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 658a1ee7362SDavid Sterba pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS); 659a1ee7362SDavid Sterba if (!pending_snapshot) 660a1ee7362SDavid Sterba return -ENOMEM; 661a1ee7362SDavid Sterba 6628257b2dcSMiao Xie atomic_inc(&root->will_be_snapshoted); 6634e857c58SPeter Zijlstra smp_mb__after_atomic(); 6649ea24bbeSFilipe Manana btrfs_wait_for_no_snapshoting_writes(root); 6658257b2dcSMiao Xie 6666a03843dSMiao Xie ret = btrfs_start_delalloc_inodes(root, 0); 6676a03843dSMiao Xie if (ret) 668a1ee7362SDavid Sterba goto dec_and_free; 6696a03843dSMiao Xie 670b0244199SMiao Xie btrfs_wait_ordered_extents(root, -1); 6716a03843dSMiao Xie 67266d8f3ddSMiao Xie btrfs_init_block_rsv(&pending_snapshot->block_rsv, 67366d8f3ddSMiao Xie BTRFS_BLOCK_RSV_TEMP); 674d5c12070SMiao Xie /* 675d5c12070SMiao Xie * 1 - parent dir inode 676d5c12070SMiao Xie * 2 - dir entries 677d5c12070SMiao Xie * 1 - root item 678d5c12070SMiao Xie * 2 - root ref/backref 679d5c12070SMiao Xie * 1 - root of snapshot 680dd5f9615SStefan Behrens * 1 - UUID item 681d5c12070SMiao Xie */ 682d5c12070SMiao Xie ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, 683dd5f9615SStefan Behrens &pending_snapshot->block_rsv, 8, 684ee3441b4SJeff Mahoney &pending_snapshot->qgroup_reserved, 685ee3441b4SJeff Mahoney false); 686d5c12070SMiao Xie if (ret) 687a1ee7362SDavid Sterba goto dec_and_free; 688d5c12070SMiao Xie 689a22285a6SYan, Zheng pending_snapshot->dentry = dentry; 690a22285a6SYan, Zheng pending_snapshot->root = root; 691b83cc969SLi Zefan pending_snapshot->readonly = readonly; 692e9662f70SMiao Xie pending_snapshot->dir = dir; 6938696c533SMiao Xie pending_snapshot->inherit = inherit; 694a22285a6SYan, Zheng 695d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 696a22285a6SYan, Zheng if (IS_ERR(trans)) { 697a22285a6SYan, Zheng ret = PTR_ERR(trans); 698a22285a6SYan, Zheng goto fail; 699a22285a6SYan, Zheng } 700a22285a6SYan, Zheng 7018351583eSJosef Bacik spin_lock(&root->fs_info->trans_lock); 702a22285a6SYan, Zheng list_add(&pending_snapshot->list, 703a22285a6SYan, Zheng &trans->transaction->pending_snapshots); 7048351583eSJosef Bacik spin_unlock(&root->fs_info->trans_lock); 70572fd032eSSage Weil if (async_transid) { 70672fd032eSSage Weil *async_transid = trans->transid; 70772fd032eSSage Weil ret = btrfs_commit_transaction_async(trans, 70872fd032eSSage Weil root->fs_info->extent_root, 1); 70900d71c9cSMiao Xie if (ret) 71000d71c9cSMiao Xie ret = btrfs_commit_transaction(trans, root); 71172fd032eSSage Weil } else { 71272fd032eSSage Weil ret = btrfs_commit_transaction(trans, 71372fd032eSSage Weil root->fs_info->extent_root); 71472fd032eSSage Weil } 715aec8030aSMiao Xie if (ret) 716c37b2b62SJosef Bacik goto fail; 717a22285a6SYan, Zheng 718a22285a6SYan, Zheng ret = pending_snapshot->error; 719f46b5a66SChristoph Hellwig if (ret) 7202e4bfab9SYan, Zheng goto fail; 721f46b5a66SChristoph Hellwig 722d3797308SChris Mason ret = btrfs_orphan_cleanup(pending_snapshot->snap); 723d3797308SChris Mason if (ret) 724d3797308SChris Mason goto fail; 725d3797308SChris Mason 7262b0143b5SDavid Howells inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry); 7272e4bfab9SYan, Zheng if (IS_ERR(inode)) { 7282e4bfab9SYan, Zheng ret = PTR_ERR(inode); 7292e4bfab9SYan, Zheng goto fail; 7302e4bfab9SYan, Zheng } 7315662344bSTsutomu Itoh 7322e4bfab9SYan, Zheng d_instantiate(dentry, inode); 7332e4bfab9SYan, Zheng ret = 0; 7342e4bfab9SYan, Zheng fail: 735d5c12070SMiao Xie btrfs_subvolume_release_metadata(BTRFS_I(dir)->root, 736d5c12070SMiao Xie &pending_snapshot->block_rsv, 737d5c12070SMiao Xie pending_snapshot->qgroup_reserved); 738a1ee7362SDavid Sterba dec_and_free: 7399ea24bbeSFilipe Manana if (atomic_dec_and_test(&root->will_be_snapshoted)) 7409ea24bbeSFilipe Manana wake_up_atomic_t(&root->will_be_snapshoted); 741a1ee7362SDavid Sterba kfree(pending_snapshot); 742a1ee7362SDavid Sterba 743f46b5a66SChristoph Hellwig return ret; 744f46b5a66SChristoph Hellwig } 745f46b5a66SChristoph Hellwig 7464260f7c7SSage Weil /* copy of may_delete in fs/namei.c() 7474260f7c7SSage Weil * Check whether we can remove a link victim from directory dir, check 7484260f7c7SSage Weil * whether the type of victim is right. 7494260f7c7SSage Weil * 1. We can't do it if dir is read-only (done in permission()) 7504260f7c7SSage Weil * 2. We should have write and exec permissions on dir 7514260f7c7SSage Weil * 3. We can't remove anything from append-only dir 7524260f7c7SSage Weil * 4. We can't do anything with immutable dir (done in permission()) 7534260f7c7SSage Weil * 5. If the sticky bit on dir is set we should either 7544260f7c7SSage Weil * a. be owner of dir, or 7554260f7c7SSage Weil * b. be owner of victim, or 7564260f7c7SSage Weil * c. have CAP_FOWNER capability 7574260f7c7SSage Weil * 6. If the victim is append-only or immutable we can't do antyhing with 7584260f7c7SSage Weil * links pointing to it. 7594260f7c7SSage Weil * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. 7604260f7c7SSage Weil * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. 7614260f7c7SSage Weil * 9. We can't remove a root or mountpoint. 7624260f7c7SSage Weil * 10. We don't allow removal of NFS sillyrenamed files; it's handled by 7634260f7c7SSage Weil * nfs_async_unlink(). 7644260f7c7SSage Weil */ 7654260f7c7SSage Weil 7664260f7c7SSage Weil static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir) 7674260f7c7SSage Weil { 7684260f7c7SSage Weil int error; 7694260f7c7SSage Weil 7702b0143b5SDavid Howells if (d_really_is_negative(victim)) 7714260f7c7SSage Weil return -ENOENT; 7724260f7c7SSage Weil 7732b0143b5SDavid Howells BUG_ON(d_inode(victim->d_parent) != dir); 7744fa6b5ecSJeff Layton audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); 7754260f7c7SSage Weil 7764260f7c7SSage Weil error = inode_permission(dir, MAY_WRITE | MAY_EXEC); 7774260f7c7SSage Weil if (error) 7784260f7c7SSage Weil return error; 7794260f7c7SSage Weil if (IS_APPEND(dir)) 7804260f7c7SSage Weil return -EPERM; 7812b0143b5SDavid Howells if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) || 7822b0143b5SDavid Howells IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim))) 7834260f7c7SSage Weil return -EPERM; 7844260f7c7SSage Weil if (isdir) { 785e36cb0b8SDavid Howells if (!d_is_dir(victim)) 7864260f7c7SSage Weil return -ENOTDIR; 7874260f7c7SSage Weil if (IS_ROOT(victim)) 7884260f7c7SSage Weil return -EBUSY; 789e36cb0b8SDavid Howells } else if (d_is_dir(victim)) 7904260f7c7SSage Weil return -EISDIR; 7914260f7c7SSage Weil if (IS_DEADDIR(dir)) 7924260f7c7SSage Weil return -ENOENT; 7934260f7c7SSage Weil if (victim->d_flags & DCACHE_NFSFS_RENAMED) 7944260f7c7SSage Weil return -EBUSY; 7954260f7c7SSage Weil return 0; 7964260f7c7SSage Weil } 7974260f7c7SSage Weil 798cb8e7090SChristoph Hellwig /* copy of may_create in fs/namei.c() */ 799cb8e7090SChristoph Hellwig static inline int btrfs_may_create(struct inode *dir, struct dentry *child) 800cb8e7090SChristoph Hellwig { 8012b0143b5SDavid Howells if (d_really_is_positive(child)) 802cb8e7090SChristoph Hellwig return -EEXIST; 803cb8e7090SChristoph Hellwig if (IS_DEADDIR(dir)) 804cb8e7090SChristoph Hellwig return -ENOENT; 805cb8e7090SChristoph Hellwig return inode_permission(dir, MAY_WRITE | MAY_EXEC); 806cb8e7090SChristoph Hellwig } 807cb8e7090SChristoph Hellwig 808cb8e7090SChristoph Hellwig /* 809cb8e7090SChristoph Hellwig * Create a new subvolume below @parent. This is largely modeled after 810cb8e7090SChristoph Hellwig * sys_mkdirat and vfs_mkdir, but we only do a single component lookup 811cb8e7090SChristoph Hellwig * inside this filesystem so it's quite a bit simpler. 812cb8e7090SChristoph Hellwig */ 81376dda93cSYan, Zheng static noinline int btrfs_mksubvol(struct path *parent, 81476dda93cSYan, Zheng char *name, int namelen, 81572fd032eSSage Weil struct btrfs_root *snap_src, 8166f72c7e2SArne Jansen u64 *async_transid, bool readonly, 8178696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 818cb8e7090SChristoph Hellwig { 8192b0143b5SDavid Howells struct inode *dir = d_inode(parent->dentry); 820cb8e7090SChristoph Hellwig struct dentry *dentry; 821cb8e7090SChristoph Hellwig int error; 822cb8e7090SChristoph Hellwig 8235c50c9b8SDavid Sterba error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT); 8245c50c9b8SDavid Sterba if (error == -EINTR) 8255c50c9b8SDavid Sterba return error; 826cb8e7090SChristoph Hellwig 827cb8e7090SChristoph Hellwig dentry = lookup_one_len(name, parent->dentry, namelen); 828cb8e7090SChristoph Hellwig error = PTR_ERR(dentry); 829cb8e7090SChristoph Hellwig if (IS_ERR(dentry)) 830cb8e7090SChristoph Hellwig goto out_unlock; 831cb8e7090SChristoph Hellwig 832cb8e7090SChristoph Hellwig error = -EEXIST; 8332b0143b5SDavid Howells if (d_really_is_positive(dentry)) 834cb8e7090SChristoph Hellwig goto out_dput; 835cb8e7090SChristoph Hellwig 83676dda93cSYan, Zheng error = btrfs_may_create(dir, dentry); 837cb8e7090SChristoph Hellwig if (error) 838a874a63eSLiu Bo goto out_dput; 839cb8e7090SChristoph Hellwig 8409c52057cSChris Mason /* 8419c52057cSChris Mason * even if this name doesn't exist, we may get hash collisions. 8429c52057cSChris Mason * check for them now when we can safely fail 8439c52057cSChris Mason */ 8449c52057cSChris Mason error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, 8459c52057cSChris Mason dir->i_ino, name, 8469c52057cSChris Mason namelen); 8479c52057cSChris Mason if (error) 8489c52057cSChris Mason goto out_dput; 8499c52057cSChris Mason 85076dda93cSYan, Zheng down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem); 85176dda93cSYan, Zheng 85276dda93cSYan, Zheng if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) 85376dda93cSYan, Zheng goto out_up_read; 85476dda93cSYan, Zheng 8553de4586cSChris Mason if (snap_src) { 856e9662f70SMiao Xie error = create_snapshot(snap_src, dir, dentry, name, namelen, 8576f72c7e2SArne Jansen async_transid, readonly, inherit); 8583de4586cSChris Mason } else { 859d5c12070SMiao Xie error = create_subvol(dir, dentry, name, namelen, 860d5c12070SMiao Xie async_transid, inherit); 8613de4586cSChris Mason } 86276dda93cSYan, Zheng if (!error) 86376dda93cSYan, Zheng fsnotify_mkdir(dir, dentry); 86476dda93cSYan, Zheng out_up_read: 86576dda93cSYan, Zheng up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem); 866cb8e7090SChristoph Hellwig out_dput: 867cb8e7090SChristoph Hellwig dput(dentry); 868cb8e7090SChristoph Hellwig out_unlock: 86976dda93cSYan, Zheng mutex_unlock(&dir->i_mutex); 870cb8e7090SChristoph Hellwig return error; 871cb8e7090SChristoph Hellwig } 872cb8e7090SChristoph Hellwig 8734cb5300bSChris Mason /* 8744cb5300bSChris Mason * When we're defragging a range, we don't want to kick it off again 8754cb5300bSChris Mason * if it is really just waiting for delalloc to send it down. 8764cb5300bSChris Mason * If we find a nice big extent or delalloc range for the bytes in the 8774cb5300bSChris Mason * file you want to defrag, we return 0 to let you know to skip this 8784cb5300bSChris Mason * part of the file 8794cb5300bSChris Mason */ 880aab110abSDavid Sterba static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh) 8814cb5300bSChris Mason { 8824cb5300bSChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 8834cb5300bSChris Mason struct extent_map *em = NULL; 8844cb5300bSChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 8854cb5300bSChris Mason u64 end; 8864cb5300bSChris Mason 8874cb5300bSChris Mason read_lock(&em_tree->lock); 8884cb5300bSChris Mason em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE); 8894cb5300bSChris Mason read_unlock(&em_tree->lock); 8904cb5300bSChris Mason 8914cb5300bSChris Mason if (em) { 8924cb5300bSChris Mason end = extent_map_end(em); 8934cb5300bSChris Mason free_extent_map(em); 8944cb5300bSChris Mason if (end - offset > thresh) 8954cb5300bSChris Mason return 0; 8964cb5300bSChris Mason } 8974cb5300bSChris Mason /* if we already have a nice delalloc here, just stop */ 8984cb5300bSChris Mason thresh /= 2; 8994cb5300bSChris Mason end = count_range_bits(io_tree, &offset, offset + thresh, 9004cb5300bSChris Mason thresh, EXTENT_DELALLOC, 1); 9014cb5300bSChris Mason if (end >= thresh) 9024cb5300bSChris Mason return 0; 9034cb5300bSChris Mason return 1; 9044cb5300bSChris Mason } 9054cb5300bSChris Mason 9064cb5300bSChris Mason /* 9074cb5300bSChris Mason * helper function to walk through a file and find extents 9084cb5300bSChris Mason * newer than a specific transid, and smaller than thresh. 9094cb5300bSChris Mason * 9104cb5300bSChris Mason * This is used by the defragging code to find new and small 9114cb5300bSChris Mason * extents 9124cb5300bSChris Mason */ 9134cb5300bSChris Mason static int find_new_extents(struct btrfs_root *root, 9144cb5300bSChris Mason struct inode *inode, u64 newer_than, 915aab110abSDavid Sterba u64 *off, u32 thresh) 9164cb5300bSChris Mason { 9174cb5300bSChris Mason struct btrfs_path *path; 9184cb5300bSChris Mason struct btrfs_key min_key; 9194cb5300bSChris Mason struct extent_buffer *leaf; 9204cb5300bSChris Mason struct btrfs_file_extent_item *extent; 9214cb5300bSChris Mason int type; 9224cb5300bSChris Mason int ret; 923a4689d2bSDavid Sterba u64 ino = btrfs_ino(inode); 9244cb5300bSChris Mason 9254cb5300bSChris Mason path = btrfs_alloc_path(); 9264cb5300bSChris Mason if (!path) 9274cb5300bSChris Mason return -ENOMEM; 9284cb5300bSChris Mason 929a4689d2bSDavid Sterba min_key.objectid = ino; 9304cb5300bSChris Mason min_key.type = BTRFS_EXTENT_DATA_KEY; 9314cb5300bSChris Mason min_key.offset = *off; 9324cb5300bSChris Mason 9334cb5300bSChris Mason while (1) { 9346174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &min_key, path, newer_than); 9354cb5300bSChris Mason if (ret != 0) 9364cb5300bSChris Mason goto none; 937f094c9bdSFilipe Manana process_slot: 938a4689d2bSDavid Sterba if (min_key.objectid != ino) 9394cb5300bSChris Mason goto none; 9404cb5300bSChris Mason if (min_key.type != BTRFS_EXTENT_DATA_KEY) 9414cb5300bSChris Mason goto none; 9424cb5300bSChris Mason 9434cb5300bSChris Mason leaf = path->nodes[0]; 9444cb5300bSChris Mason extent = btrfs_item_ptr(leaf, path->slots[0], 9454cb5300bSChris Mason struct btrfs_file_extent_item); 9464cb5300bSChris Mason 9474cb5300bSChris Mason type = btrfs_file_extent_type(leaf, extent); 9484cb5300bSChris Mason if (type == BTRFS_FILE_EXTENT_REG && 9494cb5300bSChris Mason btrfs_file_extent_num_bytes(leaf, extent) < thresh && 9504cb5300bSChris Mason check_defrag_in_cache(inode, min_key.offset, thresh)) { 9514cb5300bSChris Mason *off = min_key.offset; 9524cb5300bSChris Mason btrfs_free_path(path); 9534cb5300bSChris Mason return 0; 9544cb5300bSChris Mason } 9554cb5300bSChris Mason 956f094c9bdSFilipe Manana path->slots[0]++; 957f094c9bdSFilipe Manana if (path->slots[0] < btrfs_header_nritems(leaf)) { 958f094c9bdSFilipe Manana btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]); 959f094c9bdSFilipe Manana goto process_slot; 960f094c9bdSFilipe Manana } 961f094c9bdSFilipe Manana 9624cb5300bSChris Mason if (min_key.offset == (u64)-1) 9634cb5300bSChris Mason goto none; 9644cb5300bSChris Mason 9654cb5300bSChris Mason min_key.offset++; 9664cb5300bSChris Mason btrfs_release_path(path); 9674cb5300bSChris Mason } 9684cb5300bSChris Mason none: 9694cb5300bSChris Mason btrfs_free_path(path); 9704cb5300bSChris Mason return -ENOENT; 9714cb5300bSChris Mason } 9724cb5300bSChris Mason 9736c282eb4SLi Zefan static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start) 97417ce6ef8SLiu Bo { 97517ce6ef8SLiu Bo struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 976940100a4SChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 9776c282eb4SLi Zefan struct extent_map *em; 9786c282eb4SLi Zefan u64 len = PAGE_CACHE_SIZE; 979940100a4SChris Mason 980940100a4SChris Mason /* 981940100a4SChris Mason * hopefully we have this extent in the tree already, try without 982940100a4SChris Mason * the full extent lock 983940100a4SChris Mason */ 984940100a4SChris Mason read_lock(&em_tree->lock); 985940100a4SChris Mason em = lookup_extent_mapping(em_tree, start, len); 986940100a4SChris Mason read_unlock(&em_tree->lock); 987940100a4SChris Mason 988940100a4SChris Mason if (!em) { 989308d9800SFilipe Manana struct extent_state *cached = NULL; 990308d9800SFilipe Manana u64 end = start + len - 1; 991308d9800SFilipe Manana 992940100a4SChris Mason /* get the big lock and read metadata off disk */ 993308d9800SFilipe Manana lock_extent_bits(io_tree, start, end, 0, &cached); 994940100a4SChris Mason em = btrfs_get_extent(inode, NULL, 0, start, len, 0); 995308d9800SFilipe Manana unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS); 996940100a4SChris Mason 9976cf8bfbfSDan Carpenter if (IS_ERR(em)) 9986c282eb4SLi Zefan return NULL; 999940100a4SChris Mason } 1000940100a4SChris Mason 10016c282eb4SLi Zefan return em; 10026c282eb4SLi Zefan } 10036c282eb4SLi Zefan 10046c282eb4SLi Zefan static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em) 10056c282eb4SLi Zefan { 10066c282eb4SLi Zefan struct extent_map *next; 10076c282eb4SLi Zefan bool ret = true; 10086c282eb4SLi Zefan 10096c282eb4SLi Zefan /* this is the last extent */ 10106c282eb4SLi Zefan if (em->start + em->len >= i_size_read(inode)) 10116c282eb4SLi Zefan return false; 10126c282eb4SLi Zefan 10136c282eb4SLi Zefan next = defrag_lookup_extent(inode, em->start + em->len); 1014e9512d72SChris Mason if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE) 1015e9512d72SChris Mason ret = false; 1016e9512d72SChris Mason else if ((em->block_start + em->block_len == next->block_start) && 1017e9512d72SChris Mason (em->block_len > 128 * 1024 && next->block_len > 128 * 1024)) 10186c282eb4SLi Zefan ret = false; 10196c282eb4SLi Zefan 10206c282eb4SLi Zefan free_extent_map(next); 10216c282eb4SLi Zefan return ret; 10226c282eb4SLi Zefan } 10236c282eb4SLi Zefan 1024aab110abSDavid Sterba static int should_defrag_range(struct inode *inode, u64 start, u32 thresh, 1025a43a2111SAndrew Mahone u64 *last_len, u64 *skip, u64 *defrag_end, 1026a43a2111SAndrew Mahone int compress) 10276c282eb4SLi Zefan { 10286c282eb4SLi Zefan struct extent_map *em; 10296c282eb4SLi Zefan int ret = 1; 10306c282eb4SLi Zefan bool next_mergeable = true; 10314a3560c4SLiu Bo bool prev_mergeable = true; 10326c282eb4SLi Zefan 10336c282eb4SLi Zefan /* 10346c282eb4SLi Zefan * make sure that once we start defragging an extent, we keep on 10356c282eb4SLi Zefan * defragging it 10366c282eb4SLi Zefan */ 10376c282eb4SLi Zefan if (start < *defrag_end) 10386c282eb4SLi Zefan return 1; 10396c282eb4SLi Zefan 10406c282eb4SLi Zefan *skip = 0; 10416c282eb4SLi Zefan 10426c282eb4SLi Zefan em = defrag_lookup_extent(inode, start); 10436c282eb4SLi Zefan if (!em) 10446c282eb4SLi Zefan return 0; 10456c282eb4SLi Zefan 1046940100a4SChris Mason /* this will cover holes, and inline extents */ 104717ce6ef8SLiu Bo if (em->block_start >= EXTENT_MAP_LAST_BYTE) { 1048940100a4SChris Mason ret = 0; 104917ce6ef8SLiu Bo goto out; 105017ce6ef8SLiu Bo } 105117ce6ef8SLiu Bo 10524a3560c4SLiu Bo if (!*defrag_end) 10534a3560c4SLiu Bo prev_mergeable = false; 10544a3560c4SLiu Bo 10556c282eb4SLi Zefan next_mergeable = defrag_check_next_extent(inode, em); 1056940100a4SChris Mason /* 10576c282eb4SLi Zefan * we hit a real extent, if it is big or the next extent is not a 10586c282eb4SLi Zefan * real extent, don't bother defragging it 1059940100a4SChris Mason */ 1060a43a2111SAndrew Mahone if (!compress && (*last_len == 0 || *last_len >= thresh) && 10614a3560c4SLiu Bo (em->len >= thresh || (!next_mergeable && !prev_mergeable))) 1062940100a4SChris Mason ret = 0; 106317ce6ef8SLiu Bo out: 1064940100a4SChris Mason /* 1065940100a4SChris Mason * last_len ends up being a counter of how many bytes we've defragged. 1066940100a4SChris Mason * every time we choose not to defrag an extent, we reset *last_len 1067940100a4SChris Mason * so that the next tiny extent will force a defrag. 1068940100a4SChris Mason * 1069940100a4SChris Mason * The end result of this is that tiny extents before a single big 1070940100a4SChris Mason * extent will force at least part of that big extent to be defragged. 1071940100a4SChris Mason */ 1072940100a4SChris Mason if (ret) { 1073940100a4SChris Mason *defrag_end = extent_map_end(em); 1074940100a4SChris Mason } else { 1075940100a4SChris Mason *last_len = 0; 1076940100a4SChris Mason *skip = extent_map_end(em); 1077940100a4SChris Mason *defrag_end = 0; 1078940100a4SChris Mason } 1079940100a4SChris Mason 1080940100a4SChris Mason free_extent_map(em); 1081940100a4SChris Mason return ret; 1082940100a4SChris Mason } 1083940100a4SChris Mason 10844cb5300bSChris Mason /* 10854cb5300bSChris Mason * it doesn't do much good to defrag one or two pages 10864cb5300bSChris Mason * at a time. This pulls in a nice chunk of pages 10874cb5300bSChris Mason * to COW and defrag. 10884cb5300bSChris Mason * 10894cb5300bSChris Mason * It also makes sure the delalloc code has enough 10904cb5300bSChris Mason * dirty data to avoid making new small extents as part 10914cb5300bSChris Mason * of the defrag 10924cb5300bSChris Mason * 10934cb5300bSChris Mason * It's a good idea to start RA on this range 10944cb5300bSChris Mason * before calling this. 10954cb5300bSChris Mason */ 10964cb5300bSChris Mason static int cluster_pages_for_defrag(struct inode *inode, 10974cb5300bSChris Mason struct page **pages, 10984cb5300bSChris Mason unsigned long start_index, 1099c41570c9SJustin Maggard unsigned long num_pages) 1100f46b5a66SChristoph Hellwig { 11014cb5300bSChris Mason unsigned long file_end; 11024cb5300bSChris Mason u64 isize = i_size_read(inode); 1103f46b5a66SChristoph Hellwig u64 page_start; 1104f46b5a66SChristoph Hellwig u64 page_end; 11051f12bd06SLiu Bo u64 page_cnt; 11064cb5300bSChris Mason int ret; 11074cb5300bSChris Mason int i; 11084cb5300bSChris Mason int i_done; 11094cb5300bSChris Mason struct btrfs_ordered_extent *ordered; 11104cb5300bSChris Mason struct extent_state *cached_state = NULL; 1111600a45e1SMiao Xie struct extent_io_tree *tree; 11123b16a4e3SJosef Bacik gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping); 11134cb5300bSChris Mason 11144cb5300bSChris Mason file_end = (isize - 1) >> PAGE_CACHE_SHIFT; 11151f12bd06SLiu Bo if (!isize || start_index > file_end) 11161f12bd06SLiu Bo return 0; 11171f12bd06SLiu Bo 11181f12bd06SLiu Bo page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1); 11194cb5300bSChris Mason 11204cb5300bSChris Mason ret = btrfs_delalloc_reserve_space(inode, 1121df480633SQu Wenruo start_index << PAGE_CACHE_SHIFT, 11221f12bd06SLiu Bo page_cnt << PAGE_CACHE_SHIFT); 11234cb5300bSChris Mason if (ret) 11244cb5300bSChris Mason return ret; 11254cb5300bSChris Mason i_done = 0; 1126600a45e1SMiao Xie tree = &BTRFS_I(inode)->io_tree; 11274cb5300bSChris Mason 11284cb5300bSChris Mason /* step one, lock all the pages */ 11291f12bd06SLiu Bo for (i = 0; i < page_cnt; i++) { 11304cb5300bSChris Mason struct page *page; 1131600a45e1SMiao Xie again: 1132a94733d0SJosef Bacik page = find_or_create_page(inode->i_mapping, 11333b16a4e3SJosef Bacik start_index + i, mask); 11344cb5300bSChris Mason if (!page) 11354cb5300bSChris Mason break; 11364cb5300bSChris Mason 1137600a45e1SMiao Xie page_start = page_offset(page); 1138600a45e1SMiao Xie page_end = page_start + PAGE_CACHE_SIZE - 1; 1139600a45e1SMiao Xie while (1) { 1140308d9800SFilipe Manana lock_extent_bits(tree, page_start, page_end, 1141308d9800SFilipe Manana 0, &cached_state); 1142600a45e1SMiao Xie ordered = btrfs_lookup_ordered_extent(inode, 1143600a45e1SMiao Xie page_start); 1144308d9800SFilipe Manana unlock_extent_cached(tree, page_start, page_end, 1145308d9800SFilipe Manana &cached_state, GFP_NOFS); 1146600a45e1SMiao Xie if (!ordered) 1147600a45e1SMiao Xie break; 1148600a45e1SMiao Xie 1149600a45e1SMiao Xie unlock_page(page); 1150600a45e1SMiao Xie btrfs_start_ordered_extent(inode, ordered, 1); 1151600a45e1SMiao Xie btrfs_put_ordered_extent(ordered); 1152600a45e1SMiao Xie lock_page(page); 11531f12bd06SLiu Bo /* 11541f12bd06SLiu Bo * we unlocked the page above, so we need check if 11551f12bd06SLiu Bo * it was released or not. 11561f12bd06SLiu Bo */ 11571f12bd06SLiu Bo if (page->mapping != inode->i_mapping) { 11581f12bd06SLiu Bo unlock_page(page); 11591f12bd06SLiu Bo page_cache_release(page); 11601f12bd06SLiu Bo goto again; 11611f12bd06SLiu Bo } 1162600a45e1SMiao Xie } 1163600a45e1SMiao Xie 11644cb5300bSChris Mason if (!PageUptodate(page)) { 11654cb5300bSChris Mason btrfs_readpage(NULL, page); 11664cb5300bSChris Mason lock_page(page); 11674cb5300bSChris Mason if (!PageUptodate(page)) { 11684cb5300bSChris Mason unlock_page(page); 11694cb5300bSChris Mason page_cache_release(page); 11704cb5300bSChris Mason ret = -EIO; 11714cb5300bSChris Mason break; 11724cb5300bSChris Mason } 11734cb5300bSChris Mason } 1174600a45e1SMiao Xie 1175600a45e1SMiao Xie if (page->mapping != inode->i_mapping) { 1176600a45e1SMiao Xie unlock_page(page); 1177600a45e1SMiao Xie page_cache_release(page); 1178600a45e1SMiao Xie goto again; 1179600a45e1SMiao Xie } 1180600a45e1SMiao Xie 11814cb5300bSChris Mason pages[i] = page; 11824cb5300bSChris Mason i_done++; 11834cb5300bSChris Mason } 11844cb5300bSChris Mason if (!i_done || ret) 11854cb5300bSChris Mason goto out; 11864cb5300bSChris Mason 11874cb5300bSChris Mason if (!(inode->i_sb->s_flags & MS_ACTIVE)) 11884cb5300bSChris Mason goto out; 11894cb5300bSChris Mason 11904cb5300bSChris Mason /* 11914cb5300bSChris Mason * so now we have a nice long stream of locked 11924cb5300bSChris Mason * and up to date pages, lets wait on them 11934cb5300bSChris Mason */ 11944cb5300bSChris Mason for (i = 0; i < i_done; i++) 11954cb5300bSChris Mason wait_on_page_writeback(pages[i]); 11964cb5300bSChris Mason 11974cb5300bSChris Mason page_start = page_offset(pages[0]); 11984cb5300bSChris Mason page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE; 11994cb5300bSChris Mason 12004cb5300bSChris Mason lock_extent_bits(&BTRFS_I(inode)->io_tree, 1201d0082371SJeff Mahoney page_start, page_end - 1, 0, &cached_state); 12024cb5300bSChris Mason clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, 12034cb5300bSChris Mason page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC | 12049e8a4a8bSLiu Bo EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0, 12059e8a4a8bSLiu Bo &cached_state, GFP_NOFS); 12064cb5300bSChris Mason 12071f12bd06SLiu Bo if (i_done != page_cnt) { 12089e0baf60SJosef Bacik spin_lock(&BTRFS_I(inode)->lock); 12099e0baf60SJosef Bacik BTRFS_I(inode)->outstanding_extents++; 12109e0baf60SJosef Bacik spin_unlock(&BTRFS_I(inode)->lock); 12114cb5300bSChris Mason btrfs_delalloc_release_space(inode, 1212df480633SQu Wenruo start_index << PAGE_CACHE_SHIFT, 12131f12bd06SLiu Bo (page_cnt - i_done) << PAGE_CACHE_SHIFT); 12144cb5300bSChris Mason } 12154cb5300bSChris Mason 12164cb5300bSChris Mason 12179e8a4a8bSLiu Bo set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1, 12189e8a4a8bSLiu Bo &cached_state, GFP_NOFS); 12194cb5300bSChris Mason 12204cb5300bSChris Mason unlock_extent_cached(&BTRFS_I(inode)->io_tree, 12214cb5300bSChris Mason page_start, page_end - 1, &cached_state, 12224cb5300bSChris Mason GFP_NOFS); 12234cb5300bSChris Mason 12244cb5300bSChris Mason for (i = 0; i < i_done; i++) { 12254cb5300bSChris Mason clear_page_dirty_for_io(pages[i]); 12264cb5300bSChris Mason ClearPageChecked(pages[i]); 12274cb5300bSChris Mason set_page_extent_mapped(pages[i]); 12284cb5300bSChris Mason set_page_dirty(pages[i]); 12294cb5300bSChris Mason unlock_page(pages[i]); 12304cb5300bSChris Mason page_cache_release(pages[i]); 12314cb5300bSChris Mason } 12324cb5300bSChris Mason return i_done; 12334cb5300bSChris Mason out: 12344cb5300bSChris Mason for (i = 0; i < i_done; i++) { 12354cb5300bSChris Mason unlock_page(pages[i]); 12364cb5300bSChris Mason page_cache_release(pages[i]); 12374cb5300bSChris Mason } 12387cf5b976SQu Wenruo btrfs_delalloc_release_space(inode, 1239df480633SQu Wenruo start_index << PAGE_CACHE_SHIFT, 1240df480633SQu Wenruo page_cnt << PAGE_CACHE_SHIFT); 12414cb5300bSChris Mason return ret; 12424cb5300bSChris Mason 12434cb5300bSChris Mason } 12444cb5300bSChris Mason 12454cb5300bSChris Mason int btrfs_defrag_file(struct inode *inode, struct file *file, 12464cb5300bSChris Mason struct btrfs_ioctl_defrag_range_args *range, 12474cb5300bSChris Mason u64 newer_than, unsigned long max_to_defrag) 12484cb5300bSChris Mason { 12494cb5300bSChris Mason struct btrfs_root *root = BTRFS_I(inode)->root; 12504cb5300bSChris Mason struct file_ra_state *ra = NULL; 12514cb5300bSChris Mason unsigned long last_index; 1252151a31b2SLi Zefan u64 isize = i_size_read(inode); 1253940100a4SChris Mason u64 last_len = 0; 1254940100a4SChris Mason u64 skip = 0; 1255940100a4SChris Mason u64 defrag_end = 0; 12564cb5300bSChris Mason u64 newer_off = range->start; 1257f46b5a66SChristoph Hellwig unsigned long i; 1258008873eaSLi Zefan unsigned long ra_index = 0; 1259f46b5a66SChristoph Hellwig int ret; 12604cb5300bSChris Mason int defrag_count = 0; 12611a419d85SLi Zefan int compress_type = BTRFS_COMPRESS_ZLIB; 1262aab110abSDavid Sterba u32 extent_thresh = range->extent_thresh; 1263c41570c9SJustin Maggard unsigned long max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT; 1264c41570c9SJustin Maggard unsigned long cluster = max_cluster; 12654cb5300bSChris Mason u64 new_align = ~((u64)128 * 1024 - 1); 12664cb5300bSChris Mason struct page **pages = NULL; 12674cb5300bSChris Mason 12680abd5b17SLiu Bo if (isize == 0) 12690abd5b17SLiu Bo return 0; 12700abd5b17SLiu Bo 12710abd5b17SLiu Bo if (range->start >= isize) 12720abd5b17SLiu Bo return -EINVAL; 12731a419d85SLi Zefan 12741a419d85SLi Zefan if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) { 12751a419d85SLi Zefan if (range->compress_type > BTRFS_COMPRESS_TYPES) 12761a419d85SLi Zefan return -EINVAL; 12771a419d85SLi Zefan if (range->compress_type) 12781a419d85SLi Zefan compress_type = range->compress_type; 12791a419d85SLi Zefan } 1280f46b5a66SChristoph Hellwig 12810abd5b17SLiu Bo if (extent_thresh == 0) 12820abd5b17SLiu Bo extent_thresh = 256 * 1024; 1283f46b5a66SChristoph Hellwig 12844cb5300bSChris Mason /* 12854cb5300bSChris Mason * if we were not given a file, allocate a readahead 12864cb5300bSChris Mason * context 12874cb5300bSChris Mason */ 12884cb5300bSChris Mason if (!file) { 12894cb5300bSChris Mason ra = kzalloc(sizeof(*ra), GFP_NOFS); 12904cb5300bSChris Mason if (!ra) 12914cb5300bSChris Mason return -ENOMEM; 12924cb5300bSChris Mason file_ra_state_init(ra, inode->i_mapping); 12934cb5300bSChris Mason } else { 12944cb5300bSChris Mason ra = &file->f_ra; 12954cb5300bSChris Mason } 12964cb5300bSChris Mason 1297d9b0d9baSDulshani Gunawardhana pages = kmalloc_array(max_cluster, sizeof(struct page *), 12984cb5300bSChris Mason GFP_NOFS); 12994cb5300bSChris Mason if (!pages) { 13004cb5300bSChris Mason ret = -ENOMEM; 13014cb5300bSChris Mason goto out_ra; 13024cb5300bSChris Mason } 13034cb5300bSChris Mason 13044cb5300bSChris Mason /* find the last page to defrag */ 13051e701a32SChris Mason if (range->start + range->len > range->start) { 1306151a31b2SLi Zefan last_index = min_t(u64, isize - 1, 13071e701a32SChris Mason range->start + range->len - 1) >> PAGE_CACHE_SHIFT; 13081e701a32SChris Mason } else { 1309151a31b2SLi Zefan last_index = (isize - 1) >> PAGE_CACHE_SHIFT; 13101e701a32SChris Mason } 13111e701a32SChris Mason 13124cb5300bSChris Mason if (newer_than) { 13134cb5300bSChris Mason ret = find_new_extents(root, inode, newer_than, 13144cb5300bSChris Mason &newer_off, 64 * 1024); 13154cb5300bSChris Mason if (!ret) { 13164cb5300bSChris Mason range->start = newer_off; 13174cb5300bSChris Mason /* 13184cb5300bSChris Mason * we always align our defrag to help keep 13194cb5300bSChris Mason * the extents in the file evenly spaced 13204cb5300bSChris Mason */ 13214cb5300bSChris Mason i = (newer_off & new_align) >> PAGE_CACHE_SHIFT; 13224cb5300bSChris Mason } else 13234cb5300bSChris Mason goto out_ra; 13244cb5300bSChris Mason } else { 13251e701a32SChris Mason i = range->start >> PAGE_CACHE_SHIFT; 13264cb5300bSChris Mason } 13274cb5300bSChris Mason if (!max_to_defrag) 1328070034bdSchandan max_to_defrag = last_index - i + 1; 13294cb5300bSChris Mason 13302a0f7f57SLi Zefan /* 13312a0f7f57SLi Zefan * make writeback starts from i, so the defrag range can be 13322a0f7f57SLi Zefan * written sequentially. 13332a0f7f57SLi Zefan */ 13342a0f7f57SLi Zefan if (i < inode->i_mapping->writeback_index) 13352a0f7f57SLi Zefan inode->i_mapping->writeback_index = i; 13362a0f7f57SLi Zefan 1337f7f43cc8SChris Mason while (i <= last_index && defrag_count < max_to_defrag && 1338ed6078f7SDavid Sterba (i < DIV_ROUND_UP(i_size_read(inode), PAGE_CACHE_SIZE))) { 13394cb5300bSChris Mason /* 13404cb5300bSChris Mason * make sure we stop running if someone unmounts 13414cb5300bSChris Mason * the FS 13424cb5300bSChris Mason */ 13434cb5300bSChris Mason if (!(inode->i_sb->s_flags & MS_ACTIVE)) 13444cb5300bSChris Mason break; 13454cb5300bSChris Mason 1346210549ebSDavid Sterba if (btrfs_defrag_cancelled(root->fs_info)) { 1347f14d104dSDavid Sterba btrfs_debug(root->fs_info, "defrag_file cancelled"); 1348210549ebSDavid Sterba ret = -EAGAIN; 1349210549ebSDavid Sterba break; 1350210549ebSDavid Sterba } 1351210549ebSDavid Sterba 135266c26892SLiu Bo if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT, 13536c282eb4SLi Zefan extent_thresh, &last_len, &skip, 1354a43a2111SAndrew Mahone &defrag_end, range->flags & 1355a43a2111SAndrew Mahone BTRFS_DEFRAG_RANGE_COMPRESS)) { 1356940100a4SChris Mason unsigned long next; 1357940100a4SChris Mason /* 1358940100a4SChris Mason * the should_defrag function tells us how much to skip 1359940100a4SChris Mason * bump our counter by the suggested amount 1360940100a4SChris Mason */ 1361ed6078f7SDavid Sterba next = DIV_ROUND_UP(skip, PAGE_CACHE_SIZE); 1362940100a4SChris Mason i = max(i + 1, next); 1363940100a4SChris Mason continue; 1364940100a4SChris Mason } 1365008873eaSLi Zefan 1366008873eaSLi Zefan if (!newer_than) { 1367008873eaSLi Zefan cluster = (PAGE_CACHE_ALIGN(defrag_end) >> 1368008873eaSLi Zefan PAGE_CACHE_SHIFT) - i; 1369008873eaSLi Zefan cluster = min(cluster, max_cluster); 1370008873eaSLi Zefan } else { 1371008873eaSLi Zefan cluster = max_cluster; 1372008873eaSLi Zefan } 1373008873eaSLi Zefan 1374008873eaSLi Zefan if (i + cluster > ra_index) { 1375008873eaSLi Zefan ra_index = max(i, ra_index); 1376008873eaSLi Zefan btrfs_force_ra(inode->i_mapping, ra, file, ra_index, 1377008873eaSLi Zefan cluster); 1378e4826a5bSchandan ra_index += cluster; 1379008873eaSLi Zefan } 13804cb5300bSChris Mason 1381ecb8bea8SLiu Bo mutex_lock(&inode->i_mutex); 1382633085c7SFilipe David Borba Manana if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) 1383633085c7SFilipe David Borba Manana BTRFS_I(inode)->force_compress = compress_type; 1384008873eaSLi Zefan ret = cluster_pages_for_defrag(inode, pages, i, cluster); 1385ecb8bea8SLiu Bo if (ret < 0) { 1386ecb8bea8SLiu Bo mutex_unlock(&inode->i_mutex); 13874cb5300bSChris Mason goto out_ra; 1388ecb8bea8SLiu Bo } 13894cb5300bSChris Mason 13904cb5300bSChris Mason defrag_count += ret; 1391d0e1d66bSNamjae Jeon balance_dirty_pages_ratelimited(inode->i_mapping); 1392ecb8bea8SLiu Bo mutex_unlock(&inode->i_mutex); 13934cb5300bSChris Mason 13944cb5300bSChris Mason if (newer_than) { 13954cb5300bSChris Mason if (newer_off == (u64)-1) 13964cb5300bSChris Mason break; 13974cb5300bSChris Mason 1398e1f041e1SLiu Bo if (ret > 0) 1399e1f041e1SLiu Bo i += ret; 1400e1f041e1SLiu Bo 14014cb5300bSChris Mason newer_off = max(newer_off + 1, 14024cb5300bSChris Mason (u64)i << PAGE_CACHE_SHIFT); 14034cb5300bSChris Mason 14044cb5300bSChris Mason ret = find_new_extents(root, inode, 14054cb5300bSChris Mason newer_than, &newer_off, 14064cb5300bSChris Mason 64 * 1024); 14074cb5300bSChris Mason if (!ret) { 14084cb5300bSChris Mason range->start = newer_off; 14094cb5300bSChris Mason i = (newer_off & new_align) >> PAGE_CACHE_SHIFT; 14104cb5300bSChris Mason } else { 14114cb5300bSChris Mason break; 1412940100a4SChris Mason } 14134cb5300bSChris Mason } else { 1414008873eaSLi Zefan if (ret > 0) { 1415cbcc8326SLi Zefan i += ret; 1416008873eaSLi Zefan last_len += ret << PAGE_CACHE_SHIFT; 1417008873eaSLi Zefan } else { 1418940100a4SChris Mason i++; 1419008873eaSLi Zefan last_len = 0; 1420008873eaSLi Zefan } 1421f46b5a66SChristoph Hellwig } 14224cb5300bSChris Mason } 1423f46b5a66SChristoph Hellwig 1424dec8ef90SFilipe Manana if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) { 14251e701a32SChris Mason filemap_flush(inode->i_mapping); 1426dec8ef90SFilipe Manana if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT, 1427dec8ef90SFilipe Manana &BTRFS_I(inode)->runtime_flags)) 1428dec8ef90SFilipe Manana filemap_flush(inode->i_mapping); 1429dec8ef90SFilipe Manana } 14301e701a32SChris Mason 14311e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 14321e701a32SChris Mason /* the filemap_flush will queue IO into the worker threads, but 14331e701a32SChris Mason * we have to make sure the IO is actually started and that 14341e701a32SChris Mason * ordered extents get created before we return 14351e701a32SChris Mason */ 14361e701a32SChris Mason atomic_inc(&root->fs_info->async_submit_draining); 14371e701a32SChris Mason while (atomic_read(&root->fs_info->nr_async_submits) || 14381e701a32SChris Mason atomic_read(&root->fs_info->async_delalloc_pages)) { 14391e701a32SChris Mason wait_event(root->fs_info->async_submit_wait, 14401e701a32SChris Mason (atomic_read(&root->fs_info->nr_async_submits) == 0 && 14411e701a32SChris Mason atomic_read(&root->fs_info->async_delalloc_pages) == 0)); 14421e701a32SChris Mason } 14431e701a32SChris Mason atomic_dec(&root->fs_info->async_submit_draining); 14441e701a32SChris Mason } 14451e701a32SChris Mason 14461a419d85SLi Zefan if (range->compress_type == BTRFS_COMPRESS_LZO) { 14472b0ce2c2SMitch Harder btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO); 14481a419d85SLi Zefan } 14491a419d85SLi Zefan 145060ccf82fSDiego Calleja ret = defrag_count; 1451940100a4SChris Mason 14524cb5300bSChris Mason out_ra: 1453633085c7SFilipe David Borba Manana if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) { 1454633085c7SFilipe David Borba Manana mutex_lock(&inode->i_mutex); 1455633085c7SFilipe David Borba Manana BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE; 1456633085c7SFilipe David Borba Manana mutex_unlock(&inode->i_mutex); 1457633085c7SFilipe David Borba Manana } 14584cb5300bSChris Mason if (!file) 14594cb5300bSChris Mason kfree(ra); 14604cb5300bSChris Mason kfree(pages); 1461940100a4SChris Mason return ret; 1462f46b5a66SChristoph Hellwig } 1463f46b5a66SChristoph Hellwig 1464198605a8SMiao Xie static noinline int btrfs_ioctl_resize(struct file *file, 146576dda93cSYan, Zheng void __user *arg) 1466f46b5a66SChristoph Hellwig { 1467f46b5a66SChristoph Hellwig u64 new_size; 1468f46b5a66SChristoph Hellwig u64 old_size; 1469f46b5a66SChristoph Hellwig u64 devid = 1; 1470496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 1471f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 1472f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 1473f46b5a66SChristoph Hellwig struct btrfs_device *device = NULL; 1474f46b5a66SChristoph Hellwig char *sizestr; 14759a40f122SGui Hecheng char *retptr; 1476f46b5a66SChristoph Hellwig char *devstr = NULL; 1477f46b5a66SChristoph Hellwig int ret = 0; 1478f46b5a66SChristoph Hellwig int mod = 0; 1479f46b5a66SChristoph Hellwig 1480e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 1481e441d54dSChris Mason return -EPERM; 1482e441d54dSChris Mason 1483198605a8SMiao Xie ret = mnt_want_write_file(file); 1484198605a8SMiao Xie if (ret) 1485198605a8SMiao Xie return ret; 1486198605a8SMiao Xie 14875ac00addSStefan Behrens if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running, 14885ac00addSStefan Behrens 1)) { 148997547676SMiao Xie mnt_drop_write_file(file); 1490e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 1491c9e9f97bSIlya Dryomov } 1492c9e9f97bSIlya Dryomov 14935ac00addSStefan Behrens mutex_lock(&root->fs_info->volume_mutex); 1494dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1495c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 1496c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 1497c9e9f97bSIlya Dryomov goto out; 1498c9e9f97bSIlya Dryomov } 14995516e595SMark Fasheh 15005516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1501f46b5a66SChristoph Hellwig 1502f46b5a66SChristoph Hellwig sizestr = vol_args->name; 1503f46b5a66SChristoph Hellwig devstr = strchr(sizestr, ':'); 1504f46b5a66SChristoph Hellwig if (devstr) { 1505f46b5a66SChristoph Hellwig sizestr = devstr + 1; 1506f46b5a66SChristoph Hellwig *devstr = '\0'; 1507f46b5a66SChristoph Hellwig devstr = vol_args->name; 150858dfae63SZhangZhen ret = kstrtoull(devstr, 10, &devid); 150958dfae63SZhangZhen if (ret) 151058dfae63SZhangZhen goto out_free; 1511dfd79829SMiao Xie if (!devid) { 1512dfd79829SMiao Xie ret = -EINVAL; 1513dfd79829SMiao Xie goto out_free; 1514dfd79829SMiao Xie } 1515efe120a0SFrank Holton btrfs_info(root->fs_info, "resizing devid %llu", devid); 1516f46b5a66SChristoph Hellwig } 1517dba60f3fSMiao Xie 1518aa1b8cd4SStefan Behrens device = btrfs_find_device(root->fs_info, devid, NULL, NULL); 1519f46b5a66SChristoph Hellwig if (!device) { 1520efe120a0SFrank Holton btrfs_info(root->fs_info, "resizer unable to find device %llu", 1521c1c9ff7cSGeert Uytterhoeven devid); 1522dfd79829SMiao Xie ret = -ENODEV; 1523c9e9f97bSIlya Dryomov goto out_free; 1524f46b5a66SChristoph Hellwig } 1525dba60f3fSMiao Xie 1526dba60f3fSMiao Xie if (!device->writeable) { 1527efe120a0SFrank Holton btrfs_info(root->fs_info, 1528efe120a0SFrank Holton "resizer unable to apply on readonly device %llu", 1529c1c9ff7cSGeert Uytterhoeven devid); 1530dfd79829SMiao Xie ret = -EPERM; 15314e42ae1bSLiu Bo goto out_free; 15324e42ae1bSLiu Bo } 15334e42ae1bSLiu Bo 1534f46b5a66SChristoph Hellwig if (!strcmp(sizestr, "max")) 1535f46b5a66SChristoph Hellwig new_size = device->bdev->bd_inode->i_size; 1536f46b5a66SChristoph Hellwig else { 1537f46b5a66SChristoph Hellwig if (sizestr[0] == '-') { 1538f46b5a66SChristoph Hellwig mod = -1; 1539f46b5a66SChristoph Hellwig sizestr++; 1540f46b5a66SChristoph Hellwig } else if (sizestr[0] == '+') { 1541f46b5a66SChristoph Hellwig mod = 1; 1542f46b5a66SChristoph Hellwig sizestr++; 1543f46b5a66SChristoph Hellwig } 15449a40f122SGui Hecheng new_size = memparse(sizestr, &retptr); 15459a40f122SGui Hecheng if (*retptr != '\0' || new_size == 0) { 1546f46b5a66SChristoph Hellwig ret = -EINVAL; 1547c9e9f97bSIlya Dryomov goto out_free; 1548f46b5a66SChristoph Hellwig } 1549f46b5a66SChristoph Hellwig } 1550f46b5a66SChristoph Hellwig 155163a212abSStefan Behrens if (device->is_tgtdev_for_dev_replace) { 1552dfd79829SMiao Xie ret = -EPERM; 155363a212abSStefan Behrens goto out_free; 155463a212abSStefan Behrens } 155563a212abSStefan Behrens 15567cc8e58dSMiao Xie old_size = btrfs_device_get_total_bytes(device); 1557f46b5a66SChristoph Hellwig 1558f46b5a66SChristoph Hellwig if (mod < 0) { 1559f46b5a66SChristoph Hellwig if (new_size > old_size) { 1560f46b5a66SChristoph Hellwig ret = -EINVAL; 1561c9e9f97bSIlya Dryomov goto out_free; 1562f46b5a66SChristoph Hellwig } 1563f46b5a66SChristoph Hellwig new_size = old_size - new_size; 1564f46b5a66SChristoph Hellwig } else if (mod > 0) { 1565eb8052e0SWenliang Fan if (new_size > ULLONG_MAX - old_size) { 1566902c68a4SGui Hecheng ret = -ERANGE; 1567eb8052e0SWenliang Fan goto out_free; 1568eb8052e0SWenliang Fan } 1569f46b5a66SChristoph Hellwig new_size = old_size + new_size; 1570f46b5a66SChristoph Hellwig } 1571f46b5a66SChristoph Hellwig 1572f46b5a66SChristoph Hellwig if (new_size < 256 * 1024 * 1024) { 1573f46b5a66SChristoph Hellwig ret = -EINVAL; 1574c9e9f97bSIlya Dryomov goto out_free; 1575f46b5a66SChristoph Hellwig } 1576f46b5a66SChristoph Hellwig if (new_size > device->bdev->bd_inode->i_size) { 1577f46b5a66SChristoph Hellwig ret = -EFBIG; 1578c9e9f97bSIlya Dryomov goto out_free; 1579f46b5a66SChristoph Hellwig } 1580f46b5a66SChristoph Hellwig 1581b8b93addSDavid Sterba new_size = div_u64(new_size, root->sectorsize); 1582f46b5a66SChristoph Hellwig new_size *= root->sectorsize; 1583f46b5a66SChristoph Hellwig 1584ecaeb14bSDavid Sterba btrfs_info_in_rcu(root->fs_info, "new size for %s is %llu", 1585c1c9ff7cSGeert Uytterhoeven rcu_str_deref(device->name), new_size); 1586f46b5a66SChristoph Hellwig 1587f46b5a66SChristoph Hellwig if (new_size > old_size) { 1588a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 158998d5dc13STsutomu Itoh if (IS_ERR(trans)) { 159098d5dc13STsutomu Itoh ret = PTR_ERR(trans); 1591c9e9f97bSIlya Dryomov goto out_free; 159298d5dc13STsutomu Itoh } 1593f46b5a66SChristoph Hellwig ret = btrfs_grow_device(trans, device, new_size); 1594f46b5a66SChristoph Hellwig btrfs_commit_transaction(trans, root); 1595ece7d20eSMike Fleetwood } else if (new_size < old_size) { 1596f46b5a66SChristoph Hellwig ret = btrfs_shrink_device(device, new_size); 15970253f40eSjeff.liu } /* equal, nothing need to do */ 1598f46b5a66SChristoph Hellwig 1599c9e9f97bSIlya Dryomov out_free: 1600f46b5a66SChristoph Hellwig kfree(vol_args); 1601c9e9f97bSIlya Dryomov out: 1602c9e9f97bSIlya Dryomov mutex_unlock(&root->fs_info->volume_mutex); 16035ac00addSStefan Behrens atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0); 160418f39c41SIlya Dryomov mnt_drop_write_file(file); 1605f46b5a66SChristoph Hellwig return ret; 1606f46b5a66SChristoph Hellwig } 1607f46b5a66SChristoph Hellwig 160872fd032eSSage Weil static noinline int btrfs_ioctl_snap_create_transid(struct file *file, 16096f72c7e2SArne Jansen char *name, unsigned long fd, int subvol, 16106f72c7e2SArne Jansen u64 *transid, bool readonly, 16118696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 1612f46b5a66SChristoph Hellwig { 1613f46b5a66SChristoph Hellwig int namelen; 16143de4586cSChris Mason int ret = 0; 1615f46b5a66SChristoph Hellwig 1616a874a63eSLiu Bo ret = mnt_want_write_file(file); 1617a874a63eSLiu Bo if (ret) 1618a874a63eSLiu Bo goto out; 1619a874a63eSLiu Bo 162072fd032eSSage Weil namelen = strlen(name); 162172fd032eSSage Weil if (strchr(name, '/')) { 1622f46b5a66SChristoph Hellwig ret = -EINVAL; 1623a874a63eSLiu Bo goto out_drop_write; 1624f46b5a66SChristoph Hellwig } 1625f46b5a66SChristoph Hellwig 162616780cabSChris Mason if (name[0] == '.' && 162716780cabSChris Mason (namelen == 1 || (name[1] == '.' && namelen == 2))) { 162816780cabSChris Mason ret = -EEXIST; 1629a874a63eSLiu Bo goto out_drop_write; 163016780cabSChris Mason } 163116780cabSChris Mason 16323de4586cSChris Mason if (subvol) { 163372fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 16346f72c7e2SArne Jansen NULL, transid, readonly, inherit); 1635cb8e7090SChristoph Hellwig } else { 16362903ff01SAl Viro struct fd src = fdget(fd); 16373de4586cSChris Mason struct inode *src_inode; 16382903ff01SAl Viro if (!src.file) { 16393de4586cSChris Mason ret = -EINVAL; 1640a874a63eSLiu Bo goto out_drop_write; 16413de4586cSChris Mason } 16423de4586cSChris Mason 1643496ad9aaSAl Viro src_inode = file_inode(src.file); 1644496ad9aaSAl Viro if (src_inode->i_sb != file_inode(file)->i_sb) { 1645efe120a0SFrank Holton btrfs_info(BTRFS_I(src_inode)->root->fs_info, 1646efe120a0SFrank Holton "Snapshot src from another FS"); 164723ad5b17SKusanagi Kouichi ret = -EXDEV; 1648d0242061SDavid Sterba } else if (!inode_owner_or_capable(src_inode)) { 1649d0242061SDavid Sterba /* 1650d0242061SDavid Sterba * Subvolume creation is not restricted, but snapshots 1651d0242061SDavid Sterba * are limited to own subvolumes only 1652d0242061SDavid Sterba */ 1653d0242061SDavid Sterba ret = -EPERM; 1654ecd18815SAl Viro } else { 165572fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 165672fd032eSSage Weil BTRFS_I(src_inode)->root, 16576f72c7e2SArne Jansen transid, readonly, inherit); 1658ecd18815SAl Viro } 16592903ff01SAl Viro fdput(src); 1660cb8e7090SChristoph Hellwig } 1661a874a63eSLiu Bo out_drop_write: 1662a874a63eSLiu Bo mnt_drop_write_file(file); 1663f46b5a66SChristoph Hellwig out: 166472fd032eSSage Weil return ret; 166572fd032eSSage Weil } 166672fd032eSSage Weil 166772fd032eSSage Weil static noinline int btrfs_ioctl_snap_create(struct file *file, 1668fa0d2b9bSLi Zefan void __user *arg, int subvol) 166972fd032eSSage Weil { 1670fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args *vol_args; 167172fd032eSSage Weil int ret; 167272fd032eSSage Weil 1673fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1674fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1675fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1676fa0d2b9bSLi Zefan vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1677fa0d2b9bSLi Zefan 1678fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 1679b83cc969SLi Zefan vol_args->fd, subvol, 16806f72c7e2SArne Jansen NULL, false, NULL); 1681fa0d2b9bSLi Zefan 1682fa0d2b9bSLi Zefan kfree(vol_args); 1683fa0d2b9bSLi Zefan return ret; 1684fa0d2b9bSLi Zefan } 1685fa0d2b9bSLi Zefan 1686fa0d2b9bSLi Zefan static noinline int btrfs_ioctl_snap_create_v2(struct file *file, 1687fa0d2b9bSLi Zefan void __user *arg, int subvol) 1688fa0d2b9bSLi Zefan { 1689fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args_v2 *vol_args; 1690fa0d2b9bSLi Zefan int ret; 1691fdfb1e4fSLi Zefan u64 transid = 0; 1692fdfb1e4fSLi Zefan u64 *ptr = NULL; 1693b83cc969SLi Zefan bool readonly = false; 16946f72c7e2SArne Jansen struct btrfs_qgroup_inherit *inherit = NULL; 169572fd032eSSage Weil 1696fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1697fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1698fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1699fa0d2b9bSLi Zefan vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 1700fdfb1e4fSLi Zefan 1701b83cc969SLi Zefan if (vol_args->flags & 17026f72c7e2SArne Jansen ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY | 17036f72c7e2SArne Jansen BTRFS_SUBVOL_QGROUP_INHERIT)) { 1704b83cc969SLi Zefan ret = -EOPNOTSUPP; 1705c47ca32dSDan Carpenter goto free_args; 1706fdfb1e4fSLi Zefan } 1707fdfb1e4fSLi Zefan 1708fa0d2b9bSLi Zefan if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC) 1709fdfb1e4fSLi Zefan ptr = &transid; 1710b83cc969SLi Zefan if (vol_args->flags & BTRFS_SUBVOL_RDONLY) 1711b83cc969SLi Zefan readonly = true; 17126f72c7e2SArne Jansen if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) { 17136f72c7e2SArne Jansen if (vol_args->size > PAGE_CACHE_SIZE) { 17146f72c7e2SArne Jansen ret = -EINVAL; 1715c47ca32dSDan Carpenter goto free_args; 17166f72c7e2SArne Jansen } 17176f72c7e2SArne Jansen inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size); 17186f72c7e2SArne Jansen if (IS_ERR(inherit)) { 17196f72c7e2SArne Jansen ret = PTR_ERR(inherit); 1720c47ca32dSDan Carpenter goto free_args; 17216f72c7e2SArne Jansen } 17226f72c7e2SArne Jansen } 172375eaa0e2SSage Weil 1724fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 17256f72c7e2SArne Jansen vol_args->fd, subvol, ptr, 17268696c533SMiao Xie readonly, inherit); 1727c47ca32dSDan Carpenter if (ret) 1728c47ca32dSDan Carpenter goto free_inherit; 172975eaa0e2SSage Weil 1730c47ca32dSDan Carpenter if (ptr && copy_to_user(arg + 1731fdfb1e4fSLi Zefan offsetof(struct btrfs_ioctl_vol_args_v2, 1732c47ca32dSDan Carpenter transid), 1733c47ca32dSDan Carpenter ptr, sizeof(*ptr))) 173475eaa0e2SSage Weil ret = -EFAULT; 1735c47ca32dSDan Carpenter 1736c47ca32dSDan Carpenter free_inherit: 17376f72c7e2SArne Jansen kfree(inherit); 1738c47ca32dSDan Carpenter free_args: 1739c47ca32dSDan Carpenter kfree(vol_args); 1740f46b5a66SChristoph Hellwig return ret; 1741f46b5a66SChristoph Hellwig } 1742f46b5a66SChristoph Hellwig 17430caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_getflags(struct file *file, 17440caa102dSLi Zefan void __user *arg) 17450caa102dSLi Zefan { 1746496ad9aaSAl Viro struct inode *inode = file_inode(file); 17470caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 17480caa102dSLi Zefan int ret = 0; 17490caa102dSLi Zefan u64 flags = 0; 17500caa102dSLi Zefan 175133345d01SLi Zefan if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) 17520caa102dSLi Zefan return -EINVAL; 17530caa102dSLi Zefan 17540caa102dSLi Zefan down_read(&root->fs_info->subvol_sem); 17550caa102dSLi Zefan if (btrfs_root_readonly(root)) 17560caa102dSLi Zefan flags |= BTRFS_SUBVOL_RDONLY; 17570caa102dSLi Zefan up_read(&root->fs_info->subvol_sem); 17580caa102dSLi Zefan 17590caa102dSLi Zefan if (copy_to_user(arg, &flags, sizeof(flags))) 17600caa102dSLi Zefan ret = -EFAULT; 17610caa102dSLi Zefan 17620caa102dSLi Zefan return ret; 17630caa102dSLi Zefan } 17640caa102dSLi Zefan 17650caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_setflags(struct file *file, 17660caa102dSLi Zefan void __user *arg) 17670caa102dSLi Zefan { 1768496ad9aaSAl Viro struct inode *inode = file_inode(file); 17690caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 17700caa102dSLi Zefan struct btrfs_trans_handle *trans; 17710caa102dSLi Zefan u64 root_flags; 17720caa102dSLi Zefan u64 flags; 17730caa102dSLi Zefan int ret = 0; 17740caa102dSLi Zefan 1775bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 1776bd60ea0fSDavid Sterba return -EPERM; 1777bd60ea0fSDavid Sterba 1778b9ca0664SLiu Bo ret = mnt_want_write_file(file); 1779b9ca0664SLiu Bo if (ret) 1780b9ca0664SLiu Bo goto out; 17810caa102dSLi Zefan 1782b9ca0664SLiu Bo if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { 1783b9ca0664SLiu Bo ret = -EINVAL; 1784b9ca0664SLiu Bo goto out_drop_write; 1785b9ca0664SLiu Bo } 17860caa102dSLi Zefan 1787b9ca0664SLiu Bo if (copy_from_user(&flags, arg, sizeof(flags))) { 1788b9ca0664SLiu Bo ret = -EFAULT; 1789b9ca0664SLiu Bo goto out_drop_write; 1790b9ca0664SLiu Bo } 17910caa102dSLi Zefan 1792b9ca0664SLiu Bo if (flags & BTRFS_SUBVOL_CREATE_ASYNC) { 1793b9ca0664SLiu Bo ret = -EINVAL; 1794b9ca0664SLiu Bo goto out_drop_write; 1795b9ca0664SLiu Bo } 17960caa102dSLi Zefan 1797b9ca0664SLiu Bo if (flags & ~BTRFS_SUBVOL_RDONLY) { 1798b9ca0664SLiu Bo ret = -EOPNOTSUPP; 1799b9ca0664SLiu Bo goto out_drop_write; 1800b9ca0664SLiu Bo } 18010caa102dSLi Zefan 18020caa102dSLi Zefan down_write(&root->fs_info->subvol_sem); 18030caa102dSLi Zefan 18040caa102dSLi Zefan /* nothing to do */ 18050caa102dSLi Zefan if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) 1806b9ca0664SLiu Bo goto out_drop_sem; 18070caa102dSLi Zefan 18080caa102dSLi Zefan root_flags = btrfs_root_flags(&root->root_item); 18092c686537SDavid Sterba if (flags & BTRFS_SUBVOL_RDONLY) { 18100caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 18110caa102dSLi Zefan root_flags | BTRFS_ROOT_SUBVOL_RDONLY); 18122c686537SDavid Sterba } else { 18132c686537SDavid Sterba /* 18142c686537SDavid Sterba * Block RO -> RW transition if this subvolume is involved in 18152c686537SDavid Sterba * send 18162c686537SDavid Sterba */ 18172c686537SDavid Sterba spin_lock(&root->root_item_lock); 18182c686537SDavid Sterba if (root->send_in_progress == 0) { 18190caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 18200caa102dSLi Zefan root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); 18212c686537SDavid Sterba spin_unlock(&root->root_item_lock); 18222c686537SDavid Sterba } else { 18232c686537SDavid Sterba spin_unlock(&root->root_item_lock); 18242c686537SDavid Sterba btrfs_warn(root->fs_info, 18252c686537SDavid Sterba "Attempt to set subvolume %llu read-write during send", 18262c686537SDavid Sterba root->root_key.objectid); 18272c686537SDavid Sterba ret = -EPERM; 18282c686537SDavid Sterba goto out_drop_sem; 18292c686537SDavid Sterba } 18302c686537SDavid Sterba } 18310caa102dSLi Zefan 18320caa102dSLi Zefan trans = btrfs_start_transaction(root, 1); 18330caa102dSLi Zefan if (IS_ERR(trans)) { 18340caa102dSLi Zefan ret = PTR_ERR(trans); 18350caa102dSLi Zefan goto out_reset; 18360caa102dSLi Zefan } 18370caa102dSLi Zefan 1838b4dc2b8cSLi Zefan ret = btrfs_update_root(trans, root->fs_info->tree_root, 18390caa102dSLi Zefan &root->root_key, &root->root_item); 18400caa102dSLi Zefan 18410caa102dSLi Zefan btrfs_commit_transaction(trans, root); 18420caa102dSLi Zefan out_reset: 18430caa102dSLi Zefan if (ret) 18440caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, root_flags); 1845b9ca0664SLiu Bo out_drop_sem: 18460caa102dSLi Zefan up_write(&root->fs_info->subvol_sem); 1847b9ca0664SLiu Bo out_drop_write: 1848b9ca0664SLiu Bo mnt_drop_write_file(file); 1849b9ca0664SLiu Bo out: 18500caa102dSLi Zefan return ret; 18510caa102dSLi Zefan } 18520caa102dSLi Zefan 185376dda93cSYan, Zheng /* 185476dda93cSYan, Zheng * helper to check if the subvolume references other subvolumes 185576dda93cSYan, Zheng */ 185676dda93cSYan, Zheng static noinline int may_destroy_subvol(struct btrfs_root *root) 185776dda93cSYan, Zheng { 185876dda93cSYan, Zheng struct btrfs_path *path; 1859175a2b87SJosef Bacik struct btrfs_dir_item *di; 186076dda93cSYan, Zheng struct btrfs_key key; 1861175a2b87SJosef Bacik u64 dir_id; 186276dda93cSYan, Zheng int ret; 186376dda93cSYan, Zheng 186476dda93cSYan, Zheng path = btrfs_alloc_path(); 186576dda93cSYan, Zheng if (!path) 186676dda93cSYan, Zheng return -ENOMEM; 186776dda93cSYan, Zheng 1868175a2b87SJosef Bacik /* Make sure this root isn't set as the default subvol */ 1869175a2b87SJosef Bacik dir_id = btrfs_super_root_dir(root->fs_info->super_copy); 1870175a2b87SJosef Bacik di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path, 1871175a2b87SJosef Bacik dir_id, "default", 7, 0); 1872175a2b87SJosef Bacik if (di && !IS_ERR(di)) { 1873175a2b87SJosef Bacik btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key); 1874175a2b87SJosef Bacik if (key.objectid == root->root_key.objectid) { 187572de6b53SGuangyu Sun ret = -EPERM; 187672de6b53SGuangyu Sun btrfs_err(root->fs_info, "deleting default subvolume " 187772de6b53SGuangyu Sun "%llu is not allowed", key.objectid); 1878175a2b87SJosef Bacik goto out; 1879175a2b87SJosef Bacik } 1880175a2b87SJosef Bacik btrfs_release_path(path); 1881175a2b87SJosef Bacik } 1882175a2b87SJosef Bacik 188376dda93cSYan, Zheng key.objectid = root->root_key.objectid; 188476dda93cSYan, Zheng key.type = BTRFS_ROOT_REF_KEY; 188576dda93cSYan, Zheng key.offset = (u64)-1; 188676dda93cSYan, Zheng 188776dda93cSYan, Zheng ret = btrfs_search_slot(NULL, root->fs_info->tree_root, 188876dda93cSYan, Zheng &key, path, 0, 0); 188976dda93cSYan, Zheng if (ret < 0) 189076dda93cSYan, Zheng goto out; 189176dda93cSYan, Zheng BUG_ON(ret == 0); 189276dda93cSYan, Zheng 189376dda93cSYan, Zheng ret = 0; 189476dda93cSYan, Zheng if (path->slots[0] > 0) { 189576dda93cSYan, Zheng path->slots[0]--; 189676dda93cSYan, Zheng btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 189776dda93cSYan, Zheng if (key.objectid == root->root_key.objectid && 189876dda93cSYan, Zheng key.type == BTRFS_ROOT_REF_KEY) 189976dda93cSYan, Zheng ret = -ENOTEMPTY; 190076dda93cSYan, Zheng } 190176dda93cSYan, Zheng out: 190276dda93cSYan, Zheng btrfs_free_path(path); 190376dda93cSYan, Zheng return ret; 190476dda93cSYan, Zheng } 190576dda93cSYan, Zheng 1906ac8e9819SChris Mason static noinline int key_in_sk(struct btrfs_key *key, 1907ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk) 1908ac8e9819SChris Mason { 1909abc6e134SChris Mason struct btrfs_key test; 1910abc6e134SChris Mason int ret; 1911abc6e134SChris Mason 1912abc6e134SChris Mason test.objectid = sk->min_objectid; 1913abc6e134SChris Mason test.type = sk->min_type; 1914abc6e134SChris Mason test.offset = sk->min_offset; 1915abc6e134SChris Mason 1916abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1917abc6e134SChris Mason if (ret < 0) 1918ac8e9819SChris Mason return 0; 1919abc6e134SChris Mason 1920abc6e134SChris Mason test.objectid = sk->max_objectid; 1921abc6e134SChris Mason test.type = sk->max_type; 1922abc6e134SChris Mason test.offset = sk->max_offset; 1923abc6e134SChris Mason 1924abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1925abc6e134SChris Mason if (ret > 0) 1926ac8e9819SChris Mason return 0; 1927ac8e9819SChris Mason return 1; 1928ac8e9819SChris Mason } 1929ac8e9819SChris Mason 1930ac8e9819SChris Mason static noinline int copy_to_sk(struct btrfs_root *root, 1931ac8e9819SChris Mason struct btrfs_path *path, 1932ac8e9819SChris Mason struct btrfs_key *key, 1933ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk, 19349b6e817dSGerhard Heift size_t *buf_size, 1935ba346b35SGerhard Heift char __user *ubuf, 1936ac8e9819SChris Mason unsigned long *sk_offset, 1937ac8e9819SChris Mason int *num_found) 1938ac8e9819SChris Mason { 1939ac8e9819SChris Mason u64 found_transid; 1940ac8e9819SChris Mason struct extent_buffer *leaf; 1941ac8e9819SChris Mason struct btrfs_ioctl_search_header sh; 1942dd81d459SNaohiro Aota struct btrfs_key test; 1943ac8e9819SChris Mason unsigned long item_off; 1944ac8e9819SChris Mason unsigned long item_len; 1945ac8e9819SChris Mason int nritems; 1946ac8e9819SChris Mason int i; 1947ac8e9819SChris Mason int slot; 1948ac8e9819SChris Mason int ret = 0; 1949ac8e9819SChris Mason 1950ac8e9819SChris Mason leaf = path->nodes[0]; 1951ac8e9819SChris Mason slot = path->slots[0]; 1952ac8e9819SChris Mason nritems = btrfs_header_nritems(leaf); 1953ac8e9819SChris Mason 1954ac8e9819SChris Mason if (btrfs_header_generation(leaf) > sk->max_transid) { 1955ac8e9819SChris Mason i = nritems; 1956ac8e9819SChris Mason goto advance_key; 1957ac8e9819SChris Mason } 1958ac8e9819SChris Mason found_transid = btrfs_header_generation(leaf); 1959ac8e9819SChris Mason 1960ac8e9819SChris Mason for (i = slot; i < nritems; i++) { 1961ac8e9819SChris Mason item_off = btrfs_item_ptr_offset(leaf, i); 1962ac8e9819SChris Mason item_len = btrfs_item_size_nr(leaf, i); 1963ac8e9819SChris Mason 196403b71c6cSGabriel de Perthuis btrfs_item_key_to_cpu(leaf, key, i); 196503b71c6cSGabriel de Perthuis if (!key_in_sk(key, sk)) 196603b71c6cSGabriel de Perthuis continue; 196703b71c6cSGabriel de Perthuis 19689b6e817dSGerhard Heift if (sizeof(sh) + item_len > *buf_size) { 19698f5f6178SGerhard Heift if (*num_found) { 1970ac8e9819SChris Mason ret = 1; 19718f5f6178SGerhard Heift goto out; 19728f5f6178SGerhard Heift } 19738f5f6178SGerhard Heift 19748f5f6178SGerhard Heift /* 19758f5f6178SGerhard Heift * return one empty item back for v1, which does not 19768f5f6178SGerhard Heift * handle -EOVERFLOW 19778f5f6178SGerhard Heift */ 19788f5f6178SGerhard Heift 19799b6e817dSGerhard Heift *buf_size = sizeof(sh) + item_len; 1980ac8e9819SChris Mason item_len = 0; 19818f5f6178SGerhard Heift ret = -EOVERFLOW; 19828f5f6178SGerhard Heift } 1983ac8e9819SChris Mason 19849b6e817dSGerhard Heift if (sizeof(sh) + item_len + *sk_offset > *buf_size) { 1985ac8e9819SChris Mason ret = 1; 198625c9bc2eSGerhard Heift goto out; 1987ac8e9819SChris Mason } 1988ac8e9819SChris Mason 1989ac8e9819SChris Mason sh.objectid = key->objectid; 1990ac8e9819SChris Mason sh.offset = key->offset; 1991ac8e9819SChris Mason sh.type = key->type; 1992ac8e9819SChris Mason sh.len = item_len; 1993ac8e9819SChris Mason sh.transid = found_transid; 1994ac8e9819SChris Mason 1995ac8e9819SChris Mason /* copy search result header */ 1996ba346b35SGerhard Heift if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) { 1997ba346b35SGerhard Heift ret = -EFAULT; 1998ba346b35SGerhard Heift goto out; 1999ba346b35SGerhard Heift } 2000ba346b35SGerhard Heift 2001ac8e9819SChris Mason *sk_offset += sizeof(sh); 2002ac8e9819SChris Mason 2003ac8e9819SChris Mason if (item_len) { 2004ba346b35SGerhard Heift char __user *up = ubuf + *sk_offset; 2005ac8e9819SChris Mason /* copy the item */ 2006ba346b35SGerhard Heift if (read_extent_buffer_to_user(leaf, up, 2007ba346b35SGerhard Heift item_off, item_len)) { 2008ba346b35SGerhard Heift ret = -EFAULT; 2009ba346b35SGerhard Heift goto out; 2010ba346b35SGerhard Heift } 2011ba346b35SGerhard Heift 2012ac8e9819SChris Mason *sk_offset += item_len; 2013ac8e9819SChris Mason } 2014e2156867SHugo Mills (*num_found)++; 2015ac8e9819SChris Mason 20168f5f6178SGerhard Heift if (ret) /* -EOVERFLOW from above */ 20178f5f6178SGerhard Heift goto out; 20188f5f6178SGerhard Heift 201925c9bc2eSGerhard Heift if (*num_found >= sk->nr_items) { 202025c9bc2eSGerhard Heift ret = 1; 202125c9bc2eSGerhard Heift goto out; 202225c9bc2eSGerhard Heift } 2023ac8e9819SChris Mason } 2024ac8e9819SChris Mason advance_key: 2025ac8e9819SChris Mason ret = 0; 2026dd81d459SNaohiro Aota test.objectid = sk->max_objectid; 2027dd81d459SNaohiro Aota test.type = sk->max_type; 2028dd81d459SNaohiro Aota test.offset = sk->max_offset; 2029dd81d459SNaohiro Aota if (btrfs_comp_cpu_keys(key, &test) >= 0) 2030dd81d459SNaohiro Aota ret = 1; 2031dd81d459SNaohiro Aota else if (key->offset < (u64)-1) 2032abc6e134SChris Mason key->offset++; 2033dd81d459SNaohiro Aota else if (key->type < (u8)-1) { 2034abc6e134SChris Mason key->offset = 0; 2035abc6e134SChris Mason key->type++; 2036dd81d459SNaohiro Aota } else if (key->objectid < (u64)-1) { 2037abc6e134SChris Mason key->offset = 0; 2038abc6e134SChris Mason key->type = 0; 2039abc6e134SChris Mason key->objectid++; 2040abc6e134SChris Mason } else 2041abc6e134SChris Mason ret = 1; 204225c9bc2eSGerhard Heift out: 2043ba346b35SGerhard Heift /* 2044ba346b35SGerhard Heift * 0: all items from this leaf copied, continue with next 2045ba346b35SGerhard Heift * 1: * more items can be copied, but unused buffer is too small 2046ba346b35SGerhard Heift * * all items were found 2047ba346b35SGerhard Heift * Either way, it will stops the loop which iterates to the next 2048ba346b35SGerhard Heift * leaf 2049ba346b35SGerhard Heift * -EOVERFLOW: item was to large for buffer 2050ba346b35SGerhard Heift * -EFAULT: could not copy extent buffer back to userspace 2051ba346b35SGerhard Heift */ 2052ac8e9819SChris Mason return ret; 2053ac8e9819SChris Mason } 2054ac8e9819SChris Mason 2055ac8e9819SChris Mason static noinline int search_ioctl(struct inode *inode, 205612544442SGerhard Heift struct btrfs_ioctl_search_key *sk, 20579b6e817dSGerhard Heift size_t *buf_size, 2058ba346b35SGerhard Heift char __user *ubuf) 2059ac8e9819SChris Mason { 2060ac8e9819SChris Mason struct btrfs_root *root; 2061ac8e9819SChris Mason struct btrfs_key key; 2062ac8e9819SChris Mason struct btrfs_path *path; 2063ac8e9819SChris Mason struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info; 2064ac8e9819SChris Mason int ret; 2065ac8e9819SChris Mason int num_found = 0; 2066ac8e9819SChris Mason unsigned long sk_offset = 0; 2067ac8e9819SChris Mason 20689b6e817dSGerhard Heift if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) { 20699b6e817dSGerhard Heift *buf_size = sizeof(struct btrfs_ioctl_search_header); 207012544442SGerhard Heift return -EOVERFLOW; 20719b6e817dSGerhard Heift } 207212544442SGerhard Heift 2073ac8e9819SChris Mason path = btrfs_alloc_path(); 2074ac8e9819SChris Mason if (!path) 2075ac8e9819SChris Mason return -ENOMEM; 2076ac8e9819SChris Mason 2077ac8e9819SChris Mason if (sk->tree_id == 0) { 2078ac8e9819SChris Mason /* search the root of the inode that was passed */ 2079ac8e9819SChris Mason root = BTRFS_I(inode)->root; 2080ac8e9819SChris Mason } else { 2081ac8e9819SChris Mason key.objectid = sk->tree_id; 2082ac8e9819SChris Mason key.type = BTRFS_ROOT_ITEM_KEY; 2083ac8e9819SChris Mason key.offset = (u64)-1; 2084ac8e9819SChris Mason root = btrfs_read_fs_root_no_name(info, &key); 2085ac8e9819SChris Mason if (IS_ERR(root)) { 2086f14d104dSDavid Sterba btrfs_err(info, "could not find root %llu", 2087ac8e9819SChris Mason sk->tree_id); 2088ac8e9819SChris Mason btrfs_free_path(path); 2089ac8e9819SChris Mason return -ENOENT; 2090ac8e9819SChris Mason } 2091ac8e9819SChris Mason } 2092ac8e9819SChris Mason 2093ac8e9819SChris Mason key.objectid = sk->min_objectid; 2094ac8e9819SChris Mason key.type = sk->min_type; 2095ac8e9819SChris Mason key.offset = sk->min_offset; 2096ac8e9819SChris Mason 2097ac8e9819SChris Mason while (1) { 20986174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &key, path, sk->min_transid); 2099ac8e9819SChris Mason if (ret != 0) { 2100ac8e9819SChris Mason if (ret > 0) 2101ac8e9819SChris Mason ret = 0; 2102ac8e9819SChris Mason goto err; 2103ac8e9819SChris Mason } 2104ba346b35SGerhard Heift ret = copy_to_sk(root, path, &key, sk, buf_size, ubuf, 2105ac8e9819SChris Mason &sk_offset, &num_found); 2106b3b4aa74SDavid Sterba btrfs_release_path(path); 210725c9bc2eSGerhard Heift if (ret) 2108ac8e9819SChris Mason break; 2109ac8e9819SChris Mason 2110ac8e9819SChris Mason } 21118f5f6178SGerhard Heift if (ret > 0) 2112ac8e9819SChris Mason ret = 0; 2113ac8e9819SChris Mason err: 2114ac8e9819SChris Mason sk->nr_items = num_found; 2115ac8e9819SChris Mason btrfs_free_path(path); 2116ac8e9819SChris Mason return ret; 2117ac8e9819SChris Mason } 2118ac8e9819SChris Mason 2119ac8e9819SChris Mason static noinline int btrfs_ioctl_tree_search(struct file *file, 2120ac8e9819SChris Mason void __user *argp) 2121ac8e9819SChris Mason { 2122ba346b35SGerhard Heift struct btrfs_ioctl_search_args __user *uargs; 2123ba346b35SGerhard Heift struct btrfs_ioctl_search_key sk; 2124ac8e9819SChris Mason struct inode *inode; 2125ac8e9819SChris Mason int ret; 21269b6e817dSGerhard Heift size_t buf_size; 2127ac8e9819SChris Mason 2128ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 2129ac8e9819SChris Mason return -EPERM; 2130ac8e9819SChris Mason 2131ba346b35SGerhard Heift uargs = (struct btrfs_ioctl_search_args __user *)argp; 2132ac8e9819SChris Mason 2133ba346b35SGerhard Heift if (copy_from_user(&sk, &uargs->key, sizeof(sk))) 2134ba346b35SGerhard Heift return -EFAULT; 2135ba346b35SGerhard Heift 2136ba346b35SGerhard Heift buf_size = sizeof(uargs->buf); 2137ac8e9819SChris Mason 2138496ad9aaSAl Viro inode = file_inode(file); 2139ba346b35SGerhard Heift ret = search_ioctl(inode, &sk, &buf_size, uargs->buf); 21408f5f6178SGerhard Heift 21418f5f6178SGerhard Heift /* 21428f5f6178SGerhard Heift * In the origin implementation an overflow is handled by returning a 21438f5f6178SGerhard Heift * search header with a len of zero, so reset ret. 21448f5f6178SGerhard Heift */ 21458f5f6178SGerhard Heift if (ret == -EOVERFLOW) 21468f5f6178SGerhard Heift ret = 0; 21478f5f6178SGerhard Heift 2148ba346b35SGerhard Heift if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk))) 2149ac8e9819SChris Mason ret = -EFAULT; 2150ac8e9819SChris Mason return ret; 2151ac8e9819SChris Mason } 2152ac8e9819SChris Mason 2153cc68a8a5SGerhard Heift static noinline int btrfs_ioctl_tree_search_v2(struct file *file, 2154cc68a8a5SGerhard Heift void __user *argp) 2155cc68a8a5SGerhard Heift { 2156cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 __user *uarg; 2157cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 args; 2158cc68a8a5SGerhard Heift struct inode *inode; 2159cc68a8a5SGerhard Heift int ret; 2160cc68a8a5SGerhard Heift size_t buf_size; 2161cc68a8a5SGerhard Heift const size_t buf_limit = 16 * 1024 * 1024; 2162cc68a8a5SGerhard Heift 2163cc68a8a5SGerhard Heift if (!capable(CAP_SYS_ADMIN)) 2164cc68a8a5SGerhard Heift return -EPERM; 2165cc68a8a5SGerhard Heift 2166cc68a8a5SGerhard Heift /* copy search header and buffer size */ 2167cc68a8a5SGerhard Heift uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp; 2168cc68a8a5SGerhard Heift if (copy_from_user(&args, uarg, sizeof(args))) 2169cc68a8a5SGerhard Heift return -EFAULT; 2170cc68a8a5SGerhard Heift 2171cc68a8a5SGerhard Heift buf_size = args.buf_size; 2172cc68a8a5SGerhard Heift 2173cc68a8a5SGerhard Heift if (buf_size < sizeof(struct btrfs_ioctl_search_header)) 2174cc68a8a5SGerhard Heift return -EOVERFLOW; 2175cc68a8a5SGerhard Heift 2176cc68a8a5SGerhard Heift /* limit result size to 16MB */ 2177cc68a8a5SGerhard Heift if (buf_size > buf_limit) 2178cc68a8a5SGerhard Heift buf_size = buf_limit; 2179cc68a8a5SGerhard Heift 2180cc68a8a5SGerhard Heift inode = file_inode(file); 2181cc68a8a5SGerhard Heift ret = search_ioctl(inode, &args.key, &buf_size, 2182cc68a8a5SGerhard Heift (char *)(&uarg->buf[0])); 2183cc68a8a5SGerhard Heift if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key))) 2184cc68a8a5SGerhard Heift ret = -EFAULT; 2185cc68a8a5SGerhard Heift else if (ret == -EOVERFLOW && 2186cc68a8a5SGerhard Heift copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size))) 2187cc68a8a5SGerhard Heift ret = -EFAULT; 2188cc68a8a5SGerhard Heift 218976dda93cSYan, Zheng return ret; 219076dda93cSYan, Zheng } 219176dda93cSYan, Zheng 219298d377a0STARUISI Hiroaki /* 2193ac8e9819SChris Mason * Search INODE_REFs to identify path name of 'dirid' directory 2194ac8e9819SChris Mason * in a 'tree_id' tree. and sets path name to 'name'. 219598d377a0STARUISI Hiroaki */ 219698d377a0STARUISI Hiroaki static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, 219798d377a0STARUISI Hiroaki u64 tree_id, u64 dirid, char *name) 219898d377a0STARUISI Hiroaki { 219998d377a0STARUISI Hiroaki struct btrfs_root *root; 220098d377a0STARUISI Hiroaki struct btrfs_key key; 2201ac8e9819SChris Mason char *ptr; 220298d377a0STARUISI Hiroaki int ret = -1; 220398d377a0STARUISI Hiroaki int slot; 220498d377a0STARUISI Hiroaki int len; 220598d377a0STARUISI Hiroaki int total_len = 0; 220698d377a0STARUISI Hiroaki struct btrfs_inode_ref *iref; 220798d377a0STARUISI Hiroaki struct extent_buffer *l; 220898d377a0STARUISI Hiroaki struct btrfs_path *path; 220998d377a0STARUISI Hiroaki 221098d377a0STARUISI Hiroaki if (dirid == BTRFS_FIRST_FREE_OBJECTID) { 221198d377a0STARUISI Hiroaki name[0]='\0'; 221298d377a0STARUISI Hiroaki return 0; 221398d377a0STARUISI Hiroaki } 221498d377a0STARUISI Hiroaki 221598d377a0STARUISI Hiroaki path = btrfs_alloc_path(); 221698d377a0STARUISI Hiroaki if (!path) 221798d377a0STARUISI Hiroaki return -ENOMEM; 221898d377a0STARUISI Hiroaki 2219ac8e9819SChris Mason ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX]; 222098d377a0STARUISI Hiroaki 222198d377a0STARUISI Hiroaki key.objectid = tree_id; 222298d377a0STARUISI Hiroaki key.type = BTRFS_ROOT_ITEM_KEY; 222398d377a0STARUISI Hiroaki key.offset = (u64)-1; 222498d377a0STARUISI Hiroaki root = btrfs_read_fs_root_no_name(info, &key); 222598d377a0STARUISI Hiroaki if (IS_ERR(root)) { 2226f14d104dSDavid Sterba btrfs_err(info, "could not find root %llu", tree_id); 22278ad6fcabSChris Mason ret = -ENOENT; 22288ad6fcabSChris Mason goto out; 222998d377a0STARUISI Hiroaki } 223098d377a0STARUISI Hiroaki 223198d377a0STARUISI Hiroaki key.objectid = dirid; 223298d377a0STARUISI Hiroaki key.type = BTRFS_INODE_REF_KEY; 22338ad6fcabSChris Mason key.offset = (u64)-1; 223498d377a0STARUISI Hiroaki 223598d377a0STARUISI Hiroaki while (1) { 223698d377a0STARUISI Hiroaki ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 223798d377a0STARUISI Hiroaki if (ret < 0) 223898d377a0STARUISI Hiroaki goto out; 223918674c6cSFilipe David Borba Manana else if (ret > 0) { 224018674c6cSFilipe David Borba Manana ret = btrfs_previous_item(root, path, dirid, 224118674c6cSFilipe David Borba Manana BTRFS_INODE_REF_KEY); 224218674c6cSFilipe David Borba Manana if (ret < 0) 224318674c6cSFilipe David Borba Manana goto out; 224418674c6cSFilipe David Borba Manana else if (ret > 0) { 2245ac8e9819SChris Mason ret = -ENOENT; 224698d377a0STARUISI Hiroaki goto out; 2247ac8e9819SChris Mason } 224818674c6cSFilipe David Borba Manana } 224918674c6cSFilipe David Borba Manana 225018674c6cSFilipe David Borba Manana l = path->nodes[0]; 225118674c6cSFilipe David Borba Manana slot = path->slots[0]; 225218674c6cSFilipe David Borba Manana btrfs_item_key_to_cpu(l, &key, slot); 225398d377a0STARUISI Hiroaki 225498d377a0STARUISI Hiroaki iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); 225598d377a0STARUISI Hiroaki len = btrfs_inode_ref_name_len(l, iref); 225698d377a0STARUISI Hiroaki ptr -= len + 1; 225798d377a0STARUISI Hiroaki total_len += len + 1; 2258a696cf35SFilipe David Borba Manana if (ptr < name) { 2259a696cf35SFilipe David Borba Manana ret = -ENAMETOOLONG; 226098d377a0STARUISI Hiroaki goto out; 2261a696cf35SFilipe David Borba Manana } 226298d377a0STARUISI Hiroaki 226398d377a0STARUISI Hiroaki *(ptr + len) = '/'; 226498d377a0STARUISI Hiroaki read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len); 226598d377a0STARUISI Hiroaki 226698d377a0STARUISI Hiroaki if (key.offset == BTRFS_FIRST_FREE_OBJECTID) 226798d377a0STARUISI Hiroaki break; 226898d377a0STARUISI Hiroaki 2269b3b4aa74SDavid Sterba btrfs_release_path(path); 227098d377a0STARUISI Hiroaki key.objectid = key.offset; 22718ad6fcabSChris Mason key.offset = (u64)-1; 227298d377a0STARUISI Hiroaki dirid = key.objectid; 227398d377a0STARUISI Hiroaki } 227477906a50SLi Zefan memmove(name, ptr, total_len); 227598d377a0STARUISI Hiroaki name[total_len] = '\0'; 227698d377a0STARUISI Hiroaki ret = 0; 227798d377a0STARUISI Hiroaki out: 227898d377a0STARUISI Hiroaki btrfs_free_path(path); 2279ac8e9819SChris Mason return ret; 2280ac8e9819SChris Mason } 2281ac8e9819SChris Mason 2282ac8e9819SChris Mason static noinline int btrfs_ioctl_ino_lookup(struct file *file, 2283ac8e9819SChris Mason void __user *argp) 2284ac8e9819SChris Mason { 2285ac8e9819SChris Mason struct btrfs_ioctl_ino_lookup_args *args; 2286ac8e9819SChris Mason struct inode *inode; 228701b810b8SDavid Sterba int ret = 0; 2288ac8e9819SChris Mason 22892354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 22902354d08fSJulia Lawall if (IS_ERR(args)) 22912354d08fSJulia Lawall return PTR_ERR(args); 2292c2b96929SDan Carpenter 2293496ad9aaSAl Viro inode = file_inode(file); 2294ac8e9819SChris Mason 229501b810b8SDavid Sterba /* 229601b810b8SDavid Sterba * Unprivileged query to obtain the containing subvolume root id. The 229701b810b8SDavid Sterba * path is reset so it's consistent with btrfs_search_path_in_tree. 229801b810b8SDavid Sterba */ 22991b53ac4dSChris Mason if (args->treeid == 0) 23001b53ac4dSChris Mason args->treeid = BTRFS_I(inode)->root->root_key.objectid; 23011b53ac4dSChris Mason 230201b810b8SDavid Sterba if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) { 230301b810b8SDavid Sterba args->name[0] = 0; 230401b810b8SDavid Sterba goto out; 230501b810b8SDavid Sterba } 230601b810b8SDavid Sterba 230701b810b8SDavid Sterba if (!capable(CAP_SYS_ADMIN)) { 230801b810b8SDavid Sterba ret = -EPERM; 230901b810b8SDavid Sterba goto out; 231001b810b8SDavid Sterba } 231101b810b8SDavid Sterba 2312ac8e9819SChris Mason ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, 2313ac8e9819SChris Mason args->treeid, args->objectid, 2314ac8e9819SChris Mason args->name); 2315ac8e9819SChris Mason 231601b810b8SDavid Sterba out: 2317ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 2318ac8e9819SChris Mason ret = -EFAULT; 2319ac8e9819SChris Mason 2320ac8e9819SChris Mason kfree(args); 232198d377a0STARUISI Hiroaki return ret; 232298d377a0STARUISI Hiroaki } 232398d377a0STARUISI Hiroaki 232476dda93cSYan, Zheng static noinline int btrfs_ioctl_snap_destroy(struct file *file, 232576dda93cSYan, Zheng void __user *arg) 232676dda93cSYan, Zheng { 232754563d41SAl Viro struct dentry *parent = file->f_path.dentry; 232876dda93cSYan, Zheng struct dentry *dentry; 23292b0143b5SDavid Howells struct inode *dir = d_inode(parent); 233076dda93cSYan, Zheng struct inode *inode; 233176dda93cSYan, Zheng struct btrfs_root *root = BTRFS_I(dir)->root; 233276dda93cSYan, Zheng struct btrfs_root *dest = NULL; 233376dda93cSYan, Zheng struct btrfs_ioctl_vol_args *vol_args; 233476dda93cSYan, Zheng struct btrfs_trans_handle *trans; 2335c58aaad2SMiao Xie struct btrfs_block_rsv block_rsv; 2336521e0546SDavid Sterba u64 root_flags; 2337c58aaad2SMiao Xie u64 qgroup_reserved; 233876dda93cSYan, Zheng int namelen; 233976dda93cSYan, Zheng int ret; 234076dda93cSYan, Zheng int err = 0; 234176dda93cSYan, Zheng 234276dda93cSYan, Zheng vol_args = memdup_user(arg, sizeof(*vol_args)); 234376dda93cSYan, Zheng if (IS_ERR(vol_args)) 234476dda93cSYan, Zheng return PTR_ERR(vol_args); 234576dda93cSYan, Zheng 234676dda93cSYan, Zheng vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 234776dda93cSYan, Zheng namelen = strlen(vol_args->name); 234876dda93cSYan, Zheng if (strchr(vol_args->name, '/') || 234976dda93cSYan, Zheng strncmp(vol_args->name, "..", namelen) == 0) { 235076dda93cSYan, Zheng err = -EINVAL; 235176dda93cSYan, Zheng goto out; 235276dda93cSYan, Zheng } 235376dda93cSYan, Zheng 2354a561be71SAl Viro err = mnt_want_write_file(file); 235576dda93cSYan, Zheng if (err) 235676dda93cSYan, Zheng goto out; 235776dda93cSYan, Zheng 2358521e0546SDavid Sterba 23595c50c9b8SDavid Sterba err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT); 23605c50c9b8SDavid Sterba if (err == -EINTR) 2361e43f998eSDavid Sterba goto out_drop_write; 236276dda93cSYan, Zheng dentry = lookup_one_len(vol_args->name, parent, namelen); 236376dda93cSYan, Zheng if (IS_ERR(dentry)) { 236476dda93cSYan, Zheng err = PTR_ERR(dentry); 236576dda93cSYan, Zheng goto out_unlock_dir; 236676dda93cSYan, Zheng } 236776dda93cSYan, Zheng 23682b0143b5SDavid Howells if (d_really_is_negative(dentry)) { 236976dda93cSYan, Zheng err = -ENOENT; 237076dda93cSYan, Zheng goto out_dput; 237176dda93cSYan, Zheng } 237276dda93cSYan, Zheng 23732b0143b5SDavid Howells inode = d_inode(dentry); 23744260f7c7SSage Weil dest = BTRFS_I(inode)->root; 23754260f7c7SSage Weil if (!capable(CAP_SYS_ADMIN)) { 23764260f7c7SSage Weil /* 23774260f7c7SSage Weil * Regular user. Only allow this with a special mount 23784260f7c7SSage Weil * option, when the user has write+exec access to the 23794260f7c7SSage Weil * subvol root, and when rmdir(2) would have been 23804260f7c7SSage Weil * allowed. 23814260f7c7SSage Weil * 23824260f7c7SSage Weil * Note that this is _not_ check that the subvol is 23834260f7c7SSage Weil * empty or doesn't contain data that we wouldn't 23844260f7c7SSage Weil * otherwise be able to delete. 23854260f7c7SSage Weil * 23864260f7c7SSage Weil * Users who want to delete empty subvols should try 23874260f7c7SSage Weil * rmdir(2). 23884260f7c7SSage Weil */ 23894260f7c7SSage Weil err = -EPERM; 23904260f7c7SSage Weil if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED)) 23914260f7c7SSage Weil goto out_dput; 23924260f7c7SSage Weil 23934260f7c7SSage Weil /* 23944260f7c7SSage Weil * Do not allow deletion if the parent dir is the same 23954260f7c7SSage Weil * as the dir to be deleted. That means the ioctl 23964260f7c7SSage Weil * must be called on the dentry referencing the root 23974260f7c7SSage Weil * of the subvol, not a random directory contained 23984260f7c7SSage Weil * within it. 23994260f7c7SSage Weil */ 24004260f7c7SSage Weil err = -EINVAL; 24014260f7c7SSage Weil if (root == dest) 24024260f7c7SSage Weil goto out_dput; 24034260f7c7SSage Weil 24044260f7c7SSage Weil err = inode_permission(inode, MAY_WRITE | MAY_EXEC); 24054260f7c7SSage Weil if (err) 24064260f7c7SSage Weil goto out_dput; 24075c39da5bSMiao Xie } 24084260f7c7SSage Weil 24095c39da5bSMiao Xie /* check if subvolume may be deleted by a user */ 24104260f7c7SSage Weil err = btrfs_may_delete(dir, dentry, 1); 24114260f7c7SSage Weil if (err) 24124260f7c7SSage Weil goto out_dput; 24134260f7c7SSage Weil 241433345d01SLi Zefan if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { 241576dda93cSYan, Zheng err = -EINVAL; 241676dda93cSYan, Zheng goto out_dput; 241776dda93cSYan, Zheng } 241876dda93cSYan, Zheng 241976dda93cSYan, Zheng mutex_lock(&inode->i_mutex); 2420521e0546SDavid Sterba 2421521e0546SDavid Sterba /* 2422521e0546SDavid Sterba * Don't allow to delete a subvolume with send in progress. This is 2423521e0546SDavid Sterba * inside the i_mutex so the error handling that has to drop the bit 2424521e0546SDavid Sterba * again is not run concurrently. 2425521e0546SDavid Sterba */ 2426521e0546SDavid Sterba spin_lock(&dest->root_item_lock); 2427c55bfa67SFilipe Manana root_flags = btrfs_root_flags(&dest->root_item); 2428c55bfa67SFilipe Manana if (dest->send_in_progress == 0) { 2429c55bfa67SFilipe Manana btrfs_set_root_flags(&dest->root_item, 2430521e0546SDavid Sterba root_flags | BTRFS_ROOT_SUBVOL_DEAD); 2431521e0546SDavid Sterba spin_unlock(&dest->root_item_lock); 2432521e0546SDavid Sterba } else { 2433521e0546SDavid Sterba spin_unlock(&dest->root_item_lock); 2434521e0546SDavid Sterba btrfs_warn(root->fs_info, 2435521e0546SDavid Sterba "Attempt to delete subvolume %llu during send", 2436c55bfa67SFilipe Manana dest->root_key.objectid); 2437521e0546SDavid Sterba err = -EPERM; 2438909e26dcSOmar Sandoval goto out_unlock_inode; 2439521e0546SDavid Sterba } 2440521e0546SDavid Sterba 244176dda93cSYan, Zheng down_write(&root->fs_info->subvol_sem); 244276dda93cSYan, Zheng 244376dda93cSYan, Zheng err = may_destroy_subvol(dest); 244476dda93cSYan, Zheng if (err) 244576dda93cSYan, Zheng goto out_up_write; 244676dda93cSYan, Zheng 2447c58aaad2SMiao Xie btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); 2448c58aaad2SMiao Xie /* 2449c58aaad2SMiao Xie * One for dir inode, two for dir entries, two for root 2450c58aaad2SMiao Xie * ref/backref. 2451c58aaad2SMiao Xie */ 2452c58aaad2SMiao Xie err = btrfs_subvolume_reserve_metadata(root, &block_rsv, 2453ee3441b4SJeff Mahoney 5, &qgroup_reserved, true); 2454c58aaad2SMiao Xie if (err) 2455c58aaad2SMiao Xie goto out_up_write; 2456c58aaad2SMiao Xie 2457a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 2458a22285a6SYan, Zheng if (IS_ERR(trans)) { 2459a22285a6SYan, Zheng err = PTR_ERR(trans); 2460c58aaad2SMiao Xie goto out_release; 2461a22285a6SYan, Zheng } 2462c58aaad2SMiao Xie trans->block_rsv = &block_rsv; 2463c58aaad2SMiao Xie trans->bytes_reserved = block_rsv.size; 2464a22285a6SYan, Zheng 246576dda93cSYan, Zheng ret = btrfs_unlink_subvol(trans, root, dir, 246676dda93cSYan, Zheng dest->root_key.objectid, 246776dda93cSYan, Zheng dentry->d_name.name, 246876dda93cSYan, Zheng dentry->d_name.len); 246979787eaaSJeff Mahoney if (ret) { 247079787eaaSJeff Mahoney err = ret; 247179787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, ret); 247279787eaaSJeff Mahoney goto out_end_trans; 247379787eaaSJeff Mahoney } 247476dda93cSYan, Zheng 247576dda93cSYan, Zheng btrfs_record_root_in_trans(trans, dest); 247676dda93cSYan, Zheng 247776dda93cSYan, Zheng memset(&dest->root_item.drop_progress, 0, 247876dda93cSYan, Zheng sizeof(dest->root_item.drop_progress)); 247976dda93cSYan, Zheng dest->root_item.drop_level = 0; 248076dda93cSYan, Zheng btrfs_set_root_refs(&dest->root_item, 0); 248176dda93cSYan, Zheng 248227cdeb70SMiao Xie if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) { 248376dda93cSYan, Zheng ret = btrfs_insert_orphan_item(trans, 248476dda93cSYan, Zheng root->fs_info->tree_root, 248576dda93cSYan, Zheng dest->root_key.objectid); 248679787eaaSJeff Mahoney if (ret) { 248779787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, ret); 248879787eaaSJeff Mahoney err = ret; 248979787eaaSJeff Mahoney goto out_end_trans; 2490d68fc57bSYan, Zheng } 249179787eaaSJeff Mahoney } 2492dd5f9615SStefan Behrens 2493dd5f9615SStefan Behrens ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root, 2494dd5f9615SStefan Behrens dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL, 2495dd5f9615SStefan Behrens dest->root_key.objectid); 2496dd5f9615SStefan Behrens if (ret && ret != -ENOENT) { 2497dd5f9615SStefan Behrens btrfs_abort_transaction(trans, root, ret); 2498dd5f9615SStefan Behrens err = ret; 2499dd5f9615SStefan Behrens goto out_end_trans; 2500dd5f9615SStefan Behrens } 2501dd5f9615SStefan Behrens if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) { 2502dd5f9615SStefan Behrens ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root, 2503dd5f9615SStefan Behrens dest->root_item.received_uuid, 2504dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 2505dd5f9615SStefan Behrens dest->root_key.objectid); 2506dd5f9615SStefan Behrens if (ret && ret != -ENOENT) { 2507dd5f9615SStefan Behrens btrfs_abort_transaction(trans, root, ret); 2508dd5f9615SStefan Behrens err = ret; 2509dd5f9615SStefan Behrens goto out_end_trans; 2510dd5f9615SStefan Behrens } 2511dd5f9615SStefan Behrens } 2512dd5f9615SStefan Behrens 251379787eaaSJeff Mahoney out_end_trans: 2514c58aaad2SMiao Xie trans->block_rsv = NULL; 2515c58aaad2SMiao Xie trans->bytes_reserved = 0; 2516531cb13fSSage Weil ret = btrfs_end_transaction(trans, root); 251779787eaaSJeff Mahoney if (ret && !err) 251879787eaaSJeff Mahoney err = ret; 251976dda93cSYan, Zheng inode->i_flags |= S_DEAD; 2520c58aaad2SMiao Xie out_release: 2521c58aaad2SMiao Xie btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved); 252276dda93cSYan, Zheng out_up_write: 252376dda93cSYan, Zheng up_write(&root->fs_info->subvol_sem); 2524521e0546SDavid Sterba if (err) { 2525521e0546SDavid Sterba spin_lock(&dest->root_item_lock); 2526c55bfa67SFilipe Manana root_flags = btrfs_root_flags(&dest->root_item); 2527c55bfa67SFilipe Manana btrfs_set_root_flags(&dest->root_item, 2528521e0546SDavid Sterba root_flags & ~BTRFS_ROOT_SUBVOL_DEAD); 2529521e0546SDavid Sterba spin_unlock(&dest->root_item_lock); 2530521e0546SDavid Sterba } 2531909e26dcSOmar Sandoval out_unlock_inode: 253276dda93cSYan, Zheng mutex_unlock(&inode->i_mutex); 253376dda93cSYan, Zheng if (!err) { 253464ad6c48SOmar Sandoval d_invalidate(dentry); 253576dda93cSYan, Zheng btrfs_invalidate_inodes(dest); 253676dda93cSYan, Zheng d_delete(dentry); 253761155aa0SDavid Sterba ASSERT(dest->send_in_progress == 0); 2538fa6ac876SLiu Bo 2539fa6ac876SLiu Bo /* the last ref */ 254057cdc8dbSDavid Sterba if (dest->ino_cache_inode) { 254157cdc8dbSDavid Sterba iput(dest->ino_cache_inode); 254257cdc8dbSDavid Sterba dest->ino_cache_inode = NULL; 2543fa6ac876SLiu Bo } 254476dda93cSYan, Zheng } 254576dda93cSYan, Zheng out_dput: 254676dda93cSYan, Zheng dput(dentry); 254776dda93cSYan, Zheng out_unlock_dir: 254876dda93cSYan, Zheng mutex_unlock(&dir->i_mutex); 2549e43f998eSDavid Sterba out_drop_write: 25502a79f17eSAl Viro mnt_drop_write_file(file); 255176dda93cSYan, Zheng out: 255276dda93cSYan, Zheng kfree(vol_args); 255376dda93cSYan, Zheng return err; 255476dda93cSYan, Zheng } 255576dda93cSYan, Zheng 25561e701a32SChris Mason static int btrfs_ioctl_defrag(struct file *file, void __user *argp) 2557f46b5a66SChristoph Hellwig { 2558496ad9aaSAl Viro struct inode *inode = file_inode(file); 2559f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 25601e701a32SChris Mason struct btrfs_ioctl_defrag_range_args *range; 2561c146afadSYan Zheng int ret; 2562c146afadSYan Zheng 256325122d15SIlya Dryomov ret = mnt_want_write_file(file); 256425122d15SIlya Dryomov if (ret) 256525122d15SIlya Dryomov return ret; 2566b83cc969SLi Zefan 256725122d15SIlya Dryomov if (btrfs_root_readonly(root)) { 256825122d15SIlya Dryomov ret = -EROFS; 256925122d15SIlya Dryomov goto out; 25705ac00addSStefan Behrens } 2571f46b5a66SChristoph Hellwig 2572f46b5a66SChristoph Hellwig switch (inode->i_mode & S_IFMT) { 2573f46b5a66SChristoph Hellwig case S_IFDIR: 2574e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) { 2575e441d54dSChris Mason ret = -EPERM; 2576e441d54dSChris Mason goto out; 2577e441d54dSChris Mason } 2578de78b51aSEric Sandeen ret = btrfs_defrag_root(root); 25798929ecfaSYan, Zheng if (ret) 25808929ecfaSYan, Zheng goto out; 2581de78b51aSEric Sandeen ret = btrfs_defrag_root(root->fs_info->extent_root); 2582f46b5a66SChristoph Hellwig break; 2583f46b5a66SChristoph Hellwig case S_IFREG: 2584e441d54dSChris Mason if (!(file->f_mode & FMODE_WRITE)) { 2585e441d54dSChris Mason ret = -EINVAL; 2586e441d54dSChris Mason goto out; 2587e441d54dSChris Mason } 25881e701a32SChris Mason 25891e701a32SChris Mason range = kzalloc(sizeof(*range), GFP_KERNEL); 25901e701a32SChris Mason if (!range) { 25911e701a32SChris Mason ret = -ENOMEM; 25921e701a32SChris Mason goto out; 25931e701a32SChris Mason } 25941e701a32SChris Mason 25951e701a32SChris Mason if (argp) { 25961e701a32SChris Mason if (copy_from_user(range, argp, 25971e701a32SChris Mason sizeof(*range))) { 25981e701a32SChris Mason ret = -EFAULT; 25991e701a32SChris Mason kfree(range); 2600683be16eSDan Carpenter goto out; 26011e701a32SChris Mason } 26021e701a32SChris Mason /* compression requires us to start the IO */ 26031e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 26041e701a32SChris Mason range->flags |= BTRFS_DEFRAG_RANGE_START_IO; 26051e701a32SChris Mason range->extent_thresh = (u32)-1; 26061e701a32SChris Mason } 26071e701a32SChris Mason } else { 26081e701a32SChris Mason /* the rest are all set to zero by kzalloc */ 26091e701a32SChris Mason range->len = (u64)-1; 26101e701a32SChris Mason } 2611496ad9aaSAl Viro ret = btrfs_defrag_file(file_inode(file), file, 26124cb5300bSChris Mason range, 0, 0); 26134cb5300bSChris Mason if (ret > 0) 26144cb5300bSChris Mason ret = 0; 26151e701a32SChris Mason kfree(range); 2616f46b5a66SChristoph Hellwig break; 26178929ecfaSYan, Zheng default: 26188929ecfaSYan, Zheng ret = -EINVAL; 2619f46b5a66SChristoph Hellwig } 2620e441d54dSChris Mason out: 262125122d15SIlya Dryomov mnt_drop_write_file(file); 2622e441d54dSChris Mason return ret; 2623f46b5a66SChristoph Hellwig } 2624f46b5a66SChristoph Hellwig 2625b2950863SChristoph Hellwig static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg) 2626f46b5a66SChristoph Hellwig { 2627f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2628f46b5a66SChristoph Hellwig int ret; 2629f46b5a66SChristoph Hellwig 2630e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2631e441d54dSChris Mason return -EPERM; 2632e441d54dSChris Mason 26335ac00addSStefan Behrens if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running, 26345ac00addSStefan Behrens 1)) { 2635e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 2636c9e9f97bSIlya Dryomov } 2637c9e9f97bSIlya Dryomov 26385ac00addSStefan Behrens mutex_lock(&root->fs_info->volume_mutex); 2639dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2640c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 2641c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 2642c9e9f97bSIlya Dryomov goto out; 2643c9e9f97bSIlya Dryomov } 2644f46b5a66SChristoph Hellwig 26455516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 2646f46b5a66SChristoph Hellwig ret = btrfs_init_new_device(root, vol_args->name); 2647f46b5a66SChristoph Hellwig 264843d20761SAnand Jain if (!ret) 264943d20761SAnand Jain btrfs_info(root->fs_info, "disk added %s",vol_args->name); 265043d20761SAnand Jain 2651f46b5a66SChristoph Hellwig kfree(vol_args); 2652c9e9f97bSIlya Dryomov out: 2653c9e9f97bSIlya Dryomov mutex_unlock(&root->fs_info->volume_mutex); 26545ac00addSStefan Behrens atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0); 2655f46b5a66SChristoph Hellwig return ret; 2656f46b5a66SChristoph Hellwig } 2657f46b5a66SChristoph Hellwig 2658da24927bSMiao Xie static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg) 2659f46b5a66SChristoph Hellwig { 2660496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 2661f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2662f46b5a66SChristoph Hellwig int ret; 2663f46b5a66SChristoph Hellwig 2664e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2665e441d54dSChris Mason return -EPERM; 2666e441d54dSChris Mason 2667da24927bSMiao Xie ret = mnt_want_write_file(file); 2668da24927bSMiao Xie if (ret) 2669da24927bSMiao Xie return ret; 2670c146afadSYan Zheng 2671dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2672c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 2673c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 2674c47ca32dSDan Carpenter goto err_drop; 2675c9e9f97bSIlya Dryomov } 2676f46b5a66SChristoph Hellwig 26775516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 2678f46b5a66SChristoph Hellwig 2679183860f6SAnand Jain if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running, 2680183860f6SAnand Jain 1)) { 2681183860f6SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 2682183860f6SAnand Jain goto out; 2683183860f6SAnand Jain } 2684183860f6SAnand Jain 2685183860f6SAnand Jain mutex_lock(&root->fs_info->volume_mutex); 2686183860f6SAnand Jain ret = btrfs_rm_device(root, vol_args->name); 2687c9e9f97bSIlya Dryomov mutex_unlock(&root->fs_info->volume_mutex); 26885ac00addSStefan Behrens atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0); 2689183860f6SAnand Jain 2690ec95d491SAnand Jain if (!ret) 2691ec95d491SAnand Jain btrfs_info(root->fs_info, "disk deleted %s",vol_args->name); 2692ec95d491SAnand Jain 2693183860f6SAnand Jain out: 2694183860f6SAnand Jain kfree(vol_args); 2695c47ca32dSDan Carpenter err_drop: 26964ac20c70SIlya Dryomov mnt_drop_write_file(file); 2697f46b5a66SChristoph Hellwig return ret; 2698f46b5a66SChristoph Hellwig } 2699f46b5a66SChristoph Hellwig 2700475f6387SJan Schmidt static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg) 2701475f6387SJan Schmidt { 2702027ed2f0SLi Zefan struct btrfs_ioctl_fs_info_args *fi_args; 2703475f6387SJan Schmidt struct btrfs_device *device; 2704475f6387SJan Schmidt struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; 2705027ed2f0SLi Zefan int ret = 0; 2706475f6387SJan Schmidt 2707027ed2f0SLi Zefan fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL); 2708027ed2f0SLi Zefan if (!fi_args) 2709027ed2f0SLi Zefan return -ENOMEM; 2710027ed2f0SLi Zefan 2711f7171750SFilipe David Borba Manana mutex_lock(&fs_devices->device_list_mutex); 2712027ed2f0SLi Zefan fi_args->num_devices = fs_devices->num_devices; 2713027ed2f0SLi Zefan memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid)); 2714475f6387SJan Schmidt 2715d7641a49SByongho Lee list_for_each_entry(device, &fs_devices->devices, dev_list) { 2716027ed2f0SLi Zefan if (device->devid > fi_args->max_id) 2717027ed2f0SLi Zefan fi_args->max_id = device->devid; 2718475f6387SJan Schmidt } 2719475f6387SJan Schmidt mutex_unlock(&fs_devices->device_list_mutex); 2720475f6387SJan Schmidt 272180a773fbSDavid Sterba fi_args->nodesize = root->fs_info->super_copy->nodesize; 272280a773fbSDavid Sterba fi_args->sectorsize = root->fs_info->super_copy->sectorsize; 272380a773fbSDavid Sterba fi_args->clone_alignment = root->fs_info->super_copy->sectorsize; 272480a773fbSDavid Sterba 2725027ed2f0SLi Zefan if (copy_to_user(arg, fi_args, sizeof(*fi_args))) 2726027ed2f0SLi Zefan ret = -EFAULT; 2727475f6387SJan Schmidt 2728027ed2f0SLi Zefan kfree(fi_args); 2729027ed2f0SLi Zefan return ret; 2730475f6387SJan Schmidt } 2731475f6387SJan Schmidt 2732475f6387SJan Schmidt static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg) 2733475f6387SJan Schmidt { 2734475f6387SJan Schmidt struct btrfs_ioctl_dev_info_args *di_args; 2735475f6387SJan Schmidt struct btrfs_device *dev; 2736475f6387SJan Schmidt struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; 2737475f6387SJan Schmidt int ret = 0; 2738475f6387SJan Schmidt char *s_uuid = NULL; 2739475f6387SJan Schmidt 2740475f6387SJan Schmidt di_args = memdup_user(arg, sizeof(*di_args)); 2741475f6387SJan Schmidt if (IS_ERR(di_args)) 2742475f6387SJan Schmidt return PTR_ERR(di_args); 2743475f6387SJan Schmidt 2744dd5f9615SStefan Behrens if (!btrfs_is_empty_uuid(di_args->uuid)) 2745475f6387SJan Schmidt s_uuid = di_args->uuid; 2746475f6387SJan Schmidt 2747475f6387SJan Schmidt mutex_lock(&fs_devices->device_list_mutex); 2748aa1b8cd4SStefan Behrens dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL); 2749475f6387SJan Schmidt 2750475f6387SJan Schmidt if (!dev) { 2751475f6387SJan Schmidt ret = -ENODEV; 2752475f6387SJan Schmidt goto out; 2753475f6387SJan Schmidt } 2754475f6387SJan Schmidt 2755475f6387SJan Schmidt di_args->devid = dev->devid; 27567cc8e58dSMiao Xie di_args->bytes_used = btrfs_device_get_bytes_used(dev); 27577cc8e58dSMiao Xie di_args->total_bytes = btrfs_device_get_total_bytes(dev); 2758475f6387SJan Schmidt memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); 2759a27202fbSJim Meyering if (dev->name) { 2760606686eeSJosef Bacik struct rcu_string *name; 2761606686eeSJosef Bacik 2762606686eeSJosef Bacik rcu_read_lock(); 2763606686eeSJosef Bacik name = rcu_dereference(dev->name); 2764606686eeSJosef Bacik strncpy(di_args->path, name->str, sizeof(di_args->path)); 2765606686eeSJosef Bacik rcu_read_unlock(); 2766a27202fbSJim Meyering di_args->path[sizeof(di_args->path) - 1] = 0; 2767a27202fbSJim Meyering } else { 276899ba55adSStefan Behrens di_args->path[0] = '\0'; 2769a27202fbSJim Meyering } 2770475f6387SJan Schmidt 2771475f6387SJan Schmidt out: 277255793c0dSDavid Sterba mutex_unlock(&fs_devices->device_list_mutex); 2773475f6387SJan Schmidt if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) 2774475f6387SJan Schmidt ret = -EFAULT; 2775475f6387SJan Schmidt 2776475f6387SJan Schmidt kfree(di_args); 2777475f6387SJan Schmidt return ret; 2778475f6387SJan Schmidt } 2779475f6387SJan Schmidt 2780f4414602SMark Fasheh static struct page *extent_same_get_page(struct inode *inode, pgoff_t index) 2781416161dbSMark Fasheh { 2782416161dbSMark Fasheh struct page *page; 2783416161dbSMark Fasheh struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree; 2784416161dbSMark Fasheh 2785416161dbSMark Fasheh page = grab_cache_page(inode->i_mapping, index); 2786416161dbSMark Fasheh if (!page) 2787416161dbSMark Fasheh return NULL; 2788416161dbSMark Fasheh 2789416161dbSMark Fasheh if (!PageUptodate(page)) { 2790416161dbSMark Fasheh if (extent_read_full_page_nolock(tree, page, btrfs_get_extent, 2791416161dbSMark Fasheh 0)) 2792416161dbSMark Fasheh return NULL; 2793416161dbSMark Fasheh lock_page(page); 2794416161dbSMark Fasheh if (!PageUptodate(page)) { 2795416161dbSMark Fasheh unlock_page(page); 2796416161dbSMark Fasheh page_cache_release(page); 2797416161dbSMark Fasheh return NULL; 2798416161dbSMark Fasheh } 2799416161dbSMark Fasheh } 2800416161dbSMark Fasheh unlock_page(page); 2801416161dbSMark Fasheh 2802416161dbSMark Fasheh return page; 2803416161dbSMark Fasheh } 2804416161dbSMark Fasheh 2805f4414602SMark Fasheh static int gather_extent_pages(struct inode *inode, struct page **pages, 2806f4414602SMark Fasheh int num_pages, u64 off) 2807f4414602SMark Fasheh { 2808f4414602SMark Fasheh int i; 2809f4414602SMark Fasheh pgoff_t index = off >> PAGE_CACHE_SHIFT; 2810f4414602SMark Fasheh 2811f4414602SMark Fasheh for (i = 0; i < num_pages; i++) { 2812f4414602SMark Fasheh pages[i] = extent_same_get_page(inode, index + i); 2813f4414602SMark Fasheh if (!pages[i]) 2814f4414602SMark Fasheh return -ENOMEM; 2815f4414602SMark Fasheh } 2816f4414602SMark Fasheh return 0; 2817f4414602SMark Fasheh } 2818f4414602SMark Fasheh 281977fe20dcSMark Fasheh static inline void lock_extent_range(struct inode *inode, u64 off, u64 len) 282077fe20dcSMark Fasheh { 282177fe20dcSMark Fasheh /* do any pending delalloc/csum calc on src, one way or 282277fe20dcSMark Fasheh another, and lock file content */ 282377fe20dcSMark Fasheh while (1) { 282477fe20dcSMark Fasheh struct btrfs_ordered_extent *ordered; 282577fe20dcSMark Fasheh lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); 282677fe20dcSMark Fasheh ordered = btrfs_lookup_first_ordered_extent(inode, 282777fe20dcSMark Fasheh off + len - 1); 2828ff5df9b8SFilipe Manana if ((!ordered || 2829ff5df9b8SFilipe Manana ordered->file_offset + ordered->len <= off || 2830ff5df9b8SFilipe Manana ordered->file_offset >= off + len) && 283177fe20dcSMark Fasheh !test_range_bit(&BTRFS_I(inode)->io_tree, off, 2832ff5df9b8SFilipe Manana off + len - 1, EXTENT_DELALLOC, 0, NULL)) { 2833ff5df9b8SFilipe Manana if (ordered) 2834ff5df9b8SFilipe Manana btrfs_put_ordered_extent(ordered); 283577fe20dcSMark Fasheh break; 2836ff5df9b8SFilipe Manana } 283777fe20dcSMark Fasheh unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); 283877fe20dcSMark Fasheh if (ordered) 283977fe20dcSMark Fasheh btrfs_put_ordered_extent(ordered); 284077fe20dcSMark Fasheh btrfs_wait_ordered_range(inode, off, len); 284177fe20dcSMark Fasheh } 284277fe20dcSMark Fasheh } 284377fe20dcSMark Fasheh 2844f4414602SMark Fasheh static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2) 2845416161dbSMark Fasheh { 2846416161dbSMark Fasheh mutex_unlock(&inode1->i_mutex); 2847416161dbSMark Fasheh mutex_unlock(&inode2->i_mutex); 2848416161dbSMark Fasheh } 2849416161dbSMark Fasheh 2850f4414602SMark Fasheh static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2) 2851f4414602SMark Fasheh { 2852f4414602SMark Fasheh if (inode1 < inode2) 2853f4414602SMark Fasheh swap(inode1, inode2); 2854f4414602SMark Fasheh 2855f4414602SMark Fasheh mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT); 2856f4414602SMark Fasheh mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD); 2857f4414602SMark Fasheh } 2858f4414602SMark Fasheh 2859f4414602SMark Fasheh static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1, 2860f4414602SMark Fasheh struct inode *inode2, u64 loff2, u64 len) 2861f4414602SMark Fasheh { 2862f4414602SMark Fasheh unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); 2863f4414602SMark Fasheh unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); 2864f4414602SMark Fasheh } 2865f4414602SMark Fasheh 2866f4414602SMark Fasheh static void btrfs_double_extent_lock(struct inode *inode1, u64 loff1, 2867416161dbSMark Fasheh struct inode *inode2, u64 loff2, u64 len) 2868416161dbSMark Fasheh { 2869416161dbSMark Fasheh if (inode1 < inode2) { 2870416161dbSMark Fasheh swap(inode1, inode2); 2871416161dbSMark Fasheh swap(loff1, loff2); 2872416161dbSMark Fasheh } 2873416161dbSMark Fasheh lock_extent_range(inode1, loff1, len); 2874416161dbSMark Fasheh lock_extent_range(inode2, loff2, len); 2875416161dbSMark Fasheh } 2876f4414602SMark Fasheh 2877f4414602SMark Fasheh struct cmp_pages { 2878f4414602SMark Fasheh int num_pages; 2879f4414602SMark Fasheh struct page **src_pages; 2880f4414602SMark Fasheh struct page **dst_pages; 2881f4414602SMark Fasheh }; 2882f4414602SMark Fasheh 2883f4414602SMark Fasheh static void btrfs_cmp_data_free(struct cmp_pages *cmp) 2884f4414602SMark Fasheh { 2885f4414602SMark Fasheh int i; 2886f4414602SMark Fasheh struct page *pg; 2887f4414602SMark Fasheh 2888f4414602SMark Fasheh for (i = 0; i < cmp->num_pages; i++) { 2889f4414602SMark Fasheh pg = cmp->src_pages[i]; 2890f4414602SMark Fasheh if (pg) 2891f4414602SMark Fasheh page_cache_release(pg); 2892f4414602SMark Fasheh pg = cmp->dst_pages[i]; 2893f4414602SMark Fasheh if (pg) 2894f4414602SMark Fasheh page_cache_release(pg); 2895f4414602SMark Fasheh } 2896f4414602SMark Fasheh kfree(cmp->src_pages); 2897f4414602SMark Fasheh kfree(cmp->dst_pages); 2898f4414602SMark Fasheh } 2899f4414602SMark Fasheh 2900f4414602SMark Fasheh static int btrfs_cmp_data_prepare(struct inode *src, u64 loff, 2901f4414602SMark Fasheh struct inode *dst, u64 dst_loff, 2902f4414602SMark Fasheh u64 len, struct cmp_pages *cmp) 2903f4414602SMark Fasheh { 2904f4414602SMark Fasheh int ret; 2905f4414602SMark Fasheh int num_pages = PAGE_CACHE_ALIGN(len) >> PAGE_CACHE_SHIFT; 2906f4414602SMark Fasheh struct page **src_pgarr, **dst_pgarr; 2907f4414602SMark Fasheh 2908f4414602SMark Fasheh /* 2909f4414602SMark Fasheh * We must gather up all the pages before we initiate our 2910f4414602SMark Fasheh * extent locking. We use an array for the page pointers. Size 2911f4414602SMark Fasheh * of the array is bounded by len, which is in turn bounded by 2912f4414602SMark Fasheh * BTRFS_MAX_DEDUPE_LEN. 2913f4414602SMark Fasheh */ 2914f4414602SMark Fasheh src_pgarr = kzalloc(num_pages * sizeof(struct page *), GFP_NOFS); 2915f4414602SMark Fasheh dst_pgarr = kzalloc(num_pages * sizeof(struct page *), GFP_NOFS); 2916f4414602SMark Fasheh if (!src_pgarr || !dst_pgarr) { 2917f4414602SMark Fasheh kfree(src_pgarr); 2918f4414602SMark Fasheh kfree(dst_pgarr); 2919f4414602SMark Fasheh return -ENOMEM; 2920f4414602SMark Fasheh } 2921f4414602SMark Fasheh cmp->num_pages = num_pages; 2922f4414602SMark Fasheh cmp->src_pages = src_pgarr; 2923f4414602SMark Fasheh cmp->dst_pages = dst_pgarr; 2924f4414602SMark Fasheh 2925f4414602SMark Fasheh ret = gather_extent_pages(src, cmp->src_pages, cmp->num_pages, loff); 2926f4414602SMark Fasheh if (ret) 2927f4414602SMark Fasheh goto out; 2928f4414602SMark Fasheh 2929f4414602SMark Fasheh ret = gather_extent_pages(dst, cmp->dst_pages, cmp->num_pages, dst_loff); 2930f4414602SMark Fasheh 2931f4414602SMark Fasheh out: 2932f4414602SMark Fasheh if (ret) 2933f4414602SMark Fasheh btrfs_cmp_data_free(cmp); 2934f4414602SMark Fasheh return 0; 2935416161dbSMark Fasheh } 2936416161dbSMark Fasheh 2937416161dbSMark Fasheh static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst, 2938f4414602SMark Fasheh u64 dst_loff, u64 len, struct cmp_pages *cmp) 2939416161dbSMark Fasheh { 2940416161dbSMark Fasheh int ret = 0; 2941f4414602SMark Fasheh int i; 2942416161dbSMark Fasheh struct page *src_page, *dst_page; 2943416161dbSMark Fasheh unsigned int cmp_len = PAGE_CACHE_SIZE; 2944416161dbSMark Fasheh void *addr, *dst_addr; 2945416161dbSMark Fasheh 2946f4414602SMark Fasheh i = 0; 2947416161dbSMark Fasheh while (len) { 2948416161dbSMark Fasheh if (len < PAGE_CACHE_SIZE) 2949416161dbSMark Fasheh cmp_len = len; 2950416161dbSMark Fasheh 2951f4414602SMark Fasheh BUG_ON(i >= cmp->num_pages); 2952f4414602SMark Fasheh 2953f4414602SMark Fasheh src_page = cmp->src_pages[i]; 2954f4414602SMark Fasheh dst_page = cmp->dst_pages[i]; 2955f4414602SMark Fasheh 2956416161dbSMark Fasheh addr = kmap_atomic(src_page); 2957416161dbSMark Fasheh dst_addr = kmap_atomic(dst_page); 2958416161dbSMark Fasheh 2959416161dbSMark Fasheh flush_dcache_page(src_page); 2960416161dbSMark Fasheh flush_dcache_page(dst_page); 2961416161dbSMark Fasheh 2962416161dbSMark Fasheh if (memcmp(addr, dst_addr, cmp_len)) 2963416161dbSMark Fasheh ret = BTRFS_SAME_DATA_DIFFERS; 2964416161dbSMark Fasheh 2965416161dbSMark Fasheh kunmap_atomic(addr); 2966416161dbSMark Fasheh kunmap_atomic(dst_addr); 2967416161dbSMark Fasheh 2968416161dbSMark Fasheh if (ret) 2969416161dbSMark Fasheh break; 2970416161dbSMark Fasheh 2971416161dbSMark Fasheh len -= cmp_len; 2972f4414602SMark Fasheh i++; 2973416161dbSMark Fasheh } 2974416161dbSMark Fasheh 2975416161dbSMark Fasheh return ret; 2976416161dbSMark Fasheh } 2977416161dbSMark Fasheh 2978e1d227a4SMark Fasheh static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen, 2979e1d227a4SMark Fasheh u64 olen) 2980416161dbSMark Fasheh { 2981e1d227a4SMark Fasheh u64 len = *plen; 2982416161dbSMark Fasheh u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize; 2983416161dbSMark Fasheh 2984e1d227a4SMark Fasheh if (off + olen > inode->i_size || off + olen < off) 2985416161dbSMark Fasheh return -EINVAL; 2986e1d227a4SMark Fasheh 2987e1d227a4SMark Fasheh /* if we extend to eof, continue to block boundary */ 2988e1d227a4SMark Fasheh if (off + len == inode->i_size) 2989e1d227a4SMark Fasheh *plen = len = ALIGN(inode->i_size, bs) - off; 2990e1d227a4SMark Fasheh 2991416161dbSMark Fasheh /* Check that we are block aligned - btrfs_clone() requires this */ 2992416161dbSMark Fasheh if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs)) 2993416161dbSMark Fasheh return -EINVAL; 2994416161dbSMark Fasheh 2995416161dbSMark Fasheh return 0; 2996416161dbSMark Fasheh } 2997416161dbSMark Fasheh 2998e1d227a4SMark Fasheh static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen, 2999416161dbSMark Fasheh struct inode *dst, u64 dst_loff) 3000416161dbSMark Fasheh { 3001416161dbSMark Fasheh int ret; 3002e1d227a4SMark Fasheh u64 len = olen; 3003f4414602SMark Fasheh struct cmp_pages cmp; 30040efa9f48SMark Fasheh int same_inode = 0; 30050efa9f48SMark Fasheh u64 same_lock_start = 0; 30060efa9f48SMark Fasheh u64 same_lock_len = 0; 3007416161dbSMark Fasheh 3008416161dbSMark Fasheh if (src == dst) 30090efa9f48SMark Fasheh same_inode = 1; 3010416161dbSMark Fasheh 3011113e8283SFilipe Manana if (len == 0) 3012113e8283SFilipe Manana return 0; 3013113e8283SFilipe Manana 30140efa9f48SMark Fasheh if (same_inode) { 30150efa9f48SMark Fasheh mutex_lock(&src->i_mutex); 30160efa9f48SMark Fasheh 30170efa9f48SMark Fasheh ret = extent_same_check_offsets(src, loff, &len, olen); 30180efa9f48SMark Fasheh if (ret) 30190efa9f48SMark Fasheh goto out_unlock; 30200efa9f48SMark Fasheh 30210efa9f48SMark Fasheh /* 30220efa9f48SMark Fasheh * Single inode case wants the same checks, except we 30230efa9f48SMark Fasheh * don't want our length pushed out past i_size as 30240efa9f48SMark Fasheh * comparing that data range makes no sense. 30250efa9f48SMark Fasheh * 30260efa9f48SMark Fasheh * extent_same_check_offsets() will do this for an 30270efa9f48SMark Fasheh * unaligned length at i_size, so catch it here and 30280efa9f48SMark Fasheh * reject the request. 30290efa9f48SMark Fasheh * 30300efa9f48SMark Fasheh * This effectively means we require aligned extents 30310efa9f48SMark Fasheh * for the single-inode case, whereas the other cases 30320efa9f48SMark Fasheh * allow an unaligned length so long as it ends at 30330efa9f48SMark Fasheh * i_size. 30340efa9f48SMark Fasheh */ 30350efa9f48SMark Fasheh if (len != olen) { 30360efa9f48SMark Fasheh ret = -EINVAL; 30370efa9f48SMark Fasheh goto out_unlock; 30380efa9f48SMark Fasheh } 30390efa9f48SMark Fasheh 30400efa9f48SMark Fasheh /* Check for overlapping ranges */ 30410efa9f48SMark Fasheh if (dst_loff + len > loff && dst_loff < loff + len) { 30420efa9f48SMark Fasheh ret = -EINVAL; 30430efa9f48SMark Fasheh goto out_unlock; 30440efa9f48SMark Fasheh } 30450efa9f48SMark Fasheh 30460efa9f48SMark Fasheh same_lock_start = min_t(u64, loff, dst_loff); 30470efa9f48SMark Fasheh same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start; 30480efa9f48SMark Fasheh } else { 3049f4414602SMark Fasheh btrfs_double_inode_lock(src, dst); 3050416161dbSMark Fasheh 3051e1d227a4SMark Fasheh ret = extent_same_check_offsets(src, loff, &len, olen); 3052416161dbSMark Fasheh if (ret) 3053416161dbSMark Fasheh goto out_unlock; 3054416161dbSMark Fasheh 3055e1d227a4SMark Fasheh ret = extent_same_check_offsets(dst, dst_loff, &len, olen); 3056416161dbSMark Fasheh if (ret) 3057416161dbSMark Fasheh goto out_unlock; 30580efa9f48SMark Fasheh } 3059416161dbSMark Fasheh 3060416161dbSMark Fasheh /* don't make the dst file partly checksummed */ 3061416161dbSMark Fasheh if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != 3062416161dbSMark Fasheh (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) { 3063416161dbSMark Fasheh ret = -EINVAL; 3064416161dbSMark Fasheh goto out_unlock; 3065416161dbSMark Fasheh } 3066416161dbSMark Fasheh 3067f4414602SMark Fasheh ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, &cmp); 3068f4414602SMark Fasheh if (ret) 3069f4414602SMark Fasheh goto out_unlock; 3070f4414602SMark Fasheh 30710efa9f48SMark Fasheh if (same_inode) 30720efa9f48SMark Fasheh lock_extent_range(src, same_lock_start, same_lock_len); 30730efa9f48SMark Fasheh else 3074f4414602SMark Fasheh btrfs_double_extent_lock(src, loff, dst, dst_loff, len); 3075f4414602SMark Fasheh 3076207910ddSMark Fasheh /* pass original length for comparison so we stay within i_size */ 3077f4414602SMark Fasheh ret = btrfs_cmp_data(src, loff, dst, dst_loff, olen, &cmp); 3078416161dbSMark Fasheh if (ret == 0) 30791c919a5eSMark Fasheh ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1); 3080416161dbSMark Fasheh 30810efa9f48SMark Fasheh if (same_inode) 30820efa9f48SMark Fasheh unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start, 30830efa9f48SMark Fasheh same_lock_start + same_lock_len - 1); 30840efa9f48SMark Fasheh else 3085f4414602SMark Fasheh btrfs_double_extent_unlock(src, loff, dst, dst_loff, len); 3086f4414602SMark Fasheh 3087f4414602SMark Fasheh btrfs_cmp_data_free(&cmp); 3088416161dbSMark Fasheh out_unlock: 30890efa9f48SMark Fasheh if (same_inode) 30900efa9f48SMark Fasheh mutex_unlock(&src->i_mutex); 30910efa9f48SMark Fasheh else 3092f4414602SMark Fasheh btrfs_double_inode_unlock(src, dst); 3093416161dbSMark Fasheh 3094416161dbSMark Fasheh return ret; 3095416161dbSMark Fasheh } 3096416161dbSMark Fasheh 3097416161dbSMark Fasheh #define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024) 3098416161dbSMark Fasheh 3099416161dbSMark Fasheh static long btrfs_ioctl_file_extent_same(struct file *file, 31001c1c8747SAl Viro struct btrfs_ioctl_same_args __user *argp) 3101416161dbSMark Fasheh { 3102497b4050SFilipe Manana struct btrfs_ioctl_same_args *same = NULL; 3103cbf8b8caSMark Fasheh struct btrfs_ioctl_same_extent_info *info; 31041c1c8747SAl Viro struct inode *src = file_inode(file); 3105416161dbSMark Fasheh u64 off; 3106416161dbSMark Fasheh u64 len; 3107416161dbSMark Fasheh int i; 3108416161dbSMark Fasheh int ret; 3109cbf8b8caSMark Fasheh unsigned long size; 3110416161dbSMark Fasheh u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize; 3111416161dbSMark Fasheh bool is_admin = capable(CAP_SYS_ADMIN); 31121c1c8747SAl Viro u16 count; 3113416161dbSMark Fasheh 3114416161dbSMark Fasheh if (!(file->f_mode & FMODE_READ)) 3115416161dbSMark Fasheh return -EINVAL; 3116416161dbSMark Fasheh 3117416161dbSMark Fasheh ret = mnt_want_write_file(file); 3118416161dbSMark Fasheh if (ret) 3119416161dbSMark Fasheh return ret; 3120416161dbSMark Fasheh 31211c1c8747SAl Viro if (get_user(count, &argp->dest_count)) { 3122416161dbSMark Fasheh ret = -EFAULT; 3123416161dbSMark Fasheh goto out; 3124416161dbSMark Fasheh } 3125416161dbSMark Fasheh 31261c1c8747SAl Viro size = offsetof(struct btrfs_ioctl_same_args __user, info[count]); 3127cbf8b8caSMark Fasheh 31281c1c8747SAl Viro same = memdup_user(argp, size); 3129cbf8b8caSMark Fasheh 3130229eed43SGeyslan G. Bem if (IS_ERR(same)) { 3131229eed43SGeyslan G. Bem ret = PTR_ERR(same); 3132497b4050SFilipe Manana same = NULL; 3133cbf8b8caSMark Fasheh goto out; 3134cbf8b8caSMark Fasheh } 3135cbf8b8caSMark Fasheh 3136cbf8b8caSMark Fasheh off = same->logical_offset; 3137cbf8b8caSMark Fasheh len = same->length; 3138416161dbSMark Fasheh 3139416161dbSMark Fasheh /* 3140416161dbSMark Fasheh * Limit the total length we will dedupe for each operation. 3141416161dbSMark Fasheh * This is intended to bound the total time spent in this 3142416161dbSMark Fasheh * ioctl to something sane. 3143416161dbSMark Fasheh */ 3144416161dbSMark Fasheh if (len > BTRFS_MAX_DEDUPE_LEN) 3145416161dbSMark Fasheh len = BTRFS_MAX_DEDUPE_LEN; 3146416161dbSMark Fasheh 3147416161dbSMark Fasheh if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) { 3148416161dbSMark Fasheh /* 3149416161dbSMark Fasheh * Btrfs does not support blocksize < page_size. As a 3150416161dbSMark Fasheh * result, btrfs_cmp_data() won't correctly handle 3151416161dbSMark Fasheh * this situation without an update. 3152416161dbSMark Fasheh */ 3153416161dbSMark Fasheh ret = -EINVAL; 3154416161dbSMark Fasheh goto out; 3155416161dbSMark Fasheh } 3156416161dbSMark Fasheh 3157416161dbSMark Fasheh ret = -EISDIR; 3158416161dbSMark Fasheh if (S_ISDIR(src->i_mode)) 3159416161dbSMark Fasheh goto out; 3160416161dbSMark Fasheh 3161416161dbSMark Fasheh ret = -EACCES; 3162416161dbSMark Fasheh if (!S_ISREG(src->i_mode)) 3163416161dbSMark Fasheh goto out; 3164416161dbSMark Fasheh 3165cbf8b8caSMark Fasheh /* pre-format output fields to sane values */ 31661c1c8747SAl Viro for (i = 0; i < count; i++) { 3167cbf8b8caSMark Fasheh same->info[i].bytes_deduped = 0ULL; 3168cbf8b8caSMark Fasheh same->info[i].status = 0; 3169416161dbSMark Fasheh } 3170416161dbSMark Fasheh 31711c1c8747SAl Viro for (i = 0, info = same->info; i < count; i++, info++) { 31721c1c8747SAl Viro struct inode *dst; 31731c1c8747SAl Viro struct fd dst_file = fdget(info->fd); 31741c1c8747SAl Viro if (!dst_file.file) { 3175cbf8b8caSMark Fasheh info->status = -EBADF; 31761c1c8747SAl Viro continue; 3177416161dbSMark Fasheh } 31781c1c8747SAl Viro dst = file_inode(dst_file.file); 3179416161dbSMark Fasheh 31801c1c8747SAl Viro if (!(is_admin || (dst_file.file->f_mode & FMODE_WRITE))) { 3181cbf8b8caSMark Fasheh info->status = -EINVAL; 31821c1c8747SAl Viro } else if (file->f_path.mnt != dst_file.file->f_path.mnt) { 3183cbf8b8caSMark Fasheh info->status = -EXDEV; 31841c1c8747SAl Viro } else if (S_ISDIR(dst->i_mode)) { 3185cbf8b8caSMark Fasheh info->status = -EISDIR; 31861c1c8747SAl Viro } else if (!S_ISREG(dst->i_mode)) { 3187cbf8b8caSMark Fasheh info->status = -EACCES; 31881c1c8747SAl Viro } else { 3189cbf8b8caSMark Fasheh info->status = btrfs_extent_same(src, off, len, dst, 3190cbf8b8caSMark Fasheh info->logical_offset); 3191cbf8b8caSMark Fasheh if (info->status == 0) 3192cbf8b8caSMark Fasheh info->bytes_deduped += len; 31931c1c8747SAl Viro } 31941c1c8747SAl Viro fdput(dst_file); 3195cbf8b8caSMark Fasheh } 3196416161dbSMark Fasheh 3197cbf8b8caSMark Fasheh ret = copy_to_user(argp, same, size); 3198cbf8b8caSMark Fasheh if (ret) 3199416161dbSMark Fasheh ret = -EFAULT; 3200416161dbSMark Fasheh 3201416161dbSMark Fasheh out: 3202416161dbSMark Fasheh mnt_drop_write_file(file); 3203497b4050SFilipe Manana kfree(same); 3204416161dbSMark Fasheh return ret; 3205416161dbSMark Fasheh } 3206416161dbSMark Fasheh 3207f82a9901SFilipe Manana static int clone_finish_inode_update(struct btrfs_trans_handle *trans, 3208f82a9901SFilipe Manana struct inode *inode, 3209f82a9901SFilipe Manana u64 endoff, 3210f82a9901SFilipe Manana const u64 destoff, 32111c919a5eSMark Fasheh const u64 olen, 32121c919a5eSMark Fasheh int no_time_update) 3213f82a9901SFilipe Manana { 3214f82a9901SFilipe Manana struct btrfs_root *root = BTRFS_I(inode)->root; 3215f82a9901SFilipe Manana int ret; 3216f82a9901SFilipe Manana 3217f82a9901SFilipe Manana inode_inc_iversion(inode); 32181c919a5eSMark Fasheh if (!no_time_update) 3219f82a9901SFilipe Manana inode->i_mtime = inode->i_ctime = CURRENT_TIME; 3220f82a9901SFilipe Manana /* 3221f82a9901SFilipe Manana * We round up to the block size at eof when determining which 3222f82a9901SFilipe Manana * extents to clone above, but shouldn't round up the file size. 3223f82a9901SFilipe Manana */ 3224f82a9901SFilipe Manana if (endoff > destoff + olen) 3225f82a9901SFilipe Manana endoff = destoff + olen; 3226f82a9901SFilipe Manana if (endoff > inode->i_size) 3227f82a9901SFilipe Manana btrfs_i_size_write(inode, endoff); 3228f82a9901SFilipe Manana 3229f82a9901SFilipe Manana ret = btrfs_update_inode(trans, root, inode); 3230f82a9901SFilipe Manana if (ret) { 3231f82a9901SFilipe Manana btrfs_abort_transaction(trans, root, ret); 3232f82a9901SFilipe Manana btrfs_end_transaction(trans, root); 3233f82a9901SFilipe Manana goto out; 3234f82a9901SFilipe Manana } 3235f82a9901SFilipe Manana ret = btrfs_end_transaction(trans, root); 3236f82a9901SFilipe Manana out: 3237f82a9901SFilipe Manana return ret; 3238f82a9901SFilipe Manana } 3239f82a9901SFilipe Manana 32407ffbb598SFilipe Manana static void clone_update_extent_map(struct inode *inode, 32417ffbb598SFilipe Manana const struct btrfs_trans_handle *trans, 32427ffbb598SFilipe Manana const struct btrfs_path *path, 32437ffbb598SFilipe Manana const u64 hole_offset, 32447ffbb598SFilipe Manana const u64 hole_len) 32457ffbb598SFilipe Manana { 32467ffbb598SFilipe Manana struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 32477ffbb598SFilipe Manana struct extent_map *em; 32487ffbb598SFilipe Manana int ret; 32497ffbb598SFilipe Manana 32507ffbb598SFilipe Manana em = alloc_extent_map(); 32517ffbb598SFilipe Manana if (!em) { 32527ffbb598SFilipe Manana set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, 32537ffbb598SFilipe Manana &BTRFS_I(inode)->runtime_flags); 32547ffbb598SFilipe Manana return; 32557ffbb598SFilipe Manana } 32567ffbb598SFilipe Manana 325714f59796SFilipe Manana if (path) { 325814f59796SFilipe Manana struct btrfs_file_extent_item *fi; 325914f59796SFilipe Manana 326014f59796SFilipe Manana fi = btrfs_item_ptr(path->nodes[0], path->slots[0], 326114f59796SFilipe Manana struct btrfs_file_extent_item); 32627ffbb598SFilipe Manana btrfs_extent_item_to_extent_map(inode, path, fi, false, em); 32637ffbb598SFilipe Manana em->generation = -1; 32647ffbb598SFilipe Manana if (btrfs_file_extent_type(path->nodes[0], fi) == 32657ffbb598SFilipe Manana BTRFS_FILE_EXTENT_INLINE) 32667ffbb598SFilipe Manana set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, 32677ffbb598SFilipe Manana &BTRFS_I(inode)->runtime_flags); 32687ffbb598SFilipe Manana } else { 32697ffbb598SFilipe Manana em->start = hole_offset; 32707ffbb598SFilipe Manana em->len = hole_len; 32717ffbb598SFilipe Manana em->ram_bytes = em->len; 32727ffbb598SFilipe Manana em->orig_start = hole_offset; 32737ffbb598SFilipe Manana em->block_start = EXTENT_MAP_HOLE; 32747ffbb598SFilipe Manana em->block_len = 0; 32757ffbb598SFilipe Manana em->orig_block_len = 0; 32767ffbb598SFilipe Manana em->compress_type = BTRFS_COMPRESS_NONE; 32777ffbb598SFilipe Manana em->generation = trans->transid; 32787ffbb598SFilipe Manana } 32797ffbb598SFilipe Manana 32807ffbb598SFilipe Manana while (1) { 32817ffbb598SFilipe Manana write_lock(&em_tree->lock); 32827ffbb598SFilipe Manana ret = add_extent_mapping(em_tree, em, 1); 32837ffbb598SFilipe Manana write_unlock(&em_tree->lock); 32847ffbb598SFilipe Manana if (ret != -EEXIST) { 32857ffbb598SFilipe Manana free_extent_map(em); 32867ffbb598SFilipe Manana break; 32877ffbb598SFilipe Manana } 32887ffbb598SFilipe Manana btrfs_drop_extent_cache(inode, em->start, 32897ffbb598SFilipe Manana em->start + em->len - 1, 0); 32907ffbb598SFilipe Manana } 32917ffbb598SFilipe Manana 3292ee39b432SDavid Sterba if (ret) 32937ffbb598SFilipe Manana set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, 32947ffbb598SFilipe Manana &BTRFS_I(inode)->runtime_flags); 32957ffbb598SFilipe Manana } 32967ffbb598SFilipe Manana 32978039d87dSFilipe Manana /* 32988039d87dSFilipe Manana * Make sure we do not end up inserting an inline extent into a file that has 32998039d87dSFilipe Manana * already other (non-inline) extents. If a file has an inline extent it can 33008039d87dSFilipe Manana * not have any other extents and the (single) inline extent must start at the 33018039d87dSFilipe Manana * file offset 0. Failing to respect these rules will lead to file corruption, 33028039d87dSFilipe Manana * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc 33038039d87dSFilipe Manana * 33048039d87dSFilipe Manana * We can have extents that have been already written to disk or we can have 33058039d87dSFilipe Manana * dirty ranges still in delalloc, in which case the extent maps and items are 33068039d87dSFilipe Manana * created only when we run delalloc, and the delalloc ranges might fall outside 33078039d87dSFilipe Manana * the range we are currently locking in the inode's io tree. So we check the 33088039d87dSFilipe Manana * inode's i_size because of that (i_size updates are done while holding the 33098039d87dSFilipe Manana * i_mutex, which we are holding here). 33108039d87dSFilipe Manana * We also check to see if the inode has a size not greater than "datal" but has 33118039d87dSFilipe Manana * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are 33128039d87dSFilipe Manana * protected against such concurrent fallocate calls by the i_mutex). 33138039d87dSFilipe Manana * 33148039d87dSFilipe Manana * If the file has no extents but a size greater than datal, do not allow the 33158039d87dSFilipe Manana * copy because we would need turn the inline extent into a non-inline one (even 33168039d87dSFilipe Manana * with NO_HOLES enabled). If we find our destination inode only has one inline 33178039d87dSFilipe Manana * extent, just overwrite it with the source inline extent if its size is less 33188039d87dSFilipe Manana * than the source extent's size, or we could copy the source inline extent's 33198039d87dSFilipe Manana * data into the destination inode's inline extent if the later is greater then 33208039d87dSFilipe Manana * the former. 33218039d87dSFilipe Manana */ 33228039d87dSFilipe Manana static int clone_copy_inline_extent(struct inode *src, 33238039d87dSFilipe Manana struct inode *dst, 33248039d87dSFilipe Manana struct btrfs_trans_handle *trans, 33258039d87dSFilipe Manana struct btrfs_path *path, 33268039d87dSFilipe Manana struct btrfs_key *new_key, 33278039d87dSFilipe Manana const u64 drop_start, 33288039d87dSFilipe Manana const u64 datal, 33298039d87dSFilipe Manana const u64 skip, 33308039d87dSFilipe Manana const u64 size, 33318039d87dSFilipe Manana char *inline_data) 33328039d87dSFilipe Manana { 33338039d87dSFilipe Manana struct btrfs_root *root = BTRFS_I(dst)->root; 33348039d87dSFilipe Manana const u64 aligned_end = ALIGN(new_key->offset + datal, 33358039d87dSFilipe Manana root->sectorsize); 33368039d87dSFilipe Manana int ret; 33378039d87dSFilipe Manana struct btrfs_key key; 33388039d87dSFilipe Manana 33398039d87dSFilipe Manana if (new_key->offset > 0) 33408039d87dSFilipe Manana return -EOPNOTSUPP; 33418039d87dSFilipe Manana 33428039d87dSFilipe Manana key.objectid = btrfs_ino(dst); 33438039d87dSFilipe Manana key.type = BTRFS_EXTENT_DATA_KEY; 33448039d87dSFilipe Manana key.offset = 0; 33458039d87dSFilipe Manana ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 33468039d87dSFilipe Manana if (ret < 0) { 33478039d87dSFilipe Manana return ret; 33488039d87dSFilipe Manana } else if (ret > 0) { 33498039d87dSFilipe Manana if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { 33508039d87dSFilipe Manana ret = btrfs_next_leaf(root, path); 33518039d87dSFilipe Manana if (ret < 0) 33528039d87dSFilipe Manana return ret; 33538039d87dSFilipe Manana else if (ret > 0) 33548039d87dSFilipe Manana goto copy_inline_extent; 33558039d87dSFilipe Manana } 33568039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 33578039d87dSFilipe Manana if (key.objectid == btrfs_ino(dst) && 33588039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) { 33598039d87dSFilipe Manana ASSERT(key.offset > 0); 33608039d87dSFilipe Manana return -EOPNOTSUPP; 33618039d87dSFilipe Manana } 33628039d87dSFilipe Manana } else if (i_size_read(dst) <= datal) { 33638039d87dSFilipe Manana struct btrfs_file_extent_item *ei; 33648039d87dSFilipe Manana u64 ext_len; 33658039d87dSFilipe Manana 33668039d87dSFilipe Manana /* 33678039d87dSFilipe Manana * If the file size is <= datal, make sure there are no other 33688039d87dSFilipe Manana * extents following (can happen do to an fallocate call with 33698039d87dSFilipe Manana * the flag FALLOC_FL_KEEP_SIZE). 33708039d87dSFilipe Manana */ 33718039d87dSFilipe Manana ei = btrfs_item_ptr(path->nodes[0], path->slots[0], 33728039d87dSFilipe Manana struct btrfs_file_extent_item); 33738039d87dSFilipe Manana /* 33748039d87dSFilipe Manana * If it's an inline extent, it can not have other extents 33758039d87dSFilipe Manana * following it. 33768039d87dSFilipe Manana */ 33778039d87dSFilipe Manana if (btrfs_file_extent_type(path->nodes[0], ei) == 33788039d87dSFilipe Manana BTRFS_FILE_EXTENT_INLINE) 33798039d87dSFilipe Manana goto copy_inline_extent; 33808039d87dSFilipe Manana 33818039d87dSFilipe Manana ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei); 33828039d87dSFilipe Manana if (ext_len > aligned_end) 33838039d87dSFilipe Manana return -EOPNOTSUPP; 33848039d87dSFilipe Manana 33858039d87dSFilipe Manana ret = btrfs_next_item(root, path); 33868039d87dSFilipe Manana if (ret < 0) { 33878039d87dSFilipe Manana return ret; 33888039d87dSFilipe Manana } else if (ret == 0) { 33898039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 33908039d87dSFilipe Manana path->slots[0]); 33918039d87dSFilipe Manana if (key.objectid == btrfs_ino(dst) && 33928039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) 33938039d87dSFilipe Manana return -EOPNOTSUPP; 33948039d87dSFilipe Manana } 33958039d87dSFilipe Manana } 33968039d87dSFilipe Manana 33978039d87dSFilipe Manana copy_inline_extent: 33988039d87dSFilipe Manana /* 33998039d87dSFilipe Manana * We have no extent items, or we have an extent at offset 0 which may 34008039d87dSFilipe Manana * or may not be inlined. All these cases are dealt the same way. 34018039d87dSFilipe Manana */ 34028039d87dSFilipe Manana if (i_size_read(dst) > datal) { 34038039d87dSFilipe Manana /* 34048039d87dSFilipe Manana * If the destination inode has an inline extent... 34058039d87dSFilipe Manana * This would require copying the data from the source inline 34068039d87dSFilipe Manana * extent into the beginning of the destination's inline extent. 34078039d87dSFilipe Manana * But this is really complex, both extents can be compressed 34088039d87dSFilipe Manana * or just one of them, which would require decompressing and 34098039d87dSFilipe Manana * re-compressing data (which could increase the new compressed 34108039d87dSFilipe Manana * size, not allowing the compressed data to fit anymore in an 34118039d87dSFilipe Manana * inline extent). 34128039d87dSFilipe Manana * So just don't support this case for now (it should be rare, 34138039d87dSFilipe Manana * we are not really saving space when cloning inline extents). 34148039d87dSFilipe Manana */ 34158039d87dSFilipe Manana return -EOPNOTSUPP; 34168039d87dSFilipe Manana } 34178039d87dSFilipe Manana 34188039d87dSFilipe Manana btrfs_release_path(path); 34198039d87dSFilipe Manana ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1); 34208039d87dSFilipe Manana if (ret) 34218039d87dSFilipe Manana return ret; 34228039d87dSFilipe Manana ret = btrfs_insert_empty_item(trans, root, path, new_key, size); 34238039d87dSFilipe Manana if (ret) 34248039d87dSFilipe Manana return ret; 34258039d87dSFilipe Manana 34268039d87dSFilipe Manana if (skip) { 34278039d87dSFilipe Manana const u32 start = btrfs_file_extent_calc_inline_size(0); 34288039d87dSFilipe Manana 34298039d87dSFilipe Manana memmove(inline_data + start, inline_data + start + skip, datal); 34308039d87dSFilipe Manana } 34318039d87dSFilipe Manana 34328039d87dSFilipe Manana write_extent_buffer(path->nodes[0], inline_data, 34338039d87dSFilipe Manana btrfs_item_ptr_offset(path->nodes[0], 34348039d87dSFilipe Manana path->slots[0]), 34358039d87dSFilipe Manana size); 34368039d87dSFilipe Manana inode_add_bytes(dst, datal); 34378039d87dSFilipe Manana 34388039d87dSFilipe Manana return 0; 34398039d87dSFilipe Manana } 34408039d87dSFilipe Manana 344132b7c687SMark Fasheh /** 344232b7c687SMark Fasheh * btrfs_clone() - clone a range from inode file to another 344332b7c687SMark Fasheh * 344432b7c687SMark Fasheh * @src: Inode to clone from 344532b7c687SMark Fasheh * @inode: Inode to clone to 344632b7c687SMark Fasheh * @off: Offset within source to start clone from 344732b7c687SMark Fasheh * @olen: Original length, passed by user, of range to clone 34481c919a5eSMark Fasheh * @olen_aligned: Block-aligned value of olen 344932b7c687SMark Fasheh * @destoff: Offset within @inode to start clone 34501c919a5eSMark Fasheh * @no_time_update: Whether to update mtime/ctime on the target inode 345132b7c687SMark Fasheh */ 345232b7c687SMark Fasheh static int btrfs_clone(struct inode *src, struct inode *inode, 3453f82a9901SFilipe Manana const u64 off, const u64 olen, const u64 olen_aligned, 34541c919a5eSMark Fasheh const u64 destoff, int no_time_update) 3455f46b5a66SChristoph Hellwig { 3456f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 345732b7c687SMark Fasheh struct btrfs_path *path = NULL; 3458f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 345932b7c687SMark Fasheh struct btrfs_trans_handle *trans; 346032b7c687SMark Fasheh char *buf = NULL; 3461ae01a0abSYan Zheng struct btrfs_key key; 3462f46b5a66SChristoph Hellwig u32 nritems; 3463f46b5a66SChristoph Hellwig int slot; 3464ae01a0abSYan Zheng int ret; 3465f82a9901SFilipe Manana const u64 len = olen_aligned; 3466f82a9901SFilipe Manana u64 last_dest_end = destoff; 3467ae01a0abSYan Zheng 3468ae01a0abSYan Zheng ret = -ENOMEM; 3469707e8a07SDavid Sterba buf = vmalloc(root->nodesize); 3470ae01a0abSYan Zheng if (!buf) 347132b7c687SMark Fasheh return ret; 3472ae01a0abSYan Zheng 3473ae01a0abSYan Zheng path = btrfs_alloc_path(); 3474ae01a0abSYan Zheng if (!path) { 3475ae01a0abSYan Zheng vfree(buf); 347632b7c687SMark Fasheh return ret; 3477ae01a0abSYan Zheng } 347832b7c687SMark Fasheh 3479e4058b54SDavid Sterba path->reada = READA_FORWARD; 3480c5c9cd4dSSage Weil /* clone data */ 348133345d01SLi Zefan key.objectid = btrfs_ino(src); 3482ae01a0abSYan Zheng key.type = BTRFS_EXTENT_DATA_KEY; 34832c463823SFilipe Manana key.offset = off; 3484f46b5a66SChristoph Hellwig 3485f46b5a66SChristoph Hellwig while (1) { 3486de249e66SChris Mason u64 next_key_min_offset = key.offset + 1; 3487df858e76SFilipe Manana 3488f46b5a66SChristoph Hellwig /* 3489f46b5a66SChristoph Hellwig * note the key will change type as we walk through the 3490f46b5a66SChristoph Hellwig * tree. 3491f46b5a66SChristoph Hellwig */ 3492e4355f34SFilipe David Borba Manana path->leave_spinning = 1; 3493362a20c5SDavid Sterba ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path, 3494362a20c5SDavid Sterba 0, 0); 3495f46b5a66SChristoph Hellwig if (ret < 0) 3496f46b5a66SChristoph Hellwig goto out; 34972c463823SFilipe Manana /* 34982c463823SFilipe Manana * First search, if no extent item that starts at offset off was 34992c463823SFilipe Manana * found but the previous item is an extent item, it's possible 35002c463823SFilipe Manana * it might overlap our target range, therefore process it. 35012c463823SFilipe Manana */ 35022c463823SFilipe Manana if (key.offset == off && ret > 0 && path->slots[0] > 0) { 35032c463823SFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 35042c463823SFilipe Manana path->slots[0] - 1); 35052c463823SFilipe Manana if (key.type == BTRFS_EXTENT_DATA_KEY) 35062c463823SFilipe Manana path->slots[0]--; 35072c463823SFilipe Manana } 3508f46b5a66SChristoph Hellwig 3509ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3510e4355f34SFilipe David Borba Manana process_slot: 3511ae01a0abSYan Zheng if (path->slots[0] >= nritems) { 3512362a20c5SDavid Sterba ret = btrfs_next_leaf(BTRFS_I(src)->root, path); 3513f46b5a66SChristoph Hellwig if (ret < 0) 3514f46b5a66SChristoph Hellwig goto out; 3515f46b5a66SChristoph Hellwig if (ret > 0) 3516f46b5a66SChristoph Hellwig break; 3517ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3518f46b5a66SChristoph Hellwig } 3519f46b5a66SChristoph Hellwig leaf = path->nodes[0]; 3520f46b5a66SChristoph Hellwig slot = path->slots[0]; 3521f46b5a66SChristoph Hellwig 3522ae01a0abSYan Zheng btrfs_item_key_to_cpu(leaf, &key, slot); 3523962a298fSDavid Sterba if (key.type > BTRFS_EXTENT_DATA_KEY || 352433345d01SLi Zefan key.objectid != btrfs_ino(src)) 3525f46b5a66SChristoph Hellwig break; 3526f46b5a66SChristoph Hellwig 3527962a298fSDavid Sterba if (key.type == BTRFS_EXTENT_DATA_KEY) { 3528c5c9cd4dSSage Weil struct btrfs_file_extent_item *extent; 3529c5c9cd4dSSage Weil int type; 353031840ae1SZheng Yan u32 size; 353131840ae1SZheng Yan struct btrfs_key new_key; 3532c5c9cd4dSSage Weil u64 disko = 0, diskl = 0; 3533c5c9cd4dSSage Weil u64 datao = 0, datal = 0; 3534c5c9cd4dSSage Weil u8 comp; 3535f82a9901SFilipe Manana u64 drop_start; 353631840ae1SZheng Yan 3537c5c9cd4dSSage Weil extent = btrfs_item_ptr(leaf, slot, 3538c5c9cd4dSSage Weil struct btrfs_file_extent_item); 3539c5c9cd4dSSage Weil comp = btrfs_file_extent_compression(leaf, extent); 3540c5c9cd4dSSage Weil type = btrfs_file_extent_type(leaf, extent); 3541c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3542c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3543d397712bSChris Mason disko = btrfs_file_extent_disk_bytenr(leaf, 3544d397712bSChris Mason extent); 3545d397712bSChris Mason diskl = btrfs_file_extent_disk_num_bytes(leaf, 3546d397712bSChris Mason extent); 3547c5c9cd4dSSage Weil datao = btrfs_file_extent_offset(leaf, extent); 3548d397712bSChris Mason datal = btrfs_file_extent_num_bytes(leaf, 3549d397712bSChris Mason extent); 3550c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3551c5c9cd4dSSage Weil /* take upper bound, may be compressed */ 3552c5c9cd4dSSage Weil datal = btrfs_file_extent_ram_bytes(leaf, 3553c5c9cd4dSSage Weil extent); 3554c5c9cd4dSSage Weil } 355531840ae1SZheng Yan 35562c463823SFilipe Manana /* 35572c463823SFilipe Manana * The first search might have left us at an extent 35582c463823SFilipe Manana * item that ends before our target range's start, can 35592c463823SFilipe Manana * happen if we have holes and NO_HOLES feature enabled. 35602c463823SFilipe Manana */ 35612c463823SFilipe Manana if (key.offset + datal <= off) { 3562e4355f34SFilipe David Borba Manana path->slots[0]++; 3563e4355f34SFilipe David Borba Manana goto process_slot; 35642c463823SFilipe Manana } else if (key.offset >= off + len) { 35652c463823SFilipe Manana break; 3566e4355f34SFilipe David Borba Manana } 3567df858e76SFilipe Manana next_key_min_offset = key.offset + datal; 3568e4355f34SFilipe David Borba Manana size = btrfs_item_size_nr(leaf, slot); 3569e4355f34SFilipe David Borba Manana read_extent_buffer(leaf, buf, 3570e4355f34SFilipe David Borba Manana btrfs_item_ptr_offset(leaf, slot), 3571e4355f34SFilipe David Borba Manana size); 3572e4355f34SFilipe David Borba Manana 3573e4355f34SFilipe David Borba Manana btrfs_release_path(path); 3574e4355f34SFilipe David Borba Manana path->leave_spinning = 0; 3575c5c9cd4dSSage Weil 357631840ae1SZheng Yan memcpy(&new_key, &key, sizeof(new_key)); 357733345d01SLi Zefan new_key.objectid = btrfs_ino(inode); 35784d728ec7SLi Zefan if (off <= key.offset) 3579c5c9cd4dSSage Weil new_key.offset = key.offset + destoff - off; 35804d728ec7SLi Zefan else 35814d728ec7SLi Zefan new_key.offset = destoff; 3582c5c9cd4dSSage Weil 3583b6f3409bSSage Weil /* 3584f82a9901SFilipe Manana * Deal with a hole that doesn't have an extent item 3585f82a9901SFilipe Manana * that represents it (NO_HOLES feature enabled). 3586f82a9901SFilipe Manana * This hole is either in the middle of the cloning 3587f82a9901SFilipe Manana * range or at the beginning (fully overlaps it or 3588f82a9901SFilipe Manana * partially overlaps it). 3589f82a9901SFilipe Manana */ 3590f82a9901SFilipe Manana if (new_key.offset != last_dest_end) 3591f82a9901SFilipe Manana drop_start = last_dest_end; 3592f82a9901SFilipe Manana else 3593f82a9901SFilipe Manana drop_start = new_key.offset; 3594f82a9901SFilipe Manana 3595f82a9901SFilipe Manana /* 3596b6f3409bSSage Weil * 1 - adjusting old extent (we may have to split it) 3597b6f3409bSSage Weil * 1 - add new extent 3598b6f3409bSSage Weil * 1 - inode update 3599b6f3409bSSage Weil */ 3600b6f3409bSSage Weil trans = btrfs_start_transaction(root, 3); 3601a22285a6SYan, Zheng if (IS_ERR(trans)) { 3602a22285a6SYan, Zheng ret = PTR_ERR(trans); 3603a22285a6SYan, Zheng goto out; 3604a22285a6SYan, Zheng } 3605a22285a6SYan, Zheng 3606c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3607c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3608d72c0842SLi Zefan /* 3609d72c0842SLi Zefan * a | --- range to clone ---| b 3610d72c0842SLi Zefan * | ------------- extent ------------- | 3611d72c0842SLi Zefan */ 3612d72c0842SLi Zefan 361393915584SAntonio Ospite /* subtract range b */ 3614d72c0842SLi Zefan if (key.offset + datal > off + len) 3615d72c0842SLi Zefan datal = off + len - key.offset; 3616d72c0842SLi Zefan 361793915584SAntonio Ospite /* subtract range a */ 3618a22285a6SYan, Zheng if (off > key.offset) { 3619a22285a6SYan, Zheng datao += off - key.offset; 3620a22285a6SYan, Zheng datal -= off - key.offset; 3621a22285a6SYan, Zheng } 3622a22285a6SYan, Zheng 36235dc562c5SJosef Bacik ret = btrfs_drop_extents(trans, root, inode, 3624f82a9901SFilipe Manana drop_start, 3625a22285a6SYan, Zheng new_key.offset + datal, 36262671485dSJosef Bacik 1); 362779787eaaSJeff Mahoney if (ret) { 36283f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 362900fdf13aSLiu Bo btrfs_abort_transaction(trans, 363000fdf13aSLiu Bo root, ret); 363179787eaaSJeff Mahoney btrfs_end_transaction(trans, root); 363279787eaaSJeff Mahoney goto out; 363379787eaaSJeff Mahoney } 3634a22285a6SYan, Zheng 363531840ae1SZheng Yan ret = btrfs_insert_empty_item(trans, root, path, 363631840ae1SZheng Yan &new_key, size); 363779787eaaSJeff Mahoney if (ret) { 363879787eaaSJeff Mahoney btrfs_abort_transaction(trans, root, 363979787eaaSJeff Mahoney ret); 364079787eaaSJeff Mahoney btrfs_end_transaction(trans, root); 364179787eaaSJeff Mahoney goto out; 364279787eaaSJeff Mahoney } 364331840ae1SZheng Yan 364431840ae1SZheng Yan leaf = path->nodes[0]; 364531840ae1SZheng Yan slot = path->slots[0]; 364631840ae1SZheng Yan write_extent_buffer(leaf, buf, 364731840ae1SZheng Yan btrfs_item_ptr_offset(leaf, slot), 364831840ae1SZheng Yan size); 3649ae01a0abSYan Zheng 3650f46b5a66SChristoph Hellwig extent = btrfs_item_ptr(leaf, slot, 3651f46b5a66SChristoph Hellwig struct btrfs_file_extent_item); 3652c5c9cd4dSSage Weil 3653c5c9cd4dSSage Weil /* disko == 0 means it's a hole */ 3654c5c9cd4dSSage Weil if (!disko) 3655c5c9cd4dSSage Weil datao = 0; 3656c5c9cd4dSSage Weil 3657c5c9cd4dSSage Weil btrfs_set_file_extent_offset(leaf, extent, 3658c5c9cd4dSSage Weil datao); 3659c5c9cd4dSSage Weil btrfs_set_file_extent_num_bytes(leaf, extent, 3660c5c9cd4dSSage Weil datal); 3661fcebe456SJosef Bacik 3662c5c9cd4dSSage Weil if (disko) { 3663c5c9cd4dSSage Weil inode_add_bytes(inode, datal); 3664ae01a0abSYan Zheng ret = btrfs_inc_extent_ref(trans, root, 36655d4f98a2SYan Zheng disko, diskl, 0, 3666f46b5a66SChristoph Hellwig root->root_key.objectid, 366733345d01SLi Zefan btrfs_ino(inode), 3668b06c4bf5SFilipe Manana new_key.offset - datao); 366979787eaaSJeff Mahoney if (ret) { 367079787eaaSJeff Mahoney btrfs_abort_transaction(trans, 367179787eaaSJeff Mahoney root, 367279787eaaSJeff Mahoney ret); 367379787eaaSJeff Mahoney btrfs_end_transaction(trans, 367479787eaaSJeff Mahoney root); 367579787eaaSJeff Mahoney goto out; 367679787eaaSJeff Mahoney 367779787eaaSJeff Mahoney } 3678f46b5a66SChristoph Hellwig } 3679c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3680c5c9cd4dSSage Weil u64 skip = 0; 3681c5c9cd4dSSage Weil u64 trim = 0; 3682ed958762SFilipe Manana 3683c5c9cd4dSSage Weil if (off > key.offset) { 3684c5c9cd4dSSage Weil skip = off - key.offset; 3685c5c9cd4dSSage Weil new_key.offset += skip; 368631840ae1SZheng Yan } 3687d397712bSChris Mason 3688c5c9cd4dSSage Weil if (key.offset + datal > off + len) 3689c5c9cd4dSSage Weil trim = key.offset + datal - (off + len); 3690d397712bSChris Mason 3691c5c9cd4dSSage Weil if (comp && (skip || trim)) { 3692c5c9cd4dSSage Weil ret = -EINVAL; 3693a22285a6SYan, Zheng btrfs_end_transaction(trans, root); 3694c5c9cd4dSSage Weil goto out; 369531840ae1SZheng Yan } 3696c5c9cd4dSSage Weil size -= skip + trim; 3697c5c9cd4dSSage Weil datal -= skip + trim; 3698a22285a6SYan, Zheng 36998039d87dSFilipe Manana ret = clone_copy_inline_extent(src, inode, 37008039d87dSFilipe Manana trans, path, 37018039d87dSFilipe Manana &new_key, 3702f82a9901SFilipe Manana drop_start, 37038039d87dSFilipe Manana datal, 37048039d87dSFilipe Manana skip, size, buf); 370579787eaaSJeff Mahoney if (ret) { 37063f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 37073a29bc09SChris Mason btrfs_abort_transaction(trans, 37088039d87dSFilipe Manana root, 370979787eaaSJeff Mahoney ret); 371079787eaaSJeff Mahoney btrfs_end_transaction(trans, root); 371179787eaaSJeff Mahoney goto out; 371279787eaaSJeff Mahoney } 3713c5c9cd4dSSage Weil leaf = path->nodes[0]; 3714c5c9cd4dSSage Weil slot = path->slots[0]; 3715c5c9cd4dSSage Weil } 3716c5c9cd4dSSage Weil 37177ffbb598SFilipe Manana /* If we have an implicit hole (NO_HOLES feature). */ 37187ffbb598SFilipe Manana if (drop_start < new_key.offset) 37197ffbb598SFilipe Manana clone_update_extent_map(inode, trans, 372014f59796SFilipe Manana NULL, drop_start, 37217ffbb598SFilipe Manana new_key.offset - drop_start); 37227ffbb598SFilipe Manana 372314f59796SFilipe Manana clone_update_extent_map(inode, trans, path, 0, 0); 37247ffbb598SFilipe Manana 3725c5c9cd4dSSage Weil btrfs_mark_buffer_dirty(leaf); 3726b3b4aa74SDavid Sterba btrfs_release_path(path); 3727c5c9cd4dSSage Weil 372862e2390eSFilipe Manana last_dest_end = ALIGN(new_key.offset + datal, 372962e2390eSFilipe Manana root->sectorsize); 3730f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, 3731f82a9901SFilipe Manana last_dest_end, 37321c919a5eSMark Fasheh destoff, olen, 37331c919a5eSMark Fasheh no_time_update); 3734f82a9901SFilipe Manana if (ret) 373579787eaaSJeff Mahoney goto out; 37362c463823SFilipe Manana if (new_key.offset + datal >= destoff + len) 37372c463823SFilipe Manana break; 3738a22285a6SYan, Zheng } 3739b3b4aa74SDavid Sterba btrfs_release_path(path); 3740df858e76SFilipe Manana key.offset = next_key_min_offset; 3741ae01a0abSYan Zheng } 3742f46b5a66SChristoph Hellwig ret = 0; 374332b7c687SMark Fasheh 3744f82a9901SFilipe Manana if (last_dest_end < destoff + len) { 3745f82a9901SFilipe Manana /* 3746f82a9901SFilipe Manana * We have an implicit hole (NO_HOLES feature is enabled) that 3747f82a9901SFilipe Manana * fully or partially overlaps our cloning range at its end. 3748f82a9901SFilipe Manana */ 3749b3b4aa74SDavid Sterba btrfs_release_path(path); 3750f82a9901SFilipe Manana 3751f82a9901SFilipe Manana /* 3752f82a9901SFilipe Manana * 1 - remove extent(s) 3753f82a9901SFilipe Manana * 1 - inode update 3754f82a9901SFilipe Manana */ 3755f82a9901SFilipe Manana trans = btrfs_start_transaction(root, 2); 3756f82a9901SFilipe Manana if (IS_ERR(trans)) { 3757f82a9901SFilipe Manana ret = PTR_ERR(trans); 3758f82a9901SFilipe Manana goto out; 3759f82a9901SFilipe Manana } 3760f82a9901SFilipe Manana ret = btrfs_drop_extents(trans, root, inode, 3761f82a9901SFilipe Manana last_dest_end, destoff + len, 1); 3762f82a9901SFilipe Manana if (ret) { 3763f82a9901SFilipe Manana if (ret != -EOPNOTSUPP) 3764f82a9901SFilipe Manana btrfs_abort_transaction(trans, root, ret); 3765f82a9901SFilipe Manana btrfs_end_transaction(trans, root); 3766f82a9901SFilipe Manana goto out; 3767f82a9901SFilipe Manana } 376814f59796SFilipe Manana clone_update_extent_map(inode, trans, NULL, last_dest_end, 376914f59796SFilipe Manana destoff + len - last_dest_end); 3770f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, destoff + len, 37711c919a5eSMark Fasheh destoff, olen, no_time_update); 3772f82a9901SFilipe Manana } 3773f82a9901SFilipe Manana 3774f46b5a66SChristoph Hellwig out: 377532b7c687SMark Fasheh btrfs_free_path(path); 377632b7c687SMark Fasheh vfree(buf); 377732b7c687SMark Fasheh return ret; 377832b7c687SMark Fasheh } 377932b7c687SMark Fasheh 378032b7c687SMark Fasheh static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, 378132b7c687SMark Fasheh u64 off, u64 olen, u64 destoff) 378232b7c687SMark Fasheh { 378354563d41SAl Viro struct inode *inode = file_inode(file); 378432b7c687SMark Fasheh struct btrfs_root *root = BTRFS_I(inode)->root; 378532b7c687SMark Fasheh struct fd src_file; 378632b7c687SMark Fasheh struct inode *src; 378732b7c687SMark Fasheh int ret; 378832b7c687SMark Fasheh u64 len = olen; 378932b7c687SMark Fasheh u64 bs = root->fs_info->sb->s_blocksize; 379032b7c687SMark Fasheh int same_inode = 0; 379132b7c687SMark Fasheh 379232b7c687SMark Fasheh /* 379332b7c687SMark Fasheh * TODO: 379432b7c687SMark Fasheh * - split compressed inline extents. annoying: we need to 379532b7c687SMark Fasheh * decompress into destination's address_space (the file offset 379632b7c687SMark Fasheh * may change, so source mapping won't do), then recompress (or 379732b7c687SMark Fasheh * otherwise reinsert) a subrange. 379800fdf13aSLiu Bo * 379900fdf13aSLiu Bo * - split destination inode's inline extents. The inline extents can 380000fdf13aSLiu Bo * be either compressed or non-compressed. 380132b7c687SMark Fasheh */ 380232b7c687SMark Fasheh 380332b7c687SMark Fasheh /* the destination must be opened for writing */ 380432b7c687SMark Fasheh if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND)) 380532b7c687SMark Fasheh return -EINVAL; 380632b7c687SMark Fasheh 380732b7c687SMark Fasheh if (btrfs_root_readonly(root)) 380832b7c687SMark Fasheh return -EROFS; 380932b7c687SMark Fasheh 381032b7c687SMark Fasheh ret = mnt_want_write_file(file); 381132b7c687SMark Fasheh if (ret) 381232b7c687SMark Fasheh return ret; 381332b7c687SMark Fasheh 381432b7c687SMark Fasheh src_file = fdget(srcfd); 381532b7c687SMark Fasheh if (!src_file.file) { 381632b7c687SMark Fasheh ret = -EBADF; 381732b7c687SMark Fasheh goto out_drop_write; 381832b7c687SMark Fasheh } 381932b7c687SMark Fasheh 382032b7c687SMark Fasheh ret = -EXDEV; 382132b7c687SMark Fasheh if (src_file.file->f_path.mnt != file->f_path.mnt) 382232b7c687SMark Fasheh goto out_fput; 382332b7c687SMark Fasheh 382432b7c687SMark Fasheh src = file_inode(src_file.file); 382532b7c687SMark Fasheh 382632b7c687SMark Fasheh ret = -EINVAL; 382732b7c687SMark Fasheh if (src == inode) 382832b7c687SMark Fasheh same_inode = 1; 382932b7c687SMark Fasheh 383032b7c687SMark Fasheh /* the src must be open for reading */ 383132b7c687SMark Fasheh if (!(src_file.file->f_mode & FMODE_READ)) 383232b7c687SMark Fasheh goto out_fput; 383332b7c687SMark Fasheh 383432b7c687SMark Fasheh /* don't make the dst file partly checksummed */ 383532b7c687SMark Fasheh if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != 383632b7c687SMark Fasheh (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) 383732b7c687SMark Fasheh goto out_fput; 383832b7c687SMark Fasheh 383932b7c687SMark Fasheh ret = -EISDIR; 384032b7c687SMark Fasheh if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode)) 384132b7c687SMark Fasheh goto out_fput; 384232b7c687SMark Fasheh 384332b7c687SMark Fasheh ret = -EXDEV; 384432b7c687SMark Fasheh if (src->i_sb != inode->i_sb) 384532b7c687SMark Fasheh goto out_fput; 384632b7c687SMark Fasheh 384732b7c687SMark Fasheh if (!same_inode) { 3848293a8489SMark Fasheh btrfs_double_inode_lock(src, inode); 384932b7c687SMark Fasheh } else { 385032b7c687SMark Fasheh mutex_lock(&src->i_mutex); 385132b7c687SMark Fasheh } 385232b7c687SMark Fasheh 385332b7c687SMark Fasheh /* determine range to clone */ 385432b7c687SMark Fasheh ret = -EINVAL; 385532b7c687SMark Fasheh if (off + len > src->i_size || off + len < off) 385632b7c687SMark Fasheh goto out_unlock; 385732b7c687SMark Fasheh if (len == 0) 385832b7c687SMark Fasheh olen = len = src->i_size - off; 385932b7c687SMark Fasheh /* if we extend to eof, continue to block boundary */ 386032b7c687SMark Fasheh if (off + len == src->i_size) 386132b7c687SMark Fasheh len = ALIGN(src->i_size, bs) - off; 386232b7c687SMark Fasheh 3863ccccf3d6SFilipe Manana if (len == 0) { 3864ccccf3d6SFilipe Manana ret = 0; 3865ccccf3d6SFilipe Manana goto out_unlock; 3866ccccf3d6SFilipe Manana } 3867ccccf3d6SFilipe Manana 386832b7c687SMark Fasheh /* verify the end result is block aligned */ 386932b7c687SMark Fasheh if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) || 387032b7c687SMark Fasheh !IS_ALIGNED(destoff, bs)) 387132b7c687SMark Fasheh goto out_unlock; 387232b7c687SMark Fasheh 387332b7c687SMark Fasheh /* verify if ranges are overlapped within the same file */ 387432b7c687SMark Fasheh if (same_inode) { 387532b7c687SMark Fasheh if (destoff + len > off && destoff < off + len) 387632b7c687SMark Fasheh goto out_unlock; 387732b7c687SMark Fasheh } 387832b7c687SMark Fasheh 387932b7c687SMark Fasheh if (destoff > inode->i_size) { 388032b7c687SMark Fasheh ret = btrfs_cont_expand(inode, inode->i_size, destoff); 388132b7c687SMark Fasheh if (ret) 388232b7c687SMark Fasheh goto out_unlock; 388332b7c687SMark Fasheh } 388432b7c687SMark Fasheh 3885c125b8bfSFilipe Manana /* 3886c125b8bfSFilipe Manana * Lock the target range too. Right after we replace the file extent 3887c125b8bfSFilipe Manana * items in the fs tree (which now point to the cloned data), we might 3888c125b8bfSFilipe Manana * have a worker replace them with extent items relative to a write 3889c125b8bfSFilipe Manana * operation that was issued before this clone operation (i.e. confront 3890c125b8bfSFilipe Manana * with inode.c:btrfs_finish_ordered_io). 3891c125b8bfSFilipe Manana */ 3892c125b8bfSFilipe Manana if (same_inode) { 3893c125b8bfSFilipe Manana u64 lock_start = min_t(u64, off, destoff); 3894c125b8bfSFilipe Manana u64 lock_len = max_t(u64, off, destoff) + len - lock_start; 389532b7c687SMark Fasheh 3896c125b8bfSFilipe Manana lock_extent_range(src, lock_start, lock_len); 3897c125b8bfSFilipe Manana } else { 3898293a8489SMark Fasheh btrfs_double_extent_lock(src, off, inode, destoff, len); 3899c125b8bfSFilipe Manana } 390032b7c687SMark Fasheh 39011c919a5eSMark Fasheh ret = btrfs_clone(src, inode, off, olen, len, destoff, 0); 390232b7c687SMark Fasheh 3903c125b8bfSFilipe Manana if (same_inode) { 3904c125b8bfSFilipe Manana u64 lock_start = min_t(u64, off, destoff); 3905c125b8bfSFilipe Manana u64 lock_end = max_t(u64, off, destoff) + len - 1; 3906c125b8bfSFilipe Manana 3907c125b8bfSFilipe Manana unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end); 3908c125b8bfSFilipe Manana } else { 3909293a8489SMark Fasheh btrfs_double_extent_unlock(src, off, inode, destoff, len); 3910c125b8bfSFilipe Manana } 3911c125b8bfSFilipe Manana /* 3912c125b8bfSFilipe Manana * Truncate page cache pages so that future reads will see the cloned 3913c125b8bfSFilipe Manana * data immediately and not the previous data. 3914c125b8bfSFilipe Manana */ 3915c125b8bfSFilipe Manana truncate_inode_pages_range(&inode->i_data, destoff, 3916c125b8bfSFilipe Manana PAGE_CACHE_ALIGN(destoff + len) - 1); 3917f46b5a66SChristoph Hellwig out_unlock: 3918293a8489SMark Fasheh if (!same_inode) 3919293a8489SMark Fasheh btrfs_double_inode_unlock(src, inode); 3920293a8489SMark Fasheh else 3921f46b5a66SChristoph Hellwig mutex_unlock(&src->i_mutex); 3922f46b5a66SChristoph Hellwig out_fput: 39232903ff01SAl Viro fdput(src_file); 3924ab67b7c1SYan Zheng out_drop_write: 39252a79f17eSAl Viro mnt_drop_write_file(file); 3926f46b5a66SChristoph Hellwig return ret; 3927f46b5a66SChristoph Hellwig } 3928f46b5a66SChristoph Hellwig 39297a865e8aSChristoph Hellwig static long btrfs_ioctl_clone_range(struct file *file, void __user *argp) 3930c5c9cd4dSSage Weil { 3931c5c9cd4dSSage Weil struct btrfs_ioctl_clone_range_args args; 3932c5c9cd4dSSage Weil 39337a865e8aSChristoph Hellwig if (copy_from_user(&args, argp, sizeof(args))) 3934c5c9cd4dSSage Weil return -EFAULT; 3935c5c9cd4dSSage Weil return btrfs_ioctl_clone(file, args.src_fd, args.src_offset, 3936c5c9cd4dSSage Weil args.src_length, args.dest_offset); 3937c5c9cd4dSSage Weil } 3938c5c9cd4dSSage Weil 3939f46b5a66SChristoph Hellwig /* 3940f46b5a66SChristoph Hellwig * there are many ways the trans_start and trans_end ioctls can lead 3941f46b5a66SChristoph Hellwig * to deadlocks. They should only be used by applications that 3942f46b5a66SChristoph Hellwig * basically own the machine, and have a very in depth understanding 3943f46b5a66SChristoph Hellwig * of all the possible deadlocks and enospc problems. 3944f46b5a66SChristoph Hellwig */ 3945b2950863SChristoph Hellwig static long btrfs_ioctl_trans_start(struct file *file) 3946f46b5a66SChristoph Hellwig { 3947496ad9aaSAl Viro struct inode *inode = file_inode(file); 3948f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 3949f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 39501ab86aedSSage Weil int ret; 3951f46b5a66SChristoph Hellwig 39521ab86aedSSage Weil ret = -EPERM; 3953df5b5520SChristoph Hellwig if (!capable(CAP_SYS_ADMIN)) 3954f46b5a66SChristoph Hellwig goto out; 39551ab86aedSSage Weil 39561ab86aedSSage Weil ret = -EINPROGRESS; 39571ab86aedSSage Weil if (file->private_data) 39581ab86aedSSage Weil goto out; 39599ca9ee09SSage Weil 3960b83cc969SLi Zefan ret = -EROFS; 3961b83cc969SLi Zefan if (btrfs_root_readonly(root)) 3962b83cc969SLi Zefan goto out; 3963b83cc969SLi Zefan 3964a561be71SAl Viro ret = mnt_want_write_file(file); 3965c146afadSYan Zheng if (ret) 3966c146afadSYan Zheng goto out; 3967c146afadSYan Zheng 3968a4abeea4SJosef Bacik atomic_inc(&root->fs_info->open_ioctl_trans); 39699ca9ee09SSage Weil 3970f46b5a66SChristoph Hellwig ret = -ENOMEM; 39717a7eaa40SJosef Bacik trans = btrfs_start_ioctl_transaction(root); 3972abd30bb0STsutomu Itoh if (IS_ERR(trans)) 39731ab86aedSSage Weil goto out_drop; 39741ab86aedSSage Weil 39751ab86aedSSage Weil file->private_data = trans; 39761ab86aedSSage Weil return 0; 39771ab86aedSSage Weil 39781ab86aedSSage Weil out_drop: 3979a4abeea4SJosef Bacik atomic_dec(&root->fs_info->open_ioctl_trans); 39802a79f17eSAl Viro mnt_drop_write_file(file); 3981f46b5a66SChristoph Hellwig out: 3982f46b5a66SChristoph Hellwig return ret; 3983f46b5a66SChristoph Hellwig } 3984f46b5a66SChristoph Hellwig 39856ef5ed0dSJosef Bacik static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) 39866ef5ed0dSJosef Bacik { 3987496ad9aaSAl Viro struct inode *inode = file_inode(file); 39886ef5ed0dSJosef Bacik struct btrfs_root *root = BTRFS_I(inode)->root; 39896ef5ed0dSJosef Bacik struct btrfs_root *new_root; 39906ef5ed0dSJosef Bacik struct btrfs_dir_item *di; 39916ef5ed0dSJosef Bacik struct btrfs_trans_handle *trans; 39926ef5ed0dSJosef Bacik struct btrfs_path *path; 39936ef5ed0dSJosef Bacik struct btrfs_key location; 39946ef5ed0dSJosef Bacik struct btrfs_disk_key disk_key; 39956ef5ed0dSJosef Bacik u64 objectid = 0; 39966ef5ed0dSJosef Bacik u64 dir_id; 39973c04ce01SMiao Xie int ret; 39986ef5ed0dSJosef Bacik 39996ef5ed0dSJosef Bacik if (!capable(CAP_SYS_ADMIN)) 40006ef5ed0dSJosef Bacik return -EPERM; 40016ef5ed0dSJosef Bacik 40023c04ce01SMiao Xie ret = mnt_want_write_file(file); 40033c04ce01SMiao Xie if (ret) 40043c04ce01SMiao Xie return ret; 40053c04ce01SMiao Xie 40063c04ce01SMiao Xie if (copy_from_user(&objectid, argp, sizeof(objectid))) { 40073c04ce01SMiao Xie ret = -EFAULT; 40083c04ce01SMiao Xie goto out; 40093c04ce01SMiao Xie } 40106ef5ed0dSJosef Bacik 40116ef5ed0dSJosef Bacik if (!objectid) 40121cecf579Schandan objectid = BTRFS_FS_TREE_OBJECTID; 40136ef5ed0dSJosef Bacik 40146ef5ed0dSJosef Bacik location.objectid = objectid; 40156ef5ed0dSJosef Bacik location.type = BTRFS_ROOT_ITEM_KEY; 40166ef5ed0dSJosef Bacik location.offset = (u64)-1; 40176ef5ed0dSJosef Bacik 40186ef5ed0dSJosef Bacik new_root = btrfs_read_fs_root_no_name(root->fs_info, &location); 40193c04ce01SMiao Xie if (IS_ERR(new_root)) { 40203c04ce01SMiao Xie ret = PTR_ERR(new_root); 40213c04ce01SMiao Xie goto out; 40223c04ce01SMiao Xie } 40236ef5ed0dSJosef Bacik 40246ef5ed0dSJosef Bacik path = btrfs_alloc_path(); 40253c04ce01SMiao Xie if (!path) { 40263c04ce01SMiao Xie ret = -ENOMEM; 40273c04ce01SMiao Xie goto out; 40283c04ce01SMiao Xie } 40296ef5ed0dSJosef Bacik path->leave_spinning = 1; 40306ef5ed0dSJosef Bacik 40316ef5ed0dSJosef Bacik trans = btrfs_start_transaction(root, 1); 403298d5dc13STsutomu Itoh if (IS_ERR(trans)) { 40336ef5ed0dSJosef Bacik btrfs_free_path(path); 40343c04ce01SMiao Xie ret = PTR_ERR(trans); 40353c04ce01SMiao Xie goto out; 40366ef5ed0dSJosef Bacik } 40376ef5ed0dSJosef Bacik 40386c41761fSDavid Sterba dir_id = btrfs_super_root_dir(root->fs_info->super_copy); 40396ef5ed0dSJosef Bacik di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path, 40406ef5ed0dSJosef Bacik dir_id, "default", 7, 1); 4041cf1e99a4SDan Carpenter if (IS_ERR_OR_NULL(di)) { 40426ef5ed0dSJosef Bacik btrfs_free_path(path); 40436ef5ed0dSJosef Bacik btrfs_end_transaction(trans, root); 4044efe120a0SFrank Holton btrfs_err(new_root->fs_info, "Umm, you don't have the default dir" 4045efe120a0SFrank Holton "item, this isn't going to work"); 40463c04ce01SMiao Xie ret = -ENOENT; 40473c04ce01SMiao Xie goto out; 40486ef5ed0dSJosef Bacik } 40496ef5ed0dSJosef Bacik 40506ef5ed0dSJosef Bacik btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); 40516ef5ed0dSJosef Bacik btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); 40526ef5ed0dSJosef Bacik btrfs_mark_buffer_dirty(path->nodes[0]); 40536ef5ed0dSJosef Bacik btrfs_free_path(path); 40546ef5ed0dSJosef Bacik 40552b0ce2c2SMitch Harder btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL); 40566ef5ed0dSJosef Bacik btrfs_end_transaction(trans, root); 40573c04ce01SMiao Xie out: 40583c04ce01SMiao Xie mnt_drop_write_file(file); 40593c04ce01SMiao Xie return ret; 40606ef5ed0dSJosef Bacik } 40616ef5ed0dSJosef Bacik 40625af3e8ccSStefan Behrens void btrfs_get_block_group_info(struct list_head *groups_list, 4063bf5fc093SJosef Bacik struct btrfs_ioctl_space_info *space) 4064bf5fc093SJosef Bacik { 4065bf5fc093SJosef Bacik struct btrfs_block_group_cache *block_group; 4066bf5fc093SJosef Bacik 4067bf5fc093SJosef Bacik space->total_bytes = 0; 4068bf5fc093SJosef Bacik space->used_bytes = 0; 4069bf5fc093SJosef Bacik space->flags = 0; 4070bf5fc093SJosef Bacik list_for_each_entry(block_group, groups_list, list) { 4071bf5fc093SJosef Bacik space->flags = block_group->flags; 4072bf5fc093SJosef Bacik space->total_bytes += block_group->key.offset; 4073bf5fc093SJosef Bacik space->used_bytes += 4074bf5fc093SJosef Bacik btrfs_block_group_used(&block_group->item); 4075bf5fc093SJosef Bacik } 4076bf5fc093SJosef Bacik } 4077bf5fc093SJosef Bacik 407848a3b636SEric Sandeen static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg) 40791406e432SJosef Bacik { 40801406e432SJosef Bacik struct btrfs_ioctl_space_args space_args; 40811406e432SJosef Bacik struct btrfs_ioctl_space_info space; 40821406e432SJosef Bacik struct btrfs_ioctl_space_info *dest; 40837fde62bfSChris Mason struct btrfs_ioctl_space_info *dest_orig; 408413f2696fSDaniel J Blueman struct btrfs_ioctl_space_info __user *user_dest; 40851406e432SJosef Bacik struct btrfs_space_info *info; 4086bf5fc093SJosef Bacik u64 types[] = {BTRFS_BLOCK_GROUP_DATA, 4087bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_SYSTEM, 4088bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_METADATA, 4089bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA}; 4090bf5fc093SJosef Bacik int num_types = 4; 40917fde62bfSChris Mason int alloc_size; 40921406e432SJosef Bacik int ret = 0; 409351788b1bSDan Rosenberg u64 slot_count = 0; 4094bf5fc093SJosef Bacik int i, c; 40951406e432SJosef Bacik 40961406e432SJosef Bacik if (copy_from_user(&space_args, 40971406e432SJosef Bacik (struct btrfs_ioctl_space_args __user *)arg, 40981406e432SJosef Bacik sizeof(space_args))) 40991406e432SJosef Bacik return -EFAULT; 41001406e432SJosef Bacik 4101bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4102bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4103bf5fc093SJosef Bacik 4104bf5fc093SJosef Bacik info = NULL; 41057fde62bfSChris Mason rcu_read_lock(); 4106bf5fc093SJosef Bacik list_for_each_entry_rcu(tmp, &root->fs_info->space_info, 4107bf5fc093SJosef Bacik list) { 4108bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4109bf5fc093SJosef Bacik info = tmp; 4110bf5fc093SJosef Bacik break; 4111bf5fc093SJosef Bacik } 4112bf5fc093SJosef Bacik } 41137fde62bfSChris Mason rcu_read_unlock(); 41141406e432SJosef Bacik 4115bf5fc093SJosef Bacik if (!info) 4116bf5fc093SJosef Bacik continue; 4117bf5fc093SJosef Bacik 4118bf5fc093SJosef Bacik down_read(&info->groups_sem); 4119bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4120bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) 4121bf5fc093SJosef Bacik slot_count++; 4122bf5fc093SJosef Bacik } 4123bf5fc093SJosef Bacik up_read(&info->groups_sem); 4124bf5fc093SJosef Bacik } 4125bf5fc093SJosef Bacik 412636523e95SDavid Sterba /* 412736523e95SDavid Sterba * Global block reserve, exported as a space_info 412836523e95SDavid Sterba */ 412936523e95SDavid Sterba slot_count++; 413036523e95SDavid Sterba 41317fde62bfSChris Mason /* space_slots == 0 means they are asking for a count */ 41327fde62bfSChris Mason if (space_args.space_slots == 0) { 41337fde62bfSChris Mason space_args.total_spaces = slot_count; 41347fde62bfSChris Mason goto out; 41357fde62bfSChris Mason } 4136bf5fc093SJosef Bacik 413751788b1bSDan Rosenberg slot_count = min_t(u64, space_args.space_slots, slot_count); 4138bf5fc093SJosef Bacik 41397fde62bfSChris Mason alloc_size = sizeof(*dest) * slot_count; 4140bf5fc093SJosef Bacik 41417fde62bfSChris Mason /* we generally have at most 6 or so space infos, one for each raid 41427fde62bfSChris Mason * level. So, a whole page should be more than enough for everyone 41437fde62bfSChris Mason */ 41447fde62bfSChris Mason if (alloc_size > PAGE_CACHE_SIZE) 41457fde62bfSChris Mason return -ENOMEM; 41467fde62bfSChris Mason 41477fde62bfSChris Mason space_args.total_spaces = 0; 41487fde62bfSChris Mason dest = kmalloc(alloc_size, GFP_NOFS); 41497fde62bfSChris Mason if (!dest) 41507fde62bfSChris Mason return -ENOMEM; 41517fde62bfSChris Mason dest_orig = dest; 41527fde62bfSChris Mason 41537fde62bfSChris Mason /* now we have a buffer to copy into */ 4154bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4155bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4156bf5fc093SJosef Bacik 415751788b1bSDan Rosenberg if (!slot_count) 415851788b1bSDan Rosenberg break; 415951788b1bSDan Rosenberg 4160bf5fc093SJosef Bacik info = NULL; 41611406e432SJosef Bacik rcu_read_lock(); 4162bf5fc093SJosef Bacik list_for_each_entry_rcu(tmp, &root->fs_info->space_info, 4163bf5fc093SJosef Bacik list) { 4164bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4165bf5fc093SJosef Bacik info = tmp; 41667fde62bfSChris Mason break; 4167bf5fc093SJosef Bacik } 4168bf5fc093SJosef Bacik } 4169bf5fc093SJosef Bacik rcu_read_unlock(); 41707fde62bfSChris Mason 4171bf5fc093SJosef Bacik if (!info) 4172bf5fc093SJosef Bacik continue; 4173bf5fc093SJosef Bacik down_read(&info->groups_sem); 4174bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4175bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) { 41765af3e8ccSStefan Behrens btrfs_get_block_group_info( 41775af3e8ccSStefan Behrens &info->block_groups[c], &space); 41787fde62bfSChris Mason memcpy(dest, &space, sizeof(space)); 41791406e432SJosef Bacik dest++; 41801406e432SJosef Bacik space_args.total_spaces++; 418151788b1bSDan Rosenberg slot_count--; 41821406e432SJosef Bacik } 418351788b1bSDan Rosenberg if (!slot_count) 418451788b1bSDan Rosenberg break; 4185bf5fc093SJosef Bacik } 4186bf5fc093SJosef Bacik up_read(&info->groups_sem); 4187bf5fc093SJosef Bacik } 41881406e432SJosef Bacik 418936523e95SDavid Sterba /* 419036523e95SDavid Sterba * Add global block reserve 419136523e95SDavid Sterba */ 419236523e95SDavid Sterba if (slot_count) { 419336523e95SDavid Sterba struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv; 419436523e95SDavid Sterba 419536523e95SDavid Sterba spin_lock(&block_rsv->lock); 419636523e95SDavid Sterba space.total_bytes = block_rsv->size; 419736523e95SDavid Sterba space.used_bytes = block_rsv->size - block_rsv->reserved; 419836523e95SDavid Sterba spin_unlock(&block_rsv->lock); 419936523e95SDavid Sterba space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV; 420036523e95SDavid Sterba memcpy(dest, &space, sizeof(space)); 420136523e95SDavid Sterba space_args.total_spaces++; 420236523e95SDavid Sterba } 420336523e95SDavid Sterba 42042eec6c81SDaniel J Blueman user_dest = (struct btrfs_ioctl_space_info __user *) 42057fde62bfSChris Mason (arg + sizeof(struct btrfs_ioctl_space_args)); 42067fde62bfSChris Mason 42077fde62bfSChris Mason if (copy_to_user(user_dest, dest_orig, alloc_size)) 42087fde62bfSChris Mason ret = -EFAULT; 42097fde62bfSChris Mason 42107fde62bfSChris Mason kfree(dest_orig); 42117fde62bfSChris Mason out: 42127fde62bfSChris Mason if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) 42131406e432SJosef Bacik ret = -EFAULT; 42141406e432SJosef Bacik 42151406e432SJosef Bacik return ret; 42161406e432SJosef Bacik } 42171406e432SJosef Bacik 4218f46b5a66SChristoph Hellwig /* 4219f46b5a66SChristoph Hellwig * there are many ways the trans_start and trans_end ioctls can lead 4220f46b5a66SChristoph Hellwig * to deadlocks. They should only be used by applications that 4221f46b5a66SChristoph Hellwig * basically own the machine, and have a very in depth understanding 4222f46b5a66SChristoph Hellwig * of all the possible deadlocks and enospc problems. 4223f46b5a66SChristoph Hellwig */ 4224f46b5a66SChristoph Hellwig long btrfs_ioctl_trans_end(struct file *file) 4225f46b5a66SChristoph Hellwig { 4226496ad9aaSAl Viro struct inode *inode = file_inode(file); 4227f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 4228f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 4229f46b5a66SChristoph Hellwig 4230f46b5a66SChristoph Hellwig trans = file->private_data; 42311ab86aedSSage Weil if (!trans) 42321ab86aedSSage Weil return -EINVAL; 4233b214107eSChristoph Hellwig file->private_data = NULL; 42349ca9ee09SSage Weil 42351ab86aedSSage Weil btrfs_end_transaction(trans, root); 42361ab86aedSSage Weil 4237a4abeea4SJosef Bacik atomic_dec(&root->fs_info->open_ioctl_trans); 42389ca9ee09SSage Weil 42392a79f17eSAl Viro mnt_drop_write_file(file); 42401ab86aedSSage Weil return 0; 4241f46b5a66SChristoph Hellwig } 4242f46b5a66SChristoph Hellwig 42439a8c28beSMiao Xie static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, 42449a8c28beSMiao Xie void __user *argp) 424546204592SSage Weil { 424646204592SSage Weil struct btrfs_trans_handle *trans; 424746204592SSage Weil u64 transid; 4248db5b493aSTsutomu Itoh int ret; 424946204592SSage Weil 4250d4edf39bSMiao Xie trans = btrfs_attach_transaction_barrier(root); 4251ff7c1d33SMiao Xie if (IS_ERR(trans)) { 4252ff7c1d33SMiao Xie if (PTR_ERR(trans) != -ENOENT) 425398d5dc13STsutomu Itoh return PTR_ERR(trans); 4254ff7c1d33SMiao Xie 4255ff7c1d33SMiao Xie /* No running transaction, don't bother */ 4256ff7c1d33SMiao Xie transid = root->fs_info->last_trans_committed; 4257ff7c1d33SMiao Xie goto out; 4258ff7c1d33SMiao Xie } 425946204592SSage Weil transid = trans->transid; 4260db5b493aSTsutomu Itoh ret = btrfs_commit_transaction_async(trans, root, 0); 42618b2b2d3cSTsutomu Itoh if (ret) { 42628b2b2d3cSTsutomu Itoh btrfs_end_transaction(trans, root); 4263db5b493aSTsutomu Itoh return ret; 42648b2b2d3cSTsutomu Itoh } 4265ff7c1d33SMiao Xie out: 426646204592SSage Weil if (argp) 426746204592SSage Weil if (copy_to_user(argp, &transid, sizeof(transid))) 426846204592SSage Weil return -EFAULT; 426946204592SSage Weil return 0; 427046204592SSage Weil } 427146204592SSage Weil 42729a8c28beSMiao Xie static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root, 42739a8c28beSMiao Xie void __user *argp) 427446204592SSage Weil { 427546204592SSage Weil u64 transid; 427646204592SSage Weil 427746204592SSage Weil if (argp) { 427846204592SSage Weil if (copy_from_user(&transid, argp, sizeof(transid))) 427946204592SSage Weil return -EFAULT; 428046204592SSage Weil } else { 428146204592SSage Weil transid = 0; /* current trans */ 428246204592SSage Weil } 428346204592SSage Weil return btrfs_wait_for_commit(root, transid); 428446204592SSage Weil } 428546204592SSage Weil 4286b8e95489SMiao Xie static long btrfs_ioctl_scrub(struct file *file, void __user *arg) 4287475f6387SJan Schmidt { 4288496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 4289475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4290b8e95489SMiao Xie int ret; 4291475f6387SJan Schmidt 4292475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4293475f6387SJan Schmidt return -EPERM; 4294475f6387SJan Schmidt 4295475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4296475f6387SJan Schmidt if (IS_ERR(sa)) 4297475f6387SJan Schmidt return PTR_ERR(sa); 4298475f6387SJan Schmidt 4299b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) { 4300b8e95489SMiao Xie ret = mnt_want_write_file(file); 4301b8e95489SMiao Xie if (ret) 4302b8e95489SMiao Xie goto out; 4303b8e95489SMiao Xie } 4304b8e95489SMiao Xie 4305aa1b8cd4SStefan Behrens ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end, 430663a212abSStefan Behrens &sa->progress, sa->flags & BTRFS_SCRUB_READONLY, 430763a212abSStefan Behrens 0); 4308475f6387SJan Schmidt 4309475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 4310475f6387SJan Schmidt ret = -EFAULT; 4311475f6387SJan Schmidt 4312b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) 4313b8e95489SMiao Xie mnt_drop_write_file(file); 4314b8e95489SMiao Xie out: 4315475f6387SJan Schmidt kfree(sa); 4316475f6387SJan Schmidt return ret; 4317475f6387SJan Schmidt } 4318475f6387SJan Schmidt 4319475f6387SJan Schmidt static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg) 4320475f6387SJan Schmidt { 4321475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4322475f6387SJan Schmidt return -EPERM; 4323475f6387SJan Schmidt 4324aa1b8cd4SStefan Behrens return btrfs_scrub_cancel(root->fs_info); 4325475f6387SJan Schmidt } 4326475f6387SJan Schmidt 4327475f6387SJan Schmidt static long btrfs_ioctl_scrub_progress(struct btrfs_root *root, 4328475f6387SJan Schmidt void __user *arg) 4329475f6387SJan Schmidt { 4330475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4331475f6387SJan Schmidt int ret; 4332475f6387SJan Schmidt 4333475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4334475f6387SJan Schmidt return -EPERM; 4335475f6387SJan Schmidt 4336475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4337475f6387SJan Schmidt if (IS_ERR(sa)) 4338475f6387SJan Schmidt return PTR_ERR(sa); 4339475f6387SJan Schmidt 4340475f6387SJan Schmidt ret = btrfs_scrub_progress(root, sa->devid, &sa->progress); 4341475f6387SJan Schmidt 4342475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 4343475f6387SJan Schmidt ret = -EFAULT; 4344475f6387SJan Schmidt 4345475f6387SJan Schmidt kfree(sa); 4346475f6387SJan Schmidt return ret; 4347475f6387SJan Schmidt } 4348475f6387SJan Schmidt 4349c11d2c23SStefan Behrens static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root, 4350b27f7c0cSDavid Sterba void __user *arg) 4351c11d2c23SStefan Behrens { 4352c11d2c23SStefan Behrens struct btrfs_ioctl_get_dev_stats *sa; 4353c11d2c23SStefan Behrens int ret; 4354c11d2c23SStefan Behrens 4355c11d2c23SStefan Behrens sa = memdup_user(arg, sizeof(*sa)); 4356c11d2c23SStefan Behrens if (IS_ERR(sa)) 4357c11d2c23SStefan Behrens return PTR_ERR(sa); 4358c11d2c23SStefan Behrens 4359b27f7c0cSDavid Sterba if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) { 4360b27f7c0cSDavid Sterba kfree(sa); 4361b27f7c0cSDavid Sterba return -EPERM; 4362b27f7c0cSDavid Sterba } 4363b27f7c0cSDavid Sterba 4364b27f7c0cSDavid Sterba ret = btrfs_get_dev_stats(root, sa); 4365c11d2c23SStefan Behrens 4366c11d2c23SStefan Behrens if (copy_to_user(arg, sa, sizeof(*sa))) 4367c11d2c23SStefan Behrens ret = -EFAULT; 4368c11d2c23SStefan Behrens 4369c11d2c23SStefan Behrens kfree(sa); 4370c11d2c23SStefan Behrens return ret; 4371c11d2c23SStefan Behrens } 4372c11d2c23SStefan Behrens 43733f6bcfbdSStefan Behrens static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg) 43743f6bcfbdSStefan Behrens { 43753f6bcfbdSStefan Behrens struct btrfs_ioctl_dev_replace_args *p; 43763f6bcfbdSStefan Behrens int ret; 43773f6bcfbdSStefan Behrens 43783f6bcfbdSStefan Behrens if (!capable(CAP_SYS_ADMIN)) 43793f6bcfbdSStefan Behrens return -EPERM; 43803f6bcfbdSStefan Behrens 43813f6bcfbdSStefan Behrens p = memdup_user(arg, sizeof(*p)); 43823f6bcfbdSStefan Behrens if (IS_ERR(p)) 43833f6bcfbdSStefan Behrens return PTR_ERR(p); 43843f6bcfbdSStefan Behrens 43853f6bcfbdSStefan Behrens switch (p->cmd) { 43863f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_START: 4387adfa97cbSIlya Dryomov if (root->fs_info->sb->s_flags & MS_RDONLY) { 4388adfa97cbSIlya Dryomov ret = -EROFS; 4389adfa97cbSIlya Dryomov goto out; 4390adfa97cbSIlya Dryomov } 43913f6bcfbdSStefan Behrens if (atomic_xchg( 43923f6bcfbdSStefan Behrens &root->fs_info->mutually_exclusive_operation_running, 43933f6bcfbdSStefan Behrens 1)) { 4394e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 43953f6bcfbdSStefan Behrens } else { 43963f6bcfbdSStefan Behrens ret = btrfs_dev_replace_start(root, p); 43973f6bcfbdSStefan Behrens atomic_set( 43983f6bcfbdSStefan Behrens &root->fs_info->mutually_exclusive_operation_running, 43993f6bcfbdSStefan Behrens 0); 44003f6bcfbdSStefan Behrens } 44013f6bcfbdSStefan Behrens break; 44023f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS: 44033f6bcfbdSStefan Behrens btrfs_dev_replace_status(root->fs_info, p); 44043f6bcfbdSStefan Behrens ret = 0; 44053f6bcfbdSStefan Behrens break; 44063f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL: 44073f6bcfbdSStefan Behrens ret = btrfs_dev_replace_cancel(root->fs_info, p); 44083f6bcfbdSStefan Behrens break; 44093f6bcfbdSStefan Behrens default: 44103f6bcfbdSStefan Behrens ret = -EINVAL; 44113f6bcfbdSStefan Behrens break; 44123f6bcfbdSStefan Behrens } 44133f6bcfbdSStefan Behrens 44143f6bcfbdSStefan Behrens if (copy_to_user(arg, p, sizeof(*p))) 44153f6bcfbdSStefan Behrens ret = -EFAULT; 4416adfa97cbSIlya Dryomov out: 44173f6bcfbdSStefan Behrens kfree(p); 44183f6bcfbdSStefan Behrens return ret; 44193f6bcfbdSStefan Behrens } 44203f6bcfbdSStefan Behrens 4421d7728c96SJan Schmidt static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) 4422d7728c96SJan Schmidt { 4423d7728c96SJan Schmidt int ret = 0; 4424d7728c96SJan Schmidt int i; 4425740c3d22SChris Mason u64 rel_ptr; 4426d7728c96SJan Schmidt int size; 4427806468f8SChris Mason struct btrfs_ioctl_ino_path_args *ipa = NULL; 4428d7728c96SJan Schmidt struct inode_fs_paths *ipath = NULL; 4429d7728c96SJan Schmidt struct btrfs_path *path; 4430d7728c96SJan Schmidt 443182b22ac8SKusanagi Kouichi if (!capable(CAP_DAC_READ_SEARCH)) 4432d7728c96SJan Schmidt return -EPERM; 4433d7728c96SJan Schmidt 4434d7728c96SJan Schmidt path = btrfs_alloc_path(); 4435d7728c96SJan Schmidt if (!path) { 4436d7728c96SJan Schmidt ret = -ENOMEM; 4437d7728c96SJan Schmidt goto out; 4438d7728c96SJan Schmidt } 4439d7728c96SJan Schmidt 4440d7728c96SJan Schmidt ipa = memdup_user(arg, sizeof(*ipa)); 4441d7728c96SJan Schmidt if (IS_ERR(ipa)) { 4442d7728c96SJan Schmidt ret = PTR_ERR(ipa); 4443d7728c96SJan Schmidt ipa = NULL; 4444d7728c96SJan Schmidt goto out; 4445d7728c96SJan Schmidt } 4446d7728c96SJan Schmidt 4447d7728c96SJan Schmidt size = min_t(u32, ipa->size, 4096); 4448d7728c96SJan Schmidt ipath = init_ipath(size, root, path); 4449d7728c96SJan Schmidt if (IS_ERR(ipath)) { 4450d7728c96SJan Schmidt ret = PTR_ERR(ipath); 4451d7728c96SJan Schmidt ipath = NULL; 4452d7728c96SJan Schmidt goto out; 4453d7728c96SJan Schmidt } 4454d7728c96SJan Schmidt 4455d7728c96SJan Schmidt ret = paths_from_inode(ipa->inum, ipath); 4456d7728c96SJan Schmidt if (ret < 0) 4457d7728c96SJan Schmidt goto out; 4458d7728c96SJan Schmidt 4459d7728c96SJan Schmidt for (i = 0; i < ipath->fspath->elem_cnt; ++i) { 4460745c4d8eSJeff Mahoney rel_ptr = ipath->fspath->val[i] - 4461745c4d8eSJeff Mahoney (u64)(unsigned long)ipath->fspath->val; 4462740c3d22SChris Mason ipath->fspath->val[i] = rel_ptr; 4463d7728c96SJan Schmidt } 4464d7728c96SJan Schmidt 4465745c4d8eSJeff Mahoney ret = copy_to_user((void *)(unsigned long)ipa->fspath, 4466745c4d8eSJeff Mahoney (void *)(unsigned long)ipath->fspath, size); 4467d7728c96SJan Schmidt if (ret) { 4468d7728c96SJan Schmidt ret = -EFAULT; 4469d7728c96SJan Schmidt goto out; 4470d7728c96SJan Schmidt } 4471d7728c96SJan Schmidt 4472d7728c96SJan Schmidt out: 4473d7728c96SJan Schmidt btrfs_free_path(path); 4474d7728c96SJan Schmidt free_ipath(ipath); 4475d7728c96SJan Schmidt kfree(ipa); 4476d7728c96SJan Schmidt 4477d7728c96SJan Schmidt return ret; 4478d7728c96SJan Schmidt } 4479d7728c96SJan Schmidt 4480d7728c96SJan Schmidt static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx) 4481d7728c96SJan Schmidt { 4482d7728c96SJan Schmidt struct btrfs_data_container *inodes = ctx; 4483d7728c96SJan Schmidt const size_t c = 3 * sizeof(u64); 4484d7728c96SJan Schmidt 4485d7728c96SJan Schmidt if (inodes->bytes_left >= c) { 4486d7728c96SJan Schmidt inodes->bytes_left -= c; 4487d7728c96SJan Schmidt inodes->val[inodes->elem_cnt] = inum; 4488d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 1] = offset; 4489d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 2] = root; 4490d7728c96SJan Schmidt inodes->elem_cnt += 3; 4491d7728c96SJan Schmidt } else { 4492d7728c96SJan Schmidt inodes->bytes_missing += c - inodes->bytes_left; 4493d7728c96SJan Schmidt inodes->bytes_left = 0; 4494d7728c96SJan Schmidt inodes->elem_missed += 3; 4495d7728c96SJan Schmidt } 4496d7728c96SJan Schmidt 4497d7728c96SJan Schmidt return 0; 4498d7728c96SJan Schmidt } 4499d7728c96SJan Schmidt 4500d7728c96SJan Schmidt static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root, 4501d7728c96SJan Schmidt void __user *arg) 4502d7728c96SJan Schmidt { 4503d7728c96SJan Schmidt int ret = 0; 4504d7728c96SJan Schmidt int size; 4505d7728c96SJan Schmidt struct btrfs_ioctl_logical_ino_args *loi; 4506d7728c96SJan Schmidt struct btrfs_data_container *inodes = NULL; 4507d7728c96SJan Schmidt struct btrfs_path *path = NULL; 4508d7728c96SJan Schmidt 4509d7728c96SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4510d7728c96SJan Schmidt return -EPERM; 4511d7728c96SJan Schmidt 4512d7728c96SJan Schmidt loi = memdup_user(arg, sizeof(*loi)); 4513d7728c96SJan Schmidt if (IS_ERR(loi)) { 4514d7728c96SJan Schmidt ret = PTR_ERR(loi); 4515d7728c96SJan Schmidt loi = NULL; 4516d7728c96SJan Schmidt goto out; 4517d7728c96SJan Schmidt } 4518d7728c96SJan Schmidt 4519d7728c96SJan Schmidt path = btrfs_alloc_path(); 4520d7728c96SJan Schmidt if (!path) { 4521d7728c96SJan Schmidt ret = -ENOMEM; 4522d7728c96SJan Schmidt goto out; 4523d7728c96SJan Schmidt } 4524d7728c96SJan Schmidt 4525425d17a2SLiu Bo size = min_t(u32, loi->size, 64 * 1024); 4526d7728c96SJan Schmidt inodes = init_data_container(size); 4527d7728c96SJan Schmidt if (IS_ERR(inodes)) { 4528d7728c96SJan Schmidt ret = PTR_ERR(inodes); 4529d7728c96SJan Schmidt inodes = NULL; 4530d7728c96SJan Schmidt goto out; 4531d7728c96SJan Schmidt } 4532d7728c96SJan Schmidt 4533df031f07SLiu Bo ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path, 4534df031f07SLiu Bo build_ino_list, inodes); 4535df031f07SLiu Bo if (ret == -EINVAL) 4536d7728c96SJan Schmidt ret = -ENOENT; 4537d7728c96SJan Schmidt if (ret < 0) 4538d7728c96SJan Schmidt goto out; 4539d7728c96SJan Schmidt 4540745c4d8eSJeff Mahoney ret = copy_to_user((void *)(unsigned long)loi->inodes, 4541745c4d8eSJeff Mahoney (void *)(unsigned long)inodes, size); 4542d7728c96SJan Schmidt if (ret) 4543d7728c96SJan Schmidt ret = -EFAULT; 4544d7728c96SJan Schmidt 4545d7728c96SJan Schmidt out: 4546d7728c96SJan Schmidt btrfs_free_path(path); 4547425d17a2SLiu Bo vfree(inodes); 4548d7728c96SJan Schmidt kfree(loi); 4549d7728c96SJan Schmidt 4550d7728c96SJan Schmidt return ret; 4551d7728c96SJan Schmidt } 4552d7728c96SJan Schmidt 455319a39dceSIlya Dryomov void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock, 4554c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs) 4555c9e9f97bSIlya Dryomov { 4556c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl = fs_info->balance_ctl; 4557c9e9f97bSIlya Dryomov 4558c9e9f97bSIlya Dryomov bargs->flags = bctl->flags; 4559c9e9f97bSIlya Dryomov 4560837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_running)) 4561837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_RUNNING; 4562837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_pause_req)) 4563837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; 4564a7e99c69SIlya Dryomov if (atomic_read(&fs_info->balance_cancel_req)) 4565a7e99c69SIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; 4566837d5b6eSIlya Dryomov 4567c9e9f97bSIlya Dryomov memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); 4568c9e9f97bSIlya Dryomov memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); 4569c9e9f97bSIlya Dryomov memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); 457019a39dceSIlya Dryomov 457119a39dceSIlya Dryomov if (lock) { 457219a39dceSIlya Dryomov spin_lock(&fs_info->balance_lock); 457319a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 457419a39dceSIlya Dryomov spin_unlock(&fs_info->balance_lock); 457519a39dceSIlya Dryomov } else { 457619a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 457719a39dceSIlya Dryomov } 4578c9e9f97bSIlya Dryomov } 4579c9e9f97bSIlya Dryomov 45809ba1f6e4SLiu Bo static long btrfs_ioctl_balance(struct file *file, void __user *arg) 4581c9e9f97bSIlya Dryomov { 4582496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 4583c9e9f97bSIlya Dryomov struct btrfs_fs_info *fs_info = root->fs_info; 4584c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 4585c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl; 4586ed0fb78fSIlya Dryomov bool need_unlock; /* for mut. excl. ops lock */ 4587c9e9f97bSIlya Dryomov int ret; 4588c9e9f97bSIlya Dryomov 4589c9e9f97bSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4590c9e9f97bSIlya Dryomov return -EPERM; 4591c9e9f97bSIlya Dryomov 4592e54bfa31SLiu Bo ret = mnt_want_write_file(file); 45939ba1f6e4SLiu Bo if (ret) 45949ba1f6e4SLiu Bo return ret; 45959ba1f6e4SLiu Bo 4596ed0fb78fSIlya Dryomov again: 4597ed0fb78fSIlya Dryomov if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) { 4598c9e9f97bSIlya Dryomov mutex_lock(&fs_info->volume_mutex); 4599c9e9f97bSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4600ed0fb78fSIlya Dryomov need_unlock = true; 4601ed0fb78fSIlya Dryomov goto locked; 4602ed0fb78fSIlya Dryomov } 4603ed0fb78fSIlya Dryomov 4604ed0fb78fSIlya Dryomov /* 4605ed0fb78fSIlya Dryomov * mut. excl. ops lock is locked. Three possibilites: 4606ed0fb78fSIlya Dryomov * (1) some other op is running 4607ed0fb78fSIlya Dryomov * (2) balance is running 4608ed0fb78fSIlya Dryomov * (3) balance is paused -- special case (think resume) 4609ed0fb78fSIlya Dryomov */ 4610ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4611ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4612ed0fb78fSIlya Dryomov /* this is either (2) or (3) */ 4613ed0fb78fSIlya Dryomov if (!atomic_read(&fs_info->balance_running)) { 4614ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4615ed0fb78fSIlya Dryomov if (!mutex_trylock(&fs_info->volume_mutex)) 4616ed0fb78fSIlya Dryomov goto again; 4617ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4618ed0fb78fSIlya Dryomov 4619ed0fb78fSIlya Dryomov if (fs_info->balance_ctl && 4620ed0fb78fSIlya Dryomov !atomic_read(&fs_info->balance_running)) { 4621ed0fb78fSIlya Dryomov /* this is (3) */ 4622ed0fb78fSIlya Dryomov need_unlock = false; 4623ed0fb78fSIlya Dryomov goto locked; 4624ed0fb78fSIlya Dryomov } 4625ed0fb78fSIlya Dryomov 4626ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4627ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->volume_mutex); 4628ed0fb78fSIlya Dryomov goto again; 4629ed0fb78fSIlya Dryomov } else { 4630ed0fb78fSIlya Dryomov /* this is (2) */ 4631ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4632ed0fb78fSIlya Dryomov ret = -EINPROGRESS; 4633ed0fb78fSIlya Dryomov goto out; 4634ed0fb78fSIlya Dryomov } 4635ed0fb78fSIlya Dryomov } else { 4636ed0fb78fSIlya Dryomov /* this is (1) */ 4637ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4638e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 4639ed0fb78fSIlya Dryomov goto out; 4640ed0fb78fSIlya Dryomov } 4641ed0fb78fSIlya Dryomov 4642ed0fb78fSIlya Dryomov locked: 4643ed0fb78fSIlya Dryomov BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running)); 4644c9e9f97bSIlya Dryomov 4645c9e9f97bSIlya Dryomov if (arg) { 4646c9e9f97bSIlya Dryomov bargs = memdup_user(arg, sizeof(*bargs)); 4647c9e9f97bSIlya Dryomov if (IS_ERR(bargs)) { 4648c9e9f97bSIlya Dryomov ret = PTR_ERR(bargs); 4649ed0fb78fSIlya Dryomov goto out_unlock; 4650c9e9f97bSIlya Dryomov } 4651de322263SIlya Dryomov 4652de322263SIlya Dryomov if (bargs->flags & BTRFS_BALANCE_RESUME) { 4653de322263SIlya Dryomov if (!fs_info->balance_ctl) { 4654de322263SIlya Dryomov ret = -ENOTCONN; 4655de322263SIlya Dryomov goto out_bargs; 4656de322263SIlya Dryomov } 4657de322263SIlya Dryomov 4658de322263SIlya Dryomov bctl = fs_info->balance_ctl; 4659de322263SIlya Dryomov spin_lock(&fs_info->balance_lock); 4660de322263SIlya Dryomov bctl->flags |= BTRFS_BALANCE_RESUME; 4661de322263SIlya Dryomov spin_unlock(&fs_info->balance_lock); 4662de322263SIlya Dryomov 4663de322263SIlya Dryomov goto do_balance; 4664de322263SIlya Dryomov } 4665c9e9f97bSIlya Dryomov } else { 4666c9e9f97bSIlya Dryomov bargs = NULL; 4667c9e9f97bSIlya Dryomov } 4668c9e9f97bSIlya Dryomov 4669ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4670837d5b6eSIlya Dryomov ret = -EINPROGRESS; 4671837d5b6eSIlya Dryomov goto out_bargs; 4672837d5b6eSIlya Dryomov } 4673837d5b6eSIlya Dryomov 4674c9e9f97bSIlya Dryomov bctl = kzalloc(sizeof(*bctl), GFP_NOFS); 4675c9e9f97bSIlya Dryomov if (!bctl) { 4676c9e9f97bSIlya Dryomov ret = -ENOMEM; 4677c9e9f97bSIlya Dryomov goto out_bargs; 4678c9e9f97bSIlya Dryomov } 4679c9e9f97bSIlya Dryomov 4680c9e9f97bSIlya Dryomov bctl->fs_info = fs_info; 4681c9e9f97bSIlya Dryomov if (arg) { 4682c9e9f97bSIlya Dryomov memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); 4683c9e9f97bSIlya Dryomov memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); 4684c9e9f97bSIlya Dryomov memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); 4685c9e9f97bSIlya Dryomov 4686c9e9f97bSIlya Dryomov bctl->flags = bargs->flags; 4687f43ffb60SIlya Dryomov } else { 4688f43ffb60SIlya Dryomov /* balance everything - no filters */ 4689f43ffb60SIlya Dryomov bctl->flags |= BTRFS_BALANCE_TYPE_MASK; 4690c9e9f97bSIlya Dryomov } 4691c9e9f97bSIlya Dryomov 46928eb93459SDavid Sterba if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) { 46938eb93459SDavid Sterba ret = -EINVAL; 46940f89abf5SChristian Engelmayer goto out_bctl; 46958eb93459SDavid Sterba } 46968eb93459SDavid Sterba 4697de322263SIlya Dryomov do_balance: 4698c9e9f97bSIlya Dryomov /* 4699ed0fb78fSIlya Dryomov * Ownership of bctl and mutually_exclusive_operation_running 4700ed0fb78fSIlya Dryomov * goes to to btrfs_balance. bctl is freed in __cancel_balance, 4701ed0fb78fSIlya Dryomov * or, if restriper was paused all the way until unmount, in 4702ed0fb78fSIlya Dryomov * free_fs_info. mutually_exclusive_operation_running is 4703ed0fb78fSIlya Dryomov * cleared in __cancel_balance. 4704c9e9f97bSIlya Dryomov */ 4705ed0fb78fSIlya Dryomov need_unlock = false; 4706ed0fb78fSIlya Dryomov 4707ed0fb78fSIlya Dryomov ret = btrfs_balance(bctl, bargs); 47080f89abf5SChristian Engelmayer bctl = NULL; 4709ed0fb78fSIlya Dryomov 4710c9e9f97bSIlya Dryomov if (arg) { 4711c9e9f97bSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 4712c9e9f97bSIlya Dryomov ret = -EFAULT; 4713c9e9f97bSIlya Dryomov } 4714c9e9f97bSIlya Dryomov 47150f89abf5SChristian Engelmayer out_bctl: 47160f89abf5SChristian Engelmayer kfree(bctl); 4717c9e9f97bSIlya Dryomov out_bargs: 4718c9e9f97bSIlya Dryomov kfree(bargs); 4719ed0fb78fSIlya Dryomov out_unlock: 4720c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4721c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->volume_mutex); 4722ed0fb78fSIlya Dryomov if (need_unlock) 4723ed0fb78fSIlya Dryomov atomic_set(&fs_info->mutually_exclusive_operation_running, 0); 4724ed0fb78fSIlya Dryomov out: 4725e54bfa31SLiu Bo mnt_drop_write_file(file); 4726c9e9f97bSIlya Dryomov return ret; 4727c9e9f97bSIlya Dryomov } 4728c9e9f97bSIlya Dryomov 4729837d5b6eSIlya Dryomov static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd) 4730837d5b6eSIlya Dryomov { 4731837d5b6eSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4732837d5b6eSIlya Dryomov return -EPERM; 4733837d5b6eSIlya Dryomov 4734837d5b6eSIlya Dryomov switch (cmd) { 4735837d5b6eSIlya Dryomov case BTRFS_BALANCE_CTL_PAUSE: 4736837d5b6eSIlya Dryomov return btrfs_pause_balance(root->fs_info); 4737a7e99c69SIlya Dryomov case BTRFS_BALANCE_CTL_CANCEL: 4738a7e99c69SIlya Dryomov return btrfs_cancel_balance(root->fs_info); 4739837d5b6eSIlya Dryomov } 4740837d5b6eSIlya Dryomov 4741837d5b6eSIlya Dryomov return -EINVAL; 4742837d5b6eSIlya Dryomov } 4743837d5b6eSIlya Dryomov 474419a39dceSIlya Dryomov static long btrfs_ioctl_balance_progress(struct btrfs_root *root, 474519a39dceSIlya Dryomov void __user *arg) 474619a39dceSIlya Dryomov { 474719a39dceSIlya Dryomov struct btrfs_fs_info *fs_info = root->fs_info; 474819a39dceSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 474919a39dceSIlya Dryomov int ret = 0; 475019a39dceSIlya Dryomov 475119a39dceSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 475219a39dceSIlya Dryomov return -EPERM; 475319a39dceSIlya Dryomov 475419a39dceSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 475519a39dceSIlya Dryomov if (!fs_info->balance_ctl) { 475619a39dceSIlya Dryomov ret = -ENOTCONN; 475719a39dceSIlya Dryomov goto out; 475819a39dceSIlya Dryomov } 475919a39dceSIlya Dryomov 476019a39dceSIlya Dryomov bargs = kzalloc(sizeof(*bargs), GFP_NOFS); 476119a39dceSIlya Dryomov if (!bargs) { 476219a39dceSIlya Dryomov ret = -ENOMEM; 476319a39dceSIlya Dryomov goto out; 476419a39dceSIlya Dryomov } 476519a39dceSIlya Dryomov 476619a39dceSIlya Dryomov update_ioctl_balance_args(fs_info, 1, bargs); 476719a39dceSIlya Dryomov 476819a39dceSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 476919a39dceSIlya Dryomov ret = -EFAULT; 477019a39dceSIlya Dryomov 477119a39dceSIlya Dryomov kfree(bargs); 477219a39dceSIlya Dryomov out: 477319a39dceSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 477419a39dceSIlya Dryomov return ret; 477519a39dceSIlya Dryomov } 477619a39dceSIlya Dryomov 4777905b0ddaSMiao Xie static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) 47785d13a37bSArne Jansen { 4779496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 47805d13a37bSArne Jansen struct btrfs_ioctl_quota_ctl_args *sa; 47815d13a37bSArne Jansen struct btrfs_trans_handle *trans = NULL; 47825d13a37bSArne Jansen int ret; 47835d13a37bSArne Jansen int err; 47845d13a37bSArne Jansen 47855d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 47865d13a37bSArne Jansen return -EPERM; 47875d13a37bSArne Jansen 4788905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4789905b0ddaSMiao Xie if (ret) 4790905b0ddaSMiao Xie return ret; 47915d13a37bSArne Jansen 47925d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4793905b0ddaSMiao Xie if (IS_ERR(sa)) { 4794905b0ddaSMiao Xie ret = PTR_ERR(sa); 4795905b0ddaSMiao Xie goto drop_write; 4796905b0ddaSMiao Xie } 47975d13a37bSArne Jansen 47987708f029SWang Shilong down_write(&root->fs_info->subvol_sem); 479992f183aaSWang Shilong trans = btrfs_start_transaction(root->fs_info->tree_root, 2); 48005d13a37bSArne Jansen if (IS_ERR(trans)) { 48015d13a37bSArne Jansen ret = PTR_ERR(trans); 48025d13a37bSArne Jansen goto out; 48035d13a37bSArne Jansen } 48045d13a37bSArne Jansen 48055d13a37bSArne Jansen switch (sa->cmd) { 48065d13a37bSArne Jansen case BTRFS_QUOTA_CTL_ENABLE: 48075d13a37bSArne Jansen ret = btrfs_quota_enable(trans, root->fs_info); 48085d13a37bSArne Jansen break; 48095d13a37bSArne Jansen case BTRFS_QUOTA_CTL_DISABLE: 48105d13a37bSArne Jansen ret = btrfs_quota_disable(trans, root->fs_info); 48115d13a37bSArne Jansen break; 48125d13a37bSArne Jansen default: 48135d13a37bSArne Jansen ret = -EINVAL; 48145d13a37bSArne Jansen break; 48155d13a37bSArne Jansen } 48165d13a37bSArne Jansen 481792f183aaSWang Shilong err = btrfs_commit_transaction(trans, root->fs_info->tree_root); 48185d13a37bSArne Jansen if (err && !ret) 48195d13a37bSArne Jansen ret = err; 48205d13a37bSArne Jansen out: 48215d13a37bSArne Jansen kfree(sa); 48227708f029SWang Shilong up_write(&root->fs_info->subvol_sem); 4823905b0ddaSMiao Xie drop_write: 4824905b0ddaSMiao Xie mnt_drop_write_file(file); 48255d13a37bSArne Jansen return ret; 48265d13a37bSArne Jansen } 48275d13a37bSArne Jansen 4828905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg) 48295d13a37bSArne Jansen { 4830496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 48315d13a37bSArne Jansen struct btrfs_ioctl_qgroup_assign_args *sa; 48325d13a37bSArne Jansen struct btrfs_trans_handle *trans; 48335d13a37bSArne Jansen int ret; 48345d13a37bSArne Jansen int err; 48355d13a37bSArne Jansen 48365d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 48375d13a37bSArne Jansen return -EPERM; 48385d13a37bSArne Jansen 4839905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4840905b0ddaSMiao Xie if (ret) 4841905b0ddaSMiao Xie return ret; 48425d13a37bSArne Jansen 48435d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4844905b0ddaSMiao Xie if (IS_ERR(sa)) { 4845905b0ddaSMiao Xie ret = PTR_ERR(sa); 4846905b0ddaSMiao Xie goto drop_write; 4847905b0ddaSMiao Xie } 48485d13a37bSArne Jansen 48495d13a37bSArne Jansen trans = btrfs_join_transaction(root); 48505d13a37bSArne Jansen if (IS_ERR(trans)) { 48515d13a37bSArne Jansen ret = PTR_ERR(trans); 48525d13a37bSArne Jansen goto out; 48535d13a37bSArne Jansen } 48545d13a37bSArne Jansen 48555d13a37bSArne Jansen /* FIXME: check if the IDs really exist */ 48565d13a37bSArne Jansen if (sa->assign) { 48575d13a37bSArne Jansen ret = btrfs_add_qgroup_relation(trans, root->fs_info, 48585d13a37bSArne Jansen sa->src, sa->dst); 48595d13a37bSArne Jansen } else { 48605d13a37bSArne Jansen ret = btrfs_del_qgroup_relation(trans, root->fs_info, 48615d13a37bSArne Jansen sa->src, sa->dst); 48625d13a37bSArne Jansen } 48635d13a37bSArne Jansen 4864e082f563SQu Wenruo /* update qgroup status and info */ 4865e082f563SQu Wenruo err = btrfs_run_qgroups(trans, root->fs_info); 4866e082f563SQu Wenruo if (err < 0) 4867a4553fefSAnand Jain btrfs_std_error(root->fs_info, ret, 4868e082f563SQu Wenruo "failed to update qgroup status and info\n"); 48695d13a37bSArne Jansen err = btrfs_end_transaction(trans, root); 48705d13a37bSArne Jansen if (err && !ret) 48715d13a37bSArne Jansen ret = err; 48725d13a37bSArne Jansen 48735d13a37bSArne Jansen out: 48745d13a37bSArne Jansen kfree(sa); 4875905b0ddaSMiao Xie drop_write: 4876905b0ddaSMiao Xie mnt_drop_write_file(file); 48775d13a37bSArne Jansen return ret; 48785d13a37bSArne Jansen } 48795d13a37bSArne Jansen 4880905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg) 48815d13a37bSArne Jansen { 4882496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 48835d13a37bSArne Jansen struct btrfs_ioctl_qgroup_create_args *sa; 48845d13a37bSArne Jansen struct btrfs_trans_handle *trans; 48855d13a37bSArne Jansen int ret; 48865d13a37bSArne Jansen int err; 48875d13a37bSArne Jansen 48885d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 48895d13a37bSArne Jansen return -EPERM; 48905d13a37bSArne Jansen 4891905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4892905b0ddaSMiao Xie if (ret) 4893905b0ddaSMiao Xie return ret; 48945d13a37bSArne Jansen 48955d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4896905b0ddaSMiao Xie if (IS_ERR(sa)) { 4897905b0ddaSMiao Xie ret = PTR_ERR(sa); 4898905b0ddaSMiao Xie goto drop_write; 4899905b0ddaSMiao Xie } 49005d13a37bSArne Jansen 4901d86e56cfSMiao Xie if (!sa->qgroupid) { 4902d86e56cfSMiao Xie ret = -EINVAL; 4903d86e56cfSMiao Xie goto out; 4904d86e56cfSMiao Xie } 4905d86e56cfSMiao Xie 49065d13a37bSArne Jansen trans = btrfs_join_transaction(root); 49075d13a37bSArne Jansen if (IS_ERR(trans)) { 49085d13a37bSArne Jansen ret = PTR_ERR(trans); 49095d13a37bSArne Jansen goto out; 49105d13a37bSArne Jansen } 49115d13a37bSArne Jansen 49125d13a37bSArne Jansen /* FIXME: check if the IDs really exist */ 49135d13a37bSArne Jansen if (sa->create) { 49144087cf24SDongsheng Yang ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid); 49155d13a37bSArne Jansen } else { 49165d13a37bSArne Jansen ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid); 49175d13a37bSArne Jansen } 49185d13a37bSArne Jansen 49195d13a37bSArne Jansen err = btrfs_end_transaction(trans, root); 49205d13a37bSArne Jansen if (err && !ret) 49215d13a37bSArne Jansen ret = err; 49225d13a37bSArne Jansen 49235d13a37bSArne Jansen out: 49245d13a37bSArne Jansen kfree(sa); 4925905b0ddaSMiao Xie drop_write: 4926905b0ddaSMiao Xie mnt_drop_write_file(file); 49275d13a37bSArne Jansen return ret; 49285d13a37bSArne Jansen } 49295d13a37bSArne Jansen 4930905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg) 49315d13a37bSArne Jansen { 4932496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 49335d13a37bSArne Jansen struct btrfs_ioctl_qgroup_limit_args *sa; 49345d13a37bSArne Jansen struct btrfs_trans_handle *trans; 49355d13a37bSArne Jansen int ret; 49365d13a37bSArne Jansen int err; 49375d13a37bSArne Jansen u64 qgroupid; 49385d13a37bSArne Jansen 49395d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 49405d13a37bSArne Jansen return -EPERM; 49415d13a37bSArne Jansen 4942905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4943905b0ddaSMiao Xie if (ret) 4944905b0ddaSMiao Xie return ret; 49455d13a37bSArne Jansen 49465d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4947905b0ddaSMiao Xie if (IS_ERR(sa)) { 4948905b0ddaSMiao Xie ret = PTR_ERR(sa); 4949905b0ddaSMiao Xie goto drop_write; 4950905b0ddaSMiao Xie } 49515d13a37bSArne Jansen 49525d13a37bSArne Jansen trans = btrfs_join_transaction(root); 49535d13a37bSArne Jansen if (IS_ERR(trans)) { 49545d13a37bSArne Jansen ret = PTR_ERR(trans); 49555d13a37bSArne Jansen goto out; 49565d13a37bSArne Jansen } 49575d13a37bSArne Jansen 49585d13a37bSArne Jansen qgroupid = sa->qgroupid; 49595d13a37bSArne Jansen if (!qgroupid) { 49605d13a37bSArne Jansen /* take the current subvol as qgroup */ 49615d13a37bSArne Jansen qgroupid = root->root_key.objectid; 49625d13a37bSArne Jansen } 49635d13a37bSArne Jansen 49645d13a37bSArne Jansen /* FIXME: check if the IDs really exist */ 49655d13a37bSArne Jansen ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim); 49665d13a37bSArne Jansen 49675d13a37bSArne Jansen err = btrfs_end_transaction(trans, root); 49685d13a37bSArne Jansen if (err && !ret) 49695d13a37bSArne Jansen ret = err; 49705d13a37bSArne Jansen 49715d13a37bSArne Jansen out: 49725d13a37bSArne Jansen kfree(sa); 4973905b0ddaSMiao Xie drop_write: 4974905b0ddaSMiao Xie mnt_drop_write_file(file); 49755d13a37bSArne Jansen return ret; 49765d13a37bSArne Jansen } 49775d13a37bSArne Jansen 49782f232036SJan Schmidt static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) 49792f232036SJan Schmidt { 49806d0379ecSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 49812f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 49822f232036SJan Schmidt int ret; 49832f232036SJan Schmidt 49842f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 49852f232036SJan Schmidt return -EPERM; 49862f232036SJan Schmidt 49872f232036SJan Schmidt ret = mnt_want_write_file(file); 49882f232036SJan Schmidt if (ret) 49892f232036SJan Schmidt return ret; 49902f232036SJan Schmidt 49912f232036SJan Schmidt qsa = memdup_user(arg, sizeof(*qsa)); 49922f232036SJan Schmidt if (IS_ERR(qsa)) { 49932f232036SJan Schmidt ret = PTR_ERR(qsa); 49942f232036SJan Schmidt goto drop_write; 49952f232036SJan Schmidt } 49962f232036SJan Schmidt 49972f232036SJan Schmidt if (qsa->flags) { 49982f232036SJan Schmidt ret = -EINVAL; 49992f232036SJan Schmidt goto out; 50002f232036SJan Schmidt } 50012f232036SJan Schmidt 50022f232036SJan Schmidt ret = btrfs_qgroup_rescan(root->fs_info); 50032f232036SJan Schmidt 50042f232036SJan Schmidt out: 50052f232036SJan Schmidt kfree(qsa); 50062f232036SJan Schmidt drop_write: 50072f232036SJan Schmidt mnt_drop_write_file(file); 50082f232036SJan Schmidt return ret; 50092f232036SJan Schmidt } 50102f232036SJan Schmidt 50112f232036SJan Schmidt static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg) 50122f232036SJan Schmidt { 50136d0379ecSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 50142f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 50152f232036SJan Schmidt int ret = 0; 50162f232036SJan Schmidt 50172f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 50182f232036SJan Schmidt return -EPERM; 50192f232036SJan Schmidt 50202f232036SJan Schmidt qsa = kzalloc(sizeof(*qsa), GFP_NOFS); 50212f232036SJan Schmidt if (!qsa) 50222f232036SJan Schmidt return -ENOMEM; 50232f232036SJan Schmidt 50242f232036SJan Schmidt if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { 50252f232036SJan Schmidt qsa->flags = 1; 50262f232036SJan Schmidt qsa->progress = root->fs_info->qgroup_rescan_progress.objectid; 50272f232036SJan Schmidt } 50282f232036SJan Schmidt 50292f232036SJan Schmidt if (copy_to_user(arg, qsa, sizeof(*qsa))) 50302f232036SJan Schmidt ret = -EFAULT; 50312f232036SJan Schmidt 50322f232036SJan Schmidt kfree(qsa); 50332f232036SJan Schmidt return ret; 50342f232036SJan Schmidt } 50352f232036SJan Schmidt 503657254b6eSJan Schmidt static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg) 503757254b6eSJan Schmidt { 503854563d41SAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 503957254b6eSJan Schmidt 504057254b6eSJan Schmidt if (!capable(CAP_SYS_ADMIN)) 504157254b6eSJan Schmidt return -EPERM; 504257254b6eSJan Schmidt 504357254b6eSJan Schmidt return btrfs_qgroup_wait_for_completion(root->fs_info); 504457254b6eSJan Schmidt } 504557254b6eSJan Schmidt 5046abccd00fSHugo Mills static long _btrfs_ioctl_set_received_subvol(struct file *file, 5047abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa) 50488ea05e3aSAlexander Block { 5049496ad9aaSAl Viro struct inode *inode = file_inode(file); 50508ea05e3aSAlexander Block struct btrfs_root *root = BTRFS_I(inode)->root; 50518ea05e3aSAlexander Block struct btrfs_root_item *root_item = &root->root_item; 50528ea05e3aSAlexander Block struct btrfs_trans_handle *trans; 50538ea05e3aSAlexander Block struct timespec ct = CURRENT_TIME; 50548ea05e3aSAlexander Block int ret = 0; 5055dd5f9615SStefan Behrens int received_uuid_changed; 50568ea05e3aSAlexander Block 5057bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 5058bd60ea0fSDavid Sterba return -EPERM; 5059bd60ea0fSDavid Sterba 50608ea05e3aSAlexander Block ret = mnt_want_write_file(file); 50618ea05e3aSAlexander Block if (ret < 0) 50628ea05e3aSAlexander Block return ret; 50638ea05e3aSAlexander Block 50648ea05e3aSAlexander Block down_write(&root->fs_info->subvol_sem); 50658ea05e3aSAlexander Block 50668ea05e3aSAlexander Block if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { 50678ea05e3aSAlexander Block ret = -EINVAL; 50688ea05e3aSAlexander Block goto out; 50698ea05e3aSAlexander Block } 50708ea05e3aSAlexander Block 50718ea05e3aSAlexander Block if (btrfs_root_readonly(root)) { 50728ea05e3aSAlexander Block ret = -EROFS; 50738ea05e3aSAlexander Block goto out; 50748ea05e3aSAlexander Block } 50758ea05e3aSAlexander Block 5076dd5f9615SStefan Behrens /* 5077dd5f9615SStefan Behrens * 1 - root item 5078dd5f9615SStefan Behrens * 2 - uuid items (received uuid + subvol uuid) 5079dd5f9615SStefan Behrens */ 5080dd5f9615SStefan Behrens trans = btrfs_start_transaction(root, 3); 50818ea05e3aSAlexander Block if (IS_ERR(trans)) { 50828ea05e3aSAlexander Block ret = PTR_ERR(trans); 50838ea05e3aSAlexander Block trans = NULL; 50848ea05e3aSAlexander Block goto out; 50858ea05e3aSAlexander Block } 50868ea05e3aSAlexander Block 50878ea05e3aSAlexander Block sa->rtransid = trans->transid; 50888ea05e3aSAlexander Block sa->rtime.sec = ct.tv_sec; 50898ea05e3aSAlexander Block sa->rtime.nsec = ct.tv_nsec; 50908ea05e3aSAlexander Block 5091dd5f9615SStefan Behrens received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid, 5092dd5f9615SStefan Behrens BTRFS_UUID_SIZE); 5093dd5f9615SStefan Behrens if (received_uuid_changed && 5094dd5f9615SStefan Behrens !btrfs_is_empty_uuid(root_item->received_uuid)) 5095dd5f9615SStefan Behrens btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root, 5096dd5f9615SStefan Behrens root_item->received_uuid, 5097dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5098dd5f9615SStefan Behrens root->root_key.objectid); 50998ea05e3aSAlexander Block memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE); 51008ea05e3aSAlexander Block btrfs_set_root_stransid(root_item, sa->stransid); 51018ea05e3aSAlexander Block btrfs_set_root_rtransid(root_item, sa->rtransid); 51023cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec); 51033cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec); 51043cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec); 51053cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec); 51068ea05e3aSAlexander Block 51078ea05e3aSAlexander Block ret = btrfs_update_root(trans, root->fs_info->tree_root, 51088ea05e3aSAlexander Block &root->root_key, &root->root_item); 51098ea05e3aSAlexander Block if (ret < 0) { 51108ea05e3aSAlexander Block btrfs_end_transaction(trans, root); 51118ea05e3aSAlexander Block goto out; 5112dd5f9615SStefan Behrens } 5113dd5f9615SStefan Behrens if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { 5114dd5f9615SStefan Behrens ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root, 5115dd5f9615SStefan Behrens sa->uuid, 5116dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5117dd5f9615SStefan Behrens root->root_key.objectid); 5118dd5f9615SStefan Behrens if (ret < 0 && ret != -EEXIST) { 5119dd5f9615SStefan Behrens btrfs_abort_transaction(trans, root, ret); 5120dd5f9615SStefan Behrens goto out; 5121dd5f9615SStefan Behrens } 5122dd5f9615SStefan Behrens } 51238ea05e3aSAlexander Block ret = btrfs_commit_transaction(trans, root); 5124dd5f9615SStefan Behrens if (ret < 0) { 5125dd5f9615SStefan Behrens btrfs_abort_transaction(trans, root, ret); 51268ea05e3aSAlexander Block goto out; 51278ea05e3aSAlexander Block } 51288ea05e3aSAlexander Block 5129abccd00fSHugo Mills out: 5130abccd00fSHugo Mills up_write(&root->fs_info->subvol_sem); 5131abccd00fSHugo Mills mnt_drop_write_file(file); 5132abccd00fSHugo Mills return ret; 5133abccd00fSHugo Mills } 5134abccd00fSHugo Mills 5135abccd00fSHugo Mills #ifdef CONFIG_64BIT 5136abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol_32(struct file *file, 5137abccd00fSHugo Mills void __user *arg) 5138abccd00fSHugo Mills { 5139abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL; 5140abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *args64 = NULL; 5141abccd00fSHugo Mills int ret = 0; 5142abccd00fSHugo Mills 5143abccd00fSHugo Mills args32 = memdup_user(arg, sizeof(*args32)); 5144abccd00fSHugo Mills if (IS_ERR(args32)) { 5145abccd00fSHugo Mills ret = PTR_ERR(args32); 5146abccd00fSHugo Mills args32 = NULL; 5147abccd00fSHugo Mills goto out; 5148abccd00fSHugo Mills } 5149abccd00fSHugo Mills 5150abccd00fSHugo Mills args64 = kmalloc(sizeof(*args64), GFP_NOFS); 515184dbeb87SDan Carpenter if (!args64) { 515284dbeb87SDan Carpenter ret = -ENOMEM; 5153abccd00fSHugo Mills goto out; 5154abccd00fSHugo Mills } 5155abccd00fSHugo Mills 5156abccd00fSHugo Mills memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE); 5157abccd00fSHugo Mills args64->stransid = args32->stransid; 5158abccd00fSHugo Mills args64->rtransid = args32->rtransid; 5159abccd00fSHugo Mills args64->stime.sec = args32->stime.sec; 5160abccd00fSHugo Mills args64->stime.nsec = args32->stime.nsec; 5161abccd00fSHugo Mills args64->rtime.sec = args32->rtime.sec; 5162abccd00fSHugo Mills args64->rtime.nsec = args32->rtime.nsec; 5163abccd00fSHugo Mills args64->flags = args32->flags; 5164abccd00fSHugo Mills 5165abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, args64); 5166abccd00fSHugo Mills if (ret) 5167abccd00fSHugo Mills goto out; 5168abccd00fSHugo Mills 5169abccd00fSHugo Mills memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE); 5170abccd00fSHugo Mills args32->stransid = args64->stransid; 5171abccd00fSHugo Mills args32->rtransid = args64->rtransid; 5172abccd00fSHugo Mills args32->stime.sec = args64->stime.sec; 5173abccd00fSHugo Mills args32->stime.nsec = args64->stime.nsec; 5174abccd00fSHugo Mills args32->rtime.sec = args64->rtime.sec; 5175abccd00fSHugo Mills args32->rtime.nsec = args64->rtime.nsec; 5176abccd00fSHugo Mills args32->flags = args64->flags; 5177abccd00fSHugo Mills 5178abccd00fSHugo Mills ret = copy_to_user(arg, args32, sizeof(*args32)); 5179abccd00fSHugo Mills if (ret) 5180abccd00fSHugo Mills ret = -EFAULT; 5181abccd00fSHugo Mills 5182abccd00fSHugo Mills out: 5183abccd00fSHugo Mills kfree(args32); 5184abccd00fSHugo Mills kfree(args64); 5185abccd00fSHugo Mills return ret; 5186abccd00fSHugo Mills } 5187abccd00fSHugo Mills #endif 5188abccd00fSHugo Mills 5189abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol(struct file *file, 5190abccd00fSHugo Mills void __user *arg) 5191abccd00fSHugo Mills { 5192abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa = NULL; 5193abccd00fSHugo Mills int ret = 0; 5194abccd00fSHugo Mills 5195abccd00fSHugo Mills sa = memdup_user(arg, sizeof(*sa)); 5196abccd00fSHugo Mills if (IS_ERR(sa)) { 5197abccd00fSHugo Mills ret = PTR_ERR(sa); 5198abccd00fSHugo Mills sa = NULL; 5199abccd00fSHugo Mills goto out; 5200abccd00fSHugo Mills } 5201abccd00fSHugo Mills 5202abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, sa); 5203abccd00fSHugo Mills 5204abccd00fSHugo Mills if (ret) 5205abccd00fSHugo Mills goto out; 5206abccd00fSHugo Mills 52078ea05e3aSAlexander Block ret = copy_to_user(arg, sa, sizeof(*sa)); 52088ea05e3aSAlexander Block if (ret) 52098ea05e3aSAlexander Block ret = -EFAULT; 52108ea05e3aSAlexander Block 52118ea05e3aSAlexander Block out: 52128ea05e3aSAlexander Block kfree(sa); 52138ea05e3aSAlexander Block return ret; 52148ea05e3aSAlexander Block } 52158ea05e3aSAlexander Block 5216867ab667Sjeff.liu static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg) 5217867ab667Sjeff.liu { 52186d0379ecSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 5219a1b83ac5SAnand Jain size_t len; 5220867ab667Sjeff.liu int ret; 5221a1b83ac5SAnand Jain char label[BTRFS_LABEL_SIZE]; 5222a1b83ac5SAnand Jain 5223a1b83ac5SAnand Jain spin_lock(&root->fs_info->super_lock); 5224a1b83ac5SAnand Jain memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE); 5225a1b83ac5SAnand Jain spin_unlock(&root->fs_info->super_lock); 5226a1b83ac5SAnand Jain 5227a1b83ac5SAnand Jain len = strnlen(label, BTRFS_LABEL_SIZE); 5228867ab667Sjeff.liu 5229867ab667Sjeff.liu if (len == BTRFS_LABEL_SIZE) { 5230efe120a0SFrank Holton btrfs_warn(root->fs_info, 5231efe120a0SFrank Holton "label is too long, return the first %zu bytes", --len); 5232867ab667Sjeff.liu } 5233867ab667Sjeff.liu 5234867ab667Sjeff.liu ret = copy_to_user(arg, label, len); 5235867ab667Sjeff.liu 5236867ab667Sjeff.liu return ret ? -EFAULT : 0; 5237867ab667Sjeff.liu } 5238867ab667Sjeff.liu 5239a8bfd4abSjeff.liu static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg) 5240a8bfd4abSjeff.liu { 52416d0379ecSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 5242a8bfd4abSjeff.liu struct btrfs_super_block *super_block = root->fs_info->super_copy; 5243a8bfd4abSjeff.liu struct btrfs_trans_handle *trans; 5244a8bfd4abSjeff.liu char label[BTRFS_LABEL_SIZE]; 5245a8bfd4abSjeff.liu int ret; 5246a8bfd4abSjeff.liu 5247a8bfd4abSjeff.liu if (!capable(CAP_SYS_ADMIN)) 5248a8bfd4abSjeff.liu return -EPERM; 5249a8bfd4abSjeff.liu 5250a8bfd4abSjeff.liu if (copy_from_user(label, arg, sizeof(label))) 5251a8bfd4abSjeff.liu return -EFAULT; 5252a8bfd4abSjeff.liu 5253a8bfd4abSjeff.liu if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) { 5254efe120a0SFrank Holton btrfs_err(root->fs_info, "unable to set label with more than %d bytes", 5255a8bfd4abSjeff.liu BTRFS_LABEL_SIZE - 1); 5256a8bfd4abSjeff.liu return -EINVAL; 5257a8bfd4abSjeff.liu } 5258a8bfd4abSjeff.liu 5259a8bfd4abSjeff.liu ret = mnt_want_write_file(file); 5260a8bfd4abSjeff.liu if (ret) 5261a8bfd4abSjeff.liu return ret; 5262a8bfd4abSjeff.liu 5263a8bfd4abSjeff.liu trans = btrfs_start_transaction(root, 0); 5264a8bfd4abSjeff.liu if (IS_ERR(trans)) { 5265a8bfd4abSjeff.liu ret = PTR_ERR(trans); 5266a8bfd4abSjeff.liu goto out_unlock; 5267a8bfd4abSjeff.liu } 5268a8bfd4abSjeff.liu 5269a1b83ac5SAnand Jain spin_lock(&root->fs_info->super_lock); 5270a8bfd4abSjeff.liu strcpy(super_block->label, label); 5271a1b83ac5SAnand Jain spin_unlock(&root->fs_info->super_lock); 5272d0270acaSJeff Mahoney ret = btrfs_commit_transaction(trans, root); 5273a8bfd4abSjeff.liu 5274a8bfd4abSjeff.liu out_unlock: 5275a8bfd4abSjeff.liu mnt_drop_write_file(file); 5276a8bfd4abSjeff.liu return ret; 5277a8bfd4abSjeff.liu } 5278a8bfd4abSjeff.liu 52792eaa055fSJeff Mahoney #define INIT_FEATURE_FLAGS(suffix) \ 52802eaa055fSJeff Mahoney { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \ 52812eaa055fSJeff Mahoney .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \ 52822eaa055fSJeff Mahoney .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix } 52832eaa055fSJeff Mahoney 52842eaa055fSJeff Mahoney static int btrfs_ioctl_get_supported_features(struct file *file, 52852eaa055fSJeff Mahoney void __user *arg) 52862eaa055fSJeff Mahoney { 52874d4ab6d6SDavid Sterba static const struct btrfs_ioctl_feature_flags features[3] = { 52882eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SUPP), 52892eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_SET), 52902eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_CLEAR) 52912eaa055fSJeff Mahoney }; 52922eaa055fSJeff Mahoney 52932eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 52942eaa055fSJeff Mahoney return -EFAULT; 52952eaa055fSJeff Mahoney 52962eaa055fSJeff Mahoney return 0; 52972eaa055fSJeff Mahoney } 52982eaa055fSJeff Mahoney 52992eaa055fSJeff Mahoney static int btrfs_ioctl_get_features(struct file *file, void __user *arg) 53002eaa055fSJeff Mahoney { 53012eaa055fSJeff Mahoney struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 53022eaa055fSJeff Mahoney struct btrfs_super_block *super_block = root->fs_info->super_copy; 53032eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags features; 53042eaa055fSJeff Mahoney 53052eaa055fSJeff Mahoney features.compat_flags = btrfs_super_compat_flags(super_block); 53062eaa055fSJeff Mahoney features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block); 53072eaa055fSJeff Mahoney features.incompat_flags = btrfs_super_incompat_flags(super_block); 53082eaa055fSJeff Mahoney 53092eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 53102eaa055fSJeff Mahoney return -EFAULT; 53112eaa055fSJeff Mahoney 53122eaa055fSJeff Mahoney return 0; 53132eaa055fSJeff Mahoney } 53142eaa055fSJeff Mahoney 53153b02a68aSJeff Mahoney static int check_feature_bits(struct btrfs_root *root, 53163b02a68aSJeff Mahoney enum btrfs_feature_set set, 53172eaa055fSJeff Mahoney u64 change_mask, u64 flags, u64 supported_flags, 53182eaa055fSJeff Mahoney u64 safe_set, u64 safe_clear) 53192eaa055fSJeff Mahoney { 53203b02a68aSJeff Mahoney const char *type = btrfs_feature_set_names[set]; 53213b02a68aSJeff Mahoney char *names; 53222eaa055fSJeff Mahoney u64 disallowed, unsupported; 53232eaa055fSJeff Mahoney u64 set_mask = flags & change_mask; 53242eaa055fSJeff Mahoney u64 clear_mask = ~flags & change_mask; 53252eaa055fSJeff Mahoney 53262eaa055fSJeff Mahoney unsupported = set_mask & ~supported_flags; 53272eaa055fSJeff Mahoney if (unsupported) { 53283b02a68aSJeff Mahoney names = btrfs_printable_features(set, unsupported); 53293b02a68aSJeff Mahoney if (names) { 53303b02a68aSJeff Mahoney btrfs_warn(root->fs_info, 53313b02a68aSJeff Mahoney "this kernel does not support the %s feature bit%s", 53323b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 53333b02a68aSJeff Mahoney kfree(names); 53343b02a68aSJeff Mahoney } else 53352eaa055fSJeff Mahoney btrfs_warn(root->fs_info, 53362eaa055fSJeff Mahoney "this kernel does not support %s bits 0x%llx", 53372eaa055fSJeff Mahoney type, unsupported); 53382eaa055fSJeff Mahoney return -EOPNOTSUPP; 53392eaa055fSJeff Mahoney } 53402eaa055fSJeff Mahoney 53412eaa055fSJeff Mahoney disallowed = set_mask & ~safe_set; 53422eaa055fSJeff Mahoney if (disallowed) { 53433b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 53443b02a68aSJeff Mahoney if (names) { 53453b02a68aSJeff Mahoney btrfs_warn(root->fs_info, 53463b02a68aSJeff Mahoney "can't set the %s feature bit%s while mounted", 53473b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 53483b02a68aSJeff Mahoney kfree(names); 53493b02a68aSJeff Mahoney } else 53502eaa055fSJeff Mahoney btrfs_warn(root->fs_info, 53512eaa055fSJeff Mahoney "can't set %s bits 0x%llx while mounted", 53522eaa055fSJeff Mahoney type, disallowed); 53532eaa055fSJeff Mahoney return -EPERM; 53542eaa055fSJeff Mahoney } 53552eaa055fSJeff Mahoney 53562eaa055fSJeff Mahoney disallowed = clear_mask & ~safe_clear; 53572eaa055fSJeff Mahoney if (disallowed) { 53583b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 53593b02a68aSJeff Mahoney if (names) { 53603b02a68aSJeff Mahoney btrfs_warn(root->fs_info, 53613b02a68aSJeff Mahoney "can't clear the %s feature bit%s while mounted", 53623b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 53633b02a68aSJeff Mahoney kfree(names); 53643b02a68aSJeff Mahoney } else 53652eaa055fSJeff Mahoney btrfs_warn(root->fs_info, 53662eaa055fSJeff Mahoney "can't clear %s bits 0x%llx while mounted", 53672eaa055fSJeff Mahoney type, disallowed); 53682eaa055fSJeff Mahoney return -EPERM; 53692eaa055fSJeff Mahoney } 53702eaa055fSJeff Mahoney 53712eaa055fSJeff Mahoney return 0; 53722eaa055fSJeff Mahoney } 53732eaa055fSJeff Mahoney 53742eaa055fSJeff Mahoney #define check_feature(root, change_mask, flags, mask_base) \ 53753b02a68aSJeff Mahoney check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \ 53762eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SUPP, \ 53772eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \ 53782eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR) 53792eaa055fSJeff Mahoney 53802eaa055fSJeff Mahoney static int btrfs_ioctl_set_features(struct file *file, void __user *arg) 53812eaa055fSJeff Mahoney { 53822eaa055fSJeff Mahoney struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 53832eaa055fSJeff Mahoney struct btrfs_super_block *super_block = root->fs_info->super_copy; 53842eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags flags[2]; 53852eaa055fSJeff Mahoney struct btrfs_trans_handle *trans; 53862eaa055fSJeff Mahoney u64 newflags; 53872eaa055fSJeff Mahoney int ret; 53882eaa055fSJeff Mahoney 53892eaa055fSJeff Mahoney if (!capable(CAP_SYS_ADMIN)) 53902eaa055fSJeff Mahoney return -EPERM; 53912eaa055fSJeff Mahoney 53922eaa055fSJeff Mahoney if (copy_from_user(flags, arg, sizeof(flags))) 53932eaa055fSJeff Mahoney return -EFAULT; 53942eaa055fSJeff Mahoney 53952eaa055fSJeff Mahoney /* Nothing to do */ 53962eaa055fSJeff Mahoney if (!flags[0].compat_flags && !flags[0].compat_ro_flags && 53972eaa055fSJeff Mahoney !flags[0].incompat_flags) 53982eaa055fSJeff Mahoney return 0; 53992eaa055fSJeff Mahoney 54002eaa055fSJeff Mahoney ret = check_feature(root, flags[0].compat_flags, 54012eaa055fSJeff Mahoney flags[1].compat_flags, COMPAT); 54022eaa055fSJeff Mahoney if (ret) 54032eaa055fSJeff Mahoney return ret; 54042eaa055fSJeff Mahoney 54052eaa055fSJeff Mahoney ret = check_feature(root, flags[0].compat_ro_flags, 54062eaa055fSJeff Mahoney flags[1].compat_ro_flags, COMPAT_RO); 54072eaa055fSJeff Mahoney if (ret) 54082eaa055fSJeff Mahoney return ret; 54092eaa055fSJeff Mahoney 54102eaa055fSJeff Mahoney ret = check_feature(root, flags[0].incompat_flags, 54112eaa055fSJeff Mahoney flags[1].incompat_flags, INCOMPAT); 54122eaa055fSJeff Mahoney if (ret) 54132eaa055fSJeff Mahoney return ret; 54142eaa055fSJeff Mahoney 54158051aa1aSDavid Sterba trans = btrfs_start_transaction(root, 0); 54162eaa055fSJeff Mahoney if (IS_ERR(trans)) 54172eaa055fSJeff Mahoney return PTR_ERR(trans); 54182eaa055fSJeff Mahoney 54192eaa055fSJeff Mahoney spin_lock(&root->fs_info->super_lock); 54202eaa055fSJeff Mahoney newflags = btrfs_super_compat_flags(super_block); 54212eaa055fSJeff Mahoney newflags |= flags[0].compat_flags & flags[1].compat_flags; 54222eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags); 54232eaa055fSJeff Mahoney btrfs_set_super_compat_flags(super_block, newflags); 54242eaa055fSJeff Mahoney 54252eaa055fSJeff Mahoney newflags = btrfs_super_compat_ro_flags(super_block); 54262eaa055fSJeff Mahoney newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags; 54272eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags); 54282eaa055fSJeff Mahoney btrfs_set_super_compat_ro_flags(super_block, newflags); 54292eaa055fSJeff Mahoney 54302eaa055fSJeff Mahoney newflags = btrfs_super_incompat_flags(super_block); 54312eaa055fSJeff Mahoney newflags |= flags[0].incompat_flags & flags[1].incompat_flags; 54322eaa055fSJeff Mahoney newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags); 54332eaa055fSJeff Mahoney btrfs_set_super_incompat_flags(super_block, newflags); 54342eaa055fSJeff Mahoney spin_unlock(&root->fs_info->super_lock); 54352eaa055fSJeff Mahoney 5436d0270acaSJeff Mahoney return btrfs_commit_transaction(trans, root); 54372eaa055fSJeff Mahoney } 54382eaa055fSJeff Mahoney 5439f46b5a66SChristoph Hellwig long btrfs_ioctl(struct file *file, unsigned int 5440f46b5a66SChristoph Hellwig cmd, unsigned long arg) 5441f46b5a66SChristoph Hellwig { 5442496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 54434bcabaa3SChristoph Hellwig void __user *argp = (void __user *)arg; 5444f46b5a66SChristoph Hellwig 5445f46b5a66SChristoph Hellwig switch (cmd) { 54466cbff00fSChristoph Hellwig case FS_IOC_GETFLAGS: 54476cbff00fSChristoph Hellwig return btrfs_ioctl_getflags(file, argp); 54486cbff00fSChristoph Hellwig case FS_IOC_SETFLAGS: 54496cbff00fSChristoph Hellwig return btrfs_ioctl_setflags(file, argp); 54506cbff00fSChristoph Hellwig case FS_IOC_GETVERSION: 54516cbff00fSChristoph Hellwig return btrfs_ioctl_getversion(file, argp); 5452f7039b1dSLi Dongyang case FITRIM: 5453f7039b1dSLi Dongyang return btrfs_ioctl_fitrim(file, argp); 5454f46b5a66SChristoph Hellwig case BTRFS_IOC_SNAP_CREATE: 5455fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 0); 5456fdfb1e4fSLi Zefan case BTRFS_IOC_SNAP_CREATE_V2: 5457fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create_v2(file, argp, 0); 54583de4586cSChris Mason case BTRFS_IOC_SUBVOL_CREATE: 5459fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 1); 54606f72c7e2SArne Jansen case BTRFS_IOC_SUBVOL_CREATE_V2: 54616f72c7e2SArne Jansen return btrfs_ioctl_snap_create_v2(file, argp, 1); 546276dda93cSYan, Zheng case BTRFS_IOC_SNAP_DESTROY: 546376dda93cSYan, Zheng return btrfs_ioctl_snap_destroy(file, argp); 54640caa102dSLi Zefan case BTRFS_IOC_SUBVOL_GETFLAGS: 54650caa102dSLi Zefan return btrfs_ioctl_subvol_getflags(file, argp); 54660caa102dSLi Zefan case BTRFS_IOC_SUBVOL_SETFLAGS: 54670caa102dSLi Zefan return btrfs_ioctl_subvol_setflags(file, argp); 54686ef5ed0dSJosef Bacik case BTRFS_IOC_DEFAULT_SUBVOL: 54696ef5ed0dSJosef Bacik return btrfs_ioctl_default_subvol(file, argp); 5470f46b5a66SChristoph Hellwig case BTRFS_IOC_DEFRAG: 54711e701a32SChris Mason return btrfs_ioctl_defrag(file, NULL); 54721e701a32SChris Mason case BTRFS_IOC_DEFRAG_RANGE: 54731e701a32SChris Mason return btrfs_ioctl_defrag(file, argp); 5474f46b5a66SChristoph Hellwig case BTRFS_IOC_RESIZE: 5475198605a8SMiao Xie return btrfs_ioctl_resize(file, argp); 5476f46b5a66SChristoph Hellwig case BTRFS_IOC_ADD_DEV: 54774bcabaa3SChristoph Hellwig return btrfs_ioctl_add_dev(root, argp); 5478f46b5a66SChristoph Hellwig case BTRFS_IOC_RM_DEV: 5479da24927bSMiao Xie return btrfs_ioctl_rm_dev(file, argp); 5480475f6387SJan Schmidt case BTRFS_IOC_FS_INFO: 5481475f6387SJan Schmidt return btrfs_ioctl_fs_info(root, argp); 5482475f6387SJan Schmidt case BTRFS_IOC_DEV_INFO: 5483475f6387SJan Schmidt return btrfs_ioctl_dev_info(root, argp); 5484f46b5a66SChristoph Hellwig case BTRFS_IOC_BALANCE: 54859ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, NULL); 5486f46b5a66SChristoph Hellwig case BTRFS_IOC_CLONE: 5487c5c9cd4dSSage Weil return btrfs_ioctl_clone(file, arg, 0, 0, 0); 5488c5c9cd4dSSage Weil case BTRFS_IOC_CLONE_RANGE: 54897a865e8aSChristoph Hellwig return btrfs_ioctl_clone_range(file, argp); 5490f46b5a66SChristoph Hellwig case BTRFS_IOC_TRANS_START: 5491f46b5a66SChristoph Hellwig return btrfs_ioctl_trans_start(file); 5492f46b5a66SChristoph Hellwig case BTRFS_IOC_TRANS_END: 5493f46b5a66SChristoph Hellwig return btrfs_ioctl_trans_end(file); 5494ac8e9819SChris Mason case BTRFS_IOC_TREE_SEARCH: 5495ac8e9819SChris Mason return btrfs_ioctl_tree_search(file, argp); 5496cc68a8a5SGerhard Heift case BTRFS_IOC_TREE_SEARCH_V2: 5497cc68a8a5SGerhard Heift return btrfs_ioctl_tree_search_v2(file, argp); 5498ac8e9819SChris Mason case BTRFS_IOC_INO_LOOKUP: 5499ac8e9819SChris Mason return btrfs_ioctl_ino_lookup(file, argp); 5500d7728c96SJan Schmidt case BTRFS_IOC_INO_PATHS: 5501d7728c96SJan Schmidt return btrfs_ioctl_ino_to_path(root, argp); 5502d7728c96SJan Schmidt case BTRFS_IOC_LOGICAL_INO: 5503d7728c96SJan Schmidt return btrfs_ioctl_logical_to_ino(root, argp); 55041406e432SJosef Bacik case BTRFS_IOC_SPACE_INFO: 55051406e432SJosef Bacik return btrfs_ioctl_space_info(root, argp); 55069b199859SFilipe David Borba Manana case BTRFS_IOC_SYNC: { 55079b199859SFilipe David Borba Manana int ret; 55089b199859SFilipe David Borba Manana 55096c255e67SMiao Xie ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1); 55109b199859SFilipe David Borba Manana if (ret) 55119b199859SFilipe David Borba Manana return ret; 5512ddb52f4fSAl Viro ret = btrfs_sync_fs(file_inode(file)->i_sb, 1); 55132fad4e83SDavid Sterba /* 55142fad4e83SDavid Sterba * The transaction thread may want to do more work, 55152fad4e83SDavid Sterba * namely it pokes the cleaner ktread that will start 55162fad4e83SDavid Sterba * processing uncleaned subvols. 55172fad4e83SDavid Sterba */ 55182fad4e83SDavid Sterba wake_up_process(root->fs_info->transaction_kthread); 55199b199859SFilipe David Borba Manana return ret; 55209b199859SFilipe David Borba Manana } 552146204592SSage Weil case BTRFS_IOC_START_SYNC: 55229a8c28beSMiao Xie return btrfs_ioctl_start_sync(root, argp); 552346204592SSage Weil case BTRFS_IOC_WAIT_SYNC: 55249a8c28beSMiao Xie return btrfs_ioctl_wait_sync(root, argp); 5525475f6387SJan Schmidt case BTRFS_IOC_SCRUB: 5526b8e95489SMiao Xie return btrfs_ioctl_scrub(file, argp); 5527475f6387SJan Schmidt case BTRFS_IOC_SCRUB_CANCEL: 5528475f6387SJan Schmidt return btrfs_ioctl_scrub_cancel(root, argp); 5529475f6387SJan Schmidt case BTRFS_IOC_SCRUB_PROGRESS: 5530475f6387SJan Schmidt return btrfs_ioctl_scrub_progress(root, argp); 5531c9e9f97bSIlya Dryomov case BTRFS_IOC_BALANCE_V2: 55329ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, argp); 5533837d5b6eSIlya Dryomov case BTRFS_IOC_BALANCE_CTL: 5534837d5b6eSIlya Dryomov return btrfs_ioctl_balance_ctl(root, arg); 553519a39dceSIlya Dryomov case BTRFS_IOC_BALANCE_PROGRESS: 553619a39dceSIlya Dryomov return btrfs_ioctl_balance_progress(root, argp); 55378ea05e3aSAlexander Block case BTRFS_IOC_SET_RECEIVED_SUBVOL: 55388ea05e3aSAlexander Block return btrfs_ioctl_set_received_subvol(file, argp); 5539abccd00fSHugo Mills #ifdef CONFIG_64BIT 5540abccd00fSHugo Mills case BTRFS_IOC_SET_RECEIVED_SUBVOL_32: 5541abccd00fSHugo Mills return btrfs_ioctl_set_received_subvol_32(file, argp); 5542abccd00fSHugo Mills #endif 554331db9f7cSAlexander Block case BTRFS_IOC_SEND: 554431db9f7cSAlexander Block return btrfs_ioctl_send(file, argp); 5545c11d2c23SStefan Behrens case BTRFS_IOC_GET_DEV_STATS: 5546b27f7c0cSDavid Sterba return btrfs_ioctl_get_dev_stats(root, argp); 55475d13a37bSArne Jansen case BTRFS_IOC_QUOTA_CTL: 5548905b0ddaSMiao Xie return btrfs_ioctl_quota_ctl(file, argp); 55495d13a37bSArne Jansen case BTRFS_IOC_QGROUP_ASSIGN: 5550905b0ddaSMiao Xie return btrfs_ioctl_qgroup_assign(file, argp); 55515d13a37bSArne Jansen case BTRFS_IOC_QGROUP_CREATE: 5552905b0ddaSMiao Xie return btrfs_ioctl_qgroup_create(file, argp); 55535d13a37bSArne Jansen case BTRFS_IOC_QGROUP_LIMIT: 5554905b0ddaSMiao Xie return btrfs_ioctl_qgroup_limit(file, argp); 55552f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN: 55562f232036SJan Schmidt return btrfs_ioctl_quota_rescan(file, argp); 55572f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_STATUS: 55582f232036SJan Schmidt return btrfs_ioctl_quota_rescan_status(file, argp); 555957254b6eSJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_WAIT: 556057254b6eSJan Schmidt return btrfs_ioctl_quota_rescan_wait(file, argp); 55613f6bcfbdSStefan Behrens case BTRFS_IOC_DEV_REPLACE: 55623f6bcfbdSStefan Behrens return btrfs_ioctl_dev_replace(root, argp); 5563867ab667Sjeff.liu case BTRFS_IOC_GET_FSLABEL: 5564867ab667Sjeff.liu return btrfs_ioctl_get_fslabel(file, argp); 5565a8bfd4abSjeff.liu case BTRFS_IOC_SET_FSLABEL: 5566a8bfd4abSjeff.liu return btrfs_ioctl_set_fslabel(file, argp); 5567416161dbSMark Fasheh case BTRFS_IOC_FILE_EXTENT_SAME: 5568416161dbSMark Fasheh return btrfs_ioctl_file_extent_same(file, argp); 55692eaa055fSJeff Mahoney case BTRFS_IOC_GET_SUPPORTED_FEATURES: 55702eaa055fSJeff Mahoney return btrfs_ioctl_get_supported_features(file, argp); 55712eaa055fSJeff Mahoney case BTRFS_IOC_GET_FEATURES: 55722eaa055fSJeff Mahoney return btrfs_ioctl_get_features(file, argp); 55732eaa055fSJeff Mahoney case BTRFS_IOC_SET_FEATURES: 55742eaa055fSJeff Mahoney return btrfs_ioctl_set_features(file, argp); 5575f46b5a66SChristoph Hellwig } 5576f46b5a66SChristoph Hellwig 5577f46b5a66SChristoph Hellwig return -ENOTTY; 5578f46b5a66SChristoph Hellwig } 5579