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"; 2995c1aab1dSNick Terrell else if (fs_info->compress_type == BTRFS_COMPRESS_ZLIB) 30063541927SFilipe David Borba Manana comp = "zlib"; 3015c1aab1dSNick Terrell else 3025c1aab1dSNick Terrell comp = "zstd"; 30363541927SFilipe David Borba Manana ret = btrfs_set_prop(inode, "btrfs.compression", 30463541927SFilipe David Borba Manana comp, strlen(comp), 0); 30563541927SFilipe David Borba Manana if (ret) 30663541927SFilipe David Borba Manana goto out_drop; 30763541927SFilipe David Borba Manana 308ebcb904dSLi Zefan } else { 30978a017a2SFilipe Manana ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0); 31078a017a2SFilipe Manana if (ret && ret != -ENODATA) 31178a017a2SFilipe Manana goto out_drop; 312ebcb904dSLi Zefan ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS); 31375e7cb7fSLiu Bo } 3146cbff00fSChristoph Hellwig 3154da6f1a3SLi Zefan trans = btrfs_start_transaction(root, 1); 316f062abf0SLi Zefan if (IS_ERR(trans)) { 317f062abf0SLi Zefan ret = PTR_ERR(trans); 318f062abf0SLi Zefan goto out_drop; 319f062abf0SLi Zefan } 3206cbff00fSChristoph Hellwig 321306424ccSLi Zefan btrfs_update_iflags(inode); 3220c4d2d95SJosef Bacik inode_inc_iversion(inode); 323c2050a45SDeepa Dinamani inode->i_ctime = current_time(inode); 3246cbff00fSChristoph Hellwig ret = btrfs_update_inode(trans, root, inode); 3256cbff00fSChristoph Hellwig 3263a45bb20SJeff Mahoney btrfs_end_transaction(trans); 327f062abf0SLi Zefan out_drop: 328f062abf0SLi Zefan if (ret) { 329f062abf0SLi Zefan ip->flags = ip_oldflags; 330f062abf0SLi Zefan inode->i_flags = i_oldflags; 331f062abf0SLi Zefan } 3326cbff00fSChristoph Hellwig 3336cbff00fSChristoph Hellwig out_unlock: 3345955102cSAl Viro inode_unlock(inode); 335e7848683SJan Kara mnt_drop_write_file(file); 3362d4e6f6aSliubo return ret; 3376cbff00fSChristoph Hellwig } 3386cbff00fSChristoph Hellwig 3396cbff00fSChristoph Hellwig static int btrfs_ioctl_getversion(struct file *file, int __user *arg) 3406cbff00fSChristoph Hellwig { 341496ad9aaSAl Viro struct inode *inode = file_inode(file); 3426cbff00fSChristoph Hellwig 3436cbff00fSChristoph Hellwig return put_user(inode->i_generation, arg); 3446cbff00fSChristoph Hellwig } 345f46b5a66SChristoph Hellwig 346f7039b1dSLi Dongyang static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg) 347f7039b1dSLi Dongyang { 3480b246afaSJeff Mahoney struct inode *inode = file_inode(file); 3490b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 350f7039b1dSLi Dongyang struct btrfs_device *device; 351f7039b1dSLi Dongyang struct request_queue *q; 352f7039b1dSLi Dongyang struct fstrim_range range; 353f7039b1dSLi Dongyang u64 minlen = ULLONG_MAX; 354f7039b1dSLi Dongyang u64 num_devices = 0; 355815745cfSAl Viro u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy); 356f7039b1dSLi Dongyang int ret; 357f7039b1dSLi Dongyang 358f7039b1dSLi Dongyang if (!capable(CAP_SYS_ADMIN)) 359f7039b1dSLi Dongyang return -EPERM; 360f7039b1dSLi Dongyang 3611f78160cSXiao Guangrong rcu_read_lock(); 3621f78160cSXiao Guangrong list_for_each_entry_rcu(device, &fs_info->fs_devices->devices, 3631f78160cSXiao Guangrong dev_list) { 364f7039b1dSLi Dongyang if (!device->bdev) 365f7039b1dSLi Dongyang continue; 366f7039b1dSLi Dongyang q = bdev_get_queue(device->bdev); 367f7039b1dSLi Dongyang if (blk_queue_discard(q)) { 368f7039b1dSLi Dongyang num_devices++; 36950d0446eSSeraphime Kirkovski minlen = min_t(u64, q->limits.discard_granularity, 370f7039b1dSLi Dongyang minlen); 371f7039b1dSLi Dongyang } 372f7039b1dSLi Dongyang } 3731f78160cSXiao Guangrong rcu_read_unlock(); 374f4c697e6SLukas Czerner 375f7039b1dSLi Dongyang if (!num_devices) 376f7039b1dSLi Dongyang return -EOPNOTSUPP; 377f7039b1dSLi Dongyang if (copy_from_user(&range, arg, sizeof(range))) 378f7039b1dSLi Dongyang return -EFAULT; 379e515c18bSLukas Czerner if (range.start > total_bytes || 380e515c18bSLukas Czerner range.len < fs_info->sb->s_blocksize) 381f4c697e6SLukas Czerner return -EINVAL; 382f7039b1dSLi Dongyang 383f4c697e6SLukas Czerner range.len = min(range.len, total_bytes - range.start); 384f7039b1dSLi Dongyang range.minlen = max(range.minlen, minlen); 3852ff7e61eSJeff Mahoney ret = btrfs_trim_fs(fs_info, &range); 386f7039b1dSLi Dongyang if (ret < 0) 387f7039b1dSLi Dongyang return ret; 388f7039b1dSLi Dongyang 389f7039b1dSLi Dongyang if (copy_to_user(arg, &range, sizeof(range))) 390f7039b1dSLi Dongyang return -EFAULT; 391f7039b1dSLi Dongyang 392f7039b1dSLi Dongyang return 0; 393f7039b1dSLi Dongyang } 394f7039b1dSLi Dongyang 395dd5f9615SStefan Behrens int btrfs_is_empty_uuid(u8 *uuid) 396dd5f9615SStefan Behrens { 39746e0f66aSChris Mason int i; 39846e0f66aSChris Mason 39946e0f66aSChris Mason for (i = 0; i < BTRFS_UUID_SIZE; i++) { 40046e0f66aSChris Mason if (uuid[i]) 40146e0f66aSChris Mason return 0; 40246e0f66aSChris Mason } 40346e0f66aSChris Mason return 1; 404dd5f9615SStefan Behrens } 405dd5f9615SStefan Behrens 406d5c12070SMiao Xie static noinline int create_subvol(struct inode *dir, 407cb8e7090SChristoph Hellwig struct dentry *dentry, 40852f75f4fSDavid Sterba const char *name, int namelen, 4096f72c7e2SArne Jansen u64 *async_transid, 4108696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 411f46b5a66SChristoph Hellwig { 4120b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 413f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 414f46b5a66SChristoph Hellwig struct btrfs_key key; 41549a3c4d9SDavid Sterba struct btrfs_root_item *root_item; 416f46b5a66SChristoph Hellwig struct btrfs_inode_item *inode_item; 417f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 418d5c12070SMiao Xie struct btrfs_root *root = BTRFS_I(dir)->root; 41976dda93cSYan, Zheng struct btrfs_root *new_root; 420d5c12070SMiao Xie struct btrfs_block_rsv block_rsv; 421c2050a45SDeepa Dinamani struct timespec cur_time = current_time(dir); 4225662344bSTsutomu Itoh struct inode *inode; 423f46b5a66SChristoph Hellwig int ret; 424f46b5a66SChristoph Hellwig int err; 425f46b5a66SChristoph Hellwig u64 objectid; 426f46b5a66SChristoph Hellwig u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; 4273de4586cSChris Mason u64 index = 0; 428d5c12070SMiao Xie u64 qgroup_reserved; 4298ea05e3aSAlexander Block uuid_le new_uuid; 430f46b5a66SChristoph Hellwig 43149a3c4d9SDavid Sterba root_item = kzalloc(sizeof(*root_item), GFP_KERNEL); 43249a3c4d9SDavid Sterba if (!root_item) 43349a3c4d9SDavid Sterba return -ENOMEM; 43449a3c4d9SDavid Sterba 4350b246afaSJeff Mahoney ret = btrfs_find_free_objectid(fs_info->tree_root, &objectid); 4362fbe8c8aSAl Viro if (ret) 43749a3c4d9SDavid Sterba goto fail_free; 4386a912213SJosef Bacik 439e09fe2d2SQu Wenruo /* 440e09fe2d2SQu Wenruo * Don't create subvolume whose level is not zero. Or qgroup will be 44101327610SNicholas D Steeves * screwed up since it assumes subvolume qgroup's level to be 0. 442e09fe2d2SQu Wenruo */ 44349a3c4d9SDavid Sterba if (btrfs_qgroup_level(objectid)) { 44449a3c4d9SDavid Sterba ret = -ENOSPC; 44549a3c4d9SDavid Sterba goto fail_free; 44649a3c4d9SDavid Sterba } 447e09fe2d2SQu Wenruo 448d5c12070SMiao Xie btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); 4499ed74f2dSJosef Bacik /* 450d5c12070SMiao Xie * The same as the snapshot creation, please see the comment 451d5c12070SMiao Xie * of create_snapshot(). 4529ed74f2dSJosef Bacik */ 453d5c12070SMiao Xie ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 454dd5f9615SStefan Behrens 8, &qgroup_reserved, false); 455d5c12070SMiao Xie if (ret) 45649a3c4d9SDavid Sterba goto fail_free; 457f46b5a66SChristoph Hellwig 458d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 459d5c12070SMiao Xie if (IS_ERR(trans)) { 460d5c12070SMiao Xie ret = PTR_ERR(trans); 4617775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &block_rsv); 46249a3c4d9SDavid Sterba goto fail_free; 463d5c12070SMiao Xie } 464d5c12070SMiao Xie trans->block_rsv = &block_rsv; 465d5c12070SMiao Xie trans->bytes_reserved = block_rsv.size; 466f46b5a66SChristoph Hellwig 4670b246afaSJeff Mahoney ret = btrfs_qgroup_inherit(trans, fs_info, 0, objectid, inherit); 4686f72c7e2SArne Jansen if (ret) 4696f72c7e2SArne Jansen goto fail; 4706f72c7e2SArne Jansen 4714d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); 4728e8a1e31SJosef Bacik if (IS_ERR(leaf)) { 4738e8a1e31SJosef Bacik ret = PTR_ERR(leaf); 4748e8a1e31SJosef Bacik goto fail; 4758e8a1e31SJosef Bacik } 476f46b5a66SChristoph Hellwig 477b159fa28SDavid Sterba memzero_extent_buffer(leaf, 0, sizeof(struct btrfs_header)); 478f46b5a66SChristoph Hellwig btrfs_set_header_bytenr(leaf, leaf->start); 479f46b5a66SChristoph Hellwig btrfs_set_header_generation(leaf, trans->transid); 4805d4f98a2SYan Zheng btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 481f46b5a66SChristoph Hellwig btrfs_set_header_owner(leaf, objectid); 482f46b5a66SChristoph Hellwig 4830b246afaSJeff Mahoney write_extent_buffer_fsid(leaf, fs_info->fsid); 4840b246afaSJeff Mahoney write_extent_buffer_chunk_tree_uuid(leaf, fs_info->chunk_tree_uuid); 485f46b5a66SChristoph Hellwig btrfs_mark_buffer_dirty(leaf); 486f46b5a66SChristoph Hellwig 48749a3c4d9SDavid Sterba inode_item = &root_item->inode; 4883cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 4893cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 4903cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 491da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 4920b246afaSJeff Mahoney fs_info->nodesize); 4933cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 494f46b5a66SChristoph Hellwig 49549a3c4d9SDavid Sterba btrfs_set_root_flags(root_item, 0); 49649a3c4d9SDavid Sterba btrfs_set_root_limit(root_item, 0); 4973cae210fSQu Wenruo btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT); 49808fe4db1SLi Zefan 49949a3c4d9SDavid Sterba btrfs_set_root_bytenr(root_item, leaf->start); 50049a3c4d9SDavid Sterba btrfs_set_root_generation(root_item, trans->transid); 50149a3c4d9SDavid Sterba btrfs_set_root_level(root_item, 0); 50249a3c4d9SDavid Sterba btrfs_set_root_refs(root_item, 1); 50349a3c4d9SDavid Sterba btrfs_set_root_used(root_item, leaf->len); 50449a3c4d9SDavid Sterba btrfs_set_root_last_snapshot(root_item, 0); 505f46b5a66SChristoph Hellwig 50649a3c4d9SDavid Sterba btrfs_set_root_generation_v2(root_item, 50749a3c4d9SDavid Sterba btrfs_root_generation(root_item)); 5088ea05e3aSAlexander Block uuid_le_gen(&new_uuid); 50949a3c4d9SDavid Sterba memcpy(root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE); 51049a3c4d9SDavid Sterba btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec); 51149a3c4d9SDavid Sterba btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec); 51249a3c4d9SDavid Sterba root_item->ctime = root_item->otime; 51349a3c4d9SDavid Sterba btrfs_set_root_ctransid(root_item, trans->transid); 51449a3c4d9SDavid Sterba btrfs_set_root_otransid(root_item, trans->transid); 515f46b5a66SChristoph Hellwig 516925baeddSChris Mason btrfs_tree_unlock(leaf); 517f46b5a66SChristoph Hellwig free_extent_buffer(leaf); 518f46b5a66SChristoph Hellwig leaf = NULL; 519f46b5a66SChristoph Hellwig 52049a3c4d9SDavid Sterba btrfs_set_root_dirid(root_item, new_dirid); 521f46b5a66SChristoph Hellwig 522f46b5a66SChristoph Hellwig key.objectid = objectid; 5235d4f98a2SYan Zheng key.offset = 0; 524962a298fSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 5250b246afaSJeff Mahoney ret = btrfs_insert_root(trans, fs_info->tree_root, &key, 52649a3c4d9SDavid Sterba root_item); 527f46b5a66SChristoph Hellwig if (ret) 528f46b5a66SChristoph Hellwig goto fail; 529f46b5a66SChristoph Hellwig 53076dda93cSYan, Zheng key.offset = (u64)-1; 5310b246afaSJeff Mahoney new_root = btrfs_read_fs_root_no_name(fs_info, &key); 53279787eaaSJeff Mahoney if (IS_ERR(new_root)) { 53379787eaaSJeff Mahoney ret = PTR_ERR(new_root); 53466642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 53579787eaaSJeff Mahoney goto fail; 53679787eaaSJeff Mahoney } 53776dda93cSYan, Zheng 53876dda93cSYan, Zheng btrfs_record_root_in_trans(trans, new_root); 53976dda93cSYan, Zheng 54063541927SFilipe David Borba Manana ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid); 541ce598979SMark Fasheh if (ret) { 542ce598979SMark Fasheh /* We potentially lose an unused inode item here */ 54366642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 544ce598979SMark Fasheh goto fail; 545ce598979SMark Fasheh } 546ce598979SMark Fasheh 547f32e48e9SChandan Rajendra mutex_lock(&new_root->objectid_mutex); 548f32e48e9SChandan Rajendra new_root->highest_objectid = new_dirid; 549f32e48e9SChandan Rajendra mutex_unlock(&new_root->objectid_mutex); 550f32e48e9SChandan Rajendra 551f46b5a66SChristoph Hellwig /* 552f46b5a66SChristoph Hellwig * insert the directory item 553f46b5a66SChristoph Hellwig */ 554877574e2SNikolay Borisov ret = btrfs_set_inode_index(BTRFS_I(dir), &index); 55579787eaaSJeff Mahoney if (ret) { 55666642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 55779787eaaSJeff Mahoney goto fail; 55879787eaaSJeff Mahoney } 5593de4586cSChris Mason 5603de4586cSChris Mason ret = btrfs_insert_dir_item(trans, root, 5618e7611cfSNikolay Borisov name, namelen, BTRFS_I(dir), &key, 5623de4586cSChris Mason BTRFS_FT_DIR, index); 56379787eaaSJeff Mahoney if (ret) { 56466642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 565f46b5a66SChristoph Hellwig goto fail; 56679787eaaSJeff Mahoney } 5670660b5afSChris Mason 5686ef06d27SNikolay Borisov btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2); 56952c26179SYan Zheng ret = btrfs_update_inode(trans, root, dir); 57052c26179SYan Zheng BUG_ON(ret); 57152c26179SYan Zheng 5720b246afaSJeff Mahoney ret = btrfs_add_root_ref(trans, fs_info, 5734df27c4dSYan, Zheng objectid, root->root_key.objectid, 5744a0cc7caSNikolay Borisov btrfs_ino(BTRFS_I(dir)), index, name, namelen); 5750660b5afSChris Mason BUG_ON(ret); 5760660b5afSChris Mason 5770b246afaSJeff Mahoney ret = btrfs_uuid_tree_add(trans, fs_info, root_item->uuid, 5786bccf3abSJeff Mahoney BTRFS_UUID_KEY_SUBVOL, objectid); 579dd5f9615SStefan Behrens if (ret) 58066642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 581dd5f9615SStefan Behrens 582f46b5a66SChristoph Hellwig fail: 58349a3c4d9SDavid Sterba kfree(root_item); 584d5c12070SMiao Xie trans->block_rsv = NULL; 585d5c12070SMiao Xie trans->bytes_reserved = 0; 5867775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &block_rsv); 587de6e8200SLiu Bo 58872fd032eSSage Weil if (async_transid) { 58972fd032eSSage Weil *async_transid = trans->transid; 5903a45bb20SJeff Mahoney err = btrfs_commit_transaction_async(trans, 1); 59100d71c9cSMiao Xie if (err) 5923a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 59372fd032eSSage Weil } else { 5943a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 59572fd032eSSage Weil } 596f46b5a66SChristoph Hellwig if (err && !ret) 597f46b5a66SChristoph Hellwig ret = err; 5981a65e24bSChris Mason 5995662344bSTsutomu Itoh if (!ret) { 6005662344bSTsutomu Itoh inode = btrfs_lookup_dentry(dir, dentry); 601de6e8200SLiu Bo if (IS_ERR(inode)) 602de6e8200SLiu Bo return PTR_ERR(inode); 6035662344bSTsutomu Itoh d_instantiate(dentry, inode); 6045662344bSTsutomu Itoh } 605f46b5a66SChristoph Hellwig return ret; 60649a3c4d9SDavid Sterba 60749a3c4d9SDavid Sterba fail_free: 60849a3c4d9SDavid Sterba kfree(root_item); 60949a3c4d9SDavid Sterba return ret; 610f46b5a66SChristoph Hellwig } 611f46b5a66SChristoph Hellwig 612e9662f70SMiao Xie static int create_snapshot(struct btrfs_root *root, struct inode *dir, 61361d7e4cbSDavid Sterba struct dentry *dentry, 614e9662f70SMiao Xie u64 *async_transid, bool readonly, 615e9662f70SMiao Xie struct btrfs_qgroup_inherit *inherit) 616f46b5a66SChristoph Hellwig { 6170b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 6182e4bfab9SYan, Zheng struct inode *inode; 619f46b5a66SChristoph Hellwig struct btrfs_pending_snapshot *pending_snapshot; 620f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 6212e4bfab9SYan, Zheng int ret; 622f46b5a66SChristoph Hellwig 62327cdeb70SMiao Xie if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state)) 624f46b5a66SChristoph Hellwig return -EINVAL; 625f46b5a66SChristoph Hellwig 62623269bf5SDavid Sterba pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL); 627a1ee7362SDavid Sterba if (!pending_snapshot) 628a1ee7362SDavid Sterba return -ENOMEM; 629a1ee7362SDavid Sterba 630b0c0ea63SDavid Sterba pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item), 63123269bf5SDavid Sterba GFP_KERNEL); 6328546b570SDavid Sterba pending_snapshot->path = btrfs_alloc_path(); 6338546b570SDavid Sterba if (!pending_snapshot->root_item || !pending_snapshot->path) { 634b0c0ea63SDavid Sterba ret = -ENOMEM; 635b0c0ea63SDavid Sterba goto free_pending; 636b0c0ea63SDavid Sterba } 637b0c0ea63SDavid Sterba 638ea14b57fSDavid Sterba atomic_inc(&root->will_be_snapshotted); 6394e857c58SPeter Zijlstra smp_mb__after_atomic(); 64045bac0f3SLiu Bo /* wait for no snapshot writes */ 64145bac0f3SLiu Bo wait_event(root->subv_writers->wait, 64245bac0f3SLiu Bo percpu_counter_sum(&root->subv_writers->counter) == 0); 6438257b2dcSMiao Xie 6446a03843dSMiao Xie ret = btrfs_start_delalloc_inodes(root, 0); 6456a03843dSMiao Xie if (ret) 646a1ee7362SDavid Sterba goto dec_and_free; 6476a03843dSMiao Xie 6486374e57aSChris Mason btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1); 6496a03843dSMiao Xie 65066d8f3ddSMiao Xie btrfs_init_block_rsv(&pending_snapshot->block_rsv, 65166d8f3ddSMiao Xie BTRFS_BLOCK_RSV_TEMP); 652d5c12070SMiao Xie /* 653d5c12070SMiao Xie * 1 - parent dir inode 654d5c12070SMiao Xie * 2 - dir entries 655d5c12070SMiao Xie * 1 - root item 656d5c12070SMiao Xie * 2 - root ref/backref 657d5c12070SMiao Xie * 1 - root of snapshot 658dd5f9615SStefan Behrens * 1 - UUID item 659d5c12070SMiao Xie */ 660d5c12070SMiao Xie ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root, 661dd5f9615SStefan Behrens &pending_snapshot->block_rsv, 8, 662ee3441b4SJeff Mahoney &pending_snapshot->qgroup_reserved, 663ee3441b4SJeff Mahoney false); 664d5c12070SMiao Xie if (ret) 665a1ee7362SDavid Sterba goto dec_and_free; 666d5c12070SMiao Xie 667a22285a6SYan, Zheng pending_snapshot->dentry = dentry; 668a22285a6SYan, Zheng pending_snapshot->root = root; 669b83cc969SLi Zefan pending_snapshot->readonly = readonly; 670e9662f70SMiao Xie pending_snapshot->dir = dir; 6718696c533SMiao Xie pending_snapshot->inherit = inherit; 672a22285a6SYan, Zheng 673d5c12070SMiao Xie trans = btrfs_start_transaction(root, 0); 674a22285a6SYan, Zheng if (IS_ERR(trans)) { 675a22285a6SYan, Zheng ret = PTR_ERR(trans); 676a22285a6SYan, Zheng goto fail; 677a22285a6SYan, Zheng } 678a22285a6SYan, Zheng 6790b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 680a22285a6SYan, Zheng list_add(&pending_snapshot->list, 681a22285a6SYan, Zheng &trans->transaction->pending_snapshots); 6820b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 68372fd032eSSage Weil if (async_transid) { 68472fd032eSSage Weil *async_transid = trans->transid; 6853a45bb20SJeff Mahoney ret = btrfs_commit_transaction_async(trans, 1); 68600d71c9cSMiao Xie if (ret) 6873a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 68872fd032eSSage Weil } else { 6893a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 69072fd032eSSage Weil } 691aec8030aSMiao Xie if (ret) 692c37b2b62SJosef Bacik goto fail; 693a22285a6SYan, Zheng 694a22285a6SYan, Zheng ret = pending_snapshot->error; 695f46b5a66SChristoph Hellwig if (ret) 6962e4bfab9SYan, Zheng goto fail; 697f46b5a66SChristoph Hellwig 698d3797308SChris Mason ret = btrfs_orphan_cleanup(pending_snapshot->snap); 699d3797308SChris Mason if (ret) 700d3797308SChris Mason goto fail; 701d3797308SChris Mason 7022b0143b5SDavid Howells inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry); 7032e4bfab9SYan, Zheng if (IS_ERR(inode)) { 7042e4bfab9SYan, Zheng ret = PTR_ERR(inode); 7052e4bfab9SYan, Zheng goto fail; 7062e4bfab9SYan, Zheng } 7075662344bSTsutomu Itoh 7082e4bfab9SYan, Zheng d_instantiate(dentry, inode); 7092e4bfab9SYan, Zheng ret = 0; 7102e4bfab9SYan, Zheng fail: 7117775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &pending_snapshot->block_rsv); 712a1ee7362SDavid Sterba dec_and_free: 713ea14b57fSDavid Sterba if (atomic_dec_and_test(&root->will_be_snapshotted)) 714ea14b57fSDavid Sterba wake_up_atomic_t(&root->will_be_snapshotted); 715b0c0ea63SDavid Sterba free_pending: 716b0c0ea63SDavid Sterba kfree(pending_snapshot->root_item); 7178546b570SDavid Sterba btrfs_free_path(pending_snapshot->path); 718a1ee7362SDavid Sterba kfree(pending_snapshot); 719a1ee7362SDavid Sterba 720f46b5a66SChristoph Hellwig return ret; 721f46b5a66SChristoph Hellwig } 722f46b5a66SChristoph Hellwig 7234260f7c7SSage Weil /* copy of may_delete in fs/namei.c() 7244260f7c7SSage Weil * Check whether we can remove a link victim from directory dir, check 7254260f7c7SSage Weil * whether the type of victim is right. 7264260f7c7SSage Weil * 1. We can't do it if dir is read-only (done in permission()) 7274260f7c7SSage Weil * 2. We should have write and exec permissions on dir 7284260f7c7SSage Weil * 3. We can't remove anything from append-only dir 7294260f7c7SSage Weil * 4. We can't do anything with immutable dir (done in permission()) 7304260f7c7SSage Weil * 5. If the sticky bit on dir is set we should either 7314260f7c7SSage Weil * a. be owner of dir, or 7324260f7c7SSage Weil * b. be owner of victim, or 7334260f7c7SSage Weil * c. have CAP_FOWNER capability 73401327610SNicholas D Steeves * 6. If the victim is append-only or immutable we can't do anything with 7354260f7c7SSage Weil * links pointing to it. 7364260f7c7SSage Weil * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. 7374260f7c7SSage Weil * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. 7384260f7c7SSage Weil * 9. We can't remove a root or mountpoint. 7394260f7c7SSage Weil * 10. We don't allow removal of NFS sillyrenamed files; it's handled by 7404260f7c7SSage Weil * nfs_async_unlink(). 7414260f7c7SSage Weil */ 7424260f7c7SSage Weil 7434260f7c7SSage Weil static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir) 7444260f7c7SSage Weil { 7454260f7c7SSage Weil int error; 7464260f7c7SSage Weil 7472b0143b5SDavid Howells if (d_really_is_negative(victim)) 7484260f7c7SSage Weil return -ENOENT; 7494260f7c7SSage Weil 7502b0143b5SDavid Howells BUG_ON(d_inode(victim->d_parent) != dir); 7514fa6b5ecSJeff Layton audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); 7524260f7c7SSage Weil 7534260f7c7SSage Weil error = inode_permission(dir, MAY_WRITE | MAY_EXEC); 7544260f7c7SSage Weil if (error) 7554260f7c7SSage Weil return error; 7564260f7c7SSage Weil if (IS_APPEND(dir)) 7574260f7c7SSage Weil return -EPERM; 7582b0143b5SDavid Howells if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) || 7592b0143b5SDavid Howells IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim))) 7604260f7c7SSage Weil return -EPERM; 7614260f7c7SSage Weil if (isdir) { 762e36cb0b8SDavid Howells if (!d_is_dir(victim)) 7634260f7c7SSage Weil return -ENOTDIR; 7644260f7c7SSage Weil if (IS_ROOT(victim)) 7654260f7c7SSage Weil return -EBUSY; 766e36cb0b8SDavid Howells } else if (d_is_dir(victim)) 7674260f7c7SSage Weil return -EISDIR; 7684260f7c7SSage Weil if (IS_DEADDIR(dir)) 7694260f7c7SSage Weil return -ENOENT; 7704260f7c7SSage Weil if (victim->d_flags & DCACHE_NFSFS_RENAMED) 7714260f7c7SSage Weil return -EBUSY; 7724260f7c7SSage Weil return 0; 7734260f7c7SSage Weil } 7744260f7c7SSage Weil 775cb8e7090SChristoph Hellwig /* copy of may_create in fs/namei.c() */ 776cb8e7090SChristoph Hellwig static inline int btrfs_may_create(struct inode *dir, struct dentry *child) 777cb8e7090SChristoph Hellwig { 7782b0143b5SDavid Howells if (d_really_is_positive(child)) 779cb8e7090SChristoph Hellwig return -EEXIST; 780cb8e7090SChristoph Hellwig if (IS_DEADDIR(dir)) 781cb8e7090SChristoph Hellwig return -ENOENT; 782cb8e7090SChristoph Hellwig return inode_permission(dir, MAY_WRITE | MAY_EXEC); 783cb8e7090SChristoph Hellwig } 784cb8e7090SChristoph Hellwig 785cb8e7090SChristoph Hellwig /* 786cb8e7090SChristoph Hellwig * Create a new subvolume below @parent. This is largely modeled after 787cb8e7090SChristoph Hellwig * sys_mkdirat and vfs_mkdir, but we only do a single component lookup 788cb8e7090SChristoph Hellwig * inside this filesystem so it's quite a bit simpler. 789cb8e7090SChristoph Hellwig */ 79092872094SAl Viro static noinline int btrfs_mksubvol(const struct path *parent, 79152f75f4fSDavid Sterba const char *name, int namelen, 79272fd032eSSage Weil struct btrfs_root *snap_src, 7936f72c7e2SArne Jansen u64 *async_transid, bool readonly, 7948696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 795cb8e7090SChristoph Hellwig { 7962b0143b5SDavid Howells struct inode *dir = d_inode(parent->dentry); 7970b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb); 798cb8e7090SChristoph Hellwig struct dentry *dentry; 799cb8e7090SChristoph Hellwig int error; 800cb8e7090SChristoph Hellwig 80100235411SAl Viro error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 80200235411SAl Viro if (error == -EINTR) 80300235411SAl Viro return error; 804cb8e7090SChristoph Hellwig 805cb8e7090SChristoph Hellwig dentry = lookup_one_len(name, parent->dentry, namelen); 806cb8e7090SChristoph Hellwig error = PTR_ERR(dentry); 807cb8e7090SChristoph Hellwig if (IS_ERR(dentry)) 808cb8e7090SChristoph Hellwig goto out_unlock; 809cb8e7090SChristoph Hellwig 81076dda93cSYan, Zheng error = btrfs_may_create(dir, dentry); 811cb8e7090SChristoph Hellwig if (error) 812a874a63eSLiu Bo goto out_dput; 813cb8e7090SChristoph Hellwig 8149c52057cSChris Mason /* 8159c52057cSChris Mason * even if this name doesn't exist, we may get hash collisions. 8169c52057cSChris Mason * check for them now when we can safely fail 8179c52057cSChris Mason */ 8189c52057cSChris Mason error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root, 8199c52057cSChris Mason dir->i_ino, name, 8209c52057cSChris Mason namelen); 8219c52057cSChris Mason if (error) 8229c52057cSChris Mason goto out_dput; 8239c52057cSChris Mason 8240b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 82576dda93cSYan, Zheng 82676dda93cSYan, Zheng if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0) 82776dda93cSYan, Zheng goto out_up_read; 82876dda93cSYan, Zheng 8293de4586cSChris Mason if (snap_src) { 83061d7e4cbSDavid Sterba error = create_snapshot(snap_src, dir, dentry, 8316f72c7e2SArne Jansen async_transid, readonly, inherit); 8323de4586cSChris Mason } else { 833d5c12070SMiao Xie error = create_subvol(dir, dentry, name, namelen, 834d5c12070SMiao Xie async_transid, inherit); 8353de4586cSChris Mason } 83676dda93cSYan, Zheng if (!error) 83776dda93cSYan, Zheng fsnotify_mkdir(dir, dentry); 83876dda93cSYan, Zheng out_up_read: 8390b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 840cb8e7090SChristoph Hellwig out_dput: 841cb8e7090SChristoph Hellwig dput(dentry); 842cb8e7090SChristoph Hellwig out_unlock: 8435955102cSAl Viro inode_unlock(dir); 844cb8e7090SChristoph Hellwig return error; 845cb8e7090SChristoph Hellwig } 846cb8e7090SChristoph Hellwig 8474cb5300bSChris Mason /* 8484cb5300bSChris Mason * When we're defragging a range, we don't want to kick it off again 8494cb5300bSChris Mason * if it is really just waiting for delalloc to send it down. 8504cb5300bSChris Mason * If we find a nice big extent or delalloc range for the bytes in the 8514cb5300bSChris Mason * file you want to defrag, we return 0 to let you know to skip this 8524cb5300bSChris Mason * part of the file 8534cb5300bSChris Mason */ 854aab110abSDavid Sterba static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh) 8554cb5300bSChris Mason { 8564cb5300bSChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 8574cb5300bSChris Mason struct extent_map *em = NULL; 8584cb5300bSChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 8594cb5300bSChris Mason u64 end; 8604cb5300bSChris Mason 8614cb5300bSChris Mason read_lock(&em_tree->lock); 86209cbfeafSKirill A. Shutemov em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE); 8634cb5300bSChris Mason read_unlock(&em_tree->lock); 8644cb5300bSChris Mason 8654cb5300bSChris Mason if (em) { 8664cb5300bSChris Mason end = extent_map_end(em); 8674cb5300bSChris Mason free_extent_map(em); 8684cb5300bSChris Mason if (end - offset > thresh) 8694cb5300bSChris Mason return 0; 8704cb5300bSChris Mason } 8714cb5300bSChris Mason /* if we already have a nice delalloc here, just stop */ 8724cb5300bSChris Mason thresh /= 2; 8734cb5300bSChris Mason end = count_range_bits(io_tree, &offset, offset + thresh, 8744cb5300bSChris Mason thresh, EXTENT_DELALLOC, 1); 8754cb5300bSChris Mason if (end >= thresh) 8764cb5300bSChris Mason return 0; 8774cb5300bSChris Mason return 1; 8784cb5300bSChris Mason } 8794cb5300bSChris Mason 8804cb5300bSChris Mason /* 8814cb5300bSChris Mason * helper function to walk through a file and find extents 8824cb5300bSChris Mason * newer than a specific transid, and smaller than thresh. 8834cb5300bSChris Mason * 8844cb5300bSChris Mason * This is used by the defragging code to find new and small 8854cb5300bSChris Mason * extents 8864cb5300bSChris Mason */ 8874cb5300bSChris Mason static int find_new_extents(struct btrfs_root *root, 8884cb5300bSChris Mason struct inode *inode, u64 newer_than, 889aab110abSDavid Sterba u64 *off, u32 thresh) 8904cb5300bSChris Mason { 8914cb5300bSChris Mason struct btrfs_path *path; 8924cb5300bSChris Mason struct btrfs_key min_key; 8934cb5300bSChris Mason struct extent_buffer *leaf; 8944cb5300bSChris Mason struct btrfs_file_extent_item *extent; 8954cb5300bSChris Mason int type; 8964cb5300bSChris Mason int ret; 8974a0cc7caSNikolay Borisov u64 ino = btrfs_ino(BTRFS_I(inode)); 8984cb5300bSChris Mason 8994cb5300bSChris Mason path = btrfs_alloc_path(); 9004cb5300bSChris Mason if (!path) 9014cb5300bSChris Mason return -ENOMEM; 9024cb5300bSChris Mason 903a4689d2bSDavid Sterba min_key.objectid = ino; 9044cb5300bSChris Mason min_key.type = BTRFS_EXTENT_DATA_KEY; 9054cb5300bSChris Mason min_key.offset = *off; 9064cb5300bSChris Mason 9074cb5300bSChris Mason while (1) { 9086174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &min_key, path, newer_than); 9094cb5300bSChris Mason if (ret != 0) 9104cb5300bSChris Mason goto none; 911f094c9bdSFilipe Manana process_slot: 912a4689d2bSDavid Sterba if (min_key.objectid != ino) 9134cb5300bSChris Mason goto none; 9144cb5300bSChris Mason if (min_key.type != BTRFS_EXTENT_DATA_KEY) 9154cb5300bSChris Mason goto none; 9164cb5300bSChris Mason 9174cb5300bSChris Mason leaf = path->nodes[0]; 9184cb5300bSChris Mason extent = btrfs_item_ptr(leaf, path->slots[0], 9194cb5300bSChris Mason struct btrfs_file_extent_item); 9204cb5300bSChris Mason 9214cb5300bSChris Mason type = btrfs_file_extent_type(leaf, extent); 9224cb5300bSChris Mason if (type == BTRFS_FILE_EXTENT_REG && 9234cb5300bSChris Mason btrfs_file_extent_num_bytes(leaf, extent) < thresh && 9244cb5300bSChris Mason check_defrag_in_cache(inode, min_key.offset, thresh)) { 9254cb5300bSChris Mason *off = min_key.offset; 9264cb5300bSChris Mason btrfs_free_path(path); 9274cb5300bSChris Mason return 0; 9284cb5300bSChris Mason } 9294cb5300bSChris Mason 930f094c9bdSFilipe Manana path->slots[0]++; 931f094c9bdSFilipe Manana if (path->slots[0] < btrfs_header_nritems(leaf)) { 932f094c9bdSFilipe Manana btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]); 933f094c9bdSFilipe Manana goto process_slot; 934f094c9bdSFilipe Manana } 935f094c9bdSFilipe Manana 9364cb5300bSChris Mason if (min_key.offset == (u64)-1) 9374cb5300bSChris Mason goto none; 9384cb5300bSChris Mason 9394cb5300bSChris Mason min_key.offset++; 9404cb5300bSChris Mason btrfs_release_path(path); 9414cb5300bSChris Mason } 9424cb5300bSChris Mason none: 9434cb5300bSChris Mason btrfs_free_path(path); 9444cb5300bSChris Mason return -ENOENT; 9454cb5300bSChris Mason } 9464cb5300bSChris Mason 9476c282eb4SLi Zefan static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start) 94817ce6ef8SLiu Bo { 94917ce6ef8SLiu Bo struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 950940100a4SChris Mason struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 9516c282eb4SLi Zefan struct extent_map *em; 95209cbfeafSKirill A. Shutemov u64 len = PAGE_SIZE; 953940100a4SChris Mason 954940100a4SChris Mason /* 955940100a4SChris Mason * hopefully we have this extent in the tree already, try without 956940100a4SChris Mason * the full extent lock 957940100a4SChris Mason */ 958940100a4SChris Mason read_lock(&em_tree->lock); 959940100a4SChris Mason em = lookup_extent_mapping(em_tree, start, len); 960940100a4SChris Mason read_unlock(&em_tree->lock); 961940100a4SChris Mason 962940100a4SChris Mason if (!em) { 963308d9800SFilipe Manana struct extent_state *cached = NULL; 964308d9800SFilipe Manana u64 end = start + len - 1; 965308d9800SFilipe Manana 966940100a4SChris Mason /* get the big lock and read metadata off disk */ 967ff13db41SDavid Sterba lock_extent_bits(io_tree, start, end, &cached); 968fc4f21b1SNikolay Borisov em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len, 0); 969308d9800SFilipe Manana unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS); 970940100a4SChris Mason 9716cf8bfbfSDan Carpenter if (IS_ERR(em)) 9726c282eb4SLi Zefan return NULL; 973940100a4SChris Mason } 974940100a4SChris Mason 9756c282eb4SLi Zefan return em; 9766c282eb4SLi Zefan } 9776c282eb4SLi Zefan 9786c282eb4SLi Zefan static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em) 9796c282eb4SLi Zefan { 9806c282eb4SLi Zefan struct extent_map *next; 9816c282eb4SLi Zefan bool ret = true; 9826c282eb4SLi Zefan 9836c282eb4SLi Zefan /* this is the last extent */ 9846c282eb4SLi Zefan if (em->start + em->len >= i_size_read(inode)) 9856c282eb4SLi Zefan return false; 9866c282eb4SLi Zefan 9876c282eb4SLi Zefan next = defrag_lookup_extent(inode, em->start + em->len); 988e9512d72SChris Mason if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE) 989e9512d72SChris Mason ret = false; 990e9512d72SChris Mason else if ((em->block_start + em->block_len == next->block_start) && 991ee22184bSByongho Lee (em->block_len > SZ_128K && next->block_len > SZ_128K)) 9926c282eb4SLi Zefan ret = false; 9936c282eb4SLi Zefan 9946c282eb4SLi Zefan free_extent_map(next); 9956c282eb4SLi Zefan return ret; 9966c282eb4SLi Zefan } 9976c282eb4SLi Zefan 998aab110abSDavid Sterba static int should_defrag_range(struct inode *inode, u64 start, u32 thresh, 999a43a2111SAndrew Mahone u64 *last_len, u64 *skip, u64 *defrag_end, 1000a43a2111SAndrew Mahone int compress) 10016c282eb4SLi Zefan { 10026c282eb4SLi Zefan struct extent_map *em; 10036c282eb4SLi Zefan int ret = 1; 10046c282eb4SLi Zefan bool next_mergeable = true; 10054a3560c4SLiu Bo bool prev_mergeable = true; 10066c282eb4SLi Zefan 10076c282eb4SLi Zefan /* 10086c282eb4SLi Zefan * make sure that once we start defragging an extent, we keep on 10096c282eb4SLi Zefan * defragging it 10106c282eb4SLi Zefan */ 10116c282eb4SLi Zefan if (start < *defrag_end) 10126c282eb4SLi Zefan return 1; 10136c282eb4SLi Zefan 10146c282eb4SLi Zefan *skip = 0; 10156c282eb4SLi Zefan 10166c282eb4SLi Zefan em = defrag_lookup_extent(inode, start); 10176c282eb4SLi Zefan if (!em) 10186c282eb4SLi Zefan return 0; 10196c282eb4SLi Zefan 1020940100a4SChris Mason /* this will cover holes, and inline extents */ 102117ce6ef8SLiu Bo if (em->block_start >= EXTENT_MAP_LAST_BYTE) { 1022940100a4SChris Mason ret = 0; 102317ce6ef8SLiu Bo goto out; 102417ce6ef8SLiu Bo } 102517ce6ef8SLiu Bo 10264a3560c4SLiu Bo if (!*defrag_end) 10274a3560c4SLiu Bo prev_mergeable = false; 10284a3560c4SLiu Bo 10296c282eb4SLi Zefan next_mergeable = defrag_check_next_extent(inode, em); 1030940100a4SChris Mason /* 10316c282eb4SLi Zefan * we hit a real extent, if it is big or the next extent is not a 10326c282eb4SLi Zefan * real extent, don't bother defragging it 1033940100a4SChris Mason */ 1034a43a2111SAndrew Mahone if (!compress && (*last_len == 0 || *last_len >= thresh) && 10354a3560c4SLiu Bo (em->len >= thresh || (!next_mergeable && !prev_mergeable))) 1036940100a4SChris Mason ret = 0; 103717ce6ef8SLiu Bo out: 1038940100a4SChris Mason /* 1039940100a4SChris Mason * last_len ends up being a counter of how many bytes we've defragged. 1040940100a4SChris Mason * every time we choose not to defrag an extent, we reset *last_len 1041940100a4SChris Mason * so that the next tiny extent will force a defrag. 1042940100a4SChris Mason * 1043940100a4SChris Mason * The end result of this is that tiny extents before a single big 1044940100a4SChris Mason * extent will force at least part of that big extent to be defragged. 1045940100a4SChris Mason */ 1046940100a4SChris Mason if (ret) { 1047940100a4SChris Mason *defrag_end = extent_map_end(em); 1048940100a4SChris Mason } else { 1049940100a4SChris Mason *last_len = 0; 1050940100a4SChris Mason *skip = extent_map_end(em); 1051940100a4SChris Mason *defrag_end = 0; 1052940100a4SChris Mason } 1053940100a4SChris Mason 1054940100a4SChris Mason free_extent_map(em); 1055940100a4SChris Mason return ret; 1056940100a4SChris Mason } 1057940100a4SChris Mason 10584cb5300bSChris Mason /* 10594cb5300bSChris Mason * it doesn't do much good to defrag one or two pages 10604cb5300bSChris Mason * at a time. This pulls in a nice chunk of pages 10614cb5300bSChris Mason * to COW and defrag. 10624cb5300bSChris Mason * 10634cb5300bSChris Mason * It also makes sure the delalloc code has enough 10644cb5300bSChris Mason * dirty data to avoid making new small extents as part 10654cb5300bSChris Mason * of the defrag 10664cb5300bSChris Mason * 10674cb5300bSChris Mason * It's a good idea to start RA on this range 10684cb5300bSChris Mason * before calling this. 10694cb5300bSChris Mason */ 10704cb5300bSChris Mason static int cluster_pages_for_defrag(struct inode *inode, 10714cb5300bSChris Mason struct page **pages, 10724cb5300bSChris Mason unsigned long start_index, 1073c41570c9SJustin Maggard unsigned long num_pages) 1074f46b5a66SChristoph Hellwig { 10754cb5300bSChris Mason unsigned long file_end; 10764cb5300bSChris Mason u64 isize = i_size_read(inode); 1077f46b5a66SChristoph Hellwig u64 page_start; 1078f46b5a66SChristoph Hellwig u64 page_end; 10791f12bd06SLiu Bo u64 page_cnt; 10804cb5300bSChris Mason int ret; 10814cb5300bSChris Mason int i; 10824cb5300bSChris Mason int i_done; 10834cb5300bSChris Mason struct btrfs_ordered_extent *ordered; 10844cb5300bSChris Mason struct extent_state *cached_state = NULL; 1085600a45e1SMiao Xie struct extent_io_tree *tree; 1086364ecf36SQu Wenruo struct extent_changeset *data_reserved = NULL; 10873b16a4e3SJosef Bacik gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping); 10884cb5300bSChris Mason 108909cbfeafSKirill A. Shutemov file_end = (isize - 1) >> PAGE_SHIFT; 10901f12bd06SLiu Bo if (!isize || start_index > file_end) 10911f12bd06SLiu Bo return 0; 10921f12bd06SLiu Bo 10931f12bd06SLiu Bo page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1); 10944cb5300bSChris Mason 1095364ecf36SQu Wenruo ret = btrfs_delalloc_reserve_space(inode, &data_reserved, 109609cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 109709cbfeafSKirill A. Shutemov page_cnt << PAGE_SHIFT); 10984cb5300bSChris Mason if (ret) 10994cb5300bSChris Mason return ret; 11004cb5300bSChris Mason i_done = 0; 1101600a45e1SMiao Xie tree = &BTRFS_I(inode)->io_tree; 11024cb5300bSChris Mason 11034cb5300bSChris Mason /* step one, lock all the pages */ 11041f12bd06SLiu Bo for (i = 0; i < page_cnt; i++) { 11054cb5300bSChris Mason struct page *page; 1106600a45e1SMiao Xie again: 1107a94733d0SJosef Bacik page = find_or_create_page(inode->i_mapping, 11083b16a4e3SJosef Bacik start_index + i, mask); 11094cb5300bSChris Mason if (!page) 11104cb5300bSChris Mason break; 11114cb5300bSChris Mason 1112600a45e1SMiao Xie page_start = page_offset(page); 111309cbfeafSKirill A. Shutemov page_end = page_start + PAGE_SIZE - 1; 1114600a45e1SMiao Xie while (1) { 1115308d9800SFilipe Manana lock_extent_bits(tree, page_start, page_end, 1116ff13db41SDavid Sterba &cached_state); 1117600a45e1SMiao Xie ordered = btrfs_lookup_ordered_extent(inode, 1118600a45e1SMiao Xie page_start); 1119308d9800SFilipe Manana unlock_extent_cached(tree, page_start, page_end, 1120308d9800SFilipe Manana &cached_state, GFP_NOFS); 1121600a45e1SMiao Xie if (!ordered) 1122600a45e1SMiao Xie break; 1123600a45e1SMiao Xie 1124600a45e1SMiao Xie unlock_page(page); 1125600a45e1SMiao Xie btrfs_start_ordered_extent(inode, ordered, 1); 1126600a45e1SMiao Xie btrfs_put_ordered_extent(ordered); 1127600a45e1SMiao Xie lock_page(page); 11281f12bd06SLiu Bo /* 11291f12bd06SLiu Bo * we unlocked the page above, so we need check if 11301f12bd06SLiu Bo * it was released or not. 11311f12bd06SLiu Bo */ 11321f12bd06SLiu Bo if (page->mapping != inode->i_mapping) { 11331f12bd06SLiu Bo unlock_page(page); 113409cbfeafSKirill A. Shutemov put_page(page); 11351f12bd06SLiu Bo goto again; 11361f12bd06SLiu Bo } 1137600a45e1SMiao Xie } 1138600a45e1SMiao Xie 11394cb5300bSChris Mason if (!PageUptodate(page)) { 11404cb5300bSChris Mason btrfs_readpage(NULL, page); 11414cb5300bSChris Mason lock_page(page); 11424cb5300bSChris Mason if (!PageUptodate(page)) { 11434cb5300bSChris Mason unlock_page(page); 114409cbfeafSKirill A. Shutemov put_page(page); 11454cb5300bSChris Mason ret = -EIO; 11464cb5300bSChris Mason break; 11474cb5300bSChris Mason } 11484cb5300bSChris Mason } 1149600a45e1SMiao Xie 1150600a45e1SMiao Xie if (page->mapping != inode->i_mapping) { 1151600a45e1SMiao Xie unlock_page(page); 115209cbfeafSKirill A. Shutemov put_page(page); 1153600a45e1SMiao Xie goto again; 1154600a45e1SMiao Xie } 1155600a45e1SMiao Xie 11564cb5300bSChris Mason pages[i] = page; 11574cb5300bSChris Mason i_done++; 11584cb5300bSChris Mason } 11594cb5300bSChris Mason if (!i_done || ret) 11604cb5300bSChris Mason goto out; 11614cb5300bSChris Mason 11624cb5300bSChris Mason if (!(inode->i_sb->s_flags & MS_ACTIVE)) 11634cb5300bSChris Mason goto out; 11644cb5300bSChris Mason 11654cb5300bSChris Mason /* 11664cb5300bSChris Mason * so now we have a nice long stream of locked 11674cb5300bSChris Mason * and up to date pages, lets wait on them 11684cb5300bSChris Mason */ 11694cb5300bSChris Mason for (i = 0; i < i_done; i++) 11704cb5300bSChris Mason wait_on_page_writeback(pages[i]); 11714cb5300bSChris Mason 11724cb5300bSChris Mason page_start = page_offset(pages[0]); 117309cbfeafSKirill A. Shutemov page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE; 11744cb5300bSChris Mason 11754cb5300bSChris Mason lock_extent_bits(&BTRFS_I(inode)->io_tree, 1176ff13db41SDavid Sterba page_start, page_end - 1, &cached_state); 11774cb5300bSChris Mason clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, 11784cb5300bSChris Mason page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC | 11799e8a4a8bSLiu Bo EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0, 11809e8a4a8bSLiu Bo &cached_state, GFP_NOFS); 11814cb5300bSChris Mason 11821f12bd06SLiu Bo if (i_done != page_cnt) { 11839e0baf60SJosef Bacik spin_lock(&BTRFS_I(inode)->lock); 11849e0baf60SJosef Bacik BTRFS_I(inode)->outstanding_extents++; 11859e0baf60SJosef Bacik spin_unlock(&BTRFS_I(inode)->lock); 1186bc42bda2SQu Wenruo btrfs_delalloc_release_space(inode, data_reserved, 118709cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 118809cbfeafSKirill A. Shutemov (page_cnt - i_done) << PAGE_SHIFT); 11894cb5300bSChris Mason } 11904cb5300bSChris Mason 11914cb5300bSChris Mason 11929e8a4a8bSLiu Bo set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1, 1193018ed4f7SDavid Sterba &cached_state); 11944cb5300bSChris Mason 11954cb5300bSChris Mason unlock_extent_cached(&BTRFS_I(inode)->io_tree, 11964cb5300bSChris Mason page_start, page_end - 1, &cached_state, 11974cb5300bSChris Mason GFP_NOFS); 11984cb5300bSChris Mason 11994cb5300bSChris Mason for (i = 0; i < i_done; i++) { 12004cb5300bSChris Mason clear_page_dirty_for_io(pages[i]); 12014cb5300bSChris Mason ClearPageChecked(pages[i]); 12024cb5300bSChris Mason set_page_extent_mapped(pages[i]); 12034cb5300bSChris Mason set_page_dirty(pages[i]); 12044cb5300bSChris Mason unlock_page(pages[i]); 120509cbfeafSKirill A. Shutemov put_page(pages[i]); 12064cb5300bSChris Mason } 1207364ecf36SQu Wenruo extent_changeset_free(data_reserved); 12084cb5300bSChris Mason return i_done; 12094cb5300bSChris Mason out: 12104cb5300bSChris Mason for (i = 0; i < i_done; i++) { 12114cb5300bSChris Mason unlock_page(pages[i]); 121209cbfeafSKirill A. Shutemov put_page(pages[i]); 12134cb5300bSChris Mason } 1214bc42bda2SQu Wenruo btrfs_delalloc_release_space(inode, data_reserved, 121509cbfeafSKirill A. Shutemov start_index << PAGE_SHIFT, 121609cbfeafSKirill A. Shutemov page_cnt << PAGE_SHIFT); 1217364ecf36SQu Wenruo extent_changeset_free(data_reserved); 12184cb5300bSChris Mason return ret; 12194cb5300bSChris Mason 12204cb5300bSChris Mason } 12214cb5300bSChris Mason 12224cb5300bSChris Mason int btrfs_defrag_file(struct inode *inode, struct file *file, 12234cb5300bSChris Mason struct btrfs_ioctl_defrag_range_args *range, 12244cb5300bSChris Mason u64 newer_than, unsigned long max_to_defrag) 12254cb5300bSChris Mason { 12260b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 12274cb5300bSChris Mason struct btrfs_root *root = BTRFS_I(inode)->root; 12284cb5300bSChris Mason struct file_ra_state *ra = NULL; 12294cb5300bSChris Mason unsigned long last_index; 1230151a31b2SLi Zefan u64 isize = i_size_read(inode); 1231940100a4SChris Mason u64 last_len = 0; 1232940100a4SChris Mason u64 skip = 0; 1233940100a4SChris Mason u64 defrag_end = 0; 12344cb5300bSChris Mason u64 newer_off = range->start; 1235f46b5a66SChristoph Hellwig unsigned long i; 1236008873eaSLi Zefan unsigned long ra_index = 0; 1237f46b5a66SChristoph Hellwig int ret; 12384cb5300bSChris Mason int defrag_count = 0; 12391a419d85SLi Zefan int compress_type = BTRFS_COMPRESS_ZLIB; 1240aab110abSDavid Sterba u32 extent_thresh = range->extent_thresh; 124109cbfeafSKirill A. Shutemov unsigned long max_cluster = SZ_256K >> PAGE_SHIFT; 1242c41570c9SJustin Maggard unsigned long cluster = max_cluster; 1243ee22184bSByongho Lee u64 new_align = ~((u64)SZ_128K - 1); 12444cb5300bSChris Mason struct page **pages = NULL; 12451e2ef46dSDavid Sterba bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS; 12464cb5300bSChris Mason 12470abd5b17SLiu Bo if (isize == 0) 12480abd5b17SLiu Bo return 0; 12490abd5b17SLiu Bo 12500abd5b17SLiu Bo if (range->start >= isize) 12510abd5b17SLiu Bo return -EINVAL; 12521a419d85SLi Zefan 12531e2ef46dSDavid Sterba if (do_compress) { 12541a419d85SLi Zefan if (range->compress_type > BTRFS_COMPRESS_TYPES) 12551a419d85SLi Zefan return -EINVAL; 12561a419d85SLi Zefan if (range->compress_type) 12571a419d85SLi Zefan compress_type = range->compress_type; 12581a419d85SLi Zefan } 1259f46b5a66SChristoph Hellwig 12600abd5b17SLiu Bo if (extent_thresh == 0) 1261ee22184bSByongho Lee extent_thresh = SZ_256K; 1262f46b5a66SChristoph Hellwig 12634cb5300bSChris Mason /* 12640a52d108SDavid Sterba * If we were not given a file, allocate a readahead context. As 12650a52d108SDavid Sterba * readahead is just an optimization, defrag will work without it so 12660a52d108SDavid Sterba * we don't error out. 12674cb5300bSChris Mason */ 12684cb5300bSChris Mason if (!file) { 126963e727ecSDavid Sterba ra = kzalloc(sizeof(*ra), GFP_KERNEL); 12700a52d108SDavid Sterba if (ra) 12714cb5300bSChris Mason file_ra_state_init(ra, inode->i_mapping); 12724cb5300bSChris Mason } else { 12734cb5300bSChris Mason ra = &file->f_ra; 12744cb5300bSChris Mason } 12754cb5300bSChris Mason 127663e727ecSDavid Sterba pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL); 12774cb5300bSChris Mason if (!pages) { 12784cb5300bSChris Mason ret = -ENOMEM; 12794cb5300bSChris Mason goto out_ra; 12804cb5300bSChris Mason } 12814cb5300bSChris Mason 12824cb5300bSChris Mason /* find the last page to defrag */ 12831e701a32SChris Mason if (range->start + range->len > range->start) { 1284151a31b2SLi Zefan last_index = min_t(u64, isize - 1, 128509cbfeafSKirill A. Shutemov range->start + range->len - 1) >> PAGE_SHIFT; 12861e701a32SChris Mason } else { 128709cbfeafSKirill A. Shutemov last_index = (isize - 1) >> PAGE_SHIFT; 12881e701a32SChris Mason } 12891e701a32SChris Mason 12904cb5300bSChris Mason if (newer_than) { 12914cb5300bSChris Mason ret = find_new_extents(root, inode, newer_than, 1292ee22184bSByongho Lee &newer_off, SZ_64K); 12934cb5300bSChris Mason if (!ret) { 12944cb5300bSChris Mason range->start = newer_off; 12954cb5300bSChris Mason /* 12964cb5300bSChris Mason * we always align our defrag to help keep 12974cb5300bSChris Mason * the extents in the file evenly spaced 12984cb5300bSChris Mason */ 129909cbfeafSKirill A. Shutemov i = (newer_off & new_align) >> PAGE_SHIFT; 13004cb5300bSChris Mason } else 13014cb5300bSChris Mason goto out_ra; 13024cb5300bSChris Mason } else { 130309cbfeafSKirill A. Shutemov i = range->start >> PAGE_SHIFT; 13044cb5300bSChris Mason } 13054cb5300bSChris Mason if (!max_to_defrag) 1306070034bdSchandan max_to_defrag = last_index - i + 1; 13074cb5300bSChris Mason 13082a0f7f57SLi Zefan /* 13092a0f7f57SLi Zefan * make writeback starts from i, so the defrag range can be 13102a0f7f57SLi Zefan * written sequentially. 13112a0f7f57SLi Zefan */ 13122a0f7f57SLi Zefan if (i < inode->i_mapping->writeback_index) 13132a0f7f57SLi Zefan inode->i_mapping->writeback_index = i; 13142a0f7f57SLi Zefan 1315f7f43cc8SChris Mason while (i <= last_index && defrag_count < max_to_defrag && 131609cbfeafSKirill A. Shutemov (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) { 13174cb5300bSChris Mason /* 13184cb5300bSChris Mason * make sure we stop running if someone unmounts 13194cb5300bSChris Mason * the FS 13204cb5300bSChris Mason */ 13214cb5300bSChris Mason if (!(inode->i_sb->s_flags & MS_ACTIVE)) 13224cb5300bSChris Mason break; 13234cb5300bSChris Mason 13240b246afaSJeff Mahoney if (btrfs_defrag_cancelled(fs_info)) { 13250b246afaSJeff Mahoney btrfs_debug(fs_info, "defrag_file cancelled"); 1326210549ebSDavid Sterba ret = -EAGAIN; 1327210549ebSDavid Sterba break; 1328210549ebSDavid Sterba } 1329210549ebSDavid Sterba 133009cbfeafSKirill A. Shutemov if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT, 13316c282eb4SLi Zefan extent_thresh, &last_len, &skip, 13321e2ef46dSDavid Sterba &defrag_end, do_compress)){ 1333940100a4SChris Mason unsigned long next; 1334940100a4SChris Mason /* 1335940100a4SChris Mason * the should_defrag function tells us how much to skip 1336940100a4SChris Mason * bump our counter by the suggested amount 1337940100a4SChris Mason */ 133809cbfeafSKirill A. Shutemov next = DIV_ROUND_UP(skip, PAGE_SIZE); 1339940100a4SChris Mason i = max(i + 1, next); 1340940100a4SChris Mason continue; 1341940100a4SChris Mason } 1342008873eaSLi Zefan 1343008873eaSLi Zefan if (!newer_than) { 134409cbfeafSKirill A. Shutemov cluster = (PAGE_ALIGN(defrag_end) >> 134509cbfeafSKirill A. Shutemov PAGE_SHIFT) - i; 1346008873eaSLi Zefan cluster = min(cluster, max_cluster); 1347008873eaSLi Zefan } else { 1348008873eaSLi Zefan cluster = max_cluster; 1349008873eaSLi Zefan } 1350008873eaSLi Zefan 1351008873eaSLi Zefan if (i + cluster > ra_index) { 1352008873eaSLi Zefan ra_index = max(i, ra_index); 13530a52d108SDavid Sterba if (ra) 1354d3c0bab5SDavid Sterba page_cache_sync_readahead(inode->i_mapping, ra, 1355d3c0bab5SDavid Sterba file, ra_index, cluster); 1356e4826a5bSchandan ra_index += cluster; 1357008873eaSLi Zefan } 13584cb5300bSChris Mason 13595955102cSAl Viro inode_lock(inode); 13601e2ef46dSDavid Sterba if (do_compress) 1361eec63c65SDavid Sterba BTRFS_I(inode)->defrag_compress = compress_type; 1362008873eaSLi Zefan ret = cluster_pages_for_defrag(inode, pages, i, cluster); 1363ecb8bea8SLiu Bo if (ret < 0) { 13645955102cSAl Viro inode_unlock(inode); 13654cb5300bSChris Mason goto out_ra; 1366ecb8bea8SLiu Bo } 13674cb5300bSChris Mason 13684cb5300bSChris Mason defrag_count += ret; 1369d0e1d66bSNamjae Jeon balance_dirty_pages_ratelimited(inode->i_mapping); 13705955102cSAl Viro inode_unlock(inode); 13714cb5300bSChris Mason 13724cb5300bSChris Mason if (newer_than) { 13734cb5300bSChris Mason if (newer_off == (u64)-1) 13744cb5300bSChris Mason break; 13754cb5300bSChris Mason 1376e1f041e1SLiu Bo if (ret > 0) 1377e1f041e1SLiu Bo i += ret; 1378e1f041e1SLiu Bo 13794cb5300bSChris Mason newer_off = max(newer_off + 1, 138009cbfeafSKirill A. Shutemov (u64)i << PAGE_SHIFT); 13814cb5300bSChris Mason 1382ee22184bSByongho Lee ret = find_new_extents(root, inode, newer_than, 1383ee22184bSByongho Lee &newer_off, SZ_64K); 13844cb5300bSChris Mason if (!ret) { 13854cb5300bSChris Mason range->start = newer_off; 138609cbfeafSKirill A. Shutemov i = (newer_off & new_align) >> PAGE_SHIFT; 13874cb5300bSChris Mason } else { 13884cb5300bSChris Mason break; 1389940100a4SChris Mason } 13904cb5300bSChris Mason } else { 1391008873eaSLi Zefan if (ret > 0) { 1392cbcc8326SLi Zefan i += ret; 139309cbfeafSKirill A. Shutemov last_len += ret << PAGE_SHIFT; 1394008873eaSLi Zefan } else { 1395940100a4SChris Mason i++; 1396008873eaSLi Zefan last_len = 0; 1397008873eaSLi Zefan } 1398f46b5a66SChristoph Hellwig } 13994cb5300bSChris Mason } 1400f46b5a66SChristoph Hellwig 1401dec8ef90SFilipe Manana if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) { 14021e701a32SChris Mason filemap_flush(inode->i_mapping); 1403dec8ef90SFilipe Manana if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT, 1404dec8ef90SFilipe Manana &BTRFS_I(inode)->runtime_flags)) 1405dec8ef90SFilipe Manana filemap_flush(inode->i_mapping); 1406dec8ef90SFilipe Manana } 14071e701a32SChris Mason 14081e2ef46dSDavid Sterba if (do_compress) { 14091e701a32SChris Mason /* the filemap_flush will queue IO into the worker threads, but 14101e701a32SChris Mason * we have to make sure the IO is actually started and that 14111e701a32SChris Mason * ordered extents get created before we return 14121e701a32SChris Mason */ 14130b246afaSJeff Mahoney atomic_inc(&fs_info->async_submit_draining); 14140b246afaSJeff Mahoney while (atomic_read(&fs_info->nr_async_submits) || 14150b246afaSJeff Mahoney atomic_read(&fs_info->async_delalloc_pages)) { 14160b246afaSJeff Mahoney wait_event(fs_info->async_submit_wait, 14170b246afaSJeff Mahoney (atomic_read(&fs_info->nr_async_submits) == 0 && 14180b246afaSJeff Mahoney atomic_read(&fs_info->async_delalloc_pages) == 0)); 14191e701a32SChris Mason } 14200b246afaSJeff Mahoney atomic_dec(&fs_info->async_submit_draining); 14211e701a32SChris Mason } 14221e701a32SChris Mason 14231a419d85SLi Zefan if (range->compress_type == BTRFS_COMPRESS_LZO) { 14240b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, COMPRESS_LZO); 14255c1aab1dSNick Terrell } else if (range->compress_type == BTRFS_COMPRESS_ZSTD) { 14265c1aab1dSNick Terrell btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD); 14271a419d85SLi Zefan } 14281a419d85SLi Zefan 142960ccf82fSDiego Calleja ret = defrag_count; 1430940100a4SChris Mason 14314cb5300bSChris Mason out_ra: 14321e2ef46dSDavid Sterba if (do_compress) { 14335955102cSAl Viro inode_lock(inode); 1434eec63c65SDavid Sterba BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE; 14355955102cSAl Viro inode_unlock(inode); 1436633085c7SFilipe David Borba Manana } 14374cb5300bSChris Mason if (!file) 14384cb5300bSChris Mason kfree(ra); 14394cb5300bSChris Mason kfree(pages); 1440940100a4SChris Mason return ret; 1441f46b5a66SChristoph Hellwig } 1442f46b5a66SChristoph Hellwig 1443198605a8SMiao Xie static noinline int btrfs_ioctl_resize(struct file *file, 144476dda93cSYan, Zheng void __user *arg) 1445f46b5a66SChristoph Hellwig { 14460b246afaSJeff Mahoney struct inode *inode = file_inode(file); 14470b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 1448f46b5a66SChristoph Hellwig u64 new_size; 1449f46b5a66SChristoph Hellwig u64 old_size; 1450f46b5a66SChristoph Hellwig u64 devid = 1; 14510b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 1452f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 1453f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 1454f46b5a66SChristoph Hellwig struct btrfs_device *device = NULL; 1455f46b5a66SChristoph Hellwig char *sizestr; 14569a40f122SGui Hecheng char *retptr; 1457f46b5a66SChristoph Hellwig char *devstr = NULL; 1458f46b5a66SChristoph Hellwig int ret = 0; 1459f46b5a66SChristoph Hellwig int mod = 0; 1460f46b5a66SChristoph Hellwig 1461e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 1462e441d54dSChris Mason return -EPERM; 1463e441d54dSChris Mason 1464198605a8SMiao Xie ret = mnt_want_write_file(file); 1465198605a8SMiao Xie if (ret) 1466198605a8SMiao Xie return ret; 1467198605a8SMiao Xie 1468171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 146997547676SMiao Xie mnt_drop_write_file(file); 1470e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 1471c9e9f97bSIlya Dryomov } 1472c9e9f97bSIlya Dryomov 14730b246afaSJeff Mahoney mutex_lock(&fs_info->volume_mutex); 1474dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1475c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 1476c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 1477c9e9f97bSIlya Dryomov goto out; 1478c9e9f97bSIlya Dryomov } 14795516e595SMark Fasheh 14805516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1481f46b5a66SChristoph Hellwig 1482f46b5a66SChristoph Hellwig sizestr = vol_args->name; 1483f46b5a66SChristoph Hellwig devstr = strchr(sizestr, ':'); 1484f46b5a66SChristoph Hellwig if (devstr) { 1485f46b5a66SChristoph Hellwig sizestr = devstr + 1; 1486f46b5a66SChristoph Hellwig *devstr = '\0'; 1487f46b5a66SChristoph Hellwig devstr = vol_args->name; 148858dfae63SZhangZhen ret = kstrtoull(devstr, 10, &devid); 148958dfae63SZhangZhen if (ret) 149058dfae63SZhangZhen goto out_free; 1491dfd79829SMiao Xie if (!devid) { 1492dfd79829SMiao Xie ret = -EINVAL; 1493dfd79829SMiao Xie goto out_free; 1494dfd79829SMiao Xie } 14950b246afaSJeff Mahoney btrfs_info(fs_info, "resizing devid %llu", devid); 1496f46b5a66SChristoph Hellwig } 1497dba60f3fSMiao Xie 14980b246afaSJeff Mahoney device = btrfs_find_device(fs_info, devid, NULL, NULL); 1499f46b5a66SChristoph Hellwig if (!device) { 15000b246afaSJeff Mahoney btrfs_info(fs_info, "resizer unable to find device %llu", 1501c1c9ff7cSGeert Uytterhoeven devid); 1502dfd79829SMiao Xie ret = -ENODEV; 1503c9e9f97bSIlya Dryomov goto out_free; 1504f46b5a66SChristoph Hellwig } 1505dba60f3fSMiao Xie 1506dba60f3fSMiao Xie if (!device->writeable) { 15070b246afaSJeff Mahoney btrfs_info(fs_info, 1508efe120a0SFrank Holton "resizer unable to apply on readonly device %llu", 1509c1c9ff7cSGeert Uytterhoeven devid); 1510dfd79829SMiao Xie ret = -EPERM; 15114e42ae1bSLiu Bo goto out_free; 15124e42ae1bSLiu Bo } 15134e42ae1bSLiu Bo 1514f46b5a66SChristoph Hellwig if (!strcmp(sizestr, "max")) 1515f46b5a66SChristoph Hellwig new_size = device->bdev->bd_inode->i_size; 1516f46b5a66SChristoph Hellwig else { 1517f46b5a66SChristoph Hellwig if (sizestr[0] == '-') { 1518f46b5a66SChristoph Hellwig mod = -1; 1519f46b5a66SChristoph Hellwig sizestr++; 1520f46b5a66SChristoph Hellwig } else if (sizestr[0] == '+') { 1521f46b5a66SChristoph Hellwig mod = 1; 1522f46b5a66SChristoph Hellwig sizestr++; 1523f46b5a66SChristoph Hellwig } 15249a40f122SGui Hecheng new_size = memparse(sizestr, &retptr); 15259a40f122SGui Hecheng if (*retptr != '\0' || new_size == 0) { 1526f46b5a66SChristoph Hellwig ret = -EINVAL; 1527c9e9f97bSIlya Dryomov goto out_free; 1528f46b5a66SChristoph Hellwig } 1529f46b5a66SChristoph Hellwig } 1530f46b5a66SChristoph Hellwig 153163a212abSStefan Behrens if (device->is_tgtdev_for_dev_replace) { 1532dfd79829SMiao Xie ret = -EPERM; 153363a212abSStefan Behrens goto out_free; 153463a212abSStefan Behrens } 153563a212abSStefan Behrens 15367cc8e58dSMiao Xie old_size = btrfs_device_get_total_bytes(device); 1537f46b5a66SChristoph Hellwig 1538f46b5a66SChristoph Hellwig if (mod < 0) { 1539f46b5a66SChristoph Hellwig if (new_size > old_size) { 1540f46b5a66SChristoph Hellwig ret = -EINVAL; 1541c9e9f97bSIlya Dryomov goto out_free; 1542f46b5a66SChristoph Hellwig } 1543f46b5a66SChristoph Hellwig new_size = old_size - new_size; 1544f46b5a66SChristoph Hellwig } else if (mod > 0) { 1545eb8052e0SWenliang Fan if (new_size > ULLONG_MAX - old_size) { 1546902c68a4SGui Hecheng ret = -ERANGE; 1547eb8052e0SWenliang Fan goto out_free; 1548eb8052e0SWenliang Fan } 1549f46b5a66SChristoph Hellwig new_size = old_size + new_size; 1550f46b5a66SChristoph Hellwig } 1551f46b5a66SChristoph Hellwig 1552ee22184bSByongho Lee if (new_size < SZ_256M) { 1553f46b5a66SChristoph Hellwig ret = -EINVAL; 1554c9e9f97bSIlya Dryomov goto out_free; 1555f46b5a66SChristoph Hellwig } 1556f46b5a66SChristoph Hellwig if (new_size > device->bdev->bd_inode->i_size) { 1557f46b5a66SChristoph Hellwig ret = -EFBIG; 1558c9e9f97bSIlya Dryomov goto out_free; 1559f46b5a66SChristoph Hellwig } 1560f46b5a66SChristoph Hellwig 156147f08b96SNikolay Borisov new_size = round_down(new_size, fs_info->sectorsize); 1562f46b5a66SChristoph Hellwig 15630b246afaSJeff Mahoney btrfs_info_in_rcu(fs_info, "new size for %s is %llu", 1564c1c9ff7cSGeert Uytterhoeven rcu_str_deref(device->name), new_size); 1565f46b5a66SChristoph Hellwig 1566f46b5a66SChristoph Hellwig if (new_size > old_size) { 1567a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 156898d5dc13STsutomu Itoh if (IS_ERR(trans)) { 156998d5dc13STsutomu Itoh ret = PTR_ERR(trans); 1570c9e9f97bSIlya Dryomov goto out_free; 157198d5dc13STsutomu Itoh } 1572f46b5a66SChristoph Hellwig ret = btrfs_grow_device(trans, device, new_size); 15733a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 1574ece7d20eSMike Fleetwood } else if (new_size < old_size) { 1575f46b5a66SChristoph Hellwig ret = btrfs_shrink_device(device, new_size); 15760253f40eSjeff.liu } /* equal, nothing need to do */ 1577f46b5a66SChristoph Hellwig 1578c9e9f97bSIlya Dryomov out_free: 1579f46b5a66SChristoph Hellwig kfree(vol_args); 1580c9e9f97bSIlya Dryomov out: 15810b246afaSJeff Mahoney mutex_unlock(&fs_info->volume_mutex); 1582171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 158318f39c41SIlya Dryomov mnt_drop_write_file(file); 1584f46b5a66SChristoph Hellwig return ret; 1585f46b5a66SChristoph Hellwig } 1586f46b5a66SChristoph Hellwig 158772fd032eSSage Weil static noinline int btrfs_ioctl_snap_create_transid(struct file *file, 158852f75f4fSDavid Sterba const char *name, unsigned long fd, int subvol, 15896f72c7e2SArne Jansen u64 *transid, bool readonly, 15908696c533SMiao Xie struct btrfs_qgroup_inherit *inherit) 1591f46b5a66SChristoph Hellwig { 1592f46b5a66SChristoph Hellwig int namelen; 15933de4586cSChris Mason int ret = 0; 1594f46b5a66SChristoph Hellwig 1595325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1596325c50e3SJeff Mahoney return -ENOTDIR; 1597325c50e3SJeff Mahoney 1598a874a63eSLiu Bo ret = mnt_want_write_file(file); 1599a874a63eSLiu Bo if (ret) 1600a874a63eSLiu Bo goto out; 1601a874a63eSLiu Bo 160272fd032eSSage Weil namelen = strlen(name); 160372fd032eSSage Weil if (strchr(name, '/')) { 1604f46b5a66SChristoph Hellwig ret = -EINVAL; 1605a874a63eSLiu Bo goto out_drop_write; 1606f46b5a66SChristoph Hellwig } 1607f46b5a66SChristoph Hellwig 160816780cabSChris Mason if (name[0] == '.' && 160916780cabSChris Mason (namelen == 1 || (name[1] == '.' && namelen == 2))) { 161016780cabSChris Mason ret = -EEXIST; 1611a874a63eSLiu Bo goto out_drop_write; 161216780cabSChris Mason } 161316780cabSChris Mason 16143de4586cSChris Mason if (subvol) { 161572fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 16166f72c7e2SArne Jansen NULL, transid, readonly, inherit); 1617cb8e7090SChristoph Hellwig } else { 16182903ff01SAl Viro struct fd src = fdget(fd); 16193de4586cSChris Mason struct inode *src_inode; 16202903ff01SAl Viro if (!src.file) { 16213de4586cSChris Mason ret = -EINVAL; 1622a874a63eSLiu Bo goto out_drop_write; 16233de4586cSChris Mason } 16243de4586cSChris Mason 1625496ad9aaSAl Viro src_inode = file_inode(src.file); 1626496ad9aaSAl Viro if (src_inode->i_sb != file_inode(file)->i_sb) { 1627c79b4713SJosef Bacik btrfs_info(BTRFS_I(file_inode(file))->root->fs_info, 1628efe120a0SFrank Holton "Snapshot src from another FS"); 162923ad5b17SKusanagi Kouichi ret = -EXDEV; 1630d0242061SDavid Sterba } else if (!inode_owner_or_capable(src_inode)) { 1631d0242061SDavid Sterba /* 1632d0242061SDavid Sterba * Subvolume creation is not restricted, but snapshots 1633d0242061SDavid Sterba * are limited to own subvolumes only 1634d0242061SDavid Sterba */ 1635d0242061SDavid Sterba ret = -EPERM; 1636ecd18815SAl Viro } else { 163772fd032eSSage Weil ret = btrfs_mksubvol(&file->f_path, name, namelen, 163872fd032eSSage Weil BTRFS_I(src_inode)->root, 16396f72c7e2SArne Jansen transid, readonly, inherit); 1640ecd18815SAl Viro } 16412903ff01SAl Viro fdput(src); 1642cb8e7090SChristoph Hellwig } 1643a874a63eSLiu Bo out_drop_write: 1644a874a63eSLiu Bo mnt_drop_write_file(file); 1645f46b5a66SChristoph Hellwig out: 164672fd032eSSage Weil return ret; 164772fd032eSSage Weil } 164872fd032eSSage Weil 164972fd032eSSage Weil static noinline int btrfs_ioctl_snap_create(struct file *file, 1650fa0d2b9bSLi Zefan void __user *arg, int subvol) 165172fd032eSSage Weil { 1652fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args *vol_args; 165372fd032eSSage Weil int ret; 165472fd032eSSage Weil 1655325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1656325c50e3SJeff Mahoney return -ENOTDIR; 1657325c50e3SJeff Mahoney 1658fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1659fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1660fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1661fa0d2b9bSLi Zefan vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 1662fa0d2b9bSLi Zefan 1663fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 1664b83cc969SLi Zefan vol_args->fd, subvol, 16656f72c7e2SArne Jansen NULL, false, NULL); 1666fa0d2b9bSLi Zefan 1667fa0d2b9bSLi Zefan kfree(vol_args); 1668fa0d2b9bSLi Zefan return ret; 1669fa0d2b9bSLi Zefan } 1670fa0d2b9bSLi Zefan 1671fa0d2b9bSLi Zefan static noinline int btrfs_ioctl_snap_create_v2(struct file *file, 1672fa0d2b9bSLi Zefan void __user *arg, int subvol) 1673fa0d2b9bSLi Zefan { 1674fa0d2b9bSLi Zefan struct btrfs_ioctl_vol_args_v2 *vol_args; 1675fa0d2b9bSLi Zefan int ret; 1676fdfb1e4fSLi Zefan u64 transid = 0; 1677fdfb1e4fSLi Zefan u64 *ptr = NULL; 1678b83cc969SLi Zefan bool readonly = false; 16796f72c7e2SArne Jansen struct btrfs_qgroup_inherit *inherit = NULL; 168072fd032eSSage Weil 1681325c50e3SJeff Mahoney if (!S_ISDIR(file_inode(file)->i_mode)) 1682325c50e3SJeff Mahoney return -ENOTDIR; 1683325c50e3SJeff Mahoney 1684fa0d2b9bSLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 1685fa0d2b9bSLi Zefan if (IS_ERR(vol_args)) 1686fa0d2b9bSLi Zefan return PTR_ERR(vol_args); 1687fa0d2b9bSLi Zefan vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 1688fdfb1e4fSLi Zefan 1689b83cc969SLi Zefan if (vol_args->flags & 16906f72c7e2SArne Jansen ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY | 16916f72c7e2SArne Jansen BTRFS_SUBVOL_QGROUP_INHERIT)) { 1692b83cc969SLi Zefan ret = -EOPNOTSUPP; 1693c47ca32dSDan Carpenter goto free_args; 1694fdfb1e4fSLi Zefan } 1695fdfb1e4fSLi Zefan 1696fa0d2b9bSLi Zefan if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC) 1697fdfb1e4fSLi Zefan ptr = &transid; 1698b83cc969SLi Zefan if (vol_args->flags & BTRFS_SUBVOL_RDONLY) 1699b83cc969SLi Zefan readonly = true; 17006f72c7e2SArne Jansen if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) { 170109cbfeafSKirill A. Shutemov if (vol_args->size > PAGE_SIZE) { 17026f72c7e2SArne Jansen ret = -EINVAL; 1703c47ca32dSDan Carpenter goto free_args; 17046f72c7e2SArne Jansen } 17056f72c7e2SArne Jansen inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size); 17066f72c7e2SArne Jansen if (IS_ERR(inherit)) { 17076f72c7e2SArne Jansen ret = PTR_ERR(inherit); 1708c47ca32dSDan Carpenter goto free_args; 17096f72c7e2SArne Jansen } 17106f72c7e2SArne Jansen } 171175eaa0e2SSage Weil 1712fa0d2b9bSLi Zefan ret = btrfs_ioctl_snap_create_transid(file, vol_args->name, 17136f72c7e2SArne Jansen vol_args->fd, subvol, ptr, 17148696c533SMiao Xie readonly, inherit); 1715c47ca32dSDan Carpenter if (ret) 1716c47ca32dSDan Carpenter goto free_inherit; 171775eaa0e2SSage Weil 1718c47ca32dSDan Carpenter if (ptr && copy_to_user(arg + 1719fdfb1e4fSLi Zefan offsetof(struct btrfs_ioctl_vol_args_v2, 1720c47ca32dSDan Carpenter transid), 1721c47ca32dSDan Carpenter ptr, sizeof(*ptr))) 172275eaa0e2SSage Weil ret = -EFAULT; 1723c47ca32dSDan Carpenter 1724c47ca32dSDan Carpenter free_inherit: 17256f72c7e2SArne Jansen kfree(inherit); 1726c47ca32dSDan Carpenter free_args: 1727c47ca32dSDan Carpenter kfree(vol_args); 1728f46b5a66SChristoph Hellwig return ret; 1729f46b5a66SChristoph Hellwig } 1730f46b5a66SChristoph Hellwig 17310caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_getflags(struct file *file, 17320caa102dSLi Zefan void __user *arg) 17330caa102dSLi Zefan { 1734496ad9aaSAl Viro struct inode *inode = file_inode(file); 17350b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 17360caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 17370caa102dSLi Zefan int ret = 0; 17380caa102dSLi Zefan u64 flags = 0; 17390caa102dSLi Zefan 17404a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) 17410caa102dSLi Zefan return -EINVAL; 17420caa102dSLi Zefan 17430b246afaSJeff Mahoney down_read(&fs_info->subvol_sem); 17440caa102dSLi Zefan if (btrfs_root_readonly(root)) 17450caa102dSLi Zefan flags |= BTRFS_SUBVOL_RDONLY; 17460b246afaSJeff Mahoney up_read(&fs_info->subvol_sem); 17470caa102dSLi Zefan 17480caa102dSLi Zefan if (copy_to_user(arg, &flags, sizeof(flags))) 17490caa102dSLi Zefan ret = -EFAULT; 17500caa102dSLi Zefan 17510caa102dSLi Zefan return ret; 17520caa102dSLi Zefan } 17530caa102dSLi Zefan 17540caa102dSLi Zefan static noinline int btrfs_ioctl_subvol_setflags(struct file *file, 17550caa102dSLi Zefan void __user *arg) 17560caa102dSLi Zefan { 1757496ad9aaSAl Viro struct inode *inode = file_inode(file); 17580b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 17590caa102dSLi Zefan struct btrfs_root *root = BTRFS_I(inode)->root; 17600caa102dSLi Zefan struct btrfs_trans_handle *trans; 17610caa102dSLi Zefan u64 root_flags; 17620caa102dSLi Zefan u64 flags; 17630caa102dSLi Zefan int ret = 0; 17640caa102dSLi Zefan 1765bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 1766bd60ea0fSDavid Sterba return -EPERM; 1767bd60ea0fSDavid Sterba 1768b9ca0664SLiu Bo ret = mnt_want_write_file(file); 1769b9ca0664SLiu Bo if (ret) 1770b9ca0664SLiu Bo goto out; 17710caa102dSLi Zefan 17724a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 1773b9ca0664SLiu Bo ret = -EINVAL; 1774b9ca0664SLiu Bo goto out_drop_write; 1775b9ca0664SLiu Bo } 17760caa102dSLi Zefan 1777b9ca0664SLiu Bo if (copy_from_user(&flags, arg, sizeof(flags))) { 1778b9ca0664SLiu Bo ret = -EFAULT; 1779b9ca0664SLiu Bo goto out_drop_write; 1780b9ca0664SLiu Bo } 17810caa102dSLi Zefan 1782b9ca0664SLiu Bo if (flags & BTRFS_SUBVOL_CREATE_ASYNC) { 1783b9ca0664SLiu Bo ret = -EINVAL; 1784b9ca0664SLiu Bo goto out_drop_write; 1785b9ca0664SLiu Bo } 17860caa102dSLi Zefan 1787b9ca0664SLiu Bo if (flags & ~BTRFS_SUBVOL_RDONLY) { 1788b9ca0664SLiu Bo ret = -EOPNOTSUPP; 1789b9ca0664SLiu Bo goto out_drop_write; 1790b9ca0664SLiu Bo } 17910caa102dSLi Zefan 17920b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 17930caa102dSLi Zefan 17940caa102dSLi Zefan /* nothing to do */ 17950caa102dSLi Zefan if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root)) 1796b9ca0664SLiu Bo goto out_drop_sem; 17970caa102dSLi Zefan 17980caa102dSLi Zefan root_flags = btrfs_root_flags(&root->root_item); 17992c686537SDavid Sterba if (flags & BTRFS_SUBVOL_RDONLY) { 18000caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 18010caa102dSLi Zefan root_flags | BTRFS_ROOT_SUBVOL_RDONLY); 18022c686537SDavid Sterba } else { 18032c686537SDavid Sterba /* 18042c686537SDavid Sterba * Block RO -> RW transition if this subvolume is involved in 18052c686537SDavid Sterba * send 18062c686537SDavid Sterba */ 18072c686537SDavid Sterba spin_lock(&root->root_item_lock); 18082c686537SDavid Sterba if (root->send_in_progress == 0) { 18090caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, 18100caa102dSLi Zefan root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY); 18112c686537SDavid Sterba spin_unlock(&root->root_item_lock); 18122c686537SDavid Sterba } else { 18132c686537SDavid Sterba spin_unlock(&root->root_item_lock); 18140b246afaSJeff Mahoney btrfs_warn(fs_info, 18152c686537SDavid Sterba "Attempt to set subvolume %llu read-write during send", 18162c686537SDavid Sterba root->root_key.objectid); 18172c686537SDavid Sterba ret = -EPERM; 18182c686537SDavid Sterba goto out_drop_sem; 18192c686537SDavid Sterba } 18202c686537SDavid Sterba } 18210caa102dSLi Zefan 18220caa102dSLi Zefan trans = btrfs_start_transaction(root, 1); 18230caa102dSLi Zefan if (IS_ERR(trans)) { 18240caa102dSLi Zefan ret = PTR_ERR(trans); 18250caa102dSLi Zefan goto out_reset; 18260caa102dSLi Zefan } 18270caa102dSLi Zefan 18280b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 18290caa102dSLi Zefan &root->root_key, &root->root_item); 18309417ebc8SNikolay Borisov if (ret < 0) { 18319417ebc8SNikolay Borisov btrfs_end_transaction(trans); 18329417ebc8SNikolay Borisov goto out_reset; 18339417ebc8SNikolay Borisov } 18340caa102dSLi Zefan 18359417ebc8SNikolay Borisov ret = btrfs_commit_transaction(trans); 18369417ebc8SNikolay Borisov 18370caa102dSLi Zefan out_reset: 18380caa102dSLi Zefan if (ret) 18390caa102dSLi Zefan btrfs_set_root_flags(&root->root_item, root_flags); 1840b9ca0664SLiu Bo out_drop_sem: 18410b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 1842b9ca0664SLiu Bo out_drop_write: 1843b9ca0664SLiu Bo mnt_drop_write_file(file); 1844b9ca0664SLiu Bo out: 18450caa102dSLi Zefan return ret; 18460caa102dSLi Zefan } 18470caa102dSLi Zefan 184876dda93cSYan, Zheng /* 184976dda93cSYan, Zheng * helper to check if the subvolume references other subvolumes 185076dda93cSYan, Zheng */ 185176dda93cSYan, Zheng static noinline int may_destroy_subvol(struct btrfs_root *root) 185276dda93cSYan, Zheng { 18530b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 185476dda93cSYan, Zheng struct btrfs_path *path; 1855175a2b87SJosef Bacik struct btrfs_dir_item *di; 185676dda93cSYan, Zheng struct btrfs_key key; 1857175a2b87SJosef Bacik u64 dir_id; 185876dda93cSYan, Zheng int ret; 185976dda93cSYan, Zheng 186076dda93cSYan, Zheng path = btrfs_alloc_path(); 186176dda93cSYan, Zheng if (!path) 186276dda93cSYan, Zheng return -ENOMEM; 186376dda93cSYan, Zheng 1864175a2b87SJosef Bacik /* Make sure this root isn't set as the default subvol */ 18650b246afaSJeff Mahoney dir_id = btrfs_super_root_dir(fs_info->super_copy); 18660b246afaSJeff Mahoney di = btrfs_lookup_dir_item(NULL, fs_info->tree_root, path, 1867175a2b87SJosef Bacik dir_id, "default", 7, 0); 1868175a2b87SJosef Bacik if (di && !IS_ERR(di)) { 1869175a2b87SJosef Bacik btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key); 1870175a2b87SJosef Bacik if (key.objectid == root->root_key.objectid) { 187172de6b53SGuangyu Sun ret = -EPERM; 18720b246afaSJeff Mahoney btrfs_err(fs_info, 18735d163e0eSJeff Mahoney "deleting default subvolume %llu is not allowed", 18745d163e0eSJeff Mahoney key.objectid); 1875175a2b87SJosef Bacik goto out; 1876175a2b87SJosef Bacik } 1877175a2b87SJosef Bacik btrfs_release_path(path); 1878175a2b87SJosef Bacik } 1879175a2b87SJosef Bacik 188076dda93cSYan, Zheng key.objectid = root->root_key.objectid; 188176dda93cSYan, Zheng key.type = BTRFS_ROOT_REF_KEY; 188276dda93cSYan, Zheng key.offset = (u64)-1; 188376dda93cSYan, Zheng 18840b246afaSJeff Mahoney ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 188576dda93cSYan, Zheng if (ret < 0) 188676dda93cSYan, Zheng goto out; 188776dda93cSYan, Zheng BUG_ON(ret == 0); 188876dda93cSYan, Zheng 188976dda93cSYan, Zheng ret = 0; 189076dda93cSYan, Zheng if (path->slots[0] > 0) { 189176dda93cSYan, Zheng path->slots[0]--; 189276dda93cSYan, Zheng btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 189376dda93cSYan, Zheng if (key.objectid == root->root_key.objectid && 189476dda93cSYan, Zheng key.type == BTRFS_ROOT_REF_KEY) 189576dda93cSYan, Zheng ret = -ENOTEMPTY; 189676dda93cSYan, Zheng } 189776dda93cSYan, Zheng out: 189876dda93cSYan, Zheng btrfs_free_path(path); 189976dda93cSYan, Zheng return ret; 190076dda93cSYan, Zheng } 190176dda93cSYan, Zheng 1902ac8e9819SChris Mason static noinline int key_in_sk(struct btrfs_key *key, 1903ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk) 1904ac8e9819SChris Mason { 1905abc6e134SChris Mason struct btrfs_key test; 1906abc6e134SChris Mason int ret; 1907abc6e134SChris Mason 1908abc6e134SChris Mason test.objectid = sk->min_objectid; 1909abc6e134SChris Mason test.type = sk->min_type; 1910abc6e134SChris Mason test.offset = sk->min_offset; 1911abc6e134SChris Mason 1912abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1913abc6e134SChris Mason if (ret < 0) 1914ac8e9819SChris Mason return 0; 1915abc6e134SChris Mason 1916abc6e134SChris Mason test.objectid = sk->max_objectid; 1917abc6e134SChris Mason test.type = sk->max_type; 1918abc6e134SChris Mason test.offset = sk->max_offset; 1919abc6e134SChris Mason 1920abc6e134SChris Mason ret = btrfs_comp_cpu_keys(key, &test); 1921abc6e134SChris Mason if (ret > 0) 1922ac8e9819SChris Mason return 0; 1923ac8e9819SChris Mason return 1; 1924ac8e9819SChris Mason } 1925ac8e9819SChris Mason 1926df397565SJeff Mahoney static noinline int copy_to_sk(struct btrfs_path *path, 1927ac8e9819SChris Mason struct btrfs_key *key, 1928ac8e9819SChris Mason struct btrfs_ioctl_search_key *sk, 19299b6e817dSGerhard Heift size_t *buf_size, 1930ba346b35SGerhard Heift char __user *ubuf, 1931ac8e9819SChris Mason unsigned long *sk_offset, 1932ac8e9819SChris Mason int *num_found) 1933ac8e9819SChris Mason { 1934ac8e9819SChris Mason u64 found_transid; 1935ac8e9819SChris Mason struct extent_buffer *leaf; 1936ac8e9819SChris Mason struct btrfs_ioctl_search_header sh; 1937dd81d459SNaohiro Aota struct btrfs_key test; 1938ac8e9819SChris Mason unsigned long item_off; 1939ac8e9819SChris Mason unsigned long item_len; 1940ac8e9819SChris Mason int nritems; 1941ac8e9819SChris Mason int i; 1942ac8e9819SChris Mason int slot; 1943ac8e9819SChris Mason int ret = 0; 1944ac8e9819SChris Mason 1945ac8e9819SChris Mason leaf = path->nodes[0]; 1946ac8e9819SChris Mason slot = path->slots[0]; 1947ac8e9819SChris Mason nritems = btrfs_header_nritems(leaf); 1948ac8e9819SChris Mason 1949ac8e9819SChris Mason if (btrfs_header_generation(leaf) > sk->max_transid) { 1950ac8e9819SChris Mason i = nritems; 1951ac8e9819SChris Mason goto advance_key; 1952ac8e9819SChris Mason } 1953ac8e9819SChris Mason found_transid = btrfs_header_generation(leaf); 1954ac8e9819SChris Mason 1955ac8e9819SChris Mason for (i = slot; i < nritems; i++) { 1956ac8e9819SChris Mason item_off = btrfs_item_ptr_offset(leaf, i); 1957ac8e9819SChris Mason item_len = btrfs_item_size_nr(leaf, i); 1958ac8e9819SChris Mason 195903b71c6cSGabriel de Perthuis btrfs_item_key_to_cpu(leaf, key, i); 196003b71c6cSGabriel de Perthuis if (!key_in_sk(key, sk)) 196103b71c6cSGabriel de Perthuis continue; 196203b71c6cSGabriel de Perthuis 19639b6e817dSGerhard Heift if (sizeof(sh) + item_len > *buf_size) { 19648f5f6178SGerhard Heift if (*num_found) { 1965ac8e9819SChris Mason ret = 1; 19668f5f6178SGerhard Heift goto out; 19678f5f6178SGerhard Heift } 19688f5f6178SGerhard Heift 19698f5f6178SGerhard Heift /* 19708f5f6178SGerhard Heift * return one empty item back for v1, which does not 19718f5f6178SGerhard Heift * handle -EOVERFLOW 19728f5f6178SGerhard Heift */ 19738f5f6178SGerhard Heift 19749b6e817dSGerhard Heift *buf_size = sizeof(sh) + item_len; 1975ac8e9819SChris Mason item_len = 0; 19768f5f6178SGerhard Heift ret = -EOVERFLOW; 19778f5f6178SGerhard Heift } 1978ac8e9819SChris Mason 19799b6e817dSGerhard Heift if (sizeof(sh) + item_len + *sk_offset > *buf_size) { 1980ac8e9819SChris Mason ret = 1; 198125c9bc2eSGerhard Heift goto out; 1982ac8e9819SChris Mason } 1983ac8e9819SChris Mason 1984ac8e9819SChris Mason sh.objectid = key->objectid; 1985ac8e9819SChris Mason sh.offset = key->offset; 1986ac8e9819SChris Mason sh.type = key->type; 1987ac8e9819SChris Mason sh.len = item_len; 1988ac8e9819SChris Mason sh.transid = found_transid; 1989ac8e9819SChris Mason 1990ac8e9819SChris Mason /* copy search result header */ 1991ba346b35SGerhard Heift if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) { 1992ba346b35SGerhard Heift ret = -EFAULT; 1993ba346b35SGerhard Heift goto out; 1994ba346b35SGerhard Heift } 1995ba346b35SGerhard Heift 1996ac8e9819SChris Mason *sk_offset += sizeof(sh); 1997ac8e9819SChris Mason 1998ac8e9819SChris Mason if (item_len) { 1999ba346b35SGerhard Heift char __user *up = ubuf + *sk_offset; 2000ac8e9819SChris Mason /* copy the item */ 2001ba346b35SGerhard Heift if (read_extent_buffer_to_user(leaf, up, 2002ba346b35SGerhard Heift item_off, item_len)) { 2003ba346b35SGerhard Heift ret = -EFAULT; 2004ba346b35SGerhard Heift goto out; 2005ba346b35SGerhard Heift } 2006ba346b35SGerhard Heift 2007ac8e9819SChris Mason *sk_offset += item_len; 2008ac8e9819SChris Mason } 2009e2156867SHugo Mills (*num_found)++; 2010ac8e9819SChris Mason 20118f5f6178SGerhard Heift if (ret) /* -EOVERFLOW from above */ 20128f5f6178SGerhard Heift goto out; 20138f5f6178SGerhard Heift 201425c9bc2eSGerhard Heift if (*num_found >= sk->nr_items) { 201525c9bc2eSGerhard Heift ret = 1; 201625c9bc2eSGerhard Heift goto out; 201725c9bc2eSGerhard Heift } 2018ac8e9819SChris Mason } 2019ac8e9819SChris Mason advance_key: 2020ac8e9819SChris Mason ret = 0; 2021dd81d459SNaohiro Aota test.objectid = sk->max_objectid; 2022dd81d459SNaohiro Aota test.type = sk->max_type; 2023dd81d459SNaohiro Aota test.offset = sk->max_offset; 2024dd81d459SNaohiro Aota if (btrfs_comp_cpu_keys(key, &test) >= 0) 2025dd81d459SNaohiro Aota ret = 1; 2026dd81d459SNaohiro Aota else if (key->offset < (u64)-1) 2027abc6e134SChris Mason key->offset++; 2028dd81d459SNaohiro Aota else if (key->type < (u8)-1) { 2029abc6e134SChris Mason key->offset = 0; 2030abc6e134SChris Mason key->type++; 2031dd81d459SNaohiro Aota } else if (key->objectid < (u64)-1) { 2032abc6e134SChris Mason key->offset = 0; 2033abc6e134SChris Mason key->type = 0; 2034abc6e134SChris Mason key->objectid++; 2035abc6e134SChris Mason } else 2036abc6e134SChris Mason ret = 1; 203725c9bc2eSGerhard Heift out: 2038ba346b35SGerhard Heift /* 2039ba346b35SGerhard Heift * 0: all items from this leaf copied, continue with next 2040ba346b35SGerhard Heift * 1: * more items can be copied, but unused buffer is too small 2041ba346b35SGerhard Heift * * all items were found 2042ba346b35SGerhard Heift * Either way, it will stops the loop which iterates to the next 2043ba346b35SGerhard Heift * leaf 2044ba346b35SGerhard Heift * -EOVERFLOW: item was to large for buffer 2045ba346b35SGerhard Heift * -EFAULT: could not copy extent buffer back to userspace 2046ba346b35SGerhard Heift */ 2047ac8e9819SChris Mason return ret; 2048ac8e9819SChris Mason } 2049ac8e9819SChris Mason 2050ac8e9819SChris Mason static noinline int search_ioctl(struct inode *inode, 205112544442SGerhard Heift struct btrfs_ioctl_search_key *sk, 20529b6e817dSGerhard Heift size_t *buf_size, 2053ba346b35SGerhard Heift char __user *ubuf) 2054ac8e9819SChris Mason { 20550b246afaSJeff Mahoney struct btrfs_fs_info *info = btrfs_sb(inode->i_sb); 2056ac8e9819SChris Mason struct btrfs_root *root; 2057ac8e9819SChris Mason struct btrfs_key key; 2058ac8e9819SChris Mason struct btrfs_path *path; 2059ac8e9819SChris Mason int ret; 2060ac8e9819SChris Mason int num_found = 0; 2061ac8e9819SChris Mason unsigned long sk_offset = 0; 2062ac8e9819SChris Mason 20639b6e817dSGerhard Heift if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) { 20649b6e817dSGerhard Heift *buf_size = sizeof(struct btrfs_ioctl_search_header); 206512544442SGerhard Heift return -EOVERFLOW; 20669b6e817dSGerhard Heift } 206712544442SGerhard Heift 2068ac8e9819SChris Mason path = btrfs_alloc_path(); 2069ac8e9819SChris Mason if (!path) 2070ac8e9819SChris Mason return -ENOMEM; 2071ac8e9819SChris Mason 2072ac8e9819SChris Mason if (sk->tree_id == 0) { 2073ac8e9819SChris Mason /* search the root of the inode that was passed */ 2074ac8e9819SChris Mason root = BTRFS_I(inode)->root; 2075ac8e9819SChris Mason } else { 2076ac8e9819SChris Mason key.objectid = sk->tree_id; 2077ac8e9819SChris Mason key.type = BTRFS_ROOT_ITEM_KEY; 2078ac8e9819SChris Mason key.offset = (u64)-1; 2079ac8e9819SChris Mason root = btrfs_read_fs_root_no_name(info, &key); 2080ac8e9819SChris Mason if (IS_ERR(root)) { 2081ac8e9819SChris Mason btrfs_free_path(path); 2082ac8e9819SChris Mason return -ENOENT; 2083ac8e9819SChris Mason } 2084ac8e9819SChris Mason } 2085ac8e9819SChris Mason 2086ac8e9819SChris Mason key.objectid = sk->min_objectid; 2087ac8e9819SChris Mason key.type = sk->min_type; 2088ac8e9819SChris Mason key.offset = sk->min_offset; 2089ac8e9819SChris Mason 2090ac8e9819SChris Mason while (1) { 20916174d3cbSFilipe David Borba Manana ret = btrfs_search_forward(root, &key, path, sk->min_transid); 2092ac8e9819SChris Mason if (ret != 0) { 2093ac8e9819SChris Mason if (ret > 0) 2094ac8e9819SChris Mason ret = 0; 2095ac8e9819SChris Mason goto err; 2096ac8e9819SChris Mason } 2097df397565SJeff Mahoney ret = copy_to_sk(path, &key, sk, buf_size, ubuf, 2098ac8e9819SChris Mason &sk_offset, &num_found); 2099b3b4aa74SDavid Sterba btrfs_release_path(path); 210025c9bc2eSGerhard Heift if (ret) 2101ac8e9819SChris Mason break; 2102ac8e9819SChris Mason 2103ac8e9819SChris Mason } 21048f5f6178SGerhard Heift if (ret > 0) 2105ac8e9819SChris Mason ret = 0; 2106ac8e9819SChris Mason err: 2107ac8e9819SChris Mason sk->nr_items = num_found; 2108ac8e9819SChris Mason btrfs_free_path(path); 2109ac8e9819SChris Mason return ret; 2110ac8e9819SChris Mason } 2111ac8e9819SChris Mason 2112ac8e9819SChris Mason static noinline int btrfs_ioctl_tree_search(struct file *file, 2113ac8e9819SChris Mason void __user *argp) 2114ac8e9819SChris Mason { 2115ba346b35SGerhard Heift struct btrfs_ioctl_search_args __user *uargs; 2116ba346b35SGerhard Heift struct btrfs_ioctl_search_key sk; 2117ac8e9819SChris Mason struct inode *inode; 2118ac8e9819SChris Mason int ret; 21199b6e817dSGerhard Heift size_t buf_size; 2120ac8e9819SChris Mason 2121ac8e9819SChris Mason if (!capable(CAP_SYS_ADMIN)) 2122ac8e9819SChris Mason return -EPERM; 2123ac8e9819SChris Mason 2124ba346b35SGerhard Heift uargs = (struct btrfs_ioctl_search_args __user *)argp; 2125ac8e9819SChris Mason 2126ba346b35SGerhard Heift if (copy_from_user(&sk, &uargs->key, sizeof(sk))) 2127ba346b35SGerhard Heift return -EFAULT; 2128ba346b35SGerhard Heift 2129ba346b35SGerhard Heift buf_size = sizeof(uargs->buf); 2130ac8e9819SChris Mason 2131496ad9aaSAl Viro inode = file_inode(file); 2132ba346b35SGerhard Heift ret = search_ioctl(inode, &sk, &buf_size, uargs->buf); 21338f5f6178SGerhard Heift 21348f5f6178SGerhard Heift /* 21358f5f6178SGerhard Heift * In the origin implementation an overflow is handled by returning a 21368f5f6178SGerhard Heift * search header with a len of zero, so reset ret. 21378f5f6178SGerhard Heift */ 21388f5f6178SGerhard Heift if (ret == -EOVERFLOW) 21398f5f6178SGerhard Heift ret = 0; 21408f5f6178SGerhard Heift 2141ba346b35SGerhard Heift if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk))) 2142ac8e9819SChris Mason ret = -EFAULT; 2143ac8e9819SChris Mason return ret; 2144ac8e9819SChris Mason } 2145ac8e9819SChris Mason 2146cc68a8a5SGerhard Heift static noinline int btrfs_ioctl_tree_search_v2(struct file *file, 2147cc68a8a5SGerhard Heift void __user *argp) 2148cc68a8a5SGerhard Heift { 2149cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 __user *uarg; 2150cc68a8a5SGerhard Heift struct btrfs_ioctl_search_args_v2 args; 2151cc68a8a5SGerhard Heift struct inode *inode; 2152cc68a8a5SGerhard Heift int ret; 2153cc68a8a5SGerhard Heift size_t buf_size; 2154ee22184bSByongho Lee const size_t buf_limit = SZ_16M; 2155cc68a8a5SGerhard Heift 2156cc68a8a5SGerhard Heift if (!capable(CAP_SYS_ADMIN)) 2157cc68a8a5SGerhard Heift return -EPERM; 2158cc68a8a5SGerhard Heift 2159cc68a8a5SGerhard Heift /* copy search header and buffer size */ 2160cc68a8a5SGerhard Heift uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp; 2161cc68a8a5SGerhard Heift if (copy_from_user(&args, uarg, sizeof(args))) 2162cc68a8a5SGerhard Heift return -EFAULT; 2163cc68a8a5SGerhard Heift 2164cc68a8a5SGerhard Heift buf_size = args.buf_size; 2165cc68a8a5SGerhard Heift 2166cc68a8a5SGerhard Heift /* limit result size to 16MB */ 2167cc68a8a5SGerhard Heift if (buf_size > buf_limit) 2168cc68a8a5SGerhard Heift buf_size = buf_limit; 2169cc68a8a5SGerhard Heift 2170cc68a8a5SGerhard Heift inode = file_inode(file); 2171cc68a8a5SGerhard Heift ret = search_ioctl(inode, &args.key, &buf_size, 2172718dc5faSOmar Sandoval (char __user *)(&uarg->buf[0])); 2173cc68a8a5SGerhard Heift if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key))) 2174cc68a8a5SGerhard Heift ret = -EFAULT; 2175cc68a8a5SGerhard Heift else if (ret == -EOVERFLOW && 2176cc68a8a5SGerhard Heift copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size))) 2177cc68a8a5SGerhard Heift ret = -EFAULT; 2178cc68a8a5SGerhard Heift 217976dda93cSYan, Zheng return ret; 218076dda93cSYan, Zheng } 218176dda93cSYan, Zheng 218298d377a0STARUISI Hiroaki /* 2183ac8e9819SChris Mason * Search INODE_REFs to identify path name of 'dirid' directory 2184ac8e9819SChris Mason * in a 'tree_id' tree. and sets path name to 'name'. 218598d377a0STARUISI Hiroaki */ 218698d377a0STARUISI Hiroaki static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info, 218798d377a0STARUISI Hiroaki u64 tree_id, u64 dirid, char *name) 218898d377a0STARUISI Hiroaki { 218998d377a0STARUISI Hiroaki struct btrfs_root *root; 219098d377a0STARUISI Hiroaki struct btrfs_key key; 2191ac8e9819SChris Mason char *ptr; 219298d377a0STARUISI Hiroaki int ret = -1; 219398d377a0STARUISI Hiroaki int slot; 219498d377a0STARUISI Hiroaki int len; 219598d377a0STARUISI Hiroaki int total_len = 0; 219698d377a0STARUISI Hiroaki struct btrfs_inode_ref *iref; 219798d377a0STARUISI Hiroaki struct extent_buffer *l; 219898d377a0STARUISI Hiroaki struct btrfs_path *path; 219998d377a0STARUISI Hiroaki 220098d377a0STARUISI Hiroaki if (dirid == BTRFS_FIRST_FREE_OBJECTID) { 220198d377a0STARUISI Hiroaki name[0]='\0'; 220298d377a0STARUISI Hiroaki return 0; 220398d377a0STARUISI Hiroaki } 220498d377a0STARUISI Hiroaki 220598d377a0STARUISI Hiroaki path = btrfs_alloc_path(); 220698d377a0STARUISI Hiroaki if (!path) 220798d377a0STARUISI Hiroaki return -ENOMEM; 220898d377a0STARUISI Hiroaki 2209ac8e9819SChris Mason ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX]; 221098d377a0STARUISI Hiroaki 221198d377a0STARUISI Hiroaki key.objectid = tree_id; 221298d377a0STARUISI Hiroaki key.type = BTRFS_ROOT_ITEM_KEY; 221398d377a0STARUISI Hiroaki key.offset = (u64)-1; 221498d377a0STARUISI Hiroaki root = btrfs_read_fs_root_no_name(info, &key); 221598d377a0STARUISI Hiroaki if (IS_ERR(root)) { 2216f14d104dSDavid Sterba btrfs_err(info, "could not find root %llu", tree_id); 22178ad6fcabSChris Mason ret = -ENOENT; 22188ad6fcabSChris Mason goto out; 221998d377a0STARUISI Hiroaki } 222098d377a0STARUISI Hiroaki 222198d377a0STARUISI Hiroaki key.objectid = dirid; 222298d377a0STARUISI Hiroaki key.type = BTRFS_INODE_REF_KEY; 22238ad6fcabSChris Mason key.offset = (u64)-1; 222498d377a0STARUISI Hiroaki 222598d377a0STARUISI Hiroaki while (1) { 222698d377a0STARUISI Hiroaki ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 222798d377a0STARUISI Hiroaki if (ret < 0) 222898d377a0STARUISI Hiroaki goto out; 222918674c6cSFilipe David Borba Manana else if (ret > 0) { 223018674c6cSFilipe David Borba Manana ret = btrfs_previous_item(root, path, dirid, 223118674c6cSFilipe David Borba Manana BTRFS_INODE_REF_KEY); 223218674c6cSFilipe David Borba Manana if (ret < 0) 223318674c6cSFilipe David Borba Manana goto out; 223418674c6cSFilipe David Borba Manana else if (ret > 0) { 2235ac8e9819SChris Mason ret = -ENOENT; 223698d377a0STARUISI Hiroaki goto out; 2237ac8e9819SChris Mason } 223818674c6cSFilipe David Borba Manana } 223918674c6cSFilipe David Borba Manana 224018674c6cSFilipe David Borba Manana l = path->nodes[0]; 224118674c6cSFilipe David Borba Manana slot = path->slots[0]; 224218674c6cSFilipe David Borba Manana btrfs_item_key_to_cpu(l, &key, slot); 224398d377a0STARUISI Hiroaki 224498d377a0STARUISI Hiroaki iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref); 224598d377a0STARUISI Hiroaki len = btrfs_inode_ref_name_len(l, iref); 224698d377a0STARUISI Hiroaki ptr -= len + 1; 224798d377a0STARUISI Hiroaki total_len += len + 1; 2248a696cf35SFilipe David Borba Manana if (ptr < name) { 2249a696cf35SFilipe David Borba Manana ret = -ENAMETOOLONG; 225098d377a0STARUISI Hiroaki goto out; 2251a696cf35SFilipe David Borba Manana } 225298d377a0STARUISI Hiroaki 225398d377a0STARUISI Hiroaki *(ptr + len) = '/'; 225498d377a0STARUISI Hiroaki read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len); 225598d377a0STARUISI Hiroaki 225698d377a0STARUISI Hiroaki if (key.offset == BTRFS_FIRST_FREE_OBJECTID) 225798d377a0STARUISI Hiroaki break; 225898d377a0STARUISI Hiroaki 2259b3b4aa74SDavid Sterba btrfs_release_path(path); 226098d377a0STARUISI Hiroaki key.objectid = key.offset; 22618ad6fcabSChris Mason key.offset = (u64)-1; 226298d377a0STARUISI Hiroaki dirid = key.objectid; 226398d377a0STARUISI Hiroaki } 226477906a50SLi Zefan memmove(name, ptr, total_len); 226598d377a0STARUISI Hiroaki name[total_len] = '\0'; 226698d377a0STARUISI Hiroaki ret = 0; 226798d377a0STARUISI Hiroaki out: 226898d377a0STARUISI Hiroaki btrfs_free_path(path); 2269ac8e9819SChris Mason return ret; 2270ac8e9819SChris Mason } 2271ac8e9819SChris Mason 2272ac8e9819SChris Mason static noinline int btrfs_ioctl_ino_lookup(struct file *file, 2273ac8e9819SChris Mason void __user *argp) 2274ac8e9819SChris Mason { 2275ac8e9819SChris Mason struct btrfs_ioctl_ino_lookup_args *args; 2276ac8e9819SChris Mason struct inode *inode; 227701b810b8SDavid Sterba int ret = 0; 2278ac8e9819SChris Mason 22792354d08fSJulia Lawall args = memdup_user(argp, sizeof(*args)); 22802354d08fSJulia Lawall if (IS_ERR(args)) 22812354d08fSJulia Lawall return PTR_ERR(args); 2282c2b96929SDan Carpenter 2283496ad9aaSAl Viro inode = file_inode(file); 2284ac8e9819SChris Mason 228501b810b8SDavid Sterba /* 228601b810b8SDavid Sterba * Unprivileged query to obtain the containing subvolume root id. The 228701b810b8SDavid Sterba * path is reset so it's consistent with btrfs_search_path_in_tree. 228801b810b8SDavid Sterba */ 22891b53ac4dSChris Mason if (args->treeid == 0) 22901b53ac4dSChris Mason args->treeid = BTRFS_I(inode)->root->root_key.objectid; 22911b53ac4dSChris Mason 229201b810b8SDavid Sterba if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) { 229301b810b8SDavid Sterba args->name[0] = 0; 229401b810b8SDavid Sterba goto out; 229501b810b8SDavid Sterba } 229601b810b8SDavid Sterba 229701b810b8SDavid Sterba if (!capable(CAP_SYS_ADMIN)) { 229801b810b8SDavid Sterba ret = -EPERM; 229901b810b8SDavid Sterba goto out; 230001b810b8SDavid Sterba } 230101b810b8SDavid Sterba 2302ac8e9819SChris Mason ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info, 2303ac8e9819SChris Mason args->treeid, args->objectid, 2304ac8e9819SChris Mason args->name); 2305ac8e9819SChris Mason 230601b810b8SDavid Sterba out: 2307ac8e9819SChris Mason if (ret == 0 && copy_to_user(argp, args, sizeof(*args))) 2308ac8e9819SChris Mason ret = -EFAULT; 2309ac8e9819SChris Mason 2310ac8e9819SChris Mason kfree(args); 231198d377a0STARUISI Hiroaki return ret; 231298d377a0STARUISI Hiroaki } 231398d377a0STARUISI Hiroaki 231476dda93cSYan, Zheng static noinline int btrfs_ioctl_snap_destroy(struct file *file, 231576dda93cSYan, Zheng void __user *arg) 231676dda93cSYan, Zheng { 231754563d41SAl Viro struct dentry *parent = file->f_path.dentry; 23180b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb); 231976dda93cSYan, Zheng struct dentry *dentry; 23202b0143b5SDavid Howells struct inode *dir = d_inode(parent); 232176dda93cSYan, Zheng struct inode *inode; 232276dda93cSYan, Zheng struct btrfs_root *root = BTRFS_I(dir)->root; 232376dda93cSYan, Zheng struct btrfs_root *dest = NULL; 232476dda93cSYan, Zheng struct btrfs_ioctl_vol_args *vol_args; 232576dda93cSYan, Zheng struct btrfs_trans_handle *trans; 2326c58aaad2SMiao Xie struct btrfs_block_rsv block_rsv; 2327521e0546SDavid Sterba u64 root_flags; 2328c58aaad2SMiao Xie u64 qgroup_reserved; 232976dda93cSYan, Zheng int namelen; 233076dda93cSYan, Zheng int ret; 233176dda93cSYan, Zheng int err = 0; 233276dda93cSYan, Zheng 2333325c50e3SJeff Mahoney if (!S_ISDIR(dir->i_mode)) 2334325c50e3SJeff Mahoney return -ENOTDIR; 2335325c50e3SJeff Mahoney 233676dda93cSYan, Zheng vol_args = memdup_user(arg, sizeof(*vol_args)); 233776dda93cSYan, Zheng if (IS_ERR(vol_args)) 233876dda93cSYan, Zheng return PTR_ERR(vol_args); 233976dda93cSYan, Zheng 234076dda93cSYan, Zheng vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 234176dda93cSYan, Zheng namelen = strlen(vol_args->name); 234276dda93cSYan, Zheng if (strchr(vol_args->name, '/') || 234376dda93cSYan, Zheng strncmp(vol_args->name, "..", namelen) == 0) { 234476dda93cSYan, Zheng err = -EINVAL; 234576dda93cSYan, Zheng goto out; 234676dda93cSYan, Zheng } 234776dda93cSYan, Zheng 2348a561be71SAl Viro err = mnt_want_write_file(file); 234976dda93cSYan, Zheng if (err) 235076dda93cSYan, Zheng goto out; 235176dda93cSYan, Zheng 2352521e0546SDavid Sterba 235300235411SAl Viro err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT); 235400235411SAl Viro if (err == -EINTR) 235500235411SAl Viro goto out_drop_write; 235676dda93cSYan, Zheng dentry = lookup_one_len(vol_args->name, parent, namelen); 235776dda93cSYan, Zheng if (IS_ERR(dentry)) { 235876dda93cSYan, Zheng err = PTR_ERR(dentry); 235976dda93cSYan, Zheng goto out_unlock_dir; 236076dda93cSYan, Zheng } 236176dda93cSYan, Zheng 23622b0143b5SDavid Howells if (d_really_is_negative(dentry)) { 236376dda93cSYan, Zheng err = -ENOENT; 236476dda93cSYan, Zheng goto out_dput; 236576dda93cSYan, Zheng } 236676dda93cSYan, Zheng 23672b0143b5SDavid Howells inode = d_inode(dentry); 23684260f7c7SSage Weil dest = BTRFS_I(inode)->root; 23694260f7c7SSage Weil if (!capable(CAP_SYS_ADMIN)) { 23704260f7c7SSage Weil /* 23714260f7c7SSage Weil * Regular user. Only allow this with a special mount 23724260f7c7SSage Weil * option, when the user has write+exec access to the 23734260f7c7SSage Weil * subvol root, and when rmdir(2) would have been 23744260f7c7SSage Weil * allowed. 23754260f7c7SSage Weil * 23764260f7c7SSage Weil * Note that this is _not_ check that the subvol is 23774260f7c7SSage Weil * empty or doesn't contain data that we wouldn't 23784260f7c7SSage Weil * otherwise be able to delete. 23794260f7c7SSage Weil * 23804260f7c7SSage Weil * Users who want to delete empty subvols should try 23814260f7c7SSage Weil * rmdir(2). 23824260f7c7SSage Weil */ 23834260f7c7SSage Weil err = -EPERM; 23840b246afaSJeff Mahoney if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED)) 23854260f7c7SSage Weil goto out_dput; 23864260f7c7SSage Weil 23874260f7c7SSage Weil /* 23884260f7c7SSage Weil * Do not allow deletion if the parent dir is the same 23894260f7c7SSage Weil * as the dir to be deleted. That means the ioctl 23904260f7c7SSage Weil * must be called on the dentry referencing the root 23914260f7c7SSage Weil * of the subvol, not a random directory contained 23924260f7c7SSage Weil * within it. 23934260f7c7SSage Weil */ 23944260f7c7SSage Weil err = -EINVAL; 23954260f7c7SSage Weil if (root == dest) 23964260f7c7SSage Weil goto out_dput; 23974260f7c7SSage Weil 23984260f7c7SSage Weil err = inode_permission(inode, MAY_WRITE | MAY_EXEC); 23994260f7c7SSage Weil if (err) 24004260f7c7SSage Weil goto out_dput; 24015c39da5bSMiao Xie } 24024260f7c7SSage Weil 24035c39da5bSMiao Xie /* check if subvolume may be deleted by a user */ 24044260f7c7SSage Weil err = btrfs_may_delete(dir, dentry, 1); 24054260f7c7SSage Weil if (err) 24064260f7c7SSage Weil goto out_dput; 24074260f7c7SSage Weil 24084a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 240976dda93cSYan, Zheng err = -EINVAL; 241076dda93cSYan, Zheng goto out_dput; 241176dda93cSYan, Zheng } 241276dda93cSYan, Zheng 24135955102cSAl Viro inode_lock(inode); 2414521e0546SDavid Sterba 2415521e0546SDavid Sterba /* 2416521e0546SDavid Sterba * Don't allow to delete a subvolume with send in progress. This is 2417521e0546SDavid Sterba * inside the i_mutex so the error handling that has to drop the bit 2418521e0546SDavid Sterba * again is not run concurrently. 2419521e0546SDavid Sterba */ 2420521e0546SDavid Sterba spin_lock(&dest->root_item_lock); 2421c55bfa67SFilipe Manana root_flags = btrfs_root_flags(&dest->root_item); 2422c55bfa67SFilipe Manana if (dest->send_in_progress == 0) { 2423c55bfa67SFilipe Manana btrfs_set_root_flags(&dest->root_item, 2424521e0546SDavid Sterba root_flags | BTRFS_ROOT_SUBVOL_DEAD); 2425521e0546SDavid Sterba spin_unlock(&dest->root_item_lock); 2426521e0546SDavid Sterba } else { 2427521e0546SDavid Sterba spin_unlock(&dest->root_item_lock); 24280b246afaSJeff Mahoney btrfs_warn(fs_info, 2429521e0546SDavid Sterba "Attempt to delete subvolume %llu during send", 2430c55bfa67SFilipe Manana dest->root_key.objectid); 2431521e0546SDavid Sterba err = -EPERM; 2432909e26dcSOmar Sandoval goto out_unlock_inode; 2433521e0546SDavid Sterba } 2434521e0546SDavid Sterba 24350b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 243676dda93cSYan, Zheng 243776dda93cSYan, Zheng err = may_destroy_subvol(dest); 243876dda93cSYan, Zheng if (err) 243976dda93cSYan, Zheng goto out_up_write; 244076dda93cSYan, Zheng 2441c58aaad2SMiao Xie btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); 2442c58aaad2SMiao Xie /* 2443c58aaad2SMiao Xie * One for dir inode, two for dir entries, two for root 2444c58aaad2SMiao Xie * ref/backref. 2445c58aaad2SMiao Xie */ 2446c58aaad2SMiao Xie err = btrfs_subvolume_reserve_metadata(root, &block_rsv, 2447ee3441b4SJeff Mahoney 5, &qgroup_reserved, true); 2448c58aaad2SMiao Xie if (err) 2449c58aaad2SMiao Xie goto out_up_write; 2450c58aaad2SMiao Xie 2451a22285a6SYan, Zheng trans = btrfs_start_transaction(root, 0); 2452a22285a6SYan, Zheng if (IS_ERR(trans)) { 2453a22285a6SYan, Zheng err = PTR_ERR(trans); 2454c58aaad2SMiao Xie goto out_release; 2455a22285a6SYan, Zheng } 2456c58aaad2SMiao Xie trans->block_rsv = &block_rsv; 2457c58aaad2SMiao Xie trans->bytes_reserved = block_rsv.size; 2458a22285a6SYan, Zheng 245943663557SNikolay Borisov btrfs_record_snapshot_destroy(trans, BTRFS_I(dir)); 24602be63d5cSFilipe Manana 246176dda93cSYan, Zheng ret = btrfs_unlink_subvol(trans, root, dir, 246276dda93cSYan, Zheng dest->root_key.objectid, 246376dda93cSYan, Zheng dentry->d_name.name, 246476dda93cSYan, Zheng dentry->d_name.len); 246579787eaaSJeff Mahoney if (ret) { 246679787eaaSJeff Mahoney err = ret; 246766642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 246879787eaaSJeff Mahoney goto out_end_trans; 246979787eaaSJeff Mahoney } 247076dda93cSYan, Zheng 247176dda93cSYan, Zheng btrfs_record_root_in_trans(trans, dest); 247276dda93cSYan, Zheng 247376dda93cSYan, Zheng memset(&dest->root_item.drop_progress, 0, 247476dda93cSYan, Zheng sizeof(dest->root_item.drop_progress)); 247576dda93cSYan, Zheng dest->root_item.drop_level = 0; 247676dda93cSYan, Zheng btrfs_set_root_refs(&dest->root_item, 0); 247776dda93cSYan, Zheng 247827cdeb70SMiao Xie if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) { 247976dda93cSYan, Zheng ret = btrfs_insert_orphan_item(trans, 24800b246afaSJeff Mahoney fs_info->tree_root, 248176dda93cSYan, Zheng dest->root_key.objectid); 248279787eaaSJeff Mahoney if (ret) { 248366642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 248479787eaaSJeff Mahoney err = ret; 248579787eaaSJeff Mahoney goto out_end_trans; 2486d68fc57bSYan, Zheng } 248779787eaaSJeff Mahoney } 2488dd5f9615SStefan Behrens 24890b246afaSJeff Mahoney ret = btrfs_uuid_tree_rem(trans, fs_info, dest->root_item.uuid, 24906bccf3abSJeff Mahoney BTRFS_UUID_KEY_SUBVOL, 2491dd5f9615SStefan Behrens dest->root_key.objectid); 2492dd5f9615SStefan Behrens if (ret && ret != -ENOENT) { 249366642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 2494dd5f9615SStefan Behrens err = ret; 2495dd5f9615SStefan Behrens goto out_end_trans; 2496dd5f9615SStefan Behrens } 2497dd5f9615SStefan Behrens if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) { 24980b246afaSJeff Mahoney ret = btrfs_uuid_tree_rem(trans, fs_info, 2499dd5f9615SStefan Behrens dest->root_item.received_uuid, 2500dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 2501dd5f9615SStefan Behrens dest->root_key.objectid); 2502dd5f9615SStefan Behrens if (ret && ret != -ENOENT) { 250366642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 2504dd5f9615SStefan Behrens err = ret; 2505dd5f9615SStefan Behrens goto out_end_trans; 2506dd5f9615SStefan Behrens } 2507dd5f9615SStefan Behrens } 2508dd5f9615SStefan Behrens 250979787eaaSJeff Mahoney out_end_trans: 2510c58aaad2SMiao Xie trans->block_rsv = NULL; 2511c58aaad2SMiao Xie trans->bytes_reserved = 0; 25123a45bb20SJeff Mahoney ret = btrfs_end_transaction(trans); 251379787eaaSJeff Mahoney if (ret && !err) 251479787eaaSJeff Mahoney err = ret; 251576dda93cSYan, Zheng inode->i_flags |= S_DEAD; 2516c58aaad2SMiao Xie out_release: 25177775c818SDavid Sterba btrfs_subvolume_release_metadata(fs_info, &block_rsv); 251876dda93cSYan, Zheng out_up_write: 25190b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 2520521e0546SDavid Sterba if (err) { 2521521e0546SDavid Sterba spin_lock(&dest->root_item_lock); 2522c55bfa67SFilipe Manana root_flags = btrfs_root_flags(&dest->root_item); 2523c55bfa67SFilipe Manana btrfs_set_root_flags(&dest->root_item, 2524521e0546SDavid Sterba root_flags & ~BTRFS_ROOT_SUBVOL_DEAD); 2525521e0546SDavid Sterba spin_unlock(&dest->root_item_lock); 2526521e0546SDavid Sterba } 2527909e26dcSOmar Sandoval out_unlock_inode: 25285955102cSAl Viro inode_unlock(inode); 252976dda93cSYan, Zheng if (!err) { 253064ad6c48SOmar Sandoval d_invalidate(dentry); 253176dda93cSYan, Zheng btrfs_invalidate_inodes(dest); 253276dda93cSYan, Zheng d_delete(dentry); 253361155aa0SDavid Sterba ASSERT(dest->send_in_progress == 0); 2534fa6ac876SLiu Bo 2535fa6ac876SLiu Bo /* the last ref */ 253657cdc8dbSDavid Sterba if (dest->ino_cache_inode) { 253757cdc8dbSDavid Sterba iput(dest->ino_cache_inode); 253857cdc8dbSDavid Sterba dest->ino_cache_inode = NULL; 2539fa6ac876SLiu Bo } 254076dda93cSYan, Zheng } 254176dda93cSYan, Zheng out_dput: 254276dda93cSYan, Zheng dput(dentry); 254376dda93cSYan, Zheng out_unlock_dir: 25445955102cSAl Viro inode_unlock(dir); 254500235411SAl Viro out_drop_write: 25462a79f17eSAl Viro mnt_drop_write_file(file); 254776dda93cSYan, Zheng out: 254876dda93cSYan, Zheng kfree(vol_args); 254976dda93cSYan, Zheng return err; 255076dda93cSYan, Zheng } 255176dda93cSYan, Zheng 25521e701a32SChris Mason static int btrfs_ioctl_defrag(struct file *file, void __user *argp) 2553f46b5a66SChristoph Hellwig { 2554496ad9aaSAl Viro struct inode *inode = file_inode(file); 2555f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 25561e701a32SChris Mason struct btrfs_ioctl_defrag_range_args *range; 2557c146afadSYan Zheng int ret; 2558c146afadSYan Zheng 255925122d15SIlya Dryomov ret = mnt_want_write_file(file); 256025122d15SIlya Dryomov if (ret) 256125122d15SIlya Dryomov return ret; 2562b83cc969SLi Zefan 256325122d15SIlya Dryomov if (btrfs_root_readonly(root)) { 256425122d15SIlya Dryomov ret = -EROFS; 256525122d15SIlya Dryomov goto out; 25665ac00addSStefan Behrens } 2567f46b5a66SChristoph Hellwig 2568f46b5a66SChristoph Hellwig switch (inode->i_mode & S_IFMT) { 2569f46b5a66SChristoph Hellwig case S_IFDIR: 2570e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) { 2571e441d54dSChris Mason ret = -EPERM; 2572e441d54dSChris Mason goto out; 2573e441d54dSChris Mason } 2574de78b51aSEric Sandeen ret = btrfs_defrag_root(root); 2575f46b5a66SChristoph Hellwig break; 2576f46b5a66SChristoph Hellwig case S_IFREG: 2577e441d54dSChris Mason if (!(file->f_mode & FMODE_WRITE)) { 2578e441d54dSChris Mason ret = -EINVAL; 2579e441d54dSChris Mason goto out; 2580e441d54dSChris Mason } 25811e701a32SChris Mason 25821e701a32SChris Mason range = kzalloc(sizeof(*range), GFP_KERNEL); 25831e701a32SChris Mason if (!range) { 25841e701a32SChris Mason ret = -ENOMEM; 25851e701a32SChris Mason goto out; 25861e701a32SChris Mason } 25871e701a32SChris Mason 25881e701a32SChris Mason if (argp) { 25891e701a32SChris Mason if (copy_from_user(range, argp, 25901e701a32SChris Mason sizeof(*range))) { 25911e701a32SChris Mason ret = -EFAULT; 25921e701a32SChris Mason kfree(range); 2593683be16eSDan Carpenter goto out; 25941e701a32SChris Mason } 25951e701a32SChris Mason /* compression requires us to start the IO */ 25961e701a32SChris Mason if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) { 25971e701a32SChris Mason range->flags |= BTRFS_DEFRAG_RANGE_START_IO; 25981e701a32SChris Mason range->extent_thresh = (u32)-1; 25991e701a32SChris Mason } 26001e701a32SChris Mason } else { 26011e701a32SChris Mason /* the rest are all set to zero by kzalloc */ 26021e701a32SChris Mason range->len = (u64)-1; 26031e701a32SChris Mason } 2604496ad9aaSAl Viro ret = btrfs_defrag_file(file_inode(file), file, 26054cb5300bSChris Mason range, 0, 0); 26064cb5300bSChris Mason if (ret > 0) 26074cb5300bSChris Mason ret = 0; 26081e701a32SChris Mason kfree(range); 2609f46b5a66SChristoph Hellwig break; 26108929ecfaSYan, Zheng default: 26118929ecfaSYan, Zheng ret = -EINVAL; 2612f46b5a66SChristoph Hellwig } 2613e441d54dSChris Mason out: 261425122d15SIlya Dryomov mnt_drop_write_file(file); 2615e441d54dSChris Mason return ret; 2616f46b5a66SChristoph Hellwig } 2617f46b5a66SChristoph Hellwig 26182ff7e61eSJeff Mahoney static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg) 2619f46b5a66SChristoph Hellwig { 2620f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2621f46b5a66SChristoph Hellwig int ret; 2622f46b5a66SChristoph Hellwig 2623e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2624e441d54dSChris Mason return -EPERM; 2625e441d54dSChris Mason 2626171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) 2627e57138b3SAnand Jain return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 2628c9e9f97bSIlya Dryomov 26290b246afaSJeff Mahoney mutex_lock(&fs_info->volume_mutex); 2630dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2631c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 2632c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 2633c9e9f97bSIlya Dryomov goto out; 2634c9e9f97bSIlya Dryomov } 2635f46b5a66SChristoph Hellwig 26365516e595SMark Fasheh vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 26372ff7e61eSJeff Mahoney ret = btrfs_init_new_device(fs_info, vol_args->name); 2638f46b5a66SChristoph Hellwig 263943d20761SAnand Jain if (!ret) 26400b246afaSJeff Mahoney btrfs_info(fs_info, "disk added %s", vol_args->name); 264143d20761SAnand Jain 2642f46b5a66SChristoph Hellwig kfree(vol_args); 2643c9e9f97bSIlya Dryomov out: 26440b246afaSJeff Mahoney mutex_unlock(&fs_info->volume_mutex); 2645171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 2646f46b5a66SChristoph Hellwig return ret; 2647f46b5a66SChristoph Hellwig } 2648f46b5a66SChristoph Hellwig 26496b526ed7SAnand Jain static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg) 2650f46b5a66SChristoph Hellwig { 26510b246afaSJeff Mahoney struct inode *inode = file_inode(file); 26520b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 26536b526ed7SAnand Jain struct btrfs_ioctl_vol_args_v2 *vol_args; 2654f46b5a66SChristoph Hellwig int ret; 2655f46b5a66SChristoph Hellwig 2656e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 2657e441d54dSChris Mason return -EPERM; 2658e441d54dSChris Mason 2659da24927bSMiao Xie ret = mnt_want_write_file(file); 2660da24927bSMiao Xie if (ret) 2661da24927bSMiao Xie return ret; 2662c146afadSYan Zheng 2663dae7b665SLi Zefan vol_args = memdup_user(arg, sizeof(*vol_args)); 2664c9e9f97bSIlya Dryomov if (IS_ERR(vol_args)) { 2665c9e9f97bSIlya Dryomov ret = PTR_ERR(vol_args); 2666c47ca32dSDan Carpenter goto err_drop; 2667c9e9f97bSIlya Dryomov } 2668f46b5a66SChristoph Hellwig 26696b526ed7SAnand Jain /* Check for compatibility reject unknown flags */ 2670735654eaSDavid Sterba if (vol_args->flags & ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED) 2671735654eaSDavid Sterba return -EOPNOTSUPP; 2672f46b5a66SChristoph Hellwig 2673171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 2674183860f6SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 2675183860f6SAnand Jain goto out; 2676183860f6SAnand Jain } 2677183860f6SAnand Jain 26780b246afaSJeff Mahoney mutex_lock(&fs_info->volume_mutex); 2679735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) { 26802ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, NULL, vol_args->devid); 26816b526ed7SAnand Jain } else { 26826b526ed7SAnand Jain vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0'; 26832ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, vol_args->name, 0); 26846b526ed7SAnand Jain } 26850b246afaSJeff Mahoney mutex_unlock(&fs_info->volume_mutex); 2686171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 2687183860f6SAnand Jain 26886b526ed7SAnand Jain if (!ret) { 2689735654eaSDavid Sterba if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) 26900b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: id %llu", 26916b526ed7SAnand Jain vol_args->devid); 26926b526ed7SAnand Jain else 26930b246afaSJeff Mahoney btrfs_info(fs_info, "device deleted: %s", 26946b526ed7SAnand Jain vol_args->name); 26956b526ed7SAnand Jain } 2696183860f6SAnand Jain out: 2697183860f6SAnand Jain kfree(vol_args); 2698c47ca32dSDan Carpenter err_drop: 26994ac20c70SIlya Dryomov mnt_drop_write_file(file); 2700f46b5a66SChristoph Hellwig return ret; 2701f46b5a66SChristoph Hellwig } 2702f46b5a66SChristoph Hellwig 2703f46b5a66SChristoph Hellwig static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg) 2704f46b5a66SChristoph Hellwig { 27050b246afaSJeff Mahoney struct inode *inode = file_inode(file); 27060b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 2707f46b5a66SChristoph Hellwig struct btrfs_ioctl_vol_args *vol_args; 2708f46b5a66SChristoph Hellwig int ret; 2709f46b5a66SChristoph Hellwig 2710f46b5a66SChristoph Hellwig if (!capable(CAP_SYS_ADMIN)) 2711f46b5a66SChristoph Hellwig return -EPERM; 2712f46b5a66SChristoph Hellwig 2713f46b5a66SChristoph Hellwig ret = mnt_want_write_file(file); 2714f46b5a66SChristoph Hellwig if (ret) 2715f46b5a66SChristoph Hellwig return ret; 2716f46b5a66SChristoph Hellwig 2717171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 2718f46b5a66SChristoph Hellwig ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 271958d7bbf8SDavid Sterba goto out_drop_write; 272058d7bbf8SDavid Sterba } 272158d7bbf8SDavid Sterba 272258d7bbf8SDavid Sterba vol_args = memdup_user(arg, sizeof(*vol_args)); 272358d7bbf8SDavid Sterba if (IS_ERR(vol_args)) { 272458d7bbf8SDavid Sterba ret = PTR_ERR(vol_args); 2725f46b5a66SChristoph Hellwig goto out; 2726f46b5a66SChristoph Hellwig } 2727f46b5a66SChristoph Hellwig 272858d7bbf8SDavid Sterba vol_args->name[BTRFS_PATH_NAME_MAX] = '\0'; 27290b246afaSJeff Mahoney mutex_lock(&fs_info->volume_mutex); 27302ff7e61eSJeff Mahoney ret = btrfs_rm_device(fs_info, vol_args->name, 0); 27310b246afaSJeff Mahoney mutex_unlock(&fs_info->volume_mutex); 2732f46b5a66SChristoph Hellwig 2733f46b5a66SChristoph Hellwig if (!ret) 27340b246afaSJeff Mahoney btrfs_info(fs_info, "disk deleted %s", vol_args->name); 2735f46b5a66SChristoph Hellwig kfree(vol_args); 273658d7bbf8SDavid Sterba out: 2737171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 273858d7bbf8SDavid Sterba out_drop_write: 2739f46b5a66SChristoph Hellwig mnt_drop_write_file(file); 274058d7bbf8SDavid Sterba 2741f46b5a66SChristoph Hellwig return ret; 2742f46b5a66SChristoph Hellwig } 2743f46b5a66SChristoph Hellwig 27442ff7e61eSJeff Mahoney static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info, 27452ff7e61eSJeff Mahoney void __user *arg) 2746475f6387SJan Schmidt { 2747027ed2f0SLi Zefan struct btrfs_ioctl_fs_info_args *fi_args; 2748475f6387SJan Schmidt struct btrfs_device *device; 27490b246afaSJeff Mahoney struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 2750027ed2f0SLi Zefan int ret = 0; 2751475f6387SJan Schmidt 2752027ed2f0SLi Zefan fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL); 2753027ed2f0SLi Zefan if (!fi_args) 2754027ed2f0SLi Zefan return -ENOMEM; 2755027ed2f0SLi Zefan 2756f7171750SFilipe David Borba Manana mutex_lock(&fs_devices->device_list_mutex); 2757027ed2f0SLi Zefan fi_args->num_devices = fs_devices->num_devices; 27580b246afaSJeff Mahoney memcpy(&fi_args->fsid, fs_info->fsid, sizeof(fi_args->fsid)); 2759475f6387SJan Schmidt 2760d7641a49SByongho Lee list_for_each_entry(device, &fs_devices->devices, dev_list) { 2761027ed2f0SLi Zefan if (device->devid > fi_args->max_id) 2762027ed2f0SLi Zefan fi_args->max_id = device->devid; 2763475f6387SJan Schmidt } 2764475f6387SJan Schmidt mutex_unlock(&fs_devices->device_list_mutex); 2765475f6387SJan Schmidt 2766bea7eafdSOmar Sandoval fi_args->nodesize = fs_info->nodesize; 2767bea7eafdSOmar Sandoval fi_args->sectorsize = fs_info->sectorsize; 2768bea7eafdSOmar Sandoval fi_args->clone_alignment = fs_info->sectorsize; 276980a773fbSDavid Sterba 2770027ed2f0SLi Zefan if (copy_to_user(arg, fi_args, sizeof(*fi_args))) 2771027ed2f0SLi Zefan ret = -EFAULT; 2772475f6387SJan Schmidt 2773027ed2f0SLi Zefan kfree(fi_args); 2774027ed2f0SLi Zefan return ret; 2775475f6387SJan Schmidt } 2776475f6387SJan Schmidt 27772ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info, 27782ff7e61eSJeff Mahoney void __user *arg) 2779475f6387SJan Schmidt { 2780475f6387SJan Schmidt struct btrfs_ioctl_dev_info_args *di_args; 2781475f6387SJan Schmidt struct btrfs_device *dev; 27820b246afaSJeff Mahoney struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 2783475f6387SJan Schmidt int ret = 0; 2784475f6387SJan Schmidt char *s_uuid = NULL; 2785475f6387SJan Schmidt 2786475f6387SJan Schmidt di_args = memdup_user(arg, sizeof(*di_args)); 2787475f6387SJan Schmidt if (IS_ERR(di_args)) 2788475f6387SJan Schmidt return PTR_ERR(di_args); 2789475f6387SJan Schmidt 2790dd5f9615SStefan Behrens if (!btrfs_is_empty_uuid(di_args->uuid)) 2791475f6387SJan Schmidt s_uuid = di_args->uuid; 2792475f6387SJan Schmidt 2793475f6387SJan Schmidt mutex_lock(&fs_devices->device_list_mutex); 27940b246afaSJeff Mahoney dev = btrfs_find_device(fs_info, di_args->devid, s_uuid, NULL); 2795475f6387SJan Schmidt 2796475f6387SJan Schmidt if (!dev) { 2797475f6387SJan Schmidt ret = -ENODEV; 2798475f6387SJan Schmidt goto out; 2799475f6387SJan Schmidt } 2800475f6387SJan Schmidt 2801475f6387SJan Schmidt di_args->devid = dev->devid; 28027cc8e58dSMiao Xie di_args->bytes_used = btrfs_device_get_bytes_used(dev); 28037cc8e58dSMiao Xie di_args->total_bytes = btrfs_device_get_total_bytes(dev); 2804475f6387SJan Schmidt memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid)); 2805a27202fbSJim Meyering if (dev->name) { 2806606686eeSJosef Bacik struct rcu_string *name; 2807606686eeSJosef Bacik 2808606686eeSJosef Bacik rcu_read_lock(); 2809606686eeSJosef Bacik name = rcu_dereference(dev->name); 2810606686eeSJosef Bacik strncpy(di_args->path, name->str, sizeof(di_args->path)); 2811606686eeSJosef Bacik rcu_read_unlock(); 2812a27202fbSJim Meyering di_args->path[sizeof(di_args->path) - 1] = 0; 2813a27202fbSJim Meyering } else { 281499ba55adSStefan Behrens di_args->path[0] = '\0'; 2815a27202fbSJim Meyering } 2816475f6387SJan Schmidt 2817475f6387SJan Schmidt out: 281855793c0dSDavid Sterba mutex_unlock(&fs_devices->device_list_mutex); 2819475f6387SJan Schmidt if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args))) 2820475f6387SJan Schmidt ret = -EFAULT; 2821475f6387SJan Schmidt 2822475f6387SJan Schmidt kfree(di_args); 2823475f6387SJan Schmidt return ret; 2824475f6387SJan Schmidt } 2825475f6387SJan Schmidt 2826f4414602SMark Fasheh static struct page *extent_same_get_page(struct inode *inode, pgoff_t index) 2827416161dbSMark Fasheh { 2828416161dbSMark Fasheh struct page *page; 2829416161dbSMark Fasheh 2830416161dbSMark Fasheh page = grab_cache_page(inode->i_mapping, index); 2831416161dbSMark Fasheh if (!page) 283231314002SFilipe Manana return ERR_PTR(-ENOMEM); 2833416161dbSMark Fasheh 2834416161dbSMark Fasheh if (!PageUptodate(page)) { 283531314002SFilipe Manana int ret; 283631314002SFilipe Manana 283731314002SFilipe Manana ret = btrfs_readpage(NULL, page); 283831314002SFilipe Manana if (ret) 283931314002SFilipe Manana return ERR_PTR(ret); 2840416161dbSMark Fasheh lock_page(page); 2841416161dbSMark Fasheh if (!PageUptodate(page)) { 2842416161dbSMark Fasheh unlock_page(page); 284309cbfeafSKirill A. Shutemov put_page(page); 284431314002SFilipe Manana return ERR_PTR(-EIO); 2845416161dbSMark Fasheh } 284631314002SFilipe Manana if (page->mapping != inode->i_mapping) { 2847416161dbSMark Fasheh unlock_page(page); 284809cbfeafSKirill A. Shutemov put_page(page); 284931314002SFilipe Manana return ERR_PTR(-EAGAIN); 2850416161dbSMark Fasheh } 2851416161dbSMark Fasheh } 2852416161dbSMark Fasheh 2853416161dbSMark Fasheh return page; 2854416161dbSMark Fasheh } 2855416161dbSMark Fasheh 2856f4414602SMark Fasheh static int gather_extent_pages(struct inode *inode, struct page **pages, 2857f4414602SMark Fasheh int num_pages, u64 off) 2858f4414602SMark Fasheh { 2859f4414602SMark Fasheh int i; 286009cbfeafSKirill A. Shutemov pgoff_t index = off >> PAGE_SHIFT; 2861f4414602SMark Fasheh 2862f4414602SMark Fasheh for (i = 0; i < num_pages; i++) { 286331314002SFilipe Manana again: 2864f4414602SMark Fasheh pages[i] = extent_same_get_page(inode, index + i); 286531314002SFilipe Manana if (IS_ERR(pages[i])) { 286631314002SFilipe Manana int err = PTR_ERR(pages[i]); 286731314002SFilipe Manana 286831314002SFilipe Manana if (err == -EAGAIN) 286931314002SFilipe Manana goto again; 287031314002SFilipe Manana pages[i] = NULL; 287131314002SFilipe Manana return err; 287231314002SFilipe Manana } 2873f4414602SMark Fasheh } 2874f4414602SMark Fasheh return 0; 2875f4414602SMark Fasheh } 2876f4414602SMark Fasheh 2877e0bd70c6SFilipe Manana static int lock_extent_range(struct inode *inode, u64 off, u64 len, 2878e0bd70c6SFilipe Manana bool retry_range_locking) 287977fe20dcSMark Fasheh { 2880e0bd70c6SFilipe Manana /* 2881e0bd70c6SFilipe Manana * Do any pending delalloc/csum calculations on inode, one way or 2882e0bd70c6SFilipe Manana * another, and lock file content. 2883e0bd70c6SFilipe Manana * The locking order is: 2884e0bd70c6SFilipe Manana * 2885e0bd70c6SFilipe Manana * 1) pages 2886e0bd70c6SFilipe Manana * 2) range in the inode's io tree 2887e0bd70c6SFilipe Manana */ 288877fe20dcSMark Fasheh while (1) { 288977fe20dcSMark Fasheh struct btrfs_ordered_extent *ordered; 289077fe20dcSMark Fasheh lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); 289177fe20dcSMark Fasheh ordered = btrfs_lookup_first_ordered_extent(inode, 289277fe20dcSMark Fasheh off + len - 1); 2893ff5df9b8SFilipe Manana if ((!ordered || 2894ff5df9b8SFilipe Manana ordered->file_offset + ordered->len <= off || 2895ff5df9b8SFilipe Manana ordered->file_offset >= off + len) && 289677fe20dcSMark Fasheh !test_range_bit(&BTRFS_I(inode)->io_tree, off, 2897ff5df9b8SFilipe Manana off + len - 1, EXTENT_DELALLOC, 0, NULL)) { 2898ff5df9b8SFilipe Manana if (ordered) 2899ff5df9b8SFilipe Manana btrfs_put_ordered_extent(ordered); 290077fe20dcSMark Fasheh break; 2901ff5df9b8SFilipe Manana } 290277fe20dcSMark Fasheh unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1); 290377fe20dcSMark Fasheh if (ordered) 290477fe20dcSMark Fasheh btrfs_put_ordered_extent(ordered); 2905e0bd70c6SFilipe Manana if (!retry_range_locking) 2906e0bd70c6SFilipe Manana return -EAGAIN; 290777fe20dcSMark Fasheh btrfs_wait_ordered_range(inode, off, len); 290877fe20dcSMark Fasheh } 2909e0bd70c6SFilipe Manana return 0; 291077fe20dcSMark Fasheh } 291177fe20dcSMark Fasheh 2912f4414602SMark Fasheh static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2) 2913416161dbSMark Fasheh { 29145955102cSAl Viro inode_unlock(inode1); 29155955102cSAl Viro inode_unlock(inode2); 2916416161dbSMark Fasheh } 2917416161dbSMark Fasheh 2918f4414602SMark Fasheh static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2) 2919f4414602SMark Fasheh { 2920f4414602SMark Fasheh if (inode1 < inode2) 2921f4414602SMark Fasheh swap(inode1, inode2); 2922f4414602SMark Fasheh 29235955102cSAl Viro inode_lock_nested(inode1, I_MUTEX_PARENT); 29245955102cSAl Viro inode_lock_nested(inode2, I_MUTEX_CHILD); 2925f4414602SMark Fasheh } 2926f4414602SMark Fasheh 2927f4414602SMark Fasheh static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1, 2928f4414602SMark Fasheh struct inode *inode2, u64 loff2, u64 len) 2929f4414602SMark Fasheh { 2930f4414602SMark Fasheh unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1); 2931f4414602SMark Fasheh unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1); 2932f4414602SMark Fasheh } 2933f4414602SMark Fasheh 2934e0bd70c6SFilipe Manana static int btrfs_double_extent_lock(struct inode *inode1, u64 loff1, 2935e0bd70c6SFilipe Manana struct inode *inode2, u64 loff2, u64 len, 2936e0bd70c6SFilipe Manana bool retry_range_locking) 2937416161dbSMark Fasheh { 2938e0bd70c6SFilipe Manana int ret; 2939e0bd70c6SFilipe Manana 2940416161dbSMark Fasheh if (inode1 < inode2) { 2941416161dbSMark Fasheh swap(inode1, inode2); 2942416161dbSMark Fasheh swap(loff1, loff2); 2943416161dbSMark Fasheh } 2944e0bd70c6SFilipe Manana ret = lock_extent_range(inode1, loff1, len, retry_range_locking); 2945e0bd70c6SFilipe Manana if (ret) 2946e0bd70c6SFilipe Manana return ret; 2947e0bd70c6SFilipe Manana ret = lock_extent_range(inode2, loff2, len, retry_range_locking); 2948e0bd70c6SFilipe Manana if (ret) 2949e0bd70c6SFilipe Manana unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, 2950e0bd70c6SFilipe Manana loff1 + len - 1); 2951e0bd70c6SFilipe Manana return ret; 2952416161dbSMark Fasheh } 2953f4414602SMark Fasheh 2954f4414602SMark Fasheh struct cmp_pages { 2955f4414602SMark Fasheh int num_pages; 2956f4414602SMark Fasheh struct page **src_pages; 2957f4414602SMark Fasheh struct page **dst_pages; 2958f4414602SMark Fasheh }; 2959f4414602SMark Fasheh 2960f4414602SMark Fasheh static void btrfs_cmp_data_free(struct cmp_pages *cmp) 2961f4414602SMark Fasheh { 2962f4414602SMark Fasheh int i; 2963f4414602SMark Fasheh struct page *pg; 2964f4414602SMark Fasheh 2965f4414602SMark Fasheh for (i = 0; i < cmp->num_pages; i++) { 2966f4414602SMark Fasheh pg = cmp->src_pages[i]; 2967e0bd70c6SFilipe Manana if (pg) { 2968e0bd70c6SFilipe Manana unlock_page(pg); 296909cbfeafSKirill A. Shutemov put_page(pg); 2970e0bd70c6SFilipe Manana } 2971f4414602SMark Fasheh pg = cmp->dst_pages[i]; 2972e0bd70c6SFilipe Manana if (pg) { 2973e0bd70c6SFilipe Manana unlock_page(pg); 297409cbfeafSKirill A. Shutemov put_page(pg); 2975f4414602SMark Fasheh } 2976e0bd70c6SFilipe Manana } 2977f4414602SMark Fasheh kfree(cmp->src_pages); 2978f4414602SMark Fasheh kfree(cmp->dst_pages); 2979f4414602SMark Fasheh } 2980f4414602SMark Fasheh 2981f4414602SMark Fasheh static int btrfs_cmp_data_prepare(struct inode *src, u64 loff, 2982f4414602SMark Fasheh struct inode *dst, u64 dst_loff, 2983f4414602SMark Fasheh u64 len, struct cmp_pages *cmp) 2984f4414602SMark Fasheh { 2985f4414602SMark Fasheh int ret; 298609cbfeafSKirill A. Shutemov int num_pages = PAGE_ALIGN(len) >> PAGE_SHIFT; 2987f4414602SMark Fasheh struct page **src_pgarr, **dst_pgarr; 2988f4414602SMark Fasheh 2989f4414602SMark Fasheh /* 2990f4414602SMark Fasheh * We must gather up all the pages before we initiate our 2991f4414602SMark Fasheh * extent locking. We use an array for the page pointers. Size 2992f4414602SMark Fasheh * of the array is bounded by len, which is in turn bounded by 2993f4414602SMark Fasheh * BTRFS_MAX_DEDUPE_LEN. 2994f4414602SMark Fasheh */ 299566722f7cSDavid Sterba src_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL); 299666722f7cSDavid Sterba dst_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL); 2997f4414602SMark Fasheh if (!src_pgarr || !dst_pgarr) { 2998f4414602SMark Fasheh kfree(src_pgarr); 2999f4414602SMark Fasheh kfree(dst_pgarr); 3000f4414602SMark Fasheh return -ENOMEM; 3001f4414602SMark Fasheh } 3002f4414602SMark Fasheh cmp->num_pages = num_pages; 3003f4414602SMark Fasheh cmp->src_pages = src_pgarr; 3004f4414602SMark Fasheh cmp->dst_pages = dst_pgarr; 3005f4414602SMark Fasheh 3006b1517622SFilipe Manana /* 3007b1517622SFilipe Manana * If deduping ranges in the same inode, locking rules make it mandatory 3008b1517622SFilipe Manana * to always lock pages in ascending order to avoid deadlocks with 3009b1517622SFilipe Manana * concurrent tasks (such as starting writeback/delalloc). 3010b1517622SFilipe Manana */ 3011b1517622SFilipe Manana if (src == dst && dst_loff < loff) { 3012b1517622SFilipe Manana swap(src_pgarr, dst_pgarr); 3013b1517622SFilipe Manana swap(loff, dst_loff); 3014b1517622SFilipe Manana } 3015b1517622SFilipe Manana 3016b1517622SFilipe Manana ret = gather_extent_pages(src, src_pgarr, cmp->num_pages, loff); 3017f4414602SMark Fasheh if (ret) 3018f4414602SMark Fasheh goto out; 3019f4414602SMark Fasheh 3020b1517622SFilipe Manana ret = gather_extent_pages(dst, dst_pgarr, cmp->num_pages, dst_loff); 3021f4414602SMark Fasheh 3022f4414602SMark Fasheh out: 3023f4414602SMark Fasheh if (ret) 3024f4414602SMark Fasheh btrfs_cmp_data_free(cmp); 302578ad4ce0SNaohiro Aota return ret; 3026416161dbSMark Fasheh } 3027416161dbSMark Fasheh 30281a287cfeSDavid Sterba static int btrfs_cmp_data(u64 len, struct cmp_pages *cmp) 3029416161dbSMark Fasheh { 3030416161dbSMark Fasheh int ret = 0; 3031f4414602SMark Fasheh int i; 3032416161dbSMark Fasheh struct page *src_page, *dst_page; 303309cbfeafSKirill A. Shutemov unsigned int cmp_len = PAGE_SIZE; 3034416161dbSMark Fasheh void *addr, *dst_addr; 3035416161dbSMark Fasheh 3036f4414602SMark Fasheh i = 0; 3037416161dbSMark Fasheh while (len) { 303809cbfeafSKirill A. Shutemov if (len < PAGE_SIZE) 3039416161dbSMark Fasheh cmp_len = len; 3040416161dbSMark Fasheh 3041f4414602SMark Fasheh BUG_ON(i >= cmp->num_pages); 3042f4414602SMark Fasheh 3043f4414602SMark Fasheh src_page = cmp->src_pages[i]; 3044f4414602SMark Fasheh dst_page = cmp->dst_pages[i]; 3045e0bd70c6SFilipe Manana ASSERT(PageLocked(src_page)); 3046e0bd70c6SFilipe Manana ASSERT(PageLocked(dst_page)); 3047f4414602SMark Fasheh 3048416161dbSMark Fasheh addr = kmap_atomic(src_page); 3049416161dbSMark Fasheh dst_addr = kmap_atomic(dst_page); 3050416161dbSMark Fasheh 3051416161dbSMark Fasheh flush_dcache_page(src_page); 3052416161dbSMark Fasheh flush_dcache_page(dst_page); 3053416161dbSMark Fasheh 3054416161dbSMark Fasheh if (memcmp(addr, dst_addr, cmp_len)) 30552b3909f8SDarrick J. Wong ret = -EBADE; 3056416161dbSMark Fasheh 3057416161dbSMark Fasheh kunmap_atomic(addr); 3058416161dbSMark Fasheh kunmap_atomic(dst_addr); 3059416161dbSMark Fasheh 3060416161dbSMark Fasheh if (ret) 3061416161dbSMark Fasheh break; 3062416161dbSMark Fasheh 3063416161dbSMark Fasheh len -= cmp_len; 3064f4414602SMark Fasheh i++; 3065416161dbSMark Fasheh } 3066416161dbSMark Fasheh 3067416161dbSMark Fasheh return ret; 3068416161dbSMark Fasheh } 3069416161dbSMark Fasheh 3070e1d227a4SMark Fasheh static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen, 3071e1d227a4SMark Fasheh u64 olen) 3072416161dbSMark Fasheh { 3073e1d227a4SMark Fasheh u64 len = *plen; 3074416161dbSMark Fasheh u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize; 3075416161dbSMark Fasheh 3076e1d227a4SMark Fasheh if (off + olen > inode->i_size || off + olen < off) 3077416161dbSMark Fasheh return -EINVAL; 3078e1d227a4SMark Fasheh 3079e1d227a4SMark Fasheh /* if we extend to eof, continue to block boundary */ 3080e1d227a4SMark Fasheh if (off + len == inode->i_size) 3081e1d227a4SMark Fasheh *plen = len = ALIGN(inode->i_size, bs) - off; 3082e1d227a4SMark Fasheh 3083416161dbSMark Fasheh /* Check that we are block aligned - btrfs_clone() requires this */ 3084416161dbSMark Fasheh if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs)) 3085416161dbSMark Fasheh return -EINVAL; 3086416161dbSMark Fasheh 3087416161dbSMark Fasheh return 0; 3088416161dbSMark Fasheh } 3089416161dbSMark Fasheh 3090e1d227a4SMark Fasheh static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen, 3091416161dbSMark Fasheh struct inode *dst, u64 dst_loff) 3092416161dbSMark Fasheh { 3093416161dbSMark Fasheh int ret; 3094e1d227a4SMark Fasheh u64 len = olen; 3095f4414602SMark Fasheh struct cmp_pages cmp; 3096fc4badd9SOmar Sandoval bool same_inode = (src == dst); 30970efa9f48SMark Fasheh u64 same_lock_start = 0; 30980efa9f48SMark Fasheh u64 same_lock_len = 0; 3099416161dbSMark Fasheh 3100113e8283SFilipe Manana if (len == 0) 3101113e8283SFilipe Manana return 0; 3102113e8283SFilipe Manana 3103fc4badd9SOmar Sandoval if (same_inode) 31045955102cSAl Viro inode_lock(src); 3105fc4badd9SOmar Sandoval else 3106fc4badd9SOmar Sandoval btrfs_double_inode_lock(src, dst); 31070efa9f48SMark Fasheh 31080efa9f48SMark Fasheh ret = extent_same_check_offsets(src, loff, &len, olen); 31090efa9f48SMark Fasheh if (ret) 31100efa9f48SMark Fasheh goto out_unlock; 3111fc4badd9SOmar Sandoval 3112fc4badd9SOmar Sandoval ret = extent_same_check_offsets(dst, dst_loff, &len, olen); 3113f4dfe687SFilipe Manana if (ret) 3114f4dfe687SFilipe Manana goto out_unlock; 31150efa9f48SMark Fasheh 3116fc4badd9SOmar Sandoval if (same_inode) { 31170efa9f48SMark Fasheh /* 31180efa9f48SMark Fasheh * Single inode case wants the same checks, except we 31190efa9f48SMark Fasheh * don't want our length pushed out past i_size as 31200efa9f48SMark Fasheh * comparing that data range makes no sense. 31210efa9f48SMark Fasheh * 31220efa9f48SMark Fasheh * extent_same_check_offsets() will do this for an 31230efa9f48SMark Fasheh * unaligned length at i_size, so catch it here and 31240efa9f48SMark Fasheh * reject the request. 31250efa9f48SMark Fasheh * 31260efa9f48SMark Fasheh * This effectively means we require aligned extents 31270efa9f48SMark Fasheh * for the single-inode case, whereas the other cases 31280efa9f48SMark Fasheh * allow an unaligned length so long as it ends at 31290efa9f48SMark Fasheh * i_size. 31300efa9f48SMark Fasheh */ 31310efa9f48SMark Fasheh if (len != olen) { 31320efa9f48SMark Fasheh ret = -EINVAL; 31330efa9f48SMark Fasheh goto out_unlock; 31340efa9f48SMark Fasheh } 31350efa9f48SMark Fasheh 31360efa9f48SMark Fasheh /* Check for overlapping ranges */ 31370efa9f48SMark Fasheh if (dst_loff + len > loff && dst_loff < loff + len) { 31380efa9f48SMark Fasheh ret = -EINVAL; 31390efa9f48SMark Fasheh goto out_unlock; 31400efa9f48SMark Fasheh } 31410efa9f48SMark Fasheh 31420efa9f48SMark Fasheh same_lock_start = min_t(u64, loff, dst_loff); 31430efa9f48SMark Fasheh same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start; 31440efa9f48SMark Fasheh } 3145416161dbSMark Fasheh 3146416161dbSMark Fasheh /* don't make the dst file partly checksummed */ 3147416161dbSMark Fasheh if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != 3148416161dbSMark Fasheh (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) { 3149416161dbSMark Fasheh ret = -EINVAL; 3150416161dbSMark Fasheh goto out_unlock; 3151416161dbSMark Fasheh } 3152416161dbSMark Fasheh 3153e0bd70c6SFilipe Manana again: 3154f4414602SMark Fasheh ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, &cmp); 3155f4414602SMark Fasheh if (ret) 3156f4414602SMark Fasheh goto out_unlock; 3157f4414602SMark Fasheh 31580efa9f48SMark Fasheh if (same_inode) 3159e0bd70c6SFilipe Manana ret = lock_extent_range(src, same_lock_start, same_lock_len, 3160e0bd70c6SFilipe Manana false); 31610efa9f48SMark Fasheh else 3162e0bd70c6SFilipe Manana ret = btrfs_double_extent_lock(src, loff, dst, dst_loff, len, 3163e0bd70c6SFilipe Manana false); 3164e0bd70c6SFilipe Manana /* 3165e0bd70c6SFilipe Manana * If one of the inodes has dirty pages in the respective range or 3166e0bd70c6SFilipe Manana * ordered extents, we need to flush dellaloc and wait for all ordered 3167e0bd70c6SFilipe Manana * extents in the range. We must unlock the pages and the ranges in the 3168e0bd70c6SFilipe Manana * io trees to avoid deadlocks when flushing delalloc (requires locking 3169e0bd70c6SFilipe Manana * pages) and when waiting for ordered extents to complete (they require 3170e0bd70c6SFilipe Manana * range locking). 3171e0bd70c6SFilipe Manana */ 3172e0bd70c6SFilipe Manana if (ret == -EAGAIN) { 3173e0bd70c6SFilipe Manana /* 3174e0bd70c6SFilipe Manana * Ranges in the io trees already unlocked. Now unlock all 3175e0bd70c6SFilipe Manana * pages before waiting for all IO to complete. 3176e0bd70c6SFilipe Manana */ 3177e0bd70c6SFilipe Manana btrfs_cmp_data_free(&cmp); 3178e0bd70c6SFilipe Manana if (same_inode) { 3179e0bd70c6SFilipe Manana btrfs_wait_ordered_range(src, same_lock_start, 3180e0bd70c6SFilipe Manana same_lock_len); 3181e0bd70c6SFilipe Manana } else { 3182e0bd70c6SFilipe Manana btrfs_wait_ordered_range(src, loff, len); 3183e0bd70c6SFilipe Manana btrfs_wait_ordered_range(dst, dst_loff, len); 3184e0bd70c6SFilipe Manana } 3185e0bd70c6SFilipe Manana goto again; 3186e0bd70c6SFilipe Manana } 3187e0bd70c6SFilipe Manana ASSERT(ret == 0); 3188e0bd70c6SFilipe Manana if (WARN_ON(ret)) { 3189e0bd70c6SFilipe Manana /* ranges in the io trees already unlocked */ 3190e0bd70c6SFilipe Manana btrfs_cmp_data_free(&cmp); 3191e0bd70c6SFilipe Manana return ret; 3192e0bd70c6SFilipe Manana } 3193f4414602SMark Fasheh 3194207910ddSMark Fasheh /* pass original length for comparison so we stay within i_size */ 31951a287cfeSDavid Sterba ret = btrfs_cmp_data(olen, &cmp); 3196416161dbSMark Fasheh if (ret == 0) 31971c919a5eSMark Fasheh ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1); 3198416161dbSMark Fasheh 31990efa9f48SMark Fasheh if (same_inode) 32000efa9f48SMark Fasheh unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start, 32010efa9f48SMark Fasheh same_lock_start + same_lock_len - 1); 32020efa9f48SMark Fasheh else 3203f4414602SMark Fasheh btrfs_double_extent_unlock(src, loff, dst, dst_loff, len); 3204f4414602SMark Fasheh 3205f4414602SMark Fasheh btrfs_cmp_data_free(&cmp); 3206416161dbSMark Fasheh out_unlock: 32070efa9f48SMark Fasheh if (same_inode) 32085955102cSAl Viro inode_unlock(src); 32090efa9f48SMark Fasheh else 3210f4414602SMark Fasheh btrfs_double_inode_unlock(src, dst); 3211416161dbSMark Fasheh 3212416161dbSMark Fasheh return ret; 3213416161dbSMark Fasheh } 3214416161dbSMark Fasheh 3215ee22184bSByongho Lee #define BTRFS_MAX_DEDUPE_LEN SZ_16M 3216416161dbSMark Fasheh 32172b3909f8SDarrick J. Wong ssize_t btrfs_dedupe_file_range(struct file *src_file, u64 loff, u64 olen, 32182b3909f8SDarrick J. Wong struct file *dst_file, u64 dst_loff) 3219416161dbSMark Fasheh { 32202b3909f8SDarrick J. Wong struct inode *src = file_inode(src_file); 32212b3909f8SDarrick J. Wong struct inode *dst = file_inode(dst_file); 3222416161dbSMark Fasheh u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize; 32232b3909f8SDarrick J. Wong ssize_t res; 3224416161dbSMark Fasheh 32252b3909f8SDarrick J. Wong if (olen > BTRFS_MAX_DEDUPE_LEN) 32262b3909f8SDarrick J. Wong olen = BTRFS_MAX_DEDUPE_LEN; 3227416161dbSMark Fasheh 322809cbfeafSKirill A. Shutemov if (WARN_ON_ONCE(bs < PAGE_SIZE)) { 3229416161dbSMark Fasheh /* 3230416161dbSMark Fasheh * Btrfs does not support blocksize < page_size. As a 3231416161dbSMark Fasheh * result, btrfs_cmp_data() won't correctly handle 3232416161dbSMark Fasheh * this situation without an update. 3233416161dbSMark Fasheh */ 32342b3909f8SDarrick J. Wong return -EINVAL; 3235416161dbSMark Fasheh } 3236416161dbSMark Fasheh 32372b3909f8SDarrick J. Wong res = btrfs_extent_same(src, loff, olen, dst, dst_loff); 32382b3909f8SDarrick J. Wong if (res) 32392b3909f8SDarrick J. Wong return res; 32402b3909f8SDarrick J. Wong return olen; 3241416161dbSMark Fasheh } 3242416161dbSMark Fasheh 3243f82a9901SFilipe Manana static int clone_finish_inode_update(struct btrfs_trans_handle *trans, 3244f82a9901SFilipe Manana struct inode *inode, 3245f82a9901SFilipe Manana u64 endoff, 3246f82a9901SFilipe Manana const u64 destoff, 32471c919a5eSMark Fasheh const u64 olen, 32481c919a5eSMark Fasheh int no_time_update) 3249f82a9901SFilipe Manana { 3250f82a9901SFilipe Manana struct btrfs_root *root = BTRFS_I(inode)->root; 3251f82a9901SFilipe Manana int ret; 3252f82a9901SFilipe Manana 3253f82a9901SFilipe Manana inode_inc_iversion(inode); 32541c919a5eSMark Fasheh if (!no_time_update) 3255c2050a45SDeepa Dinamani inode->i_mtime = inode->i_ctime = current_time(inode); 3256f82a9901SFilipe Manana /* 3257f82a9901SFilipe Manana * We round up to the block size at eof when determining which 3258f82a9901SFilipe Manana * extents to clone above, but shouldn't round up the file size. 3259f82a9901SFilipe Manana */ 3260f82a9901SFilipe Manana if (endoff > destoff + olen) 3261f82a9901SFilipe Manana endoff = destoff + olen; 3262f82a9901SFilipe Manana if (endoff > inode->i_size) 32636ef06d27SNikolay Borisov btrfs_i_size_write(BTRFS_I(inode), endoff); 3264f82a9901SFilipe Manana 3265f82a9901SFilipe Manana ret = btrfs_update_inode(trans, root, inode); 3266f82a9901SFilipe Manana if (ret) { 326766642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 32683a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3269f82a9901SFilipe Manana goto out; 3270f82a9901SFilipe Manana } 32713a45bb20SJeff Mahoney ret = btrfs_end_transaction(trans); 3272f82a9901SFilipe Manana out: 3273f82a9901SFilipe Manana return ret; 3274f82a9901SFilipe Manana } 3275f82a9901SFilipe Manana 3276a2f392e4SNikolay Borisov static void clone_update_extent_map(struct btrfs_inode *inode, 32777ffbb598SFilipe Manana const struct btrfs_trans_handle *trans, 32787ffbb598SFilipe Manana const struct btrfs_path *path, 32797ffbb598SFilipe Manana const u64 hole_offset, 32807ffbb598SFilipe Manana const u64 hole_len) 32817ffbb598SFilipe Manana { 3282a2f392e4SNikolay Borisov struct extent_map_tree *em_tree = &inode->extent_tree; 32837ffbb598SFilipe Manana struct extent_map *em; 32847ffbb598SFilipe Manana int ret; 32857ffbb598SFilipe Manana 32867ffbb598SFilipe Manana em = alloc_extent_map(); 32877ffbb598SFilipe Manana if (!em) { 3288a2f392e4SNikolay Borisov set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags); 32897ffbb598SFilipe Manana return; 32907ffbb598SFilipe Manana } 32917ffbb598SFilipe Manana 329214f59796SFilipe Manana if (path) { 329314f59796SFilipe Manana struct btrfs_file_extent_item *fi; 329414f59796SFilipe Manana 329514f59796SFilipe Manana fi = btrfs_item_ptr(path->nodes[0], path->slots[0], 329614f59796SFilipe Manana struct btrfs_file_extent_item); 3297a2f392e4SNikolay Borisov btrfs_extent_item_to_extent_map(inode, path, fi, false, em); 32987ffbb598SFilipe Manana em->generation = -1; 32997ffbb598SFilipe Manana if (btrfs_file_extent_type(path->nodes[0], fi) == 33007ffbb598SFilipe Manana BTRFS_FILE_EXTENT_INLINE) 33017ffbb598SFilipe Manana set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, 3302a2f392e4SNikolay Borisov &inode->runtime_flags); 33037ffbb598SFilipe Manana } else { 33047ffbb598SFilipe Manana em->start = hole_offset; 33057ffbb598SFilipe Manana em->len = hole_len; 33067ffbb598SFilipe Manana em->ram_bytes = em->len; 33077ffbb598SFilipe Manana em->orig_start = hole_offset; 33087ffbb598SFilipe Manana em->block_start = EXTENT_MAP_HOLE; 33097ffbb598SFilipe Manana em->block_len = 0; 33107ffbb598SFilipe Manana em->orig_block_len = 0; 33117ffbb598SFilipe Manana em->compress_type = BTRFS_COMPRESS_NONE; 33127ffbb598SFilipe Manana em->generation = trans->transid; 33137ffbb598SFilipe Manana } 33147ffbb598SFilipe Manana 33157ffbb598SFilipe Manana while (1) { 33167ffbb598SFilipe Manana write_lock(&em_tree->lock); 33177ffbb598SFilipe Manana ret = add_extent_mapping(em_tree, em, 1); 33187ffbb598SFilipe Manana write_unlock(&em_tree->lock); 33197ffbb598SFilipe Manana if (ret != -EEXIST) { 33207ffbb598SFilipe Manana free_extent_map(em); 33217ffbb598SFilipe Manana break; 33227ffbb598SFilipe Manana } 3323a2f392e4SNikolay Borisov btrfs_drop_extent_cache(inode, em->start, 33247ffbb598SFilipe Manana em->start + em->len - 1, 0); 33257ffbb598SFilipe Manana } 33267ffbb598SFilipe Manana 3327ee39b432SDavid Sterba if (ret) 3328a2f392e4SNikolay Borisov set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags); 33297ffbb598SFilipe Manana } 33307ffbb598SFilipe Manana 33318039d87dSFilipe Manana /* 33328039d87dSFilipe Manana * Make sure we do not end up inserting an inline extent into a file that has 33338039d87dSFilipe Manana * already other (non-inline) extents. If a file has an inline extent it can 33348039d87dSFilipe Manana * not have any other extents and the (single) inline extent must start at the 33358039d87dSFilipe Manana * file offset 0. Failing to respect these rules will lead to file corruption, 33368039d87dSFilipe Manana * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc 33378039d87dSFilipe Manana * 33388039d87dSFilipe Manana * We can have extents that have been already written to disk or we can have 33398039d87dSFilipe Manana * dirty ranges still in delalloc, in which case the extent maps and items are 33408039d87dSFilipe Manana * created only when we run delalloc, and the delalloc ranges might fall outside 33418039d87dSFilipe Manana * the range we are currently locking in the inode's io tree. So we check the 33428039d87dSFilipe Manana * inode's i_size because of that (i_size updates are done while holding the 33438039d87dSFilipe Manana * i_mutex, which we are holding here). 33448039d87dSFilipe Manana * We also check to see if the inode has a size not greater than "datal" but has 33458039d87dSFilipe Manana * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are 33468039d87dSFilipe Manana * protected against such concurrent fallocate calls by the i_mutex). 33478039d87dSFilipe Manana * 33488039d87dSFilipe Manana * If the file has no extents but a size greater than datal, do not allow the 33498039d87dSFilipe Manana * copy because we would need turn the inline extent into a non-inline one (even 33508039d87dSFilipe Manana * with NO_HOLES enabled). If we find our destination inode only has one inline 33518039d87dSFilipe Manana * extent, just overwrite it with the source inline extent if its size is less 33528039d87dSFilipe Manana * than the source extent's size, or we could copy the source inline extent's 33538039d87dSFilipe Manana * data into the destination inode's inline extent if the later is greater then 33548039d87dSFilipe Manana * the former. 33558039d87dSFilipe Manana */ 33564a0ab9d7SDavid Sterba static int clone_copy_inline_extent(struct inode *dst, 33578039d87dSFilipe Manana struct btrfs_trans_handle *trans, 33588039d87dSFilipe Manana struct btrfs_path *path, 33598039d87dSFilipe Manana struct btrfs_key *new_key, 33608039d87dSFilipe Manana const u64 drop_start, 33618039d87dSFilipe Manana const u64 datal, 33628039d87dSFilipe Manana const u64 skip, 33638039d87dSFilipe Manana const u64 size, 33648039d87dSFilipe Manana char *inline_data) 33658039d87dSFilipe Manana { 33660b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(dst->i_sb); 33678039d87dSFilipe Manana struct btrfs_root *root = BTRFS_I(dst)->root; 33688039d87dSFilipe Manana const u64 aligned_end = ALIGN(new_key->offset + datal, 33690b246afaSJeff Mahoney fs_info->sectorsize); 33708039d87dSFilipe Manana int ret; 33718039d87dSFilipe Manana struct btrfs_key key; 33728039d87dSFilipe Manana 33738039d87dSFilipe Manana if (new_key->offset > 0) 33748039d87dSFilipe Manana return -EOPNOTSUPP; 33758039d87dSFilipe Manana 33764a0cc7caSNikolay Borisov key.objectid = btrfs_ino(BTRFS_I(dst)); 33778039d87dSFilipe Manana key.type = BTRFS_EXTENT_DATA_KEY; 33788039d87dSFilipe Manana key.offset = 0; 33798039d87dSFilipe Manana ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 33808039d87dSFilipe Manana if (ret < 0) { 33818039d87dSFilipe Manana return ret; 33828039d87dSFilipe Manana } else if (ret > 0) { 33838039d87dSFilipe Manana if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { 33848039d87dSFilipe Manana ret = btrfs_next_leaf(root, path); 33858039d87dSFilipe Manana if (ret < 0) 33868039d87dSFilipe Manana return ret; 33878039d87dSFilipe Manana else if (ret > 0) 33888039d87dSFilipe Manana goto copy_inline_extent; 33898039d87dSFilipe Manana } 33908039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 33914a0cc7caSNikolay Borisov if (key.objectid == btrfs_ino(BTRFS_I(dst)) && 33928039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) { 33938039d87dSFilipe Manana ASSERT(key.offset > 0); 33948039d87dSFilipe Manana return -EOPNOTSUPP; 33958039d87dSFilipe Manana } 33968039d87dSFilipe Manana } else if (i_size_read(dst) <= datal) { 33978039d87dSFilipe Manana struct btrfs_file_extent_item *ei; 33988039d87dSFilipe Manana u64 ext_len; 33998039d87dSFilipe Manana 34008039d87dSFilipe Manana /* 34018039d87dSFilipe Manana * If the file size is <= datal, make sure there are no other 34028039d87dSFilipe Manana * extents following (can happen do to an fallocate call with 34038039d87dSFilipe Manana * the flag FALLOC_FL_KEEP_SIZE). 34048039d87dSFilipe Manana */ 34058039d87dSFilipe Manana ei = btrfs_item_ptr(path->nodes[0], path->slots[0], 34068039d87dSFilipe Manana struct btrfs_file_extent_item); 34078039d87dSFilipe Manana /* 34088039d87dSFilipe Manana * If it's an inline extent, it can not have other extents 34098039d87dSFilipe Manana * following it. 34108039d87dSFilipe Manana */ 34118039d87dSFilipe Manana if (btrfs_file_extent_type(path->nodes[0], ei) == 34128039d87dSFilipe Manana BTRFS_FILE_EXTENT_INLINE) 34138039d87dSFilipe Manana goto copy_inline_extent; 34148039d87dSFilipe Manana 34158039d87dSFilipe Manana ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei); 34168039d87dSFilipe Manana if (ext_len > aligned_end) 34178039d87dSFilipe Manana return -EOPNOTSUPP; 34188039d87dSFilipe Manana 34198039d87dSFilipe Manana ret = btrfs_next_item(root, path); 34208039d87dSFilipe Manana if (ret < 0) { 34218039d87dSFilipe Manana return ret; 34228039d87dSFilipe Manana } else if (ret == 0) { 34238039d87dSFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 34248039d87dSFilipe Manana path->slots[0]); 34254a0cc7caSNikolay Borisov if (key.objectid == btrfs_ino(BTRFS_I(dst)) && 34268039d87dSFilipe Manana key.type == BTRFS_EXTENT_DATA_KEY) 34278039d87dSFilipe Manana return -EOPNOTSUPP; 34288039d87dSFilipe Manana } 34298039d87dSFilipe Manana } 34308039d87dSFilipe Manana 34318039d87dSFilipe Manana copy_inline_extent: 34328039d87dSFilipe Manana /* 34338039d87dSFilipe Manana * We have no extent items, or we have an extent at offset 0 which may 34348039d87dSFilipe Manana * or may not be inlined. All these cases are dealt the same way. 34358039d87dSFilipe Manana */ 34368039d87dSFilipe Manana if (i_size_read(dst) > datal) { 34378039d87dSFilipe Manana /* 34388039d87dSFilipe Manana * If the destination inode has an inline extent... 34398039d87dSFilipe Manana * This would require copying the data from the source inline 34408039d87dSFilipe Manana * extent into the beginning of the destination's inline extent. 34418039d87dSFilipe Manana * But this is really complex, both extents can be compressed 34428039d87dSFilipe Manana * or just one of them, which would require decompressing and 34438039d87dSFilipe Manana * re-compressing data (which could increase the new compressed 34448039d87dSFilipe Manana * size, not allowing the compressed data to fit anymore in an 34458039d87dSFilipe Manana * inline extent). 34468039d87dSFilipe Manana * So just don't support this case for now (it should be rare, 34478039d87dSFilipe Manana * we are not really saving space when cloning inline extents). 34488039d87dSFilipe Manana */ 34498039d87dSFilipe Manana return -EOPNOTSUPP; 34508039d87dSFilipe Manana } 34518039d87dSFilipe Manana 34528039d87dSFilipe Manana btrfs_release_path(path); 34538039d87dSFilipe Manana ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1); 34548039d87dSFilipe Manana if (ret) 34558039d87dSFilipe Manana return ret; 34568039d87dSFilipe Manana ret = btrfs_insert_empty_item(trans, root, path, new_key, size); 34578039d87dSFilipe Manana if (ret) 34588039d87dSFilipe Manana return ret; 34598039d87dSFilipe Manana 34608039d87dSFilipe Manana if (skip) { 34618039d87dSFilipe Manana const u32 start = btrfs_file_extent_calc_inline_size(0); 34628039d87dSFilipe Manana 34638039d87dSFilipe Manana memmove(inline_data + start, inline_data + start + skip, datal); 34648039d87dSFilipe Manana } 34658039d87dSFilipe Manana 34668039d87dSFilipe Manana write_extent_buffer(path->nodes[0], inline_data, 34678039d87dSFilipe Manana btrfs_item_ptr_offset(path->nodes[0], 34688039d87dSFilipe Manana path->slots[0]), 34698039d87dSFilipe Manana size); 34708039d87dSFilipe Manana inode_add_bytes(dst, datal); 34718039d87dSFilipe Manana 34728039d87dSFilipe Manana return 0; 34738039d87dSFilipe Manana } 34748039d87dSFilipe Manana 347532b7c687SMark Fasheh /** 347632b7c687SMark Fasheh * btrfs_clone() - clone a range from inode file to another 347732b7c687SMark Fasheh * 347832b7c687SMark Fasheh * @src: Inode to clone from 347932b7c687SMark Fasheh * @inode: Inode to clone to 348032b7c687SMark Fasheh * @off: Offset within source to start clone from 348132b7c687SMark Fasheh * @olen: Original length, passed by user, of range to clone 34821c919a5eSMark Fasheh * @olen_aligned: Block-aligned value of olen 348332b7c687SMark Fasheh * @destoff: Offset within @inode to start clone 34841c919a5eSMark Fasheh * @no_time_update: Whether to update mtime/ctime on the target inode 348532b7c687SMark Fasheh */ 348632b7c687SMark Fasheh static int btrfs_clone(struct inode *src, struct inode *inode, 3487f82a9901SFilipe Manana const u64 off, const u64 olen, const u64 olen_aligned, 34881c919a5eSMark Fasheh const u64 destoff, int no_time_update) 3489f46b5a66SChristoph Hellwig { 34900b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 3491f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 349232b7c687SMark Fasheh struct btrfs_path *path = NULL; 3493f46b5a66SChristoph Hellwig struct extent_buffer *leaf; 349432b7c687SMark Fasheh struct btrfs_trans_handle *trans; 349532b7c687SMark Fasheh char *buf = NULL; 3496ae01a0abSYan Zheng struct btrfs_key key; 3497f46b5a66SChristoph Hellwig u32 nritems; 3498f46b5a66SChristoph Hellwig int slot; 3499ae01a0abSYan Zheng int ret; 3500f82a9901SFilipe Manana const u64 len = olen_aligned; 3501f82a9901SFilipe Manana u64 last_dest_end = destoff; 3502ae01a0abSYan Zheng 3503ae01a0abSYan Zheng ret = -ENOMEM; 3504752ade68SMichal Hocko buf = kvmalloc(fs_info->nodesize, GFP_KERNEL); 3505ae01a0abSYan Zheng if (!buf) 350632b7c687SMark Fasheh return ret; 3507ae01a0abSYan Zheng 3508ae01a0abSYan Zheng path = btrfs_alloc_path(); 3509ae01a0abSYan Zheng if (!path) { 351015351955SDavid Sterba kvfree(buf); 351132b7c687SMark Fasheh return ret; 3512ae01a0abSYan Zheng } 351332b7c687SMark Fasheh 3514e4058b54SDavid Sterba path->reada = READA_FORWARD; 3515c5c9cd4dSSage Weil /* clone data */ 35164a0cc7caSNikolay Borisov key.objectid = btrfs_ino(BTRFS_I(src)); 3517ae01a0abSYan Zheng key.type = BTRFS_EXTENT_DATA_KEY; 35182c463823SFilipe Manana key.offset = off; 3519f46b5a66SChristoph Hellwig 3520f46b5a66SChristoph Hellwig while (1) { 3521de249e66SChris Mason u64 next_key_min_offset = key.offset + 1; 3522df858e76SFilipe Manana 3523f46b5a66SChristoph Hellwig /* 3524f46b5a66SChristoph Hellwig * note the key will change type as we walk through the 3525f46b5a66SChristoph Hellwig * tree. 3526f46b5a66SChristoph Hellwig */ 3527e4355f34SFilipe David Borba Manana path->leave_spinning = 1; 3528362a20c5SDavid Sterba ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path, 3529362a20c5SDavid Sterba 0, 0); 3530f46b5a66SChristoph Hellwig if (ret < 0) 3531f46b5a66SChristoph Hellwig goto out; 35322c463823SFilipe Manana /* 35332c463823SFilipe Manana * First search, if no extent item that starts at offset off was 35342c463823SFilipe Manana * found but the previous item is an extent item, it's possible 35352c463823SFilipe Manana * it might overlap our target range, therefore process it. 35362c463823SFilipe Manana */ 35372c463823SFilipe Manana if (key.offset == off && ret > 0 && path->slots[0] > 0) { 35382c463823SFilipe Manana btrfs_item_key_to_cpu(path->nodes[0], &key, 35392c463823SFilipe Manana path->slots[0] - 1); 35402c463823SFilipe Manana if (key.type == BTRFS_EXTENT_DATA_KEY) 35412c463823SFilipe Manana path->slots[0]--; 35422c463823SFilipe Manana } 3543f46b5a66SChristoph Hellwig 3544ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3545e4355f34SFilipe David Borba Manana process_slot: 3546ae01a0abSYan Zheng if (path->slots[0] >= nritems) { 3547362a20c5SDavid Sterba ret = btrfs_next_leaf(BTRFS_I(src)->root, path); 3548f46b5a66SChristoph Hellwig if (ret < 0) 3549f46b5a66SChristoph Hellwig goto out; 3550f46b5a66SChristoph Hellwig if (ret > 0) 3551f46b5a66SChristoph Hellwig break; 3552ae01a0abSYan Zheng nritems = btrfs_header_nritems(path->nodes[0]); 3553f46b5a66SChristoph Hellwig } 3554f46b5a66SChristoph Hellwig leaf = path->nodes[0]; 3555f46b5a66SChristoph Hellwig slot = path->slots[0]; 3556f46b5a66SChristoph Hellwig 3557ae01a0abSYan Zheng btrfs_item_key_to_cpu(leaf, &key, slot); 3558962a298fSDavid Sterba if (key.type > BTRFS_EXTENT_DATA_KEY || 35594a0cc7caSNikolay Borisov key.objectid != btrfs_ino(BTRFS_I(src))) 3560f46b5a66SChristoph Hellwig break; 3561f46b5a66SChristoph Hellwig 3562962a298fSDavid Sterba if (key.type == BTRFS_EXTENT_DATA_KEY) { 3563c5c9cd4dSSage Weil struct btrfs_file_extent_item *extent; 3564c5c9cd4dSSage Weil int type; 356531840ae1SZheng Yan u32 size; 356631840ae1SZheng Yan struct btrfs_key new_key; 3567c5c9cd4dSSage Weil u64 disko = 0, diskl = 0; 3568c5c9cd4dSSage Weil u64 datao = 0, datal = 0; 3569c5c9cd4dSSage Weil u8 comp; 3570f82a9901SFilipe Manana u64 drop_start; 357131840ae1SZheng Yan 3572c5c9cd4dSSage Weil extent = btrfs_item_ptr(leaf, slot, 3573c5c9cd4dSSage Weil struct btrfs_file_extent_item); 3574c5c9cd4dSSage Weil comp = btrfs_file_extent_compression(leaf, extent); 3575c5c9cd4dSSage Weil type = btrfs_file_extent_type(leaf, extent); 3576c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3577c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3578d397712bSChris Mason disko = btrfs_file_extent_disk_bytenr(leaf, 3579d397712bSChris Mason extent); 3580d397712bSChris Mason diskl = btrfs_file_extent_disk_num_bytes(leaf, 3581d397712bSChris Mason extent); 3582c5c9cd4dSSage Weil datao = btrfs_file_extent_offset(leaf, extent); 3583d397712bSChris Mason datal = btrfs_file_extent_num_bytes(leaf, 3584d397712bSChris Mason extent); 3585c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3586c5c9cd4dSSage Weil /* take upper bound, may be compressed */ 3587c5c9cd4dSSage Weil datal = btrfs_file_extent_ram_bytes(leaf, 3588c5c9cd4dSSage Weil extent); 3589c5c9cd4dSSage Weil } 359031840ae1SZheng Yan 35912c463823SFilipe Manana /* 35922c463823SFilipe Manana * The first search might have left us at an extent 35932c463823SFilipe Manana * item that ends before our target range's start, can 35942c463823SFilipe Manana * happen if we have holes and NO_HOLES feature enabled. 35952c463823SFilipe Manana */ 35962c463823SFilipe Manana if (key.offset + datal <= off) { 3597e4355f34SFilipe David Borba Manana path->slots[0]++; 3598e4355f34SFilipe David Borba Manana goto process_slot; 35992c463823SFilipe Manana } else if (key.offset >= off + len) { 36002c463823SFilipe Manana break; 3601e4355f34SFilipe David Borba Manana } 3602df858e76SFilipe Manana next_key_min_offset = key.offset + datal; 3603e4355f34SFilipe David Borba Manana size = btrfs_item_size_nr(leaf, slot); 3604e4355f34SFilipe David Borba Manana read_extent_buffer(leaf, buf, 3605e4355f34SFilipe David Borba Manana btrfs_item_ptr_offset(leaf, slot), 3606e4355f34SFilipe David Borba Manana size); 3607e4355f34SFilipe David Borba Manana 3608e4355f34SFilipe David Borba Manana btrfs_release_path(path); 3609e4355f34SFilipe David Borba Manana path->leave_spinning = 0; 3610c5c9cd4dSSage Weil 361131840ae1SZheng Yan memcpy(&new_key, &key, sizeof(new_key)); 36124a0cc7caSNikolay Borisov new_key.objectid = btrfs_ino(BTRFS_I(inode)); 36134d728ec7SLi Zefan if (off <= key.offset) 3614c5c9cd4dSSage Weil new_key.offset = key.offset + destoff - off; 36154d728ec7SLi Zefan else 36164d728ec7SLi Zefan new_key.offset = destoff; 3617c5c9cd4dSSage Weil 3618b6f3409bSSage Weil /* 3619f82a9901SFilipe Manana * Deal with a hole that doesn't have an extent item 3620f82a9901SFilipe Manana * that represents it (NO_HOLES feature enabled). 3621f82a9901SFilipe Manana * This hole is either in the middle of the cloning 3622f82a9901SFilipe Manana * range or at the beginning (fully overlaps it or 3623f82a9901SFilipe Manana * partially overlaps it). 3624f82a9901SFilipe Manana */ 3625f82a9901SFilipe Manana if (new_key.offset != last_dest_end) 3626f82a9901SFilipe Manana drop_start = last_dest_end; 3627f82a9901SFilipe Manana else 3628f82a9901SFilipe Manana drop_start = new_key.offset; 3629f82a9901SFilipe Manana 3630f82a9901SFilipe Manana /* 3631b6f3409bSSage Weil * 1 - adjusting old extent (we may have to split it) 3632b6f3409bSSage Weil * 1 - add new extent 3633b6f3409bSSage Weil * 1 - inode update 3634b6f3409bSSage Weil */ 3635b6f3409bSSage Weil trans = btrfs_start_transaction(root, 3); 3636a22285a6SYan, Zheng if (IS_ERR(trans)) { 3637a22285a6SYan, Zheng ret = PTR_ERR(trans); 3638a22285a6SYan, Zheng goto out; 3639a22285a6SYan, Zheng } 3640a22285a6SYan, Zheng 3641c8a894d7SChris Mason if (type == BTRFS_FILE_EXTENT_REG || 3642c8a894d7SChris Mason type == BTRFS_FILE_EXTENT_PREALLOC) { 3643d72c0842SLi Zefan /* 3644d72c0842SLi Zefan * a | --- range to clone ---| b 3645d72c0842SLi Zefan * | ------------- extent ------------- | 3646d72c0842SLi Zefan */ 3647d72c0842SLi Zefan 364893915584SAntonio Ospite /* subtract range b */ 3649d72c0842SLi Zefan if (key.offset + datal > off + len) 3650d72c0842SLi Zefan datal = off + len - key.offset; 3651d72c0842SLi Zefan 365293915584SAntonio Ospite /* subtract range a */ 3653a22285a6SYan, Zheng if (off > key.offset) { 3654a22285a6SYan, Zheng datao += off - key.offset; 3655a22285a6SYan, Zheng datal -= off - key.offset; 3656a22285a6SYan, Zheng } 3657a22285a6SYan, Zheng 36585dc562c5SJosef Bacik ret = btrfs_drop_extents(trans, root, inode, 3659f82a9901SFilipe Manana drop_start, 3660a22285a6SYan, Zheng new_key.offset + datal, 36612671485dSJosef Bacik 1); 366279787eaaSJeff Mahoney if (ret) { 36633f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 366400fdf13aSLiu Bo btrfs_abort_transaction(trans, 366566642832SJeff Mahoney ret); 36663a45bb20SJeff Mahoney btrfs_end_transaction(trans); 366779787eaaSJeff Mahoney goto out; 366879787eaaSJeff Mahoney } 3669a22285a6SYan, Zheng 367031840ae1SZheng Yan ret = btrfs_insert_empty_item(trans, root, path, 367131840ae1SZheng Yan &new_key, size); 367279787eaaSJeff Mahoney if (ret) { 367366642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 36743a45bb20SJeff Mahoney btrfs_end_transaction(trans); 367579787eaaSJeff Mahoney goto out; 367679787eaaSJeff Mahoney } 367731840ae1SZheng Yan 367831840ae1SZheng Yan leaf = path->nodes[0]; 367931840ae1SZheng Yan slot = path->slots[0]; 368031840ae1SZheng Yan write_extent_buffer(leaf, buf, 368131840ae1SZheng Yan btrfs_item_ptr_offset(leaf, slot), 368231840ae1SZheng Yan size); 3683ae01a0abSYan Zheng 3684f46b5a66SChristoph Hellwig extent = btrfs_item_ptr(leaf, slot, 3685f46b5a66SChristoph Hellwig struct btrfs_file_extent_item); 3686c5c9cd4dSSage Weil 3687c5c9cd4dSSage Weil /* disko == 0 means it's a hole */ 3688c5c9cd4dSSage Weil if (!disko) 3689c5c9cd4dSSage Weil datao = 0; 3690c5c9cd4dSSage Weil 3691c5c9cd4dSSage Weil btrfs_set_file_extent_offset(leaf, extent, 3692c5c9cd4dSSage Weil datao); 3693c5c9cd4dSSage Weil btrfs_set_file_extent_num_bytes(leaf, extent, 3694c5c9cd4dSSage Weil datal); 3695fcebe456SJosef Bacik 3696c5c9cd4dSSage Weil if (disko) { 3697c5c9cd4dSSage Weil inode_add_bytes(inode, datal); 36982ff7e61eSJeff Mahoney ret = btrfs_inc_extent_ref(trans, 36992ff7e61eSJeff Mahoney fs_info, 37005d4f98a2SYan Zheng disko, diskl, 0, 3701f46b5a66SChristoph Hellwig root->root_key.objectid, 37024a0cc7caSNikolay Borisov btrfs_ino(BTRFS_I(inode)), 3703b06c4bf5SFilipe Manana new_key.offset - datao); 370479787eaaSJeff Mahoney if (ret) { 370579787eaaSJeff Mahoney btrfs_abort_transaction(trans, 370679787eaaSJeff Mahoney ret); 37073a45bb20SJeff Mahoney btrfs_end_transaction(trans); 370879787eaaSJeff Mahoney goto out; 370979787eaaSJeff Mahoney 371079787eaaSJeff Mahoney } 3711f46b5a66SChristoph Hellwig } 3712c5c9cd4dSSage Weil } else if (type == BTRFS_FILE_EXTENT_INLINE) { 3713c5c9cd4dSSage Weil u64 skip = 0; 3714c5c9cd4dSSage Weil u64 trim = 0; 3715ed958762SFilipe Manana 3716c5c9cd4dSSage Weil if (off > key.offset) { 3717c5c9cd4dSSage Weil skip = off - key.offset; 3718c5c9cd4dSSage Weil new_key.offset += skip; 371931840ae1SZheng Yan } 3720d397712bSChris Mason 3721c5c9cd4dSSage Weil if (key.offset + datal > off + len) 3722c5c9cd4dSSage Weil trim = key.offset + datal - (off + len); 3723d397712bSChris Mason 3724c5c9cd4dSSage Weil if (comp && (skip || trim)) { 3725c5c9cd4dSSage Weil ret = -EINVAL; 37263a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3727c5c9cd4dSSage Weil goto out; 372831840ae1SZheng Yan } 3729c5c9cd4dSSage Weil size -= skip + trim; 3730c5c9cd4dSSage Weil datal -= skip + trim; 3731a22285a6SYan, Zheng 37324a0ab9d7SDavid Sterba ret = clone_copy_inline_extent(inode, 37338039d87dSFilipe Manana trans, path, 37348039d87dSFilipe Manana &new_key, 3735f82a9901SFilipe Manana drop_start, 37368039d87dSFilipe Manana datal, 37378039d87dSFilipe Manana skip, size, buf); 373879787eaaSJeff Mahoney if (ret) { 37393f9e3df8SDavid Sterba if (ret != -EOPNOTSUPP) 37403a29bc09SChris Mason btrfs_abort_transaction(trans, 374179787eaaSJeff Mahoney ret); 37423a45bb20SJeff Mahoney btrfs_end_transaction(trans); 374379787eaaSJeff Mahoney goto out; 374479787eaaSJeff Mahoney } 3745c5c9cd4dSSage Weil leaf = path->nodes[0]; 3746c5c9cd4dSSage Weil slot = path->slots[0]; 3747c5c9cd4dSSage Weil } 3748c5c9cd4dSSage Weil 37497ffbb598SFilipe Manana /* If we have an implicit hole (NO_HOLES feature). */ 37507ffbb598SFilipe Manana if (drop_start < new_key.offset) 3751a2f392e4SNikolay Borisov clone_update_extent_map(BTRFS_I(inode), trans, 375214f59796SFilipe Manana NULL, drop_start, 37537ffbb598SFilipe Manana new_key.offset - drop_start); 37547ffbb598SFilipe Manana 3755a2f392e4SNikolay Borisov clone_update_extent_map(BTRFS_I(inode), trans, 3756a2f392e4SNikolay Borisov path, 0, 0); 37577ffbb598SFilipe Manana 3758c5c9cd4dSSage Weil btrfs_mark_buffer_dirty(leaf); 3759b3b4aa74SDavid Sterba btrfs_release_path(path); 3760c5c9cd4dSSage Weil 376162e2390eSFilipe Manana last_dest_end = ALIGN(new_key.offset + datal, 37620b246afaSJeff Mahoney fs_info->sectorsize); 3763f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, 3764f82a9901SFilipe Manana last_dest_end, 37651c919a5eSMark Fasheh destoff, olen, 37661c919a5eSMark Fasheh no_time_update); 3767f82a9901SFilipe Manana if (ret) 376879787eaaSJeff Mahoney goto out; 37692c463823SFilipe Manana if (new_key.offset + datal >= destoff + len) 37702c463823SFilipe Manana break; 3771a22285a6SYan, Zheng } 3772b3b4aa74SDavid Sterba btrfs_release_path(path); 3773df858e76SFilipe Manana key.offset = next_key_min_offset; 377469ae5e44SWang Xiaoguang 377569ae5e44SWang Xiaoguang if (fatal_signal_pending(current)) { 377669ae5e44SWang Xiaoguang ret = -EINTR; 377769ae5e44SWang Xiaoguang goto out; 377869ae5e44SWang Xiaoguang } 3779ae01a0abSYan Zheng } 3780f46b5a66SChristoph Hellwig ret = 0; 378132b7c687SMark Fasheh 3782f82a9901SFilipe Manana if (last_dest_end < destoff + len) { 3783f82a9901SFilipe Manana /* 3784f82a9901SFilipe Manana * We have an implicit hole (NO_HOLES feature is enabled) that 3785f82a9901SFilipe Manana * fully or partially overlaps our cloning range at its end. 3786f82a9901SFilipe Manana */ 3787b3b4aa74SDavid Sterba btrfs_release_path(path); 3788f82a9901SFilipe Manana 3789f82a9901SFilipe Manana /* 3790f82a9901SFilipe Manana * 1 - remove extent(s) 3791f82a9901SFilipe Manana * 1 - inode update 3792f82a9901SFilipe Manana */ 3793f82a9901SFilipe Manana trans = btrfs_start_transaction(root, 2); 3794f82a9901SFilipe Manana if (IS_ERR(trans)) { 3795f82a9901SFilipe Manana ret = PTR_ERR(trans); 3796f82a9901SFilipe Manana goto out; 3797f82a9901SFilipe Manana } 3798f82a9901SFilipe Manana ret = btrfs_drop_extents(trans, root, inode, 3799f82a9901SFilipe Manana last_dest_end, destoff + len, 1); 3800f82a9901SFilipe Manana if (ret) { 3801f82a9901SFilipe Manana if (ret != -EOPNOTSUPP) 380266642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 38033a45bb20SJeff Mahoney btrfs_end_transaction(trans); 3804f82a9901SFilipe Manana goto out; 3805f82a9901SFilipe Manana } 3806a2f392e4SNikolay Borisov clone_update_extent_map(BTRFS_I(inode), trans, NULL, 3807a2f392e4SNikolay Borisov last_dest_end, 380814f59796SFilipe Manana destoff + len - last_dest_end); 3809f82a9901SFilipe Manana ret = clone_finish_inode_update(trans, inode, destoff + len, 38101c919a5eSMark Fasheh destoff, olen, no_time_update); 3811f82a9901SFilipe Manana } 3812f82a9901SFilipe Manana 3813f46b5a66SChristoph Hellwig out: 381432b7c687SMark Fasheh btrfs_free_path(path); 381515351955SDavid Sterba kvfree(buf); 381632b7c687SMark Fasheh return ret; 381732b7c687SMark Fasheh } 381832b7c687SMark Fasheh 38193db11b2eSZach Brown static noinline int btrfs_clone_files(struct file *file, struct file *file_src, 382032b7c687SMark Fasheh u64 off, u64 olen, u64 destoff) 382132b7c687SMark Fasheh { 382254563d41SAl Viro struct inode *inode = file_inode(file); 38233db11b2eSZach Brown struct inode *src = file_inode(file_src); 38240b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 382532b7c687SMark Fasheh struct btrfs_root *root = BTRFS_I(inode)->root; 382632b7c687SMark Fasheh int ret; 382732b7c687SMark Fasheh u64 len = olen; 38280b246afaSJeff Mahoney u64 bs = fs_info->sb->s_blocksize; 38293db11b2eSZach Brown int same_inode = src == inode; 383032b7c687SMark Fasheh 383132b7c687SMark Fasheh /* 383232b7c687SMark Fasheh * TODO: 383332b7c687SMark Fasheh * - split compressed inline extents. annoying: we need to 383432b7c687SMark Fasheh * decompress into destination's address_space (the file offset 383532b7c687SMark Fasheh * may change, so source mapping won't do), then recompress (or 383632b7c687SMark Fasheh * otherwise reinsert) a subrange. 383700fdf13aSLiu Bo * 383800fdf13aSLiu Bo * - split destination inode's inline extents. The inline extents can 383900fdf13aSLiu Bo * be either compressed or non-compressed. 384032b7c687SMark Fasheh */ 384132b7c687SMark Fasheh 384232b7c687SMark Fasheh if (btrfs_root_readonly(root)) 384332b7c687SMark Fasheh return -EROFS; 384432b7c687SMark Fasheh 38453db11b2eSZach Brown if (file_src->f_path.mnt != file->f_path.mnt || 38463db11b2eSZach Brown src->i_sb != inode->i_sb) 38473db11b2eSZach Brown return -EXDEV; 384832b7c687SMark Fasheh 384932b7c687SMark Fasheh /* don't make the dst file partly checksummed */ 385032b7c687SMark Fasheh if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) != 385132b7c687SMark Fasheh (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) 38523db11b2eSZach Brown return -EINVAL; 385332b7c687SMark Fasheh 385432b7c687SMark Fasheh if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode)) 38553db11b2eSZach Brown return -EISDIR; 385632b7c687SMark Fasheh 385732b7c687SMark Fasheh if (!same_inode) { 3858293a8489SMark Fasheh btrfs_double_inode_lock(src, inode); 385932b7c687SMark Fasheh } else { 38605955102cSAl Viro inode_lock(src); 386132b7c687SMark Fasheh } 386232b7c687SMark Fasheh 386332b7c687SMark Fasheh /* determine range to clone */ 386432b7c687SMark Fasheh ret = -EINVAL; 386532b7c687SMark Fasheh if (off + len > src->i_size || off + len < off) 386632b7c687SMark Fasheh goto out_unlock; 386732b7c687SMark Fasheh if (len == 0) 386832b7c687SMark Fasheh olen = len = src->i_size - off; 386932b7c687SMark Fasheh /* if we extend to eof, continue to block boundary */ 387032b7c687SMark Fasheh if (off + len == src->i_size) 387132b7c687SMark Fasheh len = ALIGN(src->i_size, bs) - off; 387232b7c687SMark Fasheh 3873ccccf3d6SFilipe Manana if (len == 0) { 3874ccccf3d6SFilipe Manana ret = 0; 3875ccccf3d6SFilipe Manana goto out_unlock; 3876ccccf3d6SFilipe Manana } 3877ccccf3d6SFilipe Manana 387832b7c687SMark Fasheh /* verify the end result is block aligned */ 387932b7c687SMark Fasheh if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) || 388032b7c687SMark Fasheh !IS_ALIGNED(destoff, bs)) 388132b7c687SMark Fasheh goto out_unlock; 388232b7c687SMark Fasheh 388332b7c687SMark Fasheh /* verify if ranges are overlapped within the same file */ 388432b7c687SMark Fasheh if (same_inode) { 388532b7c687SMark Fasheh if (destoff + len > off && destoff < off + len) 388632b7c687SMark Fasheh goto out_unlock; 388732b7c687SMark Fasheh } 388832b7c687SMark Fasheh 388932b7c687SMark Fasheh if (destoff > inode->i_size) { 389032b7c687SMark Fasheh ret = btrfs_cont_expand(inode, inode->i_size, destoff); 389132b7c687SMark Fasheh if (ret) 389232b7c687SMark Fasheh goto out_unlock; 389332b7c687SMark Fasheh } 389432b7c687SMark Fasheh 3895c125b8bfSFilipe Manana /* 3896c125b8bfSFilipe Manana * Lock the target range too. Right after we replace the file extent 3897c125b8bfSFilipe Manana * items in the fs tree (which now point to the cloned data), we might 3898c125b8bfSFilipe Manana * have a worker replace them with extent items relative to a write 3899c125b8bfSFilipe Manana * operation that was issued before this clone operation (i.e. confront 3900c125b8bfSFilipe Manana * with inode.c:btrfs_finish_ordered_io). 3901c125b8bfSFilipe Manana */ 3902c125b8bfSFilipe Manana if (same_inode) { 3903c125b8bfSFilipe Manana u64 lock_start = min_t(u64, off, destoff); 3904c125b8bfSFilipe Manana u64 lock_len = max_t(u64, off, destoff) + len - lock_start; 390532b7c687SMark Fasheh 3906e0bd70c6SFilipe Manana ret = lock_extent_range(src, lock_start, lock_len, true); 3907c125b8bfSFilipe Manana } else { 3908e0bd70c6SFilipe Manana ret = btrfs_double_extent_lock(src, off, inode, destoff, len, 3909e0bd70c6SFilipe Manana true); 3910e0bd70c6SFilipe Manana } 3911e0bd70c6SFilipe Manana ASSERT(ret == 0); 3912e0bd70c6SFilipe Manana if (WARN_ON(ret)) { 3913e0bd70c6SFilipe Manana /* ranges in the io trees already unlocked */ 3914e0bd70c6SFilipe Manana goto out_unlock; 3915c125b8bfSFilipe Manana } 391632b7c687SMark Fasheh 39171c919a5eSMark Fasheh ret = btrfs_clone(src, inode, off, olen, len, destoff, 0); 391832b7c687SMark Fasheh 3919c125b8bfSFilipe Manana if (same_inode) { 3920c125b8bfSFilipe Manana u64 lock_start = min_t(u64, off, destoff); 3921c125b8bfSFilipe Manana u64 lock_end = max_t(u64, off, destoff) + len - 1; 3922c125b8bfSFilipe Manana 3923c125b8bfSFilipe Manana unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end); 3924c125b8bfSFilipe Manana } else { 3925293a8489SMark Fasheh btrfs_double_extent_unlock(src, off, inode, destoff, len); 3926c125b8bfSFilipe Manana } 3927c125b8bfSFilipe Manana /* 3928c125b8bfSFilipe Manana * Truncate page cache pages so that future reads will see the cloned 3929c125b8bfSFilipe Manana * data immediately and not the previous data. 3930c125b8bfSFilipe Manana */ 393165bfa658SChandan Rajendra truncate_inode_pages_range(&inode->i_data, 393209cbfeafSKirill A. Shutemov round_down(destoff, PAGE_SIZE), 393309cbfeafSKirill A. Shutemov round_up(destoff + len, PAGE_SIZE) - 1); 3934f46b5a66SChristoph Hellwig out_unlock: 3935293a8489SMark Fasheh if (!same_inode) 3936293a8489SMark Fasheh btrfs_double_inode_unlock(src, inode); 3937293a8489SMark Fasheh else 39385955102cSAl Viro inode_unlock(src); 39393db11b2eSZach Brown return ret; 39403db11b2eSZach Brown } 39413db11b2eSZach Brown 394204b38d60SChristoph Hellwig int btrfs_clone_file_range(struct file *src_file, loff_t off, 394304b38d60SChristoph Hellwig struct file *dst_file, loff_t destoff, u64 len) 39443db11b2eSZach Brown { 394504b38d60SChristoph Hellwig return btrfs_clone_files(dst_file, src_file, off, len, destoff); 3946c5c9cd4dSSage Weil } 3947c5c9cd4dSSage Weil 3948f46b5a66SChristoph Hellwig /* 3949f46b5a66SChristoph Hellwig * there are many ways the trans_start and trans_end ioctls can lead 3950f46b5a66SChristoph Hellwig * to deadlocks. They should only be used by applications that 3951f46b5a66SChristoph Hellwig * basically own the machine, and have a very in depth understanding 3952f46b5a66SChristoph Hellwig * of all the possible deadlocks and enospc problems. 3953f46b5a66SChristoph Hellwig */ 3954b2950863SChristoph Hellwig static long btrfs_ioctl_trans_start(struct file *file) 3955f46b5a66SChristoph Hellwig { 3956496ad9aaSAl Viro struct inode *inode = file_inode(file); 39570b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 3958f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 3959f46b5a66SChristoph Hellwig struct btrfs_trans_handle *trans; 396023b5ec74SJosef Bacik struct btrfs_file_private *private; 39611ab86aedSSage Weil int ret; 39623558d4f8SNikolay Borisov static bool warned = false; 3963f46b5a66SChristoph Hellwig 39641ab86aedSSage Weil ret = -EPERM; 3965df5b5520SChristoph Hellwig if (!capable(CAP_SYS_ADMIN)) 3966f46b5a66SChristoph Hellwig goto out; 39671ab86aedSSage Weil 39683558d4f8SNikolay Borisov if (!warned) { 39693558d4f8SNikolay Borisov btrfs_warn(fs_info, 39703558d4f8SNikolay Borisov "Userspace transaction mechanism is considered " 39713558d4f8SNikolay Borisov "deprecated and slated to be removed in 4.17. " 39723558d4f8SNikolay Borisov "If you have a valid use case please " 39733558d4f8SNikolay Borisov "speak up on the mailing list"); 39743558d4f8SNikolay Borisov WARN_ON(1); 39753558d4f8SNikolay Borisov warned = true; 39763558d4f8SNikolay Borisov } 39773558d4f8SNikolay Borisov 39781ab86aedSSage Weil ret = -EINPROGRESS; 397923b5ec74SJosef Bacik private = file->private_data; 398023b5ec74SJosef Bacik if (private && private->trans) 39811ab86aedSSage Weil goto out; 398223b5ec74SJosef Bacik if (!private) { 398323b5ec74SJosef Bacik private = kzalloc(sizeof(struct btrfs_file_private), 398423b5ec74SJosef Bacik GFP_KERNEL); 398523b5ec74SJosef Bacik if (!private) 398623b5ec74SJosef Bacik return -ENOMEM; 398723b5ec74SJosef Bacik file->private_data = private; 398823b5ec74SJosef Bacik } 39899ca9ee09SSage Weil 3990b83cc969SLi Zefan ret = -EROFS; 3991b83cc969SLi Zefan if (btrfs_root_readonly(root)) 3992b83cc969SLi Zefan goto out; 3993b83cc969SLi Zefan 3994a561be71SAl Viro ret = mnt_want_write_file(file); 3995c146afadSYan Zheng if (ret) 3996c146afadSYan Zheng goto out; 3997c146afadSYan Zheng 39980b246afaSJeff Mahoney atomic_inc(&fs_info->open_ioctl_trans); 39999ca9ee09SSage Weil 4000f46b5a66SChristoph Hellwig ret = -ENOMEM; 40017a7eaa40SJosef Bacik trans = btrfs_start_ioctl_transaction(root); 4002abd30bb0STsutomu Itoh if (IS_ERR(trans)) 40031ab86aedSSage Weil goto out_drop; 40041ab86aedSSage Weil 400523b5ec74SJosef Bacik private->trans = trans; 40061ab86aedSSage Weil return 0; 40071ab86aedSSage Weil 40081ab86aedSSage Weil out_drop: 40090b246afaSJeff Mahoney atomic_dec(&fs_info->open_ioctl_trans); 40102a79f17eSAl Viro mnt_drop_write_file(file); 4011f46b5a66SChristoph Hellwig out: 4012f46b5a66SChristoph Hellwig return ret; 4013f46b5a66SChristoph Hellwig } 4014f46b5a66SChristoph Hellwig 40156ef5ed0dSJosef Bacik static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp) 40166ef5ed0dSJosef Bacik { 4017496ad9aaSAl Viro struct inode *inode = file_inode(file); 40180b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 40196ef5ed0dSJosef Bacik struct btrfs_root *root = BTRFS_I(inode)->root; 40206ef5ed0dSJosef Bacik struct btrfs_root *new_root; 40216ef5ed0dSJosef Bacik struct btrfs_dir_item *di; 40226ef5ed0dSJosef Bacik struct btrfs_trans_handle *trans; 40236ef5ed0dSJosef Bacik struct btrfs_path *path; 40246ef5ed0dSJosef Bacik struct btrfs_key location; 40256ef5ed0dSJosef Bacik struct btrfs_disk_key disk_key; 40266ef5ed0dSJosef Bacik u64 objectid = 0; 40276ef5ed0dSJosef Bacik u64 dir_id; 40283c04ce01SMiao Xie int ret; 40296ef5ed0dSJosef Bacik 40306ef5ed0dSJosef Bacik if (!capable(CAP_SYS_ADMIN)) 40316ef5ed0dSJosef Bacik return -EPERM; 40326ef5ed0dSJosef Bacik 40333c04ce01SMiao Xie ret = mnt_want_write_file(file); 40343c04ce01SMiao Xie if (ret) 40353c04ce01SMiao Xie return ret; 40363c04ce01SMiao Xie 40373c04ce01SMiao Xie if (copy_from_user(&objectid, argp, sizeof(objectid))) { 40383c04ce01SMiao Xie ret = -EFAULT; 40393c04ce01SMiao Xie goto out; 40403c04ce01SMiao Xie } 40416ef5ed0dSJosef Bacik 40426ef5ed0dSJosef Bacik if (!objectid) 40431cecf579Schandan objectid = BTRFS_FS_TREE_OBJECTID; 40446ef5ed0dSJosef Bacik 40456ef5ed0dSJosef Bacik location.objectid = objectid; 40466ef5ed0dSJosef Bacik location.type = BTRFS_ROOT_ITEM_KEY; 40476ef5ed0dSJosef Bacik location.offset = (u64)-1; 40486ef5ed0dSJosef Bacik 40490b246afaSJeff Mahoney new_root = btrfs_read_fs_root_no_name(fs_info, &location); 40503c04ce01SMiao Xie if (IS_ERR(new_root)) { 40513c04ce01SMiao Xie ret = PTR_ERR(new_root); 40523c04ce01SMiao Xie goto out; 40533c04ce01SMiao Xie } 40546d6d2829Ssatoru takeuchi if (!is_fstree(new_root->objectid)) { 40556d6d2829Ssatoru takeuchi ret = -ENOENT; 40566d6d2829Ssatoru takeuchi goto out; 40576d6d2829Ssatoru takeuchi } 40586ef5ed0dSJosef Bacik 40596ef5ed0dSJosef Bacik path = btrfs_alloc_path(); 40603c04ce01SMiao Xie if (!path) { 40613c04ce01SMiao Xie ret = -ENOMEM; 40623c04ce01SMiao Xie goto out; 40633c04ce01SMiao Xie } 40646ef5ed0dSJosef Bacik path->leave_spinning = 1; 40656ef5ed0dSJosef Bacik 40666ef5ed0dSJosef Bacik trans = btrfs_start_transaction(root, 1); 406798d5dc13STsutomu Itoh if (IS_ERR(trans)) { 40686ef5ed0dSJosef Bacik btrfs_free_path(path); 40693c04ce01SMiao Xie ret = PTR_ERR(trans); 40703c04ce01SMiao Xie goto out; 40716ef5ed0dSJosef Bacik } 40726ef5ed0dSJosef Bacik 40730b246afaSJeff Mahoney dir_id = btrfs_super_root_dir(fs_info->super_copy); 40740b246afaSJeff Mahoney di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path, 40756ef5ed0dSJosef Bacik dir_id, "default", 7, 1); 4076cf1e99a4SDan Carpenter if (IS_ERR_OR_NULL(di)) { 40776ef5ed0dSJosef Bacik btrfs_free_path(path); 40783a45bb20SJeff Mahoney btrfs_end_transaction(trans); 40790b246afaSJeff Mahoney btrfs_err(fs_info, 40805d163e0eSJeff Mahoney "Umm, you don't have the default diritem, this isn't going to work"); 40813c04ce01SMiao Xie ret = -ENOENT; 40823c04ce01SMiao Xie goto out; 40836ef5ed0dSJosef Bacik } 40846ef5ed0dSJosef Bacik 40856ef5ed0dSJosef Bacik btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key); 40866ef5ed0dSJosef Bacik btrfs_set_dir_item_key(path->nodes[0], di, &disk_key); 40876ef5ed0dSJosef Bacik btrfs_mark_buffer_dirty(path->nodes[0]); 40886ef5ed0dSJosef Bacik btrfs_free_path(path); 40896ef5ed0dSJosef Bacik 40900b246afaSJeff Mahoney btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL); 40913a45bb20SJeff Mahoney btrfs_end_transaction(trans); 40923c04ce01SMiao Xie out: 40933c04ce01SMiao Xie mnt_drop_write_file(file); 40943c04ce01SMiao Xie return ret; 40956ef5ed0dSJosef Bacik } 40966ef5ed0dSJosef Bacik 40975af3e8ccSStefan Behrens void btrfs_get_block_group_info(struct list_head *groups_list, 4098bf5fc093SJosef Bacik struct btrfs_ioctl_space_info *space) 4099bf5fc093SJosef Bacik { 4100bf5fc093SJosef Bacik struct btrfs_block_group_cache *block_group; 4101bf5fc093SJosef Bacik 4102bf5fc093SJosef Bacik space->total_bytes = 0; 4103bf5fc093SJosef Bacik space->used_bytes = 0; 4104bf5fc093SJosef Bacik space->flags = 0; 4105bf5fc093SJosef Bacik list_for_each_entry(block_group, groups_list, list) { 4106bf5fc093SJosef Bacik space->flags = block_group->flags; 4107bf5fc093SJosef Bacik space->total_bytes += block_group->key.offset; 4108bf5fc093SJosef Bacik space->used_bytes += 4109bf5fc093SJosef Bacik btrfs_block_group_used(&block_group->item); 4110bf5fc093SJosef Bacik } 4111bf5fc093SJosef Bacik } 4112bf5fc093SJosef Bacik 41132ff7e61eSJeff Mahoney static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info, 41142ff7e61eSJeff Mahoney void __user *arg) 41151406e432SJosef Bacik { 41161406e432SJosef Bacik struct btrfs_ioctl_space_args space_args; 41171406e432SJosef Bacik struct btrfs_ioctl_space_info space; 41181406e432SJosef Bacik struct btrfs_ioctl_space_info *dest; 41197fde62bfSChris Mason struct btrfs_ioctl_space_info *dest_orig; 412013f2696fSDaniel J Blueman struct btrfs_ioctl_space_info __user *user_dest; 41211406e432SJosef Bacik struct btrfs_space_info *info; 4122315d8e98SColin Ian King static const u64 types[] = { 4123315d8e98SColin Ian King BTRFS_BLOCK_GROUP_DATA, 4124bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_SYSTEM, 4125bf5fc093SJosef Bacik BTRFS_BLOCK_GROUP_METADATA, 4126315d8e98SColin Ian King BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA 4127315d8e98SColin Ian King }; 4128bf5fc093SJosef Bacik int num_types = 4; 41297fde62bfSChris Mason int alloc_size; 41301406e432SJosef Bacik int ret = 0; 413151788b1bSDan Rosenberg u64 slot_count = 0; 4132bf5fc093SJosef Bacik int i, c; 41331406e432SJosef Bacik 41341406e432SJosef Bacik if (copy_from_user(&space_args, 41351406e432SJosef Bacik (struct btrfs_ioctl_space_args __user *)arg, 41361406e432SJosef Bacik sizeof(space_args))) 41371406e432SJosef Bacik return -EFAULT; 41381406e432SJosef Bacik 4139bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4140bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4141bf5fc093SJosef Bacik 4142bf5fc093SJosef Bacik info = NULL; 41437fde62bfSChris Mason rcu_read_lock(); 41440b246afaSJeff Mahoney list_for_each_entry_rcu(tmp, &fs_info->space_info, 4145bf5fc093SJosef Bacik list) { 4146bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4147bf5fc093SJosef Bacik info = tmp; 4148bf5fc093SJosef Bacik break; 4149bf5fc093SJosef Bacik } 4150bf5fc093SJosef Bacik } 41517fde62bfSChris Mason rcu_read_unlock(); 41521406e432SJosef Bacik 4153bf5fc093SJosef Bacik if (!info) 4154bf5fc093SJosef Bacik continue; 4155bf5fc093SJosef Bacik 4156bf5fc093SJosef Bacik down_read(&info->groups_sem); 4157bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4158bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) 4159bf5fc093SJosef Bacik slot_count++; 4160bf5fc093SJosef Bacik } 4161bf5fc093SJosef Bacik up_read(&info->groups_sem); 4162bf5fc093SJosef Bacik } 4163bf5fc093SJosef Bacik 416436523e95SDavid Sterba /* 416536523e95SDavid Sterba * Global block reserve, exported as a space_info 416636523e95SDavid Sterba */ 416736523e95SDavid Sterba slot_count++; 416836523e95SDavid Sterba 41697fde62bfSChris Mason /* space_slots == 0 means they are asking for a count */ 41707fde62bfSChris Mason if (space_args.space_slots == 0) { 41717fde62bfSChris Mason space_args.total_spaces = slot_count; 41727fde62bfSChris Mason goto out; 41737fde62bfSChris Mason } 4174bf5fc093SJosef Bacik 417551788b1bSDan Rosenberg slot_count = min_t(u64, space_args.space_slots, slot_count); 4176bf5fc093SJosef Bacik 41777fde62bfSChris Mason alloc_size = sizeof(*dest) * slot_count; 4178bf5fc093SJosef Bacik 41797fde62bfSChris Mason /* we generally have at most 6 or so space infos, one for each raid 41807fde62bfSChris Mason * level. So, a whole page should be more than enough for everyone 41817fde62bfSChris Mason */ 418209cbfeafSKirill A. Shutemov if (alloc_size > PAGE_SIZE) 41837fde62bfSChris Mason return -ENOMEM; 41847fde62bfSChris Mason 41857fde62bfSChris Mason space_args.total_spaces = 0; 41868d2db785SDavid Sterba dest = kmalloc(alloc_size, GFP_KERNEL); 41877fde62bfSChris Mason if (!dest) 41887fde62bfSChris Mason return -ENOMEM; 41897fde62bfSChris Mason dest_orig = dest; 41907fde62bfSChris Mason 41917fde62bfSChris Mason /* now we have a buffer to copy into */ 4192bf5fc093SJosef Bacik for (i = 0; i < num_types; i++) { 4193bf5fc093SJosef Bacik struct btrfs_space_info *tmp; 4194bf5fc093SJosef Bacik 419551788b1bSDan Rosenberg if (!slot_count) 419651788b1bSDan Rosenberg break; 419751788b1bSDan Rosenberg 4198bf5fc093SJosef Bacik info = NULL; 41991406e432SJosef Bacik rcu_read_lock(); 42000b246afaSJeff Mahoney list_for_each_entry_rcu(tmp, &fs_info->space_info, 4201bf5fc093SJosef Bacik list) { 4202bf5fc093SJosef Bacik if (tmp->flags == types[i]) { 4203bf5fc093SJosef Bacik info = tmp; 42047fde62bfSChris Mason break; 4205bf5fc093SJosef Bacik } 4206bf5fc093SJosef Bacik } 4207bf5fc093SJosef Bacik rcu_read_unlock(); 42087fde62bfSChris Mason 4209bf5fc093SJosef Bacik if (!info) 4210bf5fc093SJosef Bacik continue; 4211bf5fc093SJosef Bacik down_read(&info->groups_sem); 4212bf5fc093SJosef Bacik for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 4213bf5fc093SJosef Bacik if (!list_empty(&info->block_groups[c])) { 42145af3e8ccSStefan Behrens btrfs_get_block_group_info( 42155af3e8ccSStefan Behrens &info->block_groups[c], &space); 42167fde62bfSChris Mason memcpy(dest, &space, sizeof(space)); 42171406e432SJosef Bacik dest++; 42181406e432SJosef Bacik space_args.total_spaces++; 421951788b1bSDan Rosenberg slot_count--; 42201406e432SJosef Bacik } 422151788b1bSDan Rosenberg if (!slot_count) 422251788b1bSDan Rosenberg break; 4223bf5fc093SJosef Bacik } 4224bf5fc093SJosef Bacik up_read(&info->groups_sem); 4225bf5fc093SJosef Bacik } 42261406e432SJosef Bacik 422736523e95SDavid Sterba /* 422836523e95SDavid Sterba * Add global block reserve 422936523e95SDavid Sterba */ 423036523e95SDavid Sterba if (slot_count) { 42310b246afaSJeff Mahoney struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv; 423236523e95SDavid Sterba 423336523e95SDavid Sterba spin_lock(&block_rsv->lock); 423436523e95SDavid Sterba space.total_bytes = block_rsv->size; 423536523e95SDavid Sterba space.used_bytes = block_rsv->size - block_rsv->reserved; 423636523e95SDavid Sterba spin_unlock(&block_rsv->lock); 423736523e95SDavid Sterba space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV; 423836523e95SDavid Sterba memcpy(dest, &space, sizeof(space)); 423936523e95SDavid Sterba space_args.total_spaces++; 424036523e95SDavid Sterba } 424136523e95SDavid Sterba 42422eec6c81SDaniel J Blueman user_dest = (struct btrfs_ioctl_space_info __user *) 42437fde62bfSChris Mason (arg + sizeof(struct btrfs_ioctl_space_args)); 42447fde62bfSChris Mason 42457fde62bfSChris Mason if (copy_to_user(user_dest, dest_orig, alloc_size)) 42467fde62bfSChris Mason ret = -EFAULT; 42477fde62bfSChris Mason 42487fde62bfSChris Mason kfree(dest_orig); 42497fde62bfSChris Mason out: 42507fde62bfSChris Mason if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args))) 42511406e432SJosef Bacik ret = -EFAULT; 42521406e432SJosef Bacik 42531406e432SJosef Bacik return ret; 42541406e432SJosef Bacik } 42551406e432SJosef Bacik 4256f46b5a66SChristoph Hellwig /* 4257f46b5a66SChristoph Hellwig * there are many ways the trans_start and trans_end ioctls can lead 4258f46b5a66SChristoph Hellwig * to deadlocks. They should only be used by applications that 4259f46b5a66SChristoph Hellwig * basically own the machine, and have a very in depth understanding 4260f46b5a66SChristoph Hellwig * of all the possible deadlocks and enospc problems. 4261f46b5a66SChristoph Hellwig */ 4262f46b5a66SChristoph Hellwig long btrfs_ioctl_trans_end(struct file *file) 4263f46b5a66SChristoph Hellwig { 4264496ad9aaSAl Viro struct inode *inode = file_inode(file); 4265f46b5a66SChristoph Hellwig struct btrfs_root *root = BTRFS_I(inode)->root; 426623b5ec74SJosef Bacik struct btrfs_file_private *private = file->private_data; 4267f46b5a66SChristoph Hellwig 426823b5ec74SJosef Bacik if (!private || !private->trans) 42691ab86aedSSage Weil return -EINVAL; 42709ca9ee09SSage Weil 427123b5ec74SJosef Bacik btrfs_end_transaction(private->trans); 427223b5ec74SJosef Bacik private->trans = NULL; 42731ab86aedSSage Weil 4274a4abeea4SJosef Bacik atomic_dec(&root->fs_info->open_ioctl_trans); 42759ca9ee09SSage Weil 42762a79f17eSAl Viro mnt_drop_write_file(file); 42771ab86aedSSage Weil return 0; 4278f46b5a66SChristoph Hellwig } 4279f46b5a66SChristoph Hellwig 42809a8c28beSMiao Xie static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root, 42819a8c28beSMiao Xie void __user *argp) 428246204592SSage Weil { 428346204592SSage Weil struct btrfs_trans_handle *trans; 428446204592SSage Weil u64 transid; 4285db5b493aSTsutomu Itoh int ret; 428646204592SSage Weil 4287d4edf39bSMiao Xie trans = btrfs_attach_transaction_barrier(root); 4288ff7c1d33SMiao Xie if (IS_ERR(trans)) { 4289ff7c1d33SMiao Xie if (PTR_ERR(trans) != -ENOENT) 429098d5dc13STsutomu Itoh return PTR_ERR(trans); 4291ff7c1d33SMiao Xie 4292ff7c1d33SMiao Xie /* No running transaction, don't bother */ 4293ff7c1d33SMiao Xie transid = root->fs_info->last_trans_committed; 4294ff7c1d33SMiao Xie goto out; 4295ff7c1d33SMiao Xie } 429646204592SSage Weil transid = trans->transid; 42973a45bb20SJeff Mahoney ret = btrfs_commit_transaction_async(trans, 0); 42988b2b2d3cSTsutomu Itoh if (ret) { 42993a45bb20SJeff Mahoney btrfs_end_transaction(trans); 4300db5b493aSTsutomu Itoh return ret; 43018b2b2d3cSTsutomu Itoh } 4302ff7c1d33SMiao Xie out: 430346204592SSage Weil if (argp) 430446204592SSage Weil if (copy_to_user(argp, &transid, sizeof(transid))) 430546204592SSage Weil return -EFAULT; 430646204592SSage Weil return 0; 430746204592SSage Weil } 430846204592SSage Weil 43092ff7e61eSJeff Mahoney static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info, 43109a8c28beSMiao Xie void __user *argp) 431146204592SSage Weil { 431246204592SSage Weil u64 transid; 431346204592SSage Weil 431446204592SSage Weil if (argp) { 431546204592SSage Weil if (copy_from_user(&transid, argp, sizeof(transid))) 431646204592SSage Weil return -EFAULT; 431746204592SSage Weil } else { 431846204592SSage Weil transid = 0; /* current trans */ 431946204592SSage Weil } 43202ff7e61eSJeff Mahoney return btrfs_wait_for_commit(fs_info, transid); 432146204592SSage Weil } 432246204592SSage Weil 4323b8e95489SMiao Xie static long btrfs_ioctl_scrub(struct file *file, void __user *arg) 4324475f6387SJan Schmidt { 43250b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb); 4326475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4327b8e95489SMiao Xie int ret; 4328475f6387SJan Schmidt 4329475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4330475f6387SJan Schmidt return -EPERM; 4331475f6387SJan Schmidt 4332475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4333475f6387SJan Schmidt if (IS_ERR(sa)) 4334475f6387SJan Schmidt return PTR_ERR(sa); 4335475f6387SJan Schmidt 4336b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) { 4337b8e95489SMiao Xie ret = mnt_want_write_file(file); 4338b8e95489SMiao Xie if (ret) 4339b8e95489SMiao Xie goto out; 4340b8e95489SMiao Xie } 4341b8e95489SMiao Xie 43420b246afaSJeff Mahoney ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end, 434363a212abSStefan Behrens &sa->progress, sa->flags & BTRFS_SCRUB_READONLY, 434463a212abSStefan Behrens 0); 4345475f6387SJan Schmidt 4346475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 4347475f6387SJan Schmidt ret = -EFAULT; 4348475f6387SJan Schmidt 4349b8e95489SMiao Xie if (!(sa->flags & BTRFS_SCRUB_READONLY)) 4350b8e95489SMiao Xie mnt_drop_write_file(file); 4351b8e95489SMiao Xie out: 4352475f6387SJan Schmidt kfree(sa); 4353475f6387SJan Schmidt return ret; 4354475f6387SJan Schmidt } 4355475f6387SJan Schmidt 43562ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info) 4357475f6387SJan Schmidt { 4358475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4359475f6387SJan Schmidt return -EPERM; 4360475f6387SJan Schmidt 43612ff7e61eSJeff Mahoney return btrfs_scrub_cancel(fs_info); 4362475f6387SJan Schmidt } 4363475f6387SJan Schmidt 43642ff7e61eSJeff Mahoney static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info, 4365475f6387SJan Schmidt void __user *arg) 4366475f6387SJan Schmidt { 4367475f6387SJan Schmidt struct btrfs_ioctl_scrub_args *sa; 4368475f6387SJan Schmidt int ret; 4369475f6387SJan Schmidt 4370475f6387SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4371475f6387SJan Schmidt return -EPERM; 4372475f6387SJan Schmidt 4373475f6387SJan Schmidt sa = memdup_user(arg, sizeof(*sa)); 4374475f6387SJan Schmidt if (IS_ERR(sa)) 4375475f6387SJan Schmidt return PTR_ERR(sa); 4376475f6387SJan Schmidt 43772ff7e61eSJeff Mahoney ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress); 4378475f6387SJan Schmidt 4379475f6387SJan Schmidt if (copy_to_user(arg, sa, sizeof(*sa))) 4380475f6387SJan Schmidt ret = -EFAULT; 4381475f6387SJan Schmidt 4382475f6387SJan Schmidt kfree(sa); 4383475f6387SJan Schmidt return ret; 4384475f6387SJan Schmidt } 4385475f6387SJan Schmidt 43862ff7e61eSJeff Mahoney static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info, 4387b27f7c0cSDavid Sterba void __user *arg) 4388c11d2c23SStefan Behrens { 4389c11d2c23SStefan Behrens struct btrfs_ioctl_get_dev_stats *sa; 4390c11d2c23SStefan Behrens int ret; 4391c11d2c23SStefan Behrens 4392c11d2c23SStefan Behrens sa = memdup_user(arg, sizeof(*sa)); 4393c11d2c23SStefan Behrens if (IS_ERR(sa)) 4394c11d2c23SStefan Behrens return PTR_ERR(sa); 4395c11d2c23SStefan Behrens 4396b27f7c0cSDavid Sterba if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) { 4397b27f7c0cSDavid Sterba kfree(sa); 4398b27f7c0cSDavid Sterba return -EPERM; 4399b27f7c0cSDavid Sterba } 4400b27f7c0cSDavid Sterba 44012ff7e61eSJeff Mahoney ret = btrfs_get_dev_stats(fs_info, sa); 4402c11d2c23SStefan Behrens 4403c11d2c23SStefan Behrens if (copy_to_user(arg, sa, sizeof(*sa))) 4404c11d2c23SStefan Behrens ret = -EFAULT; 4405c11d2c23SStefan Behrens 4406c11d2c23SStefan Behrens kfree(sa); 4407c11d2c23SStefan Behrens return ret; 4408c11d2c23SStefan Behrens } 4409c11d2c23SStefan Behrens 44102ff7e61eSJeff Mahoney static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info, 44112ff7e61eSJeff Mahoney void __user *arg) 44123f6bcfbdSStefan Behrens { 44133f6bcfbdSStefan Behrens struct btrfs_ioctl_dev_replace_args *p; 44143f6bcfbdSStefan Behrens int ret; 44153f6bcfbdSStefan Behrens 44163f6bcfbdSStefan Behrens if (!capable(CAP_SYS_ADMIN)) 44173f6bcfbdSStefan Behrens return -EPERM; 44183f6bcfbdSStefan Behrens 44193f6bcfbdSStefan Behrens p = memdup_user(arg, sizeof(*p)); 44203f6bcfbdSStefan Behrens if (IS_ERR(p)) 44213f6bcfbdSStefan Behrens return PTR_ERR(p); 44223f6bcfbdSStefan Behrens 44233f6bcfbdSStefan Behrens switch (p->cmd) { 44243f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_START: 4425bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 4426adfa97cbSIlya Dryomov ret = -EROFS; 4427adfa97cbSIlya Dryomov goto out; 4428adfa97cbSIlya Dryomov } 4429171938e5SDavid Sterba if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 4430e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 44313f6bcfbdSStefan Behrens } else { 44322ff7e61eSJeff Mahoney ret = btrfs_dev_replace_by_ioctl(fs_info, p); 4433171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 44343f6bcfbdSStefan Behrens } 44353f6bcfbdSStefan Behrens break; 44363f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS: 44370b246afaSJeff Mahoney btrfs_dev_replace_status(fs_info, p); 44383f6bcfbdSStefan Behrens ret = 0; 44393f6bcfbdSStefan Behrens break; 44403f6bcfbdSStefan Behrens case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL: 44410b246afaSJeff Mahoney ret = btrfs_dev_replace_cancel(fs_info, p); 44423f6bcfbdSStefan Behrens break; 44433f6bcfbdSStefan Behrens default: 44443f6bcfbdSStefan Behrens ret = -EINVAL; 44453f6bcfbdSStefan Behrens break; 44463f6bcfbdSStefan Behrens } 44473f6bcfbdSStefan Behrens 44483f6bcfbdSStefan Behrens if (copy_to_user(arg, p, sizeof(*p))) 44493f6bcfbdSStefan Behrens ret = -EFAULT; 4450adfa97cbSIlya Dryomov out: 44513f6bcfbdSStefan Behrens kfree(p); 44523f6bcfbdSStefan Behrens return ret; 44533f6bcfbdSStefan Behrens } 44543f6bcfbdSStefan Behrens 4455d7728c96SJan Schmidt static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg) 4456d7728c96SJan Schmidt { 4457d7728c96SJan Schmidt int ret = 0; 4458d7728c96SJan Schmidt int i; 4459740c3d22SChris Mason u64 rel_ptr; 4460d7728c96SJan Schmidt int size; 4461806468f8SChris Mason struct btrfs_ioctl_ino_path_args *ipa = NULL; 4462d7728c96SJan Schmidt struct inode_fs_paths *ipath = NULL; 4463d7728c96SJan Schmidt struct btrfs_path *path; 4464d7728c96SJan Schmidt 446582b22ac8SKusanagi Kouichi if (!capable(CAP_DAC_READ_SEARCH)) 4466d7728c96SJan Schmidt return -EPERM; 4467d7728c96SJan Schmidt 4468d7728c96SJan Schmidt path = btrfs_alloc_path(); 4469d7728c96SJan Schmidt if (!path) { 4470d7728c96SJan Schmidt ret = -ENOMEM; 4471d7728c96SJan Schmidt goto out; 4472d7728c96SJan Schmidt } 4473d7728c96SJan Schmidt 4474d7728c96SJan Schmidt ipa = memdup_user(arg, sizeof(*ipa)); 4475d7728c96SJan Schmidt if (IS_ERR(ipa)) { 4476d7728c96SJan Schmidt ret = PTR_ERR(ipa); 4477d7728c96SJan Schmidt ipa = NULL; 4478d7728c96SJan Schmidt goto out; 4479d7728c96SJan Schmidt } 4480d7728c96SJan Schmidt 4481d7728c96SJan Schmidt size = min_t(u32, ipa->size, 4096); 4482d7728c96SJan Schmidt ipath = init_ipath(size, root, path); 4483d7728c96SJan Schmidt if (IS_ERR(ipath)) { 4484d7728c96SJan Schmidt ret = PTR_ERR(ipath); 4485d7728c96SJan Schmidt ipath = NULL; 4486d7728c96SJan Schmidt goto out; 4487d7728c96SJan Schmidt } 4488d7728c96SJan Schmidt 4489d7728c96SJan Schmidt ret = paths_from_inode(ipa->inum, ipath); 4490d7728c96SJan Schmidt if (ret < 0) 4491d7728c96SJan Schmidt goto out; 4492d7728c96SJan Schmidt 4493d7728c96SJan Schmidt for (i = 0; i < ipath->fspath->elem_cnt; ++i) { 4494745c4d8eSJeff Mahoney rel_ptr = ipath->fspath->val[i] - 4495745c4d8eSJeff Mahoney (u64)(unsigned long)ipath->fspath->val; 4496740c3d22SChris Mason ipath->fspath->val[i] = rel_ptr; 4497d7728c96SJan Schmidt } 4498d7728c96SJan Schmidt 4499718dc5faSOmar Sandoval ret = copy_to_user((void __user *)(unsigned long)ipa->fspath, 4500718dc5faSOmar Sandoval ipath->fspath, size); 4501d7728c96SJan Schmidt if (ret) { 4502d7728c96SJan Schmidt ret = -EFAULT; 4503d7728c96SJan Schmidt goto out; 4504d7728c96SJan Schmidt } 4505d7728c96SJan Schmidt 4506d7728c96SJan Schmidt out: 4507d7728c96SJan Schmidt btrfs_free_path(path); 4508d7728c96SJan Schmidt free_ipath(ipath); 4509d7728c96SJan Schmidt kfree(ipa); 4510d7728c96SJan Schmidt 4511d7728c96SJan Schmidt return ret; 4512d7728c96SJan Schmidt } 4513d7728c96SJan Schmidt 4514d7728c96SJan Schmidt static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx) 4515d7728c96SJan Schmidt { 4516d7728c96SJan Schmidt struct btrfs_data_container *inodes = ctx; 4517d7728c96SJan Schmidt const size_t c = 3 * sizeof(u64); 4518d7728c96SJan Schmidt 4519d7728c96SJan Schmidt if (inodes->bytes_left >= c) { 4520d7728c96SJan Schmidt inodes->bytes_left -= c; 4521d7728c96SJan Schmidt inodes->val[inodes->elem_cnt] = inum; 4522d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 1] = offset; 4523d7728c96SJan Schmidt inodes->val[inodes->elem_cnt + 2] = root; 4524d7728c96SJan Schmidt inodes->elem_cnt += 3; 4525d7728c96SJan Schmidt } else { 4526d7728c96SJan Schmidt inodes->bytes_missing += c - inodes->bytes_left; 4527d7728c96SJan Schmidt inodes->bytes_left = 0; 4528d7728c96SJan Schmidt inodes->elem_missed += 3; 4529d7728c96SJan Schmidt } 4530d7728c96SJan Schmidt 4531d7728c96SJan Schmidt return 0; 4532d7728c96SJan Schmidt } 4533d7728c96SJan Schmidt 45342ff7e61eSJeff Mahoney static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info, 4535d7728c96SJan Schmidt void __user *arg) 4536d7728c96SJan Schmidt { 4537d7728c96SJan Schmidt int ret = 0; 4538d7728c96SJan Schmidt int size; 4539d7728c96SJan Schmidt struct btrfs_ioctl_logical_ino_args *loi; 4540d7728c96SJan Schmidt struct btrfs_data_container *inodes = NULL; 4541d7728c96SJan Schmidt struct btrfs_path *path = NULL; 4542d7728c96SJan Schmidt 4543d7728c96SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 4544d7728c96SJan Schmidt return -EPERM; 4545d7728c96SJan Schmidt 4546d7728c96SJan Schmidt loi = memdup_user(arg, sizeof(*loi)); 45477b9ea627SShailendra Verma if (IS_ERR(loi)) 45487b9ea627SShailendra Verma return PTR_ERR(loi); 4549d7728c96SJan Schmidt 4550d7728c96SJan Schmidt path = btrfs_alloc_path(); 4551d7728c96SJan Schmidt if (!path) { 4552d7728c96SJan Schmidt ret = -ENOMEM; 4553d7728c96SJan Schmidt goto out; 4554d7728c96SJan Schmidt } 4555d7728c96SJan Schmidt 4556ee22184bSByongho Lee size = min_t(u32, loi->size, SZ_64K); 4557d7728c96SJan Schmidt inodes = init_data_container(size); 4558d7728c96SJan Schmidt if (IS_ERR(inodes)) { 4559d7728c96SJan Schmidt ret = PTR_ERR(inodes); 4560d7728c96SJan Schmidt inodes = NULL; 4561d7728c96SJan Schmidt goto out; 4562d7728c96SJan Schmidt } 4563d7728c96SJan Schmidt 45642ff7e61eSJeff Mahoney ret = iterate_inodes_from_logical(loi->logical, fs_info, path, 4565df031f07SLiu Bo build_ino_list, inodes); 4566df031f07SLiu Bo if (ret == -EINVAL) 4567d7728c96SJan Schmidt ret = -ENOENT; 4568d7728c96SJan Schmidt if (ret < 0) 4569d7728c96SJan Schmidt goto out; 4570d7728c96SJan Schmidt 4571718dc5faSOmar Sandoval ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes, 4572718dc5faSOmar Sandoval size); 4573d7728c96SJan Schmidt if (ret) 4574d7728c96SJan Schmidt ret = -EFAULT; 4575d7728c96SJan Schmidt 4576d7728c96SJan Schmidt out: 4577d7728c96SJan Schmidt btrfs_free_path(path); 4578f54de068SDavid Sterba kvfree(inodes); 4579d7728c96SJan Schmidt kfree(loi); 4580d7728c96SJan Schmidt 4581d7728c96SJan Schmidt return ret; 4582d7728c96SJan Schmidt } 4583d7728c96SJan Schmidt 458419a39dceSIlya Dryomov void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock, 4585c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs) 4586c9e9f97bSIlya Dryomov { 4587c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl = fs_info->balance_ctl; 4588c9e9f97bSIlya Dryomov 4589c9e9f97bSIlya Dryomov bargs->flags = bctl->flags; 4590c9e9f97bSIlya Dryomov 4591837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_running)) 4592837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_RUNNING; 4593837d5b6eSIlya Dryomov if (atomic_read(&fs_info->balance_pause_req)) 4594837d5b6eSIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ; 4595a7e99c69SIlya Dryomov if (atomic_read(&fs_info->balance_cancel_req)) 4596a7e99c69SIlya Dryomov bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ; 4597837d5b6eSIlya Dryomov 4598c9e9f97bSIlya Dryomov memcpy(&bargs->data, &bctl->data, sizeof(bargs->data)); 4599c9e9f97bSIlya Dryomov memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta)); 4600c9e9f97bSIlya Dryomov memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys)); 460119a39dceSIlya Dryomov 460219a39dceSIlya Dryomov if (lock) { 460319a39dceSIlya Dryomov spin_lock(&fs_info->balance_lock); 460419a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 460519a39dceSIlya Dryomov spin_unlock(&fs_info->balance_lock); 460619a39dceSIlya Dryomov } else { 460719a39dceSIlya Dryomov memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat)); 460819a39dceSIlya Dryomov } 4609c9e9f97bSIlya Dryomov } 4610c9e9f97bSIlya Dryomov 46119ba1f6e4SLiu Bo static long btrfs_ioctl_balance(struct file *file, void __user *arg) 4612c9e9f97bSIlya Dryomov { 4613496ad9aaSAl Viro struct btrfs_root *root = BTRFS_I(file_inode(file))->root; 4614c9e9f97bSIlya Dryomov struct btrfs_fs_info *fs_info = root->fs_info; 4615c9e9f97bSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 4616c9e9f97bSIlya Dryomov struct btrfs_balance_control *bctl; 4617ed0fb78fSIlya Dryomov bool need_unlock; /* for mut. excl. ops lock */ 4618c9e9f97bSIlya Dryomov int ret; 4619c9e9f97bSIlya Dryomov 4620c9e9f97bSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4621c9e9f97bSIlya Dryomov return -EPERM; 4622c9e9f97bSIlya Dryomov 4623e54bfa31SLiu Bo ret = mnt_want_write_file(file); 46249ba1f6e4SLiu Bo if (ret) 46259ba1f6e4SLiu Bo return ret; 46269ba1f6e4SLiu Bo 4627ed0fb78fSIlya Dryomov again: 4628171938e5SDavid Sterba if (!test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) { 4629c9e9f97bSIlya Dryomov mutex_lock(&fs_info->volume_mutex); 4630c9e9f97bSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4631ed0fb78fSIlya Dryomov need_unlock = true; 4632ed0fb78fSIlya Dryomov goto locked; 4633ed0fb78fSIlya Dryomov } 4634ed0fb78fSIlya Dryomov 4635ed0fb78fSIlya Dryomov /* 463601327610SNicholas D Steeves * mut. excl. ops lock is locked. Three possibilities: 4637ed0fb78fSIlya Dryomov * (1) some other op is running 4638ed0fb78fSIlya Dryomov * (2) balance is running 4639ed0fb78fSIlya Dryomov * (3) balance is paused -- special case (think resume) 4640ed0fb78fSIlya Dryomov */ 4641ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4642ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4643ed0fb78fSIlya Dryomov /* this is either (2) or (3) */ 4644ed0fb78fSIlya Dryomov if (!atomic_read(&fs_info->balance_running)) { 4645ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4646ed0fb78fSIlya Dryomov if (!mutex_trylock(&fs_info->volume_mutex)) 4647ed0fb78fSIlya Dryomov goto again; 4648ed0fb78fSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 4649ed0fb78fSIlya Dryomov 4650ed0fb78fSIlya Dryomov if (fs_info->balance_ctl && 4651ed0fb78fSIlya Dryomov !atomic_read(&fs_info->balance_running)) { 4652ed0fb78fSIlya Dryomov /* this is (3) */ 4653ed0fb78fSIlya Dryomov need_unlock = false; 4654ed0fb78fSIlya Dryomov goto locked; 4655ed0fb78fSIlya Dryomov } 4656ed0fb78fSIlya Dryomov 4657ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4658ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->volume_mutex); 4659ed0fb78fSIlya Dryomov goto again; 4660ed0fb78fSIlya Dryomov } else { 4661ed0fb78fSIlya Dryomov /* this is (2) */ 4662ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4663ed0fb78fSIlya Dryomov ret = -EINPROGRESS; 4664ed0fb78fSIlya Dryomov goto out; 4665ed0fb78fSIlya Dryomov } 4666ed0fb78fSIlya Dryomov } else { 4667ed0fb78fSIlya Dryomov /* this is (1) */ 4668ed0fb78fSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4669e57138b3SAnand Jain ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS; 4670ed0fb78fSIlya Dryomov goto out; 4671ed0fb78fSIlya Dryomov } 4672ed0fb78fSIlya Dryomov 4673ed0fb78fSIlya Dryomov locked: 4674171938e5SDavid Sterba BUG_ON(!test_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)); 4675c9e9f97bSIlya Dryomov 4676c9e9f97bSIlya Dryomov if (arg) { 4677c9e9f97bSIlya Dryomov bargs = memdup_user(arg, sizeof(*bargs)); 4678c9e9f97bSIlya Dryomov if (IS_ERR(bargs)) { 4679c9e9f97bSIlya Dryomov ret = PTR_ERR(bargs); 4680ed0fb78fSIlya Dryomov goto out_unlock; 4681c9e9f97bSIlya Dryomov } 4682de322263SIlya Dryomov 4683de322263SIlya Dryomov if (bargs->flags & BTRFS_BALANCE_RESUME) { 4684de322263SIlya Dryomov if (!fs_info->balance_ctl) { 4685de322263SIlya Dryomov ret = -ENOTCONN; 4686de322263SIlya Dryomov goto out_bargs; 4687de322263SIlya Dryomov } 4688de322263SIlya Dryomov 4689de322263SIlya Dryomov bctl = fs_info->balance_ctl; 4690de322263SIlya Dryomov spin_lock(&fs_info->balance_lock); 4691de322263SIlya Dryomov bctl->flags |= BTRFS_BALANCE_RESUME; 4692de322263SIlya Dryomov spin_unlock(&fs_info->balance_lock); 4693de322263SIlya Dryomov 4694de322263SIlya Dryomov goto do_balance; 4695de322263SIlya Dryomov } 4696c9e9f97bSIlya Dryomov } else { 4697c9e9f97bSIlya Dryomov bargs = NULL; 4698c9e9f97bSIlya Dryomov } 4699c9e9f97bSIlya Dryomov 4700ed0fb78fSIlya Dryomov if (fs_info->balance_ctl) { 4701837d5b6eSIlya Dryomov ret = -EINPROGRESS; 4702837d5b6eSIlya Dryomov goto out_bargs; 4703837d5b6eSIlya Dryomov } 4704837d5b6eSIlya Dryomov 47058d2db785SDavid Sterba bctl = kzalloc(sizeof(*bctl), GFP_KERNEL); 4706c9e9f97bSIlya Dryomov if (!bctl) { 4707c9e9f97bSIlya Dryomov ret = -ENOMEM; 4708c9e9f97bSIlya Dryomov goto out_bargs; 4709c9e9f97bSIlya Dryomov } 4710c9e9f97bSIlya Dryomov 4711c9e9f97bSIlya Dryomov bctl->fs_info = fs_info; 4712c9e9f97bSIlya Dryomov if (arg) { 4713c9e9f97bSIlya Dryomov memcpy(&bctl->data, &bargs->data, sizeof(bctl->data)); 4714c9e9f97bSIlya Dryomov memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta)); 4715c9e9f97bSIlya Dryomov memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys)); 4716c9e9f97bSIlya Dryomov 4717c9e9f97bSIlya Dryomov bctl->flags = bargs->flags; 4718f43ffb60SIlya Dryomov } else { 4719f43ffb60SIlya Dryomov /* balance everything - no filters */ 4720f43ffb60SIlya Dryomov bctl->flags |= BTRFS_BALANCE_TYPE_MASK; 4721c9e9f97bSIlya Dryomov } 4722c9e9f97bSIlya Dryomov 47238eb93459SDavid Sterba if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) { 47248eb93459SDavid Sterba ret = -EINVAL; 47250f89abf5SChristian Engelmayer goto out_bctl; 47268eb93459SDavid Sterba } 47278eb93459SDavid Sterba 4728de322263SIlya Dryomov do_balance: 4729c9e9f97bSIlya Dryomov /* 4730171938e5SDavid Sterba * Ownership of bctl and filesystem flag BTRFS_FS_EXCL_OP 4731ed0fb78fSIlya Dryomov * goes to to btrfs_balance. bctl is freed in __cancel_balance, 4732ed0fb78fSIlya Dryomov * or, if restriper was paused all the way until unmount, in 4733171938e5SDavid Sterba * free_fs_info. The flag is cleared in __cancel_balance. 4734c9e9f97bSIlya Dryomov */ 4735ed0fb78fSIlya Dryomov need_unlock = false; 4736ed0fb78fSIlya Dryomov 4737ed0fb78fSIlya Dryomov ret = btrfs_balance(bctl, bargs); 47380f89abf5SChristian Engelmayer bctl = NULL; 4739ed0fb78fSIlya Dryomov 4740c9e9f97bSIlya Dryomov if (arg) { 4741c9e9f97bSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 4742c9e9f97bSIlya Dryomov ret = -EFAULT; 4743c9e9f97bSIlya Dryomov } 4744c9e9f97bSIlya Dryomov 47450f89abf5SChristian Engelmayer out_bctl: 47460f89abf5SChristian Engelmayer kfree(bctl); 4747c9e9f97bSIlya Dryomov out_bargs: 4748c9e9f97bSIlya Dryomov kfree(bargs); 4749ed0fb78fSIlya Dryomov out_unlock: 4750c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 4751c9e9f97bSIlya Dryomov mutex_unlock(&fs_info->volume_mutex); 4752ed0fb78fSIlya Dryomov if (need_unlock) 4753171938e5SDavid Sterba clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags); 4754ed0fb78fSIlya Dryomov out: 4755e54bfa31SLiu Bo mnt_drop_write_file(file); 4756c9e9f97bSIlya Dryomov return ret; 4757c9e9f97bSIlya Dryomov } 4758c9e9f97bSIlya Dryomov 47592ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd) 4760837d5b6eSIlya Dryomov { 4761837d5b6eSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 4762837d5b6eSIlya Dryomov return -EPERM; 4763837d5b6eSIlya Dryomov 4764837d5b6eSIlya Dryomov switch (cmd) { 4765837d5b6eSIlya Dryomov case BTRFS_BALANCE_CTL_PAUSE: 47660b246afaSJeff Mahoney return btrfs_pause_balance(fs_info); 4767a7e99c69SIlya Dryomov case BTRFS_BALANCE_CTL_CANCEL: 47680b246afaSJeff Mahoney return btrfs_cancel_balance(fs_info); 4769837d5b6eSIlya Dryomov } 4770837d5b6eSIlya Dryomov 4771837d5b6eSIlya Dryomov return -EINVAL; 4772837d5b6eSIlya Dryomov } 4773837d5b6eSIlya Dryomov 47742ff7e61eSJeff Mahoney static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info, 477519a39dceSIlya Dryomov void __user *arg) 477619a39dceSIlya Dryomov { 477719a39dceSIlya Dryomov struct btrfs_ioctl_balance_args *bargs; 477819a39dceSIlya Dryomov int ret = 0; 477919a39dceSIlya Dryomov 478019a39dceSIlya Dryomov if (!capable(CAP_SYS_ADMIN)) 478119a39dceSIlya Dryomov return -EPERM; 478219a39dceSIlya Dryomov 478319a39dceSIlya Dryomov mutex_lock(&fs_info->balance_mutex); 478419a39dceSIlya Dryomov if (!fs_info->balance_ctl) { 478519a39dceSIlya Dryomov ret = -ENOTCONN; 478619a39dceSIlya Dryomov goto out; 478719a39dceSIlya Dryomov } 478819a39dceSIlya Dryomov 47898d2db785SDavid Sterba bargs = kzalloc(sizeof(*bargs), GFP_KERNEL); 479019a39dceSIlya Dryomov if (!bargs) { 479119a39dceSIlya Dryomov ret = -ENOMEM; 479219a39dceSIlya Dryomov goto out; 479319a39dceSIlya Dryomov } 479419a39dceSIlya Dryomov 479519a39dceSIlya Dryomov update_ioctl_balance_args(fs_info, 1, bargs); 479619a39dceSIlya Dryomov 479719a39dceSIlya Dryomov if (copy_to_user(arg, bargs, sizeof(*bargs))) 479819a39dceSIlya Dryomov ret = -EFAULT; 479919a39dceSIlya Dryomov 480019a39dceSIlya Dryomov kfree(bargs); 480119a39dceSIlya Dryomov out: 480219a39dceSIlya Dryomov mutex_unlock(&fs_info->balance_mutex); 480319a39dceSIlya Dryomov return ret; 480419a39dceSIlya Dryomov } 480519a39dceSIlya Dryomov 4806905b0ddaSMiao Xie static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) 48075d13a37bSArne Jansen { 48080b246afaSJeff Mahoney struct inode *inode = file_inode(file); 48090b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 48105d13a37bSArne Jansen struct btrfs_ioctl_quota_ctl_args *sa; 48115d13a37bSArne Jansen struct btrfs_trans_handle *trans = NULL; 48125d13a37bSArne Jansen int ret; 48135d13a37bSArne Jansen int err; 48145d13a37bSArne Jansen 48155d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 48165d13a37bSArne Jansen return -EPERM; 48175d13a37bSArne Jansen 4818905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4819905b0ddaSMiao Xie if (ret) 4820905b0ddaSMiao Xie return ret; 48215d13a37bSArne Jansen 48225d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4823905b0ddaSMiao Xie if (IS_ERR(sa)) { 4824905b0ddaSMiao Xie ret = PTR_ERR(sa); 4825905b0ddaSMiao Xie goto drop_write; 4826905b0ddaSMiao Xie } 48275d13a37bSArne Jansen 48280b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 48290b246afaSJeff Mahoney trans = btrfs_start_transaction(fs_info->tree_root, 2); 48305d13a37bSArne Jansen if (IS_ERR(trans)) { 48315d13a37bSArne Jansen ret = PTR_ERR(trans); 48325d13a37bSArne Jansen goto out; 48335d13a37bSArne Jansen } 48345d13a37bSArne Jansen 48355d13a37bSArne Jansen switch (sa->cmd) { 48365d13a37bSArne Jansen case BTRFS_QUOTA_CTL_ENABLE: 48370b246afaSJeff Mahoney ret = btrfs_quota_enable(trans, fs_info); 48385d13a37bSArne Jansen break; 48395d13a37bSArne Jansen case BTRFS_QUOTA_CTL_DISABLE: 48400b246afaSJeff Mahoney ret = btrfs_quota_disable(trans, fs_info); 48415d13a37bSArne Jansen break; 48425d13a37bSArne Jansen default: 48435d13a37bSArne Jansen ret = -EINVAL; 48445d13a37bSArne Jansen break; 48455d13a37bSArne Jansen } 48465d13a37bSArne Jansen 48473a45bb20SJeff Mahoney err = btrfs_commit_transaction(trans); 48485d13a37bSArne Jansen if (err && !ret) 48495d13a37bSArne Jansen ret = err; 48505d13a37bSArne Jansen out: 48515d13a37bSArne Jansen kfree(sa); 48520b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 4853905b0ddaSMiao Xie drop_write: 4854905b0ddaSMiao Xie mnt_drop_write_file(file); 48555d13a37bSArne Jansen return ret; 48565d13a37bSArne Jansen } 48575d13a37bSArne Jansen 4858905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg) 48595d13a37bSArne Jansen { 48600b246afaSJeff Mahoney struct inode *inode = file_inode(file); 48610b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 48620b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 48635d13a37bSArne Jansen struct btrfs_ioctl_qgroup_assign_args *sa; 48645d13a37bSArne Jansen struct btrfs_trans_handle *trans; 48655d13a37bSArne Jansen int ret; 48665d13a37bSArne Jansen int err; 48675d13a37bSArne Jansen 48685d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 48695d13a37bSArne Jansen return -EPERM; 48705d13a37bSArne Jansen 4871905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4872905b0ddaSMiao Xie if (ret) 4873905b0ddaSMiao Xie return ret; 48745d13a37bSArne Jansen 48755d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4876905b0ddaSMiao Xie if (IS_ERR(sa)) { 4877905b0ddaSMiao Xie ret = PTR_ERR(sa); 4878905b0ddaSMiao Xie goto drop_write; 4879905b0ddaSMiao Xie } 48805d13a37bSArne Jansen 48815d13a37bSArne Jansen trans = btrfs_join_transaction(root); 48825d13a37bSArne Jansen if (IS_ERR(trans)) { 48835d13a37bSArne Jansen ret = PTR_ERR(trans); 48845d13a37bSArne Jansen goto out; 48855d13a37bSArne Jansen } 48865d13a37bSArne Jansen 48875d13a37bSArne Jansen if (sa->assign) { 48880b246afaSJeff Mahoney ret = btrfs_add_qgroup_relation(trans, fs_info, 48895d13a37bSArne Jansen sa->src, sa->dst); 48905d13a37bSArne Jansen } else { 48910b246afaSJeff Mahoney ret = btrfs_del_qgroup_relation(trans, fs_info, 48925d13a37bSArne Jansen sa->src, sa->dst); 48935d13a37bSArne Jansen } 48945d13a37bSArne Jansen 4895e082f563SQu Wenruo /* update qgroup status and info */ 48960b246afaSJeff Mahoney err = btrfs_run_qgroups(trans, fs_info); 4897e082f563SQu Wenruo if (err < 0) 48980b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, err, 4899ad8403dfSAnand Jain "failed to update qgroup status and info"); 49003a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 49015d13a37bSArne Jansen if (err && !ret) 49025d13a37bSArne Jansen ret = err; 49035d13a37bSArne Jansen 49045d13a37bSArne Jansen out: 49055d13a37bSArne Jansen kfree(sa); 4906905b0ddaSMiao Xie drop_write: 4907905b0ddaSMiao Xie mnt_drop_write_file(file); 49085d13a37bSArne Jansen return ret; 49095d13a37bSArne Jansen } 49105d13a37bSArne Jansen 4911905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg) 49125d13a37bSArne Jansen { 49130b246afaSJeff Mahoney struct inode *inode = file_inode(file); 49140b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 49150b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 49165d13a37bSArne Jansen struct btrfs_ioctl_qgroup_create_args *sa; 49175d13a37bSArne Jansen struct btrfs_trans_handle *trans; 49185d13a37bSArne Jansen int ret; 49195d13a37bSArne Jansen int err; 49205d13a37bSArne Jansen 49215d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 49225d13a37bSArne Jansen return -EPERM; 49235d13a37bSArne Jansen 4924905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4925905b0ddaSMiao Xie if (ret) 4926905b0ddaSMiao Xie return ret; 49275d13a37bSArne Jansen 49285d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4929905b0ddaSMiao Xie if (IS_ERR(sa)) { 4930905b0ddaSMiao Xie ret = PTR_ERR(sa); 4931905b0ddaSMiao Xie goto drop_write; 4932905b0ddaSMiao Xie } 49335d13a37bSArne Jansen 4934d86e56cfSMiao Xie if (!sa->qgroupid) { 4935d86e56cfSMiao Xie ret = -EINVAL; 4936d86e56cfSMiao Xie goto out; 4937d86e56cfSMiao Xie } 4938d86e56cfSMiao Xie 49395d13a37bSArne Jansen trans = btrfs_join_transaction(root); 49405d13a37bSArne Jansen if (IS_ERR(trans)) { 49415d13a37bSArne Jansen ret = PTR_ERR(trans); 49425d13a37bSArne Jansen goto out; 49435d13a37bSArne Jansen } 49445d13a37bSArne Jansen 49455d13a37bSArne Jansen if (sa->create) { 49460b246afaSJeff Mahoney ret = btrfs_create_qgroup(trans, fs_info, sa->qgroupid); 49475d13a37bSArne Jansen } else { 49480b246afaSJeff Mahoney ret = btrfs_remove_qgroup(trans, fs_info, sa->qgroupid); 49495d13a37bSArne Jansen } 49505d13a37bSArne Jansen 49513a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 49525d13a37bSArne Jansen if (err && !ret) 49535d13a37bSArne Jansen ret = err; 49545d13a37bSArne Jansen 49555d13a37bSArne Jansen out: 49565d13a37bSArne Jansen kfree(sa); 4957905b0ddaSMiao Xie drop_write: 4958905b0ddaSMiao Xie mnt_drop_write_file(file); 49595d13a37bSArne Jansen return ret; 49605d13a37bSArne Jansen } 49615d13a37bSArne Jansen 4962905b0ddaSMiao Xie static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg) 49635d13a37bSArne Jansen { 49640b246afaSJeff Mahoney struct inode *inode = file_inode(file); 49650b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 49660b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 49675d13a37bSArne Jansen struct btrfs_ioctl_qgroup_limit_args *sa; 49685d13a37bSArne Jansen struct btrfs_trans_handle *trans; 49695d13a37bSArne Jansen int ret; 49705d13a37bSArne Jansen int err; 49715d13a37bSArne Jansen u64 qgroupid; 49725d13a37bSArne Jansen 49735d13a37bSArne Jansen if (!capable(CAP_SYS_ADMIN)) 49745d13a37bSArne Jansen return -EPERM; 49755d13a37bSArne Jansen 4976905b0ddaSMiao Xie ret = mnt_want_write_file(file); 4977905b0ddaSMiao Xie if (ret) 4978905b0ddaSMiao Xie return ret; 49795d13a37bSArne Jansen 49805d13a37bSArne Jansen sa = memdup_user(arg, sizeof(*sa)); 4981905b0ddaSMiao Xie if (IS_ERR(sa)) { 4982905b0ddaSMiao Xie ret = PTR_ERR(sa); 4983905b0ddaSMiao Xie goto drop_write; 4984905b0ddaSMiao Xie } 49855d13a37bSArne Jansen 49865d13a37bSArne Jansen trans = btrfs_join_transaction(root); 49875d13a37bSArne Jansen if (IS_ERR(trans)) { 49885d13a37bSArne Jansen ret = PTR_ERR(trans); 49895d13a37bSArne Jansen goto out; 49905d13a37bSArne Jansen } 49915d13a37bSArne Jansen 49925d13a37bSArne Jansen qgroupid = sa->qgroupid; 49935d13a37bSArne Jansen if (!qgroupid) { 49945d13a37bSArne Jansen /* take the current subvol as qgroup */ 49955d13a37bSArne Jansen qgroupid = root->root_key.objectid; 49965d13a37bSArne Jansen } 49975d13a37bSArne Jansen 49980b246afaSJeff Mahoney ret = btrfs_limit_qgroup(trans, fs_info, qgroupid, &sa->lim); 49995d13a37bSArne Jansen 50003a45bb20SJeff Mahoney err = btrfs_end_transaction(trans); 50015d13a37bSArne Jansen if (err && !ret) 50025d13a37bSArne Jansen ret = err; 50035d13a37bSArne Jansen 50045d13a37bSArne Jansen out: 50055d13a37bSArne Jansen kfree(sa); 5006905b0ddaSMiao Xie drop_write: 5007905b0ddaSMiao Xie mnt_drop_write_file(file); 50085d13a37bSArne Jansen return ret; 50095d13a37bSArne Jansen } 50105d13a37bSArne Jansen 50112f232036SJan Schmidt static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) 50122f232036SJan Schmidt { 50130b246afaSJeff Mahoney struct inode *inode = file_inode(file); 50140b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 50152f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 50162f232036SJan Schmidt int ret; 50172f232036SJan Schmidt 50182f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 50192f232036SJan Schmidt return -EPERM; 50202f232036SJan Schmidt 50212f232036SJan Schmidt ret = mnt_want_write_file(file); 50222f232036SJan Schmidt if (ret) 50232f232036SJan Schmidt return ret; 50242f232036SJan Schmidt 50252f232036SJan Schmidt qsa = memdup_user(arg, sizeof(*qsa)); 50262f232036SJan Schmidt if (IS_ERR(qsa)) { 50272f232036SJan Schmidt ret = PTR_ERR(qsa); 50282f232036SJan Schmidt goto drop_write; 50292f232036SJan Schmidt } 50302f232036SJan Schmidt 50312f232036SJan Schmidt if (qsa->flags) { 50322f232036SJan Schmidt ret = -EINVAL; 50332f232036SJan Schmidt goto out; 50342f232036SJan Schmidt } 50352f232036SJan Schmidt 50360b246afaSJeff Mahoney ret = btrfs_qgroup_rescan(fs_info); 50372f232036SJan Schmidt 50382f232036SJan Schmidt out: 50392f232036SJan Schmidt kfree(qsa); 50402f232036SJan Schmidt drop_write: 50412f232036SJan Schmidt mnt_drop_write_file(file); 50422f232036SJan Schmidt return ret; 50432f232036SJan Schmidt } 50442f232036SJan Schmidt 50452f232036SJan Schmidt static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg) 50462f232036SJan Schmidt { 50470b246afaSJeff Mahoney struct inode *inode = file_inode(file); 50480b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 50492f232036SJan Schmidt struct btrfs_ioctl_quota_rescan_args *qsa; 50502f232036SJan Schmidt int ret = 0; 50512f232036SJan Schmidt 50522f232036SJan Schmidt if (!capable(CAP_SYS_ADMIN)) 50532f232036SJan Schmidt return -EPERM; 50542f232036SJan Schmidt 50558d2db785SDavid Sterba qsa = kzalloc(sizeof(*qsa), GFP_KERNEL); 50562f232036SJan Schmidt if (!qsa) 50572f232036SJan Schmidt return -ENOMEM; 50582f232036SJan Schmidt 50590b246afaSJeff Mahoney if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { 50602f232036SJan Schmidt qsa->flags = 1; 50610b246afaSJeff Mahoney qsa->progress = fs_info->qgroup_rescan_progress.objectid; 50622f232036SJan Schmidt } 50632f232036SJan Schmidt 50642f232036SJan Schmidt if (copy_to_user(arg, qsa, sizeof(*qsa))) 50652f232036SJan Schmidt ret = -EFAULT; 50662f232036SJan Schmidt 50672f232036SJan Schmidt kfree(qsa); 50682f232036SJan Schmidt return ret; 50692f232036SJan Schmidt } 50702f232036SJan Schmidt 507157254b6eSJan Schmidt static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg) 507257254b6eSJan Schmidt { 50730b246afaSJeff Mahoney struct inode *inode = file_inode(file); 50740b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 507557254b6eSJan Schmidt 507657254b6eSJan Schmidt if (!capable(CAP_SYS_ADMIN)) 507757254b6eSJan Schmidt return -EPERM; 507857254b6eSJan Schmidt 50790b246afaSJeff Mahoney return btrfs_qgroup_wait_for_completion(fs_info, true); 508057254b6eSJan Schmidt } 508157254b6eSJan Schmidt 5082abccd00fSHugo Mills static long _btrfs_ioctl_set_received_subvol(struct file *file, 5083abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa) 50848ea05e3aSAlexander Block { 5085496ad9aaSAl Viro struct inode *inode = file_inode(file); 50860b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 50878ea05e3aSAlexander Block struct btrfs_root *root = BTRFS_I(inode)->root; 50888ea05e3aSAlexander Block struct btrfs_root_item *root_item = &root->root_item; 50898ea05e3aSAlexander Block struct btrfs_trans_handle *trans; 5090c2050a45SDeepa Dinamani struct timespec ct = current_time(inode); 50918ea05e3aSAlexander Block int ret = 0; 5092dd5f9615SStefan Behrens int received_uuid_changed; 50938ea05e3aSAlexander Block 5094bd60ea0fSDavid Sterba if (!inode_owner_or_capable(inode)) 5095bd60ea0fSDavid Sterba return -EPERM; 5096bd60ea0fSDavid Sterba 50978ea05e3aSAlexander Block ret = mnt_want_write_file(file); 50988ea05e3aSAlexander Block if (ret < 0) 50998ea05e3aSAlexander Block return ret; 51008ea05e3aSAlexander Block 51010b246afaSJeff Mahoney down_write(&fs_info->subvol_sem); 51028ea05e3aSAlexander Block 51034a0cc7caSNikolay Borisov if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) { 51048ea05e3aSAlexander Block ret = -EINVAL; 51058ea05e3aSAlexander Block goto out; 51068ea05e3aSAlexander Block } 51078ea05e3aSAlexander Block 51088ea05e3aSAlexander Block if (btrfs_root_readonly(root)) { 51098ea05e3aSAlexander Block ret = -EROFS; 51108ea05e3aSAlexander Block goto out; 51118ea05e3aSAlexander Block } 51128ea05e3aSAlexander Block 5113dd5f9615SStefan Behrens /* 5114dd5f9615SStefan Behrens * 1 - root item 5115dd5f9615SStefan Behrens * 2 - uuid items (received uuid + subvol uuid) 5116dd5f9615SStefan Behrens */ 5117dd5f9615SStefan Behrens trans = btrfs_start_transaction(root, 3); 51188ea05e3aSAlexander Block if (IS_ERR(trans)) { 51198ea05e3aSAlexander Block ret = PTR_ERR(trans); 51208ea05e3aSAlexander Block trans = NULL; 51218ea05e3aSAlexander Block goto out; 51228ea05e3aSAlexander Block } 51238ea05e3aSAlexander Block 51248ea05e3aSAlexander Block sa->rtransid = trans->transid; 51258ea05e3aSAlexander Block sa->rtime.sec = ct.tv_sec; 51268ea05e3aSAlexander Block sa->rtime.nsec = ct.tv_nsec; 51278ea05e3aSAlexander Block 5128dd5f9615SStefan Behrens received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid, 5129dd5f9615SStefan Behrens BTRFS_UUID_SIZE); 5130dd5f9615SStefan Behrens if (received_uuid_changed && 5131dd5f9615SStefan Behrens !btrfs_is_empty_uuid(root_item->received_uuid)) 51320b246afaSJeff Mahoney btrfs_uuid_tree_rem(trans, fs_info, root_item->received_uuid, 5133dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5134dd5f9615SStefan Behrens root->root_key.objectid); 51358ea05e3aSAlexander Block memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE); 51368ea05e3aSAlexander Block btrfs_set_root_stransid(root_item, sa->stransid); 51378ea05e3aSAlexander Block btrfs_set_root_rtransid(root_item, sa->rtransid); 51383cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec); 51393cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec); 51403cae210fSQu Wenruo btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec); 51413cae210fSQu Wenruo btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec); 51428ea05e3aSAlexander Block 51430b246afaSJeff Mahoney ret = btrfs_update_root(trans, fs_info->tree_root, 51448ea05e3aSAlexander Block &root->root_key, &root->root_item); 51458ea05e3aSAlexander Block if (ret < 0) { 51463a45bb20SJeff Mahoney btrfs_end_transaction(trans); 51478ea05e3aSAlexander Block goto out; 5148dd5f9615SStefan Behrens } 5149dd5f9615SStefan Behrens if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) { 51500b246afaSJeff Mahoney ret = btrfs_uuid_tree_add(trans, fs_info, sa->uuid, 5151dd5f9615SStefan Behrens BTRFS_UUID_KEY_RECEIVED_SUBVOL, 5152dd5f9615SStefan Behrens root->root_key.objectid); 5153dd5f9615SStefan Behrens if (ret < 0 && ret != -EEXIST) { 515466642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 5155dd5f9615SStefan Behrens goto out; 5156dd5f9615SStefan Behrens } 5157dd5f9615SStefan Behrens } 51583a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 5159dd5f9615SStefan Behrens if (ret < 0) { 516066642832SJeff Mahoney btrfs_abort_transaction(trans, ret); 51618ea05e3aSAlexander Block goto out; 51628ea05e3aSAlexander Block } 51638ea05e3aSAlexander Block 5164abccd00fSHugo Mills out: 51650b246afaSJeff Mahoney up_write(&fs_info->subvol_sem); 5166abccd00fSHugo Mills mnt_drop_write_file(file); 5167abccd00fSHugo Mills return ret; 5168abccd00fSHugo Mills } 5169abccd00fSHugo Mills 5170abccd00fSHugo Mills #ifdef CONFIG_64BIT 5171abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol_32(struct file *file, 5172abccd00fSHugo Mills void __user *arg) 5173abccd00fSHugo Mills { 5174abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL; 5175abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *args64 = NULL; 5176abccd00fSHugo Mills int ret = 0; 5177abccd00fSHugo Mills 5178abccd00fSHugo Mills args32 = memdup_user(arg, sizeof(*args32)); 51797b9ea627SShailendra Verma if (IS_ERR(args32)) 51807b9ea627SShailendra Verma return PTR_ERR(args32); 5181abccd00fSHugo Mills 51828d2db785SDavid Sterba args64 = kmalloc(sizeof(*args64), GFP_KERNEL); 518384dbeb87SDan Carpenter if (!args64) { 518484dbeb87SDan Carpenter ret = -ENOMEM; 5185abccd00fSHugo Mills goto out; 5186abccd00fSHugo Mills } 5187abccd00fSHugo Mills 5188abccd00fSHugo Mills memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE); 5189abccd00fSHugo Mills args64->stransid = args32->stransid; 5190abccd00fSHugo Mills args64->rtransid = args32->rtransid; 5191abccd00fSHugo Mills args64->stime.sec = args32->stime.sec; 5192abccd00fSHugo Mills args64->stime.nsec = args32->stime.nsec; 5193abccd00fSHugo Mills args64->rtime.sec = args32->rtime.sec; 5194abccd00fSHugo Mills args64->rtime.nsec = args32->rtime.nsec; 5195abccd00fSHugo Mills args64->flags = args32->flags; 5196abccd00fSHugo Mills 5197abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, args64); 5198abccd00fSHugo Mills if (ret) 5199abccd00fSHugo Mills goto out; 5200abccd00fSHugo Mills 5201abccd00fSHugo Mills memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE); 5202abccd00fSHugo Mills args32->stransid = args64->stransid; 5203abccd00fSHugo Mills args32->rtransid = args64->rtransid; 5204abccd00fSHugo Mills args32->stime.sec = args64->stime.sec; 5205abccd00fSHugo Mills args32->stime.nsec = args64->stime.nsec; 5206abccd00fSHugo Mills args32->rtime.sec = args64->rtime.sec; 5207abccd00fSHugo Mills args32->rtime.nsec = args64->rtime.nsec; 5208abccd00fSHugo Mills args32->flags = args64->flags; 5209abccd00fSHugo Mills 5210abccd00fSHugo Mills ret = copy_to_user(arg, args32, sizeof(*args32)); 5211abccd00fSHugo Mills if (ret) 5212abccd00fSHugo Mills ret = -EFAULT; 5213abccd00fSHugo Mills 5214abccd00fSHugo Mills out: 5215abccd00fSHugo Mills kfree(args32); 5216abccd00fSHugo Mills kfree(args64); 5217abccd00fSHugo Mills return ret; 5218abccd00fSHugo Mills } 5219abccd00fSHugo Mills #endif 5220abccd00fSHugo Mills 5221abccd00fSHugo Mills static long btrfs_ioctl_set_received_subvol(struct file *file, 5222abccd00fSHugo Mills void __user *arg) 5223abccd00fSHugo Mills { 5224abccd00fSHugo Mills struct btrfs_ioctl_received_subvol_args *sa = NULL; 5225abccd00fSHugo Mills int ret = 0; 5226abccd00fSHugo Mills 5227abccd00fSHugo Mills sa = memdup_user(arg, sizeof(*sa)); 52287b9ea627SShailendra Verma if (IS_ERR(sa)) 52297b9ea627SShailendra Verma return PTR_ERR(sa); 5230abccd00fSHugo Mills 5231abccd00fSHugo Mills ret = _btrfs_ioctl_set_received_subvol(file, sa); 5232abccd00fSHugo Mills 5233abccd00fSHugo Mills if (ret) 5234abccd00fSHugo Mills goto out; 5235abccd00fSHugo Mills 52368ea05e3aSAlexander Block ret = copy_to_user(arg, sa, sizeof(*sa)); 52378ea05e3aSAlexander Block if (ret) 52388ea05e3aSAlexander Block ret = -EFAULT; 52398ea05e3aSAlexander Block 52408ea05e3aSAlexander Block out: 52418ea05e3aSAlexander Block kfree(sa); 52428ea05e3aSAlexander Block return ret; 52438ea05e3aSAlexander Block } 52448ea05e3aSAlexander Block 5245867ab667Sjeff.liu static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg) 5246867ab667Sjeff.liu { 52470b246afaSJeff Mahoney struct inode *inode = file_inode(file); 52480b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 5249a1b83ac5SAnand Jain size_t len; 5250867ab667Sjeff.liu int ret; 5251a1b83ac5SAnand Jain char label[BTRFS_LABEL_SIZE]; 5252a1b83ac5SAnand Jain 52530b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 52540b246afaSJeff Mahoney memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE); 52550b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 5256a1b83ac5SAnand Jain 5257a1b83ac5SAnand Jain len = strnlen(label, BTRFS_LABEL_SIZE); 5258867ab667Sjeff.liu 5259867ab667Sjeff.liu if (len == BTRFS_LABEL_SIZE) { 52600b246afaSJeff Mahoney btrfs_warn(fs_info, 52610b246afaSJeff Mahoney "label is too long, return the first %zu bytes", 52620b246afaSJeff Mahoney --len); 5263867ab667Sjeff.liu } 5264867ab667Sjeff.liu 5265867ab667Sjeff.liu ret = copy_to_user(arg, label, len); 5266867ab667Sjeff.liu 5267867ab667Sjeff.liu return ret ? -EFAULT : 0; 5268867ab667Sjeff.liu } 5269867ab667Sjeff.liu 5270a8bfd4abSjeff.liu static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg) 5271a8bfd4abSjeff.liu { 52720b246afaSJeff Mahoney struct inode *inode = file_inode(file); 52730b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 52740b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 52750b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 5276a8bfd4abSjeff.liu struct btrfs_trans_handle *trans; 5277a8bfd4abSjeff.liu char label[BTRFS_LABEL_SIZE]; 5278a8bfd4abSjeff.liu int ret; 5279a8bfd4abSjeff.liu 5280a8bfd4abSjeff.liu if (!capable(CAP_SYS_ADMIN)) 5281a8bfd4abSjeff.liu return -EPERM; 5282a8bfd4abSjeff.liu 5283a8bfd4abSjeff.liu if (copy_from_user(label, arg, sizeof(label))) 5284a8bfd4abSjeff.liu return -EFAULT; 5285a8bfd4abSjeff.liu 5286a8bfd4abSjeff.liu if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) { 52870b246afaSJeff Mahoney btrfs_err(fs_info, 52885d163e0eSJeff Mahoney "unable to set label with more than %d bytes", 5289a8bfd4abSjeff.liu BTRFS_LABEL_SIZE - 1); 5290a8bfd4abSjeff.liu return -EINVAL; 5291a8bfd4abSjeff.liu } 5292a8bfd4abSjeff.liu 5293a8bfd4abSjeff.liu ret = mnt_want_write_file(file); 5294a8bfd4abSjeff.liu if (ret) 5295a8bfd4abSjeff.liu return ret; 5296a8bfd4abSjeff.liu 5297a8bfd4abSjeff.liu trans = btrfs_start_transaction(root, 0); 5298a8bfd4abSjeff.liu if (IS_ERR(trans)) { 5299a8bfd4abSjeff.liu ret = PTR_ERR(trans); 5300a8bfd4abSjeff.liu goto out_unlock; 5301a8bfd4abSjeff.liu } 5302a8bfd4abSjeff.liu 53030b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 5304a8bfd4abSjeff.liu strcpy(super_block->label, label); 53050b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 53063a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 5307a8bfd4abSjeff.liu 5308a8bfd4abSjeff.liu out_unlock: 5309a8bfd4abSjeff.liu mnt_drop_write_file(file); 5310a8bfd4abSjeff.liu return ret; 5311a8bfd4abSjeff.liu } 5312a8bfd4abSjeff.liu 53132eaa055fSJeff Mahoney #define INIT_FEATURE_FLAGS(suffix) \ 53142eaa055fSJeff Mahoney { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \ 53152eaa055fSJeff Mahoney .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \ 53162eaa055fSJeff Mahoney .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix } 53172eaa055fSJeff Mahoney 5318d5131b65SDavid Sterba int btrfs_ioctl_get_supported_features(void __user *arg) 53192eaa055fSJeff Mahoney { 53204d4ab6d6SDavid Sterba static const struct btrfs_ioctl_feature_flags features[3] = { 53212eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SUPP), 53222eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_SET), 53232eaa055fSJeff Mahoney INIT_FEATURE_FLAGS(SAFE_CLEAR) 53242eaa055fSJeff Mahoney }; 53252eaa055fSJeff Mahoney 53262eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 53272eaa055fSJeff Mahoney return -EFAULT; 53282eaa055fSJeff Mahoney 53292eaa055fSJeff Mahoney return 0; 53302eaa055fSJeff Mahoney } 53312eaa055fSJeff Mahoney 53322eaa055fSJeff Mahoney static int btrfs_ioctl_get_features(struct file *file, void __user *arg) 53332eaa055fSJeff Mahoney { 53340b246afaSJeff Mahoney struct inode *inode = file_inode(file); 53350b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 53360b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 53372eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags features; 53382eaa055fSJeff Mahoney 53392eaa055fSJeff Mahoney features.compat_flags = btrfs_super_compat_flags(super_block); 53402eaa055fSJeff Mahoney features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block); 53412eaa055fSJeff Mahoney features.incompat_flags = btrfs_super_incompat_flags(super_block); 53422eaa055fSJeff Mahoney 53432eaa055fSJeff Mahoney if (copy_to_user(arg, &features, sizeof(features))) 53442eaa055fSJeff Mahoney return -EFAULT; 53452eaa055fSJeff Mahoney 53462eaa055fSJeff Mahoney return 0; 53472eaa055fSJeff Mahoney } 53482eaa055fSJeff Mahoney 53492ff7e61eSJeff Mahoney static int check_feature_bits(struct btrfs_fs_info *fs_info, 53503b02a68aSJeff Mahoney enum btrfs_feature_set set, 53512eaa055fSJeff Mahoney u64 change_mask, u64 flags, u64 supported_flags, 53522eaa055fSJeff Mahoney u64 safe_set, u64 safe_clear) 53532eaa055fSJeff Mahoney { 53543b02a68aSJeff Mahoney const char *type = btrfs_feature_set_names[set]; 53553b02a68aSJeff Mahoney char *names; 53562eaa055fSJeff Mahoney u64 disallowed, unsupported; 53572eaa055fSJeff Mahoney u64 set_mask = flags & change_mask; 53582eaa055fSJeff Mahoney u64 clear_mask = ~flags & change_mask; 53592eaa055fSJeff Mahoney 53602eaa055fSJeff Mahoney unsupported = set_mask & ~supported_flags; 53612eaa055fSJeff Mahoney if (unsupported) { 53623b02a68aSJeff Mahoney names = btrfs_printable_features(set, unsupported); 53633b02a68aSJeff Mahoney if (names) { 53640b246afaSJeff Mahoney btrfs_warn(fs_info, 53653b02a68aSJeff Mahoney "this kernel does not support the %s feature bit%s", 53663b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 53673b02a68aSJeff Mahoney kfree(names); 53683b02a68aSJeff Mahoney } else 53690b246afaSJeff Mahoney btrfs_warn(fs_info, 53702eaa055fSJeff Mahoney "this kernel does not support %s bits 0x%llx", 53712eaa055fSJeff Mahoney type, unsupported); 53722eaa055fSJeff Mahoney return -EOPNOTSUPP; 53732eaa055fSJeff Mahoney } 53742eaa055fSJeff Mahoney 53752eaa055fSJeff Mahoney disallowed = set_mask & ~safe_set; 53762eaa055fSJeff Mahoney if (disallowed) { 53773b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 53783b02a68aSJeff Mahoney if (names) { 53790b246afaSJeff Mahoney btrfs_warn(fs_info, 53803b02a68aSJeff Mahoney "can't set the %s feature bit%s while mounted", 53813b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 53823b02a68aSJeff Mahoney kfree(names); 53833b02a68aSJeff Mahoney } else 53840b246afaSJeff Mahoney btrfs_warn(fs_info, 53852eaa055fSJeff Mahoney "can't set %s bits 0x%llx while mounted", 53862eaa055fSJeff Mahoney type, disallowed); 53872eaa055fSJeff Mahoney return -EPERM; 53882eaa055fSJeff Mahoney } 53892eaa055fSJeff Mahoney 53902eaa055fSJeff Mahoney disallowed = clear_mask & ~safe_clear; 53912eaa055fSJeff Mahoney if (disallowed) { 53923b02a68aSJeff Mahoney names = btrfs_printable_features(set, disallowed); 53933b02a68aSJeff Mahoney if (names) { 53940b246afaSJeff Mahoney btrfs_warn(fs_info, 53953b02a68aSJeff Mahoney "can't clear the %s feature bit%s while mounted", 53963b02a68aSJeff Mahoney names, strchr(names, ',') ? "s" : ""); 53973b02a68aSJeff Mahoney kfree(names); 53983b02a68aSJeff Mahoney } else 53990b246afaSJeff Mahoney btrfs_warn(fs_info, 54002eaa055fSJeff Mahoney "can't clear %s bits 0x%llx while mounted", 54012eaa055fSJeff Mahoney type, disallowed); 54022eaa055fSJeff Mahoney return -EPERM; 54032eaa055fSJeff Mahoney } 54042eaa055fSJeff Mahoney 54052eaa055fSJeff Mahoney return 0; 54062eaa055fSJeff Mahoney } 54072eaa055fSJeff Mahoney 54082ff7e61eSJeff Mahoney #define check_feature(fs_info, change_mask, flags, mask_base) \ 54092ff7e61eSJeff Mahoney check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \ 54102eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SUPP, \ 54112eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \ 54122eaa055fSJeff Mahoney BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR) 54132eaa055fSJeff Mahoney 54142eaa055fSJeff Mahoney static int btrfs_ioctl_set_features(struct file *file, void __user *arg) 54152eaa055fSJeff Mahoney { 54160b246afaSJeff Mahoney struct inode *inode = file_inode(file); 54170b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 54180b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 54190b246afaSJeff Mahoney struct btrfs_super_block *super_block = fs_info->super_copy; 54202eaa055fSJeff Mahoney struct btrfs_ioctl_feature_flags flags[2]; 54212eaa055fSJeff Mahoney struct btrfs_trans_handle *trans; 54222eaa055fSJeff Mahoney u64 newflags; 54232eaa055fSJeff Mahoney int ret; 54242eaa055fSJeff Mahoney 54252eaa055fSJeff Mahoney if (!capable(CAP_SYS_ADMIN)) 54262eaa055fSJeff Mahoney return -EPERM; 54272eaa055fSJeff Mahoney 54282eaa055fSJeff Mahoney if (copy_from_user(flags, arg, sizeof(flags))) 54292eaa055fSJeff Mahoney return -EFAULT; 54302eaa055fSJeff Mahoney 54312eaa055fSJeff Mahoney /* Nothing to do */ 54322eaa055fSJeff Mahoney if (!flags[0].compat_flags && !flags[0].compat_ro_flags && 54332eaa055fSJeff Mahoney !flags[0].incompat_flags) 54342eaa055fSJeff Mahoney return 0; 54352eaa055fSJeff Mahoney 54362ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_flags, 54372eaa055fSJeff Mahoney flags[1].compat_flags, COMPAT); 54382eaa055fSJeff Mahoney if (ret) 54392eaa055fSJeff Mahoney return ret; 54402eaa055fSJeff Mahoney 54412ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].compat_ro_flags, 54422eaa055fSJeff Mahoney flags[1].compat_ro_flags, COMPAT_RO); 54432eaa055fSJeff Mahoney if (ret) 54442eaa055fSJeff Mahoney return ret; 54452eaa055fSJeff Mahoney 54462ff7e61eSJeff Mahoney ret = check_feature(fs_info, flags[0].incompat_flags, 54472eaa055fSJeff Mahoney flags[1].incompat_flags, INCOMPAT); 54482eaa055fSJeff Mahoney if (ret) 54492eaa055fSJeff Mahoney return ret; 54502eaa055fSJeff Mahoney 54517ab19625SDavid Sterba ret = mnt_want_write_file(file); 54527ab19625SDavid Sterba if (ret) 54537ab19625SDavid Sterba return ret; 54547ab19625SDavid Sterba 54558051aa1aSDavid Sterba trans = btrfs_start_transaction(root, 0); 54567ab19625SDavid Sterba if (IS_ERR(trans)) { 54577ab19625SDavid Sterba ret = PTR_ERR(trans); 54587ab19625SDavid Sterba goto out_drop_write; 54597ab19625SDavid Sterba } 54602eaa055fSJeff Mahoney 54610b246afaSJeff Mahoney spin_lock(&fs_info->super_lock); 54622eaa055fSJeff Mahoney newflags = btrfs_super_compat_flags(super_block); 54632eaa055fSJeff Mahoney newflags |= flags[0].compat_flags & flags[1].compat_flags; 54642eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags); 54652eaa055fSJeff Mahoney btrfs_set_super_compat_flags(super_block, newflags); 54662eaa055fSJeff Mahoney 54672eaa055fSJeff Mahoney newflags = btrfs_super_compat_ro_flags(super_block); 54682eaa055fSJeff Mahoney newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags; 54692eaa055fSJeff Mahoney newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags); 54702eaa055fSJeff Mahoney btrfs_set_super_compat_ro_flags(super_block, newflags); 54712eaa055fSJeff Mahoney 54722eaa055fSJeff Mahoney newflags = btrfs_super_incompat_flags(super_block); 54732eaa055fSJeff Mahoney newflags |= flags[0].incompat_flags & flags[1].incompat_flags; 54742eaa055fSJeff Mahoney newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags); 54752eaa055fSJeff Mahoney btrfs_set_super_incompat_flags(super_block, newflags); 54760b246afaSJeff Mahoney spin_unlock(&fs_info->super_lock); 54772eaa055fSJeff Mahoney 54783a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 54797ab19625SDavid Sterba out_drop_write: 54807ab19625SDavid Sterba mnt_drop_write_file(file); 54817ab19625SDavid Sterba 54827ab19625SDavid Sterba return ret; 54832eaa055fSJeff Mahoney } 54842eaa055fSJeff Mahoney 5485f46b5a66SChristoph Hellwig long btrfs_ioctl(struct file *file, unsigned int 5486f46b5a66SChristoph Hellwig cmd, unsigned long arg) 5487f46b5a66SChristoph Hellwig { 54880b246afaSJeff Mahoney struct inode *inode = file_inode(file); 54890b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 54900b246afaSJeff Mahoney struct btrfs_root *root = BTRFS_I(inode)->root; 54914bcabaa3SChristoph Hellwig void __user *argp = (void __user *)arg; 5492f46b5a66SChristoph Hellwig 5493f46b5a66SChristoph Hellwig switch (cmd) { 54946cbff00fSChristoph Hellwig case FS_IOC_GETFLAGS: 54956cbff00fSChristoph Hellwig return btrfs_ioctl_getflags(file, argp); 54966cbff00fSChristoph Hellwig case FS_IOC_SETFLAGS: 54976cbff00fSChristoph Hellwig return btrfs_ioctl_setflags(file, argp); 54986cbff00fSChristoph Hellwig case FS_IOC_GETVERSION: 54996cbff00fSChristoph Hellwig return btrfs_ioctl_getversion(file, argp); 5500f7039b1dSLi Dongyang case FITRIM: 5501f7039b1dSLi Dongyang return btrfs_ioctl_fitrim(file, argp); 5502f46b5a66SChristoph Hellwig case BTRFS_IOC_SNAP_CREATE: 5503fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 0); 5504fdfb1e4fSLi Zefan case BTRFS_IOC_SNAP_CREATE_V2: 5505fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create_v2(file, argp, 0); 55063de4586cSChris Mason case BTRFS_IOC_SUBVOL_CREATE: 5507fa0d2b9bSLi Zefan return btrfs_ioctl_snap_create(file, argp, 1); 55086f72c7e2SArne Jansen case BTRFS_IOC_SUBVOL_CREATE_V2: 55096f72c7e2SArne Jansen return btrfs_ioctl_snap_create_v2(file, argp, 1); 551076dda93cSYan, Zheng case BTRFS_IOC_SNAP_DESTROY: 551176dda93cSYan, Zheng return btrfs_ioctl_snap_destroy(file, argp); 55120caa102dSLi Zefan case BTRFS_IOC_SUBVOL_GETFLAGS: 55130caa102dSLi Zefan return btrfs_ioctl_subvol_getflags(file, argp); 55140caa102dSLi Zefan case BTRFS_IOC_SUBVOL_SETFLAGS: 55150caa102dSLi Zefan return btrfs_ioctl_subvol_setflags(file, argp); 55166ef5ed0dSJosef Bacik case BTRFS_IOC_DEFAULT_SUBVOL: 55176ef5ed0dSJosef Bacik return btrfs_ioctl_default_subvol(file, argp); 5518f46b5a66SChristoph Hellwig case BTRFS_IOC_DEFRAG: 55191e701a32SChris Mason return btrfs_ioctl_defrag(file, NULL); 55201e701a32SChris Mason case BTRFS_IOC_DEFRAG_RANGE: 55211e701a32SChris Mason return btrfs_ioctl_defrag(file, argp); 5522f46b5a66SChristoph Hellwig case BTRFS_IOC_RESIZE: 5523198605a8SMiao Xie return btrfs_ioctl_resize(file, argp); 5524f46b5a66SChristoph Hellwig case BTRFS_IOC_ADD_DEV: 55252ff7e61eSJeff Mahoney return btrfs_ioctl_add_dev(fs_info, argp); 5526f46b5a66SChristoph Hellwig case BTRFS_IOC_RM_DEV: 5527da24927bSMiao Xie return btrfs_ioctl_rm_dev(file, argp); 55286b526ed7SAnand Jain case BTRFS_IOC_RM_DEV_V2: 55296b526ed7SAnand Jain return btrfs_ioctl_rm_dev_v2(file, argp); 5530475f6387SJan Schmidt case BTRFS_IOC_FS_INFO: 55312ff7e61eSJeff Mahoney return btrfs_ioctl_fs_info(fs_info, argp); 5532475f6387SJan Schmidt case BTRFS_IOC_DEV_INFO: 55332ff7e61eSJeff Mahoney return btrfs_ioctl_dev_info(fs_info, argp); 5534f46b5a66SChristoph Hellwig case BTRFS_IOC_BALANCE: 55359ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, NULL); 5536f46b5a66SChristoph Hellwig case BTRFS_IOC_TRANS_START: 5537f46b5a66SChristoph Hellwig return btrfs_ioctl_trans_start(file); 5538f46b5a66SChristoph Hellwig case BTRFS_IOC_TRANS_END: 5539f46b5a66SChristoph Hellwig return btrfs_ioctl_trans_end(file); 5540ac8e9819SChris Mason case BTRFS_IOC_TREE_SEARCH: 5541ac8e9819SChris Mason return btrfs_ioctl_tree_search(file, argp); 5542cc68a8a5SGerhard Heift case BTRFS_IOC_TREE_SEARCH_V2: 5543cc68a8a5SGerhard Heift return btrfs_ioctl_tree_search_v2(file, argp); 5544ac8e9819SChris Mason case BTRFS_IOC_INO_LOOKUP: 5545ac8e9819SChris Mason return btrfs_ioctl_ino_lookup(file, argp); 5546d7728c96SJan Schmidt case BTRFS_IOC_INO_PATHS: 5547d7728c96SJan Schmidt return btrfs_ioctl_ino_to_path(root, argp); 5548d7728c96SJan Schmidt case BTRFS_IOC_LOGICAL_INO: 55492ff7e61eSJeff Mahoney return btrfs_ioctl_logical_to_ino(fs_info, argp); 55501406e432SJosef Bacik case BTRFS_IOC_SPACE_INFO: 55512ff7e61eSJeff Mahoney return btrfs_ioctl_space_info(fs_info, argp); 55529b199859SFilipe David Borba Manana case BTRFS_IOC_SYNC: { 55539b199859SFilipe David Borba Manana int ret; 55549b199859SFilipe David Borba Manana 55550b246afaSJeff Mahoney ret = btrfs_start_delalloc_roots(fs_info, 0, -1); 55569b199859SFilipe David Borba Manana if (ret) 55579b199859SFilipe David Borba Manana return ret; 55580b246afaSJeff Mahoney ret = btrfs_sync_fs(inode->i_sb, 1); 55592fad4e83SDavid Sterba /* 55602fad4e83SDavid Sterba * The transaction thread may want to do more work, 556101327610SNicholas D Steeves * namely it pokes the cleaner kthread that will start 55622fad4e83SDavid Sterba * processing uncleaned subvols. 55632fad4e83SDavid Sterba */ 55640b246afaSJeff Mahoney wake_up_process(fs_info->transaction_kthread); 55659b199859SFilipe David Borba Manana return ret; 55669b199859SFilipe David Borba Manana } 556746204592SSage Weil case BTRFS_IOC_START_SYNC: 55689a8c28beSMiao Xie return btrfs_ioctl_start_sync(root, argp); 556946204592SSage Weil case BTRFS_IOC_WAIT_SYNC: 55702ff7e61eSJeff Mahoney return btrfs_ioctl_wait_sync(fs_info, argp); 5571475f6387SJan Schmidt case BTRFS_IOC_SCRUB: 5572b8e95489SMiao Xie return btrfs_ioctl_scrub(file, argp); 5573475f6387SJan Schmidt case BTRFS_IOC_SCRUB_CANCEL: 55742ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_cancel(fs_info); 5575475f6387SJan Schmidt case BTRFS_IOC_SCRUB_PROGRESS: 55762ff7e61eSJeff Mahoney return btrfs_ioctl_scrub_progress(fs_info, argp); 5577c9e9f97bSIlya Dryomov case BTRFS_IOC_BALANCE_V2: 55789ba1f6e4SLiu Bo return btrfs_ioctl_balance(file, argp); 5579837d5b6eSIlya Dryomov case BTRFS_IOC_BALANCE_CTL: 55802ff7e61eSJeff Mahoney return btrfs_ioctl_balance_ctl(fs_info, arg); 558119a39dceSIlya Dryomov case BTRFS_IOC_BALANCE_PROGRESS: 55822ff7e61eSJeff Mahoney return btrfs_ioctl_balance_progress(fs_info, argp); 55838ea05e3aSAlexander Block case BTRFS_IOC_SET_RECEIVED_SUBVOL: 55848ea05e3aSAlexander Block return btrfs_ioctl_set_received_subvol(file, argp); 5585abccd00fSHugo Mills #ifdef CONFIG_64BIT 5586abccd00fSHugo Mills case BTRFS_IOC_SET_RECEIVED_SUBVOL_32: 5587abccd00fSHugo Mills return btrfs_ioctl_set_received_subvol_32(file, argp); 5588abccd00fSHugo Mills #endif 558931db9f7cSAlexander Block case BTRFS_IOC_SEND: 559031db9f7cSAlexander Block return btrfs_ioctl_send(file, argp); 5591c11d2c23SStefan Behrens case BTRFS_IOC_GET_DEV_STATS: 55922ff7e61eSJeff Mahoney return btrfs_ioctl_get_dev_stats(fs_info, argp); 55935d13a37bSArne Jansen case BTRFS_IOC_QUOTA_CTL: 5594905b0ddaSMiao Xie return btrfs_ioctl_quota_ctl(file, argp); 55955d13a37bSArne Jansen case BTRFS_IOC_QGROUP_ASSIGN: 5596905b0ddaSMiao Xie return btrfs_ioctl_qgroup_assign(file, argp); 55975d13a37bSArne Jansen case BTRFS_IOC_QGROUP_CREATE: 5598905b0ddaSMiao Xie return btrfs_ioctl_qgroup_create(file, argp); 55995d13a37bSArne Jansen case BTRFS_IOC_QGROUP_LIMIT: 5600905b0ddaSMiao Xie return btrfs_ioctl_qgroup_limit(file, argp); 56012f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN: 56022f232036SJan Schmidt return btrfs_ioctl_quota_rescan(file, argp); 56032f232036SJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_STATUS: 56042f232036SJan Schmidt return btrfs_ioctl_quota_rescan_status(file, argp); 560557254b6eSJan Schmidt case BTRFS_IOC_QUOTA_RESCAN_WAIT: 560657254b6eSJan Schmidt return btrfs_ioctl_quota_rescan_wait(file, argp); 56073f6bcfbdSStefan Behrens case BTRFS_IOC_DEV_REPLACE: 56082ff7e61eSJeff Mahoney return btrfs_ioctl_dev_replace(fs_info, argp); 5609867ab667Sjeff.liu case BTRFS_IOC_GET_FSLABEL: 5610867ab667Sjeff.liu return btrfs_ioctl_get_fslabel(file, argp); 5611a8bfd4abSjeff.liu case BTRFS_IOC_SET_FSLABEL: 5612a8bfd4abSjeff.liu return btrfs_ioctl_set_fslabel(file, argp); 56132eaa055fSJeff Mahoney case BTRFS_IOC_GET_SUPPORTED_FEATURES: 5614d5131b65SDavid Sterba return btrfs_ioctl_get_supported_features(argp); 56152eaa055fSJeff Mahoney case BTRFS_IOC_GET_FEATURES: 56162eaa055fSJeff Mahoney return btrfs_ioctl_get_features(file, argp); 56172eaa055fSJeff Mahoney case BTRFS_IOC_SET_FEATURES: 56182eaa055fSJeff Mahoney return btrfs_ioctl_set_features(file, argp); 5619f46b5a66SChristoph Hellwig } 5620f46b5a66SChristoph Hellwig 5621f46b5a66SChristoph Hellwig return -ENOTTY; 5622f46b5a66SChristoph Hellwig } 56234c63c245SLuke Dashjr 56244c63c245SLuke Dashjr #ifdef CONFIG_COMPAT 56254c63c245SLuke Dashjr long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 56264c63c245SLuke Dashjr { 56272a362249SJeff Mahoney /* 56282a362249SJeff Mahoney * These all access 32-bit values anyway so no further 56292a362249SJeff Mahoney * handling is necessary. 56302a362249SJeff Mahoney */ 56314c63c245SLuke Dashjr switch (cmd) { 56324c63c245SLuke Dashjr case FS_IOC32_GETFLAGS: 56334c63c245SLuke Dashjr cmd = FS_IOC_GETFLAGS; 56344c63c245SLuke Dashjr break; 56354c63c245SLuke Dashjr case FS_IOC32_SETFLAGS: 56364c63c245SLuke Dashjr cmd = FS_IOC_SETFLAGS; 56374c63c245SLuke Dashjr break; 56384c63c245SLuke Dashjr case FS_IOC32_GETVERSION: 56394c63c245SLuke Dashjr cmd = FS_IOC_GETVERSION; 56404c63c245SLuke Dashjr break; 56414c63c245SLuke Dashjr } 56424c63c245SLuke Dashjr 56434c63c245SLuke Dashjr return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); 56444c63c245SLuke Dashjr } 56454c63c245SLuke Dashjr #endif 5646