cb7fSLiu Bo 2146cbff00fSChristoph Hellwig static int btrfs_ioctl_setflags(struct file *file, void __user *arg) 2156cbff00fSChristoph Hellwig { 216496ad9aaSAl Viro struct inode *inode = file_inode(file); 2176cbff00fSChristoph Hellwig struct btrfs_inode *ip = BTRFS_I(inode); 2186cbff00fSChristoph Hellwig struct btrfs_root *root = ip->root; 2196cbff00fSChristoph Hellwig struct btrfs_trans_handle *trans; 2206cbff00fSChristoph Hellwig unsigned int flags, oldflags; 2216cbff00fSChristoph Hellwig int ret; 222f062abf0SLi Zefan u64 ip_oldflags; 223f062abf0SLi Zefan unsigned int i_oldflags; 2247e97b8daSDavid Sterba umode_t mode; 2256cbff00fSChristoph Hellwig 226bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 227bd60ea0fSDavid Sterba return -EPERM; 228bd60ea0fSDavid Sterba 229b83cc969SLi Zefan if (btrfs_root_readonly(root)) 230b83cc969SLi Zefan return -EROFS; 231b83cc969SLi Zefan 2326cbff00fSChristoph Hellwig if (copy_from_user(&flags, arg, sizeof(flags))) 2336cbff00fSChristoph Hellwig return -EFAULT; 2346cbff00fSChristoph Hellwig 23575e7cb7fSLiu Bo ret = check_flags(flags); 23675e7cb7fSLiu Bo if (ret) 23775e7cb7fSLiu Bo return ret; 2386cbff00fSChristoph Hellwig 239e7848683SJan Kara ret = mnt_want_write_file(file); 240e7848683SJan Kara if (ret) 241e7848683SJan Kara return ret; 242e7848683SJan Kara 2436cbff00fSChristoph Hellwig mutex_lock(&inode->i_mutex); 2446cbff00fSChristoph Hellwig 245f062abf0SLi Zefan ip_oldflags = ip->flags; 246f062abf0SLi Zefan i_oldflags = inode->i_flags; 2477e97b8daSDavid Sterba mode = inode->i_mode; 248f062abf0SLi Zefan 2496cbff00fSChristoph Hellwig flags = btrfs_mask_flags(inode->i_mode, flags); 2506cbff00fSChristoph Hellwig oldflags = btrfs_flags_to_ioctl(ip->flags); 2516cbff00fSChristoph Hellwig if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) { 2526cbff00fSChristoph Hellwig if (!capable(CAP_LINUX_IMMUTABLE)) { 2536cbff00fSChristoph Hellwig ret = -EPERM; 2546cbff00fSChristoph Hellwig goto out_unlock; 2556cbff00fSChristoph Hellwig } 2566cbff00fSChristoph Hellwig } 2576cbff00fSChristoph Hellwig 2586cbff00fSChristoph Hellwig if (flags & FS_SYNC_FL) 2596cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_SYNC; 2606cbff00fSChristoph Hellwig else 2616cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_SYNC; 2626cbff00fSChristoph Hellwig if (flags & FS_IMMUTABLE_FL) 2636cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_IMMUTABLE; 2646cbff00fSChristoph Hellwig else 2656cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_IMMUTABLE; 2666cbff00fSChristoph Hellwig if (flags & FS_APPEND_FL) 2676cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_APPEND; 2686cbff00fSChristoph Hellwig else 2696cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_APPEND; 2706cbff00fSChristoph Hellwig if (flags & FS_NODUMP_FL) 2716cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_NODUMP; 2726cbff00fSChristoph Hellwig else 2736cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_NODUMP; 2746cbff00fSChristoph Hellwig if (flags & FS_NOATIME_FL) 2756cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_NOATIME; 2766cbff00fSChristoph Hellwig else 2776cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_NOATIME; 2786cbff00fSChristoph Hellwig if (flags & FS_DIRSYNC_FL) 2796cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_DIRSYNC; 2806cbff00fSChristoph Hellwig else 2816cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_DIRSYNC; 2827e97b8daSDavid Sterba if (flags & FS_NOCOW_FL) { 2837e97b8daSDavid Sterba if (S_ISREG(mode)) { 2847e97b8daSDavid Sterba /* 2857e97b8daSDavid Sterba * It's safe to turn csums off here, no extents exist. 2867e97b8daSDavid Sterba * Otherwise we want the flag to reflect the real COW 2877e97b8daSDavid Sterba * status of the file and will not set it. 2887e97b8daSDavid Sterba */ 2897e97b8daSDavid Sterba if (inode->i_size == 0) 2907e97b8daSDavid Sterba ip->flags |= BTRFS_INODE_NODATACOW 2917e97b8daSDavid Sterba | BTRFS_INODE_NODATASUM; 2927e97b8daSDavid Sterba } else { 293e1e8fb6aSLi Zefan ip->flags |= BTRFS_INODE_NODATACOW; 2947e97b8daSDavid Sterba } 2957e97b8daSDavid Sterba } else { 2967e97b8daSDavid Sterba /* 2977e97b8daSDavid Sterba * Revert back under same assuptions as above 2987e97b8daSDavid Sterba */ 2997e97b8daSDavid Sterba if (S_ISREG(mode)) { 3007e97b8daSDavid Sterba if (inode->i_size == 0) 3017e97b8daSDavid Sterba ip->flags &= ~(BTRFS_INODE_NODATACOW 3027e97b8daSDavid Sterba | BTRFS_INODE_NODATASUM); 3037e97b8daSDavid Sterba } else { 304e1e8fb6aSLi Zefan ip->flags &= ~BTRFS_INODE_NODATACOW; 3057e97b8daSDavid Sterba } 3067e97b8daSDavid Sterba } 3076cbff00fSChristoph Hellwig 30875e7cb7fSLiu Bo /* 30975e7cb7fSLiu Bo * The COMPRESS flag can only be changed by users, while the NOCOMPRESS 31075e7cb7fSLiu Bo * flag may be changed automatically if compression code won't make 31175e7cb7fSLiu Bo * things smaller. 31275e7cb7fSLiu Bo */ 31375e7cb7fSLiu Bo if (flags & FS_NOCOMP_FL) { 31475e7cb7fSLiu Bo ip->flags &= ~BTRFS_INODE_COMPRESS; 31575e7cb7fSLiu Bo ip->flags |= BTRFS_INODE_NOCOMPRESS; 31663541927SFilipe David Borba Manana 31763541927SFilipe David Borba Manana ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0); 31863541927SFilipe David Borba Manana if (ret && ret != -ENODATA) 31963541927SFilipe David Borba Manana goto out_drop; 32075e7cb7fSLiu Bo } else if (flags & FS_COMPR_FL) { 32163541927SFilipe David Borba Manana const char *comp; 32263541927SFilipe David Borba Manana 32375e7cb7fSLiu Bo ip->flags |= BTRFS_INODE_COMPRESS; 32475e7cb7fSLiu Bo ip->flags &= ~BTRFS_INODE_NOCOMPRESS; 32563541927SFilipe David Borba Manana 32663541927SFilipe David Borba Manana if (root->fs_info->compress_type == BTRFS_COMPRESS_LZO) 32763541927SFilipe David Borba Manana comp = "lzo"; 32863541927SFilipe David Borba Manana else 32963541927SFilipe David Borba Manana comp = "zlib"; 33063541927SFilipe David Borba Manana ret = btrfs_set_prop(inode, "btrfs.compression", 33163541927SFilipe David Borba Manana comp, strlen(comp), 0); 33263541927SFilipe David Borba Manana if (ret) 33363541927SFilipe David Borba Manana goto out_drop; 33463541927SFilipe David Borba Manana 335ebcb904dSLi Zefan } else { 33678a017a2SFilipe Manana ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0); 33778a017a2SFilipe Manana if (ret && ret != -ENODATA) 33878a017a2SFilipe Manana goto out_drop; 339ebcb904dSLi Zefan ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS); 34075e7cb7fSLiu Bo } 3416cbff00fSChristoph Hellwig 3424da6f1a3SLi Zefan trans = btrfs_start_transaction(root, 1); 343f062abf0SLi Zefan if (IS_ERR(trans)) { 344f062abf0SLi Zefan ret = PTR_ERR(trans); 345f062abf0SLi Zefan goto out_drop; 346f062abf0SLi Zefan } 3476cbff00fSChristoph Hellwig 348306424ccSLi Zefan btrfs_update_iflags(inode); 3490c4d2d95SJosef Bacik inode_inc_iversion(inode); 350306424ccSLi Zefan inode->i_ctime = CURRENT_TIME; 3516cbff00fSChristoph Hellwig ret = btrfs_update_inode(trans, root, inode); 3526cbff00fSChristoph Hellwig 3536cbff00fSChristoph Hellwig btrfs_end_transaction(trans, root); 354f062abf0SLi Zefan out_drop: 355f062abf0SLi Zefan if (ret) { 356f062abf0SLi Zefan ip->flags = ip_oldflags; 357f062abf0SLi Zefan inode->i_flags = i_oldflags; 358f062abf0SLi Zefan } 3596cbff00fSChristoph Hellwig 3606cbff00fSChristoph Hellwig out_unlock: 3616cbff00fSChristoph Hellwig mutex_unlock(&inode->i_mutex); 362e7848683SJan Kara mnt_drop_write_file(file); 3632d4e6f6aSliubo return ret; 3646cbff00fSChristoph Hellwig } 3656cbff00fSChristoph Hellwig 3666cbff00fSChristoph Hellwig static int btrfs_ioctl_getversion(struct file *file, int __user *arg) 3676cbff00fSChristoph Hellwig { 368496ad9aaSAl Viro struct inode *inode = file_inode(file); 3696cbff00fSChristoph Hellwig 3706cbff00fSChristoph Hellwig return put_user(inode->i_generation, arg); 3716cbff00fSChristoph Hellwig } 372f46b5a66SChristoph Hellwig 373f7039b1dSLi Dongyang static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg) 374f7039b1dSLi Dongyang { 37554563d41SAl Viro struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb); 376f7039b1dSLi Dongyang struct btrfs_device *device; 377f7039b1dSLi Dongyang struct request_queue *q; 378f7039b1dSLi Dongyang struct fstrim_range range; 379f7039b1dSLi Dongyang u64 minlen = ULLONG_MAX; 380f7039b1dSLi Dongyang u64 num_devices = 0; 381815745cfSAl Viro u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy); 382f7039b1dSLi Dongyang int ret; 383f7039b1dSLi Dongyang 384f7039b1dSLi Dongyang if (!capable(CAP_SYS_ADMIN)) 385f7039b1dSLi Dongyang return -EPERM; 386f7039b1dSLi Dongyang 3871f78160cSXiao Guangrong rcu_read_lock(); 3881f78160cSXiao Guangrong list_for_each_entry_rcu(device, &fs_info->fs_devices->devices, 3891f78160cSXiao Guangrong dev_list) { 390f7039b1dSLi Dongyang if (!device->bdev) 391f7039b1dSLi Dongyang continue; 392f7039b1dSLi Dongyang q = bdev_get_queue(device->bdev); 393f7039b1dSLi Dongyang if (blk_queue_discard(q)) { 394f7039b1dSLi Dongyang num_devices++; 395f7039b1dSLi Dongyang minlen = min((u64)q->limits.discard_granularity, 396f7039b1dSLi Dongyang minlen); 397f7039b1dSLi Dongyang } 398f7039b1dSLi Dongyang } 3991f78160cSXiao Guangrong rcu_read_unlock(); 400f4c697e6SLukas Czerner 401f7039b1dSLi Dongyang if (!num_devices) 402f7039b1dSLi Dongyang return -EOPNOTSUPP; 403f7039b1dSLi Dongyang if (copy_from_user(&range, arg, sizeof(range))) 404f7039b1dSLi Dongyang return -EFAULT; 405e515c18bSLukas Czerner if (range.start > total_bytes || 406e515c18bSLukas Czerner range.len < fs_info->sb->s_blocksize) 407f4c697e6SLukas Czerner return -EINVAL; 408f7039b1dSLi Dongyang 409f4c697e6SLukas Czerner range.len = min(range.len, total_bytes - range.start); 410f7039b1dSLi Dongyang range.minlen = max(range.minlen, minlen); 411815745cfSAl Viro ret = btrfs_trim_fs(fs_info->tree_root, &range); 412f7039b1dSLi Dongyang if (ret < 0) 413f7039b1dSLi Dongyang return ret; 414f7039b1dSLi Dongyang 415f7039b1dSLi Dongyang if (copy_to_user(arg, &range, sizeof(range))) 416f7039b1dSLi Dongyang return -EFAULT; 417f7039b1dSLi Dongyang 418f7039b1dSLi Dongyang return 0; 419f7039b1dSLi Dongyang } 420f7039b1dSLi Dongyang 421dd5f9615SStefan Behrens int btrfs_is_empty_uuid(u8 *uuid) 422dd5f9615SStefan Behrens { 42346e0f66aSChris Mason int i; 42446e0f66aSChris Mason 42546e0f66aSChris Mason for (i = 0; i < BTRFS_UUID_SIZE; i++) { 42646e0f66aSChris Mason if (uuid[i]) 42746e0f66aSChris Mason return 0; 42846e0f66aSChris Mason } 42946e0f66aSChris Mason return 1; 430dd5f9615SStefan Behrens } 431dd5f9615SStefan Behrens 432d5c12070SMiao Xie static noinline int create_subvol(struct inode *dir, 433cb8e7090SChristoph Hellwig struct dentry *dentry, 43472fd032eSSage Weil char *name, int namelen, 4356f72c7e2SArne Jansen u64 *async_transid, 4368696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 437f46b5a66SChristoph Hellwig { 438f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 439f46b5a66SChristoph Hellwig struct btrfs_key key; 440f46b5a66SChristoph Hellwig struct btrfs_root_item root_item; 441f46b5a66SChristoph Hellwig struct btrfs_inode_item *inode_item; 442f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 443d5c12070SMiao Xie struct btrfs_root *root = BTRFS_I(dir)->root; 44476dda93cSYan, Zheng struct btrfs_root *new_root; 445d5c12070SMiao Xie struct btrfs_block_rsv block_rsv; 4468ea05e3aSAlexander Block struct timespec cur_time = CURRENT_TIME; 4475662344bSTsutomu Itoh struct inode *inode; 448f46b5a66SChristoph Hellwig int ret; 449f46b5a66SChristoph Hellwig int err; 450f46b5a66SChristoph Hellwig u64 objectid; 451f46b5a66SChristoph Hellwig u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; 4523de4586cSChris Mason u64 index = 0; 453d5c12070SMiao Xie u64 qgroup_reserved; 4548ea05e3aSAlexander Block uuid_le new_uuid; 455f46b5a66SChristoph Hellwig 456581bb050SLi Zefan ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid); 4572fbe8c8aSAl Viro if (ret) 458a22285a6SYan, Zheng return ret; 4596a912213SJosef Bacik 460e09fe2d2SQu Wenruo /* 461e09fe2d2SQu Wenruo * Don't create subvolume whose level is not zero. Or qgroup will be 462e09fe2d2SQu Wenruo * screwed up since it assume subvolme qgroup's level to be 0. 463e09fe2d2SQu Wenruo */ 464e09fe2d2SQu Wenruo if (btrfs_qgroup_level(objectid)) 465e09fe2d2SQu Wenruo return -ENOSPC; 466e09fe2d2SQu Wenruo 467d5c12070SMiao Xie btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); 4689ed74f2dSJosef Bacik /* 469d5c12070SMiao Xie * The same as the snapshot creation, please see the comment 470d5c12070SMiao Xie * of create_snapshot(). 4719ed74f2dSJosef Bacik */ 472d5c12070SMiao Xie ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 473dd5f9615SStefan Behrens 8, &qgroup_reserved, false); 474d5c12070SMiao Xie if (ret) 475d5c12070SMiao Xie return ret; 476f46b5a66SChristoph Hellwig 477d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 478d5c12070SMiao Xie if (IS_ERR(trans)) { 479d5c12070SMiao Xie ret = PTR_ERR(trans); 480de6e8200SLiu Bo btrfs_subvolume_release_metadata(root, &block_rsv, 481de6e8200SLiu Bo qgroup_reserved); 482de6e8200SLiu Bo return ret; 483d5c12070SMiao Xie } 484d5c12070SMiao Xie trans->block_rsv = &block_rsv; 485d5c12070SMiao Xie trans->bytes_reserved = block_rsv.size; 486f46b5a66SChristoph Hellwig 4878696c533SMiao Xie ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit); 4886f72c7e2SArne Jansen if (ret) 4896f72c7e2SArne Jansen goto fail; 4906f72c7e2SArne Jansen 4914d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); 4928e8a1e31SJosef Bacik if (IS_ERR(leaf)) { 4938e8a1e31SJosef Bacik ret = PTR_ERR(leaf); 4948e8a1e31SJosef Bacik goto fail; 4958e8a1e31SJosef Bacik } 496f46b5a66SChristoph Hellwig 4975d4f98a2SYan Zheng memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); 498f46b5a66SChristoph Hellwig btrfs_set_header_bytenr(leaf, leaf->start); 499f46b5a66SChristoph Hellwig btrfs_set_header_generation(leaf, trans->transid); 5005d4f98a2SYan Zheng btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 501f46b5a66SChristoph Hellwig btrfs_set_header_owner(leaf, objectid); 502f46b5a66SChristoph Hellwig 5030a4e5586SRoss Kirk write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(), 504f46b5a66SChristoph Hellwig BTRFS_FSID_SIZE); 5055d4f98a2SYan Zheng write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid, 506b308bc2fSGeert Uytterhoeven btrfs_header_chunk_tree_uuid(leaf), 5075d4f98a2SYan Zheng BTRFS_UUID_SIZE); 508f46b5a66SChristoph Hellwig btrfs_mark_buffer_dirty(leaf); 509f46b5a66SChristoph Hellwig 5108ea05e3aSAlexander Block memset(&root_item, 0, sizeof(root_item)); 5118ea05e3aSAlexander Block 512f46b5a66SChristoph Hellwig inode_item = &root_item.inode; 5133cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 5143cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 5153cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 516707e8a07SDavid Sterba btrfs_set_stack_inode_nbytes(inode_item, root->nodesize); 5173cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 518f46b5a66SChristoph Hellwig 5193cae210fSQu Wenruo btrfs_set_root_flags(&root_item, 0); 5203cae210fSQu Wenruo btrfs_set_root_limit(&root_item, 0); 5213cae210fSQu Wenruo btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT); 52208fe4db1SLi Zefan 523f46b5a66SChristoph Hellwig btrfs_set_root_bytenr(&root_item, leaf->start); 52484234f3aSYan Zheng btrfs_set_root_generation(&root_item, trans->transid); 525f46b5a66SChristoph Hellwig btrfs_set_root_level(&root_item, 0); 526f46b5a66SChristoph Hellwig btrfs_set_root_refs(&root_item, 1); 52786b9f2ecSYan, Zheng btrfs_set_root_used(&root_item, leaf->len); 52880ff3856SYan Zheng btrfs_set_root_last_snapshot(&root_item, 0); 529f46b5a66SChristoph Hellwig 5308ea05e3aSAlexander Block btrfs_set_root_generation_v2(&root_item, 5318ea05e3aSAlexander Block btrfs_root_generation(&root_item)); 5328ea05e3aSAlexander Block uuid_le_gen(&new_uuid); 5338ea05e3aSAlexander Block memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE); 5343cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec); 5353cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec); 5368ea05e3aSAlexander Block root_item.ctime = root_item.otime; 5378ea05e3aSAlexander Block btrfs_set_root_ctransid(&root_item, trans->transid); 5388ea05e3aSAlexander Block btrfs_set_root_otransid(&root_item, trans->transid); 539f46b5a66SChristoph Hellwig 540925baeddSChris Mason btrfs_tree_unlock(leaf); 541f46b5a66SChristoph Hellwig free_extent_buffer(leaf); 542f46b5a66SChristoph Hellwig leaf = NULL; 543f46b5a66SChristoph Hellwig 544f46b5a66SChristoph Hellwig btrfs_set_root_dirid(&root_item, new_dirid); 545f46b5a66SChristoph Hellwig 546f46b5a66SChristoph Hellwig key.objectid = objectid; 5475d4f98a2SYan Zheng key.offset = 0; 548962a298fSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 549f46b5a66SChristoph Hellwig ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, 550f46b5a66SChristoph Hellwig &root_item); 551f46b5a66SChristoph Hellwig if (ret) 552f46b5a66SChristoph Hellwig goto fail; 553f46b5a66SChristoph Hellwig 55476dda93cSYan, Zheng key.offset = (u64)-1; 55576dda93cSYan, Zheng new_root = btrfs_read_fs_root_no_name(root->fs_info, &key); 55679787eaaSJeff Mahoney if (IS_ERR(new_root)) { 55779787eaaSJeff Mahoney ret = PTR_ERR(new_root); 5586d13f549SDavid Sterba btrfs_abort_transaction(trans, root, ret); 55979787eaaSJeff Mahoney goto fail; 56079787eaaSJeff Mahoney } 56176dda93cSYan, Zheng 56276dda93cSYan, Zheng btrfs_record_root_in_trans(trans, new_root); 56376dda93cSYan, Zheng 56463541927SFilipe David Borba Manana ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid); 565ce598979SMark Fasheh if (ret) { 566ce598979SMark Fasheh /* We potentially lose an unused inode item here */ 56779787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, ret); 568ce598979SMark Fasheh goto fail; 569ce598979SMark Fasheh } 570ce598979SMark Fasheh 571f46b5a66SChristoph Hellwig /* 572f46b5a66SChristoph Hellwig * insert the directory item 573f46b5a66SChristoph Hellwig */ 5743de4586cSChris Mason ret = btrfs_set_inode_index(dir, &index); 57579787eaaSJeff Mahoney if (ret) { 57679787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, ret); 57779787eaaSJeff Mahoney goto fail; 57879787eaaSJeff Mahoney } 5793de4586cSChris Mason 5803de4586cSChris Mason ret = btrfs_insert_dir_item(trans, root, 58116cdcec7SMiao Xie name, namelen, dir, &key, 5823de4586cSChris Mason BTRFS_FT_DIR, index); 58379787eaaSJeff Mahoney if (ret) { 58479787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, ret); 585f46b5a66SChristoph Hellwig goto fail; 58679787eaaSJeff Mahoney } 5870660b5afSChris Mason 58852c26179SYan Zheng btrfs_i_size_write(dir, dir->i_size + namelen * 2); 58952c26179SYan Zheng ret = btrfs_update_inode(trans, root, dir); 59052c26179SYan Zheng BUG_ON(ret); 59152c26179SYan Zheng 5920660b5afSChris Mason ret = btrfs_add_root_ref(trans, root->fs_info->tree_root, 5934df27c4dSYan, Zheng objectid, root->root_key.objectid, 59433345d01SLi Zefan btrfs_ino(dir), index, name, namelen); 5950660b5afSChris Mason BUG_ON(ret); 5960660b5afSChris Mason 597dd5f9615SStefan Behrens ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root, 598dd5f9615SStefan Behrens root_item.uuid, BTRFS_UUID_KEY_SUBVOL, 599dd5f9615SStefan Behrens objectid); 600dd5f9615SStefan Behrens if (ret) 601dd5f9615SStefan Behrens btrfs_abort_transaction(trans, root, ret); 602dd5f9615SStefan Behrens 603f46b5a66SChristoph Hellwig fail: 604d5c12070SMiao Xie trans->block_rsv = NULL; 605d5c12070SMiao Xie trans->bytes_reserved = 0; 606de6e8200SLiu Bo btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved); 607de6e8200SLiu Bo 60872fd032eSSage Weil if (async_transid) { 60972fd032eSSage Weil *async_transid = trans->transid; 61072fd032eSSage Weil err = btrfs_commit_transaction_async(trans, root, 1); 61100d71c9cSMiao Xie if (err) 61200d71c9cSMiao Xie err = btrfs_commit_transaction(trans, root); 61372fd032eSSage Weil } else { 61476dda93cSYan, Zheng err = btrfs_commit_transaction(trans, root); 61572fd032eSSage Weil } 616f46b5a66SChristoph Hellwig if (err && !ret) 617f46b5a66SChristoph Hellwig ret = err; 6181a65e24bSChris Mason 6195662344bSTsutomu Itoh if (!ret) { 6205662344bSTsutomu Itoh inode = btrfs_lookup_dentry(dir, dentry); 621de6e8200SLiu Bo if (IS_ERR(inode)) 622de6e8200SLiu Bo return PTR_ERR(inode); 6235662344bSTsutomu Itoh d_instantiate(dentry, inode); 6245662344bSTsutomu Itoh } 625f46b5a66SChristoph Hellwig return ret; 626f46b5a66SChristoph Hellwig } 627f46b5a66SChristoph Hellwig 6289ea24bbeSFilipe Manana static void btrfs_wait_for_no_snapshoting_writes(struct btrfs_root *root) 6298257b2dcSMiao Xie { 6308257b2dcSMiao Xie s64 writers; 6318257b2dcSMiao Xie DEFINE_WAIT(wait); 6328257b2dcSMiao Xie 6338257b2dcSMiao Xie do { 6348257b2dcSMiao Xie prepare_to_wait(&root->subv_writers->wait, &wait, 6358257b2dcSMiao Xie TASK_UNINTERRUPTIBLE); 6368257b2dcSMiao Xie 6378257b2dcSMiao Xie writers = percpu_counter_sum(&root->subv_writers->counter); 6388257b2dcSMiao Xie if (writers) 6398257b2dcSMiao Xie schedule(); 6408257b2dcSMiao Xie 6418257b2dcSMiao Xie finish_wait(&root->subv_writers->wait, &wait); 6428257b2dcSMiao Xie } while (writers); 6438257b2dcSMiao Xie } 6448257b2dcSMiao Xie 645e9662f70SMiao Xie static int create_snapshot(struct btrfs_root *root, struct inode *dir, 646e9662f70SMiao Xie struct dentry *dentry, char *name, int namelen, 647e9662f70SMiao Xie u64 *async_transid, bool readonly, 648e9662f70SMiao Xie struct btrfs_qgroup_inherit *inherit) 649f46b5a66SChristoph Hellwig { 6502e4bfab9SYan, Zheng struct inode *inode; 651f46b5a66SChristoph Hellwig struct btrfs_pending_snapshot *pending_snapshot; 652f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 6532e4bfab9SYan, Zheng int ret; 654f46b5a66SChristoph Hellwig 65527cdeb70SMiao Xie if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state)) 656f46b5a66SChristoph Hellwig return -EINVAL; 657f46b5a66SChristoph Hellwig 6588257b2dcSMiao Xie atomic_inc(&root->will_be_snapshoted); 6594e857c58SPeter Zijlstra smp_mb__after_atomic(); 6609ea24bbeSFilipe Manana btrfs_wait_for_no_snapshoting_writes(root); 6618257b2dcSMiao Xie 6626a03843dSMiao Xie ret = btrfs_start_delalloc_inodes(root, 0); 6636a03843dSMiao Xie if (ret) 6648257b2dcSMiao Xie goto out; 6656a03843dSMiao Xie 666b0244199SMiao Xie btrfs_wait_ordered_extents(root, -1); 6676a03843dSMiao Xie 668a22285a6SYan, Zheng pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS); 6698257b2dcSMiao Xie if (!pending_snapshot) { 6708257b2dcSMiao Xie ret = -ENOMEM; 6718257b2dcSMiao Xie goto out; 6728257b2dcSMiao Xie } 673a22285a6SYan, Zheng 67466d8f3ddSMiao Xie btrfs_init_block_rsv(&pending_snapshot->block_rsv, 67566d8f3ddSMiao Xie BTRFS_BLOCK_RSV_TEMP); 676d5c12070SMiao Xie /* 677d5c12070SMiao Xie * 1 - parent dir inode 678d5c12070SMiao Xie * 2 - dir entries 679d5c12070SMiao Xie * 1 - root item 680d5c12070SMiao Xie * 2 - root ref/backref 681d5c12070SMiao Xie * 1 - root of snapshot 682dd5f9615SStefan Behrens * 1 - UUID item 683d5c12070SMiao Xie */ 684d5c12070SMiao Xie ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, 685dd5f9615SStefan Behrens &pending_snapshot->block_rsv, 8, 686ee3441b4SJeff Mahoney &pending_snapshot->qgroup_reserved, 687ee3441b4SJeff Mahoney false); 688d5c12070SMiao Xie if (ret) 6898257b2dcSMiao Xie goto free; 690d5c12070SMiao Xie 691a22285a6SYan, Zheng pending_snapshot->dentry = dentry; 692a22285a6SYan, Zheng pending_snapshot->root = root; 693b83cc969SLi Zefan pending_snapshot->readonly = readonly; 694e9662f70SMiao Xie pending_snapshot->dir = dir; 6958696c533SMiao Xie pending_snapshot->inherit = inherit; 696a22285a6SYan, Zheng 697d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 698a22285a6SYan, Zheng if (IS_ERR(trans)) { 699a22285a6SYan, Zheng ret = PTR_ERR(trans); 700a22285a6SYan, Zheng goto fail; 701a22285a6SYan, Zheng } 702a22285a6SYan, Zheng 7038351583eSJosef Bacik spin_lock(&root->fs_info->trans_lock); 704a22285a6SYan, Zheng list_add(&pending_snapshot->list, 705a22285a6SYan, Zheng &trans->transaction->pending_snapshots); 7068351583eSJosef Bacik spin_unlock(&root->fs_info->trans_lock); 70772fd032eSSage Weil if (async_transid) { 70872fd032eSSage Weil *async_transid = trans->transid; 70972fd032eSSage Weil ret = btrfs_commit_transaction_async(trans, 71072fd032eSSage Weil root->fs_info->extent_root, 1); 71100d71c9cSMiao Xie if (ret) 71200d71c9cSMiao Xie ret = btrfs_commit_transaction(trans, root); 71372fd032eSSage Weil } else { 71472fd032eSSage Weil ret = btrfs_commit_transaction(trans, 71572fd032eSSage Weil root->fs_info->extent_root); 71672fd032eSSage Weil } 717aec8030aSMiao Xie if (ret) 718c37b2b62SJosef Bacik goto fail; 719a22285a6SYan, Zheng 720a22285a6SYan, Zheng ret = pending_snapshot->error; 721f46b5a66SChristoph Hellwig if (ret) 7222e4bfab9SYan, Zheng goto fail; 723f46b5a66SChristoph Hellwig 724d3797308SChris Mason ret = btrfs_orphan_cleanup(pending_snapshot->snap); 725d3797308SChris Mason if (ret) 726d3797308SChris Mason goto fail; 727d3797308SChris Mason 7282b0143b5SDavid Howells inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry); 7292e4bfab9SYan, Zheng if (IS_ERR(inode)) { 7302e4bfab9SYan, Zheng ret = PTR_ERR(inode); 7312e4bfab9SYan, Zheng goto fail; 7322e4bfab9SYan, Zheng } 7335662344bSTsutomu Itoh 7342e4bfab9SYan, Zheng d_instantiate(dentry, inode); 7352e4bfab9SYan, Zheng ret = 0; 7362e4bfab9SYan, Zheng fail: 737d5c12070SMiao Xie btrfs_subvolume_release_metadata(BTRFS_I(dir)->root, 738d5c12070SMiao Xie &pending_snapshot->block_rsv, 739d5c12070SMiao Xie pending_snapshot->qgroup_reserved); 7408257b2dcSMiao Xie free: 741a22285a6SYan, Zheng kfree(pending_snapshot); 7428257b2dcSMiao Xie out: 7439ea24bbeSFilipe Manana if (atomic_dec_and_test(&root->will_be_snapshoted)) 7449ea24bbeSFilipe Manana wake_up_atomic_t(&root->will_be_snapshoted); 745f46b5a66SChristoph Hellwig return ret; 746f46b5a66SChristoph Hellwig } 747f46b5a66SChristoph Hellwig 7484260f7c7SSage Weil /* copy of may_delete in fs/namei.c() 7494260f7c7SSage Weil * Check whether we can remove a link victim from directory dir, check 7504260f7c7SSage Weil * whether the type of victim is right. 7514260f7c7SSage Weil * 1. We can't do it if dir is read-only (done in permission()) 7524260f7c7SSage Weil * 2. We should have write and exec permissions on dir 7534260f7c7SSage Weil * 3. We can't remove anything from append-only dir 7544260f7c7SSage Weil * 4. We can't do anything with immutable dir (done in permission()) 7554260f7c7SSage Weil * 5. If the sticky bit on dir is set we should either 7564260f7c7SSage Weil * a. be owner of dir, or 7574260f7c7SSage Weil * b. be owner of victim, or 7584260f7c7SSage Weil * c. have CAP_FOWNER capability 7594260f7c7SSage Weil * 6. If the victim is append-only or immutable we can't do antyhing with 7604260f7c7SSage Weil * links pointing to it. 7614260f7c7SSage Weil * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. 7624260f7c7SSage Weil * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. 7634260f7c7SSage Weil * 9. We can't remove a root or mountpoint. 7644260f7c7SSage Weil * 10. We don't allow removal of NFS sillyrenamed files; it's handled by 7654260f7c7SSage Weil * nfs_async_unlink(). 7664260f7c7SSage Weil */ 7674260f7c7SSage Weil 7684260f7c7SSage Weil static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir) 7694260f7c7SSage Weil { 7704260f7c7SSage Weil int error; 7714260f7c7SSage Weil 7722b0143b5SDavid Howells if (d_really_is_negative(victim)) 7734260f7c7SSage Weil return -ENOENT; 7744260f7c7SSage Weil 7752b0143b5SDavid Howells BUG_ON(d_inode(victim->d_parent) != dir); 7764fa6b5ecSJeff Layton audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); 7774260f7c7SSage Weil 7784260f7c7SSage Weil error = inode_permission(dir, MAY_WRITE | MAY_EXEC); 7794260f7c7SSage Weil if (error) 7804260f7c7SSage Weil return error; 7814260f7c7SSage Weil if (IS_APPEND(dir)) 7824260f7c7SSage Weil return -EPERM; 7832b0143b5SDavid Howells if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) || 7842b0143b5SDavid Howells IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim))) 7854260f7c7SSage Weil return -EPERM; 7864260f7c7SSage Weil if (isdir) { 787e36cb0b8SDavid Howells if (!d_is_dir(victim)) 7884260f7c7SSage Weil return -ENOTDIR; 7894260f7c7SSage Weil if (IS_ROOT(victim)) 7904260f7c7SSage Weil return -EBUSY; 791e36cb0b8SDavid Howells } else if (d_is_dir(victim)) 7924260f7c7SSage Weil return -EISDIR; 7934260f7c7SSage Weil if (IS_DEADDIR(dir)) 7944260f7c7SSage Weil return -ENOENT; 7954260f7c7SSage Weil if (victim->d_flags & DCACHE_NFSFS_RENAMED) 7964260f7c7SSage Weil return -EBUSY; 7974260f7c7SSage Weil return 0; 7984260f7c7SSage Weil } 7994260f7c7SSage Weil 800cb8e7090SChristoph Hellwig /* copy of may_create in fs/namei.c() */ 801cb8e7090SChristoph Hellwig static inline int btrfs_may_create(struct inode *dir, struct dentry *child) 802cb8e7090SChristoph Hellwig { 8032b0143b5SDavid Howells if (d_really_is_positive(child)) 804cb8e7090SChristoph Hellwig return -EEXIST; 805cb8e7090SChristoph Hellwig if (IS_DEADDIR(dir)) 806cb8e7090SChristoph Hellwig return -ENOENT; 807cb8e7090SChristoph Hellwig return inode_permission(dir, MAY_WRITE | MAY_EXEC); 808cb8e7090SChristoph Hellwig } 809cb8e7090SChristoph Hellwig 810cb8e7090SChristoph Hellwig /* 811cb8e7090SChristoph Hellwig * Create a new subvolume below @parent. This is largely modeled after 812cb8e7090SChristoph Hellwig * sys_mkdirat and vfs_mkdir, but we only do a single component lookup 813cb8e7090SChristoph Hellwig * inside this filesystem so it's quite a bit simpler. 814cb8e7090SChristoph Hellwig */ 81576dda93cSYan, Zheng static noinline int btrfs_mksubvol(struct path *parent, 81676dda93cSYan, Zheng char *name, int namelen, 81772fd032eSSage Weil struct btrfs_root *snap_src, 8186f72c7e2SArne Jansen u64 *async_transid, bool readonly, 8198696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 820cb8e7090SChristoph Hellwig { 8212b0143b5SDavid Howells struct inode *dir = d_inode(parent->dentry); 822cb8e7090SChristoph Hellwig struct dentry *dentry; 823cb8e7090SChristoph Hellwig int error; 824cb8e7090SChristoph Hellwig 8255c50c9b8SDavid Sterba error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT); 8265c50c9b8SDavid Sterba if (error == -EINTR) 8275c50c9b8SDavid Sterba return error; 828cb8e7090SChristoph Hellwig 829cb8e7090SChristoph Hellwig dentry = lookup_one_len(name, parent->dentry, namelen); 830cb8e7090SChristoph Hellwig error = PTR_ERR(dentry); 831cb8e7090SChristoph Hellwig if (IS_ERR(dentry)) 832cb8e7090SChristoph Hellwig goto out_unlock; 833cb8e7090SChristoph Hellwig 834cb8e7090SChristoph Hellwig error = -EEXIST; 8352b0143b5SDavid Howells if (d_really_is_positive(dentry)) 836cb8e7090SChristoph Hellwig goto out_dput; 837cb8e7090SChristoph Hellwig 83876dda93cSYan, Zheng error = btrfs_may_create(dir, dentry); 839cb8e7090SChristoph Hellwig if (error) 840a874a63eSLiu Bo goto out_dput; 841cb8e7090SChristoph Hellwig 8429c52057cSChris Mason /* 8439c52057cSChris Mason * even if this name doesn't exist, we may get hash collisions. 8449c52057cSChris Mason * check for them now when we can safely fail 8459c52057cSChris Mason */ 8469c52057cSChris Mason error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, 8479c52057cSChris Mason dir->i_ino, name, 8489c52057cSChris Mason namelen); 8499c52057cSChris Mason if (error) 8509c52057cSChris Mason goto out_dput; 8519c52057cSChris Mason 85276dda93cSYan, Zheng down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem); 85376dda93cSYan, Zheng 85476dda93cSYan, Zheng if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) 85576dda93cSYan, Zheng goto out_up_read; 85676dda93cSYan, Zheng 8573de4586cSChris Mason if (snap_src) { 858e9662f70SMiao Xie error = create_snapshot(snap_src, dir, dentry, name, namelen, 8596f72c7e2SArne Jansen async_transid, readonly, inherit); 8603de4586cSChris Mason } else { 861d5c12070SMiao Xie error = create_subvol(dir, dentry, name, namelen, 862d5c12070SMiao Xie async_transid, inherit); 8633de4586cSChris Mason } 86476dda93cSYan, Zheng if (!error) 86576dda93cSYan, Zheng fsnotify_mkdir(dir, dentry); 86676dda93cSYan, Zheng out_up_read: 86776dda93cSYan, Zheng up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem); 868cb8e7090SChristoph Hellwig out_dput: 869cb8e7090SChristoph Hellwig dput(dentry); 870cb8e7090SChristoph Hellwig out_unlock: 87176dda93cSYan, Zheng mutex_unlock(&dir->i_mutex); 872cb8e7090SChristoph Hellwig return error; 873cb8e7090SChristoph Hellwig } 874cb8e7090SChristoph Hellwig 8754cb5300bSChris Mason /* 8764cb5300bSChris Mason * When we're defragging a range, we don't want to kick it off again 8774cb5300bSChris Mason * if it is really just waiting for delalloc to send it down. 8784cb5300bSChris Mason * If we find a nice big extent or delalloc range for the bytes in the 8794cb5300bSChris Mason * file you want to defrag, we return 0 to let you know to skip this 8804cb5300bSChris Mason * part of the file 8814cb5300bSChris Mason */ 882aab110abSDavid Sterba static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh) 8834cb5300bSChris Mason { 8844cb5300bSChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 8854cb5300bSChris Mason struct extent_map *em = NULL; 8864cb5300bSChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 8874cb5300bSChris Mason u64 end; 8884cb5300bSChris Mason 8894cb5300bSChris Mason read_lock(&em_tree->lock); 8904cb5300bSChris Mason em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE); 8914cb5300bSChris Mason read_unlock(&em_tree->lock); 8924cb5300bSChris Mason 8934cb5300bSChris Mason if (em) { 8944cb5300bSChris Mason end = extent_map_end(em); 8954cb5300bSChris Mason free_extent_map(em); 8964cb5300bSChris Mason if (end - offset > thresh) 8974cb5300bSChris Mason return 0; 8984cb5300bSChris Mason } 8994cb5300bSChris Mason /* if we already have a nice delalloc here, just stop */ 9004cb5300bSChris Mason thresh /= 2; 9014cb5300bSChris Mason end = count_range_bits(io_tree, &offset, offset + thresh, 9024cb5300bSChris Mason thresh, EXTENT_DELALLOC, 1); 9034cb5300bSChris Mason if (end >= thresh) 9044cb5300bSChris Mason return 0; 9054cb5300bSChris Mason return 1; 9064cb5300bSChris Mason } 9074cb5300bSChris Mason 9084cb5300bSChris Mason /* 9094cb5300bSChris Mason * helper function to walk through a file and find extents 9104cb5300bSChris Mason * newer than a specific transid, and smaller than thresh. 9114cb5300bSChris Mason * 9124cb5300bSChris Mason * This is used by the defragging code to find new and small 9134cb5300bSChris Mason * extents 9144cb5300bSChris Mason */ 9154cb5300bSChris Mason static int find_new_extents(struct btrfs_root *root, 9164cb5300bSChris Mason struct inode *inode, u64 newer_than, 917aab110abSDavid Sterba u64 *off, u32 thresh) 9184cb5300bSChris Mason { 9194cb5300bSChris Mason struct btrfs_path *path; 9204cb5300bSChris Mason struct btrfs_key min_key; 9214cb5300bSChris Mason struct extent_buffer *leaf; 9224cb5300bSChris Mason struct btrfs_file_extent_item *extent; 9234cb5300bSChris Mason int type; 9244cb5300bSChris Mason int ret; 925a4689d2bSDavid Sterba u64 ino = btrfs_ino(inode); 9264cb5300bSChris Mason 9274cb5300bSChris Mason path = btrfs_alloc_path(); 9284cb5300bSChris Mason if (!path) 9294cb5300bSChris Mason return -ENOMEM; 9304cb5300bSChris Mason 931a4689d2bSDavid Sterba min_key.objectid = ino; 9324cb5300bSChris Mason min_key.type = BTRFS_EXTENT_DATA_KEY; 9334cb5300bSChris Mason min_key.offset = *off; 9344cb5300bSChris Mason 9354cb5300bSChris Mason while (1) { 9366174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &min_key, path, newer_than); 9374cb5300bSChris Mason if (ret != 0) 9384cb5300bSChris Mason goto none; 939f094c9bdSFilipe Manana process_slot: 940a4689d2bSDavid Sterba if (min_key.objectid != ino) 9414cb5300bSChris Mason goto none; 9424cb5300bSChris Mason if (min_key.type != BTRFS_EXTENT_DATA_KEY) 9434cb5300bSChris Mason goto none; 9444cb5300bSChris Mason 9454cb5300bSChris Mason leaf = path->nodes[0]; 9464cb5300bSChris Mason extent = btrfs_item_ptr(leaf, path->slots[0], 9474cb5300bSChris Mason struct btrfs_file_extent_item); 9484cb5300bSChris Mason 9494cb5300bSChris Mason type = btrfs_file_extent_type(leaf, extent); 9504cb5300bSChris Mason if (type == BTRFS_FILE_EXTENT_REG && 9514cb5300bSChris Mason btrfs_file_extent_num_bytes(leaf, extent) < thresh && 9524cb5300bSChris Mason check_defrag_in_cache(inode, min_key.offset, thresh)) { 9534cb5300bSChris Mason *off = min_key.offset; 9544cb5300bSChris Mason btrfs_free_path(path); 9554cb5300bSChris Mason return 0; 9564cb5300bSChris Mason } 9574cb5300bSChris Mason 958f094c9bdSFilipe Manana path->slots[0]++; 959f094c9bdSFilipe Manana if (path->slots[0] < btrfs_header_nritems(leaf)) { 960f094c9bdSFilipe Manana btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]); 961f094c9bdSFilipe Manana goto process_slot; 962f094c9bdSFilipe Manana } 963f094c9bdSFilipe Manana 9644cb5300bSChris Mason if (min_key.offset == (u64)-1) 9654cb5300bSChris Mason goto none; 9664cb5300bSChris Mason 9674cb5300bSChris Mason min_key.offset++; 9684cb5300bSChris Mason btrfs_release_path(path); 9694cb5300bSChris Mason } 9704cb5300bSChris Mason none: 9714cb5300bSChris Mason btrfs_free_path(path); 9724cb5300bSChris Mason return -ENOENT; 9734cb5300bSChris Mason } 9744cb5300bSChris Mason 9756c282eb4SLi Zefan static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start) 97617ce6ef8SLiu Bo { 97717ce6ef8SLiu Bo struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 978940100a4SChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 9796c282eb4SLi Zefan struct extent_map *em; 9806c282eb4SLi Zefan u64 len = PAGE_CACHE_SIZE; 981940100a4SChris Mason 982940100a4SChris Mason /* 983940100a4SChris Mason * hopefully we have this extent in the tree already, try without 984940100a4SChris Mason * the full extent lock 985940100a4SChris Mason */ 986940100a4SChris Mason read_lock(&em_tree->lock); 987940100a4SChris Mason em = lookup_extent_mapping(em_tree, start, len); 988940100a4SChris Mason read_unlock(&em_tree->lock); 989940100a4SChris Mason 990940100a4SChris Mason if (!em) { 991308d9800SFilipe Manana struct extent_state *cached = NULL; 992308d9800SFilipe Manana u64 end = start + len - 1; 993308d9800SFilipe Manana 994940100a4SChris Mason /* get the big lock and read metadata off disk */ 995308d9800SFilipe Manana lock_extent_bits(io_tree, start, end, 0, &cached); 996940100a4SChris Mason em = btrfs_get_extent(inode, NULL, 0, start, len, 0); 997308d9800SFilipe Manana unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS); 998940100a4SChris Mason 9996cf8bfbfSDan Carpenter if (IS_ERR(em)) 10006c282eb4SLi Zefan return NULL; 1001940100a4SChris Mason } 1002940100a4SChris Mason 10036c282eb4SLi Zefan return em; 10046c282eb4SLi Zefan } 10056c282eb4SLi Zefan 10066c282eb4SLi Zefan static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em) 10076c282eb4SLi Zefan { 10086c282eb4SLi Zefan struct extent_map *next; 10096c282eb4SLi Zefan bool ret = true; 10106c282eb4SLi Zefan 10116c282eb4SLi Zefan /* this is the last extent */ 10126c282eb4SLi Zefan if (em->start + em->len >= i_size_read(inode)) 10136c282eb4SLi Zefan return false; 10146c282eb4SLi Zefan 10156c282eb4SLi Zefan next = defrag_lookup_extent(inode, em->start + em->len); 1016e9512d72SChris Mason if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE) 1017e9512d72SChris Mason ret = false; 1018e9512d72SChris Mason else if ((em->block_start + em->block_len == next->block_start) && 1019e9512d72SChris Mason (em->block_len > 128 * 1024 && next->block_len > 128 * 1024)) 10206c282eb4SLi Zefan ret = false; 10216c282eb4SLi Zefan 10226c282eb4SLi Zefan free_extent_map(next); 10236c282eb4SLi Zefan return ret; 10246c282eb4SLi Zefan } 10256c282eb4SLi Zefan 1026aab110abSDavid Sterba static int should_defrag_range(struct inode *inode, u64 start, u32 thresh, 1027a43a2111SAndrew Mahone u64 *last_len, u64 *skip, u64 *defrag_end, 1028a43a2111SAndrew Mahone int compress) 10296c282eb4SLi Zefan { 10306c282eb4SLi Zefan struct extent_map *em; 10316c282eb4SLi Zefan int ret = 1; 10326c282eb4SLi Zefan bool next_mergeable = true; 10336c282eb4SLi Zefan 10346c282eb4SLi Zefan /* 10356c282eb4SLi Zefan * make sure that once we start defragging an extent, we keep on 10366c282eb4SLi Zefan * defragging it 10376c282eb4SLi Zefan */ 10386c282eb4SLi Zefan if (start < *defrag_end) 10396c282eb4SLi Zefan return 1; 10406c282eb4SLi Zefan 10416c282eb4SLi Zefan *skip = 0; 10426c282eb4SLi Zefan 10436c282eb4SLi Zefan em = defrag_lookup_extent(inode, start); 10446c282eb4SLi Zefan if (!em) 10456c282eb4SLi Zefan return 0; 10466c282eb4SLi Zefan 1047940100a4SChris Mason /* this will cover holes, and inline extents */ 104817ce6ef8SLiu Bo if (em->block_start >= EXTENT_MAP_LAST_BYTE) { 1049940100a4SChris Mason ret = 0; 105017ce6ef8SLiu Bo goto out; 105117ce6ef8SLiu Bo } 105217ce6ef8SLiu Bo 10536c282eb4SLi Zefan next_mergeable = defrag_check_next_extent(inode, em); 1054940100a4SChris Mason /* 10556c282eb4SLi Zefan * we hit a real extent, if it is big or the next extent is not a 10566c282eb4SLi Zefan * real extent, don't bother defragging it 1057940100a4SChris Mason */ 1058a43a2111SAndrew Mahone if (!compress && (*last_len == 0 || *last_len >= thresh) && 10596c282eb4SLi Zefan (em->len >= thresh || !next_mergeable)) 1060940100a4SChris Mason ret = 0; 106117ce6ef8SLiu Bo out: 1062940100a4SChris Mason /* 1063940100a4SChris Mason * last_len ends up being a counter of how many bytes we've defragged. 1064940100a4SChris Mason * every time we choose not to defrag an extent, we reset *last_len 1065940100a4SChris Mason * so that the next tiny extent will force a defrag. 1066940100a4SChris Mason * 1067940100a4SChris Mason * The end result of this is that tiny extents before a single big 1068940100a4SChris Mason * extent will force at least part of that big extent to be defragged. 1069940100a4SChris Mason */ 1070940100a4SChris Mason if (ret) { 1071940100a4SChris Mason *defrag_end = extent_map_end(em); 1072940100a4SChris Mason } else { 1073940100a4SChris Mason *last_len = 0; 1074940100a4SChris Mason *skip = extent_map_end(em); 1075940100a4SChris Mason *defrag_end = 0; 1076940100a4SChris Mason } 1077940100a4SChris Mason 1078940100a4SChris Mason free_extent_map(em); 1079940100a4SChris Mason return ret; 1080940100a4SChris Mason } 1081940100a4SChris Mason 10824cb5300bSChris Mason /* 10834cb5300bSChris Mason * it doesn't do much good to defrag one or two pages 10844cb5300bSChris Mason * at a time. This pulls in a nice chunk of pages 10854cb5300bSChris Mason * to COW and defrag. 10864cb5300bSChris Mason * 10874cb5300bSChris Mason * It also makes sure the delalloc code has enough 10884cb5300bSChris Mason * dirty data to avoid making new small extents as part 10894cb5300bSChris Mason * of the defrag 10904cb5300bSChris Mason * 10914cb5300bSChris Mason * It's a good idea to start RA on this range 10924cb5300bSChris Mason * before calling this. 10934cb5300bSChris Mason */ 10944cb5300bSChris Mason static int cluster_pages_for_defrag(struct inode *inode, 10954cb5300bSChris Mason struct page **pages, 10964cb5300bSChris Mason unsigned long start_index, 1097c41570c9SJustin Maggard unsigned long num_pages) 1098f46b5a66SChristoph Hellwig { 10994cb5300bSChris Mason unsigned long file_end; 11004cb5300bSChris Mason u64 isize = i_size_read(inode); 1101f46b5a66SChristoph Hellwig u64 page_start; 1102f46b5a66SChristoph Hellwig u64 page_end; 11031f12bd06SLiu Bo u64 page_cnt; 11044cb5300bSChris Mason int ret; 11054cb5300bSChris Mason int i; 11064cb5300bSChris Mason int i_done; 11074cb5300bSChris Mason struct btrfs_ordered_extent *ordered; 11084cb5300bSChris Mason struct extent_state *cached_state = NULL; 1109600a45e1SMiao Xie struct extent_io_tree *tree; 11103b16a4e3SJosef Bacik gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping); 11114cb5300bSChris Mason 11124cb5300bSChris Mason file_end = (isize - 1) >> PAGE_CACHE_SHIFT; 11131f12bd06SLiu Bo if (!isize || start_index > file_end) 11141f12bd06SLiu Bo return 0; 11151f12bd06SLiu Bo 11161f12bd06SLiu Bo page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1); 11174cb5300bSChris Mason 11184cb5300bSChris Mason ret = btrfs_delalloc_reserve_space(inode, 11191f12bd06SLiu Bo page_cnt << PAGE_CACHE_SHIFT); 11204cb5300bSChris Mason if (ret) 11214cb5300bSChris Mason return ret; 11224cb5300bSChris Mason i_done = 0; 1123600a45e1SMiao Xie tree = &BTRFS_I(inode)->io_tree; 11244cb5300bSChris Mason 11254cb5300bSChris Mason /* step one, lock all the pages */ 11261f12bd06SLiu Bo for (i = 0; i < page_cnt; i++) { 11274cb5300bSChris Mason struct page *page; 1128600a45e1SMiao Xie again: 1129a94733d0SJosef Bacik page = find_or_create_page(inode->i_mapping, 11303b16a4e3SJosef Bacik start_index + i, mask); 11314cb5300bSChris Mason if (!page) 11324cb5300bSChris Mason break; 11334cb5300bSChris Mason 1134600a45e1SMiao Xie page_start = page_offset(page); 1135600a45e1SMiao Xie page_end = page_start + PAGE_CACHE_SIZE - 1; 1136600a45e1SMiao Xie while (1) { 1137308d9800SFilipe Manana lock_extent_bits(tree, page_start, page_end, 1138308d9800SFilipe Manana 0, &cached_state); 1139600a45e1SMiao Xie ordered = btrfs_lookup_ordered_extent(inode, 1140600a45e1SMiao Xie page_start); 1141308d9800SFilipe Manana unlock_extent_cached(tree, page_start, page_end, 1142308d9800SFilipe Manana &cached_state, GFP_NOFS); 1143600a45e1SMiao Xie if (!ordered) 1144600a45e1SMiao Xie break; 1145600a45e1SMiao Xie 1146600a45e1SMiao Xie unlock_page(page); 1147600a45e1SMiao Xie btrfs_start_ordered_extent(inode, ordered, 1); 1148600a45e1SMiao Xie btrfs_put_ordered_extent(ordered); 1149600a45e1SMiao Xie lock_page(page); 11501f12bd06SLiu Bo /* 11511f12bd06SLiu Bo * we unlocked the page above, so we need check if 11521f12bd06SLiu Bo * it was released or not. 11531f12bd06SLiu Bo */ 11541f12bd06SLiu Bo if (page->mapping != inode->i_mapping) { 11551f12bd06SLiu Bo unlock_page(page); 11561f12bd06SLiu Bo page_cache_release(page); 11571f12bd06SLiu Bo goto again; 11581f12bd06SLiu Bo } 1159600a45e1SMiao Xie } 1160600a45e1SMiao Xie 11614cb5300bSChris Mason if (!PageUptodate(page)) { 11624cb5300bSChris Mason btrfs_readpage(NULL, page); 11634cb5300bSChris Mason lock_page(page); 11644cb5300bSChris Mason if (!PageUptodate(page)) { 11654cb5300bSChris Mason unlock_page(page); 11664cb5300bSChris Mason page_cache_release(page); 11674cb5300bSChris Mason ret = -EIO; 11684cb5300bSChris Mason break; 11694cb5300bSChris Mason } 11704cb5300bSChris Mason } 1171600a45e1SMiao Xie 1172600a45e1SMiao Xie if (page->mapping != inode->i_mapping) { 1173600a45e1SMiao Xie unlock_page(page); 1174600a45e1SMiao Xie page_cache_release(page); 1175600a45e1SMiao Xie goto again; 1176600a45e1SMiao Xie } 1177600a45e1SMiao Xie 11784cb5300bSChris Mason pages[i] = page; 11794cb5300bSChris Mason i_done++; 11804cb5300bSChris Mason } 11814cb5300bSChris Mason if (!i_done || ret) 11824cb5300bSChris Mason goto out; 11834cb5300bSChris Mason 11844cb5300bSChris Mason if (!(inode->i_sb->s_flags & MS_ACTIVE)) 11854cb5300bSChris Mason goto out; 11864cb5300bSChris Mason 11874cb5300bSChris Mason /* 11884cb5300bSChris Mason * so now we have a nice long stream of locked 11894cb5300bSChris Mason * and up to date pages, lets wait on them 11904cb5300bSChris Mason */ 11914cb5300bSChris Mason for (i = 0; i < i_done; i++) 11924cb5300bSChris Mason wait_on_page_writeback(pages[i]); 11934cb5300bSChris Mason 11944cb5300bSChris Mason page_start = page_offset(pages[0]); 11954cb5300bSChris Mason page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE; 11964cb5300bSChris Mason 11974cb5300bSChris Mason lock_extent_bits(&BTRFS_I(inode)->io_tree, 1198d0082371SJeff Mahoney page_start, page_end - 1, 0, &cached_state); 11994cb5300bSChris Mason clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, 12004cb5300bSChris Mason page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC | 12019e8a4a8bSLiu Bo EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0, 12029e8a4a8bSLiu Bo &cached_state, GFP_NOFS); 12034cb5300bSChris Mason 12041f12bd06SLiu Bo if (i_done != page_cnt) { 12059e0baf60SJosef Bacik spin_lock(&BTRFS_I(inode)->lock); 12069e0baf60SJosef Bacik BTRFS_I(inode)->outstanding_extents++; 12079e0baf60SJosef Bacik spin_unlock(&BTRFS_I(inode)->lock); 12084cb5300bSChris Mason btrfs_delalloc_release_space(inode, 12091f12bd06SLiu Bo (page_cnt - i_done) << PAGE_CACHE_SHIFT); 12104cb5300bSChris Mason } 12114cb5300bSChris Mason 12124cb5300bSChris Mason 12139e8a4a8bSLiu Bo set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1, 12149e8a4a8bSLiu Bo &cached_state, GFP_NOFS); 12154cb5300bSChris Mason 12164cb5300bSChris Mason unlock_extent_cached(&BTRFS_I(inode)->io_tree, 12174cb5300bSChris Mason page_start, page_end - 1, &cached_state, 12184cb5300bSChris Mason GFP_NOFS); 12194cb5300bSChris Mason 12204cb5300bSChris Mason for (i = 0; i < i_done; i++) { 12214cb5300bSChris Mason clear_page_dirty_for_io(pages[i]); 12224cb5300bSChris Mason ClearPageChecked(pages[i]); 12234cb5300bSChris Mason set_page_extent_mapped(pages[i]); 12244cb5300bSChris Mason set_page_dirty(pages[i]); 12254cb5300bSChris Mason unlock_page(pages[i]); 12264cb5300bSChris Mason page_cache_release(pages[i]); 12274cb5300bSChris Mason } 12284cb5300bSChris Mason return i_done; 12294cb5300bSChris Mason out: 12304cb5300bSChris Mason for (i = 0; i < i_done; i++) { 12314cb5300bSChris Mason unlock_page(pages[i]); 12324cb5300bSChris Mason page_cache_release(pages[i]); 12334cb5300bSChris Mason } 12341f12bd06SLiu Bo btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT); 12354cb5300bSChris Mason return ret; 12364cb5300bSChris Mason 12374cb5300bSChris Mason } 12384cb5300bSChris Mason 12394cb5300bSChris Mason int btrfs_defrag_file(struct inode *inode, struct file *file, 12404cb5300bSChris Mason struct btrfs_ioctl_defrag_range_args *range, 12414cb5300bSChris Mason u64 newer_than, unsigned long max_to_defrag) 12424cb5300bSChris Mason { 12434cb5300bSChris Mason struct btrfs_root *root = BTRFS_I(inode)->root; 12444cb5300bSChris Mason struct file_ra_state *ra = NULL; 12454cb5300bSChris Mason unsigned long last_index; 1246151a31b2SLi Zefan u64 isize = i_size_read(inode); 1247940100a4SChris Mason u64 last_len = 0; 1248940100a4SChris Mason u64 skip = 0; 1249940100a4SChris Mason u64 defrag_end = 0; 12504cb5300bSChris Mason u64 newer_off = range->start; 1251f46b5a66SChristoph Hellwig unsigned long i; 1252008873eaSLi Zefan unsigned long ra_index = 0; 1253f46b5a66SChristoph Hellwig int ret; 12544cb5300bSChris Mason int defrag_count = 0; 12551a419d85SLi Zefan int compress_type = BTRFS_COMPRESS_ZLIB; 1256aab110abSDavid Sterba u32 extent_thresh = range->extent_thresh; 1257c41570c9SJustin Maggard unsigned long max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT; 1258c41570c9SJustin Maggard unsigned long cluster = max_cluster; 12594cb5300bSChris Mason u64 new_align = ~((u64)128 * 1024 - 1); 12604cb5300bSChris Mason struct page **pages = NULL; 12614cb5300bSChris Mason 12620abd5b17SLiu Bo if (isize == 0) 12630abd5b17SLiu Bo return 0; 12640abd5b17SLiu Bo 12650abd5b17SLiu Bo if (range->start >= isize) 12660abd5b17SLiu Bo return -EINVAL; 12671a419d85SLi Zefan 12681a419d85SLi Zefan if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) { 12691a419d85SLi Zefan if (range->compress_type > BTRFS_COMPRESS_TYPES) 12701a419d85SLi Zefan return -EINVAL; 12711a419d85SLi Zefan if (range->compress_type) 12721a419d85SLi Zefan compress_type = range->compress_type; 12731a419d85SLi Zefan } 1274f46b5a66SChristoph Hellwig 12750abd5b17SLiu Bo if (extent_thresh == 0) 12760abd5b17SLiu Bo extent_thresh = 256 * 1024; 1277f46b5a66SChristoph Hellwig 12784cb5300bSChris Mason /* 12794cb5300bSChris Mason * if we were not given a file, allocate a readahead 12804cb5300bSChris Mason * context 12814cb5300bSChris Mason */ 12824cb5300bSChris Mason if (!file) { 12834cb5300bSChris Mason ra = kzalloc(sizeof(*ra), GFP_NOFS); 12844cb5300bSChris Mason if (!ra) 12854cb5300bSChris Mason return -ENOMEM; 12864cb5300bSChris Mason file_ra_state_init(ra, inode->i_mapping); 12874cb5300bSChris Mason } else { 12884cb5300bSChris Mason ra = &file->f_ra; 12894cb5300bSChris Mason } 12904cb5300bSChris Mason 1291d9b0d9baSDulshani Gunawardhana pages = kmalloc_array(max_cluster, sizeof(struct page *), 12924cb5300bSChris Mason GFP_NOFS); 12934cb5300bSChris Mason if (!pages) { 12944cb5300bSChris Mason ret = -ENOMEM; 12954cb5300bSChris Mason goto out_ra; 12964cb5300bSChris Mason } 12974cb5300bSChris Mason 12984cb5300bSChris Mason /* find the last page to defrag */ 12991e701a32SChris Mason if (range->start + range->len > range->start) { 1300151a31b2SLi Zefan last_index = min_t(u64, isize - 1, 13011e701a32SChris Mason range->start + range->len - 1) >> PAGE_CACHE_SHIFT; 13021e701a32SChris Mason } else { 1303151a31b2SLi Zefan last_index = (isize - 1) >> PAGE_CACHE_SHIFT; 13041e701a32SChris Mason } 13051e701a32SChris Mason 13064cb5300bSChris Mason if (newer_than) { 13074cb5300bSChris Mason ret = find_new_extents(root, inode, newer_than, 13084cb5300bSChris Mason &newer_off, 64 * 1024); 13094cb5300bSChris Mason if (!ret) { 13104cb5300bSChris Mason range->start = newer_off; 13114cb5300bSChris Mason /* 13124cb5300bSChris Mason * we always align our defrag to help keep 13134cb5300bSChris Mason * the extents in the file evenly spaced 13144cb5300bSChris Mason */ 13154cb5300bSChris Mason i = (newer_off & new_align) >> PAGE_CACHE_SHIFT; 13164cb5300bSChris Mason } else 13174cb5300bSChris Mason goto out_ra; 13184cb5300bSChris Mason } else { 13191e701a32SChris Mason i = range->start >> PAGE_CACHE_SHIFT; 13204cb5300bSChris Mason } 13214cb5300bSChris Mason if (!max_to_defrag) 1322070034bdSchandan max_to_defrag = last_index - i + 1; 13234cb5300bSChris Mason 13242a0f7f57SLi Zefan /* 13252a0f7f57SLi Zefan * make writeback starts from i, so the defrag range can be 13262a0f7f57SLi Zefan * written sequentially. 13272a0f7f57SLi Zefan */ 13282a0f7f57SLi Zefan if (i < inode->i_mapping->writeback_index) 13292a0f7f57SLi Zefan inode->i_mapping->writeback_index = i; 13302a0f7f57SLi Zefan 1331f7f43cc8SChris Mason while (i <= last_index && defrag_count < max_to_defrag && 1332ed6078f7SDavid Sterba (i < DIV_ROUND_UP(i_size_read(inode), PAGE_CACHE_SIZE))) { 13334cb5300bSChris Mason /* 13344cb5300bSChris Mason * make sure we stop running if someone unmounts 13354cb5300bSChris Mason * the FS 13364cb5300bSChris Mason */ 13374cb5300bSChris Mason if (!(inode->i_sb->s_flags & MS_ACTIVE)) 13384cb5300bSChris Mason break; 13394cb5300bSChris Mason 1340210549ebSDavid Sterba if (btrfs_defrag_cancelled(root->fs_info)) { 1341efe120a0SFrank Holton printk(KERN_DEBUG "BTRFS: defrag_file cancelled\n"); 1342210549ebSDavid Sterba ret = -EAGAIN; 1343210549ebSDavid Sterba break; 1344210549ebSDavid Sterba } 1345210549ebSDavid Sterba 134666c26892SLiu Bo if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT, 13476c282eb4SLi Zefan extent_thresh, &last_len, &skip, 1348a43a2111SAndrew Mahone &defrag_end, range->flags & 1349a43a2111SAndrew Mahone BTRFS_DEFRAG_RANGE_COMPRESS)) { 1350940100a4SChris Mason unsigned long next; 1351940100a4SChris Mason /* 1352940100a4SChris Mason * the should_defrag function tells us how much to skip 1353940100a4SChris Mason * bump our counter by the suggested amount 1354940100a4SChris Mason */ 1355ed6078f7SDavid Sterba next = DIV_ROUND_UP(skip, PAGE_CACHE_SIZE); 1356940100a4SChris Mason i = max(i + 1, next); 1357940100a4SChris Mason continue; 1358940100a4SChris Mason } 1359008873eaSLi Zefan 1360008873eaSLi Zefan if (!newer_than) { 1361008873eaSLi Zefan cluster = (PAGE_CACHE_ALIGN(defrag_end) >> 1362008873eaSLi Zefan PAGE_CACHE_SHIFT) - i; 1363008873eaSLi Zefan cluster = min(cluster, max_cluster); 1364008873eaSLi Zefan } else { 1365008873eaSLi Zefan cluster = max_cluster; 1366008873eaSLi Zefan } 1367008873eaSLi Zefan 1368008873eaSLi Zefan if (i + cluster > ra_index) { 1369008873eaSLi Zefan ra_index = max(i, ra_index); 1370008873eaSLi Zefan btrfs_force_ra(inode->i_mapping, ra, file, ra_index, 1371008873eaSLi Zefan cluster); 1372e4826a5bSchandan ra_index += cluster; 1373008873eaSLi Zefan } 13744cb5300bSChris Mason 1375ecb8bea8SLiu Bo mutex_lock(&inode->i_mutex); 1376633085c7SFilipe David Borba Manana if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) 1377633085c7SFilipe David Borba Manana BTRFS_I(inode)->force_compress = compress_type; 1378008873eaSLi Zefan ret = cluster_pages_for_defrag(inode, pages, i, cluster); 1379ecb8bea8SLiu Bo if (ret < 0) { 1380ecb8bea8SLiu Bo mutex_unlock(&inode->i_mutex); 13814cb5300bSChris Mason goto out_ra; 1382ecb8bea8SLiu Bo } 13834cb5300bSChris Mason 13844cb5300bSChris Mason defrag_count += ret; 1385d0e1d66bSNamjae Jeon balance_dirty_pages_ratelimited(inode->i_mapping); 1386ecb8bea8SLiu Bo mutex_unlock(&inode->i_mutex); 13874cb5300bSChris Mason 13884cb5300bSChris Mason if (newer_than) { 13894cb5300bSChris Mason if (newer_off == (u64)-1) 13904cb5300bSChris Mason break; 13914cb5300bSChris Mason 1392e1f041e1SLiu Bo if (ret > 0) 1393e1f041e1SLiu Bo i += ret; 1394e1f041e1SLiu Bo 13954cb5300bSChris Mason newer_off = max(newer_off + 1, 13964cb5300bSChris Mason (u64)i << PAGE_CACHE_SHIFT); 13974cb5300bSChris Mason 13984cb5300bSChris Mason ret = find_new_extents(root, inode, 13994cb5300bSChris Mason newer_than, &newer_off, 14004cb5300bSChris Mason 64 * 1024); 14014cb5300bSChris Mason if (!ret) { 14024cb5300bSChris Mason range->start = newer_off; 14034cb5300bSChris Mason i = (newer_off & new_align) >> PAGE_CACHE_SHIFT; 14044cb5300bSChris Mason } else { 14054cb5300bSChris Mason break; 1406940100a4SChris Mason } 14074cb5300bSChris Mason } else { 1408008873eaSLi Zefan if (ret > 0) { 1409cbcc8326SLi Zefan i += ret; 1410008873eaSLi Zefan last_len += ret << PAGE_CACHE_SHIFT; 1411008873eaSLi Zefan } else { 1412940100a4SChris Mason i++; 1413008873eaSLi Zefan last_len = 0; 1414008873eaSLi Zefan } 1415f46b5a66SChristoph Hellwig } 14164cb5300bSChris Mason } 1417f46b5a66SChristoph Hellwig 1418dec8ef90SFilipe Manana if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) { 14191e701a32SChris Mason filemap_flush(inode->i_mapping); 1420dec8ef90SFilipe Manana if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT, 1421dec8ef90SFilipe Manana &BTRFS_I(inode)->runtime_flags)) 1422dec8ef90SFilipe Manana filemap_flush(inode->i_mapping); 1423dec8ef90SFilipe Manana } 14241e701a32SChris Mason 14251e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 14261e701a32SChris Mason /* the filemap_flush will queue IO into the worker threads, but 14271e701a32SChris Mason * we have to make sure the IO is actually started and that 14281e701a32SChris Mason * ordered extents get created before we return 14291e701a32SChris Mason */ 14301e701a32SChris Mason atomic_inc(&root->fs_info->async_submit_draining); 14311e701a32SChris Mason while (atomic_read(&root->fs_info->nr_async_submits) || 14321e701a32SChris Mason atomic_read(&root->fs_info->async_delalloc_pages)) { 14331e701a32SChris Mason wait_event(root->fs_info->async_submit_wait, 14341e701a32SChris Mason (atomic_read(&root->fs_info->nr_async_submits) == 0 && 14351e701a32SChris Mason atomic_read(&root->fs_info->async_delalloc_pages) == 0)); 14361e701a32SChris Mason } 14371e701a32SChris Mason atomic_dec(&root->fs_info->async_submit_draining); 14381e701a32SChris Mason } 14391e701a32SChris Mason 14401a419d85SLi Zefan if (range->compress_type == BTRFS_COMPRESS_LZO) { 14412b0ce2c2SMitch Harder btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO); 14421a419d85SLi Zefan } 14431a419d85SLi Zefan 144460ccf82fSDiego Calleja ret = defrag_count; 1445940100a4SChris Mason 14464cb5300bSChris Mason out_ra: 1447633085c7SFilipe David Borba Manana if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) { 1448633085c7SFilipe David Borba Manana mutex_lock(&inode->i_mutex); 1449633085c7SFilipe David Borba Manana BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE; 1450633085c7SFilipe David Borba Manana mutex_unlock(&inode->i_mutex); 1451633085c7SFilipe David Borba Manana } 14524cb5300bSChris Mason if (!file) 14534cb5300bSChris Mason kfree(ra); 14544cb5300bSChris Mason kfree(pages); 1455940100a4SChris Mason return ret; 1456f46b5a66SChristoph Hellwig } 1457f46b5a66SChristoph Hellwig 1458198605a8SMiao Xie static noinline int btrfs_ioctl_resize(struct file *file, 145976dda93cSYan, Zheng void __user *arg) 1460f46b5a66SChristoph Hellwig { 1461f46b5a66SChristoph Hellwig u64 new_size; 1462f46b5a66SChristoph Hellwig u64 old_size; 1463f46b5a66SChristoph Hellwig u64 devid = 1; 1464496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 1465f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 1466f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 1467f46b5a66SChristoph Hellwig struct btrfs_device *device = NULL; 1468f46b5a66SChristoph Hellwig char *sizestr; 14699a40f122SGui Hecheng char *retptr; 1470f46b5a66SChristoph Hellwig char *devstr = NULL; 1471f46b5a66SChristoph Hellwig int ret = 0; 1472f46b5a66SChristoph Hellwig int mod = 0; 1473f46b5a66SChristoph Hellwig 1474e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 1475e441d54dSChris Mason return -EPERM; 1476e441d54dSChris Mason 1477198605a8SMiao Xie ret = mnt_want_write_file(file); 1478198605a8SMiao Xie if (ret) 1479198605a8SMiao Xie return ret; 1480198605a8SMiao Xie 14815ac00addSStefan Behrens if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running, 14825ac00addSStefan Behrens 1)) { 148397547676SMiao Xie mnt_drop_write_file(file); 1484e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 1485c9e9f97bSIlya Dryomov } 1486c9e9f97bSIlya Dryomov 14875ac00addSStefan Behrens mutex_lock(&root->fs_info->volume_mutex); 1488dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1489c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 1490c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 1491c9e9f97bSIlya Dryomov goto out; 1492c9e9f97bSIlya Dryomov } 14935516e595SMark Fasheh 14945516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1495f46b5a66SChristoph Hellwig 1496f46b5a66SChristoph Hellwig sizestr = vol_args->name; 1497f46b5a66SChristoph Hellwig devstr = strchr(sizestr, ':'); 1498f46b5a66SChristoph Hellwig if (devstr) { 1499f46b5a66SChristoph Hellwig sizestr = devstr + 1; 1500f46b5a66SChristoph Hellwig *devstr = '\0'; 1501f46b5a66SChristoph Hellwig devstr = vol_args->name; 150258dfae63SZhangZhen ret = kstrtoull(devstr, 10, &devid); 150358dfae63SZhangZhen if (ret) 150458dfae63SZhangZhen goto out_free; 1505dfd79829SMiao Xie if (!devid) { 1506dfd79829SMiao Xie ret = -EINVAL; 1507dfd79829SMiao Xie goto out_free; 1508dfd79829SMiao Xie } 1509efe120a0SFrank Holton btrfs_info(root->fs_info, "resizing devid %llu", devid); 1510f46b5a66SChristoph Hellwig } 1511dba60f3fSMiao Xie 1512aa1b8cd4SStefan Behrens device = btrfs_find_device(root->fs_info, devid, NULL, NULL); 1513f46b5a66SChristoph Hellwig if (!device) { 1514efe120a0SFrank Holton btrfs_info(root->fs_info, "resizer unable to find device %llu", 1515c1c9ff7cSGeert Uytterhoeven devid); 1516dfd79829SMiao Xie ret = -ENODEV; 1517c9e9f97bSIlya Dryomov goto out_free; 1518f46b5a66SChristoph Hellwig } 1519dba60f3fSMiao Xie 1520dba60f3fSMiao Xie if (!device->writeable) { 1521efe120a0SFrank Holton btrfs_info(root->fs_info, 1522efe120a0SFrank Holton "resizer unable to apply on readonly device %llu", 1523c1c9ff7cSGeert Uytterhoeven devid); 1524dfd79829SMiao Xie ret = -EPERM; 15254e42ae1bSLiu Bo goto out_free; 15264e42ae1bSLiu Bo } 15274e42ae1bSLiu Bo 1528f46b5a66SChristoph Hellwig if (!strcmp(sizestr, "max")) 1529f46b5a66SChristoph Hellwig new_size = device->bdev->bd_inode->i_size; 1530f46b5a66SChristoph Hellwig else { 1531f46b5a66SChristoph Hellwig if (sizestr[0] == '-') { 1532f46b5a66SChristoph Hellwig mod = -1; 1533f46b5a66SChristoph Hellwig sizestr++; 1534f46b5a66SChristoph Hellwig } else if (sizestr[0] == '+') { 1535f46b5a66SChristoph Hellwig mod = 1; 1536f46b5a66SChristoph Hellwig sizestr++; 1537f46b5a66SChristoph Hellwig } 15389a40f122SGui Hecheng new_size = memparse(sizestr, &retptr); 15399a40f122SGui Hecheng if (*retptr != '\0' || new_size == 0) { 1540f46b5a66SChristoph Hellwig ret = -EINVAL; 1541c9e9f97bSIlya Dryomov goto out_free; 1542f46b5a66SChristoph Hellwig } 1543f46b5a66SChristoph Hellwig } 1544f46b5a66SChristoph Hellwig 154563a212abSStefan Behrens if (device->is_tgtdev_for_dev_replace) { 1546dfd79829SMiao Xie ret = -EPERM; 154763a212abSStefan Behrens goto out_free; 154863a212abSStefan Behrens } 154963a212abSStefan Behrens 15507cc8e58dSMiao Xie old_size = btrfs_device_get_total_bytes(device); 1551f46b5a66SChristoph Hellwig 1552f46b5a66SChristoph Hellwig if (mod < 0) { 1553f46b5a66SChristoph Hellwig if (new_size > old_size) { 1554f46b5a66SChristoph Hellwig ret = -EINVAL; 1555c9e9f97bSIlya Dryomov goto out_free; 1556f46b5a66SChristoph Hellwig } 1557f46b5a66SChristoph Hellwig new_size = old_size - new_size; 1558f46b5a66SChristoph Hellwig } else if (mod > 0) { 1559eb8052e0SWenliang Fan if (new_size > ULLONG_MAX - old_size) { 1560902c68a4SGui Hecheng ret = -ERANGE; 1561eb8052e0SWenliang Fan goto out_free; 1562eb8052e0SWenliang Fan } 1563f46b5a66SChristoph Hellwig new_size = old_size + new_size; 1564f46b5a66SChristoph Hellwig } 1565f46b5a66SChristoph Hellwig 1566f46b5a66SChristoph Hellwig if (new_size < 256 * 1024 * 1024) { 1567f46b5a66SChristoph Hellwig ret = -EINVAL; 1568c9e9f97bSIlya Dryomov goto out_free; 1569f46b5a66SChristoph Hellwig } 1570f46b5a66SChristoph Hellwig if (new_size > device->bdev->bd_inode->i_size) { 1571f46b5a66SChristoph Hellwig ret = -EFBIG; 1572c9e9f97bSIlya Dryomov goto out_free; 1573f46b5a66SChristoph Hellwig } 1574f46b5a66SChristoph Hellwig 1575b8b93addSDavid Sterba new_size = div_u64(new_size, root->sectorsize); 1576f46b5a66SChristoph Hellwig new_size *= root->sectorsize; 1577f46b5a66SChristoph Hellwig 1578efe120a0SFrank Holton printk_in_rcu(KERN_INFO "BTRFS: new size for %s is %llu\n", 1579c1c9ff7cSGeert Uytterhoeven rcu_str_deref(device->name), new_size); 1580f46b5a66SChristoph Hellwig 1581f46b5a66SChristoph Hellwig if (new_size > old_size) { 1582a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 158398d5dc13STsutomu Itoh if (IS_ERR(trans)) { 158498d5dc13STsutomu Itoh ret = PTR_ERR(trans); 1585c9e9f97bSIlya Dryomov goto out_free; 158698d5dc13STsutomu Itoh } 1587f46b5a66SChristoph Hellwig ret = btrfs_grow_device(trans, device, new_size); 1588f46b5a66SChristoph Hellwig btrfs_commit_transaction(trans, root); 1589ece7d20eSMike Fleetwood } else if (new_size < old_size) { 1590f46b5a66SChristoph Hellwig ret = btrfs_shrink_device(device, new_size); 15910253f40eSjeff.liu } /* equal, nothing need to do */ 1592f46b5a66SChristoph Hellwig 1593c9e9f97bSIlya Dryomov out_free: 1594f46b5a66SChristoph Hellwig kfree(vol_args); 1595c9e9f97bSIlya Dryomov out: 1596c9e9f97bSIlya Dryomov mutex_unlock(&root->fs_info->volume_mutex); 15975ac00addSStefan Behrens atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0); 159818f39c41SIlya Dryomov mnt_drop_write_file(file); 1599f46b5a66SChristoph Hellwig return ret; 1600f46b5a66SChristoph Hellwig } 1601f46b5a66SChristoph Hellwig 160272fd032eSSage Weil static noinline int btrfs_ioctl_snap_create_transid(struct file *file, 16036f72c7e2SArne Jansen char *name, unsigned long fd, int subvol, 16046f72c7e2SArne Jansen u64 *transid, bool readonly, 16058696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 1606f46b5a66SChristoph Hellwig { 1607f46b5a66SChristoph Hellwig int namelen; 16083de4586cSChris Mason int ret = 0; 1609f46b5a66SChristoph Hellwig 1610a874a63eSLiu Bo ret = mnt_want_write_file(file); 1611a874a63eSLiu Bo if (ret) 1612a874a63eSLiu Bo goto out; 1613a874a63eSLiu Bo 161472fd032eSSage Weil namelen = strlen(name); 161572fd032eSSage Weil if (strchr(name, '/')) { 1616f46b5a66SChristoph Hellwig ret = -EINVAL; 1617a874a63eSLiu Bo goto out_drop_write; 1618f46b5a66SChristoph Hellwig } 1619f46b5a66SChristoph Hellwig 162016780cabSChris Mason if (name[0] == '.' && 162116780cabSChris Mason (namelen == 1 || (name[1] == '.' && namelen == 2))) { 162216780cabSChris Mason ret = -EEXIST; 1623a874a63eSLiu Bo goto out_drop_write; 162416780cabSChris Mason } 162516780cabSChris Mason 16263de4586cSChris Mason if (subvol) { 162772fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 16286f72c7e2SArne Jansen NULL, transid, readonly, inherit); 1629cb8e7090SChristoph Hellwig } else { 16302903ff01SAl Viro struct fd src = fdget(fd); 16313de4586cSChris Mason struct inode *src_inode; 16322903ff01SAl Viro if (!src.file) { 16333de4586cSChris Mason ret = -EINVAL; 1634a874a63eSLiu Bo goto out_drop_write; 16353de4586cSChris Mason } 16363de4586cSChris Mason 1637496ad9aaSAl Viro src_inode = file_inode(src.file); 1638496ad9aaSAl Viro if (src_inode->i_sb != file_inode(file)->i_sb) { 1639efe120a0SFrank Holton btrfs_info(BTRFS_I(src_inode)->root->fs_info, 1640efe120a0SFrank Holton "Snapshot src from another FS"); 164123ad5b17SKusanagi Kouichi ret = -EXDEV; 1642d0242061SDavid Sterba } else if (!inode_owner_or_capable(src_inode)) { 1643d0242061SDavid Sterba /* 1644d0242061SDavid Sterba * Subvolume creation is not restricted, but snapshots 1645d0242061SDavid Sterba * are limited to own subvolumes only 1646d0242061SDavid Sterba */ 1647d0242061SDavid Sterba ret = -EPERM; 1648ecd18815SAl Viro } else { 164972fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 165072fd032eSSage Weil BTRFS_I(src_inode)->root, 16516f72c7e2SArne Jansen transid, readonly, inherit); 1652ecd18815SAl Viro } 16532903ff01SAl Viro fdput(src); 1654cb8e7090SChristoph Hellwig } 1655a874a63eSLiu Bo out_drop_write: 1656a874a63eSLiu Bo mnt_drop_write_file(file); 1657f46b5a66SChristoph Hellwig out: 165872fd032eSSage Weil return ret; 165972fd032eSSage Weil } 166072fd032eSSage Weil 166172fd032eSSage Weil static noinline int btrfs_ioctl_snap_create(struct file *file, 1662fa0d2b9bSLi Zefan void __user *arg, int subvol) 166372fd032eSSage Weil { 1664fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args *vol_args; 166572fd032eSSage Weil int ret; 166672fd032eSSage Weil 1667fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1668fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1669fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1670fa0d2b9bSLi Zefan vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1671fa0d2b9bSLi Zefan 1672fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 1673b83cc969SLi Zefan vol_args->fd, subvol, 16746f72c7e2SArne Jansen NULL, false, NULL); 1675fa0d2b9bSLi Zefan 1676fa0d2b9bSLi Zefan kfree(vol_args); 1677fa0d2b9bSLi Zefan return ret; 1678fa0d2b9bSLi Zefan } 1679fa0d2b9bSLi Zefan 1680fa0d2b9bSLi Zefan static noinline int btrfs_ioctl_snap_create_v2(struct file *file, 1681fa0d2b9bSLi Zefan void __user *arg, int subvol) 1682fa0d2b9bSLi Zefan { 1683fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args_v2 *vol_args; 1684fa0d2b9bSLi Zefan int ret; 1685fdfb1e4fSLi Zefan u64 transid = 0; 1686fdfb1e4fSLi Zefan u64 *ptr = NULL; 1687b83cc969SLi Zefan bool readonly = false; 16886f72c7e2SArne Jansen struct btrfs_qgroup_inherit *inherit = NULL; 168972fd032eSSage Weil 1690fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1691fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1692fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1693fa0d2b9bSLi Zefan vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 1694fdfb1e4fSLi Zefan 1695b83cc969SLi Zefan if (vol_args->flags & 16966f72c7e2SArne Jansen ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY | 16976f72c7e2SArne Jansen BTRFS_SUBVOL_QGROUP_INHERIT)) { 1698b83cc969SLi Zefan ret = -EOPNOTSUPP; 1699c47ca32dSDan Carpenter goto free_args; 1700fdfb1e4fSLi Zefan } 1701fdfb1e4fSLi Zefan 1702fa0d2b9bSLi Zefan if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC) 1703fdfb1e4fSLi Zefan ptr = &transid; 1704b83cc969SLi Zefan if (vol_args->flags & BTRFS_SUBVOL_RDONLY) 1705b83cc969SLi Zefan readonly = true; 17066f72c7e2SArne Jansen if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) { 17076f72c7e2SArne Jansen if (vol_args->size > PAGE_CACHE_SIZE) { 17086f72c7e2SArne Jansen ret = -EINVAL; 1709c47ca32dSDan Carpenter goto free_args; 17106f72c7e2SArne Jansen } 17116f72c7e2SArne Jansen inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size); 17126f72c7e2SArne Jansen if (IS_ERR(inherit)) { 17136f72c7e2SArne Jansen ret = PTR_ERR(inherit); 1714c47ca32dSDan Carpenter goto free_args; 17156f72c7e2SArne Jansen } 17166f72c7e2SArne Jansen } 171775eaa0e2SSage Weil 1718fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 17196f72c7e2SArne Jansen vol_args->fd, subvol, ptr, 17208696c533SMiao Xie readonly, inherit); 1721c47ca32dSDan Carpenter if (ret) 1722c47ca32dSDan Carpenter goto free_inherit; 172375eaa0e2SSage Weil 1724c47ca32dSDan Carpenter if (ptr && copy_to_user(arg + 1725fdfb1e4fSLi Zefan offsetof(struct btrfs_ioctl_vol_args_v2, 1726c47ca32dSDan Carpenter transid), 1727c47ca32dSDan Carpenter ptr, sizeof(*ptr))) 172875eaa0e2SSage Weil ret = -EFAULT; 1729c47ca32dSDan Carpenter 1730c47ca32dSDan Carpenter free_inherit: 17316f72c7e2SArne Jansen kfree(inherit); 1732c47ca32dSDan Carpenter free_args: 1733c47ca32dSDan Carpenter kfree(vol_args); 1734f46b5a66SChristoph Hellwig return ret; 1735f46b5a66SChristoph Hellwig } 1736f46b5a66SChristoph Hellwig 17370caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_getflags(struct file *file, 17380caa102dSLi Zefan void __user *arg) 17390caa102dSLi Zefan { 1740496ad9aaSAl Viro struct inode *inode = file_inode(file); 17410caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 17420caa102dSLi Zefan int ret = 0; 17430caa102dSLi Zefan u64 flags = 0; 17440caa102dSLi Zefan 174533345d01SLi Zefan if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) 17460caa102dSLi Zefan return -EINVAL; 17470caa102dSLi Zefan 17480caa102dSLi Zefan down_read(&root->fs_info->subvol_sem); 17490caa102dSLi Zefan if (btrfs_root_readonly(root)) 17500caa102dSLi Zefan flags |= BTRFS_SUBVOL_RDONLY; 17510caa102dSLi Zefan up_read(&root->fs_info->subvol_sem); 17520caa102dSLi Zefan 17530caa102dSLi Zefan if (copy_to_user(arg, &flags, sizeof(flags))) 17540caa102dSLi Zefan ret = -EFAULT; 17550caa102dSLi Zefan 17560caa102dSLi Zefan return ret; 17570caa102dSLi Zefan } 17580caa102dSLi Zefan 17590caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_setflags(struct file *file, 17600caa102dSLi Zefan void __user *arg) 17610caa102dSLi Zefan { 1762496ad9aaSAl Viro struct inode *inode = file_inode(file); 17630caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 17640caa102dSLi Zefan struct btrfs_trans_handle *trans; 17650caa102dSLi Zefan u64 root_flags; 17660caa102dSLi Zefan u64 flags; 17670caa102dSLi Zefan int ret = 0; 17680caa102dSLi Zefan 1769bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 1770bd60ea0fSDavid Sterba return -EPERM; 1771bd60ea0fSDavid Sterba 1772b9ca0664SLiu Bo ret = mnt_want_write_file(file); 1773b9ca0664SLiu Bo if (ret) 1774b9ca0664SLiu Bo goto out; 17750caa102dSLi Zefan 1776b9ca0664SLiu Bo if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { 1777b9ca0664SLiu Bo ret = -EINVAL; 1778b9ca0664SLiu Bo goto out_drop_write; 1779b9ca0664SLiu Bo } 17800caa102dSLi Zefan 1781b9ca0664SLiu Bo if (copy_from_user(&flags, arg, sizeof(flags))) { 1782b9ca0664SLiu Bo ret = -EFAULT; 1783b9ca0664SLiu Bo goto out_drop_write; 1784b9ca0664SLiu Bo } 17850caa102dSLi Zefan 1786b9ca0664SLiu Bo if (flags & BTRFS_SUBVOL_CREATE_ASYNC) { 1787b9ca0664SLiu Bo ret = -EINVAL; 1788b9ca0664SLiu Bo goto out_drop_write; 1789b9ca0664SLiu Bo } 17900caa102dSLi Zefan 1791b9ca0664SLiu Bo if (flags & ~BTRFS_SUBVOL_RDONLY) { 1792b9ca0664SLiu Bo ret = -EOPNOTSUPP; 1793b9ca0664SLiu Bo goto out_drop_write; 1794b9ca0664SLiu Bo } 17950caa102dSLi Zefan 17960caa102dSLi Zefan down_write(&root->fs_info->subvol_sem); 17970caa102dSLi Zefan 17980caa102dSLi Zefan /* nothing to do */ 17990caa102dSLi Zefan if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) 1800b9ca0664SLiu Bo goto out_drop_sem; 18010caa102dSLi Zefan 18020caa102dSLi Zefan root_flags = btrfs_root_flags(&root->root_item); 18032c686537SDavid Sterba if (flags & BTRFS_SUBVOL_RDONLY) { 18040caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 18050caa102dSLi Zefan root_flags | BTRFS_ROOT_SUBVOL_RDONLY); 18062c686537SDavid Sterba } else { 18072c686537SDavid Sterba /* 18082c686537SDavid Sterba * Block RO -> RW transition if this subvolume is involved in 18092c686537SDavid Sterba * send 18102c686537SDavid Sterba */ 18112c686537SDavid Sterba spin_lock(&root->root_item_lock); 18122c686537SDavid Sterba if (root->send_in_progress == 0) { 18130caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 18140caa102dSLi Zefan root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); 18152c686537SDavid Sterba spin_unlock(&root->root_item_lock); 18162c686537SDavid Sterba } else { 18172c686537SDavid Sterba spin_unlock(&root->root_item_lock); 18182c686537SDavid Sterba btrfs_warn(root->fs_info, 18192c686537SDavid Sterba "Attempt to set subvolume %llu read-write during send", 18202c686537SDavid Sterba root->root_key.objectid); 18212c686537SDavid Sterba ret = -EPERM; 18222c686537SDavid Sterba goto out_drop_sem; 18232c686537SDavid Sterba } 18242c686537SDavid Sterba } 18250caa102dSLi Zefan 18260caa102dSLi Zefan trans = btrfs_start_transaction(root, 1); 18270caa102dSLi Zefan if (IS_ERR(trans)) { 18280caa102dSLi Zefan ret = PTR_ERR(trans); 18290caa102dSLi Zefan goto out_reset; 18300caa102dSLi Zefan } 18310caa102dSLi Zefan 1832b4dc2b8cSLi Zefan ret = btrfs_update_root(trans, root->fs_info->tree_root, 18330caa102dSLi Zefan &root->root_key, &root->root_item); 18340caa102dSLi Zefan 18350caa102dSLi Zefan btrfs_commit_transaction(trans, root); 18360caa102dSLi Zefan out_reset: 18370caa102dSLi Zefan if (ret) 18380caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, root_flags); 1839b9ca0664SLiu Bo out_drop_sem: 18400caa102dSLi Zefan up_write(&root->fs_info->subvol_sem); 1841b9ca0664SLiu Bo out_drop_write: 1842b9ca0664SLiu Bo mnt_drop_write_file(file); 1843b9ca0664SLiu Bo out: 18440caa102dSLi Zefan return ret; 18450caa102dSLi Zefan } 18460caa102dSLi Zefan 184776dda93cSYan, Zheng /* 184876dda93cSYan, Zheng * helper to check if the subvolume references other subvolumes 184976dda93cSYan, Zheng */ 185076dda93cSYan, Zheng static noinline int may_destroy_subvol(struct btrfs_root *root) 185176dda93cSYan, Zheng { 185276dda93cSYan, Zheng struct btrfs_path *path; 1853175a2b87SJosef Bacik struct btrfs_dir_item *di; 185476dda93cSYan, Zheng struct btrfs_key key; 1855175a2b87SJosef Bacik u64 dir_id; 185676dda93cSYan, Zheng int ret; 185776dda93cSYan, Zheng 185876dda93cSYan, Zheng path = btrfs_alloc_path(); 185976dda93cSYan, Zheng if (!path) 186076dda93cSYan, Zheng return -ENOMEM; 186176dda93cSYan, Zheng 1862175a2b87SJosef Bacik /* Make sure this root isn't set as the default subvol */ 1863175a2b87SJosef Bacik dir_id = btrfs_super_root_dir(root->fs_info->super_copy); 1864175a2b87SJosef Bacik di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path, 1865175a2b87SJosef Bacik dir_id, "default", 7, 0); 1866175a2b87SJosef Bacik if (di && !IS_ERR(di)) { 1867175a2b87SJosef Bacik btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key); 1868175a2b87SJosef Bacik if (key.objectid == root->root_key.objectid) { 186972de6b53SGuangyu Sun ret = -EPERM; 187072de6b53SGuangyu Sun btrfs_err(root->fs_info, "deleting default subvolume " 187172de6b53SGuangyu Sun "%llu is not allowed", key.objectid); 1872175a2b87SJosef Bacik goto out; 1873175a2b87SJosef Bacik } 1874175a2b87SJosef Bacik btrfs_release_path(path); 1875175a2b87SJosef Bacik } 1876175a2b87SJosef Bacik 187776dda93cSYan, Zheng key.objectid = root->root_key.objectid; 187876dda93cSYan, Zheng key.type = BTRFS_ROOT_REF_KEY; 187976dda93cSYan, Zheng key.offset = (u64)-1; 188076dda93cSYan, Zheng 188176dda93cSYan, Zheng ret = btrfs_search_slot(NULL, root->fs_info->tree_root, 188276dda93cSYan, Zheng &key, path, 0, 0); 188376dda93cSYan, Zheng if (ret < 0) 188476dda93cSYan, Zheng goto out; 188576dda93cSYan, Zheng BUG_ON(ret == 0); 188676dda93cSYan, Zheng 188776dda93cSYan, Zheng ret = 0; 188876dda93cSYan, Zheng if (path->slots[0] > 0) { 188976dda93cSYan, Zheng path->slots[0]--; 189076dda93cSYan, Zheng btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 189176dda93cSYan, Zheng if (key.objectid == root->root_key.objectid && 189276dda93cSYan, Zheng key.type == BTRFS_ROOT_REF_KEY) 189376dda93cSYan, Zheng ret = -ENOTEMPTY; 189476dda93cSYan, Zheng } 189576dda93cSYan, Zheng out: 189676dda93cSYan, Zheng btrfs_free_path(path); 189776dda93cSYan, Zheng return ret; 189876dda93cSYan, Zheng } 189976dda93cSYan, Zheng 1900ac8e9819SChris Mason static noinline int key_in_sk(struct btrfs_key *key, 1901ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk) 1902ac8e9819SChris Mason { 1903abc6e134SChris Mason struct btrfs_key test; 1904abc6e134SChris Mason int ret; 1905abc6e134SChris Mason 1906abc6e134SChris Mason test.objectid = sk->min_objectid; 1907abc6e134SChris Mason test.type = sk->min_type; 1908abc6e134SChris Mason test.offset = sk->min_offset; 1909abc6e134SChris Mason 1910abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1911abc6e134SChris Mason if (ret < 0) 1912ac8e9819SChris Mason return 0; 1913abc6e134SChris Mason 1914abc6e134SChris Mason test.objectid = sk->max_objectid; 1915abc6e134SChris Mason test.type = sk->max_type; 1916abc6e134SChris Mason test.offset = sk->max_offset; 1917abc6e134SChris Mason 1918abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1919abc6e134SChris Mason if (ret > 0) 1920ac8e9819SChris Mason return 0; 1921ac8e9819SChris Mason return 1; 1922ac8e9819SChris Mason } 1923ac8e9819SChris Mason 1924ac8e9819SChris Mason static noinline int copy_to_sk(struct btrfs_root *root, 1925ac8e9819SChris Mason struct btrfs_path *path, 1926ac8e9819SChris Mason struct btrfs_key *key, 1927ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk, 19289b6e817dSGerhard Heift size_t *buf_size, 1929ba346b35SGerhard Heift char __user *ubuf, 1930ac8e9819SChris Mason unsigned long *sk_offset, 1931ac8e9819SChris Mason int *num_found) 1932ac8e9819SChris Mason { 1933ac8e9819SChris Mason u64 found_transid; 1934ac8e9819SChris Mason struct extent_buffer *leaf; 1935ac8e9819SChris Mason struct btrfs_ioctl_search_header sh; 1936*dd81d459SNaohiro Aota struct btrfs_key test; 1937ac8e9819SChris Mason unsigned long item_off; 1938ac8e9819SChris Mason unsigned long item_len; 1939ac8e9819SChris Mason int nritems; 1940ac8e9819SChris Mason int i; 1941ac8e9819SChris Mason int slot; 1942ac8e9819SChris Mason int ret = 0; 1943ac8e9819SChris Mason 1944ac8e9819SChris Mason leaf = path->nodes[0]; 1945ac8e9819SChris Mason slot = path->slots[0]; 1946ac8e9819SChris Mason nritems = btrfs_header_nritems(leaf); 1947ac8e9819SChris Mason 1948ac8e9819SChris Mason if (btrfs_header_generation(leaf) > sk->max_transid) { 1949ac8e9819SChris Mason i = nritems; 1950ac8e9819SChris Mason goto advance_key; 1951ac8e9819SChris Mason } 1952ac8e9819SChris Mason found_transid = btrfs_header_generation(leaf); 1953ac8e9819SChris Mason 1954ac8e9819SChris Mason for (i = slot; i < nritems; i++) { 1955ac8e9819SChris Mason item_off = btrfs_item_ptr_offset(leaf, i); 1956ac8e9819SChris Mason item_len = btrfs_item_size_nr(leaf, i); 1957ac8e9819SChris Mason 195803b71c6cSGabriel de Perthuis btrfs_item_key_to_cpu(leaf, key, i); 195903b71c6cSGabriel de Perthuis if (!key_in_sk(key, sk)) 196003b71c6cSGabriel de Perthuis continue; 196103b71c6cSGabriel de Perthuis 19629b6e817dSGerhard Heift if (sizeof(sh) + item_len > *buf_size) { 19638f5f6178SGerhard Heift if (*num_found) { 1964ac8e9819SChris Mason ret = 1; 19658f5f6178SGerhard Heift goto out; 19668f5f6178SGerhard Heift } 19678f5f6178SGerhard Heift 19688f5f6178SGerhard Heift /* 19698f5f6178SGerhard Heift * return one empty item back for v1, which does not 19708f5f6178SGerhard Heift * handle -EOVERFLOW 19718f5f6178SGerhard Heift */ 19728f5f6178SGerhard Heift 19739b6e817dSGerhard Heift *buf_size = sizeof(sh) + item_len; 1974ac8e9819SChris Mason item_len = 0; 19758f5f6178SGerhard Heift ret = -EOVERFLOW; 19768f5f6178SGerhard Heift } 1977ac8e9819SChris Mason 19789b6e817dSGerhard Heift if (sizeof(sh) + item_len + *sk_offset > *buf_size) { 1979ac8e9819SChris Mason ret = 1; 198025c9bc2eSGerhard Heift goto out; 1981ac8e9819SChris Mason } 1982ac8e9819SChris Mason 1983ac8e9819SChris Mason sh.objectid = key->objectid; 1984ac8e9819SChris Mason sh.offset = key->offset; 1985ac8e9819SChris Mason sh.type = key->type; 1986ac8e9819SChris Mason sh.len = item_len; 1987ac8e9819SChris Mason sh.transid = found_transid; 1988ac8e9819SChris Mason 1989ac8e9819SChris Mason /* copy search result header */ 1990ba346b35SGerhard Heift if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) { 1991ba346b35SGerhard Heift ret = -EFAULT; 1992ba346b35SGerhard Heift goto out; 1993ba346b35SGerhard Heift } 1994ba346b35SGerhard Heift 1995ac8e9819SChris Mason *sk_offset += sizeof(sh); 1996ac8e9819SChris Mason 1997ac8e9819SChris Mason if (item_len) { 1998ba346b35SGerhard Heift char __user *up = ubuf + *sk_offset; 1999ac8e9819SChris Mason /* copy the item */ 2000ba346b35SGerhard Heift if (read_extent_buffer_to_user(leaf, up, 2001ba346b35SGerhard Heift item_off, item_len)) { 2002ba346b35SGerhard Heift ret = -EFAULT; 2003ba346b35SGerhard Heift goto out; 2004ba346b35SGerhard Heift } 2005ba346b35SGerhard Heift 2006ac8e9819SChris Mason *sk_offset += item_len; 2007ac8e9819SChris Mason } 2008e2156867SHugo Mills (*num_found)++; 2009ac8e9819SChris Mason 20108f5f6178SGerhard Heift if (ret) /* -EOVERFLOW from above */ 20118f5f6178SGerhard Heift goto out; 20128f5f6178SGerhard Heift 201325c9bc2eSGerhard Heift if (*num_found >= sk->nr_items) { 201425c9bc2eSGerhard Heift ret = 1; 201525c9bc2eSGerhard Heift goto out; 201625c9bc2eSGerhard Heift } 2017ac8e9819SChris Mason } 2018ac8e9819SChris Mason advance_key: 2019ac8e9819SChris Mason ret = 0; 2020*dd81d459SNaohiro Aota test.objectid = sk->max_objectid; 2021*dd81d459SNaohiro Aota test.type = sk->max_type; 2022*dd81d459SNaohiro Aota test.offset = sk->max_offset; 2023*dd81d459SNaohiro Aota if (btrfs_comp_cpu_keys(key, &test) >= 0) 2024*dd81d459SNaohiro Aota ret = 1; 2025*dd81d459SNaohiro Aota else if (key->offset < (u64)-1) 2026abc6e134SChris Mason key->offset++; 2027*dd81d459SNaohiro Aota else if (key->type < (u8)-1) { 2028abc6e134SChris Mason key->offset = 0; 2029abc6e134SChris Mason key->type++; 2030*dd81d459SNaohiro Aota } else if (key->objectid < (u64)-1) { 2031abc6e134SChris Mason key->offset = 0; 2032abc6e134SChris Mason key->type = 0; 2033abc6e134SChris Mason key->objectid++; 2034abc6e134SChris Mason } else 2035abc6e134SChris Mason ret = 1; 203625c9bc2eSGerhard Heift out: 2037ba346b35SGerhard Heift /* 2038ba346b35SGerhard Heift * 0: all items from this leaf copied, continue with next 2039ba346b35SGerhard Heift * 1: * more items can be copied, but unused buffer is too small 2040ba346b35SGerhard Heift * * all items were found 2041ba346b35SGerhard Heift * Either way, it will stops the loop which iterates to the next 2042ba346b35SGerhard Heift * leaf 2043ba346b35SGerhard Heift * -EOVERFLOW: item was to large for buffer 2044ba346b35SGerhard Heift * -EFAULT: could not copy extent buffer back to userspace 2045ba346b35SGerhard Heift */ 2046ac8e9819SChris Mason return ret; 2047ac8e9819SChris Mason } 2048ac8e9819SChris Mason 2049ac8e9819SChris Mason static noinline int search_ioctl(struct inode *inode, 205012544442SGerhard Heift struct btrfs_ioctl_search_key *sk, 20519b6e817dSGerhard Heift size_t *buf_size, 2052ba346b35SGerhard Heift char __user *ubuf) 2053ac8e9819SChris Mason { 2054ac8e9819SChris Mason struct btrfs_root *root; 2055ac8e9819SChris Mason struct btrfs_key key; 2056ac8e9819SChris Mason struct btrfs_path *path; 2057ac8e9819SChris Mason struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info; 2058ac8e9819SChris Mason int ret; 2059ac8e9819SChris Mason int num_found = 0; 2060ac8e9819SChris Mason unsigned long sk_offset = 0; 2061ac8e9819SChris Mason 20629b6e817dSGerhard Heift if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) { 20639b6e817dSGerhard Heift *buf_size = sizeof(struct btrfs_ioctl_search_header); 206412544442SGerhard Heift return -EOVERFLOW; 20659b6e817dSGerhard Heift } 206612544442SGerhard Heift 2067ac8e9819SChris Mason path = btrfs_alloc_path(); 2068ac8e9819SChris Mason if (!path) 2069ac8e9819SChris Mason return -ENOMEM; 2070ac8e9819SChris Mason 2071ac8e9819SChris Mason if (sk->tree_id == 0) { 2072ac8e9819SChris Mason /* search the root of the inode that was passed */ 2073ac8e9819SChris Mason root = BTRFS_I(inode)->root; 2074ac8e9819SChris Mason } else { 2075ac8e9819SChris Mason key.objectid = sk->tree_id; 2076ac8e9819SChris Mason key.type = BTRFS_ROOT_ITEM_KEY; 2077ac8e9819SChris Mason key.offset = (u64)-1; 2078ac8e9819SChris Mason root = btrfs_read_fs_root_no_name(info, &key); 2079ac8e9819SChris Mason if (IS_ERR(root)) { 2080efe120a0SFrank Holton printk(KERN_ERR "BTRFS: could not find root %llu\n", 2081ac8e9819SChris Mason sk->tree_id); 2082ac8e9819SChris Mason btrfs_free_path(path); 2083ac8e9819SChris Mason return -ENOENT; 2084ac8e9819SChris Mason } 2085ac8e9819SChris Mason } 2086ac8e9819SChris Mason 2087ac8e9819SChris Mason key.objectid = sk->min_objectid; 2088ac8e9819SChris Mason key.type = sk->min_type; 2089ac8e9819SChris Mason key.offset = sk->min_offset; 2090ac8e9819SChris Mason 2091ac8e9819SChris Mason while (1) { 20926174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &key, path, sk->min_transid); 2093ac8e9819SChris Mason if (ret != 0) { 2094ac8e9819SChris Mason if (ret > 0) 2095ac8e9819SChris Mason ret = 0; 2096ac8e9819SChris Mason goto err; 2097ac8e9819SChris Mason } 2098ba346b35SGerhard Heift ret = copy_to_sk(root, path, &key, sk, buf_size, ubuf, 2099ac8e9819SChris Mason &sk_offset, &num_found); 2100b3b4aa74SDavid Sterba btrfs_release_path(path); 210125c9bc2eSGerhard Heift if (ret) 2102ac8e9819SChris Mason break; 2103ac8e9819SChris Mason 2104ac8e9819SChris Mason } 21058f5f6178SGerhard Heift if (ret > 0) 2106ac8e9819SChris Mason ret = 0; 2107ac8e9819SChris Mason err: 2108ac8e9819SChris Mason sk->nr_items = num_found; 2109ac8e9819SChris Mason btrfs_free_path(path); 2110ac8e9819SChris Mason return ret; 2111ac8e9819SChris Mason } 2112ac8e9819SChris Mason 2113ac8e9819SChris Mason static noinline int btrfs_ioctl_tree_search(struct file *file, 2114ac8e9819SChris Mason void __user *argp) 2115ac8e9819SChris Mason { 2116ba346b35SGerhard Heift struct btrfs_ioctl_search_args __user *uargs; 2117ba346b35SGerhard Heift struct btrfs_ioctl_search_key sk; 2118ac8e9819SChris Mason struct inode *inode; 2119ac8e9819SChris Mason int ret; 21209b6e817dSGerhard Heift size_t buf_size; 2121ac8e9819SChris Mason 2122ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 2123ac8e9819SChris Mason return -EPERM; 2124ac8e9819SChris Mason 2125ba346b35SGerhard Heift uargs = (struct btrfs_ioctl_search_args __user *)argp; 2126ac8e9819SChris Mason 2127ba346b35SGerhard Heift if (copy_from_user(&sk, &uargs->key, sizeof(sk))) 2128ba346b35SGerhard Heift return -EFAULT; 2129ba346b35SGerhard Heift 2130ba346b35SGerhard Heift buf_size = sizeof(uargs->buf); 2131ac8e9819SChris Mason 2132496ad9aaSAl Viro inode = file_inode(file); 2133ba346b35SGerhard Heift ret = search_ioctl(inode, &sk, &buf_size, uargs->buf); 21348f5f6178SGerhard Heift 21358f5f6178SGerhard Heift /* 21368f5f6178SGerhard Heift * In the origin implementation an overflow is handled by returning a 21378f5f6178SGerhard Heift * search header with a len of zero, so reset ret. 21388f5f6178SGerhard Heift */ 21398f5f6178SGerhard Heift if (ret == -EOVERFLOW) 21408f5f6178SGerhard Heift ret = 0; 21418f5f6178SGerhard Heift 2142ba346b35SGerhard Heift if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk))) 2143ac8e9819SChris Mason ret = -EFAULT; 2144ac8e9819SChris Mason return ret; 2145ac8e9819SChris Mason } 2146ac8e9819SChris Mason 2147cc68a8a5SGerhard Heift static noinline int btrfs_ioctl_tree_search_v2(struct file *file, 2148cc68a8a5SGerhard Heift void __user *argp) 2149cc68a8a5SGerhard Heift { 2150cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 __user *uarg; 2151cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 args; 2152cc68a8a5SGerhard Heift struct inode *inode; 2153cc68a8a5SGerhard Heift int ret; 2154cc68a8a5SGerhard Heift size_t buf_size; 2155cc68a8a5SGerhard Heift const size_t buf_limit = 16 * 1024 * 1024; 2156cc68a8a5SGerhard Heift 2157cc68a8a5SGerhard Heift if (!capable(CAP_SYS_ADMIN)) 2158cc68a8a5SGerhard Heift return -EPERM; 2159cc68a8a5SGerhard Heift 2160cc68a8a5SGerhard Heift /* copy search header and buffer size */ 2161cc68a8a5SGerhard Heift uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp; 2162cc68a8a5SGerhard Heift if (copy_from_user(&args, uarg, sizeof(args))) 2163cc68a8a5SGerhard Heift return -EFAULT; 2164cc68a8a5SGerhard Heift 2165cc68a8a5SGerhard Heift buf_size = args.buf_size; 2166cc68a8a5SGerhard Heift 2167cc68a8a5SGerhard Heift if (buf_size < sizeof(struct btrfs_ioctl_search_header)) 2168cc68a8a5SGerhard Heift return -EOVERFLOW; 2169cc68a8a5SGerhard Heift 2170cc68a8a5SGerhard Heift /* limit result size to 16MB */ 2171cc68a8a5SGerhard Heift if (buf_size > buf_limit) 2172cc68a8a5SGerhard Heift buf_size = buf_limit; 2173cc68a8a5SGerhard Heift 2174cc68a8a5SGerhard Heift inode = file_inode(file); 2175cc68a8a5SGerhard Heift ret = search_ioctl(inode, &args.key, &buf_size, 2176cc68a8a5SGerhard Heift (char *)(&uarg->buf[0])); 2177cc68a8a5SGerhard Heift if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key))) 2178cc68a8a5SGerhard Heift ret = -EFAULT; 2179cc68a8a5SGerhard Heift else if (ret == -EOVERFLOW && 2180cc68a8a5SGerhard Heift copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size))) 2181cc68a8a5SGerhard Heift ret = -EFAULT; 2182cc68a8a5SGerhard Heift 218376dda93cSYan, Zheng return ret; 218476dda93cSYan, Zheng } 218576dda93cSYan, Zheng 218698d377a0STARUISI Hiroaki /* 2187ac8e9819SChris Mason * Search INODE_REFs to identify path name of 'dirid' directory 2188ac8e9819SChris Mason * in a 'tree_id' tree. and sets path name to 'name'. 218998d377a0STARUISI Hiroaki */ 219098d377a0STARUISI Hiroaki static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, 219198d377a0STARUISI Hiroaki u64 tree_id, u64 dirid, char *name) 219298d377a0STARUISI Hiroaki { 219398d377a0STARUISI Hiroaki struct btrfs_root *root; 219498d377a0STARUISI Hiroaki struct btrfs_key key; 2195ac8e9819SChris Mason char *ptr; 219698d377a0STARUISI Hiroaki int ret = -1; 219798d377a0STARUISI Hiroaki int slot; 219898d377a0STARUISI Hiroaki int len; 219998d377a0STARUISI Hiroaki int total_len = 0; 220098d377a0STARUISI Hiroaki struct btrfs_inode_ref *iref; 220198d377a0STARUISI Hiroaki struct extent_buffer *l; 220298d377a0STARUISI Hiroaki struct btrfs_path *path; 220398d377a0STARUISI Hiroaki 220498d377a0STARUISI Hiroaki if (dirid == BTRFS_FIRST_FREE_OBJECTID) { 220598d377a0STARUISI Hiroaki name[0]='\0'; 220698d377a0STARUISI Hiroaki return 0; 220798d377a0STARUISI Hiroaki } 220898d377a0STARUISI Hiroaki 220998d377a0STARUISI Hiroaki path = btrfs_alloc_path(); 221098d377a0STARUISI Hiroaki if (!path) 221198d377a0STARUISI Hiroaki return -ENOMEM; 221298d377a0STARUISI Hiroaki 2213ac8e9819SChris Mason ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX]; 221498d377a0STARUISI Hiroaki 221598d377a0STARUISI Hiroaki key.objectid = tree_id; 221698d377a0STARUISI Hiroaki key.type = BTRFS_ROOT_ITEM_KEY; 221798d377a0STARUISI Hiroaki key.offset = (u64)-1; 221898d377a0STARUISI Hiroaki root = btrfs_read_fs_root_no_name(info, &key); 221998d377a0STARUISI Hiroaki if (IS_ERR(root)) { 2220efe120a0SFrank Holton printk(KERN_ERR "BTRFS: could not find root %llu\n", tree_id); 22218ad6fcabSChris Mason ret = -ENOENT; 22228ad6fcabSChris Mason goto out; 222398d377a0STARUISI Hiroaki } 222498d377a0STARUISI Hiroaki 222598d377a0STARUISI Hiroaki key.objectid = dirid; 222698d377a0STARUISI Hiroaki key.type = BTRFS_INODE_REF_KEY; 22278ad6fcabSChris Mason key.offset = (u64)-1; 222898d377a0STARUISI Hiroaki 222998d377a0STARUISI Hiroaki while (1) { 223098d377a0STARUISI Hiroaki ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 223198d377a0STARUISI Hiroaki if (ret < 0) 223298d377a0STARUISI Hiroaki goto out; 223318674c6cSFilipe David Borba Manana else if (ret > 0) { 223418674c6cSFilipe David Borba Manana ret = btrfs_previous_item(root, path, dirid, 223518674c6cSFilipe David Borba Manana BTRFS_INODE_REF_KEY); 223618674c6cSFilipe David Borba Manana if (ret < 0) 223718674c6cSFilipe David Borba Manana goto out; 223818674c6cSFilipe David Borba Manana else if (ret > 0) { 2239ac8e9819SChris Mason ret = -ENOENT; 224098d377a0STARUISI Hiroaki goto out; 2241ac8e9819SChris Mason } 224218674c6cSFilipe David Borba Manana } 224318674c6cSFilipe David Borba Manana 224418674c6cSFilipe David Borba Manana l = path->nodes[0]; 224518674c6cSFilipe David Borba Manana slot = path->slots[0]; 224618674c6cSFilipe David Borba Manana btrfs_item_key_to_cpu(l, &key, slot); 224798d377a0STARUISI Hiroaki 224898d377a0STARUISI Hiroaki iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); 224998d377a0STARUISI Hiroaki len = btrfs_inode_ref_name_len(l, iref); 225098d377a0STARUISI Hiroaki ptr -= len + 1; 225198d377a0STARUISI Hiroaki total_len += len + 1; 2252a696cf35SFilipe David Borba Manana if (ptr < name) { 2253a696cf35SFilipe David Borba Manana ret = -ENAMETOOLONG; 225498d377a0STARUISI Hiroaki goto out; 2255a696cf35SFilipe David Borba Manana } 225698d377a0STARUISI Hiroaki 225798d377a0STARUISI Hiroaki *(ptr + len) = '/'; 225898d377a0STARUISI Hiroaki read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len); 225998d377a0STARUISI Hiroaki 226098d377a0STARUISI Hiroaki if (key.offset == BTRFS_FIRST_FREE_OBJECTID) 226198d377a0STARUISI Hiroaki break; 226298d377a0STARUISI Hiroaki 2263b3b4aa74SDavid Sterba btrfs_release_path(path); 226498d377a0STARUISI Hiroaki key.objectid = key.offset; 22658ad6fcabSChris Mason key.offset = (u64)-1; 226698d377a0STARUISI Hiroaki dirid = key.objectid; 226798d377a0STARUISI Hiroaki } 226877906a50SLi Zefan memmove(name, ptr, total_len); 226998d377a0STARUISI Hiroaki name[total_len] = '\0'; 227098d377a0STARUISI Hiroaki ret = 0; 227198d377a0STARUISI Hiroaki out: 227298d377a0STARUISI Hiroaki btrfs_free_path(path); 2273ac8e9819SChris Mason return ret; 2274ac8e9819SChris Mason } 2275ac8e9819SChris Mason 2276ac8e9819SChris Mason static noinline int btrfs_ioctl_ino_lookup(struct file *file, 2277ac8e9819SChris Mason void __user *argp) 2278ac8e9819SChris Mason { 2279ac8e9819SChris Mason struct btrfs_ioctl_ino_lookup_args *args; 2280ac8e9819SChris Mason struct inode *inode; 228101b810b8SDavid Sterba int ret = 0; 2282ac8e9819SChris Mason 22832354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 22842354d08fSJulia Lawall if (IS_ERR(args)) 22852354d08fSJulia Lawall return PTR_ERR(args); 2286c2b96929SDan Carpenter 2287496ad9aaSAl Viro inode = file_inode(file); 2288ac8e9819SChris Mason 228901b810b8SDavid Sterba /* 229001b810b8SDavid Sterba * Unprivileged query to obtain the containing subvolume root id. The 229101b810b8SDavid Sterba * path is reset so it's consistent with btrfs_search_path_in_tree. 229201b810b8SDavid Sterba */ 22931b53ac4dSChris Mason if (args->treeid == 0) 22941b53ac4dSChris Mason args->treeid = BTRFS_I(inode)->root->root_key.objectid; 22951b53ac4dSChris Mason 229601b810b8SDavid Sterba if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) { 229701b810b8SDavid Sterba args->name[0] = 0; 229801b810b8SDavid Sterba goto out; 229901b810b8SDavid Sterba } 230001b810b8SDavid Sterba 230101b810b8SDavid Sterba if (!capable(CAP_SYS_ADMIN)) { 230201b810b8SDavid Sterba ret = -EPERM; 230301b810b8SDavid Sterba goto out; 230401b810b8SDavid Sterba } 230501b810b8SDavid Sterba 2306ac8e9819SChris Mason ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, 2307ac8e9819SChris Mason args->treeid, args->objectid, 2308ac8e9819SChris Mason args->name); 2309ac8e9819SChris Mason 231001b810b8SDavid Sterba out: 2311ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 2312ac8e9819SChris Mason ret = -EFAULT; 2313ac8e9819SChris Mason 2314ac8e9819SChris Mason kfree(args); 231598d377a0STARUISI Hiroaki return ret; 231698d377a0STARUISI Hiroaki } 231798d377a0STARUISI Hiroaki 231876dda93cSYan, Zheng static noinline int btrfs_ioctl_snap_destroy(struct file *file, 231976dda93cSYan, Zheng void __user *arg) 232076dda93cSYan, Zheng { 232154563d41SAl Viro struct dentry *parent = file->f_path.dentry; 232276dda93cSYan, Zheng struct dentry *dentry; 23232b0143b5SDavid Howells struct inode *dir = d_inode(parent); 232476dda93cSYan, Zheng struct inode *inode; 232576dda93cSYan, Zheng struct btrfs_root *root = BTRFS_I(dir)->root; 232676dda93cSYan, Zheng struct btrfs_root *dest = NULL; 232776dda93cSYan, Zheng struct btrfs_ioctl_vol_args *vol_args; 232876dda93cSYan, Zheng struct btrfs_trans_handle *trans; 2329c58aaad2SMiao Xie struct btrfs_block_rsv block_rsv; 2330521e0546SDavid Sterba u64 root_flags; 2331c58aaad2SMiao Xie u64 qgroup_reserved; 233276dda93cSYan, Zheng int namelen; 233376dda93cSYan, Zheng int ret; 233476dda93cSYan, Zheng int err = 0; 233576dda93cSYan, Zheng 233676dda93cSYan, Zheng vol_args = memdup_user(arg, sizeof(*vol_args)); 233776dda93cSYan, Zheng if (IS_ERR(vol_args)) 233876dda93cSYan, Zheng return PTR_ERR(vol_args); 233976dda93cSYan, Zheng 234076dda93cSYan, Zheng vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 234176dda93cSYan, Zheng namelen = strlen(vol_args->name); 234276dda93cSYan, Zheng if (strchr(vol_args->name, '/') || 234376dda93cSYan, Zheng strncmp(vol_args->name, "..", namelen) == 0) { 234476dda93cSYan, Zheng err = -EINVAL; 234576dda93cSYan, Zheng goto out; 234676dda93cSYan, Zheng } 234776dda93cSYan, Zheng 2348a561be71SAl Viro err = mnt_want_write_file(file); 234976dda93cSYan, Zheng if (err) 235076dda93cSYan, Zheng goto out; 235176dda93cSYan, Zheng 2352521e0546SDavid Sterba 23535c50c9b8SDavid Sterba err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT); 23545c50c9b8SDavid Sterba if (err == -EINTR) 2355e43f998eSDavid Sterba goto out_drop_write; 235676dda93cSYan, Zheng dentry = lookup_one_len(vol_args->name, parent, namelen); 235776dda93cSYan, Zheng if (IS_ERR(dentry)) { 235876dda93cSYan, Zheng err = PTR_ERR(dentry); 235976dda93cSYan, Zheng goto out_unlock_dir; 236076dda93cSYan, Zheng } 236176dda93cSYan, Zheng 23622b0143b5SDavid Howells if (d_really_is_negative(dentry)) { 236376dda93cSYan, Zheng err = -ENOENT; 236476dda93cSYan, Zheng goto out_dput; 236576dda93cSYan, Zheng } 236676dda93cSYan, Zheng 23672b0143b5SDavid Howells inode = d_inode(dentry); 23684260f7c7SSage Weil dest = BTRFS_I(inode)->root; 23694260f7c7SSage Weil if (!capable(CAP_SYS_ADMIN)) { 23704260f7c7SSage Weil /* 23714260f7c7SSage Weil * Regular user. Only allow this with a special mount 23724260f7c7SSage Weil * option, when the user has write+exec access to the 23734260f7c7SSage Weil * subvol root, and when rmdir(2) would have been 23744260f7c7SSage Weil * allowed. 23754260f7c7SSage Weil * 23764260f7c7SSage Weil * Note that this is _not_ check that the subvol is 23774260f7c7SSage Weil * empty or doesn't contain data that we wouldn't 23784260f7c7SSage Weil * otherwise be able to delete. 23794260f7c7SSage Weil * 23804260f7c7SSage Weil * Users who want to delete empty subvols should try 23814260f7c7SSage Weil * rmdir(2). 23824260f7c7SSage Weil */ 23834260f7c7SSage Weil err = -EPERM; 23844260f7c7SSage Weil if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED)) 23854260f7c7SSage Weil goto out_dput; 23864260f7c7SSage Weil 23874260f7c7SSage Weil /* 23884260f7c7SSage Weil * Do not allow deletion if the parent dir is the same 23894260f7c7SSage Weil * as the dir to be deleted. That means the ioctl 23904260f7c7SSage Weil * must be called on the dentry referencing the root 23914260f7c7SSage Weil * of the subvol, not a random directory contained 23924260f7c7SSage Weil * within it. 23934260f7c7SSage Weil */ 23944260f7c7SSage Weil err = -EINVAL; 23954260f7c7SSage Weil if (root == dest) 23964260f7c7SSage Weil goto out_dput; 23974260f7c7SSage Weil 23984260f7c7SSage Weil err = inode_permission(inode, MAY_WRITE | MAY_EXEC); 23994260f7c7SSage Weil if (err) 24004260f7c7SSage Weil goto out_dput; 24015c39da5bSMiao Xie } 24024260f7c7SSage Weil 24035c39da5bSMiao Xie /* check if subvolume may be deleted by a user */ 24044260f7c7SSage Weil err = btrfs_may_delete(dir, dentry, 1); 24054260f7c7SSage Weil if (err) 24064260f7c7SSage Weil goto out_dput; 24074260f7c7SSage Weil 240833345d01SLi Zefan if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { 240976dda93cSYan, Zheng err = -EINVAL; 241076dda93cSYan, Zheng goto out_dput; 241176dda93cSYan, Zheng } 241276dda93cSYan, Zheng 241376dda93cSYan, Zheng mutex_lock(&inode->i_mutex); 2414521e0546SDavid Sterba 2415521e0546SDavid Sterba /* 2416521e0546SDavid Sterba * Don't allow to delete a subvolume with send in progress. This is 2417521e0546SDavid Sterba * inside the i_mutex so the error handling that has to drop the bit 2418521e0546SDavid Sterba * again is not run concurrently. 2419521e0546SDavid Sterba */ 2420521e0546SDavid Sterba spin_lock(&dest->root_item_lock); 2421c55bfa67SFilipe Manana root_flags = btrfs_root_flags(&dest->root_item); 2422c55bfa67SFilipe Manana if (dest->send_in_progress == 0) { 2423c55bfa67SFilipe Manana btrfs_set_root_flags(&dest->root_item, 2424521e0546SDavid Sterba root_flags | BTRFS_ROOT_SUBVOL_DEAD); 2425521e0546SDavid Sterba spin_unlock(&dest->root_item_lock); 2426521e0546SDavid Sterba } else { 2427521e0546SDavid Sterba spin_unlock(&dest->root_item_lock); 2428521e0546SDavid Sterba btrfs_warn(root->fs_info, 2429521e0546SDavid Sterba "Attempt to delete subvolume %llu during send", 2430c55bfa67SFilipe Manana dest->root_key.objectid); 2431521e0546SDavid Sterba err = -EPERM; 2432909e26dcSOmar Sandoval goto out_unlock_inode; 2433521e0546SDavid Sterba } 2434521e0546SDavid Sterba 243576dda93cSYan, Zheng down_write(&root->fs_info->subvol_sem); 243676dda93cSYan, Zheng 243776dda93cSYan, Zheng err = may_destroy_subvol(dest); 243876dda93cSYan, Zheng if (err) 243976dda93cSYan, Zheng goto out_up_write; 244076dda93cSYan, Zheng 2441c58aaad2SMiao Xie btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); 2442c58aaad2SMiao Xie /* 2443c58aaad2SMiao Xie * One for dir inode, two for dir entries, two for root 2444c58aaad2SMiao Xie * ref/backref. 2445c58aaad2SMiao Xie */ 2446c58aaad2SMiao Xie err = btrfs_subvolume_reserve_metadata(root, &block_rsv, 2447ee3441b4SJeff Mahoney 5, &qgroup_reserved, true); 2448c58aaad2SMiao Xie if (err) 2449c58aaad2SMiao Xie goto out_up_write; 2450c58aaad2SMiao Xie 2451a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 2452a22285a6SYan, Zheng if (IS_ERR(trans)) { 2453a22285a6SYan, Zheng err = PTR_ERR(trans); 2454c58aaad2SMiao Xie goto out_release; 2455a22285a6SYan, Zheng } 2456c58aaad2SMiao Xie trans->block_rsv = &block_rsv; 2457c58aaad2SMiao Xie trans->bytes_reserved = block_rsv.size; 2458a22285a6SYan, Zheng 245976dda93cSYan, Zheng ret = btrfs_unlink_subvol(trans, root, dir, 246076dda93cSYan, Zheng dest->root_key.objectid, 246176dda93cSYan, Zheng dentry->d_name.name, 246276dda93cSYan, Zheng dentry->d_name.len); 246379787eaaSJeff Mahoney if (ret) { 246479787eaaSJeff Mahoney err = ret; 246579787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, ret); 246679787eaaSJeff Mahoney goto out_end_trans; 246779787eaaSJeff Mahoney } 246876dda93cSYan, Zheng 246976dda93cSYan, Zheng btrfs_record_root_in_trans(trans, dest); 247076dda93cSYan, Zheng 247176dda93cSYan, Zheng memset(&dest->root_item.drop_progress, 0, 247276dda93cSYan, Zheng sizeof(dest->root_item.drop_progress)); 247376dda93cSYan, Zheng dest->root_item.drop_level = 0; 247476dda93cSYan, Zheng btrfs_set_root_refs(&dest->root_item, 0); 247576dda93cSYan, Zheng 247627cdeb70SMiao Xie if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) { 247776dda93cSYan, Zheng ret = btrfs_insert_orphan_item(trans, 247876dda93cSYan, Zheng root->fs_info->tree_root, 247976dda93cSYan, Zheng dest->root_key.objectid); 248079787eaaSJeff Mahoney if (ret) { 248179787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, ret); 248279787eaaSJeff Mahoney err = ret; 248379787eaaSJeff Mahoney goto out_end_trans; 2484d68fc57bSYan, Zheng } 248579787eaaSJeff Mahoney } 2486dd5f9615SStefan Behrens 2487dd5f9615SStefan Behrens ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root, 2488dd5f9615SStefan Behrens dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL, 2489dd5f9615SStefan Behrens dest->root_key.objectid); 2490dd5f9615SStefan Behrens if (ret && ret != -ENOENT) { 2491dd5f9615SStefan Behrens btrfs_abort_transaction(trans, root, ret); 2492dd5f9615SStefan Behrens err = ret; 2493dd5f9615SStefan Behrens goto out_end_trans; 2494dd5f9615SStefan Behrens } 2495dd5f9615SStefan Behrens if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) { 2496dd5f9615SStefan Behrens ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root, 2497dd5f9615SStefan Behrens dest->root_item.received_uuid, 2498dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 2499dd5f9615SStefan Behrens dest->root_key.objectid); 2500dd5f9615SStefan Behrens if (ret && ret != -ENOENT) { 2501dd5f9615SStefan Behrens btrfs_abort_transaction(trans, root, ret); 2502dd5f9615SStefan Behrens err = ret; 2503dd5f9615SStefan Behrens goto out_end_trans; 2504dd5f9615SStefan Behrens } 2505dd5f9615SStefan Behrens } 2506dd5f9615SStefan Behrens 250779787eaaSJeff Mahoney out_end_trans: 2508c58aaad2SMiao Xie trans->block_rsv = NULL; 2509c58aaad2SMiao Xie trans->bytes_reserved = 0; 2510531cb13fSSage Weil ret = btrfs_end_transaction(trans, root); 251179787eaaSJeff Mahoney if (ret && !err) 251279787eaaSJeff Mahoney err = ret; 251376dda93cSYan, Zheng inode->i_flags |= S_DEAD; 2514c58aaad2SMiao Xie out_release: 2515c58aaad2SMiao Xie btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved); 251676dda93cSYan, Zheng out_up_write: 251776dda93cSYan, Zheng up_write(&root->fs_info->subvol_sem); 2518521e0546SDavid Sterba if (err) { 2519521e0546SDavid Sterba spin_lock(&dest->root_item_lock); 2520c55bfa67SFilipe Manana root_flags = btrfs_root_flags(&dest->root_item); 2521c55bfa67SFilipe Manana btrfs_set_root_flags(&dest->root_item, 2522521e0546SDavid Sterba root_flags & ~BTRFS_ROOT_SUBVOL_DEAD); 2523521e0546SDavid Sterba spin_unlock(&dest->root_item_lock); 2524521e0546SDavid Sterba } 2525909e26dcSOmar Sandoval out_unlock_inode: 252676dda93cSYan, Zheng mutex_unlock(&inode->i_mutex); 252776dda93cSYan, Zheng if (!err) { 252864ad6c48SOmar Sandoval d_invalidate(dentry); 252976dda93cSYan, Zheng btrfs_invalidate_inodes(dest); 253076dda93cSYan, Zheng d_delete(dentry); 253161155aa0SDavid Sterba ASSERT(dest->send_in_progress == 0); 2532fa6ac876SLiu Bo 2533fa6ac876SLiu Bo /* the last ref */ 253457cdc8dbSDavid Sterba if (dest->ino_cache_inode) { 253557cdc8dbSDavid Sterba iput(dest->ino_cache_inode); 253657cdc8dbSDavid Sterba dest->ino_cache_inode = NULL; 2537fa6ac876SLiu Bo } 253876dda93cSYan, Zheng } 253976dda93cSYan, Zheng out_dput: 254076dda93cSYan, Zheng dput(dentry); 254176dda93cSYan, Zheng out_unlock_dir: 254276dda93cSYan, Zheng mutex_unlock(&dir->i_mutex); 2543e43f998eSDavid Sterba out_drop_write: 25442a79f17eSAl Viro mnt_drop_write_file(file); 254576dda93cSYan, Zheng out: 254676dda93cSYan, Zheng kfree(vol_args); 254776dda93cSYan, Zheng return err; 254876dda93cSYan, Zheng } 254976dda93cSYan, Zheng 25501e701a32SChris Mason static int btrfs_ioctl_defrag(struct file *file, void __user *argp) 2551f46b5a66SChristoph Hellwig { 2552496ad9aaSAl Viro struct inode *inode = file_inode(file); 2553f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 25541e701a32SChris Mason struct btrfs_ioctl_defrag_range_args *range; 2555c146afadSYan Zheng int ret; 2556c146afadSYan Zheng 255725122d15SIlya Dryomov ret = mnt_want_write_file(file); 255825122d15SIlya Dryomov if (ret) 255925122d15SIlya Dryomov return ret; 2560b83cc969SLi Zefan 256125122d15SIlya Dryomov if (btrfs_root_readonly(root)) { 256225122d15SIlya Dryomov ret = -EROFS; 256325122d15SIlya Dryomov goto out; 25645ac00addSStefan Behrens } 2565f46b5a66SChristoph Hellwig 2566f46b5a66SChristoph Hellwig switch (inode->i_mode & S_IFMT) { 2567f46b5a66SChristoph Hellwig case S_IFDIR: 2568e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) { 2569e441d54dSChris Mason ret = -EPERM; 2570e441d54dSChris Mason goto out; 2571e441d54dSChris Mason } 2572de78b51aSEric Sandeen ret = btrfs_defrag_root(root); 25738929ecfaSYan, Zheng if (ret) 25748929ecfaSYan, Zheng goto out; 2575de78b51aSEric Sandeen ret = btrfs_defrag_root(root->fs_info->extent_root); 2576f46b5a66SChristoph Hellwig break; 2577f46b5a66SChristoph Hellwig case S_IFREG: 2578e441d54dSChris Mason if (!(file->f_mode & FMODE_WRITE)) { 2579e441d54dSChris Mason ret = -EINVAL; 2580e441d54dSChris Mason goto out; 2581e441d54dSChris Mason } 25821e701a32SChris Mason 25831e701a32SChris Mason range = kzalloc(sizeof(*range), GFP_KERNEL); 25841e701a32SChris Mason if (!range) { 25851e701a32SChris Mason ret = -ENOMEM; 25861e701a32SChris Mason goto out; 25871e701a32SChris Mason } 25881e701a32SChris Mason 25891e701a32SChris Mason if (argp) { 25901e701a32SChris Mason if (copy_from_user(range, argp, 25911e701a32SChris Mason sizeof(*range))) { 25921e701a32SChris Mason ret = -EFAULT; 25931e701a32SChris Mason kfree(range); 2594683be16eSDan Carpenter goto out; 25951e701a32SChris Mason } 25961e701a32SChris Mason /* compression requires us to start the IO */ 25971e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 25981e701a32SChris Mason range->flags |= BTRFS_DEFRAG_RANGE_START_IO; 25991e701a32SChris Mason range->extent_thresh = (u32)-1; 26001e701a32SChris Mason } 26011e701a32SChris Mason } else { 26021e701a32SChris Mason /* the rest are all set to zero by kzalloc */ 26031e701a32SChris Mason range->len = (u64)-1; 26041e701a32SChris Mason } 2605496ad9aaSAl Viro ret = btrfs_defrag_file(file_inode(file), file, 26064cb5300bSChris Mason range, 0, 0); 26074cb5300bSChris Mason if (ret > 0) 26084cb5300bSChris Mason ret = 0; 26091e701a32SChris Mason kfree(range); 2610f46b5a66SChristoph Hellwig break; 26118929ecfaSYan, Zheng default: 26128929ecfaSYan, Zheng ret = -EINVAL; 2613f46b5a66SChristoph Hellwig } 2614e441d54dSChris Mason out: 261525122d15SIlya Dryomov mnt_drop_write_file(file); 2616e441d54dSChris Mason return ret; 2617f46b5a66SChristoph Hellwig } 2618f46b5a66SChristoph Hellwig 2619b2950863SChristoph Hellwig static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg) 2620f46b5a66SChristoph Hellwig { 2621f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2622f46b5a66SChristoph Hellwig int ret; 2623f46b5a66SChristoph Hellwig 2624e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2625e441d54dSChris Mason return -EPERM; 2626e441d54dSChris Mason 26275ac00addSStefan Behrens if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running, 26285ac00addSStefan Behrens 1)) { 2629e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 2630c9e9f97bSIlya Dryomov } 2631c9e9f97bSIlya Dryomov 26325ac00addSStefan Behrens mutex_lock(&root->fs_info->volume_mutex); 2633dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2634c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 2635c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 2636c9e9f97bSIlya Dryomov goto out; 2637c9e9f97bSIlya Dryomov } 2638f46b5a66SChristoph Hellwig 26395516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 2640f46b5a66SChristoph Hellwig ret = btrfs_init_new_device(root, vol_args->name); 2641f46b5a66SChristoph Hellwig 264243d20761SAnand Jain if (!ret) 264343d20761SAnand Jain btrfs_info(root->fs_info, "disk added %s",vol_args->name); 264443d20761SAnand Jain 2645f46b5a66SChristoph Hellwig kfree(vol_args); 2646c9e9f97bSIlya Dryomov out: 2647c9e9f97bSIlya Dryomov mutex_unlock(&root->fs_info->volume_mutex); 26485ac00addSStefan Behrens atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0); 2649f46b5a66SChristoph Hellwig return ret; 2650f46b5a66SChristoph Hellwig } 2651f46b5a66SChristoph Hellwig 2652da24927bSMiao Xie static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg) 2653f46b5a66SChristoph Hellwig { 2654496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 2655f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2656f46b5a66SChristoph Hellwig int ret; 2657f46b5a66SChristoph Hellwig 2658e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2659e441d54dSChris Mason return -EPERM; 2660e441d54dSChris Mason 2661da24927bSMiao Xie ret = mnt_want_write_file(file); 2662da24927bSMiao Xie if (ret) 2663da24927bSMiao Xie return ret; 2664c146afadSYan Zheng 2665dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2666c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 2667c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 2668c47ca32dSDan Carpenter goto err_drop; 2669c9e9f97bSIlya Dryomov } 2670f46b5a66SChristoph Hellwig 26715516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 2672f46b5a66SChristoph Hellwig 2673183860f6SAnand Jain if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running, 2674183860f6SAnand Jain 1)) { 2675183860f6SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 2676183860f6SAnand Jain goto out; 2677183860f6SAnand Jain } 2678183860f6SAnand Jain 2679183860f6SAnand Jain mutex_lock(&root->fs_info->volume_mutex); 2680183860f6SAnand Jain ret = btrfs_rm_device(root, vol_args->name); 2681c9e9f97bSIlya Dryomov mutex_unlock(&root->fs_info->volume_mutex); 26825ac00addSStefan Behrens atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0); 2683183860f6SAnand Jain 2684ec95d491SAnand Jain if (!ret) 2685ec95d491SAnand Jain btrfs_info(root->fs_info, "disk deleted %s",vol_args->name); 2686ec95d491SAnand Jain 2687183860f6SAnand Jain out: 2688183860f6SAnand Jain kfree(vol_args); 2689c47ca32dSDan Carpenter err_drop: 26904ac20c70SIlya Dryomov mnt_drop_write_file(file); 2691f46b5a66SChristoph Hellwig return ret; 2692f46b5a66SChristoph Hellwig } 2693f46b5a66SChristoph Hellwig 2694475f6387SJan Schmidt static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg) 2695475f6387SJan Schmidt { 2696027ed2f0SLi Zefan struct btrfs_ioctl_fs_info_args *fi_args; 2697475f6387SJan Schmidt struct btrfs_device *device; 2698475f6387SJan Schmidt struct btrfs_device *next; 2699475f6387SJan Schmidt struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; 2700027ed2f0SLi Zefan int ret = 0; 2701475f6387SJan Schmidt 2702027ed2f0SLi Zefan fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL); 2703027ed2f0SLi Zefan if (!fi_args) 2704027ed2f0SLi Zefan return -ENOMEM; 2705027ed2f0SLi Zefan 2706f7171750SFilipe David Borba Manana mutex_lock(&fs_devices->device_list_mutex); 2707027ed2f0SLi Zefan fi_args->num_devices = fs_devices->num_devices; 2708027ed2f0SLi Zefan memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid)); 2709475f6387SJan Schmidt 2710475f6387SJan Schmidt list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) { 2711027ed2f0SLi Zefan if (device->devid > fi_args->max_id) 2712027ed2f0SLi Zefan fi_args->max_id = device->devid; 2713475f6387SJan Schmidt } 2714475f6387SJan Schmidt mutex_unlock(&fs_devices->device_list_mutex); 2715475f6387SJan Schmidt 271680a773fbSDavid Sterba fi_args->nodesize = root->fs_info->super_copy->nodesize; 271780a773fbSDavid Sterba fi_args->sectorsize = root->fs_info->super_copy->sectorsize; 271880a773fbSDavid Sterba fi_args->clone_alignment = root->fs_info->super_copy->sectorsize; 271980a773fbSDavid Sterba 2720027ed2f0SLi Zefan if (copy_to_user(arg, fi_args, sizeof(*fi_args))) 2721027ed2f0SLi Zefan ret = -EFAULT; 2722475f6387SJan Schmidt 2723027ed2f0SLi Zefan kfree(fi_args); 2724027ed2f0SLi Zefan return ret; 2725475f6387SJan Schmidt } 2726475f6387SJan Schmidt 2727475f6387SJan Schmidt static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg) 2728475f6387SJan Schmidt { 2729475f6387SJan Schmidt struct btrfs_ioctl_dev_info_args *di_args; 2730475f6387SJan Schmidt struct btrfs_device *dev; 2731475f6387SJan Schmidt struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; 2732475f6387SJan Schmidt int ret = 0; 2733475f6387SJan Schmidt char *s_uuid = NULL; 2734475f6387SJan Schmidt 2735475f6387SJan Schmidt di_args = memdup_user(arg, sizeof(*di_args)); 2736475f6387SJan Schmidt if (IS_ERR(di_args)) 2737475f6387SJan Schmidt return PTR_ERR(di_args); 2738475f6387SJan Schmidt 2739dd5f9615SStefan Behrens if (!btrfs_is_empty_uuid(di_args->uuid)) 2740475f6387SJan Schmidt s_uuid = di_args->uuid; 2741475f6387SJan Schmidt 2742475f6387SJan Schmidt mutex_lock(&fs_devices->device_list_mutex); 2743aa1b8cd4SStefan Behrens dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL); 2744475f6387SJan Schmidt 2745475f6387SJan Schmidt if (!dev) { 2746475f6387SJan Schmidt ret = -ENODEV; 2747475f6387SJan Schmidt goto out; 2748475f6387SJan Schmidt } 2749475f6387SJan Schmidt 2750475f6387SJan Schmidt di_args->devid = dev->devid; 27517cc8e58dSMiao Xie di_args->bytes_used = btrfs_device_get_bytes_used(dev); 27527cc8e58dSMiao Xie di_args->total_bytes = btrfs_device_get_total_bytes(dev); 2753475f6387SJan Schmidt memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); 2754a27202fbSJim Meyering if (dev->name) { 2755606686eeSJosef Bacik struct rcu_string *name; 2756606686eeSJosef Bacik 2757606686eeSJosef Bacik rcu_read_lock(); 2758606686eeSJosef Bacik name = rcu_dereference(dev->name); 2759606686eeSJosef Bacik strncpy(di_args->path, name->str, sizeof(di_args->path)); 2760606686eeSJosef Bacik rcu_read_unlock(); 2761a27202fbSJim Meyering di_args->path[sizeof(di_args->path) - 1] = 0; 2762a27202fbSJim Meyering } else { 276399ba55adSStefan Behrens di_args->path[0] = '\0'; 2764a27202fbSJim Meyering } 2765475f6387SJan Schmidt 2766475f6387SJan Schmidt out: 276755793c0dSDavid Sterba mutex_unlock(&fs_devices->device_list_mutex); 2768475f6387SJan Schmidt if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) 2769475f6387SJan Schmidt ret = -EFAULT; 2770475f6387SJan Schmidt 2771475f6387SJan Schmidt kfree(di_args); 2772475f6387SJan Schmidt return ret; 2773475f6387SJan Schmidt } 2774475f6387SJan Schmidt 2775f4414602SMark Fasheh static struct page *extent_same_get_page(struct inode *inode, pgoff_t index) 2776416161dbSMark Fasheh { 2777416161dbSMark Fasheh struct page *page; 2778416161dbSMark Fasheh struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree; 2779416161dbSMark Fasheh 2780416161dbSMark Fasheh page = grab_cache_page(inode->i_mapping, index); 2781416161dbSMark Fasheh if (!page) 2782416161dbSMark Fasheh return NULL; 2783416161dbSMark Fasheh 2784416161dbSMark Fasheh if (!PageUptodate(page)) { 2785416161dbSMark Fasheh if (extent_read_full_page_nolock(tree, page, btrfs_get_extent, 2786416161dbSMark Fasheh 0)) 2787416161dbSMark Fasheh return NULL; 2788416161dbSMark Fasheh lock_page(page); 2789416161dbSMark Fasheh if (!PageUptodate(page)) { 2790416161dbSMark Fasheh unlock_page(page); 2791416161dbSMark Fasheh page_cache_release(page); 2792416161dbSMark Fasheh return NULL; 2793416161dbSMark Fasheh } 2794416161dbSMark Fasheh } 2795416161dbSMark Fasheh unlock_page(page); 2796416161dbSMark Fasheh 2797416161dbSMark Fasheh return page; 2798416161dbSMark Fasheh } 2799416161dbSMark Fasheh 2800f4414602SMark Fasheh static int gather_extent_pages(struct inode *inode, struct page **pages, 2801f4414602SMark Fasheh int num_pages, u64 off) 2802f4414602SMark Fasheh { 2803f4414602SMark Fasheh int i; 2804f4414602SMark Fasheh pgoff_t index = off >> PAGE_CACHE_SHIFT; 2805f4414602SMark Fasheh 2806f4414602SMark Fasheh for (i = 0; i < num_pages; i++) { 2807f4414602SMark Fasheh pages[i] = extent_same_get_page(inode, index + i); 2808f4414602SMark Fasheh if (!pages[i]) 2809f4414602SMark Fasheh return -ENOMEM; 2810f4414602SMark Fasheh } 2811f4414602SMark Fasheh return 0; 2812f4414602SMark Fasheh } 2813f4414602SMark Fasheh 281477fe20dcSMark Fasheh static inline void lock_extent_range(struct inode *inode, u64 off, u64 len) 281577fe20dcSMark Fasheh { 281677fe20dcSMark Fasheh /* do any pending delalloc/csum calc on src, one way or 281777fe20dcSMark Fasheh another, and lock file content */ 281877fe20dcSMark Fasheh while (1) { 281977fe20dcSMark Fasheh struct btrfs_ordered_extent *ordered; 282077fe20dcSMark Fasheh lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); 282177fe20dcSMark Fasheh ordered = btrfs_lookup_first_ordered_extent(inode, 282277fe20dcSMark Fasheh off + len - 1); 2823ff5df9b8SFilipe Manana if ((!ordered || 2824ff5df9b8SFilipe Manana ordered->file_offset + ordered->len <= off || 2825ff5df9b8SFilipe Manana ordered->file_offset >= off + len) && 282677fe20dcSMark Fasheh !test_range_bit(&BTRFS_I(inode)->io_tree, off, 2827ff5df9b8SFilipe Manana off + len - 1, EXTENT_DELALLOC, 0, NULL)) { 2828ff5df9b8SFilipe Manana if (ordered) 2829ff5df9b8SFilipe Manana btrfs_put_ordered_extent(ordered); 283077fe20dcSMark Fasheh break; 2831ff5df9b8SFilipe Manana } 283277fe20dcSMark Fasheh unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); 283377fe20dcSMark Fasheh if (ordered) 283477fe20dcSMark Fasheh btrfs_put_ordered_extent(ordered); 283577fe20dcSMark Fasheh btrfs_wait_ordered_range(inode, off, len); 283677fe20dcSMark Fasheh } 283777fe20dcSMark Fasheh } 283877fe20dcSMark Fasheh 2839f4414602SMark Fasheh static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2) 2840416161dbSMark Fasheh { 2841416161dbSMark Fasheh mutex_unlock(&inode1->i_mutex); 2842416161dbSMark Fasheh mutex_unlock(&inode2->i_mutex); 2843416161dbSMark Fasheh } 2844416161dbSMark Fasheh 2845f4414602SMark Fasheh static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2) 2846f4414602SMark Fasheh { 2847f4414602SMark Fasheh if (inode1 < inode2) 2848f4414602SMark Fasheh swap(inode1, inode2); 2849f4414602SMark Fasheh 2850f4414602SMark Fasheh mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT); 2851f4414602SMark Fasheh if (inode1 != inode2) 2852f4414602SMark Fasheh mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD); 2853f4414602SMark Fasheh } 2854f4414602SMark Fasheh 2855f4414602SMark Fasheh static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1, 2856f4414602SMark Fasheh struct inode *inode2, u64 loff2, u64 len) 2857f4414602SMark Fasheh { 2858f4414602SMark Fasheh unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); 2859f4414602SMark Fasheh unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); 2860f4414602SMark Fasheh } 2861f4414602SMark Fasheh 2862f4414602SMark Fasheh static void btrfs_double_extent_lock(struct inode *inode1, u64 loff1, 2863416161dbSMark Fasheh struct inode *inode2, u64 loff2, u64 len) 2864416161dbSMark Fasheh { 2865416161dbSMark Fasheh if (inode1 < inode2) { 2866416161dbSMark Fasheh swap(inode1, inode2); 2867416161dbSMark Fasheh swap(loff1, loff2); 2868416161dbSMark Fasheh } 2869416161dbSMark Fasheh lock_extent_range(inode1, loff1, len); 2870f4414602SMark Fasheh if (inode1 != inode2) 2871416161dbSMark Fasheh lock_extent_range(inode2, loff2, len); 2872416161dbSMark Fasheh } 2873f4414602SMark Fasheh 2874f4414602SMark Fasheh struct cmp_pages { 2875f4414602SMark Fasheh int num_pages; 2876f4414602SMark Fasheh struct page **src_pages; 2877f4414602SMark Fasheh struct page **dst_pages; 2878f4414602SMark Fasheh }; 2879f4414602SMark Fasheh 2880f4414602SMark Fasheh static void btrfs_cmp_data_free(struct cmp_pages *cmp) 2881f4414602SMark Fasheh { 2882f4414602SMark Fasheh int i; 2883f4414602SMark Fasheh struct page *pg; 2884f4414602SMark Fasheh 2885f4414602SMark Fasheh for (i = 0; i < cmp->num_pages; i++) { 2886f4414602SMark Fasheh pg = cmp->src_pages[i]; 2887f4414602SMark Fasheh if (pg) 2888f4414602SMark Fasheh page_cache_release(pg); 2889f4414602SMark Fasheh pg = cmp->dst_pages[i]; 2890f4414602SMark Fasheh if (pg) 2891f4414602SMark Fasheh page_cache_release(pg); 2892f4414602SMark Fasheh } 2893f4414602SMark Fasheh kfree(cmp->src_pages); 2894f4414602SMark Fasheh kfree(cmp->dst_pages); 2895f4414602SMark Fasheh } 2896f4414602SMark Fasheh 2897f4414602SMark Fasheh static int btrfs_cmp_data_prepare(struct inode *src, u64 loff, 2898f4414602SMark Fasheh struct inode *dst, u64 dst_loff, 2899f4414602SMark Fasheh u64 len, struct cmp_pages *cmp) 2900f4414602SMark Fasheh { 2901f4414602SMark Fasheh int ret; 2902f4414602SMark Fasheh int num_pages = PAGE_CACHE_ALIGN(len) >> PAGE_CACHE_SHIFT; 2903f4414602SMark Fasheh struct page **src_pgarr, **dst_pgarr; 2904f4414602SMark Fasheh 2905f4414602SMark Fasheh /* 2906f4414602SMark Fasheh * We must gather up all the pages before we initiate our 2907f4414602SMark Fasheh * extent locking. We use an array for the page pointers. Size 2908f4414602SMark Fasheh * of the array is bounded by len, which is in turn bounded by 2909f4414602SMark Fasheh * BTRFS_MAX_DEDUPE_LEN. 2910f4414602SMark Fasheh */ 2911f4414602SMark Fasheh src_pgarr = kzalloc(num_pages * sizeof(struct page *), GFP_NOFS); 2912f4414602SMark Fasheh dst_pgarr = kzalloc(num_pages * sizeof(struct page *), GFP_NOFS); 2913f4414602SMark Fasheh if (!src_pgarr || !dst_pgarr) { 2914f4414602SMark Fasheh kfree(src_pgarr); 2915f4414602SMark Fasheh kfree(dst_pgarr); 2916f4414602SMark Fasheh return -ENOMEM; 2917f4414602SMark Fasheh } 2918f4414602SMark Fasheh cmp->num_pages = num_pages; 2919f4414602SMark Fasheh cmp->src_pages = src_pgarr; 2920f4414602SMark Fasheh cmp->dst_pages = dst_pgarr; 2921f4414602SMark Fasheh 2922f4414602SMark Fasheh ret = gather_extent_pages(src, cmp->src_pages, cmp->num_pages, loff); 2923f4414602SMark Fasheh if (ret) 2924f4414602SMark Fasheh goto out; 2925f4414602SMark Fasheh 2926f4414602SMark Fasheh ret = gather_extent_pages(dst, cmp->dst_pages, cmp->num_pages, dst_loff); 2927f4414602SMark Fasheh 2928f4414602SMark Fasheh out: 2929f4414602SMark Fasheh if (ret) 2930f4414602SMark Fasheh btrfs_cmp_data_free(cmp); 2931f4414602SMark Fasheh return 0; 2932416161dbSMark Fasheh } 2933416161dbSMark Fasheh 2934416161dbSMark Fasheh static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst, 2935f4414602SMark Fasheh u64 dst_loff, u64 len, struct cmp_pages *cmp) 2936416161dbSMark Fasheh { 2937416161dbSMark Fasheh int ret = 0; 2938f4414602SMark Fasheh int i; 2939416161dbSMark Fasheh struct page *src_page, *dst_page; 2940416161dbSMark Fasheh unsigned int cmp_len = PAGE_CACHE_SIZE; 2941416161dbSMark Fasheh void *addr, *dst_addr; 2942416161dbSMark Fasheh 2943f4414602SMark Fasheh i = 0; 2944416161dbSMark Fasheh while (len) { 2945416161dbSMark Fasheh if (len < PAGE_CACHE_SIZE) 2946416161dbSMark Fasheh cmp_len = len; 2947416161dbSMark Fasheh 2948f4414602SMark Fasheh BUG_ON(i >= cmp->num_pages); 2949f4414602SMark Fasheh 2950f4414602SMark Fasheh src_page = cmp->src_pages[i]; 2951f4414602SMark Fasheh dst_page = cmp->dst_pages[i]; 2952f4414602SMark Fasheh 2953416161dbSMark Fasheh addr = kmap_atomic(src_page); 2954416161dbSMark Fasheh dst_addr = kmap_atomic(dst_page); 2955416161dbSMark Fasheh 2956416161dbSMark Fasheh flush_dcache_page(src_page); 2957416161dbSMark Fasheh flush_dcache_page(dst_page); 2958416161dbSMark Fasheh 2959416161dbSMark Fasheh if (memcmp(addr, dst_addr, cmp_len)) 2960416161dbSMark Fasheh ret = BTRFS_SAME_DATA_DIFFERS; 2961416161dbSMark Fasheh 2962416161dbSMark Fasheh kunmap_atomic(addr); 2963416161dbSMark Fasheh kunmap_atomic(dst_addr); 2964416161dbSMark Fasheh 2965416161dbSMark Fasheh if (ret) 2966416161dbSMark Fasheh break; 2967416161dbSMark Fasheh 2968416161dbSMark Fasheh len -= cmp_len; 2969f4414602SMark Fasheh i++; 2970416161dbSMark Fasheh } 2971416161dbSMark Fasheh 2972416161dbSMark Fasheh return ret; 2973416161dbSMark Fasheh } 2974416161dbSMark Fasheh 2975e1d227a4SMark Fasheh static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen, 2976e1d227a4SMark Fasheh u64 olen) 2977416161dbSMark Fasheh { 2978e1d227a4SMark Fasheh u64 len = *plen; 2979416161dbSMark Fasheh u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize; 2980416161dbSMark Fasheh 2981e1d227a4SMark Fasheh if (off + olen > inode->i_size || off + olen < off) 2982416161dbSMark Fasheh return -EINVAL; 2983e1d227a4SMark Fasheh 2984e1d227a4SMark Fasheh /* if we extend to eof, continue to block boundary */ 2985e1d227a4SMark Fasheh if (off + len == inode->i_size) 2986e1d227a4SMark Fasheh *plen = len = ALIGN(inode->i_size, bs) - off; 2987e1d227a4SMark Fasheh 2988416161dbSMark Fasheh /* Check that we are block aligned - btrfs_clone() requires this */ 2989416161dbSMark Fasheh if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs)) 2990416161dbSMark Fasheh return -EINVAL; 2991416161dbSMark Fasheh 2992416161dbSMark Fasheh return 0; 2993416161dbSMark Fasheh } 2994416161dbSMark Fasheh 2995e1d227a4SMark Fasheh static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen, 2996416161dbSMark Fasheh struct inode *dst, u64 dst_loff) 2997416161dbSMark Fasheh { 2998416161dbSMark Fasheh int ret; 2999e1d227a4SMark Fasheh u64 len = olen; 3000f4414602SMark Fasheh struct cmp_pages cmp; 30010efa9f48SMark Fasheh int same_inode = 0; 30020efa9f48SMark Fasheh u64 same_lock_start = 0; 30030efa9f48SMark Fasheh u64 same_lock_len = 0; 3004416161dbSMark Fasheh 3005416161dbSMark Fasheh if (src == dst) 30060efa9f48SMark Fasheh same_inode = 1; 3007416161dbSMark Fasheh 3008113e8283SFilipe Manana if (len == 0) 3009113e8283SFilipe Manana return 0; 3010113e8283SFilipe Manana 30110efa9f48SMark Fasheh if (same_inode) { 30120efa9f48SMark Fasheh mutex_lock(&src->i_mutex); 30130efa9f48SMark Fasheh 30140efa9f48SMark Fasheh ret = extent_same_check_offsets(src, loff, &len, olen); 30150efa9f48SMark Fasheh if (ret) 30160efa9f48SMark Fasheh goto out_unlock; 30170efa9f48SMark Fasheh 30180efa9f48SMark Fasheh /* 30190efa9f48SMark Fasheh * Single inode case wants the same checks, except we 30200efa9f48SMark Fasheh * don't want our length pushed out past i_size as 30210efa9f48SMark Fasheh * comparing that data range makes no sense. 30220efa9f48SMark Fasheh * 30230efa9f48SMark Fasheh * extent_same_check_offsets() will do this for an 30240efa9f48SMark Fasheh * unaligned length at i_size, so catch it here and 30250efa9f48SMark Fasheh * reject the request. 30260efa9f48SMark Fasheh * 30270efa9f48SMark Fasheh * This effectively means we require aligned extents 30280efa9f48SMark Fasheh * for the single-inode case, whereas the other cases 30290efa9f48SMark Fasheh * allow an unaligned length so long as it ends at 30300efa9f48SMark Fasheh * i_size. 30310efa9f48SMark Fasheh */ 30320efa9f48SMark Fasheh if (len != olen) { 30330efa9f48SMark Fasheh ret = -EINVAL; 30340efa9f48SMark Fasheh goto out_unlock; 30350efa9f48SMark Fasheh } 30360efa9f48SMark Fasheh 30370efa9f48SMark Fasheh /* Check for overlapping ranges */ 30380efa9f48SMark Fasheh if (dst_loff + len > loff && dst_loff < loff + len) { 30390efa9f48SMark Fasheh ret = -EINVAL; 30400efa9f48SMark Fasheh goto out_unlock; 30410efa9f48SMark Fasheh } 30420efa9f48SMark Fasheh 30430efa9f48SMark Fasheh same_lock_start = min_t(u64, loff, dst_loff); 30440efa9f48SMark Fasheh same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start; 30450efa9f48SMark Fasheh } else { 3046f4414602SMark Fasheh btrfs_double_inode_lock(src, dst); 3047416161dbSMark Fasheh 3048e1d227a4SMark Fasheh ret = extent_same_check_offsets(src, loff, &len, olen); 3049416161dbSMark Fasheh if (ret) 3050416161dbSMark Fasheh goto out_unlock; 3051416161dbSMark Fasheh 3052e1d227a4SMark Fasheh ret = extent_same_check_offsets(dst, dst_loff, &len, olen); 3053416161dbSMark Fasheh if (ret) 3054416161dbSMark Fasheh goto out_unlock; 30550efa9f48SMark Fasheh } 3056416161dbSMark Fasheh 3057416161dbSMark Fasheh /* don't make the dst file partly checksummed */ 3058416161dbSMark Fasheh if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != 3059416161dbSMark Fasheh (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) { 3060416161dbSMark Fasheh ret = -EINVAL; 3061416161dbSMark Fasheh goto out_unlock; 3062416161dbSMark Fasheh } 3063416161dbSMark Fasheh 3064f4414602SMark Fasheh ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, &cmp); 3065f4414602SMark Fasheh if (ret) 3066f4414602SMark Fasheh goto out_unlock; 3067f4414602SMark Fasheh 30680efa9f48SMark Fasheh if (same_inode) 30690efa9f48SMark Fasheh lock_extent_range(src, same_lock_start, same_lock_len); 30700efa9f48SMark Fasheh else 3071f4414602SMark Fasheh btrfs_double_extent_lock(src, loff, dst, dst_loff, len); 3072f4414602SMark Fasheh 3073207910ddSMark Fasheh /* pass original length for comparison so we stay within i_size */ 3074f4414602SMark Fasheh ret = btrfs_cmp_data(src, loff, dst, dst_loff, olen, &cmp); 3075416161dbSMark Fasheh if (ret == 0) 30761c919a5eSMark Fasheh ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1); 3077416161dbSMark Fasheh 30780efa9f48SMark Fasheh if (same_inode) 30790efa9f48SMark Fasheh unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start, 30800efa9f48SMark Fasheh same_lock_start + same_lock_len - 1); 30810efa9f48SMark Fasheh else 3082f4414602SMark Fasheh btrfs_double_extent_unlock(src, loff, dst, dst_loff, len); 3083f4414602SMark Fasheh 3084f4414602SMark Fasheh btrfs_cmp_data_free(&cmp); 3085416161dbSMark Fasheh out_unlock: 30860efa9f48SMark Fasheh if (same_inode) 30870efa9f48SMark Fasheh mutex_unlock(&src->i_mutex); 30880efa9f48SMark Fasheh else 3089f4414602SMark Fasheh btrfs_double_inode_unlock(src, dst); 3090416161dbSMark Fasheh 3091416161dbSMark Fasheh return ret; 3092416161dbSMark Fasheh } 3093416161dbSMark Fasheh 3094416161dbSMark Fasheh #define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024) 3095416161dbSMark Fasheh 3096416161dbSMark Fasheh static long btrfs_ioctl_file_extent_same(struct file *file, 30971c1c8747SAl Viro struct btrfs_ioctl_same_args __user *argp) 3098416161dbSMark Fasheh { 3099497b4050SFilipe Manana struct btrfs_ioctl_same_args *same = NULL; 3100cbf8b8caSMark Fasheh struct btrfs_ioctl_same_extent_info *info; 31011c1c8747SAl Viro struct inode *src = file_inode(file); 3102416161dbSMark Fasheh u64 off; 3103416161dbSMark Fasheh u64 len; 3104416161dbSMark Fasheh int i; 3105416161dbSMark Fasheh int ret; 3106cbf8b8caSMark Fasheh unsigned long size; 3107416161dbSMark Fasheh u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize; 3108416161dbSMark Fasheh bool is_admin = capable(CAP_SYS_ADMIN); 31091c1c8747SAl Viro u16 count; 3110416161dbSMark Fasheh 3111416161dbSMark Fasheh if (!(file->f_mode & FMODE_READ)) 3112416161dbSMark Fasheh return -EINVAL; 3113416161dbSMark Fasheh 3114416161dbSMark Fasheh ret = mnt_want_write_file(file); 3115416161dbSMark Fasheh if (ret) 3116416161dbSMark Fasheh return ret; 3117416161dbSMark Fasheh 31181c1c8747SAl Viro if (get_user(count, &argp->dest_count)) { 3119416161dbSMark Fasheh ret = -EFAULT; 3120416161dbSMark Fasheh goto out; 3121416161dbSMark Fasheh } 3122416161dbSMark Fasheh 31231c1c8747SAl Viro size = offsetof(struct btrfs_ioctl_same_args __user, info[count]); 3124cbf8b8caSMark Fasheh 31251c1c8747SAl Viro same = memdup_user(argp, size); 3126cbf8b8caSMark Fasheh 3127229eed43SGeyslan G. Bem if (IS_ERR(same)) { 3128229eed43SGeyslan G. Bem ret = PTR_ERR(same); 3129497b4050SFilipe Manana same = NULL; 3130cbf8b8caSMark Fasheh goto out; 3131cbf8b8caSMark Fasheh } 3132cbf8b8caSMark Fasheh 3133cbf8b8caSMark Fasheh off = same->logical_offset; 3134cbf8b8caSMark Fasheh len = same->length; 3135416161dbSMark Fasheh 3136416161dbSMark Fasheh /* 3137416161dbSMark Fasheh * Limit the total length we will dedupe for each operation. 3138416161dbSMark Fasheh * This is intended to bound the total time spent in this 3139416161dbSMark Fasheh * ioctl to something sane. 3140416161dbSMark Fasheh */ 3141416161dbSMark Fasheh if (len > BTRFS_MAX_DEDUPE_LEN) 3142416161dbSMark Fasheh len = BTRFS_MAX_DEDUPE_LEN; 3143416161dbSMark Fasheh 3144416161dbSMark Fasheh if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) { 3145416161dbSMark Fasheh /* 3146416161dbSMark Fasheh * Btrfs does not support blocksize < page_size. As a 3147416161dbSMark Fasheh * result, btrfs_cmp_data() won't correctly handle 3148416161dbSMark Fasheh * this situation without an update. 3149416161dbSMark Fasheh */ 3150416161dbSMark Fasheh ret = -EINVAL; 3151416161dbSMark Fasheh goto out; 3152416161dbSMark Fasheh } 3153416161dbSMark Fasheh 3154416161dbSMark Fasheh ret = -EISDIR; 3155416161dbSMark Fasheh if (S_ISDIR(src->i_mode)) 3156416161dbSMark Fasheh goto out; 3157416161dbSMark Fasheh 3158416161dbSMark Fasheh ret = -EACCES; 3159416161dbSMark Fasheh if (!S_ISREG(src->i_mode)) 3160416161dbSMark Fasheh goto out; 3161416161dbSMark Fasheh 3162cbf8b8caSMark Fasheh /* pre-format output fields to sane values */ 31631c1c8747SAl Viro for (i = 0; i < count; i++) { 3164cbf8b8caSMark Fasheh same->info[i].bytes_deduped = 0ULL; 3165cbf8b8caSMark Fasheh same->info[i].status = 0; 3166416161dbSMark Fasheh } 3167416161dbSMark Fasheh 31681c1c8747SAl Viro for (i = 0, info = same->info; i < count; i++, info++) { 31691c1c8747SAl Viro struct inode *dst; 31701c1c8747SAl Viro struct fd dst_file = fdget(info->fd); 31711c1c8747SAl Viro if (!dst_file.file) { 3172cbf8b8caSMark Fasheh info->status = -EBADF; 31731c1c8747SAl Viro continue; 3174416161dbSMark Fasheh } 31751c1c8747SAl Viro dst = file_inode(dst_file.file); 3176416161dbSMark Fasheh 31771c1c8747SAl Viro if (!(is_admin || (dst_file.file->f_mode & FMODE_WRITE))) { 3178cbf8b8caSMark Fasheh info->status = -EINVAL; 31791c1c8747SAl Viro } else if (file->f_path.mnt != dst_file.file->f_path.mnt) { 3180cbf8b8caSMark Fasheh info->status = -EXDEV; 31811c1c8747SAl Viro } else if (S_ISDIR(dst->i_mode)) { 3182cbf8b8caSMark Fasheh info->status = -EISDIR; 31831c1c8747SAl Viro } else if (!S_ISREG(dst->i_mode)) { 3184cbf8b8caSMark Fasheh info->status = -EACCES; 31851c1c8747SAl Viro } else { 3186cbf8b8caSMark Fasheh info->status = btrfs_extent_same(src, off, len, dst, 3187cbf8b8caSMark Fasheh info->logical_offset); 3188cbf8b8caSMark Fasheh if (info->status == 0) 3189cbf8b8caSMark Fasheh info->bytes_deduped += len; 31901c1c8747SAl Viro } 31911c1c8747SAl Viro fdput(dst_file); 3192cbf8b8caSMark Fasheh } 3193416161dbSMark Fasheh 3194cbf8b8caSMark Fasheh ret = copy_to_user(argp, same, size); 3195cbf8b8caSMark Fasheh if (ret) 3196416161dbSMark Fasheh ret = -EFAULT; 3197416161dbSMark Fasheh 3198416161dbSMark Fasheh out: 3199416161dbSMark Fasheh mnt_drop_write_file(file); 3200497b4050SFilipe Manana kfree(same); 3201416161dbSMark Fasheh return ret; 3202416161dbSMark Fasheh } 3203416161dbSMark Fasheh 3204fcebe456SJosef Bacik /* Helper to check and see if this root currently has a ref on the given disk 3205fcebe456SJosef Bacik * bytenr. If it does then we need to update the quota for this root. This 3206fcebe456SJosef Bacik * doesn't do anything if quotas aren't enabled. 3207fcebe456SJosef Bacik */ 3208fcebe456SJosef Bacik static int check_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, 3209fcebe456SJosef Bacik u64 disko) 3210fcebe456SJosef Bacik { 32113284da7bSDavid Sterba struct seq_list tree_mod_seq_elem = SEQ_LIST_INIT(tree_mod_seq_elem); 3212fcebe456SJosef Bacik struct ulist *roots; 3213fcebe456SJosef Bacik struct ulist_iterator uiter; 3214fcebe456SJosef Bacik struct ulist_node *root_node = NULL; 3215fcebe456SJosef Bacik int ret; 3216fcebe456SJosef Bacik 3217fcebe456SJosef Bacik if (!root->fs_info->quota_enabled) 3218fcebe456SJosef Bacik return 1; 3219fcebe456SJosef Bacik 3220fcebe456SJosef Bacik btrfs_get_tree_mod_seq(root->fs_info, &tree_mod_seq_elem); 3221fcebe456SJosef Bacik ret = btrfs_find_all_roots(trans, root->fs_info, disko, 3222fcebe456SJosef Bacik tree_mod_seq_elem.seq, &roots); 3223fcebe456SJosef Bacik if (ret < 0) 3224fcebe456SJosef Bacik goto out; 3225fcebe456SJosef Bacik ret = 0; 3226fcebe456SJosef Bacik ULIST_ITER_INIT(&uiter); 3227fcebe456SJosef Bacik while ((root_node = ulist_next(roots, &uiter))) { 3228fcebe456SJosef Bacik if (root_node->val == root->objectid) { 3229fcebe456SJosef Bacik ret = 1; 3230fcebe456SJosef Bacik break; 3231fcebe456SJosef Bacik } 3232fcebe456SJosef Bacik } 3233fcebe456SJosef Bacik ulist_free(roots); 3234fcebe456SJosef Bacik out: 3235fcebe456SJosef Bacik btrfs_put_tree_mod_seq(root->fs_info, &tree_mod_seq_elem); 3236fcebe456SJosef Bacik return ret; 3237fcebe456SJosef Bacik } 3238fcebe456SJosef Bacik 3239f82a9901SFilipe Manana static int clone_finish_inode_update(struct btrfs_trans_handle *trans, 3240f82a9901SFilipe Manana struct inode *inode, 3241f82a9901SFilipe Manana u64 endoff, 3242f82a9901SFilipe Manana const u64 destoff, 32431c919a5eSMark Fasheh const u64 olen, 32441c919a5eSMark Fasheh int no_time_update) 3245f82a9901SFilipe Manana { 3246f82a9901SFilipe Manana struct btrfs_root *root = BTRFS_I(inode)->root; 3247f82a9901SFilipe Manana int ret; 3248f82a9901SFilipe Manana 3249f82a9901SFilipe Manana inode_inc_iversion(inode); 32501c919a5eSMark Fasheh if (!no_time_update) 3251f82a9901SFilipe Manana inode->i_mtime = inode->i_ctime = CURRENT_TIME; 3252f82a9901SFilipe Manana /* 3253f82a9901SFilipe Manana * We round up to the block size at eof when determining which 3254f82a9901SFilipe Manana * extents to clone above, but shouldn't round up the file size. 3255f82a9901SFilipe Manana */ 3256f82a9901SFilipe Manana if (endoff > destoff + olen) 3257f82a9901SFilipe Manana endoff = destoff + olen; 3258f82a9901SFilipe Manana if (endoff > inode->i_size) 3259f82a9901SFilipe Manana btrfs_i_size_write(inode, endoff); 3260f82a9901SFilipe Manana 3261f82a9901SFilipe Manana ret = btrfs_update_inode(trans, root, inode); 3262f82a9901SFilipe Manana if (ret) { 3263f82a9901SFilipe Manana btrfs_abort_transaction(trans, root, ret); 3264f82a9901SFilipe Manana btrfs_end_transaction(trans, root); 3265f82a9901SFilipe Manana goto out; 3266f82a9901SFilipe Manana } 3267f82a9901SFilipe Manana ret = btrfs_end_transaction(trans, root); 3268f82a9901SFilipe Manana out: 3269f82a9901SFilipe Manana return ret; 3270f82a9901SFilipe Manana } 3271f82a9901SFilipe Manana 32727ffbb598SFilipe Manana static void clone_update_extent_map(struct inode *inode, 32737ffbb598SFilipe Manana const struct btrfs_trans_handle *trans, 32747ffbb598SFilipe Manana const struct btrfs_path *path, 32757ffbb598SFilipe Manana const u64 hole_offset, 32767ffbb598SFilipe Manana const u64 hole_len) 32777ffbb598SFilipe Manana { 32787ffbb598SFilipe Manana struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 32797ffbb598SFilipe Manana struct extent_map *em; 32807ffbb598SFilipe Manana int ret; 32817ffbb598SFilipe Manana 32827ffbb598SFilipe Manana em = alloc_extent_map(); 32837ffbb598SFilipe Manana if (!em) { 32847ffbb598SFilipe Manana set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, 32857ffbb598SFilipe Manana &BTRFS_I(inode)->runtime_flags); 32867ffbb598SFilipe Manana return; 32877ffbb598SFilipe Manana } 32887ffbb598SFilipe Manana 328914f59796SFilipe Manana if (path) { 329014f59796SFilipe Manana struct btrfs_file_extent_item *fi; 329114f59796SFilipe Manana 329214f59796SFilipe Manana fi = btrfs_item_ptr(path->nodes[0], path->slots[0], 329314f59796SFilipe Manana struct btrfs_file_extent_item); 32947ffbb598SFilipe Manana btrfs_extent_item_to_extent_map(inode, path, fi, false, em); 32957ffbb598SFilipe Manana em->generation = -1; 32967ffbb598SFilipe Manana if (btrfs_file_extent_type(path->nodes[0], fi) == 32977ffbb598SFilipe Manana BTRFS_FILE_EXTENT_INLINE) 32987ffbb598SFilipe Manana set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, 32997ffbb598SFilipe Manana &BTRFS_I(inode)->runtime_flags); 33007ffbb598SFilipe Manana } else { 33017ffbb598SFilipe Manana em->start = hole_offset; 33027ffbb598SFilipe Manana em->len = hole_len; 33037ffbb598SFilipe Manana em->ram_bytes = em->len; 33047ffbb598SFilipe Manana em->orig_start = hole_offset; 33057ffbb598SFilipe Manana em->block_start = EXTENT_MAP_HOLE; 33067ffbb598SFilipe Manana em->block_len = 0; 33077ffbb598SFilipe Manana em->orig_block_len = 0; 33087ffbb598SFilipe Manana em->compress_type = BTRFS_COMPRESS_NONE; 33097ffbb598SFilipe Manana em->generation = trans->transid; 33107ffbb598SFilipe Manana } 33117ffbb598SFilipe Manana 33127ffbb598SFilipe Manana while (1) { 33137ffbb598SFilipe Manana write_lock(&em_tree->lock); 33147ffbb598SFilipe Manana ret = add_extent_mapping(em_tree, em, 1); 33157ffbb598SFilipe Manana write_unlock(&em_tree->lock); 33167ffbb598SFilipe Manana if (ret != -EEXIST) { 33177ffbb598SFilipe Manana free_extent_map(em); 33187ffbb598SFilipe Manana break; 33197ffbb598SFilipe Manana } 33207ffbb598SFilipe Manana btrfs_drop_extent_cache(inode, em->start, 33217ffbb598SFilipe Manana em->start + em->len - 1, 0); 33227ffbb598SFilipe Manana } 33237ffbb598SFilipe Manana 3324ee39b432SDavid Sterba if (ret) 33257ffbb598SFilipe Manana set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, 33267ffbb598SFilipe Manana &BTRFS_I(inode)->runtime_flags); 33277ffbb598SFilipe Manana } 33287ffbb598SFilipe Manana 332932b7c687SMark Fasheh /** 333032b7c687SMark Fasheh * btrfs_clone() - clone a range from inode file to another 333132b7c687SMark Fasheh * 333232b7c687SMark Fasheh * @src: Inode to clone from 333332b7c687SMark Fasheh * @inode: Inode to clone to 333432b7c687SMark Fasheh * @off: Offset within source to start clone from 333532b7c687SMark Fasheh * @olen: Original length, passed by user, of range to clone 33361c919a5eSMark Fasheh * @olen_aligned: Block-aligned value of olen 333732b7c687SMark Fasheh * @destoff: Offset within @inode to start clone 33381c919a5eSMark Fasheh * @no_time_update: Whether to update mtime/ctime on the target inode 333932b7c687SMark Fasheh */ 334032b7c687SMark Fasheh static int btrfs_clone(struct inode *src, struct inode *inode, 3341f82a9901SFilipe Manana const u64 off, const u64 olen, const u64 olen_aligned, 33421c919a5eSMark Fasheh const u64 destoff, int no_time_update) 3343f46b5a66SChristoph Hellwig { 3344f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 334532b7c687SMark Fasheh struct btrfs_path *path = NULL; 3346f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 334732b7c687SMark Fasheh struct btrfs_trans_handle *trans; 334832b7c687SMark Fasheh char *buf = NULL; 3349ae01a0abSYan Zheng struct btrfs_key key; 3350f46b5a66SChristoph Hellwig u32 nritems; 3351f46b5a66SChristoph Hellwig int slot; 3352ae01a0abSYan Zheng int ret; 3353fcebe456SJosef Bacik int no_quota; 3354f82a9901SFilipe Manana const u64 len = olen_aligned; 3355fcebe456SJosef Bacik u64 last_disko = 0; 3356f82a9901SFilipe Manana u64 last_dest_end = destoff; 3357ae01a0abSYan Zheng 3358ae01a0abSYan Zheng ret = -ENOMEM; 3359707e8a07SDavid Sterba buf = vmalloc(root->nodesize); 3360ae01a0abSYan Zheng if (!buf) 336132b7c687SMark Fasheh return ret; 3362ae01a0abSYan Zheng 3363ae01a0abSYan Zheng path = btrfs_alloc_path(); 3364ae01a0abSYan Zheng if (!path) { 3365ae01a0abSYan Zheng vfree(buf); 336632b7c687SMark Fasheh return ret; 3367ae01a0abSYan Zheng } 336832b7c687SMark Fasheh 3369ae01a0abSYan Zheng path->reada = 2; 3370c5c9cd4dSSage Weil /* clone data */ 337133345d01SLi Zefan key.objectid = btrfs_ino(src); 3372ae01a0abSYan Zheng key.type = BTRFS_EXTENT_DATA_KEY; 33732c463823SFilipe Manana key.offset = off; 3374f46b5a66SChristoph Hellwig 3375f46b5a66SChristoph Hellwig while (1) { 3376de249e66SChris Mason u64 next_key_min_offset = key.offset + 1; 3377df858e76SFilipe Manana 3378f46b5a66SChristoph Hellwig /* 3379f46b5a66SChristoph Hellwig * note the key will change type as we walk through the 3380f46b5a66SChristoph Hellwig * tree. 3381f46b5a66SChristoph Hellwig */ 3382e4355f34SFilipe David Borba Manana path->leave_spinning = 1; 3383362a20c5SDavid Sterba ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path, 3384362a20c5SDavid Sterba 0, 0); 3385f46b5a66SChristoph Hellwig if (ret < 0) 3386f46b5a66SChristoph Hellwig goto out; 33872c463823SFilipe Manana /* 33882c463823SFilipe Manana * First search, if no extent item that starts at offset off was 33892c463823SFilipe Manana * found but the previous item is an extent item, it's possible 33902c463823SFilipe Manana * it might overlap our target range, therefore process it. 33912c463823SFilipe Manana */ 33922c463823SFilipe Manana if (key.offset == off && ret > 0 && path->slots[0] > 0) { 33932c463823SFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 33942c463823SFilipe Manana path->slots[0] - 1); 33952c463823SFilipe Manana if (key.type == BTRFS_EXTENT_DATA_KEY) 33962c463823SFilipe Manana path->slots[0]--; 33972c463823SFilipe Manana } 3398f46b5a66SChristoph Hellwig 3399ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3400e4355f34SFilipe David Borba Manana process_slot: 3401fcebe456SJosef Bacik no_quota = 1; 3402ae01a0abSYan Zheng if (path->slots[0] >= nritems) { 3403362a20c5SDavid Sterba ret = btrfs_next_leaf(BTRFS_I(src)->root, path); 3404f46b5a66SChristoph Hellwig if (ret < 0) 3405f46b5a66SChristoph Hellwig goto out; 3406f46b5a66SChristoph Hellwig if (ret > 0) 3407f46b5a66SChristoph Hellwig break; 3408ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3409f46b5a66SChristoph Hellwig } 3410f46b5a66SChristoph Hellwig leaf = path->nodes[0]; 3411f46b5a66SChristoph Hellwig slot = path->slots[0]; 3412f46b5a66SChristoph Hellwig 3413ae01a0abSYan Zheng btrfs_item_key_to_cpu(leaf, &key, slot); 3414962a298fSDavid Sterba if (key.type > BTRFS_EXTENT_DATA_KEY || 341533345d01SLi Zefan key.objectid != btrfs_ino(src)) 3416f46b5a66SChristoph Hellwig break; 3417f46b5a66SChristoph Hellwig 3418962a298fSDavid Sterba if (key.type == BTRFS_EXTENT_DATA_KEY) { 3419c5c9cd4dSSage Weil struct btrfs_file_extent_item *extent; 3420c5c9cd4dSSage Weil int type; 342131840ae1SZheng Yan u32 size; 342231840ae1SZheng Yan struct btrfs_key new_key; 3423c5c9cd4dSSage Weil u64 disko = 0, diskl = 0; 3424c5c9cd4dSSage Weil u64 datao = 0, datal = 0; 3425c5c9cd4dSSage Weil u8 comp; 3426f82a9901SFilipe Manana u64 drop_start; 342731840ae1SZheng Yan 3428c5c9cd4dSSage Weil extent = btrfs_item_ptr(leaf, slot, 3429c5c9cd4dSSage Weil struct btrfs_file_extent_item); 3430c5c9cd4dSSage Weil comp = btrfs_file_extent_compression(leaf, extent); 3431c5c9cd4dSSage Weil type = btrfs_file_extent_type(leaf, extent); 3432c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3433c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3434d397712bSChris Mason disko = btrfs_file_extent_disk_bytenr(leaf, 3435d397712bSChris Mason extent); 3436d397712bSChris Mason diskl = btrfs_file_extent_disk_num_bytes(leaf, 3437d397712bSChris Mason extent); 3438c5c9cd4dSSage Weil datao = btrfs_file_extent_offset(leaf, extent); 3439d397712bSChris Mason datal = btrfs_file_extent_num_bytes(leaf, 3440d397712bSChris Mason extent); 3441c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3442c5c9cd4dSSage Weil /* take upper bound, may be compressed */ 3443c5c9cd4dSSage Weil datal = btrfs_file_extent_ram_bytes(leaf, 3444c5c9cd4dSSage Weil extent); 3445c5c9cd4dSSage Weil } 344631840ae1SZheng Yan 34472c463823SFilipe Manana /* 34482c463823SFilipe Manana * The first search might have left us at an extent 34492c463823SFilipe Manana * item that ends before our target range's start, can 34502c463823SFilipe Manana * happen if we have holes and NO_HOLES feature enabled. 34512c463823SFilipe Manana */ 34522c463823SFilipe Manana if (key.offset + datal <= off) { 3453e4355f34SFilipe David Borba Manana path->slots[0]++; 3454e4355f34SFilipe David Borba Manana goto process_slot; 34552c463823SFilipe Manana } else if (key.offset >= off + len) { 34562c463823SFilipe Manana break; 3457e4355f34SFilipe David Borba Manana } 3458df858e76SFilipe Manana next_key_min_offset = key.offset + datal; 3459e4355f34SFilipe David Borba Manana size = btrfs_item_size_nr(leaf, slot); 3460e4355f34SFilipe David Borba Manana read_extent_buffer(leaf, buf, 3461e4355f34SFilipe David Borba Manana btrfs_item_ptr_offset(leaf, slot), 3462e4355f34SFilipe David Borba Manana size); 3463e4355f34SFilipe David Borba Manana 3464e4355f34SFilipe David Borba Manana btrfs_release_path(path); 3465e4355f34SFilipe David Borba Manana path->leave_spinning = 0; 3466c5c9cd4dSSage Weil 346731840ae1SZheng Yan memcpy(&new_key, &key, sizeof(new_key)); 346833345d01SLi Zefan new_key.objectid = btrfs_ino(inode); 34694d728ec7SLi Zefan if (off <= key.offset) 3470c5c9cd4dSSage Weil new_key.offset = key.offset + destoff - off; 34714d728ec7SLi Zefan else 34724d728ec7SLi Zefan new_key.offset = destoff; 3473c5c9cd4dSSage Weil 3474b6f3409bSSage Weil /* 3475f82a9901SFilipe Manana * Deal with a hole that doesn't have an extent item 3476f82a9901SFilipe Manana * that represents it (NO_HOLES feature enabled). 3477f82a9901SFilipe Manana * This hole is either in the middle of the cloning 3478f82a9901SFilipe Manana * range or at the beginning (fully overlaps it or 3479f82a9901SFilipe Manana * partially overlaps it). 3480f82a9901SFilipe Manana */ 3481f82a9901SFilipe Manana if (new_key.offset != last_dest_end) 3482f82a9901SFilipe Manana drop_start = last_dest_end; 3483f82a9901SFilipe Manana else 3484f82a9901SFilipe Manana drop_start = new_key.offset; 3485f82a9901SFilipe Manana 3486f82a9901SFilipe Manana /* 3487b6f3409bSSage Weil * 1 - adjusting old extent (we may have to split it) 3488b6f3409bSSage Weil * 1 - add new extent 3489b6f3409bSSage Weil * 1 - inode update 3490b6f3409bSSage Weil */ 3491b6f3409bSSage Weil trans = btrfs_start_transaction(root, 3); 3492a22285a6SYan, Zheng if (IS_ERR(trans)) { 3493a22285a6SYan, Zheng ret = PTR_ERR(trans); 3494a22285a6SYan, Zheng goto out; 3495a22285a6SYan, Zheng } 3496a22285a6SYan, Zheng 3497c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3498c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3499d72c0842SLi Zefan /* 3500d72c0842SLi Zefan * a | --- range to clone ---| b 3501d72c0842SLi Zefan * | ------------- extent ------------- | 3502d72c0842SLi Zefan */ 3503d72c0842SLi Zefan 350493915584SAntonio Ospite /* subtract range b */ 3505d72c0842SLi Zefan if (key.offset + datal > off + len) 3506d72c0842SLi Zefan datal = off + len - key.offset; 3507d72c0842SLi Zefan 350893915584SAntonio Ospite /* subtract range a */ 3509a22285a6SYan, Zheng if (off > key.offset) { 3510a22285a6SYan, Zheng datao += off - key.offset; 3511a22285a6SYan, Zheng datal -= off - key.offset; 3512a22285a6SYan, Zheng } 3513a22285a6SYan, Zheng 35145dc562c5SJosef Bacik ret = btrfs_drop_extents(trans, root, inode, 3515f82a9901SFilipe Manana drop_start, 3516a22285a6SYan, Zheng new_key.offset + datal, 35172671485dSJosef Bacik 1); 351879787eaaSJeff Mahoney if (ret) { 35193f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 352000fdf13aSLiu Bo btrfs_abort_transaction(trans, 352100fdf13aSLiu Bo root, ret); 352279787eaaSJeff Mahoney btrfs_end_transaction(trans, root); 352379787eaaSJeff Mahoney goto out; 352479787eaaSJeff Mahoney } 3525a22285a6SYan, Zheng 352631840ae1SZheng Yan ret = btrfs_insert_empty_item(trans, root, path, 352731840ae1SZheng Yan &new_key, size); 352879787eaaSJeff Mahoney if (ret) { 352979787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, 353079787eaaSJeff Mahoney ret); 353179787eaaSJeff Mahoney btrfs_end_transaction(trans, root); 353279787eaaSJeff Mahoney goto out; 353379787eaaSJeff Mahoney } 353431840ae1SZheng Yan 353531840ae1SZheng Yan leaf = path->nodes[0]; 353631840ae1SZheng Yan slot = path->slots[0]; 353731840ae1SZheng Yan write_extent_buffer(leaf, buf, 353831840ae1SZheng Yan btrfs_item_ptr_offset(leaf, slot), 353931840ae1SZheng Yan size); 3540ae01a0abSYan Zheng 3541f46b5a66SChristoph Hellwig extent = btrfs_item_ptr(leaf, slot, 3542f46b5a66SChristoph Hellwig struct btrfs_file_extent_item); 3543c5c9cd4dSSage Weil 3544c5c9cd4dSSage Weil /* disko == 0 means it's a hole */ 3545c5c9cd4dSSage Weil if (!disko) 3546c5c9cd4dSSage Weil datao = 0; 3547c5c9cd4dSSage Weil 3548c5c9cd4dSSage Weil btrfs_set_file_extent_offset(leaf, extent, 3549c5c9cd4dSSage Weil datao); 3550c5c9cd4dSSage Weil btrfs_set_file_extent_num_bytes(leaf, extent, 3551c5c9cd4dSSage Weil datal); 3552fcebe456SJosef Bacik 3553fcebe456SJosef Bacik /* 3554fcebe456SJosef Bacik * We need to look up the roots that point at 3555fcebe456SJosef Bacik * this bytenr and see if the new root does. If 3556fcebe456SJosef Bacik * it does not we need to make sure we update 3557fcebe456SJosef Bacik * quotas appropriately. 3558fcebe456SJosef Bacik */ 3559fcebe456SJosef Bacik if (disko && root != BTRFS_I(src)->root && 3560fcebe456SJosef Bacik disko != last_disko) { 3561fcebe456SJosef Bacik no_quota = check_ref(trans, root, 3562fcebe456SJosef Bacik disko); 3563fcebe456SJosef Bacik if (no_quota < 0) { 3564fcebe456SJosef Bacik btrfs_abort_transaction(trans, 3565fcebe456SJosef Bacik root, 3566fcebe456SJosef Bacik ret); 3567fcebe456SJosef Bacik btrfs_end_transaction(trans, 3568fcebe456SJosef Bacik root); 3569fcebe456SJosef Bacik ret = no_quota; 3570fcebe456SJosef Bacik goto out; 3571fcebe456SJosef Bacik } 3572fcebe456SJosef Bacik } 3573fcebe456SJosef Bacik 3574c5c9cd4dSSage Weil if (disko) { 3575c5c9cd4dSSage Weil inode_add_bytes(inode, datal); 3576ae01a0abSYan Zheng ret = btrfs_inc_extent_ref(trans, root, 35775d4f98a2SYan Zheng disko, diskl, 0, 3578f46b5a66SChristoph Hellwig root->root_key.objectid, 357933345d01SLi Zefan btrfs_ino(inode), 358066d7e7f0SArne Jansen new_key.offset - datao, 3581fcebe456SJosef Bacik no_quota); 358279787eaaSJeff Mahoney if (ret) { 358379787eaaSJeff Mahoney btrfs_abort_transaction(trans, 358479787eaaSJeff Mahoney root, 358579787eaaSJeff Mahoney ret); 358679787eaaSJeff Mahoney btrfs_end_transaction(trans, 358779787eaaSJeff Mahoney root); 358879787eaaSJeff Mahoney goto out; 358979787eaaSJeff Mahoney 359079787eaaSJeff Mahoney } 3591f46b5a66SChristoph Hellwig } 3592c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3593c5c9cd4dSSage Weil u64 skip = 0; 3594c5c9cd4dSSage Weil u64 trim = 0; 3595d3ecfcdfSLiu Bo u64 aligned_end = 0; 3596d3ecfcdfSLiu Bo 3597ed958762SFilipe Manana /* 3598ed958762SFilipe Manana * Don't copy an inline extent into an offset 3599ed958762SFilipe Manana * greater than zero. Having an inline extent 3600ed958762SFilipe Manana * at such an offset results in chaos as btrfs 3601ed958762SFilipe Manana * isn't prepared for such cases. Just skip 3602ed958762SFilipe Manana * this case for the same reasons as commented 3603ed958762SFilipe Manana * at btrfs_ioctl_clone(). 3604ed958762SFilipe Manana */ 3605ed958762SFilipe Manana if (last_dest_end > 0) { 3606ed958762SFilipe Manana ret = -EOPNOTSUPP; 3607ed958762SFilipe Manana btrfs_end_transaction(trans, root); 3608ed958762SFilipe Manana goto out; 3609ed958762SFilipe Manana } 3610ed958762SFilipe Manana 3611c5c9cd4dSSage Weil if (off > key.offset) { 3612c5c9cd4dSSage Weil skip = off - key.offset; 3613c5c9cd4dSSage Weil new_key.offset += skip; 361431840ae1SZheng Yan } 3615d397712bSChris Mason 3616c5c9cd4dSSage Weil if (key.offset + datal > off + len) 3617c5c9cd4dSSage Weil trim = key.offset + datal - (off + len); 3618d397712bSChris Mason 3619c5c9cd4dSSage Weil if (comp && (skip || trim)) { 3620c5c9cd4dSSage Weil ret = -EINVAL; 3621a22285a6SYan, Zheng btrfs_end_transaction(trans, root); 3622c5c9cd4dSSage Weil goto out; 362331840ae1SZheng Yan } 3624c5c9cd4dSSage Weil size -= skip + trim; 3625c5c9cd4dSSage Weil datal -= skip + trim; 3626a22285a6SYan, Zheng 3627d3ecfcdfSLiu Bo aligned_end = ALIGN(new_key.offset + datal, 3628d3ecfcdfSLiu Bo root->sectorsize); 36295dc562c5SJosef Bacik ret = btrfs_drop_extents(trans, root, inode, 3630f82a9901SFilipe Manana drop_start, 3631d3ecfcdfSLiu Bo aligned_end, 36322671485dSJosef Bacik 1); 363379787eaaSJeff Mahoney if (ret) { 36343f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 36353a29bc09SChris Mason btrfs_abort_transaction(trans, 36363a29bc09SChris Mason root, ret); 363779787eaaSJeff Mahoney btrfs_end_transaction(trans, root); 363879787eaaSJeff Mahoney goto out; 363979787eaaSJeff Mahoney } 3640a22285a6SYan, Zheng 3641c5c9cd4dSSage Weil ret = btrfs_insert_empty_item(trans, root, path, 3642c5c9cd4dSSage Weil &new_key, size); 364379787eaaSJeff Mahoney if (ret) { 364479787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, 364579787eaaSJeff Mahoney ret); 364679787eaaSJeff Mahoney btrfs_end_transaction(trans, root); 364779787eaaSJeff Mahoney goto out; 364879787eaaSJeff Mahoney } 3649c5c9cd4dSSage Weil 3650c5c9cd4dSSage Weil if (skip) { 3651d397712bSChris Mason u32 start = 3652d397712bSChris Mason btrfs_file_extent_calc_inline_size(0); 3653c5c9cd4dSSage Weil memmove(buf+start, buf+start+skip, 3654c5c9cd4dSSage Weil datal); 3655c5c9cd4dSSage Weil } 3656c5c9cd4dSSage Weil 3657c5c9cd4dSSage Weil leaf = path->nodes[0]; 3658c5c9cd4dSSage Weil slot = path->slots[0]; 3659c5c9cd4dSSage Weil write_extent_buffer(leaf, buf, 3660c5c9cd4dSSage Weil btrfs_item_ptr_offset(leaf, slot), 3661c5c9cd4dSSage Weil size); 3662c5c9cd4dSSage Weil inode_add_bytes(inode, datal); 3663c5c9cd4dSSage Weil } 3664c5c9cd4dSSage Weil 36657ffbb598SFilipe Manana /* If we have an implicit hole (NO_HOLES feature). */ 36667ffbb598SFilipe Manana if (drop_start < new_key.offset) 36677ffbb598SFilipe Manana clone_update_extent_map(inode, trans, 366814f59796SFilipe Manana NULL, drop_start, 36697ffbb598SFilipe Manana new_key.offset - drop_start); 36707ffbb598SFilipe Manana 367114f59796SFilipe Manana clone_update_extent_map(inode, trans, path, 0, 0); 36727ffbb598SFilipe Manana 3673c5c9cd4dSSage Weil btrfs_mark_buffer_dirty(leaf); 3674b3b4aa74SDavid Sterba btrfs_release_path(path); 3675c5c9cd4dSSage Weil 367662e2390eSFilipe Manana last_dest_end = ALIGN(new_key.offset + datal, 367762e2390eSFilipe Manana root->sectorsize); 3678f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, 3679f82a9901SFilipe Manana last_dest_end, 36801c919a5eSMark Fasheh destoff, olen, 36811c919a5eSMark Fasheh no_time_update); 3682f82a9901SFilipe Manana if (ret) 368379787eaaSJeff Mahoney goto out; 36842c463823SFilipe Manana if (new_key.offset + datal >= destoff + len) 36852c463823SFilipe Manana break; 3686a22285a6SYan, Zheng } 3687b3b4aa74SDavid Sterba btrfs_release_path(path); 3688df858e76SFilipe Manana key.offset = next_key_min_offset; 3689ae01a0abSYan Zheng } 3690f46b5a66SChristoph Hellwig ret = 0; 369132b7c687SMark Fasheh 3692f82a9901SFilipe Manana if (last_dest_end < destoff + len) { 3693f82a9901SFilipe Manana /* 3694f82a9901SFilipe Manana * We have an implicit hole (NO_HOLES feature is enabled) that 3695f82a9901SFilipe Manana * fully or partially overlaps our cloning range at its end. 3696f82a9901SFilipe Manana */ 3697b3b4aa74SDavid Sterba btrfs_release_path(path); 3698f82a9901SFilipe Manana 3699f82a9901SFilipe Manana /* 3700f82a9901SFilipe Manana * 1 - remove extent(s) 3701f82a9901SFilipe Manana * 1 - inode update 3702f82a9901SFilipe Manana */ 3703f82a9901SFilipe Manana trans = btrfs_start_transaction(root, 2); 3704f82a9901SFilipe Manana if (IS_ERR(trans)) { 3705f82a9901SFilipe Manana ret = PTR_ERR(trans); 3706f82a9901SFilipe Manana goto out; 3707f82a9901SFilipe Manana } 3708f82a9901SFilipe Manana ret = btrfs_drop_extents(trans, root, inode, 3709f82a9901SFilipe Manana last_dest_end, destoff + len, 1); 3710f82a9901SFilipe Manana if (ret) { 3711f82a9901SFilipe Manana if (ret != -EOPNOTSUPP) 3712f82a9901SFilipe Manana btrfs_abort_transaction(trans, root, ret); 3713f82a9901SFilipe Manana btrfs_end_transaction(trans, root); 3714f82a9901SFilipe Manana goto out; 3715f82a9901SFilipe Manana } 371614f59796SFilipe Manana clone_update_extent_map(inode, trans, NULL, last_dest_end, 371714f59796SFilipe Manana destoff + len - last_dest_end); 3718f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, destoff + len, 37191c919a5eSMark Fasheh destoff, olen, no_time_update); 3720f82a9901SFilipe Manana } 3721f82a9901SFilipe Manana 3722f46b5a66SChristoph Hellwig out: 372332b7c687SMark Fasheh btrfs_free_path(path); 372432b7c687SMark Fasheh vfree(buf); 372532b7c687SMark Fasheh return ret; 372632b7c687SMark Fasheh } 372732b7c687SMark Fasheh 372832b7c687SMark Fasheh static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, 372932b7c687SMark Fasheh u64 off, u64 olen, u64 destoff) 373032b7c687SMark Fasheh { 373154563d41SAl Viro struct inode *inode = file_inode(file); 373232b7c687SMark Fasheh struct btrfs_root *root = BTRFS_I(inode)->root; 373332b7c687SMark Fasheh struct fd src_file; 373432b7c687SMark Fasheh struct inode *src; 373532b7c687SMark Fasheh int ret; 373632b7c687SMark Fasheh u64 len = olen; 373732b7c687SMark Fasheh u64 bs = root->fs_info->sb->s_blocksize; 373832b7c687SMark Fasheh int same_inode = 0; 373932b7c687SMark Fasheh 374032b7c687SMark Fasheh /* 374132b7c687SMark Fasheh * TODO: 374232b7c687SMark Fasheh * - split compressed inline extents. annoying: we need to 374332b7c687SMark Fasheh * decompress into destination's address_space (the file offset 374432b7c687SMark Fasheh * may change, so source mapping won't do), then recompress (or 374532b7c687SMark Fasheh * otherwise reinsert) a subrange. 374600fdf13aSLiu Bo * 374700fdf13aSLiu Bo * - split destination inode's inline extents. The inline extents can 374800fdf13aSLiu Bo * be either compressed or non-compressed. 374932b7c687SMark Fasheh */ 375032b7c687SMark Fasheh 375132b7c687SMark Fasheh /* the destination must be opened for writing */ 375232b7c687SMark Fasheh if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND)) 375332b7c687SMark Fasheh return -EINVAL; 375432b7c687SMark Fasheh 375532b7c687SMark Fasheh if (btrfs_root_readonly(root)) 375632b7c687SMark Fasheh return -EROFS; 375732b7c687SMark Fasheh 375832b7c687SMark Fasheh ret = mnt_want_write_file(file); 375932b7c687SMark Fasheh if (ret) 376032b7c687SMark Fasheh return ret; 376132b7c687SMark Fasheh 376232b7c687SMark Fasheh src_file = fdget(srcfd); 376332b7c687SMark Fasheh if (!src_file.file) { 376432b7c687SMark Fasheh ret = -EBADF; 376532b7c687SMark Fasheh goto out_drop_write; 376632b7c687SMark Fasheh } 376732b7c687SMark Fasheh 376832b7c687SMark Fasheh ret = -EXDEV; 376932b7c687SMark Fasheh if (src_file.file->f_path.mnt != file->f_path.mnt) 377032b7c687SMark Fasheh goto out_fput; 377132b7c687SMark Fasheh 377232b7c687SMark Fasheh src = file_inode(src_file.file); 377332b7c687SMark Fasheh 377432b7c687SMark Fasheh ret = -EINVAL; 377532b7c687SMark Fasheh if (src == inode) 377632b7c687SMark Fasheh same_inode = 1; 377732b7c687SMark Fasheh 377832b7c687SMark Fasheh /* the src must be open for reading */ 377932b7c687SMark Fasheh if (!(src_file.file->f_mode & FMODE_READ)) 378032b7c687SMark Fasheh goto out_fput; 378132b7c687SMark Fasheh 378232b7c687SMark Fasheh /* don't make the dst file partly checksummed */ 378332b7c687SMark Fasheh if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != 378432b7c687SMark Fasheh (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) 378532b7c687SMark Fasheh goto out_fput; 378632b7c687SMark Fasheh 378732b7c687SMark Fasheh ret = -EISDIR; 378832b7c687SMark Fasheh if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode)) 378932b7c687SMark Fasheh goto out_fput; 379032b7c687SMark Fasheh 379132b7c687SMark Fasheh ret = -EXDEV; 379232b7c687SMark Fasheh if (src->i_sb != inode->i_sb) 379332b7c687SMark Fasheh goto out_fput; 379432b7c687SMark Fasheh 379532b7c687SMark Fasheh if (!same_inode) { 379632b7c687SMark Fasheh if (inode < src) { 379732b7c687SMark Fasheh mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT); 379832b7c687SMark Fasheh mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD); 379932b7c687SMark Fasheh } else { 380032b7c687SMark Fasheh mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT); 380132b7c687SMark Fasheh mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD); 380232b7c687SMark Fasheh } 380332b7c687SMark Fasheh } else { 380432b7c687SMark Fasheh mutex_lock(&src->i_mutex); 380532b7c687SMark Fasheh } 380632b7c687SMark Fasheh 380732b7c687SMark Fasheh /* determine range to clone */ 380832b7c687SMark Fasheh ret = -EINVAL; 380932b7c687SMark Fasheh if (off + len > src->i_size || off + len < off) 381032b7c687SMark Fasheh goto out_unlock; 381132b7c687SMark Fasheh if (len == 0) 381232b7c687SMark Fasheh olen = len = src->i_size - off; 381332b7c687SMark Fasheh /* if we extend to eof, continue to block boundary */ 381432b7c687SMark Fasheh if (off + len == src->i_size) 381532b7c687SMark Fasheh len = ALIGN(src->i_size, bs) - off; 381632b7c687SMark Fasheh 3817ccccf3d6SFilipe Manana if (len == 0) { 3818ccccf3d6SFilipe Manana ret = 0; 3819ccccf3d6SFilipe Manana goto out_unlock; 3820ccccf3d6SFilipe Manana } 3821ccccf3d6SFilipe Manana 382232b7c687SMark Fasheh /* verify the end result is block aligned */ 382332b7c687SMark Fasheh if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) || 382432b7c687SMark Fasheh !IS_ALIGNED(destoff, bs)) 382532b7c687SMark Fasheh goto out_unlock; 382632b7c687SMark Fasheh 382732b7c687SMark Fasheh /* verify if ranges are overlapped within the same file */ 382832b7c687SMark Fasheh if (same_inode) { 382932b7c687SMark Fasheh if (destoff + len > off && destoff < off + len) 383032b7c687SMark Fasheh goto out_unlock; 383132b7c687SMark Fasheh } 383232b7c687SMark Fasheh 383332b7c687SMark Fasheh if (destoff > inode->i_size) { 383432b7c687SMark Fasheh ret = btrfs_cont_expand(inode, inode->i_size, destoff); 383532b7c687SMark Fasheh if (ret) 383632b7c687SMark Fasheh goto out_unlock; 383732b7c687SMark Fasheh } 383832b7c687SMark Fasheh 3839c125b8bfSFilipe Manana /* 3840c125b8bfSFilipe Manana * Lock the target range too. Right after we replace the file extent 3841c125b8bfSFilipe Manana * items in the fs tree (which now point to the cloned data), we might 3842c125b8bfSFilipe Manana * have a worker replace them with extent items relative to a write 3843c125b8bfSFilipe Manana * operation that was issued before this clone operation (i.e. confront 3844c125b8bfSFilipe Manana * with inode.c:btrfs_finish_ordered_io). 3845c125b8bfSFilipe Manana */ 3846c125b8bfSFilipe Manana if (same_inode) { 3847c125b8bfSFilipe Manana u64 lock_start = min_t(u64, off, destoff); 3848c125b8bfSFilipe Manana u64 lock_len = max_t(u64, off, destoff) + len - lock_start; 384932b7c687SMark Fasheh 3850c125b8bfSFilipe Manana lock_extent_range(src, lock_start, lock_len); 3851c125b8bfSFilipe Manana } else { 385232b7c687SMark Fasheh lock_extent_range(src, off, len); 3853c125b8bfSFilipe Manana lock_extent_range(inode, destoff, len); 3854c125b8bfSFilipe Manana } 385532b7c687SMark Fasheh 38561c919a5eSMark Fasheh ret = btrfs_clone(src, inode, off, olen, len, destoff, 0); 385732b7c687SMark Fasheh 3858c125b8bfSFilipe Manana if (same_inode) { 3859c125b8bfSFilipe Manana u64 lock_start = min_t(u64, off, destoff); 3860c125b8bfSFilipe Manana u64 lock_end = max_t(u64, off, destoff) + len - 1; 3861c125b8bfSFilipe Manana 3862c125b8bfSFilipe Manana unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end); 3863c125b8bfSFilipe Manana } else { 3864aa42ffd9SLiu Bo unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1); 3865c125b8bfSFilipe Manana unlock_extent(&BTRFS_I(inode)->io_tree, destoff, 3866c125b8bfSFilipe Manana destoff + len - 1); 3867c125b8bfSFilipe Manana } 3868c125b8bfSFilipe Manana /* 3869c125b8bfSFilipe Manana * Truncate page cache pages so that future reads will see the cloned 3870c125b8bfSFilipe Manana * data immediately and not the previous data. 3871c125b8bfSFilipe Manana */ 3872c125b8bfSFilipe Manana truncate_inode_pages_range(&inode->i_data, destoff, 3873c125b8bfSFilipe Manana PAGE_CACHE_ALIGN(destoff + len) - 1); 3874f46b5a66SChristoph Hellwig out_unlock: 3875c57c2b3eSFilipe David Borba Manana if (!same_inode) { 3876c57c2b3eSFilipe David Borba Manana if (inode < src) { 3877f46b5a66SChristoph Hellwig mutex_unlock(&src->i_mutex); 3878f46b5a66SChristoph Hellwig mutex_unlock(&inode->i_mutex); 3879c57c2b3eSFilipe David Borba Manana } else { 3880c57c2b3eSFilipe David Borba Manana mutex_unlock(&inode->i_mutex); 3881c57c2b3eSFilipe David Borba Manana mutex_unlock(&src->i_mutex); 3882c57c2b3eSFilipe David Borba Manana } 3883c57c2b3eSFilipe David Borba Manana } else { 3884c57c2b3eSFilipe David Borba Manana mutex_unlock(&src->i_mutex); 3885c57c2b3eSFilipe David Borba Manana } 3886f46b5a66SChristoph Hellwig out_fput: 38872903ff01SAl Viro fdput(src_file); 3888ab67b7c1SYan Zheng out_drop_write: 38892a79f17eSAl Viro mnt_drop_write_file(file); 3890f46b5a66SChristoph Hellwig return ret; 3891f46b5a66SChristoph Hellwig } 3892f46b5a66SChristoph Hellwig 38937a865e8aSChristoph Hellwig static long btrfs_ioctl_clone_range(struct file *file, void __user *argp) 3894c5c9cd4dSSage Weil { 3895c5c9cd4dSSage Weil struct btrfs_ioctl_clone_range_args args; 3896c5c9cd4dSSage Weil 38977a865e8aSChristoph Hellwig if (copy_from_user(&args, argp, sizeof(args))) 3898c5c9cd4dSSage Weil return -EFAULT; 3899c5c9cd4dSSage Weil return btrfs_ioctl_clone(file, args.src_fd, args.src_offset, 3900c5c9cd4dSSage Weil args.src_length, args.dest_offset); 3901c5c9cd4dSSage Weil } 3902c5c9cd4dSSage Weil 3903f46b5a66SChristoph Hellwig /* 3904f46b5a66SChristoph Hellwig * there are many ways the trans_start and trans_end ioctls can lead 3905f46b5a66SChristoph Hellwig * to deadlocks. They should only be used by applications that 3906f46b5a66SChristoph Hellwig * basically own the machine, and have a very in depth understanding 3907f46b5a66SChristoph Hellwig * of all the possible deadlocks and enospc problems. 3908f46b5a66SChristoph Hellwig */ 3909b2950863SChristoph Hellwig static long btrfs_ioctl_trans_start(struct file *file) 3910f46b5a66SChristoph Hellwig { 3911496ad9aaSAl Viro struct inode *inode = file_inode(file); 3912f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 3913f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 39141ab86aedSSage Weil int ret; 3915f46b5a66SChristoph Hellwig 39161ab86aedSSage Weil ret = -EPERM; 3917df5b5520SChristoph Hellwig if (!capable(CAP_SYS_ADMIN)) 3918f46b5a66SChristoph Hellwig goto out; 39191ab86aedSSage Weil 39201ab86aedSSage Weil ret = -EINPROGRESS; 39211ab86aedSSage Weil if (file->private_data) 39221ab86aedSSage Weil goto out; 39239ca9ee09SSage Weil 3924b83cc969SLi Zefan ret = -EROFS; 3925b83cc969SLi Zefan if (btrfs_root_readonly(root)) 3926b83cc969SLi Zefan goto out; 3927b83cc969SLi Zefan 3928a561be71SAl Viro ret = mnt_want_write_file(file); 3929c146afadSYan Zheng if (ret) 3930c146afadSYan Zheng goto out; 3931c146afadSYan Zheng 3932a4abeea4SJosef Bacik atomic_inc(&root->fs_info->open_ioctl_trans); 39339ca9ee09SSage Weil 3934f46b5a66SChristoph Hellwig ret = -ENOMEM; 39357a7eaa40SJosef Bacik trans = btrfs_start_ioctl_transaction(root); 3936abd30bb0STsutomu Itoh if (IS_ERR(trans)) 39371ab86aedSSage Weil goto out_drop; 39381ab86aedSSage Weil 39391ab86aedSSage Weil file->private_data = trans; 39401ab86aedSSage Weil return 0; 39411ab86aedSSage Weil 39421ab86aedSSage Weil out_drop: 3943a4abeea4SJosef Bacik atomic_dec(&root->fs_info->open_ioctl_trans); 39442a79f17eSAl Viro mnt_drop_write_file(file); 3945f46b5a66SChristoph Hellwig out: 3946f46b5a66SChristoph Hellwig return ret; 3947f46b5a66SChristoph Hellwig } 3948f46b5a66SChristoph Hellwig 39496ef5ed0dSJosef Bacik static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) 39506ef5ed0dSJosef Bacik { 3951496ad9aaSAl Viro struct inode *inode = file_inode(file); 39526ef5ed0dSJosef Bacik struct btrfs_root *root = BTRFS_I(inode)->root; 39536ef5ed0dSJosef Bacik struct btrfs_root *new_root; 39546ef5ed0dSJosef Bacik struct btrfs_dir_item *di; 39556ef5ed0dSJosef Bacik struct btrfs_trans_handle *trans; 39566ef5ed0dSJosef Bacik struct btrfs_path *path; 39576ef5ed0dSJosef Bacik struct btrfs_key location; 39586ef5ed0dSJosef Bacik struct btrfs_disk_key disk_key; 39596ef5ed0dSJosef Bacik u64 objectid = 0; 39606ef5ed0dSJosef Bacik u64 dir_id; 39613c04ce01SMiao Xie int ret; 39626ef5ed0dSJosef Bacik 39636ef5ed0dSJosef Bacik if (!capable(CAP_SYS_ADMIN)) 39646ef5ed0dSJosef Bacik return -EPERM; 39656ef5ed0dSJosef Bacik 39663c04ce01SMiao Xie ret = mnt_want_write_file(file); 39673c04ce01SMiao Xie if (ret) 39683c04ce01SMiao Xie return ret; 39693c04ce01SMiao Xie 39703c04ce01SMiao Xie if (copy_from_user(&objectid, argp, sizeof(objectid))) { 39713c04ce01SMiao Xie ret = -EFAULT; 39723c04ce01SMiao Xie goto out; 39733c04ce01SMiao Xie } 39746ef5ed0dSJosef Bacik 39756ef5ed0dSJosef Bacik if (!objectid) 39761cecf579Schandan objectid = BTRFS_FS_TREE_OBJECTID; 39776ef5ed0dSJosef Bacik 39786ef5ed0dSJosef Bacik location.objectid = objectid; 39796ef5ed0dSJosef Bacik location.type = BTRFS_ROOT_ITEM_KEY; 39806ef5ed0dSJosef Bacik location.offset = (u64)-1; 39816ef5ed0dSJosef Bacik 39826ef5ed0dSJosef Bacik new_root = btrfs_read_fs_root_no_name(root->fs_info, &location); 39833c04ce01SMiao Xie if (IS_ERR(new_root)) { 39843c04ce01SMiao Xie ret = PTR_ERR(new_root); 39853c04ce01SMiao Xie goto out; 39863c04ce01SMiao Xie } 39876ef5ed0dSJosef Bacik 39886ef5ed0dSJosef Bacik path = btrfs_alloc_path(); 39893c04ce01SMiao Xie if (!path) { 39903c04ce01SMiao Xie ret = -ENOMEM; 39913c04ce01SMiao Xie goto out; 39923c04ce01SMiao Xie } 39936ef5ed0dSJosef Bacik path->leave_spinning = 1; 39946ef5ed0dSJosef Bacik 39956ef5ed0dSJosef Bacik trans = btrfs_start_transaction(root, 1); 399698d5dc13STsutomu Itoh if (IS_ERR(trans)) { 39976ef5ed0dSJosef Bacik btrfs_free_path(path); 39983c04ce01SMiao Xie ret = PTR_ERR(trans); 39993c04ce01SMiao Xie goto out; 40006ef5ed0dSJosef Bacik } 40016ef5ed0dSJosef Bacik 40026c41761fSDavid Sterba dir_id = btrfs_super_root_dir(root->fs_info->super_copy); 40036ef5ed0dSJosef Bacik di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path, 40046ef5ed0dSJosef Bacik dir_id, "default", 7, 1); 4005cf1e99a4SDan Carpenter if (IS_ERR_OR_NULL(di)) { 40066ef5ed0dSJosef Bacik btrfs_free_path(path); 40076ef5ed0dSJosef Bacik btrfs_end_transaction(trans, root); 4008efe120a0SFrank Holton btrfs_err(new_root->fs_info, "Umm, you don't have the default dir" 4009efe120a0SFrank Holton "item, this isn't going to work"); 40103c04ce01SMiao Xie ret = -ENOENT; 40113c04ce01SMiao Xie goto out; 40126ef5ed0dSJosef Bacik } 40136ef5ed0dSJosef Bacik 40146ef5ed0dSJosef Bacik btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); 40156ef5ed0dSJosef Bacik btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); 40166ef5ed0dSJosef Bacik btrfs_mark_buffer_dirty(path->nodes[0]); 40176ef5ed0dSJosef Bacik btrfs_free_path(path); 40186ef5ed0dSJosef Bacik 40192b0ce2c2SMitch Harder btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL); 40206ef5ed0dSJosef Bacik btrfs_end_transaction(trans, root); 40213c04ce01SMiao Xie out: 40223c04ce01SMiao Xie mnt_drop_write_file(file); 40233c04ce01SMiao Xie return ret; 40246ef5ed0dSJosef Bacik } 40256ef5ed0dSJosef Bacik 40265af3e8ccSStefan Behrens void btrfs_get_block_group_info(struct list_head *groups_list, 4027bf5fc093SJosef Bacik struct btrfs_ioctl_space_info *space) 4028bf5fc093SJosef Bacik { 4029bf5fc093SJosef Bacik struct btrfs_block_group_cache *block_group; 4030bf5fc093SJosef Bacik 4031bf5fc093SJosef Bacik space->total_bytes = 0; 4032bf5fc093SJosef Bacik space->used_bytes = 0; 4033bf5fc093SJosef Bacik space->flags = 0; 4034bf5fc093SJosef Bacik list_for_each_entry(block_group, groups_list, list) { 4035bf5fc093SJosef Bacik space->flags = block_group->flags; 4036bf5fc093SJosef Bacik space->total_bytes += block_group->key.offset; 4037bf5fc093SJosef Bacik space->used_bytes += 4038bf5fc093SJosef Bacik btrfs_block_group_used(&block_group->item); 4039bf5fc093SJosef Bacik } 4040bf5fc093SJosef Bacik } 4041bf5fc093SJosef Bacik 404248a3b636SEric Sandeen static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg) 40431406e432SJosef Bacik { 40441406e432SJosef Bacik struct btrfs_ioctl_space_args space_args; 40451406e432SJosef Bacik struct btrfs_ioctl_space_info space; 40461406e432SJosef Bacik struct btrfs_ioctl_space_info *dest; 40477fde62bfSChris Mason struct btrfs_ioctl_space_info *dest_orig; 404813f2696fSDaniel J Blueman struct btrfs_ioctl_space_info __user *user_dest; 40491406e432SJosef Bacik struct btrfs_space_info *info; 4050bf5fc093SJosef Bacik u64 types[] = {BTRFS_BLOCK_GROUP_DATA, 4051bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_SYSTEM, 4052bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_METADATA, 4053bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA}; 4054bf5fc093SJosef Bacik int num_types = 4; 40557fde62bfSChris Mason int alloc_size; 40561406e432SJosef Bacik int ret = 0; 405751788b1bSDan Rosenberg u64 slot_count = 0; 4058bf5fc093SJosef Bacik int i, c; 40591406e432SJosef Bacik 40601406e432SJosef Bacik if (copy_from_user(&space_args, 40611406e432SJosef Bacik (struct btrfs_ioctl_space_args __user *)arg, 40621406e432SJosef Bacik sizeof(space_args))) 40631406e432SJosef Bacik return -EFAULT; 40641406e432SJosef Bacik 4065bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4066bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4067bf5fc093SJosef Bacik 4068bf5fc093SJosef Bacik info = NULL; 40697fde62bfSChris Mason rcu_read_lock(); 4070bf5fc093SJosef Bacik list_for_each_entry_rcu(tmp, &root->fs_info->space_info, 4071bf5fc093SJosef Bacik list) { 4072bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4073bf5fc093SJosef Bacik info = tmp; 4074bf5fc093SJosef Bacik break; 4075bf5fc093SJosef Bacik } 4076bf5fc093SJosef Bacik } 40777fde62bfSChris Mason rcu_read_unlock(); 40781406e432SJosef Bacik 4079bf5fc093SJosef Bacik if (!info) 4080bf5fc093SJosef Bacik continue; 4081bf5fc093SJosef Bacik 4082bf5fc093SJosef Bacik down_read(&info->groups_sem); 4083bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4084bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) 4085bf5fc093SJosef Bacik slot_count++; 4086bf5fc093SJosef Bacik } 4087bf5fc093SJosef Bacik up_read(&info->groups_sem); 4088bf5fc093SJosef Bacik } 4089bf5fc093SJosef Bacik 409036523e95SDavid Sterba /* 409136523e95SDavid Sterba * Global block reserve, exported as a space_info 409236523e95SDavid Sterba */ 409336523e95SDavid Sterba slot_count++; 409436523e95SDavid Sterba 40957fde62bfSChris Mason /* space_slots == 0 means they are asking for a count */ 40967fde62bfSChris Mason if (space_args.space_slots == 0) { 40977fde62bfSChris Mason space_args.total_spaces = slot_count; 40987fde62bfSChris Mason goto out; 40997fde62bfSChris Mason } 4100bf5fc093SJosef Bacik 410151788b1bSDan Rosenberg slot_count = min_t(u64, space_args.space_slots, slot_count); 4102bf5fc093SJosef Bacik 41037fde62bfSChris Mason alloc_size = sizeof(*dest) * slot_count; 4104bf5fc093SJosef Bacik 41057fde62bfSChris Mason /* we generally have at most 6 or so space infos, one for each raid 41067fde62bfSChris Mason * level. So, a whole page should be more than enough for everyone 41077fde62bfSChris Mason */ 41087fde62bfSChris Mason if (alloc_size > PAGE_CACHE_SIZE) 41097fde62bfSChris Mason return -ENOMEM; 41107fde62bfSChris Mason 41117fde62bfSChris Mason space_args.total_spaces = 0; 41127fde62bfSChris Mason dest = kmalloc(alloc_size, GFP_NOFS); 41137fde62bfSChris Mason if (!dest) 41147fde62bfSChris Mason return -ENOMEM; 41157fde62bfSChris Mason dest_orig = dest; 41167fde62bfSChris Mason 41177fde62bfSChris Mason /* now we have a buffer to copy into */ 4118bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4119bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4120bf5fc093SJosef Bacik 412151788b1bSDan Rosenberg if (!slot_count) 412251788b1bSDan Rosenberg break; 412351788b1bSDan Rosenberg 4124bf5fc093SJosef Bacik info = NULL; 41251406e432SJosef Bacik rcu_read_lock(); 4126bf5fc093SJosef Bacik list_for_each_entry_rcu(tmp, &root->fs_info->space_info, 4127bf5fc093SJosef Bacik list) { 4128bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4129bf5fc093SJosef Bacik info = tmp; 41307fde62bfSChris Mason break; 4131bf5fc093SJosef Bacik } 4132bf5fc093SJosef Bacik } 4133bf5fc093SJosef Bacik rcu_read_unlock(); 41347fde62bfSChris Mason 4135bf5fc093SJosef Bacik if (!info) 4136bf5fc093SJosef Bacik continue; 4137bf5fc093SJosef Bacik down_read(&info->groups_sem); 4138bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4139bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) { 41405af3e8ccSStefan Behrens btrfs_get_block_group_info( 41415af3e8ccSStefan Behrens &info->block_groups[c], &space); 41427fde62bfSChris Mason memcpy(dest, &space, sizeof(space)); 41431406e432SJosef Bacik dest++; 41441406e432SJosef Bacik space_args.total_spaces++; 414551788b1bSDan Rosenberg slot_count--; 41461406e432SJosef Bacik } 414751788b1bSDan Rosenberg if (!slot_count) 414851788b1bSDan Rosenberg break; 4149bf5fc093SJosef Bacik } 4150bf5fc093SJosef Bacik up_read(&info->groups_sem); 4151bf5fc093SJosef Bacik } 41521406e432SJosef Bacik 415336523e95SDavid Sterba /* 415436523e95SDavid Sterba * Add global block reserve 415536523e95SDavid Sterba */ 415636523e95SDavid Sterba if (slot_count) { 415736523e95SDavid Sterba struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv; 415836523e95SDavid Sterba 415936523e95SDavid Sterba spin_lock(&block_rsv->lock); 416036523e95SDavid Sterba space.total_bytes = block_rsv->size; 416136523e95SDavid Sterba space.used_bytes = block_rsv->size - block_rsv->reserved; 416236523e95SDavid Sterba spin_unlock(&block_rsv->lock); 416336523e95SDavid Sterba space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV; 416436523e95SDavid Sterba memcpy(dest, &space, sizeof(space)); 416536523e95SDavid Sterba space_args.total_spaces++; 416636523e95SDavid Sterba } 416736523e95SDavid Sterba 41682eec6c81SDaniel J Blueman user_dest = (struct btrfs_ioctl_space_info __user *) 41697fde62bfSChris Mason (arg + sizeof(struct btrfs_ioctl_space_args)); 41707fde62bfSChris Mason 41717fde62bfSChris Mason if (copy_to_user(user_dest, dest_orig, alloc_size)) 41727fde62bfSChris Mason ret = -EFAULT; 41737fde62bfSChris Mason 41747fde62bfSChris Mason kfree(dest_orig); 41757fde62bfSChris Mason out: 41767fde62bfSChris Mason if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) 41771406e432SJosef Bacik ret = -EFAULT; 41781406e432SJosef Bacik 41791406e432SJosef Bacik return ret; 41801406e432SJosef Bacik } 41811406e432SJosef Bacik 4182f46b5a66SChristoph Hellwig /* 4183f46b5a66SChristoph Hellwig * there are many ways the trans_start and trans_end ioctls can lead 4184f46b5a66SChristoph Hellwig * to deadlocks. They should only be used by applications that 4185f46b5a66SChristoph Hellwig * basically own the machine, and have a very in depth understanding 4186f46b5a66SChristoph Hellwig * of all the possible deadlocks and enospc problems. 4187f46b5a66SChristoph Hellwig */ 4188f46b5a66SChristoph Hellwig long btrfs_ioctl_trans_end(struct file *file) 4189f46b5a66SChristoph Hellwig { 4190496ad9aaSAl Viro struct inode *inode = file_inode(file); 4191f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 4192f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 4193f46b5a66SChristoph Hellwig 4194f46b5a66SChristoph Hellwig trans = file->private_data; 41951ab86aedSSage Weil if (!trans) 41961ab86aedSSage Weil return -EINVAL; 4197b214107eSChristoph Hellwig file->private_data = NULL; 41989ca9ee09SSage Weil 41991ab86aedSSage Weil btrfs_end_transaction(trans, root); 42001ab86aedSSage Weil 4201a4abeea4SJosef Bacik atomic_dec(&root->fs_info->open_ioctl_trans); 42029ca9ee09SSage Weil 42032a79f17eSAl Viro mnt_drop_write_file(file); 42041ab86aedSSage Weil return 0; 4205f46b5a66SChristoph Hellwig } 4206f46b5a66SChristoph Hellwig 42079a8c28beSMiao Xie static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, 42089a8c28beSMiao Xie void __user *argp) 420946204592SSage Weil { 421046204592SSage Weil struct btrfs_trans_handle *trans; 421146204592SSage Weil u64 transid; 4212db5b493aSTsutomu Itoh int ret; 421346204592SSage Weil 4214d4edf39bSMiao Xie trans = btrfs_attach_transaction_barrier(root); 4215ff7c1d33SMiao Xie if (IS_ERR(trans)) { 4216ff7c1d33SMiao Xie if (PTR_ERR(trans) != -ENOENT) 421798d5dc13STsutomu Itoh return PTR_ERR(trans); 4218ff7c1d33SMiao Xie 4219ff7c1d33SMiao Xie /* No running transaction, don't bother */ 4220ff7c1d33SMiao Xie transid = root->fs_info->last_trans_committed; 4221ff7c1d33SMiao Xie goto out; 4222ff7c1d33SMiao Xie } 422346204592SSage Weil transid = trans->transid; 4224db5b493aSTsutomu Itoh ret = btrfs_commit_transaction_async(trans, root, 0); 42258b2b2d3cSTsutomu Itoh if (ret) { 42268b2b2d3cSTsutomu Itoh btrfs_end_transaction(trans, root); 4227db5b493aSTsutomu Itoh return ret; 42288b2b2d3cSTsutomu Itoh } 4229ff7c1d33SMiao Xie out: 423046204592SSage Weil if (argp) 423146204592SSage Weil if (copy_to_user(argp, &transid, sizeof(transid))) 423246204592SSage Weil return -EFAULT; 423346204592SSage Weil return 0; 423446204592SSage Weil } 423546204592SSage Weil 42369a8c28beSMiao Xie static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root, 42379a8c28beSMiao Xie void __user *argp) 423846204592SSage Weil { 423946204592SSage Weil u64 transid; 424046204592SSage Weil 424146204592SSage Weil if (argp) { 424246204592SSage Weil if (copy_from_user(&transid, argp, sizeof(transid))) 424346204592SSage Weil return -EFAULT; 424446204592SSage Weil } else { 424546204592SSage Weil transid = 0; /* current trans */ 424646204592SSage Weil } 424746204592SSage Weil return btrfs_wait_for_commit(root, transid); 424846204592SSage Weil } 424946204592SSage Weil 4250b8e95489SMiao Xie static long btrfs_ioctl_scrub(struct file *file, void __user *arg) 4251475f6387SJan Schmidt { 4252496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 4253475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4254b8e95489SMiao Xie int ret; 4255475f6387SJan Schmidt 4256475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4257475f6387SJan Schmidt return -EPERM; 4258475f6387SJan Schmidt 4259475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4260475f6387SJan Schmidt if (IS_ERR(sa)) 4261475f6387SJan Schmidt return PTR_ERR(sa); 4262475f6387SJan Schmidt 4263b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) { 4264b8e95489SMiao Xie ret = mnt_want_write_file(file); 4265b8e95489SMiao Xie if (ret) 4266b8e95489SMiao Xie goto out; 4267b8e95489SMiao Xie } 4268b8e95489SMiao Xie 4269aa1b8cd4SStefan Behrens ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end, 427063a212abSStefan Behrens &sa->progress, sa->flags & BTRFS_SCRUB_READONLY, 427163a212abSStefan Behrens 0); 4272475f6387SJan Schmidt 4273475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 4274475f6387SJan Schmidt ret = -EFAULT; 4275475f6387SJan Schmidt 4276b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) 4277b8e95489SMiao Xie mnt_drop_write_file(file); 4278b8e95489SMiao Xie out: 4279475f6387SJan Schmidt kfree(sa); 4280475f6387SJan Schmidt return ret; 4281475f6387SJan Schmidt } 4282475f6387SJan Schmidt 4283475f6387SJan Schmidt static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg) 4284475f6387SJan Schmidt { 4285475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4286475f6387SJan Schmidt return -EPERM; 4287475f6387SJan Schmidt 4288aa1b8cd4SStefan Behrens return btrfs_scrub_cancel(root->fs_info); 4289475f6387SJan Schmidt } 4290475f6387SJan Schmidt 4291475f6387SJan Schmidt static long btrfs_ioctl_scrub_progress(struct btrfs_root *root, 4292475f6387SJan Schmidt void __user *arg) 4293475f6387SJan Schmidt { 4294475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4295475f6387SJan Schmidt int ret; 4296475f6387SJan Schmidt 4297475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4298475f6387SJan Schmidt return -EPERM; 4299475f6387SJan Schmidt 4300475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4301475f6387SJan Schmidt if (IS_ERR(sa)) 4302475f6387SJan Schmidt return PTR_ERR(sa); 4303475f6387SJan Schmidt 4304475f6387SJan Schmidt ret = btrfs_scrub_progress(root, sa->devid, &sa->progress); 4305475f6387SJan Schmidt 4306475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 4307475f6387SJan Schmidt ret = -EFAULT; 4308475f6387SJan Schmidt 4309475f6387SJan Schmidt kfree(sa); 4310475f6387SJan Schmidt return ret; 4311475f6387SJan Schmidt } 4312475f6387SJan Schmidt 4313c11d2c23SStefan Behrens static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root, 4314b27f7c0cSDavid Sterba void __user *arg) 4315c11d2c23SStefan Behrens { 4316c11d2c23SStefan Behrens struct btrfs_ioctl_get_dev_stats *sa; 4317c11d2c23SStefan Behrens int ret; 4318c11d2c23SStefan Behrens 4319c11d2c23SStefan Behrens sa = memdup_user(arg, sizeof(*sa)); 4320c11d2c23SStefan Behrens if (IS_ERR(sa)) 4321c11d2c23SStefan Behrens return PTR_ERR(sa); 4322c11d2c23SStefan Behrens 4323b27f7c0cSDavid Sterba if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) { 4324b27f7c0cSDavid Sterba kfree(sa); 4325b27f7c0cSDavid Sterba return -EPERM; 4326b27f7c0cSDavid Sterba } 4327b27f7c0cSDavid Sterba 4328b27f7c0cSDavid Sterba ret = btrfs_get_dev_stats(root, sa); 4329c11d2c23SStefan Behrens 4330c11d2c23SStefan Behrens if (copy_to_user(arg, sa, sizeof(*sa))) 4331c11d2c23SStefan Behrens ret = -EFAULT; 4332c11d2c23SStefan Behrens 4333c11d2c23SStefan Behrens kfree(sa); 4334c11d2c23SStefan Behrens return ret; 4335c11d2c23SStefan Behrens } 4336c11d2c23SStefan Behrens 43373f6bcfbdSStefan Behrens static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg) 43383f6bcfbdSStefan Behrens { 43393f6bcfbdSStefan Behrens struct btrfs_ioctl_dev_replace_args *p; 43403f6bcfbdSStefan Behrens int ret; 43413f6bcfbdSStefan Behrens 43423f6bcfbdSStefan Behrens if (!capable(CAP_SYS_ADMIN)) 43433f6bcfbdSStefan Behrens return -EPERM; 43443f6bcfbdSStefan Behrens 43453f6bcfbdSStefan Behrens p = memdup_user(arg, sizeof(*p)); 43463f6bcfbdSStefan Behrens if (IS_ERR(p)) 43473f6bcfbdSStefan Behrens return PTR_ERR(p); 43483f6bcfbdSStefan Behrens 43493f6bcfbdSStefan Behrens switch (p->cmd) { 43503f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_START: 4351adfa97cbSIlya Dryomov if (root->fs_info->sb->s_flags & MS_RDONLY) { 4352adfa97cbSIlya Dryomov ret = -EROFS; 4353adfa97cbSIlya Dryomov goto out; 4354adfa97cbSIlya Dryomov } 43553f6bcfbdSStefan Behrens if (atomic_xchg( 43563f6bcfbdSStefan Behrens &root->fs_info->mutually_exclusive_operation_running, 43573f6bcfbdSStefan Behrens 1)) { 4358e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 43593f6bcfbdSStefan Behrens } else { 43603f6bcfbdSStefan Behrens ret = btrfs_dev_replace_start(root, p); 43613f6bcfbdSStefan Behrens atomic_set( 43623f6bcfbdSStefan Behrens &root->fs_info->mutually_exclusive_operation_running, 43633f6bcfbdSStefan Behrens 0); 43643f6bcfbdSStefan Behrens } 43653f6bcfbdSStefan Behrens break; 43663f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS: 43673f6bcfbdSStefan Behrens btrfs_dev_replace_status(root->fs_info, p); 43683f6bcfbdSStefan Behrens ret = 0; 43693f6bcfbdSStefan Behrens break; 43703f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL: 43713f6bcfbdSStefan Behrens ret = btrfs_dev_replace_cancel(root->fs_info, p); 43723f6bcfbdSStefan Behrens break; 43733f6bcfbdSStefan Behrens default: 43743f6bcfbdSStefan Behrens ret = -EINVAL; 43753f6bcfbdSStefan Behrens break; 43763f6bcfbdSStefan Behrens } 43773f6bcfbdSStefan Behrens 43783f6bcfbdSStefan Behrens if (copy_to_user(arg, p, sizeof(*p))) 43793f6bcfbdSStefan Behrens ret = -EFAULT; 4380adfa97cbSIlya Dryomov out: 43813f6bcfbdSStefan Behrens kfree(p); 43823f6bcfbdSStefan Behrens return ret; 43833f6bcfbdSStefan Behrens } 43843f6bcfbdSStefan Behrens 4385d7728c96SJan Schmidt static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) 4386d7728c96SJan Schmidt { 4387d7728c96SJan Schmidt int ret = 0; 4388d7728c96SJan Schmidt int i; 4389740c3d22SChris Mason u64 rel_ptr; 4390d7728c96SJan Schmidt int size; 4391806468f8SChris Mason struct btrfs_ioctl_ino_path_args *ipa = NULL; 4392d7728c96SJan Schmidt struct inode_fs_paths *ipath = NULL; 4393d7728c96SJan Schmidt struct btrfs_path *path; 4394d7728c96SJan Schmidt 439582b22ac8SKusanagi Kouichi if (!capable(CAP_DAC_READ_SEARCH)) 4396d7728c96SJan Schmidt return -EPERM; 4397d7728c96SJan Schmidt 4398d7728c96SJan Schmidt path = btrfs_alloc_path(); 4399d7728c96SJan Schmidt if (!path) { 4400d7728c96SJan Schmidt ret = -ENOMEM; 4401d7728c96SJan Schmidt goto out; 4402d7728c96SJan Schmidt } 4403d7728c96SJan Schmidt 4404d7728c96SJan Schmidt ipa = memdup_user(arg, sizeof(*ipa)); 4405d7728c96SJan Schmidt if (IS_ERR(ipa)) { 4406d7728c96SJan Schmidt ret = PTR_ERR(ipa); 4407d7728c96SJan Schmidt ipa = NULL; 4408d7728c96SJan Schmidt goto out; 4409d7728c96SJan Schmidt } 4410d7728c96SJan Schmidt 4411d7728c96SJan Schmidt size = min_t(u32, ipa->size, 4096); 4412d7728c96SJan Schmidt ipath = init_ipath(size, root, path); 4413d7728c96SJan Schmidt if (IS_ERR(ipath)) { 4414d7728c96SJan Schmidt ret = PTR_ERR(ipath); 4415d7728c96SJan Schmidt ipath = NULL; 4416d7728c96SJan Schmidt goto out; 4417d7728c96SJan Schmidt } 4418d7728c96SJan Schmidt 4419d7728c96SJan Schmidt ret = paths_from_inode(ipa->inum, ipath); 4420d7728c96SJan Schmidt if (ret < 0) 4421d7728c96SJan Schmidt goto out; 4422d7728c96SJan Schmidt 4423d7728c96SJan Schmidt for (i = 0; i < ipath->fspath->elem_cnt; ++i) { 4424745c4d8eSJeff Mahoney rel_ptr = ipath->fspath->val[i] - 4425745c4d8eSJeff Mahoney (u64)(unsigned long)ipath->fspath->val; 4426740c3d22SChris Mason ipath->fspath->val[i] = rel_ptr; 4427d7728c96SJan Schmidt } 4428d7728c96SJan Schmidt 4429745c4d8eSJeff Mahoney ret = copy_to_user((void *)(unsigned long)ipa->fspath, 4430745c4d8eSJeff Mahoney (void *)(unsigned long)ipath->fspath, size); 4431d7728c96SJan Schmidt if (ret) { 4432d7728c96SJan Schmidt ret = -EFAULT; 4433d7728c96SJan Schmidt goto out; 4434d7728c96SJan Schmidt } 4435d7728c96SJan Schmidt 4436d7728c96SJan Schmidt out: 4437d7728c96SJan Schmidt btrfs_free_path(path); 4438d7728c96SJan Schmidt free_ipath(ipath); 4439d7728c96SJan Schmidt kfree(ipa); 4440d7728c96SJan Schmidt 4441d7728c96SJan Schmidt return ret; 4442d7728c96SJan Schmidt } 4443d7728c96SJan Schmidt 4444d7728c96SJan Schmidt static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx) 4445d7728c96SJan Schmidt { 4446d7728c96SJan Schmidt struct btrfs_data_container *inodes = ctx; 4447d7728c96SJan Schmidt const size_t c = 3 * sizeof(u64); 4448d7728c96SJan Schmidt 4449d7728c96SJan Schmidt if (inodes->bytes_left >= c) { 4450d7728c96SJan Schmidt inodes->bytes_left -= c; 4451d7728c96SJan Schmidt inodes->val[inodes->elem_cnt] = inum; 4452d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 1] = offset; 4453d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 2] = root; 4454d7728c96SJan Schmidt inodes->elem_cnt += 3; 4455d7728c96SJan Schmidt } else { 4456d7728c96SJan Schmidt inodes->bytes_missing += c - inodes->bytes_left; 4457d7728c96SJan Schmidt inodes->bytes_left = 0; 4458d7728c96SJan Schmidt inodes->elem_missed += 3; 4459d7728c96SJan Schmidt } 4460d7728c96SJan Schmidt 4461d7728c96SJan Schmidt return 0; 4462d7728c96SJan Schmidt } 4463d7728c96SJan Schmidt 4464d7728c96SJan Schmidt static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root, 4465d7728c96SJan Schmidt void __user *arg) 4466d7728c96SJan Schmidt { 4467d7728c96SJan Schmidt int ret = 0; 4468d7728c96SJan Schmidt int size; 4469d7728c96SJan Schmidt struct btrfs_ioctl_logical_ino_args *loi; 4470d7728c96SJan Schmidt struct btrfs_data_container *inodes = NULL; 4471d7728c96SJan Schmidt struct btrfs_path *path = NULL; 4472d7728c96SJan Schmidt 4473d7728c96SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4474d7728c96SJan Schmidt return -EPERM; 4475d7728c96SJan Schmidt 4476d7728c96SJan Schmidt loi = memdup_user(arg, sizeof(*loi)); 4477d7728c96SJan Schmidt if (IS_ERR(loi)) { 4478d7728c96SJan Schmidt ret = PTR_ERR(loi); 4479d7728c96SJan Schmidt loi = NULL; 4480d7728c96SJan Schmidt goto out; 4481d7728c96SJan Schmidt } 4482d7728c96SJan Schmidt 4483d7728c96SJan Schmidt path = btrfs_alloc_path(); 4484d7728c96SJan Schmidt if (!path) { 4485d7728c96SJan Schmidt ret = -ENOMEM; 4486d7728c96SJan Schmidt goto out; 4487d7728c96SJan Schmidt } 4488d7728c96SJan Schmidt 4489425d17a2SLiu Bo size = min_t(u32, loi->size, 64 * 1024); 4490d7728c96SJan Schmidt inodes = init_data_container(size); 4491d7728c96SJan Schmidt if (IS_ERR(inodes)) { 4492d7728c96SJan Schmidt ret = PTR_ERR(inodes); 4493d7728c96SJan Schmidt inodes = NULL; 4494d7728c96SJan Schmidt goto out; 4495d7728c96SJan Schmidt } 4496d7728c96SJan Schmidt 4497df031f07SLiu Bo ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path, 4498df031f07SLiu Bo build_ino_list, inodes); 4499df031f07SLiu Bo if (ret == -EINVAL) 4500d7728c96SJan Schmidt ret = -ENOENT; 4501d7728c96SJan Schmidt if (ret < 0) 4502d7728c96SJan Schmidt goto out; 4503d7728c96SJan Schmidt 4504745c4d8eSJeff Mahoney ret = copy_to_user((void *)(unsigned long)loi->inodes, 4505745c4d8eSJeff Mahoney (void *)(unsigned long)inodes, size); 4506d7728c96SJan Schmidt if (ret) 4507d7728c96SJan Schmidt ret = -EFAULT; 4508d7728c96SJan Schmidt 4509d7728c96SJan Schmidt out: 4510d7728c96SJan Schmidt btrfs_free_path(path); 4511425d17a2SLiu Bo vfree(inodes); 4512d7728c96SJan Schmidt kfree(loi); 4513d7728c96SJan Schmidt 4514d7728c96SJan Schmidt return ret; 4515d7728c96SJan Schmidt } 4516d7728c96SJan Schmidt 451719a39dceSIlya Dryomov void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock, 4518c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs) 4519c9e9f97bSIlya Dryomov { 4520c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl = fs_info->balance_ctl; 4521c9e9f97bSIlya Dryomov 4522c9e9f97bSIlya Dryomov bargs->flags = bctl->flags; 4523c9e9f97bSIlya Dryomov 4524837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_running)) 4525837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_RUNNING; 4526837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_pause_req)) 4527837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; 4528a7e99c69SIlya Dryomov if (atomic_read(&fs_info->balance_cancel_req)) 4529a7e99c69SIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; 4530837d5b6eSIlya Dryomov 4531c9e9f97bSIlya Dryomov memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); 4532c9e9f97bSIlya Dryomov memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); 4533c9e9f97bSIlya Dryomov memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); 453419a39dceSIlya Dryomov 453519a39dceSIlya Dryomov if (lock) { 453619a39dceSIlya Dryomov spin_lock(&fs_info->balance_lock); 453719a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 453819a39dceSIlya Dryomov spin_unlock(&fs_info->balance_lock); 453919a39dceSIlya Dryomov } else { 454019a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 454119a39dceSIlya Dryomov } 4542c9e9f97bSIlya Dryomov } 4543c9e9f97bSIlya Dryomov 45449ba1f6e4SLiu Bo static long btrfs_ioctl_balance(struct file *file, void __user *arg) 4545c9e9f97bSIlya Dryomov { 4546496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 4547c9e9f97bSIlya Dryomov struct btrfs_fs_info *fs_info = root->fs_info; 4548c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 4549c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl; 4550ed0fb78fSIlya Dryomov bool need_unlock; /* for mut. excl. ops lock */ 4551c9e9f97bSIlya Dryomov int ret; 4552c9e9f97bSIlya Dryomov 4553c9e9f97bSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4554c9e9f97bSIlya Dryomov return -EPERM; 4555c9e9f97bSIlya Dryomov 4556e54bfa31SLiu Bo ret = mnt_want_write_file(file); 45579ba1f6e4SLiu Bo if (ret) 45589ba1f6e4SLiu Bo return ret; 45599ba1f6e4SLiu Bo 4560ed0fb78fSIlya Dryomov again: 4561ed0fb78fSIlya Dryomov if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) { 4562c9e9f97bSIlya Dryomov mutex_lock(&fs_info->volume_mutex); 4563c9e9f97bSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4564ed0fb78fSIlya Dryomov need_unlock = true; 4565ed0fb78fSIlya Dryomov goto locked; 4566ed0fb78fSIlya Dryomov } 4567ed0fb78fSIlya Dryomov 4568ed0fb78fSIlya Dryomov /* 4569ed0fb78fSIlya Dryomov * mut. excl. ops lock is locked. Three possibilites: 4570ed0fb78fSIlya Dryomov * (1) some other op is running 4571ed0fb78fSIlya Dryomov * (2) balance is running 4572ed0fb78fSIlya Dryomov * (3) balance is paused -- special case (think resume) 4573ed0fb78fSIlya Dryomov */ 4574ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4575ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4576ed0fb78fSIlya Dryomov /* this is either (2) or (3) */ 4577ed0fb78fSIlya Dryomov if (!atomic_read(&fs_info->balance_running)) { 4578ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4579ed0fb78fSIlya Dryomov if (!mutex_trylock(&fs_info->volume_mutex)) 4580ed0fb78fSIlya Dryomov goto again; 4581ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4582ed0fb78fSIlya Dryomov 4583ed0fb78fSIlya Dryomov if (fs_info->balance_ctl && 4584ed0fb78fSIlya Dryomov !atomic_read(&fs_info->balance_running)) { 4585ed0fb78fSIlya Dryomov /* this is (3) */ 4586ed0fb78fSIlya Dryomov need_unlock = false; 4587ed0fb78fSIlya Dryomov goto locked; 4588ed0fb78fSIlya Dryomov } 4589ed0fb78fSIlya Dryomov 4590ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4591ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->volume_mutex); 4592ed0fb78fSIlya Dryomov goto again; 4593ed0fb78fSIlya Dryomov } else { 4594ed0fb78fSIlya Dryomov /* this is (2) */ 4595ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4596ed0fb78fSIlya Dryomov ret = -EINPROGRESS; 4597ed0fb78fSIlya Dryomov goto out; 4598ed0fb78fSIlya Dryomov } 4599ed0fb78fSIlya Dryomov } else { 4600ed0fb78fSIlya Dryomov /* this is (1) */ 4601ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4602e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 4603ed0fb78fSIlya Dryomov goto out; 4604ed0fb78fSIlya Dryomov } 4605ed0fb78fSIlya Dryomov 4606ed0fb78fSIlya Dryomov locked: 4607ed0fb78fSIlya Dryomov BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running)); 4608c9e9f97bSIlya Dryomov 4609c9e9f97bSIlya Dryomov if (arg) { 4610c9e9f97bSIlya Dryomov bargs = memdup_user(arg, sizeof(*bargs)); 4611c9e9f97bSIlya Dryomov if (IS_ERR(bargs)) { 4612c9e9f97bSIlya Dryomov ret = PTR_ERR(bargs); 4613ed0fb78fSIlya Dryomov goto out_unlock; 4614c9e9f97bSIlya Dryomov } 4615de322263SIlya Dryomov 4616de322263SIlya Dryomov if (bargs->flags & BTRFS_BALANCE_RESUME) { 4617de322263SIlya Dryomov if (!fs_info->balance_ctl) { 4618de322263SIlya Dryomov ret = -ENOTCONN; 4619de322263SIlya Dryomov goto out_bargs; 4620de322263SIlya Dryomov } 4621de322263SIlya Dryomov 4622de322263SIlya Dryomov bctl = fs_info->balance_ctl; 4623de322263SIlya Dryomov spin_lock(&fs_info->balance_lock); 4624de322263SIlya Dryomov bctl->flags |= BTRFS_BALANCE_RESUME; 4625de322263SIlya Dryomov spin_unlock(&fs_info->balance_lock); 4626de322263SIlya Dryomov 4627de322263SIlya Dryomov goto do_balance; 4628de322263SIlya Dryomov } 4629c9e9f97bSIlya Dryomov } else { 4630c9e9f97bSIlya Dryomov bargs = NULL; 4631c9e9f97bSIlya Dryomov } 4632c9e9f97bSIlya Dryomov 4633ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4634837d5b6eSIlya Dryomov ret = -EINPROGRESS; 4635837d5b6eSIlya Dryomov goto out_bargs; 4636837d5b6eSIlya Dryomov } 4637837d5b6eSIlya Dryomov 4638c9e9f97bSIlya Dryomov bctl = kzalloc(sizeof(*bctl), GFP_NOFS); 4639c9e9f97bSIlya Dryomov if (!bctl) { 4640c9e9f97bSIlya Dryomov ret = -ENOMEM; 4641c9e9f97bSIlya Dryomov goto out_bargs; 4642c9e9f97bSIlya Dryomov } 4643c9e9f97bSIlya Dryomov 4644c9e9f97bSIlya Dryomov bctl->fs_info = fs_info; 4645c9e9f97bSIlya Dryomov if (arg) { 4646c9e9f97bSIlya Dryomov memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); 4647c9e9f97bSIlya Dryomov memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); 4648c9e9f97bSIlya Dryomov memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); 4649c9e9f97bSIlya Dryomov 4650c9e9f97bSIlya Dryomov bctl->flags = bargs->flags; 4651f43ffb60SIlya Dryomov } else { 4652f43ffb60SIlya Dryomov /* balance everything - no filters */ 4653f43ffb60SIlya Dryomov bctl->flags |= BTRFS_BALANCE_TYPE_MASK; 4654c9e9f97bSIlya Dryomov } 4655c9e9f97bSIlya Dryomov 4656de322263SIlya Dryomov do_balance: 4657c9e9f97bSIlya Dryomov /* 4658ed0fb78fSIlya Dryomov * Ownership of bctl and mutually_exclusive_operation_running 4659ed0fb78fSIlya Dryomov * goes to to btrfs_balance. bctl is freed in __cancel_balance, 4660ed0fb78fSIlya Dryomov * or, if restriper was paused all the way until unmount, in 4661ed0fb78fSIlya Dryomov * free_fs_info. mutually_exclusive_operation_running is 4662ed0fb78fSIlya Dryomov * cleared in __cancel_balance. 4663c9e9f97bSIlya Dryomov */ 4664ed0fb78fSIlya Dryomov need_unlock = false; 4665ed0fb78fSIlya Dryomov 4666ed0fb78fSIlya Dryomov ret = btrfs_balance(bctl, bargs); 4667ed0fb78fSIlya Dryomov 4668c9e9f97bSIlya Dryomov if (arg) { 4669c9e9f97bSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 4670c9e9f97bSIlya Dryomov ret = -EFAULT; 4671c9e9f97bSIlya Dryomov } 4672c9e9f97bSIlya Dryomov 4673c9e9f97bSIlya Dryomov out_bargs: 4674c9e9f97bSIlya Dryomov kfree(bargs); 4675ed0fb78fSIlya Dryomov out_unlock: 4676c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4677c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->volume_mutex); 4678ed0fb78fSIlya Dryomov if (need_unlock) 4679ed0fb78fSIlya Dryomov atomic_set(&fs_info->mutually_exclusive_operation_running, 0); 4680ed0fb78fSIlya Dryomov out: 4681e54bfa31SLiu Bo mnt_drop_write_file(file); 4682c9e9f97bSIlya Dryomov return ret; 4683c9e9f97bSIlya Dryomov } 4684c9e9f97bSIlya Dryomov 4685837d5b6eSIlya Dryomov static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd) 4686837d5b6eSIlya Dryomov { 4687837d5b6eSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4688837d5b6eSIlya Dryomov return -EPERM; 4689837d5b6eSIlya Dryomov 4690837d5b6eSIlya Dryomov switch (cmd) { 4691837d5b6eSIlya Dryomov case BTRFS_BALANCE_CTL_PAUSE: 4692837d5b6eSIlya Dryomov return btrfs_pause_balance(root->fs_info); 4693a7e99c69SIlya Dryomov case BTRFS_BALANCE_CTL_CANCEL: 4694a7e99c69SIlya Dryomov return btrfs_cancel_balance(root->fs_info); 4695837d5b6eSIlya Dryomov } 4696837d5b6eSIlya Dryomov 4697837d5b6eSIlya Dryomov return -EINVAL; 4698837d5b6eSIlya Dryomov } 4699837d5b6eSIlya Dryomov 470019a39dceSIlya Dryomov static long btrfs_ioctl_balance_progress(struct btrfs_root *root, 470119a39dceSIlya Dryomov void __user *arg) 470219a39dceSIlya Dryomov { 470319a39dceSIlya Dryomov struct btrfs_fs_info *fs_info = root->fs_info; 470419a39dceSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 470519a39dceSIlya Dryomov int ret = 0; 470619a39dceSIlya Dryomov 470719a39dceSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 470819a39dceSIlya Dryomov return -EPERM; 470919a39dceSIlya Dryomov 471019a39dceSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 471119a39dceSIlya Dryomov if (!fs_info->balance_ctl) { 471219a39dceSIlya Dryomov ret = -ENOTCONN; 471319a39dceSIlya Dryomov goto out; 471419a39dceSIlya Dryomov } 471519a39dceSIlya Dryomov 471619a39dceSIlya Dryomov bargs = kzalloc(sizeof(*bargs), GFP_NOFS); 471719a39dceSIlya Dryomov if (!bargs) { 471819a39dceSIlya Dryomov ret = -ENOMEM; 471919a39dceSIlya Dryomov goto out; 472019a39dceSIlya Dryomov } 472119a39dceSIlya Dryomov 472219a39dceSIlya Dryomov update_ioctl_balance_args(fs_info, 1, bargs); 472319a39dceSIlya Dryomov 472419a39dceSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 472519a39dceSIlya Dryomov ret = -EFAULT; 472619a39dceSIlya Dryomov 472719a39dceSIlya Dryomov kfree(bargs); 472819a39dceSIlya Dryomov out: 472919a39dceSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 473019a39dceSIlya Dryomov return ret; 473119a39dceSIlya Dryomov } 473219a39dceSIlya Dryomov 4733905b0ddaSMiao Xie static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) 47345d13a37bSArne Jansen { 4735496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 47365d13a37bSArne Jansen struct btrfs_ioctl_quota_ctl_args *sa; 47375d13a37bSArne Jansen struct btrfs_trans_handle *trans = NULL; 47385d13a37bSArne Jansen int ret; 47395d13a37bSArne Jansen int err; 47405d13a37bSArne Jansen 47415d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 47425d13a37bSArne Jansen return -EPERM; 47435d13a37bSArne Jansen 4744905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4745905b0ddaSMiao Xie if (ret) 4746905b0ddaSMiao Xie return ret; 47475d13a37bSArne Jansen 47485d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4749905b0ddaSMiao Xie if (IS_ERR(sa)) { 4750905b0ddaSMiao Xie ret = PTR_ERR(sa); 4751905b0ddaSMiao Xie goto drop_write; 4752905b0ddaSMiao Xie } 47535d13a37bSArne Jansen 47547708f029SWang Shilong down_write(&root->fs_info->subvol_sem); 475592f183aaSWang Shilong trans = btrfs_start_transaction(root->fs_info->tree_root, 2); 47565d13a37bSArne Jansen if (IS_ERR(trans)) { 47575d13a37bSArne Jansen ret = PTR_ERR(trans); 47585d13a37bSArne Jansen goto out; 47595d13a37bSArne Jansen } 47605d13a37bSArne Jansen 47615d13a37bSArne Jansen switch (sa->cmd) { 47625d13a37bSArne Jansen case BTRFS_QUOTA_CTL_ENABLE: 47635d13a37bSArne Jansen ret = btrfs_quota_enable(trans, root->fs_info); 47645d13a37bSArne Jansen break; 47655d13a37bSArne Jansen case BTRFS_QUOTA_CTL_DISABLE: 47665d13a37bSArne Jansen ret = btrfs_quota_disable(trans, root->fs_info); 47675d13a37bSArne Jansen break; 47685d13a37bSArne Jansen default: 47695d13a37bSArne Jansen ret = -EINVAL; 47705d13a37bSArne Jansen break; 47715d13a37bSArne Jansen } 47725d13a37bSArne Jansen 477392f183aaSWang Shilong err = btrfs_commit_transaction(trans, root->fs_info->tree_root); 47745d13a37bSArne Jansen if (err && !ret) 47755d13a37bSArne Jansen ret = err; 47765d13a37bSArne Jansen out: 47775d13a37bSArne Jansen kfree(sa); 47787708f029SWang Shilong up_write(&root->fs_info->subvol_sem); 4779905b0ddaSMiao Xie drop_write: 4780905b0ddaSMiao Xie mnt_drop_write_file(file); 47815d13a37bSArne Jansen return ret; 47825d13a37bSArne Jansen } 47835d13a37bSArne Jansen 4784905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg) 47855d13a37bSArne Jansen { 4786496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 47875d13a37bSArne Jansen struct btrfs_ioctl_qgroup_assign_args *sa; 47885d13a37bSArne Jansen struct btrfs_trans_handle *trans; 47895d13a37bSArne Jansen int ret; 47905d13a37bSArne Jansen int err; 47915d13a37bSArne Jansen 47925d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 47935d13a37bSArne Jansen return -EPERM; 47945d13a37bSArne Jansen 4795905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4796905b0ddaSMiao Xie if (ret) 4797905b0ddaSMiao Xie return ret; 47985d13a37bSArne Jansen 47995d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4800905b0ddaSMiao Xie if (IS_ERR(sa)) { 4801905b0ddaSMiao Xie ret = PTR_ERR(sa); 4802905b0ddaSMiao Xie goto drop_write; 4803905b0ddaSMiao Xie } 48045d13a37bSArne Jansen 48055d13a37bSArne Jansen trans = btrfs_join_transaction(root); 48065d13a37bSArne Jansen if (IS_ERR(trans)) { 48075d13a37bSArne Jansen ret = PTR_ERR(trans); 48085d13a37bSArne Jansen goto out; 48095d13a37bSArne Jansen } 48105d13a37bSArne Jansen 48115d13a37bSArne Jansen /* FIXME: check if the IDs really exist */ 48125d13a37bSArne Jansen if (sa->assign) { 48135d13a37bSArne Jansen ret = btrfs_add_qgroup_relation(trans, root->fs_info, 48145d13a37bSArne Jansen sa->src, sa->dst); 48155d13a37bSArne Jansen } else { 48165d13a37bSArne Jansen ret = btrfs_del_qgroup_relation(trans, root->fs_info, 48175d13a37bSArne Jansen sa->src, sa->dst); 48185d13a37bSArne Jansen } 48195d13a37bSArne Jansen 4820e082f563SQu Wenruo /* update qgroup status and info */ 4821e082f563SQu Wenruo err = btrfs_run_qgroups(trans, root->fs_info); 4822e082f563SQu Wenruo if (err < 0) 4823e082f563SQu Wenruo btrfs_error(root->fs_info, ret, 4824e082f563SQu Wenruo "failed to update qgroup status and info\n"); 48255d13a37bSArne Jansen err = btrfs_end_transaction(trans, root); 48265d13a37bSArne Jansen if (err && !ret) 48275d13a37bSArne Jansen ret = err; 48285d13a37bSArne Jansen 48295d13a37bSArne Jansen out: 48305d13a37bSArne Jansen kfree(sa); 4831905b0ddaSMiao Xie drop_write: 4832905b0ddaSMiao Xie mnt_drop_write_file(file); 48335d13a37bSArne Jansen return ret; 48345d13a37bSArne Jansen } 48355d13a37bSArne Jansen 4836905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg) 48375d13a37bSArne Jansen { 4838496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 48395d13a37bSArne Jansen struct btrfs_ioctl_qgroup_create_args *sa; 48405d13a37bSArne Jansen struct btrfs_trans_handle *trans; 48415d13a37bSArne Jansen int ret; 48425d13a37bSArne Jansen int err; 48435d13a37bSArne Jansen 48445d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 48455d13a37bSArne Jansen return -EPERM; 48465d13a37bSArne Jansen 4847905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4848905b0ddaSMiao Xie if (ret) 4849905b0ddaSMiao Xie return ret; 48505d13a37bSArne Jansen 48515d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4852905b0ddaSMiao Xie if (IS_ERR(sa)) { 4853905b0ddaSMiao Xie ret = PTR_ERR(sa); 4854905b0ddaSMiao Xie goto drop_write; 4855905b0ddaSMiao Xie } 48565d13a37bSArne Jansen 4857d86e56cfSMiao Xie if (!sa->qgroupid) { 4858d86e56cfSMiao Xie ret = -EINVAL; 4859d86e56cfSMiao Xie goto out; 4860d86e56cfSMiao Xie } 4861d86e56cfSMiao Xie 48625d13a37bSArne Jansen trans = btrfs_join_transaction(root); 48635d13a37bSArne Jansen if (IS_ERR(trans)) { 48645d13a37bSArne Jansen ret = PTR_ERR(trans); 48655d13a37bSArne Jansen goto out; 48665d13a37bSArne Jansen } 48675d13a37bSArne Jansen 48685d13a37bSArne Jansen /* FIXME: check if the IDs really exist */ 48695d13a37bSArne Jansen if (sa->create) { 48704087cf24SDongsheng Yang ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid); 48715d13a37bSArne Jansen } else { 48725d13a37bSArne Jansen ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid); 48735d13a37bSArne Jansen } 48745d13a37bSArne Jansen 48755d13a37bSArne Jansen err = btrfs_end_transaction(trans, root); 48765d13a37bSArne Jansen if (err && !ret) 48775d13a37bSArne Jansen ret = err; 48785d13a37bSArne Jansen 48795d13a37bSArne Jansen out: 48805d13a37bSArne Jansen kfree(sa); 4881905b0ddaSMiao Xie drop_write: 4882905b0ddaSMiao Xie mnt_drop_write_file(file); 48835d13a37bSArne Jansen return ret; 48845d13a37bSArne Jansen } 48855d13a37bSArne Jansen 4886905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg) 48875d13a37bSArne Jansen { 4888496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 48895d13a37bSArne Jansen struct btrfs_ioctl_qgroup_limit_args *sa; 48905d13a37bSArne Jansen struct btrfs_trans_handle *trans; 48915d13a37bSArne Jansen int ret; 48925d13a37bSArne Jansen int err; 48935d13a37bSArne Jansen u64 qgroupid; 48945d13a37bSArne Jansen 48955d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 48965d13a37bSArne Jansen return -EPERM; 48975d13a37bSArne Jansen 4898905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4899905b0ddaSMiao Xie if (ret) 4900905b0ddaSMiao Xie return ret; 49015d13a37bSArne Jansen 49025d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4903905b0ddaSMiao Xie if (IS_ERR(sa)) { 4904905b0ddaSMiao Xie ret = PTR_ERR(sa); 4905905b0ddaSMiao Xie goto drop_write; 4906905b0ddaSMiao Xie } 49075d13a37bSArne Jansen 49085d13a37bSArne Jansen trans = btrfs_join_transaction(root); 49095d13a37bSArne Jansen if (IS_ERR(trans)) { 49105d13a37bSArne Jansen ret = PTR_ERR(trans); 49115d13a37bSArne Jansen goto out; 49125d13a37bSArne Jansen } 49135d13a37bSArne Jansen 49145d13a37bSArne Jansen qgroupid = sa->qgroupid; 49155d13a37bSArne Jansen if (!qgroupid) { 49165d13a37bSArne Jansen /* take the current subvol as qgroup */ 49175d13a37bSArne Jansen qgroupid = root->root_key.objectid; 49185d13a37bSArne Jansen } 49195d13a37bSArne Jansen 49205d13a37bSArne Jansen /* FIXME: check if the IDs really exist */ 49215d13a37bSArne Jansen ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim); 49225d13a37bSArne Jansen 49235d13a37bSArne Jansen err = btrfs_end_transaction(trans, root); 49245d13a37bSArne Jansen if (err && !ret) 49255d13a37bSArne Jansen ret = err; 49265d13a37bSArne Jansen 49275d13a37bSArne Jansen out: 49285d13a37bSArne Jansen kfree(sa); 4929905b0ddaSMiao Xie drop_write: 4930905b0ddaSMiao Xie mnt_drop_write_file(file); 49315d13a37bSArne Jansen return ret; 49325d13a37bSArne Jansen } 49335d13a37bSArne Jansen 49342f232036SJan Schmidt static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) 49352f232036SJan Schmidt { 49366d0379ecSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 49372f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 49382f232036SJan Schmidt int ret; 49392f232036SJan Schmidt 49402f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 49412f232036SJan Schmidt return -EPERM; 49422f232036SJan Schmidt 49432f232036SJan Schmidt ret = mnt_want_write_file(file); 49442f232036SJan Schmidt if (ret) 49452f232036SJan Schmidt return ret; 49462f232036SJan Schmidt 49472f232036SJan Schmidt qsa = memdup_user(arg, sizeof(*qsa)); 49482f232036SJan Schmidt if (IS_ERR(qsa)) { 49492f232036SJan Schmidt ret = PTR_ERR(qsa); 49502f232036SJan Schmidt goto drop_write; 49512f232036SJan Schmidt } 49522f232036SJan Schmidt 49532f232036SJan Schmidt if (qsa->flags) { 49542f232036SJan Schmidt ret = -EINVAL; 49552f232036SJan Schmidt goto out; 49562f232036SJan Schmidt } 49572f232036SJan Schmidt 49582f232036SJan Schmidt ret = btrfs_qgroup_rescan(root->fs_info); 49592f232036SJan Schmidt 49602f232036SJan Schmidt out: 49612f232036SJan Schmidt kfree(qsa); 49622f232036SJan Schmidt drop_write: 49632f232036SJan Schmidt mnt_drop_write_file(file); 49642f232036SJan Schmidt return ret; 49652f232036SJan Schmidt } 49662f232036SJan Schmidt 49672f232036SJan Schmidt static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg) 49682f232036SJan Schmidt { 49696d0379ecSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 49702f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 49712f232036SJan Schmidt int ret = 0; 49722f232036SJan Schmidt 49732f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 49742f232036SJan Schmidt return -EPERM; 49752f232036SJan Schmidt 49762f232036SJan Schmidt qsa = kzalloc(sizeof(*qsa), GFP_NOFS); 49772f232036SJan Schmidt if (!qsa) 49782f232036SJan Schmidt return -ENOMEM; 49792f232036SJan Schmidt 49802f232036SJan Schmidt if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { 49812f232036SJan Schmidt qsa->flags = 1; 49822f232036SJan Schmidt qsa->progress = root->fs_info->qgroup_rescan_progress.objectid; 49832f232036SJan Schmidt } 49842f232036SJan Schmidt 49852f232036SJan Schmidt if (copy_to_user(arg, qsa, sizeof(*qsa))) 49862f232036SJan Schmidt ret = -EFAULT; 49872f232036SJan Schmidt 49882f232036SJan Schmidt kfree(qsa); 49892f232036SJan Schmidt return ret; 49902f232036SJan Schmidt } 49912f232036SJan Schmidt 499257254b6eSJan Schmidt static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg) 499357254b6eSJan Schmidt { 499454563d41SAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 499557254b6eSJan Schmidt 499657254b6eSJan Schmidt if (!capable(CAP_SYS_ADMIN)) 499757254b6eSJan Schmidt return -EPERM; 499857254b6eSJan Schmidt 499957254b6eSJan Schmidt return btrfs_qgroup_wait_for_completion(root->fs_info); 500057254b6eSJan Schmidt } 500157254b6eSJan Schmidt 5002abccd00fSHugo Mills static long _btrfs_ioctl_set_received_subvol(struct file *file, 5003abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa) 50048ea05e3aSAlexander Block { 5005496ad9aaSAl Viro struct inode *inode = file_inode(file); 50068ea05e3aSAlexander Block struct btrfs_root *root = BTRFS_I(inode)->root; 50078ea05e3aSAlexander Block struct btrfs_root_item *root_item = &root->root_item; 50088ea05e3aSAlexander Block struct btrfs_trans_handle *trans; 50098ea05e3aSAlexander Block struct timespec ct = CURRENT_TIME; 50108ea05e3aSAlexander Block int ret = 0; 5011dd5f9615SStefan Behrens int received_uuid_changed; 50128ea05e3aSAlexander Block 5013bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 5014bd60ea0fSDavid Sterba return -EPERM; 5015bd60ea0fSDavid Sterba 50168ea05e3aSAlexander Block ret = mnt_want_write_file(file); 50178ea05e3aSAlexander Block if (ret < 0) 50188ea05e3aSAlexander Block return ret; 50198ea05e3aSAlexander Block 50208ea05e3aSAlexander Block down_write(&root->fs_info->subvol_sem); 50218ea05e3aSAlexander Block 50228ea05e3aSAlexander Block if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { 50238ea05e3aSAlexander Block ret = -EINVAL; 50248ea05e3aSAlexander Block goto out; 50258ea05e3aSAlexander Block } 50268ea05e3aSAlexander Block 50278ea05e3aSAlexander Block if (btrfs_root_readonly(root)) { 50288ea05e3aSAlexander Block ret = -EROFS; 50298ea05e3aSAlexander Block goto out; 50308ea05e3aSAlexander Block } 50318ea05e3aSAlexander Block 5032dd5f9615SStefan Behrens /* 5033dd5f9615SStefan Behrens * 1 - root item 5034dd5f9615SStefan Behrens * 2 - uuid items (received uuid + subvol uuid) 5035dd5f9615SStefan Behrens */ 5036dd5f9615SStefan Behrens trans = btrfs_start_transaction(root, 3); 50378ea05e3aSAlexander Block if (IS_ERR(trans)) { 50388ea05e3aSAlexander Block ret = PTR_ERR(trans); 50398ea05e3aSAlexander Block trans = NULL; 50408ea05e3aSAlexander Block goto out; 50418ea05e3aSAlexander Block } 50428ea05e3aSAlexander Block 50438ea05e3aSAlexander Block sa->rtransid = trans->transid; 50448ea05e3aSAlexander Block sa->rtime.sec = ct.tv_sec; 50458ea05e3aSAlexander Block sa->rtime.nsec = ct.tv_nsec; 50468ea05e3aSAlexander Block 5047dd5f9615SStefan Behrens received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid, 5048dd5f9615SStefan Behrens BTRFS_UUID_SIZE); 5049dd5f9615SStefan Behrens if (received_uuid_changed && 5050dd5f9615SStefan Behrens !btrfs_is_empty_uuid(root_item->received_uuid)) 5051dd5f9615SStefan Behrens btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root, 5052dd5f9615SStefan Behrens root_item->received_uuid, 5053dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5054dd5f9615SStefan Behrens root->root_key.objectid); 50558ea05e3aSAlexander Block memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE); 50568ea05e3aSAlexander Block btrfs_set_root_stransid(root_item, sa->stransid); 50578ea05e3aSAlexander Block btrfs_set_root_rtransid(root_item, sa->rtransid); 50583cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec); 50593cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec); 50603cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec); 50613cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec); 50628ea05e3aSAlexander Block 50638ea05e3aSAlexander Block ret = btrfs_update_root(trans, root->fs_info->tree_root, 50648ea05e3aSAlexander Block &root->root_key, &root->root_item); 50658ea05e3aSAlexander Block if (ret < 0) { 50668ea05e3aSAlexander Block btrfs_end_transaction(trans, root); 50678ea05e3aSAlexander Block goto out; 5068dd5f9615SStefan Behrens } 5069dd5f9615SStefan Behrens if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { 5070dd5f9615SStefan Behrens ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root, 5071dd5f9615SStefan Behrens sa->uuid, 5072dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5073dd5f9615SStefan Behrens root->root_key.objectid); 5074dd5f9615SStefan Behrens if (ret < 0 && ret != -EEXIST) { 5075dd5f9615SStefan Behrens btrfs_abort_transaction(trans, root, ret); 5076dd5f9615SStefan Behrens goto out; 5077dd5f9615SStefan Behrens } 5078dd5f9615SStefan Behrens } 50798ea05e3aSAlexander Block ret = btrfs_commit_transaction(trans, root); 5080dd5f9615SStefan Behrens if (ret < 0) { 5081dd5f9615SStefan Behrens btrfs_abort_transaction(trans, root, ret); 50828ea05e3aSAlexander Block goto out; 50838ea05e3aSAlexander Block } 50848ea05e3aSAlexander Block 5085abccd00fSHugo Mills out: 5086abccd00fSHugo Mills up_write(&root->fs_info->subvol_sem); 5087abccd00fSHugo Mills mnt_drop_write_file(file); 5088abccd00fSHugo Mills return ret; 5089abccd00fSHugo Mills } 5090abccd00fSHugo Mills 5091abccd00fSHugo Mills #ifdef CONFIG_64BIT 5092abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol_32(struct file *file, 5093abccd00fSHugo Mills void __user *arg) 5094abccd00fSHugo Mills { 5095abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL; 5096abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *args64 = NULL; 5097abccd00fSHugo Mills int ret = 0; 5098abccd00fSHugo Mills 5099abccd00fSHugo Mills args32 = memdup_user(arg, sizeof(*args32)); 5100abccd00fSHugo Mills if (IS_ERR(args32)) { 5101abccd00fSHugo Mills ret = PTR_ERR(args32); 5102abccd00fSHugo Mills args32 = NULL; 5103abccd00fSHugo Mills goto out; 5104abccd00fSHugo Mills } 5105abccd00fSHugo Mills 5106abccd00fSHugo Mills args64 = kmalloc(sizeof(*args64), GFP_NOFS); 510784dbeb87SDan Carpenter if (!args64) { 510884dbeb87SDan Carpenter ret = -ENOMEM; 5109abccd00fSHugo Mills goto out; 5110abccd00fSHugo Mills } 5111abccd00fSHugo Mills 5112abccd00fSHugo Mills memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE); 5113abccd00fSHugo Mills args64->stransid = args32->stransid; 5114abccd00fSHugo Mills args64->rtransid = args32->rtransid; 5115abccd00fSHugo Mills args64->stime.sec = args32->stime.sec; 5116abccd00fSHugo Mills args64->stime.nsec = args32->stime.nsec; 5117abccd00fSHugo Mills args64->rtime.sec = args32->rtime.sec; 5118abccd00fSHugo Mills args64->rtime.nsec = args32->rtime.nsec; 5119abccd00fSHugo Mills args64->flags = args32->flags; 5120abccd00fSHugo Mills 5121abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, args64); 5122abccd00fSHugo Mills if (ret) 5123abccd00fSHugo Mills goto out; 5124abccd00fSHugo Mills 5125abccd00fSHugo Mills memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE); 5126abccd00fSHugo Mills args32->stransid = args64->stransid; 5127abccd00fSHugo Mills args32->rtransid = args64->rtransid; 5128abccd00fSHugo Mills args32->stime.sec = args64->stime.sec; 5129abccd00fSHugo Mills args32->stime.nsec = args64->stime.nsec; 5130abccd00fSHugo Mills args32->rtime.sec = args64->rtime.sec; 5131abccd00fSHugo Mills args32->rtime.nsec = args64->rtime.nsec; 5132abccd00fSHugo Mills args32->flags = args64->flags; 5133abccd00fSHugo Mills 5134abccd00fSHugo Mills ret = copy_to_user(arg, args32, sizeof(*args32)); 5135abccd00fSHugo Mills if (ret) 5136abccd00fSHugo Mills ret = -EFAULT; 5137abccd00fSHugo Mills 5138abccd00fSHugo Mills out: 5139abccd00fSHugo Mills kfree(args32); 5140abccd00fSHugo Mills kfree(args64); 5141abccd00fSHugo Mills return ret; 5142abccd00fSHugo Mills } 5143abccd00fSHugo Mills #endif 5144abccd00fSHugo Mills 5145abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol(struct file *file, 5146abccd00fSHugo Mills void __user *arg) 5147abccd00fSHugo Mills { 5148abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa = NULL; 5149abccd00fSHugo Mills int ret = 0; 5150abccd00fSHugo Mills 5151abccd00fSHugo Mills sa = memdup_user(arg, sizeof(*sa)); 5152abccd00fSHugo Mills if (IS_ERR(sa)) { 5153abccd00fSHugo Mills ret = PTR_ERR(sa); 5154abccd00fSHugo Mills sa = NULL; 5155abccd00fSHugo Mills goto out; 5156abccd00fSHugo Mills } 5157abccd00fSHugo Mills 5158abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, sa); 5159abccd00fSHugo Mills 5160abccd00fSHugo Mills if (ret) 5161abccd00fSHugo Mills goto out; 5162abccd00fSHugo Mills 51638ea05e3aSAlexander Block ret = copy_to_user(arg, sa, sizeof(*sa)); 51648ea05e3aSAlexander Block if (ret) 51658ea05e3aSAlexander Block ret = -EFAULT; 51668ea05e3aSAlexander Block 51678ea05e3aSAlexander Block out: 51688ea05e3aSAlexander Block kfree(sa); 51698ea05e3aSAlexander Block return ret; 51708ea05e3aSAlexander Block } 51718ea05e3aSAlexander Block 5172867ab667Sjeff.liu static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg) 5173867ab667Sjeff.liu { 51746d0379ecSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 5175a1b83ac5SAnand Jain size_t len; 5176867ab667Sjeff.liu int ret; 5177a1b83ac5SAnand Jain char label[BTRFS_LABEL_SIZE]; 5178a1b83ac5SAnand Jain 5179a1b83ac5SAnand Jain spin_lock(&root->fs_info->super_lock); 5180a1b83ac5SAnand Jain memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE); 5181a1b83ac5SAnand Jain spin_unlock(&root->fs_info->super_lock); 5182a1b83ac5SAnand Jain 5183a1b83ac5SAnand Jain len = strnlen(label, BTRFS_LABEL_SIZE); 5184867ab667Sjeff.liu 5185867ab667Sjeff.liu if (len == BTRFS_LABEL_SIZE) { 5186efe120a0SFrank Holton btrfs_warn(root->fs_info, 5187efe120a0SFrank Holton "label is too long, return the first %zu bytes", --len); 5188867ab667Sjeff.liu } 5189867ab667Sjeff.liu 5190867ab667Sjeff.liu ret = copy_to_user(arg, label, len); 5191867ab667Sjeff.liu 5192867ab667Sjeff.liu return ret ? -EFAULT : 0; 5193867ab667Sjeff.liu } 5194867ab667Sjeff.liu 5195a8bfd4abSjeff.liu static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg) 5196a8bfd4abSjeff.liu { 51976d0379ecSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 5198a8bfd4abSjeff.liu struct btrfs_super_block *super_block = root->fs_info->super_copy; 5199a8bfd4abSjeff.liu struct btrfs_trans_handle *trans; 5200a8bfd4abSjeff.liu char label[BTRFS_LABEL_SIZE]; 5201a8bfd4abSjeff.liu int ret; 5202a8bfd4abSjeff.liu 5203a8bfd4abSjeff.liu if (!capable(CAP_SYS_ADMIN)) 5204a8bfd4abSjeff.liu return -EPERM; 5205a8bfd4abSjeff.liu 5206a8bfd4abSjeff.liu if (copy_from_user(label, arg, sizeof(label))) 5207a8bfd4abSjeff.liu return -EFAULT; 5208a8bfd4abSjeff.liu 5209a8bfd4abSjeff.liu if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) { 5210efe120a0SFrank Holton btrfs_err(root->fs_info, "unable to set label with more than %d bytes", 5211a8bfd4abSjeff.liu BTRFS_LABEL_SIZE - 1); 5212a8bfd4abSjeff.liu return -EINVAL; 5213a8bfd4abSjeff.liu } 5214a8bfd4abSjeff.liu 5215a8bfd4abSjeff.liu ret = mnt_want_write_file(file); 5216a8bfd4abSjeff.liu if (ret) 5217a8bfd4abSjeff.liu return ret; 5218a8bfd4abSjeff.liu 5219a8bfd4abSjeff.liu trans = btrfs_start_transaction(root, 0); 5220a8bfd4abSjeff.liu if (IS_ERR(trans)) { 5221a8bfd4abSjeff.liu ret = PTR_ERR(trans); 5222a8bfd4abSjeff.liu goto out_unlock; 5223a8bfd4abSjeff.liu } 5224a8bfd4abSjeff.liu 5225a1b83ac5SAnand Jain spin_lock(&root->fs_info->super_lock); 5226a8bfd4abSjeff.liu strcpy(super_block->label, label); 5227a1b83ac5SAnand Jain spin_unlock(&root->fs_info->super_lock); 5228d0270acaSJeff Mahoney ret = btrfs_commit_transaction(trans, root); 5229a8bfd4abSjeff.liu 5230a8bfd4abSjeff.liu out_unlock: 5231a8bfd4abSjeff.liu mnt_drop_write_file(file); 5232a8bfd4abSjeff.liu return ret; 5233a8bfd4abSjeff.liu } 5234a8bfd4abSjeff.liu 52352eaa055fSJeff Mahoney #define INIT_FEATURE_FLAGS(suffix) \ 52362eaa055fSJeff Mahoney { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \ 52372eaa055fSJeff Mahoney .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \ 52382eaa055fSJeff Mahoney .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix } 52392eaa055fSJeff Mahoney 52402eaa055fSJeff Mahoney static int btrfs_ioctl_get_supported_features(struct file *file, 52412eaa055fSJeff Mahoney void __user *arg) 52422eaa055fSJeff Mahoney { 52432eaa055fSJeff Mahoney static struct btrfs_ioctl_feature_flags features[3] = { 52442eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SUPP), 52452eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_SET), 52462eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_CLEAR) 52472eaa055fSJeff Mahoney }; 52482eaa055fSJeff Mahoney 52492eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 52502eaa055fSJeff Mahoney return -EFAULT; 52512eaa055fSJeff Mahoney 52522eaa055fSJeff Mahoney return 0; 52532eaa055fSJeff Mahoney } 52542eaa055fSJeff Mahoney 52552eaa055fSJeff Mahoney static int btrfs_ioctl_get_features(struct file *file, void __user *arg) 52562eaa055fSJeff Mahoney { 52572eaa055fSJeff Mahoney struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 52582eaa055fSJeff Mahoney struct btrfs_super_block *super_block = root->fs_info->super_copy; 52592eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags features; 52602eaa055fSJeff Mahoney 52612eaa055fSJeff Mahoney features.compat_flags = btrfs_super_compat_flags(super_block); 52622eaa055fSJeff Mahoney features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block); 52632eaa055fSJeff Mahoney features.incompat_flags = btrfs_super_incompat_flags(super_block); 52642eaa055fSJeff Mahoney 52652eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 52662eaa055fSJeff Mahoney return -EFAULT; 52672eaa055fSJeff Mahoney 52682eaa055fSJeff Mahoney return 0; 52692eaa055fSJeff Mahoney } 52702eaa055fSJeff Mahoney 52713b02a68aSJeff Mahoney static int check_feature_bits(struct btrfs_root *root, 52723b02a68aSJeff Mahoney enum btrfs_feature_set set, 52732eaa055fSJeff Mahoney u64 change_mask, u64 flags, u64 supported_flags, 52742eaa055fSJeff Mahoney u64 safe_set, u64 safe_clear) 52752eaa055fSJeff Mahoney { 52763b02a68aSJeff Mahoney const char *type = btrfs_feature_set_names[set]; 52773b02a68aSJeff Mahoney char *names; 52782eaa055fSJeff Mahoney u64 disallowed, unsupported; 52792eaa055fSJeff Mahoney u64 set_mask = flags & change_mask; 52802eaa055fSJeff Mahoney u64 clear_mask = ~flags & change_mask; 52812eaa055fSJeff Mahoney 52822eaa055fSJeff Mahoney unsupported = set_mask & ~supported_flags; 52832eaa055fSJeff Mahoney if (unsupported) { 52843b02a68aSJeff Mahoney names = btrfs_printable_features(set, unsupported); 52853b02a68aSJeff Mahoney if (names) { 52863b02a68aSJeff Mahoney btrfs_warn(root->fs_info, 52873b02a68aSJeff Mahoney "this kernel does not support the %s feature bit%s", 52883b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 52893b02a68aSJeff Mahoney kfree(names); 52903b02a68aSJeff Mahoney } else 52912eaa055fSJeff Mahoney btrfs_warn(root->fs_info, 52922eaa055fSJeff Mahoney "this kernel does not support %s bits 0x%llx", 52932eaa055fSJeff Mahoney type, unsupported); 52942eaa055fSJeff Mahoney return -EOPNOTSUPP; 52952eaa055fSJeff Mahoney } 52962eaa055fSJeff Mahoney 52972eaa055fSJeff Mahoney disallowed = set_mask & ~safe_set; 52982eaa055fSJeff Mahoney if (disallowed) { 52993b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 53003b02a68aSJeff Mahoney if (names) { 53013b02a68aSJeff Mahoney btrfs_warn(root->fs_info, 53023b02a68aSJeff Mahoney "can't set the %s feature bit%s while mounted", 53033b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 53043b02a68aSJeff Mahoney kfree(names); 53053b02a68aSJeff Mahoney } else 53062eaa055fSJeff Mahoney btrfs_warn(root->fs_info, 53072eaa055fSJeff Mahoney "can't set %s bits 0x%llx while mounted", 53082eaa055fSJeff Mahoney type, disallowed); 53092eaa055fSJeff Mahoney return -EPERM; 53102eaa055fSJeff Mahoney } 53112eaa055fSJeff Mahoney 53122eaa055fSJeff Mahoney disallowed = clear_mask & ~safe_clear; 53132eaa055fSJeff Mahoney if (disallowed) { 53143b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 53153b02a68aSJeff Mahoney if (names) { 53163b02a68aSJeff Mahoney btrfs_warn(root->fs_info, 53173b02a68aSJeff Mahoney "can't clear the %s feature bit%s while mounted", 53183b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 53193b02a68aSJeff Mahoney kfree(names); 53203b02a68aSJeff Mahoney } else 53212eaa055fSJeff Mahoney btrfs_warn(root->fs_info, 53222eaa055fSJeff Mahoney "can't clear %s bits 0x%llx while mounted", 53232eaa055fSJeff Mahoney type, disallowed); 53242eaa055fSJeff Mahoney return -EPERM; 53252eaa055fSJeff Mahoney } 53262eaa055fSJeff Mahoney 53272eaa055fSJeff Mahoney return 0; 53282eaa055fSJeff Mahoney } 53292eaa055fSJeff Mahoney 53302eaa055fSJeff Mahoney #define check_feature(root, change_mask, flags, mask_base) \ 53313b02a68aSJeff Mahoney check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \ 53322eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SUPP, \ 53332eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \ 53342eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR) 53352eaa055fSJeff Mahoney 53362eaa055fSJeff Mahoney static int btrfs_ioctl_set_features(struct file *file, void __user *arg) 53372eaa055fSJeff Mahoney { 53382eaa055fSJeff Mahoney struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 53392eaa055fSJeff Mahoney struct btrfs_super_block *super_block = root->fs_info->super_copy; 53402eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags flags[2]; 53412eaa055fSJeff Mahoney struct btrfs_trans_handle *trans; 53422eaa055fSJeff Mahoney u64 newflags; 53432eaa055fSJeff Mahoney int ret; 53442eaa055fSJeff Mahoney 53452eaa055fSJeff Mahoney if (!capable(CAP_SYS_ADMIN)) 53462eaa055fSJeff Mahoney return -EPERM; 53472eaa055fSJeff Mahoney 53482eaa055fSJeff Mahoney if (copy_from_user(flags, arg, sizeof(flags))) 53492eaa055fSJeff Mahoney return -EFAULT; 53502eaa055fSJeff Mahoney 53512eaa055fSJeff Mahoney /* Nothing to do */ 53522eaa055fSJeff Mahoney if (!flags[0].compat_flags && !flags[0].compat_ro_flags && 53532eaa055fSJeff Mahoney !flags[0].incompat_flags) 53542eaa055fSJeff Mahoney return 0; 53552eaa055fSJeff Mahoney 53562eaa055fSJeff Mahoney ret = check_feature(root, flags[0].compat_flags, 53572eaa055fSJeff Mahoney flags[1].compat_flags, COMPAT); 53582eaa055fSJeff Mahoney if (ret) 53592eaa055fSJeff Mahoney return ret; 53602eaa055fSJeff Mahoney 53612eaa055fSJeff Mahoney ret = check_feature(root, flags[0].compat_ro_flags, 53622eaa055fSJeff Mahoney flags[1].compat_ro_flags, COMPAT_RO); 53632eaa055fSJeff Mahoney if (ret) 53642eaa055fSJeff Mahoney return ret; 53652eaa055fSJeff Mahoney 53662eaa055fSJeff Mahoney ret = check_feature(root, flags[0].incompat_flags, 53672eaa055fSJeff Mahoney flags[1].incompat_flags, INCOMPAT); 53682eaa055fSJeff Mahoney if (ret) 53692eaa055fSJeff Mahoney return ret; 53702eaa055fSJeff Mahoney 53718051aa1aSDavid Sterba trans = btrfs_start_transaction(root, 0); 53722eaa055fSJeff Mahoney if (IS_ERR(trans)) 53732eaa055fSJeff Mahoney return PTR_ERR(trans); 53742eaa055fSJeff Mahoney 53752eaa055fSJeff Mahoney spin_lock(&root->fs_info->super_lock); 53762eaa055fSJeff Mahoney newflags = btrfs_super_compat_flags(super_block); 53772eaa055fSJeff Mahoney newflags |= flags[0].compat_flags & flags[1].compat_flags; 53782eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags); 53792eaa055fSJeff Mahoney btrfs_set_super_compat_flags(super_block, newflags); 53802eaa055fSJeff Mahoney 53812eaa055fSJeff Mahoney newflags = btrfs_super_compat_ro_flags(super_block); 53822eaa055fSJeff Mahoney newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags; 53832eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags); 53842eaa055fSJeff Mahoney btrfs_set_super_compat_ro_flags(super_block, newflags); 53852eaa055fSJeff Mahoney 53862eaa055fSJeff Mahoney newflags = btrfs_super_incompat_flags(super_block); 53872eaa055fSJeff Mahoney newflags |= flags[0].incompat_flags & flags[1].incompat_flags; 53882eaa055fSJeff Mahoney newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags); 53892eaa055fSJeff Mahoney btrfs_set_super_incompat_flags(super_block, newflags); 53902eaa055fSJeff Mahoney spin_unlock(&root->fs_info->super_lock); 53912eaa055fSJeff Mahoney 5392d0270acaSJeff Mahoney return btrfs_commit_transaction(trans, root); 53932eaa055fSJeff Mahoney } 53942eaa055fSJeff Mahoney 5395f46b5a66SChristoph Hellwig long btrfs_ioctl(struct file *file, unsigned int 5396f46b5a66SChristoph Hellwig cmd, unsigned long arg) 5397f46b5a66SChristoph Hellwig { 5398496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 53994bcabaa3SChristoph Hellwig void __user *argp = (void __user *)arg; 5400f46b5a66SChristoph Hellwig 5401f46b5a66SChristoph Hellwig switch (cmd) { 54026cbff00fSChristoph Hellwig case FS_IOC_GETFLAGS: 54036cbff00fSChristoph Hellwig return btrfs_ioctl_getflags(file, argp); 54046cbff00fSChristoph Hellwig case FS_IOC_SETFLAGS: 54056cbff00fSChristoph Hellwig return btrfs_ioctl_setflags(file, argp); 54066cbff00fSChristoph Hellwig case FS_IOC_GETVERSION: 54076cbff00fSChristoph Hellwig return btrfs_ioctl_getversion(file, argp); 5408f7039b1dSLi Dongyang case FITRIM: 5409f7039b1dSLi Dongyang return btrfs_ioctl_fitrim(file, argp); 5410f46b5a66SChristoph Hellwig case BTRFS_IOC_SNAP_CREATE: 5411fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 0); 5412fdfb1e4fSLi Zefan case BTRFS_IOC_SNAP_CREATE_V2: 5413fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create_v2(file, argp, 0); 54143de4586cSChris Mason case BTRFS_IOC_SUBVOL_CREATE: 5415fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 1); 54166f72c7e2SArne Jansen case BTRFS_IOC_SUBVOL_CREATE_V2: 54176f72c7e2SArne Jansen return btrfs_ioctl_snap_create_v2(file, argp, 1); 541876dda93cSYan, Zheng case BTRFS_IOC_SNAP_DESTROY: 541976dda93cSYan, Zheng return btrfs_ioctl_snap_destroy(file, argp); 54200caa102dSLi Zefan case BTRFS_IOC_SUBVOL_GETFLAGS: 54210caa102dSLi Zefan return btrfs_ioctl_subvol_getflags(file, argp); 54220caa102dSLi Zefan case BTRFS_IOC_SUBVOL_SETFLAGS: 54230caa102dSLi Zefan return btrfs_ioctl_subvol_setflags(file, argp); 54246ef5ed0dSJosef Bacik case BTRFS_IOC_DEFAULT_SUBVOL: 54256ef5ed0dSJosef Bacik return btrfs_ioctl_default_subvol(file, argp); 5426f46b5a66SChristoph Hellwig case BTRFS_IOC_DEFRAG: 54271e701a32SChris Mason return btrfs_ioctl_defrag(file, NULL); 54281e701a32SChris Mason case BTRFS_IOC_DEFRAG_RANGE: 54291e701a32SChris Mason return btrfs_ioctl_defrag(file, argp); 5430f46b5a66SChristoph Hellwig case BTRFS_IOC_RESIZE: 5431198605a8SMiao Xie return btrfs_ioctl_resize(file, argp); 5432f46b5a66SChristoph Hellwig case BTRFS_IOC_ADD_DEV: 54334bcabaa3SChristoph Hellwig return btrfs_ioctl_add_dev(root, argp); 5434f46b5a66SChristoph Hellwig case BTRFS_IOC_RM_DEV: 5435da24927bSMiao Xie return btrfs_ioctl_rm_dev(file, argp); 5436475f6387SJan Schmidt case BTRFS_IOC_FS_INFO: 5437475f6387SJan Schmidt return btrfs_ioctl_fs_info(root, argp); 5438475f6387SJan Schmidt case BTRFS_IOC_DEV_INFO: 5439475f6387SJan Schmidt return btrfs_ioctl_dev_info(root, argp); 5440f46b5a66SChristoph Hellwig case BTRFS_IOC_BALANCE: 54419ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, NULL); 5442f46b5a66SChristoph Hellwig case BTRFS_IOC_CLONE: 5443c5c9cd4dSSage Weil return btrfs_ioctl_clone(file, arg, 0, 0, 0); 5444c5c9cd4dSSage Weil case BTRFS_IOC_CLONE_RANGE: 54457a865e8aSChristoph Hellwig return btrfs_ioctl_clone_range(file, argp); 5446f46b5a66SChristoph Hellwig case BTRFS_IOC_TRANS_START: 5447f46b5a66SChristoph Hellwig return btrfs_ioctl_trans_start(file); 5448f46b5a66SChristoph Hellwig case BTRFS_IOC_TRANS_END: 5449f46b5a66SChristoph Hellwig return btrfs_ioctl_trans_end(file); 5450ac8e9819SChris Mason case BTRFS_IOC_TREE_SEARCH: 5451ac8e9819SChris Mason return btrfs_ioctl_tree_search(file, argp); 5452cc68a8a5SGerhard Heift case BTRFS_IOC_TREE_SEARCH_V2: 5453cc68a8a5SGerhard Heift return btrfs_ioctl_tree_search_v2(file, argp); 5454ac8e9819SChris Mason case BTRFS_IOC_INO_LOOKUP: 5455ac8e9819SChris Mason return btrfs_ioctl_ino_lookup(file, argp); 5456d7728c96SJan Schmidt case BTRFS_IOC_INO_PATHS: 5457d7728c96SJan Schmidt return btrfs_ioctl_ino_to_path(root, argp); 5458d7728c96SJan Schmidt case BTRFS_IOC_LOGICAL_INO: 5459d7728c96SJan Schmidt return btrfs_ioctl_logical_to_ino(root, argp); 54601406e432SJosef Bacik case BTRFS_IOC_SPACE_INFO: 54611406e432SJosef Bacik return btrfs_ioctl_space_info(root, argp); 54629b199859SFilipe David Borba Manana case BTRFS_IOC_SYNC: { 54639b199859SFilipe David Borba Manana int ret; 54649b199859SFilipe David Borba Manana 54656c255e67SMiao Xie ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1); 54669b199859SFilipe David Borba Manana if (ret) 54679b199859SFilipe David Borba Manana return ret; 5468ddb52f4fSAl Viro ret = btrfs_sync_fs(file_inode(file)->i_sb, 1); 54692fad4e83SDavid Sterba /* 54702fad4e83SDavid Sterba * The transaction thread may want to do more work, 54712fad4e83SDavid Sterba * namely it pokes the cleaner ktread that will start 54722fad4e83SDavid Sterba * processing uncleaned subvols. 54732fad4e83SDavid Sterba */ 54742fad4e83SDavid Sterba wake_up_process(root->fs_info->transaction_kthread); 54759b199859SFilipe David Borba Manana return ret; 54769b199859SFilipe David Borba Manana } 547746204592SSage Weil case BTRFS_IOC_START_SYNC: 54789a8c28beSMiao Xie return btrfs_ioctl_start_sync(root, argp); 547946204592SSage Weil case BTRFS_IOC_WAIT_SYNC: 54809a8c28beSMiao Xie return btrfs_ioctl_wait_sync(root, argp); 5481475f6387SJan Schmidt case BTRFS_IOC_SCRUB: 5482b8e95489SMiao Xie return btrfs_ioctl_scrub(file, argp); 5483475f6387SJan Schmidt case BTRFS_IOC_SCRUB_CANCEL: 5484475f6387SJan Schmidt return btrfs_ioctl_scrub_cancel(root, argp); 5485475f6387SJan Schmidt case BTRFS_IOC_SCRUB_PROGRESS: 5486475f6387SJan Schmidt return btrfs_ioctl_scrub_progress(root, argp); 5487c9e9f97bSIlya Dryomov case BTRFS_IOC_BALANCE_V2: 54889ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, argp); 5489837d5b6eSIlya Dryomov case BTRFS_IOC_BALANCE_CTL: 5490837d5b6eSIlya Dryomov return btrfs_ioctl_balance_ctl(root, arg); 549119a39dceSIlya Dryomov case BTRFS_IOC_BALANCE_PROGRESS: 549219a39dceSIlya Dryomov return btrfs_ioctl_balance_progress(root, argp); 54938ea05e3aSAlexander Block case BTRFS_IOC_SET_RECEIVED_SUBVOL: 54948ea05e3aSAlexander Block return btrfs_ioctl_set_received_subvol(file, argp); 5495abccd00fSHugo Mills #ifdef CONFIG_64BIT 5496abccd00fSHugo Mills case BTRFS_IOC_SET_RECEIVED_SUBVOL_32: 5497abccd00fSHugo Mills return btrfs_ioctl_set_received_subvol_32(file, argp); 5498abccd00fSHugo Mills #endif 549931db9f7cSAlexander Block case BTRFS_IOC_SEND: 550031db9f7cSAlexander Block return btrfs_ioctl_send(file, argp); 5501c11d2c23SStefan Behrens case BTRFS_IOC_GET_DEV_STATS: 5502b27f7c0cSDavid Sterba return btrfs_ioctl_get_dev_stats(root, argp); 55035d13a37bSArne Jansen case BTRFS_IOC_QUOTA_CTL: 5504905b0ddaSMiao Xie return btrfs_ioctl_quota_ctl(file, argp); 55055d13a37bSArne Jansen case BTRFS_IOC_QGROUP_ASSIGN: 5506905b0ddaSMiao Xie return btrfs_ioctl_qgroup_assign(file, argp); 55075d13a37bSArne Jansen case BTRFS_IOC_QGROUP_CREATE: 5508905b0ddaSMiao Xie return btrfs_ioctl_qgroup_create(file, argp); 55095d13a37bSArne Jansen case BTRFS_IOC_QGROUP_LIMIT: 5510905b0ddaSMiao Xie return btrfs_ioctl_qgroup_limit(file, argp); 55112f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN: 55122f232036SJan Schmidt return btrfs_ioctl_quota_rescan(file, argp); 55132f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_STATUS: 55142f232036SJan Schmidt return btrfs_ioctl_quota_rescan_status(file, argp); 551557254b6eSJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_WAIT: 551657254b6eSJan Schmidt return btrfs_ioctl_quota_rescan_wait(file, argp); 55173f6bcfbdSStefan Behrens case BTRFS_IOC_DEV_REPLACE: 55183f6bcfbdSStefan Behrens return btrfs_ioctl_dev_replace(root, argp); 5519867ab667Sjeff.liu case BTRFS_IOC_GET_FSLABEL: 5520867ab667Sjeff.liu return btrfs_ioctl_get_fslabel(file, argp); 5521a8bfd4abSjeff.liu case BTRFS_IOC_SET_FSLABEL: 5522a8bfd4abSjeff.liu return btrfs_ioctl_set_fslabel(file, argp); 5523416161dbSMark Fasheh case BTRFS_IOC_FILE_EXTENT_SAME: 5524416161dbSMark Fasheh return btrfs_ioctl_file_extent_same(file, argp); 55252eaa055fSJeff Mahoney case BTRFS_IOC_GET_SUPPORTED_FEATURES: 55262eaa055fSJeff Mahoney return btrfs_ioctl_get_supported_features(file, argp); 55272eaa055fSJeff Mahoney case BTRFS_IOC_GET_FEATURES: 55282eaa055fSJeff Mahoney return btrfs_ioctl_get_features(file, argp); 55292eaa055fSJeff Mahoney case BTRFS_IOC_SET_FEATURES: 55302eaa055fSJeff Mahoney return btrfs_ioctl_set_features(file, argp); 5531f46b5a66SChristoph Hellwig } 5532f46b5a66SChristoph Hellwig 5533f46b5a66SChristoph Hellwig return -ENOTTY; 5534f46b5a66SChristoph Hellwig } 5535