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/compat.h> 37f46b5a66SChristoph Hellwig #include <linux/bit_spinlock.h> 38cb8e7090SChristoph Hellwig #include <linux/security.h> 39f46b5a66SChristoph Hellwig #include <linux/xattr.h> 40f54de068SDavid Sterba #include <linux/mm.h> 415a0e3ad6STejun Heo #include <linux/slab.h> 42f7039b1dSLi Dongyang #include <linux/blkdev.h> 438ea05e3aSAlexander Block #include <linux/uuid.h> 4455e301fdSFilipe Brandenburger #include <linux/btrfs.h> 45416161dbSMark Fasheh #include <linux/uaccess.h> 46f46b5a66SChristoph Hellwig #include "ctree.h" 47f46b5a66SChristoph Hellwig #include "disk-io.h" 48f46b5a66SChristoph Hellwig #include "transaction.h" 49f46b5a66SChristoph Hellwig #include "btrfs_inode.h" 50f46b5a66SChristoph Hellwig #include "print-tree.h" 51f46b5a66SChristoph Hellwig #include "volumes.h" 52925baeddSChris Mason #include "locking.h" 53581bb050SLi Zefan #include "inode-map.h" 54d7728c96SJan Schmidt #include "backref.h" 55606686eeSJosef Bacik #include "rcu-string.h" 5631db9f7cSAlexander Block #include "send.h" 573f6bcfbdSStefan Behrens #include "dev-replace.h" 5863541927SFilipe David Borba Manana #include "props.h" 593b02a68aSJeff Mahoney #include "sysfs.h" 60fcebe456SJosef Bacik #include "qgroup.h" 611ec9a1aeSFilipe Manana #include "tree-log.h" 62ebb8765bSAnand Jain #include "compression.h" 63f46b5a66SChristoph Hellwig 64abccd00fSHugo Mills #ifdef CONFIG_64BIT 65abccd00fSHugo Mills /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI 66abccd00fSHugo Mills * structures are incorrect, as the timespec structure from userspace 67abccd00fSHugo Mills * is 4 bytes too small. We define these alternatives here to teach 68abccd00fSHugo Mills * the kernel about the 32-bit struct packing. 69abccd00fSHugo Mills */ 70abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 { 71abccd00fSHugo Mills __u64 sec; 72abccd00fSHugo Mills __u32 nsec; 73abccd00fSHugo Mills } __attribute__ ((__packed__)); 74abccd00fSHugo Mills 75abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 { 76abccd00fSHugo Mills char uuid[BTRFS_UUID_SIZE]; /* in */ 77abccd00fSHugo Mills __u64 stransid; /* in */ 78abccd00fSHugo Mills __u64 rtransid; /* out */ 79abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 stime; /* in */ 80abccd00fSHugo Mills struct btrfs_ioctl_timespec_32 rtime; /* out */ 81abccd00fSHugo Mills __u64 flags; /* in */ 82abccd00fSHugo Mills __u64 reserved[16]; /* in */ 83abccd00fSHugo Mills } __attribute__ ((__packed__)); 84abccd00fSHugo Mills 85abccd00fSHugo Mills #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \ 86abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32) 87abccd00fSHugo Mills #endif 88abccd00fSHugo Mills 89abccd00fSHugo Mills 90416161dbSMark Fasheh static int btrfs_clone(struct inode *src, struct inode *inode, 911c919a5eSMark Fasheh u64 off, u64 olen, u64 olen_aligned, u64 destoff, 921c919a5eSMark Fasheh int no_time_update); 93416161dbSMark Fasheh 946cbff00fSChristoph Hellwig /* Mask out flags that are inappropriate for the given type of inode. */ 956cbff00fSChristoph Hellwig static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags) 966cbff00fSChristoph Hellwig { 976cbff00fSChristoph Hellwig if (S_ISDIR(mode)) 986cbff00fSChristoph Hellwig return flags; 996cbff00fSChristoph Hellwig else if (S_ISREG(mode)) 1006cbff00fSChristoph Hellwig return flags & ~FS_DIRSYNC_FL; 1016cbff00fSChristoph Hellwig else 1026cbff00fSChristoph Hellwig return flags & (FS_NODUMP_FL | FS_NOATIME_FL); 1036cbff00fSChristoph Hellwig } 104f46b5a66SChristoph Hellwig 1056cbff00fSChristoph Hellwig /* 1066cbff00fSChristoph Hellwig * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl. 1076cbff00fSChristoph Hellwig */ 1086cbff00fSChristoph Hellwig static unsigned int btrfs_flags_to_ioctl(unsigned int flags) 1096cbff00fSChristoph Hellwig { 1106cbff00fSChristoph Hellwig unsigned int iflags = 0; 1116cbff00fSChristoph Hellwig 1126cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_SYNC) 1136cbff00fSChristoph Hellwig iflags |= FS_SYNC_FL; 1146cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_IMMUTABLE) 1156cbff00fSChristoph Hellwig iflags |= FS_IMMUTABLE_FL; 1166cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_APPEND) 1176cbff00fSChristoph Hellwig iflags |= FS_APPEND_FL; 1186cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NODUMP) 1196cbff00fSChristoph Hellwig iflags |= FS_NODUMP_FL; 1206cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_NOATIME) 1216cbff00fSChristoph Hellwig iflags |= FS_NOATIME_FL; 1226cbff00fSChristoph Hellwig if (flags & BTRFS_INODE_DIRSYNC) 1236cbff00fSChristoph Hellwig iflags |= FS_DIRSYNC_FL; 124d0092bddSLi Zefan if (flags & BTRFS_INODE_NODATACOW) 125d0092bddSLi Zefan iflags |= FS_NOCOW_FL; 126d0092bddSLi Zefan 12713f48dc9SSatoru Takeuchi if (flags & BTRFS_INODE_NOCOMPRESS) 128d0092bddSLi Zefan iflags |= FS_NOCOMP_FL; 12913f48dc9SSatoru Takeuchi else if (flags & BTRFS_INODE_COMPRESS) 13013f48dc9SSatoru Takeuchi iflags |= FS_COMPR_FL; 1316cbff00fSChristoph Hellwig 1326cbff00fSChristoph Hellwig return iflags; 1336cbff00fSChristoph Hellwig } 1346cbff00fSChristoph Hellwig 1356cbff00fSChristoph Hellwig /* 1366cbff00fSChristoph Hellwig * Update inode->i_flags based on the btrfs internal flags. 1376cbff00fSChristoph Hellwig */ 1386cbff00fSChristoph Hellwig void btrfs_update_iflags(struct inode *inode) 1396cbff00fSChristoph Hellwig { 1406cbff00fSChristoph Hellwig struct btrfs_inode *ip = BTRFS_I(inode); 1413cc79392SFilipe Manana unsigned int new_fl = 0; 1426cbff00fSChristoph Hellwig 1436cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_SYNC) 1443cc79392SFilipe Manana new_fl |= S_SYNC; 1456cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_IMMUTABLE) 1463cc79392SFilipe Manana new_fl |= S_IMMUTABLE; 1476cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_APPEND) 1483cc79392SFilipe Manana new_fl |= S_APPEND; 1496cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_NOATIME) 1503cc79392SFilipe Manana new_fl |= S_NOATIME; 1516cbff00fSChristoph Hellwig if (ip->flags & BTRFS_INODE_DIRSYNC) 1523cc79392SFilipe Manana new_fl |= S_DIRSYNC; 1533cc79392SFilipe Manana 1543cc79392SFilipe Manana set_mask_bits(&inode->i_flags, 1553cc79392SFilipe Manana S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC, 1563cc79392SFilipe Manana new_fl); 1576cbff00fSChristoph Hellwig } 1586cbff00fSChristoph Hellwig 1596cbff00fSChristoph Hellwig static int btrfs_ioctl_getflags(struct file *file, void __user *arg) 1606cbff00fSChristoph Hellwig { 161496ad9aaSAl Viro struct btrfs_inode *ip = BTRFS_I(file_inode(file)); 1626cbff00fSChristoph Hellwig unsigned int flags = btrfs_flags_to_ioctl(ip->flags); 1636cbff00fSChristoph Hellwig 1646cbff00fSChristoph Hellwig if (copy_to_user(arg, &flags, sizeof(flags))) 1656cbff00fSChristoph Hellwig return -EFAULT; 1666cbff00fSChristoph Hellwig return 0; 1676cbff00fSChristoph Hellwig } 1686cbff00fSChristoph Hellwig 16975e7cb7fSLiu Bo static int check_flags(unsigned int flags) 17075e7cb7fSLiu Bo { 17175e7cb7fSLiu Bo if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \ 17275e7cb7fSLiu Bo FS_NOATIME_FL | FS_NODUMP_FL | \ 17375e7cb7fSLiu Bo FS_SYNC_FL | FS_DIRSYNC_FL | \ 174e1e8fb6aSLi Zefan FS_NOCOMP_FL | FS_COMPR_FL | 175e1e8fb6aSLi Zefan FS_NOCOW_FL)) 17675e7cb7fSLiu Bo return -EOPNOTSUPP; 17775e7cb7fSLiu Bo 17875e7cb7fSLiu Bo if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL)) 17975e7cb7fSLiu Bo return -EINVAL; 18075e7cb7fSLiu Bo 18175e7cb7fSLiu Bo return 0; 18275e7cb7fSLiu Bo } 18375e7cb7fSLiu Bo 1846cbff00fSChristoph Hellwig static int btrfs_ioctl_setflags(struct file *file, void __user *arg) 1856cbff00fSChristoph Hellwig { 186496ad9aaSAl Viro struct inode *inode = file_inode(file); 1870b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 1886cbff00fSChristoph Hellwig struct btrfs_inode *ip = BTRFS_I(inode); 1896cbff00fSChristoph Hellwig struct btrfs_root *root = ip->root; 1906cbff00fSChristoph Hellwig struct btrfs_trans_handle *trans; 1916cbff00fSChristoph Hellwig unsigned int flags, oldflags; 1926cbff00fSChristoph Hellwig int ret; 193f062abf0SLi Zefan u64 ip_oldflags; 194f062abf0SLi Zefan unsigned int i_oldflags; 1957e97b8daSDavid Sterba umode_t mode; 1966cbff00fSChristoph Hellwig 197bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 198bd60ea0fSDavid Sterba return -EPERM; 199bd60ea0fSDavid Sterba 200b83cc969SLi Zefan if (btrfs_root_readonly(root)) 201b83cc969SLi Zefan return -EROFS; 202b83cc969SLi Zefan 2036cbff00fSChristoph Hellwig if (copy_from_user(&flags, arg, sizeof(flags))) 2046cbff00fSChristoph Hellwig return -EFAULT; 2056cbff00fSChristoph Hellwig 20675e7cb7fSLiu Bo ret = check_flags(flags); 20775e7cb7fSLiu Bo if (ret) 20875e7cb7fSLiu Bo return ret; 2096cbff00fSChristoph Hellwig 210e7848683SJan Kara ret = mnt_want_write_file(file); 211e7848683SJan Kara if (ret) 212e7848683SJan Kara return ret; 213e7848683SJan Kara 2145955102cSAl Viro inode_lock(inode); 2156cbff00fSChristoph Hellwig 216f062abf0SLi Zefan ip_oldflags = ip->flags; 217f062abf0SLi Zefan i_oldflags = inode->i_flags; 2187e97b8daSDavid Sterba mode = inode->i_mode; 219f062abf0SLi Zefan 2206cbff00fSChristoph Hellwig flags = btrfs_mask_flags(inode->i_mode, flags); 2216cbff00fSChristoph Hellwig oldflags = btrfs_flags_to_ioctl(ip->flags); 2226cbff00fSChristoph Hellwig if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) { 2236cbff00fSChristoph Hellwig if (!capable(CAP_LINUX_IMMUTABLE)) { 2246cbff00fSChristoph Hellwig ret = -EPERM; 2256cbff00fSChristoph Hellwig goto out_unlock; 2266cbff00fSChristoph Hellwig } 2276cbff00fSChristoph Hellwig } 2286cbff00fSChristoph Hellwig 2296cbff00fSChristoph Hellwig if (flags & FS_SYNC_FL) 2306cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_SYNC; 2316cbff00fSChristoph Hellwig else 2326cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_SYNC; 2336cbff00fSChristoph Hellwig if (flags & FS_IMMUTABLE_FL) 2346cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_IMMUTABLE; 2356cbff00fSChristoph Hellwig else 2366cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_IMMUTABLE; 2376cbff00fSChristoph Hellwig if (flags & FS_APPEND_FL) 2386cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_APPEND; 2396cbff00fSChristoph Hellwig else 2406cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_APPEND; 2416cbff00fSChristoph Hellwig if (flags & FS_NODUMP_FL) 2426cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_NODUMP; 2436cbff00fSChristoph Hellwig else 2446cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_NODUMP; 2456cbff00fSChristoph Hellwig if (flags & FS_NOATIME_FL) 2466cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_NOATIME; 2476cbff00fSChristoph Hellwig else 2486cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_NOATIME; 2496cbff00fSChristoph Hellwig if (flags & FS_DIRSYNC_FL) 2506cbff00fSChristoph Hellwig ip->flags |= BTRFS_INODE_DIRSYNC; 2516cbff00fSChristoph Hellwig else 2526cbff00fSChristoph Hellwig ip->flags &= ~BTRFS_INODE_DIRSYNC; 2537e97b8daSDavid Sterba if (flags & FS_NOCOW_FL) { 2547e97b8daSDavid Sterba if (S_ISREG(mode)) { 2557e97b8daSDavid Sterba /* 2567e97b8daSDavid Sterba * It's safe to turn csums off here, no extents exist. 2577e97b8daSDavid Sterba * Otherwise we want the flag to reflect the real COW 2587e97b8daSDavid Sterba * status of the file and will not set it. 2597e97b8daSDavid Sterba */ 2607e97b8daSDavid Sterba if (inode->i_size == 0) 2617e97b8daSDavid Sterba ip->flags |= BTRFS_INODE_NODATACOW 2627e97b8daSDavid Sterba | BTRFS_INODE_NODATASUM; 2637e97b8daSDavid Sterba } else { 264e1e8fb6aSLi Zefan ip->flags |= BTRFS_INODE_NODATACOW; 2657e97b8daSDavid Sterba } 2667e97b8daSDavid Sterba } else { 2677e97b8daSDavid Sterba /* 26801327610SNicholas D Steeves * Revert back under same assumptions as above 2697e97b8daSDavid Sterba */ 2707e97b8daSDavid Sterba if (S_ISREG(mode)) { 2717e97b8daSDavid Sterba if (inode->i_size == 0) 2727e97b8daSDavid Sterba ip->flags &= ~(BTRFS_INODE_NODATACOW 2737e97b8daSDavid Sterba | BTRFS_INODE_NODATASUM); 2747e97b8daSDavid Sterba } else { 275e1e8fb6aSLi Zefan ip->flags &= ~BTRFS_INODE_NODATACOW; 2767e97b8daSDavid Sterba } 2777e97b8daSDavid Sterba } 2786cbff00fSChristoph Hellwig 27975e7cb7fSLiu Bo /* 28075e7cb7fSLiu Bo * The COMPRESS flag can only be changed by users, while the NOCOMPRESS 28175e7cb7fSLiu Bo * flag may be changed automatically if compression code won't make 28275e7cb7fSLiu Bo * things smaller. 28375e7cb7fSLiu Bo */ 28475e7cb7fSLiu Bo if (flags & FS_NOCOMP_FL) { 28575e7cb7fSLiu Bo ip->flags &= ~BTRFS_INODE_COMPRESS; 28675e7cb7fSLiu Bo ip->flags |= BTRFS_INODE_NOCOMPRESS; 28763541927SFilipe David Borba Manana 28863541927SFilipe David Borba Manana ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0); 28963541927SFilipe David Borba Manana if (ret && ret != -ENODATA) 29063541927SFilipe David Borba Manana goto out_drop; 29175e7cb7fSLiu Bo } else if (flags & FS_COMPR_FL) { 29263541927SFilipe David Borba Manana const char *comp; 29363541927SFilipe David Borba Manana 29475e7cb7fSLiu Bo ip->flags |= BTRFS_INODE_COMPRESS; 29575e7cb7fSLiu Bo ip->flags &= ~BTRFS_INODE_NOCOMPRESS; 29663541927SFilipe David Borba Manana 2970b246afaSJeff Mahoney if (fs_info->compress_type == BTRFS_COMPRESS_LZO) 29863541927SFilipe David Borba Manana comp = "lzo"; 29963541927SFilipe David Borba Manana else 30063541927SFilipe David Borba Manana comp = "zlib"; 30163541927SFilipe David Borba Manana ret = btrfs_set_prop(inode, "btrfs.compression", 30263541927SFilipe David Borba Manana comp, strlen(comp), 0); 30363541927SFilipe David Borba Manana if (ret) 30463541927SFilipe David Borba Manana goto out_drop; 30563541927SFilipe David Borba Manana 306ebcb904dSLi Zefan } else { 30778a017a2SFilipe Manana ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0); 30878a017a2SFilipe Manana if (ret && ret != -ENODATA) 30978a017a2SFilipe Manana goto out_drop; 310ebcb904dSLi Zefan ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS); 31175e7cb7fSLiu Bo } 3126cbff00fSChristoph Hellwig 3134da6f1a3SLi Zefan trans = btrfs_start_transaction(root, 1); 314f062abf0SLi Zefan if (IS_ERR(trans)) { 315f062abf0SLi Zefan ret = PTR_ERR(trans); 316f062abf0SLi Zefan goto out_drop; 317f062abf0SLi Zefan } 3186cbff00fSChristoph Hellwig 319306424ccSLi Zefan btrfs_update_iflags(inode); 3200c4d2d95SJosef Bacik inode_inc_iversion(inode); 321c2050a45SDeepa Dinamani inode->i_ctime = current_time(inode); 3226cbff00fSChristoph Hellwig ret = btrfs_update_inode(trans, root, inode); 3236cbff00fSChristoph Hellwig 3243a45bb20SJeff Mahoney btrfs_end_transaction(trans); 325f062abf0SLi Zefan out_drop: 326f062abf0SLi Zefan if (ret) { 327f062abf0SLi Zefan ip->flags = ip_oldflags; 328f062abf0SLi Zefan inode->i_flags = i_oldflags; 329f062abf0SLi Zefan } 3306cbff00fSChristoph Hellwig 3316cbff00fSChristoph Hellwig out_unlock: 3325955102cSAl Viro inode_unlock(inode); 333e7848683SJan Kara mnt_drop_write_file(file); 3342d4e6f6aSliubo return ret; 3356cbff00fSChristoph Hellwig } 3366cbff00fSChristoph Hellwig 3376cbff00fSChristoph Hellwig static int btrfs_ioctl_getversion(struct file *file, int __user *arg) 3386cbff00fSChristoph Hellwig { 339496ad9aaSAl Viro struct inode *inode = file_inode(file); 3406cbff00fSChristoph Hellwig 3416cbff00fSChristoph Hellwig return put_user(inode->i_generation, arg); 3426cbff00fSChristoph Hellwig } 343f46b5a66SChristoph Hellwig 344f7039b1dSLi Dongyang static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg) 345f7039b1dSLi Dongyang { 3460b246afaSJeff Mahoney struct inode *inode = file_inode(file); 3470b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 348f7039b1dSLi Dongyang struct btrfs_device *device; 349f7039b1dSLi Dongyang struct request_queue *q; 350f7039b1dSLi Dongyang struct fstrim_range range; 351f7039b1dSLi Dongyang u64 minlen = ULLONG_MAX; 352f7039b1dSLi Dongyang u64 num_devices = 0; 353815745cfSAl Viro u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy); 354f7039b1dSLi Dongyang int ret; 355f7039b1dSLi Dongyang 356f7039b1dSLi Dongyang if (!capable(CAP_SYS_ADMIN)) 357f7039b1dSLi Dongyang return -EPERM; 358f7039b1dSLi Dongyang 3591f78160cSXiao Guangrong rcu_read_lock(); 3601f78160cSXiao Guangrong list_for_each_entry_rcu(device, &fs_info->fs_devices->devices, 3611f78160cSXiao Guangrong dev_list) { 362f7039b1dSLi Dongyang if (!device->bdev) 363f7039b1dSLi Dongyang continue; 364f7039b1dSLi Dongyang q = bdev_get_queue(device->bdev); 365f7039b1dSLi Dongyang if (blk_queue_discard(q)) { 366f7039b1dSLi Dongyang num_devices++; 36750d0446eSSeraphime Kirkovski minlen = min_t(u64, q->limits.discard_granularity, 368f7039b1dSLi Dongyang minlen); 369f7039b1dSLi Dongyang } 370f7039b1dSLi Dongyang } 3711f78160cSXiao Guangrong rcu_read_unlock(); 372f4c697e6SLukas Czerner 373f7039b1dSLi Dongyang if (!num_devices) 374f7039b1dSLi Dongyang return -EOPNOTSUPP; 375f7039b1dSLi Dongyang if (copy_from_user(&range, arg, sizeof(range))) 376f7039b1dSLi Dongyang return -EFAULT; 377e515c18bSLukas Czerner if (range.start > total_bytes || 378e515c18bSLukas Czerner range.len < fs_info->sb->s_blocksize) 379f4c697e6SLukas Czerner return -EINVAL; 380f7039b1dSLi Dongyang 381f4c697e6SLukas Czerner range.len = min(range.len, total_bytes - range.start); 382f7039b1dSLi Dongyang range.minlen = max(range.minlen, minlen); 3832ff7e61eSJeff Mahoney ret = btrfs_trim_fs(fs_info, &range); 384f7039b1dSLi Dongyang if (ret < 0) 385f7039b1dSLi Dongyang return ret; 386f7039b1dSLi Dongyang 387f7039b1dSLi Dongyang if (copy_to_user(arg, &range, sizeof(range))) 388f7039b1dSLi Dongyang return -EFAULT; 389f7039b1dSLi Dongyang 390f7039b1dSLi Dongyang return 0; 391f7039b1dSLi Dongyang } 392f7039b1dSLi Dongyang 393dd5f9615SStefan Behrens int btrfs_is_empty_uuid(u8 *uuid) 394dd5f9615SStefan Behrens { 39546e0f66aSChris Mason int i; 39646e0f66aSChris Mason 39746e0f66aSChris Mason for (i = 0; i < BTRFS_UUID_SIZE; i++) { 39846e0f66aSChris Mason if (uuid[i]) 39946e0f66aSChris Mason return 0; 40046e0f66aSChris Mason } 40146e0f66aSChris Mason return 1; 402dd5f9615SStefan Behrens } 403dd5f9615SStefan Behrens 404d5c12070SMiao Xie static noinline int create_subvol(struct inode *dir, 405cb8e7090SChristoph Hellwig struct dentry *dentry, 40652f75f4fSDavid Sterba const char *name, int namelen, 4076f72c7e2SArne Jansen u64 *async_transid, 4088696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 409f46b5a66SChristoph Hellwig { 4100b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 411f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 412f46b5a66SChristoph Hellwig struct btrfs_key key; 41349a3c4d9SDavid Sterba struct btrfs_root_item *root_item; 414f46b5a66SChristoph Hellwig struct btrfs_inode_item *inode_item; 415f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 416d5c12070SMiao Xie struct btrfs_root *root = BTRFS_I(dir)->root; 41776dda93cSYan, Zheng struct btrfs_root *new_root; 418d5c12070SMiao Xie struct btrfs_block_rsv block_rsv; 419c2050a45SDeepa Dinamani struct timespec cur_time = current_time(dir); 4205662344bSTsutomu Itoh struct inode *inode; 421f46b5a66SChristoph Hellwig int ret; 422f46b5a66SChristoph Hellwig int err; 423f46b5a66SChristoph Hellwig u64 objectid; 424f46b5a66SChristoph Hellwig u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; 4253de4586cSChris Mason u64 index = 0; 426d5c12070SMiao Xie u64 qgroup_reserved; 4278ea05e3aSAlexander Block uuid_le new_uuid; 428f46b5a66SChristoph Hellwig 42949a3c4d9SDavid Sterba root_item = kzalloc(sizeof(*root_item), GFP_KERNEL); 43049a3c4d9SDavid Sterba if (!root_item) 43149a3c4d9SDavid Sterba return -ENOMEM; 43249a3c4d9SDavid Sterba 4330b246afaSJeff Mahoney ret = btrfs_find_free_objectid(fs_info->tree_root, &objectid); 4342fbe8c8aSAl Viro if (ret) 43549a3c4d9SDavid Sterba goto fail_free; 4366a912213SJosef Bacik 437e09fe2d2SQu Wenruo /* 438e09fe2d2SQu Wenruo * Don't create subvolume whose level is not zero. Or qgroup will be 43901327610SNicholas D Steeves * screwed up since it assumes subvolume qgroup's level to be 0. 440e09fe2d2SQu Wenruo */ 44149a3c4d9SDavid Sterba if (btrfs_qgroup_level(objectid)) { 44249a3c4d9SDavid Sterba ret = -ENOSPC; 44349a3c4d9SDavid Sterba goto fail_free; 44449a3c4d9SDavid Sterba } 445e09fe2d2SQu Wenruo 446d5c12070SMiao Xie btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); 4479ed74f2dSJosef Bacik /* 448d5c12070SMiao Xie * The same as the snapshot creation, please see the comment 449d5c12070SMiao Xie * of create_snapshot(). 4509ed74f2dSJosef Bacik */ 451d5c12070SMiao Xie ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 452dd5f9615SStefan Behrens 8, &qgroup_reserved, false); 453d5c12070SMiao Xie if (ret) 45449a3c4d9SDavid Sterba goto fail_free; 455f46b5a66SChristoph Hellwig 456d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 457d5c12070SMiao Xie if (IS_ERR(trans)) { 458d5c12070SMiao Xie ret = PTR_ERR(trans); 4597775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &block_rsv); 46049a3c4d9SDavid Sterba goto fail_free; 461d5c12070SMiao Xie } 462d5c12070SMiao Xie trans->block_rsv = &block_rsv; 463d5c12070SMiao Xie trans->bytes_reserved = block_rsv.size; 464f46b5a66SChristoph Hellwig 4650b246afaSJeff Mahoney ret = btrfs_qgroup_inherit(trans, fs_info, 0, objectid, inherit); 4666f72c7e2SArne Jansen if (ret) 4676f72c7e2SArne Jansen goto fail; 4686f72c7e2SArne Jansen 4694d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); 4708e8a1e31SJosef Bacik if (IS_ERR(leaf)) { 4718e8a1e31SJosef Bacik ret = PTR_ERR(leaf); 4728e8a1e31SJosef Bacik goto fail; 4738e8a1e31SJosef Bacik } 474f46b5a66SChristoph Hellwig 475b159fa28SDavid Sterba memzero_extent_buffer(leaf, 0, sizeof(struct btrfs_header)); 476f46b5a66SChristoph Hellwig btrfs_set_header_bytenr(leaf, leaf->start); 477f46b5a66SChristoph Hellwig btrfs_set_header_generation(leaf, trans->transid); 4785d4f98a2SYan Zheng btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 479f46b5a66SChristoph Hellwig btrfs_set_header_owner(leaf, objectid); 480f46b5a66SChristoph Hellwig 4810b246afaSJeff Mahoney write_extent_buffer_fsid(leaf, fs_info->fsid); 4820b246afaSJeff Mahoney write_extent_buffer_chunk_tree_uuid(leaf, fs_info->chunk_tree_uuid); 483f46b5a66SChristoph Hellwig btrfs_mark_buffer_dirty(leaf); 484f46b5a66SChristoph Hellwig 48549a3c4d9SDavid Sterba inode_item = &root_item->inode; 4863cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 4873cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 4883cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 489da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 4900b246afaSJeff Mahoney fs_info->nodesize); 4913cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 492f46b5a66SChristoph Hellwig 49349a3c4d9SDavid Sterba btrfs_set_root_flags(root_item, 0); 49449a3c4d9SDavid Sterba btrfs_set_root_limit(root_item, 0); 4953cae210fSQu Wenruo btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT); 49608fe4db1SLi Zefan 49749a3c4d9SDavid Sterba btrfs_set_root_bytenr(root_item, leaf->start); 49849a3c4d9SDavid Sterba btrfs_set_root_generation(root_item, trans->transid); 49949a3c4d9SDavid Sterba btrfs_set_root_level(root_item, 0); 50049a3c4d9SDavid Sterba btrfs_set_root_refs(root_item, 1); 50149a3c4d9SDavid Sterba btrfs_set_root_used(root_item, leaf->len); 50249a3c4d9SDavid Sterba btrfs_set_root_last_snapshot(root_item, 0); 503f46b5a66SChristoph Hellwig 50449a3c4d9SDavid Sterba btrfs_set_root_generation_v2(root_item, 50549a3c4d9SDavid Sterba btrfs_root_generation(root_item)); 5068ea05e3aSAlexander Block uuid_le_gen(&new_uuid); 50749a3c4d9SDavid Sterba memcpy(root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE); 50849a3c4d9SDavid Sterba btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec); 50949a3c4d9SDavid Sterba btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec); 51049a3c4d9SDavid Sterba root_item->ctime = root_item->otime; 51149a3c4d9SDavid Sterba btrfs_set_root_ctransid(root_item, trans->transid); 51249a3c4d9SDavid Sterba btrfs_set_root_otransid(root_item, trans->transid); 513f46b5a66SChristoph Hellwig 514925baeddSChris Mason btrfs_tree_unlock(leaf); 515f46b5a66SChristoph Hellwig free_extent_buffer(leaf); 516f46b5a66SChristoph Hellwig leaf = NULL; 517f46b5a66SChristoph Hellwig 51849a3c4d9SDavid Sterba btrfs_set_root_dirid(root_item, new_dirid); 519f46b5a66SChristoph Hellwig 520f46b5a66SChristoph Hellwig key.objectid = objectid; 5215d4f98a2SYan Zheng key.offset = 0; 522962a298fSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 5230b246afaSJeff Mahoney ret = btrfs_insert_root(trans, fs_info->tree_root, &key, 52449a3c4d9SDavid Sterba root_item); 525f46b5a66SChristoph Hellwig if (ret) 526f46b5a66SChristoph Hellwig goto fail; 527f46b5a66SChristoph Hellwig 52876dda93cSYan, Zheng key.offset = (u64)-1; 5290b246afaSJeff Mahoney new_root = btrfs_read_fs_root_no_name(fs_info, &key); 53079787eaaSJeff Mahoney if (IS_ERR(new_root)) { 53179787eaaSJeff Mahoney ret = PTR_ERR(new_root); 53266642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 53379787eaaSJeff Mahoney goto fail; 53479787eaaSJeff Mahoney } 53576dda93cSYan, Zheng 53676dda93cSYan, Zheng btrfs_record_root_in_trans(trans, new_root); 53776dda93cSYan, Zheng 53863541927SFilipe David Borba Manana ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid); 539ce598979SMark Fasheh if (ret) { 540ce598979SMark Fasheh /* We potentially lose an unused inode item here */ 54166642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 542ce598979SMark Fasheh goto fail; 543ce598979SMark Fasheh } 544ce598979SMark Fasheh 545f32e48e9SChandan Rajendra mutex_lock(&new_root->objectid_mutex); 546f32e48e9SChandan Rajendra new_root->highest_objectid = new_dirid; 547f32e48e9SChandan Rajendra mutex_unlock(&new_root->objectid_mutex); 548f32e48e9SChandan Rajendra 549f46b5a66SChristoph Hellwig /* 550f46b5a66SChristoph Hellwig * insert the directory item 551f46b5a66SChristoph Hellwig */ 552877574e2SNikolay Borisov ret = btrfs_set_inode_index(BTRFS_I(dir), &index); 55379787eaaSJeff Mahoney if (ret) { 55466642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 55579787eaaSJeff Mahoney goto fail; 55679787eaaSJeff Mahoney } 5573de4586cSChris Mason 5583de4586cSChris Mason ret = btrfs_insert_dir_item(trans, root, 5598e7611cfSNikolay Borisov name, namelen, BTRFS_I(dir), &key, 5603de4586cSChris Mason BTRFS_FT_DIR, index); 56179787eaaSJeff Mahoney if (ret) { 56266642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 563f46b5a66SChristoph Hellwig goto fail; 56479787eaaSJeff Mahoney } 5650660b5afSChris Mason 5666ef06d27SNikolay Borisov btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2); 56752c26179SYan Zheng ret = btrfs_update_inode(trans, root, dir); 56852c26179SYan Zheng BUG_ON(ret); 56952c26179SYan Zheng 5700b246afaSJeff Mahoney ret = btrfs_add_root_ref(trans, fs_info, 5714df27c4dSYan, Zheng objectid, root->root_key.objectid, 5724a0cc7caSNikolay Borisov btrfs_ino(BTRFS_I(dir)), index, name, namelen); 5730660b5afSChris Mason BUG_ON(ret); 5740660b5afSChris Mason 5750b246afaSJeff Mahoney ret = btrfs_uuid_tree_add(trans, fs_info, root_item->uuid, 5766bccf3abSJeff Mahoney BTRFS_UUID_KEY_SUBVOL, objectid); 577dd5f9615SStefan Behrens if (ret) 57866642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 579dd5f9615SStefan Behrens 580f46b5a66SChristoph Hellwig fail: 58149a3c4d9SDavid Sterba kfree(root_item); 582d5c12070SMiao Xie trans->block_rsv = NULL; 583d5c12070SMiao Xie trans->bytes_reserved = 0; 5847775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &block_rsv); 585de6e8200SLiu Bo 58672fd032eSSage Weil if (async_transid) { 58772fd032eSSage Weil *async_transid = trans->transid; 5883a45bb20SJeff Mahoney err = btrfs_commit_transaction_async(trans, 1); 58900d71c9cSMiao Xie if (err) 5903a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 59172fd032eSSage Weil } else { 5923a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 59372fd032eSSage Weil } 594f46b5a66SChristoph Hellwig if (err && !ret) 595f46b5a66SChristoph Hellwig ret = err; 5961a65e24bSChris Mason 5975662344bSTsutomu Itoh if (!ret) { 5985662344bSTsutomu Itoh inode = btrfs_lookup_dentry(dir, dentry); 599de6e8200SLiu Bo if (IS_ERR(inode)) 600de6e8200SLiu Bo return PTR_ERR(inode); 6015662344bSTsutomu Itoh d_instantiate(dentry, inode); 6025662344bSTsutomu Itoh } 603f46b5a66SChristoph Hellwig return ret; 60449a3c4d9SDavid Sterba 60549a3c4d9SDavid Sterba fail_free: 60649a3c4d9SDavid Sterba kfree(root_item); 60749a3c4d9SDavid Sterba return ret; 608f46b5a66SChristoph Hellwig } 609f46b5a66SChristoph Hellwig 6109ea24bbeSFilipe Manana static void btrfs_wait_for_no_snapshoting_writes(struct btrfs_root *root) 6118257b2dcSMiao Xie { 6128257b2dcSMiao Xie s64 writers; 6138257b2dcSMiao Xie DEFINE_WAIT(wait); 6148257b2dcSMiao Xie 6158257b2dcSMiao Xie do { 6168257b2dcSMiao Xie prepare_to_wait(&root->subv_writers->wait, &wait, 6178257b2dcSMiao Xie TASK_UNINTERRUPTIBLE); 6188257b2dcSMiao Xie 6198257b2dcSMiao Xie writers = percpu_counter_sum(&root->subv_writers->counter); 6208257b2dcSMiao Xie if (writers) 6218257b2dcSMiao Xie schedule(); 6228257b2dcSMiao Xie 6238257b2dcSMiao Xie finish_wait(&root->subv_writers->wait, &wait); 6248257b2dcSMiao Xie } while (writers); 6258257b2dcSMiao Xie } 6268257b2dcSMiao Xie 627e9662f70SMiao Xie static int create_snapshot(struct btrfs_root *root, struct inode *dir, 62861d7e4cbSDavid Sterba struct dentry *dentry, 629e9662f70SMiao Xie u64 *async_transid, bool readonly, 630e9662f70SMiao Xie struct btrfs_qgroup_inherit *inherit) 631f46b5a66SChristoph Hellwig { 6320b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 6332e4bfab9SYan, Zheng struct inode *inode; 634f46b5a66SChristoph Hellwig struct btrfs_pending_snapshot *pending_snapshot; 635f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 6362e4bfab9SYan, Zheng int ret; 637f46b5a66SChristoph Hellwig 63827cdeb70SMiao Xie if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state)) 639f46b5a66SChristoph Hellwig return -EINVAL; 640f46b5a66SChristoph Hellwig 64123269bf5SDavid Sterba pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL); 642a1ee7362SDavid Sterba if (!pending_snapshot) 643a1ee7362SDavid Sterba return -ENOMEM; 644a1ee7362SDavid Sterba 645b0c0ea63SDavid Sterba pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item), 64623269bf5SDavid Sterba GFP_KERNEL); 6478546b570SDavid Sterba pending_snapshot->path = btrfs_alloc_path(); 6488546b570SDavid Sterba if (!pending_snapshot->root_item || !pending_snapshot->path) { 649b0c0ea63SDavid Sterba ret = -ENOMEM; 650b0c0ea63SDavid Sterba goto free_pending; 651b0c0ea63SDavid Sterba } 652b0c0ea63SDavid Sterba 6538257b2dcSMiao Xie atomic_inc(&root->will_be_snapshoted); 6544e857c58SPeter Zijlstra smp_mb__after_atomic(); 6559ea24bbeSFilipe Manana btrfs_wait_for_no_snapshoting_writes(root); 6568257b2dcSMiao Xie 6576a03843dSMiao Xie ret = btrfs_start_delalloc_inodes(root, 0); 6586a03843dSMiao Xie if (ret) 659a1ee7362SDavid Sterba goto dec_and_free; 6606a03843dSMiao Xie 6616374e57aSChris Mason btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1); 6626a03843dSMiao Xie 66366d8f3ddSMiao Xie btrfs_init_block_rsv(&pending_snapshot->block_rsv, 66466d8f3ddSMiao Xie BTRFS_BLOCK_RSV_TEMP); 665d5c12070SMiao Xie /* 666d5c12070SMiao Xie * 1 - parent dir inode 667d5c12070SMiao Xie * 2 - dir entries 668d5c12070SMiao Xie * 1 - root item 669d5c12070SMiao Xie * 2 - root ref/backref 670d5c12070SMiao Xie * 1 - root of snapshot 671dd5f9615SStefan Behrens * 1 - UUID item 672d5c12070SMiao Xie */ 673d5c12070SMiao Xie ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, 674dd5f9615SStefan Behrens &pending_snapshot->block_rsv, 8, 675ee3441b4SJeff Mahoney &pending_snapshot->qgroup_reserved, 676ee3441b4SJeff Mahoney false); 677d5c12070SMiao Xie if (ret) 678a1ee7362SDavid Sterba goto dec_and_free; 679d5c12070SMiao Xie 680a22285a6SYan, Zheng pending_snapshot->dentry = dentry; 681a22285a6SYan, Zheng pending_snapshot->root = root; 682b83cc969SLi Zefan pending_snapshot->readonly = readonly; 683e9662f70SMiao Xie pending_snapshot->dir = dir; 6848696c533SMiao Xie pending_snapshot->inherit = inherit; 685a22285a6SYan, Zheng 686d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 687a22285a6SYan, Zheng if (IS_ERR(trans)) { 688a22285a6SYan, Zheng ret = PTR_ERR(trans); 689a22285a6SYan, Zheng goto fail; 690a22285a6SYan, Zheng } 691a22285a6SYan, Zheng 6920b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 693a22285a6SYan, Zheng list_add(&pending_snapshot->list, 694a22285a6SYan, Zheng &trans->transaction->pending_snapshots); 6950b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 69672fd032eSSage Weil if (async_transid) { 69772fd032eSSage Weil *async_transid = trans->transid; 6983a45bb20SJeff Mahoney ret = btrfs_commit_transaction_async(trans, 1); 69900d71c9cSMiao Xie if (ret) 7003a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 70172fd032eSSage Weil } else { 7023a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 70372fd032eSSage Weil } 704aec8030aSMiao Xie if (ret) 705c37b2b62SJosef Bacik goto fail; 706a22285a6SYan, Zheng 707a22285a6SYan, Zheng ret = pending_snapshot->error; 708f46b5a66SChristoph Hellwig if (ret) 7092e4bfab9SYan, Zheng goto fail; 710f46b5a66SChristoph Hellwig 711d3797308SChris Mason ret = btrfs_orphan_cleanup(pending_snapshot->snap); 712d3797308SChris Mason if (ret) 713d3797308SChris Mason goto fail; 714d3797308SChris Mason 7152b0143b5SDavid Howells inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry); 7162e4bfab9SYan, Zheng if (IS_ERR(inode)) { 7172e4bfab9SYan, Zheng ret = PTR_ERR(inode); 7182e4bfab9SYan, Zheng goto fail; 7192e4bfab9SYan, Zheng } 7205662344bSTsutomu Itoh 7212e4bfab9SYan, Zheng d_instantiate(dentry, inode); 7222e4bfab9SYan, Zheng ret = 0; 7232e4bfab9SYan, Zheng fail: 7247775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &pending_snapshot->block_rsv); 725a1ee7362SDavid Sterba dec_and_free: 7269ea24bbeSFilipe Manana if (atomic_dec_and_test(&root->will_be_snapshoted)) 7279ea24bbeSFilipe Manana wake_up_atomic_t(&root->will_be_snapshoted); 728b0c0ea63SDavid Sterba free_pending: 729b0c0ea63SDavid Sterba kfree(pending_snapshot->root_item); 7308546b570SDavid Sterba btrfs_free_path(pending_snapshot->path); 731a1ee7362SDavid Sterba kfree(pending_snapshot); 732a1ee7362SDavid Sterba 733f46b5a66SChristoph Hellwig return ret; 734f46b5a66SChristoph Hellwig } 735f46b5a66SChristoph Hellwig 7364260f7c7SSage Weil /* copy of may_delete in fs/namei.c() 7374260f7c7SSage Weil * Check whether we can remove a link victim from directory dir, check 7384260f7c7SSage Weil * whether the type of victim is right. 7394260f7c7SSage Weil * 1. We can't do it if dir is read-only (done in permission()) 7404260f7c7SSage Weil * 2. We should have write and exec permissions on dir 7414260f7c7SSage Weil * 3. We can't remove anything from append-only dir 7424260f7c7SSage Weil * 4. We can't do anything with immutable dir (done in permission()) 7434260f7c7SSage Weil * 5. If the sticky bit on dir is set we should either 7444260f7c7SSage Weil * a. be owner of dir, or 7454260f7c7SSage Weil * b. be owner of victim, or 7464260f7c7SSage Weil * c. have CAP_FOWNER capability 74701327610SNicholas D Steeves * 6. If the victim is append-only or immutable we can't do anything with 7484260f7c7SSage Weil * links pointing to it. 7494260f7c7SSage Weil * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. 7504260f7c7SSage Weil * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. 7514260f7c7SSage Weil * 9. We can't remove a root or mountpoint. 7524260f7c7SSage Weil * 10. We don't allow removal of NFS sillyrenamed files; it's handled by 7534260f7c7SSage Weil * nfs_async_unlink(). 7544260f7c7SSage Weil */ 7554260f7c7SSage Weil 7564260f7c7SSage Weil static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir) 7574260f7c7SSage Weil { 7584260f7c7SSage Weil int error; 7594260f7c7SSage Weil 7602b0143b5SDavid Howells if (d_really_is_negative(victim)) 7614260f7c7SSage Weil return -ENOENT; 7624260f7c7SSage Weil 7632b0143b5SDavid Howells BUG_ON(d_inode(victim->d_parent) != dir); 7644fa6b5ecSJeff Layton audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); 7654260f7c7SSage Weil 7664260f7c7SSage Weil error = inode_permission(dir, MAY_WRITE | MAY_EXEC); 7674260f7c7SSage Weil if (error) 7684260f7c7SSage Weil return error; 7694260f7c7SSage Weil if (IS_APPEND(dir)) 7704260f7c7SSage Weil return -EPERM; 7712b0143b5SDavid Howells if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) || 7722b0143b5SDavid Howells IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim))) 7734260f7c7SSage Weil return -EPERM; 7744260f7c7SSage Weil if (isdir) { 775e36cb0b8SDavid Howells if (!d_is_dir(victim)) 7764260f7c7SSage Weil return -ENOTDIR; 7774260f7c7SSage Weil if (IS_ROOT(victim)) 7784260f7c7SSage Weil return -EBUSY; 779e36cb0b8SDavid Howells } else if (d_is_dir(victim)) 7804260f7c7SSage Weil return -EISDIR; 7814260f7c7SSage Weil if (IS_DEADDIR(dir)) 7824260f7c7SSage Weil return -ENOENT; 7834260f7c7SSage Weil if (victim->d_flags & DCACHE_NFSFS_RENAMED) 7844260f7c7SSage Weil return -EBUSY; 7854260f7c7SSage Weil return 0; 7864260f7c7SSage Weil } 7874260f7c7SSage Weil 788cb8e7090SChristoph Hellwig /* copy of may_create in fs/namei.c() */ 789cb8e7090SChristoph Hellwig static inline int btrfs_may_create(struct inode *dir, struct dentry *child) 790cb8e7090SChristoph Hellwig { 7912b0143b5SDavid Howells if (d_really_is_positive(child)) 792cb8e7090SChristoph Hellwig return -EEXIST; 793cb8e7090SChristoph Hellwig if (IS_DEADDIR(dir)) 794cb8e7090SChristoph Hellwig return -ENOENT; 795cb8e7090SChristoph Hellwig return inode_permission(dir, MAY_WRITE | MAY_EXEC); 796cb8e7090SChristoph Hellwig } 797cb8e7090SChristoph Hellwig 798cb8e7090SChristoph Hellwig /* 799cb8e7090SChristoph Hellwig * Create a new subvolume below @parent. This is largely modeled after 800cb8e7090SChristoph Hellwig * sys_mkdirat and vfs_mkdir, but we only do a single component lookup 801cb8e7090SChristoph Hellwig * inside this filesystem so it's quite a bit simpler. 802cb8e7090SChristoph Hellwig */ 80392872094SAl Viro static noinline int btrfs_mksubvol(const struct path *parent, 80452f75f4fSDavid Sterba const char *name, int namelen, 80572fd032eSSage Weil struct btrfs_root *snap_src, 8066f72c7e2SArne Jansen u64 *async_transid, bool readonly, 8078696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 808cb8e7090SChristoph Hellwig { 8092b0143b5SDavid Howells struct inode *dir = d_inode(parent->dentry); 8100b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 811cb8e7090SChristoph Hellwig struct dentry *dentry; 812cb8e7090SChristoph Hellwig int error; 813cb8e7090SChristoph Hellwig 81400235411SAl Viro error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 81500235411SAl Viro if (error == -EINTR) 81600235411SAl Viro return error; 817cb8e7090SChristoph Hellwig 818cb8e7090SChristoph Hellwig dentry = lookup_one_len(name, parent->dentry, namelen); 819cb8e7090SChristoph Hellwig error = PTR_ERR(dentry); 820cb8e7090SChristoph Hellwig if (IS_ERR(dentry)) 821cb8e7090SChristoph Hellwig goto out_unlock; 822cb8e7090SChristoph Hellwig 82376dda93cSYan, Zheng error = btrfs_may_create(dir, dentry); 824cb8e7090SChristoph Hellwig if (error) 825a874a63eSLiu Bo goto out_dput; 826cb8e7090SChristoph Hellwig 8279c52057cSChris Mason /* 8289c52057cSChris Mason * even if this name doesn't exist, we may get hash collisions. 8299c52057cSChris Mason * check for them now when we can safely fail 8309c52057cSChris Mason */ 8319c52057cSChris Mason error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, 8329c52057cSChris Mason dir->i_ino, name, 8339c52057cSChris Mason namelen); 8349c52057cSChris Mason if (error) 8359c52057cSChris Mason goto out_dput; 8369c52057cSChris Mason 8370b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 83876dda93cSYan, Zheng 83976dda93cSYan, Zheng if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) 84076dda93cSYan, Zheng goto out_up_read; 84176dda93cSYan, Zheng 8423de4586cSChris Mason if (snap_src) { 84361d7e4cbSDavid Sterba error = create_snapshot(snap_src, dir, dentry, 8446f72c7e2SArne Jansen async_transid, readonly, inherit); 8453de4586cSChris Mason } else { 846d5c12070SMiao Xie error = create_subvol(dir, dentry, name, namelen, 847d5c12070SMiao Xie async_transid, inherit); 8483de4586cSChris Mason } 84976dda93cSYan, Zheng if (!error) 85076dda93cSYan, Zheng fsnotify_mkdir(dir, dentry); 85176dda93cSYan, Zheng out_up_read: 8520b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 853cb8e7090SChristoph Hellwig out_dput: 854cb8e7090SChristoph Hellwig dput(dentry); 855cb8e7090SChristoph Hellwig out_unlock: 8565955102cSAl Viro inode_unlock(dir); 857cb8e7090SChristoph Hellwig return error; 858cb8e7090SChristoph Hellwig } 859cb8e7090SChristoph Hellwig 8604cb5300bSChris Mason /* 8614cb5300bSChris Mason * When we're defragging a range, we don't want to kick it off again 8624cb5300bSChris Mason * if it is really just waiting for delalloc to send it down. 8634cb5300bSChris Mason * If we find a nice big extent or delalloc range for the bytes in the 8644cb5300bSChris Mason * file you want to defrag, we return 0 to let you know to skip this 8654cb5300bSChris Mason * part of the file 8664cb5300bSChris Mason */ 867aab110abSDavid Sterba static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh) 8684cb5300bSChris Mason { 8694cb5300bSChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 8704cb5300bSChris Mason struct extent_map *em = NULL; 8714cb5300bSChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 8724cb5300bSChris Mason u64 end; 8734cb5300bSChris Mason 8744cb5300bSChris Mason read_lock(&em_tree->lock); 87509cbfeafSKirill A. Shutemov em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE); 8764cb5300bSChris Mason read_unlock(&em_tree->lock); 8774cb5300bSChris Mason 8784cb5300bSChris Mason if (em) { 8794cb5300bSChris Mason end = extent_map_end(em); 8804cb5300bSChris Mason free_extent_map(em); 8814cb5300bSChris Mason if (end - offset > thresh) 8824cb5300bSChris Mason return 0; 8834cb5300bSChris Mason } 8844cb5300bSChris Mason /* if we already have a nice delalloc here, just stop */ 8854cb5300bSChris Mason thresh /= 2; 8864cb5300bSChris Mason end = count_range_bits(io_tree, &offset, offset + thresh, 8874cb5300bSChris Mason thresh, EXTENT_DELALLOC, 1); 8884cb5300bSChris Mason if (end >= thresh) 8894cb5300bSChris Mason return 0; 8904cb5300bSChris Mason return 1; 8914cb5300bSChris Mason } 8924cb5300bSChris Mason 8934cb5300bSChris Mason /* 8944cb5300bSChris Mason * helper function to walk through a file and find extents 8954cb5300bSChris Mason * newer than a specific transid, and smaller than thresh. 8964cb5300bSChris Mason * 8974cb5300bSChris Mason * This is used by the defragging code to find new and small 8984cb5300bSChris Mason * extents 8994cb5300bSChris Mason */ 9004cb5300bSChris Mason static int find_new_extents(struct btrfs_root *root, 9014cb5300bSChris Mason struct inode *inode, u64 newer_than, 902aab110abSDavid Sterba u64 *off, u32 thresh) 9034cb5300bSChris Mason { 9044cb5300bSChris Mason struct btrfs_path *path; 9054cb5300bSChris Mason struct btrfs_key min_key; 9064cb5300bSChris Mason struct extent_buffer *leaf; 9074cb5300bSChris Mason struct btrfs_file_extent_item *extent; 9084cb5300bSChris Mason int type; 9094cb5300bSChris Mason int ret; 9104a0cc7caSNikolay Borisov u64 ino = btrfs_ino(BTRFS_I(inode)); 9114cb5300bSChris Mason 9124cb5300bSChris Mason path = btrfs_alloc_path(); 9134cb5300bSChris Mason if (!path) 9144cb5300bSChris Mason return -ENOMEM; 9154cb5300bSChris Mason 916a4689d2bSDavid Sterba min_key.objectid = ino; 9174cb5300bSChris Mason min_key.type = BTRFS_EXTENT_DATA_KEY; 9184cb5300bSChris Mason min_key.offset = *off; 9194cb5300bSChris Mason 9204cb5300bSChris Mason while (1) { 9216174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &min_key, path, newer_than); 9224cb5300bSChris Mason if (ret != 0) 9234cb5300bSChris Mason goto none; 924f094c9bdSFilipe Manana process_slot: 925a4689d2bSDavid Sterba if (min_key.objectid != ino) 9264cb5300bSChris Mason goto none; 9274cb5300bSChris Mason if (min_key.type != BTRFS_EXTENT_DATA_KEY) 9284cb5300bSChris Mason goto none; 9294cb5300bSChris Mason 9304cb5300bSChris Mason leaf = path->nodes[0]; 9314cb5300bSChris Mason extent = btrfs_item_ptr(leaf, path->slots[0], 9324cb5300bSChris Mason struct btrfs_file_extent_item); 9334cb5300bSChris Mason 9344cb5300bSChris Mason type = btrfs_file_extent_type(leaf, extent); 9354cb5300bSChris Mason if (type == BTRFS_FILE_EXTENT_REG && 9364cb5300bSChris Mason btrfs_file_extent_num_bytes(leaf, extent) < thresh && 9374cb5300bSChris Mason check_defrag_in_cache(inode, min_key.offset, thresh)) { 9384cb5300bSChris Mason *off = min_key.offset; 9394cb5300bSChris Mason btrfs_free_path(path); 9404cb5300bSChris Mason return 0; 9414cb5300bSChris Mason } 9424cb5300bSChris Mason 943f094c9bdSFilipe Manana path->slots[0]++; 944f094c9bdSFilipe Manana if (path->slots[0] < btrfs_header_nritems(leaf)) { 945f094c9bdSFilipe Manana btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]); 946f094c9bdSFilipe Manana goto process_slot; 947f094c9bdSFilipe Manana } 948f094c9bdSFilipe Manana 9494cb5300bSChris Mason if (min_key.offset == (u64)-1) 9504cb5300bSChris Mason goto none; 9514cb5300bSChris Mason 9524cb5300bSChris Mason min_key.offset++; 9534cb5300bSChris Mason btrfs_release_path(path); 9544cb5300bSChris Mason } 9554cb5300bSChris Mason none: 9564cb5300bSChris Mason btrfs_free_path(path); 9574cb5300bSChris Mason return -ENOENT; 9584cb5300bSChris Mason } 9594cb5300bSChris Mason 9606c282eb4SLi Zefan static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start) 96117ce6ef8SLiu Bo { 96217ce6ef8SLiu Bo struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 963940100a4SChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 9646c282eb4SLi Zefan struct extent_map *em; 96509cbfeafSKirill A. Shutemov u64 len = PAGE_SIZE; 966940100a4SChris Mason 967940100a4SChris Mason /* 968940100a4SChris Mason * hopefully we have this extent in the tree already, try without 969940100a4SChris Mason * the full extent lock 970940100a4SChris Mason */ 971940100a4SChris Mason read_lock(&em_tree->lock); 972940100a4SChris Mason em = lookup_extent_mapping(em_tree, start, len); 973940100a4SChris Mason read_unlock(&em_tree->lock); 974940100a4SChris Mason 975940100a4SChris Mason if (!em) { 976308d9800SFilipe Manana struct extent_state *cached = NULL; 977308d9800SFilipe Manana u64 end = start + len - 1; 978308d9800SFilipe Manana 979940100a4SChris Mason /* get the big lock and read metadata off disk */ 980ff13db41SDavid Sterba lock_extent_bits(io_tree, start, end, &cached); 981fc4f21b1SNikolay Borisov em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len, 0); 982308d9800SFilipe Manana unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS); 983940100a4SChris Mason 9846cf8bfbfSDan Carpenter if (IS_ERR(em)) 9856c282eb4SLi Zefan return NULL; 986940100a4SChris Mason } 987940100a4SChris Mason 9886c282eb4SLi Zefan return em; 9896c282eb4SLi Zefan } 9906c282eb4SLi Zefan 9916c282eb4SLi Zefan static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em) 9926c282eb4SLi Zefan { 9936c282eb4SLi Zefan struct extent_map *next; 9946c282eb4SLi Zefan bool ret = true; 9956c282eb4SLi Zefan 9966c282eb4SLi Zefan /* this is the last extent */ 9976c282eb4SLi Zefan if (em->start + em->len >= i_size_read(inode)) 9986c282eb4SLi Zefan return false; 9996c282eb4SLi Zefan 10006c282eb4SLi Zefan next = defrag_lookup_extent(inode, em->start + em->len); 1001e9512d72SChris Mason if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE) 1002e9512d72SChris Mason ret = false; 1003e9512d72SChris Mason else if ((em->block_start + em->block_len == next->block_start) && 1004ee22184bSByongho Lee (em->block_len > SZ_128K && next->block_len > SZ_128K)) 10056c282eb4SLi Zefan ret = false; 10066c282eb4SLi Zefan 10076c282eb4SLi Zefan free_extent_map(next); 10086c282eb4SLi Zefan return ret; 10096c282eb4SLi Zefan } 10106c282eb4SLi Zefan 1011aab110abSDavid Sterba static int should_defrag_range(struct inode *inode, u64 start, u32 thresh, 1012a43a2111SAndrew Mahone u64 *last_len, u64 *skip, u64 *defrag_end, 1013a43a2111SAndrew Mahone int compress) 10146c282eb4SLi Zefan { 10156c282eb4SLi Zefan struct extent_map *em; 10166c282eb4SLi Zefan int ret = 1; 10176c282eb4SLi Zefan bool next_mergeable = true; 10184a3560c4SLiu Bo bool prev_mergeable = true; 10196c282eb4SLi Zefan 10206c282eb4SLi Zefan /* 10216c282eb4SLi Zefan * make sure that once we start defragging an extent, we keep on 10226c282eb4SLi Zefan * defragging it 10236c282eb4SLi Zefan */ 10246c282eb4SLi Zefan if (start < *defrag_end) 10256c282eb4SLi Zefan return 1; 10266c282eb4SLi Zefan 10276c282eb4SLi Zefan *skip = 0; 10286c282eb4SLi Zefan 10296c282eb4SLi Zefan em = defrag_lookup_extent(inode, start); 10306c282eb4SLi Zefan if (!em) 10316c282eb4SLi Zefan return 0; 10326c282eb4SLi Zefan 1033940100a4SChris Mason /* this will cover holes, and inline extents */ 103417ce6ef8SLiu Bo if (em->block_start >= EXTENT_MAP_LAST_BYTE) { 1035940100a4SChris Mason ret = 0; 103617ce6ef8SLiu Bo goto out; 103717ce6ef8SLiu Bo } 103817ce6ef8SLiu Bo 10394a3560c4SLiu Bo if (!*defrag_end) 10404a3560c4SLiu Bo prev_mergeable = false; 10414a3560c4SLiu Bo 10426c282eb4SLi Zefan next_mergeable = defrag_check_next_extent(inode, em); 1043940100a4SChris Mason /* 10446c282eb4SLi Zefan * we hit a real extent, if it is big or the next extent is not a 10456c282eb4SLi Zefan * real extent, don't bother defragging it 1046940100a4SChris Mason */ 1047a43a2111SAndrew Mahone if (!compress && (*last_len == 0 || *last_len >= thresh) && 10484a3560c4SLiu Bo (em->len >= thresh || (!next_mergeable && !prev_mergeable))) 1049940100a4SChris Mason ret = 0; 105017ce6ef8SLiu Bo out: 1051940100a4SChris Mason /* 1052940100a4SChris Mason * last_len ends up being a counter of how many bytes we've defragged. 1053940100a4SChris Mason * every time we choose not to defrag an extent, we reset *last_len 1054940100a4SChris Mason * so that the next tiny extent will force a defrag. 1055940100a4SChris Mason * 1056940100a4SChris Mason * The end result of this is that tiny extents before a single big 1057940100a4SChris Mason * extent will force at least part of that big extent to be defragged. 1058940100a4SChris Mason */ 1059940100a4SChris Mason if (ret) { 1060940100a4SChris Mason *defrag_end = extent_map_end(em); 1061940100a4SChris Mason } else { 1062940100a4SChris Mason *last_len = 0; 1063940100a4SChris Mason *skip = extent_map_end(em); 1064940100a4SChris Mason *defrag_end = 0; 1065940100a4SChris Mason } 1066940100a4SChris Mason 1067940100a4SChris Mason free_extent_map(em); 1068940100a4SChris Mason return ret; 1069940100a4SChris Mason } 1070940100a4SChris Mason 10714cb5300bSChris Mason /* 10724cb5300bSChris Mason * it doesn't do much good to defrag one or two pages 10734cb5300bSChris Mason * at a time. This pulls in a nice chunk of pages 10744cb5300bSChris Mason * to COW and defrag. 10754cb5300bSChris Mason * 10764cb5300bSChris Mason * It also makes sure the delalloc code has enough 10774cb5300bSChris Mason * dirty data to avoid making new small extents as part 10784cb5300bSChris Mason * of the defrag 10794cb5300bSChris Mason * 10804cb5300bSChris Mason * It's a good idea to start RA on this range 10814cb5300bSChris Mason * before calling this. 10824cb5300bSChris Mason */ 10834cb5300bSChris Mason static int cluster_pages_for_defrag(struct inode *inode, 10844cb5300bSChris Mason struct page **pages, 10854cb5300bSChris Mason unsigned long start_index, 1086c41570c9SJustin Maggard unsigned long num_pages) 1087f46b5a66SChristoph Hellwig { 10884cb5300bSChris Mason unsigned long file_end; 10894cb5300bSChris Mason u64 isize = i_size_read(inode); 1090f46b5a66SChristoph Hellwig u64 page_start; 1091f46b5a66SChristoph Hellwig u64 page_end; 10921f12bd06SLiu Bo u64 page_cnt; 10934cb5300bSChris Mason int ret; 10944cb5300bSChris Mason int i; 10954cb5300bSChris Mason int i_done; 10964cb5300bSChris Mason struct btrfs_ordered_extent *ordered; 10974cb5300bSChris Mason struct extent_state *cached_state = NULL; 1098600a45e1SMiao Xie struct extent_io_tree *tree; 1099364ecf36SQu Wenruo struct extent_changeset *data_reserved = NULL; 11003b16a4e3SJosef Bacik gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping); 11014cb5300bSChris Mason 110209cbfeafSKirill A. Shutemov file_end = (isize - 1) >> PAGE_SHIFT; 11031f12bd06SLiu Bo if (!isize || start_index > file_end) 11041f12bd06SLiu Bo return 0; 11051f12bd06SLiu Bo 11061f12bd06SLiu Bo page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1); 11074cb5300bSChris Mason 1108364ecf36SQu Wenruo ret = btrfs_delalloc_reserve_space(inode, &data_reserved, 110909cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 111009cbfeafSKirill A. Shutemov page_cnt << PAGE_SHIFT); 11114cb5300bSChris Mason if (ret) 11124cb5300bSChris Mason return ret; 11134cb5300bSChris Mason i_done = 0; 1114600a45e1SMiao Xie tree = &BTRFS_I(inode)->io_tree; 11154cb5300bSChris Mason 11164cb5300bSChris Mason /* step one, lock all the pages */ 11171f12bd06SLiu Bo for (i = 0; i < page_cnt; i++) { 11184cb5300bSChris Mason struct page *page; 1119600a45e1SMiao Xie again: 1120a94733d0SJosef Bacik page = find_or_create_page(inode->i_mapping, 11213b16a4e3SJosef Bacik start_index + i, mask); 11224cb5300bSChris Mason if (!page) 11234cb5300bSChris Mason break; 11244cb5300bSChris Mason 1125600a45e1SMiao Xie page_start = page_offset(page); 112609cbfeafSKirill A. Shutemov page_end = page_start + PAGE_SIZE - 1; 1127600a45e1SMiao Xie while (1) { 1128308d9800SFilipe Manana lock_extent_bits(tree, page_start, page_end, 1129ff13db41SDavid Sterba &cached_state); 1130600a45e1SMiao Xie ordered = btrfs_lookup_ordered_extent(inode, 1131600a45e1SMiao Xie page_start); 1132308d9800SFilipe Manana unlock_extent_cached(tree, page_start, page_end, 1133308d9800SFilipe Manana &cached_state, GFP_NOFS); 1134600a45e1SMiao Xie if (!ordered) 1135600a45e1SMiao Xie break; 1136600a45e1SMiao Xie 1137600a45e1SMiao Xie unlock_page(page); 1138600a45e1SMiao Xie btrfs_start_ordered_extent(inode, ordered, 1); 1139600a45e1SMiao Xie btrfs_put_ordered_extent(ordered); 1140600a45e1SMiao Xie lock_page(page); 11411f12bd06SLiu Bo /* 11421f12bd06SLiu Bo * we unlocked the page above, so we need check if 11431f12bd06SLiu Bo * it was released or not. 11441f12bd06SLiu Bo */ 11451f12bd06SLiu Bo if (page->mapping != inode->i_mapping) { 11461f12bd06SLiu Bo unlock_page(page); 114709cbfeafSKirill A. Shutemov put_page(page); 11481f12bd06SLiu Bo goto again; 11491f12bd06SLiu Bo } 1150600a45e1SMiao Xie } 1151600a45e1SMiao Xie 11524cb5300bSChris Mason if (!PageUptodate(page)) { 11534cb5300bSChris Mason btrfs_readpage(NULL, page); 11544cb5300bSChris Mason lock_page(page); 11554cb5300bSChris Mason if (!PageUptodate(page)) { 11564cb5300bSChris Mason unlock_page(page); 115709cbfeafSKirill A. Shutemov put_page(page); 11584cb5300bSChris Mason ret = -EIO; 11594cb5300bSChris Mason break; 11604cb5300bSChris Mason } 11614cb5300bSChris Mason } 1162600a45e1SMiao Xie 1163600a45e1SMiao Xie if (page->mapping != inode->i_mapping) { 1164600a45e1SMiao Xie unlock_page(page); 116509cbfeafSKirill A. Shutemov put_page(page); 1166600a45e1SMiao Xie goto again; 1167600a45e1SMiao Xie } 1168600a45e1SMiao Xie 11694cb5300bSChris Mason pages[i] = page; 11704cb5300bSChris Mason i_done++; 11714cb5300bSChris Mason } 11724cb5300bSChris Mason if (!i_done || ret) 11734cb5300bSChris Mason goto out; 11744cb5300bSChris Mason 11754cb5300bSChris Mason if (!(inode->i_sb->s_flags & MS_ACTIVE)) 11764cb5300bSChris Mason goto out; 11774cb5300bSChris Mason 11784cb5300bSChris Mason /* 11794cb5300bSChris Mason * so now we have a nice long stream of locked 11804cb5300bSChris Mason * and up to date pages, lets wait on them 11814cb5300bSChris Mason */ 11824cb5300bSChris Mason for (i = 0; i < i_done; i++) 11834cb5300bSChris Mason wait_on_page_writeback(pages[i]); 11844cb5300bSChris Mason 11854cb5300bSChris Mason page_start = page_offset(pages[0]); 118609cbfeafSKirill A. Shutemov page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE; 11874cb5300bSChris Mason 11884cb5300bSChris Mason lock_extent_bits(&BTRFS_I(inode)->io_tree, 1189ff13db41SDavid Sterba page_start, page_end - 1, &cached_state); 11904cb5300bSChris Mason clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, 11914cb5300bSChris Mason page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC | 11929e8a4a8bSLiu Bo EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0, 11939e8a4a8bSLiu Bo &cached_state, GFP_NOFS); 11944cb5300bSChris Mason 11951f12bd06SLiu Bo if (i_done != page_cnt) { 11969e0baf60SJosef Bacik spin_lock(&BTRFS_I(inode)->lock); 11979e0baf60SJosef Bacik BTRFS_I(inode)->outstanding_extents++; 11989e0baf60SJosef Bacik spin_unlock(&BTRFS_I(inode)->lock); 1199bc42bda2SQu Wenruo btrfs_delalloc_release_space(inode, data_reserved, 120009cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 120109cbfeafSKirill A. Shutemov (page_cnt - i_done) << PAGE_SHIFT); 12024cb5300bSChris Mason } 12034cb5300bSChris Mason 12044cb5300bSChris Mason 12059e8a4a8bSLiu Bo set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1, 1206018ed4f7SDavid Sterba &cached_state); 12074cb5300bSChris Mason 12084cb5300bSChris Mason unlock_extent_cached(&BTRFS_I(inode)->io_tree, 12094cb5300bSChris Mason page_start, page_end - 1, &cached_state, 12104cb5300bSChris Mason GFP_NOFS); 12114cb5300bSChris Mason 12124cb5300bSChris Mason for (i = 0; i < i_done; i++) { 12134cb5300bSChris Mason clear_page_dirty_for_io(pages[i]); 12144cb5300bSChris Mason ClearPageChecked(pages[i]); 12154cb5300bSChris Mason set_page_extent_mapped(pages[i]); 12164cb5300bSChris Mason set_page_dirty(pages[i]); 12174cb5300bSChris Mason unlock_page(pages[i]); 121809cbfeafSKirill A. Shutemov put_page(pages[i]); 12194cb5300bSChris Mason } 1220364ecf36SQu Wenruo extent_changeset_free(data_reserved); 12214cb5300bSChris Mason return i_done; 12224cb5300bSChris Mason out: 12234cb5300bSChris Mason for (i = 0; i < i_done; i++) { 12244cb5300bSChris Mason unlock_page(pages[i]); 122509cbfeafSKirill A. Shutemov put_page(pages[i]); 12264cb5300bSChris Mason } 1227bc42bda2SQu Wenruo btrfs_delalloc_release_space(inode, data_reserved, 122809cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 122909cbfeafSKirill A. Shutemov page_cnt << PAGE_SHIFT); 1230364ecf36SQu Wenruo extent_changeset_free(data_reserved); 12314cb5300bSChris Mason return ret; 12324cb5300bSChris Mason 12334cb5300bSChris Mason } 12344cb5300bSChris Mason 12354cb5300bSChris Mason int btrfs_defrag_file(struct inode *inode, struct file *file, 12364cb5300bSChris Mason struct btrfs_ioctl_defrag_range_args *range, 12374cb5300bSChris Mason u64 newer_than, unsigned long max_to_defrag) 12384cb5300bSChris Mason { 12390b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 12404cb5300bSChris Mason struct btrfs_root *root = BTRFS_I(inode)->root; 12414cb5300bSChris Mason struct file_ra_state *ra = NULL; 12424cb5300bSChris Mason unsigned long last_index; 1243151a31b2SLi Zefan u64 isize = i_size_read(inode); 1244940100a4SChris Mason u64 last_len = 0; 1245940100a4SChris Mason u64 skip = 0; 1246940100a4SChris Mason u64 defrag_end = 0; 12474cb5300bSChris Mason u64 newer_off = range->start; 1248f46b5a66SChristoph Hellwig unsigned long i; 1249008873eaSLi Zefan unsigned long ra_index = 0; 1250f46b5a66SChristoph Hellwig int ret; 12514cb5300bSChris Mason int defrag_count = 0; 12521a419d85SLi Zefan int compress_type = BTRFS_COMPRESS_ZLIB; 1253aab110abSDavid Sterba u32 extent_thresh = range->extent_thresh; 125409cbfeafSKirill A. Shutemov unsigned long max_cluster = SZ_256K >> PAGE_SHIFT; 1255c41570c9SJustin Maggard unsigned long cluster = max_cluster; 1256ee22184bSByongho Lee u64 new_align = ~((u64)SZ_128K - 1); 12574cb5300bSChris Mason struct page **pages = NULL; 12584cb5300bSChris Mason 12590abd5b17SLiu Bo if (isize == 0) 12600abd5b17SLiu Bo return 0; 12610abd5b17SLiu Bo 12620abd5b17SLiu Bo if (range->start >= isize) 12630abd5b17SLiu Bo return -EINVAL; 12641a419d85SLi Zefan 12651a419d85SLi Zefan if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) { 12661a419d85SLi Zefan if (range->compress_type > BTRFS_COMPRESS_TYPES) 12671a419d85SLi Zefan return -EINVAL; 12681a419d85SLi Zefan if (range->compress_type) 12691a419d85SLi Zefan compress_type = range->compress_type; 12701a419d85SLi Zefan } 1271f46b5a66SChristoph Hellwig 12720abd5b17SLiu Bo if (extent_thresh == 0) 1273ee22184bSByongho Lee extent_thresh = SZ_256K; 1274f46b5a66SChristoph Hellwig 12754cb5300bSChris Mason /* 12764cb5300bSChris Mason * if we were not given a file, allocate a readahead 12774cb5300bSChris Mason * context 12784cb5300bSChris Mason */ 12794cb5300bSChris Mason if (!file) { 12804cb5300bSChris Mason ra = kzalloc(sizeof(*ra), GFP_NOFS); 12814cb5300bSChris Mason if (!ra) 12824cb5300bSChris Mason return -ENOMEM; 12834cb5300bSChris Mason file_ra_state_init(ra, inode->i_mapping); 12844cb5300bSChris Mason } else { 12854cb5300bSChris Mason ra = &file->f_ra; 12864cb5300bSChris Mason } 12874cb5300bSChris Mason 1288d9b0d9baSDulshani Gunawardhana pages = kmalloc_array(max_cluster, sizeof(struct page *), 12894cb5300bSChris Mason GFP_NOFS); 12904cb5300bSChris Mason if (!pages) { 12914cb5300bSChris Mason ret = -ENOMEM; 12924cb5300bSChris Mason goto out_ra; 12934cb5300bSChris Mason } 12944cb5300bSChris Mason 12954cb5300bSChris Mason /* find the last page to defrag */ 12961e701a32SChris Mason if (range->start + range->len > range->start) { 1297151a31b2SLi Zefan last_index = min_t(u64, isize - 1, 129809cbfeafSKirill A. Shutemov range->start + range->len - 1) >> PAGE_SHIFT; 12991e701a32SChris Mason } else { 130009cbfeafSKirill A. Shutemov last_index = (isize - 1) >> PAGE_SHIFT; 13011e701a32SChris Mason } 13021e701a32SChris Mason 13034cb5300bSChris Mason if (newer_than) { 13044cb5300bSChris Mason ret = find_new_extents(root, inode, newer_than, 1305ee22184bSByongho Lee &newer_off, SZ_64K); 13064cb5300bSChris Mason if (!ret) { 13074cb5300bSChris Mason range->start = newer_off; 13084cb5300bSChris Mason /* 13094cb5300bSChris Mason * we always align our defrag to help keep 13104cb5300bSChris Mason * the extents in the file evenly spaced 13114cb5300bSChris Mason */ 131209cbfeafSKirill A. Shutemov i = (newer_off & new_align) >> PAGE_SHIFT; 13134cb5300bSChris Mason } else 13144cb5300bSChris Mason goto out_ra; 13154cb5300bSChris Mason } else { 131609cbfeafSKirill A. Shutemov i = range->start >> PAGE_SHIFT; 13174cb5300bSChris Mason } 13184cb5300bSChris Mason if (!max_to_defrag) 1319070034bdSchandan max_to_defrag = last_index - i + 1; 13204cb5300bSChris Mason 13212a0f7f57SLi Zefan /* 13222a0f7f57SLi Zefan * make writeback starts from i, so the defrag range can be 13232a0f7f57SLi Zefan * written sequentially. 13242a0f7f57SLi Zefan */ 13252a0f7f57SLi Zefan if (i < inode->i_mapping->writeback_index) 13262a0f7f57SLi Zefan inode->i_mapping->writeback_index = i; 13272a0f7f57SLi Zefan 1328f7f43cc8SChris Mason while (i <= last_index && defrag_count < max_to_defrag && 132909cbfeafSKirill A. Shutemov (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) { 13304cb5300bSChris Mason /* 13314cb5300bSChris Mason * make sure we stop running if someone unmounts 13324cb5300bSChris Mason * the FS 13334cb5300bSChris Mason */ 13344cb5300bSChris Mason if (!(inode->i_sb->s_flags & MS_ACTIVE)) 13354cb5300bSChris Mason break; 13364cb5300bSChris Mason 13370b246afaSJeff Mahoney if (btrfs_defrag_cancelled(fs_info)) { 13380b246afaSJeff Mahoney btrfs_debug(fs_info, "defrag_file cancelled"); 1339210549ebSDavid Sterba ret = -EAGAIN; 1340210549ebSDavid Sterba break; 1341210549ebSDavid Sterba } 1342210549ebSDavid Sterba 134309cbfeafSKirill A. Shutemov if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT, 13446c282eb4SLi Zefan extent_thresh, &last_len, &skip, 1345a43a2111SAndrew Mahone &defrag_end, range->flags & 1346a43a2111SAndrew Mahone BTRFS_DEFRAG_RANGE_COMPRESS)) { 1347940100a4SChris Mason unsigned long next; 1348940100a4SChris Mason /* 1349940100a4SChris Mason * the should_defrag function tells us how much to skip 1350940100a4SChris Mason * bump our counter by the suggested amount 1351940100a4SChris Mason */ 135209cbfeafSKirill A. Shutemov next = DIV_ROUND_UP(skip, PAGE_SIZE); 1353940100a4SChris Mason i = max(i + 1, next); 1354940100a4SChris Mason continue; 1355940100a4SChris Mason } 1356008873eaSLi Zefan 1357008873eaSLi Zefan if (!newer_than) { 135809cbfeafSKirill A. Shutemov cluster = (PAGE_ALIGN(defrag_end) >> 135909cbfeafSKirill A. Shutemov PAGE_SHIFT) - i; 1360008873eaSLi Zefan cluster = min(cluster, max_cluster); 1361008873eaSLi Zefan } else { 1362008873eaSLi Zefan cluster = max_cluster; 1363008873eaSLi Zefan } 1364008873eaSLi Zefan 1365008873eaSLi Zefan if (i + cluster > ra_index) { 1366008873eaSLi Zefan ra_index = max(i, ra_index); 1367008873eaSLi Zefan btrfs_force_ra(inode->i_mapping, ra, file, ra_index, 1368008873eaSLi Zefan cluster); 1369e4826a5bSchandan ra_index += cluster; 1370008873eaSLi Zefan } 13714cb5300bSChris Mason 13725955102cSAl Viro inode_lock(inode); 1373633085c7SFilipe David Borba Manana if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) 1374633085c7SFilipe David Borba Manana BTRFS_I(inode)->force_compress = compress_type; 1375008873eaSLi Zefan ret = cluster_pages_for_defrag(inode, pages, i, cluster); 1376ecb8bea8SLiu Bo if (ret < 0) { 13775955102cSAl Viro inode_unlock(inode); 13784cb5300bSChris Mason goto out_ra; 1379ecb8bea8SLiu Bo } 13804cb5300bSChris Mason 13814cb5300bSChris Mason defrag_count += ret; 1382d0e1d66bSNamjae Jeon balance_dirty_pages_ratelimited(inode->i_mapping); 13835955102cSAl Viro inode_unlock(inode); 13844cb5300bSChris Mason 13854cb5300bSChris Mason if (newer_than) { 13864cb5300bSChris Mason if (newer_off == (u64)-1) 13874cb5300bSChris Mason break; 13884cb5300bSChris Mason 1389e1f041e1SLiu Bo if (ret > 0) 1390e1f041e1SLiu Bo i += ret; 1391e1f041e1SLiu Bo 13924cb5300bSChris Mason newer_off = max(newer_off + 1, 139309cbfeafSKirill A. Shutemov (u64)i << PAGE_SHIFT); 13944cb5300bSChris Mason 1395ee22184bSByongho Lee ret = find_new_extents(root, inode, newer_than, 1396ee22184bSByongho Lee &newer_off, SZ_64K); 13974cb5300bSChris Mason if (!ret) { 13984cb5300bSChris Mason range->start = newer_off; 139909cbfeafSKirill A. Shutemov i = (newer_off & new_align) >> PAGE_SHIFT; 14004cb5300bSChris Mason } else { 14014cb5300bSChris Mason break; 1402940100a4SChris Mason } 14034cb5300bSChris Mason } else { 1404008873eaSLi Zefan if (ret > 0) { 1405cbcc8326SLi Zefan i += ret; 140609cbfeafSKirill A. Shutemov last_len += ret << PAGE_SHIFT; 1407008873eaSLi Zefan } else { 1408940100a4SChris Mason i++; 1409008873eaSLi Zefan last_len = 0; 1410008873eaSLi Zefan } 1411f46b5a66SChristoph Hellwig } 14124cb5300bSChris Mason } 1413f46b5a66SChristoph Hellwig 1414dec8ef90SFilipe Manana if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) { 14151e701a32SChris Mason filemap_flush(inode->i_mapping); 1416dec8ef90SFilipe Manana if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT, 1417dec8ef90SFilipe Manana &BTRFS_I(inode)->runtime_flags)) 1418dec8ef90SFilipe Manana filemap_flush(inode->i_mapping); 1419dec8ef90SFilipe Manana } 14201e701a32SChris Mason 14211e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 14221e701a32SChris Mason /* the filemap_flush will queue IO into the worker threads, but 14231e701a32SChris Mason * we have to make sure the IO is actually started and that 14241e701a32SChris Mason * ordered extents get created before we return 14251e701a32SChris Mason */ 14260b246afaSJeff Mahoney atomic_inc(&fs_info->async_submit_draining); 14270b246afaSJeff Mahoney while (atomic_read(&fs_info->nr_async_submits) || 14280b246afaSJeff Mahoney atomic_read(&fs_info->async_delalloc_pages)) { 14290b246afaSJeff Mahoney wait_event(fs_info->async_submit_wait, 14300b246afaSJeff Mahoney (atomic_read(&fs_info->nr_async_submits) == 0 && 14310b246afaSJeff Mahoney atomic_read(&fs_info->async_delalloc_pages) == 0)); 14321e701a32SChris Mason } 14330b246afaSJeff Mahoney atomic_dec(&fs_info->async_submit_draining); 14341e701a32SChris Mason } 14351e701a32SChris Mason 14361a419d85SLi Zefan if (range->compress_type == BTRFS_COMPRESS_LZO) { 14370b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, COMPRESS_LZO); 14381a419d85SLi Zefan } 14391a419d85SLi Zefan 144060ccf82fSDiego Calleja ret = defrag_count; 1441940100a4SChris Mason 14424cb5300bSChris Mason out_ra: 1443633085c7SFilipe David Borba Manana if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) { 14445955102cSAl Viro inode_lock(inode); 1445633085c7SFilipe David Borba Manana BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE; 14465955102cSAl Viro inode_unlock(inode); 1447633085c7SFilipe David Borba Manana } 14484cb5300bSChris Mason if (!file) 14494cb5300bSChris Mason kfree(ra); 14504cb5300bSChris Mason kfree(pages); 1451940100a4SChris Mason return ret; 1452f46b5a66SChristoph Hellwig } 1453f46b5a66SChristoph Hellwig 1454198605a8SMiao Xie static noinline int btrfs_ioctl_resize(struct file *file, 145576dda93cSYan, Zheng void __user *arg) 1456f46b5a66SChristoph Hellwig { 14570b246afaSJeff Mahoney struct inode *inode = file_inode(file); 14580b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 1459f46b5a66SChristoph Hellwig u64 new_size; 1460f46b5a66SChristoph Hellwig u64 old_size; 1461f46b5a66SChristoph Hellwig u64 devid = 1; 14620b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 1463f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 1464f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 1465f46b5a66SChristoph Hellwig struct btrfs_device *device = NULL; 1466f46b5a66SChristoph Hellwig char *sizestr; 14679a40f122SGui Hecheng char *retptr; 1468f46b5a66SChristoph Hellwig char *devstr = NULL; 1469f46b5a66SChristoph Hellwig int ret = 0; 1470f46b5a66SChristoph Hellwig int mod = 0; 1471f46b5a66SChristoph Hellwig 1472e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 1473e441d54dSChris Mason return -EPERM; 1474e441d54dSChris Mason 1475198605a8SMiao Xie ret = mnt_want_write_file(file); 1476198605a8SMiao Xie if (ret) 1477198605a8SMiao Xie return ret; 1478198605a8SMiao Xie 1479171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 148097547676SMiao Xie mnt_drop_write_file(file); 1481e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 1482c9e9f97bSIlya Dryomov } 1483c9e9f97bSIlya Dryomov 14840b246afaSJeff Mahoney mutex_lock(&fs_info->volume_mutex); 1485dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1486c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 1487c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 1488c9e9f97bSIlya Dryomov goto out; 1489c9e9f97bSIlya Dryomov } 14905516e595SMark Fasheh 14915516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1492f46b5a66SChristoph Hellwig 1493f46b5a66SChristoph Hellwig sizestr = vol_args->name; 1494f46b5a66SChristoph Hellwig devstr = strchr(sizestr, ':'); 1495f46b5a66SChristoph Hellwig if (devstr) { 1496f46b5a66SChristoph Hellwig sizestr = devstr + 1; 1497f46b5a66SChristoph Hellwig *devstr = '\0'; 1498f46b5a66SChristoph Hellwig devstr = vol_args->name; 149958dfae63SZhangZhen ret = kstrtoull(devstr, 10, &devid); 150058dfae63SZhangZhen if (ret) 150158dfae63SZhangZhen goto out_free; 1502dfd79829SMiao Xie if (!devid) { 1503dfd79829SMiao Xie ret = -EINVAL; 1504dfd79829SMiao Xie goto out_free; 1505dfd79829SMiao Xie } 15060b246afaSJeff Mahoney btrfs_info(fs_info, "resizing devid %llu", devid); 1507f46b5a66SChristoph Hellwig } 1508dba60f3fSMiao Xie 15090b246afaSJeff Mahoney device = btrfs_find_device(fs_info, devid, NULL, NULL); 1510f46b5a66SChristoph Hellwig if (!device) { 15110b246afaSJeff Mahoney btrfs_info(fs_info, "resizer unable to find device %llu", 1512c1c9ff7cSGeert Uytterhoeven devid); 1513dfd79829SMiao Xie ret = -ENODEV; 1514c9e9f97bSIlya Dryomov goto out_free; 1515f46b5a66SChristoph Hellwig } 1516dba60f3fSMiao Xie 1517dba60f3fSMiao Xie if (!device->writeable) { 15180b246afaSJeff Mahoney btrfs_info(fs_info, 1519efe120a0SFrank Holton "resizer unable to apply on readonly device %llu", 1520c1c9ff7cSGeert Uytterhoeven devid); 1521dfd79829SMiao Xie ret = -EPERM; 15224e42ae1bSLiu Bo goto out_free; 15234e42ae1bSLiu Bo } 15244e42ae1bSLiu Bo 1525f46b5a66SChristoph Hellwig if (!strcmp(sizestr, "max")) 1526f46b5a66SChristoph Hellwig new_size = device->bdev->bd_inode->i_size; 1527f46b5a66SChristoph Hellwig else { 1528f46b5a66SChristoph Hellwig if (sizestr[0] == '-') { 1529f46b5a66SChristoph Hellwig mod = -1; 1530f46b5a66SChristoph Hellwig sizestr++; 1531f46b5a66SChristoph Hellwig } else if (sizestr[0] == '+') { 1532f46b5a66SChristoph Hellwig mod = 1; 1533f46b5a66SChristoph Hellwig sizestr++; 1534f46b5a66SChristoph Hellwig } 15359a40f122SGui Hecheng new_size = memparse(sizestr, &retptr); 15369a40f122SGui Hecheng if (*retptr != '\0' || new_size == 0) { 1537f46b5a66SChristoph Hellwig ret = -EINVAL; 1538c9e9f97bSIlya Dryomov goto out_free; 1539f46b5a66SChristoph Hellwig } 1540f46b5a66SChristoph Hellwig } 1541f46b5a66SChristoph Hellwig 154263a212abSStefan Behrens if (device->is_tgtdev_for_dev_replace) { 1543dfd79829SMiao Xie ret = -EPERM; 154463a212abSStefan Behrens goto out_free; 154563a212abSStefan Behrens } 154663a212abSStefan Behrens 15477cc8e58dSMiao Xie old_size = btrfs_device_get_total_bytes(device); 1548f46b5a66SChristoph Hellwig 1549f46b5a66SChristoph Hellwig if (mod < 0) { 1550f46b5a66SChristoph Hellwig if (new_size > old_size) { 1551f46b5a66SChristoph Hellwig ret = -EINVAL; 1552c9e9f97bSIlya Dryomov goto out_free; 1553f46b5a66SChristoph Hellwig } 1554f46b5a66SChristoph Hellwig new_size = old_size - new_size; 1555f46b5a66SChristoph Hellwig } else if (mod > 0) { 1556eb8052e0SWenliang Fan if (new_size > ULLONG_MAX - old_size) { 1557902c68a4SGui Hecheng ret = -ERANGE; 1558eb8052e0SWenliang Fan goto out_free; 1559eb8052e0SWenliang Fan } 1560f46b5a66SChristoph Hellwig new_size = old_size + new_size; 1561f46b5a66SChristoph Hellwig } 1562f46b5a66SChristoph Hellwig 1563ee22184bSByongho Lee if (new_size < SZ_256M) { 1564f46b5a66SChristoph Hellwig ret = -EINVAL; 1565c9e9f97bSIlya Dryomov goto out_free; 1566f46b5a66SChristoph Hellwig } 1567f46b5a66SChristoph Hellwig if (new_size > device->bdev->bd_inode->i_size) { 1568f46b5a66SChristoph Hellwig ret = -EFBIG; 1569c9e9f97bSIlya Dryomov goto out_free; 1570f46b5a66SChristoph Hellwig } 1571f46b5a66SChristoph Hellwig 157247f08b96SNikolay Borisov new_size = round_down(new_size, fs_info->sectorsize); 1573f46b5a66SChristoph Hellwig 15740b246afaSJeff Mahoney btrfs_info_in_rcu(fs_info, "new size for %s is %llu", 1575c1c9ff7cSGeert Uytterhoeven rcu_str_deref(device->name), new_size); 1576f46b5a66SChristoph Hellwig 1577f46b5a66SChristoph Hellwig if (new_size > old_size) { 1578a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 157998d5dc13STsutomu Itoh if (IS_ERR(trans)) { 158098d5dc13STsutomu Itoh ret = PTR_ERR(trans); 1581c9e9f97bSIlya Dryomov goto out_free; 158298d5dc13STsutomu Itoh } 1583f46b5a66SChristoph Hellwig ret = btrfs_grow_device(trans, device, new_size); 15843a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 1585ece7d20eSMike Fleetwood } else if (new_size < old_size) { 1586f46b5a66SChristoph Hellwig ret = btrfs_shrink_device(device, new_size); 15870253f40eSjeff.liu } /* equal, nothing need to do */ 1588f46b5a66SChristoph Hellwig 1589c9e9f97bSIlya Dryomov out_free: 1590f46b5a66SChristoph Hellwig kfree(vol_args); 1591c9e9f97bSIlya Dryomov out: 15920b246afaSJeff Mahoney mutex_unlock(&fs_info->volume_mutex); 1593171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 159418f39c41SIlya Dryomov mnt_drop_write_file(file); 1595f46b5a66SChristoph Hellwig return ret; 1596f46b5a66SChristoph Hellwig } 1597f46b5a66SChristoph Hellwig 159872fd032eSSage Weil static noinline int btrfs_ioctl_snap_create_transid(struct file *file, 159952f75f4fSDavid Sterba const char *name, unsigned long fd, int subvol, 16006f72c7e2SArne Jansen u64 *transid, bool readonly, 16018696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 1602f46b5a66SChristoph Hellwig { 1603f46b5a66SChristoph Hellwig int namelen; 16043de4586cSChris Mason int ret = 0; 1605f46b5a66SChristoph Hellwig 1606325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1607325c50e3SJeff Mahoney return -ENOTDIR; 1608325c50e3SJeff Mahoney 1609a874a63eSLiu Bo ret = mnt_want_write_file(file); 1610a874a63eSLiu Bo if (ret) 1611a874a63eSLiu Bo goto out; 1612a874a63eSLiu Bo 161372fd032eSSage Weil namelen = strlen(name); 161472fd032eSSage Weil if (strchr(name, '/')) { 1615f46b5a66SChristoph Hellwig ret = -EINVAL; 1616a874a63eSLiu Bo goto out_drop_write; 1617f46b5a66SChristoph Hellwig } 1618f46b5a66SChristoph Hellwig 161916780cabSChris Mason if (name[0] == '.' && 162016780cabSChris Mason (namelen == 1 || (name[1] == '.' && namelen == 2))) { 162116780cabSChris Mason ret = -EEXIST; 1622a874a63eSLiu Bo goto out_drop_write; 162316780cabSChris Mason } 162416780cabSChris Mason 16253de4586cSChris Mason if (subvol) { 162672fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 16276f72c7e2SArne Jansen NULL, transid, readonly, inherit); 1628cb8e7090SChristoph Hellwig } else { 16292903ff01SAl Viro struct fd src = fdget(fd); 16303de4586cSChris Mason struct inode *src_inode; 16312903ff01SAl Viro if (!src.file) { 16323de4586cSChris Mason ret = -EINVAL; 1633a874a63eSLiu Bo goto out_drop_write; 16343de4586cSChris Mason } 16353de4586cSChris Mason 1636496ad9aaSAl Viro src_inode = file_inode(src.file); 1637496ad9aaSAl Viro if (src_inode->i_sb != file_inode(file)->i_sb) { 1638c79b4713SJosef Bacik btrfs_info(BTRFS_I(file_inode(file))->root->fs_info, 1639efe120a0SFrank Holton "Snapshot src from another FS"); 164023ad5b17SKusanagi Kouichi ret = -EXDEV; 1641d0242061SDavid Sterba } else if (!inode_owner_or_capable(src_inode)) { 1642d0242061SDavid Sterba /* 1643d0242061SDavid Sterba * Subvolume creation is not restricted, but snapshots 1644d0242061SDavid Sterba * are limited to own subvolumes only 1645d0242061SDavid Sterba */ 1646d0242061SDavid Sterba ret = -EPERM; 1647ecd18815SAl Viro } else { 164872fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 164972fd032eSSage Weil BTRFS_I(src_inode)->root, 16506f72c7e2SArne Jansen transid, readonly, inherit); 1651ecd18815SAl Viro } 16522903ff01SAl Viro fdput(src); 1653cb8e7090SChristoph Hellwig } 1654a874a63eSLiu Bo out_drop_write: 1655a874a63eSLiu Bo mnt_drop_write_file(file); 1656f46b5a66SChristoph Hellwig out: 165772fd032eSSage Weil return ret; 165872fd032eSSage Weil } 165972fd032eSSage Weil 166072fd032eSSage Weil static noinline int btrfs_ioctl_snap_create(struct file *file, 1661fa0d2b9bSLi Zefan void __user *arg, int subvol) 166272fd032eSSage Weil { 1663fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args *vol_args; 166472fd032eSSage Weil int ret; 166572fd032eSSage Weil 1666325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1667325c50e3SJeff Mahoney return -ENOTDIR; 1668325c50e3SJeff Mahoney 1669fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1670fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1671fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1672fa0d2b9bSLi Zefan vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1673fa0d2b9bSLi Zefan 1674fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 1675b83cc969SLi Zefan vol_args->fd, subvol, 16766f72c7e2SArne Jansen NULL, false, NULL); 1677fa0d2b9bSLi Zefan 1678fa0d2b9bSLi Zefan kfree(vol_args); 1679fa0d2b9bSLi Zefan return ret; 1680fa0d2b9bSLi Zefan } 1681fa0d2b9bSLi Zefan 1682fa0d2b9bSLi Zefan static noinline int btrfs_ioctl_snap_create_v2(struct file *file, 1683fa0d2b9bSLi Zefan void __user *arg, int subvol) 1684fa0d2b9bSLi Zefan { 1685fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args_v2 *vol_args; 1686fa0d2b9bSLi Zefan int ret; 1687fdfb1e4fSLi Zefan u64 transid = 0; 1688fdfb1e4fSLi Zefan u64 *ptr = NULL; 1689b83cc969SLi Zefan bool readonly = false; 16906f72c7e2SArne Jansen struct btrfs_qgroup_inherit *inherit = NULL; 169172fd032eSSage Weil 1692325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1693325c50e3SJeff Mahoney return -ENOTDIR; 1694325c50e3SJeff Mahoney 1695fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1696fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1697fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1698fa0d2b9bSLi Zefan vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 1699fdfb1e4fSLi Zefan 1700b83cc969SLi Zefan if (vol_args->flags & 17016f72c7e2SArne Jansen ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY | 17026f72c7e2SArne Jansen BTRFS_SUBVOL_QGROUP_INHERIT)) { 1703b83cc969SLi Zefan ret = -EOPNOTSUPP; 1704c47ca32dSDan Carpenter goto free_args; 1705fdfb1e4fSLi Zefan } 1706fdfb1e4fSLi Zefan 1707fa0d2b9bSLi Zefan if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC) 1708fdfb1e4fSLi Zefan ptr = &transid; 1709b83cc969SLi Zefan if (vol_args->flags & BTRFS_SUBVOL_RDONLY) 1710b83cc969SLi Zefan readonly = true; 17116f72c7e2SArne Jansen if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) { 171209cbfeafSKirill A. Shutemov if (vol_args->size > PAGE_SIZE) { 17136f72c7e2SArne Jansen ret = -EINVAL; 1714c47ca32dSDan Carpenter goto free_args; 17156f72c7e2SArne Jansen } 17166f72c7e2SArne Jansen inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size); 17176f72c7e2SArne Jansen if (IS_ERR(inherit)) { 17186f72c7e2SArne Jansen ret = PTR_ERR(inherit); 1719c47ca32dSDan Carpenter goto free_args; 17206f72c7e2SArne Jansen } 17216f72c7e2SArne Jansen } 172275eaa0e2SSage Weil 1723fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 17246f72c7e2SArne Jansen vol_args->fd, subvol, ptr, 17258696c533SMiao Xie readonly, inherit); 1726c47ca32dSDan Carpenter if (ret) 1727c47ca32dSDan Carpenter goto free_inherit; 172875eaa0e2SSage Weil 1729c47ca32dSDan Carpenter if (ptr && copy_to_user(arg + 1730fdfb1e4fSLi Zefan offsetof(struct btrfs_ioctl_vol_args_v2, 1731c47ca32dSDan Carpenter transid), 1732c47ca32dSDan Carpenter ptr, sizeof(*ptr))) 173375eaa0e2SSage Weil ret = -EFAULT; 1734c47ca32dSDan Carpenter 1735c47ca32dSDan Carpenter free_inherit: 17366f72c7e2SArne Jansen kfree(inherit); 1737c47ca32dSDan Carpenter free_args: 1738c47ca32dSDan Carpenter kfree(vol_args); 1739f46b5a66SChristoph Hellwig return ret; 1740f46b5a66SChristoph Hellwig } 1741f46b5a66SChristoph Hellwig 17420caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_getflags(struct file *file, 17430caa102dSLi Zefan void __user *arg) 17440caa102dSLi Zefan { 1745496ad9aaSAl Viro struct inode *inode = file_inode(file); 17460b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 17470caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 17480caa102dSLi Zefan int ret = 0; 17490caa102dSLi Zefan u64 flags = 0; 17500caa102dSLi Zefan 17514a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) 17520caa102dSLi Zefan return -EINVAL; 17530caa102dSLi Zefan 17540b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 17550caa102dSLi Zefan if (btrfs_root_readonly(root)) 17560caa102dSLi Zefan flags |= BTRFS_SUBVOL_RDONLY; 17570b246afaSJeff Mahoney up_read(&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); 17690b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 17700caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 17710caa102dSLi Zefan struct btrfs_trans_handle *trans; 17720caa102dSLi Zefan u64 root_flags; 17730caa102dSLi Zefan u64 flags; 17740caa102dSLi Zefan int ret = 0; 17750caa102dSLi Zefan 1776bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 1777bd60ea0fSDavid Sterba return -EPERM; 1778bd60ea0fSDavid Sterba 1779b9ca0664SLiu Bo ret = mnt_want_write_file(file); 1780b9ca0664SLiu Bo if (ret) 1781b9ca0664SLiu Bo goto out; 17820caa102dSLi Zefan 17834a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 1784b9ca0664SLiu Bo ret = -EINVAL; 1785b9ca0664SLiu Bo goto out_drop_write; 1786b9ca0664SLiu Bo } 17870caa102dSLi Zefan 1788b9ca0664SLiu Bo if (copy_from_user(&flags, arg, sizeof(flags))) { 1789b9ca0664SLiu Bo ret = -EFAULT; 1790b9ca0664SLiu Bo goto out_drop_write; 1791b9ca0664SLiu Bo } 17920caa102dSLi Zefan 1793b9ca0664SLiu Bo if (flags & BTRFS_SUBVOL_CREATE_ASYNC) { 1794b9ca0664SLiu Bo ret = -EINVAL; 1795b9ca0664SLiu Bo goto out_drop_write; 1796b9ca0664SLiu Bo } 17970caa102dSLi Zefan 1798b9ca0664SLiu Bo if (flags & ~BTRFS_SUBVOL_RDONLY) { 1799b9ca0664SLiu Bo ret = -EOPNOTSUPP; 1800b9ca0664SLiu Bo goto out_drop_write; 1801b9ca0664SLiu Bo } 18020caa102dSLi Zefan 18030b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 18040caa102dSLi Zefan 18050caa102dSLi Zefan /* nothing to do */ 18060caa102dSLi Zefan if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) 1807b9ca0664SLiu Bo goto out_drop_sem; 18080caa102dSLi Zefan 18090caa102dSLi Zefan root_flags = btrfs_root_flags(&root->root_item); 18102c686537SDavid Sterba if (flags & BTRFS_SUBVOL_RDONLY) { 18110caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 18120caa102dSLi Zefan root_flags | BTRFS_ROOT_SUBVOL_RDONLY); 18132c686537SDavid Sterba } else { 18142c686537SDavid Sterba /* 18152c686537SDavid Sterba * Block RO -> RW transition if this subvolume is involved in 18162c686537SDavid Sterba * send 18172c686537SDavid Sterba */ 18182c686537SDavid Sterba spin_lock(&root->root_item_lock); 18192c686537SDavid Sterba if (root->send_in_progress == 0) { 18200caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 18210caa102dSLi Zefan root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); 18222c686537SDavid Sterba spin_unlock(&root->root_item_lock); 18232c686537SDavid Sterba } else { 18242c686537SDavid Sterba spin_unlock(&root->root_item_lock); 18250b246afaSJeff Mahoney btrfs_warn(fs_info, 18262c686537SDavid Sterba "Attempt to set subvolume %llu read-write during send", 18272c686537SDavid Sterba root->root_key.objectid); 18282c686537SDavid Sterba ret = -EPERM; 18292c686537SDavid Sterba goto out_drop_sem; 18302c686537SDavid Sterba } 18312c686537SDavid Sterba } 18320caa102dSLi Zefan 18330caa102dSLi Zefan trans = btrfs_start_transaction(root, 1); 18340caa102dSLi Zefan if (IS_ERR(trans)) { 18350caa102dSLi Zefan ret = PTR_ERR(trans); 18360caa102dSLi Zefan goto out_reset; 18370caa102dSLi Zefan } 18380caa102dSLi Zefan 18390b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 18400caa102dSLi Zefan &root->root_key, &root->root_item); 18410caa102dSLi Zefan 18423a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 18430caa102dSLi Zefan out_reset: 18440caa102dSLi Zefan if (ret) 18450caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, root_flags); 1846b9ca0664SLiu Bo out_drop_sem: 18470b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 1848b9ca0664SLiu Bo out_drop_write: 1849b9ca0664SLiu Bo mnt_drop_write_file(file); 1850b9ca0664SLiu Bo out: 18510caa102dSLi Zefan return ret; 18520caa102dSLi Zefan } 18530caa102dSLi Zefan 185476dda93cSYan, Zheng /* 185576dda93cSYan, Zheng * helper to check if the subvolume references other subvolumes 185676dda93cSYan, Zheng */ 185776dda93cSYan, Zheng static noinline int may_destroy_subvol(struct btrfs_root *root) 185876dda93cSYan, Zheng { 18590b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 186076dda93cSYan, Zheng struct btrfs_path *path; 1861175a2b87SJosef Bacik struct btrfs_dir_item *di; 186276dda93cSYan, Zheng struct btrfs_key key; 1863175a2b87SJosef Bacik u64 dir_id; 186476dda93cSYan, Zheng int ret; 186576dda93cSYan, Zheng 186676dda93cSYan, Zheng path = btrfs_alloc_path(); 186776dda93cSYan, Zheng if (!path) 186876dda93cSYan, Zheng return -ENOMEM; 186976dda93cSYan, Zheng 1870175a2b87SJosef Bacik /* Make sure this root isn't set as the default subvol */ 18710b246afaSJeff Mahoney dir_id = btrfs_super_root_dir(fs_info->super_copy); 18720b246afaSJeff Mahoney di = btrfs_lookup_dir_item(NULL, fs_info->tree_root, path, 1873175a2b87SJosef Bacik dir_id, "default", 7, 0); 1874175a2b87SJosef Bacik if (di && !IS_ERR(di)) { 1875175a2b87SJosef Bacik btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key); 1876175a2b87SJosef Bacik if (key.objectid == root->root_key.objectid) { 187772de6b53SGuangyu Sun ret = -EPERM; 18780b246afaSJeff Mahoney btrfs_err(fs_info, 18795d163e0eSJeff Mahoney "deleting default subvolume %llu is not allowed", 18805d163e0eSJeff Mahoney key.objectid); 1881175a2b87SJosef Bacik goto out; 1882175a2b87SJosef Bacik } 1883175a2b87SJosef Bacik btrfs_release_path(path); 1884175a2b87SJosef Bacik } 1885175a2b87SJosef Bacik 188676dda93cSYan, Zheng key.objectid = root->root_key.objectid; 188776dda93cSYan, Zheng key.type = BTRFS_ROOT_REF_KEY; 188876dda93cSYan, Zheng key.offset = (u64)-1; 188976dda93cSYan, Zheng 18900b246afaSJeff Mahoney ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 189176dda93cSYan, Zheng if (ret < 0) 189276dda93cSYan, Zheng goto out; 189376dda93cSYan, Zheng BUG_ON(ret == 0); 189476dda93cSYan, Zheng 189576dda93cSYan, Zheng ret = 0; 189676dda93cSYan, Zheng if (path->slots[0] > 0) { 189776dda93cSYan, Zheng path->slots[0]--; 189876dda93cSYan, Zheng btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 189976dda93cSYan, Zheng if (key.objectid == root->root_key.objectid && 190076dda93cSYan, Zheng key.type == BTRFS_ROOT_REF_KEY) 190176dda93cSYan, Zheng ret = -ENOTEMPTY; 190276dda93cSYan, Zheng } 190376dda93cSYan, Zheng out: 190476dda93cSYan, Zheng btrfs_free_path(path); 190576dda93cSYan, Zheng return ret; 190676dda93cSYan, Zheng } 190776dda93cSYan, Zheng 1908ac8e9819SChris Mason static noinline int key_in_sk(struct btrfs_key *key, 1909ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk) 1910ac8e9819SChris Mason { 1911abc6e134SChris Mason struct btrfs_key test; 1912abc6e134SChris Mason int ret; 1913abc6e134SChris Mason 1914abc6e134SChris Mason test.objectid = sk->min_objectid; 1915abc6e134SChris Mason test.type = sk->min_type; 1916abc6e134SChris Mason test.offset = sk->min_offset; 1917abc6e134SChris Mason 1918abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1919abc6e134SChris Mason if (ret < 0) 1920ac8e9819SChris Mason return 0; 1921abc6e134SChris Mason 1922abc6e134SChris Mason test.objectid = sk->max_objectid; 1923abc6e134SChris Mason test.type = sk->max_type; 1924abc6e134SChris Mason test.offset = sk->max_offset; 1925abc6e134SChris Mason 1926abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1927abc6e134SChris Mason if (ret > 0) 1928ac8e9819SChris Mason return 0; 1929ac8e9819SChris Mason return 1; 1930ac8e9819SChris Mason } 1931ac8e9819SChris Mason 1932df397565SJeff Mahoney static noinline int copy_to_sk(struct btrfs_path *path, 1933ac8e9819SChris Mason struct btrfs_key *key, 1934ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk, 19359b6e817dSGerhard Heift size_t *buf_size, 1936ba346b35SGerhard Heift char __user *ubuf, 1937ac8e9819SChris Mason unsigned long *sk_offset, 1938ac8e9819SChris Mason int *num_found) 1939ac8e9819SChris Mason { 1940ac8e9819SChris Mason u64 found_transid; 1941ac8e9819SChris Mason struct extent_buffer *leaf; 1942ac8e9819SChris Mason struct btrfs_ioctl_search_header sh; 1943dd81d459SNaohiro Aota struct btrfs_key test; 1944ac8e9819SChris Mason unsigned long item_off; 1945ac8e9819SChris Mason unsigned long item_len; 1946ac8e9819SChris Mason int nritems; 1947ac8e9819SChris Mason int i; 1948ac8e9819SChris Mason int slot; 1949ac8e9819SChris Mason int ret = 0; 1950ac8e9819SChris Mason 1951ac8e9819SChris Mason leaf = path->nodes[0]; 1952ac8e9819SChris Mason slot = path->slots[0]; 1953ac8e9819SChris Mason nritems = btrfs_header_nritems(leaf); 1954ac8e9819SChris Mason 1955ac8e9819SChris Mason if (btrfs_header_generation(leaf) > sk->max_transid) { 1956ac8e9819SChris Mason i = nritems; 1957ac8e9819SChris Mason goto advance_key; 1958ac8e9819SChris Mason } 1959ac8e9819SChris Mason found_transid = btrfs_header_generation(leaf); 1960ac8e9819SChris Mason 1961ac8e9819SChris Mason for (i = slot; i < nritems; i++) { 1962ac8e9819SChris Mason item_off = btrfs_item_ptr_offset(leaf, i); 1963ac8e9819SChris Mason item_len = btrfs_item_size_nr(leaf, i); 1964ac8e9819SChris Mason 196503b71c6cSGabriel de Perthuis btrfs_item_key_to_cpu(leaf, key, i); 196603b71c6cSGabriel de Perthuis if (!key_in_sk(key, sk)) 196703b71c6cSGabriel de Perthuis continue; 196803b71c6cSGabriel de Perthuis 19699b6e817dSGerhard Heift if (sizeof(sh) + item_len > *buf_size) { 19708f5f6178SGerhard Heift if (*num_found) { 1971ac8e9819SChris Mason ret = 1; 19728f5f6178SGerhard Heift goto out; 19738f5f6178SGerhard Heift } 19748f5f6178SGerhard Heift 19758f5f6178SGerhard Heift /* 19768f5f6178SGerhard Heift * return one empty item back for v1, which does not 19778f5f6178SGerhard Heift * handle -EOVERFLOW 19788f5f6178SGerhard Heift */ 19798f5f6178SGerhard Heift 19809b6e817dSGerhard Heift *buf_size = sizeof(sh) + item_len; 1981ac8e9819SChris Mason item_len = 0; 19828f5f6178SGerhard Heift ret = -EOVERFLOW; 19838f5f6178SGerhard Heift } 1984ac8e9819SChris Mason 19859b6e817dSGerhard Heift if (sizeof(sh) + item_len + *sk_offset > *buf_size) { 1986ac8e9819SChris Mason ret = 1; 198725c9bc2eSGerhard Heift goto out; 1988ac8e9819SChris Mason } 1989ac8e9819SChris Mason 1990ac8e9819SChris Mason sh.objectid = key->objectid; 1991ac8e9819SChris Mason sh.offset = key->offset; 1992ac8e9819SChris Mason sh.type = key->type; 1993ac8e9819SChris Mason sh.len = item_len; 1994ac8e9819SChris Mason sh.transid = found_transid; 1995ac8e9819SChris Mason 1996ac8e9819SChris Mason /* copy search result header */ 1997ba346b35SGerhard Heift if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) { 1998ba346b35SGerhard Heift ret = -EFAULT; 1999ba346b35SGerhard Heift goto out; 2000ba346b35SGerhard Heift } 2001ba346b35SGerhard Heift 2002ac8e9819SChris Mason *sk_offset += sizeof(sh); 2003ac8e9819SChris Mason 2004ac8e9819SChris Mason if (item_len) { 2005ba346b35SGerhard Heift char __user *up = ubuf + *sk_offset; 2006ac8e9819SChris Mason /* copy the item */ 2007ba346b35SGerhard Heift if (read_extent_buffer_to_user(leaf, up, 2008ba346b35SGerhard Heift item_off, item_len)) { 2009ba346b35SGerhard Heift ret = -EFAULT; 2010ba346b35SGerhard Heift goto out; 2011ba346b35SGerhard Heift } 2012ba346b35SGerhard Heift 2013ac8e9819SChris Mason *sk_offset += item_len; 2014ac8e9819SChris Mason } 2015e2156867SHugo Mills (*num_found)++; 2016ac8e9819SChris Mason 20178f5f6178SGerhard Heift if (ret) /* -EOVERFLOW from above */ 20188f5f6178SGerhard Heift goto out; 20198f5f6178SGerhard Heift 202025c9bc2eSGerhard Heift if (*num_found >= sk->nr_items) { 202125c9bc2eSGerhard Heift ret = 1; 202225c9bc2eSGerhard Heift goto out; 202325c9bc2eSGerhard Heift } 2024ac8e9819SChris Mason } 2025ac8e9819SChris Mason advance_key: 2026ac8e9819SChris Mason ret = 0; 2027dd81d459SNaohiro Aota test.objectid = sk->max_objectid; 2028dd81d459SNaohiro Aota test.type = sk->max_type; 2029dd81d459SNaohiro Aota test.offset = sk->max_offset; 2030dd81d459SNaohiro Aota if (btrfs_comp_cpu_keys(key, &test) >= 0) 2031dd81d459SNaohiro Aota ret = 1; 2032dd81d459SNaohiro Aota else if (key->offset < (u64)-1) 2033abc6e134SChris Mason key->offset++; 2034dd81d459SNaohiro Aota else if (key->type < (u8)-1) { 2035abc6e134SChris Mason key->offset = 0; 2036abc6e134SChris Mason key->type++; 2037dd81d459SNaohiro Aota } else if (key->objectid < (u64)-1) { 2038abc6e134SChris Mason key->offset = 0; 2039abc6e134SChris Mason key->type = 0; 2040abc6e134SChris Mason key->objectid++; 2041abc6e134SChris Mason } else 2042abc6e134SChris Mason ret = 1; 204325c9bc2eSGerhard Heift out: 2044ba346b35SGerhard Heift /* 2045ba346b35SGerhard Heift * 0: all items from this leaf copied, continue with next 2046ba346b35SGerhard Heift * 1: * more items can be copied, but unused buffer is too small 2047ba346b35SGerhard Heift * * all items were found 2048ba346b35SGerhard Heift * Either way, it will stops the loop which iterates to the next 2049ba346b35SGerhard Heift * leaf 2050ba346b35SGerhard Heift * -EOVERFLOW: item was to large for buffer 2051ba346b35SGerhard Heift * -EFAULT: could not copy extent buffer back to userspace 2052ba346b35SGerhard Heift */ 2053ac8e9819SChris Mason return ret; 2054ac8e9819SChris Mason } 2055ac8e9819SChris Mason 2056ac8e9819SChris Mason static noinline int search_ioctl(struct inode *inode, 205712544442SGerhard Heift struct btrfs_ioctl_search_key *sk, 20589b6e817dSGerhard Heift size_t *buf_size, 2059ba346b35SGerhard Heift char __user *ubuf) 2060ac8e9819SChris Mason { 20610b246afaSJeff Mahoney struct btrfs_fs_info *info = btrfs_sb(inode->i_sb); 2062ac8e9819SChris Mason struct btrfs_root *root; 2063ac8e9819SChris Mason struct btrfs_key key; 2064ac8e9819SChris Mason struct btrfs_path *path; 2065ac8e9819SChris Mason int ret; 2066ac8e9819SChris Mason int num_found = 0; 2067ac8e9819SChris Mason unsigned long sk_offset = 0; 2068ac8e9819SChris Mason 20699b6e817dSGerhard Heift if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) { 20709b6e817dSGerhard Heift *buf_size = sizeof(struct btrfs_ioctl_search_header); 207112544442SGerhard Heift return -EOVERFLOW; 20729b6e817dSGerhard Heift } 207312544442SGerhard Heift 2074ac8e9819SChris Mason path = btrfs_alloc_path(); 2075ac8e9819SChris Mason if (!path) 2076ac8e9819SChris Mason return -ENOMEM; 2077ac8e9819SChris Mason 2078ac8e9819SChris Mason if (sk->tree_id == 0) { 2079ac8e9819SChris Mason /* search the root of the inode that was passed */ 2080ac8e9819SChris Mason root = BTRFS_I(inode)->root; 2081ac8e9819SChris Mason } else { 2082ac8e9819SChris Mason key.objectid = sk->tree_id; 2083ac8e9819SChris Mason key.type = BTRFS_ROOT_ITEM_KEY; 2084ac8e9819SChris Mason key.offset = (u64)-1; 2085ac8e9819SChris Mason root = btrfs_read_fs_root_no_name(info, &key); 2086ac8e9819SChris Mason if (IS_ERR(root)) { 2087ac8e9819SChris Mason btrfs_free_path(path); 2088ac8e9819SChris Mason return -ENOENT; 2089ac8e9819SChris Mason } 2090ac8e9819SChris Mason } 2091ac8e9819SChris Mason 2092ac8e9819SChris Mason key.objectid = sk->min_objectid; 2093ac8e9819SChris Mason key.type = sk->min_type; 2094ac8e9819SChris Mason key.offset = sk->min_offset; 2095ac8e9819SChris Mason 2096ac8e9819SChris Mason while (1) { 20976174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &key, path, sk->min_transid); 2098ac8e9819SChris Mason if (ret != 0) { 2099ac8e9819SChris Mason if (ret > 0) 2100ac8e9819SChris Mason ret = 0; 2101ac8e9819SChris Mason goto err; 2102ac8e9819SChris Mason } 2103df397565SJeff Mahoney ret = copy_to_sk(path, &key, sk, buf_size, ubuf, 2104ac8e9819SChris Mason &sk_offset, &num_found); 2105b3b4aa74SDavid Sterba btrfs_release_path(path); 210625c9bc2eSGerhard Heift if (ret) 2107ac8e9819SChris Mason break; 2108ac8e9819SChris Mason 2109ac8e9819SChris Mason } 21108f5f6178SGerhard Heift if (ret > 0) 2111ac8e9819SChris Mason ret = 0; 2112ac8e9819SChris Mason err: 2113ac8e9819SChris Mason sk->nr_items = num_found; 2114ac8e9819SChris Mason btrfs_free_path(path); 2115ac8e9819SChris Mason return ret; 2116ac8e9819SChris Mason } 2117ac8e9819SChris Mason 2118ac8e9819SChris Mason static noinline int btrfs_ioctl_tree_search(struct file *file, 2119ac8e9819SChris Mason void __user *argp) 2120ac8e9819SChris Mason { 2121ba346b35SGerhard Heift struct btrfs_ioctl_search_args __user *uargs; 2122ba346b35SGerhard Heift struct btrfs_ioctl_search_key sk; 2123ac8e9819SChris Mason struct inode *inode; 2124ac8e9819SChris Mason int ret; 21259b6e817dSGerhard Heift size_t buf_size; 2126ac8e9819SChris Mason 2127ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 2128ac8e9819SChris Mason return -EPERM; 2129ac8e9819SChris Mason 2130ba346b35SGerhard Heift uargs = (struct btrfs_ioctl_search_args __user *)argp; 2131ac8e9819SChris Mason 2132ba346b35SGerhard Heift if (copy_from_user(&sk, &uargs->key, sizeof(sk))) 2133ba346b35SGerhard Heift return -EFAULT; 2134ba346b35SGerhard Heift 2135ba346b35SGerhard Heift buf_size = sizeof(uargs->buf); 2136ac8e9819SChris Mason 2137496ad9aaSAl Viro inode = file_inode(file); 2138ba346b35SGerhard Heift ret = search_ioctl(inode, &sk, &buf_size, uargs->buf); 21398f5f6178SGerhard Heift 21408f5f6178SGerhard Heift /* 21418f5f6178SGerhard Heift * In the origin implementation an overflow is handled by returning a 21428f5f6178SGerhard Heift * search header with a len of zero, so reset ret. 21438f5f6178SGerhard Heift */ 21448f5f6178SGerhard Heift if (ret == -EOVERFLOW) 21458f5f6178SGerhard Heift ret = 0; 21468f5f6178SGerhard Heift 2147ba346b35SGerhard Heift if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk))) 2148ac8e9819SChris Mason ret = -EFAULT; 2149ac8e9819SChris Mason return ret; 2150ac8e9819SChris Mason } 2151ac8e9819SChris Mason 2152cc68a8a5SGerhard Heift static noinline int btrfs_ioctl_tree_search_v2(struct file *file, 2153cc68a8a5SGerhard Heift void __user *argp) 2154cc68a8a5SGerhard Heift { 2155cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 __user *uarg; 2156cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 args; 2157cc68a8a5SGerhard Heift struct inode *inode; 2158cc68a8a5SGerhard Heift int ret; 2159cc68a8a5SGerhard Heift size_t buf_size; 2160ee22184bSByongho Lee const size_t buf_limit = SZ_16M; 2161cc68a8a5SGerhard Heift 2162cc68a8a5SGerhard Heift if (!capable(CAP_SYS_ADMIN)) 2163cc68a8a5SGerhard Heift return -EPERM; 2164cc68a8a5SGerhard Heift 2165cc68a8a5SGerhard Heift /* copy search header and buffer size */ 2166cc68a8a5SGerhard Heift uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp; 2167cc68a8a5SGerhard Heift if (copy_from_user(&args, uarg, sizeof(args))) 2168cc68a8a5SGerhard Heift return -EFAULT; 2169cc68a8a5SGerhard Heift 2170cc68a8a5SGerhard Heift buf_size = args.buf_size; 2171cc68a8a5SGerhard Heift 2172cc68a8a5SGerhard Heift if (buf_size < sizeof(struct btrfs_ioctl_search_header)) 2173cc68a8a5SGerhard Heift return -EOVERFLOW; 2174cc68a8a5SGerhard Heift 2175cc68a8a5SGerhard Heift /* limit result size to 16MB */ 2176cc68a8a5SGerhard Heift if (buf_size > buf_limit) 2177cc68a8a5SGerhard Heift buf_size = buf_limit; 2178cc68a8a5SGerhard Heift 2179cc68a8a5SGerhard Heift inode = file_inode(file); 2180cc68a8a5SGerhard Heift ret = search_ioctl(inode, &args.key, &buf_size, 2181cc68a8a5SGerhard Heift (char *)(&uarg->buf[0])); 2182cc68a8a5SGerhard Heift if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key))) 2183cc68a8a5SGerhard Heift ret = -EFAULT; 2184cc68a8a5SGerhard Heift else if (ret == -EOVERFLOW && 2185cc68a8a5SGerhard Heift copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size))) 2186cc68a8a5SGerhard Heift ret = -EFAULT; 2187cc68a8a5SGerhard Heift 218876dda93cSYan, Zheng return ret; 218976dda93cSYan, Zheng } 219076dda93cSYan, Zheng 219198d377a0STARUISI Hiroaki /* 2192ac8e9819SChris Mason * Search INODE_REFs to identify path name of 'dirid' directory 2193ac8e9819SChris Mason * in a 'tree_id' tree. and sets path name to 'name'. 219498d377a0STARUISI Hiroaki */ 219598d377a0STARUISI Hiroaki static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, 219698d377a0STARUISI Hiroaki u64 tree_id, u64 dirid, char *name) 219798d377a0STARUISI Hiroaki { 219898d377a0STARUISI Hiroaki struct btrfs_root *root; 219998d377a0STARUISI Hiroaki struct btrfs_key key; 2200ac8e9819SChris Mason char *ptr; 220198d377a0STARUISI Hiroaki int ret = -1; 220298d377a0STARUISI Hiroaki int slot; 220398d377a0STARUISI Hiroaki int len; 220498d377a0STARUISI Hiroaki int total_len = 0; 220598d377a0STARUISI Hiroaki struct btrfs_inode_ref *iref; 220698d377a0STARUISI Hiroaki struct extent_buffer *l; 220798d377a0STARUISI Hiroaki struct btrfs_path *path; 220898d377a0STARUISI Hiroaki 220998d377a0STARUISI Hiroaki if (dirid == BTRFS_FIRST_FREE_OBJECTID) { 221098d377a0STARUISI Hiroaki name[0]='\0'; 221198d377a0STARUISI Hiroaki return 0; 221298d377a0STARUISI Hiroaki } 221398d377a0STARUISI Hiroaki 221498d377a0STARUISI Hiroaki path = btrfs_alloc_path(); 221598d377a0STARUISI Hiroaki if (!path) 221698d377a0STARUISI Hiroaki return -ENOMEM; 221798d377a0STARUISI Hiroaki 2218ac8e9819SChris Mason ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX]; 221998d377a0STARUISI Hiroaki 222098d377a0STARUISI Hiroaki key.objectid = tree_id; 222198d377a0STARUISI Hiroaki key.type = BTRFS_ROOT_ITEM_KEY; 222298d377a0STARUISI Hiroaki key.offset = (u64)-1; 222398d377a0STARUISI Hiroaki root = btrfs_read_fs_root_no_name(info, &key); 222498d377a0STARUISI Hiroaki if (IS_ERR(root)) { 2225f14d104dSDavid Sterba btrfs_err(info, "could not find root %llu", tree_id); 22268ad6fcabSChris Mason ret = -ENOENT; 22278ad6fcabSChris Mason goto out; 222898d377a0STARUISI Hiroaki } 222998d377a0STARUISI Hiroaki 223098d377a0STARUISI Hiroaki key.objectid = dirid; 223198d377a0STARUISI Hiroaki key.type = BTRFS_INODE_REF_KEY; 22328ad6fcabSChris Mason key.offset = (u64)-1; 223398d377a0STARUISI Hiroaki 223498d377a0STARUISI Hiroaki while (1) { 223598d377a0STARUISI Hiroaki ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 223698d377a0STARUISI Hiroaki if (ret < 0) 223798d377a0STARUISI Hiroaki goto out; 223818674c6cSFilipe David Borba Manana else if (ret > 0) { 223918674c6cSFilipe David Borba Manana ret = btrfs_previous_item(root, path, dirid, 224018674c6cSFilipe David Borba Manana BTRFS_INODE_REF_KEY); 224118674c6cSFilipe David Borba Manana if (ret < 0) 224218674c6cSFilipe David Borba Manana goto out; 224318674c6cSFilipe David Borba Manana else if (ret > 0) { 2244ac8e9819SChris Mason ret = -ENOENT; 224598d377a0STARUISI Hiroaki goto out; 2246ac8e9819SChris Mason } 224718674c6cSFilipe David Borba Manana } 224818674c6cSFilipe David Borba Manana 224918674c6cSFilipe David Borba Manana l = path->nodes[0]; 225018674c6cSFilipe David Borba Manana slot = path->slots[0]; 225118674c6cSFilipe David Borba Manana btrfs_item_key_to_cpu(l, &key, slot); 225298d377a0STARUISI Hiroaki 225398d377a0STARUISI Hiroaki iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); 225498d377a0STARUISI Hiroaki len = btrfs_inode_ref_name_len(l, iref); 225598d377a0STARUISI Hiroaki ptr -= len + 1; 225698d377a0STARUISI Hiroaki total_len += len + 1; 2257a696cf35SFilipe David Borba Manana if (ptr < name) { 2258a696cf35SFilipe David Borba Manana ret = -ENAMETOOLONG; 225998d377a0STARUISI Hiroaki goto out; 2260a696cf35SFilipe David Borba Manana } 226198d377a0STARUISI Hiroaki 226298d377a0STARUISI Hiroaki *(ptr + len) = '/'; 226398d377a0STARUISI Hiroaki read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len); 226498d377a0STARUISI Hiroaki 226598d377a0STARUISI Hiroaki if (key.offset == BTRFS_FIRST_FREE_OBJECTID) 226698d377a0STARUISI Hiroaki break; 226798d377a0STARUISI Hiroaki 2268b3b4aa74SDavid Sterba btrfs_release_path(path); 226998d377a0STARUISI Hiroaki key.objectid = key.offset; 22708ad6fcabSChris Mason key.offset = (u64)-1; 227198d377a0STARUISI Hiroaki dirid = key.objectid; 227298d377a0STARUISI Hiroaki } 227377906a50SLi Zefan memmove(name, ptr, total_len); 227498d377a0STARUISI Hiroaki name[total_len] = '\0'; 227598d377a0STARUISI Hiroaki ret = 0; 227698d377a0STARUISI Hiroaki out: 227798d377a0STARUISI Hiroaki btrfs_free_path(path); 2278ac8e9819SChris Mason return ret; 2279ac8e9819SChris Mason } 2280ac8e9819SChris Mason 2281ac8e9819SChris Mason static noinline int btrfs_ioctl_ino_lookup(struct file *file, 2282ac8e9819SChris Mason void __user *argp) 2283ac8e9819SChris Mason { 2284ac8e9819SChris Mason struct btrfs_ioctl_ino_lookup_args *args; 2285ac8e9819SChris Mason struct inode *inode; 228601b810b8SDavid Sterba int ret = 0; 2287ac8e9819SChris Mason 22882354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 22892354d08fSJulia Lawall if (IS_ERR(args)) 22902354d08fSJulia Lawall return PTR_ERR(args); 2291c2b96929SDan Carpenter 2292496ad9aaSAl Viro inode = file_inode(file); 2293ac8e9819SChris Mason 229401b810b8SDavid Sterba /* 229501b810b8SDavid Sterba * Unprivileged query to obtain the containing subvolume root id. The 229601b810b8SDavid Sterba * path is reset so it's consistent with btrfs_search_path_in_tree. 229701b810b8SDavid Sterba */ 22981b53ac4dSChris Mason if (args->treeid == 0) 22991b53ac4dSChris Mason args->treeid = BTRFS_I(inode)->root->root_key.objectid; 23001b53ac4dSChris Mason 230101b810b8SDavid Sterba if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) { 230201b810b8SDavid Sterba args->name[0] = 0; 230301b810b8SDavid Sterba goto out; 230401b810b8SDavid Sterba } 230501b810b8SDavid Sterba 230601b810b8SDavid Sterba if (!capable(CAP_SYS_ADMIN)) { 230701b810b8SDavid Sterba ret = -EPERM; 230801b810b8SDavid Sterba goto out; 230901b810b8SDavid Sterba } 231001b810b8SDavid Sterba 2311ac8e9819SChris Mason ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, 2312ac8e9819SChris Mason args->treeid, args->objectid, 2313ac8e9819SChris Mason args->name); 2314ac8e9819SChris Mason 231501b810b8SDavid Sterba out: 2316ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 2317ac8e9819SChris Mason ret = -EFAULT; 2318ac8e9819SChris Mason 2319ac8e9819SChris Mason kfree(args); 232098d377a0STARUISI Hiroaki return ret; 232198d377a0STARUISI Hiroaki } 232298d377a0STARUISI Hiroaki 232376dda93cSYan, Zheng static noinline int btrfs_ioctl_snap_destroy(struct file *file, 232476dda93cSYan, Zheng void __user *arg) 232576dda93cSYan, Zheng { 232654563d41SAl Viro struct dentry *parent = file->f_path.dentry; 23270b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb); 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 2342325c50e3SJeff Mahoney if (!S_ISDIR(dir->i_mode)) 2343325c50e3SJeff Mahoney return -ENOTDIR; 2344325c50e3SJeff Mahoney 234576dda93cSYan, Zheng vol_args = memdup_user(arg, sizeof(*vol_args)); 234676dda93cSYan, Zheng if (IS_ERR(vol_args)) 234776dda93cSYan, Zheng return PTR_ERR(vol_args); 234876dda93cSYan, Zheng 234976dda93cSYan, Zheng vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 235076dda93cSYan, Zheng namelen = strlen(vol_args->name); 235176dda93cSYan, Zheng if (strchr(vol_args->name, '/') || 235276dda93cSYan, Zheng strncmp(vol_args->name, "..", namelen) == 0) { 235376dda93cSYan, Zheng err = -EINVAL; 235476dda93cSYan, Zheng goto out; 235576dda93cSYan, Zheng } 235676dda93cSYan, Zheng 2357a561be71SAl Viro err = mnt_want_write_file(file); 235876dda93cSYan, Zheng if (err) 235976dda93cSYan, Zheng goto out; 236076dda93cSYan, Zheng 2361521e0546SDavid Sterba 236200235411SAl Viro err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 236300235411SAl Viro if (err == -EINTR) 236400235411SAl Viro goto out_drop_write; 236576dda93cSYan, Zheng dentry = lookup_one_len(vol_args->name, parent, namelen); 236676dda93cSYan, Zheng if (IS_ERR(dentry)) { 236776dda93cSYan, Zheng err = PTR_ERR(dentry); 236876dda93cSYan, Zheng goto out_unlock_dir; 236976dda93cSYan, Zheng } 237076dda93cSYan, Zheng 23712b0143b5SDavid Howells if (d_really_is_negative(dentry)) { 237276dda93cSYan, Zheng err = -ENOENT; 237376dda93cSYan, Zheng goto out_dput; 237476dda93cSYan, Zheng } 237576dda93cSYan, Zheng 23762b0143b5SDavid Howells inode = d_inode(dentry); 23774260f7c7SSage Weil dest = BTRFS_I(inode)->root; 23784260f7c7SSage Weil if (!capable(CAP_SYS_ADMIN)) { 23794260f7c7SSage Weil /* 23804260f7c7SSage Weil * Regular user. Only allow this with a special mount 23814260f7c7SSage Weil * option, when the user has write+exec access to the 23824260f7c7SSage Weil * subvol root, and when rmdir(2) would have been 23834260f7c7SSage Weil * allowed. 23844260f7c7SSage Weil * 23854260f7c7SSage Weil * Note that this is _not_ check that the subvol is 23864260f7c7SSage Weil * empty or doesn't contain data that we wouldn't 23874260f7c7SSage Weil * otherwise be able to delete. 23884260f7c7SSage Weil * 23894260f7c7SSage Weil * Users who want to delete empty subvols should try 23904260f7c7SSage Weil * rmdir(2). 23914260f7c7SSage Weil */ 23924260f7c7SSage Weil err = -EPERM; 23930b246afaSJeff Mahoney if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED)) 23944260f7c7SSage Weil goto out_dput; 23954260f7c7SSage Weil 23964260f7c7SSage Weil /* 23974260f7c7SSage Weil * Do not allow deletion if the parent dir is the same 23984260f7c7SSage Weil * as the dir to be deleted. That means the ioctl 23994260f7c7SSage Weil * must be called on the dentry referencing the root 24004260f7c7SSage Weil * of the subvol, not a random directory contained 24014260f7c7SSage Weil * within it. 24024260f7c7SSage Weil */ 24034260f7c7SSage Weil err = -EINVAL; 24044260f7c7SSage Weil if (root == dest) 24054260f7c7SSage Weil goto out_dput; 24064260f7c7SSage Weil 24074260f7c7SSage Weil err = inode_permission(inode, MAY_WRITE | MAY_EXEC); 24084260f7c7SSage Weil if (err) 24094260f7c7SSage Weil goto out_dput; 24105c39da5bSMiao Xie } 24114260f7c7SSage Weil 24125c39da5bSMiao Xie /* check if subvolume may be deleted by a user */ 24134260f7c7SSage Weil err = btrfs_may_delete(dir, dentry, 1); 24144260f7c7SSage Weil if (err) 24154260f7c7SSage Weil goto out_dput; 24164260f7c7SSage Weil 24174a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 241876dda93cSYan, Zheng err = -EINVAL; 241976dda93cSYan, Zheng goto out_dput; 242076dda93cSYan, Zheng } 242176dda93cSYan, Zheng 24225955102cSAl Viro inode_lock(inode); 2423521e0546SDavid Sterba 2424521e0546SDavid Sterba /* 2425521e0546SDavid Sterba * Don't allow to delete a subvolume with send in progress. This is 2426521e0546SDavid Sterba * inside the i_mutex so the error handling that has to drop the bit 2427521e0546SDavid Sterba * again is not run concurrently. 2428521e0546SDavid Sterba */ 2429521e0546SDavid Sterba spin_lock(&dest->root_item_lock); 2430c55bfa67SFilipe Manana root_flags = btrfs_root_flags(&dest->root_item); 2431c55bfa67SFilipe Manana if (dest->send_in_progress == 0) { 2432c55bfa67SFilipe Manana btrfs_set_root_flags(&dest->root_item, 2433521e0546SDavid Sterba root_flags | BTRFS_ROOT_SUBVOL_DEAD); 2434521e0546SDavid Sterba spin_unlock(&dest->root_item_lock); 2435521e0546SDavid Sterba } else { 2436521e0546SDavid Sterba spin_unlock(&dest->root_item_lock); 24370b246afaSJeff Mahoney btrfs_warn(fs_info, 2438521e0546SDavid Sterba "Attempt to delete subvolume %llu during send", 2439c55bfa67SFilipe Manana dest->root_key.objectid); 2440521e0546SDavid Sterba err = -EPERM; 2441909e26dcSOmar Sandoval goto out_unlock_inode; 2442521e0546SDavid Sterba } 2443521e0546SDavid Sterba 24440b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 244576dda93cSYan, Zheng 244676dda93cSYan, Zheng err = may_destroy_subvol(dest); 244776dda93cSYan, Zheng if (err) 244876dda93cSYan, Zheng goto out_up_write; 244976dda93cSYan, Zheng 2450c58aaad2SMiao Xie btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); 2451c58aaad2SMiao Xie /* 2452c58aaad2SMiao Xie * One for dir inode, two for dir entries, two for root 2453c58aaad2SMiao Xie * ref/backref. 2454c58aaad2SMiao Xie */ 2455c58aaad2SMiao Xie err = btrfs_subvolume_reserve_metadata(root, &block_rsv, 2456ee3441b4SJeff Mahoney 5, &qgroup_reserved, true); 2457c58aaad2SMiao Xie if (err) 2458c58aaad2SMiao Xie goto out_up_write; 2459c58aaad2SMiao Xie 2460a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 2461a22285a6SYan, Zheng if (IS_ERR(trans)) { 2462a22285a6SYan, Zheng err = PTR_ERR(trans); 2463c58aaad2SMiao Xie goto out_release; 2464a22285a6SYan, Zheng } 2465c58aaad2SMiao Xie trans->block_rsv = &block_rsv; 2466c58aaad2SMiao Xie trans->bytes_reserved = block_rsv.size; 2467a22285a6SYan, Zheng 246843663557SNikolay Borisov btrfs_record_snapshot_destroy(trans, BTRFS_I(dir)); 24692be63d5cSFilipe Manana 247076dda93cSYan, Zheng ret = btrfs_unlink_subvol(trans, root, dir, 247176dda93cSYan, Zheng dest->root_key.objectid, 247276dda93cSYan, Zheng dentry->d_name.name, 247376dda93cSYan, Zheng dentry->d_name.len); 247479787eaaSJeff Mahoney if (ret) { 247579787eaaSJeff Mahoney err = ret; 247666642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 247779787eaaSJeff Mahoney goto out_end_trans; 247879787eaaSJeff Mahoney } 247976dda93cSYan, Zheng 248076dda93cSYan, Zheng btrfs_record_root_in_trans(trans, dest); 248176dda93cSYan, Zheng 248276dda93cSYan, Zheng memset(&dest->root_item.drop_progress, 0, 248376dda93cSYan, Zheng sizeof(dest->root_item.drop_progress)); 248476dda93cSYan, Zheng dest->root_item.drop_level = 0; 248576dda93cSYan, Zheng btrfs_set_root_refs(&dest->root_item, 0); 248676dda93cSYan, Zheng 248727cdeb70SMiao Xie if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) { 248876dda93cSYan, Zheng ret = btrfs_insert_orphan_item(trans, 24890b246afaSJeff Mahoney fs_info->tree_root, 249076dda93cSYan, Zheng dest->root_key.objectid); 249179787eaaSJeff Mahoney if (ret) { 249266642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 249379787eaaSJeff Mahoney err = ret; 249479787eaaSJeff Mahoney goto out_end_trans; 2495d68fc57bSYan, Zheng } 249679787eaaSJeff Mahoney } 2497dd5f9615SStefan Behrens 24980b246afaSJeff Mahoney ret = btrfs_uuid_tree_rem(trans, fs_info, dest->root_item.uuid, 24996bccf3abSJeff Mahoney BTRFS_UUID_KEY_SUBVOL, 2500dd5f9615SStefan Behrens dest->root_key.objectid); 2501dd5f9615SStefan Behrens if (ret && ret != -ENOENT) { 250266642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 2503dd5f9615SStefan Behrens err = ret; 2504dd5f9615SStefan Behrens goto out_end_trans; 2505dd5f9615SStefan Behrens } 2506dd5f9615SStefan Behrens if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) { 25070b246afaSJeff Mahoney ret = btrfs_uuid_tree_rem(trans, fs_info, 2508dd5f9615SStefan Behrens dest->root_item.received_uuid, 2509dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 2510dd5f9615SStefan Behrens dest->root_key.objectid); 2511dd5f9615SStefan Behrens if (ret && ret != -ENOENT) { 251266642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 2513dd5f9615SStefan Behrens err = ret; 2514dd5f9615SStefan Behrens goto out_end_trans; 2515dd5f9615SStefan Behrens } 2516dd5f9615SStefan Behrens } 2517dd5f9615SStefan Behrens 251879787eaaSJeff Mahoney out_end_trans: 2519c58aaad2SMiao Xie trans->block_rsv = NULL; 2520c58aaad2SMiao Xie trans->bytes_reserved = 0; 25213a45bb20SJeff Mahoney ret = btrfs_end_transaction(trans); 252279787eaaSJeff Mahoney if (ret && !err) 252379787eaaSJeff Mahoney err = ret; 252476dda93cSYan, Zheng inode->i_flags |= S_DEAD; 2525c58aaad2SMiao Xie out_release: 25267775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &block_rsv); 252776dda93cSYan, Zheng out_up_write: 25280b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 2529521e0546SDavid Sterba if (err) { 2530521e0546SDavid Sterba spin_lock(&dest->root_item_lock); 2531c55bfa67SFilipe Manana root_flags = btrfs_root_flags(&dest->root_item); 2532c55bfa67SFilipe Manana btrfs_set_root_flags(&dest->root_item, 2533521e0546SDavid Sterba root_flags & ~BTRFS_ROOT_SUBVOL_DEAD); 2534521e0546SDavid Sterba spin_unlock(&dest->root_item_lock); 2535521e0546SDavid Sterba } 2536909e26dcSOmar Sandoval out_unlock_inode: 25375955102cSAl Viro inode_unlock(inode); 253876dda93cSYan, Zheng if (!err) { 253964ad6c48SOmar Sandoval d_invalidate(dentry); 254076dda93cSYan, Zheng btrfs_invalidate_inodes(dest); 254176dda93cSYan, Zheng d_delete(dentry); 254261155aa0SDavid Sterba ASSERT(dest->send_in_progress == 0); 2543fa6ac876SLiu Bo 2544fa6ac876SLiu Bo /* the last ref */ 254557cdc8dbSDavid Sterba if (dest->ino_cache_inode) { 254657cdc8dbSDavid Sterba iput(dest->ino_cache_inode); 254757cdc8dbSDavid Sterba dest->ino_cache_inode = NULL; 2548fa6ac876SLiu Bo } 254976dda93cSYan, Zheng } 255076dda93cSYan, Zheng out_dput: 255176dda93cSYan, Zheng dput(dentry); 255276dda93cSYan, Zheng out_unlock_dir: 25535955102cSAl Viro inode_unlock(dir); 255400235411SAl Viro out_drop_write: 25552a79f17eSAl Viro mnt_drop_write_file(file); 255676dda93cSYan, Zheng out: 255776dda93cSYan, Zheng kfree(vol_args); 255876dda93cSYan, Zheng return err; 255976dda93cSYan, Zheng } 256076dda93cSYan, Zheng 25611e701a32SChris Mason static int btrfs_ioctl_defrag(struct file *file, void __user *argp) 2562f46b5a66SChristoph Hellwig { 2563496ad9aaSAl Viro struct inode *inode = file_inode(file); 2564f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 25651e701a32SChris Mason struct btrfs_ioctl_defrag_range_args *range; 2566c146afadSYan Zheng int ret; 2567c146afadSYan Zheng 256825122d15SIlya Dryomov ret = mnt_want_write_file(file); 256925122d15SIlya Dryomov if (ret) 257025122d15SIlya Dryomov return ret; 2571b83cc969SLi Zefan 257225122d15SIlya Dryomov if (btrfs_root_readonly(root)) { 257325122d15SIlya Dryomov ret = -EROFS; 257425122d15SIlya Dryomov goto out; 25755ac00addSStefan Behrens } 2576f46b5a66SChristoph Hellwig 2577f46b5a66SChristoph Hellwig switch (inode->i_mode & S_IFMT) { 2578f46b5a66SChristoph Hellwig case S_IFDIR: 2579e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) { 2580e441d54dSChris Mason ret = -EPERM; 2581e441d54dSChris Mason goto out; 2582e441d54dSChris Mason } 2583de78b51aSEric Sandeen ret = btrfs_defrag_root(root); 2584f46b5a66SChristoph Hellwig break; 2585f46b5a66SChristoph Hellwig case S_IFREG: 2586e441d54dSChris Mason if (!(file->f_mode & FMODE_WRITE)) { 2587e441d54dSChris Mason ret = -EINVAL; 2588e441d54dSChris Mason goto out; 2589e441d54dSChris Mason } 25901e701a32SChris Mason 25911e701a32SChris Mason range = kzalloc(sizeof(*range), GFP_KERNEL); 25921e701a32SChris Mason if (!range) { 25931e701a32SChris Mason ret = -ENOMEM; 25941e701a32SChris Mason goto out; 25951e701a32SChris Mason } 25961e701a32SChris Mason 25971e701a32SChris Mason if (argp) { 25981e701a32SChris Mason if (copy_from_user(range, argp, 25991e701a32SChris Mason sizeof(*range))) { 26001e701a32SChris Mason ret = -EFAULT; 26011e701a32SChris Mason kfree(range); 2602683be16eSDan Carpenter goto out; 26031e701a32SChris Mason } 26041e701a32SChris Mason /* compression requires us to start the IO */ 26051e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 26061e701a32SChris Mason range->flags |= BTRFS_DEFRAG_RANGE_START_IO; 26071e701a32SChris Mason range->extent_thresh = (u32)-1; 26081e701a32SChris Mason } 26091e701a32SChris Mason } else { 26101e701a32SChris Mason /* the rest are all set to zero by kzalloc */ 26111e701a32SChris Mason range->len = (u64)-1; 26121e701a32SChris Mason } 2613496ad9aaSAl Viro ret = btrfs_defrag_file(file_inode(file), file, 26144cb5300bSChris Mason range, 0, 0); 26154cb5300bSChris Mason if (ret > 0) 26164cb5300bSChris Mason ret = 0; 26171e701a32SChris Mason kfree(range); 2618f46b5a66SChristoph Hellwig break; 26198929ecfaSYan, Zheng default: 26208929ecfaSYan, Zheng ret = -EINVAL; 2621f46b5a66SChristoph Hellwig } 2622e441d54dSChris Mason out: 262325122d15SIlya Dryomov mnt_drop_write_file(file); 2624e441d54dSChris Mason return ret; 2625f46b5a66SChristoph Hellwig } 2626f46b5a66SChristoph Hellwig 26272ff7e61eSJeff Mahoney static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg) 2628f46b5a66SChristoph Hellwig { 2629f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2630f46b5a66SChristoph Hellwig int ret; 2631f46b5a66SChristoph Hellwig 2632e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2633e441d54dSChris Mason return -EPERM; 2634e441d54dSChris Mason 2635171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) 2636e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 2637c9e9f97bSIlya Dryomov 26380b246afaSJeff Mahoney mutex_lock(&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'; 26462ff7e61eSJeff Mahoney ret = btrfs_init_new_device(fs_info, vol_args->name); 2647f46b5a66SChristoph Hellwig 264843d20761SAnand Jain if (!ret) 26490b246afaSJeff Mahoney btrfs_info(fs_info, "disk added %s", vol_args->name); 265043d20761SAnand Jain 2651f46b5a66SChristoph Hellwig kfree(vol_args); 2652c9e9f97bSIlya Dryomov out: 26530b246afaSJeff Mahoney mutex_unlock(&fs_info->volume_mutex); 2654171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 2655f46b5a66SChristoph Hellwig return ret; 2656f46b5a66SChristoph Hellwig } 2657f46b5a66SChristoph Hellwig 26586b526ed7SAnand Jain static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg) 2659f46b5a66SChristoph Hellwig { 26600b246afaSJeff Mahoney struct inode *inode = file_inode(file); 26610b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 26626b526ed7SAnand Jain struct btrfs_ioctl_vol_args_v2 *vol_args; 2663f46b5a66SChristoph Hellwig int ret; 2664f46b5a66SChristoph Hellwig 2665e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2666e441d54dSChris Mason return -EPERM; 2667e441d54dSChris Mason 2668da24927bSMiao Xie ret = mnt_want_write_file(file); 2669da24927bSMiao Xie if (ret) 2670da24927bSMiao Xie return ret; 2671c146afadSYan Zheng 2672dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2673c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 2674c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 2675c47ca32dSDan Carpenter goto err_drop; 2676c9e9f97bSIlya Dryomov } 2677f46b5a66SChristoph Hellwig 26786b526ed7SAnand Jain /* Check for compatibility reject unknown flags */ 2679735654eaSDavid Sterba if (vol_args->flags & ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED) 2680735654eaSDavid Sterba return -EOPNOTSUPP; 2681f46b5a66SChristoph Hellwig 2682171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 2683183860f6SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 2684183860f6SAnand Jain goto out; 2685183860f6SAnand Jain } 2686183860f6SAnand Jain 26870b246afaSJeff Mahoney mutex_lock(&fs_info->volume_mutex); 2688735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) { 26892ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, NULL, vol_args->devid); 26906b526ed7SAnand Jain } else { 26916b526ed7SAnand Jain vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 26922ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, vol_args->name, 0); 26936b526ed7SAnand Jain } 26940b246afaSJeff Mahoney mutex_unlock(&fs_info->volume_mutex); 2695171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 2696183860f6SAnand Jain 26976b526ed7SAnand Jain if (!ret) { 2698735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) 26990b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: id %llu", 27006b526ed7SAnand Jain vol_args->devid); 27016b526ed7SAnand Jain else 27020b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: %s", 27036b526ed7SAnand Jain vol_args->name); 27046b526ed7SAnand Jain } 2705183860f6SAnand Jain out: 2706183860f6SAnand Jain kfree(vol_args); 2707c47ca32dSDan Carpenter err_drop: 27084ac20c70SIlya Dryomov mnt_drop_write_file(file); 2709f46b5a66SChristoph Hellwig return ret; 2710f46b5a66SChristoph Hellwig } 2711f46b5a66SChristoph Hellwig 2712f46b5a66SChristoph Hellwig static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg) 2713f46b5a66SChristoph Hellwig { 27140b246afaSJeff Mahoney struct inode *inode = file_inode(file); 27150b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 2716f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2717f46b5a66SChristoph Hellwig int ret; 2718f46b5a66SChristoph Hellwig 2719f46b5a66SChristoph Hellwig if (!capable(CAP_SYS_ADMIN)) 2720f46b5a66SChristoph Hellwig return -EPERM; 2721f46b5a66SChristoph Hellwig 2722f46b5a66SChristoph Hellwig ret = mnt_want_write_file(file); 2723f46b5a66SChristoph Hellwig if (ret) 2724f46b5a66SChristoph Hellwig return ret; 2725f46b5a66SChristoph Hellwig 2726171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 2727f46b5a66SChristoph Hellwig ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 272858d7bbf8SDavid Sterba goto out_drop_write; 272958d7bbf8SDavid Sterba } 273058d7bbf8SDavid Sterba 273158d7bbf8SDavid Sterba vol_args = memdup_user(arg, sizeof(*vol_args)); 273258d7bbf8SDavid Sterba if (IS_ERR(vol_args)) { 273358d7bbf8SDavid Sterba ret = PTR_ERR(vol_args); 2734f46b5a66SChristoph Hellwig goto out; 2735f46b5a66SChristoph Hellwig } 2736f46b5a66SChristoph Hellwig 273758d7bbf8SDavid Sterba vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 27380b246afaSJeff Mahoney mutex_lock(&fs_info->volume_mutex); 27392ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, vol_args->name, 0); 27400b246afaSJeff Mahoney mutex_unlock(&fs_info->volume_mutex); 2741f46b5a66SChristoph Hellwig 2742f46b5a66SChristoph Hellwig if (!ret) 27430b246afaSJeff Mahoney btrfs_info(fs_info, "disk deleted %s", vol_args->name); 2744f46b5a66SChristoph Hellwig kfree(vol_args); 274558d7bbf8SDavid Sterba out: 2746171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 274758d7bbf8SDavid Sterba out_drop_write: 2748f46b5a66SChristoph Hellwig mnt_drop_write_file(file); 274958d7bbf8SDavid Sterba 2750f46b5a66SChristoph Hellwig return ret; 2751f46b5a66SChristoph Hellwig } 2752f46b5a66SChristoph Hellwig 27532ff7e61eSJeff Mahoney static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info, 27542ff7e61eSJeff Mahoney void __user *arg) 2755475f6387SJan Schmidt { 2756027ed2f0SLi Zefan struct btrfs_ioctl_fs_info_args *fi_args; 2757475f6387SJan Schmidt struct btrfs_device *device; 27580b246afaSJeff Mahoney struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 2759027ed2f0SLi Zefan int ret = 0; 2760475f6387SJan Schmidt 2761027ed2f0SLi Zefan fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL); 2762027ed2f0SLi Zefan if (!fi_args) 2763027ed2f0SLi Zefan return -ENOMEM; 2764027ed2f0SLi Zefan 2765f7171750SFilipe David Borba Manana mutex_lock(&fs_devices->device_list_mutex); 2766027ed2f0SLi Zefan fi_args->num_devices = fs_devices->num_devices; 27670b246afaSJeff Mahoney memcpy(&fi_args->fsid, fs_info->fsid, sizeof(fi_args->fsid)); 2768475f6387SJan Schmidt 2769d7641a49SByongho Lee list_for_each_entry(device, &fs_devices->devices, dev_list) { 2770027ed2f0SLi Zefan if (device->devid > fi_args->max_id) 2771027ed2f0SLi Zefan fi_args->max_id = device->devid; 2772475f6387SJan Schmidt } 2773475f6387SJan Schmidt mutex_unlock(&fs_devices->device_list_mutex); 2774475f6387SJan Schmidt 27750b246afaSJeff Mahoney fi_args->nodesize = fs_info->super_copy->nodesize; 27760b246afaSJeff Mahoney fi_args->sectorsize = fs_info->super_copy->sectorsize; 27770b246afaSJeff Mahoney fi_args->clone_alignment = fs_info->super_copy->sectorsize; 277880a773fbSDavid Sterba 2779027ed2f0SLi Zefan if (copy_to_user(arg, fi_args, sizeof(*fi_args))) 2780027ed2f0SLi Zefan ret = -EFAULT; 2781475f6387SJan Schmidt 2782027ed2f0SLi Zefan kfree(fi_args); 2783027ed2f0SLi Zefan return ret; 2784475f6387SJan Schmidt } 2785475f6387SJan Schmidt 27862ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info, 27872ff7e61eSJeff Mahoney void __user *arg) 2788475f6387SJan Schmidt { 2789475f6387SJan Schmidt struct btrfs_ioctl_dev_info_args *di_args; 2790475f6387SJan Schmidt struct btrfs_device *dev; 27910b246afaSJeff Mahoney struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 2792475f6387SJan Schmidt int ret = 0; 2793475f6387SJan Schmidt char *s_uuid = NULL; 2794475f6387SJan Schmidt 2795475f6387SJan Schmidt di_args = memdup_user(arg, sizeof(*di_args)); 2796475f6387SJan Schmidt if (IS_ERR(di_args)) 2797475f6387SJan Schmidt return PTR_ERR(di_args); 2798475f6387SJan Schmidt 2799dd5f9615SStefan Behrens if (!btrfs_is_empty_uuid(di_args->uuid)) 2800475f6387SJan Schmidt s_uuid = di_args->uuid; 2801475f6387SJan Schmidt 2802475f6387SJan Schmidt mutex_lock(&fs_devices->device_list_mutex); 28030b246afaSJeff Mahoney dev = btrfs_find_device(fs_info, di_args->devid, s_uuid, NULL); 2804475f6387SJan Schmidt 2805475f6387SJan Schmidt if (!dev) { 2806475f6387SJan Schmidt ret = -ENODEV; 2807475f6387SJan Schmidt goto out; 2808475f6387SJan Schmidt } 2809475f6387SJan Schmidt 2810475f6387SJan Schmidt di_args->devid = dev->devid; 28117cc8e58dSMiao Xie di_args->bytes_used = btrfs_device_get_bytes_used(dev); 28127cc8e58dSMiao Xie di_args->total_bytes = btrfs_device_get_total_bytes(dev); 2813475f6387SJan Schmidt memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); 2814a27202fbSJim Meyering if (dev->name) { 2815606686eeSJosef Bacik struct rcu_string *name; 2816606686eeSJosef Bacik 2817606686eeSJosef Bacik rcu_read_lock(); 2818606686eeSJosef Bacik name = rcu_dereference(dev->name); 2819606686eeSJosef Bacik strncpy(di_args->path, name->str, sizeof(di_args->path)); 2820606686eeSJosef Bacik rcu_read_unlock(); 2821a27202fbSJim Meyering di_args->path[sizeof(di_args->path) - 1] = 0; 2822a27202fbSJim Meyering } else { 282399ba55adSStefan Behrens di_args->path[0] = '\0'; 2824a27202fbSJim Meyering } 2825475f6387SJan Schmidt 2826475f6387SJan Schmidt out: 282755793c0dSDavid Sterba mutex_unlock(&fs_devices->device_list_mutex); 2828475f6387SJan Schmidt if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) 2829475f6387SJan Schmidt ret = -EFAULT; 2830475f6387SJan Schmidt 2831475f6387SJan Schmidt kfree(di_args); 2832475f6387SJan Schmidt return ret; 2833475f6387SJan Schmidt } 2834475f6387SJan Schmidt 2835f4414602SMark Fasheh static struct page *extent_same_get_page(struct inode *inode, pgoff_t index) 2836416161dbSMark Fasheh { 2837416161dbSMark Fasheh struct page *page; 2838416161dbSMark Fasheh 2839416161dbSMark Fasheh page = grab_cache_page(inode->i_mapping, index); 2840416161dbSMark Fasheh if (!page) 284131314002SFilipe Manana return ERR_PTR(-ENOMEM); 2842416161dbSMark Fasheh 2843416161dbSMark Fasheh if (!PageUptodate(page)) { 284431314002SFilipe Manana int ret; 284531314002SFilipe Manana 284631314002SFilipe Manana ret = btrfs_readpage(NULL, page); 284731314002SFilipe Manana if (ret) 284831314002SFilipe Manana return ERR_PTR(ret); 2849416161dbSMark Fasheh lock_page(page); 2850416161dbSMark Fasheh if (!PageUptodate(page)) { 2851416161dbSMark Fasheh unlock_page(page); 285209cbfeafSKirill A. Shutemov put_page(page); 285331314002SFilipe Manana return ERR_PTR(-EIO); 2854416161dbSMark Fasheh } 285531314002SFilipe Manana if (page->mapping != inode->i_mapping) { 2856416161dbSMark Fasheh unlock_page(page); 285709cbfeafSKirill A. Shutemov put_page(page); 285831314002SFilipe Manana return ERR_PTR(-EAGAIN); 2859416161dbSMark Fasheh } 2860416161dbSMark Fasheh } 2861416161dbSMark Fasheh 2862416161dbSMark Fasheh return page; 2863416161dbSMark Fasheh } 2864416161dbSMark Fasheh 2865f4414602SMark Fasheh static int gather_extent_pages(struct inode *inode, struct page **pages, 2866f4414602SMark Fasheh int num_pages, u64 off) 2867f4414602SMark Fasheh { 2868f4414602SMark Fasheh int i; 286909cbfeafSKirill A. Shutemov pgoff_t index = off >> PAGE_SHIFT; 2870f4414602SMark Fasheh 2871f4414602SMark Fasheh for (i = 0; i < num_pages; i++) { 287231314002SFilipe Manana again: 2873f4414602SMark Fasheh pages[i] = extent_same_get_page(inode, index + i); 287431314002SFilipe Manana if (IS_ERR(pages[i])) { 287531314002SFilipe Manana int err = PTR_ERR(pages[i]); 287631314002SFilipe Manana 287731314002SFilipe Manana if (err == -EAGAIN) 287831314002SFilipe Manana goto again; 287931314002SFilipe Manana pages[i] = NULL; 288031314002SFilipe Manana return err; 288131314002SFilipe Manana } 2882f4414602SMark Fasheh } 2883f4414602SMark Fasheh return 0; 2884f4414602SMark Fasheh } 2885f4414602SMark Fasheh 2886e0bd70c6SFilipe Manana static int lock_extent_range(struct inode *inode, u64 off, u64 len, 2887e0bd70c6SFilipe Manana bool retry_range_locking) 288877fe20dcSMark Fasheh { 2889e0bd70c6SFilipe Manana /* 2890e0bd70c6SFilipe Manana * Do any pending delalloc/csum calculations on inode, one way or 2891e0bd70c6SFilipe Manana * another, and lock file content. 2892e0bd70c6SFilipe Manana * The locking order is: 2893e0bd70c6SFilipe Manana * 2894e0bd70c6SFilipe Manana * 1) pages 2895e0bd70c6SFilipe Manana * 2) range in the inode's io tree 2896e0bd70c6SFilipe Manana */ 289777fe20dcSMark Fasheh while (1) { 289877fe20dcSMark Fasheh struct btrfs_ordered_extent *ordered; 289977fe20dcSMark Fasheh lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); 290077fe20dcSMark Fasheh ordered = btrfs_lookup_first_ordered_extent(inode, 290177fe20dcSMark Fasheh off + len - 1); 2902ff5df9b8SFilipe Manana if ((!ordered || 2903ff5df9b8SFilipe Manana ordered->file_offset + ordered->len <= off || 2904ff5df9b8SFilipe Manana ordered->file_offset >= off + len) && 290577fe20dcSMark Fasheh !test_range_bit(&BTRFS_I(inode)->io_tree, off, 2906ff5df9b8SFilipe Manana off + len - 1, EXTENT_DELALLOC, 0, NULL)) { 2907ff5df9b8SFilipe Manana if (ordered) 2908ff5df9b8SFilipe Manana btrfs_put_ordered_extent(ordered); 290977fe20dcSMark Fasheh break; 2910ff5df9b8SFilipe Manana } 291177fe20dcSMark Fasheh unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); 291277fe20dcSMark Fasheh if (ordered) 291377fe20dcSMark Fasheh btrfs_put_ordered_extent(ordered); 2914e0bd70c6SFilipe Manana if (!retry_range_locking) 2915e0bd70c6SFilipe Manana return -EAGAIN; 291677fe20dcSMark Fasheh btrfs_wait_ordered_range(inode, off, len); 291777fe20dcSMark Fasheh } 2918e0bd70c6SFilipe Manana return 0; 291977fe20dcSMark Fasheh } 292077fe20dcSMark Fasheh 2921f4414602SMark Fasheh static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2) 2922416161dbSMark Fasheh { 29235955102cSAl Viro inode_unlock(inode1); 29245955102cSAl Viro inode_unlock(inode2); 2925416161dbSMark Fasheh } 2926416161dbSMark Fasheh 2927f4414602SMark Fasheh static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2) 2928f4414602SMark Fasheh { 2929f4414602SMark Fasheh if (inode1 < inode2) 2930f4414602SMark Fasheh swap(inode1, inode2); 2931f4414602SMark Fasheh 29325955102cSAl Viro inode_lock_nested(inode1, I_MUTEX_PARENT); 29335955102cSAl Viro inode_lock_nested(inode2, I_MUTEX_CHILD); 2934f4414602SMark Fasheh } 2935f4414602SMark Fasheh 2936f4414602SMark Fasheh static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1, 2937f4414602SMark Fasheh struct inode *inode2, u64 loff2, u64 len) 2938f4414602SMark Fasheh { 2939f4414602SMark Fasheh unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); 2940f4414602SMark Fasheh unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); 2941f4414602SMark Fasheh } 2942f4414602SMark Fasheh 2943e0bd70c6SFilipe Manana static int btrfs_double_extent_lock(struct inode *inode1, u64 loff1, 2944e0bd70c6SFilipe Manana struct inode *inode2, u64 loff2, u64 len, 2945e0bd70c6SFilipe Manana bool retry_range_locking) 2946416161dbSMark Fasheh { 2947e0bd70c6SFilipe Manana int ret; 2948e0bd70c6SFilipe Manana 2949416161dbSMark Fasheh if (inode1 < inode2) { 2950416161dbSMark Fasheh swap(inode1, inode2); 2951416161dbSMark Fasheh swap(loff1, loff2); 2952416161dbSMark Fasheh } 2953e0bd70c6SFilipe Manana ret = lock_extent_range(inode1, loff1, len, retry_range_locking); 2954e0bd70c6SFilipe Manana if (ret) 2955e0bd70c6SFilipe Manana return ret; 2956e0bd70c6SFilipe Manana ret = lock_extent_range(inode2, loff2, len, retry_range_locking); 2957e0bd70c6SFilipe Manana if (ret) 2958e0bd70c6SFilipe Manana unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, 2959e0bd70c6SFilipe Manana loff1 + len - 1); 2960e0bd70c6SFilipe Manana return ret; 2961416161dbSMark Fasheh } 2962f4414602SMark Fasheh 2963f4414602SMark Fasheh struct cmp_pages { 2964f4414602SMark Fasheh int num_pages; 2965f4414602SMark Fasheh struct page **src_pages; 2966f4414602SMark Fasheh struct page **dst_pages; 2967f4414602SMark Fasheh }; 2968f4414602SMark Fasheh 2969f4414602SMark Fasheh static void btrfs_cmp_data_free(struct cmp_pages *cmp) 2970f4414602SMark Fasheh { 2971f4414602SMark Fasheh int i; 2972f4414602SMark Fasheh struct page *pg; 2973f4414602SMark Fasheh 2974f4414602SMark Fasheh for (i = 0; i < cmp->num_pages; i++) { 2975f4414602SMark Fasheh pg = cmp->src_pages[i]; 2976e0bd70c6SFilipe Manana if (pg) { 2977e0bd70c6SFilipe Manana unlock_page(pg); 297809cbfeafSKirill A. Shutemov put_page(pg); 2979e0bd70c6SFilipe Manana } 2980f4414602SMark Fasheh pg = cmp->dst_pages[i]; 2981e0bd70c6SFilipe Manana if (pg) { 2982e0bd70c6SFilipe Manana unlock_page(pg); 298309cbfeafSKirill A. Shutemov put_page(pg); 2984f4414602SMark Fasheh } 2985e0bd70c6SFilipe Manana } 2986f4414602SMark Fasheh kfree(cmp->src_pages); 2987f4414602SMark Fasheh kfree(cmp->dst_pages); 2988f4414602SMark Fasheh } 2989f4414602SMark Fasheh 2990f4414602SMark Fasheh static int btrfs_cmp_data_prepare(struct inode *src, u64 loff, 2991f4414602SMark Fasheh struct inode *dst, u64 dst_loff, 2992f4414602SMark Fasheh u64 len, struct cmp_pages *cmp) 2993f4414602SMark Fasheh { 2994f4414602SMark Fasheh int ret; 299509cbfeafSKirill A. Shutemov int num_pages = PAGE_ALIGN(len) >> PAGE_SHIFT; 2996f4414602SMark Fasheh struct page **src_pgarr, **dst_pgarr; 2997f4414602SMark Fasheh 2998f4414602SMark Fasheh /* 2999f4414602SMark Fasheh * We must gather up all the pages before we initiate our 3000f4414602SMark Fasheh * extent locking. We use an array for the page pointers. Size 3001f4414602SMark Fasheh * of the array is bounded by len, which is in turn bounded by 3002f4414602SMark Fasheh * BTRFS_MAX_DEDUPE_LEN. 3003f4414602SMark Fasheh */ 300466722f7cSDavid Sterba src_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL); 300566722f7cSDavid Sterba dst_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL); 3006f4414602SMark Fasheh if (!src_pgarr || !dst_pgarr) { 3007f4414602SMark Fasheh kfree(src_pgarr); 3008f4414602SMark Fasheh kfree(dst_pgarr); 3009f4414602SMark Fasheh return -ENOMEM; 3010f4414602SMark Fasheh } 3011f4414602SMark Fasheh cmp->num_pages = num_pages; 3012f4414602SMark Fasheh cmp->src_pages = src_pgarr; 3013f4414602SMark Fasheh cmp->dst_pages = dst_pgarr; 3014f4414602SMark Fasheh 3015b1517622SFilipe Manana /* 3016b1517622SFilipe Manana * If deduping ranges in the same inode, locking rules make it mandatory 3017b1517622SFilipe Manana * to always lock pages in ascending order to avoid deadlocks with 3018b1517622SFilipe Manana * concurrent tasks (such as starting writeback/delalloc). 3019b1517622SFilipe Manana */ 3020b1517622SFilipe Manana if (src == dst && dst_loff < loff) { 3021b1517622SFilipe Manana swap(src_pgarr, dst_pgarr); 3022b1517622SFilipe Manana swap(loff, dst_loff); 3023b1517622SFilipe Manana } 3024b1517622SFilipe Manana 3025b1517622SFilipe Manana ret = gather_extent_pages(src, src_pgarr, cmp->num_pages, loff); 3026f4414602SMark Fasheh if (ret) 3027f4414602SMark Fasheh goto out; 3028f4414602SMark Fasheh 3029b1517622SFilipe Manana ret = gather_extent_pages(dst, dst_pgarr, cmp->num_pages, dst_loff); 3030f4414602SMark Fasheh 3031f4414602SMark Fasheh out: 3032f4414602SMark Fasheh if (ret) 3033f4414602SMark Fasheh btrfs_cmp_data_free(cmp); 3034f4414602SMark Fasheh return 0; 3035416161dbSMark Fasheh } 3036416161dbSMark Fasheh 30371a287cfeSDavid Sterba static int btrfs_cmp_data(u64 len, struct cmp_pages *cmp) 3038416161dbSMark Fasheh { 3039416161dbSMark Fasheh int ret = 0; 3040f4414602SMark Fasheh int i; 3041416161dbSMark Fasheh struct page *src_page, *dst_page; 304209cbfeafSKirill A. Shutemov unsigned int cmp_len = PAGE_SIZE; 3043416161dbSMark Fasheh void *addr, *dst_addr; 3044416161dbSMark Fasheh 3045f4414602SMark Fasheh i = 0; 3046416161dbSMark Fasheh while (len) { 304709cbfeafSKirill A. Shutemov if (len < PAGE_SIZE) 3048416161dbSMark Fasheh cmp_len = len; 3049416161dbSMark Fasheh 3050f4414602SMark Fasheh BUG_ON(i >= cmp->num_pages); 3051f4414602SMark Fasheh 3052f4414602SMark Fasheh src_page = cmp->src_pages[i]; 3053f4414602SMark Fasheh dst_page = cmp->dst_pages[i]; 3054e0bd70c6SFilipe Manana ASSERT(PageLocked(src_page)); 3055e0bd70c6SFilipe Manana ASSERT(PageLocked(dst_page)); 3056f4414602SMark Fasheh 3057416161dbSMark Fasheh addr = kmap_atomic(src_page); 3058416161dbSMark Fasheh dst_addr = kmap_atomic(dst_page); 3059416161dbSMark Fasheh 3060416161dbSMark Fasheh flush_dcache_page(src_page); 3061416161dbSMark Fasheh flush_dcache_page(dst_page); 3062416161dbSMark Fasheh 3063416161dbSMark Fasheh if (memcmp(addr, dst_addr, cmp_len)) 30642b3909f8SDarrick J. Wong ret = -EBADE; 3065416161dbSMark Fasheh 3066416161dbSMark Fasheh kunmap_atomic(addr); 3067416161dbSMark Fasheh kunmap_atomic(dst_addr); 3068416161dbSMark Fasheh 3069416161dbSMark Fasheh if (ret) 3070416161dbSMark Fasheh break; 3071416161dbSMark Fasheh 3072416161dbSMark Fasheh len -= cmp_len; 3073f4414602SMark Fasheh i++; 3074416161dbSMark Fasheh } 3075416161dbSMark Fasheh 3076416161dbSMark Fasheh return ret; 3077416161dbSMark Fasheh } 3078416161dbSMark Fasheh 3079e1d227a4SMark Fasheh static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen, 3080e1d227a4SMark Fasheh u64 olen) 3081416161dbSMark Fasheh { 3082e1d227a4SMark Fasheh u64 len = *plen; 3083416161dbSMark Fasheh u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize; 3084416161dbSMark Fasheh 3085e1d227a4SMark Fasheh if (off + olen > inode->i_size || off + olen < off) 3086416161dbSMark Fasheh return -EINVAL; 3087e1d227a4SMark Fasheh 3088e1d227a4SMark Fasheh /* if we extend to eof, continue to block boundary */ 3089e1d227a4SMark Fasheh if (off + len == inode->i_size) 3090e1d227a4SMark Fasheh *plen = len = ALIGN(inode->i_size, bs) - off; 3091e1d227a4SMark Fasheh 3092416161dbSMark Fasheh /* Check that we are block aligned - btrfs_clone() requires this */ 3093416161dbSMark Fasheh if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs)) 3094416161dbSMark Fasheh return -EINVAL; 3095416161dbSMark Fasheh 3096416161dbSMark Fasheh return 0; 3097416161dbSMark Fasheh } 3098416161dbSMark Fasheh 3099e1d227a4SMark Fasheh static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen, 3100416161dbSMark Fasheh struct inode *dst, u64 dst_loff) 3101416161dbSMark Fasheh { 3102416161dbSMark Fasheh int ret; 3103e1d227a4SMark Fasheh u64 len = olen; 3104f4414602SMark Fasheh struct cmp_pages cmp; 3105fc4badd9SOmar Sandoval bool same_inode = (src == dst); 31060efa9f48SMark Fasheh u64 same_lock_start = 0; 31070efa9f48SMark Fasheh u64 same_lock_len = 0; 3108416161dbSMark Fasheh 3109113e8283SFilipe Manana if (len == 0) 3110113e8283SFilipe Manana return 0; 3111113e8283SFilipe Manana 3112fc4badd9SOmar Sandoval if (same_inode) 31135955102cSAl Viro inode_lock(src); 3114fc4badd9SOmar Sandoval else 3115fc4badd9SOmar Sandoval btrfs_double_inode_lock(src, dst); 31160efa9f48SMark Fasheh 31170efa9f48SMark Fasheh ret = extent_same_check_offsets(src, loff, &len, olen); 31180efa9f48SMark Fasheh if (ret) 31190efa9f48SMark Fasheh goto out_unlock; 3120fc4badd9SOmar Sandoval 3121fc4badd9SOmar Sandoval ret = extent_same_check_offsets(dst, dst_loff, &len, olen); 3122f4dfe687SFilipe Manana if (ret) 3123f4dfe687SFilipe Manana goto out_unlock; 31240efa9f48SMark Fasheh 3125fc4badd9SOmar Sandoval if (same_inode) { 31260efa9f48SMark Fasheh /* 31270efa9f48SMark Fasheh * Single inode case wants the same checks, except we 31280efa9f48SMark Fasheh * don't want our length pushed out past i_size as 31290efa9f48SMark Fasheh * comparing that data range makes no sense. 31300efa9f48SMark Fasheh * 31310efa9f48SMark Fasheh * extent_same_check_offsets() will do this for an 31320efa9f48SMark Fasheh * unaligned length at i_size, so catch it here and 31330efa9f48SMark Fasheh * reject the request. 31340efa9f48SMark Fasheh * 31350efa9f48SMark Fasheh * This effectively means we require aligned extents 31360efa9f48SMark Fasheh * for the single-inode case, whereas the other cases 31370efa9f48SMark Fasheh * allow an unaligned length so long as it ends at 31380efa9f48SMark Fasheh * i_size. 31390efa9f48SMark Fasheh */ 31400efa9f48SMark Fasheh if (len != olen) { 31410efa9f48SMark Fasheh ret = -EINVAL; 31420efa9f48SMark Fasheh goto out_unlock; 31430efa9f48SMark Fasheh } 31440efa9f48SMark Fasheh 31450efa9f48SMark Fasheh /* Check for overlapping ranges */ 31460efa9f48SMark Fasheh if (dst_loff + len > loff && dst_loff < loff + len) { 31470efa9f48SMark Fasheh ret = -EINVAL; 31480efa9f48SMark Fasheh goto out_unlock; 31490efa9f48SMark Fasheh } 31500efa9f48SMark Fasheh 31510efa9f48SMark Fasheh same_lock_start = min_t(u64, loff, dst_loff); 31520efa9f48SMark Fasheh same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start; 31530efa9f48SMark Fasheh } 3154416161dbSMark Fasheh 3155416161dbSMark Fasheh /* don't make the dst file partly checksummed */ 3156416161dbSMark Fasheh if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != 3157416161dbSMark Fasheh (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) { 3158416161dbSMark Fasheh ret = -EINVAL; 3159416161dbSMark Fasheh goto out_unlock; 3160416161dbSMark Fasheh } 3161416161dbSMark Fasheh 3162e0bd70c6SFilipe Manana again: 3163f4414602SMark Fasheh ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, &cmp); 3164f4414602SMark Fasheh if (ret) 3165f4414602SMark Fasheh goto out_unlock; 3166f4414602SMark Fasheh 31670efa9f48SMark Fasheh if (same_inode) 3168e0bd70c6SFilipe Manana ret = lock_extent_range(src, same_lock_start, same_lock_len, 3169e0bd70c6SFilipe Manana false); 31700efa9f48SMark Fasheh else 3171e0bd70c6SFilipe Manana ret = btrfs_double_extent_lock(src, loff, dst, dst_loff, len, 3172e0bd70c6SFilipe Manana false); 3173e0bd70c6SFilipe Manana /* 3174e0bd70c6SFilipe Manana * If one of the inodes has dirty pages in the respective range or 3175e0bd70c6SFilipe Manana * ordered extents, we need to flush dellaloc and wait for all ordered 3176e0bd70c6SFilipe Manana * extents in the range. We must unlock the pages and the ranges in the 3177e0bd70c6SFilipe Manana * io trees to avoid deadlocks when flushing delalloc (requires locking 3178e0bd70c6SFilipe Manana * pages) and when waiting for ordered extents to complete (they require 3179e0bd70c6SFilipe Manana * range locking). 3180e0bd70c6SFilipe Manana */ 3181e0bd70c6SFilipe Manana if (ret == -EAGAIN) { 3182e0bd70c6SFilipe Manana /* 3183e0bd70c6SFilipe Manana * Ranges in the io trees already unlocked. Now unlock all 3184e0bd70c6SFilipe Manana * pages before waiting for all IO to complete. 3185e0bd70c6SFilipe Manana */ 3186e0bd70c6SFilipe Manana btrfs_cmp_data_free(&cmp); 3187e0bd70c6SFilipe Manana if (same_inode) { 3188e0bd70c6SFilipe Manana btrfs_wait_ordered_range(src, same_lock_start, 3189e0bd70c6SFilipe Manana same_lock_len); 3190e0bd70c6SFilipe Manana } else { 3191e0bd70c6SFilipe Manana btrfs_wait_ordered_range(src, loff, len); 3192e0bd70c6SFilipe Manana btrfs_wait_ordered_range(dst, dst_loff, len); 3193e0bd70c6SFilipe Manana } 3194e0bd70c6SFilipe Manana goto again; 3195e0bd70c6SFilipe Manana } 3196e0bd70c6SFilipe Manana ASSERT(ret == 0); 3197e0bd70c6SFilipe Manana if (WARN_ON(ret)) { 3198e0bd70c6SFilipe Manana /* ranges in the io trees already unlocked */ 3199e0bd70c6SFilipe Manana btrfs_cmp_data_free(&cmp); 3200e0bd70c6SFilipe Manana return ret; 3201e0bd70c6SFilipe Manana } 3202f4414602SMark Fasheh 3203207910ddSMark Fasheh /* pass original length for comparison so we stay within i_size */ 32041a287cfeSDavid Sterba ret = btrfs_cmp_data(olen, &cmp); 3205416161dbSMark Fasheh if (ret == 0) 32061c919a5eSMark Fasheh ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1); 3207416161dbSMark Fasheh 32080efa9f48SMark Fasheh if (same_inode) 32090efa9f48SMark Fasheh unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start, 32100efa9f48SMark Fasheh same_lock_start + same_lock_len - 1); 32110efa9f48SMark Fasheh else 3212f4414602SMark Fasheh btrfs_double_extent_unlock(src, loff, dst, dst_loff, len); 3213f4414602SMark Fasheh 3214f4414602SMark Fasheh btrfs_cmp_data_free(&cmp); 3215416161dbSMark Fasheh out_unlock: 32160efa9f48SMark Fasheh if (same_inode) 32175955102cSAl Viro inode_unlock(src); 32180efa9f48SMark Fasheh else 3219f4414602SMark Fasheh btrfs_double_inode_unlock(src, dst); 3220416161dbSMark Fasheh 3221416161dbSMark Fasheh return ret; 3222416161dbSMark Fasheh } 3223416161dbSMark Fasheh 3224ee22184bSByongho Lee #define BTRFS_MAX_DEDUPE_LEN SZ_16M 3225416161dbSMark Fasheh 32262b3909f8SDarrick J. Wong ssize_t btrfs_dedupe_file_range(struct file *src_file, u64 loff, u64 olen, 32272b3909f8SDarrick J. Wong struct file *dst_file, u64 dst_loff) 3228416161dbSMark Fasheh { 32292b3909f8SDarrick J. Wong struct inode *src = file_inode(src_file); 32302b3909f8SDarrick J. Wong struct inode *dst = file_inode(dst_file); 3231416161dbSMark Fasheh u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize; 32322b3909f8SDarrick J. Wong ssize_t res; 3233416161dbSMark Fasheh 32342b3909f8SDarrick J. Wong if (olen > BTRFS_MAX_DEDUPE_LEN) 32352b3909f8SDarrick J. Wong olen = BTRFS_MAX_DEDUPE_LEN; 3236416161dbSMark Fasheh 323709cbfeafSKirill A. Shutemov if (WARN_ON_ONCE(bs < PAGE_SIZE)) { 3238416161dbSMark Fasheh /* 3239416161dbSMark Fasheh * Btrfs does not support blocksize < page_size. As a 3240416161dbSMark Fasheh * result, btrfs_cmp_data() won't correctly handle 3241416161dbSMark Fasheh * this situation without an update. 3242416161dbSMark Fasheh */ 32432b3909f8SDarrick J. Wong return -EINVAL; 3244416161dbSMark Fasheh } 3245416161dbSMark Fasheh 32462b3909f8SDarrick J. Wong res = btrfs_extent_same(src, loff, olen, dst, dst_loff); 32472b3909f8SDarrick J. Wong if (res) 32482b3909f8SDarrick J. Wong return res; 32492b3909f8SDarrick J. Wong return olen; 3250416161dbSMark Fasheh } 3251416161dbSMark Fasheh 3252f82a9901SFilipe Manana static int clone_finish_inode_update(struct btrfs_trans_handle *trans, 3253f82a9901SFilipe Manana struct inode *inode, 3254f82a9901SFilipe Manana u64 endoff, 3255f82a9901SFilipe Manana const u64 destoff, 32561c919a5eSMark Fasheh const u64 olen, 32571c919a5eSMark Fasheh int no_time_update) 3258f82a9901SFilipe Manana { 3259f82a9901SFilipe Manana struct btrfs_root *root = BTRFS_I(inode)->root; 3260f82a9901SFilipe Manana int ret; 3261f82a9901SFilipe Manana 3262f82a9901SFilipe Manana inode_inc_iversion(inode); 32631c919a5eSMark Fasheh if (!no_time_update) 3264c2050a45SDeepa Dinamani inode->i_mtime = inode->i_ctime = current_time(inode); 3265f82a9901SFilipe Manana /* 3266f82a9901SFilipe Manana * We round up to the block size at eof when determining which 3267f82a9901SFilipe Manana * extents to clone above, but shouldn't round up the file size. 3268f82a9901SFilipe Manana */ 3269f82a9901SFilipe Manana if (endoff > destoff + olen) 3270f82a9901SFilipe Manana endoff = destoff + olen; 3271f82a9901SFilipe Manana if (endoff > inode->i_size) 32726ef06d27SNikolay Borisov btrfs_i_size_write(BTRFS_I(inode), endoff); 3273f82a9901SFilipe Manana 3274f82a9901SFilipe Manana ret = btrfs_update_inode(trans, root, inode); 3275f82a9901SFilipe Manana if (ret) { 327666642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 32773a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3278f82a9901SFilipe Manana goto out; 3279f82a9901SFilipe Manana } 32803a45bb20SJeff Mahoney ret = btrfs_end_transaction(trans); 3281f82a9901SFilipe Manana out: 3282f82a9901SFilipe Manana return ret; 3283f82a9901SFilipe Manana } 3284f82a9901SFilipe Manana 3285a2f392e4SNikolay Borisov static void clone_update_extent_map(struct btrfs_inode *inode, 32867ffbb598SFilipe Manana const struct btrfs_trans_handle *trans, 32877ffbb598SFilipe Manana const struct btrfs_path *path, 32887ffbb598SFilipe Manana const u64 hole_offset, 32897ffbb598SFilipe Manana const u64 hole_len) 32907ffbb598SFilipe Manana { 3291a2f392e4SNikolay Borisov struct extent_map_tree *em_tree = &inode->extent_tree; 32927ffbb598SFilipe Manana struct extent_map *em; 32937ffbb598SFilipe Manana int ret; 32947ffbb598SFilipe Manana 32957ffbb598SFilipe Manana em = alloc_extent_map(); 32967ffbb598SFilipe Manana if (!em) { 3297a2f392e4SNikolay Borisov set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags); 32987ffbb598SFilipe Manana return; 32997ffbb598SFilipe Manana } 33007ffbb598SFilipe Manana 330114f59796SFilipe Manana if (path) { 330214f59796SFilipe Manana struct btrfs_file_extent_item *fi; 330314f59796SFilipe Manana 330414f59796SFilipe Manana fi = btrfs_item_ptr(path->nodes[0], path->slots[0], 330514f59796SFilipe Manana struct btrfs_file_extent_item); 3306a2f392e4SNikolay Borisov btrfs_extent_item_to_extent_map(inode, path, fi, false, em); 33077ffbb598SFilipe Manana em->generation = -1; 33087ffbb598SFilipe Manana if (btrfs_file_extent_type(path->nodes[0], fi) == 33097ffbb598SFilipe Manana BTRFS_FILE_EXTENT_INLINE) 33107ffbb598SFilipe Manana set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, 3311a2f392e4SNikolay Borisov &inode->runtime_flags); 33127ffbb598SFilipe Manana } else { 33137ffbb598SFilipe Manana em->start = hole_offset; 33147ffbb598SFilipe Manana em->len = hole_len; 33157ffbb598SFilipe Manana em->ram_bytes = em->len; 33167ffbb598SFilipe Manana em->orig_start = hole_offset; 33177ffbb598SFilipe Manana em->block_start = EXTENT_MAP_HOLE; 33187ffbb598SFilipe Manana em->block_len = 0; 33197ffbb598SFilipe Manana em->orig_block_len = 0; 33207ffbb598SFilipe Manana em->compress_type = BTRFS_COMPRESS_NONE; 33217ffbb598SFilipe Manana em->generation = trans->transid; 33227ffbb598SFilipe Manana } 33237ffbb598SFilipe Manana 33247ffbb598SFilipe Manana while (1) { 33257ffbb598SFilipe Manana write_lock(&em_tree->lock); 33267ffbb598SFilipe Manana ret = add_extent_mapping(em_tree, em, 1); 33277ffbb598SFilipe Manana write_unlock(&em_tree->lock); 33287ffbb598SFilipe Manana if (ret != -EEXIST) { 33297ffbb598SFilipe Manana free_extent_map(em); 33307ffbb598SFilipe Manana break; 33317ffbb598SFilipe Manana } 3332a2f392e4SNikolay Borisov btrfs_drop_extent_cache(inode, em->start, 33337ffbb598SFilipe Manana em->start + em->len - 1, 0); 33347ffbb598SFilipe Manana } 33357ffbb598SFilipe Manana 3336ee39b432SDavid Sterba if (ret) 3337a2f392e4SNikolay Borisov set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags); 33387ffbb598SFilipe Manana } 33397ffbb598SFilipe Manana 33408039d87dSFilipe Manana /* 33418039d87dSFilipe Manana * Make sure we do not end up inserting an inline extent into a file that has 33428039d87dSFilipe Manana * already other (non-inline) extents. If a file has an inline extent it can 33438039d87dSFilipe Manana * not have any other extents and the (single) inline extent must start at the 33448039d87dSFilipe Manana * file offset 0. Failing to respect these rules will lead to file corruption, 33458039d87dSFilipe Manana * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc 33468039d87dSFilipe Manana * 33478039d87dSFilipe Manana * We can have extents that have been already written to disk or we can have 33488039d87dSFilipe Manana * dirty ranges still in delalloc, in which case the extent maps and items are 33498039d87dSFilipe Manana * created only when we run delalloc, and the delalloc ranges might fall outside 33508039d87dSFilipe Manana * the range we are currently locking in the inode's io tree. So we check the 33518039d87dSFilipe Manana * inode's i_size because of that (i_size updates are done while holding the 33528039d87dSFilipe Manana * i_mutex, which we are holding here). 33538039d87dSFilipe Manana * We also check to see if the inode has a size not greater than "datal" but has 33548039d87dSFilipe Manana * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are 33558039d87dSFilipe Manana * protected against such concurrent fallocate calls by the i_mutex). 33568039d87dSFilipe Manana * 33578039d87dSFilipe Manana * If the file has no extents but a size greater than datal, do not allow the 33588039d87dSFilipe Manana * copy because we would need turn the inline extent into a non-inline one (even 33598039d87dSFilipe Manana * with NO_HOLES enabled). If we find our destination inode only has one inline 33608039d87dSFilipe Manana * extent, just overwrite it with the source inline extent if its size is less 33618039d87dSFilipe Manana * than the source extent's size, or we could copy the source inline extent's 33628039d87dSFilipe Manana * data into the destination inode's inline extent if the later is greater then 33638039d87dSFilipe Manana * the former. 33648039d87dSFilipe Manana */ 33654a0ab9d7SDavid Sterba static int clone_copy_inline_extent(struct inode *dst, 33668039d87dSFilipe Manana struct btrfs_trans_handle *trans, 33678039d87dSFilipe Manana struct btrfs_path *path, 33688039d87dSFilipe Manana struct btrfs_key *new_key, 33698039d87dSFilipe Manana const u64 drop_start, 33708039d87dSFilipe Manana const u64 datal, 33718039d87dSFilipe Manana const u64 skip, 33728039d87dSFilipe Manana const u64 size, 33738039d87dSFilipe Manana char *inline_data) 33748039d87dSFilipe Manana { 33750b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dst->i_sb); 33768039d87dSFilipe Manana struct btrfs_root *root = BTRFS_I(dst)->root; 33778039d87dSFilipe Manana const u64 aligned_end = ALIGN(new_key->offset + datal, 33780b246afaSJeff Mahoney fs_info->sectorsize); 33798039d87dSFilipe Manana int ret; 33808039d87dSFilipe Manana struct btrfs_key key; 33818039d87dSFilipe Manana 33828039d87dSFilipe Manana if (new_key->offset > 0) 33838039d87dSFilipe Manana return -EOPNOTSUPP; 33848039d87dSFilipe Manana 33854a0cc7caSNikolay Borisov key.objectid = btrfs_ino(BTRFS_I(dst)); 33868039d87dSFilipe Manana key.type = BTRFS_EXTENT_DATA_KEY; 33878039d87dSFilipe Manana key.offset = 0; 33888039d87dSFilipe Manana ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 33898039d87dSFilipe Manana if (ret < 0) { 33908039d87dSFilipe Manana return ret; 33918039d87dSFilipe Manana } else if (ret > 0) { 33928039d87dSFilipe Manana if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { 33938039d87dSFilipe Manana ret = btrfs_next_leaf(root, path); 33948039d87dSFilipe Manana if (ret < 0) 33958039d87dSFilipe Manana return ret; 33968039d87dSFilipe Manana else if (ret > 0) 33978039d87dSFilipe Manana goto copy_inline_extent; 33988039d87dSFilipe Manana } 33998039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 34004a0cc7caSNikolay Borisov if (key.objectid == btrfs_ino(BTRFS_I(dst)) && 34018039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) { 34028039d87dSFilipe Manana ASSERT(key.offset > 0); 34038039d87dSFilipe Manana return -EOPNOTSUPP; 34048039d87dSFilipe Manana } 34058039d87dSFilipe Manana } else if (i_size_read(dst) <= datal) { 34068039d87dSFilipe Manana struct btrfs_file_extent_item *ei; 34078039d87dSFilipe Manana u64 ext_len; 34088039d87dSFilipe Manana 34098039d87dSFilipe Manana /* 34108039d87dSFilipe Manana * If the file size is <= datal, make sure there are no other 34118039d87dSFilipe Manana * extents following (can happen do to an fallocate call with 34128039d87dSFilipe Manana * the flag FALLOC_FL_KEEP_SIZE). 34138039d87dSFilipe Manana */ 34148039d87dSFilipe Manana ei = btrfs_item_ptr(path->nodes[0], path->slots[0], 34158039d87dSFilipe Manana struct btrfs_file_extent_item); 34168039d87dSFilipe Manana /* 34178039d87dSFilipe Manana * If it's an inline extent, it can not have other extents 34188039d87dSFilipe Manana * following it. 34198039d87dSFilipe Manana */ 34208039d87dSFilipe Manana if (btrfs_file_extent_type(path->nodes[0], ei) == 34218039d87dSFilipe Manana BTRFS_FILE_EXTENT_INLINE) 34228039d87dSFilipe Manana goto copy_inline_extent; 34238039d87dSFilipe Manana 34248039d87dSFilipe Manana ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei); 34258039d87dSFilipe Manana if (ext_len > aligned_end) 34268039d87dSFilipe Manana return -EOPNOTSUPP; 34278039d87dSFilipe Manana 34288039d87dSFilipe Manana ret = btrfs_next_item(root, path); 34298039d87dSFilipe Manana if (ret < 0) { 34308039d87dSFilipe Manana return ret; 34318039d87dSFilipe Manana } else if (ret == 0) { 34328039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 34338039d87dSFilipe Manana path->slots[0]); 34344a0cc7caSNikolay Borisov if (key.objectid == btrfs_ino(BTRFS_I(dst)) && 34358039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) 34368039d87dSFilipe Manana return -EOPNOTSUPP; 34378039d87dSFilipe Manana } 34388039d87dSFilipe Manana } 34398039d87dSFilipe Manana 34408039d87dSFilipe Manana copy_inline_extent: 34418039d87dSFilipe Manana /* 34428039d87dSFilipe Manana * We have no extent items, or we have an extent at offset 0 which may 34438039d87dSFilipe Manana * or may not be inlined. All these cases are dealt the same way. 34448039d87dSFilipe Manana */ 34458039d87dSFilipe Manana if (i_size_read(dst) > datal) { 34468039d87dSFilipe Manana /* 34478039d87dSFilipe Manana * If the destination inode has an inline extent... 34488039d87dSFilipe Manana * This would require copying the data from the source inline 34498039d87dSFilipe Manana * extent into the beginning of the destination's inline extent. 34508039d87dSFilipe Manana * But this is really complex, both extents can be compressed 34518039d87dSFilipe Manana * or just one of them, which would require decompressing and 34528039d87dSFilipe Manana * re-compressing data (which could increase the new compressed 34538039d87dSFilipe Manana * size, not allowing the compressed data to fit anymore in an 34548039d87dSFilipe Manana * inline extent). 34558039d87dSFilipe Manana * So just don't support this case for now (it should be rare, 34568039d87dSFilipe Manana * we are not really saving space when cloning inline extents). 34578039d87dSFilipe Manana */ 34588039d87dSFilipe Manana return -EOPNOTSUPP; 34598039d87dSFilipe Manana } 34608039d87dSFilipe Manana 34618039d87dSFilipe Manana btrfs_release_path(path); 34628039d87dSFilipe Manana ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1); 34638039d87dSFilipe Manana if (ret) 34648039d87dSFilipe Manana return ret; 34658039d87dSFilipe Manana ret = btrfs_insert_empty_item(trans, root, path, new_key, size); 34668039d87dSFilipe Manana if (ret) 34678039d87dSFilipe Manana return ret; 34688039d87dSFilipe Manana 34698039d87dSFilipe Manana if (skip) { 34708039d87dSFilipe Manana const u32 start = btrfs_file_extent_calc_inline_size(0); 34718039d87dSFilipe Manana 34728039d87dSFilipe Manana memmove(inline_data + start, inline_data + start + skip, datal); 34738039d87dSFilipe Manana } 34748039d87dSFilipe Manana 34758039d87dSFilipe Manana write_extent_buffer(path->nodes[0], inline_data, 34768039d87dSFilipe Manana btrfs_item_ptr_offset(path->nodes[0], 34778039d87dSFilipe Manana path->slots[0]), 34788039d87dSFilipe Manana size); 34798039d87dSFilipe Manana inode_add_bytes(dst, datal); 34808039d87dSFilipe Manana 34818039d87dSFilipe Manana return 0; 34828039d87dSFilipe Manana } 34838039d87dSFilipe Manana 348432b7c687SMark Fasheh /** 348532b7c687SMark Fasheh * btrfs_clone() - clone a range from inode file to another 348632b7c687SMark Fasheh * 348732b7c687SMark Fasheh * @src: Inode to clone from 348832b7c687SMark Fasheh * @inode: Inode to clone to 348932b7c687SMark Fasheh * @off: Offset within source to start clone from 349032b7c687SMark Fasheh * @olen: Original length, passed by user, of range to clone 34911c919a5eSMark Fasheh * @olen_aligned: Block-aligned value of olen 349232b7c687SMark Fasheh * @destoff: Offset within @inode to start clone 34931c919a5eSMark Fasheh * @no_time_update: Whether to update mtime/ctime on the target inode 349432b7c687SMark Fasheh */ 349532b7c687SMark Fasheh static int btrfs_clone(struct inode *src, struct inode *inode, 3496f82a9901SFilipe Manana const u64 off, const u64 olen, const u64 olen_aligned, 34971c919a5eSMark Fasheh const u64 destoff, int no_time_update) 3498f46b5a66SChristoph Hellwig { 34990b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 3500f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 350132b7c687SMark Fasheh struct btrfs_path *path = NULL; 3502f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 350332b7c687SMark Fasheh struct btrfs_trans_handle *trans; 350432b7c687SMark Fasheh char *buf = NULL; 3505ae01a0abSYan Zheng struct btrfs_key key; 3506f46b5a66SChristoph Hellwig u32 nritems; 3507f46b5a66SChristoph Hellwig int slot; 3508ae01a0abSYan Zheng int ret; 3509f82a9901SFilipe Manana const u64 len = olen_aligned; 3510f82a9901SFilipe Manana u64 last_dest_end = destoff; 3511ae01a0abSYan Zheng 3512ae01a0abSYan Zheng ret = -ENOMEM; 3513752ade68SMichal Hocko buf = kvmalloc(fs_info->nodesize, GFP_KERNEL); 3514ae01a0abSYan Zheng if (!buf) 351532b7c687SMark Fasheh return ret; 3516ae01a0abSYan Zheng 3517ae01a0abSYan Zheng path = btrfs_alloc_path(); 3518ae01a0abSYan Zheng if (!path) { 351915351955SDavid Sterba kvfree(buf); 352032b7c687SMark Fasheh return ret; 3521ae01a0abSYan Zheng } 352232b7c687SMark Fasheh 3523e4058b54SDavid Sterba path->reada = READA_FORWARD; 3524c5c9cd4dSSage Weil /* clone data */ 35254a0cc7caSNikolay Borisov key.objectid = btrfs_ino(BTRFS_I(src)); 3526ae01a0abSYan Zheng key.type = BTRFS_EXTENT_DATA_KEY; 35272c463823SFilipe Manana key.offset = off; 3528f46b5a66SChristoph Hellwig 3529f46b5a66SChristoph Hellwig while (1) { 3530de249e66SChris Mason u64 next_key_min_offset = key.offset + 1; 3531df858e76SFilipe Manana 3532f46b5a66SChristoph Hellwig /* 3533f46b5a66SChristoph Hellwig * note the key will change type as we walk through the 3534f46b5a66SChristoph Hellwig * tree. 3535f46b5a66SChristoph Hellwig */ 3536e4355f34SFilipe David Borba Manana path->leave_spinning = 1; 3537362a20c5SDavid Sterba ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path, 3538362a20c5SDavid Sterba 0, 0); 3539f46b5a66SChristoph Hellwig if (ret < 0) 3540f46b5a66SChristoph Hellwig goto out; 35412c463823SFilipe Manana /* 35422c463823SFilipe Manana * First search, if no extent item that starts at offset off was 35432c463823SFilipe Manana * found but the previous item is an extent item, it's possible 35442c463823SFilipe Manana * it might overlap our target range, therefore process it. 35452c463823SFilipe Manana */ 35462c463823SFilipe Manana if (key.offset == off && ret > 0 && path->slots[0] > 0) { 35472c463823SFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 35482c463823SFilipe Manana path->slots[0] - 1); 35492c463823SFilipe Manana if (key.type == BTRFS_EXTENT_DATA_KEY) 35502c463823SFilipe Manana path->slots[0]--; 35512c463823SFilipe Manana } 3552f46b5a66SChristoph Hellwig 3553ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3554e4355f34SFilipe David Borba Manana process_slot: 3555ae01a0abSYan Zheng if (path->slots[0] >= nritems) { 3556362a20c5SDavid Sterba ret = btrfs_next_leaf(BTRFS_I(src)->root, path); 3557f46b5a66SChristoph Hellwig if (ret < 0) 3558f46b5a66SChristoph Hellwig goto out; 3559f46b5a66SChristoph Hellwig if (ret > 0) 3560f46b5a66SChristoph Hellwig break; 3561ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3562f46b5a66SChristoph Hellwig } 3563f46b5a66SChristoph Hellwig leaf = path->nodes[0]; 3564f46b5a66SChristoph Hellwig slot = path->slots[0]; 3565f46b5a66SChristoph Hellwig 3566ae01a0abSYan Zheng btrfs_item_key_to_cpu(leaf, &key, slot); 3567962a298fSDavid Sterba if (key.type > BTRFS_EXTENT_DATA_KEY || 35684a0cc7caSNikolay Borisov key.objectid != btrfs_ino(BTRFS_I(src))) 3569f46b5a66SChristoph Hellwig break; 3570f46b5a66SChristoph Hellwig 3571962a298fSDavid Sterba if (key.type == BTRFS_EXTENT_DATA_KEY) { 3572c5c9cd4dSSage Weil struct btrfs_file_extent_item *extent; 3573c5c9cd4dSSage Weil int type; 357431840ae1SZheng Yan u32 size; 357531840ae1SZheng Yan struct btrfs_key new_key; 3576c5c9cd4dSSage Weil u64 disko = 0, diskl = 0; 3577c5c9cd4dSSage Weil u64 datao = 0, datal = 0; 3578c5c9cd4dSSage Weil u8 comp; 3579f82a9901SFilipe Manana u64 drop_start; 358031840ae1SZheng Yan 3581c5c9cd4dSSage Weil extent = btrfs_item_ptr(leaf, slot, 3582c5c9cd4dSSage Weil struct btrfs_file_extent_item); 3583c5c9cd4dSSage Weil comp = btrfs_file_extent_compression(leaf, extent); 3584c5c9cd4dSSage Weil type = btrfs_file_extent_type(leaf, extent); 3585c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3586c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3587d397712bSChris Mason disko = btrfs_file_extent_disk_bytenr(leaf, 3588d397712bSChris Mason extent); 3589d397712bSChris Mason diskl = btrfs_file_extent_disk_num_bytes(leaf, 3590d397712bSChris Mason extent); 3591c5c9cd4dSSage Weil datao = btrfs_file_extent_offset(leaf, extent); 3592d397712bSChris Mason datal = btrfs_file_extent_num_bytes(leaf, 3593d397712bSChris Mason extent); 3594c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3595c5c9cd4dSSage Weil /* take upper bound, may be compressed */ 3596c5c9cd4dSSage Weil datal = btrfs_file_extent_ram_bytes(leaf, 3597c5c9cd4dSSage Weil extent); 3598c5c9cd4dSSage Weil } 359931840ae1SZheng Yan 36002c463823SFilipe Manana /* 36012c463823SFilipe Manana * The first search might have left us at an extent 36022c463823SFilipe Manana * item that ends before our target range's start, can 36032c463823SFilipe Manana * happen if we have holes and NO_HOLES feature enabled. 36042c463823SFilipe Manana */ 36052c463823SFilipe Manana if (key.offset + datal <= off) { 3606e4355f34SFilipe David Borba Manana path->slots[0]++; 3607e4355f34SFilipe David Borba Manana goto process_slot; 36082c463823SFilipe Manana } else if (key.offset >= off + len) { 36092c463823SFilipe Manana break; 3610e4355f34SFilipe David Borba Manana } 3611df858e76SFilipe Manana next_key_min_offset = key.offset + datal; 3612e4355f34SFilipe David Borba Manana size = btrfs_item_size_nr(leaf, slot); 3613e4355f34SFilipe David Borba Manana read_extent_buffer(leaf, buf, 3614e4355f34SFilipe David Borba Manana btrfs_item_ptr_offset(leaf, slot), 3615e4355f34SFilipe David Borba Manana size); 3616e4355f34SFilipe David Borba Manana 3617e4355f34SFilipe David Borba Manana btrfs_release_path(path); 3618e4355f34SFilipe David Borba Manana path->leave_spinning = 0; 3619c5c9cd4dSSage Weil 362031840ae1SZheng Yan memcpy(&new_key, &key, sizeof(new_key)); 36214a0cc7caSNikolay Borisov new_key.objectid = btrfs_ino(BTRFS_I(inode)); 36224d728ec7SLi Zefan if (off <= key.offset) 3623c5c9cd4dSSage Weil new_key.offset = key.offset + destoff - off; 36244d728ec7SLi Zefan else 36254d728ec7SLi Zefan new_key.offset = destoff; 3626c5c9cd4dSSage Weil 3627b6f3409bSSage Weil /* 3628f82a9901SFilipe Manana * Deal with a hole that doesn't have an extent item 3629f82a9901SFilipe Manana * that represents it (NO_HOLES feature enabled). 3630f82a9901SFilipe Manana * This hole is either in the middle of the cloning 3631f82a9901SFilipe Manana * range or at the beginning (fully overlaps it or 3632f82a9901SFilipe Manana * partially overlaps it). 3633f82a9901SFilipe Manana */ 3634f82a9901SFilipe Manana if (new_key.offset != last_dest_end) 3635f82a9901SFilipe Manana drop_start = last_dest_end; 3636f82a9901SFilipe Manana else 3637f82a9901SFilipe Manana drop_start = new_key.offset; 3638f82a9901SFilipe Manana 3639f82a9901SFilipe Manana /* 3640b6f3409bSSage Weil * 1 - adjusting old extent (we may have to split it) 3641b6f3409bSSage Weil * 1 - add new extent 3642b6f3409bSSage Weil * 1 - inode update 3643b6f3409bSSage Weil */ 3644b6f3409bSSage Weil trans = btrfs_start_transaction(root, 3); 3645a22285a6SYan, Zheng if (IS_ERR(trans)) { 3646a22285a6SYan, Zheng ret = PTR_ERR(trans); 3647a22285a6SYan, Zheng goto out; 3648a22285a6SYan, Zheng } 3649a22285a6SYan, Zheng 3650c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3651c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3652d72c0842SLi Zefan /* 3653d72c0842SLi Zefan * a | --- range to clone ---| b 3654d72c0842SLi Zefan * | ------------- extent ------------- | 3655d72c0842SLi Zefan */ 3656d72c0842SLi Zefan 365793915584SAntonio Ospite /* subtract range b */ 3658d72c0842SLi Zefan if (key.offset + datal > off + len) 3659d72c0842SLi Zefan datal = off + len - key.offset; 3660d72c0842SLi Zefan 366193915584SAntonio Ospite /* subtract range a */ 3662a22285a6SYan, Zheng if (off > key.offset) { 3663a22285a6SYan, Zheng datao += off - key.offset; 3664a22285a6SYan, Zheng datal -= off - key.offset; 3665a22285a6SYan, Zheng } 3666a22285a6SYan, Zheng 36675dc562c5SJosef Bacik ret = btrfs_drop_extents(trans, root, inode, 3668f82a9901SFilipe Manana drop_start, 3669a22285a6SYan, Zheng new_key.offset + datal, 36702671485dSJosef Bacik 1); 367179787eaaSJeff Mahoney if (ret) { 36723f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 367300fdf13aSLiu Bo btrfs_abort_transaction(trans, 367466642832SJeff Mahoney ret); 36753a45bb20SJeff Mahoney btrfs_end_transaction(trans); 367679787eaaSJeff Mahoney goto out; 367779787eaaSJeff Mahoney } 3678a22285a6SYan, Zheng 367931840ae1SZheng Yan ret = btrfs_insert_empty_item(trans, root, path, 368031840ae1SZheng Yan &new_key, size); 368179787eaaSJeff Mahoney if (ret) { 368266642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 36833a45bb20SJeff Mahoney btrfs_end_transaction(trans); 368479787eaaSJeff Mahoney goto out; 368579787eaaSJeff Mahoney } 368631840ae1SZheng Yan 368731840ae1SZheng Yan leaf = path->nodes[0]; 368831840ae1SZheng Yan slot = path->slots[0]; 368931840ae1SZheng Yan write_extent_buffer(leaf, buf, 369031840ae1SZheng Yan btrfs_item_ptr_offset(leaf, slot), 369131840ae1SZheng Yan size); 3692ae01a0abSYan Zheng 3693f46b5a66SChristoph Hellwig extent = btrfs_item_ptr(leaf, slot, 3694f46b5a66SChristoph Hellwig struct btrfs_file_extent_item); 3695c5c9cd4dSSage Weil 3696c5c9cd4dSSage Weil /* disko == 0 means it's a hole */ 3697c5c9cd4dSSage Weil if (!disko) 3698c5c9cd4dSSage Weil datao = 0; 3699c5c9cd4dSSage Weil 3700c5c9cd4dSSage Weil btrfs_set_file_extent_offset(leaf, extent, 3701c5c9cd4dSSage Weil datao); 3702c5c9cd4dSSage Weil btrfs_set_file_extent_num_bytes(leaf, extent, 3703c5c9cd4dSSage Weil datal); 3704fcebe456SJosef Bacik 3705c5c9cd4dSSage Weil if (disko) { 3706c5c9cd4dSSage Weil inode_add_bytes(inode, datal); 37072ff7e61eSJeff Mahoney ret = btrfs_inc_extent_ref(trans, 37082ff7e61eSJeff Mahoney fs_info, 37095d4f98a2SYan Zheng disko, diskl, 0, 3710f46b5a66SChristoph Hellwig root->root_key.objectid, 37114a0cc7caSNikolay Borisov btrfs_ino(BTRFS_I(inode)), 3712b06c4bf5SFilipe Manana new_key.offset - datao); 371379787eaaSJeff Mahoney if (ret) { 371479787eaaSJeff Mahoney btrfs_abort_transaction(trans, 371579787eaaSJeff Mahoney ret); 37163a45bb20SJeff Mahoney btrfs_end_transaction(trans); 371779787eaaSJeff Mahoney goto out; 371879787eaaSJeff Mahoney 371979787eaaSJeff Mahoney } 3720f46b5a66SChristoph Hellwig } 3721c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3722c5c9cd4dSSage Weil u64 skip = 0; 3723c5c9cd4dSSage Weil u64 trim = 0; 3724ed958762SFilipe Manana 3725c5c9cd4dSSage Weil if (off > key.offset) { 3726c5c9cd4dSSage Weil skip = off - key.offset; 3727c5c9cd4dSSage Weil new_key.offset += skip; 372831840ae1SZheng Yan } 3729d397712bSChris Mason 3730c5c9cd4dSSage Weil if (key.offset + datal > off + len) 3731c5c9cd4dSSage Weil trim = key.offset + datal - (off + len); 3732d397712bSChris Mason 3733c5c9cd4dSSage Weil if (comp && (skip || trim)) { 3734c5c9cd4dSSage Weil ret = -EINVAL; 37353a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3736c5c9cd4dSSage Weil goto out; 373731840ae1SZheng Yan } 3738c5c9cd4dSSage Weil size -= skip + trim; 3739c5c9cd4dSSage Weil datal -= skip + trim; 3740a22285a6SYan, Zheng 37414a0ab9d7SDavid Sterba ret = clone_copy_inline_extent(inode, 37428039d87dSFilipe Manana trans, path, 37438039d87dSFilipe Manana &new_key, 3744f82a9901SFilipe Manana drop_start, 37458039d87dSFilipe Manana datal, 37468039d87dSFilipe Manana skip, size, buf); 374779787eaaSJeff Mahoney if (ret) { 37483f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 37493a29bc09SChris Mason btrfs_abort_transaction(trans, 375079787eaaSJeff Mahoney ret); 37513a45bb20SJeff Mahoney btrfs_end_transaction(trans); 375279787eaaSJeff Mahoney goto out; 375379787eaaSJeff Mahoney } 3754c5c9cd4dSSage Weil leaf = path->nodes[0]; 3755c5c9cd4dSSage Weil slot = path->slots[0]; 3756c5c9cd4dSSage Weil } 3757c5c9cd4dSSage Weil 37587ffbb598SFilipe Manana /* If we have an implicit hole (NO_HOLES feature). */ 37597ffbb598SFilipe Manana if (drop_start < new_key.offset) 3760a2f392e4SNikolay Borisov clone_update_extent_map(BTRFS_I(inode), trans, 376114f59796SFilipe Manana NULL, drop_start, 37627ffbb598SFilipe Manana new_key.offset - drop_start); 37637ffbb598SFilipe Manana 3764a2f392e4SNikolay Borisov clone_update_extent_map(BTRFS_I(inode), trans, 3765a2f392e4SNikolay Borisov path, 0, 0); 37667ffbb598SFilipe Manana 3767c5c9cd4dSSage Weil btrfs_mark_buffer_dirty(leaf); 3768b3b4aa74SDavid Sterba btrfs_release_path(path); 3769c5c9cd4dSSage Weil 377062e2390eSFilipe Manana last_dest_end = ALIGN(new_key.offset + datal, 37710b246afaSJeff Mahoney fs_info->sectorsize); 3772f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, 3773f82a9901SFilipe Manana last_dest_end, 37741c919a5eSMark Fasheh destoff, olen, 37751c919a5eSMark Fasheh no_time_update); 3776f82a9901SFilipe Manana if (ret) 377779787eaaSJeff Mahoney goto out; 37782c463823SFilipe Manana if (new_key.offset + datal >= destoff + len) 37792c463823SFilipe Manana break; 3780a22285a6SYan, Zheng } 3781b3b4aa74SDavid Sterba btrfs_release_path(path); 3782df858e76SFilipe Manana key.offset = next_key_min_offset; 378369ae5e44SWang Xiaoguang 378469ae5e44SWang Xiaoguang if (fatal_signal_pending(current)) { 378569ae5e44SWang Xiaoguang ret = -EINTR; 378669ae5e44SWang Xiaoguang goto out; 378769ae5e44SWang Xiaoguang } 3788ae01a0abSYan Zheng } 3789f46b5a66SChristoph Hellwig ret = 0; 379032b7c687SMark Fasheh 3791f82a9901SFilipe Manana if (last_dest_end < destoff + len) { 3792f82a9901SFilipe Manana /* 3793f82a9901SFilipe Manana * We have an implicit hole (NO_HOLES feature is enabled) that 3794f82a9901SFilipe Manana * fully or partially overlaps our cloning range at its end. 3795f82a9901SFilipe Manana */ 3796b3b4aa74SDavid Sterba btrfs_release_path(path); 3797f82a9901SFilipe Manana 3798f82a9901SFilipe Manana /* 3799f82a9901SFilipe Manana * 1 - remove extent(s) 3800f82a9901SFilipe Manana * 1 - inode update 3801f82a9901SFilipe Manana */ 3802f82a9901SFilipe Manana trans = btrfs_start_transaction(root, 2); 3803f82a9901SFilipe Manana if (IS_ERR(trans)) { 3804f82a9901SFilipe Manana ret = PTR_ERR(trans); 3805f82a9901SFilipe Manana goto out; 3806f82a9901SFilipe Manana } 3807f82a9901SFilipe Manana ret = btrfs_drop_extents(trans, root, inode, 3808f82a9901SFilipe Manana last_dest_end, destoff + len, 1); 3809f82a9901SFilipe Manana if (ret) { 3810f82a9901SFilipe Manana if (ret != -EOPNOTSUPP) 381166642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 38123a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3813f82a9901SFilipe Manana goto out; 3814f82a9901SFilipe Manana } 3815a2f392e4SNikolay Borisov clone_update_extent_map(BTRFS_I(inode), trans, NULL, 3816a2f392e4SNikolay Borisov last_dest_end, 381714f59796SFilipe Manana destoff + len - last_dest_end); 3818f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, destoff + len, 38191c919a5eSMark Fasheh destoff, olen, no_time_update); 3820f82a9901SFilipe Manana } 3821f82a9901SFilipe Manana 3822f46b5a66SChristoph Hellwig out: 382332b7c687SMark Fasheh btrfs_free_path(path); 382415351955SDavid Sterba kvfree(buf); 382532b7c687SMark Fasheh return ret; 382632b7c687SMark Fasheh } 382732b7c687SMark Fasheh 38283db11b2eSZach Brown static noinline int btrfs_clone_files(struct file *file, struct file *file_src, 382932b7c687SMark Fasheh u64 off, u64 olen, u64 destoff) 383032b7c687SMark Fasheh { 383154563d41SAl Viro struct inode *inode = file_inode(file); 38323db11b2eSZach Brown struct inode *src = file_inode(file_src); 38330b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 383432b7c687SMark Fasheh struct btrfs_root *root = BTRFS_I(inode)->root; 383532b7c687SMark Fasheh int ret; 383632b7c687SMark Fasheh u64 len = olen; 38370b246afaSJeff Mahoney u64 bs = fs_info->sb->s_blocksize; 38383db11b2eSZach Brown int same_inode = src == inode; 383932b7c687SMark Fasheh 384032b7c687SMark Fasheh /* 384132b7c687SMark Fasheh * TODO: 384232b7c687SMark Fasheh * - split compressed inline extents. annoying: we need to 384332b7c687SMark Fasheh * decompress into destination's address_space (the file offset 384432b7c687SMark Fasheh * may change, so source mapping won't do), then recompress (or 384532b7c687SMark Fasheh * otherwise reinsert) a subrange. 384600fdf13aSLiu Bo * 384700fdf13aSLiu Bo * - split destination inode's inline extents. The inline extents can 384800fdf13aSLiu Bo * be either compressed or non-compressed. 384932b7c687SMark Fasheh */ 385032b7c687SMark Fasheh 385132b7c687SMark Fasheh if (btrfs_root_readonly(root)) 385232b7c687SMark Fasheh return -EROFS; 385332b7c687SMark Fasheh 38543db11b2eSZach Brown if (file_src->f_path.mnt != file->f_path.mnt || 38553db11b2eSZach Brown src->i_sb != inode->i_sb) 38563db11b2eSZach Brown return -EXDEV; 385732b7c687SMark Fasheh 385832b7c687SMark Fasheh /* don't make the dst file partly checksummed */ 385932b7c687SMark Fasheh if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != 386032b7c687SMark Fasheh (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) 38613db11b2eSZach Brown return -EINVAL; 386232b7c687SMark Fasheh 386332b7c687SMark Fasheh if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode)) 38643db11b2eSZach Brown return -EISDIR; 386532b7c687SMark Fasheh 386632b7c687SMark Fasheh if (!same_inode) { 3867293a8489SMark Fasheh btrfs_double_inode_lock(src, inode); 386832b7c687SMark Fasheh } else { 38695955102cSAl Viro inode_lock(src); 387032b7c687SMark Fasheh } 387132b7c687SMark Fasheh 387232b7c687SMark Fasheh /* determine range to clone */ 387332b7c687SMark Fasheh ret = -EINVAL; 387432b7c687SMark Fasheh if (off + len > src->i_size || off + len < off) 387532b7c687SMark Fasheh goto out_unlock; 387632b7c687SMark Fasheh if (len == 0) 387732b7c687SMark Fasheh olen = len = src->i_size - off; 387832b7c687SMark Fasheh /* if we extend to eof, continue to block boundary */ 387932b7c687SMark Fasheh if (off + len == src->i_size) 388032b7c687SMark Fasheh len = ALIGN(src->i_size, bs) - off; 388132b7c687SMark Fasheh 3882ccccf3d6SFilipe Manana if (len == 0) { 3883ccccf3d6SFilipe Manana ret = 0; 3884ccccf3d6SFilipe Manana goto out_unlock; 3885ccccf3d6SFilipe Manana } 3886ccccf3d6SFilipe Manana 388732b7c687SMark Fasheh /* verify the end result is block aligned */ 388832b7c687SMark Fasheh if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) || 388932b7c687SMark Fasheh !IS_ALIGNED(destoff, bs)) 389032b7c687SMark Fasheh goto out_unlock; 389132b7c687SMark Fasheh 389232b7c687SMark Fasheh /* verify if ranges are overlapped within the same file */ 389332b7c687SMark Fasheh if (same_inode) { 389432b7c687SMark Fasheh if (destoff + len > off && destoff < off + len) 389532b7c687SMark Fasheh goto out_unlock; 389632b7c687SMark Fasheh } 389732b7c687SMark Fasheh 389832b7c687SMark Fasheh if (destoff > inode->i_size) { 389932b7c687SMark Fasheh ret = btrfs_cont_expand(inode, inode->i_size, destoff); 390032b7c687SMark Fasheh if (ret) 390132b7c687SMark Fasheh goto out_unlock; 390232b7c687SMark Fasheh } 390332b7c687SMark Fasheh 3904c125b8bfSFilipe Manana /* 3905c125b8bfSFilipe Manana * Lock the target range too. Right after we replace the file extent 3906c125b8bfSFilipe Manana * items in the fs tree (which now point to the cloned data), we might 3907c125b8bfSFilipe Manana * have a worker replace them with extent items relative to a write 3908c125b8bfSFilipe Manana * operation that was issued before this clone operation (i.e. confront 3909c125b8bfSFilipe Manana * with inode.c:btrfs_finish_ordered_io). 3910c125b8bfSFilipe Manana */ 3911c125b8bfSFilipe Manana if (same_inode) { 3912c125b8bfSFilipe Manana u64 lock_start = min_t(u64, off, destoff); 3913c125b8bfSFilipe Manana u64 lock_len = max_t(u64, off, destoff) + len - lock_start; 391432b7c687SMark Fasheh 3915e0bd70c6SFilipe Manana ret = lock_extent_range(src, lock_start, lock_len, true); 3916c125b8bfSFilipe Manana } else { 3917e0bd70c6SFilipe Manana ret = btrfs_double_extent_lock(src, off, inode, destoff, len, 3918e0bd70c6SFilipe Manana true); 3919e0bd70c6SFilipe Manana } 3920e0bd70c6SFilipe Manana ASSERT(ret == 0); 3921e0bd70c6SFilipe Manana if (WARN_ON(ret)) { 3922e0bd70c6SFilipe Manana /* ranges in the io trees already unlocked */ 3923e0bd70c6SFilipe Manana goto out_unlock; 3924c125b8bfSFilipe Manana } 392532b7c687SMark Fasheh 39261c919a5eSMark Fasheh ret = btrfs_clone(src, inode, off, olen, len, destoff, 0); 392732b7c687SMark Fasheh 3928c125b8bfSFilipe Manana if (same_inode) { 3929c125b8bfSFilipe Manana u64 lock_start = min_t(u64, off, destoff); 3930c125b8bfSFilipe Manana u64 lock_end = max_t(u64, off, destoff) + len - 1; 3931c125b8bfSFilipe Manana 3932c125b8bfSFilipe Manana unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end); 3933c125b8bfSFilipe Manana } else { 3934293a8489SMark Fasheh btrfs_double_extent_unlock(src, off, inode, destoff, len); 3935c125b8bfSFilipe Manana } 3936c125b8bfSFilipe Manana /* 3937c125b8bfSFilipe Manana * Truncate page cache pages so that future reads will see the cloned 3938c125b8bfSFilipe Manana * data immediately and not the previous data. 3939c125b8bfSFilipe Manana */ 394065bfa658SChandan Rajendra truncate_inode_pages_range(&inode->i_data, 394109cbfeafSKirill A. Shutemov round_down(destoff, PAGE_SIZE), 394209cbfeafSKirill A. Shutemov round_up(destoff + len, PAGE_SIZE) - 1); 3943f46b5a66SChristoph Hellwig out_unlock: 3944293a8489SMark Fasheh if (!same_inode) 3945293a8489SMark Fasheh btrfs_double_inode_unlock(src, inode); 3946293a8489SMark Fasheh else 39475955102cSAl Viro inode_unlock(src); 39483db11b2eSZach Brown return ret; 39493db11b2eSZach Brown } 39503db11b2eSZach Brown 395104b38d60SChristoph Hellwig int btrfs_clone_file_range(struct file *src_file, loff_t off, 395204b38d60SChristoph Hellwig struct file *dst_file, loff_t destoff, u64 len) 39533db11b2eSZach Brown { 395404b38d60SChristoph Hellwig return btrfs_clone_files(dst_file, src_file, off, len, destoff); 3955c5c9cd4dSSage Weil } 3956c5c9cd4dSSage Weil 3957f46b5a66SChristoph Hellwig /* 3958f46b5a66SChristoph Hellwig * there are many ways the trans_start and trans_end ioctls can lead 3959f46b5a66SChristoph Hellwig * to deadlocks. They should only be used by applications that 3960f46b5a66SChristoph Hellwig * basically own the machine, and have a very in depth understanding 3961f46b5a66SChristoph Hellwig * of all the possible deadlocks and enospc problems. 3962f46b5a66SChristoph Hellwig */ 3963b2950863SChristoph Hellwig static long btrfs_ioctl_trans_start(struct file *file) 3964f46b5a66SChristoph Hellwig { 3965496ad9aaSAl Viro struct inode *inode = file_inode(file); 39660b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 3967f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 3968f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 39691ab86aedSSage Weil int ret; 3970f46b5a66SChristoph Hellwig 39711ab86aedSSage Weil ret = -EPERM; 3972df5b5520SChristoph Hellwig if (!capable(CAP_SYS_ADMIN)) 3973f46b5a66SChristoph Hellwig goto out; 39741ab86aedSSage Weil 39751ab86aedSSage Weil ret = -EINPROGRESS; 39761ab86aedSSage Weil if (file->private_data) 39771ab86aedSSage Weil goto out; 39789ca9ee09SSage Weil 3979b83cc969SLi Zefan ret = -EROFS; 3980b83cc969SLi Zefan if (btrfs_root_readonly(root)) 3981b83cc969SLi Zefan goto out; 3982b83cc969SLi Zefan 3983a561be71SAl Viro ret = mnt_want_write_file(file); 3984c146afadSYan Zheng if (ret) 3985c146afadSYan Zheng goto out; 3986c146afadSYan Zheng 39870b246afaSJeff Mahoney atomic_inc(&fs_info->open_ioctl_trans); 39889ca9ee09SSage Weil 3989f46b5a66SChristoph Hellwig ret = -ENOMEM; 39907a7eaa40SJosef Bacik trans = btrfs_start_ioctl_transaction(root); 3991abd30bb0STsutomu Itoh if (IS_ERR(trans)) 39921ab86aedSSage Weil goto out_drop; 39931ab86aedSSage Weil 39941ab86aedSSage Weil file->private_data = trans; 39951ab86aedSSage Weil return 0; 39961ab86aedSSage Weil 39971ab86aedSSage Weil out_drop: 39980b246afaSJeff Mahoney atomic_dec(&fs_info->open_ioctl_trans); 39992a79f17eSAl Viro mnt_drop_write_file(file); 4000f46b5a66SChristoph Hellwig out: 4001f46b5a66SChristoph Hellwig return ret; 4002f46b5a66SChristoph Hellwig } 4003f46b5a66SChristoph Hellwig 40046ef5ed0dSJosef Bacik static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) 40056ef5ed0dSJosef Bacik { 4006496ad9aaSAl Viro struct inode *inode = file_inode(file); 40070b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 40086ef5ed0dSJosef Bacik struct btrfs_root *root = BTRFS_I(inode)->root; 40096ef5ed0dSJosef Bacik struct btrfs_root *new_root; 40106ef5ed0dSJosef Bacik struct btrfs_dir_item *di; 40116ef5ed0dSJosef Bacik struct btrfs_trans_handle *trans; 40126ef5ed0dSJosef Bacik struct btrfs_path *path; 40136ef5ed0dSJosef Bacik struct btrfs_key location; 40146ef5ed0dSJosef Bacik struct btrfs_disk_key disk_key; 40156ef5ed0dSJosef Bacik u64 objectid = 0; 40166ef5ed0dSJosef Bacik u64 dir_id; 40173c04ce01SMiao Xie int ret; 40186ef5ed0dSJosef Bacik 40196ef5ed0dSJosef Bacik if (!capable(CAP_SYS_ADMIN)) 40206ef5ed0dSJosef Bacik return -EPERM; 40216ef5ed0dSJosef Bacik 40223c04ce01SMiao Xie ret = mnt_want_write_file(file); 40233c04ce01SMiao Xie if (ret) 40243c04ce01SMiao Xie return ret; 40253c04ce01SMiao Xie 40263c04ce01SMiao Xie if (copy_from_user(&objectid, argp, sizeof(objectid))) { 40273c04ce01SMiao Xie ret = -EFAULT; 40283c04ce01SMiao Xie goto out; 40293c04ce01SMiao Xie } 40306ef5ed0dSJosef Bacik 40316ef5ed0dSJosef Bacik if (!objectid) 40321cecf579Schandan objectid = BTRFS_FS_TREE_OBJECTID; 40336ef5ed0dSJosef Bacik 40346ef5ed0dSJosef Bacik location.objectid = objectid; 40356ef5ed0dSJosef Bacik location.type = BTRFS_ROOT_ITEM_KEY; 40366ef5ed0dSJosef Bacik location.offset = (u64)-1; 40376ef5ed0dSJosef Bacik 40380b246afaSJeff Mahoney new_root = btrfs_read_fs_root_no_name(fs_info, &location); 40393c04ce01SMiao Xie if (IS_ERR(new_root)) { 40403c04ce01SMiao Xie ret = PTR_ERR(new_root); 40413c04ce01SMiao Xie goto out; 40423c04ce01SMiao Xie } 40436ef5ed0dSJosef Bacik 40446ef5ed0dSJosef Bacik path = btrfs_alloc_path(); 40453c04ce01SMiao Xie if (!path) { 40463c04ce01SMiao Xie ret = -ENOMEM; 40473c04ce01SMiao Xie goto out; 40483c04ce01SMiao Xie } 40496ef5ed0dSJosef Bacik path->leave_spinning = 1; 40506ef5ed0dSJosef Bacik 40516ef5ed0dSJosef Bacik trans = btrfs_start_transaction(root, 1); 405298d5dc13STsutomu Itoh if (IS_ERR(trans)) { 40536ef5ed0dSJosef Bacik btrfs_free_path(path); 40543c04ce01SMiao Xie ret = PTR_ERR(trans); 40553c04ce01SMiao Xie goto out; 40566ef5ed0dSJosef Bacik } 40576ef5ed0dSJosef Bacik 40580b246afaSJeff Mahoney dir_id = btrfs_super_root_dir(fs_info->super_copy); 40590b246afaSJeff Mahoney di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path, 40606ef5ed0dSJosef Bacik dir_id, "default", 7, 1); 4061cf1e99a4SDan Carpenter if (IS_ERR_OR_NULL(di)) { 40626ef5ed0dSJosef Bacik btrfs_free_path(path); 40633a45bb20SJeff Mahoney btrfs_end_transaction(trans); 40640b246afaSJeff Mahoney btrfs_err(fs_info, 40655d163e0eSJeff Mahoney "Umm, you don't have the default diritem, this isn't going to work"); 40663c04ce01SMiao Xie ret = -ENOENT; 40673c04ce01SMiao Xie goto out; 40686ef5ed0dSJosef Bacik } 40696ef5ed0dSJosef Bacik 40706ef5ed0dSJosef Bacik btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); 40716ef5ed0dSJosef Bacik btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); 40726ef5ed0dSJosef Bacik btrfs_mark_buffer_dirty(path->nodes[0]); 40736ef5ed0dSJosef Bacik btrfs_free_path(path); 40746ef5ed0dSJosef Bacik 40750b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL); 40763a45bb20SJeff Mahoney btrfs_end_transaction(trans); 40773c04ce01SMiao Xie out: 40783c04ce01SMiao Xie mnt_drop_write_file(file); 40793c04ce01SMiao Xie return ret; 40806ef5ed0dSJosef Bacik } 40816ef5ed0dSJosef Bacik 40825af3e8ccSStefan Behrens void btrfs_get_block_group_info(struct list_head *groups_list, 4083bf5fc093SJosef Bacik struct btrfs_ioctl_space_info *space) 4084bf5fc093SJosef Bacik { 4085bf5fc093SJosef Bacik struct btrfs_block_group_cache *block_group; 4086bf5fc093SJosef Bacik 4087bf5fc093SJosef Bacik space->total_bytes = 0; 4088bf5fc093SJosef Bacik space->used_bytes = 0; 4089bf5fc093SJosef Bacik space->flags = 0; 4090bf5fc093SJosef Bacik list_for_each_entry(block_group, groups_list, list) { 4091bf5fc093SJosef Bacik space->flags = block_group->flags; 4092bf5fc093SJosef Bacik space->total_bytes += block_group->key.offset; 4093bf5fc093SJosef Bacik space->used_bytes += 4094bf5fc093SJosef Bacik btrfs_block_group_used(&block_group->item); 4095bf5fc093SJosef Bacik } 4096bf5fc093SJosef Bacik } 4097bf5fc093SJosef Bacik 40982ff7e61eSJeff Mahoney static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info, 40992ff7e61eSJeff Mahoney void __user *arg) 41001406e432SJosef Bacik { 41011406e432SJosef Bacik struct btrfs_ioctl_space_args space_args; 41021406e432SJosef Bacik struct btrfs_ioctl_space_info space; 41031406e432SJosef Bacik struct btrfs_ioctl_space_info *dest; 41047fde62bfSChris Mason struct btrfs_ioctl_space_info *dest_orig; 410513f2696fSDaniel J Blueman struct btrfs_ioctl_space_info __user *user_dest; 41061406e432SJosef Bacik struct btrfs_space_info *info; 4107bf5fc093SJosef Bacik u64 types[] = {BTRFS_BLOCK_GROUP_DATA, 4108bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_SYSTEM, 4109bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_METADATA, 4110bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA}; 4111bf5fc093SJosef Bacik int num_types = 4; 41127fde62bfSChris Mason int alloc_size; 41131406e432SJosef Bacik int ret = 0; 411451788b1bSDan Rosenberg u64 slot_count = 0; 4115bf5fc093SJosef Bacik int i, c; 41161406e432SJosef Bacik 41171406e432SJosef Bacik if (copy_from_user(&space_args, 41181406e432SJosef Bacik (struct btrfs_ioctl_space_args __user *)arg, 41191406e432SJosef Bacik sizeof(space_args))) 41201406e432SJosef Bacik return -EFAULT; 41211406e432SJosef Bacik 4122bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4123bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4124bf5fc093SJosef Bacik 4125bf5fc093SJosef Bacik info = NULL; 41267fde62bfSChris Mason rcu_read_lock(); 41270b246afaSJeff Mahoney list_for_each_entry_rcu(tmp, &fs_info->space_info, 4128bf5fc093SJosef Bacik list) { 4129bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4130bf5fc093SJosef Bacik info = tmp; 4131bf5fc093SJosef Bacik break; 4132bf5fc093SJosef Bacik } 4133bf5fc093SJosef Bacik } 41347fde62bfSChris Mason rcu_read_unlock(); 41351406e432SJosef Bacik 4136bf5fc093SJosef Bacik if (!info) 4137bf5fc093SJosef Bacik continue; 4138bf5fc093SJosef Bacik 4139bf5fc093SJosef Bacik down_read(&info->groups_sem); 4140bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4141bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) 4142bf5fc093SJosef Bacik slot_count++; 4143bf5fc093SJosef Bacik } 4144bf5fc093SJosef Bacik up_read(&info->groups_sem); 4145bf5fc093SJosef Bacik } 4146bf5fc093SJosef Bacik 414736523e95SDavid Sterba /* 414836523e95SDavid Sterba * Global block reserve, exported as a space_info 414936523e95SDavid Sterba */ 415036523e95SDavid Sterba slot_count++; 415136523e95SDavid Sterba 41527fde62bfSChris Mason /* space_slots == 0 means they are asking for a count */ 41537fde62bfSChris Mason if (space_args.space_slots == 0) { 41547fde62bfSChris Mason space_args.total_spaces = slot_count; 41557fde62bfSChris Mason goto out; 41567fde62bfSChris Mason } 4157bf5fc093SJosef Bacik 415851788b1bSDan Rosenberg slot_count = min_t(u64, space_args.space_slots, slot_count); 4159bf5fc093SJosef Bacik 41607fde62bfSChris Mason alloc_size = sizeof(*dest) * slot_count; 4161bf5fc093SJosef Bacik 41627fde62bfSChris Mason /* we generally have at most 6 or so space infos, one for each raid 41637fde62bfSChris Mason * level. So, a whole page should be more than enough for everyone 41647fde62bfSChris Mason */ 416509cbfeafSKirill A. Shutemov if (alloc_size > PAGE_SIZE) 41667fde62bfSChris Mason return -ENOMEM; 41677fde62bfSChris Mason 41687fde62bfSChris Mason space_args.total_spaces = 0; 41698d2db785SDavid Sterba dest = kmalloc(alloc_size, GFP_KERNEL); 41707fde62bfSChris Mason if (!dest) 41717fde62bfSChris Mason return -ENOMEM; 41727fde62bfSChris Mason dest_orig = dest; 41737fde62bfSChris Mason 41747fde62bfSChris Mason /* now we have a buffer to copy into */ 4175bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4176bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4177bf5fc093SJosef Bacik 417851788b1bSDan Rosenberg if (!slot_count) 417951788b1bSDan Rosenberg break; 418051788b1bSDan Rosenberg 4181bf5fc093SJosef Bacik info = NULL; 41821406e432SJosef Bacik rcu_read_lock(); 41830b246afaSJeff Mahoney list_for_each_entry_rcu(tmp, &fs_info->space_info, 4184bf5fc093SJosef Bacik list) { 4185bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4186bf5fc093SJosef Bacik info = tmp; 41877fde62bfSChris Mason break; 4188bf5fc093SJosef Bacik } 4189bf5fc093SJosef Bacik } 4190bf5fc093SJosef Bacik rcu_read_unlock(); 41917fde62bfSChris Mason 4192bf5fc093SJosef Bacik if (!info) 4193bf5fc093SJosef Bacik continue; 4194bf5fc093SJosef Bacik down_read(&info->groups_sem); 4195bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4196bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) { 41975af3e8ccSStefan Behrens btrfs_get_block_group_info( 41985af3e8ccSStefan Behrens &info->block_groups[c], &space); 41997fde62bfSChris Mason memcpy(dest, &space, sizeof(space)); 42001406e432SJosef Bacik dest++; 42011406e432SJosef Bacik space_args.total_spaces++; 420251788b1bSDan Rosenberg slot_count--; 42031406e432SJosef Bacik } 420451788b1bSDan Rosenberg if (!slot_count) 420551788b1bSDan Rosenberg break; 4206bf5fc093SJosef Bacik } 4207bf5fc093SJosef Bacik up_read(&info->groups_sem); 4208bf5fc093SJosef Bacik } 42091406e432SJosef Bacik 421036523e95SDavid Sterba /* 421136523e95SDavid Sterba * Add global block reserve 421236523e95SDavid Sterba */ 421336523e95SDavid Sterba if (slot_count) { 42140b246afaSJeff Mahoney struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv; 421536523e95SDavid Sterba 421636523e95SDavid Sterba spin_lock(&block_rsv->lock); 421736523e95SDavid Sterba space.total_bytes = block_rsv->size; 421836523e95SDavid Sterba space.used_bytes = block_rsv->size - block_rsv->reserved; 421936523e95SDavid Sterba spin_unlock(&block_rsv->lock); 422036523e95SDavid Sterba space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV; 422136523e95SDavid Sterba memcpy(dest, &space, sizeof(space)); 422236523e95SDavid Sterba space_args.total_spaces++; 422336523e95SDavid Sterba } 422436523e95SDavid Sterba 42252eec6c81SDaniel J Blueman user_dest = (struct btrfs_ioctl_space_info __user *) 42267fde62bfSChris Mason (arg + sizeof(struct btrfs_ioctl_space_args)); 42277fde62bfSChris Mason 42287fde62bfSChris Mason if (copy_to_user(user_dest, dest_orig, alloc_size)) 42297fde62bfSChris Mason ret = -EFAULT; 42307fde62bfSChris Mason 42317fde62bfSChris Mason kfree(dest_orig); 42327fde62bfSChris Mason out: 42337fde62bfSChris Mason if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) 42341406e432SJosef Bacik ret = -EFAULT; 42351406e432SJosef Bacik 42361406e432SJosef Bacik return ret; 42371406e432SJosef Bacik } 42381406e432SJosef Bacik 4239f46b5a66SChristoph Hellwig /* 4240f46b5a66SChristoph Hellwig * there are many ways the trans_start and trans_end ioctls can lead 4241f46b5a66SChristoph Hellwig * to deadlocks. They should only be used by applications that 4242f46b5a66SChristoph Hellwig * basically own the machine, and have a very in depth understanding 4243f46b5a66SChristoph Hellwig * of all the possible deadlocks and enospc problems. 4244f46b5a66SChristoph Hellwig */ 4245f46b5a66SChristoph Hellwig long btrfs_ioctl_trans_end(struct file *file) 4246f46b5a66SChristoph Hellwig { 4247496ad9aaSAl Viro struct inode *inode = file_inode(file); 4248f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 4249f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 4250f46b5a66SChristoph Hellwig 4251f46b5a66SChristoph Hellwig trans = file->private_data; 42521ab86aedSSage Weil if (!trans) 42531ab86aedSSage Weil return -EINVAL; 4254b214107eSChristoph Hellwig file->private_data = NULL; 42559ca9ee09SSage Weil 42563a45bb20SJeff Mahoney btrfs_end_transaction(trans); 42571ab86aedSSage Weil 4258a4abeea4SJosef Bacik atomic_dec(&root->fs_info->open_ioctl_trans); 42599ca9ee09SSage Weil 42602a79f17eSAl Viro mnt_drop_write_file(file); 42611ab86aedSSage Weil return 0; 4262f46b5a66SChristoph Hellwig } 4263f46b5a66SChristoph Hellwig 42649a8c28beSMiao Xie static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, 42659a8c28beSMiao Xie void __user *argp) 426646204592SSage Weil { 426746204592SSage Weil struct btrfs_trans_handle *trans; 426846204592SSage Weil u64 transid; 4269db5b493aSTsutomu Itoh int ret; 427046204592SSage Weil 4271d4edf39bSMiao Xie trans = btrfs_attach_transaction_barrier(root); 4272ff7c1d33SMiao Xie if (IS_ERR(trans)) { 4273ff7c1d33SMiao Xie if (PTR_ERR(trans) != -ENOENT) 427498d5dc13STsutomu Itoh return PTR_ERR(trans); 4275ff7c1d33SMiao Xie 4276ff7c1d33SMiao Xie /* No running transaction, don't bother */ 4277ff7c1d33SMiao Xie transid = root->fs_info->last_trans_committed; 4278ff7c1d33SMiao Xie goto out; 4279ff7c1d33SMiao Xie } 428046204592SSage Weil transid = trans->transid; 42813a45bb20SJeff Mahoney ret = btrfs_commit_transaction_async(trans, 0); 42828b2b2d3cSTsutomu Itoh if (ret) { 42833a45bb20SJeff Mahoney btrfs_end_transaction(trans); 4284db5b493aSTsutomu Itoh return ret; 42858b2b2d3cSTsutomu Itoh } 4286ff7c1d33SMiao Xie out: 428746204592SSage Weil if (argp) 428846204592SSage Weil if (copy_to_user(argp, &transid, sizeof(transid))) 428946204592SSage Weil return -EFAULT; 429046204592SSage Weil return 0; 429146204592SSage Weil } 429246204592SSage Weil 42932ff7e61eSJeff Mahoney static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info, 42949a8c28beSMiao Xie void __user *argp) 429546204592SSage Weil { 429646204592SSage Weil u64 transid; 429746204592SSage Weil 429846204592SSage Weil if (argp) { 429946204592SSage Weil if (copy_from_user(&transid, argp, sizeof(transid))) 430046204592SSage Weil return -EFAULT; 430146204592SSage Weil } else { 430246204592SSage Weil transid = 0; /* current trans */ 430346204592SSage Weil } 43042ff7e61eSJeff Mahoney return btrfs_wait_for_commit(fs_info, transid); 430546204592SSage Weil } 430646204592SSage Weil 4307b8e95489SMiao Xie static long btrfs_ioctl_scrub(struct file *file, void __user *arg) 4308475f6387SJan Schmidt { 43090b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb); 4310475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4311b8e95489SMiao Xie int ret; 4312475f6387SJan Schmidt 4313475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4314475f6387SJan Schmidt return -EPERM; 4315475f6387SJan Schmidt 4316475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4317475f6387SJan Schmidt if (IS_ERR(sa)) 4318475f6387SJan Schmidt return PTR_ERR(sa); 4319475f6387SJan Schmidt 4320b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) { 4321b8e95489SMiao Xie ret = mnt_want_write_file(file); 4322b8e95489SMiao Xie if (ret) 4323b8e95489SMiao Xie goto out; 4324b8e95489SMiao Xie } 4325b8e95489SMiao Xie 43260b246afaSJeff Mahoney ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end, 432763a212abSStefan Behrens &sa->progress, sa->flags & BTRFS_SCRUB_READONLY, 432863a212abSStefan Behrens 0); 4329475f6387SJan Schmidt 4330475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 4331475f6387SJan Schmidt ret = -EFAULT; 4332475f6387SJan Schmidt 4333b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) 4334b8e95489SMiao Xie mnt_drop_write_file(file); 4335b8e95489SMiao Xie out: 4336475f6387SJan Schmidt kfree(sa); 4337475f6387SJan Schmidt return ret; 4338475f6387SJan Schmidt } 4339475f6387SJan Schmidt 43402ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info) 4341475f6387SJan Schmidt { 4342475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4343475f6387SJan Schmidt return -EPERM; 4344475f6387SJan Schmidt 43452ff7e61eSJeff Mahoney return btrfs_scrub_cancel(fs_info); 4346475f6387SJan Schmidt } 4347475f6387SJan Schmidt 43482ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info, 4349475f6387SJan Schmidt void __user *arg) 4350475f6387SJan Schmidt { 4351475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4352475f6387SJan Schmidt int ret; 4353475f6387SJan Schmidt 4354475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4355475f6387SJan Schmidt return -EPERM; 4356475f6387SJan Schmidt 4357475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4358475f6387SJan Schmidt if (IS_ERR(sa)) 4359475f6387SJan Schmidt return PTR_ERR(sa); 4360475f6387SJan Schmidt 43612ff7e61eSJeff Mahoney ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress); 4362475f6387SJan Schmidt 4363475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 4364475f6387SJan Schmidt ret = -EFAULT; 4365475f6387SJan Schmidt 4366475f6387SJan Schmidt kfree(sa); 4367475f6387SJan Schmidt return ret; 4368475f6387SJan Schmidt } 4369475f6387SJan Schmidt 43702ff7e61eSJeff Mahoney static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info, 4371b27f7c0cSDavid Sterba void __user *arg) 4372c11d2c23SStefan Behrens { 4373c11d2c23SStefan Behrens struct btrfs_ioctl_get_dev_stats *sa; 4374c11d2c23SStefan Behrens int ret; 4375c11d2c23SStefan Behrens 4376c11d2c23SStefan Behrens sa = memdup_user(arg, sizeof(*sa)); 4377c11d2c23SStefan Behrens if (IS_ERR(sa)) 4378c11d2c23SStefan Behrens return PTR_ERR(sa); 4379c11d2c23SStefan Behrens 4380b27f7c0cSDavid Sterba if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) { 4381b27f7c0cSDavid Sterba kfree(sa); 4382b27f7c0cSDavid Sterba return -EPERM; 4383b27f7c0cSDavid Sterba } 4384b27f7c0cSDavid Sterba 43852ff7e61eSJeff Mahoney ret = btrfs_get_dev_stats(fs_info, sa); 4386c11d2c23SStefan Behrens 4387c11d2c23SStefan Behrens if (copy_to_user(arg, sa, sizeof(*sa))) 4388c11d2c23SStefan Behrens ret = -EFAULT; 4389c11d2c23SStefan Behrens 4390c11d2c23SStefan Behrens kfree(sa); 4391c11d2c23SStefan Behrens return ret; 4392c11d2c23SStefan Behrens } 4393c11d2c23SStefan Behrens 43942ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info, 43952ff7e61eSJeff Mahoney void __user *arg) 43963f6bcfbdSStefan Behrens { 43973f6bcfbdSStefan Behrens struct btrfs_ioctl_dev_replace_args *p; 43983f6bcfbdSStefan Behrens int ret; 43993f6bcfbdSStefan Behrens 44003f6bcfbdSStefan Behrens if (!capable(CAP_SYS_ADMIN)) 44013f6bcfbdSStefan Behrens return -EPERM; 44023f6bcfbdSStefan Behrens 44033f6bcfbdSStefan Behrens p = memdup_user(arg, sizeof(*p)); 44043f6bcfbdSStefan Behrens if (IS_ERR(p)) 44053f6bcfbdSStefan Behrens return PTR_ERR(p); 44063f6bcfbdSStefan Behrens 44073f6bcfbdSStefan Behrens switch (p->cmd) { 44083f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_START: 44090b246afaSJeff Mahoney if (fs_info->sb->s_flags & MS_RDONLY) { 4410adfa97cbSIlya Dryomov ret = -EROFS; 4411adfa97cbSIlya Dryomov goto out; 4412adfa97cbSIlya Dryomov } 4413171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 4414e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 44153f6bcfbdSStefan Behrens } else { 44162ff7e61eSJeff Mahoney ret = btrfs_dev_replace_by_ioctl(fs_info, p); 4417171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 44183f6bcfbdSStefan Behrens } 44193f6bcfbdSStefan Behrens break; 44203f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS: 44210b246afaSJeff Mahoney btrfs_dev_replace_status(fs_info, p); 44223f6bcfbdSStefan Behrens ret = 0; 44233f6bcfbdSStefan Behrens break; 44243f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL: 44250b246afaSJeff Mahoney ret = btrfs_dev_replace_cancel(fs_info, p); 44263f6bcfbdSStefan Behrens break; 44273f6bcfbdSStefan Behrens default: 44283f6bcfbdSStefan Behrens ret = -EINVAL; 44293f6bcfbdSStefan Behrens break; 44303f6bcfbdSStefan Behrens } 44313f6bcfbdSStefan Behrens 44323f6bcfbdSStefan Behrens if (copy_to_user(arg, p, sizeof(*p))) 44333f6bcfbdSStefan Behrens ret = -EFAULT; 4434adfa97cbSIlya Dryomov out: 44353f6bcfbdSStefan Behrens kfree(p); 44363f6bcfbdSStefan Behrens return ret; 44373f6bcfbdSStefan Behrens } 44383f6bcfbdSStefan Behrens 4439d7728c96SJan Schmidt static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) 4440d7728c96SJan Schmidt { 4441d7728c96SJan Schmidt int ret = 0; 4442d7728c96SJan Schmidt int i; 4443740c3d22SChris Mason u64 rel_ptr; 4444d7728c96SJan Schmidt int size; 4445806468f8SChris Mason struct btrfs_ioctl_ino_path_args *ipa = NULL; 4446d7728c96SJan Schmidt struct inode_fs_paths *ipath = NULL; 4447d7728c96SJan Schmidt struct btrfs_path *path; 4448d7728c96SJan Schmidt 444982b22ac8SKusanagi Kouichi if (!capable(CAP_DAC_READ_SEARCH)) 4450d7728c96SJan Schmidt return -EPERM; 4451d7728c96SJan Schmidt 4452d7728c96SJan Schmidt path = btrfs_alloc_path(); 4453d7728c96SJan Schmidt if (!path) { 4454d7728c96SJan Schmidt ret = -ENOMEM; 4455d7728c96SJan Schmidt goto out; 4456d7728c96SJan Schmidt } 4457d7728c96SJan Schmidt 4458d7728c96SJan Schmidt ipa = memdup_user(arg, sizeof(*ipa)); 4459d7728c96SJan Schmidt if (IS_ERR(ipa)) { 4460d7728c96SJan Schmidt ret = PTR_ERR(ipa); 4461d7728c96SJan Schmidt ipa = NULL; 4462d7728c96SJan Schmidt goto out; 4463d7728c96SJan Schmidt } 4464d7728c96SJan Schmidt 4465d7728c96SJan Schmidt size = min_t(u32, ipa->size, 4096); 4466d7728c96SJan Schmidt ipath = init_ipath(size, root, path); 4467d7728c96SJan Schmidt if (IS_ERR(ipath)) { 4468d7728c96SJan Schmidt ret = PTR_ERR(ipath); 4469d7728c96SJan Schmidt ipath = NULL; 4470d7728c96SJan Schmidt goto out; 4471d7728c96SJan Schmidt } 4472d7728c96SJan Schmidt 4473d7728c96SJan Schmidt ret = paths_from_inode(ipa->inum, ipath); 4474d7728c96SJan Schmidt if (ret < 0) 4475d7728c96SJan Schmidt goto out; 4476d7728c96SJan Schmidt 4477d7728c96SJan Schmidt for (i = 0; i < ipath->fspath->elem_cnt; ++i) { 4478745c4d8eSJeff Mahoney rel_ptr = ipath->fspath->val[i] - 4479745c4d8eSJeff Mahoney (u64)(unsigned long)ipath->fspath->val; 4480740c3d22SChris Mason ipath->fspath->val[i] = rel_ptr; 4481d7728c96SJan Schmidt } 4482d7728c96SJan Schmidt 4483745c4d8eSJeff Mahoney ret = copy_to_user((void *)(unsigned long)ipa->fspath, 4484745c4d8eSJeff Mahoney (void *)(unsigned long)ipath->fspath, size); 4485d7728c96SJan Schmidt if (ret) { 4486d7728c96SJan Schmidt ret = -EFAULT; 4487d7728c96SJan Schmidt goto out; 4488d7728c96SJan Schmidt } 4489d7728c96SJan Schmidt 4490d7728c96SJan Schmidt out: 4491d7728c96SJan Schmidt btrfs_free_path(path); 4492d7728c96SJan Schmidt free_ipath(ipath); 4493d7728c96SJan Schmidt kfree(ipa); 4494d7728c96SJan Schmidt 4495d7728c96SJan Schmidt return ret; 4496d7728c96SJan Schmidt } 4497d7728c96SJan Schmidt 4498d7728c96SJan Schmidt static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx) 4499d7728c96SJan Schmidt { 4500d7728c96SJan Schmidt struct btrfs_data_container *inodes = ctx; 4501d7728c96SJan Schmidt const size_t c = 3 * sizeof(u64); 4502d7728c96SJan Schmidt 4503d7728c96SJan Schmidt if (inodes->bytes_left >= c) { 4504d7728c96SJan Schmidt inodes->bytes_left -= c; 4505d7728c96SJan Schmidt inodes->val[inodes->elem_cnt] = inum; 4506d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 1] = offset; 4507d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 2] = root; 4508d7728c96SJan Schmidt inodes->elem_cnt += 3; 4509d7728c96SJan Schmidt } else { 4510d7728c96SJan Schmidt inodes->bytes_missing += c - inodes->bytes_left; 4511d7728c96SJan Schmidt inodes->bytes_left = 0; 4512d7728c96SJan Schmidt inodes->elem_missed += 3; 4513d7728c96SJan Schmidt } 4514d7728c96SJan Schmidt 4515d7728c96SJan Schmidt return 0; 4516d7728c96SJan Schmidt } 4517d7728c96SJan Schmidt 45182ff7e61eSJeff Mahoney static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info, 4519d7728c96SJan Schmidt void __user *arg) 4520d7728c96SJan Schmidt { 4521d7728c96SJan Schmidt int ret = 0; 4522d7728c96SJan Schmidt int size; 4523d7728c96SJan Schmidt struct btrfs_ioctl_logical_ino_args *loi; 4524d7728c96SJan Schmidt struct btrfs_data_container *inodes = NULL; 4525d7728c96SJan Schmidt struct btrfs_path *path = NULL; 4526d7728c96SJan Schmidt 4527d7728c96SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4528d7728c96SJan Schmidt return -EPERM; 4529d7728c96SJan Schmidt 4530d7728c96SJan Schmidt loi = memdup_user(arg, sizeof(*loi)); 45317b9ea627SShailendra Verma if (IS_ERR(loi)) 45327b9ea627SShailendra Verma return PTR_ERR(loi); 4533d7728c96SJan Schmidt 4534d7728c96SJan Schmidt path = btrfs_alloc_path(); 4535d7728c96SJan Schmidt if (!path) { 4536d7728c96SJan Schmidt ret = -ENOMEM; 4537d7728c96SJan Schmidt goto out; 4538d7728c96SJan Schmidt } 4539d7728c96SJan Schmidt 4540ee22184bSByongho Lee size = min_t(u32, loi->size, SZ_64K); 4541d7728c96SJan Schmidt inodes = init_data_container(size); 4542d7728c96SJan Schmidt if (IS_ERR(inodes)) { 4543d7728c96SJan Schmidt ret = PTR_ERR(inodes); 4544d7728c96SJan Schmidt inodes = NULL; 4545d7728c96SJan Schmidt goto out; 4546d7728c96SJan Schmidt } 4547d7728c96SJan Schmidt 45482ff7e61eSJeff Mahoney ret = iterate_inodes_from_logical(loi->logical, fs_info, path, 4549df031f07SLiu Bo build_ino_list, inodes); 4550df031f07SLiu Bo if (ret == -EINVAL) 4551d7728c96SJan Schmidt ret = -ENOENT; 4552d7728c96SJan Schmidt if (ret < 0) 4553d7728c96SJan Schmidt goto out; 4554d7728c96SJan Schmidt 4555745c4d8eSJeff Mahoney ret = copy_to_user((void *)(unsigned long)loi->inodes, 4556745c4d8eSJeff Mahoney (void *)(unsigned long)inodes, size); 4557d7728c96SJan Schmidt if (ret) 4558d7728c96SJan Schmidt ret = -EFAULT; 4559d7728c96SJan Schmidt 4560d7728c96SJan Schmidt out: 4561d7728c96SJan Schmidt btrfs_free_path(path); 4562f54de068SDavid Sterba kvfree(inodes); 4563d7728c96SJan Schmidt kfree(loi); 4564d7728c96SJan Schmidt 4565d7728c96SJan Schmidt return ret; 4566d7728c96SJan Schmidt } 4567d7728c96SJan Schmidt 456819a39dceSIlya Dryomov void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock, 4569c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs) 4570c9e9f97bSIlya Dryomov { 4571c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl = fs_info->balance_ctl; 4572c9e9f97bSIlya Dryomov 4573c9e9f97bSIlya Dryomov bargs->flags = bctl->flags; 4574c9e9f97bSIlya Dryomov 4575837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_running)) 4576837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_RUNNING; 4577837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_pause_req)) 4578837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; 4579a7e99c69SIlya Dryomov if (atomic_read(&fs_info->balance_cancel_req)) 4580a7e99c69SIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; 4581837d5b6eSIlya Dryomov 4582c9e9f97bSIlya Dryomov memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); 4583c9e9f97bSIlya Dryomov memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); 4584c9e9f97bSIlya Dryomov memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); 458519a39dceSIlya Dryomov 458619a39dceSIlya Dryomov if (lock) { 458719a39dceSIlya Dryomov spin_lock(&fs_info->balance_lock); 458819a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 458919a39dceSIlya Dryomov spin_unlock(&fs_info->balance_lock); 459019a39dceSIlya Dryomov } else { 459119a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 459219a39dceSIlya Dryomov } 4593c9e9f97bSIlya Dryomov } 4594c9e9f97bSIlya Dryomov 45959ba1f6e4SLiu Bo static long btrfs_ioctl_balance(struct file *file, void __user *arg) 4596c9e9f97bSIlya Dryomov { 4597496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 4598c9e9f97bSIlya Dryomov struct btrfs_fs_info *fs_info = root->fs_info; 4599c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 4600c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl; 4601ed0fb78fSIlya Dryomov bool need_unlock; /* for mut. excl. ops lock */ 4602c9e9f97bSIlya Dryomov int ret; 4603c9e9f97bSIlya Dryomov 4604c9e9f97bSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4605c9e9f97bSIlya Dryomov return -EPERM; 4606c9e9f97bSIlya Dryomov 4607e54bfa31SLiu Bo ret = mnt_want_write_file(file); 46089ba1f6e4SLiu Bo if (ret) 46099ba1f6e4SLiu Bo return ret; 46109ba1f6e4SLiu Bo 4611ed0fb78fSIlya Dryomov again: 4612171938e5SDavid Sterba if (!test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 4613c9e9f97bSIlya Dryomov mutex_lock(&fs_info->volume_mutex); 4614c9e9f97bSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4615ed0fb78fSIlya Dryomov need_unlock = true; 4616ed0fb78fSIlya Dryomov goto locked; 4617ed0fb78fSIlya Dryomov } 4618ed0fb78fSIlya Dryomov 4619ed0fb78fSIlya Dryomov /* 462001327610SNicholas D Steeves * mut. excl. ops lock is locked. Three possibilities: 4621ed0fb78fSIlya Dryomov * (1) some other op is running 4622ed0fb78fSIlya Dryomov * (2) balance is running 4623ed0fb78fSIlya Dryomov * (3) balance is paused -- special case (think resume) 4624ed0fb78fSIlya Dryomov */ 4625ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4626ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4627ed0fb78fSIlya Dryomov /* this is either (2) or (3) */ 4628ed0fb78fSIlya Dryomov if (!atomic_read(&fs_info->balance_running)) { 4629ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4630ed0fb78fSIlya Dryomov if (!mutex_trylock(&fs_info->volume_mutex)) 4631ed0fb78fSIlya Dryomov goto again; 4632ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4633ed0fb78fSIlya Dryomov 4634ed0fb78fSIlya Dryomov if (fs_info->balance_ctl && 4635ed0fb78fSIlya Dryomov !atomic_read(&fs_info->balance_running)) { 4636ed0fb78fSIlya Dryomov /* this is (3) */ 4637ed0fb78fSIlya Dryomov need_unlock = false; 4638ed0fb78fSIlya Dryomov goto locked; 4639ed0fb78fSIlya Dryomov } 4640ed0fb78fSIlya Dryomov 4641ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4642ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->volume_mutex); 4643ed0fb78fSIlya Dryomov goto again; 4644ed0fb78fSIlya Dryomov } else { 4645ed0fb78fSIlya Dryomov /* this is (2) */ 4646ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4647ed0fb78fSIlya Dryomov ret = -EINPROGRESS; 4648ed0fb78fSIlya Dryomov goto out; 4649ed0fb78fSIlya Dryomov } 4650ed0fb78fSIlya Dryomov } else { 4651ed0fb78fSIlya Dryomov /* this is (1) */ 4652ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4653e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 4654ed0fb78fSIlya Dryomov goto out; 4655ed0fb78fSIlya Dryomov } 4656ed0fb78fSIlya Dryomov 4657ed0fb78fSIlya Dryomov locked: 4658171938e5SDavid Sterba BUG_ON(!test_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)); 4659c9e9f97bSIlya Dryomov 4660c9e9f97bSIlya Dryomov if (arg) { 4661c9e9f97bSIlya Dryomov bargs = memdup_user(arg, sizeof(*bargs)); 4662c9e9f97bSIlya Dryomov if (IS_ERR(bargs)) { 4663c9e9f97bSIlya Dryomov ret = PTR_ERR(bargs); 4664ed0fb78fSIlya Dryomov goto out_unlock; 4665c9e9f97bSIlya Dryomov } 4666de322263SIlya Dryomov 4667de322263SIlya Dryomov if (bargs->flags & BTRFS_BALANCE_RESUME) { 4668de322263SIlya Dryomov if (!fs_info->balance_ctl) { 4669de322263SIlya Dryomov ret = -ENOTCONN; 4670de322263SIlya Dryomov goto out_bargs; 4671de322263SIlya Dryomov } 4672de322263SIlya Dryomov 4673de322263SIlya Dryomov bctl = fs_info->balance_ctl; 4674de322263SIlya Dryomov spin_lock(&fs_info->balance_lock); 4675de322263SIlya Dryomov bctl->flags |= BTRFS_BALANCE_RESUME; 4676de322263SIlya Dryomov spin_unlock(&fs_info->balance_lock); 4677de322263SIlya Dryomov 4678de322263SIlya Dryomov goto do_balance; 4679de322263SIlya Dryomov } 4680c9e9f97bSIlya Dryomov } else { 4681c9e9f97bSIlya Dryomov bargs = NULL; 4682c9e9f97bSIlya Dryomov } 4683c9e9f97bSIlya Dryomov 4684ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4685837d5b6eSIlya Dryomov ret = -EINPROGRESS; 4686837d5b6eSIlya Dryomov goto out_bargs; 4687837d5b6eSIlya Dryomov } 4688837d5b6eSIlya Dryomov 46898d2db785SDavid Sterba bctl = kzalloc(sizeof(*bctl), GFP_KERNEL); 4690c9e9f97bSIlya Dryomov if (!bctl) { 4691c9e9f97bSIlya Dryomov ret = -ENOMEM; 4692c9e9f97bSIlya Dryomov goto out_bargs; 4693c9e9f97bSIlya Dryomov } 4694c9e9f97bSIlya Dryomov 4695c9e9f97bSIlya Dryomov bctl->fs_info = fs_info; 4696c9e9f97bSIlya Dryomov if (arg) { 4697c9e9f97bSIlya Dryomov memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); 4698c9e9f97bSIlya Dryomov memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); 4699c9e9f97bSIlya Dryomov memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); 4700c9e9f97bSIlya Dryomov 4701c9e9f97bSIlya Dryomov bctl->flags = bargs->flags; 4702f43ffb60SIlya Dryomov } else { 4703f43ffb60SIlya Dryomov /* balance everything - no filters */ 4704f43ffb60SIlya Dryomov bctl->flags |= BTRFS_BALANCE_TYPE_MASK; 4705c9e9f97bSIlya Dryomov } 4706c9e9f97bSIlya Dryomov 47078eb93459SDavid Sterba if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) { 47088eb93459SDavid Sterba ret = -EINVAL; 47090f89abf5SChristian Engelmayer goto out_bctl; 47108eb93459SDavid Sterba } 47118eb93459SDavid Sterba 4712de322263SIlya Dryomov do_balance: 4713c9e9f97bSIlya Dryomov /* 4714171938e5SDavid Sterba * Ownership of bctl and filesystem flag BTRFS_FS_EXCL_OP 4715ed0fb78fSIlya Dryomov * goes to to btrfs_balance. bctl is freed in __cancel_balance, 4716ed0fb78fSIlya Dryomov * or, if restriper was paused all the way until unmount, in 4717171938e5SDavid Sterba * free_fs_info. The flag is cleared in __cancel_balance. 4718c9e9f97bSIlya Dryomov */ 4719ed0fb78fSIlya Dryomov need_unlock = false; 4720ed0fb78fSIlya Dryomov 4721ed0fb78fSIlya Dryomov ret = btrfs_balance(bctl, bargs); 47220f89abf5SChristian Engelmayer bctl = NULL; 4723ed0fb78fSIlya Dryomov 4724c9e9f97bSIlya Dryomov if (arg) { 4725c9e9f97bSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 4726c9e9f97bSIlya Dryomov ret = -EFAULT; 4727c9e9f97bSIlya Dryomov } 4728c9e9f97bSIlya Dryomov 47290f89abf5SChristian Engelmayer out_bctl: 47300f89abf5SChristian Engelmayer kfree(bctl); 4731c9e9f97bSIlya Dryomov out_bargs: 4732c9e9f97bSIlya Dryomov kfree(bargs); 4733ed0fb78fSIlya Dryomov out_unlock: 4734c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4735c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->volume_mutex); 4736ed0fb78fSIlya Dryomov if (need_unlock) 4737171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 4738ed0fb78fSIlya Dryomov out: 4739e54bfa31SLiu Bo mnt_drop_write_file(file); 4740c9e9f97bSIlya Dryomov return ret; 4741c9e9f97bSIlya Dryomov } 4742c9e9f97bSIlya Dryomov 47432ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd) 4744837d5b6eSIlya Dryomov { 4745837d5b6eSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4746837d5b6eSIlya Dryomov return -EPERM; 4747837d5b6eSIlya Dryomov 4748837d5b6eSIlya Dryomov switch (cmd) { 4749837d5b6eSIlya Dryomov case BTRFS_BALANCE_CTL_PAUSE: 47500b246afaSJeff Mahoney return btrfs_pause_balance(fs_info); 4751a7e99c69SIlya Dryomov case BTRFS_BALANCE_CTL_CANCEL: 47520b246afaSJeff Mahoney return btrfs_cancel_balance(fs_info); 4753837d5b6eSIlya Dryomov } 4754837d5b6eSIlya Dryomov 4755837d5b6eSIlya Dryomov return -EINVAL; 4756837d5b6eSIlya Dryomov } 4757837d5b6eSIlya Dryomov 47582ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info, 475919a39dceSIlya Dryomov void __user *arg) 476019a39dceSIlya Dryomov { 476119a39dceSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 476219a39dceSIlya Dryomov int ret = 0; 476319a39dceSIlya Dryomov 476419a39dceSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 476519a39dceSIlya Dryomov return -EPERM; 476619a39dceSIlya Dryomov 476719a39dceSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 476819a39dceSIlya Dryomov if (!fs_info->balance_ctl) { 476919a39dceSIlya Dryomov ret = -ENOTCONN; 477019a39dceSIlya Dryomov goto out; 477119a39dceSIlya Dryomov } 477219a39dceSIlya Dryomov 47738d2db785SDavid Sterba bargs = kzalloc(sizeof(*bargs), GFP_KERNEL); 477419a39dceSIlya Dryomov if (!bargs) { 477519a39dceSIlya Dryomov ret = -ENOMEM; 477619a39dceSIlya Dryomov goto out; 477719a39dceSIlya Dryomov } 477819a39dceSIlya Dryomov 477919a39dceSIlya Dryomov update_ioctl_balance_args(fs_info, 1, bargs); 478019a39dceSIlya Dryomov 478119a39dceSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 478219a39dceSIlya Dryomov ret = -EFAULT; 478319a39dceSIlya Dryomov 478419a39dceSIlya Dryomov kfree(bargs); 478519a39dceSIlya Dryomov out: 478619a39dceSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 478719a39dceSIlya Dryomov return ret; 478819a39dceSIlya Dryomov } 478919a39dceSIlya Dryomov 4790905b0ddaSMiao Xie static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) 47915d13a37bSArne Jansen { 47920b246afaSJeff Mahoney struct inode *inode = file_inode(file); 47930b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 47945d13a37bSArne Jansen struct btrfs_ioctl_quota_ctl_args *sa; 47955d13a37bSArne Jansen struct btrfs_trans_handle *trans = NULL; 47965d13a37bSArne Jansen int ret; 47975d13a37bSArne Jansen int err; 47985d13a37bSArne Jansen 47995d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 48005d13a37bSArne Jansen return -EPERM; 48015d13a37bSArne Jansen 4802905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4803905b0ddaSMiao Xie if (ret) 4804905b0ddaSMiao Xie return ret; 48055d13a37bSArne Jansen 48065d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4807905b0ddaSMiao Xie if (IS_ERR(sa)) { 4808905b0ddaSMiao Xie ret = PTR_ERR(sa); 4809905b0ddaSMiao Xie goto drop_write; 4810905b0ddaSMiao Xie } 48115d13a37bSArne Jansen 48120b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 48130b246afaSJeff Mahoney trans = btrfs_start_transaction(fs_info->tree_root, 2); 48145d13a37bSArne Jansen if (IS_ERR(trans)) { 48155d13a37bSArne Jansen ret = PTR_ERR(trans); 48165d13a37bSArne Jansen goto out; 48175d13a37bSArne Jansen } 48185d13a37bSArne Jansen 48195d13a37bSArne Jansen switch (sa->cmd) { 48205d13a37bSArne Jansen case BTRFS_QUOTA_CTL_ENABLE: 48210b246afaSJeff Mahoney ret = btrfs_quota_enable(trans, fs_info); 48225d13a37bSArne Jansen break; 48235d13a37bSArne Jansen case BTRFS_QUOTA_CTL_DISABLE: 48240b246afaSJeff Mahoney ret = btrfs_quota_disable(trans, fs_info); 48255d13a37bSArne Jansen break; 48265d13a37bSArne Jansen default: 48275d13a37bSArne Jansen ret = -EINVAL; 48285d13a37bSArne Jansen break; 48295d13a37bSArne Jansen } 48305d13a37bSArne Jansen 48313a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 48325d13a37bSArne Jansen if (err && !ret) 48335d13a37bSArne Jansen ret = err; 48345d13a37bSArne Jansen out: 48355d13a37bSArne Jansen kfree(sa); 48360b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 4837905b0ddaSMiao Xie drop_write: 4838905b0ddaSMiao Xie mnt_drop_write_file(file); 48395d13a37bSArne Jansen return ret; 48405d13a37bSArne Jansen } 48415d13a37bSArne Jansen 4842905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg) 48435d13a37bSArne Jansen { 48440b246afaSJeff Mahoney struct inode *inode = file_inode(file); 48450b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 48460b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 48475d13a37bSArne Jansen struct btrfs_ioctl_qgroup_assign_args *sa; 48485d13a37bSArne Jansen struct btrfs_trans_handle *trans; 48495d13a37bSArne Jansen int ret; 48505d13a37bSArne Jansen int err; 48515d13a37bSArne Jansen 48525d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 48535d13a37bSArne Jansen return -EPERM; 48545d13a37bSArne Jansen 4855905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4856905b0ddaSMiao Xie if (ret) 4857905b0ddaSMiao Xie return ret; 48585d13a37bSArne Jansen 48595d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4860905b0ddaSMiao Xie if (IS_ERR(sa)) { 4861905b0ddaSMiao Xie ret = PTR_ERR(sa); 4862905b0ddaSMiao Xie goto drop_write; 4863905b0ddaSMiao Xie } 48645d13a37bSArne Jansen 48655d13a37bSArne Jansen trans = btrfs_join_transaction(root); 48665d13a37bSArne Jansen if (IS_ERR(trans)) { 48675d13a37bSArne Jansen ret = PTR_ERR(trans); 48685d13a37bSArne Jansen goto out; 48695d13a37bSArne Jansen } 48705d13a37bSArne Jansen 48715d13a37bSArne Jansen if (sa->assign) { 48720b246afaSJeff Mahoney ret = btrfs_add_qgroup_relation(trans, fs_info, 48735d13a37bSArne Jansen sa->src, sa->dst); 48745d13a37bSArne Jansen } else { 48750b246afaSJeff Mahoney ret = btrfs_del_qgroup_relation(trans, fs_info, 48765d13a37bSArne Jansen sa->src, sa->dst); 48775d13a37bSArne Jansen } 48785d13a37bSArne Jansen 4879e082f563SQu Wenruo /* update qgroup status and info */ 48800b246afaSJeff Mahoney err = btrfs_run_qgroups(trans, fs_info); 4881e082f563SQu Wenruo if (err < 0) 48820b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, err, 4883ad8403dfSAnand Jain "failed to update qgroup status and info"); 48843a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 48855d13a37bSArne Jansen if (err && !ret) 48865d13a37bSArne Jansen ret = err; 48875d13a37bSArne Jansen 48885d13a37bSArne Jansen out: 48895d13a37bSArne Jansen kfree(sa); 4890905b0ddaSMiao Xie drop_write: 4891905b0ddaSMiao Xie mnt_drop_write_file(file); 48925d13a37bSArne Jansen return ret; 48935d13a37bSArne Jansen } 48945d13a37bSArne Jansen 4895905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg) 48965d13a37bSArne Jansen { 48970b246afaSJeff Mahoney struct inode *inode = file_inode(file); 48980b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 48990b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 49005d13a37bSArne Jansen struct btrfs_ioctl_qgroup_create_args *sa; 49015d13a37bSArne Jansen struct btrfs_trans_handle *trans; 49025d13a37bSArne Jansen int ret; 49035d13a37bSArne Jansen int err; 49045d13a37bSArne Jansen 49055d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 49065d13a37bSArne Jansen return -EPERM; 49075d13a37bSArne Jansen 4908905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4909905b0ddaSMiao Xie if (ret) 4910905b0ddaSMiao Xie return ret; 49115d13a37bSArne Jansen 49125d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4913905b0ddaSMiao Xie if (IS_ERR(sa)) { 4914905b0ddaSMiao Xie ret = PTR_ERR(sa); 4915905b0ddaSMiao Xie goto drop_write; 4916905b0ddaSMiao Xie } 49175d13a37bSArne Jansen 4918d86e56cfSMiao Xie if (!sa->qgroupid) { 4919d86e56cfSMiao Xie ret = -EINVAL; 4920d86e56cfSMiao Xie goto out; 4921d86e56cfSMiao Xie } 4922d86e56cfSMiao Xie 49235d13a37bSArne Jansen trans = btrfs_join_transaction(root); 49245d13a37bSArne Jansen if (IS_ERR(trans)) { 49255d13a37bSArne Jansen ret = PTR_ERR(trans); 49265d13a37bSArne Jansen goto out; 49275d13a37bSArne Jansen } 49285d13a37bSArne Jansen 49295d13a37bSArne Jansen if (sa->create) { 49300b246afaSJeff Mahoney ret = btrfs_create_qgroup(trans, fs_info, sa->qgroupid); 49315d13a37bSArne Jansen } else { 49320b246afaSJeff Mahoney ret = btrfs_remove_qgroup(trans, fs_info, sa->qgroupid); 49335d13a37bSArne Jansen } 49345d13a37bSArne Jansen 49353a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 49365d13a37bSArne Jansen if (err && !ret) 49375d13a37bSArne Jansen ret = err; 49385d13a37bSArne Jansen 49395d13a37bSArne Jansen out: 49405d13a37bSArne Jansen kfree(sa); 4941905b0ddaSMiao Xie drop_write: 4942905b0ddaSMiao Xie mnt_drop_write_file(file); 49435d13a37bSArne Jansen return ret; 49445d13a37bSArne Jansen } 49455d13a37bSArne Jansen 4946905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg) 49475d13a37bSArne Jansen { 49480b246afaSJeff Mahoney struct inode *inode = file_inode(file); 49490b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 49500b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 49515d13a37bSArne Jansen struct btrfs_ioctl_qgroup_limit_args *sa; 49525d13a37bSArne Jansen struct btrfs_trans_handle *trans; 49535d13a37bSArne Jansen int ret; 49545d13a37bSArne Jansen int err; 49555d13a37bSArne Jansen u64 qgroupid; 49565d13a37bSArne Jansen 49575d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 49585d13a37bSArne Jansen return -EPERM; 49595d13a37bSArne Jansen 4960905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4961905b0ddaSMiao Xie if (ret) 4962905b0ddaSMiao Xie return ret; 49635d13a37bSArne Jansen 49645d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4965905b0ddaSMiao Xie if (IS_ERR(sa)) { 4966905b0ddaSMiao Xie ret = PTR_ERR(sa); 4967905b0ddaSMiao Xie goto drop_write; 4968905b0ddaSMiao Xie } 49695d13a37bSArne Jansen 49705d13a37bSArne Jansen trans = btrfs_join_transaction(root); 49715d13a37bSArne Jansen if (IS_ERR(trans)) { 49725d13a37bSArne Jansen ret = PTR_ERR(trans); 49735d13a37bSArne Jansen goto out; 49745d13a37bSArne Jansen } 49755d13a37bSArne Jansen 49765d13a37bSArne Jansen qgroupid = sa->qgroupid; 49775d13a37bSArne Jansen if (!qgroupid) { 49785d13a37bSArne Jansen /* take the current subvol as qgroup */ 49795d13a37bSArne Jansen qgroupid = root->root_key.objectid; 49805d13a37bSArne Jansen } 49815d13a37bSArne Jansen 49820b246afaSJeff Mahoney ret = btrfs_limit_qgroup(trans, fs_info, qgroupid, &sa->lim); 49835d13a37bSArne Jansen 49843a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 49855d13a37bSArne Jansen if (err && !ret) 49865d13a37bSArne Jansen ret = err; 49875d13a37bSArne Jansen 49885d13a37bSArne Jansen out: 49895d13a37bSArne Jansen kfree(sa); 4990905b0ddaSMiao Xie drop_write: 4991905b0ddaSMiao Xie mnt_drop_write_file(file); 49925d13a37bSArne Jansen return ret; 49935d13a37bSArne Jansen } 49945d13a37bSArne Jansen 49952f232036SJan Schmidt static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) 49962f232036SJan Schmidt { 49970b246afaSJeff Mahoney struct inode *inode = file_inode(file); 49980b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 49992f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 50002f232036SJan Schmidt int ret; 50012f232036SJan Schmidt 50022f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 50032f232036SJan Schmidt return -EPERM; 50042f232036SJan Schmidt 50052f232036SJan Schmidt ret = mnt_want_write_file(file); 50062f232036SJan Schmidt if (ret) 50072f232036SJan Schmidt return ret; 50082f232036SJan Schmidt 50092f232036SJan Schmidt qsa = memdup_user(arg, sizeof(*qsa)); 50102f232036SJan Schmidt if (IS_ERR(qsa)) { 50112f232036SJan Schmidt ret = PTR_ERR(qsa); 50122f232036SJan Schmidt goto drop_write; 50132f232036SJan Schmidt } 50142f232036SJan Schmidt 50152f232036SJan Schmidt if (qsa->flags) { 50162f232036SJan Schmidt ret = -EINVAL; 50172f232036SJan Schmidt goto out; 50182f232036SJan Schmidt } 50192f232036SJan Schmidt 50200b246afaSJeff Mahoney ret = btrfs_qgroup_rescan(fs_info); 50212f232036SJan Schmidt 50222f232036SJan Schmidt out: 50232f232036SJan Schmidt kfree(qsa); 50242f232036SJan Schmidt drop_write: 50252f232036SJan Schmidt mnt_drop_write_file(file); 50262f232036SJan Schmidt return ret; 50272f232036SJan Schmidt } 50282f232036SJan Schmidt 50292f232036SJan Schmidt static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg) 50302f232036SJan Schmidt { 50310b246afaSJeff Mahoney struct inode *inode = file_inode(file); 50320b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 50332f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 50342f232036SJan Schmidt int ret = 0; 50352f232036SJan Schmidt 50362f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 50372f232036SJan Schmidt return -EPERM; 50382f232036SJan Schmidt 50398d2db785SDavid Sterba qsa = kzalloc(sizeof(*qsa), GFP_KERNEL); 50402f232036SJan Schmidt if (!qsa) 50412f232036SJan Schmidt return -ENOMEM; 50422f232036SJan Schmidt 50430b246afaSJeff Mahoney if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { 50442f232036SJan Schmidt qsa->flags = 1; 50450b246afaSJeff Mahoney qsa->progress = fs_info->qgroup_rescan_progress.objectid; 50462f232036SJan Schmidt } 50472f232036SJan Schmidt 50482f232036SJan Schmidt if (copy_to_user(arg, qsa, sizeof(*qsa))) 50492f232036SJan Schmidt ret = -EFAULT; 50502f232036SJan Schmidt 50512f232036SJan Schmidt kfree(qsa); 50522f232036SJan Schmidt return ret; 50532f232036SJan Schmidt } 50542f232036SJan Schmidt 505557254b6eSJan Schmidt static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg) 505657254b6eSJan Schmidt { 50570b246afaSJeff Mahoney struct inode *inode = file_inode(file); 50580b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 505957254b6eSJan Schmidt 506057254b6eSJan Schmidt if (!capable(CAP_SYS_ADMIN)) 506157254b6eSJan Schmidt return -EPERM; 506257254b6eSJan Schmidt 50630b246afaSJeff Mahoney return btrfs_qgroup_wait_for_completion(fs_info, true); 506457254b6eSJan Schmidt } 506557254b6eSJan Schmidt 5066abccd00fSHugo Mills static long _btrfs_ioctl_set_received_subvol(struct file *file, 5067abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa) 50688ea05e3aSAlexander Block { 5069496ad9aaSAl Viro struct inode *inode = file_inode(file); 50700b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 50718ea05e3aSAlexander Block struct btrfs_root *root = BTRFS_I(inode)->root; 50728ea05e3aSAlexander Block struct btrfs_root_item *root_item = &root->root_item; 50738ea05e3aSAlexander Block struct btrfs_trans_handle *trans; 5074c2050a45SDeepa Dinamani struct timespec ct = current_time(inode); 50758ea05e3aSAlexander Block int ret = 0; 5076dd5f9615SStefan Behrens int received_uuid_changed; 50778ea05e3aSAlexander Block 5078bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 5079bd60ea0fSDavid Sterba return -EPERM; 5080bd60ea0fSDavid Sterba 50818ea05e3aSAlexander Block ret = mnt_want_write_file(file); 50828ea05e3aSAlexander Block if (ret < 0) 50838ea05e3aSAlexander Block return ret; 50848ea05e3aSAlexander Block 50850b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 50868ea05e3aSAlexander Block 50874a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 50888ea05e3aSAlexander Block ret = -EINVAL; 50898ea05e3aSAlexander Block goto out; 50908ea05e3aSAlexander Block } 50918ea05e3aSAlexander Block 50928ea05e3aSAlexander Block if (btrfs_root_readonly(root)) { 50938ea05e3aSAlexander Block ret = -EROFS; 50948ea05e3aSAlexander Block goto out; 50958ea05e3aSAlexander Block } 50968ea05e3aSAlexander Block 5097dd5f9615SStefan Behrens /* 5098dd5f9615SStefan Behrens * 1 - root item 5099dd5f9615SStefan Behrens * 2 - uuid items (received uuid + subvol uuid) 5100dd5f9615SStefan Behrens */ 5101dd5f9615SStefan Behrens trans = btrfs_start_transaction(root, 3); 51028ea05e3aSAlexander Block if (IS_ERR(trans)) { 51038ea05e3aSAlexander Block ret = PTR_ERR(trans); 51048ea05e3aSAlexander Block trans = NULL; 51058ea05e3aSAlexander Block goto out; 51068ea05e3aSAlexander Block } 51078ea05e3aSAlexander Block 51088ea05e3aSAlexander Block sa->rtransid = trans->transid; 51098ea05e3aSAlexander Block sa->rtime.sec = ct.tv_sec; 51108ea05e3aSAlexander Block sa->rtime.nsec = ct.tv_nsec; 51118ea05e3aSAlexander Block 5112dd5f9615SStefan Behrens received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid, 5113dd5f9615SStefan Behrens BTRFS_UUID_SIZE); 5114dd5f9615SStefan Behrens if (received_uuid_changed && 5115dd5f9615SStefan Behrens !btrfs_is_empty_uuid(root_item->received_uuid)) 51160b246afaSJeff Mahoney btrfs_uuid_tree_rem(trans, fs_info, root_item->received_uuid, 5117dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5118dd5f9615SStefan Behrens root->root_key.objectid); 51198ea05e3aSAlexander Block memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE); 51208ea05e3aSAlexander Block btrfs_set_root_stransid(root_item, sa->stransid); 51218ea05e3aSAlexander Block btrfs_set_root_rtransid(root_item, sa->rtransid); 51223cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec); 51233cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec); 51243cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec); 51253cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec); 51268ea05e3aSAlexander Block 51270b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 51288ea05e3aSAlexander Block &root->root_key, &root->root_item); 51298ea05e3aSAlexander Block if (ret < 0) { 51303a45bb20SJeff Mahoney btrfs_end_transaction(trans); 51318ea05e3aSAlexander Block goto out; 5132dd5f9615SStefan Behrens } 5133dd5f9615SStefan Behrens if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { 51340b246afaSJeff Mahoney ret = btrfs_uuid_tree_add(trans, fs_info, sa->uuid, 5135dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5136dd5f9615SStefan Behrens root->root_key.objectid); 5137dd5f9615SStefan Behrens if (ret < 0 && ret != -EEXIST) { 513866642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 5139dd5f9615SStefan Behrens goto out; 5140dd5f9615SStefan Behrens } 5141dd5f9615SStefan Behrens } 51423a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 5143dd5f9615SStefan Behrens if (ret < 0) { 514466642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 51458ea05e3aSAlexander Block goto out; 51468ea05e3aSAlexander Block } 51478ea05e3aSAlexander Block 5148abccd00fSHugo Mills out: 51490b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 5150abccd00fSHugo Mills mnt_drop_write_file(file); 5151abccd00fSHugo Mills return ret; 5152abccd00fSHugo Mills } 5153abccd00fSHugo Mills 5154abccd00fSHugo Mills #ifdef CONFIG_64BIT 5155abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol_32(struct file *file, 5156abccd00fSHugo Mills void __user *arg) 5157abccd00fSHugo Mills { 5158abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL; 5159abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *args64 = NULL; 5160abccd00fSHugo Mills int ret = 0; 5161abccd00fSHugo Mills 5162abccd00fSHugo Mills args32 = memdup_user(arg, sizeof(*args32)); 51637b9ea627SShailendra Verma if (IS_ERR(args32)) 51647b9ea627SShailendra Verma return PTR_ERR(args32); 5165abccd00fSHugo Mills 51668d2db785SDavid Sterba args64 = kmalloc(sizeof(*args64), GFP_KERNEL); 516784dbeb87SDan Carpenter if (!args64) { 516884dbeb87SDan Carpenter ret = -ENOMEM; 5169abccd00fSHugo Mills goto out; 5170abccd00fSHugo Mills } 5171abccd00fSHugo Mills 5172abccd00fSHugo Mills memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE); 5173abccd00fSHugo Mills args64->stransid = args32->stransid; 5174abccd00fSHugo Mills args64->rtransid = args32->rtransid; 5175abccd00fSHugo Mills args64->stime.sec = args32->stime.sec; 5176abccd00fSHugo Mills args64->stime.nsec = args32->stime.nsec; 5177abccd00fSHugo Mills args64->rtime.sec = args32->rtime.sec; 5178abccd00fSHugo Mills args64->rtime.nsec = args32->rtime.nsec; 5179abccd00fSHugo Mills args64->flags = args32->flags; 5180abccd00fSHugo Mills 5181abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, args64); 5182abccd00fSHugo Mills if (ret) 5183abccd00fSHugo Mills goto out; 5184abccd00fSHugo Mills 5185abccd00fSHugo Mills memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE); 5186abccd00fSHugo Mills args32->stransid = args64->stransid; 5187abccd00fSHugo Mills args32->rtransid = args64->rtransid; 5188abccd00fSHugo Mills args32->stime.sec = args64->stime.sec; 5189abccd00fSHugo Mills args32->stime.nsec = args64->stime.nsec; 5190abccd00fSHugo Mills args32->rtime.sec = args64->rtime.sec; 5191abccd00fSHugo Mills args32->rtime.nsec = args64->rtime.nsec; 5192abccd00fSHugo Mills args32->flags = args64->flags; 5193abccd00fSHugo Mills 5194abccd00fSHugo Mills ret = copy_to_user(arg, args32, sizeof(*args32)); 5195abccd00fSHugo Mills if (ret) 5196abccd00fSHugo Mills ret = -EFAULT; 5197abccd00fSHugo Mills 5198abccd00fSHugo Mills out: 5199abccd00fSHugo Mills kfree(args32); 5200abccd00fSHugo Mills kfree(args64); 5201abccd00fSHugo Mills return ret; 5202abccd00fSHugo Mills } 5203abccd00fSHugo Mills #endif 5204abccd00fSHugo Mills 5205abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol(struct file *file, 5206abccd00fSHugo Mills void __user *arg) 5207abccd00fSHugo Mills { 5208abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa = NULL; 5209abccd00fSHugo Mills int ret = 0; 5210abccd00fSHugo Mills 5211abccd00fSHugo Mills sa = memdup_user(arg, sizeof(*sa)); 52127b9ea627SShailendra Verma if (IS_ERR(sa)) 52137b9ea627SShailendra Verma return PTR_ERR(sa); 5214abccd00fSHugo Mills 5215abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, sa); 5216abccd00fSHugo Mills 5217abccd00fSHugo Mills if (ret) 5218abccd00fSHugo Mills goto out; 5219abccd00fSHugo Mills 52208ea05e3aSAlexander Block ret = copy_to_user(arg, sa, sizeof(*sa)); 52218ea05e3aSAlexander Block if (ret) 52228ea05e3aSAlexander Block ret = -EFAULT; 52238ea05e3aSAlexander Block 52248ea05e3aSAlexander Block out: 52258ea05e3aSAlexander Block kfree(sa); 52268ea05e3aSAlexander Block return ret; 52278ea05e3aSAlexander Block } 52288ea05e3aSAlexander Block 5229867ab667Sjeff.liu static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg) 5230867ab667Sjeff.liu { 52310b246afaSJeff Mahoney struct inode *inode = file_inode(file); 52320b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 5233a1b83ac5SAnand Jain size_t len; 5234867ab667Sjeff.liu int ret; 5235a1b83ac5SAnand Jain char label[BTRFS_LABEL_SIZE]; 5236a1b83ac5SAnand Jain 52370b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 52380b246afaSJeff Mahoney memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE); 52390b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 5240a1b83ac5SAnand Jain 5241a1b83ac5SAnand Jain len = strnlen(label, BTRFS_LABEL_SIZE); 5242867ab667Sjeff.liu 5243867ab667Sjeff.liu if (len == BTRFS_LABEL_SIZE) { 52440b246afaSJeff Mahoney btrfs_warn(fs_info, 52450b246afaSJeff Mahoney "label is too long, return the first %zu bytes", 52460b246afaSJeff Mahoney --len); 5247867ab667Sjeff.liu } 5248867ab667Sjeff.liu 5249867ab667Sjeff.liu ret = copy_to_user(arg, label, len); 5250867ab667Sjeff.liu 5251867ab667Sjeff.liu return ret ? -EFAULT : 0; 5252867ab667Sjeff.liu } 5253867ab667Sjeff.liu 5254a8bfd4abSjeff.liu static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg) 5255a8bfd4abSjeff.liu { 52560b246afaSJeff Mahoney struct inode *inode = file_inode(file); 52570b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 52580b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 52590b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 5260a8bfd4abSjeff.liu struct btrfs_trans_handle *trans; 5261a8bfd4abSjeff.liu char label[BTRFS_LABEL_SIZE]; 5262a8bfd4abSjeff.liu int ret; 5263a8bfd4abSjeff.liu 5264a8bfd4abSjeff.liu if (!capable(CAP_SYS_ADMIN)) 5265a8bfd4abSjeff.liu return -EPERM; 5266a8bfd4abSjeff.liu 5267a8bfd4abSjeff.liu if (copy_from_user(label, arg, sizeof(label))) 5268a8bfd4abSjeff.liu return -EFAULT; 5269a8bfd4abSjeff.liu 5270a8bfd4abSjeff.liu if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) { 52710b246afaSJeff Mahoney btrfs_err(fs_info, 52725d163e0eSJeff Mahoney "unable to set label with more than %d bytes", 5273a8bfd4abSjeff.liu BTRFS_LABEL_SIZE - 1); 5274a8bfd4abSjeff.liu return -EINVAL; 5275a8bfd4abSjeff.liu } 5276a8bfd4abSjeff.liu 5277a8bfd4abSjeff.liu ret = mnt_want_write_file(file); 5278a8bfd4abSjeff.liu if (ret) 5279a8bfd4abSjeff.liu return ret; 5280a8bfd4abSjeff.liu 5281a8bfd4abSjeff.liu trans = btrfs_start_transaction(root, 0); 5282a8bfd4abSjeff.liu if (IS_ERR(trans)) { 5283a8bfd4abSjeff.liu ret = PTR_ERR(trans); 5284a8bfd4abSjeff.liu goto out_unlock; 5285a8bfd4abSjeff.liu } 5286a8bfd4abSjeff.liu 52870b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 5288a8bfd4abSjeff.liu strcpy(super_block->label, label); 52890b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 52903a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 5291a8bfd4abSjeff.liu 5292a8bfd4abSjeff.liu out_unlock: 5293a8bfd4abSjeff.liu mnt_drop_write_file(file); 5294a8bfd4abSjeff.liu return ret; 5295a8bfd4abSjeff.liu } 5296a8bfd4abSjeff.liu 52972eaa055fSJeff Mahoney #define INIT_FEATURE_FLAGS(suffix) \ 52982eaa055fSJeff Mahoney { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \ 52992eaa055fSJeff Mahoney .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \ 53002eaa055fSJeff Mahoney .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix } 53012eaa055fSJeff Mahoney 5302d5131b65SDavid Sterba int btrfs_ioctl_get_supported_features(void __user *arg) 53032eaa055fSJeff Mahoney { 53044d4ab6d6SDavid Sterba static const struct btrfs_ioctl_feature_flags features[3] = { 53052eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SUPP), 53062eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_SET), 53072eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_CLEAR) 53082eaa055fSJeff Mahoney }; 53092eaa055fSJeff Mahoney 53102eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 53112eaa055fSJeff Mahoney return -EFAULT; 53122eaa055fSJeff Mahoney 53132eaa055fSJeff Mahoney return 0; 53142eaa055fSJeff Mahoney } 53152eaa055fSJeff Mahoney 53162eaa055fSJeff Mahoney static int btrfs_ioctl_get_features(struct file *file, void __user *arg) 53172eaa055fSJeff Mahoney { 53180b246afaSJeff Mahoney struct inode *inode = file_inode(file); 53190b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 53200b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 53212eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags features; 53222eaa055fSJeff Mahoney 53232eaa055fSJeff Mahoney features.compat_flags = btrfs_super_compat_flags(super_block); 53242eaa055fSJeff Mahoney features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block); 53252eaa055fSJeff Mahoney features.incompat_flags = btrfs_super_incompat_flags(super_block); 53262eaa055fSJeff Mahoney 53272eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 53282eaa055fSJeff Mahoney return -EFAULT; 53292eaa055fSJeff Mahoney 53302eaa055fSJeff Mahoney return 0; 53312eaa055fSJeff Mahoney } 53322eaa055fSJeff Mahoney 53332ff7e61eSJeff Mahoney static int check_feature_bits(struct btrfs_fs_info *fs_info, 53343b02a68aSJeff Mahoney enum btrfs_feature_set set, 53352eaa055fSJeff Mahoney u64 change_mask, u64 flags, u64 supported_flags, 53362eaa055fSJeff Mahoney u64 safe_set, u64 safe_clear) 53372eaa055fSJeff Mahoney { 53383b02a68aSJeff Mahoney const char *type = btrfs_feature_set_names[set]; 53393b02a68aSJeff Mahoney char *names; 53402eaa055fSJeff Mahoney u64 disallowed, unsupported; 53412eaa055fSJeff Mahoney u64 set_mask = flags & change_mask; 53422eaa055fSJeff Mahoney u64 clear_mask = ~flags & change_mask; 53432eaa055fSJeff Mahoney 53442eaa055fSJeff Mahoney unsupported = set_mask & ~supported_flags; 53452eaa055fSJeff Mahoney if (unsupported) { 53463b02a68aSJeff Mahoney names = btrfs_printable_features(set, unsupported); 53473b02a68aSJeff Mahoney if (names) { 53480b246afaSJeff Mahoney btrfs_warn(fs_info, 53493b02a68aSJeff Mahoney "this kernel does not support the %s feature bit%s", 53503b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 53513b02a68aSJeff Mahoney kfree(names); 53523b02a68aSJeff Mahoney } else 53530b246afaSJeff Mahoney btrfs_warn(fs_info, 53542eaa055fSJeff Mahoney "this kernel does not support %s bits 0x%llx", 53552eaa055fSJeff Mahoney type, unsupported); 53562eaa055fSJeff Mahoney return -EOPNOTSUPP; 53572eaa055fSJeff Mahoney } 53582eaa055fSJeff Mahoney 53592eaa055fSJeff Mahoney disallowed = set_mask & ~safe_set; 53602eaa055fSJeff Mahoney if (disallowed) { 53613b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 53623b02a68aSJeff Mahoney if (names) { 53630b246afaSJeff Mahoney btrfs_warn(fs_info, 53643b02a68aSJeff Mahoney "can't set the %s feature bit%s while mounted", 53653b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 53663b02a68aSJeff Mahoney kfree(names); 53673b02a68aSJeff Mahoney } else 53680b246afaSJeff Mahoney btrfs_warn(fs_info, 53692eaa055fSJeff Mahoney "can't set %s bits 0x%llx while mounted", 53702eaa055fSJeff Mahoney type, disallowed); 53712eaa055fSJeff Mahoney return -EPERM; 53722eaa055fSJeff Mahoney } 53732eaa055fSJeff Mahoney 53742eaa055fSJeff Mahoney disallowed = clear_mask & ~safe_clear; 53752eaa055fSJeff Mahoney if (disallowed) { 53763b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 53773b02a68aSJeff Mahoney if (names) { 53780b246afaSJeff Mahoney btrfs_warn(fs_info, 53793b02a68aSJeff Mahoney "can't clear the %s feature bit%s while mounted", 53803b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 53813b02a68aSJeff Mahoney kfree(names); 53823b02a68aSJeff Mahoney } else 53830b246afaSJeff Mahoney btrfs_warn(fs_info, 53842eaa055fSJeff Mahoney "can't clear %s bits 0x%llx while mounted", 53852eaa055fSJeff Mahoney type, disallowed); 53862eaa055fSJeff Mahoney return -EPERM; 53872eaa055fSJeff Mahoney } 53882eaa055fSJeff Mahoney 53892eaa055fSJeff Mahoney return 0; 53902eaa055fSJeff Mahoney } 53912eaa055fSJeff Mahoney 53922ff7e61eSJeff Mahoney #define check_feature(fs_info, change_mask, flags, mask_base) \ 53932ff7e61eSJeff Mahoney check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \ 53942eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SUPP, \ 53952eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \ 53962eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR) 53972eaa055fSJeff Mahoney 53982eaa055fSJeff Mahoney static int btrfs_ioctl_set_features(struct file *file, void __user *arg) 53992eaa055fSJeff Mahoney { 54000b246afaSJeff Mahoney struct inode *inode = file_inode(file); 54010b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 54020b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 54030b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 54042eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags flags[2]; 54052eaa055fSJeff Mahoney struct btrfs_trans_handle *trans; 54062eaa055fSJeff Mahoney u64 newflags; 54072eaa055fSJeff Mahoney int ret; 54082eaa055fSJeff Mahoney 54092eaa055fSJeff Mahoney if (!capable(CAP_SYS_ADMIN)) 54102eaa055fSJeff Mahoney return -EPERM; 54112eaa055fSJeff Mahoney 54122eaa055fSJeff Mahoney if (copy_from_user(flags, arg, sizeof(flags))) 54132eaa055fSJeff Mahoney return -EFAULT; 54142eaa055fSJeff Mahoney 54152eaa055fSJeff Mahoney /* Nothing to do */ 54162eaa055fSJeff Mahoney if (!flags[0].compat_flags && !flags[0].compat_ro_flags && 54172eaa055fSJeff Mahoney !flags[0].incompat_flags) 54182eaa055fSJeff Mahoney return 0; 54192eaa055fSJeff Mahoney 54202ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_flags, 54212eaa055fSJeff Mahoney flags[1].compat_flags, COMPAT); 54222eaa055fSJeff Mahoney if (ret) 54232eaa055fSJeff Mahoney return ret; 54242eaa055fSJeff Mahoney 54252ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_ro_flags, 54262eaa055fSJeff Mahoney flags[1].compat_ro_flags, COMPAT_RO); 54272eaa055fSJeff Mahoney if (ret) 54282eaa055fSJeff Mahoney return ret; 54292eaa055fSJeff Mahoney 54302ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].incompat_flags, 54312eaa055fSJeff Mahoney flags[1].incompat_flags, INCOMPAT); 54322eaa055fSJeff Mahoney if (ret) 54332eaa055fSJeff Mahoney return ret; 54342eaa055fSJeff Mahoney 54357ab19625SDavid Sterba ret = mnt_want_write_file(file); 54367ab19625SDavid Sterba if (ret) 54377ab19625SDavid Sterba return ret; 54387ab19625SDavid Sterba 54398051aa1aSDavid Sterba trans = btrfs_start_transaction(root, 0); 54407ab19625SDavid Sterba if (IS_ERR(trans)) { 54417ab19625SDavid Sterba ret = PTR_ERR(trans); 54427ab19625SDavid Sterba goto out_drop_write; 54437ab19625SDavid Sterba } 54442eaa055fSJeff Mahoney 54450b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 54462eaa055fSJeff Mahoney newflags = btrfs_super_compat_flags(super_block); 54472eaa055fSJeff Mahoney newflags |= flags[0].compat_flags & flags[1].compat_flags; 54482eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags); 54492eaa055fSJeff Mahoney btrfs_set_super_compat_flags(super_block, newflags); 54502eaa055fSJeff Mahoney 54512eaa055fSJeff Mahoney newflags = btrfs_super_compat_ro_flags(super_block); 54522eaa055fSJeff Mahoney newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags; 54532eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags); 54542eaa055fSJeff Mahoney btrfs_set_super_compat_ro_flags(super_block, newflags); 54552eaa055fSJeff Mahoney 54562eaa055fSJeff Mahoney newflags = btrfs_super_incompat_flags(super_block); 54572eaa055fSJeff Mahoney newflags |= flags[0].incompat_flags & flags[1].incompat_flags; 54582eaa055fSJeff Mahoney newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags); 54592eaa055fSJeff Mahoney btrfs_set_super_incompat_flags(super_block, newflags); 54600b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 54612eaa055fSJeff Mahoney 54623a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 54637ab19625SDavid Sterba out_drop_write: 54647ab19625SDavid Sterba mnt_drop_write_file(file); 54657ab19625SDavid Sterba 54667ab19625SDavid Sterba return ret; 54672eaa055fSJeff Mahoney } 54682eaa055fSJeff Mahoney 5469f46b5a66SChristoph Hellwig long btrfs_ioctl(struct file *file, unsigned int 5470f46b5a66SChristoph Hellwig cmd, unsigned long arg) 5471f46b5a66SChristoph Hellwig { 54720b246afaSJeff Mahoney struct inode *inode = file_inode(file); 54730b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 54740b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 54754bcabaa3SChristoph Hellwig void __user *argp = (void __user *)arg; 5476f46b5a66SChristoph Hellwig 5477f46b5a66SChristoph Hellwig switch (cmd) { 54786cbff00fSChristoph Hellwig case FS_IOC_GETFLAGS: 54796cbff00fSChristoph Hellwig return btrfs_ioctl_getflags(file, argp); 54806cbff00fSChristoph Hellwig case FS_IOC_SETFLAGS: 54816cbff00fSChristoph Hellwig return btrfs_ioctl_setflags(file, argp); 54826cbff00fSChristoph Hellwig case FS_IOC_GETVERSION: 54836cbff00fSChristoph Hellwig return btrfs_ioctl_getversion(file, argp); 5484f7039b1dSLi Dongyang case FITRIM: 5485f7039b1dSLi Dongyang return btrfs_ioctl_fitrim(file, argp); 5486f46b5a66SChristoph Hellwig case BTRFS_IOC_SNAP_CREATE: 5487fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 0); 5488fdfb1e4fSLi Zefan case BTRFS_IOC_SNAP_CREATE_V2: 5489fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create_v2(file, argp, 0); 54903de4586cSChris Mason case BTRFS_IOC_SUBVOL_CREATE: 5491fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 1); 54926f72c7e2SArne Jansen case BTRFS_IOC_SUBVOL_CREATE_V2: 54936f72c7e2SArne Jansen return btrfs_ioctl_snap_create_v2(file, argp, 1); 549476dda93cSYan, Zheng case BTRFS_IOC_SNAP_DESTROY: 549576dda93cSYan, Zheng return btrfs_ioctl_snap_destroy(file, argp); 54960caa102dSLi Zefan case BTRFS_IOC_SUBVOL_GETFLAGS: 54970caa102dSLi Zefan return btrfs_ioctl_subvol_getflags(file, argp); 54980caa102dSLi Zefan case BTRFS_IOC_SUBVOL_SETFLAGS: 54990caa102dSLi Zefan return btrfs_ioctl_subvol_setflags(file, argp); 55006ef5ed0dSJosef Bacik case BTRFS_IOC_DEFAULT_SUBVOL: 55016ef5ed0dSJosef Bacik return btrfs_ioctl_default_subvol(file, argp); 5502f46b5a66SChristoph Hellwig case BTRFS_IOC_DEFRAG: 55031e701a32SChris Mason return btrfs_ioctl_defrag(file, NULL); 55041e701a32SChris Mason case BTRFS_IOC_DEFRAG_RANGE: 55051e701a32SChris Mason return btrfs_ioctl_defrag(file, argp); 5506f46b5a66SChristoph Hellwig case BTRFS_IOC_RESIZE: 5507198605a8SMiao Xie return btrfs_ioctl_resize(file, argp); 5508f46b5a66SChristoph Hellwig case BTRFS_IOC_ADD_DEV: 55092ff7e61eSJeff Mahoney return btrfs_ioctl_add_dev(fs_info, argp); 5510f46b5a66SChristoph Hellwig case BTRFS_IOC_RM_DEV: 5511da24927bSMiao Xie return btrfs_ioctl_rm_dev(file, argp); 55126b526ed7SAnand Jain case BTRFS_IOC_RM_DEV_V2: 55136b526ed7SAnand Jain return btrfs_ioctl_rm_dev_v2(file, argp); 5514475f6387SJan Schmidt case BTRFS_IOC_FS_INFO: 55152ff7e61eSJeff Mahoney return btrfs_ioctl_fs_info(fs_info, argp); 5516475f6387SJan Schmidt case BTRFS_IOC_DEV_INFO: 55172ff7e61eSJeff Mahoney return btrfs_ioctl_dev_info(fs_info, argp); 5518f46b5a66SChristoph Hellwig case BTRFS_IOC_BALANCE: 55199ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, NULL); 5520f46b5a66SChristoph Hellwig case BTRFS_IOC_TRANS_START: 5521f46b5a66SChristoph Hellwig return btrfs_ioctl_trans_start(file); 5522f46b5a66SChristoph Hellwig case BTRFS_IOC_TRANS_END: 5523f46b5a66SChristoph Hellwig return btrfs_ioctl_trans_end(file); 5524ac8e9819SChris Mason case BTRFS_IOC_TREE_SEARCH: 5525ac8e9819SChris Mason return btrfs_ioctl_tree_search(file, argp); 5526cc68a8a5SGerhard Heift case BTRFS_IOC_TREE_SEARCH_V2: 5527cc68a8a5SGerhard Heift return btrfs_ioctl_tree_search_v2(file, argp); 5528ac8e9819SChris Mason case BTRFS_IOC_INO_LOOKUP: 5529ac8e9819SChris Mason return btrfs_ioctl_ino_lookup(file, argp); 5530d7728c96SJan Schmidt case BTRFS_IOC_INO_PATHS: 5531d7728c96SJan Schmidt return btrfs_ioctl_ino_to_path(root, argp); 5532d7728c96SJan Schmidt case BTRFS_IOC_LOGICAL_INO: 55332ff7e61eSJeff Mahoney return btrfs_ioctl_logical_to_ino(fs_info, argp); 55341406e432SJosef Bacik case BTRFS_IOC_SPACE_INFO: 55352ff7e61eSJeff Mahoney return btrfs_ioctl_space_info(fs_info, argp); 55369b199859SFilipe David Borba Manana case BTRFS_IOC_SYNC: { 55379b199859SFilipe David Borba Manana int ret; 55389b199859SFilipe David Borba Manana 55390b246afaSJeff Mahoney ret = btrfs_start_delalloc_roots(fs_info, 0, -1); 55409b199859SFilipe David Borba Manana if (ret) 55419b199859SFilipe David Borba Manana return ret; 55420b246afaSJeff Mahoney ret = btrfs_sync_fs(inode->i_sb, 1); 55432fad4e83SDavid Sterba /* 55442fad4e83SDavid Sterba * The transaction thread may want to do more work, 554501327610SNicholas D Steeves * namely it pokes the cleaner kthread that will start 55462fad4e83SDavid Sterba * processing uncleaned subvols. 55472fad4e83SDavid Sterba */ 55480b246afaSJeff Mahoney wake_up_process(fs_info->transaction_kthread); 55499b199859SFilipe David Borba Manana return ret; 55509b199859SFilipe David Borba Manana } 555146204592SSage Weil case BTRFS_IOC_START_SYNC: 55529a8c28beSMiao Xie return btrfs_ioctl_start_sync(root, argp); 555346204592SSage Weil case BTRFS_IOC_WAIT_SYNC: 55542ff7e61eSJeff Mahoney return btrfs_ioctl_wait_sync(fs_info, argp); 5555475f6387SJan Schmidt case BTRFS_IOC_SCRUB: 5556b8e95489SMiao Xie return btrfs_ioctl_scrub(file, argp); 5557475f6387SJan Schmidt case BTRFS_IOC_SCRUB_CANCEL: 55582ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_cancel(fs_info); 5559475f6387SJan Schmidt case BTRFS_IOC_SCRUB_PROGRESS: 55602ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_progress(fs_info, argp); 5561c9e9f97bSIlya Dryomov case BTRFS_IOC_BALANCE_V2: 55629ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, argp); 5563837d5b6eSIlya Dryomov case BTRFS_IOC_BALANCE_CTL: 55642ff7e61eSJeff Mahoney return btrfs_ioctl_balance_ctl(fs_info, arg); 556519a39dceSIlya Dryomov case BTRFS_IOC_BALANCE_PROGRESS: 55662ff7e61eSJeff Mahoney return btrfs_ioctl_balance_progress(fs_info, argp); 55678ea05e3aSAlexander Block case BTRFS_IOC_SET_RECEIVED_SUBVOL: 55688ea05e3aSAlexander Block return btrfs_ioctl_set_received_subvol(file, argp); 5569abccd00fSHugo Mills #ifdef CONFIG_64BIT 5570abccd00fSHugo Mills case BTRFS_IOC_SET_RECEIVED_SUBVOL_32: 5571abccd00fSHugo Mills return btrfs_ioctl_set_received_subvol_32(file, argp); 5572abccd00fSHugo Mills #endif 557331db9f7cSAlexander Block case BTRFS_IOC_SEND: 557431db9f7cSAlexander Block return btrfs_ioctl_send(file, argp); 5575c11d2c23SStefan Behrens case BTRFS_IOC_GET_DEV_STATS: 55762ff7e61eSJeff Mahoney return btrfs_ioctl_get_dev_stats(fs_info, argp); 55775d13a37bSArne Jansen case BTRFS_IOC_QUOTA_CTL: 5578905b0ddaSMiao Xie return btrfs_ioctl_quota_ctl(file, argp); 55795d13a37bSArne Jansen case BTRFS_IOC_QGROUP_ASSIGN: 5580905b0ddaSMiao Xie return btrfs_ioctl_qgroup_assign(file, argp); 55815d13a37bSArne Jansen case BTRFS_IOC_QGROUP_CREATE: 5582905b0ddaSMiao Xie return btrfs_ioctl_qgroup_create(file, argp); 55835d13a37bSArne Jansen case BTRFS_IOC_QGROUP_LIMIT: 5584905b0ddaSMiao Xie return btrfs_ioctl_qgroup_limit(file, argp); 55852f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN: 55862f232036SJan Schmidt return btrfs_ioctl_quota_rescan(file, argp); 55872f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_STATUS: 55882f232036SJan Schmidt return btrfs_ioctl_quota_rescan_status(file, argp); 558957254b6eSJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_WAIT: 559057254b6eSJan Schmidt return btrfs_ioctl_quota_rescan_wait(file, argp); 55913f6bcfbdSStefan Behrens case BTRFS_IOC_DEV_REPLACE: 55922ff7e61eSJeff Mahoney return btrfs_ioctl_dev_replace(fs_info, argp); 5593867ab667Sjeff.liu case BTRFS_IOC_GET_FSLABEL: 5594867ab667Sjeff.liu return btrfs_ioctl_get_fslabel(file, argp); 5595a8bfd4abSjeff.liu case BTRFS_IOC_SET_FSLABEL: 5596a8bfd4abSjeff.liu return btrfs_ioctl_set_fslabel(file, argp); 55972eaa055fSJeff Mahoney case BTRFS_IOC_GET_SUPPORTED_FEATURES: 5598d5131b65SDavid Sterba return btrfs_ioctl_get_supported_features(argp); 55992eaa055fSJeff Mahoney case BTRFS_IOC_GET_FEATURES: 56002eaa055fSJeff Mahoney return btrfs_ioctl_get_features(file, argp); 56012eaa055fSJeff Mahoney case BTRFS_IOC_SET_FEATURES: 56022eaa055fSJeff Mahoney return btrfs_ioctl_set_features(file, argp); 5603f46b5a66SChristoph Hellwig } 5604f46b5a66SChristoph Hellwig 5605f46b5a66SChristoph Hellwig return -ENOTTY; 5606f46b5a66SChristoph Hellwig } 56074c63c245SLuke Dashjr 56084c63c245SLuke Dashjr #ifdef CONFIG_COMPAT 56094c63c245SLuke Dashjr long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 56104c63c245SLuke Dashjr { 56112a362249SJeff Mahoney /* 56122a362249SJeff Mahoney * These all access 32-bit values anyway so no further 56132a362249SJeff Mahoney * handling is necessary. 56142a362249SJeff Mahoney */ 56154c63c245SLuke Dashjr switch (cmd) { 56164c63c245SLuke Dashjr case FS_IOC32_GETFLAGS: 56174c63c245SLuke Dashjr cmd = FS_IOC_GETFLAGS; 56184c63c245SLuke Dashjr break; 56194c63c245SLuke Dashjr case FS_IOC32_SETFLAGS: 56204c63c245SLuke Dashjr cmd = FS_IOC_SETFLAGS; 56214c63c245SLuke Dashjr break; 56224c63c245SLuke Dashjr case FS_IOC32_GETVERSION: 56234c63c245SLuke Dashjr cmd = FS_IOC_GETVERSION; 56244c63c245SLuke Dashjr break; 56254c63c245SLuke Dashjr } 56264c63c245SLuke Dashjr 56274c63c245SLuke Dashjr return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 56284c63c245SLuke Dashjr } 56294c63c245SLuke Dashjr #endif 5630